[RFC PATCH for 4.16 00/21] Restartable sequences and CPU op vector

[RFC PATCH for 4.16 00/21] Restartable sequences and CPU op vector

[Mathieu Desnoyers]

Hi,

Following changes based on the last round of review, I'm respinning this
series for another RFC round. It is based on the v4.15-rc3 tag. I am
now targeting the 4.16 merge window.

This series contains:

- Restartable sequences system call (x86 32/64, powerpc 32/64, arm 32),
- CPU operation vector system call (x86 32/64, powerpc 32/64, arm 32).

I have moved the membarrier patches to a different series.

The main changes introduced in this updated version are:

- Introducing a user-space "percpu-op.h" API, which wraps fast and slow
  paths into a single inline function,
- Changes to the cpu_opv page pinning, making it robust with respect to
  concurrent PTE changes,
- Fixed a rseq bug where the rseq_cs pointer was cleared too soon,
  causing hard-to-reproduce races triggered by NUMA page protection.
- Extended rseq selftests to add a "compare twice" approach validating
  invaritants, which catches races more quickly and closer to the
  race.

This updated series documents a major upside of the cpu_opv system call:
it allows performing operations on remote per-cpu data (e.g. migration
of memory between per-cpu memory pools, task migration across per-cpu
work queues, or summing up and clearning per-cpu statistics counters)
by simply passing the CPU number on which the operation must be
performed as argument to the system call.

Doing this without cpu_opv would require use of sched_setaffinity
alongside with restartable sequences, which is not safe against CPU
hot-unplug.

Feedback is welcome!

Thanks,

Mathieu

Boqun Feng (2):
  powerpc: Add support for restartable sequences
  powerpc: Wire up restartable sequences system call

Mathieu Desnoyers (19):
  uapi headers: Provide types_32_64.h
  rseq: Introduce restartable sequences system call (v12)
  arm: Add restartable sequences support
  arm: Wire up restartable sequences system call
  x86: Add support for restartable sequences
  x86: Wire up restartable sequence system call
  sched: Implement push_task_to_cpu
  cpu_opv: Provide cpu_opv system call (v5)
  x86: Wire up cpu_opv system call
  powerpc: Wire up cpu_opv system call
  arm: Wire up cpu_opv system call
  selftests: lib.mk: Introduce OVERRIDE_TARGETS
  cpu_opv: selftests: Implement selftests (v6)
  rseq: selftests: Provide rseq library (v5)
  rseq: selftests: Provide percpu_op API
  rseq: selftests: Provide basic test
  rseq: selftests: Provide basic percpu ops test
  rseq: selftests: Provide parametrized tests
  rseq: selftests: Provide Makefile, scripts, gitignore

 MAINTAINERS                                        |   20 +
 arch/Kconfig                                       |    7 +
 arch/arm/Kconfig                                   |    1 +
 arch/arm/kernel/signal.c                           |    7 +
 arch/arm/tools/syscall.tbl                         |    2 +
 arch/powerpc/Kconfig                               |    1 +
 arch/powerpc/include/asm/systbl.h                  |    2 +
 arch/powerpc/include/asm/unistd.h                  |    2 +-
 arch/powerpc/include/uapi/asm/unistd.h             |    2 +
 arch/powerpc/kernel/signal.c                       |    3 +
 arch/x86/Kconfig                                   |    1 +
 arch/x86/entry/common.c                            |    1 +
 arch/x86/entry/syscalls/syscall_32.tbl             |    2 +
 arch/x86/entry/syscalls/syscall_64.tbl             |    2 +
 arch/x86/kernel/signal.c                           |    6 +
 fs/exec.c                                          |    1 +
 include/linux/sched.h                              |  109 ++
 include/linux/syscalls.h                           |    6 +
 include/trace/events/rseq.h                        |   56 +
 include/uapi/linux/cpu_opv.h                       |  114 ++
 include/uapi/linux/rseq.h                          |  150 +++
 include/uapi/linux/types_32_64.h                   |   67 ++
 init/Kconfig                                       |   30 +
 kernel/Makefile                                    |    2 +
 kernel/cpu_opv.c                                   | 1078 +++++++++++++++++
 kernel/fork.c                                      |    2 +
 kernel/rseq.c                                      |  358 ++++++
 kernel/sched/core.c                                |   38 +
 kernel/sched/sched.h                               |   10 +
 kernel/sys_ni.c                                    |    4 +
 tools/testing/selftests/Makefile                   |    2 +
 tools/testing/selftests/cpu-opv/.gitignore         |    1 +
 tools/testing/selftests/cpu-opv/Makefile           |   17 +
 .../testing/selftests/cpu-opv/basic_cpu_opv_test.c | 1258 ++++++++++++++++++++
 tools/testing/selftests/cpu-opv/cpu-op.c           |  352 ++++++
 tools/testing/selftests/cpu-opv/cpu-op.h           |   59 +
 tools/testing/selftests/lib.mk                     |    4 +
 tools/testing/selftests/rseq/.gitignore            |    7 +
 tools/testing/selftests/rseq/Makefile              |   37 +
 .../testing/selftests/rseq/basic_percpu_ops_test.c |  296 +++++
 tools/testing/selftests/rseq/basic_test.c          |   55 +
 tools/testing/selftests/rseq/param_test.c          | 1163 ++++++++++++++++++
 tools/testing/selftests/rseq/percpu-op.h           |  163 +++
 tools/testing/selftests/rseq/rseq-arm.h            |  732 ++++++++++++
 tools/testing/selftests/rseq/rseq-ppc.h            |  688 +++++++++++
 tools/testing/selftests/rseq/rseq-skip.h           |   82 ++
 tools/testing/selftests/rseq/rseq-x86.h            | 1149 ++++++++++++++++++
 tools/testing/selftests/rseq/rseq.c                |  116 ++
 tools/testing/selftests/rseq/rseq.h                |  164 +++
 tools/testing/selftests/rseq/run_param_test.sh     |  130 ++
 50 files changed, 8558 insertions(+), 1 deletion(-)
 create mode 100644 include/trace/events/rseq.h
 create mode 100644 include/uapi/linux/cpu_opv.h
 create mode 100644 include/uapi/linux/rseq.h
 create mode 100644 include/uapi/linux/types_32_64.h
 create mode 100644 kernel/cpu_opv.c
 create mode 100644 kernel/rseq.c
 create mode 100644 tools/testing/selftests/cpu-opv/.gitignore
 create mode 100644 tools/testing/selftests/cpu-opv/Makefile
 create mode 100644 tools/testing/selftests/cpu-opv/basic_cpu_opv_test.c
 create mode 100644 tools/testing/selftests/cpu-opv/cpu-op.c
 create mode 100644 tools/testing/selftests/cpu-opv/cpu-op.h
 create mode 100644 tools/testing/selftests/rseq/.gitignore
 create mode 100644 tools/testing/selftests/rseq/Makefile
 create mode 100644 tools/testing/selftests/rseq/basic_percpu_ops_test.c
 create mode 100644 tools/testing/selftests/rseq/basic_test.c
 create mode 100644 tools/testing/selftests/rseq/param_test.c
 create mode 100644 tools/testing/selftests/rseq/percpu-op.h
 create mode 100644 tools/testing/selftests/rseq/rseq-arm.h
 create mode 100644 tools/testing/selftests/rseq/rseq-ppc.h
 create mode 100644 tools/testing/selftests/rseq/rseq-skip.h
 create mode 100644 tools/testing/selftests/rseq/rseq-x86.h
 create mode 100644 tools/testing/selftests/rseq/rseq.c
 create mode 100644 tools/testing/selftests/rseq/rseq.h
 create mode 100755 tools/testing/selftests/rseq/run_param_test.sh

-- 
2.11.0

[RFC PATCH for 4.16 01/21] uapi headers: Provide types_32_64.h

[Mathieu Desnoyers]

Provide helper macros for fields which represent pointers in
kernel-userspace ABI. This facilitates handling of 32-bit
user-space by 64-bit kernels by defining those fields as
32-bit 0-padding and 32-bit integer on 32-bit architectures,
which allows the kernel to treat those as 64-bit integers.
The order of padding and 32-bit integer depends on the
endianness.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Paul Turner <pjt@google.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Andrew Hunter <ahh@google.com>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Michael Kerrisk <mtk.manpages@gmail.com>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-api@vger.kernel.org
---
 include/uapi/linux/types_32_64.h | 67 ++++++++++++++++++++++++++++++++++++++++
 1 file changed, 67 insertions(+)
 create mode 100644 include/uapi/linux/types_32_64.h

diff --git a/include/uapi/linux/types_32_64.h b/include/uapi/linux/types_32_64.h
new file mode 100644
index 000000000000..18dc8808d026
--- /dev/null
+++ b/include/uapi/linux/types_32_64.h
@@ -0,0 +1,67 @@
+#ifndef _UAPI_LINUX_TYPES_32_64_H
+#define _UAPI_LINUX_TYPES_32_64_H
+
+/*
+ * linux/types_32_64.h
+ *
+ * Integer type declaration for pointers across 32-bit and 64-bit systems.
+ *
+ * Copyright (c) 2015-2017 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifdef __KERNEL__
+# include <linux/types.h>
+#else
+# include <stdint.h>
+#endif
+
+#include <asm/byteorder.h>
+
+#ifdef __BYTE_ORDER
+# if (__BYTE_ORDER == __BIG_ENDIAN)
+#  define LINUX_BYTE_ORDER_BIG_ENDIAN
+# else
+#  define LINUX_BYTE_ORDER_LITTLE_ENDIAN
+# endif
+#else
+# ifdef __BIG_ENDIAN
+#  define LINUX_BYTE_ORDER_BIG_ENDIAN
+# else
+#  define LINUX_BYTE_ORDER_LITTLE_ENDIAN
+# endif
+#endif
+
+#ifdef __LP64__
+# define LINUX_FIELD_u32_u64(field)			uint64_t field
+# define LINUX_FIELD_u32_u64_INIT_ONSTACK(field, v)	field = (intptr_t)v
+#else
+# ifdef LINUX_BYTE_ORDER_BIG_ENDIAN
+#  define LINUX_FIELD_u32_u64(field)	uint32_t field ## _padding, field
+#  define LINUX_FIELD_u32_u64_INIT_ONSTACK(field, v)	\
+	field ## _padding = 0, field = (intptr_t)v
+# else
+#  define LINUX_FIELD_u32_u64(field)	uint32_t field, field ## _padding
+#  define LINUX_FIELD_u32_u64_INIT_ONSTACK(field, v)	\
+	field = (intptr_t)v, field ## _padding = 0
+# endif
+#endif
+
+#endif /* _UAPI_LINUX_TYPES_32_64_H */
-- 
2.11.0

[RFC PATCH for 4.16 02/21] rseq: Introduce restartable sequences system call (v12)

[Mathieu Desnoyers]

Expose a new system call allowing each thread to register one userspace
memory area to be used as an ABI between kernel and user-space for two
purposes: user-space restartable sequences and quick access to read the
current CPU number value from user-space.

* Restartable sequences (per-cpu atomics)

Restartables sequences allow user-space to perform update operations on
per-cpu data without requiring heavy-weight atomic operations.

The restartable critical sections (percpu atomics) work has been started
by Paul Turner and Andrew Hunter. It lets the kernel handle restart of
critical sections. [1] [2] The re-implementation proposed here brings a
few simplifications to the ABI which facilitates porting to other
architectures and speeds up the user-space fast path. A second system
call, cpu_opv(), is proposed as fallback to deal with debugger
single-stepping. cpu_opv() executes a sequence of operations on behalf
of user-space with preemption disabled.

Here are benchmarks of various rseq use-cases.

Test hardware:

arm32: ARMv7 Processor rev 4 (v7l) "Cubietruck", 2-core
x86-64: Intel E5-2630 v3@2.40GHz, 16-core, hyperthreading

The following benchmarks were all performed on a single thread.

* Per-CPU statistic counter increment

                getcpu+atomic (ns/op)    rseq (ns/op)    speedup
arm32:                344.0                 31.4          11.0
x86-64:                15.3                  2.0           7.7

* LTTng-UST: write event 32-bit header, 32-bit payload into tracer
             per-cpu buffer

                getcpu+atomic (ns/op)    rseq (ns/op)    speedup
arm32:               2502.0                 2250.0         1.1
x86-64:               117.4                   98.0         1.2

* liburcu percpu: lock-unlock pair, dereference, read/compare word

                getcpu+atomic (ns/op)    rseq (ns/op)    speedup
arm32:                751.0                 128.5          5.8
x86-64:                53.4                  28.6          1.9

* jemalloc memory allocator adapted to use rseq

Using rseq with per-cpu memory pools in jemalloc at Facebook (based on
rseq 2016 implementation):

The production workload response-time has 1-2% gain avg. latency, and
the P99 overall latency drops by 2-3%.

* Reading the current CPU number

Speeding up reading the current CPU number on which the caller thread is
running is done by keeping the current CPU number up do date within the
cpu_id field of the memory area registered by the thread. This is done
by making scheduler preemption set the TIF_NOTIFY_RESUME flag on the
current thread. Upon return to user-space, a notify-resume handler
updates the current CPU value within the registered user-space memory
area. User-space can then read the current CPU number directly from
memory.

Keeping the current cpu id in a memory area shared between kernel and
user-space is an improvement over current mechanisms available to read
the current CPU number, which has the following benefits over
alternative approaches:

- 35x speedup on ARM vs system call through glibc
- 20x speedup on x86 compared to calling glibc, which calls vdso
  executing a "lsl" instruction,
- 14x speedup on x86 compared to inlined "lsl" instruction,
- Unlike vdso approaches, this cpu_id value can be read from an inline
  assembly, which makes it a useful building block for restartable
  sequences.
- The approach of reading the cpu id through memory mapping shared
  between kernel and user-space is portable (e.g. ARM), which is not the
  case for the lsl-based x86 vdso.

On x86, yet another possible approach would be to use the gs segment
selector to point to user-space per-cpu data. This approach performs
similarly to the cpu id cache, but it has two disadvantages: it is
not portable, and it is incompatible with existing applications already
using the gs segment selector for other purposes.

Benchmarking various approaches for reading the current CPU number:

ARMv7 Processor rev 4 (v7l)
Machine model: Cubietruck
- Baseline (empty loop):                                    8.4 ns
- Read CPU from rseq cpu_id:                               16.7 ns
- Read CPU from rseq cpu_id (lazy register):               19.8 ns
- glibc 2.19-0ubuntu6.6 getcpu:                           301.8 ns
- getcpu system call:                                     234.9 ns

x86-64 Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz:
- Baseline (empty loop):                                    0.8 ns
- Read CPU from rseq cpu_id:                                0.8 ns
- Read CPU from rseq cpu_id (lazy register):                0.8 ns
- Read using gs segment selector:                           0.8 ns
- "lsl" inline assembly:                                   13.0 ns
- glibc 2.19-0ubuntu6 getcpu:                              16.6 ns
- getcpu system call:                                      53.9 ns

- Speed (benchmark taken on v8 of patchset)

Running 10 runs of hackbench -l 100000 seems to indicate, contrary to
expectations, that enabling CONFIG_RSEQ slightly accelerates the
scheduler:

Configuration: 2 sockets * 8-core Intel(R) Xeon(R) CPU E5-2630 v3 @
2.40GHz (directly on hardware, hyperthreading disabled in BIOS, energy
saving disabled in BIOS, turboboost disabled in BIOS, cpuidle.off=1
kernel parameter), with a Linux v4.6 defconfig+localyesconfig,
restartable sequences series applied.

* CONFIG_RSEQ=n

avg.:      41.37 s
std.dev.:   0.36 s

* CONFIG_RSEQ=y

avg.:      40.46 s
std.dev.:   0.33 s

- Size

On x86-64, between CONFIG_RSEQ=n/y, the text size increase of vmlinux is
567 bytes, and the data size increase of vmlinux is 5696 bytes.

On x86-64, between CONFIG_CPU_OPV=n/y, the text size increase of vmlinux is
5576 bytes, and the data size increase of vmlinux is 6164 bytes.

[1] https://lwn.net/Articles/650333/
[2] http://www.linuxplumbersconf.org/2013/ocw/system/presentations/1695/original/LPC%20-%20PerCpu%20Atomics.pdf

Link: http://lkml.kernel.org/r/20151027235635.16059.11630.stgit@pjt-glaptop.roam.corp.google.com
Link: http://lkml.kernel.org/r/20150624222609.6116.86035.stgit@kitami.mtv.corp.google.com
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Paul Turner <pjt@google.com>
CC: Andrew Hunter <ahh@google.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Michael Kerrisk <mtk.manpages@gmail.com>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: Alexander Viro <viro@zeniv.linux.org.uk>
CC: linux-api@vger.kernel.org
---

Changes since v1:
- Return -1, errno=EINVAL if cpu_cache pointer is not aligned on
  sizeof(int32_t).
- Update man page to describe the pointer alignement requirements and
  update atomicity guarantees.
- Add MAINTAINERS file GETCPU_CACHE entry.
- Remove dynamic memory allocation: go back to having a single
  getcpu_cache entry per thread. Update documentation accordingly.
- Rebased on Linux 4.4.

Changes since v2:
- Introduce a "cmd" argument, along with an enum with GETCPU_CACHE_GET
  and GETCPU_CACHE_SET. Introduce a uapi header linux/getcpu_cache.h
  defining this enumeration.
- Split resume notifier architecture implementation from the system call
  wire up in the following arch-specific patches.
- Man pages updates.
- Handle 32-bit compat pointers.
- Simplify handling of getcpu_cache GETCPU_CACHE_SET compiler barrier:
  set the current cpu cache pointer before doing the cache update, and
  set it back to NULL if the update fails. Setting it back to NULL on
  error ensures that no resume notifier will trigger a SIGSEGV if a
  migration happened concurrently.

Changes since v3:
- Fix __user annotations in compat code,
- Update memory ordering comments.
- Rebased on kernel v4.5-rc5.

Changes since v4:
- Inline getcpu_cache_fork, getcpu_cache_execve, and getcpu_cache_exit.
- Add new line between if() and switch() to improve readability.
- Added sched switch benchmarks (hackbench) and size overhead comparison
  to change log.

Changes since v5:
- Rename "getcpu_cache" to "thread_local_abi", allowing to extend
  this system call to cover future features such as restartable critical
  sections. Generalizing this system call ensures that we can add
  features similar to the cpu_id field within the same cache-line
  without having to track one pointer per feature within the task
  struct.
- Add a tlabi_nr parameter to the system call, thus allowing to extend
  the ABI beyond the initial 64-byte structure by registering structures
  with tlabi_nr greater than 0. The initial ABI structure is associated
  with tlabi_nr 0.
- Rebased on kernel v4.5.

Changes since v6:
- Integrate "restartable sequences" v2 patchset from Paul Turner.
- Add handling of single-stepping purely in user-space, with a
  fallback to locking after 2 rseq failures to ensure progress, and
  by exposing a __rseq_table section to debuggers so they know where
  to put breakpoints when dealing with rseq assembly blocks which
  can be aborted at any point.
- make the code and ABI generic: porting the kernel implementation
  simply requires to wire up the signal handler and return to user-space
  hooks, and allocate the syscall number.
- extend testing with a fully configurable test program. See
  param_spinlock_test -h for details.
- handling of rseq ENOSYS in user-space, also with a fallback
  to locking.
- modify Paul Turner's rseq ABI to only require a single TLS store on
  the user-space fast-path, removing the need to populate two additional
  registers. This is made possible by introducing struct rseq_cs into
  the ABI to describe a critical section start_ip, post_commit_ip, and
  abort_ip.
- Rebased on kernel v4.7-rc7.

Changes since v7:
- Documentation updates.
- Integrated powerpc architecture support.
- Compare rseq critical section start_ip, allows shriking the user-space
  fast-path code size.
- Added Peter Zijlstra, Paul E. McKenney and Boqun Feng as
  co-maintainers.
- Added do_rseq2 and do_rseq_memcpy to test program helper library.
- Code cleanup based on review from Peter Zijlstra, Andy Lutomirski and
  Boqun Feng.
- Rebase on kernel v4.8-rc2.

Changes since v8:
- clear rseq_cs even if non-nested. Speeds up user-space fast path by
  removing the final "rseq_cs=NULL" assignment.
- add enum rseq_flags: critical sections and threads can set migration,
  preemption and signal "disable" flags to inhibit rseq behavior.
- rseq_event_counter needs to be updated with a pre-increment: Otherwise
  misses an increment after exec (when TLS and in-kernel states are
  initially 0).

Changes since v9:
- Update changelog.
- Fold instrumentation patch.
- check abort-ip signature: Add a signature before the abort-ip landing
  address. This signature is also received as a new parameter to the
  rseq system call. The kernel uses it ensures that rseq cannot be used
  as an exploit vector to redirect execution to arbitrary code.
- Use rseq pointer for both register and unregister. This is more
  symmetric, and eventually allow supporting a linked list of rseq
  struct per thread if needed in the future.
- Unregistration of a rseq structure is now done with
  RSEQ_FLAG_UNREGISTER.
- Remove reference counting. Return "EBUSY" to the caller if rseq is
  already registered for the current thread. This simplifies
  implementation while still allowing user-space to perform lazy
  registration in multi-lib use-cases. (suggested by Ben Maurer)
- Clear rseq_cs upon unregister.
- Set cpu_id back to -1 on unregister, so if rseq user libraries follow
  an unregister, and they expect to lazily register rseq, they can do
  so.
- Document rseq_cs clear requirement: JIT should reset the rseq_cs
  pointer before reclaiming memory of rseq_cs structure.
- Introduce rseq_len syscall parameter, rseq_cs version field:
  Allow keeping track of the registered rseq struct length, for future
  extensions. Add rseq_cs version as first field. Will allow future
  extensions.
- Use offset and unsigned arithmetic to save a branch:  Save a
  conditional branch when comparing instruction pointer against a
  rseq_cs descriptor's address range by having post_commit_ip as an
  offset from start_ip, and using unsigned integer comparison.
  Suggested by Ben Maurer.
- Remove event counter from ABI. Suggested by Andy Lutomirski.
- Add INIT_ONSTACK macro: Introduce the
  RSEQ_FIELD_u32_u64_INIT_ONSTACK() macros to ensure that users
  correctly initialize the upper bits of RSEQ_FIELD_u32_u64() on their
  stack to 0 on 32-bit architectures.
- Select MEMBARRIER: Allows user-space rseq fast-paths to use the value
  of cpu_id field (inherently required by the rseq algorithm) to figure
  out whether membarrier can be expected to be available.
  This effectively allows user-space fast-paths to remove extra
  comparisons and branch testing whether membarrier is enabled, and thus
  whether a full barrier is required (e.g. in userspace RCU
  implementation after rcu_read_lock/before rcu_read_unlock).
- Expose cpu_id_start field: Checking whether the (cpu_id < 0) in the C
  preparation part of the rseq fast-path brings significant overhead at
  least on arm32. We can remove this extra comparison by exposing two
  distinct cpu_id fields in the rseq TLS:

  The field cpu_id_start always contain a *possible* cpu number, although
  it may not be the current one if, for instance, rseq is not initialized
  for the current thread. cpu_id_start is meant to be used in the C code
  for the pointer chasing to figure out which per-cpu data structure
  should be passed to the rseq asm sequence.

  The field cpu_id values -1 means rseq is not initialized, and -2 means
  initialization failed. That field is used in the rseq asm sequence to
  confirm that the cpu_id_start value was indeed the current cpu number.
  It also ends up confirming that rseq is initialized for the current
  thread, because values -1 and -2 will never match the cpu_id_start
  possible cpu number values.

  This allows checking the current CPU number and rseq initialization
  state with a single comparison on the fast-path.

Changes since v10:

- Update rseq.c comment, removing reference to event_counter.

Changes since v11:

- Replace task struct rseq_preempt, rseq_signal, and rseq_migrate
  bool by u32 rseq_event_mask.
- Add missing sys_rseq() asmlinkage declaration to
  include/linux/syscalls.h.
- Copy event mask on process fork, set to 0 on exec and thread-fork.
- Cleanups based on review from Peter Zijlstra.
- Cleanups based on review from Thomas Gleixner.
- Fix: rseq_cs needs to be cleared only when:
  - Nested over non-critical-section userspace code,
  - Nested over rseq_cs _and_ handling abort.
  Basically, we should never clear rseq_cs when the rseq resume to
  userspace handler is called and it is not handling abort: the
  problematic case is if any of the __get_user()/__put_user done
  by the handler trigger a page fault (e.g. page protection
  done by NUMA page migration work), which triggers preemption:
  the next call to the rseq resume to userspace handler needs to
  perform the abort.
- Perform rseq event mask updates atomically wrt preemption,
- Move rseq_migrate to __set_task_cpu(), thus catching migration
  scenario that bypass set_task_cpu(): fork and wake_up_new_task.
- Merge content of rseq_sched_out into rseq_preempt. There is no
  need to have two hook sites. Both setting the rseq event mask
  preempt bit and setting the notify resume thread flag can be
  done from rseq_preempt().
- Issue rseq_preempt() from fork(), thus ensuring that we handle
  abort if needed.

Man page associated:

RSEQ(2)                Linux Programmer's Manual               RSEQ(2)

NAME
       rseq - Restartable sequences and cpu number cache

SYNOPSIS
       #include <linux/rseq.h>

       int rseq(struct rseq * rseq, uint32_t rseq_len, int flags, uint32_t sig);

DESCRIPTION
       The  rseq()  ABI  accelerates  user-space operations on per-cpu
       data by defining a shared data structure ABI between each user-
       space thread and the kernel.

       It  allows  user-space  to perform update operations on per-cpu
       data without requiring heavy-weight atomic operations.

       Restartable sequences are atomic  with  respect  to  preemption
       (making  it atomic with respect to other threads running on the
       same CPU), as well as  signal  delivery  (user-space  execution
       contexts nested over the same thread).

       It is suited for update operations on per-cpu data.

       It can be used on data structures shared between threads within
       a process, and on data structures shared between threads across
       different processes.

       Some examples of operations that can be accelerated or improved
       by this ABI:

       · Memory allocator per-cpu free-lists,

       · Querying the current CPU number,

       · Incrementing per-CPU counters,

       · Modifying data protected by per-CPU spinlocks,

       · Inserting/removing elements in per-CPU linked-lists,

       · Writing/reading per-CPU ring buffers content.

       · Accurately reading performance monitoring unit counters  with
         respect to thread migration.

       The  rseq argument is a pointer to the thread-local rseq struc‐
       ture to be shared between kernel and user-space.  A  NULL  rseq
       value unregisters the current thread rseq structure.

       The layout of struct rseq is as follows:

       Structure alignment
              This structure is aligned on multiples of 32 bytes.

       Structure size
              This  structure  is  extensible.  Its  size is passed as
              parameter to the rseq system call.

       Fields

           cpu_id_start
              Optimistic cache of the CPU number on which the  current
              thread  is running. Its value is guaranteed to always be
              a possible CPU number, even when rseq  is  not  initial‐
              ized.  The  value it contains should always be confirmed
              by reading the cpu_id field.

           cpu_id
              Cache of the CPU number on which the current  thread  is
              running.  -1 if uninitialized.

           rseq_cs
              The  rseq_cs  field is a pointer to a struct rseq_cs. Is
              is NULL when no rseq assembly block critical section  is
              active for the current thread.  Setting it to point to a
              critical section descriptor (struct rseq_cs)  marks  the
              beginning of the critical section.

           flags
              Flags  indicating  the  restart behavior for the current
              thread. This is mainly used for debugging purposes.  Can
              be either:

       ·      RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT

       ·      RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL

       ·      RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE

       The layout of struct rseq_cs version 0 is as follows:

       Structure alignment
              This structure is aligned on multiples of 32 bytes.

       Structure size
              This structure has a fixed size of 32 bytes.

       Fields

           version
              Version of this structure.

           flags
              Flags indicating the restart behavior of this structure.
              Can be either:

       ·      RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT

       ·      RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL

       ·      RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE

           start_ip
              Instruction pointer address of the first instruction  of
              the sequence of consecutive assembly instructions.

           post_commit_offset
              Offset  (from start_ip address) of the address after the
              last instruction of the sequence of consecutive assembly
              instructions.

           abort_ip
              Instruction  pointer address where to move the execution
              flow in case of abort of  the  sequence  of  consecutive
              assembly instructions.

       The  rseq_len argument is the size of the struct rseq to regis‐
       ter.

       The flags argument is 0 for registration, and  RSEQ_FLAG_UNREG‐
       ISTER for unregistration.

       The  sig argument is the 32-bit signature to be expected before
       the abort handler code.

       A single library per process should keep the rseq structure  in
       a  thread-local  storage  variable.  The cpu_id field should be
       initialized to -1, and the cpu_id_start field  should  be  ini‐
       tialized to a possible CPU value (typically 0).

       Each  thread  is  responsible for registering and unregistering
       its rseq structure. No more than one rseq structure address can
       be registered per thread at a given time.

       In  a  typical  usage scenario, the thread registering the rseq
       structure will be performing  loads  and  stores  from/to  that
       structure.  It  is  however also allowed to read that structure
       from other threads.  The rseq field updates  performed  by  the
       kernel  provide  relaxed  atomicity  semantics, which guarantee
       that other threads performing relaxed atomic reads of  the  cpu
       number cache will always observe a consistent value.

RETURN VALUE
       A  return  value  of  0  indicates  success.  On  error,  -1 is
       returned, and errno is set appropriately.

ERRORS
       EINVAL Either flags contains an invalid value, or rseq contains
              an  address  which  is  not  appropriately  aligned,  or
              rseq_len contains a size that does not  match  the  size
              received on registration.

       ENOSYS The  rseq()  system call is not implemented by this ker‐
              nel.

       EFAULT rseq is an invalid address.

       EBUSY  Restartable sequence  is  already  registered  for  this
              thread.

       EPERM  The  sig  argument  on unregistration does not match the
              signature received on registration.

VERSIONS
       The rseq() system call was added in Linux 4.X (TODO).

CONFORMING TO
       rseq() is Linux-specific.

SEE ALSO
       sched_getcpu(3)

Linux                         2017-11-06                       RSEQ(2)
---
 MAINTAINERS                 |  11 ++
 arch/Kconfig                |   7 +
 fs/exec.c                   |   1 +
 include/linux/sched.h       | 109 ++++++++++++++
 include/linux/syscalls.h    |   3 +
 include/trace/events/rseq.h |  56 +++++++
 include/uapi/linux/rseq.h   | 150 +++++++++++++++++++
 init/Kconfig                |  14 ++
 kernel/Makefile             |   1 +
 kernel/fork.c               |   2 +
 kernel/rseq.c               | 358 ++++++++++++++++++++++++++++++++++++++++++++
 kernel/sched/core.c         |   1 +
 kernel/sched/sched.h        |   1 +
 kernel/sys_ni.c             |   3 +
 14 files changed, 717 insertions(+)
 create mode 100644 include/trace/events/rseq.h
 create mode 100644 include/uapi/linux/rseq.h
 create mode 100644 kernel/rseq.c

diff --git a/MAINTAINERS b/MAINTAINERS
index 82ad0eabce4f..4ede6c16d49f 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -11620,6 +11620,17 @@ F:	include/dt-bindings/reset/
 F:	include/linux/reset.h
 F:	include/linux/reset-controller.h
 
+RESTARTABLE SEQUENCES SUPPORT
+M:	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+M:	Peter Zijlstra <peterz@infradead.org>
+M:	"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
+M:	Boqun Feng <boqun.feng@gmail.com>
+L:	linux-kernel@vger.kernel.org
+S:	Supported
+F:	kernel/rseq.c
+F:	include/uapi/linux/rseq.h
+F:	include/trace/events/rseq.h
+
 RFKILL
 M:	Johannes Berg <johannes@sipsolutions.net>
 L:	linux-wireless@vger.kernel.org
diff --git a/arch/Kconfig b/arch/Kconfig
index 400b9e1b2f27..2d7f54a5784b 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -258,6 +258,13 @@ config HAVE_REGS_AND_STACK_ACCESS_API
 	  declared in asm/ptrace.h
 	  For example the kprobes-based event tracer needs this API.
 
+config HAVE_RSEQ
+	bool
+	depends on HAVE_REGS_AND_STACK_ACCESS_API
+	help
+	  This symbol should be selected by an architecture if it
+	  supports an implementation of restartable sequences.
+
 config HAVE_CLK
 	bool
 	help
diff --git a/fs/exec.c b/fs/exec.c
index 6be2aa0ab26f..8d6eac1e20d8 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -1808,6 +1808,7 @@ static int do_execveat_common(int fd, struct filename *filename,
 	current->fs->in_exec = 0;
 	current->in_execve = 0;
 	membarrier_execve(current);
+	rseq_execve(current);
 	acct_update_integrals(current);
 	task_numa_free(current);
 	free_bprm(bprm);
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 21991d668d35..503f076d6f21 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -27,6 +27,7 @@
 #include <linux/signal_types.h>
 #include <linux/mm_types_task.h>
 #include <linux/task_io_accounting.h>
+#include <linux/rseq.h>
 
 /* task_struct member predeclarations (sorted alphabetically): */
 struct audit_context;
@@ -978,6 +979,17 @@ struct task_struct {
 	unsigned long			numa_pages_migrated;
 #endif /* CONFIG_NUMA_BALANCING */
 
+#ifdef CONFIG_RSEQ
+	struct rseq __user *rseq;
+	u32 rseq_len;
+	u32 rseq_sig;
+	/*
+	 * RmW on rseq_event_mask must be performed atomically
+	 * with respect to preemption.
+	 */
+	unsigned long rseq_event_mask;
+#endif
+
 	struct tlbflush_unmap_batch	tlb_ubc;
 
 	struct rcu_head			rcu;
@@ -1668,4 +1680,101 @@ extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
 #define TASK_SIZE_OF(tsk)	TASK_SIZE
 #endif
 
+#ifdef CONFIG_RSEQ
+/*
+ * Map the event mask on the user-space ABI enum rseq_cs_flags
+ * for direct mask checks.
+ */
+
+enum rseq_event_mask_bits {
+	RSEQ_EVENT_PREEMPT_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
+	RSEQ_EVENT_SIGNAL_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
+	RSEQ_EVENT_MIGRATE_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
+};
+
+
+enum rseq_event_mask {
+	RSEQ_EVENT_PREEMPT	= (1U << RSEQ_EVENT_PREEMPT_BIT),
+	RSEQ_EVENT_SIGNAL	= (1U << RSEQ_EVENT_SIGNAL_BIT),
+	RSEQ_EVENT_MIGRATE	= (1U << RSEQ_EVENT_MIGRATE_BIT),
+};
+
+static inline void rseq_set_notify_resume(struct task_struct *t)
+{
+	if (t->rseq)
+		set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
+}
+void __rseq_handle_notify_resume(struct pt_regs *regs);
+static inline void rseq_handle_notify_resume(struct pt_regs *regs)
+{
+	if (current->rseq)
+		__rseq_handle_notify_resume(regs);
+}
+static inline void rseq_signal_deliver(struct pt_regs *regs)
+{
+	set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
+	rseq_handle_notify_resume(regs);
+}
+static inline void rseq_preempt(struct task_struct *t)
+{
+	set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
+	rseq_set_notify_resume(t);
+}
+static inline void rseq_migrate(struct task_struct *t)
+{
+	set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
+	rseq_set_notify_resume(t);
+}
+/*
+ * If parent process has a registered restartable sequences area, the
+ * child inherits. Only applies when forking a process, not a thread. In
+ * case a parent fork() in the middle of a restartable sequence, set the
+ * resume notifier to force the child to retry.
+ */
+static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
+{
+	if (clone_flags & CLONE_THREAD) {
+		t->rseq = NULL;
+		t->rseq_len = 0;
+		t->rseq_sig = 0;
+		t->rseq_event_mask = 0;
+	} else {
+		t->rseq = current->rseq;
+		t->rseq_len = current->rseq_len;
+		t->rseq_sig = current->rseq_sig;
+		t->rseq_event_mask = current->rseq_event_mask;
+		rseq_preempt(t);
+	}
+}
+static inline void rseq_execve(struct task_struct *t)
+{
+	t->rseq = NULL;
+	t->rseq_len = 0;
+	t->rseq_sig = 0;
+	t->rseq_event_mask = 0;
+}
+#else
+static inline void rseq_set_notify_resume(struct task_struct *t)
+{
+}
+static inline void rseq_handle_notify_resume(struct pt_regs *regs)
+{
+}
+static inline void rseq_signal_deliver(struct pt_regs *regs)
+{
+}
+static inline void rseq_preempt(struct task_struct *t)
+{
+}
+static inline void rseq_migrate(struct task_struct *t)
+{
+}
+static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
+{
+}
+static inline void rseq_execve(struct task_struct *t)
+{
+}
+#endif
+
 #endif
diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
index a78186d826d7..340650b4ec54 100644
--- a/include/linux/syscalls.h
+++ b/include/linux/syscalls.h
@@ -66,6 +66,7 @@ struct old_linux_dirent;
 struct perf_event_attr;
 struct file_handle;
 struct sigaltstack;
+struct rseq;
 union bpf_attr;
 
 #include <linux/types.h>
@@ -940,5 +941,7 @@ asmlinkage long sys_pkey_alloc(unsigned long flags, unsigned long init_val);
 asmlinkage long sys_pkey_free(int pkey);
 asmlinkage long sys_statx(int dfd, const char __user *path, unsigned flags,
 			  unsigned mask, struct statx __user *buffer);
+asmlinkage long sys_rseq(struct rseq __user *rseq, uint32_t rseq_len,
+			int flags, uint32_t sig);
 
 #endif
diff --git a/include/trace/events/rseq.h b/include/trace/events/rseq.h
new file mode 100644
index 000000000000..c4609a3f5008
--- /dev/null
+++ b/include/trace/events/rseq.h
@@ -0,0 +1,56 @@
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM rseq
+
+#if !defined(_TRACE_RSEQ_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_RSEQ_H
+
+#include <linux/tracepoint.h>
+#include <linux/types.h>
+
+TRACE_EVENT(rseq_update,
+
+	TP_PROTO(struct task_struct *t),
+
+	TP_ARGS(t),
+
+	TP_STRUCT__entry(
+		__field(s32, cpu_id)
+	),
+
+	TP_fast_assign(
+		__entry->cpu_id = raw_smp_processor_id();
+	),
+
+	TP_printk("cpu_id=%d", __entry->cpu_id)
+);
+
+TRACE_EVENT(rseq_ip_fixup,
+
+	TP_PROTO(unsigned long regs_ip, unsigned long start_ip,
+		unsigned long post_commit_offset, unsigned long abort_ip),
+
+	TP_ARGS(regs_ip, start_ip, post_commit_offset, abort_ip),
+
+	TP_STRUCT__entry(
+		__field(unsigned long, regs_ip)
+		__field(unsigned long, start_ip)
+		__field(unsigned long, post_commit_offset)
+		__field(unsigned long, abort_ip)
+	),
+
+	TP_fast_assign(
+		__entry->regs_ip = regs_ip;
+		__entry->start_ip = start_ip;
+		__entry->post_commit_offset = post_commit_offset;
+		__entry->abort_ip = abort_ip;
+	),
+
+	TP_printk("regs_ip=0x%lx start_ip=0x%lx post_commit_offset=%lu abort_ip=0x%lx",
+		__entry->regs_ip, __entry->start_ip,
+		__entry->post_commit_offset, __entry->abort_ip)
+);
+
+#endif /* _TRACE_SOCK_H */
+
+/* This part must be outside protection */
+#include <trace/define_trace.h>
diff --git a/include/uapi/linux/rseq.h b/include/uapi/linux/rseq.h
new file mode 100644
index 000000000000..3895ec940059
--- /dev/null
+++ b/include/uapi/linux/rseq.h
@@ -0,0 +1,150 @@
+#ifndef _UAPI_LINUX_RSEQ_H
+#define _UAPI_LINUX_RSEQ_H
+
+/*
+ * linux/rseq.h
+ *
+ * Restartable sequences system call API
+ *
+ * Copyright (c) 2015-2016 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifdef __KERNEL__
+# include <linux/types.h>
+#else
+# include <stdint.h>
+#endif
+
+#include <linux/types_32_64.h>
+
+enum rseq_cpu_id_state {
+	RSEQ_CPU_ID_UNINITIALIZED		= -1,
+	RSEQ_CPU_ID_REGISTRATION_FAILED		= -2,
+};
+
+enum rseq_flags {
+	RSEQ_FLAG_UNREGISTER = (1 << 0),
+};
+
+enum rseq_cs_flags_bit {
+	RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT	= 0,
+	RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT	= 1,
+	RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT	= 2,
+};
+
+enum rseq_cs_flags {
+	RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT	=
+		(1U << RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT),
+	RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL	=
+		(1U << RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT),
+	RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE	=
+		(1U << RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT),
+};
+
+/*
+ * struct rseq_cs is aligned on 4 * 8 bytes to ensure it is always
+ * contained within a single cache-line. It is usually declared as
+ * link-time constant data.
+ */
+struct rseq_cs {
+	/* Version of this structure. */
+	uint32_t version;
+	/* enum rseq_cs_flags */
+	uint32_t flags;
+	LINUX_FIELD_u32_u64(start_ip);
+	/* Offset from start_ip. */
+	LINUX_FIELD_u32_u64(post_commit_offset);
+	LINUX_FIELD_u32_u64(abort_ip);
+} __attribute__((aligned(4 * sizeof(uint64_t))));
+
+/*
+ * struct rseq is aligned on 4 * 8 bytes to ensure it is always
+ * contained within a single cache-line.
+ *
+ * A single struct rseq per thread is allowed.
+ */
+struct rseq {
+	/*
+	 * Restartable sequences cpu_id_start field. Updated by the
+	 * kernel, and read by user-space with single-copy atomicity
+	 * semantics. Aligned on 32-bit. Always contains a value in the
+	 * range of possible CPUs, although the value may not be the
+	 * actual current CPU (e.g. if rseq is not initialized). This
+	 * CPU number value should always be compared against the value
+	 * of the cpu_id field before performing a rseq commit or
+	 * returning a value read from a data structure indexed using
+	 * the cpu_id_start value.
+	 */
+	uint32_t cpu_id_start;
+	/*
+	 * Restartable sequences cpu_id field. Updated by the kernel,
+	 * and read by user-space with single-copy atomicity semantics.
+	 * Aligned on 32-bit. Values RSEQ_CPU_ID_UNINITIALIZED and
+	 * RSEQ_CPU_ID_REGISTRATION_FAILED have a special semantic: the
+	 * former means "rseq uninitialized", and latter means "rseq
+	 * initialization failed". This value is meant to be read within
+	 * rseq critical sections and compared with the cpu_id_start
+	 * value previously read, before performing the commit instruction,
+	 * or read and compared with the cpu_id_start value before returning
+	 * a value loaded from a data structure indexed using the
+	 * cpu_id_start value.
+	 */
+	uint32_t cpu_id;
+	/*
+	 * Restartable sequences rseq_cs field.
+	 *
+	 * Contains NULL when no critical section is active for the current
+	 * thread, or holds a pointer to the currently active struct rseq_cs.
+	 *
+	 * Updated by user-space, which sets the address of the currently
+	 * active rseq_cs at the beginning of assembly instruction sequence
+	 * block, and set to NULL by the kernel when it restarts an assembly
+	 * instruction sequence block, as well as when the kernel detects that
+	 * it is preempting or delivering a signal outside of the range
+	 * targeted by the rseq_cs. Also needs to be set to NULL by user-space
+	 * before reclaiming memory that contains the targeted struct rseq_cs.
+	 *
+	 * Read and set by the kernel with single-copy atomicity semantics.
+	 * Set by user-space with single-copy atomicity semantics. Aligned
+	 * on 64-bit.
+	 */
+	LINUX_FIELD_u32_u64(rseq_cs);
+	/*
+	 * - RSEQ_DISABLE flag:
+	 *
+	 * Fallback fast-track flag for single-stepping.
+	 * Set by user-space if lack of progress is detected.
+	 * Cleared by user-space after rseq finish.
+	 * Read by the kernel.
+	 * - RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT
+	 *     Inhibit instruction sequence block restart and event
+	 *     counter increment on preemption for this thread.
+	 * - RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL
+	 *     Inhibit instruction sequence block restart and event
+	 *     counter increment on signal delivery for this thread.
+	 * - RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE
+	 *     Inhibit instruction sequence block restart and event
+	 *     counter increment on migration for this thread.
+	 */
+	uint32_t flags;
+} __attribute__((aligned(4 * sizeof(uint64_t))));
+
+#endif /* _UAPI_LINUX_RSEQ_H */
diff --git a/init/Kconfig b/init/Kconfig
index 2934249fba46..88e36395390f 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -1400,6 +1400,20 @@ config USERFAULTFD
 	  Enable the userfaultfd() system call that allows to intercept and
 	  handle page faults in userland.
 
+config RSEQ
+	bool "Enable rseq() system call" if EXPERT
+	default y
+	depends on HAVE_RSEQ
+	select MEMBARRIER
+	help
+	  Enable the restartable sequences system call. It provides a
+	  user-space cache for the current CPU number value, which
+	  speeds up getting the current CPU number from user-space,
+	  as well as an ABI to speed up user-space operations on
+	  per-CPU data.
+
+	  If unsure, say Y.
+
 config EMBEDDED
 	bool "Embedded system"
 	option allnoconfig_y
diff --git a/kernel/Makefile b/kernel/Makefile
index 172d151d429c..3574669dafd9 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -112,6 +112,7 @@ obj-$(CONFIG_CONTEXT_TRACKING) += context_tracking.o
 obj-$(CONFIG_TORTURE_TEST) += torture.o
 
 obj-$(CONFIG_HAS_IOMEM) += memremap.o
+obj-$(CONFIG_RSEQ) += rseq.o
 
 $(obj)/configs.o: $(obj)/config_data.h
 
diff --git a/kernel/fork.c b/kernel/fork.c
index 432eadf6b58c..e903ee4f21ba 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -1858,6 +1858,8 @@ static __latent_entropy struct task_struct *copy_process(
 	 */
 	copy_seccomp(p);
 
+	rseq_fork(p, clone_flags);
+
 	/*
 	 * Process group and session signals need to be delivered to just the
 	 * parent before the fork or both the parent and the child after the
diff --git a/kernel/rseq.c b/kernel/rseq.c
new file mode 100644
index 000000000000..93f3f169e112
--- /dev/null
+++ b/kernel/rseq.c
@@ -0,0 +1,358 @@
+/*
+ * Restartable sequences system call
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * Copyright (C) 2015, Google, Inc.,
+ * Paul Turner <pjt@google.com> and Andrew Hunter <ahh@google.com>
+ * Copyright (C) 2015-2016, EfficiOS Inc.,
+ * Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ */
+
+#include <linux/sched.h>
+#include <linux/uaccess.h>
+#include <linux/syscalls.h>
+#include <linux/rseq.h>
+#include <linux/types.h>
+#include <asm/ptrace.h>
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/rseq.h>
+
+/*
+ *
+ * Restartable sequences are a lightweight interface that allows
+ * user-level code to be executed atomically relative to scheduler
+ * preemption and signal delivery. Typically used for implementing
+ * per-cpu operations.
+ *
+ * It allows user-space to perform update operations on per-cpu data
+ * without requiring heavy-weight atomic operations.
+ *
+ * Detailed algorithm of rseq user-space assembly sequences:
+ *
+ *                     init(rseq_cs)
+ *                     cpu = TLS->rseq::cpu_id_start
+ *   [1]               TLS->rseq::rseq_cs = rseq_cs
+ *   [start_ip]        ----------------------------
+ *   [2]               if (cpu != TLS->rseq::cpu_id)
+ *                             goto abort_ip;
+ *   [3]               <last_instruction_in_cs>
+ *   [post_commit_ip]  ----------------------------
+ *
+ *   The address of jump target abort_ip must be outside the critical
+ *   region, i.e.:
+ *
+ *     [abort_ip] < [start_ip]  || [abort_ip] >= [post_commit_ip]
+ *
+ *   Steps [2]-[3] (inclusive) need to be a sequence of instructions in
+ *   userspace that can handle being interrupted between any of those
+ *   instructions, and then resumed to the abort_ip.
+ *
+ *   1.  Userspace stores the address of the struct rseq_cs assembly
+ *       block descriptor into the rseq_cs field of the registered
+ *       struct rseq TLS area. This update is performed through a single
+ *       store within the inline assembly instruction sequence.
+ *       [start_ip]
+ *
+ *   2.  Userspace tests to check whether the current cpu_id field match
+ *       the cpu number loaded before start_ip, branching to abort_ip
+ *       in case of a mismatch.
+ *
+ *       If the sequence is preempted or interrupted by a signal
+ *       at or after start_ip and before post_commit_ip, then the kernel
+ *       clears TLS->__rseq_abi::rseq_cs, and sets the user-space return
+ *       ip to abort_ip before returning to user-space, so the preempted
+ *       execution resumes at abort_ip.
+ *
+ *   3.  Userspace critical section final instruction before
+ *       post_commit_ip is the commit. The critical section is
+ *       self-terminating.
+ *       [post_commit_ip]
+ *
+ *   4.  <success>
+ *
+ *   On failure at [2], or if interrupted by preempt or signal delivery
+ *   between [1] and [3]:
+ *
+ *       [abort_ip]
+ *   F1. <failure>
+ */
+
+static int rseq_update_cpu_id(struct task_struct *t)
+{
+	uint32_t cpu_id = raw_smp_processor_id();
+
+	if (__put_user(cpu_id, &t->rseq->cpu_id_start))
+		return -EFAULT;
+	if (__put_user(cpu_id, &t->rseq->cpu_id))
+		return -EFAULT;
+	trace_rseq_update(t);
+	return 0;
+}
+
+static int rseq_reset_rseq_cpu_id(struct task_struct *t)
+{
+	uint32_t cpu_id_start = 0, cpu_id = RSEQ_CPU_ID_UNINITIALIZED;
+
+	/*
+	 * Reset cpu_id_start to its initial state (0).
+	 */
+	if (__put_user(cpu_id_start, &t->rseq->cpu_id_start))
+		return -EFAULT;
+	/*
+	 * Reset cpu_id to RSEQ_CPU_ID_UNINITIALIZED, so any user coming
+	 * in after unregistration can figure out that rseq needs to be
+	 * registered again.
+	 */
+	if (__put_user(cpu_id, &t->rseq->cpu_id))
+		return -EFAULT;
+	return 0;
+}
+
+static int rseq_get_rseq_cs(struct task_struct *t,
+			    unsigned long *start_ip,
+			    unsigned long *post_commit_offset,
+			    unsigned long *abort_ip,
+			    uint32_t *cs_flags)
+{
+	struct rseq_cs __user *urseq_cs;
+	struct rseq_cs rseq_cs;
+	unsigned long ptr;
+	u32 __user *usig;
+	u32 sig;
+	int ret;
+
+	ret = __get_user(ptr, &t->rseq->rseq_cs);
+	if (ret)
+		return ret;
+	if (!ptr)
+		return 0;
+	urseq_cs = (struct rseq_cs __user *)ptr;
+	if (copy_from_user(&rseq_cs, urseq_cs, sizeof(rseq_cs)))
+		return -EFAULT;
+	if (rseq_cs.version > 0)
+		return -EINVAL;
+
+	/* Ensure that abort_ip is not in the critical section. */
+	if (rseq_cs.abort_ip - rseq_cs.start_ip < rseq_cs.post_commit_offset)
+		return -EINVAL;
+
+	*cs_flags = rseq_cs.flags;
+	*start_ip = rseq_cs.start_ip;
+	*post_commit_offset = rseq_cs.post_commit_offset;
+	*abort_ip = rseq_cs.abort_ip;
+
+	usig = (u32 __user *)(rseq_cs.abort_ip - sizeof(u32));
+	ret = get_user(sig, usig);
+	if (ret)
+		return ret;
+
+	if (current->rseq_sig != sig) {
+		printk_ratelimited(KERN_WARNING
+			"Possible attack attempt. Unexpected rseq signature 0x%x, expecting 0x%x (pid=%d, addr=%p).\n",
+			sig, current->rseq_sig, current->pid, usig);
+		return -EPERM;
+	}
+	return 0;
+}
+
+static int rseq_need_restart(struct task_struct *t, uint32_t cs_flags)
+{
+	uint32_t flags, event_mask;
+	int ret;
+
+	/* Get thread flags. */
+	ret = __get_user(flags, &t->rseq->flags);
+	if (ret)
+		return ret;
+
+	/* Take critical section flags into account. */
+	flags |= cs_flags;
+
+	/*
+	 * Restart on signal can only be inhibited when restart on
+	 * preempt and restart on migrate are inhibited too. Otherwise,
+	 * a preempted signal handler could fail to restart the prior
+	 * execution context on sigreturn.
+	 */
+	if (unlikely(flags & RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL)) {
+		if ((flags & (RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE
+		    | RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT)) !=
+		    (RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE
+		     | RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT))
+			return -EINVAL;
+	}
+
+	/*
+	 * Load and clear event mask atomically with respect to
+	 * scheduler preemption.
+	 */
+	preempt_disable();
+	event_mask = t->rseq_event_mask;
+	t->rseq_event_mask = 0;
+	preempt_enable();
+
+	event_mask &= ~flags;
+	if (event_mask)
+		return 1;
+	return 0;
+}
+
+static int clear_rseq_cs(struct task_struct *t)
+{
+	unsigned long ptr = 0;
+
+	/*
+	 * The rseq_cs field is set to NULL on preemption or signal
+	 * delivery on top of rseq assembly block, as well as on top
+	 * of code outside of the rseq assembly block. This performs
+	 * a lazy clear of the rseq_cs field.
+	 *
+	 * Set rseq_cs to NULL with single-copy atomicity.
+	 */
+	return __put_user(ptr, &t->rseq->rseq_cs);
+}
+
+static int rseq_ip_fixup(struct pt_regs *regs)
+{
+	unsigned long ip = instruction_pointer(regs), start_ip = 0,
+		post_commit_offset = 0, abort_ip = 0;
+	struct task_struct *t = current;
+	uint32_t cs_flags = 0;
+	bool in_rseq_cs = false;
+	int ret;
+
+	ret = rseq_get_rseq_cs(t, &start_ip, &post_commit_offset, &abort_ip,
+			&cs_flags);
+	if (ret)
+		return ret;
+
+	/*
+	 * Handle potentially not being within a critical section.
+	 * Unsigned comparison will be true when
+	 * ip >= start_ip, and when ip < start_ip + post_commit_offset.
+	 */
+	if (ip - start_ip < post_commit_offset)
+		in_rseq_cs = true;
+
+	/*
+	 * If not nested over a rseq critical section, restart is
+	 * useless. Clear the rseq_cs pointer and return.
+	 */
+	if (!in_rseq_cs)
+		return clear_rseq_cs(t);
+	ret = rseq_need_restart(t, cs_flags);
+	if (ret <= 0)
+		return ret;
+	ret = clear_rseq_cs(t);
+	if (ret)
+		return ret;
+	trace_rseq_ip_fixup(ip, start_ip, post_commit_offset, abort_ip);
+	instruction_pointer_set(regs, (unsigned long)abort_ip);
+	return 0;
+}
+
+/*
+ * This resume handler must always be executed between any of:
+ * - preemption,
+ * - signal delivery,
+ * and return to user-space.
+ *
+ * This is how we can ensure that the entire rseq critical section,
+ * consisting of both the C part and the assembly instruction sequence,
+ * will issue the commit instruction only if executed atomically with
+ * respect to other threads scheduled on the same CPU, and with respect
+ * to signal handlers.
+ */
+void __rseq_handle_notify_resume(struct pt_regs *regs)
+{
+	struct task_struct *t = current;
+	int ret;
+
+	if (unlikely(t->flags & PF_EXITING))
+		return;
+	if (unlikely(!access_ok(VERIFY_WRITE, t->rseq, sizeof(*t->rseq))))
+		goto error;
+	ret = rseq_ip_fixup(regs);
+	if (unlikely(ret < 0))
+		goto error;
+	if (unlikely(rseq_update_cpu_id(t)))
+		goto error;
+	return;
+
+error:
+	force_sig(SIGSEGV, t);
+}
+
+/*
+ * sys_rseq - setup restartable sequences for caller thread.
+ */
+SYSCALL_DEFINE4(rseq, struct rseq __user *, rseq, uint32_t, rseq_len,
+		int, flags, uint32_t, sig)
+{
+	int ret;
+
+	if (flags & RSEQ_FLAG_UNREGISTER) {
+		/* Unregister rseq for current thread. */
+		if (current->rseq != rseq || !current->rseq)
+			return -EINVAL;
+		if (current->rseq_len != rseq_len)
+			return -EINVAL;
+		if (current->rseq_sig != sig)
+			return -EPERM;
+		ret = rseq_reset_rseq_cpu_id(current);
+		if (ret)
+			return ret;
+		current->rseq = NULL;
+		current->rseq_len = 0;
+		current->rseq_sig = 0;
+		return 0;
+	}
+
+	if (unlikely(flags))
+		return -EINVAL;
+
+	if (current->rseq) {
+		/*
+		 * If rseq is already registered, check whether
+		 * the provided address differs from the prior
+		 * one.
+		 */
+		if (current->rseq != rseq || current->rseq_len != rseq_len)
+			return -EINVAL;
+		if (current->rseq_sig != sig)
+			return -EPERM;
+		/* Already registered. */
+		return -EBUSY;
+	}
+
+	/*
+	 * If there was no rseq previously registered,
+	 * ensure the provided rseq is properly aligned and valid.
+	 */
+	if (!IS_ALIGNED((unsigned long)rseq, __alignof__(*rseq)) ||
+	    rseq_len != sizeof(*rseq))
+		return -EINVAL;
+	if (!access_ok(VERIFY_WRITE, rseq, rseq_len))
+		return -EFAULT;
+	current->rseq = rseq;
+	current->rseq_len = rseq_len;
+	current->rseq_sig = sig;
+	/*
+	 * If rseq was previously inactive, and has just been
+	 * registered, ensure the cpu_id_start and cpu_id fields
+	 * are updated before returning to user-space.
+	 */
+	rseq_set_notify_resume(current);
+
+	return 0;
+}
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 75554f366fd3..a7cc81d1fcb6 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2590,6 +2590,7 @@ prepare_task_switch(struct rq *rq, struct task_struct *prev,
 {
 	sched_info_switch(rq, prev, next);
 	perf_event_task_sched_out(prev, next);
+	rseq_preempt(prev);
 	fire_sched_out_preempt_notifiers(prev, next);
 	prepare_lock_switch(rq, next);
 	prepare_arch_switch(next);
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index b19552a212de..85e7a622ee88 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -1221,6 +1221,7 @@ static inline void __set_task_cpu(struct task_struct *p, unsigned int cpu)
 #endif
 	p->wake_cpu = cpu;
 #endif
+	rseq_migrate(p);
 }
 
 /*
diff --git a/kernel/sys_ni.c b/kernel/sys_ni.c
index b5189762d275..bfa1ee1bf669 100644
--- a/kernel/sys_ni.c
+++ b/kernel/sys_ni.c
@@ -259,3 +259,6 @@ cond_syscall(sys_membarrier);
 cond_syscall(sys_pkey_mprotect);
 cond_syscall(sys_pkey_alloc);
 cond_syscall(sys_pkey_free);
+
+/* restartable sequence */
+cond_syscall(sys_rseq);
-- 
2.11.0

[RFC PATCH for 4.16 03/21] arm: Add restartable sequences support

[Mathieu Desnoyers]

Call the rseq_handle_notify_resume() function on return to
userspace if TIF_NOTIFY_RESUME thread flag is set.

Increment the event counter and perform fixup on the pre-signal frame
when a signal is delivered on top of a restartable sequence critical
section.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Paul Turner <pjt@google.com>
CC: Andrew Hunter <ahh@google.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-api@vger.kernel.org
---
 arch/arm/Kconfig         | 1 +
 arch/arm/kernel/signal.c | 7 +++++++
 2 files changed, 8 insertions(+)

diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
index 51c8df561077..556e9b2225c6 100644
--- a/arch/arm/Kconfig
+++ b/arch/arm/Kconfig
@@ -88,6 +88,7 @@ config ARM
 	select HAVE_PERF_USER_STACK_DUMP
 	select HAVE_RCU_TABLE_FREE if (SMP && ARM_LPAE)
 	select HAVE_REGS_AND_STACK_ACCESS_API
+	select HAVE_RSEQ
 	select HAVE_SYSCALL_TRACEPOINTS
 	select HAVE_UID16
 	select HAVE_VIRT_CPU_ACCOUNTING_GEN
diff --git a/arch/arm/kernel/signal.c b/arch/arm/kernel/signal.c
index bd8810d4acb3..5879ab3f53c1 100644
--- a/arch/arm/kernel/signal.c
+++ b/arch/arm/kernel/signal.c
@@ -541,6 +541,12 @@ static void handle_signal(struct ksignal *ksig, struct pt_regs *regs)
 	int ret;
 
 	/*
+	 * Increment event counter and perform fixup for the pre-signal
+	 * frame.
+	 */
+	rseq_signal_deliver(regs);
+
+	/*
 	 * Set up the stack frame
 	 */
 	if (ksig->ka.sa.sa_flags & SA_SIGINFO)
@@ -660,6 +666,7 @@ do_work_pending(struct pt_regs *regs, unsigned int thread_flags, int syscall)
 			} else {
 				clear_thread_flag(TIF_NOTIFY_RESUME);
 				tracehook_notify_resume(regs);
+				rseq_handle_notify_resume(regs);
 			}
 		}
 		local_irq_disable();
-- 
2.11.0

[RFC PATCH for 4.16 04/21] arm: Wire up restartable sequences system call

[Mathieu Desnoyers]

Wire up the rseq system call on 32-bit ARM.

This provides an ABI improving the speed of a user-space getcpu
operation on ARM by skipping the getcpu system call on the fast path, as
well as improving the speed of user-space operations on per-cpu data
compared to using load-linked/store-conditional.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Paul Turner <pjt@google.com>
CC: Andrew Hunter <ahh@google.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-api@vger.kernel.org
---
 arch/arm/tools/syscall.tbl | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/arm/tools/syscall.tbl b/arch/arm/tools/syscall.tbl
index 0bb0e9c6376c..fbc74b5fa3ed 100644
--- a/arch/arm/tools/syscall.tbl
+++ b/arch/arm/tools/syscall.tbl
@@ -412,3 +412,4 @@
 395	common	pkey_alloc		sys_pkey_alloc
 396	common	pkey_free		sys_pkey_free
 397	common	statx			sys_statx
+398	common	rseq			sys_rseq
-- 
2.11.0

[RFC PATCH for 4.16 05/21] x86: Add support for restartable sequences

[Mathieu Desnoyers]

Call the rseq_handle_notify_resume() function on return to userspace if
TIF_NOTIFY_RESUME thread flag is set.

Increment the event counter and perform fixup on the pre-signal frame
when a signal is delivered on top of a restartable sequence critical
section.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Paul Turner <pjt@google.com>
CC: Andrew Hunter <ahh@google.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-api@vger.kernel.org
---
 arch/x86/Kconfig         | 1 +
 arch/x86/entry/common.c  | 1 +
 arch/x86/kernel/signal.c | 6 ++++++
 3 files changed, 8 insertions(+)

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 8eed3f94bfc7..aa86dab90a95 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -174,6 +174,7 @@ config X86
 	select HAVE_REGS_AND_STACK_ACCESS_API
 	select HAVE_RELIABLE_STACKTRACE		if X86_64 && UNWINDER_FRAME_POINTER && STACK_VALIDATION
 	select HAVE_STACK_VALIDATION		if X86_64
+	select HAVE_RSEQ
 	select HAVE_SYSCALL_TRACEPOINTS
 	select HAVE_UNSTABLE_SCHED_CLOCK
 	select HAVE_USER_RETURN_NOTIFIER
diff --git a/arch/x86/entry/common.c b/arch/x86/entry/common.c
index d7d3cc24baf4..d65595b201f4 100644
--- a/arch/x86/entry/common.c
+++ b/arch/x86/entry/common.c
@@ -160,6 +160,7 @@ static void exit_to_usermode_loop(struct pt_regs *regs, u32 cached_flags)
 		if (cached_flags & _TIF_NOTIFY_RESUME) {
 			clear_thread_flag(TIF_NOTIFY_RESUME);
 			tracehook_notify_resume(regs);
+			rseq_handle_notify_resume(regs);
 		}
 
 		if (cached_flags & _TIF_USER_RETURN_NOTIFY)
diff --git a/arch/x86/kernel/signal.c b/arch/x86/kernel/signal.c
index b9e00e8f1c9b..991017d26d00 100644
--- a/arch/x86/kernel/signal.c
+++ b/arch/x86/kernel/signal.c
@@ -687,6 +687,12 @@ setup_rt_frame(struct ksignal *ksig, struct pt_regs *regs)
 	sigset_t *set = sigmask_to_save();
 	compat_sigset_t *cset = (compat_sigset_t *) set;
 
+	/*
+	 * Increment event counter and perform fixup for the pre-signal
+	 * frame.
+	 */
+	rseq_signal_deliver(regs);
+
 	/* Set up the stack frame */
 	if (is_ia32_frame(ksig)) {
 		if (ksig->ka.sa.sa_flags & SA_SIGINFO)
-- 
2.11.0

[RFC PATCH for 4.16 06/21] x86: Wire up restartable sequence system call

[Mathieu Desnoyers]

Wire up the rseq system call on x86 32/64.

This provides an ABI improving the speed of a user-space getcpu
operation on x86 by removing the need to perform a function call, "lsl"
instruction, or system call on the fast path, as well as improving the
speed of user-space operations on per-cpu data.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Paul Turner <pjt@google.com>
CC: Andrew Hunter <ahh@google.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-api@vger.kernel.org
---
 arch/x86/entry/syscalls/syscall_32.tbl | 1 +
 arch/x86/entry/syscalls/syscall_64.tbl | 1 +
 2 files changed, 2 insertions(+)

diff --git a/arch/x86/entry/syscalls/syscall_32.tbl b/arch/x86/entry/syscalls/syscall_32.tbl
index 448ac2161112..ba43ee75e425 100644
--- a/arch/x86/entry/syscalls/syscall_32.tbl
+++ b/arch/x86/entry/syscalls/syscall_32.tbl
@@ -391,3 +391,4 @@
 382	i386	pkey_free		sys_pkey_free
 383	i386	statx			sys_statx
 384	i386	arch_prctl		sys_arch_prctl			compat_sys_arch_prctl
+385	i386	rseq			sys_rseq
diff --git a/arch/x86/entry/syscalls/syscall_64.tbl b/arch/x86/entry/syscalls/syscall_64.tbl
index 5aef183e2f85..3ad03495bbb9 100644
--- a/arch/x86/entry/syscalls/syscall_64.tbl
+++ b/arch/x86/entry/syscalls/syscall_64.tbl
@@ -339,6 +339,7 @@
 330	common	pkey_alloc		sys_pkey_alloc
 331	common	pkey_free		sys_pkey_free
 332	common	statx			sys_statx
+333	common	rseq			sys_rseq
 
 #
 # x32-specific system call numbers start at 512 to avoid cache impact
-- 
2.11.0

[RFC PATCH for 4.16 07/21] powerpc: Add support for restartable sequences

[Mathieu Desnoyers]

From: Boqun Feng <boqun.feng-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>

Call the rseq_handle_notify_resume() function on return to userspace if
TIF_NOTIFY_RESUME thread flag is set.

Increment the event counter and perform fixup on the pre-signal when a
signal is delivered on top of a restartable sequence critical section.

Signed-off-by: Boqun Feng <boqun.feng-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
CC: Benjamin Herrenschmidt <benh-XVmvHMARGAS8U2dJNN8I7kB+6BGkLq7r@public.gmane.org>
CC: Paul Mackerras <paulus-eUNUBHrolfbYtjvyW6yDsg@public.gmane.org>
CC: Michael Ellerman <mpe-Gsx/Oe8HsFggBc27wqDAHg@public.gmane.org>
CC: Peter Zijlstra <peterz-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org>
CC: "Paul E. McKenney" <paulmck-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
CC: linuxppc-dev-uLR06cmDAlY/bJ5BZ2RsiQ@public.gmane.org
---
 arch/powerpc/Kconfig         | 1 +
 arch/powerpc/kernel/signal.c | 3 +++
 2 files changed, 4 insertions(+)

diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index c51e6ce42e7a..e9992f80819c 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig
@@ -221,6 +221,7 @@ config PPC
 	select HAVE_SYSCALL_TRACEPOINTS
 	select HAVE_VIRT_CPU_ACCOUNTING
 	select HAVE_IRQ_TIME_ACCOUNTING
+	select HAVE_RSEQ
 	select IRQ_DOMAIN
 	select IRQ_FORCED_THREADING
 	select MODULES_USE_ELF_RELA
diff --git a/arch/powerpc/kernel/signal.c b/arch/powerpc/kernel/signal.c
index 3d7539b90010..a7b95f7bcaf1 100644
--- a/arch/powerpc/kernel/signal.c
+++ b/arch/powerpc/kernel/signal.c
@@ -133,6 +133,8 @@ static void do_signal(struct task_struct *tsk)
 	/* Re-enable the breakpoints for the signal stack */
 	thread_change_pc(tsk, tsk->thread.regs);
 
+	rseq_signal_deliver(tsk->thread.regs);
+
 	if (is32) {
         	if (ksig.ka.sa.sa_flags & SA_SIGINFO)
 			ret = handle_rt_signal32(&ksig, oldset, tsk);
@@ -161,6 +163,7 @@ void do_notify_resume(struct pt_regs *regs, unsigned long thread_info_flags)
 	if (thread_info_flags & _TIF_NOTIFY_RESUME) {
 		clear_thread_flag(TIF_NOTIFY_RESUME);
 		tracehook_notify_resume(regs);
+		rseq_handle_notify_resume(regs);
 	}
 
 	if (thread_info_flags & _TIF_PATCH_PENDING)
-- 
2.11.0

[RFC PATCH for 4.16 08/21] powerpc: Wire up restartable sequences system call

[Mathieu Desnoyers]

From: Boqun Feng <boqun.feng@gmail.com>

Wire up the rseq system call on powerpc.

This provides an ABI improving the speed of a user-space getcpu
operation on powerpc by skipping the getcpu system call on the fast
path, as well as improving the speed of user-space operations on per-cpu
data compared to using load-reservation/store-conditional atomics.

Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Benjamin Herrenschmidt <benh@kernel.crashing.org>
CC: Paul Mackerras <paulus@samba.org>
CC: Michael Ellerman <mpe@ellerman.id.au>
CC: Peter Zijlstra <peterz@infradead.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: linuxppc-dev@lists.ozlabs.org
---
 arch/powerpc/include/asm/systbl.h      | 1 +
 arch/powerpc/include/asm/unistd.h      | 2 +-
 arch/powerpc/include/uapi/asm/unistd.h | 1 +
 3 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/include/asm/systbl.h b/arch/powerpc/include/asm/systbl.h
index 449912f057f6..964321a5799c 100644
--- a/arch/powerpc/include/asm/systbl.h
+++ b/arch/powerpc/include/asm/systbl.h
@@ -389,3 +389,4 @@ COMPAT_SYS_SPU(preadv2)
 COMPAT_SYS_SPU(pwritev2)
 SYSCALL(kexec_file_load)
 SYSCALL(statx)
+SYSCALL(rseq)
diff --git a/arch/powerpc/include/asm/unistd.h b/arch/powerpc/include/asm/unistd.h
index 9ba11dbcaca9..e76bd5601ea4 100644
--- a/arch/powerpc/include/asm/unistd.h
+++ b/arch/powerpc/include/asm/unistd.h
@@ -12,7 +12,7 @@
 #include <uapi/asm/unistd.h>
 
 
-#define NR_syscalls		384
+#define NR_syscalls		385
 
 #define __NR__exit __NR_exit
 
diff --git a/arch/powerpc/include/uapi/asm/unistd.h b/arch/powerpc/include/uapi/asm/unistd.h
index df8684f31919..b1980fcd56d5 100644
--- a/arch/powerpc/include/uapi/asm/unistd.h
+++ b/arch/powerpc/include/uapi/asm/unistd.h
@@ -395,5 +395,6 @@
 #define __NR_pwritev2		381
 #define __NR_kexec_file_load	382
 #define __NR_statx		383
+#define __NR_rseq		384
 
 #endif /* _UAPI_ASM_POWERPC_UNISTD_H_ */
-- 
2.11.0

[RFC PATCH for 4.16 09/21] sched: Implement push_task_to_cpu

[Mathieu Desnoyers]

Implement push_task_to_cpu(), which moves the task received as argument
to the destination cpu's runqueue. It only does so if the CPU is within
the CPU allowed mask of the task, else it returns -EINVAL.

It does not change the CPU allowed mask, and can therefore be used
within applications which rely on owning the sched_setaffinity() state.

It does not pin the task to the destination CPU, which means that the
scheduler may choose to move the task away from that CPU before the
task executes. Code invoking push_task_to_cpu() must be prepared to
retry in that case.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Paul Turner <pjt@google.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Andrew Hunter <ahh@google.com>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Michael Kerrisk <mtk.manpages@gmail.com>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-api@vger.kernel.org
---
 kernel/sched/core.c  | 37 +++++++++++++++++++++++++++++++++++++
 kernel/sched/sched.h |  9 +++++++++
 2 files changed, 46 insertions(+)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index a7cc81d1fcb6..58a8f93949d8 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -1061,6 +1061,43 @@ void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
 		set_curr_task(rq, p);
 }
 
+int push_task_to_cpu(struct task_struct *p, unsigned int dest_cpu)
+{
+	struct rq_flags rf;
+	struct rq *rq;
+	int ret = 0;
+
+	rq = task_rq_lock(p, &rf);
+	update_rq_clock(rq);
+
+	if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed)) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (task_cpu(p) == dest_cpu)
+		goto out;
+
+	if (task_running(rq, p) || p->state == TASK_WAKING) {
+		struct migration_arg arg = { p, dest_cpu };
+		/* Need help from migration thread: drop lock and wait. */
+		task_rq_unlock(rq, p, &rf);
+		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
+		tlb_migrate_finish(p->mm);
+		return 0;
+	} else if (task_on_rq_queued(p)) {
+		/*
+		 * OK, since we're going to drop the lock immediately
+		 * afterwards anyway.
+		 */
+		rq = move_queued_task(rq, &rf, p, dest_cpu);
+	}
+out:
+	task_rq_unlock(rq, p, &rf);
+
+	return ret;
+}
+
 /*
  * Change a given task's CPU affinity. Migrate the thread to a
  * proper CPU and schedule it away if the CPU it's executing on
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 85e7a622ee88..85f409a18772 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -1224,6 +1224,15 @@ static inline void __set_task_cpu(struct task_struct *p, unsigned int cpu)
 	rseq_migrate(p);
 }
 
+#ifdef CONFIG_SMP
+int push_task_to_cpu(struct task_struct *p, unsigned int dest_cpu);
+#else
+static inline int push_task_to_cpu(struct task_struct *p, unsigned int dest_cpu)
+{
+	return 0;
+}
+#endif
+
 /*
  * Tunables that become constants when CONFIG_SCHED_DEBUG is off:
  */
-- 
2.11.0

[RFC PATCH for 4.16 10/21] cpu_opv: Provide cpu_opv system call (v5)

[Mathieu Desnoyers]

The cpu_opv system call executes a vector of operations on behalf of
user-space on a specific CPU with preemption disabled. It is inspired
by readv() and writev() system calls which take a "struct iovec"
array as argument.

The operations available are: comparison, memcpy, add, or, and, xor,
left shift, right shift, and memory barrier. The system call receives
a CPU number from user-space as argument, which is the CPU on which
those operations need to be performed.  All pointers in the ops must
have been set up to point to the per CPU memory of the CPU on which
the operations should be executed. The "comparison" operation can be
used to check that the data used in the preparation step did not
change between preparation of system call inputs and operation
execution within the preempt-off critical section.

The reason why we require all pointer offsets to be calculated by
user-space beforehand is because we need to use get_user_pages_fast()
to first pin all pages touched by each operation. This takes care of
faulting-in the pages. Then, preemption is disabled, and the
operations are performed atomically with respect to other thread
execution on that CPU, without generating any page fault.

An overall maximum of 4216 bytes in enforced on the sum of operation
length within an operation vector, so user-space cannot generate a
too long preempt-off critical section (cache cold critical section
duration measured as 4.7µs on x86-64). Each operation is also limited
a length of 4096 bytes, meaning that an operation can touch a
maximum of 4 pages (memcpy: 2 pages for source, 2 pages for
destination if addresses are not aligned on page boundaries).

If the thread is not running on the requested CPU, it is migrated to
it.

**** Justification for cpu_opv ****

Here are a few reasons justifying why the cpu_opv system call is
needed in addition to rseq:

1) Allow algorithms to perform per-cpu data migration without relying on
   sched_setaffinity()

The use-cases are migrating memory between per-cpu memory free-lists, or
stealing tasks from other per-cpu work queues: each require that
accesses to remote per-cpu data structures are performed.

Just rseq is not enough to cover those use-cases without additionally
relying on sched_setaffinity, which is unfortunately not
CPU-hotplug-safe.

The cpu_opv system call receives a CPU number as argument, and migrates
the current task to the right CPU to perform the operation sequence. If
the requested CPU is offline, it performs the operations from the
current CPU while preventing CPU hotplug, and with a mutex held.

2) Handling single-stepping from tools

Tools like debuggers, and simulators use single-stepping to run through
existing programs. If core libraries start to use restartable sequences
for e.g. memory allocation, this means pre-existing programs cannot be
single-stepped, simply because the underlying glibc or jemalloc has
changed.

The rseq user-space does expose a __rseq_table section for the sake of
debuggers, so they can skip over the rseq critical sections if they
want.  However, this requires upgrading tools, and still breaks
single-stepping in case where glibc or jemalloc is updated, but not the
tooling.

Having a performance-related library improvement break tooling is likely
to cause a big push-back against wide adoption of rseq.

3) Forward-progress guarantee

Having a piece of user-space code that stops progressing due to external
conditions is pretty bad. Developers are used to think of fast-path and
slow-path (e.g. for locking), where the contended vs uncontended cases
have different performance characteristics, but each need to provide
some level of progress guarantees.

There are concerns about proposing just "rseq" without the associated
slow-path (cpu_opv) that guarantees progress. It's just asking for
trouble when real-life will happen: page faults, uprobes, and other
unforeseen conditions that would seldom cause a rseq fast-path to never
progress.

4) Handling page faults

It's pretty easy to come up with corner-case scenarios where rseq does
not progress without the help from cpu_opv. For instance, a system with
swap enabled which is under high memory pressure could trigger page
faults at pretty much every rseq attempt. Although this scenario
is extremely unlikely, rseq becomes the weak link of the chain.

5) Comparison with LL/SC

The layman versed in the load-link/store-conditional instructions in
RISC architectures will notice the similarity between rseq and LL/SC
critical sections. The comparison can even be pushed further: since
debuggers can handle those LL/SC critical sections, they should be
able to handle rseq c.s. in the same way.

First, the way gdb recognises LL/SC c.s. patterns is very fragile:
it's limited to specific common patterns, and will miss the pattern
in all other cases. But fear not, having the rseq c.s. expose a
__rseq_table to debuggers removes that guessing part.

The main difference between LL/SC and rseq is that debuggers had
to support single-stepping through LL/SC critical sections from the
get go in order to support a given architecture. For rseq, we're
adding critical sections into pre-existing applications/libraries,
so the user expectation is that tools don't break due to a library
optimization.

6) Perform maintenance operations on per-cpu data

rseq c.s. are quite limited feature-wise: they need to end with a
*single* commit instruction that updates a memory location. On the other
hand, the cpu_opv system call can combine a sequence of operations that
need to be executed with preemption disabled. While slower than rseq,
this allows for more complex maintenance operations to be performed on
per-cpu data concurrently with rseq fast-paths, in cases where it's not
possible to map those sequences of ops to a rseq.

7) Use cpu_opv as generic implementation for architectures not
   implementing rseq assembly code

rseq critical sections require architecture-specific user-space code to
be crafted in order to port an algorithm to a given architecture.  In
addition, it requires that the kernel architecture implementation adds
hooks into signal delivery and resume to user-space.

In order to facilitate integration of rseq into user-space, cpu_opv can
provide a (relatively slower) architecture-agnostic implementation of
rseq. This means that user-space code can be ported to all architectures
through use of cpu_opv initially, and have the fast-path use rseq
whenever the asm code is implemented.

8) Allow libraries with multi-part algorithms to work on same per-cpu
   data without affecting the allowed cpu mask

The lttng-ust tracer presents an interesting use-case for per-cpu
buffers: the algorithm needs to update a "reserve" counter, serialize
data into the buffer, and then update a "commit" counter _on the same
per-cpu buffer_. Using rseq for both reserve and commit can bring
significant performance benefits.

Clearly, if rseq reserve fails, the algorithm can retry on a different
per-cpu buffer. However, it's not that easy for the commit. It needs to
be performed on the same per-cpu buffer as the reserve.

The cpu_opv system call solves that problem by receiving the cpu number
on which the operation needs to be performed as argument. It can push
the task to the right CPU if needed, and perform the operations there
with preemption disabled.

Changing the allowed cpu mask for the current thread is not an
acceptable alternative for a tracing library, because the application
being traced does not expect that mask to be changed by libraries.

9) Ensure that data structures don't need store-release/load-acquire
   semantic to handle fall-back

cpu_opv performs the fall-back on the requested CPU by migrating the
task to that CPU. Executing the slow-path on the right CPU ensures that
store-release/load-acquire semantic is not required neither on the
fast-path nor slow-path.

**** rseq and cpu_opv use-cases ****

1) per-cpu spinlock

A per-cpu spinlock can be implemented as a rseq consisting of a
comparison operation (== 0) on a word, and a word store (1), followed
by an acquire barrier after control dependency. The unlock path can be
performed with a simple store-release of 0 to the word, which does
not require rseq.

The cpu_opv fallback requires a single-word comparison (== 0) and a
single-word store (1).

2) per-cpu statistics counters

A per-cpu statistics counters can be implemented as a rseq consisting
of a final "add" instruction on a word as commit.

The cpu_opv fallback can be implemented as a "ADD" operation.

Besides statistics tracking, these counters can be used to implement
user-space RCU per-cpu grace period tracking for both single and
multi-process user-space RCU.

3) per-cpu LIFO linked-list (unlimited size stack)

A per-cpu LIFO linked-list has a "push" and "pop" operation,
which respectively adds an item to the list, and removes an
item from the list.

The "push" operation can be implemented as a rseq consisting of
a word comparison instruction against head followed by a word store
(commit) to head. Its cpu_opv fallback can be implemented as a
word-compare followed by word-store as well.

The "pop" operation can be implemented as a rseq consisting of
loading head, comparing it against NULL, loading the next pointer
at the right offset within the head item, and the next pointer as
a new head, returning the old head on success.

The cpu_opv fallback for "pop" differs from its rseq algorithm:
considering that cpu_opv requires to know all pointers at system
call entry so it can pin all pages, so cpu_opv cannot simply load
head and then load the head->next address within the preempt-off
critical section. User-space needs to pass the head and head->next
addresses to the kernel, and the kernel needs to check that the
head address is unchanged since it has been loaded by user-space.
However, when accessing head->next in a ABA situation, it's
possible that head is unchanged, but loading head->next can
result in a page fault due to a concurrently freed head object.
This is why the "expect_fault" operation field is introduced: if a
fault is triggered by this access, "-EAGAIN" will be returned by
cpu_opv rather than -EFAULT, thus indicating the the operation
vector should be attempted again. The "pop" operation can thus be
implemented as a word comparison of head against the head loaded
by user-space, followed by a load of the head->next pointer (which
may fault), and a store of that pointer as a new head.

4) per-cpu LIFO ring buffer with pointers to objects (fixed-sized stack)

This structure is useful for passing around allocated objects
by passing pointers through per-cpu fixed-sized stack.

The "push" side can be implemented with a check of the current
offset against the maximum buffer length, followed by a rseq
consisting of a comparison of the previously loaded offset
against the current offset, a word "try store" operation into the
next ring buffer array index (it's OK to abort after a try-store,
since it's not the commit, and its side-effect can be overwritten),
then followed by a word-store to increment the current offset (commit).

The "push" cpu_opv fallback can be done with the comparison, and
two consecutive word stores, all within the preempt-off section.

The "pop" side can be implemented with a check that offset is not
0 (whether the buffer is empty), a load of the "head" pointer before the
offset array index, followed by a rseq consisting of a word
comparison checking that the offset is unchanged since previously
loaded, another check ensuring that the "head" pointer is unchanged,
followed by a store decrementing the current offset.

The cpu_opv "pop" can be implemented with the same algorithm
as the rseq fast-path (compare, compare, store).

5) per-cpu LIFO ring buffer with pointers to objects (fixed-sized stack)
   supporting "peek" from remote CPU

In order to implement work queues with work-stealing between CPUs, it is
useful to ensure the offset "commit" in scenario 4) "push" have a
store-release semantic, thus allowing remote CPU to load the offset
with acquire semantic, and load the top pointer, in order to check if
work-stealing should be performed. The task (work queue item) existence
should be protected by other means, e.g. RCU.

If the peek operation notices that work-stealing should indeed be
performed, a thread can use cpu_opv to move the task between per-cpu
workqueues, by first invoking cpu_opv passing the remote work queue
cpu number as argument to pop the task, and then again as "push" with
the target work queue CPU number.

6) per-cpu LIFO ring buffer with data copy (fixed-sized stack)
   (with and without acquire-release)

This structure is useful for passing around data without requiring
memory allocation by copying the data content into per-cpu fixed-sized
stack.

The "push" operation is performed with an offset comparison against
the buffer size (figuring out if the buffer is full), followed by
a rseq consisting of a comparison of the offset, a try-memcpy attempting
to copy the data content into the buffer (which can be aborted and
overwritten), and a final store incrementing the offset.

The cpu_opv fallback needs to same operations, except that the memcpy
is guaranteed to complete, given that it is performed with preemption
disabled. This requires a memcpy operation supporting length up to 4kB.

The "pop" operation is similar to the "push, except that the offset
is first compared to 0 to ensure the buffer is not empty. The
copy source is the ring buffer, and the destination is an output
buffer.

7) per-cpu FIFO ring buffer (fixed-sized queue)

This structure is useful wherever a FIFO behavior (queue) is needed.
One major use-case is tracer ring buffer.

An implementation of this ring buffer has a "reserve", followed by
serialization of multiple bytes into the buffer, ended by a "commit".
The "reserve" can be implemented as a rseq consisting of a word
comparison followed by a word store. The reserve operation moves the
producer "head". The multi-byte serialization can be performed
non-atomically. Finally, the "commit" update can be performed with
a rseq "add" commit instruction with store-release semantic. The
ring buffer consumer reads the commit value with load-acquire
semantic to know whenever it is safe to read from the ring buffer.

This use-case requires that both "reserve" and "commit" operations
be performed on the same per-cpu ring buffer, even if a migration
happens between those operations. In the typical case, both operations
will happens on the same CPU and use rseq. In the unlikely event of a
migration, the cpu_opv system call will ensure the commit can be
performed on the right CPU by migrating the task to that CPU.

On the consumer side, an alternative to using store-release and
load-acquire on the commit counter would be to use cpu_opv to
ensure the commit counter load is performed on the right CPU. This
effectively allows moving a consumer thread between CPUs to execute
close to the ring buffer cache lines it will read.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Paul Turner <pjt@google.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Andrew Hunter <ahh@google.com>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Michael Kerrisk <mtk.manpages@gmail.com>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-api@vger.kernel.org
---
Changes since v1:
- handle CPU hotplug,
- cleanup implementation using function pointers: We can use function
  pointers to implement the operations rather than duplicating all the
  user-access code.
- refuse device pages: Performing cpu_opv operations on io map'd pages
  with preemption disabled could generate long preempt-off critical
  sections, which leads to unwanted scheduler latency. Return EFAULT if
  a device page is received as parameter
- restrict op vector to 4216 bytes length sum: Restrict the operation
  vector to length sum of:
  - 4096 bytes (typical page size on most architectures, should be
    enough for a string, or structures)
  - 15 * 8 bytes (typical operations on integers or pointers).
  The goal here is to keep the duration of preempt off critical section
  short, so we don't add significant scheduler latency.
- Add INIT_ONSTACK macro: Introduce the
  CPU_OP_FIELD_u32_u64_INIT_ONSTACK() macros to ensure that users
  correctly initialize the upper bits of CPU_OP_FIELD_u32_u64() on their
  stack to 0 on 32-bit architectures.
- Add CPU_MB_OP operation:
  Use-cases with:
  - two consecutive stores,
  - a mempcy followed by a store,
  require a memory barrier before the final store operation. A typical
  use-case is a store-release on the final store. Given that this is a
  slow path, just providing an explicit full barrier instruction should
  be sufficient.
- Add expect fault field:
  The use-case of list_pop brings interesting challenges. With rseq, we
  can use rseq_cmpnev_storeoffp_load(), and therefore load a pointer,
  compare it against NULL, add an offset, and load the target "next"
  pointer from the object, all within a single req critical section.

  Life is not so easy for cpu_opv in this use-case, mainly because we
  need to pin all pages we are going to touch in the preempt-off
  critical section beforehand. So we need to know the target object (in
  which we apply an offset to fetch the next pointer) when we pin pages
  before disabling preemption.

  So the approach is to load the head pointer and compare it against
  NULL in user-space, before doing the cpu_opv syscall. User-space can
  then compute the address of the head->next field, *without loading it*.

  The cpu_opv system call will first need to pin all pages associated
  with input data. This includes the page backing the head->next object,
  which may have been concurrently deallocated and unmapped. Therefore,
  in this case, getting -EFAULT when trying to pin those pages may
  happen: it just means they have been concurrently unmapped. This is
  an expected situation, and should just return -EAGAIN to user-space,
  to user-space can distinguish between "should retry" type of
  situations and actual errors that should be handled with extreme
  prejudice to the program (e.g. abort()).

  Therefore, add "expect_fault" fields along with op input address
  pointers, so user-space can identify whether a fault when getting a
  field should return EAGAIN rather than EFAULT.
- Add compiler barrier between operations: Adding a compiler barrier
  between store operations in a cpu_opv sequence can be useful when
  paired with membarrier system call.

  An algorithm with a paired slow path and fast path can use
  sys_membarrier on the slow path to replace fast-path memory barriers
  by compiler barrier.

  Adding an explicit compiler barrier between operations allows
  cpu_opv to be used as fallback for operations meant to match
  the membarrier system call.

Changes since v2:

- Fix memory leak by introducing struct cpu_opv_pinned_pages.
  Suggested by Boqun Feng.
- Cast argument 1 passed to access_ok from integer to void __user *,
  fixing sparse warning.

Changes since v3:

- Fix !SMP by adding push_task_to_cpu() empty static inline.
- Add missing sys_cpu_opv() asmlinkage declaration to
  include/linux/syscalls.h.

Changes since v4:

- Cleanup based on Thomas Gleixner's feedback.
- Handle retry in case where the scheduler migrates the thread away
  from the target CPU after migration within the syscall rather than
  returning EAGAIN to user-space.
- Move push_task_to_cpu() to its own patch.
- New scheme for touching user-space memory:
   1) get_user_pages_fast() to pin/get all pages (which can sleep),
   2) vm_map_ram() those pages
   3) grab mmap_sem (read lock)
   4) __get_user_pages_fast() (or get_user_pages() on failure)
      -> Confirm that the same page pointers are returned. This
         catches cases where COW mappings are changed concurrently.
      -> If page pointers differ, or on gup failure, release mmap_sem,
         vm_unmap_ram/put_page and retry from step (1).
      -> perform put_page on the extra reference immediately for each
         page.
   5) preempt disable
   6) Perform operations on vmap. Those operations are normal
      loads/stores/memcpy.
   7) preempt enable
   8) release mmap_sem
   9) vm_unmap_ram() all virtual addresses
  10) put_page() all pages
- Handle architectures with VIVT caches along with vmap(): call
  flush_kernel_vmap_range() after each "write" operation. This
  ensures that the user-space mapping and vmap reach a consistent
  state between each operation.
- Depend on MMU for is_zero_pfn(). e.g. Blackfin and SH architectures
  don't provide the zero_pfn symbol.

---
Man page associated:

CPU_OPV(2)              Linux Programmer's Manual             CPU_OPV(2)

NAME
       cpu_opv - CPU preempt-off operation vector system call

SYNOPSIS
       #include <linux/cpu_opv.h>

       int cpu_opv(struct cpu_op * cpu_opv, int cpuopcnt, int cpu, int flags);

DESCRIPTION
       The cpu_opv system call executes a vector of operations on behalf
       of user-space on a specific CPU with preemption disabled.

       The operations available are: comparison, memcpy, add,  or,  and,
       xor, left shift, right shift, and memory barrier. The system call
       receives a CPU number from user-space as argument, which  is  the
       CPU on which those operations need to be performed.  All pointers
       in the ops must have been set up to point to the per  CPU  memory
       of  the CPU on which the operations should be executed. The "com‐
       parison" operation can be used to check that the data used in the
       preparation  step  did  not  change between preparation of system
       call inputs and operation execution within the preempt-off criti‐
       cal section.

       An overall maximum of 4216 bytes in enforced on the sum of opera‐
       tion length within an operation vector, so user-space cannot gen‐
       erate  a too long preempt-off critical section. Each operation is
       also limited a length of 4096 bytes. A maximum limit of 16 opera‐
       tions per cpu_opv syscall invocation is enforced.

       If the thread is not running on the requested CPU, it is migrated
       to it.

       The layout of struct cpu_opv is as follows:

       Fields

           op Operation of type enum cpu_op_type to perform. This opera‐
              tion type selects the associated "u" union field.

           len
              Length (in bytes) of data to consider for this operation.

           u.compare_op
              For a CPU_COMPARE_EQ_OP , and CPU_COMPARE_NE_OP , contains
              the  a  and  b pointers to compare. The expect_fault_a and
              expect_fault_b fields indicate whether a page fault should
              be expected for each of those pointers.  If expect_fault_a
              , or expect_fault_b is set, EAGAIN is returned  on  fault,
              else  EFAULT is returned. The len field is allowed to take
              values from 0 to 4096 for comparison operations.

           u.memcpy_op
              For a CPU_MEMCPY_OP , contains the dst and  src  pointers,
              expressing  a  copy  of src into dst. The expect_fault_dst
              and expect_fault_src fields indicate whether a page  fault
              should  be  expected  for  each  of  those  pointers.   If
              expect_fault_dst , or expect_fault_src is set,  EAGAIN  is
              returned  on fault, else EFAULT is returned. The len field
              is allowed to take values from 0 to 4096 for memcpy opera‐
              tions.

           u.arithmetic_op
              For   a  CPU_ADD_OP  ,  contains  the  p  ,  count  ,  and
              expect_fault_p fields, which are respectively a pointer to
              the  memory location to increment, the 64-bit signed inte‐
              ger value to add, and  whether  a  page  fault  should  be
              expected  for  p  .   If  expect_fault_p is set, EAGAIN is
              returned on fault, else EFAULT is returned. The len  field
              is  allowed  to take values of 1, 2, 4, 8 bytes for arith‐
              metic operations.

           u.bitwise_op
              For a CPU_OR_OP , CPU_AND_OP , and CPU_XOR_OP  ,  contains
              the  p  ,  mask  ,  and  expect_fault_p  fields, which are
              respectively a pointer to the memory location  to  target,
              the  mask  to  apply,  and  whether a page fault should be
              expected for p .  If  expect_fault_p  is  set,  EAGAIN  is
              returned  on fault, else EFAULT is returned. The len field
              is allowed to take values of 1, 2, 4, 8 bytes for  bitwise
              operations.

           u.shift_op
              For a CPU_LSHIFT_OP , and CPU_RSHIFT_OP , contains the p ,
              bits , and expect_fault_p fields, which are respectively a
              pointer  to  the  memory location to target, the number of
              bits to shift either left of right,  and  whether  a  page
              fault  should  be  expected  for p .  If expect_fault_p is
              set, EAGAIN is returned on fault, else EFAULT is returned.
              The  len  field  is  allowed  to take values of 1, 2, 4, 8
              bytes for shift operations. The bits field is  allowed  to
              take values between 0 and 63.

       The enum cpu_op_types contains the following operations:

       · CPU_COMPARE_EQ_OP:  Compare  whether  two  memory locations are
         equal,

       · CPU_COMPARE_NE_OP: Compare whether two memory locations differ,

       · CPU_MEMCPY_OP: Copy a source memory location  into  a  destina‐
         tion,

       · CPU_ADD_OP:  Increment  a  target  memory  location  of a given
         count,

       · CPU_OR_OP: Apply a "or" mask to a memory location,

       · CPU_AND_OP: Apply a "and" mask to a memory location,

       · CPU_XOR_OP: Apply a "xor" mask to a memory location,

       · CPU_LSHIFT_OP: Shift a memory location left of a  given  number
         of bits,

       · CPU_RSHIFT_OP:  Shift a memory location right of a given number
         of bits.

       · CPU_MB_OP: Issue a memory barrier.

         All of the operations above provide single-copy atomicity guar‐
         antees  for  word-sized, word-aligned target pointers, for both
         loads and stores.

       The cpuopcnt argument is the number of elements  in  the  cpu_opv
       array. It can take values from 0 to 16.

       The  cpu  argument  is  the  CPU  number  on  which the operation
       sequence needs to be executed.

       The flags argument is expected to be 0.

RETURN VALUE
       A return value of 0 indicates success. On error, -1 is  returned,
       and  errno is set appropriately. If a comparison operation fails,
       execution of the operation vector  is  stopped,  and  the  return
       value is the index after the comparison operation (values between
       1 and 16).

ERRORS
       EAGAIN cpu_opv() system call should be attempted again.

       EINVAL Either flags contains an invalid value, or cpu contains an
              invalid  value  or  a  value  not  allowed  by the current
              thread's allowed cpu mask, or cpuopcnt contains an invalid
              value, or the cpu_opv operation vector contains an invalid
              op value, or the  cpu_opv  operation  vector  contains  an
              invalid  len value, or the cpu_opv operation vector sum of
              len values is too large.

       ENOSYS The cpu_opv() system call is not implemented by this  ker‐
              nel.

       EFAULT cpu_opv  is  an  invalid  address,  or a pointer contained
              within an  operation  is  invalid  (and  a  fault  is  not
              expected for that pointer).

VERSIONS
       The cpu_opv() system call was added in Linux 4.X (TODO).

CONFORMING TO
       cpu_opv() is Linux-specific.

SEE ALSO
       membarrier(2), rseq(2)

Linux                          2017-11-10                     CPU_OPV(2)
---
 MAINTAINERS                  |    7 +
 include/linux/syscalls.h     |    3 +
 include/uapi/linux/cpu_opv.h |  114 +++++
 init/Kconfig                 |   16 +
 kernel/Makefile              |    1 +
 kernel/cpu_opv.c             | 1078 ++++++++++++++++++++++++++++++++++++++++++
 kernel/sys_ni.c              |    1 +
 7 files changed, 1220 insertions(+)
 create mode 100644 include/uapi/linux/cpu_opv.h
 create mode 100644 kernel/cpu_opv.c

diff --git a/MAINTAINERS b/MAINTAINERS
index 4ede6c16d49f..36c5246b385b 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3732,6 +3732,13 @@ B:	https://bugzilla.kernel.org
 F:	drivers/cpuidle/*
 F:	include/linux/cpuidle.h
 
+CPU NON-PREEMPTIBLE OPERATION VECTOR SUPPORT
+M:	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+L:	linux-kernel@vger.kernel.org
+S:	Supported
+F:	kernel/cpu_opv.c
+F:	include/uapi/linux/cpu_opv.h
+
 CRAMFS FILESYSTEM
 M:	Nicolas Pitre <nico@linaro.org>
 S:	Maintained
diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
index 340650b4ec54..32d289f41f62 100644
--- a/include/linux/syscalls.h
+++ b/include/linux/syscalls.h
@@ -67,6 +67,7 @@ struct perf_event_attr;
 struct file_handle;
 struct sigaltstack;
 struct rseq;
+struct cpu_op;
 union bpf_attr;
 
 #include <linux/types.h>
@@ -943,5 +944,7 @@ asmlinkage long sys_statx(int dfd, const char __user *path, unsigned flags,
 			  unsigned mask, struct statx __user *buffer);
 asmlinkage long sys_rseq(struct rseq __user *rseq, uint32_t rseq_len,
 			int flags, uint32_t sig);
+asmlinkage long sys_cpu_opv(struct cpu_op __user *ucpuopv, int cpuopcnt,
+			int cpu, int flags);
 
 #endif
diff --git a/include/uapi/linux/cpu_opv.h b/include/uapi/linux/cpu_opv.h
new file mode 100644
index 000000000000..ccd8167fc189
--- /dev/null
+++ b/include/uapi/linux/cpu_opv.h
@@ -0,0 +1,114 @@
+#ifndef _UAPI_LINUX_CPU_OPV_H
+#define _UAPI_LINUX_CPU_OPV_H
+
+/*
+ * linux/cpu_opv.h
+ *
+ * CPU preempt-off operation vector system call API
+ *
+ * Copyright (c) 2017 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifdef __KERNEL__
+# include <linux/types.h>
+#else
+# include <stdint.h>
+#endif
+
+#include <linux/types_32_64.h>
+
+#define CPU_OP_VEC_LEN_MAX		16
+#define CPU_OP_ARG_LEN_MAX		24
+/* Maximum data len per operation. */
+#define CPU_OP_DATA_LEN_MAX		4096
+/*
+ * Maximum data len for overall vector. Restrict the amount of user-space
+ * data touched by the kernel in non-preemptible context, so it does not
+ * introduce long scheduler latencies.
+ * This allows one copy of up to 4096 bytes, and 15 operations touching 8
+ * bytes each.
+ * This limit is applied to the sum of length specified for all operations
+ * in a vector.
+ */
+#define CPU_OP_MEMCPY_EXPECT_LEN	4096
+#define CPU_OP_EXPECT_LEN		8
+#define CPU_OP_VEC_DATA_LEN_MAX		\
+	(CPU_OP_MEMCPY_EXPECT_LEN +	\
+	 (CPU_OP_VEC_LEN_MAX - 1) * CPU_OP_EXPECT_LEN)
+
+enum cpu_op_type {
+	/* compare */
+	CPU_COMPARE_EQ_OP,
+	CPU_COMPARE_NE_OP,
+	/* memcpy */
+	CPU_MEMCPY_OP,
+	/* arithmetic */
+	CPU_ADD_OP,
+	/* bitwise */
+	CPU_OR_OP,
+	CPU_AND_OP,
+	CPU_XOR_OP,
+	/* shift */
+	CPU_LSHIFT_OP,
+	CPU_RSHIFT_OP,
+	/* memory barrier */
+	CPU_MB_OP,
+};
+
+/* Vector of operations to perform. Limited to 16. */
+struct cpu_op {
+	/* enum cpu_op_type. */
+	int32_t op;
+	/* data length, in bytes. */
+	uint32_t len;
+	union {
+		struct {
+			LINUX_FIELD_u32_u64(a);
+			LINUX_FIELD_u32_u64(b);
+			uint8_t expect_fault_a;
+			uint8_t expect_fault_b;
+		} compare_op;
+		struct {
+			LINUX_FIELD_u32_u64(dst);
+			LINUX_FIELD_u32_u64(src);
+			uint8_t expect_fault_dst;
+			uint8_t expect_fault_src;
+		} memcpy_op;
+		struct {
+			LINUX_FIELD_u32_u64(p);
+			int64_t count;
+			uint8_t expect_fault_p;
+		} arithmetic_op;
+		struct {
+			LINUX_FIELD_u32_u64(p);
+			uint64_t mask;
+			uint8_t expect_fault_p;
+		} bitwise_op;
+		struct {
+			LINUX_FIELD_u32_u64(p);
+			uint32_t bits;
+			uint8_t expect_fault_p;
+		} shift_op;
+		char __padding[CPU_OP_ARG_LEN_MAX];
+	} u;
+};
+
+#endif /* _UAPI_LINUX_CPU_OPV_H */
diff --git a/init/Kconfig b/init/Kconfig
index 88e36395390f..8a4995ed1d19 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -1404,6 +1404,7 @@ config RSEQ
 	bool "Enable rseq() system call" if EXPERT
 	default y
 	depends on HAVE_RSEQ
+	select CPU_OPV
 	select MEMBARRIER
 	help
 	  Enable the restartable sequences system call. It provides a
@@ -1414,6 +1415,21 @@ config RSEQ
 
 	  If unsure, say Y.
 
