On 07/14/14 at 08:27am, Vivek Goyal wrote:
On Mon, Jul 14, 2014 at 03:57:12PM +0800, WANG Chao wrote:
> On 07/10/14 at 11:28am, Vivek Goyal wrote:
> > On Tue, Jun 17, 2014 at 03:07:58PM +0800, WANG Chao wrote:
> > > Currently in kdump initramfs, mount units specified in /etc/fstab are
> > > required by "local-fs.target". When any of these mounts fails,
> > > local-fs.target fails.
> > >
> > > For kdump initramfs, we need to isolate to emergency service on any of
> > > the mount failure,
> >
> > Hi Chao,
> >
> > I can't remember that why do we need to isolate to emergency service on
> > any of the mount failures?
>
> If we don't isolate, system will continue to boot and causing emergency
> service to be shut down at some point later.
So let us say that somebody is dumping to nfs. Now /etc/fstab will have
two entries. One for nfs and other for root. If root, mount fails, kdump
will still succeed and dump to nfs and system will reboot. So we should
not have this strict requirement that if any of the mount in /etc/fstab
fails we isolate to emergency service.
We will only have a single entry, the nfs mount. We never put root mount
into /etc/fstab in any case. Because sysroot.mount unit is generated
automatically according to the "root=" arg of /proc/cmdline.
The bottom line is only the mount of the dump target will be put in
/etc/fstab. And we only have one dump target for now. Even if we have
multiple dump targets down the line, we will append "x-initrd.mount" to
each of them.
Though if root fails, systemd might not make any progress.
Anyway, just put in changelogs that why isolating to emergency service is
improtant.
Yeah of course.
Thanks
WANG Chao