Created attachment 661502 [details]
udev rules patch

Description of problem:
When we have a raid volume and one of disks fails, the failed disk is still present in volume and container. The raid volume is in normal state (should be  degraded) and rebuild cannot start. 

How reproducible:
Always

Steps to Reproduce:
mdadm -Ss
mdadm --zero-superblock /dev/sd[b-d]
mdadm -C /dev/md/imsm0 -amd -e imsm -n 3 /dev/sdb /dev/sdc /dev/sdd -R
mdadm -C /dev/md/raid5 -amd -l5 -n 3 /dev/sdb /dev/sdc /dev/sdd -R
mdadm --wait /dev/md/raid5
# power off a raid member disk (e.g. /dev/sdd)

Actual results:
The failed disk is still present in container/volume. State in 'mdadm -D /dev/md/raid5' output is 'clean'.

Expected results:
The failed disk should disappear from container and volume. State in 'mdadm -D /dev/md/raid5' output should be 'clean, degraded'.

Additional info:
When one of disks fails udev runs "/usr/lib/udev/rules.d/65-md-incremental.rules". The "65-md-incremental.rules" script has following rules:
SUBSYSTEM=="block", ACTION=="remove", ENV{ID_FS_TYPE}=="linux_raid_member", \
        RUN+="/sbin/mdadm -If $name --path $env{ID_PATH}"
SUBSYSTEM=="block", ACTION=="remove", ENV{ID_FS_TYPE}=="isw_raid_member", \
        RUN+="/sbin/mdadm -If $name --path $env{ID_PATH}"

If a disk fails and "$env{ID_PATH}" is null, udev runs "/sbin/mdadm -If sdd --path" (what does nothing - it is an incorrect mdadm's command), instead of "/sbin/mdadm -If sdd".

The following patch fixes this bug:
correct-65-md-incremental-rules-in-case-a-raid-disk-fails.patch