+# CPU_OPV depends on MMU for is_zero_pfn()
+config CPU_OPV
+	bool "Enable cpu_opv() system call" if EXPERT
+	default y
+	depends on MMU
+	help
+	  Enable the CPU preempt-off operation vector system call.
+	  It allows user-space to perform a sequence of operations on
+	  per-cpu data with preemption disabled. Useful as
+	  single-stepping fall-back for restartable sequences, and for
+	  performing more complex operations on per-cpu data that would
+	  not be otherwise possible to do with restartable sequences.
+
+	  If unsure, say Y.
+
 config EMBEDDED
 	bool "Embedded system"
 	option allnoconfig_y
diff --git a/kernel/Makefile b/kernel/Makefile
index 3574669dafd9..cac8855196ff 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -113,6 +113,7 @@ obj-$(CONFIG_TORTURE_TEST) += torture.o
 
 obj-$(CONFIG_HAS_IOMEM) += memremap.o
 obj-$(CONFIG_RSEQ) += rseq.o
+obj-$(CONFIG_CPU_OPV) += cpu_opv.o
 
 $(obj)/configs.o: $(obj)/config_data.h
 
diff --git a/kernel/cpu_opv.c b/kernel/cpu_opv.c
new file mode 100644
index 000000000000..965fbf0a86b0
--- /dev/null
+++ b/kernel/cpu_opv.c
@@ -0,0 +1,1078 @@
+/*
+ * CPU preempt-off operation vector system call
+ *
+ * It allows user-space to perform a sequence of operations on per-cpu
+ * data with preemption disabled. Useful as single-stepping fall-back
+ * for restartable sequences, and for performing more complex operations
+ * on per-cpu data that would not be otherwise possible to do with
+ * restartable sequences.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * Copyright (C) 2017, EfficiOS Inc.,
+ * Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ */
+
+#include <linux/sched.h>
+#include <linux/uaccess.h>
+#include <linux/syscalls.h>
+#include <linux/cpu_opv.h>
+#include <linux/types.h>
+#include <linux/mutex.h>
+#include <linux/pagemap.h>
+#include <linux/mm.h>
+#include <asm/ptrace.h>
+#include <asm/byteorder.h>
+#include <asm/cacheflush.h>
+
+#include "sched/sched.h"
+
+/*
+ * Typical invocation of cpu_opv need few virtual address pointers. Keep
+ * those in an array on the stack of the cpu_opv system call up to
+ * this limit, beyond which the array is dynamically allocated.
+ */
+#define NR_VADDR_ON_STACK		8
+
+/* Maximum pages per op. */
+#define CPU_OP_MAX_PAGES		4
+
+/* Maximum number of virtual addresses per op. */
+#define CPU_OP_VEC_MAX_ADDR		(2 * CPU_OP_VEC_LEN_MAX)
+
+union op_fn_data {
+	uint8_t _u8;
+	uint16_t _u16;
+	uint32_t _u32;
+	uint64_t _u64;
+#if (BITS_PER_LONG < 64)
+	uint32_t _u64_split[2];
+#endif
+};
+
+struct vaddr {
+	unsigned long mem;
+	unsigned long uaddr;
+	struct page *pages[2];
+	unsigned int nr_pages;
+	int write;
+};
+
+struct cpu_opv_vaddr {
+	struct vaddr *addr;
+	size_t nr_vaddr;
+	bool is_kmalloc;
+};
+
+typedef int (*op_fn_t)(union op_fn_data *data, uint64_t v, uint32_t len);
+
+/*
+ * Provide mutual exclution for threads executing a cpu_opv against an
+ * offline CPU.
+ */
+static DEFINE_MUTEX(cpu_opv_offline_lock);
+
+/*
+ * The cpu_opv system call executes a vector of operations on behalf of
+ * user-space on a specific CPU with preemption disabled. It is inspired
+ * by readv() and writev() system calls which take a "struct iovec"
+ * array as argument.
+ *
+ * The operations available are: comparison, memcpy, add, or, and, xor,
+ * left shift, right shift, and memory barrier. The system call receives
+ * a CPU number from user-space as argument, which is the CPU on which
+ * those operations need to be performed.  All pointers in the ops must
+ * have been set up to point to the per CPU memory of the CPU on which
+ * the operations should be executed. The "comparison" operation can be
+ * used to check that the data used in the preparation step did not
+ * change between preparation of system call inputs and operation
+ * execution within the preempt-off critical section.
+ *
+ * The reason why we require all pointer offsets to be calculated by
+ * user-space beforehand is because we need to use get_user_pages_fast()
+ * to first pin all pages touched by each operation. This takes care of
+ * faulting-in the pages. Then, preemption is disabled, and the
+ * operations are performed atomically with respect to other thread
+ * execution on that CPU, without generating any page fault.
+ *
+ * An overall maximum of 4216 bytes in enforced on the sum of operation
+ * length within an operation vector, so user-space cannot generate a
+ * too long preempt-off critical section (cache cold critical section
+ * duration measured as 4.7µs on x86-64). Each operation is also limited
+ * a length of 4096 bytes, meaning that an operation can touch a
+ * maximum of 4 pages (memcpy: 2 pages for source, 2 pages for
+ * destination if addresses are not aligned on page boundaries).
+ *
+ * If the thread is not running on the requested CPU, it is migrated to
+ * it.
+ */
+
+static unsigned long cpu_op_range_nr_pages(unsigned long addr,
+					   unsigned long len)
+{
+	return ((addr + len - 1) >> PAGE_SHIFT) - (addr >> PAGE_SHIFT) + 1;
+}
+
+static int cpu_op_count_pages(unsigned long addr, unsigned long len)
+{
+	unsigned long nr_pages;
+
+	if (!len)
+		return 0;
+	nr_pages = cpu_op_range_nr_pages(addr, len);
+	if (nr_pages > 2) {
+		WARN_ON(1);
+		return -EINVAL;
+	}
+	return nr_pages;
+}
+
+static struct vaddr *cpu_op_alloc_vaddr_vector(int nr_vaddr)
+{
+	return kzalloc(nr_vaddr * sizeof(struct vaddr), GFP_KERNEL);
+}
+
+/*
+ * Check operation types and length parameters. Count number of pages.
+ */
+static int cpu_opv_check_op(struct cpu_op *op, int *nr_vaddr, uint32_t *sum)
+{
+	int ret;
+
+	switch (op->op) {
+	case CPU_MB_OP:
+		break;
+	default:
+		*sum += op->len;
+	}
+
+	/* Validate inputs. */
+	switch (op->op) {
+	case CPU_COMPARE_EQ_OP:
+	case CPU_COMPARE_NE_OP:
+	case CPU_MEMCPY_OP:
+		if (op->len > CPU_OP_DATA_LEN_MAX)
+			return -EINVAL;
+		break;
+	case CPU_ADD_OP:
+	case CPU_OR_OP:
+	case CPU_AND_OP:
+	case CPU_XOR_OP:
+		switch (op->len) {
+		case 1:
+		case 2:
+		case 4:
+		case 8:
+			break;
+		default:
+			return -EINVAL;
+		}
+		break;
+	case CPU_LSHIFT_OP:
+	case CPU_RSHIFT_OP:
+		switch (op->len) {
+		case 1:
+			if (op->u.shift_op.bits > 7)
+				return -EINVAL;
+			break;
+		case 2:
+			if (op->u.shift_op.bits > 15)
+				return -EINVAL;
+			break;
+		case 4:
+			if (op->u.shift_op.bits > 31)
+				return -EINVAL;
+			break;
+		case 8:
+			if (op->u.shift_op.bits > 63)
+				return -EINVAL;
+			break;
+		default:
+			return -EINVAL;
+		}
+		break;
+	case CPU_MB_OP:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* Count pages and virtual addresses. */
+	switch (op->op) {
+	case CPU_COMPARE_EQ_OP:
+	case CPU_COMPARE_NE_OP:
+		ret = cpu_op_count_pages(op->u.compare_op.a, op->len);
+		if (ret < 0)
+			return ret;
+		ret = cpu_op_count_pages(op->u.compare_op.b, op->len);
+		if (ret < 0)
+			return ret;
+		*nr_vaddr += 2;
+		break;
+	case CPU_MEMCPY_OP:
+		ret = cpu_op_count_pages(op->u.memcpy_op.dst, op->len);
+		if (ret < 0)
+			return ret;
+		ret = cpu_op_count_pages(op->u.memcpy_op.src, op->len);
+		if (ret < 0)
+			return ret;
+		*nr_vaddr += 2;
+		break;
+	case CPU_ADD_OP:
+		ret = cpu_op_count_pages(op->u.arithmetic_op.p, op->len);
+		if (ret < 0)
+			return ret;
+		(*nr_vaddr)++;
+		break;
+	case CPU_OR_OP:
+	case CPU_AND_OP:
+	case CPU_XOR_OP:
+		ret = cpu_op_count_pages(op->u.bitwise_op.p, op->len);
+		if (ret < 0)
+			return ret;
+		(*nr_vaddr)++;
+		break;
+	case CPU_LSHIFT_OP:
+	case CPU_RSHIFT_OP:
+		ret = cpu_op_count_pages(op->u.shift_op.p, op->len);
+		if (ret < 0)
+			return ret;
+		(*nr_vaddr)++;
+		break;
+	case CPU_MB_OP:
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+/*
+ * Check operation types and length parameters. Count number of pages.
+ */
+static int cpu_opv_check(struct cpu_op *cpuopv, int cpuopcnt, int *nr_vaddr)
+{
+	uint32_t sum = 0;
+	int i, ret;
+
+	for (i = 0; i < cpuopcnt; i++) {
+		ret = cpu_opv_check_op(&cpuopv[i], nr_vaddr, &sum);
+		if (ret)
+			return ret;
+	}
+	if (sum > CPU_OP_VEC_DATA_LEN_MAX)
+		return -EINVAL;
+	return 0;
+}
+
+static int cpu_op_check_page(struct page *page, int write)
+{
+	struct address_space *mapping;
+
+	if (is_zone_device_page(page))
+		return -EFAULT;
+
+	/*
+	 * The page lock protects many things but in this context the page
+	 * lock stabilizes mapping, prevents inode freeing in the shared
+	 * file-backed region case and guards against movement to swap
+	 * cache.
+	 *
+	 * Strictly speaking the page lock is not needed in all cases being
+	 * considered here and page lock forces unnecessarily serialization
+	 * From this point on, mapping will be re-verified if necessary and
+	 * page lock will be acquired only if it is unavoidable
+	 *
+	 * Mapping checks require the head page for any compound page so the
+	 * head page and mapping is looked up now.
+	 */
+	page = compound_head(page);
+	mapping = READ_ONCE(page->mapping);
+
+	/*
+	 * If page->mapping is NULL, then it cannot be a PageAnon page;
+	 * but it might be the ZERO_PAGE (which is OK to read from), or
+	 * in the gate area or in a special mapping (for which this
+	 * check should fail); or it may have been a good file page when
+	 * get_user_pages_fast found it, but truncated or holepunched or
+	 * subjected to invalidate_complete_page2 before the page lock
+	 * is acquired (also cases which should fail). Given that a
+	 * reference to the page is currently held, refcount care in
+	 * invalidate_complete_page's remove_mapping prevents
+	 * drop_caches from setting mapping to NULL concurrently.
+	 *
+	 * The case to guard against is when memory pressure cause
+	 * shmem_writepage to move the page from filecache to swapcache
+	 * concurrently: an unlikely race, but a retry for page->mapping
+	 * is required in that situation.
+	 */
+	if (!mapping) {
+		int shmem_swizzled;
+
+		/*
+		 * Check again with page lock held to guard against
+		 * memory pressure making shmem_writepage move the page
+		 * from filecache to swapcache.
+		 */
+		lock_page(page);
+		shmem_swizzled = PageSwapCache(page) || page->mapping;
+		unlock_page(page);
+		if (shmem_swizzled)
+			return -EAGAIN;
+		/*
+		 * It is valid to read from, but invalid to write to the
+		 * ZERO_PAGE.
+		 */
+		if (!(is_zero_pfn(page_to_pfn(page)) ||
+		      is_huge_zero_page(page)) || write) {
+			return -EFAULT;
+		}
+	}
+	return 0;
+}
+
+static int cpu_op_check_pages(struct page **pages,
+			      unsigned long nr_pages,
+			      int write)
+{
+	unsigned long i;
+
+	for (i = 0; i < nr_pages; i++) {
+		int ret;
+
+		ret = cpu_op_check_page(pages[i], write);
+		if (ret)
+			return ret;
+	}
+	return 0;
+}
+
+static int cpu_op_pin_pages(unsigned long addr, unsigned long len,
+			    struct cpu_opv_vaddr *vaddr_ptrs,
+			    unsigned long *vaddr, int write)
+{
+	struct page *pages[2];
+	int ret, nr_pages, nr_put_pages, n;
+	unsigned long _vaddr;
+	struct vaddr *va;
+
+	nr_pages = cpu_op_count_pages(addr, len);
+	if (!nr_pages)
+		return 0;
+again:
+	ret = get_user_pages_fast(addr, nr_pages, write, pages);
+	if (ret < nr_pages) {
+		if (ret >= 0) {
+			nr_put_pages = ret;
+			ret = -EFAULT;
+		} else {
+			nr_put_pages = 0;
+		}
+		goto error;
+	}
+	ret = cpu_op_check_pages(pages, nr_pages, write);
+	if (ret) {
+		nr_put_pages = nr_pages;
+		goto error;
+	}
+	va = &vaddr_ptrs->addr[vaddr_ptrs->nr_vaddr++];
+	_vaddr = (unsigned long)vm_map_ram(pages, nr_pages, numa_node_id(),
+					   PAGE_KERNEL);
+	if (!_vaddr) {
+		nr_put_pages = nr_pages;
+		ret = -ENOMEM;
+		goto error;
+	}
+	va->mem = _vaddr;
+	va->uaddr = addr;
+	for (n = 0; n < nr_pages; n++)
+		va->pages[n] = pages[n];
+	va->nr_pages = nr_pages;
+	va->write = write;
+	*vaddr = _vaddr + (addr & ~PAGE_MASK);
+	return 0;
+
+error:
+	for (n = 0; n < nr_put_pages; n++)
+		put_page(pages[n]);
+	/*
+	 * Retry if a page has been faulted in, or is being swapped in.
+	 */
+	if (ret == -EAGAIN)
+		goto again;
+	return ret;
+}
+
+static int cpu_opv_pin_pages_op(struct cpu_op *op,
+				struct cpu_opv_vaddr *vaddr_ptrs,
+				bool *expect_fault)
+{
+	int ret;
+	unsigned long vaddr = 0;
+
+	switch (op->op) {
+	case CPU_COMPARE_EQ_OP:
+	case CPU_COMPARE_NE_OP:
+		ret = -EFAULT;
+		*expect_fault = op->u.compare_op.expect_fault_a;
+		if (!access_ok(VERIFY_READ,
+			       (void __user *)op->u.compare_op.a,
+			       op->len))
+			return ret;
+		ret = cpu_op_pin_pages(op->u.compare_op.a, op->len,
+				       vaddr_ptrs, &vaddr, 0);
+		if (ret)
+			return ret;
+		op->u.compare_op.a = vaddr;
+		ret = -EFAULT;
+		*expect_fault = op->u.compare_op.expect_fault_b;
+		if (!access_ok(VERIFY_READ,
+			       (void __user *)op->u.compare_op.b,
+			       op->len))
+			return ret;
+		ret = cpu_op_pin_pages(op->u.compare_op.b, op->len,
+				       vaddr_ptrs, &vaddr, 0);
+		if (ret)
+			return ret;
+		op->u.compare_op.b = vaddr;
+		break;
+	case CPU_MEMCPY_OP:
+		ret = -EFAULT;
+		*expect_fault = op->u.memcpy_op.expect_fault_dst;
+		if (!access_ok(VERIFY_WRITE,
+			       (void __user *)op->u.memcpy_op.dst,
+			       op->len))
+			return ret;
+		ret = cpu_op_pin_pages(op->u.memcpy_op.dst, op->len,
+				       vaddr_ptrs, &vaddr, 1);
+		if (ret)
+			return ret;
+		op->u.memcpy_op.dst = vaddr;
+		ret = -EFAULT;
+		*expect_fault = op->u.memcpy_op.expect_fault_src;
+		if (!access_ok(VERIFY_READ,
+			       (void __user *)op->u.memcpy_op.src,
+			       op->len))
+			return ret;
+		ret = cpu_op_pin_pages(op->u.memcpy_op.src, op->len,
+				       vaddr_ptrs, &vaddr, 0);
+		if (ret)
+			return ret;
+		op->u.memcpy_op.src = vaddr;
+		break;
+	case CPU_ADD_OP:
+		ret = -EFAULT;
+		*expect_fault = op->u.arithmetic_op.expect_fault_p;
+		if (!access_ok(VERIFY_WRITE,
+			       (void __user *)op->u.arithmetic_op.p,
+			       op->len))
+			return ret;
+		ret = cpu_op_pin_pages(op->u.arithmetic_op.p, op->len,
+				       vaddr_ptrs, &vaddr, 1);
+		if (ret)
+			return ret;
+		op->u.arithmetic_op.p = vaddr;
+		break;
+	case CPU_OR_OP:
+	case CPU_AND_OP:
+	case CPU_XOR_OP:
+		ret = -EFAULT;
+		*expect_fault = op->u.bitwise_op.expect_fault_p;
+		if (!access_ok(VERIFY_WRITE,
+			       (void __user *)op->u.bitwise_op.p,
+			       op->len))
+			return ret;
+		ret = cpu_op_pin_pages(op->u.bitwise_op.p, op->len,
+				       vaddr_ptrs, &vaddr, 1);
+		if (ret)
+			return ret;
+		op->u.bitwise_op.p = vaddr;
+		break;
+	case CPU_LSHIFT_OP:
+	case CPU_RSHIFT_OP:
+		ret = -EFAULT;
+		*expect_fault = op->u.shift_op.expect_fault_p;
+		if (!access_ok(VERIFY_WRITE,
+			       (void __user *)op->u.shift_op.p,
+			       op->len))
+			return ret;
+		ret = cpu_op_pin_pages(op->u.shift_op.p, op->len,
+				       vaddr_ptrs, &vaddr, 1);
+		if (ret)
+			return ret;
+		op->u.shift_op.p = vaddr;
+		break;
+	case CPU_MB_OP:
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int cpu_opv_pin_pages(struct cpu_op *cpuop, int cpuopcnt,
+			     struct cpu_opv_vaddr *vaddr_ptrs)
+{
+	int ret, i;
+	bool expect_fault = false;
+
+	/* Check access, pin pages. */
+	for (i = 0; i < cpuopcnt; i++) {
+		ret = cpu_opv_pin_pages_op(&cpuop[i], vaddr_ptrs,
+					   &expect_fault);
+		if (ret)
+			goto error;
+	}
+	return 0;
+
+error:
+	/*
+	 * If faulting access is expected, return EAGAIN to user-space.
+	 * It allows user-space to distinguish between a fault caused by
+	 * an access which is expect to fault (e.g. due to concurrent
+	 * unmapping of underlying memory) from an unexpected fault from
+	 * which a retry would not recover.
+	 */
+	if (ret == -EFAULT && expect_fault)
+		return -EAGAIN;
+	return ret;
+}
+
+static int __op_get(union op_fn_data *data, void *p, size_t len)
+{
+	switch (len) {
+	case 1:
+		data->_u8 = READ_ONCE(*(uint8_t *)p);
+		break;
+	case 2:
+		data->_u16 = READ_ONCE(*(uint16_t *)p);
+		break;
+	case 4:
+		data->_u32 = READ_ONCE(*(uint32_t *)p);
+		break;
+	case 8:
+#if (BITS_PER_LONG == 64)
+		data->_u64 = READ_ONCE(*(uint64_t *)p);
+#else
+	{
+		data->_u64_split[0] = READ_ONCE(*(uint32_t *)p);
+		data->_u64_split[1] = READ_ONCE(*((uint32_t *)p + 1));
+	}
+#endif
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int __op_put(union op_fn_data *data, void *p, size_t len)
+{
+	switch (len) {
+	case 1:
+		WRITE_ONCE(*(uint8_t *)p, data->_u8);
+		break;
+	case 2:
+		WRITE_ONCE(*(uint16_t *)p, data->_u16);
+		break;
+	case 4:
+		WRITE_ONCE(*(uint32_t *)p, data->_u32);
+		break;
+	case 8:
+#if (BITS_PER_LONG == 64)
+		WRITE_ONCE(*(uint64_t *)p, data->_u64);
+#else
+	{
+		WRITE_ONCE(*(uint32_t *)p, data->_u64_split[0]);
+		WRITE_ONCE(*((uint32_t *)p + 1), data->_u64_split[1]);
+	}
+#endif
+		break;
+	default:
+		return -EINVAL;
+	}
+	flush_kernel_vmap_range(p, len);
+	return 0;
+}
+
+/* Return 0 if same, > 0 if different, < 0 on error. */
+static int do_cpu_op_compare(unsigned long _a, unsigned long _b, uint32_t len)
+{
+	void *a = (void *)_a;
+	void *b = (void *)_b;
+	union op_fn_data tmp[2];
+	int ret;
+
+	switch (len) {
+	case 1:
+	case 2:
+	case 4:
+	case 8:
+		if (!IS_ALIGNED(_a, len) || !IS_ALIGNED(_b, len))
+			goto memcmp;
+		break;
+	default:
+		goto memcmp;
+	}
+
+	ret = __op_get(&tmp[0], a, len);
+	if (ret)
+		return ret;
+	ret = __op_get(&tmp[1], b, len);
+	if (ret)
+		return ret;
+
+	switch (len) {
+	case 1:
+		ret = !!(tmp[0]._u8 != tmp[1]._u8);
+		break;
+	case 2:
+		ret = !!(tmp[0]._u16 != tmp[1]._u16);
+		break;
+	case 4:
+		ret = !!(tmp[0]._u32 != tmp[1]._u32);
+		break;
+	case 8:
+		ret = !!(tmp[0]._u64 != tmp[1]._u64);
+		break;
+	default:
+		return -EINVAL;
+	}
+	return ret;
+
+memcmp:
+	if (memcmp(a, b, len))
+		return 1;
+	return 0;
+}
+
+/* Return 0 on success, < 0 on error. */
+static int do_cpu_op_memcpy(unsigned long _dst, unsigned long _src,
+			    uint32_t len)
+{
+	void *dst = (void *)_dst;
+	void *src = (void *)_src;
+	union op_fn_data tmp;
+	int ret;
+
+	switch (len) {
+	case 1:
+	case 2:
+	case 4:
+	case 8:
+		if (!IS_ALIGNED(_dst, len) || !IS_ALIGNED(_src, len))
+			goto memcpy;
+		break;
+	default:
+		goto memcpy;
+	}
+
+	ret = __op_get(&tmp, src, len);
+	if (ret)
+		return ret;
+	return __op_put(&tmp, dst, len);
+
+memcpy:
+	memcpy(dst, src, len);
+	flush_kernel_vmap_range(dst, len);
+	return 0;
+}
+
+static int op_add_fn(union op_fn_data *data, uint64_t count, uint32_t len)
+{
+	switch (len) {
+	case 1:
+		data->_u8 += (uint8_t)count;
+		break;
+	case 2:
+		data->_u16 += (uint16_t)count;
+		break;
+	case 4:
+		data->_u32 += (uint32_t)count;
+		break;
+	case 8:
+		data->_u64 += (uint64_t)count;
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int op_or_fn(union op_fn_data *data, uint64_t mask, uint32_t len)
+{
+	switch (len) {
+	case 1:
+		data->_u8 |= (uint8_t)mask;
+		break;
+	case 2:
+		data->_u16 |= (uint16_t)mask;
+		break;
+	case 4:
+		data->_u32 |= (uint32_t)mask;
+		break;
+	case 8:
+		data->_u64 |= (uint64_t)mask;
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int op_and_fn(union op_fn_data *data, uint64_t mask, uint32_t len)
+{
+	switch (len) {
+	case 1:
+		data->_u8 &= (uint8_t)mask;
+		break;
+	case 2:
+		data->_u16 &= (uint16_t)mask;
+		break;
+	case 4:
+		data->_u32 &= (uint32_t)mask;
+		break;
+	case 8:
+		data->_u64 &= (uint64_t)mask;
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int op_xor_fn(union op_fn_data *data, uint64_t mask, uint32_t len)
+{
+	switch (len) {
+	case 1:
+		data->_u8 ^= (uint8_t)mask;
+		break;
+	case 2:
+		data->_u16 ^= (uint16_t)mask;
+		break;
+	case 4:
+		data->_u32 ^= (uint32_t)mask;
+		break;
+	case 8:
+		data->_u64 ^= (uint64_t)mask;
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int op_lshift_fn(union op_fn_data *data, uint64_t bits, uint32_t len)
+{
+	switch (len) {
+	case 1:
+		data->_u8 <<= (uint8_t)bits;
+		break;
+	case 2:
+		data->_u16 <<= (uint16_t)bits;
+		break;
+	case 4:
+		data->_u32 <<= (uint32_t)bits;
+		break;
+	case 8:
+		data->_u64 <<= (uint64_t)bits;
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int op_rshift_fn(union op_fn_data *data, uint64_t bits, uint32_t len)
+{
+	switch (len) {
+	case 1:
+		data->_u8 >>= (uint8_t)bits;
+		break;
+	case 2:
+		data->_u16 >>= (uint16_t)bits;
+		break;
+	case 4:
+		data->_u32 >>= (uint32_t)bits;
+		break;
+	case 8:
+		data->_u64 >>= (uint64_t)bits;
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+/* Return 0 on success, < 0 on error. */
+static int do_cpu_op_fn(op_fn_t op_fn, unsigned long _p, uint64_t v,
+			uint32_t len)
+{
+	union op_fn_data tmp;
+	void *p = (void *)_p;
+	int ret;
+
+	ret = __op_get(&tmp, p, len);
+	if (ret)
+		return ret;
+	ret = op_fn(&tmp, v, len);
+	if (ret)
+		return ret;
+	ret = __op_put(&tmp, p, len);
+	if (ret)
+		return ret;
+	return 0;
+}
+
+/*
+ * Return negative value on error, positive value if comparison
+ * fails, 0 on success.
+ */
+static int __do_cpu_opv_op(struct cpu_op *op)
+{
+	/* Guarantee a compiler barrier between each operation. */
+	barrier();
+
+	switch (op->op) {
+	case CPU_COMPARE_EQ_OP:
+		return do_cpu_op_compare(op->u.compare_op.a,
+					 op->u.compare_op.b,
+					 op->len);
+	case CPU_COMPARE_NE_OP:
+	{
+		int ret;
+
+		ret = do_cpu_op_compare(op->u.compare_op.a,
+					op->u.compare_op.b,
+					op->len);
+		if (ret < 0)
+			return ret;
+		/*
+		 * Stop execution, return positive value if comparison
+		 * is identical.
+		 */
+		if (ret == 0)
+			return 1;
+		return 0;
+	}
+	case CPU_MEMCPY_OP:
+		return do_cpu_op_memcpy(op->u.memcpy_op.dst,
+					op->u.memcpy_op.src,
+					op->len);
+	case CPU_ADD_OP:
+		return do_cpu_op_fn(op_add_fn, op->u.arithmetic_op.p,
+				    op->u.arithmetic_op.count, op->len);
+	case CPU_OR_OP:
+		return do_cpu_op_fn(op_or_fn, op->u.bitwise_op.p,
+				    op->u.bitwise_op.mask, op->len);
+	case CPU_AND_OP:
+		return do_cpu_op_fn(op_and_fn, op->u.bitwise_op.p,
+				    op->u.bitwise_op.mask, op->len);
+	case CPU_XOR_OP:
+		return do_cpu_op_fn(op_xor_fn, op->u.bitwise_op.p,
+				    op->u.bitwise_op.mask, op->len);
+	case CPU_LSHIFT_OP:
+		return do_cpu_op_fn(op_lshift_fn, op->u.shift_op.p,
+				    op->u.shift_op.bits, op->len);
+	case CPU_RSHIFT_OP:
+		return do_cpu_op_fn(op_rshift_fn, op->u.shift_op.p,
+				    op->u.shift_op.bits, op->len);
+	case CPU_MB_OP:
+		/* Memory barrier provided by this operation. */
+		smp_mb();
+		return 0;
+	default:
+		return -EINVAL;
+	}
+}
+
+static int __do_cpu_opv(struct cpu_op *cpuop, int cpuopcnt)
+{
+	int i, ret;
+
+	for (i = 0; i < cpuopcnt; i++) {
+		ret = __do_cpu_opv_op(&cpuop[i]);
+		/* If comparison fails, stop execution and return index + 1. */
+		if (ret > 0)
+			return i + 1;
+		/* On error, stop execution. */
+		if (ret < 0)
+			return ret;
+	}
+	return 0;
+}
+
+/*
+ * Check that the page pointers pinned by get_user_pages_fast()
+ * are still in the page table. Invoked with mmap_sem held.
+ * Return 0 if pointers match, -EAGAIN if they don't.
+ */
+static int vaddr_check(struct vaddr *vaddr)
+{
+	struct page *pages[2];
+	int ret, n;
+
+	ret = __get_user_pages_fast(vaddr->uaddr, vaddr->nr_pages,
+				    vaddr->write, pages);
+	for (n = 0; n < ret; n++)
+		put_page(pages[n]);
+	if (ret < vaddr->nr_pages) {
+		ret = get_user_pages(vaddr->uaddr, vaddr->nr_pages,
+				     vaddr->write ? FOLL_WRITE : 0,
+				     pages, NULL);
+		if (ret < 0)
+			return -EAGAIN;
+		for (n = 0; n < ret; n++)
+			put_page(pages[n]);
+		if (ret < vaddr->nr_pages)
+			return -EAGAIN;
+	}
+	for (n = 0; n < vaddr->nr_pages; n++) {
+		if (pages[n] != vaddr->pages[n])
+			return -EAGAIN;
+	}
+	return 0;
+}
+
+static int vaddr_ptrs_check(struct cpu_opv_vaddr *vaddr_ptrs)
+{
+	int i;
+
+	for (i = 0; i < vaddr_ptrs->nr_vaddr; i++) {
+		int ret;
+
+		ret = vaddr_check(&vaddr_ptrs->addr[i]);
+		if (ret)
+			return ret;
+	}
+	return 0;
+}
+
+static int do_cpu_opv(struct cpu_op *cpuop, int cpuopcnt,
+		      struct cpu_opv_vaddr *vaddr_ptrs, int cpu)
+{
+	struct mm_struct *mm = current->mm;
+	int ret;
+
+retry:
+	if (cpu != raw_smp_processor_id()) {
+		ret = push_task_to_cpu(current, cpu);
+		if (ret)
+			goto check_online;
+	}
+	down_read(&mm->mmap_sem);
+	ret = vaddr_ptrs_check(vaddr_ptrs);
+	if (ret)
+		goto end;
+	preempt_disable();
+	if (cpu != smp_processor_id()) {
+		preempt_enable();
+		up_read(&mm->mmap_sem);
+		goto retry;
+	}
+	ret = __do_cpu_opv(cpuop, cpuopcnt);
+	preempt_enable();
+end:
+	up_read(&mm->mmap_sem);
+	return ret;
+
+check_online:
+	if (!cpu_possible(cpu))
+		return -EINVAL;
+	get_online_cpus();
+	if (cpu_online(cpu)) {
+		put_online_cpus();
+		goto retry;
+	}
+	/*
+	 * CPU is offline. Perform operation from the current CPU with
+	 * cpu_online read lock held, preventing that CPU from coming online,
+	 * and with mutex held, providing mutual exclusion against other
+	 * CPUs also finding out about an offline CPU.
+	 */
+	down_read(&mm->mmap_sem);
+	ret = vaddr_ptrs_check(vaddr_ptrs);
+	if (ret)
+		goto offline_end;
+	mutex_lock(&cpu_opv_offline_lock);
+	ret = __do_cpu_opv(cpuop, cpuopcnt);
+	mutex_unlock(&cpu_opv_offline_lock);
+offline_end:
+	up_read(&mm->mmap_sem);
+	put_online_cpus();
+	return ret;
+}
+
+/*
+ * cpu_opv - execute operation vector on a given CPU with preempt off.
+ *
+ * Userspace should pass current CPU number as parameter.
+ */
+SYSCALL_DEFINE4(cpu_opv, struct cpu_op __user *, ucpuopv, int, cpuopcnt,
+		int, cpu, int, flags)
+{
+	struct vaddr vaddr_on_stack[NR_VADDR_ON_STACK];
+	struct cpu_op cpuopv[CPU_OP_VEC_LEN_MAX];
+	struct cpu_opv_vaddr vaddr_ptrs = {
+		.addr = vaddr_on_stack,
+		.nr_vaddr = 0,
+		.is_kmalloc = false,
+	};
+	int ret, i, nr_vaddr = 0;
+	bool retry = false;
+
+	if (unlikely(flags))
+		return -EINVAL;
+	if (unlikely(cpu < 0))
+		return -EINVAL;
+	if (cpuopcnt < 0 || cpuopcnt > CPU_OP_VEC_LEN_MAX)
+		return -EINVAL;
+	if (copy_from_user(cpuopv, ucpuopv, cpuopcnt * sizeof(struct cpu_op)))
+		return -EFAULT;
+	ret = cpu_opv_check(cpuopv, cpuopcnt, &nr_vaddr);
+	if (ret)
+		return ret;
+	if (nr_vaddr > NR_VADDR_ON_STACK) {
+		vaddr_ptrs.addr = cpu_op_alloc_vaddr_vector(nr_vaddr);
+		if (!vaddr_ptrs.addr) {
+			ret = -ENOMEM;
+			goto end;
+		}
+		vaddr_ptrs.is_kmalloc = true;
+	}
+again:
+	ret = cpu_opv_pin_pages(cpuopv, cpuopcnt, &vaddr_ptrs);
+	if (ret)
+		goto end;
+	ret = do_cpu_opv(cpuopv, cpuopcnt, &vaddr_ptrs, cpu);
+	if (ret == -EAGAIN)
+		retry = true;
+end:
+	for (i = 0; i < vaddr_ptrs.nr_vaddr; i++) {
+		struct vaddr *vaddr = &vaddr_ptrs.addr[i];
+		int j;
+
+		vm_unmap_ram((void *)vaddr->mem, vaddr->nr_pages);
+		for (j = 0; j < vaddr->nr_pages; j++) {
+			if (vaddr->write)
+				set_page_dirty(vaddr->pages[j]);
+			put_page(vaddr->pages[j]);
+		}
+	}
+	if (retry) {
+		retry = false;
+		vaddr_ptrs.nr_vaddr = 0;
+		goto again;
+	}
+	if (vaddr_ptrs.is_kmalloc)
+		kfree(vaddr_ptrs.addr);
+	return ret;
+}
diff --git a/kernel/sys_ni.c b/kernel/sys_ni.c
index bfa1ee1bf669..59e622296dc3 100644
--- a/kernel/sys_ni.c
+++ b/kernel/sys_ni.c
@@ -262,3 +262,4 @@ cond_syscall(sys_pkey_free);
 
 /* restartable sequence */
 cond_syscall(sys_rseq);
+cond_syscall(sys_cpu_opv);
-- 
2.11.0

[RFC PATCH for 4.16 11/21] x86: Wire up cpu_opv system call

[Mathieu Desnoyers]

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Paul Turner <pjt@google.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Andrew Hunter <ahh@google.com>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Michael Kerrisk <mtk.manpages@gmail.com>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-api@vger.kernel.org
---
 arch/x86/entry/syscalls/syscall_32.tbl | 1 +
 arch/x86/entry/syscalls/syscall_64.tbl | 1 +
 2 files changed, 2 insertions(+)

diff --git a/arch/x86/entry/syscalls/syscall_32.tbl b/arch/x86/entry/syscalls/syscall_32.tbl
index ba43ee75e425..afc6988fb2c8 100644
--- a/arch/x86/entry/syscalls/syscall_32.tbl
+++ b/arch/x86/entry/syscalls/syscall_32.tbl
@@ -392,3 +392,4 @@
 383	i386	statx			sys_statx
 384	i386	arch_prctl		sys_arch_prctl			compat_sys_arch_prctl
 385	i386	rseq			sys_rseq
+386	i386	cpu_opv			sys_cpu_opv
diff --git a/arch/x86/entry/syscalls/syscall_64.tbl b/arch/x86/entry/syscalls/syscall_64.tbl
index 3ad03495bbb9..ab5d1f9f9396 100644
--- a/arch/x86/entry/syscalls/syscall_64.tbl
+++ b/arch/x86/entry/syscalls/syscall_64.tbl
@@ -340,6 +340,7 @@
 331	common	pkey_free		sys_pkey_free
 332	common	statx			sys_statx
 333	common	rseq			sys_rseq
+334	common	cpu_opv			sys_cpu_opv
 
 #
 # x32-specific system call numbers start at 512 to avoid cache impact
-- 
2.11.0

[RFC PATCH for 4.16 12/21] powerpc: Wire up cpu_opv system call

[Mathieu Desnoyers]

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Benjamin Herrenschmidt <benh@kernel.crashing.org>
CC: Paul Mackerras <paulus@samba.org>
CC: Michael Ellerman <mpe@ellerman.id.au>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: linuxppc-dev@lists.ozlabs.org
---
 arch/powerpc/include/asm/systbl.h      | 1 +
 arch/powerpc/include/asm/unistd.h      | 2 +-
 arch/powerpc/include/uapi/asm/unistd.h | 1 +
 3 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/include/asm/systbl.h b/arch/powerpc/include/asm/systbl.h
index 964321a5799c..f9cdb896fbaa 100644
--- a/arch/powerpc/include/asm/systbl.h
+++ b/arch/powerpc/include/asm/systbl.h
@@ -390,3 +390,4 @@ COMPAT_SYS_SPU(pwritev2)
 SYSCALL(kexec_file_load)
 SYSCALL(statx)
 SYSCALL(rseq)
+SYSCALL(cpu_opv)
diff --git a/arch/powerpc/include/asm/unistd.h b/arch/powerpc/include/asm/unistd.h
index e76bd5601ea4..48f80f452e31 100644
--- a/arch/powerpc/include/asm/unistd.h
+++ b/arch/powerpc/include/asm/unistd.h
@@ -12,7 +12,7 @@
 #include <uapi/asm/unistd.h>
 
 
-#define NR_syscalls		385
+#define NR_syscalls		386
 
 #define __NR__exit __NR_exit
 
diff --git a/arch/powerpc/include/uapi/asm/unistd.h b/arch/powerpc/include/uapi/asm/unistd.h
index b1980fcd56d5..972a7d68c143 100644
--- a/arch/powerpc/include/uapi/asm/unistd.h
+++ b/arch/powerpc/include/uapi/asm/unistd.h
@@ -396,5 +396,6 @@
 #define __NR_kexec_file_load	382
 #define __NR_statx		383
 #define __NR_rseq		384
+#define __NR_cpu_opv		385
 
 #endif /* _UAPI_ASM_POWERPC_UNISTD_H_ */
-- 
2.11.0

[RFC PATCH for 4.16 13/21] arm: Wire up cpu_opv system call

[Mathieu Desnoyers]

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
CC: Russell King <linux-lFZ/pmaqli7XmaaqVzeoHQ@public.gmane.org>
CC: Catalin Marinas <catalin.marinas-5wv7dgnIgG8@public.gmane.org>
CC: Will Deacon <will.deacon-5wv7dgnIgG8@public.gmane.org>
CC: Thomas Gleixner <tglx-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org>
CC: Paul Turner <pjt-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>
CC: Andrew Hunter <ahh-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>
CC: Peter Zijlstra <peterz-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org>
CC: Andy Lutomirski <luto-kltTT9wpgjJwATOyAt5JVQ@public.gmane.org>
CC: Andi Kleen <andi-Vw/NltI1exuRpAAqCnN02g@public.gmane.org>
CC: Dave Watson <davejwatson-b10kYP2dOMg@public.gmane.org>
CC: Chris Lameter <cl-vYTEC60ixJUAvxtiuMwx3w@public.gmane.org>
CC: Ingo Molnar <mingo-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
CC: Ben Maurer <bmaurer-b10kYP2dOMg@public.gmane.org>
CC: Steven Rostedt <rostedt-nx8X9YLhiw1AfugRpC6u6w@public.gmane.org>
CC: "Paul E. McKenney" <paulmck-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
CC: Josh Triplett <josh-iaAMLnmF4UmaiuxdJuQwMA@public.gmane.org>
CC: Linus Torvalds <torvalds-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>
CC: Andrew Morton <akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>
CC: Boqun Feng <boqun.feng-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
CC: linux-api-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
---
 arch/arm/tools/syscall.tbl | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/arm/tools/syscall.tbl b/arch/arm/tools/syscall.tbl
index fbc74b5fa3ed..213ccfc2c437 100644
--- a/arch/arm/tools/syscall.tbl
+++ b/arch/arm/tools/syscall.tbl
@@ -413,3 +413,4 @@
 396	common	pkey_free		sys_pkey_free
 397	common	statx			sys_statx
 398	common	rseq			sys_rseq
+399	common	cpu_opv			sys_cpu_opv
-- 
2.11.0

[RFC PATCH for 4.16 14/21] selftests: lib.mk: Introduce OVERRIDE_TARGETS

[Mathieu Desnoyers]

Introduce OVERRIDE_TARGETS to allow tests to express dependencies on
header files and .so, which require to override the selftests lib.mk
targets.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Acked-by: Shuah Khan <shuahkh@osg.samsung.com>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Paul Turner <pjt@google.com>
CC: Andrew Hunter <ahh@google.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-kselftest@vger.kernel.org
CC: linux-api@vger.kernel.org
---
 tools/testing/selftests/lib.mk | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/tools/testing/selftests/lib.mk b/tools/testing/selftests/lib.mk
index 5bef05d6ba39..441d7bc63bb7 100644
--- a/tools/testing/selftests/lib.mk
+++ b/tools/testing/selftests/lib.mk
@@ -105,6 +105,9 @@ COMPILE.S = $(CC) $(ASFLAGS) $(CPPFLAGS) $(TARGET_ARCH) -c
 LINK.S = $(CC) $(ASFLAGS) $(CPPFLAGS) $(LDFLAGS) $(TARGET_ARCH)
 endif
 
+# Selftest makefiles can override those targets by setting
+# OVERRIDE_TARGETS = 1.
+ifeq ($(OVERRIDE_TARGETS),)
 $(OUTPUT)/%:%.c
 	$(LINK.c) $^ $(LDLIBS) -o $@
 
@@ -113,5 +116,6 @@ $(OUTPUT)/%.o:%.S
 
 $(OUTPUT)/%:%.S
 	$(LINK.S) $^ $(LDLIBS) -o $@
+endif
 
 .PHONY: run_tests all clean install emit_tests
-- 
2.11.0

[RFC PATCH for 4.16 15/21] cpu_opv: selftests: Implement selftests (v6)

[Mathieu Desnoyers]

Implement cpu_opv selftests. It needs to express dependencies on
header files and .so, which require to override the selftests
lib.mk targets. Use OVERRIDE_TARGETS define for this.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Acked-by: Shuah Khan <shuahkh@osg.samsung.com>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Paul Turner <pjt@google.com>
CC: Andrew Hunter <ahh@google.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-kselftest@vger.kernel.org
CC: linux-api@vger.kernel.org
---
Changes since v1:

- Expose similar library API as rseq:  Expose library API closely
  matching the rseq APIs, following removal of the event counter from
  the rseq kernel API.
- Update makefile to fix make run_tests dependency on "all".
- Introduce a OVERRIDE_TARGETS.

Changes since v2:

- Test page faults.

Changes since v3:

- Move lib.mk OVERRIDE_TARGETS change to its own patch.
- Printout TAP output.

Changes since v4:

- Retry internally within cpu_op_cmpnev_storeoffp_load().

Changes since v5:

- Test huge pages.
---
 MAINTAINERS                                        |    1 +
 tools/testing/selftests/Makefile                   |    1 +
 tools/testing/selftests/cpu-opv/.gitignore         |    1 +
 tools/testing/selftests/cpu-opv/Makefile           |   17 +
 .../testing/selftests/cpu-opv/basic_cpu_opv_test.c | 1258 ++++++++++++++++++++
 tools/testing/selftests/cpu-opv/cpu-op.c           |  352 ++++++
 tools/testing/selftests/cpu-opv/cpu-op.h           |   59 +
 7 files changed, 1689 insertions(+)
 create mode 100644 tools/testing/selftests/cpu-opv/.gitignore
 create mode 100644 tools/testing/selftests/cpu-opv/Makefile
 create mode 100644 tools/testing/selftests/cpu-opv/basic_cpu_opv_test.c
 create mode 100644 tools/testing/selftests/cpu-opv/cpu-op.c
 create mode 100644 tools/testing/selftests/cpu-opv/cpu-op.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 36c5246b385b..dce21dea5a63 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3738,6 +3738,7 @@ L:	linux-kernel@vger.kernel.org
 S:	Supported
 F:	kernel/cpu_opv.c
 F:	include/uapi/linux/cpu_opv.h
+F:	tools/testing/selftests/cpu-opv/
 
