Hello,
How is the boot process of RAID (using a Debian supplied kernel that
doesn't have RAID autodetect compiled in) meant to work with udev?
What seems to happen (at least with recent testing/sarge based
system):
1. initrd script starts RAID for swap and /. No problems here. The
initrd script does the right thing (after I renamed devfs names to
non-devfs names in /etc/raidtab that is).
2. kernel boots and transfers control to userland.
3. udev, I believe (not checked) starts early in the boot process. It
creates /dev/md entries for the activated RAID partitions (swap and
/), but nothing else (since the RAID hasn't been initialised for
these devices yet).
4. /etc/init.d/raid2 attempts to initialise the other RAID partitions
but fails to do so because the /dev/md* entries do not exist.
As I hacked solution, I have added to the very start of my
/etc/init.d/raid2 (not tested yet in automatic boot, but worked when I
run it manually):
for i in 12 14 20 30 32
do
mknod /dev/md/$i b 9 $i
ln -s md/$i /dev/md$i
done
I think this will solve the problem.
Now I know what happens in practise, what is meant to happen in
theory?
I haven't observed anything like this behaviour with devfs, so I
suspect this is udev specific.
--
Brian May <bam@debian.org>