Btrfs: add snapshot/subvolume destroy ioctl

This patch adds snapshot/subvolume destroy ioctl.  A subvolume that isn't being
used and doesn't contains links to other subvolumes can be destroyed.

Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

12 files changed, 605 insertions, 233 deletions

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 6ade48b227e9..bc57e236ac64 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -839,9 +839,7 @@ struct btrfs_fs_info {
         struct mutex transaction_kthread_mutex;
         struct mutex cleaner_mutex;
         struct mutex chunk_mutex;
-        struct mutex drop_mutex;
         struct mutex volume_mutex;
-        struct mutex tree_reloc_mutex;
         /*
          * this protects the ordered operations list only while we are
          * processing all of the entries on it.  This way we make
@@ -852,6 +850,10 @@ struct btrfs_fs_info {
         struct mutex ordered_operations_mutex;
         struct rw_semaphore extent_commit_sem;
 
+        struct rw_semaphore subvol_sem;
+
+        struct srcu_struct subvol_srcu;
+
         struct list_head trans_list;
         struct list_head hashers;
         struct list_head dead_roots;
@@ -2142,6 +2144,7 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid, struct
 int btrfs_search_root(struct btrfs_root *root, u64 search_start,
                       u64 *found_objectid);
 int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid);
+int btrfs_find_orphan_roots(struct btrfs_root *tree_root);
 int btrfs_set_root_node(struct btrfs_root_item *item,
                         struct extent_buffer *node);
 /* dir-item.c */
@@ -2273,7 +2276,7 @@ int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end);
 int btrfs_writepages(struct address_space *mapping,
                      struct writeback_control *wbc);
 int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
-                             struct btrfs_root *new_root, struct dentry *dentry,
+                             struct btrfs_root *new_root,
                              u64 new_dirid, u64 alloc_hint);
 int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
                          size_t size, struct bio *bio, unsigned long bio_flags);
@@ -2289,6 +2292,7 @@ int btrfs_write_inode(struct inode *inode, int wait);
 void btrfs_dirty_inode(struct inode *inode);
 struct inode *btrfs_alloc_inode(struct super_block *sb);
 void btrfs_destroy_inode(struct inode *inode);
+void btrfs_drop_inode(struct inode *inode);
 int btrfs_init_cachep(void);
 void btrfs_destroy_cachep(void);
 long btrfs_ioctl_trans_end(struct file *file);
@@ -2306,6 +2310,8 @@ int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode);
 int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode);
 void btrfs_orphan_cleanup(struct btrfs_root *root);
 int btrfs_cont_expand(struct inode *inode, loff_t size);
+int btrfs_invalidate_inodes(struct btrfs_root *root);
+extern struct dentry_operations btrfs_dentry_operations;
 
 /* ioctl.c */
 long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index a4f531047c4a..a0d41e713f3c 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1378,8 +1378,10 @@ static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
 
         err = bdi_register(bdi, NULL, "btrfs-%d",
                                 atomic_inc_return(&btrfs_bdi_num));
-        if (err)
+        if (err) {
+                bdi_destroy(bdi);
                 return err;
+        }
 
         bdi->ra_pages   = default_backing_dev_info.ra_pages;
         bdi->unplug_io_fn       = btrfs_unplug_io_fn;
@@ -1469,9 +1471,12 @@ static int cleaner_kthread(void *arg)
                         break;
 
                 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
-                mutex_lock(&root->fs_info->cleaner_mutex);
-                btrfs_clean_old_snapshots(root);
-                mutex_unlock(&root->fs_info->cleaner_mutex);
+
+                if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
+                    mutex_trylock(&root->fs_info->cleaner_mutex)) {
+                        btrfs_clean_old_snapshots(root);
+                        mutex_unlock(&root->fs_info->cleaner_mutex);
+                }
 
                 if (freezing(current)) {
                         refrigerator();
@@ -1576,7 +1581,26 @@ struct btrfs_root *open_ctree(struct super_block *sb,
                 err = -ENOMEM;
                 goto fail;
         }
-        INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
+
+        ret = init_srcu_struct(&fs_info->subvol_srcu);
+        if (ret) {
+                err = ret;
+                goto fail;
+        }
+
+        ret = setup_bdi(fs_info, &fs_info->bdi);
+        if (ret) {
+                err = ret;
+                goto fail_srcu;
+        }
+
+        fs_info->btree_inode = new_inode(sb);
+        if (!fs_info->btree_inode) {
+                err = -ENOMEM;
+                goto fail_bdi;
+        }
+
+        INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
         INIT_LIST_HEAD(&fs_info->trans_list);
         INIT_LIST_HEAD(&fs_info->dead_roots);
         INIT_LIST_HEAD(&fs_info->hashers);
@@ -1586,6 +1610,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
         spin_lock_init(&fs_info->delalloc_lock);
         spin_lock_init(&fs_info->new_trans_lock);
         spin_lock_init(&fs_info->ref_cache_lock);
+        spin_lock_init(&fs_info->fs_roots_radix_lock);
 
         init_completion(&fs_info->kobj_unregister);
         fs_info->tree_root = tree_root;
@@ -1604,11 +1629,6 @@ struct btrfs_root *open_ctree(struct super_block *sb,
         fs_info->sb = sb;
         fs_info->max_extent = (u64)-1;
         fs_info->max_inline = 8192 * 1024;
-        if (setup_bdi(fs_info, &fs_info->bdi))
-                goto fail_bdi;
-        fs_info->btree_inode = new_inode(sb);
-        fs_info->btree_inode->i_ino = 1;
-        fs_info->btree_inode->i_nlink = 1;
         fs_info->metadata_ratio = 8;
 
         fs_info->thread_pool_size = min_t(unsigned long,
@@ -1620,6 +1640,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
         sb->s_blocksize = 4096;
         sb->s_blocksize_bits = blksize_bits(4096);
 
