Btrfs: remove transaction from send
authorJosef Bacik <jbacik@fb.com>
Thu, 13 Mar 2014 19:42:13 +0000 (15:42 -0400)
committerChris Mason <clm@fb.com>
Mon, 7 Apr 2014 00:39:30 +0000 (17:39 -0700)
Lets try this again.  We can deadlock the box if we send on a box and try to
write onto the same fs with the app that is trying to listen to the send pipe.
This is because the writer could get stuck waiting for a transaction commit
which is being blocked by the send.  So fix this by making sure looking at the
commit roots is always going to be consistent.  We do this by keeping track of
which roots need to have their commit roots swapped during commit, and then
taking the commit_root_sem and swapping them all at once.  Then make sure we
take a read lock on the commit_root_sem in cases where we search the commit root
to make sure we're always looking at a consistent view of the commit roots.
Previously we had problems with this because we would swap a fs tree commit root
and then swap the extent tree commit root independently which would cause the
backref walking code to screw up sometimes.  With this patch we no longer
deadlock and pass all the weird send/receive corner cases.  Thanks,

Reportedy-by: Hugo Mills <hugo@carfax.org.uk>
Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Chris Mason <clm@fb.com>
fs/btrfs/backref.c
fs/btrfs/ctree.c
fs/btrfs/ctree.h
fs/btrfs/disk-io.c
fs/btrfs/extent-tree.c
fs/btrfs/inode-map.c
fs/btrfs/send.c
fs/btrfs/transaction.c
fs/btrfs/transaction.h

index aad7201ad11bb767cbe760df38f88572fe788cf5..10db21fa09263786db6d094c64ee7b16a1c2c2ae 100644 (file)
@@ -330,7 +330,10 @@ static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
                goto out;
        }
 
-       root_level = btrfs_old_root_level(root, time_seq);
+       if (path->search_commit_root)
+               root_level = btrfs_header_level(root->commit_root);
+       else
+               root_level = btrfs_old_root_level(root, time_seq);
 
        if (root_level + 1 == level) {
                srcu_read_unlock(&fs_info->subvol_srcu, index);
@@ -1099,9 +1102,9 @@ static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
  *
  * returns 0 on success, < 0 on error.
  */
-int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
-                               struct btrfs_fs_info *fs_info, u64 bytenr,
-                               u64 time_seq, struct ulist **roots)
+static int __btrfs_find_all_roots(struct btrfs_trans_handle *trans,
+                                 struct btrfs_fs_info *fs_info, u64 bytenr,
+                                 u64 time_seq, struct ulist **roots)
 {
        struct ulist *tmp;
        struct ulist_node *node = NULL;
@@ -1137,6 +1140,20 @@ int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
        return 0;
 }
 
+int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
+                        struct btrfs_fs_info *fs_info, u64 bytenr,
+                        u64 time_seq, struct ulist **roots)
+{
+       int ret;
+
+       if (!trans)
+               down_read(&fs_info->commit_root_sem);
+       ret = __btrfs_find_all_roots(trans, fs_info, bytenr, time_seq, roots);
+       if (!trans)
+               up_read(&fs_info->commit_root_sem);
+       return ret;
+}
+
 /*
  * this makes the path point to (inum INODE_ITEM ioff)
  */
@@ -1516,6 +1533,8 @@ int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
                if (IS_ERR(trans))
                        return PTR_ERR(trans);
                btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem);
+       } else {
+               down_read(&fs_info->commit_root_sem);
        }
 
        ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
@@ -1526,8 +1545,8 @@ int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
 
        ULIST_ITER_INIT(&ref_uiter);
        while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
-               ret = btrfs_find_all_roots(trans, fs_info, ref_node->val,
-                                          tree_mod_seq_elem.seq, &roots);
+               ret = __btrfs_find_all_roots(trans, fs_info, ref_node->val,
+                                            tree_mod_seq_elem.seq, &roots);
                if (ret)
                        break;
                ULIST_ITER_INIT(&root_uiter);
@@ -1549,6 +1568,8 @@ out:
        if (!search_commit_root) {
                btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem);
                btrfs_end_transaction(trans, fs_info->extent_root);
+       } else {
+               up_read(&fs_info->commit_root_sem);
        }
 
        return ret;
index 88d1b1eedc9cd3a9679758dc1d209e0bf88767aa..9d89c1648e8ed48c677bc21d7c3e66032f261fe6 100644 (file)
@@ -5360,7 +5360,6 @@ int btrfs_compare_trees(struct btrfs_root *left_root,
 {
        int ret;
        int cmp;
-       struct btrfs_trans_handle *trans = NULL;
        struct btrfs_path *left_path = NULL;
        struct btrfs_path *right_path = NULL;
        struct btrfs_key left_key;
@@ -5378,9 +5377,6 @@ int btrfs_compare_trees(struct btrfs_root *left_root,
        u64 right_blockptr;
        u64 left_gen;
        u64 right_gen;
-       u64 left_start_ctransid;
-       u64 right_start_ctransid;
-       u64 ctransid;
 
        left_path = btrfs_alloc_path();
        if (!left_path) {
@@ -5404,21 +5400,6 @@ int btrfs_compare_trees(struct btrfs_root *left_root,
        right_path->search_commit_root = 1;
        right_path->skip_locking = 1;
 
-       spin_lock(&left_root->root_item_lock);
-       left_start_ctransid = btrfs_root_ctransid(&left_root->root_item);
-       spin_unlock(&left_root->root_item_lock);
-
-       spin_lock(&right_root->root_item_lock);
-       right_start_ctransid = btrfs_root_ctransid(&right_root->root_item);
-       spin_unlock(&right_root->root_item_lock);
-
-       trans = btrfs_join_transaction(left_root);
-       if (IS_ERR(trans)) {
-               ret = PTR_ERR(trans);
-               trans = NULL;
-               goto out;
-       }
-
        /*
         * Strategy: Go to the first items of both trees. Then do
         *
@@ -5482,67 +5463,6 @@ int btrfs_compare_trees(struct btrfs_root *left_root,
        advance_left = advance_right = 0;
 
        while (1) {
-               /*
-                * We need to make sure the transaction does not get committed
-                * while we do anything on commit roots. This means, we need to
-                * join and leave transactions for every item that we process.
-                */
-               if (trans && btrfs_should_end_transaction(trans, left_root)) {
-                       btrfs_release_path(left_path);
-                       btrfs_release_path(right_path);
-
-                       ret = btrfs_end_transaction(trans, left_root);
-                       trans = NULL;
-                       if (ret < 0)
-                               goto out;
-               }
-               /* now rejoin the transaction */
-               if (!trans) {
-                       trans = btrfs_join_transaction(left_root);
-                       if (IS_ERR(trans)) {
-                               ret = PTR_ERR(trans);
-                               trans = NULL;
-                               goto out;
-                       }
-
-                       spin_lock(&left_root->root_item_lock);
-                       ctransid = btrfs_root_ctransid(&left_root->root_item);
-                       spin_unlock(&left_root->root_item_lock);
-                       if (ctransid != left_start_ctransid)
-                               left_start_ctransid = 0;
-
-                       spin_lock(&right_root->root_item_lock);
-                       ctransid = btrfs_root_ctransid(&right_root->root_item);
-                       spin_unlock(&right_root->root_item_lock);
-                       if (ctransid != right_start_ctransid)
-                               right_start_ctransid = 0;
-
-                       if (!left_start_ctransid || !right_start_ctransid) {
-                               WARN(1, KERN_WARNING
-                                       "BTRFS: btrfs_compare_tree detected "
-                                       "a change in one of the trees while "
-                                       "iterating. This is probably a "
-                                       "bug.\n");
-                               ret = -EIO;
-                               goto out;
-                       }
-
-                       /*
-                        * the commit root may have changed, so start again
-                        * where we stopped
-                        */
-                       left_path->lowest_level = left_level;
-                       right_path->lowest_level = right_level;
-                       ret = btrfs_search_slot(NULL, left_root,
-                                       &left_key, left_path, 0, 0);
-                       if (ret < 0)
-                               goto out;
-                       ret = btrfs_search_slot(NULL, right_root,
-                                       &right_key, right_path, 0, 0);
-                       if (ret < 0)
-                               goto out;
-               }
-
                if (advance_left && !left_end_reached) {
                        ret = tree_advance(left_root, left_path, &left_level,
                                        left_root_level,
@@ -5672,14 +5592,6 @@ out:
        btrfs_free_path(left_path);
        btrfs_free_path(right_path);
        kfree(tmp_buf);
-
-       if (trans) {
-               if (!ret)
-                       ret = btrfs_end_transaction(trans, left_root);
-               else
-                       btrfs_end_transaction(trans, left_root);
-       }
-
        return ret;
 }
 
