sched: Fix update_cfs_load() synchronization
authorPaul Turner <pjt@google.com>
Mon, 15 Nov 2010 23:47:04 +0000 (15:47 -0800)
committerIngo Molnar <mingo@elte.hu>
Thu, 18 Nov 2010 12:27:48 +0000 (13:27 +0100)
commite33078baa4d30ad1d0e46d1f62b9e5a63a3e6ee3
tree13d1a075c9cd41ce666cf21b0b209bd3d121f566
parentf0d7442a5924a802b66eef79b3708f77297bfb35
sched: Fix update_cfs_load() synchronization

Using cfs_rq->nr_running is not sufficient to synchronize update_cfs_load with
the put path since nr_running accounting occurs at deactivation.

It's also not safe to make the removal decision based on load_avg as this fails
with both high periods and low shares.  Resolve this by clipping history after
4 periods without activity.

Note: the above will always occur from update_shares() since in the
last-task-sleep-case that task will still be cfs_rq->curr when update_cfs_load
is called.

Signed-off-by: Paul Turner <pjt@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20101115234937.933428187@google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
kernel/sched.c
kernel/sched_fair.c