+        fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
+        fs_info->btree_inode->i_nlink = 1;
         /*
          * we set the i_size on the btree inode to the max possible int.
          * the real end of the address space is determined by all of
@@ -1638,6 +1660,11 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 
         BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
 
+        BTRFS_I(fs_info->btree_inode)->root = tree_root;
+        memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
+               sizeof(struct btrfs_key));
+        BTRFS_I(fs_info->btree_inode)->dummy_inode = 1;
+
         spin_lock_init(&fs_info->block_group_cache_lock);
         fs_info->block_group_cache_tree.rb_node = NULL;
 
@@ -1648,21 +1675,16 @@ struct btrfs_root *open_ctree(struct super_block *sb,
         fs_info->pinned_extents = &fs_info->freed_extents[0];
         fs_info->do_barriers = 1;
 
-        BTRFS_I(fs_info->btree_inode)->root = tree_root;
-        memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
-               sizeof(struct btrfs_key));
-        insert_inode_hash(fs_info->btree_inode);
 
         mutex_init(&fs_info->trans_mutex);
         mutex_init(&fs_info->ordered_operations_mutex);
         mutex_init(&fs_info->tree_log_mutex);
-        mutex_init(&fs_info->drop_mutex);
         mutex_init(&fs_info->chunk_mutex);
         mutex_init(&fs_info->transaction_kthread_mutex);
         mutex_init(&fs_info->cleaner_mutex);
         mutex_init(&fs_info->volume_mutex);
-        mutex_init(&fs_info->tree_reloc_mutex);
         init_rwsem(&fs_info->extent_commit_sem);
+        init_rwsem(&fs_info->subvol_sem);
 
         btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
         btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
@@ -1941,6 +1963,9 @@ printk("thread pool is %d\n", fs_info->thread_pool_size);
                 }
         }
 
+        ret = btrfs_find_orphan_roots(tree_root);
+        BUG_ON(ret);
+
         if (!(sb->s_flags & MS_RDONLY)) {
                 ret = btrfs_recover_relocation(tree_root);
                 BUG_ON(ret);
@@ -2000,6 +2025,8 @@ fail_iput:
         btrfs_mapping_tree_free(&fs_info->mapping_tree);
 fail_bdi:
         bdi_destroy(&fs_info->bdi);
+fail_srcu:
+        cleanup_srcu_struct(&fs_info->subvol_srcu);
 fail:
         kfree(extent_root);
         kfree(tree_root);
@@ -2263,6 +2290,10 @@ int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
         radix_tree_delete(&fs_info->fs_roots_radix,
                           (unsigned long)root->root_key.objectid);
         spin_unlock(&fs_info->fs_roots_radix_lock);
+
+        if (btrfs_root_refs(&root->root_item) == 0)
+                synchronize_srcu(&fs_info->subvol_srcu);
+
         free_fs_root(root);
         return 0;
 }
@@ -2286,6 +2317,20 @@ static int del_fs_roots(struct btrfs_fs_info *fs_info)
         struct btrfs_root *gang[8];
         int i;
 
+        while (!list_empty(&fs_info->dead_roots)) {
+                gang[0] = list_entry(fs_info->dead_roots.next,
+                                     struct btrfs_root, root_list);
+                list_del(&gang[0]->root_list);
+
+                if (gang[0]->in_radix) {
+                        btrfs_free_fs_root(fs_info, gang[0]);
+                } else {
+                        free_extent_buffer(gang[0]->node);
+                        free_extent_buffer(gang[0]->commit_root);
+                        kfree(gang[0]);
+                }
+        }
+
         while (1) {
                 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
                                              (void **)gang, 0,
@@ -2315,9 +2360,6 @@ int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
                 root_objectid = gang[ret - 1]->root_key.objectid + 1;
                 for (i = 0; i < ret; i++) {
                         root_objectid = gang[i]->root_key.objectid;
-                        ret = btrfs_find_dead_roots(fs_info->tree_root,
-                                                    root_objectid);
-                        BUG_ON(ret);
                         btrfs_orphan_cleanup(gang[i]);
                 }
                 root_objectid++;
@@ -2405,6 +2447,7 @@ int close_ctree(struct btrfs_root *root)
         btrfs_mapping_tree_free(&fs_info->mapping_tree);
 
         bdi_destroy(&fs_info->bdi);
+        cleanup_srcu_struct(&fs_info->subvol_srcu);
 
         kfree(fs_info->extent_root);
         kfree(fs_info->tree_root);
diff --git a/fs/btrfs/export.c b/fs/btrfs/export.c
index 9596b40caa4e..ba5c3fd5ab8c 100644
--- a/fs/btrfs/export.c
+++ b/fs/btrfs/export.c
@@ -28,7 +28,7 @@ static int btrfs_encode_fh(struct dentry *dentry, u32 *fh, int *max_len,
         len  = BTRFS_FID_SIZE_NON_CONNECTABLE;
         type = FILEID_BTRFS_WITHOUT_PARENT;
 
-        fid->objectid = BTRFS_I(inode)->location.objectid;
+        fid->objectid = inode->i_ino;
         fid->root_objectid = BTRFS_I(inode)->root->objectid;
         fid->gen = inode->i_generation;
 
