Revert "sched: Update_cfs_shares at period edge"
authorLinus Torvalds <torvalds@linux-foundation.org>
Fri, 14 Dec 2012 15:20:43 +0000 (07:20 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 14 Dec 2012 15:20:43 +0000 (07:20 -0800)
commit17bc14b767cf0692420c43dbe5310ae98a5a7836
tree9b0f339e5d9769a51acdeb7a246f6e7522efa966
parent7313264b899bbf3988841296265a6e0e8a7b6521
Revert "sched: Update_cfs_shares at period edge"

This reverts commit f269ae0469fc882332bdfb5db15d3c1315fe2a10.

It turns out it causes a very noticeable interactivity regression with
CONFIG_SCHED_AUTOGROUP (test-case: "make -j32" of the kernel in a
terminal window, while scrolling in a browser - the autogrouping means
that the two end up in separate cgroups, and the browser should be
smooth as silk despite the high load).

Says Paul Turner:
 "It seems that the update-throttling on the wake-side is reducing the
  interactive tasks' ability to preempt.  While I suspect the right
  longer term answer here is force these updates only in the
  cross-cgroup case; this is less trivial.  For this release I believe
  the right answer is either going to be a revert or restore the updates
  on the enqueue-side."

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Bisected-by: Mike Galbraith <efault@gmx.de>
Acked-by: Paul Turner <pjt@google.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
kernel/sched/fair.c