Re: CFS: some bad numbers with Java/database threading [FIXED]

[Ingo Molnar <mingo@elte.hu>]

* Ingo Molnar <mingo@elte.hu> wrote:

> Peter, what do you think?
> 
> Linus, if this would be acceptable for .23 then i'll push it out into 
> sched.git together with another fix that Hiroshi Shimamoto just posted 
> to lkml.

it's getting late here so i've pushed the current version of those two 
patches out to:

   git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-sched.git

I'll redo this git tree if we want some other solution for yield. (but i 
think this is the safest approach for 2.6.23 - some apps will complain 
about too strong yield, some apps will complain about too weak yield. So 
by providing the two extremes we at least cover the practical range of 
behavior.)

there's nothing else pending for 2.6.23 otherwise at the moment, 
scheduler-wise.

	Ingo

------------------>
Hiroshi Shimamoto (1):
      sched: fix invalid sched_class use

Ingo Molnar (1):
      sched: add /proc/sys/kernel/sched_compat_yield

 include/linux/sched.h |    1 
 kernel/sched.c        |   10 ++++---
 kernel/sched_fair.c   |   63 +++++++++++++++++++++++++++++++++++++++++++++-----
 kernel/sysctl.c       |    8 ++++++
 4 files changed, 72 insertions(+), 10 deletions(-)