@@ -60,34 +60,61 @@ static int btrfs_encode_fh(struct dentry *dentry, u32 *fh, int *max_len,
 }
 
 static struct dentry *btrfs_get_dentry(struct super_block *sb, u64 objectid,
-                                       u64 root_objectid, u32 generation)
+                                       u64 root_objectid, u32 generation,
+                                       int check_generation)
 {
+        struct btrfs_fs_info *fs_info = btrfs_sb(sb)->fs_info;
         struct btrfs_root *root;
+        struct dentry *dentry;
         struct inode *inode;
         struct btrfs_key key;
+        int index;
+        int err = 0;
+
+        if (objectid < BTRFS_FIRST_FREE_OBJECTID)
+                return ERR_PTR(-ESTALE);
 
         key.objectid = root_objectid;
         btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
         key.offset = (u64)-1;
 
-        root = btrfs_read_fs_root_no_name(btrfs_sb(sb)->fs_info, &key);
-        if (IS_ERR(root))
-                return ERR_CAST(root);
+        index = srcu_read_lock(&fs_info->subvol_srcu);
+
+        root = btrfs_read_fs_root_no_name(fs_info, &key);
+        if (IS_ERR(root)) {
+                err = PTR_ERR(root);
+                goto fail;
+        }
+
+        if (btrfs_root_refs(&root->root_item) == 0) {
+                err = -ENOENT;
+                goto fail;
+        }
 
         key.objectid = objectid;
         btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
         key.offset = 0;
 
         inode = btrfs_iget(sb, &key, root);
-        if (IS_ERR(inode))
-                return (void *)inode;
+        if (IS_ERR(inode)) {
+                err = PTR_ERR(inode);
+                goto fail;
+        }
+
+        srcu_read_unlock(&fs_info->subvol_srcu, index);
 
-        if (generation != inode->i_generation) {
+        if (check_generation && generation != inode->i_generation) {
                 iput(inode);
                 return ERR_PTR(-ESTALE);
         }
 
-        return d_obtain_alias(inode);
+        dentry = d_obtain_alias(inode);
+        if (!IS_ERR(dentry))
+                dentry->d_op = &btrfs_dentry_operations;
+        return dentry;
+fail:
+        srcu_read_unlock(&fs_info->subvol_srcu, index);
+        return ERR_PTR(err);
 }
 
 static struct dentry *btrfs_fh_to_parent(struct super_block *sb, struct fid *fh,
@@ -111,7 +138,7 @@ static struct dentry *btrfs_fh_to_parent(struct super_block *sb, struct fid *fh,
         objectid = fid->parent_objectid;
         generation = fid->parent_gen;
 
-        return btrfs_get_dentry(sb, objectid, root_objectid, generation);
+        return btrfs_get_dentry(sb, objectid, root_objectid, generation, 1);
 }
 
 static struct dentry *btrfs_fh_to_dentry(struct super_block *sb, struct fid *fh,
@@ -133,66 +160,76 @@ static struct dentry *btrfs_fh_to_dentry(struct super_block *sb, struct fid *fh,
         root_objectid = fid->root_objectid;
         generation = fid->gen;
 
-        return btrfs_get_dentry(sb, objectid, root_objectid, generation);
+        return btrfs_get_dentry(sb, objectid, root_objectid, generation, 1);
 }
 
 static struct dentry *btrfs_get_parent(struct dentry *child)
 {
         struct inode *dir = child->d_inode;
+        static struct dentry *dentry;
         struct btrfs_root *root = BTRFS_I(dir)->root;
-        struct btrfs_key key;
         struct btrfs_path *path;
         struct extent_buffer *leaf;
-        int slot;
-        u64 objectid;
+        struct btrfs_root_ref *ref;
+        struct btrfs_key key;
+        struct btrfs_key found_key;
         int ret;
 
         path = btrfs_alloc_path();
 
-        key.objectid = dir->i_ino;
-        btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
-        key.offset = (u64)-1;
+        if (dir->i_ino == BTRFS_FIRST_FREE_OBJECTID) {
+                key.objectid = root->root_key.objectid;
+                key.type = BTRFS_ROOT_BACKREF_KEY;
+                key.offset = (u64)-1;
+                root = root->fs_info->tree_root;
+        } else {
+                key.objectid = dir->i_ino;
+                key.type = BTRFS_INODE_REF_KEY;
+                key.offset = (u64)-1;
+        }
 
         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
-        if (ret < 0) {
-                /* Error */
-                btrfs_free_path(path);
-                return ERR_PTR(ret);
+        if (ret < 0)
+                goto fail;
+
+        BUG_ON(ret == 0);
+        if (path->slots[0] == 0) {
+                ret = -ENOENT;
+                goto fail;
         }
+
+        path->slots[0]--;
         leaf = path->nodes[0];
-        slot = path->slots[0];
-        if (ret) {
-                /* btrfs_search_slot() returns the slot where we'd want to
-                   insert a backref for parent inode #0xFFFFFFFFFFFFFFFF.
-                   The _real_ backref, telling us what the parent inode
-                   _actually_ is, will be in the slot _before_ the one
-                   that btrfs_search_slot() returns. */
-                if (!slot) {
-                        /* Unless there is _no_ key in the tree before... */
-                        btrfs_free_path(path);
-                        return ERR_PTR(-EIO);
-                }
-                slot--;
+
+        btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+        if (found_key.objectid != key.objectid || found_key.type != key.type) {
+                ret = -ENOENT;
+                goto fail;
         }
 
-        btrfs_item_key_to_cpu(leaf, &key, slot);
+        if (found_key.type == BTRFS_ROOT_BACKREF_KEY) {
+                ref = btrfs_item_ptr(leaf, path->slots[0],
+                                     struct btrfs_root_ref);
+                key.objectid = btrfs_root_ref_dirid(leaf, ref);
+        } else {
+                key.objectid = found_key.offset;
+        }
         btrfs_free_path(path);
 
