]> git.kernelconcepts.de Git - karo-tx-linux.git/commit
RAID5: revert e9e4c377e2f563 to fix a livelock
authorShaohua Li <shli@fb.com>
Fri, 26 Feb 2016 00:24:42 +0000 (16:24 -0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 12 Apr 2016 16:08:57 +0000 (09:08 -0700)
commitad072f606586a70eab8fd168465a6df4c695a9e9
treef09fdb7b9968ecaa1e04c23f89cb5868ce9e739d
parent8568767fe1d63fbbdeacf2f69da6f84ad277d4f7
RAID5: revert e9e4c377e2f563 to fix a livelock

commit 6ab2a4b806ae21b6c3e47c5ff1285ec06d505325 upstream.

Revert commit
e9e4c377e2f563(md/raid5: per hash value and exclusive wait_for_stripe)

The problem is raid5_get_active_stripe waits on
conf->wait_for_stripe[hash]. Assume hash is 0. My test release stripes
in this order:
- release all stripes with hash 0
- raid5_get_active_stripe still sleeps since active_stripes >
  max_nr_stripes * 3 / 4
- release all stripes with hash other than 0. active_stripes becomes 0
- raid5_get_active_stripe still sleeps, since nobody wakes up
  wait_for_stripe[0]
The system live locks. The problem is active_stripes isn't a per-hash
count. Revert the patch makes the live lock go away.

Cc: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Cc: NeilBrown <neilb@suse.de>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/md/raid5.c
drivers/md/raid5.h