index 2a9d32e193a5f9c7b10f8f8f1d3b8abb8ede9bb6..4253ab257c9c0abf1b63993a8c1e2a97fa00b441 100644 (file)
@@ -1440,7 +1440,7 @@ struct btrfs_fs_info {
         */
        struct mutex ordered_extent_flush_mutex;
 
-       struct rw_semaphore extent_commit_sem;
+       struct rw_semaphore commit_root_sem;
 
        struct rw_semaphore cleanup_work_sem;
 
@@ -1711,7 +1711,6 @@ struct btrfs_root {
        struct btrfs_block_rsv *block_rsv;
 
        /* free ino cache stuff */
-       struct mutex fs_commit_mutex;
        struct btrfs_free_space_ctl *free_ino_ctl;
        enum btrfs_caching_type cached;
        spinlock_t cache_lock;
index 98fe701933976c71473f16eed8d571c3576940a2..6d1ac7d46f8193c843e056fd7b6b36a24ef520d9 100644 (file)
@@ -1567,7 +1567,6 @@ int btrfs_init_fs_root(struct btrfs_root *root)
        root->subv_writers = writers;
 
        btrfs_init_free_ino_ctl(root);
-       mutex_init(&root->fs_commit_mutex);
        spin_lock_init(&root->cache_lock);
        init_waitqueue_head(&root->cache_wait);
 
@@ -2345,7 +2344,7 @@ int open_ctree(struct super_block *sb,
        mutex_init(&fs_info->transaction_kthread_mutex);
        mutex_init(&fs_info->cleaner_mutex);
        mutex_init(&fs_info->volume_mutex);
-       init_rwsem(&fs_info->extent_commit_sem);
+       init_rwsem(&fs_info->commit_root_sem);
        init_rwsem(&fs_info->cleanup_work_sem);
        init_rwsem(&fs_info->subvol_sem);
        sema_init(&fs_info->uuid_tree_rescan_sem, 1);
index e09db2c2f3b48d578cefd5bc528b2b3e9bcba333..4d2508bbf6f0f6cb45bc94ecf5b993bf590a393f 100644 (file)
@@ -419,7 +419,7 @@ static noinline void caching_thread(struct btrfs_work *work)
 again:
        mutex_lock(&caching_ctl->mutex);
        /* need to make sure the commit_root doesn't disappear */
-       down_read(&fs_info->extent_commit_sem);
+       down_read(&fs_info->commit_root_sem);
 
 next:
        ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
@@ -443,10 +443,10 @@ next:
                                break;
 
                        if (need_resched() ||
-                           rwsem_is_contended(&fs_info->extent_commit_sem)) {
+                           rwsem_is_contended(&fs_info->commit_root_sem)) {
                                caching_ctl->progress = last;
                                btrfs_release_path(path);
-                               up_read(&fs_info->extent_commit_sem);
+                               up_read(&fs_info->commit_root_sem);
                                mutex_unlock(&caching_ctl->mutex);
                                cond_resched();
                                goto again;
@@ -513,7 +513,7 @@ next:
 
 err:
        btrfs_free_path(path);
-       up_read(&fs_info->extent_commit_sem);
+       up_read(&fs_info->commit_root_sem);
 
        free_excluded_extents(extent_root, block_group);
 
@@ -633,10 +633,10 @@ static int cache_block_group(struct btrfs_block_group_cache *cache,
                return 0;
        }
 
