/sbin/kexec became zero size... Anyone else?

[Neil Horman]

On Tue, Sep 23, 2008 at 01:32:19PM -0700, Jay Lan wrote:
> Neil Horman wrote:
> > On Tue, Sep 23, 2008 at 09:41:50AM -0700, Jay Lan wrote:
> >> Simon Horman wrote:
> >>> On Wed, Sep 17, 2008 at 02:07:15PM -0700, Jay Lan wrote:
> >>>> Hi,
> >>>>
> >>>> My root disk was populated with sles10sp2, but the kernel was
> >>>> 2.6.27-rc5 and /sbin/kexec was built from 2.0.0 version.
> >>>>
> >>>> Many times when kdump kernel failed early i found  after reboot that
> >>>> /sbin/kexec became zero size. There was no warning on executing
> >>>> '/etc/init.d/kdump start' when /sbin/kexec is zero size. I have
> >>>> no idea when and how that happened.
> >>>>
> >>>> Has anyone else seen this problem?
> >>> Is it possible that this is caused by filesystem corruption?
> >> I found the problems, which i can not reliably reproduce, happened
> >> when i ran into kdump kernel boot failure (yet not every time.)
> >> What puzzled me was why only /sbin/kexec was affected.
> >>
> > This is why most kdump implementations capture vmcore from an initramfs, since
> > a crash means that you may not be able to trust the integrity of the filesystems
> > that were mounted before the panic.
> > 
> > 
> >> I can not explain what caused it (otherwise i would be able to
> >> reproduce it) and do not know where my finger should point to. ;)
> >> The good news is since i fixed the kernel boot problem it has not
> >> happened any more.
> >>
> >> The distros probably can add a check against zero-sized kexec
> >> in /etc/init.d/kdump script though.
> >>
> > Theres really not a need to.  Filesystems can be corrupted by crashes, thats why
> > we capture cores from the initrd.  We need to be able to trust the rest of the
> > system to scan filesystems and do their best to recover when filsystems loose
> > data.
> 
> Hi Neil,
> 
> Which rhel release starts doing this?
> 
Not sure what you mean.  Which rhel releases attempt to recover filesystems on
boot?  Linux has done that with fsck for years.  Which RHEL releases have
allowed vmcore capture from an initrd with kdump?  Read the man page, its been
available since GA.

Neil

> Thanks,
>  - jay
> 
> 
> > 
> > Regards
> > Neil
> >  
> >> Thanks,
> >>  - jay
> >>
> >>
> >> _______________________________________________
> >> kexec mailing list
> >> kexec at lists.infradead.org
> >> http://lists.infradead.org/mailman/listinfo/kexec
> > 
> 

-- 
/***************************************************
 *Neil Horman
 *Senior Software Engineer
 *Red Hat, Inc.
 *nhorman at redhat.com
 *gpg keyid: 1024D / 0x92A74FA1
 *http://pgp.mit.edu
 ***************************************************/