Merge tag 'dm-3.19-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:
 "Two stable fixes for dm-cache and one 3.19 DM core fix:

   - fix potential for dm-cache metadata corruption via stale metadata
     buffers being used when switching an inactive cache table to
     active; this could occur due to each table having it's own bufio
     client rather than sharing the client between tables.

   - fix dm-cache target to properly account for discard IO while
     suspending otherwise IO quiescing could complete prematurely.

   - fix DM core's handling of multiple internal suspends by maintaining
     an 'internal_suspend_count' and only resuming the device when this
     count drops to zero"

* tag 'dm-3.19-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm: fix handling of multiple internal suspends
  dm cache: fix problematic dual use of a single migration count variable
  dm cache: share cache-metadata object across inactive and active DM tables

3 files changed, 152 insertions, 47 deletions

diff --git a/drivers/md/dm-cache-metadata.c b/drivers/md/dm-cache-metadata.c
index 9fc616c2755e..21b156242e42 100644
--- a/drivers/md/dm-cache-metadata.c
+++ b/drivers/md/dm-cache-metadata.c
@@ -94,6 +94,9 @@ struct cache_disk_superblock {
 } __packed;
 
 struct dm_cache_metadata {
+	atomic_t ref_count;
+	struct list_head list;
+
 	struct block_device *bdev;
 	struct dm_block_manager *bm;
 	struct dm_space_map *metadata_sm;
@@ -669,10 +672,10 @@ static void unpack_value(__le64 value_le, dm_oblock_t *block, unsigned *flags)
 
 /*----------------------------------------------------------------*/
 
-struct dm_cache_metadata *dm_cache_metadata_open(struct block_device *bdev,
-						 sector_t data_block_size,
-						 bool may_format_device,
-						 size_t policy_hint_size)
+static struct dm_cache_metadata *metadata_open(struct block_device *bdev,
+					       sector_t data_block_size,
+					       bool may_format_device,
+					       size_t policy_hint_size)
 {
 	int r;
 	struct dm_cache_metadata *cmd;
@@ -683,6 +686,7 @@ struct dm_cache_metadata *dm_cache_metadata_open(struct block_device *bdev,
 		return NULL;
 	}
 
+	atomic_set(&cmd->ref_count, 1);
 	init_rwsem(&cmd->root_lock);
 	cmd->bdev = bdev;
 	cmd->data_block_size = data_block_size;
@@ -705,10 +709,95 @@ struct dm_cache_metadata *dm_cache_metadata_open(struct block_device *bdev,
 	return cmd;
 }
 
+/*
+ * We keep a little list of ref counted metadata objects to prevent two
+ * different target instances creating separate bufio instances.  This is
+ * an issue if a table is reloaded before the suspend.
+ */
+static DEFINE_MUTEX(table_lock);
+static LIST_HEAD(table);
+
+static struct dm_cache_metadata *lookup(struct block_device *bdev)
+{
+	struct dm_cache_metadata *cmd;
+
+	list_for_each_entry(cmd, &table, list)
+		if (cmd->bdev == bdev) {
+			atomic_inc(&cmd->ref_count);
+			return cmd;
+		}
+
+	return NULL;
+}
+
+static struct dm_cache_metadata *lookup_or_open(struct block_device *bdev,
+						sector_t data_block_size,
+						bool may_format_device,
+						size_t policy_hint_size)
+{
+	struct dm_cache_metadata *cmd, *cmd2;
+
+	mutex_lock(&table_lock);
+	cmd = lookup(bdev);
+	mutex_unlock(&table_lock);
+
+	if (cmd)
+		return cmd;
+
+	cmd = metadata_open(bdev, data_block_size, may_format_device, policy_hint_size);
+	if (cmd) {
+		mutex_lock(&table_lock);
+		cmd2 = lookup(bdev);
+		if (cmd2) {
+			mutex_unlock(&table_lock);
+			__destroy_persistent_data_objects(cmd);
+			kfree(cmd);
+			return cmd2;
+		}
+		list_add(&cmd->list, &table);
+		mutex_unlock(&table_lock);
+	}
+
+	return cmd;
+}
+
+static bool same_params(struct dm_cache_metadata *cmd, sector_t data_block_size)
+{
+	if (cmd->data_block_size != data_block_size) {
+		DMERR("data_block_size (%llu) different from that in metadata (%llu)\n",
+		      (unsigned long long) data_block_size,
+		      (unsigned long long) cmd->data_block_size);
+		return false;
+	}
+
+	return true;
+}
+
+struct dm_cache_metadata *dm_cache_metadata_open(struct block_device *bdev,
+						 sector_t data_block_size,
+						 bool may_format_device,
+						 size_t policy_hint_size)
+{
+	struct dm_cache_metadata *cmd = lookup_or_open(bdev, data_block_size,
+						       may_format_device, policy_hint_size);
+	if (cmd && !same_params(cmd, data_block_size)) {
+		dm_cache_metadata_close(cmd);
+		return NULL;
+	}
+
+	return cmd;
+}
+
 void dm_cache_metadata_close(struct dm_cache_metadata *cmd)
 {
-	__destroy_persistent_data_objects(cmd);
-	kfree(cmd);
+	if (atomic_dec_and_test(&cmd->ref_count)) {
+		mutex_lock(&table_lock);
+		list_del(&cmd->list);
+		mutex_unlock(&table_lock);
+
+		__destroy_persistent_data_objects(cmd);
+		kfree(cmd);
+	}
 }
 
