Sched: fix skip_clock_update optimization
authorMike Galbraith <efault@gmx.de>
Wed, 8 Dec 2010 10:05:42 +0000 (11:05 +0100)
committerIngo Molnar <mingo@elte.hu>
Wed, 8 Dec 2010 19:15:06 +0000 (20:15 +0100)
commitf26f9aff6aaf67e9a430d16c266f91b13a5bff64
tree84e39b539b42655dbae6b80acbe14e10772711fe
parent0f004f5a696a9434b7214d0d3cbd0525ee77d428
Sched: fix skip_clock_update optimization

idle_balance() drops/retakes rq->lock, leaving the previous task
vulnerable to set_tsk_need_resched().  Clear it after we return
from balancing instead, and in setup_thread_stack() as well, so
no successfully descheduled or never scheduled task has it set.

Need resched confused the skip_clock_update logic, which assumes
that the next call to update_rq_clock() will come nearly immediately
after being set.  Make the optimization robust against the waking
a sleeper before it sucessfully deschedules case by checking that
the current task has not been dequeued before setting the flag,
since it is that useless clock update we're trying to save, and
clear unconditionally in schedule() proper instead of conditionally
in put_prev_task().

Signed-off-by: Mike Galbraith <efault@gmx.de>
Reported-by: Bjoern B. Brandenburg <bbb.lst@gmail.com>
Tested-by: Yong Zhang <yong.zhang0@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: stable@kernel.org
LKML-Reference: <1291802742.1417.9.camel@marge.simson.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
kernel/fork.c
kernel/sched.c