sched: adjust scheduler cpu power for stolen time
authorGlauber Costa <glommer@redhat.com>
Mon, 11 Jul 2011 19:28:18 +0000 (15:28 -0400)
committerAvi Kivity <avi@redhat.com>
Thu, 14 Jul 2011 09:59:47 +0000 (12:59 +0300)
commit095c0aa83e52d6c3dd7168610746703921f570af
tree7461604a6040f1724bdf66c6f87e3abe60e2ad50
parente6e6685accfa81f509fadfc9624bc7c3862d75c4
sched: adjust scheduler cpu power for stolen time

This patch makes update_rq_clock() aware of steal time.
The mechanism of operation is not different from irq_time,
and follows the same principles. This lives in a CONFIG
option itself, and can be compiled out independently of
the rest of steal time reporting. The effect of disabling it
is that the scheduler will still report steal time (that cannot be
disabled), but won't use this information for cpu power adjustments.

Everytime update_rq_clock_task() is invoked, we query information
about how much time was stolen since last call, and feed it into
sched_rt_avg_update().

Although steal time reporting in account_process_tick() keeps
track of the last time we read the steal clock, in prev_steal_time,
this patch do it independently using another field,
prev_steal_time_rq. This is because otherwise, information about time
accounted in update_process_tick() would never reach us in update_rq_clock().

Signed-off-by: Glauber Costa <glommer@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Tested-by: Eric B Munson <emunson@mgebm.net>
CC: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
CC: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
arch/x86/Kconfig
kernel/sched.c
kernel/sched_features.h
This page took 0.024591 seconds and 5 git commands to generate.