dm thin: provide userspace access to pool metadata

This patch implements two new messages that can be sent to the thin
pool target allowing it to take a snapshot of the _metadata_.  This,
read-only snapshot can be accessed by userland, concurrently with the
live target.

Only one metadata snapshot can be held at a time.  The pool's status
line will give the block location for the current msnap.

Since version 0.1.5 of the userland thin provisioning tools, the
thin_dump program displays the msnap as follows:

    thin_dump -m <msnap root> <metadata dev>

Available here: https://github.com/jthornber/thin-provisioning-tools

Now that userland can access the metadata we can do various things
that have traditionally been kernel side tasks:

     i) Incremental backups.

     By using metadata snapshots we can work out what blocks have
     changed over time.  Combined with data snapshots we can ensure
     the data doesn't change while we back it up.

     A short proof of concept script can be found here:

     https://github.com/jthornber/thinp-test-suite/blob/master/incremental_backup_example.rb

     ii) Migration of thin devices from one pool to another.

     iii) Merging snapshots back into an external origin.

     iv) Asyncronous replication.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

5 files changed, 193 insertions, 11 deletions

diff --git a/Documentation/device-mapper/thin-provisioning.txt b/Documentation/device-mapper/thin-provisioning.txt
index 3370bc4d7b9..f5cfc62b7ad 100644
--- a/Documentation/device-mapper/thin-provisioning.txt
+++ b/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
 -------------
 
diff --git a/drivers/md/dm-thin-metadata.c b/drivers/md/dm-thin-metadata.c
index 737d38865b6..3e2907f0bc4 100644
--- a/drivers/md/dm-thin-metadata.c
+++ b/drivers/md/dm-thin-metadata.c
@@ -1082,12 +1082,89 @@ 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);
@@ -1095,18 +1172,65 @@ static int __get_held_metadata_root(struct dm_pool_metadata *pmd,
                 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;
diff --git a/drivers/md/dm-thin-metadata.h b/drivers/md/dm-thin-metadata.h
index ed4725e67c9..b88918ccdaf 100644
--- a/drivers/md/dm-thin-metadata.h
+++ b/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.
diff --git a/drivers/md/dm-thin.c b/drivers/md/dm-thin.c
index db1b041ce97..37fdaf81bd1 100644
--- a/drivers/md/dm-thin.c
+++ b/drivers/md/dm-thin.c
@@ -2284,6 +2284,36 @@ static int process_set_transaction_id_mesg(unsigned argc, char **argv, struct po
         return 0;
 }
 
+static int process_reserve_metadata_snap_mesg(unsigned argc, char **argv, struct pool *pool)
+{
+        int r;
+
+        r = check_arg_count(argc, 1);
+        if (r)
+                return r;
+
+        r = dm_pool_reserve_metadata_snap(pool->pmd);
+        if (r)
+                DMWARN("reserve_metadata_snap message failed.");
+
+        return r;
+}
+
+static int process_release_metadata_snap_mesg(unsigned argc, char **argv, struct pool *pool)
+{
+        int r;
+
+        r = check_arg_count(argc, 1);
+        if (r)
+                return r;
+
+        r = dm_pool_release_metadata_snap(pool->pmd);
+        if (r)
+                DMWARN("release_metadata_snap message failed.");
+
+        return r;
+}
+
 /*
  * Messages supported:
  *   create_thin        <dev_id>
@@ -2291,6 +2321,8 @@ static int process_set_transaction_id_mesg(unsigned argc, char **argv, struct po
  *   delete             <dev_id>
  *   trim               <dev_id> <new_size_in_sectors>
  *   set_transaction_id <current_trans_id> <new_trans_id>
+ *   reserve_metadata_snap
+ *   release_metadata_snap
  */
 static int pool_message(struct dm_target *ti, unsigned argc, char **argv)
 {
@@ -2310,6 +2342,12 @@ static int pool_message(struct dm_target *ti, unsigned argc, char **argv)
         else if (!strcasecmp(argv[0], "set_transaction_id"))
                 r = process_set_transaction_id_mesg(argc, argv, pool);
 
+        else if (!strcasecmp(argv[0], "reserve_metadata_snap"))
+                r = process_reserve_metadata_snap_mesg(argc, argv, pool);
+
+        else if (!strcasecmp(argv[0], "release_metadata_snap"))
+                r = process_release_metadata_snap_mesg(argc, argv, pool);
+
         else
                 DMWARN("Unrecognised thin pool target message received: %s", argv[0]);
 
@@ -2369,7 +2407,7 @@ static int pool_status(struct dm_target *ti, status_type_t type,
                 if (r)
                         return r;
 
-                r = dm_pool_get_held_metadata_root(pool->pmd, &held_root);
+                r = dm_pool_get_metadata_snap(pool->pmd, &held_root);
                 if (r)
                         return r;
 
@@ -2465,7 +2503,7 @@ static struct target_type pool_target = {
         .name = "thin-pool",
         .features = DM_TARGET_SINGLETON | DM_TARGET_ALWAYS_WRITEABLE |
                     DM_TARGET_IMMUTABLE,
-        .version = {1, 1, 0},
+        .version = {1, 2, 0},
         .module = THIS_MODULE,
         .ctr = pool_ctr,
         .dtr = pool_dtr,
diff --git a/drivers/md/persistent-data/dm-transaction-manager.c b/drivers/md/persistent-data/dm-transaction-manager.c
index 6f8d38747d7..400fe144c0c 100644
--- a/drivers/md/persistent-data/dm-transaction-manager.c
+++ b/drivers/md/persistent-data/dm-transaction-manager.c
@@ -249,6 +249,7 @@ int dm_tm_shadow_block(struct dm_transaction_manager *tm, dm_block_t orig,
 
         return r;
 }
+EXPORT_SYMBOL_GPL(dm_tm_shadow_block);
 
 int dm_tm_read_lock(struct dm_transaction_manager *tm, dm_block_t b,
                     struct dm_block_validator *v,
@@ -259,6 +260,7 @@ int dm_tm_read_lock(struct dm_transaction_manager *tm, dm_block_t b,
 
         return dm_bm_read_lock(tm->bm, b, v, blk);
 }
+EXPORT_SYMBOL_GPL(dm_tm_read_lock);
 
 int dm_tm_unlock(struct dm_transaction_manager *tm, struct dm_block *b)
 {