Re: [PATCH] avoid race condition in pick_next_task_fair in kernel/sched_fair.c

[Peter Zijlstra <peterz@infradead.org>]

On Wed, 2010-12-22 at 18:16 +0100, Ingo Molnar wrote:
> * Peter Zijlstra <peterz@infradead.org> wrote:
> 
> > On Wed, 2010-12-22 at 16:14 +0100, Miklos Vajna wrote:
> > > kernel-build is a git build using the config I already sent and after a
> > > 'git checkout v2.6.36'. I can try to build master as well, so far what I
> > > saw is that the bug occurs there as well, but less frequently, so maybe
> > > that's a bit harder to debug. 
> > 
> > Right, so I've got v2.6.36 exploding, current -git and -tip won't
> > explode for me, tried both about 20 times. So I'll try and figure out
> > wth makes .36 explode and then see if further kernel still suffer that
> > problem.
> 
> Does it explode if you change some of the CONFIG_CC flags, such as 
> CONFIG_CC_OPTIMIZE_FOR_SIZE? I.e. is the crash dependent on the precise kernel image 
> layout perhaps? If it's sensitive to that then this might keep you from chasing 
> shadows ...

I've since changed the .config slightly (enabled frame pointers, added
debug info, enabled the function tracer) and its still exploding. Also,
I suspect Miklos and me have different gcc versions.

So I suspect its really something weird, I just can't really make sense
of it yet.