md: avoid races when stopping resync.
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
Cc: stable@kernel.org
Signed-off-by: NeilBrown <neilb@suse.de>
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index e1feb87..7301631 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -1236,6 +1236,7 @@
/* 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,
@@ -1243,7 +1244,6 @@
*/
reschedule_retry(r10_bio);
}
- rdev_dec_pending(conf->mirrors[d].rdev, conf->mddev);
}
static void end_sync_write(struct bio *bio, int error)
@@ -1264,11 +1264,13 @@
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;
@@ -1276,7 +1278,6 @@
r10_bio = r10_bio2;
}
}
- rdev_dec_pending(conf->mirrors[d].rdev, mddev);
}
/*