Red Hat Bugzilla – Bug 234471
multipath devices (dm-?) will not be created in /dev/
Last modified: 2007-11-30 17:12:00 EST
From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1; .NET CLR 2.0.50727)
Description of problem:
Hardware: SUN X4100 2AMD 285 16GB ISCSI over Silverback iSNAP HBA 2path
when I boot, the multipath is working fine and the devices are created well under /dev/mapper/mpath[0-n] and I can see underlying devices in cat /proc/partitions
I do see as well the within that list (proc/partitions) that there are multipath-devices dm-0, dm-1, dm2 .... but this devices are not created under /dev/
So, I can create the nodes with mknod dm-0 b 253 0 and so on, and mount the multipathed device then, but this is not boot safe. After a boot this manually created devicenodes are gone and I have to do it by myself again. Really strange is, that this "dm-n" devices are there and named as you can see in multipath -ll or in cat /proc/partitions but the nodes are never created.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
2. cat /proc/partitions | grep dm-
3. ls -la /dev/dm*
after 2.) I see dm-0, dm-1 and dm-2 created
after 3.) I get a no such file or directory
after 2.) OK
after 3.) I would expect a devicenode
In fc6, the udev rules for device-mapper devices don't create /dev/dm-* devices.
This is actually a good thing. There is no guarantee that these device names will
always point to the same device. For instance if on one boot /dev/dm-1 points to
your LVM root volume, on another boot, it could be that /dev/dm-2 points to it.
So referencing devices by these names is not safe. Further, the /dev/dm-* names
are not created until udev runs, so if you need to access a device early in the
boot process, you must use the /dev/mapper name. Really, there is no good reason
to ever use the /dev/dm-* names. You should always use the /dev/mapper ones. It
is unfortunate that /proc/partitions references them by the /dev/dm-* name.