Re: [patch] latency tracer, 2.6.15-rc7

[Lee Revell <rlrevell@joe-job.com>]

On Mon, 2006-01-02 at 21:14 +0100, Ingo Molnar wrote:
> * Linus Torvalds <torvalds@osdl.org> wrote:
> 
> > > have you applied the zlib patches too? In particular this one should 
> > > make a difference:
> > > 
> > >     http://redhat.com/~mingo/latency-tracing-patches/patches/reduce-zlib-stack-hack.patch
> > > 
> > > If you didnt have this applied, could you apply it and retry with 
> > > stack-footprint-debugging again?
> > 
> > Ingo, instead of having a static work area and using locking, why not 
> > just move those fields into the "z_stream" structure, and thus make 
> > them be per-stream? The z_stream structure should already be allocated 
> > with kmalloc() or similar by the caller, so that also gets it off the 
> > stack without the locking thing.
> > 
> > Hmm?
> 
> yeah, i'll implement that. The above one was just a quick hack that 
> somehow survived. (and i am definitely a wimp when it comes to changing 
> zlib internals :-)

Ingo,

If you get a chance would you mind porting this to 2.6.16-rc1?  I only
get 6 failures when applying the 2.6.15-rc7 patch set.

I'd like to be able to catch any 2.6.16 latency regressions while there
is time to do something about them, to avoid a repeat of the unmap_vmas
problem in 2.6.15.

Lee