Merge tag 'dm-3.18-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:

 - stable fix for dm-thin that avoids normal IO racing with discard

 - stable fix for a dm-cache related bug in dm-btree walking code that
   results from using very large fast device (eg 4T) with a very small
   cache blocksize (eg 32K) -- this is a very uncommon configuration

 - a couple fixes for dm-raid (one for stable and the other addresses a
   crash in 3.18-rc1 code)

 - stable fix for dm-thinp that addresses a very rare dm-bufio bug
   having to do with memory reclaimation (via shrinker) when using
   dm-thinp ontop of loopback devices

 - fix a leak in dm-stripe target constructor's error path

* tag 'dm-3.18-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm btree: fix a recursion depth bug in btree walking code
  dm thin: grab a virtual cell before looking up the mapping
  dm raid: fix inaccessible superblocks causing oops in configure_discard_support
  dm raid: ensure superblock's size matches device's logical block size
  dm bufio: change __GFP_IO to __GFP_FS in shrinker callbacks
  dm stripe: fix potential for leak in stripe_ctr error path

drivers/md/dm-bufio.c

@@ -1434,9 +1434,9 @@ static void drop_buffers(struct dm_bufio_client *c)
 
 /*
  * Test if the buffer is unused and too old, and commit it.
- * At if noio is set, we must not do any I/O because we hold
- * dm_bufio_clients_lock and we would risk deadlock if the I/O gets rerouted to
- * different bufio client.
+ * And if GFP_NOFS is used, we must not do any I/O because we hold
+ * dm_bufio_clients_lock and we would risk deadlock if the I/O gets
+ * rerouted to different bufio client.
  */
 static int __cleanup_old_buffer(struct dm_buffer *b, gfp_t gfp,
 				unsigned long max_jiffies)
@@ -1444,7 +1444,7 @@ static int __cleanup_old_buffer(struct dm_buffer *b, gfp_t gfp,
 	if (jiffies - b->last_accessed < max_jiffies)
 		return 0;
 
-	if (!(gfp & __GFP_IO)) {
+	if (!(gfp & __GFP_FS)) {
 		if (test_bit(B_READING, &b->state) ||
 		    test_bit(B_WRITING, &b->state) ||
 		    test_bit(B_DIRTY, &b->state))
@@ -1486,7 +1486,7 @@ dm_bufio_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
 	unsigned long freed;
 
 	c = container_of(shrink, struct dm_bufio_client, shrinker);
-	if (sc->gfp_mask & __GFP_IO)
+	if (sc->gfp_mask & __GFP_FS)
 		dm_bufio_lock(c);
 	else if (!dm_bufio_trylock(c))
 		return SHRINK_STOP;
@@ -1503,7 +1503,7 @@ dm_bufio_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
 	unsigned long count;
 
 	c = container_of(shrink, struct dm_bufio_client, shrinker);
-	if (sc->gfp_mask & __GFP_IO)
+	if (sc->gfp_mask & __GFP_FS)
 		dm_bufio_lock(c);
 	else if (!dm_bufio_trylock(c))
 		return 0;

drivers/md/dm-raid.c

@@ -789,8 +789,7 @@ struct dm_raid_superblock {
 	__le32 layout;
 	__le32 stripe_sectors;
 
-	__u8 pad[452];		/* Round struct to 512 bytes. */
-				/* Always set to 0 when writing. */
+	/* Remainder of a logical block is zero-filled when writing (see super_sync()). */
 } __packed;
 
 static int read_disk_sb(struct md_rdev *rdev, int size)
@@ -827,7 +826,7 @@ static void super_sync(struct mddev *mddev, struct md_rdev *rdev)
 		    test_bit(Faulty, &(rs->dev[i].rdev.flags)))
 			failed_devices |= (1ULL << i);
 
-	memset(sb, 0, sizeof(*sb));
+	memset(sb + 1, 0, rdev->sb_size - sizeof(*sb));
 
 	sb->magic = cpu_to_le32(DM_RAID_MAGIC);
 	sb->features = cpu_to_le32(0);	/* No features yet */
@@ -862,7 +861,11 @@ static int super_load(struct md_rdev *rdev, struct md_rdev *refdev)
 	uint64_t events_sb, events_refsb;
 
 	rdev->sb_start = 0;
-	rdev->sb_size = sizeof(*sb);
+	rdev->sb_size = bdev_logical_block_size(rdev->meta_bdev);
+	if (rdev->sb_size < sizeof(*sb) || rdev->sb_size > PAGE_SIZE) {
+		DMERR("superblock size of a logical block is no longer valid");
+		return -EINVAL;
+	}
 
 	ret = read_disk_sb(rdev, rdev->sb_size);
 	if (ret)
@@ -1169,8 +1172,12 @@ static void configure_discard_support(struct dm_target *ti, struct raid_set *rs)
 	raid456 = (rs->md.level == 4 || rs->md.level == 5 || rs->md.level == 6);
 
 	for (i = 0; i < rs->md.raid_disks; i++) {
-		struct request_queue *q = bdev_get_queue(rs->dev[i].rdev.bdev);
+		struct request_queue *q;
+
+		if (!rs->dev[i].rdev.bdev)
+			continue;
 
