[PATCHSET v4] sched: Implement BPF extensible scheduler class

[PATCHSET v4] sched: Implement BPF extensible scheduler class

[Tejun Heo]

Changes
-------

This is v4 of sched_ext (SCX) patchset. The followings are changes from v3
(https://lkml.kernel.org/r/20230317213333.2174969-1-tj@kernel.org):

- There aren't any significant changes to the sched_ext API even though we
  kept experimenting heavily with a couple BPF scheduler implementations
  indicating that the core API reached a level of maturity.

- 0002-sched-Encapsulate-task-attribute-change-sequence-int.patch which
  implemented custom guard scope for scheduler attribute changes dropped as
  upstream is moving towards a more generic implementation.

- Build fixes with different CONFIG combinations.

- Core code cleanups and improvements including how idle CPU is selected and
  disabling ttwu_queue for tasks on SCX to avoid confusing BPF schedulers
  expecting ->select_cpu() call. See
  0012-sched_ext-Implement-BPF-extensible-scheduler-class.patch for more
  details.

- "_example" dropped from the example schedulers as the distinction between
  the example-only and practically-useful isn't black-and-white. Instead,
  each scheduler has detailed comments and there's also a README file.

- scx_central, scx_pair and scx_flatcg are moved into their own patches as
  suggested by Josh Don.

- scx_atropos received sustantial updates including fixes for bugs that
  could cause temporary stalls and improvements in load balancing and wakeup
  target CPU selection. For details, See
  0034-sched_ext-Add-a-rust-userspace-hybrid-example-schedu.patch.

v2 (http://lkml.kernel.org/r/20230128001639.3510083-1-tj@kernel.org) -> v3:

- ops.set_weight() added to allow BPF schedulers to track weight changes
  without polling p->scx.weight.

- scx_bpf_task_cgroup() kfunc added to allow BPF scheduler to reliably
  determine the current cpu cgroup under rq lock protection. This required
  improving the kf_mask SCX operation verification mechanism and adding
  0023-sched_ext-Track-tasks-that-are-subjects-of-the-in-fl.patch.

- Updated to use the latest BPF improvements including KF_RCU and the inline
  iterator.

- scx_example_flatcg added to 0024-sched_ext-Add-cgroup-support.patch. It
  uses the new BPF RB tree support to implement flattened cgroup hierarchy.

- A DSQ now also contains an rbtree so that it can be used to implement
  vtime based scheduling among tasks sharing a DSQ conveniently and
  efficiently. For more details, see
  0029-sched_ext-Add-vtime-ordered-priority-queue-to-dispat.patch. All
  eligible example schedulers are updated to default to weighted vtime
  scheduilng.

- atropos scheduler's userspace code is substantially restructred and
  rewritten. The binary is renamed to scx_atropos and can auto-config the
  domains according to the cache topology.

- Various other example scheduler updates including scx_example_dummy being
  renamed to scx_example_simple, the example schedulers defaulting to
  enabling switch_all and clarifying performance expectation of each example
  scheduler.

- A bunch of fixes and improvements. Please refer to each patch for details.

v1 (http://lkml.kernel.org/r/20221130082313.3241517-1-tj@kernel.org) -> v2:

- Rebased on top of bpf/for-next - a5f6b9d577eb ("Merge branch 'Enable
  struct_ops programs to be sleepable'"). There were several missing
  features including generic cpumask helpers and sleepable struct_ops
  operation support that v1 was working around. The rebase gets rid of all
  SCX specific temporary helpers.

- Some kfunc helpers are context-sensitive and can only be called from
  specific operations. v1 didn't restrict kfunc accesses allowing them to be
  misused which can lead to crashes and other malfunctions. v2 makes more
  kfuncs safe to be called from anywhere and implements per-task mask based
  runtime access control for the rest. The longer-term plan is to make the
  BPF verifier enforce these restrictions. Combined with the above, sans
  mistakes and bugs, it shouldn't be possible to crash the machine through
  SCX and its helpers.

- Core-sched support. While v1 implemented the pick_task operation, there
  were multiple missing pieces for working core-sched support. v2 adds
  0027-sched_ext-Implement-core-sched-support.patch. SCX by default
  implements global FIFO ordering and allows the BPF schedulers to implement
  custom ordering via scx_ops.core_sched_before(). scx_example_qmap is
  updated so that the five queues' relative priorities are correctly
  reflected when core-sched is enabled.

- Dropped balance_scx_on_up() which was called from put_prev_task_balance().
  UP support is now contained in SCX proper.

- 0002-sched-Encapsulate-task-attribute-change-sequence-int.patch adds
  SCHED_CHANGE_BLOCK() which encapsulates the preparation and restoration
  sequences used for task attribute changes. For SCX, this replaces
  sched_deq_and_put_task() and sched_enq_and_set_task() from v1.

- 0011-sched-Add-reason-to-sched_move_task.patch dropped from v1. SCX now
  distinguishes cgroup and autogroup tg's using task_group_is_autogroup().

- Other misc changes including fixes for bugs that Julia Lawall noticed and
  patch descriptions updates with more details on how the introduced changes
  are going to be used.

- MAINTAINERS entries added.

The followings are discussion points which were raised but didn't result in
code changes in this iteration.

- There were discussions around exposing __setscheduler_prio() and, in v2,
  SCHED_CHANGE_BLOCK() in kernel/sched/sched.h. Switching scheduler
  implementations is innate for SCX. At the very least, it needs to be able
  to turn on and off the BPF scheduler which requires something equivalent
  to SCHED_CHANGE_BLOCK(). The use of __setscheduler_prio() depends on the
  behavior we want to present to userspace. The current one of using CFS as
  the fallback when BPF scheduler is not available seems more friendly and
  less error-prone to other options.

- Another discussion point was around for_each_active_class() and friends
  which skip over CFS or SCX when it's known that the sched_class must be
  empty. I left it as-is for now as it seems to be cleaner and more robust
  than trying to plug each operation which may added unnecessary overheads.

Overview
--------

This patch set proposes a new scheduler class called ‘ext_sched_class’, or
sched_ext, which allows scheduling policies to be implemented as BPF programs.

More details will be provided on the overall architecture of sched_ext
throughout the various patches in this set, as well as in the “How” section
below. We realize that this patch set is a significant proposal, so we will be
going into depth in the following “Motivation” section to explain why we think
it’s justified. That section is laid out as follows, touching on three main
axes where we believe that sched_ext provides significant value:

1. Ease of experimentation and exploration: Enabling rapid iteration of new
   scheduling policies.

2. Customization: Building application-specific schedulers which implement
   policies that are not applicable to general-purpose schedulers.

3. Rapid scheduler deployments: Non-disruptive swap outs of scheduling
   policies in production environments.

After the motivation section, we’ll provide a more detailed (but still
high-level) overview of how sched_ext works.

Motivation
----------

1. Ease of experimentation and exploration

*Why is exploration important?*

Scheduling is a challenging problem space. Small changes in scheduling
behavior can have a significant impact on various components of a system, with
the corresponding effects varying widely across different platforms,
architectures, and workloads.

While complexities have always existed in scheduling, they have increased
dramatically over the past 10-15 years. In the mid-late 2000s, cores were
typically homogeneous and further apart from each other, with the criteria for
scheduling being roughly the same across the entire die.

Systems in the modern age are by comparison much more complex. Modern CPU
designs, where the total power budget of all CPU cores often far exceeds the
power budget of the socket, with dynamic frequency scaling, and with or
without chiplets, have significantly expanded the scheduling problem space.
Cache hierarchies have become less uniform, with Core Complex (CCX) designs
such as recent AMD processors having multiple shared L3 caches within a single
socket. Such topologies resemble NUMA sans persistent NUMA node stickiness.

Use-cases have become increasingly complex and diverse as well. Applications
such as mobile and VR have strict latency requirements to avoid missing
deadlines that impact user experience. Stacking workloads in servers is
constantly pushing the demands on the scheduler in terms of workload isolation
and resource distribution.

Experimentation and exploration are important for any non-trivial problem
space. However, given the recent hardware and software developments, we
believe that experimentation and exploration are not just important, but
_critical_ in the scheduling problem space.

Indeed, other approaches in industry are already being explored. AMD has
proposed an experimental patch set [0] which enables userspace to provide
hints to the scheduler via “Userspace Hinting”. The approach adds a prctl()
API which allows callers to set a numerical “hint” value on a struct
task_struct. This hint is then optionally read by the scheduler to adjust the
cost calculus for various scheduling decisions.

[0]: https://lore.kernel.org/lkml/20220910105326.1797-1-kprateek.nayak@amd.com/

Huawei have also expressed interest [1] in enabling some form of programmable
scheduling. While we’re unaware of any patch sets which have been sent to the
upstream list for this proposal, it similarly illustrates the need for more
flexibility in the scheduler.

[1]: https://lore.kernel.org/bpf/dedc7b72-9da4-91d0-d81d-75360c177188@huawei.com/

Additionally, Google has developed ghOSt [2] with the goal of enabling custom,
userspace driven scheduling policies. Prior presentations at LPC [3] have
discussed ghOSt and how BPF can be used to accelerate scheduling.

[2]: https://dl.acm.org/doi/pdf/10.1145/3477132.3483542
[3]: https://lpc.events/event/16/contributions/1365/

*Why can’t we just explore directly with CFS?*

Experimenting with CFS directly or implementing a new sched_class from scratch
is of course possible, but is often difficult and time consuming. Newcomers to
the scheduler often require years to understand the codebase and become
productive contributors. Even for seasoned kernel engineers, experimenting
with and upstreaming features can take a very long time. The iteration process
itself is also time consuming, as testing scheduler changes on real hardware
requires reinstalling the kernel and rebooting the host.
 
Core scheduling is an example of a feature that took a significant amount of
time and effort to integrate into the kernel. Part of the difficulty with core
scheduling was the inherent mismatch in abstraction between the desire to
perform core-wide scheduling, and the per-cpu design of the kernel scheduler.
This caused issues, for example ensuring proper fairness between the
independent runqueues of SMT siblings.

The high barrier to entry for working on the scheduler is an impediment to
academia as well. Master’s/PhD candidates who are interested in improving the
scheduler will spend years ramping-up, only to complete their degrees just as
they’re finally ready to make significant changes. A lower entrance barrier
would allow researchers to more quickly ramp up, test out hypotheses, and
iterate on novel ideas. Research methodology is also severely hampered by the
high barrier of entry to make modifications; for example, the Shenango [4] and
Shinjuku scheduling policies used sched affinity to replicate the desired
policy semantics, due to the difficulty of incorporating these policies into
the kernel directly.

[4]: https://www.usenix.org/system/files/nsdi19-ousterhout.pdf

The iterative process itself also imposes a significant cost to working on the
scheduler. Testing changes requires developers to recompile and reinstall the
kernel, reboot their machines, rewarm their workloads, and then finally rerun
their benchmarks. Though some of this overhead could potentially be mitigated
by enabling schedulers to be implemented as kernel modules, a machine crash or
subtle system state corruption is always only one innocuous mistake away.
These problems are exacerbated when testing production workloads in a
datacenter environment as well, where multiple hosts may be involved in an
experiment; requiring a significantly longer ramp up time. Warming up memcache
instances in the Meta production environment takes hours, for example.

*How does sched_ext help with exploration?*

sched_ext attempts to address all of the problems described above. In this
section, we’ll describe the benefits to experimentation and exploration that
are afforded by sched_ext, provide real-world examples of those benefits, and
discuss some of the trade-offs and considerations in our design choices.

One of our main goals was to lower the barrier to entry for experimenting
with the scheduler. sched_ext provides ergonomic callbacks and helpers to
ease common operations such as managing idle CPUs, scheduling tasks on
arbitrary CPUs, handling preemptions from other scheduling classes, and
more. While sched_ext does require some ramp-up, the complexity is
self-contained, and the learning curve gradual. Developers can ramp up by
first implementing simple policies such as global weighted vtime scheduling
in only tens of lines of code, and then continue to learn the APIs and
building blocks available with sched_ext as they build more featureful and
complex schedulers.

Another critical advantage provided by sched_ext is the use of BPF. BPF
provides strong safety guarantees by statically analyzing programs at load
time to ensure that they cannot corrupt or crash the system. sched_ext
guarantees system integrity no matter what BPF scheduler is loaded, and
provides mechanisms to safely disable the current BPF scheduler and migrate
tasks back to a trusted scheduler. For example, we also implement in-kernel
safety mechanisms to guarantee that a misbehaving scheduler cannot
indefinitely starve tasks. BPF also enables sched_ext to significantly improve
iteration speed for running experiments. Loading and unloading a BPF scheduler
is simply a matter of running and terminating a sched_ext binary.

BPF also provides programs with a rich set of APIs, such as maps, kfuncs,
and BPF helpers. In addition to providing useful building blocks to programs
that run entirely in kernel space (such as many of our example schedulers),
these APIs also allow programs to leverage user space in making scheduling
decisions. Specifically, the Atropos sample scheduler has a relatively
simple weighted vtime or FIFO scheduling layer in BPF, paired with a load
balancing component in userspace written in Rust. As described in more
detail below, we also built a more general user-space scheduling framework
called “rhone” by leveraging various BPF features.

On the other hand, BPF does have shortcomings, as can be plainly seen from the
complexity in some of the example schedulers. scx_example_pair.bpf.c
illustrates this point well. To start, it requires a good amount of code to
emulate cgroup-local-storage. In the kernel proper, this would simply be a
matter of adding another pointer to the struct cgroup, but in BPF, it requires
a complex juggling of data amongst multiple different maps, a good amount of
boilerplate code, and some unwieldy bpf_loop()‘s and atomics. The code is also
littered with explicit and often unnecessary sanity checks to appease the
verifier.

That being said, BPF is being rapidly improved. For example, Yonghong Song
recently upstreamed a patch set [5] to add a cgroup local storage map type,
allowing scx_example_pair.bpf.c to be simplified. There are plans to address
other issues as well, such as providing statically-verified locking, and
avoiding the need for unnecessary sanity checks. Addressing these shortcomings
is a high priority for BPF, and as progress continues to be made, we expect
most deficiencies to be addressed in the not-too-distant future.

[5]: https://lore.kernel.org/bpf/20221026042835.672317-1-yhs@fb.com/

Yet another exploration advantage of sched_ext is helping widening the scope
of experiments. For example, sched_ext makes it easy to defer CPU assignment
until a task starts executing, allowing schedulers to share scheduling queues
at any granularity (hyper-twin, CCX and so on). Additionally, higher level
frameworks can be built on top to further widen the scope. For example, the
aforementioned “rhone” [6] library allows implementing scheduling policies in
user-space by encapsulating the complexity around communicating scheduling
decisions with the kernel. This allows taking advantage of a richer
programming environment in user-space, enabling experimenting with, for
instance, more complex mathematical models.

[6]: https://github.com/Decave/rhone

sched_ext also allows developers to leverage machine learning. At Meta, we
experimented with using machine learning to predict whether a running task
would soon yield its CPU. These predictions can be used to aid the scheduler
in deciding whether to keep a runnable task on its current CPU rather than
migrating it to an idle CPU, with the hope of avoiding unnecessary cache
misses. Using a tiny neural net model with only one hidden layer of size 16,
and a decaying count of 64 syscalls as a feature, we were able to achieve a
15% throughput improvement on an Nginx benchmark, with an 87% inference
accuracy.

2. Customization

This section discusses how sched_ext can enable users to run workloads on
application-specific schedulers.

*Why deploy custom schedulers rather than improving CFS?*

Implementing application-specific schedulers and improving CFS are not
conflicting goals. Scheduling features explored with sched_ext which yield
beneficial results, and which are sufficiently generalizable, can and should
be integrated into CFS. However, CFS is fundamentally designed to be a general
purpose scheduler, and thus is not conducive to being extended with some
highly targeted application or hardware specific changes.

Targeted, bespoke scheduling has many potential use cases. For example, VM
scheduling can make certain optimizations that are infeasible in CFS due to
the constrained problem space (scheduling a static number of long-running
VCPUs versus an arbitrary number of threads). Additionally, certain
applications might want to make targeted policy decisions based on hints
directly from the application (for example, a service that knows the different
deadlines of incoming RPCs).

Google has also experimented with some promising, novel scheduling policies.
One example is “central” scheduling, wherein a single CPU makes all scheduling
decisions for the entire system. This allows most cores on the system to be
fully dedicated to running workloads, and can have significant performance
improvements for certain use cases. For example, central scheduling with VCPUs
can avoid expensive vmexits and cache flushes, by instead delegating the
responsibility of preemption checks from the tick to a single CPU. See
scx_example_central.bpf.c for a simple example of a central scheduling policy
built in sched_ext.

Some workloads also have non-generalizable constraints which enable
optimizations in a scheduling policy which would otherwise not be feasible.
For example,VM workloads at Google typically have a low overcommit ratio
compared to the number of physical CPUs. This allows the scheduler to support
bounded tail latencies, as well as longer blocks of uninterrupted time.

Yet another interesting use case is the scx_example_flatcg scheduler, which
is in 0024-sched_ext-Add-cgroup-support.patch and provides a flattened
hierarchical vtree for cgroups. This scheduler does not account for
thundering herd problems among cgroups, and therefore may not be suitable
for inclusion in CFS. However, in a simple benchmark using wrk[8] on apache
serving a CGI script calculating sha1sum of a small file, it outperformed
CFS by ~3% with CPU controller disabled and by ~10% with two apache
instances competing with 2:1 weight ratio nested four level deep.

[7] https://github.com/wg/wrk

Certain industries require specific scheduling behaviors that do not apply
broadly. For example, ARINC 653 defines scheduling behavior that is widely
used by avionic software, and some out-of-tree implementations
(https://ieeexplore.ieee.org/document/7005306) have been built. While the
upstream community may decide to merge one such implementation in the future,
it would also be entirely reasonable to not do so given the narrowness of
use-case, and non-generalizable, strict requirements. Such cases can be well
served by sched_ext in all stages of the software development lifecycle --
development, testing, deployment and maintenance.

There are also classes of policy exploration, such as machine learning, or
responding in real-time to application hints, that are significantly harder
(and not necessarily appropriate) to integrate within the kernel itself.

*Won’t this increase fragmentation?*

We acknowledge that to some degree, sched_ext does run the risk of increasing
the fragmentation of scheduler implementations. As a result of exploration,
however, we believe that enabling the larger ecosystem to innovate will
ultimately accelerate the overall development and performance of Linux.
Additionally, our licensing and API stability policies should incentivize
users to upstream their schedulers.

BPF programs are required to be GPLv2, which is enforced by the verifier on
program loads. With regards to API stability, just as with other semi-internal
interfaces such as BPF kfuncs, we won’t be providing any API stability
guarantees to BPF schedulers. While we intend to make an effort to provide
compatibility when possible, we will not provide any explicit, strong
guarantees as the kernel typically does with e.g. UAPI headers. For users who
decide to keep their schedulers out-of-tree,the licensing and maintenance
overheads will be fundamentally the same as for carrying out-of-tree patches.

With regards to the schedulers included in this patch set, and any other
schedulers we implement in the future, both Meta and Google will open-source
all of the schedulers we implement which have any relevance to the broader
upstream community. We expect that some of these, such as the example
schedulers and scx_example_flatcg scheduler, will be upstreamed as part of
the kernel tree. Distros will be able to package and release these
schedulers with the kernel, allowing users to utilize these schedulers
out-of-the-box without requiring any additional work or dependencies such as
clang or building the scheduler programs themselves. Other schedulers and
scheduling frameworks such as rhone may be open-sourced through separate
per-project repos.

3. Rapid scheduler deployments

Rolling out kernel upgrades is a slow and iterative process. At a large scale
it can take months to roll a new kernel out to a fleet of servers. While this
latency is expected and inevitable for normal kernel upgrades, it can become
highly problematic when kernel changes are required to fix bugs. Livepatch [8]
is available to quickly roll out critical security fixes to large fleets, but
the scope of changes that can be applied with livepatching is fairly limited,
and would likely not be usable for patching scheduling policies. With
sched_ext, new scheduling policies can be rapidly rolled out to production
environments.

[8]: https://www.kernel.org/doc/html/latest/livepatch/livepatch.html

As an example, one of the variants of the L1 Terminal Fault (L1TF) [9]
vulnerability allows a VCPU running a VM to read arbitrary host kernel
memory for pages in L1 data cache. The solution was to implement core
scheduling, which ensures that tasks running as hypertwins have the same
“cookie”.

[9]: https://www.intel.com/content/www/us/en/architecture-and-technology/l1tf.html

While core scheduling works well, it took a long time to finalize and land
upstream. This long rollout period was painful, and required organizations to
make difficult choices amongst a bad set of options. Some companies such as
Google chose to implement and use their own custom L1TF-safe scheduler, others
chose to run without hyper-threading enabled, and yet others left
hyper-threading enabled and crossed their fingers.

Once core scheduling was upstream, organizations had to upgrade the kernels on
their entire fleets. As downtime is not an option for many, these upgrades had
to be gradually rolled out, which can take a very long time for large fleets.

An example of an sched_ext scheduler that illustrates core scheduling
semantics is scx_example_pair.bpf.c, which co-schedules pairs of tasks from
the same cgroup, and is resilient to L1TF vulnerabilities. While this example
scheduler is certainly not suitable for production in its current form, a
similar scheduler that is more performant and featureful could be written and
deployed if necessary.

Rapid scheduling deployments can similarly be useful to quickly roll-out new
scheduling features without requiring kernel upgrades. At Google, for example,
it was observed that some low-priority workloads were causing degraded
performance for higher-priority workloads due to consuming a disproportionate
share of memory bandwidth. While a temporary mitigation was to use sched
affinity to limit the footprint of this low-priority workload to a small
subset of CPUs, a preferable solution would be to implement a more featureful
task-priority mechanism which automatically throttles lower-priority tasks
which are causing memory contention for the rest of the system. Implementing
this in CFS and rolling it out to the fleet could take a very long time.

sched_ext would directly address these gaps. If another hardware bug or
resource contention issue comes in that requires scheduler support to
mitigate, sched_ext can be used to experiment with and test different
policies. Once a scheduler is available, it can quickly be rolled out to as
many hosts as necessary, and function as a stop-gap solution until a
longer-term mitigation is upstreamed.

How
---

sched_ext is a new sched_class which allows scheduling policies to be
implemented in BPF programs.

sched_ext leverages BPF’s struct_ops feature to define a structure which
exports function callbacks and flags to BPF programs that wish to implement
scheduling policies. The struct_ops structure exported by sched_ext is struct
sched_ext_ops, and is conceptually similar to struct sched_class. The role of
sched_ext is to map the complex sched_class callbacks to the more simple and
ergonomic struct sched_ext_ops callbacks.

Unlike some other BPF program types which have ABI requirements due to
exporting UAPIs, struct_ops has no ABI requirements whatsoever. This provides
us with the flexibility to change the APIs provided to schedulers as
necessary. BPF struct_ops is also already being used successfully in other
subsystems, such as in support of TCP congestion control.

The only struct_ops field that is required to be specified by a scheduler is
the ‘name’ field. Otherwise, sched_ext will provide sane default behavior,
such as automatically choosing an idle CPU on the task wakeup path if
.select_cpu() is missing.

*Dispatch queues*

To bridge the workflow imbalance between the scheduler core and sched_ext_ops
callbacks, sched_ext uses simple FIFOs called dispatch queues (dsq's). By
default, there is one global dsq (SCX_DSQ_GLOBAL), and one local per-CPU dsq
(SCX_DSQ_LOCAL). SCX_DSQ_GLOBAL is provided for convenience and need not be
used by a scheduler that doesn't require it. As described in more detail
below, SCX_DSQ_LOCAL is the per-CPU FIFO that sched_ext pulls from when
putting the next task on the CPU. The BPF scheduler can manage an arbitrary
number of dsq's using scx_bpf_create_dsq() and scx_bpf_destroy_dsq().

*Scheduling cycle*

The following briefly shows a typical workflow for how a waking task is
scheduled and executed.

1. When a task is waking up, .select_cpu() is the first operation invoked.
   This serves two purposes. It both allows a scheduler to optimize task
   placement by specifying a CPU where it expects the task to eventually be
   scheduled, and the latter is that the selected CPU will be woken if it’s
   idle.

2. Once the target CPU is selected, .enqueue() is invoked. It can make one of
   the following decisions:

   - Immediately dispatch the task to either the global dsq (SCX_DSQ_GLOBAL)
     or the current CPU’s local dsq (SCX_DSQ_LOCAL).

   - Immediately dispatch the task to a user-created dispatch queue.

   - Queue the task on the BPF side, e.g. in an rbtree map for a vruntime
     scheduler, with the intention of dispatching it at a later time from
     .dispatch().

3. When a CPU is ready to schedule, it first looks at its local dsq. If empty,
   it invokes .consume() which should make one or more scx_bpf_consume() calls
   to consume tasks from dsq's. If a scx_bpf_consume() call succeeds, the CPU
   has the next task to run and .consume() can return. If .consume() is not
   defined, sched_ext will by-default consume from only the built-in
   SCX_DSQ_GLOBAL dsq.

4. If there's still no task to run, .dispatch() is invoked which should make
   one or more scx_bpf_dispatch() calls to dispatch tasks from the BPF
   scheduler to one of the dsq's. If more than one task has been dispatched,
   go back to the previous consumption step.

*Verifying callback behavior*

sched_ext always verifies that any value returned from a callback is valid,
and will issue an error and unload the scheduler if it is not. For example, if
.select_cpu() returns an invalid CPU, or if an attempt is made to invoke the
scx_bpf_dispatch() with invalid enqueue flags. Furthermore, if a task remains
runnable for too long without being scheduled, sched_ext will detect it and
error-out the scheduler.

Closing Thoughts
----------------

Both Meta and Google have experimented quite a lot with schedulers in the last
several years. Google has benchmarked various workloads using user space
scheduling, and have achieved performance wins by trading off generality for
application specific needs. At Meta, we have not yet deployed sched_ext on any
production workloads, though our preliminary experiments indicate that
sched_ext would provide significant performance wins when deployed at scale.
If successfully upstreamed, we expect to leverage it extensively to run
various experiments and develop customized schedulers for a number of critical
workloads.

In closing, both Meta and Google believe that sched_ext will significantly
evolve how the broader community explores the scheduling problem space,
empowering continued improvement to the in-kernel scheduler, while also
enabling targeted policies for custom applications. We’ll be able to
experiment easier and faster, explore uncharted areas, and deploy emergency
scheduler changes when necessary. The same applies to anyone who wants to work
on the scheduler, including academia and specialized industries. sched_ext
will push forward the state of the art when it comes to scheduling and
performance in Linux.

Written By
----------

David Vernet <dvernet@meta.com>
Josh Don <joshdon@google.com>
Tejun Heo <tj@kernel.org>
Barret Rhoden <brho@google.com>

Supported By
------------

Paul Turner <pjt@google.com>
Neel Natu <neelnatu@google.com>
Patrick Bellasi <derkling@google.com>
Hao Luo <haoluo@google.com>
Dimitrios Skarlatos <dskarlat@cs.cmu.edu>

Patchset
--------

This patchset is on top of bpf/for-next as of 2023-06-07:

 67faabbde36b ("selftests/bpf: Add missing prototypes for several test kfuncs")

and contains the following patches:

NOTE: The doc added by 0032 contains a high-level overview and might be good
      place to start.

 0001-cgroup-Implement-cgroup_show_cftypes.patch
 0002-sched-Restructure-sched_class-order-sanity-checks-in.patch
 0003-sched-Allow-sched_cgroup_fork-to-fail-and-introduce-.patch
 0004-sched-Add-sched_class-reweight_task.patch
 0005-sched-Add-sched_class-switching_to-and-expose-check_.patch
 0006-sched-Factor-out-cgroup-weight-conversion-functions.patch
 0007-sched-Expose-css_tg-and-__setscheduler_prio.patch
 0008-sched-Enumerate-CPU-cgroup-file-types.patch
 0009-sched-Add-reason-to-sched_class-rq_-on-off-line.patch
 0010-sched-Add-normal_policy.patch
 0011-sched_ext-Add-boilerplate-for-extensible-scheduler-c.patch
 0012-sched_ext-Implement-BPF-extensible-scheduler-class.patch
 0013-sched_ext-Add-scx_simple-and-scx_example_qmap-exampl.patch
 0014-sched_ext-Add-sysrq-S-which-disables-the-BPF-schedul.patch
 0015-sched_ext-Implement-runnable-task-stall-watchdog.patch
 0016-sched_ext-Allow-BPF-schedulers-to-disallow-specific-.patch
 0017-sched_ext-Allow-BPF-schedulers-to-switch-all-eligibl.patch
 0018-sched_ext-Implement-scx_bpf_kick_cpu-and-task-preemp.patch
 0019-sched_ext-Add-a-central-scheduler-which-makes-all-sc.patch
 0020-sched_ext-Make-watchdog-handle-ops.dispatch-looping-.patch
 0021-sched_ext-Add-task-state-tracking-operations.patch
 0022-sched_ext-Implement-tickless-support.patch
 0023-sched_ext-Track-tasks-that-are-subjects-of-the-in-fl.patch
 0024-sched_ext-Add-cgroup-support.patch
 0025-sched_ext-Add-a-cgroup-based-core-scheduling-schedul.patch
 0026-sched_ext-Add-a-cgroup-scheduler-which-uses-flattene.patch
 0027-sched_ext-Implement-SCX_KICK_WAIT.patch
 0028-sched_ext-Implement-sched_ext_ops.cpu_acquire-releas.patch
 0029-sched_ext-Implement-sched_ext_ops.cpu_online-offline.patch
 0030-sched_ext-Implement-core-sched-support.patch
 0031-sched_ext-Add-vtime-ordered-priority-queue-to-dispat.patch
 0032-sched_ext-Documentation-scheduler-Document-extensibl.patch
 0033-sched_ext-Add-a-basic-userland-vruntime-scheduler.patch
 0034-sched_ext-Add-a-rust-userspace-hybrid-example-schedu.patch

0001     : Cgroup prep.

0002-0010: Scheduler prep.

0011-0013: sched_ext core implementation and a couple example BPF scheduler.

0014-0017: Utility features including safety mechanisms and switch-all.

0018-0023: Kicking and preempting other CPUs, task state transition tracking
           and tickless support. Demonstrated with an example central
           scheduler which makes all scheduling decisions on one CPU.

0024-0028: cgroup support and the ability to wait for other CPUs after
           kicking them. Demonstrated with an example pair scheduler which
           guarantees that a hyperthread pair always executes tasks from the
           same cgroup at any given time.

0029     : Add CPU hotplug callbacks.

0030     : Add core-sched support.

0031     : Add DSQ rbtree support.

0032     : Add documentation.

0033-0034: Add two example schedulers. One demonstrating deferring most
           scheduling decisions to userland. The other demonstrating a
           hybrid approach where load balancing decisions are made by
           userspace written in rust.

The patchset is also available in the following git branch:

 https://github.com/htejun/sched_ext sched_ext-v4

diffstat follows.

 Documentation/scheduler/index.rst                 |    1
 Documentation/scheduler/sched-ext.rst             |  229 ++
 MAINTAINERS                                       |    3
 drivers/tty/sysrq.c                               |    1
 include/asm-generic/vmlinux.lds.h                 |    1
 include/linux/cgroup-defs.h                       |    8
 include/linux/cgroup.h                            |    5
 include/linux/sched.h                             |    5
 include/linux/sched/ext.h                         |  708 +++++++
 include/linux/sched/task.h                        |    3
 include/uapi/linux/sched.h                        |    1
 init/Kconfig                                      |    5
 init/init_task.c                                  |   12
 kernel/Kconfig.preempt                            |   24
 kernel/bpf/bpf_struct_ops_types.h                 |    4
 kernel/cgroup/cgroup.c                            |   97 -
 kernel/fork.c                                     |   17
 kernel/sched/build_policy.c                       |    5
 kernel/sched/core.c                               |  315 ++-
 kernel/sched/deadline.c                           |    4
 kernel/sched/debug.c                              |    6
 kernel/sched/ext.c                                | 4427 ++++++++++++++++++++++++++++++++++++++++++++++++
 kernel/sched/ext.h                                |  266 ++
 kernel/sched/fair.c                               |    9
 kernel/sched/idle.c                               |    2
 kernel/sched/rt.c                                 |    4
 kernel/sched/sched.h                              |  117 +
 kernel/sched/topology.c                           |    4
 tools/sched_ext/.gitignore                        |    9
 tools/sched_ext/Makefile                          |  213 ++
 tools/sched_ext/README                            |  264 ++
 tools/sched_ext/gnu/stubs.h                       |    1
 tools/sched_ext/scx_atropos/.gitignore            |    3
 tools/sched_ext/scx_atropos/Cargo.toml            |   27
 tools/sched_ext/scx_atropos/build.rs              |   70
 tools/sched_ext/scx_atropos/rustfmt.toml          |    8
 tools/sched_ext/scx_atropos/src/atropos_sys.rs    |   10
 tools/sched_ext/scx_atropos/src/bpf/atropos.bpf.c |  978 ++++++++++
 tools/sched_ext/scx_atropos/src/bpf/atropos.h     |   64
 tools/sched_ext/scx_atropos/src/main.rs           | 1196 ++++++++++++
 tools/sched_ext/scx_central.bpf.c                 |  334 +++
 tools/sched_ext/scx_central.c                     |   94 +
 tools/sched_ext/scx_common.bpf.h                  |  293 +++
 tools/sched_ext/scx_flatcg.bpf.c                  |  904 +++++++++
 tools/sched_ext/scx_flatcg.c                      |  232 ++
 tools/sched_ext/scx_flatcg.h                      |   49
 tools/sched_ext/scx_pair.bpf.c                    |  627 ++++++
 tools/sched_ext/scx_pair.c                        |  164 +
 tools/sched_ext/scx_pair.h                        |   10
 tools/sched_ext/scx_qmap.bpf.c                    |  401 ++++
 tools/sched_ext/scx_qmap.c                        |  107 +
 tools/sched_ext/scx_simple.bpf.c                  |  135 +
 tools/sched_ext/scx_simple.c                      |  101 +
 tools/sched_ext/scx_userland.bpf.c                |  262 ++
 tools/sched_ext/scx_userland.c                    |  402 ++++
 tools/sched_ext/scx_userland.h                    |   19
 tools/sched_ext/user_exit_info.h                  |   50
 57 files changed, 13207 insertions(+), 103 deletions(-)

Thanks.

--
tejun

[PATCH 01/34] cgroup: Implement cgroup_show_cftypes()

[Tejun Heo]

Implement cgroup_show_cftypes() which shows and hides all cgroup files
associated with the specified set of cgroup file types. CFTYPE_HIDDEN flag
is added so that files can be created hidden from the get-go.

cgroup_show_cftypes() can be used whether the cftypes are added or not. It
also combines with cgroup_show_file() so that a given file is visible iff
both its cftype and cfile are visible.

This will be used by a new sched_class to selectively show and hide CPU
controller interface files depending on whether they're supported.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 include/linux/cgroup-defs.h |  8 +++
 include/linux/cgroup.h      |  1 +
 kernel/cgroup/cgroup.c      | 97 ++++++++++++++++++++++++++++++++++---
 3 files changed, 99 insertions(+), 7 deletions(-)

diff --git a/include/linux/cgroup-defs.h b/include/linux/cgroup-defs.h
index 8a0d5466c7be..8af1e7d487cb 100644
--- a/include/linux/cgroup-defs.h
+++ b/include/linux/cgroup-defs.h
@@ -127,12 +127,18 @@ enum {
 	CFTYPE_WORLD_WRITABLE	= (1 << 4),	/* (DON'T USE FOR NEW FILES) S_IWUGO */
 	CFTYPE_DEBUG		= (1 << 5),	/* create when cgroup_debug */
 
+	CFTYPE_HIDDEN		= (1 << 6),	/* file type hidden, see cgroup_show_cftypes() */
+
 	/* internal flags, do not use outside cgroup core proper */
 	__CFTYPE_ONLY_ON_DFL	= (1 << 16),	/* only on default hierarchy */
 	__CFTYPE_NOT_ON_DFL	= (1 << 17),	/* not on default hierarchy */
 	__CFTYPE_ADDED		= (1 << 18),
 };
 
+enum cfile_flags {
+	CFILE_HIDDEN		= (1 << 0),	/* file instance hidden */
+};
+
 /*
  * cgroup_file is the handle for a file instance created in a cgroup which
  * is used, for example, to generate file changed notifications.  This can
@@ -140,7 +146,9 @@ enum {
  */
 struct cgroup_file {
 	/* do not access any fields from outside cgroup core */
+	struct cftype *cft;
 	struct kernfs_node *kn;
+	unsigned int flags;
 	unsigned long notified_at;
 	struct timer_list notify_timer;
 };
diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h
index 885f5395fcd0..5d1e8432d9af 100644
--- a/include/linux/cgroup.h
+++ b/include/linux/cgroup.h
@@ -115,6 +115,7 @@ int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from);
 int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
 int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
 int cgroup_rm_cftypes(struct cftype *cfts);
+void cgroup_show_cftype(struct cftype *cft, bool show);
 void cgroup_file_notify(struct cgroup_file *cfile);
 void cgroup_file_show(struct cgroup_file *cfile, bool show);
 
diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
index 625d7483951c..bd04bdd5caeb 100644
--- a/kernel/cgroup/cgroup.c
+++ b/kernel/cgroup/cgroup.c
@@ -4206,10 +4206,13 @@ static int cgroup_add_file(struct cgroup_subsys_state *css, struct cgroup *cgrp,
 		return ret;
 	}
 
+	kernfs_show(kn, !(cft->flags & CFTYPE_HIDDEN));
+
 	if (cft->file_offset) {
 		struct cgroup_file *cfile = (void *)css + cft->file_offset;
 
 		timer_setup(&cfile->notify_timer, cgroup_file_notify_timer, 0);
+		cfile->cft = cft;
 
 		spin_lock_irq(&cgroup_file_kn_lock);
 		cfile->kn = kn;
@@ -4488,6 +4491,24 @@ void cgroup_file_notify(struct cgroup_file *cfile)
 	spin_unlock_irqrestore(&cgroup_file_kn_lock, flags);
 }
 
+static struct kernfs_node *cfile_kn_get(struct cgroup_file *cfile)
+{
+	struct kernfs_node *kn;
+
+	spin_lock_irq(&cgroup_file_kn_lock);
+	kn = cfile->kn;
+	kernfs_get(kn);
+	spin_unlock_irq(&cgroup_file_kn_lock);
+
+	return kn;
+}
+
+static bool cfile_visible(struct cgroup_file *cfile)
+{
+	return !(cfile->cft->flags & CFTYPE_HIDDEN) &&
+		!(cfile->flags & CFILE_HIDDEN);
+}
+
 /**
  * cgroup_file_show - show or hide a hidden cgroup file
  * @cfile: target cgroup_file obtained by setting cftype->file_offset
@@ -4497,15 +4518,20 @@ void cgroup_file_show(struct cgroup_file *cfile, bool show)
 {
 	struct kernfs_node *kn;
 
-	spin_lock_irq(&cgroup_file_kn_lock);
-	kn = cfile->kn;
-	kernfs_get(kn);
-	spin_unlock_irq(&cgroup_file_kn_lock);
+	mutex_lock(&cgroup_mutex);
 
-	if (kn)
-		kernfs_show(kn, show);
+	if (show)
+		cfile->flags &= ~CFILE_HIDDEN;
+	else
+		cfile->flags |= CFILE_HIDDEN;
 
-	kernfs_put(kn);
+	kn = cfile_kn_get(cfile);
+	if (kn) {
+		kernfs_show(kn, cfile_visible(cfile));
+		kernfs_put(kn);
+	}
+
+	mutex_unlock(&cgroup_mutex);
 }
 
 /**
@@ -5519,6 +5545,63 @@ static void offline_css(struct cgroup_subsys_state *css)
 	wake_up_all(&css->cgroup->offline_waitq);
 }
 
+/**
+ * cgroup_show_cftype - show or hide a cgroup file type
+ * @cft: cftype to show or hide
+ * @show: whether to show or hide
+ *
+ * Sets %CFTYPE_HIDDEN and shows/hides the matching files according to @show.
+ * @cft may or may not be added at the time of this call. After hiding, it's
+ * guaranteed that there are no in-flight operations on the hidden files.
+ */
+void cgroup_show_cftype(struct cftype *cft, bool show)
+{
+	struct cgroup_subsys *ss = cft->ss;
+	struct cgroup *root = ss ? &ss->root->cgrp : &cgrp_dfl_root.cgrp;
+	struct cgroup_subsys_state *css;
+
+	mutex_lock(&cgroup_mutex);
+
+	if (show)
+		cft->flags &= ~CFTYPE_HIDDEN;
+	else
+		cft->flags |= CFTYPE_HIDDEN;
+
+	if (!(cft->flags & __CFTYPE_ADDED))
+		goto out_unlock;
+
+	css_for_each_descendant_pre(css, cgroup_css(root, ss)) {
+		struct cgroup *cgrp = css->cgroup;
+		struct kernfs_node *kn;
+
+		if (!(css->flags & CSS_VISIBLE))
+			continue;
+
+		if (cft->file_offset) {
+			struct cgroup_file *cfile =
+				(void *)css + cft->file_offset;
+
+			kn = cfile_kn_get(cfile);
+			if (kn) {
+				kernfs_show(kn, cfile_visible(cfile));
+				kernfs_put(kn);
+			}
+		} else {
+			char buf[CGROUP_FILE_NAME_MAX];
+
+			kn = kernfs_find_and_get(cgrp->kn,
+					cgroup_file_name(cgrp, cft, buf));
+			if (kn) {
+				kernfs_show(kn, show);
+				kernfs_put(kn);
+			}
+		}
+	}
+
+out_unlock:
+	mutex_unlock(&cgroup_mutex);
+}
+
 /**
  * css_create - create a cgroup_subsys_state
  * @cgrp: the cgroup new css will be associated with
-- 
2.41.0

[PATCH 02/34] sched: Restructure sched_class order sanity checks in sched_init()

[Tejun Heo]

Currently, sched_init() checks that the sched_class'es are in the expected
order by testing each adjacency which is a bit brittle and makes it
cumbersome to add optional sched_class'es. Instead, let's verify whether
they're in the expected order using sched_class_above() which is what
matters.

Signed-off-by: Tejun Heo <tj@kernel.org>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Reviewed-by: David Vernet <dvernet@meta.com>
---
 kernel/sched/core.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index a68d1276bab0..173a42336d54 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -9878,12 +9878,12 @@ void __init sched_init(void)
 	int i;
 
 	/* Make sure the linker didn't screw up */
-	BUG_ON(&idle_sched_class != &fair_sched_class + 1 ||
-	       &fair_sched_class != &rt_sched_class + 1 ||
-	       &rt_sched_class   != &dl_sched_class + 1);
 #ifdef CONFIG_SMP
-	BUG_ON(&dl_sched_class != &stop_sched_class + 1);
+	BUG_ON(!sched_class_above(&stop_sched_class, &dl_sched_class));
 #endif
+	BUG_ON(!sched_class_above(&dl_sched_class, &rt_sched_class));
+	BUG_ON(!sched_class_above(&rt_sched_class, &fair_sched_class));
+	BUG_ON(!sched_class_above(&fair_sched_class, &idle_sched_class));
 
 	wait_bit_init();
 
-- 
2.41.0

[PATCH 03/34] sched: Allow sched_cgroup_fork() to fail and introduce sched_cancel_fork()

[Tejun Heo]

A new BPF extensible sched_class will need more control over the forking
process. It wants to be able to fail from sched_cgroup_fork() after the new
task's sched_task_group is initialized so that the loaded BPF program can
prepare the task with its cgroup association is established and reject fork
if e.g. allocation fails.

Allow sched_cgroup_fork() to fail by making it return int instead of void
and adding sched_cancel_fork() to undo sched_fork() in the error path.

sched_cgroup_fork() doesn't fail yet and this patch shouldn't cause any
behavior changes.

v2: Patch description updated to detail the expected use.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 include/linux/sched/task.h |  3 ++-
 kernel/fork.c              | 15 ++++++++++-----
 kernel/sched/core.c        |  8 +++++++-
 3 files changed, 19 insertions(+), 7 deletions(-)

diff --git a/include/linux/sched/task.h b/include/linux/sched/task.h
index 537cbf9a2ade..5431b350ed96 100644
--- a/include/linux/sched/task.h
+++ b/include/linux/sched/task.h
@@ -62,7 +62,8 @@ extern asmlinkage void schedule_tail(struct task_struct *prev);
 extern void init_idle(struct task_struct *idle, int cpu);
 
 extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
-extern void sched_cgroup_fork(struct task_struct *p, struct kernel_clone_args *kargs);
+extern int sched_cgroup_fork(struct task_struct *p, struct kernel_clone_args *kargs);
+extern void sched_cancel_fork(struct task_struct *p);
 extern void sched_post_fork(struct task_struct *p);
 extern void sched_dead(struct task_struct *p);
 
diff --git a/kernel/fork.c b/kernel/fork.c
index ed4e01daccaa..0a9c8c03889f 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -2480,7 +2480,7 @@ __latent_entropy struct task_struct *copy_process(
 
 	retval = perf_event_init_task(p, clone_flags);
 	if (retval)
-		goto bad_fork_cleanup_policy;
+		goto bad_fork_sched_cancel_fork;
 	retval = audit_alloc(p);
 	if (retval)
 		goto bad_fork_cleanup_perf;
@@ -2615,7 +2615,9 @@ __latent_entropy struct task_struct *copy_process(
 	 * cgroup specific, it unconditionally needs to place the task on a
 	 * runqueue.
 	 */
-	sched_cgroup_fork(p, args);
+	retval = sched_cgroup_fork(p, args);
+	if (retval)
+		goto bad_fork_cancel_cgroup;
 
 	/*
 	 * From this point on we must avoid any synchronous user-space
@@ -2661,13 +2663,13 @@ __latent_entropy struct task_struct *copy_process(
 	/* Don't start children in a dying pid namespace */
 	if (unlikely(!(ns_of_pid(pid)->pid_allocated & PIDNS_ADDING))) {
 		retval = -ENOMEM;
-		goto bad_fork_cancel_cgroup;
+		goto bad_fork_core_free;
 	}
 
 	/* Let kill terminate clone/fork in the middle */
 	if (fatal_signal_pending(current)) {
 		retval = -EINTR;
-		goto bad_fork_cancel_cgroup;
+		goto bad_fork_core_free;
 	}
 
 	/* No more failure paths after this point. */
@@ -2743,10 +2745,11 @@ __latent_entropy struct task_struct *copy_process(
 
 	return p;
 
-bad_fork_cancel_cgroup:
+bad_fork_core_free:
 	sched_core_free(p);
 	spin_unlock(&current->sighand->siglock);
 	write_unlock_irq(&tasklist_lock);
+bad_fork_cancel_cgroup:
 	cgroup_cancel_fork(p, args);
 bad_fork_put_pidfd:
 	if (clone_flags & CLONE_PIDFD) {
@@ -2785,6 +2788,8 @@ __latent_entropy struct task_struct *copy_process(
 	audit_free(p);
 bad_fork_cleanup_perf:
 	perf_event_free_task(p);
+bad_fork_sched_cancel_fork:
+	sched_cancel_fork(p);
 bad_fork_cleanup_policy:
 	lockdep_free_task(p);
 #ifdef CONFIG_NUMA
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 173a42336d54..a869236d0735 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4749,7 +4749,7 @@ int sched_fork(unsigned long clone_flags, struct task_struct *p)
 	return 0;
 }
 
-void sched_cgroup_fork(struct task_struct *p, struct kernel_clone_args *kargs)
+int sched_cgroup_fork(struct task_struct *p, struct kernel_clone_args *kargs)
 {
 	unsigned long flags;
 
@@ -4776,6 +4776,12 @@ void sched_cgroup_fork(struct task_struct *p, struct kernel_clone_args *kargs)
 	if (p->sched_class->task_fork)
 		p->sched_class->task_fork(p);
 	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
+
+	return 0;
+}
+
+void sched_cancel_fork(struct task_struct *p)
+{
 }
 
 void sched_post_fork(struct task_struct *p)
-- 
2.41.0

[PATCH 04/34] sched: Add sched_class->reweight_task()

[Tejun Heo]

Currently, during a task weight change, sched core directly calls
reweight_task() defined in fair.c if @p is on CFS. Let's make it a proper
sched_class operation instead. CFS's reweight_task() is renamed to
reweight_task_fair() and now called through sched_class.

While it turns a direct call into an indirect one, set_load_weight() isn't
called from a hot path and this change shouldn't cause any noticeable
difference. This will be used to implement reweight_task for a new BPF
extensible sched_class so that it can keep its cached task weight
up-to-date.

This will be used by a new sched_class to track weight changes.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 kernel/sched/core.c  | 4 ++--
 kernel/sched/fair.c  | 3 ++-
 kernel/sched/sched.h | 4 ++--
 3 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index a869236d0735..a205d00f0669 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -1297,8 +1297,8 @@ static void set_load_weight(struct task_struct *p, bool update_load)
 	 * SCHED_OTHER tasks have to update their load when changing their
 	 * weight
 	 */
-	if (update_load && p->sched_class == &fair_sched_class) {
-		reweight_task(p, prio);
+	if (update_load && p->sched_class->reweight_task) {
+		p->sched_class->reweight_task(task_rq(p), p, prio);
 	} else {
 		load->weight = scale_load(sched_prio_to_weight[prio]);
 		load->inv_weight = sched_prio_to_wmult[prio];
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 373ff5f55884..c0818324a9cd 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -3399,7 +3399,7 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
 
 }
 
-void reweight_task(struct task_struct *p, int prio)
+static void reweight_task_fair(struct rq *rq, struct task_struct *p, int prio)
 {
 	struct sched_entity *se = &p->se;
 	struct cfs_rq *cfs_rq = cfs_rq_of(se);
@@ -12663,6 +12663,7 @@ DEFINE_SCHED_CLASS(fair) = {
 	.task_tick		= task_tick_fair,
 	.task_fork		= task_fork_fair,
 
+	.reweight_task		= reweight_task_fair,
 	.prio_changed		= prio_changed_fair,
 	.switched_from		= switched_from_fair,
 	.switched_to		= switched_to_fair,
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index ec7b3e0a2b20..1a12a0d9ea40 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -2213,6 +2213,8 @@ struct sched_class {
 	 */
 	void (*switched_from)(struct rq *this_rq, struct task_struct *task);
 	void (*switched_to)  (struct rq *this_rq, struct task_struct *task);
+	void (*reweight_task)(struct rq *this_rq, struct task_struct *task,
+			      int newprio);
 	void (*prio_changed) (struct rq *this_rq, struct task_struct *task,
 			      int oldprio);
 
@@ -2369,8 +2371,6 @@ extern void init_sched_dl_class(void);
 extern void init_sched_rt_class(void);
 extern void init_sched_fair_class(void);
 
-extern void reweight_task(struct task_struct *p, int prio);
-
 extern void resched_curr(struct rq *rq);
 extern void resched_cpu(int cpu);
 
-- 
2.41.0

[PATCH 05/34] sched: Add sched_class->switching_to() and expose check_class_changing/changed()

[Tejun Heo]

When a task switches to a new sched_class, the prev and new classes are
notified through ->switched_from() and ->switched_to(), respectively, after
the switching is done.

A new BPF extensible sched_class will have callbacks that allow the BPF
scheduler to keep track of relevant task states (like priority and cpumask).
Those callbacks aren't called while a task is on a different sched_class.
When a task comes back, we wanna tell the BPF progs the up-to-date state
before the task gets enqueued, so we need a hook which is called before the
switching is committed.

This patch adds ->switching_to() which is called during sched_class switch
through check_class_changing() before the task is restored. Also, this patch
exposes check_class_changing/changed() in kernel/sched/sched.h. They will be
used by the new BPF extensible sched_class to implement implicit sched_class
switching which is used e.g. when falling back to CFS when the BPF scheduler
fails or unloads.

This is a prep patch and doesn't cause any behavior changes. The new
operation and exposed functions aren't used yet.

v2: Improve patch description w/ details on planned use.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 kernel/sched/core.c  | 19 ++++++++++++++++---
 kernel/sched/sched.h |  7 +++++++
 2 files changed, 23 insertions(+), 3 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index a205d00f0669..29f75a31b414 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2178,6 +2178,17 @@ inline int task_curr(const struct task_struct *p)
 	return cpu_curr(task_cpu(p)) == p;
 }
 
+/*
+ * ->switching_to() is called with the pi_lock and rq_lock held and must not
+ * mess with locking.
+ */
+void check_class_changing(struct rq *rq, struct task_struct *p,
+			  const struct sched_class *prev_class)
+{
+	if (prev_class != p->sched_class && p->sched_class->switching_to)
+		p->sched_class->switching_to(rq, p);
+}
+
 /*
  * switched_from, switched_to and prio_changed must _NOT_ drop rq->lock,
  * use the balance_callback list if you want balancing.
@@ -2185,9 +2196,9 @@ inline int task_curr(const struct task_struct *p)
  * this means any call to check_class_changed() must be followed by a call to
  * balance_callback().
  */
-static inline void check_class_changed(struct rq *rq, struct task_struct *p,
-				       const struct sched_class *prev_class,
-				       int oldprio)
+void check_class_changed(struct rq *rq, struct task_struct *p,
+			 const struct sched_class *prev_class,
+			 int oldprio)
 {
 	if (prev_class != p->sched_class) {
 		if (prev_class->switched_from)
@@ -7140,6 +7151,7 @@ void rt_mutex_setprio(struct task_struct *p, struct task_struct *pi_task)
 	}
 
 	__setscheduler_prio(p, prio);
+	check_class_changing(rq, p, prev_class);
 
 	if (queued)
 		enqueue_task(rq, p, queue_flag);
@@ -7768,6 +7780,7 @@ static int __sched_setscheduler(struct task_struct *p,
 		__setscheduler_prio(p, newprio);
 	}
 	__setscheduler_uclamp(p, attr);
+	check_class_changing(rq, p, prev_class);
 
 	if (queued) {
 		/*
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 1a12a0d9ea40..0a25bcd8d9d1 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -2211,6 +2211,7 @@ struct sched_class {
 	 * cannot assume the switched_from/switched_to pair is serialized by
 	 * rq->lock. They are however serialized by p->pi_lock.
 	 */
+	void (*switching_to) (struct rq *this_rq, struct task_struct *task);
 	void (*switched_from)(struct rq *this_rq, struct task_struct *task);
 	void (*switched_to)  (struct rq *this_rq, struct task_struct *task);
 	void (*reweight_task)(struct rq *this_rq, struct task_struct *task,
@@ -2451,6 +2452,12 @@ static inline void sub_nr_running(struct rq *rq, unsigned count)
 extern void activate_task(struct rq *rq, struct task_struct *p, int flags);
 extern void deactivate_task(struct rq *rq, struct task_struct *p, int flags);
 
+extern void check_class_changing(struct rq *rq, struct task_struct *p,
+				 const struct sched_class *prev_class);
+extern void check_class_changed(struct rq *rq, struct task_struct *p,
+				const struct sched_class *prev_class,
+				int oldprio);
+
 extern void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags);
 
 #ifdef CONFIG_PREEMPT_RT
-- 
2.41.0

[PATCH 06/34] sched: Factor out cgroup weight conversion functions

[Tejun Heo]

Factor out sched_weight_from/to_cgroup() which convert between scheduler
shares and cgroup weight. No functional change. The factored out functions
will be used by a new BPF extensible sched_class so that the weights can be
exposed to the BPF programs in a way which is consistent cgroup weights and
easier to interpret.

The weight conversions will be used regardless of cgroup usage. It's just
borrowing the cgroup weight range as it's more intuitive.
CGROUP_WEIGHT_MIN/DFL/MAX constants are moved outside CONFIG_CGROUPS so that
the conversion helpers can always be defined.

v2: The helpers are now defined regardless of COFNIG_CGROUPS.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 include/linux/cgroup.h |  4 ++--
 kernel/sched/core.c    | 28 +++++++++++++---------------
 kernel/sched/sched.h   | 18 ++++++++++++++++++
 3 files changed, 33 insertions(+), 17 deletions(-)

diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h
index 5d1e8432d9af..96d81dfb143f 100644
--- a/include/linux/cgroup.h
+++ b/include/linux/cgroup.h
@@ -29,8 +29,6 @@
 
 struct kernel_clone_args;
 
-#ifdef CONFIG_CGROUPS
-
 /*
  * All weight knobs on the default hierarchy should use the following min,
  * default and max values.  The default value is the logarithmic center of
@@ -40,6 +38,8 @@ struct kernel_clone_args;
 #define CGROUP_WEIGHT_DFL		100
 #define CGROUP_WEIGHT_MAX		10000
 
+#ifdef CONFIG_CGROUPS
+
 /* walk only threadgroup leaders */
 #define CSS_TASK_ITER_PROCS		(1U << 0)
 /* walk all threaded css_sets in the domain */
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 29f75a31b414..00d6751b0f8d 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -11255,29 +11255,27 @@ static int cpu_extra_stat_show(struct seq_file *sf,
 }
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
+
+static unsigned long tg_weight(struct task_group *tg)
+{
+	return scale_load_down(tg->shares);
+}
+
 static u64 cpu_weight_read_u64(struct cgroup_subsys_state *css,
 			       struct cftype *cft)
 {
-	struct task_group *tg = css_tg(css);
-	u64 weight = scale_load_down(tg->shares);
-
-	return DIV_ROUND_CLOSEST_ULL(weight * CGROUP_WEIGHT_DFL, 1024);
+	return sched_weight_to_cgroup(tg_weight(css_tg(css)));
 }
 
 static int cpu_weight_write_u64(struct cgroup_subsys_state *css,
-				struct cftype *cft, u64 weight)
+				struct cftype *cft, u64 cgrp_weight)
 {
-	/*
-	 * cgroup weight knobs should use the common MIN, DFL and MAX
-	 * values which are 1, 100 and 10000 respectively.  While it loses
-	 * a bit of range on both ends, it maps pretty well onto the shares
-	 * value used by scheduler and the round-trip conversions preserve
-	 * the original value over the entire range.
-	 */
-	if (weight < CGROUP_WEIGHT_MIN || weight > CGROUP_WEIGHT_MAX)
+	unsigned long weight;
+
+	if (cgrp_weight < CGROUP_WEIGHT_MIN || cgrp_weight > CGROUP_WEIGHT_MAX)
 		return -ERANGE;
 
-	weight = DIV_ROUND_CLOSEST_ULL(weight * 1024, CGROUP_WEIGHT_DFL);
+	weight = sched_weight_from_cgroup(cgrp_weight);
 
 	return sched_group_set_shares(css_tg(css), scale_load(weight));
 }
@@ -11285,7 +11283,7 @@ static int cpu_weight_write_u64(struct cgroup_subsys_state *css,
 static s64 cpu_weight_nice_read_s64(struct cgroup_subsys_state *css,
 				    struct cftype *cft)
 {
-	unsigned long weight = scale_load_down(css_tg(css)->shares);
+	unsigned long weight = tg_weight(css_tg(css));
 	int last_delta = INT_MAX;
 	int prio, delta;
 
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 0a25bcd8d9d1..2eb5759f4be9 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -232,6 +232,24 @@ static inline void update_avg(u64 *avg, u64 sample)
 #define shr_bound(val, shift)							\
 	(val >> min_t(typeof(shift), shift, BITS_PER_TYPE(typeof(val)) - 1))
 
+/*
+ * cgroup weight knobs should use the common MIN, DFL and MAX values which are
+ * 1, 100 and 10000 respectively. While it loses a bit of range on both ends, it
+ * maps pretty well onto the shares value used by scheduler and the round-trip
+ * conversions preserve the original value over the entire range.
+ */
+static inline unsigned long sched_weight_from_cgroup(unsigned long cgrp_weight)
+{
+	return DIV_ROUND_CLOSEST_ULL(cgrp_weight * 1024, CGROUP_WEIGHT_DFL);
+}
+
+static inline unsigned long sched_weight_to_cgroup(unsigned long weight)
+{
+	return clamp_t(unsigned long,
+		       DIV_ROUND_CLOSEST_ULL(weight * CGROUP_WEIGHT_DFL, 1024),
+		       CGROUP_WEIGHT_MIN, CGROUP_WEIGHT_MAX);
+}
+
 /*
  * !! For sched_setattr_nocheck() (kernel) only !!
  *
-- 
2.41.0

[PATCH 07/34] sched: Expose css_tg() and __setscheduler_prio()

[Tejun Heo]

These will be used by a new BPF extensible sched_class.

css_tg() will be used in the init and exit paths to visit all task_groups by
walking cgroups.

__setscheduler_prio() is used to pick the sched_class matching the current
prio of the task. For the new BPF extensible sched_class, the mapping from
the task configuration to sched_class isn't static and depends on a few
factors - e.g. whether the BPF progs implementing the scheduler are loaded
and in a serviceable state. That mapping logic will be added to
__setscheduler_prio().

When the BPF scheduler progs get loaded and unloaded, the mapping changes
and the new sched_class will walk the tasks applying the new mapping using
__setscheduler_prio().

v3: Dropped SCHED_CHANGE_BLOCK() as upstream is adding more generic cleanup
    mechanism.

v2: Expose SCHED_CHANGE_BLOCK() too and update the description.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
Reported-by: kernel test robot <lkp@intel.com>
---
 kernel/sched/core.c  | 7 +------
 kernel/sched/sched.h | 7 +++++++
 2 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 00d6751b0f8d..6818ed1a7a42 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -7011,7 +7011,7 @@ int default_wake_function(wait_queue_entry_t *curr, unsigned mode, int wake_flag
 }
 EXPORT_SYMBOL(default_wake_function);
 
-static void __setscheduler_prio(struct task_struct *p, int prio)
+void __setscheduler_prio(struct task_struct *p, int prio)
 {
 	if (dl_prio(prio))
 		p->sched_class = &dl_sched_class;
@@ -10523,11 +10523,6 @@ void sched_move_task(struct task_struct *tsk)
 	task_rq_unlock(rq, tsk, &rf);
 }
 
-static inline struct task_group *css_tg(struct cgroup_subsys_state *css)
-{
-	return css ? container_of(css, struct task_group, css) : NULL;
-}
-
 static struct cgroup_subsys_state *
 cpu_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
 {
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 2eb5759f4be9..71343472e6b1 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -469,6 +469,11 @@ static inline int walk_tg_tree(tg_visitor down, tg_visitor up, void *data)
 	return walk_tg_tree_from(&root_task_group, down, up, data);
 }
 
+static inline struct task_group *css_tg(struct cgroup_subsys_state *css)
+{
+	return css ? container_of(css, struct task_group, css) : NULL;
+}
+
 extern int tg_nop(struct task_group *tg, void *data);
 
 extern void free_fair_sched_group(struct task_group *tg);
@@ -2390,6 +2395,8 @@ extern void init_sched_dl_class(void);
 extern void init_sched_rt_class(void);
 extern void init_sched_fair_class(void);
 
+extern void __setscheduler_prio(struct task_struct *p, int prio);
+
 extern void resched_curr(struct rq *rq);
 extern void resched_cpu(int cpu);
 
-- 
2.41.0

[PATCH 08/34] sched: Enumerate CPU cgroup file types

[Tejun Heo]

Rename cpu[_legacy]_files to cpu[_legacy]_cftypes for clarity and add
cpu_cftype_id which enumerates every cgroup2 interface file type. This
doesn't make any functional difference now. The enums will be used to access
specific cftypes by a new BPF extensible sched_class to selectively show and
hide CPU controller interface files depending on the capability of the
currently loaded BPF scheduler progs.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 kernel/sched/core.c  | 22 +++++++++++-----------
 kernel/sched/sched.h | 21 +++++++++++++++++++++
 2 files changed, 32 insertions(+), 11 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 6818ed1a7a42..f7e7a25a0fdf 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -11160,7 +11160,7 @@ static int cpu_idle_write_s64(struct cgroup_subsys_state *css,
 }
 #endif
 
-static struct cftype cpu_legacy_files[] = {
+static struct cftype cpu_legacy_cftypes[] = {
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	{
 		.name = "shares",
@@ -11367,21 +11367,21 @@ static ssize_t cpu_max_write(struct kernfs_open_file *of,
 }
 #endif
 
-static struct cftype cpu_files[] = {
+struct cftype cpu_cftypes[CPU_CFTYPE_CNT + 1] = {
 #ifdef CONFIG_FAIR_GROUP_SCHED
-	{
+	[CPU_CFTYPE_WEIGHT] = {
 		.name = "weight",
 		.flags = CFTYPE_NOT_ON_ROOT,
 		.read_u64 = cpu_weight_read_u64,
 		.write_u64 = cpu_weight_write_u64,
 	},
-	{
+	[CPU_CFTYPE_WEIGHT_NICE] = {
 		.name = "weight.nice",
 		.flags = CFTYPE_NOT_ON_ROOT,
 		.read_s64 = cpu_weight_nice_read_s64,
 		.write_s64 = cpu_weight_nice_write_s64,
 	},
-	{
+	[CPU_CFTYPE_IDLE] = {
 		.name = "idle",
 		.flags = CFTYPE_NOT_ON_ROOT,
 		.read_s64 = cpu_idle_read_s64,
@@ -11389,13 +11389,13 @@ static struct cftype cpu_files[] = {
 	},
 #endif
 #ifdef CONFIG_CFS_BANDWIDTH
-	{
+	[CPU_CFTYPE_MAX] = {
 		.name = "max",
 		.flags = CFTYPE_NOT_ON_ROOT,
 		.seq_show = cpu_max_show,
 		.write = cpu_max_write,
 	},
-	{
+	[CPU_CFTYPE_MAX_BURST] = {
 		.name = "max.burst",
 		.flags = CFTYPE_NOT_ON_ROOT,
 		.read_u64 = cpu_cfs_burst_read_u64,
@@ -11403,13 +11403,13 @@ static struct cftype cpu_files[] = {
 	},
 #endif
 #ifdef CONFIG_UCLAMP_TASK_GROUP
-	{
+	[CPU_CFTYPE_UCLAMP_MIN] = {
 		.name = "uclamp.min",
 		.flags = CFTYPE_NOT_ON_ROOT,
 		.seq_show = cpu_uclamp_min_show,
 		.write = cpu_uclamp_min_write,
 	},
-	{
+	[CPU_CFTYPE_UCLAMP_MAX] = {
 		.name = "uclamp.max",
 		.flags = CFTYPE_NOT_ON_ROOT,
 		.seq_show = cpu_uclamp_max_show,
@@ -11429,8 +11429,8 @@ struct cgroup_subsys cpu_cgrp_subsys = {
 	.can_attach	= cpu_cgroup_can_attach,
 #endif
 	.attach		= cpu_cgroup_attach,
-	.legacy_cftypes	= cpu_legacy_files,
-	.dfl_cftypes	= cpu_files,
+	.legacy_cftypes	= cpu_legacy_cftypes,
+	.dfl_cftypes	= cpu_cftypes,
 	.early_init	= true,
 	.threaded	= true,
 };
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 71343472e6b1..eddc3775cc92 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -3519,4 +3519,25 @@ static inline void task_tick_mm_cid(struct rq *rq, struct task_struct *curr) { }
 static inline void init_sched_mm_cid(struct task_struct *t) { }
 #endif
 
+#ifdef CONFIG_CGROUP_SCHED
+enum cpu_cftype_id {
+#ifdef CONFIG_FAIR_GROUP_SCHED
+	CPU_CFTYPE_WEIGHT,
+	CPU_CFTYPE_WEIGHT_NICE,
+	CPU_CFTYPE_IDLE,
+#endif
+#ifdef CONFIG_CFS_BANDWIDTH
+	CPU_CFTYPE_MAX,
+	CPU_CFTYPE_MAX_BURST,
+#endif
+#ifdef CONFIG_UCLAMP_TASK_GROUP
+	CPU_CFTYPE_UCLAMP_MIN,
+	CPU_CFTYPE_UCLAMP_MAX,
+#endif
+	CPU_CFTYPE_CNT,
+};
+
+extern struct cftype cpu_cftypes[CPU_CFTYPE_CNT + 1];
+#endif /* CONFIG_CGROUP_SCHED */
+
 #endif /* _KERNEL_SCHED_SCHED_H */
-- 
2.41.0

[PATCH 09/34] sched: Add @reason to sched_class->rq_{on|off}line()

[Tejun Heo]

->rq_{on|off}line are called either during CPU hotplug or cpuset partition
updates. A planned BPF extensible sched_class wants to tell the BPF
scheduler progs about CPU hotplug events in a way that's synchronized with
rq state changes.

As the BPF scheduler progs aren't necessarily affected by cpuset partition
updates, we need a way to distinguish the two types of events. Let's add an
argument to tell them apart.

v2: Patch description updated to detail the expected use.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 kernel/sched/core.c     | 12 ++++++------
 kernel/sched/deadline.c |  4 ++--
 kernel/sched/fair.c     |  4 ++--
 kernel/sched/rt.c       |  4 ++--
 kernel/sched/sched.h    | 13 +++++++++----
 kernel/sched/topology.c |  4 ++--
 6 files changed, 23 insertions(+), 18 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index f7e7a25a0fdf..297740c4a5bc 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -9547,7 +9547,7 @@ static inline void balance_hotplug_wait(void)
 
 #endif /* CONFIG_HOTPLUG_CPU */
 
-void set_rq_online(struct rq *rq)
+void set_rq_online(struct rq *rq, enum rq_onoff_reason reason)
 {
 	if (!rq->online) {
 		const struct sched_class *class;
@@ -9557,19 +9557,19 @@ void set_rq_online(struct rq *rq)
 
 		for_each_class(class) {
 			if (class->rq_online)
-				class->rq_online(rq);
+				class->rq_online(rq, reason);
 		}
 	}
 }
 
-void set_rq_offline(struct rq *rq)
+void set_rq_offline(struct rq *rq, enum rq_onoff_reason reason)
 {
 	if (rq->online) {
 		const struct sched_class *class;
 
 		for_each_class(class) {
 			if (class->rq_offline)
-				class->rq_offline(rq);
+				class->rq_offline(rq, reason);
 		}
 
 		cpumask_clear_cpu(rq->cpu, rq->rd->online);
@@ -9665,7 +9665,7 @@ int sched_cpu_activate(unsigned int cpu)
 	rq_lock_irqsave(rq, &rf);
 	if (rq->rd) {
 		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
-		set_rq_online(rq);
+		set_rq_online(rq, RQ_ONOFF_HOTPLUG);
 	}
 	rq_unlock_irqrestore(rq, &rf);
 
@@ -9710,7 +9710,7 @@ int sched_cpu_deactivate(unsigned int cpu)
 	if (rq->rd) {
 		update_rq_clock(rq);
 		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
-		set_rq_offline(rq);
+		set_rq_offline(rq, RQ_ONOFF_HOTPLUG);
 	}
 	rq_unlock_irqrestore(rq, &rf);
 
diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 5a9a4b81c972..56b5fc9b0530 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -2519,7 +2519,7 @@ static void set_cpus_allowed_dl(struct task_struct *p,
 }
 
 /* Assumes rq->lock is held */
-static void rq_online_dl(struct rq *rq)
+static void rq_online_dl(struct rq *rq, enum rq_onoff_reason reason)
 {
 	if (rq->dl.overloaded)
 		dl_set_overload(rq);
@@ -2530,7 +2530,7 @@ static void rq_online_dl(struct rq *rq)
 }
 
 /* Assumes rq->lock is held */
-static void rq_offline_dl(struct rq *rq)
+static void rq_offline_dl(struct rq *rq, enum rq_onoff_reason reason)
 {
 	if (rq->dl.overloaded)
 		dl_clear_overload(rq);
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index c0818324a9cd..9838c82230e1 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -11909,14 +11909,14 @@ void trigger_load_balance(struct rq *rq)
 	nohz_balancer_kick(rq);
 }
 
-static void rq_online_fair(struct rq *rq)
+static void rq_online_fair(struct rq *rq, enum rq_onoff_reason reason)
 {
 	update_sysctl();
 
 	update_runtime_enabled(rq);
 }
 
-static void rq_offline_fair(struct rq *rq)
+static void rq_offline_fair(struct rq *rq, enum rq_onoff_reason reason)
 {
 	update_sysctl();
 
diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 00e0e5074115..5c15d017f762 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -2477,7 +2477,7 @@ static void task_woken_rt(struct rq *rq, struct task_struct *p)
 }
 
 /* Assumes rq->lock is held */
-static void rq_online_rt(struct rq *rq)
+static void rq_online_rt(struct rq *rq, enum rq_onoff_reason reason)
 {
 	if (rq->rt.overloaded)
 		rt_set_overload(rq);
@@ -2488,7 +2488,7 @@ static void rq_online_rt(struct rq *rq)
 }
 
 /* Assumes rq->lock is held */
-static void rq_offline_rt(struct rq *rq)
+static void rq_offline_rt(struct rq *rq, enum rq_onoff_reason reason)
 {
 	if (rq->rt.overloaded)
 		rt_clear_overload(rq);
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index eddc3775cc92..cd11bf9de7f9 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -2183,6 +2183,11 @@ extern const u32		sched_prio_to_wmult[40];
 
 #define RETRY_TASK		((void *)-1UL)
 
+enum rq_onoff_reason {
+	RQ_ONOFF_HOTPLUG,		/* CPU is going on/offline */
+	RQ_ONOFF_TOPOLOGY,		/* sched domain topology update */
+};
+
 struct affinity_context {
 	const struct cpumask *new_mask;
 	struct cpumask *user_mask;
@@ -2219,8 +2224,8 @@ struct sched_class {
 
 	void (*set_cpus_allowed)(struct task_struct *p, struct affinity_context *ctx);
 
-	void (*rq_online)(struct rq *rq);
-	void (*rq_offline)(struct rq *rq);
+	void (*rq_online)(struct rq *rq, enum rq_onoff_reason reason);
+	void (*rq_offline)(struct rq *rq, enum rq_onoff_reason reason);
 
 	struct rq *(*find_lock_rq)(struct task_struct *p, struct rq *rq);
 #endif
@@ -2750,8 +2755,8 @@ static inline void double_rq_unlock(struct rq *rq1, struct rq *rq2)
 	raw_spin_rq_unlock(rq1);
 }
 
-extern void set_rq_online (struct rq *rq);
-extern void set_rq_offline(struct rq *rq);
+extern void set_rq_online (struct rq *rq, enum rq_onoff_reason reason);
+extern void set_rq_offline(struct rq *rq, enum rq_onoff_reason reason);
 extern bool sched_smp_initialized;
 
 #else /* CONFIG_SMP */
diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c
index 6682535e37c8..e9a1d61eb186 100644
--- a/kernel/sched/topology.c
+++ b/kernel/sched/topology.c
@@ -495,7 +495,7 @@ void rq_attach_root(struct rq *rq, struct root_domain *rd)
 		old_rd = rq->rd;
 
 		if (cpumask_test_cpu(rq->cpu, old_rd->online))
-			set_rq_offline(rq);
+			set_rq_offline(rq, RQ_ONOFF_TOPOLOGY);
 
 		cpumask_clear_cpu(rq->cpu, old_rd->span);
 
@@ -513,7 +513,7 @@ void rq_attach_root(struct rq *rq, struct root_domain *rd)
 
 	cpumask_set_cpu(rq->cpu, rd->span);
 	if (cpumask_test_cpu(rq->cpu, cpu_active_mask))
-		set_rq_online(rq);
+		set_rq_online(rq, RQ_ONOFF_TOPOLOGY);
 
 	raw_spin_rq_unlock_irqrestore(rq, flags);
 
-- 
2.41.0

[PATCH 10/34] sched: Add normal_policy()

[Tejun Heo]

A new BPF extensible sched_class will need to dynamically change how a task
picks its sched_class. For example, if the loaded BPF scheduler progs fail,
the tasks will be forced back on CFS even if the task's policy is set to the
new sched_class. To support such mapping, add normal_policy() which wraps
testing for %SCHED_NORMAL. This doesn't cause any behavior changes.

v2: Update the description with more details on the expected use.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 kernel/sched/fair.c  | 2 +-
 kernel/sched/sched.h | 8 +++++++-
 2 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 9838c82230e1..5ba7e1f2e7c7 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -7900,7 +7900,7 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_
 	 * Batch and idle tasks do not preempt non-idle tasks (their preemption
 	 * is driven by the tick):
 	 */
-	if (unlikely(p->policy != SCHED_NORMAL) || !sched_feat(WAKEUP_PREEMPTION))
+	if (unlikely(!normal_policy(p->policy)) || !sched_feat(WAKEUP_PREEMPTION))
 		return;
 
 	find_matching_se(&se, &pse);
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index cd11bf9de7f9..7eee863942f4 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -182,9 +182,15 @@ static inline int idle_policy(int policy)
 {
 	return policy == SCHED_IDLE;
 }
+
+static inline int normal_policy(int policy)
+{
+	return policy == SCHED_NORMAL;
+}
+
 static inline int fair_policy(int policy)
 {
-	return policy == SCHED_NORMAL || policy == SCHED_BATCH;
+	return normal_policy(policy) || policy == SCHED_BATCH;
 }
 
 static inline int rt_policy(int policy)
-- 
2.41.0

[PATCH 11/34] sched_ext: Add boilerplate for extensible scheduler class

[Tejun Heo]

This adds dummy implementations of sched_ext interfaces which interact with
the scheduler core and hook them in the correct places. As they're all
dummies, this doesn't cause any behavior changes. This is split out to help
reviewing.

v2: balance_scx_on_up() dropped. This will be handled in sched_ext proper.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 include/linux/sched/ext.h | 12 ++++++++++++
 kernel/fork.c             |  2 ++
 kernel/sched/core.c       | 32 ++++++++++++++++++++++++--------
 kernel/sched/ext.h        | 24 ++++++++++++++++++++++++
 kernel/sched/idle.c       |  2 ++
 kernel/sched/sched.h      |  2 ++
 6 files changed, 66 insertions(+), 8 deletions(-)
 create mode 100644 include/linux/sched/ext.h
 create mode 100644 kernel/sched/ext.h

diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
new file mode 100644
index 000000000000..a05dfcf533b0
--- /dev/null
+++ b/include/linux/sched/ext.h
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_SCHED_EXT_H
+#define _LINUX_SCHED_EXT_H
+
+#ifdef CONFIG_SCHED_CLASS_EXT
+#error "NOT IMPLEMENTED YET"
+#else	/* !CONFIG_SCHED_CLASS_EXT */
+
+static inline void sched_ext_free(struct task_struct *p) {}
+
+#endif	/* CONFIG_SCHED_CLASS_EXT */
+#endif	/* _LINUX_SCHED_EXT_H */
diff --git a/kernel/fork.c b/kernel/fork.c
index 0a9c8c03889f..d285d1a91228 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -23,6 +23,7 @@
 #include <linux/sched/task.h>
 #include <linux/sched/task_stack.h>
 #include <linux/sched/cputime.h>
+#include <linux/sched/ext.h>
 #include <linux/seq_file.h>
 #include <linux/rtmutex.h>
 #include <linux/init.h>
@@ -975,6 +976,7 @@ void __put_task_struct(struct task_struct *tsk)
 	WARN_ON(refcount_read(&tsk->usage));
 	WARN_ON(tsk == current);
 
+	sched_ext_free(tsk);
 	io_uring_free(tsk);
 	cgroup_free(tsk);
 	task_numa_free(tsk, true);
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 297740c4a5bc..c8a2c99248b7 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4699,6 +4699,8 @@ late_initcall(sched_core_sysctl_init);
  */
 int sched_fork(unsigned long clone_flags, struct task_struct *p)
 {
+	int ret;
+
 	__sched_fork(clone_flags, p);
 	/*
 	 * We mark the process as NEW here. This guarantees that
@@ -4735,12 +4737,16 @@ int sched_fork(unsigned long clone_flags, struct task_struct *p)
 		p->sched_reset_on_fork = 0;
 	}
 
-	if (dl_prio(p->prio))
-		return -EAGAIN;
-	else if (rt_prio(p->prio))
+	scx_pre_fork(p);
+
+	if (dl_prio(p->prio)) {
+		ret = -EAGAIN;
+		goto out_cancel;
+	} else if (rt_prio(p->prio)) {
 		p->sched_class = &rt_sched_class;
-	else
+	} else {
 		p->sched_class = &fair_sched_class;
+	}
 
 	init_entity_runnable_average(&p->se);
 
@@ -4758,6 +4764,10 @@ int sched_fork(unsigned long clone_flags, struct task_struct *p)
 	RB_CLEAR_NODE(&p->pushable_dl_tasks);
 #endif
 	return 0;
+
+out_cancel:
+	scx_cancel_fork(p);
+	return ret;
 }
 
 int sched_cgroup_fork(struct task_struct *p, struct kernel_clone_args *kargs)
@@ -4788,16 +4798,18 @@ int sched_cgroup_fork(struct task_struct *p, struct kernel_clone_args *kargs)
 		p->sched_class->task_fork(p);
 	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
 
-	return 0;
+	return scx_fork(p);
 }
 
 void sched_cancel_fork(struct task_struct *p)
 {
+	scx_cancel_fork(p);
 }
 
 void sched_post_fork(struct task_struct *p)
 {
 	uclamp_post_fork(p);
+	scx_post_fork(p);
 }
 
 unsigned long to_ratio(u64 period, u64 runtime)
@@ -5959,7 +5971,7 @@ static void put_prev_task_balance(struct rq *rq, struct task_struct *prev,
 	 * We can terminate the balance pass as soon as we know there is
 	 * a runnable task of @class priority or higher.
 	 */
-	for_class_range(class, prev->sched_class, &idle_sched_class) {
+	for_balance_class_range(class, prev->sched_class, &idle_sched_class) {
 		if (class->balance(rq, prev, rf))
 			break;
 	}
@@ -5977,6 +5989,9 @@ __pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 	const struct sched_class *class;
 	struct task_struct *p;
 
+	if (scx_enabled())
+		goto restart;
+
 	/*
 	 * Optimization: we know that if all tasks are in the fair class we can
 	 * call that function directly, but only if the @prev task wasn't of a
@@ -6002,7 +6017,7 @@ __pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 restart:
 	put_prev_task_balance(rq, prev, rf);
 
-	for_each_class(class) {
+	for_each_active_class(class) {
 		p = class->pick_next_task(rq);
 		if (p)
 			return p;
@@ -6035,7 +6050,7 @@ static inline struct task_struct *pick_task(struct rq *rq)
 	const struct sched_class *class;
 	struct task_struct *p;
 
-	for_each_class(class) {
+	for_each_active_class(class) {
 		p = class->pick_task(rq);
 		if (p)
 			return p;
@@ -10072,6 +10087,7 @@ void __init sched_init(void)
 	balance_push_set(smp_processor_id(), false);
 #endif
 	init_sched_fair_class();
+	init_sched_ext_class();
 
 	psi_init();
 
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
new file mode 100644
index 000000000000..6a93c4825339
--- /dev/null
+++ b/kernel/sched/ext.h
@@ -0,0 +1,24 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifdef CONFIG_SCHED_CLASS_EXT
+#error "NOT IMPLEMENTED YET"
+#else	/* CONFIG_SCHED_CLASS_EXT */
+
+#define scx_enabled()		false
+
+static inline void scx_pre_fork(struct task_struct *p) {}
+static inline int scx_fork(struct task_struct *p) { return 0; }
+static inline void scx_post_fork(struct task_struct *p) {}
+static inline void scx_cancel_fork(struct task_struct *p) {}
+static inline void init_sched_ext_class(void) {}
+
+#define for_each_active_class		for_each_class
+#define for_balance_class_range		for_class_range
+
+#endif	/* CONFIG_SCHED_CLASS_EXT */
+
+#if defined(CONFIG_SCHED_CLASS_EXT) && defined(CONFIG_SMP)
+#error "NOT IMPLEMENTED YET"
+#else
+static inline void scx_update_idle(struct rq *rq, bool idle) {}
+#endif
diff --git a/kernel/sched/idle.c b/kernel/sched/idle.c
index 342f58a329f5..8767d97fcfa5 100644
--- a/kernel/sched/idle.c
+++ b/kernel/sched/idle.c
@@ -407,11 +407,13 @@ static void check_preempt_curr_idle(struct rq *rq, struct task_struct *p, int fl
 
 static void put_prev_task_idle(struct rq *rq, struct task_struct *prev)
 {
+	scx_update_idle(rq, false);
 }
 
 static void set_next_task_idle(struct rq *rq, struct task_struct *next, bool first)
 {
 	update_idle_core(rq);
+	scx_update_idle(rq, true);
 	schedstat_inc(rq->sched_goidle);
 }
 
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 7eee863942f4..17bd277cf27a 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -3551,4 +3551,6 @@ enum cpu_cftype_id {
 extern struct cftype cpu_cftypes[CPU_CFTYPE_CNT + 1];
 #endif /* CONFIG_CGROUP_SCHED */
 
+#include "ext.h"
+
 #endif /* _KERNEL_SCHED_SCHED_H */
-- 
2.41.0

[PATCH 12/34] sched_ext: Implement BPF extensible scheduler class

[Tejun Heo]

Implement a new scheduler class sched_ext (SCX), which allows scheduling
policies to be implemented as BPF programs to achieve the following:

1. Ease of experimentation and exploration: Enabling rapid iteration of new
   scheduling policies.

2. Customization: Building application-specific schedulers which implement
   policies that are not applicable to general-purpose schedulers.

3. Rapid scheduler deployments: Non-disruptive swap outs of scheduling
   policies in production environments.

sched_ext leverages BPF’s struct_ops feature to define a structure which
exports function callbacks and flags to BPF programs that wish to implement
scheduling policies. The struct_ops structure exported by sched_ext is
struct sched_ext_ops, and is conceptually similar to struct sched_class. The
role of sched_ext is to map the complex sched_class callbacks to the more
simple and ergonomic struct sched_ext_ops callbacks.

For more detailed discussion on the motivations and overview, please refer
to the cover letter.

Later patches will also add several example schedulers and documentation.

This patch implements the minimum core framework to enable implementation of
BPF schedulers. Subsequent patches will gradually add functionalities
including safety guarantee mechanisms, nohz and cgroup support.

include/linux/sched/ext.h defines struct sched_ext_ops. With the comment on
top, each operation should be self-explanatory. The followings are worth
noting:

* Both "sched_ext" and its shorthand "scx" are used. If the identifier
  already has "sched" in it, "ext" is used; otherwise, "scx".

* In sched_ext_ops, only .name is mandatory. Every operation is optional and
  if omitted a simple but functional default behavior is provided.

* A new policy constant SCHED_EXT is added and a task can select sched_ext
  by invoking sched_setscheduler(2) with the new policy constant. However,
  if the BPF scheduler is not loaded, SCHED_EXT is the same as SCHED_NORMAL
  and the task is scheduled by CFS. When the BPF scheduler is loaded, all
  tasks which have the SCHED_EXT policy are switched to sched_ext.

* To bridge the workflow imbalance between the scheduler core and
  sched_ext_ops callbacks, sched_ext uses simple FIFOs called dispatch
  queues (dsq's). By default, there is one global dsq (SCX_DSQ_GLOBAL), and
  one local per-CPU dsq (SCX_DSQ_LOCAL). SCX_DSQ_GLOBAL is provided for
  convenience and need not be used by a scheduler that doesn't require it.
  SCX_DSQ_LOCAL is the per-CPU FIFO that sched_ext pulls from when putting
  the next task on the CPU. The BPF scheduler can manage an arbitrary number
  of dsq's using scx_bpf_create_dsq() and scx_bpf_destroy_dsq().

* sched_ext guarantees system integrity no matter what the BPF scheduler
  does. To enable this, each task's ownership is tracked through
  p->scx.ops_state and all tasks are put on scx_tasks list. The disable path
  can always recover and revert all tasks back to CFS. See p->scx.ops_state
  and scx_tasks.

* A task is not tied to its rq while enqueued. This decouples CPU selection
  from queueing and allows sharing a scheduling queue across an arbitrary
  subset of CPUs. This adds some complexities as a task may need to be
  bounced between rq's right before it starts executing. See
  dispatch_to_local_dsq() and move_task_to_local_dsq().

* One complication that arises from the above weak association between task
  and rq is that synchronizing with dequeue() gets complicated as dequeue()
  may happen anytime while the task is enqueued and the dispatch path might
  need to release the rq lock to transfer the task. Solving this requires a
  bit of complexity. See the logic around p->scx.sticky_cpu and
  p->scx.ops_qseq.

* Both enable and disable paths are a bit complicated. The enable path
  switches all tasks without blocking to avoid issues which can arise from
  partially switched states (e.g. the switching task itself being starved).
  The disable path can't trust the BPF scheduler at all, so it also has to
  guarantee forward progress without blocking. See scx_ops_enable() and
  scx_ops_disable_workfn().

* When sched_ext is disabled, static_branches are used to shut down the
  entry points from hot paths.

v4: * SCHED_CHANGE_BLOCK replaced with the previous
      sched_deq_and_put_task()/sched_enq_and_set_tsak() pair. This is
      because upstream is adaopting a different generic cleanup mechanism.
      Once that lands, the code will be adapted accordingly.

    * task_on_scx() used to test whether a task should be switched into SCX,
      which is confusing. Renamed to task_should_scx(). task_on_scx() now
      tests whether a task is currently on SCX.

    * scx_has_idle_cpus is barely used anymore and replaced with direct
      check on the idle cpumask.

    * SCX_PICK_IDLE_CORE added and scx_pick_idle_cpu() improved to prefer
      fully idle cores.

    * ops.enable() now sees up-to-date p->scx.weight value.

    * ttwu_queue path is disabled for tasks on SCX to avoid confusing BPF
      schedulers expecting ->select_cpu() call.

    * Use cpu_smt_mask() instead of topology_sibling_cpumask() like the rest
      of the scheduler.

v3: * ops.set_weight() added to allow BPF schedulers to track weight changes
      without polling p->scx.weight.

    * move_task_to_local_dsq() was losing SCX-specific enq_flags when
      enqueueing the task on the target dsq because it goes through
      activate_task() which loses the upper 32bit of the flags. Carry the
      flags through rq->scx.extra_enq_flags.

    * scx_bpf_dispatch(), scx_bpf_pick_idle_cpu(), scx_bpf_task_running()
      and scx_bpf_task_cpu() now use the new KF_RCU instead of
      KF_TRUSTED_ARGS to make it easier for BPF schedulers to call them.

    * The kfunc helper access control mechanism implemented through
      sched_ext_entity.kf_mask is improved. Now SCX_CALL_OP*() is always
      used when invoking scx_ops operations.

v2: * balance_scx_on_up() is dropped. Instead, on UP, balance_scx() is
      called from put_prev_taks_scx() and pick_next_task_scx() as necessary.
      To determine whether balance_scx() should be called from
      put_prev_task_scx(), SCX_TASK_DEQD_FOR_SLEEP flag is added. See the
      comment in put_prev_task_scx() for details.

    * sched_deq_and_put_task() / sched_enq_and_set_task() sequences replaced
      with SCHED_CHANGE_BLOCK().

    * Unused all_dsqs list removed. This was a left-over from previous
      iterations.

    * p->scx.kf_mask is added to track and enforce which kfunc helpers are
      allowed. Also, init/exit sequences are updated to make some kfuncs
      always safe to call regardless of the current BPF scheduler state.
      Combined, this should make all the kfuncs safe.

    * BPF now supports sleepable struct_ops operations. Hacky workaround
      removed and operations and kfunc helpers are tagged appropriately.

    * BPF now supports bitmask / cpumask helpers. scx_bpf_get_idle_cpumask()
      and friends are added so that BPF schedulers can use the idle masks
      with the generic helpers. This replaces the hacky kfunc helpers added
      by a separate patch in V1.

    * CONFIG_SCHED_CLASS_EXT can no longer be enabled if SCHED_CORE is
      enabled. This restriction will be removed by a later patch which adds
      core-sched support.

    * Add MAINTAINERS entries and other misc changes.

Signed-off-by: Tejun Heo <tj@kernel.org>
Co-authored-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
Cc: Andrea Righi <andrea.righi@canonical.com>
---
 MAINTAINERS                       |    3 +
 include/asm-generic/vmlinux.lds.h |    1 +
 include/linux/sched.h             |    5 +
 include/linux/sched/ext.h         |  400 +++-
 include/uapi/linux/sched.h        |    1 +
 init/init_task.c                  |   10 +
 kernel/Kconfig.preempt            |   22 +-
 kernel/bpf/bpf_struct_ops_types.h |    4 +
 kernel/sched/build_policy.c       |    4 +
 kernel/sched/core.c               |   70 +
 kernel/sched/debug.c              |    6 +
 kernel/sched/ext.c                | 3140 +++++++++++++++++++++++++++++
 kernel/sched/ext.h                |  118 +-
 kernel/sched/sched.h              |   16 +
 14 files changed, 3796 insertions(+), 4 deletions(-)
 create mode 100644 kernel/sched/ext.c

diff --git a/MAINTAINERS b/MAINTAINERS
index c904dba1733b..5c301e22ff74 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -18767,6 +18767,8 @@ R:	Ben Segall <bsegall@google.com> (CONFIG_CFS_BANDWIDTH)
 R:	Mel Gorman <mgorman@suse.de> (CONFIG_NUMA_BALANCING)
 R:	Daniel Bristot de Oliveira <bristot@redhat.com> (SCHED_DEADLINE)
 R:	Valentin Schneider <vschneid@redhat.com> (TOPOLOGY)
+R:	Tejun Heo <tj@kernel.org> (SCHED_EXT)
+R:	David Vernet <void@manifault.com> (SCHED_EXT)
 L:	linux-kernel@vger.kernel.org
 S:	Maintained
 T:	git git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git sched/core
@@ -18775,6 +18777,7 @@ F:	include/linux/sched.h
 F:	include/linux/wait.h
 F:	include/uapi/linux/sched.h
 F:	kernel/sched/
+F:	tools/sched_ext/
 
 SCSI RDMA PROTOCOL (SRP) INITIATOR
 M:	Bart Van Assche <bvanassche@acm.org>
diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h
index d1f57e4868ed..cd5a718ba49f 100644
--- a/include/asm-generic/vmlinux.lds.h
+++ b/include/asm-generic/vmlinux.lds.h
@@ -131,6 +131,7 @@
 	*(__dl_sched_class)			\
 	*(__rt_sched_class)			\
 	*(__fair_sched_class)			\
+	*(__ext_sched_class)			\
 	*(__idle_sched_class)			\
 	__sched_class_lowest = .;
 
diff --git a/include/linux/sched.h b/include/linux/sched.h
index eed5d65b8d1f..00d4ce3af52a 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -72,6 +72,8 @@ struct task_delay_info;
 struct task_group;
 struct user_event_mm;
 
+#include <linux/sched/ext.h>
+
 /*
  * Task state bitmask. NOTE! These bits are also
  * encoded in fs/proc/array.c: get_task_state().
@@ -790,6 +792,9 @@ struct task_struct {
 	struct sched_entity		se;
 	struct sched_rt_entity		rt;
 	struct sched_dl_entity		dl;
+#ifdef CONFIG_SCHED_CLASS_EXT
+	struct sched_ext_entity		scx;
+#endif
 	const struct sched_class	*sched_class;
 
 #ifdef CONFIG_SCHED_CORE
diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index a05dfcf533b0..92011a63cc15 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -1,9 +1,407 @@
 /* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
 #ifndef _LINUX_SCHED_EXT_H
 #define _LINUX_SCHED_EXT_H
 
 #ifdef CONFIG_SCHED_CLASS_EXT
-#error "NOT IMPLEMENTED YET"
+
+#include <linux/rhashtable.h>
+#include <linux/llist.h>
+
+enum scx_consts {
+	SCX_OPS_NAME_LEN	= 128,
+	SCX_EXIT_REASON_LEN	= 128,
+	SCX_EXIT_BT_LEN		= 64,
+	SCX_EXIT_MSG_LEN	= 1024,
+
+	SCX_SLICE_DFL		= 20 * NSEC_PER_MSEC,
+};
+
+/*
+ * DSQ (dispatch queue) IDs are 64bit of the format:
+ *
+ *   Bits: [63] [62 ..  0]
+ *         [ B] [   ID   ]
+ *
+ *    B: 1 for IDs for built-in DSQs, 0 for ops-created user DSQs
+ *   ID: 63 bit ID
+ *
+ * Built-in IDs:
+ *
+ *   Bits: [63] [62] [61..32] [31 ..  0]
+ *         [ 1] [ L] [   R  ] [    V   ]
+ *
+ *    1: 1 for built-in DSQs.
+ *    L: 1 for LOCAL_ON DSQ IDs, 0 for others
+ *    V: For LOCAL_ON DSQ IDs, a CPU number. For others, a pre-defined value.
+ */
+enum scx_dsq_id_flags {
+	SCX_DSQ_FLAG_BUILTIN	= 1LLU << 63,
+	SCX_DSQ_FLAG_LOCAL_ON	= 1LLU << 62,
+
+	SCX_DSQ_INVALID		= SCX_DSQ_FLAG_BUILTIN | 0,
+	SCX_DSQ_GLOBAL		= SCX_DSQ_FLAG_BUILTIN | 1,
+	SCX_DSQ_LOCAL		= SCX_DSQ_FLAG_BUILTIN | 2,
+	SCX_DSQ_LOCAL_ON	= SCX_DSQ_FLAG_BUILTIN | SCX_DSQ_FLAG_LOCAL_ON,
+	SCX_DSQ_LOCAL_CPU_MASK	= 0xffffffffLLU,
+};
+
+enum scx_exit_type {
+	SCX_EXIT_NONE,
+	SCX_EXIT_DONE,
+
+	SCX_EXIT_UNREG = 64,	/* BPF unregistration */
+
+	SCX_EXIT_ERROR = 1024,	/* runtime error, error msg contains details */
+	SCX_EXIT_ERROR_BPF,	/* ERROR but triggered through scx_bpf_error() */
+};
+
+/*
+ * scx_exit_info is passed to ops.exit() to describe why the BPF scheduler is
+ * being disabled.
+ */
+struct scx_exit_info {
+	/* %SCX_EXIT_* - broad category of the exit reason */
+	enum scx_exit_type	type;
+	/* textual representation of the above */
+	char			reason[SCX_EXIT_REASON_LEN];
+	/* number of entries in the backtrace */
+	u32			bt_len;
+	/* backtrace if exiting due to an error */
+	unsigned long		bt[SCX_EXIT_BT_LEN];
+	/* extra message */
+	char			msg[SCX_EXIT_MSG_LEN];
+};
+
+/* sched_ext_ops.flags */
+enum scx_ops_flags {
+	/*
+	 * Keep built-in idle tracking even if ops.update_idle() is implemented.
+	 */
+	SCX_OPS_KEEP_BUILTIN_IDLE = 1LLU << 0,
+
+	/*
+	 * By default, if there are no other task to run on the CPU, ext core
+	 * keeps running the current task even after its slice expires. If this
+	 * flag is specified, such tasks are passed to ops.enqueue() with
+	 * %SCX_ENQ_LAST. See the comment above %SCX_ENQ_LAST for more info.
+	 */
+	SCX_OPS_ENQ_LAST	= 1LLU << 1,
+
+	/*
+	 * An exiting task may schedule after PF_EXITING is set. In such cases,
+	 * bpf_task_from_pid() may not be able to find the task and if the BPF
+	 * scheduler depends on pid lookup for dispatching, the task will be
+	 * lost leading to various issues including RCU grace period stalls.
+	 *
+	 * To mask this problem, by default, unhashed tasks are automatically
+	 * dispatched to the local DSQ on enqueue. If the BPF scheduler doesn't
+	 * depend on pid lookups and wants to handle these tasks directly, the
+	 * following flag can be used.
+	 */
+	SCX_OPS_ENQ_EXITING	= 1LLU << 2,
+
+	SCX_OPS_ALL_FLAGS	= SCX_OPS_KEEP_BUILTIN_IDLE |
+				  SCX_OPS_ENQ_LAST |
+				  SCX_OPS_ENQ_EXITING,
+};
+
+/* argument container for ops.enable() and friends */
+struct scx_enable_args {
+	/* empty for now */
+};
+
+/**
+ * struct sched_ext_ops - Operation table for BPF scheduler implementation
+ *
+ * Userland can implement an arbitrary scheduling policy by implementing and
+ * loading operations in this table.
+ */
+struct sched_ext_ops {
+	/**
+	 * select_cpu - Pick the target CPU for a task which is being woken up
+	 * @p: task being woken up
+	 * @prev_cpu: the cpu @p was on before sleeping
+	 * @wake_flags: SCX_WAKE_*
+	 *
+	 * Decision made here isn't final. @p may be moved to any CPU while it
+	 * is getting dispatched for execution later. However, as @p is not on
+	 * the rq at this point, getting the eventual execution CPU right here
+	 * saves a small bit of overhead down the line.
+	 *
+	 * If an idle CPU is returned, the CPU is kicked and will try to
+	 * dispatch. While an explicit custom mechanism can be added,
+	 * select_cpu() serves as the default way to wake up idle CPUs.
+	 */
+	s32 (*select_cpu)(struct task_struct *p, s32 prev_cpu, u64 wake_flags);
+
+	/**
+	 * enqueue - Enqueue a task on the BPF scheduler
+	 * @p: task being enqueued
+	 * @enq_flags: %SCX_ENQ_*
+	 *
+	 * @p is ready to run. Dispatch directly by calling scx_bpf_dispatch()
+	 * or enqueue on the BPF scheduler. If not directly dispatched, the bpf
+	 * scheduler owns @p and if it fails to dispatch @p, the task will
+	 * stall.
+	 */
+	void (*enqueue)(struct task_struct *p, u64 enq_flags);
+
+	/**
+	 * dequeue - Remove a task from the BPF scheduler
+	 * @p: task being dequeued
+	 * @deq_flags: %SCX_DEQ_*
+	 *
+	 * Remove @p from the BPF scheduler. This is usually called to isolate
+	 * the task while updating its scheduling properties (e.g. priority).
+	 *
+	 * The ext core keeps track of whether the BPF side owns a given task or
+	 * not and can gracefully ignore spurious dispatches from BPF side,
+	 * which makes it safe to not implement this method. However, depending
+	 * on the scheduling logic, this can lead to confusing behaviors - e.g.
+	 * scheduling position not being updated across a priority change.
+	 */
+	void (*dequeue)(struct task_struct *p, u64 deq_flags);
+
+	/**
+	 * dispatch - Dispatch tasks from the BPF scheduler and/or consume DSQs
+	 * @cpu: CPU to dispatch tasks for
+	 * @prev: previous task being switched out
+	 *
+	 * Called when a CPU's local dsq is empty. The operation should dispatch
+	 * one or more tasks from the BPF scheduler into the DSQs using
+	 * scx_bpf_dispatch() and/or consume user DSQs into the local DSQ using
+	 * scx_bpf_consume().
+	 *
+	 * The maximum number of times scx_bpf_dispatch() can be called without
+	 * an intervening scx_bpf_consume() is specified by
+	 * ops.dispatch_max_batch. See the comments on top of the two functions
+	 * for more details.
+	 *
+	 * When not %NULL, @prev is an SCX task with its slice depleted. If
+	 * @prev is still runnable as indicated by set %SCX_TASK_QUEUED in
+	 * @prev->scx.flags, it is not enqueued yet and will be enqueued after
+	 * ops.dispatch() returns. To keep executing @prev, return without
+	 * dispatching or consuming any tasks. Also see %SCX_OPS_ENQ_LAST.
+	 */
+	void (*dispatch)(s32 cpu, struct task_struct *prev);
+
+	/**
+	 * yield - Yield CPU
+	 * @from: yielding task
+	 * @to: optional yield target task
+	 *
+	 * If @to is NULL, @from is yielding the CPU to other runnable tasks.
+	 * The BPF scheduler should ensure that other available tasks are
+	 * dispatched before the yielding task. Return value is ignored in this
+	 * case.
+	 *
+	 * If @to is not-NULL, @from wants to yield the CPU to @to. If the bpf
+	 * scheduler can implement the request, return %true; otherwise, %false.
+	 */
+	bool (*yield)(struct task_struct *from, struct task_struct *to);
+
+	/**
+	 * set_weight - Set task weight
+	 * @p: task to set weight for
+	 * @weight: new eight [1..10000]
+	 *
+	 * Update @p's weight to @weight.
+	 */
+	void (*set_weight)(struct task_struct *p, u32 weight);
+
+	/**
+	 * set_cpumask - Set CPU affinity
+	 * @p: task to set CPU affinity for
+	 * @cpumask: cpumask of cpus that @p can run on
+	 *
+	 * Update @p's CPU affinity to @cpumask.
+	 */
+	void (*set_cpumask)(struct task_struct *p, struct cpumask *cpumask);
+
+	/**
+	 * update_idle - Update the idle state of a CPU
+	 * @cpu: CPU to udpate the idle state for
+	 * @idle: whether entering or exiting the idle state
+	 *
+	 * This operation is called when @rq's CPU goes or leaves the idle
+	 * state. By default, implementing this operation disables the built-in
+	 * idle CPU tracking and the following helpers become unavailable:
+	 *
+	 * - scx_bpf_select_cpu_dfl()
+	 * - scx_bpf_test_and_clear_cpu_idle()
+	 * - scx_bpf_pick_idle_cpu()
+	 *
+	 * The user also must implement ops.select_cpu() as the default
+	 * implementation relies on scx_bpf_select_cpu_dfl().
+	 *
+	 * Specify the %SCX_OPS_KEEP_BUILTIN_IDLE flag to keep the built-in idle
+	 * tracking.
+	 */
+	void (*update_idle)(s32 cpu, bool idle);
+
+	/**
+	 * prep_enable - Prepare to enable BPF scheduling for a task
+	 * @p: task to prepare BPF scheduling for
+	 * @args: enable arguments, see the struct definition
+	 *
+	 * Either we're loading a BPF scheduler or a new task is being forked.
+	 * Prepare BPF scheduling for @p. This operation may block and can be
+	 * used for allocations.
+	 *
+	 * Return 0 for success, -errno for failure. An error return while
+	 * loading will abort loading of the BPF scheduler. During a fork, will
+	 * abort the specific fork.
+	 */
+	s32 (*prep_enable)(struct task_struct *p, struct scx_enable_args *args);
+
+	/**
+	 * enable - Enable BPF scheduling for a task
+	 * @p: task to enable BPF scheduling for
+	 * @args: enable arguments, see the struct definition
+	 *
+	 * Enable @p for BPF scheduling. @p will start running soon.
+	 */
+	void (*enable)(struct task_struct *p, struct scx_enable_args *args);
+
+	/**
+	 * cancel_enable - Cancel prep_enable()
+	 * @p: task being canceled
+	 * @args: enable arguments, see the struct definition
+	 *
+	 * @p was prep_enable()'d but failed before reaching enable(). Undo the
+	 * preparation.
+	 */
+	void (*cancel_enable)(struct task_struct *p,
+			      struct scx_enable_args *args);
+
+	/**
+	 * disable - Disable BPF scheduling for a task
+	 * @p: task to disable BPF scheduling for
+	 *
+	 * @p is exiting, leaving SCX or the BPF scheduler is being unloaded.
+	 * Disable BPF scheduling for @p.
+	 */
+	void (*disable)(struct task_struct *p);
+
+	/*
+	 * All online ops must come before ops.init().
+	 */
+
+	/**
+	 * init - Initialize the BPF scheduler
+	 */
+	s32 (*init)(void);
+
+	/**
+	 * exit - Clean up after the BPF scheduler
+	 * @info: Exit info
+	 */
+	void (*exit)(struct scx_exit_info *info);
+
+	/**
+	 * dispatch_max_batch - Max nr of tasks that dispatch() can dispatch
+	 */
+	u32 dispatch_max_batch;
+
+	/**
+	 * flags - %SCX_OPS_* flags
+	 */
+	u64 flags;
+
+	/**
+	 * name - BPF scheduler's name
+	 *
+	 * Must be a non-zero valid BPF object name including only isalnum(),
+	 * '_' and '.' chars. Shows up in kernel.sched_ext_ops sysctl while the
+	 * BPF scheduler is enabled.
+	 */
+	char name[SCX_OPS_NAME_LEN];
+};
+
+/*
+ * Dispatch queue (dsq) is a simple FIFO which is used to buffer between the
+ * scheduler core and the BPF scheduler. See the documentation for more details.
+ */
+struct scx_dispatch_q {
+	raw_spinlock_t		lock;
+	struct list_head	fifo;	/* processed in dispatching order */
+	u32			nr;
+	u64			id;
+	struct rhash_head	hash_node;
+	struct llist_node	free_node;
+	struct rcu_head		rcu;
+};
+
+/* scx_entity.flags */
+enum scx_ent_flags {
+	SCX_TASK_QUEUED		= 1 << 0, /* on ext runqueue */
+	SCX_TASK_BAL_KEEP	= 1 << 1, /* balance decided to keep current */
+	SCX_TASK_ENQ_LOCAL	= 1 << 2, /* used by scx_select_cpu_dfl() to set SCX_ENQ_LOCAL */
+
+	SCX_TASK_OPS_PREPPED	= 1 << 8, /* prepared for BPF scheduler enable */
+	SCX_TASK_OPS_ENABLED	= 1 << 9, /* task has BPF scheduler enabled */
+
+	SCX_TASK_DEQD_FOR_SLEEP	= 1 << 17, /* last dequeue was for SLEEP */
+
+	SCX_TASK_CURSOR		= 1 << 31, /* iteration cursor, not a task */
+};
+
+/*
+ * Mask bits for scx_entity.kf_mask. Not all kfuncs can be called from
+ * everywhere and the following bits track which kfunc sets are currently
+ * allowed for %current. This simple per-task tracking works because SCX ops
+ * nest in a limited way. BPF will likely implement a way to allow and disallow
+ * kfuncs depending on the calling context which will replace this manual
+ * mechanism. See scx_kf_allow().
+ */
+enum scx_kf_mask {
+	SCX_KF_UNLOCKED		= 0,	  /* not sleepable, not rq locked */
+	/* all non-sleepables may be nested inside INIT and SLEEPABLE */
+	SCX_KF_INIT		= 1 << 0, /* running ops.init() */
+	SCX_KF_SLEEPABLE	= 1 << 1, /* other sleepable init operations */
+	/* ops.dequeue (in REST) may be nested inside DISPATCH */
+	SCX_KF_DISPATCH		= 1 << 3, /* ops.dispatch() */
+	SCX_KF_ENQUEUE		= 1 << 4, /* ops.enqueue() */
+	SCX_KF_REST		= 1 << 5, /* other rq-locked operations */
+
+	__SCX_KF_RQ_LOCKED	= SCX_KF_DISPATCH | SCX_KF_ENQUEUE | SCX_KF_REST,
+};
+
+/*
+ * The following is embedded in task_struct and contains all fields necessary
+ * for a task to be scheduled by SCX.
+ */
+struct sched_ext_entity {
+	struct scx_dispatch_q	*dsq;
+	struct list_head	dsq_node;
+	u32			flags;		/* protected by rq lock */
+	u32			weight;
+	s32			sticky_cpu;
+	s32			holding_cpu;
+	u32			kf_mask;	/* see scx_kf_mask above */
+	atomic64_t		ops_state;
+
+	/* BPF scheduler modifiable fields */
+
+	/*
+	 * Runtime budget in nsecs. This is usually set through
+	 * scx_bpf_dispatch() but can also be modified directly by the BPF
+	 * scheduler. Automatically decreased by SCX as the task executes. On
+	 * depletion, a scheduling event is triggered.
+	 */
+	u64			slice;
+
+	/* cold fields */
+	struct list_head	tasks_node;
+};
+
+void sched_ext_free(struct task_struct *p);
+
 #else	/* !CONFIG_SCHED_CLASS_EXT */
 
 static inline void sched_ext_free(struct task_struct *p) {}
diff --git a/include/uapi/linux/sched.h b/include/uapi/linux/sched.h
index 3bac0a8ceab2..359a14cc76a4 100644
--- a/include/uapi/linux/sched.h
+++ b/include/uapi/linux/sched.h
@@ -118,6 +118,7 @@ struct clone_args {
 /* SCHED_ISO: reserved but not implemented yet */
 #define SCHED_IDLE		5
 #define SCHED_DEADLINE		6
+#define SCHED_EXT		7
 
 /* Can be ORed in to make sure the process is reverted back to SCHED_NORMAL on fork */
 #define SCHED_RESET_ON_FORK     0x40000000
diff --git a/init/init_task.c b/init/init_task.c
index ff6c4b9bfe6b..bdbc663107bf 100644
--- a/init/init_task.c
+++ b/init/init_task.c
@@ -6,6 +6,7 @@
 #include <linux/sched/sysctl.h>
 #include <linux/sched/rt.h>
 #include <linux/sched/task.h>
+#include <linux/sched/ext.h>
 #include <linux/init.h>
 #include <linux/fs.h>
 #include <linux/mm.h>
@@ -101,6 +102,15 @@ struct task_struct init_task
 #endif
 #ifdef CONFIG_CGROUP_SCHED
 	.sched_task_group = &root_task_group,
+#endif
+#ifdef CONFIG_SCHED_CLASS_EXT
+	.scx		= {
+		.dsq_node	= LIST_HEAD_INIT(init_task.scx.dsq_node),
+		.sticky_cpu	= -1,
+		.holding_cpu	= -1,
+		.ops_state	= ATOMIC_INIT(0),
+		.slice		= SCX_SLICE_DFL,
+	},
 #endif
 	.ptraced	= LIST_HEAD_INIT(init_task.ptraced),
 	.ptrace_entry	= LIST_HEAD_INIT(init_task.ptrace_entry),
diff --git a/kernel/Kconfig.preempt b/kernel/Kconfig.preempt
index c2f1fd95a821..0afcda19bc50 100644
--- a/kernel/Kconfig.preempt
+++ b/kernel/Kconfig.preempt
@@ -133,4 +133,24 @@ config SCHED_CORE
 	  which is the likely usage by Linux distributions, there should
 	  be no measurable impact on performance.
 
