Subject: Re: [vserver] opteron server dies with vserver patch.
From: Pawel Sikora <pluto@pld-linux.org>
Date: Mon, 08 Aug 2011 17:08:53 +0200

On Monday 08 of August 2011 16:07:54 Herbert Poetzl wrote:
> On Mon, Aug 08, 2011 at 01:24:48PM +0200, Pawel Sikora wrote:
> > Hi,
> 
> > i have few dual opteron-6128 machines which are part of eda
> > software auto-test farm. these machines don't host any vserver
> > guests but installed kernel has vserver-patch applied by
> > vendor. on vanilla kernel 2.6.38 (and newer) with vserver
> > patch applied machines lock in some way after few hours of
> > quite heavy computing + i/o (avg. load ~16). on 2.6.37.y with
> > vserver everything works fine and on pure vanilla kernels also
> > everything works fine.
> 
> > on the ipmi console i've logged only kernel flood:
> > https://lkml.org/lkml/2011/5/23/398 this crash is reproducible
> > on 2.6.38.y/2.6.39.y/3.0.y with the latest vserver patches.
> 
> the lkml posting says vserver/grsec, now does this mean
>  o you are testing with mainline + Linux-VServer patch
>  o you are testing with mainline + grsec patch
>  o you are testing with a combined patch?

i've tested mainline / mainline+grsec / mainline+vs to isolate a bad patchset.
now i'm sure that vserver patchset causes server locking on 2.6.38 and newer.

> > # cat /proc/version
> > Linux version 3.0.0-vs2.3.1-pre8-dirty (pawels@hal) (gcc version 4.5.3 20110515
(release) (PLD-Linux) ) #2 SMP Fri Aug 5 09:48:25 CEST 2011
> 
> what does -dirty mean here?

the kernel makefiles reports local uncommited changes (vserver pathset) as '-dirty'.