Merge tag 'dm-3.14-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:

 - dm-cache memory allocation failure fix
 - fix DM's Kconfig identation
 - dm-snapshot metadata corruption fix for bug introduced in 3.14-rc1
 - important refcount < 0 fix for the DM persistent data library's space
   map metadata interface which fixes corruption reported by a few
   dm-thinp users

and last but not least:

 - more extensive fixes than ideal for dm-thinp's data resize capability
   (which has had growing pain much like we've seen from -ENOSPC
   handling of filesystems that mature).

   The end result is dm-thinp now handles metadata operation failure and
   no data space error conditions much better than before.

* tag 'dm-3.14-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm space map metadata: fix refcount decrement below 0 which caused corruption
  dm thin: fix Documentation for held metadata root feature
  dm thin: fix noflush suspend IO queueing
  dm thin: fix deadlock in __requeue_bio_list
  dm thin: fix out of data space handling
  dm thin: ensure user takes action to validate data and metadata consistency
  dm thin: synchronize the pool mode during suspend
  dm snapshot: fix metadata corruption
  dm: fix Kconfig indentation
  dm cache mq: fix memory allocation failure for large cache devices

Documentation/device-mapper/cache.txt

@@ -124,12 +124,11 @@ the default being 204800 sectors (or 100MB).
 Updating on-disk metadata
 -------------------------
 
-On-disk metadata is committed every time a REQ_SYNC or REQ_FUA bio is
-written.  If no such requests are made then commits will occur every
-second.  This means the cache behaves like a physical disk that has a
-write cache (the same is true of the thin-provisioning target).  If
-power is lost you may lose some recent writes.  The metadata should
-always be consistent in spite of any crash.
+On-disk metadata is committed every time a FLUSH or FUA bio is written.
+If no such requests are made then commits will occur every second.  This
+means the cache behaves like a physical disk that has a volatile write
+cache.  If power is lost you may lose some recent writes.  The metadata
+should always be consistent in spite of any crash.
 
 The 'dirty' state for a cache block changes far too frequently for us
 to keep updating it on the fly.  So we treat it as a hint.  In normal

Documentation/device-mapper/thin-provisioning.txt

@@ -116,6 +116,35 @@ Resuming a device with a new table itself triggers an event so the
 userspace daemon can use this to detect a situation where a new table
 already exceeds the threshold.
 
+A low water mark for the metadata device is maintained in the kernel and
+will trigger a dm event if free space on the metadata device drops below
+it.
+
+Updating on-disk metadata
+-------------------------
+
+On-disk metadata is committed every time a FLUSH or FUA bio is written.
+If no such requests are made then commits will occur every second.  This
+means the thin-provisioning target behaves like a physical disk that has
+a volatile write cache.  If power is lost you may lose some recent
+writes.  The metadata should always be consistent in spite of any crash.
+
+If data space is exhausted the pool will either error or queue IO
+according to the configuration (see: error_if_no_space).  If metadata
+space is exhausted or a metadata operation fails: the pool will error IO
+until the pool is taken offline and repair is performed to 1) fix any
+potential inconsistencies and 2) clear the flag that imposes repair.
+Once the pool's metadata device is repaired it may be resized, which
+will allow the pool to return to normal operation.  Note that if a pool
+is flagged as needing repair, the pool's data and metadata devices
+cannot be resized until repair is performed.  It should also be noted
+that when the pool's metadata space is exhausted the current metadata
+transaction is aborted.  Given that the pool will cache IO whose
+completion may have already been acknowledged to upper IO layers
+(e.g. filesystem) it is strongly suggested that consistency checks
+(e.g. fsck) be performed on those layers when repair of the pool is
+required.
+
 Thin provisioning
 -----------------
 
@@ -258,10 +287,9 @@ ii) Status
 	should register for the event and then check the target's status.
 
     held metadata root:
-	The location, in sectors, of the metadata root that has been
+	The location, in blocks, of the metadata root that has been
 	'held' for userspace read access.  '-' indicates there is no
-	held root.  This feature is not yet implemented so '-' is
-	always returned.
+	held root.
 
     discard_passdown|no_discard_passdown
 	Whether or not discards are actually being passed down to the

drivers/md/Kconfig

@@ -254,16 +254,6 @@ config DM_THIN_PROVISIONING
        ---help---
          Provides thin provisioning and snapshots that share a data store.
 
