Merge tag 'dm-3.5-changes-1' of git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-dm

Pull device-mapper updates from Alasdair G Kergon:
 "Improve multipath's retrying mechanism in some defined circumstances
  and provide a simple reserve/release mechanism for userspace tools to
  access thin provisioning metadata while the pool is in use."

* tag 'dm-3.5-changes-1' of git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-dm:
  dm thin: provide userspace access to pool metadata
  dm thin: use slab mempools
  dm mpath: allow ioctls to trigger pg init
  dm mpath: delay retry of bypassed pg
  dm mpath: reduce size of struct multipath

Author: torvalds

Documentation/device-mapper/thin-provisioning.txt

@@ -287,6 +287,17 @@ iii) Messages
 	the current transaction id is when you change it with this
 	compare-and-swap message.
 
+    reserve_metadata_snap
+
+        Reserve a copy of the data mapping btree for use by userland.
+        This allows userland to inspect the mappings as they were when
+        this message was executed.  Use the pool's status command to
+        get the root block associated with the metadata snapshot.
+
+    release_metadata_snap
+
+        Release a previously reserved copy of the data mapping btree.
+
 'thin' target
 -------------
 

drivers/md/dm-mpath.c

@@ -18,6 +18,7 @@
 #include <linux/slab.h>
 #include <linux/time.h>
 #include <linux/workqueue.h>
+#include <linux/delay.h>
 #include <scsi/scsi_dh.h>
 #include <linux/atomic.h>
 
@@ -61,11 +62,11 @@ struct multipath {
 	struct list_head list;
 	struct dm_target *ti;
 
-	spinlock_t lock;
-
 	const char *hw_handler_name;
 	char *hw_handler_params;
 
+	spinlock_t lock;
+
 	unsigned nr_priority_groups;
 	struct list_head priority_groups;
 
@@ -81,16 +82,17 @@ struct multipath {
 	struct priority_group *next_pg;	/* Switch to this PG if set */
 	unsigned repeat_count;		/* I/Os left before calling PS again */
 
-	unsigned queue_io;		/* Must we queue all I/O? */
-	unsigned queue_if_no_path;	/* Queue I/O if last path fails? */
-	unsigned saved_queue_if_no_path;/* Saved state during suspension */
+	unsigned queue_io:1;		/* Must we queue all I/O? */
+	unsigned queue_if_no_path:1;	/* Queue I/O if last path fails? */
+	unsigned saved_queue_if_no_path:1; /* Saved state during suspension */
+
 	unsigned pg_init_retries;	/* Number of times to retry pg_init */
 	unsigned pg_init_count;		/* Number of times pg_init called */
 	unsigned pg_init_delay_msecs;	/* Number of msecs before pg_init retry */
 
+	unsigned queue_size;
 	struct work_struct process_queued_ios;
 	struct list_head queued_ios;
-	unsigned queue_size;
 
 	struct work_struct trigger_event;
 
@@ -328,14 +330,18 @@ static void __choose_pgpath(struct multipath *m, size_t nr_bytes)
 	/*
 	 * Loop through priority groups until we find a valid path.
 	 * First time we skip PGs marked 'bypassed'.
-	 * Second time we only try the ones we skipped.
+	 * Second time we only try the ones we skipped, but set
+	 * pg_init_delay_retry so we do not hammer controllers.
 	 */
 	do {
 		list_for_each_entry(pg, &m->priority_groups, list) {
 			if (pg->bypassed == bypassed)
 				continue;
-			if (!__choose_path_in_pg(m, pg, nr_bytes))
+			if (!__choose_path_in_pg(m, pg, nr_bytes)) {
+				if (!bypassed)
+					m->pg_init_delay_retry = 1;
 				return;
+			}
 		}
 	} while (bypassed--);
 
@@ -481,9 +487,6 @@ static void process_queued_ios(struct work_struct *work)
 
 	spin_lock_irqsave(&m->lock, flags);
 
-	if (!m->queue_size)
-		goto out;
-
 	if (!m->current_pgpath)
 		__choose_pgpath(m, 0);
 
@@ -496,7 +499,6 @@ static void process_queued_ios(struct work_struct *work)
 	if (m->pg_init_required && !m->pg_init_in_progress && pgpath)
 		__pg_init_all_paths(m);
 