 CRAMFS FILESYSTEM
 M:	Nicolas Pitre <nico@linaro.org>
diff --git a/tools/testing/selftests/Makefile b/tools/testing/selftests/Makefile
index eaf599dc2137..fc1eba0e0130 100644
--- a/tools/testing/selftests/Makefile
+++ b/tools/testing/selftests/Makefile
@@ -5,6 +5,7 @@ TARGETS += breakpoints
 TARGETS += capabilities
 TARGETS += cpufreq
 TARGETS += cpu-hotplug
+TARGETS += cpu-opv
 TARGETS += efivarfs
 TARGETS += exec
 TARGETS += firmware
diff --git a/tools/testing/selftests/cpu-opv/.gitignore b/tools/testing/selftests/cpu-opv/.gitignore
new file mode 100644
index 000000000000..c7186eb95cf5
--- /dev/null
+++ b/tools/testing/selftests/cpu-opv/.gitignore
@@ -0,0 +1 @@
+basic_cpu_opv_test
diff --git a/tools/testing/selftests/cpu-opv/Makefile b/tools/testing/selftests/cpu-opv/Makefile
new file mode 100644
index 000000000000..21e63545d521
--- /dev/null
+++ b/tools/testing/selftests/cpu-opv/Makefile
@@ -0,0 +1,17 @@
+CFLAGS += -O2 -Wall -g -I./ -I../../../../usr/include/ -L./ -Wl,-rpath=./
+
+# Own dependencies because we only want to build against 1st prerequisite, but
+# still track changes to header files and depend on shared object.
+OVERRIDE_TARGETS = 1
+
+TEST_GEN_PROGS = basic_cpu_opv_test
+
+TEST_GEN_PROGS_EXTENDED = libcpu-op.so
+
+include ../lib.mk
+
+$(OUTPUT)/libcpu-op.so: cpu-op.c cpu-op.h
+	$(CC) $(CFLAGS) -shared -fPIC $< -o $@
+
+$(OUTPUT)/%: %.c $(TEST_GEN_PROGS_EXTENDED) cpu-op.h
+	$(CC) $(CFLAGS) $< -lcpu-op -o $@
diff --git a/tools/testing/selftests/cpu-opv/basic_cpu_opv_test.c b/tools/testing/selftests/cpu-opv/basic_cpu_opv_test.c
new file mode 100644
index 000000000000..e33daa0b683d
--- /dev/null
+++ b/tools/testing/selftests/cpu-opv/basic_cpu_opv_test.c
@@ -0,0 +1,1258 @@
+/*
+ * Basic test coverage for cpu_opv system call.
+ */
+
+#define _GNU_SOURCE
+#include <assert.h>
+#include <sched.h>
+#include <signal.h>
+#include <stdio.h>
+#include <string.h>
+#include <errno.h>
+#include <stdlib.h>
+#include <sys/time.h>
+#include <sys/mman.h>
+
+#include "../kselftest.h"
+
+#include "cpu-op.h"
+
+#define ARRAY_SIZE(arr)	(sizeof(arr) / sizeof((arr)[0]))
+
+#define TESTBUFLEN	4096
+#define TESTBUFLEN_CMP	16
+
+#define TESTBUFLEN_PAGE_MAX	65536
+
+#define NR_PF_ARRAY	16384
+#define PF_ARRAY_LEN	4096
+
+#define NR_HUGE_ARRAY	512
+#define HUGEMAPLEN	(NR_HUGE_ARRAY * PF_ARRAY_LEN)
+
+/* 64 MB arrays for page fault testing. */
+char pf_array_dst[NR_PF_ARRAY][PF_ARRAY_LEN];
+char pf_array_src[NR_PF_ARRAY][PF_ARRAY_LEN];
+
+static int test_compare_eq_op(char *a, char *b, size_t len)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_COMPARE_EQ_OP,
+			.len = len,
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.compare_op.a, a),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.compare_op.b, b),
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+	};
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_compare_eq_same(void)
+{
+	int i, ret;
+	char buf1[TESTBUFLEN];
+	char buf2[TESTBUFLEN];
+	const char *test_name = "test_compare_eq same";
+
+	/* Test compare_eq */
+	for (i = 0; i < TESTBUFLEN; i++)
+		buf1[i] = (char)i;
+	for (i = 0; i < TESTBUFLEN; i++)
+		buf2[i] = (char)i;
+	ret = test_compare_eq_op(buf2, buf1, TESTBUFLEN);
+	if (ret < 0) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (ret > 0) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, ret, 0);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_compare_eq_diff(void)
+{
+	int i, ret;
+	char buf1[TESTBUFLEN];
+	char buf2[TESTBUFLEN];
+	const char *test_name = "test_compare_eq different";
+
+	for (i = 0; i < TESTBUFLEN; i++)
+		buf1[i] = (char)i;
+	memset(buf2, 0, TESTBUFLEN);
+	ret = test_compare_eq_op(buf2, buf1, TESTBUFLEN);
+	if (ret < 0) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (ret == 0) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, ret, 1);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_compare_ne_op(char *a, char *b, size_t len)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_COMPARE_NE_OP,
+			.len = len,
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.compare_op.a, a),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.compare_op.b, b),
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+	};
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_compare_ne_same(void)
+{
+	int i, ret;
+	char buf1[TESTBUFLEN];
+	char buf2[TESTBUFLEN];
+	const char *test_name = "test_compare_ne same";
+
+	/* Test compare_ne */
+	for (i = 0; i < TESTBUFLEN; i++)
+		buf1[i] = (char)i;
+	for (i = 0; i < TESTBUFLEN; i++)
+		buf2[i] = (char)i;
+	ret = test_compare_ne_op(buf2, buf1, TESTBUFLEN);
+	if (ret < 0) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (ret == 0) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, ret, 1);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_compare_ne_diff(void)
+{
+	int i, ret;
+	char buf1[TESTBUFLEN];
+	char buf2[TESTBUFLEN];
+	const char *test_name = "test_compare_ne different";
+
+	for (i = 0; i < TESTBUFLEN; i++)
+		buf1[i] = (char)i;
+	memset(buf2, 0, TESTBUFLEN);
+	ret = test_compare_ne_op(buf2, buf1, TESTBUFLEN);
+	if (ret < 0) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (ret != 0) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, ret, 0);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_2compare_eq_op(char *a, char *b, char *c, char *d,
+		size_t len)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_COMPARE_EQ_OP,
+			.len = len,
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.compare_op.a, a),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.compare_op.b, b),
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+		[1] = {
+			.op = CPU_COMPARE_EQ_OP,
+			.len = len,
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.compare_op.a, c),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.compare_op.b, d),
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+	};
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_2compare_eq_index(void)
+{
+	int i, ret;
+	char buf1[TESTBUFLEN_CMP];
+	char buf2[TESTBUFLEN_CMP];
+	char buf3[TESTBUFLEN_CMP];
+	char buf4[TESTBUFLEN_CMP];
+	const char *test_name = "test_2compare_eq index";
+
+	for (i = 0; i < TESTBUFLEN_CMP; i++)
+		buf1[i] = (char)i;
+	memset(buf2, 0, TESTBUFLEN_CMP);
+	memset(buf3, 0, TESTBUFLEN_CMP);
+	memset(buf4, 0, TESTBUFLEN_CMP);
+
+	/* First compare failure is op[0], expect 1. */
+	ret = test_2compare_eq_op(buf2, buf1, buf4, buf3, TESTBUFLEN_CMP);
+	if (ret < 0) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (ret != 1) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, ret, 1);
+		return -1;
+	}
+
+	/* All compares succeed. */
+	for (i = 0; i < TESTBUFLEN_CMP; i++)
+		buf2[i] = (char)i;
+	ret = test_2compare_eq_op(buf2, buf1, buf4, buf3, TESTBUFLEN_CMP);
+	if (ret < 0) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (ret != 0) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, ret, 0);
+		return -1;
+	}
+
+	/* First compare failure is op[1], expect 2. */
+	for (i = 0; i < TESTBUFLEN_CMP; i++)
+		buf3[i] = (char)i;
+	ret = test_2compare_eq_op(buf2, buf1, buf4, buf3, TESTBUFLEN_CMP);
+	if (ret < 0) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (ret != 2) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, ret, 2);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_2compare_ne_op(char *a, char *b, char *c, char *d,
+		size_t len)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_COMPARE_NE_OP,
+			.len = len,
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.compare_op.a, a),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.compare_op.b, b),
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+		[1] = {
+			.op = CPU_COMPARE_NE_OP,
+			.len = len,
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.compare_op.a, c),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.compare_op.b, d),
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+	};
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_2compare_ne_index(void)
+{
+	int i, ret;
+	char buf1[TESTBUFLEN_CMP];
+	char buf2[TESTBUFLEN_CMP];
+	char buf3[TESTBUFLEN_CMP];
+	char buf4[TESTBUFLEN_CMP];
+	const char *test_name = "test_2compare_ne index";
+
+	memset(buf1, 0, TESTBUFLEN_CMP);
+	memset(buf2, 0, TESTBUFLEN_CMP);
+	memset(buf3, 0, TESTBUFLEN_CMP);
+	memset(buf4, 0, TESTBUFLEN_CMP);
+
+	/* First compare ne failure is op[0], expect 1. */
+	ret = test_2compare_ne_op(buf2, buf1, buf4, buf3, TESTBUFLEN_CMP);
+	if (ret < 0) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (ret != 1) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, ret, 1);
+		return -1;
+	}
+
+	/* All compare ne succeed. */
+	for (i = 0; i < TESTBUFLEN_CMP; i++)
+		buf1[i] = (char)i;
+	for (i = 0; i < TESTBUFLEN_CMP; i++)
+		buf3[i] = (char)i;
+	ret = test_2compare_ne_op(buf2, buf1, buf4, buf3, TESTBUFLEN_CMP);
+	if (ret < 0) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (ret != 0) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, ret, 0);
+		return -1;
+	}
+
+	/* First compare failure is op[1], expect 2. */
+	for (i = 0; i < TESTBUFLEN_CMP; i++)
+		buf4[i] = (char)i;
+	ret = test_2compare_ne_op(buf2, buf1, buf4, buf3, TESTBUFLEN_CMP);
+	if (ret < 0) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (ret != 2) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, ret, 2);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_memcpy_op(void *dst, void *src, size_t len)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_MEMCPY_OP,
+			.len = len,
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.memcpy_op.dst, dst),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.memcpy_op.src, src),
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+	};
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_memcpy(void)
+{
+	int i, ret;
+	char buf1[TESTBUFLEN];
+	char buf2[TESTBUFLEN];
+	const char *test_name = "test_memcpy";
+
+	/* Test memcpy */
+	for (i = 0; i < TESTBUFLEN; i++)
+		buf1[i] = (char)i;
+	memset(buf2, 0, TESTBUFLEN);
+	ret = test_memcpy_op(buf2, buf1, TESTBUFLEN);
+	if (ret) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	for (i = 0; i < TESTBUFLEN; i++) {
+		if (buf2[i] != (char)i) {
+			ksft_exit_fail_msg("%s test: unexpected value at offset %d. Found %d. Should be %d.\n",
+				   test_name, i, buf2[i], (char)i);
+			return -1;
+		}
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_memcpy_u32(void)
+{
+	int ret;
+	uint32_t v1, v2;
+	const char *test_name = "test_memcpy_u32";
+
+	/* Test memcpy_u32 */
+	v1 = 42;
+	v2 = 0;
+	ret = test_memcpy_op(&v2, &v1, sizeof(v1));
+	if (ret) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (v1 != v2) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, v2, v1);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_memcpy_mb_memcpy_op(void *dst1, void *src1,
+		void *dst2, void *src2, size_t len)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_MEMCPY_OP,
+			.len = len,
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.memcpy_op.dst, dst1),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.memcpy_op.src, src1),
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+		[1] = {
+			.op = CPU_MB_OP,
+		},
+		[2] = {
+			.op = CPU_MEMCPY_OP,
+			.len = len,
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.memcpy_op.dst, dst2),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.memcpy_op.src, src2),
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+	};
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_memcpy_mb_memcpy(void)
+{
+	int ret;
+	int v1, v2, v3;
+	const char *test_name = "test_memcpy_mb_memcpy";
+
+	/* Test memcpy */
+	v1 = 42;
+	v2 = v3 = 0;
+	ret = test_memcpy_mb_memcpy_op(&v2, &v1, &v3, &v2, sizeof(int));
+	if (ret) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (v3 != v1) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, v3, v1);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_add_op(int *v, int64_t increment)
+{
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_op_add(v, increment, sizeof(*v), cpu);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_add(void)
+{
+	int orig_v = 42, v, ret;
+	int increment = 1;
+	const char *test_name = "test_add";
+
+	v = orig_v;
+	ret = test_add_op(&v, increment);
+	if (ret) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (v != orig_v + increment) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, v,
+				   orig_v + increment);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_two_add_op(int *v, int64_t *increments)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_ADD_OP,
+			.len = sizeof(*v),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(
+				.u.arithmetic_op.p, v),
+			.u.arithmetic_op.count = increments[0],
+			.u.arithmetic_op.expect_fault_p = 0,
+		},
+		[1] = {
+			.op = CPU_ADD_OP,
+			.len = sizeof(*v),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(
+				.u.arithmetic_op.p, v),
+			.u.arithmetic_op.count = increments[1],
+			.u.arithmetic_op.expect_fault_p = 0,
+		},
+	};
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_two_add(void)
+{
+	int orig_v = 42, v, ret;
+	int64_t increments[2] = { 99, 123 };
+	const char *test_name = "test_two_add";
+
+	v = orig_v;
+	ret = test_two_add_op(&v, increments);
+	if (ret) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (v != orig_v + increments[0] + increments[1]) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, v,
+				   orig_v + increments[0] + increments[1]);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_or_op(int *v, uint64_t mask)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_OR_OP,
+			.len = sizeof(*v),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(
+				.u.bitwise_op.p, v),
+			.u.bitwise_op.mask = mask,
+			.u.bitwise_op.expect_fault_p = 0,
+		},
+	};
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_or(void)
+{
+	int orig_v = 0xFF00000, v, ret;
+	uint32_t mask = 0xFFF;
+	const char *test_name = "test_or";
+
+	v = orig_v;
+	ret = test_or_op(&v, mask);
+	if (ret) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (v != (orig_v | mask)) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, v, orig_v | mask);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_and_op(int *v, uint64_t mask)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_AND_OP,
+			.len = sizeof(*v),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(
+				.u.bitwise_op.p, v),
+			.u.bitwise_op.mask = mask,
+			.u.bitwise_op.expect_fault_p = 0,
+		},
+	};
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_and(void)
+{
+	int orig_v = 0xF00, v, ret;
+	uint32_t mask = 0xFFF;
+	const char *test_name = "test_and";
+
+	v = orig_v;
+	ret = test_and_op(&v, mask);
+	if (ret) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (v != (orig_v & mask)) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, v, orig_v & mask);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_xor_op(int *v, uint64_t mask)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_XOR_OP,
+			.len = sizeof(*v),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(
+				.u.bitwise_op.p, v),
+			.u.bitwise_op.mask = mask,
+			.u.bitwise_op.expect_fault_p = 0,
+		},
+	};
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_xor(void)
+{
+	int orig_v = 0xF00, v, ret;
+	uint32_t mask = 0xFFF;
+	const char *test_name = "test_xor";
+
+	v = orig_v;
+	ret = test_xor_op(&v, mask);
+	if (ret) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (v != (orig_v ^ mask)) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, v, orig_v ^ mask);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_lshift_op(int *v, uint32_t bits)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_LSHIFT_OP,
+			.len = sizeof(*v),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(
+				.u.shift_op.p, v),
+			.u.shift_op.bits = bits,
+			.u.shift_op.expect_fault_p = 0,
+		},
+	};
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_lshift(void)
+{
+	int orig_v = 0xF00, v, ret;
+	uint32_t bits = 5;
+	const char *test_name = "test_lshift";
+
+	v = orig_v;
+	ret = test_lshift_op(&v, bits);
+	if (ret) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (v != (orig_v << bits)) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, v, orig_v << bits);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_rshift_op(int *v, uint32_t bits)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_RSHIFT_OP,
+			.len = sizeof(*v),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(
+				.u.shift_op.p, v),
+			.u.shift_op.bits = bits,
+			.u.shift_op.expect_fault_p = 0,
+		},
+	};
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_rshift(void)
+{
+	int orig_v = 0xF00, v, ret;
+	uint32_t bits = 5;
+	const char *test_name = "test_rshift";
+
+	v = orig_v;
+	ret = test_rshift_op(&v, bits);
+	if (ret) {
+		ksft_exit_fail_msg("%s test: returned with %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (v != (orig_v >> bits)) {
+		ksft_exit_fail_msg("%s test: unexpected value %d. Should be %d.\n",
+				   test_name, v, orig_v >> bits);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_cmpxchg_op(void *v, void *expect, void *old, void *n,
+		size_t len)
+{
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_op_cmpxchg(v, expect, old, n, len, cpu);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_cmpxchg_success(void)
+{
+	int ret;
+	uint64_t orig_v = 1, v, expect = 1, old = 0, n = 3;
+	const char *test_name = "test_cmpxchg success";
+
+	v = orig_v;
+	ret = test_cmpxchg_op(&v, &expect, &old, &n, sizeof(uint64_t));
+	if (ret < 0) {
+		ksft_exit_fail_msg("%s test: ret = %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (ret) {
+		ksft_exit_fail_msg("%s returned %d, expecting %d\n",
+				   test_name, ret, 0);
+		return -1;
+	}
+	if (v != n) {
+		ksft_exit_fail_msg("%s v is %lld, expecting %lld\n",
+				   test_name, (long long)v, (long long)n);
+		return -1;
+	}
+	if (old != orig_v) {
+		ksft_exit_fail_msg("%s old is %lld, expecting %lld\n",
+				   test_name, (long long)old,
+				   (long long)orig_v);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_cmpxchg_fail(void)
+{
+	int ret;
+	uint64_t orig_v = 1, v, expect = 123, old = 0, n = 3;
+	const char *test_name = "test_cmpxchg fail";
+
+	v = orig_v;
+	ret = test_cmpxchg_op(&v, &expect, &old, &n, sizeof(uint64_t));
+	if (ret < 0) {
+		ksft_exit_fail_msg("%s test: ret = %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	if (ret == 0) {
+		ksft_exit_fail_msg("%s returned %d, expecting %d\n",
+				   test_name, ret, 1);
+		return -1;
+	}
+	if (v == n) {
+		ksft_exit_fail_msg("%s returned %lld, expecting %lld\n",
+				   test_name, (long long)v, (long long)orig_v);
+		return -1;
+	}
+	if (old != orig_v) {
+		ksft_exit_fail_msg("%s old is %lld, expecting %lld\n",
+				   test_name, (long long)old,
+				   (long long)orig_v);
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_memcpy_expect_fault_op(void *dst, void *src, size_t len)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_MEMCPY_OP,
+			.len = len,
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.memcpy_op.dst, dst),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.memcpy_op.src, src),
+			.u.memcpy_op.expect_fault_dst = 0,
+			/* Return EAGAIN on fault. */
+			.u.memcpy_op.expect_fault_src = 1,
+		},
+	};
+	int cpu;
+
+	cpu = cpu_op_get_current_cpu();
+	return cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+}
+
+static int test_memcpy_fault(void)
+{
+	int ret;
+	char buf1[TESTBUFLEN];
+	const char *test_name = "test_memcpy_fault";
+
+	/* Test memcpy */
+	ret = test_memcpy_op(buf1, NULL, TESTBUFLEN);
+	if (!ret || (ret < 0 && errno != EFAULT)) {
+		ksft_exit_fail_msg("%s test: ret = %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	/* Test memcpy expect fault */
+	ret = test_memcpy_expect_fault_op(buf1, NULL, TESTBUFLEN);
+	if (!ret || (ret < 0 && errno != EAGAIN)) {
+		ksft_exit_fail_msg("%s test: ret = %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int do_test_unknown_op(void)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = -1,	/* Unknown */
+			.len = 0,
+		},
+	};
+	int cpu;
+
+	cpu = cpu_op_get_current_cpu();
+	return cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+}
+
+static int test_unknown_op(void)
+{
+	int ret;
+	const char *test_name = "test_unknown_op";
+
+	ret = do_test_unknown_op();
+	if (!ret || (ret < 0 && errno != EINVAL)) {
+		ksft_exit_fail_msg("%s test: ret = %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int do_test_max_ops(void)
+{
+	struct cpu_op opvec[] = {
+		[0] = { .op = CPU_MB_OP, },
+		[1] = { .op = CPU_MB_OP, },
+		[2] = { .op = CPU_MB_OP, },
+		[3] = { .op = CPU_MB_OP, },
+		[4] = { .op = CPU_MB_OP, },
+		[5] = { .op = CPU_MB_OP, },
+		[6] = { .op = CPU_MB_OP, },
+		[7] = { .op = CPU_MB_OP, },
+		[8] = { .op = CPU_MB_OP, },
+		[9] = { .op = CPU_MB_OP, },
+		[10] = { .op = CPU_MB_OP, },
+		[11] = { .op = CPU_MB_OP, },
+		[12] = { .op = CPU_MB_OP, },
+		[13] = { .op = CPU_MB_OP, },
+		[14] = { .op = CPU_MB_OP, },
+		[15] = { .op = CPU_MB_OP, },
+	};
+	int cpu;
+
+	cpu = cpu_op_get_current_cpu();
+	return cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+}
+
+static int test_max_ops(void)
+{
+	int ret;
+	const char *test_name = "test_max_ops";
+
+	ret = do_test_max_ops();
+	if (ret < 0) {
+		ksft_exit_fail_msg("%s test: ret = %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int do_test_too_many_ops(void)
+{
+	struct cpu_op opvec[] = {
+		[0] = { .op = CPU_MB_OP, },
+		[1] = { .op = CPU_MB_OP, },
+		[2] = { .op = CPU_MB_OP, },
+		[3] = { .op = CPU_MB_OP, },
+		[4] = { .op = CPU_MB_OP, },
+		[5] = { .op = CPU_MB_OP, },
+		[6] = { .op = CPU_MB_OP, },
+		[7] = { .op = CPU_MB_OP, },
+		[8] = { .op = CPU_MB_OP, },
+		[9] = { .op = CPU_MB_OP, },
+		[10] = { .op = CPU_MB_OP, },
+		[11] = { .op = CPU_MB_OP, },
+		[12] = { .op = CPU_MB_OP, },
+		[13] = { .op = CPU_MB_OP, },
+		[14] = { .op = CPU_MB_OP, },
+		[15] = { .op = CPU_MB_OP, },
+		[16] = { .op = CPU_MB_OP, },
+	};
+	int cpu;
+
+	cpu = cpu_op_get_current_cpu();
+	return cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+}
+
+static int test_too_many_ops(void)
+{
+	int ret;
+	const char *test_name = "test_too_many_ops";
+
+	ret = do_test_too_many_ops();
+	if (!ret || (ret < 0 && errno != EINVAL)) {
+		ksft_exit_fail_msg("%s test: ret = %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+/* Use 64kB len, largest page size known on Linux. */
+static int test_memcpy_single_too_large(void)
+{
+	int i, ret;
+	char buf1[TESTBUFLEN_PAGE_MAX + 1];
+	char buf2[TESTBUFLEN_PAGE_MAX + 1];
+	const char *test_name = "test_memcpy_single_too_large";
+
+	/* Test memcpy */
+	for (i = 0; i < TESTBUFLEN_PAGE_MAX + 1; i++)
+		buf1[i] = (char)i;
+	memset(buf2, 0, TESTBUFLEN_PAGE_MAX + 1);
+	ret = test_memcpy_op(buf2, buf1, TESTBUFLEN_PAGE_MAX + 1);
+	if (!ret || (ret < 0 && errno != EINVAL)) {
+		ksft_exit_fail_msg("%s test: ret = %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+static int test_memcpy_single_ok_sum_too_large_op(void *dst, void *src,
+						  size_t len)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_MEMCPY_OP,
+			.len = len,
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.memcpy_op.dst, dst),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.memcpy_op.src, src),
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+		[1] = {
+			.op = CPU_MEMCPY_OP,
+			.len = len,
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.memcpy_op.dst, dst),
+			LINUX_FIELD_u32_u64_INIT_ONSTACK(.u.memcpy_op.src, src),
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+	};
+	int ret, cpu;
+
+	do {
+		cpu = cpu_op_get_current_cpu();
+		ret = cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+	} while (ret == -1 && errno == EAGAIN);
+
+	return ret;
+}
+
+static int test_memcpy_single_ok_sum_too_large(void)
+{
+	int i, ret;
+	char buf1[TESTBUFLEN];
+	char buf2[TESTBUFLEN];
+	const char *test_name = "test_memcpy_single_ok_sum_too_large";
+
+	/* Test memcpy */
+	for (i = 0; i < TESTBUFLEN; i++)
+		buf1[i] = (char)i;
+	memset(buf2, 0, TESTBUFLEN);
+	ret = test_memcpy_single_ok_sum_too_large_op(buf2, buf1, TESTBUFLEN);
+	if (!ret || (ret < 0 && errno != EINVAL)) {
+		ksft_exit_fail_msg("%s test: ret = %d, errno = %s\n",
+				   test_name, ret, strerror(errno));
+		return -1;
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+/*
+ * Iterate over large uninitialized arrays to trigger page faults.
+ * This includes reading from zero pages.
+ */
+int test_page_fault(void)
+{
+	int ret = 0;
+	uint64_t i;
+	const char *test_name = "test_page_fault";
+
+	for (i = 0; i < NR_PF_ARRAY; i++) {
+		ret = test_memcpy_op(pf_array_dst[i],
+				     pf_array_src[i],
+				     PF_ARRAY_LEN);
+		if (ret) {
+			ksft_exit_fail_msg("%s test: ret = %d, errno = %s\n",
+					   test_name, ret, strerror(errno));
+			return ret;
+		}
+	}
+	ksft_test_result_pass("%s test\n", test_name);
+	return 0;
+}
+
+/*
+ * Try to use 2MB huge pages.
+ */
+int test_hugetlb(void)
+{
+	int ret = 0;
+	uint64_t i;
+	const char *test_name = "test_hugetlb";
+	int *dst, *src;
+
+	dst = mmap(NULL, HUGEMAPLEN, PROT_READ | PROT_WRITE,
+		   MAP_HUGETLB | MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
+	if (dst == MAP_FAILED) {
+		switch (errno) {
+		case ENOMEM:
+		case ENOENT:
+		case EINVAL:
+			ksft_test_result_skip("%s test.\n", test_name);
+			goto end;
+		default:
+			break;
+		}
+		perror("mmap");
+		abort();
+	}
+	src = mmap(NULL, HUGEMAPLEN, PROT_READ | PROT_WRITE,
+		   MAP_HUGETLB | MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
+	if (src == MAP_FAILED) {
+		if (errno == ENOMEM) {
+			ksft_test_result_skip("%s test.\n", test_name);
+			goto unmap_dst;
+		}
+		perror("mmap");
+		abort();
+	}
+
+	/* Read/write from/to huge zero pages. */
+	for (i = 0; i < NR_HUGE_ARRAY; i++) {
+		ret = test_memcpy_op(dst + (i * PF_ARRAY_LEN / sizeof(int)),
+				     src + (i * PF_ARRAY_LEN / sizeof(int)),
+				     PF_ARRAY_LEN);
+		if (ret) {
+			ksft_exit_fail_msg("%s test: ret = %d, errno = %s\n",
+					   test_name, ret, strerror(errno));
+			return ret;
+		}
+	}
+	for (i = 0; i < NR_HUGE_ARRAY * (PF_ARRAY_LEN / sizeof(int)); i++)
+		src[i] = i;
+
+	for (i = 0; i < NR_HUGE_ARRAY; i++) {
+		ret = test_memcpy_op(dst + (i * PF_ARRAY_LEN / sizeof(int)),
+				     src + (i * PF_ARRAY_LEN / sizeof(int)),
+				     PF_ARRAY_LEN);
+		if (ret) {
+			ksft_exit_fail_msg("%s test: ret = %d, errno = %s\n",
+					   test_name, ret, strerror(errno));
+			return ret;
+		}
+	}
+
+	for (i = 0; i < NR_HUGE_ARRAY * (PF_ARRAY_LEN / sizeof(int)); i++) {
+		if (dst[i] != i) {
+			ksft_exit_fail_msg("%s mismatch, expect %d, got %d\n",
+					   test_name, i, dst[i]);
+			return ret;
+		}
+	}
+
+	ksft_test_result_pass("%s test\n", test_name);
+
+	if (munmap(src, HUGEMAPLEN)) {
+		perror("munmap");
+		abort();
+	}
+unmap_dst:
+	if (munmap(dst, HUGEMAPLEN)) {
+		perror("munmap");
+		abort();
+	}
+end:
+	return 0;
+}
+
+int main(int argc, char **argv)
+{
+	ksft_print_header();
+
+	test_compare_eq_same();
+	test_compare_eq_diff();
+	test_compare_ne_same();
+	test_compare_ne_diff();
+	test_2compare_eq_index();
+	test_2compare_ne_index();
+	test_memcpy();
+	test_memcpy_u32();
+	test_memcpy_mb_memcpy();
+	test_add();
+	test_two_add();
+	test_or();
+	test_and();
+	test_xor();
+	test_lshift();
+	test_rshift();
+	test_cmpxchg_success();
+	test_cmpxchg_fail();
+	test_memcpy_fault();
+	test_unknown_op();
+	test_max_ops();
+	test_too_many_ops();
+	test_memcpy_single_too_large();
+	test_memcpy_single_ok_sum_too_large();
+	test_page_fault();
+	test_hugetlb();
+
+	return ksft_exit_pass();
+}
diff --git a/tools/testing/selftests/cpu-opv/cpu-op.c b/tools/testing/selftests/cpu-opv/cpu-op.c
new file mode 100644
index 000000000000..5981895df25a
--- /dev/null
+++ b/tools/testing/selftests/cpu-opv/cpu-op.c
@@ -0,0 +1,352 @@
+/*
+ * cpu-op.c
+ *
+ * Copyright (C) 2017 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; only
+ * version 2.1 of the License.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ */
+
+#define _GNU_SOURCE
+#include <errno.h>
+#include <sched.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <syscall.h>
+#include <assert.h>
+#include <signal.h>
+
+#include "cpu-op.h"
+
+#define ARRAY_SIZE(arr)	(sizeof(arr) / sizeof((arr)[0]))
+
+#define ACCESS_ONCE(x)		(*(__volatile__  __typeof__(x) *)&(x))
+#define WRITE_ONCE(x, v)	__extension__ ({ ACCESS_ONCE(x) = (v); })
+#define READ_ONCE(x)		ACCESS_ONCE(x)
+
+int cpu_opv(struct cpu_op *cpu_opv, int cpuopcnt, int cpu, int flags)
+{
+	return syscall(__NR_cpu_opv, cpu_opv, cpuopcnt, cpu, flags);
+}
+
+int cpu_op_get_current_cpu(void)
+{
+	int cpu;
+
+	cpu = sched_getcpu();
+	if (cpu < 0) {
+		perror("sched_getcpu()");
+		abort();
+	}
+	return cpu;
+}
+
+int cpu_op_cmpxchg(void *v, void *expect, void *old, void *n, size_t len,
+		   int cpu)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_MEMCPY_OP,
+			.len = len,
+			.u.memcpy_op.dst = (unsigned long)old,
+			.u.memcpy_op.src = (unsigned long)v,
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+		[1] = {
+			.op = CPU_COMPARE_EQ_OP,
+			.len = len,
+			.u.compare_op.a = (unsigned long)v,
+			.u.compare_op.b = (unsigned long)expect,
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+		[2] = {
+			.op = CPU_MEMCPY_OP,
+			.len = len,
+			.u.memcpy_op.dst = (unsigned long)v,
+			.u.memcpy_op.src = (unsigned long)n,
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+	};
+
+	return cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+}
+
+int cpu_op_add(void *v, int64_t count, size_t len, int cpu)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_ADD_OP,
+			.len = len,
+			.u.arithmetic_op.p = (unsigned long)v,
+			.u.arithmetic_op.count = count,
+			.u.arithmetic_op.expect_fault_p = 0,
+		},
+	};
+
+	return cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+}
+
+int cpu_op_cmpeqv_storev(intptr_t *v, intptr_t expect, intptr_t newv,
+			 int cpu)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_COMPARE_EQ_OP,
+			.len = sizeof(intptr_t),
+			.u.compare_op.a = (unsigned long)v,
+			.u.compare_op.b = (unsigned long)&expect,
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+		[1] = {
+			.op = CPU_MEMCPY_OP,
+			.len = sizeof(intptr_t),
+			.u.memcpy_op.dst = (unsigned long)v,
+			.u.memcpy_op.src = (unsigned long)&newv,
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+	};
+
+	return cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+}
+
+static int cpu_op_cmpeqv_storep_expect_fault(intptr_t *v, intptr_t expect,
+					     intptr_t *newp, int cpu)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_COMPARE_EQ_OP,
+			.len = sizeof(intptr_t),
+			.u.compare_op.a = (unsigned long)v,
+			.u.compare_op.b = (unsigned long)&expect,
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+		[1] = {
+			.op = CPU_MEMCPY_OP,
+			.len = sizeof(intptr_t),
+			.u.memcpy_op.dst = (unsigned long)v,
+			.u.memcpy_op.src = (unsigned long)newp,
+			.u.memcpy_op.expect_fault_dst = 0,
+			/* Return EAGAIN on src fault. */
+			.u.memcpy_op.expect_fault_src = 1,
+		},
+	};
+
+	return cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+}
+
+int cpu_op_cmpnev_storeoffp_load(intptr_t *v, intptr_t expectnot,
+				 off_t voffp, intptr_t *load, int cpu)
+{
+	int ret;
+
+	do {
+		intptr_t oldv = READ_ONCE(*v);
+		intptr_t *newp = (intptr_t *)(oldv + voffp);
+
+		if (oldv == expectnot)
+			return 1;
+		ret = cpu_op_cmpeqv_storep_expect_fault(v, oldv, newp, cpu);
+		if (!ret) {
+			*load = oldv;
+			return 0;
+		}
+	} while (ret > 0);
+
+	return -1;
+}
+
+int cpu_op_cmpeqv_storev_storev(intptr_t *v, intptr_t expect,
+				intptr_t *v2, intptr_t newv2,
+				intptr_t newv, int cpu)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_COMPARE_EQ_OP,
+			.len = sizeof(intptr_t),
+			.u.compare_op.a = (unsigned long)v,
+			.u.compare_op.b = (unsigned long)&expect,
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+		[1] = {
+			.op = CPU_MEMCPY_OP,
+			.len = sizeof(intptr_t),
+			.u.memcpy_op.dst = (unsigned long)v2,
+			.u.memcpy_op.src = (unsigned long)&newv2,
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+		[2] = {
+			.op = CPU_MEMCPY_OP,
+			.len = sizeof(intptr_t),
+			.u.memcpy_op.dst = (unsigned long)v,
+			.u.memcpy_op.src = (unsigned long)&newv,
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+	};
+
+	return cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+}
+
+int cpu_op_cmpeqv_storev_mb_storev(intptr_t *v, intptr_t expect,
+				   intptr_t *v2, intptr_t newv2,
+				   intptr_t newv, int cpu)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_COMPARE_EQ_OP,
+			.len = sizeof(intptr_t),
+			.u.compare_op.a = (unsigned long)v,
+			.u.compare_op.b = (unsigned long)&expect,
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+		[1] = {
+			.op = CPU_MEMCPY_OP,
+			.len = sizeof(intptr_t),
+			.u.memcpy_op.dst = (unsigned long)v2,
+			.u.memcpy_op.src = (unsigned long)&newv2,
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+		[2] = {
+			.op = CPU_MB_OP,
+		},
+		[3] = {
+			.op = CPU_MEMCPY_OP,
+			.len = sizeof(intptr_t),
+			.u.memcpy_op.dst = (unsigned long)v,
+			.u.memcpy_op.src = (unsigned long)&newv,
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+	};
+
+	return cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+}
+
+int cpu_op_cmpeqv_cmpeqv_storev(intptr_t *v, intptr_t expect,
+				intptr_t *v2, intptr_t expect2,
+				intptr_t newv, int cpu)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_COMPARE_EQ_OP,
+			.len = sizeof(intptr_t),
+			.u.compare_op.a = (unsigned long)v,
+			.u.compare_op.b = (unsigned long)&expect,
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+		[1] = {
+			.op = CPU_COMPARE_EQ_OP,
+			.len = sizeof(intptr_t),
+			.u.compare_op.a = (unsigned long)v2,
+			.u.compare_op.b = (unsigned long)&expect2,
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+		[2] = {
+			.op = CPU_MEMCPY_OP,
+			.len = sizeof(intptr_t),
+			.u.memcpy_op.dst = (unsigned long)v,
+			.u.memcpy_op.src = (unsigned long)&newv,
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+	};
+
+	return cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+}
+
+int cpu_op_cmpeqv_memcpy_storev(intptr_t *v, intptr_t expect,
+				void *dst, void *src, size_t len,
+				intptr_t newv, int cpu)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_COMPARE_EQ_OP,
+			.len = sizeof(intptr_t),
+			.u.compare_op.a = (unsigned long)v,
+			.u.compare_op.b = (unsigned long)&expect,
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+		[1] = {
+			.op = CPU_MEMCPY_OP,
+			.len = len,
+			.u.memcpy_op.dst = (unsigned long)dst,
+			.u.memcpy_op.src = (unsigned long)src,
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+		[2] = {
+			.op = CPU_MEMCPY_OP,
+			.len = sizeof(intptr_t),
+			.u.memcpy_op.dst = (unsigned long)v,
+			.u.memcpy_op.src = (unsigned long)&newv,
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+	};
+
+	return cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+}
+
+int cpu_op_cmpeqv_memcpy_mb_storev(intptr_t *v, intptr_t expect,
+				   void *dst, void *src, size_t len,
+				   intptr_t newv, int cpu)
+{
+	struct cpu_op opvec[] = {
+		[0] = {
+			.op = CPU_COMPARE_EQ_OP,
+			.len = sizeof(intptr_t),
+			.u.compare_op.a = (unsigned long)v,
+			.u.compare_op.b = (unsigned long)&expect,
+			.u.compare_op.expect_fault_a = 0,
+			.u.compare_op.expect_fault_b = 0,
+		},
+		[1] = {
+			.op = CPU_MEMCPY_OP,
+			.len = len,
+			.u.memcpy_op.dst = (unsigned long)dst,
+			.u.memcpy_op.src = (unsigned long)src,
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+		[2] = {
+			.op = CPU_MB_OP,
+		},
+		[3] = {
+			.op = CPU_MEMCPY_OP,
+			.len = sizeof(intptr_t),
+			.u.memcpy_op.dst = (unsigned long)v,
+			.u.memcpy_op.src = (unsigned long)&newv,
+			.u.memcpy_op.expect_fault_dst = 0,
+			.u.memcpy_op.expect_fault_src = 0,
+		},
+	};
+
+	return cpu_opv(opvec, ARRAY_SIZE(opvec), cpu, 0);
+}
+
+int cpu_op_addv(intptr_t *v, int64_t count, int cpu)
+{
+	return cpu_op_add(v, count, sizeof(intptr_t), cpu);
+}
diff --git a/tools/testing/selftests/cpu-opv/cpu-op.h b/tools/testing/selftests/cpu-opv/cpu-op.h
new file mode 100644
index 000000000000..762a38d6e0d0
--- /dev/null
+++ b/tools/testing/selftests/cpu-opv/cpu-op.h
@@ -0,0 +1,59 @@
+/*
+ * cpu-op.h
+ *
+ * (C) Copyright 2017 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef CPU_OPV_H
+#define CPU_OPV_H
+
+#include <stdlib.h>
+#include <stdint.h>
+#include <linux/cpu_opv.h>
+
+int cpu_opv(struct cpu_op *cpuopv, int cpuopcnt, int cpu, int flags);
+int cpu_op_get_current_cpu(void);
+
+int cpu_op_cmpxchg(void *v, void *expect, void *old, void *_new, size_t len,
+		   int cpu);
+int cpu_op_add(void *v, int64_t count, size_t len, int cpu);
+
+int cpu_op_cmpeqv_storev(intptr_t *v, intptr_t expect, intptr_t newv, int cpu);
+int cpu_op_cmpnev_storeoffp_load(intptr_t *v, intptr_t expectnot,
+				 off_t voffp, intptr_t *load, int cpu);
+int cpu_op_cmpeqv_storev_storev(intptr_t *v, intptr_t expect,
+				intptr_t *v2, intptr_t newv2,
+				intptr_t newv, int cpu);
+int cpu_op_cmpeqv_storev_mb_storev(intptr_t *v, intptr_t expect,
+				   intptr_t *v2, intptr_t newv2,
+				   intptr_t newv, int cpu);
+int cpu_op_cmpeqv_cmpeqv_storev(intptr_t *v, intptr_t expect,
+				intptr_t *v2, intptr_t expect2,
+				intptr_t newv, int cpu);
+int cpu_op_cmpeqv_memcpy_storev(intptr_t *v, intptr_t expect,
+				void *dst, void *src, size_t len,
+				intptr_t newv, int cpu);
+int cpu_op_cmpeqv_memcpy_mb_storev(intptr_t *v, intptr_t expect,
+				   void *dst, void *src, size_t len,
+				   intptr_t newv, int cpu);
+int cpu_op_addv(intptr_t *v, int64_t count, int cpu);
+
+#endif  /* CPU_OPV_H_ */
-- 
2.11.0

[RFC PATCH for 4.16 16/21] rseq: selftests: Provide rseq library (v5)

[Mathieu Desnoyers]

This rseq helper library provides a user-space API to the rseq()
system call.

The rseq fast-path exposes the instruction pointer addresses where the
rseq assembly blocks begin and end, as well as the associated abort
instruction pointer, in the __rseq_table section. This section allows
debuggers may know where to place breakpoints when single-stepping
through assembly blocks which may be aborted at any point by the kernel.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Shuah Khan <shuahkh@osg.samsung.com>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Paul Turner <pjt@google.com>
CC: Andrew Hunter <ahh@google.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-kselftest@vger.kernel.org
CC: linux-api@vger.kernel.org
---
Changes since v1:
- Provide abort-ip signature: The abort-ip signature is located just
  before the abort-ip target. It is currently hardcoded, but a
  user-space application could use the __rseq_table to iterate on all
  abort-ip targets and use a random value as signature if needed in the
  future.
- Add rseq_prepare_unload(): Libraries and JIT code using rseq critical
  sections need to issue rseq_prepare_unload() on each thread at least
  once before reclaim of struct rseq_cs.
- Use initial-exec TLS model, non-weak symbol: The initial-exec model is
  signal-safe, whereas the global-dynamic model is not.  Remove the
  "weak" symbol attribute from the __rseq_abi in rseq.c. The rseq.so
  library will have ownership of that symbol, and there is not reason for
  an application or user library to try to define that symbol.
  The expected use is to link against libreq.so, which owns and provide
  that symbol.
- Set cpu_id to -2 on register error
- Add rseq_len syscall parameter, rseq_cs version
- Ensure disassember-friendly signature: x86 32/64 disassembler have a
  hard time decoding the instruction stream after a bad instruction. Use
  a nopl instruction to encode the signature. Suggested by Andy Lutomirski.
- Exercise parametrized tests variants in a shell scripts.
- Restartable sequences selftests: Remove use of event counter.
- Use cpu_id_start field:  With the cpu_id_start field, the C
  preparation phase of the fast-path does not need to compare cpu_id < 0
  anymore.
- Signal-safe registration and refcounting: Allow libraries using
  librseq.so to register it from signal handlers.
- Use OVERRIDE_TARGETS in makefile.
- Use "m" constraints for rseq_cs field.

Changes since v2:
- Update based on Thomas Gleixner's comments.

Changes since v3:
- Generate param_test_skip_fastpath and param_test_benchmark with
  -DSKIP_FASTPATH and -DBENCHMARK (respectively). Add param_test_fastpath
  to run_param_test.sh.

Changes since v4:
- Fold arm: workaround gcc asm size guess,
- Namespace barrier() -> rseq_barrier() in library header,
- Take into account coding style feedback from Peter Zijlstra,
- Split rseq selftests into logical commits.
---
 tools/testing/selftests/rseq/rseq-arm.h  |  732 +++++++++++++++++++
 tools/testing/selftests/rseq/rseq-ppc.h  |  688 ++++++++++++++++++
 tools/testing/selftests/rseq/rseq-skip.h |   82 +++
 tools/testing/selftests/rseq/rseq-x86.h  | 1149 ++++++++++++++++++++++++++++++
 tools/testing/selftests/rseq/rseq.c      |  116 +++
 tools/testing/selftests/rseq/rseq.h      |  164 +++++
 6 files changed, 2931 insertions(+)
 create mode 100644 tools/testing/selftests/rseq/rseq-arm.h
 create mode 100644 tools/testing/selftests/rseq/rseq-ppc.h
 create mode 100644 tools/testing/selftests/rseq/rseq-skip.h
 create mode 100644 tools/testing/selftests/rseq/rseq-x86.h
 create mode 100644 tools/testing/selftests/rseq/rseq.c
 create mode 100644 tools/testing/selftests/rseq/rseq.h