-
+config SCHED_CLASS_EXT
+	bool "Extensible Scheduling Class"
+	depends on BPF_SYSCALL && BPF_JIT && !SCHED_CORE
+	help
+	  This option enables a new scheduler class sched_ext (SCX), which
+	  allows scheduling policies to be implemented as BPF programs to
+	  achieve the following:
+
+	  - Ease of experimentation and exploration: Enabling rapid
+	    iteration of new scheduling policies.
+	  - Customization: Building application-specific schedulers which
+	    implement policies that are not applicable to general-purpose
+	    schedulers.
+	  - Rapid scheduler deployments: Non-disruptive swap outs of
+	    scheduling policies in production environments.
+
+	  sched_ext leverages BPF’s struct_ops feature to define a structure
+	  which exports function callbacks and flags to BPF programs that
+	  wish to implement scheduling policies. The struct_ops structure
+	  exported by sched_ext is struct sched_ext_ops, and is conceptually
+	  similar to struct sched_class.
diff --git a/kernel/bpf/bpf_struct_ops_types.h b/kernel/bpf/bpf_struct_ops_types.h
index 5678a9ddf817..3618769d853d 100644
--- a/kernel/bpf/bpf_struct_ops_types.h
+++ b/kernel/bpf/bpf_struct_ops_types.h
@@ -9,4 +9,8 @@ BPF_STRUCT_OPS_TYPE(bpf_dummy_ops)
 #include <net/tcp.h>
 BPF_STRUCT_OPS_TYPE(tcp_congestion_ops)
 #endif
+#ifdef CONFIG_SCHED_CLASS_EXT
+#include <linux/sched/ext.h>
+BPF_STRUCT_OPS_TYPE(sched_ext_ops)
+#endif
 #endif
diff --git a/kernel/sched/build_policy.c b/kernel/sched/build_policy.c
index d9dc9ab3773f..4c658b21f603 100644
--- a/kernel/sched/build_policy.c
+++ b/kernel/sched/build_policy.c
@@ -28,6 +28,7 @@
 #include <linux/suspend.h>
 #include <linux/tsacct_kern.h>
 #include <linux/vtime.h>
+#include <linux/percpu-rwsem.h>
 
 #include <uapi/linux/sched/types.h>
 
@@ -52,3 +53,6 @@
 #include "cputime.c"
 #include "deadline.c"
 
+#ifdef CONFIG_SCHED_CLASS_EXT
+# include "ext.c"
+#endif
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index c8a2c99248b7..a182a9adec3e 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -3925,6 +3925,15 @@ bool cpus_share_cache(int this_cpu, int that_cpu)
 
 static inline bool ttwu_queue_cond(struct task_struct *p, int cpu)
 {
+	/*
+	 * The BPF scheduler may depend on select_task_rq() being invoked during
+	 * wakeups. In addition, @p may end up executing on a different CPU
+	 * regardless of what happens in the wakeup path making the ttwu_queue
+	 * optimization less meaningful. Skip if on SCX.
+	 */
+	if (task_on_scx(p))
+		return false;
+
 	/*
 	 * Do not complicate things with the async wake_list while the CPU is
 	 * in hotplug state.
@@ -4496,6 +4505,18 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
 	p->rt.on_rq		= 0;
 	p->rt.on_list		= 0;
 
+#ifdef CONFIG_SCHED_CLASS_EXT
+	p->scx.dsq		= NULL;
+	INIT_LIST_HEAD(&p->scx.dsq_node);
+	p->scx.flags		= 0;
+	p->scx.weight		= 0;
+	p->scx.sticky_cpu	= -1;
+	p->scx.holding_cpu	= -1;
+	p->scx.kf_mask		= 0;
+	atomic64_set(&p->scx.ops_state, 0);
+	p->scx.slice		= SCX_SLICE_DFL;
+#endif
+
 #ifdef CONFIG_PREEMPT_NOTIFIERS
 	INIT_HLIST_HEAD(&p->preempt_notifiers);
 #endif
@@ -4744,6 +4765,10 @@ int sched_fork(unsigned long clone_flags, struct task_struct *p)
 		goto out_cancel;
 	} else if (rt_prio(p->prio)) {
 		p->sched_class = &rt_sched_class;
+#ifdef CONFIG_SCHED_CLASS_EXT
+	} else if (task_should_scx(p)) {
+		p->sched_class = &ext_sched_class;
+#endif
 	} else {
 		p->sched_class = &fair_sched_class;
 	}
@@ -7032,6 +7057,10 @@ void __setscheduler_prio(struct task_struct *p, int prio)
 		p->sched_class = &dl_sched_class;
 	else if (rt_prio(prio))
 		p->sched_class = &rt_sched_class;
+#ifdef CONFIG_SCHED_CLASS_EXT
+	else if (task_should_scx(p))
+		p->sched_class = &ext_sched_class;
+#endif
 	else
 		p->sched_class = &fair_sched_class;
 
@@ -9036,6 +9065,7 @@ SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
 	case SCHED_NORMAL:
 	case SCHED_BATCH:
 	case SCHED_IDLE:
+	case SCHED_EXT:
 		ret = 0;
 		break;
 	}
@@ -9063,6 +9093,7 @@ SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
 	case SCHED_NORMAL:
 	case SCHED_BATCH:
 	case SCHED_IDLE:
+	case SCHED_EXT:
 		ret = 0;
 	}
 	return ret;
@@ -9918,6 +9949,10 @@ void __init sched_init(void)
 	BUG_ON(!sched_class_above(&dl_sched_class, &rt_sched_class));
 	BUG_ON(!sched_class_above(&rt_sched_class, &fair_sched_class));
 	BUG_ON(!sched_class_above(&fair_sched_class, &idle_sched_class));
+#ifdef CONFIG_SCHED_CLASS_EXT
+	BUG_ON(!sched_class_above(&fair_sched_class, &ext_sched_class));
+	BUG_ON(!sched_class_above(&ext_sched_class, &idle_sched_class));
+#endif
 
 	wait_bit_init();
 
@@ -12046,3 +12081,38 @@ void sched_mm_cid_fork(struct task_struct *t)
 	t->mm_cid_active = 1;
 }
 #endif
+
+#ifdef CONFIG_SCHED_CLASS_EXT
+void sched_deq_and_put_task(struct task_struct *p, int queue_flags,
+			    struct sched_enq_and_set_ctx *ctx)
+{
+	struct rq *rq = task_rq(p);
+
+	lockdep_assert_rq_held(rq);
+
+	*ctx = (struct sched_enq_and_set_ctx){
+		.p = p,
+		.queue_flags = queue_flags,
+		.queued = task_on_rq_queued(p),
+		.running = task_current(rq, p),
+	};
+
+	update_rq_clock(rq);
+	if (ctx->queued)
+		dequeue_task(rq, p, queue_flags | DEQUEUE_NOCLOCK);
+	if (ctx->running)
+		put_prev_task(rq, p);
+}
+
+void sched_enq_and_set_task(struct sched_enq_and_set_ctx *ctx)
+{
+	struct rq *rq = task_rq(ctx->p);
+
+	lockdep_assert_rq_held(rq);
+
+	if (ctx->queued)
+		enqueue_task(rq, ctx->p, ctx->queue_flags | ENQUEUE_NOCLOCK);
+	if (ctx->running)
+		set_next_task(rq, ctx->p);
+}
+#endif	/* CONFIG_SCHED_CLASS_EXT */
diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index 0b2340a79b65..79fac9c92a22 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -377,6 +377,9 @@ static __init int sched_init_debug(void)
 
 	debugfs_create_file("debug", 0444, debugfs_sched, NULL, &sched_debug_fops);
 
+#ifdef CONFIG_SCHED_CLASS_EXT
+	debugfs_create_file("ext", 0444, debugfs_sched, NULL, &sched_ext_fops);
+#endif
 	return 0;
 }
 late_initcall(sched_init_debug);
@@ -1093,6 +1096,9 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
 		P(dl.runtime);
 		P(dl.deadline);
 	}
+#ifdef CONFIG_SCHED_CLASS_EXT
+	__PS("ext.enabled", task_on_scx(p));
+#endif
 #undef PN_SCHEDSTAT
 #undef P_SCHEDSTAT
 
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
new file mode 100644
index 000000000000..51d77459d208
--- /dev/null
+++ b/kernel/sched/ext.c
@@ -0,0 +1,3140 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+#define SCX_OP_IDX(op)		(offsetof(struct sched_ext_ops, op) / sizeof(void (*)(void)))
+
+enum scx_internal_consts {
+	SCX_NR_ONLINE_OPS	= SCX_OP_IDX(init),
+	SCX_DSP_DFL_MAX_BATCH	= 32,
+};
+
+enum scx_ops_enable_state {
+	SCX_OPS_PREPPING,
+	SCX_OPS_ENABLING,
+	SCX_OPS_ENABLED,
+	SCX_OPS_DISABLING,
+	SCX_OPS_DISABLED,
+};
+
+/*
+ * sched_ext_entity->ops_state
+ *
+ * Used to track the task ownership between the SCX core and the BPF scheduler.
+ * State transitions look as follows:
+ *
+ * NONE -> QUEUEING -> QUEUED -> DISPATCHING
+ *   ^              |                 |
+ *   |              v                 v
+ *   \-------------------------------/
+ *
+ * QUEUEING and DISPATCHING states can be waited upon. See wait_ops_state() call
+ * sites for explanations on the conditions being waited upon and why they are
+ * safe. Transitions out of them into NONE or QUEUED must store_release and the
+ * waiters should load_acquire.
+ *
+ * Tracking scx_ops_state enables sched_ext core to reliably determine whether
+ * any given task can be dispatched by the BPF scheduler at all times and thus
+ * relaxes the requirements on the BPF scheduler. This allows the BPF scheduler
+ * to try to dispatch any task anytime regardless of its state as the SCX core
+ * can safely reject invalid dispatches.
+ */
+enum scx_ops_state {
+	SCX_OPSS_NONE,		/* owned by the SCX core */
+	SCX_OPSS_QUEUEING,	/* in transit to the BPF scheduler */
+	SCX_OPSS_QUEUED,	/* owned by the BPF scheduler */
+	SCX_OPSS_DISPATCHING,	/* in transit back to the SCX core */
+
+	/*
+	 * QSEQ brands each QUEUED instance so that, when dispatch races
+	 * dequeue/requeue, the dispatcher can tell whether it still has a claim
+	 * on the task being dispatched.
+	 */
+	SCX_OPSS_QSEQ_SHIFT	= 2,
+	SCX_OPSS_STATE_MASK	= (1LLU << SCX_OPSS_QSEQ_SHIFT) - 1,
+	SCX_OPSS_QSEQ_MASK	= ~SCX_OPSS_STATE_MASK,
+};
+
+/*
+ * During exit, a task may schedule after losing its PIDs. When disabling the
+ * BPF scheduler, we need to be able to iterate tasks in every state to
+ * guarantee system safety. Maintain a dedicated task list which contains every
+ * task between its fork and eventual free.
+ */
+static DEFINE_SPINLOCK(scx_tasks_lock);
+static LIST_HEAD(scx_tasks);
+
+/* ops enable/disable */
+static struct kthread_worker *scx_ops_helper;
+static DEFINE_MUTEX(scx_ops_enable_mutex);
+DEFINE_STATIC_KEY_FALSE(__scx_ops_enabled);
+DEFINE_STATIC_PERCPU_RWSEM(scx_fork_rwsem);
+static atomic_t scx_ops_enable_state_var = ATOMIC_INIT(SCX_OPS_DISABLED);
+static struct sched_ext_ops scx_ops;
+static bool scx_warned_zero_slice;
+
+static DEFINE_STATIC_KEY_FALSE(scx_ops_enq_last);
+static DEFINE_STATIC_KEY_FALSE(scx_ops_enq_exiting);
+static DEFINE_STATIC_KEY_FALSE(scx_builtin_idle_enabled);
+
+struct static_key_false scx_has_op[SCX_NR_ONLINE_OPS] =
+	{ [0 ... SCX_NR_ONLINE_OPS-1] = STATIC_KEY_FALSE_INIT };
+
+static atomic_t scx_exit_type = ATOMIC_INIT(SCX_EXIT_DONE);
+static struct scx_exit_info scx_exit_info;
+
+static atomic64_t scx_nr_rejected = ATOMIC64_INIT(0);
+
+/* idle tracking */
+#ifdef CONFIG_SMP
+#ifdef CONFIG_CPUMASK_OFFSTACK
+#define CL_ALIGNED_IF_ONSTACK
+#else
+#define CL_ALIGNED_IF_ONSTACK __cacheline_aligned_in_smp
+#endif
+
+static struct {
+	cpumask_var_t cpu;
+	cpumask_var_t smt;
+} idle_masks CL_ALIGNED_IF_ONSTACK;
+
+#endif	/* CONFIG_SMP */
+
+/*
+ * Direct dispatch marker.
+ *
+ * Non-NULL values are used for direct dispatch from enqueue path. A valid
+ * pointer points to the task currently being enqueued. An ERR_PTR value is used
+ * to indicate that direct dispatch has already happened.
+ */
+static DEFINE_PER_CPU(struct task_struct *, direct_dispatch_task);
+
+/* dispatch queues */
+static struct scx_dispatch_q __cacheline_aligned_in_smp scx_dsq_global;
+
+static const struct rhashtable_params dsq_hash_params = {
+	.key_len		= 8,
+	.key_offset		= offsetof(struct scx_dispatch_q, id),
+	.head_offset		= offsetof(struct scx_dispatch_q, hash_node),
+};
+
+static struct rhashtable dsq_hash;
+static LLIST_HEAD(dsqs_to_free);
+
+/* dispatch buf */
+struct scx_dsp_buf_ent {
+	struct task_struct	*task;
+	u64			qseq;
+	u64			dsq_id;
+	u64			enq_flags;
+};
+
+static u32 scx_dsp_max_batch;
+static struct scx_dsp_buf_ent __percpu *scx_dsp_buf;
+
+struct scx_dsp_ctx {
+	struct rq		*rq;
+	struct rq_flags		*rf;
+	u32			buf_cursor;
+	u32			nr_tasks;
+};
+
+static DEFINE_PER_CPU(struct scx_dsp_ctx, scx_dsp_ctx);
+
+void scx_bpf_dispatch(struct task_struct *p, u64 dsq_id, u64 slice,
+		      u64 enq_flags);
+__printf(2, 3) static void scx_ops_error_type(enum scx_exit_type type,
+					      const char *fmt, ...);
+#define scx_ops_error(fmt, args...)						\
+	scx_ops_error_type(SCX_EXIT_ERROR, fmt, ##args)
+
+struct scx_task_iter {
+	struct sched_ext_entity		cursor;
+	struct task_struct		*locked;
+	struct rq			*rq;
+	struct rq_flags			rf;
+};
+
+#define SCX_HAS_OP(op)	static_branch_likely(&scx_has_op[SCX_OP_IDX(op)])
+
+/* if the highest set bit is N, return a mask with bits [N+1, 31] set */
+static u32 higher_bits(u32 flags)
+{
+	return ~((1 << fls(flags)) - 1);
+}
+
+/* return the mask with only the highest bit set */
+static u32 highest_bit(u32 flags)
+{
+	int bit = fls(flags);
+	return bit ? 1 << (bit - 1) : 0;
+}
+
+/*
+ * scx_kf_mask enforcement. Some kfuncs can only be called from specific SCX
+ * ops. When invoking SCX ops, SCX_CALL_OP[_RET]() should be used to indicate
+ * the allowed kfuncs and those kfuncs should use scx_kf_allowed() to check
+ * whether it's running from an allowed context.
+ *
+ * @mask is constant, always inline to cull the mask calculations.
+ */
+static __always_inline void scx_kf_allow(u32 mask)
+{
+	/* nesting is allowed only in increasing scx_kf_mask order */
+	WARN_ONCE((mask | higher_bits(mask)) & current->scx.kf_mask,
+		  "invalid nesting current->scx.kf_mask=0x%x mask=0x%x\n",
+		  current->scx.kf_mask, mask);
+	current->scx.kf_mask |= mask;
+}
+
+static void scx_kf_disallow(u32 mask)
+{
+	current->scx.kf_mask &= ~mask;
+}
+
+#define SCX_CALL_OP(mask, op, args...)						\
+do {										\
+	if (mask) {								\
+		scx_kf_allow(mask);						\
+		scx_ops.op(args);						\
+		scx_kf_disallow(mask);						\
+	} else {								\
+		scx_ops.op(args);						\
+	}									\
+} while (0)
+
+#define SCX_CALL_OP_RET(mask, op, args...)					\
+({										\
+	__typeof__(scx_ops.op(args)) __ret;					\
+	if (mask) {								\
+		scx_kf_allow(mask);						\
+		__ret = scx_ops.op(args);					\
+		scx_kf_disallow(mask);						\
+	} else {								\
+		__ret = scx_ops.op(args);					\
+	}									\
+	__ret;									\
+})
+
+/* @mask is constant, always inline to cull unnecessary branches */
+static __always_inline bool scx_kf_allowed(u32 mask)
+{
+	if (unlikely(!(current->scx.kf_mask & mask))) {
+		scx_ops_error("kfunc with mask 0x%x called from an operation only allowing 0x%x",
+			      mask, current->scx.kf_mask);
+		return false;
+	}
+
+	if (unlikely((mask & (SCX_KF_INIT | SCX_KF_SLEEPABLE)) &&
+		     in_interrupt())) {
+		scx_ops_error("sleepable kfunc called from non-sleepable context");
+		return false;
+	}
+
+	/*
+	 * Enforce nesting boundaries. e.g. A kfunc which can be called from
+	 * DISPATCH must not be called if we're running DEQUEUE which is nested
+	 * inside ops.dispatch(). We don't need to check the SCX_KF_SLEEPABLE
+	 * boundary thanks to the above in_interrupt() check.
+	 */
+	if (unlikely(highest_bit(mask) == SCX_KF_DISPATCH &&
+		     (current->scx.kf_mask & higher_bits(SCX_KF_DISPATCH)))) {
+		scx_ops_error("dispatch kfunc called from a nested operation");
+		return false;
+	}
+
+	return true;
+}
+
+/**
+ * scx_task_iter_init - Initialize a task iterator
+ * @iter: iterator to init
+ *
+ * Initialize @iter. Must be called with scx_tasks_lock held. Once initialized,
+ * @iter must eventually be exited with scx_task_iter_exit().
+ *
+ * scx_tasks_lock may be released between this and the first next() call or
+ * between any two next() calls. If scx_tasks_lock is released between two
+ * next() calls, the caller is responsible for ensuring that the task being
+ * iterated remains accessible either through RCU read lock or obtaining a
+ * reference count.
+ *
+ * All tasks which existed when the iteration started are guaranteed to be
+ * visited as long as they still exist.
+ */
+static void scx_task_iter_init(struct scx_task_iter *iter)
+{
+	lockdep_assert_held(&scx_tasks_lock);
+
+	iter->cursor = (struct sched_ext_entity){ .flags = SCX_TASK_CURSOR };
+	list_add(&iter->cursor.tasks_node, &scx_tasks);
+	iter->locked = NULL;
+}
+
+/**
+ * scx_task_iter_exit - Exit a task iterator
+ * @iter: iterator to exit
+ *
+ * Exit a previously initialized @iter. Must be called with scx_tasks_lock held.
+ * If the iterator holds a task's rq lock, that rq lock is released. See
+ * scx_task_iter_init() for details.
+ */
+static void scx_task_iter_exit(struct scx_task_iter *iter)
+{
+	struct list_head *cursor = &iter->cursor.tasks_node;
+
+	lockdep_assert_held(&scx_tasks_lock);
+
+	if (iter->locked) {
+		task_rq_unlock(iter->rq, iter->locked, &iter->rf);
+		iter->locked = NULL;
+	}
+
+	if (list_empty(cursor))
+		return;
+
+	list_del_init(cursor);
+}
+
+/**
+ * scx_task_iter_next - Next task
+ * @iter: iterator to walk
+ *
+ * Visit the next task. See scx_task_iter_init() for details.
+ */
+static struct task_struct *scx_task_iter_next(struct scx_task_iter *iter)
+{
+	struct list_head *cursor = &iter->cursor.tasks_node;
+	struct sched_ext_entity *pos;
+
+	lockdep_assert_held(&scx_tasks_lock);
+
+	list_for_each_entry(pos, cursor, tasks_node) {
+		if (&pos->tasks_node == &scx_tasks)
+			return NULL;
+		if (!(pos->flags & SCX_TASK_CURSOR)) {
+			list_move(cursor, &pos->tasks_node);
+			return container_of(pos, struct task_struct, scx);
+		}
+	}
+
+	/* can't happen, should always terminate at scx_tasks above */
+	BUG();
+}
+
+/**
+ * scx_task_iter_next_filtered - Next non-idle task
+ * @iter: iterator to walk
+ *
+ * Visit the next non-idle task. See scx_task_iter_init() for details.
+ */
+static struct task_struct *
+scx_task_iter_next_filtered(struct scx_task_iter *iter)
+{
+	struct task_struct *p;
+
+	while ((p = scx_task_iter_next(iter))) {
+		if (!is_idle_task(p))
+			return p;
+	}
+	return NULL;
+}
+
+/**
+ * scx_task_iter_next_filtered_locked - Next non-idle task with its rq locked
+ * @iter: iterator to walk
+ *
+ * Visit the next non-idle task with its rq lock held. See scx_task_iter_init()
+ * for details.
+ */
+static struct task_struct *
+scx_task_iter_next_filtered_locked(struct scx_task_iter *iter)
+{
+	struct task_struct *p;
+
+	if (iter->locked) {
+		task_rq_unlock(iter->rq, iter->locked, &iter->rf);
+		iter->locked = NULL;
+	}
+
+	p = scx_task_iter_next_filtered(iter);
+	if (!p)
+		return NULL;
+
+	iter->rq = task_rq_lock(p, &iter->rf);
+	iter->locked = p;
+	return p;
+}
+
+static enum scx_ops_enable_state scx_ops_enable_state(void)
+{
+	return atomic_read(&scx_ops_enable_state_var);
+}
+
+static enum scx_ops_enable_state
+scx_ops_set_enable_state(enum scx_ops_enable_state to)
+{
+	return atomic_xchg(&scx_ops_enable_state_var, to);
+}
+
+static bool scx_ops_tryset_enable_state(enum scx_ops_enable_state to,
+					enum scx_ops_enable_state from)
+{
+	int from_v = from;
+
+	return atomic_try_cmpxchg(&scx_ops_enable_state_var, &from_v, to);
+}
+
+static bool scx_ops_disabling(void)
+{
+	return unlikely(scx_ops_enable_state() == SCX_OPS_DISABLING);
+}
+
+/**
+ * wait_ops_state - Busy-wait the specified ops state to end
+ * @p: target task
+ * @opss: state to wait the end of
+ *
+ * Busy-wait for @p to transition out of @opss. This can only be used when the
+ * state part of @opss is %SCX_QUEUEING or %SCX_DISPATCHING. This function also
+ * has load_acquire semantics to ensure that the caller can see the updates made
+ * in the enqueueing and dispatching paths.
+ */
+static void wait_ops_state(struct task_struct *p, u64 opss)
+{
+	do {
+		cpu_relax();
+	} while (atomic64_read_acquire(&p->scx.ops_state) == opss);
+}
+
+/**
+ * ops_cpu_valid - Verify a cpu number
+ * @cpu: cpu number which came from a BPF ops
+ *
+ * @cpu is a cpu number which came from the BPF scheduler and can be any value.
+ * Verify that it is in range and one of the possible cpus.
+ */
+static bool ops_cpu_valid(s32 cpu)
+{
+	return likely(cpu >= 0 && cpu < nr_cpu_ids && cpu_possible(cpu));
+}
+
+/**
+ * ops_sanitize_err - Sanitize a -errno value
+ * @ops_name: operation to blame on failure
+ * @err: -errno value to sanitize
+ *
+ * Verify @err is a valid -errno. If not, trigger scx_ops_error() and return
+ * -%EPROTO. This is necessary because returning a rogue -errno up the chain can
+ * cause misbehaviors. For an example, a large negative return from
+ * ops.prep_enable() triggers an oops when passed up the call chain because the
+ * value fails IS_ERR() test after being encoded with ERR_PTR() and then is
+ * handled as a pointer.
+ */
+static int ops_sanitize_err(const char *ops_name, s32 err)
+{
+	if (err < 0 && err >= -MAX_ERRNO)
+		return err;
+
+	scx_ops_error("ops.%s() returned an invalid errno %d", ops_name, err);
+	return -EPROTO;
+}
+
+static void update_curr_scx(struct rq *rq)
+{
+	struct task_struct *curr = rq->curr;
+	u64 now = rq_clock_task(rq);
+	u64 delta_exec;
+
+	if (time_before_eq64(now, curr->se.exec_start))
+		return;
+
+	delta_exec = now - curr->se.exec_start;
+	curr->se.exec_start = now;
+	curr->se.sum_exec_runtime += delta_exec;
+	account_group_exec_runtime(curr, delta_exec);
+	cgroup_account_cputime(curr, delta_exec);
+
+	curr->scx.slice -= min(curr->scx.slice, delta_exec);
+}
+
+static void dispatch_enqueue(struct scx_dispatch_q *dsq, struct task_struct *p,
+			     u64 enq_flags)
+{
+	bool is_local = dsq->id == SCX_DSQ_LOCAL;
+
+	WARN_ON_ONCE(p->scx.dsq || !list_empty(&p->scx.dsq_node));
+
+	if (!is_local) {
+		raw_spin_lock(&dsq->lock);
+		if (unlikely(dsq->id == SCX_DSQ_INVALID)) {
+			scx_ops_error("attempting to dispatch to a destroyed dsq");
+			/* fall back to the global dsq */
+			raw_spin_unlock(&dsq->lock);
+			dsq = &scx_dsq_global;
+			raw_spin_lock(&dsq->lock);
+		}
+	}
+
+	if (enq_flags & SCX_ENQ_HEAD)
+		list_add(&p->scx.dsq_node, &dsq->fifo);
+	else
+		list_add_tail(&p->scx.dsq_node, &dsq->fifo);
+	dsq->nr++;
+	p->scx.dsq = dsq;
+
+	/*
+	 * We're transitioning out of QUEUEING or DISPATCHING. store_release to
+	 * match waiters' load_acquire.
+	 */
+	if (enq_flags & SCX_ENQ_CLEAR_OPSS)
+		atomic64_set_release(&p->scx.ops_state, SCX_OPSS_NONE);
+
+	if (is_local) {
+		struct rq *rq = container_of(dsq, struct rq, scx.local_dsq);
+
+		if (sched_class_above(&ext_sched_class, rq->curr->sched_class))
+			resched_curr(rq);
+	} else {
+		raw_spin_unlock(&dsq->lock);
+	}
+}
+
+static void dispatch_dequeue(struct scx_rq *scx_rq, struct task_struct *p)
+{
+	struct scx_dispatch_q *dsq = p->scx.dsq;
+	bool is_local = dsq == &scx_rq->local_dsq;
+
+	if (!dsq) {
+		WARN_ON_ONCE(!list_empty(&p->scx.dsq_node));
+		/*
+		 * When dispatching directly from the BPF scheduler to a local
+		 * DSQ, the task isn't associated with any DSQ but
+		 * @p->scx.holding_cpu may be set under the protection of
+		 * %SCX_OPSS_DISPATCHING.
+		 */
+		if (p->scx.holding_cpu >= 0)
+			p->scx.holding_cpu = -1;
+		return;
+	}
+
+	if (!is_local)
+		raw_spin_lock(&dsq->lock);
+
+	/*
+	 * Now that we hold @dsq->lock, @p->holding_cpu and @p->scx.dsq_node
+	 * can't change underneath us.
+	*/
+	if (p->scx.holding_cpu < 0) {
+		/* @p must still be on @dsq, dequeue */
+		WARN_ON_ONCE(list_empty(&p->scx.dsq_node));
+		list_del_init(&p->scx.dsq_node);
+		dsq->nr--;
+	} else {
+		/*
+		 * We're racing against dispatch_to_local_dsq() which already
+		 * removed @p from @dsq and set @p->scx.holding_cpu. Clear the
+		 * holding_cpu which tells dispatch_to_local_dsq() that it lost
+		 * the race.
+		 */
+		WARN_ON_ONCE(!list_empty(&p->scx.dsq_node));
+		p->scx.holding_cpu = -1;
+	}
+	p->scx.dsq = NULL;
+
+	if (!is_local)
+		raw_spin_unlock(&dsq->lock);
+}
+
+static struct scx_dispatch_q *find_non_local_dsq(u64 dsq_id)
+{
+	lockdep_assert(rcu_read_lock_any_held());
+
+	if (dsq_id == SCX_DSQ_GLOBAL)
+		return &scx_dsq_global;
+	else
+		return rhashtable_lookup_fast(&dsq_hash, &dsq_id,
+					      dsq_hash_params);
+}
+
+static struct scx_dispatch_q *find_dsq_for_dispatch(struct rq *rq, u64 dsq_id,
+						    struct task_struct *p)
+{
+	struct scx_dispatch_q *dsq;
+
+	if (dsq_id == SCX_DSQ_LOCAL)
+		return &rq->scx.local_dsq;
+
+	dsq = find_non_local_dsq(dsq_id);
+	if (unlikely(!dsq)) {
+		scx_ops_error("non-existent DSQ 0x%llx for %s[%d]",
+			      dsq_id, p->comm, p->pid);
+		return &scx_dsq_global;
+	}
+
+	return dsq;
+}
+
+static void direct_dispatch(struct task_struct *ddsp_task, struct task_struct *p,
+			    u64 dsq_id, u64 enq_flags)
+{
+	struct scx_dispatch_q *dsq;
+
+	/* @p must match the task which is being enqueued */
+	if (unlikely(p != ddsp_task)) {
+		if (IS_ERR(ddsp_task))
+			scx_ops_error("%s[%d] already direct-dispatched",
+				      p->comm, p->pid);
+		else
+			scx_ops_error("enqueueing %s[%d] but trying to direct-dispatch %s[%d]",
+				      ddsp_task->comm, ddsp_task->pid,
+				      p->comm, p->pid);
+		return;
+	}
+
+	/*
+	 * %SCX_DSQ_LOCAL_ON is not supported during direct dispatch because
+	 * dispatching to the local DSQ of a different CPU requires unlocking
+	 * the current rq which isn't allowed in the enqueue path. Use
+	 * ops.select_cpu() to be on the target CPU and then %SCX_DSQ_LOCAL.
+	 */
+	if (unlikely((dsq_id & SCX_DSQ_LOCAL_ON) == SCX_DSQ_LOCAL_ON)) {
+		scx_ops_error("SCX_DSQ_LOCAL_ON can't be used for direct-dispatch");
+		return;
+	}
+
+	dsq = find_dsq_for_dispatch(task_rq(p), dsq_id, p);
+	dispatch_enqueue(dsq, p, enq_flags | SCX_ENQ_CLEAR_OPSS);
+
+	/*
+	 * Mark that dispatch already happened by spoiling direct_dispatch_task
+	 * with a non-NULL value which can never match a valid task pointer.
+	 */
+	__this_cpu_write(direct_dispatch_task, ERR_PTR(-ESRCH));
+}
+
+static bool test_rq_online(struct rq *rq)
+{
+#ifdef CONFIG_SMP
+	return rq->online;
+#else
+	return true;
+#endif
+}
+
+static void do_enqueue_task(struct rq *rq, struct task_struct *p, u64 enq_flags,
+			    int sticky_cpu)
+{
+	struct task_struct **ddsp_taskp;
+	u64 qseq;
+
+	WARN_ON_ONCE(!(p->scx.flags & SCX_TASK_QUEUED));
+
+	if (p->scx.flags & SCX_TASK_ENQ_LOCAL) {
+		enq_flags |= SCX_ENQ_LOCAL;
+		p->scx.flags &= ~SCX_TASK_ENQ_LOCAL;
+	}
+
+	/* rq migration */
+	if (sticky_cpu == cpu_of(rq))
+		goto local_norefill;
+
+	/*
+	 * If !rq->online, we already told the BPF scheduler that the CPU is
+	 * offline. We're just trying to on/offline the CPU. Don't bother the
+	 * BPF scheduler.
+	 */
+	if (unlikely(!test_rq_online(rq)))
+		goto local;
+
+	/* see %SCX_OPS_ENQ_EXITING */
+	if (!static_branch_unlikely(&scx_ops_enq_exiting) &&
+	    unlikely(p->flags & PF_EXITING))
+		goto local;
+
+	/* see %SCX_OPS_ENQ_LAST */
+	if (!static_branch_unlikely(&scx_ops_enq_last) &&
+	    (enq_flags & SCX_ENQ_LAST))
+		goto local;
+
+	if (!SCX_HAS_OP(enqueue)) {
+		if (enq_flags & SCX_ENQ_LOCAL)
+			goto local;
+		else
+			goto global;
+	}
+
+	/* DSQ bypass didn't trigger, enqueue on the BPF scheduler */
+	qseq = rq->scx.ops_qseq++ << SCX_OPSS_QSEQ_SHIFT;
+
+	WARN_ON_ONCE(atomic64_read(&p->scx.ops_state) != SCX_OPSS_NONE);
+	atomic64_set(&p->scx.ops_state, SCX_OPSS_QUEUEING | qseq);
+
+	ddsp_taskp = this_cpu_ptr(&direct_dispatch_task);
+	WARN_ON_ONCE(*ddsp_taskp);
+	*ddsp_taskp = p;
+
+	SCX_CALL_OP(SCX_KF_ENQUEUE, enqueue, p, enq_flags);
+
+	/*
+	 * If not directly dispatched, QUEUEING isn't clear yet and dispatch or
+	 * dequeue may be waiting. The store_release matches their load_acquire.
+	 */
+	if (*ddsp_taskp == p)
+		atomic64_set_release(&p->scx.ops_state, SCX_OPSS_QUEUED | qseq);
+	*ddsp_taskp = NULL;
+	return;
+
+local:
+	p->scx.slice = SCX_SLICE_DFL;
+local_norefill:
+	dispatch_enqueue(&rq->scx.local_dsq, p, enq_flags);
+	return;
+
+global:
+	p->scx.slice = SCX_SLICE_DFL;
+	dispatch_enqueue(&scx_dsq_global, p, enq_flags);
+}
+
+static void enqueue_task_scx(struct rq *rq, struct task_struct *p, int enq_flags)
+{
+	int sticky_cpu = p->scx.sticky_cpu;
+
+	enq_flags |= rq->scx.extra_enq_flags;
+
+	if (sticky_cpu >= 0)
+		p->scx.sticky_cpu = -1;
+
+	/*
+	 * Restoring a running task will be immediately followed by
+	 * set_next_task_scx() which expects the task to not be on the BPF
+	 * scheduler as tasks can only start running through local DSQs. Force
+	 * direct-dispatch into the local DSQ by setting the sticky_cpu.
+	 */
+	if (unlikely(enq_flags & ENQUEUE_RESTORE) && task_current(rq, p))
+		sticky_cpu = cpu_of(rq);
+
+	if (p->scx.flags & SCX_TASK_QUEUED)
+		return;
+
+	p->scx.flags |= SCX_TASK_QUEUED;
+	rq->scx.nr_running++;
+	add_nr_running(rq, 1);
+
+	do_enqueue_task(rq, p, enq_flags, sticky_cpu);
+}
+
+static void ops_dequeue(struct task_struct *p, u64 deq_flags)
+{
+	u64 opss;
+
+	/* acquire ensures that we see the preceding updates on QUEUED */
+	opss = atomic64_read_acquire(&p->scx.ops_state);
+
+	switch (opss & SCX_OPSS_STATE_MASK) {
+	case SCX_OPSS_NONE:
+		break;
+	case SCX_OPSS_QUEUEING:
+		/*
+		 * QUEUEING is started and finished while holding @p's rq lock.
+		 * As we're holding the rq lock now, we shouldn't see QUEUEING.
+		 */
+		BUG();
+	case SCX_OPSS_QUEUED:
+		if (SCX_HAS_OP(dequeue))
+			SCX_CALL_OP(SCX_KF_REST, dequeue, p, deq_flags);
+
+		if (atomic64_try_cmpxchg(&p->scx.ops_state, &opss,
+					 SCX_OPSS_NONE))
+			break;
+		fallthrough;
+	case SCX_OPSS_DISPATCHING:
+		/*
+		 * If @p is being dispatched from the BPF scheduler to a DSQ,
+		 * wait for the transfer to complete so that @p doesn't get
+		 * added to its DSQ after dequeueing is complete.
+		 *
+		 * As we're waiting on DISPATCHING with the rq locked, the
+		 * dispatching side shouldn't try to lock the rq while
+		 * DISPATCHING is set. See dispatch_to_local_dsq().
+		 *
+		 * DISPATCHING shouldn't have qseq set and control can reach
+		 * here with NONE @opss from the above QUEUED case block.
+		 * Explicitly wait on %SCX_OPSS_DISPATCHING instead of @opss.
+		 */
+		wait_ops_state(p, SCX_OPSS_DISPATCHING);
+		BUG_ON(atomic64_read(&p->scx.ops_state) != SCX_OPSS_NONE);
+		break;
+	}
+}
+
+static void dequeue_task_scx(struct rq *rq, struct task_struct *p, int deq_flags)
+{
+	struct scx_rq *scx_rq = &rq->scx;
+
+	if (!(p->scx.flags & SCX_TASK_QUEUED))
+		return;
+
+	ops_dequeue(p, deq_flags);
+
+	if (deq_flags & SCX_DEQ_SLEEP)
+		p->scx.flags |= SCX_TASK_DEQD_FOR_SLEEP;
+	else
+		p->scx.flags &= ~SCX_TASK_DEQD_FOR_SLEEP;
+
+	p->scx.flags &= ~SCX_TASK_QUEUED;
+	scx_rq->nr_running--;
+	sub_nr_running(rq, 1);
+
+	dispatch_dequeue(scx_rq, p);
+}
+
+static void yield_task_scx(struct rq *rq)
+{
+	struct task_struct *p = rq->curr;
+
+	if (SCX_HAS_OP(yield))
+		SCX_CALL_OP_RET(SCX_KF_REST, yield, p, NULL);
+	else
+		p->scx.slice = 0;
+}
+
+static bool yield_to_task_scx(struct rq *rq, struct task_struct *to)
+{
+	struct task_struct *from = rq->curr;
+
+	if (SCX_HAS_OP(yield))
+		return SCX_CALL_OP_RET(SCX_KF_REST, yield, from, to);
+	else
+		return false;
+}
+
+#ifdef CONFIG_SMP
+/**
+ * move_task_to_local_dsq - Move a task from a different rq to a local DSQ
+ * @rq: rq to move the task into, currently locked
+ * @p: task to move
+ * @enq_flags: %SCX_ENQ_*
+ *
+ * Move @p which is currently on a different rq to @rq's local DSQ. The caller
+ * must:
+ *
+ * 1. Start with exclusive access to @p either through its DSQ lock or
+ *    %SCX_OPSS_DISPATCHING flag.
+ *
+ * 2. Set @p->scx.holding_cpu to raw_smp_processor_id().
+ *
+ * 3. Remember task_rq(@p). Release the exclusive access so that we don't
+ *    deadlock with dequeue.
+ *
+ * 4. Lock @rq and the task_rq from #3.
+ *
+ * 5. Call this function.
+ *
+ * Returns %true if @p was successfully moved. %false after racing dequeue and
+ * losing.
+ */
+static bool move_task_to_local_dsq(struct rq *rq, struct task_struct *p,
+				   u64 enq_flags)
+{
+	struct rq *task_rq;
+
+	lockdep_assert_rq_held(rq);
+
+	/*
+	 * If dequeue got to @p while we were trying to lock both rq's, it'd
+	 * have cleared @p->scx.holding_cpu to -1. While other cpus may have
+	 * updated it to different values afterwards, as this operation can't be
+	 * preempted or recurse, @p->scx.holding_cpu can never become
+	 * raw_smp_processor_id() again before we're done. Thus, we can tell
+	 * whether we lost to dequeue by testing whether @p->scx.holding_cpu is
+	 * still raw_smp_processor_id().
+	 *
+	 * See dispatch_dequeue() for the counterpart.
+	 */
+	if (unlikely(p->scx.holding_cpu != raw_smp_processor_id()))
+		return false;
+
+	/* @p->rq couldn't have changed if we're still the holding cpu */
+	task_rq = task_rq(p);
+	lockdep_assert_rq_held(task_rq);
+
+	WARN_ON_ONCE(!cpumask_test_cpu(cpu_of(rq), p->cpus_ptr));
+	deactivate_task(task_rq, p, 0);
+	set_task_cpu(p, cpu_of(rq));
+	p->scx.sticky_cpu = cpu_of(rq);
+
+	/*
+	 * We want to pass scx-specific enq_flags but activate_task() will
+	 * truncate the upper 32 bit. As we own @rq, we can pass them through
+	 * @rq->scx.extra_enq_flags instead.
+	 */
+	WARN_ON_ONCE(rq->scx.extra_enq_flags);
+	rq->scx.extra_enq_flags = enq_flags;
+	activate_task(rq, p, 0);
+	rq->scx.extra_enq_flags = 0;
+
+	return true;
+}
+
+/**
+ * dispatch_to_local_dsq_lock - Ensure source and desitnation rq's are locked
+ * @rq: current rq which is locked
+ * @rf: rq_flags to use when unlocking @rq
+ * @src_rq: rq to move task from
+ * @dst_rq: rq to move task to
+ *
+ * We're holding @rq lock and trying to dispatch a task from @src_rq to
+ * @dst_rq's local DSQ and thus need to lock both @src_rq and @dst_rq. Whether
+ * @rq stays locked isn't important as long as the state is restored after
+ * dispatch_to_local_dsq_unlock().
+ */
+static void dispatch_to_local_dsq_lock(struct rq *rq, struct rq_flags *rf,
+				       struct rq *src_rq, struct rq *dst_rq)
+{
+	rq_unpin_lock(rq, rf);
+
+	if (src_rq == dst_rq) {
+		raw_spin_rq_unlock(rq);
+		raw_spin_rq_lock(dst_rq);
+	} else if (rq == src_rq) {
+		double_lock_balance(rq, dst_rq);
+		rq_repin_lock(rq, rf);
+	} else if (rq == dst_rq) {
+		double_lock_balance(rq, src_rq);
+		rq_repin_lock(rq, rf);
+	} else {
+		raw_spin_rq_unlock(rq);
+		double_rq_lock(src_rq, dst_rq);
+	}
+}
+
+/**
+ * dispatch_to_local_dsq_unlock - Undo dispatch_to_local_dsq_lock()
+ * @rq: current rq which is locked
+ * @rf: rq_flags to use when unlocking @rq
+ * @src_rq: rq to move task from
+ * @dst_rq: rq to move task to
+ *
+ * Unlock @src_rq and @dst_rq and ensure that @rq is locked on return.
+ */
+static void dispatch_to_local_dsq_unlock(struct rq *rq, struct rq_flags *rf,
+					 struct rq *src_rq, struct rq *dst_rq)
+{
+	if (src_rq == dst_rq) {
+		raw_spin_rq_unlock(dst_rq);
+		raw_spin_rq_lock(rq);
+		rq_repin_lock(rq, rf);
+	} else if (rq == src_rq) {
+		double_unlock_balance(rq, dst_rq);
+	} else if (rq == dst_rq) {
+		double_unlock_balance(rq, src_rq);
+	} else {
+		double_rq_unlock(src_rq, dst_rq);
+		raw_spin_rq_lock(rq);
+		rq_repin_lock(rq, rf);
+	}
+}
+#endif	/* CONFIG_SMP */
+
+
+static bool consume_dispatch_q(struct rq *rq, struct rq_flags *rf,
+			       struct scx_dispatch_q *dsq)
+{
+	struct scx_rq *scx_rq = &rq->scx;
+	struct task_struct *p;
+	struct rq *task_rq;
+	bool moved = false;
+retry:
+	if (list_empty(&dsq->fifo))
+		return false;
+
+	raw_spin_lock(&dsq->lock);
+	list_for_each_entry(p, &dsq->fifo, scx.dsq_node) {
+		task_rq = task_rq(p);
+		if (rq == task_rq)
+			goto this_rq;
+		if (likely(test_rq_online(rq)) && !is_migration_disabled(p) &&
+		    cpumask_test_cpu(cpu_of(rq), p->cpus_ptr))
+			goto remote_rq;
+	}
+	raw_spin_unlock(&dsq->lock);
+	return false;
+
+this_rq:
+	/* @dsq is locked and @p is on this rq */
+	WARN_ON_ONCE(p->scx.holding_cpu >= 0);
+	list_move_tail(&p->scx.dsq_node, &scx_rq->local_dsq.fifo);
+	dsq->nr--;
+	scx_rq->local_dsq.nr++;
+	p->scx.dsq = &scx_rq->local_dsq;
+	raw_spin_unlock(&dsq->lock);
+	return true;
+
+remote_rq:
+#ifdef CONFIG_SMP
+	/*
+	 * @dsq is locked and @p is on a remote rq. @p is currently protected by
+	 * @dsq->lock. We want to pull @p to @rq but may deadlock if we grab
+	 * @task_rq while holding @dsq and @rq locks. As dequeue can't drop the
+	 * rq lock or fail, do a little dancing from our side. See
+	 * move_task_to_local_dsq().
+	 */
+	WARN_ON_ONCE(p->scx.holding_cpu >= 0);
+	list_del_init(&p->scx.dsq_node);
+	dsq->nr--;
+	p->scx.holding_cpu = raw_smp_processor_id();
+	raw_spin_unlock(&dsq->lock);
+
+	rq_unpin_lock(rq, rf);
+	double_lock_balance(rq, task_rq);
+	rq_repin_lock(rq, rf);
+
+	moved = move_task_to_local_dsq(rq, p, 0);
+
+	double_unlock_balance(rq, task_rq);
+#endif /* CONFIG_SMP */
+	if (likely(moved))
+		return true;
+	goto retry;
+}
+
+enum dispatch_to_local_dsq_ret {
+	DTL_DISPATCHED,		/* successfully dispatched */
+	DTL_LOST,		/* lost race to dequeue */
+	DTL_NOT_LOCAL,		/* destination is not a local DSQ */
+	DTL_INVALID,		/* invalid local dsq_id */
+};
+
+/**
+ * dispatch_to_local_dsq - Dispatch a task to a local dsq
+ * @rq: current rq which is locked
+ * @rf: rq_flags to use when unlocking @rq
+ * @dsq_id: destination dsq ID
+ * @p: task to dispatch
+ * @enq_flags: %SCX_ENQ_*
+ *
+ * We're holding @rq lock and want to dispatch @p to the local DSQ identified by
+ * @dsq_id. This function performs all the synchronization dancing needed
+ * because local DSQs are protected with rq locks.
+ *
+ * The caller must have exclusive ownership of @p (e.g. through
+ * %SCX_OPSS_DISPATCHING).
+ */
+static enum dispatch_to_local_dsq_ret
+dispatch_to_local_dsq(struct rq *rq, struct rq_flags *rf, u64 dsq_id,
+		      struct task_struct *p, u64 enq_flags)
+{
+	struct rq *src_rq = task_rq(p);
+	struct rq *dst_rq;
+
+	/*
+	 * We're synchronized against dequeue through DISPATCHING. As @p can't
+	 * be dequeued, its task_rq and cpus_allowed are stable too.
+	 */
+	if (dsq_id == SCX_DSQ_LOCAL) {
+		dst_rq = rq;
+	} else if ((dsq_id & SCX_DSQ_LOCAL_ON) == SCX_DSQ_LOCAL_ON) {
+		s32 cpu = dsq_id & SCX_DSQ_LOCAL_CPU_MASK;
+
+		if (!ops_cpu_valid(cpu)) {
+			scx_ops_error("invalid cpu %d in SCX_DSQ_LOCAL_ON verdict for %s[%d]",
+				      cpu, p->comm, p->pid);
+			return DTL_INVALID;
+		}
+		dst_rq = cpu_rq(cpu);
+	} else {
+		return DTL_NOT_LOCAL;
+	}
+
+	/* if dispatching to @rq that @p is already on, no lock dancing needed */
+	if (rq == src_rq && rq == dst_rq) {
+		dispatch_enqueue(&dst_rq->scx.local_dsq, p,
+				 enq_flags | SCX_ENQ_CLEAR_OPSS);
+		return DTL_DISPATCHED;
+	}
+
+#ifdef CONFIG_SMP
+	if (cpumask_test_cpu(cpu_of(dst_rq), p->cpus_ptr)) {
+		struct rq *locked_dst_rq = dst_rq;
+		bool dsp;
+
+		/*
+		 * @p is on a possibly remote @src_rq which we need to lock to
+		 * move the task. If dequeue is in progress, it'd be locking
+		 * @src_rq and waiting on DISPATCHING, so we can't grab @src_rq
+		 * lock while holding DISPATCHING.
+		 *
+		 * As DISPATCHING guarantees that @p is wholly ours, we can
+		 * pretend that we're moving from a DSQ and use the same
+		 * mechanism - mark the task under transfer with holding_cpu,
+		 * release DISPATCHING and then follow the same protocol.
+		 */
+		p->scx.holding_cpu = raw_smp_processor_id();
+
+		/* store_release ensures that dequeue sees the above */
+		atomic64_set_release(&p->scx.ops_state, SCX_OPSS_NONE);
+
+		dispatch_to_local_dsq_lock(rq, rf, src_rq, locked_dst_rq);
+
+		/*
+		 * We don't require the BPF scheduler to avoid dispatching to
+		 * offline CPUs mostly for convenience but also because CPUs can
+		 * go offline between scx_bpf_dispatch() calls and here. If @p
+		 * is destined to an offline CPU, queue it on its current CPU
+		 * instead, which should always be safe. As this is an allowed
+		 * behavior, don't trigger an ops error.
+		 */
+		if (unlikely(!test_rq_online(dst_rq)))
+			dst_rq = src_rq;
+
+		if (src_rq == dst_rq) {
+			/*
+			 * As @p is staying on the same rq, there's no need to
+			 * go through the full deactivate/activate cycle.
+			 * Optimize by abbreviating the operations in
+			 * move_task_to_local_dsq().
+			 */
+			dsp = p->scx.holding_cpu == raw_smp_processor_id();
+			if (likely(dsp)) {
+				p->scx.holding_cpu = -1;
+				dispatch_enqueue(&dst_rq->scx.local_dsq, p,
+						 enq_flags);
+			}
+		} else {
+			dsp = move_task_to_local_dsq(dst_rq, p, enq_flags);
+		}
+
+		/* if the destination CPU is idle, wake it up */
+		if (dsp && p->sched_class > dst_rq->curr->sched_class)
+			resched_curr(dst_rq);
+
+		dispatch_to_local_dsq_unlock(rq, rf, src_rq, locked_dst_rq);
+
+		return dsp ? DTL_DISPATCHED : DTL_LOST;
+	}
+#endif /* CONFIG_SMP */
+
+	scx_ops_error("SCX_DSQ_LOCAL[_ON] verdict target cpu %d not allowed for %s[%d]",
+		      cpu_of(dst_rq), p->comm, p->pid);
+	return DTL_INVALID;
+}
+
+/**
+ * finish_dispatch - Asynchronously finish dispatching a task
+ * @rq: current rq which is locked
+ * @rf: rq_flags to use when unlocking @rq
+ * @p: task to finish dispatching
+ * @qseq_at_dispatch: qseq when @p started getting dispatched
+ * @dsq_id: destination DSQ ID
+ * @enq_flags: %SCX_ENQ_*
+ *
+ * Dispatching to local DSQs may need to wait for queueing to complete or
+ * require rq lock dancing. As we don't wanna do either while inside
+ * ops.dispatch() to avoid locking order inversion, we split dispatching into
+ * two parts. scx_bpf_dispatch() which is called by ops.dispatch() records the
+ * task and its qseq. Once ops.dispatch() returns, this function is called to
+ * finish up.
+ *
+ * There is no guarantee that @p is still valid for dispatching or even that it
+ * was valid in the first place. Make sure that the task is still owned by the
+ * BPF scheduler and claim the ownership before dispatching.
+ */
+static void finish_dispatch(struct rq *rq, struct rq_flags *rf,
+			    struct task_struct *p, u64 qseq_at_dispatch,
+			    u64 dsq_id, u64 enq_flags)
+{
+	struct scx_dispatch_q *dsq;
+	u64 opss;
+
+retry:
+	/*
+	 * No need for _acquire here. @p is accessed only after a successful
+	 * try_cmpxchg to DISPATCHING.
+	 */
+	opss = atomic64_read(&p->scx.ops_state);
+
+	switch (opss & SCX_OPSS_STATE_MASK) {
+	case SCX_OPSS_DISPATCHING:
+	case SCX_OPSS_NONE:
+		/* someone else already got to it */
+		return;
+	case SCX_OPSS_QUEUED:
+		/*
+		 * If qseq doesn't match, @p has gone through at least one
+		 * dispatch/dequeue and re-enqueue cycle between
+		 * scx_bpf_dispatch() and here and we have no claim on it.
+		 */
+		if ((opss & SCX_OPSS_QSEQ_MASK) != qseq_at_dispatch)
+			return;
+
+		/*
+		 * While we know @p is accessible, we don't yet have a claim on
+		 * it - the BPF scheduler is allowed to dispatch tasks
+		 * spuriously and there can be a racing dequeue attempt. Let's
+		 * claim @p by atomically transitioning it from QUEUED to
+		 * DISPATCHING.
+		 */
+		if (likely(atomic64_try_cmpxchg(&p->scx.ops_state, &opss,
+						SCX_OPSS_DISPATCHING)))
+			break;
+		goto retry;
+	case SCX_OPSS_QUEUEING:
+		/*
+		 * do_enqueue_task() is in the process of transferring the task
+		 * to the BPF scheduler while holding @p's rq lock. As we aren't
+		 * holding any kernel or BPF resource that the enqueue path may
+		 * depend upon, it's safe to wait.
+		 */
+		wait_ops_state(p, opss);
+		goto retry;
+	}
+
+	BUG_ON(!(p->scx.flags & SCX_TASK_QUEUED));
+
+	switch (dispatch_to_local_dsq(rq, rf, dsq_id, p, enq_flags)) {
+	case DTL_DISPATCHED:
+		break;
+	case DTL_LOST:
+		break;
+	case DTL_INVALID:
+		dsq_id = SCX_DSQ_GLOBAL;
+		fallthrough;
+	case DTL_NOT_LOCAL:
+		dsq = find_dsq_for_dispatch(cpu_rq(raw_smp_processor_id()),
+					    dsq_id, p);
+		dispatch_enqueue(dsq, p, enq_flags | SCX_ENQ_CLEAR_OPSS);
+		break;
+	}
+}
+
+static void flush_dispatch_buf(struct rq *rq, struct rq_flags *rf)
+{
+	struct scx_dsp_ctx *dspc = this_cpu_ptr(&scx_dsp_ctx);
+	u32 u;
+
+	for (u = 0; u < dspc->buf_cursor; u++) {
+		struct scx_dsp_buf_ent *ent = &this_cpu_ptr(scx_dsp_buf)[u];
+
+		finish_dispatch(rq, rf, ent->task, ent->qseq, ent->dsq_id,
+				ent->enq_flags);
+	}
+
+	dspc->nr_tasks += dspc->buf_cursor;
+	dspc->buf_cursor = 0;
+}
+
+static int balance_scx(struct rq *rq, struct task_struct *prev,
+		       struct rq_flags *rf)
+{
+	struct scx_rq *scx_rq = &rq->scx;
+	struct scx_dsp_ctx *dspc = this_cpu_ptr(&scx_dsp_ctx);
+	bool prev_on_scx = prev->sched_class == &ext_sched_class;
+
+	lockdep_assert_rq_held(rq);
+
+	if (prev_on_scx) {
+		WARN_ON_ONCE(prev->scx.flags & SCX_TASK_BAL_KEEP);
+		update_curr_scx(rq);
+
+		/*
+		 * If @prev is runnable & has slice left, it has priority and
+		 * fetching more just increases latency for the fetched tasks.
+		 * Tell put_prev_task_scx() to put @prev on local_dsq.
+		 *
+		 * See scx_ops_disable_workfn() for the explanation on the
+		 * disabling() test.
+		 */
+		if ((prev->scx.flags & SCX_TASK_QUEUED) &&
+		    prev->scx.slice && !scx_ops_disabling()) {
+			prev->scx.flags |= SCX_TASK_BAL_KEEP;
+			return 1;
+		}
+	}
+
+	/* if there already are tasks to run, nothing to do */
+	if (scx_rq->local_dsq.nr)
+		return 1;
+
+	if (consume_dispatch_q(rq, rf, &scx_dsq_global))
+		return 1;
+
+	if (!SCX_HAS_OP(dispatch))
+		return 0;
+
+	dspc->rq = rq;
+	dspc->rf = rf;
+
+	/*
+	 * The dispatch loop. Because flush_dispatch_buf() may drop the rq lock,
+	 * the local DSQ might still end up empty after a successful
+	 * ops.dispatch(). If the local DSQ is empty even after ops.dispatch()
+	 * produced some tasks, retry. The BPF scheduler may depend on this
+	 * looping behavior to simplify its implementation.
+	 */
+	do {
+		dspc->nr_tasks = 0;
+
+		SCX_CALL_OP(SCX_KF_DISPATCH, dispatch, cpu_of(rq),
+			    prev_on_scx ? prev : NULL);
+
+		flush_dispatch_buf(rq, rf);
+
+		if (scx_rq->local_dsq.nr)
+			return 1;
+		if (consume_dispatch_q(rq, rf, &scx_dsq_global))
+			return 1;
+	} while (dspc->nr_tasks);
+
+	return 0;
+}
+
+static void set_next_task_scx(struct rq *rq, struct task_struct *p, bool first)
+{
+	if (p->scx.flags & SCX_TASK_QUEUED) {
+		WARN_ON_ONCE(atomic64_read(&p->scx.ops_state) != SCX_OPSS_NONE);
+		dispatch_dequeue(&rq->scx, p);
+	}
+
+	p->se.exec_start = rq_clock_task(rq);
+}
+
+static void put_prev_task_scx(struct rq *rq, struct task_struct *p)
+{
+#ifndef CONFIG_SMP
+	/*
+	 * UP workaround.
+	 *
+	 * Because SCX may transfer tasks across CPUs during dispatch, dispatch
+	 * is performed from its balance operation which isn't called in UP.
+	 * Let's work around by calling it from the operations which come right
+	 * after.
+	 *
+	 * 1. If the prev task is on SCX, pick_next_task() calls
+	 *    .put_prev_task() right after. As .put_prev_task() is also called
+	 *    from other places, we need to distinguish the calls which can be
+	 *    done by looking at the previous task's state - if still queued or
+	 *    dequeued with %SCX_DEQ_SLEEP, the caller must be pick_next_task().
+	 *    This case is handled here.
+	 *
+	 * 2. If the prev task is not on SCX, the first following call into SCX
+	 *    will be .pick_next_task(), which is covered by calling
+	 *    balance_scx() from pick_next_task_scx().
+	 *
+	 * Note that we can't merge the first case into the second as
+	 * balance_scx() must be called before the previous SCX task goes
+	 * through put_prev_task_scx().
+	 *
+	 * As UP doesn't transfer tasks around, balance_scx() doesn't need @rf.
+	 * Pass in %NULL.
+	 */
+	if (p->scx.flags & (SCX_TASK_QUEUED | SCX_TASK_DEQD_FOR_SLEEP))
+		balance_scx(rq, p, NULL);
+#endif
+
+	update_curr_scx(rq);
+
+	/*
+	 * If we're being called from put_prev_task_balance(), balance_scx() may
+	 * have decided that @p should keep running.
+	 */
+	if (p->scx.flags & SCX_TASK_BAL_KEEP) {
+		p->scx.flags &= ~SCX_TASK_BAL_KEEP;
+		dispatch_enqueue(&rq->scx.local_dsq, p, SCX_ENQ_HEAD);
+		return;
+	}
+
+	if (p->scx.flags & SCX_TASK_QUEUED) {
+		/*
+		 * If @p has slice left and balance_scx() didn't tag it for
+		 * keeping, @p is getting preempted by a higher priority
+		 * scheduler class. Leave it at the head of the local DSQ.
+		 */
+		if (p->scx.slice && !scx_ops_disabling()) {
+			dispatch_enqueue(&rq->scx.local_dsq, p, SCX_ENQ_HEAD);
+			return;
+		}
+
+		/*
+		 * If we're in the pick_next_task path, balance_scx() should
+		 * have already populated the local DSQ if there are any other
+		 * available tasks. If empty, tell ops.enqueue() that @p is the
+		 * only one available for this cpu. ops.enqueue() should put it
+		 * on the local DSQ so that the subsequent pick_next_task_scx()
+		 * can find the task unless it wants to trigger a separate
+		 * follow-up scheduling event.
+		 */
+		if (list_empty(&rq->scx.local_dsq.fifo))
+			do_enqueue_task(rq, p, SCX_ENQ_LAST | SCX_ENQ_LOCAL, -1);
+		else
+			do_enqueue_task(rq, p, 0, -1);
+	}
+}
+
+static struct task_struct *first_local_task(struct rq *rq)
+{
+	return list_first_entry_or_null(&rq->scx.local_dsq.fifo,
+					struct task_struct, scx.dsq_node);
+}
+
+static struct task_struct *pick_next_task_scx(struct rq *rq)
+{
+	struct task_struct *p;
+
+#ifndef CONFIG_SMP
+	/* UP workaround - see the comment at the head of put_prev_task_scx() */
+	if (unlikely(rq->curr->sched_class != &ext_sched_class))
+		balance_scx(rq, rq->curr, NULL);
+#endif
+
+	p = first_local_task(rq);
+	if (!p)
+		return NULL;
+
+	if (unlikely(!p->scx.slice)) {
+		if (!scx_ops_disabling() && !scx_warned_zero_slice) {
+			printk_deferred(KERN_WARNING "sched_ext: %s[%d] has zero slice in pick_next_task_scx()\n",
+					p->comm, p->pid);
+			scx_warned_zero_slice = true;
+		}
+		p->scx.slice = SCX_SLICE_DFL;
+	}
+
+	set_next_task_scx(rq, p, true);
+
+	return p;
+}
+
+#ifdef CONFIG_SMP
+
+static bool test_and_clear_cpu_idle(int cpu)
+{
+#ifdef CONFIG_SCHED_SMT
+	/*
+	 * SMT mask should be cleared whether we can claim @cpu or not. The SMT
+	 * cluster is not wholly idle either way. This also prevents
+	 * scx_pick_idle_cpu() from getting caught in an infinite loop.
+	 */
+	if (sched_smt_active()) {
+		const struct cpumask *smt = cpu_smt_mask(cpu);
+
+		/*
+		 * If offline, @cpu is not its own sibling and
+		 * scx_pick_idle_cpu() can get caught in an infinite loop as
+		 * @cpu is never cleared from idle_masks.smt. Ensure that @cpu
+		 * is eventually cleared.
+		 */
+		if (cpumask_intersects(smt, idle_masks.smt))
+			cpumask_andnot(idle_masks.smt, idle_masks.smt, smt);
+		else if (cpumask_test_cpu(cpu, idle_masks.smt))
+			__cpumask_clear_cpu(cpu, idle_masks.smt);
+	}
+#endif
+	return cpumask_test_and_clear_cpu(cpu, idle_masks.cpu);
+}
+
+static s32 scx_pick_idle_cpu(const struct cpumask *cpus_allowed, u64 flags)
+{
+	int cpu;
+
+retry:
+	if (sched_smt_active()) {
+		cpu = cpumask_any_and_distribute(idle_masks.smt, cpus_allowed);
+		if (cpu < nr_cpu_ids)
+			goto found;
+
+		if (flags & SCX_PICK_IDLE_CORE)
+			return -EBUSY;
+	}
+
+	cpu = cpumask_any_and_distribute(idle_masks.cpu, cpus_allowed);
+	if (cpu >= nr_cpu_ids)
+		return -EBUSY;
+
+found:
+	if (test_and_clear_cpu_idle(cpu))
+		return cpu;
+	else
+		goto retry;
+}
+
+static s32 scx_select_cpu_dfl(struct task_struct *p, s32 prev_cpu, u64 wake_flags)
+{
+	s32 cpu;
+
+	if (!static_branch_likely(&scx_builtin_idle_enabled)) {
+		scx_ops_error("built-in idle tracking is disabled");
+		return prev_cpu;
+	}
+
+	/*
+	 * If WAKE_SYNC and the machine isn't fully saturated, wake up @p to the
+	 * local DSQ of the waker.
+	 */
+	if ((wake_flags & SCX_WAKE_SYNC) && p->nr_cpus_allowed > 1 &&
+	    !cpumask_empty(idle_masks.cpu) && !(current->flags & PF_EXITING)) {
+		cpu = smp_processor_id();
+		if (cpumask_test_cpu(cpu, p->cpus_ptr)) {
+			p->scx.flags |= SCX_TASK_ENQ_LOCAL;
+			return cpu;
+		}
+	}
+
+	if (p->nr_cpus_allowed == 1)
+		return prev_cpu;
+
+	/*
+	 * If CPU has SMT, any wholly idle CPU is likely a better pick than
+	 * partially idle @prev_cpu.
+	 */
+	if (sched_smt_active()) {
+		if (cpumask_test_cpu(prev_cpu, idle_masks.smt) &&
+		    test_and_clear_cpu_idle(prev_cpu)) {
+			p->scx.flags |= SCX_TASK_ENQ_LOCAL;
+			return prev_cpu;
+		}
+
+		cpu = scx_pick_idle_cpu(p->cpus_ptr, SCX_PICK_IDLE_CORE);
+		if (cpu >= 0) {
+			p->scx.flags |= SCX_TASK_ENQ_LOCAL;
+			return cpu;
+		}
+	}
+
+	if (test_and_clear_cpu_idle(prev_cpu)) {
+		p->scx.flags |= SCX_TASK_ENQ_LOCAL;
+		return prev_cpu;
+	}
+
+	cpu = scx_pick_idle_cpu(p->cpus_ptr, 0);
+	if (cpu >= 0) {
+		p->scx.flags |= SCX_TASK_ENQ_LOCAL;
+		return cpu;
+	}
+
+	return prev_cpu;
+}
+
+static int select_task_rq_scx(struct task_struct *p, int prev_cpu, int wake_flags)
+{
+	if (SCX_HAS_OP(select_cpu)) {
+		s32 cpu;
+
+		cpu = SCX_CALL_OP_RET(SCX_KF_REST, select_cpu, p, prev_cpu,
+				      wake_flags);
+		if (ops_cpu_valid(cpu)) {
+			return cpu;
+		} else {
+			scx_ops_error("select_cpu returned invalid cpu %d", cpu);
+			return prev_cpu;
+		}
+	} else {
+		return scx_select_cpu_dfl(p, prev_cpu, wake_flags);
+	}
+}
+
+static void set_cpus_allowed_scx(struct task_struct *p,
+				 struct affinity_context *ac)
+{
+	set_cpus_allowed_common(p, ac);
+
+	/*
+	 * The effective cpumask is stored in @p->cpus_ptr which may temporarily
+	 * differ from the configured one in @p->cpus_mask. Always tell the bpf
+	 * scheduler the effective one.
+	 *
+	 * Fine-grained memory write control is enforced by BPF making the const
+	 * designation pointless. Cast it away when calling the operation.
+	 */
+	if (SCX_HAS_OP(set_cpumask))
+		SCX_CALL_OP(SCX_KF_REST, set_cpumask, p,
+			    (struct cpumask *)p->cpus_ptr);
+}
+
+static void reset_idle_masks(void)
+{
+	/* consider all cpus idle, should converge to the actual state quickly */
+	cpumask_setall(idle_masks.cpu);
+	cpumask_setall(idle_masks.smt);
+}
+
+void __scx_update_idle(struct rq *rq, bool idle)
+{
+	int cpu = cpu_of(rq);
+
+	if (SCX_HAS_OP(update_idle)) {
+		SCX_CALL_OP(SCX_KF_REST, update_idle, cpu_of(rq), idle);
+		if (!static_branch_unlikely(&scx_builtin_idle_enabled))
+			return;
+	}
+
+	if (idle)
+		cpumask_set_cpu(cpu, idle_masks.cpu);
+	else
+		cpumask_clear_cpu(cpu, idle_masks.cpu);
+
+#ifdef CONFIG_SCHED_SMT
+	if (sched_smt_active()) {
+		const struct cpumask *smt = cpu_smt_mask(cpu);
+
+		if (idle) {
+			/*
+			 * idle_masks.smt handling is racy but that's fine as
+			 * it's only for optimization and self-correcting.
+			 */
+			for_each_cpu(cpu, smt) {
+				if (!cpumask_test_cpu(cpu, idle_masks.cpu))
+					return;
+			}
+			cpumask_or(idle_masks.smt, idle_masks.smt, smt);
+		} else {
+			cpumask_andnot(idle_masks.smt, idle_masks.smt, smt);
+		}
+	}
+#endif
+}
+
+#else /* !CONFIG_SMP */
+
+static bool test_and_clear_cpu_idle(int cpu) { return false; }
+static s32 scx_pick_idle_cpu(const struct cpumask *cpus_allowed, u64 flags) { return -EBUSY; }
+static void reset_idle_masks(void) {}
+
+#endif /* CONFIG_SMP */
+
+static void task_tick_scx(struct rq *rq, struct task_struct *curr, int queued)
+{
+	update_curr_scx(rq);
+
+	/*
+	 * While disabling, always resched as we can't trust the slice
+	 * management.
+	 */
+	if (scx_ops_disabling())
+		curr->scx.slice = 0;
+
+	if (!curr->scx.slice)
+		resched_curr(rq);
+}
+
+static int scx_ops_prepare_task(struct task_struct *p, struct task_group *tg)
+{
+	int ret;
+
+	WARN_ON_ONCE(p->scx.flags & SCX_TASK_OPS_PREPPED);
+
+	if (SCX_HAS_OP(prep_enable)) {
+		struct scx_enable_args args = { };
+
+		ret = SCX_CALL_OP_RET(SCX_KF_SLEEPABLE, prep_enable, p, &args);
+		if (unlikely(ret)) {
+			ret = ops_sanitize_err("prep_enable", ret);
+			return ret;
+		}
+	}
+
+	p->scx.flags |= SCX_TASK_OPS_PREPPED;
+	return 0;
+}
+
+static void scx_ops_enable_task(struct task_struct *p)
+{
+	lockdep_assert_rq_held(task_rq(p));
+	WARN_ON_ONCE(!(p->scx.flags & SCX_TASK_OPS_PREPPED));
+
+	if (SCX_HAS_OP(enable)) {
+		struct scx_enable_args args = { };
+		SCX_CALL_OP(SCX_KF_REST, enable, p, &args);
+	}
+	p->scx.flags &= ~SCX_TASK_OPS_PREPPED;
+	p->scx.flags |= SCX_TASK_OPS_ENABLED;
+}
+
+static void scx_ops_disable_task(struct task_struct *p)
+{
+	lockdep_assert_rq_held(task_rq(p));
+
+	if (p->scx.flags & SCX_TASK_OPS_PREPPED) {
+		if (SCX_HAS_OP(cancel_enable)) {
+			struct scx_enable_args args = { };
+			SCX_CALL_OP(SCX_KF_REST, cancel_enable, p, &args);
+		}
+		p->scx.flags &= ~SCX_TASK_OPS_PREPPED;
+	} else if (p->scx.flags & SCX_TASK_OPS_ENABLED) {
+		if (SCX_HAS_OP(disable))
+			SCX_CALL_OP(SCX_KF_REST, disable, p);
+		p->scx.flags &= ~SCX_TASK_OPS_ENABLED;
+	}
+}
+
+static void set_task_scx_weight(struct task_struct *p)
+{
+	u32 weight = sched_prio_to_weight[p->static_prio - MAX_RT_PRIO];
+
+	p->scx.weight = sched_weight_to_cgroup(weight);
+}
+
+/**
+ * refresh_scx_weight - Refresh a task's ext weight
+ * @p: task to refresh ext weight for
+ *
+ * @p->scx.weight carries the task's static priority in cgroup weight scale to
+ * enable easy access from the BPF scheduler. To keep it synchronized with the
+ * current task priority, this function should be called when a new task is
+ * created, priority is changed for a task on sched_ext, and a task is switched
+ * to sched_ext from other classes.
+ */
+static void refresh_scx_weight(struct task_struct *p)
+{
+	lockdep_assert_rq_held(task_rq(p));
+	set_task_scx_weight(p);
+	if (SCX_HAS_OP(set_weight))
+		SCX_CALL_OP(SCX_KF_REST, set_weight, p, p->scx.weight);
+}
+
+void scx_pre_fork(struct task_struct *p)
+{
+	/*
+	 * BPF scheduler enable/disable paths want to be able to iterate and
+	 * update all tasks which can become complex when racing forks. As
+	 * enable/disable are very cold paths, let's use a percpu_rwsem to
+	 * exclude forks.
+	 */
+	percpu_down_read(&scx_fork_rwsem);
+}
+
+int scx_fork(struct task_struct *p)
+{
+	percpu_rwsem_assert_held(&scx_fork_rwsem);
+
+	if (scx_enabled())
+		return scx_ops_prepare_task(p, task_group(p));
+	else
+		return 0;
+}
+
+void scx_post_fork(struct task_struct *p)
+{
+	if (scx_enabled()) {
+		struct rq_flags rf;
+		struct rq *rq;
+
+		rq = task_rq_lock(p, &rf);
+		/*
+		 * Set the weight manually before calling ops.enable() so that
+		 * the scheduler doesn't see a stale value if they inspect the
+		 * task struct. We'll invoke ops.set_weight() afterwards, as it
+		 * would be odd to receive a callback on the task before we
+		 * tell the scheduler that it's been fully enabled.
+		 */
+		set_task_scx_weight(p);
+		scx_ops_enable_task(p);
+		refresh_scx_weight(p);
+		task_rq_unlock(rq, p, &rf);
+	}
+
+	spin_lock_irq(&scx_tasks_lock);
+	list_add_tail(&p->scx.tasks_node, &scx_tasks);
+	spin_unlock_irq(&scx_tasks_lock);
+
+	percpu_up_read(&scx_fork_rwsem);
+}
+
+void scx_cancel_fork(struct task_struct *p)
+{
+	if (scx_enabled())
+		scx_ops_disable_task(p);
+	percpu_up_read(&scx_fork_rwsem);
+}
+
+void sched_ext_free(struct task_struct *p)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&scx_tasks_lock, flags);
+	list_del_init(&p->scx.tasks_node);
+	spin_unlock_irqrestore(&scx_tasks_lock, flags);
+
+	/*
+	 * @p is off scx_tasks and wholly ours. scx_ops_enable()'s PREPPED ->
+	 * ENABLED transitions can't race us. Disable ops for @p.
+	 */
+	if (p->scx.flags & (SCX_TASK_OPS_PREPPED | SCX_TASK_OPS_ENABLED)) {
+		struct rq_flags rf;
+		struct rq *rq;
+
+		rq = task_rq_lock(p, &rf);
+		scx_ops_disable_task(p);
+		task_rq_unlock(rq, p, &rf);
+	}
+}
+
+static void reweight_task_scx(struct rq *rq, struct task_struct *p, int newprio)
+{
+	refresh_scx_weight(p);
+}
+
+static void prio_changed_scx(struct rq *rq, struct task_struct *p, int oldprio)
+{
+}
+
+static void switching_to_scx(struct rq *rq, struct task_struct *p)
+{
+	refresh_scx_weight(p);
+
+	/*
+	 * set_cpus_allowed_scx() is not called while @p is associated with a
+	 * different scheduler class. Keep the BPF scheduler up-to-date.
+	 */
+	if (SCX_HAS_OP(set_cpumask))
+		SCX_CALL_OP(SCX_KF_REST, set_cpumask, p,
+			    (struct cpumask *)p->cpus_ptr);
+}
+
+static void check_preempt_curr_scx(struct rq *rq, struct task_struct *p,int wake_flags) {}
+static void switched_to_scx(struct rq *rq, struct task_struct *p) {}
+
+/*
+ * Omitted operations:
+ *
+ * - check_preempt_curr: NOOP as it isn't useful in the wakeup path because the
+ *   task isn't tied to the CPU at that point.
+ *
+ * - migrate_task_rq: Unncessary as task to cpu mapping is transient.
+ *
+ * - task_fork/dead: We need fork/dead notifications for all tasks regardless of
+ *   their current sched_class. Call them directly from sched core instead.
+ *
+ * - task_woken, switched_from: Unnecessary.
+ */
+DEFINE_SCHED_CLASS(ext) = {
+	.enqueue_task		= enqueue_task_scx,
+	.dequeue_task		= dequeue_task_scx,
+	.yield_task		= yield_task_scx,
+	.yield_to_task		= yield_to_task_scx,
+
+	.check_preempt_curr	= check_preempt_curr_scx,
+
+	.pick_next_task		= pick_next_task_scx,
+
+	.put_prev_task		= put_prev_task_scx,
+	.set_next_task          = set_next_task_scx,
+
+#ifdef CONFIG_SMP
+	.balance		= balance_scx,
+	.select_task_rq		= select_task_rq_scx,
+	.set_cpus_allowed	= set_cpus_allowed_scx,
+#endif
+
+	.task_tick		= task_tick_scx,
+
+	.switching_to		= switching_to_scx,
+	.switched_to		= switched_to_scx,
+	.reweight_task		= reweight_task_scx,
+	.prio_changed		= prio_changed_scx,
+
+	.update_curr		= update_curr_scx,
+
+#ifdef CONFIG_UCLAMP_TASK
+	.uclamp_enabled		= 0,
+#endif
+};
+
+static void init_dsq(struct scx_dispatch_q *dsq, u64 dsq_id)
+{
+	memset(dsq, 0, sizeof(*dsq));
+
+	raw_spin_lock_init(&dsq->lock);
+	INIT_LIST_HEAD(&dsq->fifo);
+	dsq->id = dsq_id;
+}
+
+static struct scx_dispatch_q *create_dsq(u64 dsq_id, int node)
+{
+	struct scx_dispatch_q *dsq;
+	int ret;
+
+	if (dsq_id & SCX_DSQ_FLAG_BUILTIN)
+		return ERR_PTR(-EINVAL);
+
+	dsq = kmalloc_node(sizeof(*dsq), GFP_KERNEL, node);
+	if (!dsq)
+		return ERR_PTR(-ENOMEM);
+
+	init_dsq(dsq, dsq_id);
+
+	ret = rhashtable_insert_fast(&dsq_hash, &dsq->hash_node,
+				     dsq_hash_params);
+	if (ret) {
+		kfree(dsq);
+		return ERR_PTR(ret);
+	}
+	return dsq;
+}
+
+static void free_dsq_irq_workfn(struct irq_work *irq_work)
+{
+	struct llist_node *to_free = llist_del_all(&dsqs_to_free);
+	struct scx_dispatch_q *dsq, *tmp_dsq;
+
+	llist_for_each_entry_safe(dsq, tmp_dsq, to_free, free_node)
+		kfree_rcu(dsq);
+}
+
+static DEFINE_IRQ_WORK(free_dsq_irq_work, free_dsq_irq_workfn);
+
+static void destroy_dsq(u64 dsq_id)
+{
+	struct scx_dispatch_q *dsq;
+	unsigned long flags;
+
+	rcu_read_lock();
+
+	dsq = rhashtable_lookup_fast(&dsq_hash, &dsq_id, dsq_hash_params);
+	if (!dsq)
+		goto out_unlock_rcu;
+
+	raw_spin_lock_irqsave(&dsq->lock, flags);
+
+	if (dsq->nr) {
+		scx_ops_error("attempting to destroy in-use dsq 0x%016llx (nr=%u)",
+			      dsq->id, dsq->nr);
+		goto out_unlock_dsq;
+	}
+
+	if (rhashtable_remove_fast(&dsq_hash, &dsq->hash_node, dsq_hash_params))
+		goto out_unlock_dsq;
+
+	/*
+	 * Mark dead by invalidating ->id to prevent dispatch_enqueue() from
+	 * queueing more tasks. As this function can be called from anywhere,
+	 * freeing is bounced through an irq work to avoid nesting RCU
+	 * operations inside scheduler locks.
+	 */
+	dsq->id = SCX_DSQ_INVALID;
+	llist_add(&dsq->free_node, &dsqs_to_free);
+	irq_work_queue(&free_dsq_irq_work);
+
+out_unlock_dsq:
+	raw_spin_unlock_irqrestore(&dsq->lock, flags);
+out_unlock_rcu:
+	rcu_read_unlock();
+}
+
+/*
+ * Used by sched_fork() and __setscheduler_prio() to pick the matching
+ * sched_class. dl/rt are already handled.
+ */
+bool task_should_scx(struct task_struct *p)
+{
+	if (!scx_enabled() || scx_ops_disabling())
+		return false;
+	return p->policy == SCHED_EXT;
+}
+
+static void scx_ops_fallback_enqueue(struct task_struct *p, u64 enq_flags)
+{
+	if (enq_flags & SCX_ENQ_LAST)
+		scx_bpf_dispatch(p, SCX_DSQ_LOCAL, SCX_SLICE_DFL, enq_flags);
+	else
+		scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, SCX_SLICE_DFL, enq_flags);
+}
+
+static void scx_ops_fallback_dispatch(s32 cpu, struct task_struct *prev) {}
+
+static void scx_ops_disable_workfn(struct kthread_work *work)
+{
+	struct scx_exit_info *ei = &scx_exit_info;
+	struct scx_task_iter sti;
+	struct task_struct *p;
+	struct rhashtable_iter rht_iter;
+	struct scx_dispatch_q *dsq;
+	const char *reason;
+	int i, type;
+
+	type = atomic_read(&scx_exit_type);
+	while (true) {
+		/*
+		 * NONE indicates that a new scx_ops has been registered since
+		 * disable was scheduled - don't kill the new ops. DONE
+		 * indicates that the ops has already been disabled.
+		 */
+		if (type == SCX_EXIT_NONE || type == SCX_EXIT_DONE)
+			return;
+		if (atomic_try_cmpxchg(&scx_exit_type, &type, SCX_EXIT_DONE))
+			break;
+	}
+
+	switch (type) {
+	case SCX_EXIT_UNREG:
+		reason = "BPF scheduler unregistered";
+		break;
+	case SCX_EXIT_ERROR:
+		reason = "runtime error";
+		break;
+	case SCX_EXIT_ERROR_BPF:
+		reason = "scx_bpf_error";
+		break;
+	default:
+		reason = "<UNKNOWN>";
+	}
+
+	ei->type = type;
+	strlcpy(ei->reason, reason, sizeof(ei->reason));
+
+	switch (scx_ops_set_enable_state(SCX_OPS_DISABLING)) {
+	case SCX_OPS_DISABLED:
+		pr_warn("sched_ext: ops error detected without ops (%s)\n",
+			scx_exit_info.msg);
+		WARN_ON_ONCE(scx_ops_set_enable_state(SCX_OPS_DISABLED) !=
+			     SCX_OPS_DISABLING);
+		return;
+	case SCX_OPS_PREPPING:
+		goto forward_progress_guaranteed;
+	case SCX_OPS_DISABLING:
+		/* shouldn't happen but handle it like ENABLING if it does */
+		WARN_ONCE(true, "sched_ext: duplicate disabling instance?");
+		fallthrough;
+	case SCX_OPS_ENABLING:
+	case SCX_OPS_ENABLED:
+		break;
+	}
+
+	/*
+	 * DISABLING is set and ops was either ENABLING or ENABLED indicating
+	 * that the ops and static branches are set.
+	 *
+	 * We must guarantee that all runnable tasks make forward progress
+	 * without trusting the BPF scheduler. We can't grab any mutexes or
+	 * rwsems as they might be held by tasks that the BPF scheduler is
+	 * forgetting to run, which unfortunately also excludes toggling the
+	 * static branches.
+	 *
+	 * Let's work around by overriding a couple ops and modifying behaviors
+	 * based on the DISABLING state and then cycling the tasks through
+	 * dequeue/enqueue to force global FIFO scheduling.
+	 *
+	 * a. ops.enqueue() and .dispatch() are overridden for simple global
+	 *    FIFO scheduling.
+	 *
+	 * b. balance_scx() never sets %SCX_TASK_BAL_KEEP as the slice value
+	 *    can't be trusted. Whenever a tick triggers, the running task is
+	 *    rotated to the tail of the queue.
+	 *
+	 * c. pick_next_task() suppresses zero slice warning.
+	 */
+	scx_ops.enqueue = scx_ops_fallback_enqueue;
+	scx_ops.dispatch = scx_ops_fallback_dispatch;
+
+	spin_lock_irq(&scx_tasks_lock);
+	scx_task_iter_init(&sti);
+	while ((p = scx_task_iter_next_filtered_locked(&sti))) {
+		if (READ_ONCE(p->__state) != TASK_DEAD) {
+			struct sched_enq_and_set_ctx ctx;
+
+			/* cycling deq/enq is enough, see above */
+			sched_deq_and_put_task(p, DEQUEUE_SAVE | DEQUEUE_MOVE, &ctx);
+			sched_enq_and_set_task(&ctx);
+		}
+	}
+	scx_task_iter_exit(&sti);
+	spin_unlock_irq(&scx_tasks_lock);
+
+forward_progress_guaranteed:
+	/*
+	 * Here, every runnable task is guaranteed to make forward progress and
+	 * we can safely use blocking synchronization constructs. Actually
+	 * disable ops.
+	 */
+	mutex_lock(&scx_ops_enable_mutex);
+
+	/* avoid racing against fork */
+	cpus_read_lock();
+	percpu_down_write(&scx_fork_rwsem);
+
+	spin_lock_irq(&scx_tasks_lock);
+	scx_task_iter_init(&sti);
+	while ((p = scx_task_iter_next_filtered_locked(&sti))) {
+		const struct sched_class *old_class = p->sched_class;
+		struct sched_enq_and_set_ctx ctx;
+		bool alive = READ_ONCE(p->__state) != TASK_DEAD;
+
+		sched_deq_and_put_task(p, DEQUEUE_SAVE | DEQUEUE_MOVE, &ctx);
+
+		p->scx.slice = min_t(u64, p->scx.slice, SCX_SLICE_DFL);
+
+		__setscheduler_prio(p, p->prio);
+		if (alive)
+			check_class_changing(task_rq(p), p, old_class);
+
+		sched_enq_and_set_task(&ctx);
+
+		if (alive)
+			check_class_changed(task_rq(p), p, old_class, p->prio);
+
+		scx_ops_disable_task(p);
+	}
+	scx_task_iter_exit(&sti);
+	spin_unlock_irq(&scx_tasks_lock);
+
+	/* no task is on scx, turn off all the switches and flush in-progress calls */
+	static_branch_disable_cpuslocked(&__scx_ops_enabled);
+	for (i = 0; i < SCX_NR_ONLINE_OPS; i++)
+		static_branch_disable_cpuslocked(&scx_has_op[i]);
+	static_branch_disable_cpuslocked(&scx_ops_enq_last);
+	static_branch_disable_cpuslocked(&scx_ops_enq_exiting);
+	static_branch_disable_cpuslocked(&scx_builtin_idle_enabled);
+	synchronize_rcu();
+
+	percpu_up_write(&scx_fork_rwsem);
+	cpus_read_unlock();
+
+	if (ei->type >= SCX_EXIT_ERROR) {
+		printk(KERN_ERR "sched_ext: BPF scheduler \"%s\" errored, disabling\n", scx_ops.name);
+
+		if (ei->msg[0] == '\0')
+			printk(KERN_ERR "sched_ext: %s\n", ei->reason);
+		else
+			printk(KERN_ERR "sched_ext: %s (%s)\n", ei->reason, ei->msg);
+
+		stack_trace_print(ei->bt, ei->bt_len, 2);
+	}
+
+	if (scx_ops.exit)
+		SCX_CALL_OP(SCX_KF_UNLOCKED, exit, ei);
+
+	memset(&scx_ops, 0, sizeof(scx_ops));
+
+	rhashtable_walk_enter(&dsq_hash, &rht_iter);
+	do {
+		rhashtable_walk_start(&rht_iter);
+
+		while ((dsq = rhashtable_walk_next(&rht_iter)) && !IS_ERR(dsq))
+			destroy_dsq(dsq->id);
+
+		rhashtable_walk_stop(&rht_iter);
+	} while (dsq == ERR_PTR(-EAGAIN));
+	rhashtable_walk_exit(&rht_iter);
+
+	free_percpu(scx_dsp_buf);
+	scx_dsp_buf = NULL;
+	scx_dsp_max_batch = 0;
+
+	mutex_unlock(&scx_ops_enable_mutex);
+
+	WARN_ON_ONCE(scx_ops_set_enable_state(SCX_OPS_DISABLED) !=
+		     SCX_OPS_DISABLING);
+}
+
+static DEFINE_KTHREAD_WORK(scx_ops_disable_work, scx_ops_disable_workfn);
+
+static void schedule_scx_ops_disable_work(void)
+{
+	struct kthread_worker *helper = READ_ONCE(scx_ops_helper);
+
+	/*
+	 * We may be called spuriously before the first bpf_sched_ext_reg(). If
+	 * scx_ops_helper isn't set up yet, there's nothing to do.
+	 */
+	if (helper)
+		kthread_queue_work(helper, &scx_ops_disable_work);
+}
+
+static void scx_ops_disable(enum scx_exit_type type)
+{
+	int none = SCX_EXIT_NONE;
+
+	if (WARN_ON_ONCE(type == SCX_EXIT_NONE || type == SCX_EXIT_DONE))
+		type = SCX_EXIT_ERROR;
+
+	atomic_try_cmpxchg(&scx_exit_type, &none, type);
+
+	schedule_scx_ops_disable_work();
+}
+
+static void scx_ops_error_irq_workfn(struct irq_work *irq_work)
+{
+	schedule_scx_ops_disable_work();
+}
+
+static DEFINE_IRQ_WORK(scx_ops_error_irq_work, scx_ops_error_irq_workfn);
+
+__printf(2, 3) static void scx_ops_error_type(enum scx_exit_type type,
+					      const char *fmt, ...)
+{
+	struct scx_exit_info *ei = &scx_exit_info;
+	int none = SCX_EXIT_NONE;
+	va_list args;
+
+	if (!atomic_try_cmpxchg(&scx_exit_type, &none, type))
+		return;
+
+	ei->bt_len = stack_trace_save(ei->bt, ARRAY_SIZE(ei->bt), 1);
+
+	va_start(args, fmt);
+	vscnprintf(ei->msg, ARRAY_SIZE(ei->msg), fmt, args);
+	va_end(args);
+
+	irq_work_queue(&scx_ops_error_irq_work);
+}
+
+static struct kthread_worker *scx_create_rt_helper(const char *name)
+{
+	struct kthread_worker *helper;
+
+	helper = kthread_create_worker(0, name);
+	if (helper)
+		sched_set_fifo(helper->task);
+	return helper;
+}
+
+static int scx_ops_enable(struct sched_ext_ops *ops)
+{
+	struct scx_task_iter sti;
+	struct task_struct *p;
+	int i, ret;
+
+	mutex_lock(&scx_ops_enable_mutex);
+
+	if (!scx_ops_helper) {
+		WRITE_ONCE(scx_ops_helper,
+			   scx_create_rt_helper("sched_ext_ops_helper"));
+		if (!scx_ops_helper) {
+			ret = -ENOMEM;
+			goto err_unlock;
+		}
+	}
+
+	if (scx_ops_enable_state() != SCX_OPS_DISABLED) {
+		ret = -EBUSY;
+		goto err_unlock;
+	}
+
+	/*
+	 * Set scx_ops, transition to PREPPING and clear exit info to arm the
+	 * disable path. Failure triggers full disabling from here on.
+	 */
+	scx_ops = *ops;
+
+	WARN_ON_ONCE(scx_ops_set_enable_state(SCX_OPS_PREPPING) !=
+		     SCX_OPS_DISABLED);
+
+	memset(&scx_exit_info, 0, sizeof(scx_exit_info));
+	atomic_set(&scx_exit_type, SCX_EXIT_NONE);
+	scx_warned_zero_slice = false;
+
+	atomic64_set(&scx_nr_rejected, 0);
+
+	/*
+	 * Keep CPUs stable during enable so that the BPF scheduler can track
+	 * online CPUs by watching ->on/offline_cpu() after ->init().
+	 */
+	cpus_read_lock();
+
+	if (scx_ops.init) {
+		ret = SCX_CALL_OP_RET(SCX_KF_INIT, init);
+		if (ret) {
+			ret = ops_sanitize_err("init", ret);
+			goto err_disable;
+		}
+
+		/*
+		 * Exit early if ops.init() triggered scx_bpf_error(). Not
+		 * strictly necessary as we'll fail transitioning into ENABLING
+		 * later but that'd be after calling ops.prep_enable() on all
+		 * tasks and with -EBUSY which isn't very intuitive. Let's exit
+		 * early with success so that the condition is notified through
+		 * ops.exit() like other scx_bpf_error() invocations.
+		 */
+		if (atomic_read(&scx_exit_type) != SCX_EXIT_NONE)
+			goto err_disable;
+	}
+
+	WARN_ON_ONCE(scx_dsp_buf);
+	scx_dsp_max_batch = ops->dispatch_max_batch ?: SCX_DSP_DFL_MAX_BATCH;
+	scx_dsp_buf = __alloc_percpu(sizeof(scx_dsp_buf[0]) * scx_dsp_max_batch,
+				     __alignof__(scx_dsp_buf[0]));
+	if (!scx_dsp_buf) {
+		ret = -ENOMEM;
+		goto err_disable;
+	}
+
+	/*
+	 * Lock out forks before opening the floodgate so that they don't wander
+	 * into the operations prematurely.
+	 */
+	percpu_down_write(&scx_fork_rwsem);
+
+	for (i = 0; i < SCX_NR_ONLINE_OPS; i++)
+		if (((void (**)(void))ops)[i])
+			static_branch_enable_cpuslocked(&scx_has_op[i]);
+
+	if (ops->flags & SCX_OPS_ENQ_LAST)
+		static_branch_enable_cpuslocked(&scx_ops_enq_last);
+
+	if (ops->flags & SCX_OPS_ENQ_EXITING)
+		static_branch_enable_cpuslocked(&scx_ops_enq_exiting);
+
+	if (!ops->update_idle || (ops->flags & SCX_OPS_KEEP_BUILTIN_IDLE)) {
+		reset_idle_masks();
+		static_branch_enable_cpuslocked(&scx_builtin_idle_enabled);
+	} else {
+		static_branch_disable_cpuslocked(&scx_builtin_idle_enabled);
+	}
+
+	static_branch_enable_cpuslocked(&__scx_ops_enabled);
+
+	/*
+	 * Enable ops for every task. Fork is excluded by scx_fork_rwsem
+	 * preventing new tasks from being added. No need to exclude tasks
+	 * leaving as sched_ext_free() can handle both prepped and enabled
+	 * tasks. Prep all tasks first and then enable them with preemption
+	 * disabled.
+	 */
+	spin_lock_irq(&scx_tasks_lock);
+
+	scx_task_iter_init(&sti);
+	while ((p = scx_task_iter_next_filtered(&sti))) {
+		get_task_struct(p);
+		spin_unlock_irq(&scx_tasks_lock);
+
+		ret = scx_ops_prepare_task(p, task_group(p));
+		if (ret) {
+			put_task_struct(p);
+			spin_lock_irq(&scx_tasks_lock);
+			scx_task_iter_exit(&sti);
+			spin_unlock_irq(&scx_tasks_lock);
+			pr_err("sched_ext: ops.prep_enable() failed (%d) for %s[%d] while loading\n",
+			       ret, p->comm, p->pid);
+			goto err_disable_unlock;
+		}
+
+		put_task_struct(p);
+		spin_lock_irq(&scx_tasks_lock);
+	}
+	scx_task_iter_exit(&sti);
+
+	/*
+	 * All tasks are prepped but are still ops-disabled. Ensure that
+	 * %current can't be scheduled out and switch everyone.
+	 * preempt_disable() is necessary because we can't guarantee that
+	 * %current won't be starved if scheduled out while switching.
+	 */
+	preempt_disable();
+
+	/*
+	 * From here on, the disable path must assume that tasks have ops
+	 * enabled and need to be recovered.
+	 */
+	if (!scx_ops_tryset_enable_state(SCX_OPS_ENABLING, SCX_OPS_PREPPING)) {
+		preempt_enable();
+		spin_unlock_irq(&scx_tasks_lock);
+		ret = -EBUSY;
+		goto err_disable_unlock;
+	}
+
+	/*
+	 * We're fully committed and can't fail. The PREPPED -> ENABLED
+	 * transitions here are synchronized against sched_ext_free() through
+	 * scx_tasks_lock.
+	 */
+	scx_task_iter_init(&sti);
+	while ((p = scx_task_iter_next_filtered_locked(&sti))) {
+		if (READ_ONCE(p->__state) != TASK_DEAD) {
+			const struct sched_class *old_class = p->sched_class;
+			struct sched_enq_and_set_ctx ctx;
+
+			sched_deq_and_put_task(p, DEQUEUE_SAVE | DEQUEUE_MOVE,
+					       &ctx);
+
+			scx_ops_enable_task(p);
+			__setscheduler_prio(p, p->prio);
+			check_class_changing(task_rq(p), p, old_class);
+
+			sched_enq_and_set_task(&ctx);
+
+			check_class_changed(task_rq(p), p, old_class, p->prio);
+		} else {
+			scx_ops_disable_task(p);
+		}
+	}
+	scx_task_iter_exit(&sti);
+
+	spin_unlock_irq(&scx_tasks_lock);
+	preempt_enable();
+	percpu_up_write(&scx_fork_rwsem);
+
+	if (!scx_ops_tryset_enable_state(SCX_OPS_ENABLED, SCX_OPS_ENABLING)) {
+		ret = -EBUSY;
+		goto err_disable;
+	}
+
+	cpus_read_unlock();
+	mutex_unlock(&scx_ops_enable_mutex);
+
+	return 0;
+
+err_unlock:
+	mutex_unlock(&scx_ops_enable_mutex);
+	return ret;
+
+err_disable_unlock:
+	percpu_up_write(&scx_fork_rwsem);
+err_disable:
+	cpus_read_unlock();
+	mutex_unlock(&scx_ops_enable_mutex);
+	/* must be fully disabled before returning */
+	scx_ops_disable(SCX_EXIT_ERROR);
+	kthread_flush_work(&scx_ops_disable_work);
+	return ret;
+}
+
+#ifdef CONFIG_SCHED_DEBUG
+static const char *scx_ops_enable_state_str[] = {
+	[SCX_OPS_PREPPING]	= "prepping",
+	[SCX_OPS_ENABLING]	= "enabling",
+	[SCX_OPS_ENABLED]	= "enabled",
+	[SCX_OPS_DISABLING]	= "disabling",
+	[SCX_OPS_DISABLED]	= "disabled",
+};
+
+static int scx_debug_show(struct seq_file *m, void *v)
+{
+	mutex_lock(&scx_ops_enable_mutex);
+	seq_printf(m, "%-30s: %s\n", "ops", scx_ops.name);
+	seq_printf(m, "%-30s: %ld\n", "enabled", scx_enabled());
+	seq_printf(m, "%-30s: %s\n", "enable_state",
+		   scx_ops_enable_state_str[scx_ops_enable_state()]);
+	seq_printf(m, "%-30s: %llu\n", "nr_rejected",
+		   atomic64_read(&scx_nr_rejected));
+	mutex_unlock(&scx_ops_enable_mutex);
+	return 0;
+}
+
+static int scx_debug_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, scx_debug_show, NULL);
+}
+
+const struct file_operations sched_ext_fops = {
+	.open		= scx_debug_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+#endif
+
+/********************************************************************************
+ * bpf_struct_ops plumbing.
+ */
+#include <linux/bpf_verifier.h>
+#include <linux/bpf.h>
+#include <linux/btf.h>
+
+extern struct btf *btf_vmlinux;
+static const struct btf_type *task_struct_type;
+
+static bool bpf_scx_is_valid_access(int off, int size,
+				    enum bpf_access_type type,
+				    const struct bpf_prog *prog,
+				    struct bpf_insn_access_aux *info)
+{
+	if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS)
+		return false;
+	if (type != BPF_READ)
+		return false;
+	if (off % size != 0)
+		return false;
+
+	return btf_ctx_access(off, size, type, prog, info);
+}
+
+static int bpf_scx_btf_struct_access(struct bpf_verifier_log *log,
+				     const struct bpf_reg_state *reg, int off,
+				     int size)
+{
+	const struct btf_type *t;
+
+	t = btf_type_by_id(reg->btf, reg->btf_id);
+	if (t == task_struct_type) {
+		if (off >= offsetof(struct task_struct, scx.slice) &&
+		    off + size <= offsetofend(struct task_struct, scx.slice))
+			return SCALAR_VALUE;
+	}
+
+	return 0;
+}
+
+static const struct bpf_func_proto *
+bpf_scx_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
+{
+	switch (func_id) {
+	case BPF_FUNC_task_storage_get:
+		return &bpf_task_storage_get_proto;
+	case BPF_FUNC_task_storage_delete:
+		return &bpf_task_storage_delete_proto;
+	default:
+		return bpf_base_func_proto(func_id);
+	}
+}
+
+const struct bpf_verifier_ops bpf_scx_verifier_ops = {
+	.get_func_proto = bpf_scx_get_func_proto,
+	.is_valid_access = bpf_scx_is_valid_access,
+	.btf_struct_access = bpf_scx_btf_struct_access,
+};
+
+static int bpf_scx_init_member(const struct btf_type *t,
+			       const struct btf_member *member,
+			       void *kdata, const void *udata)
+{
+	const struct sched_ext_ops *uops = udata;
+	struct sched_ext_ops *ops = kdata;
+	u32 moff = __btf_member_bit_offset(t, member) / 8;
+	int ret;
+
+	switch (moff) {
+	case offsetof(struct sched_ext_ops, dispatch_max_batch):
+		if (*(u32 *)(udata + moff) > INT_MAX)
+			return -E2BIG;
+		ops->dispatch_max_batch = *(u32 *)(udata + moff);
+		return 1;
+	case offsetof(struct sched_ext_ops, flags):
+		if (*(u64 *)(udata + moff) & ~SCX_OPS_ALL_FLAGS)
+			return -EINVAL;
+		ops->flags = *(u64 *)(udata + moff);
+		return 1;
+	case offsetof(struct sched_ext_ops, name):
+		ret = bpf_obj_name_cpy(ops->name, uops->name,
+				       sizeof(ops->name));
+		if (ret < 0)
+			return ret;
+		if (ret == 0)
+			return -EINVAL;
+		return 1;
+	}
+
+	return 0;
+}
+
+static int bpf_scx_check_member(const struct btf_type *t,
+				const struct btf_member *member,
+				const struct bpf_prog *prog)
+{
+	u32 moff = __btf_member_bit_offset(t, member) / 8;
+
+	switch (moff) {
+	case offsetof(struct sched_ext_ops, prep_enable):
+	case offsetof(struct sched_ext_ops, init):
+	case offsetof(struct sched_ext_ops, exit):
+		break;
+	default:
+		if (prog->aux->sleepable)
+			return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int bpf_scx_reg(void *kdata)
+{
+	return scx_ops_enable(kdata);
+}
+
+static void bpf_scx_unreg(void *kdata)
+{
+	scx_ops_disable(SCX_EXIT_UNREG);
+	kthread_flush_work(&scx_ops_disable_work);
+}
+
+static int bpf_scx_init(struct btf *btf)
+{
+	u32 type_id;
+
+	type_id = btf_find_by_name_kind(btf, "task_struct", BTF_KIND_STRUCT);
+	if (type_id < 0)
+		return -EINVAL;
+	task_struct_type = btf_type_by_id(btf, type_id);
+
+	return 0;
+}
+
+static int bpf_scx_update(void *kdata, void *old_kdata)
+{
+	/*
+	 * sched_ext does not support updating the actively-loaded BPF
+	 * scheduler, as registering a BPF scheduler can always fail if the
+	 * scheduler returns an error code for e.g. ops.init(),
+	 * ops.prep_enable(), etc. Similarly, we can always race with
+	 * unregistration happening elsewhere, such as with sysrq.
+	 */
+	return -EOPNOTSUPP;
+}
+
+static int bpf_scx_validate(void *kdata)
+{
+	return 0;
+}
+
+/* "extern" to avoid sparse warning, only used in this file */
+extern struct bpf_struct_ops bpf_sched_ext_ops;
+
+struct bpf_struct_ops bpf_sched_ext_ops = {
+	.verifier_ops = &bpf_scx_verifier_ops,
+	.reg = bpf_scx_reg,
+	.unreg = bpf_scx_unreg,
+	.check_member = bpf_scx_check_member,
+	.init_member = bpf_scx_init_member,
+	.init = bpf_scx_init,
+	.update = bpf_scx_update,
+	.validate = bpf_scx_validate,
+	.name = "sched_ext_ops",
+};
+
+void __init init_sched_ext_class(void)
+{
+	int cpu;
+	u32 v;
+
+	/*
+	 * The following is to prevent the compiler from optimizing out the enum
+	 * definitions so that BPF scheduler implementations can use them
+	 * through the generated vmlinux.h.
+	 */
+	WRITE_ONCE(v, SCX_WAKE_EXEC | SCX_ENQ_WAKEUP | SCX_DEQ_SLEEP);
+
+	BUG_ON(rhashtable_init(&dsq_hash, &dsq_hash_params));
+	init_dsq(&scx_dsq_global, SCX_DSQ_GLOBAL);
+#ifdef CONFIG_SMP
+	BUG_ON(!alloc_cpumask_var(&idle_masks.cpu, GFP_KERNEL));
+	BUG_ON(!alloc_cpumask_var(&idle_masks.smt, GFP_KERNEL));
+#endif
+	for_each_possible_cpu(cpu) {
+		struct rq *rq = cpu_rq(cpu);
+
+		init_dsq(&rq->scx.local_dsq, SCX_DSQ_LOCAL);
+	}
+}
+
+
+/********************************************************************************
+ * Helpers that can be called from the BPF scheduler.
+ */
+#include <linux/btf_ids.h>
+
+/* Disables missing prototype warnings for kfuncs */
+__diag_push();
+__diag_ignore_all("-Wmissing-prototypes",
+		  "Global functions as their definitions will be in vmlinux BTF");
+
+/**
+ * scx_bpf_create_dsq - Create a custom DSQ
+ * @dsq_id: DSQ to create
+ * @node: NUMA node to allocate from
+ *
+ * Create a custom DSQ identified by @dsq_id. Can be called from ops.init() and
+ * ops.prep_enable().
+ */
+s32 scx_bpf_create_dsq(u64 dsq_id, s32 node)
+{
+	if (!scx_kf_allowed(SCX_KF_INIT | SCX_KF_SLEEPABLE))
+		return -EINVAL;
+
+	if (unlikely(node >= (int)nr_node_ids ||
+		     (node < 0 && node != NUMA_NO_NODE)))
+		return -EINVAL;
+	return PTR_ERR_OR_ZERO(create_dsq(dsq_id, node));
+}
+
+BTF_SET8_START(scx_kfunc_ids_sleepable)
+BTF_ID_FLAGS(func, scx_bpf_create_dsq, KF_SLEEPABLE)
+BTF_SET8_END(scx_kfunc_ids_sleepable)
+
+static const struct btf_kfunc_id_set scx_kfunc_set_sleepable = {
+	.owner			= THIS_MODULE,
+	.set			= &scx_kfunc_ids_sleepable,
+};
+
+static bool scx_dispatch_preamble(struct task_struct *p, u64 enq_flags)
+{
+	if (!scx_kf_allowed(SCX_KF_ENQUEUE | SCX_KF_DISPATCH))
+		return false;
+
+	lockdep_assert_irqs_disabled();
+
+	if (unlikely(!p)) {
+		scx_ops_error("called with NULL task");
+		return false;
+	}
+
+	if (unlikely(enq_flags & __SCX_ENQ_INTERNAL_MASK)) {
+		scx_ops_error("invalid enq_flags 0x%llx", enq_flags);
+		return false;
+	}
+
+	return true;
+}
+
+static void scx_dispatch_commit(struct task_struct *p, u64 dsq_id, u64 enq_flags)
+{
+	struct task_struct *ddsp_task;
+	int idx;
+
+	ddsp_task = __this_cpu_read(direct_dispatch_task);
+	if (ddsp_task) {
+		direct_dispatch(ddsp_task, p, dsq_id, enq_flags);
+		return;
+	}
+
+	idx = __this_cpu_read(scx_dsp_ctx.buf_cursor);
+	if (unlikely(idx >= scx_dsp_max_batch)) {
+		scx_ops_error("dispatch buffer overflow");
+		return;
+	}
+
+	this_cpu_ptr(scx_dsp_buf)[idx] = (struct scx_dsp_buf_ent){
+		.task = p,
+		.qseq = atomic64_read(&p->scx.ops_state) & SCX_OPSS_QSEQ_MASK,
+		.dsq_id = dsq_id,
+		.enq_flags = enq_flags,
+	};
+	__this_cpu_inc(scx_dsp_ctx.buf_cursor);
+}
+
+/**
+ * scx_bpf_dispatch - Dispatch a task into the FIFO queue of a DSQ
+ * @p: task_struct to dispatch
+ * @dsq_id: DSQ to dispatch to
+ * @slice: duration @p can run for in nsecs
+ * @enq_flags: SCX_ENQ_*
+ *
+ * Dispatch @p into the FIFO queue of the DSQ identified by @dsq_id. It is safe
+ * to call this function spuriously. Can be called from ops.enqueue() and
+ * ops.dispatch().
+ *
+ * When called from ops.enqueue(), it's for direct dispatch and @p must match
+ * the task being enqueued. Also, %SCX_DSQ_LOCAL_ON can't be used to target the
+ * local DSQ of a CPU other than the enqueueing one. Use ops.select_cpu() to be
+ * on the target CPU in the first place.
+ *
+ * When called from ops.dispatch(), there are no restrictions on @p or @dsq_id
+ * and this function can be called upto ops.dispatch_max_batch times to dispatch
+ * multiple tasks. scx_bpf_dispatch_nr_slots() returns the number of the
+ * remaining slots. scx_bpf_consume() flushes the batch and resets the counter.
+ *
+ * This function doesn't have any locking restrictions and may be called under
+ * BPF locks (in the future when BPF introduces more flexible locking).
+ *
+ * @p is allowed to run for @slice. The scheduling path is triggered on slice
+ * exhaustion. If zero, the current residual slice is maintained. If
+ * %SCX_SLICE_INF, @p never expires and the BPF scheduler must kick the CPU with
+ * scx_bpf_kick_cpu() to trigger scheduling.
+ */
+void scx_bpf_dispatch(struct task_struct *p, u64 dsq_id, u64 slice,
+		      u64 enq_flags)
+{
+	if (!scx_dispatch_preamble(p, enq_flags))
+		return;
+
+	if (slice)
+		p->scx.slice = slice;
+	else
+		p->scx.slice = p->scx.slice ?: 1;
+
+	scx_dispatch_commit(p, dsq_id, enq_flags);
+}
+
+BTF_SET8_START(scx_kfunc_ids_enqueue_dispatch)
+BTF_ID_FLAGS(func, scx_bpf_dispatch, KF_RCU)
+BTF_SET8_END(scx_kfunc_ids_enqueue_dispatch)
+
+static const struct btf_kfunc_id_set scx_kfunc_set_enqueue_dispatch = {
+	.owner			= THIS_MODULE,
+	.set			= &scx_kfunc_ids_enqueue_dispatch,
+};
+
+/**
+ * scx_bpf_dispatch_nr_slots - Return the number of remaining dispatch slots
+ *
+ * Can only be called from ops.dispatch().
+ */
+u32 scx_bpf_dispatch_nr_slots(void)
+{
+	if (!scx_kf_allowed(SCX_KF_DISPATCH))
+		return 0;
+
+	return scx_dsp_max_batch - __this_cpu_read(scx_dsp_ctx.buf_cursor);
+}
+
+/**
+ * scx_bpf_consume - Transfer a task from a DSQ to the current CPU's local DSQ
+ * @dsq_id: DSQ to consume
+ *
+ * Consume a task from the non-local DSQ identified by @dsq_id and transfer it
+ * to the current CPU's local DSQ for execution. Can only be called from
+ * ops.dispatch().
+ *
+ * This function flushes the in-flight dispatches from scx_bpf_dispatch() before
+ * trying to consume the specified DSQ. It may also grab rq locks and thus can't
+ * be called under any BPF locks.
+ *
+ * Returns %true if a task has been consumed, %false if there isn't any task to
+ * consume.
+ */
+bool scx_bpf_consume(u64 dsq_id)
+{
+	struct scx_dsp_ctx *dspc = this_cpu_ptr(&scx_dsp_ctx);
+	struct scx_dispatch_q *dsq;
+
+	if (!scx_kf_allowed(SCX_KF_DISPATCH))
+		return false;
+
+	flush_dispatch_buf(dspc->rq, dspc->rf);
+
+	dsq = find_non_local_dsq(dsq_id);
+	if (unlikely(!dsq)) {
+		scx_ops_error("invalid DSQ ID 0x%016llx", dsq_id);
+		return false;
+	}
+
+	if (consume_dispatch_q(dspc->rq, dspc->rf, dsq)) {
+		/*
+		 * A successfully consumed task can be dequeued before it starts
+		 * running while the CPU is trying to migrate other dispatched
+		 * tasks. Bump nr_tasks to tell balance_scx() to retry on empty
+		 * local DSQ.
+		 */
+		dspc->nr_tasks++;
+		return true;
+	} else {
+		return false;
+	}
+}
+
+BTF_SET8_START(scx_kfunc_ids_dispatch)
+BTF_ID_FLAGS(func, scx_bpf_dispatch_nr_slots)
+BTF_ID_FLAGS(func, scx_bpf_consume)
+BTF_SET8_END(scx_kfunc_ids_dispatch)
+
+static const struct btf_kfunc_id_set scx_kfunc_set_dispatch = {
+	.owner			= THIS_MODULE,
+	.set			= &scx_kfunc_ids_dispatch,
+};
+
+/**
+ * scx_bpf_dsq_nr_queued - Return the number of queued tasks
+ * @dsq_id: id of the DSQ
+ *
+ * Return the number of tasks in the DSQ matching @dsq_id. If not found,
+ * -%ENOENT is returned. Can be called from any non-sleepable online scx_ops
+ * operations.
+ */
+s32 scx_bpf_dsq_nr_queued(u64 dsq_id)
+{
+	struct scx_dispatch_q *dsq;
+
+	lockdep_assert(rcu_read_lock_any_held());
+
+	if (dsq_id == SCX_DSQ_LOCAL) {
+		return this_rq()->scx.local_dsq.nr;
+	} else if ((dsq_id & SCX_DSQ_LOCAL_ON) == SCX_DSQ_LOCAL_ON) {
+		s32 cpu = dsq_id & SCX_DSQ_LOCAL_CPU_MASK;
+
+		if (ops_cpu_valid(cpu))
+			return cpu_rq(cpu)->scx.local_dsq.nr;
+	} else {
+		dsq = find_non_local_dsq(dsq_id);
+		if (dsq)
+			return dsq->nr;
+	}
+	return -ENOENT;
+}
+
+/**
+ * scx_bpf_test_and_clear_cpu_idle - Test and clear @cpu's idle state
+ * @cpu: cpu to test and clear idle for
+ *
+ * Returns %true if @cpu was idle and its idle state was successfully cleared.
+ * %false otherwise.
+ *
+ * Unavailable if ops.update_idle() is implemented and
+ * %SCX_OPS_KEEP_BUILTIN_IDLE is not set.
+ */
+bool scx_bpf_test_and_clear_cpu_idle(s32 cpu)
+{
+	if (!static_branch_likely(&scx_builtin_idle_enabled)) {
+		scx_ops_error("built-in idle tracking is disabled");
+		return false;
+	}
+
+	if (ops_cpu_valid(cpu))
+		return test_and_clear_cpu_idle(cpu);
+	else
+		return false;
+}
+
+/**
+ * scx_bpf_pick_idle_cpu - Pick and claim an idle cpu
+ * @cpus_allowed: Allowed cpumask
+ * @flags: %SCX_PICK_IDLE_CPU_* flags
+ *
+ * Pick and claim an idle cpu in @cpus_allowed. Returns the picked idle cpu
+ * number on success. -%EBUSY if no matching cpu was found.
+ *
+ * Idle CPU tracking may race against CPU scheduling state transitions. For
+ * example, this function may return -%EBUSY as CPUs are transitioning into the
+ * idle state. If the caller then assumes that there will be dispatch events on
+ * the CPUs as they were all busy, the scheduler may end up stalling with CPUs
+ * idling while there are pending tasks. Use scx_bpf_pick_any_cpu() and
+ * scx_bpf_kick_cpu() to guarantee that there will be at least one dispatch
+ * event in the near future.
+ *
+ * Unavailable if ops.update_idle() is implemented and
+ * %SCX_OPS_KEEP_BUILTIN_IDLE is not set.
+ */
+s32 scx_bpf_pick_idle_cpu(const struct cpumask *cpus_allowed, u64 flags)
+{
+	if (!static_branch_likely(&scx_builtin_idle_enabled)) {
+		scx_ops_error("built-in idle tracking is disabled");
+		return -EBUSY;
+	}
+
+	return scx_pick_idle_cpu(cpus_allowed, flags);
+}
+
+/**
+ * scx_bpf_pick_any_cpu - Pick and claim an idle cpu if available or pick any CPU
+ * @cpus_allowed: Allowed cpumask
+ * @flags: %SCX_PICK_IDLE_CPU_* flags
+ *
+ * Pick and claim an idle cpu in @cpus_allowed. If none is available, pick any
+ * CPU in @cpus_allowed. Guaranteed to succeed and returns the picked idle cpu
+ * number if @cpus_allowed is not empty. -%EBUSY is returned if @cpus_allowed is
+ * empty.
+ *
+ * If ops.update_idle() is implemented and %SCX_OPS_KEEP_BUILTIN_IDLE is not
+ * set, this function can't tell which CPUs are idle and will always pick any
+ * CPU.
+ */
+s32 scx_bpf_pick_any_cpu(const struct cpumask *cpus_allowed, u64 flags)
+{
+	s32 cpu;
+
+	if (static_branch_likely(&scx_builtin_idle_enabled)) {
+		cpu = scx_pick_idle_cpu(cpus_allowed, flags);
+		if (cpu >= 0)
+			return cpu;
+	}
+
+	cpu = cpumask_any_distribute(cpus_allowed);
+	if (cpu < nr_cpu_ids)
+		return cpu;
+	else
+		return -EBUSY;
+}
+
+/**
+ * scx_bpf_get_idle_cpumask - Get a referenced kptr to the idle-tracking
+ * per-CPU cpumask.
+ *
+ * Returns NULL if idle tracking is not enabled, or running on a UP kernel.
+ */
+const struct cpumask *scx_bpf_get_idle_cpumask(void)
+{
+	if (!static_branch_likely(&scx_builtin_idle_enabled)) {
+		scx_ops_error("built-in idle tracking is disabled");
+		return cpu_none_mask;
+	}
+
+#ifdef CONFIG_SMP
+	return idle_masks.cpu;
+#else
+	return cpu_none_mask;
+#endif
+}
+
+/**
+ * scx_bpf_get_idle_smtmask - Get a referenced kptr to the idle-tracking,
+ * per-physical-core cpumask. Can be used to determine if an entire physical
+ * core is free.
+ *
+ * Returns NULL if idle tracking is not enabled, or running on a UP kernel.
+ */
+const struct cpumask *scx_bpf_get_idle_smtmask(void)
+{
+	if (!static_branch_likely(&scx_builtin_idle_enabled)) {
+		scx_ops_error("built-in idle tracking is disabled");
+		return cpu_none_mask;
+	}
+
+#ifdef CONFIG_SMP
+	if (sched_smt_active())
+		return idle_masks.smt;
+	else
+		return idle_masks.cpu;
+#else
+	return cpu_none_mask;
+#endif
+}
+
+/**
+ * scx_bpf_put_idle_cpumask - Release a previously acquired referenced kptr to
+ * either the percpu, or SMT idle-tracking cpumask.
+ */
+void scx_bpf_put_idle_cpumask(const struct cpumask *idle_mask)
+{
+	/*
+	 * Empty function body because we aren't actually acquiring or
+	 * releasing a reference to a global idle cpumask, which is read-only
+	 * in the caller and is never released. The acquire / release semantics
+	 * here are just used to make the cpumask is a trusted pointer in the
+	 * caller.
+	 */
+}
+
+struct scx_bpf_error_bstr_bufs {
+	u64			data[MAX_BPRINTF_VARARGS];
+	char			msg[SCX_EXIT_MSG_LEN];
+};
+
+static DEFINE_PER_CPU(struct scx_bpf_error_bstr_bufs, scx_bpf_error_bstr_bufs);
+
+/**
+ * scx_bpf_error_bstr - Indicate fatal error
+ * @fmt: error message format string
+ * @data: format string parameters packaged using ___bpf_fill() macro
+ * @data__sz: @data len, must end in '__sz' for the verifier
+ *
+ * Indicate that the BPF scheduler encountered a fatal error and initiate ops
+ * disabling.
+ */
+void scx_bpf_error_bstr(char *fmt, unsigned long long *data, u32 data__sz)
+{
+	struct bpf_bprintf_data bprintf_data = { .get_bin_args = true };
+	struct scx_bpf_error_bstr_bufs *bufs;
+	unsigned long flags;
+	int ret;
+
+	local_irq_save(flags);
+	bufs = this_cpu_ptr(&scx_bpf_error_bstr_bufs);
+
+	if (data__sz % 8 || data__sz > MAX_BPRINTF_VARARGS * 8 ||
+	    (data__sz && !data)) {
+		scx_ops_error("invalid data=%p and data__sz=%u",
+			      (void *)data, data__sz);
+		goto out_restore;
+	}
+
+	ret = copy_from_kernel_nofault(bufs->data, data, data__sz);
+	if (ret) {
+		scx_ops_error("failed to read data fields (%d)", ret);
+		goto out_restore;
+	}
+
+	ret = bpf_bprintf_prepare(fmt, UINT_MAX, bufs->data, data__sz / 8,
+				  &bprintf_data);
+	if (ret < 0) {
+		scx_ops_error("failed to format prepration (%d)", ret);
+		goto out_restore;
+	}
+
+	ret = bstr_printf(bufs->msg, sizeof(bufs->msg), fmt,
+			  bprintf_data.bin_args);
+	bpf_bprintf_cleanup(&bprintf_data);
+	if (ret < 0) {
+		scx_ops_error("scx_ops_error(\"%s\", %p, %u) failed to format",
+			      fmt, data, data__sz);
+		goto out_restore;
+	}
+
+	scx_ops_error_type(SCX_EXIT_ERROR_BPF, "%s", bufs->msg);
+out_restore:
+	local_irq_restore(flags);
+}
+
+/**
+ * scx_bpf_destroy_dsq - Destroy a custom DSQ
+ * @dsq_id: DSQ to destroy
+ *
+ * Destroy the custom DSQ identified by @dsq_id. Only DSQs created with
+ * scx_bpf_create_dsq() can be destroyed. The caller must ensure that the DSQ is
+ * empty and no further tasks are dispatched to it. Ignored if called on a DSQ
+ * which doesn't exist. Can be called from any online scx_ops operations.
+ */
+void scx_bpf_destroy_dsq(u64 dsq_id)
+{
+	destroy_dsq(dsq_id);
+}
+
+/**
+ * scx_bpf_task_running - Is task currently running?
+ * @p: task of interest
+ */
+bool scx_bpf_task_running(const struct task_struct *p)
+{
+	return task_rq(p)->curr == p;
+}
+
+/**
+ * scx_bpf_task_cpu - CPU a task is currently associated with
+ * @p: task of interest
+ */
+s32 scx_bpf_task_cpu(const struct task_struct *p)
+{
+	return task_cpu(p);
+}
+
+BTF_SET8_START(scx_kfunc_ids_any)
+BTF_ID_FLAGS(func, scx_bpf_dsq_nr_queued)
+BTF_ID_FLAGS(func, scx_bpf_test_and_clear_cpu_idle)
+BTF_ID_FLAGS(func, scx_bpf_pick_idle_cpu, KF_RCU)
+BTF_ID_FLAGS(func, scx_bpf_pick_any_cpu, KF_RCU)
+BTF_ID_FLAGS(func, scx_bpf_get_idle_cpumask, KF_ACQUIRE)
+BTF_ID_FLAGS(func, scx_bpf_get_idle_smtmask, KF_ACQUIRE)
+BTF_ID_FLAGS(func, scx_bpf_put_idle_cpumask, KF_RELEASE)
+BTF_ID_FLAGS(func, scx_bpf_error_bstr, KF_TRUSTED_ARGS)
+BTF_ID_FLAGS(func, scx_bpf_destroy_dsq)
+BTF_ID_FLAGS(func, scx_bpf_task_running, KF_RCU)
+BTF_ID_FLAGS(func, scx_bpf_task_cpu, KF_RCU)
+BTF_SET8_END(scx_kfunc_ids_any)
+
+static const struct btf_kfunc_id_set scx_kfunc_set_any = {
+	.owner			= THIS_MODULE,
+	.set			= &scx_kfunc_ids_any,
+};
+
+__diag_pop();
+
+/*
+ * This can't be done from init_sched_ext_class() as register_btf_kfunc_id_set()
+ * needs most of the system to be up.
+ */
+static int __init register_ext_kfuncs(void)
+{
+	int ret;
+
+	/*
+	 * Some kfuncs are context-sensitive and can only be called from
+	 * specific SCX ops. They are grouped into BTF sets accordingly.
+	 * Unfortunately, BPF currently doesn't have a way of enforcing such
+	 * restrictions. Eventually, the verifier should be able to enforce
+	 * them. For now, register them the same and make each kfunc explicitly
+	 * check using scx_kf_allowed().
+	 */
+	if ((ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
+					     &scx_kfunc_set_sleepable)) ||
+	    (ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
+					     &scx_kfunc_set_enqueue_dispatch)) ||
+	    (ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
+					     &scx_kfunc_set_dispatch)) ||
+	    (ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
+					     &scx_kfunc_set_any))) {
+		pr_err("sched_ext: failed to register kfunc sets (%d)\n", ret);
+		return ret;
+	}
+
+	return 0;
+}
+__initcall(register_ext_kfuncs);
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index 6a93c4825339..d78d151fdbf8 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -1,11 +1,119 @@
 /* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+enum scx_wake_flags {
+	/* expose select WF_* flags as enums */
+	SCX_WAKE_EXEC		= WF_EXEC,
+	SCX_WAKE_FORK		= WF_FORK,
+	SCX_WAKE_TTWU		= WF_TTWU,
+	SCX_WAKE_SYNC		= WF_SYNC,
+};
+
+enum scx_enq_flags {
+	/* expose select ENQUEUE_* flags as enums */
+	SCX_ENQ_WAKEUP		= ENQUEUE_WAKEUP,
+	SCX_ENQ_HEAD		= ENQUEUE_HEAD,
+
+	/* high 32bits are SCX specific */
+
+	/*
+	 * The task being enqueued is the only task available for the cpu. By
+	 * default, ext core keeps executing such tasks but when
+	 * %SCX_OPS_ENQ_LAST is specified, they're ops.enqueue()'d with
+	 * %SCX_ENQ_LAST and %SCX_ENQ_LOCAL flags set.
+	 *
+	 * If the BPF scheduler wants to continue executing the task,
+	 * ops.enqueue() should dispatch the task to %SCX_DSQ_LOCAL immediately.
+	 * If the task gets queued on a different dsq or the BPF side, the BPF
+	 * scheduler is responsible for triggering a follow-up scheduling event.
+	 * Otherwise, Execution may stall.
+	 */
+	SCX_ENQ_LAST		= 1LLU << 41,
+
+	/*
+	 * A hint indicating that it's advisable to enqueue the task on the
+	 * local dsq of the currently selected CPU. Currently used by
+	 * select_cpu_dfl() and together with %SCX_ENQ_LAST.
+	 */
+	SCX_ENQ_LOCAL		= 1LLU << 42,
+
+	/* high 8 bits are internal */
+	__SCX_ENQ_INTERNAL_MASK	= 0xffLLU << 56,
+
+	SCX_ENQ_CLEAR_OPSS	= 1LLU << 56,
+};
+
+enum scx_deq_flags {
+	/* expose select DEQUEUE_* flags as enums */
+	SCX_DEQ_SLEEP		= DEQUEUE_SLEEP,
+};
+
+enum scx_pick_idle_cpu_flags {
+	SCX_PICK_IDLE_CORE	= 1LLU << 0,	/* pick a CPU whose SMT siblings are also idle */
+};
 
 #ifdef CONFIG_SCHED_CLASS_EXT
-#error "NOT IMPLEMENTED YET"
+
+struct sched_enq_and_set_ctx {
+	struct task_struct	*p;
+	int			queue_flags;
+	bool			queued;
+	bool			running;
+};
+
+void sched_deq_and_put_task(struct task_struct *p, int queue_flags,
+			    struct sched_enq_and_set_ctx *ctx);
+void sched_enq_and_set_task(struct sched_enq_and_set_ctx *ctx);
+
+extern const struct sched_class ext_sched_class;
+extern const struct bpf_verifier_ops bpf_sched_ext_verifier_ops;
+extern const struct file_operations sched_ext_fops;
+
+DECLARE_STATIC_KEY_FALSE(__scx_ops_enabled);
+#define scx_enabled()		static_branch_unlikely(&__scx_ops_enabled)
+
+static inline bool task_on_scx(struct task_struct *p)
+{
+	return scx_enabled() && p->sched_class == &ext_sched_class;
+}
+
+bool task_should_scx(struct task_struct *p);
+void scx_pre_fork(struct task_struct *p);
+int scx_fork(struct task_struct *p);
+void scx_post_fork(struct task_struct *p);
+void scx_cancel_fork(struct task_struct *p);
+void init_sched_ext_class(void);
+
+static inline const struct sched_class *next_active_class(const struct sched_class *class)
+{
+	class++;
+	if (!scx_enabled() && class == &ext_sched_class)
+		class++;
+	return class;
+}
+
+#define for_active_class_range(class, _from, _to)				\
+	for (class = (_from); class != (_to); class = next_active_class(class))
+
+#define for_each_active_class(class)						\
+	for_active_class_range(class, __sched_class_highest, __sched_class_lowest)
+
+/*
+ * SCX requires a balance() call before every pick_next_task() call including
+ * when waking up from idle.
+ */
+#define for_balance_class_range(class, prev_class, end_class)			\
+	for_active_class_range(class, (prev_class) > &ext_sched_class ?		\
+			       &ext_sched_class : (prev_class), (end_class))
+
 #else	/* CONFIG_SCHED_CLASS_EXT */
 
 #define scx_enabled()		false
 
+static inline bool task_on_scx(struct task_struct *p) { return false; }
 static inline void scx_pre_fork(struct task_struct *p) {}
 static inline int scx_fork(struct task_struct *p) { return 0; }
 static inline void scx_post_fork(struct task_struct *p) {}
@@ -18,7 +126,13 @@ static inline void init_sched_ext_class(void) {}
 #endif	/* CONFIG_SCHED_CLASS_EXT */
 
 #if defined(CONFIG_SCHED_CLASS_EXT) && defined(CONFIG_SMP)
-#error "NOT IMPLEMENTED YET"
+void __scx_update_idle(struct rq *rq, bool idle);
+
+static inline void scx_update_idle(struct rq *rq, bool idle)
+{
+	if (scx_enabled())
+		__scx_update_idle(rq, idle);
+}
 #else
 static inline void scx_update_idle(struct rq *rq, bool idle) {}
 #endif
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 17bd277cf27a..666166908eb6 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -185,6 +185,10 @@ static inline int idle_policy(int policy)
 
 static inline int normal_policy(int policy)
 {
+#ifdef CONFIG_SCHED_CLASS_EXT
+	if (policy == SCHED_EXT)
+		return true;
+#endif
 	return policy == SCHED_NORMAL;
 }
 
@@ -681,6 +685,15 @@ struct cfs_rq {
 #endif /* CONFIG_FAIR_GROUP_SCHED */
 };
 
+#ifdef CONFIG_SCHED_CLASS_EXT
+struct scx_rq {
+	struct scx_dispatch_q	local_dsq;
+	u64			ops_qseq;
+	u64			extra_enq_flags;	/* see move_task_to_local_dsq() */
+	u32			nr_running;
+};
+#endif /* CONFIG_SCHED_CLASS_EXT */
+
 static inline int rt_bandwidth_enabled(void)
 {
 	return sysctl_sched_rt_runtime >= 0;
@@ -1022,6 +1035,9 @@ struct rq {
 	struct cfs_rq		cfs;
 	struct rt_rq		rt;
 	struct dl_rq		dl;
+#ifdef CONFIG_SCHED_CLASS_EXT
+	struct scx_rq		scx;
+#endif
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	/* list of leaf cfs_rq on this CPU: */
-- 
2.41.0

[PATCH 13/34] sched_ext: Add scx_simple and scx_example_qmap example schedulers

[Tejun Heo]

Add two simple example BPF schedulers - simple and qmap.

* simple: In terms of scheduling, it behaves identical to not having any
  operation implemented at all. The two operations it implements are only to
  improve visibility and exit handling. On certain homogeneous
  configurations, this actually can perform pretty well.

* qmap: A fixed five level priority scheduler to demonstrate queueing PIDs
  on BPF maps for scheduling. While not very practical, this is useful as a
  simple example and will be used to demonstrate different features.

v4: * Dropped _example prefix from scheduler names.

v3: * Rename scx_example_dummy to scx_example_simple and restructure a bit
      to ease later additions. Comment updates.

