sched: Make dl_task_time() use task_rq_lock()
authorPeter Zijlstra <peterz@infradead.org>
Tue, 17 Feb 2015 12:22:25 +0000 (13:22 +0100)
committerIngo Molnar <mingo@kernel.org>
Wed, 18 Feb 2015 13:27:30 +0000 (14:27 +0100)
commit3960c8c0c7891dfc0f7be687cbdabb0d6916d614
treecfd3caba8980a19d579a74e57d220d9fe9850732
parent74b8a4cb6ce3685049ee124243a52238c5cabe55
sched: Make dl_task_time() use task_rq_lock()

Kirill reported that a dl task can be throttled and dequeued at the
same time. This happens, when it becomes throttled in schedule(),
which is called to go to sleep:

current->state = TASK_INTERRUPTIBLE;
schedule()
    deactivate_task()
        dequeue_task_dl()
            update_curr_dl()
                start_dl_timer()
            __dequeue_task_dl()
    prev->on_rq = 0;

This invalidates the assumption from commit 0f397f2c90ce ("sched/dl:
Fix race in dl_task_timer()"):

  "The only reason we don't strictly need ->pi_lock now is because
   we're guaranteed to have p->state == TASK_RUNNING here and are
   thus free of ttwu races".

And therefore we have to use the full task_rq_lock() here.

This further amends the fact that we forgot to update the rq lock loop
for TASK_ON_RQ_MIGRATE, from commit cca26e8009d1 ("sched: Teach
scheduler to understand TASK_ON_RQ_MIGRATING state").

Reported-by: Kirill Tkhai <ktkhai@parallels.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Juri Lelli <juri.lelli@arm.com>
Link: http://lkml.kernel.org/r/20150217123139.GN5029@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/sched/core.c
kernel/sched/deadline.c
kernel/sched/sched.h