Bug#335230: More Code also Random, but less :-)

[Marco Amadori <marco.amadori@gmail.com>]

Alle 11:15, mercoledì 2 novembre 2005, hai scritto:
> On Wed, Nov 02, 2005 at 11:11:24AM +0100, Marco Amadori wrote:
> > But you missed to clear us if mdadm and vgchange are needed to evms to
> > shows up in case of compatibility volumes, but I think so anyway so I
> > produced this random code:
>
> I actually do not think so. I'm unsure for LVM2, but for the MD volumes,
> the device node /dev/evms/ has exactly the same major and minor as the one
> in /dev.

Exactly, for md seems to do not need it, but evms lvm2 needs vgchange to 
proper activate all volumes, and the devices numbers differs:

# ls -l /dev/evms/lvm2/sicuro/usr64
brw-r-----  1 root disk 253, 41 2005-11-02 11:05 /dev/evms/lvm2/sicuro/usr64
# ls -l /dev/mapper/*usr64*
brw-------  1 root root 253, 41 2005-11-02 11:05 /dev/mapper/lvm2|sicuro|usr64
brw-------  1 root root 253, 28 2005-11-02 11:05 /dev/mapper/sicuro-usr64

different 

and
ls -l /dev/md3 /dev/evms/md/md3
brw-r-----  1 root disk 9, 3 2005-11-02 11:05 /dev/evms/md/md3
brw-rw----  1 root disk 9, 3 2005-11-02 11:05 /dev/md3

same

>
> Note that /dev/evms/lvm2/ can AFAIK contain EVMS volumes, it's just the LVM
> that isn't managed directly by EVMS.

So we need vgchange on initramfs. 
This cut the head at the bull (not translatable :-) ).

Testing this solution right now.


-- 
ESC:wq