    * Added declarations for BPF inline iterators. In the future, hopefully,
      these will be consolidated into a generic BPF header so that they
      don't need to be replicated here.

v2: * Updated with the generic BPF cpumask helpers.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 tools/sched_ext/.gitignore       |   5 +
 tools/sched_ext/Makefile         | 188 ++++++++++++++++++++
 tools/sched_ext/gnu/stubs.h      |   1 +
 tools/sched_ext/scx_common.bpf.h | 288 +++++++++++++++++++++++++++++++
 tools/sched_ext/scx_qmap.bpf.c   | 241 ++++++++++++++++++++++++++
 tools/sched_ext/scx_qmap.c       |  84 +++++++++
 tools/sched_ext/scx_simple.bpf.c |  56 ++++++
 tools/sched_ext/scx_simple.c     |  93 ++++++++++
 tools/sched_ext/user_exit_info.h |  50 ++++++
 9 files changed, 1006 insertions(+)
 create mode 100644 tools/sched_ext/.gitignore
 create mode 100644 tools/sched_ext/Makefile
 create mode 100644 tools/sched_ext/gnu/stubs.h
 create mode 100644 tools/sched_ext/scx_common.bpf.h
 create mode 100644 tools/sched_ext/scx_qmap.bpf.c
 create mode 100644 tools/sched_ext/scx_qmap.c
 create mode 100644 tools/sched_ext/scx_simple.bpf.c
 create mode 100644 tools/sched_ext/scx_simple.c
 create mode 100644 tools/sched_ext/user_exit_info.h

diff --git a/tools/sched_ext/.gitignore b/tools/sched_ext/.gitignore
new file mode 100644
index 000000000000..b1dd7580a5b4
--- /dev/null
+++ b/tools/sched_ext/.gitignore
@@ -0,0 +1,5 @@
+scx_simple
+scx_qmap
+*.skel.h
+*.subskel.h
+/tools/
diff --git a/tools/sched_ext/Makefile b/tools/sched_ext/Makefile
new file mode 100644
index 000000000000..f2723a0cde8b
--- /dev/null
+++ b/tools/sched_ext/Makefile
@@ -0,0 +1,188 @@
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+include ../build/Build.include
+include ../scripts/Makefile.arch
+include ../scripts/Makefile.include
+
+ifneq ($(LLVM),)
+ifneq ($(filter %/,$(LLVM)),)
+LLVM_PREFIX := $(LLVM)
+else ifneq ($(filter -%,$(LLVM)),)
+LLVM_SUFFIX := $(LLVM)
+endif
+
+CLANG_TARGET_FLAGS_arm          := arm-linux-gnueabi
+CLANG_TARGET_FLAGS_arm64        := aarch64-linux-gnu
+CLANG_TARGET_FLAGS_hexagon      := hexagon-linux-musl
+CLANG_TARGET_FLAGS_m68k         := m68k-linux-gnu
+CLANG_TARGET_FLAGS_mips         := mipsel-linux-gnu
+CLANG_TARGET_FLAGS_powerpc      := powerpc64le-linux-gnu
+CLANG_TARGET_FLAGS_riscv        := riscv64-linux-gnu
+CLANG_TARGET_FLAGS_s390         := s390x-linux-gnu
+CLANG_TARGET_FLAGS_x86          := x86_64-linux-gnu
+CLANG_TARGET_FLAGS              := $(CLANG_TARGET_FLAGS_$(ARCH))
+
+ifeq ($(CROSS_COMPILE),)
+ifeq ($(CLANG_TARGET_FLAGS),)
+$(error Specify CROSS_COMPILE or add '--target=' option to lib.mk
+else
+CLANG_FLAGS     += --target=$(CLANG_TARGET_FLAGS)
+endif # CLANG_TARGET_FLAGS
+else
+CLANG_FLAGS     += --target=$(notdir $(CROSS_COMPILE:%-=%))
+endif # CROSS_COMPILE
+
+CC := $(LLVM_PREFIX)clang$(LLVM_SUFFIX) $(CLANG_FLAGS) -fintegrated-as
+else
+CC := $(CROSS_COMPILE)gcc
+endif # LLVM
+
+CURDIR := $(abspath .)
+TOOLSDIR := $(abspath ..)
+LIBDIR := $(TOOLSDIR)/lib
+BPFDIR := $(LIBDIR)/bpf
+TOOLSINCDIR := $(TOOLSDIR)/include
+BPFTOOLDIR := $(TOOLSDIR)/bpf/bpftool
+APIDIR := $(TOOLSINCDIR)/uapi
+GENDIR := $(abspath ../../include/generated)
+GENHDR := $(GENDIR)/autoconf.h
+
+SCRATCH_DIR := $(CURDIR)/tools
+BUILD_DIR := $(SCRATCH_DIR)/build
+INCLUDE_DIR := $(SCRATCH_DIR)/include
+BPFOBJ_DIR := $(BUILD_DIR)/libbpf
+BPFOBJ := $(BPFOBJ_DIR)/libbpf.a
+ifneq ($(CROSS_COMPILE),)
+HOST_BUILD_DIR		:= $(BUILD_DIR)/host
+HOST_SCRATCH_DIR	:= host-tools
+HOST_INCLUDE_DIR	:= $(HOST_SCRATCH_DIR)/include
+else
+HOST_BUILD_DIR		:= $(BUILD_DIR)
+HOST_SCRATCH_DIR	:= $(SCRATCH_DIR)
+HOST_INCLUDE_DIR	:= $(INCLUDE_DIR)
+endif
+HOST_BPFOBJ := $(HOST_BUILD_DIR)/libbpf/libbpf.a
+RESOLVE_BTFIDS := $(HOST_BUILD_DIR)/resolve_btfids/resolve_btfids
+DEFAULT_BPFTOOL := $(HOST_SCRATCH_DIR)/sbin/bpftool
+
+VMLINUX_BTF_PATHS ?= $(if $(O),$(O)/vmlinux)					\
+		     $(if $(KBUILD_OUTPUT),$(KBUILD_OUTPUT)/vmlinux)		\
+		     ../../vmlinux						\
+		     /sys/kernel/btf/vmlinux					\
+		     /boot/vmlinux-$(shell uname -r)
+VMLINUX_BTF ?= $(abspath $(firstword $(wildcard $(VMLINUX_BTF_PATHS))))
+ifeq ($(VMLINUX_BTF),)
+$(error Cannot find a vmlinux for VMLINUX_BTF at any of "$(VMLINUX_BTF_PATHS)")
+endif
+
+BPFTOOL ?= $(DEFAULT_BPFTOOL)
+
+ifneq ($(wildcard $(GENHDR)),)
+  GENFLAGS := -DHAVE_GENHDR
+endif
+
+CFLAGS += -g -O2 -rdynamic -pthread -Wall -Werror $(GENFLAGS)			\
+	  -I$(INCLUDE_DIR) -I$(GENDIR) -I$(LIBDIR)				\
+	  -I$(TOOLSINCDIR) -I$(APIDIR)
+
+# Silence some warnings when compiled with clang
+ifneq ($(LLVM),)
+CFLAGS += -Wno-unused-command-line-argument
+endif
+
+LDFLAGS = -lelf -lz -lpthread
+
+IS_LITTLE_ENDIAN = $(shell $(CC) -dM -E - </dev/null |				\
+			grep 'define __BYTE_ORDER__ __ORDER_LITTLE_ENDIAN__')
+
+# Get Clang's default includes on this system, as opposed to those seen by
+# '-target bpf'. This fixes "missing" files on some architectures/distros,
+# such as asm/byteorder.h, asm/socket.h, asm/sockios.h, sys/cdefs.h etc.
+#
+# Use '-idirafter': Don't interfere with include mechanics except where the
+# build would have failed anyways.
+define get_sys_includes
+$(shell $(1) -v -E - </dev/null 2>&1 \
+	| sed -n '/<...> search starts here:/,/End of search list./{ s| \(/.*\)|-idirafter \1|p }') \
+$(shell $(1) -dM -E - </dev/null | grep '__riscv_xlen ' | awk '{printf("-D__riscv_xlen=%d -D__BITS_PER_LONG=%d", $$3, $$3)}')
+endef
+
+BPF_CFLAGS = -g -D__TARGET_ARCH_$(SRCARCH)					\
+	     $(if $(IS_LITTLE_ENDIAN),-mlittle-endian,-mbig-endian)		\
+	     -I$(INCLUDE_DIR) -I$(CURDIR) -I$(APIDIR)				\
+	     -I../../include							\
+	     $(call get_sys_includes,$(CLANG))					\
+	     -Wall -Wno-compare-distinct-pointer-types				\
+	     -O2 -mcpu=v3
+
+all: scx_simple scx_qmap
+
+# sort removes libbpf duplicates when not cross-building
+MAKE_DIRS := $(sort $(BUILD_DIR)/libbpf $(HOST_BUILD_DIR)/libbpf		\
+	       $(HOST_BUILD_DIR)/bpftool $(HOST_BUILD_DIR)/resolve_btfids	\
+	       $(INCLUDE_DIR))
+
+$(MAKE_DIRS):
+	$(call msg,MKDIR,,$@)
+	$(Q)mkdir -p $@
+
+$(BPFOBJ): $(wildcard $(BPFDIR)/*.[ch] $(BPFDIR)/Makefile)			\
+	   $(APIDIR)/linux/bpf.h						\
+	   | $(BUILD_DIR)/libbpf
+	$(Q)$(MAKE) $(submake_extras) -C $(BPFDIR) OUTPUT=$(BUILD_DIR)/libbpf/	\
+		    EXTRA_CFLAGS='-g -O0 -fPIC'					\
+		    DESTDIR=$(SCRATCH_DIR) prefix= all install_headers
+
+$(DEFAULT_BPFTOOL): $(wildcard $(BPFTOOLDIR)/*.[ch] $(BPFTOOLDIR)/Makefile)	\
+		    $(HOST_BPFOBJ) | $(HOST_BUILD_DIR)/bpftool
+	$(Q)$(MAKE) $(submake_extras)  -C $(BPFTOOLDIR)				\
+		    ARCH= CROSS_COMPILE= CC=$(HOSTCC) LD=$(HOSTLD)		\
+		    EXTRA_CFLAGS='-g -O0'					\
+		    OUTPUT=$(HOST_BUILD_DIR)/bpftool/				\
+		    LIBBPF_OUTPUT=$(HOST_BUILD_DIR)/libbpf/			\
+		    LIBBPF_DESTDIR=$(HOST_SCRATCH_DIR)/				\
+		    prefix= DESTDIR=$(HOST_SCRATCH_DIR)/ install-bin
+
+$(INCLUDE_DIR)/vmlinux.h: $(VMLINUX_BTF) $(BPFTOOL) | $(INCLUDE_DIR)
+ifeq ($(VMLINUX_H),)
+	$(call msg,GEN,,$@)
+	$(Q)$(BPFTOOL) btf dump file $(VMLINUX_BTF) format c > $@
+else
+	$(call msg,CP,,$@)
+	$(Q)cp "$(VMLINUX_H)" $@
+endif
+
+%.bpf.o: %.bpf.c $(INCLUDE_DIR)/vmlinux.h scx_common.bpf.h user_exit_info.h	\
+	| $(BPFOBJ)
+	$(call msg,CLNG-BPF,,$@)
+	$(Q)$(CLANG) $(BPF_CFLAGS) -target bpf -c $< -o $@
+
+%.skel.h: %.bpf.o $(BPFTOOL)
+	$(call msg,GEN-SKEL,,$@)
+	$(Q)$(BPFTOOL) gen object $(<:.o=.linked1.o) $<
+	$(Q)$(BPFTOOL) gen object $(<:.o=.linked2.o) $(<:.o=.linked1.o)
+	$(Q)$(BPFTOOL) gen object $(<:.o=.linked3.o) $(<:.o=.linked2.o)
+	$(Q)diff $(<:.o=.linked2.o) $(<:.o=.linked3.o)
+	$(Q)$(BPFTOOL) gen skeleton $(<:.o=.linked3.o) name $(<:.bpf.o=) > $@
+	$(Q)$(BPFTOOL) gen subskeleton $(<:.o=.linked3.o) name $(<:.bpf.o=) > $(@:.skel.h=.subskel.h)
+
+scx_simple: scx_simple.c scx_simple.skel.h user_exit_info.h
+	$(CC) $(CFLAGS) -c $< -o $@.o
+	$(CC) -o $@ $@.o $(HOST_BPFOBJ) $(LDFLAGS)
+
+scx_qmap: scx_qmap.c scx_qmap.skel.h user_exit_info.h
+	$(CC) $(CFLAGS) -c $< -o $@.o
+	$(CC) -o $@ $@.o $(HOST_BPFOBJ) $(LDFLAGS)
+
+clean:
+	rm -rf $(SCRATCH_DIR) $(HOST_SCRATCH_DIR)
+	rm -f *.o *.bpf.o *.skel.h *.subskel.h
+	rm -f scx_simple scx_qmap
+
+.PHONY: all clean
+
+# delete failed targets
+.DELETE_ON_ERROR:
+
+# keep intermediate (.skel.h, .bpf.o, etc) targets
+.SECONDARY:
diff --git a/tools/sched_ext/gnu/stubs.h b/tools/sched_ext/gnu/stubs.h
new file mode 100644
index 000000000000..719225b16626
--- /dev/null
+++ b/tools/sched_ext/gnu/stubs.h
@@ -0,0 +1 @@
+/* dummy .h to trick /usr/include/features.h to work with 'clang -target bpf' */
diff --git a/tools/sched_ext/scx_common.bpf.h b/tools/sched_ext/scx_common.bpf.h
new file mode 100644
index 000000000000..8b8502afcce3
--- /dev/null
+++ b/tools/sched_ext/scx_common.bpf.h
@@ -0,0 +1,288 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+#ifndef __SCHED_EXT_COMMON_BPF_H
+#define __SCHED_EXT_COMMON_BPF_H
+
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include <linux/errno.h>
+#include "user_exit_info.h"
+
+#define PF_WQ_WORKER			0x00000020	/* I'm a workqueue worker */
+#define PF_KTHREAD			0x00200000	/* I am a kernel thread */
+#define PF_EXITING			0x00000004
+#define CLOCK_MONOTONIC			1
+
+/*
+ * Earlier versions of clang/pahole lost upper 32bits in 64bit enums which can
+ * lead to really confusing misbehaviors. Let's trigger a build failure.
+ */
+static inline void ___vmlinux_h_sanity_check___(void)
+{
+	_Static_assert(SCX_DSQ_FLAG_BUILTIN,
+		       "bpftool generated vmlinux.h is missing high bits for 64bit enums, upgrade clang and pahole");
+}
+
+void scx_bpf_error_bstr(char *fmt, unsigned long long *data, u32 data_len) __ksym;
+
+static inline __attribute__((format(printf, 1, 2)))
+void ___scx_bpf_error_format_checker(const char *fmt, ...) {}
+
+/*
+ * scx_bpf_error() wraps the scx_bpf_error_bstr() kfunc with variadic arguments
+ * instead of an array of u64. Note that __param[] must have at least one
+ * element to keep the verifier happy.
+ */
+#define scx_bpf_error(fmt, args...)						\
+({										\
+	static char ___fmt[] = fmt;						\
+	unsigned long long ___param[___bpf_narg(args) ?: 1] = {};		\
+										\
+	_Pragma("GCC diagnostic push")						\
+	_Pragma("GCC diagnostic ignored \"-Wint-conversion\"")			\
+	___bpf_fill(___param, args);						\
+	_Pragma("GCC diagnostic pop")						\
+										\
+	scx_bpf_error_bstr(___fmt, ___param, sizeof(___param));			\
+										\
+	___scx_bpf_error_format_checker(fmt, ##args);				\
+})
+
+s32 scx_bpf_create_dsq(u64 dsq_id, s32 node) __ksym;
+bool scx_bpf_consume(u64 dsq_id) __ksym;
+u32 scx_bpf_dispatch_nr_slots(void) __ksym;
+void scx_bpf_dispatch(struct task_struct *p, u64 dsq_id, u64 slice, u64 enq_flags) __ksym;
+s32 scx_bpf_dsq_nr_queued(u64 dsq_id) __ksym;
+bool scx_bpf_test_and_clear_cpu_idle(s32 cpu) __ksym;
+s32 scx_bpf_pick_idle_cpu(const cpumask_t *cpus_allowed, u64 flags) __ksym;
+s32 scx_bpf_pick_any_cpu(const cpumask_t *cpus_allowed, u64 flags) __ksym;
+const struct cpumask *scx_bpf_get_idle_cpumask(void) __ksym;
+const struct cpumask *scx_bpf_get_idle_smtmask(void) __ksym;
+void scx_bpf_put_idle_cpumask(const struct cpumask *cpumask) __ksym;
+void scx_bpf_destroy_dsq(u64 dsq_id) __ksym;
+bool scx_bpf_task_running(const struct task_struct *p) __ksym;
+s32 scx_bpf_task_cpu(const struct task_struct *p) __ksym;
+
+#define BPF_STRUCT_OPS(name, args...)						\
+SEC("struct_ops/"#name)								\
+BPF_PROG(name, ##args)
+
+#define BPF_STRUCT_OPS_SLEEPABLE(name, args...)					\
+SEC("struct_ops.s/"#name)							\
+BPF_PROG(name, ##args)
+
+/**
+ * MEMBER_VPTR - Obtain the verified pointer to a struct or array member
+ * @base: struct or array to index
+ * @member: dereferenced member (e.g. ->field, [idx0][idx1], ...)
+ *
+ * The verifier often gets confused by the instruction sequence the compiler
+ * generates for indexing struct fields or arrays. This macro forces the
+ * compiler to generate a code sequence which first calculates the byte offset,
+ * checks it against the struct or array size and add that byte offset to
+ * generate the pointer to the member to help the verifier.
+ *
+ * Ideally, we want to abort if the calculated offset is out-of-bounds. However,
+ * BPF currently doesn't support abort, so evaluate to NULL instead. The caller
+ * must check for NULL and take appropriate action to appease the verifier. To
+ * avoid confusing the verifier, it's best to check for NULL and dereference
+ * immediately.
+ *
+ *	vptr = MEMBER_VPTR(my_array, [i][j]);
+ *	if (!vptr)
+ *		return error;
+ *	*vptr = new_value;
+ */
+#define MEMBER_VPTR(base, member) (typeof(base member) *)({			\
+	u64 __base = (u64)base;							\
+	u64 __addr = (u64)&(base member) - __base;				\
+	asm volatile (								\
+		"if %0 <= %[max] goto +2\n"					\
+		"%0 = 0\n"							\
+		"goto +1\n"							\
+		"%0 += %1\n"							\
+		: "+r"(__addr)							\
+		: "r"(__base),							\
+		  [max]"i"(sizeof(base) - sizeof(base member)));		\
+	__addr;									\
+})
+
+/*
+ * BPF core and other generic helpers
+ */
+
+/* list and rbtree */
+#define __contains(name, node) __attribute__((btf_decl_tag("contains:" #name ":" #node)))
+#define private(name) SEC(".data." #name) __hidden __attribute__((aligned(8)))
+
+void *bpf_obj_new_impl(__u64 local_type_id, void *meta) __ksym;
+void bpf_obj_drop_impl(void *kptr, void *meta) __ksym;
+
+#define bpf_obj_new(type) ((type *)bpf_obj_new_impl(bpf_core_type_id_local(type), NULL))
+#define bpf_obj_drop(kptr) bpf_obj_drop_impl(kptr, NULL)
+
+void bpf_list_push_front(struct bpf_list_head *head, struct bpf_list_node *node) __ksym;
+void bpf_list_push_back(struct bpf_list_head *head, struct bpf_list_node *node) __ksym;
+struct bpf_list_node *bpf_list_pop_front(struct bpf_list_head *head) __ksym;
+struct bpf_list_node *bpf_list_pop_back(struct bpf_list_head *head) __ksym;
+struct bpf_rb_node *bpf_rbtree_remove(struct bpf_rb_root *root,
+				      struct bpf_rb_node *node) __ksym;
+int bpf_rbtree_add_impl(struct bpf_rb_root *root, struct bpf_rb_node *node,
+			bool (less)(struct bpf_rb_node *a, const struct bpf_rb_node *b),
+			void *meta, __u64 off) __ksym;
+#define bpf_rbtree_add(head, node, less) bpf_rbtree_add_impl(head, node, less, NULL, 0)
+
+struct bpf_rb_node *bpf_rbtree_first(struct bpf_rb_root *root) __ksym;
+
+/* task */
+struct task_struct *bpf_task_from_pid(s32 pid) __ksym;
+struct task_struct *bpf_task_acquire(struct task_struct *p) __ksym;
+void bpf_task_release(struct task_struct *p) __ksym;
+
+/* cgroup */
+struct cgroup *bpf_cgroup_ancestor(struct cgroup *cgrp, int level) __ksym;
+void bpf_cgroup_release(struct cgroup *cgrp) __ksym;
+struct cgroup *bpf_cgroup_from_id(u64 cgid) __ksym;
+
+/* cpumask */
+struct bpf_cpumask *bpf_cpumask_create(void) __ksym;
+struct bpf_cpumask *bpf_cpumask_acquire(struct bpf_cpumask *cpumask) __ksym;
+void bpf_cpumask_release(struct bpf_cpumask *cpumask) __ksym;
+u32 bpf_cpumask_first(const struct cpumask *cpumask) __ksym;
+u32 bpf_cpumask_first_zero(const struct cpumask *cpumask) __ksym;
+void bpf_cpumask_set_cpu(u32 cpu, struct bpf_cpumask *cpumask) __ksym;
+void bpf_cpumask_clear_cpu(u32 cpu, struct bpf_cpumask *cpumask) __ksym;
+bool bpf_cpumask_test_cpu(u32 cpu, const struct cpumask *cpumask) __ksym;
+bool bpf_cpumask_test_and_set_cpu(u32 cpu, struct bpf_cpumask *cpumask) __ksym;
+bool bpf_cpumask_test_and_clear_cpu(u32 cpu, struct bpf_cpumask *cpumask) __ksym;
+void bpf_cpumask_setall(struct bpf_cpumask *cpumask) __ksym;
+void bpf_cpumask_clear(struct bpf_cpumask *cpumask) __ksym;
+bool bpf_cpumask_and(struct bpf_cpumask *dst, const struct cpumask *src1,
+		     const struct cpumask *src2) __ksym;
+void bpf_cpumask_or(struct bpf_cpumask *dst, const struct cpumask *src1,
+		    const struct cpumask *src2) __ksym;
+void bpf_cpumask_xor(struct bpf_cpumask *dst, const struct cpumask *src1,
+		     const struct cpumask *src2) __ksym;
+bool bpf_cpumask_equal(const struct cpumask *src1, const struct cpumask *src2) __ksym;
+bool bpf_cpumask_intersects(const struct cpumask *src1, const struct cpumask *src2) __ksym;
+bool bpf_cpumask_subset(const struct cpumask *src1, const struct cpumask *src2) __ksym;
+bool bpf_cpumask_empty(const struct cpumask *cpumask) __ksym;
+bool bpf_cpumask_full(const struct cpumask *cpumask) __ksym;
+void bpf_cpumask_copy(struct bpf_cpumask *dst, const struct cpumask *src) __ksym;
+u32 bpf_cpumask_any(const struct cpumask *cpumask) __ksym;
+u32 bpf_cpumask_any_and(const struct cpumask *src1, const struct cpumask *src2) __ksym;
+
+/* rcu */
+void bpf_rcu_read_lock(void) __ksym;
+void bpf_rcu_read_unlock(void) __ksym;
+
+/* BPF core iterators from tools/testing/selftests/bpf/progs/bpf_misc.h */
+struct bpf_iter_num;
+
+extern int bpf_iter_num_new(struct bpf_iter_num *it, int start, int end) __ksym;
+extern int *bpf_iter_num_next(struct bpf_iter_num *it) __ksym;
+extern void bpf_iter_num_destroy(struct bpf_iter_num *it) __ksym;
+
+#ifndef bpf_for_each
+/* bpf_for_each(iter_type, cur_elem, args...) provides generic construct for
+ * using BPF open-coded iterators without having to write mundane explicit
+ * low-level loop logic. Instead, it provides for()-like generic construct
+ * that can be used pretty naturally. E.g., for some hypothetical cgroup
+ * iterator, you'd write:
+ *
+ * struct cgroup *cg, *parent_cg = <...>;
+ *
+ * bpf_for_each(cgroup, cg, parent_cg, CG_ITER_CHILDREN) {
+ *     bpf_printk("Child cgroup id = %d", cg->cgroup_id);
+ *     if (cg->cgroup_id == 123)
+ *         break;
+ * }
+ *
+ * I.e., it looks almost like high-level for each loop in other languages,
+ * supports continue/break, and is verifiable by BPF verifier.
+ *
+ * For iterating integers, the difference betwen bpf_for_each(num, i, N, M)
+ * and bpf_for(i, N, M) is in that bpf_for() provides additional proof to
+ * verifier that i is in [N, M) range, and in bpf_for_each() case i is `int
+ * *`, not just `int`. So for integers bpf_for() is more convenient.
+ *
+ * Note: this macro relies on C99 feature of allowing to declare variables
+ * inside for() loop, bound to for() loop lifetime. It also utilizes GCC
+ * extension: __attribute__((cleanup(<func>))), supported by both GCC and
+ * Clang.
+ */
+#define bpf_for_each(type, cur, args...) for (							\
+	/* initialize and define destructor */							\
+	struct bpf_iter_##type ___it __attribute__((aligned(8), /* enforce, just in case */,	\
+						    cleanup(bpf_iter_##type##_destroy))),	\
+	/* ___p pointer is just to call bpf_iter_##type##_new() *once* to init ___it */		\
+			       *___p __attribute__((unused)) = (				\
+					bpf_iter_##type##_new(&___it, ##args),			\
+	/* this is a workaround for Clang bug: it currently doesn't emit BTF */			\
+	/* for bpf_iter_##type##_destroy() when used from cleanup() attribute */		\
+					(void)bpf_iter_##type##_destroy, (void *)0);		\
+	/* iteration and termination check */							\
+	(((cur) = bpf_iter_##type##_next(&___it)));						\
+)
+#endif /* bpf_for_each */
+
+#ifndef bpf_for
+/* bpf_for(i, start, end) implements a for()-like looping construct that sets
+ * provided integer variable *i* to values starting from *start* through,
+ * but not including, *end*. It also proves to BPF verifier that *i* belongs
+ * to range [start, end), so this can be used for accessing arrays without
+ * extra checks.
+ *
+ * Note: *start* and *end* are assumed to be expressions with no side effects
+ * and whose values do not change throughout bpf_for() loop execution. They do
+ * not have to be statically known or constant, though.
+ *
+ * Note: similarly to bpf_for_each(), it relies on C99 feature of declaring for()
+ * loop bound variables and cleanup attribute, supported by GCC and Clang.
+ */
+#define bpf_for(i, start, end) for (								\
+	/* initialize and define destructor */							\
+	struct bpf_iter_num ___it __attribute__((aligned(8), /* enforce, just in case */	\
+						 cleanup(bpf_iter_num_destroy))),		\
+	/* ___p pointer is necessary to call bpf_iter_num_new() *once* to init ___it */		\
+			    *___p __attribute__((unused)) = (					\
+				bpf_iter_num_new(&___it, (start), (end)),			\
+	/* this is a workaround for Clang bug: it currently doesn't emit BTF */			\
+	/* for bpf_iter_num_destroy() when used from cleanup() attribute */			\
+				(void)bpf_iter_num_destroy, (void *)0);				\
+	({											\
+		/* iteration step */								\
+		int *___t = bpf_iter_num_next(&___it);						\
+		/* termination and bounds check */						\
+		(___t && ((i) = *___t, (i) >= (start) && (i) < (end)));				\
+	});											\
+)
+#endif /* bpf_for */
+
+#ifndef bpf_repeat
+/* bpf_repeat(N) performs N iterations without exposing iteration number
+ *
+ * Note: similarly to bpf_for_each(), it relies on C99 feature of declaring for()
+ * loop bound variables and cleanup attribute, supported by GCC and Clang.
+ */
+#define bpf_repeat(N) for (									\
+	/* initialize and define destructor */							\
+	struct bpf_iter_num ___it __attribute__((aligned(8), /* enforce, just in case */	\
+						 cleanup(bpf_iter_num_destroy))),		\
+	/* ___p pointer is necessary to call bpf_iter_num_new() *once* to init ___it */		\
+			    *___p __attribute__((unused)) = (					\
+				bpf_iter_num_new(&___it, 0, (N)),				\
+	/* this is a workaround for Clang bug: it currently doesn't emit BTF */			\
+	/* for bpf_iter_num_destroy() when used from cleanup() attribute */			\
+				(void)bpf_iter_num_destroy, (void *)0);				\
+	bpf_iter_num_next(&___it);								\
+	/* nothing here  */									\
+)
+#endif /* bpf_repeat */
+
+#endif	/* __SCHED_EXT_COMMON_BPF_H */
diff --git a/tools/sched_ext/scx_qmap.bpf.c b/tools/sched_ext/scx_qmap.bpf.c
new file mode 100644
index 000000000000..686681e2008a
--- /dev/null
+++ b/tools/sched_ext/scx_qmap.bpf.c
@@ -0,0 +1,241 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * A simple five-level FIFO queue scheduler.
+ *
+ * There are five FIFOs implemented using BPF_MAP_TYPE_QUEUE. A task gets
+ * assigned to one depending on its compound weight. Each CPU round robins
+ * through the FIFOs and dispatches more from FIFOs with higher indices - 1 from
+ * queue0, 2 from queue1, 4 from queue2 and so on.
+ *
+ * This scheduler demonstrates:
+ *
+ * - BPF-side queueing using PIDs.
+ * - Sleepable per-task storage allocation using ops.prep_enable().
+ *
+ * This scheduler is primarily for demonstration and testing of sched_ext
+ * features and unlikely to be useful for actual workloads.
+ *
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+#include "scx_common.bpf.h"
+#include <linux/sched/prio.h>
+
+char _license[] SEC("license") = "GPL";
+
+const volatile u64 slice_ns = SCX_SLICE_DFL;
+
+u32 test_error_cnt;
+
+struct user_exit_info uei;
+
+struct qmap {
+	__uint(type, BPF_MAP_TYPE_QUEUE);
+	__uint(max_entries, 4096);
+	__type(value, u32);
+} queue0 SEC(".maps"),
+  queue1 SEC(".maps"),
+  queue2 SEC(".maps"),
+  queue3 SEC(".maps"),
+  queue4 SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_ARRAY_OF_MAPS);
+	__uint(max_entries, 5);
+	__type(key, int);
+	__array(values, struct qmap);
+} queue_arr SEC(".maps") = {
+	.values = {
+		[0] = &queue0,
+		[1] = &queue1,
+		[2] = &queue2,
+		[3] = &queue3,
+		[4] = &queue4,
+	},
+};
+
+/* Per-task scheduling context */
+struct task_ctx {
+	bool	force_local;	/* Dispatch directly to local_dsq */
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_TASK_STORAGE);
+	__uint(map_flags, BPF_F_NO_PREALLOC);
+	__type(key, int);
+	__type(value, struct task_ctx);
+} task_ctx_stor SEC(".maps");
+
+/* Per-cpu dispatch index and remaining count */
+struct {
+	__uint(type, BPF_MAP_TYPE_PERCPU_ARRAY);
+	__uint(max_entries, 2);
+	__type(key, u32);
+	__type(value, u64);
+} dispatch_idx_cnt SEC(".maps");
+
+/* Statistics */
+unsigned long nr_enqueued, nr_dispatched, nr_dequeued;
+
+s32 BPF_STRUCT_OPS(qmap_select_cpu, struct task_struct *p,
+		   s32 prev_cpu, u64 wake_flags)
+{
+	struct task_ctx *tctx;
+	s32 cpu;
+
+	tctx = bpf_task_storage_get(&task_ctx_stor, p, 0, 0);
+	if (!tctx) {
+		scx_bpf_error("task_ctx lookup failed");
+		return -ESRCH;
+	}
+
+	if (p->nr_cpus_allowed == 1 ||
+	    scx_bpf_test_and_clear_cpu_idle(prev_cpu)) {
+		tctx->force_local = true;
+		return prev_cpu;
+	}
+
+	cpu = scx_bpf_pick_idle_cpu(p->cpus_ptr, 0);
+	if (cpu >= 0)
+		return cpu;
+
+	return prev_cpu;
+}
+
+static int weight_to_idx(u32 weight)
+{
+	/* Coarsely map the compound weight to a FIFO. */
+	if (weight <= 25)
+		return 0;
+	else if (weight <= 50)
+		return 1;
+	else if (weight < 200)
+		return 2;
+	else if (weight < 400)
+		return 3;
+	else
+		return 4;
+}
+
+void BPF_STRUCT_OPS(qmap_enqueue, struct task_struct *p, u64 enq_flags)
+{
+	struct task_ctx *tctx;
+	u32 pid = p->pid;
+	int idx = weight_to_idx(p->scx.weight);
+	void *ring;
+
+	if (test_error_cnt && !--test_error_cnt)
+		scx_bpf_error("test triggering error");
+
+	tctx = bpf_task_storage_get(&task_ctx_stor, p, 0, 0);
+	if (!tctx) {
+		scx_bpf_error("task_ctx lookup failed");
+		return;
+	}
+
+	/* Is select_cpu() is telling us to enqueue locally? */
+	if (tctx->force_local) {
+		tctx->force_local = false;
+		scx_bpf_dispatch(p, SCX_DSQ_LOCAL, slice_ns, enq_flags);
+		return;
+	}
+
+	ring = bpf_map_lookup_elem(&queue_arr, &idx);
+	if (!ring) {
+		scx_bpf_error("failed to find ring %d", idx);
+		return;
+	}
+
+	/* Queue on the selected FIFO. If the FIFO overflows, punt to global. */
+	if (bpf_map_push_elem(ring, &pid, 0)) {
+		scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, slice_ns, enq_flags);
+		return;
+	}
+
+	__sync_fetch_and_add(&nr_enqueued, 1);
+}
+
+/*
+ * The BPF queue map doesn't support removal and sched_ext can handle spurious
+ * dispatches. qmap_dequeue() is only used to collect statistics.
+ */
+void BPF_STRUCT_OPS(qmap_dequeue, struct task_struct *p, u64 deq_flags)
+{
+	__sync_fetch_and_add(&nr_dequeued, 1);
+}
+
+void BPF_STRUCT_OPS(qmap_dispatch, s32 cpu, struct task_struct *prev)
+{
+	u32 zero = 0, one = 1;
+	u64 *idx = bpf_map_lookup_elem(&dispatch_idx_cnt, &zero);
+	u64 *cnt = bpf_map_lookup_elem(&dispatch_idx_cnt, &one);
+	void *fifo;
+	s32 pid;
+	int i;
+
+	if (!idx || !cnt) {
+		scx_bpf_error("failed to lookup idx[%p], cnt[%p]", idx, cnt);
+		return;
+	}
+
+	for (i = 0; i < 5; i++) {
+		/* Advance the dispatch cursor and pick the fifo. */
+		if (!*cnt) {
+			*idx = (*idx + 1) % 5;
+			*cnt = 1 << *idx;
+		}
+		(*cnt)--;
+
+		fifo = bpf_map_lookup_elem(&queue_arr, idx);
+		if (!fifo) {
+			scx_bpf_error("failed to find ring %llu", *idx);
+			return;
+		}
+
+		/* Dispatch or advance. */
+		if (!bpf_map_pop_elem(fifo, &pid)) {
+			struct task_struct *p;
+
+			p = bpf_task_from_pid(pid);
+			if (p) {
+				__sync_fetch_and_add(&nr_dispatched, 1);
+				scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, slice_ns, 0);
+				bpf_task_release(p);
+				return;
+			}
+		}
+
+		*cnt = 0;
+	}
+}
+
+s32 BPF_STRUCT_OPS(qmap_prep_enable, struct task_struct *p,
+		   struct scx_enable_args *args)
+{
+	/*
+	 * @p is new. Let's ensure that its task_ctx is available. We can sleep
+	 * in this function and the following will automatically use GFP_KERNEL.
+	 */
+	if (bpf_task_storage_get(&task_ctx_stor, p, 0,
+				 BPF_LOCAL_STORAGE_GET_F_CREATE))
+		return 0;
+	else
+		return -ENOMEM;
+}
+
+void BPF_STRUCT_OPS(qmap_exit, struct scx_exit_info *ei)
+{
+	uei_record(&uei, ei);
+}
+
+SEC(".struct_ops.link")
+struct sched_ext_ops qmap_ops = {
+	.select_cpu		= (void *)qmap_select_cpu,
+	.enqueue		= (void *)qmap_enqueue,
+	.dequeue		= (void *)qmap_dequeue,
+	.dispatch		= (void *)qmap_dispatch,
+	.prep_enable		= (void *)qmap_prep_enable,
+	.exit			= (void *)qmap_exit,
+	.name			= "qmap",
+};
diff --git a/tools/sched_ext/scx_qmap.c b/tools/sched_ext/scx_qmap.c
new file mode 100644
index 000000000000..7e96d8e96f73
--- /dev/null
+++ b/tools/sched_ext/scx_qmap.c
@@ -0,0 +1,84 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <signal.h>
+#include <assert.h>
+#include <libgen.h>
+#include <bpf/bpf.h>
+#include "user_exit_info.h"
+#include "scx_qmap.skel.h"
+
+const char help_fmt[] =
+"A simple five-level FIFO queue sched_ext scheduler.\n"
+"\n"
+"See the top-level comment in .bpf.c for more details.\n"
+"\n"
+"Usage: %s [-s SLICE_US] [-e COUNT]\n"
+"\n"
+"  -s SLICE_US   Override slice duration\n"
+"  -e COUNT      Trigger scx_bpf_error() after COUNT enqueues\n"
+"  -h            Display this help and exit\n";
+
+static volatile int exit_req;
+
+static void sigint_handler(int dummy)
+{
+	exit_req = 1;
+}
+
+int main(int argc, char **argv)
+{
+	struct scx_qmap *skel;
+	struct bpf_link *link;
+	int opt;
+
+	signal(SIGINT, sigint_handler);
+	signal(SIGTERM, sigint_handler);
+
+	libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
+
+	skel = scx_qmap__open();
+	assert(skel);
+
+	while ((opt = getopt(argc, argv, "s:e:tTd:h")) != -1) {
+		switch (opt) {
+		case 's':
+			skel->rodata->slice_ns = strtoull(optarg, NULL, 0) * 1000;
+			break;
+		case 'e':
+			skel->bss->test_error_cnt = strtoul(optarg, NULL, 0);
+			break;
+		default:
+			fprintf(stderr, help_fmt, basename(argv[0]));
+			return opt != 'h';
+		}
+	}
+
+	assert(!scx_qmap__load(skel));
+
+	link = bpf_map__attach_struct_ops(skel->maps.qmap_ops);
+	assert(link);
+
+	while (!exit_req && !uei_exited(&skel->bss->uei)) {
+		long nr_enqueued = skel->bss->nr_enqueued;
+		long nr_dispatched = skel->bss->nr_dispatched;
+
+		printf("enq=%lu, dsp=%lu, delta=%ld, deq=%lu\n",
+		       nr_enqueued, nr_dispatched, nr_enqueued - nr_dispatched,
+		       skel->bss->nr_dequeued);
+		fflush(stdout);
+		sleep(1);
+	}
+
+	bpf_link__destroy(link);
+	uei_print(&skel->bss->uei);
+	scx_qmap__destroy(skel);
+	return 0;
+}
diff --git a/tools/sched_ext/scx_simple.bpf.c b/tools/sched_ext/scx_simple.bpf.c
new file mode 100644
index 000000000000..9326124a32fa
--- /dev/null
+++ b/tools/sched_ext/scx_simple.bpf.c
@@ -0,0 +1,56 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * A simple scheduler.
+ *
+ * A simple global FIFO scheduler. It also demonstrates the following niceties.
+ *
+ * - Statistics tracking how many tasks are queued to local and global dsq's.
+ * - Termination notification for userspace.
+ *
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+#include "scx_common.bpf.h"
+
+char _license[] SEC("license") = "GPL";
+
+struct user_exit_info uei;
+
+struct {
+	__uint(type, BPF_MAP_TYPE_PERCPU_ARRAY);
+	__uint(key_size, sizeof(u32));
+	__uint(value_size, sizeof(u64));
+	__uint(max_entries, 2);			/* [local, global] */
+} stats SEC(".maps");
+
+static void stat_inc(u32 idx)
+{
+	u64 *cnt_p = bpf_map_lookup_elem(&stats, &idx);
+	if (cnt_p)
+		(*cnt_p)++;
+}
+
+void BPF_STRUCT_OPS(simple_enqueue, struct task_struct *p, u64 enq_flags)
+{
+	if (enq_flags & SCX_ENQ_LOCAL) {
+		stat_inc(0);	/* count local queueing */
+		scx_bpf_dispatch(p, SCX_DSQ_LOCAL, SCX_SLICE_DFL, enq_flags);
+		return;
+	}
+
+	stat_inc(1);	/* count global queueing */
+	scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, SCX_SLICE_DFL, enq_flags);
+}
+
+void BPF_STRUCT_OPS(simple_exit, struct scx_exit_info *ei)
+{
+	uei_record(&uei, ei);
+}
+
+SEC(".struct_ops.link")
+struct sched_ext_ops simple_ops = {
+	.enqueue		= (void *)simple_enqueue,
+	.exit			= (void *)simple_exit,
+	.name			= "simple",
+};
diff --git a/tools/sched_ext/scx_simple.c b/tools/sched_ext/scx_simple.c
new file mode 100644
index 000000000000..9ba38ba1e71f
--- /dev/null
+++ b/tools/sched_ext/scx_simple.c
@@ -0,0 +1,93 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <unistd.h>
+#include <signal.h>
+#include <assert.h>
+#include <libgen.h>
+#include <bpf/bpf.h>
+#include "user_exit_info.h"
+#include "scx_simple.skel.h"
+
+const char help_fmt[] =
+"A simple sched_ext scheduler.\n"
+"\n"
+"See the top-level comment in .bpf.c for more details.\n"
+"\n"
+"Usage: %s\n"
+"\n"
+"  -h            Display this help and exit\n";
+
+static volatile int exit_req;
+
+static void sigint_handler(int simple)
+{
+	exit_req = 1;
+}
+
+static void read_stats(struct scx_simple *skel, u64 *stats)
+{
+	int nr_cpus = libbpf_num_possible_cpus();
+	u64 cnts[2][nr_cpus];
+	u32 idx;
+
+	memset(stats, 0, sizeof(stats[0]) * 2);
+
+	for (idx = 0; idx < 2; idx++) {
+		int ret, cpu;
+
+		ret = bpf_map_lookup_elem(bpf_map__fd(skel->maps.stats),
+					  &idx, cnts[idx]);
+		if (ret < 0)
+			continue;
+		for (cpu = 0; cpu < nr_cpus; cpu++)
+			stats[idx] += cnts[idx][cpu];
+	}
+}
+
+int main(int argc, char **argv)
+{
+	struct scx_simple *skel;
+	struct bpf_link *link;
+	u32 opt;
+
+	signal(SIGINT, sigint_handler);
+	signal(SIGTERM, sigint_handler);
+
+	libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
+
+	skel = scx_simple__open();
+	assert(skel);
+
+	while ((opt = getopt(argc, argv, "h")) != -1) {
+		switch (opt) {
+		default:
+			fprintf(stderr, help_fmt, basename(argv[0]));
+			return opt != 'h';
+		}
+	}
+
+	assert(!scx_simple__load(skel));
+
+	link = bpf_map__attach_struct_ops(skel->maps.simple_ops);
+	assert(link);
+
+	while (!exit_req && !uei_exited(&skel->bss->uei)) {
+		u64 stats[2];
+
+		read_stats(skel, stats);
+		printf("local=%lu global=%lu\n", stats[0], stats[1]);
+		fflush(stdout);
+		sleep(1);
+	}
+
+	bpf_link__destroy(link);
+	uei_print(&skel->bss->uei);
+	scx_simple__destroy(skel);
+	return 0;
+}
diff --git a/tools/sched_ext/user_exit_info.h b/tools/sched_ext/user_exit_info.h
new file mode 100644
index 000000000000..e701ef0e0b86
--- /dev/null
+++ b/tools/sched_ext/user_exit_info.h
@@ -0,0 +1,50 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Define struct user_exit_info which is shared between BPF and userspace parts
+ * to communicate exit status and other information.
+ *
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+#ifndef __USER_EXIT_INFO_H
+#define __USER_EXIT_INFO_H
+
+struct user_exit_info {
+	int		type;
+	char		reason[128];
+	char		msg[1024];
+};
+
+#ifdef __bpf__
+
+#include "vmlinux.h"
+#include <bpf/bpf_core_read.h>
+
+static inline void uei_record(struct user_exit_info *uei,
+			      const struct scx_exit_info *ei)
+{
+	bpf_probe_read_kernel_str(uei->reason, sizeof(uei->reason), ei->reason);
+	bpf_probe_read_kernel_str(uei->msg, sizeof(uei->msg), ei->msg);
+	/* use __sync to force memory barrier */
+	__sync_val_compare_and_swap(&uei->type, uei->type, ei->type);
+}
+
+#else	/* !__bpf__ */
+
+static inline bool uei_exited(struct user_exit_info *uei)
+{
+	/* use __sync to force memory barrier */
+	return __sync_val_compare_and_swap(&uei->type, -1, -1);
+}
+
+static inline void uei_print(const struct user_exit_info *uei)
+{
+	fprintf(stderr, "EXIT: %s", uei->reason);
+	if (uei->msg[0] != '\0')
+		fprintf(stderr, " (%s)", uei->msg);
+	fputs("\n", stderr);
+}
+
+#endif	/* __bpf__ */
+#endif	/* __USER_EXIT_INFO_H */
-- 
2.41.0

[PATCH 14/34] sched_ext: Add sysrq-S which disables the BPF scheduler

[Tejun Heo]

This enables the admin to abort the BPF scheduler and revert to CFS anytime.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 drivers/tty/sysrq.c         |  1 +
 include/linux/sched/ext.h   |  1 +
 kernel/sched/build_policy.c |  1 +
 kernel/sched/ext.c          | 20 ++++++++++++++++++++
 4 files changed, 23 insertions(+)

diff --git a/drivers/tty/sysrq.c b/drivers/tty/sysrq.c
index b6e70c5cfa17..ddfcdb6aecd7 100644
--- a/drivers/tty/sysrq.c
+++ b/drivers/tty/sysrq.c
@@ -520,6 +520,7 @@ static const struct sysrq_key_op *sysrq_key_table[62] = {
 	NULL,				/* P */
 	NULL,				/* Q */
 	NULL,				/* R */
+	/* S: May be registered by sched_ext for resetting */
 	NULL,				/* S */
 	NULL,				/* T */
 	NULL,				/* U */
diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 92011a63cc15..2f24642bbf2e 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -55,6 +55,7 @@ enum scx_exit_type {
 	SCX_EXIT_DONE,
 
 	SCX_EXIT_UNREG = 64,	/* BPF unregistration */
+	SCX_EXIT_SYSRQ,		/* requested by 'S' sysrq */
 
 	SCX_EXIT_ERROR = 1024,	/* runtime error, error msg contains details */
 	SCX_EXIT_ERROR_BPF,	/* ERROR but triggered through scx_bpf_error() */
diff --git a/kernel/sched/build_policy.c b/kernel/sched/build_policy.c
index 4c658b21f603..005025f55bea 100644
--- a/kernel/sched/build_policy.c
+++ b/kernel/sched/build_policy.c
@@ -28,6 +28,7 @@
 #include <linux/suspend.h>
 #include <linux/tsacct_kern.h>
 #include <linux/vtime.h>
+#include <linux/sysrq.h>
 #include <linux/percpu-rwsem.h>
 
 #include <uapi/linux/sched/types.h>
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 51d77459d208..f33df61c24cb 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -1971,6 +1971,9 @@ static void scx_ops_disable_workfn(struct kthread_work *work)
 	case SCX_EXIT_UNREG:
 		reason = "BPF scheduler unregistered";
 		break;
+	case SCX_EXIT_SYSRQ:
+		reason = "disabled by sysrq-S";
+		break;
 	case SCX_EXIT_ERROR:
 		reason = "runtime error";
 		break;
@@ -2593,6 +2596,21 @@ struct bpf_struct_ops bpf_sched_ext_ops = {
 	.name = "sched_ext_ops",
 };
 
+static void sysrq_handle_sched_ext_reset(int key)
+{
+	if (scx_ops_helper)
+		scx_ops_disable(SCX_EXIT_SYSRQ);
+	else
+		pr_info("sched_ext: BPF scheduler not yet used\n");
+}
+
+static const struct sysrq_key_op sysrq_sched_ext_reset_op = {
+	.handler	= sysrq_handle_sched_ext_reset,
+	.help_msg	= "reset-sched-ext(S)",
+	.action_msg	= "Disable sched_ext and revert all tasks to CFS",
+	.enable_mask	= SYSRQ_ENABLE_RTNICE,
+};
+
 void __init init_sched_ext_class(void)
 {
 	int cpu;
@@ -2616,6 +2634,8 @@ void __init init_sched_ext_class(void)
 
 		init_dsq(&rq->scx.local_dsq, SCX_DSQ_LOCAL);
 	}
+
+	register_sysrq_key('S', &sysrq_sched_ext_reset_op);
 }
 
 
-- 
2.41.0

[PATCH 15/34] sched_ext: Implement runnable task stall watchdog

[Tejun Heo]

From: David Vernet <dvernet@meta.com>

The most common and critical way that a BPF scheduler can misbehave is by
failing to run runnable tasks for too long. This patch implements a
watchdog.

* All tasks record when they become runnable.

* A watchdog work periodically scans all runnable tasks. If any task has
  stayed runnable for too long, the BPF scheduler is aborted.

* scheduler_tick() monitors whether the watchdog itself is stuck. If so, the
  BPF scheduler is aborted.

Because the watchdog only scans the tasks which are currently runnable and
usually very infrequently, the overhead should be negligible.
scx_qmap is updated so that it can be told to stall user and/or
kernel tasks.

A detected task stall looks like the following:

 sched_ext: BPF scheduler "qmap" errored, disabling
 sched_ext: runnable task stall (dbus-daemon[953] failed to run for 6.478s)
    scx_check_timeout_workfn+0x10e/0x1b0
    process_one_work+0x287/0x560
    worker_thread+0x234/0x420
    kthread+0xe9/0x100
    ret_from_fork+0x1f/0x30

A detected watchdog stall:

 sched_ext: BPF scheduler "qmap" errored, disabling
 sched_ext: runnable task stall (watchdog failed to check in for 5.001s)
    scheduler_tick+0x2eb/0x340
    update_process_times+0x7a/0x90
    tick_sched_timer+0xd8/0x130
    __hrtimer_run_queues+0x178/0x3b0
    hrtimer_interrupt+0xfc/0x390
    __sysvec_apic_timer_interrupt+0xb7/0x2b0
    sysvec_apic_timer_interrupt+0x90/0xb0
    asm_sysvec_apic_timer_interrupt+0x1b/0x20
    default_idle+0x14/0x20
    arch_cpu_idle+0xf/0x20
    default_idle_call+0x50/0x90
    do_idle+0xe8/0x240
    cpu_startup_entry+0x1d/0x20
    kernel_init+0x0/0x190
    start_kernel+0x0/0x392
    start_kernel+0x324/0x392
    x86_64_start_reservations+0x2a/0x2c
    x86_64_start_kernel+0x104/0x109
    secondary_startup_64_no_verify+0xce/0xdb

Note that this patch exposes scx_ops_error[_type]() in kernel/sched/ext.h to
inline scx_notify_sched_tick().

v2: Julia Lawall noticed that the watchdog code was mixing msecs and
    jiffies. Fix by using jiffies for everything.

Signed-off-by: David Vernet <dvernet@meta.com>
Reviewed-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
Cc: Julia Lawall <julia.lawall@inria.fr>
---
 include/linux/sched/ext.h      |  13 ++++
 init/init_task.c               |   2 +
 kernel/sched/core.c            |   3 +
 kernel/sched/ext.c             | 128 ++++++++++++++++++++++++++++++---
 kernel/sched/ext.h             |  25 +++++++
 kernel/sched/sched.h           |   1 +
 tools/sched_ext/scx_qmap.bpf.c |  12 ++++
 tools/sched_ext/scx_qmap.c     |  12 +++-
 8 files changed, 185 insertions(+), 11 deletions(-)

diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 2f24642bbf2e..1588e4496afa 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -59,6 +59,7 @@ enum scx_exit_type {
 
 	SCX_EXIT_ERROR = 1024,	/* runtime error, error msg contains details */
 	SCX_EXIT_ERROR_BPF,	/* ERROR but triggered through scx_bpf_error() */
+	SCX_EXIT_ERROR_STALL,	/* watchdog detected stalled runnable tasks */
 };
 
 /*
@@ -314,6 +315,15 @@ struct sched_ext_ops {
 	 */
 	u64 flags;
 
+	/**
+	 * timeout_ms - The maximum amount of time, in milliseconds, that a
+	 * runnable task should be able to wait before being scheduled. The
+	 * maximum timeout may not exceed the default timeout of 30 seconds.
+	 *
+	 * Defaults to the maximum allowed timeout value of 30 seconds.
+	 */
+	u32 timeout_ms;
+
 	/**
 	 * name - BPF scheduler's name
 	 *
@@ -347,6 +357,7 @@ enum scx_ent_flags {
 	SCX_TASK_OPS_PREPPED	= 1 << 8, /* prepared for BPF scheduler enable */
 	SCX_TASK_OPS_ENABLED	= 1 << 9, /* task has BPF scheduler enabled */
 
+	SCX_TASK_WATCHDOG_RESET = 1 << 16, /* task watchdog counter should be reset */
 	SCX_TASK_DEQD_FOR_SLEEP	= 1 << 17, /* last dequeue was for SLEEP */
 
 	SCX_TASK_CURSOR		= 1 << 31, /* iteration cursor, not a task */
@@ -380,12 +391,14 @@ enum scx_kf_mask {
 struct sched_ext_entity {
 	struct scx_dispatch_q	*dsq;
 	struct list_head	dsq_node;
+	struct list_head	watchdog_node;
 	u32			flags;		/* protected by rq lock */
 	u32			weight;
 	s32			sticky_cpu;
 	s32			holding_cpu;
 	u32			kf_mask;	/* see scx_kf_mask above */
 	atomic64_t		ops_state;
+	unsigned long		runnable_at;
 
 	/* BPF scheduler modifiable fields */
 
diff --git a/init/init_task.c b/init/init_task.c
index bdbc663107bf..913194aab623 100644
--- a/init/init_task.c
+++ b/init/init_task.c
@@ -106,9 +106,11 @@ struct task_struct init_task
 #ifdef CONFIG_SCHED_CLASS_EXT
 	.scx		= {
 		.dsq_node	= LIST_HEAD_INIT(init_task.scx.dsq_node),
+		.watchdog_node	= LIST_HEAD_INIT(init_task.scx.watchdog_node),
 		.sticky_cpu	= -1,
 		.holding_cpu	= -1,
 		.ops_state	= ATOMIC_INIT(0),
+		.runnable_at	= INITIAL_JIFFIES,
 		.slice		= SCX_SLICE_DFL,
 	},
 #endif
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index a182a9adec3e..9d97c0e6442a 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4508,12 +4508,14 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
 #ifdef CONFIG_SCHED_CLASS_EXT
 	p->scx.dsq		= NULL;
 	INIT_LIST_HEAD(&p->scx.dsq_node);
+	INIT_LIST_HEAD(&p->scx.watchdog_node);
 	p->scx.flags		= 0;
 	p->scx.weight		= 0;
 	p->scx.sticky_cpu	= -1;
 	p->scx.holding_cpu	= -1;
 	p->scx.kf_mask		= 0;
 	atomic64_set(&p->scx.ops_state, 0);
+	p->scx.runnable_at	= INITIAL_JIFFIES;
 	p->scx.slice		= SCX_SLICE_DFL;
 #endif
 
@@ -5684,6 +5686,7 @@ void scheduler_tick(void)
 	if (sched_feat(LATENCY_WARN) && resched_latency)
 		resched_latency_warn(cpu, resched_latency);
 
+	scx_notify_sched_tick();
 	perf_event_task_tick();
 
 #ifdef CONFIG_SMP
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index f33df61c24cb..eca1790375b8 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -9,6 +9,7 @@
 enum scx_internal_consts {
 	SCX_NR_ONLINE_OPS	= SCX_OP_IDX(init),
 	SCX_DSP_DFL_MAX_BATCH	= 32,
+	SCX_WATCHDOG_MAX_TIMEOUT = 30 * HZ,
 };
 
 enum scx_ops_enable_state {
@@ -87,6 +88,23 @@ static struct scx_exit_info scx_exit_info;
 
 static atomic64_t scx_nr_rejected = ATOMIC64_INIT(0);
 
+/*
+ * The maximum amount of time in jiffies that a task may be runnable without
+ * being scheduled on a CPU. If this timeout is exceeded, it will trigger
+ * scx_ops_error().
+ */
+unsigned long scx_watchdog_timeout;
+
+/*
+ * The last time the delayed work was run. This delayed work relies on
+ * ksoftirqd being able to run to service timer interrupts, so it's possible
+ * that this work itself could get wedged. To account for this, we check that
+ * it's not stalled in the timer tick, and trigger an error if it is.
+ */
+unsigned long scx_watchdog_timestamp = INITIAL_JIFFIES;
+
+static struct delayed_work scx_watchdog_work;
+
 /* idle tracking */
 #ifdef CONFIG_SMP
 #ifdef CONFIG_CPUMASK_OFFSTACK
@@ -145,10 +163,6 @@ static DEFINE_PER_CPU(struct scx_dsp_ctx, scx_dsp_ctx);
 
 void scx_bpf_dispatch(struct task_struct *p, u64 dsq_id, u64 slice,
 		      u64 enq_flags);
-__printf(2, 3) static void scx_ops_error_type(enum scx_exit_type type,
-					      const char *fmt, ...);
-#define scx_ops_error(fmt, args...)						\
-	scx_ops_error_type(SCX_EXIT_ERROR, fmt, ##args)
 
 struct scx_task_iter {
 	struct sched_ext_entity		cursor;
@@ -698,6 +712,27 @@ static void do_enqueue_task(struct rq *rq, struct task_struct *p, u64 enq_flags,
 	dispatch_enqueue(&scx_dsq_global, p, enq_flags);
 }
 
+static bool watchdog_task_watched(const struct task_struct *p)
+{
+	return !list_empty(&p->scx.watchdog_node);
+}
+
+static void watchdog_watch_task(struct rq *rq, struct task_struct *p)
+{
+	lockdep_assert_rq_held(rq);
+	if (p->scx.flags & SCX_TASK_WATCHDOG_RESET)
+		p->scx.runnable_at = jiffies;
+	p->scx.flags &= ~SCX_TASK_WATCHDOG_RESET;
+	list_add_tail(&p->scx.watchdog_node, &rq->scx.watchdog_list);
+}
+
+static void watchdog_unwatch_task(struct task_struct *p, bool reset_timeout)
+{
+	list_del_init(&p->scx.watchdog_node);
+	if (reset_timeout)
+		p->scx.flags |= SCX_TASK_WATCHDOG_RESET;
+}
+
 static void enqueue_task_scx(struct rq *rq, struct task_struct *p, int enq_flags)
 {
 	int sticky_cpu = p->scx.sticky_cpu;
@@ -716,9 +751,12 @@ static void enqueue_task_scx(struct rq *rq, struct task_struct *p, int enq_flags
 	if (unlikely(enq_flags & ENQUEUE_RESTORE) && task_current(rq, p))
 		sticky_cpu = cpu_of(rq);
 
-	if (p->scx.flags & SCX_TASK_QUEUED)
+	if (p->scx.flags & SCX_TASK_QUEUED) {
+		WARN_ON_ONCE(!watchdog_task_watched(p));
 		return;
+	}
 
+	watchdog_watch_task(rq, p);
 	p->scx.flags |= SCX_TASK_QUEUED;
 	rq->scx.nr_running++;
 	add_nr_running(rq, 1);
@@ -730,6 +768,8 @@ static void ops_dequeue(struct task_struct *p, u64 deq_flags)
 {
 	u64 opss;
 
+	watchdog_unwatch_task(p, false);
+
 	/* acquire ensures that we see the preceding updates on QUEUED */
 	opss = atomic64_read_acquire(&p->scx.ops_state);
 
@@ -774,8 +814,10 @@ static void dequeue_task_scx(struct rq *rq, struct task_struct *p, int deq_flags
 {
 	struct scx_rq *scx_rq = &rq->scx;
 
-	if (!(p->scx.flags & SCX_TASK_QUEUED))
+	if (!(p->scx.flags & SCX_TASK_QUEUED)) {
+		WARN_ON_ONCE(watchdog_task_watched(p));
 		return;
+	}
 
 	ops_dequeue(p, deq_flags);
 
@@ -1299,6 +1341,8 @@ static void set_next_task_scx(struct rq *rq, struct task_struct *p, bool first)
 	}
 
 	p->se.exec_start = rq_clock_task(rq);
+
+	watchdog_unwatch_task(p, true);
 }
 
 static void put_prev_task_scx(struct rq *rq, struct task_struct *p)
@@ -1342,11 +1386,14 @@ static void put_prev_task_scx(struct rq *rq, struct task_struct *p)
 	 */
 	if (p->scx.flags & SCX_TASK_BAL_KEEP) {
 		p->scx.flags &= ~SCX_TASK_BAL_KEEP;
+		watchdog_watch_task(rq, p);
 		dispatch_enqueue(&rq->scx.local_dsq, p, SCX_ENQ_HEAD);
 		return;
 	}
 
 	if (p->scx.flags & SCX_TASK_QUEUED) {
+		watchdog_watch_task(rq, p);
+
 		/*
 		 * If @p has slice left and balance_scx() didn't tag it for
 		 * keeping, @p is getting preempted by a higher priority
@@ -1604,6 +1651,49 @@ static void reset_idle_masks(void) {}
 
 #endif /* CONFIG_SMP */
 
+static bool check_rq_for_timeouts(struct rq *rq)
+{
+	struct task_struct *p;
+	struct rq_flags rf;
+	bool timed_out = false;
+
+	rq_lock_irqsave(rq, &rf);
+	list_for_each_entry(p, &rq->scx.watchdog_list, scx.watchdog_node) {
+		unsigned long last_runnable = p->scx.runnable_at;
+
+		if (unlikely(time_after(jiffies,
+					last_runnable + scx_watchdog_timeout))) {
+			u32 dur_ms = jiffies_to_msecs(jiffies - last_runnable);
+
+			scx_ops_error_type(SCX_EXIT_ERROR_STALL,
+					   "%s[%d] failed to run for %u.%03us",
+					   p->comm, p->pid,
+					   dur_ms / 1000, dur_ms % 1000);
+			timed_out = true;
+			break;
+		}
+	}
+	rq_unlock_irqrestore(rq, &rf);
+
+	return timed_out;
+}
+
+static void scx_watchdog_workfn(struct work_struct *work)
+{
+	int cpu;
+
+	scx_watchdog_timestamp = jiffies;
+
+	for_each_online_cpu(cpu) {
+		if (unlikely(check_rq_for_timeouts(cpu_rq(cpu))))
+			break;
+
+		cond_resched();
+	}
+	queue_delayed_work(system_unbound_wq, to_delayed_work(work),
+			   scx_watchdog_timeout / 2);
+}
+
 static void task_tick_scx(struct rq *rq, struct task_struct *curr, int queued)
 {
 	update_curr_scx(rq);
@@ -1635,7 +1725,7 @@ static int scx_ops_prepare_task(struct task_struct *p, struct task_group *tg)
 		}
 	}
 
-	p->scx.flags |= SCX_TASK_OPS_PREPPED;
+	p->scx.flags |= (SCX_TASK_OPS_PREPPED | SCX_TASK_WATCHDOG_RESET);
 	return 0;
 }
 
@@ -1967,6 +2057,8 @@ static void scx_ops_disable_workfn(struct kthread_work *work)
 			break;
 	}
 
+	cancel_delayed_work_sync(&scx_watchdog_work);
+
 	switch (type) {
 	case SCX_EXIT_UNREG:
 		reason = "BPF scheduler unregistered";
@@ -1980,6 +2072,9 @@ static void scx_ops_disable_workfn(struct kthread_work *work)
 	case SCX_EXIT_ERROR_BPF:
 		reason = "scx_bpf_error";
 		break;
+	case SCX_EXIT_ERROR_STALL:
+		reason = "runnable task stall";
+		break;
 	default:
 		reason = "<UNKNOWN>";
 	}
@@ -2164,8 +2259,8 @@ static void scx_ops_error_irq_workfn(struct irq_work *irq_work)
 
 static DEFINE_IRQ_WORK(scx_ops_error_irq_work, scx_ops_error_irq_workfn);
 
-__printf(2, 3) static void scx_ops_error_type(enum scx_exit_type type,
-					      const char *fmt, ...)
+__printf(2, 3) void scx_ops_error_type(enum scx_exit_type type,
+				       const char *fmt, ...)
 {
 	struct scx_exit_info *ei = &scx_exit_info;
 	int none = SCX_EXIT_NONE;
@@ -2264,6 +2359,14 @@ static int scx_ops_enable(struct sched_ext_ops *ops)
 		goto err_disable;
 	}
 
+	scx_watchdog_timeout = SCX_WATCHDOG_MAX_TIMEOUT;
+	if (ops->timeout_ms)
+		scx_watchdog_timeout = msecs_to_jiffies(ops->timeout_ms);
+
+	scx_watchdog_timestamp = jiffies;
+	queue_delayed_work(system_unbound_wq, &scx_watchdog_work,
+			   scx_watchdog_timeout / 2);
+
 	/*
 	 * Lock out forks before opening the floodgate so that they don't wander
 	 * into the operations prematurely.
@@ -2517,6 +2620,11 @@ static int bpf_scx_init_member(const struct btf_type *t,
 		if (ret == 0)
 			return -EINVAL;
 		return 1;
+	case offsetof(struct sched_ext_ops, timeout_ms):
+		if (*(u32 *)(udata + moff) > SCX_WATCHDOG_MAX_TIMEOUT)
+			return -E2BIG;
+		ops->timeout_ms = *(u32 *)(udata + moff);
+		return 1;
 	}
 
 	return 0;
@@ -2633,9 +2741,11 @@ void __init init_sched_ext_class(void)
 		struct rq *rq = cpu_rq(cpu);
 
 		init_dsq(&rq->scx.local_dsq, SCX_DSQ_LOCAL);
+		INIT_LIST_HEAD(&rq->scx.watchdog_list);
 	}
 
 	register_sysrq_key('S', &sysrq_sched_ext_reset_op);
+	INIT_DELAYED_WORK(&scx_watchdog_work, scx_watchdog_workfn);
 }
 
 
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index d78d151fdbf8..e6e7869c6f8d 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -71,6 +71,8 @@ void sched_enq_and_set_task(struct sched_enq_and_set_ctx *ctx);
 extern const struct sched_class ext_sched_class;
 extern const struct bpf_verifier_ops bpf_sched_ext_verifier_ops;
 extern const struct file_operations sched_ext_fops;
+extern unsigned long scx_watchdog_timeout;
+extern unsigned long scx_watchdog_timestamp;
 
 DECLARE_STATIC_KEY_FALSE(__scx_ops_enabled);
 #define scx_enabled()		static_branch_unlikely(&__scx_ops_enabled)
@@ -87,6 +89,28 @@ void scx_post_fork(struct task_struct *p);
 void scx_cancel_fork(struct task_struct *p);
 void init_sched_ext_class(void);
 
+__printf(2, 3) void scx_ops_error_type(enum scx_exit_type type,
+				       const char *fmt, ...);
+#define scx_ops_error(fmt, args...)						\
+	scx_ops_error_type(SCX_EXIT_ERROR, fmt, ##args)
+
+static inline void scx_notify_sched_tick(void)
+{
+	unsigned long last_check;
+
+	if (!scx_enabled())
+		return;
+
+	last_check = scx_watchdog_timestamp;
+	if (unlikely(time_after(jiffies, last_check + scx_watchdog_timeout))) {
+		u32 dur_ms = jiffies_to_msecs(jiffies - last_check);
+
+		scx_ops_error_type(SCX_EXIT_ERROR_STALL,
+				   "watchdog failed to check in for %u.%03us",
+				   dur_ms / 1000, dur_ms % 1000);
+	}
+}
+
 static inline const struct sched_class *next_active_class(const struct sched_class *class)
 {
 	class++;
@@ -119,6 +143,7 @@ static inline int scx_fork(struct task_struct *p) { return 0; }
 static inline void scx_post_fork(struct task_struct *p) {}
 static inline void scx_cancel_fork(struct task_struct *p) {}
 static inline void init_sched_ext_class(void) {}
+static inline void scx_notify_sched_tick(void) {}
 
 #define for_each_active_class		for_each_class
 #define for_balance_class_range		for_class_range
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 666166908eb6..c77fefae1694 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -688,6 +688,7 @@ struct cfs_rq {
 #ifdef CONFIG_SCHED_CLASS_EXT
 struct scx_rq {
 	struct scx_dispatch_q	local_dsq;
+	struct list_head	watchdog_list;
 	u64			ops_qseq;
 	u64			extra_enq_flags;	/* see move_task_to_local_dsq() */
 	u32			nr_running;
diff --git a/tools/sched_ext/scx_qmap.bpf.c b/tools/sched_ext/scx_qmap.bpf.c
index 686681e2008a..a1bdb5774a63 100644
--- a/tools/sched_ext/scx_qmap.bpf.c
+++ b/tools/sched_ext/scx_qmap.bpf.c
@@ -25,6 +25,8 @@
 char _license[] SEC("license") = "GPL";
 
 const volatile u64 slice_ns = SCX_SLICE_DFL;
+const volatile u32 stall_user_nth;
+const volatile u32 stall_kernel_nth;
 
 u32 test_error_cnt;
 
@@ -120,11 +122,20 @@ static int weight_to_idx(u32 weight)
 
 void BPF_STRUCT_OPS(qmap_enqueue, struct task_struct *p, u64 enq_flags)
 {
+	static u32 user_cnt, kernel_cnt;
 	struct task_ctx *tctx;
 	u32 pid = p->pid;
 	int idx = weight_to_idx(p->scx.weight);
 	void *ring;
 
+	if (p->flags & PF_KTHREAD) {
+		if (stall_kernel_nth && !(++kernel_cnt % stall_kernel_nth))
+			return;
+	} else {
+		if (stall_user_nth && !(++user_cnt % stall_user_nth))
+			return;
+	}
+
 	if (test_error_cnt && !--test_error_cnt)
 		scx_bpf_error("test triggering error");
 
@@ -237,5 +248,6 @@ struct sched_ext_ops qmap_ops = {
 	.dispatch		= (void *)qmap_dispatch,
 	.prep_enable		= (void *)qmap_prep_enable,
 	.exit			= (void *)qmap_exit,
+	.timeout_ms		= 5000U,
 	.name			= "qmap",
 };
diff --git a/tools/sched_ext/scx_qmap.c b/tools/sched_ext/scx_qmap.c
index 7e96d8e96f73..93add1156def 100644
--- a/tools/sched_ext/scx_qmap.c
+++ b/tools/sched_ext/scx_qmap.c
@@ -20,10 +20,12 @@ const char help_fmt[] =
 "\n"
 "See the top-level comment in .bpf.c for more details.\n"
 "\n"
-"Usage: %s [-s SLICE_US] [-e COUNT]\n"
+"Usage: %s [-s SLICE_US] [-e COUNT] [-t COUNT] [-T COUNT]\n"
 "\n"
 "  -s SLICE_US   Override slice duration\n"
 "  -e COUNT      Trigger scx_bpf_error() after COUNT enqueues\n"
+"  -t COUNT      Stall every COUNT'th user thread\n"
+"  -T COUNT      Stall every COUNT'th kernel thread\n"
 "  -h            Display this help and exit\n";
 
 static volatile int exit_req;
@@ -47,7 +49,7 @@ int main(int argc, char **argv)
 	skel = scx_qmap__open();
 	assert(skel);
 
-	while ((opt = getopt(argc, argv, "s:e:tTd:h")) != -1) {
+	while ((opt = getopt(argc, argv, "s:e:t:T:d:h")) != -1) {
 		switch (opt) {
 		case 's':
 			skel->rodata->slice_ns = strtoull(optarg, NULL, 0) * 1000;
@@ -55,6 +57,12 @@ int main(int argc, char **argv)
 		case 'e':
 			skel->bss->test_error_cnt = strtoul(optarg, NULL, 0);
 			break;
+		case 't':
+			skel->rodata->stall_user_nth = strtoul(optarg, NULL, 0);
+			break;
+		case 'T':
+			skel->rodata->stall_kernel_nth = strtoul(optarg, NULL, 0);
+			break;
 		default:
 			fprintf(stderr, help_fmt, basename(argv[0]));
 			return opt != 'h';
-- 
2.41.0

[PATCH 16/34] sched_ext: Allow BPF schedulers to disallow specific tasks from joining SCHED_EXT

[Tejun Heo]

BPF schedulers might not want to schedule certain tasks - e.g. kernel
threads. This patch adds p->scx.disallow which can be set by BPF schedulers
in such cases. The field can be changed anytime and setting it in
ops.prep_enable() guarantees that the task can never be scheduled by
sched_ext.

scx_qmap is updated with the -d option to disallow a specific PID:

  # echo $$
  1092
  # egrep '(policy)|(ext\.enabled)' /proc/self/sched
  policy                                       :                    0
  ext.enabled                                  :                    0
  # ./set-scx 1092
  # egrep '(policy)|(ext\.enabled)' /proc/self/sched
  policy                                       :                    7
  ext.enabled                                  :                    0

Run "scx_qmap -d 1092" in another terminal.

  # grep rejected /sys/kernel/debug/sched/ext
  nr_rejected                   : 1
  # egrep '(policy)|(ext\.enabled)' /proc/self/sched
  policy                                       :                    0
  ext.enabled                                  :                    0
  # ./set-scx 1092
  setparam failed for 1092 (Permission denied)

Signed-off-by: Tejun Heo <tj@kernel.org>
Suggested-by: Barret Rhoden <brho@google.com>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 include/linux/sched/ext.h      | 12 +++++++++++
 kernel/sched/core.c            |  4 ++++
 kernel/sched/ext.c             | 38 ++++++++++++++++++++++++++++++++++
 kernel/sched/ext.h             |  3 +++
 tools/sched_ext/scx_qmap.bpf.c |  4 ++++
 tools/sched_ext/scx_qmap.c     |  8 ++++++-
 6 files changed, 68 insertions(+), 1 deletion(-)

diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 1588e4496afa..772d84033155 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -410,6 +410,18 @@ struct sched_ext_entity {
 	 */
 	u64			slice;
 
+	/*
+	 * If set, reject future sched_setscheduler(2) calls updating the policy
+	 * to %SCHED_EXT with -%EACCES.
+	 *
+	 * If set from ops.prep_enable() and the task's policy is already
+	 * %SCHED_EXT, which can happen while the BPF scheduler is being loaded
+	 * or by inhering the parent's policy during fork, the task's policy is
+	 * rejected and forcefully reverted to %SCHED_NORMAL. The number of such
+	 * events are reported through /sys/kernel/debug/sched_ext::nr_rejected.
+	 */
+	bool			disallow;	/* reject switching into SCX */
+
 	/* cold fields */
 	struct list_head	tasks_node;
 };
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 9d97c0e6442a..c8528cbfeb57 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -7725,6 +7725,10 @@ static int __sched_setscheduler(struct task_struct *p,
 		goto unlock;
 	}
 
+	retval = scx_check_setscheduler(p, policy);
+	if (retval)
+		goto unlock;
+
 	/*
 	 * If not changing anything there's no need to proceed further,
 	 * but store a possible modification of reset_on_fork.
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index eca1790375b8..d5ef8809e05f 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -1715,6 +1715,8 @@ static int scx_ops_prepare_task(struct task_struct *p, struct task_group *tg)
 
 	WARN_ON_ONCE(p->scx.flags & SCX_TASK_OPS_PREPPED);
 
+	p->scx.disallow = false;
+
 	if (SCX_HAS_OP(prep_enable)) {
 		struct scx_enable_args args = { };
 
@@ -1725,6 +1727,27 @@ static int scx_ops_prepare_task(struct task_struct *p, struct task_group *tg)
 		}
 	}
 
+	if (p->scx.disallow) {
+		struct rq *rq;
+		struct rq_flags rf;
+
+		rq = task_rq_lock(p, &rf);
+
+		/*
+		 * We're either in fork or load path and @p->policy will be
+		 * applied right after. Reverting @p->policy here and rejecting
+		 * %SCHED_EXT transitions from scx_check_setscheduler()
+		 * guarantees that if ops.prep_enable() sets @p->disallow, @p
+		 * can never be in SCX.
+		 */
+		if (p->policy == SCHED_EXT) {
+			p->policy = SCHED_NORMAL;
+			atomic64_inc(&scx_nr_rejected);
+		}
+
+		task_rq_unlock(rq, p, &rf);
+	}
+
 	p->scx.flags |= (SCX_TASK_OPS_PREPPED | SCX_TASK_WATCHDOG_RESET);
 	return 0;
 }
@@ -1886,6 +1909,18 @@ static void switching_to_scx(struct rq *rq, struct task_struct *p)
 static void check_preempt_curr_scx(struct rq *rq, struct task_struct *p,int wake_flags) {}
 static void switched_to_scx(struct rq *rq, struct task_struct *p) {}
 
+int scx_check_setscheduler(struct task_struct *p, int policy)
+{
+	lockdep_assert_rq_held(task_rq(p));
+
+	/* if disallow, reject transitioning into SCX */
+	if (scx_enabled() && READ_ONCE(p->scx.disallow) &&
+	    p->policy != policy && policy == SCHED_EXT)
+		return -EACCES;
+
+	return 0;
+}
+
 /*
  * Omitted operations:
  *
@@ -2568,6 +2603,9 @@ static int bpf_scx_btf_struct_access(struct bpf_verifier_log *log,
 		if (off >= offsetof(struct task_struct, scx.slice) &&
 		    off + size <= offsetofend(struct task_struct, scx.slice))
 			return SCALAR_VALUE;
+		if (off >= offsetof(struct task_struct, scx.disallow) &&
+		    off + size <= offsetofend(struct task_struct, scx.disallow))
+			return SCALAR_VALUE;
 	}
 
 	return 0;
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index e6e7869c6f8d..444a917d27b1 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -87,6 +87,7 @@ void scx_pre_fork(struct task_struct *p);
 int scx_fork(struct task_struct *p);
 void scx_post_fork(struct task_struct *p);
 void scx_cancel_fork(struct task_struct *p);
+int scx_check_setscheduler(struct task_struct *p, int policy);
 void init_sched_ext_class(void);
 
 __printf(2, 3) void scx_ops_error_type(enum scx_exit_type type,
@@ -142,6 +143,8 @@ static inline void scx_pre_fork(struct task_struct *p) {}
 static inline int scx_fork(struct task_struct *p) { return 0; }
 static inline void scx_post_fork(struct task_struct *p) {}
 static inline void scx_cancel_fork(struct task_struct *p) {}
+static inline int scx_check_setscheduler(struct task_struct *p,
+					 int policy) { return 0; }
 static inline void init_sched_ext_class(void) {}
 static inline void scx_notify_sched_tick(void) {}
 
diff --git a/tools/sched_ext/scx_qmap.bpf.c b/tools/sched_ext/scx_qmap.bpf.c
index a1bdb5774a63..d0bc67095062 100644
--- a/tools/sched_ext/scx_qmap.bpf.c
+++ b/tools/sched_ext/scx_qmap.bpf.c
@@ -27,6 +27,7 @@ char _license[] SEC("license") = "GPL";
 const volatile u64 slice_ns = SCX_SLICE_DFL;
 const volatile u32 stall_user_nth;
 const volatile u32 stall_kernel_nth;
+const volatile s32 disallow_tgid;
 
 u32 test_error_cnt;
 
@@ -224,6 +225,9 @@ void BPF_STRUCT_OPS(qmap_dispatch, s32 cpu, struct task_struct *prev)
 s32 BPF_STRUCT_OPS(qmap_prep_enable, struct task_struct *p,
 		   struct scx_enable_args *args)
 {
+	if (p->tgid == disallow_tgid)
+		p->scx.disallow = true;
+
 	/*
 	 * @p is new. Let's ensure that its task_ctx is available. We can sleep
 	 * in this function and the following will automatically use GFP_KERNEL.
diff --git a/tools/sched_ext/scx_qmap.c b/tools/sched_ext/scx_qmap.c
index 93add1156def..5f50f889ea18 100644
--- a/tools/sched_ext/scx_qmap.c
+++ b/tools/sched_ext/scx_qmap.c
@@ -20,12 +20,13 @@ const char help_fmt[] =
 "\n"
 "See the top-level comment in .bpf.c for more details.\n"
 "\n"
-"Usage: %s [-s SLICE_US] [-e COUNT] [-t COUNT] [-T COUNT]\n"
+"Usage: %s [-s SLICE_US] [-e COUNT] [-t COUNT] [-T COUNT] [-d PID]\n"
 "\n"
 "  -s SLICE_US   Override slice duration\n"
 "  -e COUNT      Trigger scx_bpf_error() after COUNT enqueues\n"
 "  -t COUNT      Stall every COUNT'th user thread\n"
 "  -T COUNT      Stall every COUNT'th kernel thread\n"
+"  -d PID        Disallow a process from switching into SCHED_EXT (-1 for self)\n"
 "  -h            Display this help and exit\n";
 
 static volatile int exit_req;
@@ -63,6 +64,11 @@ int main(int argc, char **argv)
 		case 'T':
 			skel->rodata->stall_kernel_nth = strtoul(optarg, NULL, 0);
 			break;
+		case 'd':
+			skel->rodata->disallow_tgid = strtol(optarg, NULL, 0);
+			if (skel->rodata->disallow_tgid < 0)
+				skel->rodata->disallow_tgid = getpid();
+			break;
 		default:
 			fprintf(stderr, help_fmt, basename(argv[0]));
 			return opt != 'h';
-- 
2.41.0

[PATCH 17/34] sched_ext: Allow BPF schedulers to switch all eligible tasks into sched_ext

[Tejun Heo]

Currently, to use sched_ext, each task has to be put into sched_ext using
sched_setscheduler(2). However, some BPF schedulers and use cases might
prefer to service all eligible tasks.

This patch adds a new kfunc helper, scx_bpf_switch_all(), that BPF
schedulers can call from ops.init() to switch all SCHED_NORMAL, SCHED_BATCH
and SCHED_IDLE tasks into sched_ext. This has the benefit that the scheduler
swaps are transparent to the users and applications. As we know that CFS is
not being used when scx_bpf_switch_all() is used, we can also disable hot
path entry points with static_branches.

Both the simple and qmap example schedulers are updated to switch all tasks
by default to ease testing. '-p' option is added which enables the original
behavior of switching only tasks which are explicitly on SCHED_EXT.

v2: In the example schedulers, switch all tasks by default.

Signed-off-by: Tejun Heo <tj@kernel.org>
Suggested-by: Barret Rhoden <brho@google.com>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 kernel/sched/core.c              |  8 +++---
 kernel/sched/ext.c               | 43 ++++++++++++++++++++++++++++++++
 kernel/sched/ext.h               |  5 ++++
 tools/sched_ext/scx_common.bpf.h |  1 +
 tools/sched_ext/scx_qmap.bpf.c   |  9 +++++++
 tools/sched_ext/scx_qmap.c       |  8 ++++--
 tools/sched_ext/scx_simple.bpf.c | 10 ++++++++
 tools/sched_ext/scx_simple.c     |  8 ++++--
 8 files changed, 85 insertions(+), 7 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index c8528cbfeb57..c976a36dd642 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -1226,7 +1226,7 @@ bool sched_can_stop_tick(struct rq *rq)
 	 * if there's more than one we need the tick for involuntary
 	 * preemption.
 	 */
-	if (rq->nr_running > 1)
+	if (!scx_switched_all() && rq->nr_running > 1)
 		return false;
 
 	return true;
@@ -5690,8 +5690,10 @@ void scheduler_tick(void)
 	perf_event_task_tick();
 
 #ifdef CONFIG_SMP
-	rq->idle_balance = idle_cpu(cpu);
-	trigger_load_balance(rq);
+	if (!scx_switched_all()) {
+		rq->idle_balance = idle_cpu(cpu);
+		trigger_load_balance(rq);
+	}
 #endif
 }
 
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index d5ef8809e05f..6cb3412cee9f 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -73,6 +73,10 @@ static DEFINE_MUTEX(scx_ops_enable_mutex);
 DEFINE_STATIC_KEY_FALSE(__scx_ops_enabled);
 DEFINE_STATIC_PERCPU_RWSEM(scx_fork_rwsem);
 static atomic_t scx_ops_enable_state_var = ATOMIC_INIT(SCX_OPS_DISABLED);
+static bool scx_switch_all_req;
+static bool scx_switching_all;
+DEFINE_STATIC_KEY_FALSE(__scx_switched_all);
+
 static struct sched_ext_ops scx_ops;
 static bool scx_warned_zero_slice;
 
@@ -2056,6 +2060,8 @@ bool task_should_scx(struct task_struct *p)
 {
 	if (!scx_enabled() || scx_ops_disabling())
 		return false;
+	if (READ_ONCE(scx_switching_all))
+		return true;
 	return p->policy == SCHED_EXT;
 }
 
@@ -2183,6 +2189,9 @@ static void scx_ops_disable_workfn(struct kthread_work *work)
 	 */
 	mutex_lock(&scx_ops_enable_mutex);
 
+	static_branch_disable(&__scx_switched_all);
+	WRITE_ONCE(scx_switching_all, false);
+
 	/* avoid racing against fork */
 	cpus_read_lock();
 	percpu_down_write(&scx_fork_rwsem);
@@ -2366,6 +2375,7 @@ static int scx_ops_enable(struct sched_ext_ops *ops)
 	 */
 	cpus_read_lock();
 
+	scx_switch_all_req = false;
 	if (scx_ops.init) {
 		ret = SCX_CALL_OP_RET(SCX_KF_INIT, init);
 		if (ret) {
@@ -2481,6 +2491,8 @@ static int scx_ops_enable(struct sched_ext_ops *ops)
 	 * transitions here are synchronized against sched_ext_free() through
 	 * scx_tasks_lock.
 	 */
+	WRITE_ONCE(scx_switching_all, scx_switch_all_req);
+
 	scx_task_iter_init(&sti);
 	while ((p = scx_task_iter_next_filtered_locked(&sti))) {
 		if (READ_ONCE(p->__state) != TASK_DEAD) {
@@ -2512,6 +2524,9 @@ static int scx_ops_enable(struct sched_ext_ops *ops)
 		goto err_disable;
 	}
 
+	if (scx_switch_all_req)
+		static_branch_enable_cpuslocked(&__scx_switched_all);
+
 	cpus_read_unlock();
 	mutex_unlock(&scx_ops_enable_mutex);
 
@@ -2546,6 +2561,9 @@ static int scx_debug_show(struct seq_file *m, void *v)
 	mutex_lock(&scx_ops_enable_mutex);
 	seq_printf(m, "%-30s: %s\n", "ops", scx_ops.name);
 	seq_printf(m, "%-30s: %ld\n", "enabled", scx_enabled());
+	seq_printf(m, "%-30s: %d\n", "switching_all",
+		   READ_ONCE(scx_switching_all));
+	seq_printf(m, "%-30s: %ld\n", "switched_all", scx_switched_all());
 	seq_printf(m, "%-30s: %s\n", "enable_state",
 		   scx_ops_enable_state_str[scx_ops_enable_state()]);
 	seq_printf(m, "%-30s: %llu\n", "nr_rejected",
@@ -2797,6 +2815,29 @@ __diag_push();
 __diag_ignore_all("-Wmissing-prototypes",
 		  "Global functions as their definitions will be in vmlinux BTF");
 
+/**
+ * scx_bpf_switch_all - Switch all tasks into SCX
+ *
+ * Switch all existing and future non-dl/rt tasks to SCX. This can only be
+ * called from ops.init(), and actual switching is performed asynchronously.
+ */
+void scx_bpf_switch_all(void)
+{
+	if (!scx_kf_allowed(SCX_KF_INIT))
+		return;
+
+	scx_switch_all_req = true;
+}
+
+BTF_SET8_START(scx_kfunc_ids_init)
+BTF_ID_FLAGS(func, scx_bpf_switch_all)
+BTF_SET8_END(scx_kfunc_ids_init)
+
+static const struct btf_kfunc_id_set scx_kfunc_set_init = {
+	.owner			= THIS_MODULE,
+	.set			= &scx_kfunc_ids_init,
+};
+
 /**
  * scx_bpf_create_dsq - Create a custom DSQ
  * @dsq_id: DSQ to create
@@ -3292,6 +3333,8 @@ static int __init register_ext_kfuncs(void)
 	 * check using scx_kf_allowed().
 	 */
 	if ((ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
+					     &scx_kfunc_set_init)) ||
+	    (ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
 					     &scx_kfunc_set_sleepable)) ||
 	    (ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
 					     &scx_kfunc_set_enqueue_dispatch)) ||
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index 444a917d27b1..c32f14bb981c 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -75,7 +75,9 @@ extern unsigned long scx_watchdog_timeout;
 extern unsigned long scx_watchdog_timestamp;
 
 DECLARE_STATIC_KEY_FALSE(__scx_ops_enabled);
+DECLARE_STATIC_KEY_FALSE(__scx_switched_all);
 #define scx_enabled()		static_branch_unlikely(&__scx_ops_enabled)
+#define scx_switched_all()	static_branch_unlikely(&__scx_switched_all)
 
 static inline bool task_on_scx(struct task_struct *p)
 {
@@ -115,6 +117,8 @@ static inline void scx_notify_sched_tick(void)
 static inline const struct sched_class *next_active_class(const struct sched_class *class)
 {
 	class++;
+	if (scx_switched_all() && class == &fair_sched_class)
+		class++;
 	if (!scx_enabled() && class == &ext_sched_class)
 		class++;
 	return class;
@@ -137,6 +141,7 @@ static inline const struct sched_class *next_active_class(const struct sched_cla
 #else	/* CONFIG_SCHED_CLASS_EXT */
 
 #define scx_enabled()		false
+#define scx_switched_all()	false
 
 static inline bool task_on_scx(struct task_struct *p) { return false; }
 static inline void scx_pre_fork(struct task_struct *p) {}
diff --git a/tools/sched_ext/scx_common.bpf.h b/tools/sched_ext/scx_common.bpf.h
index 8b8502afcce3..0e6d7d3e2d27 100644
--- a/tools/sched_ext/scx_common.bpf.h
+++ b/tools/sched_ext/scx_common.bpf.h
@@ -53,6 +53,7 @@ void ___scx_bpf_error_format_checker(const char *fmt, ...) {}
 	___scx_bpf_error_format_checker(fmt, ##args);				\
 })
 
+void scx_bpf_switch_all(void) __ksym;
 s32 scx_bpf_create_dsq(u64 dsq_id, s32 node) __ksym;
 bool scx_bpf_consume(u64 dsq_id) __ksym;
 u32 scx_bpf_dispatch_nr_slots(void) __ksym;
diff --git a/tools/sched_ext/scx_qmap.bpf.c b/tools/sched_ext/scx_qmap.bpf.c
index d0bc67095062..da43f962ab4e 100644
--- a/tools/sched_ext/scx_qmap.bpf.c
+++ b/tools/sched_ext/scx_qmap.bpf.c
@@ -25,6 +25,7 @@
 char _license[] SEC("license") = "GPL";
 
 const volatile u64 slice_ns = SCX_SLICE_DFL;
+const volatile bool switch_partial;
 const volatile u32 stall_user_nth;
 const volatile u32 stall_kernel_nth;
 const volatile s32 disallow_tgid;
@@ -239,6 +240,13 @@ s32 BPF_STRUCT_OPS(qmap_prep_enable, struct task_struct *p,
 		return -ENOMEM;
 }
 
+s32 BPF_STRUCT_OPS(qmap_init)
+{
+	if (!switch_partial)
+		scx_bpf_switch_all();
+	return 0;
+}
+
 void BPF_STRUCT_OPS(qmap_exit, struct scx_exit_info *ei)
 {
 	uei_record(&uei, ei);
@@ -251,6 +259,7 @@ struct sched_ext_ops qmap_ops = {
 	.dequeue		= (void *)qmap_dequeue,
 	.dispatch		= (void *)qmap_dispatch,
 	.prep_enable		= (void *)qmap_prep_enable,
+	.init			= (void *)qmap_init,
 	.exit			= (void *)qmap_exit,
 	.timeout_ms		= 5000U,
 	.name			= "qmap",
diff --git a/tools/sched_ext/scx_qmap.c b/tools/sched_ext/scx_qmap.c
index 5f50f889ea18..3444e3597b19 100644
--- a/tools/sched_ext/scx_qmap.c
+++ b/tools/sched_ext/scx_qmap.c
@@ -20,13 +20,14 @@ const char help_fmt[] =
 "\n"
 "See the top-level comment in .bpf.c for more details.\n"
 "\n"
-"Usage: %s [-s SLICE_US] [-e COUNT] [-t COUNT] [-T COUNT] [-d PID]\n"
+"Usage: %s [-s SLICE_US] [-e COUNT] [-t COUNT] [-T COUNT] [-d PID] [-p]\n"
 "\n"
 "  -s SLICE_US   Override slice duration\n"
 "  -e COUNT      Trigger scx_bpf_error() after COUNT enqueues\n"
 "  -t COUNT      Stall every COUNT'th user thread\n"
 "  -T COUNT      Stall every COUNT'th kernel thread\n"
 "  -d PID        Disallow a process from switching into SCHED_EXT (-1 for self)\n"
+"  -p            Switch only tasks on SCHED_EXT policy intead of all\n"
 "  -h            Display this help and exit\n";
 
 static volatile int exit_req;
@@ -50,7 +51,7 @@ int main(int argc, char **argv)
 	skel = scx_qmap__open();
 	assert(skel);
 
-	while ((opt = getopt(argc, argv, "s:e:t:T:d:h")) != -1) {
+	while ((opt = getopt(argc, argv, "s:e:t:T:d:ph")) != -1) {
 		switch (opt) {
 		case 's':
 			skel->rodata->slice_ns = strtoull(optarg, NULL, 0) * 1000;
@@ -69,6 +70,9 @@ int main(int argc, char **argv)
 			if (skel->rodata->disallow_tgid < 0)
 				skel->rodata->disallow_tgid = getpid();
 			break;
+		case 'p':
+			skel->rodata->switch_partial = true;
+			break;
 		default:
 			fprintf(stderr, help_fmt, basename(argv[0]));
 			return opt != 'h';
diff --git a/tools/sched_ext/scx_simple.bpf.c b/tools/sched_ext/scx_simple.bpf.c
index 9326124a32fa..6302a4ea9ea5 100644
--- a/tools/sched_ext/scx_simple.bpf.c
+++ b/tools/sched_ext/scx_simple.bpf.c
@@ -15,6 +15,8 @@
 
 char _license[] SEC("license") = "GPL";
 
+const volatile bool switch_partial;
+
 struct user_exit_info uei;
 
 struct {
@@ -43,6 +45,13 @@ void BPF_STRUCT_OPS(simple_enqueue, struct task_struct *p, u64 enq_flags)
 	scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, SCX_SLICE_DFL, enq_flags);
 }
 
+s32 BPF_STRUCT_OPS(simple_init)
+{
+	if (!switch_partial)
+		scx_bpf_switch_all();
+	return 0;
+}
+
 void BPF_STRUCT_OPS(simple_exit, struct scx_exit_info *ei)
 {
 	uei_record(&uei, ei);
@@ -51,6 +60,7 @@ void BPF_STRUCT_OPS(simple_exit, struct scx_exit_info *ei)
 SEC(".struct_ops.link")
 struct sched_ext_ops simple_ops = {
 	.enqueue		= (void *)simple_enqueue,
+	.init			= (void *)simple_init,
 	.exit			= (void *)simple_exit,
 	.name			= "simple",
 };
diff --git a/tools/sched_ext/scx_simple.c b/tools/sched_ext/scx_simple.c
index 9ba38ba1e71f..1e507c0a35cd 100644
--- a/tools/sched_ext/scx_simple.c
+++ b/tools/sched_ext/scx_simple.c
@@ -19,8 +19,9 @@ const char help_fmt[] =
 "\n"
 "See the top-level comment in .bpf.c for more details.\n"
 "\n"
-"Usage: %s\n"
+"Usage: %s [-p]\n"
 "\n"
+"  -p            Switch only tasks on SCHED_EXT policy intead of all\n"
 "  -h            Display this help and exit\n";
 
 static volatile int exit_req;
@@ -64,8 +65,11 @@ int main(int argc, char **argv)
 	skel = scx_simple__open();
 	assert(skel);
 
-	while ((opt = getopt(argc, argv, "h")) != -1) {
+	while ((opt = getopt(argc, argv, "ph")) != -1) {
 		switch (opt) {
+		case 'p':
+			skel->rodata->switch_partial = true;
+			break;
 		default:
 			fprintf(stderr, help_fmt, basename(argv[0]));
 			return opt != 'h';
-- 
2.41.0

[PATCH 18/34] sched_ext: Implement scx_bpf_kick_cpu() and task preemption support

[Tejun Heo]

It's often useful to wake up and/or trigger reschedule on other CPUs. This
patch adds scx_bpf_kick_cpu() kfunc helper that BPF scheduler can call to
kick the target CPU into the scheduling path.

As a sched_ext task relinquishes its CPU only after its slice is depleted,
this patch also adds SCX_KICK_PREEMPT and SCX_ENQ_PREEMPT which clears the
slice of the target CPU's current task to guarantee that sched_ext's
scheduling path runs on the CPU.

v4: * Move example scheduler to its own patch.

v3: * Make scx_example_central switch all tasks by default.

    * Convert to BPF inline iterators.

v2: * Julia Lawall reported that scx_example_central can overflow the
      dispatch buffer and malfunction. As scheduling for other CPUs can't be
      handled by the automatic retry mechanism, fix by implementing an
      explicit overflow and retry handling.

    * Updated to use generic BPF cpumask helpers.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 include/linux/sched/ext.h        |  4 ++
 kernel/sched/ext.c               | 82 ++++++++++++++++++++++++++++++--
 kernel/sched/ext.h               | 12 +++++
 kernel/sched/sched.h             |  3 ++
 tools/sched_ext/scx_common.bpf.h |  1 +
 5 files changed, 99 insertions(+), 3 deletions(-)

diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 772d84033155..5c2df7ccc0a6 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -407,6 +407,10 @@ struct sched_ext_entity {
 	 * scx_bpf_dispatch() but can also be modified directly by the BPF
 	 * scheduler. Automatically decreased by SCX as the task executes. On
 	 * depletion, a scheduling event is triggered.
+	 *
+	 * This value is cleared to zero if the task is preempted by
+	 * %SCX_KICK_PREEMPT and shouldn't be used to determine how long the
+	 * task ran. Use p->se.sum_exec_runtime instead.
 	 */
 	u64			slice;
 
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 6cb3412cee9f..9e8f9f9fcb3d 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -496,7 +496,7 @@ static void dispatch_enqueue(struct scx_dispatch_q *dsq, struct task_struct *p,
 		}
 	}
 
-	if (enq_flags & SCX_ENQ_HEAD)
+	if (enq_flags & (SCX_ENQ_HEAD | SCX_ENQ_PREEMPT))
 		list_add(&p->scx.dsq_node, &dsq->fifo);
 	else
 		list_add_tail(&p->scx.dsq_node, &dsq->fifo);
@@ -512,8 +512,16 @@ static void dispatch_enqueue(struct scx_dispatch_q *dsq, struct task_struct *p,
 
 	if (is_local) {
 		struct rq *rq = container_of(dsq, struct rq, scx.local_dsq);
+		bool preempt = false;
 
-		if (sched_class_above(&ext_sched_class, rq->curr->sched_class))
+		if ((enq_flags & SCX_ENQ_PREEMPT) && p != rq->curr &&
+		    rq->curr->sched_class == &ext_sched_class) {
+			rq->curr->scx.slice = 0;
+			preempt = true;
+		}
+
+		if (preempt || sched_class_above(&ext_sched_class,
+						 rq->curr->sched_class))
 			resched_curr(rq);
 	} else {
 		raw_spin_unlock(&dsq->lock);
@@ -1929,7 +1937,9 @@ int scx_check_setscheduler(struct task_struct *p, int policy)
  * Omitted operations:
  *
  * - check_preempt_curr: NOOP as it isn't useful in the wakeup path because the
- *   task isn't tied to the CPU at that point.
+ *   task isn't tied to the CPU at that point. Preemption is implemented by
+ *   resetting the victim task's slice to 0 and triggering reschedule on the
+ *   target CPU.
  *
  * - migrate_task_rq: Unncessary as task to cpu mapping is transient.
  *
@@ -2775,6 +2785,32 @@ static const struct sysrq_key_op sysrq_sched_ext_reset_op = {
 	.enable_mask	= SYSRQ_ENABLE_RTNICE,
 };
 
+static void kick_cpus_irq_workfn(struct irq_work *irq_work)
+{
+	struct rq *this_rq = this_rq();
+	int this_cpu = cpu_of(this_rq);
+	int cpu;
+
+	for_each_cpu(cpu, this_rq->scx.cpus_to_kick) {
+		struct rq *rq = cpu_rq(cpu);
+		unsigned long flags;
+
+		raw_spin_rq_lock_irqsave(rq, flags);
+
+		if (cpu_online(cpu) || cpu == this_cpu) {
+			if (cpumask_test_cpu(cpu, this_rq->scx.cpus_to_preempt) &&
+			    rq->curr->sched_class == &ext_sched_class)
+				rq->curr->scx.slice = 0;
+			resched_curr(rq);
+		}
+
+		raw_spin_rq_unlock_irqrestore(rq, flags);
+	}
+
+	cpumask_clear(this_rq->scx.cpus_to_kick);
+	cpumask_clear(this_rq->scx.cpus_to_preempt);
+}
+
 void __init init_sched_ext_class(void)
 {
 	int cpu;
@@ -2798,6 +2834,10 @@ void __init init_sched_ext_class(void)
 
 		init_dsq(&rq->scx.local_dsq, SCX_DSQ_LOCAL);
 		INIT_LIST_HEAD(&rq->scx.watchdog_list);
+
+		BUG_ON(!zalloc_cpumask_var(&rq->scx.cpus_to_kick, GFP_KERNEL));
+		BUG_ON(!zalloc_cpumask_var(&rq->scx.cpus_to_preempt, GFP_KERNEL));
+		init_irq_work(&rq->scx.kick_cpus_irq_work, kick_cpus_irq_workfn);
 	}
 
 	register_sysrq_key('S', &sysrq_sched_ext_reset_op);
@@ -3032,6 +3072,41 @@ static const struct btf_kfunc_id_set scx_kfunc_set_dispatch = {
 	.set			= &scx_kfunc_ids_dispatch,
 };
 
+/**
+ * scx_bpf_kick_cpu - Trigger reschedule on a CPU
+ * @cpu: cpu to kick
+ * @flags: %SCX_KICK_* flags
+ *
+ * Kick @cpu into rescheduling. This can be used to wake up an idle CPU or
+ * trigger rescheduling on a busy CPU. This can be called from any online
+ * scx_ops operation and the actual kicking is performed asynchronously through
+ * an irq work.
+ */
+void scx_bpf_kick_cpu(s32 cpu, u64 flags)
+{
+	struct rq *rq;
+
+	if (!ops_cpu_valid(cpu)) {
+		scx_ops_error("invalid cpu %d", cpu);
+		return;
+	}
+
+	preempt_disable();
+	rq = this_rq();
+
+	/*
+	 * Actual kicking is bounced to kick_cpus_irq_workfn() to avoid nesting
+	 * rq locks. We can probably be smarter and avoid bouncing if called
+	 * from ops which don't hold a rq lock.
+	 */
+	cpumask_set_cpu(cpu, rq->scx.cpus_to_kick);
+	if (flags & SCX_KICK_PREEMPT)
+		cpumask_set_cpu(cpu, rq->scx.cpus_to_preempt);
+
+	irq_work_queue(&rq->scx.kick_cpus_irq_work);
+	preempt_enable();
+}
+
 /**
  * scx_bpf_dsq_nr_queued - Return the number of queued tasks
  * @dsq_id: id of the DSQ
@@ -3296,6 +3371,7 @@ s32 scx_bpf_task_cpu(const struct task_struct *p)
 }
 
 BTF_SET8_START(scx_kfunc_ids_any)
+BTF_ID_FLAGS(func, scx_bpf_kick_cpu)
 BTF_ID_FLAGS(func, scx_bpf_dsq_nr_queued)
 BTF_ID_FLAGS(func, scx_bpf_test_and_clear_cpu_idle)
 BTF_ID_FLAGS(func, scx_bpf_pick_idle_cpu, KF_RCU)
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index c32f14bb981c..f535ec39e660 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -19,6 +19,14 @@ enum scx_enq_flags {
 
 	/* high 32bits are SCX specific */
 
+	/*
+	 * Set the following to trigger preemption when calling
+	 * scx_bpf_dispatch() with a local dsq as the target. The slice of the
+	 * current task is cleared to zero and the CPU is kicked into the
+	 * scheduling path. Implies %SCX_ENQ_HEAD.
+	 */
+	SCX_ENQ_PREEMPT		= 1LLU << 32,
+
 	/*
 	 * The task being enqueued is the only task available for the cpu. By
 	 * default, ext core keeps executing such tasks but when
@@ -55,6 +63,10 @@ enum scx_pick_idle_cpu_flags {
 	SCX_PICK_IDLE_CORE	= 1LLU << 0,	/* pick a CPU whose SMT siblings are also idle */
 };
 
+enum scx_kick_flags {
+	SCX_KICK_PREEMPT	= 1LLU << 0,	/* force scheduling on the CPU */
+};
+
 #ifdef CONFIG_SCHED_CLASS_EXT
 
 struct sched_enq_and_set_ctx {
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index c77fefae1694..ed544ed5ecd6 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -692,6 +692,9 @@ struct scx_rq {
 	u64			ops_qseq;
 	u64			extra_enq_flags;	/* see move_task_to_local_dsq() */
 	u32			nr_running;
+	cpumask_var_t		cpus_to_kick;
+	cpumask_var_t		cpus_to_preempt;
+	struct irq_work		kick_cpus_irq_work;
 };
 #endif /* CONFIG_SCHED_CLASS_EXT */
 
diff --git a/tools/sched_ext/scx_common.bpf.h b/tools/sched_ext/scx_common.bpf.h
index 0e6d7d3e2d27..1a0184024eaf 100644
--- a/tools/sched_ext/scx_common.bpf.h
+++ b/tools/sched_ext/scx_common.bpf.h
@@ -58,6 +58,7 @@ s32 scx_bpf_create_dsq(u64 dsq_id, s32 node) __ksym;
 bool scx_bpf_consume(u64 dsq_id) __ksym;
 u32 scx_bpf_dispatch_nr_slots(void) __ksym;
 void scx_bpf_dispatch(struct task_struct *p, u64 dsq_id, u64 slice, u64 enq_flags) __ksym;
+void scx_bpf_kick_cpu(s32 cpu, u64 flags) __ksym;
 s32 scx_bpf_dsq_nr_queued(u64 dsq_id) __ksym;
 bool scx_bpf_test_and_clear_cpu_idle(s32 cpu) __ksym;
 s32 scx_bpf_pick_idle_cpu(const cpumask_t *cpus_allowed, u64 flags) __ksym;
-- 
2.41.0

[PATCH 19/34] sched_ext: Add a central scheduler which makes all scheduling decisions on one CPU

[Tejun Heo]

This patch adds a new example scheduler, scx_central, which demonstrates
central scheduling where one CPU is responsible for making all scheduling
decisions in the system using scx_bpf_kick_cpu(). The central CPU makes
scheduling decisions for all CPUs in the system, queues tasks on the
appropriate local dsq's and preempts the worker CPUs. The worker CPUs in
turn preempt the central CPU when it needs tasks to run.

Currently, every CPU depends on its own tick to expire the current task. A
follow-up patch implementing tickless support for sched_ext will allow the
worker CPUs to go full tickless so that they can run completely undisturbed.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
Cc: Julia Lawall <julia.lawall@inria.fr>
---
 tools/sched_ext/.gitignore        |   1 +
 tools/sched_ext/Makefile          |   8 +-
 tools/sched_ext/scx_central.bpf.c | 225 ++++++++++++++++++++++++++++++
 tools/sched_ext/scx_central.c     |  93 ++++++++++++
 4 files changed, 325 insertions(+), 2 deletions(-)
 create mode 100644 tools/sched_ext/scx_central.bpf.c
 create mode 100644 tools/sched_ext/scx_central.c

diff --git a/tools/sched_ext/.gitignore b/tools/sched_ext/.gitignore
index b1dd7580a5b4..7e5dec30a87e 100644
--- a/tools/sched_ext/.gitignore
+++ b/tools/sched_ext/.gitignore
@@ -1,5 +1,6 @@
 scx_simple
 scx_qmap
+scx_central
 *.skel.h
 *.subskel.h
 /tools/
diff --git a/tools/sched_ext/Makefile b/tools/sched_ext/Makefile
index f2723a0cde8b..be0445e071ef 100644
--- a/tools/sched_ext/Makefile
+++ b/tools/sched_ext/Makefile
@@ -115,7 +115,7 @@ BPF_CFLAGS = -g -D__TARGET_ARCH_$(SRCARCH)					\
 	     -Wall -Wno-compare-distinct-pointer-types				\
 	     -O2 -mcpu=v3
 
-all: scx_simple scx_qmap
+all: scx_simple scx_qmap scx_central
 
 # sort removes libbpf duplicates when not cross-building
 MAKE_DIRS := $(sort $(BUILD_DIR)/libbpf $(HOST_BUILD_DIR)/libbpf		\
@@ -174,10 +174,14 @@ scx_qmap: scx_qmap.c scx_qmap.skel.h user_exit_info.h
 	$(CC) $(CFLAGS) -c $< -o $@.o
 	$(CC) -o $@ $@.o $(HOST_BPFOBJ) $(LDFLAGS)
 
+scx_central: scx_central.c scx_central.skel.h user_exit_info.h
+	$(CC) $(CFLAGS) -c $< -o $@.o
+	$(CC) -o $@ $@.o $(HOST_BPFOBJ) $(LDFLAGS)
+
 clean:
 	rm -rf $(SCRATCH_DIR) $(HOST_SCRATCH_DIR)
 	rm -f *.o *.bpf.o *.skel.h *.subskel.h
-	rm -f scx_simple scx_qmap
+	rm -f scx_simple scx_qmap scx_central
 
 .PHONY: all clean
 
diff --git a/tools/sched_ext/scx_central.bpf.c b/tools/sched_ext/scx_central.bpf.c
new file mode 100644
index 000000000000..094e089300a2
--- /dev/null
+++ b/tools/sched_ext/scx_central.bpf.c
@@ -0,0 +1,225 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * A central FIFO sched_ext scheduler which demonstrates the followings:
+ *
+ * a. Making all scheduling decisions from one CPU:
+ *
+ *    The central CPU is the only one making scheduling decisions. All other
+ *    CPUs kick the central CPU when they run out of tasks to run.
+ *
+ *    There is one global BPF queue and the central CPU schedules all CPUs by
+ *    dispatching from the global queue to each CPU's local dsq from dispatch().
+ *    This isn't the most straightforward. e.g. It'd be easier to bounce
+ *    through per-CPU BPF queues. The current design is chosen to maximally
+ *    utilize and verify various SCX mechanisms such as LOCAL_ON dispatching.
+ *
+ * b. Preemption
+ *
+ *    SCX_KICK_PREEMPT is used to trigger scheduling and CPUs to move to the
+ *    next tasks.
+ *
+ * This scheduler is designed to maximize usage of various SCX mechanisms. A
+ * more practical implementation would likely put the scheduling loop outside
+ * the central CPU's dispatch() path and add some form of priority mechanism.
+ *
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+#include "scx_common.bpf.h"
+
+char _license[] SEC("license") = "GPL";
+
+enum {
+	FALLBACK_DSQ_ID		= 0,
+	MAX_CPUS		= 4096,
+	MS_TO_NS		= 1000LLU * 1000,
+	TIMER_INTERVAL_NS	= 1 * MS_TO_NS,
+};
+
+const volatile bool switch_partial;
+const volatile s32 central_cpu;
+const volatile u32 nr_cpu_ids = 64;	/* !0 for veristat, set during init */
+
+u64 nr_total, nr_locals, nr_queued, nr_lost_pids;
+u64 nr_dispatches, nr_mismatches, nr_retries;
+u64 nr_overflows;
+
+struct user_exit_info uei;
+
+struct {
+	__uint(type, BPF_MAP_TYPE_QUEUE);
+	__uint(max_entries, 4096);
+	__type(value, s32);
+} central_q SEC(".maps");
+
+/* can't use percpu map due to bad lookups */
+static bool cpu_gimme_task[MAX_CPUS];
+
+struct central_timer {
+	struct bpf_timer timer;
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_ARRAY);
+	__uint(max_entries, 1);
+	__type(key, u32);
+	__type(value, struct central_timer);
+} central_timer SEC(".maps");
+
+s32 BPF_STRUCT_OPS(central_select_cpu, struct task_struct *p,
+		   s32 prev_cpu, u64 wake_flags)
+{
+	/*
+	 * Steer wakeups to the central CPU as much as possible to avoid
+	 * disturbing other CPUs. It's safe to blindly return the central cpu as
+	 * select_cpu() is a hint and if @p can't be on it, the kernel will
+	 * automatically pick a fallback CPU.
+	 */
+	return central_cpu;
+}
+
+void BPF_STRUCT_OPS(central_enqueue, struct task_struct *p, u64 enq_flags)
+{
+	s32 pid = p->pid;
+
+	__sync_fetch_and_add(&nr_total, 1);
+
+	if (bpf_map_push_elem(&central_q, &pid, 0)) {
+		__sync_fetch_and_add(&nr_overflows, 1);
+		scx_bpf_dispatch(p, FALLBACK_DSQ_ID, SCX_SLICE_DFL, enq_flags);
+		return;
+	}
+
+	__sync_fetch_and_add(&nr_queued, 1);
+
+	if (!scx_bpf_task_running(p))
+		scx_bpf_kick_cpu(central_cpu, SCX_KICK_PREEMPT);
+}
+
+static bool dispatch_to_cpu(s32 cpu)
+{
+	struct task_struct *p;
+	s32 pid;
+
+	bpf_repeat(BPF_MAX_LOOPS) {
+		if (bpf_map_pop_elem(&central_q, &pid))
+			break;
+
+		__sync_fetch_and_sub(&nr_queued, 1);
+
+		p = bpf_task_from_pid(pid);
+		if (!p) {
+			__sync_fetch_and_add(&nr_lost_pids, 1);
+			continue;
+		}
+
+		/*
+		 * If we can't run the task at the top, do the dumb thing and
+		 * bounce it to the fallback dsq.
+		 */
+		if (!bpf_cpumask_test_cpu(cpu, p->cpus_ptr)) {
+			__sync_fetch_and_add(&nr_mismatches, 1);
+			scx_bpf_dispatch(p, FALLBACK_DSQ_ID, SCX_SLICE_DFL, 0);
+			bpf_task_release(p);
+			continue;
+		}
+
+		/* dispatch to local and mark that @cpu doesn't need more */
+		scx_bpf_dispatch(p, SCX_DSQ_LOCAL_ON | cpu, SCX_SLICE_DFL, 0);
+
+		if (cpu != central_cpu)
+			scx_bpf_kick_cpu(cpu, 0);
+
+		bpf_task_release(p);
+		return true;
+	}
+
+	return false;
+}
+
+void BPF_STRUCT_OPS(central_dispatch, s32 cpu, struct task_struct *prev)
+{
+	if (cpu == central_cpu) {
+		/* dispatch for all other CPUs first */
+		__sync_fetch_and_add(&nr_dispatches, 1);
+
+		bpf_for(cpu, 0, nr_cpu_ids) {
+			bool *gimme;
+
+			if (!scx_bpf_dispatch_nr_slots())
+				break;
+
+			/* central's gimme is never set */
+			gimme = MEMBER_VPTR(cpu_gimme_task, [cpu]);
+			if (gimme && !*gimme)
+				continue;
+
+			if (dispatch_to_cpu(cpu))
+				*gimme = false;
+		}
+
+		/*
+		 * Retry if we ran out of dispatch buffer slots as we might have
+		 * skipped some CPUs and also need to dispatch for self. The ext
+		 * core automatically retries if the local dsq is empty but we
+		 * can't rely on that as we're dispatching for other CPUs too.
+		 * Kick self explicitly to retry.
+		 */
+		if (!scx_bpf_dispatch_nr_slots()) {
+			__sync_fetch_and_add(&nr_retries, 1);
+			scx_bpf_kick_cpu(central_cpu, SCX_KICK_PREEMPT);
+			return;
+		}
+
+		/* look for a task to run on the central CPU */
+		if (scx_bpf_consume(FALLBACK_DSQ_ID))
+			return;
+		dispatch_to_cpu(central_cpu);
+	} else {
+		bool *gimme;
+
+		if (scx_bpf_consume(FALLBACK_DSQ_ID))
+			return;
+
+		gimme = MEMBER_VPTR(cpu_gimme_task, [cpu]);
+		if (gimme)
+			*gimme = true;
+
+		/*
+		 * Force dispatch on the scheduling CPU so that it finds a task
+		 * to run for us.
+		 */
+		scx_bpf_kick_cpu(central_cpu, SCX_KICK_PREEMPT);
+	}
+}
+
+int BPF_STRUCT_OPS_SLEEPABLE(central_init)
+{
+	if (!switch_partial)
+		scx_bpf_switch_all();
+
+	return scx_bpf_create_dsq(FALLBACK_DSQ_ID, -1);
+}
+
+void BPF_STRUCT_OPS(central_exit, struct scx_exit_info *ei)
+{
+	uei_record(&uei, ei);
+}
+
+SEC(".struct_ops.link")
+struct sched_ext_ops central_ops = {
+	/*
+	 * We are offloading all scheduling decisions to the central CPU and
+	 * thus being the last task on a given CPU doesn't mean anything
+	 * special. Enqueue the last tasks like any other tasks.
+	 */
+	.flags			= SCX_OPS_ENQ_LAST,
+
+	.select_cpu		= (void *)central_select_cpu,
+	.enqueue		= (void *)central_enqueue,
+	.dispatch		= (void *)central_dispatch,
+	.init			= (void *)central_init,
+	.exit			= (void *)central_exit,
+	.name			= "central",
+};
diff --git a/tools/sched_ext/scx_central.c b/tools/sched_ext/scx_central.c
new file mode 100644
index 000000000000..09cb759984a6
--- /dev/null
+++ b/tools/sched_ext/scx_central.c
@@ -0,0 +1,93 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <unistd.h>
+#include <signal.h>
+#include <assert.h>
+#include <libgen.h>
+#include <bpf/bpf.h>
+#include "user_exit_info.h"
+#include "scx_central.skel.h"
+
+const char help_fmt[] =
+"A central FIFO sched_ext scheduler.\n"
+"\n"
+"See the top-level comment in .bpf.c for more details.\n"
+"\n"
+"Usage: %s [-c CPU] [-p]\n"
+"\n"
+"  -c CPU        Override the central CPU (default: 0)\n"
+"  -p            Switch only tasks on SCHED_EXT policy intead of all\n"
+"  -h            Display this help and exit\n";
+
+static volatile int exit_req;
+
+static void sigint_handler(int dummy)
+{
+	exit_req = 1;
+}
+
+int main(int argc, char **argv)
+{
+	struct scx_central *skel;
+	struct bpf_link *link;
+	u64 seq = 0;
+	s32 opt;
+
+	signal(SIGINT, sigint_handler);
+	signal(SIGTERM, sigint_handler);
+
+	libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
+
+	skel = scx_central__open();
+	assert(skel);
+
+	skel->rodata->central_cpu = 0;
+	skel->rodata->nr_cpu_ids = libbpf_num_possible_cpus();
+
+	while ((opt = getopt(argc, argv, "c:ph")) != -1) {
+		switch (opt) {
+		case 'c':
+			skel->rodata->central_cpu = strtoul(optarg, NULL, 0);
+			break;
+		case 'p':
+			skel->rodata->switch_partial = true;
+			break;
+		default:
+			fprintf(stderr, help_fmt, basename(argv[0]));
+			return opt != 'h';
+		}
+	}
+
+	assert(!scx_central__load(skel));
+
+	link = bpf_map__attach_struct_ops(skel->maps.central_ops);
+	assert(link);
+
+	while (!exit_req && !uei_exited(&skel->bss->uei)) {
+		printf("[SEQ %lu]\n", seq++);
+		printf("total   :%10lu    local:%10lu   queued:%10lu  lost:%10lu\n",
+		       skel->bss->nr_total,
+		       skel->bss->nr_locals,
+		       skel->bss->nr_queued,
+		       skel->bss->nr_lost_pids);
+		printf("                    dispatch:%10lu mismatch:%10lu retry:%10lu\n",
+		       skel->bss->nr_dispatches,
+		       skel->bss->nr_mismatches,
+		       skel->bss->nr_retries);
+		printf("overflow:%10lu\n",
+		       skel->bss->nr_overflows);
+		fflush(stdout);
+		sleep(1);
+	}
+
+	bpf_link__destroy(link);
+	uei_print(&skel->bss->uei);
+	scx_central__destroy(skel);
+	return 0;
+}
-- 
2.41.0

[PATCH 20/34] sched_ext: Make watchdog handle ops.dispatch() looping stall

[Tejun Heo]

The dispatch path retries if the local DSQ is still empty after
ops.dispatch() either dispatched or consumed a task. This is both out of
necessity and for convenience. It has to retry because the dispatch path
might lose the tasks to dequeue while the rq lock is released while trying
to migrate tasks across CPUs, and the retry mechanism makes ops.dispatch()
implementation easier as it only needs to make some forward progress each
iteration.

However, this makes it possible for ops.dispatch() to stall CPUs by
repeatedly dispatching ineligible tasks. If all CPUs are stalled that way,
the watchdog or sysrq handler can't run and the system can't be saved. Let's
address the issue by breaking out of the dispatch loop after 32 iterations.

It is unlikely but not impossible for ops.dispatch() to legitimately go over
the iteration limit. We want to come back to the dispatch path in such cases
as not doing so risks stalling the CPU by idling with runnable tasks
pending. As the previous task is still current in balance_scx(),
resched_curr() doesn't do anything - it will just get cleared. Let's instead
use scx_kick_bpf() which will trigger reschedule after switching to the next
task which will likely be the idle task.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
---
 kernel/sched/ext.c             | 17 +++++++++++++++++
 tools/sched_ext/scx_qmap.bpf.c | 17 +++++++++++++++++
 tools/sched_ext/scx_qmap.c     |  8 ++++++--
 3 files changed, 40 insertions(+), 2 deletions(-)

diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 9e8f9f9fcb3d..48e27d59e621 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -9,6 +9,7 @@
 enum scx_internal_consts {
 	SCX_NR_ONLINE_OPS	= SCX_OP_IDX(init),
 	SCX_DSP_DFL_MAX_BATCH	= 32,
+	SCX_DSP_MAX_LOOPS	= 32,
 	SCX_WATCHDOG_MAX_TIMEOUT = 30 * HZ,
 };
 
@@ -167,6 +168,7 @@ static DEFINE_PER_CPU(struct scx_dsp_ctx, scx_dsp_ctx);
 
 void scx_bpf_dispatch(struct task_struct *p, u64 dsq_id, u64 slice,
 		      u64 enq_flags);
+void scx_bpf_kick_cpu(s32 cpu, u64 flags);
 
 struct scx_task_iter {
 	struct sched_ext_entity		cursor;
@@ -1286,6 +1288,7 @@ static int balance_scx(struct rq *rq, struct task_struct *prev,
 	struct scx_rq *scx_rq = &rq->scx;
 	struct scx_dsp_ctx *dspc = this_cpu_ptr(&scx_dsp_ctx);
 	bool prev_on_scx = prev->sched_class == &ext_sched_class;
+	int nr_loops = SCX_DSP_MAX_LOOPS;
 
 	lockdep_assert_rq_held(rq);
 
@@ -1340,6 +1343,20 @@ static int balance_scx(struct rq *rq, struct task_struct *prev,
 			return 1;
 		if (consume_dispatch_q(rq, rf, &scx_dsq_global))
 			return 1;
+
+		/*
+		 * ops.dispatch() can trap us in this loop by repeatedly
+		 * dispatching ineligible tasks. Break out once in a while to
+		 * allow the watchdog to run. As IRQ can't be enabled in
+		 * balance(), we want to complete this scheduling cycle and then
+		 * start a new one. IOW, we want to call resched_curr() on the
+		 * next, most likely idle, task, not the current one. Use
+		 * scx_bpf_kick_cpu() for deferred kicking.
+		 */
+		if (unlikely(!--nr_loops)) {
+			scx_bpf_kick_cpu(cpu_of(rq), 0);
+			break;
+		}
 	} while (dspc->nr_tasks);
 
 	return 0;
diff --git a/tools/sched_ext/scx_qmap.bpf.c b/tools/sched_ext/scx_qmap.bpf.c
index da43f962ab4e..1c3a7d050e32 100644
--- a/tools/sched_ext/scx_qmap.bpf.c
+++ b/tools/sched_ext/scx_qmap.bpf.c
@@ -28,6 +28,7 @@ const volatile u64 slice_ns = SCX_SLICE_DFL;
 const volatile bool switch_partial;
 const volatile u32 stall_user_nth;
 const volatile u32 stall_kernel_nth;
+const volatile u32 dsp_inf_loop_after;
 const volatile s32 disallow_tgid;
 
 u32 test_error_cnt;
@@ -187,6 +188,22 @@ void BPF_STRUCT_OPS(qmap_dispatch, s32 cpu, struct task_struct *prev)
 	s32 pid;
 	int i;
 
+	if (dsp_inf_loop_after && nr_dispatched > dsp_inf_loop_after) {
+		struct task_struct *p;
+
+		/*
+		 * PID 2 should be kthreadd which should mostly be idle and off
+		 * the scheduler. Let's keep dispatching it to force the kernel
+		 * to call this function over and over again.
+		 */
+		p = bpf_task_from_pid(2);
+		if (p) {
+			scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, slice_ns, 0);
+			bpf_task_release(p);
+			return;
+		}
+	}
+
 	if (!idx || !cnt) {
 		scx_bpf_error("failed to lookup idx[%p], cnt[%p]", idx, cnt);
 		return;
diff --git a/tools/sched_ext/scx_qmap.c b/tools/sched_ext/scx_qmap.c
index 3444e3597b19..805ac453698f 100644
--- a/tools/sched_ext/scx_qmap.c
+++ b/tools/sched_ext/scx_qmap.c
@@ -20,12 +20,13 @@ const char help_fmt[] =
 "\n"
 "See the top-level comment in .bpf.c for more details.\n"
 "\n"
-"Usage: %s [-s SLICE_US] [-e COUNT] [-t COUNT] [-T COUNT] [-d PID] [-p]\n"
+"Usage: %s [-s SLICE_US] [-e COUNT] [-t COUNT] [-T COUNT] [-l COUNT] [-d PID] [-p]\n"
 "\n"
 "  -s SLICE_US   Override slice duration\n"
 "  -e COUNT      Trigger scx_bpf_error() after COUNT enqueues\n"
 "  -t COUNT      Stall every COUNT'th user thread\n"
 "  -T COUNT      Stall every COUNT'th kernel thread\n"
+"  -l COUNT      Trigger dispatch infinite looping after COUNT dispatches\n"
 "  -d PID        Disallow a process from switching into SCHED_EXT (-1 for self)\n"
 "  -p            Switch only tasks on SCHED_EXT policy intead of all\n"
 "  -h            Display this help and exit\n";
@@ -51,7 +52,7 @@ int main(int argc, char **argv)
 	skel = scx_qmap__open();
 	assert(skel);
 
-	while ((opt = getopt(argc, argv, "s:e:t:T:d:ph")) != -1) {
+	while ((opt = getopt(argc, argv, "s:e:t:T:l:d:ph")) != -1) {
 		switch (opt) {
 		case 's':
 			skel->rodata->slice_ns = strtoull(optarg, NULL, 0) * 1000;
@@ -65,6 +66,9 @@ int main(int argc, char **argv)
 		case 'T':
 			skel->rodata->stall_kernel_nth = strtoul(optarg, NULL, 0);
 			break;
+		case 'l':
+			skel->rodata->dsp_inf_loop_after = strtoul(optarg, NULL, 0);
+			break;
 		case 'd':
 			skel->rodata->disallow_tgid = strtol(optarg, NULL, 0);
 			if (skel->rodata->disallow_tgid < 0)
-- 
2.41.0

[PATCH 21/34] sched_ext: Add task state tracking operations

[Tejun Heo]

Being able to track the task runnable and running state transitions are
useful for a variety of purposes including latency tracking and load factor
calculation.

Currently, BPF schedulers don't have a good way of tracking these
transitions. Becoming runnable can be determined from ops.enqueue() but
becoming quiescent can only be inferred from the lack of subsequent enqueue.
Also, as the local dsq can have multiple tasks and some events are handled
in the sched_ext core, it's difficult to determine when a given task starts
and stops executing.

This patch adds sched_ext_ops.runnable(), .running(), .stopping() and
.quiescent() operations to track the task runnable and running state
transitions. They're mostly self explanatory; however, we want to ensure
that running <-> stopping transitions are always contained within runnable
<-> quiescent transitions which is a bit different from how the scheduler
core behaves. This adds a bit of complication. See the comment in
dequeue_task_scx().