-out:
 	spin_unlock_irqrestore(&m->lock, flags);
 	if (!must_queue)
 		dispatch_queued_ios(m);
@@ -1517,11 +1519,16 @@ static int multipath_message(struct dm_target *ti, unsigned argc, char **argv)
 static int multipath_ioctl(struct dm_target *ti, unsigned int cmd,
 			   unsigned long arg)
 {
-	struct multipath *m = (struct multipath *) ti->private;
-	struct block_device *bdev = NULL;
-	fmode_t mode = 0;
+	struct multipath *m = ti->private;
+	struct block_device *bdev;
+	fmode_t mode;
 	unsigned long flags;
-	int r = 0;
+	int r;
+
+again:
+	bdev = NULL;
+	mode = 0;
+	r = 0;
 
 	spin_lock_irqsave(&m->lock, flags);
 
@@ -1546,6 +1553,12 @@ static int multipath_ioctl(struct dm_target *ti, unsigned int cmd,
 	if (!r && ti->len != i_size_read(bdev->bd_inode) >> SECTOR_SHIFT)
 		r = scsi_verify_blk_ioctl(NULL, cmd);
 
+	if (r == -EAGAIN && !fatal_signal_pending(current)) {
+		queue_work(kmultipathd, &m->process_queued_ios);
+		msleep(10);
+		goto again;
+	}
+
 	return r ? : __blkdev_driver_ioctl(bdev, mode, cmd, arg);
 }
 
@@ -1643,7 +1656,7 @@ static int multipath_busy(struct dm_target *ti)
  *---------------------------------------------------------------*/
 static struct target_type multipath_target = {
 	.name = "multipath",
-	.version = {1, 3, 0},
+	.version = {1, 4, 0},
 	.module = THIS_MODULE,
 	.ctr = multipath_ctr,
 	.dtr = multipath_dtr,

drivers/md/dm-thin-metadata.c

@@ -1082,31 +1082,155 @@ int dm_pool_get_metadata_transaction_id(struct dm_pool_metadata *pmd,
 	return 0;
 }
 
-static int __get_held_metadata_root(struct dm_pool_metadata *pmd,
-				    dm_block_t *result)
+static int __reserve_metadata_snap(struct dm_pool_metadata *pmd)
+{
+	int r, inc;
+	struct thin_disk_superblock *disk_super;
+	struct dm_block *copy, *sblock;
+	dm_block_t held_root;
+
+	/*
+	 * Copy the superblock.
+	 */
+	dm_sm_inc_block(pmd->metadata_sm, THIN_SUPERBLOCK_LOCATION);
+	r = dm_tm_shadow_block(pmd->tm, THIN_SUPERBLOCK_LOCATION,
+			       &sb_validator, &copy, &inc);
+	if (r)
+		return r;
+
+	BUG_ON(!inc);
+
+	held_root = dm_block_location(copy);
+	disk_super = dm_block_data(copy);
+
+	if (le64_to_cpu(disk_super->held_root)) {
+		DMWARN("Pool metadata snapshot already exists: release this before taking another.");
+
+		dm_tm_dec(pmd->tm, held_root);
+		dm_tm_unlock(pmd->tm, copy);
+		pmd->need_commit = 1;
+
+		return -EBUSY;
+	}
+
+	/*
+	 * Wipe the spacemap since we're not publishing this.
+	 */
+	memset(&disk_super->data_space_map_root, 0,
+	       sizeof(disk_super->data_space_map_root));
+	memset(&disk_super->metadata_space_map_root, 0,
+	       sizeof(disk_super->metadata_space_map_root));
+
+	/*
+	 * Increment the data structures that need to be preserved.
+	 */
+	dm_tm_inc(pmd->tm, le64_to_cpu(disk_super->data_mapping_root));
+	dm_tm_inc(pmd->tm, le64_to_cpu(disk_super->device_details_root));
+	dm_tm_unlock(pmd->tm, copy);
+
+	/*
+	 * Write the held root into the superblock.
+	 */
+	r = dm_bm_write_lock(pmd->bm, THIN_SUPERBLOCK_LOCATION,
+			     &sb_validator, &sblock);
+	if (r) {
+		dm_tm_dec(pmd->tm, held_root);
+		pmd->need_commit = 1;
+		return r;
+	}
+
+	disk_super = dm_block_data(sblock);
+	disk_super->held_root = cpu_to_le64(held_root);
+	dm_bm_unlock(sblock);
+
+	pmd->need_commit = 1;
+
+	return 0;
+}
+
+int dm_pool_reserve_metadata_snap(struct dm_pool_metadata *pmd)
+{
+	int r;
+
+	down_write(&pmd->root_lock);
+	r = __reserve_metadata_snap(pmd);
+	up_write(&pmd->root_lock);
+
+	return r;
+}
+
+static int __release_metadata_snap(struct dm_pool_metadata *pmd)
 {
 	int r;
 	struct thin_disk_superblock *disk_super;
-	struct dm_block *sblock;
+	struct dm_block *sblock, *copy;
+	dm_block_t held_root;
 