diff --git a/tools/testing/selftests/rseq/rseq-arm.h b/tools/testing/selftests/rseq/rseq-arm.h
new file mode 100644
index 000000000000..adcaa6cbbd01
--- /dev/null
+++ b/tools/testing/selftests/rseq/rseq-arm.h
@@ -0,0 +1,732 @@
+/*
+ * rseq-arm.h
+ *
+ * (C) Copyright 2016 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#define RSEQ_SIG	0x53053053
+
+#define rseq_smp_mb()	__asm__ __volatile__ ("dmb" ::: "memory", "cc")
+#define rseq_smp_rmb()	__asm__ __volatile__ ("dmb" ::: "memory", "cc")
+#define rseq_smp_wmb()	__asm__ __volatile__ ("dmb" ::: "memory", "cc")
+
+#define rseq_smp_load_acquire(p)					\
+__extension__ ({							\
+	__typeof(*p) ____p1 = RSEQ_READ_ONCE(*p);			\
+	rseq_smp_mb();							\
+	____p1;								\
+})
+
+#define rseq_smp_acquire__after_ctrl_dep()	rseq_smp_rmb()
+
+#define rseq_smp_store_release(p, v)					\
+do {									\
+	rseq_smp_mb();							\
+	RSEQ_WRITE_ONCE(*p, v);						\
+} while (0)
+
+#ifdef RSEQ_SKIP_FASTPATH
+#include "rseq-skip.h"
+#else /* !RSEQ_SKIP_FASTPATH */
+
+#define __RSEQ_ASM_DEFINE_TABLE(version, flags,	start_ip,		\
+				post_commit_offset, abort_ip)		\
+		".pushsection __rseq_table, \"aw\"\n\t"			\
+		".balign 32\n\t"					\
+		".word " __rseq_str(version) ", " __rseq_str(flags) "\n\t" \
+		".word " __rseq_str(start_ip) ", 0x0, " __rseq_str(post_commit_offset) ", 0x0, " __rseq_str(abort_ip) ", 0x0\n\t" \
+		".popsection\n\t"
+
+#define RSEQ_ASM_DEFINE_TABLE(start_ip, post_commit_ip, abort_ip)	\
+	__RSEQ_ASM_DEFINE_TABLE(0x0, 0x0, start_ip,			\
+				(post_commit_ip - start_ip), abort_ip)
+
+#define RSEQ_ASM_STORE_RSEQ_CS(label, cs_label, rseq_cs)		\
+		RSEQ_INJECT_ASM(1)					\
+		"adr r0, " __rseq_str(cs_label) "\n\t"			\
+		"str r0, %[" __rseq_str(rseq_cs) "]\n\t"		\
+		__rseq_str(label) ":\n\t"
+
+#define RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, label)		\
+		RSEQ_INJECT_ASM(2)					\
+		"ldr r0, %[" __rseq_str(current_cpu_id) "]\n\t"	\
+		"cmp %[" __rseq_str(cpu_id) "], r0\n\t"		\
+		"bne " __rseq_str(label) "\n\t"
+
+#define __RSEQ_ASM_DEFINE_ABORT(table_label, label, teardown,		\
+				abort_label, version, flags,		\
+				start_ip, post_commit_offset, abort_ip)	\
+		__rseq_str(table_label) ":\n\t"				\
+		".word " __rseq_str(version) ", " __rseq_str(flags) "\n\t" \
+		".word " __rseq_str(start_ip) ", 0x0, " __rseq_str(post_commit_offset) ", 0x0, " __rseq_str(abort_ip) ", 0x0\n\t" \
+		".word " __rseq_str(RSEQ_SIG) "\n\t"			\
+		__rseq_str(label) ":\n\t"				\
+		teardown						\
+		"b %l[" __rseq_str(abort_label) "]\n\t"
+
+#define RSEQ_ASM_DEFINE_ABORT(table_label, label, teardown, abort_label, \
+			      start_ip, post_commit_ip, abort_ip)	\
+	__RSEQ_ASM_DEFINE_ABORT(table_label, label, teardown,		\
+				abort_label, 0x0, 0x0, start_ip,	\
+				(post_commit_ip - start_ip), abort_ip)
+
+#define RSEQ_ASM_DEFINE_CMPFAIL(label, teardown, cmpfail_label)		\
+		__rseq_str(label) ":\n\t"				\
+		teardown						\
+		"b %l[" __rseq_str(cmpfail_label) "]\n\t"
+
+#define rseq_workaround_gcc_asm_size_guess()	__asm__ __volatile__("")
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_storev(intptr_t *v, intptr_t expect, intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	rseq_workaround_gcc_asm_size_guess();
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3f, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"ldr r0, %[v]\n\t"
+		"cmp %[expect], r0\n\t"
+		"bne %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"ldr r0, %[v]\n\t"
+		"cmp %[expect], r0\n\t"
+		"bne %l[error2]\n\t"
+#endif
+		/* final store */
+		"str %[newv], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(5)
+		"b 5f\n\t"
+		RSEQ_ASM_DEFINE_ABORT(3, 4, "", abort, 1b, 2b, 4f)
+		"5:\n\t"
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv)
+		  RSEQ_INJECT_INPUT
+		: "r0", "memory", "cc"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	rseq_workaround_gcc_asm_size_guess();
+	return 0;
+abort:
+	rseq_workaround_gcc_asm_size_guess();
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	rseq_workaround_gcc_asm_size_guess();
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpnev_storeoffp_load(intptr_t *v, intptr_t expectnot,
+			       off_t voffp, intptr_t *load, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	rseq_workaround_gcc_asm_size_guess();
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3f, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"ldr r0, %[v]\n\t"
+		"cmp %[expectnot], r0\n\t"
+		"beq %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"ldr r0, %[v]\n\t"
+		"cmp %[expectnot], r0\n\t"
+		"beq %l[error2]\n\t"
+#endif
+		"str r0, %[load]\n\t"
+		"add r0, %[voffp]\n\t"
+		"ldr r0, [r0]\n\t"
+		/* final store */
+		"str r0, %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(5)
+		"b 5f\n\t"
+		RSEQ_ASM_DEFINE_ABORT(3, 4, "", abort, 1b, 2b, 4f)
+		"5:\n\t"
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expectnot]		"r" (expectnot),
+		  [voffp]		"Ir" (voffp),
+		  [load]		"m" (*load)
+		  RSEQ_INJECT_INPUT
+		: "r0", "memory", "cc"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	rseq_workaround_gcc_asm_size_guess();
+	return 0;
+abort:
+	rseq_workaround_gcc_asm_size_guess();
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	rseq_workaround_gcc_asm_size_guess();
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_addv(intptr_t *v, intptr_t count, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	rseq_workaround_gcc_asm_size_guess();
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3f, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+#endif
+		"ldr r0, %[v]\n\t"
+		"add r0, %[count]\n\t"
+		/* final store */
+		"str r0, %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(4)
+		"b 5f\n\t"
+		RSEQ_ASM_DEFINE_ABORT(3, 4, "", abort, 1b, 2b, 4f)
+		"5:\n\t"
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  [v]			"m" (*v),
+		  [count]		"Ir" (count)
+		  RSEQ_INJECT_INPUT
+		: "r0", "memory", "cc"
+		  RSEQ_INJECT_CLOBBER
+		: abort
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1
+#endif
+	);
+	rseq_workaround_gcc_asm_size_guess();
+	return 0;
+abort:
+	rseq_workaround_gcc_asm_size_guess();
+	RSEQ_INJECT_FAILED
+	return -1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trystorev_storev(intptr_t *v, intptr_t expect,
+				 intptr_t *v2, intptr_t newv2,
+				 intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	rseq_workaround_gcc_asm_size_guess();
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3f, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"ldr r0, %[v]\n\t"
+		"cmp %[expect], r0\n\t"
+		"bne %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"ldr r0, %[v]\n\t"
+		"cmp %[expect], r0\n\t"
+		"bne %l[error2]\n\t"
+#endif
+		/* try store */
+		"str %[newv2], %[v2]\n\t"
+		RSEQ_INJECT_ASM(5)
+		/* final store */
+		"str %[newv], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(6)
+		"b 5f\n\t"
+		RSEQ_ASM_DEFINE_ABORT(3, 4, "", abort, 1b, 2b, 4f)
+		"5:\n\t"
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* try store input */
+		  [v2]			"m" (*v2),
+		  [newv2]		"r" (newv2),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv)
+		  RSEQ_INJECT_INPUT
+		: "r0", "memory", "cc"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	rseq_workaround_gcc_asm_size_guess();
+	return 0;
+abort:
+	rseq_workaround_gcc_asm_size_guess();
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	rseq_workaround_gcc_asm_size_guess();
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trystorev_storev_release(intptr_t *v, intptr_t expect,
+					 intptr_t *v2, intptr_t newv2,
+					 intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	rseq_workaround_gcc_asm_size_guess();
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3f, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"ldr r0, %[v]\n\t"
+		"cmp %[expect], r0\n\t"
+		"bne %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"ldr r0, %[v]\n\t"
+		"cmp %[expect], r0\n\t"
+		"bne %l[error2]\n\t"
+#endif
+		/* try store */
+		"str %[newv2], %[v2]\n\t"
+		RSEQ_INJECT_ASM(5)
+		"dmb\n\t"	/* full mb provides store-release */
+		/* final store */
+		"str %[newv], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(6)
+		"b 5f\n\t"
+		RSEQ_ASM_DEFINE_ABORT(3, 4, "", abort, 1b, 2b, 4f)
+		"5:\n\t"
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* try store input */
+		  [v2]			"m" (*v2),
+		  [newv2]		"r" (newv2),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv)
+		  RSEQ_INJECT_INPUT
+		: "r0", "memory", "cc"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	rseq_workaround_gcc_asm_size_guess();
+	return 0;
+abort:
+	rseq_workaround_gcc_asm_size_guess();
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	rseq_workaround_gcc_asm_size_guess();
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_cmpeqv_storev(intptr_t *v, intptr_t expect,
+			      intptr_t *v2, intptr_t expect2,
+			      intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	rseq_workaround_gcc_asm_size_guess();
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3f, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"ldr r0, %[v]\n\t"
+		"cmp %[expect], r0\n\t"
+		"bne %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+		"ldr r0, %[v2]\n\t"
+		"cmp %[expect2], r0\n\t"
+		"bne %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(5)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"ldr r0, %[v]\n\t"
+		"cmp %[expect], r0\n\t"
+		"bne %l[error2]\n\t"
+		"ldr r0, %[v2]\n\t"
+		"cmp %[expect2], r0\n\t"
+		"bne %l[error3]\n\t"
+#endif
+		/* final store */
+		"str %[newv], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(6)
+		"b 5f\n\t"
+		RSEQ_ASM_DEFINE_ABORT(3, 4, "", abort, 1b, 2b, 4f)
+		"5:\n\t"
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* cmp2 input */
+		  [v2]			"m" (*v2),
+		  [expect2]		"r" (expect2),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv)
+		  RSEQ_INJECT_INPUT
+		: "r0", "memory", "cc"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2, error3
+#endif
+	);
+	rseq_workaround_gcc_asm_size_guess();
+	return 0;
+abort:
+	rseq_workaround_gcc_asm_size_guess();
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	rseq_workaround_gcc_asm_size_guess();
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("1st expected value comparison failed");
+error3:
+	rseq_bug("2nd expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trymemcpy_storev(intptr_t *v, intptr_t expect,
+				 void *dst, void *src, size_t len,
+				 intptr_t newv, int cpu)
+{
+	uint32_t rseq_scratch[3];
+
+	RSEQ_INJECT_C(9)
+
+	rseq_workaround_gcc_asm_size_guess();
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(1f, 2f, 4f) /* start, commit, abort */
+		"str %[src], %[rseq_scratch0]\n\t"
+		"str %[dst], %[rseq_scratch1]\n\t"
+		"str %[len], %[rseq_scratch2]\n\t"
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3f, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"ldr r0, %[v]\n\t"
+		"cmp %[expect], r0\n\t"
+		"bne 5f\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 6f)
+		"ldr r0, %[v]\n\t"
+		"cmp %[expect], r0\n\t"
+		"bne 7f\n\t"
+#endif
+		/* try memcpy */
+		"cmp %[len], #0\n\t" \
+		"beq 333f\n\t" \
+		"222:\n\t" \
+		"ldrb %%r0, [%[src]]\n\t" \
+		"strb %%r0, [%[dst]]\n\t" \
+		"adds %[src], #1\n\t" \
+		"adds %[dst], #1\n\t" \
+		"subs %[len], #1\n\t" \
+		"bne 222b\n\t" \
+		"333:\n\t" \
+		RSEQ_INJECT_ASM(5)
+		/* final store */
+		"str %[newv], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(6)
+		/* teardown */
+		"ldr %[len], %[rseq_scratch2]\n\t"
+		"ldr %[dst], %[rseq_scratch1]\n\t"
+		"ldr %[src], %[rseq_scratch0]\n\t"
+		"b 8f\n\t"
+		RSEQ_ASM_DEFINE_ABORT(3, 4,
+				      /* teardown */
+				      "ldr %[len], %[rseq_scratch2]\n\t"
+				      "ldr %[dst], %[rseq_scratch1]\n\t"
+				      "ldr %[src], %[rseq_scratch0]\n\t",
+				      abort, 1b, 2b, 4f)
+		RSEQ_ASM_DEFINE_CMPFAIL(5,
+					/* teardown */
+					"ldr %[len], %[rseq_scratch2]\n\t"
+					"ldr %[dst], %[rseq_scratch1]\n\t"
+					"ldr %[src], %[rseq_scratch0]\n\t",
+					cmpfail)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_DEFINE_CMPFAIL(6,
+					/* teardown */
+					"ldr %[len], %[rseq_scratch2]\n\t"
+					"ldr %[dst], %[rseq_scratch1]\n\t"
+					"ldr %[src], %[rseq_scratch0]\n\t",
+					error1)
+		RSEQ_ASM_DEFINE_CMPFAIL(7,
+					/* teardown */
+					"ldr %[len], %[rseq_scratch2]\n\t"
+					"ldr %[dst], %[rseq_scratch1]\n\t"
+					"ldr %[src], %[rseq_scratch0]\n\t",
+					error2)
+#endif
+		"8:\n\t"
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv),
+		  /* try memcpy input */
+		  [dst]			"r" (dst),
+		  [src]			"r" (src),
+		  [len]			"r" (len),
+		  [rseq_scratch0]	"m" (rseq_scratch[0]),
+		  [rseq_scratch1]	"m" (rseq_scratch[1]),
+		  [rseq_scratch2]	"m" (rseq_scratch[2])
+		  RSEQ_INJECT_INPUT
+		: "r0", "memory", "cc"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	rseq_workaround_gcc_asm_size_guess();
+	return 0;
+abort:
+	rseq_workaround_gcc_asm_size_guess();
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	rseq_workaround_gcc_asm_size_guess();
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_workaround_gcc_asm_size_guess();
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_workaround_gcc_asm_size_guess();
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trymemcpy_storev_release(intptr_t *v, intptr_t expect,
+					 void *dst, void *src, size_t len,
+					 intptr_t newv, int cpu)
+{
+	uint32_t rseq_scratch[3];
+
+	RSEQ_INJECT_C(9)
+
+	rseq_workaround_gcc_asm_size_guess();
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(1f, 2f, 4f) /* start, commit, abort */
+		"str %[src], %[rseq_scratch0]\n\t"
+		"str %[dst], %[rseq_scratch1]\n\t"
+		"str %[len], %[rseq_scratch2]\n\t"
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3f, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"ldr r0, %[v]\n\t"
+		"cmp %[expect], r0\n\t"
+		"bne 5f\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 6f)
+		"ldr r0, %[v]\n\t"
+		"cmp %[expect], r0\n\t"
+		"bne 7f\n\t"
+#endif
+		/* try memcpy */
+		"cmp %[len], #0\n\t" \
+		"beq 333f\n\t" \
+		"222:\n\t" \
+		"ldrb %%r0, [%[src]]\n\t" \
+		"strb %%r0, [%[dst]]\n\t" \
+		"adds %[src], #1\n\t" \
+		"adds %[dst], #1\n\t" \
+		"subs %[len], #1\n\t" \
+		"bne 222b\n\t" \
+		"333:\n\t" \
+		RSEQ_INJECT_ASM(5)
+		"dmb\n\t"	/* full mb provides store-release */
+		/* final store */
+		"str %[newv], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(6)
+		/* teardown */
+		"ldr %[len], %[rseq_scratch2]\n\t"
+		"ldr %[dst], %[rseq_scratch1]\n\t"
+		"ldr %[src], %[rseq_scratch0]\n\t"
+		"b 8f\n\t"
+		RSEQ_ASM_DEFINE_ABORT(3, 4,
+				      /* teardown */
+				      "ldr %[len], %[rseq_scratch2]\n\t"
+				      "ldr %[dst], %[rseq_scratch1]\n\t"
+				      "ldr %[src], %[rseq_scratch0]\n\t",
+				      abort, 1b, 2b, 4f)
+		RSEQ_ASM_DEFINE_CMPFAIL(5,
+					/* teardown */
+					"ldr %[len], %[rseq_scratch2]\n\t"
+					"ldr %[dst], %[rseq_scratch1]\n\t"
+					"ldr %[src], %[rseq_scratch0]\n\t",
+					cmpfail)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_DEFINE_CMPFAIL(6,
+					/* teardown */
+					"ldr %[len], %[rseq_scratch2]\n\t"
+					"ldr %[dst], %[rseq_scratch1]\n\t"
+					"ldr %[src], %[rseq_scratch0]\n\t",
+					error1)
+		RSEQ_ASM_DEFINE_CMPFAIL(7,
+					/* teardown */
+					"ldr %[len], %[rseq_scratch2]\n\t"
+					"ldr %[dst], %[rseq_scratch1]\n\t"
+					"ldr %[src], %[rseq_scratch0]\n\t",
+					error2)
+#endif
+		"8:\n\t"
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv),
+		  /* try memcpy input */
+		  [dst]			"r" (dst),
+		  [src]			"r" (src),
+		  [len]			"r" (len),
+		  [rseq_scratch0]	"m" (rseq_scratch[0]),
+		  [rseq_scratch1]	"m" (rseq_scratch[1]),
+		  [rseq_scratch2]	"m" (rseq_scratch[2])
+		  RSEQ_INJECT_INPUT
+		: "r0", "memory", "cc"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	rseq_workaround_gcc_asm_size_guess();
+	return 0;
+abort:
+	rseq_workaround_gcc_asm_size_guess();
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	rseq_workaround_gcc_asm_size_guess();
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_workaround_gcc_asm_size_guess();
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_workaround_gcc_asm_size_guess();
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+#endif /* !RSEQ_SKIP_FASTPATH */
diff --git a/tools/testing/selftests/rseq/rseq-ppc.h b/tools/testing/selftests/rseq/rseq-ppc.h
new file mode 100644
index 000000000000..2d6d2dfa1235
--- /dev/null
+++ b/tools/testing/selftests/rseq/rseq-ppc.h
@@ -0,0 +1,688 @@
+/*
+ * rseq-ppc.h
+ *
+ * (C) Copyright 2016 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ * (C) Copyright 2016 - Boqun Feng <boqun.feng@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#define RSEQ_SIG	0x53053053
+
+#define rseq_smp_mb()		__asm__ __volatile__ ("sync"	::: "memory", "cc")
+#define rseq_smp_lwsync()	__asm__ __volatile__ ("lwsync"	::: "memory", "cc")
+#define rseq_smp_rmb()		rseq_smp_lwsync()
+#define rseq_smp_wmb()		rseq_smp_lwsync()
+
+#define rseq_smp_load_acquire(p)					\
+__extension__ ({							\
+	__typeof(*p) ____p1 = RSEQ_READ_ONCE(*p);			\
+	rseq_smp_lwsync();						\
+	____p1;								\
+})
+
+#define rseq_smp_acquire__after_ctrl_dep()	rseq_smp_lwsync()
+
+#define rseq_smp_store_release(p, v)					\
+do {									\
+	rseq_smp_lwsync();						\
+	RSEQ_WRITE_ONCE(*p, v);						\
+} while (0)
+
+#ifdef RSEQ_SKIP_FASTPATH
+#include "rseq-skip.h"
+#else /* !RSEQ_SKIP_FASTPATH */
+
+/*
+ * The __rseq_table section can be used by debuggers to better handle
+ * single-stepping through the restartable critical sections.
+ */
+
+#ifdef __PPC64__
+
+#define STORE_WORD	"std "
+#define LOAD_WORD	"ld "
+#define LOADX_WORD	"ldx "
+#define CMP_WORD	"cmpd "
+
+#define __RSEQ_ASM_DEFINE_TABLE(label, version, flags,				\
+			start_ip, post_commit_offset, abort_ip)			\
+		".pushsection __rseq_table, \"aw\"\n\t"				\
+		".balign 32\n\t"						\
+		__rseq_str(label) ":\n\t"					\
+		".long " __rseq_str(version) ", " __rseq_str(flags) "\n\t"	\
+		".quad " __rseq_str(start_ip) ", " __rseq_str(post_commit_offset) ", " __rseq_str(abort_ip) "\n\t" \
+		".popsection\n\t"
+
+#define RSEQ_ASM_STORE_RSEQ_CS(label, cs_label, rseq_cs)			\
+		RSEQ_INJECT_ASM(1)						\
+		"lis %%r17, (" __rseq_str(cs_label) ")@highest\n\t"		\
+		"ori %%r17, %%r17, (" __rseq_str(cs_label) ")@higher\n\t"	\
+		"rldicr %%r17, %%r17, 32, 31\n\t"				\
+		"oris %%r17, %%r17, (" __rseq_str(cs_label) ")@high\n\t"	\
+		"ori %%r17, %%r17, (" __rseq_str(cs_label) ")@l\n\t"		\
+		"std %%r17, %[" __rseq_str(rseq_cs) "]\n\t"			\
+		__rseq_str(label) ":\n\t"
+
+#else /* #ifdef __PPC64__ */
+
+#define STORE_WORD	"stw "
+#define LOAD_WORD	"lwz "
+#define LOADX_WORD	"lwzx "
+#define CMP_WORD	"cmpw "
+
+#define __RSEQ_ASM_DEFINE_TABLE(label, version, flags,				\
+			start_ip, post_commit_offset, abort_ip)			\
+		".pushsection __rseq_table, \"aw\"\n\t"				\
+		".balign 32\n\t"						\
+		__rseq_str(label) ":\n\t"					\
+		".long " __rseq_str(version) ", " __rseq_str(flags) "\n\t"	\
+		/* 32-bit only supported on BE */				\
+		".long 0x0, " __rseq_str(start_ip) ", 0x0, " __rseq_str(post_commit_offset) ", 0x0, " __rseq_str(abort_ip) "\n\t" \
+		".popsection\n\t"
+
+#define RSEQ_ASM_STORE_RSEQ_CS(label, cs_label, rseq_cs)			\
+		RSEQ_INJECT_ASM(1)						\
+		"lis %%r17, (" __rseq_str(cs_label) ")@ha\n\t"			\
+		"addi %%r17, %%r17, (" __rseq_str(cs_label) ")@l\n\t"		\
+		"stw %%r17, %[" __rseq_str(rseq_cs) "]\n\t"			\
+		__rseq_str(label) ":\n\t"
+
+#endif /* #ifdef __PPC64__ */
+
+#define RSEQ_ASM_DEFINE_TABLE(label, start_ip, post_commit_ip, abort_ip)	\
+		__RSEQ_ASM_DEFINE_TABLE(label, 0x0, 0x0, start_ip,		\
+					(post_commit_ip - start_ip), abort_ip)
+
+#define RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, label)			\
+		RSEQ_INJECT_ASM(2)						\
+		"lwz %%r17, %[" __rseq_str(current_cpu_id) "]\n\t"		\
+		"cmpw cr7, %[" __rseq_str(cpu_id) "], %%r17\n\t"		\
+		"bne- cr7, " __rseq_str(label) "\n\t"
+
+#define RSEQ_ASM_DEFINE_ABORT(label, abort_label)				\
+		".pushsection __rseq_failure, \"ax\"\n\t"			\
+		".long " __rseq_str(RSEQ_SIG) "\n\t"				\
+		__rseq_str(label) ":\n\t"					\
+		"b %l[" __rseq_str(abort_label) "]\n\t"				\
+		".popsection\n\t"
+
+/*
+ * RSEQ_ASM_OPs: asm operations for rseq
+ * 	RSEQ_ASM_OP_R_*: has hard-code registers in it
+ * 	RSEQ_ASM_OP_* (else): doesn't have hard-code registers(unless cr7)
+ */
+#define RSEQ_ASM_OP_CMPEQ(var, expect, label)					\
+		LOAD_WORD "%%r17, %[" __rseq_str(var) "]\n\t"			\
+		CMP_WORD "cr7, %%r17, %[" __rseq_str(expect) "]\n\t"		\
+		"bne- cr7, " __rseq_str(label) "\n\t"
+
+#define RSEQ_ASM_OP_CMPNE(var, expectnot, label)				\
+		LOAD_WORD "%%r17, %[" __rseq_str(var) "]\n\t"			\
+		CMP_WORD "cr7, %%r17, %[" __rseq_str(expectnot) "]\n\t"		\
+		"beq- cr7, " __rseq_str(label) "\n\t"
+
+#define RSEQ_ASM_OP_STORE(value, var)						\
+		STORE_WORD "%[" __rseq_str(value) "], %[" __rseq_str(var) "]\n\t"
+
+/* Load @var to r17 */
+#define RSEQ_ASM_OP_R_LOAD(var)							\
+		LOAD_WORD "%%r17, %[" __rseq_str(var) "]\n\t"
+
+/* Store r17 to @var */
+#define RSEQ_ASM_OP_R_STORE(var)						\
+		STORE_WORD "%%r17, %[" __rseq_str(var) "]\n\t"
+
+/* Add @count to r17 */
+#define RSEQ_ASM_OP_R_ADD(count)						\
+		"add %%r17, %[" __rseq_str(count) "], %%r17\n\t"
+
+/* Load (r17 + voffp) to r17 */
+#define RSEQ_ASM_OP_R_LOADX(voffp)						\
+		LOADX_WORD "%%r17, %[" __rseq_str(voffp) "], %%r17\n\t"
+
+/* TODO: implement a faster memcpy. */
+#define RSEQ_ASM_OP_R_MEMCPY() \
+		"cmpdi %%r19, 0\n\t" \
+		"beq 333f\n\t" \
+		"addi %%r20, %%r20, -1\n\t" \
+		"addi %%r21, %%r21, -1\n\t" \
+		"222:\n\t" \
+		"lbzu %%r18, 1(%%r20)\n\t" \
+		"stbu %%r18, 1(%%r21)\n\t" \
+		"addi %%r19, %%r19, -1\n\t" \
+		"cmpdi %%r19, 0\n\t" \
+		"bne 222b\n\t" \
+		"333:\n\t" \
+
+#define RSEQ_ASM_OP_R_FINAL_STORE(var, post_commit_label)			\
+		STORE_WORD "%%r17, %[" __rseq_str(var) "]\n\t"			\
+		__rseq_str(post_commit_label) ":\n\t"
+
+#define RSEQ_ASM_OP_FINAL_STORE(value, var, post_commit_label)			\
+		STORE_WORD "%[" __rseq_str(value) "], %[" __rseq_str(var) "]\n\t" \
+		__rseq_str(post_commit_label) ":\n\t"
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_storev(intptr_t *v, intptr_t expect, intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		/* cmp @v equal to @expect */
+		RSEQ_ASM_OP_CMPEQ(v, expect, %l[cmpfail])
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		/* cmp @v equal to @expect */
+		RSEQ_ASM_OP_CMPEQ(v, expect, %l[error2])
+#endif
+		/* final store */
+		RSEQ_ASM_OP_FINAL_STORE(newv, v, 2)
+		RSEQ_INJECT_ASM(5)
+		RSEQ_ASM_DEFINE_ABORT(4, abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv)
+		  RSEQ_INJECT_INPUT
+		: "memory", "cc", "r17"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpnev_storeoffp_load(intptr_t *v, intptr_t expectnot,
+			       off_t voffp, intptr_t *load, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		/* cmp @v not equal to @expectnot */
+		RSEQ_ASM_OP_CMPNE(v, expectnot, %l[cmpfail])
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		/* cmp @v not equal to @expectnot */
+		RSEQ_ASM_OP_CMPNE(v, expectnot, %l[error2])
+#endif
+		/* load the value of @v */
+		RSEQ_ASM_OP_R_LOAD(v)
+		/* store it in @load */
+		RSEQ_ASM_OP_R_STORE(load)
+		/* dereference voffp(v) */
+		RSEQ_ASM_OP_R_LOADX(voffp)
+		/* final store the value at voffp(v) */
+		RSEQ_ASM_OP_R_FINAL_STORE(v, 2)
+		RSEQ_INJECT_ASM(5)
+		RSEQ_ASM_DEFINE_ABORT(4, abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expectnot]		"r" (expectnot),
+		  [voffp]		"b" (voffp),
+		  [load]		"m" (*load)
+		  RSEQ_INJECT_INPUT
+		: "memory", "cc", "r17"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_addv(intptr_t *v, intptr_t count, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+#ifdef RSEQ_COMPARE_TWICE
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+#endif
+		/* load the value of @v */
+		RSEQ_ASM_OP_R_LOAD(v)
+		/* add @count to it */
+		RSEQ_ASM_OP_R_ADD(count)
+		/* final store */
+		RSEQ_ASM_OP_R_FINAL_STORE(v, 2)
+		RSEQ_INJECT_ASM(4)
+		RSEQ_ASM_DEFINE_ABORT(4, abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [count]		"r" (count)
+		  RSEQ_INJECT_INPUT
+		: "memory", "cc", "r17"
+		  RSEQ_INJECT_CLOBBER
+		: abort
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trystorev_storev(intptr_t *v, intptr_t expect,
+				 intptr_t *v2, intptr_t newv2,
+				 intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		/* cmp @v equal to @expect */
+		RSEQ_ASM_OP_CMPEQ(v, expect, %l[cmpfail])
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		/* cmp @v equal to @expect */
+		RSEQ_ASM_OP_CMPEQ(v, expect, %l[error2])
+#endif
+		/* try store */
+		RSEQ_ASM_OP_STORE(newv2, v2)
+		RSEQ_INJECT_ASM(5)
+		/* final store */
+		RSEQ_ASM_OP_FINAL_STORE(newv, v, 2)
+		RSEQ_INJECT_ASM(6)
+		RSEQ_ASM_DEFINE_ABORT(4, abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* try store input */
+		  [v2]			"m" (*v2),
+		  [newv2]		"r" (newv2),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv)
+		  RSEQ_INJECT_INPUT
+		: "memory", "cc", "r17"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trystorev_storev_release(intptr_t *v, intptr_t expect,
+					 intptr_t *v2, intptr_t newv2,
+					 intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		/* cmp @v equal to @expect */
+		RSEQ_ASM_OP_CMPEQ(v, expect, %l[cmpfail])
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		/* cmp @v equal to @expect */
+		RSEQ_ASM_OP_CMPEQ(v, expect, %l[error2])
+#endif
+		/* try store */
+		RSEQ_ASM_OP_STORE(newv2, v2)
+		RSEQ_INJECT_ASM(5)
+		/* for 'release' */
+		"lwsync\n\t"
+		/* final store */
+		RSEQ_ASM_OP_FINAL_STORE(newv, v, 2)
+		RSEQ_INJECT_ASM(6)
+		RSEQ_ASM_DEFINE_ABORT(4, abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* try store input */
+		  [v2]			"m" (*v2),
+		  [newv2]		"r" (newv2),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv)
+		  RSEQ_INJECT_INPUT
+		: "memory", "cc", "r17"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_cmpeqv_storev(intptr_t *v, intptr_t expect,
+			      intptr_t *v2, intptr_t expect2,
+			      intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		/* cmp @v equal to @expect */
+		RSEQ_ASM_OP_CMPEQ(v, expect, %l[cmpfail])
+		RSEQ_INJECT_ASM(4)
+		/* cmp @v2 equal to @expct2 */
+		RSEQ_ASM_OP_CMPEQ(v2, expect2, %l[cmpfail])
+		RSEQ_INJECT_ASM(5)
+#ifdef RSEQ_COMPARE_TWICE
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		/* cmp @v equal to @expect */
+		RSEQ_ASM_OP_CMPEQ(v, expect, %l[error2])
+		/* cmp @v2 equal to @expct2 */
+		RSEQ_ASM_OP_CMPEQ(v2, expect2, %l[error3])
+#endif
+		/* final store */
+		RSEQ_ASM_OP_FINAL_STORE(newv, v, 2)
+		RSEQ_INJECT_ASM(6)
+		RSEQ_ASM_DEFINE_ABORT(4, abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* cmp2 input */
+		  [v2]			"m" (*v2),
+		  [expect2]		"r" (expect2),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv)
+		  RSEQ_INJECT_INPUT
+		: "memory", "cc", "r17"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2, error3
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("1st expected value comparison failed");
+error3:
+	rseq_bug("2nd expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trymemcpy_storev(intptr_t *v, intptr_t expect,
+				 void *dst, void *src, size_t len,
+				 intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* setup for mempcy */
+		"mr %%r19, %[len]\n\t"
+		"mr %%r20, %[src]\n\t"
+		"mr %%r21, %[dst]\n\t"
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		/* cmp @v equal to @expect */
+		RSEQ_ASM_OP_CMPEQ(v, expect, %l[cmpfail])
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		/* cmp @v equal to @expect */
+		RSEQ_ASM_OP_CMPEQ(v, expect, %l[error2])
+#endif
+		/* try memcpy */
+		RSEQ_ASM_OP_R_MEMCPY()
+		RSEQ_INJECT_ASM(5)
+		/* final store */
+		RSEQ_ASM_OP_FINAL_STORE(newv, v, 2)
+		RSEQ_INJECT_ASM(6)
+		/* teardown */
+		RSEQ_ASM_DEFINE_ABORT(4, abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv),
+		  /* try memcpy input */
+		  [dst]			"r" (dst),
+		  [src]			"r" (src),
+		  [len]			"r" (len)
+		  RSEQ_INJECT_INPUT
+		: "memory", "cc", "r17", "r18", "r19", "r20", "r21"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trymemcpy_storev_release(intptr_t *v, intptr_t expect,
+					 void *dst, void *src, size_t len,
+					 intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* setup for mempcy */
+		"mr %%r19, %[len]\n\t"
+		"mr %%r20, %[src]\n\t"
+		"mr %%r21, %[dst]\n\t"
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		/* cmp @v equal to @expect */
+		RSEQ_ASM_OP_CMPEQ(v, expect, %l[cmpfail])
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		/* cmp cpuid */
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		/* cmp @v equal to @expect */
+		RSEQ_ASM_OP_CMPEQ(v, expect, %l[error2])
+#endif
+		/* try memcpy */
+		RSEQ_ASM_OP_R_MEMCPY()
+		RSEQ_INJECT_ASM(5)
+		/* for 'release' */
+		"lwsync\n\t"
+		/* final store */
+		RSEQ_ASM_OP_FINAL_STORE(newv, v, 2)
+		RSEQ_INJECT_ASM(6)
+		/* teardown */
+		RSEQ_ASM_DEFINE_ABORT(4, abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv),
+		  /* try memcpy input */
+		  [dst]			"r" (dst),
+		  [src]			"r" (src),
+		  [len]			"r" (len)
+		  RSEQ_INJECT_INPUT
+		: "memory", "cc", "r17", "r18", "r19", "r20", "r21"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+#undef STORE_WORD
+#undef LOAD_WORD
+#undef LOADX_WORD
+#undef CMP_WORD
+
+#endif /* !RSEQ_SKIP_FASTPATH */
diff --git a/tools/testing/selftests/rseq/rseq-skip.h b/tools/testing/selftests/rseq/rseq-skip.h
new file mode 100644
index 000000000000..dc8f8e74b737
--- /dev/null
+++ b/tools/testing/selftests/rseq/rseq-skip.h
@@ -0,0 +1,82 @@
+/*
+ * rseq-skip.h
+ *
+ * (C) Copyright 2017 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_storev(intptr_t *v, intptr_t expect, intptr_t newv, int cpu)
+{
+	return -1;
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpnev_storeoffp_load(intptr_t *v, intptr_t expectnot,
+			       off_t voffp, intptr_t *load, int cpu)
+{
+	return -1;
+}
+
+static inline __attribute__((always_inline))
+int rseq_addv(intptr_t *v, intptr_t count, int cpu)
+{
+	return -1;
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trystorev_storev(intptr_t *v, intptr_t expect,
+				 intptr_t *v2, intptr_t newv2,
+				 intptr_t newv, int cpu)
+{
+	return -1;
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trystorev_storev_release(intptr_t *v, intptr_t expect,
+					 intptr_t *v2, intptr_t newv2,
+					 intptr_t newv, int cpu)
+{
+	return -1;
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_cmpeqv_storev(intptr_t *v, intptr_t expect,
+			      intptr_t *v2, intptr_t expect2,
+			      intptr_t newv, int cpu)
+{
+	return -1;
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trymemcpy_storev(intptr_t *v, intptr_t expect,
+				 void *dst, void *src, size_t len,
+				 intptr_t newv, int cpu)
+{
+	return -1;
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trymemcpy_storev_release(intptr_t *v, intptr_t expect,
+					 void *dst, void *src, size_t len,
+					 intptr_t newv, int cpu)
+{
+	return -1;
+}
diff --git a/tools/testing/selftests/rseq/rseq-x86.h b/tools/testing/selftests/rseq/rseq-x86.h
new file mode 100644
index 000000000000..02d853ae2ce1
--- /dev/null
+++ b/tools/testing/selftests/rseq/rseq-x86.h
@@ -0,0 +1,1149 @@
+/*
+ * rseq-x86.h
+ *
+ * (C) Copyright 2016 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <stdint.h>
+
+#define RSEQ_SIG	0x53053053
+
+#ifdef __x86_64__
+
+#define rseq_smp_mb()	\
+	__asm__ __volatile__ ("lock; addl $0,-128(%%rsp)" ::: "memory", "cc")
+#define rseq_smp_rmb()	rseq_barrier()
+#define rseq_smp_wmb()	rseq_barrier()
+
+#define rseq_smp_load_acquire(p)					\
+__extension__ ({							\
+	__typeof(*p) ____p1 = RSEQ_READ_ONCE(*p);			\
+	rseq_barrier();							\
+	____p1;								\
+})
+
+#define rseq_smp_acquire__after_ctrl_dep()	rseq_smp_rmb()
+
+#define rseq_smp_store_release(p, v)					\
+do {									\
+	rseq_barrier();							\
+	RSEQ_WRITE_ONCE(*p, v);						\
+} while (0)
+
+#ifdef RSEQ_SKIP_FASTPATH
+#include "rseq-skip.h"
+#else /* !RSEQ_SKIP_FASTPATH */
+
+#define __RSEQ_ASM_DEFINE_TABLE(label, version, flags,			\
+				start_ip, post_commit_offset, abort_ip)	\
+		".pushsection __rseq_table, \"aw\"\n\t"			\
+		".balign 32\n\t"					\
+		__rseq_str(label) ":\n\t"				\
+		".long " __rseq_str(version) ", " __rseq_str(flags) "\n\t" \
+		".quad " __rseq_str(start_ip) ", " __rseq_str(post_commit_offset) ", " __rseq_str(abort_ip) "\n\t" \
+		".popsection\n\t"
+
+#define RSEQ_ASM_DEFINE_TABLE(label, start_ip, post_commit_ip, abort_ip) \
+	__RSEQ_ASM_DEFINE_TABLE(label, 0x0, 0x0, start_ip,		\
+				(post_commit_ip - start_ip), abort_ip)
+
+#define RSEQ_ASM_STORE_RSEQ_CS(label, cs_label, rseq_cs)		\
+		RSEQ_INJECT_ASM(1)					\
+		"leaq " __rseq_str(cs_label) "(%%rip), %%rax\n\t"	\
+		"movq %%rax, %[" __rseq_str(rseq_cs) "]\n\t"		\
+		__rseq_str(label) ":\n\t"
+
+#define RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, label)		\
+		RSEQ_INJECT_ASM(2)					\
+		"cmpl %[" __rseq_str(cpu_id) "], %[" __rseq_str(current_cpu_id) "]\n\t" \
+		"jnz " __rseq_str(label) "\n\t"
+
+#define RSEQ_ASM_DEFINE_ABORT(label, teardown, abort_label)		\
+		".pushsection __rseq_failure, \"ax\"\n\t"		\
+		/* Disassembler-friendly signature: nopl <sig>(%rip). */\
+		".byte 0x0f, 0x1f, 0x05\n\t"				\
+		".long " __rseq_str(RSEQ_SIG) "\n\t"			\
+		__rseq_str(label) ":\n\t"				\
+		teardown						\
+		"jmp %l[" __rseq_str(abort_label) "]\n\t"		\
+		".popsection\n\t"
+
+#define RSEQ_ASM_DEFINE_CMPFAIL(label, teardown, cmpfail_label)		\
+		".pushsection __rseq_failure, \"ax\"\n\t"		\
+		__rseq_str(label) ":\n\t"				\
+		teardown						\
+		"jmp %l[" __rseq_str(cmpfail_label) "]\n\t"		\
+		".popsection\n\t"
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_storev(intptr_t *v, intptr_t expect, intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"cmpq %[v], %[expect]\n\t"
+		"jnz %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"cmpq %[v], %[expect]\n\t"
+		"jnz %l[error2]\n\t"
+#endif
+		/* final store */
+		"movq %[newv], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(5)
+		RSEQ_ASM_DEFINE_ABORT(4, "", abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv)
+		: "memory", "cc", "rax"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+/*
+ * Compare @v against @expectnot. When it does _not_ match, load @v
+ * into @load, and store the content of *@v + voffp into @v.
+ */
+static inline __attribute__((always_inline))
+int rseq_cmpnev_storeoffp_load(intptr_t *v, intptr_t expectnot,
+			       off_t voffp, intptr_t *load, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"movq %[v], %%rbx\n\t"
+		"cmpq %%rbx, %[expectnot]\n\t"
+		"je %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"movq %[v], %%rbx\n\t"
+		"cmpq %%rbx, %[expectnot]\n\t"
+		"je %l[error2]\n\t"
+#endif
+		"movq %%rbx, %[load]\n\t"
+		"addq %[voffp], %%rbx\n\t"
+		"movq (%%rbx), %%rbx\n\t"
+		/* final store */
+		"movq %%rbx, %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(5)
+		RSEQ_ASM_DEFINE_ABORT(4, "", abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expectnot]		"r" (expectnot),
+		  [voffp]		"er" (voffp),
+		  [load]		"m" (*load)
+		: "memory", "cc", "rax", "rbx"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_addv(intptr_t *v, intptr_t count, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+#endif
+		/* final store */
+		"addq %[count], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(4)
+		RSEQ_ASM_DEFINE_ABORT(4, "", abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [count]		"er" (count)
+		: "memory", "cc", "rax"
+		  RSEQ_INJECT_CLOBBER
+		: abort
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trystorev_storev(intptr_t *v, intptr_t expect,
+				 intptr_t *v2, intptr_t newv2,
+				 intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"cmpq %[v], %[expect]\n\t"
+		"jnz %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"cmpq %[v], %[expect]\n\t"
+		"jnz %l[error2]\n\t"
+#endif
+		/* try store */
+		"movq %[newv2], %[v2]\n\t"
+		RSEQ_INJECT_ASM(5)
+		/* final store */
+		"movq %[newv], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(6)
+		RSEQ_ASM_DEFINE_ABORT(4, "", abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* try store input */
+		  [v2]			"m" (*v2),
+		  [newv2]		"r" (newv2),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv)
+		: "memory", "cc", "rax"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+/* x86-64 is TSO. */
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trystorev_storev_release(intptr_t *v, intptr_t expect,
+					 intptr_t *v2, intptr_t newv2,
+					 intptr_t newv, int cpu)
+{
+	return rseq_cmpeqv_trystorev_storev(v, expect, v2, newv2, newv, cpu);
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_cmpeqv_storev(intptr_t *v, intptr_t expect,
+			      intptr_t *v2, intptr_t expect2,
+			      intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"cmpq %[v], %[expect]\n\t"
+		"jnz %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+		"cmpq %[v2], %[expect2]\n\t"
+		"jnz %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(5)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"cmpq %[v], %[expect]\n\t"
+		"jnz %l[error2]\n\t"
+		"cmpq %[v2], %[expect2]\n\t"
+		"jnz %l[error3]\n\t"
+#endif
+		/* final store */
+		"movq %[newv], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(6)
+		RSEQ_ASM_DEFINE_ABORT(4, "", abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* cmp2 input */
+		  [v2]			"m" (*v2),
+		  [expect2]		"r" (expect2),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv)
+		: "memory", "cc", "rax"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2, error3
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("1st expected value comparison failed");
+error3:
+	rseq_bug("2nd expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trymemcpy_storev(intptr_t *v, intptr_t expect,
+				 void *dst, void *src, size_t len,
+				 intptr_t newv, int cpu)
+{
+	uint64_t rseq_scratch[3];
+
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		"movq %[src], %[rseq_scratch0]\n\t"
+		"movq %[dst], %[rseq_scratch1]\n\t"
+		"movq %[len], %[rseq_scratch2]\n\t"
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"cmpq %[v], %[expect]\n\t"
+		"jnz 5f\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 6f)
+		"cmpq %[v], %[expect]\n\t"
+		"jnz 7f\n\t"
+#endif
+		/* try memcpy */
+		"test %[len], %[len]\n\t" \
+		"jz 333f\n\t" \
+		"222:\n\t" \
+		"movb (%[src]), %%al\n\t" \
+		"movb %%al, (%[dst])\n\t" \
+		"inc %[src]\n\t" \
+		"inc %[dst]\n\t" \
+		"dec %[len]\n\t" \
+		"jnz 222b\n\t" \
+		"333:\n\t" \
+		RSEQ_INJECT_ASM(5)
+		/* final store */
+		"movq %[newv], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(6)
+		/* teardown */
+		"movq %[rseq_scratch2], %[len]\n\t"
+		"movq %[rseq_scratch1], %[dst]\n\t"
+		"movq %[rseq_scratch0], %[src]\n\t"
+		RSEQ_ASM_DEFINE_ABORT(4,
+			"movq %[rseq_scratch2], %[len]\n\t"
+			"movq %[rseq_scratch1], %[dst]\n\t"
+			"movq %[rseq_scratch0], %[src]\n\t",
+			abort)
+		RSEQ_ASM_DEFINE_CMPFAIL(5,
+			"movq %[rseq_scratch2], %[len]\n\t"
+			"movq %[rseq_scratch1], %[dst]\n\t"
+			"movq %[rseq_scratch0], %[src]\n\t",
+			cmpfail)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_DEFINE_CMPFAIL(6,
+			"movq %[rseq_scratch2], %[len]\n\t"
+			"movq %[rseq_scratch1], %[dst]\n\t"
+			"movq %[rseq_scratch0], %[src]\n\t",
+			error1)
+		RSEQ_ASM_DEFINE_CMPFAIL(7,
+			"movq %[rseq_scratch2], %[len]\n\t"
+			"movq %[rseq_scratch1], %[dst]\n\t"
+			"movq %[rseq_scratch0], %[src]\n\t",
+			error2)
+#endif
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv),
+		  /* try memcpy input */
+		  [dst]			"r" (dst),
+		  [src]			"r" (src),
+		  [len]			"r" (len),
+		  [rseq_scratch0]	"m" (rseq_scratch[0]),
+		  [rseq_scratch1]	"m" (rseq_scratch[1]),
+		  [rseq_scratch2]	"m" (rseq_scratch[2])
+		: "memory", "cc", "rax"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+/* x86-64 is TSO. */
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trymemcpy_storev_release(intptr_t *v, intptr_t expect,
+					 void *dst, void *src, size_t len,
+					 intptr_t newv, int cpu)
+{
+	return rseq_cmpeqv_trymemcpy_storev(v, expect, dst, src, len,
+					    newv, cpu);
+}
+
+#endif /* !RSEQ_SKIP_FASTPATH */
+
+#elif __i386__
+
+#define rseq_smp_mb()	\
+	__asm__ __volatile__ ("lock; addl $0,-128(%%esp)" ::: "memory", "cc")
+#define rseq_smp_rmb()	\
+	__asm__ __volatile__ ("lock; addl $0,-128(%%esp)" ::: "memory", "cc")
+#define rseq_smp_wmb()	\
+	__asm__ __volatile__ ("lock; addl $0,-128(%%esp)" ::: "memory", "cc")
+
+#define rseq_smp_load_acquire(p)					\
+__extension__ ({							\
+	__typeof(*p) ____p1 = RSEQ_READ_ONCE(*p);			\
+	rseq_smp_mb();							\
+	____p1;								\
+})
+
+#define rseq_smp_acquire__after_ctrl_dep()	rseq_smp_rmb()
+
+#define rseq_smp_store_release(p, v)					\
+do {									\
+	rseq_smp_mb();							\
+	RSEQ_WRITE_ONCE(*p, v);						\
+} while (0)
+
+#ifdef RSEQ_SKIP_FASTPATH
+#include "rseq-skip.h"
+#else /* !RSEQ_SKIP_FASTPATH */
+
+/*
+ * Use eax as scratch register and take memory operands as input to
+ * lessen register pressure. Especially needed when compiling in O0.
+ */
+#define __RSEQ_ASM_DEFINE_TABLE(label, version, flags,			\
+				start_ip, post_commit_offset, abort_ip)	\
+		".pushsection __rseq_table, \"aw\"\n\t"			\
+		".balign 32\n\t"					\
+		__rseq_str(label) ":\n\t"				\
+		".long " __rseq_str(version) ", " __rseq_str(flags) "\n\t" \
+		".long " __rseq_str(start_ip) ", 0x0, " __rseq_str(post_commit_offset) ", 0x0, " __rseq_str(abort_ip) ", 0x0\n\t" \
+		".popsection\n\t"
+
+#define RSEQ_ASM_DEFINE_TABLE(label, start_ip, post_commit_ip, abort_ip) \
+	__RSEQ_ASM_DEFINE_TABLE(label, 0x0, 0x0, start_ip,		\
+				(post_commit_ip - start_ip), abort_ip)
+
+#define RSEQ_ASM_STORE_RSEQ_CS(label, cs_label, rseq_cs)		\
+		RSEQ_INJECT_ASM(1)					\
+		"movl $" __rseq_str(cs_label) ", %[rseq_cs]\n\t"	\
+		__rseq_str(label) ":\n\t"
+
+#define RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, label)		\
+		RSEQ_INJECT_ASM(2)					\
+		"cmpl %[" __rseq_str(cpu_id) "], %[" __rseq_str(current_cpu_id) "]\n\t" \
+		"jnz " __rseq_str(label) "\n\t"
+
+#define RSEQ_ASM_DEFINE_ABORT(label, teardown, abort_label)		\
+		".pushsection __rseq_failure, \"ax\"\n\t"		\
+		/* Disassembler-friendly signature: nopl <sig>. */	\
+		".byte 0x0f, 0x1f, 0x05\n\t"				\
+		".long " __rseq_str(RSEQ_SIG) "\n\t"			\
+		__rseq_str(label) ":\n\t"				\
+		teardown						\
+		"jmp %l[" __rseq_str(abort_label) "]\n\t"		\
+		".popsection\n\t"
+
+#define RSEQ_ASM_DEFINE_CMPFAIL(label, teardown, cmpfail_label)		\
+		".pushsection __rseq_failure, \"ax\"\n\t"		\
+		__rseq_str(label) ":\n\t"				\
+		teardown						\
+		"jmp %l[" __rseq_str(cmpfail_label) "]\n\t"		\
+		".popsection\n\t"
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_storev(intptr_t *v, intptr_t expect, intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"cmpl %[v], %[expect]\n\t"
+		"jnz %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"cmpl %[v], %[expect]\n\t"
+		"jnz %l[error2]\n\t"
+#endif
+		/* final store */
+		"movl %[newv], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(5)
+		RSEQ_ASM_DEFINE_ABORT(4, "", abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv)
+		: "memory", "cc", "eax"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+/*
+ * Compare @v against @expectnot. When it does _not_ match, load @v
+ * into @load, and store the content of *@v + voffp into @v.
+ */
+static inline __attribute__((always_inline))
+int rseq_cmpnev_storeoffp_load(intptr_t *v, intptr_t expectnot,
+			       off_t voffp, intptr_t *load, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"movl %[v], %%ebx\n\t"
+		"cmpl %%ebx, %[expectnot]\n\t"
+		"je %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"movl %[v], %%ebx\n\t"
+		"cmpl %%ebx, %[expectnot]\n\t"
+		"je %l[error2]\n\t"
+#endif
+		"movl %%ebx, %[load]\n\t"
+		"addl %[voffp], %%ebx\n\t"
+		"movl (%%ebx), %%ebx\n\t"
+		/* final store */
+		"movl %%ebx, %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(5)
+		RSEQ_ASM_DEFINE_ABORT(4, "", abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expectnot]		"r" (expectnot),
+		  [voffp]		"ir" (voffp),
+		  [load]		"m" (*load)
+		: "memory", "cc", "eax", "ebx"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_addv(intptr_t *v, intptr_t count, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+#endif
+		/* final store */
+		"addl %[count], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(4)
+		RSEQ_ASM_DEFINE_ABORT(4, "", abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [count]		"ir" (count)
+		: "memory", "cc", "eax"
+		  RSEQ_INJECT_CLOBBER
+		: abort
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trystorev_storev(intptr_t *v, intptr_t expect,
+				 intptr_t *v2, intptr_t newv2,
+				 intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"cmpl %[v], %[expect]\n\t"
+		"jnz %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"cmpl %[v], %[expect]\n\t"
+		"jnz %l[error2]\n\t"
+#endif
+		/* try store */
+		"movl %[newv2], %%eax\n\t"
+		"movl %%eax, %[v2]\n\t"
+		RSEQ_INJECT_ASM(5)
+		/* final store */
+		"movl %[newv], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(6)
+		RSEQ_ASM_DEFINE_ABORT(4, "", abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* try store input */
+		  [v2]			"m" (*v2),
+		  [newv2]		"m" (newv2),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"r" (newv)
+		: "memory", "cc", "eax"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trystorev_storev_release(intptr_t *v, intptr_t expect,
+					 intptr_t *v2, intptr_t newv2,
+					 intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"movl %[expect], %%eax\n\t"
+		"cmpl %[v], %%eax\n\t"
+		"jnz %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"movl %[expect], %%eax\n\t"
+		"cmpl %[v], %%eax\n\t"
+		"jnz %l[error2]\n\t"
+#endif
+		/* try store */
+		"movl %[newv2], %[v2]\n\t"
+		RSEQ_INJECT_ASM(5)
+		"lock; addl $0,-128(%%esp)\n\t"
+		/* final store */
+		"movl %[newv], %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(6)
+		RSEQ_ASM_DEFINE_ABORT(4, "", abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* try store input */
+		  [v2]			"m" (*v2),
+		  [newv2]		"r" (newv2),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"m" (expect),
+		  [newv]		"r" (newv)
+		: "memory", "cc", "eax"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+
+}
+
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_cmpeqv_storev(intptr_t *v, intptr_t expect,
+			      intptr_t *v2, intptr_t expect2,
+			      intptr_t newv, int cpu)
+{
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"cmpl %[v], %[expect]\n\t"
+		"jnz %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(4)
+		"cmpl %[expect2], %[v2]\n\t"
+		"jnz %l[cmpfail]\n\t"
+		RSEQ_INJECT_ASM(5)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, %l[error1])
+		"cmpl %[v], %[expect]\n\t"
+		"jnz %l[error2]\n\t"
+		"cmpl %[expect2], %[v2]\n\t"
+		"jnz %l[error3]\n\t"
+#endif
+		"movl %[newv], %%eax\n\t"
+		/* final store */
+		"movl %%eax, %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(6)
+		RSEQ_ASM_DEFINE_ABORT(4, "", abort)
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* cmp2 input */
+		  [v2]			"m" (*v2),
+		  [expect2]		"r" (expect2),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"r" (expect),
+		  [newv]		"m" (newv)
+		: "memory", "cc", "eax"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2, error3
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("1st expected value comparison failed");
+error3:
+	rseq_bug("2nd expected value comparison failed");
+#endif
+}
+
+/* TODO: implement a faster memcpy. */
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trymemcpy_storev(intptr_t *v, intptr_t expect,
+				 void *dst, void *src, size_t len,
+				 intptr_t newv, int cpu)
+{
+	uint32_t rseq_scratch[3];
+
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		"movl %[src], %[rseq_scratch0]\n\t"
+		"movl %[dst], %[rseq_scratch1]\n\t"
+		"movl %[len], %[rseq_scratch2]\n\t"
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"movl %[expect], %%eax\n\t"
+		"cmpl %%eax, %[v]\n\t"
+		"jnz 5f\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 6f)
+		"movl %[expect], %%eax\n\t"
+		"cmpl %%eax, %[v]\n\t"
+		"jnz 7f\n\t"
+#endif
+		/* try memcpy */
+		"test %[len], %[len]\n\t" \
+		"jz 333f\n\t" \
+		"222:\n\t" \
+		"movb (%[src]), %%al\n\t" \
+		"movb %%al, (%[dst])\n\t" \
+		"inc %[src]\n\t" \
+		"inc %[dst]\n\t" \
+		"dec %[len]\n\t" \
+		"jnz 222b\n\t" \
+		"333:\n\t" \
+		RSEQ_INJECT_ASM(5)
+		"movl %[newv], %%eax\n\t"
+		/* final store */
+		"movl %%eax, %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(6)
+		/* teardown */
+		"movl %[rseq_scratch2], %[len]\n\t"
+		"movl %[rseq_scratch1], %[dst]\n\t"
+		"movl %[rseq_scratch0], %[src]\n\t"
+		RSEQ_ASM_DEFINE_ABORT(4,
+			"movl %[rseq_scratch2], %[len]\n\t"
+			"movl %[rseq_scratch1], %[dst]\n\t"
+			"movl %[rseq_scratch0], %[src]\n\t",
+			abort)
+		RSEQ_ASM_DEFINE_CMPFAIL(5,
+			"movl %[rseq_scratch2], %[len]\n\t"
+			"movl %[rseq_scratch1], %[dst]\n\t"
+			"movl %[rseq_scratch0], %[src]\n\t",
+			cmpfail)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_DEFINE_CMPFAIL(6,
+			"movl %[rseq_scratch2], %[len]\n\t"
+			"movl %[rseq_scratch1], %[dst]\n\t"
+			"movl %[rseq_scratch0], %[src]\n\t",
+			error1)
+		RSEQ_ASM_DEFINE_CMPFAIL(7,
+			"movl %[rseq_scratch2], %[len]\n\t"
+			"movl %[rseq_scratch1], %[dst]\n\t"
+			"movl %[rseq_scratch0], %[src]\n\t",
+			error2)
+#endif
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"m" (expect),
+		  [newv]		"m" (newv),
+		  /* try memcpy input */
+		  [dst]			"r" (dst),
+		  [src]			"r" (src),
+		  [len]			"r" (len),
+		  [rseq_scratch0]	"m" (rseq_scratch[0]),
+		  [rseq_scratch1]	"m" (rseq_scratch[1]),
+		  [rseq_scratch2]	"m" (rseq_scratch[2])
+		: "memory", "cc", "eax"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+/* TODO: implement a faster memcpy. */
+static inline __attribute__((always_inline))
+int rseq_cmpeqv_trymemcpy_storev_release(intptr_t *v, intptr_t expect,
+					 void *dst, void *src, size_t len,
+					 intptr_t newv, int cpu)
+{
+	uint32_t rseq_scratch[3];
+
+	RSEQ_INJECT_C(9)
+
+	__asm__ __volatile__ goto (
+		RSEQ_ASM_DEFINE_TABLE(3, 1f, 2f, 4f) /* start, commit, abort */
+		"movl %[src], %[rseq_scratch0]\n\t"
+		"movl %[dst], %[rseq_scratch1]\n\t"
+		"movl %[len], %[rseq_scratch2]\n\t"
+		/* Start rseq by storing table entry pointer into rseq_cs. */
+		RSEQ_ASM_STORE_RSEQ_CS(1, 3b, rseq_cs)
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
+		RSEQ_INJECT_ASM(3)
+		"movl %[expect], %%eax\n\t"
+		"cmpl %%eax, %[v]\n\t"
+		"jnz 5f\n\t"
+		RSEQ_INJECT_ASM(4)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 6f)
+		"movl %[expect], %%eax\n\t"
+		"cmpl %%eax, %[v]\n\t"
+		"jnz 7f\n\t"
+#endif
+		/* try memcpy */
+		"test %[len], %[len]\n\t" \
+		"jz 333f\n\t" \
+		"222:\n\t" \
+		"movb (%[src]), %%al\n\t" \
+		"movb %%al, (%[dst])\n\t" \
+		"inc %[src]\n\t" \
+		"inc %[dst]\n\t" \
+		"dec %[len]\n\t" \
+		"jnz 222b\n\t" \
+		"333:\n\t" \
+		RSEQ_INJECT_ASM(5)
+		"lock; addl $0,-128(%%esp)\n\t"
+		"movl %[newv], %%eax\n\t"
+		/* final store */
+		"movl %%eax, %[v]\n\t"
+		"2:\n\t"
+		RSEQ_INJECT_ASM(6)
+		/* teardown */
+		"movl %[rseq_scratch2], %[len]\n\t"
+		"movl %[rseq_scratch1], %[dst]\n\t"
+		"movl %[rseq_scratch0], %[src]\n\t"
+		RSEQ_ASM_DEFINE_ABORT(4,
+			"movl %[rseq_scratch2], %[len]\n\t"
+			"movl %[rseq_scratch1], %[dst]\n\t"
+			"movl %[rseq_scratch0], %[src]\n\t",
+			abort)
+		RSEQ_ASM_DEFINE_CMPFAIL(5,
+			"movl %[rseq_scratch2], %[len]\n\t"
+			"movl %[rseq_scratch1], %[dst]\n\t"
+			"movl %[rseq_scratch0], %[src]\n\t",
+			cmpfail)
+#ifdef RSEQ_COMPARE_TWICE
+		RSEQ_ASM_DEFINE_CMPFAIL(6,
+			"movl %[rseq_scratch2], %[len]\n\t"
+			"movl %[rseq_scratch1], %[dst]\n\t"
+			"movl %[rseq_scratch0], %[src]\n\t",
+			error1)
+		RSEQ_ASM_DEFINE_CMPFAIL(7,
+			"movl %[rseq_scratch2], %[len]\n\t"
+			"movl %[rseq_scratch1], %[dst]\n\t"
+			"movl %[rseq_scratch0], %[src]\n\t",
+			error2)
+#endif
+		: /* gcc asm goto does not allow outputs */
+		: [cpu_id]		"r" (cpu),
+		  [current_cpu_id]	"m" (__rseq_abi.cpu_id),
+		  [rseq_cs]		"m" (__rseq_abi.rseq_cs),
+		  /* final store input */
+		  [v]			"m" (*v),
+		  [expect]		"m" (expect),
+		  [newv]		"m" (newv),
+		  /* try memcpy input */
+		  [dst]			"r" (dst),
+		  [src]			"r" (src),
+		  [len]			"r" (len),
+		  [rseq_scratch0]	"m" (rseq_scratch[0]),
+		  [rseq_scratch1]	"m" (rseq_scratch[1]),
+		  [rseq_scratch2]	"m" (rseq_scratch[2])
+		: "memory", "cc", "eax"
+		  RSEQ_INJECT_CLOBBER
+		: abort, cmpfail
+#ifdef RSEQ_COMPARE_TWICE
+		  , error1, error2
+#endif
+	);
+	return 0;
+abort:
+	RSEQ_INJECT_FAILED
+	return -1;
+cmpfail:
+	return 1;
+#ifdef RSEQ_COMPARE_TWICE
+error1:
+	rseq_bug("cpu_id comparison failed");
+error2:
+	rseq_bug("expected value comparison failed");
+#endif
+}
+
+#endif /* !RSEQ_SKIP_FASTPATH */
+
+#endif
diff --git a/tools/testing/selftests/rseq/rseq.c b/tools/testing/selftests/rseq/rseq.c
new file mode 100644
index 000000000000..ba65c14ef3ae
--- /dev/null
+++ b/tools/testing/selftests/rseq/rseq.c
@@ -0,0 +1,116 @@
+/*
+ * rseq.c
+ *
+ * Copyright (C) 2016 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; only
+ * version 2.1 of the License.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ */
+
+#define _GNU_SOURCE
+#include <errno.h>
+#include <sched.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <syscall.h>
+#include <assert.h>
+#include <signal.h>
+
+#include "rseq.h"
+
+#define ARRAY_SIZE(arr)	(sizeof(arr) / sizeof((arr)[0]))
+
+__attribute__((tls_model("initial-exec"))) __thread
+volatile struct rseq __rseq_abi = {
+	.cpu_id = RSEQ_CPU_ID_UNINITIALIZED,
+};
+
+static __attribute__((tls_model("initial-exec"))) __thread
+volatile int refcount;
+
+static void signal_off_save(sigset_t *oldset)
+{
+	sigset_t set;
+	int ret;
+
+	sigfillset(&set);
+	ret = pthread_sigmask(SIG_BLOCK, &set, oldset);
+	if (ret)
+		abort();
+}
+
+static void signal_restore(sigset_t oldset)
+{
+	int ret;
+
+	ret = pthread_sigmask(SIG_SETMASK, &oldset, NULL);
+	if (ret)
+		abort();
+}
+
+static int sys_rseq(volatile struct rseq *rseq_abi, uint32_t rseq_len,
+		    int flags, uint32_t sig)
+{
+	return syscall(__NR_rseq, rseq_abi, rseq_len, flags, sig);
+}
+
+int rseq_register_current_thread(void)
+{
+	int rc, ret = 0;
+	sigset_t oldset;
+
+	signal_off_save(&oldset);
+	if (refcount++)
+		goto end;
+	rc = sys_rseq(&__rseq_abi, sizeof(struct rseq), 0, RSEQ_SIG);
+	if (!rc) {
+		assert(rseq_current_cpu_raw() >= 0);
+		goto end;
+	}
+	if (errno != EBUSY)
+		__rseq_abi.cpu_id = -2;
+	ret = -1;
+	refcount--;
+end:
+	signal_restore(oldset);
+	return ret;
+}
+
+int rseq_unregister_current_thread(void)
+{
+	int rc, ret = 0;
+	sigset_t oldset;
+
+	signal_off_save(&oldset);
+	if (--refcount)
+		goto end;
+	rc = sys_rseq(&__rseq_abi, sizeof(struct rseq),
+		      RSEQ_FLAG_UNREGISTER, RSEQ_SIG);
+	if (!rc)
+		goto end;
+	ret = -1;
+end:
+	signal_restore(oldset);
+	return ret;
+}
+
+int32_t rseq_fallback_current_cpu(void)
+{
+	int32_t cpu;
+
+	cpu = sched_getcpu();
+	if (cpu < 0) {
+		perror("sched_getcpu()");
+		abort();
+	}
+	return cpu;
+}
diff --git a/tools/testing/selftests/rseq/rseq.h b/tools/testing/selftests/rseq/rseq.h
new file mode 100644
index 000000000000..d199f0beee1b
--- /dev/null
+++ b/tools/testing/selftests/rseq/rseq.h
@@ -0,0 +1,164 @@
+/*
+ * rseq.h
+ *
+ * (C) Copyright 2016 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef RSEQ_H
+#define RSEQ_H
+
+#include <stdint.h>
+#include <stdbool.h>
+#include <pthread.h>
+#include <signal.h>
+#include <sched.h>
+#include <errno.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <sched.h>
+#include <linux/rseq.h>
+
+/*
+ * Empty code injection macros, override when testing.
+ * It is important to consider that the ASM injection macros need to be
+ * fully reentrant (e.g. do not modify the stack).
+ */
+#ifndef RSEQ_INJECT_ASM
+#define RSEQ_INJECT_ASM(n)
+#endif
+
+#ifndef RSEQ_INJECT_C
+#define RSEQ_INJECT_C(n)
+#endif
+
+#ifndef RSEQ_INJECT_INPUT
+#define RSEQ_INJECT_INPUT
+#endif
+
+#ifndef RSEQ_INJECT_CLOBBER
+#define RSEQ_INJECT_CLOBBER
+#endif
+
+#ifndef RSEQ_INJECT_FAILED
+#define RSEQ_INJECT_FAILED
+#endif
+
+extern __thread volatile struct rseq __rseq_abi;
+
+#define rseq_likely(x)		__builtin_expect(!!(x), 1)
+#define rseq_unlikely(x)	__builtin_expect(!!(x), 0)
+#define rseq_barrier()		__asm__ __volatile__("" : : : "memory")
+
+#define RSEQ_ACCESS_ONCE(x)	(*(__volatile__  __typeof__(x) *)&(x))
+#define RSEQ_WRITE_ONCE(x, v)	__extension__ ({ RSEQ_ACCESS_ONCE(x) = (v); })
+#define RSEQ_READ_ONCE(x)	RSEQ_ACCESS_ONCE(x)
+
+#define __rseq_str_1(x)	#x
+#define __rseq_str(x)		__rseq_str_1(x)
+
+#define rseq_log(fmt, args...)						       \
+	fprintf(stderr, fmt "(in %s() at " __FILE__ ":" __rseq_str(__LINE__)"\n", \
+		## args, __func__)
+
+#define rseq_bug(fmt, args...)		\
+	do {				\
+		rseq_log(fmt, ##args);	\
+		abort();		\
+	} while (0)
+
+#if defined(__x86_64__) || defined(__i386__)
+#include <rseq-x86.h>
+#elif defined(__ARMEL__)
+#include <rseq-arm.h>
+#elif defined(__PPC__)
+#include <rseq-ppc.h>
+#else
+#error unsupported target
+#endif
+
+/*
+ * Register rseq for the current thread. This needs to be called once
+ * by any thread which uses restartable sequences, before they start
+ * using restartable sequences, to ensure restartable sequences
+ * succeed. A restartable sequence executed from a non-registered
+ * thread will always fail.
+ */
+int rseq_register_current_thread(void);
+
+/*
+ * Unregister rseq for current thread.
+ */
+int rseq_unregister_current_thread(void);
+
+/*
+ * Restartable sequence fallback for reading the current CPU number.
+ */
+int32_t rseq_fallback_current_cpu(void);
+
+/*
+ * Values returned can be either the current CPU number, -1 (rseq is
+ * uninitialized), or -2 (rseq initialization has failed).
+ */
+static inline int32_t rseq_current_cpu_raw(void)
+{
+	return RSEQ_ACCESS_ONCE(__rseq_abi.cpu_id);
+}
+
+/*
+ * Returns a possible CPU number, which is typically the current CPU.
+ * The returned CPU number can be used to prepare for an rseq critical
+ * section, which will confirm whether the cpu number is indeed the
+ * current one, and whether rseq is initialized.
+ *
+ * The CPU number returned by rseq_cpu_start should always be validated
+ * by passing it to a rseq asm sequence, or by comparing it to the
+ * return value of rseq_current_cpu_raw() if the rseq asm sequence
+ * does not need to be invoked.
+ */
+static inline uint32_t rseq_cpu_start(void)
+{
+	return RSEQ_ACCESS_ONCE(__rseq_abi.cpu_id_start);
+}
+
+static inline uint32_t rseq_current_cpu(void)
+{
+	int32_t cpu;
+
+	cpu = rseq_current_cpu_raw();
+	if (rseq_unlikely(cpu < 0))
+		cpu = rseq_fallback_current_cpu();
+	return cpu;
+}
+
+/*
+ * rseq_prepare_unload() should be invoked by each thread using rseq_finish*()
+ * at least once between their last rseq_finish*() and library unload of the
+ * library defining the rseq critical section (struct rseq_cs). This also
+ * applies to use of rseq in code generated by JIT: rseq_prepare_unload()
+ * should be invoked at least once by each thread using rseq_finish*() before
+ * reclaim of the memory holding the struct rseq_cs.
+ */
+static inline void rseq_prepare_unload(void)
+{
+	__rseq_abi.rseq_cs = 0;
+}
+
+#endif  /* RSEQ_H_ */
-- 
2.11.0

[RFC PATCH for 4.16 17/21] rseq: selftests: Provide percpu_op API

[Mathieu Desnoyers]

Introduce percpu-op.h API. It uses rseq internally as fast-path if
invoked from the right CPU, else cpu_opv as slow-path if called
from the wrong CPU or if rseq fails.

