Hi, I see you commented in #401393 linking to #402511.
On Thu, Jan 18, 2007, Frans Pop wrote: > Further analysis in http://bugs.debian.org/406697 has shown that the > reason the /dev/dm-* devices are not removed, is the fact that udevd has > already been killed by the time the crypto and lvm scripts are run on > shutdown and thus the events created by the kernel are no longer > processed as they would be if the system was running normally. So what's happening in #402511 is that some process is working on all devices such as /dev/mapper/* and /dev/dm-*, when it processes /dev/mapper/*, it causes some devices to disappear in the kernel, but /dev/mapper/* nodes and /dev/dm-* are not removed by udev. This is IMO different from #401393 where the problem seems to be that lilo decides to call dm operations on device names such as "dm-0" but this is never supported; instead, either dm-0 should be converted to device-mapper names or operations should be done via major/minor (which is the patch I just proposed). Bye, -- Loïc Minier <[EMAIL PROTECTED]>