+		q = bdev_get_queue(rs->dev[i].rdev.bdev);
 		if (!q || !blk_queue_discard(q))
 			return;
 

drivers/md/dm-stripe.c

@@ -159,8 +159,10 @@ static int stripe_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		sc->stripes_shift = __ffs(stripes);
 
 	r = dm_set_target_max_io_len(ti, chunk_size);
-	if (r)
+	if (r) {
+		kfree(sc);
 		return r;
+	}
 
 	ti->num_flush_bios = stripes;
 	ti->num_discard_bios = stripes;

drivers/md/dm-thin.c

@@ -1936,6 +1936,14 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio)
 		return DM_MAPIO_SUBMITTED;
 	}
 
+	/*
+	 * We must hold the virtual cell before doing the lookup, otherwise
+	 * there's a race with discard.
+	 */
+	build_virtual_key(tc->td, block, &key);
+	if (dm_bio_detain(tc->pool->prison, &key, bio, &cell1, &cell_result))
+		return DM_MAPIO_SUBMITTED;
+
 	r = dm_thin_find_block(td, block, 0, &result);
 
 	/*
@@ -1959,13 +1967,10 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio)
 			 * shared flag will be set in their case.
 			 */
 			thin_defer_bio(tc, bio);
+			cell_defer_no_holder_no_free(tc, &cell1);
 			return DM_MAPIO_SUBMITTED;
 		}
 
-		build_virtual_key(tc->td, block, &key);
-		if (dm_bio_detain(tc->pool->prison, &key, bio, &cell1, &cell_result))
-			return DM_MAPIO_SUBMITTED;
-
 		build_data_key(tc->td, result.block, &key);
 		if (dm_bio_detain(tc->pool->prison, &key, bio, &cell2, &cell_result)) {
 			cell_defer_no_holder_no_free(tc, &cell1);
@@ -1986,6 +1991,7 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio)
 			 * of doing so.
 			 */
 			handle_unserviceable_bio(tc->pool, bio);
+			cell_defer_no_holder_no_free(tc, &cell1);
 			return DM_MAPIO_SUBMITTED;
 		}
 		/* fall through */
@@ -1996,6 +2002,7 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio)
 		 * provide the hint to load the metadata into cache.
 		 */
 		thin_defer_bio(tc, bio);
+		cell_defer_no_holder_no_free(tc, &cell1);
 		return DM_MAPIO_SUBMITTED;
 
 	default:
@@ -2005,6 +2012,7 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio)
 		 * pool is switched to fail-io mode.
 		 */
 		bio_io_error(bio);
+		cell_defer_no_holder_no_free(tc, &cell1);
 		return DM_MAPIO_SUBMITTED;
 	}
 }

drivers/md/persistent-data/dm-btree-internal.h

@@ -42,6 +42,12 @@ struct btree_node {
 } __packed;
 
 
+/*
+ * Locks a block using the btree node validator.
+ */
+int bn_read_lock(struct dm_btree_info *info, dm_block_t b,
+		 struct dm_block **result);
+
 void inc_children(struct dm_transaction_manager *tm, struct btree_node *n,
 		  struct dm_btree_value_type *vt);
 

drivers/md/persistent-data/dm-btree-spine.c

@@ -92,7 +92,7 @@ struct dm_block_validator btree_node_validator = {
 
 /*----------------------------------------------------------------*/
 
-static int bn_read_lock(struct dm_btree_info *info, dm_block_t b,
+int bn_read_lock(struct dm_btree_info *info, dm_block_t b,
 		 struct dm_block **result)
 {
 	return dm_tm_read_lock(info->tm, b, &btree_node_validator, result);

drivers/md/persistent-data/dm-btree.c

@@ -847,22 +847,26 @@ EXPORT_SYMBOL_GPL(dm_btree_find_lowest_key);
  * FIXME: We shouldn't use a recursive algorithm when we have limited stack
  * space.  Also this only works for single level trees.
  */
-static int walk_node(struct ro_spine *s, dm_block_t block,
+static int walk_node(struct dm_btree_info *info, dm_block_t block,
 		     int (*fn)(void *context, uint64_t *keys, void *leaf),
 		     void *context)
 {
 	int r;
 	unsigned i, nr;
+	struct dm_block *node;
 	struct btree_node *n;
 	uint64_t keys;
 
-	r = ro_step(s, block);
-	n = ro_node(s);
+	r = bn_read_lock(info, block, &node);
+	if (r)
+		return r;
+
+	n = dm_block_data(node);
 
 	nr = le32_to_cpu(n->header.nr_entries);
 	for (i = 0; i < nr; i++) {
 		if (le32_to_cpu(n->header.flags) & INTERNAL_NODE) {
-			r = walk_node(s, value64(n, i), fn, context);
+			r = walk_node(info, value64(n, i), fn, context);
 			if (r)
 				goto out;
 		} else {
@@ -874,7 +878,7 @@ static int walk_node(struct ro_spine *s, dm_block_t block,
 	}
 
 out:
-	ro_pop(s);
+	dm_tm_unlock(info->tm, node);
 	return r;
 }
 
@@ -882,15 +886,7 @@ int dm_btree_walk(struct dm_btree_info *info, dm_block_t root,
 		  int (*fn)(void *context, uint64_t *keys, void *leaf),
 		  void *context)
 {
-	int r;
-	struct ro_spine spine;
-
 	BUG_ON(info->levels > 1);
-
-	init_ro_spine(&spine, info);
-	r = walk_node(&spine, root, fn, context);
-	exit_ro_spine(&spine);
-
-	return r;
+	return walk_node(info, root, fn, context);
 }
 EXPORT_SYMBOL_GPL(dm_btree_walk);