Merge git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm

* git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm: (36 commits)
  dm: set queue ordered mode
  dm: move wait queue declaration
  dm: merge pushback and deferred bio lists
  dm: allow uninterruptible wait for pending io
  dm: merge __flush_deferred_io into caller
  dm: move bio_io_error into __split_and_process_bio
  dm: rename __split_bio
  dm: remove unnecessary struct dm_wq_req
  dm: remove unnecessary work queue context field
  dm: remove unnecessary work queue type field
  dm: bio list add bio_list_add_head
  dm snapshot: persistent fix dtr cleanup
  dm snapshot: move status to exception store
  dm snapshot: move ctr parsing to exception store
  dm snapshot: use DMEMIT macro for status
  dm snapshot: remove dm_snap header
  dm snapshot: remove dm_snap header use
  dm exception store: move cow pointer
  dm exception store: move chunk_fields
  dm exception store: move dm_target pointer
  ...

drivers/md/dm-bio-list.h

@@ -52,6 +52,16 @@ static inline void bio_list_add(struct bio_list *bl, struct bio *bio)
 	bl->tail = bio;
 }
 
+static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio)
+{
+	bio->bi_next = bl->head;
+
+	bl->head = bio;
+
+	if (!bl->tail)
+		bl->tail = bio;
+}
+
 static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2)
 {
 	if (!bl2->head)

drivers/md/dm-bio-record.h

@@ -16,30 +16,56 @@
  * functions in this file help the target record and restore the
  * original bio state.
  */
+
+struct dm_bio_vec_details {
+#if PAGE_SIZE < 65536
+	__u16 bv_len;
+	__u16 bv_offset;
+#else
+	unsigned bv_len;
+	unsigned bv_offset;
+#endif
+};
+
 struct dm_bio_details {
 	sector_t bi_sector;
 	struct block_device *bi_bdev;
 	unsigned int bi_size;
 	unsigned short bi_idx;
 	unsigned long bi_flags;
+	struct dm_bio_vec_details bi_io_vec[BIO_MAX_PAGES];
 };
 
 static inline void dm_bio_record(struct dm_bio_details *bd, struct bio *bio)
 {
+	unsigned i;
+
 	bd->bi_sector = bio->bi_sector;
 	bd->bi_bdev = bio->bi_bdev;
 	bd->bi_size = bio->bi_size;
 	bd->bi_idx = bio->bi_idx;
 	bd->bi_flags = bio->bi_flags;
+
+	for (i = 0; i < bio->bi_vcnt; i++) {
+		bd->bi_io_vec[i].bv_len = bio->bi_io_vec[i].bv_len;
+		bd->bi_io_vec[i].bv_offset = bio->bi_io_vec[i].bv_offset;
+	}
 }
 
 static inline void dm_bio_restore(struct dm_bio_details *bd, struct bio *bio)
 {
+	unsigned i;
+
 	bio->bi_sector = bd->bi_sector;
 	bio->bi_bdev = bd->bi_bdev;
 	bio->bi_size = bd->bi_size;
 	bio->bi_idx = bd->bi_idx;
 	bio->bi_flags = bd->bi_flags;
+
+	for (i = 0; i < bio->bi_vcnt; i++) {
+		bio->bi_io_vec[i].bv_len = bd->bi_io_vec[i].bv_len;
+		bio->bi_io_vec[i].bv_offset = bd->bi_io_vec[i].bv_offset;
+	}
 }
 
 #endif

drivers/md/dm-crypt.c

@@ -1156,8 +1156,7 @@ bad_ivmode:
 	crypto_free_ablkcipher(tfm);
 bad_cipher:
 	/* Must zero key material before freeing */
-	memset(cc, 0, sizeof(*cc) + cc->key_size * sizeof(u8));
-	kfree(cc);
+	kzfree(cc);
 	return -EINVAL;
 }
 
@@ -1183,8 +1182,7 @@ static void crypt_dtr(struct dm_target *ti)
 	dm_put_device(ti, cc->dev);
 
 	/* Must zero key material before freeing */
-	memset(cc, 0, sizeof(*cc) + cc->key_size * sizeof(u8));
-	kfree(cc);
+	kzfree(cc);
 }
 
 static int crypt_map(struct dm_target *ti, struct bio *bio,

drivers/md/dm-exception-store.c

@@ -7,6 +7,7 @@
 
 #include "dm-exception-store.h"
 
+#include <linux/ctype.h>
 #include <linux/mm.h>
 #include <linux/pagemap.h>
 #include <linux/vmalloc.h>
@@ -14,6 +15,257 @@
 
 #define DM_MSG_PREFIX "snapshot exception stores"
 
+static LIST_HEAD(_exception_store_types);
+static DEFINE_SPINLOCK(_lock);
+
+static struct dm_exception_store_type *__find_exception_store_type(const char *name)
+{
+	struct dm_exception_store_type *type;
+
+	list_for_each_entry(type, &_exception_store_types, list)
+		if (!strcmp(name, type->name))
+			return type;
+
+	return NULL;
+}
+
+static struct dm_exception_store_type *_get_exception_store_type(const char *name)
+{
+	struct dm_exception_store_type *type;
+
+	spin_lock(&_lock);
+
+	type = __find_exception_store_type(name);
+
+	if (type && !try_module_get(type->module))
+		type = NULL;
+
+	spin_unlock(&_lock);
+
+	return type;
+}
+
+/*
+ * get_type
+ * @type_name
+ *
+ * Attempt to retrieve the dm_exception_store_type by name.  If not already
+ * available, attempt to load the appropriate module.
+ *
+ * Exstore modules are named "dm-exstore-" followed by the 'type_name'.
+ * Modules may contain multiple types.
+ * This function will first try the module "dm-exstore-<type_name>",
+ * then truncate 'type_name' on the last '-' and try again.
+ *
+ * For example, if type_name was "clustered-shared", it would search
+ * 'dm-exstore-clustered-shared' then 'dm-exstore-clustered'.
+ *
+ * 'dm-exception-store-<type_name>' is too long of a name in my
+ * opinion, which is why I've chosen to have the files
+ * containing exception store implementations be 'dm-exstore-<type_name>'.
+ * If you want your module to be autoloaded, you will follow this
+ * naming convention.
+ *
+ * Returns: dm_exception_store_type* on success, NULL on failure
+ */
+static struct dm_exception_store_type *get_type(const char *type_name)
+{
+	char *p, *type_name_dup;
+	struct dm_exception_store_type *type;
+
+	type = _get_exception_store_type(type_name);
+	if (type)
+		return type;
+
+	type_name_dup = kstrdup(type_name, GFP_KERNEL);
+	if (!type_name_dup) {
+		DMERR("No memory left to attempt load for \"%s\"", type_name);
+		return NULL;
+	}
+
+	while (request_module("dm-exstore-%s", type_name_dup) ||
+	       !(type = _get_exception_store_type(type_name))) {
+		p = strrchr(type_name_dup, '-');
+		if (!p)
+			break;
+		p[0] = '\0';
+	}
+
+	if (!type)
+		DMWARN("Module for exstore type \"%s\" not found.", type_name);
+
+	kfree(type_name_dup);
+
+	return type;
+}
+
+static void put_type(struct dm_exception_store_type *type)
+{
+	spin_lock(&_lock);
+	module_put(type->module);
+	spin_unlock(&_lock);
+}
+
+int dm_exception_store_type_register(struct dm_exception_store_type *type)
+{
+	int r = 0;
+
+	spin_lock(&_lock);
+	if (!__find_exception_store_type(type->name))
+		list_add(&type->list, &_exception_store_types);
+	else
+		r = -EEXIST;
+	spin_unlock(&_lock);
+
+	return r;
+}
+EXPORT_SYMBOL(dm_exception_store_type_register);
+
+int dm_exception_store_type_unregister(struct dm_exception_store_type *type)
+{
+	spin_lock(&_lock);
+
+	if (!__find_exception_store_type(type->name)) {
+		spin_unlock(&_lock);
+		return -EINVAL;
+	}
+
+	list_del(&type->list);
+
+	spin_unlock(&_lock);
+
+	return 0;
+}
+EXPORT_SYMBOL(dm_exception_store_type_unregister);
+
+/*
+ * Round a number up to the nearest 'size' boundary.  size must
+ * be a power of 2.
+ */
+static ulong round_up(ulong n, ulong size)
+{
+	size--;
+	return (n + size) & ~size;
+}
+
+static int set_chunk_size(struct dm_exception_store *store,
+			  const char *chunk_size_arg, char **error)
+{
+	unsigned long chunk_size_ulong;
+	char *value;
+
+	chunk_size_ulong = simple_strtoul(chunk_size_arg, &value, 10);
+	if (*chunk_size_arg == '\0' || *value != '\0') {
+		*error = "Invalid chunk size";
+		return -EINVAL;
+	}
+
+	if (!chunk_size_ulong) {
+		store->chunk_size = store->chunk_mask = store->chunk_shift = 0;
+		return 0;
+	}
+
+	/*
+	 * Chunk size must be multiple of page size.  Silently
+	 * round up if it's not.
+	 */
+	chunk_size_ulong = round_up(chunk_size_ulong, PAGE_SIZE >> 9);
+
+	/* Check chunk_size is a power of 2 */
+	if (!is_power_of_2(chunk_size_ulong)) {
+		*error = "Chunk size is not a power of 2";
+		return -EINVAL;
+	}
+
+	/* Validate the chunk size against the device block size */
+	if (chunk_size_ulong % (bdev_hardsect_size(store->cow->bdev) >> 9)) {
+		*error = "Chunk size is not a multiple of device blocksize";
+		return -EINVAL;
+	}
+
+	store->chunk_size = chunk_size_ulong;
+	store->chunk_mask = chunk_size_ulong - 1;
+	store->chunk_shift = ffs(chunk_size_ulong) - 1;
+
+	return 0;
+}
+
+int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
+			      unsigned *args_used,
+			      struct dm_exception_store **store)
+{
+	int r = 0;
+	struct dm_exception_store_type *type;
+	struct dm_exception_store *tmp_store;
+	char persistent;
+
+	if (argc < 3) {
+		ti->error = "Insufficient exception store arguments";
+		return -EINVAL;
+	}
+
+	tmp_store = kmalloc(sizeof(*tmp_store), GFP_KERNEL);
+	if (!tmp_store) {
+		ti->error = "Exception store allocation failed";
+		return -ENOMEM;
+	}
+
+	persistent = toupper(*argv[1]);
+	if (persistent != 'P' && persistent != 'N') {
+		ti->error = "Persistent flag is not P or N";
+		return -EINVAL;
+	}
+
+	type = get_type(argv[1]);
+	if (!type) {
+		ti->error = "Exception store type not recognised";
+		r = -EINVAL;
+		goto bad_type;
+	}
+
+	tmp_store->type = type;
+	tmp_store->ti = ti;
+
+	r = dm_get_device(ti, argv[0], 0, 0,
+			  FMODE_READ | FMODE_WRITE, &tmp_store->cow);
+	if (r) {
+		ti->error = "Cannot get COW device";
+		goto bad_cow;
+	}
+
+	r = set_chunk_size(tmp_store, argv[2], &ti->error);
+	if (r)
+		goto bad_cow;
+
+	r = type->ctr(tmp_store, 0, NULL);
+	if (r) {
+		ti->error = "Exception store type constructor failed";
+		goto bad_ctr;
+	}
+
+	*args_used = 3;
+	*store = tmp_store;
+	return 0;
+
+bad_ctr:
+	dm_put_device(ti, tmp_store->cow);
+bad_cow:
+	put_type(type);
+bad_type:
+	kfree(tmp_store);
+	return r;
+}
+EXPORT_SYMBOL(dm_exception_store_create);
+
+void dm_exception_store_destroy(struct dm_exception_store *store)
+{
+	store->type->dtr(store);
+	dm_put_device(store->ti, store->cow);
+	put_type(store->type);
+	kfree(store);
+}
+EXPORT_SYMBOL(dm_exception_store_destroy);
+
 int dm_exception_store_init(void)
 {
 	int r;

