Merge tag 'dm-3.17-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm

Pull device mapper changes from Mike Snitzer:

- Allow the thin target to paired with any size external origin; also
allow thin snapshots to be larger than the external origin.

- Add support for quickly loading a repetitive pattern into the
dm-switch target.

- Use per-bio data in the dm-crypt target instead of always using a
mempool for each allocation. Required switching to kmalloc alignment
for the bio slab.

- Fix DM core to properly stack the QUEUE_FLAG_NO_SG_MERGE flag

- Fix the dm-cache and dm-thin targets' export of the minimum_io_size
to match the data block size -- this fixes an issue where mkfs.xfs
would improperly infer raid striping was in place on the underlying
storage.

- Small cleanups in dm-io, dm-mpath and dm-cache

* tag 'dm-3.17-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
dm table: propagate QUEUE_FLAG_NO_SG_MERGE
dm switch: efficiently support repetitive patterns
dm switch: factor out switch_region_table_read
dm cache: set minimum_io_size to cache's data block size
dm thin: set minimum_io_size to pool's data block size
dm crypt: use per-bio data
block: use kmalloc alignment for bio slab
dm table: make dm_table_supports_discards static
dm cache metadata: use dm-space-map-metadata.h defined size limits
dm cache: fail migrations in the do_worker error path
dm cache: simplify deferred set reference count increments
dm thin: relax external origin size constraints
dm thin: switch to an atomic_t for tracking pending new block preparations
dm mpath: eliminate pg_ready() wrapper
dm io: simplify dec_count and sync_io

Documentation/device-mapper/switch.txt

···
     The path number in the range 0 ... (<num_paths> - 1).
     Expressed in hexadecimal (WITHOUT any prefix like 0x).
 
+R<n>,<m>
+    This parameter allows repetitive patterns to be loaded quickly. <n> and <m>
+    are hexadecimal numbers. The last <n> mappings are repeated in the next <m>
+    slots.
+
 Status
 ======
 
···
 Set mappings for the first 7 entries to point to devices switch0, switch1,
 switch2, switch0, switch1, switch2, switch1:
     dmsetup message switch 0 set_region_mappings 0:0 :1 :2 :0 :1 :2 :1
+
+Set repetitive mapping. This command:
+    dmsetup message switch 0 set_region_mappings 1000:1 :2 R2,10
+is equivalent to:
+    dmsetup message switch 0 set_region_mappings 1000:1 :2 :1 :2 :1 :2 :1 :2 \
+	:1 :2 :1 :2 :1 :2 :1 :2 :1 :2
+

block/bio.c

···
 	bslab = &bio_slabs[entry];
 
 	snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry);
-	slab = kmem_cache_create(bslab->name, sz, 0, SLAB_HWCACHE_ALIGN, NULL);
+	slab = kmem_cache_create(bslab->name, sz, ARCH_KMALLOC_MINALIGN,
+				 SLAB_HWCACHE_ALIGN, NULL);
 	if (!slab)
 		goto out_unlock;
 

drivers/md/dm-cache-metadata.c

···
 	disk_super->discard_root = cpu_to_le64(cmd->discard_root);
 	disk_super->discard_block_size = cpu_to_le64(cmd->discard_block_size);
 	disk_super->discard_nr_blocks = cpu_to_le64(from_oblock(cmd->discard_nr_blocks));
-	disk_super->metadata_block_size = cpu_to_le32(DM_CACHE_METADATA_BLOCK_SIZE >> SECTOR_SHIFT);
+	disk_super->metadata_block_size = cpu_to_le32(DM_CACHE_METADATA_BLOCK_SIZE);
 	disk_super->data_block_size = cpu_to_le32(cmd->data_block_size);
 	disk_super->cache_blocks = cpu_to_le32(0);
 
···
 					    bool may_format_device)
 {
 	int r;
-	cmd->bm = dm_block_manager_create(cmd->bdev, DM_CACHE_METADATA_BLOCK_SIZE,
+	cmd->bm = dm_block_manager_create(cmd->bdev, DM_CACHE_METADATA_BLOCK_SIZE << SECTOR_SHIFT,
 					  CACHE_METADATA_CACHE_SIZE,
 					  CACHE_MAX_CONCURRENT_LOCKS);
 	if (IS_ERR(cmd->bm)) {

drivers/md/dm-cache-metadata.h

···
 
 #include "dm-cache-block-types.h"
 #include "dm-cache-policy-internal.h"
+#include "persistent-data/dm-space-map-metadata.h"
 
 /*----------------------------------------------------------------*/
 
-#define DM_CACHE_METADATA_BLOCK_SIZE 4096
+#define DM_CACHE_METADATA_BLOCK_SIZE DM_SM_METADATA_BLOCK_SIZE
 
 /* FIXME: remove this restriction */
 /*
  * The metadata device is currently limited in size.
- *
- * We have one block of index, which can hold 255 index entries.  Each
- * index entry contains allocation info about 16k metadata blocks.
  */
-#define DM_CACHE_METADATA_MAX_SECTORS (255 * (1 << 14) * (DM_CACHE_METADATA_BLOCK_SIZE / (1 << SECTOR_SHIFT)))
+#define DM_CACHE_METADATA_MAX_SECTORS DM_SM_METADATA_MAX_SECTORS
 
 /*
  * A metadata device larger than 16GB triggers a warning.

drivers/md/dm-cache-target.c

···
 	return bio->bi_rw & (REQ_FLUSH | REQ_FUA);
 }
 
+/*
+ * You must increment the deferred set whilst the prison cell is held.  To
+ * encourage this, we ask for 'cell' to be passed in.
+ */
+static void inc_ds(struct cache *cache, struct bio *bio,
+		   struct dm_bio_prison_cell *cell)
+{
+	size_t pb_data_size = get_per_bio_data_size(cache);
+	struct per_bio_data *pb = get_per_bio_data(bio, pb_data_size);
+
+	BUG_ON(!cell);
+	BUG_ON(pb->all_io_entry);
+
+	pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
+}
+
 static void issue(struct cache *cache, struct bio *bio)
 {
 	unsigned long flags;
···
 	cache->commit_requested = true;
 	bio_list_add(&cache->deferred_flush_bios, bio);
 	spin_unlock_irqrestore(&cache->lock, flags);
+}
+
+static void inc_and_issue(struct cache *cache, struct bio *bio, struct dm_bio_prison_cell *cell)
+{
+	inc_ds(cache, bio, cell);
+	issue(cache, bio);
 }
 
 static void defer_writethrough_bio(struct cache *cache, struct bio *bio)
···
 
 	dm_hook_bio(&pb->hook_info, bio, overwrite_endio, mg);
 	remap_to_cache_dirty(mg->cache, bio, mg->new_oblock, mg->cblock);
+
+	/*
+	 * No need to inc_ds() here, since the cell will be held for the
+	 * duration of the io.
+	 */
 	generic_make_request(bio);
 }
 
···
 		return;
 
 	INIT_LIST_HEAD(&work);
-	if (pb->all_io_entry)
-		dm_deferred_entry_dec(pb->all_io_entry, &work);
+	dm_deferred_entry_dec(pb->all_io_entry, &work);
 
 	if (!list_empty(&work))
 		queue_quiesced_migrations(cache, &work);
···
 	else
 		remap_to_cache(cache, bio, 0);
 
+	/*
+	 * REQ_FLUSH is not directed at any particular block so we don't
+	 * need to inc_ds().  REQ_FUA's are split into a write + REQ_FLUSH
+	 * by dm-core.
+	 */
 	issue(cache, bio);
 }
 
···
 		   &cache->stats.read_miss : &cache->stats.write_miss);
 }
 
-static void issue_cache_bio(struct cache *cache, struct bio *bio,
-			    struct per_bio_data *pb,
-			    dm_oblock_t oblock, dm_cblock_t cblock)
-{
-	pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
-	remap_to_cache_dirty(cache, bio, oblock, cblock);
-	issue(cache, bio);
-}
-
 static void process_bio(struct cache *cache, struct prealloc *structs,
 			struct bio *bio)
 {
···
 	dm_oblock_t block = get_bio_block(cache, bio);
 	struct dm_bio_prison_cell *cell_prealloc, *old_ocell, *new_ocell;
 	struct policy_result lookup_result;
-	size_t pb_data_size = get_per_bio_data_size(cache);
-	struct per_bio_data *pb = get_per_bio_data(bio, pb_data_size);
 	bool discarded_block = is_discarded_oblock(cache, block);
 	bool passthrough = passthrough_mode(&cache->features);
 	bool can_migrate = !passthrough && (discarded_block || spare_migration_bandwidth(cache));
···
 
 			} else {
 				/* FIXME: factor out issue_origin() */
-				pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
 				remap_to_origin_clear_discard(cache, bio, block);
-				issue(cache, bio);
+				inc_and_issue(cache, bio, new_ocell);
 			}
 		} else {
 			inc_hit_counter(cache, bio);
···
 			if (bio_data_dir(bio) == WRITE &&
 			    writethrough_mode(&cache->features) &&
 			    !is_dirty(cache, lookup_result.cblock)) {
-				pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
 				remap_to_origin_then_cache(cache, bio, block, lookup_result.cblock);
-				issue(cache, bio);
-			} else
-				issue_cache_bio(cache, bio, pb, block, lookup_result.cblock);
+				inc_and_issue(cache, bio, new_ocell);
+
+			} else  {
+				remap_to_cache_dirty(cache, bio, block, lookup_result.cblock);
+				inc_and_issue(cache, bio, new_ocell);
+			}
 		}
 
 		break;
 
 	case POLICY_MISS:
 		inc_miss_counter(cache, bio);
-		pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
 		remap_to_origin_clear_discard(cache, bio, block);
-		issue(cache, bio);
+		inc_and_issue(cache, bio, new_ocell);
 		break;
 
 	case POLICY_NEW:
···
 	bio_list_init(&cache->deferred_flush_bios);
 	spin_unlock_irqrestore(&cache->lock, flags);
 
+	/*
+	 * These bios have already been through inc_ds()
+	 */
 	while ((bio = bio_list_pop(&bios)))
 		submit_bios ? generic_make_request(bio) : bio_io_error(bio);
 }
···
 	bio_list_init(&cache->deferred_writethrough_bios);
 	spin_unlock_irqrestore(&cache->lock, flags);
 
+	/*
+	 * These bios have already been through inc_ds()
+	 */
 	while ((bio = bio_list_pop(&bios)))
 		generic_make_request(bio);
 }
···
 
 		if (commit_if_needed(cache)) {
 			process_deferred_flush_bios(cache, false);
+			process_migrations(cache, &cache->need_commit_migrations, migration_failure);
 
 			/*
 			 * FIXME: rollback metadata or just go into a
···
 	return r;
 }
 
-static int cache_map(struct dm_target *ti, struct bio *bio)
+static int __cache_map(struct cache *cache, struct bio *bio, struct dm_bio_prison_cell **cell)
 {
-	struct cache *cache = ti->private;
-
 	int r;
 	dm_oblock_t block = get_bio_block(cache, bio);
 	size_t pb_data_size = get_per_bio_data_size(cache);
 	bool can_migrate = false;
 	bool discarded_block;
-	struct dm_bio_prison_cell *cell;
 	struct policy_result lookup_result;
 	struct per_bio_data *pb = init_per_bio_data(bio, pb_data_size);
 
···
 	/*
 	 * Check to see if that block is currently migrating.
 	 */
-	cell = alloc_prison_cell(cache);
-	if (!cell) {
+	*cell = alloc_prison_cell(cache);
+	if (!*cell) {
 		defer_bio(cache, bio);
 		return DM_MAPIO_SUBMITTED;
 	}
 
-	r = bio_detain(cache, block, bio, cell,
+	r = bio_detain(cache, block, bio, *cell,
 		       (cell_free_fn) free_prison_cell,
-		       cache, &cell);
+		       cache, cell);
 	if (r) {
 		if (r < 0)
 			defer_bio(cache, bio);
···
 	r = policy_map(cache->policy, block, false, can_migrate, discarded_block,
 		       bio, &lookup_result);
 	if (r == -EWOULDBLOCK) {
-		cell_defer(cache, cell, true);
+		cell_defer(cache, *cell, true);
 		return DM_MAPIO_SUBMITTED;
 
 	} else if (r) {
 		DMERR_LIMIT("Unexpected return from cache replacement policy: %d", r);
+		cell_defer(cache, *cell, false);
 		bio_io_error(bio);
 		return DM_MAPIO_SUBMITTED;
 	}
···
 				 * We need to invalidate this block, so
 				 * defer for the worker thread.
 				 */
-				cell_defer(cache, cell, true);
+				cell_defer(cache, *cell, true);
 				r = DM_MAPIO_SUBMITTED;
 
 			} else {
-				pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
 				inc_miss_counter(cache, bio);
 				remap_to_origin_clear_discard(cache, bio, block);
-
-				cell_defer(cache, cell, false);
 			}
 
 		} else {
 			inc_hit_counter(cache, bio);
-			pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
-
 			if (bio_data_dir(bio) == WRITE && writethrough_mode(&cache->features) &&
 			    !is_dirty(cache, lookup_result.cblock))
 				remap_to_origin_then_cache(cache, bio, block, lookup_result.cblock);
 			else
 				remap_to_cache_dirty(cache, bio, block, lookup_result.cblock);
-
-			cell_defer(cache, cell, false);
 		}
 		break;
 
 	case POLICY_MISS:
 		inc_miss_counter(cache, bio);
-		pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
-
 		if (pb->req_nr != 0) {
 			/*
 			 * This is a duplicate writethrough io that is no
 			 * longer needed because the block has been demoted.
 			 */
 			bio_endio(bio, 0);
-			cell_defer(cache, cell, false);
-			return DM_MAPIO_SUBMITTED;
-		} else {
+			cell_defer(cache, *cell, false);
+			r = DM_MAPIO_SUBMITTED;
+
+		} else
 			remap_to_origin_clear_discard(cache, bio, block);
-			cell_defer(cache, cell, false);
-		}
+
 		break;
 
 	default:
 		DMERR_LIMIT("%s: erroring bio: unknown policy op: %u", __func__,
 			    (unsigned) lookup_result.op);
+		cell_defer(cache, *cell, false);
 		bio_io_error(bio);
 		r = DM_MAPIO_SUBMITTED;
+	}
+
+	return r;
+}
+
+static int cache_map(struct dm_target *ti, struct bio *bio)
+{
+	int r;
+	struct dm_bio_prison_cell *cell;
+	struct cache *cache = ti->private;
+
+	r = __cache_map(cache, bio, &cell);
+	if (r == DM_MAPIO_REMAPPED) {
+		inc_ds(cache, bio, cell);
+		cell_defer(cache, cell, false);
 	}
 
 	return r;
···
 		residency = policy_residency(cache->policy);
 
 		DMEMIT("%u %llu/%llu %u %llu/%llu %u %u %u %u %u %u %lu ",
-		       (unsigned)(DM_CACHE_METADATA_BLOCK_SIZE >> SECTOR_SHIFT),
+		       (unsigned)DM_CACHE_METADATA_BLOCK_SIZE,
 		       (unsigned long long)(nr_blocks_metadata - nr_free_blocks_metadata),
 		       (unsigned long long)nr_blocks_metadata,
 		       cache->sectors_per_block,
···
 	 */
 	if (io_opt_sectors < cache->sectors_per_block ||
 	    do_div(io_opt_sectors, cache->sectors_per_block)) {
-		blk_limits_io_min(limits, 0);
+		blk_limits_io_min(limits, cache->sectors_per_block << SECTOR_SHIFT);
 		blk_limits_io_opt(limits, cache->sectors_per_block << SECTOR_SHIFT);
 	}
 	set_discard_limits(cache, limits);
···
 
 static struct target_type cache_target = {
 	.name = "cache",
-	.version = {1, 4, 0},
+	.version = {1, 5, 0},
 	.module = THIS_MODULE,
 	.ctr = cache_ctr,
 	.dtr = cache_dtr,

drivers/md/dm-crypt.c

···
 	int error;
 	sector_t sector;
 	struct dm_crypt_io *base_io;
-};
+} CRYPTO_MINALIGN_ATTR;
 
 struct dm_crypt_request {
 	struct convert_context *ctx;
···
 	 * correctly aligned.
 	 */
 	unsigned int dmreq_start;
+
+	unsigned int per_bio_data_size;
 
 	unsigned long flags;
 	unsigned int key_size;
···
 	    kcryptd_async_done, dmreq_of_req(cc, ctx->req));
 }
 
+static void crypt_free_req(struct crypt_config *cc,
+			   struct ablkcipher_request *req, struct bio *base_bio)
+{
+	struct dm_crypt_io *io = dm_per_bio_data(base_bio, cc->per_bio_data_size);
+
+	if ((struct ablkcipher_request *)(io + 1) != req)
+		mempool_free(req, cc->req_pool);
+}
+
 /*
  * Encrypt / decrypt data from one bio to another one (can be the same one)
  */
···
 	}
 }
 
-static struct dm_crypt_io *crypt_io_alloc(struct crypt_config *cc,
-					  struct bio *bio, sector_t sector)
+static void crypt_io_init(struct dm_crypt_io *io, struct crypt_config *cc,
+			  struct bio *bio, sector_t sector)
 {
-	struct dm_crypt_io *io;
-
-	io = mempool_alloc(cc->io_pool, GFP_NOIO);
 	io->cc = cc;
 	io->base_bio = bio;
 	io->sector = sector;
···
 	io->base_io = NULL;
 	io->ctx.req = NULL;
 	atomic_set(&io->io_pending, 0);
-
-	return io;
 }
 
 static void crypt_inc_pending(struct dm_crypt_io *io)
···
 		return;
 
 	if (io->ctx.req)
-		mempool_free(io->ctx.req, cc->req_pool);
-	mempool_free(io, cc->io_pool);
+		crypt_free_req(cc, io->ctx.req, base_bio);
+	if (io != dm_per_bio_data(base_bio, cc->per_bio_data_size))
+		mempool_free(io, cc->io_pool);
 
 	if (likely(!base_io))
 		bio_endio(base_bio, error);
···
 		 * between fragments, so switch to a new dm_crypt_io structure.
 		 */
 		if (unlikely(!crypt_finished && remaining)) {
-			new_io = crypt_io_alloc(io->cc, io->base_bio,
-						sector);
+			new_io = mempool_alloc(cc->io_pool, GFP_NOIO);
+			crypt_io_init(new_io, io->cc, io->base_bio, sector);
 			crypt_inc_pending(new_io);
 			crypt_convert_init(cc, &new_io->ctx, NULL,
 					   io->base_bio, sector);
···
 	if (error < 0)
 		io->error = -EIO;
 
-	mempool_free(req_of_dmreq(cc, dmreq), cc->req_pool);
+	crypt_free_req(cc, req_of_dmreq(cc, dmreq), io->base_bio);
 
 	if (!atomic_dec_and_test(&ctx->cc_pending))
 		return;
···
 		goto bad;
 	}
 
+	cc->per_bio_data_size = ti->per_bio_data_size =
+				sizeof(struct dm_crypt_io) + cc->dmreq_start +
+				sizeof(struct dm_crypt_request) + cc->iv_size;
+
 	cc->page_pool = mempool_create_page_pool(MIN_POOL_PAGES, 0);
 	if (!cc->page_pool) {
 		ti->error = "Cannot allocate page mempool";
···
 		return DM_MAPIO_REMAPPED;
 	}
 
-	io = crypt_io_alloc(cc, bio, dm_target_offset(ti, bio->bi_iter.bi_sector));
+	io = dm_per_bio_data(bio, cc->per_bio_data_size);
+	crypt_io_init(io, cc, bio, dm_target_offset(ti, bio->bi_iter.bi_sector));
+	io->ctx.req = (struct ablkcipher_request *)(io + 1);
 
 	if (bio_data_dir(io->base_bio) == READ) {
 		if (kcryptd_io_read(io, GFP_NOWAIT))

drivers/md/dm-io.c

···
 struct io {
 	unsigned long error_bits;
 	atomic_t count;
-	struct completion *wait;
 	struct dm_io_client *client;
 	io_notify_fn callback;
 	void *context;
···
  * We need an io object to keep track of the number of bios that
  * have been dispatched for a particular io.
  *---------------------------------------------------------------*/
+static void complete_io(struct io *io)
+{
+	unsigned long error_bits = io->error_bits;
+	io_notify_fn fn = io->callback;
+	void *context = io->context;
+
+	if (io->vma_invalidate_size)
+		invalidate_kernel_vmap_range(io->vma_invalidate_address,
+					     io->vma_invalidate_size);
+
+	mempool_free(io, io->client->pool);
+	fn(error_bits, context);
+}
+
 static void dec_count(struct io *io, unsigned int region, int error)
 {
 	if (error)
 		set_bit(region, &io->error_bits);
 
-	if (atomic_dec_and_test(&io->count)) {
-		if (io->vma_invalidate_size)
-			invalidate_kernel_vmap_range(io->vma_invalidate_address,
-						     io->vma_invalidate_size);
-
-		if (io->wait)
-			complete(io->wait);
-
-		else {
-			unsigned long r = io->error_bits;
-			io_notify_fn fn = io->callback;
-			void *context = io->context;
-
-			mempool_free(io, io->client->pool);
-			fn(r, context);
-		}
-	}
+	if (atomic_dec_and_test(&io->count))
+		complete_io(io);
 }
 
 static void endio(struct bio *bio, int error)
···
 	dec_count(io, 0, 0);
 }
 
+struct sync_io {
+	unsigned long error_bits;
+	struct completion wait;
+};
+
+static void sync_io_complete(unsigned long error, void *context)
+{
+	struct sync_io *sio = context;
+
+	sio->error_bits = error;
+	complete(&sio->wait);
+}
+
 static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 		   struct dm_io_region *where, int rw, struct dpages *dp,
 		   unsigned long *error_bits)
 {
-	/*
-	 * gcc <= 4.3 can't do the alignment for stack variables, so we must
-	 * align it on our own.
-	 * volatile prevents the optimizer from removing or reusing
-	 * "io_" field from the stack frame (allowed in ANSI C).
-	 */
-	volatile char io_[sizeof(struct io) + __alignof__(struct io) - 1];
-	struct io *io = (struct io *)PTR_ALIGN(&io_, __alignof__(struct io));
-	DECLARE_COMPLETION_ONSTACK(wait);
+	struct io *io;
+	struct sync_io sio;
 
 	if (num_regions > 1 && (rw & RW_MASK) != WRITE) {
 		WARN_ON(1);
 		return -EIO;
 	}
 
+	init_completion(&sio.wait);
+
+	io = mempool_alloc(client->pool, GFP_NOIO);
 	io->error_bits = 0;
 	atomic_set(&io->count, 1); /* see dispatch_io() */
-	io->wait = &wait;
 	io->client = client;
+	io->callback = sync_io_complete;
+	io->context = &sio;
 
 	io->vma_invalidate_address = dp->vma_invalidate_address;
 	io->vma_invalidate_size = dp->vma_invalidate_size;
 
 	dispatch_io(rw, num_regions, where, dp, io, 1);
 
-	wait_for_completion_io(&wait);
+	wait_for_completion_io(&sio.wait);
 
 	if (error_bits)
-		*error_bits = io->error_bits;
+		*error_bits = sio.error_bits;
 
-	return io->error_bits ? -EIO : 0;
+	return sio.error_bits ? -EIO : 0;
 }
 
 static int async_io(struct dm_io_client *client, unsigned int num_regions,
···
 	io = mempool_alloc(client->pool, GFP_NOIO);
 	io->error_bits = 0;
 	atomic_set(&io->count, 1); /* see dispatch_io() */
-	io->wait = NULL;
 	io->client = client;
 	io->callback = fn;
 	io->context = context;
···
  * New collapsed (a)synchronous interface.
  *
  * If the IO is asynchronous (i.e. it has notify.fn), you must either unplug
- * the queue with blk_unplug() some time later or set REQ_SYNC in
-io_req->bi_rw. If you fail to do one of these, the IO will be submitted to
- * the disk after q->unplug_delay, which defaults to 3ms in blk-settings.c.
+ * the queue with blk_unplug() some time later or set REQ_SYNC in io_req->bi_rw.
+ * If you fail to do one of these, the IO will be submitted to the disk after
+ * q->unplug_delay, which defaults to 3ms in blk-settings.c.
  */
 int dm_io(struct dm_io_request *io_req, unsigned num_regions,
 	  struct dm_io_region *where, unsigned long *sync_error_bits)

drivers/md/dm-mpath.c

···
 		 dm_noflush_suspending(m->ti)));
 }
 
-#define pg_ready(m) (!(m)->queue_io && !(m)->pg_init_required)
-
 /*
  * Map cloned requests
  */
···
 		if (!__must_push_back(m))
 			r = -EIO;	/* Failed */
 		goto out_unlock;
-	}
-	if (!pg_ready(m)) {
+	} else if (m->queue_io || m->pg_init_required) {
 		__pg_init_all_paths(m);
 		goto out_unlock;
 	}
+
 	if (set_mapinfo(m, map_context) < 0)
 		/* ENOMEM, requeue */
 		goto out_unlock;

drivers/md/dm-switch.c

···
 	*bit *= sctx->region_table_entry_bits;
 }
 
+static unsigned switch_region_table_read(struct switch_ctx *sctx, unsigned long region_nr)
+{
+	unsigned long region_index;
+	unsigned bit;
+
+	switch_get_position(sctx, region_nr, &region_index, &bit);
+
+	return (ACCESS_ONCE(sctx->region_table[region_index]) >> bit) &
+		((1 << sctx->region_table_entry_bits) - 1);
+}
+
 /*
  * Find which path to use at given offset.
  */
 static unsigned switch_get_path_nr(struct switch_ctx *sctx, sector_t offset)
 {
-	unsigned long region_index;
-	unsigned bit, path_nr;
+	unsigned path_nr;
 	sector_t p;
 
 	p = offset;
···
 	else
 		sector_div(p, sctx->region_size);
 
-	switch_get_position(sctx, p, &region_index, &bit);
-	path_nr = (ACCESS_ONCE(sctx->region_table[region_index]) >> bit) &
-	       ((1 << sctx->region_table_entry_bits) - 1);
+	path_nr = switch_region_table_read(sctx, p);
 
 	/* This can only happen if the processor uses non-atomic stores. */
 	if (unlikely(path_nr >= sctx->nr_paths))
···
 }
 
 static int process_set_region_mappings(struct switch_ctx *sctx,
-			     unsigned argc, char **argv)
+				       unsigned argc, char **argv)
 {
 	unsigned i;
 	unsigned long region_index = 0;
···
 	for (i = 1; i < argc; i++) {
 		unsigned long path_nr;
 		const char *string = argv[i];
+
+		if ((*string & 0xdf) == 'R') {
+			unsigned long cycle_length, num_write;
+
+			string++;
+			if (unlikely(*string == ',')) {
+				DMWARN("invalid set_region_mappings argument: '%s'", argv[i]);
+				return -EINVAL;
+			}
+			cycle_length = parse_hex(&string);
+			if (unlikely(*string != ',')) {
+				DMWARN("invalid set_region_mappings argument: '%s'", argv[i]);
+				return -EINVAL;
+			}
+			string++;
+			if (unlikely(!*string)) {
+				DMWARN("invalid set_region_mappings argument: '%s'", argv[i]);
+				return -EINVAL;
+			}
+			num_write = parse_hex(&string);
+			if (unlikely(*string)) {
+				DMWARN("invalid set_region_mappings argument: '%s'", argv[i]);
+				return -EINVAL;
+			}
+
+			if (unlikely(!cycle_length) || unlikely(cycle_length - 1 > region_index)) {
+				DMWARN("invalid set_region_mappings cycle length: %lu > %lu",
+				       cycle_length - 1, region_index);
+				return -EINVAL;
+			}
+			if (unlikely(region_index + num_write < region_index) ||
+			    unlikely(region_index + num_write >= sctx->nr_regions)) {
+				DMWARN("invalid set_region_mappings region number: %lu + %lu >= %lu",
+				       region_index, num_write, sctx->nr_regions);
+				return -EINVAL;
+			}
+
+			while (num_write--) {
+				region_index++;
+				path_nr = switch_region_table_read(sctx, region_index - cycle_length);
+				switch_region_table_write(sctx, region_index, path_nr);
+			}
+
+			continue;
+		}
 
 		if (*string == ':')
 			region_index++;
···
 
 static struct target_type switch_target = {
 	.name = "switch",
-	.version = {1, 0, 0},
+	.version = {1, 1, 0},
 	.module = THIS_MODULE,
 	.ctr = switch_ctr,
 	.dtr = switch_dtr,

drivers/md/dm-table.c

···
 	return q && !blk_queue_add_random(q);
 }
 
+static int queue_supports_sg_merge(struct dm_target *ti, struct dm_dev *dev,
+				   sector_t start, sector_t len, void *data)
+{
+	struct request_queue *q = bdev_get_queue(dev->bdev);
+
+	return q && !test_bit(QUEUE_FLAG_NO_SG_MERGE, &q->queue_flags);
+}
+
 static bool dm_table_all_devices_attribute(struct dm_table *t,
 					   iterate_devices_callout_fn func)
 {
···
 	return true;
 }
 
+static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev,
+				  sector_t start, sector_t len, void *data)
+{
+	struct request_queue *q = bdev_get_queue(dev->bdev);
+
+	return q && blk_queue_discard(q);
+}
+
+static bool dm_table_supports_discards(struct dm_table *t)
+{
+	struct dm_target *ti;
+	unsigned i = 0;
+
+	/*
+	 * Unless any target used by the table set discards_supported,
+	 * require at least one underlying device to support discards.
+	 * t->devices includes internal dm devices such as mirror logs
+	 * so we need to use iterate_devices here, which targets
+	 * supporting discard selectively must provide.
+	 */
+	while (i < dm_table_get_num_targets(t)) {
+		ti = dm_table_get_target(t, i++);
+
+		if (!ti->num_discard_bios)
+			continue;
+
+		if (ti->discards_supported)
+			return 1;
+
+		if (ti->type->iterate_devices &&
+		    ti->type->iterate_devices(ti, device_discard_capable, NULL))
+			return 1;
+	}
+
+	return 0;
+}
+
 void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
 			       struct queue_limits *limits)
 {
···
 
 	if (!dm_table_supports_write_same(t))
 		q->limits.max_write_same_sectors = 0;
+
+	if (dm_table_all_devices_attribute(t, queue_supports_sg_merge))
+		queue_flag_clear_unlocked(QUEUE_FLAG_NO_SG_MERGE, q);
+	else
+		queue_flag_set_unlocked(QUEUE_FLAG_NO_SG_MERGE, q);
 
 	dm_table_set_integrity(t);
 
···
 }
 EXPORT_SYMBOL(dm_table_run_md_queue_async);
 
-static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev,
-				  sector_t start, sector_t len, void *data)
-{
-	struct request_queue *q = bdev_get_queue(dev->bdev);
-
-	return q && blk_queue_discard(q);
-}
-
-bool dm_table_supports_discards(struct dm_table *t)
-{
-	struct dm_target *ti;
-	unsigned i = 0;
-
-	/*
-	 * Unless any target used by the table set discards_supported,
-	 * require at least one underlying device to support discards.
-	 * t->devices includes internal dm devices such as mirror logs
-	 * so we need to use iterate_devices here, which targets
-	 * supporting discard selectively must provide.
-	 */
-	while (i < dm_table_get_num_targets(t)) {
-		ti = dm_table_get_target(t, i++);
-
-		if (!ti->num_discard_bios)
-			continue;
-
-		if (ti->discards_supported)
-			return 1;
-
-		if (ti->type->iterate_devices &&
-		    ti->type->iterate_devices(ti, device_discard_capable, NULL))
-			return 1;
-	}
-
-	return 0;
-}

drivers/md/dm-thin.c

···
 	struct list_head list;
 	struct dm_dev *pool_dev;
 	struct dm_dev *origin_dev;
+	sector_t origin_size;
 	dm_thin_id dev_id;
 
 	struct pool *pool;
···
 struct dm_thin_new_mapping {
 	struct list_head list;
 
-	bool quiesced:1;
-	bool prepared:1;
 	bool pass_discard:1;
 	bool definitely_not_shared:1;
+
+	/*
+	 * Track quiescing, copying and zeroing preparation actions.  When this
+	 * counter hits zero the block is prepared and can be inserted into the
+	 * btree.
+	 */
+	atomic_t prepare_actions;
 
 	int err;
 	struct thin_c *tc;
···
 	bio_end_io_t *saved_bi_end_io;
 };
 
-static void __maybe_add_mapping(struct dm_thin_new_mapping *m)
+static void __complete_mapping_preparation(struct dm_thin_new_mapping *m)
 {
 	struct pool *pool = m->tc->pool;
 
-	if (m->quiesced && m->prepared) {
+	if (atomic_dec_and_test(&m->prepare_actions)) {
 		list_add_tail(&m->list, &pool->prepared_mappings);
 		wake_worker(pool);
 	}
 }
 
-static void copy_complete(int read_err, unsigned long write_err, void *context)
+static void complete_mapping_preparation(struct dm_thin_new_mapping *m)
 {
 	unsigned long flags;
-	struct dm_thin_new_mapping *m = context;
 	struct pool *pool = m->tc->pool;
 
-	m->err = read_err || write_err ? -EIO : 0;
-
 	spin_lock_irqsave(&pool->lock, flags);
-	m->prepared = true;
-	__maybe_add_mapping(m);
+	__complete_mapping_preparation(m);
 	spin_unlock_irqrestore(&pool->lock, flags);
+}
+
+static void copy_complete(int read_err, unsigned long write_err, void *context)
+{
+	struct dm_thin_new_mapping *m = context;
+
+	m->err = read_err || write_err ? -EIO : 0;
+	complete_mapping_preparation(m);
 }
 
 static void overwrite_endio(struct bio *bio, int err)
 {
-	unsigned long flags;
 	struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
 	struct dm_thin_new_mapping *m = h->overwrite_mapping;
-	struct pool *pool = m->tc->pool;
 
 	m->err = err;
-
-	spin_lock_irqsave(&pool->lock, flags);
-	m->prepared = true;
-	__maybe_add_mapping(m);
-	spin_unlock_irqrestore(&pool->lock, flags);
+	complete_mapping_preparation(m);
 }
 
 /*----------------------------------------------------------------*/
···
 	return m;
 }
 
+static void ll_zero(struct thin_c *tc, struct dm_thin_new_mapping *m,
+		    sector_t begin, sector_t end)
+{
+	int r;
+	struct dm_io_region to;
+
+	to.bdev = tc->pool_dev->bdev;
+	to.sector = begin;
+	to.count = end - begin;
+
+	r = dm_kcopyd_zero(tc->pool->copier, 1, &to, 0, copy_complete, m);
+	if (r < 0) {
+		DMERR_LIMIT("dm_kcopyd_zero() failed");
+		copy_complete(1, 1, m);
+	}
+}
+
+/*
+ * A partial copy also needs to zero the uncopied region.
+ */
 static void schedule_copy(struct thin_c *tc, dm_block_t virt_block,
 			  struct dm_dev *origin, dm_block_t data_origin,
 			  dm_block_t data_dest,
-			  struct dm_bio_prison_cell *cell, struct bio *bio)
+			  struct dm_bio_prison_cell *cell, struct bio *bio,
+			  sector_t len)
 {
 	int r;
 	struct pool *pool = tc->pool;
···
 	m->data_block = data_dest;
 	m->cell = cell;
 
+	/*
+	 * quiesce action + copy action + an extra reference held for the
+	 * duration of this function (we may need to inc later for a
+	 * partial zero).
+	 */
+	atomic_set(&m->prepare_actions, 3);
+
 	if (!dm_deferred_set_add_work(pool->shared_read_ds, &m->list))
-		m->quiesced = true;
+		complete_mapping_preparation(m); /* already quiesced */
 
 	/*
 	 * IO to pool_dev remaps to the pool target's data_dev.
···
 
 		from.bdev = origin->bdev;
 		from.sector = data_origin * pool->sectors_per_block;
-		from.count = pool->sectors_per_block;
+		from.count = len;
 
 		to.bdev = tc->pool_dev->bdev;
 		to.sector = data_dest * pool->sectors_per_block;
-		to.count = pool->sectors_per_block;
+		to.count = len;
 
 		r = dm_kcopyd_copy(pool->copier, &from, 1, &to,
 				   0, copy_complete, m);
 		if (r < 0) {
-			mempool_free(m, pool->mapping_pool);
 			DMERR_LIMIT("dm_kcopyd_copy() failed");
-			cell_error(pool, cell);
+			copy_complete(1, 1, m);
+
+			/*
+			 * We allow the zero to be issued, to simplify the
+			 * error path.  Otherwise we'd need to start
+			 * worrying about decrementing the prepare_actions
+			 * counter.
+			 */
+		}
+
+		/*
+		 * Do we need to zero a tail region?
+		 */
+		if (len < pool->sectors_per_block && pool->pf.zero_new_blocks) {
+			atomic_inc(&m->prepare_actions);
+			ll_zero(tc, m,
+				data_dest * pool->sectors_per_block + len,
+				(data_dest + 1) * pool->sectors_per_block);
 		}
 	}
+
+	complete_mapping_preparation(m); /* drop our ref */
 }
 
 static void schedule_internal_copy(struct thin_c *tc, dm_block_t virt_block,
···
 				   struct dm_bio_prison_cell *cell, struct bio *bio)
 {
 	schedule_copy(tc, virt_block, tc->pool_dev,
-		      data_origin, data_dest, cell, bio);
-}
-
-static void schedule_external_copy(struct thin_c *tc, dm_block_t virt_block,
-				   dm_block_t data_dest,
-				   struct dm_bio_prison_cell *cell, struct bio *bio)
-{
-	schedule_copy(tc, virt_block, tc->origin_dev,
-		      virt_block, data_dest, cell, bio);
+		      data_origin, data_dest, cell, bio,
+		      tc->pool->sectors_per_block);
 }
 
 static void schedule_zero(struct thin_c *tc, dm_block_t virt_block,
···
 	struct pool *pool = tc->pool;
 	struct dm_thin_new_mapping *m = get_next_mapping(pool);
 
-	m->quiesced = true;
-	m->prepared = false;
+	atomic_set(&m->prepare_actions, 1); /* no need to quiesce */
 	m->tc = tc;
 	m->virt_block = virt_block;
 	m->data_block = data_block;
···
 		save_and_set_endio(bio, &m->saved_bi_end_io, overwrite_endio);
 		inc_all_io_entry(pool, bio);
 		remap_and_issue(tc, bio, data_block);
-	} else {
-		int r;
-		struct dm_io_region to;
 
-		to.bdev = tc->pool_dev->bdev;
-		to.sector = data_block * pool->sectors_per_block;
-		to.count = pool->sectors_per_block;
+	} else
+		ll_zero(tc, m,
+			data_block * pool->sectors_per_block,
+			(data_block + 1) * pool->sectors_per_block);
+}
 
-		r = dm_kcopyd_zero(pool->copier, 1, &to, 0, copy_complete, m);
-		if (r < 0) {
-			mempool_free(m, pool->mapping_pool);
-			DMERR_LIMIT("dm_kcopyd_zero() failed");
-			cell_error(pool, cell);
-		}
-	}
+static void schedule_external_copy(struct thin_c *tc, dm_block_t virt_block,
+				   dm_block_t data_dest,
+				   struct dm_bio_prison_cell *cell, struct bio *bio)
+{
+	struct pool *pool = tc->pool;
+	sector_t virt_block_begin = virt_block * pool->sectors_per_block;
+	sector_t virt_block_end = (virt_block + 1) * pool->sectors_per_block;
+
+	if (virt_block_end <= tc->origin_size)
+		schedule_copy(tc, virt_block, tc->origin_dev,
+			      virt_block, data_dest, cell, bio,
+			      pool->sectors_per_block);
+
+	else if (virt_block_begin < tc->origin_size)
+		schedule_copy(tc, virt_block, tc->origin_dev,
+			      virt_block, data_dest, cell, bio,
+			      tc->origin_size - virt_block_begin);
+
+	else
+		schedule_zero(tc, virt_block, data_dest, cell, bio);
 }
 
 /*
···
 			inc_all_io_entry(pool, bio);
 			cell_defer_no_holder(tc, cell);
 
-			remap_to_origin_and_issue(tc, bio);
+			if (bio_end_sector(bio) <= tc->origin_size)
+				remap_to_origin_and_issue(tc, bio);
+
+			else if (bio->bi_iter.bi_sector < tc->origin_size) {
+				zero_fill_bio(bio);
+				bio->bi_iter.bi_size = (tc->origin_size - bio->bi_iter.bi_sector) << SECTOR_SHIFT;
+				remap_to_origin_and_issue(tc, bio);
+
+			} else {
+				zero_fill_bio(bio);
+				bio_endio(bio, 0);
+			}
 		} else
 			provision_block(tc, bio, block, cell);
 		break;
···
 	 */
 	if (io_opt_sectors < pool->sectors_per_block ||
 	    do_div(io_opt_sectors, pool->sectors_per_block)) {
-		blk_limits_io_min(limits, 0);
+		blk_limits_io_min(limits, pool->sectors_per_block << SECTOR_SHIFT);
 		blk_limits_io_opt(limits, pool->sectors_per_block << SECTOR_SHIFT);
 	}
 
···
 	.name = "thin-pool",
 	.features = DM_TARGET_SINGLETON | DM_TARGET_ALWAYS_WRITEABLE |
 		    DM_TARGET_IMMUTABLE,
-	.version = {1, 12, 0},
+	.version = {1, 13, 0},
 	.module = THIS_MODULE,
 	.ctr = pool_ctr,
 	.dtr = pool_dtr,
···
 		spin_lock_irqsave(&pool->lock, flags);
 		list_for_each_entry_safe(m, tmp, &work, list) {
 			list_del(&m->list);
-			m->quiesced = true;
-			__maybe_add_mapping(m);
+			__complete_mapping_preparation(m);
 		}
 		spin_unlock_irqrestore(&pool->lock, flags);
 	}
···
 	 * unfortunately we must always run this.
 	 */
 	noflush_work(tc, do_noflush_stop);
+}
+
+static int thin_preresume(struct dm_target *ti)
+{
+	struct thin_c *tc = ti->private;
+
+	if (tc->origin_dev)
+		tc->origin_size = get_dev_size(tc->origin_dev->bdev);
+
+	return 0;
 }
 
 /*
···
 
 static struct target_type thin_target = {
 	.name = "thin",
-	.version = {1, 12, 0},
+	.version = {1, 13, 0},
 	.module	= THIS_MODULE,
 	.ctr = thin_ctr,
 	.dtr = thin_dtr,
 	.map = thin_map,
 	.end_io = thin_endio,
+	.preresume = thin_preresume,
 	.presuspend = thin_presuspend,
 	.postsuspend = thin_postsuspend,
 	.status = thin_status,

drivers/md/dm.h

···
 unsigned dm_table_get_type(struct dm_table *t);
 struct target_type *dm_table_get_immutable_target_type(struct dm_table *t);
 bool dm_table_request_based(struct dm_table *t);
-bool dm_table_supports_discards(struct dm_table *t);
 void dm_table_free_md_mempools(struct dm_table *t);
 struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t);