Re: udev and md/dm

[Kay Sievers <kay.sievers@vrfy.org>]

On Sat, 2007-06-16 at 00:31 -0700, Joel Becker wrote:
> On Sat, Jun 16, 2007 at 04:32:49AM +0200, Kay Sievers wrote:
> > On 6/16/07, Joel Becker <Joel.Becker@oracle.com> wrote:
> > >        Why does udev ignore md* and dm-* devices for persistent naming?
> > >I was utterly surprised to find that an ext3 on /dev/md0 or /dev/dm-0
> > >won't show up in /dev/disk/by-uuid.  Given that dm and md devices can
> > >change names, isn't this something we'd really like to have?
> > 
> > Sure, they just have their own rules files. Both have specific
> > requirements/lifetime-rules and plug into "change" events.
> 
> 	I'm looking at rhel5, debian, and raw udev-105, and I can't find
> anywhere in the rules that use vol_id to map md or dm devices to
> /dev/disk/by-label and /dev/disk/by-uuid.  The vol_id using files
> (persistent.rules) skips those devices.  I can see why perhaps you'd say
> "loop and ramdisk may be garbage", but dm*, md*, and similar are
> expected persistent environment things.  And their names can certainly
> change.

dm: (requires a simple patch to dmsetup)
  http://git.kernel.org/?p=linux/hotplug/udev.git;a=blob;f=etc/udev/suse/64-device-mapper.rules;hb=HEAD


md: (requires the latest mdadm)
  http://git.kernel.org/?p=linux/hotplug/udev.git;a=blob;f=etc/udev/suse/64-md-raid.rules;hb=HEAD

> 	I ask because I am looking at implementing a robust "how to
> determine what device is meant by UUID xxxxx".  Robust in the face of
> catching the raid1 and multipath and so on.  So /dev/disk/by-uuid/xxxxx
> wouldn't point to /dev/sdd1 if it really should be /dev/md0.  I had
> hoped udev + vol_id would do this for me, as vol_id seems to have
> explicit ordering to make this work right.

Right, vol_id recognizes md-metadata and doesn't return uuid/label for
raid members.

Udev has priorities for the symlink-stack (multiple devices can claim
the same name, but only one can actually own the symlink), and the
multipath rules for dm-devices specify a higher priority, so that
symlinks requested from the individual paths are overwritten.

> But the udev rules skip
> checking uuids and labels on the aggregate devices.
> 	Please, show me where I'm wrong :-)

All of that is not in RHEL or Debian at the moment. :)

Kay


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel