9 files changed, 262 insertions, 47 deletions

diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c
index ab6a61db63c..13956437bc8 100644
--- a/drivers/md/dm-crypt.c
+++ b/drivers/md/dm-crypt.c
@@ -1216,9 +1216,24 @@ error:
 	return -EINVAL;
 }
 
+static int crypt_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
+		       struct bio_vec *biovec, int max_size)
+{
+	struct crypt_config *cc = ti->private;
+	struct request_queue *q = bdev_get_queue(cc->dev->bdev);
+
+	if (!q->merge_bvec_fn)
+		return max_size;
+
+	bvm->bi_bdev = cc->dev->bdev;
+	bvm->bi_sector = cc->start + bvm->bi_sector - ti->begin;
+
+	return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
+}
+
 static struct target_type crypt_target = {
 	.name   = "crypt",
-	.version= {1, 5, 0},
+	.version= {1, 6, 0},
 	.module = THIS_MODULE,
 	.ctr    = crypt_ctr,
 	.dtr    = crypt_dtr,
@@ -1228,6 +1243,7 @@ static struct target_type crypt_target = {
 	.preresume = crypt_preresume,
 	.resume = crypt_resume,
 	.message = crypt_message,
+	.merge  = crypt_merge,
 };
 
 static int __init dm_crypt_init(void)
diff --git a/drivers/md/dm-linear.c b/drivers/md/dm-linear.c
index 17753d80ad2..6449bcdf84c 100644
--- a/drivers/md/dm-linear.c
+++ b/drivers/md/dm-linear.c
@@ -69,13 +69,25 @@ static void linear_dtr(struct dm_target *ti)
 	kfree(lc);
 }
 
-static int linear_map(struct dm_target *ti, struct bio *bio,
-		      union map_info *map_context)
+static sector_t linear_map_sector(struct dm_target *ti, sector_t bi_sector)
 {
-	struct linear_c *lc = (struct linear_c *) ti->private;
+	struct linear_c *lc = ti->private;
+
+	return lc->start + (bi_sector - ti->begin);
+}
+
+static void linear_map_bio(struct dm_target *ti, struct bio *bio)
+{
+	struct linear_c *lc = ti->private;
 
 	bio->bi_bdev = lc->dev->bdev;
-	bio->bi_sector = lc->start + (bio->bi_sector - ti->begin);
+	bio->bi_sector = linear_map_sector(ti, bio->bi_sector);
+}
+
+static int linear_map(struct dm_target *ti, struct bio *bio,
+		      union map_info *map_context)
+{
+	linear_map_bio(ti, bio);
 
 	return DM_MAPIO_REMAPPED;
 }
@@ -114,15 +126,31 @@ static int linear_ioctl(struct dm_target *ti, struct inode *inode,
 	return blkdev_driver_ioctl(bdev->bd_inode, &fake_file, bdev->bd_disk, cmd, arg);
 }
 
+static int linear_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
+			struct bio_vec *biovec, int max_size)
+{
+	struct linear_c *lc = ti->private;
+	struct request_queue *q = bdev_get_queue(lc->dev->bdev);
+
+	if (!q->merge_bvec_fn)
+		return max_size;
+
+	bvm->bi_bdev = lc->dev->bdev;
+	bvm->bi_sector = linear_map_sector(ti, bvm->bi_sector);
+
+	return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
+}
+
 static struct target_type linear_target = {
 	.name   = "linear",
-	.version= {1, 0, 2},
+	.version= {1, 0, 3},
 	.module = THIS_MODULE,
 	.ctr    = linear_ctr,
 	.dtr    = linear_dtr,
 	.map    = linear_map,
 	.status = linear_status,
 	.ioctl  = linear_ioctl,
+	.merge  = linear_merge,
 };
 
 int __init dm_linear_init(void)
diff --git a/drivers/md/dm-log.c b/drivers/md/dm-log.c
index 67a6f31b7fc..5b48478c79f 100644
--- a/drivers/md/dm-log.c
+++ b/drivers/md/dm-log.c
@@ -831,7 +831,7 @@ static struct dm_dirty_log_type _disk_type = {
 	.status = disk_status,
 };
 
-int __init dm_dirty_log_init(void)
+static int __init dm_dirty_log_init(void)
 {
 	int r;
 