-config DM_DEBUG_BLOCK_STACK_TRACING
-	boolean "Keep stack trace of persistent data block lock holders"
-	depends on STACKTRACE_SUPPORT && DM_PERSISTENT_DATA
-	select STACKTRACE
-	---help---
-	  Enable this for messages that may help debug problems with the
-	  block manager locking used by thin provisioning and caching.
-
-	  If unsure, say N.
-
 config DM_CACHE
        tristate "Cache target (EXPERIMENTAL)"
        depends on BLK_DEV_DM

drivers/md/dm-cache-policy-mq.c

@@ -872,7 +872,7 @@ static void mq_destroy(struct dm_cache_policy *p)
 {
 	struct mq_policy *mq = to_mq_policy(p);
 
-	kfree(mq->table);
+	vfree(mq->table);
 	epool_exit(&mq->cache_pool);
 	epool_exit(&mq->pre_cache_pool);
 	kfree(mq);
@@ -1245,7 +1245,7 @@ static struct dm_cache_policy *mq_create(dm_cblock_t cache_size,
 
 	mq->nr_buckets = next_power(from_cblock(cache_size) / 2, 16);
 	mq->hash_bits = ffs(mq->nr_buckets) - 1;
-	mq->table = kzalloc(sizeof(*mq->table) * mq->nr_buckets, GFP_KERNEL);
+	mq->table = vzalloc(sizeof(*mq->table) * mq->nr_buckets);
 	if (!mq->table)
 		goto bad_alloc_table;
 

drivers/md/dm-snap-persistent.c

@@ -546,6 +546,9 @@ static int read_exceptions(struct pstore *ps,
 		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);

drivers/md/dm-thin-metadata.c

@@ -76,7 +76,7 @@
 
 #define THIN_SUPERBLOCK_MAGIC 27022010
 #define THIN_SUPERBLOCK_LOCATION 0
-#define THIN_VERSION 1
+#define THIN_VERSION 2
 #define THIN_METADATA_CACHE_SIZE 64
 #define SECTOR_TO_BLOCK_SHIFT 3
 
@@ -1755,3 +1755,38 @@ int dm_pool_register_metadata_threshold(struct dm_pool_metadata *pmd,
 
 	return r;
 }
+
+int dm_pool_metadata_set_needs_check(struct dm_pool_metadata *pmd)
+{
+	int r;
+	struct dm_block *sblock;
+	struct thin_disk_superblock *disk_super;
+
+	down_write(&pmd->root_lock);
+	pmd->flags |= THIN_METADATA_NEEDS_CHECK_FLAG;
+
+	r = superblock_lock(pmd, &sblock);
+	if (r) {
+		DMERR("couldn't read superblock");
+		goto out;
+	}
+
+	disk_super = dm_block_data(sblock);
+	disk_super->flags = cpu_to_le32(pmd->flags);
+
+	dm_bm_unlock(sblock);
+out:
+	up_write(&pmd->root_lock);
+	return r;
+}
+
+bool dm_pool_metadata_needs_check(struct dm_pool_metadata *pmd)
+{
+	bool needs_check;
+
+	down_read(&pmd->root_lock);
+	needs_check = pmd->flags & THIN_METADATA_NEEDS_CHECK_FLAG;
+	up_read(&pmd->root_lock);
+
+	return needs_check;
+}

drivers/md/dm-thin-metadata.h

@@ -25,6 +25,11 @@
 
 /*----------------------------------------------------------------*/
 
+/*
+ * Thin metadata superblock flags.
+ */
+#define THIN_METADATA_NEEDS_CHECK_FLAG (1 << 0)
+
 struct dm_pool_metadata;
 struct dm_thin_device;
 
