Merge tag 'dm-3.10-changes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-dm

Pull device-mapper updates from Alasdair Kergon:
 "Allow devices that hold metadata for the device-mapper thin
  provisioning target to be extended easily; allow WRITE SAME on
  multipath devices; an assortment of little fixes and clean-ups."

* tag 'dm-3.10-changes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-dm: (21 commits)
  dm cache: set config value
  dm cache: move config fns
  dm thin: generate event when metadata threshold passed
  dm persistent metadata: add space map threshold callback
  dm persistent data: add threshold callback to space map
  dm thin: detect metadata device resizing
  dm persistent data: support space map resizing
  dm thin: open dev read only when possible
  dm thin: refactor data dev resize
  dm cache: replace memcpy with struct assignment
  dm cache: fix typos in comments
  dm cache policy: fix description of lookup fn
  dm: document iterate_devices
  dm persistent data: fix error message typos
  dm cache: tune migration throttling
  dm mpath: enable WRITE SAME support
  dm table: fix write same support
  dm bufio: avoid a possible __vmalloc deadlock
  dm snapshot: fix error return code in snapshot_ctr
  dm cache: fix error return code in cache_create
  ...

drivers/md/dm-bufio.c

@@ -319,6 +319,9 @@ static void __cache_size_refresh(void)
 static void *alloc_buffer_data(struct dm_bufio_client *c, gfp_t gfp_mask,
 			       enum data_mode *data_mode)
 {
+	unsigned noio_flag;
+	void *ptr;
+
 	if (c->block_size <= DM_BUFIO_BLOCK_SIZE_SLAB_LIMIT) {
 		*data_mode = DATA_MODE_SLAB;
 		return kmem_cache_alloc(DM_BUFIO_CACHE(c), gfp_mask);
@@ -332,7 +335,26 @@ static void *alloc_buffer_data(struct dm_bufio_client *c, gfp_t gfp_mask,
 	}
 
 	*data_mode = DATA_MODE_VMALLOC;
-	return __vmalloc(c->block_size, gfp_mask, PAGE_KERNEL);
+
+	/*
+	 * __vmalloc allocates the data pages and auxiliary structures with
+	 * gfp_flags that were specified, but pagetables are always allocated
+	 * with GFP_KERNEL, no matter what was specified as gfp_mask.
+	 *
+	 * Consequently, we must set per-process flag PF_MEMALLOC_NOIO so that
+	 * all allocations done by this process (including pagetables) are done
+	 * as if GFP_NOIO was specified.
+	 */
+
+	if (gfp_mask & __GFP_NORETRY)
+		noio_flag = memalloc_noio_save();
+
+	ptr = __vmalloc(c->block_size, gfp_mask, PAGE_KERNEL);
+
+	if (gfp_mask & __GFP_NORETRY)
+		memalloc_noio_restore(noio_flag);
+
+	return ptr;
 }
 
 /*

drivers/md/dm-cache-metadata.c

@@ -1044,7 +1044,7 @@ void dm_cache_metadata_get_stats(struct dm_cache_metadata *cmd,
 				 struct dm_cache_statistics *stats)
 {
 	down_read(&cmd->root_lock);
-	memcpy(stats, &cmd->stats, sizeof(*stats));
+	*stats = cmd->stats;
 	up_read(&cmd->root_lock);
 }
 
@@ -1052,7 +1052,7 @@ void dm_cache_metadata_set_stats(struct dm_cache_metadata *cmd,
 				 struct dm_cache_statistics *stats)
 {
 	down_write(&cmd->root_lock);
-	memcpy(&cmd->stats, stats, sizeof(*stats));
+	cmd->stats = *stats;
 	up_write(&cmd->root_lock);
 }
 

drivers/md/dm-cache-policy.h

@@ -130,8 +130,8 @@ struct dm_cache_policy {
 	 *
 	 * Must not block.
 	 *
-	 * Returns 1 iff in cache, 0 iff not, < 0 on error (-EWOULDBLOCK
-	 * would be typical).
+	 * Returns 0 if in cache, -ENOENT if not, < 0 for other errors
+	 * (-EWOULDBLOCK would be typical).
 	 */
 	int (*lookup)(struct dm_cache_policy *p, dm_oblock_t oblock, dm_cblock_t *cblock);
 

drivers/md/dm-cache-target.c

@@ -205,7 +205,7 @@ struct per_bio_data {
 	/*
 	 * writethrough fields.  These MUST remain at the end of this
 	 * structure and the 'cache' member must be the first as it
-	 * is used to determine the offsetof the writethrough fields.
+	 * is used to determine the offset of the writethrough fields.
 	 */
 	struct cache *cache;
 	dm_cblock_t cblock;
@@ -393,7 +393,7 @@ static int get_cell(struct cache *cache,
 	return r;
 }
 
- /*----------------------------------------------------------------*/
+/*----------------------------------------------------------------*/
 
 static bool is_dirty(struct cache *cache, dm_cblock_t b)
 {
@@ -419,6 +419,7 @@ static void clear_dirty(struct cache *cache, dm_oblock_t oblock, dm_cblock_t cbl
 }
 