drivers/md/dm-exception-store.h

@@ -37,11 +37,18 @@ struct dm_snap_exception {
  * Abstraction to handle the meta/layout of exception stores (the
  * COW device).
  */
-struct dm_exception_store {
+struct dm_exception_store;
+struct dm_exception_store_type {
+	const char *name;
+	struct module *module;
+
+	int (*ctr) (struct dm_exception_store *store,
+		    unsigned argc, char **argv);
+
 	/*
 	 * Destroys this object when you've finished with it.
 	 */
-	void (*destroy) (struct dm_exception_store *store);
+	void (*dtr) (struct dm_exception_store *store);
 
 	/*
 	 * The target shouldn't read the COW device until this is
@@ -72,8 +79,9 @@ struct dm_exception_store {
 	 */
 	void (*drop_snapshot) (struct dm_exception_store *store);
 
-	int (*status) (struct dm_exception_store *store, status_type_t status,
-		       char *result, unsigned int maxlen);
+	unsigned (*status) (struct dm_exception_store *store,
+			    status_type_t status, char *result,
+			    unsigned maxlen);
 
 	/*
 	 * Return how full the snapshot is.
@@ -82,7 +90,21 @@ struct dm_exception_store {
 			       sector_t *numerator,
 			       sector_t *denominator);
 
-	struct dm_snapshot *snap;
+	/* For internal device-mapper use only. */
+	struct list_head list;
+};
+
+struct dm_exception_store {
+	struct dm_exception_store_type *type;
+	struct dm_target *ti;
+
+	struct dm_dev *cow;
+
+	/* Size of data blocks saved - must be a power of 2 */
+	chunk_t chunk_size;
+	chunk_t chunk_mask;
+	chunk_t chunk_shift;
+
 	void *context;
 };
 
@@ -129,6 +151,28 @@ static inline void dm_consecutive_chunk_count_inc(struct dm_snap_exception *e)
 
 #  endif
 
+/*
+ * Return the number of sectors in the device.
+ */
+static inline sector_t get_dev_size(struct block_device *bdev)
+{
+	return bdev->bd_inode->i_size >> SECTOR_SHIFT;
+}
+
+static inline chunk_t sector_to_chunk(struct dm_exception_store *store,
+				      sector_t sector)
+{
+	return (sector & ~store->chunk_mask) >> store->chunk_shift;
+}
+
+int dm_exception_store_type_register(struct dm_exception_store_type *type);
+int dm_exception_store_type_unregister(struct dm_exception_store_type *type);
+
+int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
+			      unsigned *args_used,
+			      struct dm_exception_store **store);
+void dm_exception_store_destroy(struct dm_exception_store *store);
+
 int dm_exception_store_init(void);
 void dm_exception_store_exit(void);
 
@@ -141,8 +185,4 @@ void dm_persistent_snapshot_exit(void);
 int dm_transient_snapshot_init(void);
 void dm_transient_snapshot_exit(void);
 
-int dm_create_persistent(struct dm_exception_store *store);
-
-int dm_create_transient(struct dm_exception_store *store);
-
 #endif /* _LINUX_DM_EXCEPTION_STORE */

drivers/md/dm-io.c

@@ -370,16 +370,13 @@ static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 	while (1) {
 		set_current_state(TASK_UNINTERRUPTIBLE);
 
-		if (!atomic_read(&io.count) || signal_pending(current))
+		if (!atomic_read(&io.count))
 			break;
 
 		io_schedule();
 	}
 	set_current_state(TASK_RUNNING);
 
-	if (atomic_read(&io.count))
-		return -EINTR;
-
 	if (error_bits)
 		*error_bits = io.error_bits;
 

drivers/md/dm-log.c

@@ -16,40 +16,29 @@
 
 #define DM_MSG_PREFIX "dirty region log"
 
-struct dm_dirty_log_internal {
-	struct dm_dirty_log_type *type;
-
-	struct list_head list;
-	long use;
-};
-
 static LIST_HEAD(_log_types);
 static DEFINE_SPINLOCK(_lock);
 
-static struct dm_dirty_log_internal *__find_dirty_log_type(const char *name)
+static struct dm_dirty_log_type *__find_dirty_log_type(const char *name)
 {
-	struct dm_dirty_log_internal *log_type;
+	struct dm_dirty_log_type *log_type;
 
 	list_for_each_entry(log_type, &_log_types, list)
-		if (!strcmp(name, log_type->type->name))
+		if (!strcmp(name, log_type->name))
 			return log_type;
 
 	return NULL;
 }
 
-static struct dm_dirty_log_internal *_get_dirty_log_type(const char *name)
+static struct dm_dirty_log_type *_get_dirty_log_type(const char *name)
 {
-	struct dm_dirty_log_internal *log_type;
+	struct dm_dirty_log_type *log_type;
 
 	spin_lock(&_lock);
 
 	log_type = __find_dirty_log_type(name);
-	if (log_type) {
-		if (!log_type->use && !try_module_get(log_type->type->module))
-			log_type = NULL;
-		else
-			log_type->use++;
-	}
+	if (log_type && !try_module_get(log_type->module))
+		log_type = NULL;
 
 	spin_unlock(&_lock);
 
@@ -76,14 +65,14 @@ static struct dm_dirty_log_internal *_get_dirty_log_type(const char *name)
 static struct dm_dirty_log_type *get_type(const char *type_name)
 {
 	char *p, *type_name_dup;
-	struct dm_dirty_log_internal *log_type;
+	struct dm_dirty_log_type *log_type;
 
 	if (!type_name)
 		return NULL;
 
 	log_type = _get_dirty_log_type(type_name);
 	if (log_type)
-		return log_type->type;
+		return log_type;
 
 	type_name_dup = kstrdup(type_name, GFP_KERNEL);
 	if (!type_name_dup) {
@@ -105,56 +94,33 @@ static struct dm_dirty_log_type *get_type(const char *type_name)
 
 	kfree(type_name_dup);
 
-	return log_type ? log_type->type : NULL;
+	return log_type;
 }
 
 static void put_type(struct dm_dirty_log_type *type)
 {
-	struct dm_dirty_log_internal *log_type;
-
 	if (!type)
 		return;
 
 	spin_lock(&_lock);
-	log_type = __find_dirty_log_type(type->name);
-	if (!log_type)
+	if (!__find_dirty_log_type(type->name))
 		goto out;
 
-	if (!--log_type->use)
-		module_put(type->module);
-
-	BUG_ON(log_type->use < 0);
+	module_put(type->module);
 
 out:
 	spin_unlock(&_lock);
 }
 
-static struct dm_dirty_log_internal *_alloc_dirty_log_type(struct dm_dirty_log_type *type)
-{
-	struct dm_dirty_log_internal *log_type = kzalloc(sizeof(*log_type),
-							 GFP_KERNEL);
-
-	if (log_type)
-		log_type->type = type;
-
-	return log_type;
-}
-
 int dm_dirty_log_type_register(struct dm_dirty_log_type *type)
 {
-	struct dm_dirty_log_internal *log_type = _alloc_dirty_log_type(type);
 	int r = 0;
 
-	if (!log_type)
-		return -ENOMEM;
-
 	spin_lock(&_lock);
 	if (!__find_dirty_log_type(type->name))
-		list_add(&log_type->list, &_log_types);
-	else {
-		kfree(log_type);
+		list_add(&type->list, &_log_types);
+	else
 		r = -EEXIST;
-	}
 	spin_unlock(&_lock);
 
 	return r;
@@ -163,25 +129,16 @@ EXPORT_SYMBOL(dm_dirty_log_type_register);
 
 int dm_dirty_log_type_unregister(struct dm_dirty_log_type *type)
 {
-	struct dm_dirty_log_internal *log_type;
-
 	spin_lock(&_lock);
 
-	log_type = __find_dirty_log_type(type->name);
-	if (!log_type) {
+	if (!__find_dirty_log_type(type->name)) {
 		spin_unlock(&_lock);
 		return -EINVAL;
 	}
 
-	if (log_type->use) {
-		spin_unlock(&_lock);
-		return -ETXTBSY;
-	}
-
-	list_del(&log_type->list);
+	list_del(&type->list);
 
 	spin_unlock(&_lock);
-	kfree(log_type);
 
 	return 0;
 }

drivers/md/dm-path-selector.c

@@ -17,9 +17,7 @@
 
 struct ps_internal {
 	struct path_selector_type pst;
-
 	struct list_head list;
-	long use;
 };
 
 #define pst_to_psi(__pst) container_of((__pst), struct ps_internal, pst)
@@ -45,12 +43,8 @@ static struct ps_internal *get_path_selector(const char *name)
 
 	down_read(&_ps_lock);
 	psi = __find_path_selector_type(name);
-	if (psi) {
-		if ((psi->use == 0) && !try_module_get(psi->pst.module))
-			psi = NULL;
-		else
-			psi->use++;
-	}
+	if (psi && !try_module_get(psi->pst.module))
+		psi = NULL;
 	up_read(&_ps_lock);
 
 	return psi;
@@ -84,11 +78,7 @@ void dm_put_path_selector(struct path_selector_type *pst)
 	if (!psi)
 		goto out;
 
-	if (--psi->use == 0)
-		module_put(psi->pst.module);
-
-	BUG_ON(psi->use < 0);
-
+	module_put(psi->pst.module);
 out:
 	up_read(&_ps_lock);
 }
@@ -136,11 +126,6 @@ int dm_unregister_path_selector(struct path_selector_type *pst)
 		return -EINVAL;
 	}
 
