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: (80 commits)
  dm snapshot: use merge origin if snapshot invalid
  dm snapshot: report merge failure in status
  dm snapshot: merge consecutive chunks together
  dm snapshot: trigger exceptions in remaining snapshots during merge
  dm snapshot: delay merging a chunk until writes to it complete
  dm snapshot: queue writes to chunks being merged
  dm snapshot: add merging
  dm snapshot: permit only one merge at once
  dm snapshot: support barriers in snapshot merge target
  dm snapshot: avoid allocating exceptions in merge
  dm snapshot: rework writing to origin
  dm snapshot: add merge target
  dm exception store: add merge specific methods
  dm snapshot: create function for chunk_is_tracked wait
  dm snapshot: make bio optional in __origin_write
  dm mpath: reject messages when device is suspended
  dm: export suspended state to targets
  dm: rename dm_suspended to dm_suspended_md
  dm: swap target postsuspend call and setting suspended flag
  dm crypt: add plain64 iv
  ...

Documentation/device-mapper/snapshot.txt

@@ -8,13 +8,19 @@ the block device which are also writable without interfering with the
 original content;
 *) To create device "forks", i.e. multiple different versions of the
 same data stream.
+*) To merge a snapshot of a block device back into the snapshot's origin
+device.
 
+In the first two cases, dm copies only the chunks of data that get
+changed and uses a separate copy-on-write (COW) block device for
+storage.
 
-In both cases, dm copies only the chunks of data that get changed and
-uses a separate copy-on-write (COW) block device for storage.
+For snapshot merge the contents of the COW storage are merged back into
+the origin device.
 
 
-There are two dm targets available: snapshot and snapshot-origin.
+There are three dm targets available:
+snapshot, snapshot-origin, and snapshot-merge.
 
 *) snapshot-origin <origin>
 
@@ -40,8 +46,25 @@ The difference is that for transient snapshots less metadata must be
 saved on disk - they can be kept in memory by the kernel.
 
 
-How this is used by LVM2
-========================
+* snapshot-merge <origin> <COW device> <persistent> <chunksize>
+
+takes the same table arguments as the snapshot target except it only
+works with persistent snapshots.  This target assumes the role of the
+"snapshot-origin" target and must not be loaded if the "snapshot-origin"
+is still present for <origin>.
+
+Creates a merging snapshot that takes control of the changed chunks
+stored in the <COW device> of an existing snapshot, through a handover
+procedure, and merges these chunks back into the <origin>.  Once merging
+has started (in the background) the <origin> may be opened and the merge
+will continue while I/O is flowing to it.  Changes to the <origin> are
+deferred until the merging snapshot's corresponding chunk(s) have been
+merged.  Once merging has started the snapshot device, associated with
+the "snapshot" target, will return -EIO when accessed.
+
+
+How snapshot is used by LVM2
+============================
 When you create the first LVM2 snapshot of a volume, four dm devices are used:
 
 1) a device containing the original mapping table of the source volume;
@@ -72,3 +95,30 @@ brw-------  1 root root 254, 12 29 ago 18:15 /dev/mapper/volumeGroup-snap-cow
 brw-------  1 root root 254, 13 29 ago 18:15 /dev/mapper/volumeGroup-snap
 brw-------  1 root root 254, 10 29 ago 18:14 /dev/mapper/volumeGroup-base
 
+
+How snapshot-merge is used by LVM2
+==================================
+A merging snapshot assumes the role of the "snapshot-origin" while
+merging.  As such the "snapshot-origin" is replaced with
+"snapshot-merge".  The "-real" device is not changed and the "-cow"
+device is renamed to <origin name>-cow to aid LVM2's cleanup of the
+merging snapshot after it completes.  The "snapshot" that hands over its
+COW device to the "snapshot-merge" is deactivated (unless using lvchange
+--refresh); but if it is left active it will simply return I/O errors.
+
+A snapshot will merge into its origin with the following command:
+
+lvconvert --merge volumeGroup/snap
+
+we'll now have this situation:
+
+# dmsetup table|grep volumeGroup
+
+volumeGroup-base-real: 0 2097152 linear 8:19 384
+volumeGroup-base-cow: 0 204800 linear 8:19 2097536
+volumeGroup-base: 0 2097152 snapshot-merge 254:11 254:12 P 16
+
+# ls -lL /dev/mapper/volumeGroup-*
+brw-------  1 root root 254, 11 29 ago 18:15 /dev/mapper/volumeGroup-base-real
+brw-------  1 root root 254, 12 29 ago 18:16 /dev/mapper/volumeGroup-base-cow
+brw-------  1 root root 254, 10 29 ago 18:16 /dev/mapper/volumeGroup-base

drivers/md/dm-exception-store.c

@@ -172,7 +172,8 @@ int dm_exception_store_set_chunk_size(struct dm_exception_store *store,
 	}
 
 	/* Validate the chunk size against the device block size */
-	if (chunk_size % (bdev_logical_block_size(store->cow->bdev) >> 9)) {
+	if (chunk_size %
+	    (bdev_logical_block_size(dm_snap_cow(store->snap)->bdev) >> 9)) {
 		*error = "Chunk size is not a multiple of device blocksize";
 		return -EINVAL;
 	}
@@ -190,6 +191,7 @@ int dm_exception_store_set_chunk_size(struct dm_exception_store *store,
 }
 
 int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
+			      struct dm_snapshot *snap,
 			      unsigned *args_used,
 			      struct dm_exception_store **store)
 {
@@ -198,7 +200,7 @@ int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
 	struct dm_exception_store *tmp_store;
 	char persistent;
 
-	if (argc < 3) {
+	if (argc < 2) {
 		ti->error = "Insufficient exception store arguments";
 		return -EINVAL;
 	}
@@ -209,14 +211,15 @@ int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
 		return -ENOMEM;
 	}
 
-	persistent = toupper(*argv[1]);
+	persistent = toupper(*argv[0]);
 	if (persistent == 'P')
 		type = get_type("P");
 	else if (persistent == 'N')
 		type = get_type("N");
 	else {
 		ti->error = "Persistent flag is not P or N";
-		return -EINVAL;
+		r = -EINVAL;
+		goto bad_type;
 	}
 
 	if (!type) {
@@ -226,32 +229,23 @@ int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
 	}
 
 	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;
-	}
+	tmp_store->snap = snap;
 
-	r = set_chunk_size(tmp_store, argv[2], &ti->error);
+	r = set_chunk_size(tmp_store, argv[1], &ti->error);
 	if (r)
-		goto bad_ctr;
+		goto bad;
 
 	r = type->ctr(tmp_store, 0, NULL);
 	if (r) {
 		ti->error = "Exception store type constructor failed";
-		goto bad_ctr;
+		goto bad;
 	}
 
-	*args_used = 3;
+	*args_used = 2;
 	*store = tmp_store;
 	return 0;
 
-bad_ctr:
-	dm_put_device(ti, tmp_store->cow);
-bad_cow:
+bad:
 	put_type(type);
 bad_type:
 	kfree(tmp_store);
@@ -262,7 +256,6 @@ 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);
 }

drivers/md/dm-exception-store.h

@@ -26,7 +26,7 @@ typedef sector_t chunk_t;
  * of chunks that follow contiguously.  Remaining bits hold the number of the
  * chunk within the device.
  */
-struct dm_snap_exception {
+struct dm_exception {
 	struct list_head hash_list;
 
 	chunk_t old_chunk;
@@ -64,16 +64,33 @@ struct dm_exception_store_type {
 	 * Find somewhere to store the next exception.
 	 */
 	int (*prepare_exception) (struct dm_exception_store *store,
-				  struct dm_snap_exception *e);
+				  struct dm_exception *e);
 
 	/*
 	 * Update the metadata with this exception.
 	 */
 	void (*commit_exception) (struct dm_exception_store *store,
-				  struct dm_snap_exception *e,
+				  struct dm_exception *e,
 				  void (*callback) (void *, int success),
 				  void *callback_context);
 
+	/*
+	 * Returns 0 if the exception store is empty.
+	 *
+	 * If there are exceptions still to be merged, sets
+	 * *last_old_chunk and *last_new_chunk to the most recent
+	 * still-to-be-merged chunk and returns the number of
+	 * consecutive previous ones.
+	 */
+	int (*prepare_merge) (struct dm_exception_store *store,
+			      chunk_t *last_old_chunk, chunk_t *last_new_chunk);
+
+	/*
+	 * Clear the last n exceptions.
+	 * nr_merged must be <= the value returned by prepare_merge.
+	 */
+	int (*commit_merge) (struct dm_exception_store *store, int nr_merged);
+
 	/*
 	 * The snapshot is invalid, note this in the metadata.
 	 */
@@ -86,19 +103,19 @@ struct dm_exception_store_type {
 	/*
 	 * Return how full the snapshot is.
 	 */
-	void (*fraction_full) (struct dm_exception_store *store,
-			       sector_t *numerator,
-			       sector_t *denominator);
+	void (*usage) (struct dm_exception_store *store,
+		       sector_t *total_sectors, sector_t *sectors_allocated,
+		       sector_t *metadata_sectors);
 