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 include/linux/sched/ext.h | 65 +++++++++++++++++++++++++++++++++++++++
 kernel/sched/ext.c        | 31 +++++++++++++++++++
 2 files changed, 96 insertions(+)

diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 5c2df7ccc0a6..5cc37bc30352 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -192,6 +192,71 @@ struct sched_ext_ops {
 	 */
 	void (*dispatch)(s32 cpu, struct task_struct *prev);
 
+	/**
+	 * runnable - A task is becoming runnable on its associated CPU
+	 * @p: task becoming runnable
+	 * @enq_flags: %SCX_ENQ_*
+	 *
+	 * This and the following three functions can be used to track a task's
+	 * execution state transitions. A task becomes ->runnable() on a CPU,
+	 * and then goes through one or more ->running() and ->stopping() pairs
+	 * as it runs on the CPU, and eventually becomes ->quiescent() when it's
+	 * done running on the CPU.
+	 *
+	 * @p is becoming runnable on the CPU because it's
+	 *
+	 * - waking up (%SCX_ENQ_WAKEUP)
+	 * - being moved from another CPU
+	 * - being restored after temporarily taken off the queue for an
+	 *   attribute change.
+	 *
+	 * This and ->enqueue() are related but not coupled. This operation
+	 * notifies @p's state transition and may not be followed by ->enqueue()
+	 * e.g. when @p is being dispatched to a remote CPU. Likewise, a task
+	 * may be ->enqueue()'d without being preceded by this operation e.g.
+	 * after exhausting its slice.
+	 */
+	void (*runnable)(struct task_struct *p, u64 enq_flags);
+
+	/**
+	 * running - A task is starting to run on its associated CPU
+	 * @p: task starting to run
+	 *
+	 * See ->runnable() for explanation on the task state notifiers.
+	 */
+	void (*running)(struct task_struct *p);
+
+	/**
+	 * stopping - A task is stopping execution
+	 * @p: task stopping to run
+	 * @runnable: is task @p still runnable?
+	 *
+	 * See ->runnable() for explanation on the task state notifiers. If
+	 * !@runnable, ->quiescent() will be invoked after this operation
+	 * returns.
+	 */
+	void (*stopping)(struct task_struct *p, bool runnable);
+
+	/**
+	 * quiescent - A task is becoming not runnable on its associated CPU
+	 * @p: task becoming not runnable
+	 * @deq_flags: %SCX_DEQ_*
+	 *
+	 * See ->runnable() for explanation on the task state notifiers.
+	 *
+	 * @p is becoming quiescent on the CPU because it's
+	 *
+	 * - sleeping (%SCX_DEQ_SLEEP)
+	 * - being moved to another CPU
+	 * - being temporarily taken off the queue for an attribute change
+	 *   (%SCX_DEQ_SAVE)
+	 *
+	 * This and ->dequeue() are related but not coupled. This operation
+	 * notifies @p's state transition and may not be preceded by ->dequeue()
+	 * e.g. when @p is being dispatched to a remote CPU.
+	 */
+	void (*quiescent)(struct task_struct *p, u64 deq_flags);
+
 	/**
 	 * yield - Yield CPU
 	 * @from: yielding task
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 48e27d59e621..2951200cd81a 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -775,6 +775,9 @@ static void enqueue_task_scx(struct rq *rq, struct task_struct *p, int enq_flags
 	rq->scx.nr_running++;
 	add_nr_running(rq, 1);
 
+	if (SCX_HAS_OP(runnable))
+		SCX_CALL_OP(SCX_KF_REST, runnable, p, enq_flags);
+
 	do_enqueue_task(rq, p, enq_flags, sticky_cpu);
 }
 
@@ -835,6 +838,26 @@ static void dequeue_task_scx(struct rq *rq, struct task_struct *p, int deq_flags
 
 	ops_dequeue(p, deq_flags);
 
+	/*
+	 * A currently running task which is going off @rq first gets dequeued
+	 * and then stops running. As we want running <-> stopping transitions
+	 * to be contained within runnable <-> quiescent transitions, trigger
+	 * ->stopping() early here instead of in put_prev_task_scx().
+	 *
+	 * @p may go through multiple stopping <-> running transitions between
+	 * here and put_prev_task_scx() if task attribute changes occur while
+	 * balance_scx() leaves @rq unlocked. However, they don't contain any
+	 * information meaningful to the BPF scheduler and can be suppressed by
+	 * skipping the callbacks if the task is !QUEUED.
+	 */
+	if (SCX_HAS_OP(stopping) && task_current(rq, p)) {
+		update_curr_scx(rq);
+		SCX_CALL_OP(SCX_KF_REST, stopping, p, false);
+	}
+
+	if (SCX_HAS_OP(quiescent))
+		SCX_CALL_OP(SCX_KF_REST, quiescent, p, deq_flags);
+
 	if (deq_flags & SCX_DEQ_SLEEP)
 		p->scx.flags |= SCX_TASK_DEQD_FOR_SLEEP;
 	else
@@ -1371,6 +1394,10 @@ static void set_next_task_scx(struct rq *rq, struct task_struct *p, bool first)
 
 	p->se.exec_start = rq_clock_task(rq);
 
+	/* see dequeue_task_scx() on why we skip when !QUEUED */
+	if (SCX_HAS_OP(running) && (p->scx.flags & SCX_TASK_QUEUED))
+		SCX_CALL_OP(SCX_KF_REST, running, p);
+
 	watchdog_unwatch_task(p, true);
 }
 
@@ -1409,6 +1436,10 @@ static void put_prev_task_scx(struct rq *rq, struct task_struct *p)
 
 	update_curr_scx(rq);
 
+	/* see dequeue_task_scx() on why we skip when !QUEUED */
+	if (SCX_HAS_OP(stopping) && (p->scx.flags & SCX_TASK_QUEUED))
+		SCX_CALL_OP(SCX_KF_REST, stopping, p, true);
+
 	/*
 	 * If we're being called from put_prev_task_balance(), balance_scx() may
 	 * have decided that @p should keep running.
-- 
2.41.0

[PATCH 22/34] sched_ext: Implement tickless support

[Tejun Heo]

Allow BPF schedulers to indicate tickless operation by setting p->scx.slice
to SCX_SLICE_INF. A CPU whose current task has infinte slice goes into
tickless operation.

scx_central is updated to use tickless operations for all tasks and
instead use a BPF timer to expire slices. This also uses the SCX_ENQ_PREEMPT
and task state tracking added by the previous patches.

Currently, there is no way to pin the timer on the central CPU, so it may
end up on one of the worker CPUs; however, outside of that, the worker CPUs
can go tickless both while running sched_ext tasks and idling.

With schbench running, scx_central shows:

  root@test ~# grep ^LOC /proc/interrupts; sleep 10; grep ^LOC /proc/interrupts
  LOC:     142024        656        664        449   Local timer interrupts
  LOC:     161663        663        665        449   Local timer interrupts

Without it:

  root@test ~ [SIGINT]# grep ^LOC /proc/interrupts; sleep 10; grep ^LOC /proc/interrupts
  LOC:     188778       3142       3793       3993   Local timer interrupts
  LOC:     198993       5314       6323       6438   Local timer interrupts

While scx_central itself is too barebone to be useful as a
production scheduler, a more featureful central scheduler can be built using
the same approach. Google's experience shows that such an approach can have
significant benefits for certain applications such as VM hosting.

v2: * Convert to BPF inline iterators.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 include/linux/sched/ext.h         |   1 +
 kernel/sched/core.c               |   9 ++-
 kernel/sched/ext.c                |  43 ++++++++++-
 kernel/sched/ext.h                |   2 +
 kernel/sched/sched.h              |   6 ++
 tools/sched_ext/scx_central.bpf.c | 121 ++++++++++++++++++++++++++++--
 tools/sched_ext/scx_central.c     |   3 +-
 7 files changed, 173 insertions(+), 12 deletions(-)

diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 5cc37bc30352..676259b09ac4 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -19,6 +19,7 @@ enum scx_consts {
 	SCX_EXIT_MSG_LEN	= 1024,
 
 	SCX_SLICE_DFL		= 20 * NSEC_PER_MSEC,
+	SCX_SLICE_INF		= U64_MAX,	/* infinite, implies nohz */
 };
 
 /*
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index c976a36dd642..30603331c2d7 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -1222,13 +1222,16 @@ bool sched_can_stop_tick(struct rq *rq)
 		return true;
 
 	/*
-	 * If there are no DL,RR/FIFO tasks, there must only be CFS tasks left;
-	 * if there's more than one we need the tick for involuntary
-	 * preemption.
+	 * If there are no DL,RR/FIFO tasks, there must only be CFS or SCX tasks
+	 * left. For CFS, if there's more than one we need the tick for
+	 * involuntary preemption. For SCX, ask.
 	 */
 	if (!scx_switched_all() && rq->nr_running > 1)
 		return false;
 
+	if (scx_enabled() && !scx_can_stop_tick(rq))
+		return false;
+
 	return true;
 }
 #endif /* CONFIG_NO_HZ_FULL */
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 2951200cd81a..7632bb7a016d 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -477,7 +477,8 @@ static void update_curr_scx(struct rq *rq)
 	account_group_exec_runtime(curr, delta_exec);
 	cgroup_account_cputime(curr, delta_exec);
 
-	curr->scx.slice -= min(curr->scx.slice, delta_exec);
+	if (curr->scx.slice != SCX_SLICE_INF)
+		curr->scx.slice -= min(curr->scx.slice, delta_exec);
 }
 
 static void dispatch_enqueue(struct scx_dispatch_q *dsq, struct task_struct *p,
@@ -1399,6 +1400,20 @@ static void set_next_task_scx(struct rq *rq, struct task_struct *p, bool first)
 		SCX_CALL_OP(SCX_KF_REST, running, p);
 
 	watchdog_unwatch_task(p, true);
+
+	/*
+	 * @p is getting newly scheduled or got kicked after someone updated its
+	 * slice. Refresh whether tick can be stopped. See can_stop_tick_scx().
+	 */
+	if ((p->scx.slice == SCX_SLICE_INF) !=
+	    (bool)(rq->scx.flags & SCX_RQ_CAN_STOP_TICK)) {
+		if (p->scx.slice == SCX_SLICE_INF)
+			rq->scx.flags |= SCX_RQ_CAN_STOP_TICK;
+		else
+			rq->scx.flags &= ~SCX_RQ_CAN_STOP_TICK;
+
+		sched_update_tick_dependency(rq);
+	}
 }
 
 static void put_prev_task_scx(struct rq *rq, struct task_struct *p)
@@ -1981,6 +1996,26 @@ int scx_check_setscheduler(struct task_struct *p, int policy)
 	return 0;
 }
 
+#ifdef CONFIG_NO_HZ_FULL
+bool scx_can_stop_tick(struct rq *rq)
+{
+	struct task_struct *p = rq->curr;
+
+	if (scx_ops_disabling())
+		return false;
+
+	if (p->sched_class != &ext_sched_class)
+		return true;
+
+	/*
+	 * @rq can dispatch from different DSQs, so we can't tell whether it
+	 * needs the tick or not by looking at nr_running. Allow stopping ticks
+	 * iff the BPF scheduler indicated so. See set_next_task_scx().
+	 */
+	return rq->scx.flags & SCX_RQ_CAN_STOP_TICK;
+}
+#endif
+
 /*
  * Omitted operations:
  *
@@ -2141,7 +2176,7 @@ static void scx_ops_disable_workfn(struct kthread_work *work)
 	struct rhashtable_iter rht_iter;
 	struct scx_dispatch_q *dsq;
 	const char *reason;
-	int i, type;
+	int i, cpu, type;
 
 	type = atomic_read(&scx_exit_type);
 	while (true) {
@@ -2239,6 +2274,10 @@ static void scx_ops_disable_workfn(struct kthread_work *work)
 	scx_task_iter_exit(&sti);
 	spin_unlock_irq(&scx_tasks_lock);
 
+	/* kick all CPUs to restore ticks */
+	for_each_possible_cpu(cpu)
+		resched_cpu(cpu);
+
 forward_progress_guaranteed:
 	/*
 	 * Here, every runnable task is guaranteed to make forward progress and
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index f535ec39e660..0953091750a6 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -102,6 +102,7 @@ int scx_fork(struct task_struct *p);
 void scx_post_fork(struct task_struct *p);
 void scx_cancel_fork(struct task_struct *p);
 int scx_check_setscheduler(struct task_struct *p, int policy);
+bool scx_can_stop_tick(struct rq *rq);
 void init_sched_ext_class(void);
 
 __printf(2, 3) void scx_ops_error_type(enum scx_exit_type type,
@@ -162,6 +163,7 @@ static inline void scx_post_fork(struct task_struct *p) {}
 static inline void scx_cancel_fork(struct task_struct *p) {}
 static inline int scx_check_setscheduler(struct task_struct *p,
 					 int policy) { return 0; }
+static inline bool scx_can_stop_tick(struct rq *rq) { return true; }
 static inline void init_sched_ext_class(void) {}
 static inline void scx_notify_sched_tick(void) {}
 
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index ed544ed5ecd6..f79fa633e343 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -686,12 +686,18 @@ struct cfs_rq {
 };
 
 #ifdef CONFIG_SCHED_CLASS_EXT
+/* scx_rq->flags, protected by the rq lock */
+enum scx_rq_flags {
+	SCX_RQ_CAN_STOP_TICK	= 1 << 0,
+};
+
 struct scx_rq {
 	struct scx_dispatch_q	local_dsq;
 	struct list_head	watchdog_list;
 	u64			ops_qseq;
 	u64			extra_enq_flags;	/* see move_task_to_local_dsq() */
 	u32			nr_running;
+	u32			flags;
 	cpumask_var_t		cpus_to_kick;
 	cpumask_var_t		cpus_to_preempt;
 	struct irq_work		kick_cpus_irq_work;
diff --git a/tools/sched_ext/scx_central.bpf.c b/tools/sched_ext/scx_central.bpf.c
index 094e089300a2..f44b9365a177 100644
--- a/tools/sched_ext/scx_central.bpf.c
+++ b/tools/sched_ext/scx_central.bpf.c
@@ -13,7 +13,26 @@
  *    through per-CPU BPF queues. The current design is chosen to maximally
  *    utilize and verify various SCX mechanisms such as LOCAL_ON dispatching.
  *
- * b. Preemption
+ * b. Tickless operation
+ *
+ *    All tasks are dispatched with the infinite slice which allows stopping the
+ *    ticks on CONFIG_NO_HZ_FULL kernels running with the proper nohz_full
+ *    parameter. The tickless operation can be observed through
+ *    /proc/interrupts.
+ *
+ *    Periodic switching is enforced by a periodic timer checking all CPUs and
+ *    preempting them as necessary. Unfortunately, BPF timer currently doesn't
+ *    have a way to pin to a specific CPU, so the periodic timer isn't pinned to
+ *    the central CPU.
+ *
+ * c. Preemption
+ *
+ *    Kthreads are unconditionally queued to the head of a matching local dsq
+ *    and dispatched with SCX_DSQ_PREEMPT. This ensures that a kthread is always
+ *    prioritized over user threads, which is required for ensuring forward
+ *    progress as e.g. the periodic timer may run on a ksoftirqd and if the
+ *    ksoftirqd gets starved by a user thread, there may not be anything else to
+ *    vacate that user thread.
  *
  *    SCX_KICK_PREEMPT is used to trigger scheduling and CPUs to move to the
  *    next tasks.
@@ -42,7 +61,7 @@ const volatile s32 central_cpu;
 const volatile u32 nr_cpu_ids = 64;	/* !0 for veristat, set during init */
 
 u64 nr_total, nr_locals, nr_queued, nr_lost_pids;
-u64 nr_dispatches, nr_mismatches, nr_retries;
+u64 nr_timers, nr_dispatches, nr_mismatches, nr_retries;
 u64 nr_overflows;
 
 struct user_exit_info uei;
@@ -55,6 +74,7 @@ struct {
 
 /* can't use percpu map due to bad lookups */
 static bool cpu_gimme_task[MAX_CPUS];
+static u64 cpu_started_at[MAX_CPUS];
 
 struct central_timer {
 	struct bpf_timer timer;
@@ -67,6 +87,11 @@ struct {
 	__type(value, struct central_timer);
 } central_timer SEC(".maps");
 
+static bool vtime_before(u64 a, u64 b)
+{
+	return (s64)(a - b) < 0;
+}
+
 s32 BPF_STRUCT_OPS(central_select_cpu, struct task_struct *p,
 		   s32 prev_cpu, u64 wake_flags)
 {
@@ -85,9 +110,22 @@ void BPF_STRUCT_OPS(central_enqueue, struct task_struct *p, u64 enq_flags)
 
 	__sync_fetch_and_add(&nr_total, 1);
 
+	/*
+	 * Push per-cpu kthreads at the head of local dsq's and preempt the
+	 * corresponding CPU. This ensures that e.g. ksoftirqd isn't blocked
+	 * behind other threads which is necessary for forward progress
+	 * guarantee as we depend on the BPF timer which may run from ksoftirqd.
+	 */
+	if ((p->flags & PF_KTHREAD) && p->nr_cpus_allowed == 1) {
+		__sync_fetch_and_add(&nr_locals, 1);
+		scx_bpf_dispatch(p, SCX_DSQ_LOCAL, SCX_SLICE_INF,
+				 enq_flags | SCX_ENQ_PREEMPT);
+		return;
+	}
+
 	if (bpf_map_push_elem(&central_q, &pid, 0)) {
 		__sync_fetch_and_add(&nr_overflows, 1);
-		scx_bpf_dispatch(p, FALLBACK_DSQ_ID, SCX_SLICE_DFL, enq_flags);
+		scx_bpf_dispatch(p, FALLBACK_DSQ_ID, SCX_SLICE_INF, enq_flags);
 		return;
 	}
 
@@ -120,13 +158,13 @@ static bool dispatch_to_cpu(s32 cpu)
 		 */
 		if (!bpf_cpumask_test_cpu(cpu, p->cpus_ptr)) {
 			__sync_fetch_and_add(&nr_mismatches, 1);
-			scx_bpf_dispatch(p, FALLBACK_DSQ_ID, SCX_SLICE_DFL, 0);
+			scx_bpf_dispatch(p, FALLBACK_DSQ_ID, SCX_SLICE_INF, 0);
 			bpf_task_release(p);
 			continue;
 		}
 
 		/* dispatch to local and mark that @cpu doesn't need more */
-		scx_bpf_dispatch(p, SCX_DSQ_LOCAL_ON | cpu, SCX_SLICE_DFL, 0);
+		scx_bpf_dispatch(p, SCX_DSQ_LOCAL_ON | cpu, SCX_SLICE_INF, 0);
 
 		if (cpu != central_cpu)
 			scx_bpf_kick_cpu(cpu, 0);
@@ -194,12 +232,81 @@ void BPF_STRUCT_OPS(central_dispatch, s32 cpu, struct task_struct *prev)
 	}
 }
 
+void BPF_STRUCT_OPS(central_running, struct task_struct *p)
+{
+	s32 cpu = scx_bpf_task_cpu(p);
+	u64 *started_at = MEMBER_VPTR(cpu_started_at, [cpu]);
+	if (started_at)
+		*started_at = bpf_ktime_get_ns() ?: 1;	/* 0 indicates idle */
+}
+
+void BPF_STRUCT_OPS(central_stopping, struct task_struct *p, bool runnable)
+{
+	s32 cpu = scx_bpf_task_cpu(p);
+	u64 *started_at = MEMBER_VPTR(cpu_started_at, [cpu]);
+	if (started_at)
+		*started_at = 0;
+}
+
+static int central_timerfn(void *map, int *key, struct bpf_timer *timer)
+{
+	u64 now = bpf_ktime_get_ns();
+	u64 nr_to_kick = nr_queued;
+	s32 i;
+
+	bpf_for(i, 0, nr_cpu_ids) {
+		s32 cpu = (nr_timers + i) % nr_cpu_ids;
+		u64 *started_at;
+
+		if (cpu == central_cpu)
+			continue;
+
+		/* kick iff the current one exhausted its slice */
+		started_at = MEMBER_VPTR(cpu_started_at, [cpu]);
+		if (started_at && *started_at &&
+		    vtime_before(now, *started_at + SCX_SLICE_DFL))
+			continue;
+
+		/* and there's something pending */
+		if (scx_bpf_dsq_nr_queued(FALLBACK_DSQ_ID) ||
+		    scx_bpf_dsq_nr_queued(SCX_DSQ_LOCAL_ON | cpu))
+			;
+		else if (nr_to_kick)
+			nr_to_kick--;
+		else
+			continue;
+
+		scx_bpf_kick_cpu(cpu, SCX_KICK_PREEMPT);
+	}
+
+	scx_bpf_kick_cpu(central_cpu, SCX_KICK_PREEMPT);
+
+	bpf_timer_start(timer, TIMER_INTERVAL_NS, 0);
+	__sync_fetch_and_add(&nr_timers, 1);
+	return 0;
+}
+
 int BPF_STRUCT_OPS_SLEEPABLE(central_init)
 {
+	u32 key = 0;
+	struct bpf_timer *timer;
+	int ret;
+
 	if (!switch_partial)
 		scx_bpf_switch_all();
 
-	return scx_bpf_create_dsq(FALLBACK_DSQ_ID, -1);
+	ret = scx_bpf_create_dsq(FALLBACK_DSQ_ID, -1);
+	if (ret)
+		return ret;
+
+	timer = bpf_map_lookup_elem(&central_timer, &key);
+	if (!timer)
+		return -ESRCH;
+
+	bpf_timer_init(timer, &central_timer, CLOCK_MONOTONIC);
+	bpf_timer_set_callback(timer, central_timerfn);
+	ret = bpf_timer_start(timer, TIMER_INTERVAL_NS, 0);
+	return ret;
 }
 
 void BPF_STRUCT_OPS(central_exit, struct scx_exit_info *ei)
@@ -219,6 +326,8 @@ struct sched_ext_ops central_ops = {
 	.select_cpu		= (void *)central_select_cpu,
 	.enqueue		= (void *)central_enqueue,
 	.dispatch		= (void *)central_dispatch,
+	.running		= (void *)central_running,
+	.stopping		= (void *)central_stopping,
 	.init			= (void *)central_init,
 	.exit			= (void *)central_exit,
 	.name			= "central",
diff --git a/tools/sched_ext/scx_central.c b/tools/sched_ext/scx_central.c
index 09cb759984a6..7481d3c9123a 100644
--- a/tools/sched_ext/scx_central.c
+++ b/tools/sched_ext/scx_central.c
@@ -76,7 +76,8 @@ int main(int argc, char **argv)
 		       skel->bss->nr_locals,
 		       skel->bss->nr_queued,
 		       skel->bss->nr_lost_pids);
-		printf("                    dispatch:%10lu mismatch:%10lu retry:%10lu\n",
+		printf("timer   :%10lu dispatch:%10lu mismatch:%10lu retry:%10lu\n",
+		       skel->bss->nr_timers,
 		       skel->bss->nr_dispatches,
 		       skel->bss->nr_mismatches,
 		       skel->bss->nr_retries);
-- 
2.41.0

[PATCH 23/34] sched_ext: Track tasks that are subjects of the in-flight SCX operation

[Tejun Heo]

When some SCX operations are in flight, it is known that the subject task's
rq lock is held throughout which makes it safe to access certain fields of
the task - e.g. its current task_group. We want to add SCX kfunc helpers
that can make use of this guarantee - e.g. to help determining the currently
associated CPU cgroup from the task's current task_group.

As it'd be dangerous call such a helper on a task which isn't rq lock
protected, the helper should be able to verify the input task and reject
accordingly. This patch adds sched_ext_entity.kf_tasks[] that track the
tasks which are currently being operated on by a terminal SCX operation. The
new SCX_CALL_OP_[2]TASK[_RET]() can be used when invoking SCX operations
which take tasks as arguments and the scx_kf_allowed_on_arg_tasks() can be
used by kfunc helpers to verify the input task status.

Note that as sched_ext_entity.kf_tasks[] can't handle nesting, the tracking
is currently only limited to terminal SCX operations. If needed in the
future, this restriction can be removed by moving the tracking to the task
side with a couple per-task counters.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
---
 include/linux/sched/ext.h |  2 +
 kernel/sched/ext.c        | 91 +++++++++++++++++++++++++++++++--------
 2 files changed, 76 insertions(+), 17 deletions(-)

diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 676259b09ac4..d2146cbbd9c2 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -448,6 +448,7 @@ enum scx_kf_mask {
 	SCX_KF_REST		= 1 << 5, /* other rq-locked operations */
 
 	__SCX_KF_RQ_LOCKED	= SCX_KF_DISPATCH | SCX_KF_ENQUEUE | SCX_KF_REST,
+	__SCX_KF_TERMINAL	= SCX_KF_ENQUEUE | SCX_KF_REST,
 };
 
 /*
@@ -463,6 +464,7 @@ struct sched_ext_entity {
 	s32			sticky_cpu;
 	s32			holding_cpu;
 	u32			kf_mask;	/* see scx_kf_mask above */
+	struct task_struct	*kf_tasks[2];	/* see SCX_CALL_OP_TASK() */
 	atomic64_t		ops_state;
 	unsigned long		runnable_at;
 
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 7632bb7a016d..490df7a43b70 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -238,6 +238,47 @@ do {										\
 	__ret;									\
 })
 
+/*
+ * Some kfuncs are allowed only on the tasks that are subjects of the
+ * in-progress scx_ops operation for, e.g., locking guarantees. To enforce such
+ * restrictions, the following SCX_CALL_OP_*() variants should be used when
+ * invoking scx_ops operations that take task arguments. These can only be used
+ * for non-nesting operations due to the way the tasks are tracked.
+ *
+ * kfuncs which can only operate on such tasks can in turn use
+ * scx_kf_allowed_on_arg_tasks() to test whether the invocation is allowed on
+ * the specific task.
+ */
+#define SCX_CALL_OP_TASK(mask, op, task, args...)				\
+do {										\
+	BUILD_BUG_ON(mask & ~__SCX_KF_TERMINAL);				\
+	current->scx.kf_tasks[0] = task;					\
+	SCX_CALL_OP(mask, op, task, ##args);					\
+	current->scx.kf_tasks[0] = NULL;					\
+} while (0)
+
+#define SCX_CALL_OP_TASK_RET(mask, op, task, args...)				\
+({										\
+	__typeof__(scx_ops.op(task, ##args)) __ret;				\
+	BUILD_BUG_ON(mask & ~__SCX_KF_TERMINAL);				\
+	current->scx.kf_tasks[0] = task;					\
+	__ret = SCX_CALL_OP_RET(mask, op, task, ##args);			\
+	current->scx.kf_tasks[0] = NULL;					\
+	__ret;									\
+})
+
+#define SCX_CALL_OP_2TASKS_RET(mask, op, task0, task1, args...)			\
+({										\
+	__typeof__(scx_ops.op(task0, task1, ##args)) __ret;			\
+	BUILD_BUG_ON(mask & ~__SCX_KF_TERMINAL);				\
+	current->scx.kf_tasks[0] = task0;					\
+	current->scx.kf_tasks[1] = task1;					\
+	__ret = SCX_CALL_OP_RET(mask, op, task0, task1, ##args);		\
+	current->scx.kf_tasks[0] = NULL;					\
+	current->scx.kf_tasks[1] = NULL;					\
+	__ret;									\
+})
+
 /* @mask is constant, always inline to cull unnecessary branches */
 static __always_inline bool scx_kf_allowed(u32 mask)
 {
@@ -268,6 +309,22 @@ static __always_inline bool scx_kf_allowed(u32 mask)
 	return true;
 }
 
+/* see SCX_CALL_OP_TASK() */
+static __always_inline bool scx_kf_allowed_on_arg_tasks(u32 mask,
+							struct task_struct *p)
+{
+	if (!scx_kf_allowed(__SCX_KF_RQ_LOCKED))
+		return false;
+
+	if (unlikely((p != current->scx.kf_tasks[0] &&
+		      p != current->scx.kf_tasks[1]))) {
+		scx_ops_error("called on a task not being operated on");
+		return false;
+	}
+
+	return true;
+}
+
 /**
  * scx_task_iter_init - Initialize a task iterator
  * @iter: iterator to init
@@ -705,7 +762,7 @@ static void do_enqueue_task(struct rq *rq, struct task_struct *p, u64 enq_flags,
 	WARN_ON_ONCE(*ddsp_taskp);
 	*ddsp_taskp = p;
 
-	SCX_CALL_OP(SCX_KF_ENQUEUE, enqueue, p, enq_flags);
+	SCX_CALL_OP_TASK(SCX_KF_ENQUEUE, enqueue, p, enq_flags);
 
 	/*
 	 * If not directly dispatched, QUEUEING isn't clear yet and dispatch or
@@ -777,7 +834,7 @@ static void enqueue_task_scx(struct rq *rq, struct task_struct *p, int enq_flags
 	add_nr_running(rq, 1);
 
 	if (SCX_HAS_OP(runnable))
-		SCX_CALL_OP(SCX_KF_REST, runnable, p, enq_flags);
+		SCX_CALL_OP_TASK(SCX_KF_REST, runnable, p, enq_flags);
 
 	do_enqueue_task(rq, p, enq_flags, sticky_cpu);
 }
@@ -802,7 +859,7 @@ static void ops_dequeue(struct task_struct *p, u64 deq_flags)
 		BUG();
 	case SCX_OPSS_QUEUED:
 		if (SCX_HAS_OP(dequeue))
-			SCX_CALL_OP(SCX_KF_REST, dequeue, p, deq_flags);
+			SCX_CALL_OP_TASK(SCX_KF_REST, dequeue, p, deq_flags);
 
 		if (atomic64_try_cmpxchg(&p->scx.ops_state, &opss,
 					 SCX_OPSS_NONE))
@@ -853,11 +910,11 @@ static void dequeue_task_scx(struct rq *rq, struct task_struct *p, int deq_flags
 	 */
 	if (SCX_HAS_OP(stopping) && task_current(rq, p)) {
 		update_curr_scx(rq);
-		SCX_CALL_OP(SCX_KF_REST, stopping, p, false);
+		SCX_CALL_OP_TASK(SCX_KF_REST, stopping, p, false);
 	}
 
 	if (SCX_HAS_OP(quiescent))
-		SCX_CALL_OP(SCX_KF_REST, quiescent, p, deq_flags);
+		SCX_CALL_OP_TASK(SCX_KF_REST, quiescent, p, deq_flags);
 
 	if (deq_flags & SCX_DEQ_SLEEP)
 		p->scx.flags |= SCX_TASK_DEQD_FOR_SLEEP;
@@ -876,7 +933,7 @@ static void yield_task_scx(struct rq *rq)
 	struct task_struct *p = rq->curr;
 
 	if (SCX_HAS_OP(yield))
-		SCX_CALL_OP_RET(SCX_KF_REST, yield, p, NULL);
+		SCX_CALL_OP_2TASKS_RET(SCX_KF_REST, yield, p, NULL);
 	else
 		p->scx.slice = 0;
 }
@@ -886,7 +943,7 @@ static bool yield_to_task_scx(struct rq *rq, struct task_struct *to)
 	struct task_struct *from = rq->curr;
 
 	if (SCX_HAS_OP(yield))
-		return SCX_CALL_OP_RET(SCX_KF_REST, yield, from, to);
+		return SCX_CALL_OP_2TASKS_RET(SCX_KF_REST, yield, from, to);
 	else
 		return false;
 }
@@ -1397,7 +1454,7 @@ static void set_next_task_scx(struct rq *rq, struct task_struct *p, bool first)
 
 	/* see dequeue_task_scx() on why we skip when !QUEUED */
 	if (SCX_HAS_OP(running) && (p->scx.flags & SCX_TASK_QUEUED))
-		SCX_CALL_OP(SCX_KF_REST, running, p);
+		SCX_CALL_OP_TASK(SCX_KF_REST, running, p);
 
 	watchdog_unwatch_task(p, true);
 
@@ -1453,7 +1510,7 @@ static void put_prev_task_scx(struct rq *rq, struct task_struct *p)
 
 	/* see dequeue_task_scx() on why we skip when !QUEUED */
 	if (SCX_HAS_OP(stopping) && (p->scx.flags & SCX_TASK_QUEUED))
-		SCX_CALL_OP(SCX_KF_REST, stopping, p, true);
+		SCX_CALL_OP_TASK(SCX_KF_REST, stopping, p, true);
 
 	/*
 	 * If we're being called from put_prev_task_balance(), balance_scx() may
@@ -1644,8 +1701,8 @@ static int select_task_rq_scx(struct task_struct *p, int prev_cpu, int wake_flag
 	if (SCX_HAS_OP(select_cpu)) {
 		s32 cpu;
 
-		cpu = SCX_CALL_OP_RET(SCX_KF_REST, select_cpu, p, prev_cpu,
-				      wake_flags);
+		cpu = SCX_CALL_OP_TASK_RET(SCX_KF_REST, select_cpu, p, prev_cpu,
+					   wake_flags);
 		if (ops_cpu_valid(cpu)) {
 			return cpu;
 		} else {
@@ -1671,8 +1728,8 @@ static void set_cpus_allowed_scx(struct task_struct *p,
 	 * designation pointless. Cast it away when calling the operation.
 	 */
 	if (SCX_HAS_OP(set_cpumask))
-		SCX_CALL_OP(SCX_KF_REST, set_cpumask, p,
-			    (struct cpumask *)p->cpus_ptr);
+		SCX_CALL_OP_TASK(SCX_KF_REST, set_cpumask, p,
+				 (struct cpumask *)p->cpus_ptr);
 }
 
 static void reset_idle_masks(void)
@@ -1834,7 +1891,7 @@ static void scx_ops_enable_task(struct task_struct *p)
 
 	if (SCX_HAS_OP(enable)) {
 		struct scx_enable_args args = { };
-		SCX_CALL_OP(SCX_KF_REST, enable, p, &args);
+		SCX_CALL_OP_TASK(SCX_KF_REST, enable, p, &args);
 	}
 	p->scx.flags &= ~SCX_TASK_OPS_PREPPED;
 	p->scx.flags |= SCX_TASK_OPS_ENABLED;
@@ -1879,7 +1936,7 @@ static void refresh_scx_weight(struct task_struct *p)
 	lockdep_assert_rq_held(task_rq(p));
 	set_task_scx_weight(p);
 	if (SCX_HAS_OP(set_weight))
-		SCX_CALL_OP(SCX_KF_REST, set_weight, p, p->scx.weight);
+		SCX_CALL_OP_TASK(SCX_KF_REST, set_weight, p, p->scx.weight);
 }
 
 void scx_pre_fork(struct task_struct *p)
@@ -1977,8 +2034,8 @@ static void switching_to_scx(struct rq *rq, struct task_struct *p)
 	 * different scheduler class. Keep the BPF scheduler up-to-date.
 	 */
 	if (SCX_HAS_OP(set_cpumask))
-		SCX_CALL_OP(SCX_KF_REST, set_cpumask, p,
-			    (struct cpumask *)p->cpus_ptr);
+		SCX_CALL_OP_TASK(SCX_KF_REST, set_cpumask, p,
+				 (struct cpumask *)p->cpus_ptr);
 }
 
 static void check_preempt_curr_scx(struct rq *rq, struct task_struct *p,int wake_flags) {}
-- 
2.41.0

[PATCH 24/34] sched_ext: Add cgroup support

[Tejun Heo]

Add sched_ext_ops operations to init/exit cgroups, and track task migrations
and config changes. Because different BPF schedulers may implement different
subsets of CPU control features, allow BPF schedulers to pick which cgroup
interface files to enable using SCX_OPS_CGROUP_KNOB_* flags. For now, only
the weight knobs are supported but adding more should be straightforward.

While a BPF scheduler is being enabled and disabled, relevant cgroup
operations are locked out using scx_cgroup_rwsem. This avoids situations
like task prep taking place while the task is being moved across cgroups,
making things easier for BPF schedulers.

v4: * Example schedulers moved into their own patches.

    * Fix build failure when !CONFIG_CGROUP_SCHED, reported by Andrea Righi.

v3: * Make scx_example_pair switch all tasks by default.

    * Convert to BPF inline iterators.

    * scx_bpf_task_cgroup() is added to determine the current cgroup from
      CPU controller's POV. This allows BPF schedulers to accurately track
      CPU cgroup membership.

    * scx_example_flatcg added. This demonstrates flattened hierarchy
      implementation of CPU cgroup control and shows significant performance
      improvement when cgroups which are nested multiple levels are under
      competition.

v2: * Build fixes for different CONFIG combinations.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
Reported-by: kernel test robot <lkp@intel.com>
Cc: Andrea Righi <andrea.righi@canonical.com>
---
 include/linux/sched/ext.h        | 100 +++++++-
 init/Kconfig                     |   5 +
 kernel/sched/core.c              |  70 +++++-
 kernel/sched/ext.c               | 395 ++++++++++++++++++++++++++++++-
 kernel/sched/ext.h               |  25 ++
 kernel/sched/sched.h             |  12 +-
 tools/sched_ext/scx_common.bpf.h |   1 +
 7 files changed, 586 insertions(+), 22 deletions(-)

diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index d2146cbbd9c2..2d6da27bdd74 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -12,6 +12,8 @@
 #include <linux/rhashtable.h>
 #include <linux/llist.h>
 
+struct cgroup;
+
 enum scx_consts {
 	SCX_OPS_NAME_LEN	= 128,
 	SCX_EXIT_REASON_LEN	= 128,
@@ -108,14 +110,29 @@ enum scx_ops_flags {
 	 */
 	SCX_OPS_ENQ_EXITING	= 1LLU << 2,
 
+	/*
+	 * CPU cgroup knob enable flags
+	 */
+	SCX_OPS_CGROUP_KNOB_WEIGHT = 1LLU << 16,	/* cpu.weight */
+
 	SCX_OPS_ALL_FLAGS	= SCX_OPS_KEEP_BUILTIN_IDLE |
 				  SCX_OPS_ENQ_LAST |
-				  SCX_OPS_ENQ_EXITING,
+				  SCX_OPS_ENQ_EXITING |
+				  SCX_OPS_CGROUP_KNOB_WEIGHT,
 };
 
 /* argument container for ops.enable() and friends */
 struct scx_enable_args {
-	/* empty for now */
+#ifdef CONFIG_EXT_GROUP_SCHED
+	/* the cgroup the task is joining */
+	struct cgroup		*cgroup;
+#endif
+};
+
+/* argument container for ops->cgroup_init() */
+struct scx_cgroup_init_args {
+	/* the weight of the cgroup [1..10000] */
+	u32			weight;
 };
 
 /**
@@ -332,7 +349,8 @@ struct sched_ext_ops {
 	 * @p: task to enable BPF scheduling for
 	 * @args: enable arguments, see the struct definition
 	 *
-	 * Enable @p for BPF scheduling. @p will start running soon.
+	 * Enable @p for BPF scheduling. @p is now in the cgroup specified for
+	 * the preceding prep_enable() and will start running soon.
 	 */
 	void (*enable)(struct task_struct *p, struct scx_enable_args *args);
 
@@ -356,6 +374,79 @@ struct sched_ext_ops {
 	 */
 	void (*disable)(struct task_struct *p);
 
+#ifdef CONFIG_EXT_GROUP_SCHED
+	/**
+	 * cgroup_init - Initialize a cgroup
+	 * @cgrp: cgroup being initialized
+	 * @args: init arguments, see the struct definition
+	 *
+	 * Either the BPF scheduler is being loaded or @cgrp created, initialize
+	 * @cgrp for sched_ext. This operation may block.
+	 *
+	 * Return 0 for success, -errno for failure. An error return while
+	 * loading will abort loading of the BPF scheduler. During cgroup
+	 * creation, it will abort the specific cgroup creation.
+	 */
+	s32 (*cgroup_init)(struct cgroup *cgrp,
+			   struct scx_cgroup_init_args *args);
+
+	/**
+	 * cgroup_exit - Exit a cgroup
+	 * @cgrp: cgroup being exited
+	 *
+	 * Either the BPF scheduler is being unloaded or @cgrp destroyed, exit
+	 * @cgrp for sched_ext. This operation my block.
+	 */
+	void (*cgroup_exit)(struct cgroup *cgrp);
+
+	/**
+	 * cgroup_prep_move - Prepare a task to be moved to a different cgroup
+	 * @p: task being moved
+	 * @from: cgroup @p is being moved from
+	 * @to: cgroup @p is being moved to
+	 *
+	 * Prepare @p for move from cgroup @from to @to. This operation may
+	 * block and can be used for allocations.
+	 *
+	 * Return 0 for success, -errno for failure. An error return aborts the
+	 * migration.
+	 */
+	s32 (*cgroup_prep_move)(struct task_struct *p,
+				struct cgroup *from, struct cgroup *to);
+
+	/**
+	 * cgroup_move - Commit cgroup move
+	 * @p: task being moved
+	 * @from: cgroup @p is being moved from
+	 * @to: cgroup @p is being moved to
+	 *
+	 * Commit the move. @p is dequeued during this operation.
+	 */
+	void (*cgroup_move)(struct task_struct *p,
+			    struct cgroup *from, struct cgroup *to);
+
+	/**
+	 * cgroup_cancel_move - Cancel cgroup move
+	 * @p: task whose cgroup move is being canceled
+	 * @from: cgroup @p was being moved from
+	 * @to: cgroup @p was being moved to
+	 *
+	 * @p was cgroup_prep_move()'d but failed before reaching cgroup_move().
+	 * Undo the preparation.
+	 */
+	void (*cgroup_cancel_move)(struct task_struct *p,
+				   struct cgroup *from, struct cgroup *to);
+
+	/**
+	 * cgroup_set_weight - A cgroup's weight is being changed
+	 * @cgrp: cgroup whose weight is being updated
+	 * @weight: new weight [1..10000]
+	 *
+	 * Update @tg's weight to @weight.
+	 */
+	void (*cgroup_set_weight)(struct cgroup *cgrp, u32 weight);
+#endif	/* CONFIG_CGROUPS */
+
 	/*
 	 * All online ops must come before ops.init().
 	 */
@@ -496,6 +587,9 @@ struct sched_ext_entity {
 
 	/* cold fields */
 	struct list_head	tasks_node;
+#ifdef CONFIG_EXT_GROUP_SCHED
+	struct cgroup		*cgrp_moving_from;
+#endif
 };
 
 void sched_ext_free(struct task_struct *p);
diff --git a/init/Kconfig b/init/Kconfig
index 32c24950c4ce..d4823d842f78 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -1011,6 +1011,11 @@ config RT_GROUP_SCHED
 	  realtime bandwidth for them.
 	  See Documentation/scheduler/sched-rt-group.rst for more information.
 
+config EXT_GROUP_SCHED
+	bool
+	depends on SCHED_CLASS_EXT && CGROUP_SCHED
+	default y
+
 endif #CGROUP_SCHED
 
 config SCHED_MM_CID
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 30603331c2d7..77eb4ee4f759 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -9987,6 +9987,9 @@ void __init sched_init(void)
 		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
 		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
 #endif /* CONFIG_FAIR_GROUP_SCHED */
+#ifdef CONFIG_EXT_GROUP_SCHED
+		root_task_group.scx_weight = CGROUP_WEIGHT_DFL;
+#endif /* CONFIG_EXT_GROUP_SCHED */
 #ifdef CONFIG_RT_GROUP_SCHED
 		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
 		ptr += nr_cpu_ids * sizeof(void **);
@@ -10443,6 +10446,7 @@ struct task_group *sched_create_group(struct task_group *parent)
 	if (!alloc_rt_sched_group(tg, parent))
 		goto err;
 
+	scx_group_set_weight(tg, CGROUP_WEIGHT_DFL);
 	alloc_uclamp_sched_group(tg, parent);
 
 	return tg;
@@ -10569,6 +10573,7 @@ void sched_move_task(struct task_struct *tsk)
 		put_prev_task(rq, tsk);
 
 	sched_change_group(tsk, group);
+	scx_move_task(tsk);
 
 	if (queued)
 		enqueue_task(rq, tsk, queue_flags);
@@ -10609,6 +10614,11 @@ static int cpu_cgroup_css_online(struct cgroup_subsys_state *css)
 {
 	struct task_group *tg = css_tg(css);
 	struct task_group *parent = css_tg(css->parent);
+	int ret;
+
+	ret = scx_tg_online(tg);
+	if (ret)
+		return ret;
 
 	if (parent)
 		sched_online_group(tg, parent);
@@ -10625,6 +10635,13 @@ static int cpu_cgroup_css_online(struct cgroup_subsys_state *css)
 	return 0;
 }
 
+static void cpu_cgroup_css_offline(struct cgroup_subsys_state *css)
+{
+	struct task_group *tg = css_tg(css);
+
+	scx_tg_offline(tg);
+}
+
 static void cpu_cgroup_css_released(struct cgroup_subsys_state *css)
 {
 	struct task_group *tg = css_tg(css);
@@ -10642,9 +10659,10 @@ static void cpu_cgroup_css_free(struct cgroup_subsys_state *css)
 	sched_unregister_group(tg);
 }
 
-#ifdef CONFIG_RT_GROUP_SCHED
+#if defined(CONFIG_RT_GROUP_SCHED) || defined(CONFIG_EXT_GROUP_SCHED)
 static int cpu_cgroup_can_attach(struct cgroup_taskset *tset)
 {
+#ifdef CONFIG_RT_GROUP_SCHED
 	struct task_struct *task;
 	struct cgroup_subsys_state *css;
 
@@ -10652,7 +10670,8 @@ static int cpu_cgroup_can_attach(struct cgroup_taskset *tset)
 		if (!sched_rt_can_attach(css_tg(css), task))
 			return -EINVAL;
 	}
-	return 0;
+#endif
+	return scx_cgroup_can_attach(tset);
 }
 #endif
 
@@ -10663,8 +10682,17 @@ static void cpu_cgroup_attach(struct cgroup_taskset *tset)
 
 	cgroup_taskset_for_each(task, css, tset)
 		sched_move_task(task);
+
+	scx_cgroup_finish_attach();
 }
 
+#ifdef CONFIG_EXT_GROUP_SCHED
+static void cpu_cgroup_cancel_attach(struct cgroup_taskset *tset)
+{
+	scx_cgroup_cancel_attach(tset);
+}
+#endif
+
 #ifdef CONFIG_UCLAMP_TASK_GROUP
 static void cpu_util_update_eff(struct cgroup_subsys_state *css)
 {
@@ -10846,9 +10874,15 @@ static int cpu_uclamp_max_show(struct seq_file *sf, void *v)
 static int cpu_shares_write_u64(struct cgroup_subsys_state *css,
 				struct cftype *cftype, u64 shareval)
 {
+	int ret;
+
 	if (shareval > scale_load_down(ULONG_MAX))
 		shareval = MAX_SHARES;
-	return sched_group_set_shares(css_tg(css), scale_load(shareval));
+	ret = sched_group_set_shares(css_tg(css), scale_load(shareval));
+	if (!ret)
+		scx_group_set_weight(css_tg(css),
+				     sched_weight_to_cgroup(shareval));
+	return ret;
 }
 
 static u64 cpu_shares_read_u64(struct cgroup_subsys_state *css,
@@ -11312,11 +11346,15 @@ static int cpu_extra_stat_show(struct seq_file *sf,
 	return 0;
 }
 
-#ifdef CONFIG_FAIR_GROUP_SCHED
+#if defined(CONFIG_FAIR_GROUP_SCHED) || defined(CONFIG_EXT_GROUP_SCHED)
 
 static unsigned long tg_weight(struct task_group *tg)
 {
+#ifdef CONFIG_FAIR_GROUP_SCHED
 	return scale_load_down(tg->shares);
+#else
+	return sched_weight_from_cgroup(tg->scx_weight);
+#endif
 }
 
 static u64 cpu_weight_read_u64(struct cgroup_subsys_state *css,
@@ -11329,13 +11367,17 @@ static int cpu_weight_write_u64(struct cgroup_subsys_state *css,
 				struct cftype *cft, u64 cgrp_weight)
 {
 	unsigned long weight;
+	int ret;
 
 	if (cgrp_weight < CGROUP_WEIGHT_MIN || cgrp_weight > CGROUP_WEIGHT_MAX)
 		return -ERANGE;
 
 	weight = sched_weight_from_cgroup(cgrp_weight);
 
-	return sched_group_set_shares(css_tg(css), scale_load(weight));
+	ret = sched_group_set_shares(css_tg(css), scale_load(weight));
+	if (!ret)
+		scx_group_set_weight(css_tg(css), cgrp_weight);
+	return ret;
 }
 
 static s64 cpu_weight_nice_read_s64(struct cgroup_subsys_state *css,
@@ -11360,7 +11402,7 @@ static int cpu_weight_nice_write_s64(struct cgroup_subsys_state *css,
 				     struct cftype *cft, s64 nice)
 {
 	unsigned long weight;
-	int idx;
+	int idx, ret;
 
 	if (nice < MIN_NICE || nice > MAX_NICE)
 		return -ERANGE;
@@ -11369,7 +11411,11 @@ static int cpu_weight_nice_write_s64(struct cgroup_subsys_state *css,
 	idx = array_index_nospec(idx, 40);
 	weight = sched_prio_to_weight[idx];
 
-	return sched_group_set_shares(css_tg(css), scale_load(weight));
+	ret = sched_group_set_shares(css_tg(css), scale_load(weight));
+	if (!ret)
+		scx_group_set_weight(css_tg(css),
+				     sched_weight_to_cgroup(weight));
+	return ret;
 }
 #endif
 
@@ -11431,7 +11477,7 @@ static ssize_t cpu_max_write(struct kernfs_open_file *of,
 #endif
 
 struct cftype cpu_cftypes[CPU_CFTYPE_CNT + 1] = {
-#ifdef CONFIG_FAIR_GROUP_SCHED
+#if defined(CONFIG_FAIR_GROUP_SCHED) || defined(CONFIG_EXT_GROUP_SCHED)
 	[CPU_CFTYPE_WEIGHT] = {
 		.name = "weight",
 		.flags = CFTYPE_NOT_ON_ROOT,
@@ -11444,6 +11490,8 @@ struct cftype cpu_cftypes[CPU_CFTYPE_CNT + 1] = {
 		.read_s64 = cpu_weight_nice_read_s64,
 		.write_s64 = cpu_weight_nice_write_s64,
 	},
+#endif
+#ifdef CONFIG_FAIR_GROUP_SCHED
 	[CPU_CFTYPE_IDLE] = {
 		.name = "idle",
 		.flags = CFTYPE_NOT_ON_ROOT,
@@ -11485,13 +11533,17 @@ struct cftype cpu_cftypes[CPU_CFTYPE_CNT + 1] = {
 struct cgroup_subsys cpu_cgrp_subsys = {
 	.css_alloc	= cpu_cgroup_css_alloc,
 	.css_online	= cpu_cgroup_css_online,
+	.css_offline	= cpu_cgroup_css_offline,
 	.css_released	= cpu_cgroup_css_released,
 	.css_free	= cpu_cgroup_css_free,
 	.css_extra_stat_show = cpu_extra_stat_show,
-#ifdef CONFIG_RT_GROUP_SCHED
+#if defined(CONFIG_RT_GROUP_SCHED) || defined(CONFIG_EXT_GROUP_SCHED)
 	.can_attach	= cpu_cgroup_can_attach,
 #endif
 	.attach		= cpu_cgroup_attach,
+#ifdef CONFIG_EXT_GROUP_SCHED
+	.cancel_attach	= cpu_cgroup_cancel_attach,
+#endif
 	.legacy_cftypes	= cpu_legacy_cftypes,
 	.dfl_cftypes	= cpu_cftypes,
 	.early_init	= true,
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 490df7a43b70..5862e8290207 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -1841,6 +1841,28 @@ static void task_tick_scx(struct rq *rq, struct task_struct *curr, int queued)
 		resched_curr(rq);
 }
 
+#ifdef CONFIG_EXT_GROUP_SCHED
+static struct cgroup *tg_cgrp(struct task_group *tg)
+{
+	/*
+	 * If CGROUP_SCHED is disabled, @tg is NULL. If @tg is an autogroup,
+	 * @tg->css.cgroup is NULL. In both cases, @tg can be treated as the
+	 * root cgroup.
+	 */
+	if (tg && tg->css.cgroup)
+		return tg->css.cgroup;
+	else
+		return &cgrp_dfl_root.cgrp;
+}
+
+#define SCX_ENABLE_ARGS_INIT_CGROUP(tg)		.cgroup = tg_cgrp(tg),
+
+#else	/* CONFIG_EXT_GROUP_SCHED */
+
+#define SCX_ENABLE_ARGS_INIT_CGROUP(tg)
+
+#endif	/* CONFIG_EXT_GROUP_SCHED */
+
 static int scx_ops_prepare_task(struct task_struct *p, struct task_group *tg)
 {
 	int ret;
@@ -1850,7 +1872,9 @@ static int scx_ops_prepare_task(struct task_struct *p, struct task_group *tg)
 	p->scx.disallow = false;
 
 	if (SCX_HAS_OP(prep_enable)) {
-		struct scx_enable_args args = { };
+		struct scx_enable_args args = {
+			SCX_ENABLE_ARGS_INIT_CGROUP(tg)
+		};
 
 		ret = SCX_CALL_OP_RET(SCX_KF_SLEEPABLE, prep_enable, p, &args);
 		if (unlikely(ret)) {
@@ -1890,7 +1914,9 @@ static void scx_ops_enable_task(struct task_struct *p)
 	WARN_ON_ONCE(!(p->scx.flags & SCX_TASK_OPS_PREPPED));
 
 	if (SCX_HAS_OP(enable)) {
-		struct scx_enable_args args = { };
+		struct scx_enable_args args = {
+			SCX_ENABLE_ARGS_INIT_CGROUP(task_group(p))
+		};
 		SCX_CALL_OP_TASK(SCX_KF_REST, enable, p, &args);
 	}
 	p->scx.flags &= ~SCX_TASK_OPS_PREPPED;
@@ -1903,7 +1929,9 @@ static void scx_ops_disable_task(struct task_struct *p)
 
 	if (p->scx.flags & SCX_TASK_OPS_PREPPED) {
 		if (SCX_HAS_OP(cancel_enable)) {
-			struct scx_enable_args args = { };
+			struct scx_enable_args args = {
+				SCX_ENABLE_ARGS_INIT_CGROUP(task_group(p))
+			};
 			SCX_CALL_OP(SCX_KF_REST, cancel_enable, p, &args);
 		}
 		p->scx.flags &= ~SCX_TASK_OPS_PREPPED;
@@ -2073,6 +2101,166 @@ bool scx_can_stop_tick(struct rq *rq)
 }
 #endif
 
+#ifdef CONFIG_EXT_GROUP_SCHED
+
+DEFINE_STATIC_PERCPU_RWSEM(scx_cgroup_rwsem);
+
+int scx_tg_online(struct task_group *tg)
+{
+	int ret = 0;
+
+	WARN_ON_ONCE(tg->scx_flags & (SCX_TG_ONLINE | SCX_TG_INITED));
+
+	percpu_down_read(&scx_cgroup_rwsem);
+
+	if (SCX_HAS_OP(cgroup_init)) {
+		struct scx_cgroup_init_args args = { .weight = tg->scx_weight };
+
+		ret = SCX_CALL_OP_RET(SCX_KF_SLEEPABLE, cgroup_init,
+				      tg->css.cgroup, &args);
+		if (!ret)
+			tg->scx_flags |= SCX_TG_ONLINE | SCX_TG_INITED;
+		else
+			ret = ops_sanitize_err("cgroup_init", ret);
+	} else {
+		tg->scx_flags |= SCX_TG_ONLINE;
+	}
+
+	percpu_up_read(&scx_cgroup_rwsem);
+	return ret;
+}
+
+void scx_tg_offline(struct task_group *tg)
+{
+	WARN_ON_ONCE(!(tg->scx_flags & SCX_TG_ONLINE));
+
+	percpu_down_read(&scx_cgroup_rwsem);
+
+	if (SCX_HAS_OP(cgroup_exit) && (tg->scx_flags & SCX_TG_INITED))
+		SCX_CALL_OP(SCX_KF_SLEEPABLE, cgroup_exit, tg->css.cgroup);
+	tg->scx_flags &= ~(SCX_TG_ONLINE | SCX_TG_INITED);
+
+	percpu_up_read(&scx_cgroup_rwsem);
+}
+
+int scx_cgroup_can_attach(struct cgroup_taskset *tset)
+{
+	struct cgroup_subsys_state *css;
+	struct task_struct *p;
+	int ret;
+
+	/* released in scx_finish/cancel_attach() */
+	percpu_down_read(&scx_cgroup_rwsem);
+
+	if (!scx_enabled())
+		return 0;
+
+	cgroup_taskset_for_each(p, css, tset) {
+		struct cgroup *from = tg_cgrp(task_group(p));
+
+		if (SCX_HAS_OP(cgroup_prep_move)) {
+			ret = SCX_CALL_OP_RET(SCX_KF_SLEEPABLE, cgroup_prep_move,
+					      p, from, css->cgroup);
+			if (ret)
+				goto err;
+		}
+
+		WARN_ON_ONCE(p->scx.cgrp_moving_from);
+		p->scx.cgrp_moving_from = from;
+	}
+
+	return 0;
+
+err:
+	cgroup_taskset_for_each(p, css, tset) {
+		if (!p->scx.cgrp_moving_from)
+			break;
+		if (SCX_HAS_OP(cgroup_cancel_move))
+			SCX_CALL_OP(SCX_KF_SLEEPABLE, cgroup_cancel_move, p,
+				    p->scx.cgrp_moving_from, css->cgroup);
+		p->scx.cgrp_moving_from = NULL;
+	}
+
+	percpu_up_read(&scx_cgroup_rwsem);
+	return ops_sanitize_err("cgroup_prep_move", ret);
+}
+
+void scx_move_task(struct task_struct *p)
+{
+	/*
+	 * We're called from sched_move_task() which handles both cgroup and
+	 * autogroup moves. Ignore the latter.
+	 */
+	if (task_group_is_autogroup(task_group(p)))
+		return;
+
+	if (!scx_enabled())
+		return;
+
+	if (SCX_HAS_OP(cgroup_move)) {
+		WARN_ON_ONCE(!p->scx.cgrp_moving_from);
+		SCX_CALL_OP_TASK(SCX_KF_UNLOCKED, cgroup_move, p,
+			p->scx.cgrp_moving_from, tg_cgrp(task_group(p)));
+	}
+	p->scx.cgrp_moving_from = NULL;
+}
+
+void scx_cgroup_finish_attach(void)
+{
+	percpu_up_read(&scx_cgroup_rwsem);
+}
+
+void scx_cgroup_cancel_attach(struct cgroup_taskset *tset)
+{
+	struct cgroup_subsys_state *css;
+	struct task_struct *p;
+
+	if (!scx_enabled())
+		goto out_unlock;
+
+	cgroup_taskset_for_each(p, css, tset) {
+		if (SCX_HAS_OP(cgroup_cancel_move)) {
+			WARN_ON_ONCE(!p->scx.cgrp_moving_from);
+			SCX_CALL_OP(SCX_KF_SLEEPABLE, cgroup_cancel_move, p,
+				    p->scx.cgrp_moving_from, css->cgroup);
+		}
+		p->scx.cgrp_moving_from = NULL;
+	}
+out_unlock:
+	percpu_up_read(&scx_cgroup_rwsem);
+}
+
+void scx_group_set_weight(struct task_group *tg, unsigned long weight)
+{
+	percpu_down_read(&scx_cgroup_rwsem);
+
+	if (tg->scx_weight != weight) {
+		if (SCX_HAS_OP(cgroup_set_weight))
+			SCX_CALL_OP(SCX_KF_SLEEPABLE, cgroup_set_weight,
+				    tg_cgrp(tg), weight);
+		tg->scx_weight = weight;
+	}
+
+	percpu_up_read(&scx_cgroup_rwsem);
+}
+
+static void scx_cgroup_lock(void)
+{
+	percpu_down_write(&scx_cgroup_rwsem);
+}
+
+static void scx_cgroup_unlock(void)
+{
+	percpu_up_write(&scx_cgroup_rwsem);
+}
+
+#else	/* CONFIG_EXT_GROUP_SCHED */
+
+static inline void scx_cgroup_lock(void) {}
+static inline void scx_cgroup_unlock(void) {}
+
+#endif	/* CONFIG_EXT_GROUP_SCHED */
+
 /*
  * Omitted operations:
  *
@@ -2202,6 +2390,131 @@ static void destroy_dsq(u64 dsq_id)
 	rcu_read_unlock();
 }
 
+#ifdef CONFIG_EXT_GROUP_SCHED
+static void scx_cgroup_exit(void)
+{
+	struct cgroup_subsys_state *css;
+
+	percpu_rwsem_assert_held(&scx_cgroup_rwsem);
+
+	/*
+	 * scx_tg_on/offline() are excluded through scx_cgroup_rwsem. If we walk
+	 * cgroups and exit all the inited ones, all online cgroups are exited.
+	 */
+	rcu_read_lock();
+	css_for_each_descendant_post(css, &root_task_group.css) {
+		struct task_group *tg = css_tg(css);
+
+		if (!(tg->scx_flags & SCX_TG_INITED))
+			continue;
+		tg->scx_flags &= ~SCX_TG_INITED;
+
+		if (!scx_ops.cgroup_exit)
+			continue;
+
+		if (WARN_ON_ONCE(!css_tryget(css)))
+			continue;
+		rcu_read_unlock();
+
+		SCX_CALL_OP(SCX_KF_UNLOCKED, cgroup_exit, css->cgroup);
+
+		rcu_read_lock();
+		css_put(css);
+	}
+	rcu_read_unlock();
+}
+
+static int scx_cgroup_init(void)
+{
+	struct cgroup_subsys_state *css;
+	int ret;
+
+	percpu_rwsem_assert_held(&scx_cgroup_rwsem);
+
+	/*
+	 * scx_tg_on/offline() are excluded thorugh scx_cgroup_rwsem. If we walk
+	 * cgroups and init, all online cgroups are initialized.
+	 */
+	rcu_read_lock();
+	css_for_each_descendant_pre(css, &root_task_group.css) {
+		struct task_group *tg = css_tg(css);
+		struct scx_cgroup_init_args args = { .weight = tg->scx_weight };
+
+		if ((tg->scx_flags &
+		     (SCX_TG_ONLINE | SCX_TG_INITED)) != SCX_TG_ONLINE)
+			continue;
+
+		if (!scx_ops.cgroup_init) {
+			tg->scx_flags |= SCX_TG_INITED;
+			continue;
+		}
+
+		if (WARN_ON_ONCE(!css_tryget(css)))
+			continue;
+		rcu_read_unlock();
+
+		ret = SCX_CALL_OP_RET(SCX_KF_SLEEPABLE, cgroup_init,
+				      css->cgroup, &args);
+		if (ret) {
+			css_put(css);
+			return ret;
+		}
+		tg->scx_flags |= SCX_TG_INITED;
+
+		rcu_read_lock();
+		css_put(css);
+	}
+	rcu_read_unlock();
+
+	return 0;
+}
+
+static void scx_cgroup_config_knobs(void)
+{
+	static DEFINE_MUTEX(cgintf_mutex);
+	DECLARE_BITMAP(mask, CPU_CFTYPE_CNT) = { };
+	u64 knob_flags;
+	int i;
+
+	/*
+	 * Called from both class switch and ops enable/disable paths,
+	 * synchronize internally.
+	 */
+	mutex_lock(&cgintf_mutex);
+
+	/* if fair is in use, all knobs should be shown */
+	if (!scx_switched_all()) {
+		bitmap_fill(mask, CPU_CFTYPE_CNT);
+		goto apply;
+	}
+
+	/*
+	 * On ext, only show the supported knobs. Otherwise, show all possible
+	 * knobs so that configuration attempts succeed and the states are
+	 * remembered while ops is not loaded.
+	 */
+	if (scx_enabled())
+		knob_flags = scx_ops.flags;
+	else
+		knob_flags = SCX_OPS_ALL_FLAGS;
+
+	if (knob_flags & SCX_OPS_CGROUP_KNOB_WEIGHT) {
+		__set_bit(CPU_CFTYPE_WEIGHT, mask);
+		__set_bit(CPU_CFTYPE_WEIGHT_NICE, mask);
+	}
+apply:
+	for (i = 0; i < CPU_CFTYPE_CNT; i++)
+		cgroup_show_cftype(&cpu_cftypes[i], test_bit(i, mask));
+
+	mutex_unlock(&cgintf_mutex);
+}
+
+#else
+static void scx_cgroup_exit(void) {}
+static int scx_cgroup_init(void) { return 0; }
+static void scx_cgroup_config_knobs(void) {}
+#endif
+
 /*
  * Used by sched_fork() and __setscheduler_prio() to pick the matching
  * sched_class. dl/rt are already handled.
@@ -2346,9 +2659,10 @@ static void scx_ops_disable_workfn(struct kthread_work *work)
 	static_branch_disable(&__scx_switched_all);
 	WRITE_ONCE(scx_switching_all, false);
 
-	/* avoid racing against fork */
+	/* avoid racing against fork and cgroup changes */
 	cpus_read_lock();
 	percpu_down_write(&scx_fork_rwsem);
+	scx_cgroup_lock();
 
 	spin_lock_irq(&scx_tasks_lock);
 	scx_task_iter_init(&sti);
@@ -2384,6 +2698,9 @@ static void scx_ops_disable_workfn(struct kthread_work *work)
 	static_branch_disable_cpuslocked(&scx_builtin_idle_enabled);
 	synchronize_rcu();
 
+	scx_cgroup_exit();
+
+	scx_cgroup_unlock();
 	percpu_up_write(&scx_fork_rwsem);
 	cpus_read_unlock();
 
@@ -2422,6 +2739,8 @@ static void scx_ops_disable_workfn(struct kthread_work *work)
 
 	WARN_ON_ONCE(scx_ops_set_enable_state(SCX_OPS_DISABLED) !=
 		     SCX_OPS_DISABLING);
+
+	scx_cgroup_config_knobs();
 }
 
 static DEFINE_KTHREAD_WORK(scx_ops_disable_work, scx_ops_disable_workfn);
@@ -2567,10 +2886,11 @@ static int scx_ops_enable(struct sched_ext_ops *ops)
 			   scx_watchdog_timeout / 2);
 
 	/*
-	 * Lock out forks before opening the floodgate so that they don't wander
-	 * into the operations prematurely.
+	 * Lock out forks, cgroup on/offlining and moves before opening the
+	 * floodgate so that they don't wander into the operations prematurely.
 	 */
 	percpu_down_write(&scx_fork_rwsem);
+	scx_cgroup_lock();
 
 	for (i = 0; i < SCX_NR_ONLINE_OPS; i++)
 		if (((void (**)(void))ops)[i])
@@ -2589,6 +2909,14 @@ static int scx_ops_enable(struct sched_ext_ops *ops)
 		static_branch_disable_cpuslocked(&scx_builtin_idle_enabled);
 	}
 
+	/*
+	 * All cgroups should be initialized before letting in tasks. cgroup
+	 * on/offlining and task migrations are already locked out.
+	 */
+	ret = scx_cgroup_init();
+	if (ret)
+		goto err_disable_unlock;
+
 	static_branch_enable_cpuslocked(&__scx_ops_enabled);
 
 	/*
@@ -2671,6 +2999,7 @@ static int scx_ops_enable(struct sched_ext_ops *ops)
 
 	spin_unlock_irq(&scx_tasks_lock);
 	preempt_enable();
+	scx_cgroup_unlock();
 	percpu_up_write(&scx_fork_rwsem);
 
 	if (!scx_ops_tryset_enable_state(SCX_OPS_ENABLED, SCX_OPS_ENABLING)) {
@@ -2684,6 +3013,8 @@ static int scx_ops_enable(struct sched_ext_ops *ops)
 	cpus_read_unlock();
 	mutex_unlock(&scx_ops_enable_mutex);
 
+	scx_cgroup_config_knobs();
+
 	return 0;
 
 err_unlock:
@@ -2691,6 +3022,7 @@ static int scx_ops_enable(struct sched_ext_ops *ops)
 	return ret;
 
 err_disable_unlock:
+	scx_cgroup_unlock();
 	percpu_up_write(&scx_fork_rwsem);
 err_disable:
 	cpus_read_unlock();
@@ -2848,6 +3180,11 @@ static int bpf_scx_check_member(const struct btf_type *t,
 
 	switch (moff) {
 	case offsetof(struct sched_ext_ops, prep_enable):
+#ifdef CONFIG_EXT_GROUP_SCHED
+	case offsetof(struct sched_ext_ops, cgroup_init):
+	case offsetof(struct sched_ext_ops, cgroup_exit):
+	case offsetof(struct sched_ext_ops, cgroup_prep_move):
+#endif
 	case offsetof(struct sched_ext_ops, init):
 	case offsetof(struct sched_ext_ops, exit):
 		break;
@@ -2965,7 +3302,8 @@ void __init init_sched_ext_class(void)
 	 * definitions so that BPF scheduler implementations can use them
 	 * through the generated vmlinux.h.
 	 */
-	WRITE_ONCE(v, SCX_WAKE_EXEC | SCX_ENQ_WAKEUP | SCX_DEQ_SLEEP);
+	WRITE_ONCE(v, SCX_WAKE_EXEC | SCX_ENQ_WAKEUP | SCX_DEQ_SLEEP |
+		   SCX_TG_ONLINE);
 
 	BUG_ON(rhashtable_init(&dsq_hash, &dsq_hash_params));
 	init_dsq(&scx_dsq_global, SCX_DSQ_GLOBAL);
@@ -2986,6 +3324,7 @@ void __init init_sched_ext_class(void)
 
 	register_sysrq_key('S', &sysrq_sched_ext_reset_op);
 	INIT_DELAYED_WORK(&scx_watchdog_work, scx_watchdog_workfn);
+	scx_cgroup_config_knobs();
 }
 
 
@@ -3027,8 +3366,8 @@ static const struct btf_kfunc_id_set scx_kfunc_set_init = {
  * @dsq_id: DSQ to create
  * @node: NUMA node to allocate from
  *
- * Create a custom DSQ identified by @dsq_id. Can be called from ops.init() and
- * ops.prep_enable().
+ * Create a custom DSQ identified by @dsq_id. Can be called from ops.init(),
+ * ops.prep_enable(), ops.cgroup_init() and ops.cgroup_prep_move().
  */
 s32 scx_bpf_create_dsq(u64 dsq_id, s32 node)
 {
@@ -3514,6 +3853,41 @@ s32 scx_bpf_task_cpu(const struct task_struct *p)
 	return task_cpu(p);
 }
 
+/**
+ * scx_bpf_task_cgroup - Return the sched cgroup of a task
+ * @p: task of interest
+ *
+ * @p->sched_task_group->css.cgroup represents the cgroup @p is associated with
+ * from the scheduler's POV. SCX operations should use this function to
+ * determine @p's current cgroup as, unlike following @p->cgroups,
+ * @p->sched_task_group is protected by @p's rq lock and thus atomic w.r.t. all
+ * rq-locked operations. Can be called on the parameter tasks of rq-locked
+ * operations. The restriction guarantees that @p's rq is locked by the caller.
+ */
+#ifdef CONFIG_CGROUP_SCHED
+struct cgroup *scx_bpf_task_cgroup(struct task_struct *p)
+{
+	struct task_group *tg = p->sched_task_group;
+	struct cgroup *cgrp = &cgrp_dfl_root.cgrp;
+
+	if (!scx_kf_allowed_on_arg_tasks(__SCX_KF_RQ_LOCKED, p))
+		goto out;
+
+	/*
+	 * A task_group may either be a cgroup or an autogroup. In the latter
+	 * case, @tg->css.cgroup is %NULL. A task_group can't become the other
+	 * kind once created.
+	 */
+	if (tg && tg->css.cgroup)
+		cgrp = tg->css.cgroup;
+	else
+		cgrp = &cgrp_dfl_root.cgrp;
+out:
+	cgroup_get(cgrp);
+	return cgrp;
+}
+#endif
+
 BTF_SET8_START(scx_kfunc_ids_any)
 BTF_ID_FLAGS(func, scx_bpf_kick_cpu)
 BTF_ID_FLAGS(func, scx_bpf_dsq_nr_queued)
@@ -3527,6 +3901,9 @@ BTF_ID_FLAGS(func, scx_bpf_error_bstr, KF_TRUSTED_ARGS)
 BTF_ID_FLAGS(func, scx_bpf_destroy_dsq)
 BTF_ID_FLAGS(func, scx_bpf_task_running, KF_RCU)
 BTF_ID_FLAGS(func, scx_bpf_task_cpu, KF_RCU)
+#ifdef CONFIG_CGROUP_SCHED
+BTF_ID_FLAGS(func, scx_bpf_task_cgroup, KF_RCU | KF_ACQUIRE)
+#endif
 BTF_SET8_END(scx_kfunc_ids_any)
 
 static const struct btf_kfunc_id_set scx_kfunc_set_any = {
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index 0953091750a6..c3404a0a7637 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -67,6 +67,11 @@ enum scx_kick_flags {
 	SCX_KICK_PREEMPT	= 1LLU << 0,	/* force scheduling on the CPU */
 };
 
+enum scx_tg_flags {
+	SCX_TG_ONLINE		= 1U << 0,
+	SCX_TG_INITED		= 1U << 1,
+};
+
 #ifdef CONFIG_SCHED_CLASS_EXT
 
 struct sched_enq_and_set_ctx {
@@ -183,3 +188,23 @@ static inline void scx_update_idle(struct rq *rq, bool idle)
 #else
 static inline void scx_update_idle(struct rq *rq, bool idle) {}
 #endif
+
+#ifdef CONFIG_CGROUP_SCHED
+#ifdef CONFIG_EXT_GROUP_SCHED
+int scx_tg_online(struct task_group *tg);
+void scx_tg_offline(struct task_group *tg);
+int scx_cgroup_can_attach(struct cgroup_taskset *tset);
+void scx_move_task(struct task_struct *p);
+void scx_cgroup_finish_attach(void);
+void scx_cgroup_cancel_attach(struct cgroup_taskset *tset);
+void scx_group_set_weight(struct task_group *tg, unsigned long cgrp_weight);
+#else	/* CONFIG_EXT_GROUP_SCHED */
+static inline int scx_tg_online(struct task_group *tg) { return 0; }
+static inline void scx_tg_offline(struct task_group *tg) {}
+static inline int scx_cgroup_can_attach(struct cgroup_taskset *tset) { return 0; }
+static inline void scx_move_task(struct task_struct *p) {}
+static inline void scx_cgroup_finish_attach(void) {}
+static inline void scx_cgroup_cancel_attach(struct cgroup_taskset *tset) {}
+static inline void scx_group_set_weight(struct task_group *tg, unsigned long cgrp_weight) {}
+#endif	/* CONFIG_EXT_GROUP_SCHED */
+#endif	/* CONFIG_CGROUP_SCHED */
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index f79fa633e343..00bf33fdbd64 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -424,6 +424,11 @@ struct task_group {
 	struct rt_bandwidth	rt_bandwidth;
 #endif
 
+#ifdef CONFIG_EXT_GROUP_SCHED
+	u32			scx_flags;	/* SCX_TG_* */
+	u32			scx_weight;
+#endif
+
 	struct rcu_head		rcu;
 	struct list_head	list;
 
@@ -528,6 +533,11 @@ extern void set_task_rq_fair(struct sched_entity *se,
 static inline void set_task_rq_fair(struct sched_entity *se,
 			     struct cfs_rq *prev, struct cfs_rq *next) { }
 #endif /* CONFIG_SMP */
+#else /* CONFIG_FAIR_GROUP_SCHED */
+static inline int sched_group_set_shares(struct task_group *tg, unsigned long shares)
+{
+	return 0;
+}
 #endif /* CONFIG_FAIR_GROUP_SCHED */
 
 #else /* CONFIG_CGROUP_SCHED */
@@ -3558,7 +3568,7 @@ static inline void init_sched_mm_cid(struct task_struct *t) { }
 
 #ifdef CONFIG_CGROUP_SCHED
 enum cpu_cftype_id {
-#ifdef CONFIG_FAIR_GROUP_SCHED
+#if defined(CONFIG_FAIR_GROUP_SCHED) || defined(CONFIG_EXT_GROUP_SCHED)
 	CPU_CFTYPE_WEIGHT,
 	CPU_CFTYPE_WEIGHT_NICE,
 	CPU_CFTYPE_IDLE,
diff --git a/tools/sched_ext/scx_common.bpf.h b/tools/sched_ext/scx_common.bpf.h
index 1a0184024eaf..4f8e447f66d3 100644
--- a/tools/sched_ext/scx_common.bpf.h
+++ b/tools/sched_ext/scx_common.bpf.h
@@ -69,6 +69,7 @@ void scx_bpf_put_idle_cpumask(const struct cpumask *cpumask) __ksym;
 void scx_bpf_destroy_dsq(u64 dsq_id) __ksym;
 bool scx_bpf_task_running(const struct task_struct *p) __ksym;
 s32 scx_bpf_task_cpu(const struct task_struct *p) __ksym;
+struct cgroup *scx_bpf_task_cgroup(struct task_struct *p) __ksym;
 
 #define BPF_STRUCT_OPS(name, args...)						\
 SEC("struct_ops/"#name)								\
-- 
2.41.0

[PATCH 25/34] sched_ext: Add a cgroup-based core-scheduling scheduler

[Tejun Heo]

This patch adds scx_pair example scheduler which implements a variant of
core scheduling where a hyperthread pair only run tasks from the same
cgroup. The BPF scheduler achieves this by putting tasks into per-cgroup
queues, time-slicing the cgroup to run for each pair first, and then
scheduling within the cgroup. See the header comment in scx_pair.bpf.c for
more details.

Note that scx_pair's cgroup-boundary guarantee breaks down for tasks running
in higher priority scheduler classes. This will be addressed by a followup
patch which implements a mechanism to track CPU preemption.

v2: * Improved stride parameter input verification.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 tools/sched_ext/.gitignore     |   1 +
 tools/sched_ext/Makefile       |   8 +-
 tools/sched_ext/scx_pair.bpf.c | 536 +++++++++++++++++++++++++++++++++
 tools/sched_ext/scx_pair.c     | 164 ++++++++++
 tools/sched_ext/scx_pair.h     |  10 +
 5 files changed, 717 insertions(+), 2 deletions(-)
 create mode 100644 tools/sched_ext/scx_pair.bpf.c
 create mode 100644 tools/sched_ext/scx_pair.c
 create mode 100644 tools/sched_ext/scx_pair.h

diff --git a/tools/sched_ext/.gitignore b/tools/sched_ext/.gitignore
index 7e5dec30a87e..4659a15b8daf 100644
--- a/tools/sched_ext/.gitignore
+++ b/tools/sched_ext/.gitignore
@@ -1,6 +1,7 @@
 scx_simple
 scx_qmap
 scx_central
+scx_pair
 *.skel.h
 *.subskel.h
 /tools/
diff --git a/tools/sched_ext/Makefile b/tools/sched_ext/Makefile
index be0445e071ef..f3f8f083de16 100644
--- a/tools/sched_ext/Makefile
+++ b/tools/sched_ext/Makefile
@@ -115,7 +115,7 @@ BPF_CFLAGS = -g -D__TARGET_ARCH_$(SRCARCH)					\
 	     -Wall -Wno-compare-distinct-pointer-types				\
 	     -O2 -mcpu=v3
 
-all: scx_simple scx_qmap scx_central
+all: scx_simple scx_qmap scx_central scx_pair
 
 # sort removes libbpf duplicates when not cross-building
 MAKE_DIRS := $(sort $(BUILD_DIR)/libbpf $(HOST_BUILD_DIR)/libbpf		\
@@ -178,10 +178,14 @@ scx_central: scx_central.c scx_central.skel.h user_exit_info.h
 	$(CC) $(CFLAGS) -c $< -o $@.o
 	$(CC) -o $@ $@.o $(HOST_BPFOBJ) $(LDFLAGS)
 
+scx_pair: scx_pair.c scx_pair.skel.h user_exit_info.h
+	$(CC) $(CFLAGS) -c $< -o $@.o
+	$(CC) -o $@ $@.o $(HOST_BPFOBJ) $(LDFLAGS)
+
 clean:
 	rm -rf $(SCRATCH_DIR) $(HOST_SCRATCH_DIR)
 	rm -f *.o *.bpf.o *.skel.h *.subskel.h
-	rm -f scx_simple scx_qmap scx_central
+	rm -f scx_simple scx_qmap scx_central scx_pair scx_flatcg
 
 .PHONY: all clean
 
diff --git a/tools/sched_ext/scx_pair.bpf.c b/tools/sched_ext/scx_pair.bpf.c
new file mode 100644
index 000000000000..f980d301dbbe
--- /dev/null
+++ b/tools/sched_ext/scx_pair.bpf.c
@@ -0,0 +1,536 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * A demo sched_ext core-scheduler which always makes every sibling CPU pair
+ * execute from the same CPU cgroup.
+ *
+ * This scheduler is a minimal implementation and would need some form of
+ * priority handling both inside each cgroup and across the cgroups to be
+ * practically useful.
+ *
+ * Each CPU in the system is paired with exactly one other CPU, according to a
+ * "stride" value that can be specified when the BPF scheduler program is first
+ * loaded. Throughout the runtime of the scheduler, these CPU pairs guarantee
+ * that they will only ever schedule tasks that belong to the same CPU cgroup.
+ *
+ * Scheduler Initialization
+ * ------------------------
+ *
+ * The scheduler BPF program is first initialized from user space, before it is
+ * enabled. During this initialization process, each CPU on the system is
+ * assigned several values that are constant throughout its runtime:
+ *
+ * 1. *Pair CPU*: The CPU that it synchronizes with when making scheduling
+ *		  decisions. Paired CPUs always schedule tasks from the same
+ *		  CPU cgroup, and synchronize with each other to guarantee
+ *		  that this constraint is not violated.
+ * 2. *Pair ID*:  Each CPU pair is assigned a Pair ID, which is used to access
+ *		  a struct pair_ctx object that is shared between the pair.
+ * 3. *In-pair-index*: An index, 0 or 1, that is assigned to each core in the
+ *		       pair. Each struct pair_ctx has an active_mask field,
+ *		       which is a bitmap used to indicate whether each core
+ *		       in the pair currently has an actively running task.
+ *		       This index specifies which entry in the bitmap corresponds
+ *		       to each CPU in the pair.
+ *
+ * During this initialization, the CPUs are paired according to a "stride" that
+ * may be specified when invoking the user space program that initializes and
+ * loads the scheduler. By default, the stride is 1/2 the total number of CPUs.
+ *
+ * Tasks and cgroups
+ * -----------------
+ *
+ * Every cgroup in the system is registered with the scheduler using the
+ * pair_cgroup_init() callback, and every task in the system is associated with
+ * exactly one cgroup. At a high level, the idea with the pair scheduler is to
+ * always schedule tasks from the same cgroup within a given CPU pair. When a
+ * task is enqueued (i.e. passed to the pair_enqueue() callback function), its
+ * cgroup ID is read from its task struct, and then a corresponding queue map
+ * is used to FIFO-enqueue the task for that cgroup.
+ *
+ * If you look through the implementation of the scheduler, you'll notice that
+ * there is quite a bit of complexity involved with looking up the per-cgroup
+ * FIFO queue that we enqueue tasks in. For example, there is a cgrp_q_idx_hash
+ * BPF hash map that is used to map a cgroup ID to a globally unique ID that's
+ * allocated in the BPF program. This is done because we use separate maps to
+ * store the FIFO queue of tasks, and the length of that map, per cgroup. This
+ * complexity is only present because of current deficiencies in BPF that will
+ * soon be addressed. The main point to keep in mind is that newly enqueued
+ * tasks are added to their cgroup's FIFO queue.
+ *
+ * Dispatching tasks
+ * -----------------
+ *
+ * This section will describe how enqueued tasks are dispatched and scheduled.
+ * Tasks are dispatched in pair_dispatch(), and at a high level the workflow is
+ * as follows:
+ *
+ * 1. Fetch the struct pair_ctx for the current CPU. As mentioned above, this is
+ *    the structure that's used to synchronize amongst the two pair CPUs in their
+ *    scheduling decisions. After any of the following events have occurred:
+ *
+ * - The cgroup's slice run has expired, or
+ * - The cgroup becomes empty, or
+ * - Either CPU in the pair is preempted by a higher priority scheduling class
+ *
+ * The cgroup transitions to the draining state and stops executing new tasks
+ * from the cgroup.
+ *
+ * 2. If the pair is still executing a task, mark the pair_ctx as draining, and
+ *    wait for the pair CPU to be preempted.
+ *
+ * 3. Otherwise, if the pair CPU is not running a task, we can move onto
+ *    scheduling new tasks. Pop the next cgroup id from the top_q queue.
+ *
+ * 4. Pop a task from that cgroup's FIFO task queue, and begin executing it.
+ *
+ * Note again that this scheduling behavior is simple, but the implementation
+ * is complex mostly because this it hits several BPF shortcomings and has to
+ * work around in often awkward ways. Most of the shortcomings are expected to
+ * be resolved in the near future which should allow greatly simplifying this
+ * scheduler.
+ *
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+#include "scx_common.bpf.h"
+#include "scx_pair.h"
+
+char _license[] SEC("license") = "GPL";
+
+const volatile bool switch_partial;
+
+/* !0 for veristat, set during init */
+const volatile u32 nr_cpu_ids = 64;
+
+/* a pair of CPUs stay on a cgroup for this duration */
+const volatile u32 pair_batch_dur_ns = SCX_SLICE_DFL;
+
+/* cpu ID -> pair cpu ID */
+const volatile s32 pair_cpu[MAX_CPUS] = { [0 ... MAX_CPUS - 1] = -1 };
+
+/* cpu ID -> pair_id */
+const volatile u32 pair_id[MAX_CPUS];
+
+/* CPU ID -> CPU # in the pair (0 or 1) */
+const volatile u32 in_pair_idx[MAX_CPUS];
+
+struct pair_ctx {
+	struct bpf_spin_lock	lock;
+
+	/* the cgroup the pair is currently executing */
+	u64			cgid;
+
+	/* the pair started executing the current cgroup at */
+	u64			started_at;
+
+	/* whether the current cgroup is draining */
+	bool			draining;
+
+	/* the CPUs that are currently active on the cgroup */
+	u32			active_mask;
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_ARRAY);
+	__uint(max_entries, MAX_CPUS / 2);
+	__type(key, u32);
+	__type(value, struct pair_ctx);
+} pair_ctx SEC(".maps");
+
+/* queue of cgrp_q's possibly with tasks on them */
+struct {
+	__uint(type, BPF_MAP_TYPE_QUEUE);
+	/*
+	 * Because it's difficult to build strong synchronization encompassing
+	 * multiple non-trivial operations in BPF, this queue is managed in an
+	 * opportunistic way so that we guarantee that a cgroup w/ active tasks
+	 * is always on it but possibly multiple times. Once we have more robust
+	 * synchronization constructs and e.g. linked list, we should be able to
+	 * do this in a prettier way but for now just size it big enough.
+	 */
+	__uint(max_entries, 4 * MAX_CGRPS);
+	__type(value, u64);
+} top_q SEC(".maps");
+
+/* per-cgroup q which FIFOs the tasks from the cgroup */
+struct cgrp_q {
+	__uint(type, BPF_MAP_TYPE_QUEUE);
+	__uint(max_entries, MAX_QUEUED);
+	__type(value, u32);
+};
+
+/*
+ * Ideally, we want to allocate cgrp_q and cgrq_q_len in the cgroup local
+ * storage; however, a cgroup local storage can only be accessed from the BPF
+ * progs attached to the cgroup. For now, work around by allocating array of
+ * cgrp_q's and then allocating per-cgroup indices.
+ *
+ * Another caveat: It's difficult to populate a large array of maps statically
+ * or from BPF. Initialize it from userland.
+ */
+struct {
+	__uint(type, BPF_MAP_TYPE_ARRAY_OF_MAPS);
+	__uint(max_entries, MAX_CGRPS);
+	__type(key, s32);
+	__array(values, struct cgrp_q);
+} cgrp_q_arr SEC(".maps");
+
+static u64 cgrp_q_len[MAX_CGRPS];
+
+/*
+ * This and cgrp_q_idx_hash combine into a poor man's IDR. This likely would be
+ * useful to have as a map type.
+ */
+static u32 cgrp_q_idx_cursor;
+static u64 cgrp_q_idx_busy[MAX_CGRPS];
+
+/*
+ * All added up, the following is what we do:
+ *
+ * 1. When a cgroup is enabled, RR cgroup_q_idx_busy array doing cmpxchg looking
+ *    for a free ID. If not found, fail cgroup creation with -EBUSY.
+ *
+ * 2. Hash the cgroup ID to the allocated cgrp_q_idx in the following
+ *    cgrp_q_idx_hash.
+ *
+ * 3. Whenever a cgrp_q needs to be accessed, first look up the cgrp_q_idx from
+ *    cgrp_q_idx_hash and then access the corresponding entry in cgrp_q_arr.
+ *
+ * This is sadly complicated for something pretty simple. Hopefully, we should
+ * be able to simplify in the future.
+ */
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(max_entries, MAX_CGRPS);
+	__uint(key_size, sizeof(u64));		/* cgrp ID */
+	__uint(value_size, sizeof(s32));	/* cgrp_q idx */
+} cgrp_q_idx_hash SEC(".maps");
+
+/* statistics */
+u64 nr_total, nr_dispatched, nr_missing, nr_kicks, nr_preemptions;
+u64 nr_exps, nr_exp_waits, nr_exp_empty;
+u64 nr_cgrp_next, nr_cgrp_coll, nr_cgrp_empty;
+
+struct user_exit_info uei;
+
+static bool time_before(u64 a, u64 b)
+{
+	return (s64)(a - b) < 0;
+}
+
+void BPF_STRUCT_OPS(pair_enqueue, struct task_struct *p, u64 enq_flags)
+{
+	struct cgroup *cgrp;
+	struct cgrp_q *cgq;
+	s32 pid = p->pid;
+	u64 cgid;
+	u32 *q_idx;
+	u64 *cgq_len;
+
+	__sync_fetch_and_add(&nr_total, 1);
+
+	cgrp = scx_bpf_task_cgroup(p);
+	cgid = cgrp->kn->id;
+	bpf_cgroup_release(cgrp);
+
+	/* find the cgroup's q and push @p into it */
+	q_idx = bpf_map_lookup_elem(&cgrp_q_idx_hash, &cgid);
+	if (!q_idx) {
+		scx_bpf_error("failed to lookup q_idx for cgroup[%llu]", cgid);
+		return;
+	}
+
+	cgq = bpf_map_lookup_elem(&cgrp_q_arr, q_idx);
+	if (!cgq) {
+		scx_bpf_error("failed to lookup q_arr for cgroup[%llu] q_idx[%u]",
+			      cgid, *q_idx);
+		return;
+	}
+
+	if (bpf_map_push_elem(cgq, &pid, 0)) {
+		scx_bpf_error("cgroup[%llu] queue overflow", cgid);
+		return;
+	}
+
+	/* bump q len, if going 0 -> 1, queue cgroup into the top_q */
+	cgq_len = MEMBER_VPTR(cgrp_q_len, [*q_idx]);
+	if (!cgq_len) {
+		scx_bpf_error("MEMBER_VTPR malfunction");
+		return;
+	}
+
+	if (!__sync_fetch_and_add(cgq_len, 1) &&
+	    bpf_map_push_elem(&top_q, &cgid, 0)) {
+		scx_bpf_error("top_q overflow");
+		return;
+	}
+}
+
+static int lookup_pairc_and_mask(s32 cpu, struct pair_ctx **pairc, u32 *mask)
+{
+	u32 *vptr;
+
+	vptr = (u32 *)MEMBER_VPTR(pair_id, [cpu]);
+	if (!vptr)
+		return -EINVAL;
+
+	*pairc = bpf_map_lookup_elem(&pair_ctx, vptr);
+	if (!(*pairc))
+		return -EINVAL;
+
+	vptr = (u32 *)MEMBER_VPTR(in_pair_idx, [cpu]);
+	if (!vptr)
+		return -EINVAL;
+
+	*mask = 1U << *vptr;
+
+	return 0;
+}
+
+static int try_dispatch(s32 cpu)
+{
+	struct pair_ctx *pairc;
+	struct bpf_map *cgq_map;
+	struct task_struct *p;
+	u64 now = bpf_ktime_get_ns();
+	bool kick_pair = false;
+	bool expired;
+	u32 *vptr, in_pair_mask;
+	s32 pid, q_idx;
+	u64 cgid;
+	int ret;
+
+	ret = lookup_pairc_and_mask(cpu, &pairc, &in_pair_mask);
+	if (ret) {
+		scx_bpf_error("failed to lookup pairc and in_pair_mask for cpu[%d]",
+			      cpu);
+		return -ENOENT;
+	}
+
+	bpf_spin_lock(&pairc->lock);
+	pairc->active_mask &= ~in_pair_mask;
+
+	expired = time_before(pairc->started_at + pair_batch_dur_ns, now);
+	if (expired || pairc->draining) {
+		u64 new_cgid = 0;
+
+		__sync_fetch_and_add(&nr_exps, 1);
+
+		/*
+		 * We're done with the current cgid. An obvious optimization
+		 * would be not draining if the next cgroup is the current one.
+		 * For now, be dumb and always expire.
+		 */
+		pairc->draining = true;
+
+		if (pairc->active_mask) {
+			/*
+			 * The other CPU is still active We want to wait until
+			 * this cgroup expires.
+			 *
+			 * If the pair controls its CPU, and the time already
+			 * expired, kick.  When the other CPU arrives at
+			 * dispatch and clears its active mask, it'll push the
+			 * pair to the next cgroup and kick this CPU.
+			 */
+			__sync_fetch_and_add(&nr_exp_waits, 1);
+			bpf_spin_unlock(&pairc->lock);
+			if (expired)
+				kick_pair = true;
+			goto out_maybe_kick;
+		}
+
+		bpf_spin_unlock(&pairc->lock);
+
+		/*
+		 * Pick the next cgroup. It'd be easier / cleaner to not drop
+		 * pairc->lock and use stronger synchronization here especially
+		 * given that we'll be switching cgroups significantly less
+		 * frequently than tasks. Unfortunately, bpf_spin_lock can't
+		 * really protect anything non-trivial. Let's do opportunistic
+		 * operations instead.
+		 */
+		bpf_repeat(BPF_MAX_LOOPS) {
+			u32 *q_idx;
+			u64 *cgq_len;
+
+			if (bpf_map_pop_elem(&top_q, &new_cgid)) {
+				/* no active cgroup, go idle */
+				__sync_fetch_and_add(&nr_exp_empty, 1);
+				return 0;
+			}
+
+			q_idx = bpf_map_lookup_elem(&cgrp_q_idx_hash, &new_cgid);
+			if (!q_idx)
+				continue;
+
+			/*
+			 * This is the only place where empty cgroups are taken
+			 * off the top_q.
+			 */
+			cgq_len = MEMBER_VPTR(cgrp_q_len, [*q_idx]);
+			if (!cgq_len || !*cgq_len)
+				continue;
+
+			/*
+			 * If it has any tasks, requeue as we may race and not
+			 * execute it.
+			 */
+			bpf_map_push_elem(&top_q, &new_cgid, 0);
+			break;
+		}
+
+		bpf_spin_lock(&pairc->lock);
+
+		/*
+		 * The other CPU may already have started on a new cgroup while
+		 * we dropped the lock. Make sure that we're still draining and
+		 * start on the new cgroup.
+		 */
+		if (pairc->draining && !pairc->active_mask) {
+			__sync_fetch_and_add(&nr_cgrp_next, 1);
+			pairc->cgid = new_cgid;
+			pairc->started_at = now;
+			pairc->draining = false;
+			kick_pair = true;
+		} else {
+			__sync_fetch_and_add(&nr_cgrp_coll, 1);
+		}
+	}
+
+	cgid = pairc->cgid;
+	pairc->active_mask |= in_pair_mask;
+	bpf_spin_unlock(&pairc->lock);
+
+	/* again, it'd be better to do all these with the lock held, oh well */
+	vptr = bpf_map_lookup_elem(&cgrp_q_idx_hash, &cgid);
+	if (!vptr) {
+		scx_bpf_error("failed to lookup q_idx for cgroup[%llu]", cgid);
+		return -ENOENT;
+	}
+	q_idx = *vptr;
+
+	/* claim one task from cgrp_q w/ q_idx */
+	bpf_repeat(BPF_MAX_LOOPS) {
+		u64 *cgq_len, len;
+
+		cgq_len = MEMBER_VPTR(cgrp_q_len, [q_idx]);
+		if (!cgq_len || !(len = *(volatile u64 *)cgq_len)) {
+			/* the cgroup must be empty, expire and repeat */
+			__sync_fetch_and_add(&nr_cgrp_empty, 1);
+			bpf_spin_lock(&pairc->lock);
+			pairc->draining = true;
+			pairc->active_mask &= ~in_pair_mask;
+			bpf_spin_unlock(&pairc->lock);
+			return -EAGAIN;
+		}
+
+		if (__sync_val_compare_and_swap(cgq_len, len, len - 1) != len)
+			continue;
+
+		break;
+	}
+
+	cgq_map = bpf_map_lookup_elem(&cgrp_q_arr, &q_idx);
+	if (!cgq_map) {
+		scx_bpf_error("failed to lookup cgq_map for cgroup[%llu] q_idx[%d]",
+			      cgid, q_idx);
+		return -ENOENT;
+	}
+
+	if (bpf_map_pop_elem(cgq_map, &pid)) {
+		scx_bpf_error("cgq_map is empty for cgroup[%llu] q_idx[%d]",
+			      cgid, q_idx);
+		return -ENOENT;
+	}
+
+	p = bpf_task_from_pid(pid);
+	if (p) {
+		__sync_fetch_and_add(&nr_dispatched, 1);
+		scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, SCX_SLICE_DFL, 0);
+		bpf_task_release(p);
+	} else {
+		/* we don't handle dequeues, retry on lost tasks */
+		__sync_fetch_and_add(&nr_missing, 1);
+		return -EAGAIN;
+	}
+
+out_maybe_kick:
+	if (kick_pair) {
+		s32 *pair = (s32 *)MEMBER_VPTR(pair_cpu, [cpu]);
+		if (pair) {
+			__sync_fetch_and_add(&nr_kicks, 1);
+			scx_bpf_kick_cpu(*pair, SCX_KICK_PREEMPT);
+		}
+	}
+	return 0;
+}
+
+void BPF_STRUCT_OPS(pair_dispatch, s32 cpu, struct task_struct *prev)
+{
+	bpf_repeat(BPF_MAX_LOOPS) {
+		if (try_dispatch(cpu) != -EAGAIN)
+			break;
+	}
+}
+
+s32 BPF_STRUCT_OPS(pair_cgroup_init, struct cgroup *cgrp)
+{
+	u64 cgid = cgrp->kn->id;
+	s32 i, q_idx;
+
+	bpf_for(i, 0, MAX_CGRPS) {
+		q_idx = __sync_fetch_and_add(&cgrp_q_idx_cursor, 1) % MAX_CGRPS;
+		if (!__sync_val_compare_and_swap(&cgrp_q_idx_busy[q_idx], 0, 1))
+			break;
+	}
+	if (i == MAX_CGRPS)
+		return -EBUSY;
+
+	if (bpf_map_update_elem(&cgrp_q_idx_hash, &cgid, &q_idx, BPF_ANY)) {
+		u64 *busy = MEMBER_VPTR(cgrp_q_idx_busy, [q_idx]);
+		if (busy)
+			*busy = 0;
+		return -EBUSY;
+	}
+
+	return 0;
+}
+
+void BPF_STRUCT_OPS(pair_cgroup_exit, struct cgroup *cgrp)
+{
+	u64 cgid = cgrp->kn->id;
+	s32 *q_idx;
+
+	q_idx = bpf_map_lookup_elem(&cgrp_q_idx_hash, &cgid);
+	if (q_idx) {
+		u64 *busy = MEMBER_VPTR(cgrp_q_idx_busy, [*q_idx]);
+		if (busy)
+			*busy = 0;
+		bpf_map_delete_elem(&cgrp_q_idx_hash, &cgid);
+	}
+}
+
+s32 BPF_STRUCT_OPS(pair_init)
+{
+	if (!switch_partial)
+		scx_bpf_switch_all();
+	return 0;
+}
+
+void BPF_STRUCT_OPS(pair_exit, struct scx_exit_info *ei)
+{
+	uei_record(&uei, ei);
+}
+
+SEC(".struct_ops.link")
+struct sched_ext_ops pair_ops = {
+	.enqueue		= (void *)pair_enqueue,
+	.dispatch		= (void *)pair_dispatch,
+	.cgroup_init		= (void *)pair_cgroup_init,
+	.cgroup_exit		= (void *)pair_cgroup_exit,
+	.init			= (void *)pair_init,
+	.exit			= (void *)pair_exit,
+	.name			= "pair",
+};
diff --git a/tools/sched_ext/scx_pair.c b/tools/sched_ext/scx_pair.c
new file mode 100644
index 000000000000..4d24fcedc2cd
--- /dev/null
+++ b/tools/sched_ext/scx_pair.c
@@ -0,0 +1,164 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <unistd.h>
+#include <signal.h>
+#include <assert.h>
+#include <libgen.h>
+#include <bpf/bpf.h>
+#include "user_exit_info.h"
+#include "scx_pair.h"
+#include "scx_pair.skel.h"
+
+const char help_fmt[] =
+"A demo sched_ext core-scheduler which always makes every sibling CPU pair\n"
+"execute from the same CPU cgroup.\n"
+"\n"
+"See the top-level comment in .bpf.c for more details.\n"
+"\n"
+"Usage: %s [-S STRIDE] [-p]\n"
+"\n"
+"  -S STRIDE     Override CPU pair stride (default: nr_cpus_ids / 2)\n"
+"  -p            Switch only tasks on SCHED_EXT policy intead of all\n"
+"  -h            Display this help and exit\n";
+
+static volatile int exit_req;
+
+static void sigint_handler(int dummy)
+{
+	exit_req = 1;
+}
+
+int main(int argc, char **argv)
+{
+	struct scx_pair *skel;
+	struct bpf_link *link;
+	u64 seq = 0;
+	s32 stride, i, opt, outer_fd;
+
+	signal(SIGINT, sigint_handler);
+	signal(SIGTERM, sigint_handler);
+
+	libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
+
+	skel = scx_pair__open();
+	assert(skel);
+
+	skel->rodata->nr_cpu_ids = libbpf_num_possible_cpus();
+
+	/* pair up the earlier half to the latter by default, override with -s */
+	stride = skel->rodata->nr_cpu_ids / 2;
+
+	while ((opt = getopt(argc, argv, "S:ph")) != -1) {
+		switch (opt) {
+		case 'S':
+			stride = strtoul(optarg, NULL, 0);
+			break;
+		case 'p':
+			skel->rodata->switch_partial = true;
+			break;
+		default:
+			fprintf(stderr, help_fmt, basename(argv[0]));
+			return opt != 'h';
+		}
+	}
+
+	printf("Pairs: ");
+	for (i = 0; i < skel->rodata->nr_cpu_ids; i++) {
+		int j = (i + stride) % skel->rodata->nr_cpu_ids;
+
+		if (skel->rodata->pair_cpu[i] >= 0)
+			continue;
+
+		if (i == j) {
+			printf("\n");
+			fprintf(stderr, "Invalid stride %d - CPU%d wants to be its own pair\n",
+				stride, i);
+			return 1;
+		}
+
+		if (skel->rodata->pair_cpu[j] >= 0) {
+			printf("\n");
+			fprintf(stderr, "Invalid stride %d - three CPUs (%d, %d, %d) want to be a pair\n",
+				stride, i, j, skel->rodata->pair_cpu[j]);
+			return 1;
+		}
+
+		skel->rodata->pair_cpu[i] = j;
+		skel->rodata->pair_cpu[j] = i;
+		skel->rodata->pair_id[i] = i;
+		skel->rodata->pair_id[j] = i;
+		skel->rodata->in_pair_idx[i] = 0;
+		skel->rodata->in_pair_idx[j] = 1;
+
+		printf("[%d, %d] ", i, j);
+	}
+	printf("\n");
+
+	assert(!scx_pair__load(skel));
+
+	/*
+	 * Populate the cgrp_q_arr map which is an array containing per-cgroup
+	 * queues. It'd probably be better to do this from BPF but there are too
+	 * many to initialize statically and there's no way to dynamically
+	 * populate from BPF.
+	 */
+	outer_fd = bpf_map__fd(skel->maps.cgrp_q_arr);
+	assert(outer_fd >= 0);
+
+	printf("Initializing");
+        for (i = 0; i < MAX_CGRPS; i++) {
+		s32 inner_fd;
+
+		if (exit_req)
+			break;
+
+		inner_fd = bpf_map_create(BPF_MAP_TYPE_QUEUE, NULL, 0,
+					  sizeof(u32), MAX_QUEUED, NULL);
+		assert(inner_fd >= 0);
+		assert(!bpf_map_update_elem(outer_fd, &i, &inner_fd, BPF_ANY));
+		close(inner_fd);
+
+		if (!(i % 10))
+			printf(".");
+		fflush(stdout);
+        }
+	printf("\n");
+
+	/*
+	 * Fully initialized, attach and run.
+	 */
+	link = bpf_map__attach_struct_ops(skel->maps.pair_ops);
+	assert(link);
+
+	while (!exit_req && !uei_exited(&skel->bss->uei)) {
+		printf("[SEQ %lu]\n", seq++);
+		printf(" total:%10lu dispatch:%10lu   missing:%10lu\n",
+		       skel->bss->nr_total,
+		       skel->bss->nr_dispatched,
+		       skel->bss->nr_missing);
+		printf(" kicks:%10lu preemptions:%7lu\n",
+		       skel->bss->nr_kicks,
+		       skel->bss->nr_preemptions);
+		printf("   exp:%10lu exp_wait:%10lu exp_empty:%10lu\n",
+		       skel->bss->nr_exps,
+		       skel->bss->nr_exp_waits,
+		       skel->bss->nr_exp_empty);
+		printf("cgnext:%10lu   cgcoll:%10lu   cgempty:%10lu\n",
+		       skel->bss->nr_cgrp_next,
+		       skel->bss->nr_cgrp_coll,
+		       skel->bss->nr_cgrp_empty);
+		fflush(stdout);
+		sleep(1);
+	}
+
+	bpf_link__destroy(link);
+	uei_print(&skel->bss->uei);
+	scx_pair__destroy(skel);
+	return 0;
+}
diff --git a/tools/sched_ext/scx_pair.h b/tools/sched_ext/scx_pair.h
new file mode 100644
index 000000000000..f60b824272f7
--- /dev/null
+++ b/tools/sched_ext/scx_pair.h
@@ -0,0 +1,10 @@
+#ifndef __SCX_EXAMPLE_PAIR_H
+#define __SCX_EXAMPLE_PAIR_H
+
+enum {
+	MAX_CPUS		= 4096,
+	MAX_QUEUED		= 4096,
+	MAX_CGRPS		= 4096,
+};
+
+#endif /* __SCX_EXAMPLE_PAIR_H */
-- 
2.41.0

[PATCH 26/34] sched_ext: Add a cgroup scheduler which uses flattened hierarchy

[Tejun Heo]

This patch adds scx_flatcg example scheduler which implements hierarchical
weight-based cgroup CPU control by flattening the cgroup hierarchy into a
single layer by compounding the active weight share at each level.

This flattening of hierarchy can bring a substantial performance gain when
the cgroup hierarchy is nested multiple levels. in a simple benchmark using
wrk[8] on apache serving a CGI script calculating sha1sum of a small file,
it outperforms CFS by ~3% with CPU controller disabled and by ~10% with two
apache instances competing with 2:1 weight ratio nested four level deep.

However, the gain comes at the cost of not being able to properly handle
thundering herd of cgroups. For example, if many cgroups which are nested
behind a low priority parent cgroup wake up around the same time, they may
be able to consume more CPU cycles than they are entitled to. In many use
cases, this isn't a real concern especially given the performance gain.
Also, there are ways to mitigate the problem further by e.g. introducing an
extra scheduling layer on cgroup delegation boundaries.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 tools/sched_ext/.gitignore       |   1 +
 tools/sched_ext/Makefile         |   6 +-
 tools/sched_ext/scx_flatcg.bpf.c | 834 +++++++++++++++++++++++++++++++
 tools/sched_ext/scx_flatcg.c     | 228 +++++++++
 tools/sched_ext/scx_flatcg.h     |  49 ++
 5 files changed, 1117 insertions(+), 1 deletion(-)
 create mode 100644 tools/sched_ext/scx_flatcg.bpf.c
 create mode 100644 tools/sched_ext/scx_flatcg.c
 create mode 100644 tools/sched_ext/scx_flatcg.h

diff --git a/tools/sched_ext/.gitignore b/tools/sched_ext/.gitignore
index 4659a15b8daf..d5a4923919ce 100644
--- a/tools/sched_ext/.gitignore
+++ b/tools/sched_ext/.gitignore
@@ -2,6 +2,7 @@ scx_simple
 scx_qmap
 scx_central
 scx_pair
+scx_flatcg
 *.skel.h
 *.subskel.h
 /tools/
diff --git a/tools/sched_ext/Makefile b/tools/sched_ext/Makefile
index f3f8f083de16..a0dacea5993c 100644
--- a/tools/sched_ext/Makefile
+++ b/tools/sched_ext/Makefile
@@ -115,7 +115,7 @@ BPF_CFLAGS = -g -D__TARGET_ARCH_$(SRCARCH)					\
 	     -Wall -Wno-compare-distinct-pointer-types				\
 	     -O2 -mcpu=v3
 
-all: scx_simple scx_qmap scx_central scx_pair
+all: scx_simple scx_qmap scx_central scx_pair scx_flatcg
 
 # sort removes libbpf duplicates when not cross-building
 MAKE_DIRS := $(sort $(BUILD_DIR)/libbpf $(HOST_BUILD_DIR)/libbpf		\
@@ -182,6 +182,10 @@ scx_pair: scx_pair.c scx_pair.skel.h user_exit_info.h
 	$(CC) $(CFLAGS) -c $< -o $@.o
 	$(CC) -o $@ $@.o $(HOST_BPFOBJ) $(LDFLAGS)
 