-	if (psi->use) {
-		up_write(&_ps_lock);
-		return -ETXTBSY;
-	}
-
 	list_del(&psi->list);
 
 	up_write(&_ps_lock);

drivers/md/dm-raid1.c

@@ -145,6 +145,8 @@ struct dm_raid1_read_record {
 	struct dm_bio_details details;
 };
 
+static struct kmem_cache *_dm_raid1_read_record_cache;
+
 /*
  * Every mirror should look like this one.
  */
@@ -586,6 +588,9 @@ static void do_writes(struct mirror_set *ms, struct bio_list *writes)
 	int state;
 	struct bio *bio;
 	struct bio_list sync, nosync, recover, *this_list = NULL;
+	struct bio_list requeue;
+	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
+	region_t region;
 
 	if (!writes->head)
 		return;
@@ -596,10 +601,18 @@ static void do_writes(struct mirror_set *ms, struct bio_list *writes)
 	bio_list_init(&sync);
 	bio_list_init(&nosync);
 	bio_list_init(&recover);
+	bio_list_init(&requeue);
 
 	while ((bio = bio_list_pop(writes))) {
-		state = dm_rh_get_state(ms->rh,
-					dm_rh_bio_to_region(ms->rh, bio), 1);
+		region = dm_rh_bio_to_region(ms->rh, bio);
+
+		if (log->type->is_remote_recovering &&
+		    log->type->is_remote_recovering(log, region)) {
+			bio_list_add(&requeue, bio);
+			continue;
+		}
+
+		state = dm_rh_get_state(ms->rh, region, 1);
 		switch (state) {
 		case DM_RH_CLEAN:
 		case DM_RH_DIRTY:
@@ -618,6 +631,16 @@ static void do_writes(struct mirror_set *ms, struct bio_list *writes)
 		bio_list_add(this_list, bio);
 	}
 
+	/*
+	 * Add bios that are delayed due to remote recovery
+	 * back on to the write queue
+	 */
+	if (unlikely(requeue.head)) {
+		spin_lock_irq(&ms->lock);
+		bio_list_merge(&ms->writes, &requeue);
+		spin_unlock_irq(&ms->lock);
+	}
+
 	/*
 	 * Increment the pending counts for any regions that will
 	 * be written to (writes to recover regions are going to
@@ -764,9 +787,9 @@ static struct mirror_set *alloc_context(unsigned int nr_mirrors,
 	atomic_set(&ms->suspend, 0);
 	atomic_set(&ms->default_mirror, DEFAULT_MIRROR);
 
-	len = sizeof(struct dm_raid1_read_record);
-	ms->read_record_pool = mempool_create_kmalloc_pool(MIN_READ_RECORDS,
-							   len);
+	ms->read_record_pool = mempool_create_slab_pool(MIN_READ_RECORDS,
+						_dm_raid1_read_record_cache);
+
 	if (!ms->read_record_pool) {
 		ti->error = "Error creating mirror read_record_pool";
 		kfree(ms);
@@ -1279,16 +1302,31 @@ static int __init dm_mirror_init(void)
 {
 	int r;
 
+	_dm_raid1_read_record_cache = KMEM_CACHE(dm_raid1_read_record, 0);
+	if (!_dm_raid1_read_record_cache) {
+		DMERR("Can't allocate dm_raid1_read_record cache");
+		r = -ENOMEM;
+		goto bad_cache;
+	}
+
 	r = dm_register_target(&mirror_target);
-	if (r < 0)
+	if (r < 0) {
 		DMERR("Failed to register mirror target");
+		goto bad_target;
+	}
+
+	return 0;
 
+bad_target:
+	kmem_cache_destroy(_dm_raid1_read_record_cache);
+bad_cache:
 	return r;
 }
 
 static void __exit dm_mirror_exit(void)
 {
 	dm_unregister_target(&mirror_target);
+	kmem_cache_destroy(_dm_raid1_read_record_cache);
 }
 
 /* Module hooks */

drivers/md/dm-snap-persistent.c

@@ -6,7 +6,6 @@
  */
 
 #include "dm-exception-store.h"
-#include "dm-snap.h"
 
 #include <linux/mm.h>
 #include <linux/pagemap.h>
@@ -89,7 +88,7 @@ struct commit_callback {
  * The top level structure for a persistent exception store.
  */
 struct pstore {
-	struct dm_snapshot *snap;	/* up pointer to my snapshot */
+	struct dm_exception_store *store;
 	int version;
 	int valid;
 	uint32_t exceptions_per_area;
@@ -141,7 +140,7 @@ static int alloc_area(struct pstore *ps)
 	int r = -ENOMEM;
 	size_t len;
 
-	len = ps->snap->chunk_size << SECTOR_SHIFT;
+	len = ps->store->chunk_size << SECTOR_SHIFT;
 
 	/*
 	 * Allocate the chunk_size block of memory that will hold
@@ -163,9 +162,12 @@ static int alloc_area(struct pstore *ps)
 
 static void free_area(struct pstore *ps)
 {
-	vfree(ps->area);
+	if (ps->area)
+		vfree(ps->area);
 	ps->area = NULL;
-	vfree(ps->zero_area);
+
+	if (ps->zero_area)
+		vfree(ps->zero_area);
 	ps->zero_area = NULL;
 }
 
@@ -189,9 +191,9 @@ static void do_metadata(struct work_struct *work)
 static int chunk_io(struct pstore *ps, chunk_t chunk, int rw, int metadata)
 {
 	struct dm_io_region where = {
-		.bdev = ps->snap->cow->bdev,
-		.sector = ps->snap->chunk_size * chunk,
-		.count = ps->snap->chunk_size,
+		.bdev = ps->store->cow->bdev,
+		.sector = ps->store->chunk_size * chunk,
+		.count = ps->store->chunk_size,
 	};
 	struct dm_io_request io_req = {
 		.bi_rw = rw,
@@ -247,15 +249,15 @@ static int area_io(struct pstore *ps, int rw)
 
 static void zero_memory_area(struct pstore *ps)
 {
-	memset(ps->area, 0, ps->snap->chunk_size << SECTOR_SHIFT);
+	memset(ps->area, 0, ps->store->chunk_size << SECTOR_SHIFT);
 }
 
 static int zero_disk_area(struct pstore *ps, chunk_t area)
 {
 	struct dm_io_region where = {
-		.bdev = ps->snap->cow->bdev,
-		.sector = ps->snap->chunk_size * area_location(ps, area),
-		.count = ps->snap->chunk_size,
+		.bdev = ps->store->cow->bdev,
+		.sector = ps->store->chunk_size * area_location(ps, area),
+		.count = ps->store->chunk_size,
 	};
 	struct dm_io_request io_req = {
 		.bi_rw = WRITE,
@@ -278,15 +280,15 @@ static int read_header(struct pstore *ps, int *new_snapshot)
 	/*
 	 * Use default chunk size (or hardsect_size, if larger) if none supplied
 	 */
-	if (!ps->snap->chunk_size) {
-		ps->snap->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS,
-		    bdev_hardsect_size(ps->snap->cow->bdev) >> 9);
-		ps->snap->chunk_mask = ps->snap->chunk_size - 1;
-		ps->snap->chunk_shift = ffs(ps->snap->chunk_size) - 1;
+	if (!ps->store->chunk_size) {
+		ps->store->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS,
+		    bdev_hardsect_size(ps->store->cow->bdev) >> 9);
+		ps->store->chunk_mask = ps->store->chunk_size - 1;
+		ps->store->chunk_shift = ffs(ps->store->chunk_size) - 1;
 		chunk_size_supplied = 0;
 	}
 
-	ps->io_client = dm_io_client_create(sectors_to_pages(ps->snap->
+	ps->io_client = dm_io_client_create(sectors_to_pages(ps->store->
 							     chunk_size));
 	if (IS_ERR(ps->io_client))
 		return PTR_ERR(ps->io_client);
@@ -317,22 +319,22 @@ static int read_header(struct pstore *ps, int *new_snapshot)
 	ps->version = le32_to_cpu(dh->version);
 	chunk_size = le32_to_cpu(dh->chunk_size);
 
-	if (!chunk_size_supplied || ps->snap->chunk_size == chunk_size)
+	if (!chunk_size_supplied || ps->store->chunk_size == chunk_size)
 		return 0;
 
 	DMWARN("chunk size %llu in device metadata overrides "
 	       "table chunk size of %llu.",
 	       (unsigned long long)chunk_size,
-	       (unsigned long long)ps->snap->chunk_size);
+	       (unsigned long long)ps->store->chunk_size);
 
 	/* We had a bogus chunk_size. Fix stuff up. */
 	free_area(ps);
 
-	ps->snap->chunk_size = chunk_size;
-	ps->snap->chunk_mask = chunk_size - 1;
-	ps->snap->chunk_shift = ffs(chunk_size) - 1;
+	ps->store->chunk_size = chunk_size;
+	ps->store->chunk_mask = chunk_size - 1;
+	ps->store->chunk_shift = ffs(chunk_size) - 1;
 
-	r = dm_io_client_resize(sectors_to_pages(ps->snap->chunk_size),
+	r = dm_io_client_resize(sectors_to_pages(ps->store->chunk_size),
 				ps->io_client);
 	if (r)
 		return r;
@@ -349,13 +351,13 @@ static int write_header(struct pstore *ps)
 {
 	struct disk_header *dh;
 
-	memset(ps->area, 0, ps->snap->chunk_size << SECTOR_SHIFT);
+	memset(ps->area, 0, ps->store->chunk_size << SECTOR_SHIFT);
 
 	dh = (struct disk_header *) ps->area;
 	dh->magic = cpu_to_le32(SNAP_MAGIC);
 	dh->valid = cpu_to_le32(ps->valid);
 	dh->version = cpu_to_le32(ps->version);
-	dh->chunk_size = cpu_to_le32(ps->snap->chunk_size);
+	dh->chunk_size = cpu_to_le32(ps->store->chunk_size);
 
 	return chunk_io(ps, 0, WRITE, 1);
 }
@@ -474,18 +476,25 @@ static struct pstore *get_info(struct dm_exception_store *store)
 static void persistent_fraction_full(struct dm_exception_store *store,
 				     sector_t *numerator, sector_t *denominator)
 {
-	*numerator = get_info(store)->next_free * store->snap->chunk_size;
-	*denominator = get_dev_size(store->snap->cow->bdev);
+	*numerator = get_info(store)->next_free * store->chunk_size;
+	*denominator = get_dev_size(store->cow->bdev);
 }
 
-static void persistent_destroy(struct dm_exception_store *store)
+static void persistent_dtr(struct dm_exception_store *store)
 {
 	struct pstore *ps = get_info(store);
 
 	destroy_workqueue(ps->metadata_wq);
-	dm_io_client_destroy(ps->io_client);
-	vfree(ps->callbacks);
+
+	/* Created in read_header */
+	if (ps->io_client)
+		dm_io_client_destroy(ps->io_client);
 	free_area(ps);
+
+	/* Allocated in persistent_read_metadata */
+	if (ps->callbacks)
+		vfree(ps->callbacks);
+
 	kfree(ps);
 }
 
@@ -507,7 +516,7 @@ static int persistent_read_metadata(struct dm_exception_store *store,
 	/*
 	 * Now we know correct chunk_size, complete the initialisation.
 	 */
-	ps->exceptions_per_area = (ps->snap->chunk_size << SECTOR_SHIFT) /
+	ps->exceptions_per_area = (ps->store->chunk_size << SECTOR_SHIFT) /
 				  sizeof(struct disk_exception);
 	ps->callbacks = dm_vcalloc(ps->exceptions_per_area,
 			sizeof(*ps->callbacks));
@@ -564,10 +573,10 @@ static int persistent_prepare_exception(struct dm_exception_store *store,
 	struct pstore *ps = get_info(store);
 	uint32_t stride;
 	chunk_t next_free;
-	sector_t size = get_dev_size(store->snap->cow->bdev);
+	sector_t size = get_dev_size(store->cow->bdev);
 
 	/* Is there enough room ? */
-	if (size < ((ps->next_free + 1) * store->snap->chunk_size))
+	if (size < ((ps->next_free + 1) * store->chunk_size))
 		return -ENOSPC;
 
 	e->new_chunk = ps->next_free;
@@ -656,16 +665,17 @@ static void persistent_drop_snapshot(struct dm_exception_store *store)
 		DMWARN("write header failed");
 }
 
-int dm_create_persistent(struct dm_exception_store *store)
+static int persistent_ctr(struct dm_exception_store *store,
+			  unsigned argc, char **argv)
 {
 	struct pstore *ps;
 
 	/* allocate the pstore */
-	ps = kmalloc(sizeof(*ps), GFP_KERNEL);
+	ps = kzalloc(sizeof(*ps), GFP_KERNEL);
 	if (!ps)
 		return -ENOMEM;
 
-	ps->snap = store->snap;
+	ps->store = store;
 	ps->valid = 1;
 	ps->version = SNAPSHOT_DISK_VERSION;
 	ps->area = NULL;
@@ -683,22 +693,77 @@ int dm_create_persistent(struct dm_exception_store *store)
 		return -ENOMEM;
 	}
 
-	store->destroy = persistent_destroy;
-	store->read_metadata = persistent_read_metadata;
-	store->prepare_exception = persistent_prepare_exception;
-	store->commit_exception = persistent_commit_exception;
-	store->drop_snapshot = persistent_drop_snapshot;
-	store->fraction_full = persistent_fraction_full;
 	store->context = ps;
 
 	return 0;
 }
 
+static unsigned persistent_status(struct dm_exception_store *store,
+				  status_type_t status, char *result,
+				  unsigned maxlen)
+{
+	unsigned sz = 0;
+
+	switch (status) {
+	case STATUSTYPE_INFO:
+		break;
+	case STATUSTYPE_TABLE:
+		DMEMIT(" %s P %llu", store->cow->name,
+		       (unsigned long long)store->chunk_size);
+	}
+
+	return sz;
+}
+
+static struct dm_exception_store_type _persistent_type = {
+	.name = "persistent",
+	.module = THIS_MODULE,
+	.ctr = persistent_ctr,
+	.dtr = persistent_dtr,
+	.read_metadata = persistent_read_metadata,
+	.prepare_exception = persistent_prepare_exception,
+	.commit_exception = persistent_commit_exception,
+	.drop_snapshot = persistent_drop_snapshot,
+	.fraction_full = persistent_fraction_full,
+	.status = persistent_status,
+};
+
+static struct dm_exception_store_type _persistent_compat_type = {
+	.name = "P",
+	.module = THIS_MODULE,
+	.ctr = persistent_ctr,
+	.dtr = persistent_dtr,
+	.read_metadata = persistent_read_metadata,
+	.prepare_exception = persistent_prepare_exception,
+	.commit_exception = persistent_commit_exception,
+	.drop_snapshot = persistent_drop_snapshot,
+	.fraction_full = persistent_fraction_full,
+	.status = persistent_status,
+};
+
 int dm_persistent_snapshot_init(void)
 {
-	return 0;
+	int r;
+
+	r = dm_exception_store_type_register(&_persistent_type);
+	if (r) {
+		DMERR("Unable to register persistent exception store type");
+		return r;
+	}
+
+	r = dm_exception_store_type_register(&_persistent_compat_type);
+	if (r) {
+		DMERR("Unable to register old-style persistent exception "
+		      "store type");
+		dm_exception_store_type_unregister(&_persistent_type);
+		return r;
+	}
+
+	return r;
 }
 
 void dm_persistent_snapshot_exit(void)
 {
+	dm_exception_store_type_unregister(&_persistent_type);
+	dm_exception_store_type_unregister(&_persistent_compat_type);
 }

drivers/md/dm-snap-transient.c

@@ -6,7 +6,6 @@
  */
 
 #include "dm-exception-store.h"
-#include "dm-snap.h"
 
 #include <linux/mm.h>
 #include <linux/pagemap.h>
@@ -23,7 +22,7 @@ struct transient_c {
 	sector_t next_free;
 };
 
-static void transient_destroy(struct dm_exception_store *store)
+static void transient_dtr(struct dm_exception_store *store)
 {
 	kfree(store->context);
 }
@@ -39,14 +38,14 @@ static int transient_read_metadata(struct dm_exception_store *store,
 static int transient_prepare_exception(struct dm_exception_store *store,
 				       struct dm_snap_exception *e)
 {
-	struct transient_c *tc = (struct transient_c *) store->context;
-	sector_t size = get_dev_size(store->snap->cow->bdev);
+	struct transient_c *tc = store->context;
+	sector_t size = get_dev_size(store->cow->bdev);
 
-	if (size < (tc->next_free + store->snap->chunk_size))
+	if (size < (tc->next_free + store->chunk_size))
 		return -1;
 
-	e->new_chunk = sector_to_chunk(store->snap, tc->next_free);
-	tc->next_free += store->snap->chunk_size;
+	e->new_chunk = sector_to_chunk(store, tc->next_free);
+	tc->next_free += store->chunk_size;
 
 	return 0;
 }
@@ -64,20 +63,14 @@ static void transient_fraction_full(struct dm_exception_store *store,
 				    sector_t *numerator, sector_t *denominator)
 {
 	*numerator = ((struct transient_c *) store->context)->next_free;
-	*denominator = get_dev_size(store->snap->cow->bdev);
+	*denominator = get_dev_size(store->cow->bdev);
 }
 
-int dm_create_transient(struct dm_exception_store *store)
+static int transient_ctr(struct dm_exception_store *store,
+			 unsigned argc, char **argv)
 {
 	struct transient_c *tc;
 
-	store->destroy = transient_destroy;
-	store->read_metadata = transient_read_metadata;
-	store->prepare_exception = transient_prepare_exception;
-	store->commit_exception = transient_commit_exception;
-	store->drop_snapshot = NULL;
-	store->fraction_full = transient_fraction_full;
-
 	tc = kmalloc(sizeof(struct transient_c), GFP_KERNEL);
 	if (!tc)
 		return -ENOMEM;
@@ -88,11 +81,70 @@ int dm_create_transient(struct dm_exception_store *store)
 	return 0;
 }
 
+static unsigned transient_status(struct dm_exception_store *store,
+				 status_type_t status, char *result,
+				 unsigned maxlen)
+{
+	unsigned sz = 0;
+
+	switch (status) {
+	case STATUSTYPE_INFO:
+		break;
+	case STATUSTYPE_TABLE:
+		DMEMIT(" %s N %llu", store->cow->name,
+		       (unsigned long long)store->chunk_size);
+	}
+
+	return sz;
+}
+
+static struct dm_exception_store_type _transient_type = {
+	.name = "transient",
+	.module = THIS_MODULE,
+	.ctr = transient_ctr,
+	.dtr = transient_dtr,
+	.read_metadata = transient_read_metadata,
+	.prepare_exception = transient_prepare_exception,
+	.commit_exception = transient_commit_exception,
+	.fraction_full = transient_fraction_full,
+	.status = transient_status,
+};
+
+static struct dm_exception_store_type _transient_compat_type = {
+	.name = "N",
+	.module = THIS_MODULE,
+	.ctr = transient_ctr,
+	.dtr = transient_dtr,
+	.read_metadata = transient_read_metadata,
+	.prepare_exception = transient_prepare_exception,
+	.commit_exception = transient_commit_exception,
+	.fraction_full = transient_fraction_full,
+	.status = transient_status,
+};
+
 int dm_transient_snapshot_init(void)
 {
-	return 0;
+	int r;
+
+	r = dm_exception_store_type_register(&_transient_type);
+	if (r) {
+		DMWARN("Unable to register transient exception store type");
+		return r;
+	}
+
+	r = dm_exception_store_type_register(&_transient_compat_type);
+	if (r) {
+		DMWARN("Unable to register old-style transient "
+		       "exception store type");
+		dm_exception_store_type_unregister(&_transient_type);
+		return r;
+	}
+
+	return r;
 }
 
 void dm_transient_snapshot_exit(void)
 {
+	dm_exception_store_type_unregister(&_transient_type);
+	dm_exception_store_type_unregister(&_transient_compat_type);
 }

drivers/md/dm-snap.c

@@ -7,7 +7,6 @@
  */
 
 #include <linux/blkdev.h>
-#include <linux/ctype.h>
 #include <linux/device-mapper.h>
 #include <linux/delay.h>
 #include <linux/fs.h>
@@ -20,9 +19,9 @@
 #include <linux/vmalloc.h>
 #include <linux/log2.h>
 #include <linux/dm-kcopyd.h>
+#include <linux/workqueue.h>
 
 #include "dm-exception-store.h"
-#include "dm-snap.h"
 #include "dm-bio-list.h"
 
 #define DM_MSG_PREFIX "snapshots"
@@ -47,9 +46,76 @@
  */
 #define MIN_IOS 256
 
+#define DM_TRACKED_CHUNK_HASH_SIZE	16
+#define DM_TRACKED_CHUNK_HASH(x)	((unsigned long)(x) & \
+					 (DM_TRACKED_CHUNK_HASH_SIZE - 1))
+
+struct exception_table {
+	uint32_t hash_mask;
+	unsigned hash_shift;
+	struct list_head *table;
+};
+
+struct dm_snapshot {
+	struct rw_semaphore lock;
+
+	struct dm_dev *origin;
+
+	/* List of snapshots per Origin */
+	struct list_head list;
+
+	/* You can't use a snapshot if this is 0 (e.g. if full) */
+	int valid;
+
+	/* Origin writes don't trigger exceptions until this is set */
+	int active;
+
+	mempool_t *pending_pool;
+
+	atomic_t pending_exceptions_count;
+
+	struct exception_table pending;
+	struct exception_table complete;
+
+	/*
+	 * pe_lock protects all pending_exception operations and access
+	 * as well as the snapshot_bios list.
+	 */
+	spinlock_t pe_lock;
+
+	/* The on disk metadata handler */
+	struct dm_exception_store *store;
+
+	struct dm_kcopyd_client *kcopyd_client;
+
+	/* Queue of snapshot writes for ksnapd to flush */
+	struct bio_list queued_bios;
+	struct work_struct queued_bios_work;
+
+	/* Chunks with outstanding reads */
+	mempool_t *tracked_chunk_pool;
+	spinlock_t tracked_chunk_lock;
+	struct hlist_head tracked_chunk_hash[DM_TRACKED_CHUNK_HASH_SIZE];
+};
+
 static struct workqueue_struct *ksnapd;
 static void flush_queued_bios(struct work_struct *work);
 
+static sector_t chunk_to_sector(struct dm_exception_store *store,
+				chunk_t chunk)
+{
+	return chunk << store->chunk_shift;
+}
+
+static int bdev_equal(struct block_device *lhs, struct block_device *rhs)
+{
+	/*
+	 * There is only ever one instance of a particular block
+	 * device so we can compare pointers safely.
+	 */
+	return lhs == rhs;
+}
+
 struct dm_snap_pending_exception {
 	struct dm_snap_exception e;
 
@@ -476,11 +542,11 @@ static int init_hash_tables(struct dm_snapshot *s)
 	 * Calculate based on the size of the original volume or
 	 * the COW volume...
 	 */
-	cow_dev_size = get_dev_size(s->cow->bdev);
+	cow_dev_size = get_dev_size(s->store->cow->bdev);
 	origin_dev_size = get_dev_size(s->origin->bdev);
 	max_buckets = calc_max_buckets();
 
-	hash_size = min(origin_dev_size, cow_dev_size) >> s->chunk_shift;
+	hash_size = min(origin_dev_size, cow_dev_size) >> s->store->chunk_shift;
 	hash_size = min(hash_size, max_buckets);
 
 	hash_size = rounddown_pow_of_two(hash_size);
@@ -504,58 +570,6 @@ static int init_hash_tables(struct dm_snapshot *s)
 	return 0;
 }
 
-/*
- * Round a number up to the nearest 'size' boundary.  size must
- * be a power of 2.
- */
-static ulong round_up(ulong n, ulong size)
-{
-	size--;
-	return (n + size) & ~size;
-}
-
-static int set_chunk_size(struct dm_snapshot *s, const char *chunk_size_arg,
-			  char **error)
-{
-	unsigned long chunk_size;
-	char *value;
-
-	chunk_size = simple_strtoul(chunk_size_arg, &value, 10);
-	if (*chunk_size_arg == '\0' || *value != '\0') {
-		*error = "Invalid chunk size";
-		return -EINVAL;
-	}
-
-	if (!chunk_size) {
-		s->chunk_size = s->chunk_mask = s->chunk_shift = 0;
-		return 0;
-	}
-
-	/*
-	 * Chunk size must be multiple of page size.  Silently
-	 * round up if it's not.
-	 */
-	chunk_size = round_up(chunk_size, PAGE_SIZE >> 9);
-
-	/* Check chunk_size is a power of 2 */
-	if (!is_power_of_2(chunk_size)) {
-		*error = "Chunk size is not a power of 2";
-		return -EINVAL;
-	}
-
-	/* Validate the chunk size against the device block size */
-	if (chunk_size % (bdev_hardsect_size(s->cow->bdev) >> 9)) {
-		*error = "Chunk size is not a multiple of device blocksize";
-		return -EINVAL;
-	}
-
-	s->chunk_size = chunk_size;
-	s->chunk_mask = chunk_size - 1;
-	s->chunk_shift = ffs(chunk_size) - 1;
-
-	return 0;
-}
-
 /*
  * Construct a snapshot mapping: <origin_dev> <COW-dev> <p/n> <chunk-size>
  */
@@ -564,91 +578,68 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	struct dm_snapshot *s;
 	int i;
 	int r = -EINVAL;
-	char persistent;
 	char *origin_path;
-	char *cow_path;
+	struct dm_exception_store *store;
+	unsigned args_used;
 
 	if (argc != 4) {
 		ti->error = "requires exactly 4 arguments";
 		r = -EINVAL;
-		goto bad1;
+		goto bad_args;
 	}
 
 	origin_path = argv[0];
-	cow_path = argv[1];
-	persistent = toupper(*argv[2]);
+	argv++;
+	argc--;
 
-	if (persistent != 'P' && persistent != 'N') {
-		ti->error = "Persistent flag is not P or N";
+	r = dm_exception_store_create(ti, argc, argv, &args_used, &store);
+	if (r) {
+		ti->error = "Couldn't create exception store";
 		r = -EINVAL;
-		goto bad1;
+		goto bad_args;
 	}
 
+	argv += args_used;
+	argc -= args_used;
+
 	s = kmalloc(sizeof(*s), GFP_KERNEL);
-	if (s == NULL) {
+	if (!s) {
 		ti->error = "Cannot allocate snapshot context private "
 		    "structure";
 		r = -ENOMEM;
-		goto bad1;
+		goto bad_snap;
 	}
 
 	r = dm_get_device(ti, origin_path, 0, ti->len, FMODE_READ, &s->origin);
 	if (r) {
 		ti->error = "Cannot get origin device";
-		goto bad2;
-	}
-
-	r = dm_get_device(ti, cow_path, 0, 0,
-			  FMODE_READ | FMODE_WRITE, &s->cow);
-	if (r) {
-		dm_put_device(ti, s->origin);
-		ti->error = "Cannot get COW device";
-		goto bad2;
+		goto bad_origin;
 	}
 
-	r = set_chunk_size(s, argv[3], &ti->error);
-	if (r)
-		goto bad3;
-
-	s->type = persistent;
-
+	s->store = store;
 	s->valid = 1;
 	s->active = 0;
 	atomic_set(&s->pending_exceptions_count, 0);
 	init_rwsem(&s->lock);
 	spin_lock_init(&s->pe_lock);
-	s->ti = ti;
 
 	/* Allocate hash table for COW data */
 	if (init_hash_tables(s)) {
 		ti->error = "Unable to allocate hash table space";
 		r = -ENOMEM;
-		goto bad3;
-	}
-
-	s->store.snap = s;
-
-	if (persistent == 'P')
-		r = dm_create_persistent(&s->store);
-	else
-		r = dm_create_transient(&s->store);
-
-	if (r) {
-		ti->error = "Couldn't create exception store";
-		r = -EINVAL;
-		goto bad4;
+		goto bad_hash_tables;
 	}
 
 	r = dm_kcopyd_client_create(SNAPSHOT_PAGES, &s->kcopyd_client);
 	if (r) {
 		ti->error = "Could not create kcopyd client";
-		goto bad5;
+		goto bad_kcopyd;
 	}
 
 	s->pending_pool = mempool_create_slab_pool(MIN_IOS, pending_cache);
 	if (!s->pending_pool) {
 		ti->error = "Could not allocate mempool for pending exceptions";
-		goto bad6;
+		goto bad_pending_pool;
 	}
 
 	s->tracked_chunk_pool = mempool_create_slab_pool(MIN_IOS,
@@ -665,7 +656,8 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	spin_lock_init(&s->tracked_chunk_lock);
 
 	/* Metadata must only be loaded into one table at once */
-	r = s->store.read_metadata(&s->store, dm_add_exception, (void *)s);
+	r = s->store->type->read_metadata(s->store, dm_add_exception,
+					  (void *)s);
 	if (r < 0) {
 		ti->error = "Failed to read snapshot metadata";
 		goto bad_load_and_register;
@@ -686,34 +678,33 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	}
 
 	ti->private = s;
-	ti->split_io = s->chunk_size;
+	ti->split_io = s->store->chunk_size;
 
 	return 0;
 
- bad_load_and_register:
+bad_load_and_register:
 	mempool_destroy(s->tracked_chunk_pool);
 
- bad_tracked_chunk_pool:
+bad_tracked_chunk_pool:
 	mempool_destroy(s->pending_pool);
 
- bad6:
+bad_pending_pool:
 	dm_kcopyd_client_destroy(s->kcopyd_client);
 
- bad5:
-	s->store.destroy(&s->store);
-
- bad4:
+bad_kcopyd:
 	exit_exception_table(&s->pending, pending_cache);
 	exit_exception_table(&s->complete, exception_cache);
 
- bad3:
-	dm_put_device(ti, s->cow);
+bad_hash_tables:
 	dm_put_device(ti, s->origin);
 
- bad2:
+bad_origin:
 	kfree(s);
 
- bad1:
+bad_snap:
+	dm_exception_store_destroy(store);
+
+bad_args:
 	return r;
 }
 
@@ -724,8 +715,6 @@ static void __free_exceptions(struct dm_snapshot *s)
 
 	exit_exception_table(&s->pending, pending_cache);
 	exit_exception_table(&s->complete, exception_cache);
-
-	s->store.destroy(&s->store);
 }
 
 static void snapshot_dtr(struct dm_target *ti)
@@ -761,7 +750,8 @@ static void snapshot_dtr(struct dm_target *ti)
 	mempool_destroy(s->pending_pool);
 
 	dm_put_device(ti, s->origin);
-	dm_put_device(ti, s->cow);
+
+	dm_exception_store_destroy(s->store);
 
 	kfree(s);
 }
@@ -820,12 +810,12 @@ static void __invalidate_snapshot(struct dm_snapshot *s, int err)
 	else if (err == -ENOMEM)
 		DMERR("Invalidating snapshot: Unable to allocate exception.");
 
-	if (s->store.drop_snapshot)
-		s->store.drop_snapshot(&s->store);
+	if (s->store->type->drop_snapshot)
+		s->store->type->drop_snapshot(s->store);
 
 	s->valid = 0;
 
-	dm_table_event(s->ti->table);
+	dm_table_event(s->store->ti->table);
 }
 
 static void get_pending_exception(struct dm_snap_pending_exception *pe)
@@ -943,8 +933,8 @@ static void copy_callback(int read_err, unsigned long write_err, void *context)
 
 	else
 		/* Update the metadata if we are persistent */
-		s->store.commit_exception(&s->store, &pe->e, commit_callback,
-					  pe);
+		s->store->type->commit_exception(s->store, &pe->e,
+						 commit_callback, pe);
 }
 
 /*
@@ -960,11 +950,11 @@ static void start_copy(struct dm_snap_pending_exception *pe)
 	dev_size = get_dev_size(bdev);
 
 	src.bdev = bdev;
-	src.sector = chunk_to_sector(s, pe->e.old_chunk);
-	src.count = min(s->chunk_size, dev_size - src.sector);
+	src.sector = chunk_to_sector(s->store, pe->e.old_chunk);
+	src.count = min(s->store->chunk_size, dev_size - src.sector);
 
-	dest.bdev = s->cow->bdev;
-	dest.sector = chunk_to_sector(s, pe->e.new_chunk);
+	dest.bdev = s->store->cow->bdev;
+	dest.sector = chunk_to_sector(s->store, pe->e.new_chunk);
 	dest.count = src.count;
 
 	/* Hand over to kcopyd */
@@ -972,6 +962,17 @@ static void start_copy(struct dm_snap_pending_exception *pe)
 		    &src, 1, &dest, 0, copy_callback, pe);
 }
 
+static struct dm_snap_pending_exception *
+__lookup_pending_exception(struct dm_snapshot *s, chunk_t chunk)
+{
+	struct dm_snap_exception *e = lookup_exception(&s->pending, chunk);
+
+	if (!e)
+		return NULL;
+
+	return container_of(e, struct dm_snap_pending_exception, e);
+}
+
 /*
  * Looks to see if this snapshot already has a pending exception
  * for this chunk, otherwise it allocates a new one and inserts
@@ -981,40 +982,15 @@ static void start_copy(struct dm_snap_pending_exception *pe)
  * this.
  */
 static struct dm_snap_pending_exception *
-__find_pending_exception(struct dm_snapshot *s, struct bio *bio)
+__find_pending_exception(struct dm_snapshot *s,
+			 struct dm_snap_pending_exception *pe, chunk_t chunk)
 {
-	struct dm_snap_exception *e;
-	struct dm_snap_pending_exception *pe;
-	chunk_t chunk = sector_to_chunk(s, bio->bi_sector);
+	struct dm_snap_pending_exception *pe2;
 
-	/*
-	 * Is there a pending exception for this already ?
-	 */
-	e = lookup_exception(&s->pending, chunk);
-	if (e) {
-		/* cast the exception to a pending exception */
-		pe = container_of(e, struct dm_snap_pending_exception, e);
-		goto out;
-	}
-
-	/*
-	 * Create a new pending exception, we don't want
-	 * to hold the lock while we do this.
-	 */
-	up_write(&s->lock);
-	pe = alloc_pending_exception(s);
-	down_write(&s->lock);
-
-	if (!s->valid) {
-		free_pending_exception(pe);
-		return NULL;
-	}
-
-	e = lookup_exception(&s->pending, chunk);
-	if (e) {
+	pe2 = __lookup_pending_exception(s, chunk);
+	if (pe2) {
 		free_pending_exception(pe);
-		pe = container_of(e, struct dm_snap_pending_exception, e);
-		goto out;
+		return pe2;
 	}
 
 	pe->e.old_chunk = chunk;
@@ -1024,7 +1000,7 @@ __find_pending_exception(struct dm_snapshot *s, struct bio *bio)
 	atomic_set(&pe->ref_count, 0);
 	pe->started = 0;
 
-	if (s->store.prepare_exception(&s->store, &pe->e)) {
+	if (s->store->type->prepare_exception(s->store, &pe->e)) {
 		free_pending_exception(pe);
 		return NULL;
 	}
@@ -1032,17 +1008,18 @@ __find_pending_exception(struct dm_snapshot *s, struct bio *bio)
 	get_pending_exception(pe);
 	insert_exception(&s->pending, &pe->e);
 
- out:
 	return pe;
 }
 
 static void remap_exception(struct dm_snapshot *s, struct dm_snap_exception *e,
 			    struct bio *bio, chunk_t chunk)
 {
-	bio->bi_bdev = s->cow->bdev;
-	bio->bi_sector = chunk_to_sector(s, dm_chunk_number(e->new_chunk) +
-			 (chunk - e->old_chunk)) +
-			 (bio->bi_sector & s->chunk_mask);
+	bio->bi_bdev = s->store->cow->bdev;
+	bio->bi_sector = chunk_to_sector(s->store,
+					 dm_chunk_number(e->new_chunk) +
+					 (chunk - e->old_chunk)) +
+					 (bio->bi_sector &
+					  s->store->chunk_mask);
 }
 
 static int snapshot_map(struct dm_target *ti, struct bio *bio,
@@ -1054,7 +1031,7 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio,
 	chunk_t chunk;
 	struct dm_snap_pending_exception *pe = NULL;
 
-	chunk = sector_to_chunk(s, bio->bi_sector);
+	chunk = sector_to_chunk(s->store, bio->bi_sector);
 
 	/* Full snapshots are not usable */
 	/* To get here the table must be live so s->active is always set. */
@@ -1083,11 +1060,31 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio,
 	 * writeable.
 	 */
 	if (bio_rw(bio) == WRITE) {
-		pe = __find_pending_exception(s, bio);
+		pe = __lookup_pending_exception(s, chunk);
 		if (!pe) {
-			__invalidate_snapshot(s, -ENOMEM);
-			r = -EIO;
-			goto out_unlock;
+			up_write(&s->lock);
+			pe = alloc_pending_exception(s);
+			down_write(&s->lock);
+
+			if (!s->valid) {
+				free_pending_exception(pe);
+				r = -EIO;
+				goto out_unlock;
+			}
+
+			e = lookup_exception(&s->complete, chunk);
+			if (e) {
+				free_pending_exception(pe);
+				remap_exception(s, e, bio, chunk);
+				goto out_unlock;
+			}
+
+			pe = __find_pending_exception(s, pe, chunk);
+			if (!pe) {
+				__invalidate_snapshot(s, -ENOMEM);
+				r = -EIO;
+				goto out_unlock;
+			}
 		}
 
 		remap_exception(s, &pe->e, bio, chunk);
@@ -1137,24 +1134,25 @@ static void snapshot_resume(struct dm_target *ti)
 static int snapshot_status(struct dm_target *ti, status_type_t type,
 			   char *result, unsigned int maxlen)
 {
+	unsigned sz = 0;
 	struct dm_snapshot *snap = ti->private;
 
 	switch (type) {
 	case STATUSTYPE_INFO:
 		if (!snap->valid)
-			snprintf(result, maxlen, "Invalid");
+			DMEMIT("Invalid");
 		else {
-			if (snap->store.fraction_full) {
+			if (snap->store->type->fraction_full) {
 				sector_t numerator, denominator;
-				snap->store.fraction_full(&snap->store,
-							  &numerator,
-							  &denominator);
-				snprintf(result, maxlen, "%llu/%llu",
-					(unsigned long long)numerator,
-					(unsigned long long)denominator);
+				snap->store->type->fraction_full(snap->store,
+								 &numerator,
+								 &denominator);
+				DMEMIT("%llu/%llu",
+				       (unsigned long long)numerator,
+				       (unsigned long long)denominator);
 			}
 			else
-				snprintf(result, maxlen, "Unknown");
+				DMEMIT("Unknown");
 		}
 		break;
 
@@ -1164,10 +1162,9 @@ static int snapshot_status(struct dm_target *ti, status_type_t type,
 		 * to make private copies if the output is to
 		 * make sense.
 		 */
-		snprintf(result, maxlen, "%s %s %c %llu",
-			 snap->origin->name, snap->cow->name,
-			 snap->type,
-			 (unsigned long long)snap->chunk_size);
+		DMEMIT("%s", snap->origin->name);
+		snap->store->type->status(snap->store, type, result + sz,
+					  maxlen - sz);
 		break;
 	}
 
@@ -1196,14 +1193,14 @@ static int __origin_write(struct list_head *snapshots, struct bio *bio)
 			goto next_snapshot;
 
 		/* Nothing to do if writing beyond end of snapshot */
-		if (bio->bi_sector >= dm_table_get_size(snap->ti->table))
+		if (bio->bi_sector >= dm_table_get_size(snap->store->ti->table))
 			goto next_snapshot;
 
 		/*
 		 * Remember, different snapshots can have
 		 * different chunk sizes.
 		 */
-		chunk = sector_to_chunk(snap, bio->bi_sector);
+		chunk = sector_to_chunk(snap->store, bio->bi_sector);
 
 		/*
 		 * Check exception table to see if block
@@ -1217,10 +1214,28 @@ static int __origin_write(struct list_head *snapshots, struct bio *bio)
 		if (e)
 			goto next_snapshot;
 
-		pe = __find_pending_exception(snap, bio);
+		pe = __lookup_pending_exception(snap, chunk);
 		if (!pe) {
-			__invalidate_snapshot(snap, -ENOMEM);
-			goto next_snapshot;
+			up_write(&snap->lock);
+			pe = alloc_pending_exception(snap);
+			down_write(&snap->lock);
+
+			if (!snap->valid) {
+				free_pending_exception(pe);
+				goto next_snapshot;
+			}
+
+			e = lookup_exception(&snap->complete, chunk);
+			if (e) {
+				free_pending_exception(pe);
+				goto next_snapshot;
+			}
+
+			pe = __find_pending_exception(snap, pe, chunk);
+			if (!pe) {
+				__invalidate_snapshot(snap, -ENOMEM);
+				goto next_snapshot;
+			}
 		}
 
 		if (!primary_pe) {
@@ -1360,7 +1375,8 @@ static void origin_resume(struct dm_target *ti)
 	o = __lookup_origin(dev->bdev);
 	if (o)
 		list_for_each_entry (snap, &o->snapshots, list)
-			chunk_size = min_not_zero(chunk_size, snap->chunk_size);
+			chunk_size = min_not_zero(chunk_size,
+						  snap->store->chunk_size);
 	up_read(&_origins_lock);
 
 	ti->split_io = chunk_size;

drivers/md/dm-snap.h

@@ -1,105 +0,0 @@
-/*
- * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
- *
- * This file is released under the GPL.
- */
-
-#ifndef DM_SNAPSHOT_H
-#define DM_SNAPSHOT_H
-
-#include <linux/device-mapper.h>
-#include "dm-exception-store.h"
-#include "dm-bio-list.h"
-#include <linux/blkdev.h>
-#include <linux/workqueue.h>
-
-struct exception_table {
-	uint32_t hash_mask;
-	unsigned hash_shift;
-	struct list_head *table;
-};
-
-#define DM_TRACKED_CHUNK_HASH_SIZE	16
-#define DM_TRACKED_CHUNK_HASH(x)	((unsigned long)(x) & \
-					 (DM_TRACKED_CHUNK_HASH_SIZE - 1))
-
-struct dm_snapshot {
-	struct rw_semaphore lock;
-	struct dm_target *ti;
-
-	struct dm_dev *origin;
-	struct dm_dev *cow;
-
-	/* List of snapshots per Origin */
-	struct list_head list;
-
-	/* Size of data blocks saved - must be a power of 2 */
-	chunk_t chunk_size;
-	chunk_t chunk_mask;
-	chunk_t chunk_shift;
-
-	/* You can't use a snapshot if this is 0 (e.g. if full) */
-	int valid;
-
-	/* Origin writes don't trigger exceptions until this is set */
-	int active;
-
-	/* Used for display of table */
-	char type;
-
-	mempool_t *pending_pool;
-
-	atomic_t pending_exceptions_count;
-
-	struct exception_table pending;
-	struct exception_table complete;
-
-	/*
-	 * pe_lock protects all pending_exception operations and access
-	 * as well as the snapshot_bios list.
-	 */
-	spinlock_t pe_lock;
-
-	/* The on disk metadata handler */
-	struct dm_exception_store store;
-
-	struct dm_kcopyd_client *kcopyd_client;
-
-	/* Queue of snapshot writes for ksnapd to flush */
-	struct bio_list queued_bios;
-	struct work_struct queued_bios_work;
-
-	/* Chunks with outstanding reads */
-	mempool_t *tracked_chunk_pool;
-	spinlock_t tracked_chunk_lock;
-	struct hlist_head tracked_chunk_hash[DM_TRACKED_CHUNK_HASH_SIZE];
-};
-
-/*
- * Return the number of sectors in the device.
- */
-static inline sector_t get_dev_size(struct block_device *bdev)
-{
-	return bdev->bd_inode->i_size >> SECTOR_SHIFT;
-}
-
-static inline chunk_t sector_to_chunk(struct dm_snapshot *s, sector_t sector)
-{
-	return (sector & ~s->chunk_mask) >> s->chunk_shift;
-}
-
-static inline sector_t chunk_to_sector(struct dm_snapshot *s, chunk_t chunk)
-{
-	return chunk << s->chunk_shift;
-}
-
-static inline int bdev_equal(struct block_device *lhs, struct block_device *rhs)
-{
-	/*
-	 * There is only ever one instance of a particular block
-	 * device so we can compare pointers safely.
-	 */
-	return lhs == rhs;
-}
-
-#endif

drivers/md/dm-table.c

@@ -399,28 +399,30 @@ static int check_device_area(struct dm_dev_internal *dd, sector_t start,
 }
 
 /*
- * This upgrades the mode on an already open dm_dev.  Being
+ * This upgrades the mode on an already open dm_dev, being
  * careful to leave things as they were if we fail to reopen the
- * device.
+ * device and not to touch the existing bdev field in case
+ * it is accessed concurrently inside dm_table_any_congested().
  */
 static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode,
 			struct mapped_device *md)
 {
 	int r;
-	struct dm_dev_internal dd_copy;
-	dev_t dev = dd->dm_dev.bdev->bd_dev;
+	struct dm_dev_internal dd_new, dd_old;
 
-	dd_copy = *dd;
+	dd_new = dd_old = *dd;
+
+	dd_new.dm_dev.mode |= new_mode;
+	dd_new.dm_dev.bdev = NULL;
+
+	r = open_dev(&dd_new, dd->dm_dev.bdev->bd_dev, md);
+	if (r)
+		return r;
 
 	dd->dm_dev.mode |= new_mode;
-	dd->dm_dev.bdev = NULL;
-	r = open_dev(dd, dev, md);
-	if (!r)
-		close_dev(&dd_copy, md);
-	else
-		*dd = dd_copy;
+	close_dev(&dd_old, md);
 
-	return r;
+	return 0;
 }
 
 /*

drivers/md/dm-target.c

@@ -14,45 +14,34 @@
 
 #define DM_MSG_PREFIX "target"
 
-struct tt_internal {
-	struct target_type tt;
-
-	struct list_head list;
-	long use;
-};
-
 static LIST_HEAD(_targets);
 static DECLARE_RWSEM(_lock);
 
 #define DM_MOD_NAME_SIZE 32
 
-static inline struct tt_internal *__find_target_type(const char *name)
+static inline struct target_type *__find_target_type(const char *name)
 {
-	struct tt_internal *ti;
+	struct target_type *tt;
 
-	list_for_each_entry (ti, &_targets, list)
-		if (!strcmp(name, ti->tt.name))
-			return ti;
+	list_for_each_entry(tt, &_targets, list)
+		if (!strcmp(name, tt->name))
+			return tt;
 
 	return NULL;
 }
 
-static struct tt_internal *get_target_type(const char *name)
+static struct target_type *get_target_type(const char *name)
 {
-	struct tt_internal *ti;
+	struct target_type *tt;
 
 	down_read(&_lock);
 
-	ti = __find_target_type(name);
-	if (ti) {
-		if ((ti->use == 0) && !try_module_get(ti->tt.module))
-			ti = NULL;
-		else
-			ti->use++;
-	}
+	tt = __find_target_type(name);
+	if (tt && !try_module_get(tt->module))
+		tt = NULL;
 
 	up_read(&_lock);
-	return ti;
+	return tt;
 }
 
 static void load_module(const char *name)
@@ -62,92 +51,59 @@ static void load_module(const char *name)
 
 struct target_type *dm_get_target_type(const char *name)
 {
-	struct tt_internal *ti = get_target_type(name);
+	struct target_type *tt = get_target_type(name);
 
-	if (!ti) {
+	if (!tt) {
 		load_module(name);
-		ti = get_target_type(name);
+		tt = get_target_type(name);
 	}
 
-	return ti ? &ti->tt : NULL;
+	return tt;
 }
 
-void dm_put_target_type(struct target_type *t)
+void dm_put_target_type(struct target_type *tt)
 {
-	struct tt_internal *ti = (struct tt_internal *) t;
-
 	down_read(&_lock);
-	if (--ti->use == 0)
-		module_put(ti->tt.module);
-
-	BUG_ON(ti->use < 0);
+	module_put(tt->module);
 	up_read(&_lock);
-
-	return;
-}
-
-static struct tt_internal *alloc_target(struct target_type *t)
-{
-	struct tt_internal *ti = kzalloc(sizeof(*ti), GFP_KERNEL);
-
-	if (ti)
-		ti->tt = *t;
-
-	return ti;
 }
 
-
 int dm_target_iterate(void (*iter_func)(struct target_type *tt,
 					void *param), void *param)
 {
-	struct tt_internal *ti;
+	struct target_type *tt;
 
 	down_read(&_lock);
-	list_for_each_entry (ti, &_targets, list)
-		iter_func(&ti->tt, param);
+	list_for_each_entry(tt, &_targets, list)
+		iter_func(tt, param);
 	up_read(&_lock);
 
 	return 0;
 }
 
-int dm_register_target(struct target_type *t)
+int dm_register_target(struct target_type *tt)
 {
 	int rv = 0;
-	struct tt_internal *ti = alloc_target(t);
-
-	if (!ti)
-		return -ENOMEM;
 
 	down_write(&_lock);
-	if (__find_target_type(t->name))
+	if (__find_target_type(tt->name))
 		rv = -EEXIST;
 	else
-		list_add(&ti->list, &_targets);
+		list_add(&tt->list, &_targets);
 
 	up_write(&_lock);
-	if (rv)
-		kfree(ti);
 	return rv;
 }
 
-void dm_unregister_target(struct target_type *t)
+void dm_unregister_target(struct target_type *tt)
 {
-	struct tt_internal *ti;
-
 	down_write(&_lock);
-	if (!(ti = __find_target_type(t->name))) {
-		DMCRIT("Unregistering unrecognised target: %s", t->name);
-		BUG();
-	}
-
-	if (ti->use) {
-		DMCRIT("Attempt to unregister target still in use: %s",
-		       t->name);
+	if (!__find_target_type(tt->name)) {
+		DMCRIT("Unregistering unrecognised target: %s", tt->name);
 		BUG();
 	}
 
-	list_del(&ti->list);
-	kfree(ti);
+	list_del(&tt->list);
 
 	up_write(&_lock);
 }
@@ -156,17 +112,17 @@ void dm_unregister_target(struct target_type *t)
  * io-err: always fails an io, useful for bringing
  * up LVs that have holes in them.
  */
-static int io_err_ctr(struct dm_target *ti, unsigned int argc, char **args)
+static int io_err_ctr(struct dm_target *tt, unsigned int argc, char **args)
 {
 	return 0;
 }
 
-static void io_err_dtr(struct dm_target *ti)
+static void io_err_dtr(struct dm_target *tt)
 {
 	/* empty */
 }
 
-static int io_err_map(struct dm_target *ti, struct bio *bio,
+static int io_err_map(struct dm_target *tt, struct bio *bio,
 		      union map_info *map_context)
 {
 	return -EIO;

drivers/md/dm.c

@@ -99,19 +99,9 @@ union map_info *dm_get_mapinfo(struct bio *bio)
 /*
  * Work processed by per-device workqueue.
  */
-struct dm_wq_req {
-	enum {
-		DM_WQ_FLUSH_DEFERRED,
-	} type;
-	struct work_struct work;
-	struct mapped_device *md;
-	void *context;
-};
-
 struct mapped_device {
 	struct rw_semaphore io_lock;
 	struct mutex suspend_lock;
-	spinlock_t pushback_lock;
 	rwlock_t map_lock;
 	atomic_t holders;
 	atomic_t open_count;
@@ -129,8 +119,9 @@ struct mapped_device {
 	 */
 	atomic_t pending;
 	wait_queue_head_t wait;
+	struct work_struct work;
 	struct bio_list deferred;
-	struct bio_list pushback;
+	spinlock_t deferred_lock;
 
 	/*
 	 * Processing queue (flush/barriers)
@@ -453,7 +444,9 @@ static int queue_io(struct mapped_device *md, struct bio *bio)
 		return 1;
 	}
 