 /*
diff --git a/drivers/md/dm-cache-target.c b/drivers/md/dm-cache-target.c
index 1e96d7889f51..e1650539cc2f 100644
--- a/drivers/md/dm-cache-target.c
+++ b/drivers/md/dm-cache-target.c
@@ -221,7 +221,13 @@ struct cache {
 	struct list_head need_commit_migrations;
 	sector_t migration_threshold;
 	wait_queue_head_t migration_wait;
-	atomic_t nr_migrations;
+	atomic_t nr_allocated_migrations;
+
+	/*
+	 * The number of in flight migrations that are performing
+	 * background io. eg, promotion, writeback.
+	 */
+	atomic_t nr_io_migrations;
 
 	wait_queue_head_t quiescing_wait;
 	atomic_t quiescing;
@@ -258,7 +264,6 @@ struct cache {
 	struct dm_deferred_set *all_io_ds;
 
 	mempool_t *migration_pool;
-	struct dm_cache_migration *next_migration;
 
 	struct dm_cache_policy *policy;
 	unsigned policy_nr_args;
@@ -350,10 +355,31 @@ static void free_prison_cell(struct cache *cache, struct dm_bio_prison_cell *cel
 	dm_bio_prison_free_cell(cache->prison, cell);
 }
 
+static struct dm_cache_migration *alloc_migration(struct cache *cache)
+{
+	struct dm_cache_migration *mg;
+
+	mg = mempool_alloc(cache->migration_pool, GFP_NOWAIT);
+	if (mg) {
+		mg->cache = cache;
+		atomic_inc(&mg->cache->nr_allocated_migrations);
+	}
+
+	return mg;
+}
+
+static void free_migration(struct dm_cache_migration *mg)
+{
+	if (atomic_dec_and_test(&mg->cache->nr_allocated_migrations))
+		wake_up(&mg->cache->migration_wait);
+
+	mempool_free(mg, mg->cache->migration_pool);
+}
+
 static int prealloc_data_structs(struct cache *cache, struct prealloc *p)
 {
 	if (!p->mg) {
-		p->mg = mempool_alloc(cache->migration_pool, GFP_NOWAIT);
+		p->mg = alloc_migration(cache);
 		if (!p->mg)
 			return -ENOMEM;
 	}
@@ -382,7 +408,7 @@ static void prealloc_free_structs(struct cache *cache, struct prealloc *p)
 		free_prison_cell(cache, p->cell1);
 
 	if (p->mg)
-		mempool_free(p->mg, cache->migration_pool);
+		free_migration(p->mg);
 }
 
 static struct dm_cache_migration *prealloc_get_migration(struct prealloc *p)
@@ -854,24 +880,14 @@ static void remap_to_origin_then_cache(struct cache *cache, struct bio *bio,
  * Migration covers moving data from the origin device to the cache, or
  * vice versa.
  *--------------------------------------------------------------*/
-static void free_migration(struct dm_cache_migration *mg)
-{
-	mempool_free(mg, mg->cache->migration_pool);
-}
-
-static void inc_nr_migrations(struct cache *cache)
+static void inc_io_migrations(struct cache *cache)
 {
-	atomic_inc(&cache->nr_migrations);
+	atomic_inc(&cache->nr_io_migrations);
 }
 
