sched: Cure more NO_HZ load average woes
authorPeter Zijlstra <a.p.zijlstra@chello.nl>
Tue, 30 Nov 2010 18:48:45 +0000 (19:48 +0100)
committerIngo Molnar <mingo@elte.hu>
Wed, 8 Dec 2010 19:15:04 +0000 (20:15 +0100)
commit0f004f5a696a9434b7214d0d3cbd0525ee77d428
tree274b3bb92469789284d864314d46e902c70e8384
parent6313e3c21743cc88bb5bd8aa72948ee1e83937b6
sched: Cure more NO_HZ load average woes

There's a long-running regression that proved difficult to fix and
which is hitting certain people and is rather annoying in its effects.

Damien reported that after 74f5187ac8 (sched: Cure load average vs
NO_HZ woes) his load average is unnaturally high, he also noted that
even with that patch reverted the load avgerage numbers are not
correct.

The problem is that the previous patch only solved half the NO_HZ
problem, it addressed the part of going into NO_HZ mode, not of
comming out of NO_HZ mode. This patch implements that missing half.

When comming out of NO_HZ mode there are two important things to take
care of:

 - Folding the pending idle delta into the global active count.
 - Correctly aging the averages for the idle-duration.

So with this patch the NO_HZ interaction should be complete and
behaviour between CONFIG_NO_HZ=[yn] should be equivalent.

Furthermore, this patch slightly changes the load average computation
by adding a rounding term to the fixed point multiplication.

Reported-by: Damien Wyart <damien.wyart@free.fr>
Reported-by: Tim McGrath <tmhikaru@gmail.com>
Tested-by: Damien Wyart <damien.wyart@free.fr>
Tested-by: Orion Poplawski <orion@cora.nwra.com>
Tested-by: Kyle McMartin <kyle@mcmartin.ca>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: stable@kernel.org
Cc: Chase Douglas <chase.douglas@canonical.com>
LKML-Reference: <1291129145.32004.874.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
include/linux/sched.h
kernel/sched.c
kernel/timer.c