-       down_write(&fs_info->extent_commit_sem);
+       down_write(&fs_info->commit_root_sem);
        atomic_inc(&caching_ctl->count);
        list_add_tail(&caching_ctl->list, &fs_info->caching_block_groups);
-       up_write(&fs_info->extent_commit_sem);
+       up_write(&fs_info->commit_root_sem);
 
        btrfs_get_block_group(cache);
 
@@ -5471,7 +5471,7 @@ void btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
        struct btrfs_block_group_cache *cache;
        struct btrfs_space_info *space_info;
 
-       down_write(&fs_info->extent_commit_sem);
+       down_write(&fs_info->commit_root_sem);
 
        list_for_each_entry_safe(caching_ctl, next,
                                 &fs_info->caching_block_groups, list) {
@@ -5490,7 +5490,7 @@ void btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
        else
                fs_info->pinned_extents = &fs_info->freed_extents[0];
 
-       up_write(&fs_info->extent_commit_sem);
+       up_write(&fs_info->commit_root_sem);
 
        list_for_each_entry_rcu(space_info, &fs_info->space_info, list)
                percpu_counter_set(&space_info->total_bytes_pinned, 0);
@@ -8256,14 +8256,14 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info)
        struct btrfs_caching_control *caching_ctl;
        struct rb_node *n;
 
-       down_write(&info->extent_commit_sem);
+       down_write(&info->commit_root_sem);
        while (!list_empty(&info->caching_block_groups)) {
                caching_ctl = list_entry(info->caching_block_groups.next,
                                         struct btrfs_caching_control, list);
                list_del(&caching_ctl->list);
                put_caching_control(caching_ctl);
        }
-       up_write(&info->extent_commit_sem);
+       up_write(&info->commit_root_sem);
 
        spin_lock(&info->block_group_cache_lock);
        while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
index ab485e57b6fe6c7ea14f580976dd46c7efd88ffb..cc8ca193d830f62ec5202933f49d9867b1427969 100644 (file)
@@ -55,7 +55,7 @@ static int caching_kthread(void *data)
        key.type = BTRFS_INODE_ITEM_KEY;
 again:
        /* need to make sure the commit_root doesn't disappear */
-       mutex_lock(&root->fs_commit_mutex);
+       down_read(&fs_info->commit_root_sem);
 
        ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
        if (ret < 0)
@@ -88,7 +88,7 @@ again:
                                btrfs_item_key_to_cpu(leaf, &key, 0);
                                btrfs_release_path(path);
                                root->cache_progress = last;
-                               mutex_unlock(&root->fs_commit_mutex);
+                               up_read(&fs_info->commit_root_sem);
                                schedule_timeout(1);
                                goto again;
                        } else
@@ -127,7 +127,7 @@ next:
        btrfs_unpin_free_ino(root);
 out:
        wake_up(&root->cache_wait);
-       mutex_unlock(&root->fs_commit_mutex);
+       up_read(&fs_info->commit_root_sem);
 
        btrfs_free_path(path);
 
@@ -223,11 +223,11 @@ again:
                 * or the caching work is done.
                 */
 
-               mutex_lock(&root->fs_commit_mutex);
+               down_write(&root->fs_info->commit_root_sem);
                spin_lock(&root->cache_lock);
                if (root->cached == BTRFS_CACHE_FINISHED) {
                        spin_unlock(&root->cache_lock);
-                       mutex_unlock(&root->fs_commit_mutex);
+                       up_write(&root->fs_info->commit_root_sem);
                        goto again;
                }
                spin_unlock(&root->cache_lock);
@@ -240,7 +240,7 @@ again:
                else
                        __btrfs_add_free_space(pinned, objectid, 1);
 
-               mutex_unlock(&root->fs_commit_mutex);
+               up_write(&root->fs_info->commit_root_sem);
        }
 }
 
@@ -250,7 +250,7 @@ again:
  * and others will just be dropped, because the commit root we were
  * searching has changed.
  *
- * Must be called with root->fs_commit_mutex held
+ * Must be called with root->fs_info->commit_root_sem held
  */
 void btrfs_unpin_free_ino(struct btrfs_root *root)
 {
index d00534b78382df62eb7f2a310f64ffd6747f2d27..6b5f13659317b100c4f3126a24390026839d301b 100644 (file)
@@ -1268,8 +1268,10 @@ static int find_extent_clone(struct send_ctx *sctx,
        }
        logical = disk_byte + btrfs_file_extent_offset(eb, fi);
 