 	r = dm_bm_write_lock(pmd->bm, THIN_SUPERBLOCK_LOCATION,
 			     &sb_validator, &sblock);
 	if (r)
 		return r;
 
+	disk_super = dm_block_data(sblock);
+	held_root = le64_to_cpu(disk_super->held_root);
+	disk_super->held_root = cpu_to_le64(0);
+	pmd->need_commit = 1;
+
+	dm_bm_unlock(sblock);
+
+	if (!held_root) {
+		DMWARN("No pool metadata snapshot found: nothing to release.");
+		return -EINVAL;
+	}
+
+	r = dm_tm_read_lock(pmd->tm, held_root, &sb_validator, &copy);
+	if (r)
+		return r;
+
+	disk_super = dm_block_data(copy);
+	dm_sm_dec_block(pmd->metadata_sm, le64_to_cpu(disk_super->data_mapping_root));
+	dm_sm_dec_block(pmd->metadata_sm, le64_to_cpu(disk_super->device_details_root));
+	dm_sm_dec_block(pmd->metadata_sm, held_root);
+
+	return dm_tm_unlock(pmd->tm, copy);
+}
+
+int dm_pool_release_metadata_snap(struct dm_pool_metadata *pmd)
+{
+	int r;
+
+	down_write(&pmd->root_lock);
+	r = __release_metadata_snap(pmd);
+	up_write(&pmd->root_lock);
+
+	return r;
+}
+
+static int __get_metadata_snap(struct dm_pool_metadata *pmd,
+			       dm_block_t *result)
+{
+	int r;
+	struct thin_disk_superblock *disk_super;
+	struct dm_block *sblock;
+
+	r = dm_bm_read_lock(pmd->bm, THIN_SUPERBLOCK_LOCATION,
+			    &sb_validator, &sblock);
+	if (r)
+		return r;
+
 	disk_super = dm_block_data(sblock);
 	*result = le64_to_cpu(disk_super->held_root);
 
 	return dm_bm_unlock(sblock);
 }
 
-int dm_pool_get_held_metadata_root(struct dm_pool_metadata *pmd,
-				   dm_block_t *result)
+int dm_pool_get_metadata_snap(struct dm_pool_metadata *pmd,
+			      dm_block_t *result)
 {
 	int r;
 
 	down_read(&pmd->root_lock);
-	r = __get_held_metadata_root(pmd, result);
+	r = __get_metadata_snap(pmd, result);
 	up_read(&pmd->root_lock);
 
 	return r;

drivers/md/dm-thin-metadata.h

@@ -90,11 +90,18 @@ int dm_pool_get_metadata_transaction_id(struct dm_pool_metadata *pmd,
 
 /*
  * Hold/get root for userspace transaction.
+ *
+ * The metadata snapshot is a copy of the current superblock (minus the
+ * space maps).  Userland can access the data structures for READ
+ * operations only.  A small performance hit is incurred by providing this
+ * copy of the metadata to userland due to extra copy-on-write operations
+ * on the metadata nodes.  Release this as soon as you finish with it.
  */
-int dm_pool_hold_metadata_root(struct dm_pool_metadata *pmd);
+int dm_pool_reserve_metadata_snap(struct dm_pool_metadata *pmd);
+int dm_pool_release_metadata_snap(struct dm_pool_metadata *pmd);
 
-int dm_pool_get_held_metadata_root(struct dm_pool_metadata *pmd,
-				   dm_block_t *result);
+int dm_pool_get_metadata_snap(struct dm_pool_metadata *pmd,
+			      dm_block_t *result);
 
 /*
  * Actions on a single virtual device.