@@ -202,6 +207,12 @@ int dm_pool_register_metadata_threshold(struct dm_pool_metadata *pmd,
 					dm_sm_threshold_fn fn,
 					void *context);
 
+/*
+ * Updates the superblock immediately.
+ */
+int dm_pool_metadata_set_needs_check(struct dm_pool_metadata *pmd);
+bool dm_pool_metadata_needs_check(struct dm_pool_metadata *pmd);
+
 /*----------------------------------------------------------------*/
 
 #endif

drivers/md/persistent-data/Kconfig

@@ -6,3 +6,13 @@ config DM_PERSISTENT_DATA
        ---help---
 	 Library providing immutable on-disk data structure support for
 	 device-mapper targets such as the thin provisioning target.
+
+config DM_DEBUG_BLOCK_STACK_TRACING
+       boolean "Keep stack trace of persistent data block lock holders"
+       depends on STACKTRACE_SUPPORT && DM_PERSISTENT_DATA
+       select STACKTRACE
+       ---help---
+	 Enable this for messages that may help debug problems with the
+	 block manager locking used by thin provisioning and caching.
+
+	 If unsure, say N.

drivers/md/persistent-data/dm-space-map-metadata.c

@@ -91,6 +91,69 @@ struct block_op {
 	dm_block_t block;
 };
 
+struct bop_ring_buffer {
+	unsigned begin;
+	unsigned end;
+	struct block_op bops[MAX_RECURSIVE_ALLOCATIONS + 1];
+};
+
+static void brb_init(struct bop_ring_buffer *brb)
+{
+	brb->begin = 0;
+	brb->end = 0;
+}
+
+static bool brb_empty(struct bop_ring_buffer *brb)
+{
+	return brb->begin == brb->end;
+}
+
+static unsigned brb_next(struct bop_ring_buffer *brb, unsigned old)
+{
+	unsigned r = old + 1;
+	return (r >= (sizeof(brb->bops) / sizeof(*brb->bops))) ? 0 : r;
+}
+
+static int brb_push(struct bop_ring_buffer *brb,
+		    enum block_op_type type, dm_block_t b)
+{
+	struct block_op *bop;
+	unsigned next = brb_next(brb, brb->end);
+
+	/*
+	 * We don't allow the last bop to be filled, this way we can
+	 * differentiate between full and empty.
+	 */
+	if (next == brb->begin)
+		return -ENOMEM;
+
+	bop = brb->bops + brb->end;
+	bop->type = type;
+	bop->block = b;
+
+	brb->end = next;
+
+	return 0;
+}
+
+static int brb_pop(struct bop_ring_buffer *brb, struct block_op *result)
+{
+	struct block_op *bop;
+
+	if (brb_empty(brb))
+		return -ENODATA;
+
+	bop = brb->bops + brb->begin;
+	result->type = bop->type;
+	result->block = bop->block;
+
+	brb->begin = brb_next(brb, brb->begin);
+
+	return 0;
+}
+
+/*----------------------------------------------------------------*/
+
 struct sm_metadata {
 	struct dm_space_map sm;
 
@@ -101,25 +164,20 @@ struct sm_metadata {
 
 	unsigned recursion_count;
 	unsigned allocated_this_transaction;
-	unsigned nr_uncommitted;
-	struct block_op uncommitted[MAX_RECURSIVE_ALLOCATIONS];
+	struct bop_ring_buffer uncommitted;
 
 	struct threshold threshold;
 };
 
 static int add_bop(struct sm_metadata *smm, enum block_op_type type, dm_block_t b)
 {
-	struct block_op *op;
+	int r = brb_push(&smm->uncommitted, type, b);
 
-	if (smm->nr_uncommitted == MAX_RECURSIVE_ALLOCATIONS) {
+	if (r) {
 		DMERR("too many recursive allocations");
 		return -ENOMEM;
 	}
 