+       down_read(&sctx->send_root->fs_info->commit_root_sem);
        ret = extent_from_logical(sctx->send_root->fs_info, disk_byte, tmp_path,
                                  &found_key, &flags);
+       up_read(&sctx->send_root->fs_info->commit_root_sem);
        btrfs_release_path(tmp_path);
 
        if (ret < 0)
@@ -5367,57 +5369,21 @@ out:
 static int full_send_tree(struct send_ctx *sctx)
 {
        int ret;
-       struct btrfs_trans_handle *trans = NULL;
        struct btrfs_root *send_root = sctx->send_root;
        struct btrfs_key key;
        struct btrfs_key found_key;
        struct btrfs_path *path;
        struct extent_buffer *eb;
        int slot;
-       u64 start_ctransid;
-       u64 ctransid;
 
        path = alloc_path_for_send();
        if (!path)
                return -ENOMEM;
 
-       spin_lock(&send_root->root_item_lock);
-       start_ctransid = btrfs_root_ctransid(&send_root->root_item);
-       spin_unlock(&send_root->root_item_lock);
-
        key.objectid = BTRFS_FIRST_FREE_OBJECTID;
        key.type = BTRFS_INODE_ITEM_KEY;
        key.offset = 0;
 
-join_trans:
-       /*
-        * We need to make sure the transaction does not get committed
-        * while we do anything on commit roots. Join a transaction to prevent
-        * this.
-        */
-       trans = btrfs_join_transaction(send_root);
-       if (IS_ERR(trans)) {
-               ret = PTR_ERR(trans);
-               trans = NULL;
-               goto out;
-       }
-
-       /*
-        * Make sure the tree has not changed after re-joining. We detect this
-        * by comparing start_ctransid and ctransid. They should always match.
-        */
-       spin_lock(&send_root->root_item_lock);
-       ctransid = btrfs_root_ctransid(&send_root->root_item);
-       spin_unlock(&send_root->root_item_lock);
-
-       if (ctransid != start_ctransid) {
-               WARN(1, KERN_WARNING "BTRFS: the root that you're trying to "
-                                    "send was modified in between. This is "
-                                    "probably a bug.\n");
-               ret = -EIO;
-               goto out;
-       }
-
        ret = btrfs_search_slot_for_read(send_root, &key, path, 1, 0);
        if (ret < 0)
                goto out;
@@ -5425,19 +5391,6 @@ join_trans:
                goto out_finish;
 
        while (1) {
-               /*
-                * When someone want to commit while we iterate, end the
-                * joined transaction and rejoin.
-                */
-               if (btrfs_should_end_transaction(trans, send_root)) {
-                       ret = btrfs_end_transaction(trans, send_root);
-                       trans = NULL;
-                       if (ret < 0)
-                               goto out;
-                       btrfs_release_path(path);
-                       goto join_trans;
-               }
-
                eb = path->nodes[0];
                slot = path->slots[0];
                btrfs_item_key_to_cpu(eb, &found_key, slot);
@@ -5465,12 +5418,6 @@ out_finish:
 
 out:
        btrfs_free_path(path);
-       if (trans) {
-               if (!ret)
-                       ret = btrfs_end_transaction(trans, send_root);
-               else
-                       btrfs_end_transaction(trans, send_root);
-       }
        return ret;
 }
 
index 038177cfabbb3da52535e64273b43d21775ee3a4..7579f6d0b8549525e6460f36af8666e1ca980a23 100644 (file)
@@ -75,10 +75,21 @@ void btrfs_put_transaction(struct btrfs_transaction *transaction)
        }
 }
 
