diff options
author | NeilBrown <neilb@suse.de> | 2009-02-25 13:18:47 +1100 |
---|---|---|
committer | Greg Kroah-Hartman <gregkh@suse.de> | 2009-03-16 17:52:54 -0700 |
commit | a9785b107920670a8ec8ef1744e0e0c59dde9eb4 (patch) | |
tree | 37cb0a3b2a75543918226ec892fa2d0801ffe7b8 | |
parent | 285ed5b1cedb3f043fa90f31deaa11f7c3ff4c76 (diff) | |
download | lwn-a9785b107920670a8ec8ef1744e0e0c59dde9eb4.tar.gz lwn-a9785b107920670a8ec8ef1744e0e0c59dde9eb4.zip |
md: avoid races when stopping resync.
commit 73d5c38a9536142e062c35997b044e89166e063b upstream.
There has been a race in raid10 and raid1 for a long time
which has only recently started showing up due to a scheduler changed.
When a sync_read request finishes, as soon as reschedule_retry
is called, another thread can mark the resync request as having
completed, so md_do_sync can finish, ->stop can be called, and
->conf can be freed. So using conf after reschedule_retry is not
safe.
Similarly, when finishing a sync_write, calling md_done_sync must be
the last thing we do, as it allows a chain of events which will free
conf and other data structures.
The first of these requires action in raid10.c
The second requires action in raid1.c and raid10.c
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
-rw-r--r-- | drivers/md/raid1.c | 3 | ||||
-rw-r--r-- | drivers/md/raid10.c | 7 |
2 files changed, 6 insertions, 4 deletions
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 03a5ab705c20..2b510a305082 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -1229,8 +1229,9 @@ static void end_sync_write(struct bio *bio, int error) update_head_pos(mirror, r1_bio); if (atomic_dec_and_test(&r1_bio->remaining)) { - md_done_sync(mddev, r1_bio->sectors, uptodate); + sector_t s = r1_bio->sectors; put_buf(r1_bio); + md_done_sync(mddev, s, uptodate); } } diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index 941576d4eceb..310493ff55ff 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -1231,6 +1231,7 @@ static void end_sync_read(struct bio *bio, int error) /* for reconstruct, we always reschedule after a read. * for resync, only after all reads */ + rdev_dec_pending(conf->mirrors[d].rdev, conf->mddev); if (test_bit(R10BIO_IsRecover, &r10_bio->state) || atomic_dec_and_test(&r10_bio->remaining)) { /* we have read all the blocks, @@ -1238,7 +1239,6 @@ static void end_sync_read(struct bio *bio, int error) */ reschedule_retry(r10_bio); } - rdev_dec_pending(conf->mirrors[d].rdev, conf->mddev); } static void end_sync_write(struct bio *bio, int error) @@ -1259,11 +1259,13 @@ static void end_sync_write(struct bio *bio, int error) update_head_pos(i, r10_bio); + rdev_dec_pending(conf->mirrors[d].rdev, mddev); while (atomic_dec_and_test(&r10_bio->remaining)) { if (r10_bio->master_bio == NULL) { /* the primary of several recovery bios */ - md_done_sync(mddev, r10_bio->sectors, 1); + sector_t s = r10_bio->sectors; put_buf(r10_bio); + md_done_sync(mddev, s, 1); break; } else { r10bio_t *r10_bio2 = (r10bio_t *)r10_bio->master_bio; @@ -1271,7 +1273,6 @@ static void end_sync_write(struct bio *bio, int error) r10_bio = r10_bio2; } } - rdev_dec_pending(conf->mirrors[d].rdev, mddev); } /* |