From: Steven Rostedt <rostedt@goodmis.org>
To: linux-kernel@vger.kernel.org,
linux-rt-users <linux-rt-users@vger.kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>,
Carsten Emde <C.Emde@osadl.org>,
Sebastian Andrzej Siewior <bigeasy@linutronix.de>,
John Kacur <jkacur@redhat.com>, Daniel Wagner <wagi@monom.org>,
Tom Zanussi <zanussi@kernel.org>,
"Srivatsa S. Bhat" <srivatsa@csail.mit.edu>,
stable@kernel.org,
"Peter Zijlstra (Intel)" <peterz@infradead.org>,
Ingo Molnar <mingo@kernel.org>,
Valentin Schneider <valentin.schneider@arm.com>,
Paul Gortmaker <paul.gortmaker@windriver.com>
Subject: [PATCH RT 1/8] sched: Fix migration_cpu_stop() requeueing
Date: Fri, 09 Jul 2021 17:59:54 -0400 [thread overview]
Message-ID: <20210709220017.262194093@goodmis.org> (raw)
In-Reply-To: 20210709215953.122804544@goodmis.org
5.10.47-rt46-rc1 stable review patch.
If anyone has any objections, please let me know.
------------------
From: Peter Zijlstra <peterz@infradead.org>
commit 8a6edb5257e2a84720fe78cb179eca58ba76126f upstream.
When affine_move_task(p) is called on a running task @p, which is not
otherwise already changing affinity, we'll first set
p->migration_pending and then do:
stop_one_cpu(cpu_of_rq(rq), migration_cpu_stop, &arg);
This then gets us to migration_cpu_stop() running on the CPU that was
previously running our victim task @p.
If we find that our task is no longer on that runqueue (this can
happen because of a concurrent migration due to load-balance etc.),
then we'll end up at the:
} else if (dest_cpu < 1 || pending) {
branch. Which we'll take because we set pending earlier. Here we first
check if the task @p has already satisfied the affinity constraints,
if so we bail early [A]. Otherwise we'll reissue migration_cpu_stop()
onto the CPU that is now hosting our task @p:
stop_one_cpu_nowait(cpu_of(rq), migration_cpu_stop,
&pending->arg, &pending->stop_work);
Except, we've never initialized pending->arg, which will be all 0s.
This then results in running migration_cpu_stop() on the next CPU with
arg->p == NULL, which gives the by now obvious result of fireworks.
The cure is to change affine_move_task() to always use pending->arg,
furthermore we can use the exact same pattern as the
SCA_MIGRATE_ENABLE case, since we'll block on the pending->done
completion anyway, no point in adding yet another completion in
stop_one_cpu().
This then gives a clear distinction between the two
migration_cpu_stop() use cases:
- sched_exec() / migrate_task_to() : arg->pending == NULL
- affine_move_task() : arg->pending != NULL;
And we can have it ignore p->migration_pending when !arg->pending. Any
stop work from sched_exec() / migrate_task_to() is in addition to stop
works from affine_move_task(), which will be sufficient to issue the
completion.
Fixes: 6d337eab041d ("sched: Fix migrate_disable() vs set_cpus_allowed_ptr()")
Cc: stable@kernel.org
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Valentin Schneider <valentin.schneider@arm.com>
Link: https://lkml.kernel.org/r/20210224131355.357743989@infradead.org
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
---
kernel/sched/core.c | 39 ++++++++++++++++++++++++++++-----------
1 file changed, 28 insertions(+), 11 deletions(-)
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 3d3aa9db1548..a3dea38f410a 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -1958,6 +1958,24 @@ static int migration_cpu_stop(void *data)
rq_lock(rq, &rf);
pending = p->migration_pending;
+ if (pending && !arg->pending) {
+ /*
+ * This happens from sched_exec() and migrate_task_to(),
+ * neither of them care about pending and just want a task to
+ * maybe move about.
+ *
+ * Even if there is a pending, we can ignore it, since
+ * affine_move_task() will have it's own stop_work's in flight
+ * which will manage the completion.
+ *
+ * Notably, pending doesn't need to match arg->pending. This can
+ * happen when tripple concurrent affine_move_task() first sets
+ * pending, then clears pending and eventually sets another
+ * pending.