@@ -848,7 +848,7 @@ int __init dm_dirty_log_init(void)
 	return r;
 }
 
-void __exit dm_dirty_log_exit(void)
+static void __exit dm_dirty_log_exit(void)
 {
 	dm_dirty_log_type_unregister(&_disk_type);
 	dm_dirty_log_type_unregister(&_core_type);
diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
index 9f7302d4878..fea966d66f9 100644
--- a/drivers/md/dm-mpath.c
+++ b/drivers/md/dm-mpath.c
@@ -525,8 +525,10 @@ static int parse_path_selector(struct arg_set *as, struct priority_group *pg,
 	}
 
 	r = read_param(_params, shift(as), &ps_argc, &ti->error);
-	if (r)
+	if (r) {
+		dm_put_path_selector(pst);
 		return -EINVAL;
+	}
 
 	r = pst->create(&pg->ps, ps_argc, as->argv);
 	if (r) {
@@ -623,8 +625,10 @@ static struct priority_group *parse_priority_group(struct arg_set *as,
 		struct pgpath *pgpath;
 		struct arg_set path_args;
 
-		if (as->argc < nr_params)
+		if (as->argc < nr_params) {
+			ti->error = "not enough path parameters";
 			goto bad;
+		}
 
 		path_args.argc = nr_params;
 		path_args.argv = as->argv;
@@ -867,7 +871,7 @@ static int reinstate_path(struct pgpath *pgpath)
 	if (pgpath->path.is_active)
 		goto out;
 
-	if (!pgpath->pg->ps.type) {
+	if (!pgpath->pg->ps.type->reinstate_path) {
 		DMWARN("Reinstate path not supported by path selector %s",
 		       pgpath->pg->ps.type->name);
 		r = -EINVAL;
diff --git a/drivers/md/dm-snap.c b/drivers/md/dm-snap.c
index 1ba8a47d61b..6e5528aecc9 100644
--- a/drivers/md/dm-snap.c
+++ b/drivers/md/dm-snap.c
@@ -40,6 +40,11 @@
  */
 #define SNAPSHOT_PAGES (((1UL << 20) >> PAGE_SHIFT) ? : 1)
 
+/*
+ * The size of the mempool used to track chunks in use.
+ */
+#define MIN_IOS 256
+
 static struct workqueue_struct *ksnapd;
 static void flush_queued_bios(struct work_struct *work);
 
@@ -91,7 +96,63 @@ struct dm_snap_pending_exception {
  */
 static struct kmem_cache *exception_cache;
 static struct kmem_cache *pending_cache;
-static mempool_t *pending_pool;
+
+struct dm_snap_tracked_chunk {
+	struct hlist_node node;
+	chunk_t chunk;
+};
+
+static struct kmem_cache *tracked_chunk_cache;
+
+static struct dm_snap_tracked_chunk *track_chunk(struct dm_snapshot *s,
+						 chunk_t chunk)
+{
+	struct dm_snap_tracked_chunk *c = mempool_alloc(s->tracked_chunk_pool,
+							GFP_NOIO);
+	unsigned long flags;
+
+	c->chunk = chunk;
+
+	spin_lock_irqsave(&s->tracked_chunk_lock, flags);
+	hlist_add_head(&c->node,
+		       &s->tracked_chunk_hash[DM_TRACKED_CHUNK_HASH(chunk)]);
+	spin_unlock_irqrestore(&s->tracked_chunk_lock, flags);
+
+	return c;
+}
+
+static void stop_tracking_chunk(struct dm_snapshot *s,
+				struct dm_snap_tracked_chunk *c)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&s->tracked_chunk_lock, flags);
+	hlist_del(&c->node);
+	spin_unlock_irqrestore(&s->tracked_chunk_lock, flags);
+
+	mempool_free(c, s->tracked_chunk_pool);
+}
+
+static int __chunk_is_tracked(struct dm_snapshot *s, chunk_t chunk)
+{
+	struct dm_snap_tracked_chunk *c;
+	struct hlist_node *hn;
+	int found = 0;
+
+	spin_lock_irq(&s->tracked_chunk_lock);
+
+	hlist_for_each_entry(c, hn,
+	    &s->tracked_chunk_hash[DM_TRACKED_CHUNK_HASH(chunk)], node) {
+		if (c->chunk == chunk) {
+			found = 1;
+			break;
+		}
+	}
+
+	spin_unlock_irq(&s->tracked_chunk_lock);
+
+	return found;
+}
 