-static noinline void switch_commit_root(struct btrfs_root *root)
+static noinline void switch_commit_roots(struct btrfs_transaction *trans,
+                                        struct btrfs_fs_info *fs_info)
 {
-       free_extent_buffer(root->commit_root);
-       root->commit_root = btrfs_root_node(root);
+       struct btrfs_root *root, *tmp;
+
+       down_write(&fs_info->commit_root_sem);
+       list_for_each_entry_safe(root, tmp, &trans->switch_commits,
+                                dirty_list) {
+               list_del_init(&root->dirty_list);
+               free_extent_buffer(root->commit_root);
+               root->commit_root = btrfs_root_node(root);
+               if (is_fstree(root->objectid))
+                       btrfs_unpin_free_ino(root);
+       }
+       up_write(&fs_info->commit_root_sem);
 }
 
 static inline void extwriter_counter_inc(struct btrfs_transaction *trans,
@@ -208,6 +219,7 @@ loop:
        INIT_LIST_HEAD(&cur_trans->pending_snapshots);
        INIT_LIST_HEAD(&cur_trans->ordered_operations);
        INIT_LIST_HEAD(&cur_trans->pending_chunks);
+       INIT_LIST_HEAD(&cur_trans->switch_commits);
        list_add_tail(&cur_trans->list, &fs_info->trans_list);
        extent_io_tree_init(&cur_trans->dirty_pages,
                             fs_info->btree_inode->i_mapping);
@@ -920,9 +932,6 @@ static int update_cowonly_root(struct btrfs_trans_handle *trans,
                        return ret;
        }
 
-       if (root != root->fs_info->extent_root)
-               switch_commit_root(root);
-
        return 0;
 }
 
@@ -978,15 +987,16 @@ static noinline int commit_cowonly_roots(struct btrfs_trans_handle *trans,
                list_del_init(next);
                root = list_entry(next, struct btrfs_root, dirty_list);
 
+               if (root != fs_info->extent_root)
+                       list_add_tail(&root->dirty_list,
+                                     &trans->transaction->switch_commits);
                ret = update_cowonly_root(trans, root);
                if (ret)
                        return ret;
        }
 
-       down_write(&fs_info->extent_commit_sem);
-       switch_commit_root(fs_info->extent_root);
-       up_write(&fs_info->extent_commit_sem);
-
+       list_add_tail(&fs_info->extent_root->dirty_list,
+                     &trans->transaction->switch_commits);
        btrfs_after_dev_replace_commit(fs_info);
 
        return 0;
@@ -1043,11 +1053,8 @@ static noinline int commit_fs_roots(struct btrfs_trans_handle *trans,
                        smp_wmb();
 
                        if (root->commit_root != root->node) {
-                               mutex_lock(&root->fs_commit_mutex);
-                               switch_commit_root(root);
-                               btrfs_unpin_free_ino(root);
-                               mutex_unlock(&root->fs_commit_mutex);
-
+                               list_add_tail(&root->dirty_list,
+                                       &trans->transaction->switch_commits);
                                btrfs_set_root_node(&root->root_item,
                                                    root->node);
                        }
@@ -1858,11 +1865,15 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 
        btrfs_set_root_node(&root->fs_info->tree_root->root_item,
                            root->fs_info->tree_root->node);
-       switch_commit_root(root->fs_info->tree_root);
+       list_add_tail(&root->fs_info->tree_root->dirty_list,
+                     &cur_trans->switch_commits);
 
        btrfs_set_root_node(&root->fs_info->chunk_root->root_item,
                            root->fs_info->chunk_root->node);
-       switch_commit_root(root->fs_info->chunk_root);
+       list_add_tail(&root->fs_info->chunk_root->dirty_list,
+                     &cur_trans->switch_commits);
+
+       switch_commit_roots(cur_trans, root->fs_info);
 
        assert_qgroups_uptodate(trans);
        update_super_roots(root);
index 2bcba89d952e877d93ff924130c3258227fcc986..b57b924e8e03ee66fd98bc24b0e53de672be5b3b 100644 (file)
@@ -57,6 +57,7 @@ struct btrfs_transaction {
        struct list_head pending_snapshots;
        struct list_head ordered_operations;
        struct list_head pending_chunks;
+       struct list_head switch_commits;
        struct btrfs_delayed_ref_root delayed_refs;
        int aborted;
 };