-        if (key.objectid != dir->i_ino || key.type != BTRFS_INODE_REF_KEY)
-                return ERR_PTR(-EINVAL);
-
-        objectid = key.offset;
-
-        /* If we are already at the root of a subvol, return the real root */
-        if (objectid == dir->i_ino)
-                return dget(dir->i_sb->s_root);
+        if (found_key.type == BTRFS_ROOT_BACKREF_KEY) {
+                return btrfs_get_dentry(root->fs_info->sb, key.objectid,
+                                        found_key.offset, 0, 0);
+        }
 
-        /* Build a new key for the inode item */
-        key.objectid = objectid;
-        btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
+        key.type = BTRFS_INODE_ITEM_KEY;
         key.offset = 0;
-
-        return d_obtain_alias(btrfs_iget(root->fs_info->sb, &key, root));
+        dentry = d_obtain_alias(btrfs_iget(root->fs_info->sb, &key, root));
+        if (!IS_ERR(dentry))
+                dentry->d_op = &btrfs_dentry_operations;
+        return dentry;
+fail:
+        btrfs_free_path(path);
+        return ERR_PTR(ret);
 }
 
 const struct export_operations btrfs_export_ops = {
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 8fc922982183..4bd04f3fa8bb 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -5463,9 +5463,24 @@ int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref)
         ret = btrfs_del_root(trans, tree_root, &root->root_key);
         BUG_ON(ret);
 
-        free_extent_buffer(root->node);
-        free_extent_buffer(root->commit_root);
-        kfree(root);
+        if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
+                ret = btrfs_find_last_root(tree_root, root->root_key.objectid,
+                                           NULL, NULL);
+                BUG_ON(ret < 0);
+                if (ret > 0) {
+                        ret = btrfs_del_orphan_item(trans, tree_root,
+                                                    root->root_key.objectid);
+                        BUG_ON(ret);
+                }
+        }
+
+        if (root->in_radix) {
+                btrfs_free_fs_root(tree_root->fs_info, root);
+        } else {
+                free_extent_buffer(root->node);
+                free_extent_buffer(root->commit_root);
+                kfree(root);
+        }
 out:
         btrfs_end_transaction(trans, tree_root);
         kfree(wc);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 6036b36789cc..db9cbd91eb4c 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -3089,6 +3089,11 @@ void btrfs_delete_inode(struct inode *inode)
         }
         btrfs_wait_ordered_range(inode, 0, (u64)-1);
 
+        if (inode->i_nlink > 0) {
+                BUG_ON(btrfs_root_refs(&root->root_item) != 0);
+                goto no_delete;
+        }
+
         btrfs_i_size_write(inode, 0);
         trans = btrfs_join_transaction(root, 1);
 
@@ -3225,11 +3230,13 @@ static void inode_tree_add(struct inode *inode)
         struct btrfs_inode *entry;
         struct rb_node **p;
         struct rb_node *parent;
-
 again:
         p = &root->inode_tree.rb_node;
         parent = NULL;
 
+        if (hlist_unhashed(&inode->i_hash))
+                return;
+
         spin_lock(&root->inode_lock);
         while (*p) {
                 parent = *p;
@@ -3256,13 +3263,87 @@ again:
 static void inode_tree_del(struct inode *inode)
 {
         struct btrfs_root *root = BTRFS_I(inode)->root;
+        int empty = 0;
 
         spin_lock(&root->inode_lock);
         if (!RB_EMPTY_NODE(&BTRFS_I(inode)->rb_node)) {
                 rb_erase(&BTRFS_I(inode)->rb_node, &root->inode_tree);
                 RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node);
+                empty = RB_EMPTY_ROOT(&root->inode_tree);
         }
         spin_unlock(&root->inode_lock);
+
+        if (empty && btrfs_root_refs(&root->root_item) == 0) {
+                synchronize_srcu(&root->fs_info->subvol_srcu);
+                spin_lock(&root->inode_lock);
+                empty = RB_EMPTY_ROOT(&root->inode_tree);
+                spin_unlock(&root->inode_lock);
+                if (empty)
+                        btrfs_add_dead_root(root);
+        }
+}
+
+int btrfs_invalidate_inodes(struct btrfs_root *root)
+{
+        struct rb_node *node;
+        struct rb_node *prev;
+        struct btrfs_inode *entry;
+        struct inode *inode;
+        u64 objectid = 0;
+
+        WARN_ON(btrfs_root_refs(&root->root_item) != 0);
+
+        spin_lock(&root->inode_lock);
+again:
+        node = root->inode_tree.rb_node;
+        prev = NULL;
+        while (node) {
+                prev = node;
+                entry = rb_entry(node, struct btrfs_inode, rb_node);
+
+                if (objectid < entry->vfs_inode.i_ino)
+                        node = node->rb_left;
+                else if (objectid > entry->vfs_inode.i_ino)
+                        node = node->rb_right;
+                else
+                        break;
+        }
+        if (!node) {
+                while (prev) {
+                        entry = rb_entry(prev, struct btrfs_inode, rb_node);
+                        if (objectid <= entry->vfs_inode.i_ino) {
+                                node = prev;
+                                break;
+                        }
+                        prev = rb_next(prev);
+                }
+        }
+        while (node) {
+                entry = rb_entry(node, struct btrfs_inode, rb_node);
+                objectid = entry->vfs_inode.i_ino + 1;
+                inode = igrab(&entry->vfs_inode);
+                if (inode) {
+                        spin_unlock(&root->inode_lock);
+                        if (atomic_read(&inode->i_count) > 1)
+                                d_prune_aliases(inode);
+                        /*
+                         * btrfs_drop_inode will remove it from
+                         * the inode cache when its usage count
+                         * hits zero.
+                         */
+                        iput(inode);
+                        cond_resched();
+                        spin_lock(&root->inode_lock);
+                        goto again;
+                }
+
+                if (cond_resched_lock(&root->inode_lock))
+                        goto again;
+
+                node = rb_next(node);
+        }
+        spin_unlock(&root->inode_lock);
+        return 0;
 }
 
 static noinline void init_btrfs_i(struct inode *inode)
@@ -3379,8 +3460,11 @@ struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
         struct btrfs_root *root = BTRFS_I(dir)->root;
         struct btrfs_root *sub_root = root;
         struct btrfs_key location;
+        int index;
         int ret;
 
