[pkg-cryptsetup-devel] Bug#554506: Bug#554506: Bug#554506: cryptsetup: unable to find root volume

[Jonas Meurer]

hey,

On 05/11/2009 Max Vozeler wrote:
> On Thu, Nov 05, 2009 at 12:44:36AM +0100, Philip Frei wrote:
> > volume group "debian" not found.
> > skipping volume group debian.
> > unable to find LVM volume debian/root
> > volume group "debian" not found.
> > cryptsetup: lvm device name (dev/disk/by-uuid/$FOO) does not begin
> > with /dev/mapper
> 
> We recently changed d-i to use UUID in crypttab entries,
> which is likely to be causing this problem.
> 
> Jonas, do you think the initramfs script can be adapted to
> allow UUID or should partman-crypto go back to /dev/mapper
> for the root backing device for now?
> 
> (No time to look deeper into it right now, sorry.)

i take a deeper look now, and figued out what the real problem is here:

in case that the setup is LUKS on LVM, the cryptroot initramfs script
needs to determine the volume group to start in order to make the
underlying logical volume available before unlocking it with cryptsetup.

this is done by parsing the cryptsetup source device. for example the
source device "/dev/mapper/vg00-root" leads to volume group "vg00" being
activated.

when the UUID is given instead, no information about the volume group
can be extracted.

i don't know what the propper fix is yet. i might change the cryptroot
hook script to always transfer UUID to path for /conf/conf.d/cryptroot
in initramfs. but then i don't see a reason to use UUIDs in initramfs at
all. any other solution that i see now (e.g. to detect LUKS over LVM in
cryptroot hook and save the volume group name in /conf/conf.d/cryptroot)
would require intrusive changes to the cryptroot implementation.

greetings,
 jonas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 490 bytes
Desc: Digital signature
URL: <http://lists.alioth.debian.org/pipermail/pkg-cryptsetup-devel/attachments/20100225/f6195c6b/attachment.pgp>