 	/* For internal device-mapper use only. */
 	struct list_head list;
 };
 
+struct dm_snapshot;
+
 struct dm_exception_store {
 	struct dm_exception_store_type *type;
-	struct dm_target *ti;
-
-	struct dm_dev *cow;
+	struct dm_snapshot *snap;
 
 	/* Size of data blocks saved - must be a power of 2 */
 	unsigned chunk_size;
@@ -108,6 +125,11 @@ struct dm_exception_store {
 	void *context;
 };
 
+/*
+ * Obtain the cow device used by a given snapshot.
+ */
+struct dm_dev *dm_snap_cow(struct dm_snapshot *snap);
+
 /*
  * Funtions to manipulate consecutive chunks
  */
@@ -120,18 +142,25 @@ static inline chunk_t dm_chunk_number(chunk_t chunk)
 	return chunk & (chunk_t)((1ULL << DM_CHUNK_NUMBER_BITS) - 1ULL);
 }
 
-static inline unsigned dm_consecutive_chunk_count(struct dm_snap_exception *e)
+static inline unsigned dm_consecutive_chunk_count(struct dm_exception *e)
 {
 	return e->new_chunk >> DM_CHUNK_NUMBER_BITS;
 }
 
-static inline void dm_consecutive_chunk_count_inc(struct dm_snap_exception *e)
+static inline void dm_consecutive_chunk_count_inc(struct dm_exception *e)
 {
 	e->new_chunk += (1ULL << DM_CHUNK_NUMBER_BITS);
 
 	BUG_ON(!dm_consecutive_chunk_count(e));
 }
 
+static inline void dm_consecutive_chunk_count_dec(struct dm_exception *e)
+{
+	BUG_ON(!dm_consecutive_chunk_count(e));
+
+	e->new_chunk -= (1ULL << DM_CHUNK_NUMBER_BITS);
+}
+
 #  else
 #    define DM_CHUNK_CONSECUTIVE_BITS 0
 
@@ -140,12 +169,16 @@ static inline chunk_t dm_chunk_number(chunk_t chunk)
 	return chunk;
 }
 
-static inline unsigned dm_consecutive_chunk_count(struct dm_snap_exception *e)
+static inline unsigned dm_consecutive_chunk_count(struct dm_exception *e)
 {
 	return 0;
 }
 
-static inline void dm_consecutive_chunk_count_inc(struct dm_snap_exception *e)
+static inline void dm_consecutive_chunk_count_inc(struct dm_exception *e)
+{
+}
+
+static inline void dm_consecutive_chunk_count_dec(struct dm_exception *e)
 {
 }
 
@@ -162,7 +195,7 @@ static inline sector_t get_dev_size(struct block_device *bdev)
 static inline chunk_t sector_to_chunk(struct dm_exception_store *store,
 				      sector_t sector)
 {
-	return (sector & ~store->chunk_mask) >> store->chunk_shift;
+	return sector >> store->chunk_shift;
 }
 
 int dm_exception_store_type_register(struct dm_exception_store_type *type);
@@ -173,6 +206,7 @@ int dm_exception_store_set_chunk_size(struct dm_exception_store *store,
 				      char **error);
 
 int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
+			      struct dm_snapshot *snap,
 			      unsigned *args_used,
 			      struct dm_exception_store **store);
 void dm_exception_store_destroy(struct dm_exception_store *store);

drivers/md/dm-io.c

@@ -5,6 +5,8 @@
  * This file is released under the GPL.
  */
 
+#include "dm.h"
+
 #include <linux/device-mapper.h>
 
 #include <linux/bio.h>
@@ -14,12 +16,19 @@
 #include <linux/slab.h>
 #include <linux/dm-io.h>
 
+#define DM_MSG_PREFIX "io"
+
+#define DM_IO_MAX_REGIONS	BITS_PER_LONG
+
 struct dm_io_client {
 	mempool_t *pool;
 	struct bio_set *bios;
 };
 
-/* FIXME: can we shrink this ? */
+/*
+ * Aligning 'struct io' reduces the number of bits required to store
+ * its address.  Refer to store_io_and_region_in_bio() below.
+ */
 struct io {
 	unsigned long error_bits;
 	unsigned long eopnotsupp_bits;
@@ -28,7 +37,9 @@ struct io {
 	struct dm_io_client *client;
 	io_notify_fn callback;
 	void *context;
-};
+} __attribute__((aligned(DM_IO_MAX_REGIONS)));
+
+static struct kmem_cache *_dm_io_cache;
 
 /*
  * io contexts are only dynamically allocated for asynchronous
@@ -53,7 +64,7 @@ struct dm_io_client *dm_io_client_create(unsigned num_pages)
 	if (!client)
 		return ERR_PTR(-ENOMEM);
 
-	client->pool = mempool_create_kmalloc_pool(ios, sizeof(struct io));
+	client->pool = mempool_create_slab_pool(ios, _dm_io_cache);
 	if (!client->pool)
 		goto bad;
 
@@ -88,18 +99,29 @@ EXPORT_SYMBOL(dm_io_client_destroy);
 
 /*-----------------------------------------------------------------
  * We need to keep track of which region a bio is doing io for.
- * In order to save a memory allocation we store this the last
- * bvec which we know is unused (blech).
- * XXX This is ugly and can OOPS with some configs... find another way.
+ * To avoid a memory allocation to store just 5 or 6 bits, we
+ * ensure the 'struct io' pointer is aligned so enough low bits are
+ * always zero and then combine it with the region number directly in
+ * bi_private.
  *---------------------------------------------------------------*/
-static inline void bio_set_region(struct bio *bio, unsigned region)
+static void store_io_and_region_in_bio(struct bio *bio, struct io *io,
+				       unsigned region)
 {
-	bio->bi_io_vec[bio->bi_max_vecs].bv_len = region;
+	if (unlikely(!IS_ALIGNED((unsigned long)io, DM_IO_MAX_REGIONS))) {
+		DMCRIT("Unaligned struct io pointer %p", io);
+		BUG();
+	}
+
+	bio->bi_private = (void *)((unsigned long)io | region);
 }
 
-static inline unsigned bio_get_region(struct bio *bio)
+static void retrieve_io_and_region_from_bio(struct bio *bio, struct io **io,
+				       unsigned *region)
 {
-	return bio->bi_io_vec[bio->bi_max_vecs].bv_len;
+	unsigned long val = (unsigned long)bio->bi_private;
+
+	*io = (void *)(val & -(unsigned long)DM_IO_MAX_REGIONS);
+	*region = val & (DM_IO_MAX_REGIONS - 1);
 }
 
 /*-----------------------------------------------------------------
@@ -140,10 +162,8 @@ static void endio(struct bio *bio, int error)
 	/*
 	 * The bio destructor in bio_put() may use the io object.
 	 */
-	io = bio->bi_private;
-	region = bio_get_region(bio);
+	retrieve_io_and_region_from_bio(bio, &io, &region);
 
-	bio->bi_max_vecs++;
 	bio_put(bio);
 
 	dec_count(io, region, error);
@@ -243,7 +263,10 @@ static void vm_dp_init(struct dpages *dp, void *data)
 
 static void dm_bio_destructor(struct bio *bio)
 {
-	struct io *io = bio->bi_private;
+	unsigned region;
+	struct io *io;
+
+	retrieve_io_and_region_from_bio(bio, &io, &region);
 
 	bio_free(bio, io->client->bios);
 }
@@ -286,26 +309,23 @@ static void do_region(int rw, unsigned region, struct dm_io_region *where,
 	unsigned num_bvecs;
 	sector_t remaining = where->count;
 