-	op = smm->uncommitted + smm->nr_uncommitted++;
-	op->type = type;
-	op->block = b;
-
 	return 0;
 }
 
@@ -158,11 +216,17 @@ static int out(struct sm_metadata *smm)
 		return -ENOMEM;
 	}
 
-	if (smm->recursion_count == 1 && smm->nr_uncommitted) {
-		while (smm->nr_uncommitted && !r) {
-			smm->nr_uncommitted--;
-			r = commit_bop(smm, smm->uncommitted +
-				       smm->nr_uncommitted);
+	if (smm->recursion_count == 1) {
+		while (!brb_empty(&smm->uncommitted)) {
+			struct block_op bop;
+
+			r = brb_pop(&smm->uncommitted, &bop);
+			if (r) {
+				DMERR("bug in bop ring buffer");
+				break;
+			}
+
+			r = commit_bop(smm, &bop);
 			if (r)
 				break;
 		}
@@ -217,7 +281,8 @@ static int sm_metadata_get_nr_free(struct dm_space_map *sm, dm_block_t *count)
 static int sm_metadata_get_count(struct dm_space_map *sm, dm_block_t b,
 				 uint32_t *result)
 {
-	int r, i;
+	int r;
+	unsigned i;
 	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
 	unsigned adjustment = 0;
 
@@ -225,8 +290,10 @@ static int sm_metadata_get_count(struct dm_space_map *sm, dm_block_t b,
 	 * We may have some uncommitted adjustments to add.  This list
 	 * should always be really short.
 	 */
-	for (i = 0; i < smm->nr_uncommitted; i++) {
-		struct block_op *op = smm->uncommitted + i;
+	for (i = smm->uncommitted.begin;
+	     i != smm->uncommitted.end;
+	     i = brb_next(&smm->uncommitted, i)) {
+		struct block_op *op = smm->uncommitted.bops + i;
 
 		if (op->block != b)
 			continue;
@@ -254,7 +321,8 @@ static int sm_metadata_get_count(struct dm_space_map *sm, dm_block_t b,
 static int sm_metadata_count_is_more_than_one(struct dm_space_map *sm,
 					      dm_block_t b, int *result)
 {
-	int r, i, adjustment = 0;
+	int r, adjustment = 0;
+	unsigned i;
 	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
 	uint32_t rc;
 
@@ -262,8 +330,11 @@ static int sm_metadata_count_is_more_than_one(struct dm_space_map *sm,
 	 * We may have some uncommitted adjustments to add.  This list
 	 * should always be really short.
 	 */
-	for (i = 0; i < smm->nr_uncommitted; i++) {
-		struct block_op *op = smm->uncommitted + i;
+	for (i = smm->uncommitted.begin;
+	     i != smm->uncommitted.end;
+	     i = brb_next(&smm->uncommitted, i)) {
+
+		struct block_op *op = smm->uncommitted.bops + i;
 
 		if (op->block != b)
 			continue;
@@ -671,7 +742,7 @@ int dm_sm_metadata_create(struct dm_space_map *sm,
 	smm->begin = superblock + 1;
 	smm->recursion_count = 0;
 	smm->allocated_this_transaction = 0;
-	smm->nr_uncommitted = 0;
+	brb_init(&smm->uncommitted);
 	threshold_init(&smm->threshold);
 
 	memcpy(&smm->sm, &bootstrap_ops, sizeof(smm->sm));
@@ -715,7 +786,7 @@ int dm_sm_metadata_open(struct dm_space_map *sm,
 	smm->begin = 0;
 	smm->recursion_count = 0;
 	smm->allocated_this_transaction = 0;
-	smm->nr_uncommitted = 0;
+	brb_init(&smm->uncommitted);
 	threshold_init(&smm->threshold);
 
 	memcpy(&smm->old_ll, &smm->ll, sizeof(smm->old_ll));