+scx_flatcg: scx_flatcg.c scx_flatcg.skel.h user_exit_info.h
+	$(CC) $(CFLAGS) -c $< -o $@.o
+	$(CC) -o $@ $@.o $(HOST_BPFOBJ) $(LDFLAGS)
+
 clean:
 	rm -rf $(SCRATCH_DIR) $(HOST_SCRATCH_DIR)
 	rm -f *.o *.bpf.o *.skel.h *.subskel.h
diff --git a/tools/sched_ext/scx_flatcg.bpf.c b/tools/sched_ext/scx_flatcg.bpf.c
new file mode 100644
index 000000000000..ab7cff4da7da
--- /dev/null
+++ b/tools/sched_ext/scx_flatcg.bpf.c
@@ -0,0 +1,834 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * A demo sched_ext flattened cgroup hierarchy scheduler. It implements
+ * hierarchical weight-based cgroup CPU control by flattening the cgroup
+ * hierarchy into a single layer by compounding the active weight share at each
+ * level. Consider the following hierarchy with weights in parentheses:
+ *
+ * R + A (100) + B (100)
+ *   |         \ C (100)
+ *   \ D (200)
+ *
+ * Ignoring the root and threaded cgroups, only B, C and D can contain tasks.
+ * Let's say all three have runnable tasks. The total share that each of these
+ * three cgroups is entitled to can be calculated by compounding its share at
+ * each level.
+ *
+ * For example, B is competing against C and in that competition its share is
+ * 100/(100+100) == 1/2. At its parent level, A is competing against D and A's
+ * share in that competition is 200/(200+100) == 1/3. B's eventual share in the
+ * system can be calculated by multiplying the two shares, 1/2 * 1/3 == 1/6. C's
+ * eventual shaer is the same at 1/6. D is only competing at the top level and
+ * its share is 200/(100+200) == 2/3.
+ *
+ * So, instead of hierarchically scheduling level-by-level, we can consider it
+ * as B, C and D competing each other with respective share of 1/6, 1/6 and 2/3
+ * and keep updating the eventual shares as the cgroups' runnable states change.
+ *
+ * This flattening of hierarchy can bring a substantial performance gain when
+ * the cgroup hierarchy is nested multiple levels. in a simple benchmark using
+ * wrk[8] on apache serving a CGI script calculating sha1sum of a small file, it
+ * outperforms CFS by ~3% with CPU controller disabled and by ~10% with two
+ * apache instances competing with 2:1 weight ratio nested four level deep.
+ *
+ * However, the gain comes at the cost of not being able to properly handle
+ * thundering herd of cgroups. For example, if many cgroups which are nested
+ * behind a low priority parent cgroup wake up around the same time, they may be
+ * able to consume more CPU cycles than they are entitled to. In many use cases,
+ * this isn't a real concern especially given the performance gain. Also, there
+ * are ways to mitigate the problem further by e.g. introducing an extra
+ * scheduling layer on cgroup delegation boundaries.
+ */
+#include "scx_common.bpf.h"
+#include "user_exit_info.h"
+#include "scx_flatcg.h"
+
+char _license[] SEC("license") = "GPL";
+
+const volatile u32 nr_cpus = 32;	/* !0 for veristat, set during init */
+const volatile u64 cgrp_slice_ns = SCX_SLICE_DFL;
+const volatile bool switch_partial;
+
+u64 cvtime_now;
+struct user_exit_info uei;
+
+struct {
+	__uint(type, BPF_MAP_TYPE_PERCPU_ARRAY);
+	__type(key, u32);
+	__type(value, u64);
+	__uint(max_entries, FCG_NR_STATS);
+} stats SEC(".maps");
+
+static void stat_inc(enum fcg_stat_idx idx)
+{
+	u32 idx_v = idx;
+
+	u64 *cnt_p = bpf_map_lookup_elem(&stats, &idx_v);
+	if (cnt_p)
+		(*cnt_p)++;
+}
+
+struct fcg_cpu_ctx {
+	u64			cur_cgid;
+	u64			cur_at;
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_PERCPU_ARRAY);
+	__type(key, u32);
+	__type(value, struct fcg_cpu_ctx);
+	__uint(max_entries, 1);
+} cpu_ctx SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_CGRP_STORAGE);
+	__uint(map_flags, BPF_F_NO_PREALLOC);
+	__type(key, int);
+	__type(value, struct fcg_cgrp_ctx);
+} cgrp_ctx SEC(".maps");
+
+struct cgv_node {
+	struct bpf_rb_node	rb_node;
+	__u64			cvtime;
+	__u64			cgid;
+};
+
+private(CGV_TREE) struct bpf_spin_lock cgv_tree_lock;
+private(CGV_TREE) struct bpf_rb_root cgv_tree __contains(cgv_node, rb_node);
+
+struct cgv_node_stash {
+	struct cgv_node __kptr *node;
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(max_entries, 16384);
+	__type(key, __u64);
+	__type(value, struct cgv_node_stash);
+} cgv_node_stash SEC(".maps");
+
+struct fcg_task_ctx {
+	u64		bypassed_at;
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_TASK_STORAGE);
+	__uint(map_flags, BPF_F_NO_PREALLOC);
+	__type(key, int);
+	__type(value, struct fcg_task_ctx);
+} task_ctx SEC(".maps");
+
+/* gets inc'd on weight tree changes to expire the cached hweights */
+unsigned long hweight_gen = 1;
+
+static u64 div_round_up(u64 dividend, u64 divisor)
+{
+	return (dividend + divisor - 1) / divisor;
+}
+
+static bool vtime_before(u64 a, u64 b)
+{
+	return (s64)(a - b) < 0;
+}
+
+static bool cgv_node_less(struct bpf_rb_node *a, const struct bpf_rb_node *b)
+{
+	struct cgv_node *cgc_a, *cgc_b;
+
+	cgc_a = container_of(a, struct cgv_node, rb_node);
+	cgc_b = container_of(b, struct cgv_node, rb_node);
+
+	return cgc_a->cvtime < cgc_b->cvtime;
+}
+
+static struct fcg_cpu_ctx *find_cpu_ctx(void)
+{
+	struct fcg_cpu_ctx *cpuc;
+	u32 idx = 0;
+
+	cpuc = bpf_map_lookup_elem(&cpu_ctx, &idx);
+	if (!cpuc) {
+		scx_bpf_error("cpu_ctx lookup failed");
+		return NULL;
+	}
+	return cpuc;
+}
+
+static struct fcg_cgrp_ctx *find_cgrp_ctx(struct cgroup *cgrp)
+{
+	struct fcg_cgrp_ctx *cgc;
+
+	cgc = bpf_cgrp_storage_get(&cgrp_ctx, cgrp, 0, 0);
+	if (!cgc) {
+		scx_bpf_error("cgrp_ctx lookup failed for cgid %llu", cgrp->kn->id);
+		return NULL;
+	}
+	return cgc;
+}
+
+static struct fcg_cgrp_ctx *find_ancestor_cgrp_ctx(struct cgroup *cgrp, int level)
+{
+	struct fcg_cgrp_ctx *cgc;
+
+	cgrp = bpf_cgroup_ancestor(cgrp, level);
+	if (!cgrp) {
+		scx_bpf_error("ancestor cgroup lookup failed");
+		return NULL;
+	}
+
+	cgc = find_cgrp_ctx(cgrp);
+	if (!cgc)
+		scx_bpf_error("ancestor cgrp_ctx lookup failed");
+	bpf_cgroup_release(cgrp);
+	return cgc;
+}
+
+static void cgrp_refresh_hweight(struct cgroup *cgrp, struct fcg_cgrp_ctx *cgc)
+{
+	int level;
+
+	if (!cgc->nr_active) {
+		stat_inc(FCG_STAT_HWT_SKIP);
+		return;
+	}
+
+	if (cgc->hweight_gen == hweight_gen) {
+		stat_inc(FCG_STAT_HWT_CACHE);
+		return;
+	}
+
+	stat_inc(FCG_STAT_HWT_UPDATES);
+	bpf_for(level, 0, cgrp->level + 1) {
+		struct fcg_cgrp_ctx *cgc;
+		bool is_active;
+
+		cgc = find_ancestor_cgrp_ctx(cgrp, level);
+		if (!cgc)
+			break;
+
+		if (!level) {
+			cgc->hweight = FCG_HWEIGHT_ONE;
+			cgc->hweight_gen = hweight_gen;
+		} else {
+			struct fcg_cgrp_ctx *pcgc;
+
+			pcgc = find_ancestor_cgrp_ctx(cgrp, level - 1);
+			if (!pcgc)
+				break;
+
+			/*
+			 * We can be oppotunistic here and not grab the
+			 * cgv_tree_lock and deal with the occasional races.
+			 * However, hweight updates are already cached and
+			 * relatively low-frequency. Let's just do the
+			 * straightforward thing.
+			 */
+			bpf_spin_lock(&cgv_tree_lock);
+			is_active = cgc->nr_active;
+			if (is_active) {
+				cgc->hweight_gen = pcgc->hweight_gen;
+				cgc->hweight =
+					div_round_up(pcgc->hweight * cgc->weight,
+						     pcgc->child_weight_sum);
+			}
+			bpf_spin_unlock(&cgv_tree_lock);
+
+			if (!is_active) {
+				stat_inc(FCG_STAT_HWT_RACE);
+				break;
+			}
+		}
+	}
+}
+
+static void cgrp_cap_budget(struct cgv_node *cgv_node, struct fcg_cgrp_ctx *cgc)
+{
+	u64 delta, cvtime, max_budget;
+
+	/*
+	 * A node which is on the rbtree can't be pointed to from elsewhere yet
+	 * and thus can't be updated and repositioned. Instead, we collect the
+	 * vtime deltas separately and apply it asynchronously here.
+	 */
+	delta = cgc->cvtime_delta;
+	__sync_fetch_and_sub(&cgc->cvtime_delta, delta);
+	cvtime = cgv_node->cvtime + delta;
+
+	/*
+	 * Allow a cgroup to carry the maximum budget proportional to its
+	 * hweight such that a full-hweight cgroup can immediately take up half
+	 * of the CPUs at the most while staying at the front of the rbtree.
+	 */
+	max_budget = (cgrp_slice_ns * nr_cpus * cgc->hweight) /
+		(2 * FCG_HWEIGHT_ONE);
+	if (vtime_before(cvtime, cvtime_now - max_budget))
+		cvtime = cvtime_now - max_budget;
+
+	cgv_node->cvtime = cvtime;
+}
+
+static void cgrp_enqueued(struct cgroup *cgrp, struct fcg_cgrp_ctx *cgc)
+{
+	struct cgv_node_stash *stash;
+	struct cgv_node *cgv_node;
+	u64 cgid = cgrp->kn->id;
+
+	/* paired with cmpxchg in try_pick_next_cgroup() */
+	if (__sync_val_compare_and_swap(&cgc->queued, 0, 1)) {
+		stat_inc(FCG_STAT_ENQ_SKIP);
+		return;
+	}
+
+	stash = bpf_map_lookup_elem(&cgv_node_stash, &cgid);
+	if (!stash) {
+		scx_bpf_error("cgv_node lookup failed for cgid %llu", cgid);
+		return;
+	}
+
+	/* NULL if the node is already on the rbtree */
+	cgv_node = bpf_kptr_xchg(&stash->node, NULL);
+	if (!cgv_node) {
+		stat_inc(FCG_STAT_ENQ_RACE);
+		return;
+	}
+
+	bpf_spin_lock(&cgv_tree_lock);
+	cgrp_cap_budget(cgv_node, cgc);
+	bpf_rbtree_add(&cgv_tree, &cgv_node->rb_node, cgv_node_less);
+	bpf_spin_unlock(&cgv_tree_lock);
+}
+
+void BPF_STRUCT_OPS(fcg_enqueue, struct task_struct *p, u64 enq_flags)
+{
+	struct fcg_task_ctx *taskc;
+	struct cgroup *cgrp;
+	struct fcg_cgrp_ctx *cgc;
+
+	taskc = bpf_task_storage_get(&task_ctx, p, 0, 0);
+	if (!taskc) {
+		scx_bpf_error("task_ctx lookup failed");
+		return;
+	}
+
+	/*
+	 * If select_cpu_dfl() is recommending local enqueue, the target CPU is
+	 * idle. Follow it and charge the cgroup later in fcg_stopping() after
+	 * the fact. Use the same mechanism to deal with tasks with custom
+	 * affinities so that we don't have to worry about per-cgroup dq's
+	 * containing tasks that can't be executed from some CPUs.
+	 */
+	if ((enq_flags & SCX_ENQ_LOCAL) || p->nr_cpus_allowed != nr_cpus) {
+		/*
+		 * Tell fcg_stopping() that this bypassed the regular scheduling
+		 * path and should be force charged to the cgroup. 0 is used to
+		 * indicate that the task isn't bypassing, so if the current
+		 * runtime is 0, go back by one nanosecond.
+		 */
+		taskc->bypassed_at = p->se.sum_exec_runtime ?: (u64)-1;
+
+		/*
+		 * The global dq is deprioritized as we don't want to let tasks
+		 * to boost themselves by constraining its cpumask. The
+		 * deprioritization is rather severe, so let's not apply that to
+		 * per-cpu kernel threads. This is ham-fisted. We probably wanna
+		 * implement per-cgroup fallback dq's instead so that we have
+		 * more control over when tasks with custom cpumask get issued.
+		 */
+		if ((enq_flags & SCX_ENQ_LOCAL) ||
+		    (p->nr_cpus_allowed == 1 && (p->flags & PF_KTHREAD))) {
+			stat_inc(FCG_STAT_LOCAL);
+			scx_bpf_dispatch(p, SCX_DSQ_LOCAL, SCX_SLICE_DFL, enq_flags);
+		} else {
+			stat_inc(FCG_STAT_GLOBAL);
+			scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, SCX_SLICE_DFL, enq_flags);
+		}
+		return;
+	}
+
+	cgrp = scx_bpf_task_cgroup(p);
+	cgc = find_cgrp_ctx(cgrp);
+	if (!cgc)
+		goto out_release;
+
+	scx_bpf_dispatch(p, cgrp->kn->id, SCX_SLICE_DFL, enq_flags);
+
+	cgrp_enqueued(cgrp, cgc);
+out_release:
+	bpf_cgroup_release(cgrp);
+}
+
+/*
+ * Walk the cgroup tree to update the active weight sums as tasks wake up and
+ * sleep. The weight sums are used as the base when calculating the proportion a
+ * given cgroup or task is entitled to at each level.
+ */
+static void update_active_weight_sums(struct cgroup *cgrp, bool runnable)
+{
+	struct fcg_cgrp_ctx *cgc;
+	bool updated = false;
+	int idx;
+
+	cgc = find_cgrp_ctx(cgrp);
+	if (!cgc)
+		return;
+
+	/*
+	 * In most cases, a hot cgroup would have multiple threads going to
+	 * sleep and waking up while the whole cgroup stays active. In leaf
+	 * cgroups, ->nr_runnable which is updated with __sync operations gates
+	 * ->nr_active updates, so that we don't have to grab the cgv_tree_lock
+	 * repeatedly for a busy cgroup which is staying active.
+	 */
+	if (runnable) {
+		if (__sync_fetch_and_add(&cgc->nr_runnable, 1))
+			return;
+		stat_inc(FCG_STAT_ACT);
+	} else {
+		if (__sync_sub_and_fetch(&cgc->nr_runnable, 1))
+			return;
+		stat_inc(FCG_STAT_DEACT);
+	}
+
+	/*
+	 * If @cgrp is becoming runnable, its hweight should be refreshed after
+	 * it's added to the weight tree so that enqueue has the up-to-date
+	 * value. If @cgrp is becoming quiescent, the hweight should be
+	 * refreshed before it's removed from the weight tree so that the usage
+	 * charging which happens afterwards has access to the latest value.
+	 */
+	if (!runnable)
+		cgrp_refresh_hweight(cgrp, cgc);
+
+	/* propagate upwards */
+	bpf_for(idx, 0, cgrp->level) {
+		int level = cgrp->level - idx;
+		struct fcg_cgrp_ctx *cgc, *pcgc = NULL;
+		bool propagate = false;
+
+		cgc = find_ancestor_cgrp_ctx(cgrp, level);
+		if (!cgc)
+			break;
+		if (level) {
+			pcgc = find_ancestor_cgrp_ctx(cgrp, level - 1);
+			if (!pcgc)
+				break;
+		}
+
+		/*
+		 * We need the propagation protected by a lock to synchronize
+		 * against weight changes. There's no reason to drop the lock at
+		 * each level but bpf_spin_lock() doesn't want any function
+		 * calls while locked.
+		 */
+		bpf_spin_lock(&cgv_tree_lock);
+
+		if (runnable) {
+			if (!cgc->nr_active++) {
+				updated = true;
+				if (pcgc) {
+					propagate = true;
+					pcgc->child_weight_sum += cgc->weight;
+				}
+			}
+		} else {
+			if (!--cgc->nr_active) {
+				updated = true;
+				if (pcgc) {
+					propagate = true;
+					pcgc->child_weight_sum -= cgc->weight;
+				}
+			}
+		}
+
+		bpf_spin_unlock(&cgv_tree_lock);
+
+		if (!propagate)
+			break;
+	}
+
+	if (updated)
+		__sync_fetch_and_add(&hweight_gen, 1);
+
+	if (runnable)
+		cgrp_refresh_hweight(cgrp, cgc);
+}
+
+void BPF_STRUCT_OPS(fcg_runnable, struct task_struct *p, u64 enq_flags)
+{
+	struct cgroup *cgrp;
+
+	cgrp = scx_bpf_task_cgroup(p);
+	update_active_weight_sums(cgrp, true);
+	bpf_cgroup_release(cgrp);
+}
+
+void BPF_STRUCT_OPS(fcg_stopping, struct task_struct *p, bool runnable)
+{
+	struct fcg_task_ctx *taskc;
+	struct cgroup *cgrp;
+	struct fcg_cgrp_ctx *cgc;
+
+	taskc = bpf_task_storage_get(&task_ctx, p, 0, 0);
+	if (!taskc) {
+		scx_bpf_error("task_ctx lookup failed");
+		return;
+	}
+
+	if (!taskc->bypassed_at)
+		return;
+
+	cgrp = scx_bpf_task_cgroup(p);
+	cgc = find_cgrp_ctx(cgrp);
+	if (cgc) {
+		__sync_fetch_and_add(&cgc->cvtime_delta,
+				     p->se.sum_exec_runtime - taskc->bypassed_at);
+		taskc->bypassed_at = 0;
+	}
+	bpf_cgroup_release(cgrp);
+}
+
+void BPF_STRUCT_OPS(fcg_quiescent, struct task_struct *p, u64 deq_flags)
+{
+	struct cgroup *cgrp;
+
+	cgrp = scx_bpf_task_cgroup(p);
+	update_active_weight_sums(cgrp, false);
+	bpf_cgroup_release(cgrp);
+}
+
+void BPF_STRUCT_OPS(fcg_cgroup_set_weight, struct cgroup *cgrp, u32 weight)
+{
+	struct fcg_cgrp_ctx *cgc, *pcgc = NULL;
+
+	cgc = find_cgrp_ctx(cgrp);
+	if (!cgc)
+		return;
+
+	if (cgrp->level) {
+		pcgc = find_ancestor_cgrp_ctx(cgrp, cgrp->level - 1);
+		if (!pcgc)
+			return;
+	}
+
+	bpf_spin_lock(&cgv_tree_lock);
+	if (pcgc && cgc->nr_active)
+		pcgc->child_weight_sum += (s64)weight - cgc->weight;
+	cgc->weight = weight;
+	bpf_spin_unlock(&cgv_tree_lock);
+}
+
+static bool try_pick_next_cgroup(u64 *cgidp)
+{
+	struct bpf_rb_node *rb_node;
+	struct cgv_node_stash *stash;
+	struct cgv_node *cgv_node;
+	struct fcg_cgrp_ctx *cgc;
+	struct cgroup *cgrp;
+	u64 cgid;
+
+	/* pop the front cgroup and wind cvtime_now accordingly */
+	bpf_spin_lock(&cgv_tree_lock);
+
+	rb_node = bpf_rbtree_first(&cgv_tree);
+	if (!rb_node) {
+		bpf_spin_unlock(&cgv_tree_lock);
+		stat_inc(FCG_STAT_PNC_NO_CGRP);
+		*cgidp = 0;
+		return true;
+	}
+
+	rb_node = bpf_rbtree_remove(&cgv_tree, rb_node);
+	bpf_spin_unlock(&cgv_tree_lock);
+
+	if (!rb_node) {
+		/*
+		 * This should never happen. bpf_rbtree_first() was called
+		 * above while the tree lock was held, so the node should
+		 * always be present.
+		 */
+		scx_bpf_error("node could not be removed");
+		return true;
+	}
+
+	cgv_node = container_of(rb_node, struct cgv_node, rb_node);
+	cgid = cgv_node->cgid;
+
+	if (vtime_before(cvtime_now, cgv_node->cvtime))
+		cvtime_now = cgv_node->cvtime;
+
+	/*
+	 * If lookup fails, the cgroup's gone. Free and move on. See
+	 * fcg_cgroup_exit().
+	 */
+	cgrp = bpf_cgroup_from_id(cgid);
+	if (!cgrp) {
+		stat_inc(FCG_STAT_PNC_GONE);
+		goto out_free;
+	}
+
+	cgc = bpf_cgrp_storage_get(&cgrp_ctx, cgrp, 0, 0);
+	if (!cgc) {
+		bpf_cgroup_release(cgrp);
+		stat_inc(FCG_STAT_PNC_GONE);
+		goto out_free;
+	}
+
+	if (!scx_bpf_consume(cgid)) {
+		bpf_cgroup_release(cgrp);
+		stat_inc(FCG_STAT_PNC_EMPTY);
+		goto out_stash;
+	}
+
+	/*
+	 * Successfully consumed from the cgroup. This will be our current
+	 * cgroup for the new slice. Refresh its hweight.
+	 */
+	cgrp_refresh_hweight(cgrp, cgc);
+
+	bpf_cgroup_release(cgrp);
+
+	/*
+	 * As the cgroup may have more tasks, add it back to the rbtree. Note
+	 * that here we charge the full slice upfront and then exact later
+	 * according to the actual consumption. This prevents lowpri thundering
+	 * herd from saturating the machine.
+	 */
+	bpf_spin_lock(&cgv_tree_lock);
+	cgv_node->cvtime += cgrp_slice_ns * FCG_HWEIGHT_ONE / (cgc->hweight ?: 1);
+	cgrp_cap_budget(cgv_node, cgc);
+	bpf_rbtree_add(&cgv_tree, &cgv_node->rb_node, cgv_node_less);
+	bpf_spin_unlock(&cgv_tree_lock);
+
+	*cgidp = cgid;
+	stat_inc(FCG_STAT_PNC_NEXT);
+	return true;
+
+out_stash:
+	stash = bpf_map_lookup_elem(&cgv_node_stash, &cgid);
+	if (!stash) {
+		stat_inc(FCG_STAT_PNC_GONE);
+		goto out_free;
+	}
+
+	/*
+	 * Paired with cmpxchg in cgrp_enqueued(). If they see the following
+	 * transition, they'll enqueue the cgroup. If they are earlier, we'll
+	 * see their task in the dq below and requeue the cgroup.
+	 */
+	__sync_val_compare_and_swap(&cgc->queued, 1, 0);
+
+	if (scx_bpf_dsq_nr_queued(cgid)) {
+		bpf_spin_lock(&cgv_tree_lock);
+		bpf_rbtree_add(&cgv_tree, &cgv_node->rb_node, cgv_node_less);
+		bpf_spin_unlock(&cgv_tree_lock);
+	} else {
+		cgv_node = bpf_kptr_xchg(&stash->node, cgv_node);
+		if (cgv_node) {
+			scx_bpf_error("unexpected !NULL cgv_node stash");
+			goto out_free;
+		}
+	}
+
+	return false;
+
+out_free:
+	bpf_obj_drop(cgv_node);
+	return false;
+}
+
+void BPF_STRUCT_OPS(fcg_dispatch, s32 cpu, struct task_struct *prev)
+{
+	struct fcg_cpu_ctx *cpuc;
+	struct fcg_cgrp_ctx *cgc;
+	struct cgroup *cgrp;
+	u64 now = bpf_ktime_get_ns();
+
+	cpuc = find_cpu_ctx();
+	if (!cpuc)
+		return;
+
+	if (!cpuc->cur_cgid)
+		goto pick_next_cgroup;
+
+	if (vtime_before(now, cpuc->cur_at + cgrp_slice_ns)) {
+		if (scx_bpf_consume(cpuc->cur_cgid)) {
+			stat_inc(FCG_STAT_CNS_KEEP);
+			return;
+		}
+		stat_inc(FCG_STAT_CNS_EMPTY);
+	} else {
+		stat_inc(FCG_STAT_CNS_EXPIRE);
+	}
+
+	/*
+	 * The current cgroup is expiring. It was already charged a full slice.
+	 * Calculate the actual usage and accumulate the delta.
+	 */
+	cgrp = bpf_cgroup_from_id(cpuc->cur_cgid);
+	if (!cgrp) {
+		stat_inc(FCG_STAT_CNS_GONE);
+		goto pick_next_cgroup;
+	}
+
+	cgc = bpf_cgrp_storage_get(&cgrp_ctx, cgrp, 0, 0);
+	if (cgc) {
+		/*
+		 * We want to update the vtime delta and then look for the next
+		 * cgroup to execute but the latter needs to be done in a loop
+		 * and we can't keep the lock held. Oh well...
+		 */
+		bpf_spin_lock(&cgv_tree_lock);
+		__sync_fetch_and_add(&cgc->cvtime_delta,
+				     (cpuc->cur_at + cgrp_slice_ns - now) *
+				     FCG_HWEIGHT_ONE / (cgc->hweight ?: 1));
+		bpf_spin_unlock(&cgv_tree_lock);
+	} else {
+		stat_inc(FCG_STAT_CNS_GONE);
+	}
+
+	bpf_cgroup_release(cgrp);
+
+pick_next_cgroup:
+	cpuc->cur_at = now;
+
+	if (scx_bpf_consume(SCX_DSQ_GLOBAL)) {
+		cpuc->cur_cgid = 0;
+		return;
+	}
+
+	bpf_repeat(BPF_MAX_LOOPS) {
+		if (try_pick_next_cgroup(&cpuc->cur_cgid))
+			break;
+	}
+}
+
+s32 BPF_STRUCT_OPS(fcg_prep_enable, struct task_struct *p,
+		   struct scx_enable_args *args)
+{
+	struct fcg_task_ctx *taskc;
+
+	/*
+	 * @p is new. Let's ensure that its task_ctx is available. We can sleep
+	 * in this function and the following will automatically use GFP_KERNEL.
+	 */
+	taskc = bpf_task_storage_get(&task_ctx, p, 0,
+				     BPF_LOCAL_STORAGE_GET_F_CREATE);
+	if (!taskc)
+		return -ENOMEM;
+
+	taskc->bypassed_at = 0;
+	return 0;
+}
+
+int BPF_STRUCT_OPS_SLEEPABLE(fcg_cgroup_init, struct cgroup *cgrp,
+			     struct scx_cgroup_init_args *args)
+{
+	struct fcg_cgrp_ctx *cgc;
+	struct cgv_node *cgv_node;
+	struct cgv_node_stash empty_stash = {}, *stash;
+	u64 cgid = cgrp->kn->id;
+	int ret;
+
+	/*
+	 * Technically incorrect as cgroup ID is full 64bit while dq ID is
+	 * 63bit. Should not be a problem in practice and easy to spot in the
+	 * unlikely case that it breaks.
+	 */
+	ret = scx_bpf_create_dsq(cgid, -1);
+	if (ret)
+		return ret;
+
+	cgc = bpf_cgrp_storage_get(&cgrp_ctx, cgrp, 0,
+				   BPF_LOCAL_STORAGE_GET_F_CREATE);
+	if (!cgc) {
+		ret = -ENOMEM;
+		goto err_destroy_dsq;
+	}
+
+	cgc->weight = args->weight;
+	cgc->hweight = FCG_HWEIGHT_ONE;
+
+	ret = bpf_map_update_elem(&cgv_node_stash, &cgid, &empty_stash,
+				  BPF_NOEXIST);
+	if (ret) {
+		if (ret != -ENOMEM)
+			scx_bpf_error("unexpected stash creation error (%d)",
+				      ret);
+		goto err_destroy_dsq;
+	}
+
+	stash = bpf_map_lookup_elem(&cgv_node_stash, &cgid);
+	if (!stash) {
+		scx_bpf_error("unexpected cgv_node stash lookup failure");
+		ret = -ENOENT;
+		goto err_destroy_dsq;
+	}
+
+	cgv_node = bpf_obj_new(struct cgv_node);
+	if (!cgv_node) {
+		ret = -ENOMEM;
+		goto err_del_cgv_node;
+	}
+
+	cgv_node->cgid = cgid;
+	cgv_node->cvtime = cvtime_now;
+
+	cgv_node = bpf_kptr_xchg(&stash->node, cgv_node);
+	if (cgv_node) {
+		scx_bpf_error("unexpected !NULL cgv_node stash");
+		ret = -EBUSY;
+		goto err_drop;
+	}
+
+	return 0;
+
+err_drop:
+	bpf_obj_drop(cgv_node);
+err_del_cgv_node:
+	bpf_map_delete_elem(&cgv_node_stash, &cgid);
+err_destroy_dsq:
+	scx_bpf_destroy_dsq(cgid);
+	return ret;
+}
+
+void BPF_STRUCT_OPS(fcg_cgroup_exit, struct cgroup *cgrp)
+{
+	u64 cgid = cgrp->kn->id;
+
+	/*
+	 * For now, there's no way find and remove the cgv_node if it's on the
+	 * cgv_tree. Let's drain them in the dispatch path as they get popped
+	 * off the front of the tree.
+	 */
+	bpf_map_delete_elem(&cgv_node_stash, &cgid);
+	scx_bpf_destroy_dsq(cgid);
+}
+
+s32 BPF_STRUCT_OPS(fcg_init)
+{
+	if (!switch_partial)
+		scx_bpf_switch_all();
+	return 0;
+}
+
+void BPF_STRUCT_OPS(fcg_exit, struct scx_exit_info *ei)
+{
+	uei_record(&uei, ei);
+}
+
+SEC(".struct_ops.link")
+struct sched_ext_ops flatcg_ops = {
+	.enqueue		= (void *)fcg_enqueue,
+	.dispatch		= (void *)fcg_dispatch,
+	.runnable		= (void *)fcg_runnable,
+	.stopping		= (void *)fcg_stopping,
+	.quiescent		= (void *)fcg_quiescent,
+	.prep_enable		= (void *)fcg_prep_enable,
+	.cgroup_set_weight	= (void *)fcg_cgroup_set_weight,
+	.cgroup_init		= (void *)fcg_cgroup_init,
+	.cgroup_exit		= (void *)fcg_cgroup_exit,
+	.init			= (void *)fcg_init,
+	.exit			= (void *)fcg_exit,
+	.flags			= SCX_OPS_CGROUP_KNOB_WEIGHT | SCX_OPS_ENQ_EXITING,
+	.name			= "flatcg",
+};
diff --git a/tools/sched_ext/scx_flatcg.c b/tools/sched_ext/scx_flatcg.c
new file mode 100644
index 000000000000..82afaa98d7a7
--- /dev/null
+++ b/tools/sched_ext/scx_flatcg.c
@@ -0,0 +1,228 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2023 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2023 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2023 David Vernet <dvernet@meta.com>
+ */
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <signal.h>
+#include <unistd.h>
+#include <limits.h>
+#include <fcntl.h>
+#include <time.h>
+#include <assert.h>
+#include <bpf/bpf.h>
+#include "user_exit_info.h"
+#include "scx_flatcg.h"
+#include "scx_flatcg.skel.h"
+
+#ifndef FILEID_KERNFS
+#define FILEID_KERNFS		0xfe
+#endif
+
+const char help_fmt[] =
+"A flattened cgroup hierarchy sched_ext scheduler.\n"
+"\n"
+"See the top-level comment in .bpf.c for more details.\n"
+"\n"
+"Usage: %s [-s SLICE_US] [-i INTERVAL] [-p]\n"
+"\n"
+"  -s SLICE_US   Override slice duration\n"
+"  -i INTERVAL   Report interval\n"
+"  -p            Switch only tasks on SCHED_EXT policy intead of all\n"
+"  -h            Display this help and exit\n";
+
+static volatile int exit_req;
+
+static void sigint_handler(int dummy)
+{
+	exit_req = 1;
+}
+
+static float read_cpu_util(__u64 *last_sum, __u64 *last_idle)
+{
+	FILE *fp;
+	char buf[4096];
+	char *line, *cur = NULL, *tok;
+	__u64 sum = 0, idle = 0;
+	__u64 delta_sum, delta_idle;
+	int idx;
+
+	fp = fopen("/proc/stat", "r");
+	if (!fp) {
+		perror("fopen(\"/proc/stat\")");
+		return 0.0;
+	}
+
+	if (!fgets(buf, sizeof(buf), fp)) {
+		perror("fgets(\"/proc/stat\")");
+		fclose(fp);
+		return 0.0;
+	}
+	fclose(fp);
+
+	line = buf;
+	for (idx = 0; (tok = strtok_r(line, " \n", &cur)); idx++) {
+		char *endp = NULL;
+		__u64 v;
+
+		if (idx == 0) {
+			line = NULL;
+			continue;
+		}
+		v = strtoull(tok, &endp, 0);
+		if (!endp || *endp != '\0') {
+			fprintf(stderr, "failed to parse %dth field of /proc/stat (\"%s\")\n",
+				idx, tok);
+			continue;
+		}
+		sum += v;
+		if (idx == 4)
+			idle = v;
+	}
+
+	delta_sum = sum - *last_sum;
+	delta_idle = idle - *last_idle;
+	*last_sum = sum;
+	*last_idle = idle;
+
+	return delta_sum ? (float)(delta_sum - delta_idle) / delta_sum : 0.0;
+}
+
+static void fcg_read_stats(struct scx_flatcg *skel, __u64 *stats)
+{
+	__u64 cnts[FCG_NR_STATS][skel->rodata->nr_cpus];
+	__u32 idx;
+
+	memset(stats, 0, sizeof(stats[0]) * FCG_NR_STATS);
+
+	for (idx = 0; idx < FCG_NR_STATS; idx++) {
+		int ret, cpu;
+
+		ret = bpf_map_lookup_elem(bpf_map__fd(skel->maps.stats),
+					  &idx, cnts[idx]);
+		if (ret < 0)
+			continue;
+		for (cpu = 0; cpu < skel->rodata->nr_cpus; cpu++)
+			stats[idx] += cnts[idx][cpu];
+	}
+}
+
+int main(int argc, char **argv)
+{
+	struct scx_flatcg *skel;
+	struct bpf_link *link;
+	struct timespec intv_ts = { .tv_sec = 2, .tv_nsec = 0 };
+	bool dump_cgrps = false;
+	__u64 last_cpu_sum = 0, last_cpu_idle = 0;
+	__u64 last_stats[FCG_NR_STATS] = {};
+	unsigned long seq = 0;
+	s32 opt;
+
+	signal(SIGINT, sigint_handler);
+	signal(SIGTERM, sigint_handler);
+
+	libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
+
+	skel = scx_flatcg__open();
+	if (!skel) {
+		fprintf(stderr, "Failed to open: %s\n", strerror(errno));
+		return 1;
+	}
+
+	skel->rodata->nr_cpus = libbpf_num_possible_cpus();
+
+	while ((opt = getopt(argc, argv, "s:i:dfph")) != -1) {
+		double v;
+
+		switch (opt) {
+		case 's':
+			v = strtod(optarg, NULL);
+			skel->rodata->cgrp_slice_ns = v * 1000;
+			break;
+		case 'i':
+			v = strtod(optarg, NULL);
+			intv_ts.tv_sec = v;
+			intv_ts.tv_nsec = (v - (float)intv_ts.tv_sec) * 1000000000;
+			break;
+		case 'd':
+			dump_cgrps = true;
+			break;
+		case 'p':
+			skel->rodata->switch_partial = true;
+			break;
+		case 'h':
+		default:
+			fprintf(stderr, help_fmt, basename(argv[0]));
+			return opt != 'h';
+		}
+	}
+
+	printf("slice=%.1lfms intv=%.1lfs dump_cgrps=%d",
+	       (double)skel->rodata->cgrp_slice_ns / 1000000.0,
+	       (double)intv_ts.tv_sec + (double)intv_ts.tv_nsec / 1000000000.0,
+	       dump_cgrps);
+
+	if (scx_flatcg__load(skel)) {
+		fprintf(stderr, "Failed to load: %s\n", strerror(errno));
+		return 1;
+	}
+
+	link = bpf_map__attach_struct_ops(skel->maps.flatcg_ops);
+	if (!link) {
+		fprintf(stderr, "Failed to attach_struct_ops: %s\n",
+			strerror(errno));
+		return 1;
+	}
+
+	while (!exit_req && !uei_exited(&skel->bss->uei)) {
+		__u64 acc_stats[FCG_NR_STATS];
+		__u64 stats[FCG_NR_STATS];
+		float cpu_util;
+		int i;
+
+		cpu_util = read_cpu_util(&last_cpu_sum, &last_cpu_idle);
+
+		fcg_read_stats(skel, acc_stats);
+		for (i = 0; i < FCG_NR_STATS; i++)
+			stats[i] = acc_stats[i] - last_stats[i];
+
+		memcpy(last_stats, acc_stats, sizeof(acc_stats));
+
+		printf("\n[SEQ %6lu cpu=%5.1lf hweight_gen=%lu]\n",
+		       seq++, cpu_util * 100.0, skel->data->hweight_gen);
+		printf("       act:%6llu  deact:%6llu local:%6llu global:%6llu\n",
+		       stats[FCG_STAT_ACT],
+		       stats[FCG_STAT_DEACT],
+		       stats[FCG_STAT_LOCAL],
+		       stats[FCG_STAT_GLOBAL]);
+		printf("HWT   skip:%6llu   race:%6llu cache:%6llu update:%6llu\n",
+		       stats[FCG_STAT_HWT_SKIP],
+		       stats[FCG_STAT_HWT_RACE],
+		       stats[FCG_STAT_HWT_CACHE],
+		       stats[FCG_STAT_HWT_UPDATES]);
+		printf("ENQ   skip:%6llu   race:%6llu\n",
+		       stats[FCG_STAT_ENQ_SKIP],
+		       stats[FCG_STAT_ENQ_RACE]);
+		printf("CNS   keep:%6llu expire:%6llu empty:%6llu   gone:%6llu\n",
+		       stats[FCG_STAT_CNS_KEEP],
+		       stats[FCG_STAT_CNS_EXPIRE],
+		       stats[FCG_STAT_CNS_EMPTY],
+		       stats[FCG_STAT_CNS_GONE]);
+		printf("PNC nocgrp:%6llu   next:%6llu empty:%6llu   gone:%6llu\n",
+		       stats[FCG_STAT_PNC_NO_CGRP],
+		       stats[FCG_STAT_PNC_NEXT],
+		       stats[FCG_STAT_PNC_EMPTY],
+		       stats[FCG_STAT_PNC_GONE]);
+		printf("BAD remove:%6llu\n",
+		       acc_stats[FCG_STAT_BAD_REMOVAL]);
+
+		nanosleep(&intv_ts, NULL);
+	}
+
+	bpf_link__destroy(link);
+	uei_print(&skel->bss->uei);
+	scx_flatcg__destroy(skel);
+	return 0;
+}
diff --git a/tools/sched_ext/scx_flatcg.h b/tools/sched_ext/scx_flatcg.h
new file mode 100644
index 000000000000..490758ed41f0
--- /dev/null
+++ b/tools/sched_ext/scx_flatcg.h
@@ -0,0 +1,49 @@
+#ifndef __SCX_EXAMPLE_FLATCG_H
+#define __SCX_EXAMPLE_FLATCG_H
+
+enum {
+	FCG_HWEIGHT_ONE		= 1LLU << 16,
+};
+
+enum fcg_stat_idx {
+	FCG_STAT_ACT,
+	FCG_STAT_DEACT,
+	FCG_STAT_LOCAL,
+	FCG_STAT_GLOBAL,
+
+	FCG_STAT_HWT_UPDATES,
+	FCG_STAT_HWT_CACHE,
+	FCG_STAT_HWT_SKIP,
+	FCG_STAT_HWT_RACE,
+
+	FCG_STAT_ENQ_SKIP,
+	FCG_STAT_ENQ_RACE,
+
+	FCG_STAT_CNS_KEEP,
+	FCG_STAT_CNS_EXPIRE,
+	FCG_STAT_CNS_EMPTY,
+	FCG_STAT_CNS_GONE,
+
+	FCG_STAT_PNC_NO_CGRP,
+	FCG_STAT_PNC_NEXT,
+	FCG_STAT_PNC_EMPTY,
+	FCG_STAT_PNC_GONE,
+
+	FCG_STAT_BAD_REMOVAL,
+
+	FCG_NR_STATS,
+};
+
+struct fcg_cgrp_ctx {
+	u32			nr_active;
+	u32			nr_runnable;
+	u32			queued;
+	u32			weight;
+	u32			hweight;
+	u64			child_weight_sum;
+	u64			hweight_gen;
+	s64			cvtime_delta;
+	u64			tvtime_now;
+};
+
+#endif /* __SCX_EXAMPLE_FLATCG_H */
-- 
2.41.0

[PATCH 27/34] sched_ext: Implement SCX_KICK_WAIT

[Tejun Heo]

From: David Vernet <dvernet@meta.com>

If set when calling scx_bpf_kick_cpu(), the invoking CPU will busy wait for
the kicked cpu to enter the scheduler. This will be used to improve the
exclusion guarantees in scx_pair.

Signed-off-by: David Vernet <dvernet@meta.com>
Reviewed-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 kernel/sched/core.c  |  4 +++-
 kernel/sched/ext.c   | 33 ++++++++++++++++++++++++++++++++-
 kernel/sched/ext.h   | 20 ++++++++++++++++++++
 kernel/sched/sched.h |  2 ++
 4 files changed, 57 insertions(+), 2 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 77eb4ee4f759..878e84694a6e 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -6052,8 +6052,10 @@ __pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 
 	for_each_active_class(class) {
 		p = class->pick_next_task(rq);
-		if (p)
+		if (p) {
+			scx_notify_pick_next_task(rq, p, class);
 			return p;
+		}
 	}
 
 	BUG(); /* The idle class should always have a runnable task. */
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 5862e8290207..48a8881ff01f 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -125,6 +125,9 @@ static struct {
 
 #endif	/* CONFIG_SMP */
 
+/* for %SCX_KICK_WAIT */
+static u64 __percpu *scx_kick_cpus_pnt_seqs;
+
 /*
  * Direct dispatch marker.
  *
@@ -3269,6 +3272,7 @@ static const struct sysrq_key_op sysrq_sched_ext_reset_op = {
 static void kick_cpus_irq_workfn(struct irq_work *irq_work)
 {
 	struct rq *this_rq = this_rq();
+	u64 *pseqs = this_cpu_ptr(scx_kick_cpus_pnt_seqs);
 	int this_cpu = cpu_of(this_rq);
 	int cpu;
 
@@ -3282,14 +3286,32 @@ static void kick_cpus_irq_workfn(struct irq_work *irq_work)
 			if (cpumask_test_cpu(cpu, this_rq->scx.cpus_to_preempt) &&
 			    rq->curr->sched_class == &ext_sched_class)
 				rq->curr->scx.slice = 0;
+			pseqs[cpu] = rq->scx.pnt_seq;
 			resched_curr(rq);
+		} else {
+			cpumask_clear_cpu(cpu, this_rq->scx.cpus_to_wait);
 		}
 
 		raw_spin_rq_unlock_irqrestore(rq, flags);
 	}
 
+	for_each_cpu_andnot(cpu, this_rq->scx.cpus_to_wait,
+			    cpumask_of(this_cpu)) {
+		/*
+		 * Pairs with smp_store_release() issued by this CPU in
+		 * scx_notify_pick_next_task() on the resched path.
+		 *
+		 * We busy-wait here to guarantee that no other task can be
+		 * scheduled on our core before the target CPU has entered the
+		 * resched path.
+		 */
+		while (smp_load_acquire(&cpu_rq(cpu)->scx.pnt_seq) == pseqs[cpu])
+			cpu_relax();
+	}
+
 	cpumask_clear(this_rq->scx.cpus_to_kick);
 	cpumask_clear(this_rq->scx.cpus_to_preempt);
+	cpumask_clear(this_rq->scx.cpus_to_wait);
 }
 
 void __init init_sched_ext_class(void)
@@ -3303,7 +3325,7 @@ void __init init_sched_ext_class(void)
 	 * through the generated vmlinux.h.
 	 */
 	WRITE_ONCE(v, SCX_WAKE_EXEC | SCX_ENQ_WAKEUP | SCX_DEQ_SLEEP |
-		   SCX_TG_ONLINE);
+		   SCX_TG_ONLINE | SCX_KICK_PREEMPT);
 
 	BUG_ON(rhashtable_init(&dsq_hash, &dsq_hash_params));
 	init_dsq(&scx_dsq_global, SCX_DSQ_GLOBAL);
@@ -3311,6 +3333,12 @@ void __init init_sched_ext_class(void)
 	BUG_ON(!alloc_cpumask_var(&idle_masks.cpu, GFP_KERNEL));
 	BUG_ON(!alloc_cpumask_var(&idle_masks.smt, GFP_KERNEL));
 #endif
+	scx_kick_cpus_pnt_seqs =
+		__alloc_percpu(sizeof(scx_kick_cpus_pnt_seqs[0]) *
+			       num_possible_cpus(),
+			       __alignof__(scx_kick_cpus_pnt_seqs[0]));
+	BUG_ON(!scx_kick_cpus_pnt_seqs);
+
 	for_each_possible_cpu(cpu) {
 		struct rq *rq = cpu_rq(cpu);
 
@@ -3319,6 +3347,7 @@ void __init init_sched_ext_class(void)
 
 		BUG_ON(!zalloc_cpumask_var(&rq->scx.cpus_to_kick, GFP_KERNEL));
 		BUG_ON(!zalloc_cpumask_var(&rq->scx.cpus_to_preempt, GFP_KERNEL));
+		BUG_ON(!zalloc_cpumask_var(&rq->scx.cpus_to_wait, GFP_KERNEL));
 		init_irq_work(&rq->scx.kick_cpus_irq_work, kick_cpus_irq_workfn);
 	}
 
@@ -3585,6 +3614,8 @@ void scx_bpf_kick_cpu(s32 cpu, u64 flags)
 	cpumask_set_cpu(cpu, rq->scx.cpus_to_kick);
 	if (flags & SCX_KICK_PREEMPT)
 		cpumask_set_cpu(cpu, rq->scx.cpus_to_preempt);
+	if (flags & SCX_KICK_WAIT)
+		cpumask_set_cpu(cpu, rq->scx.cpus_to_wait);
 
 	irq_work_queue(&rq->scx.kick_cpus_irq_work);
 	preempt_enable();
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index c3404a0a7637..abb283ac3bc7 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -65,6 +65,7 @@ enum scx_pick_idle_cpu_flags {
 
 enum scx_kick_flags {
 	SCX_KICK_PREEMPT	= 1LLU << 0,	/* force scheduling on the CPU */
+	SCX_KICK_WAIT		= 1LLU << 1,	/* wait for the CPU to be rescheduled */
 };
 
 enum scx_tg_flags {
@@ -115,6 +116,22 @@ __printf(2, 3) void scx_ops_error_type(enum scx_exit_type type,
 #define scx_ops_error(fmt, args...)						\
 	scx_ops_error_type(SCX_EXIT_ERROR, fmt, ##args)
 
+static inline void scx_notify_pick_next_task(struct rq *rq,
+					     const struct task_struct *p,
+					     const struct sched_class *active)
+{
+#ifdef CONFIG_SMP
+	if (!scx_enabled())
+		return;
+	/*
+	 * Pairs with the smp_load_acquire() issued by a CPU in
+	 * kick_cpus_irq_workfn() who is waiting for this CPU to perform a
+	 * resched.
+	 */
+	smp_store_release(&rq->scx.pnt_seq, rq->scx.pnt_seq + 1);
+#endif
+}
+
 static inline void scx_notify_sched_tick(void)
 {
 	unsigned long last_check;
@@ -170,6 +187,9 @@ static inline int scx_check_setscheduler(struct task_struct *p,
 					 int policy) { return 0; }
 static inline bool scx_can_stop_tick(struct rq *rq) { return true; }
 static inline void init_sched_ext_class(void) {}
+static inline void scx_notify_pick_next_task(struct rq *rq,
+					     const struct task_struct *p,
+					     const struct sched_class *active) {}
 static inline void scx_notify_sched_tick(void) {}
 
 #define for_each_active_class		for_each_class
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 00bf33fdbd64..ce6e0a73135b 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -710,6 +710,8 @@ struct scx_rq {
 	u32			flags;
 	cpumask_var_t		cpus_to_kick;
 	cpumask_var_t		cpus_to_preempt;
+	cpumask_var_t		cpus_to_wait;
+	u64			pnt_seq;
 	struct irq_work		kick_cpus_irq_work;
 };
 #endif /* CONFIG_SCHED_CLASS_EXT */
-- 
2.41.0

[PATCH 28/34] sched_ext: Implement sched_ext_ops.cpu_acquire/release()

[Tejun Heo]

From: David Vernet <dvernet@meta.com>

Scheduler classes are strictly ordered and when a higher priority class has
tasks to run, the lower priority ones lose access to the CPU. Being able to
monitor and act on these events are necessary for use cases includling
strict core-scheduling and latency management.

This patch adds two operations ops.cpu_acquire() and .cpu_release(). The
former is invoked when a CPU becomes available to the BPF scheduler and the
opposite for the latter. This patch also implements
scx_bpf_reenqueue_local() which can be called from .cpu_release() to trigger
requeueing of all tasks in the local dsq of the CPU so that the tasks can be
reassigned to other available CPUs.

scx_pair is updated to use .cpu_acquire/release() along with
%SCX_KICK_WAIT to make the pair scheduling guarantee strict even when a CPU
is preempted by a higher priority scheduler class.

scx_qmap is updated to use .cpu_acquire/release() to empty the local
dsq of a preempted CPU. A similar approach can be adopted by BPF schedulers
that want to have a tight control over latency.

v3: * Drop the const qualifier from scx_cpu_release_args.task. BPF enforces
      access control through the verifier, so the qualifier isn't actually
      operative and only gets in the way when interacting with various
      helpers.

v2: * Add p->scx.kf_mask annotation to allow calling
      scx_bpf_reenqueue_local() from ops.cpu_release() nested inside
      ops.init() and other sleepable operations.

Signed-off-by: David Vernet <dvernet@meta.com>
Reviewed-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 include/linux/sched/ext.h        |  53 ++++++++++++-
 kernel/sched/ext.c               | 131 ++++++++++++++++++++++++++++++-
 kernel/sched/ext.h               |  24 +++++-
 kernel/sched/sched.h             |   1 +
 tools/sched_ext/scx_common.bpf.h |   1 +
 tools/sched_ext/scx_pair.bpf.c   | 101 ++++++++++++++++++++++--
 tools/sched_ext/scx_qmap.bpf.c   |  37 ++++++++-
 tools/sched_ext/scx_qmap.c       |   4 +-
 8 files changed, 340 insertions(+), 12 deletions(-)

diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 2d6da27bdd74..8a275ec05ee1 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -135,6 +135,32 @@ struct scx_cgroup_init_args {
 	u32			weight;
 };
 
+enum scx_cpu_preempt_reason {
+	/* next task is being scheduled by &sched_class_rt */
+        SCX_CPU_PREEMPT_RT,
+	/* next task is being scheduled by &sched_class_dl */
+        SCX_CPU_PREEMPT_DL,
+	/* next task is being scheduled by &sched_class_stop */
+        SCX_CPU_PREEMPT_STOP,
+	/* unknown reason for SCX being preempted */
+        SCX_CPU_PREEMPT_UNKNOWN,
+};
+
+/*
+ * Argument container for ops->cpu_acquire(). Currently empty, but may be
+ * expanded in the future.
+ */
+struct scx_cpu_acquire_args {};
+
+/* argument container for ops->cpu_release() */
+struct scx_cpu_release_args {
+	/* the reason the CPU was preempted */
+	enum scx_cpu_preempt_reason reason;
+
+	/* the task that's going to be scheduled on the CPU */
+	struct task_struct *task;
+};
+
 /**
  * struct sched_ext_ops - Operation table for BPF scheduler implementation
  *
@@ -329,6 +355,28 @@ struct sched_ext_ops {
 	 */
 	void (*update_idle)(s32 cpu, bool idle);
 
+	/**
+	 * cpu_acquire - A CPU is becoming available to the BPF scheduler
+	 * @cpu: The CPU being acquired by the BPF scheduler.
+	 * @args: Acquire arguments, see the struct definition.
+	 *
+	 * A CPU that was previously released from the BPF scheduler is now once
+	 * again under its control.
+	 */
+	void (*cpu_acquire)(s32 cpu, struct scx_cpu_acquire_args *args);
+
+	/**
+	 * cpu_release - A CPU is taken away from the BPF scheduler
+	 * @cpu: The CPU being released by the BPF scheduler.
+	 * @args: Release arguments, see the struct definition.
+	 *
+	 * The specified CPU is no longer under the control of the BPF
+	 * scheduler. This could be because it was preempted by a higher
+	 * priority sched_class, though there may be other reasons as well. The
+	 * caller should consult @args->reason to determine the cause.
+	 */
+	void (*cpu_release)(s32 cpu, struct scx_cpu_release_args *args);
+
 	/**
 	 * prep_enable - Prepare to enable BPF scheduling for a task
 	 * @p: task to prepare BPF scheduling for
@@ -533,12 +581,15 @@ enum scx_kf_mask {
 	/* all non-sleepables may be nested inside INIT and SLEEPABLE */
 	SCX_KF_INIT		= 1 << 0, /* running ops.init() */
 	SCX_KF_SLEEPABLE	= 1 << 1, /* other sleepable init operations */
+	/* ENQUEUE and DISPATCH may be nested inside CPU_RELEASE */
+	SCX_KF_CPU_RELEASE	= 1 << 2, /* ops.cpu_release() */
 	/* ops.dequeue (in REST) may be nested inside DISPATCH */
 	SCX_KF_DISPATCH		= 1 << 3, /* ops.dispatch() */
 	SCX_KF_ENQUEUE		= 1 << 4, /* ops.enqueue() */
 	SCX_KF_REST		= 1 << 5, /* other rq-locked operations */
 
-	__SCX_KF_RQ_LOCKED	= SCX_KF_DISPATCH | SCX_KF_ENQUEUE | SCX_KF_REST,
+	__SCX_KF_RQ_LOCKED	= SCX_KF_CPU_RELEASE | SCX_KF_DISPATCH |
+				  SCX_KF_ENQUEUE | SCX_KF_REST,
 	__SCX_KF_TERMINAL	= SCX_KF_ENQUEUE | SCX_KF_REST,
 };
 
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 48a8881ff01f..1b83dddbdf10 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -83,6 +83,7 @@ static bool scx_warned_zero_slice;
 
 static DEFINE_STATIC_KEY_FALSE(scx_ops_enq_last);
 static DEFINE_STATIC_KEY_FALSE(scx_ops_enq_exiting);
+DEFINE_STATIC_KEY_FALSE(scx_ops_cpu_preempt);
 static DEFINE_STATIC_KEY_FALSE(scx_builtin_idle_enabled);
 
 struct static_key_false scx_has_op[SCX_NR_ONLINE_OPS] =
@@ -303,6 +304,12 @@ static __always_inline bool scx_kf_allowed(u32 mask)
 	 * inside ops.dispatch(). We don't need to check the SCX_KF_SLEEPABLE
 	 * boundary thanks to the above in_interrupt() check.
 	 */
+	if (unlikely(highest_bit(mask) == SCX_KF_CPU_RELEASE &&
+		     (current->scx.kf_mask & higher_bits(SCX_KF_CPU_RELEASE)))) {
+		scx_ops_error("cpu_release kfunc called from a nested operation");
+		return false;
+	}
+
 	if (unlikely(highest_bit(mask) == SCX_KF_DISPATCH &&
 		     (current->scx.kf_mask & higher_bits(SCX_KF_DISPATCH)))) {
 		scx_ops_error("dispatch kfunc called from a nested operation");
@@ -1376,6 +1383,19 @@ static int balance_scx(struct rq *rq, struct task_struct *prev,
 
 	lockdep_assert_rq_held(rq);
 
+	if (static_branch_unlikely(&scx_ops_cpu_preempt) &&
+	    unlikely(rq->scx.cpu_released)) {
+		/*
+		 * If the previous sched_class for the current CPU was not SCX,
+		 * notify the BPF scheduler that it again has control of the
+		 * core. This callback complements ->cpu_release(), which is
+		 * emitted in scx_notify_pick_next_task().
+		 */
+		if (SCX_HAS_OP(cpu_acquire))
+			SCX_CALL_OP(0, cpu_acquire, cpu_of(rq), NULL);
+		rq->scx.cpu_released = false;
+	}
+
 	if (prev_on_scx) {
 		WARN_ON_ONCE(prev->scx.flags & SCX_TASK_BAL_KEEP);
 		update_curr_scx(rq);
@@ -1383,7 +1403,9 @@ static int balance_scx(struct rq *rq, struct task_struct *prev,
 		/*
 		 * If @prev is runnable & has slice left, it has priority and
 		 * fetching more just increases latency for the fetched tasks.
-		 * Tell put_prev_task_scx() to put @prev on local_dsq.
+		 * Tell put_prev_task_scx() to put @prev on local_dsq. If the
+		 * BPF scheduler wants to handle this explicitly, it should
+		 * implement ->cpu_released().
 		 *
 		 * See scx_ops_disable_workfn() for the explanation on the
 		 * disabling() test.
@@ -1589,6 +1611,58 @@ static struct task_struct *pick_next_task_scx(struct rq *rq)
 	return p;
 }
 
+static enum scx_cpu_preempt_reason
+preempt_reason_from_class(const struct sched_class *class)
+{
+#ifdef CONFIG_SMP
+	if (class == &stop_sched_class)
+		return SCX_CPU_PREEMPT_STOP;
+#endif
+	if (class == &dl_sched_class)
+		return SCX_CPU_PREEMPT_DL;
+	if (class == &rt_sched_class)
+		return SCX_CPU_PREEMPT_RT;
+	return SCX_CPU_PREEMPT_UNKNOWN;
+}
+
+void __scx_notify_pick_next_task(struct rq *rq, struct task_struct *task,
+				 const struct sched_class *active)
+{
+	lockdep_assert_rq_held(rq);
+
+	/*
+	 * The callback is conceptually meant to convey that the CPU is no
+	 * longer under the control of SCX. Therefore, don't invoke the
+	 * callback if the CPU is is staying on SCX, or going idle (in which
+	 * case the SCX scheduler has actively decided not to schedule any
+	 * tasks on the CPU).
+	 */
+	if (likely(active >= &ext_sched_class))
+		return;
+
+	/*
+	 * At this point we know that SCX was preempted by a higher priority
+	 * sched_class, so invoke the ->cpu_release() callback if we have not
+	 * done so already. We only send the callback once between SCX being
+	 * preempted, and it regaining control of the CPU.
+	 *
+	 * ->cpu_release() complements ->cpu_acquire(), which is emitted the
+	 *  next time that balance_scx() is invoked.
+	 */
+	if (!rq->scx.cpu_released) {
+		if (SCX_HAS_OP(cpu_release)) {
+			struct scx_cpu_release_args args = {
+				.reason = preempt_reason_from_class(active),
+				.task = task,
+			};
+
+			SCX_CALL_OP(SCX_KF_CPU_RELEASE,
+				    cpu_release, cpu_of(rq), &args);
+		}
+		rq->scx.cpu_released = true;
+	}
+}
+
 #ifdef CONFIG_SMP
 
 static bool test_and_clear_cpu_idle(int cpu)
@@ -2698,6 +2772,7 @@ static void scx_ops_disable_workfn(struct kthread_work *work)
 		static_branch_disable_cpuslocked(&scx_has_op[i]);
 	static_branch_disable_cpuslocked(&scx_ops_enq_last);
 	static_branch_disable_cpuslocked(&scx_ops_enq_exiting);
+	static_branch_disable_cpuslocked(&scx_ops_cpu_preempt);
 	static_branch_disable_cpuslocked(&scx_builtin_idle_enabled);
 	synchronize_rcu();
 
@@ -2904,6 +2979,8 @@ static int scx_ops_enable(struct sched_ext_ops *ops)
 
 	if (ops->flags & SCX_OPS_ENQ_EXITING)
 		static_branch_enable_cpuslocked(&scx_ops_enq_exiting);
+	if (scx_ops.cpu_acquire || scx_ops.cpu_release)
+		static_branch_enable_cpuslocked(&scx_ops_cpu_preempt);
 
 	if (!ops->update_idle || (ops->flags & SCX_OPS_KEEP_BUILTIN_IDLE)) {
 		reset_idle_masks();
@@ -3584,6 +3661,56 @@ static const struct btf_kfunc_id_set scx_kfunc_set_dispatch = {
 	.set			= &scx_kfunc_ids_dispatch,
 };
 
+/**
+ * scx_bpf_reenqueue_local - Re-enqueue tasks on a local DSQ
+ *
+ * Iterate over all of the tasks currently enqueued on the local DSQ of the
+ * caller's CPU, and re-enqueue them in the BPF scheduler. Returns the number of
+ * processed tasks. Can only be called from ops.cpu_release().
+ */
+u32 scx_bpf_reenqueue_local(void)
+{
+	u32 nr_enqueued, i;
+	struct rq *rq;
+	struct scx_rq *scx_rq;
+
+	if (!scx_kf_allowed(SCX_KF_CPU_RELEASE))
+		return 0;
+
+	rq = cpu_rq(smp_processor_id());
+	lockdep_assert_rq_held(rq);
+	scx_rq = &rq->scx;
+
+	/*
+	 * Get the number of tasks on the local DSQ before iterating over it to
+	 * pull off tasks. The enqueue callback below can signal that it wants
+	 * the task to stay on the local DSQ, and we want to prevent the BPF
+	 * scheduler from causing us to loop indefinitely.
+	 */
+	nr_enqueued = scx_rq->local_dsq.nr;
+	for (i = 0; i < nr_enqueued; i++) {
+		struct task_struct *p;
+
+		p = first_local_task(rq);
+		WARN_ON_ONCE(atomic64_read(&p->scx.ops_state) != SCX_OPSS_NONE);
+		WARN_ON_ONCE(!(p->scx.flags & SCX_TASK_QUEUED));
+		WARN_ON_ONCE(p->scx.holding_cpu != -1);
+		dispatch_dequeue(scx_rq, p);
+		do_enqueue_task(rq, p, SCX_ENQ_REENQ, -1);
+	}
+
+	return nr_enqueued;
+}
+
+BTF_SET8_START(scx_kfunc_ids_cpu_release)
+BTF_ID_FLAGS(func, scx_bpf_reenqueue_local)
+BTF_SET8_END(scx_kfunc_ids_cpu_release)
+
+static const struct btf_kfunc_id_set scx_kfunc_set_cpu_release = {
+	.owner			= THIS_MODULE,
+	.set			= &scx_kfunc_ids_cpu_release,
+};
+
 /**
  * scx_bpf_kick_cpu - Trigger reschedule on a CPU
  * @cpu: cpu to kick
@@ -3968,6 +4095,8 @@ static int __init register_ext_kfuncs(void)
 					     &scx_kfunc_set_enqueue_dispatch)) ||
 	    (ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
 					     &scx_kfunc_set_dispatch)) ||
+	    (ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
+					     &scx_kfunc_set_cpu_release)) ||
 	    (ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
 					     &scx_kfunc_set_any))) {
 		pr_err("sched_ext: failed to register kfunc sets (%d)\n", ret);
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index abb283ac3bc7..c71bf641f7a5 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -27,6 +27,17 @@ enum scx_enq_flags {
 	 */
 	SCX_ENQ_PREEMPT		= 1LLU << 32,
 
+	/*
+	 * The task being enqueued was previously enqueued on the current CPU's
+	 * %SCX_DSQ_LOCAL, but was removed from it in a call to the
+	 * bpf_scx_reenqueue_local() kfunc. If bpf_scx_reenqueue_local() was
+	 * invoked in a ->cpu_release() callback, and the task is again
+	 * dispatched back to %SCX_LOCAL_DSQ by this current ->enqueue(), the
+	 * task will not be scheduled on the CPU until at least the next invocation
+	 * of the ->cpu_acquire() callback.
+	 */
+	SCX_ENQ_REENQ		= 1LLU << 40,
+
 	/*
 	 * The task being enqueued is the only task available for the cpu. By
 	 * default, ext core keeps executing such tasks but when
@@ -97,6 +108,8 @@ DECLARE_STATIC_KEY_FALSE(__scx_switched_all);
 #define scx_enabled()		static_branch_unlikely(&__scx_ops_enabled)
 #define scx_switched_all()	static_branch_unlikely(&__scx_switched_all)
 
+DECLARE_STATIC_KEY_FALSE(scx_ops_cpu_preempt);
+
 static inline bool task_on_scx(struct task_struct *p)
 {
 	return scx_enabled() && p->sched_class == &ext_sched_class;
@@ -116,13 +129,17 @@ __printf(2, 3) void scx_ops_error_type(enum scx_exit_type type,
 #define scx_ops_error(fmt, args...)						\
 	scx_ops_error_type(SCX_EXIT_ERROR, fmt, ##args)
 
+void __scx_notify_pick_next_task(struct rq *rq,
+				 struct task_struct *p,
+				 const struct sched_class *active);
+
 static inline void scx_notify_pick_next_task(struct rq *rq,
-					     const struct task_struct *p,
+					     struct task_struct *p,
 					     const struct sched_class *active)
 {
-#ifdef CONFIG_SMP
 	if (!scx_enabled())
 		return;
+#ifdef CONFIG_SMP
 	/*
 	 * Pairs with the smp_load_acquire() issued by a CPU in
 	 * kick_cpus_irq_workfn() who is waiting for this CPU to perform a
@@ -130,6 +147,9 @@ static inline void scx_notify_pick_next_task(struct rq *rq,
 	 */
 	smp_store_release(&rq->scx.pnt_seq, rq->scx.pnt_seq + 1);
 #endif
+	if (!static_branch_unlikely(&scx_ops_cpu_preempt))
+		return;
+	__scx_notify_pick_next_task(rq, p, active);
 }
 
 static inline void scx_notify_sched_tick(void)
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index ce6e0a73135b..e7b15bd7adbc 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -708,6 +708,7 @@ struct scx_rq {
 	u64			extra_enq_flags;	/* see move_task_to_local_dsq() */
 	u32			nr_running;
 	u32			flags;
+	bool			cpu_released;
 	cpumask_var_t		cpus_to_kick;
 	cpumask_var_t		cpus_to_preempt;
 	cpumask_var_t		cpus_to_wait;
diff --git a/tools/sched_ext/scx_common.bpf.h b/tools/sched_ext/scx_common.bpf.h
index 4f8e447f66d3..8a210ab0eaab 100644
--- a/tools/sched_ext/scx_common.bpf.h
+++ b/tools/sched_ext/scx_common.bpf.h
@@ -70,6 +70,7 @@ void scx_bpf_destroy_dsq(u64 dsq_id) __ksym;
 bool scx_bpf_task_running(const struct task_struct *p) __ksym;
 s32 scx_bpf_task_cpu(const struct task_struct *p) __ksym;
 struct cgroup *scx_bpf_task_cgroup(struct task_struct *p) __ksym;
+u32 scx_bpf_reenqueue_local(void) __ksym;
 
 #define BPF_STRUCT_OPS(name, args...)						\
 SEC("struct_ops/"#name)								\
diff --git a/tools/sched_ext/scx_pair.bpf.c b/tools/sched_ext/scx_pair.bpf.c
index f980d301dbbe..cda126980ed5 100644
--- a/tools/sched_ext/scx_pair.bpf.c
+++ b/tools/sched_ext/scx_pair.bpf.c
@@ -89,6 +89,28 @@
  * be resolved in the near future which should allow greatly simplifying this
  * scheduler.
  *
+ * Dealing with preemption
+ * -----------------------
+ *
+ * SCX is the lowest priority sched_class, and could be preempted by them at
+ * any time. To address this, the scheduler implements pair_cpu_release() and
+ * pair_cpu_acquire() callbacks which are invoked by the core scheduler when
+ * the scheduler loses and gains control of the CPU respectively.
+ *
+ * In pair_cpu_release(), we mark the pair_ctx as having been preempted, and
+ * then invoke:
+ *
+ * scx_bpf_kick_cpu(pair_cpu, SCX_KICK_PREEMPT | SCX_KICK_WAIT);
+ *
+ * This preempts the pair CPU, and waits until it has re-entered the scheduler
+ * before returning. This is necessary to ensure that the higher priority
+ * sched_class that preempted our scheduler does not schedule a task
+ * concurrently with our pair CPU.
+ *
+ * When the CPU is re-acquired in pair_cpu_acquire(), we unmark the preemption
+ * in the pair_ctx, and send another resched IPI to the pair CPU to re-enable
+ * pair scheduling.
+ *
  * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
  * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
  * Copyright (c) 2022 David Vernet <dvernet@meta.com>
@@ -129,6 +151,12 @@ struct pair_ctx {
 
 	/* the CPUs that are currently active on the cgroup */
 	u32			active_mask;
+
+	/*
+	 * the CPUs that are currently preempted and running tasks in a
+	 * different scheduler.
+	 */
+	u32			preempted_mask;
 };
 
 struct {
@@ -295,7 +323,7 @@ static int try_dispatch(s32 cpu)
 	struct task_struct *p;
 	u64 now = bpf_ktime_get_ns();
 	bool kick_pair = false;
-	bool expired;
+	bool expired, pair_preempted;
 	u32 *vptr, in_pair_mask;
 	s32 pid, q_idx;
 	u64 cgid;
@@ -324,10 +352,14 @@ static int try_dispatch(s32 cpu)
 		 */
 		pairc->draining = true;
 
-		if (pairc->active_mask) {
+		pair_preempted = pairc->preempted_mask;
+		if (pairc->active_mask || pair_preempted) {
 			/*
-			 * The other CPU is still active We want to wait until
-			 * this cgroup expires.
+			 * The other CPU is still active, or is no longer under
+			 * our control due to e.g. being preempted by a higher
+			 * priority sched_class. We want to wait until this
+			 * cgroup expires, or until control of our pair CPU has
+			 * been returned to us.
 			 *
 			 * If the pair controls its CPU, and the time already
 			 * expired, kick.  When the other CPU arrives at
@@ -336,7 +368,7 @@ static int try_dispatch(s32 cpu)
 			 */
 			__sync_fetch_and_add(&nr_exp_waits, 1);
 			bpf_spin_unlock(&pairc->lock);
-			if (expired)
+			if (expired && !pair_preempted)
 				kick_pair = true;
 			goto out_maybe_kick;
 		}
@@ -475,6 +507,63 @@ void BPF_STRUCT_OPS(pair_dispatch, s32 cpu, struct task_struct *prev)
 	}
 }
 
+void BPF_STRUCT_OPS(pair_cpu_acquire, s32 cpu, struct scx_cpu_acquire_args *args)
+{
+	int ret;
+	u32 in_pair_mask;
+	struct pair_ctx *pairc;
+	bool kick_pair;
+
+	ret = lookup_pairc_and_mask(cpu, &pairc, &in_pair_mask);
+	if (ret)
+		return;
+
+	bpf_spin_lock(&pairc->lock);
+	pairc->preempted_mask &= ~in_pair_mask;
+	/* Kick the pair CPU, unless it was also preempted. */
+	kick_pair = !pairc->preempted_mask;
+	bpf_spin_unlock(&pairc->lock);
+
+	if (kick_pair) {
+		s32 *pair = (s32 *)MEMBER_VPTR(pair_cpu, [cpu]);
+
+		if (pair) {
+			__sync_fetch_and_add(&nr_kicks, 1);
+			scx_bpf_kick_cpu(*pair, SCX_KICK_PREEMPT);
+		}
+	}
+}
+
+void BPF_STRUCT_OPS(pair_cpu_release, s32 cpu, struct scx_cpu_release_args *args)
+{
+	int ret;
+	u32 in_pair_mask;
+	struct pair_ctx *pairc;
+	bool kick_pair;
+
+	ret = lookup_pairc_and_mask(cpu, &pairc, &in_pair_mask);
+	if (ret)
+		return;
+
+	bpf_spin_lock(&pairc->lock);
+	pairc->preempted_mask |= in_pair_mask;
+	pairc->active_mask &= ~in_pair_mask;
+	/* Kick the pair CPU if it's still running. */
+	kick_pair = pairc->active_mask;
+	pairc->draining = true;
+	bpf_spin_unlock(&pairc->lock);
+
+	if (kick_pair) {
+		s32 *pair = (s32 *)MEMBER_VPTR(pair_cpu, [cpu]);
+
+		if (pair) {
+			__sync_fetch_and_add(&nr_kicks, 1);
+			scx_bpf_kick_cpu(*pair, SCX_KICK_PREEMPT | SCX_KICK_WAIT);
+		}
+	}
+	__sync_fetch_and_add(&nr_preemptions, 1);
+}
+
 s32 BPF_STRUCT_OPS(pair_cgroup_init, struct cgroup *cgrp)
 {
 	u64 cgid = cgrp->kn->id;
@@ -528,6 +617,8 @@ SEC(".struct_ops.link")
 struct sched_ext_ops pair_ops = {
 	.enqueue		= (void *)pair_enqueue,
 	.dispatch		= (void *)pair_dispatch,
+	.cpu_acquire		= (void *)pair_cpu_acquire,
+	.cpu_release		= (void *)pair_cpu_release,
 	.cgroup_init		= (void *)pair_cgroup_init,
 	.cgroup_exit		= (void *)pair_cgroup_exit,
 	.init			= (void *)pair_init,
diff --git a/tools/sched_ext/scx_qmap.bpf.c b/tools/sched_ext/scx_qmap.bpf.c
index 1c3a7d050e32..bfffbfd3368b 100644
--- a/tools/sched_ext/scx_qmap.bpf.c
+++ b/tools/sched_ext/scx_qmap.bpf.c
@@ -11,6 +11,8 @@
  *
  * - BPF-side queueing using PIDs.
  * - Sleepable per-task storage allocation using ops.prep_enable().
+ * - Using ops.cpu_release() to handle a higher priority scheduling class taking
+ *   the CPU away.
  *
  * This scheduler is primarily for demonstration and testing of sched_ext
  * features and unlikely to be useful for actual workloads.
@@ -81,7 +83,7 @@ struct {
 } dispatch_idx_cnt SEC(".maps");
 
 /* Statistics */
-unsigned long nr_enqueued, nr_dispatched, nr_dequeued;
+unsigned long nr_enqueued, nr_dispatched, nr_reenqueued, nr_dequeued;
 
 s32 BPF_STRUCT_OPS(qmap_select_cpu, struct task_struct *p,
 		   s32 prev_cpu, u64 wake_flags)
@@ -155,6 +157,22 @@ void BPF_STRUCT_OPS(qmap_enqueue, struct task_struct *p, u64 enq_flags)
 		return;
 	}
 
+	/*
+	 * If the task was re-enqueued due to the CPU being preempted by a
+	 * higher priority scheduling class, just re-enqueue the task directly
+	 * on the global DSQ. As we want another CPU to pick it up, find and
+	 * kick an idle CPU.
+	 */
+	if (enq_flags & SCX_ENQ_REENQ) {
+		s32 cpu;
+
+		scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, 0, enq_flags);
+		cpu = scx_bpf_pick_idle_cpu(p->cpus_ptr, 0);
+		if (cpu >= 0)
+			scx_bpf_kick_cpu(cpu, 0);
+		return;
+	}
+
 	ring = bpf_map_lookup_elem(&queue_arr, &idx);
 	if (!ring) {
 		scx_bpf_error("failed to find ring %d", idx);
@@ -240,6 +258,22 @@ void BPF_STRUCT_OPS(qmap_dispatch, s32 cpu, struct task_struct *prev)
 	}
 }
 
+void BPF_STRUCT_OPS(qmap_cpu_release, s32 cpu, struct scx_cpu_release_args *args)
+{
+	u32 cnt;
+
+	/*
+	 * Called when @cpu is taken by a higher priority scheduling class. This
+	 * makes @cpu no longer available for executing sched_ext tasks. As we
+	 * don't want the tasks in @cpu's local dsq to sit there until @cpu
+	 * becomes available again, re-enqueue them into the global dsq. See
+	 * %SCX_ENQ_REENQ handling in qmap_enqueue().
+	 */
+	cnt = scx_bpf_reenqueue_local();
+	if (cnt)
+		__sync_fetch_and_add(&nr_reenqueued, cnt);
+}
+
 s32 BPF_STRUCT_OPS(qmap_prep_enable, struct task_struct *p,
 		   struct scx_enable_args *args)
 {
@@ -275,6 +309,7 @@ struct sched_ext_ops qmap_ops = {
 	.enqueue		= (void *)qmap_enqueue,
 	.dequeue		= (void *)qmap_dequeue,
 	.dispatch		= (void *)qmap_dispatch,
+	.cpu_release		= (void *)qmap_cpu_release,
 	.prep_enable		= (void *)qmap_prep_enable,
 	.init			= (void *)qmap_init,
 	.exit			= (void *)qmap_exit,
diff --git a/tools/sched_ext/scx_qmap.c b/tools/sched_ext/scx_qmap.c
index 805ac453698f..b47be72ee66e 100644
--- a/tools/sched_ext/scx_qmap.c
+++ b/tools/sched_ext/scx_qmap.c
@@ -92,9 +92,9 @@ int main(int argc, char **argv)
 		long nr_enqueued = skel->bss->nr_enqueued;
 		long nr_dispatched = skel->bss->nr_dispatched;
 
-		printf("enq=%lu, dsp=%lu, delta=%ld, deq=%lu\n",
+		printf("enq=%lu, dsp=%lu, delta=%ld, reenq=%lu, deq=%lu\n",
 		       nr_enqueued, nr_dispatched, nr_enqueued - nr_dispatched,
-		       skel->bss->nr_dequeued);
+		       skel->bss->nr_reenqueued, skel->bss->nr_dequeued);
 		fflush(stdout);
 		sleep(1);
 	}
-- 
2.41.0

[PATCH 29/34] sched_ext: Implement sched_ext_ops.cpu_online/offline()

[Tejun Heo]

Add ops.cpu_online/offline() which are invoked when CPUs come online and
offline respectively. As the enqueue path already automatically bypasses
tasks to the local dsq on a deactivated CPU, BPF schedulers are guaranteed
to see tasks only on CPUs which are between online() and offline().

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
---
 include/linux/sched/ext.h | 18 ++++++++++++++++++
 kernel/sched/ext.c        | 18 +++++++++++++++++-
 2 files changed, 35 insertions(+), 1 deletion(-)

diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 8a275ec05ee1..c17957bd75df 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -377,6 +377,24 @@ struct sched_ext_ops {
 	 */
 	void (*cpu_release)(s32 cpu, struct scx_cpu_release_args *args);
 
+	/**
+	 * cpu_online - A CPU became online
+	 * @cpu: CPU which just came up
+	 *
+	 * @cpu just came online. @cpu doesn't call ops.enqueue() or run tasks
+	 * associated with other CPUs beforehand.
+	 */
+	void (*cpu_online)(s32 cpu);
+
+	/**
+	 * cpu_offline - A CPU is going offline
+	 * @cpu: CPU which is going offline
+	 *
+	 * @cpu is going offline. @cpu doesn't call ops.enqueue() or run tasks
+	 * associated with other CPUs afterwards.
+	 */
+	void (*cpu_offline)(s32 cpu);
+
 	/**
 	 * prep_enable - Prepare to enable BPF scheduling for a task
 	 * @p: task to prepare BPF scheduling for
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 1b83dddbdf10..e54d8c7d19a9 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -1392,7 +1392,8 @@ static int balance_scx(struct rq *rq, struct task_struct *prev,
 		 * emitted in scx_notify_pick_next_task().
 		 */
 		if (SCX_HAS_OP(cpu_acquire))
-			SCX_CALL_OP(0, cpu_acquire, cpu_of(rq), NULL);
+			SCX_CALL_OP(SCX_KF_UNLOCKED, cpu_acquire, cpu_of(rq),
+				    NULL);
 		rq->scx.cpu_released = false;
 	}
 
@@ -1852,6 +1853,18 @@ void __scx_update_idle(struct rq *rq, bool idle)
 #endif
 }
 
+static void rq_online_scx(struct rq *rq, enum rq_onoff_reason reason)
+{
+	if (SCX_HAS_OP(cpu_online) && reason == RQ_ONOFF_HOTPLUG)
+		SCX_CALL_OP(SCX_KF_REST, cpu_online, cpu_of(rq));
+}
+
+static void rq_offline_scx(struct rq *rq, enum rq_onoff_reason reason)
+{
+	if (SCX_HAS_OP(cpu_offline) && reason == RQ_ONOFF_HOTPLUG)
+		SCX_CALL_OP(SCX_KF_REST, cpu_offline, cpu_of(rq));
+}
+
 #else /* !CONFIG_SMP */
 
 static bool test_and_clear_cpu_idle(int cpu) { return false; }
@@ -2370,6 +2383,9 @@ DEFINE_SCHED_CLASS(ext) = {
 	.balance		= balance_scx,
 	.select_task_rq		= select_task_rq_scx,
 	.set_cpus_allowed	= set_cpus_allowed_scx,
+
+	.rq_online		= rq_online_scx,
+	.rq_offline		= rq_offline_scx,
 #endif
 
 	.task_tick		= task_tick_scx,
-- 
2.41.0

[PATCH 30/34] sched_ext: Implement core-sched support

[Tejun Heo]

The core-sched support is composed of the following parts:

* task_struct->scx.core_sched_at is added. This is a timestamp which can be
  used to order tasks. Depending on whether the BPF scheduler implements
  custom ordering, it tracks either global FIFO ordering of all tasks or
  local-DSQ ordering within the dispatched tasks on a CPU.

* prio_less() is updated to call scx_prio_less() when comparing SCX tasks.
  scx_prio_less() calls ops.core_sched_before() if available or uses the
  core_sched_at timestamp. For global FIFO ordering, the BPF scheduler
  doesn't need to do anything. Otherwise, it should implement
  ops.core_sched_before() which reflects the ordering.

* When core-sched is enabled, balance_scx() balances all SMT siblings so
  that they all have tasks dispatched if necessary before pick_task_scx() is
  called. pick_task_scx() picks between the current task and the first
  dispatched task on the local DSQ based on availability and the
  core_sched_at timestamps. Note that FIFO ordering is expected among the
  already dispatched tasks whether running or on the local DSQ, so this path
  always compares core_sched_at instead of calling into
  ops.core_sched_before().

qmap_core_sched_before() is added to scx_qmap. It scales the
distances from the heads of the queues to compare the tasks across different
priority queues and seems to behave as expected.

v3: * Fixed build error when !CONFIG_SCHED_SMT reported by Andrea Righi.

v2: * Sched core added the const qualifiers to prio_less task arguments.
      Explicitly drop them for ops.core_sched_before() task arguments. BPF
      enforces access control through the verifier, so the qualifier isn't
      actually operative and only gets in the way when interacting with
      various helpers.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Reviewed-by: Josh Don <joshdon@google.com>
Cc: Andrea Righi <andrea.righi@canonical.com>
---
 include/linux/sched/ext.h      |  21 ++++
 kernel/Kconfig.preempt         |   2 +-
 kernel/sched/core.c            |  10 +-
 kernel/sched/ext.c             | 220 +++++++++++++++++++++++++++++++--
 kernel/sched/ext.h             |  13 ++
 tools/sched_ext/scx_qmap.bpf.c |  87 ++++++++++++-
 tools/sched_ext/scx_qmap.c     |   5 +-
 7 files changed, 341 insertions(+), 17 deletions(-)

diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index c17957bd75df..9ef4c3a90199 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -316,6 +316,24 @@ struct sched_ext_ops {
 	 */
 	bool (*yield)(struct task_struct *from, struct task_struct *to);
 
+	/**
+	 * core_sched_before - Task ordering for core-sched
+	 * @a: task A
+	 * @b: task B
+	 *
+	 * Used by core-sched to determine the ordering between two tasks. See
+	 * Documentation/admin-guide/hw-vuln/core-scheduling.rst for details on
+	 * core-sched.
+	 *
+	 * Both @a and @b are runnable and may or may not currently be queued on
+	 * the BPF scheduler. Should return %true if @a should run before @b.
+	 * %false if there's no required ordering or @b should run before @a.
+	 *
+	 * If not specified, the default is ordering them according to when they
+	 * became runnable.
+	 */
+	bool (*core_sched_before)(struct task_struct *a,struct task_struct *b);
+
 	/**
 	 * set_weight - Set task weight
 	 * @p: task to set weight for
@@ -627,6 +645,9 @@ struct sched_ext_entity {
 	struct task_struct	*kf_tasks[2];	/* see SCX_CALL_OP_TASK() */
 	atomic64_t		ops_state;
 	unsigned long		runnable_at;
+#ifdef CONFIG_SCHED_CORE
+	u64			core_sched_at;	/* see scx_prio_less() */
+#endif
 
 	/* BPF scheduler modifiable fields */
 
diff --git a/kernel/Kconfig.preempt b/kernel/Kconfig.preempt
index 0afcda19bc50..e12a057ead7b 100644
--- a/kernel/Kconfig.preempt
+++ b/kernel/Kconfig.preempt
@@ -135,7 +135,7 @@ config SCHED_CORE
 
 config SCHED_CLASS_EXT
 	bool "Extensible Scheduling Class"
-	depends on BPF_SYSCALL && BPF_JIT && !SCHED_CORE
+	depends on BPF_SYSCALL && BPF_JIT
 	help
 	  This option enables a new scheduler class sched_ext (SCX), which
 	  allows scheduling policies to be implemented as BPF programs to
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 878e84694a6e..972996c05263 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -167,7 +167,10 @@ static inline int __task_prio(const struct task_struct *p)
 	if (p->sched_class == &idle_sched_class)
 		return MAX_RT_PRIO + NICE_WIDTH; /* 140 */
 
-	return MAX_RT_PRIO + MAX_NICE; /* 120, squash fair */
+	if (task_on_scx(p))
+		return MAX_RT_PRIO + MAX_NICE + 1; /* 120, squash ext */
+
+	return MAX_RT_PRIO + MAX_NICE; /* 119, squash fair */
 }
 
 /*
@@ -196,6 +199,11 @@ static inline bool prio_less(const struct task_struct *a,
 	if (pa == MAX_RT_PRIO + MAX_NICE)	/* fair */
 		return cfs_prio_less(a, b, in_fi);
 
+#ifdef CONFIG_SCHED_CLASS_EXT
+	if (pa == MAX_RT_PRIO + MAX_NICE + 1)	/* ext */
+		return scx_prio_less(a, b, in_fi);
+#endif
+
 	return false;
 }
 
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index e54d8c7d19a9..115d4d76cbc6 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -529,6 +529,49 @@ static int ops_sanitize_err(const char *ops_name, s32 err)
 	return -EPROTO;
 }
 
+/**
+ * touch_core_sched - Update timestamp used for core-sched task ordering
+ * @rq: rq to read clock from, must be locked
+ * @p: task to update the timestamp for
+ *
+ * Update @p->scx.core_sched_at timestamp. This is used by scx_prio_less() to
+ * implement global or local-DSQ FIFO ordering for core-sched. Should be called
+ * when a task becomes runnable and its turn on the CPU ends (e.g. slice
+ * exhaustion).
+ */
+static void touch_core_sched(struct rq *rq, struct task_struct *p)
+{
+#ifdef CONFIG_SCHED_CORE
+	/*
+	 * It's okay to update the timestamp spuriously. Use
+	 * sched_core_disabled() which is cheaper than enabled().
+	 */
+	if (!sched_core_disabled())
+		p->scx.core_sched_at = rq_clock_task(rq);
+#endif
+}
+
+/**
+ * touch_core_sched_dispatch - Update core-sched timestamp on dispatch
+ * @rq: rq to read clock from, must be locked
+ * @p: task being dispatched
+ *
+ * If the BPF scheduler implements custom core-sched ordering via
+ * ops.core_sched_before(), @p->scx.core_sched_at is used to implement FIFO
+ * ordering within each local DSQ. This function is called from dispatch paths
+ * and updates @p->scx.core_sched_at if custom core-sched ordering is in effect.
+ */
+static void touch_core_sched_dispatch(struct rq *rq, struct task_struct *p)
+{
+	lockdep_assert_rq_held(rq);
+	assert_clock_updated(rq);
+
+#ifdef CONFIG_SCHED_CORE
+	if (SCX_HAS_OP(core_sched_before))
+		touch_core_sched(rq, p);
+#endif
+}
+
 static void update_curr_scx(struct rq *rq)
 {
 	struct task_struct *curr = rq->curr;
@@ -544,8 +587,11 @@ static void update_curr_scx(struct rq *rq)
 	account_group_exec_runtime(curr, delta_exec);
 	cgroup_account_cputime(curr, delta_exec);
 
-	if (curr->scx.slice != SCX_SLICE_INF)
+	if (curr->scx.slice != SCX_SLICE_INF) {
 		curr->scx.slice -= min(curr->scx.slice, delta_exec);
+		if (!curr->scx.slice)
+			touch_core_sched(rq, curr);
+	}
 }
 
 static void dispatch_enqueue(struct scx_dispatch_q *dsq, struct task_struct *p,
@@ -701,6 +747,8 @@ static void direct_dispatch(struct task_struct *ddsp_task, struct task_struct *p
 		return;
 	}
 
+	touch_core_sched_dispatch(task_rq(p), p);
+
 	dsq = find_dsq_for_dispatch(task_rq(p), dsq_id, p);
 	dispatch_enqueue(dsq, p, enq_flags | SCX_ENQ_CLEAR_OPSS);
 
@@ -784,12 +832,19 @@ static void do_enqueue_task(struct rq *rq, struct task_struct *p, u64 enq_flags,
 	return;
 
 local:
+	/*
+	 * For task-ordering, slice refill must be treated as implying the end
+	 * of the current slice. Otherwise, the longer @p stays on the CPU, the
+	 * higher priority it becomes from scx_prio_less()'s POV.
+	 */
+	touch_core_sched(rq, p);
 	p->scx.slice = SCX_SLICE_DFL;
 local_norefill:
 	dispatch_enqueue(&rq->scx.local_dsq, p, enq_flags);
 	return;
 
 global:
+	touch_core_sched(rq, p);	/* see the comment in local: */
 	p->scx.slice = SCX_SLICE_DFL;
 	dispatch_enqueue(&scx_dsq_global, p, enq_flags);
 }
@@ -846,6 +901,9 @@ static void enqueue_task_scx(struct rq *rq, struct task_struct *p, int enq_flags
 	if (SCX_HAS_OP(runnable))
 		SCX_CALL_OP_TASK(SCX_KF_REST, runnable, p, enq_flags);
 
+	if (enq_flags & SCX_ENQ_WAKEUP)
+		touch_core_sched(rq, p);
+
 	do_enqueue_task(rq, p, enq_flags, sticky_cpu);
 }
 
@@ -1296,6 +1354,7 @@ static void finish_dispatch(struct rq *rq, struct rq_flags *rf,
 	struct scx_dispatch_q *dsq;
 	u64 opss;
 
+	touch_core_sched_dispatch(rq, p);
 retry:
 	/*
 	 * No need for _acquire here. @p is accessed only after a successful
@@ -1373,8 +1432,8 @@ static void flush_dispatch_buf(struct rq *rq, struct rq_flags *rf)
 	dspc->buf_cursor = 0;
 }
 
-static int balance_scx(struct rq *rq, struct task_struct *prev,
-		       struct rq_flags *rf)
+static int balance_one(struct rq *rq, struct task_struct *prev,
+		       struct rq_flags *rf, bool local)
 {
 	struct scx_rq *scx_rq = &rq->scx;
 	struct scx_dsp_ctx *dspc = this_cpu_ptr(&scx_dsp_ctx);
@@ -1398,7 +1457,7 @@ static int balance_scx(struct rq *rq, struct task_struct *prev,
 	}
 
 	if (prev_on_scx) {
-		WARN_ON_ONCE(prev->scx.flags & SCX_TASK_BAL_KEEP);
+		WARN_ON_ONCE(local && (prev->scx.flags & SCX_TASK_BAL_KEEP));
 		update_curr_scx(rq);
 
 		/*
@@ -1410,10 +1469,16 @@ static int balance_scx(struct rq *rq, struct task_struct *prev,
 		 *
 		 * See scx_ops_disable_workfn() for the explanation on the
 		 * disabling() test.
+		 *
+		 * When balancing a remote CPU for core-sched, there won't be a
+		 * following put_prev_task_scx() call and we don't own
+		 * %SCX_TASK_BAL_KEEP. Instead, pick_task_scx() will test the
+		 * same conditions later and pick @rq->curr accordingly.
 		 */
 		if ((prev->scx.flags & SCX_TASK_QUEUED) &&
 		    prev->scx.slice && !scx_ops_disabling()) {
-			prev->scx.flags |= SCX_TASK_BAL_KEEP;
+			if (local)
+				prev->scx.flags |= SCX_TASK_BAL_KEEP;
 			return 1;
 		}
 	}
@@ -1469,10 +1534,56 @@ static int balance_scx(struct rq *rq, struct task_struct *prev,
 	return 0;
 }
 
+static int balance_scx(struct rq *rq, struct task_struct *prev,
+		       struct rq_flags *rf)
+{
+	int ret;
+
+	ret = balance_one(rq, prev, rf, true);
+
+#ifdef CONFIG_SCHED_SMT
+	/*
+	 * When core-sched is enabled, this ops.balance() call will be followed
+	 * by put_prev_scx() and pick_task_scx() on this CPU and pick_task_scx()
+	 * on the SMT siblings. Balance the siblings too.
+	 */
+	if (sched_core_enabled(rq)) {
+		const struct cpumask *smt_mask = cpu_smt_mask(cpu_of(rq));
+		int scpu;
+
+		for_each_cpu_andnot(scpu, smt_mask, cpumask_of(cpu_of(rq))) {
+			struct rq *srq = cpu_rq(scpu);
+			struct rq_flags srf;
+			struct task_struct *sprev = srq->curr;
+
+			/*
+			 * While core-scheduling, rq lock is shared among
+			 * siblings but the debug annotations and rq clock
+			 * aren't. Do pinning dance to transfer the ownership.
+			 */
+			WARN_ON_ONCE(__rq_lockp(rq) != __rq_lockp(srq));
+			rq_unpin_lock(rq, rf);
+			rq_pin_lock(srq, &srf);
+
+			update_rq_clock(srq);
+			balance_one(srq, sprev, &srf, false);
+
+			rq_unpin_lock(srq, &srf);
+			rq_repin_lock(rq, rf);
+		}
+	}
+#endif
+	return ret;
+}
+
 static void set_next_task_scx(struct rq *rq, struct task_struct *p, bool first)
 {
 	if (p->scx.flags & SCX_TASK_QUEUED) {
-		WARN_ON_ONCE(atomic64_read(&p->scx.ops_state) != SCX_OPSS_NONE);
+		/*
+		 * Core-sched might decide to execute @p before it is
+		 * dispatched. Call ops_dequeue() to notify the BPF scheduler.
+		 */
+		ops_dequeue(p, SCX_DEQ_CORE_SCHED_EXEC);
 		dispatch_dequeue(&rq->scx, p);
 	}
 
@@ -1555,7 +1666,8 @@ static void put_prev_task_scx(struct rq *rq, struct task_struct *p)
 		/*
 		 * If @p has slice left and balance_scx() didn't tag it for
 		 * keeping, @p is getting preempted by a higher priority
-		 * scheduler class. Leave it at the head of the local DSQ.
+		 * scheduler class or core-sched forcing a different task. Leave
+		 * it at the head of the local DSQ.
 		 */
 		if (p->scx.slice && !scx_ops_disabling()) {
 			dispatch_enqueue(&rq->scx.local_dsq, p, SCX_ENQ_HEAD);
@@ -1612,6 +1724,84 @@ static struct task_struct *pick_next_task_scx(struct rq *rq)
 	return p;
 }
 
+#ifdef CONFIG_SCHED_CORE
+/**
+ * scx_prio_less - Task ordering for core-sched
+ * @a: task A
+ * @b: task B
+ *
+ * Core-sched is implemented as an additional scheduling layer on top of the
+ * usual sched_class'es and needs to find out the expected task ordering. For
+ * SCX, core-sched calls this function to interrogate the task ordering.
+ *
+ * Unless overridden by ops.core_sched_before(), @p->scx.core_sched_at is used
+ * to implement the default task ordering. The older the timestamp, the higher
+ * prority the task - the global FIFO ordering matching the default scheduling
+ * behavior.
+ *
+ * When ops.core_sched_before() is enabled, @p->scx.core_sched_at is used to
+ * implement FIFO ordering within each local DSQ. See pick_task_scx().
+ */
+bool scx_prio_less(const struct task_struct *a, const struct task_struct *b,
+		   bool in_fi)
+{
+	/*
+	 * The const qualifiers are dropped from task_struct pointers when
+	 * calling ops.core_sched_before(). Accesses are controlled by the
+	 * verifier.
+	 */
+	if (SCX_HAS_OP(core_sched_before) && !scx_ops_disabling())
+		return SCX_CALL_OP_2TASKS_RET(SCX_KF_REST, core_sched_before,
+					      (struct task_struct *)a,
+					      (struct task_struct *)b);
+	else
+		return time_after64(a->scx.core_sched_at, b->scx.core_sched_at);
+}
+
+/**
+ * pick_task_scx - Pick a candidate task for core-sched
+ * @rq: rq to pick the candidate task from
+ *
+ * Core-sched calls this function on each SMT sibling to determine the next
+ * tasks to run on the SMT siblings. balance_one() has been called on all
+ * siblings and put_prev_task_scx() has been called only for the current CPU.
+ *
+ * As put_prev_task_scx() hasn't been called on remote CPUs, we can't just look
+ * at the first task in the local dsq. @rq->curr has to be considered explicitly
+ * to mimic %SCX_TASK_BAL_KEEP.
+ */
+static struct task_struct *pick_task_scx(struct rq *rq)
+{
+	struct task_struct *curr = rq->curr;
+	struct task_struct *first = first_local_task(rq);
+
+	if (curr->scx.flags & SCX_TASK_QUEUED) {
+		/* is curr the only runnable task? */
+		if (!first)
+			return curr;
+
+		/*
+		 * Does curr trump first? We can always go by core_sched_at for
+		 * this comparison as it represents global FIFO ordering when
+		 * the default core-sched ordering is used and local-DSQ FIFO
+		 * ordering otherwise.
+		 *
+		 * We can have a task with an earlier timestamp on the DSQ. For
+		 * example, when a current task is preempted by a sibling
+		 * picking a different cookie, the task would be requeued at the
+		 * head of the local DSQ with an earlier timestamp than the
+		 * core-sched picked next task. Besides, the BPF scheduler may
+		 * dispatch any tasks to the local DSQ anytime.
+		 */
+		if (curr->scx.slice && time_before64(curr->scx.core_sched_at,
+						     first->scx.core_sched_at))
+			return curr;
+	}
+
+	return first;	/* this may be %NULL */
+}
+#endif	/* CONFIG_SCHED_CORE */
+
 static enum scx_cpu_preempt_reason
 preempt_reason_from_class(const struct sched_class *class)
 {
@@ -1921,11 +2111,13 @@ static void task_tick_scx(struct rq *rq, struct task_struct *curr, int queued)
 	update_curr_scx(rq);
 
 	/*
-	 * While disabling, always resched as we can't trust the slice
-	 * management.
+	 * While disabling, always resched and refresh core-sched timestamp as
+	 * we can't trust the slice management or ops.core_sched_before().
 	 */
-	if (scx_ops_disabling())
+	if (scx_ops_disabling()) {
 		curr->scx.slice = 0;
+		touch_core_sched(rq, curr);
+	}
 
 	if (!curr->scx.slice)
 		resched_curr(rq);
@@ -2388,6 +2580,10 @@ DEFINE_SCHED_CLASS(ext) = {
 	.rq_offline		= rq_offline_scx,
 #endif
 
+#ifdef CONFIG_SCHED_CORE
+	.pick_task		= pick_task_scx,
+#endif
+
 	.task_tick		= task_tick_scx,
 
 	.switching_to		= switching_to_scx,
@@ -2716,9 +2912,11 @@ static void scx_ops_disable_workfn(struct kthread_work *work)
 	 *
 	 * b. balance_scx() never sets %SCX_TASK_BAL_KEEP as the slice value
 	 *    can't be trusted. Whenever a tick triggers, the running task is
-	 *    rotated to the tail of the queue.
+	 *    rotated to the tail of the queue with core_sched_at touched.
 	 *
 	 * c. pick_next_task() suppresses zero slice warning.
+	 *
+	 * d. scx_prio_less() reverts to the default core_sched_at order.
 	 */
 	scx_ops.enqueue = scx_ops_fallback_enqueue;
 	scx_ops.dispatch = scx_ops_fallback_dispatch;
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index c71bf641f7a5..cd14970c239d 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -68,6 +68,14 @@ enum scx_enq_flags {
 enum scx_deq_flags {
 	/* expose select DEQUEUE_* flags as enums */
 	SCX_DEQ_SLEEP		= DEQUEUE_SLEEP,
+
+	/* high 32bits are SCX specific */
+
+	/*
+	 * The generic core-sched layer decided to execute the task even though
+	 * it hasn't been dispatched yet. Dequeue from the BPF side.
+	 */
+	SCX_DEQ_CORE_SCHED_EXEC	= 1LLU << 32,
 };
 
 enum scx_pick_idle_cpu_flags {
@@ -193,6 +201,11 @@ static inline const struct sched_class *next_active_class(const struct sched_cla
 	for_active_class_range(class, (prev_class) > &ext_sched_class ?		\
 			       &ext_sched_class : (prev_class), (end_class))
 
+#ifdef CONFIG_SCHED_CORE
+bool scx_prio_less(const struct task_struct *a, const struct task_struct *b,
+		   bool in_fi);
+#endif
+
 #else	/* CONFIG_SCHED_CLASS_EXT */
 
 #define scx_enabled()		false
diff --git a/tools/sched_ext/scx_qmap.bpf.c b/tools/sched_ext/scx_qmap.bpf.c
index bfffbfd3368b..b6365df0fb64 100644
--- a/tools/sched_ext/scx_qmap.bpf.c
+++ b/tools/sched_ext/scx_qmap.bpf.c
@@ -13,6 +13,7 @@
  * - Sleepable per-task storage allocation using ops.prep_enable().
  * - Using ops.cpu_release() to handle a higher priority scheduling class taking
  *   the CPU away.
+ * - Core-sched support.
  *
  * This scheduler is primarily for demonstration and testing of sched_ext
  * features and unlikely to be useful for actual workloads.
@@ -62,9 +63,21 @@ struct {
 	},
 };
 
+/*
+ * Per-queue sequence numbers to implement core-sched ordering.
+ *
+ * Tail seq is assigned to each queued task and incremented. Head seq tracks the
+ * sequence number of the latest dispatched task. The distance between the a
+ * task's seq and the associated queue's head seq is called the queue distance
+ * and used when comparing two tasks for ordering. See qmap_core_sched_before().
+ */
+static u64 core_sched_head_seqs[5];
+static u64 core_sched_tail_seqs[5];
+
 /* Per-task scheduling context */
 struct task_ctx {
 	bool	force_local;	/* Dispatch directly to local_dsq */
+	u64	core_sched_seq;
 };
 
 struct {
@@ -84,6 +97,7 @@ struct {
 
 /* Statistics */
 unsigned long nr_enqueued, nr_dispatched, nr_reenqueued, nr_dequeued;
+unsigned long nr_core_sched_execed;
 
 s32 BPF_STRUCT_OPS(qmap_select_cpu, struct task_struct *p,
 		   s32 prev_cpu, u64 wake_flags)
@@ -150,8 +164,18 @@ void BPF_STRUCT_OPS(qmap_enqueue, struct task_struct *p, u64 enq_flags)
 		return;
 	}
 
-	/* Is select_cpu() is telling us to enqueue locally? */
-	if (tctx->force_local) {
+	/*
+	 * All enqueued tasks must have their core_sched_seq updated for correct
+	 * core-sched ordering, which is why %SCX_OPS_ENQ_LAST is specified in
+	 * qmap_ops.flags.
+	 */
+	tctx->core_sched_seq = core_sched_tail_seqs[idx]++;
+
+	/*
+	 * If qmap_select_cpu() is telling us to or this is the last runnable
+	 * task on the CPU, enqueue locally.
+	 */
+	if (tctx->force_local || (enq_flags & SCX_ENQ_LAST)) {
 		tctx->force_local = false;
 		scx_bpf_dispatch(p, SCX_DSQ_LOCAL, slice_ns, enq_flags);
 		return;
@@ -195,6 +219,19 @@ void BPF_STRUCT_OPS(qmap_enqueue, struct task_struct *p, u64 enq_flags)
 void BPF_STRUCT_OPS(qmap_dequeue, struct task_struct *p, u64 deq_flags)
 {
 	__sync_fetch_and_add(&nr_dequeued, 1);
+	if (deq_flags & SCX_DEQ_CORE_SCHED_EXEC)
+		__sync_fetch_and_add(&nr_core_sched_execed, 1);
+}
+
+static void update_core_sched_head_seq(struct task_struct *p)
+{
+	struct task_ctx *tctx = bpf_task_storage_get(&task_ctx_stor, p, 0, 0);
+	int idx = weight_to_idx(p->scx.weight);
+
+	if (tctx)
+		core_sched_head_seqs[idx] = tctx->core_sched_seq;
+	else
+		scx_bpf_error("task_ctx lookup failed");
 }
 
 void BPF_STRUCT_OPS(qmap_dispatch, s32 cpu, struct task_struct *prev)
@@ -247,6 +284,7 @@ void BPF_STRUCT_OPS(qmap_dispatch, s32 cpu, struct task_struct *prev)
 
 			p = bpf_task_from_pid(pid);
 			if (p) {
+				update_core_sched_head_seq(p);
 				__sync_fetch_and_add(&nr_dispatched, 1);
 				scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, slice_ns, 0);
 				bpf_task_release(p);
@@ -258,6 +296,49 @@ void BPF_STRUCT_OPS(qmap_dispatch, s32 cpu, struct task_struct *prev)
 	}
 }
 
+/*
+ * The distance from the head of the queue scaled by the weight of the queue.
+ * The lower the number, the older the task and the higher the priority.
+ */
+static s64 task_qdist(struct task_struct *p)
+{
+	int idx = weight_to_idx(p->scx.weight);
+	struct task_ctx *tctx;
+	s64 qdist;
+
+	tctx = bpf_task_storage_get(&task_ctx_stor, p, 0, 0);
+	if (!tctx) {
+		scx_bpf_error("task_ctx lookup failed");
+		return 0;
+	}
+
+	qdist = tctx->core_sched_seq - core_sched_head_seqs[idx];
+
+	/*
+	 * As queue index increments, the priority doubles. The queue w/ index 3
+	 * is dispatched twice more frequently than 2. Reflect the difference by
+	 * scaling qdists accordingly. Note that the shift amount needs to be
+	 * flipped depending on the sign to avoid flipping priority direction.
+	 */
+	if (qdist >= 0)
+		return qdist << (4 - idx);
+	else
+		return qdist << idx;
+}
+
+/*
+ * This is called to determine the task ordering when core-sched is picking
+ * tasks to execute on SMT siblings and should encode about the same ordering as
+ * the regular scheduling path. Use the priority-scaled distances from the head
+ * of the queues to compare the two tasks which should be consistent with the
+ * dispatch path behavior.
+ */
+bool BPF_STRUCT_OPS(qmap_core_sched_before,
+		    struct task_struct *a, struct task_struct *b)
+{
+	return task_qdist(a) > task_qdist(b);
+}
+
 void BPF_STRUCT_OPS(qmap_cpu_release, s32 cpu, struct scx_cpu_release_args *args)
 {
 	u32 cnt;
@@ -309,10 +390,12 @@ struct sched_ext_ops qmap_ops = {
 	.enqueue		= (void *)qmap_enqueue,
 	.dequeue		= (void *)qmap_dequeue,
 	.dispatch		= (void *)qmap_dispatch,
+	.core_sched_before	= (void *)qmap_core_sched_before,
 	.cpu_release		= (void *)qmap_cpu_release,
 	.prep_enable		= (void *)qmap_prep_enable,
 	.init			= (void *)qmap_init,
 	.exit			= (void *)qmap_exit,
+	.flags			= SCX_OPS_ENQ_LAST,
 	.timeout_ms		= 5000U,
 	.name			= "qmap",
 };
diff --git a/tools/sched_ext/scx_qmap.c b/tools/sched_ext/scx_qmap.c
index b47be72ee66e..0a02aa166b47 100644
--- a/tools/sched_ext/scx_qmap.c
+++ b/tools/sched_ext/scx_qmap.c
@@ -92,9 +92,10 @@ int main(int argc, char **argv)
 		long nr_enqueued = skel->bss->nr_enqueued;
 		long nr_dispatched = skel->bss->nr_dispatched;
 
-		printf("enq=%lu, dsp=%lu, delta=%ld, reenq=%lu, deq=%lu\n",
+		printf("enq=%lu, dsp=%lu, delta=%ld, reenq=%lu, deq=%lu, core=%lu\n",
 		       nr_enqueued, nr_dispatched, nr_enqueued - nr_dispatched,
-		       skel->bss->nr_reenqueued, skel->bss->nr_dequeued);
+		       skel->bss->nr_reenqueued, skel->bss->nr_dequeued,
+		       skel->bss->nr_core_sched_execed);
 		fflush(stdout);
 		sleep(1);
 	}
-- 
2.41.0

[PATCH 31/34] sched_ext: Add vtime-ordered priority queue to dispatch_q's

[Tejun Heo]

Currently, a dsq is always a FIFO. A task which is dispatched earlier gets
consumed or executed earlier. While this is sufficient when dsq's are used
for simple staging areas for tasks which are ready to execute, it'd make
dsq's a lot more useful if they can implement custom ordering.

This patch adds a vtime-ordered priority queue to dsq's. When the BPF
scheduler dispatches a task with the new scx_bpf_dispatch_vtime() helper, it
can specify the vtime tha the task should be inserted at and the task is
inserted into the priority queue in the dsq which is ordered according to
time_before64() comparison of the vtime values. When executing or consuming
the dsq, the FIFO is always processed first and the priority queue is
processed iff the FIFO is empty.

The design decision was made to allow both FIFO and priority queue to be
available at the same timeq for all dsq's for three reasons. First, the new
priority queue is useful for the local dsq's too but they also need the FIFO
when consuming tasks from other dsq's as the vtimes may not be comparable
across them. Second, the interface surface is smaller this way - the only
additional interface necessary is scx_bpf_dispsatch_vtime(). Third, the
overhead isn't meaningfully different whether they're available at the same
time or not.

This makes it very easy for the BPF schedulers to implement proper vtime
based scheduling within each dsq very easy and efficient at a negligible
cost in terms of code complexity and overhead.

scx_simple and scx_example_flatcg are updated to default to weighted
vtime scheduling (the latter within each cgroup). FIFO scheduling can be
selected with -f option.

v2: p->scx.dsq_vtime was not initialized on load or across cgroup migrations
    leading to some tasks being stalled for extended period of time
    depending on how saturated the machine is. Fixed.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
---
 include/linux/sched/ext.h        |  16 +++-
 init/init_task.c                 |   2 +-
 kernel/sched/core.c              |   3 +-
 kernel/sched/ext.c               | 137 +++++++++++++++++++++++++++----
 kernel/sched/ext.h               |   1 +
 tools/sched_ext/scx_common.bpf.h |   1 +
 tools/sched_ext/scx_flatcg.bpf.c |  72 +++++++++++++++-
 tools/sched_ext/scx_flatcg.c     |   6 +-
 tools/sched_ext/scx_simple.bpf.c |  73 +++++++++++++++-
 tools/sched_ext/scx_simple.c     |   8 +-
 10 files changed, 293 insertions(+), 26 deletions(-)

diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 9ef4c3a90199..446821073bd1 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -582,6 +582,7 @@ struct sched_ext_ops {
 struct scx_dispatch_q {
 	raw_spinlock_t		lock;
 	struct list_head	fifo;	/* processed in dispatching order */
+	struct rb_root_cached	priq;	/* processed in p->scx.dsq_vtime order */
 	u32			nr;
 	u64			id;
 	struct rhash_head	hash_node;
@@ -594,6 +595,7 @@ enum scx_ent_flags {
 	SCX_TASK_QUEUED		= 1 << 0, /* on ext runqueue */
 	SCX_TASK_BAL_KEEP	= 1 << 1, /* balance decided to keep current */
 	SCX_TASK_ENQ_LOCAL	= 1 << 2, /* used by scx_select_cpu_dfl() to set SCX_ENQ_LOCAL */
+	SCX_TASK_ON_DSQ_PRIQ	= 1 << 3, /* task is queued on the priority queue of a dsq */
 
 	SCX_TASK_OPS_PREPPED	= 1 << 8, /* prepared for BPF scheduler enable */
 	SCX_TASK_OPS_ENABLED	= 1 << 9, /* task has BPF scheduler enabled */
@@ -635,7 +637,10 @@ enum scx_kf_mask {
  */
 struct sched_ext_entity {
 	struct scx_dispatch_q	*dsq;
-	struct list_head	dsq_node;
+	struct {
+		struct list_head	fifo;	/* dispatch order */
+		struct rb_node		priq;	/* p->scx.dsq_vtime order */
+	} dsq_node;
 	struct list_head	watchdog_node;
 	u32			flags;		/* protected by rq lock */
 	u32			weight;
@@ -663,6 +668,15 @@ struct sched_ext_entity {
 	 */
 	u64			slice;
 
+	/*
+	 * Used to order tasks when dispatching to the vtime-ordered priority
+	 * queue of a dsq. This is usually set through scx_bpf_dispatch_vtime()
+	 * but can also be modified directly by the BPF scheduler. Modifying it
+	 * while a task is queued on a dsq may mangle the ordering and is not
+	 * recommended.
+	 */
+	u64			dsq_vtime;
+
 	/*
 	 * If set, reject future sched_setscheduler(2) calls updating the policy
 	 * to %SCHED_EXT with -%EACCES.
diff --git a/init/init_task.c b/init/init_task.c
index 913194aab623..7ea89ccd0cf1 100644
--- a/init/init_task.c
+++ b/init/init_task.c
@@ -105,7 +105,7 @@ struct task_struct init_task
 #endif
 #ifdef CONFIG_SCHED_CLASS_EXT
 	.scx		= {
-		.dsq_node	= LIST_HEAD_INIT(init_task.scx.dsq_node),
+		.dsq_node.fifo	= LIST_HEAD_INIT(init_task.scx.dsq_node.fifo),
 		.watchdog_node	= LIST_HEAD_INIT(init_task.scx.watchdog_node),
 		.sticky_cpu	= -1,
 		.holding_cpu	= -1,
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 972996c05263..9128160b6264 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4518,7 +4518,8 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
 
 #ifdef CONFIG_SCHED_CLASS_EXT
 	p->scx.dsq		= NULL;
-	INIT_LIST_HEAD(&p->scx.dsq_node);
+	INIT_LIST_HEAD(&p->scx.dsq_node.fifo);
+	RB_CLEAR_NODE(&p->scx.dsq_node.priq);
 	INIT_LIST_HEAD(&p->scx.watchdog_node);
 	p->scx.flags		= 0;
 	p->scx.weight		= 0;
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 115d4d76cbc6..44ba457263bd 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -594,12 +594,25 @@ static void update_curr_scx(struct rq *rq)
 	}
 }
 
+static bool scx_dsq_priq_less(struct rb_node *node_a,
+			      const struct rb_node *node_b)
+{
+	const struct task_struct *a =
+		container_of(node_a, struct task_struct, scx.dsq_node.priq);
+	const struct task_struct *b =
+		container_of(node_b, struct task_struct, scx.dsq_node.priq);
+
+	return time_before64(a->scx.dsq_vtime, b->scx.dsq_vtime);
+}
+
 static void dispatch_enqueue(struct scx_dispatch_q *dsq, struct task_struct *p,
 			     u64 enq_flags)
 {
 	bool is_local = dsq->id == SCX_DSQ_LOCAL;
 
-	WARN_ON_ONCE(p->scx.dsq || !list_empty(&p->scx.dsq_node));
+	WARN_ON_ONCE(p->scx.dsq || !list_empty(&p->scx.dsq_node.fifo));
+	WARN_ON_ONCE((p->scx.flags & SCX_TASK_ON_DSQ_PRIQ) ||
+		     !RB_EMPTY_NODE(&p->scx.dsq_node.priq));
 
 	if (!is_local) {
 		raw_spin_lock(&dsq->lock);
@@ -612,10 +625,16 @@ static void dispatch_enqueue(struct scx_dispatch_q *dsq, struct task_struct *p,
 		}
 	}
 
-	if (enq_flags & (SCX_ENQ_HEAD | SCX_ENQ_PREEMPT))
-		list_add(&p->scx.dsq_node, &dsq->fifo);
-	else
-		list_add_tail(&p->scx.dsq_node, &dsq->fifo);
+	if (enq_flags & SCX_ENQ_DSQ_PRIQ) {
+		p->scx.flags |= SCX_TASK_ON_DSQ_PRIQ;
+		rb_add_cached(&p->scx.dsq_node.priq, &dsq->priq,
+			      scx_dsq_priq_less);
+	} else {
+		if (enq_flags & (SCX_ENQ_HEAD | SCX_ENQ_PREEMPT))
+			list_add(&p->scx.dsq_node.fifo, &dsq->fifo);
+		else
+			list_add_tail(&p->scx.dsq_node.fifo, &dsq->fifo);
+	}
 	dsq->nr++;
 	p->scx.dsq = dsq;
 
@@ -644,13 +663,31 @@ static void dispatch_enqueue(struct scx_dispatch_q *dsq, struct task_struct *p,
 	}
 }
 
+static void task_unlink_from_dsq(struct task_struct *p,
+				 struct scx_dispatch_q *dsq)
+{
+	if (p->scx.flags & SCX_TASK_ON_DSQ_PRIQ) {
+		rb_erase_cached(&p->scx.dsq_node.priq, &dsq->priq);
+		RB_CLEAR_NODE(&p->scx.dsq_node.priq);
+		p->scx.flags &= ~SCX_TASK_ON_DSQ_PRIQ;
+	} else {
+		list_del_init(&p->scx.dsq_node.fifo);
+	}
+}
+
+static bool task_linked_on_dsq(struct task_struct *p)
+{
+	return !list_empty(&p->scx.dsq_node.fifo) ||
+		!RB_EMPTY_NODE(&p->scx.dsq_node.priq);
+}
+
 static void dispatch_dequeue(struct scx_rq *scx_rq, struct task_struct *p)
 {
 	struct scx_dispatch_q *dsq = p->scx.dsq;
 	bool is_local = dsq == &scx_rq->local_dsq;
 
 	if (!dsq) {
-		WARN_ON_ONCE(!list_empty(&p->scx.dsq_node));
+		WARN_ON_ONCE(task_linked_on_dsq(p));
 		/*
 		 * When dispatching directly from the BPF scheduler to a local
 		 * DSQ, the task isn't associated with any DSQ but
@@ -671,8 +708,8 @@ static void dispatch_dequeue(struct scx_rq *scx_rq, struct task_struct *p)
 	*/
 	if (p->scx.holding_cpu < 0) {
 		/* @p must still be on @dsq, dequeue */
-		WARN_ON_ONCE(list_empty(&p->scx.dsq_node));
-		list_del_init(&p->scx.dsq_node);
+		WARN_ON_ONCE(!task_linked_on_dsq(p));
+		task_unlink_from_dsq(p, dsq);
 		dsq->nr--;
 	} else {
 		/*
@@ -681,7 +718,7 @@ static void dispatch_dequeue(struct scx_rq *scx_rq, struct task_struct *p)
 		 * holding_cpu which tells dispatch_to_local_dsq() that it lost
 		 * the race.
 		 */
-		WARN_ON_ONCE(!list_empty(&p->scx.dsq_node));
+		WARN_ON_ONCE(task_linked_on_dsq(p));
 		p->scx.holding_cpu = -1;
 	}
 	p->scx.dsq = NULL;
@@ -1145,33 +1182,52 @@ static void dispatch_to_local_dsq_unlock(struct rq *rq, struct rq_flags *rf,
 #endif	/* CONFIG_SMP */
 
 
+static bool task_can_run_on_rq(struct task_struct *p, struct rq *rq)
+{
+	return likely(test_rq_online(rq)) && !is_migration_disabled(p) &&
+		cpumask_test_cpu(cpu_of(rq), p->cpus_ptr);
+}
+
 static bool consume_dispatch_q(struct rq *rq, struct rq_flags *rf,
 			       struct scx_dispatch_q *dsq)
 {
 	struct scx_rq *scx_rq = &rq->scx;
 	struct task_struct *p;
+	struct rb_node *rb_node;
 	struct rq *task_rq;
 	bool moved = false;
 retry:
-	if (list_empty(&dsq->fifo))
+	if (list_empty(&dsq->fifo) && !rb_first_cached(&dsq->priq))
 		return false;
 
 	raw_spin_lock(&dsq->lock);
-	list_for_each_entry(p, &dsq->fifo, scx.dsq_node) {
+
+	list_for_each_entry(p, &dsq->fifo, scx.dsq_node.fifo) {
+		task_rq = task_rq(p);
+		if (rq == task_rq)
+			goto this_rq;
+		if (task_can_run_on_rq(p, rq))
+			goto remote_rq;
+	}
+
+	for (rb_node = rb_first_cached(&dsq->priq); rb_node;
+	     rb_node = rb_next(rb_node)) {
+		p = container_of(rb_node, struct task_struct, scx.dsq_node.priq);
 		task_rq = task_rq(p);
 		if (rq == task_rq)
 			goto this_rq;
-		if (likely(test_rq_online(rq)) && !is_migration_disabled(p) &&
-		    cpumask_test_cpu(cpu_of(rq), p->cpus_ptr))
+		if (task_can_run_on_rq(p, rq))
 			goto remote_rq;
 	}
+
 	raw_spin_unlock(&dsq->lock);
 	return false;
 
 this_rq:
 	/* @dsq is locked and @p is on this rq */
 	WARN_ON_ONCE(p->scx.holding_cpu >= 0);
-	list_move_tail(&p->scx.dsq_node, &scx_rq->local_dsq.fifo);
+	task_unlink_from_dsq(p, dsq);
+	list_add_tail(&p->scx.dsq_node.fifo, &scx_rq->local_dsq.fifo);
 	dsq->nr--;
 	scx_rq->local_dsq.nr++;
 	p->scx.dsq = &scx_rq->local_dsq;
@@ -1188,7 +1244,7 @@ static bool consume_dispatch_q(struct rq *rq, struct rq_flags *rf,
 	 * move_task_to_local_dsq().
 	 */
 	WARN_ON_ONCE(p->scx.holding_cpu >= 0);
-	list_del_init(&p->scx.dsq_node);
+	task_unlink_from_dsq(p, dsq);
 	dsq->nr--;
 	p->scx.holding_cpu = raw_smp_processor_id();
 	raw_spin_unlock(&dsq->lock);
@@ -1692,8 +1748,18 @@ static void put_prev_task_scx(struct rq *rq, struct task_struct *p)
 
 static struct task_struct *first_local_task(struct rq *rq)
 {
-	return list_first_entry_or_null(&rq->scx.local_dsq.fifo,
-					struct task_struct, scx.dsq_node);
+	struct rb_node *rb_node;
+
+	if (!list_empty(&rq->scx.local_dsq.fifo))
+		return list_first_entry(&rq->scx.local_dsq.fifo,
+					struct task_struct, scx.dsq_node.fifo);
+
+	rb_node = rb_first_cached(&rq->scx.local_dsq.priq);
+	if (rb_node)
+		return container_of(rb_node,
+				    struct task_struct, scx.dsq_node.priq);
+
+	return NULL;
 }
 
 static struct task_struct *pick_next_task_scx(struct rq *rq)
@@ -3401,6 +3467,9 @@ static int bpf_scx_btf_struct_access(struct bpf_verifier_log *log,
 		if (off >= offsetof(struct task_struct, scx.slice) &&
 		    off + size <= offsetofend(struct task_struct, scx.slice))
 			return SCALAR_VALUE;
+		if (off >= offsetof(struct task_struct, scx.dsq_vtime) &&
+		    off + size <= offsetofend(struct task_struct, scx.dsq_vtime))
+			return SCALAR_VALUE;
 		if (off >= offsetof(struct task_struct, scx.disallow) &&
 		    off + size <= offsetofend(struct task_struct, scx.disallow))
 			return SCALAR_VALUE;
@@ -3798,8 +3867,42 @@ void scx_bpf_dispatch(struct task_struct *p, u64 dsq_id, u64 slice,
 	scx_dispatch_commit(p, dsq_id, enq_flags);
 }
 
+/**
+ * scx_bpf_dispatch_vtime - Dispatch a task into the vtime priority queue of a DSQ
+ * @p: task_struct to dispatch
+ * @dsq_id: DSQ to dispatch to
+ * @slice: duration @p can run for in nsecs
+ * @vtime: @p's ordering inside the vtime-sorted queue of the target DSQ
+ * @enq_flags: SCX_ENQ_*
+ *
+ * Dispatch @p into the vtime priority queue of the DSQ identified by @dsq_id.
+ * Tasks queued into the priority queue are ordered by @vtime and always
+ * consumed after the tasks in the FIFO queue. All other aspects are identical
+ * to scx_bpf_dispatch().
+ *
+ * @vtime ordering is according to time_before64() which considers wrapping. A
+ * numerically larger vtime may indicate an earlier position in the ordering and
+ * vice-versa.
+ */
+void scx_bpf_dispatch_vtime(struct task_struct *p, u64 dsq_id, u64 slice,
+			    u64 vtime, u64 enq_flags)
+{
+	if (!scx_dispatch_preamble(p, enq_flags))
+		return;
+
+	if (slice)
+		p->scx.slice = slice;
+	else
+		p->scx.slice = p->scx.slice ?: 1;
+
+	p->scx.dsq_vtime = vtime;
+
+	scx_dispatch_commit(p, dsq_id, enq_flags | SCX_ENQ_DSQ_PRIQ);
+}
+
 BTF_SET8_START(scx_kfunc_ids_enqueue_dispatch)
 BTF_ID_FLAGS(func, scx_bpf_dispatch, KF_RCU)
+BTF_ID_FLAGS(func, scx_bpf_dispatch_vtime, KF_RCU)
 BTF_SET8_END(scx_kfunc_ids_enqueue_dispatch)
 
 static const struct btf_kfunc_id_set scx_kfunc_set_enqueue_dispatch = {
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index cd14970c239d..821515bb6580 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -63,6 +63,7 @@ enum scx_enq_flags {
 	__SCX_ENQ_INTERNAL_MASK	= 0xffLLU << 56,
 
 	SCX_ENQ_CLEAR_OPSS	= 1LLU << 56,
+	SCX_ENQ_DSQ_PRIQ	= 1LLU << 57,
 };
 
 enum scx_deq_flags {
diff --git a/tools/sched_ext/scx_common.bpf.h b/tools/sched_ext/scx_common.bpf.h
index 8a210ab0eaab..5a6136dceb4d 100644
--- a/tools/sched_ext/scx_common.bpf.h
+++ b/tools/sched_ext/scx_common.bpf.h
@@ -58,6 +58,7 @@ s32 scx_bpf_create_dsq(u64 dsq_id, s32 node) __ksym;
 bool scx_bpf_consume(u64 dsq_id) __ksym;
 u32 scx_bpf_dispatch_nr_slots(void) __ksym;
 void scx_bpf_dispatch(struct task_struct *p, u64 dsq_id, u64 slice, u64 enq_flags) __ksym;
+void scx_bpf_dispatch_vtime(struct task_struct *p, u64 dsq_id, u64 slice, u64 vtime, u64 enq_flags) __ksym;
 void scx_bpf_kick_cpu(s32 cpu, u64 flags) __ksym;
 s32 scx_bpf_dsq_nr_queued(u64 dsq_id) __ksym;
 bool scx_bpf_test_and_clear_cpu_idle(s32 cpu) __ksym;
diff --git a/tools/sched_ext/scx_flatcg.bpf.c b/tools/sched_ext/scx_flatcg.bpf.c
index ab7cff4da7da..6d8c6f396577 100644
--- a/tools/sched_ext/scx_flatcg.bpf.c
+++ b/tools/sched_ext/scx_flatcg.bpf.c
@@ -38,6 +38,10 @@
  * this isn't a real concern especially given the performance gain. Also, there
  * are ways to mitigate the problem further by e.g. introducing an extra
  * scheduling layer on cgroup delegation boundaries.
+ *
+ * The scheduler first picks the cgroup to run and then schedule the tasks
+ * within by using nested weighted vtime scheduling by default. The
+ * cgroup-internal scheduling can be switched to FIFO with the -f option.
  */
 #include "scx_common.bpf.h"
 #include "user_exit_info.h"
@@ -47,6 +51,7 @@ char _license[] SEC("license") = "GPL";
 
 const volatile u32 nr_cpus = 32;	/* !0 for veristat, set during init */
 const volatile u64 cgrp_slice_ns = SCX_SLICE_DFL;
+const volatile bool fifo_sched;
 const volatile bool switch_partial;
 
 u64 cvtime_now;
@@ -350,7 +355,21 @@ void BPF_STRUCT_OPS(fcg_enqueue, struct task_struct *p, u64 enq_flags)
 	if (!cgc)
 		goto out_release;
 
-	scx_bpf_dispatch(p, cgrp->kn->id, SCX_SLICE_DFL, enq_flags);
+	if (fifo_sched) {
+		scx_bpf_dispatch(p, cgrp->kn->id, SCX_SLICE_DFL, enq_flags);
+	} else {
+		u64 tvtime = p->scx.dsq_vtime;
+
+		/*
+		 * Limit the amount of budget that an idling task can accumulate
+		 * to one slice.
+		 */
+		if (vtime_before(tvtime, cgc->tvtime_now - SCX_SLICE_DFL))
+			tvtime = cgc->tvtime_now - SCX_SLICE_DFL;
+
+		scx_bpf_dispatch_vtime(p, cgrp->kn->id, SCX_SLICE_DFL,
+				       tvtime, enq_flags);
+	}
 
 	cgrp_enqueued(cgrp, cgc);
 out_release:
@@ -462,12 +481,40 @@ void BPF_STRUCT_OPS(fcg_runnable, struct task_struct *p, u64 enq_flags)
 	bpf_cgroup_release(cgrp);
 }
 
+void BPF_STRUCT_OPS(fcg_running, struct task_struct *p)
+{
+	struct cgroup *cgrp;
+	struct fcg_cgrp_ctx *cgc;
+
+	if (fifo_sched)
+		return;
+
+	cgrp = scx_bpf_task_cgroup(p);
+	cgc = find_cgrp_ctx(cgrp);
+	if (cgc) {
+		/*
+		 * @cgc->tvtime_now always progresses forward as tasks start
+		 * executing. The test and update can be performed concurrently
+		 * from multiple CPUs and thus racy. Any error should be
+		 * contained and temporary. Let's just live with it.
+		 */
+		if (vtime_before(cgc->tvtime_now, p->scx.dsq_vtime))
+			cgc->tvtime_now = p->scx.dsq_vtime;
+	}
+	bpf_cgroup_release(cgrp);
+}
+
 void BPF_STRUCT_OPS(fcg_stopping, struct task_struct *p, bool runnable)
 {
 	struct fcg_task_ctx *taskc;
 	struct cgroup *cgrp;
 	struct fcg_cgrp_ctx *cgc;
 
+	/* scale the execution time by the inverse of the weight and charge */
+	if (!fifo_sched)
+		p->scx.dsq_vtime +=
+			(SCX_SLICE_DFL - p->scx.slice) * 100 / p->scx.weight;
+
 	taskc = bpf_task_storage_get(&task_ctx, p, 0, 0);
 	if (!taskc) {
 		scx_bpf_error("task_ctx lookup failed");
@@ -706,6 +753,7 @@ s32 BPF_STRUCT_OPS(fcg_prep_enable, struct task_struct *p,
 		   struct scx_enable_args *args)
 {
 	struct fcg_task_ctx *taskc;
+	struct fcg_cgrp_ctx *cgc;
 
 	/*
 	 * @p is new. Let's ensure that its task_ctx is available. We can sleep
@@ -717,6 +765,12 @@ s32 BPF_STRUCT_OPS(fcg_prep_enable, struct task_struct *p,
 		return -ENOMEM;
 
 	taskc->bypassed_at = 0;
+
+	if (!(cgc = find_cgrp_ctx(args->cgroup)))
+		return -ENOENT;
+
+	p->scx.dsq_vtime = cgc->tvtime_now;
+
 	return 0;
 }
 
@@ -804,6 +858,20 @@ void BPF_STRUCT_OPS(fcg_cgroup_exit, struct cgroup *cgrp)
 	scx_bpf_destroy_dsq(cgid);
 }
 
+void BPF_STRUCT_OPS(fcg_cgroup_move, struct task_struct *p,
+		    struct cgroup *from, struct cgroup *to)
+{
+	struct fcg_cgrp_ctx *from_cgc, *to_cgc;
+	s64 vtime_delta;
+
+	/* find_cgrp_ctx() triggers scx_ops_error() on lookup failures */
+	if (!(from_cgc = find_cgrp_ctx(from)) || !(to_cgc = find_cgrp_ctx(to)))
+		return;
+
+	vtime_delta = p->scx.dsq_vtime - from_cgc->tvtime_now;
+	p->scx.dsq_vtime = to_cgc->tvtime_now + vtime_delta;
+}
+
 s32 BPF_STRUCT_OPS(fcg_init)
 {
 	if (!switch_partial)
@@ -821,12 +889,14 @@ struct sched_ext_ops flatcg_ops = {
 	.enqueue		= (void *)fcg_enqueue,
 	.dispatch		= (void *)fcg_dispatch,
 	.runnable		= (void *)fcg_runnable,
+	.running		= (void *)fcg_running,
 	.stopping		= (void *)fcg_stopping,
 	.quiescent		= (void *)fcg_quiescent,
 	.prep_enable		= (void *)fcg_prep_enable,
 	.cgroup_set_weight	= (void *)fcg_cgroup_set_weight,
 	.cgroup_init		= (void *)fcg_cgroup_init,
 	.cgroup_exit		= (void *)fcg_cgroup_exit,
+	.cgroup_move		= (void *)fcg_cgroup_move,
 	.init			= (void *)fcg_init,
 	.exit			= (void *)fcg_exit,
 	.flags			= SCX_OPS_CGROUP_KNOB_WEIGHT | SCX_OPS_ENQ_EXITING,
diff --git a/tools/sched_ext/scx_flatcg.c b/tools/sched_ext/scx_flatcg.c
index 82afaa98d7a7..40aa464c55b1 100644
--- a/tools/sched_ext/scx_flatcg.c
+++ b/tools/sched_ext/scx_flatcg.c
@@ -26,10 +26,11 @@ const char help_fmt[] =
 "\n"
 "See the top-level comment in .bpf.c for more details.\n"
 "\n"
-"Usage: %s [-s SLICE_US] [-i INTERVAL] [-p]\n"
+"Usage: %s [-s SLICE_US] [-i INTERVAL] [-f] [-p]\n"
 "\n"
 "  -s SLICE_US   Override slice duration\n"
 "  -i INTERVAL   Report interval\n"
+"  -f            Use FIFO scheduling instead of weighted vtime scheduling\n"
 "  -p            Switch only tasks on SCHED_EXT policy intead of all\n"
 "  -h            Display this help and exit\n";
 
@@ -149,6 +150,9 @@ int main(int argc, char **argv)
 		case 'd':
 			dump_cgrps = true;
 			break;
+		case 'f':
+			skel->rodata->fifo_sched = true;
+			break;
 		case 'p':
 			skel->rodata->switch_partial = true;
 			break;
diff --git a/tools/sched_ext/scx_simple.bpf.c b/tools/sched_ext/scx_simple.bpf.c
index 6302a4ea9ea5..d4528c7da450 100644
--- a/tools/sched_ext/scx_simple.bpf.c
+++ b/tools/sched_ext/scx_simple.bpf.c
@@ -2,11 +2,20 @@
 /*
  * A simple scheduler.
  *
- * A simple global FIFO scheduler. It also demonstrates the following niceties.
+ * By default, it operates as a simple global weighted vtime scheduler and can
+ * be switched to FIFO scheduling. It also demonstrates the following niceties.
  *
  * - Statistics tracking how many tasks are queued to local and global dsq's.
  * - Termination notification for userspace.
  *
+ * While very simple, this scheduler should work reasonably well on CPUs with a
+ * uniform L3 cache topology. While preemption is not implemented, the fact that
+ * the scheduling queue is shared across all CPUs means that whatever is at the
+ * front of the queue is likely to be executed fairly quickly given enough
+ * number of CPUs. The FIFO scheduling mode may be beneficial to some workloads
+ * but comes with the usual problems with FIFO scheduling where saturating
+ * threads can easily drown out interactive ones.
+ *
  * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
  * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
  * Copyright (c) 2022 David Vernet <dvernet@meta.com>
@@ -15,8 +24,10 @@
 
 char _license[] SEC("license") = "GPL";
 
+const volatile bool fifo_sched;
 const volatile bool switch_partial;
 
+static u64 vtime_now;
 struct user_exit_info uei;
 
 struct {
@@ -33,8 +44,18 @@ static void stat_inc(u32 idx)
 		(*cnt_p)++;
 }
 
+static inline bool vtime_before(u64 a, u64 b)
+{
+	return (s64)(a - b) < 0;
+}
+
 void BPF_STRUCT_OPS(simple_enqueue, struct task_struct *p, u64 enq_flags)
 {
+	/*
+	 * If scx_select_cpu_dfl() is setting %SCX_ENQ_LOCAL, it indicates that
+	 * running @p on its CPU directly shouldn't affect fairness. Just queue
+	 * it on the local FIFO.
+	 */
 	if (enq_flags & SCX_ENQ_LOCAL) {
 		stat_inc(0);	/* count local queueing */
 		scx_bpf_dispatch(p, SCX_DSQ_LOCAL, SCX_SLICE_DFL, enq_flags);
@@ -42,7 +63,52 @@ void BPF_STRUCT_OPS(simple_enqueue, struct task_struct *p, u64 enq_flags)
 	}
 
 	stat_inc(1);	/* count global queueing */
-	scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, SCX_SLICE_DFL, enq_flags);
+
+	if (fifo_sched) {
+		scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, SCX_SLICE_DFL, enq_flags);
+	} else {
+		u64 vtime = p->scx.dsq_vtime;
+
+		/*
+		 * Limit the amount of budget that an idling task can accumulate
+		 * to one slice.
+		 */
+		if (vtime_before(vtime, vtime_now - SCX_SLICE_DFL))
+			vtime = vtime_now - SCX_SLICE_DFL;
+
+		scx_bpf_dispatch_vtime(p, SCX_DSQ_GLOBAL, SCX_SLICE_DFL, vtime,
+				       enq_flags);
+	}
+}
+
+void BPF_STRUCT_OPS(simple_running, struct task_struct *p)
+{
+	if (fifo_sched)
+		return;
+
+	/*
+	 * Global vtime always progresses forward as tasks start executing. The
+	 * test and update can be performed concurrently from multiple CPUs and
+	 * thus racy. Any error should be contained and temporary. Let's just
+	 * live with it.
+	 */
+	if (vtime_before(vtime_now, p->scx.dsq_vtime))
+		vtime_now = p->scx.dsq_vtime;
+}
+
+void BPF_STRUCT_OPS(simple_stopping, struct task_struct *p, bool runnable)
+{
+	if (fifo_sched)
+		return;
+
+	/* scale the execution time by the inverse of the weight and charge */
+	p->scx.dsq_vtime += (SCX_SLICE_DFL - p->scx.slice) * 100 / p->scx.weight;
+}
+
+void BPF_STRUCT_OPS(simple_enable, struct task_struct *p,
+		    struct scx_enable_args *args)
+{
+	p->scx.dsq_vtime = vtime_now;
 }
 
 s32 BPF_STRUCT_OPS(simple_init)
@@ -60,6 +126,9 @@ void BPF_STRUCT_OPS(simple_exit, struct scx_exit_info *ei)
 SEC(".struct_ops.link")
 struct sched_ext_ops simple_ops = {
 	.enqueue		= (void *)simple_enqueue,
+	.running		= (void *)simple_running,
+	.stopping		= (void *)simple_stopping,
+	.enable			= (void *)simple_enable,
 	.init			= (void *)simple_init,
 	.exit			= (void *)simple_exit,
 	.name			= "simple",
diff --git a/tools/sched_ext/scx_simple.c b/tools/sched_ext/scx_simple.c
index 1e507c0a35cd..4b2f0c16a9d1 100644
--- a/tools/sched_ext/scx_simple.c
+++ b/tools/sched_ext/scx_simple.c
@@ -19,8 +19,9 @@ const char help_fmt[] =
 "\n"
 "See the top-level comment in .bpf.c for more details.\n"
 "\n"
-"Usage: %s [-p]\n"
+"Usage: %s [-f] [-p]\n"
 "\n"
+"  -f            Use FIFO scheduling instead of weighted vtime scheduling\n"
 "  -p            Switch only tasks on SCHED_EXT policy intead of all\n"
 "  -h            Display this help and exit\n";
 
@@ -65,8 +66,11 @@ int main(int argc, char **argv)
 	skel = scx_simple__open();
 	assert(skel);
 
-	while ((opt = getopt(argc, argv, "ph")) != -1) {
+	while ((opt = getopt(argc, argv, "fph")) != -1) {
 		switch (opt) {
+		case 'f':
+			skel->rodata->fifo_sched = true;
+			break;
 		case 'p':
 			skel->rodata->switch_partial = true;
 			break;
-- 
2.41.0

[PATCH 32/34] sched_ext: Documentation: scheduler: Document extensible scheduler class

[Tejun Heo]

Add Documentation/scheduler/sched-ext.rst which gives a high-level overview
and pointers to the examples.

v3: * Added tools/sched_ext/README.

