Re: Fixing PR port-i386/45946 (Kernel locks up in VMEM system)

[Matthias Scheler <tron%zhadum.org.uk@localhost>]

On Fri, Feb 10, 2012 at 02:47:04PM +0100, Lars Heidieker wrote:
> > Index: sys/arch/i386/include/param.h
> > ===================================================================
> > RCS file: /cvsroot/src/sys/arch/i386/include/param.h,v
> > retrieving revision 1.75
> > diff -u -r1.75 param.h
> > --- sys/arch/i386/include/param.h       4 Feb 2012 17:56:16 -0000       1.75
> > +++ sys/arch/i386/include/param.h       10 Feb 2012 13:27:42 -0000
> > @@ -129,7 +129,7 @@
> >  * logical pages.
> >  */
> >  #define        NKMEMPAGES_MIN_DEFAULT  ((8 * 1024 * 1024) >> PAGE_SHIFT)
> > -#define        NKMEMPAGES_MAX_DEFAULT  ((280 * 1024 * 1024) >> PAGE_SHIFT)
> > +#define        NKMEMPAGES_MAX_DEFAULT  ((360 * 1024 * 1024) >> PAGE_SHIFT)
> >
> >  /*
> >  * Mach derived conversion macros
> >
> > And this change fixes all the problems I have observed.
> >
> > Is there any reason not to commit that? I'm particular wondering what
> > influence the patch will have on machines with low amounts of memory.
> >
> there is no influence on machines with low memory.

Excellent.

> It should go in.

I completely agree for selfish reasons. ;-)

> I'll make a patch this evening including such adjustments for macppc
> etc. and reducing the quantum caching in the kmem_arena a bit.

Thanks. I've assigned the PR to you.

        Kind regards

-- 
Matthias Scheler                                  http://zhadum.org.uk/