This allows acting on per-cpu data from various CPUs transparently from
user-space: cpu_opv will take care of migrating the thread to the
requested CPU. Use-cases such as rebalancing memory across per-cpu
memory pools, or migrating tasks for a user-space scheduler, are thus
facilitated. This also handles debugger single-stepping.

The use from userspace is, e.g. for a counter increment:

    int cpu, ret;

    cpu = rseq_cpu_start();
    ret = percpu_addv(&data->c[cpu].count, 1, cpu);
    if (unlikely(ret)) {
         perror("percpu_addv");
         return -1;
    }
    return 0;

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
CC: Shuah Khan <shuahkh-JPH+aEBZ4P+UEJcrhfAQsw@public.gmane.org>
CC: Russell King <linux-lFZ/pmaqli7XmaaqVzeoHQ@public.gmane.org>
CC: Catalin Marinas <catalin.marinas-5wv7dgnIgG8@public.gmane.org>
CC: Will Deacon <will.deacon-5wv7dgnIgG8@public.gmane.org>
CC: Thomas Gleixner <tglx-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org>
CC: Paul Turner <pjt-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>
CC: Andrew Hunter <ahh-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>
CC: Peter Zijlstra <peterz-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org>
CC: Andy Lutomirski <luto-kltTT9wpgjJwATOyAt5JVQ@public.gmane.org>
CC: Andi Kleen <andi-Vw/NltI1exuRpAAqCnN02g@public.gmane.org>
CC: Dave Watson <davejwatson-b10kYP2dOMg@public.gmane.org>
CC: Chris Lameter <cl-vYTEC60ixJUAvxtiuMwx3w@public.gmane.org>
CC: Ingo Molnar <mingo-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
CC: "H. Peter Anvin" <hpa-YMNOUZJC4hwAvxtiuMwx3w@public.gmane.org>
CC: Ben Maurer <bmaurer-b10kYP2dOMg@public.gmane.org>
CC: Steven Rostedt <rostedt-nx8X9YLhiw1AfugRpC6u6w@public.gmane.org>
CC: "Paul E. McKenney" <paulmck-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
CC: Josh Triplett <josh-iaAMLnmF4UmaiuxdJuQwMA@public.gmane.org>
CC: Linus Torvalds <torvalds-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>
CC: Andrew Morton <akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>
CC: Boqun Feng <boqun.feng-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
CC: linux-kselftest-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
CC: linux-api-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
---
 tools/testing/selftests/rseq/percpu-op.h | 163 +++++++++++++++++++++++++++++++
 1 file changed, 163 insertions(+)
 create mode 100644 tools/testing/selftests/rseq/percpu-op.h

diff --git a/tools/testing/selftests/rseq/percpu-op.h b/tools/testing/selftests/rseq/percpu-op.h
new file mode 100644
index 000000000000..c17d165438a6
--- /dev/null
+++ b/tools/testing/selftests/rseq/percpu-op.h
@@ -0,0 +1,163 @@
+/*
+ * percpu-op.h
+ *
+ * (C) Copyright 2017 - Mathieu Desnoyers <mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef PERCPU_OP_H
+#define PERCPU_OP_H
+
+#include <stdint.h>
+#include <stdbool.h>
+#include <errno.h>
+#include <stdlib.h>
+#include "rseq.h"
+#include "cpu-op.h"
+
+static inline __attribute__((always_inline))
+int percpu_cmpeqv_storev(intptr_t *v, intptr_t expect, intptr_t newv,
+			 int cpu)
+{
+	int ret;
+
+	ret = rseq_cmpeqv_storev(v, expect, newv, cpu);
+	if (rseq_unlikely(ret)) {
+		if (ret > 0)
+			return ret;
+		return cpu_op_cmpeqv_storev(v, expect, newv, cpu);
+	}
+	return 0;
+}
+
+static inline __attribute__((always_inline))
+int percpu_cmpnev_storeoffp_load(intptr_t *v, intptr_t expectnot,
+			       off_t voffp, intptr_t *load, int cpu)
+{
+	int ret;
+
+	ret = rseq_cmpnev_storeoffp_load(v, expectnot, voffp, load, cpu);
+	if (rseq_unlikely(ret)) {
+		if (ret > 0)
+			return ret;
+		return cpu_op_cmpnev_storeoffp_load(v, expectnot, voffp,
+						    load, cpu);
+	}
+	return 0;
+}
+
+static inline __attribute__((always_inline))
+int percpu_addv(intptr_t *v, intptr_t count, int cpu)
+{
+	if (rseq_unlikely(rseq_addv(v, count, cpu)))
+		return cpu_op_addv(v, count, cpu);
+	return 0;
+}
+
+static inline __attribute__((always_inline))
+int percpu_cmpeqv_storev_storev(intptr_t *v, intptr_t expect,
+				intptr_t *v2, intptr_t newv2,
+				intptr_t newv, int cpu)
+{
+	int ret;
+
+	ret = rseq_cmpeqv_trystorev_storev(v, expect, v2, newv2,
+					   newv, cpu);
+	if (rseq_unlikely(ret)) {
+		if (ret > 0)
+			return ret;
+		return cpu_op_cmpeqv_storev_storev(v, expect, v2, newv2,
+						   newv, cpu);
+	}
+	return 0;
+}
+
+static inline __attribute__((always_inline))
+int percpu_cmpeqv_storev_storev_release(intptr_t *v, intptr_t expect,
+					intptr_t *v2, intptr_t newv2,
+					intptr_t newv, int cpu)
+{
+	int ret;
+
+	ret = rseq_cmpeqv_trystorev_storev_release(v, expect, v2, newv2,
+						   newv, cpu);
+	if (rseq_unlikely(ret)) {
+		if (ret > 0)
+			return ret;
+		return cpu_op_cmpeqv_storev_mb_storev(v, expect, v2, newv2,
+						      newv, cpu);
+	}
+	return 0;
+}
+
+static inline __attribute__((always_inline))
+int percpu_cmpeqv_cmpeqv_storev(intptr_t *v, intptr_t expect,
+				intptr_t *v2, intptr_t expect2,
+				intptr_t newv, int cpu)
+{
+	int ret;
+
+	ret = rseq_cmpeqv_cmpeqv_storev(v, expect, v2, expect2, newv, cpu);
+	if (rseq_unlikely(ret)) {
+		if (ret > 0)
+			return ret;
+		return cpu_op_cmpeqv_cmpeqv_storev(v, expect, v2, expect2,
+						   newv, cpu);
+	}
+	return 0;
+}
+
+static inline __attribute__((always_inline))
+int percpu_cmpeqv_memcpy_storev(intptr_t *v, intptr_t expect,
+				void *dst, void *src, size_t len,
+				intptr_t newv, int cpu)
+{
+	int ret;
+
+	ret = rseq_cmpeqv_trymemcpy_storev(v, expect, dst, src, len,
+					   newv, cpu);
+	if (rseq_unlikely(ret)) {
+		if (ret > 0)
+			return ret;
+		return cpu_op_cmpeqv_memcpy_storev(v, expect, dst, src, len,
+						   newv, cpu);
+	}
+	return 0;
+}
+
+static inline __attribute__((always_inline))
+int percpu_cmpeqv_memcpy_storev_release(intptr_t *v, intptr_t expect,
+					void *dst, void *src, size_t len,
+					intptr_t newv, int cpu)
+{
+	int ret;
+
+	ret = rseq_cmpeqv_trymemcpy_storev_release(v, expect, dst, src, len,
+						   newv, cpu);
+	if (rseq_unlikely(ret)) {
+		if (ret > 0)
+			return ret;
+		return cpu_op_cmpeqv_memcpy_mb_storev(v, expect, dst, src, len,
+						      newv, cpu);
+	}
+	return 0;
+}
+
+#endif  /* PERCPU_OP_H_ */
-- 
2.11.0

[RFC PATCH for 4.16 18/21] rseq: selftests: Provide basic test

[Mathieu Desnoyers]

"basic_test" only asserts that RSEQ works moderately correctly. E.g.
that the CPUID pointer works.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Shuah Khan <shuahkh@osg.samsung.com>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Paul Turner <pjt@google.com>
CC: Andrew Hunter <ahh@google.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-kselftest@vger.kernel.org
CC: linux-api@vger.kernel.org
---
 tools/testing/selftests/rseq/basic_test.c | 55 +++++++++++++++++++++++++++++++
 1 file changed, 55 insertions(+)
 create mode 100644 tools/testing/selftests/rseq/basic_test.c

diff --git a/tools/testing/selftests/rseq/basic_test.c b/tools/testing/selftests/rseq/basic_test.c
new file mode 100644
index 000000000000..e2086b3885d7
--- /dev/null
+++ b/tools/testing/selftests/rseq/basic_test.c
@@ -0,0 +1,55 @@
+/*
+ * Basic test coverage for critical regions and rseq_current_cpu().
+ */
+
+#define _GNU_SOURCE
+#include <assert.h>
+#include <sched.h>
+#include <signal.h>
+#include <stdio.h>
+#include <string.h>
+#include <sys/time.h>
+
+#include "rseq.h"
+
+void test_cpu_pointer(void)
+{
+	cpu_set_t affinity, test_affinity;
+	int i;
+
+	sched_getaffinity(0, sizeof(affinity), &affinity);
+	CPU_ZERO(&test_affinity);
+	for (i = 0; i < CPU_SETSIZE; i++) {
+		if (CPU_ISSET(i, &affinity)) {
+			CPU_SET(i, &test_affinity);
+			sched_setaffinity(0, sizeof(test_affinity),
+					&test_affinity);
+			assert(sched_getcpu() == i);
+			assert(rseq_current_cpu() == i);
+			assert(rseq_current_cpu_raw() == i);
+			assert(rseq_cpu_start() == i);
+			CPU_CLR(i, &test_affinity);
+		}
+	}
+	sched_setaffinity(0, sizeof(affinity), &affinity);
+}
+
+int main(int argc, char **argv)
+{
+	if (rseq_register_current_thread()) {
+		fprintf(stderr, "Error: rseq_register_current_thread(...) failed(%d): %s\n",
+			errno, strerror(errno));
+		goto init_thread_error;
+	}
+	printf("testing current cpu\n");
+	test_cpu_pointer();
+	if (rseq_unregister_current_thread()) {
+		fprintf(stderr, "Error: rseq_unregister_current_thread(...) failed(%d): %s\n",
+			errno, strerror(errno));
+		goto init_thread_error;
+	}
+	return 0;
+
+init_thread_error:
+	return -1;
+}
-- 
2.11.0

[RFC PATCH for 4.16 19/21] rseq: selftests: Provide basic percpu ops test

[Mathieu Desnoyers]

"basic_percpu_ops_test" is a slightly more "realistic" variant,
implementing a few simple per-cpu operations and testing their
correctness.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Shuah Khan <shuahkh@osg.samsung.com>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Paul Turner <pjt@google.com>
CC: Andrew Hunter <ahh@google.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-kselftest@vger.kernel.org
CC: linux-api@vger.kernel.org
---
 .../testing/selftests/rseq/basic_percpu_ops_test.c | 296 +++++++++++++++++++++
 1 file changed, 296 insertions(+)
 create mode 100644 tools/testing/selftests/rseq/basic_percpu_ops_test.c

diff --git a/tools/testing/selftests/rseq/basic_percpu_ops_test.c b/tools/testing/selftests/rseq/basic_percpu_ops_test.c
new file mode 100644
index 000000000000..e585bba0bf8d
--- /dev/null
+++ b/tools/testing/selftests/rseq/basic_percpu_ops_test.c
@@ -0,0 +1,296 @@
+#define _GNU_SOURCE
+#include <assert.h>
+#include <pthread.h>
+#include <sched.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stddef.h>
+
+#include "percpu-op.h"
+
+#define ARRAY_SIZE(arr)	(sizeof(arr) / sizeof((arr)[0]))
+
+struct percpu_lock_entry {
+	intptr_t v;
+} __attribute__((aligned(128)));
+
+struct percpu_lock {
+	struct percpu_lock_entry c[CPU_SETSIZE];
+};
+
+struct test_data_entry {
+	intptr_t count;
+} __attribute__((aligned(128)));
+
+struct spinlock_test_data {
+	struct percpu_lock lock;
+	struct test_data_entry c[CPU_SETSIZE];
+	int reps;
+};
+
+struct percpu_list_node {
+	intptr_t data;
+	struct percpu_list_node *next;
+};
+
+struct percpu_list_entry {
+	struct percpu_list_node *head;
+} __attribute__((aligned(128)));
+
+struct percpu_list {
+	struct percpu_list_entry c[CPU_SETSIZE];
+};
+
+/* A simple percpu spinlock.  Returns the cpu lock was acquired on. */
+int rseq_percpu_lock(struct percpu_lock *lock)
+{
+	int cpu;
+
+	for (;;) {
+		int ret;
+
+		cpu = rseq_cpu_start();
+		ret = percpu_cmpeqv_storev(&lock->c[cpu].v,
+					   0, 1, cpu);
+		if (rseq_likely(!ret))
+			break;
+		if (rseq_unlikely(ret < 0)) {
+			perror("cpu_opv");
+			abort();
+		}
+		/* Retry if comparison fails. */
+	}
+	/*
+	 * Acquire semantic when taking lock after control dependency.
+	 * Matches rseq_smp_store_release().
+	 */
+	rseq_smp_acquire__after_ctrl_dep();
+	return cpu;
+}
+
+void rseq_percpu_unlock(struct percpu_lock *lock, int cpu)
+{
+	assert(lock->c[cpu].v == 1);
+	/*
+	 * Release lock, with release semantic. Matches
+	 * rseq_smp_acquire__after_ctrl_dep().
+	 */
+	rseq_smp_store_release(&lock->c[cpu].v, 0);
+}
+
+void *test_percpu_spinlock_thread(void *arg)
+{
+	struct spinlock_test_data *data = arg;
+	int i, cpu;
+
+	if (rseq_register_current_thread()) {
+		fprintf(stderr, "Error: rseq_register_current_thread(...) failed(%d): %s\n",
+			errno, strerror(errno));
+		abort();
+	}
+	for (i = 0; i < data->reps; i++) {
+		cpu = rseq_percpu_lock(&data->lock);
+		data->c[cpu].count++;
+		rseq_percpu_unlock(&data->lock, cpu);
+	}
+	if (rseq_unregister_current_thread()) {
+		fprintf(stderr, "Error: rseq_unregister_current_thread(...) failed(%d): %s\n",
+			errno, strerror(errno));
+		abort();
+	}
+
+	return NULL;
+}
+
+/*
+ * A simple test which implements a sharded counter using a per-cpu
+ * lock.  Obviously real applications might prefer to simply use a
+ * per-cpu increment; however, this is reasonable for a test and the
+ * lock can be extended to synchronize more complicated operations.
+ */
+void test_percpu_spinlock(void)
+{
+	const int num_threads = 200;
+	int i;
+	uint64_t sum;
+	pthread_t test_threads[num_threads];
+	struct spinlock_test_data data;
+
+	memset(&data, 0, sizeof(data));
+	data.reps = 5000;
+
+	for (i = 0; i < num_threads; i++)
+		pthread_create(&test_threads[i], NULL,
+			       test_percpu_spinlock_thread, &data);
+
+	for (i = 0; i < num_threads; i++)
+		pthread_join(test_threads[i], NULL);
+
+	sum = 0;
+	for (i = 0; i < CPU_SETSIZE; i++)
+		sum += data.c[i].count;
+
+	assert(sum == (uint64_t)data.reps * num_threads);
+}
+
+int percpu_list_push(struct percpu_list *list, struct percpu_list_node *node,
+		     int cpu)
+{
+	for (;;) {
+		intptr_t *targetptr, newval, expect;
+		int ret;
+
+		/* Load list->c[cpu].head with single-copy atomicity. */
+		expect = (intptr_t)RSEQ_READ_ONCE(list->c[cpu].head);
+		newval = (intptr_t)node;
+		targetptr = (intptr_t *)&list->c[cpu].head;
+		node->next = (struct percpu_list_node *)expect;
+		ret = percpu_cmpeqv_storev(targetptr, expect, newval, cpu);
+		if (rseq_likely(!ret))
+			break;
+		if (rseq_unlikely(ret < 0)) {
+			perror("cpu_opv");
+			abort();
+		}
+		/* Retry if comparison fails. */
+	}
+	return cpu;
+}
+
+/*
+ * Unlike a traditional lock-less linked list; the availability of a
+ * rseq primitive allows us to implement pop without concerns over
+ * ABA-type races.
+ */
+struct percpu_list_node *percpu_list_pop(struct percpu_list *list,
+					 int cpu)
+{
+	struct percpu_list_node *head;
+	intptr_t *targetptr, expectnot, *load;
+	off_t offset;
+	int ret;
+
+	targetptr = (intptr_t *)&list->c[cpu].head;
+	expectnot = (intptr_t)NULL;
+	offset = offsetof(struct percpu_list_node, next);
+	load = (intptr_t *)&head;
+	ret = percpu_cmpnev_storeoffp_load(targetptr, expectnot,
+					   offset, load, cpu);
+	if (rseq_unlikely(ret < 0)) {
+		perror("cpu_opv");
+		abort();
+	}
+	if (ret > 0)
+		return NULL;
+	return head;
+}
+
+void *test_percpu_list_thread(void *arg)
+{
+	int i;
+	struct percpu_list *list = (struct percpu_list *)arg;
+
+	if (rseq_register_current_thread()) {
+		fprintf(stderr, "Error: rseq_register_current_thread(...) failed(%d): %s\n",
+			errno, strerror(errno));
+		abort();
+	}
+
+	for (i = 0; i < 100000; i++) {
+		struct percpu_list_node *node;
+
+		node = percpu_list_pop(list, rseq_cpu_start());
+		sched_yield();  /* encourage shuffling */
+		if (node)
+			percpu_list_push(list, node, rseq_cpu_start());
+	}
+
+	if (rseq_unregister_current_thread()) {
+		fprintf(stderr, "Error: rseq_unregister_current_thread(...) failed(%d): %s\n",
+			errno, strerror(errno));
+		abort();
+	}
+
+	return NULL;
+}
+
+/* Simultaneous modification to a per-cpu linked list from many threads.  */
+void test_percpu_list(void)
+{
+	int i, j;
+	uint64_t sum = 0, expected_sum = 0;
+	struct percpu_list list;
+	pthread_t test_threads[200];
+	cpu_set_t allowed_cpus;
+
+	memset(&list, 0, sizeof(list));
+
+	/* Generate list entries for every usable cpu. */
+	sched_getaffinity(0, sizeof(allowed_cpus), &allowed_cpus);
+	for (i = 0; i < CPU_SETSIZE; i++) {
+		if (!CPU_ISSET(i, &allowed_cpus))
+			continue;
+		for (j = 1; j <= 100; j++) {
+			struct percpu_list_node *node;
+
+			expected_sum += j;
+
+			node = malloc(sizeof(*node));
+			assert(node);
+			node->data = j;
+			node->next = list.c[i].head;
+			list.c[i].head = node;
+		}
+	}
+
+	for (i = 0; i < 200; i++)
+		pthread_create(&test_threads[i], NULL,
+		       test_percpu_list_thread, &list);
+
+	for (i = 0; i < 200; i++)
+		pthread_join(test_threads[i], NULL);
+
+	for (i = 0; i < CPU_SETSIZE; i++) {
+		struct percpu_list_node *node;
+
+		if (!CPU_ISSET(i, &allowed_cpus))
+			continue;
+
+		while ((node = percpu_list_pop(&list, i))) {
+			sum += node->data;
+			free(node);
+		}
+	}
+
+	/*
+	 * All entries should now be accounted for (unless some external
+	 * actor is interfering with our allowed affinity while this
+	 * test is running).
+	 */
+	assert(sum == expected_sum);
+}
+
+int main(int argc, char **argv)
+{
+	if (rseq_register_current_thread()) {
+		fprintf(stderr, "Error: rseq_register_current_thread(...) failed(%d): %s\n",
+			errno, strerror(errno));
+		goto error;
+	}
+	printf("spinlock\n");
+	test_percpu_spinlock();
+	printf("percpu_list\n");
+	test_percpu_list();
+	if (rseq_unregister_current_thread()) {
+		fprintf(stderr, "Error: rseq_unregister_current_thread(...) failed(%d): %s\n",
+			errno, strerror(errno));
+		goto error;
+	}
+	return 0;
+
+error:
+	return -1;
+}
+
-- 
2.11.0

[RFC PATCH for 4.16 20/21] rseq: selftests: Provide parametrized tests

[Mathieu Desnoyers]

"param_test" is a parametrizable restartable sequences test. See
the "--help" output for usage.

"param_test_benchmark" is the same as "param_test", but it removes
testing book-keeping code to allow accurate benchmarks.

"param_test_skip_fastpath" is the same as "param_test", but it skips
the rseq fast-path, and only calls the cpu_opv slow path.

"param_test_compare_twice" is the same as "param_test", but it performs
each comparison within rseq critical section twice, thus validating
invariants. If any of the second comparisons fails, an error message
is printed and the test aborts.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Shuah Khan <shuahkh@osg.samsung.com>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Paul Turner <pjt@google.com>
CC: Andrew Hunter <ahh@google.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-kselftest@vger.kernel.org
CC: linux-api@vger.kernel.org
---
 tools/testing/selftests/rseq/param_test.c | 1163 +++++++++++++++++++++++++++++
 1 file changed, 1163 insertions(+)
 create mode 100644 tools/testing/selftests/rseq/param_test.c

diff --git a/tools/testing/selftests/rseq/param_test.c b/tools/testing/selftests/rseq/param_test.c
new file mode 100644
index 000000000000..0a7c05f506be
--- /dev/null
+++ b/tools/testing/selftests/rseq/param_test.c
@@ -0,0 +1,1163 @@
+#define _GNU_SOURCE
+#include <assert.h>
+#include <pthread.h>
+#include <sched.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <syscall.h>
+#include <unistd.h>
+#include <poll.h>
+#include <sys/types.h>
+#include <signal.h>
+#include <errno.h>
+#include <stddef.h>
+
+static inline pid_t gettid(void)
+{
+	return syscall(__NR_gettid);
+}
+
+#define NR_INJECT	9
+static int loop_cnt[NR_INJECT + 1];
+
+static int loop_cnt_1 asm("asm_loop_cnt_1") __attribute__((used));
+static int loop_cnt_2 asm("asm_loop_cnt_2") __attribute__((used));
+static int loop_cnt_3 asm("asm_loop_cnt_3") __attribute__((used));
+static int loop_cnt_4 asm("asm_loop_cnt_4") __attribute__((used));
+static int loop_cnt_5 asm("asm_loop_cnt_5") __attribute__((used));
+static int loop_cnt_6 asm("asm_loop_cnt_6") __attribute__((used));
+
+static int opt_modulo, verbose;
+
+static int opt_yield, opt_signal, opt_sleep,
+		opt_disable_rseq, opt_threads = 200,
+		opt_disable_mod = 0, opt_test = 's', opt_mb = 0;
+
+static long long opt_reps = 5000;
+
+static __thread __attribute__((tls_model("initial-exec")))
+unsigned int signals_delivered;
+
+#ifndef BENCHMARK
+
+static __thread __attribute__((tls_model("initial-exec"), unused))
+unsigned int yield_mod_cnt, nr_abort;
+
+#define printf_verbose(fmt, ...)			\
+	do {						\
+		if (verbose)				\
+			printf(fmt, ## __VA_ARGS__);	\
+	} while (0)
+
+#if defined(__x86_64__) || defined(__i386__)
+
+#define INJECT_ASM_REG	"eax"
+
+#define RSEQ_INJECT_CLOBBER \
+	, INJECT_ASM_REG
+
+#define RSEQ_INJECT_ASM(n) \
+	"mov asm_loop_cnt_" #n ", %%" INJECT_ASM_REG "\n\t" \
+	"test %%" INJECT_ASM_REG ",%%" INJECT_ASM_REG "\n\t" \
+	"jz 333f\n\t" \
+	"222:\n\t" \
+	"dec %%" INJECT_ASM_REG "\n\t" \
+	"jnz 222b\n\t" \
+	"333:\n\t"
+
+#elif defined(__ARMEL__)
+
+#define RSEQ_INJECT_INPUT \
+	, [loop_cnt_1]"m"(loop_cnt[1]) \
+	, [loop_cnt_2]"m"(loop_cnt[2]) \
+	, [loop_cnt_3]"m"(loop_cnt[3]) \
+	, [loop_cnt_4]"m"(loop_cnt[4]) \
+	, [loop_cnt_5]"m"(loop_cnt[5]) \
+	, [loop_cnt_6]"m"(loop_cnt[6])
+
+#define INJECT_ASM_REG	"r4"
+
+#define RSEQ_INJECT_CLOBBER \
+	, INJECT_ASM_REG
+
+#define RSEQ_INJECT_ASM(n) \
+	"ldr " INJECT_ASM_REG ", %[loop_cnt_" #n "]\n\t" \
+	"cmp " INJECT_ASM_REG ", #0\n\t" \
+	"beq 333f\n\t" \
+	"222:\n\t" \
+	"subs " INJECT_ASM_REG ", #1\n\t" \
+	"bne 222b\n\t" \
+	"333:\n\t"
+
+#elif __PPC__
+
+#define RSEQ_INJECT_INPUT \
+	, [loop_cnt_1]"m"(loop_cnt[1]) \
+	, [loop_cnt_2]"m"(loop_cnt[2]) \
+	, [loop_cnt_3]"m"(loop_cnt[3]) \
+	, [loop_cnt_4]"m"(loop_cnt[4]) \
+	, [loop_cnt_5]"m"(loop_cnt[5]) \
+	, [loop_cnt_6]"m"(loop_cnt[6])
+
+#define INJECT_ASM_REG	"r18"
+
+#define RSEQ_INJECT_CLOBBER \
+	, INJECT_ASM_REG
+
+#define RSEQ_INJECT_ASM(n) \
+	"lwz %%" INJECT_ASM_REG ", %[loop_cnt_" #n "]\n\t" \
+	"cmpwi %%" INJECT_ASM_REG ", 0\n\t" \
+	"beq 333f\n\t" \
+	"222:\n\t" \
+	"subic. %%" INJECT_ASM_REG ", %%" INJECT_ASM_REG ", 1\n\t" \
+	"bne 222b\n\t" \
+	"333:\n\t"
+#else
+#error unsupported target
+#endif
+
+#define RSEQ_INJECT_FAILED \
+	nr_abort++;
+
+#define RSEQ_INJECT_C(n) \
+{ \
+	int loc_i, loc_nr_loops = loop_cnt[n]; \
+	\
+	for (loc_i = 0; loc_i < loc_nr_loops; loc_i++) { \
+		rseq_barrier(); \
+	} \
+	if (loc_nr_loops == -1 && opt_modulo) { \
+		if (yield_mod_cnt == opt_modulo - 1) { \
+			if (opt_sleep > 0) \
+				poll(NULL, 0, opt_sleep); \
+			if (opt_yield) \
+				sched_yield(); \
+			if (opt_signal) \
+				raise(SIGUSR1); \
+			yield_mod_cnt = 0; \
+		} else { \
+			yield_mod_cnt++; \
+		} \
+	} \
+}
+
+#else
+
+#define printf_verbose(fmt, ...)
+
+#endif /* BENCHMARK */
+
+#include "percpu-op.h"
+
+struct percpu_lock_entry {
+	intptr_t v;
+} __attribute__((aligned(128)));
+
+struct percpu_lock {
+	struct percpu_lock_entry c[CPU_SETSIZE];
+};
+
+struct test_data_entry {
+	intptr_t count;
+} __attribute__((aligned(128)));
+
+struct spinlock_test_data {
+	struct percpu_lock lock;
+	struct test_data_entry c[CPU_SETSIZE];
+};
+
+struct spinlock_thread_test_data {
+	struct spinlock_test_data *data;
+	long long reps;
+	int reg;
+};
+
+struct inc_test_data {
+	struct test_data_entry c[CPU_SETSIZE];
+};
+
+struct inc_thread_test_data {
+	struct inc_test_data *data;
+	long long reps;
+	int reg;
+};
+
+struct percpu_list_node {
+	intptr_t data;
+	struct percpu_list_node *next;
+};
+
+struct percpu_list_entry {
+	struct percpu_list_node *head;
+} __attribute__((aligned(128)));
+
+struct percpu_list {
+	struct percpu_list_entry c[CPU_SETSIZE];
+};
+
+#define BUFFER_ITEM_PER_CPU	100
+
+struct percpu_buffer_node {
+	intptr_t data;
+};
+
+struct percpu_buffer_entry {
+	intptr_t offset;
+	intptr_t buflen;
+	struct percpu_buffer_node **array;
+} __attribute__((aligned(128)));
+
+struct percpu_buffer {
+	struct percpu_buffer_entry c[CPU_SETSIZE];
+};
+
+#define MEMCPY_BUFFER_ITEM_PER_CPU	100
+
+struct percpu_memcpy_buffer_node {
+	intptr_t data1;
+	uint64_t data2;
+};
+
+struct percpu_memcpy_buffer_entry {
+	intptr_t offset;
+	intptr_t buflen;
+	struct percpu_memcpy_buffer_node *array;
+} __attribute__((aligned(128)));
+
+struct percpu_memcpy_buffer {
+	struct percpu_memcpy_buffer_entry c[CPU_SETSIZE];
+};
+
+/* A simple percpu spinlock. */
+static void rseq_percpu_lock(struct percpu_lock *lock, int cpu)
+{
+	for (;;) {
+		int ret;
+
+		ret = percpu_cmpeqv_storev(&lock->c[cpu].v,
+					   0, 1, cpu);
+		if (rseq_likely(!ret))
+			break;
+		if (rseq_unlikely(ret < 0)) {
+			perror("cpu_opv");
+			abort();
+		}
+		/* Retry if comparison fails. */
+	}
+	/*
+	 * Acquire semantic when taking lock after control dependency.
+	 * Matches rseq_smp_store_release().
+	 */
+	rseq_smp_acquire__after_ctrl_dep();
+}
+
+static void rseq_percpu_unlock(struct percpu_lock *lock, int cpu)
+{
+	assert(lock->c[cpu].v == 1);
+	/*
+	 * Release lock, with release semantic. Matches
+	 * rseq_smp_acquire__after_ctrl_dep().
+	 */
+	rseq_smp_store_release(&lock->c[cpu].v, 0);
+}
+
+void *test_percpu_spinlock_thread(void *arg)
+{
+	struct spinlock_thread_test_data *thread_data = arg;
+	struct spinlock_test_data *data = thread_data->data;
+	long long i, reps;
+
+	if (!opt_disable_rseq && thread_data->reg &&
+	    rseq_register_current_thread())
+		abort();
+	reps = thread_data->reps;
+	for (i = 0; i < reps; i++) {
+		int cpu = rseq_cpu_start();
+
+		rseq_percpu_lock(&data->lock, cpu);
+		data->c[cpu].count++;
+		rseq_percpu_unlock(&data->lock, cpu);
+#ifndef BENCHMARK
+		if (i != 0 && !(i % (reps / 10)))
+			printf_verbose("tid %d: count %lld\n", (int) gettid(), i);
+#endif
+	}
+	printf_verbose("tid %d: number of rseq abort: %d, signals delivered: %u\n",
+		       (int) gettid(), nr_abort, signals_delivered);
+	if (!opt_disable_rseq && thread_data->reg &&
+	    rseq_unregister_current_thread())
+		abort();
+	return NULL;
+}
+
+/*
+ * A simple test which implements a sharded counter using a per-cpu
+ * lock.  Obviously real applications might prefer to simply use a
+ * per-cpu increment; however, this is reasonable for a test and the
+ * lock can be extended to synchronize more complicated operations.
+ */
+void test_percpu_spinlock(void)
+{
+	const int num_threads = opt_threads;
+	int i, ret;
+	uint64_t sum;
+	pthread_t test_threads[num_threads];
+	struct spinlock_test_data data;
+	struct spinlock_thread_test_data thread_data[num_threads];
+
+	memset(&data, 0, sizeof(data));
+	for (i = 0; i < num_threads; i++) {
+		thread_data[i].reps = opt_reps;
+		if (opt_disable_mod <= 0 || (i % opt_disable_mod))
+			thread_data[i].reg = 1;
+		else
+			thread_data[i].reg = 0;
+		thread_data[i].data = &data;
+		ret = pthread_create(&test_threads[i], NULL,
+				     test_percpu_spinlock_thread,
+				     &thread_data[i]);
+		if (ret) {
+			errno = ret;
+			perror("pthread_create");
+			abort();
+		}
+	}
+
+	for (i = 0; i < num_threads; i++) {
+		ret = pthread_join(test_threads[i], NULL);
+		if (ret) {
+			errno = ret;
+			perror("pthread_join");
+			abort();
+		}
+	}
+
+	sum = 0;
+	for (i = 0; i < CPU_SETSIZE; i++)
+		sum += data.c[i].count;
+
+	assert(sum == (uint64_t)opt_reps * num_threads);
+}
+
+void *test_percpu_inc_thread(void *arg)
+{
+	struct inc_thread_test_data *thread_data = arg;
+	struct inc_test_data *data = thread_data->data;
+	long long i, reps;
+
+	if (!opt_disable_rseq && thread_data->reg &&
+	    rseq_register_current_thread())
+		abort();
+	reps = thread_data->reps;
+	for (i = 0; i < reps; i++) {
+		int cpu, ret;
+
+		cpu = rseq_cpu_start();
+		ret = percpu_addv(&data->c[cpu].count, 1, cpu);
+		if (rseq_unlikely(ret)) {
+			perror("cpu_opv");
+			abort();
+		}
+#ifndef BENCHMARK
+		if (i != 0 && !(i % (reps / 10)))
+			printf_verbose("tid %d: count %lld\n", (int) gettid(), i);
+#endif
+	}
+	printf_verbose("tid %d: number of rseq abort: %d, signals delivered: %u\n",
+		       (int) gettid(), nr_abort, signals_delivered);
+	if (!opt_disable_rseq && thread_data->reg &&
+	    rseq_unregister_current_thread())
+		abort();
+	return NULL;
+}
+
+void test_percpu_inc(void)
+{
+	const int num_threads = opt_threads;
+	int i, ret;
+	uint64_t sum;
+	pthread_t test_threads[num_threads];
+	struct inc_test_data data;
+	struct inc_thread_test_data thread_data[num_threads];
+
+	memset(&data, 0, sizeof(data));
+	for (i = 0; i < num_threads; i++) {
+		thread_data[i].reps = opt_reps;
+		if (opt_disable_mod <= 0 || (i % opt_disable_mod))
+			thread_data[i].reg = 1;
+		else
+			thread_data[i].reg = 0;
+		thread_data[i].data = &data;
+		ret = pthread_create(&test_threads[i], NULL,
+				     test_percpu_inc_thread,
+				     &thread_data[i]);
+		if (ret) {
+			errno = ret;
+			perror("pthread_create");
+			abort();
+		}
+	}
+
+	for (i = 0; i < num_threads; i++) {
+		ret = pthread_join(test_threads[i], NULL);
+		if (ret) {
+			errno = ret;
+			perror("pthread_join");
+			abort();
+		}
+	}
+
+	sum = 0;
+	for (i = 0; i < CPU_SETSIZE; i++)
+		sum += data.c[i].count;
+
+	assert(sum == (uint64_t)opt_reps * num_threads);
+}
+
+void percpu_list_push(struct percpu_list *list,
+		      struct percpu_list_node *node,
+		      int cpu)
+{
+	for (;;) {
+		intptr_t *targetptr, newval, expect;
+		int ret;
+
+		/* Load list->c[cpu].head with single-copy atomicity. */
+		expect = (intptr_t)RSEQ_READ_ONCE(list->c[cpu].head);
+		newval = (intptr_t)node;
+		targetptr = (intptr_t *)&list->c[cpu].head;
+		node->next = (struct percpu_list_node *)expect;
+		ret = percpu_cmpeqv_storev(targetptr, expect, newval, cpu);
+		if (rseq_likely(!ret))
+			break;
+		if (rseq_unlikely(ret < 0)) {
+			perror("cpu_opv");
+			abort();
+		}
+		/* Retry if comparison fails. */
+	}
+}
+
+/*
+ * Unlike a traditional lock-less linked list; the availability of a
+ * rseq primitive allows us to implement pop without concerns over
+ * ABA-type races.
+ */
+struct percpu_list_node *percpu_list_pop(struct percpu_list *list,
+					 int cpu)
+{
+	struct percpu_list_node *head;
+	intptr_t *targetptr, expectnot, *load;
+	off_t offset;
+	int ret;
+
+	targetptr = (intptr_t *)&list->c[cpu].head;
+	expectnot = (intptr_t)NULL;
+	offset = offsetof(struct percpu_list_node, next);
+	load = (intptr_t *)&head;
+	ret = percpu_cmpnev_storeoffp_load(targetptr, expectnot,
+					   offset, load, cpu);
+	if (rseq_unlikely(ret < 0)) {
+		perror("cpu_opv");
+		abort();
+	}
+	if (ret > 0)
+		return NULL;
+	return head;
+}
+
+void *test_percpu_list_thread(void *arg)
+{
+	long long i, reps;
+	struct percpu_list *list = (struct percpu_list *)arg;
+
+	if (!opt_disable_rseq && rseq_register_current_thread())
+		abort();
+
+	reps = opt_reps;
+	for (i = 0; i < reps; i++) {
+		struct percpu_list_node *node;
+
+		node = percpu_list_pop(list, rseq_cpu_start());
+		if (opt_yield)
+			sched_yield();  /* encourage shuffling */
+		if (node)
+			percpu_list_push(list, node, rseq_cpu_start());
+	}
+
+	printf_verbose("tid %d: number of rseq abort: %d, signals delivered: %u\n",
+		       (int) gettid(), nr_abort, signals_delivered);
+	if (!opt_disable_rseq && rseq_unregister_current_thread())
+		abort();
+
+	return NULL;
+}
+
+/* Simultaneous modification to a per-cpu linked list from many threads.  */
+void test_percpu_list(void)
+{
+	const int num_threads = opt_threads;
+	int i, j, ret;
+	uint64_t sum = 0, expected_sum = 0;
+	struct percpu_list list;
+	pthread_t test_threads[num_threads];
+	cpu_set_t allowed_cpus;
+
+	memset(&list, 0, sizeof(list));
+
+	/* Generate list entries for every usable cpu. */
+	sched_getaffinity(0, sizeof(allowed_cpus), &allowed_cpus);
+	for (i = 0; i < CPU_SETSIZE; i++) {
+		if (!CPU_ISSET(i, &allowed_cpus))
+			continue;
+		for (j = 1; j <= 100; j++) {
+			struct percpu_list_node *node;
+
+			expected_sum += j;
+
+			node = malloc(sizeof(*node));
+			assert(node);
+			node->data = j;
+			node->next = list.c[i].head;
+			list.c[i].head = node;
+		}
+	}
+
+	for (i = 0; i < num_threads; i++) {
+		ret = pthread_create(&test_threads[i], NULL,
+				     test_percpu_list_thread, &list);
+		if (ret) {
+			errno = ret;
+			perror("pthread_create");
+			abort();
+		}
+	}
+
+	for (i = 0; i < num_threads; i++) {
+		ret = pthread_join(test_threads[i], NULL);
+		if (ret) {
+			errno = ret;
+			perror("pthread_join");
+			abort();
+		}
+	}
+
+	for (i = 0; i < CPU_SETSIZE; i++) {
+		struct percpu_list_node *node;
+
+		if (!CPU_ISSET(i, &allowed_cpus))
+			continue;
+
+		while ((node = percpu_list_pop(&list, i))) {
+			sum += node->data;
+			free(node);
+		}
+	}
+
+	/*
+	 * All entries should now be accounted for (unless some external
+	 * actor is interfering with our allowed affinity while this
+	 * test is running).
+	 */
+	assert(sum == expected_sum);
+}
+
+bool percpu_buffer_push(struct percpu_buffer *buffer,
+			struct percpu_buffer_node *node,
+			int cpu)
+{
+	for (;;) {
+		intptr_t *targetptr_spec, newval_spec;
+		intptr_t *targetptr_final, newval_final;
+		intptr_t offset;
+		int ret;
+
+		offset = RSEQ_READ_ONCE(buffer->c[cpu].offset);
+		if (offset == buffer->c[cpu].buflen)
+			return false;
+		newval_spec = (intptr_t)node;
+		targetptr_spec = (intptr_t *)&buffer->c[cpu].array[offset];
+		newval_final = offset + 1;
+		targetptr_final = &buffer->c[cpu].offset;
+		if (opt_mb)
+			ret = percpu_cmpeqv_storev_storev_release(
+				targetptr_final, offset, targetptr_spec,
+				newval_spec, newval_final, cpu);
+		else
+			ret = percpu_cmpeqv_storev_storev(targetptr_final,
+				offset, targetptr_spec, newval_spec,
+				newval_final, cpu);
+		if (rseq_likely(!ret))
+			break;
+		if (rseq_unlikely(ret < 0)) {
+			perror("cpu_opv");
+			abort();
+		}
+		/* Retry if comparison fails. */
+	}
+	return true;
+}
+
+struct percpu_buffer_node *percpu_buffer_pop(struct percpu_buffer *buffer,
+					     int cpu)
+{
+	struct percpu_buffer_node *head;
+
+	for (;;) {
+		intptr_t *targetptr, newval;
+		intptr_t offset;
+		int ret;
+
+		/* Load offset with single-copy atomicity. */
+		offset = RSEQ_READ_ONCE(buffer->c[cpu].offset);
+		if (offset == 0)
+			return NULL;
+		head = RSEQ_READ_ONCE(buffer->c[cpu].array[offset - 1]);
+		newval = offset - 1;
+		targetptr = (intptr_t *)&buffer->c[cpu].offset;
+		ret = percpu_cmpeqv_cmpeqv_storev(targetptr, offset,
+			(intptr_t *)&buffer->c[cpu].array[offset - 1],
+			(intptr_t)head, newval, cpu);
+		if (rseq_likely(!ret))
+			break;
+		if (rseq_unlikely(ret < 0)) {
+			perror("cpu_opv");
+			abort();
+		}
+		/* Retry if comparison fails. */
+	}
+	return head;
+}
+
+void *test_percpu_buffer_thread(void *arg)
+{
+	long long i, reps;
+	struct percpu_buffer *buffer = (struct percpu_buffer *)arg;
+
+	if (!opt_disable_rseq && rseq_register_current_thread())
+		abort();
+
+	reps = opt_reps;
+	for (i = 0; i < reps; i++) {
+		struct percpu_buffer_node *node;
+
+		node = percpu_buffer_pop(buffer, rseq_cpu_start());
+		if (opt_yield)
+			sched_yield();  /* encourage shuffling */
+		if (node) {
+			if (!percpu_buffer_push(buffer, node,
+						rseq_cpu_start())) {
+				/* Should increase buffer size. */
+				abort();
+			}
+		}
+	}
+
+	printf_verbose("tid %d: number of rseq abort: %d, signals delivered: %u\n",
+		       (int) gettid(), nr_abort, signals_delivered);
+	if (!opt_disable_rseq && rseq_unregister_current_thread())
+		abort();
+
+	return NULL;
+}
+
+/* Simultaneous modification to a per-cpu buffer from many threads.  */
+void test_percpu_buffer(void)
+{
+	const int num_threads = opt_threads;
+	int i, j, ret;
+	uint64_t sum = 0, expected_sum = 0;
+	struct percpu_buffer buffer;
+	pthread_t test_threads[num_threads];
+	cpu_set_t allowed_cpus;
+
+	memset(&buffer, 0, sizeof(buffer));
+
+	/* Generate list entries for every usable cpu. */
+	sched_getaffinity(0, sizeof(allowed_cpus), &allowed_cpus);
+	for (i = 0; i < CPU_SETSIZE; i++) {
+		if (!CPU_ISSET(i, &allowed_cpus))
+			continue;
+		/* Worse-case is every item in same CPU. */
+		buffer.c[i].array =
+			malloc(sizeof(*buffer.c[i].array) * CPU_SETSIZE *
+			       BUFFER_ITEM_PER_CPU);
+		assert(buffer.c[i].array);
+		buffer.c[i].buflen = CPU_SETSIZE * BUFFER_ITEM_PER_CPU;
+		for (j = 1; j <= BUFFER_ITEM_PER_CPU; j++) {
+			struct percpu_buffer_node *node;
+
+			expected_sum += j;
+
+			/*
+			 * We could theoretically put the word-sized
+			 * "data" directly in the buffer. However, we
+			 * want to model objects that would not fit
+			 * within a single word, so allocate an object
+			 * for each node.
+			 */
+			node = malloc(sizeof(*node));
+			assert(node);
+			node->data = j;
+			buffer.c[i].array[j - 1] = node;
+			buffer.c[i].offset++;
+		}
+	}
+
+	for (i = 0; i < num_threads; i++) {
+		ret = pthread_create(&test_threads[i], NULL,
+				     test_percpu_buffer_thread, &buffer);
+		if (ret) {
+			errno = ret;
+			perror("pthread_create");
+			abort();
+		}
+	}
+
+	for (i = 0; i < num_threads; i++) {
+		ret = pthread_join(test_threads[i], NULL);
+		if (ret) {
+			errno = ret;
+			perror("pthread_join");
+			abort();
+		}
+	}
+
+	for (i = 0; i < CPU_SETSIZE; i++) {
+		struct percpu_buffer_node *node;
+
+		if (!CPU_ISSET(i, &allowed_cpus))
+			continue;
+
+		while ((node = percpu_buffer_pop(&buffer, i))) {
+			sum += node->data;
+			free(node);
+		}
+		free(buffer.c[i].array);
+	}
+
+	/*
+	 * All entries should now be accounted for (unless some external
+	 * actor is interfering with our allowed affinity while this
+	 * test is running).
+	 */
+	assert(sum == expected_sum);
+}
+
+bool percpu_memcpy_buffer_push(struct percpu_memcpy_buffer *buffer,
+			       struct percpu_memcpy_buffer_node item, int cpu)
+{
+	for (;;) {
+		intptr_t *targetptr_final, newval_final, offset;
+		char *destptr, *srcptr;
+		size_t copylen;
+		int ret;
+
+		/* Load offset with single-copy atomicity. */
+		offset = RSEQ_READ_ONCE(buffer->c[cpu].offset);
+		if (offset == buffer->c[cpu].buflen)
+			return false;
+		destptr = (char *)&buffer->c[cpu].array[offset];
+		srcptr = (char *)&item;
+		/* copylen must be <= 4kB. */
+		copylen = sizeof(item);
+		newval_final = offset + 1;
+		targetptr_final = &buffer->c[cpu].offset;
+		if (opt_mb)
+			ret = percpu_cmpeqv_memcpy_storev_release(
+				targetptr_final, offset,
+				destptr, srcptr, copylen,
+				newval_final, cpu);
+		else
+			ret = percpu_cmpeqv_memcpy_storev(targetptr_final,
+				offset, destptr, srcptr, copylen,
+				newval_final, cpu);
+		if (rseq_likely(!ret))
+			break;
+		if (rseq_unlikely(ret < 0)) {
+			perror("cpu_opv");
+			abort();
+		}
+		/* Retry if comparison fails. */
+	}
+	return true;
+}
+
+bool percpu_memcpy_buffer_pop(struct percpu_memcpy_buffer *buffer,
+		struct percpu_memcpy_buffer_node *item, int cpu)
+{
+	for (;;) {
+		intptr_t *targetptr_final, newval_final, offset;
+		char *destptr, *srcptr;
+		size_t copylen;
+		int ret;
+
+		/* Load offset with single-copy atomicity. */
+		offset = RSEQ_READ_ONCE(buffer->c[cpu].offset);
+		if (offset == 0)
+			return false;
+		destptr = (char *)item;
+		srcptr = (char *)&buffer->c[cpu].array[offset - 1];
+		/* copylen must be <= 4kB. */
+		copylen = sizeof(*item);
+		newval_final = offset - 1;
+		targetptr_final = &buffer->c[cpu].offset;
+		ret = percpu_cmpeqv_memcpy_storev(targetptr_final,
+			offset, destptr, srcptr, copylen,
+			newval_final, cpu);
+		if (rseq_likely(!ret))
+			break;
+		if (rseq_unlikely(ret < 0)) {
+			perror("cpu_opv");
+			abort();
+		}
+		/* Retry if comparison fails. */
+	}
+	return true;
+}
+
+void *test_percpu_memcpy_buffer_thread(void *arg)
+{
+	long long i, reps;
+	struct percpu_memcpy_buffer *buffer = (struct percpu_memcpy_buffer *)arg;
+
+	if (!opt_disable_rseq && rseq_register_current_thread())
+		abort();
+
+	reps = opt_reps;
+	for (i = 0; i < reps; i++) {
+		struct percpu_memcpy_buffer_node item;
+		bool result;
+
+		result = percpu_memcpy_buffer_pop(buffer, &item,
+						  rseq_cpu_start());
+		if (opt_yield)
+			sched_yield();  /* encourage shuffling */
+		if (result) {
+			if (!percpu_memcpy_buffer_push(buffer, item,
+						       rseq_cpu_start())) {
+				/* Should increase buffer size. */
+				abort();
+			}
+		}
+	}
+
+	printf_verbose("tid %d: number of rseq abort: %d, signals delivered: %u\n",
+		       (int) gettid(), nr_abort, signals_delivered);
+	if (!opt_disable_rseq && rseq_unregister_current_thread())
+		abort();
+
+	return NULL;
+}
+
+/* Simultaneous modification to a per-cpu buffer from many threads.  */
+void test_percpu_memcpy_buffer(void)
+{
+	const int num_threads = opt_threads;
+	int i, j, ret;
+	uint64_t sum = 0, expected_sum = 0;
+	struct percpu_memcpy_buffer buffer;
+	pthread_t test_threads[num_threads];
+	cpu_set_t allowed_cpus;
+
+	memset(&buffer, 0, sizeof(buffer));
+
+	/* Generate list entries for every usable cpu. */
+	sched_getaffinity(0, sizeof(allowed_cpus), &allowed_cpus);
+	for (i = 0; i < CPU_SETSIZE; i++) {
+		if (!CPU_ISSET(i, &allowed_cpus))
+			continue;
+		/* Worse-case is every item in same CPU. */
+		buffer.c[i].array =
+			malloc(sizeof(*buffer.c[i].array) * CPU_SETSIZE *
+			       MEMCPY_BUFFER_ITEM_PER_CPU);
+		assert(buffer.c[i].array);
+		buffer.c[i].buflen = CPU_SETSIZE * MEMCPY_BUFFER_ITEM_PER_CPU;
+		for (j = 1; j <= MEMCPY_BUFFER_ITEM_PER_CPU; j++) {
+			expected_sum += 2 * j + 1;
+
+			/*
+			 * We could theoretically put the word-sized
+			 * "data" directly in the buffer. However, we
+			 * want to model objects that would not fit
+			 * within a single word, so allocate an object
+			 * for each node.
+			 */
+			buffer.c[i].array[j - 1].data1 = j;
+			buffer.c[i].array[j - 1].data2 = j + 1;
+			buffer.c[i].offset++;
+		}
+	}
+
+	for (i = 0; i < num_threads; i++) {
+		ret = pthread_create(&test_threads[i], NULL,
+				     test_percpu_memcpy_buffer_thread,
+				     &buffer);
+		if (ret) {
+			errno = ret;
+			perror("pthread_create");
+			abort();
+		}
+	}
+
+	for (i = 0; i < num_threads; i++) {
+		ret = pthread_join(test_threads[i], NULL);
+		if (ret) {
+			errno = ret;
+			perror("pthread_join");
+			abort();
+		}
+	}
+
+	for (i = 0; i < CPU_SETSIZE; i++) {
+		struct percpu_memcpy_buffer_node item;
+
+		if (!CPU_ISSET(i, &allowed_cpus))
+			continue;
+
+		while (percpu_memcpy_buffer_pop(&buffer, &item, i)) {
+			sum += item.data1;
+			sum += item.data2;
+		}
+		free(buffer.c[i].array);
+	}
+
+	/*
+	 * All entries should now be accounted for (unless some external
+	 * actor is interfering with our allowed affinity while this
+	 * test is running).
+	 */
+	assert(sum == expected_sum);
+}
+
+static void test_signal_interrupt_handler(int signo)
+{
+	signals_delivered++;
+}
+
+static int set_signal_handler(void)
+{
+	int ret = 0;
+	struct sigaction sa;
+	sigset_t sigset;
+
+	ret = sigemptyset(&sigset);
+	if (ret < 0) {
+		perror("sigemptyset");
+		return ret;
+	}
+
+	sa.sa_handler = test_signal_interrupt_handler;
+	sa.sa_mask = sigset;
+	sa.sa_flags = 0;
+	ret = sigaction(SIGUSR1, &sa, NULL);
+	if (ret < 0) {
+		perror("sigaction");
+		return ret;
+	}
+
+	printf_verbose("Signal handler set for SIGUSR1\n");
+
+	return ret;
+}
+
+static void show_usage(int argc, char **argv)
+{
+	printf("Usage : %s <OPTIONS>\n",
+		argv[0]);
+	printf("OPTIONS:\n");
+	printf("	[-1 loops] Number of loops for delay injection 1\n");
+	printf("	[-2 loops] Number of loops for delay injection 2\n");
+	printf("	[-3 loops] Number of loops for delay injection 3\n");
+	printf("	[-4 loops] Number of loops for delay injection 4\n");
+	printf("	[-5 loops] Number of loops for delay injection 5\n");
+	printf("	[-6 loops] Number of loops for delay injection 6\n");
+	printf("	[-7 loops] Number of loops for delay injection 7 (-1 to enable -m)\n");
+	printf("	[-8 loops] Number of loops for delay injection 8 (-1 to enable -m)\n");
+	printf("	[-9 loops] Number of loops for delay injection 9 (-1 to enable -m)\n");
+	printf("	[-m N] Yield/sleep/kill every modulo N (default 0: disabled) (>= 0)\n");
+	printf("	[-y] Yield\n");
+	printf("	[-k] Kill thread with signal\n");
+	printf("	[-s S] S: =0: disabled (default), >0: sleep time (ms)\n");
+	printf("	[-t N] Number of threads (default 200)\n");
+	printf("	[-r N] Number of repetitions per thread (default 5000)\n");
+	printf("	[-d] Disable rseq system call (no initialization)\n");
+	printf("	[-D M] Disable rseq for each M threads\n");
+	printf("	[-T test] Choose test: (s)pinlock, (l)ist, (b)uffer, (m)emcpy, (i)ncrement\n");
+	printf("	[-M] Push into buffer and memcpy buffer with memory barriers.\n");
+	printf("	[-v] Verbose output.\n");
+	printf("	[-h] Show this help.\n");
+	printf("\n");
+}
+
+int main(int argc, char **argv)
+{
+	int i;
+
+	for (i = 1; i < argc; i++) {
+		if (argv[i][0] != '-')
+			continue;
+		switch (argv[i][1]) {
+		case '1':
+		case '2':
+		case '3':
+		case '4':
+		case '5':
+		case '6':
+		case '7':
+		case '8':
+		case '9':
+			if (argc < i + 2) {
+				show_usage(argc, argv);
+				goto error;
+			}
+			loop_cnt[argv[i][1] - '0'] = atol(argv[i + 1]);
+			i++;
+			break;
+		case 'm':
+			if (argc < i + 2) {
+				show_usage(argc, argv);
+				goto error;
+			}
+			opt_modulo = atol(argv[i + 1]);
+			if (opt_modulo < 0) {
+				show_usage(argc, argv);
+				goto error;
+			}
+			i++;
+			break;
+		case 's':
+			if (argc < i + 2) {
+				show_usage(argc, argv);
+				goto error;
+			}
+			opt_sleep = atol(argv[i + 1]);
+			if (opt_sleep < 0) {
+				show_usage(argc, argv);
+				goto error;
+			}
+			i++;
+			break;
+		case 'y':
+			opt_yield = 1;
+			break;
+		case 'k':
+			opt_signal = 1;
+			break;
+		case 'd':
+			opt_disable_rseq = 1;
+			break;
+		case 'D':
+			if (argc < i + 2) {
+				show_usage(argc, argv);
+				goto error;
+			}
+			opt_disable_mod = atol(argv[i + 1]);
+			if (opt_disable_mod < 0) {
+				show_usage(argc, argv);
+				goto error;
+			}
+			i++;
+			break;
+		case 't':
+			if (argc < i + 2) {
+				show_usage(argc, argv);
+				goto error;
+			}
+			opt_threads = atol(argv[i + 1]);
+			if (opt_threads < 0) {
+				show_usage(argc, argv);
+				goto error;
+			}
+			i++;
+			break;
+		case 'r':
+			if (argc < i + 2) {
+				show_usage(argc, argv);
+				goto error;
+			}
+			opt_reps = atoll(argv[i + 1]);
+			if (opt_reps < 0) {
+				show_usage(argc, argv);
+				goto error;
+			}
+			i++;
+			break;
+		case 'h':
+			show_usage(argc, argv);
+			goto end;
+		case 'T':
+			if (argc < i + 2) {
+				show_usage(argc, argv);
+				goto error;
+			}
+			opt_test = *argv[i + 1];
+			switch (opt_test) {
+			case 's':
+			case 'l':
+			case 'i':
+			case 'b':
+			case 'm':
+				break;
+			default:
+				show_usage(argc, argv);
+				goto error;
+			}
+			i++;
+			break;
+		case 'v':
+			verbose = 1;
+			break;
+		case 'M':
+			opt_mb = 1;
+			break;
+		default:
+			show_usage(argc, argv);
+			goto error;
+		}
+	}
+
+	loop_cnt_1 = loop_cnt[1];
+	loop_cnt_2 = loop_cnt[2];
+	loop_cnt_3 = loop_cnt[3];
+	loop_cnt_4 = loop_cnt[4];
+	loop_cnt_5 = loop_cnt[5];
+	loop_cnt_6 = loop_cnt[6];
+
+	if (set_signal_handler())
+		goto error;
+
+	if (!opt_disable_rseq && rseq_register_current_thread())
+		goto error;
+	switch (opt_test) {
+	case 's':
+		printf_verbose("spinlock\n");
+		test_percpu_spinlock();
+		break;
+	case 'l':
+		printf_verbose("linked list\n");
+		test_percpu_list();
+		break;
+	case 'b':
+		printf_verbose("buffer\n");
+		test_percpu_buffer();
+		break;
+	case 'm':
+		printf_verbose("memcpy buffer\n");
+		test_percpu_memcpy_buffer();
+		break;
+	case 'i':
+		printf_verbose("counter increment\n");
+		test_percpu_inc();
+		break;
+	}
+	if (!opt_disable_rseq && rseq_unregister_current_thread())
+		abort();
+end:
+	return 0;
+
+error:
+	return -1;
+}
-- 
2.11.0

