dm cache: share cache-metadata object across inactive and active DM tables

If a DM table is reloaded with an inactive table when the device is not
suspended (normal procedure for LVM2), then there will be two dm-bufio
objects that can diverge.  This can lead to a situation where the
inactive table uses bufio to read metadata at the same time the active
table writes metadata -- resulting in the inactive table having stale
metadata buffers once it is promoted to the active table slot.

Fix this by using reference counting and a global list of cache metadata
objects to ensure there is only one metadata object per metadata device.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org

1 files changed, 95 insertions, 6 deletions

diff --git a/drivers/md/dm-cache-metadata.c b/drivers/md/dm-cache-metadata.c
index 9fc616c2755e..21b156242e42 100644
--- a/drivers/md/dm-cache-metadata.c
+++ b/drivers/md/dm-cache-metadata.c
@@ -94,6 +94,9 @@ struct cache_disk_superblock {
 } __packed;
 
 struct dm_cache_metadata {
+        atomic_t ref_count;
+        struct list_head list;
+
         struct block_device *bdev;
         struct dm_block_manager *bm;
         struct dm_space_map *metadata_sm;
@@ -669,10 +672,10 @@ static void unpack_value(__le64 value_le, dm_oblock_t *block, unsigned *flags)
 
 /*----------------------------------------------------------------*/
 
-struct dm_cache_metadata *dm_cache_metadata_open(struct block_device *bdev,
-                                                 sector_t data_block_size,
-                                                 bool may_format_device,
-                                                 size_t policy_hint_size)
+static struct dm_cache_metadata *metadata_open(struct block_device *bdev,
+                                               sector_t data_block_size,
+                                               bool may_format_device,
+                                               size_t policy_hint_size)
 {
         int r;
         struct dm_cache_metadata *cmd;
@@ -683,6 +686,7 @@ struct dm_cache_metadata *dm_cache_metadata_open(struct block_device *bdev,
                 return NULL;
         }
 
+        atomic_set(&cmd->ref_count, 1);
         init_rwsem(&cmd->root_lock);
         cmd->bdev = bdev;
         cmd->data_block_size = data_block_size;
@@ -705,10 +709,95 @@ struct dm_cache_metadata *dm_cache_metadata_open(struct block_device *bdev,
         return cmd;
 }
 
+/*
+ * We keep a little list of ref counted metadata objects to prevent two
+ * different target instances creating separate bufio instances.  This is
+ * an issue if a table is reloaded before the suspend.
+ */
+static DEFINE_MUTEX(table_lock);
+static LIST_HEAD(table);
+
+static struct dm_cache_metadata *lookup(struct block_device *bdev)
+{
+        struct dm_cache_metadata *cmd;
+
+        list_for_each_entry(cmd, &table, list)
+                if (cmd->bdev == bdev) {
+                        atomic_inc(&cmd->ref_count);
+                        return cmd;
+                }
+
+        return NULL;
+}
+
+static struct dm_cache_metadata *lookup_or_open(struct block_device *bdev,
+                                                sector_t data_block_size,
+                                                bool may_format_device,
+                                                size_t policy_hint_size)
+{
+        struct dm_cache_metadata *cmd, *cmd2;
+
+        mutex_lock(&table_lock);
+        cmd = lookup(bdev);
+        mutex_unlock(&table_lock);
+
+        if (cmd)
+                return cmd;
+
+        cmd = metadata_open(bdev, data_block_size, may_format_device, policy_hint_size);
+        if (cmd) {
+                mutex_lock(&table_lock);
+                cmd2 = lookup(bdev);
+                if (cmd2) {
+                        mutex_unlock(&table_lock);
+                        __destroy_persistent_data_objects(cmd);
+                        kfree(cmd);
+                        return cmd2;
+                }
+                list_add(&cmd->list, &table);
+                mutex_unlock(&table_lock);
+        }
+
+        return cmd;
+}
+
+static bool same_params(struct dm_cache_metadata *cmd, sector_t data_block_size)
+{
+        if (cmd->data_block_size != data_block_size) {
+                DMERR("data_block_size (%llu) different from that in metadata (%llu)\n",
+                      (unsigned long long) data_block_size,
+                      (unsigned long long) cmd->data_block_size);
+                return false;
+        }
+
+        return true;
+}
+
+struct dm_cache_metadata *dm_cache_metadata_open(struct block_device *bdev,
+                                                 sector_t data_block_size,
+                                                 bool may_format_device,
+                                                 size_t policy_hint_size)
+{
+        struct dm_cache_metadata *cmd = lookup_or_open(bdev, data_block_size,
+                                                       may_format_device, policy_hint_size);
+        if (cmd && !same_params(cmd, data_block_size)) {
+                dm_cache_metadata_close(cmd);
+                return NULL;
+        }
+
+        return cmd;
+}
+
 void dm_cache_metadata_close(struct dm_cache_metadata *cmd)
 {
-        __destroy_persistent_data_objects(cmd);
-        kfree(cmd);
+        if (atomic_dec_and_test(&cmd->ref_count)) {
+                mutex_lock(&table_lock);
+                list_del(&cmd->list);
+                mutex_unlock(&table_lock);
+
+                __destroy_persistent_data_objects(cmd);
+                kfree(cmd);
+        }
 }
 
 /*