-	while (remaining) {
+	/*
+	 * where->count may be zero if rw holds a write barrier and we
+	 * need to send a zero-sized barrier.
+	 */
+	do {
 		/*
-		 * Allocate a suitably sized-bio: we add an extra
-		 * bvec for bio_get/set_region() and decrement bi_max_vecs
-		 * to hide it from bio_add_page().
+		 * Allocate a suitably sized-bio.
 		 */
 		num_bvecs = dm_sector_div_up(remaining,
 					     (PAGE_SIZE >> SECTOR_SHIFT));
-		num_bvecs = 1 + min_t(int, bio_get_nr_vecs(where->bdev),
-				      num_bvecs);
-		if (unlikely(num_bvecs > BIO_MAX_PAGES))
-			num_bvecs = BIO_MAX_PAGES;
+		num_bvecs = min_t(int, bio_get_nr_vecs(where->bdev), num_bvecs);
 		bio = bio_alloc_bioset(GFP_NOIO, num_bvecs, io->client->bios);
 		bio->bi_sector = where->sector + (where->count - remaining);
 		bio->bi_bdev = where->bdev;
 		bio->bi_end_io = endio;
-		bio->bi_private = io;
 		bio->bi_destructor = dm_bio_destructor;
-		bio->bi_max_vecs--;
-		bio_set_region(bio, region);
+		store_io_and_region_in_bio(bio, io, region);
 
 		/*
 		 * Try and add as many pages as possible.
@@ -323,7 +343,7 @@ static void do_region(int rw, unsigned region, struct dm_io_region *where,
 
 		atomic_inc(&io->count);
 		submit_bio(rw, bio);
-	}
+	} while (remaining);
 }
 
 static void dispatch_io(int rw, unsigned int num_regions,
@@ -333,6 +353,8 @@ static void dispatch_io(int rw, unsigned int num_regions,
 	int i;
 	struct dpages old_pages = *dp;
 
+	BUG_ON(num_regions > DM_IO_MAX_REGIONS);
+
 	if (sync)
 		rw |= (1 << BIO_RW_SYNCIO) | (1 << BIO_RW_UNPLUG);
 
@@ -342,7 +364,7 @@ static void dispatch_io(int rw, unsigned int num_regions,
 	 */
 	for (i = 0; i < num_regions; i++) {
 		*dp = old_pages;
-		if (where[i].count)
+		if (where[i].count || (rw & (1 << BIO_RW_BARRIER)))
 			do_region(rw, i, where + i, dp, io);
 	}
 
@@ -357,7 +379,14 @@ 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)
 {
-	struct io io;
+	/*
+	 * 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));
 
 	if (num_regions > 1 && (rw & RW_MASK) != WRITE) {
 		WARN_ON(1);
@@ -365,33 +394,33 @@ static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 	}
 
 retry:
-	io.error_bits = 0;
-	io.eopnotsupp_bits = 0;
-	atomic_set(&io.count, 1); /* see dispatch_io() */
-	io.sleeper = current;
-	io.client = client;
+	io->error_bits = 0;
+	io->eopnotsupp_bits = 0;
+	atomic_set(&io->count, 1); /* see dispatch_io() */
+	io->sleeper = current;
+	io->client = client;
 
-	dispatch_io(rw, num_regions, where, dp, &io, 1);
+	dispatch_io(rw, num_regions, where, dp, io, 1);
 
 	while (1) {
 		set_current_state(TASK_UNINTERRUPTIBLE);
 
-		if (!atomic_read(&io.count))
+		if (!atomic_read(&io->count))
 			break;
 
 		io_schedule();
 	}
 	set_current_state(TASK_RUNNING);
 
-	if (io.eopnotsupp_bits && (rw & (1 << BIO_RW_BARRIER))) {
+	if (io->eopnotsupp_bits && (rw & (1 << BIO_RW_BARRIER))) {
 		rw &= ~(1 << BIO_RW_BARRIER);
 		goto retry;
 	}
 
 	if (error_bits)
-		*error_bits = io.error_bits;
+		*error_bits = io->error_bits;
 
-	return io.error_bits ? -EIO : 0;
+	return io->error_bits ? -EIO : 0;
 }
 
 static int async_io(struct dm_io_client *client, unsigned int num_regions,
@@ -472,3 +501,18 @@ int dm_io(struct dm_io_request *io_req, unsigned num_regions,
 			&dp, io_req->notify.fn, io_req->notify.context);
 }
 EXPORT_SYMBOL(dm_io);
+
+int __init dm_io_init(void)
+{
+	_dm_io_cache = KMEM_CACHE(io, 0);
+	if (!_dm_io_cache)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void dm_io_exit(void)
+{
+	kmem_cache_destroy(_dm_io_cache);
+	_dm_io_cache = NULL;
+}

drivers/md/dm-ioctl.c

@@ -56,6 +56,11 @@ static void dm_hash_remove_all(int keep_open_devices);
  */
 static DECLARE_RWSEM(_hash_lock);
 
+/*
+ * Protects use of mdptr to obtain hash cell name and uuid from mapped device.
+ */
+static DEFINE_MUTEX(dm_hash_cells_mutex);
+
 static void init_buckets(struct list_head *buckets)
 {
 	unsigned int i;
@@ -206,7 +211,9 @@ static int dm_hash_insert(const char *name, const char *uuid, struct mapped_devi
 		list_add(&cell->uuid_list, _uuid_buckets + hash_str(uuid));
 	}
 	dm_get(md);
+	mutex_lock(&dm_hash_cells_mutex);
 	dm_set_mdptr(md, cell);
+	mutex_unlock(&dm_hash_cells_mutex);
 	up_write(&_hash_lock);
 
 	return 0;
@@ -224,9 +231,11 @@ static void __hash_remove(struct hash_cell *hc)
 	/* remove from the dev hash */
 	list_del(&hc->uuid_list);
 	list_del(&hc->name_list);
+	mutex_lock(&dm_hash_cells_mutex);
 	dm_set_mdptr(hc->md, NULL);
+	mutex_unlock(&dm_hash_cells_mutex);
 
-	table = dm_get_table(hc->md);
+	table = dm_get_live_table(hc->md);
 	if (table) {
 		dm_table_event(table);
 		dm_table_put(table);
@@ -321,13 +330,15 @@ static int dm_hash_rename(uint32_t cookie, const char *old, const char *new)
 	 */
 	list_del(&hc->name_list);
 	old_name = hc->name;
+	mutex_lock(&dm_hash_cells_mutex);
 	hc->name = new_name;
+	mutex_unlock(&dm_hash_cells_mutex);
 	list_add(&hc->name_list, _name_buckets + hash_str(new_name));
 
 	/*
 	 * Wake up any dm event waiters.
 	 */
-	table = dm_get_table(hc->md);
+	table = dm_get_live_table(hc->md);
 	if (table) {
 		dm_table_event(table);
 		dm_table_put(table);
@@ -512,8 +523,6 @@ static int list_versions(struct dm_ioctl *param, size_t param_size)
 	return 0;
 }
 
-
-
 static int check_name(const char *name)
 {
 	if (strchr(name, '/')) {
@@ -524,6 +533,40 @@ static int check_name(const char *name)
 	return 0;
 }
 
+/*
+ * On successful return, the caller must not attempt to acquire
+ * _hash_lock without first calling dm_table_put, because dm_table_destroy
+ * waits for this dm_table_put and could be called under this lock.
+ */
+static struct dm_table *dm_get_inactive_table(struct mapped_device *md)
+{
+	struct hash_cell *hc;
+	struct dm_table *table = NULL;
+
+	down_read(&_hash_lock);
+	hc = dm_get_mdptr(md);
+	if (!hc || hc->md != md) {
+		DMWARN("device has been removed from the dev hash table.");
+		goto out;
+	}
+
+	table = hc->new_map;
+	if (table)
+		dm_table_get(table);
+
+out:
+	up_read(&_hash_lock);
+
+	return table;
+}
+
+static struct dm_table *dm_get_live_or_inactive_table(struct mapped_device *md,
+						      struct dm_ioctl *param)
+{
+	return (param->flags & DM_QUERY_INACTIVE_TABLE_FLAG) ?
+		dm_get_inactive_table(md) : dm_get_live_table(md);
+}
+
 /*
  * Fills in a dm_ioctl structure, ready for sending back to
  * userland.
@@ -536,7 +579,7 @@ static int __dev_status(struct mapped_device *md, struct dm_ioctl *param)
 	param->flags &= ~(DM_SUSPEND_FLAG | DM_READONLY_FLAG |
 			  DM_ACTIVE_PRESENT_FLAG);
 
-	if (dm_suspended(md))
+	if (dm_suspended_md(md))
 		param->flags |= DM_SUSPEND_FLAG;
 
 	param->dev = huge_encode_dev(disk_devt(disk));
@@ -548,18 +591,30 @@ static int __dev_status(struct mapped_device *md, struct dm_ioctl *param)
 	 */
 	param->open_count = dm_open_count(md);
 
-	if (get_disk_ro(disk))
-		param->flags |= DM_READONLY_FLAG;
-
 	param->event_nr = dm_get_event_nr(md);
+	param->target_count = 0;
 
-	table = dm_get_table(md);
+	table = dm_get_live_table(md);
 	if (table) {
-		param->flags |= DM_ACTIVE_PRESENT_FLAG;
-		param->target_count = dm_table_get_num_targets(table);
+		if (!(param->flags & DM_QUERY_INACTIVE_TABLE_FLAG)) {
+			if (get_disk_ro(disk))
+				param->flags |= DM_READONLY_FLAG;
+			param->target_count = dm_table_get_num_targets(table);
+		}
 		dm_table_put(table);
-	} else
-		param->target_count = 0;
+
+		param->flags |= DM_ACTIVE_PRESENT_FLAG;
+	}
+
+	if (param->flags & DM_QUERY_INACTIVE_TABLE_FLAG) {
+		table = dm_get_inactive_table(md);
+		if (table) {
+			if (!(dm_table_get_mode(table) & FMODE_WRITE))
+				param->flags |= DM_READONLY_FLAG;
+			param->target_count = dm_table_get_num_targets(table);
+			dm_table_put(table);
+		}
+	}
 
 	return 0;
 }
@@ -634,9 +689,9 @@ static struct mapped_device *find_device(struct dm_ioctl *param)
 		 * Sneakily write in both the name and the uuid
 		 * while we have the cell.
 		 */
-		strncpy(param->name, hc->name, sizeof(param->name));
+		strlcpy(param->name, hc->name, sizeof(param->name));
 		if (hc->uuid)
-			strncpy(param->uuid, hc->uuid, sizeof(param->uuid)-1);
+			strlcpy(param->uuid, hc->uuid, sizeof(param->uuid));
 		else
 			param->uuid[0] = '\0';
 
@@ -784,7 +839,7 @@ static int do_suspend(struct dm_ioctl *param)
 	if (param->flags & DM_NOFLUSH_FLAG)
 		suspend_flags |= DM_SUSPEND_NOFLUSH_FLAG;
 
-	if (!dm_suspended(md))
+	if (!dm_suspended_md(md))
 		r = dm_suspend(md, suspend_flags);
 
 	if (!r)
@@ -800,7 +855,7 @@ static int do_resume(struct dm_ioctl *param)
 	unsigned suspend_flags = DM_SUSPEND_LOCKFS_FLAG;
 	struct hash_cell *hc;
 	struct mapped_device *md;
-	struct dm_table *new_map;
+	struct dm_table *new_map, *old_map = NULL;
 
 	down_write(&_hash_lock);
 
@@ -826,14 +881,14 @@ static int do_resume(struct dm_ioctl *param)
 			suspend_flags &= ~DM_SUSPEND_LOCKFS_FLAG;
 		if (param->flags & DM_NOFLUSH_FLAG)
 			suspend_flags |= DM_SUSPEND_NOFLUSH_FLAG;
-		if (!dm_suspended(md))
+		if (!dm_suspended_md(md))
 			dm_suspend(md, suspend_flags);
 
-		r = dm_swap_table(md, new_map);
-		if (r) {
+		old_map = dm_swap_table(md, new_map);
+		if (IS_ERR(old_map)) {
 			dm_table_destroy(new_map);
 			dm_put(md);
-			return r;
+			return PTR_ERR(old_map);
 		}
 
 		if (dm_table_get_mode(new_map) & FMODE_WRITE)
@@ -842,9 +897,11 @@ static int do_resume(struct dm_ioctl *param)
 			set_disk_ro(dm_disk(md), 1);
 	}
 
-	if (dm_suspended(md))
+	if (dm_suspended_md(md))
 		r = dm_resume(md);
 
+	if (old_map)
+		dm_table_destroy(old_map);
 
 	if (!r) {
 		dm_kobject_uevent(md, KOBJ_CHANGE, param->event_nr);
@@ -982,7 +1039,7 @@ static int dev_wait(struct dm_ioctl *param, size_t param_size)
 	if (r)
 		goto out;
 
-	table = dm_get_table(md);
+	table = dm_get_live_or_inactive_table(md, param);
 	if (table) {
 		retrieve_status(table, param, param_size);
 		dm_table_put(table);
@@ -1215,7 +1272,7 @@ static int table_deps(struct dm_ioctl *param, size_t param_size)
 	if (r)
 		goto out;
 
-	table = dm_get_table(md);
+	table = dm_get_live_or_inactive_table(md, param);
 	if (table) {
 		retrieve_deps(table, param, param_size);
 		dm_table_put(table);
@@ -1244,13 +1301,13 @@ static int table_status(struct dm_ioctl *param, size_t param_size)
 	if (r)
 		goto out;
 
-	table = dm_get_table(md);
+	table = dm_get_live_or_inactive_table(md, param);
 	if (table) {
 		retrieve_status(table, param, param_size);
 		dm_table_put(table);
 	}
 
- out:
+out:
 	dm_put(md);
 	return r;
 }
@@ -1288,10 +1345,15 @@ static int target_message(struct dm_ioctl *param, size_t param_size)
 		goto out;
 	}
 
-	table = dm_get_table(md);
+	table = dm_get_live_table(md);
 	if (!table)
 		goto out_argv;
 
+	if (dm_deleting_md(md)) {
+		r = -ENXIO;
+		goto out_table;
+	}
+
 	ti = dm_table_find_target(table, tmsg->sector);
 	if (!dm_target_is_valid(ti)) {
 		DMWARN("Target message sector outside device.");
@@ -1303,6 +1365,7 @@ static int target_message(struct dm_ioctl *param, size_t param_size)
 		r = -EINVAL;
 	}
 
+ out_table:
 	dm_table_put(table);
  out_argv:
 	kfree(argv);
@@ -1582,8 +1645,7 @@ int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid)
 	if (!md)
 		return -ENXIO;
 
-	dm_get(md);
-	down_read(&_hash_lock);
+	mutex_lock(&dm_hash_cells_mutex);
 	hc = dm_get_mdptr(md);
 	if (!hc || hc->md != md) {
 		r = -ENXIO;
@@ -1596,8 +1658,7 @@ int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid)
 		strcpy(uuid, hc->uuid ? : "");
 
 out:
-	up_read(&_hash_lock);
-	dm_put(md);
+	mutex_unlock(&dm_hash_cells_mutex);
 
 	return r;
 }

drivers/md/dm-kcopyd.c

@@ -450,7 +450,10 @@ static void dispatch_job(struct kcopyd_job *job)
 {
 	struct dm_kcopyd_client *kc = job->kc;
 	atomic_inc(&kc->nr_jobs);
-	push(&kc->pages_jobs, job);
+	if (unlikely(!job->source.count))
+		push(&kc->complete_jobs, job);
+	else
+		push(&kc->pages_jobs, job);
 	wake(kc);
 }
 

drivers/md/dm-log.c

@@ -145,8 +145,9 @@ int dm_dirty_log_type_unregister(struct dm_dirty_log_type *type)
 EXPORT_SYMBOL(dm_dirty_log_type_unregister);
 
 struct dm_dirty_log *dm_dirty_log_create(const char *type_name,
-					 struct dm_target *ti,
-					 unsigned int argc, char **argv)
+			struct dm_target *ti,
+			int (*flush_callback_fn)(struct dm_target *ti),
+			unsigned int argc, char **argv)
 {
 	struct dm_dirty_log_type *type;
 	struct dm_dirty_log *log;
@@ -161,6 +162,7 @@ struct dm_dirty_log *dm_dirty_log_create(const char *type_name,
 		return NULL;
 	}
 
+	log->flush_callback_fn = flush_callback_fn;
 	log->type = type;
 	if (type->ctr(log, ti, argc, argv)) {
 		kfree(log);
@@ -208,7 +210,9 @@ struct log_header {
 
 struct log_c {
 	struct dm_target *ti;
-	int touched;
+	int touched_dirtied;
+	int touched_cleaned;
+	int flush_failed;
 	uint32_t region_size;
 	unsigned int region_count;
 	region_t sync_count;
@@ -233,6 +237,7 @@ struct log_c {
 	 * Disk log fields
 	 */
 	int log_dev_failed;
+	int log_dev_flush_failed;
 	struct dm_dev *log_dev;
 	struct log_header header;
 
@@ -253,14 +258,14 @@ static inline void log_set_bit(struct log_c *l,
 			       uint32_t *bs, unsigned bit)
 {
 	ext2_set_bit(bit, (unsigned long *) bs);
-	l->touched = 1;
+	l->touched_cleaned = 1;
 }
 
 static inline void log_clear_bit(struct log_c *l,
 				 uint32_t *bs, unsigned bit)
 {
 	ext2_clear_bit(bit, (unsigned long *) bs);
-	l->touched = 1;
+	l->touched_dirtied = 1;
 }
 
 /*----------------------------------------------------------------
@@ -287,6 +292,19 @@ static int rw_header(struct log_c *lc, int rw)
 	return dm_io(&lc->io_req, 1, &lc->header_location, NULL);
 }
 
+static int flush_header(struct log_c *lc)
+{
+	struct dm_io_region null_location = {
+		.bdev = lc->header_location.bdev,
+		.sector = 0,
+		.count = 0,
+	};
+
+	lc->io_req.bi_rw = WRITE_BARRIER;
+
+	return dm_io(&lc->io_req, 1, &null_location, NULL);
+}
+
 static int read_header(struct log_c *log)
 {
 	int r;
@@ -378,7 +396,9 @@ static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
 	}
 
 	lc->ti = ti;
-	lc->touched = 0;
+	lc->touched_dirtied = 0;
+	lc->touched_cleaned = 0;
+	lc->flush_failed = 0;
 	lc->region_size = region_size;
 	lc->region_count = region_count;
 	lc->sync = sync;
@@ -406,6 +426,7 @@ static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
 	} else {
 		lc->log_dev = dev;
 		lc->log_dev_failed = 0;
+		lc->log_dev_flush_failed = 0;
 		lc->header_location.bdev = lc->log_dev->bdev;
 		lc->header_location.sector = 0;
 
@@ -614,6 +635,11 @@ static int disk_resume(struct dm_dirty_log *log)
 
 	/* write the new header */
 	r = rw_header(lc, WRITE);
+	if (!r) {
+		r = flush_header(lc);
+		if (r)
+			lc->log_dev_flush_failed = 1;
+	}
 	if (r) {
 		DMWARN("%s: Failed to write header on dirty region log device",
 		       lc->log_dev->name);
@@ -656,18 +682,40 @@ static int core_flush(struct dm_dirty_log *log)
 
 static int disk_flush(struct dm_dirty_log *log)
 {
-	int r;
-	struct log_c *lc = (struct log_c *) log->context;
+	int r, i;
+	struct log_c *lc = log->context;
 
 	/* only write if the log has changed */
-	if (!lc->touched)
+	if (!lc->touched_cleaned && !lc->touched_dirtied)
 		return 0;
 
+	if (lc->touched_cleaned && log->flush_callback_fn &&
+	    log->flush_callback_fn(lc->ti)) {
+		/*
+		 * At this point it is impossible to determine which
+		 * regions are clean and which are dirty (without
+		 * re-reading the log off disk). So mark all of them
+		 * dirty.
+		 */
+		lc->flush_failed = 1;
+		for (i = 0; i < lc->region_count; i++)
+			log_clear_bit(lc, lc->clean_bits, i);
+	}
+
 	r = rw_header(lc, WRITE);
 	if (r)
 		fail_log_device(lc);
-	else
-		lc->touched = 0;
+	else {
+		if (lc->touched_dirtied) {
+			r = flush_header(lc);
+			if (r) {
+				lc->log_dev_flush_failed = 1;
+				fail_log_device(lc);
+			} else
+				lc->touched_dirtied = 0;
+		}
+		lc->touched_cleaned = 0;
+	}
 
 	return r;
 }
@@ -681,7 +729,8 @@ static void core_mark_region(struct dm_dirty_log *log, region_t region)
 static void core_clear_region(struct dm_dirty_log *log, region_t region)
 {
 	struct log_c *lc = (struct log_c *) log->context;
-	log_set_bit(lc, lc->clean_bits, region);
+	if (likely(!lc->flush_failed))
+		log_set_bit(lc, lc->clean_bits, region);
 }
 
 static int core_get_resync_work(struct dm_dirty_log *log, region_t *region)
@@ -762,7 +811,9 @@ static int disk_status(struct dm_dirty_log *log, status_type_t status,
 	switch(status) {
 	case STATUSTYPE_INFO:
 		DMEMIT("3 %s %s %c", log->type->name, lc->log_dev->name,
-		       lc->log_dev_failed ? 'D' : 'A');
+		       lc->log_dev_flush_failed ? 'F' :
+		       lc->log_dev_failed ? 'D' :
+		       'A');
 		break;
 
 	case STATUSTYPE_TABLE:

drivers/md/dm-mpath.c

@@ -93,6 +93,10 @@ struct multipath {
 	 * can resubmit bios on error.
 	 */
 	mempool_t *mpio_pool;
+
+	struct mutex work_mutex;
+
+	unsigned suspended;	/* Don't create new I/O internally when set. */
 };
 
 /*
@@ -198,6 +202,7 @@ static struct multipath *alloc_multipath(struct dm_target *ti)
 		m->queue_io = 1;
 		INIT_WORK(&m->process_queued_ios, process_queued_ios);
 		INIT_WORK(&m->trigger_event, trigger_event);
+		mutex_init(&m->work_mutex);
 		m->mpio_pool = mempool_create_slab_pool(MIN_IOS, _mpio_cache);
 		if (!m->mpio_pool) {
 			kfree(m);
@@ -885,13 +890,18 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc,
 	return r;
 }
 
-static void multipath_dtr(struct dm_target *ti)
+static void flush_multipath_work(void)
 {
-	struct multipath *m = (struct multipath *) ti->private;
-
 	flush_workqueue(kmpath_handlerd);
 	flush_workqueue(kmultipathd);
 	flush_scheduled_work();
+}
+
+static void multipath_dtr(struct dm_target *ti)
+{
+	struct multipath *m = ti->private;
+
+	flush_multipath_work();
 	free_multipath(m);
 }
 
@@ -1261,6 +1271,16 @@ static void multipath_presuspend(struct dm_target *ti)
 	queue_if_no_path(m, 0, 1);
 }
 
+static void multipath_postsuspend(struct dm_target *ti)
+{
+	struct multipath *m = ti->private;
+
+	mutex_lock(&m->work_mutex);
+	m->suspended = 1;
+	flush_multipath_work();
+	mutex_unlock(&m->work_mutex);
+}
+
 /*
  * Restore the queue_if_no_path setting.
  */
@@ -1269,6 +1289,10 @@ static void multipath_resume(struct dm_target *ti)
 	struct multipath *m = (struct multipath *) ti->private;
 	unsigned long flags;
 
+	mutex_lock(&m->work_mutex);
+	m->suspended = 0;
+	mutex_unlock(&m->work_mutex);
+
 	spin_lock_irqsave(&m->lock, flags);
 	m->queue_if_no_path = m->saved_queue_if_no_path;
 	spin_unlock_irqrestore(&m->lock, flags);
@@ -1397,51 +1421,71 @@ static int multipath_status(struct dm_target *ti, status_type_t type,
 
 static int multipath_message(struct dm_target *ti, unsigned argc, char **argv)
 {
-	int r;
+	int r = -EINVAL;
 	struct dm_dev *dev;
 	struct multipath *m = (struct multipath *) ti->private;
 	action_fn action;
 
+	mutex_lock(&m->work_mutex);
+
+	if (m->suspended) {
+		r = -EBUSY;
+		goto out;
+	}
+
+	if (dm_suspended(ti)) {
+		r = -EBUSY;
+		goto out;
+	}
+
 	if (argc == 1) {
-		if (!strnicmp(argv[0], MESG_STR("queue_if_no_path")))
-			return queue_if_no_path(m, 1, 0);
-		else if (!strnicmp(argv[0], MESG_STR("fail_if_no_path")))
-			return queue_if_no_path(m, 0, 0);
+		if (!strnicmp(argv[0], MESG_STR("queue_if_no_path"))) {
+			r = queue_if_no_path(m, 1, 0);
+			goto out;
+		} else if (!strnicmp(argv[0], MESG_STR("fail_if_no_path"))) {
+			r = queue_if_no_path(m, 0, 0);
+			goto out;
+		}
 	}
 
-	if (argc != 2)
-		goto error;
+	if (argc != 2) {
+		DMWARN("Unrecognised multipath message received.");
+		goto out;
+	}
 
-	if (!strnicmp(argv[0], MESG_STR("disable_group")))
-		return bypass_pg_num(m, argv[1], 1);
-	else if (!strnicmp(argv[0], MESG_STR("enable_group")))
-		return bypass_pg_num(m, argv[1], 0);
-	else if (!strnicmp(argv[0], MESG_STR("switch_group")))
-		return switch_pg_num(m, argv[1]);
-	else if (!strnicmp(argv[0], MESG_STR("reinstate_path")))
+	if (!strnicmp(argv[0], MESG_STR("disable_group"))) {
+		r = bypass_pg_num(m, argv[1], 1);
+		goto out;
+	} else if (!strnicmp(argv[0], MESG_STR("enable_group"))) {
+		r = bypass_pg_num(m, argv[1], 0);
+		goto out;
+	} else if (!strnicmp(argv[0], MESG_STR("switch_group"))) {
+		r = switch_pg_num(m, argv[1]);
+		goto out;
+	} else if (!strnicmp(argv[0], MESG_STR("reinstate_path")))
 		action = reinstate_path;
 	else if (!strnicmp(argv[0], MESG_STR("fail_path")))
 		action = fail_path;
-	else
-		goto error;
+	else {
+		DMWARN("Unrecognised multipath message received.");
+		goto out;
+	}
 
 	r = dm_get_device(ti, argv[1], ti->begin, ti->len,
 			  dm_table_get_mode(ti->table), &dev);
 	if (r) {
 		DMWARN("message: error getting device %s",
 		       argv[1]);
-		return -EINVAL;
+		goto out;
 	}
 
 	r = action_dev(m, dev, action);
 
 	dm_put_device(ti, dev);
 
+out:
+	mutex_unlock(&m->work_mutex);
 	return r;
-
-error:
-	DMWARN("Unrecognised multipath message received.");
-	return -EINVAL;
 }
 
 static int multipath_ioctl(struct dm_target *ti, unsigned int cmd,
@@ -1567,13 +1611,14 @@ out:
  *---------------------------------------------------------------*/
 static struct target_type multipath_target = {
 	.name = "multipath",
-	.version = {1, 1, 0},
+	.version = {1, 1, 1},
 	.module = THIS_MODULE,
 	.ctr = multipath_ctr,
 	.dtr = multipath_dtr,
 	.map_rq = multipath_map,
 	.rq_end_io = multipath_end_io,
 	.presuspend = multipath_presuspend,
+	.postsuspend = multipath_postsuspend,
 	.resume = multipath_resume,
 	.status = multipath_status,
 	.message = multipath_message,

drivers/md/dm-region-hash.c

@@ -79,6 +79,11 @@ struct dm_region_hash {
 	struct list_head recovered_regions;
 	struct list_head failed_recovered_regions;
 
+	/*
+	 * If there was a barrier failure no regions can be marked clean.
+	 */
+	int barrier_failure;
+
 	void *context;
 	sector_t target_begin;
 
@@ -211,6 +216,7 @@ struct dm_region_hash *dm_region_hash_create(
 	INIT_LIST_HEAD(&rh->quiesced_regions);
 	INIT_LIST_HEAD(&rh->recovered_regions);
 	INIT_LIST_HEAD(&rh->failed_recovered_regions);
+	rh->barrier_failure = 0;
 
 	rh->region_pool = mempool_create_kmalloc_pool(MIN_REGIONS,
 						      sizeof(struct dm_region));
@@ -377,8 +383,6 @@ static void complete_resync_work(struct dm_region *reg, int success)
 /* dm_rh_mark_nosync
  * @ms
  * @bio
- * @done
- * @error
  *
  * The bio was written on some mirror(s) but failed on other mirror(s).
  * We can successfully endio the bio but should avoid the region being
@@ -386,8 +390,7 @@ static void complete_resync_work(struct dm_region *reg, int success)
  *
  * This function is _not_ safe in interrupt context!
  */
-void dm_rh_mark_nosync(struct dm_region_hash *rh,
-		       struct bio *bio, unsigned done, int error)
+void dm_rh_mark_nosync(struct dm_region_hash *rh, struct bio *bio)
 {
 	unsigned long flags;
 	struct dm_dirty_log *log = rh->log;
@@ -395,6 +398,11 @@ void dm_rh_mark_nosync(struct dm_region_hash *rh,
 	region_t region = dm_rh_bio_to_region(rh, bio);
 	int recovering = 0;
 
+	if (bio_empty_barrier(bio)) {
+		rh->barrier_failure = 1;
+		return;
+	}
+
 	/* We must inform the log that the sync count has changed. */
 	log->type->set_region_sync(log, region, 0);
 
@@ -419,7 +427,6 @@ void dm_rh_mark_nosync(struct dm_region_hash *rh,
 	BUG_ON(!list_empty(&reg->list));
 	spin_unlock_irqrestore(&rh->region_lock, flags);
 
-	bio_endio(bio, error);
 	if (recovering)
 		complete_resync_work(reg, 0);
 }
@@ -515,8 +522,11 @@ void dm_rh_inc_pending(struct dm_region_hash *rh, struct bio_list *bios)
 {
 	struct bio *bio;
 
-	for (bio = bios->head; bio; bio = bio->bi_next)
+	for (bio = bios->head; bio; bio = bio->bi_next) {
+		if (bio_empty_barrier(bio))
+			continue;
 		rh_inc(rh, dm_rh_bio_to_region(rh, bio));
+	}
 }
 EXPORT_SYMBOL_GPL(dm_rh_inc_pending);
 
@@ -544,7 +554,14 @@ void dm_rh_dec(struct dm_region_hash *rh, region_t region)
 		 */
 
 		/* do nothing for DM_RH_NOSYNC */
-		if (reg->state == DM_RH_RECOVERING) {
+		if (unlikely(rh->barrier_failure)) {
+			/*
+			 * If a write barrier failed some time ago, we
+			 * don't know whether or not this write made it
+			 * to the disk, so we must resync the device.
+			 */
+			reg->state = DM_RH_NOSYNC;
+		} else if (reg->state == DM_RH_RECOVERING) {
 			list_add_tail(&reg->list, &rh->quiesced_regions);
 		} else if (reg->state == DM_RH_DIRTY) {
 			reg->state = DM_RH_CLEAN;

drivers/md/dm-snap-persistent.c

@@ -55,6 +55,8 @@
  */
 #define SNAPSHOT_DISK_VERSION 1
 
+#define NUM_SNAPSHOT_HDR_CHUNKS 1
+
 struct disk_header {
 	uint32_t magic;
 
@@ -120,7 +122,22 @@ struct pstore {
 
 	/*
 	 * The next free chunk for an exception.
+	 *
+	 * When creating exceptions, all the chunks here and above are
+	 * free.  It holds the next chunk to be allocated.  On rare
+	 * occasions (e.g. after a system crash) holes can be left in
+	 * the exception store because chunks can be committed out of
+	 * order.
+	 *
+	 * When merging exceptions, it does not necessarily mean all the
+	 * chunks here and above are free.  It holds the value it would
+	 * have held if all chunks had been committed in order of
+	 * allocation.  Consequently the value may occasionally be
+	 * slightly too low, but since it's only used for 'status' and
+	 * it can never reach its minimum value too early this doesn't
+	 * matter.
 	 */
+
 	chunk_t next_free;
 
 	/*
@@ -214,7 +231,7 @@ static int chunk_io(struct pstore *ps, void *area, chunk_t chunk, int rw,
 		    int metadata)
 {
 	struct dm_io_region where = {
-		.bdev = ps->store->cow->bdev,
+		.bdev = dm_snap_cow(ps->store->snap)->bdev,
 		.sector = ps->store->chunk_size * chunk,
 		.count = ps->store->chunk_size,
 	};
@@ -294,7 +311,8 @@ static int read_header(struct pstore *ps, int *new_snapshot)
 	 */
 	if (!ps->store->chunk_size) {
 		ps->store->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS,
-		    bdev_logical_block_size(ps->store->cow->bdev) >> 9);
+		    bdev_logical_block_size(dm_snap_cow(ps->store->snap)->
+					    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;
@@ -408,6 +426,15 @@ static void write_exception(struct pstore *ps,
 	e->new_chunk = cpu_to_le64(de->new_chunk);
 }
 
+static void clear_exception(struct pstore *ps, uint32_t index)
+{
+	struct disk_exception *e = get_exception(ps, index);
+
+	/* clear it */
+	e->old_chunk = 0;
+	e->new_chunk = 0;
+}
+
 /*
  * Registers the exceptions that are present in the current area.
  * 'full' is filled in to indicate if the area has been
@@ -489,11 +516,23 @@ static struct pstore *get_info(struct dm_exception_store *store)
 	return (struct pstore *) store->context;
 }
 
-static void persistent_fraction_full(struct dm_exception_store *store,
-				     sector_t *numerator, sector_t *denominator)
+static void persistent_usage(struct dm_exception_store *store,
+			     sector_t *total_sectors,
+			     sector_t *sectors_allocated,
+			     sector_t *metadata_sectors)
 {
-	*numerator = get_info(store)->next_free * store->chunk_size;
-	*denominator = get_dev_size(store->cow->bdev);
+	struct pstore *ps = get_info(store);
+
+	*sectors_allocated = ps->next_free * store->chunk_size;
+	*total_sectors = get_dev_size(dm_snap_cow(store->snap)->bdev);
+
+	/*
+	 * First chunk is the fixed header.
+	 * Then there are (ps->current_area + 1) metadata chunks, each one
+	 * separated from the next by ps->exceptions_per_area data chunks.
+	 */
+	*metadata_sectors = (ps->current_area + 1 + NUM_SNAPSHOT_HDR_CHUNKS) *
+			    store->chunk_size;
 }
 
 static void persistent_dtr(struct dm_exception_store *store)
@@ -552,44 +591,40 @@ static int persistent_read_metadata(struct dm_exception_store *store,
 		ps->current_area = 0;
 		zero_memory_area(ps);
 		r = zero_disk_area(ps, 0);
-		if (r) {
+		if (r)
 			DMWARN("zero_disk_area(0) failed");
-			return r;
-		}
-	} else {
-		/*
-		 * Sanity checks.
-		 */
-		if (ps->version != SNAPSHOT_DISK_VERSION) {
-			DMWARN("unable to handle snapshot disk version %d",
-			       ps->version);
-			return -EINVAL;
-		}
+		return r;
+	}
+	/*
+	 * Sanity checks.
+	 */
+	if (ps->version != SNAPSHOT_DISK_VERSION) {
+		DMWARN("unable to handle snapshot disk version %d",
+		       ps->version);
+		return -EINVAL;
+	}
 
-		/*
-		 * Metadata are valid, but snapshot is invalidated
-		 */
-		if (!ps->valid)
-			return 1;
+	/*
+	 * Metadata are valid, but snapshot is invalidated
+	 */
+	if (!ps->valid)
+		return 1;
 
-		/*
-		 * Read the metadata.
-		 */
-		r = read_exceptions(ps, callback, callback_context);
-		if (r)
-			return r;
-	}
+	/*
+	 * Read the metadata.
+	 */
+	r = read_exceptions(ps, callback, callback_context);
 
-	return 0;
+	return r;
 }
 
 static int persistent_prepare_exception(struct dm_exception_store *store,
-					struct dm_snap_exception *e)
+					struct dm_exception *e)
 {
 	struct pstore *ps = get_info(store);
 	uint32_t stride;
 	chunk_t next_free;
-	sector_t size = get_dev_size(store->cow->bdev);
+	sector_t size = get_dev_size(dm_snap_cow(store->snap)->bdev);
 
 	/* Is there enough room ? */
 	if (size < ((ps->next_free + 1) * store->chunk_size))
@@ -611,7 +646,7 @@ static int persistent_prepare_exception(struct dm_exception_store *store,
 }
 
 static void persistent_commit_exception(struct dm_exception_store *store,
-					struct dm_snap_exception *e,
+					struct dm_exception *e,
 					void (*callback) (void *, int success),
 					void *callback_context)
 {
@@ -672,6 +707,85 @@ static void persistent_commit_exception(struct dm_exception_store *store,
 	ps->callback_count = 0;
 }
 
+static int persistent_prepare_merge(struct dm_exception_store *store,
+				    chunk_t *last_old_chunk,
+				    chunk_t *last_new_chunk)
+{
+	struct pstore *ps = get_info(store);
+	struct disk_exception de;
+	int nr_consecutive;
+	int r;
+
+	/*
+	 * When current area is empty, move back to preceding area.
+	 */
+	if (!ps->current_committed) {
+		/*
+		 * Have we finished?
+		 */
+		if (!ps->current_area)
+			return 0;
+
+		ps->current_area--;
+		r = area_io(ps, READ);
+		if (r < 0)
+			return r;
+		ps->current_committed = ps->exceptions_per_area;
+	}
+
+	read_exception(ps, ps->current_committed - 1, &de);
+	*last_old_chunk = de.old_chunk;
+	*last_new_chunk = de.new_chunk;
+
+	/*
+	 * Find number of consecutive chunks within the current area,
+	 * working backwards.
+	 */
+	for (nr_consecutive = 1; nr_consecutive < ps->current_committed;
+	     nr_consecutive++) {
+		read_exception(ps, ps->current_committed - 1 - nr_consecutive,
+			       &de);
+		if (de.old_chunk != *last_old_chunk - nr_consecutive ||
+		    de.new_chunk != *last_new_chunk - nr_consecutive)
+			break;
+	}
+
+	return nr_consecutive;
+}
+
+static int persistent_commit_merge(struct dm_exception_store *store,
+				   int nr_merged)
+{
+	int r, i;
+	struct pstore *ps = get_info(store);
+
+	BUG_ON(nr_merged > ps->current_committed);
+
+	for (i = 0; i < nr_merged; i++)
+		clear_exception(ps, ps->current_committed - 1 - i);
+
+	r = area_io(ps, WRITE);
+	if (r < 0)
+		return r;
+
+	ps->current_committed -= nr_merged;
+
+	/*
+	 * At this stage, only persistent_usage() uses ps->next_free, so
+	 * we make no attempt to keep ps->next_free strictly accurate
+	 * as exceptions may have been committed out-of-order originally.
+	 * Once a snapshot has become merging, we set it to the value it
+	 * would have held had all the exceptions been committed in order.
+	 *
+	 * ps->current_area does not get reduced by prepare_merge() until
+	 * after commit_merge() has removed the nr_merged previous exceptions.
+	 */
+	ps->next_free = (area_location(ps, ps->current_area) - 1) +
+			(ps->current_committed + 1) + NUM_SNAPSHOT_HDR_CHUNKS;
+
+	return 0;
+}
+
 static void persistent_drop_snapshot(struct dm_exception_store *store)
 {
 	struct pstore *ps = get_info(store);
@@ -697,7 +811,7 @@ static int persistent_ctr(struct dm_exception_store *store,
 	ps->area = NULL;
 	ps->zero_area = NULL;
 	ps->header_area = NULL;
-	ps->next_free = 2;	/* skipping the header and first area */
+	ps->next_free = NUM_SNAPSHOT_HDR_CHUNKS + 1; /* header and 1st area */
 	ps->current_committed = 0;
 
 	ps->callback_count = 0;
@@ -726,8 +840,7 @@ static unsigned persistent_status(struct dm_exception_store *store,
 	case STATUSTYPE_INFO:
 		break;
 	case STATUSTYPE_TABLE:
-		DMEMIT(" %s P %llu", store->cow->name,
-		       (unsigned long long)store->chunk_size);
+		DMEMIT(" P %llu", (unsigned long long)store->chunk_size);
 	}
 
 	return sz;
@@ -741,8 +854,10 @@ static struct dm_exception_store_type _persistent_type = {
 	.read_metadata = persistent_read_metadata,
 	.prepare_exception = persistent_prepare_exception,
 	.commit_exception = persistent_commit_exception,
+	.prepare_merge = persistent_prepare_merge,
+	.commit_merge = persistent_commit_merge,
 	.drop_snapshot = persistent_drop_snapshot,
-	.fraction_full = persistent_fraction_full,
+	.usage = persistent_usage,
 	.status = persistent_status,
 };
 
@@ -754,8 +869,10 @@ static struct dm_exception_store_type _persistent_compat_type = {
 	.read_metadata = persistent_read_metadata,
 	.prepare_exception = persistent_prepare_exception,
 	.commit_exception = persistent_commit_exception,
+	.prepare_merge = persistent_prepare_merge,
+	.commit_merge = persistent_commit_merge,
 	.drop_snapshot = persistent_drop_snapshot,
-	.fraction_full = persistent_fraction_full,
+	.usage = persistent_usage,
 	.status = persistent_status,
 };
 

drivers/md/dm-snap-transient.c

@@ -36,10 +36,10 @@ 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 dm_exception *e)
 {
 	struct transient_c *tc = store->context;
-	sector_t size = get_dev_size(store->cow->bdev);
+	sector_t size = get_dev_size(dm_snap_cow(store->snap)->bdev);
 
 	if (size < (tc->next_free + store->chunk_size))
 		return -1;
@@ -51,7 +51,7 @@ static int transient_prepare_exception(struct dm_exception_store *store,
 }
 
 static void transient_commit_exception(struct dm_exception_store *store,
-				       struct dm_snap_exception *e,
+				       struct dm_exception *e,
 				       void (*callback) (void *, int success),
 				       void *callback_context)
 {
@@ -59,11 +59,14 @@ static void transient_commit_exception(struct dm_exception_store *store,
 	callback(callback_context, 1);
 }
 
