The patch 55494bf2947dccdf2d98b62374fea7365dfead84 broke snapshots. When loading the list of exceptions, the old code read exception areas one by one into ps->area and inserted the exceptions into the hash table. The patch 55494bf2 changed it so that dm-bufio with prefetch is used to load exceptions. It loads the exceptions correctly, but leaves ps->area uninitialized. When a new exception is allocated, it is stored in this uninitialized ps->area location and full ps->area is written to the disk. This causes metadata corruption. This patch fixes the bug by copying the last area that was read via dm-bufio into ps->area. Signed-off-by: Mikulas Patocka <mpatocka@xxxxxxxxxx> --- drivers/md/dm-snap-persistent.c | 3 +++ 1 file changed, 3 insertions(+) Index: linux-3.14-rc5/drivers/md/dm-snap-persistent.c =================================================================== --- linux-3.14-rc5.orig/drivers/md/dm-snap-persistent.c 2014-03-03 22:42:32.000000000 +0100 +++ linux-3.14-rc5/drivers/md/dm-snap-persistent.c 2014-03-03 22:45:45.000000000 +0100 @@ -546,6 +546,9 @@ static int read_exceptions(struct pstore r = insert_exceptions(ps, area, callback, callback_context, &full); + if (!full) + memcpy(ps->area, area, ps->store->chunk_size << SECTOR_SHIFT); + dm_bufio_release(bp); dm_bufio_forget(client, chunk); -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