    * Dropped _example prefix from scheduler names.

v2: * Apply minor edits suggested by Bagas. Caveats section dropped as all
      of them are addressed.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
Cc: Bagas Sanjaya <bagasdotme@gmail.com>
---
 Documentation/scheduler/index.rst     |   1 +
 Documentation/scheduler/sched-ext.rst | 229 ++++++++++++++++++++++
 include/linux/sched/ext.h             |   2 +
 kernel/Kconfig.preempt                |   2 +
 kernel/sched/ext.c                    |   2 +
 kernel/sched/ext.h                    |   2 +
 tools/sched_ext/README                | 264 ++++++++++++++++++++++++++
 7 files changed, 502 insertions(+)
 create mode 100644 Documentation/scheduler/sched-ext.rst
 create mode 100644 tools/sched_ext/README

diff --git a/Documentation/scheduler/index.rst b/Documentation/scheduler/index.rst
index 3170747226f6..0b650bb550e6 100644
--- a/Documentation/scheduler/index.rst
+++ b/Documentation/scheduler/index.rst
@@ -19,6 +19,7 @@ Scheduler
     sched-nice-design
     sched-rt-group
     sched-stats
+    sched-ext
     sched-debug
 
     text_files
diff --git a/Documentation/scheduler/sched-ext.rst b/Documentation/scheduler/sched-ext.rst
new file mode 100644
index 000000000000..25ddb535c297
--- /dev/null
+++ b/Documentation/scheduler/sched-ext.rst
@@ -0,0 +1,229 @@
+==========================
+Extensible Scheduler Class
+==========================
+
+sched_ext is a scheduler class whose behavior can be defined by a set of BPF
+programs - the BPF scheduler.
+
+* sched_ext exports a full scheduling interface so that any scheduling
+  algorithm can be implemented on top.
+
+* The BPF scheduler can group CPUs however it sees fit and schedule them
+  together, as tasks aren't tied to specific CPUs at the time of wakeup.
+
+* The BPF scheduler can be turned on and off dynamically anytime.
+
+* The system integrity is maintained no matter what the BPF scheduler does.
+  The default scheduling behavior is restored anytime an error is detected,
+  a runnable task stalls, or on invoking the SysRq key sequence
+  :kbd:`SysRq-S`.
+
+Switching to and from sched_ext
+===============================
+
+``CONFIG_SCHED_CLASS_EXT`` is the config option to enable sched_ext and
+``tools/sched_ext`` contains the example schedulers.
+
+sched_ext is used only when the BPF scheduler is loaded and running.
+
+If a task explicitly sets its scheduling policy to ``SCHED_EXT``, it will be
+treated as ``SCHED_NORMAL`` and scheduled by CFS until the BPF scheduler is
+loaded. On load, such tasks will be switched to and scheduled by sched_ext.
+
+The BPF scheduler can choose to schedule all normal and lower class tasks by
+calling ``scx_bpf_switch_all()`` from its ``init()`` operation. In this
+case, all ``SCHED_NORMAL``, ``SCHED_BATCH``, ``SCHED_IDLE`` and
+``SCHED_EXT`` tasks are scheduled by sched_ext. In the example schedulers,
+this mode can be selected with the ``-a`` option.
+
+Terminating the sched_ext scheduler program, triggering :kbd:`SysRq-S`, or
+detection of any internal error including stalled runnable tasks aborts the
+BPF scheduler and reverts all tasks back to CFS.
+
+.. code-block:: none
+
+    # make -j16 -C tools/sched_ext
+    # tools/sched_ext/scx_simple
+    local=0 global=3
+    local=5 global=24
+    local=9 global=44
+    local=13 global=56
+    local=17 global=72
+    ^CEXIT: BPF scheduler unregistered
+
+If ``CONFIG_SCHED_DEBUG`` is set, the current status of the BPF scheduler
+and whether a given task is on sched_ext can be determined as follows:
+
+.. code-block:: none
+
+    # cat /sys/kernel/debug/sched/ext
+    ops                           : simple
+    enabled                       : 1
+    switching_all                 : 1
+    switched_all                  : 1
+    enable_state                  : enabled
+
+    # grep ext /proc/self/sched
+    ext.enabled                                  :                    1
+
+The Basics
+==========
+
+Userspace can implement an arbitrary BPF scheduler by loading a set of BPF
+programs that implement ``struct sched_ext_ops``. The only mandatory field
+is ``ops.name`` which must be a valid BPF object name. All operations are
+optional. The following modified excerpt is from
+``tools/sched/scx_simple.bpf.c`` showing a minimal global FIFO scheduler.
+
+.. code-block:: c
+
+    s32 BPF_STRUCT_OPS(simple_init)
+    {
+            if (!switch_partial)
+                    scx_bpf_switch_all();
+            return 0;
+    }
+
+    void BPF_STRUCT_OPS(simple_enqueue, struct task_struct *p, u64 enq_flags)
+    {
+            if (enq_flags & SCX_ENQ_LOCAL)
+                    scx_bpf_dispatch(p, SCX_DSQ_LOCAL, SCX_SLICE_DFL, enq_flags);
+            else
+                    scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, SCX_SLICE_DFL, enq_flags);
+    }
+
+    void BPF_STRUCT_OPS(simple_exit, struct scx_exit_info *ei)
+    {
+            exit_type = ei->type;
+    }
+
+    SEC(".struct_ops")
+    struct sched_ext_ops simple_ops = {
+            .enqueue                = (void *)simple_enqueue,
+            .init                   = (void *)simple_init,
+            .exit                   = (void *)simple_exit,
+            .name                   = "simple",
+    };
+
+Dispatch Queues
+---------------
+
+To match the impedance between the scheduler core and the BPF scheduler,
+sched_ext uses DSQs (dispatch queues) which can operate as both a FIFO and a
+priority queue. By default, there is one global FIFO (``SCX_DSQ_GLOBAL``),
+and one local dsq per CPU (``SCX_DSQ_LOCAL``). The BPF scheduler can manage
+an arbitrary number of dsq's using ``scx_bpf_create_dsq()`` and
+``scx_bpf_destroy_dsq()``.
+
+A CPU always executes a task from its local DSQ. A task is "dispatched" to a
+DSQ. A non-local DSQ is "consumed" to transfer a task to the consuming CPU's
+local DSQ.
+
+When a CPU is looking for the next task to run, if the local DSQ is not
+empty, the first task is picked. Otherwise, the CPU tries to consume the
+global DSQ. If that doesn't yield a runnable task either, ``ops.dispatch()``
+is invoked.
+
+Scheduling Cycle
+----------------
+
+The following briefly shows how a waking task is scheduled and executed.
+
+1. When a task is waking up, ``ops.select_cpu()`` is the first operation
+   invoked. This serves two purposes. First, CPU selection optimization
+   hint. Second, waking up the selected CPU if idle.
+
+   The CPU selected by ``ops.select_cpu()`` is an optimization hint and not
+   binding. The actual decision is made at the last step of scheduling.
+   However, there is a small performance gain if the CPU
+   ``ops.select_cpu()`` returns matches the CPU the task eventually runs on.
+
+   A side-effect of selecting a CPU is waking it up from idle. While a BPF
+   scheduler can wake up any cpu using the ``scx_bpf_kick_cpu()`` helper,
+   using ``ops.select_cpu()`` judiciously can be simpler and more efficient.
+
+   Note that the scheduler core will ignore an invalid CPU selection, for
+   example, if it's outside the allowed cpumask of the task.
+
+2. Once the target CPU is selected, ``ops.enqueue()`` is invoked. It can
+   make one of the following decisions:
+
+   * Immediately dispatch the task to either the global or local DSQ by
+     calling ``scx_bpf_dispatch()`` with ``SCX_DSQ_GLOBAL`` or
+     ``SCX_DSQ_LOCAL``, respectively.
+
+   * Immediately dispatch the task to a custom DSQ by calling
+     ``scx_bpf_dispatch()`` with a DSQ ID which is smaller than 2^63.
+
+   * Queue the task on the BPF side.
+
+3. When a CPU is ready to schedule, it first looks at its local DSQ. If
+   empty, it then looks at the global DSQ. If there still isn't a task to
+   run, ``ops.dispatch()`` is invoked which can use the following two
+   functions to populate the local DSQ.
+
+   * ``scx_bpf_dispatch()`` dispatches a task to a DSQ. Any target DSQ can
+     be used - ``SCX_DSQ_LOCAL``, ``SCX_DSQ_LOCAL_ON | cpu``,
+     ``SCX_DSQ_GLOBAL`` or a custom DSQ. While ``scx_bpf_dispatch()``
+     currently can't be called with BPF locks held, this is being worked on
+     and will be supported. ``scx_bpf_dispatch()`` schedules dispatching
+     rather than performing them immediately. There can be up to
+     ``ops.dispatch_max_batch`` pending tasks.
+
+   * ``scx_bpf_consume()`` tranfers a task from the specified non-local DSQ
+     to the dispatching DSQ. This function cannot be called with any BPF
+     locks held. ``scx_bpf_consume()`` flushes the pending dispatched tasks
+     before trying to consume the specified DSQ.
+
+4. After ``ops.dispatch()`` returns, if there are tasks in the local DSQ,
+   the CPU runs the first one. If empty, the following steps are taken:
+
+   * Try to consume the global DSQ. If successful, run the task.
+
+   * If ``ops.dispatch()`` has dispatched any tasks, retry #3.
+
+   * If the previous task is an SCX task and still runnable, keep executing
+     it (see ``SCX_OPS_ENQ_LAST``).
+
+   * Go idle.
+
+Note that the BPF scheduler can always choose to dispatch tasks immediately
+in ``ops.enqueue()`` as illustrated in the above simple example. If only the
+built-in DSQs are used, there is no need to implement ``ops.dispatch()`` as
+a task is never queued on the BPF scheduler and both the local and global
+DSQs are consumed automatically.
+
+``scx_bpf_dispatch()`` queues the task on the FIFO of the target DSQ. Use
+``scx_bpf_dispatch_vtime()`` for the priority queue. See the function
+documentation and usage in ``tools/sched_ext/scx_simple.bpf.c`` for more
+information.
+
+Where to Look
+=============
+
+* ``include/linux/sched/ext.h`` defines the core data structures, ops table
+  and constants.
+
+* ``kernel/sched/ext.c`` contains sched_ext core implementation and helpers.
+  The functions prefixed with ``scx_bpf_`` can be called from the BPF
+  scheduler.
+
+* ``tools/sched_ext/`` hosts example BPF scheduler implementations.
+
+  * ``scx_simple[.bpf].c``: Minimal global FIFO scheduler example using a
+    custom DSQ.
+
+  * ``scx_qmap[.bpf].c``: A multi-level FIFO scheduler supporting five
+    levels of priority implemented with ``BPF_MAP_TYPE_QUEUE``.
+
+ABI Instability
+===============
+
+The APIs provided by sched_ext to BPF schedulers programs have no stability
+guarantees. This includes the ops table callbacks and constants defined in
+``include/linux/sched/ext.h``, as well as the ``scx_bpf_`` kfuncs defined in
+``kernel/sched/ext.c``.
+
+While we will attempt to provide a relatively stable API surface when
+possible, they are subject to change without warning between kernel
+versions.
diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 446821073bd1..26537b2f6c95 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -1,5 +1,7 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 /*
+ * BPF extensible scheduler class: Documentation/scheduler/sched-ext.rst
+ *
  * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
  * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
  * Copyright (c) 2022 David Vernet <dvernet@meta.com>
diff --git a/kernel/Kconfig.preempt b/kernel/Kconfig.preempt
index e12a057ead7b..bae49b743834 100644
--- a/kernel/Kconfig.preempt
+++ b/kernel/Kconfig.preempt
@@ -154,3 +154,5 @@ config SCHED_CLASS_EXT
 	  wish to implement scheduling policies. The struct_ops structure
 	  exported by sched_ext is struct sched_ext_ops, and is conceptually
 	  similar to struct sched_class.
+
+	  See Documentation/scheduler/sched-ext.rst for more details.
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 44ba457263bd..0db6d400d340 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -1,5 +1,7 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 /*
+ * BPF extensible scheduler class: Documentation/scheduler/sched-ext.rst
+ *
  * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
  * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
  * Copyright (c) 2022 David Vernet <dvernet@meta.com>
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index 821515bb6580..405037a4e6ce 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -1,5 +1,7 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 /*
+ * BPF extensible scheduler class: Documentation/scheduler/sched-ext.rst
+ *
  * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
  * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
  * Copyright (c) 2022 David Vernet <dvernet@meta.com>
diff --git a/tools/sched_ext/README b/tools/sched_ext/README
new file mode 100644
index 000000000000..33f413f8a403
--- /dev/null
+++ b/tools/sched_ext/README
@@ -0,0 +1,264 @@
+		============================
+		SCHED_EXT EXAMPLE SCHEDULERS
+		============================
+
+Introduction
+============
+
+This directory contains a number of example sched_ext schedulers. These
+schedulers are meant to provide examples of different types of schedulers
+that can be built using sched_ext, and illustrate how various features of
+sched_ext can be used.
+
+Some of the examples are performant, production-ready schedulers. That is, for
+the correct workload and with the correct tuning, they may be deployed in a
+production environment with acceptable or possibly even improved performance.
+Others are just examples that in practice, would not provide acceptable
+performance (though they could be improved to get there).
+
+This README will describe these example schedulers, including describing the
+types of workloads or scenarios they're designed to accommodate, and whether or
+not they're production ready. For more details on any of these schedulers,
+please see the header comment in their .bpf.c file.
+
+
+Compiling the examples
+======================
+
+There are a few toolchain dependencies for compiling the example schedulers.
+
+Toolchain dependencies
+----------------------
+
+1. clang >= 17.0
+
+The schedulers are BPF programs, and therefore must be compiled with clang. gcc
+is actively working on adding a BPF backend compiler as well, but are still
+missing some features such as BTF type tags which are necessary for using
+kptrs.
+
+clang 17.0 has not yet been released, so you'll need to compile it yourself if
+you want to compile the benchmarks.
+
+2. rustup nightly
+
+Atropos's user space load balancing component is written in Rust, and uses
+nightly features. You'll need to use the nightly build from rustup in order to
+compile it.
+
+There are other requirements as well, such as make, but these are the main /
+non-trivial ones.
+
+Compiling the schedulers
+------------------------
+
+Once you have your toolchain setup, you can compile the schedulers as follows:
+
+$ make CC=clang LLVM=1 -j
+
+See Documentation/scheduler/sched-ext.rst for a description of the config
+options required to compile a sched_ext kernel.
+
+Schedulers
+==========
+
+This section lists, in alphabetical order, all of the current example
+schedulers.
+
+--------------------------------------------------------------------------------
+
+Atropos
+-------
+
+Overview
+~~~~~~~~
+
+A multi-domain, BPF / user space hybrid scheduler. The BPF portion of the
+scheduler does a simple round robin in each domain, and the user space portion
+(written in Rust) calculates the load factor of each domain, and informs BPF of
+how tasks should be load balanced accordingly.
+
+Typical Use Case
+~~~~~~~~~~~~~~~~
+
+Atropos is designed to be flexible, and accommodate different architectures and
+workloads. Various load balancing thresholds (e.g. greediness, frequenty, etc),
+as well as how Atropos should partition the system into scheduling domains, can
+be tuned to achieve the optimal configuration for any given system or workload.
+
+Production Ready?
+~~~~~~~~~~~~~~~~~
+
+Yes. If tuned correctly, Atropos should be performant across various CPU
+architectures and workloads.
+
+--------------------------------------------------------------------------------
+
+scx_central
+-----------
+
+Overview
+~~~~~~~~
+
+A "central" scheduler where scheduling decisions are made from a single CPU.
+This scheduler illustrates how scheduling decisions can be dispatched from a
+single CPU, allowing other cores to run with infinite slices, without timer
+ticks, and without having to incur the overhead of making scheduling decisions.
+
+Typical Use Case
+~~~~~~~~~~~~~~~~
+
+This scheduler could theoretically be useful for any workload that benefits
+from minimizing scheduling overhead and timer ticks. An example of where this
+could be particularly useful is running VMs, where running with infinite slices
+and no timer ticks allows the VM to avoid unnecessary expensive vmexits.
+
+Production Ready?
+~~~~~~~~~~~~~~~~~
+
+Not yet. While tasks are run with an infinite slice (SCX_SLICE_INF), they're
+preempted every 20ms in a timer callback. The scheduler also puts the core
+schedling logic inside of the central / scheduling CPU's ops.dispatch() path,
+and does not yet have any kind of priority mechanism.
+
+--------------------------------------------------------------------------------
+
+scx_flatcg
+----------
+
+Overview
+~~~~~~~~
+
+A flattened cgroup hierarchy scheduler. This scheduler implements hierarchical
+weight-based cgroup CPU control by flattening the cgroup hierarchy into a
+single layer, by compounding the active weight share at each level. The effect
+of this is a much more performant CPU controller, which does not need to
+descend down cgroup trees in order to properly compute a cgroup's share.
+
+Typical Use Case
+~~~~~~~~~~~~~~~~
+
+This scheduler could be useful for any typical workload requiring a CPU
+controller, but which cannot tolerate the higher overheads of the fair CPU
+controller.
+
+Production Ready?
+~~~~~~~~~~~~~~~~~
+
+Yes, though the scheduler (currently) does not adequately accommodate
+thundering herds of cgroups. If, for example, many cgroups which are nested
+behind a low-priority cgroup were to wake up around the same time, they may be
+able to consume more CPU cycles than they are entitled to.
+
+--------------------------------------------------------------------------------
+
+scx_pair
+--------
+
+Overview
+~~~~~~~~
+
+A sibling scheduler which ensures that tasks will only ever be co-located on a
+physical core if they're in the same cgroup. It illustrates how a scheduling
+policy could be implemented to mitigate CPU bugs, such as L1TF, and also shows
+how some useful kfuncs such as scx_bpf_kick_cpu() can be utilized.
+
+Typical Use Case
+~~~~~~~~~~~~~~~~
+
+While this scheduler is only meant to be used to illustrate certain sched_ext
+features, with a bit more work (e.g. by adding some form of priority handling
+inside and across cgroups), it could have been used as a way to quickly
+mitigate L1TF before core scheduling was implemented and rolled out.
+
+Production Ready?
+~~~~~~~~~~~~~~~~~
+
+No
+
+--------------------------------------------------------------------------------
+
+scx_qmap
+--------
+
+Overview
+~~~~~~~~
+
+Another simple, yet slightly more complex scheduler that provides an example of
+a basic weighted FIFO queuing policy. It also provides examples of some common
+useful BPF features, such as sleepable per-task storage allocation in the
+ops.prep_enable() callback, and using the BPF_MAP_TYPE_QUEUE map type to
+enqueue tasks. It also illustrates how core-sched support could be implemented.
+
+Typical Use Case
+~~~~~~~~~~~~~~~~
+
+Purely used to illustrate sched_ext features.
+
+**Production Ready?**
+
+No
+
+--------------------------------------------------------------------------------
+
+scx_simple
+----------
+
+Overview
+~~~~~~~~
+
+A simple scheduler that provides an example of a minimal sched_ext
+scheduler. scx_simple can be run in either global weighted vtime mode, or
+FIFO mode.
+
+Typical Use Case
+~~~~~~~~~~~~~~~~
+
+Though very simple, this scheduler should perform reasonably well on
+single-socket CPUs with a uniform L3 cache topology. Note that while running in
+global FIFO mode may work well for some workloads, saturating threads can
+easily drown out inactive ones.
+
+Production Ready?
+~~~~~~~~~~~~~~~~~
+
+This scheduler could be used in a production environment, assuming the hardware
+constraints enumerated above, and assuming the workload can accommodate a
+simple scheduling policy.
+
+--------------------------------------------------------------------------------
+
+scx_userland
+------------
+
+Overview
+~~~~~~~~
+
+A simple weighted vtime scheduler where all scheduling decisions take place in
+user space. This is in contrast to Atropos, where load balancing lives in user
+space, but scheduling decisions are still made in the kernel.
+
+Typical Use Case
+~~~~~~~~~~~~~~~~
+
+There are many advantages to writing schedulers in user space. For example, you
+can use a debugger, you can write the scheduler in Rust, and you can use data
+structures bundled with your favorite library.
+
+On the other hand, user space scheduling can be hard to get right. You can
+potentially deadlock due to not scheduling a task that's required for the
+scheduler itself to make forward progress (though the sched_ext watchdog will
+protect the system by unloading your scheduler after a timeout if that
+happens). You also have to bootstrap some communication protocol between the
+kernel and user space.
+
+A more robust solution to this would be building a user space scheduling
+framework that abstracts much of this complexity away from you.
+
+Production Ready?
+~~~~~~~~~~~~~~~~~
+
+No. This scheduler uses an ordered list for vtime scheduling, and is stricly
+less performant than just using something like `scx_simple`. It is purely
+meant to illustrate that it's possible to build a user space scheduler on
+top of sched_ext.
-- 
2.41.0

[PATCH 33/34] sched_ext: Add a basic, userland vruntime scheduler

[Tejun Heo]

From: David Vernet <dvernet@meta.com>

This patch adds a new scx_userland BPF scheduler that implements a
fairly unsophisticated sorted-list vruntime scheduler in userland to
demonstrate how most scheduling decisions can be delegated to userland. The
scheduler doesn't implement load balancing, and treats all tasks as part of
a single domain.

v2: * Converted to BPF inline iterators.

Signed-off-by: David Vernet <dvernet@meta.com>
Reviewed-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 tools/sched_ext/.gitignore         |   1 +
 tools/sched_ext/Makefile           |   8 +-
 tools/sched_ext/scx_userland.bpf.c | 262 +++++++++++++++++++
 tools/sched_ext/scx_userland.c     | 402 +++++++++++++++++++++++++++++
 tools/sched_ext/scx_userland.h     |  19 ++
 5 files changed, 690 insertions(+), 2 deletions(-)
 create mode 100644 tools/sched_ext/scx_userland.bpf.c
 create mode 100644 tools/sched_ext/scx_userland.c
 create mode 100644 tools/sched_ext/scx_userland.h

diff --git a/tools/sched_ext/.gitignore b/tools/sched_ext/.gitignore
index d5a4923919ce..c63ee5e4f4bb 100644
--- a/tools/sched_ext/.gitignore
+++ b/tools/sched_ext/.gitignore
@@ -3,6 +3,7 @@ scx_qmap
 scx_central
 scx_pair
 scx_flatcg
+scx_userland
 *.skel.h
 *.subskel.h
 /tools/
diff --git a/tools/sched_ext/Makefile b/tools/sched_ext/Makefile
index a0dacea5993c..092c3859228f 100644
--- a/tools/sched_ext/Makefile
+++ b/tools/sched_ext/Makefile
@@ -115,7 +115,7 @@ BPF_CFLAGS = -g -D__TARGET_ARCH_$(SRCARCH)					\
 	     -Wall -Wno-compare-distinct-pointer-types				\
 	     -O2 -mcpu=v3
 
-all: scx_simple scx_qmap scx_central scx_pair scx_flatcg
+all: scx_simple scx_qmap scx_central scx_pair scx_flatcg scx_userland
 
 # sort removes libbpf duplicates when not cross-building
 MAKE_DIRS := $(sort $(BUILD_DIR)/libbpf $(HOST_BUILD_DIR)/libbpf		\
@@ -186,10 +186,14 @@ scx_flatcg: scx_flatcg.c scx_flatcg.skel.h user_exit_info.h
 	$(CC) $(CFLAGS) -c $< -o $@.o
 	$(CC) -o $@ $@.o $(HOST_BPFOBJ) $(LDFLAGS)
 
+scx_userland: scx_userland.c scx_userland.skel.h scx_userland.h user_exit_info.h
+	$(CC) $(CFLAGS) -c $< -o $@.o
+	$(CC) -o $@ $@.o $(HOST_BPFOBJ) $(LDFLAGS)
+
 clean:
 	rm -rf $(SCRATCH_DIR) $(HOST_SCRATCH_DIR)
 	rm -f *.o *.bpf.o *.skel.h *.subskel.h
-	rm -f scx_simple scx_qmap scx_central scx_pair scx_flatcg
+	rm -f scx_simple scx_qmap scx_central scx_pair scx_flatcg scx_userland
 
 .PHONY: all clean
 
diff --git a/tools/sched_ext/scx_userland.bpf.c b/tools/sched_ext/scx_userland.bpf.c
new file mode 100644
index 000000000000..9e107a874a92
--- /dev/null
+++ b/tools/sched_ext/scx_userland.bpf.c
@@ -0,0 +1,262 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * A minimal userland scheduler.
+ *
+ * In terms of scheduling, this provides two different types of behaviors:
+ * 1. A global FIFO scheduling order for _any_ tasks that have CPU affinity.
+ *    All such tasks are direct-dispatched from the kernel, and are never
+ *    enqueued in user space.
+ * 2. A primitive vruntime scheduler that is implemented in user space, for all
+ *    other tasks.
+ *
+ * Some parts of this example user space scheduler could be implemented more
+ * efficiently using more complex and sophisticated data structures. For
+ * example, rather than using BPF_MAP_TYPE_QUEUE's,
+ * BPF_MAP_TYPE_{USER_}RINGBUF's could be used for exchanging messages between
+ * user space and kernel space. Similarly, we use a simple vruntime-sorted list
+ * in user space, but an rbtree could be used instead.
+ *
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+#include <string.h>
+#include "scx_common.bpf.h"
+#include "scx_userland.h"
+
+char _license[] SEC("license") = "GPL";
+
+const volatile bool switch_partial;
+const volatile s32 usersched_pid;
+
+/* !0 for veristat, set during init */
+const volatile u32 num_possible_cpus = 64;
+
+/* Stats that are printed by user space. */
+u64 nr_failed_enqueues, nr_kernel_enqueues, nr_user_enqueues;
+
+struct user_exit_info uei;
+
+/*
+ * Whether the user space scheduler needs to be scheduled due to a task being
+ * enqueued in user space.
+ */
+static bool usersched_needed;
+
+/*
+ * The map containing tasks that are enqueued in user space from the kernel.
+ *
+ * This map is drained by the user space scheduler.
+ */
+struct {
+	__uint(type, BPF_MAP_TYPE_QUEUE);
+	__uint(max_entries, USERLAND_MAX_TASKS);
+	__type(value, struct scx_userland_enqueued_task);
+} enqueued SEC(".maps");
+
+/*
+ * The map containing tasks that are dispatched to the kernel from user space.
+ *
+ * Drained by the kernel in userland_dispatch().
+ */
+struct {
+	__uint(type, BPF_MAP_TYPE_QUEUE);
+	__uint(max_entries, USERLAND_MAX_TASKS);
+	__type(value, s32);
+} dispatched SEC(".maps");
+
+/* Per-task scheduling context */
+struct task_ctx {
+	bool force_local; /* Dispatch directly to local DSQ */
+};
+
+/* Map that contains task-local storage. */
+struct {
+	__uint(type, BPF_MAP_TYPE_TASK_STORAGE);
+	__uint(map_flags, BPF_F_NO_PREALLOC);
+	__type(key, int);
+	__type(value, struct task_ctx);
+} task_ctx_stor SEC(".maps");
+
+static bool is_usersched_task(const struct task_struct *p)
+{
+	return p->pid == usersched_pid;
+}
+
+static bool keep_in_kernel(const struct task_struct *p)
+{
+	return p->nr_cpus_allowed < num_possible_cpus;
+}
+
+static struct task_struct *usersched_task(void)
+{
+	struct task_struct *p;
+
+	p = bpf_task_from_pid(usersched_pid);
+	/*
+	 * Should never happen -- the usersched task should always be managed
+	 * by sched_ext.
+	 */
+	if (!p)
+		scx_bpf_error("Failed to find usersched task %d", usersched_pid);
+
+	return p;
+}
+
+s32 BPF_STRUCT_OPS(userland_select_cpu, struct task_struct *p,
+		   s32 prev_cpu, u64 wake_flags)
+{
+	if (keep_in_kernel(p)) {
+		s32 cpu;
+		struct task_ctx *tctx;
+
+		tctx = bpf_task_storage_get(&task_ctx_stor, p, 0, 0);
+		if (!tctx) {
+			scx_bpf_error("Failed to look up task-local storage for %s", p->comm);
+			return -ESRCH;
+		}
+
+		if (p->nr_cpus_allowed == 1 ||
+		    scx_bpf_test_and_clear_cpu_idle(prev_cpu)) {
+			tctx->force_local = true;
+			return prev_cpu;
+		}
+
+		cpu = scx_bpf_pick_idle_cpu(p->cpus_ptr, 0);
+		if (cpu >= 0) {
+			tctx->force_local = true;
+			return cpu;
+		}
+	}
+
+	return prev_cpu;
+}
+
+static void dispatch_user_scheduler(void)
+{
+	struct task_struct *p;
+
+	usersched_needed = false;
+	p = usersched_task();
+	if (p) {
+		scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, SCX_SLICE_DFL, 0);
+		bpf_task_release(p);
+	}
+}
+
+static void enqueue_task_in_user_space(struct task_struct *p, u64 enq_flags)
+{
+	struct scx_userland_enqueued_task task;
+
+	memset(&task, 0, sizeof(task));
+	task.pid = p->pid;
+	task.sum_exec_runtime = p->se.sum_exec_runtime;
+	task.weight = p->scx.weight;
+
+	if (bpf_map_push_elem(&enqueued, &task, 0)) {
+		/*
+		 * If we fail to enqueue the task in user space, put it
+		 * directly on the global DSQ.
+		 */
+		__sync_fetch_and_add(&nr_failed_enqueues, 1);
+		scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, SCX_SLICE_DFL, enq_flags);
+	} else {
+		__sync_fetch_and_add(&nr_user_enqueues, 1);
+		usersched_needed = true;
+	}
+}
+
+void BPF_STRUCT_OPS(userland_enqueue, struct task_struct *p, u64 enq_flags)
+{
+	if (keep_in_kernel(p)) {
+		u64 dsq_id = SCX_DSQ_GLOBAL;
+		struct task_ctx *tctx;
+
+		tctx = bpf_task_storage_get(&task_ctx_stor, p, 0, 0);
+		if (!tctx) {
+			scx_bpf_error("Failed to lookup task ctx for %s", p->comm);
+			return;
+		}
+
+		if (tctx->force_local)
+			dsq_id = SCX_DSQ_LOCAL;
+		tctx->force_local = false;
+		scx_bpf_dispatch(p, dsq_id, SCX_SLICE_DFL, enq_flags);
+		__sync_fetch_and_add(&nr_kernel_enqueues, 1);
+		return;
+	} else if (!is_usersched_task(p)) {
+		enqueue_task_in_user_space(p, enq_flags);
+	}
+}
+
+void BPF_STRUCT_OPS(userland_dispatch, s32 cpu, struct task_struct *prev)
+{
+	if (usersched_needed)
+		dispatch_user_scheduler();
+
+	bpf_repeat(4096) {
+		s32 pid;
+		struct task_struct *p;
+
+		if (bpf_map_pop_elem(&dispatched, &pid))
+			break;
+
+		/*
+		 * The task could have exited by the time we get around to
+		 * dispatching it. Treat this as a normal occurrence, and simply
+		 * move onto the next iteration.
+		 */
+		p = bpf_task_from_pid(pid);
+		if (!p)
+			continue;
+
+		scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, SCX_SLICE_DFL, 0);
+		bpf_task_release(p);
+	}
+}
+
+s32 BPF_STRUCT_OPS(userland_prep_enable, struct task_struct *p,
+		   struct scx_enable_args *args)
+{
+	if (bpf_task_storage_get(&task_ctx_stor, p, 0,
+				 BPF_LOCAL_STORAGE_GET_F_CREATE))
+		return 0;
+	else
+		return -ENOMEM;
+}
+
+s32 BPF_STRUCT_OPS(userland_init)
+{
+	if (num_possible_cpus == 0) {
+		scx_bpf_error("User scheduler # CPUs uninitialized (%d)",
+			      num_possible_cpus);
+		return -EINVAL;
+	}
+
+	if (usersched_pid <= 0) {
+		scx_bpf_error("User scheduler pid uninitialized (%d)",
+			      usersched_pid);
+		return -EINVAL;
+	}
+
+	if (!switch_partial)
+		scx_bpf_switch_all();
+	return 0;
+}
+
+void BPF_STRUCT_OPS(userland_exit, struct scx_exit_info *ei)
+{
+	uei_record(&uei, ei);
+}
+
+SEC(".struct_ops.link")
+struct sched_ext_ops userland_ops = {
+	.select_cpu		= (void *)userland_select_cpu,
+	.enqueue		= (void *)userland_enqueue,
+	.dispatch		= (void *)userland_dispatch,
+	.prep_enable		= (void *)userland_prep_enable,
+	.init			= (void *)userland_init,
+	.exit			= (void *)userland_exit,
+	.timeout_ms		= 3000,
+	.name			= "userland",
+};
diff --git a/tools/sched_ext/scx_userland.c b/tools/sched_ext/scx_userland.c
new file mode 100644
index 000000000000..a63adae74f21
--- /dev/null
+++ b/tools/sched_ext/scx_userland.c
@@ -0,0 +1,402 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * A demo sched_ext user space scheduler which provides vruntime semantics
+ * using a simple ordered-list implementation.
+ *
+ * Each CPU in the system resides in a single, global domain. This precludes
+ * the need to do any load balancing between domains. The scheduler could
+ * easily be extended to support multiple domains, with load balancing
+ * happening in user space.
+ *
+ * Any task which has any CPU affinity is scheduled entirely in BPF. This
+ * program only schedules tasks which may run on any CPU.
+ *
+ * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
+ * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
+ * Copyright (c) 2022 David Vernet <dvernet@meta.com>
+ */
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <unistd.h>
+#include <sched.h>
+#include <signal.h>
+#include <assert.h>
+#include <libgen.h>
+#include <pthread.h>
+#include <bpf/bpf.h>
+#include <sys/mman.h>
+#include <sys/queue.h>
+#include <sys/syscall.h>
+
+#include "user_exit_info.h"
+#include "scx_userland.h"
+#include "scx_userland.skel.h"
+
+const char help_fmt[] =
+"A minimal userland sched_ext scheduler.\n"
+"\n"
+"See the top-level comment in .bpf.c for more details.\n"
+"\n"
+"Usage: %s [-b BATCH] [-p]\n"
+"\n"
+"  -b BATCH      The number of tasks to batch when dispatching (default: 8)\n"
+"  -p            Don't switch all, switch only tasks on SCHED_EXT policy\n"
+"  -h            Display this help and exit\n";
+
+/* Defined in UAPI */
+#define SCHED_EXT 7
+
+/* Number of tasks to batch when dispatching to user space. */
+static __u32 batch_size = 8;
+
+static volatile int exit_req;
+static int enqueued_fd, dispatched_fd;
+
+static struct scx_userland *skel;
+static struct bpf_link *ops_link;
+
+/* Stats collected in user space. */
+static __u64 nr_vruntime_enqueues, nr_vruntime_dispatches;
+
+/* The data structure containing tasks that are enqueued in user space. */
+struct enqueued_task {
+	LIST_ENTRY(enqueued_task) entries;
+	__u64 sum_exec_runtime;
+	double vruntime;
+};
+
+/*
+ * Use a vruntime-sorted list to store tasks. This could easily be extended to
+ * a more optimal data structure, such as an rbtree as is done in CFS. We
+ * currently elect to use a sorted list to simplify the example for
+ * illustrative purposes.
+ */
+LIST_HEAD(listhead, enqueued_task);
+
+/*
+ * A vruntime-sorted list of tasks. The head of the list contains the task with
+ * the lowest vruntime. That is, the task that has the "highest" claim to be
+ * scheduled.
+ */
+static struct listhead vruntime_head = LIST_HEAD_INITIALIZER(vruntime_head);
+
+/*
+ * The statically allocated array of tasks. We use a statically allocated list
+ * here to avoid having to allocate on the enqueue path, which could cause a
+ * deadlock. A more substantive user space scheduler could e.g. provide a hook
+ * for newly enabled tasks that are passed to the scheduler from the
+ * .prep_enable() callback to allows the scheduler to allocate on safe paths.
+ */
+struct enqueued_task tasks[USERLAND_MAX_TASKS];
+
+static double min_vruntime;
+
+static void sigint_handler(int userland)
+{
+	exit_req = 1;
+}
+
+static __u32 task_pid(const struct enqueued_task *task)
+{
+	return ((uintptr_t)task - (uintptr_t)tasks) / sizeof(*task);
+}
+
+static int dispatch_task(s32 pid)
+{
+	int err;
+
+	err = bpf_map_update_elem(dispatched_fd, NULL, &pid, 0);
+	if (err) {
+		fprintf(stderr, "Failed to dispatch task %d\n", pid);
+		exit_req = 1;
+	} else {
+		nr_vruntime_dispatches++;
+	}
+
+	return err;
+}
+
+static struct enqueued_task *get_enqueued_task(__s32 pid)
+{
+	if (pid >= USERLAND_MAX_TASKS)
+		return NULL;
+
+	return &tasks[pid];
+}
+
+static double calc_vruntime_delta(__u64 weight, __u64 delta)
+{
+	double weight_f = (double)weight / 100.0;
+	double delta_f = (double)delta;
+
+	return delta_f / weight_f;
+}
+
+static void update_enqueued(struct enqueued_task *enqueued, const struct scx_userland_enqueued_task *bpf_task)
+{
+	__u64 delta;
+
+	delta = bpf_task->sum_exec_runtime - enqueued->sum_exec_runtime;
+
+	enqueued->vruntime += calc_vruntime_delta(bpf_task->weight, delta);
+	if (min_vruntime > enqueued->vruntime)
+		enqueued->vruntime = min_vruntime;
+	enqueued->sum_exec_runtime = bpf_task->sum_exec_runtime;
+}
+
+static int vruntime_enqueue(const struct scx_userland_enqueued_task *bpf_task)
+{
+	struct enqueued_task *curr, *enqueued, *prev;
+
+	curr = get_enqueued_task(bpf_task->pid);
+	if (!curr)
+		return ENOENT;
+
+	update_enqueued(curr, bpf_task);
+	nr_vruntime_enqueues++;
+
+	/*
+	 * Enqueue the task in a vruntime-sorted list. A more optimal data
+	 * structure such as an rbtree could easily be used as well. We elect
+	 * to use a list here simply because it's less code, and thus the
+	 * example is less convoluted and better serves to illustrate what a
+	 * user space scheduler could look like.
+	 */
+
+	if (LIST_EMPTY(&vruntime_head)) {
+		LIST_INSERT_HEAD(&vruntime_head, curr, entries);
+		return 0;
+	}
+
+	LIST_FOREACH(enqueued, &vruntime_head, entries) {
+		if (curr->vruntime <= enqueued->vruntime) {
+			LIST_INSERT_BEFORE(enqueued, curr, entries);
+			return 0;
+		}
+		prev = enqueued;
+	}
+
+	LIST_INSERT_AFTER(prev, curr, entries);
+
+	return 0;
+}
+
+static void drain_enqueued_map(void)
+{
+	while (1) {
+		struct scx_userland_enqueued_task task;
+		int err;
+
+		if (bpf_map_lookup_and_delete_elem(enqueued_fd, NULL, &task))
+			return;
+
+		err = vruntime_enqueue(&task);
+		if (err) {
+			fprintf(stderr, "Failed to enqueue task %d: %s\n",
+				task.pid, strerror(err));
+			exit_req = 1;
+			return;
+		}
+	}
+}
+
+static void dispatch_batch(void)
+{
+	__u32 i;
+
+	for (i = 0; i < batch_size; i++) {
+		struct enqueued_task *task;
+		int err;
+		__s32 pid;
+
+		task = LIST_FIRST(&vruntime_head);
+		if (!task)
+			return;
+
+		min_vruntime = task->vruntime;
+		pid = task_pid(task);
+		LIST_REMOVE(task, entries);
+		err = dispatch_task(pid);
+		if (err) {
+			fprintf(stderr, "Failed to dispatch task %d in %u\n",
+				pid, i);
+			return;
+		}
+	}
+}
+
+static void *run_stats_printer(void *arg)
+{
+	while (!exit_req) {
+		__u64 nr_failed_enqueues, nr_kernel_enqueues, nr_user_enqueues, total;
+
+		nr_failed_enqueues = skel->bss->nr_failed_enqueues;
+		nr_kernel_enqueues = skel->bss->nr_kernel_enqueues;
+		nr_user_enqueues = skel->bss->nr_user_enqueues;
+		total = nr_failed_enqueues + nr_kernel_enqueues + nr_user_enqueues;
+
+		printf("o-----------------------o\n");
+		printf("| BPF ENQUEUES          |\n");
+		printf("|-----------------------|\n");
+		printf("|  kern:     %10llu |\n", nr_kernel_enqueues);
+		printf("|  user:     %10llu |\n", nr_user_enqueues);
+		printf("|  failed:   %10llu |\n", nr_failed_enqueues);
+		printf("|  -------------------- |\n");
+		printf("|  total:    %10llu |\n", total);
+		printf("|                       |\n");
+		printf("|-----------------------|\n");
+		printf("| VRUNTIME / USER       |\n");
+		printf("|-----------------------|\n");
+		printf("|  enq:      %10llu |\n", nr_vruntime_enqueues);
+		printf("|  disp:     %10llu |\n", nr_vruntime_dispatches);
+		printf("o-----------------------o\n");
+		printf("\n\n");
+		sleep(1);
+	}
+
+	return NULL;
+}
+
+static int spawn_stats_thread(void)
+{
+	pthread_t stats_printer;
+
+	return pthread_create(&stats_printer, NULL, run_stats_printer, NULL);
+}
+
+static int bootstrap(int argc, char **argv)
+{
+	int err;
+	__u32 opt;
+	struct sched_param sched_param = {
+		.sched_priority = sched_get_priority_max(SCHED_EXT),
+	};
+	bool switch_partial = false;
+
+	signal(SIGINT, sigint_handler);
+	signal(SIGTERM, sigint_handler);
+	libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
+
+	/*
+	 * Enforce that the user scheduler task is managed by sched_ext. The
+	 * task eagerly drains the list of enqueued tasks in its main work
+	 * loop, and then yields the CPU. The BPF scheduler only schedules the
+	 * user space scheduler task when at least one other task in the system
+	 * needs to be scheduled.
+	 */
+	err = syscall(__NR_sched_setscheduler, getpid(), SCHED_EXT, &sched_param);
+	if (err) {
+		fprintf(stderr, "Failed to set scheduler to SCHED_EXT: %s\n", strerror(err));
+		return err;
+	}
+
+	while ((opt = getopt(argc, argv, "b:ph")) != -1) {
+		switch (opt) {
+		case 'b':
+			batch_size = strtoul(optarg, NULL, 0);
+			break;
+		case 'p':
+			switch_partial = true;
+			break;
+		default:
+			fprintf(stderr, help_fmt, basename(argv[0]));
+			exit(opt != 'h');
+		}
+	}
+
+	/*
+	 * It's not always safe to allocate in a user space scheduler, as an
+	 * enqueued task could hold a lock that we require in order to be able
+	 * to allocate.
+	 */
+	err = mlockall(MCL_CURRENT | MCL_FUTURE);
+	if (err) {
+		fprintf(stderr, "Failed to prefault and lock address space: %s\n",
+			strerror(err));
+		return err;
+	}
+
+	skel = scx_userland__open();
+	if (!skel) {
+		fprintf(stderr, "Failed to open scheduler: %s\n", strerror(errno));
+		return errno;
+	}
+	skel->rodata->num_possible_cpus = libbpf_num_possible_cpus();
+	assert(skel->rodata->num_possible_cpus > 0);
+	skel->rodata->usersched_pid = getpid();
+	assert(skel->rodata->usersched_pid > 0);
+	skel->rodata->switch_partial = switch_partial;
+
+	err = scx_userland__load(skel);
+	if (err) {
+		fprintf(stderr, "Failed to load scheduler: %s\n", strerror(err));
+		goto destroy_skel;
+	}
+
+	enqueued_fd = bpf_map__fd(skel->maps.enqueued);
+	dispatched_fd = bpf_map__fd(skel->maps.dispatched);
+	assert(enqueued_fd > 0);
+	assert(dispatched_fd > 0);
+
+	err = spawn_stats_thread();
+	if (err) {
+		fprintf(stderr, "Failed to spawn stats thread: %s\n", strerror(err));
+		goto destroy_skel;
+	}
+
+	ops_link = bpf_map__attach_struct_ops(skel->maps.userland_ops);
+	if (!ops_link) {
+		fprintf(stderr, "Failed to attach struct ops: %s\n", strerror(errno));
+		err = errno;
+		goto destroy_skel;
+	}
+
+	return 0;
+
+destroy_skel:
+	scx_userland__destroy(skel);
+	exit_req = 1;
+	return err;
+}
+
+static void sched_main_loop(void)
+{
+	while (!exit_req) {
+		/*
+		 * Perform the following work in the main user space scheduler
+		 * loop:
+		 *
+		 * 1. Drain all tasks from the enqueued map, and enqueue them
+		 *    to the vruntime sorted list.
+		 *
+		 * 2. Dispatch a batch of tasks from the vruntime sorted list
+		 *    down to the kernel.
+		 *
+		 * 3. Yield the CPU back to the system. The BPF scheduler will
+		 *    reschedule the user space scheduler once another task has
+		 *    been enqueued to user space.
+		 */
+		drain_enqueued_map();
+		dispatch_batch();
+		sched_yield();
+	}
+}
+
+int main(int argc, char **argv)
+{
+	int err;
+
+	err = bootstrap(argc, argv);
+	if (err) {
+		fprintf(stderr, "Failed to bootstrap scheduler: %s\n", strerror(err));
+		return err;
+	}
+
+	sched_main_loop();
+
+	exit_req = 1;
+	bpf_link__destroy(ops_link);
+	uei_print(&skel->bss->uei);
+	scx_userland__destroy(skel);
+	return 0;
+}
diff --git a/tools/sched_ext/scx_userland.h b/tools/sched_ext/scx_userland.h
new file mode 100644
index 000000000000..639c6809c5ff
--- /dev/null
+++ b/tools/sched_ext/scx_userland.h
@@ -0,0 +1,19 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2022 Meta, Inc */
+
+#ifndef __SCX_USERLAND_COMMON_H
+#define __SCX_USERLAND_COMMON_H
+
+#define USERLAND_MAX_TASKS 8192
+
+/*
+ * An instance of a task that has been enqueued by the kernel for consumption
+ * by a user space global scheduler thread.
+ */
+struct scx_userland_enqueued_task {
+	__s32 pid;
+	u64 sum_exec_runtime;
+	u64 weight;
+};
+
+#endif  // __SCX_USERLAND_COMMON_H
-- 
2.41.0

[PATCH 34/34] sched_ext: Add a rust userspace hybrid example scheduler

[Tejun Heo]

From: Dan Schatzberg <dschatzberg@meta.com>

Atropos is a multi-domain BPF / userspace hybrid scheduler where the BPF
part does simple round robin in each domain and the userspace part
calculates the load factor of each domain and tells the BPF part how to load
balance the domains.

This scheduler demonstrates dividing scheduling logic between BPF and
userspace and using rust to build the userspace part. An earlier variant of
this scheduler was used to balance across six domains, each representing a
chiplet in a six-chiplet AMD processor, and could match the performance of
production setup using CFS.

v4: * tools/sched_ext/atropos renamed to tools/sched_ext/scx_atropos for
      consistency.

    * LoadBalancer sometimes couldn't converge on balanced state due to
      restrictions it put on each balancing operation. Fixed.

    * Topology information refactored into struct Topology and Tuner is
      added. Tuner runs in shorter cycles (100ms) than LoadBalancer and
      dynamically adjusts scheduling behaviors, currently, based on the
      per-domain utilization states.

    * ->select_cpu() has been revamped. Combined with other improvements,
      this allows atropos to outperform CFS in various sub-saturation
      scenarios when tested with fio over dm-crypt.

    * Many minor code cleanups and improvements.

v3: * The userspace code is substantially restructured and rewritten. The
      binary is renamed to scx_atropos and can now figure out the domain
      topology automatically based on L3 cache configuration. The LB logic
      which was rather broken in the previous postings are revamped and
      should behave better.

    * Updated to support weighted vtime scheduling (can be turned off with
      --fifo-sched). Added a couple options (--slice_us, --kthreads-local)
      to modify scheduling behaviors.

    * Converted to use BPF inline iterators.

v2: * Updated to use generic BPF cpumask helpers.

Signed-off-by: Dan Schatzberg <dschatzberg@meta.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 tools/sched_ext/Makefile                      |   13 +-
 tools/sched_ext/scx_atropos/.gitignore        |    3 +
 tools/sched_ext/scx_atropos/Cargo.toml        |   27 +
 tools/sched_ext/scx_atropos/build.rs          |   70 +
 tools/sched_ext/scx_atropos/rustfmt.toml      |    8 +
 .../sched_ext/scx_atropos/src/atropos_sys.rs  |   10 +
 .../scx_atropos/src/bpf/atropos.bpf.c         |  978 ++++++++++++++
 tools/sched_ext/scx_atropos/src/bpf/atropos.h |   64 +
 tools/sched_ext/scx_atropos/src/main.rs       | 1196 +++++++++++++++++
 9 files changed, 2367 insertions(+), 2 deletions(-)
 create mode 100644 tools/sched_ext/scx_atropos/.gitignore
 create mode 100644 tools/sched_ext/scx_atropos/Cargo.toml
 create mode 100644 tools/sched_ext/scx_atropos/build.rs
 create mode 100644 tools/sched_ext/scx_atropos/rustfmt.toml
 create mode 100644 tools/sched_ext/scx_atropos/src/atropos_sys.rs
 create mode 100644 tools/sched_ext/scx_atropos/src/bpf/atropos.bpf.c
 create mode 100644 tools/sched_ext/scx_atropos/src/bpf/atropos.h
 create mode 100644 tools/sched_ext/scx_atropos/src/main.rs

diff --git a/tools/sched_ext/Makefile b/tools/sched_ext/Makefile
index 092c3859228f..1515ff9cce7f 100644
--- a/tools/sched_ext/Makefile
+++ b/tools/sched_ext/Makefile
@@ -85,6 +85,8 @@ CFLAGS += -g -O2 -rdynamic -pthread -Wall -Werror $(GENFLAGS)			\
 	  -I$(INCLUDE_DIR) -I$(GENDIR) -I$(LIBDIR)				\
 	  -I$(TOOLSINCDIR) -I$(APIDIR)
 
+CARGOFLAGS := --release
+
 # Silence some warnings when compiled with clang
 ifneq ($(LLVM),)
 CFLAGS += -Wno-unused-command-line-argument
@@ -115,7 +117,7 @@ BPF_CFLAGS = -g -D__TARGET_ARCH_$(SRCARCH)					\
 	     -Wall -Wno-compare-distinct-pointer-types				\
 	     -O2 -mcpu=v3
 
-all: scx_simple scx_qmap scx_central scx_pair scx_flatcg scx_userland
+all: scx_simple scx_qmap scx_central scx_pair scx_flatcg scx_userland scx_atropos
 
 # sort removes libbpf duplicates when not cross-building
 MAKE_DIRS := $(sort $(BUILD_DIR)/libbpf $(HOST_BUILD_DIR)/libbpf		\
@@ -190,12 +192,19 @@ scx_userland: scx_userland.c scx_userland.skel.h scx_userland.h user_exit_info.h
 	$(CC) $(CFLAGS) -c $< -o $@.o
 	$(CC) -o $@ $@.o $(HOST_BPFOBJ) $(LDFLAGS)
 
+scx_atropos: export RUSTFLAGS = -C link-args=-lzstd -C link-args=-lz -C link-args=-lelf -L $(BPFOBJ_DIR)
+scx_atropos: export ATROPOS_CLANG = $(CLANG)
+scx_atropos: export ATROPOS_BPF_CFLAGS = $(BPF_CFLAGS)
+scx_atropos: $(INCLUDE_DIR)/vmlinux.h
+	cargo build --manifest-path=scx_atropos/Cargo.toml $(CARGOFLAGS)
+
 clean:
+	cargo clean --manifest-path=scx_atropos/Cargo.toml
 	rm -rf $(SCRATCH_DIR) $(HOST_SCRATCH_DIR)
 	rm -f *.o *.bpf.o *.skel.h *.subskel.h
 	rm -f scx_simple scx_qmap scx_central scx_pair scx_flatcg scx_userland
 
-.PHONY: all clean
+.PHONY: all scx_atropos clean
 
 # delete failed targets
 .DELETE_ON_ERROR:
diff --git a/tools/sched_ext/scx_atropos/.gitignore b/tools/sched_ext/scx_atropos/.gitignore
new file mode 100644
index 000000000000..186dba259ec2
--- /dev/null
+++ b/tools/sched_ext/scx_atropos/.gitignore
@@ -0,0 +1,3 @@
+src/bpf/.output
+Cargo.lock
+target
diff --git a/tools/sched_ext/scx_atropos/Cargo.toml b/tools/sched_ext/scx_atropos/Cargo.toml
new file mode 100644
index 000000000000..a5ab02cb55f8
--- /dev/null
+++ b/tools/sched_ext/scx_atropos/Cargo.toml
@@ -0,0 +1,27 @@
+[package]
+name = "scx_atropos"
+version = "0.5.0"
+authors = ["Dan Schatzberg <dschatzberg@meta.com>", "Meta"]
+edition = "2021"
+description = "Userspace scheduling with BPF"
+license = "GPL-2.0-only"
+
+[dependencies]
+anyhow = "1.0.65"
+bitvec = { version = "1.0", features = ["serde"] }
+clap = { version = "4.1", features = ["derive", "env", "unicode", "wrap_help"] }
+ctrlc = { version = "3.1", features = ["termination"] }
+hex = "0.4.3"
+libbpf-rs = "0.19.1"
+libbpf-sys = { version = "1.0.4", features = ["novendor", "static"] }
+libc = "0.2.137"
+log = "0.4.17"
+ordered-float = "3.4.0"
+simplelog = "0.12.0"
+
+[build-dependencies]
+bindgen = { version = "0.61.0" }
+libbpf-cargo = "0.13.0"
+
+[features]
+enable_backtrace = []
diff --git a/tools/sched_ext/scx_atropos/build.rs b/tools/sched_ext/scx_atropos/build.rs
new file mode 100644
index 000000000000..26e792c5e17e
--- /dev/null
+++ b/tools/sched_ext/scx_atropos/build.rs
@@ -0,0 +1,70 @@
+// Copyright (c) Meta Platforms, Inc. and affiliates.
+
+// This software may be used and distributed according to the terms of the
+// GNU General Public License version 2.
+extern crate bindgen;
+
+use std::env;
+use std::fs::create_dir_all;
+use std::path::Path;
+use std::path::PathBuf;
+
+use libbpf_cargo::SkeletonBuilder;
+
+const HEADER_PATH: &str = "src/bpf/atropos.h";
+
+fn bindgen_atropos() {
+    // Tell cargo to invalidate the built crate whenever the wrapper changes
+    println!("cargo:rerun-if-changed={}", HEADER_PATH);
+
+    // The bindgen::Builder is the main entry point
+    // to bindgen, and lets you build up options for
+    // the resulting bindings.
+    let bindings = bindgen::Builder::default()
+        // The input header we would like to generate
+        // bindings for.
+        .header(HEADER_PATH)
+        // Tell cargo to invalidate the built crate whenever any of the
+        // included header files changed.
+        .parse_callbacks(Box::new(bindgen::CargoCallbacks))
+        // Finish the builder and generate the bindings.
+        .generate()
+        // Unwrap the Result and panic on failure.
+        .expect("Unable to generate bindings");
+
+    // Write the bindings to the $OUT_DIR/bindings.rs file.
+    let out_path = PathBuf::from(env::var("OUT_DIR").unwrap());
+    bindings
+        .write_to_file(out_path.join("atropos-sys.rs"))
+        .expect("Couldn't write bindings!");
+}
+
+fn gen_bpf_sched(name: &str) {
+    let bpf_cflags = env::var("ATROPOS_BPF_CFLAGS").unwrap();
+    let clang = env::var("ATROPOS_CLANG").unwrap();
+    eprintln!("{}", clang);
+    let outpath = format!("./src/bpf/.output/{}.skel.rs", name);
+    let skel = Path::new(&outpath);
+    let src = format!("./src/bpf/{}.bpf.c", name);
+    SkeletonBuilder::new()
+        .source(src.clone())
+        .clang(clang)
+        .clang_args(bpf_cflags)
+        .build_and_generate(&skel)
+        .unwrap();
+    println!("cargo:rerun-if-changed={}", src);
+}
+
+fn main() {
+    bindgen_atropos();
+    // It's unfortunate we cannot use `OUT_DIR` to store the generated skeleton.
+    // Reasons are because the generated skeleton contains compiler attributes
+    // that cannot be `include!()`ed via macro. And we cannot use the `#[path = "..."]`
+    // trick either because you cannot yet `concat!(env!("OUT_DIR"), "/skel.rs")` inside
+    // the path attribute either (see https://github.com/rust-lang/rust/pull/83366).
+    //
+    // However, there is hope! When the above feature stabilizes we can clean this
+    // all up.
+    create_dir_all("./src/bpf/.output").unwrap();
+    gen_bpf_sched("atropos");
+}
diff --git a/tools/sched_ext/scx_atropos/rustfmt.toml b/tools/sched_ext/scx_atropos/rustfmt.toml
new file mode 100644
index 000000000000..b7258ed0a8d8
--- /dev/null
+++ b/tools/sched_ext/scx_atropos/rustfmt.toml
@@ -0,0 +1,8 @@
+# Get help on options with `rustfmt --help=config`
+# Please keep these in alphabetical order.
+edition = "2021"
+group_imports = "StdExternalCrate"
+imports_granularity = "Item"
+merge_derives = false
+use_field_init_shorthand = true
+version = "Two"
diff --git a/tools/sched_ext/scx_atropos/src/atropos_sys.rs b/tools/sched_ext/scx_atropos/src/atropos_sys.rs
new file mode 100644
index 000000000000..bbeaf856d40e
--- /dev/null
+++ b/tools/sched_ext/scx_atropos/src/atropos_sys.rs
@@ -0,0 +1,10 @@
+// Copyright (c) Meta Platforms, Inc. and affiliates.
+
+// This software may be used and distributed according to the terms of the
+// GNU General Public License version 2.
+#![allow(non_upper_case_globals)]
+#![allow(non_camel_case_types)]
+#![allow(non_snake_case)]
+#![allow(dead_code)]
+
+include!(concat!(env!("OUT_DIR"), "/atropos-sys.rs"));
diff --git a/tools/sched_ext/scx_atropos/src/bpf/atropos.bpf.c b/tools/sched_ext/scx_atropos/src/bpf/atropos.bpf.c
new file mode 100644
index 000000000000..118fe728e886
--- /dev/null
+++ b/tools/sched_ext/scx_atropos/src/bpf/atropos.bpf.c
@@ -0,0 +1,978 @@
+/* Copyright (c) Meta Platforms, Inc. and affiliates. */
+/*
+ * This software may be used and distributed according to the terms of the
+ * GNU General Public License version 2.
+ *
+ * Atropos is a multi-domain BPF / userspace hybrid scheduler where the BPF
+ * part does simple round robin in each domain and the userspace part
+ * calculates the load factor of each domain and tells the BPF part how to load
+ * balance the domains.
+ *
+ * Every task has an entry in the task_data map which lists which domain the
+ * task belongs to. When a task first enters the system (atropos_prep_enable),
+ * they are round-robined to a domain.
+ *
+ * atropos_select_cpu is the primary scheduling logic, invoked when a task
+ * becomes runnable. The lb_data map is populated by userspace to inform the BPF
+ * scheduler that a task should be migrated to a new domain. Otherwise, the task
+ * is scheduled in priority order as follows:
+ * * The current core if the task was woken up synchronously and there are idle
+ *   cpus in the system
+ * * The previous core, if idle
+ * * The pinned-to core if the task is pinned to a specific core
+ * * Any idle cpu in the domain
+ *
+ * If none of the above conditions are met, then the task is enqueued to a
+ * dispatch queue corresponding to the domain (atropos_enqueue).
+ *
+ * atropos_dispatch will attempt to consume a task from its domain's
+ * corresponding dispatch queue (this occurs after scheduling any tasks directly
+ * assigned to it due to the logic in atropos_select_cpu). If no task is found,
+ * then greedy load stealing will attempt to find a task on another dispatch
+ * queue to run.
+ *
+ * Load balancing is almost entirely handled by userspace. BPF populates the
+ * task weight, dom mask and current dom in the task_data map and executes the
+ * load balance based on userspace populating the lb_data map.
+ */
+#include "../../../scx_common.bpf.h"
+#include "atropos.h"
+
+#include <errno.h>
+#include <stdbool.h>
+#include <string.h>
+#include <bpf/bpf_core_read.h>
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+
+char _license[] SEC("license") = "GPL";
+
+/*
+ * const volatiles are set during initialization and treated as consts by the
+ * jit compiler.
+ */
+
+/*
+ * Domains and cpus
+ */
+const volatile __u32 nr_doms = 32;	/* !0 for veristat, set during init */
+const volatile __u32 nr_cpus = 64;	/* !0 for veristat, set during init */
+const volatile __u32 cpu_dom_id_map[MAX_CPUS];
+const volatile __u64 dom_cpumasks[MAX_DOMS][MAX_CPUS / 64];
+
+const volatile bool kthreads_local;
+const volatile bool fifo_sched;
+const volatile bool switch_partial;
+const volatile __u32 greedy_threshold;
+
+/* base slice duration */
+const volatile __u64 slice_ns = SCX_SLICE_DFL;
+
+/*
+ * Exit info
+ */
+int exit_type = SCX_EXIT_NONE;
+char exit_msg[SCX_EXIT_MSG_LEN];
+
+/*
+ * Per-CPU context
+ */
+struct pcpu_ctx {
+	__u32 dom_rr_cur; /* used when scanning other doms */
+
+	/* libbpf-rs does not respect the alignment, so pad out the struct explicitly */
+	__u8 _padding[CACHELINE_SIZE - sizeof(u32)];
+} __attribute__((aligned(CACHELINE_SIZE)));
+
+struct pcpu_ctx pcpu_ctx[MAX_CPUS];
+
+/*
+ * Domain context
+ */
+struct dom_ctx {
+	struct bpf_cpumask __kptr *cpumask;
+	struct bpf_cpumask __kptr *direct_greedy_cpumask;
+	u64 vtime_now;
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_ARRAY);
+	__type(key, u32);
+	__type(value, struct dom_ctx);
+	__uint(max_entries, MAX_DOMS);
+	__uint(map_flags, 0);
+} dom_ctx SEC(".maps");
+
+/*
+ * Statistics
+ */
+struct {
+	__uint(type, BPF_MAP_TYPE_PERCPU_ARRAY);
+	__uint(key_size, sizeof(u32));
+	__uint(value_size, sizeof(u64));
+	__uint(max_entries, ATROPOS_NR_STATS);
+} stats SEC(".maps");
+
+static inline void stat_add(enum stat_idx idx, u64 addend)
+{
+	u32 idx_v = idx;
+
+	u64 *cnt_p = bpf_map_lookup_elem(&stats, &idx_v);
+	if (cnt_p)
+		(*cnt_p) += addend;
+}
+
+/* Map pid -> task_ctx */
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__type(key, pid_t);
+	__type(value, struct task_ctx);
+	__uint(max_entries, 1000000);
+	__uint(map_flags, 0);
+} task_data SEC(".maps");
+
+/*
+ * This is populated from userspace to indicate which pids should be reassigned
+ * to new doms.
+ */
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__type(key, pid_t);
+	__type(value, u32);
+	__uint(max_entries, 1000);
+	__uint(map_flags, 0);
+} lb_data SEC(".maps");
+
+/*
+ * Userspace tuner will frequently update the following struct with tuning
+ * parameters and bump its gen. refresh_tune_params() converts them into forms
+ * that can be used directly in the scheduling paths.
+ */
+struct tune_input{
+	__u64 gen;
+	__u64 direct_greedy_cpumask[MAX_CPUS / 64];
+	__u64 kick_greedy_cpumask[MAX_CPUS / 64];
+} tune_input;
+
+__u64 tune_params_gen;
+private(A) struct bpf_cpumask __kptr *direct_greedy_cpumask;
+private(A) struct bpf_cpumask __kptr *kick_greedy_cpumask;
+
+static inline bool vtime_before(u64 a, u64 b)
+{
+	return (s64)(a - b) < 0;
+}
+
+static u32 cpu_to_dom_id(s32 cpu)
+{
+	const volatile u32 *dom_idp;
+
+	if (nr_doms <= 1)
+		return 0;
+
+	dom_idp = MEMBER_VPTR(cpu_dom_id_map, [cpu]);
+	if (!dom_idp)
+		return MAX_DOMS;
+
+	return *dom_idp;
+}
+
+static void refresh_tune_params(void)
+{
+	s32 cpu;
+
+	if (tune_params_gen == tune_input.gen)
+		return;
+
+	tune_params_gen = tune_input.gen;
+
+	bpf_for(cpu, 0, nr_cpus) {
+		u32 dom_id = cpu_to_dom_id(cpu);
+		struct dom_ctx *domc;
+
+		if (!(domc = bpf_map_lookup_elem(&dom_ctx, &dom_id))) {
+			scx_bpf_error("Failed to lookup dom[%u]", dom_id);
+			return;
+		}
+
+		if (tune_input.direct_greedy_cpumask[cpu / 64] & (1LLU << (cpu % 64))) {
+			if (direct_greedy_cpumask)
+				bpf_cpumask_set_cpu(cpu, direct_greedy_cpumask);
+			if (domc->direct_greedy_cpumask)
+				bpf_cpumask_set_cpu(cpu, domc->direct_greedy_cpumask);
+		} else {
+			if (direct_greedy_cpumask)
+				bpf_cpumask_clear_cpu(cpu, direct_greedy_cpumask);
+			if (domc->direct_greedy_cpumask)
+				bpf_cpumask_clear_cpu(cpu, domc->direct_greedy_cpumask);
+		}
+
+		if (tune_input.kick_greedy_cpumask[cpu / 64] & (1LLU << (cpu % 64))) {
+			if (kick_greedy_cpumask)
+				bpf_cpumask_set_cpu(cpu, kick_greedy_cpumask);
+		} else {
+			if (kick_greedy_cpumask)
+				bpf_cpumask_clear_cpu(cpu, kick_greedy_cpumask);
+		}
+	}
+}
+
+static bool task_set_domain(struct task_ctx *task_ctx, struct task_struct *p,
+			    u32 new_dom_id, bool init_dsq_vtime)
+{
+	struct dom_ctx *old_domc, *new_domc;
+	struct bpf_cpumask *d_cpumask, *t_cpumask;
+	u32 old_dom_id = task_ctx->dom_id;
+	s64 vtime_delta;
+
+	old_domc = bpf_map_lookup_elem(&dom_ctx, &old_dom_id);
+	if (!old_domc) {
+		scx_bpf_error("Failed to lookup old dom%u", old_dom_id);
+		return false;
+	}
+
+	if (init_dsq_vtime)
+		vtime_delta = 0;
+	else
+		vtime_delta = p->scx.dsq_vtime - old_domc->vtime_now;
+
+	new_domc = bpf_map_lookup_elem(&dom_ctx, &new_dom_id);
+	if (!new_domc) {
+		scx_bpf_error("Failed to lookup new dom%u", new_dom_id);
+		return false;
+	}
+
+	d_cpumask = new_domc->cpumask;
+	if (!d_cpumask) {
+		scx_bpf_error("Failed to get dom%u cpumask kptr",
+			      new_dom_id);
+		return false;
+	}
+
+	t_cpumask = task_ctx->cpumask;
+	if (!t_cpumask) {
+		scx_bpf_error("Failed to look up task cpumask");
+		return false;
+	}
+
+	/*
+	 * set_cpumask might have happened between userspace requesting LB and
+	 * here and @p might not be able to run in @dom_id anymore. Verify.
+	 */
+	if (bpf_cpumask_intersects((const struct cpumask *)d_cpumask,
+				   p->cpus_ptr)) {
+		p->scx.dsq_vtime = new_domc->vtime_now + vtime_delta;
+		task_ctx->dom_id = new_dom_id;
+		bpf_cpumask_and(t_cpumask, (const struct cpumask *)d_cpumask,
+				p->cpus_ptr);
+	}
+
+	return task_ctx->dom_id == new_dom_id;
+}
+
+s32 BPF_STRUCT_OPS(atropos_select_cpu, struct task_struct *p, s32 prev_cpu,
+		   u64 wake_flags)
+{
+	struct cpumask *idle_smtmask = scx_bpf_get_idle_smtmask();
+	struct task_ctx *task_ctx;
+	struct bpf_cpumask *p_cpumask;
+	pid_t pid = p->pid;
+	bool prev_domestic, has_idle_cores;
+	s32 cpu;
+
+	refresh_tune_params();
+
+	if (!(task_ctx = bpf_map_lookup_elem(&task_data, &pid)) ||
+	    !(p_cpumask = task_ctx->cpumask))
+		goto enoent;
+
+	if (kthreads_local &&
+	    (p->flags & PF_KTHREAD) && p->nr_cpus_allowed == 1) {
+		cpu = prev_cpu;
+		stat_add(ATROPOS_STAT_DIRECT_DISPATCH, 1);
+		goto direct;
+	}
+
+	/*
+	 * If WAKE_SYNC and the machine isn't fully saturated, wake up @p to the
+	 * local dsq of the waker.
+	 */
+	if (p->nr_cpus_allowed > 1 && (wake_flags & SCX_WAKE_SYNC)) {
+		struct task_struct *current = (void *)bpf_get_current_task();
+
+		if (!(BPF_CORE_READ(current, flags) & PF_EXITING) &&
+		    task_ctx->dom_id < MAX_DOMS) {
+			struct dom_ctx *domc;
+			struct bpf_cpumask *d_cpumask;
+			const struct cpumask *idle_cpumask;
+			bool has_idle;
+
+			domc = bpf_map_lookup_elem(&dom_ctx, &task_ctx->dom_id);
+			if (!domc) {
+				scx_bpf_error("Failed to find dom%u",
+					      task_ctx->dom_id);
+				goto enoent;
+			}
+			d_cpumask = domc->cpumask;
+			if (!d_cpumask) {
+				scx_bpf_error("Failed to acquire dom%u cpumask kptr",
+					      task_ctx->dom_id);
+				goto enoent;
+			}
+
+			idle_cpumask = scx_bpf_get_idle_cpumask();
+
+			has_idle = bpf_cpumask_intersects((const struct cpumask *)d_cpumask,
+							  idle_cpumask);
+
+			scx_bpf_put_idle_cpumask(idle_cpumask);
+
+			if (has_idle) {
+				cpu = bpf_get_smp_processor_id();
+				if (bpf_cpumask_test_cpu(cpu, p->cpus_ptr)) {
+					stat_add(ATROPOS_STAT_WAKE_SYNC, 1);
+					goto direct;
+				}
+			}
+		}
+	}
+
+	/* If only one CPU is allowed, dispatch */
+	if (p->nr_cpus_allowed == 1) {
+		stat_add(ATROPOS_STAT_PINNED, 1);
+		cpu = prev_cpu;
+		goto direct;
+	}
+
+	has_idle_cores = !bpf_cpumask_empty(idle_smtmask);
+
+	/* did @p get pulled out to a foreign domain by e.g. greedy execution? */
+	prev_domestic = bpf_cpumask_test_cpu(prev_cpu,
+					     (const struct cpumask *)p_cpumask);
+
+	/*
+	 * See if we want to keep @prev_cpu. We want to keep @prev_cpu if the
+	 * whole physical core is idle. If the sibling[s] are busy, it's likely
+	 * more advantageous to look for wholly idle cores first.
+	 */
+	if (prev_domestic) {
+		if (bpf_cpumask_test_cpu(prev_cpu, idle_smtmask) &&
+		    scx_bpf_test_and_clear_cpu_idle(prev_cpu)) {
+			stat_add(ATROPOS_STAT_PREV_IDLE, 1);
+			cpu = prev_cpu;
+			goto direct;
+		}
+	} else {
+		/*
+		 * @prev_cpu is foreign. Linger iff the domain isn't too busy as
+		 * indicated by direct_greedy_cpumask. There may also be an idle
+		 * CPU in the domestic domain
+		 */
+		if (direct_greedy_cpumask &&
+		    bpf_cpumask_test_cpu(prev_cpu, (const struct cpumask *)
+					 direct_greedy_cpumask) &&
+		    bpf_cpumask_test_cpu(prev_cpu, idle_smtmask) &&
+		    scx_bpf_test_and_clear_cpu_idle(prev_cpu)) {
+			stat_add(ATROPOS_STAT_GREEDY_IDLE, 1);
+			cpu = prev_cpu;
+			goto direct;
+		}
+	}
+
+	/*
+	 * @prev_cpu didn't work out. Let's see whether there's an idle CPU @p
+	 * can be directly dispatched to. We'll first try to find the best idle
+	 * domestic CPU and then move onto foreign.
+	 */
+
+	/* If there is a domestic idle core, dispatch directly */
+	if (has_idle_cores) {
+		cpu = scx_bpf_pick_idle_cpu((const struct cpumask *)p_cpumask,
+					    SCX_PICK_IDLE_CORE);
+		if (cpu >= 0) {
+			stat_add(ATROPOS_STAT_DIRECT_DISPATCH, 1);
+			goto direct;
+		}
+	}
+
+	/*
+	 * If @prev_cpu was domestic and is idle itself even though the core
+	 * isn't, picking @prev_cpu may improve L1/2 locality.
+	 */
+	if (prev_domestic && scx_bpf_test_and_clear_cpu_idle(prev_cpu)) {
+		stat_add(ATROPOS_STAT_DIRECT_DISPATCH, 1);
+		cpu = prev_cpu;
+		goto direct;
+	}
+
+	/* If there is any domestic idle CPU, dispatch directly */
+	cpu = scx_bpf_pick_idle_cpu((const struct cpumask *)p_cpumask, 0);
+	if (cpu >= 0) {
+		stat_add(ATROPOS_STAT_DIRECT_DISPATCH, 1);
+		goto direct;
+	}
+
+	/*
+	 * Domestic domain is fully booked. If there are CPUs which are idle and
+	 * under-utilized, ignore domain boundaries and push the task there. Try
+	 * to find an idle core first.
+	 */
+	if (task_ctx->all_cpus && direct_greedy_cpumask &&
+	    !bpf_cpumask_empty((const struct cpumask *)direct_greedy_cpumask)) {
+		u32 dom_id = cpu_to_dom_id(prev_cpu);
+		struct dom_ctx *domc;
+
+		if (!(domc = bpf_map_lookup_elem(&dom_ctx, &dom_id))) {
+			scx_bpf_error("Failed to lookup dom[%u]", dom_id);
+			goto enoent;
+		}
+
+		/* Try to find an idle core in the previous and then any domain */
+		if (has_idle_cores) {
+			if (domc->direct_greedy_cpumask) {
+				cpu = scx_bpf_pick_idle_cpu((const struct cpumask *)
+							    domc->direct_greedy_cpumask,
+							    SCX_PICK_IDLE_CORE);
+				if (cpu >= 0) {
+					stat_add(ATROPOS_STAT_DIRECT_GREEDY, 1);
+					goto direct;
+				}
+			}
+
+			if (direct_greedy_cpumask) {
+				cpu = scx_bpf_pick_idle_cpu((const struct cpumask *)
+							    direct_greedy_cpumask,
+							    SCX_PICK_IDLE_CORE);
+				if (cpu >= 0) {
+					stat_add(ATROPOS_STAT_DIRECT_GREEDY_FAR, 1);
+					goto direct;
+				}
+			}
+		}
+
+		/*
+		 * No idle core. Is there any idle CPU?
+		 */
+		if (domc->direct_greedy_cpumask) {
+			cpu = scx_bpf_pick_idle_cpu((const struct cpumask *)
+						    domc->direct_greedy_cpumask, 0);
+			if (cpu >= 0) {
+				stat_add(ATROPOS_STAT_DIRECT_GREEDY, 1);
+				goto direct;
+			}
+		}
+
+		if (direct_greedy_cpumask) {
+			cpu = scx_bpf_pick_idle_cpu((const struct cpumask *)
+						    direct_greedy_cpumask, 0);
+			if (cpu >= 0) {
+				stat_add(ATROPOS_STAT_DIRECT_GREEDY_FAR, 1);
+				goto direct;
+			}
+		}
+	}
+
+	/*
+	 * We're going to queue on the domestic domain's DSQ. @prev_cpu may be
+	 * in a different domain. Returning an out-of-domain CPU can lead to
+	 * stalls as all in-domain CPUs may be idle by the time @p gets
+	 * enqueued.
+	 */
+	if (prev_domestic)
+		cpu = prev_cpu;
+	else
+		cpu = scx_bpf_pick_any_cpu((const struct cpumask *)p_cpumask, 0);
+
+	scx_bpf_put_idle_cpumask(idle_smtmask);
+	return cpu;
+
+direct:
+	task_ctx->dispatch_local = true;
+	scx_bpf_put_idle_cpumask(idle_smtmask);
+	return cpu;
+
+enoent:
+	scx_bpf_put_idle_cpumask(idle_smtmask);
+	return -ENOENT;
+}
+
+void BPF_STRUCT_OPS(atropos_enqueue, struct task_struct *p, u64 enq_flags)
+{
+	struct task_ctx *task_ctx;
+	struct bpf_cpumask *p_cpumask;
+	pid_t pid = p->pid;
+	u32 *new_dom;
+	s32 cpu;
+
+	if (!(task_ctx = bpf_map_lookup_elem(&task_data, &pid)) ||
+	    !(p_cpumask = task_ctx->cpumask)) {
+		scx_bpf_error("Failed to lookup task_ctx or cpumask");
+		return;
+	}
+
+	/*
+	 * Migrate @p to a new domain if requested by userland through lb_data.
+	 */
+	new_dom = bpf_map_lookup_elem(&lb_data, &pid);
+	if (new_dom && *new_dom != task_ctx->dom_id &&
+	    task_set_domain(task_ctx, p, *new_dom, false)) {
+		stat_add(ATROPOS_STAT_LOAD_BALANCE, 1);
+		task_ctx->dispatch_local = false;
+		cpu = scx_bpf_pick_any_cpu((const struct cpumask *)p_cpumask, 0);
+		if (cpu >= 0)
+			scx_bpf_kick_cpu(cpu, 0);
+		goto dom_queue;
+	}
+
+	if (task_ctx->dispatch_local) {
+		task_ctx->dispatch_local = false;
+		scx_bpf_dispatch(p, SCX_DSQ_LOCAL, slice_ns, enq_flags);
+		return;
+	}
+
+	/*
+	 * @p is about to be queued on its domain's dsq. However, @p may be on a
+	 * foreign CPU due to a greedy execution and not have gone through
+	 * ->select_cpu() if it's being enqueued e.g. after slice exhaustion. If
+	 * so, @p would be queued on its domain's dsq but none of the CPUs in
+	 * the domain would be woken up which can induce temporary execution
+	 * stalls. Kick a domestic CPU if @p is on a foreign domain.
+	 */
+	if (!bpf_cpumask_test_cpu(scx_bpf_task_cpu(p), (const struct cpumask *)p_cpumask)) {
+		cpu = scx_bpf_pick_any_cpu((const struct cpumask *)p_cpumask, 0);
+		scx_bpf_kick_cpu(cpu, 0);
+		stat_add(ATROPOS_STAT_REPATRIATE, 1);
+	}
+
+dom_queue:
+	if (fifo_sched) {
+		scx_bpf_dispatch(p, task_ctx->dom_id, slice_ns,
+				 enq_flags);
+	} else {
+		u64 vtime = p->scx.dsq_vtime;
+		u32 dom_id = task_ctx->dom_id;
+		struct dom_ctx *domc;
+
+		domc = bpf_map_lookup_elem(&dom_ctx, &dom_id);
+		if (!domc) {
+			scx_bpf_error("Failed to lookup dom[%u]", dom_id);
+			return;
+		}
+
+		/*
+		 * Limit the amount of budget that an idling task can accumulate
+		 * to one slice.
+		 */
+		if (vtime_before(vtime, domc->vtime_now - slice_ns))
+			vtime = domc->vtime_now - slice_ns;
+
+		scx_bpf_dispatch_vtime(p, task_ctx->dom_id, slice_ns, vtime,
+				       enq_flags);
+	}
+
+	/*
+	 * If there are CPUs which are idle and not saturated, wake them up to
+	 * see whether they'd be able to steal the just queued task. This path
+	 * is taken only if DIRECT_GREEDY didn't trigger in select_cpu().
+	 *
+	 * While both mechanisms serve very similar purposes, DIRECT_GREEDY
+	 * emplaces the task in a foreign CPU directly while KICK_GREEDY just
+	 * wakes up a foreign CPU which will then first try to execute from its
+	 * domestic domain first before snooping foreign ones.
+	 *
+	 * While KICK_GREEDY is a more expensive way of accelerating greedy
+	 * execution, DIRECT_GREEDY shows negative performance impacts when the
+	 * CPUs are highly loaded while KICK_GREEDY doesn't. Even under fairly
+	 * high utilization, KICK_GREEDY can slightly improve work-conservation.
+	 */
+	if (task_ctx->all_cpus && kick_greedy_cpumask) {
+		cpu = scx_bpf_pick_idle_cpu((const struct cpumask *)
+					    kick_greedy_cpumask, 0);
+		if (cpu >= 0) {
+			stat_add(ATROPOS_STAT_KICK_GREEDY, 1);
+			scx_bpf_kick_cpu(cpu, 0);
+		}
+	}
+}
+
+static bool cpumask_intersects_domain(const struct cpumask *cpumask, u32 dom_id)
+{
+	s32 cpu;
+
+	if (dom_id >= MAX_DOMS)
+		return false;
+
+	bpf_for(cpu, 0, nr_cpus) {
+		if (bpf_cpumask_test_cpu(cpu, cpumask) &&
+		    (dom_cpumasks[dom_id][cpu / 64] & (1LLU << (cpu % 64))))
+			return true;
+	}
+	return false;
+}
+
+static u32 dom_rr_next(s32 cpu)
+{
+	struct pcpu_ctx *pcpuc;
+	u32 dom_id;
+
+	pcpuc = MEMBER_VPTR(pcpu_ctx, [cpu]);
+	if (!pcpuc)
+		return 0;
+
+	dom_id = (pcpuc->dom_rr_cur + 1) % nr_doms;
+
+	if (dom_id == cpu_to_dom_id(cpu))
+		dom_id = (dom_id + 1) % nr_doms;
+
+	pcpuc->dom_rr_cur = dom_id;
+	return dom_id;
+}
+
+void BPF_STRUCT_OPS(atropos_dispatch, s32 cpu, struct task_struct *prev)
+{
+	u32 dom = cpu_to_dom_id(cpu);
+
+	if (scx_bpf_consume(dom)) {
+		stat_add(ATROPOS_STAT_DSQ_DISPATCH, 1);
+		return;
+	}
+
+	if (!greedy_threshold)
+		return;
+
+	bpf_repeat(nr_doms - 1) {
+		u32 dom_id = dom_rr_next(cpu);
+
+		if (scx_bpf_dsq_nr_queued(dom_id) >= greedy_threshold &&
+		    scx_bpf_consume(dom_id)) {
+			stat_add(ATROPOS_STAT_GREEDY, 1);
+			break;
+		}
+	}
+}
+
+void BPF_STRUCT_OPS(atropos_runnable, struct task_struct *p, u64 enq_flags)
+{
+	struct task_ctx *task_ctx;
+	pid_t pid = p->pid;
+
+	if (!(task_ctx = bpf_map_lookup_elem(&task_data, &pid))) {
+		scx_bpf_error("Failed to lookup task_ctx");
+		return;
+	}
+
+	task_ctx->runnable_at = bpf_ktime_get_ns();
+	task_ctx->is_kworker = p->flags & PF_WQ_WORKER;
+}
+
+void BPF_STRUCT_OPS(atropos_running, struct task_struct *p)
+{
+	struct task_ctx *taskc;
+	struct dom_ctx *domc;
+	pid_t pid = p->pid;
+	u32 dom_id;
+
+	if (fifo_sched)
+		return;
+
+	taskc = bpf_map_lookup_elem(&task_data, &pid);
+	if (!taskc) {
+		scx_bpf_error("Failed to lookup task_ctx");
+		return;
+	}
+	dom_id = taskc->dom_id;
+
+	domc = bpf_map_lookup_elem(&dom_ctx, &dom_id);
+	if (!domc) {
+		scx_bpf_error("Failed to lookup dom[%u]", dom_id);
+		return;
+	}
+
+	/*
+	 * Global vtime always progresses forward as tasks start executing. The
+	 * test and update can be performed concurrently from multiple CPUs and
+	 * thus racy. Any error should be contained and temporary. Let's just
+	 * live with it.
+	 */
+	if (vtime_before(domc->vtime_now, p->scx.dsq_vtime))
+		domc->vtime_now = p->scx.dsq_vtime;
+}
+
+void BPF_STRUCT_OPS(atropos_stopping, struct task_struct *p, bool runnable)
+{
+	if (fifo_sched)
+		return;
+
+	/* scale the execution time by the inverse of the weight and charge */
+	p->scx.dsq_vtime += (slice_ns - p->scx.slice) * 100 / p->scx.weight;
+}
+
+void BPF_STRUCT_OPS(atropos_quiescent, struct task_struct *p, u64 deq_flags)
+{
+	struct task_ctx *task_ctx;
+	pid_t pid = p->pid;
+
+	if (!(task_ctx = bpf_map_lookup_elem(&task_data, &pid))) {
+		scx_bpf_error("Failed to lookup task_ctx");
+		return;
+	}
+
+	task_ctx->runnable_for += bpf_ktime_get_ns() - task_ctx->runnable_at;
+	task_ctx->runnable_at = 0;
+}
+
+void BPF_STRUCT_OPS(atropos_set_weight, struct task_struct *p, u32 weight)
+{
+	struct task_ctx *task_ctx;
+	pid_t pid = p->pid;
+
+	if (!(task_ctx = bpf_map_lookup_elem(&task_data, &pid))) {
+		scx_bpf_error("Failed to lookup task_ctx");
+		return;
+	}
+
+	task_ctx->weight = weight;
+}
+
+static u32 task_pick_domain(struct task_ctx *task_ctx, struct task_struct *p,
+			    const struct cpumask *cpumask)
+{
+	s32 cpu = bpf_get_smp_processor_id();
+	u32 first_dom = MAX_DOMS, dom;
+
+	if (cpu < 0 || cpu >= MAX_CPUS)
+		return MAX_DOMS;
+
+	task_ctx->dom_mask = 0;
+
+	dom = pcpu_ctx[cpu].dom_rr_cur++;
+	bpf_repeat(nr_doms) {
+		dom = (dom + 1) % nr_doms;
+		if (cpumask_intersects_domain(cpumask, dom)) {
+			task_ctx->dom_mask |= 1LLU << dom;
+			/*
+			 * AsThe starting point is round-robin'd and the first
+			 * match should be spread across all the domains.
+			 */
+			if (first_dom == MAX_DOMS)
+				first_dom = dom;
+		}
+	}
+
+	return first_dom;
+}
+
+static void task_pick_and_set_domain(struct task_ctx *task_ctx,
+				     struct task_struct *p,
+				     const struct cpumask *cpumask,
+				     bool init_dsq_vtime)
+{
+	u32 dom_id = 0;
+
+	if (nr_doms > 1)
+		dom_id = task_pick_domain(task_ctx, p, cpumask);
+
+	if (!task_set_domain(task_ctx, p, dom_id, init_dsq_vtime))
+		scx_bpf_error("Failed to set dom%d for %s[%d]",
+			      dom_id, p->comm, p->pid);
+}
+
+void BPF_STRUCT_OPS(atropos_set_cpumask, struct task_struct *p,
+		    const struct cpumask *cpumask)
+{
+	struct task_ctx *task_ctx;
+	pid_t pid = p->pid;
+
+	if (!(task_ctx = bpf_map_lookup_elem(&task_data, &pid))) {
+		scx_bpf_error("Failed to lookup task_ctx for %s[%d]",
+			      p->comm, pid);
+		return;
+	}
+
+	task_pick_and_set_domain(task_ctx, p, cpumask, false);
+	task_ctx->all_cpus = bpf_cpumask_full(cpumask);
+}
+
+s32 BPF_STRUCT_OPS(atropos_prep_enable, struct task_struct *p,
+		   struct scx_enable_args *args)
+{
+	struct bpf_cpumask *cpumask;
+	struct task_ctx task_ctx, *map_value;
+	long ret;
+	pid_t pid;
+
+	memset(&task_ctx, 0, sizeof(task_ctx));
+
+	pid = p->pid;
+	ret = bpf_map_update_elem(&task_data, &pid, &task_ctx, BPF_NOEXIST);
+	if (ret) {
+		stat_add(ATROPOS_STAT_TASK_GET_ERR, 1);
+		return ret;
+	}
+
+	/*
+	 * Read the entry from the map immediately so we can add the cpumask
+	 * with bpf_kptr_xchg().
+	 */
+	map_value = bpf_map_lookup_elem(&task_data, &pid);
+	if (!map_value)
+		/* Should never happen -- it was just inserted above. */
+		return -EINVAL;
+
+	cpumask = bpf_cpumask_create();
+	if (!cpumask) {
+		bpf_map_delete_elem(&task_data, &pid);
+		return -ENOMEM;
+	}
+
+	cpumask = bpf_kptr_xchg(&map_value->cpumask, cpumask);
+	if (cpumask) {
+		/* Should never happen as we just inserted it above. */
+		bpf_cpumask_release(cpumask);
+		bpf_map_delete_elem(&task_data, &pid);
+		return -EINVAL;
+	}
+
+	task_pick_and_set_domain(map_value, p, p->cpus_ptr, true);
+
+	return 0;
+}
+
+void BPF_STRUCT_OPS(atropos_disable, struct task_struct *p)
+{
+	pid_t pid = p->pid;
+	long ret = bpf_map_delete_elem(&task_data, &pid);
+	if (ret) {
+		stat_add(ATROPOS_STAT_TASK_GET_ERR, 1);
+		return;
+	}
+}
+
+static s32 create_dom(u32 dom_id)
+{
+	struct dom_ctx domc_init = {}, *domc;
+	struct bpf_cpumask *cpumask;
+	u32 cpu;
+	s32 ret;
+
+	ret = scx_bpf_create_dsq(dom_id, -1);
+	if (ret < 0) {
+		scx_bpf_error("Failed to create dsq %u (%d)", dom_id, ret);
+		return ret;
+	}
+
+	ret = bpf_map_update_elem(&dom_ctx, &dom_id, &domc_init, 0);
+	if (ret) {
+		scx_bpf_error("Failed to add dom_ctx entry %u (%d)", dom_id, ret);
+		return ret;
+	}
+
+	domc = bpf_map_lookup_elem(&dom_ctx, &dom_id);
+	if (!domc) {
+		/* Should never happen, we just inserted it above. */
+		scx_bpf_error("No dom%u", dom_id);
+		return -ENOENT;
+	}
+
+	cpumask = bpf_cpumask_create();
+	if (!cpumask) {
+		scx_bpf_error("Failed to create BPF cpumask for domain %u", dom_id);
+		return -ENOMEM;
+	}
+
+	for (cpu = 0; cpu < MAX_CPUS; cpu++) {
+		const volatile __u64 *dmask;
+
+		dmask = MEMBER_VPTR(dom_cpumasks, [dom_id][cpu / 64]);
+		if (!dmask) {
+			scx_bpf_error("array index error");
+			bpf_cpumask_release(cpumask);
+			return -ENOENT;
+		}
+
+		if (*dmask & (1LLU << (cpu % 64)))
+			bpf_cpumask_set_cpu(cpu, cpumask);
+	}
+
+	cpumask = bpf_kptr_xchg(&domc->cpumask, cpumask);
+	if (cpumask) {
+		scx_bpf_error("Domain %u cpumask already present", dom_id);
+		bpf_cpumask_release(cpumask);
+		return -EEXIST;
+	}
+
+	cpumask = bpf_cpumask_create();
+	if (!cpumask) {
+		scx_bpf_error("Failed to create BPF cpumask for domain %u",
+			      dom_id);
+		return -ENOMEM;
+	}
+
+	cpumask = bpf_kptr_xchg(&domc->direct_greedy_cpumask, cpumask);
+	if (cpumask) {
+		scx_bpf_error("Domain %u direct_greedy_cpumask already present",
+			      dom_id);
+		bpf_cpumask_release(cpumask);
+		return -EEXIST;
+	}
+
+	return 0;
+}
+
+s32 BPF_STRUCT_OPS_SLEEPABLE(atropos_init)
+{
+	struct bpf_cpumask *cpumask;
+	s32 i, ret;
+
+	if (!switch_partial)
+		scx_bpf_switch_all();
+
+	bpf_for(i, 0, nr_doms) {
+		ret = create_dom(i);
+		if (ret)
+			return ret;
+	}
+
+	for (u32 i = 0; i < nr_cpus; i++)
+		pcpu_ctx[i].dom_rr_cur = i;
+
+	cpumask = bpf_cpumask_create();
+	if (!cpumask)
+		return -ENOMEM;
+	cpumask = bpf_kptr_xchg(&direct_greedy_cpumask, cpumask);
+	if (cpumask)
+		bpf_cpumask_release(cpumask);
+
+	cpumask = bpf_cpumask_create();
+	if (!cpumask)
+		return -ENOMEM;
+	cpumask = bpf_kptr_xchg(&kick_greedy_cpumask, cpumask);
+	if (cpumask)
+		bpf_cpumask_release(cpumask);
+
+	return 0;
+}
+
+void BPF_STRUCT_OPS(atropos_exit, struct scx_exit_info *ei)
+{
+	bpf_probe_read_kernel_str(exit_msg, sizeof(exit_msg), ei->msg);
+	exit_type = ei->type;
+}
+
+SEC(".struct_ops.link")
+struct sched_ext_ops atropos = {
+	.select_cpu		= (void *)atropos_select_cpu,
+	.enqueue		= (void *)atropos_enqueue,
+	.dispatch		= (void *)atropos_dispatch,
+	.runnable		= (void *)atropos_runnable,
+	.running		= (void *)atropos_running,
+	.stopping		= (void *)atropos_stopping,
+	.quiescent		= (void *)atropos_quiescent,
+	.set_weight		= (void *)atropos_set_weight,
+	.set_cpumask		= (void *)atropos_set_cpumask,
+	.prep_enable		= (void *)atropos_prep_enable,
+	.disable		= (void *)atropos_disable,
+	.init			= (void *)atropos_init,
+	.exit			= (void *)atropos_exit,
+	.name			= "atropos",
+};
diff --git a/tools/sched_ext/scx_atropos/src/bpf/atropos.h b/tools/sched_ext/scx_atropos/src/bpf/atropos.h
new file mode 100644
index 000000000000..894782e32fa1
--- /dev/null
+++ b/tools/sched_ext/scx_atropos/src/bpf/atropos.h
@@ -0,0 +1,64 @@
+// Copyright (c) Meta Platforms, Inc. and affiliates.
+
+// This software may be used and distributed according to the terms of the
+// GNU General Public License version 2.
+#ifndef __ATROPOS_H
+#define __ATROPOS_H
+
+#include <stdbool.h>
+#ifndef __kptr
+#ifdef __KERNEL__
+#error "__kptr_ref not defined in the kernel"
+#endif
+#define __kptr
+#endif
+
+#define	MAX_CPUS 512
+#define	MAX_DOMS 64 /* limited to avoid complex bitmask ops */
+#define	CACHELINE_SIZE 64
+
+/* Statistics */
+enum stat_idx {
+	/* The following fields add up to all dispatched tasks */
+	ATROPOS_STAT_WAKE_SYNC,
+	ATROPOS_STAT_PREV_IDLE,
+	ATROPOS_STAT_GREEDY_IDLE,
+	ATROPOS_STAT_PINNED,
+	ATROPOS_STAT_DIRECT_DISPATCH,
+	ATROPOS_STAT_DIRECT_GREEDY,
+	ATROPOS_STAT_DIRECT_GREEDY_FAR,
+	ATROPOS_STAT_DSQ_DISPATCH,
+	ATROPOS_STAT_GREEDY,
+
+	/* Extra stats that don't contribute to total */
+	ATROPOS_STAT_REPATRIATE,
+	ATROPOS_STAT_KICK_GREEDY,
+	ATROPOS_STAT_LOAD_BALANCE,
+
+	/* Errors */
+	ATROPOS_STAT_TASK_GET_ERR,
+
+	ATROPOS_NR_STATS,
+};
+
+struct task_ctx {
+	/* The domains this task can run on */
+	unsigned long long dom_mask;
+
+	struct bpf_cpumask __kptr *cpumask;
+	unsigned int dom_id;
+	unsigned int weight;
+	unsigned long long runnable_at;
+	unsigned long long runnable_for;
+
+	/* The task is a workqueue worker thread */
+	bool is_kworker;
+
+	/* Allowed on all CPUs and eligible for DIRECT_GREEDY optimization */
+	bool all_cpus;
+
+	/* select_cpu() telling enqueue() to queue directly on the DSQ */
+	bool dispatch_local;
+};
+
+#endif /* __ATROPOS_H */
diff --git a/tools/sched_ext/scx_atropos/src/main.rs b/tools/sched_ext/scx_atropos/src/main.rs
new file mode 100644
index 000000000000..6d8ea6f4ef3c
--- /dev/null
+++ b/tools/sched_ext/scx_atropos/src/main.rs
@@ -0,0 +1,1196 @@
+// Copyright (c) Meta Platforms, Inc. and affiliates.
+
+// This software may be used and distributed according to the terms of the
+// GNU General Public License version 2.
+#[path = "bpf/.output/atropos.skel.rs"]
+mod atropos;
+pub use atropos::*;
+pub mod atropos_sys;
+
+use std::cell::Cell;
+use std::collections::BTreeMap;
+use std::collections::BTreeSet;
+use std::ffi::CStr;
+use std::ops::Bound::Included;
+use std::ops::Bound::Unbounded;
+use std::sync::atomic::AtomicBool;
+use std::sync::atomic::Ordering;
+use std::sync::Arc;
+use std::time::Duration;
+use std::time::Instant;
+
+use anyhow::anyhow;
+use anyhow::bail;
+use anyhow::Context;
+use anyhow::Result;
+use bitvec::prelude::*;
+use clap::Parser;
+use log::info;
+use log::trace;
+use log::warn;
+use ordered_float::OrderedFloat;
+
+/// Atropos is a multi-domain BPF / userspace hybrid scheduler where the BPF
+/// part does simple round robin in each domain and the userspace part
+/// calculates the load factor of each domain and tells the BPF part how to load
+/// balance the domains.
+///
+/// This scheduler demonstrates dividing scheduling logic between BPF and
+/// userspace and using rust to build the userspace part. An earlier variant of
+/// this scheduler was used to balance across six domains, each representing a
+/// chiplet in a six-chiplet AMD processor, and could match the performance of
+/// production setup using CFS.
+///
+/// WARNING: Atropos currently assumes that all domains have equal
+/// processing power and at similar distances from each other. This
+/// limitation will be removed in the future.
+#[derive(Debug, Parser)]
+struct Opts {
+    /// Scheduling slice duration in microseconds.
+    #[clap(short = 's', long, default_value = "20000")]
+    slice_us: u64,
+
+    /// Monitoring and load balance interval in seconds.
+    #[clap(short = 'i', long, default_value = "2.0")]
+    interval: f64,
+
+    /// Tuner runs at higher frequency than the load balancer to dynamically
+    /// tune scheduling behavior. Tuning interval in seconds.
+    #[clap(short = 'I', long, default_value = "0.1")]
+    tune_interval: f64,
+
+    /// Build domains according to how CPUs are grouped at this cache level
+    /// as determined by /sys/devices/system/cpu/cpuX/cache/indexI/id.
+    #[clap(short = 'c', long, default_value = "3")]
+    cache_level: u32,
+
+    /// Instead of using cache locality, set the cpumask for each domain
+    /// manually, provide multiple --cpumasks, one for each domain. E.g.
+    /// --cpumasks 0xff_00ff --cpumasks 0xff00 will create two domains with
+    /// the corresponding CPUs belonging to each domain. Each CPU must
+    /// belong to precisely one domain.
+    #[clap(short = 'C', long, num_args = 1.., conflicts_with = "cache_level")]
+    cpumasks: Vec<String>,
+
+    /// When non-zero, enable greedy task stealing. When a domain is idle, a
+    /// cpu will attempt to steal tasks from a domain with at least
+    /// greedy_threshold tasks enqueued. These tasks aren't permanently
+    /// stolen from the domain.
+    #[clap(short = 'g', long, default_value = "1")]
+    greedy_threshold: u32,
+
+    /// The load decay factor. Every interval, the existing load is decayed
+    /// by this factor and new load is added. Must be in the range [0.0,
+    /// 0.99]. The smaller the value, the more sensitive load calculation
+    /// is to recent changes. When 0.0, history is ignored and the load
+    /// value from the latest period is used directly.
+    #[clap(long, default_value = "0.5")]
+    load_decay_factor: f64,
+
+    /// Disable load balancing. Unless disabled, periodically userspace will
+    /// calculate the load factor of each domain and instruct BPF which
+    /// processes to move.
+    #[clap(long, action = clap::ArgAction::SetTrue)]
+    no_load_balance: bool,
+
+    /// Put per-cpu kthreads directly into local dsq's.
+    #[clap(short = 'k', long, action = clap::ArgAction::SetTrue)]
+    kthreads_local: bool,
+
+    /// In recent kernels (>=v6.6), the kernel is responsible for balancing
+    /// kworkers across L3 cache domains. Exclude them from load-balancing
+    /// to avoid conflicting operations. Greedy executions still apply.
+    #[clap(short = 'b', long, action = clap::ArgAction::SetTrue)]
+    balanced_kworkers: bool,
+
+    /// Use FIFO scheduling instead of weighted vtime scheduling.
+    #[clap(short = 'f', long, action = clap::ArgAction::SetTrue)]
+    fifo_sched: bool,
+
+    /// Idle CPUs with utilization lower than this will get remote tasks
+    /// directly pushed on them. 0 disables, 100 enables always.
+    #[clap(short = 'D', long, default_value = "90.0")]
+    direct_greedy_under: f64,
+
+    /// Idle CPUs with utilization lower than this may get kicked to
+    /// accelerate stealing when a task is queued on a saturated remote
+    /// domain. 0 disables, 100 enables always.
+    #[clap(short = 'K', long, default_value = "100.0")]
+    kick_greedy_under: f64,
+
+    /// If specified, only tasks which have their scheduling policy set to
+    /// SCHED_EXT using sched_setscheduler(2) are switched. Otherwise, all
+    /// tasks are switched.
+    #[clap(short = 'p', long, action = clap::ArgAction::SetTrue)]
+    partial: bool,
+
+    /// Enable verbose output including libbpf details. Specify multiple
+    /// times to increase verbosity.
+    #[clap(short = 'v', long, action = clap::ArgAction::Count)]
+    verbose: u8,
+}
+
+fn now_monotonic() -> u64 {
+    let mut time = libc::timespec {
+        tv_sec: 0,
+        tv_nsec: 0,
+    };
+    let ret = unsafe { libc::clock_gettime(libc::CLOCK_MONOTONIC, &mut time) };
+    assert!(ret == 0);
+    time.tv_sec as u64 * 1_000_000_000 + time.tv_nsec as u64
+}
+
+fn clear_map(map: &mut libbpf_rs::Map) {
+    // XXX: libbpf_rs has some design flaw that make it impossible to
+    // delete while iterating despite it being safe so we alias it here
+    let deleter: &mut libbpf_rs::Map = unsafe { &mut *(map as *mut _) };
+    for key in map.keys() {
+        let _ = deleter.delete(&key);
+    }
+}
+
+fn format_cpumask(cpumask: &[u64], nr_cpus: usize) -> String {
+    cpumask
+        .iter()
+        .take((nr_cpus + 64) / 64)
+        .rev()
+        .fold(String::new(), |acc, x| format!("{} {:016X}", acc, x))
+}
+
+// Neither procfs or fb_procfs can determine per-CPU utilization reliably
+// with CPU hot[un]plugs. Roll our own.
+//
+// https://github.com/eminence/procfs/issues/274
+// https://github.com/facebookincubator/below/issues/8190
+#[derive(Clone, Debug, Default)]
+struct MyCpuStat {
+    user: u64,
+    nice: u64,
+    system: u64,
+    idle: u64,
+    iowait: u64,
+    irq: u64,
+    softirq: u64,
+    steal: u64,
+}
+
+impl MyCpuStat {
+    fn busy_and_total(&self) -> (u64, u64) {
+        let busy = self.user + self.system + self.nice + self.irq + self.softirq + self.steal;
+        (busy, self.idle + busy + self.iowait)
+    }
+
+    fn calc_util(&self, prev: &MyCpuStat) -> f64 {
+        let (curr_busy, curr_total) = self.busy_and_total();
+        let (prev_busy, prev_total) = prev.busy_and_total();
+        let busy = curr_busy - prev_busy;
+        let total = curr_total - prev_total;
+        if total > 0 {
+            ((busy as f64) / (total as f64)).clamp(0.0, 1.0)
+        } else {
+            1.0
+        }
+    }
+}
+
+#[derive(Clone, Debug, Default)]
+struct MyProcStat {
+    total: MyCpuStat,
+    cpus: BTreeMap<usize, MyCpuStat>,
+}
+
+impl MyProcStat {
+    fn read() -> Result<Self> {
+        let mut result: MyProcStat = Default::default();
+        for line in std::fs::read_to_string("/proc/stat")?.lines() {
+            let mut toks = line.split_whitespace();
+
+            let key = toks.next().ok_or(anyhow!("no key"))?;
+            if !key.starts_with("cpu") {
+                break;
+            }
+
+            let cputime = MyCpuStat {
+                user: toks.next().ok_or(anyhow!("missing"))?.parse::<u64>()?,
+                nice: toks.next().ok_or(anyhow!("missing"))?.parse::<u64>()?,
+                system: toks.next().ok_or(anyhow!("missing"))?.parse::<u64>()?,
+                idle: toks.next().ok_or(anyhow!("missing"))?.parse::<u64>()?,
+                iowait: toks.next().ok_or(anyhow!("missing"))?.parse::<u64>()?,
+                irq: toks.next().ok_or(anyhow!("missing"))?.parse::<u64>()?,
+                softirq: toks.next().ok_or(anyhow!("missing"))?.parse::<u64>()?,
+                steal: toks.next().ok_or(anyhow!("missing"))?.parse::<u64>()?,
+            };
+
+            if key.len() == 3 {
+                result.total = cputime;
+            } else {
+                result.cpus.insert(key[3..].parse::<usize>()?, cputime);
+            }
+        }
+        Ok(result)
+    }
+}
+
+#[derive(Debug)]
+struct Topology {
+    nr_cpus: usize,
+    nr_doms: usize,
+    dom_cpus: Vec<BitVec<u64, Lsb0>>,
+    cpu_dom: Vec<Option<usize>>,
+}
+
+impl Topology {
+    fn from_cpumasks(cpumasks: &[String], nr_cpus: usize) -> Result<Self> {
+        if cpumasks.len() > atropos_sys::MAX_DOMS as usize {
+            bail!(
+                "Number of requested domains ({}) is greater than MAX_DOMS ({})",
+                cpumasks.len(),
+                atropos_sys::MAX_DOMS
+            );
+        }
+        let mut cpu_dom = vec![None; nr_cpus];
+        let mut dom_cpus =
+            vec![bitvec![u64, Lsb0; 0; atropos_sys::MAX_CPUS as usize]; cpumasks.len()];
+        for (dom, cpumask) in cpumasks.iter().enumerate() {
+            let hex_str = {
+                let mut tmp_str = cpumask
+                    .strip_prefix("0x")
+                    .unwrap_or(cpumask)
+                    .replace('_', "");
+                if tmp_str.len() % 2 != 0 {
+                    tmp_str = "0".to_string() + &tmp_str;
+                }
+                tmp_str
+            };
+            let byte_vec = hex::decode(&hex_str)
+                .with_context(|| format!("Failed to parse cpumask: {}", cpumask))?;
+
+            for (index, &val) in byte_vec.iter().rev().enumerate() {
+                let mut v = val;
+                while v != 0 {
+                    let lsb = v.trailing_zeros() as usize;
+                    v &= !(1 << lsb);
+                    let cpu = index * 8 + lsb;
+                    if cpu > nr_cpus {
+                        bail!(
+                            concat!(
+                                "Found cpu ({}) in cpumask ({}) which is larger",
+                                " than the number of cpus on the machine ({})"
+                            ),
+                            cpu,
+                            cpumask,
+                            nr_cpus
+                        );
+                    }
+                    if let Some(other_dom) = cpu_dom[cpu] {
+                        bail!(
+                            "Found cpu ({}) with domain ({}) but also in cpumask ({})",
+                            cpu,
+                            other_dom,
+                            cpumask
+                        );
+                    }
+                    cpu_dom[cpu] = Some(dom);
+                    dom_cpus[dom].set(cpu, true);
+                }
+            }
+            dom_cpus[dom].set_uninitialized(false);
+        }
+
+        for (cpu, dom) in cpu_dom.iter().enumerate() {
+            if dom.is_none() {
+                bail!(
+                    "CPU {} not assigned to any domain. Make sure it is covered by some --cpumasks argument.",
+                    cpu
+                );
+            }
+        }
+
+        Ok(Self {
+            nr_cpus,
+            nr_doms: dom_cpus.len(),
+            dom_cpus,
+            cpu_dom,
+        })
+    }
+
+    fn from_cache_level(level: u32, nr_cpus: usize) -> Result<Self> {
+        let mut cpu_to_cache = vec![]; // (cpu_id, Option<cache_id>)
+        let mut cache_ids = BTreeSet::<usize>::new();
+        let mut nr_offline = 0;
+
+        // Build cpu -> cache ID mapping.
+        for cpu in 0..nr_cpus {
+            let path = format!("/sys/devices/system/cpu/cpu{}/cache/index{}/id", cpu, level);
+            let id = match std::fs::read_to_string(&path) {
+                Ok(val) => Some(val.trim().parse::<usize>().with_context(|| {
+                    format!("Failed to parse {:?}'s content {:?}", &path, &val)
+                })?),
+                Err(e) if e.kind() == std::io::ErrorKind::NotFound => {
+                    nr_offline += 1;
+                    None
+                }
+                Err(e) => return Err(e).with_context(|| format!("Failed to open {:?}", &path)),
+            };
+
+            cpu_to_cache.push(id);
+            if id.is_some() {
+                cache_ids.insert(id.unwrap());
+            }
+        }
+
+        info!(
+            "CPUs: online/possible = {}/{}",
+            nr_cpus - nr_offline,
+            nr_cpus
+        );
+
+        // Cache IDs may have holes. Assign consecutive domain IDs to
+        // existing cache IDs.
+        let mut cache_to_dom = BTreeMap::<usize, usize>::new();
+        let mut nr_doms = 0;
+        for cache_id in cache_ids.iter() {
+            cache_to_dom.insert(*cache_id, nr_doms);
+            nr_doms += 1;
+        }
+
+        if nr_doms > atropos_sys::MAX_DOMS as usize {
+            bail!(
+                "Total number of doms {} is greater than MAX_DOMS ({})",
+                nr_doms,
+                atropos_sys::MAX_DOMS
+            );
+        }
+
+        // Build and return dom -> cpumask and cpu -> dom mappings.
+        let mut dom_cpus =
+            vec![bitvec![u64, Lsb0; 0; atropos_sys::MAX_CPUS as usize]; nr_doms as usize];
+        let mut cpu_dom = vec![];
+
+        for cpu in 0..nr_cpus {
+            match cpu_to_cache[cpu] {
+                Some(cache_id) => {
+                    let dom_id = cache_to_dom[&cache_id];
+                    dom_cpus[dom_id].set(cpu, true);
+                    cpu_dom.push(Some(dom_id));
+                }
+                None => {
+                    dom_cpus[0].set(cpu, true);
+                    cpu_dom.push(None);
+                }
+            }
+        }
+
+        Ok(Self {
+            nr_cpus,
+            nr_doms: dom_cpus.len(),
+            dom_cpus,
+            cpu_dom,
+        })
+    }
+}
+
+struct Tuner {
+    top: Arc<Topology>,
+    direct_greedy_under: f64,
+    kick_greedy_under: f64,
+    prev_cpu_stats: BTreeMap<usize, MyCpuStat>,
+    dom_utils: Vec<f64>,
+}
+
+impl Tuner {
+    fn new(top: Arc<Topology>, opts: &Opts) -> Result<Self> {
+        Ok(Self {
+            direct_greedy_under: opts.direct_greedy_under / 100.0,
+            kick_greedy_under: opts.kick_greedy_under / 100.0,
+            prev_cpu_stats: MyProcStat::read()?.cpus,
+            dom_utils: vec![0.0; top.nr_doms],
+            top,
+        })
+    }
+
+    fn step(&mut self, skel: &mut AtroposSkel) -> Result<()> {
+        let curr_cpu_stats = MyProcStat::read()?.cpus;
+        let ti = &mut skel.bss().tune_input;
+        let mut dom_nr_cpus = vec![0; self.top.nr_doms];
+        let mut dom_util_sum = vec![0.0; self.top.nr_doms];
+
+        for cpu in 0..self.top.nr_cpus {
+            // None domain indicates the CPU was offline during
+            // initialization and None MyCpuStat indicates the CPU has gone
+            // down since then. Ignore both.
+            if let (Some(dom), Some(curr), Some(prev)) = (
+                self.top.cpu_dom[cpu],
+                curr_cpu_stats.get(&cpu),
+                self.prev_cpu_stats.get(&cpu),
+            ) {
+                dom_nr_cpus[dom] += 1;
+                dom_util_sum[dom] += curr.calc_util(prev);
+            }
+        }
+
+        for dom in 0..self.top.nr_doms {
+            // Calculate the domain avg util. If there are no active CPUs,
+            // it doesn't really matter. Go with 0.0 as that's less likely
+            // to confuse users.
+            let util = match dom_nr_cpus[dom] {
+                0 => 0.0,
+                nr => dom_util_sum[dom] / nr as f64,
+            };
+
+            self.dom_utils[dom] = util;
+
+            // This could be implemented better.
+            let update_dom_bits = |target: &mut [u64; 8], val: bool| {
+                for cpu in 0..self.top.nr_cpus {
+                    if let Some(cdom) = self.top.cpu_dom[cpu] {
+                        if cdom == dom {
+                            if val {
+                                target[cpu / 64] |= 1u64 << (cpu % 64);
+                            } else {
+                                target[cpu / 64] &= !(1u64 << (cpu % 64));
+                            }
+                        }
+                    }
+                }
+            };
+
+            update_dom_bits(
+                &mut ti.direct_greedy_cpumask,
+                self.direct_greedy_under > 0.99999 || util < self.direct_greedy_under,
+            );
+            update_dom_bits(
+                &mut ti.kick_greedy_cpumask,
+                self.kick_greedy_under > 0.99999 || util < self.kick_greedy_under,
+            );
+        }
+
+        ti.gen += 1;
+        self.prev_cpu_stats = curr_cpu_stats;
+        Ok(())
+    }
+}
+
+#[derive(Debug)]
+struct TaskLoad {
+    runnable_for: u64,
+    load: f64,
+}
+
+#[derive(Debug)]
+struct TaskInfo {
+    pid: i32,
+    dom_mask: u64,
+    migrated: Cell<bool>,
+    is_kworker: bool,
+}
+
+struct LoadBalancer<'a, 'b, 'c> {
+    maps: AtroposMapsMut<'a>,
+    top: Arc<Topology>,
+    task_loads: &'b mut BTreeMap<i32, TaskLoad>,
+    load_decay_factor: f64,
+    skip_kworkers: bool,
+
+    tasks_by_load: Vec<BTreeMap<OrderedFloat<f64>, TaskInfo>>,
+    load_avg: f64,
+    dom_loads: Vec<f64>,
+
+    imbal: Vec<f64>,
+    doms_to_push: BTreeMap<OrderedFloat<f64>, u32>,
+    doms_to_pull: BTreeMap<OrderedFloat<f64>, u32>,
+
+    nr_lb_data_errors: &'c mut u64,
+}
+
+impl<'a, 'b, 'c> LoadBalancer<'a, 'b, 'c> {
+    // If imbalance gets higher than this ratio, try to balance the loads.
+    const LOAD_IMBAL_HIGH_RATIO: f64 = 0.10;
+
+    // Aim to transfer this fraction of the imbalance on each round. We want
+    // to be gradual to avoid unnecessary oscillations. While this can delay
+    // convergence, greedy execution should be able to bridge the temporary
+    // gap.
+    const LOAD_IMBAL_XFER_TARGET_RATIO: f64 = 0.50;
+
+    // Don't push out more than this ratio of load on each round. While this
+    // overlaps with XFER_TARGET_RATIO, XFER_TARGET_RATIO only defines the
+    // target and doesn't limit the total load. As long as the transfer
+    // reduces load imbalance between the two involved domains, it'd happily
+    // transfer whatever amount that can be transferred. This limit is used
+    // as the safety cap to avoid draining a given domain too much in a
+    // single round.
+    const LOAD_IMBAL_PUSH_MAX_RATIO: f64 = 0.50;
+
+    fn new(
+        maps: AtroposMapsMut<'a>,
+        top: Arc<Topology>,
+        task_loads: &'b mut BTreeMap<i32, TaskLoad>,
+        load_decay_factor: f64,
+        skip_kworkers: bool,
+        nr_lb_data_errors: &'c mut u64,
+    ) -> Self {
+        Self {
+            maps,
+            task_loads,
+            load_decay_factor,
+            skip_kworkers,
+
+            tasks_by_load: (0..top.nr_doms).map(|_| BTreeMap::<_, _>::new()).collect(),
+            load_avg: 0f64,
+            dom_loads: vec![0.0; top.nr_doms],
+
+            imbal: vec![0.0; top.nr_doms],
+            doms_to_pull: BTreeMap::new(),
+            doms_to_push: BTreeMap::new(),
+
+            nr_lb_data_errors,
+
+            top,
+        }
+    }
+
+    fn read_task_loads(&mut self, period: Duration) -> Result<()> {
+        let now_mono = now_monotonic();
+        let task_data = self.maps.task_data();
+        let mut this_task_loads = BTreeMap::<i32, TaskLoad>::new();
+        let mut load_sum = 0.0f64;
+        self.dom_loads = vec![0f64; self.top.nr_doms];
+
+        for key in task_data.keys() {
+            if let Some(task_ctx_vec) = task_data
+                .lookup(&key, libbpf_rs::MapFlags::ANY)
+                .context("Failed to lookup task_data")?
+            {
+                let task_ctx =
+                    unsafe { &*(task_ctx_vec.as_slice().as_ptr() as *const atropos_sys::task_ctx) };
+                let pid = i32::from_ne_bytes(
+                    key.as_slice()
+                        .try_into()
+                        .context("Invalid key length in task_data map")?,
+                );
+
+                let (this_at, this_for, weight) = unsafe {
+                    (
+                        std::ptr::read_volatile(&task_ctx.runnable_at as *const u64),
+                        std::ptr::read_volatile(&task_ctx.runnable_for as *const u64),
+                        std::ptr::read_volatile(&task_ctx.weight as *const u32),
+                    )
+                };
+
+                let (mut delta, prev_load) = match self.task_loads.get(&pid) {
+                    Some(prev) => (this_for - prev.runnable_for, Some(prev.load)),
+                    None => (this_for, None),
+                };
+
+                // Non-zero this_at indicates that the task is currently
+                // runnable. Note that we read runnable_at and runnable_for
+                // without any synchronization and there is a small window
+                // where we end up misaccounting. While this can cause
+                // temporary error, it's unlikely to cause any noticeable
+                // misbehavior especially given the load value clamping.
+                if this_at > 0 && this_at < now_mono {
+                    delta += now_mono - this_at;
+                }
+
+                delta = delta.min(period.as_nanos() as u64);
+                let this_load = (weight as f64 * delta as f64 / period.as_nanos() as f64)
+                    .clamp(0.0, weight as f64);
+
+                let this_load = match prev_load {
+                    Some(prev_load) => {
+                        prev_load * self.load_decay_factor
+                            + this_load * (1.0 - self.load_decay_factor)
+                    }
+                    None => this_load,
+                };
+
+                this_task_loads.insert(
+                    pid,
+                    TaskLoad {
+                        runnable_for: this_for,
+                        load: this_load,
+                    },
+                );
+
+                load_sum += this_load;
+                self.dom_loads[task_ctx.dom_id as usize] += this_load;
+                // Only record pids that are eligible for load balancing
+                if task_ctx.dom_mask == (1u64 << task_ctx.dom_id) {
+                    continue;
+                }
+                self.tasks_by_load[task_ctx.dom_id as usize].insert(
+                    OrderedFloat(this_load),
+                    TaskInfo {
+                        pid,
+                        dom_mask: task_ctx.dom_mask,
+                        migrated: Cell::new(false),
+                        is_kworker: task_ctx.is_kworker,
+                    },
+                );
+            }
+        }
+
+        self.load_avg = load_sum / self.top.nr_doms as f64;
+        *self.task_loads = this_task_loads;
+        Ok(())
+    }
+
+    // To balance dom loads we identify doms with lower and higher load than average
+    fn calculate_dom_load_balance(&mut self) -> Result<()> {
+        for (dom, dom_load) in self.dom_loads.iter().enumerate() {
+            let imbal = dom_load - self.load_avg;
+            if imbal.abs() >= self.load_avg * Self::LOAD_IMBAL_HIGH_RATIO {
+                if imbal > 0f64 {
+                    self.doms_to_push.insert(OrderedFloat(imbal), dom as u32);
+                } else {
+                    self.doms_to_pull.insert(OrderedFloat(-imbal), dom as u32);
+                }
+                self.imbal[dom] = imbal;
+            }
+        }
+        Ok(())
+    }
+
+    // Find the first candidate pid which hasn't already been migrated and
+    // can run in @pull_dom.
+    fn find_first_candidate<'d, I>(
+        tasks_by_load: I,
+        pull_dom: u32,
+        skip_kworkers: bool,
+    ) -> Option<(f64, &'d TaskInfo)>
+    where
+        I: IntoIterator<Item = (&'d OrderedFloat<f64>, &'d TaskInfo)>,
+    {
+        match tasks_by_load
+            .into_iter()
+            .skip_while(|(_, task)| {
+                task.migrated.get()
+                    || (task.dom_mask & (1 << pull_dom) == 0)
+                    || (skip_kworkers && task.is_kworker)
+            })
+            .next()
+        {
+            Some((OrderedFloat(load), task)) => Some((*load, task)),
+            None => None,
+        }
+    }
+
+    fn pick_victim(
+        &self,
+        (push_dom, to_push): (u32, f64),
+        (pull_dom, to_pull): (u32, f64),
+    ) -> Option<(&TaskInfo, f64)> {
+        let to_xfer = to_pull.min(to_push) * Self::LOAD_IMBAL_XFER_TARGET_RATIO;
+
+        trace!(
+            "considering dom {}@{:.2} -> {}@{:.2}",
+            push_dom,
+            to_push,
+            pull_dom,
+            to_pull
+        );
+
+        let calc_new_imbal = |xfer: f64| (to_push - xfer).abs() + (to_pull - xfer).abs();
+
+        trace!(
+            "to_xfer={:.2} tasks_by_load={:?}",
+            to_xfer,
+            &self.tasks_by_load[push_dom as usize]
+        );
+
+        // We want to pick a task to transfer from push_dom to pull_dom to
+        // reduce the load imbalance between the two closest to $to_xfer.
+        // IOW, pick a task which has the closest load value to $to_xfer
+        // that can be migrated. Find such task by locating the first
+        // migratable task while scanning left from $to_xfer and the
+        // counterpart while scanning right and picking the better of the
+        // two.
+        let (load, task, new_imbal) = match (
+            Self::find_first_candidate(
+                self.tasks_by_load[push_dom as usize]
+                    .range((Unbounded, Included(&OrderedFloat(to_xfer))))
+                    .rev(),
+                pull_dom,
+                self.skip_kworkers,
+            ),
+            Self::find_first_candidate(
+                self.tasks_by_load[push_dom as usize]
+                    .range((Included(&OrderedFloat(to_xfer)), Unbounded)),
+                pull_dom,
+                self.skip_kworkers,
+            ),
+        ) {
+            (None, None) => return None,
+            (Some((load, task)), None) | (None, Some((load, task))) => {
+                (load, task, calc_new_imbal(load))
+            }
+            (Some((load0, task0)), Some((load1, task1))) => {
+                let (new_imbal0, new_imbal1) = (calc_new_imbal(load0), calc_new_imbal(load1));
+                if new_imbal0 <= new_imbal1 {
+                    (load0, task0, new_imbal0)
+                } else {
+                    (load1, task1, new_imbal1)
+                }
+            }
+        };
+
+        // If the best candidate can't reduce the imbalance, there's nothing
+        // to do for this pair.
+        let old_imbal = to_push + to_pull;
+        if old_imbal < new_imbal {
+            trace!(
+                "skipping pid {}, dom {} -> {} won't improve imbal {:.2} -> {:.2}",
+                task.pid,
+                push_dom,
+                pull_dom,
+                old_imbal,
+                new_imbal
+            );
+            return None;
+        }
+
+        trace!(
+            "migrating pid {}, dom {} -> {}, imbal={:.2} -> {:.2}",
+            task.pid,
+            push_dom,
+            pull_dom,
+            old_imbal,
+            new_imbal,
+        );
+
+        Some((task, load))
+    }
+
+    // Actually execute the load balancing. Concretely this writes pid -> dom
+    // entries into the lb_data map for bpf side to consume.
+    fn load_balance(&mut self) -> Result<()> {
+        clear_map(self.maps.lb_data());
+
+        trace!("imbal={:?}", &self.imbal);
+        trace!("doms_to_push={:?}", &self.doms_to_push);
+        trace!("doms_to_pull={:?}", &self.doms_to_pull);
+
+        // Push from the most imbalanced to least.
+        while let Some((OrderedFloat(mut to_push), push_dom)) = self.doms_to_push.pop_last() {
+            let push_max = self.dom_loads[push_dom as usize] * Self::LOAD_IMBAL_PUSH_MAX_RATIO;
+            let mut pushed = 0f64;
+
+            // Transfer tasks from push_dom to reduce imbalance.
+            loop {
+                let last_pushed = pushed;
+
+                // Pull from the most imbalaned to least.
+                let mut doms_to_pull = BTreeMap::<_, _>::new();
+                std::mem::swap(&mut self.doms_to_pull, &mut doms_to_pull);
+                let mut pull_doms = doms_to_pull.into_iter().rev().collect::<Vec<(_, _)>>();
+
+                for (to_pull, pull_dom) in pull_doms.iter_mut() {
+                    if let Some((task, load)) =
+                        self.pick_victim((push_dom, to_push), (*pull_dom, f64::from(*to_pull)))
+                    {
+                        // Execute migration.
+                        task.migrated.set(true);
+                        to_push -= load;
+                        *to_pull -= load;
+                        pushed += load;
+
+                        // Ask BPF code to execute the migration.
+                        let pid = task.pid;
+                        let cpid = (pid as libc::pid_t).to_ne_bytes();
+                        if let Err(e) = self.maps.lb_data().update(
+                            &cpid,
+                            &pull_dom.to_ne_bytes(),
+                            libbpf_rs::MapFlags::NO_EXIST,
+                        ) {
+                            warn!(
+                                "Failed to update lb_data map for pid={} error={:?}",
+                                pid, &e
+                            );
+                            *self.nr_lb_data_errors += 1;
+                        }
+
+                        // Always break after a successful migration so that
+                        // the pulling domains are always considered in the
+                        // descending imbalance order.
+                        break;
+                    }
+                }
+
+                pull_doms
+                    .into_iter()
+                    .map(|(k, v)| self.doms_to_pull.insert(k, v))
+                    .count();
+
+                // Stop repeating if nothing got transferred or pushed enough.
+                if pushed == last_pushed || pushed >= push_max {
+                    break;
+                }
+            }
+        }
+        Ok(())
+    }
+}
+
+struct Scheduler<'a> {
+    skel: AtroposSkel<'a>,
+    struct_ops: Option<libbpf_rs::Link>,
+
+    sched_interval: Duration,
+    tune_interval: Duration,
+    load_decay_factor: f64,
+    balance_load: bool,
+    balanced_kworkers: bool,
+
+    top: Arc<Topology>,
+
+    prev_at: Instant,
+    prev_total_cpu: MyCpuStat,
+    task_loads: BTreeMap<i32, TaskLoad>,
+
+    nr_lb_data_errors: u64,
+
+    tuner: Tuner,
+}
+
+impl<'a> Scheduler<'a> {
+    fn init(opts: &Opts) -> Result<Self> {
+        // Open the BPF prog first for verification.
+        let mut skel_builder = AtroposSkelBuilder::default();
+        skel_builder.obj_builder.debug(opts.verbose > 0);
+        let mut skel = skel_builder.open().context("Failed to open BPF program")?;
+
+        let nr_cpus = libbpf_rs::num_possible_cpus().unwrap();
+        if nr_cpus > atropos_sys::MAX_CPUS as usize {
+            bail!(
+                "nr_cpus ({}) is greater than MAX_CPUS ({})",
+                nr_cpus,
+                atropos_sys::MAX_CPUS
+            );
+        }
+
+        // Initialize skel according to @opts.
+        let top = Arc::new(if opts.cpumasks.len() > 0 {
+            Topology::from_cpumasks(&opts.cpumasks, nr_cpus)?
+        } else {
+            Topology::from_cache_level(opts.cache_level, nr_cpus)?
+        });
+
+        skel.rodata().nr_doms = top.nr_doms as u32;
+        skel.rodata().nr_cpus = top.nr_cpus as u32;
+
+        for (cpu, dom) in top.cpu_dom.iter().enumerate() {
+            skel.rodata().cpu_dom_id_map[cpu] = dom.unwrap_or(0) as u32;
+        }
+
+        for (dom, cpus) in top.dom_cpus.iter().enumerate() {
+            let raw_cpus_slice = cpus.as_raw_slice();
+            let dom_cpumask_slice = &mut skel.rodata().dom_cpumasks[dom];
+            let (left, _) = dom_cpumask_slice.split_at_mut(raw_cpus_slice.len());
+            left.clone_from_slice(cpus.as_raw_slice());
+            info!(
+                "DOM[{:02}] cpumask{} ({} cpus)",
+                dom,
+                &format_cpumask(dom_cpumask_slice, nr_cpus),
+                cpus.count_ones()
+            );
+        }
+
+        skel.rodata().slice_ns = opts.slice_us * 1000;
+        skel.rodata().kthreads_local = opts.kthreads_local;
+        skel.rodata().fifo_sched = opts.fifo_sched;
+        skel.rodata().switch_partial = opts.partial;
+        skel.rodata().greedy_threshold = opts.greedy_threshold;
+
+        // Attach.
+        let mut skel = skel.load().context("Failed to load BPF program")?;
+        skel.attach().context("Failed to attach BPF program")?;
+        let struct_ops = Some(
+            skel.maps_mut()
+                .atropos()
+                .attach_struct_ops()
+                .context("Failed to attach atropos struct ops")?,
+        );
+        info!("Atropos Scheduler Attached");
+
+        // Other stuff.
+        let prev_total_cpu = MyProcStat::read()?.total;
+
+        Ok(Self {
+            skel,
+            struct_ops, // should be held to keep it attached
+
+            sched_interval: Duration::from_secs_f64(opts.interval),
+            tune_interval: Duration::from_secs_f64(opts.tune_interval),
+            load_decay_factor: opts.load_decay_factor.clamp(0.0, 0.99),
+            balance_load: !opts.no_load_balance,
+            balanced_kworkers: opts.balanced_kworkers,
+
+            top: top.clone(),
+
+            prev_at: Instant::now(),
+            prev_total_cpu,
+            task_loads: BTreeMap::new(),
+
+            nr_lb_data_errors: 0,
+
+            tuner: Tuner::new(top, opts)?,
+        })
+    }
+
+    fn get_cpu_busy(&mut self) -> Result<f64> {
+        let total_cpu = MyProcStat::read()?.total;
+        let busy = total_cpu.calc_util(&self.prev_total_cpu);
+        self.prev_total_cpu = total_cpu;
+        Ok(busy)
+    }
+
+    fn read_bpf_stats(&mut self) -> Result<Vec<u64>> {
+        let mut maps = self.skel.maps_mut();
+        let stats_map = maps.stats();
+        let mut stats: Vec<u64> = Vec::new();
+        let zero_vec = vec![vec![0u8; stats_map.value_size() as usize]; self.top.nr_cpus];
+
+        for stat in 0..atropos_sys::stat_idx_ATROPOS_NR_STATS {
+            let cpu_stat_vec = stats_map
+                .lookup_percpu(&(stat as u32).to_ne_bytes(), libbpf_rs::MapFlags::ANY)
+                .with_context(|| format!("Failed to lookup stat {}", stat))?
+                .expect("per-cpu stat should exist");
+            let sum = cpu_stat_vec
+                .iter()
+                .map(|val| {
+                    u64::from_ne_bytes(
+                        val.as_slice()
+                            .try_into()
+                            .expect("Invalid value length in stat map"),
+                    )
+                })
+                .sum();
+            stats_map
+                .update_percpu(
+                    &(stat as u32).to_ne_bytes(),
+                    &zero_vec,
+                    libbpf_rs::MapFlags::ANY,
+                )
+                .context("Failed to zero stat")?;
+            stats.push(sum);
+        }
+        Ok(stats)
+    }
+
+    fn report(
+        &mut self,
+        stats: &Vec<u64>,
+        cpu_busy: f64,
+        processing_dur: Duration,
+        load_avg: f64,
+        dom_loads: &Vec<f64>,
+        imbal: &Vec<f64>,
+    ) {
+        let stat = |idx| stats[idx as usize];
+        let total = stat(atropos_sys::stat_idx_ATROPOS_STAT_WAKE_SYNC)
+            + stat(atropos_sys::stat_idx_ATROPOS_STAT_PREV_IDLE)
+            + stat(atropos_sys::stat_idx_ATROPOS_STAT_GREEDY_IDLE)
+            + stat(atropos_sys::stat_idx_ATROPOS_STAT_PINNED)
+            + stat(atropos_sys::stat_idx_ATROPOS_STAT_DIRECT_DISPATCH)
+            + stat(atropos_sys::stat_idx_ATROPOS_STAT_DIRECT_GREEDY)
+            + stat(atropos_sys::stat_idx_ATROPOS_STAT_DIRECT_GREEDY_FAR)
+            + stat(atropos_sys::stat_idx_ATROPOS_STAT_DSQ_DISPATCH)
+            + stat(atropos_sys::stat_idx_ATROPOS_STAT_GREEDY);
+
+        info!(
+            "cpu={:7.2} bal={} load_avg={:8.2} task_err={} lb_data_err={} proc={:?}ms",
+            cpu_busy * 100.0,
+            stats[atropos_sys::stat_idx_ATROPOS_STAT_LOAD_BALANCE as usize],
+            load_avg,
+            stats[atropos_sys::stat_idx_ATROPOS_STAT_TASK_GET_ERR as usize],
+            self.nr_lb_data_errors,
+            processing_dur.as_millis(),
+        );
+
+        let stat_pct = |idx| stat(idx) as f64 / total as f64 * 100.0;
+
+        info!(
+            "tot={:7} wsync={:5.2} prev_idle={:5.2} greedy_idle={:5.2} pin={:5.2}",
+            total,
+            stat_pct(atropos_sys::stat_idx_ATROPOS_STAT_WAKE_SYNC),
+            stat_pct(atropos_sys::stat_idx_ATROPOS_STAT_PREV_IDLE),
+            stat_pct(atropos_sys::stat_idx_ATROPOS_STAT_GREEDY_IDLE),
+            stat_pct(atropos_sys::stat_idx_ATROPOS_STAT_PINNED),
+        );
+
+        info!(
+            "dir={:5.2} dir_greedy={:5.2} dir_greedy_far={:5.2}",
+            stat_pct(atropos_sys::stat_idx_ATROPOS_STAT_DIRECT_DISPATCH),
+            stat_pct(atropos_sys::stat_idx_ATROPOS_STAT_DIRECT_GREEDY),
+            stat_pct(atropos_sys::stat_idx_ATROPOS_STAT_DIRECT_GREEDY_FAR),
+        );
+
+        info!(
+            "dsq={:5.2} greedy={:5.2} kick_greedy={:5.2} rep={:5.2}",
+            stat_pct(atropos_sys::stat_idx_ATROPOS_STAT_DSQ_DISPATCH),
+            stat_pct(atropos_sys::stat_idx_ATROPOS_STAT_GREEDY),
+            stat_pct(atropos_sys::stat_idx_ATROPOS_STAT_KICK_GREEDY),
+            stat_pct(atropos_sys::stat_idx_ATROPOS_STAT_REPATRIATE),
+        );
+
+        let ti = &self.skel.bss().tune_input;
+        info!(
+            "direct_greedy_cpumask={}",
+            format_cpumask(&ti.direct_greedy_cpumask, self.top.nr_cpus)
+        );
+        info!(
+            "  kick_greedy_cpumask={}",
+            format_cpumask(&ti.kick_greedy_cpumask, self.top.nr_cpus)
+        );
+
+        for i in 0..self.top.nr_doms {
+            info!(
+                "DOM[{:02}] util={:6.2} load={:8.2} imbal={}",
+                i,
+                self.tuner.dom_utils[i] * 100.0,
+                dom_loads[i],
+                if imbal[i] == 0.0 {
+                    format!("{:9.2}", 0.0)
+                } else {
+                    format!("{:+9.2}", imbal[i])
+                },
+            );
+        }
+    }
+
+    fn lb_step(&mut self) -> Result<()> {
+        let started_at = Instant::now();
+        let bpf_stats = self.read_bpf_stats()?;
+        let cpu_busy = self.get_cpu_busy()?;
+
+        let mut lb = LoadBalancer::new(
+            self.skel.maps_mut(),
+            self.top.clone(),
+            &mut self.task_loads,
+            self.load_decay_factor,
+            self.balanced_kworkers,
+            &mut self.nr_lb_data_errors,
+        );
+
+        lb.read_task_loads(started_at.duration_since(self.prev_at))?;
+        lb.calculate_dom_load_balance()?;
+
+        if self.balance_load {
+            lb.load_balance()?;
+        }
+
+        // Extract fields needed for reporting and drop lb to release
+        // mutable borrows.
+        let (load_avg, dom_loads, imbal) = (lb.load_avg, lb.dom_loads, lb.imbal);
+
+        self.report(
+            &bpf_stats,
+            cpu_busy,
+            Instant::now().duration_since(started_at),
+            load_avg,
+            &dom_loads,
+            &imbal,
+        );
+
+        self.prev_at = started_at;
+        Ok(())
+    }
+
+    fn read_bpf_exit_type(&mut self) -> i32 {
+        unsafe { std::ptr::read_volatile(&self.skel.bss().exit_type as *const _) }
+    }
+
+    fn report_bpf_exit_type(&mut self) -> Result<()> {
+        // Report msg if EXT_OPS_EXIT_ERROR.
+        match self.read_bpf_exit_type() {
+            0 => Ok(()),
+            etype if etype == 2 => {
+                let cstr = unsafe { CStr::from_ptr(self.skel.bss().exit_msg.as_ptr() as *const _) };
+                let msg = cstr
+                    .to_str()
+                    .context("Failed to convert exit msg to string")
+                    .unwrap();
+                bail!("BPF exit_type={} msg={}", etype, msg);
+            }
+            etype => {
+                info!("BPF exit_type={}", etype);
+                Ok(())
+            }
+        }
+    }
+
+    fn run(&mut self, shutdown: Arc<AtomicBool>) -> Result<()> {
+        let now = Instant::now();
+        let mut next_tune_at = now + self.tune_interval;
+        let mut next_sched_at = now + self.sched_interval;
+
+        while !shutdown.load(Ordering::Relaxed) && self.read_bpf_exit_type() == 0 {
+            let now = Instant::now();
+
+            if now >= next_tune_at {
+                self.tuner.step(&mut self.skel)?;
+                next_tune_at += self.tune_interval;
+                if next_tune_at < now {
+                    next_tune_at = now + self.tune_interval;
+                }
+            }
+
+            if now >= next_sched_at {
+                self.lb_step()?;
+                next_sched_at += self.sched_interval;
+                if next_sched_at < now {
+                    next_sched_at = now + self.sched_interval;
+                }
+            }
+
+            std::thread::sleep(
+                next_sched_at
+                    .min(next_tune_at)
+                    .duration_since(Instant::now()),
+            );
+        }
+
+        self.report_bpf_exit_type()
+    }
+}
+
+impl<'a> Drop for Scheduler<'a> {
+    fn drop(&mut self) {
+        if let Some(struct_ops) = self.struct_ops.take() {
+            drop(struct_ops);
+        }
+    }
+}
+
+fn main() -> Result<()> {
+    let opts = Opts::parse();
+
+    let llv = match opts.verbose {
+        0 => simplelog::LevelFilter::Info,
+        1 => simplelog::LevelFilter::Debug,
+        _ => simplelog::LevelFilter::Trace,
+    };
+    let mut lcfg = simplelog::ConfigBuilder::new();
+    lcfg.set_time_level(simplelog::LevelFilter::Error)
+        .set_location_level(simplelog::LevelFilter::Off)
+        .set_target_level(simplelog::LevelFilter::Off)
+        .set_thread_level(simplelog::LevelFilter::Off);
+    simplelog::TermLogger::init(
+        llv,
+        lcfg.build(),
+        simplelog::TerminalMode::Stderr,
+        simplelog::ColorChoice::Auto,
+    )?;
+
+    let mut sched = Scheduler::init(&opts)?;
+
+    let shutdown = Arc::new(AtomicBool::new(false));
+    let shutdown_clone = shutdown.clone();
+    ctrlc::set_handler(move || {
+        shutdown_clone.store(true, Ordering::Relaxed);
+    })
+    .context("Error setting Ctrl-C handler")?;
+
+    sched.run(shutdown)
+}
-- 
2.41.0

Re: [PATCH 12/34] sched_ext: Implement BPF extensible scheduler class

[Andrea Righi]

On Mon, Jul 10, 2023 at 03:13:30PM -1000, Tejun Heo wrote:
...
> +static void free_dsq_irq_workfn(struct irq_work *irq_work)
> +{
> +	struct llist_node *to_free = llist_del_all(&dsqs_to_free);
> +	struct scx_dispatch_q *dsq, *tmp_dsq;
> +
> +	llist_for_each_entry_safe(dsq, tmp_dsq, to_free, free_node)
> +		kfree_rcu(dsq);

Maybe kfree_rcu(dsq, rcu)?