-static void transient_fraction_full(struct dm_exception_store *store,
-				    sector_t *numerator, sector_t *denominator)
+static void transient_usage(struct dm_exception_store *store,
+			    sector_t *total_sectors,
+			    sector_t *sectors_allocated,
+			    sector_t *metadata_sectors)
 {
-	*numerator = ((struct transient_c *) store->context)->next_free;
-	*denominator = get_dev_size(store->cow->bdev);
+	*sectors_allocated = ((struct transient_c *) store->context)->next_free;
+	*total_sectors = get_dev_size(dm_snap_cow(store->snap)->bdev);
+	*metadata_sectors = 0;
 }
 
 static int transient_ctr(struct dm_exception_store *store,
@@ -91,8 +94,7 @@ static unsigned transient_status(struct dm_exception_store *store,
 	case STATUSTYPE_INFO:
 		break;
 	case STATUSTYPE_TABLE:
-		DMEMIT(" %s N %llu", store->cow->name,
-		       (unsigned long long)store->chunk_size);
+		DMEMIT(" N %llu", (unsigned long long)store->chunk_size);
 	}
 
 	return sz;
@@ -106,7 +108,7 @@ static struct dm_exception_store_type _transient_type = {
 	.read_metadata = transient_read_metadata,
 	.prepare_exception = transient_prepare_exception,
 	.commit_exception = transient_commit_exception,
-	.fraction_full = transient_fraction_full,
+	.usage = transient_usage,
 	.status = transient_status,
 };
 
@@ -118,7 +120,7 @@ static struct dm_exception_store_type _transient_compat_type = {
 	.read_metadata = transient_read_metadata,
 	.prepare_exception = transient_prepare_exception,
 	.commit_exception = transient_commit_exception,
-	.fraction_full = transient_fraction_full,
+	.usage = transient_usage,
 	.status = transient_status,
 };
 

drivers/md/dm-sysfs.c

@@ -59,7 +59,7 @@ static ssize_t dm_attr_uuid_show(struct mapped_device *md, char *buf)
 
 static ssize_t dm_attr_suspended_show(struct mapped_device *md, char *buf)
 {
-	sprintf(buf, "%d\n", dm_suspended(md));
+	sprintf(buf, "%d\n", dm_suspended_md(md));
 
 	return strlen(buf);
 }
@@ -79,6 +79,13 @@ static struct sysfs_ops dm_sysfs_ops = {
 	.show	= dm_attr_show,
 };
 
+/*
+ * The sysfs structure is embedded in md struct, nothing to do here
+ */
+static void dm_sysfs_release(struct kobject *kobj)
+{
+}
+
 /*
  * dm kobject is embedded in mapped_device structure
  * no need to define release function here
@@ -86,6 +93,7 @@ static struct sysfs_ops dm_sysfs_ops = {
 static struct kobj_type dm_ktype = {
 	.sysfs_ops	= &dm_sysfs_ops,
 	.default_attrs	= dm_attrs,
+	.release	= dm_sysfs_release
 };
 
 /*

drivers/md/dm-table.c

@@ -238,6 +238,9 @@ void dm_table_destroy(struct dm_table *t)
 {
 	unsigned int i;
 
+	if (!t)
+		return;
+
 	while (atomic_read(&t->holders))
 		msleep(1);
 	smp_mb();

drivers/md/dm-uevent.c

@@ -139,14 +139,13 @@ void dm_send_uevents(struct list_head *events, struct kobject *kobj)
 		list_del_init(&event->elist);
 
 		/*
-		 * Need to call dm_copy_name_and_uuid from here for now.
-		 * Context of previous var adds and locking used for
-		 * hash_cell not compatable.
+		 * When a device is being removed this copy fails and we
+		 * discard these unsent events.
 		 */
 		if (dm_copy_name_and_uuid(event->md, event->name,
 					  event->uuid)) {
-			DMERR("%s: dm_copy_name_and_uuid() failed",
-			      __func__);
+			DMINFO("%s: skipping sending uevent for lost device",
+			       __func__);
 			goto uevent_free;
 		}
 

drivers/md/dm.h

@@ -88,6 +88,16 @@ int dm_target_iterate(void (*iter_func)(struct target_type *tt,
 
 int dm_split_args(int *argc, char ***argvp, char *input);
 
+/*
+ * Is this mapped_device being deleted?
+ */
+int dm_deleting_md(struct mapped_device *md);
+
+/*
+ * Is this mapped_device suspended?
+ */
+int dm_suspended_md(struct mapped_device *md);
+
 /*
  * The device-mapper can be driven through one of two interfaces;
  * ioctl or filesystem, depending which patch you have applied.
@@ -118,6 +128,9 @@ int dm_lock_for_deletion(struct mapped_device *md);
 void dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
 		       unsigned cookie);
 
+int dm_io_init(void);
+void dm_io_exit(void);
+
 int dm_kcopyd_init(void);
 void dm_kcopyd_exit(void);
 

include/linux/device-mapper.h

@@ -235,7 +235,7 @@ void dm_uevent_add(struct mapped_device *md, struct list_head *elist);
 const char *dm_device_name(struct mapped_device *md);
 int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid);
 struct gendisk *dm_disk(struct mapped_device *md);
-int dm_suspended(struct mapped_device *md);
+int dm_suspended(struct dm_target *ti);
 int dm_noflush_suspending(struct dm_target *ti);
 union map_info *dm_get_mapinfo(struct bio *bio);
 union map_info *dm_get_rq_mapinfo(struct request *rq);
@@ -276,7 +276,7 @@ void dm_table_unplug_all(struct dm_table *t);
 /*
  * Table reference counting.
  */
-struct dm_table *dm_get_table(struct mapped_device *md);
+struct dm_table *dm_get_live_table(struct mapped_device *md);
 void dm_table_get(struct dm_table *t);
 void dm_table_put(struct dm_table *t);
 
@@ -295,8 +295,10 @@ void dm_table_event(struct dm_table *t);
 
 /*
  * The device must be suspended before calling this method.
+ * Returns the previous table, which the caller must destroy.
  */
-int dm_swap_table(struct mapped_device *md, struct dm_table *t);
+struct dm_table *dm_swap_table(struct mapped_device *md,
+			       struct dm_table *t);
 
 /*
  * A wrapper around vmalloc.

include/linux/dm-dirty-log.h

@@ -21,6 +21,7 @@ struct dm_dirty_log_type;
 
 struct dm_dirty_log {
 	struct dm_dirty_log_type *type;
+	int (*flush_callback_fn)(struct dm_target *ti);
 	void *context;
 };
 
@@ -136,8 +137,9 @@ int dm_dirty_log_type_unregister(struct dm_dirty_log_type *type);
  * type->constructor/destructor() directly.
  */
 struct dm_dirty_log *dm_dirty_log_create(const char *type_name,
-					 struct dm_target *ti,
-					 unsigned argc, char **argv);
+			struct dm_target *ti,
+			int (*flush_callback_fn)(struct dm_target *ti),
+			unsigned argc, char **argv);
 void dm_dirty_log_destroy(struct dm_dirty_log *log);
 
 #endif	/* __KERNEL__ */

include/linux/dm-ioctl.h

 /*
  * Copyright (C) 2001 - 2003 Sistina Software (UK) Limited.
- * Copyright (C) 2004 - 2005 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2004 - 2009 Red Hat, Inc. All rights reserved.
  *
  * This file is released under the LGPL.
  */
@@ -266,9 +266,9 @@ enum {
 #define DM_DEV_SET_GEOMETRY	_IOWR(DM_IOCTL, DM_DEV_SET_GEOMETRY_CMD, struct dm_ioctl)
 
 #define DM_VERSION_MAJOR	4
-#define DM_VERSION_MINOR	15
+#define DM_VERSION_MINOR	16
 #define DM_VERSION_PATCHLEVEL	0
-#define DM_VERSION_EXTRA	"-ioctl (2009-04-01)"
+#define DM_VERSION_EXTRA	"-ioctl (2009-11-05)"
 
 /* Status bits */
 #define DM_READONLY_FLAG	(1 << 0) /* In/Out */
@@ -309,4 +309,11 @@ enum {
  */
 #define DM_NOFLUSH_FLAG		(1 << 11) /* In */
 
+/*
+ * If set, any table information returned will relate to the inactive
+ * table instead of the live one.  Always check DM_INACTIVE_PRESENT_FLAG
+ * is set before using the data returned.
+ */
+#define DM_QUERY_INACTIVE_TABLE_FLAG	(1 << 12) /* In */
+
 #endif				/* _LINUX_DM_IOCTL_H */

include/linux/dm-region-hash.h

@@ -78,8 +78,7 @@ void dm_rh_dec(struct dm_region_hash *rh, region_t region);
 /* Delay bios on regions. */
 void dm_rh_delay(struct dm_region_hash *rh, struct bio *bio);
 
-void dm_rh_mark_nosync(struct dm_region_hash *rh,
-		       struct bio *bio, unsigned done, int error);
+void dm_rh_mark_nosync(struct dm_region_hash *rh, struct bio *bio);
 
 /*
  * Region recovery control.