Re: memory.c - bad pmd - x86_64

[Ivan Adzhubey <iadzhubey rics bwh harvard edu>]

On Tue, 31 May 2005 22:30:43 +0200
Christoph Franke <news thefranke net> wrote:

> Dave Jones - Tue, May 24 2005 12:57:15 -0400:
> 
> > On Tue, May 24, 2005 at 07:02:40AM -0700, Peter J. Stieber wrote:
> >  > DJ = Dave Jones
> >  > DJ> Hrmph. No, I screwed up the patch.
> >  > DJ> -29 is rebuilding, it'll appear in the FC3/ subdir of my
> >  > DJ> people page a little while after its done building.
> >  > 
> >  > here's what I get with 2.6.11-1.29_FC3smp.
> >  > 
> >  > May 24 06:55:43 maggie kernel: collect2:5519 free pmd
> >  > ffff8100777f1000 freed by 0xffffffffffffffff
> >  > May 24 06:55:43 maggie kernel: mm/memory.c:109: bad pmd 
> >  > ffff8100777f1000(0000000000000064).
> 
> [snip]
> 
> I noticed a -30 build on your webspace at people.redhat.com. Is this
> based on 2.6.11.11, and does it nail the bad pmd bug?

Christoph,

I am not Dave, sorry ;). Testing Dave's .30 build right now on my
compute nodes and yet to see any errors, even with memory-intensive
multi-threaded jobs. But I can't test with my fileserver at the moment
since jobs already running that need NFS access. The fileserver box is
the one that gave me most headaches before so I will try rebooting it
under .30 build in the next 24 hours and post results ASAP.

Cheers,
Ivan