 /*----------------------------------------------------------------*/
+
 static bool block_size_is_power_of_two(struct cache *cache)
 {
 	return cache->sectors_per_block_shift >= 0;
@@ -667,7 +668,7 @@ static void writethrough_endio(struct bio *bio, int err)
 
 	/*
 	 * We can't issue this bio directly, since we're in interrupt
-	 * context.  So it get's put on a bio list for processing by the
+	 * context.  So it gets put on a bio list for processing by the
 	 * worker thread.
 	 */
 	defer_writethrough_bio(pb->cache, bio);
@@ -1445,6 +1446,7 @@ static void do_worker(struct work_struct *ws)
 static void do_waker(struct work_struct *ws)
 {
 	struct cache *cache = container_of(to_delayed_work(ws), struct cache, waker);
+	policy_tick(cache->policy);
 	wake_worker(cache);
 	queue_delayed_work(cache->wq, &cache->waker, COMMIT_PERIOD);
 }
@@ -1809,7 +1811,37 @@ static int parse_cache_args(struct cache_args *ca, int argc, char **argv,
 
 static struct kmem_cache *migration_cache;
 
-static int set_config_values(struct dm_cache_policy *p, int argc, const char **argv)
+#define NOT_CORE_OPTION 1
+
+static int process_config_option(struct cache *cache, const char *key, const char *value)
+{
+	unsigned long tmp;
+
+	if (!strcasecmp(key, "migration_threshold")) {
+		if (kstrtoul(value, 10, &tmp))
+			return -EINVAL;
+
+		cache->migration_threshold = tmp;
+		return 0;
+	}
+
+	return NOT_CORE_OPTION;
+}
+
+static int set_config_value(struct cache *cache, const char *key, const char *value)
+{
+	int r = process_config_option(cache, key, value);
+
+	if (r == NOT_CORE_OPTION)
+		r = policy_set_config_value(cache->policy, key, value);
+
+	if (r)
+		DMWARN("bad config value for %s: %s", key, value);
+
+	return r;
+}
+
+static int set_config_values(struct cache *cache, int argc, const char **argv)
 {
 	int r = 0;
 
@@ -1819,12 +1851,9 @@ static int set_config_values(struct dm_cache_policy *p, int argc, const char **a
 	}
 
 	while (argc) {
-		r = policy_set_config_value(p, argv[0], argv[1]);
-		if (r) {
-			DMWARN("policy_set_config_value failed: key = '%s', value = '%s'",
-			       argv[0], argv[1]);
-			return r;
-		}
+		r = set_config_value(cache, argv[0], argv[1]);
+		if (r)
+			break;
 
 		argc -= 2;
 		argv += 2;
@@ -1836,8 +1865,6 @@ static int set_config_values(struct dm_cache_policy *p, int argc, const char **a
 static int create_cache_policy(struct cache *cache, struct cache_args *ca,
 			       char **error)
 {
-	int r;
-
 	cache->policy =	dm_cache_policy_create(ca->policy_name,
 					       cache->cache_size,
 					       cache->origin_sectors,
@@ -1847,14 +1874,7 @@ static int create_cache_policy(struct cache *cache, struct cache_args *ca,
 		return -ENOMEM;
 	}
 
-	r = set_config_values(cache->policy, ca->policy_argc, ca->policy_argv);
-	if (r) {
-		*error = "Error setting cache policy's config values";
-		dm_cache_policy_destroy(cache->policy);
-		cache->policy = NULL;
-	}
-
-	return r;
+	return 0;
 }
 
 /*
@@ -1886,7 +1906,7 @@ static sector_t calculate_discard_block_size(sector_t cache_block_size,
 	return discard_block_size;
 }
 
-#define DEFAULT_MIGRATION_THRESHOLD (2048 * 100)
+#define DEFAULT_MIGRATION_THRESHOLD 2048
 
 static int cache_create(struct cache_args *ca, struct cache **result)
 {
@@ -1911,7 +1931,7 @@ static int cache_create(struct cache_args *ca, struct cache **result)
 	ti->discards_supported = true;
 	ti->discard_zeroes_data_unsupported = true;
 
-	memcpy(&cache->features, &ca->features, sizeof(cache->features));
+	cache->features = ca->features;
 	ti->per_bio_data_size = get_per_bio_data_size(cache);
 
 	cache->callbacks.congested_fn = cache_is_congested;
@@ -1948,7 +1968,15 @@ static int cache_create(struct cache_args *ca, struct cache **result)
 	r = create_cache_policy(cache, ca, error);
 	if (r)
 		goto bad;
+
 	cache->policy_nr_args = ca->policy_argc;
+	cache->migration_threshold = DEFAULT_MIGRATION_THRESHOLD;
+
+	r = set_config_values(cache, ca->policy_argc, ca->policy_argv);
+	if (r) {
+		*error = "Error setting cache policy's config values";
+		goto bad;
+	}
 
 	cmd = dm_cache_metadata_open(cache->metadata_dev->bdev,
 				     ca->block_size, may_format,
@@ -1967,10 +1995,10 @@ 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);
-	cache->migration_threshold = DEFAULT_MIGRATION_THRESHOLD;
 	atomic_set(&cache->nr_migrations, 0);
 	init_waitqueue_head(&cache->migration_wait);
 
+	r = -ENOMEM;
 	cache->nr_dirty = 0;
 	cache->dirty_bitset = alloc_bitset(from_cblock(cache->cache_size));
 	if (!cache->dirty_bitset) {
@@ -2517,23 +2545,6 @@ err:
 	DMEMIT("Error");
 }
 
-#define NOT_CORE_OPTION 1
-
-static int process_config_option(struct cache *cache, char **argv)
-{
-	unsigned long tmp;
-
-	if (!strcasecmp(argv[0], "migration_threshold")) {
-		if (kstrtoul(argv[1], 10, &tmp))
-			return -EINVAL;
-
-		cache->migration_threshold = tmp;
-		return 0;
-	}
-
-	return NOT_CORE_OPTION;
-}
-
 /*
  * Supports <key> <value>.
  *
@@ -2541,17 +2552,12 @@ static int process_config_option(struct cache *cache, char **argv)
  */
 static int cache_message(struct dm_target *ti, unsigned argc, char **argv)
 {
-	int r;
 	struct cache *cache = ti->private;
 
 	if (argc != 2)
 		return -EINVAL;
 
-	r = process_config_option(cache, argv);
-	if (r == NOT_CORE_OPTION)
-		return policy_set_config_value(cache->policy, argv[0], argv[1]);
-
-	return r;
+	return set_config_value(cache, argv[0], argv[1]);
 }
 
 static int cache_iterate_devices(struct dm_target *ti,
@@ -2609,7 +2615,7 @@ static void cache_io_hints(struct dm_target *ti, struct queue_limits *limits)
 
 static struct target_type cache_target = {
 	.name = "cache",
-	.version = {1, 1, 0},
+	.version = {1, 1, 1},
 	.module = THIS_MODULE,
 	.ctr = cache_ctr,
 	.dtr = cache_dtr,

drivers/md/dm-mpath.c

@@ -907,6 +907,7 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc,
 
 	ti->num_flush_bios = 1;
 	ti->num_discard_bios = 1;
+	ti->num_write_same_bios = 1;
 
 	return 0;
 

drivers/md/dm-snap.c

@@ -1121,6 +1121,7 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	s->pending_pool = mempool_create_slab_pool(MIN_IOS, pending_cache);
 	if (!s->pending_pool) {
 		ti->error = "Could not allocate mempool for pending exceptions";
+		r = -ENOMEM;
 		goto bad_pending_pool;
 	}
 

drivers/md/dm-stripe.c

@@ -94,7 +94,7 @@ static int get_stripe(struct dm_target *ti, struct stripe_c *sc,
 static int stripe_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 {
 	struct stripe_c *sc;
-	sector_t width;
+	sector_t width, tmp_len;
 	uint32_t stripes;
 	uint32_t chunk_size;
 	int r;
@@ -116,15 +116,16 @@ static int stripe_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	}
 
 	width = ti->len;
-	if (sector_div(width, chunk_size)) {
+	if (sector_div(width, stripes)) {
 		ti->error = "Target length not divisible by "
-		    "chunk size";
+		    "number of stripes";
 		return -EINVAL;
 	}
 
-	if (sector_div(width, stripes)) {
+	tmp_len = width;
+	if (sector_div(tmp_len, chunk_size)) {
 		ti->error = "Target length not divisible by "
-		    "number of stripes";
+		    "chunk size";
 		return -EINVAL;
 	}
 

drivers/md/dm-table.c

@@ -1442,7 +1442,7 @@ static bool dm_table_supports_write_same(struct dm_table *t)
 			return false;
 
 		if (!ti->type->iterate_devices ||
-		    !ti->type->iterate_devices(ti, device_not_write_same_capable, NULL))
+		    ti->type->iterate_devices(ti, device_not_write_same_capable, NULL))
 			return false;
 	}
 

drivers/md/dm-thin-metadata.c

@@ -1645,12 +1645,12 @@ int dm_thin_get_highest_mapped_block(struct dm_thin_device *td,
 	return r;
 }
 
-static int __resize_data_dev(struct dm_pool_metadata *pmd, dm_block_t new_count)
+static int __resize_space_map(struct dm_space_map *sm, dm_block_t new_count)
 {
 	int r;
 	dm_block_t old_count;
 
-	r = dm_sm_get_nr_blocks(pmd->data_sm, &old_count);
+	r = dm_sm_get_nr_blocks(sm, &old_count);
 	if (r)
 		return r;
 
@@ -1658,11 +1658,11 @@ static int __resize_data_dev(struct dm_pool_metadata *pmd, dm_block_t new_count)
 		return 0;
 
 	if (new_count < old_count) {
-		DMERR("cannot reduce size of data device");
+		DMERR("cannot reduce size of space map");
 		return -EINVAL;
 	}
 
-	return dm_sm_extend(pmd->data_sm, new_count - old_count);
+	return dm_sm_extend(sm, new_count - old_count);
 }
 
 int dm_pool_resize_data_dev(struct dm_pool_metadata *pmd, dm_block_t new_count)
@@ -1671,7 +1671,19 @@ int dm_pool_resize_data_dev(struct dm_pool_metadata *pmd, dm_block_t new_count)
 
 	down_write(&pmd->root_lock);
 	if (!pmd->fail_io)
-		r = __resize_data_dev(pmd, new_count);
+		r = __resize_space_map(pmd->data_sm, new_count);
+	up_write(&pmd->root_lock);
+
+	return r;
+}
+
+int dm_pool_resize_metadata_dev(struct dm_pool_metadata *pmd, dm_block_t new_count)
+{
+	int r = -EINVAL;
+
+	down_write(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = __resize_space_map(pmd->metadata_sm, new_count);
 	up_write(&pmd->root_lock);
 
 	return r;
@@ -1684,3 +1696,17 @@ void dm_pool_metadata_read_only(struct dm_pool_metadata *pmd)
 	dm_bm_set_read_only(pmd->bm);
 	up_write(&pmd->root_lock);
 }
+
+int dm_pool_register_metadata_threshold(struct dm_pool_metadata *pmd,
+					dm_block_t threshold,
+					dm_sm_threshold_fn fn,
+					void *context)
+{
+	int r;
+
+	down_write(&pmd->root_lock);
+	r = dm_sm_register_threshold_callback(pmd->metadata_sm, threshold, fn, context);
+	up_write(&pmd->root_lock);
+
+	return r;
+}

drivers/md/dm-thin-metadata.h

@@ -8,6 +8,7 @@
 #define DM_THIN_METADATA_H
 
 #include "persistent-data/dm-block-manager.h"
+#include "persistent-data/dm-space-map.h"
 
 #define THIN_METADATA_BLOCK_SIZE 4096
 
@@ -185,6 +186,7 @@ int dm_pool_get_data_dev_size(struct dm_pool_metadata *pmd, dm_block_t *result);
  * blocks would be lost.
  */
 int dm_pool_resize_data_dev(struct dm_pool_metadata *pmd, dm_block_t new_size);
+int dm_pool_resize_metadata_dev(struct dm_pool_metadata *pmd, dm_block_t new_size);
 
 /*
  * Flicks the underlying block manager into read only mode, so you know
@@ -192,6 +194,11 @@ int dm_pool_resize_data_dev(struct dm_pool_metadata *pmd, dm_block_t new_size);
  */
 void dm_pool_metadata_read_only(struct dm_pool_metadata *pmd);
 
+int dm_pool_register_metadata_threshold(struct dm_pool_metadata *pmd,
+					dm_block_t threshold,
+					dm_sm_threshold_fn fn,
+					void *context);
+
 /*----------------------------------------------------------------*/
 
 #endif

drivers/md/dm-thin.c

@@ -922,7 +922,7 @@ static int alloc_data_block(struct thin_c *tc, dm_block_t *result)
 		return r;
 
 	if (free_blocks <= pool->low_water_blocks && !pool->low_water_triggered) {
-		DMWARN("%s: reached low water mark, sending event.",
+		DMWARN("%s: reached low water mark for data device: sending event.",
 		       dm_device_name(pool->pool_md));
 		spin_lock_irqsave(&pool->lock, flags);
 		pool->low_water_triggered = 1;
@@ -1281,6 +1281,10 @@ static void process_bio_fail(struct thin_c *tc, struct bio *bio)
 	bio_io_error(bio);
 }
 
+/*
+ * FIXME: should we also commit due to size of transaction, measured in
+ * metadata blocks?
+ */
 static int need_commit_due_to_time(struct pool *pool)
 {
 	return jiffies < pool->last_commit_jiffies ||
@@ -1909,6 +1913,56 @@ static int parse_pool_features(struct dm_arg_set *as, struct pool_features *pf,
 	return r;
 }
 
+static void metadata_low_callback(void *context)
+{
+	struct pool *pool = context;
+
+	DMWARN("%s: reached low water mark for metadata device: sending event.",
+	       dm_device_name(pool->pool_md));
+
+	dm_table_event(pool->ti->table);
+}
+
+static sector_t get_metadata_dev_size(struct block_device *bdev)
+{
+	sector_t metadata_dev_size = i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
+	char buffer[BDEVNAME_SIZE];
+
+	if (metadata_dev_size > THIN_METADATA_MAX_SECTORS_WARNING) {
+		DMWARN("Metadata device %s is larger than %u sectors: excess space will not be used.",
+		       bdevname(bdev, buffer), THIN_METADATA_MAX_SECTORS);
+		metadata_dev_size = THIN_METADATA_MAX_SECTORS_WARNING;
+	}
+
+	return metadata_dev_size;
+}
+
+static dm_block_t get_metadata_dev_size_in_blocks(struct block_device *bdev)
+{
+	sector_t metadata_dev_size = get_metadata_dev_size(bdev);
+
+	sector_div(metadata_dev_size, THIN_METADATA_BLOCK_SIZE >> SECTOR_SHIFT);
+
+	return metadata_dev_size;
+}
+
+/*
+ * When a metadata threshold is crossed a dm event is triggered, and
+ * userland should respond by growing the metadata device.  We could let
+ * userland set the threshold, like we do with the data threshold, but I'm
+ * not sure they know enough to do this well.
+ */
+static dm_block_t calc_metadata_threshold(struct pool_c *pt)
+{
+	/*
+	 * 4M is ample for all ops with the possible exception of thin
+	 * device deletion which is harmless if it fails (just retry the
+	 * delete after you've grown the device).
+	 */
+	dm_block_t quarter = get_metadata_dev_size_in_blocks(pt->metadata_dev->bdev) / 4;
+	return min((dm_block_t)1024ULL /* 4M */, quarter);
+}
+
 /*
  * thin-pool <metadata dev> <data dev>
  *	     <data block size (sectors)>
@@ -1931,8 +1985,7 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv)
 	unsigned long block_size;
 	dm_block_t low_water_blocks;
 	struct dm_dev *metadata_dev;
-	sector_t metadata_dev_size;
-	char b[BDEVNAME_SIZE];
+	fmode_t metadata_mode;
 
 	/*
 	 * FIXME Remove validation from scope of lock.
@@ -1944,19 +1997,32 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv)
 		r = -EINVAL;
 		goto out_unlock;
 	}
+
 	as.argc = argc;
 	as.argv = argv;
 
-	r = dm_get_device(ti, argv[0], FMODE_READ | FMODE_WRITE, &metadata_dev);
+	/*
+	 * Set default pool features.
+	 */
+	pool_features_init(&pf);
+
+	dm_consume_args(&as, 4);
+	r = parse_pool_features(&as, &pf, ti);
+	if (r)
+		goto out_unlock;
+
+	metadata_mode = FMODE_READ | ((pf.mode == PM_READ_ONLY) ? 0 : FMODE_WRITE);
+	r = dm_get_device(ti, argv[0], metadata_mode, &metadata_dev);
 	if (r) {
 		ti->error = "Error opening metadata block device";
 		goto out_unlock;
 	}
 
-	metadata_dev_size = i_size_read(metadata_dev->bdev->bd_inode) >> SECTOR_SHIFT;
-	if (metadata_dev_size > THIN_METADATA_MAX_SECTORS_WARNING)
-		DMWARN("Metadata device %s is larger than %u sectors: excess space will not be used.",
-		       bdevname(metadata_dev->bdev, b), THIN_METADATA_MAX_SECTORS);
+	/*
+	 * Run for the side-effect of possibly issuing a warning if the
+	 * device is too big.
+	 */
+	(void) get_metadata_dev_size(metadata_dev->bdev);
 
 	r = dm_get_device(ti, argv[1], FMODE_READ | FMODE_WRITE, &data_dev);
 	if (r) {
@@ -1979,16 +2045,6 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv)
 		goto out;
 	}
 
-	/*
-	 * Set default pool features.
-	 */
-	pool_features_init(&pf);
-
-	dm_consume_args(&as, 4);
-	r = parse_pool_features(&as, &pf, ti);
-	if (r)
-		goto out;
-
 	pt = kzalloc(sizeof(*pt), GFP_KERNEL);
 	if (!pt) {
 		r = -ENOMEM;
@@ -2040,6 +2096,13 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv)
 	}
 	ti->private = pt;
 
+	r = dm_pool_register_metadata_threshold(pt->pool->pmd,
+						calc_metadata_threshold(pt),
+						metadata_low_callback,
+						pool);
+	if (r)
+		goto out_free_pt;
+
 	pt->callbacks.congested_fn = pool_is_congested;
 	dm_table_add_target_callbacks(ti->table, &pt->callbacks);
 
@@ -2079,18 +2142,7 @@ static int pool_map(struct dm_target *ti, struct bio *bio)
 	return r;
 }
 
-/*
- * Retrieves the number of blocks of the data device from
- * the superblock and compares it to the actual device size,
- * thus resizing the data device in case it has grown.
- *
- * This both copes with opening preallocated data devices in the ctr
- * being followed by a resume
- * -and-
- * calling the resume method individually after userspace has
- * grown the data device in reaction to a table event.
- */
-static int pool_preresume(struct dm_target *ti)
+static int maybe_resize_data_dev(struct dm_target *ti, bool *need_commit)
 {
 	int r;
 	struct pool_c *pt = ti->private;
@@ -2098,12 +2150,7 @@ static int pool_preresume(struct dm_target *ti)
 	sector_t data_size = ti->len;
 	dm_block_t sb_data_size;
 
-	/*
-	 * Take control of the pool object.
-	 */
-	r = bind_control_target(pool, ti);
-	if (r)
-		return r;
+	*need_commit = false;
 
 	(void) sector_div(data_size, pool->sectors_per_block);
 
@@ -2114,7 +2161,7 @@ static int pool_preresume(struct dm_target *ti)
 	}
 
 	if (data_size < sb_data_size) {
-		DMERR("pool target too small, is %llu blocks (expected %llu)",
+		DMERR("pool target (%llu blocks) too small: expected %llu",
 		      (unsigned long long)data_size, sb_data_size);
 		return -EINVAL;
 
@@ -2122,17 +2169,90 @@ static int pool_preresume(struct dm_target *ti)
 		r = dm_pool_resize_data_dev(pool->pmd, data_size);
 		if (r) {
 			DMERR("failed to resize data device");
-			/* FIXME Stricter than necessary: Rollback transaction instead here */
 			set_pool_mode(pool, PM_READ_ONLY);
 			return r;
 		}
 
-		(void) commit_or_fallback(pool);
+		*need_commit = true;
 	}
 
 	return 0;
 }
 
+static int maybe_resize_metadata_dev(struct dm_target *ti, bool *need_commit)
+{
+	int r;
+	struct pool_c *pt = ti->private;
+	struct pool *pool = pt->pool;
+	dm_block_t metadata_dev_size, sb_metadata_dev_size;
+
+	*need_commit = false;
+
+	metadata_dev_size = get_metadata_dev_size(pool->md_dev);
+
+	r = dm_pool_get_metadata_dev_size(pool->pmd, &sb_metadata_dev_size);
+	if (r) {
+		DMERR("failed to retrieve data device size");
+		return r;
+	}
+
+	if (metadata_dev_size < sb_metadata_dev_size) {
+		DMERR("metadata device (%llu sectors) too small: expected %llu",
+		      metadata_dev_size, sb_metadata_dev_size);
+		return -EINVAL;
+
+	} else if (metadata_dev_size > sb_metadata_dev_size) {
+		r = dm_pool_resize_metadata_dev(pool->pmd, metadata_dev_size);
+		if (r) {
+			DMERR("failed to resize metadata device");
+			return r;
+		}
+
+		*need_commit = true;
+	}
+
+	return 0;
+}
+
+/*
+ * Retrieves the number of blocks of the data device from
+ * the superblock and compares it to the actual device size,
+ * thus resizing the data device in case it has grown.
+ *
+ * This both copes with opening preallocated data devices in the ctr
+ * being followed by a resume
+ * -and-
+ * calling the resume method individually after userspace has
+ * grown the data device in reaction to a table event.
+ */
+static int pool_preresume(struct dm_target *ti)
+{
+	int r;
+	bool need_commit1, need_commit2;
+	struct pool_c *pt = ti->private;
+	struct pool *pool = pt->pool;
+
+	/*
+	 * Take control of the pool object.
+	 */
+	r = bind_control_target(pool, ti);
+	if (r)
+		return r;
+
+	r = maybe_resize_data_dev(ti, &need_commit1);
+	if (r)
+		return r;
+
+	r = maybe_resize_metadata_dev(ti, &need_commit2);
+	if (r)
+		return r;
+
+	if (need_commit1 || need_commit2)
+		(void) commit_or_fallback(pool);
+
+	return 0;
+}
+
 static void pool_resume(struct dm_target *ti)
 {
 	struct pool_c *pt = ti->private;
@@ -2549,7 +2669,7 @@ static struct target_type pool_target = {
 	.name = "thin-pool",
 	.features = DM_TARGET_SINGLETON | DM_TARGET_ALWAYS_WRITEABLE |
 		    DM_TARGET_IMMUTABLE,
-	.version = {1, 7, 0},
+	.version = {1, 8, 0},
 	.module = THIS_MODULE,
 	.ctr = pool_ctr,
 	.dtr = pool_dtr,

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

@@ -248,7 +248,8 @@ static struct dm_space_map ops = {
 	.new_block = sm_disk_new_block,
 	.commit = sm_disk_commit,
 	.root_size = sm_disk_root_size,
-	.copy_root = sm_disk_copy_root
+	.copy_root = sm_disk_copy_root,
+	.register_threshold_callback = NULL
 };
 
 struct dm_space_map *dm_sm_disk_create(struct dm_transaction_manager *tm,

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

@@ -16,6 +16,55 @@
 
 /*----------------------------------------------------------------*/
 
+/*
+ * An edge triggered threshold.
+ */
+struct threshold {
+	bool threshold_set;
+	bool value_set;
+	dm_block_t threshold;
+	dm_block_t current_value;
+	dm_sm_threshold_fn fn;
+	void *context;
+};
+
+static void threshold_init(struct threshold *t)
+{
+	t->threshold_set = false;
+	t->value_set = false;
+}
+
+static void set_threshold(struct threshold *t, dm_block_t value,
+			  dm_sm_threshold_fn fn, void *context)
+{
+	t->threshold_set = true;
+	t->threshold = value;
+	t->fn = fn;
+	t->context = context;
+}
+
+static bool below_threshold(struct threshold *t, dm_block_t value)
+{
+	return t->threshold_set && value <= t->threshold;
+}
+
+static bool threshold_already_triggered(struct threshold *t)
+{
+	return t->value_set && below_threshold(t, t->current_value);
+}
+
+static void check_threshold(struct threshold *t, dm_block_t value)
+{
+	if (below_threshold(t, value) &&
+	    !threshold_already_triggered(t))
+		t->fn(t->context);
+
+	t->value_set = true;
+	t->current_value = value;
+}
+
+/*----------------------------------------------------------------*/
+
 /*
  * Space map interface.
  *
@@ -54,6 +103,8 @@ struct sm_metadata {
 	unsigned allocated_this_transaction;
 	unsigned nr_uncommitted;
 	struct block_op uncommitted[MAX_RECURSIVE_ALLOCATIONS];
+
+	struct threshold threshold;
 };
 
 static int add_bop(struct sm_metadata *smm, enum block_op_type type, dm_block_t b)
@@ -144,12 +195,6 @@ static void sm_metadata_destroy(struct dm_space_map *sm)
 	kfree(smm);
 }
 
-static int sm_metadata_extend(struct dm_space_map *sm, dm_block_t extra_blocks)
-{
-	DMERR("doesn't support extend");
-	return -EINVAL;
-}
-
 static int sm_metadata_get_nr_blocks(struct dm_space_map *sm, dm_block_t *count)
 {
 	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
@@ -335,9 +380,19 @@ static int sm_metadata_new_block_(struct dm_space_map *sm, dm_block_t *b)
 
 static int sm_metadata_new_block(struct dm_space_map *sm, dm_block_t *b)
 {
+	dm_block_t count;
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
 	int r = sm_metadata_new_block_(sm, b);
 	if (r)
 		DMERR("unable to allocate new metadata block");
+
+	r = sm_metadata_get_nr_free(sm, &count);
+	if (r)
+		DMERR("couldn't get free block count");
+
+	check_threshold(&smm->threshold, count);
+
 	return r;
 }
 
@@ -357,6 +412,18 @@ static int sm_metadata_commit(struct dm_space_map *sm)
 	return 0;
 }
 
+static int sm_metadata_register_threshold_callback(struct dm_space_map *sm,
+						   dm_block_t threshold,
+						   dm_sm_threshold_fn fn,
+						   void *context)
+{
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	set_threshold(&smm->threshold, threshold, fn, context);
+
+	return 0;
+}
+
 static int sm_metadata_root_size(struct dm_space_map *sm, size_t *result)
 {
 	*result = sizeof(struct disk_sm_root);
@@ -382,6 +449,8 @@ static int sm_metadata_copy_root(struct dm_space_map *sm, void *where_le, size_t
 	return 0;
 }
 
+static int sm_metadata_extend(struct dm_space_map *sm, dm_block_t extra_blocks);
+
 static struct dm_space_map ops = {
 	.destroy = sm_metadata_destroy,
 	.extend = sm_metadata_extend,
@@ -395,7 +464,8 @@ static struct dm_space_map ops = {
 	.new_block = sm_metadata_new_block,
 	.commit = sm_metadata_commit,
 	.root_size = sm_metadata_root_size,
-	.copy_root = sm_metadata_copy_root
+	.copy_root = sm_metadata_copy_root,
+	.register_threshold_callback = sm_metadata_register_threshold_callback
 };
 
 /*----------------------------------------------------------------*/
@@ -410,7 +480,7 @@ static void sm_bootstrap_destroy(struct dm_space_map *sm)
 
 static int sm_bootstrap_extend(struct dm_space_map *sm, dm_block_t extra_blocks)
 {
-	DMERR("boostrap doesn't support extend");
+	DMERR("bootstrap doesn't support extend");
 
 	return -EINVAL;
 }
@@ -450,7 +520,7 @@ static int sm_bootstrap_count_is_more_than_one(struct dm_space_map *sm,
 static int sm_bootstrap_set_count(struct dm_space_map *sm, dm_block_t b,
 				  uint32_t count)
 {
-	DMERR("boostrap doesn't support set_count");
+	DMERR("bootstrap doesn't support set_count");
 
 	return -EINVAL;
 }
@@ -491,7 +561,7 @@ static int sm_bootstrap_commit(struct dm_space_map *sm)
 
 static int sm_bootstrap_root_size(struct dm_space_map *sm, size_t *result)
 {
-	DMERR("boostrap doesn't support root_size");
+	DMERR("bootstrap doesn't support root_size");
 
 	return -EINVAL;
 }
@@ -499,7 +569,7 @@ static int sm_bootstrap_root_size(struct dm_space_map *sm, size_t *result)
 static int sm_bootstrap_copy_root(struct dm_space_map *sm, void *where,
 				  size_t max)
 {
-	DMERR("boostrap doesn't support copy_root");
+	DMERR("bootstrap doesn't support copy_root");
 
 	return -EINVAL;
 }
@@ -517,11 +587,42 @@ static struct dm_space_map bootstrap_ops = {
 	.new_block = sm_bootstrap_new_block,
 	.commit = sm_bootstrap_commit,
 	.root_size = sm_bootstrap_root_size,
-	.copy_root = sm_bootstrap_copy_root
+	.copy_root = sm_bootstrap_copy_root,
+	.register_threshold_callback = NULL
 };
 
 /*----------------------------------------------------------------*/
 
+static int sm_metadata_extend(struct dm_space_map *sm, dm_block_t extra_blocks)
+{
+	int r, i;
+	enum allocation_event ev;
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+	dm_block_t old_len = smm->ll.nr_blocks;
+
+	/*
+	 * Flick into a mode where all blocks get allocated in the new area.
+	 */
+	smm->begin = old_len;
+	memcpy(&smm->sm, &bootstrap_ops, sizeof(smm->sm));
+
+	/*
+	 * Extend.
+	 */
+	r = sm_ll_extend(&smm->ll, extra_blocks);
+
+	/*
+	 * Switch back to normal behaviour.
+	 */
+	memcpy(&smm->sm, &ops, sizeof(smm->sm));
+	for (i = old_len; !r && i < smm->begin; i++)
+		r = sm_ll_inc(&smm->ll, i, &ev);
+
+	return r;
+}
+
+/*----------------------------------------------------------------*/
+
 struct dm_space_map *dm_sm_metadata_init(void)
 {
 	struct sm_metadata *smm;
@@ -549,6 +650,7 @@ int dm_sm_metadata_create(struct dm_space_map *sm,
 	smm->recursion_count = 0;
 	smm->allocated_this_transaction = 0;
 	smm->nr_uncommitted = 0;
+	threshold_init(&smm->threshold);
 
 	memcpy(&smm->sm, &bootstrap_ops, sizeof(smm->sm));
 
@@ -590,6 +692,7 @@ int dm_sm_metadata_open(struct dm_space_map *sm,
 	smm->recursion_count = 0;
 	smm->allocated_this_transaction = 0;
 	smm->nr_uncommitted = 0;
+	threshold_init(&smm->threshold);
 
 	memcpy(&smm->old_ll, &smm->ll, sizeof(smm->old_ll));
 	return 0;

drivers/md/persistent-data/dm-space-map.h

@@ -9,6 +9,8 @@
 
 #include "dm-block-manager.h"
 
+typedef void (*dm_sm_threshold_fn)(void *context);
+
 /*
  * struct dm_space_map keeps a record of how many times each block in a device
  * is referenced.  It needs to be fixed on disk as part of the transaction.
@@ -59,6 +61,15 @@ struct dm_space_map {
 	 */
 	int (*root_size)(struct dm_space_map *sm, size_t *result);
 	int (*copy_root)(struct dm_space_map *sm, void *copy_to_here_le, size_t len);
+
+	/*
+	 * You can register one threshold callback which is edge-triggered
+	 * when the free space in the space map drops below the threshold.
+	 */
+	int (*register_threshold_callback)(struct dm_space_map *sm,
+					   dm_block_t threshold,
+					   dm_sm_threshold_fn fn,
+					   void *context);
 };
 
 /*----------------------------------------------------------------*/
@@ -131,4 +142,16 @@ static inline int dm_sm_copy_root(struct dm_space_map *sm, void *copy_to_here_le
 	return sm->copy_root(sm, copy_to_here_le, len);
 }
 
+static inline int dm_sm_register_threshold_callback(struct dm_space_map *sm,
+						    dm_block_t threshold,
+						    dm_sm_threshold_fn fn,
+						    void *context)
+{
+	if (sm->register_threshold_callback)
+		return sm->register_threshold_callback(sm, threshold, fn, context);
+
+	return -EINVAL;
+}
+
+
 #endif	/* _LINUX_DM_SPACE_MAP_H */

include/linux/device-mapper.h

@@ -79,11 +79,26 @@ typedef int (*dm_ioctl_fn) (struct dm_target *ti, unsigned int cmd,
 typedef int (*dm_merge_fn) (struct dm_target *ti, struct bvec_merge_data *bvm,
 			    struct bio_vec *biovec, int max_size);
 
+/*
+ * These iteration functions are typically used to check (and combine)
+ * properties of underlying devices.
+ * E.g. Does at least one underlying device support flush?
+ *      Does any underlying device not support WRITE_SAME?
+ *
+ * The callout function is called once for each contiguous section of
+ * an underlying device.  State can be maintained in *data.
+ * Return non-zero to stop iterating through any further devices.
+ */
 typedef int (*iterate_devices_callout_fn) (struct dm_target *ti,
 					   struct dm_dev *dev,
 					   sector_t start, sector_t len,
 					   void *data);
 
+/*
+ * This function must iterate through each section of device used by the
+ * target until it encounters a non-zero return code, which it then returns.
+ * Returns zero if no callout returned non-zero.
+ */
 typedef int (*dm_iterate_devices_fn) (struct dm_target *ti,
 				      iterate_devices_callout_fn fn,
 				      void *data);