]> git.karo-electronics.de Git - karo-tx-linux.git/commit
cputimer: Cure lock inversion
authorPeter Zijlstra <a.p.zijlstra@chello.nl>
Mon, 17 Oct 2011 09:50:30 +0000 (11:50 +0200)
committerGreg Kroah-Hartman <gregkh@suse.de>
Tue, 25 Oct 2011 05:10:14 +0000 (07:10 +0200)
commit607ce3ed1cf6aa5af82a12faf1adf0de5d2f583e
treec362e60a92b358b460147092ea479bb3f52055c8
parentf62f4cad504f32e9d4e407ea5ba8a743da399a8a
cputimer: Cure lock inversion

commit bcd5cff7216f9b2de0a148cc355eac199dc6f1cf upstream.

There's a lock inversion between the cputimer->lock and rq->lock;
notably the two callchains involved are:

 update_rlimit_cpu()
   sighand->siglock
   set_process_cpu_timer()
     cpu_timer_sample_group()
       thread_group_cputimer()
         cputimer->lock
         thread_group_cputime()
           task_sched_runtime()
             ->pi_lock
             rq->lock

 scheduler_tick()
   rq->lock
   task_tick_fair()
     update_curr()
       account_group_exec()
         cputimer->lock

Where the first one is enabling a CLOCK_PROCESS_CPUTIME_ID timer, and
the second one is keeping up-to-date.

This problem was introduced by e8abccb7193 ("posix-cpu-timers: Cure
SMP accounting oddities").

Cure the problem by removing the cputimer->lock and rq->lock nesting,
this leaves concurrent enablers doing duplicate work, but the time
wasted should be on the same order otherwise wasted spinning on the
lock and the greater-than assignment filter should ensure we preserve
monotonicity.

Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Simon Kirby <sim@hostway.ca>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Link: http://lkml.kernel.org/r/1318928713.21167.4.camel@twins
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
kernel/posix-cpu-timers.c