[RFC PATCH for 4.16 21/21] rseq: selftests: Provide Makefile, scripts, gitignore

[Mathieu Desnoyers]

A run_param_test.sh script runs many variants of the parametrizable
tests.

Wire up the rseq Makefile, add directory entry into MAINTAINERS file.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Shuah Khan <shuahkh@osg.samsung.com>
CC: Russell King <linux@arm.linux.org.uk>
CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Paul Turner <pjt@google.com>
CC: Andrew Hunter <ahh@google.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Andy Lutomirski <luto@amacapital.net>
CC: Andi Kleen <andi@firstfloor.org>
CC: Dave Watson <davejwatson@fb.com>
CC: Chris Lameter <cl@linux.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: "H. Peter Anvin" <hpa@zytor.com>
CC: Ben Maurer <bmaurer@fb.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Linus Torvalds <torvalds@linux-foundation.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Boqun Feng <boqun.feng@gmail.com>
CC: linux-kselftest@vger.kernel.org
CC: linux-api@vger.kernel.org
---
 MAINTAINERS                                    |   1 +
 tools/testing/selftests/Makefile               |   1 +
 tools/testing/selftests/rseq/.gitignore        |   7 ++
 tools/testing/selftests/rseq/Makefile          |  37 +++++++
 tools/testing/selftests/rseq/run_param_test.sh | 130 +++++++++++++++++++++++++
 5 files changed, 176 insertions(+)
 create mode 100644 tools/testing/selftests/rseq/.gitignore
 create mode 100644 tools/testing/selftests/rseq/Makefile
 create mode 100755 tools/testing/selftests/rseq/run_param_test.sh

diff --git a/MAINTAINERS b/MAINTAINERS
index dce21dea5a63..64422f253aa9 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -11638,6 +11638,7 @@ S:	Supported
 F:	kernel/rseq.c
 F:	include/uapi/linux/rseq.h
 F:	include/trace/events/rseq.h
+F:	tools/testing/selftests/rseq/
 
 RFKILL
 M:	Johannes Berg <johannes@sipsolutions.net>
diff --git a/tools/testing/selftests/Makefile b/tools/testing/selftests/Makefile
index fc1eba0e0130..fc314334628a 100644
--- a/tools/testing/selftests/Makefile
+++ b/tools/testing/selftests/Makefile
@@ -26,6 +26,7 @@ TARGETS += nsfs
 TARGETS += powerpc
 TARGETS += pstore
 TARGETS += ptrace
+TARGETS += rseq
 TARGETS += seccomp
 TARGETS += sigaltstack
 TARGETS += size
diff --git a/tools/testing/selftests/rseq/.gitignore b/tools/testing/selftests/rseq/.gitignore
new file mode 100644
index 000000000000..73dcf05b69fd
--- /dev/null
+++ b/tools/testing/selftests/rseq/.gitignore
@@ -0,0 +1,7 @@
+basic_percpu_ops_test
+basic_test
+basic_rseq_op_test
+param_test
+param_test_benchmark
+param_test_compare_twice
+param_test_skip_fastpath
diff --git a/tools/testing/selftests/rseq/Makefile b/tools/testing/selftests/rseq/Makefile
new file mode 100644
index 000000000000..69f90d718774
--- /dev/null
+++ b/tools/testing/selftests/rseq/Makefile
@@ -0,0 +1,37 @@
+CFLAGS += -O2 -Wall -g -I./ -I../cpu-opv/ -I../../../../usr/include/ -L./ -Wl,-rpath=./
+LDLIBS += -lpthread
+
+# Own dependencies because we only want to build against 1st prerequisite, but
+# still track changes to header files and depend on shared object.
+OVERRIDE_TARGETS = 1
+
+TEST_GEN_PROGS = basic_test basic_percpu_ops_test \
+		param_test param_test_skip_fastpath \
+		param_test_benchmark param_test_compare_twice
+
+TEST_GEN_PROGS_EXTENDED = librseq.so libcpu-op.so
+
+TEST_PROGS = run_param_test.sh
+
+include ../lib.mk
+
+$(OUTPUT)/librseq.so: rseq.c rseq.h rseq-*.h
+	$(CC) $(CFLAGS) -shared -fPIC $< $(LDLIBS) -o $@
+
+$(OUTPUT)/libcpu-op.so: ../cpu-opv/cpu-op.c ../cpu-opv/cpu-op.h
+	$(CC) $(CFLAGS) -shared -fPIC $< $(LDLIBS) -o $@
+
+$(OUTPUT)/%: %.c $(TEST_GEN_PROGS_EXTENDED) rseq.h rseq-*.h ../cpu-opv/cpu-op.h percpu-op.h
+	$(CC) $(CFLAGS) $< $(LDLIBS) -lrseq -lcpu-op -o $@
+
+$(OUTPUT)/param_test_skip_fastpath: param_test.c $(TEST_GEN_PROGS_EXTENDED) \
+					rseq.h rseq-*.h ../cpu-opv/cpu-op.h percpu-op.h
+	$(CC) $(CFLAGS) -DRSEQ_SKIP_FASTPATH $< $(LDLIBS) -lrseq -lcpu-op -o $@
+
+$(OUTPUT)/param_test_benchmark: param_test.c $(TEST_GEN_PROGS_EXTENDED) \
+					rseq.h rseq-*.h ../cpu-opv/cpu-op.h percpu-op.h
+	$(CC) $(CFLAGS) -DBENCHMARK $< $(LDLIBS) -lrseq -lcpu-op -o $@
+
+$(OUTPUT)/param_test_compare_twice: param_test.c $(TEST_GEN_PROGS_EXTENDED) \
+					rseq.h rseq-*.h ../cpu-opv/cpu-op.h percpu-op.h
+	$(CC) $(CFLAGS) -DRSEQ_COMPARE_TWICE $< $(LDLIBS) -lrseq -lcpu-op -o $@
diff --git a/tools/testing/selftests/rseq/run_param_test.sh b/tools/testing/selftests/rseq/run_param_test.sh
new file mode 100755
index 000000000000..64d4c7d722c9
--- /dev/null
+++ b/tools/testing/selftests/rseq/run_param_test.sh
@@ -0,0 +1,130 @@
+#!/bin/bash
+
+EXTRA_ARGS=${@}
+
+OLDIFS="$IFS"
+IFS=$'\n'
+TEST_LIST=(
+	"-T s"
+	"-T l"
+	"-T b"
+	"-T b -M"
+	"-T m"
+	"-T m -M"
+	"-T i"
+)
+
+TEST_NAME=(
+	"spinlock"
+	"list"
+	"buffer"
+	"buffer with barrier"
+	"memcpy"
+	"memcpy with barrier"
+	"increment"
+)
+IFS="$OLDIFS"
+
+REPS=1000
+
+function do_tests()
+{
+	local i=0
+	while [ "$i" -lt "${#TEST_LIST[@]}" ]; do
+		echo "Running test ${TEST_NAME[$i]}"
+		./param_test ${TEST_LIST[$i]} -r ${REPS} ${@} ${EXTRA_ARGS} || exit 1
+		echo "Running skip fast-path test ${TEST_NAME[$i]}"
+		./param_test_skip_fastpath ${TEST_LIST[$i]} -r ${REPS} ${@} ${EXTRA_ARGS} || exit 1
+		echo "Running compare-twice test ${TEST_NAME[$i]}"
+		./param_test_compare_twice ${TEST_LIST[$i]} -r ${REPS} ${@} ${EXTRA_ARGS} || exit 1
+		let "i++"
+	done
+}
+
+echo "Default parameters"
+do_tests
+
+echo "Loop injection: 10000 loops"
+
+OLDIFS="$IFS"
+IFS=$'\n'
+INJECT_LIST=(
+	"1"
+	"2"
+	"3"
+	"4"
+	"5"
+	"6"
+	"7"
+	"8"
+	"9"
+)
+IFS="$OLDIFS"
+
+NR_LOOPS=10000
+
+i=0
+while [ "$i" -lt "${#INJECT_LIST[@]}" ]; do
+	echo "Injecting at <${INJECT_LIST[$i]}>"
+	do_tests -${INJECT_LIST[i]} ${NR_LOOPS}
+	let "i++"
+done
+NR_LOOPS=
+
+function inject_blocking()
+{
+	OLDIFS="$IFS"
+	IFS=$'\n'
+	INJECT_LIST=(
+		"7"
+		"8"
+		"9"
+	)
+	IFS="$OLDIFS"
+
+	NR_LOOPS=-1
+
+	i=0
+	while [ "$i" -lt "${#INJECT_LIST[@]}" ]; do
+		echo "Injecting at <${INJECT_LIST[$i]}>"
+		do_tests -${INJECT_LIST[i]} -1 ${@}
+		let "i++"
+	done
+	NR_LOOPS=
+}
+
+echo "Yield injection (25%)"
+inject_blocking -m 4 -y
+
+echo "Yield injection (50%)"
+inject_blocking -m 2 -y
+
+echo "Yield injection (100%)"
+inject_blocking -m 1 -y
+
+echo "Kill injection (25%)"
+inject_blocking -m 4 -k
+
+echo "Kill injection (50%)"
+inject_blocking -m 2 -k
+
+echo "Kill injection (100%)"
+inject_blocking -m 1 -k
+
+echo "Sleep injection (1ms, 25%)"
+inject_blocking -m 4 -s 1
+
+echo "Sleep injection (1ms, 50%)"
+inject_blocking -m 2 -s 1
+
+echo "Sleep injection (1ms, 100%)"
+inject_blocking -m 1 -s 1
+
+echo "Disable rseq for 25% threads"
+do_tests -D 4
+
+echo "Disable rseq for 50% threads"
+do_tests -D 2
+
+echo "Disable rseq"
+do_tests -d
-- 
2.11.0

Re: [RFC PATCH for 4.16 02/21] rseq: Introduce restartable sequences system call (v12)

[Christopher Lameter]

On Thu, 14 Dec 2017, Mathieu Desnoyers wrote:

> On x86, yet another possible approach would be to use the gs segment
> selector to point to user-space per-cpu data. This approach performs
> similarly to the cpu id cache, but it has two disadvantages: it is
> not portable, and it is incompatible with existing applications already
> using the gs segment selector for other purposes.

I think the proper way to think about gs and fs on x86 is as base
registers. They are essentially values in registers added to the address
generated in an instruction. As such the approach is transferable to other
processor architecture. Many support base register and base register
relative processing. If a processor can do RMV instructions base register
relative then you have something similar.

In a restartable sequence you could increase efficieny by avoiding full
atomic instructions. This would be similar to the lockless RMV available
on x86 then. And in that form it is portable.

A context switch to another processors would mean that the value of the
base register has changed and that we therefore are accessing another per
cpu segment. Restarting the sequence will yield a correct result without
any reloading of registers.

Re: [RFC PATCH for 4.16 02/21] rseq: Introduce restartable sequences system call (v12)

[Mathieu Desnoyers]

----- On Dec 14, 2017, at 11:44 AM, Chris Lameter cl-vYTEC60ixJUAvxtiuMwx3w@public.gmane.org wrote:

> On Thu, 14 Dec 2017, Mathieu Desnoyers wrote:
> 
>> On x86, yet another possible approach would be to use the gs segment
>> selector to point to user-space per-cpu data. This approach performs
>> similarly to the cpu id cache, but it has two disadvantages: it is
>> not portable, and it is incompatible with existing applications already
>> using the gs segment selector for other purposes.
> 
> I think the proper way to think about gs and fs on x86 is as base
> registers. They are essentially values in registers added to the address
> generated in an instruction. As such the approach is transferable to other
> processor architecture. Many support base register and base register
> relative processing. If a processor can do RMV instructions base register
> relative then you have something similar.

How would you do it on ARM32 ?

> 
> In a restartable sequence you could increase efficieny by avoiding full
> atomic instructions. This would be similar to the lockless RMV available
> on x86 then. And in that form it is portable.
> 
> A context switch to another processors would mean that the value of the
> base register has changed and that we therefore are accessing another per
> cpu segment. Restarting the sequence will yield a correct result without
> any reloading of registers.

As a concrete example, let's try to apply your proposal on a common use-case:
a compare-and-store on user-space per-cpu data.

With my rseq proposal the fast-path pseudo-code boils down to:

load TLS::cpu_id_start into reg_X
add reg_X offset to base to find target v
store pointer to TLS::rseq_cs
compare reg_X against TLS::cpu_id
jne abort
cmp *v, value
jne cmpfail
store newval to *v

My benchmark on Intel x86-64 E5-2630 shows that it takes 1.9 ns/iteration
for a test-case incrementing a counter with this rseq compare-and-store
sequence.

Let's assume we can reserve the gs segment selector for use in user-space,
and that the per-cpu data layout allows using this segment selector as offset.
The compare-and-store use-case would require a "cmpxchg" instruction with
a gs segment selector.

A single-threaded test-case which uses non-lock-prefixed cmpxchg in a loop
on a E5-2630, I get 2.8 ns/iteration. (no per-cpu data involved, done on a single
global value)

One benefit of your proposal is to lessen the number of retired instructions,
but if we take the IPC into account, it is slower than rseq in my benchmark. What
benefits do you expect from using segment selectors and non-lock-prefixed atomic
instructions on the fast-path ?

Thanks,

Mathieu

-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com

Re: [RFC PATCH for 4.16 02/21] rseq: Introduce restartable sequences system call (v12)

[Christopher Lameter]

On Thu, 14 Dec 2017, Mathieu Desnoyers wrote:

> > I think the proper way to think about gs and fs on x86 is as base
> > registers. They are essentially values in registers added to the address
> > generated in an instruction. As such the approach is transferable to other
> > processor architecture. Many support base register and base register
> > relative processing. If a processor can do RMV instructions base register
> > relative then you have something similar.
>
> How would you do it on ARM32 ?

Actually you do not really need RMV instructions. The data is cpu specific
so within a restartable sequence you would have exclusive access right?

F.e. a increment would be

1. Load base register relative
2. add 1
3. Store base register relative

The main overhead would be the registeration of the sequence.

The advantage on x86 is that you do not need a restartable sequence
since a single lockless RMV instruction can do this (this_cpu_inc f.e.)

> One benefit of your proposal is to lessen the number of retired instructions,
> but if we take the IPC into account, it is slower than rseq in my benchmark. What
> benefits do you expect from using segment selectors and non-lock-prefixed atomic
> instructions on the fast-path ?

Ultimately I wish fast increments like done by this_cpu_inc() could be
implemented in an efficient way on non x86 platforms that do not have
cheap instructions like that.

If cmpxchg local is slower than a group of instructions to do the same
then there is an obvious question to the cpu architects why we would need
the instruction at all (aside from the fact that we do not need a
restartable sequence for these instructions).

Re: [RFC PATCH for 4.16 02/21] rseq: Introduce restartable sequences system call (v12)

[Mathieu Desnoyers]

----- On Dec 14, 2017, at 1:50 PM, Chris Lameter cl-vYTEC60ixJUAvxtiuMwx3w@public.gmane.org wrote:

> On Thu, 14 Dec 2017, Mathieu Desnoyers wrote:
> 
>> > I think the proper way to think about gs and fs on x86 is as base
>> > registers. They are essentially values in registers added to the address
>> > generated in an instruction. As such the approach is transferable to other
>> > processor architecture. Many support base register and base register
>> > relative processing. If a processor can do RMV instructions base register
>> > relative then you have something similar.
>>
>> How would you do it on ARM32 ?
> 
> Actually you do not really need RMV instructions. The data is cpu specific
> so within a restartable sequence you would have exclusive access right?

Yep.

> 
> F.e. a increment would be
> 
> 1. Load base register relative
> 2. add 1
> 3. Store base register relative

Actually, for the increment case, rseq headers provide a "add" API,
which uses a "add" instruction on x86. On arm 32, it does indeed:

RSEQ_ASM_STORE_RSEQ_CS(1, 3f, rseq_cs)
RSEQ_ASM_CMP_CPU_ID(cpu_id, current_cpu_id, 4f)
"ldr r0, %[v]\n\t"
"add r0, %[count]\n\t"
/* final store */
"str r0, %[v]\n\t"

> 
> The main overhead would be the registeration of the sequence.

Registering a rseq is a single store to a TLS (in user-space),
which really isn't that expensive.

If we port this concept to kernel-space (as I start to understand
would be your wish), then a simple pointer store to the current
task_struct would suffice.

> 
> The advantage on x86 is that you do not need a restartable sequence
> since a single lockless RMV instruction can do this (this_cpu_inc f.e.)

Indeed, on x86, for the specific case of counter increment, the single-instruction
"add" or "inc" with a segment-selector prefix can save setting up the rseq
(a pointer store), and offsetting from a base using the cpu number.

If your wish is to do this at kernel level, where we have full control over
the gs segment, this makes sense. I'm worried that applying this to user-space
might create conflicts wrt who owns that segment selector register wrt
pre-existing applications.

> 
>> One benefit of your proposal is to lessen the number of retired instructions,
>> but if we take the IPC into account, it is slower than rseq in my benchmark.
>> What
>> benefits do you expect from using segment selectors and non-lock-prefixed atomic
>> instructions on the fast-path ?
> 
> Ultimately I wish fast increments like done by this_cpu_inc() could be
> implemented in an efficient way on non x86 platforms that do not have
> cheap instructions like that.

My understanding is that your focus is mainly on kernel code, right ? Or is
your aim to port this_cpu_inc() to userspace as well ?

Indeed, the concepts behind rseq could be ported to kernel code eventually.
The immediate gain is much higher by exposing this to user-space though,
given that there is no good way to perform per-cpu operations efficiently
at all there, whereas kernel code can always disable preemption.

> 
> If cmpxchg local is slower than a group of instructions to do the same
> then there is an obvious question to the cpu architects why we would need
> the instruction at all (aside from the fact that we do not need a
> restartable sequence for these instructions).

I'm not a specialist in CPU instruction scheduling, so I won't speculate
on this topic. ;-)

Thanks,

Mathieu

-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com

Re: [RFC PATCH for 4.16 02/21] rseq: Introduce restartable sequences system call (v12)

[Peter Zijlstra]

On Thu, Dec 14, 2017 at 12:50:13PM -0600, Christopher Lameter wrote:
> Ultimately I wish fast increments like done by this_cpu_inc() could be
> implemented in an efficient way on non x86 platforms that do not have
> cheap instructions like that.

So the problem isn't migration; for that we could wrap the operation in
preempt_disable() which is not more expensive than rseq would be. And a
lot more deterministic.

The problem instead is interrupts, which can result in nested load-store
operations, and that comes apart. This then means having to disable
interrupts over these things and _that_ is expensive.

Re: [RFC PATCH for 4.16 02/21] rseq: Introduce restartable sequences system call (v12)

[Mathieu Desnoyers]

----- On Dec 14, 2017, at 2:48 PM, Peter Zijlstra peterz-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org wrote:

> On Thu, Dec 14, 2017 at 12:50:13PM -0600, Christopher Lameter wrote:
>> Ultimately I wish fast increments like done by this_cpu_inc() could be
>> implemented in an efficient way on non x86 platforms that do not have
>> cheap instructions like that.
> 
> So the problem isn't migration; for that we could wrap the operation in
> preempt_disable() which is not more expensive than rseq would be. And a
> lot more deterministic.
> 
> The problem instead is interrupts, which can result in nested load-store
> operations, and that comes apart. This then means having to disable
> interrupts over these things and _that_ is expensive.

Then could we consider checking a per task-struct rseq_cs pointer when
returning from interrupt handler ? This rseq_cs pointer would track
kernel restartable sequences. This would also work for NMI handlers.

Thanks,

Mathieu


-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com

Re: [RFC PATCH for 4.16 02/21] rseq: Introduce restartable sequences system call (v12)

[Peter Zijlstra]

On Thu, Dec 14, 2017 at 07:57:08PM +0000, Mathieu Desnoyers wrote:
> ----- On Dec 14, 2017, at 2:48 PM, Peter Zijlstra peterz-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org wrote:
> 
> > On Thu, Dec 14, 2017 at 12:50:13PM -0600, Christopher Lameter wrote:
> >> Ultimately I wish fast increments like done by this_cpu_inc() could be
> >> implemented in an efficient way on non x86 platforms that do not have
> >> cheap instructions like that.
> > 
> > So the problem isn't migration; for that we could wrap the operation in
> > preempt_disable() which is not more expensive than rseq would be. And a
> > lot more deterministic.
> > 
> > The problem instead is interrupts, which can result in nested load-store
> > operations, and that comes apart. This then means having to disable
> > interrupts over these things and _that_ is expensive.
> 
> Then could we consider checking a per task-struct rseq_cs pointer when
> returning from interrupt handler ? This rseq_cs pointer would track
> kernel restartable sequences. This would also work for NMI handlers.

I really don't much like making the interrupt handlers more expensive
for this.

And I don't think NMIs are a real worry, you should be very careful when
you share state with any of them in any case.

Also; what you can do is soft interrupt disable, which is effectively
the oppose approach, instead of restarting the sequence, you delay the
interrupt handler. And that has the obvious benefit of making all the
local_irq_disable/enable crud much faster all over.

This is something PowerPC already does.

Re: [RFC PATCH for 4.16 02/21] rseq: Introduce restartable sequences system call (v12)

[Christopher Lameter]

On Thu, 14 Dec 2017, Mathieu Desnoyers wrote:

> If we port this concept to kernel-space (as I start to understand
> would be your wish), then a simple pointer store to the current
> task_struct would suffice.

Certainly such a port would be beneficial for non x86 archs.

But my company has extensive user space code that maintains a lot of
counters and does other tricks to get full performance out of the
hardware. Such a mechanism would also be good from user space. Why keep
the good stuff only inside the kernel?

Re: [RFC PATCH for 4.16 02/21] rseq: Introduce restartable sequences system call (v12)

[Peter Zijlstra]

On Thu, Dec 14, 2017 at 03:14:00PM -0600, Christopher Lameter wrote:
> On Thu, 14 Dec 2017, Mathieu Desnoyers wrote:
> 
> > If we port this concept to kernel-space (as I start to understand
> > would be your wish), then a simple pointer store to the current
> > task_struct would suffice.
> 
> Certainly such a port would be beneficial for non x86 archs.
> 
> But my company has extensive user space code that maintains a lot of
> counters and does other tricks to get full performance out of the
> hardware. Such a mechanism would also be good from user space. Why keep
> the good stuff only inside the kernel?

Mathieu's proposal is for userspace, _only_ userspace.

Re: [RFC PATCH for 4.16 02/21] rseq: Introduce restartable sequences system call (v12)

[Christopher Lameter]

On Thu, 14 Dec 2017, Peter Zijlstra wrote:

> > But my company has extensive user space code that maintains a lot of
> > counters and does other tricks to get full performance out of the
> > hardware. Such a mechanism would also be good from user space. Why keep
> > the good stuff only inside the kernel?
>
> Mathieu's proposal is for userspace, _only_ userspace.

But what we were talking about are instructions that work effectively in
kernel space whose efficiency restartable sequences could bring to user
space.

Re: [RFC PATCH for 4.16 02/21] rseq: Introduce restartable sequences system call (v12)

[Mathieu Desnoyers]

----- On Dec 15, 2017, at 10:05 AM, Chris Lameter cl-vYTEC60ixJUAvxtiuMwx3w@public.gmane.org wrote:

> On Thu, 14 Dec 2017, Peter Zijlstra wrote:
> 
>> > But my company has extensive user space code that maintains a lot of
>> > counters and does other tricks to get full performance out of the
>> > hardware. Such a mechanism would also be good from user space. Why keep
>> > the good stuff only inside the kernel?
>>
>> Mathieu's proposal is for userspace, _only_ userspace.
> 
> But what we were talking about are instructions that work effectively in
> kernel space whose efficiency restartable sequences could bring to user
> space.

It can be worthwhile to recap my understanding of this thread so far:

AFAIU, Chris' proposal is to use the "gs" segment selector as instruction
prefix on x86 rather than explicitly loading CPU number and calculating
offsets.

This can turn sequences of rseq operations like this cmpxchg:

Registers:

  R1: return value
  R2: expected value
  R3: new value
  R4: cpu_id

rseq cmpxchg:

  load TLS::cpu_id_start into R4
  calculate offset of v
  fs:mov (store rseq descriptor address into TLS::rseq_cs)
  compare R4 against TLS::cpu_id
  jne abort
  mov (load v into R1)
  compare R1 against R2
  jne cmpfail
  mov (store R3 into *v)

into:

  fs:mov (store rseq descriptor address into TLS::rseq_cs)
  gs:mov (load *v+off into R1)
  compare R1 against R2
  jne cmpfail
  gs:mov (store R3 into *v+off)

My first concern with this approach is the lack of flexibility of the segment
selector method wrt variety of schemes user-space has to deal with for memory
allocation. In the kernel, this is achieved by ensuring that all per-cpu data
layout is segment-selector-prefix friendly.

Another aspect that worries me is applications using the gs segment selector
for other purposes. Suddenly reserving the gs segment selector for use by a
library like glibc may lead to incompatibilities with applications already
using it.

Thanks,

Mathieu

-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com

Re: [RFC PATCH for 4.16 02/21] rseq: Introduce restartable sequences system call (v12)

[Christopher Lameter]

On Fri, 15 Dec 2017, Mathieu Desnoyers wrote:

> Another aspect that worries me is applications using the gs segment selector
> for other purposes. Suddenly reserving the gs segment selector for use by a
> library like glibc may lead to incompatibilities with applications already
> using it.

fs/gs seems to be reserved for thread local storage. So it would
be shared in user space like the corresponding cpu segment register in
kernel space where multiple subsystems share %gs.

The same can be done in user space. Ulrich Drepper has a writeup on this

https://www.akkadia.org/drepper/tls.pdf

Savings in execution time could come about because there would not be the
need to determine the address of the processor specific memory area in
each restartable sequence and there would be memory free of contention for
such a sequence in order f.e. to realize fast counters.

Re: [RFC PATCH for 4.16 10/21] cpu_opv: Provide cpu_opv system call (v5)

[Mathieu Desnoyers]

Hi Al,

Your feedback on this new cpu_opv system call would be welcome. This series
is now aiming at the next merge window (4.17).

The whole restartable sequences series can be fetched at:

https://git.kernel.org/pub/scm/linux/kernel/git/rseq/linux-rseq.git/
tag: v4.15-rc9-rseq-20180122

Thanks!

Mathieu

----- On Dec 14, 2017, at 11:13 AM, Mathieu Desnoyers mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org wrote:

> The cpu_opv system call executes a vector of operations on behalf of
> user-space on a specific CPU with preemption disabled. It is inspired
> by readv() and writev() system calls which take a "struct iovec"
> array as argument.
> 
> The operations available are: comparison, memcpy, add, or, and, xor,
> left shift, right shift, and memory barrier. The system call receives
> a CPU number from user-space as argument, which is the CPU on which
> those operations need to be performed.  All pointers in the ops must
> have been set up to point to the per CPU memory of the CPU on which
> the operations should be executed. The "comparison" operation can be
> used to check that the data used in the preparation step did not
> change between preparation of system call inputs and operation
> execution within the preempt-off critical section.
> 
> The reason why we require all pointer offsets to be calculated by
> user-space beforehand is because we need to use get_user_pages_fast()
> to first pin all pages touched by each operation. This takes care of
> faulting-in the pages. Then, preemption is disabled, and the
> operations are performed atomically with respect to other thread
> execution on that CPU, without generating any page fault.
> 
> An overall maximum of 4216 bytes in enforced on the sum of operation
> length within an operation vector, so user-space cannot generate a
> too long preempt-off critical section (cache cold critical section
> duration measured as 4.7µs on x86-64). Each operation is also limited
> a length of 4096 bytes, meaning that an operation can touch a
> maximum of 4 pages (memcpy: 2 pages for source, 2 pages for
> destination if addresses are not aligned on page boundaries).
> 
> If the thread is not running on the requested CPU, it is migrated to
> it.
> 
> **** Justification for cpu_opv ****
> 
> Here are a few reasons justifying why the cpu_opv system call is
> needed in addition to rseq:
> 
> 1) Allow algorithms to perform per-cpu data migration without relying on
>   sched_setaffinity()
> 
> The use-cases are migrating memory between per-cpu memory free-lists, or
> stealing tasks from other per-cpu work queues: each require that
> accesses to remote per-cpu data structures are performed.
> 
> Just rseq is not enough to cover those use-cases without additionally
> relying on sched_setaffinity, which is unfortunately not
> CPU-hotplug-safe.
> 
> The cpu_opv system call receives a CPU number as argument, and migrates
> the current task to the right CPU to perform the operation sequence. If
> the requested CPU is offline, it performs the operations from the
> current CPU while preventing CPU hotplug, and with a mutex held.
> 
> 2) Handling single-stepping from tools
> 
> Tools like debuggers, and simulators use single-stepping to run through
> existing programs. If core libraries start to use restartable sequences
> for e.g. memory allocation, this means pre-existing programs cannot be
> single-stepped, simply because the underlying glibc or jemalloc has
> changed.
> 
> The rseq user-space does expose a __rseq_table section for the sake of
> debuggers, so they can skip over the rseq critical sections if they
> want.  However, this requires upgrading tools, and still breaks
> single-stepping in case where glibc or jemalloc is updated, but not the
> tooling.
> 
> Having a performance-related library improvement break tooling is likely
> to cause a big push-back against wide adoption of rseq.
> 
> 3) Forward-progress guarantee
> 
> Having a piece of user-space code that stops progressing due to external
> conditions is pretty bad. Developers are used to think of fast-path and
> slow-path (e.g. for locking), where the contended vs uncontended cases
> have different performance characteristics, but each need to provide
> some level of progress guarantees.
> 
> There are concerns about proposing just "rseq" without the associated
> slow-path (cpu_opv) that guarantees progress. It's just asking for
> trouble when real-life will happen: page faults, uprobes, and other
> unforeseen conditions that would seldom cause a rseq fast-path to never
> progress.
> 
> 4) Handling page faults
> 
> It's pretty easy to come up with corner-case scenarios where rseq does
> not progress without the help from cpu_opv. For instance, a system with
> swap enabled which is under high memory pressure could trigger page
> faults at pretty much every rseq attempt. Although this scenario
> is extremely unlikely, rseq becomes the weak link of the chain.
> 
> 5) Comparison with LL/SC
> 
> The layman versed in the load-link/store-conditional instructions in
> RISC architectures will notice the similarity between rseq and LL/SC
> critical sections. The comparison can even be pushed further: since
> debuggers can handle those LL/SC critical sections, they should be
> able to handle rseq c.s. in the same way.
> 
> First, the way gdb recognises LL/SC c.s. patterns is very fragile:
> it's limited to specific common patterns, and will miss the pattern
> in all other cases. But fear not, having the rseq c.s. expose a
> __rseq_table to debuggers removes that guessing part.
> 
> The main difference between LL/SC and rseq is that debuggers had
> to support single-stepping through LL/SC critical sections from the
> get go in order to support a given architecture. For rseq, we're
> adding critical sections into pre-existing applications/libraries,
> so the user expectation is that tools don't break due to a library
> optimization.
> 
> 6) Perform maintenance operations on per-cpu data
> 
> rseq c.s. are quite limited feature-wise: they need to end with a
> *single* commit instruction that updates a memory location. On the other
> hand, the cpu_opv system call can combine a sequence of operations that
> need to be executed with preemption disabled. While slower than rseq,
> this allows for more complex maintenance operations to be performed on
> per-cpu data concurrently with rseq fast-paths, in cases where it's not
> possible to map those sequences of ops to a rseq.
> 
> 7) Use cpu_opv as generic implementation for architectures not
>   implementing rseq assembly code
> 
> rseq critical sections require architecture-specific user-space code to
> be crafted in order to port an algorithm to a given architecture.  In
> addition, it requires that the kernel architecture implementation adds
> hooks into signal delivery and resume to user-space.
> 
> In order to facilitate integration of rseq into user-space, cpu_opv can
> provide a (relatively slower) architecture-agnostic implementation of
> rseq. This means that user-space code can be ported to all architectures
> through use of cpu_opv initially, and have the fast-path use rseq
> whenever the asm code is implemented.
> 
> 8) Allow libraries with multi-part algorithms to work on same per-cpu
>   data without affecting the allowed cpu mask
> 
> The lttng-ust tracer presents an interesting use-case for per-cpu
> buffers: the algorithm needs to update a "reserve" counter, serialize
> data into the buffer, and then update a "commit" counter _on the same
> per-cpu buffer_. Using rseq for both reserve and commit can bring
> significant performance benefits.
> 
> Clearly, if rseq reserve fails, the algorithm can retry on a different
> per-cpu buffer. However, it's not that easy for the commit. It needs to
> be performed on the same per-cpu buffer as the reserve.
> 
> The cpu_opv system call solves that problem by receiving the cpu number
> on which the operation needs to be performed as argument. It can push
> the task to the right CPU if needed, and perform the operations there
> with preemption disabled.
> 
> Changing the allowed cpu mask for the current thread is not an
> acceptable alternative for a tracing library, because the application
> being traced does not expect that mask to be changed by libraries.
> 
> 9) Ensure that data structures don't need store-release/load-acquire
>   semantic to handle fall-back
> 
> cpu_opv performs the fall-back on the requested CPU by migrating the
> task to that CPU. Executing the slow-path on the right CPU ensures that
> store-release/load-acquire semantic is not required neither on the
> fast-path nor slow-path.
> 
> **** rseq and cpu_opv use-cases ****
> 
> 1) per-cpu spinlock
> 
> A per-cpu spinlock can be implemented as a rseq consisting of a
> comparison operation (== 0) on a word, and a word store (1), followed
> by an acquire barrier after control dependency. The unlock path can be
> performed with a simple store-release of 0 to the word, which does
> not require rseq.
> 
> The cpu_opv fallback requires a single-word comparison (== 0) and a
> single-word store (1).
> 
> 2) per-cpu statistics counters
> 
> A per-cpu statistics counters can be implemented as a rseq consisting
> of a final "add" instruction on a word as commit.
> 
> The cpu_opv fallback can be implemented as a "ADD" operation.
> 
> Besides statistics tracking, these counters can be used to implement
> user-space RCU per-cpu grace period tracking for both single and
> multi-process user-space RCU.
> 
> 3) per-cpu LIFO linked-list (unlimited size stack)
> 
> A per-cpu LIFO linked-list has a "push" and "pop" operation,
> which respectively adds an item to the list, and removes an
> item from the list.
> 
> The "push" operation can be implemented as a rseq consisting of
> a word comparison instruction against head followed by a word store
> (commit) to head. Its cpu_opv fallback can be implemented as a
> word-compare followed by word-store as well.
> 
> The "pop" operation can be implemented as a rseq consisting of
> loading head, comparing it against NULL, loading the next pointer
> at the right offset within the head item, and the next pointer as
> a new head, returning the old head on success.
> 
> The cpu_opv fallback for "pop" differs from its rseq algorithm:
> considering that cpu_opv requires to know all pointers at system
> call entry so it can pin all pages, so cpu_opv cannot simply load
> head and then load the head->next address within the preempt-off
> critical section. User-space needs to pass the head and head->next
> addresses to the kernel, and the kernel needs to check that the
> head address is unchanged since it has been loaded by user-space.
> However, when accessing head->next in a ABA situation, it's
> possible that head is unchanged, but loading head->next can
> result in a page fault due to a concurrently freed head object.
> This is why the "expect_fault" operation field is introduced: if a
> fault is triggered by this access, "-EAGAIN" will be returned by
> cpu_opv rather than -EFAULT, thus indicating the the operation
> vector should be attempted again. The "pop" operation can thus be
> implemented as a word comparison of head against the head loaded
> by user-space, followed by a load of the head->next pointer (which
> may fault), and a store of that pointer as a new head.
> 
> 4) per-cpu LIFO ring buffer with pointers to objects (fixed-sized stack)
> 
> This structure is useful for passing around allocated objects
> by passing pointers through per-cpu fixed-sized stack.
> 
> The "push" side can be implemented with a check of the current
> offset against the maximum buffer length, followed by a rseq
> consisting of a comparison of the previously loaded offset
> against the current offset, a word "try store" operation into the
> next ring buffer array index (it's OK to abort after a try-store,
> since it's not the commit, and its side-effect can be overwritten),
> then followed by a word-store to increment the current offset (commit).
> 
> The "push" cpu_opv fallback can be done with the comparison, and
> two consecutive word stores, all within the preempt-off section.
> 
> The "pop" side can be implemented with a check that offset is not
> 0 (whether the buffer is empty), a load of the "head" pointer before the
> offset array index, followed by a rseq consisting of a word
> comparison checking that the offset is unchanged since previously
> loaded, another check ensuring that the "head" pointer is unchanged,
> followed by a store decrementing the current offset.
> 
> The cpu_opv "pop" can be implemented with the same algorithm
> as the rseq fast-path (compare, compare, store).
> 
> 5) per-cpu LIFO ring buffer with pointers to objects (fixed-sized stack)
>   supporting "peek" from remote CPU
> 
> In order to implement work queues with work-stealing between CPUs, it is
> useful to ensure the offset "commit" in scenario 4) "push" have a
> store-release semantic, thus allowing remote CPU to load the offset
> with acquire semantic, and load the top pointer, in order to check if
> work-stealing should be performed. The task (work queue item) existence
> should be protected by other means, e.g. RCU.
> 
> If the peek operation notices that work-stealing should indeed be
> performed, a thread can use cpu_opv to move the task between per-cpu
> workqueues, by first invoking cpu_opv passing the remote work queue
> cpu number as argument to pop the task, and then again as "push" with
> the target work queue CPU number.
> 
> 6) per-cpu LIFO ring buffer with data copy (fixed-sized stack)
>   (with and without acquire-release)
> 
> This structure is useful for passing around data without requiring
> memory allocation by copying the data content into per-cpu fixed-sized
> stack.
> 
> The "push" operation is performed with an offset comparison against
> the buffer size (figuring out if the buffer is full), followed by
> a rseq consisting of a comparison of the offset, a try-memcpy attempting
> to copy the data content into the buffer (which can be aborted and
> overwritten), and a final store incrementing the offset.
> 
> The cpu_opv fallback needs to same operations, except that the memcpy
> is guaranteed to complete, given that it is performed with preemption
> disabled. This requires a memcpy operation supporting length up to 4kB.
> 
> The "pop" operation is similar to the "push, except that the offset
> is first compared to 0 to ensure the buffer is not empty. The
> copy source is the ring buffer, and the destination is an output
> buffer.
> 
> 7) per-cpu FIFO ring buffer (fixed-sized queue)
> 
> This structure is useful wherever a FIFO behavior (queue) is needed.
> One major use-case is tracer ring buffer.
> 
> An implementation of this ring buffer has a "reserve", followed by
> serialization of multiple bytes into the buffer, ended by a "commit".
> The "reserve" can be implemented as a rseq consisting of a word
> comparison followed by a word store. The reserve operation moves the
> producer "head". The multi-byte serialization can be performed
> non-atomically. Finally, the "commit" update can be performed with
> a rseq "add" commit instruction with store-release semantic. The
> ring buffer consumer reads the commit value with load-acquire
> semantic to know whenever it is safe to read from the ring buffer.
> 
> This use-case requires that both "reserve" and "commit" operations
> be performed on the same per-cpu ring buffer, even if a migration
> happens between those operations. In the typical case, both operations
> will happens on the same CPU and use rseq. In the unlikely event of a
> migration, the cpu_opv system call will ensure the commit can be
> performed on the right CPU by migrating the task to that CPU.
> 
> On the consumer side, an alternative to using store-release and
> load-acquire on the commit counter would be to use cpu_opv to
> ensure the commit counter load is performed on the right CPU. This
> effectively allows moving a consumer thread between CPUs to execute
> close to the ring buffer cache lines it will read.
> 
> Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
> CC: "Paul E. McKenney" <paulmck-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
> CC: Peter Zijlstra <peterz-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org>
> CC: Paul Turner <pjt-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>
> CC: Thomas Gleixner <tglx-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org>
> CC: Andrew Hunter <ahh-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>
> CC: Andy Lutomirski <luto-kltTT9wpgjJwATOyAt5JVQ@public.gmane.org>
> CC: Andi Kleen <andi-Vw/NltI1exuRpAAqCnN02g@public.gmane.org>
> CC: Dave Watson <davejwatson-b10kYP2dOMg@public.gmane.org>
> CC: Chris Lameter <cl-vYTEC60ixJUAvxtiuMwx3w@public.gmane.org>
> CC: Ingo Molnar <mingo-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
> CC: "H. Peter Anvin" <hpa-YMNOUZJC4hwAvxtiuMwx3w@public.gmane.org>
> CC: Ben Maurer <bmaurer-b10kYP2dOMg@public.gmane.org>
> CC: Steven Rostedt <rostedt-nx8X9YLhiw1AfugRpC6u6w@public.gmane.org>
> CC: Josh Triplett <josh-iaAMLnmF4UmaiuxdJuQwMA@public.gmane.org>
> CC: Linus Torvalds <torvalds-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>
> CC: Andrew Morton <akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>
> CC: Russell King <linux-lFZ/pmaqli7XmaaqVzeoHQ@public.gmane.org>
> CC: Catalin Marinas <catalin.marinas-5wv7dgnIgG8@public.gmane.org>
> CC: Will Deacon <will.deacon-5wv7dgnIgG8@public.gmane.org>
> CC: Michael Kerrisk <mtk.manpages-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> CC: Boqun Feng <boqun.feng-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> CC: linux-api-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> ---
> Changes since v1:
> - handle CPU hotplug,
> - cleanup implementation using function pointers: We can use function
>  pointers to implement the operations rather than duplicating all the
>  user-access code.
> - refuse device pages: Performing cpu_opv operations on io map'd pages
>  with preemption disabled could generate long preempt-off critical
>  sections, which leads to unwanted scheduler latency. Return EFAULT if
>  a device page is received as parameter
> - restrict op vector to 4216 bytes length sum: Restrict the operation
>  vector to length sum of:
>  - 4096 bytes (typical page size on most architectures, should be
>    enough for a string, or structures)
>  - 15 * 8 bytes (typical operations on integers or pointers).
>  The goal here is to keep the duration of preempt off critical section
>  short, so we don't add significant scheduler latency.
> - Add INIT_ONSTACK macro: Introduce the
>  CPU_OP_FIELD_u32_u64_INIT_ONSTACK() macros to ensure that users
>  correctly initialize the upper bits of CPU_OP_FIELD_u32_u64() on their
>  stack to 0 on 32-bit architectures.
> - Add CPU_MB_OP operation:
>  Use-cases with:
>  - two consecutive stores,
>  - a mempcy followed by a store,
>  require a memory barrier before the final store operation. A typical
>  use-case is a store-release on the final store. Given that this is a
>  slow path, just providing an explicit full barrier instruction should
>  be sufficient.
> - Add expect fault field:
>  The use-case of list_pop brings interesting challenges. With rseq, we
>  can use rseq_cmpnev_storeoffp_load(), and therefore load a pointer,
>  compare it against NULL, add an offset, and load the target "next"
>  pointer from the object, all within a single req critical section.
> 
>  Life is not so easy for cpu_opv in this use-case, mainly because we
>  need to pin all pages we are going to touch in the preempt-off
>  critical section beforehand. So we need to know the target object (in
>  which we apply an offset to fetch the next pointer) when we pin pages
>  before disabling preemption.
> 
>  So the approach is to load the head pointer and compare it against
>  NULL in user-space, before doing the cpu_opv syscall. User-space can
>  then compute the address of the head->next field, *without loading it*.
> 
>  The cpu_opv system call will first need to pin all pages associated
>  with input data. This includes the page backing the head->next object,
>  which may have been concurrently deallocated and unmapped. Therefore,
>  in this case, getting -EFAULT when trying to pin those pages may
>  happen: it just means they have been concurrently unmapped. This is
>  an expected situation, and should just return -EAGAIN to user-space,
>  to user-space can distinguish between "should retry" type of
>  situations and actual errors that should be handled with extreme
>  prejudice to the program (e.g. abort()).
> 
>  Therefore, add "expect_fault" fields along with op input address
>  pointers, so user-space can identify whether a fault when getting a
>  field should return EAGAIN rather than EFAULT.
> - Add compiler barrier between operations: Adding a compiler barrier
>  between store operations in a cpu_opv sequence can be useful when
>  paired with membarrier system call.
> 
>  An algorithm with a paired slow path and fast path can use
>  sys_membarrier on the slow path to replace fast-path memory barriers
>  by compiler barrier.
> 
>  Adding an explicit compiler barrier between operations allows
>  cpu_opv to be used as fallback for operations meant to match
>  the membarrier system call.
> 
> Changes since v2:
> 
> - Fix memory leak by introducing struct cpu_opv_pinned_pages.
>  Suggested by Boqun Feng.
> - Cast argument 1 passed to access_ok from integer to void __user *,
>  fixing sparse warning.
> 
> Changes since v3:
> 
> - Fix !SMP by adding push_task_to_cpu() empty static inline.
> - Add missing sys_cpu_opv() asmlinkage declaration to
>  include/linux/syscalls.h.
> 
> Changes since v4:
> 
> - Cleanup based on Thomas Gleixner's feedback.
> - Handle retry in case where the scheduler migrates the thread away
>  from the target CPU after migration within the syscall rather than
>  returning EAGAIN to user-space.
> - Move push_task_to_cpu() to its own patch.
> - New scheme for touching user-space memory:
>   1) get_user_pages_fast() to pin/get all pages (which can sleep),
>   2) vm_map_ram() those pages
>   3) grab mmap_sem (read lock)
>   4) __get_user_pages_fast() (or get_user_pages() on failure)
>      -> Confirm that the same page pointers are returned. This
>         catches cases where COW mappings are changed concurrently.
>      -> If page pointers differ, or on gup failure, release mmap_sem,
>         vm_unmap_ram/put_page and retry from step (1).
>      -> perform put_page on the extra reference immediately for each
>         page.
>   5) preempt disable
>   6) Perform operations on vmap. Those operations are normal
>      loads/stores/memcpy.
>   7) preempt enable
>   8) release mmap_sem
>   9) vm_unmap_ram() all virtual addresses
>  10) put_page() all pages
> - Handle architectures with VIVT caches along with vmap(): call
>  flush_kernel_vmap_range() after each "write" operation. This
>  ensures that the user-space mapping and vmap reach a consistent
>  state between each operation.
> - Depend on MMU for is_zero_pfn(). e.g. Blackfin and SH architectures
>  don't provide the zero_pfn symbol.
> 
> ---
> Man page associated:
> 
> CPU_OPV(2)              Linux Programmer's Manual             CPU_OPV(2)
> 
> NAME
>       cpu_opv - CPU preempt-off operation vector system call
> 
> SYNOPSIS
>       #include <linux/cpu_opv.h>
> 
>       int cpu_opv(struct cpu_op * cpu_opv, int cpuopcnt, int cpu, int flags);
> 
> DESCRIPTION
>       The cpu_opv system call executes a vector of operations on behalf
>       of user-space on a specific CPU with preemption disabled.
> 
>       The operations available are: comparison, memcpy, add,  or,  and,
>       xor, left shift, right shift, and memory barrier. The system call
>       receives a CPU number from user-space as argument, which  is  the
>       CPU on which those operations need to be performed.  All pointers
>       in the ops must have been set up to point to the per  CPU  memory
>       of  the CPU on which the operations should be executed. The "com‐
>       parison" operation can be used to check that the data used in the
>       preparation  step  did  not  change between preparation of system
>       call inputs and operation execution within the preempt-off criti‐
>       cal section.
> 
>       An overall maximum of 4216 bytes in enforced on the sum of opera‐
>       tion length within an operation vector, so user-space cannot gen‐
>       erate  a too long preempt-off critical section. Each operation is
>       also limited a length of 4096 bytes. A maximum limit of 16 opera‐
>       tions per cpu_opv syscall invocation is enforced.
> 
>       If the thread is not running on the requested CPU, it is migrated
>       to it.
> 
>       The layout of struct cpu_opv is as follows:
> 
>       Fields
> 
>           op Operation of type enum cpu_op_type to perform. This opera‐
>              tion type selects the associated "u" union field.
> 
>           len
>              Length (in bytes) of data to consider for this operation.
> 
>           u.compare_op
>              For a CPU_COMPARE_EQ_OP , and CPU_COMPARE_NE_OP , contains
>              the  a  and  b pointers to compare. The expect_fault_a and
>              expect_fault_b fields indicate whether a page fault should
>              be expected for each of those pointers.  If expect_fault_a
>              , or expect_fault_b is set, EAGAIN is returned  on  fault,
>              else  EFAULT is returned. The len field is allowed to take
>              values from 0 to 4096 for comparison operations.
> 
>           u.memcpy_op
>              For a CPU_MEMCPY_OP , contains the dst and  src  pointers,
>              expressing  a  copy  of src into dst. The expect_fault_dst
>              and expect_fault_src fields indicate whether a page  fault
>              should  be  expected  for  each  of  those  pointers.   If
>              expect_fault_dst , or expect_fault_src is set,  EAGAIN  is
>              returned  on fault, else EFAULT is returned. The len field
>              is allowed to take values from 0 to 4096 for memcpy opera‐
>              tions.
> 
>           u.arithmetic_op
>              For   a  CPU_ADD_OP  ,  contains  the  p  ,  count  ,  and
>              expect_fault_p fields, which are respectively a pointer to
>              the  memory location to increment, the 64-bit signed inte‐
>              ger value to add, and  whether  a  page  fault  should  be
>              expected  for  p  .   If  expect_fault_p is set, EAGAIN is
>              returned on fault, else EFAULT is returned. The len  field
>              is  allowed  to take values of 1, 2, 4, 8 bytes for arith‐
>              metic operations.
> 
>           u.bitwise_op
>              For a CPU_OR_OP , CPU_AND_OP , and CPU_XOR_OP  ,  contains
>              the  p  ,  mask  ,  and  expect_fault_p  fields, which are
>              respectively a pointer to the memory location  to  target,
>              the  mask  to  apply,  and  whether a page fault should be
>              expected for p .  If  expect_fault_p  is  set,  EAGAIN  is
>              returned  on fault, else EFAULT is returned. The len field
>              is allowed to take values of 1, 2, 4, 8 bytes for  bitwise
>              operations.
> 
>           u.shift_op
>              For a CPU_LSHIFT_OP , and CPU_RSHIFT_OP , contains the p ,
>              bits , and expect_fault_p fields, which are respectively a
>              pointer  to  the  memory location to target, the number of
>              bits to shift either left of right,  and  whether  a  page
>              fault  should  be  expected  for p .  If expect_fault_p is
>              set, EAGAIN is returned on fault, else EFAULT is returned.
>              The  len  field  is  allowed  to take values of 1, 2, 4, 8
>              bytes for shift operations. The bits field is  allowed  to
>              take values between 0 and 63.
> 
>       The enum cpu_op_types contains the following operations:
> 
>       · CPU_COMPARE_EQ_OP:  Compare  whether  two  memory locations are
>         equal,
> 
>       · CPU_COMPARE_NE_OP: Compare whether two memory locations differ,
> 
>       · CPU_MEMCPY_OP: Copy a source memory location  into  a  destina‐
>         tion,
> 
>       · CPU_ADD_OP:  Increment  a  target  memory  location  of a given
>         count,
> 
>       · CPU_OR_OP: Apply a "or" mask to a memory location,
> 
>       · CPU_AND_OP: Apply a "and" mask to a memory location,
> 
>       · CPU_XOR_OP: Apply a "xor" mask to a memory location,
> 
>       · CPU_LSHIFT_OP: Shift a memory location left of a  given  number
>         of bits,
> 
>       · CPU_RSHIFT_OP:  Shift a memory location right of a given number
>         of bits.
> 
>       · CPU_MB_OP: Issue a memory barrier.
> 
>         All of the operations above provide single-copy atomicity guar‐
>         antees  for  word-sized, word-aligned target pointers, for both
>         loads and stores.
> 
>       The cpuopcnt argument is the number of elements  in  the  cpu_opv
>       array. It can take values from 0 to 16.
> 
>       The  cpu  argument  is  the  CPU  number  on  which the operation
>       sequence needs to be executed.
> 
>       The flags argument is expected to be 0.
> 
> RETURN VALUE
>       A return value of 0 indicates success. On error, -1 is  returned,
>       and  errno is set appropriately. If a comparison operation fails,
>       execution of the operation vector  is  stopped,  and  the  return
>       value is the index after the comparison operation (values between
>       1 and 16).
> 
> ERRORS
>       EAGAIN cpu_opv() system call should be attempted again.
> 
>       EINVAL Either flags contains an invalid value, or cpu contains an
>              invalid  value  or  a  value  not  allowed  by the current
>              thread's allowed cpu mask, or cpuopcnt contains an invalid
>              value, or the cpu_opv operation vector contains an invalid
>              op value, or the  cpu_opv  operation  vector  contains  an
>              invalid  len value, or the cpu_opv operation vector sum of
>              len values is too large.
> 
>       ENOSYS The cpu_opv() system call is not implemented by this  ker‐
>              nel.
> 
>       EFAULT cpu_opv  is  an  invalid  address,  or a pointer contained
>              within an  operation  is  invalid  (and  a  fault  is  not
>              expected for that pointer).
> 
> VERSIONS
>       The cpu_opv() system call was added in Linux 4.X (TODO).
> 
> CONFORMING TO
>       cpu_opv() is Linux-specific.
> 
> SEE ALSO
>       membarrier(2), rseq(2)
> 
> Linux                          2017-11-10                     CPU_OPV(2)
> ---
> MAINTAINERS                  |    7 +
> include/linux/syscalls.h     |    3 +
> include/uapi/linux/cpu_opv.h |  114 +++++
> init/Kconfig                 |   16 +
> kernel/Makefile              |    1 +
> kernel/cpu_opv.c             | 1078 ++++++++++++++++++++++++++++++++++++++++++
> kernel/sys_ni.c              |    1 +
> 7 files changed, 1220 insertions(+)
> create mode 100644 include/uapi/linux/cpu_opv.h
> create mode 100644 kernel/cpu_opv.c
> 
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 4ede6c16d49f..36c5246b385b 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -3732,6 +3732,13 @@ B:	https://bugzilla.kernel.org
> F:	drivers/cpuidle/*
> F:	include/linux/cpuidle.h
> 
> +CPU NON-PREEMPTIBLE OPERATION VECTOR SUPPORT
> +M:	Mathieu Desnoyers <mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
> +L:	linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> +S:	Supported
> +F:	kernel/cpu_opv.c
> +F:	include/uapi/linux/cpu_opv.h
> +
> CRAMFS FILESYSTEM
> M:	Nicolas Pitre <nico-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>
> S:	Maintained
> diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
> index 340650b4ec54..32d289f41f62 100644
> --- a/include/linux/syscalls.h
> +++ b/include/linux/syscalls.h
> @@ -67,6 +67,7 @@ struct perf_event_attr;
> struct file_handle;
> struct sigaltstack;
> struct rseq;
> +struct cpu_op;
> union bpf_attr;
> 
> #include <linux/types.h>
> @@ -943,5 +944,7 @@ asmlinkage long sys_statx(int dfd, const char __user *path,
> unsigned flags,
> 			  unsigned mask, struct statx __user *buffer);
> asmlinkage long sys_rseq(struct rseq __user *rseq, uint32_t rseq_len,
> 			int flags, uint32_t sig);
> +asmlinkage long sys_cpu_opv(struct cpu_op __user *ucpuopv, int cpuopcnt,
> +			int cpu, int flags);
> 
> #endif
> diff --git a/include/uapi/linux/cpu_opv.h b/include/uapi/linux/cpu_opv.h
> new file mode 100644
> index 000000000000..ccd8167fc189
> --- /dev/null
> +++ b/include/uapi/linux/cpu_opv.h
> @@ -0,0 +1,114 @@
> +#ifndef _UAPI_LINUX_CPU_OPV_H
> +#define _UAPI_LINUX_CPU_OPV_H
> +
> +/*
> + * linux/cpu_opv.h
> + *
> + * CPU preempt-off operation vector system call API
> + *
> + * Copyright (c) 2017 Mathieu Desnoyers <mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a copy
> + * of this software and associated documentation files (the "Software"), to
> deal
> + * in the Software without restriction, including without limitation the rights
> + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
> + * copies of the Software, and to permit persons to whom the Software is
> + * furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice shall be included in
> + * all copies or substantial portions of the Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
> + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> FROM,
> + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
> THE
> + * SOFTWARE.
> + */
> +
> +#ifdef __KERNEL__
> +# include <linux/types.h>
> +#else
> +# include <stdint.h>
> +#endif
> +
> +#include <linux/types_32_64.h>
> +
> +#define CPU_OP_VEC_LEN_MAX		16
> +#define CPU_OP_ARG_LEN_MAX		24
> +/* Maximum data len per operation. */
> +#define CPU_OP_DATA_LEN_MAX		4096
> +/*
> + * Maximum data len for overall vector. Restrict the amount of user-space
> + * data touched by the kernel in non-preemptible context, so it does not
> + * introduce long scheduler latencies.
> + * This allows one copy of up to 4096 bytes, and 15 operations touching 8
> + * bytes each.
> + * This limit is applied to the sum of length specified for all operations
> + * in a vector.
> + */
> +#define CPU_OP_MEMCPY_EXPECT_LEN	4096
> +#define CPU_OP_EXPECT_LEN		8
> +#define CPU_OP_VEC_DATA_LEN_MAX		\
> +	(CPU_OP_MEMCPY_EXPECT_LEN +	\
> +	 (CPU_OP_VEC_LEN_MAX - 1) * CPU_OP_EXPECT_LEN)
> +
> +enum cpu_op_type {
> +	/* compare */
> +	CPU_COMPARE_EQ_OP,
> +	CPU_COMPARE_NE_OP,
> +	/* memcpy */
> +	CPU_MEMCPY_OP,
> +	/* arithmetic */
> +	CPU_ADD_OP,
> +	/* bitwise */
> +	CPU_OR_OP,
> +	CPU_AND_OP,
> +	CPU_XOR_OP,
> +	/* shift */
> +	CPU_LSHIFT_OP,
> +	CPU_RSHIFT_OP,
> +	/* memory barrier */
> +	CPU_MB_OP,
> +};
> +
> +/* Vector of operations to perform. Limited to 16. */
> +struct cpu_op {
> +	/* enum cpu_op_type. */
> +	int32_t op;
> +	/* data length, in bytes. */
> +	uint32_t len;
> +	union {
> +		struct {
> +			LINUX_FIELD_u32_u64(a);
> +			LINUX_FIELD_u32_u64(b);
> +			uint8_t expect_fault_a;
> +			uint8_t expect_fault_b;
> +		} compare_op;
> +		struct {
> +			LINUX_FIELD_u32_u64(dst);
> +			LINUX_FIELD_u32_u64(src);
> +			uint8_t expect_fault_dst;
> +			uint8_t expect_fault_src;
> +		} memcpy_op;
> +		struct {
> +			LINUX_FIELD_u32_u64(p);
> +			int64_t count;
> +			uint8_t expect_fault_p;
> +		} arithmetic_op;
> +		struct {
> +			LINUX_FIELD_u32_u64(p);
> +			uint64_t mask;
> +			uint8_t expect_fault_p;
> +		} bitwise_op;
> +		struct {
> +			LINUX_FIELD_u32_u64(p);
> +			uint32_t bits;
> +			uint8_t expect_fault_p;
> +		} shift_op;
> +		char __padding[CPU_OP_ARG_LEN_MAX];
> +	} u;
> +};
> +
> +#endif /* _UAPI_LINUX_CPU_OPV_H */
> diff --git a/init/Kconfig b/init/Kconfig
> index 88e36395390f..8a4995ed1d19 100644
> --- a/init/Kconfig
> +++ b/init/Kconfig
> @@ -1404,6 +1404,7 @@ config RSEQ
> 	bool "Enable rseq() system call" if EXPERT
> 	default y
> 	depends on HAVE_RSEQ
> +	select CPU_OPV
> 	select MEMBARRIER
> 	help
> 	  Enable the restartable sequences system call. It provides a
> @@ -1414,6 +1415,21 @@ config RSEQ
> 
> 	  If unsure, say Y.
> 
> +# CPU_OPV depends on MMU for is_zero_pfn()
> +config CPU_OPV
> +	bool "Enable cpu_opv() system call" if EXPERT
> +	default y
> +	depends on MMU
> +	help
> +	  Enable the CPU preempt-off operation vector system call.
> +	  It allows user-space to perform a sequence of operations on
> +	  per-cpu data with preemption disabled. Useful as
> +	  single-stepping fall-back for restartable sequences, and for
> +	  performing more complex operations on per-cpu data that would
> +	  not be otherwise possible to do with restartable sequences.
> +
> +	  If unsure, say Y.
> +
> config EMBEDDED
> 	bool "Embedded system"
> 	option allnoconfig_y
> diff --git a/kernel/Makefile b/kernel/Makefile
> index 3574669dafd9..cac8855196ff 100644
> --- a/kernel/Makefile
> +++ b/kernel/Makefile
> @@ -113,6 +113,7 @@ obj-$(CONFIG_TORTURE_TEST) += torture.o
> 
> obj-$(CONFIG_HAS_IOMEM) += memremap.o
> obj-$(CONFIG_RSEQ) += rseq.o
> +obj-$(CONFIG_CPU_OPV) += cpu_opv.o
> 
> $(obj)/configs.o: $(obj)/config_data.h
> 
> diff --git a/kernel/cpu_opv.c b/kernel/cpu_opv.c
> new file mode 100644
> index 000000000000..965fbf0a86b0
> --- /dev/null
> +++ b/kernel/cpu_opv.c
> @@ -0,0 +1,1078 @@
> +/*
> + * CPU preempt-off operation vector system call
> + *
> + * It allows user-space to perform a sequence of operations on per-cpu
> + * data with preemption disabled. Useful as single-stepping fall-back
> + * for restartable sequences, and for performing more complex operations
> + * on per-cpu data that would not be otherwise possible to do with
> + * restartable sequences.
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; either version 2 of the License, or
> + * (at your option) any later version.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> + * GNU General Public License for more details.
> + *
> + * Copyright (C) 2017, EfficiOS Inc.,
> + * Mathieu Desnoyers <mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
> + */
> +
> +#include <linux/sched.h>
> +#include <linux/uaccess.h>
> +#include <linux/syscalls.h>
> +#include <linux/cpu_opv.h>
> +#include <linux/types.h>
> +#include <linux/mutex.h>
> +#include <linux/pagemap.h>
> +#include <linux/mm.h>
> +#include <asm/ptrace.h>
> +#include <asm/byteorder.h>
> +#include <asm/cacheflush.h>
> +
> +#include "sched/sched.h"
> +
> +/*
> + * Typical invocation of cpu_opv need few virtual address pointers. Keep
> + * those in an array on the stack of the cpu_opv system call up to
> + * this limit, beyond which the array is dynamically allocated.
> + */
> +#define NR_VADDR_ON_STACK		8
> +
> +/* Maximum pages per op. */
> +#define CPU_OP_MAX_PAGES		4
> +
> +/* Maximum number of virtual addresses per op. */
> +#define CPU_OP_VEC_MAX_ADDR		(2 * CPU_OP_VEC_LEN_MAX)
> +
> +union op_fn_data {
> +	uint8_t _u8;
> +	uint16_t _u16;
> +	uint32_t _u32;
> +	uint64_t _u64;
> +#if (BITS_PER_LONG < 64)
> +	uint32_t _u64_split[2];
> +#endif
> +};
> +
> +struct vaddr {
> +	unsigned long mem;
> +	unsigned long uaddr;
> +	struct page *pages[2];
> +	unsigned int nr_pages;
> +	int write;
> +};
> +
> +struct cpu_opv_vaddr {
> +	struct vaddr *addr;
> +	size_t nr_vaddr;
> +	bool is_kmalloc;
> +};
> +
> +typedef int (*op_fn_t)(union op_fn_data *data, uint64_t v, uint32_t len);
> +
> +/*
> + * Provide mutual exclution for threads executing a cpu_opv against an
> + * offline CPU.
> + */
> +static DEFINE_MUTEX(cpu_opv_offline_lock);
> +
> +/*
> + * The cpu_opv system call executes a vector of operations on behalf of
> + * user-space on a specific CPU with preemption disabled. It is inspired
> + * by readv() and writev() system calls which take a "struct iovec"
> + * array as argument.
> + *
> + * The operations available are: comparison, memcpy, add, or, and, xor,
> + * left shift, right shift, and memory barrier. The system call receives
> + * a CPU number from user-space as argument, which is the CPU on which
> + * those operations need to be performed.  All pointers in the ops must
> + * have been set up to point to the per CPU memory of the CPU on which
> + * the operations should be executed. The "comparison" operation can be
> + * used to check that the data used in the preparation step did not
> + * change between preparation of system call inputs and operation
> + * execution within the preempt-off critical section.
> + *
> + * The reason why we require all pointer offsets to be calculated by
> + * user-space beforehand is because we need to use get_user_pages_fast()
> + * to first pin all pages touched by each operation. This takes care of
> + * faulting-in the pages. Then, preemption is disabled, and the
> + * operations are performed atomically with respect to other thread
> + * execution on that CPU, without generating any page fault.
> + *
> + * An overall maximum of 4216 bytes in enforced on the sum of operation
> + * length within an operation vector, so user-space cannot generate a
> + * too long preempt-off critical section (cache cold critical section
> + * duration measured as 4.7µs on x86-64). Each operation is also limited
> + * a length of 4096 bytes, meaning that an operation can touch a
> + * maximum of 4 pages (memcpy: 2 pages for source, 2 pages for
> + * destination if addresses are not aligned on page boundaries).
> + *
> + * If the thread is not running on the requested CPU, it is migrated to
> + * it.
> + */
> +
> +static unsigned long cpu_op_range_nr_pages(unsigned long addr,
> +					   unsigned long len)
> +{
> +	return ((addr + len - 1) >> PAGE_SHIFT) - (addr >> PAGE_SHIFT) + 1;
> +}
> +
> +static int cpu_op_count_pages(unsigned long addr, unsigned long len)
> +{
> +	unsigned long nr_pages;
> +
> +	if (!len)
> +		return 0;
> +	nr_pages = cpu_op_range_nr_pages(addr, len);
> +	if (nr_pages > 2) {
> +		WARN_ON(1);
> +		return -EINVAL;
> +	}
> +	return nr_pages;
> +}
> +
> +static struct vaddr *cpu_op_alloc_vaddr_vector(int nr_vaddr)
> +{
> +	return kzalloc(nr_vaddr * sizeof(struct vaddr), GFP_KERNEL);
> +}
> +
> +/*
> + * Check operation types and length parameters. Count number of pages.
> + */
> +static int cpu_opv_check_op(struct cpu_op *op, int *nr_vaddr, uint32_t *sum)
> +{
> +	int ret;
> +
> +	switch (op->op) {
> +	case CPU_MB_OP:
> +		break;
> +	default:
> +		*sum += op->len;
> +	}
> +
> +	/* Validate inputs. */
> +	switch (op->op) {
> +	case CPU_COMPARE_EQ_OP:
> +	case CPU_COMPARE_NE_OP:
> +	case CPU_MEMCPY_OP:
> +		if (op->len > CPU_OP_DATA_LEN_MAX)
> +			return -EINVAL;
> +		break;
> +	case CPU_ADD_OP:
> +	case CPU_OR_OP:
> +	case CPU_AND_OP:
> +	case CPU_XOR_OP:
> +		switch (op->len) {
> +		case 1:
> +		case 2:
> +		case 4:
> +		case 8:
> +			break;
> +		default:
> +			return -EINVAL;
> +		}
> +		break;
> +	case CPU_LSHIFT_OP:
> +	case CPU_RSHIFT_OP:
> +		switch (op->len) {
> +		case 1:
> +			if (op->u.shift_op.bits > 7)
> +				return -EINVAL;
> +			break;
> +		case 2:
> +			if (op->u.shift_op.bits > 15)
> +				return -EINVAL;
> +			break;
> +		case 4:
> +			if (op->u.shift_op.bits > 31)
> +				return -EINVAL;
> +			break;
> +		case 8:
> +			if (op->u.shift_op.bits > 63)
> +				return -EINVAL;
> +			break;
> +		default:
> +			return -EINVAL;
> +		}
> +		break;
> +	case CPU_MB_OP:
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +
> +	/* Count pages and virtual addresses. */
> +	switch (op->op) {
> +	case CPU_COMPARE_EQ_OP:
> +	case CPU_COMPARE_NE_OP:
> +		ret = cpu_op_count_pages(op->u.compare_op.a, op->len);
> +		if (ret < 0)
> +			return ret;
> +		ret = cpu_op_count_pages(op->u.compare_op.b, op->len);
> +		if (ret < 0)
> +			return ret;
> +		*nr_vaddr += 2;
> +		break;
> +	case CPU_MEMCPY_OP:
> +		ret = cpu_op_count_pages(op->u.memcpy_op.dst, op->len);
> +		if (ret < 0)
> +			return ret;
> +		ret = cpu_op_count_pages(op->u.memcpy_op.src, op->len);
> +		if (ret < 0)
> +			return ret;
> +		*nr_vaddr += 2;
> +		break;
> +	case CPU_ADD_OP:
> +		ret = cpu_op_count_pages(op->u.arithmetic_op.p, op->len);
> +		if (ret < 0)
> +			return ret;
> +		(*nr_vaddr)++;
> +		break;
> +	case CPU_OR_OP:
> +	case CPU_AND_OP:
> +	case CPU_XOR_OP:
> +		ret = cpu_op_count_pages(op->u.bitwise_op.p, op->len);
> +		if (ret < 0)
> +			return ret;
> +		(*nr_vaddr)++;
> +		break;
> +	case CPU_LSHIFT_OP:
> +	case CPU_RSHIFT_OP:
> +		ret = cpu_op_count_pages(op->u.shift_op.p, op->len);
> +		if (ret < 0)
> +			return ret;
> +		(*nr_vaddr)++;
> +		break;
> +	case CPU_MB_OP:
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +	return 0;
> +}
> +
> +/*
> + * Check operation types and length parameters. Count number of pages.
> + */
> +static int cpu_opv_check(struct cpu_op *cpuopv, int cpuopcnt, int *nr_vaddr)
> +{
> +	uint32_t sum = 0;
> +	int i, ret;
> +
> +	for (i = 0; i < cpuopcnt; i++) {
> +		ret = cpu_opv_check_op(&cpuopv[i], nr_vaddr, &sum);
> +		if (ret)
> +			return ret;
> +	}
> +	if (sum > CPU_OP_VEC_DATA_LEN_MAX)
> +		return -EINVAL;
> +	return 0;
> +}
> +
> +static int cpu_op_check_page(struct page *page, int write)
> +{
> +	struct address_space *mapping;
> +
> +	if (is_zone_device_page(page))
> +		return -EFAULT;
> +
> +	/*
> +	 * The page lock protects many things but in this context the page
> +	 * lock stabilizes mapping, prevents inode freeing in the shared
> +	 * file-backed region case and guards against movement to swap
> +	 * cache.
> +	 *
> +	 * Strictly speaking the page lock is not needed in all cases being
> +	 * considered here and page lock forces unnecessarily serialization
> +	 * From this point on, mapping will be re-verified if necessary and
> +	 * page lock will be acquired only if it is unavoidable
> +	 *
> +	 * Mapping checks require the head page for any compound page so the
> +	 * head page and mapping is looked up now.
> +	 */
> +	page = compound_head(page);
> +	mapping = READ_ONCE(page->mapping);
> +
> +	/*
> +	 * If page->mapping is NULL, then it cannot be a PageAnon page;
> +	 * but it might be the ZERO_PAGE (which is OK to read from), or
> +	 * in the gate area or in a special mapping (for which this
> +	 * check should fail); or it may have been a good file page when
> +	 * get_user_pages_fast found it, but truncated or holepunched or
> +	 * subjected to invalidate_complete_page2 before the page lock
> +	 * is acquired (also cases which should fail). Given that a
> +	 * reference to the page is currently held, refcount care in
> +	 * invalidate_complete_page's remove_mapping prevents
> +	 * drop_caches from setting mapping to NULL concurrently.
> +	 *
> +	 * The case to guard against is when memory pressure cause
> +	 * shmem_writepage to move the page from filecache to swapcache
> +	 * concurrently: an unlikely race, but a retry for page->mapping
> +	 * is required in that situation.
> +	 */
> +	if (!mapping) {
> +		int shmem_swizzled;
> +
> +		/*
> +		 * Check again with page lock held to guard against
> +		 * memory pressure making shmem_writepage move the page
> +		 * from filecache to swapcache.
> +		 */
> +		lock_page(page);
> +		shmem_swizzled = PageSwapCache(page) || page->mapping;
> +		unlock_page(page);
> +		if (shmem_swizzled)
> +			return -EAGAIN;
> +		/*
> +		 * It is valid to read from, but invalid to write to the
> +		 * ZERO_PAGE.
> +		 */
> +		if (!(is_zero_pfn(page_to_pfn(page)) ||
> +		      is_huge_zero_page(page)) || write) {
> +			return -EFAULT;
> +		}
> +	}
> +	return 0;
> +}
> +
> +static int cpu_op_check_pages(struct page **pages,
> +			      unsigned long nr_pages,
> +			      int write)
> +{
> +	unsigned long i;
> +
> +	for (i = 0; i < nr_pages; i++) {
> +		int ret;
> +
> +		ret = cpu_op_check_page(pages[i], write);
> +		if (ret)
> +			return ret;
> +	}
> +	return 0;
> +}
> +
> +static int cpu_op_pin_pages(unsigned long addr, unsigned long len,
> +			    struct cpu_opv_vaddr *vaddr_ptrs,
> +			    unsigned long *vaddr, int write)
> +{
> +	struct page *pages[2];
> +	int ret, nr_pages, nr_put_pages, n;
> +	unsigned long _vaddr;
> +	struct vaddr *va;
> +
> +	nr_pages = cpu_op_count_pages(addr, len);
> +	if (!nr_pages)
> +		return 0;
> +again:
> +	ret = get_user_pages_fast(addr, nr_pages, write, pages);
> +	if (ret < nr_pages) {
> +		if (ret >= 0) {
> +			nr_put_pages = ret;
> +			ret = -EFAULT;
> +		} else {
> +			nr_put_pages = 0;
> +		}
> +		goto error;
> +	}
> +	ret = cpu_op_check_pages(pages, nr_pages, write);
> +	if (ret) {
> +		nr_put_pages = nr_pages;
> +		goto error;
> +	}
> +	va = &vaddr_ptrs->addr[vaddr_ptrs->nr_vaddr++];
> +	_vaddr = (unsigned long)vm_map_ram(pages, nr_pages, numa_node_id(),
> +					   PAGE_KERNEL);
> +	if (!_vaddr) {
> +		nr_put_pages = nr_pages;
> +		ret = -ENOMEM;
> +		goto error;
> +	}
> +	va->mem = _vaddr;
> +	va->uaddr = addr;
> +	for (n = 0; n < nr_pages; n++)
> +		va->pages[n] = pages[n];
> +	va->nr_pages = nr_pages;
> +	va->write = write;
> +	*vaddr = _vaddr + (addr & ~PAGE_MASK);
> +	return 0;
> +
> +error:
> +	for (n = 0; n < nr_put_pages; n++)
> +		put_page(pages[n]);
> +	/*
> +	 * Retry if a page has been faulted in, or is being swapped in.
> +	 */
> +	if (ret == -EAGAIN)
> +		goto again;
> +	return ret;
> +}
> +
> +static int cpu_opv_pin_pages_op(struct cpu_op *op,
> +				struct cpu_opv_vaddr *vaddr_ptrs,
> +				bool *expect_fault)
> +{
> +	int ret;
> +	unsigned long vaddr = 0;
> +
> +	switch (op->op) {
> +	case CPU_COMPARE_EQ_OP:
> +	case CPU_COMPARE_NE_OP:
> +		ret = -EFAULT;
> +		*expect_fault = op->u.compare_op.expect_fault_a;
> +		if (!access_ok(VERIFY_READ,
> +			       (void __user *)op->u.compare_op.a,
> +			       op->len))
> +			return ret;
> +		ret = cpu_op_pin_pages(op->u.compare_op.a, op->len,
> +				       vaddr_ptrs, &vaddr, 0);
> +		if (ret)
> +			return ret;
> +		op->u.compare_op.a = vaddr;
> +		ret = -EFAULT;
> +		*expect_fault = op->u.compare_op.expect_fault_b;
> +		if (!access_ok(VERIFY_READ,
> +			       (void __user *)op->u.compare_op.b,
> +			       op->len))
> +			return ret;
> +		ret = cpu_op_pin_pages(op->u.compare_op.b, op->len,
> +				       vaddr_ptrs, &vaddr, 0);
> +		if (ret)
> +			return ret;
> +		op->u.compare_op.b = vaddr;
> +		break;
> +	case CPU_MEMCPY_OP:
> +		ret = -EFAULT;
> +		*expect_fault = op->u.memcpy_op.expect_fault_dst;
> +		if (!access_ok(VERIFY_WRITE,
> +			       (void __user *)op->u.memcpy_op.dst,
> +			       op->len))
> +			return ret;
> +		ret = cpu_op_pin_pages(op->u.memcpy_op.dst, op->len,
> +				       vaddr_ptrs, &vaddr, 1);
> +		if (ret)
> +			return ret;
> +		op->u.memcpy_op.dst = vaddr;
> +		ret = -EFAULT;
> +		*expect_fault = op->u.memcpy_op.expect_fault_src;
> +		if (!access_ok(VERIFY_READ,
> +			       (void __user *)op->u.memcpy_op.src,
> +			       op->len))
> +			return ret;
> +		ret = cpu_op_pin_pages(op->u.memcpy_op.src, op->len,
> +				       vaddr_ptrs, &vaddr, 0);
> +		if (ret)
> +			return ret;
> +		op->u.memcpy_op.src = vaddr;
> +		break;
> +	case CPU_ADD_OP:
> +		ret = -EFAULT;
> +		*expect_fault = op->u.arithmetic_op.expect_fault_p;
> +		if (!access_ok(VERIFY_WRITE,
> +			       (void __user *)op->u.arithmetic_op.p,
> +			       op->len))
> +			return ret;
> +		ret = cpu_op_pin_pages(op->u.arithmetic_op.p, op->len,
> +				       vaddr_ptrs, &vaddr, 1);
> +		if (ret)
> +			return ret;
> +		op->u.arithmetic_op.p = vaddr;
> +		break;
> +	case CPU_OR_OP:
> +	case CPU_AND_OP:
> +	case CPU_XOR_OP:
> +		ret = -EFAULT;
> +		*expect_fault = op->u.bitwise_op.expect_fault_p;
> +		if (!access_ok(VERIFY_WRITE,
> +			       (void __user *)op->u.bitwise_op.p,
> +			       op->len))
> +			return ret;
> +		ret = cpu_op_pin_pages(op->u.bitwise_op.p, op->len,
> +				       vaddr_ptrs, &vaddr, 1);
> +		if (ret)
> +			return ret;
> +		op->u.bitwise_op.p = vaddr;
> +		break;
> +	case CPU_LSHIFT_OP:
> +	case CPU_RSHIFT_OP:
> +		ret = -EFAULT;
> +		*expect_fault = op->u.shift_op.expect_fault_p;
> +		if (!access_ok(VERIFY_WRITE,
> +			       (void __user *)op->u.shift_op.p,
> +			       op->len))
> +			return ret;
> +		ret = cpu_op_pin_pages(op->u.shift_op.p, op->len,
> +				       vaddr_ptrs, &vaddr, 1);
> +		if (ret)
> +			return ret;
> +		op->u.shift_op.p = vaddr;
> +		break;
> +	case CPU_MB_OP:
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +	return 0;
> +}
> +
> +static int cpu_opv_pin_pages(struct cpu_op *cpuop, int cpuopcnt,
> +			     struct cpu_opv_vaddr *vaddr_ptrs)
> +{
> +	int ret, i;
> +	bool expect_fault = false;
> +
> +	/* Check access, pin pages. */
> +	for (i = 0; i < cpuopcnt; i++) {
> +		ret = cpu_opv_pin_pages_op(&cpuop[i], vaddr_ptrs,
> +					   &expect_fault);
> +		if (ret)
> +			goto error;
> +	}
> +	return 0;
> +
> +error:
> +	/*
> +	 * If faulting access is expected, return EAGAIN to user-space.
> +	 * It allows user-space to distinguish between a fault caused by
> +	 * an access which is expect to fault (e.g. due to concurrent
> +	 * unmapping of underlying memory) from an unexpected fault from
> +	 * which a retry would not recover.
> +	 */
> +	if (ret == -EFAULT && expect_fault)
> +		return -EAGAIN;
> +	return ret;
> +}
> +
> +static int __op_get(union op_fn_data *data, void *p, size_t len)
> +{
> +	switch (len) {
> +	case 1:
> +		data->_u8 = READ_ONCE(*(uint8_t *)p);
> +		break;
> +	case 2:
> +		data->_u16 = READ_ONCE(*(uint16_t *)p);
> +		break;
> +	case 4:
> +		data->_u32 = READ_ONCE(*(uint32_t *)p);
> +		break;
> +	case 8:
> +#if (BITS_PER_LONG == 64)
> +		data->_u64 = READ_ONCE(*(uint64_t *)p);
> +#else
> +	{
> +		data->_u64_split[0] = READ_ONCE(*(uint32_t *)p);
> +		data->_u64_split[1] = READ_ONCE(*((uint32_t *)p + 1));
> +	}
> +#endif
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +	return 0;
> +}
> +
> +static int __op_put(union op_fn_data *data, void *p, size_t len)
> +{
> +	switch (len) {
> +	case 1:
> +		WRITE_ONCE(*(uint8_t *)p, data->_u8);
> +		break;
> +	case 2:
> +		WRITE_ONCE(*(uint16_t *)p, data->_u16);
> +		break;
> +	case 4:
> +		WRITE_ONCE(*(uint32_t *)p, data->_u32);
> +		break;
> +	case 8:
> +#if (BITS_PER_LONG == 64)
> +		WRITE_ONCE(*(uint64_t *)p, data->_u64);
> +#else
> +	{
> +		WRITE_ONCE(*(uint32_t *)p, data->_u64_split[0]);
> +		WRITE_ONCE(*((uint32_t *)p + 1), data->_u64_split[1]);
> +	}
> +#endif
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +	flush_kernel_vmap_range(p, len);
> +	return 0;
> +}
> +
> +/* Return 0 if same, > 0 if different, < 0 on error. */
> +static int do_cpu_op_compare(unsigned long _a, unsigned long _b, uint32_t len)
> +{
> +	void *a = (void *)_a;
> +	void *b = (void *)_b;
> +	union op_fn_data tmp[2];
> +	int ret;
> +
> +	switch (len) {
> +	case 1:
> +	case 2:
> +	case 4:
> +	case 8:
> +		if (!IS_ALIGNED(_a, len) || !IS_ALIGNED(_b, len))
> +			goto memcmp;
> +		break;
> +	default:
> +		goto memcmp;
> +	}
> +
> +	ret = __op_get(&tmp[0], a, len);
> +	if (ret)
> +		return ret;
> +	ret = __op_get(&tmp[1], b, len);
> +	if (ret)
> +		return ret;
> +
> +	switch (len) {
> +	case 1:
> +		ret = !!(tmp[0]._u8 != tmp[1]._u8);
> +		break;
> +	case 2:
> +		ret = !!(tmp[0]._u16 != tmp[1]._u16);
> +		break;
> +	case 4:
> +		ret = !!(tmp[0]._u32 != tmp[1]._u32);
> +		break;
> +	case 8:
> +		ret = !!(tmp[0]._u64 != tmp[1]._u64);
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +	return ret;
> +
> +memcmp:
> +	if (memcmp(a, b, len))
> +		return 1;
> +	return 0;
> +}
> +
> +/* Return 0 on success, < 0 on error. */
> +static int do_cpu_op_memcpy(unsigned long _dst, unsigned long _src,
> +			    uint32_t len)
> +{
> +	void *dst = (void *)_dst;
> +	void *src = (void *)_src;
> +	union op_fn_data tmp;
> +	int ret;
> +
> +	switch (len) {
> +	case 1:
> +	case 2:
> +	case 4:
> +	case 8:
> +		if (!IS_ALIGNED(_dst, len) || !IS_ALIGNED(_src, len))
> +			goto memcpy;
> +		break;
> +	default:
> +		goto memcpy;
> +	}
> +
> +	ret = __op_get(&tmp, src, len);
> +	if (ret)
> +		return ret;
> +	return __op_put(&tmp, dst, len);
> +
> +memcpy:
> +	memcpy(dst, src, len);
> +	flush_kernel_vmap_range(dst, len);
> +	return 0;
> +}
> +
> +static int op_add_fn(union op_fn_data *data, uint64_t count, uint32_t len)
> +{
> +	switch (len) {
> +	case 1:
> +		data->_u8 += (uint8_t)count;
> +		break;
> +	case 2:
> +		data->_u16 += (uint16_t)count;
> +		break;
> +	case 4:
> +		data->_u32 += (uint32_t)count;
> +		break;
> +	case 8:
> +		data->_u64 += (uint64_t)count;
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +	return 0;
> +}
> +
> +static int op_or_fn(union op_fn_data *data, uint64_t mask, uint32_t len)
> +{
> +	switch (len) {
> +	case 1:
> +		data->_u8 |= (uint8_t)mask;
> +		break;
> +	case 2:
> +		data->_u16 |= (uint16_t)mask;
> +		break;
> +	case 4:
> +		data->_u32 |= (uint32_t)mask;
> +		break;
> +	case 8:
> +		data->_u64 |= (uint64_t)mask;
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +	return 0;
> +}
> +
> +static int op_and_fn(union op_fn_data *data, uint64_t mask, uint32_t len)
> +{
> +	switch (len) {
> +	case 1:
> +		data->_u8 &= (uint8_t)mask;
> +		break;
> +	case 2:
> +		data->_u16 &= (uint16_t)mask;
> +		break;
> +	case 4:
> +		data->_u32 &= (uint32_t)mask;
> +		break;
> +	case 8:
> +		data->_u64 &= (uint64_t)mask;
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +	return 0;
> +}
> +
> +static int op_xor_fn(union op_fn_data *data, uint64_t mask, uint32_t len)
> +{
> +	switch (len) {
> +	case 1:
> +		data->_u8 ^= (uint8_t)mask;
> +		break;
> +	case 2:
> +		data->_u16 ^= (uint16_t)mask;
> +		break;
> +	case 4:
> +		data->_u32 ^= (uint32_t)mask;
> +		break;
> +	case 8:
> +		data->_u64 ^= (uint64_t)mask;
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +	return 0;
> +}
> +
> +static int op_lshift_fn(union op_fn_data *data, uint64_t bits, uint32_t len)
> +{
> +	switch (len) {
> +	case 1:
> +		data->_u8 <<= (uint8_t)bits;
> +		break;
> +	case 2:
> +		data->_u16 <<= (uint16_t)bits;
> +		break;
> +	case 4:
> +		data->_u32 <<= (uint32_t)bits;
> +		break;
> +	case 8:
> +		data->_u64 <<= (uint64_t)bits;
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +	return 0;
> +}
> +
> +static int op_rshift_fn(union op_fn_data *data, uint64_t bits, uint32_t len)
> +{
> +	switch (len) {
> +	case 1:
> +		data->_u8 >>= (uint8_t)bits;
> +		break;
> +	case 2:
> +		data->_u16 >>= (uint16_t)bits;
> +		break;
> +	case 4:
> +		data->_u32 >>= (uint32_t)bits;
> +		break;
> +	case 8:
> +		data->_u64 >>= (uint64_t)bits;
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +	return 0;
> +}
> +
> +/* Return 0 on success, < 0 on error. */
> +static int do_cpu_op_fn(op_fn_t op_fn, unsigned long _p, uint64_t v,
> +			uint32_t len)
> +{
> +	union op_fn_data tmp;
> +	void *p = (void *)_p;
> +	int ret;
> +
> +	ret = __op_get(&tmp, p, len);
> +	if (ret)
> +		return ret;
> +	ret = op_fn(&tmp, v, len);
> +	if (ret)
> +		return ret;
> +	ret = __op_put(&tmp, p, len);
> +	if (ret)
> +		return ret;
> +	return 0;
> +}
> +
> +/*
> + * Return negative value on error, positive value if comparison
> + * fails, 0 on success.
> + */
> +static int __do_cpu_opv_op(struct cpu_op *op)
> +{
> +	/* Guarantee a compiler barrier between each operation. */
> +	barrier();
> +
> +	switch (op->op) {
> +	case CPU_COMPARE_EQ_OP:
> +		return do_cpu_op_compare(op->u.compare_op.a,
> +					 op->u.compare_op.b,
> +					 op->len);
> +	case CPU_COMPARE_NE_OP:
> +	{
> +		int ret;
> +
> +		ret = do_cpu_op_compare(op->u.compare_op.a,
> +					op->u.compare_op.b,
> +					op->len);
> +		if (ret < 0)
> +			return ret;
> +		/*
> +		 * Stop execution, return positive value if comparison
> +		 * is identical.
> +		 */
> +		if (ret == 0)
> +			return 1;
> +		return 0;
> +	}
> +	case CPU_MEMCPY_OP:
> +		return do_cpu_op_memcpy(op->u.memcpy_op.dst,
> +					op->u.memcpy_op.src,
> +					op->len);
> +	case CPU_ADD_OP:
> +		return do_cpu_op_fn(op_add_fn, op->u.arithmetic_op.p,
> +				    op->u.arithmetic_op.count, op->len);
> +	case CPU_OR_OP:
> +		return do_cpu_op_fn(op_or_fn, op->u.bitwise_op.p,
> +				    op->u.bitwise_op.mask, op->len);
> +	case CPU_AND_OP:
> +		return do_cpu_op_fn(op_and_fn, op->u.bitwise_op.p,
> +				    op->u.bitwise_op.mask, op->len);
> +	case CPU_XOR_OP:
> +		return do_cpu_op_fn(op_xor_fn, op->u.bitwise_op.p,
> +				    op->u.bitwise_op.mask, op->len);
> +	case CPU_LSHIFT_OP:
> +		return do_cpu_op_fn(op_lshift_fn, op->u.shift_op.p,
> +				    op->u.shift_op.bits, op->len);
> +	case CPU_RSHIFT_OP:
> +		return do_cpu_op_fn(op_rshift_fn, op->u.shift_op.p,
> +				    op->u.shift_op.bits, op->len);
> +	case CPU_MB_OP:
> +		/* Memory barrier provided by this operation. */
> +		smp_mb();
> +		return 0;
> +	default:
> +		return -EINVAL;
> +	}
> +}
> +
> +static int __do_cpu_opv(struct cpu_op *cpuop, int cpuopcnt)
> +{
> +	int i, ret;
> +
> +	for (i = 0; i < cpuopcnt; i++) {
> +		ret = __do_cpu_opv_op(&cpuop[i]);
> +		/* If comparison fails, stop execution and return index + 1. */
> +		if (ret > 0)
> +			return i + 1;
> +		/* On error, stop execution. */
> +		if (ret < 0)
> +			return ret;
> +	}
> +	return 0;
> +}
> +
> +/*
> + * Check that the page pointers pinned by get_user_pages_fast()
> + * are still in the page table. Invoked with mmap_sem held.
> + * Return 0 if pointers match, -EAGAIN if they don't.
> + */
> +static int vaddr_check(struct vaddr *vaddr)
> +{
> +	struct page *pages[2];
> +	int ret, n;
> +
> +	ret = __get_user_pages_fast(vaddr->uaddr, vaddr->nr_pages,
> +				    vaddr->write, pages);
> +	for (n = 0; n < ret; n++)
> +		put_page(pages[n]);
> +	if (ret < vaddr->nr_pages) {
> +		ret = get_user_pages(vaddr->uaddr, vaddr->nr_pages,
> +				     vaddr->write ? FOLL_WRITE : 0,
> +				     pages, NULL);
> +		if (ret < 0)
> +			return -EAGAIN;
> +		for (n = 0; n < ret; n++)
> +			put_page(pages[n]);
> +		if (ret < vaddr->nr_pages)
> +			return -EAGAIN;
> +	}
> +	for (n = 0; n < vaddr->nr_pages; n++) {
> +		if (pages[n] != vaddr->pages[n])
> +			return -EAGAIN;
> +	}
> +	return 0;
> +}
> +
> +static int vaddr_ptrs_check(struct cpu_opv_vaddr *vaddr_ptrs)
> +{
> +	int i;
> +
> +	for (i = 0; i < vaddr_ptrs->nr_vaddr; i++) {
> +		int ret;
> +
> +		ret = vaddr_check(&vaddr_ptrs->addr[i]);
> +		if (ret)
> +			return ret;
> +	}
> +	return 0;
> +}
> +
> +static int do_cpu_opv(struct cpu_op *cpuop, int cpuopcnt,
> +		      struct cpu_opv_vaddr *vaddr_ptrs, int cpu)
> +{
> +	struct mm_struct *mm = current->mm;
> +	int ret;
> +
> +retry:
> +	if (cpu != raw_smp_processor_id()) {
> +		ret = push_task_to_cpu(current, cpu);
> +		if (ret)
> +			goto check_online;
> +	}
> +	down_read(&mm->mmap_sem);
> +	ret = vaddr_ptrs_check(vaddr_ptrs);
> +	if (ret)
> +		goto end;
> +	preempt_disable();
> +	if (cpu != smp_processor_id()) {
> +		preempt_enable();
> +		up_read(&mm->mmap_sem);
> +		goto retry;
> +	}
> +	ret = __do_cpu_opv(cpuop, cpuopcnt);
> +	preempt_enable();
> +end:
> +	up_read(&mm->mmap_sem);
> +	return ret;
> +
> +check_online:
> +	if (!cpu_possible(cpu))
> +		return -EINVAL;
> +	get_online_cpus();
> +	if (cpu_online(cpu)) {
> +		put_online_cpus();
> +		goto retry;
> +	}
> +	/*
> +	 * CPU is offline. Perform operation from the current CPU with
> +	 * cpu_online read lock held, preventing that CPU from coming online,
> +	 * and with mutex held, providing mutual exclusion against other
> +	 * CPUs also finding out about an offline CPU.
> +	 */
> +	down_read(&mm->mmap_sem);
> +	ret = vaddr_ptrs_check(vaddr_ptrs);
> +	if (ret)
> +		goto offline_end;
> +	mutex_lock(&cpu_opv_offline_lock);
> +	ret = __do_cpu_opv(cpuop, cpuopcnt);
> +	mutex_unlock(&cpu_opv_offline_lock);
> +offline_end:
> +	up_read(&mm->mmap_sem);
> +	put_online_cpus();
> +	return ret;
> +}
> +
> +/*
> + * cpu_opv - execute operation vector on a given CPU with preempt off.
> + *
> + * Userspace should pass current CPU number as parameter.
> + */
> +SYSCALL_DEFINE4(cpu_opv, struct cpu_op __user *, ucpuopv, int, cpuopcnt,
> +		int, cpu, int, flags)
> +{
> +	struct vaddr vaddr_on_stack[NR_VADDR_ON_STACK];
> +	struct cpu_op cpuopv[CPU_OP_VEC_LEN_MAX];
> +	struct cpu_opv_vaddr vaddr_ptrs = {
> +		.addr = vaddr_on_stack,
> +		.nr_vaddr = 0,
> +		.is_kmalloc = false,
> +	};
> +	int ret, i, nr_vaddr = 0;
> +	bool retry = false;
> +
> +	if (unlikely(flags))
> +		return -EINVAL;
> +	if (unlikely(cpu < 0))
> +		return -EINVAL;
> +	if (cpuopcnt < 0 || cpuopcnt > CPU_OP_VEC_LEN_MAX)
> +		return -EINVAL;
> +	if (copy_from_user(cpuopv, ucpuopv, cpuopcnt * sizeof(struct cpu_op)))
> +		return -EFAULT;
> +	ret = cpu_opv_check(cpuopv, cpuopcnt, &nr_vaddr);
> +	if (ret)
> +		return ret;
> +	if (nr_vaddr > NR_VADDR_ON_STACK) {
> +		vaddr_ptrs.addr = cpu_op_alloc_vaddr_vector(nr_vaddr);
> +		if (!vaddr_ptrs.addr) {
> +			ret = -ENOMEM;
> +			goto end;
> +		}
> +		vaddr_ptrs.is_kmalloc = true;
> +	}
> +again:
> +	ret = cpu_opv_pin_pages(cpuopv, cpuopcnt, &vaddr_ptrs);
> +	if (ret)
> +		goto end;
> +	ret = do_cpu_opv(cpuopv, cpuopcnt, &vaddr_ptrs, cpu);
> +	if (ret == -EAGAIN)
> +		retry = true;
> +end:
> +	for (i = 0; i < vaddr_ptrs.nr_vaddr; i++) {
> +		struct vaddr *vaddr = &vaddr_ptrs.addr[i];
> +		int j;
> +
> +		vm_unmap_ram((void *)vaddr->mem, vaddr->nr_pages);
> +		for (j = 0; j < vaddr->nr_pages; j++) {
> +			if (vaddr->write)
> +				set_page_dirty(vaddr->pages[j]);
> +			put_page(vaddr->pages[j]);
> +		}
> +	}
> +	if (retry) {
> +		retry = false;
> +		vaddr_ptrs.nr_vaddr = 0;
> +		goto again;
> +	}
> +	if (vaddr_ptrs.is_kmalloc)
> +		kfree(vaddr_ptrs.addr);
> +	return ret;
> +}
> diff --git a/kernel/sys_ni.c b/kernel/sys_ni.c
> index bfa1ee1bf669..59e622296dc3 100644
> --- a/kernel/sys_ni.c
> +++ b/kernel/sys_ni.c
> @@ -262,3 +262,4 @@ cond_syscall(sys_pkey_free);
> 
> /* restartable sequence */
> cond_syscall(sys_rseq);
> +cond_syscall(sys_cpu_opv);
> --
> 2.11.0

-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com