+        dentry->d_op = &btrfs_dentry_operations;
+
         if (dentry->d_name.len > BTRFS_NAME_LEN)
                 return ERR_PTR(-ENAMETOOLONG);
 
@@ -3399,6 +3483,7 @@ struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
 
         BUG_ON(location.type != BTRFS_ROOT_ITEM_KEY);
 
+        index = srcu_read_lock(&root->fs_info->subvol_srcu);
         ret = fixup_tree_root_location(root, dir, dentry,
                                        &location, &sub_root);
         if (ret < 0) {
@@ -3409,9 +3494,24 @@ struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
         } else {
                 inode = btrfs_iget(dir->i_sb, &location, sub_root);
         }
+        srcu_read_unlock(&root->fs_info->subvol_srcu, index);
+
         return inode;
 }
 
+static int btrfs_dentry_delete(struct dentry *dentry)
+{
+        struct btrfs_root *root;
+
+        if (!dentry->d_inode)
+                return 0;
+
+        root = BTRFS_I(dentry->d_inode)->root;
+        if (btrfs_root_refs(&root->root_item) == 0)
+                return 1;
+        return 0;
+}
+
 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
                                    struct nameidata *nd)
 {
@@ -4773,11 +4873,11 @@ out:
  * create a new subvolume directory/inode (helper for the ioctl).
  */
 int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
-                             struct btrfs_root *new_root, struct dentry *dentry,
+                             struct btrfs_root *new_root,
                              u64 new_dirid, u64 alloc_hint)
 {
         struct inode *inode;
-        int error;
+        int err;
         u64 index = 0;
 
         inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid,
@@ -4790,11 +4890,10 @@ int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
         inode->i_nlink = 1;
         btrfs_i_size_write(inode, 0);
 
-        error = btrfs_update_inode(trans, new_root, inode);
-        if (error)
-                return error;
+        err = btrfs_update_inode(trans, new_root, inode);
+        BUG_ON(err);
 
-        d_instantiate(dentry, inode);
+        iput(inode);
         return 0;
 }
 
@@ -4872,6 +4971,16 @@ void btrfs_destroy_inode(struct inode *inode)
         kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
 }
 
+void btrfs_drop_inode(struct inode *inode)
+{
+        struct btrfs_root *root = BTRFS_I(inode)->root;
+
+        if (inode->i_nlink > 0 && btrfs_root_refs(&root->root_item) == 0)
+                generic_delete_inode(inode);
+        else
+                generic_drop_inode(inode);
+}
+
 static void init_once(void *foo)
 {
         struct btrfs_inode *ei = (struct btrfs_inode *) foo;
@@ -4973,6 +5082,10 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
             old_inode->i_size > BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT)
                 filemap_flush(old_inode->i_mapping);
 
+        /* close the racy window with snapshot create/destroy ioctl */
+        if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
+                down_read(&root->fs_info->subvol_sem);
+
         trans = btrfs_start_transaction(root, 1);
 
         if (dest != root)
@@ -5062,6 +5175,8 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 
         btrfs_end_transaction_throttle(trans, root);
 
+        if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
+                up_read(&root->fs_info->subvol_sem);
         return ret;
 }
 
@@ -5420,6 +5535,7 @@ static struct inode_operations btrfs_dir_ro_inode_operations = {
         .lookup         = btrfs_lookup,
         .permission     = btrfs_permission,
 };
+
 static struct file_operations btrfs_dir_file_operations = {
         .llseek         = generic_file_llseek,
         .read           = generic_read_dir,
@@ -5506,3 +5622,7 @@ static struct inode_operations btrfs_symlink_inode_operations = {
         .listxattr      = btrfs_listxattr,
         .removexattr    = btrfs_removexattr,
 };
+
+struct dentry_operations btrfs_dentry_operations = {
+        .d_delete       = btrfs_dentry_delete,
+};
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 9b3a88755e51..a13fd556db74 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -230,8 +230,8 @@ static noinline int create_subvol(struct btrfs_root *root,
         struct btrfs_root_item root_item;
         struct btrfs_inode_item *inode_item;
         struct extent_buffer *leaf;
-        struct btrfs_root *new_root = root;
-        struct inode *dir;
+        struct btrfs_root *new_root;
+        struct inode *dir = dentry->d_parent->d_inode;
         int ret;
         int err;
         u64 objectid;
@@ -241,7 +241,7 @@ static noinline int create_subvol(struct btrfs_root *root,
 
         ret = btrfs_check_metadata_free_space(root);
         if (ret)
-                goto fail_commit;
+                return ret;
 
         trans = btrfs_start_transaction(root, 1);
         BUG_ON(!trans);
@@ -304,11 +304,17 @@ static noinline int create_subvol(struct btrfs_root *root,
         if (ret)
                 goto fail;
 
+        key.offset = (u64)-1;
+        new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
+        BUG_ON(IS_ERR(new_root));
+
+        btrfs_record_root_in_trans(trans, new_root);
+
+        ret = btrfs_create_subvol_root(trans, new_root, new_dirid,
+                                       BTRFS_I(dir)->block_group);
         /*
          * insert the directory item
          */
-        key.offset = (u64)-1;
-        dir = dentry->d_parent->d_inode;
         ret = btrfs_set_inode_index(dir, &index);
         BUG_ON(ret);
 