-static void dec_nr_migrations(struct cache *cache)
+static void dec_io_migrations(struct cache *cache)
 {
-	atomic_dec(&cache->nr_migrations);
-
-	/*
-	 * Wake the worker in case we're suspending the target.
-	 */
-	wake_up(&cache->migration_wait);
+	atomic_dec(&cache->nr_io_migrations);
 }
 
 static void __cell_defer(struct cache *cache, struct dm_bio_prison_cell *cell,
@@ -894,11 +910,10 @@ static void cell_defer(struct cache *cache, struct dm_bio_prison_cell *cell,
 	wake_worker(cache);
 }
 
-static void cleanup_migration(struct dm_cache_migration *mg)
+static void free_io_migration(struct dm_cache_migration *mg)
 {
-	struct cache *cache = mg->cache;
+	dec_io_migrations(mg->cache);
 	free_migration(mg);
-	dec_nr_migrations(cache);
 }
 
 static void migration_failure(struct dm_cache_migration *mg)
@@ -923,7 +938,7 @@ static void migration_failure(struct dm_cache_migration *mg)
 		cell_defer(cache, mg->new_ocell, true);
 	}
 
-	cleanup_migration(mg);
+	free_io_migration(mg);
 }
 
 static void migration_success_pre_commit(struct dm_cache_migration *mg)
@@ -934,7 +949,7 @@ static void migration_success_pre_commit(struct dm_cache_migration *mg)
 	if (mg->writeback) {
 		clear_dirty(cache, mg->old_oblock, mg->cblock);
 		cell_defer(cache, mg->old_ocell, false);
-		cleanup_migration(mg);
+		free_io_migration(mg);
 		return;
 
 	} else if (mg->demote) {
@@ -944,14 +959,14 @@ static void migration_success_pre_commit(struct dm_cache_migration *mg)
 					     mg->old_oblock);
 			if (mg->promote)
 				cell_defer(cache, mg->new_ocell, true);
-			cleanup_migration(mg);
+			free_io_migration(mg);
 			return;
 		}
 	} else {
 		if (dm_cache_insert_mapping(cache->cmd, mg->cblock, mg->new_oblock)) {
 			DMWARN_LIMIT("promotion failed; couldn't update on disk metadata");
 			policy_remove_mapping(cache->policy, mg->new_oblock);
-			cleanup_migration(mg);
+			free_io_migration(mg);
 			return;
 		}
 	}
@@ -984,7 +999,7 @@ static void migration_success_post_commit(struct dm_cache_migration *mg)
 		} else {
 			if (mg->invalidate)
 				policy_remove_mapping(cache->policy, mg->old_oblock);
-			cleanup_migration(mg);
+			free_io_migration(mg);
 		}
 
 	} else {
@@ -999,7 +1014,7 @@ static void migration_success_post_commit(struct dm_cache_migration *mg)
 			bio_endio(mg->new_ocell->holder, 0);
 			cell_defer(cache, mg->new_ocell, false);
 		}
-		cleanup_migration(mg);
+		free_io_migration(mg);
 	}
 }
 
@@ -1251,7 +1266,7 @@ static void promote(struct cache *cache, struct prealloc *structs,
 	mg->new_ocell = cell;
 	mg->start_jiffies = jiffies;
 
-	inc_nr_migrations(cache);
+	inc_io_migrations(cache);
 	quiesce_migration(mg);
 }
 
@@ -1275,7 +1290,7 @@ static void writeback(struct cache *cache, struct prealloc *structs,
 	mg->new_ocell = NULL;
 	mg->start_jiffies = jiffies;
 
