On Mon, Sep 03, 2001 at 03:44:04PM +0400, Borsenkow Andrej wrote:
>
> This comes because of mknod in vgscan.
Correct.
> From devfs point of view, both
> mknod and devfs_register means the same.
Meaning on a system with devfs mounted, the kernel reroutes (or does
similar) mknod(2) to devfs_register when it's done on the devfs
filesystem?
> And you are not expected to use
> both at the same time (for the same device in any case).
Well, that would make sense.
> Here activating
> VG tries to register devfs names that were already created by vgscan.
> Strictly speaking, it is harmless.
I don't so much care about the errors as much as I care that I don't
have /dev/<volume group>/* entries. If I had the latter, and not the
former I would write it off to old lvm code and hope that it goes away
when mdk updates.
> Yes, but still devfs exists just once. Mounting it just makes it
> user-visible - but as long as devfs is conpiled in kernel, *all* device
> operations use devfs. It does not depend on its mounted state. So
> remounting devfs should not (module bug) make any difference.
But does any nodes created after devfs is mounted disappear when it is
(unmounted and) remounted?
> > The system boots fine onto the LVM based root filesystem. The problem
> > is that the device nodes in /dev/VG (group, lvol1 and 1) are missing
> > an uncreatable.
> >
>
> To make sure - is it root device?
What do you mean? What test do you want me to run. Your question
might make more sense to me when I see what test you want to see the
results of.
> Then it could explain something.
> Is it possible to download your installer with LVM root support? I would
> give it a try.
Theoretically, I can package something for you. I would give you the
files to replace in your cooker to do lvm-root install. It is not
complete however as I am still trying to figure out the gyrations
needed for the initrd.
You need to replace some installer files, the lilo rpm and eventually,
the mkinitrd rpm.
b.
--
Brian J. Murrell