@@ -325,30 +331,15 @@ static noinline int create_subvol(struct btrfs_root *root,
         ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
                                  objectid, root->root_key.objectid,
                                  dir->i_ino, index, name, namelen);
-        BUG_ON(ret);
 
-        ret = btrfs_commit_transaction(trans, root);
-        if (ret)
-                goto fail_commit;
-
-        new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
-        BUG_ON(!new_root);
-
-        trans = btrfs_start_transaction(new_root, 1);
-        BUG_ON(!trans);
-
-        ret = btrfs_create_subvol_root(trans, new_root, dentry, new_dirid,
-                                       BTRFS_I(dir)->block_group);
-        if (ret)
-                goto fail;
+        BUG_ON(ret);
 
+        d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
 fail:
         nr = trans->blocks_used;
-        err = btrfs_commit_transaction(trans, new_root);
+        err = btrfs_commit_transaction(trans, root);
         if (err && !ret)
                 ret = err;
-fail_commit:
-        btrfs_btree_balance_dirty(root, nr);
         return ret;
 }
 
@@ -409,14 +400,15 @@ static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
  * inside this filesystem so it's quite a bit simpler.
  */
-static noinline int btrfs_mksubvol(struct path *parent, char *name,
-                                   int mode, int namelen,
+static noinline int btrfs_mksubvol(struct path *parent,
+                                   char *name, int namelen,
                                    struct btrfs_root *snap_src)
 {
+        struct inode *dir  = parent->dentry->d_inode;
         struct dentry *dentry;
         int error;
 
-        mutex_lock_nested(&parent->dentry->d_inode->i_mutex, I_MUTEX_PARENT);
+        mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
 
         dentry = lookup_one_len(name, parent->dentry, namelen);
         error = PTR_ERR(dentry);
@@ -427,99 +419,39 @@ static noinline int btrfs_mksubvol(struct path *parent, char *name,
         if (dentry->d_inode)
                 goto out_dput;
 
-        if (!IS_POSIXACL(parent->dentry->d_inode))
-                mode &= ~current_umask();
-
         error = mnt_want_write(parent->mnt);
         if (error)
                 goto out_dput;
 
-        error = btrfs_may_create(parent->dentry->d_inode, dentry);
+        error = btrfs_may_create(dir, dentry);
         if (error)
                 goto out_drop_write;
 
-        /*
-         * Actually perform the low-level subvolume creation after all
-         * this VFS fuzz.
-         *
-         * Eventually we want to pass in an inode under which we create this
-         * subvolume, but for now all are under the filesystem root.
-         *
-         * Also we should pass on the mode eventually to allow creating new
-         * subvolume with specific mode bits.
-         */
+        down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
+
+        if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
+                goto out_up_read;
+
         if (snap_src) {
-                struct dentry *dir = dentry->d_parent;
-                struct dentry *test = dir->d_parent;
-                struct btrfs_path *path = btrfs_alloc_path();
-                int ret;
-                u64 test_oid;
-                u64 parent_oid = BTRFS_I(dir->d_inode)->root->root_key.objectid;
-
-                test_oid = snap_src->root_key.objectid;
-
-                ret = btrfs_find_root_ref(snap_src->fs_info->tree_root,
-                                          path, parent_oid, test_oid);
-                if (ret == 0)
-                        goto create;
-                btrfs_release_path(snap_src->fs_info->tree_root, path);
-
-                /* we need to make sure we aren't creating a directory loop
-                 * by taking a snapshot of something that has our current
-                 * subvol in its directory tree.  So, this loops through
-                 * the dentries and checks the forward refs for each subvolume
-                 * to see if is references the subvolume where we are
-                 * placing this new snapshot.
-                 */
-                while (1) {
-                        if (!test ||
-                            dir == snap_src->fs_info->sb->s_root ||
-                            test == snap_src->fs_info->sb->s_root ||
-                            test->d_inode->i_sb != snap_src->fs_info->sb) {
-                                break;
-                        }
-                        if (S_ISLNK(test->d_inode->i_mode)) {
-                                printk(KERN_INFO "Btrfs symlink in snapshot "
-                                       "path, failed\n");
-                                error = -EMLINK;
-                                btrfs_free_path(path);
-                                goto out_drop_write;
-                        }
-                        test_oid =
-                                BTRFS_I(test->d_inode)->root->root_key.objectid;
-                        ret = btrfs_find_root_ref(snap_src->fs_info->tree_root,
-                                  path, test_oid, parent_oid);
-                        if (ret == 0) {
-                                printk(KERN_INFO "Btrfs snapshot creation "
-                                       "failed, looping\n");
-                                error = -EMLINK;
-                                btrfs_free_path(path);
-                                goto out_drop_write;
-                        }
-                        btrfs_release_path(snap_src->fs_info->tree_root, path);
-                        test = test->d_parent;
-                }
-create:
-                btrfs_free_path(path);
-                error = create_snapshot(snap_src, dentry, name, namelen);
+                error = create_snapshot(snap_src, dentry,
+                                        name, namelen);
         } else {
-                error = create_subvol(BTRFS_I(parent->dentry->d_inode)->root,
-                                      dentry, name, namelen);
+                error = create_subvol(BTRFS_I(dir)->root, dentry,
+                                      name, namelen);
         }
-        if (error)
-                goto out_drop_write;
-
-        fsnotify_mkdir(parent->dentry->d_inode, dentry);
+        if (!error)
+                fsnotify_mkdir(dir, dentry);
+out_up_read:
+        up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 out_drop_write:
         mnt_drop_write(parent->mnt);
 out_dput:
         dput(dentry);
 out_unlock:
-        mutex_unlock(&parent->dentry->d_inode->i_mutex);
+        mutex_unlock(&dir->i_mutex);
         return error;
 }
 