With 7e3f926bf453 ("rcu/kvfree: Eliminate k[v]free_rcu() single argument macro")
we don't allow single argument kfree_rcu() anymore and I don't think we
want to use kfree_rcu_mightsleep() here...

-Andrea

Re: [PATCH 12/34] sched_ext: Implement BPF extensible scheduler class

[Tejun Heo]

On Tue, Jul 11, 2023 at 11:21:48AM +0200, Andrea Righi wrote:
> On Mon, Jul 10, 2023 at 03:13:30PM -1000, Tejun Heo wrote:
> ...
> > +static void free_dsq_irq_workfn(struct irq_work *irq_work)
> > +{
> > +	struct llist_node *to_free = llist_del_all(&dsqs_to_free);
> > +	struct scx_dispatch_q *dsq, *tmp_dsq;
> > +
> > +	llist_for_each_entry_safe(dsq, tmp_dsq, to_free, free_node)
> > +		kfree_rcu(dsq);
> 
> Maybe kfree_rcu(dsq, rcu)?
> 
> With 7e3f926bf453 ("rcu/kvfree: Eliminate k[v]free_rcu() single argument macro")
> we don't allow single argument kfree_rcu() anymore and I don't think we
> want to use kfree_rcu_mightsleep() here...

Oh, thanks for pointing that out. I'll update.

Thanks.

-- 
tejun

Re: [PATCH 27/34] sched_ext: Implement SCX_KICK_WAIT

[Andrea Righi]

On Mon, Jul 10, 2023 at 03:13:45PM -1000, Tejun Heo wrote:
...
> +	for_each_cpu_andnot(cpu, this_rq->scx.cpus_to_wait,
> +			    cpumask_of(this_cpu)) {
> +		/*
> +		 * Pairs with smp_store_release() issued by this CPU in
> +		 * scx_notify_pick_next_task() on the resched path.
> +		 *
> +		 * We busy-wait here to guarantee that no other task can be
> +		 * scheduled on our core before the target CPU has entered the
> +		 * resched path.
> +		 */
> +		while (smp_load_acquire(&cpu_rq(cpu)->scx.pnt_seq) == pseqs[cpu])
> +			cpu_relax();
> +	}
> +

...

> +static inline void scx_notify_pick_next_task(struct rq *rq,
> +					     const struct task_struct *p,
> +					     const struct sched_class *active)
> +{
> +#ifdef CONFIG_SMP
> +	if (!scx_enabled())
> +		return;
> +	/*
> +	 * Pairs with the smp_load_acquire() issued by a CPU in
> +	 * kick_cpus_irq_workfn() who is waiting for this CPU to perform a
> +	 * resched.
> +	 */
> +	smp_store_release(&rq->scx.pnt_seq, rq->scx.pnt_seq + 1);
> +#endif
> +}

We can't use smp_load_acquire()/smp_store_release() with a u64 on
32-bit architectures.

For example, on armhf the build is broken:

In function ‘scx_notify_pick_next_task’,
    inlined from ‘__pick_next_task’ at /<<PKGBUILDDIR>>/kernel/sched/core.c:6106:4,
    inlined from ‘pick_next_task’ at /<<PKGBUILDDIR>>/kernel/sched/core.c:6605:9,
    inlined from ‘__schedule’ at /<<PKGBUILDDIR>>/kernel/sched/core.c:6750:9:
/<<PKGBUILDDIR>>/include/linux/compiler_types.h:397:45: error: call to ‘__compiletime_assert_597’ declared with attribute error: Need native word sized stores/loads for atomicity.
  397 |         _compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
      |                                             ^
/<<PKGBUILDDIR>>/include/linux/compiler_types.h:378:25: note: in definition of macro ‘__compiletime_assert’
  378 |                         prefix ## suffix();                             \
      |                         ^~~~~~
/<<PKGBUILDDIR>>/include/linux/compiler_types.h:397:9: note: in expansion of macro ‘_compiletime_assert’
  397 |         _compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
      |         ^~~~~~~~~~~~~~~~~~~
/<<PKGBUILDDIR>>/include/linux/compiler_types.h:400:9: note: in expansion of macro ‘compiletime_assert’
  400 |         compiletime_assert(__native_word(t),                            \
      |         ^~~~~~~~~~~~~~~~~~
/<<PKGBUILDDIR>>/include/asm-generic/barrier.h:141:9: note: in expansion of macro ‘compiletime_assert_atomic_type’
  141 |         compiletime_assert_atomic_type(*p);                             \
      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/<<PKGBUILDDIR>>/include/asm-generic/barrier.h:172:55: note: in expansion of macro ‘__smp_store_release’
  172 | #define smp_store_release(p, v) do { kcsan_release(); __smp_store_release(p, v); } while (0)
      |                                                       ^~~~~~~~~~~~~~~~~~~
/<<PKGBUILDDIR>>/kernel/sched/ext.h:159:9: note: in expansion of macro ‘smp_store_release’
  159 |         smp_store_release(&rq->scx.pnt_seq, rq->scx.pnt_seq + 1);

There's probably a better way to fix this, but for now I've temporarily
solved this using cmpxchg64() - see patch below.

I'm not sure if we already have an equivalent of
smp_store_release_u64/smp_load_acquire_u64(). Otherwise, it may be worth
to add them to a more generic place.

-Andrea

diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 051c79fa25f7..5da72b1cf88d 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -3667,7 +3667,7 @@ static void kick_cpus_irq_workfn(struct irq_work *irq_work)
 		 * scheduled on our core before the target CPU has entered the
 		 * resched path.
 		 */
-		while (smp_load_acquire(&cpu_rq(cpu)->scx.pnt_seq) == pseqs[cpu])
+		while (smp_load_acquire_u64(&cpu_rq(cpu)->scx.pnt_seq) == pseqs[cpu])
 			cpu_relax();
 	}
 
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index 405037a4e6ce..ef4a24d77d30 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -144,6 +144,40 @@ void __scx_notify_pick_next_task(struct rq *rq,
 				 struct task_struct *p,
 				 const struct sched_class *active);
 
+#ifdef CONFIG_64BIT
+static inline u64 smp_load_acquire_u64(u64 *ptr)
+{
+	return smp_load_acquire(ptr);
+}
+
+static inline void smp_store_release_u64(u64 *ptr, u64 val)
+{
+	smp_store_release(ptr, val);
+}
+#else
+static inline u64 smp_load_acquire_u64(u64 *ptr)
+{
+	u64 prev, next;
+
+	do {
+		prev = *ptr;
+		next = prev;
+	} while (cmpxchg64(ptr, prev, next) != prev);
+
+	return prev;
+}
+
+static inline void smp_store_release_u64(u64 *ptr, u64 val)
+{
+	u64 prev, next;
+
+	do {
+		prev = *ptr;
+		next = val;
+	} while (cmpxchg64(ptr, prev, next) != prev);
+}
+#endif
+
 static inline void scx_notify_pick_next_task(struct rq *rq,
 					     struct task_struct *p,
 					     const struct sched_class *active)
@@ -156,7 +190,7 @@ static inline void scx_notify_pick_next_task(struct rq *rq,
 	 * kick_cpus_irq_workfn() who is waiting for this CPU to perform a
 	 * resched.
 	 */
-	smp_store_release(&rq->scx.pnt_seq, rq->scx.pnt_seq + 1);
+	smp_store_release_u64(&rq->scx.pnt_seq, rq->scx.pnt_seq + 1);
 #endif
 	if (!static_branch_unlikely(&scx_ops_cpu_preempt))
 		return;
-- 
2.40.1

Re: [PATCH 27/34] sched_ext: Implement SCX_KICK_WAIT

[Linus Torvalds]

On Thu, 13 Jul 2023 at 06:46, Andrea Righi <andrea.righi@canonical.com> wrote:
>
> I'm not sure if we already have an equivalent of
> smp_store_release_u64/smp_load_acquire_u64(). Otherwise, it may be worth
> to add them to a more generic place.

Yeah, a 64-bit atomic load/store is not necessarily even possible on
32-bit architectures.

And when it *is* possible, it might be very very expensive indeed (eg
on 32-bit x86, the way to do a 64-bit load would be with "cmpxchg8b",
which is ridiculously slow)

              Linus

Re: [PATCH 27/34] sched_ext: Implement SCX_KICK_WAIT

[Tejun Heo]

Hello,

On Thu, Jul 13, 2023 at 11:32:37AM -0700, Linus Torvalds wrote:
> On Thu, 13 Jul 2023 at 06:46, Andrea Righi <andrea.righi@canonical.com> wrote:
> >
> > I'm not sure if we already have an equivalent of
> > smp_store_release_u64/smp_load_acquire_u64(). Otherwise, it may be worth
> > to add them to a more generic place.
> 
> Yeah, a 64-bit atomic load/store is not necessarily even possible on
> 32-bit architectures.
> 
> And when it *is* possible, it might be very very expensive indeed (eg
> on 32-bit x86, the way to do a 64-bit load would be with "cmpxchg8b",
> which is ridiculously slow)

There are two places where sched_ext is depending on atomic load/store.
One's this pnt_seq which is using smp_store_release/load_acquire(). The
other is task_struct->scx.ops_state which uses atomic64_read_acquire() and
atomic64_store_release(). atomic64's are implemented with spinlocks on
32bits by default which is probably why Andrea didn't hit it.

pnt_seq is a per-cpu counter for successful pick_next_task's from sched_ext
and used to tell "has at least one pick_next_task() succeeded after my
kicking that CPU".

p->scx_ops.state has embedded qseq counter (2bits for state flags, the rest
for the counter. I gotta change the masks to macros too.) which is used to
detect whether the task has been dequeued and re-enqueued between while a
CPU is trying to double lock rq's for task migration.

As both are used to detect races in very short and immediate time windows,
using, respectively, 32bit and 30bit, should be safe practically. e.g. while
it's theoretically possible for the task to be dequeued and re-enqueued
exactly 2^30 times while the CPU is trying to switch rq locks, I don't think
that's practically possible without something going very wrong with the
machine (e.g. NMI / SMI).

I'll note the above and change both to unsigned longs.

Thanks.

-- 
tejun

Re: [PATCHSET v4] sched: Implement BPF extensible scheduler class

[Tejun Heo]

Hello,

It's been more than half a year since the initial posting of the patchset
and we are now at the fourth iteration. There have been some reviews around
specifics (should be all addressed now except for the ones Andrea raised on
this iteration) but none at high level. There also were some in-person and
off-list discussions. Some, I believe, are addressed by the cover letter but
it'd be nonetheless useful to delve into them on-list.

On our side, we've been diligently experimenting. A lot of our earlier
experiments were focused on improving work conservation which led to the CFS
shared runqueue patchset that David Vernet is currently working on. The
workqueue experiments which led to PeterZ's SIS_NODE patch was also informed
by the same work conservation experiments. We also played with soft affinity
which is inspired by Julia's nest scheduler. Another thing we recently
learned was that at least for our web workload, L1/2 locality doesn't matter
much. This is an area of on-going experiments.

We are comfortable with the current API. Everything we tried fit pretty
well. It will continue to evolve but sched_ext now seems mature enough for
initial inclusion. I suppose lack of response doesn't indicate tacit
agreement from everyone, so what are you guys all thinking?

Thanks.

--
tejun

Re: [PATCHSET v4] sched: Implement BPF extensible scheduler class

[Barret Rhoden]

Hi -

On 7/21/23 14:37, Tejun Heo wrote:
> Hello,
> 
> It's been more than half a year since the initial posting of the patchset
> and we are now at the fourth iteration. There have been some reviews around
> specifics (should be all addressed now except for the ones Andrea raised on
> this iteration) but none at high level. There also were some in-person and
> off-list discussions. Some, I believe, are addressed by the cover letter but
> it'd be nonetheless useful to delve into them on-list.
> 
> On our side, we've been diligently experimenting. 

On the google side, we're still experimenting and developing schedulers 
based on ghost, which we think we can port over to sched_ext.

Specifically, I've been working on a framework to write multicore 
schedulers in BPF called 'Flux'.  The idea in brief is to compose a 
scheduler as a hierarchy of "subschedulers", where cpus allocations go 
up and down the tree.

Flux is open-source, but it needs the ghost kernel and our BPF 
extensions currently (which are also open source, but harder to use for 
people).  I'll send a proposal to talk about it at LPC in case people 
are interested - if not the scheduler framework itself, then as a "this 
is some crazy stuff people can do with BPF".

As far as results go, I wrote a custom scheduler with Flux for our 
Search app and have been testing it on our single-leaf loadtester.  The 
initial results out of the box were pretty great: 17% QPS increase, 43% 
p99 decrease (default settings for the loadtester).  But the loadtester 
varies a bit, so it's hard to get reliable numbers out of it for an A/B 
comparison of schedulers.  Overall, we run equal or better than CFS.  I 
did a sweep across various offered loads, and we got 5% better QPS on 
average, 30% better p99 latency, 6% lower utilization.  The better 
numbers come under higher load, as you'd expect, when there are more 
threads competing for the cpu.

The big caveat to those numbers is the single-leaf loadtester isn't a 
representative test.  It's more of a microbenchmark.  Our next step is 
to run a full cluster load test, which will give us a better signal.

Anyway, this scheduler is highly specific to our app, including shared 
memory regions where the app's threads can tell us stuff like RPC 
deadlines.  It's the sort of thing you could only reasonably do with a 
pluggable scheduler like sched_ext or ghost.


> We are comfortable with the current API. Everything we tried fit pretty
> well. It will continue to evolve but sched_ext now seems mature enough for
> initial inclusion. I suppose lack of response doesn't indicate tacit
> agreement from everyone, so what are you guys all thinking?

Btw, I backported your patchset to our "franken-kernel".  I was able to 
boot it on one of our nodes, and run the search loadtest on CFS. 
Nothing broke, performance was the same, etc.  Not a huge surprise, 
since I didn't turn on sched_ext.  I haven't been able to get a 
sched_ext scheduler to work yet with our kernel - there's more patch 
backporting needed for your schedulers themselves (the bpf_for iterators 
and whatnot).  I'll report back if/when I can get it running.

Thanks,

Barret

Re: [PATCHSET v4] sched: Implement BPF extensible scheduler class

[Peter Zijlstra]

On Fri, Jul 21, 2023 at 08:37:41AM -1000, Tejun Heo wrote:

> We are comfortable with the current API. Everything we tried fit pretty
> well. It will continue to evolve but sched_ext now seems mature enough for
> initial inclusion. I suppose lack of response doesn't indicate tacit
> agreement from everyone, so what are you guys all thinking?

I'm still hating the whole thing with a passion.

As can be seen from the wide-spread SCHED_DEBUG abuse; people are, in
general, not interested in doing the right thing. They prod random
numbers (as in really, some are just completely insane) until their
workload improves and call it a day.

There is not a single doubt in my mind that if I were to merge this,
there will be Enterprise software out there that will mandate its own
BPF sched thing, or else it won't work.

They will not care, they will not contribute, they might even pull a
RedHat and only share the code to customers.

We all loose in that scenario. Not least me, because I get the
additional maintenance burden.

I also don't see upsides to merging this. You all can play with
schedulers out-of-tree just fine and then submit what actually works.

So, since you wanted it in writing, here goes:

NAK

Re: [PATCHSET v4] sched: Implement BPF extensible scheduler class

[Tejun Heo]

Hello, Peter.

On Wed, Jul 26, 2023 at 11:17:52AM +0200, Peter Zijlstra wrote:
> On Fri, Jul 21, 2023 at 08:37:41AM -1000, Tejun Heo wrote:
> > We are comfortable with the current API. Everything we tried fit pretty
> > well. It will continue to evolve but sched_ext now seems mature enough for
> > initial inclusion. I suppose lack of response doesn't indicate tacit
> > agreement from everyone, so what are you guys all thinking?
> 
> I'm still hating the whole thing with a passion.
> 
> As can be seen from the wide-spread SCHED_DEBUG abuse; people are, in
> general, not interested in doing the right thing. They prod random
> numbers (as in really, some are just completely insane) until their
> workload improves and call it a day.

I think it'd be useful to add some details to what's going on in situations
like above. This of course wouldn't apply directly to everyone but I suspect
many will recognize at least some parts of it.

In many production setups, there are aspects of workload behaviors that are
difficult to understand comprehensively. The workloads are often massively
complex, constantly being developed by many people, and dynamically
interacting with external entities. As with any sufficiently complex system,
there are many emergent properties which are difficult to untangle
completely.

Add to that multiple generations of divergent hardware and most of the
software stack coming from third parties (including kernel from application
team's POV), people often and justifiably feel as if they're swimming in the
sea of black boxes and emergent properties.

Scheduling, naturally, is one of the areas that people look into when trying
to optimize system performance. Vast majority of people don't know scheduler
code base well enough to hack on it. Even when they do, it's often not easy
to set up benchmarks in production environments and cycle through different
kernels. We (Meta) are a lot better now than a couple years ago, but even
now swapping kernels and ramping workloads back up can take a long time for
certain workloads.

Given the circumstances, it's not surprising that people go for tunable
knobs when they're trying to find out whether changing scheduling behaviors
would improve performance for their workloads. That's often the only option
available and tuning the knobs frequently leads to some gains. Most people
aren't scheduling experts and the causal relationships between changes and
results may not be direct or intuitive. So, that's often where things end.
Given that nobody has found scheduling behavior which is optimal for every
workload and the SCHED_DEBUG knobs are what people can access, it is an
expected outcome.

If a consistent pattern is repeated across multiple workloads, we can
sometimes work back why tuning certain way makes sense and generalize that,
which is to some degree how we ended up focusing on recent work-conservation
related projects.

Maybe the situation is not ideal but I don't think it's people not being
interested in doing the right thing. They are doing what they can within the
confines of available mechanisms, expertise, and time & effort they can
afford to invest.

One of the impediments when trying to connect these disparate data points
into something meaningful is the difficulty in experimentation. The trials
are confined to whatever combinations that can be achieved with SCHED_DEBUG
knobs which are both limiting and obscuring. I believe we're a lot more
likely to learn more about scheduling with sched_ext widely available than
without as it would allow easier and wider-in-scope experimentations.

> There is not a single doubt in my mind that if I were to merge this,
> there will be Enterprise software out there that will mandate its own
> BPF sched thing, or else it won't work.
>
> They will not care, they will not contribute, they might even pull a
> RedHat and only share the code to customers.

I'm sure some will behave in a way which isn't the most conducive to
collective improvement of the upstream kernel. That said, I don't see how
this will be noticeably worsened by inclusion of sched_ext. Most mobile
kernels and some production kernels in cloud environments already carry
significant custom modifications, and they're often addressing real problems
for their use cases.

It'd be ideal if everyone had the commitment and bandwidth to try their best
to merge back their changes but it's also understandable why that can't
always be the case. Sometimes, it's too specific or underdeveloped. At other
times, time and resources just aren't there. We can incentivize and coerce
but that can be pushed only so far. However, we do have an a lot easier time
learning about what people are doing thanks to GPL which all sched_ext
programs would need to follow exactly like the rest of the kernel.

At least relatively speaking, scheduling doesn't seem like an area which is
particularly starved for developer bandwidth although one can always hope
for more. Actual insights and an easy way to experiment and collaborate to
discover them seem like a bigger bottleneck. Hopefully, sched_ext will widen
the scope of things that people will try. Even when they don't directly
contribute those changes back to CFS, if a strategy is effective and general
enough, others can learn from them and apply to improve scheduling for
everyone.

Both Meta and Google are committed to sharing what we learn, both in terms
of code and insights. The example schedulers in the posting are all we
(Meta) have been experimenting with except for really hacky soft affinity
trials which will be generalized and shared too. David has also been
actively working to apply the shared runqueue changes to CFS which came from
earlier sched_ext experiments. Google has been open-sourcing their ghOSt
framework and schedulers built on top of it which will be ported to
sched_ext in the future. Google is starting to see promising results with
search and will share their findings in code and through other venues
including conferences.

> We all loose in that scenario. Not least me, because I get the
> additional maintenance burden.

sched_ext isn't that invasive to the core code and its interactions with
other scheduling classes are very limited. This would make changing
scheduling core API a bit more burdensome but they have been relatively
stable and both David and I would be on the hook if anything is in your way.
I don't see why this would significantly increase your maintenance burden.
It's a thing but it's a thing in its own corner.

> I also don't see upsides to merging this. You all can play with
> schedulers out-of-tree just fine and then submit what actually works.

There is a huge difference between having a common framework upstream and
not having one. If in kernel, everyone knows that it's widely available and
will remain so for a very long time. It removes the risk of investing energy
and effort into something which may or may not exist next year.

It also has the standardizing effect where different parties can exchange
code and ideas easily. It's so much more effective to be able to directly
build upon other people's work than trying to reimplement everything on your
own or navigate maze of different frameworks and patches with different
baseline kernel versions and so on. I mean, these are the reasons that we
want things upstreamed, right?

Thanks.

-- 
tejun

Re: [PATCHSET v4] sched: Implement BPF extensible scheduler class

[Tejun Heo]

Hello,

A gentle ping. I think it warrants further discussions.

Thanks.

-- 
tejun

Re: [PATCHSET v4] sched: Implement BPF extensible scheduler class

[Tejun Heo]

Another ping.

Thanks.

-- 
tejun

Re: [PATCH 12/34] sched_ext: Implement BPF extensible scheduler class

[Vishal Chourasia]

On Mon, Jul 10, 2023 at 03:13:30PM -1000, Tejun Heo wrote:
> Implement a new scheduler class sched_ext (SCX), which allows scheduling
> policies to be implemented as BPF programs to achieve the following:
> 
> 1. Ease of experimentation and exploration: Enabling rapid iteration of new
>    scheduling policies.
> 
> 2. Customization: Building application-specific schedulers which implement
>    policies that are not applicable to general-purpose schedulers.
> 
> 3. Rapid scheduler deployments: Non-disruptive swap outs of scheduling
>    policies in production environments.
> 
> sched_ext leverages BPF’s struct_ops feature to define a structure which
> exports function callbacks and flags to BPF programs that wish to implement
> scheduling policies. The struct_ops structure exported by sched_ext is
> struct sched_ext_ops, and is conceptually similar to struct sched_class. The
> role of sched_ext is to map the complex sched_class callbacks to the more
> simple and ergonomic struct sched_ext_ops callbacks.
> 
> For more detailed discussion on the motivations and overview, please refer
> to the cover letter.
> 
> Later patches will also add several example schedulers and documentation.
> 
> This patch implements the minimum core framework to enable implementation of
> BPF schedulers. Subsequent patches will gradually add functionalities
> including safety guarantee mechanisms, nohz and cgroup support.
> 
> include/linux/sched/ext.h defines struct sched_ext_ops. With the comment on
> top, each operation should be self-explanatory. The followings are worth
> noting:
> 
> * Both "sched_ext" and its shorthand "scx" are used. If the identifier
>   already has "sched" in it, "ext" is used; otherwise, "scx".
> 
> * In sched_ext_ops, only .name is mandatory. Every operation is optional and
>   if omitted a simple but functional default behavior is provided.
> 
> * A new policy constant SCHED_EXT is added and a task can select sched_ext
>   by invoking sched_setscheduler(2) with the new policy constant. However,
>   if the BPF scheduler is not loaded, SCHED_EXT is the same as SCHED_NORMAL
>   and the task is scheduled by CFS. When the BPF scheduler is loaded, all
>   tasks which have the SCHED_EXT policy are switched to sched_ext.
> 
> * To bridge the workflow imbalance between the scheduler core and
>   sched_ext_ops callbacks, sched_ext uses simple FIFOs called dispatch
>   queues (dsq's). By default, there is one global dsq (SCX_DSQ_GLOBAL), and
>   one local per-CPU dsq (SCX_DSQ_LOCAL). SCX_DSQ_GLOBAL is provided for
>   convenience and need not be used by a scheduler that doesn't require it.
>   SCX_DSQ_LOCAL is the per-CPU FIFO that sched_ext pulls from when putting
>   the next task on the CPU. The BPF scheduler can manage an arbitrary number
>   of dsq's using scx_bpf_create_dsq() and scx_bpf_destroy_dsq().
> 
> * sched_ext guarantees system integrity no matter what the BPF scheduler
>   does. To enable this, each task's ownership is tracked through
>   p->scx.ops_state and all tasks are put on scx_tasks list. The disable path
>   can always recover and revert all tasks back to CFS. See p->scx.ops_state
>   and scx_tasks.
> 
> * A task is not tied to its rq while enqueued. This decouples CPU selection
>   from queueing and allows sharing a scheduling queue across an arbitrary
>   subset of CPUs. This adds some complexities as a task may need to be
>   bounced between rq's right before it starts executing. See
>   dispatch_to_local_dsq() and move_task_to_local_dsq().
> 
> * One complication that arises from the above weak association between task
>   and rq is that synchronizing with dequeue() gets complicated as dequeue()
>   may happen anytime while the task is enqueued and the dispatch path might
>   need to release the rq lock to transfer the task. Solving this requires a
>   bit of complexity. See the logic around p->scx.sticky_cpu and
>   p->scx.ops_qseq.
> 
> * Both enable and disable paths are a bit complicated. The enable path
>   switches all tasks without blocking to avoid issues which can arise from
>   partially switched states (e.g. the switching task itself being starved).
>   The disable path can't trust the BPF scheduler at all, so it also has to
>   guarantee forward progress without blocking. See scx_ops_enable() and
>   scx_ops_disable_workfn().
> 
> * When sched_ext is disabled, static_branches are used to shut down the
>   entry points from hot paths.
> 
> v4: * SCHED_CHANGE_BLOCK replaced with the previous
>       sched_deq_and_put_task()/sched_enq_and_set_tsak() pair. This is
>       because upstream is adaopting a different generic cleanup mechanism.
>       Once that lands, the code will be adapted accordingly.
> 
>     * task_on_scx() used to test whether a task should be switched into SCX,
>       which is confusing. Renamed to task_should_scx(). task_on_scx() now
>       tests whether a task is currently on SCX.
> 
>     * scx_has_idle_cpus is barely used anymore and replaced with direct
>       check on the idle cpumask.
> 
>     * SCX_PICK_IDLE_CORE added and scx_pick_idle_cpu() improved to prefer
>       fully idle cores.
> 
>     * ops.enable() now sees up-to-date p->scx.weight value.
> 
>     * ttwu_queue path is disabled for tasks on SCX to avoid confusing BPF
>       schedulers expecting ->select_cpu() call.
> 
>     * Use cpu_smt_mask() instead of topology_sibling_cpumask() like the rest
>       of the scheduler.
> 
> v3: * ops.set_weight() added to allow BPF schedulers to track weight changes
>       without polling p->scx.weight.
> 
>     * move_task_to_local_dsq() was losing SCX-specific enq_flags when
>       enqueueing the task on the target dsq because it goes through
>       activate_task() which loses the upper 32bit of the flags. Carry the
>       flags through rq->scx.extra_enq_flags.
> 
>     * scx_bpf_dispatch(), scx_bpf_pick_idle_cpu(), scx_bpf_task_running()
>       and scx_bpf_task_cpu() now use the new KF_RCU instead of
>       KF_TRUSTED_ARGS to make it easier for BPF schedulers to call them.
> 
>     * The kfunc helper access control mechanism implemented through
>       sched_ext_entity.kf_mask is improved. Now SCX_CALL_OP*() is always
>       used when invoking scx_ops operations.
> 
> v2: * balance_scx_on_up() is dropped. Instead, on UP, balance_scx() is
>       called from put_prev_taks_scx() and pick_next_task_scx() as necessary.
>       To determine whether balance_scx() should be called from
>       put_prev_task_scx(), SCX_TASK_DEQD_FOR_SLEEP flag is added. See the
>       comment in put_prev_task_scx() for details.
> 
>     * sched_deq_and_put_task() / sched_enq_and_set_task() sequences replaced
>       with SCHED_CHANGE_BLOCK().
> 
>     * Unused all_dsqs list removed. This was a left-over from previous
>       iterations.
> 
>     * p->scx.kf_mask is added to track and enforce which kfunc helpers are
>       allowed. Also, init/exit sequences are updated to make some kfuncs
>       always safe to call regardless of the current BPF scheduler state.
>       Combined, this should make all the kfuncs safe.
> 
>     * BPF now supports sleepable struct_ops operations. Hacky workaround
>       removed and operations and kfunc helpers are tagged appropriately.
> 
>     * BPF now supports bitmask / cpumask helpers. scx_bpf_get_idle_cpumask()
>       and friends are added so that BPF schedulers can use the idle masks
>       with the generic helpers. This replaces the hacky kfunc helpers added
>       by a separate patch in V1.
> 
>     * CONFIG_SCHED_CLASS_EXT can no longer be enabled if SCHED_CORE is
>       enabled. This restriction will be removed by a later patch which adds
>       core-sched support.
> 
>     * Add MAINTAINERS entries and other misc changes.
> 
> Signed-off-by: Tejun Heo <tj@kernel.org>
> Co-authored-by: David Vernet <dvernet@meta.com>
> Acked-by: Josh Don <joshdon@google.com>
> Acked-by: Hao Luo <haoluo@google.com>
> Acked-by: Barret Rhoden <brho@google.com>
> Cc: Andrea Righi <andrea.righi@canonical.com>
> ---
>  MAINTAINERS                       |    3 +
>  include/asm-generic/vmlinux.lds.h |    1 +
>  include/linux/sched.h             |    5 +
>  include/linux/sched/ext.h         |  400 +++-
>  include/uapi/linux/sched.h        |    1 +
>  init/init_task.c                  |   10 +
>  kernel/Kconfig.preempt            |   22 +-
>  kernel/bpf/bpf_struct_ops_types.h |    4 +
>  kernel/sched/build_policy.c       |    4 +
>  kernel/sched/core.c               |   70 +
>  kernel/sched/debug.c              |    6 +
>  kernel/sched/ext.c                | 3140 +++++++++++++++++++++++++++++
>  kernel/sched/ext.h                |  118 +-
>  kernel/sched/sched.h              |   16 +
>  14 files changed, 3796 insertions(+), 4 deletions(-)
>  create mode 100644 kernel/sched/ext.c
> 
> diff --git a/MAINTAINERS b/MAINTAINERS
> index c904dba1733b..5c301e22ff74 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -18767,6 +18767,8 @@ R:	Ben Segall <bsegall@google.com> (CONFIG_CFS_BANDWIDTH)
>  R:	Mel Gorman <mgorman@suse.de> (CONFIG_NUMA_BALANCING)
>  R:	Daniel Bristot de Oliveira <bristot@redhat.com> (SCHED_DEADLINE)
>  R:	Valentin Schneider <vschneid@redhat.com> (TOPOLOGY)
> +R:	Tejun Heo <tj@kernel.org> (SCHED_EXT)
> +R:	David Vernet <void@manifault.com> (SCHED_EXT)
>  L:	linux-kernel@vger.kernel.org
>  S:	Maintained
>  T:	git git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git sched/core
> @@ -18775,6 +18777,7 @@ F:	include/linux/sched.h
>  F:	include/linux/wait.h
>  F:	include/uapi/linux/sched.h
>  F:	kernel/sched/
> +F:	tools/sched_ext/
>  
>  SCSI RDMA PROTOCOL (SRP) INITIATOR
>  M:	Bart Van Assche <bvanassche@acm.org>
> diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h
> index d1f57e4868ed..cd5a718ba49f 100644
> --- a/include/asm-generic/vmlinux.lds.h
> +++ b/include/asm-generic/vmlinux.lds.h
> @@ -131,6 +131,7 @@
>  	*(__dl_sched_class)			\
>  	*(__rt_sched_class)			\
>  	*(__fair_sched_class)			\
> +	*(__ext_sched_class)			\
>  	*(__idle_sched_class)			\
>  	__sched_class_lowest = .;
>  
> diff --git a/include/linux/sched.h b/include/linux/sched.h
> index eed5d65b8d1f..00d4ce3af52a 100644
> --- a/include/linux/sched.h
> +++ b/include/linux/sched.h
> @@ -72,6 +72,8 @@ struct task_delay_info;
>  struct task_group;
>  struct user_event_mm;
>  
> +#include <linux/sched/ext.h>
> +
>  /*
>   * Task state bitmask. NOTE! These bits are also
>   * encoded in fs/proc/array.c: get_task_state().
> @@ -790,6 +792,9 @@ struct task_struct {
>  	struct sched_entity		se;
>  	struct sched_rt_entity		rt;
>  	struct sched_dl_entity		dl;
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +	struct sched_ext_entity		scx;
> +#endif
>  	const struct sched_class	*sched_class;
>  
>  #ifdef CONFIG_SCHED_CORE
> diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
> index a05dfcf533b0..92011a63cc15 100644
> --- a/include/linux/sched/ext.h
> +++ b/include/linux/sched/ext.h
> @@ -1,9 +1,407 @@
>  /* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
> + * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
> + * Copyright (c) 2022 David Vernet <dvernet@meta.com>
> + */
>  #ifndef _LINUX_SCHED_EXT_H
>  #define _LINUX_SCHED_EXT_H
>  
>  #ifdef CONFIG_SCHED_CLASS_EXT
> -#error "NOT IMPLEMENTED YET"
> +
> +#include <linux/rhashtable.h>
> +#include <linux/llist.h>
> +
> +enum scx_consts {
> +	SCX_OPS_NAME_LEN	= 128,
> +	SCX_EXIT_REASON_LEN	= 128,
> +	SCX_EXIT_BT_LEN		= 64,
> +	SCX_EXIT_MSG_LEN	= 1024,
> +
> +	SCX_SLICE_DFL		= 20 * NSEC_PER_MSEC,
> +};
> +
> +/*
> + * DSQ (dispatch queue) IDs are 64bit of the format:
> + *
> + *   Bits: [63] [62 ..  0]
> + *         [ B] [   ID   ]
> + *
> + *    B: 1 for IDs for built-in DSQs, 0 for ops-created user DSQs
> + *   ID: 63 bit ID
> + *
> + * Built-in IDs:
> + *
> + *   Bits: [63] [62] [61..32] [31 ..  0]
> + *         [ 1] [ L] [   R  ] [    V   ]
> + *
> + *    1: 1 for built-in DSQs.
> + *    L: 1 for LOCAL_ON DSQ IDs, 0 for others
> + *    V: For LOCAL_ON DSQ IDs, a CPU number. For others, a pre-defined value.
> + */
> +enum scx_dsq_id_flags {
> +	SCX_DSQ_FLAG_BUILTIN	= 1LLU << 63,
> +	SCX_DSQ_FLAG_LOCAL_ON	= 1LLU << 62,
> +
> +	SCX_DSQ_INVALID		= SCX_DSQ_FLAG_BUILTIN | 0,
> +	SCX_DSQ_GLOBAL		= SCX_DSQ_FLAG_BUILTIN | 1,
> +	SCX_DSQ_LOCAL		= SCX_DSQ_FLAG_BUILTIN | 2,
> +	SCX_DSQ_LOCAL_ON	= SCX_DSQ_FLAG_BUILTIN | SCX_DSQ_FLAG_LOCAL_ON,
> +	SCX_DSQ_LOCAL_CPU_MASK	= 0xffffffffLLU,
> +};
> +
> +enum scx_exit_type {
> +	SCX_EXIT_NONE,
> +	SCX_EXIT_DONE,
> +
> +	SCX_EXIT_UNREG = 64,	/* BPF unregistration */
> +
> +	SCX_EXIT_ERROR = 1024,	/* runtime error, error msg contains details */
> +	SCX_EXIT_ERROR_BPF,	/* ERROR but triggered through scx_bpf_error() */
> +};
> +
> +/*
> + * scx_exit_info is passed to ops.exit() to describe why the BPF scheduler is
> + * being disabled.
> + */
> +struct scx_exit_info {
> +	/* %SCX_EXIT_* - broad category of the exit reason */
> +	enum scx_exit_type	type;
> +	/* textual representation of the above */
> +	char			reason[SCX_EXIT_REASON_LEN];
> +	/* number of entries in the backtrace */
> +	u32			bt_len;
> +	/* backtrace if exiting due to an error */
> +	unsigned long		bt[SCX_EXIT_BT_LEN];
> +	/* extra message */
> +	char			msg[SCX_EXIT_MSG_LEN];
> +};
> +
> +/* sched_ext_ops.flags */
> +enum scx_ops_flags {
> +	/*
> +	 * Keep built-in idle tracking even if ops.update_idle() is implemented.
> +	 */
> +	SCX_OPS_KEEP_BUILTIN_IDLE = 1LLU << 0,
> +
> +	/*
> +	 * By default, if there are no other task to run on the CPU, ext core
> +	 * keeps running the current task even after its slice expires. If this
> +	 * flag is specified, such tasks are passed to ops.enqueue() with
> +	 * %SCX_ENQ_LAST. See the comment above %SCX_ENQ_LAST for more info.
> +	 */
> +	SCX_OPS_ENQ_LAST	= 1LLU << 1,
> +
> +	/*
> +	 * An exiting task may schedule after PF_EXITING is set. In such cases,
> +	 * bpf_task_from_pid() may not be able to find the task and if the BPF
> +	 * scheduler depends on pid lookup for dispatching, the task will be
> +	 * lost leading to various issues including RCU grace period stalls.
> +	 *
> +	 * To mask this problem, by default, unhashed tasks are automatically
> +	 * dispatched to the local DSQ on enqueue. If the BPF scheduler doesn't
> +	 * depend on pid lookups and wants to handle these tasks directly, the
> +	 * following flag can be used.
> +	 */
> +	SCX_OPS_ENQ_EXITING	= 1LLU << 2,
> +
> +	SCX_OPS_ALL_FLAGS	= SCX_OPS_KEEP_BUILTIN_IDLE |
> +				  SCX_OPS_ENQ_LAST |
> +				  SCX_OPS_ENQ_EXITING,
> +};
> +
> +/* argument container for ops.enable() and friends */
> +struct scx_enable_args {
> +	/* empty for now */
> +};
> +
> +/**
> + * struct sched_ext_ops - Operation table for BPF scheduler implementation
> + *
> + * Userland can implement an arbitrary scheduling policy by implementing and
> + * loading operations in this table.
> + */
> +struct sched_ext_ops {
> +	/**
> +	 * select_cpu - Pick the target CPU for a task which is being woken up
> +	 * @p: task being woken up
> +	 * @prev_cpu: the cpu @p was on before sleeping
> +	 * @wake_flags: SCX_WAKE_*
> +	 *
> +	 * Decision made here isn't final. @p may be moved to any CPU while it
> +	 * is getting dispatched for execution later. However, as @p is not on
> +	 * the rq at this point, getting the eventual execution CPU right here
> +	 * saves a small bit of overhead down the line.
> +	 *
> +	 * If an idle CPU is returned, the CPU is kicked and will try to
> +	 * dispatch. While an explicit custom mechanism can be added,
> +	 * select_cpu() serves as the default way to wake up idle CPUs.
> +	 */
> +	s32 (*select_cpu)(struct task_struct *p, s32 prev_cpu, u64 wake_flags);
> +
> +	/**
> +	 * enqueue - Enqueue a task on the BPF scheduler
> +	 * @p: task being enqueued
> +	 * @enq_flags: %SCX_ENQ_*
> +	 *
> +	 * @p is ready to run. Dispatch directly by calling scx_bpf_dispatch()
> +	 * or enqueue on the BPF scheduler. If not directly dispatched, the bpf
> +	 * scheduler owns @p and if it fails to dispatch @p, the task will
> +	 * stall.
> +	 */
> +	void (*enqueue)(struct task_struct *p, u64 enq_flags);
> +
> +	/**
> +	 * dequeue - Remove a task from the BPF scheduler
> +	 * @p: task being dequeued
> +	 * @deq_flags: %SCX_DEQ_*
> +	 *
> +	 * Remove @p from the BPF scheduler. This is usually called to isolate
> +	 * the task while updating its scheduling properties (e.g. priority).
> +	 *
> +	 * The ext core keeps track of whether the BPF side owns a given task or
> +	 * not and can gracefully ignore spurious dispatches from BPF side,
> +	 * which makes it safe to not implement this method. However, depending
> +	 * on the scheduling logic, this can lead to confusing behaviors - e.g.
> +	 * scheduling position not being updated across a priority change.
> +	 */
> +	void (*dequeue)(struct task_struct *p, u64 deq_flags);
> +
> +	/**
> +	 * dispatch - Dispatch tasks from the BPF scheduler and/or consume DSQs
> +	 * @cpu: CPU to dispatch tasks for
> +	 * @prev: previous task being switched out
> +	 *
> +	 * Called when a CPU's local dsq is empty. The operation should dispatch
> +	 * one or more tasks from the BPF scheduler into the DSQs using
> +	 * scx_bpf_dispatch() and/or consume user DSQs into the local DSQ using
> +	 * scx_bpf_consume().
> +	 *
> +	 * The maximum number of times scx_bpf_dispatch() can be called without
> +	 * an intervening scx_bpf_consume() is specified by
> +	 * ops.dispatch_max_batch. See the comments on top of the two functions
> +	 * for more details.
> +	 *
> +	 * When not %NULL, @prev is an SCX task with its slice depleted. If
> +	 * @prev is still runnable as indicated by set %SCX_TASK_QUEUED in
> +	 * @prev->scx.flags, it is not enqueued yet and will be enqueued after
> +	 * ops.dispatch() returns. To keep executing @prev, return without
> +	 * dispatching or consuming any tasks. Also see %SCX_OPS_ENQ_LAST.
> +	 */
> +	void (*dispatch)(s32 cpu, struct task_struct *prev);
> +
> +	/**
> +	 * yield - Yield CPU
> +	 * @from: yielding task
> +	 * @to: optional yield target task
> +	 *
> +	 * If @to is NULL, @from is yielding the CPU to other runnable tasks.
> +	 * The BPF scheduler should ensure that other available tasks are
> +	 * dispatched before the yielding task. Return value is ignored in this
> +	 * case.
> +	 *
> +	 * If @to is not-NULL, @from wants to yield the CPU to @to. If the bpf
> +	 * scheduler can implement the request, return %true; otherwise, %false.
> +	 */
> +	bool (*yield)(struct task_struct *from, struct task_struct *to);
> +
> +	/**
> +	 * set_weight - Set task weight
> +	 * @p: task to set weight for
> +	 * @weight: new eight [1..10000]
> +	 *
> +	 * Update @p's weight to @weight.
> +	 */
> +	void (*set_weight)(struct task_struct *p, u32 weight);
> +
> +	/**
> +	 * set_cpumask - Set CPU affinity
> +	 * @p: task to set CPU affinity for
> +	 * @cpumask: cpumask of cpus that @p can run on
> +	 *
> +	 * Update @p's CPU affinity to @cpumask.
> +	 */
> +	void (*set_cpumask)(struct task_struct *p, struct cpumask *cpumask);
> +
> +	/**
> +	 * update_idle - Update the idle state of a CPU
> +	 * @cpu: CPU to udpate the idle state for
> +	 * @idle: whether entering or exiting the idle state
> +	 *
> +	 * This operation is called when @rq's CPU goes or leaves the idle
> +	 * state. By default, implementing this operation disables the built-in
> +	 * idle CPU tracking and the following helpers become unavailable:
> +	 *
> +	 * - scx_bpf_select_cpu_dfl()
> +	 * - scx_bpf_test_and_clear_cpu_idle()
> +	 * - scx_bpf_pick_idle_cpu()
> +	 *
> +	 * The user also must implement ops.select_cpu() as the default
> +	 * implementation relies on scx_bpf_select_cpu_dfl().
> +	 *
> +	 * Specify the %SCX_OPS_KEEP_BUILTIN_IDLE flag to keep the built-in idle
> +	 * tracking.
> +	 */
> +	void (*update_idle)(s32 cpu, bool idle);
> +
> +	/**
> +	 * prep_enable - Prepare to enable BPF scheduling for a task
> +	 * @p: task to prepare BPF scheduling for
> +	 * @args: enable arguments, see the struct definition
> +	 *
> +	 * Either we're loading a BPF scheduler or a new task is being forked.
> +	 * Prepare BPF scheduling for @p. This operation may block and can be
> +	 * used for allocations.
> +	 *
> +	 * Return 0 for success, -errno for failure. An error return while
> +	 * loading will abort loading of the BPF scheduler. During a fork, will
> +	 * abort the specific fork.
> +	 */
> +	s32 (*prep_enable)(struct task_struct *p, struct scx_enable_args *args);
> +
> +	/**
> +	 * enable - Enable BPF scheduling for a task
> +	 * @p: task to enable BPF scheduling for
> +	 * @args: enable arguments, see the struct definition
> +	 *
> +	 * Enable @p for BPF scheduling. @p will start running soon.
> +	 */
> +	void (*enable)(struct task_struct *p, struct scx_enable_args *args);
> +
> +	/**
> +	 * cancel_enable - Cancel prep_enable()
> +	 * @p: task being canceled
> +	 * @args: enable arguments, see the struct definition
> +	 *
> +	 * @p was prep_enable()'d but failed before reaching enable(). Undo the
> +	 * preparation.
> +	 */
> +	void (*cancel_enable)(struct task_struct *p,
> +			      struct scx_enable_args *args);
> +
> +	/**
> +	 * disable - Disable BPF scheduling for a task
> +	 * @p: task to disable BPF scheduling for
> +	 *
> +	 * @p is exiting, leaving SCX or the BPF scheduler is being unloaded.
> +	 * Disable BPF scheduling for @p.
> +	 */
> +	void (*disable)(struct task_struct *p);
> +
> +	/*
> +	 * All online ops must come before ops.init().
> +	 */
> +
> +	/**
> +	 * init - Initialize the BPF scheduler
> +	 */
> +	s32 (*init)(void);
> +
> +	/**
> +	 * exit - Clean up after the BPF scheduler
> +	 * @info: Exit info
> +	 */
> +	void (*exit)(struct scx_exit_info *info);
> +
> +	/**
> +	 * dispatch_max_batch - Max nr of tasks that dispatch() can dispatch
> +	 */
> +	u32 dispatch_max_batch;
> +
> +	/**
> +	 * flags - %SCX_OPS_* flags
> +	 */
> +	u64 flags;
> +
> +	/**
> +	 * name - BPF scheduler's name
> +	 *
> +	 * Must be a non-zero valid BPF object name including only isalnum(),
> +	 * '_' and '.' chars. Shows up in kernel.sched_ext_ops sysctl while the
> +	 * BPF scheduler is enabled.
> +	 */
> +	char name[SCX_OPS_NAME_LEN];
> +};
> +
> +/*
> + * Dispatch queue (dsq) is a simple FIFO which is used to buffer between the
> + * scheduler core and the BPF scheduler. See the documentation for more details.
> + */
> +struct scx_dispatch_q {
> +	raw_spinlock_t		lock;
> +	struct list_head	fifo;	/* processed in dispatching order */
> +	u32			nr;
> +	u64			id;
> +	struct rhash_head	hash_node;
> +	struct llist_node	free_node;
> +	struct rcu_head		rcu;
> +};
> +
> +/* scx_entity.flags */
> +enum scx_ent_flags {
> +	SCX_TASK_QUEUED		= 1 << 0, /* on ext runqueue */
> +	SCX_TASK_BAL_KEEP	= 1 << 1, /* balance decided to keep current */
> +	SCX_TASK_ENQ_LOCAL	= 1 << 2, /* used by scx_select_cpu_dfl() to set SCX_ENQ_LOCAL */
> +
> +	SCX_TASK_OPS_PREPPED	= 1 << 8, /* prepared for BPF scheduler enable */
> +	SCX_TASK_OPS_ENABLED	= 1 << 9, /* task has BPF scheduler enabled */
> +
> +	SCX_TASK_DEQD_FOR_SLEEP	= 1 << 17, /* last dequeue was for SLEEP */
> +
> +	SCX_TASK_CURSOR		= 1 << 31, /* iteration cursor, not a task */
> +};
> +
> +/*
> + * Mask bits for scx_entity.kf_mask. Not all kfuncs can be called from
> + * everywhere and the following bits track which kfunc sets are currently
> + * allowed for %current. This simple per-task tracking works because SCX ops
> + * nest in a limited way. BPF will likely implement a way to allow and disallow
> + * kfuncs depending on the calling context which will replace this manual
> + * mechanism. See scx_kf_allow().
> + */
> +enum scx_kf_mask {
> +	SCX_KF_UNLOCKED		= 0,	  /* not sleepable, not rq locked */
> +	/* all non-sleepables may be nested inside INIT and SLEEPABLE */
> +	SCX_KF_INIT		= 1 << 0, /* running ops.init() */
> +	SCX_KF_SLEEPABLE	= 1 << 1, /* other sleepable init operations */
> +	/* ops.dequeue (in REST) may be nested inside DISPATCH */
> +	SCX_KF_DISPATCH		= 1 << 3, /* ops.dispatch() */
> +	SCX_KF_ENQUEUE		= 1 << 4, /* ops.enqueue() */
> +	SCX_KF_REST		= 1 << 5, /* other rq-locked operations */
> +
> +	__SCX_KF_RQ_LOCKED	= SCX_KF_DISPATCH | SCX_KF_ENQUEUE | SCX_KF_REST,
> +};
> +
> +/*
> + * The following is embedded in task_struct and contains all fields necessary
> + * for a task to be scheduled by SCX.
> + */
> +struct sched_ext_entity {
> +	struct scx_dispatch_q	*dsq;
> +	struct list_head	dsq_node;
> +	u32			flags;		/* protected by rq lock */
> +	u32			weight;
> +	s32			sticky_cpu;
> +	s32			holding_cpu;
> +	u32			kf_mask;	/* see scx_kf_mask above */
> +	atomic64_t		ops_state;
> +
> +	/* BPF scheduler modifiable fields */
> +
> +	/*
> +	 * Runtime budget in nsecs. This is usually set through
> +	 * scx_bpf_dispatch() but can also be modified directly by the BPF
> +	 * scheduler. Automatically decreased by SCX as the task executes. On
> +	 * depletion, a scheduling event is triggered.
> +	 */
> +	u64			slice;
> +
> +	/* cold fields */
> +	struct list_head	tasks_node;
> +};
> +
> +void sched_ext_free(struct task_struct *p);
> +
>  #else	/* !CONFIG_SCHED_CLASS_EXT */
>  
>  static inline void sched_ext_free(struct task_struct *p) {}
> diff --git a/include/uapi/linux/sched.h b/include/uapi/linux/sched.h
> index 3bac0a8ceab2..359a14cc76a4 100644
> --- a/include/uapi/linux/sched.h
> +++ b/include/uapi/linux/sched.h
> @@ -118,6 +118,7 @@ struct clone_args {
>  /* SCHED_ISO: reserved but not implemented yet */
>  #define SCHED_IDLE		5
>  #define SCHED_DEADLINE		6
> +#define SCHED_EXT		7
>  
>  /* Can be ORed in to make sure the process is reverted back to SCHED_NORMAL on fork */
>  #define SCHED_RESET_ON_FORK     0x40000000
> diff --git a/init/init_task.c b/init/init_task.c
> index ff6c4b9bfe6b..bdbc663107bf 100644
> --- a/init/init_task.c
> +++ b/init/init_task.c
> @@ -6,6 +6,7 @@
>  #include <linux/sched/sysctl.h>
>  #include <linux/sched/rt.h>
>  #include <linux/sched/task.h>
> +#include <linux/sched/ext.h>
>  #include <linux/init.h>
>  #include <linux/fs.h>
>  #include <linux/mm.h>
> @@ -101,6 +102,15 @@ struct task_struct init_task
>  #endif
>  #ifdef CONFIG_CGROUP_SCHED
>  	.sched_task_group = &root_task_group,
> +#endif
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +	.scx		= {
> +		.dsq_node	= LIST_HEAD_INIT(init_task.scx.dsq_node),
> +		.sticky_cpu	= -1,
> +		.holding_cpu	= -1,
> +		.ops_state	= ATOMIC_INIT(0),
> +		.slice		= SCX_SLICE_DFL,
> +	},
>  #endif
>  	.ptraced	= LIST_HEAD_INIT(init_task.ptraced),
>  	.ptrace_entry	= LIST_HEAD_INIT(init_task.ptrace_entry),
> diff --git a/kernel/Kconfig.preempt b/kernel/Kconfig.preempt
> index c2f1fd95a821..0afcda19bc50 100644
> --- a/kernel/Kconfig.preempt
> +++ b/kernel/Kconfig.preempt
> @@ -133,4 +133,24 @@ config SCHED_CORE
>  	  which is the likely usage by Linux distributions, there should
>  	  be no measurable impact on performance.
>  
> -
> +config SCHED_CLASS_EXT
> +	bool "Extensible Scheduling Class"
> +	depends on BPF_SYSCALL && BPF_JIT && !SCHED_CORE
> +	help
> +	  This option enables a new scheduler class sched_ext (SCX), which
> +	  allows scheduling policies to be implemented as BPF programs to
> +	  achieve the following:
> +
> +	  - Ease of experimentation and exploration: Enabling rapid
> +	    iteration of new scheduling policies.
> +	  - Customization: Building application-specific schedulers which
> +	    implement policies that are not applicable to general-purpose
> +	    schedulers.
> +	  - Rapid scheduler deployments: Non-disruptive swap outs of
> +	    scheduling policies in production environments.
> +
> +	  sched_ext leverages BPF’s struct_ops feature to define a structure
> +	  which exports function callbacks and flags to BPF programs that
> +	  wish to implement scheduling policies. The struct_ops structure
> +	  exported by sched_ext is struct sched_ext_ops, and is conceptually
> +	  similar to struct sched_class.
> diff --git a/kernel/bpf/bpf_struct_ops_types.h b/kernel/bpf/bpf_struct_ops_types.h
> index 5678a9ddf817..3618769d853d 100644
> --- a/kernel/bpf/bpf_struct_ops_types.h
> +++ b/kernel/bpf/bpf_struct_ops_types.h
> @@ -9,4 +9,8 @@ BPF_STRUCT_OPS_TYPE(bpf_dummy_ops)
>  #include <net/tcp.h>
>  BPF_STRUCT_OPS_TYPE(tcp_congestion_ops)
>  #endif
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +#include <linux/sched/ext.h>
> +BPF_STRUCT_OPS_TYPE(sched_ext_ops)
> +#endif
>  #endif
> diff --git a/kernel/sched/build_policy.c b/kernel/sched/build_policy.c
> index d9dc9ab3773f..4c658b21f603 100644
> --- a/kernel/sched/build_policy.c
> +++ b/kernel/sched/build_policy.c
> @@ -28,6 +28,7 @@
>  #include <linux/suspend.h>
>  #include <linux/tsacct_kern.h>
>  #include <linux/vtime.h>
> +#include <linux/percpu-rwsem.h>
>  
>  #include <uapi/linux/sched/types.h>
>  
> @@ -52,3 +53,6 @@
>  #include "cputime.c"
>  #include "deadline.c"
>  
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +# include "ext.c"
> +#endif
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index c8a2c99248b7..a182a9adec3e 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -3925,6 +3925,15 @@ bool cpus_share_cache(int this_cpu, int that_cpu)
>  
>  static inline bool ttwu_queue_cond(struct task_struct *p, int cpu)
>  {
> +	/*
> +	 * The BPF scheduler may depend on select_task_rq() being invoked during
> +	 * wakeups. In addition, @p may end up executing on a different CPU
> +	 * regardless of what happens in the wakeup path making the ttwu_queue
> +	 * optimization less meaningful. Skip if on SCX.
> +	 */
> +	if (task_on_scx(p))
> +		return false;
> +
>  	/*
>  	 * Do not complicate things with the async wake_list while the CPU is
>  	 * in hotplug state.
> @@ -4496,6 +4505,18 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
>  	p->rt.on_rq		= 0;
>  	p->rt.on_list		= 0;
>  
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +	p->scx.dsq		= NULL;
> +	INIT_LIST_HEAD(&p->scx.dsq_node);
> +	p->scx.flags		= 0;
> +	p->scx.weight		= 0;
> +	p->scx.sticky_cpu	= -1;
> +	p->scx.holding_cpu	= -1;
> +	p->scx.kf_mask		= 0;
> +	atomic64_set(&p->scx.ops_state, 0);
> +	p->scx.slice		= SCX_SLICE_DFL;
> +#endif
> +
>  #ifdef CONFIG_PREEMPT_NOTIFIERS
>  	INIT_HLIST_HEAD(&p->preempt_notifiers);
>  #endif
> @@ -4744,6 +4765,10 @@ int sched_fork(unsigned long clone_flags, struct task_struct *p)
>  		goto out_cancel;
>  	} else if (rt_prio(p->prio)) {
>  		p->sched_class = &rt_sched_class;
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +	} else if (task_should_scx(p)) {
> +		p->sched_class = &ext_sched_class;
> +#endif
>  	} else {
>  		p->sched_class = &fair_sched_class;
>  	}
> @@ -7032,6 +7057,10 @@ void __setscheduler_prio(struct task_struct *p, int prio)
>  		p->sched_class = &dl_sched_class;
>  	else if (rt_prio(prio))
>  		p->sched_class = &rt_sched_class;
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +	else if (task_should_scx(p))
> +		p->sched_class = &ext_sched_class;
> +#endif
>  	else
>  		p->sched_class = &fair_sched_class;
>  
> @@ -9036,6 +9065,7 @@ SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
>  	case SCHED_NORMAL:
>  	case SCHED_BATCH:
>  	case SCHED_IDLE:
> +	case SCHED_EXT:
>  		ret = 0;
>  		break;
>  	}
> @@ -9063,6 +9093,7 @@ SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
>  	case SCHED_NORMAL:
>  	case SCHED_BATCH:
>  	case SCHED_IDLE:
> +	case SCHED_EXT:
>  		ret = 0;
>  	}
>  	return ret;
> @@ -9918,6 +9949,10 @@ void __init sched_init(void)
>  	BUG_ON(!sched_class_above(&dl_sched_class, &rt_sched_class));
>  	BUG_ON(!sched_class_above(&rt_sched_class, &fair_sched_class));
>  	BUG_ON(!sched_class_above(&fair_sched_class, &idle_sched_class));
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +	BUG_ON(!sched_class_above(&fair_sched_class, &ext_sched_class));
> +	BUG_ON(!sched_class_above(&ext_sched_class, &idle_sched_class));
> +#endif
>  
>  	wait_bit_init();
>  
> @@ -12046,3 +12081,38 @@ void sched_mm_cid_fork(struct task_struct *t)
>  	t->mm_cid_active = 1;
>  }
>  #endif
> +
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +void sched_deq_and_put_task(struct task_struct *p, int queue_flags,
> +			    struct sched_enq_and_set_ctx *ctx)
> +{
> +	struct rq *rq = task_rq(p);
> +
> +	lockdep_assert_rq_held(rq);
> +
> +	*ctx = (struct sched_enq_and_set_ctx){
> +		.p = p,
> +		.queue_flags = queue_flags,
> +		.queued = task_on_rq_queued(p),
> +		.running = task_current(rq, p),
> +	};
> +
> +	update_rq_clock(rq);
> +	if (ctx->queued)
> +		dequeue_task(rq, p, queue_flags | DEQUEUE_NOCLOCK);
> +	if (ctx->running)
> +		put_prev_task(rq, p);
> +}
> +
> +void sched_enq_and_set_task(struct sched_enq_and_set_ctx *ctx)
> +{
> +	struct rq *rq = task_rq(ctx->p);
> +
> +	lockdep_assert_rq_held(rq);
> +
> +	if (ctx->queued)
> +		enqueue_task(rq, ctx->p, ctx->queue_flags | ENQUEUE_NOCLOCK);
> +	if (ctx->running)
> +		set_next_task(rq, ctx->p);
> +}
> +#endif	/* CONFIG_SCHED_CLASS_EXT */
> diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
> index 0b2340a79b65..79fac9c92a22 100644
> --- a/kernel/sched/debug.c
> +++ b/kernel/sched/debug.c
> @@ -377,6 +377,9 @@ static __init int sched_init_debug(void)
>  
>  	debugfs_create_file("debug", 0444, debugfs_sched, NULL, &sched_debug_fops);
>  
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +	debugfs_create_file("ext", 0444, debugfs_sched, NULL, &sched_ext_fops);
> +#endif
>  	return 0;
>  }
>  late_initcall(sched_init_debug);
> @@ -1093,6 +1096,9 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
>  		P(dl.runtime);
>  		P(dl.deadline);
>  	}
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +	__PS("ext.enabled", task_on_scx(p));
> +#endif
>  #undef PN_SCHEDSTAT
>  #undef P_SCHEDSTAT
>  
> diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
> new file mode 100644
> index 000000000000..51d77459d208
> --- /dev/null
> +++ b/kernel/sched/ext.c
> @@ -0,0 +1,3140 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
> + * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
> + * Copyright (c) 2022 David Vernet <dvernet@meta.com>
> + */
> +#define SCX_OP_IDX(op)		(offsetof(struct sched_ext_ops, op) / sizeof(void (*)(void)))
> +
> +enum scx_internal_consts {
> +	SCX_NR_ONLINE_OPS	= SCX_OP_IDX(init),
> +	SCX_DSP_DFL_MAX_BATCH	= 32,
> +};
> +
> +enum scx_ops_enable_state {
> +	SCX_OPS_PREPPING,
> +	SCX_OPS_ENABLING,
> +	SCX_OPS_ENABLED,
> +	SCX_OPS_DISABLING,
> +	SCX_OPS_DISABLED,
> +};
> +
> +/*
> + * sched_ext_entity->ops_state
> + *
> + * Used to track the task ownership between the SCX core and the BPF scheduler.
> + * State transitions look as follows:
> + *
> + * NONE -> QUEUEING -> QUEUED -> DISPATCHING
> + *   ^              |                 |
> + *   |              v                 v
> + *   \-------------------------------/
> + *
> + * QUEUEING and DISPATCHING states can be waited upon. See wait_ops_state() call
> + * sites for explanations on the conditions being waited upon and why they are
> + * safe. Transitions out of them into NONE or QUEUED must store_release and the
> + * waiters should load_acquire.
> + *
> + * Tracking scx_ops_state enables sched_ext core to reliably determine whether
> + * any given task can be dispatched by the BPF scheduler at all times and thus
> + * relaxes the requirements on the BPF scheduler. This allows the BPF scheduler
> + * to try to dispatch any task anytime regardless of its state as the SCX core
> + * can safely reject invalid dispatches.
> + */
> +enum scx_ops_state {
> +	SCX_OPSS_NONE,		/* owned by the SCX core */
> +	SCX_OPSS_QUEUEING,	/* in transit to the BPF scheduler */
> +	SCX_OPSS_QUEUED,	/* owned by the BPF scheduler */
> +	SCX_OPSS_DISPATCHING,	/* in transit back to the SCX core */
> +
> +	/*
> +	 * QSEQ brands each QUEUED instance so that, when dispatch races
> +	 * dequeue/requeue, the dispatcher can tell whether it still has a claim
> +	 * on the task being dispatched.
> +	 */
> +	SCX_OPSS_QSEQ_SHIFT	= 2,
> +	SCX_OPSS_STATE_MASK	= (1LLU << SCX_OPSS_QSEQ_SHIFT) - 1,
> +	SCX_OPSS_QSEQ_MASK	= ~SCX_OPSS_STATE_MASK,
> +};
> +
> +/*
> + * During exit, a task may schedule after losing its PIDs. When disabling the
> + * BPF scheduler, we need to be able to iterate tasks in every state to
> + * guarantee system safety. Maintain a dedicated task list which contains every
> + * task between its fork and eventual free.
> + */
> +static DEFINE_SPINLOCK(scx_tasks_lock);
> +static LIST_HEAD(scx_tasks);
> +
> +/* ops enable/disable */
> +static struct kthread_worker *scx_ops_helper;
> +static DEFINE_MUTEX(scx_ops_enable_mutex);
> +DEFINE_STATIC_KEY_FALSE(__scx_ops_enabled);
> +DEFINE_STATIC_PERCPU_RWSEM(scx_fork_rwsem);
> +static atomic_t scx_ops_enable_state_var = ATOMIC_INIT(SCX_OPS_DISABLED);
> +static struct sched_ext_ops scx_ops;
> +static bool scx_warned_zero_slice;
> +
> +static DEFINE_STATIC_KEY_FALSE(scx_ops_enq_last);
> +static DEFINE_STATIC_KEY_FALSE(scx_ops_enq_exiting);
> +static DEFINE_STATIC_KEY_FALSE(scx_builtin_idle_enabled);
> +
> +struct static_key_false scx_has_op[SCX_NR_ONLINE_OPS] =
> +	{ [0 ... SCX_NR_ONLINE_OPS-1] = STATIC_KEY_FALSE_INIT };
> +
> +static atomic_t scx_exit_type = ATOMIC_INIT(SCX_EXIT_DONE);
> +static struct scx_exit_info scx_exit_info;
> +
> +static atomic64_t scx_nr_rejected = ATOMIC64_INIT(0);
> +
> +/* idle tracking */
> +#ifdef CONFIG_SMP
> +#ifdef CONFIG_CPUMASK_OFFSTACK
> +#define CL_ALIGNED_IF_ONSTACK
> +#else
> +#define CL_ALIGNED_IF_ONSTACK __cacheline_aligned_in_smp
> +#endif
> +
> +static struct {
> +	cpumask_var_t cpu;
> +	cpumask_var_t smt;
> +} idle_masks CL_ALIGNED_IF_ONSTACK;
> +
> +#endif	/* CONFIG_SMP */
> +
> +/*
> + * Direct dispatch marker.
> + *
> + * Non-NULL values are used for direct dispatch from enqueue path. A valid
> + * pointer points to the task currently being enqueued. An ERR_PTR value is used
> + * to indicate that direct dispatch has already happened.
> + */
> +static DEFINE_PER_CPU(struct task_struct *, direct_dispatch_task);
> +
> +/* dispatch queues */
> +static struct scx_dispatch_q __cacheline_aligned_in_smp scx_dsq_global;
> +
> +static const struct rhashtable_params dsq_hash_params = {
> +	.key_len		= 8,
> +	.key_offset		= offsetof(struct scx_dispatch_q, id),
> +	.head_offset		= offsetof(struct scx_dispatch_q, hash_node),
> +};
> +
> +static struct rhashtable dsq_hash;
> +static LLIST_HEAD(dsqs_to_free);
> +
> +/* dispatch buf */
> +struct scx_dsp_buf_ent {
> +	struct task_struct	*task;
> +	u64			qseq;
> +	u64			dsq_id;
> +	u64			enq_flags;
> +};
> +
> +static u32 scx_dsp_max_batch;
> +static struct scx_dsp_buf_ent __percpu *scx_dsp_buf;
> +
> +struct scx_dsp_ctx {
> +	struct rq		*rq;
> +	struct rq_flags		*rf;
> +	u32			buf_cursor;
> +	u32			nr_tasks;
> +};
> +
> +static DEFINE_PER_CPU(struct scx_dsp_ctx, scx_dsp_ctx);
> +
> +void scx_bpf_dispatch(struct task_struct *p, u64 dsq_id, u64 slice,
> +		      u64 enq_flags);
> +__printf(2, 3) static void scx_ops_error_type(enum scx_exit_type type,
> +					      const char *fmt, ...);
> +#define scx_ops_error(fmt, args...)						\
> +	scx_ops_error_type(SCX_EXIT_ERROR, fmt, ##args)
> +
> +struct scx_task_iter {
> +	struct sched_ext_entity		cursor;
> +	struct task_struct		*locked;
> +	struct rq			*rq;
> +	struct rq_flags			rf;
> +};
> +
> +#define SCX_HAS_OP(op)	static_branch_likely(&scx_has_op[SCX_OP_IDX(op)])
> +
> +/* if the highest set bit is N, return a mask with bits [N+1, 31] set */
> +static u32 higher_bits(u32 flags)
> +{
> +	return ~((1 << fls(flags)) - 1);
> +}
> +
> +/* return the mask with only the highest bit set */
> +static u32 highest_bit(u32 flags)
> +{
> +	int bit = fls(flags);
> +	return bit ? 1 << (bit - 1) : 0;
> +}
> +
> +/*
> + * scx_kf_mask enforcement. Some kfuncs can only be called from specific SCX
> + * ops. When invoking SCX ops, SCX_CALL_OP[_RET]() should be used to indicate
> + * the allowed kfuncs and those kfuncs should use scx_kf_allowed() to check
> + * whether it's running from an allowed context.
> + *
> + * @mask is constant, always inline to cull the mask calculations.
> + */
> +static __always_inline void scx_kf_allow(u32 mask)
> +{
> +	/* nesting is allowed only in increasing scx_kf_mask order */
> +	WARN_ONCE((mask | higher_bits(mask)) & current->scx.kf_mask,
> +		  "invalid nesting current->scx.kf_mask=0x%x mask=0x%x\n",
> +		  current->scx.kf_mask, mask);
> +	current->scx.kf_mask |= mask;
> +}
> +
> +static void scx_kf_disallow(u32 mask)
> +{
> +	current->scx.kf_mask &= ~mask;
> +}
> +
> +#define SCX_CALL_OP(mask, op, args...)						\
> +do {										\
> +	if (mask) {								\
> +		scx_kf_allow(mask);						\
> +		scx_ops.op(args);						\
> +		scx_kf_disallow(mask);						\
> +	} else {								\
> +		scx_ops.op(args);						\
> +	}									\
> +} while (0)
> +
> +#define SCX_CALL_OP_RET(mask, op, args...)					\
> +({										\
> +	__typeof__(scx_ops.op(args)) __ret;					\
> +	if (mask) {								\
> +		scx_kf_allow(mask);						\
> +		__ret = scx_ops.op(args);					\
> +		scx_kf_disallow(mask);						\
> +	} else {								\
> +		__ret = scx_ops.op(args);					\
> +	}									\
> +	__ret;									\
> +})
> +
> +/* @mask is constant, always inline to cull unnecessary branches */
> +static __always_inline bool scx_kf_allowed(u32 mask)
> +{
> +	if (unlikely(!(current->scx.kf_mask & mask))) {
> +		scx_ops_error("kfunc with mask 0x%x called from an operation only allowing 0x%x",
> +			      mask, current->scx.kf_mask);
> +		return false;
> +	}
> +
> +	if (unlikely((mask & (SCX_KF_INIT | SCX_KF_SLEEPABLE)) &&
> +		     in_interrupt())) {
> +		scx_ops_error("sleepable kfunc called from non-sleepable context");
> +		return false;
> +	}
> +
> +	/*
> +	 * Enforce nesting boundaries. e.g. A kfunc which can be called from
> +	 * DISPATCH must not be called if we're running DEQUEUE which is nested
> +	 * inside ops.dispatch(). We don't need to check the SCX_KF_SLEEPABLE
> +	 * boundary thanks to the above in_interrupt() check.
> +	 */
> +	if (unlikely(highest_bit(mask) == SCX_KF_DISPATCH &&
> +		     (current->scx.kf_mask & higher_bits(SCX_KF_DISPATCH)))) {
> +		scx_ops_error("dispatch kfunc called from a nested operation");
> +		return false;
> +	}
> +
> +	return true;
> +}
> +
> +/**
> + * scx_task_iter_init - Initialize a task iterator
> + * @iter: iterator to init
> + *
> + * Initialize @iter. Must be called with scx_tasks_lock held. Once initialized,
> + * @iter must eventually be exited with scx_task_iter_exit().
> + *
> + * scx_tasks_lock may be released between this and the first next() call or
> + * between any two next() calls. If scx_tasks_lock is released between two
> + * next() calls, the caller is responsible for ensuring that the task being
> + * iterated remains accessible either through RCU read lock or obtaining a
> + * reference count.
> + *
> + * All tasks which existed when the iteration started are guaranteed to be
> + * visited as long as they still exist.
> + */
> +static void scx_task_iter_init(struct scx_task_iter *iter)
> +{
> +	lockdep_assert_held(&scx_tasks_lock);
> +
> +	iter->cursor = (struct sched_ext_entity){ .flags = SCX_TASK_CURSOR };
> +	list_add(&iter->cursor.tasks_node, &scx_tasks);
> +	iter->locked = NULL;
> +}
> +
> +/**
> + * scx_task_iter_exit - Exit a task iterator
> + * @iter: iterator to exit
> + *
> + * Exit a previously initialized @iter. Must be called with scx_tasks_lock held.
> + * If the iterator holds a task's rq lock, that rq lock is released. See
> + * scx_task_iter_init() for details.
> + */
> +static void scx_task_iter_exit(struct scx_task_iter *iter)
> +{
> +	struct list_head *cursor = &iter->cursor.tasks_node;
> +
> +	lockdep_assert_held(&scx_tasks_lock);
> +
> +	if (iter->locked) {
> +		task_rq_unlock(iter->rq, iter->locked, &iter->rf);
> +		iter->locked = NULL;
> +	}
> +
> +	if (list_empty(cursor))
> +		return;
> +
> +	list_del_init(cursor);
> +}
> +
> +/**
> + * scx_task_iter_next - Next task
> + * @iter: iterator to walk
> + *
> + * Visit the next task. See scx_task_iter_init() for details.
> + */
> +static struct task_struct *scx_task_iter_next(struct scx_task_iter *iter)
> +{
> +	struct list_head *cursor = &iter->cursor.tasks_node;
> +	struct sched_ext_entity *pos;
> +
> +	lockdep_assert_held(&scx_tasks_lock);
> +
> +	list_for_each_entry(pos, cursor, tasks_node) {
> +		if (&pos->tasks_node == &scx_tasks)
> +			return NULL;
> +		if (!(pos->flags & SCX_TASK_CURSOR)) {
> +			list_move(cursor, &pos->tasks_node);
> +			return container_of(pos, struct task_struct, scx);
> +		}
> +	}
> +
> +	/* can't happen, should always terminate at scx_tasks above */
> +	BUG();
> +}
> +
> +/**
> + * scx_task_iter_next_filtered - Next non-idle task
> + * @iter: iterator to walk
> + *
> + * Visit the next non-idle task. See scx_task_iter_init() for details.
> + */
> +static struct task_struct *
> +scx_task_iter_next_filtered(struct scx_task_iter *iter)
> +{
> +	struct task_struct *p;
> +
> +	while ((p = scx_task_iter_next(iter))) {
> +		if (!is_idle_task(p))
> +			return p;
> +	}
> +	return NULL;
> +}
> +
> +/**
> + * scx_task_iter_next_filtered_locked - Next non-idle task with its rq locked
> + * @iter: iterator to walk
> + *
> + * Visit the next non-idle task with its rq lock held. See scx_task_iter_init()
> + * for details.
> + */
> +static struct task_struct *
> +scx_task_iter_next_filtered_locked(struct scx_task_iter *iter)
> +{
> +	struct task_struct *p;
> +
> +	if (iter->locked) {
> +		task_rq_unlock(iter->rq, iter->locked, &iter->rf);
> +		iter->locked = NULL;
> +	}
> +
> +	p = scx_task_iter_next_filtered(iter);
> +	if (!p)
> +		return NULL;
> +
> +	iter->rq = task_rq_lock(p, &iter->rf);
> +	iter->locked = p;
> +	return p;
> +}
> +
> +static enum scx_ops_enable_state scx_ops_enable_state(void)
> +{
> +	return atomic_read(&scx_ops_enable_state_var);
> +}
> +
> +static enum scx_ops_enable_state
> +scx_ops_set_enable_state(enum scx_ops_enable_state to)
> +{
> +	return atomic_xchg(&scx_ops_enable_state_var, to);
> +}
> +
> +static bool scx_ops_tryset_enable_state(enum scx_ops_enable_state to,
> +					enum scx_ops_enable_state from)
> +{
> +	int from_v = from;
> +
> +	return atomic_try_cmpxchg(&scx_ops_enable_state_var, &from_v, to);
> +}
> +
> +static bool scx_ops_disabling(void)
> +{
> +	return unlikely(scx_ops_enable_state() == SCX_OPS_DISABLING);
> +}
> +
> +/**
> + * wait_ops_state - Busy-wait the specified ops state to end
> + * @p: target task
> + * @opss: state to wait the end of
> + *
> + * Busy-wait for @p to transition out of @opss. This can only be used when the
> + * state part of @opss is %SCX_QUEUEING or %SCX_DISPATCHING. This function also
> + * has load_acquire semantics to ensure that the caller can see the updates made
> + * in the enqueueing and dispatching paths.
> + */
> +static void wait_ops_state(struct task_struct *p, u64 opss)
> +{
> +	do {
> +		cpu_relax();
> +	} while (atomic64_read_acquire(&p->scx.ops_state) == opss);
> +}
> +
> +/**
> + * ops_cpu_valid - Verify a cpu number
> + * @cpu: cpu number which came from a BPF ops
> + *
> + * @cpu is a cpu number which came from the BPF scheduler and can be any value.
> + * Verify that it is in range and one of the possible cpus.
> + */
> +static bool ops_cpu_valid(s32 cpu)
> +{
> +	return likely(cpu >= 0 && cpu < nr_cpu_ids && cpu_possible(cpu));
> +}
> +
> +/**
> + * ops_sanitize_err - Sanitize a -errno value
> + * @ops_name: operation to blame on failure
> + * @err: -errno value to sanitize
> + *
> + * Verify @err is a valid -errno. If not, trigger scx_ops_error() and return
> + * -%EPROTO. This is necessary because returning a rogue -errno up the chain can
> + * cause misbehaviors. For an example, a large negative return from
> + * ops.prep_enable() triggers an oops when passed up the call chain because the
> + * value fails IS_ERR() test after being encoded with ERR_PTR() and then is
> + * handled as a pointer.
> + */
> +static int ops_sanitize_err(const char *ops_name, s32 err)
> +{
> +	if (err < 0 && err >= -MAX_ERRNO)
> +		return err;
> +
> +	scx_ops_error("ops.%s() returned an invalid errno %d", ops_name, err);
> +	return -EPROTO;
> +}
> +
> +static void update_curr_scx(struct rq *rq)
> +{
> +	struct task_struct *curr = rq->curr;
> +	u64 now = rq_clock_task(rq);
> +	u64 delta_exec;
> +
> +	if (time_before_eq64(now, curr->se.exec_start))
> +		return;
> +
> +	delta_exec = now - curr->se.exec_start;
> +	curr->se.exec_start = now;
> +	curr->se.sum_exec_runtime += delta_exec;
> +	account_group_exec_runtime(curr, delta_exec);
> +	cgroup_account_cputime(curr, delta_exec);
> +
> +	curr->scx.slice -= min(curr->scx.slice, delta_exec);
> +}
> +
> +static void dispatch_enqueue(struct scx_dispatch_q *dsq, struct task_struct *p,
> +			     u64 enq_flags)
> +{
> +	bool is_local = dsq->id == SCX_DSQ_LOCAL;
> +
> +	WARN_ON_ONCE(p->scx.dsq || !list_empty(&p->scx.dsq_node));
> +
> +	if (!is_local) {
> +		raw_spin_lock(&dsq->lock);
> +		if (unlikely(dsq->id == SCX_DSQ_INVALID)) {
> +			scx_ops_error("attempting to dispatch to a destroyed dsq");
> +			/* fall back to the global dsq */
> +			raw_spin_unlock(&dsq->lock);
> +			dsq = &scx_dsq_global;
> +			raw_spin_lock(&dsq->lock);
> +		}
> +	}
> +
> +	if (enq_flags & SCX_ENQ_HEAD)
> +		list_add(&p->scx.dsq_node, &dsq->fifo);
> +	else
> +		list_add_tail(&p->scx.dsq_node, &dsq->fifo);
> +	dsq->nr++;
> +	p->scx.dsq = dsq;
> +
> +	/*
> +	 * We're transitioning out of QUEUEING or DISPATCHING. store_release to
> +	 * match waiters' load_acquire.
> +	 */
> +	if (enq_flags & SCX_ENQ_CLEAR_OPSS)
> +		atomic64_set_release(&p->scx.ops_state, SCX_OPSS_NONE);
> +
> +	if (is_local) {
> +		struct rq *rq = container_of(dsq, struct rq, scx.local_dsq);
> +
> +		if (sched_class_above(&ext_sched_class, rq->curr->sched_class))
> +			resched_curr(rq);
> +	} else {
> +		raw_spin_unlock(&dsq->lock);
> +	}
> +}
> +
> +static void dispatch_dequeue(struct scx_rq *scx_rq, struct task_struct *p)
> +{
> +	struct scx_dispatch_q *dsq = p->scx.dsq;
> +	bool is_local = dsq == &scx_rq->local_dsq;
> +
> +	if (!dsq) {
> +		WARN_ON_ONCE(!list_empty(&p->scx.dsq_node));
> +		/*
> +		 * When dispatching directly from the BPF scheduler to a local
> +		 * DSQ, the task isn't associated with any DSQ but
> +		 * @p->scx.holding_cpu may be set under the protection of
> +		 * %SCX_OPSS_DISPATCHING.
> +		 */
> +		if (p->scx.holding_cpu >= 0)
> +			p->scx.holding_cpu = -1;
> +		return;
> +	}
> +
> +	if (!is_local)
> +		raw_spin_lock(&dsq->lock);
> +
> +	/*
> +	 * Now that we hold @dsq->lock, @p->holding_cpu and @p->scx.dsq_node
> +	 * can't change underneath us.
> +	*/
> +	if (p->scx.holding_cpu < 0) {
> +		/* @p must still be on @dsq, dequeue */
> +		WARN_ON_ONCE(list_empty(&p->scx.dsq_node));
> +		list_del_init(&p->scx.dsq_node);
> +		dsq->nr--;
> +	} else {
> +		/*
> +		 * We're racing against dispatch_to_local_dsq() which already
> +		 * removed @p from @dsq and set @p->scx.holding_cpu. Clear the
> +		 * holding_cpu which tells dispatch_to_local_dsq() that it lost
> +		 * the race.
> +		 */
> +		WARN_ON_ONCE(!list_empty(&p->scx.dsq_node));
> +		p->scx.holding_cpu = -1;
> +	}
> +	p->scx.dsq = NULL;
> +
> +	if (!is_local)
> +		raw_spin_unlock(&dsq->lock);
> +}
> +
> +static struct scx_dispatch_q *find_non_local_dsq(u64 dsq_id)
> +{
> +	lockdep_assert(rcu_read_lock_any_held());
> +
> +	if (dsq_id == SCX_DSQ_GLOBAL)
> +		return &scx_dsq_global;
> +	else
> +		return rhashtable_lookup_fast(&dsq_hash, &dsq_id,
> +					      dsq_hash_params);
> +}
> +
> +static struct scx_dispatch_q *find_dsq_for_dispatch(struct rq *rq, u64 dsq_id,
> +						    struct task_struct *p)
> +{
> +	struct scx_dispatch_q *dsq;
> +
> +	if (dsq_id == SCX_DSQ_LOCAL)
> +		return &rq->scx.local_dsq;
> +
> +	dsq = find_non_local_dsq(dsq_id);
> +	if (unlikely(!dsq)) {
> +		scx_ops_error("non-existent DSQ 0x%llx for %s[%d]",
> +			      dsq_id, p->comm, p->pid);
> +		return &scx_dsq_global;
> +	}
> +
> +	return dsq;
> +}
> +
> +static void direct_dispatch(struct task_struct *ddsp_task, struct task_struct *p,
> +			    u64 dsq_id, u64 enq_flags)
> +{
> +	struct scx_dispatch_q *dsq;
> +
> +	/* @p must match the task which is being enqueued */
> +	if (unlikely(p != ddsp_task)) {
> +		if (IS_ERR(ddsp_task))
> +			scx_ops_error("%s[%d] already direct-dispatched",
> +				      p->comm, p->pid);
> +		else
> +			scx_ops_error("enqueueing %s[%d] but trying to direct-dispatch %s[%d]",
> +				      ddsp_task->comm, ddsp_task->pid,
> +				      p->comm, p->pid);
> +		return;
> +	}
> +
> +	/*
> +	 * %SCX_DSQ_LOCAL_ON is not supported during direct dispatch because
> +	 * dispatching to the local DSQ of a different CPU requires unlocking
> +	 * the current rq which isn't allowed in the enqueue path. Use
> +	 * ops.select_cpu() to be on the target CPU and then %SCX_DSQ_LOCAL.
> +	 */
> +	if (unlikely((dsq_id & SCX_DSQ_LOCAL_ON) == SCX_DSQ_LOCAL_ON)) {
> +		scx_ops_error("SCX_DSQ_LOCAL_ON can't be used for direct-dispatch");
> +		return;
> +	}
> +
> +	dsq = find_dsq_for_dispatch(task_rq(p), dsq_id, p);
> +	dispatch_enqueue(dsq, p, enq_flags | SCX_ENQ_CLEAR_OPSS);
> +
> +	/*
> +	 * Mark that dispatch already happened by spoiling direct_dispatch_task
> +	 * with a non-NULL value which can never match a valid task pointer.
> +	 */
> +	__this_cpu_write(direct_dispatch_task, ERR_PTR(-ESRCH));
> +}
> +
> +static bool test_rq_online(struct rq *rq)
> +{
> +#ifdef CONFIG_SMP
> +	return rq->online;
> +#else
> +	return true;
> +#endif
> +}
> +
> +static void do_enqueue_task(struct rq *rq, struct task_struct *p, u64 enq_flags,
> +			    int sticky_cpu)
> +{
> +	struct task_struct **ddsp_taskp;
> +	u64 qseq;
> +
> +	WARN_ON_ONCE(!(p->scx.flags & SCX_TASK_QUEUED));
> +
> +	if (p->scx.flags & SCX_TASK_ENQ_LOCAL) {
> +		enq_flags |= SCX_ENQ_LOCAL;
> +		p->scx.flags &= ~SCX_TASK_ENQ_LOCAL;
> +	}
> +
> +	/* rq migration */
> +	if (sticky_cpu == cpu_of(rq))
> +		goto local_norefill;
> +
> +	/*
> +	 * If !rq->online, we already told the BPF scheduler that the CPU is
> +	 * offline. We're just trying to on/offline the CPU. Don't bother the
> +	 * BPF scheduler.
> +	 */
> +	if (unlikely(!test_rq_online(rq)))
> +		goto local;
> +
> +	/* see %SCX_OPS_ENQ_EXITING */
> +	if (!static_branch_unlikely(&scx_ops_enq_exiting) &&
> +	    unlikely(p->flags & PF_EXITING))
> +		goto local;
> +
> +	/* see %SCX_OPS_ENQ_LAST */
> +	if (!static_branch_unlikely(&scx_ops_enq_last) &&
> +	    (enq_flags & SCX_ENQ_LAST))
> +		goto local;
> +
> +	if (!SCX_HAS_OP(enqueue)) {
> +		if (enq_flags & SCX_ENQ_LOCAL)
> +			goto local;
> +		else
> +			goto global;
> +	}
> +
> +	/* DSQ bypass didn't trigger, enqueue on the BPF scheduler */
> +	qseq = rq->scx.ops_qseq++ << SCX_OPSS_QSEQ_SHIFT;
> +
> +	WARN_ON_ONCE(atomic64_read(&p->scx.ops_state) != SCX_OPSS_NONE);
> +	atomic64_set(&p->scx.ops_state, SCX_OPSS_QUEUEING | qseq);
> +
> +	ddsp_taskp = this_cpu_ptr(&direct_dispatch_task);
> +	WARN_ON_ONCE(*ddsp_taskp);
> +	*ddsp_taskp = p;
> +
> +	SCX_CALL_OP(SCX_KF_ENQUEUE, enqueue, p, enq_flags);
> +
> +	/*
> +	 * If not directly dispatched, QUEUEING isn't clear yet and dispatch or
> +	 * dequeue may be waiting. The store_release matches their load_acquire.
> +	 */
> +	if (*ddsp_taskp == p)
> +		atomic64_set_release(&p->scx.ops_state, SCX_OPSS_QUEUED | qseq);
> +	*ddsp_taskp = NULL;
> +	return;
> +
> +local:
> +	p->scx.slice = SCX_SLICE_DFL;
> +local_norefill:
> +	dispatch_enqueue(&rq->scx.local_dsq, p, enq_flags);
> +	return;
> +
> +global:
> +	p->scx.slice = SCX_SLICE_DFL;
> +	dispatch_enqueue(&scx_dsq_global, p, enq_flags);
> +}
> +
> +static void enqueue_task_scx(struct rq *rq, struct task_struct *p, int enq_flags)
> +{
> +	int sticky_cpu = p->scx.sticky_cpu;
> +
> +	enq_flags |= rq->scx.extra_enq_flags;
> +
> +	if (sticky_cpu >= 0)
> +		p->scx.sticky_cpu = -1;
> +
> +	/*
> +	 * Restoring a running task will be immediately followed by
> +	 * set_next_task_scx() which expects the task to not be on the BPF
> +	 * scheduler as tasks can only start running through local DSQs. Force
> +	 * direct-dispatch into the local DSQ by setting the sticky_cpu.
> +	 */
> +	if (unlikely(enq_flags & ENQUEUE_RESTORE) && task_current(rq, p))
> +		sticky_cpu = cpu_of(rq);
> +
> +	if (p->scx.flags & SCX_TASK_QUEUED)
> +		return;
> +
> +	p->scx.flags |= SCX_TASK_QUEUED;
> +	rq->scx.nr_running++;
> +	add_nr_running(rq, 1);
> +
> +	do_enqueue_task(rq, p, enq_flags, sticky_cpu);
> +}
> +
> +static void ops_dequeue(struct task_struct *p, u64 deq_flags)
> +{
> +	u64 opss;
> +
> +	/* acquire ensures that we see the preceding updates on QUEUED */
> +	opss = atomic64_read_acquire(&p->scx.ops_state);
> +
> +	switch (opss & SCX_OPSS_STATE_MASK) {
> +	case SCX_OPSS_NONE:
> +		break;
> +	case SCX_OPSS_QUEUEING:
> +		/*
> +		 * QUEUEING is started and finished while holding @p's rq lock.
> +		 * As we're holding the rq lock now, we shouldn't see QUEUEING.
> +		 */
> +		BUG();
> +	case SCX_OPSS_QUEUED:
> +		if (SCX_HAS_OP(dequeue))
> +			SCX_CALL_OP(SCX_KF_REST, dequeue, p, deq_flags);
> +
> +		if (atomic64_try_cmpxchg(&p->scx.ops_state, &opss,
> +					 SCX_OPSS_NONE))
> +			break;
> +		fallthrough;
> +	case SCX_OPSS_DISPATCHING:
> +		/*
> +		 * If @p is being dispatched from the BPF scheduler to a DSQ,
> +		 * wait for the transfer to complete so that @p doesn't get
> +		 * added to its DSQ after dequeueing is complete.
> +		 *
> +		 * As we're waiting on DISPATCHING with the rq locked, the
> +		 * dispatching side shouldn't try to lock the rq while
> +		 * DISPATCHING is set. See dispatch_to_local_dsq().
> +		 *
> +		 * DISPATCHING shouldn't have qseq set and control can reach
> +		 * here with NONE @opss from the above QUEUED case block.
> +		 * Explicitly wait on %SCX_OPSS_DISPATCHING instead of @opss.
> +		 */
> +		wait_ops_state(p, SCX_OPSS_DISPATCHING);
> +		BUG_ON(atomic64_read(&p->scx.ops_state) != SCX_OPSS_NONE);
> +		break;
> +	}
> +}
> +
> +static void dequeue_task_scx(struct rq *rq, struct task_struct *p, int deq_flags)
> +{
> +	struct scx_rq *scx_rq = &rq->scx;
> +
> +	if (!(p->scx.flags & SCX_TASK_QUEUED))
> +		return;
> +
> +	ops_dequeue(p, deq_flags);
> +
> +	if (deq_flags & SCX_DEQ_SLEEP)
> +		p->scx.flags |= SCX_TASK_DEQD_FOR_SLEEP;
> +	else
> +		p->scx.flags &= ~SCX_TASK_DEQD_FOR_SLEEP;
> +
> +	p->scx.flags &= ~SCX_TASK_QUEUED;
> +	scx_rq->nr_running--;
> +	sub_nr_running(rq, 1);
> +
> +	dispatch_dequeue(scx_rq, p);
> +}
> +
> +static void yield_task_scx(struct rq *rq)
> +{
> +	struct task_struct *p = rq->curr;
> +
> +	if (SCX_HAS_OP(yield))
> +		SCX_CALL_OP_RET(SCX_KF_REST, yield, p, NULL);
> +	else
> +		p->scx.slice = 0;
> +}
> +
> +static bool yield_to_task_scx(struct rq *rq, struct task_struct *to)
> +{
> +	struct task_struct *from = rq->curr;
> +
> +	if (SCX_HAS_OP(yield))
> +		return SCX_CALL_OP_RET(SCX_KF_REST, yield, from, to);
> +	else
> +		return false;
> +}
> +
> +#ifdef CONFIG_SMP
> +/**
> + * move_task_to_local_dsq - Move a task from a different rq to a local DSQ
> + * @rq: rq to move the task into, currently locked
> + * @p: task to move
> + * @enq_flags: %SCX_ENQ_*
> + *
> + * Move @p which is currently on a different rq to @rq's local DSQ. The caller
> + * must:
> + *
> + * 1. Start with exclusive access to @p either through its DSQ lock or
> + *    %SCX_OPSS_DISPATCHING flag.
> + *
> + * 2. Set @p->scx.holding_cpu to raw_smp_processor_id().
> + *
> + * 3. Remember task_rq(@p). Release the exclusive access so that we don't
> + *    deadlock with dequeue.
> + *
> + * 4. Lock @rq and the task_rq from #3.
> + *
> + * 5. Call this function.
> + *
> + * Returns %true if @p was successfully moved. %false after racing dequeue and
> + * losing.
> + */
> +static bool move_task_to_local_dsq(struct rq *rq, struct task_struct *p,
> +				   u64 enq_flags)
> +{
> +	struct rq *task_rq;
> +
> +	lockdep_assert_rq_held(rq);
> +
> +	/*
> +	 * If dequeue got to @p while we were trying to lock both rq's, it'd
> +	 * have cleared @p->scx.holding_cpu to -1. While other cpus may have
> +	 * updated it to different values afterwards, as this operation can't be
> +	 * preempted or recurse, @p->scx.holding_cpu can never become
> +	 * raw_smp_processor_id() again before we're done. Thus, we can tell
> +	 * whether we lost to dequeue by testing whether @p->scx.holding_cpu is
> +	 * still raw_smp_processor_id().
> +	 *
> +	 * See dispatch_dequeue() for the counterpart.
> +	 */
> +	if (unlikely(p->scx.holding_cpu != raw_smp_processor_id()))
> +		return false;
> +
> +	/* @p->rq couldn't have changed if we're still the holding cpu */
> +	task_rq = task_rq(p);
> +	lockdep_assert_rq_held(task_rq);
> +
> +	WARN_ON_ONCE(!cpumask_test_cpu(cpu_of(rq), p->cpus_ptr));
> +	deactivate_task(task_rq, p, 0);
> +	set_task_cpu(p, cpu_of(rq));
> +	p->scx.sticky_cpu = cpu_of(rq);
> +
> +	/*
> +	 * We want to pass scx-specific enq_flags but activate_task() will
> +	 * truncate the upper 32 bit. As we own @rq, we can pass them through
> +	 * @rq->scx.extra_enq_flags instead.
> +	 */
> +	WARN_ON_ONCE(rq->scx.extra_enq_flags);
> +	rq->scx.extra_enq_flags = enq_flags;
> +	activate_task(rq, p, 0);
> +	rq->scx.extra_enq_flags = 0;
> +
> +	return true;
> +}
> +
> +/**
> + * dispatch_to_local_dsq_lock - Ensure source and desitnation rq's are locked
> + * @rq: current rq which is locked
> + * @rf: rq_flags to use when unlocking @rq
> + * @src_rq: rq to move task from
> + * @dst_rq: rq to move task to
> + *
> + * We're holding @rq lock and trying to dispatch a task from @src_rq to
> + * @dst_rq's local DSQ and thus need to lock both @src_rq and @dst_rq. Whether
> + * @rq stays locked isn't important as long as the state is restored after
> + * dispatch_to_local_dsq_unlock().
> + */
> +static void dispatch_to_local_dsq_lock(struct rq *rq, struct rq_flags *rf,
> +				       struct rq *src_rq, struct rq *dst_rq)
> +{
> +	rq_unpin_lock(rq, rf);
> +
> +	if (src_rq == dst_rq) {
> +		raw_spin_rq_unlock(rq);
> +		raw_spin_rq_lock(dst_rq);
> +	} else if (rq == src_rq) {
> +		double_lock_balance(rq, dst_rq);
> +		rq_repin_lock(rq, rf);
> +	} else if (rq == dst_rq) {
> +		double_lock_balance(rq, src_rq);
> +		rq_repin_lock(rq, rf);
> +	} else {
> +		raw_spin_rq_unlock(rq);
> +		double_rq_lock(src_rq, dst_rq);
> +	}
> +}
> +
> +/**
> + * dispatch_to_local_dsq_unlock - Undo dispatch_to_local_dsq_lock()
> + * @rq: current rq which is locked
> + * @rf: rq_flags to use when unlocking @rq
> + * @src_rq: rq to move task from
> + * @dst_rq: rq to move task to
> + *
> + * Unlock @src_rq and @dst_rq and ensure that @rq is locked on return.
> + */
> +static void dispatch_to_local_dsq_unlock(struct rq *rq, struct rq_flags *rf,
> +					 struct rq *src_rq, struct rq *dst_rq)
> +{
> +	if (src_rq == dst_rq) {
> +		raw_spin_rq_unlock(dst_rq);
> +		raw_spin_rq_lock(rq);
> +		rq_repin_lock(rq, rf);
> +	} else if (rq == src_rq) {
> +		double_unlock_balance(rq, dst_rq);
> +	} else if (rq == dst_rq) {
> +		double_unlock_balance(rq, src_rq);
> +	} else {
> +		double_rq_unlock(src_rq, dst_rq);
> +		raw_spin_rq_lock(rq);
> +		rq_repin_lock(rq, rf);
> +	}
> +}
> +#endif	/* CONFIG_SMP */
> +
> +
> +static bool consume_dispatch_q(struct rq *rq, struct rq_flags *rf,
> +			       struct scx_dispatch_q *dsq)
> +{
> +	struct scx_rq *scx_rq = &rq->scx;
> +	struct task_struct *p;
> +	struct rq *task_rq;
> +	bool moved = false;
> +retry:
> +	if (list_empty(&dsq->fifo))
> +		return false;
> +
> +	raw_spin_lock(&dsq->lock);
> +	list_for_each_entry(p, &dsq->fifo, scx.dsq_node) {
> +		task_rq = task_rq(p);
> +		if (rq == task_rq)
> +			goto this_rq;
> +		if (likely(test_rq_online(rq)) && !is_migration_disabled(p) &&
> +		    cpumask_test_cpu(cpu_of(rq), p->cpus_ptr))
> +			goto remote_rq;
> +	}
> +	raw_spin_unlock(&dsq->lock);
> +	return false;
> +
> +this_rq:
> +	/* @dsq is locked and @p is on this rq */
> +	WARN_ON_ONCE(p->scx.holding_cpu >= 0);
> +	list_move_tail(&p->scx.dsq_node, &scx_rq->local_dsq.fifo);
> +	dsq->nr--;
> +	scx_rq->local_dsq.nr++;
> +	p->scx.dsq = &scx_rq->local_dsq;
> +	raw_spin_unlock(&dsq->lock);
> +	return true;
> +
> +remote_rq:
> +#ifdef CONFIG_SMP
> +	/*
> +	 * @dsq is locked and @p is on a remote rq. @p is currently protected by
> +	 * @dsq->lock. We want to pull @p to @rq but may deadlock if we grab
> +	 * @task_rq while holding @dsq and @rq locks. As dequeue can't drop the
> +	 * rq lock or fail, do a little dancing from our side. See
> +	 * move_task_to_local_dsq().
> +	 */
> +	WARN_ON_ONCE(p->scx.holding_cpu >= 0);
> +	list_del_init(&p->scx.dsq_node);
> +	dsq->nr--;
> +	p->scx.holding_cpu = raw_smp_processor_id();
> +	raw_spin_unlock(&dsq->lock);
> +
> +	rq_unpin_lock(rq, rf);
> +	double_lock_balance(rq, task_rq);
> +	rq_repin_lock(rq, rf);
> +
> +	moved = move_task_to_local_dsq(rq, p, 0);
> +
> +	double_unlock_balance(rq, task_rq);
> +#endif /* CONFIG_SMP */
> +	if (likely(moved))
> +		return true;
> +	goto retry;
> +}
> +
> +enum dispatch_to_local_dsq_ret {
> +	DTL_DISPATCHED,		/* successfully dispatched */
> +	DTL_LOST,		/* lost race to dequeue */
> +	DTL_NOT_LOCAL,		/* destination is not a local DSQ */
> +	DTL_INVALID,		/* invalid local dsq_id */
> +};
> +
> +/**
> + * dispatch_to_local_dsq - Dispatch a task to a local dsq
> + * @rq: current rq which is locked
> + * @rf: rq_flags to use when unlocking @rq
> + * @dsq_id: destination dsq ID
> + * @p: task to dispatch
> + * @enq_flags: %SCX_ENQ_*
> + *
> + * We're holding @rq lock and want to dispatch @p to the local DSQ identified by
> + * @dsq_id. This function performs all the synchronization dancing needed
> + * because local DSQs are protected with rq locks.
> + *
> + * The caller must have exclusive ownership of @p (e.g. through
> + * %SCX_OPSS_DISPATCHING).
> + */
> +static enum dispatch_to_local_dsq_ret
> +dispatch_to_local_dsq(struct rq *rq, struct rq_flags *rf, u64 dsq_id,
> +		      struct task_struct *p, u64 enq_flags)
> +{
> +	struct rq *src_rq = task_rq(p);
> +	struct rq *dst_rq;
> +
> +	/*
> +	 * We're synchronized against dequeue through DISPATCHING. As @p can't
> +	 * be dequeued, its task_rq and cpus_allowed are stable too.
> +	 */
> +	if (dsq_id == SCX_DSQ_LOCAL) {
> +		dst_rq = rq;
> +	} else if ((dsq_id & SCX_DSQ_LOCAL_ON) == SCX_DSQ_LOCAL_ON) {
> +		s32 cpu = dsq_id & SCX_DSQ_LOCAL_CPU_MASK;
> +
> +		if (!ops_cpu_valid(cpu)) {
> +			scx_ops_error("invalid cpu %d in SCX_DSQ_LOCAL_ON verdict for %s[%d]",
> +				      cpu, p->comm, p->pid);
> +			return DTL_INVALID;
> +		}
> +		dst_rq = cpu_rq(cpu);
> +	} else {
> +		return DTL_NOT_LOCAL;
> +	}
> +
> +	/* if dispatching to @rq that @p is already on, no lock dancing needed */
> +	if (rq == src_rq && rq == dst_rq) {
> +		dispatch_enqueue(&dst_rq->scx.local_dsq, p,
> +				 enq_flags | SCX_ENQ_CLEAR_OPSS);
> +		return DTL_DISPATCHED;
> +	}
> +
> +#ifdef CONFIG_SMP
> +	if (cpumask_test_cpu(cpu_of(dst_rq), p->cpus_ptr)) {
> +		struct rq *locked_dst_rq = dst_rq;
> +		bool dsp;
> +
> +		/*
> +		 * @p is on a possibly remote @src_rq which we need to lock to
> +		 * move the task. If dequeue is in progress, it'd be locking
> +		 * @src_rq and waiting on DISPATCHING, so we can't grab @src_rq
> +		 * lock while holding DISPATCHING.
> +		 *
> +		 * As DISPATCHING guarantees that @p is wholly ours, we can
> +		 * pretend that we're moving from a DSQ and use the same
> +		 * mechanism - mark the task under transfer with holding_cpu,
> +		 * release DISPATCHING and then follow the same protocol.
> +		 */
> +		p->scx.holding_cpu = raw_smp_processor_id();
> +
> +		/* store_release ensures that dequeue sees the above */
> +		atomic64_set_release(&p->scx.ops_state, SCX_OPSS_NONE);
> +
> +		dispatch_to_local_dsq_lock(rq, rf, src_rq, locked_dst_rq);
> +
> +		/*
> +		 * We don't require the BPF scheduler to avoid dispatching to
> +		 * offline CPUs mostly for convenience but also because CPUs can
> +		 * go offline between scx_bpf_dispatch() calls and here. If @p
> +		 * is destined to an offline CPU, queue it on its current CPU
> +		 * instead, which should always be safe. As this is an allowed
> +		 * behavior, don't trigger an ops error.
> +		 */
> +		if (unlikely(!test_rq_online(dst_rq)))
> +			dst_rq = src_rq;
> +
> +		if (src_rq == dst_rq) {
> +			/*
> +			 * As @p is staying on the same rq, there's no need to
> +			 * go through the full deactivate/activate cycle.
> +			 * Optimize by abbreviating the operations in
> +			 * move_task_to_local_dsq().
> +			 */
> +			dsp = p->scx.holding_cpu == raw_smp_processor_id();
> +			if (likely(dsp)) {
> +				p->scx.holding_cpu = -1;
> +				dispatch_enqueue(&dst_rq->scx.local_dsq, p,
> +						 enq_flags);
> +			}
> +		} else {
> +			dsp = move_task_to_local_dsq(dst_rq, p, enq_flags);
> +		}
> +
> +		/* if the destination CPU is idle, wake it up */
> +		if (dsp && p->sched_class > dst_rq->curr->sched_class)
> +			resched_curr(dst_rq);
> +
> +		dispatch_to_local_dsq_unlock(rq, rf, src_rq, locked_dst_rq);
> +
> +		return dsp ? DTL_DISPATCHED : DTL_LOST;
> +	}
> +#endif /* CONFIG_SMP */
> +
> +	scx_ops_error("SCX_DSQ_LOCAL[_ON] verdict target cpu %d not allowed for %s[%d]",
> +		      cpu_of(dst_rq), p->comm, p->pid);
> +	return DTL_INVALID;
> +}
> +
> +/**
> + * finish_dispatch - Asynchronously finish dispatching a task
> + * @rq: current rq which is locked
> + * @rf: rq_flags to use when unlocking @rq
> + * @p: task to finish dispatching
> + * @qseq_at_dispatch: qseq when @p started getting dispatched
> + * @dsq_id: destination DSQ ID
> + * @enq_flags: %SCX_ENQ_*
> + *
> + * Dispatching to local DSQs may need to wait for queueing to complete or
> + * require rq lock dancing. As we don't wanna do either while inside
> + * ops.dispatch() to avoid locking order inversion, we split dispatching into
> + * two parts. scx_bpf_dispatch() which is called by ops.dispatch() records the
> + * task and its qseq. Once ops.dispatch() returns, this function is called to
> + * finish up.
> + *
> + * There is no guarantee that @p is still valid for dispatching or even that it
> + * was valid in the first place. Make sure that the task is still owned by the
> + * BPF scheduler and claim the ownership before dispatching.
> + */
> +static void finish_dispatch(struct rq *rq, struct rq_flags *rf,
> +			    struct task_struct *p, u64 qseq_at_dispatch,
> +			    u64 dsq_id, u64 enq_flags)
> +{
> +	struct scx_dispatch_q *dsq;
> +	u64 opss;
> +
> +retry:
> +	/*
> +	 * No need for _acquire here. @p is accessed only after a successful
> +	 * try_cmpxchg to DISPATCHING.
> +	 */
> +	opss = atomic64_read(&p->scx.ops_state);
> +
> +	switch (opss & SCX_OPSS_STATE_MASK) {
> +	case SCX_OPSS_DISPATCHING:
> +	case SCX_OPSS_NONE:
> +		/* someone else already got to it */
> +		return;
> +	case SCX_OPSS_QUEUED:
> +		/*
> +		 * If qseq doesn't match, @p has gone through at least one
> +		 * dispatch/dequeue and re-enqueue cycle between
> +		 * scx_bpf_dispatch() and here and we have no claim on it.
> +		 */
> +		if ((opss & SCX_OPSS_QSEQ_MASK) != qseq_at_dispatch)
> +			return;
> +
> +		/*
> +		 * While we know @p is accessible, we don't yet have a claim on
> +		 * it - the BPF scheduler is allowed to dispatch tasks
> +		 * spuriously and there can be a racing dequeue attempt. Let's
> +		 * claim @p by atomically transitioning it from QUEUED to
> +		 * DISPATCHING.
> +		 */
> +		if (likely(atomic64_try_cmpxchg(&p->scx.ops_state, &opss,
> +						SCX_OPSS_DISPATCHING)))
> +			break;
> +		goto retry;
> +	case SCX_OPSS_QUEUEING:
> +		/*
> +		 * do_enqueue_task() is in the process of transferring the task
> +		 * to the BPF scheduler while holding @p's rq lock. As we aren't
> +		 * holding any kernel or BPF resource that the enqueue path may
> +		 * depend upon, it's safe to wait.
> +		 */
> +		wait_ops_state(p, opss);
> +		goto retry;
> +	}
> +
> +	BUG_ON(!(p->scx.flags & SCX_TASK_QUEUED));
> +
> +	switch (dispatch_to_local_dsq(rq, rf, dsq_id, p, enq_flags)) {
> +	case DTL_DISPATCHED:
> +		break;
> +	case DTL_LOST:
> +		break;
> +	case DTL_INVALID:
> +		dsq_id = SCX_DSQ_GLOBAL;
> +		fallthrough;
> +	case DTL_NOT_LOCAL:
> +		dsq = find_dsq_for_dispatch(cpu_rq(raw_smp_processor_id()),
> +					    dsq_id, p);
> +		dispatch_enqueue(dsq, p, enq_flags | SCX_ENQ_CLEAR_OPSS);
> +		break;
> +	}
> +}
> +
> +static void flush_dispatch_buf(struct rq *rq, struct rq_flags *rf)
> +{
> +	struct scx_dsp_ctx *dspc = this_cpu_ptr(&scx_dsp_ctx);
> +	u32 u;
> +
> +	for (u = 0; u < dspc->buf_cursor; u++) {
> +		struct scx_dsp_buf_ent *ent = &this_cpu_ptr(scx_dsp_buf)[u];
> +
> +		finish_dispatch(rq, rf, ent->task, ent->qseq, ent->dsq_id,
> +				ent->enq_flags);
> +	}
> +
> +	dspc->nr_tasks += dspc->buf_cursor;
> +	dspc->buf_cursor = 0;
> +}
> +
> +static int balance_scx(struct rq *rq, struct task_struct *prev,
> +		       struct rq_flags *rf)
> +{
> +	struct scx_rq *scx_rq = &rq->scx;
> +	struct scx_dsp_ctx *dspc = this_cpu_ptr(&scx_dsp_ctx);
> +	bool prev_on_scx = prev->sched_class == &ext_sched_class;
> +
> +	lockdep_assert_rq_held(rq);
> +
> +	if (prev_on_scx) {
> +		WARN_ON_ONCE(prev->scx.flags & SCX_TASK_BAL_KEEP);
> +		update_curr_scx(rq);
> +
> +		/*
> +		 * If @prev is runnable & has slice left, it has priority and
> +		 * fetching more just increases latency for the fetched tasks.
> +		 * Tell put_prev_task_scx() to put @prev on local_dsq.
> +		 *
> +		 * See scx_ops_disable_workfn() for the explanation on the
> +		 * disabling() test.
> +		 */
> +		if ((prev->scx.flags & SCX_TASK_QUEUED) &&
> +		    prev->scx.slice && !scx_ops_disabling()) {
> +			prev->scx.flags |= SCX_TASK_BAL_KEEP;
> +			return 1;
> +		}
> +	}
> +
> +	/* if there already are tasks to run, nothing to do */
> +	if (scx_rq->local_dsq.nr)
> +		return 1;
> +
> +	if (consume_dispatch_q(rq, rf, &scx_dsq_global))
> +		return 1;
> +
> +	if (!SCX_HAS_OP(dispatch))
> +		return 0;
> +
> +	dspc->rq = rq;
> +	dspc->rf = rf;
> +
> +	/*
> +	 * The dispatch loop. Because flush_dispatch_buf() may drop the rq lock,
> +	 * the local DSQ might still end up empty after a successful
> +	 * ops.dispatch(). If the local DSQ is empty even after ops.dispatch()
> +	 * produced some tasks, retry. The BPF scheduler may depend on this
> +	 * looping behavior to simplify its implementation.
> +	 */
> +	do {
> +		dspc->nr_tasks = 0;
> +
> +		SCX_CALL_OP(SCX_KF_DISPATCH, dispatch, cpu_of(rq),
> +			    prev_on_scx ? prev : NULL);
> +
> +		flush_dispatch_buf(rq, rf);
> +
> +		if (scx_rq->local_dsq.nr)
> +			return 1;
> +		if (consume_dispatch_q(rq, rf, &scx_dsq_global))
> +			return 1;
> +	} while (dspc->nr_tasks);
> +
> +	return 0;
> +}
> +
> +static void set_next_task_scx(struct rq *rq, struct task_struct *p, bool first)
> +{
> +	if (p->scx.flags & SCX_TASK_QUEUED) {
> +		WARN_ON_ONCE(atomic64_read(&p->scx.ops_state) != SCX_OPSS_NONE);
> +		dispatch_dequeue(&rq->scx, p);
> +	}
> +
> +	p->se.exec_start = rq_clock_task(rq);
> +}
> +
> +static void put_prev_task_scx(struct rq *rq, struct task_struct *p)
> +{
> +#ifndef CONFIG_SMP
> +	/*
> +	 * UP workaround.
> +	 *
> +	 * Because SCX may transfer tasks across CPUs during dispatch, dispatch
> +	 * is performed from its balance operation which isn't called in UP.
> +	 * Let's work around by calling it from the operations which come right
> +	 * after.
> +	 *
> +	 * 1. If the prev task is on SCX, pick_next_task() calls
> +	 *    .put_prev_task() right after. As .put_prev_task() is also called
> +	 *    from other places, we need to distinguish the calls which can be
> +	 *    done by looking at the previous task's state - if still queued or
> +	 *    dequeued with %SCX_DEQ_SLEEP, the caller must be pick_next_task().
> +	 *    This case is handled here.
> +	 *
> +	 * 2. If the prev task is not on SCX, the first following call into SCX
> +	 *    will be .pick_next_task(), which is covered by calling
> +	 *    balance_scx() from pick_next_task_scx().
> +	 *
> +	 * Note that we can't merge the first case into the second as
> +	 * balance_scx() must be called before the previous SCX task goes
> +	 * through put_prev_task_scx().
> +	 *
> +	 * As UP doesn't transfer tasks around, balance_scx() doesn't need @rf.
> +	 * Pass in %NULL.
> +	 */
> +	if (p->scx.flags & (SCX_TASK_QUEUED | SCX_TASK_DEQD_FOR_SLEEP))
> +		balance_scx(rq, p, NULL);
> +#endif
> +
> +	update_curr_scx(rq);
> +
> +	/*
> +	 * If we're being called from put_prev_task_balance(), balance_scx() may
> +	 * have decided that @p should keep running.
> +	 */
> +	if (p->scx.flags & SCX_TASK_BAL_KEEP) {
> +		p->scx.flags &= ~SCX_TASK_BAL_KEEP;
> +		dispatch_enqueue(&rq->scx.local_dsq, p, SCX_ENQ_HEAD);
> +		return;
> +	}
> +
> +	if (p->scx.flags & SCX_TASK_QUEUED) {
> +		/*
> +		 * If @p has slice left and balance_scx() didn't tag it for
> +		 * keeping, @p is getting preempted by a higher priority
> +		 * scheduler class. Leave it at the head of the local DSQ.
> +		 */
> +		if (p->scx.slice && !scx_ops_disabling()) {
> +			dispatch_enqueue(&rq->scx.local_dsq, p, SCX_ENQ_HEAD);
> +			return;
> +		}
> +
> +		/*
> +		 * If we're in the pick_next_task path, balance_scx() should
> +		 * have already populated the local DSQ if there are any other
> +		 * available tasks. If empty, tell ops.enqueue() that @p is the
> +		 * only one available for this cpu. ops.enqueue() should put it
> +		 * on the local DSQ so that the subsequent pick_next_task_scx()
> +		 * can find the task unless it wants to trigger a separate
> +		 * follow-up scheduling event.
> +		 */
> +		if (list_empty(&rq->scx.local_dsq.fifo))
> +			do_enqueue_task(rq, p, SCX_ENQ_LAST | SCX_ENQ_LOCAL, -1);
> +		else
> +			do_enqueue_task(rq, p, 0, -1);
> +	}
> +}
> +
> +static struct task_struct *first_local_task(struct rq *rq)
> +{
> +	return list_first_entry_or_null(&rq->scx.local_dsq.fifo,
> +					struct task_struct, scx.dsq_node);
> +}
> +
> +static struct task_struct *pick_next_task_scx(struct rq *rq)
> +{
> +	struct task_struct *p;
> +
> +#ifndef CONFIG_SMP
> +	/* UP workaround - see the comment at the head of put_prev_task_scx() */
> +	if (unlikely(rq->curr->sched_class != &ext_sched_class))
> +		balance_scx(rq, rq->curr, NULL);
> +#endif
> +
> +	p = first_local_task(rq);
> +	if (!p)
> +		return NULL;
> +
> +	if (unlikely(!p->scx.slice)) {
> +		if (!scx_ops_disabling() && !scx_warned_zero_slice) {
> +			printk_deferred(KERN_WARNING "sched_ext: %s[%d] has zero slice in pick_next_task_scx()\n",
> +					p->comm, p->pid);
> +			scx_warned_zero_slice = true;
> +		}
> +		p->scx.slice = SCX_SLICE_DFL;
> +	}
> +
> +	set_next_task_scx(rq, p, true);
> +
> +	return p;
> +}
> +
> +#ifdef CONFIG_SMP
> +
> +static bool test_and_clear_cpu_idle(int cpu)
> +{
> +#ifdef CONFIG_SCHED_SMT
> +	/*
> +	 * SMT mask should be cleared whether we can claim @cpu or not. The SMT
> +	 * cluster is not wholly idle either way. This also prevents
> +	 * scx_pick_idle_cpu() from getting caught in an infinite loop.
> +	 */
> +	if (sched_smt_active()) {
> +		const struct cpumask *smt = cpu_smt_mask(cpu);
> +
> +		/*
> +		 * If offline, @cpu is not its own sibling and
> +		 * scx_pick_idle_cpu() can get caught in an infinite loop as
> +		 * @cpu is never cleared from idle_masks.smt. Ensure that @cpu
> +		 * is eventually cleared.
> +		 */
> +		if (cpumask_intersects(smt, idle_masks.smt))
> +			cpumask_andnot(idle_masks.smt, idle_masks.smt, smt);
> +		else if (cpumask_test_cpu(cpu, idle_masks.smt))
> +			__cpumask_clear_cpu(cpu, idle_masks.smt);
> +	}
> +#endif
> +	return cpumask_test_and_clear_cpu(cpu, idle_masks.cpu);
> +}
> +
> +static s32 scx_pick_idle_cpu(const struct cpumask *cpus_allowed, u64 flags)
> +{
> +	int cpu;
> +
> +retry:
> +	if (sched_smt_active()) {
> +		cpu = cpumask_any_and_distribute(idle_masks.smt, cpus_allowed);
> +		if (cpu < nr_cpu_ids)
> +			goto found;
> +
> +		if (flags & SCX_PICK_IDLE_CORE)
> +			return -EBUSY;
> +	}
> +
> +	cpu = cpumask_any_and_distribute(idle_masks.cpu, cpus_allowed);
> +	if (cpu >= nr_cpu_ids)
> +		return -EBUSY;
> +
> +found:
> +	if (test_and_clear_cpu_idle(cpu))
> +		return cpu;
> +	else
> +		goto retry;
> +}
> +
> +static s32 scx_select_cpu_dfl(struct task_struct *p, s32 prev_cpu, u64 wake_flags)
> +{
> +	s32 cpu;
> +
> +	if (!static_branch_likely(&scx_builtin_idle_enabled)) {
> +		scx_ops_error("built-in idle tracking is disabled");
> +		return prev_cpu;
> +	}
> +
> +	/*
> +	 * If WAKE_SYNC and the machine isn't fully saturated, wake up @p to the
> +	 * local DSQ of the waker.
> +	 */
> +	if ((wake_flags & SCX_WAKE_SYNC) && p->nr_cpus_allowed > 1 &&
> +	    !cpumask_empty(idle_masks.cpu) && !(current->flags & PF_EXITING)) {
> +		cpu = smp_processor_id();
> +		if (cpumask_test_cpu(cpu, p->cpus_ptr)) {
> +			p->scx.flags |= SCX_TASK_ENQ_LOCAL;
> +			return cpu;
> +		}
> +	}
> +
> +	if (p->nr_cpus_allowed == 1)
> +		return prev_cpu;
> +
> +	/*
> +	 * If CPU has SMT, any wholly idle CPU is likely a better pick than
> +	 * partially idle @prev_cpu.
> +	 */
> +	if (sched_smt_active()) {
> +		if (cpumask_test_cpu(prev_cpu, idle_masks.smt) &&
> +		    test_and_clear_cpu_idle(prev_cpu)) {
> +			p->scx.flags |= SCX_TASK_ENQ_LOCAL;
> +			return prev_cpu;
> +		}
> +
> +		cpu = scx_pick_idle_cpu(p->cpus_ptr, SCX_PICK_IDLE_CORE);
> +		if (cpu >= 0) {
> +			p->scx.flags |= SCX_TASK_ENQ_LOCAL;
> +			return cpu;
> +		}
> +	}
> +
> +	if (test_and_clear_cpu_idle(prev_cpu)) {
> +		p->scx.flags |= SCX_TASK_ENQ_LOCAL;
> +		return prev_cpu;
> +	}
> +
> +	cpu = scx_pick_idle_cpu(p->cpus_ptr, 0);
> +	if (cpu >= 0) {
> +		p->scx.flags |= SCX_TASK_ENQ_LOCAL;
> +		return cpu;
> +	}
> +
> +	return prev_cpu;
> +}
> +
> +static int select_task_rq_scx(struct task_struct *p, int prev_cpu, int wake_flags)
> +{
> +	if (SCX_HAS_OP(select_cpu)) {
> +		s32 cpu;
> +
> +		cpu = SCX_CALL_OP_RET(SCX_KF_REST, select_cpu, p, prev_cpu,
> +				      wake_flags);
> +		if (ops_cpu_valid(cpu)) {
> +			return cpu;
> +		} else {
> +			scx_ops_error("select_cpu returned invalid cpu %d", cpu);
> +			return prev_cpu;
> +		}
> +	} else {
> +		return scx_select_cpu_dfl(p, prev_cpu, wake_flags);
> +	}
> +}
> +
> +static void set_cpus_allowed_scx(struct task_struct *p,
> +				 struct affinity_context *ac)
> +{
> +	set_cpus_allowed_common(p, ac);
> +
> +	/*
> +	 * The effective cpumask is stored in @p->cpus_ptr which may temporarily
> +	 * differ from the configured one in @p->cpus_mask. Always tell the bpf
> +	 * scheduler the effective one.
> +	 *
> +	 * Fine-grained memory write control is enforced by BPF making the const
> +	 * designation pointless. Cast it away when calling the operation.
> +	 */
> +	if (SCX_HAS_OP(set_cpumask))
> +		SCX_CALL_OP(SCX_KF_REST, set_cpumask, p,
> +			    (struct cpumask *)p->cpus_ptr);
> +}
> +
> +static void reset_idle_masks(void)
> +{
> +	/* consider all cpus idle, should converge to the actual state quickly */
> +	cpumask_setall(idle_masks.cpu);
> +	cpumask_setall(idle_masks.smt);
> +}
> +
> +void __scx_update_idle(struct rq *rq, bool idle)
> +{
> +	int cpu = cpu_of(rq);
> +
> +	if (SCX_HAS_OP(update_idle)) {
> +		SCX_CALL_OP(SCX_KF_REST, update_idle, cpu_of(rq), idle);
> +		if (!static_branch_unlikely(&scx_builtin_idle_enabled))
> +			return;
> +	}
> +
> +	if (idle)
> +		cpumask_set_cpu(cpu, idle_masks.cpu);
> +	else
> +		cpumask_clear_cpu(cpu, idle_masks.cpu);
> +
> +#ifdef CONFIG_SCHED_SMT
> +	if (sched_smt_active()) {
> +		const struct cpumask *smt = cpu_smt_mask(cpu);
> +
> +		if (idle) {
> +			/*
> +			 * idle_masks.smt handling is racy but that's fine as
> +			 * it's only for optimization and self-correcting.
> +			 */
> +			for_each_cpu(cpu, smt) {
> +				if (!cpumask_test_cpu(cpu, idle_masks.cpu))
> +					return;
> +			}
> +			cpumask_or(idle_masks.smt, idle_masks.smt, smt);
> +		} else {
> +			cpumask_andnot(idle_masks.smt, idle_masks.smt, smt);
> +		}
> +	}
> +#endif
> +}
> +
> +#else /* !CONFIG_SMP */
> +
> +static bool test_and_clear_cpu_idle(int cpu) { return false; }
> +static s32 scx_pick_idle_cpu(const struct cpumask *cpus_allowed, u64 flags) { return -EBUSY; }
> +static void reset_idle_masks(void) {}
> +
> +#endif /* CONFIG_SMP */
> +
> +static void task_tick_scx(struct rq *rq, struct task_struct *curr, int queued)
> +{
> +	update_curr_scx(rq);
> +
> +	/*
> +	 * While disabling, always resched as we can't trust the slice
> +	 * management.
> +	 */
> +	if (scx_ops_disabling())
> +		curr->scx.slice = 0;
> +
> +	if (!curr->scx.slice)
> +		resched_curr(rq);
> +}
> +
> +static int scx_ops_prepare_task(struct task_struct *p, struct task_group *tg)
> +{
> +	int ret;
> +
> +	WARN_ON_ONCE(p->scx.flags & SCX_TASK_OPS_PREPPED);
> +
> +	if (SCX_HAS_OP(prep_enable)) {
> +		struct scx_enable_args args = { };
> +
> +		ret = SCX_CALL_OP_RET(SCX_KF_SLEEPABLE, prep_enable, p, &args);
> +		if (unlikely(ret)) {
> +			ret = ops_sanitize_err("prep_enable", ret);
> +			return ret;
> +		}
> +	}
> +
> +	p->scx.flags |= SCX_TASK_OPS_PREPPED;
> +	return 0;
> +}
> +
> +static void scx_ops_enable_task(struct task_struct *p)
> +{
> +	lockdep_assert_rq_held(task_rq(p));
> +	WARN_ON_ONCE(!(p->scx.flags & SCX_TASK_OPS_PREPPED));
> +
> +	if (SCX_HAS_OP(enable)) {
> +		struct scx_enable_args args = { };
> +		SCX_CALL_OP(SCX_KF_REST, enable, p, &args);
> +	}
> +	p->scx.flags &= ~SCX_TASK_OPS_PREPPED;
> +	p->scx.flags |= SCX_TASK_OPS_ENABLED;
> +}
> +
> +static void scx_ops_disable_task(struct task_struct *p)
> +{
> +	lockdep_assert_rq_held(task_rq(p));
> +
> +	if (p->scx.flags & SCX_TASK_OPS_PREPPED) {
> +		if (SCX_HAS_OP(cancel_enable)) {
> +			struct scx_enable_args args = { };
> +			SCX_CALL_OP(SCX_KF_REST, cancel_enable, p, &args);
> +		}
> +		p->scx.flags &= ~SCX_TASK_OPS_PREPPED;
> +	} else if (p->scx.flags & SCX_TASK_OPS_ENABLED) {
> +		if (SCX_HAS_OP(disable))
> +			SCX_CALL_OP(SCX_KF_REST, disable, p);
> +		p->scx.flags &= ~SCX_TASK_OPS_ENABLED;
> +	}
> +}
> +
> +static void set_task_scx_weight(struct task_struct *p)
> +{
> +	u32 weight = sched_prio_to_weight[p->static_prio - MAX_RT_PRIO];
> +
> +	p->scx.weight = sched_weight_to_cgroup(weight);
> +}
> +
> +/**
> + * refresh_scx_weight - Refresh a task's ext weight
> + * @p: task to refresh ext weight for
> + *
> + * @p->scx.weight carries the task's static priority in cgroup weight scale to
> + * enable easy access from the BPF scheduler. To keep it synchronized with the
> + * current task priority, this function should be called when a new task is
> + * created, priority is changed for a task on sched_ext, and a task is switched
> + * to sched_ext from other classes.
> + */
> +static void refresh_scx_weight(struct task_struct *p)
> +{
> +	lockdep_assert_rq_held(task_rq(p));
> +	set_task_scx_weight(p);
> +	if (SCX_HAS_OP(set_weight))
> +		SCX_CALL_OP(SCX_KF_REST, set_weight, p, p->scx.weight);
> +}
> +
> +void scx_pre_fork(struct task_struct *p)
> +{
> +	/*
> +	 * BPF scheduler enable/disable paths want to be able to iterate and
> +	 * update all tasks which can become complex when racing forks. As
> +	 * enable/disable are very cold paths, let's use a percpu_rwsem to
> +	 * exclude forks.
> +	 */
> +	percpu_down_read(&scx_fork_rwsem);
> +}
> +
> +int scx_fork(struct task_struct *p)
> +{
> +	percpu_rwsem_assert_held(&scx_fork_rwsem);
> +
> +	if (scx_enabled())
> +		return scx_ops_prepare_task(p, task_group(p));
> +	else
> +		return 0;
> +}
> +
> +void scx_post_fork(struct task_struct *p)
> +{
> +	if (scx_enabled()) {
> +		struct rq_flags rf;
> +		struct rq *rq;
> +
> +		rq = task_rq_lock(p, &rf);
> +		/*
> +		 * Set the weight manually before calling ops.enable() so that
> +		 * the scheduler doesn't see a stale value if they inspect the
> +		 * task struct. We'll invoke ops.set_weight() afterwards, as it
> +		 * would be odd to receive a callback on the task before we
> +		 * tell the scheduler that it's been fully enabled.
> +		 */
> +		set_task_scx_weight(p);
> +		scx_ops_enable_task(p);
> +		refresh_scx_weight(p);
> +		task_rq_unlock(rq, p, &rf);
> +	}
> +
> +	spin_lock_irq(&scx_tasks_lock);
> +	list_add_tail(&p->scx.tasks_node, &scx_tasks);
> +	spin_unlock_irq(&scx_tasks_lock);
> +
> +	percpu_up_read(&scx_fork_rwsem);
> +}
> +
> +void scx_cancel_fork(struct task_struct *p)
> +{
> +	if (scx_enabled())
> +		scx_ops_disable_task(p);
> +	percpu_up_read(&scx_fork_rwsem);
> +}
> +
> +void sched_ext_free(struct task_struct *p)
> +{
> +	unsigned long flags;
> +
> +	spin_lock_irqsave(&scx_tasks_lock, flags);
> +	list_del_init(&p->scx.tasks_node);
> +	spin_unlock_irqrestore(&scx_tasks_lock, flags);
> +
> +	/*
> +	 * @p is off scx_tasks and wholly ours. scx_ops_enable()'s PREPPED ->
> +	 * ENABLED transitions can't race us. Disable ops for @p.
> +	 */
> +	if (p->scx.flags & (SCX_TASK_OPS_PREPPED | SCX_TASK_OPS_ENABLED)) {
> +		struct rq_flags rf;
> +		struct rq *rq;
> +
> +		rq = task_rq_lock(p, &rf);
> +		scx_ops_disable_task(p);
> +		task_rq_unlock(rq, p, &rf);
> +	}
> +}
> +
> +static void reweight_task_scx(struct rq *rq, struct task_struct *p, int newprio)
> +{
> +	refresh_scx_weight(p);
> +}
> +
> +static void prio_changed_scx(struct rq *rq, struct task_struct *p, int oldprio)
> +{
> +}
> +
> +static void switching_to_scx(struct rq *rq, struct task_struct *p)
> +{
> +	refresh_scx_weight(p);
> +
> +	/*
> +	 * set_cpus_allowed_scx() is not called while @p is associated with a
> +	 * different scheduler class. Keep the BPF scheduler up-to-date.
> +	 */
> +	if (SCX_HAS_OP(set_cpumask))
> +		SCX_CALL_OP(SCX_KF_REST, set_cpumask, p,
> +			    (struct cpumask *)p->cpus_ptr);
> +}
> +
> +static void check_preempt_curr_scx(struct rq *rq, struct task_struct *p,int wake_flags) {}
> +static void switched_to_scx(struct rq *rq, struct task_struct *p) {}
> +
> +/*
> + * Omitted operations:
> + *
> + * - check_preempt_curr: NOOP as it isn't useful in the wakeup path because the
> + *   task isn't tied to the CPU at that point.
> + *
> + * - migrate_task_rq: Unncessary as task to cpu mapping is transient.
> + *
> + * - task_fork/dead: We need fork/dead notifications for all tasks regardless of
> + *   their current sched_class. Call them directly from sched core instead.
> + *
> + * - task_woken, switched_from: Unnecessary.
> + */
> +DEFINE_SCHED_CLASS(ext) = {
> +	.enqueue_task		= enqueue_task_scx,
> +	.dequeue_task		= dequeue_task_scx,
> +	.yield_task		= yield_task_scx,
> +	.yield_to_task		= yield_to_task_scx,
> +
> +	.check_preempt_curr	= check_preempt_curr_scx,
> +
> +	.pick_next_task		= pick_next_task_scx,
> +
> +	.put_prev_task		= put_prev_task_scx,
> +	.set_next_task          = set_next_task_scx,
> +
> +#ifdef CONFIG_SMP
> +	.balance		= balance_scx,
> +	.select_task_rq		= select_task_rq_scx,
> +	.set_cpus_allowed	= set_cpus_allowed_scx,
> +#endif
> +
> +	.task_tick		= task_tick_scx,
> +
> +	.switching_to		= switching_to_scx,
> +	.switched_to		= switched_to_scx,
> +	.reweight_task		= reweight_task_scx,
> +	.prio_changed		= prio_changed_scx,
> +
> +	.update_curr		= update_curr_scx,
> +
> +#ifdef CONFIG_UCLAMP_TASK
> +	.uclamp_enabled		= 0,
> +#endif
> +};
> +
> +static void init_dsq(struct scx_dispatch_q *dsq, u64 dsq_id)
> +{
> +	memset(dsq, 0, sizeof(*dsq));
> +
> +	raw_spin_lock_init(&dsq->lock);
> +	INIT_LIST_HEAD(&dsq->fifo);
> +	dsq->id = dsq_id;
> +}
> +
> +static struct scx_dispatch_q *create_dsq(u64 dsq_id, int node)
> +{
> +	struct scx_dispatch_q *dsq;
> +	int ret;
> +
> +	if (dsq_id & SCX_DSQ_FLAG_BUILTIN)
> +		return ERR_PTR(-EINVAL);
> +
> +	dsq = kmalloc_node(sizeof(*dsq), GFP_KERNEL, node);
> +	if (!dsq)
> +		return ERR_PTR(-ENOMEM);
> +
> +	init_dsq(dsq, dsq_id);
> +
> +	ret = rhashtable_insert_fast(&dsq_hash, &dsq->hash_node,
> +				     dsq_hash_params);
> +	if (ret) {
> +		kfree(dsq);
> +		return ERR_PTR(ret);
> +	}
> +	return dsq;
> +}
> +
> +static void free_dsq_irq_workfn(struct irq_work *irq_work)
> +{
> +	struct llist_node *to_free = llist_del_all(&dsqs_to_free);
> +	struct scx_dispatch_q *dsq, *tmp_dsq;
> +
> +	llist_for_each_entry_safe(dsq, tmp_dsq, to_free, free_node)
> +		kfree_rcu(dsq);
> +}
> +
> +static DEFINE_IRQ_WORK(free_dsq_irq_work, free_dsq_irq_workfn);
> +
> +static void destroy_dsq(u64 dsq_id)
> +{
> +	struct scx_dispatch_q *dsq;
> +	unsigned long flags;
> +
> +	rcu_read_lock();
> +
> +	dsq = rhashtable_lookup_fast(&dsq_hash, &dsq_id, dsq_hash_params);
> +	if (!dsq)
> +		goto out_unlock_rcu;
> +
> +	raw_spin_lock_irqsave(&dsq->lock, flags);
> +
> +	if (dsq->nr) {
> +		scx_ops_error("attempting to destroy in-use dsq 0x%016llx (nr=%u)",
> +			      dsq->id, dsq->nr);
> +		goto out_unlock_dsq;
> +	}
> +
> +	if (rhashtable_remove_fast(&dsq_hash, &dsq->hash_node, dsq_hash_params))
> +		goto out_unlock_dsq;
> +
> +	/*
> +	 * Mark dead by invalidating ->id to prevent dispatch_enqueue() from
> +	 * queueing more tasks. As this function can be called from anywhere,
> +	 * freeing is bounced through an irq work to avoid nesting RCU
> +	 * operations inside scheduler locks.
> +	 */
> +	dsq->id = SCX_DSQ_INVALID;
> +	llist_add(&dsq->free_node, &dsqs_to_free);
> +	irq_work_queue(&free_dsq_irq_work);
> +
> +out_unlock_dsq:
> +	raw_spin_unlock_irqrestore(&dsq->lock, flags);
> +out_unlock_rcu:
> +	rcu_read_unlock();
> +}
> +
> +/*
> + * Used by sched_fork() and __setscheduler_prio() to pick the matching
> + * sched_class. dl/rt are already handled.
> + */
> +bool task_should_scx(struct task_struct *p)
> +{
> +	if (!scx_enabled() || scx_ops_disabling())
> +		return false;
> +	return p->policy == SCHED_EXT;
> +}
> +
> +static void scx_ops_fallback_enqueue(struct task_struct *p, u64 enq_flags)
> +{
> +	if (enq_flags & SCX_ENQ_LAST)
> +		scx_bpf_dispatch(p, SCX_DSQ_LOCAL, SCX_SLICE_DFL, enq_flags);
> +	else
> +		scx_bpf_dispatch(p, SCX_DSQ_GLOBAL, SCX_SLICE_DFL, enq_flags);
> +}
> +
> +static void scx_ops_fallback_dispatch(s32 cpu, struct task_struct *prev) {}
> +
> +static void scx_ops_disable_workfn(struct kthread_work *work)
> +{
> +	struct scx_exit_info *ei = &scx_exit_info;
> +	struct scx_task_iter sti;
> +	struct task_struct *p;
> +	struct rhashtable_iter rht_iter;
> +	struct scx_dispatch_q *dsq;
> +	const char *reason;
> +	int i, type;
> +
> +	type = atomic_read(&scx_exit_type);
> +	while (true) {
> +		/*
> +		 * NONE indicates that a new scx_ops has been registered since
> +		 * disable was scheduled - don't kill the new ops. DONE
> +		 * indicates that the ops has already been disabled.
> +		 */
> +		if (type == SCX_EXIT_NONE || type == SCX_EXIT_DONE)
> +			return;
> +		if (atomic_try_cmpxchg(&scx_exit_type, &type, SCX_EXIT_DONE))
> +			break;
> +	}
> +
> +	switch (type) {
> +	case SCX_EXIT_UNREG:
> +		reason = "BPF scheduler unregistered";
> +		break;
> +	case SCX_EXIT_ERROR:
> +		reason = "runtime error";
> +		break;
> +	case SCX_EXIT_ERROR_BPF:
> +		reason = "scx_bpf_error";
> +		break;
> +	default:
> +		reason = "<UNKNOWN>";
> +	}
> +
> +	ei->type = type;
> +	strlcpy(ei->reason, reason, sizeof(ei->reason));
> +
> +	switch (scx_ops_set_enable_state(SCX_OPS_DISABLING)) {
> +	case SCX_OPS_DISABLED:
> +		pr_warn("sched_ext: ops error detected without ops (%s)\n",
> +			scx_exit_info.msg);
> +		WARN_ON_ONCE(scx_ops_set_enable_state(SCX_OPS_DISABLED) !=
> +			     SCX_OPS_DISABLING);
> +		return;
> +	case SCX_OPS_PREPPING:
> +		goto forward_progress_guaranteed;
> +	case SCX_OPS_DISABLING:
> +		/* shouldn't happen but handle it like ENABLING if it does */
> +		WARN_ONCE(true, "sched_ext: duplicate disabling instance?");
> +		fallthrough;
> +	case SCX_OPS_ENABLING:
> +	case SCX_OPS_ENABLED:
> +		break;
> +	}
> +
> +	/*
> +	 * DISABLING is set and ops was either ENABLING or ENABLED indicating
> +	 * that the ops and static branches are set.
> +	 *
> +	 * We must guarantee that all runnable tasks make forward progress
> +	 * without trusting the BPF scheduler. We can't grab any mutexes or
> +	 * rwsems as they might be held by tasks that the BPF scheduler is
> +	 * forgetting to run, which unfortunately also excludes toggling the
> +	 * static branches.
> +	 *
> +	 * Let's work around by overriding a couple ops and modifying behaviors
> +	 * based on the DISABLING state and then cycling the tasks through
> +	 * dequeue/enqueue to force global FIFO scheduling.
> +	 *
> +	 * a. ops.enqueue() and .dispatch() are overridden for simple global
> +	 *    FIFO scheduling.
> +	 *
> +	 * b. balance_scx() never sets %SCX_TASK_BAL_KEEP as the slice value
> +	 *    can't be trusted. Whenever a tick triggers, the running task is
> +	 *    rotated to the tail of the queue.
> +	 *
> +	 * c. pick_next_task() suppresses zero slice warning.
> +	 */
> +	scx_ops.enqueue = scx_ops_fallback_enqueue;
> +	scx_ops.dispatch = scx_ops_fallback_dispatch;
> +
> +	spin_lock_irq(&scx_tasks_lock);
> +	scx_task_iter_init(&sti);
> +	while ((p = scx_task_iter_next_filtered_locked(&sti))) {
> +		if (READ_ONCE(p->__state) != TASK_DEAD) {
> +			struct sched_enq_and_set_ctx ctx;
> +
> +			/* cycling deq/enq is enough, see above */
> +			sched_deq_and_put_task(p, DEQUEUE_SAVE | DEQUEUE_MOVE, &ctx);
> +			sched_enq_and_set_task(&ctx);
> +		}
> +	}
> +	scx_task_iter_exit(&sti);
> +	spin_unlock_irq(&scx_tasks_lock);
> +
> +forward_progress_guaranteed:
> +	/*
> +	 * Here, every runnable task is guaranteed to make forward progress and
> +	 * we can safely use blocking synchronization constructs. Actually
> +	 * disable ops.
> +	 */
> +	mutex_lock(&scx_ops_enable_mutex);
> +
> +	/* avoid racing against fork */
> +	cpus_read_lock();
> +	percpu_down_write(&scx_fork_rwsem);
> +
> +	spin_lock_irq(&scx_tasks_lock);
> +	scx_task_iter_init(&sti);
> +	while ((p = scx_task_iter_next_filtered_locked(&sti))) {
> +		const struct sched_class *old_class = p->sched_class;
> +		struct sched_enq_and_set_ctx ctx;
> +		bool alive = READ_ONCE(p->__state) != TASK_DEAD;
> +
> +		sched_deq_and_put_task(p, DEQUEUE_SAVE | DEQUEUE_MOVE, &ctx);
> +
> +		p->scx.slice = min_t(u64, p->scx.slice, SCX_SLICE_DFL);
> +
> +		__setscheduler_prio(p, p->prio);
> +		if (alive)
> +			check_class_changing(task_rq(p), p, old_class);
> +
> +		sched_enq_and_set_task(&ctx);
> +
> +		if (alive)
> +			check_class_changed(task_rq(p), p, old_class, p->prio);
> +
> +		scx_ops_disable_task(p);
> +	}
> +	scx_task_iter_exit(&sti);
> +	spin_unlock_irq(&scx_tasks_lock);
> +
> +	/* no task is on scx, turn off all the switches and flush in-progress calls */
> +	static_branch_disable_cpuslocked(&__scx_ops_enabled);
> +	for (i = 0; i < SCX_NR_ONLINE_OPS; i++)
> +		static_branch_disable_cpuslocked(&scx_has_op[i]);
> +	static_branch_disable_cpuslocked(&scx_ops_enq_last);
> +	static_branch_disable_cpuslocked(&scx_ops_enq_exiting);
> +	static_branch_disable_cpuslocked(&scx_builtin_idle_enabled);
> +	synchronize_rcu();
> +
> +	percpu_up_write(&scx_fork_rwsem);
> +	cpus_read_unlock();
> +
> +	if (ei->type >= SCX_EXIT_ERROR) {
> +		printk(KERN_ERR "sched_ext: BPF scheduler \"%s\" errored, disabling\n", scx_ops.name);
> +
> +		if (ei->msg[0] == '\0')
> +			printk(KERN_ERR "sched_ext: %s\n", ei->reason);
> +		else
> +			printk(KERN_ERR "sched_ext: %s (%s)\n", ei->reason, ei->msg);
> +
> +		stack_trace_print(ei->bt, ei->bt_len, 2);
> +	}
> +
> +	if (scx_ops.exit)
> +		SCX_CALL_OP(SCX_KF_UNLOCKED, exit, ei);
> +
> +	memset(&scx_ops, 0, sizeof(scx_ops));
> +
> +	rhashtable_walk_enter(&dsq_hash, &rht_iter);
> +	do {
> +		rhashtable_walk_start(&rht_iter);
> +
> +		while ((dsq = rhashtable_walk_next(&rht_iter)) && !IS_ERR(dsq))
> +			destroy_dsq(dsq->id);
> +
> +		rhashtable_walk_stop(&rht_iter);
> +	} while (dsq == ERR_PTR(-EAGAIN));
> +	rhashtable_walk_exit(&rht_iter);
> +
> +	free_percpu(scx_dsp_buf);
> +	scx_dsp_buf = NULL;
> +	scx_dsp_max_batch = 0;
> +
> +	mutex_unlock(&scx_ops_enable_mutex);
> +
> +	WARN_ON_ONCE(scx_ops_set_enable_state(SCX_OPS_DISABLED) !=
> +		     SCX_OPS_DISABLING);
> +}
> +
> +static DEFINE_KTHREAD_WORK(scx_ops_disable_work, scx_ops_disable_workfn);
> +
> +static void schedule_scx_ops_disable_work(void)
> +{
> +	struct kthread_worker *helper = READ_ONCE(scx_ops_helper);
> +
> +	/*
> +	 * We may be called spuriously before the first bpf_sched_ext_reg(). If
> +	 * scx_ops_helper isn't set up yet, there's nothing to do.
> +	 */
> +	if (helper)
> +		kthread_queue_work(helper, &scx_ops_disable_work);
> +}
> +
> +static void scx_ops_disable(enum scx_exit_type type)
> +{
> +	int none = SCX_EXIT_NONE;
> +
> +	if (WARN_ON_ONCE(type == SCX_EXIT_NONE || type == SCX_EXIT_DONE))
> +		type = SCX_EXIT_ERROR;
> +
> +	atomic_try_cmpxchg(&scx_exit_type, &none, type);
> +
> +	schedule_scx_ops_disable_work();
> +}
> +
> +static void scx_ops_error_irq_workfn(struct irq_work *irq_work)
> +{
> +	schedule_scx_ops_disable_work();
> +}
> +
> +static DEFINE_IRQ_WORK(scx_ops_error_irq_work, scx_ops_error_irq_workfn);
> +
> +__printf(2, 3) static void scx_ops_error_type(enum scx_exit_type type,
> +					      const char *fmt, ...)
> +{
> +	struct scx_exit_info *ei = &scx_exit_info;
> +	int none = SCX_EXIT_NONE;
> +	va_list args;
> +
> +	if (!atomic_try_cmpxchg(&scx_exit_type, &none, type))
> +		return;
> +
> +	ei->bt_len = stack_trace_save(ei->bt, ARRAY_SIZE(ei->bt), 1);
> +
> +	va_start(args, fmt);
> +	vscnprintf(ei->msg, ARRAY_SIZE(ei->msg), fmt, args);
> +	va_end(args);
> +
> +	irq_work_queue(&scx_ops_error_irq_work);
> +}
> +
> +static struct kthread_worker *scx_create_rt_helper(const char *name)
> +{
> +	struct kthread_worker *helper;
> +
> +	helper = kthread_create_worker(0, name);
> +	if (helper)
> +		sched_set_fifo(helper->task);
> +	return helper;
> +}
> +
> +static int scx_ops_enable(struct sched_ext_ops *ops)
> +{
> +	struct scx_task_iter sti;
> +	struct task_struct *p;
> +	int i, ret;
> +
> +	mutex_lock(&scx_ops_enable_mutex);
> +
> +	if (!scx_ops_helper) {
> +		WRITE_ONCE(scx_ops_helper,
> +			   scx_create_rt_helper("sched_ext_ops_helper"));
> +		if (!scx_ops_helper) {
> +			ret = -ENOMEM;
> +			goto err_unlock;
> +		}
> +	}
> +
> +	if (scx_ops_enable_state() != SCX_OPS_DISABLED) {
> +		ret = -EBUSY;
> +		goto err_unlock;
> +	}
> +
> +	/*
> +	 * Set scx_ops, transition to PREPPING and clear exit info to arm the
> +	 * disable path. Failure triggers full disabling from here on.
> +	 */
> +	scx_ops = *ops;
> +
> +	WARN_ON_ONCE(scx_ops_set_enable_state(SCX_OPS_PREPPING) !=
> +		     SCX_OPS_DISABLED);
> +
> +	memset(&scx_exit_info, 0, sizeof(scx_exit_info));
> +	atomic_set(&scx_exit_type, SCX_EXIT_NONE);
> +	scx_warned_zero_slice = false;
> +
> +	atomic64_set(&scx_nr_rejected, 0);
> +
> +	/*
> +	 * Keep CPUs stable during enable so that the BPF scheduler can track
> +	 * online CPUs by watching ->on/offline_cpu() after ->init().
> +	 */
> +	cpus_read_lock();
> +
> +	if (scx_ops.init) {
> +		ret = SCX_CALL_OP_RET(SCX_KF_INIT, init);
> +		if (ret) {
> +			ret = ops_sanitize_err("init", ret);
> +			goto err_disable;
> +		}
> +
> +		/*
> +		 * Exit early if ops.init() triggered scx_bpf_error(). Not
> +		 * strictly necessary as we'll fail transitioning into ENABLING
> +		 * later but that'd be after calling ops.prep_enable() on all
> +		 * tasks and with -EBUSY which isn't very intuitive. Let's exit
> +		 * early with success so that the condition is notified through
> +		 * ops.exit() like other scx_bpf_error() invocations.
> +		 */
> +		if (atomic_read(&scx_exit_type) != SCX_EXIT_NONE)
> +			goto err_disable;
> +	}
> +
> +	WARN_ON_ONCE(scx_dsp_buf);
> +	scx_dsp_max_batch = ops->dispatch_max_batch ?: SCX_DSP_DFL_MAX_BATCH;
> +	scx_dsp_buf = __alloc_percpu(sizeof(scx_dsp_buf[0]) * scx_dsp_max_batch,
> +				     __alignof__(scx_dsp_buf[0]));
> +	if (!scx_dsp_buf) {
> +		ret = -ENOMEM;
> +		goto err_disable;
> +	}
> +
> +	/*
> +	 * Lock out forks before opening the floodgate so that they don't wander
> +	 * into the operations prematurely.
> +	 */
> +	percpu_down_write(&scx_fork_rwsem);
> +
> +	for (i = 0; i < SCX_NR_ONLINE_OPS; i++)
> +		if (((void (**)(void))ops)[i])
> +			static_branch_enable_cpuslocked(&scx_has_op[i]);
> +
> +	if (ops->flags & SCX_OPS_ENQ_LAST)
> +		static_branch_enable_cpuslocked(&scx_ops_enq_last);
> +
> +	if (ops->flags & SCX_OPS_ENQ_EXITING)
> +		static_branch_enable_cpuslocked(&scx_ops_enq_exiting);
> +
> +	if (!ops->update_idle || (ops->flags & SCX_OPS_KEEP_BUILTIN_IDLE)) {
> +		reset_idle_masks();
> +		static_branch_enable_cpuslocked(&scx_builtin_idle_enabled);
> +	} else {
> +		static_branch_disable_cpuslocked(&scx_builtin_idle_enabled);
> +	}
> +
> +	static_branch_enable_cpuslocked(&__scx_ops_enabled);
> +
> +	/*
> +	 * Enable ops for every task. Fork is excluded by scx_fork_rwsem
> +	 * preventing new tasks from being added. No need to exclude tasks
> +	 * leaving as sched_ext_free() can handle both prepped and enabled
> +	 * tasks. Prep all tasks first and then enable them with preemption
> +	 * disabled.
> +	 */
> +	spin_lock_irq(&scx_tasks_lock);
> +
> +	scx_task_iter_init(&sti);
> +	while ((p = scx_task_iter_next_filtered(&sti))) {
> +		get_task_struct(p);
> +		spin_unlock_irq(&scx_tasks_lock);
> +
> +		ret = scx_ops_prepare_task(p, task_group(p));
> +		if (ret) {
> +			put_task_struct(p);
> +			spin_lock_irq(&scx_tasks_lock);
> +			scx_task_iter_exit(&sti);
> +			spin_unlock_irq(&scx_tasks_lock);
> +			pr_err("sched_ext: ops.prep_enable() failed (%d) for %s[%d] while loading\n",
> +			       ret, p->comm, p->pid);
> +			goto err_disable_unlock;
> +		}
> +
> +		put_task_struct(p);
> +		spin_lock_irq(&scx_tasks_lock);
> +	}
> +	scx_task_iter_exit(&sti);
> +
> +	/*
> +	 * All tasks are prepped but are still ops-disabled. Ensure that
> +	 * %current can't be scheduled out and switch everyone.
> +	 * preempt_disable() is necessary because we can't guarantee that
> +	 * %current won't be starved if scheduled out while switching.
> +	 */
> +	preempt_disable();
> +
> +	/*
> +	 * From here on, the disable path must assume that tasks have ops
> +	 * enabled and need to be recovered.
> +	 */
> +	if (!scx_ops_tryset_enable_state(SCX_OPS_ENABLING, SCX_OPS_PREPPING)) {
> +		preempt_enable();
> +		spin_unlock_irq(&scx_tasks_lock);
> +		ret = -EBUSY;
> +		goto err_disable_unlock;
> +	}
> +
> +	/*
> +	 * We're fully committed and can't fail. The PREPPED -> ENABLED
> +	 * transitions here are synchronized against sched_ext_free() through
> +	 * scx_tasks_lock.
> +	 */
> +	scx_task_iter_init(&sti);
> +	while ((p = scx_task_iter_next_filtered_locked(&sti))) {
> +		if (READ_ONCE(p->__state) != TASK_DEAD) {
> +			const struct sched_class *old_class = p->sched_class;
> +			struct sched_enq_and_set_ctx ctx;
> +
> +			sched_deq_and_put_task(p, DEQUEUE_SAVE | DEQUEUE_MOVE,
> +					       &ctx);
> +
> +			scx_ops_enable_task(p);
> +			__setscheduler_prio(p, p->prio);
> +			check_class_changing(task_rq(p), p, old_class);
> +
> +			sched_enq_and_set_task(&ctx);
> +
> +			check_class_changed(task_rq(p), p, old_class, p->prio);
> +		} else {
> +			scx_ops_disable_task(p);
> +		}
> +	}
> +	scx_task_iter_exit(&sti);
> +
> +	spin_unlock_irq(&scx_tasks_lock);
> +	preempt_enable();
> +	percpu_up_write(&scx_fork_rwsem);
> +
> +	if (!scx_ops_tryset_enable_state(SCX_OPS_ENABLED, SCX_OPS_ENABLING)) {
> +		ret = -EBUSY;
> +		goto err_disable;
> +	}
> +
> +	cpus_read_unlock();
> +	mutex_unlock(&scx_ops_enable_mutex);
> +
> +	return 0;
> +
> +err_unlock:
> +	mutex_unlock(&scx_ops_enable_mutex);
> +	return ret;
> +
> +err_disable_unlock:
> +	percpu_up_write(&scx_fork_rwsem);
> +err_disable:
> +	cpus_read_unlock();
> +	mutex_unlock(&scx_ops_enable_mutex);
> +	/* must be fully disabled before returning */
> +	scx_ops_disable(SCX_EXIT_ERROR);
> +	kthread_flush_work(&scx_ops_disable_work);
> +	return ret;
> +}
> +
> +#ifdef CONFIG_SCHED_DEBUG
> +static const char *scx_ops_enable_state_str[] = {
> +	[SCX_OPS_PREPPING]	= "prepping",
> +	[SCX_OPS_ENABLING]	= "enabling",
> +	[SCX_OPS_ENABLED]	= "enabled",
> +	[SCX_OPS_DISABLING]	= "disabling",
> +	[SCX_OPS_DISABLED]	= "disabled",
> +};
> +
> +static int scx_debug_show(struct seq_file *m, void *v)
> +{
> +	mutex_lock(&scx_ops_enable_mutex);
> +	seq_printf(m, "%-30s: %s\n", "ops", scx_ops.name);
> +	seq_printf(m, "%-30s: %ld\n", "enabled", scx_enabled());
> +	seq_printf(m, "%-30s: %s\n", "enable_state",
> +		   scx_ops_enable_state_str[scx_ops_enable_state()]);
> +	seq_printf(m, "%-30s: %llu\n", "nr_rejected",
> +		   atomic64_read(&scx_nr_rejected));
> +	mutex_unlock(&scx_ops_enable_mutex);
> +	return 0;
> +}
> +
> +static int scx_debug_open(struct inode *inode, struct file *file)
> +{
> +	return single_open(file, scx_debug_show, NULL);
> +}
> +
> +const struct file_operations sched_ext_fops = {
> +	.open		= scx_debug_open,
> +	.read		= seq_read,
> +	.llseek		= seq_lseek,
> +	.release	= single_release,
> +};
> +#endif
> +
> +/********************************************************************************
> + * bpf_struct_ops plumbing.
> + */
> +#include <linux/bpf_verifier.h>
> +#include <linux/bpf.h>
> +#include <linux/btf.h>
> +
> +extern struct btf *btf_vmlinux;
> +static const struct btf_type *task_struct_type;
> +
> +static bool bpf_scx_is_valid_access(int off, int size,
> +				    enum bpf_access_type type,
> +				    const struct bpf_prog *prog,
> +				    struct bpf_insn_access_aux *info)
> +{
> +	if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS)
> +		return false;
> +	if (type != BPF_READ)
> +		return false;
> +	if (off % size != 0)
> +		return false;
> +
> +	return btf_ctx_access(off, size, type, prog, info);
> +}
> +
> +static int bpf_scx_btf_struct_access(struct bpf_verifier_log *log,
> +				     const struct bpf_reg_state *reg, int off,
> +				     int size)
> +{
> +	const struct btf_type *t;
> +
> +	t = btf_type_by_id(reg->btf, reg->btf_id);
> +	if (t == task_struct_type) {
> +		if (off >= offsetof(struct task_struct, scx.slice) &&
> +		    off + size <= offsetofend(struct task_struct, scx.slice))
> +			return SCALAR_VALUE;
> +	}
> +
> +	return 0;
> +}
> +
> +static const struct bpf_func_proto *
> +bpf_scx_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
> +{
> +	switch (func_id) {
> +	case BPF_FUNC_task_storage_get:
> +		return &bpf_task_storage_get_proto;
> +	case BPF_FUNC_task_storage_delete:
> +		return &bpf_task_storage_delete_proto;
> +	default:
> +		return bpf_base_func_proto(func_id);
> +	}
> +}
> +
> +const struct bpf_verifier_ops bpf_scx_verifier_ops = {
> +	.get_func_proto = bpf_scx_get_func_proto,
> +	.is_valid_access = bpf_scx_is_valid_access,
> +	.btf_struct_access = bpf_scx_btf_struct_access,
> +};
> +
> +static int bpf_scx_init_member(const struct btf_type *t,
> +			       const struct btf_member *member,
> +			       void *kdata, const void *udata)
> +{
> +	const struct sched_ext_ops *uops = udata;
> +	struct sched_ext_ops *ops = kdata;
> +	u32 moff = __btf_member_bit_offset(t, member) / 8;
> +	int ret;
> +
> +	switch (moff) {
> +	case offsetof(struct sched_ext_ops, dispatch_max_batch):
> +		if (*(u32 *)(udata + moff) > INT_MAX)
> +			return -E2BIG;
> +		ops->dispatch_max_batch = *(u32 *)(udata + moff);
> +		return 1;
> +	case offsetof(struct sched_ext_ops, flags):
> +		if (*(u64 *)(udata + moff) & ~SCX_OPS_ALL_FLAGS)
> +			return -EINVAL;
> +		ops->flags = *(u64 *)(udata + moff);
> +		return 1;
> +	case offsetof(struct sched_ext_ops, name):
> +		ret = bpf_obj_name_cpy(ops->name, uops->name,
> +				       sizeof(ops->name));
> +		if (ret < 0)
> +			return ret;
> +		if (ret == 0)
> +			return -EINVAL;
> +		return 1;
> +	}
> +
> +	return 0;
> +}
> +
> +static int bpf_scx_check_member(const struct btf_type *t,
> +				const struct btf_member *member,
> +				const struct bpf_prog *prog)
> +{
> +	u32 moff = __btf_member_bit_offset(t, member) / 8;
> +
> +	switch (moff) {
> +	case offsetof(struct sched_ext_ops, prep_enable):
> +	case offsetof(struct sched_ext_ops, init):
> +	case offsetof(struct sched_ext_ops, exit):
> +		break;
> +	default:
> +		if (prog->aux->sleepable)
> +			return -EINVAL;
> +	}
> +
> +	return 0;
> +}
> +
> +static int bpf_scx_reg(void *kdata)
> +{
> +	return scx_ops_enable(kdata);
> +}
> +
> +static void bpf_scx_unreg(void *kdata)
> +{
> +	scx_ops_disable(SCX_EXIT_UNREG);
> +	kthread_flush_work(&scx_ops_disable_work);
> +}
> +
> +static int bpf_scx_init(struct btf *btf)
> +{
> +	u32 type_id;
> +
> +	type_id = btf_find_by_name_kind(btf, "task_struct", BTF_KIND_STRUCT);
> +	if (type_id < 0)
> +		return -EINVAL;
> +	task_struct_type = btf_type_by_id(btf, type_id);
> +
> +	return 0;
> +}
> +
> +static int bpf_scx_update(void *kdata, void *old_kdata)
> +{
> +	/*
> +	 * sched_ext does not support updating the actively-loaded BPF
> +	 * scheduler, as registering a BPF scheduler can always fail if the
> +	 * scheduler returns an error code for e.g. ops.init(),
> +	 * ops.prep_enable(), etc. Similarly, we can always race with
> +	 * unregistration happening elsewhere, such as with sysrq.
> +	 */
> +	return -EOPNOTSUPP;
> +}
> +
> +static int bpf_scx_validate(void *kdata)
> +{
> +	return 0;
> +}
> +
> +/* "extern" to avoid sparse warning, only used in this file */
> +extern struct bpf_struct_ops bpf_sched_ext_ops;
> +
> +struct bpf_struct_ops bpf_sched_ext_ops = {
> +	.verifier_ops = &bpf_scx_verifier_ops,
> +	.reg = bpf_scx_reg,
> +	.unreg = bpf_scx_unreg,
> +	.check_member = bpf_scx_check_member,
> +	.init_member = bpf_scx_init_member,
> +	.init = bpf_scx_init,
> +	.update = bpf_scx_update,
> +	.validate = bpf_scx_validate,
> +	.name = "sched_ext_ops",
> +};
> +
> +void __init init_sched_ext_class(void)
> +{
> +	int cpu;
> +	u32 v;
> +
> +	/*
> +	 * The following is to prevent the compiler from optimizing out the enum
> +	 * definitions so that BPF scheduler implementations can use them
> +	 * through the generated vmlinux.h.
> +	 */
> +	WRITE_ONCE(v, SCX_WAKE_EXEC | SCX_ENQ_WAKEUP | SCX_DEQ_SLEEP);
> +
> +	BUG_ON(rhashtable_init(&dsq_hash, &dsq_hash_params));
> +	init_dsq(&scx_dsq_global, SCX_DSQ_GLOBAL);
> +#ifdef CONFIG_SMP
> +	BUG_ON(!alloc_cpumask_var(&idle_masks.cpu, GFP_KERNEL));
> +	BUG_ON(!alloc_cpumask_var(&idle_masks.smt, GFP_KERNEL));
> +#endif
> +	for_each_possible_cpu(cpu) {
> +		struct rq *rq = cpu_rq(cpu);
> +
> +		init_dsq(&rq->scx.local_dsq, SCX_DSQ_LOCAL);
> +	}
> +}
> +
> +
> +/********************************************************************************
> + * Helpers that can be called from the BPF scheduler.
> + */
> +#include <linux/btf_ids.h>
> +
> +/* Disables missing prototype warnings for kfuncs */
> +__diag_push();
> +__diag_ignore_all("-Wmissing-prototypes",
> +		  "Global functions as their definitions will be in vmlinux BTF");
> +
> +/**
> + * scx_bpf_create_dsq - Create a custom DSQ
> + * @dsq_id: DSQ to create
> + * @node: NUMA node to allocate from
> + *
> + * Create a custom DSQ identified by @dsq_id. Can be called from ops.init() and
> + * ops.prep_enable().
> + */
> +s32 scx_bpf_create_dsq(u64 dsq_id, s32 node)
> +{
> +	if (!scx_kf_allowed(SCX_KF_INIT | SCX_KF_SLEEPABLE))
> +		return -EINVAL;
> +
> +	if (unlikely(node >= (int)nr_node_ids ||
> +		     (node < 0 && node != NUMA_NO_NODE)))
> +		return -EINVAL;
> +	return PTR_ERR_OR_ZERO(create_dsq(dsq_id, node));
> +}
> +
> +BTF_SET8_START(scx_kfunc_ids_sleepable)
> +BTF_ID_FLAGS(func, scx_bpf_create_dsq, KF_SLEEPABLE)
> +BTF_SET8_END(scx_kfunc_ids_sleepable)
> +
> +static const struct btf_kfunc_id_set scx_kfunc_set_sleepable = {
> +	.owner			= THIS_MODULE,
> +	.set			= &scx_kfunc_ids_sleepable,
> +};
> +
> +static bool scx_dispatch_preamble(struct task_struct *p, u64 enq_flags)
> +{
> +	if (!scx_kf_allowed(SCX_KF_ENQUEUE | SCX_KF_DISPATCH))
> +		return false;
> +
> +	lockdep_assert_irqs_disabled();
> +
> +	if (unlikely(!p)) {
> +		scx_ops_error("called with NULL task");
> +		return false;
> +	}
> +
> +	if (unlikely(enq_flags & __SCX_ENQ_INTERNAL_MASK)) {
> +		scx_ops_error("invalid enq_flags 0x%llx", enq_flags);
> +		return false;
> +	}
> +
> +	return true;
> +}
> +
> +static void scx_dispatch_commit(struct task_struct *p, u64 dsq_id, u64 enq_flags)
> +{
> +	struct task_struct *ddsp_task;
> +	int idx;
> +
> +	ddsp_task = __this_cpu_read(direct_dispatch_task);
> +	if (ddsp_task) {
> +		direct_dispatch(ddsp_task, p, dsq_id, enq_flags);
> +		return;
> +	}
> +
> +	idx = __this_cpu_read(scx_dsp_ctx.buf_cursor);
> +	if (unlikely(idx >= scx_dsp_max_batch)) {
> +		scx_ops_error("dispatch buffer overflow");
> +		return;
> +	}
> +
> +	this_cpu_ptr(scx_dsp_buf)[idx] = (struct scx_dsp_buf_ent){
> +		.task = p,
> +		.qseq = atomic64_read(&p->scx.ops_state) & SCX_OPSS_QSEQ_MASK,
> +		.dsq_id = dsq_id,
> +		.enq_flags = enq_flags,
> +	};
> +	__this_cpu_inc(scx_dsp_ctx.buf_cursor);
> +}
> +
> +/**
> + * scx_bpf_dispatch - Dispatch a task into the FIFO queue of a DSQ
> + * @p: task_struct to dispatch
> + * @dsq_id: DSQ to dispatch to
> + * @slice: duration @p can run for in nsecs
> + * @enq_flags: SCX_ENQ_*
> + *
> + * Dispatch @p into the FIFO queue of the DSQ identified by @dsq_id. It is safe
> + * to call this function spuriously. Can be called from ops.enqueue() and
> + * ops.dispatch().
> + *
> + * When called from ops.enqueue(), it's for direct dispatch and @p must match
> + * the task being enqueued. Also, %SCX_DSQ_LOCAL_ON can't be used to target the
> + * local DSQ of a CPU other than the enqueueing one. Use ops.select_cpu() to be
> + * on the target CPU in the first place.
> + *
> + * When called from ops.dispatch(), there are no restrictions on @p or @dsq_id
> + * and this function can be called upto ops.dispatch_max_batch times to dispatch
> + * multiple tasks. scx_bpf_dispatch_nr_slots() returns the number of the
> + * remaining slots. scx_bpf_consume() flushes the batch and resets the counter.
> + *
> + * This function doesn't have any locking restrictions and may be called under
> + * BPF locks (in the future when BPF introduces more flexible locking).
> + *
> + * @p is allowed to run for @slice. The scheduling path is triggered on slice
> + * exhaustion. If zero, the current residual slice is maintained. If
> + * %SCX_SLICE_INF, @p never expires and the BPF scheduler must kick the CPU with
> + * scx_bpf_kick_cpu() to trigger scheduling.
> + */
> +void scx_bpf_dispatch(struct task_struct *p, u64 dsq_id, u64 slice,
> +		      u64 enq_flags)
> +{
> +	if (!scx_dispatch_preamble(p, enq_flags))
> +		return;
> +
> +	if (slice)
> +		p->scx.slice = slice;
> +	else
> +		p->scx.slice = p->scx.slice ?: 1;
> +
> +	scx_dispatch_commit(p, dsq_id, enq_flags);
> +}
> +
> +BTF_SET8_START(scx_kfunc_ids_enqueue_dispatch)
> +BTF_ID_FLAGS(func, scx_bpf_dispatch, KF_RCU)
> +BTF_SET8_END(scx_kfunc_ids_enqueue_dispatch)
> +
> +static const struct btf_kfunc_id_set scx_kfunc_set_enqueue_dispatch = {
> +	.owner			= THIS_MODULE,
> +	.set			= &scx_kfunc_ids_enqueue_dispatch,
> +};
> +
> +/**
> + * scx_bpf_dispatch_nr_slots - Return the number of remaining dispatch slots
> + *
> + * Can only be called from ops.dispatch().
> + */
> +u32 scx_bpf_dispatch_nr_slots(void)
> +{
> +	if (!scx_kf_allowed(SCX_KF_DISPATCH))
> +		return 0;
> +
> +	return scx_dsp_max_batch - __this_cpu_read(scx_dsp_ctx.buf_cursor);
> +}
> +
> +/**
> + * scx_bpf_consume - Transfer a task from a DSQ to the current CPU's local DSQ
> + * @dsq_id: DSQ to consume
> + *
> + * Consume a task from the non-local DSQ identified by @dsq_id and transfer it
> + * to the current CPU's local DSQ for execution. Can only be called from
> + * ops.dispatch().
> + *
> + * This function flushes the in-flight dispatches from scx_bpf_dispatch() before
> + * trying to consume the specified DSQ. It may also grab rq locks and thus can't
> + * be called under any BPF locks.
> + *
> + * Returns %true if a task has been consumed, %false if there isn't any task to
> + * consume.
> + */
> +bool scx_bpf_consume(u64 dsq_id)
> +{
> +	struct scx_dsp_ctx *dspc = this_cpu_ptr(&scx_dsp_ctx);
> +	struct scx_dispatch_q *dsq;
> +
> +	if (!scx_kf_allowed(SCX_KF_DISPATCH))
> +		return false;
> +
> +	flush_dispatch_buf(dspc->rq, dspc->rf);
> +
> +	dsq = find_non_local_dsq(dsq_id);
> +	if (unlikely(!dsq)) {
> +		scx_ops_error("invalid DSQ ID 0x%016llx", dsq_id);
> +		return false;
> +	}
> +
> +	if (consume_dispatch_q(dspc->rq, dspc->rf, dsq)) {
> +		/*
> +		 * A successfully consumed task can be dequeued before it starts
> +		 * running while the CPU is trying to migrate other dispatched
> +		 * tasks. Bump nr_tasks to tell balance_scx() to retry on empty
> +		 * local DSQ.
> +		 */
> +		dspc->nr_tasks++;
> +		return true;
> +	} else {
> +		return false;
> +	}
> +}
> +
> +BTF_SET8_START(scx_kfunc_ids_dispatch)
> +BTF_ID_FLAGS(func, scx_bpf_dispatch_nr_slots)
> +BTF_ID_FLAGS(func, scx_bpf_consume)
> +BTF_SET8_END(scx_kfunc_ids_dispatch)
> +
> +static const struct btf_kfunc_id_set scx_kfunc_set_dispatch = {
> +	.owner			= THIS_MODULE,
> +	.set			= &scx_kfunc_ids_dispatch,
> +};
> +
> +/**
> + * scx_bpf_dsq_nr_queued - Return the number of queued tasks
> + * @dsq_id: id of the DSQ
> + *
> + * Return the number of tasks in the DSQ matching @dsq_id. If not found,
> + * -%ENOENT is returned. Can be called from any non-sleepable online scx_ops
> + * operations.
> + */
> +s32 scx_bpf_dsq_nr_queued(u64 dsq_id)
> +{
> +	struct scx_dispatch_q *dsq;
> +
> +	lockdep_assert(rcu_read_lock_any_held());
> +
> +	if (dsq_id == SCX_DSQ_LOCAL) {
> +		return this_rq()->scx.local_dsq.nr;
> +	} else if ((dsq_id & SCX_DSQ_LOCAL_ON) == SCX_DSQ_LOCAL_ON) {
> +		s32 cpu = dsq_id & SCX_DSQ_LOCAL_CPU_MASK;
> +
> +		if (ops_cpu_valid(cpu))
> +			return cpu_rq(cpu)->scx.local_dsq.nr;
> +	} else {
> +		dsq = find_non_local_dsq(dsq_id);
> +		if (dsq)
> +			return dsq->nr;
> +	}
> +	return -ENOENT;
> +}
> +
> +/**
> + * scx_bpf_test_and_clear_cpu_idle - Test and clear @cpu's idle state
> + * @cpu: cpu to test and clear idle for
> + *
> + * Returns %true if @cpu was idle and its idle state was successfully cleared.
> + * %false otherwise.
> + *
> + * Unavailable if ops.update_idle() is implemented and
> + * %SCX_OPS_KEEP_BUILTIN_IDLE is not set.
> + */
> +bool scx_bpf_test_and_clear_cpu_idle(s32 cpu)
> +{
> +	if (!static_branch_likely(&scx_builtin_idle_enabled)) {
> +		scx_ops_error("built-in idle tracking is disabled");
> +		return false;
> +	}
> +
> +	if (ops_cpu_valid(cpu))
> +		return test_and_clear_cpu_idle(cpu);
> +	else
> +		return false;
> +}
> +
> +/**
> + * scx_bpf_pick_idle_cpu - Pick and claim an idle cpu
> + * @cpus_allowed: Allowed cpumask
> + * @flags: %SCX_PICK_IDLE_CPU_* flags
> + *
> + * Pick and claim an idle cpu in @cpus_allowed. Returns the picked idle cpu
> + * number on success. -%EBUSY if no matching cpu was found.
> + *
> + * Idle CPU tracking may race against CPU scheduling state transitions. For
> + * example, this function may return -%EBUSY as CPUs are transitioning into the
> + * idle state. If the caller then assumes that there will be dispatch events on
> + * the CPUs as they were all busy, the scheduler may end up stalling with CPUs
> + * idling while there are pending tasks. Use scx_bpf_pick_any_cpu() and
> + * scx_bpf_kick_cpu() to guarantee that there will be at least one dispatch
> + * event in the near future.
> + *
> + * Unavailable if ops.update_idle() is implemented and
> + * %SCX_OPS_KEEP_BUILTIN_IDLE is not set.
> + */
> +s32 scx_bpf_pick_idle_cpu(const struct cpumask *cpus_allowed, u64 flags)
> +{
> +	if (!static_branch_likely(&scx_builtin_idle_enabled)) {
> +		scx_ops_error("built-in idle tracking is disabled");
> +		return -EBUSY;
> +	}
> +
> +	return scx_pick_idle_cpu(cpus_allowed, flags);
> +}
> +
> +/**
> + * scx_bpf_pick_any_cpu - Pick and claim an idle cpu if available or pick any CPU
> + * @cpus_allowed: Allowed cpumask
> + * @flags: %SCX_PICK_IDLE_CPU_* flags
> + *
> + * Pick and claim an idle cpu in @cpus_allowed. If none is available, pick any
> + * CPU in @cpus_allowed. Guaranteed to succeed and returns the picked idle cpu
> + * number if @cpus_allowed is not empty. -%EBUSY is returned if @cpus_allowed is
> + * empty.
> + *
> + * If ops.update_idle() is implemented and %SCX_OPS_KEEP_BUILTIN_IDLE is not
> + * set, this function can't tell which CPUs are idle and will always pick any
> + * CPU.
> + */
> +s32 scx_bpf_pick_any_cpu(const struct cpumask *cpus_allowed, u64 flags)
> +{
> +	s32 cpu;
> +
> +	if (static_branch_likely(&scx_builtin_idle_enabled)) {
> +		cpu = scx_pick_idle_cpu(cpus_allowed, flags);
> +		if (cpu >= 0)
> +			return cpu;
> +	}
> +
> +	cpu = cpumask_any_distribute(cpus_allowed);
> +	if (cpu < nr_cpu_ids)
> +		return cpu;
> +	else
> +		return -EBUSY;
> +}
> +
> +/**
> + * scx_bpf_get_idle_cpumask - Get a referenced kptr to the idle-tracking
> + * per-CPU cpumask.
> + *
> + * Returns NULL if idle tracking is not enabled, or running on a UP kernel.
> + */
> +const struct cpumask *scx_bpf_get_idle_cpumask(void)
> +{
> +	if (!static_branch_likely(&scx_builtin_idle_enabled)) {
> +		scx_ops_error("built-in idle tracking is disabled");
> +		return cpu_none_mask;
> +	}
> +
> +#ifdef CONFIG_SMP
> +	return idle_masks.cpu;
> +#else
> +	return cpu_none_mask;
> +#endif
> +}
> +
> +/**
> + * scx_bpf_get_idle_smtmask - Get a referenced kptr to the idle-tracking,
> + * per-physical-core cpumask. Can be used to determine if an entire physical
> + * core is free.
> + *
> + * Returns NULL if idle tracking is not enabled, or running on a UP kernel.
> + */
> +const struct cpumask *scx_bpf_get_idle_smtmask(void)
> +{
> +	if (!static_branch_likely(&scx_builtin_idle_enabled)) {
> +		scx_ops_error("built-in idle tracking is disabled");
> +		return cpu_none_mask;
> +	}
> +
> +#ifdef CONFIG_SMP
> +	if (sched_smt_active())
> +		return idle_masks.smt;
> +	else
> +		return idle_masks.cpu;
> +#else
> +	return cpu_none_mask;
> +#endif
> +}
> +
> +/**
> + * scx_bpf_put_idle_cpumask - Release a previously acquired referenced kptr to
> + * either the percpu, or SMT idle-tracking cpumask.
> + */
> +void scx_bpf_put_idle_cpumask(const struct cpumask *idle_mask)
> +{
> +	/*
> +	 * Empty function body because we aren't actually acquiring or
> +	 * releasing a reference to a global idle cpumask, which is read-only
> +	 * in the caller and is never released. The acquire / release semantics
> +	 * here are just used to make the cpumask is a trusted pointer in the
> +	 * caller.
> +	 */
> +}
> +
> +struct scx_bpf_error_bstr_bufs {
> +	u64			data[MAX_BPRINTF_VARARGS];
> +	char			msg[SCX_EXIT_MSG_LEN];
> +};
> +
> +static DEFINE_PER_CPU(struct scx_bpf_error_bstr_bufs, scx_bpf_error_bstr_bufs);
> +
> +/**
> + * scx_bpf_error_bstr - Indicate fatal error
> + * @fmt: error message format string
> + * @data: format string parameters packaged using ___bpf_fill() macro
> + * @data__sz: @data len, must end in '__sz' for the verifier
> + *
> + * Indicate that the BPF scheduler encountered a fatal error and initiate ops
> + * disabling.
> + */
> +void scx_bpf_error_bstr(char *fmt, unsigned long long *data, u32 data__sz)
> +{
> +	struct bpf_bprintf_data bprintf_data = { .get_bin_args = true };
> +	struct scx_bpf_error_bstr_bufs *bufs;
> +	unsigned long flags;
> +	int ret;
> +
> +	local_irq_save(flags);
> +	bufs = this_cpu_ptr(&scx_bpf_error_bstr_bufs);
> +
> +	if (data__sz % 8 || data__sz > MAX_BPRINTF_VARARGS * 8 ||
> +	    (data__sz && !data)) {
> +		scx_ops_error("invalid data=%p and data__sz=%u",
> +			      (void *)data, data__sz);
> +		goto out_restore;
> +	}
> +
> +	ret = copy_from_kernel_nofault(bufs->data, data, data__sz);
> +	if (ret) {
> +		scx_ops_error("failed to read data fields (%d)", ret);
> +		goto out_restore;
> +	}
> +
> +	ret = bpf_bprintf_prepare(fmt, UINT_MAX, bufs->data, data__sz / 8,
> +				  &bprintf_data);
> +	if (ret < 0) {
> +		scx_ops_error("failed to format prepration (%d)", ret);
> +		goto out_restore;
> +	}
> +
> +	ret = bstr_printf(bufs->msg, sizeof(bufs->msg), fmt,
> +			  bprintf_data.bin_args);
> +	bpf_bprintf_cleanup(&bprintf_data);
> +	if (ret < 0) {
> +		scx_ops_error("scx_ops_error(\"%s\", %p, %u) failed to format",
> +			      fmt, data, data__sz);
> +		goto out_restore;
> +	}
> +
> +	scx_ops_error_type(SCX_EXIT_ERROR_BPF, "%s", bufs->msg);
> +out_restore:
> +	local_irq_restore(flags);
> +}
> +
> +/**
> + * scx_bpf_destroy_dsq - Destroy a custom DSQ
> + * @dsq_id: DSQ to destroy
> + *
> + * Destroy the custom DSQ identified by @dsq_id. Only DSQs created with
> + * scx_bpf_create_dsq() can be destroyed. The caller must ensure that the DSQ is
> + * empty and no further tasks are dispatched to it. Ignored if called on a DSQ
> + * which doesn't exist. Can be called from any online scx_ops operations.
> + */
> +void scx_bpf_destroy_dsq(u64 dsq_id)
> +{
> +	destroy_dsq(dsq_id);
> +}
> +
> +/**
> + * scx_bpf_task_running - Is task currently running?
> + * @p: task of interest
> + */
> +bool scx_bpf_task_running(const struct task_struct *p)
> +{
> +	return task_rq(p)->curr == p;
> +}
> +
> +/**
> + * scx_bpf_task_cpu - CPU a task is currently associated with
> + * @p: task of interest
> + */
> +s32 scx_bpf_task_cpu(const struct task_struct *p)
> +{
> +	return task_cpu(p);
> +}
> +
> +BTF_SET8_START(scx_kfunc_ids_any)
> +BTF_ID_FLAGS(func, scx_bpf_dsq_nr_queued)
> +BTF_ID_FLAGS(func, scx_bpf_test_and_clear_cpu_idle)
> +BTF_ID_FLAGS(func, scx_bpf_pick_idle_cpu, KF_RCU)
> +BTF_ID_FLAGS(func, scx_bpf_pick_any_cpu, KF_RCU)
> +BTF_ID_FLAGS(func, scx_bpf_get_idle_cpumask, KF_ACQUIRE)
> +BTF_ID_FLAGS(func, scx_bpf_get_idle_smtmask, KF_ACQUIRE)
> +BTF_ID_FLAGS(func, scx_bpf_put_idle_cpumask, KF_RELEASE)
> +BTF_ID_FLAGS(func, scx_bpf_error_bstr, KF_TRUSTED_ARGS)
> +BTF_ID_FLAGS(func, scx_bpf_destroy_dsq)
> +BTF_ID_FLAGS(func, scx_bpf_task_running, KF_RCU)
> +BTF_ID_FLAGS(func, scx_bpf_task_cpu, KF_RCU)
> +BTF_SET8_END(scx_kfunc_ids_any)
> +
> +static const struct btf_kfunc_id_set scx_kfunc_set_any = {
> +	.owner			= THIS_MODULE,
> +	.set			= &scx_kfunc_ids_any,
> +};
> +
> +__diag_pop();
> +
> +/*
> + * This can't be done from init_sched_ext_class() as register_btf_kfunc_id_set()
> + * needs most of the system to be up.
> + */
> +static int __init register_ext_kfuncs(void)
> +{
> +	int ret;
> +
> +	/*
> +	 * Some kfuncs are context-sensitive and can only be called from
> +	 * specific SCX ops. They are grouped into BTF sets accordingly.
> +	 * Unfortunately, BPF currently doesn't have a way of enforcing such
> +	 * restrictions. Eventually, the verifier should be able to enforce
> +	 * them. For now, register them the same and make each kfunc explicitly
> +	 * check using scx_kf_allowed().
> +	 */
> +	if ((ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
> +					     &scx_kfunc_set_sleepable)) ||
> +	    (ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
> +					     &scx_kfunc_set_enqueue_dispatch)) ||
> +	    (ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
> +					     &scx_kfunc_set_dispatch)) ||
> +	    (ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS,
> +					     &scx_kfunc_set_any))) {
> +		pr_err("sched_ext: failed to register kfunc sets (%d)\n", ret);
> +		return ret;
> +	}
> +
> +	return 0;
> +}
> +__initcall(register_ext_kfuncs);
> diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
> index 6a93c4825339..d78d151fdbf8 100644
> --- a/kernel/sched/ext.h
> +++ b/kernel/sched/ext.h
> @@ -1,11 +1,119 @@
>  /* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
> + * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
> + * Copyright (c) 2022 David Vernet <dvernet@meta.com>
> + */
> +enum scx_wake_flags {
> +	/* expose select WF_* flags as enums */
> +	SCX_WAKE_EXEC		= WF_EXEC,
> +	SCX_WAKE_FORK		= WF_FORK,
> +	SCX_WAKE_TTWU		= WF_TTWU,
> +	SCX_WAKE_SYNC		= WF_SYNC,
> +};
> +
> +enum scx_enq_flags {
> +	/* expose select ENQUEUE_* flags as enums */
> +	SCX_ENQ_WAKEUP		= ENQUEUE_WAKEUP,
> +	SCX_ENQ_HEAD		= ENQUEUE_HEAD,
> +
> +	/* high 32bits are SCX specific */
> +
> +	/*
> +	 * The task being enqueued is the only task available for the cpu. By
> +	 * default, ext core keeps executing such tasks but when
> +	 * %SCX_OPS_ENQ_LAST is specified, they're ops.enqueue()'d with
> +	 * %SCX_ENQ_LAST and %SCX_ENQ_LOCAL flags set.
> +	 *
> +	 * If the BPF scheduler wants to continue executing the task,
> +	 * ops.enqueue() should dispatch the task to %SCX_DSQ_LOCAL immediately.
> +	 * If the task gets queued on a different dsq or the BPF side, the BPF
> +	 * scheduler is responsible for triggering a follow-up scheduling event.
> +	 * Otherwise, Execution may stall.
> +	 */
> +	SCX_ENQ_LAST		= 1LLU << 41,
> +
> +	/*
> +	 * A hint indicating that it's advisable to enqueue the task on the
> +	 * local dsq of the currently selected CPU. Currently used by
> +	 * select_cpu_dfl() and together with %SCX_ENQ_LAST.
> +	 */
> +	SCX_ENQ_LOCAL		= 1LLU << 42,
> +
> +	/* high 8 bits are internal */
> +	__SCX_ENQ_INTERNAL_MASK	= 0xffLLU << 56,
> +
> +	SCX_ENQ_CLEAR_OPSS	= 1LLU << 56,
> +};
> +
> +enum scx_deq_flags {
> +	/* expose select DEQUEUE_* flags as enums */
> +	SCX_DEQ_SLEEP		= DEQUEUE_SLEEP,
> +};
> +
> +enum scx_pick_idle_cpu_flags {
> +	SCX_PICK_IDLE_CORE	= 1LLU << 0,	/* pick a CPU whose SMT siblings are also idle */
> +};
>  
>  #ifdef CONFIG_SCHED_CLASS_EXT
> -#error "NOT IMPLEMENTED YET"
> +
> +struct sched_enq_and_set_ctx {
> +	struct task_struct	*p;
> +	int			queue_flags;
> +	bool			queued;
> +	bool			running;
> +};
> +
> +void sched_deq_and_put_task(struct task_struct *p, int queue_flags,
> +			    struct sched_enq_and_set_ctx *ctx);
> +void sched_enq_and_set_task(struct sched_enq_and_set_ctx *ctx);
> +
> +extern const struct sched_class ext_sched_class;
> +extern const struct bpf_verifier_ops bpf_sched_ext_verifier_ops;
> +extern const struct file_operations sched_ext_fops;
> +
> +DECLARE_STATIC_KEY_FALSE(__scx_ops_enabled);
> +#define scx_enabled()		static_branch_unlikely(&__scx_ops_enabled)
> +
> +static inline bool task_on_scx(struct task_struct *p)
> +{
> +	return scx_enabled() && p->sched_class == &ext_sched_class;
> +}
While building the kernel, I encountered the following warning:

{KERNEL_SRC}/kernel/sched/core.c: In function ‘__task_prio’:
{KERNEL_SRC}/kernel/sched/core.c:170:25: warning: passing argument 1 of ‘task_on_scx’ discards ‘const’ qualifier from pointer target type [-Wdiscarded-qualifiers]
  170 |         if (task_on_scx(p))
      |                         ^
In file included from {KERNEL_SRC}/kernel/sched/sched.h:3593,
                 from {KERNEL_SRC}/kernel/sched/core.c:86:
{KERNEL_SRC}/kernel/sched/ext.h:124:52: note: expected ‘struct task_struct *’ but argument is of type ‘const struct task_struct *’
  124 | static inline bool task_on_scx(struct task_struct *p)
      |                                ~~~~~~~~~~~~~~~~~~~~^

To address this warning, I'd suggest modifying the signature of `task_on_scx` to
accept `task_struct` argument as `const`. The proposed change is as follows: 

diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index 405037a4e6ce..e9c699a87770 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -121,7 +121,7 @@ DECLARE_STATIC_KEY_FALSE(__scx_switched_all);
 
 DECLARE_STATIC_KEY_FALSE(scx_ops_cpu_preempt);
 
-static inline bool task_on_scx(struct task_struct *p)
+static inline bool task_on_scx(const struct task_struct *p)
 {
        return scx_enabled() && p->sched_class == &ext_sched_class;
 }
@@ -214,7 +214,7 @@ bool scx_prio_less(const struct task_struct *a, const struct task_struct *b,
 #define scx_enabled()          false
 #define scx_switched_all()     false
 
-static inline bool task_on_scx(struct task_struct *p) { return false; }
+static inline bool task_on_scx(const struct task_struct *p) { return false; }
 static inline void scx_pre_fork(struct task_struct *p) {}
 static inline int scx_fork(struct task_struct *p) { return 0; }
 static inline void scx_post_fork(struct task_struct *p) {}
 
> +
> +bool task_should_scx(struct task_struct *p);
> +void scx_pre_fork(struct task_struct *p);
> +int scx_fork(struct task_struct *p);
> +void scx_post_fork(struct task_struct *p);
> +void scx_cancel_fork(struct task_struct *p);
> +void init_sched_ext_class(void);
> +
> +static inline const struct sched_class *next_active_class(const struct sched_class *class)
> +{
> +	class++;
> +	if (!scx_enabled() && class == &ext_sched_class)
> +		class++;
> +	return class;
> +}
> +
> +#define for_active_class_range(class, _from, _to)				\
> +	for (class = (_from); class != (_to); class = next_active_class(class))
> +
> +#define for_each_active_class(class)						\
> +	for_active_class_range(class, __sched_class_highest, __sched_class_lowest)
> +
> +/*
> + * SCX requires a balance() call before every pick_next_task() call including
> + * when waking up from idle.
> + */
> +#define for_balance_class_range(class, prev_class, end_class)			\
> +	for_active_class_range(class, (prev_class) > &ext_sched_class ?		\
> +			       &ext_sched_class : (prev_class), (end_class))
> +
>  #else	/* CONFIG_SCHED_CLASS_EXT */
>  
>  #define scx_enabled()		false
>  
> +static inline bool task_on_scx(struct task_struct *p) { return false; }
>  static inline void scx_pre_fork(struct task_struct *p) {}
>  static inline int scx_fork(struct task_struct *p) { return 0; }
>  static inline void scx_post_fork(struct task_struct *p) {}
> @@ -18,7 +126,13 @@ static inline void init_sched_ext_class(void) {}
>  #endif	/* CONFIG_SCHED_CLASS_EXT */
>  
>  #if defined(CONFIG_SCHED_CLASS_EXT) && defined(CONFIG_SMP)
> -#error "NOT IMPLEMENTED YET"
> +void __scx_update_idle(struct rq *rq, bool idle);
> +
> +static inline void scx_update_idle(struct rq *rq, bool idle)
> +{
> +	if (scx_enabled())
> +		__scx_update_idle(rq, idle);
> +}
>  #else
>  static inline void scx_update_idle(struct rq *rq, bool idle) {}
>  #endif
> diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
> index 17bd277cf27a..666166908eb6 100644
> --- a/kernel/sched/sched.h
> +++ b/kernel/sched/sched.h
> @@ -185,6 +185,10 @@ static inline int idle_policy(int policy)
>  
>  static inline int normal_policy(int policy)
>  {
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +	if (policy == SCHED_EXT)
> +		return true;
> +#endif
>  	return policy == SCHED_NORMAL;
>  }
>  
> @@ -681,6 +685,15 @@ struct cfs_rq {
>  #endif /* CONFIG_FAIR_GROUP_SCHED */
>  };
>  
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +struct scx_rq {
> +	struct scx_dispatch_q	local_dsq;
> +	u64			ops_qseq;
> +	u64			extra_enq_flags;	/* see move_task_to_local_dsq() */
> +	u32			nr_running;
> +};
> +#endif /* CONFIG_SCHED_CLASS_EXT */
> +
>  static inline int rt_bandwidth_enabled(void)
>  {
>  	return sysctl_sched_rt_runtime >= 0;
> @@ -1022,6 +1035,9 @@ struct rq {
>  	struct cfs_rq		cfs;
>  	struct rt_rq		rt;
>  	struct dl_rq		dl;
> +#ifdef CONFIG_SCHED_CLASS_EXT
> +	struct scx_rq		scx;
> +#endif
>  
>  #ifdef CONFIG_FAIR_GROUP_SCHED
>  	/* list of leaf cfs_rq on this CPU: */
> -- 
> 2.41.0
> 

Re: [PATCH 12/34] sched_ext: Implement BPF extensible scheduler class

[Tejun Heo]

Hello, Vishal.

On Wed, Aug 16, 2023 at 05:15:43PM +0530, Vishal Chourasia wrote:
> > +static inline bool task_on_scx(struct task_struct *p)
> > +{
> > +	return scx_enabled() && p->sched_class == &ext_sched_class;
> > +}
> While building the kernel, I encountered the following warning:
> 
> {KERNEL_SRC}/kernel/sched/core.c: In function ‘__task_prio’:
> {KERNEL_SRC}/kernel/sched/core.c:170:25: warning: passing argument 1 of ‘task_on_scx’ discards ‘const’ qualifier from pointer target type [-Wdiscarded-qualifiers]
>   170 |         if (task_on_scx(p))
>       |                         ^
> In file included from {KERNEL_SRC}/kernel/sched/sched.h:3593,
>                  from {KERNEL_SRC}/kernel/sched/core.c:86:
> {KERNEL_SRC}/kernel/sched/ext.h:124:52: note: expected ‘struct task_struct *’ but argument is of type ‘const struct task_struct *’
>   124 | static inline bool task_on_scx(struct task_struct *p)
>       |                                ~~~~~~~~~~~~~~~~~~~~^
> 
> To address this warning, I'd suggest modifying the signature of `task_on_scx` to
> accept `task_struct` argument as `const`. The proposed change is as follows: 
> 
> diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
> index 405037a4e6ce..e9c699a87770 100644
> --- a/kernel/sched/ext.h
> +++ b/kernel/sched/ext.h
> @@ -121,7 +121,7 @@ DECLARE_STATIC_KEY_FALSE(__scx_switched_all);
>  
>  DECLARE_STATIC_KEY_FALSE(scx_ops_cpu_preempt);
>  
> -static inline bool task_on_scx(struct task_struct *p)
> +static inline bool task_on_scx(const struct task_struct *p)
>  {
>         return scx_enabled() && p->sched_class == &ext_sched_class;
>  }
> @@ -214,7 +214,7 @@ bool scx_prio_less(const struct task_struct *a, const struct task_struct *b,
>  #define scx_enabled()          false
>  #define scx_switched_all()     false
>  
> -static inline bool task_on_scx(struct task_struct *p) { return false; }
> +static inline bool task_on_scx(const struct task_struct *p) { return false; }
>  static inline void scx_pre_fork(struct task_struct *p) {}
>  static inline int scx_fork(struct task_struct *p) { return 0; }
>  static inline void scx_post_fork(struct task_struct *p) {}

Yeah, this is already fixed in the github repo by the following commit:

 https://github.com/sched-ext/sched_ext/commit/56b278fa8b5136457993f7389e34070d35f17e8a

The fix will be included in the next iteration.

Thanks.

-- 
tejun

Re: [PATCHSET v4] sched: Implement BPF extensible scheduler class

[Mel Gorman]

On Thu, Jul 27, 2023 at 02:12:05PM -1000, Tejun Heo wrote:
> Hello, Peter.
> 
> On Wed, Jul 26, 2023 at 11:17:52AM +0200, Peter Zijlstra wrote:
> > On Fri, Jul 21, 2023 at 08:37:41AM -1000, Tejun Heo wrote:
> > > We are comfortable with the current API. Everything we tried fit pretty
> > > well. It will continue to evolve but sched_ext now seems mature enough for
> > > initial inclusion. I suppose lack of response doesn't indicate tacit
> > > agreement from everyone, so what are you guys all thinking?
> > 
> > I'm still hating the whole thing with a passion.
> > 
> > As can be seen from the wide-spread SCHED_DEBUG abuse; people are, in
> > general, not interested in doing the right thing. They prod random
> > numbers (as in really, some are just completely insane) until their
> > workload improves and call it a day.
> 
> I think it'd be useful to add some details to what's going on in situations
> like above. This of course wouldn't apply directly to everyone but I suspect
> many will recognize at least some parts of it.
> 

I cannot speak for Peter but I've seen numerous cases of similar abuse.
While there is selection bias, in that I only see bugs about failures and
not stories about successes, I've yet to see an instance where debugfs
tunables were used correctly. This includes the "tuned" tool where,
depending on the circumstance, a symbolic name for the tuning set may
have the opposite effect to what is desired in a specific circumstance.
In almost all cases where I recommended a debug tuning for a user, it
was to workaround a scheduler bug in an old kernel that was then fixed
upstream. In many cases, the backport was not feasible as the risk for
other regressions was too substantial.

The reasons for the mistakes vary. Sometimes it was because there was a
complete misunderstanding of what the parameter did. Others it was because
the side-effects were not taken into account. While I cannot remember
specifically, I think there were a few instances where a tuning parameter
that used to work did not work the same way in a later kernel. In at least
one case, the "tuning" happened to help a benchmark (which had nothing to
do with their workload) because it caused starvation issues which happened
to produce good numbers.

> In many production setups, there are aspects of workload behaviors that are
> difficult to understand comprehensively. The workloads are often massively
> complex, constantly being developed by many people, and dynamically
> interacting with external entities. As with any sufficiently complex system,
> there are many emergent properties which are difficult to untangle
> completely.
> 

While this is true, the same is also true for some of the scheduler
paramters.

> Add to that multiple generations of divergent hardware and most of the
> software stack coming from third parties (including kernel from application
> team's POV), people often and justifiably feel as if they're swimming in the
> sea of black boxes and emergent properties.
> 

Also true of the tuning parameters. In some cases, changes in hardware or
the software stack violate so many of the original assumptions that tuning
is not sufficient.

> Scheduling, naturally, is one of the areas that people look into when trying
> to optimize system performance. Vast majority of people don't know scheduler
> code base well enough to hack on it. Even when they do, it's often not easy
> to set up benchmarks in production environments and cycle through different
> kernels. We (Meta) are a lot better now than a couple years ago, but even
> now swapping kernels and ramping workloads back up can take a long time for
> certain workloads.
> 

But this complexity is not necessarily solved by introducing pluggable
schedulers. The level of expertise required to hack on the scheduler or
hack on a pluggable scheduler both require a solid understanding of
scheduling in general and the ability to analyse what is going on.

> Given the circumstances, it's not surprising that people go for tunable
> knobs when they're trying to find out whether changing scheduling behaviors
> would improve performance for their workloads. That's often the only option
> available and tuning the knobs frequently leads to some gains. Most people
> aren't scheduling experts and the causal relationships between changes and
> results may not be direct or intuitive. So, that's often where things end.
> Given that nobody has found scheduling behavior which is optimal for every
> workload and the SCHED_DEBUG knobs are what people can access, it is an
> expected outcome.
> 

And while those tuning options are used, usually by searching for a stock
tuning guide for an arbitrary workload/platform and hoping for the best,
it's often the case that any gain is co-incidental. Worse, I've seen far
too many cases where there was no test for significance before/after to
ensure the tuning parameter is actually helping. Worse again, I've seen
tuning for some completely random workload and then hoping that this is
somehow relevant to the target workload (spoiler: it almost never is).

> If a consistent pattern is repeated across multiple workloads, we can
> sometimes work back why tuning certain way makes sense and generalize that,
> which is to some degree how we ended up focusing on recent work-conservation
> related projects.
> 
> Maybe the situation is not ideal but I don't think it's people not being
> interested in doing the right thing. They are doing what they can within the
> confines of available mechanisms, expertise, and time & effort they can
> afford to invest.
> 
> One of the impediments when trying to connect these disparate data points
> into something meaningful is the difficulty in experimentation. The trials
> are confined to whatever combinations that can be achieved with SCHED_DEBUG
> knobs which are both limiting and obscuring. I believe we're a lot more
> likely to learn more about scheduling with sched_ext widely available than
> without as it would allow easier and wider-in-scope experimentations.
> 

An apparent assumption there is "the impact of sched debug tuning is
too hard to understand but sched_ext and writing your own scheduler is
easier". That's .... a big leap.

> > There is not a single doubt in my mind that if I were to merge this,
> > there will be Enterprise software out there that will mandate its own
> > BPF sched thing, or else it won't work.
> >
> > They will not care, they will not contribute, they might even pull a
> > RedHat and only share the code to customers.
> 
> I'm sure some will behave in a way which isn't the most conducive to
> collective improvement of the upstream kernel. That said, I don't see how
> this will be noticeably worsened by inclusion of sched_ext. Most mobile
> kernels and some production kernels in cloud environments already carry
> significant custom modifications, and they're often addressing real problems
> for their use cases.
> 

Peter answered this

	There is not a single doubt in my mind that if I were to merge this,
	there will be Enterprise software out there that will mandate its
	own BPF sched thing, or else it won't work.

While I have not seen this class of problem for scheduler in particular,
I've deal with too many bugs that required out-of-tree components to be
loaded and a requirement that the bug be fixed in that context.
Scheduler bugs are already very tricky without adding opaque blobs.

> It'd be ideal if everyone had the commitment and bandwidth to try their best
> to merge back their changes but it's also understandable why that can't
> always be the case. Sometimes, it's too specific or underdeveloped. At other
> times, time and resources just aren't there. We can incentivize and coerce
> but that can be pushed only so far. However, we do have an a lot easier time
> learning about what people are doing thanks to GPL which all sched_ext
> programs would need to follow exactly like the rest of the kernel.
> 

By this logic, you expect sched_ext usage to grow with special cases
that may never be compatible with the core scheduler.

> At least relatively speaking, scheduling doesn't seem like an area which is
> particularly starved for developer bandwidth although one can always hope
> for more.

There are a fair few active developers but I know my own backlog for
scheduler review is severe and I doubt it's just me. The focus is also
diffuse as scheduler has a lot of parts. There might be a lot of activity
in EAS for example at a point in time, but that's a specific use case.

> > <SNIP>
> > We all loose in that scenario. Not least me, because I get the
> > additional maintenance burden.
> 
> sched_ext isn't that invasive to the core code and its interactions with
> other scheduling classes are very limited.

On the flip side, there are some layering violations between the fair
scheduler and core so modifying those without breaking an arbitrary
external scheduler may be problematic. Maintenance also isn't just code.
Dealing with regressions in scheduler or even just changes in hardware
is already quite difficult, particularly as it's not always possible to
get high quality data or reproduce the environment. That is orders of
magnitude more difficult if an unknown scheduler is involved.

There is also the possibility that running within containers becomes harder
if different containers require different schedulers or a custom scheduler
helps one container and hurts another.

> This would make changing
> scheduling core API a bit more burdensome but they have been relatively
> stable and both David and I would be on the hook if anything is in your way.
> I don't see why this would significantly increase your maintenance burden.
> It's a thing but it's a thing in its own corner.
> 

I disagree about the maintenance burden. I think it's inevitable that some
enterprise software will require it and are unwilling or unable to modify the
upstream scheduler to meet their requirements. There may be even be incentive
for them to use a custom scheduler to game benchmark comparisons and not
contribute it back. Distributions would be faced with the stark choice of
"disable sched_ext and hope some major software stack does not require it"
or "try support arbitrary black box schedulers when regressions occur". Both
choices are unpalatable and while this is not directly a maintenance issue
upstream, it's an indirect one when dealing with regressions after a kernel
version upgrade. The problems faced by distributions are not the same as
those faced by Google or Meta because distributions have limited to no
control over what software stack runs on the OS.

While I don't have the authority to NAK a custom scheduler framework, I
am opposed to it conceptually because I suspect supporting it in the
field will be a nightmare.

-- 
Mel Gorman
SUSE Labs

Re: [PATCHSET v4] sched: Implement BPF extensible scheduler class

[Tejun Heo]

Hello, Mel.

On Thu, Aug 17, 2023 at 01:44:57PM +0100, Mel Gorman wrote:
> > > As can be seen from the wide-spread SCHED_DEBUG abuse; people are, in
> > > general, not interested in doing the right thing. They prod random
> > > numbers (as in really, some are just completely insane) until their
> > > workload improves and call it a day.
> > 
> > I think it'd be useful to add some details to what's going on in situations
> > like above. This of course wouldn't apply directly to everyone but I suspect
> > many will recognize at least some parts of it.
> 
> I cannot speak for Peter but I've seen numerous cases of similar abuse.
> While there is selection bias, in that I only see bugs about failures and
> not stories about successes, I've yet to see an instance where debugfs
> tunables were used correctly. This includes the "tuned" tool where,
> depending on the circumstance, a symbolic name for the tuning set may
> have the opposite effect to what is desired in a specific circumstance.
> In almost all cases where I recommended a debug tuning for a user, it
> was to workaround a scheduler bug in an old kernel that was then fixed
> upstream. In many cases, the backport was not feasible as the risk for
> other regressions was too substantial.

I'm sure there are ample examples of knobs being misused as with anything.
However, given that there are quite a few heuristics involved in making
scheduling decisions, which are inherently parameterized and have to make
trade-offs, claiming that the scheduler can do away with exposing these
parameters sounds impractically optimistic.

For example, migration_cost_ns determines the trade-off between cache
locality and work conservation. It materially impacts tail latencies,
maximum bandwidth, and power consumption. The optimal point doesn't just
depend on the hardware, the type of workload or its current load level but
also what matters to the specific use case of the given workload.

Also, while the default parameters are the results of many benchmarks and
years of experience, the kernel is used for all kinds of things and we are
always subject to exposure bias. Whoever is working on the code and/or
posting results gets to influence how the code is structured and these
parameters are determined. There's nothing wrong with that but it does point
to the fact that there are going to be sizable holes and biases in our
evaluations.

One may argue that problems like the above can be solved by the scheduler
being adaptive with the applications and admins providing enough hints on
the intents. Maybe that is possible but it certainly doesn't seem like we're
anywhere close. As an ideal, it may make sense but I'm skeptical it's
tracking reality.

> The reasons for the mistakes vary. Sometimes it was because there was a
> complete misunderstanding of what the parameter did. Others it was because
> the side-effects were not taken into account. While I cannot remember
> specifically, I think there were a few instances where a tuning parameter
> that used to work did not work the same way in a later kernel. In at least
> one case, the "tuning" happened to help a benchmark (which had nothing to
> do with their workload) because it caused starvation issues which happened
> to produce good numbers.

It is difficult to compare anecdotes with much objectivity but I'd like to
offer some counter-points.

What you're describing is very different from what I see. I often see
talented engineers working tirelessly to establish practical performance and
reliability metrics, constantly improve the tools to measure and run
experiments, and optimize their workloads and configuration accordingly.
People of course have different areas of expertise, differing levels of
insights and make mistakes which make for good stories, but I don't
recognize what you're describing as the prevalent case.

Both of us could be right. Different segments of the industry operate in
different manners and maybe more importantly interact in their specific
ways. Depending on one's exposure, I can imagine the general impression
being swayed one way or the other.

While we don't want to unnecessarily encourage users to shoot their feet, we
do want what's presented to userspace to be true to the technical and
practical reality. Otherwise, we run the risk of getting in the way of users
doing things. We don't have the perfect scheduler for all, or even most,
users. We do have a pretty good one with reasonable default behaviors and a
bunch of tunable parameters. What's presented to the world should match that
reality.

Speaking for Meta, these knobs really are critical. We see significant
bandwidth and latency gains (bandwidth usually few percents, larger tail
latency impacts) across multiple major workloads. We ate the operational
costs of moving our infra to using debugfs instead of sysctls (where
appropriate) because the capacity losses were untenable otherwise. While
this wasn't the end of the world, it was perplexing why this change was
made, as it could easily have broken just as many valid users as it might
block misuses. If something is needed, users will adapt to use the feature,
making such churn pointless. I suppose this is a reflection of how we
perceive the reality differently.

...
> > Scheduling, naturally, is one of the areas that people look into when trying
> > to optimize system performance. Vast majority of people don't know scheduler
> > code base well enough to hack on it. Even when they do, it's often not easy
> > to set up benchmarks in production environments and cycle through different
> > kernels. We (Meta) are a lot better now than a couple years ago, but even
> > now swapping kernels and ramping workloads back up can take a long time for
> > certain workloads.
> 
> But this complexity is not necessarily solved by introducing pluggable
> schedulers. The level of expertise required to hack on the scheduler or
> hack on a pluggable scheduler both require a solid understanding of
> scheduling in general and the ability to analyse what is going on.

I think there's some disconnect on how sched_ext can improve situations like
this. As this comes up again later, let's discuss there.

> > Given the circumstances, it's not surprising that people go for tunable
> > knobs when they're trying to find out whether changing scheduling behaviors
> > would improve performance for their workloads. That's often the only option
> > available and tuning the knobs frequently leads to some gains. Most people
> > aren't scheduling experts and the causal relationships between changes and
> > results may not be direct or intuitive. So, that's often where things end.
> > Given that nobody has found scheduling behavior which is optimal for every
> > workload and the SCHED_DEBUG knobs are what people can access, it is an
> > expected outcome.
> 
> And while those tuning options are used, usually by searching for a stock
> tuning guide for an arbitrary workload/platform and hoping for the best,
> it's often the case that any gain is co-incidental. Worse, I've seen far
> too many cases where there was no test for significance before/after to
> ensure the tuning parameter is actually helping. Worse again, I've seen
> tuning for some completely random workload and then hoping that this is
> somehow relevant to the target workload (spoiler: it almost never is).

As before, the above doesn't agree with what I see at all. Even small
performance differences count for huge eventual impacts at scale for both
servers and personal devices including e.g. VR headsets. There are a lot of
smart engineers poring over scheduling traces and profiling outputs trying
to optimize things as much as they can with the tools they have available.

We could be looking at the two extreme ends of the user competence spectrum.
The incompetent end is more entertaining but emphasizing that too much can
lead to inflated evaluation of ourselves and underestimation of others,
which is a dangerous place to be. While I don't have statistical data to
prove which perception is correct, it may still be useful to imagine what
the reality would be like if it were to match what you're describing.

We both agree that there are evidences showing that users often try to and
actually tune the scheduler parameters. If the perception that most, if not
nearly all, of these cases are misguided and don't even achieve anything
were to be true, the wider industry would have to be filled mostly with
incompetence and cargo-culting. That's a very stark view of the world.

Wouldn't a more reasonable interpretation be that what's suggested is a real
and persistent need that many use-cases encounter and that the effectiveness
of the actions that people take follow some normal-like distribution? There
will be ample silliness at one tail end, clever feats of engineering at the
other, with most cases doing something which isn't quite optimal but not
completely stupid either.

...
> > One of the impediments when trying to connect these disparate data points
> > into something meaningful is the difficulty in experimentation. The trials
> > are confined to whatever combinations that can be achieved with SCHED_DEBUG
> > knobs which are both limiting and obscuring. I believe we're a lot more
> > likely to learn more about scheduling with sched_ext widely available than
> > without as it would allow easier and wider-in-scope experimentations.
> 
> An apparent assumption there is "the impact of sched debug tuning is
> too hard to understand but sched_ext and writing your own scheduler is
> easier". That's .... a big leap.

The followings are what I understood as your arguments from the earlier
paragraph on difficulties of working on the scheduler and the above one.
Please correct me if I got them wrong:

1. Understanding and working on the scheduler requires a high level of
   expertise.

2. Pluggable schedulers won't contribute to improving scheduling in general
   because most won't be able to work on it anyway.

3. Pluggable schedulers won't add much in terms of making understanding and
   working on schedulers easier. Specifically, understanding the scheduling
   behavior of a sched_ext scheduler would not be easier than understanding
   how to competently apply CFS debug knobs.

Scheduling is a specialized area and possibly a more challenging one than
most. However, I don't think it's so much more difficult than others to the
point where wider participation is unrealistic. Here are some counterpoints
to the claim that not many people would be able to use or benefit from
pluggable schedulers (#1 and #2):

* Filesystems are difficult but people come up with new ideas and
  implementations all the time. If anything, we have too many, not too few.
  With all the problems that entails, this is a huge net benefit in the long
  term. That active experimentation and competition mean that there are many
  more and better options for filesystems now than there were 20 years ago.

* How a subsystem is structured and operated has a strong influence on how
  many would be active in the area. Scheduling is pretty exclusive. Even
  then, we constantly see attempts at implementing something new from
  different people. Even just the currently active full implementations
  include MuQSS, PDS, BMQ and TT and there are more focused efforts like the
  nest scheduler or BORE.

* Many application engineers have a very detailed understanding of how their
  application behaves, sometimes down to specific scheduling misbehaviors
  they want to address. Most wouldn't work on scheduler code directly but
  people who are more familiar with the scheduler can partner with them.
  Also, providing them more canned tools and space to explore would lead to
  discovery of useful strategies.

Writing a scheduler is difficult (#2 and #3). However, there are ways to
make these things easier, safer and more accessible. With sched_ext:

* Scheduler implementation can't crash the machine. There are multiple ways
  to safely revert to the default scheduler, and a watchdog exists which
  will boot the scheduler out if it’s failing to schedule tasks in a timely
  manner.

* The API is structured so that it's easier to write schedulers. One can get
  a basic vtime based scheduler going with a few tens of lines of code which
  can be good enough for personal or even some production use. It really
  doesn't take much to experiment in areas like CPU selection and ad-hoc
  soft-affinity.

* Iterating through implementations is as easy as changing-code, compiling
  and reloading. The worst that can happen to the machine is hanging due to
  some threads being forgotten or starved. You can easily restore with a
  sysrq or wait a bit until the watchdog kicks in and try again.

* People don't have to start from scratch. We can build up repertoire of
  examples, best practices and libraries to make it even easier.

Combined, I believe sched_ext significantly lowers the barrier of entry.

As for the argument that scheduling behavior won't be easier to understand
with sched_ext (#3), I wonder whether drawing parallels to a debugging
session would be useful.

When debugging a subtle and difficult bug caused by a combination of dynamic
behaviors, the ability to modify behaviors is critical. I want to know
what's happening and what happens if I change a part of the logic in
question. Less so with better visibility tools nowadays but debugging
sessions can be repeated iterations of adding [trace_]printk()'s and
eliminating possibilities by modifying code piecemeal.

Identifying and fixing lost performance due to suboptimal, complex
scheduling behaviors can have a lot of similarities. If you only have three
knobs you can twiddle, there's only so much you can learn. The configuration
space afforded by the knobs is unlikely to fully cover what you want to try.
However, if you can change and modulate every behavior safely and quickly,
you can learn a lot more a lot faster.

...
> > I'm sure some will behave in a way which isn't the most conducive to
> > collective improvement of the upstream kernel. That said, I don't see how
> > this will be noticeably worsened by inclusion of sched_ext. Most mobile
> > kernels and some production kernels in cloud environments already carry
> > significant custom modifications, and they're often addressing real problems
> > for their use cases.
> 
> Peter answered this
> 
> 	There is not a single doubt in my mind that if I were to merge this,
> 	there will be Enterprise software out there that will mandate its
> 	own BPF sched thing, or else it won't work.
> 
> While I have not seen this class of problem for scheduler in particular,
> I've deal with too many bugs that required out-of-tree components to be
> loaded and a requirement that the bug be fixed in that context.
> Scheduler bugs are already very tricky without adding opaque blobs.

This seems to stem from distro POV more than upstream (although not
completely). As this comes up again later, let's discuss there.

> > It'd be ideal if everyone had the commitment and bandwidth to try their best
> > to merge back their changes but it's also understandable why that can't
> > always be the case. Sometimes, it's too specific or underdeveloped. At other
> > times, time and resources just aren't there. We can incentivize and coerce
> > but that can be pushed only so far. However, we do have an a lot easier time
> > learning about what people are doing thanks to GPL which all sched_ext
> > programs would need to follow exactly like the rest of the kernel.
> 
> By this logic, you expect sched_ext usage to grow with special cases
> that may never be compatible with the core scheduler.

Prediction is always challenging but my gut feeling is that most users
wouldn't care to adopt custom schedulers. It just won't make meaningful
differences. In some quarters tho, custom schedulers may become more
popular. For example, it's already more prevalent to run custom schedulers
in gaming communities because their requirements aren't fully served by the
default scheduler. There also are use cases which require lower latency at
the cost of everything else or even ones which must implement specific
scheduling behavior incompatible with generic use such as ARINC 653 for
avionics.

Given how widely Linux is used these days, the fact that one size doesn't
fit all shouldn't be too surprising and I don't believe this necessarily
hampers the default scheduler. Code should be just as available as any
kernel modifications, and, when implementing the same logic, the builtin
scheduler would always have some advantages.

The hardware is changing really fast with increased core counts, more
complex cache topologies with core distances that weren't common before,
more and more dynamic clock scaling and heterogeneous CPUs. The use cases
keep becoming wider and more diverse too. The problem space is expanding
faster than what can be reasonably mapped out with a single strictly
controlled scheduler implementation. In the long term, I strongly believe
widening the field would benefit everyone including the users of the default
scheduler.

...
> > sched_ext isn't that invasive to the core code and its interactions with
> > other scheduling classes are very limited.
> 
> On the flip side, there are some layering violations between the fair
> scheduler and core so modifying those without breaking an arbitrary
> external scheduler may be problematic. Maintenance also isn't just code.

There's quite a bit of abstraction between the scheduler core API and the
API sched_ext exposes. The only part that needs to be maintained as core
code changes should be the sched_ext core, not the BPF scheduler
implementations.

> Dealing with regressions in scheduler or even just changes in hardware
> is already quite difficult, particularly as it's not always possible to
> get high quality data or reproduce the environment. That is orders of
> magnitude more difficult if an unknown scheduler is involved.

I don't quite follow. If any scheduling related issue is suspected, wouldn't
the first step of debugging be excluding that by switching the scheduler?
And if the problem goes away without the scheduler, the problem is for that
scheduler to solve, right? It may add some triaging overhead but I don't
quite understand how it would make things orders of magnitude more
difficult.

Maybe it's not the best analogy but the existence of btrfs doesn't make ext4
problems significantly more difficult to debug. With sched_ext, isolating
scheduling problems can even become easier in some cases as the schedulers
can be switched dynamically.

> There is also the possibility that running within containers becomes harder
> if different containers require different schedulers or a custom scheduler
> helps one container and hurts another.

This is definitely something we've been thinking about given how
containerized our environments are. Google's ghOSt scheduler has the concept
of enclaves where different schedulers can be used on different CPUs.
sched_ext didn't go that route mostly because this seemed like a problem
which can be solved at the BPF layer where the framework layer itself which
manages multiple active schedulers is implemented in BPF. This would have
the advantage of more flexibility - e.g. by allowing custom stealing
mechanism across partitions to improve work conservation.

While there are some existing examples in networking where this kind of BPF
program layering is used, it's likely that BPF would need more features to
properly support such layering for sched_ext. It is also possible that
supporting enclave equivalent at the sched_ext core layer like ghOSt will
eventually be chosen, which has a lot of pros too.

So, yeah, this is definitely an area of future development.

> > This would make changing
> > scheduling core API a bit more burdensome but they have been relatively
> > stable and both David and I would be on the hook if anything is in your way.
> > I don't see why this would significantly increase your maintenance burden.
> > It's a thing but it's a thing in its own corner.
> 
> I disagree about the maintenance burden. I think it's inevitable that some
> enterprise software will require it and are unwilling or unable to modify the
> upstream scheduler to meet their requirements. There may be even be incentive
> for them to use a custom scheduler to game benchmark comparisons and not
> contribute it back. Distributions would be faced with the stark choice of
> "disable sched_ext and hope some major software stack does not require it"
> or "try support arbitrary black box schedulers when regressions occur". Both
> choices are unpalatable and while this is not directly a maintenance issue
> upstream, it's an indirect one when dealing with regressions after a kernel
> version upgrade. The problems faced by distributions are not the same as
> those faced by Google or Meta because distributions have limited to no
> control over what software stack runs on the OS.

That is a very good point. It's probably useful to distinguish the impacts
on upstream, distros and the big users that have better control of software
stack which obviously includes the server side of Google and Meta but also
encompasses Android, ChromeOS, VR headsets, gaming devices and so on.

Upstream, we've already discussed. sched_ext shouldn't create huge problems
for big users as they can opt-in and out as they see fit. Especially if
you're already deploying a lot of BPF, much of the infrastructure and
institutional knowledge is already there.

Distros are in a harder spot as sched_ext may add a debug surface which is
unfamiliar and distros are often forced (or rather paid) to solve these
problems. However, it might be useful to take a step back and take a wider
view. Usually, things don't just have costs. They have benefits too.

I'm sure there are pathological cases, but if users and enterprise software
vendors want to use sched_ext, which will probably be a small subset, it's
likely that there are benefits that they see. They would do so because it
enables them to do things which weren't easily possible before. The
ecosystem is split differently on company boundaries but if big users like
Meta and Google find applications which can benefit from custom scheduling
(e.g. remote scheduling for VM-only CPUs), there probably are distro users
who can benefit in a similar fashion.

In principle, enabling customers to achieve more is the role that distros
play and are paid for. So, yes, this would add support cost, but that'd be
because the customers can do more, thus incentivizing them to use the
distro.

This isn't something special about sched_ext. Any kernel subsystem used in
production isn't free, but like any tool, one has to decide if the benefits
outweigh the costs. Meta is using btrfs and BPF heavily, so there are
substantial support costs for both. However, we happily pay them because the
benefits are far greater. For example, at this point, we wouldn't be able to
run our fleet without BPF because the capacity loss would be too drastic.

I'm sure distro users would see similar gains (e.g. w/ firewalls) and
request their distros to support BPF. If a distro chooses to support btrfs,
ext4 or BPF, the distro has to have the capability, and, in general, such
decisions should and would be mutually beneficial.

Thanks.

-- 
tejun

Re: [PATCHSET v4] sched: Implement BPF extensible scheduler class

[Josh Don]

Hello,

On Fri, Jul 21, 2023 at 11:37 AM Tejun Heo <tj@kernel.org> wrote:
[snip]
> We are comfortable with the current API. Everything we tried fit pretty
> well. It will continue to evolve but sched_ext now seems mature enough for
> initial inclusion. I suppose lack of response doesn't indicate tacit
> agreement from everyone, so what are you guys all thinking?

I want to reiterate Google’s support for this proposal.

We’ve been experimenting with pluggable scheduling via our ghOSt
framework (https://github.com/google/ghost-kernel) for quite a while
now. A few things have become clearly evident.

(1) There is a very non-trivial level of headroom that can be taken
advantage of by directed policy that more closely specializes to the
types of workloads deployed on a machine. I can provide two direct
examples.
In Search, the backend application has intimate knowledge of thread
workloads and RPC deadlines, which it immediately communicates to our
BPF scheduler via BPF maps. We've used this info to construct a policy
that reduces context switches, decreases p99 latency, and increases
QPS by 5% in testing. The flexibility of expressiveness in terms of
priority goes far beyond what niceness or cpu.shares could achieve.

For VM workloads, we’ve been testing a policy that has virtually
eliminated our >10ms latency tails via a combination of deadline and
fair scheduling, using an approach inspired by Tableau
(https://arpangujarati.github.io/pdfs/eurosys2018_paper.pdf). I find
this case particularly appealing from a pluggable scheduling
perspective because it highlights an area in which specialization to
the type of workload (VMs, which prefer longer, gang scheduled,
uninterrupted, and predictable low-latency access to CPU) provides
clear benefits, yet is not appropriate for a general-purpose scheduler
like CFS.

(2) Sched knobs are incredibly useful, and tuning has real effects.
The scheduler exports various debugfs knobs to control behavior, such
as minimum granularity, overall sched latency, and migration cost.
These mostly get baked into the kernel with semi-arbitrary values.
But, experimentally, it makes a lot of sense to (for example) lower
the migration cost on certain combinations of hardware and workload,
taking a tradeoff to increase migration rate but reduce non-work
conserving behavior.

We’ve taken this idea further with an ML based system to automatically
find the best combination of sched knobs for a given workload, given a
goal such as to maximize QPS. This has resulted in gains of 2-5%; a
lot of performance to leave on the table simply due to using preset
defaults. Pluggable scheduling would further increase the surface area
of experimentation, and yield additional insight into what other
kernel heuristics could be improved. It was from the ML work that we
gleaned that migrating tasks in smaller batches, but more frequently,
was a better tradeoff than the default configuration.

(3) There are a number of really interesting scheduling ideas that
would be difficult or infeasible to bake into the kernel. One clear
example is core scheduling, which was quite a complex addition to the
kernel (for example, due to managing fairness of tasks spanning the
logical cpus of a core), but which has relatively straightforward
implementation in sched_ext and ghOSt (for example, in ghOSt, a single
cpu can issue a transaction to run tasks on both itself and its
sibling, achieving the needed security property of core scheduling.
Fairness follows easily because runqueues can easily be any shape in
userspace, such as per-core.).
Another interesting idea is to offload scheduling entirely from VM
cores in order to keep ticks stopped with NOHZ independent of the task
count, since preemptive scheduling can be driven by a remote core.


Moving forward, we’re planning to redesign our ghOSt userspace
infrastructure to work on top of the sched_ext kernel infrastructure.
We think there’s a lot of benefit to the sched_ext design, especially
the very tight BPF integration. We’re committed to the idea of
pluggable scheduling, and are in close collaboration with Meta to
advance this work while we simultaneously deploy it internally.

Best,
Josh

Re: [PATCHSET v4] sched: Implement BPF extensible scheduler class

[Tejun Heo]

Hello, Mel.

I don't think the discussion has reached a point where the points of
disagreements are sufficiently laid out from both sides. Do you have any
further thoughts?

Thanks.

-- 
tejun

Re: [PATCHSET v4] sched: Implement BPF extensible scheduler class

[Mel Gorman]

On Tue, Sep 19, 2023 at 07:56:01AM -1000, Tejun Heo wrote:
> Hello, Mel.
> 
> I don't think the discussion has reached a point where the points of
> disagreements are sufficiently laid out from both sides. Do you have any
> further thoughts?
> 

Plenty, but I'm not sure how to reconcile this. I view pluggable scheduler
as something that would be a future maintenance nightmare and our "lived
experience" or "exposure bias" with respect to the expertise of users differs
drastically. Some developers will be mostly dealing with users that have
extensive relevant expertise, a strong incentive to maximise performance
and full control of their stack, others do not and the time cost of
supporting such users is high. While I can see advantages to having specific
schedulers targeting either a specific workload or hardware configuration,
the proliferation of such schedulers and the inevitable need to avoid
introducing any new regressions in deployed schedulers will be cumbersome.

I generally worry that certain things may not have existed in the shipped
scheduler if plugging was an option including EAS, throttling control,
schedutil integration, big.Little, adapting to chiplets and picking preferred
SMT siblings for turbo boost. In each case, integrating support was time
consuming painful and a pluggable scheduler would have been a relatively
easy out that would ultimately cost us if it was never properly integrated.
While no one wants the pain, a few of us also want to avoid the problem
of vendors publishing a hacky scheduler for their specific hardware and
discontinuing the work at that point.

I see that some friction with the current state is due to tuning knobs
moving to debugfs. FWIW, I didn't 100% agree with that move either and
carried an out-of-tree revert that displayed warnings for a time but I
understood the logic behind it. However, if the tuning parameters are
insufficient, and there is good reason to change them then the answer
is to add tuning knobs with defined semantics and document them -- not
pluggable schedulers. We've seen something along those lines recently
with nice_latency even if it turned into EEVDF instead of a new interface,
so I guess we'll see how that pans out.

I get most of your points. Maybe most users will not care about a pluggable
scheduler but *some will* and they will the maintenance burden. I get your
point as well that if there is a bug and the pluggable scheduler then the
first step would be "reproduce without the pluggable scheduler" and again,
you'd be right, that is a great first step *except* sometimes they can't or
sometimes they simply won't without significant proof and that's incurs a
maintenance burden. Even if the pluggable schedulers are GPL, there still
is a burden to understood any scheduler that is loaded to see if it's the
source of a problem which means. Instead of understanding a defined number
of schedulers that are developed over time with the history in changelogs,
we may have to understand N schedulers that may be popular and that also
is painful. That's leaving aside the difficulty of what happens when
more than 1 can be loaded and interacting once containers are involved
assuming that such support would exist in the future. It's already known
that interacting IO schedulers are a nightmare so presumably interacting
CPU schedulers within the same host would also be zero fun.

Pluggable schedulers are effectively a change that we cannot walk back
from if it turns out to be a bad idea because it potentially comes under
the "you cannot break userspace" rule if a particular pluggable scheduler
becomes popular. As I strongly believe it will be a nightmare to support
within distributions where there is almost no control over the software
stack of managing user expectations, I'm opposed to crossing that line with
pluggable schedulers. While my nightmare scenarios may never be realised
and could be overblown, it'll be hard to convince me it'll not kick me in
the face eventually.

-- 
Mel Gorman
SUSE Labs

Re: [PATCHSET v4] sched: Implement BPF extensible scheduler class

[Tejun Heo]

Hello, Mel.

On Tue, Sep 26, 2023 at 10:20:20AM +0100, Mel Gorman wrote:
> Plenty, but I'm not sure how to reconcile this. I view pluggable scheduler
> as something that would be a future maintenance nightmare and our "lived
> experience" or "exposure bias" with respect to the expertise of users differs
> drastically. Some developers will be mostly dealing with users that have
> extensive relevant expertise, a strong incentive to maximise performance
> and full control of their stack, others do not and the time cost of
> supporting such users is high.

My experience working for distros is shorter and less extensive than yours
but I believe I can appreciate the pain points to some degree. Getting
dropped into an enterprise software problem where most of the environment is
walled off can get really frustrating and pluggable schedulers can add
unfamiliar problems on top.

That said, I'd like to reiterate two counter points:

First, there are multiple substantial non-distro communities that use and
work on the kernel. Hyper-scalers that operate huge fleets of machines such
as Google and Meta among others, public cloud operators such as Amazon and
Google, hardware projects including smartphones, watches, VR headsets and so
on.

Because these communities tend to have more vertical integration across the
software stack and extremely high usage fan-out, development resource
allocation and the boundaries of what can be reasonably tried are different
from distros. In addition, the cost of failing to exploit optimization
opportunities is very direct and often extremely high. The cost-benefit
balance is pretty one sided in the favorable direction for these
communities.

Second, if a distro chooses to support sched_ext, there will be new kinds of
problems but that wouldn't be for nothing. While there will be stupid cases,
overall, the new problems will be there because users can now do things that
they couldn't do before. As I wrote before, in principle at least, adding
more capabilities should be a mutually beneficial proposition for both the
distros and their users.

> While I can see advantages to having specific
> schedulers targeting either a specific workload or hardware configuration,
> the proliferation of such schedulers and the inevitable need to avoid
> introducing any new regressions in deployed schedulers will be cumbersome.

I'm having a bit of a hard time following here. If someone (or group) writes
a custom scheduler, they'd be responsible for maintaining that, right? It
shouldn't significantly increase the maintenance burden somewhere centrally.
FUSE can be an imperfect analogy. There are numerous FUSE implementations.
While they sometimes expose shared underlying issues, the proliferation
doesn't usually lead to huge extra maintenance overhead.

If you're referring to the maintenance of sched_ext itself, yes, people
would get sad if things get slower or break and we'd want to avoid such
situations if we reasonably can. However, we absolutely can make breaking
changes if necessary. Here, BCC tools can serve as the imperfect analogy
(other BPF use cases often share similar characteristics). BCC tools are
coupled with the kernel in more fragile and opportunistic ways than
sched_ext, and some of them do break as the kernel code changes. However,
they get patched up pretty quickly and don't have any problem maintaining a
thriving ecosystem. This sort of arrangement isn't rare - browser, desktop,
even editor plugins are often like this.

Another point which may be worth considering is that there is an innate
drive towards consolidation even in open ecosystems. In most cases, over
time, a large number of experiments will be boiled down to a small number of
winners. Many people end up having similar needs and sooner or later
something becomes dominant in the area. While I don't have a crystal ball, I
don't think sched_ext schedulers are going to be much different - there will
probably be a handful that are both useful in some niches and well
maintained against the backdrop of many one-off and experimental ones. A bit
messy maybe, but not overwhelming. Most projects flourish in similar
situations.

> I generally worry that certain things may not have existed in the shipped
> scheduler if plugging was an option including EAS, throttling control,
> schedutil integration, big.Little, adapting to chiplets and picking preferred
> SMT siblings for turbo boost. In each case, integrating support was time
> consuming painful and a pluggable scheduler would have been a relatively
> easy out that would ultimately cost us if it was never properly integrated.
> While no one wants the pain, a few of us also want to avoid the problem
> of vendors publishing a hacky scheduler for their specific hardware and
> discontinuing the work at that point.

There are two diametric approaches. Currently, in the scheduler, everyone is
forced to work on a single implementation so that all the efforts and energy
can be captured in one place. Let's call this the funneling model. The
opposite would be the open model, where a common framework is laid out and
there are multiple competing implementations.

As with anything, there are pros and cons to both approaches and the long
term outcome can sometimes be counter-intuitive. For example, if a project
is funneled too hard, it's unlikely to grow a large and diverse pool of
contributors as there just isn't enough space to accommodate them. This
limits the contributor pool which in turn fortifies the justification for
funneling everyone into one thing, creating a feedback loop.

When development effort is perceived as finite zero-sum resource, this way
of describing the situation makes sense - "Had that developer not been
forced to work on this code base, we would have lost this feature."

Putting the other end into a similar proposition may be useful in seeing the
inherent trade-off - "What new ideas and talents are we missing out on by
constricting the contributor pool?"

As I mentioned before, the closest analogy I can think of is filesystems.
It's not an apples-to-apples comparison of course but there easily are an
order of magnitude more developers working on filesystems compared to
scheduling. Sure, a lot of efforts are diffused and duplicated but in the
long run we benefit so much more by letting people explore the problem
space, compete, copy from and improve upon each other.

Given how much both the machines and workloads have changed, we should be
trying wilder things a lot more and we can't do that with the funneled
model.

> I see that some friction with the current state is due to tuning knobs
> moving to debugfs. FWIW, I didn't 100% agree with that move either and
> carried an out-of-tree revert that displayed warnings for a time but I
> understood the logic behind it. However, if the tuning parameters are
> insufficient, and there is good reason to change them then the answer
> is to add tuning knobs with defined semantics and document them -- not
> pluggable schedulers. We've seen something along those lines recently
> with nice_latency even if it turned into EEVDF instead of a new interface,
> so I guess we'll see how that pans out.

That only works if we assume that most of what we want to try can be
reasonably covered by parameterizing EEVDF. The argument for sched_ext is
that there are a lot more radical things we can and should be trying. For
example, if we have many dozens of CPUs and workloads which are logically
grouped, it may make more sense to allocate CPUs to workloads rather than
scheduling each task's slice. Or given the close distance within a LLC
domain and non-uniform LLC layouts on some chiplet processors, task-to-CPU
stickiness maybe doesn’t matter anymore while spilling across multiple CCXs
can benefit from application-side hinting. Or given that CPU time has become
a lot flimsier as a way to measure CPU utilization, we should experiment
with different metrics which may or may not include wallclock times. Or,
given the proliferation of async frameworks in userspace, working closely
with async runtime could be pretty interesting.

That's too big a surface to cover by parametrizing EEVDF in any reasonable
way. Once we know what's really useful, some part can likely be incorporated
with well-defined interface but we don't know what this should look like yet
and it's pretty difficult to find out without an easy way to experiment.

> I get most of your points. Maybe most users will not care about a pluggable
> scheduler but *some will* and they will the maintenance burden. I get your
> point as well that if there is a bug and the pluggable scheduler then the
> first step would be "reproduce without the pluggable scheduler" and again,
> you'd be right, that is a great first step *except* sometimes they can't or
> sometimes they simply won't without significant proof and that's incurs a
> maintenance burden. Even if the pluggable schedulers are GPL, there still
> is a burden to understood any scheduler that is loaded to see if it's the
> source of a problem which means. Instead of understanding a defined number
> of schedulers that are developed over time with the history in changelogs,
> we may have to understand N schedulers that may be popular and that also
> is painful. That's leaving aside the difficulty of what happens when
> more than 1 can be loaded and interacting once containers are involved
> assuming that such support would exist in the future. It's already known
> that interacting IO schedulers are a nightmare so presumably interacting
> CPU schedulers within the same host would also be zero fun.

We should do a better job of clarifying and enforcing which IO controllers
and schedulers can be employed together, but here's the same question turned
around. What'd it be like if cfq had been deemed the only allowable IO
scheduler while the underlying hardware was going through rapid
developments? The IO schedulers and controllers being pluggable makes the
overall scene more complicated but that's also what allowed the block layer
to quickly adapt to the new hardware reality.

> Pluggable schedulers are effectively a change that we cannot walk back
> from if it turns out to be a bad idea because it potentially comes under
> the "you cannot break userspace" rule if a particular pluggable scheduler
> becomes popular.

I believe the track record of BPF use cases provides pretty strong evidence
against this. There are a lot more tools and usages which depend on kernel
internals now than several years ago, but that hasn't really gotten in the
way of kernel evolving in any meaningful way. We need to manage the
expectations as necessary but I believe the recent history has shown that
this concern doesn't really match reality.

> As I strongly believe it will be a nightmare to support
> within distributions where there is almost no control over the software
> stack of managing user expectations, I'm opposed to crossing that line with
> pluggable schedulers. While my nightmare scenarios may never be realised
> and could be overblown, it'll be hard to convince me it'll not kick me in
> the face eventually.

I agree that this may add support cost to distros and we should take
measures to limit that (e.g. make it evident which scheduler\[s\] have been
active and we’d be happy to take any suggestions), but we can't just look at
the costs. What benefits other parts of the industry are going to benefit
enterprise users too. If enterprise customers want to use sched_ext, their
reasons wouldn't be all that different from other industries.

Taking a further step back. You're asking "What's the cost that sched_ext
will add in terms of maintenance and support?" The question I'd like to
counter with is "How much are we missing out on by making it difficult to
explore and experiment in the scheduling space?" Opportunity costs are
usually more challenging to factor in but they're costs all the same at the
end of the day.

Thanks.

-- 
tejun