-	inc_nr_migrations(cache);
+	inc_io_migrations(cache);
 	quiesce_migration(mg);
 }
 
@@ -1302,7 +1317,7 @@ static void demote_then_promote(struct cache *cache, struct prealloc *structs,
 	mg->new_ocell = new_ocell;
 	mg->start_jiffies = jiffies;
 
-	inc_nr_migrations(cache);
+	inc_io_migrations(cache);
 	quiesce_migration(mg);
 }
 
@@ -1330,7 +1345,7 @@ static void invalidate(struct cache *cache, struct prealloc *structs,
 	mg->new_ocell = NULL;
 	mg->start_jiffies = jiffies;
 
-	inc_nr_migrations(cache);
+	inc_io_migrations(cache);
 	quiesce_migration(mg);
 }
 
@@ -1412,7 +1427,7 @@ static void process_discard_bio(struct cache *cache, struct prealloc *structs,
 
 static bool spare_migration_bandwidth(struct cache *cache)
 {
-	sector_t current_volume = (atomic_read(&cache->nr_migrations) + 1) *
+	sector_t current_volume = (atomic_read(&cache->nr_io_migrations) + 1) *
 		cache->sectors_per_block;
 	return current_volume < cache->migration_threshold;
 }
@@ -1764,7 +1779,7 @@ static void stop_quiescing(struct cache *cache)
 
 static void wait_for_migrations(struct cache *cache)
 {
-	wait_event(cache->migration_wait, !atomic_read(&cache->nr_migrations));
+	wait_event(cache->migration_wait, !atomic_read(&cache->nr_allocated_migrations));
 }
 
 static void stop_worker(struct cache *cache)
@@ -1876,9 +1891,6 @@ static void destroy(struct cache *cache)
 {
 	unsigned i;
 
-	if (cache->next_migration)
-		mempool_free(cache->next_migration, cache->migration_pool);
-
 	if (cache->migration_pool)
 		mempool_destroy(cache->migration_pool);
 
@@ -2424,7 +2436,8 @@ static int cache_create(struct cache_args *ca, struct cache **result)
 	INIT_LIST_HEAD(&cache->quiesced_migrations);
 	INIT_LIST_HEAD(&cache->completed_migrations);
 	INIT_LIST_HEAD(&cache->need_commit_migrations);
-	atomic_set(&cache->nr_migrations, 0);
+	atomic_set(&cache->nr_allocated_migrations, 0);
+	atomic_set(&cache->nr_io_migrations, 0);
 	init_waitqueue_head(&cache->migration_wait);
 
 	init_waitqueue_head(&cache->quiescing_wait);
@@ -2487,8 +2500,6 @@ static int cache_create(struct cache_args *ca, struct cache **result)
 		goto bad;
 	}
 
-	cache->next_migration = NULL;
-
 	cache->need_tick_bio = true;
 	cache->sized = false;
 	cache->invalidate = false;
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index b98cd9d84435..2caf5b374649 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -206,6 +206,9 @@ struct mapped_device {
 	/* zero-length flush that will be cloned and submitted to targets */
 	struct bio flush_bio;
 
+	/* the number of internal suspends */
+	unsigned internal_suspend_count;
+
 	struct dm_stats stats;
 };
 
@@ -2928,7 +2931,7 @@ static void __dm_internal_suspend(struct mapped_device *md, unsigned suspend_fla
 {
 	struct dm_table *map = NULL;
 
-	if (dm_suspended_internally_md(md))
+	if (md->internal_suspend_count++)
 		return; /* nested internal suspend */
 
 	if (dm_suspended_md(md)) {
@@ -2953,7 +2956,9 @@ static void __dm_internal_suspend(struct mapped_device *md, unsigned suspend_fla
 
 static void __dm_internal_resume(struct mapped_device *md)
 {
-	if (!dm_suspended_internally_md(md))
+	BUG_ON(!md->internal_suspend_count);
+
+	if (--md->internal_suspend_count)
 		return; /* resume from nested internal suspend */
 
 	if (dm_suspended_md(md))