-
 static int btrfs_defrag_file(struct file *file)
 {
         struct inode *inode = fdentry(file)->d_inode;
@@ -597,7 +529,8 @@ out_unlock:
         return 0;
 }
 
-static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
+static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
+                                        void __user *arg)
 {
         u64 new_size;
         u64 old_size;
@@ -706,10 +639,7 @@ static noinline int btrfs_ioctl_snap_create(struct file *file,
 {
         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
         struct btrfs_ioctl_vol_args *vol_args;
-        struct btrfs_dir_item *di;
-        struct btrfs_path *path;
         struct file *src_file;
-        u64 root_dirid;
         int namelen;
         int ret = 0;
 
@@ -727,32 +657,9 @@ static noinline int btrfs_ioctl_snap_create(struct file *file,
                 goto out;
         }
 
-        path = btrfs_alloc_path();
-        if (!path) {
-                ret = -ENOMEM;
-                goto out;
-        }
-
-        root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
-        di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
-                            path, root_dirid,
-                            vol_args->name, namelen, 0);
-        btrfs_free_path(path);
-
-        if (di && !IS_ERR(di)) {
-                ret = -EEXIST;
-                goto out;
-        }
-
-        if (IS_ERR(di)) {
-                ret = PTR_ERR(di);
-                goto out;
-        }
-
         if (subvol) {
-                ret = btrfs_mksubvol(&file->f_path, vol_args->name,
-                                     file->f_path.dentry->d_inode->i_mode,
-                                     namelen, NULL);
+                ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
+                                     NULL);
         } else {
                 struct inode *src_inode;
                 src_file = fget(vol_args->fd);
@@ -769,17 +676,156 @@ static noinline int btrfs_ioctl_snap_create(struct file *file,
                         fput(src_file);
                         goto out;
                 }
-                ret = btrfs_mksubvol(&file->f_path, vol_args->name,
-                             file->f_path.dentry->d_inode->i_mode,
-                             namelen, BTRFS_I(src_inode)->root);
+                ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
+                                     BTRFS_I(src_inode)->root);
                 fput(src_file);
         }
-
 out:
         kfree(vol_args);
         return ret;
 }
 
+/*
+ * helper to check if the subvolume references other subvolumes
+ */
+static noinline int may_destroy_subvol(struct btrfs_root *root)
+{
+        struct btrfs_path *path;
+        struct btrfs_key key;
+        int ret;
+
+        path = btrfs_alloc_path();
+        if (!path)
+                return -ENOMEM;
+
+        key.objectid = root->root_key.objectid;
+        key.type = BTRFS_ROOT_REF_KEY;
+        key.offset = (u64)-1;
+
+        ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
+                                &key, path, 0, 0);
+        if (ret < 0)
+                goto out;
+        BUG_ON(ret == 0);
+
+        ret = 0;
+        if (path->slots[0] > 0) {
+                path->slots[0]--;
+                btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
+                if (key.objectid == root->root_key.objectid &&
+                    key.type == BTRFS_ROOT_REF_KEY)
+                        ret = -ENOTEMPTY;
+        }
+out:
+        btrfs_free_path(path);
+        return ret;
+}
+
+static noinline int btrfs_ioctl_snap_destroy(struct file *file,
+                                             void __user *arg)
+{
+        struct dentry *parent = fdentry(file);
+        struct dentry *dentry;
+        struct inode *dir = parent->d_inode;
+        struct inode *inode;
+        struct btrfs_root *root = BTRFS_I(dir)->root;
+        struct btrfs_root *dest = NULL;
+        struct btrfs_ioctl_vol_args *vol_args;
+        struct btrfs_trans_handle *trans;
+        int namelen;
+        int ret;
+        int err = 0;
+
+        if (!capable(CAP_SYS_ADMIN))
+                return -EPERM;
+
+        vol_args = memdup_user(arg, sizeof(*vol_args));
+        if (IS_ERR(vol_args))
+                return PTR_ERR(vol_args);
+
+        vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
+        namelen = strlen(vol_args->name);
+        if (strchr(vol_args->name, '/') ||
+            strncmp(vol_args->name, "..", namelen) == 0) {
+                err = -EINVAL;
+                goto out;
+        }
+
+        err = mnt_want_write(file->f_path.mnt);
+        if (err)
+                goto out;
+
+        mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
+        dentry = lookup_one_len(vol_args->name, parent, namelen);
+        if (IS_ERR(dentry)) {
+                err = PTR_ERR(dentry);
+                goto out_unlock_dir;
+        }
+
+        if (!dentry->d_inode) {
+                err = -ENOENT;
+                goto out_dput;
+        }
+
+        inode = dentry->d_inode;
+        if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
+                err = -EINVAL;
+                goto out_dput;
+        }
+
+        dest = BTRFS_I(inode)->root;
+
+        mutex_lock(&inode->i_mutex);
+        err = d_invalidate(dentry);
+        if (err)
+                goto out_unlock;
+
+        down_write(&root->fs_info->subvol_sem);
+
+        err = may_destroy_subvol(dest);
+        if (err)
+                goto out_up_write;
+
+        trans = btrfs_start_transaction(root, 1);
+        ret = btrfs_unlink_subvol(trans, root, dir,
+                                dest->root_key.objectid,
+                                dentry->d_name.name,
+                                dentry->d_name.len);
+        BUG_ON(ret);
+
+        btrfs_record_root_in_trans(trans, dest);
+
+        memset(&dest->root_item.drop_progress, 0,
+                sizeof(dest->root_item.drop_progress));
+        dest->root_item.drop_level = 0;
+        btrfs_set_root_refs(&dest->root_item, 0);
+
+        ret = btrfs_insert_orphan_item(trans,
+                                root->fs_info->tree_root,
+                                dest->root_key.objectid);
+        BUG_ON(ret);
+
+        ret = btrfs_commit_transaction(trans, root);
+        BUG_ON(ret);
+        inode->i_flags |= S_DEAD;
+out_up_write:
+        up_write(&root->fs_info->subvol_sem);
+out_unlock:
+        mutex_unlock(&inode->i_mutex);
+        if (!err) {
+                btrfs_invalidate_inodes(dest);
+                d_delete(dentry);
+        }
+out_dput:
+        dput(dentry);
+out_unlock_dir:
+        mutex_unlock(&dir->i_mutex);
+        mnt_drop_write(file->f_path.mnt);
+out:
+        kfree(vol_args);
+        return err;
+}
+
 static int btrfs_ioctl_defrag(struct file *file)
 {
         struct inode *inode = fdentry(file)->d_inode;
@@ -853,8 +899,8 @@ static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
         return ret;
 }
 