 /*
  * One of these per registered origin, held in the snapshot_origins hash
@@ -302,14 +363,19 @@ static void free_exception(struct dm_snap_exception *e)
 	kmem_cache_free(exception_cache, e);
 }
 
-static struct dm_snap_pending_exception *alloc_pending_exception(void)
+static struct dm_snap_pending_exception *alloc_pending_exception(struct dm_snapshot *s)
 {
-	return mempool_alloc(pending_pool, GFP_NOIO);
+	struct dm_snap_pending_exception *pe = mempool_alloc(s->pending_pool,
+							     GFP_NOIO);
+
+	pe->snap = s;
+
+	return pe;
 }
 
 static void free_pending_exception(struct dm_snap_pending_exception *pe)
 {
-	mempool_free(pe, pending_pool);
+	mempool_free(pe, pe->snap->pending_pool);
 }
 
 static void insert_completed_exception(struct dm_snapshot *s,
@@ -482,6 +548,7 @@ static int set_chunk_size(struct dm_snapshot *s, const char *chunk_size_arg,
 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;
@@ -564,11 +631,30 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		goto bad5;
 	}
 
+	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;
+	}
+
+	s->tracked_chunk_pool = mempool_create_slab_pool(MIN_IOS,
+							 tracked_chunk_cache);
+	if (!s->tracked_chunk_pool) {
+		ti->error = "Could not allocate tracked_chunk mempool for "
+			    "tracking reads";
+		goto bad_tracked_chunk_pool;
+	}
+
+	for (i = 0; i < DM_TRACKED_CHUNK_HASH_SIZE; i++)
+		INIT_HLIST_HEAD(&s->tracked_chunk_hash[i]);
+
+	spin_lock_init(&s->tracked_chunk_lock);
+
 	/* Metadata must only be loaded into one table at once */
 	r = s->store.read_metadata(&s->store);
 	if (r < 0) {
 		ti->error = "Failed to read snapshot metadata";
-		goto bad6;
+		goto bad_load_and_register;
 	} else if (r > 0) {
 		s->valid = 0;
 		DMWARN("Snapshot is marked invalid.");
@@ -582,7 +668,7 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	if (register_snapshot(s)) {
 		r = -EINVAL;
 		ti->error = "Cannot register snapshot origin";
-		goto bad6;
+		goto bad_load_and_register;
 	}
 
 	ti->private = s;
@@ -590,6 +676,12 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 
 	return 0;
 
+ bad_load_and_register:
+	mempool_destroy(s->tracked_chunk_pool);
+
+ bad_tracked_chunk_pool:
+	mempool_destroy(s->pending_pool);
+
  bad6:
 	dm_kcopyd_client_destroy(s->kcopyd_client);
 
@@ -624,6 +716,9 @@ static void __free_exceptions(struct dm_snapshot *s)
 
 static void snapshot_dtr(struct dm_target *ti)
 {
+#ifdef CONFIG_DM_DEBUG
+	int i;
+#endif
 	struct dm_snapshot *s = ti->private;
 
 	flush_workqueue(ksnapd);
@@ -632,8 +727,17 @@ static void snapshot_dtr(struct dm_target *ti)
 	/* After this returns there can be no new kcopyd jobs. */
 	unregister_snapshot(s);
 
+#ifdef CONFIG_DM_DEBUG
+	for (i = 0; i < DM_TRACKED_CHUNK_HASH_SIZE; i++)
+		BUG_ON(!hlist_empty(&s->tracked_chunk_hash[i]));
+#endif
+
+	mempool_destroy(s->tracked_chunk_pool);
+
 	__free_exceptions(s);
 
+	mempool_destroy(s->pending_pool);
+
 	dm_put_device(ti, s->origin);
 	dm_put_device(ti, s->cow);
 
@@ -772,6 +876,13 @@ static void pending_complete(struct dm_snap_pending_exception *pe, int success)
 	}
 