+ */
+ pending = NULL;
+ }
+
/*
* If task_rq(p) != rq, it cannot be migrated here, because we're
* holding rq->lock, if p->on_rq == 0 it cannot get enqueued because
@@ -2230,10 +2248,6 @@ static int affine_move_task(struct rq *rq, struct task_struct *p, struct rq_flag
int dest_cpu, unsigned int flags)
{
struct set_affinity_pending my_pending = { }, *pending = NULL;
- struct migration_arg arg = {
- .task = p,
- .dest_cpu = dest_cpu,
- };
bool complete = false;
/* Can the task run on the task's current CPU? If so, we're done */
@@ -2271,6 +2285,12 @@ static int affine_move_task(struct rq *rq, struct task_struct *p, struct rq_flag
/* Install the request */
refcount_set(&my_pending.refs, 1);
init_completion(&my_pending.done);
+ my_pending.arg = (struct migration_arg) {
+ .task = p,
+ .dest_cpu = -1, /* any */
+ .pending = &my_pending,
+ };
+
p->migration_pending = &my_pending;
} else {
pending = p->migration_pending;
@@ -2301,12 +2321,6 @@ static int affine_move_task(struct rq *rq, struct task_struct *p, struct rq_flag
p->migration_flags &= ~MDF_PUSH;
task_rq_unlock(rq, p, rf);
- pending->arg = (struct migration_arg) {
- .task = p,
- .dest_cpu = -1,
- .pending = pending,
- };
-
stop_one_cpu_nowait(cpu_of(rq), migration_cpu_stop,
&pending->arg, &pending->stop_work);
@@ -2319,8 +2333,11 @@ static int affine_move_task(struct rq *rq, struct task_struct *p, struct rq_flag
* is_migration_disabled(p) checks to the stopper, which will
* run on the same CPU as said p.
*/
+ refcount_inc(&pending->refs); /* pending->{arg,stop_work} */
task_rq_unlock(rq, p, rf);
- stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
+
+ stop_one_cpu_nowait(cpu_of(rq), migration_cpu_stop,
+ &pending->arg, &pending->stop_work);
} else {
--
2.30.2
next prev parent reply other threads:[~2021-07-09 22:00 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-07-09 21:59 [PATCH RT 0/8] Linux 5.10.47-rt46-rc1 Steven Rostedt
2021-07-09 21:59 ` Steven Rostedt [this message]
2021-07-09 21:59 ` [PATCH RT 2/8] sched: Simplify migration_cpu_stop() Steven Rostedt
2021-07-09 21:59 ` [PATCH RT 3/8] sched: Collate affine_move_task() stoppers Steven Rostedt
2021-07-09 21:59 ` [PATCH RT 4/8] sched: Optimize migration_cpu_stop() Steven Rostedt
2021-07-09 21:59 ` [PATCH RT 5/8] sched: Fix affine_move_task() self-concurrency Steven Rostedt
2021-07-25 5:03 ` Pavel Machek
2021-07-26 13:39 ` Valentin Schneider
2021-07-26 16:07 ` Paul Gortmaker
2021-07-09 21:59 ` [PATCH RT 6/8] sched: Simplify set_affinity_pending refcounts Steven Rostedt
2021-07-09 22:00 ` [PATCH RT 7/8] sched: Dont defer CPU pick to migration_cpu_stop() Steven Rostedt
2021-07-09 22:00 ` [PATCH RT 8/8] Linux 5.10.47-rt46-rc1 Steven Rostedt
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210709220017.262194093@goodmis.org \
--to=rostedt@goodmis.org \
--cc=C.Emde@osadl.org \
--cc=bigeasy@linutronix.de \
--cc=jkacur@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-rt-users@vger.kernel.org \
--cc=mingo@kernel.org \
--cc=paul.gortmaker@windriver.com \
--cc=peterz@infradead.org \
--cc=srivatsa@csail.mit.edu \
--cc=stable@kernel.org \
--cc=tglx@linutronix.de \
--cc=valentin.schneider@arm.com \
--cc=wagi@monom.org \
--cc=zanussi@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).