Even if lvm worked well before with, afaik, the same behaviour of dmraid, after digging a bit, I tend to blame dmraid probably a bit more than lvm.
There are some issues with /sbin/dmraid-activate, which tries to parse /proc/cmdline and conclude if the device is root, and if so, using the -Z switch. This fails with "root=/dev/mapper/some-vg-root" and comparing this to "/dev/sda", of course. There are also other issues, like when /dev/sda is param, this goes into $NodeName, and then operating on /dev/$NodeName (which then is /dev/dev/$NodeName of course) and failing. Would it be advisable to file another bug for dmraid and link these two somehow? jo. PS: For people needing a quick'n'dirty workaround: put the -Z in the line of /sbin/dmraid-activate where there's a dmraid -i -ay without it, and do an update-initramfs -k all -u. (I don't believe the underlying partitions will be missed a lot. Correct me if I'm wrong.)