 	/*
+	 * Check for conflicting reads. This is extremely improbable,
+	 * so yield() is sufficient and there is no need for a wait queue.
+	 */
+	while (__chunk_is_tracked(s, pe->e.old_chunk))
+		yield();
+
+	/*
 	 * Add a proper exception, and remove the
 	 * in-flight exception from the list.
 	 */
@@ -873,7 +984,7 @@ __find_pending_exception(struct dm_snapshot *s, struct bio *bio)
 	 * to hold the lock while we do this.
 	 */
 	up_write(&s->lock);
-	pe = alloc_pending_exception();
+	pe = alloc_pending_exception(s);
 	down_write(&s->lock);
 
 	if (!s->valid) {
@@ -893,7 +1004,6 @@ __find_pending_exception(struct dm_snapshot *s, struct bio *bio)
 	bio_list_init(&pe->snapshot_bios);
 	pe->primary_pe = NULL;
 	atomic_set(&pe->ref_count, 0);
-	pe->snap = s;
 	pe->started = 0;
 
 	if (s->store.prepare_exception(&s->store, &pe->e)) {
@@ -974,14 +1084,10 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio,
 			start_copy(pe);
 			goto out;
 		}
-	} else
-		/*
-		 * FIXME: this read path scares me because we
-		 * always use the origin when we have a pending
-		 * exception.  However I can't think of a
-		 * situation where this is wrong - ejt.
-		 */
+	} else {
 		bio->bi_bdev = s->origin->bdev;
+		map_context->ptr = track_chunk(s, chunk);
+	}
 
  out_unlock:
 	up_write(&s->lock);
@@ -989,6 +1095,18 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio,
 	return r;
 }
 
+static int snapshot_end_io(struct dm_target *ti, struct bio *bio,
+			   int error, union map_info *map_context)
+{
+	struct dm_snapshot *s = ti->private;
+	struct dm_snap_tracked_chunk *c = map_context->ptr;
+
+	if (c)
+		stop_tracking_chunk(s, c);
+
+	return 0;
+}
+
 static void snapshot_resume(struct dm_target *ti)
 {
 	struct dm_snapshot *s = ti->private;
@@ -1266,6 +1384,7 @@ static struct target_type snapshot_target = {
 	.ctr     = snapshot_ctr,
 	.dtr     = snapshot_dtr,
 	.map     = snapshot_map,
+	.end_io  = snapshot_end_io,
 	.resume  = snapshot_resume,
 	.status  = snapshot_status,
 };
@@ -1306,9 +1425,9 @@ static int __init dm_snapshot_init(void)
 		goto bad4;
 	}
 
-	pending_pool = mempool_create_slab_pool(128, pending_cache);
-	if (!pending_pool) {
-		DMERR("Couldn't create pending pool.");
+	tracked_chunk_cache = KMEM_CACHE(dm_snap_tracked_chunk, 0);
+	if (!tracked_chunk_cache) {
+		DMERR("Couldn't create cache to track chunks in use.");
 		r = -ENOMEM;
 		goto bad5;
 	}
@@ -1317,13 +1436,13 @@ static int __init dm_snapshot_init(void)
 	if (!ksnapd) {
 		DMERR("Failed to create ksnapd workqueue.");
 		r = -ENOMEM;
-		goto bad6;
+		goto bad_pending_pool;
 	}
 
 	return 0;
 
-      bad6:
-	mempool_destroy(pending_pool);
+      bad_pending_pool:
+	kmem_cache_destroy(tracked_chunk_cache);
       bad5:
 	kmem_cache_destroy(pending_cache);
       bad4:
@@ -1352,9 +1471,9 @@ static void __exit dm_snapshot_exit(void)
 		DMERR("origin unregister failed %d", r);
 
 	exit_origin_hash();
-	mempool_destroy(pending_pool);
 	kmem_cache_destroy(pending_cache);
 	kmem_cache_destroy(exception_cache);
+	kmem_cache_destroy(tracked_chunk_cache);
 }
 
 /* Module hooks */
diff --git a/drivers/md/dm-snap.h b/drivers/md/dm-snap.h
index 24f9fb73b98..292c15609ae 100644
--- a/drivers/md/dm-snap.h
+++ b/drivers/md/dm-snap.h
@@ -130,6 +130,10 @@ struct exception_store {
 	void *context;
 };
 