+	spin_lock_irq(&md->deferred_lock);
 	bio_list_add(&md->deferred, bio);
+	spin_unlock_irq(&md->deferred_lock);
 
 	up_write(&md->io_lock);
 	return 0;		/* deferred successfully */
@@ -537,16 +530,14 @@ static void dec_pending(struct dm_io *io, int error)
 		if (io->error == DM_ENDIO_REQUEUE) {
 			/*
 			 * Target requested pushing back the I/O.
-			 * This must be handled before the sleeper on
-			 * suspend queue merges the pushback list.
 			 */
-			spin_lock_irqsave(&md->pushback_lock, flags);
+			spin_lock_irqsave(&md->deferred_lock, flags);
 			if (__noflush_suspending(md))
-				bio_list_add(&md->pushback, io->bio);
+				bio_list_add(&md->deferred, io->bio);
 			else
 				/* noflush suspend was interrupted. */
 				io->error = -EIO;
-			spin_unlock_irqrestore(&md->pushback_lock, flags);
+			spin_unlock_irqrestore(&md->deferred_lock, flags);
 		}
 
 		end_io_acct(io);
@@ -834,20 +825,22 @@ static int __clone_and_map(struct clone_info *ci)
 }
 
 /*
- * Split the bio into several clones.
+ * Split the bio into several clones and submit it to targets.
  */
-static int __split_bio(struct mapped_device *md, struct bio *bio)
+static void __split_and_process_bio(struct mapped_device *md, struct bio *bio)
 {
 	struct clone_info ci;
 	int error = 0;
 
 	ci.map = dm_get_table(md);
-	if (unlikely(!ci.map))
-		return -EIO;
+	if (unlikely(!ci.map)) {
+		bio_io_error(bio);
+		return;
+	}
 	if (unlikely(bio_barrier(bio) && !dm_table_barrier_ok(ci.map))) {
 		dm_table_put(ci.map);
 		bio_endio(bio, -EOPNOTSUPP);
-		return 0;
+		return;
 	}
 	ci.md = md;
 	ci.bio = bio;
@@ -867,8 +860,6 @@ static int __split_bio(struct mapped_device *md, struct bio *bio)
 	/* drop the extra reference count */
 	dec_pending(ci.io, error);
 	dm_table_put(ci.map);
-
-	return 0;
 }
 /*-----------------------------------------------------------------
  * CRUD END
@@ -959,8 +950,9 @@ static int dm_request(struct request_queue *q, struct bio *bio)
 		down_read(&md->io_lock);
 	}
 
-	r = __split_bio(md, bio);
+	__split_and_process_bio(md, bio);
 	up_read(&md->io_lock);
+	return 0;
 
 out_req:
 	if (r < 0)
@@ -1074,6 +1066,8 @@ out:
 
 static struct block_device_operations dm_blk_dops;
 
+static void dm_wq_work(struct work_struct *work);
+
 /*
  * Allocate and initialise a blank device with a given minor.
  */
@@ -1101,7 +1095,7 @@ static struct mapped_device *alloc_dev(int minor)
 
 	init_rwsem(&md->io_lock);
 	mutex_init(&md->suspend_lock);
-	spin_lock_init(&md->pushback_lock);
+	spin_lock_init(&md->deferred_lock);
 	rwlock_init(&md->map_lock);
 	atomic_set(&md->holders, 1);
 	atomic_set(&md->open_count, 0);
@@ -1118,6 +1112,7 @@ static struct mapped_device *alloc_dev(int minor)
 	md->queue->backing_dev_info.congested_fn = dm_any_congested;
 	md->queue->backing_dev_info.congested_data = md;
 	blk_queue_make_request(md->queue, dm_request);
+	blk_queue_ordered(md->queue, QUEUE_ORDERED_DRAIN, NULL);
 	blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY);
 	md->queue->unplug_fn = dm_unplug_all;
 	blk_queue_merge_bvec(md->queue, dm_merge_bvec);
@@ -1140,6 +1135,7 @@ static struct mapped_device *alloc_dev(int minor)
 
 	atomic_set(&md->pending, 0);
 	init_waitqueue_head(&md->wait);
+	INIT_WORK(&md->work, dm_wq_work);
 	init_waitqueue_head(&md->eventq);
 
 	md->disk->major = _major;
@@ -1379,18 +1375,24 @@ void dm_put(struct mapped_device *md)
 }
 EXPORT_SYMBOL_GPL(dm_put);
 
-static int dm_wait_for_completion(struct mapped_device *md)
+static int dm_wait_for_completion(struct mapped_device *md, int interruptible)
 {
 	int r = 0;
+	DECLARE_WAITQUEUE(wait, current);
+
+	dm_unplug_all(md->queue);
+
+	add_wait_queue(&md->wait, &wait);
 
 	while (1) {
-		set_current_state(TASK_INTERRUPTIBLE);
+		set_current_state(interruptible);
 
 		smp_mb();
 		if (!atomic_read(&md->pending))
 			break;
 
-		if (signal_pending(current)) {
+		if (interruptible == TASK_INTERRUPTIBLE &&
+		    signal_pending(current)) {
 			r = -EINTR;
 			break;
 		}
@@ -1399,67 +1401,40 @@ static int dm_wait_for_completion(struct mapped_device *md)
 	}
 	set_current_state(TASK_RUNNING);
 
+	remove_wait_queue(&md->wait, &wait);
+
 	return r;
 }
 
 /*
  * Process the deferred bios
  */
-static void __flush_deferred_io(struct mapped_device *md)
+static void dm_wq_work(struct work_struct *work)
 {
+	struct mapped_device *md = container_of(work, struct mapped_device,
+						work);
 	struct bio *c;
 
-	while ((c = bio_list_pop(&md->deferred))) {
-		if (__split_bio(md, c))
-			bio_io_error(c);
-	}
-
-	clear_bit(DMF_BLOCK_IO, &md->flags);
-}
+	down_write(&md->io_lock);
 
-static void __merge_pushback_list(struct mapped_device *md)
-{
-	unsigned long flags;
+next_bio:
+	spin_lock_irq(&md->deferred_lock);
+	c = bio_list_pop(&md->deferred);
+	spin_unlock_irq(&md->deferred_lock);
 
-	spin_lock_irqsave(&md->pushback_lock, flags);
-	clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
-	bio_list_merge_head(&md->deferred, &md->pushback);
-	bio_list_init(&md->pushback);
-	spin_unlock_irqrestore(&md->pushback_lock, flags);
-}
+	if (c) {
+		__split_and_process_bio(md, c);
+		goto next_bio;
+	}
 
-static void dm_wq_work(struct work_struct *work)
-{
-	struct dm_wq_req *req = container_of(work, struct dm_wq_req, work);
-	struct mapped_device *md = req->md;
+	clear_bit(DMF_BLOCK_IO, &md->flags);
 
-	down_write(&md->io_lock);
-	switch (req->type) {
-	case DM_WQ_FLUSH_DEFERRED:
-		__flush_deferred_io(md);
-		break;
-	default:
-		DMERR("dm_wq_work: unrecognised work type %d", req->type);
-		BUG();
-	}
 	up_write(&md->io_lock);
 }
 
-static void dm_wq_queue(struct mapped_device *md, int type, void *context,
-			struct dm_wq_req *req)
-{
-	req->type = type;
-	req->md = md;
-	req->context = context;
-	INIT_WORK(&req->work, dm_wq_work);
-	queue_work(md->wq, &req->work);
-}
-
-static void dm_queue_flush(struct mapped_device *md, int type, void *context)
+static void dm_queue_flush(struct mapped_device *md)
 {
-	struct dm_wq_req req;
-
-	dm_wq_queue(md, type, context, &req);
+	queue_work(md->wq, &md->work);
 	flush_workqueue(md->wq);
 }
 
@@ -1534,7 +1509,6 @@ static void unlock_fs(struct mapped_device *md)
 int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
 {
 	struct dm_table *map = NULL;
-	DECLARE_WAITQUEUE(wait, current);
 	int r = 0;
 	int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0;
 	int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0;
@@ -1584,28 +1558,22 @@ int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
 	down_write(&md->io_lock);
 	set_bit(DMF_BLOCK_IO, &md->flags);
 
-	add_wait_queue(&md->wait, &wait);
 	up_write(&md->io_lock);
 
-	/* unplug */
-	if (map)
-		dm_table_unplug_all(map);
-
 	/*
 	 * Wait for the already-mapped ios to complete.
 	 */
-	r = dm_wait_for_completion(md);
+	r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE);
 
 	down_write(&md->io_lock);
-	remove_wait_queue(&md->wait, &wait);
 
 	if (noflush)
-		__merge_pushback_list(md);
+		clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
 	up_write(&md->io_lock);
 
 	/* were we interrupted ? */
 	if (r < 0) {
-		dm_queue_flush(md, DM_WQ_FLUSH_DEFERRED, NULL);
+		dm_queue_flush(md);
 
 		unlock_fs(md);
 		goto out; /* pushback list is already flushed, so skip flush */
@@ -1645,7 +1613,7 @@ int dm_resume(struct mapped_device *md)
 	if (r)
 		goto out;
 
-	dm_queue_flush(md, DM_WQ_FLUSH_DEFERRED, NULL);
+	dm_queue_flush(md);
 
 	unlock_fs(md);
 

drivers/md/dm.h

@@ -60,7 +60,7 @@ int dm_table_barrier_ok(struct dm_table *t);
 int dm_target_init(void);
 void dm_target_exit(void);
 struct target_type *dm_get_target_type(const char *name);
-void dm_put_target_type(struct target_type *t);
+void dm_put_target_type(struct target_type *tt);
 int dm_target_iterate(void (*iter_func)(struct target_type *tt,
 					void *param), void *param);
 

include/linux/device-mapper.h

@@ -139,6 +139,9 @@ struct target_type {
 	dm_ioctl_fn ioctl;
 	dm_merge_fn merge;
 	dm_busy_fn busy;
+
+	/* For internal device-mapper use. */
+	struct list_head list;
 };
 
 struct io_restrictions {

include/linux/dm-dirty-log.h

@@ -28,6 +28,9 @@ struct dm_dirty_log_type {
 	const char *name;
 	struct module *module;
 
+	/* For internal device-mapper use */
+	struct list_head list;
+
 	int (*ctr)(struct dm_dirty_log *log, struct dm_target *ti,
 		   unsigned argc, char **argv);
 	void (*dtr)(struct dm_dirty_log *log);
@@ -113,6 +116,16 @@ struct dm_dirty_log_type {
 	 */
 	int (*status)(struct dm_dirty_log *log, status_type_t status_type,
 		      char *result, unsigned maxlen);
+
+	/*
+	 * is_remote_recovering is necessary for cluster mirroring. It provides
+	 * a way to detect recovery on another node, so we aren't writing
+	 * concurrently.  This function is likely to block (when a cluster log
+	 * is used).
+	 *
+	 * Returns: 0, 1
+	 */
+	int (*is_remote_recovering)(struct dm_dirty_log *log, region_t region);
 };
 
 int dm_dirty_log_type_register(struct dm_dirty_log_type *type);