sched: 'Annotate' migrate_tasks()

Kernel testing triggered this warning:

| WARNING: CPU: 0 PID: 13 at kernel/sched/core.c:1156 do_set_cpus_allowed+0x7e/0x80()
| Modules linked in:
| CPU: 0 PID: 13 Comm: migration/0 Not tainted 4.2.0-rc1-00049-g25834c7 #2
| Call Trace:
|   dump_stack+0x4b/0x75
|   warn_slowpath_common+0x8b/0xc0
|   warn_slowpath_null+0x22/0x30
|   do_set_cpus_allowed+0x7e/0x80
|   cpuset_cpus_allowed_fallback+0x7c/0x170
|   select_fallback_rq+0x221/0x280
|   migration_call+0xe3/0x250
|   notifier_call_chain+0x53/0x70
|   __raw_notifier_call_chain+0x1e/0x30
|   cpu_notify+0x28/0x50
|   take_cpu_down+0x22/0x40
|   multi_cpu_stop+0xd5/0x140
|   cpu_stopper_thread+0xbc/0x170
|   smpboot_thread_fn+0x174/0x2f0
|   kthread+0xc4/0xe0
|   ret_from_kernel_thread+0x21/0x30

As Peterz pointed out:

| So the normal rules for changing task_struct::cpus_allowed are holding
| both pi_lock and rq->lock, such that holding either stabilizes the mask.
|
| This is so that wakeup can happen without rq->lock and load-balance
| without pi_lock.
|
| From this we already get the relaxation that we can omit acquiring
| rq->lock if the task is not on the rq, because in that case
| load-balancing will not apply to it.
|
| ** these are the rules currently tested in do_set_cpus_allowed() **
|
| Now, since __set_cpus_allowed_ptr() uses task_rq_lock() which
| unconditionally acquires both locks, we could get away with holding just
| rq->lock when on_rq for modification because that'd still exclude
| __set_cpus_allowed_ptr(), it would also work against
| __kthread_bind_mask() because that assumes !on_rq.
|
| That said, this is all somewhat fragile.
|
| Now, I don't think dropping rq->lock is quite as disastrous as it
| usually is because !cpu_active at this point, which means load-balance
| will not interfere, but that too is somewhat fragile.
|
| So we end up with a choice of two fragile..

This patch fixes it by following the rules for changing
task_struct::cpus_allowed with both pi_lock and rq->lock held.

Reported-by: kernel test robot <ying.huang@intel.com>
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Wanpeng Li <wanpeng.li@hotmail.com>
[ Modified changelog and patch. ]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/BLU436-SMTP1660820490DE202E3934ED3806E0@phx.gbl
Signed-off-by: Ingo Molnar <mingo@kernel.org>

1 files changed, 26 insertions, 3 deletions

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 0902e4d72671..9b786704d34b 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -5183,24 +5183,47 @@ static void migrate_tasks(struct rq *dead_rq)
 			break;
 
 		/*
-		 * Ensure rq->lock covers the entire task selection
-		 * until the migration.
+		 * pick_next_task assumes pinned rq->lock.
 		 */
 		lockdep_pin_lock(&rq->lock);
 		next = pick_next_task(rq, &fake_task);
 		BUG_ON(!next);
 		next->sched_class->put_prev_task(rq, next);
 
+		/*
+		 * Rules for changing task_struct::cpus_allowed are holding
+		 * both pi_lock and rq->lock, such that holding either
+		 * stabilizes the mask.
+		 *
+		 * Drop rq->lock is not quite as disastrous as it usually is
+		 * because !cpu_active at this point, which means load-balance
+		 * will not interfere. Also, stop-machine.
+		 */
+		lockdep_unpin_lock(&rq->lock);
+		raw_spin_unlock(&rq->lock);
+		raw_spin_lock(&next->pi_lock);
+		raw_spin_lock(&rq->lock);
+
+		/*
+		 * Since we're inside stop-machine, _nothing_ should have
+		 * changed the task, WARN if weird stuff happened, because in
+		 * that case the above rq->lock drop is a fail too.
+		 */
+		if (WARN_ON(task_rq(next) != rq || !task_on_rq_queued(next))) {
+			raw_spin_unlock(&next->pi_lock);
+			continue;
+		}
+
 		/* Find suitable destination for @next, with force if needed. */
 		dest_cpu = select_fallback_rq(dead_rq->cpu, next);
 
-		lockdep_unpin_lock(&rq->lock);
 		rq = __migrate_task(rq, next, dest_cpu);
 		if (rq != dead_rq) {
 			raw_spin_unlock(&rq->lock);
 			rq = dead_rq;
 			raw_spin_lock(&rq->lock);
 		}
+		raw_spin_unlock(&next->pi_lock);
 	}
 
 	rq->stop = stop;