+#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;
@@ -157,6 +161,8 @@ struct dm_snapshot {
 	/* The last percentage we notified */
 	int last_percent;
 
+	mempool_t *pending_pool;
+
 	struct exception_table pending;
 	struct exception_table complete;
 
@@ -174,6 +180,11 @@ struct dm_snapshot {
 	/* 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];
 };
 
 /*
diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c
index 94116eaf470..798e468103b 100644
--- a/drivers/md/dm-table.c
+++ b/drivers/md/dm-table.c
@@ -506,14 +506,13 @@ void dm_set_device_limits(struct dm_target *ti, struct block_device *bdev)
 	rs->max_sectors =
 		min_not_zero(rs->max_sectors, q->max_sectors);
 
-	/* FIXME: Device-Mapper on top of RAID-0 breaks because DM
-	 *        currently doesn't honor MD's merge_bvec_fn routine.
-	 *        In this case, we'll force DM to use PAGE_SIZE or
-	 *        smaller I/O, just to be safe. A better fix is in the
-	 *        works, but add this for the time being so it will at
-	 *        least operate correctly.
+	/*
+	 * Check if merge fn is supported.
+	 * If not we'll force DM to use PAGE_SIZE or
+	 * smaller I/O, just to be safe.
 	 */
-	if (q->merge_bvec_fn)
+
+	if (q->merge_bvec_fn && !ti->type->merge)
 		rs->max_sectors =
 			min_not_zero(rs->max_sectors,
 				     (unsigned int) (PAGE_SIZE >> 9));
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index 372369b1cc2..bca448e1187 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -37,8 +37,8 @@ static DEFINE_SPINLOCK(_minor_lock);
 struct dm_io {
 	struct mapped_device *md;
 	int error;
-	struct bio *bio;
 	atomic_t io_count;
+	struct bio *bio;
 	unsigned long start_time;
 };
 
@@ -829,6 +829,49 @@ static int __split_bio(struct mapped_device *md, struct bio *bio)
  * CRUD END
  *---------------------------------------------------------------*/
 
+static int dm_merge_bvec(struct request_queue *q,
+			 struct bvec_merge_data *bvm,
+			 struct bio_vec *biovec)
+{
+	struct mapped_device *md = q->queuedata;
+	struct dm_table *map = dm_get_table(md);
+	struct dm_target *ti;
+	sector_t max_sectors;
+	int max_size;
+
+	if (unlikely(!map))
+		return 0;
+
+	ti = dm_table_find_target(map, bvm->bi_sector);
+
+	/*
+	 * Find maximum amount of I/O that won't need splitting
+	 */
+	max_sectors = min(max_io_len(md, bvm->bi_sector, ti),
+			  (sector_t) BIO_MAX_SECTORS);
+	max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size;
+	if (max_size < 0)
+		max_size = 0;
+
+	/*
+	 * merge_bvec_fn() returns number of bytes
+	 * it can accept at this offset
+	 * max is precomputed maximal io size
+	 */
+	if (max_size && ti->type->merge)
+		max_size = ti->type->merge(ti, bvm, biovec, max_size);
+
+	/*
+	 * Always allow an entire first page
+	 */
+	if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT))
+		max_size = biovec->bv_len;
+
+	dm_table_put(map);
+
+	return max_size;
+}
+
 /*
  * The request function that just remaps the bio built up by
  * dm_merge_bvec.
@@ -1032,6 +1075,7 @@ static struct mapped_device *alloc_dev(int minor)
 	blk_queue_make_request(md->queue, dm_request);
 	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);
 
 	md->io_pool = mempool_create_slab_pool(MIN_IOS, _io_cache);
 	if (!md->io_pool)
diff --git a/drivers/md/dm.h b/drivers/md/dm.h
index 8c03b634e62..1e59a0b0a78 100644
--- a/drivers/md/dm.h
+++ b/drivers/md/dm.h
@@ -100,12 +100,6 @@ int dm_lock_for_deletion(struct mapped_device *md);
 
 void dm_kobject_uevent(struct mapped_device *md);
 
-/*
- * Dirty log
- */
-int dm_dirty_log_init(void);
-void dm_dirty_log_exit(void);
-
 int dm_kcopyd_init(void);
 void dm_kcopyd_exit(void);