Re: kexec?

[Coywolf Qi Hunt <coywolf@gmail.com>]

On 5/11/05, Borislav Petkov <petkov@uni-muenster.de> wrote:
> On Wednesday 11 May 2005 08:04, Maneesh Soni wrote:
> <snip>
> > > > [root@zmei]: kexec -p vmlinux --args-linux --append="root=/dev/hda1
> > > > maxcpus=1 init 1"
> > >
> > >  kexec-tools-1.101 loads for me, but if cmdline is used, it hangs up
> > > after "Starting new kernel"
> >
> > Thanks for trying this out. As Vivek mentioned can you please try with
> > bulding second or dump capture kernel with CONFIG_SMP=N and _without_
> > maxcpus= option. Basically the second kernel's job is just to save the dump
> > and it doesnot need to be a SMP kernel. There are some issues with booting
> > SMP kernel as dump capture kernel.
> 
> Hm, without 'maxcpus' seems to work. However, when booting into the new
> kernel, the rootfs had to be fsck'ed due to "/ was not cleanly unmounted,
> check forced." and then was forced to reboot linux due to inconsistency in
> the fs. I simply did kexec -l <vmlinux> --args-linux --append="root=/dev/hda1
> init 1" and then kexec -e to execute the loaded image. It seems that the
> filesystems are not unmounted properly before loading the second kernel, (or
> I am missing something..., which is more likely :))

kexec is like a bare reboot. 
Add kexec -l and -e just above the reboot line in your
/etc/init.d/reboot script,
or umount manually( sysrq+s sysrq+u ).

-- 
Coywolf Qi Hunt
http://sosdg.org/~coywolf/