sched/rt: Fix lockdep annotation within find_lock_lowest_rq()
authorPeter Zijlstra <a.p.zijlstra@chello.nl>
Thu, 17 May 2012 19:19:46 +0000 (21:19 +0200)
committerIngo Molnar <mingo@kernel.org>
Wed, 6 Jun 2012 14:52:26 +0000 (16:52 +0200)
commit7f1b43936f0ecad14770634c021cf4a929aec74d
treeea200fcf70af41839e049c59a2dc922407e7495b
parent10717dcde10d09f9fcee53a12a4236af1a82b484
sched/rt: Fix lockdep annotation within find_lock_lowest_rq()

Roland Dreier reported spurious, hard to trigger lockdep warnings
within the scheduler - without any real lockup.

This bit gives us the right clue:

> [89945.640512]  [<ffffffff8103fa1a>] double_lock_balance+0x5a/0x90
> [89945.640568]  [<ffffffff8104c546>] push_rt_task+0xc6/0x290

if you look at that code you'll find the double_lock_balance() in
question is the one in find_lock_lowest_rq() [yay for inlining].

Now find_lock_lowest_rq() has a bug.. it fails to use
double_unlock_balance() in one exit path, if this results in a retry in
push_rt_task() we'll call double_lock_balance() again, at which point
we'll run into said lockdep confusion.

Reported-by: Roland Dreier <roland@kernel.org>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1337282386.4281.77.camel@twins
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/sched/rt.c