-static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
-                u64 off, u64 olen, u64 destoff)
+static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
+                                       u64 off, u64 olen, u64 destoff)
 {
         struct inode *inode = fdentry(file)->d_inode;
         struct btrfs_root *root = BTRFS_I(inode)->root;
@@ -1246,6 +1292,8 @@ long btrfs_ioctl(struct file *file, unsigned int
                 return btrfs_ioctl_snap_create(file, argp, 0);
         case BTRFS_IOC_SUBVOL_CREATE:
                 return btrfs_ioctl_snap_create(file, argp, 1);
+        case BTRFS_IOC_SNAP_DESTROY:
+                return btrfs_ioctl_snap_destroy(file, argp);
         case BTRFS_IOC_DEFRAG:
                 return btrfs_ioctl_defrag(file);
         case BTRFS_IOC_RESIZE:
diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h
index b320b103fa13..bc49914475eb 100644
--- a/fs/btrfs/ioctl.h
+++ b/fs/btrfs/ioctl.h
@@ -65,5 +65,6 @@ struct btrfs_ioctl_clone_range_args {
 
 #define BTRFS_IOC_SUBVOL_CREATE _IOW(BTRFS_IOCTL_MAGIC, 14, \
                                    struct btrfs_ioctl_vol_args)
-
+#define BTRFS_IOC_SNAP_DESTROY _IOW(BTRFS_IOCTL_MAGIC, 15, \
+                                struct btrfs_ioctl_vol_args)
 #endif
diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index 3be16ccc7eea..48a504260635 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -3203,6 +3203,7 @@ static int check_extent_flags(u64 flags)
         return 0;
 }
 
+
 static noinline_for_stack int relocate_block_group(struct reloc_control *rc)
 {
         struct rb_root blocks = RB_ROOT;
@@ -3220,6 +3221,9 @@ static noinline_for_stack int relocate_block_group(struct reloc_control *rc)
         if (!path)
                 return -ENOMEM;
 
+        rc->extents_found = 0;
+        rc->extents_skipped = 0;
+
         rc->search_start = rc->block_group->key.objectid;
         clear_extent_bits(&rc->processed_blocks, 0, (u64)-1, EXTENT_DIRTY,
                           GFP_NOFS);
@@ -3475,14 +3479,15 @@ int btrfs_relocate_block_group(struct btrfs_root *extent_root, u64 group_start)
         btrfs_wait_ordered_extents(fs_info->tree_root, 0);
 
         while (1) {
-                mutex_lock(&fs_info->cleaner_mutex);
-                btrfs_clean_old_snapshots(fs_info->tree_root);
-                mutex_unlock(&fs_info->cleaner_mutex);
-
                 rc->extents_found = 0;
                 rc->extents_skipped = 0;
 
+                mutex_lock(&fs_info->cleaner_mutex);
+
+                btrfs_clean_old_snapshots(fs_info->tree_root);
                 ret = relocate_block_group(rc);
+
+                mutex_unlock(&fs_info->cleaner_mutex);
                 if (ret < 0) {
                         err = ret;
                         break;
@@ -3530,6 +3535,26 @@ out:
         return err;
 }
 
+static noinline_for_stack int mark_garbage_root(struct btrfs_root *root)
+{
+        struct btrfs_trans_handle *trans;
+        int ret;
+
+        trans = btrfs_start_transaction(root->fs_info->tree_root, 1);
+
+        memset(&root->root_item.drop_progress, 0,
+                sizeof(root->root_item.drop_progress));
+        root->root_item.drop_level = 0;
+        btrfs_set_root_refs(&root->root_item, 0);
+        ret = btrfs_update_root(trans, root->fs_info->tree_root,
+                                &root->root_key, &root->root_item);
+        BUG_ON(ret);
+
+        ret = btrfs_end_transaction(trans, root->fs_info->tree_root);
+        BUG_ON(ret);
+        return 0;
+}
+
 /*
  * recover relocation interrupted by system crash.
  *
@@ -3589,8 +3614,12 @@ int btrfs_recover_relocation(struct btrfs_root *root)
                         fs_root = read_fs_root(root->fs_info,
                                                reloc_root->root_key.offset);
                         if (IS_ERR(fs_root)) {
-                                err = PTR_ERR(fs_root);
-                                goto out;
+                                ret = PTR_ERR(fs_root);
+                                if (ret != -ENOENT) {
+                                        err = ret;
+                                        goto out;
+                                }
+                                mark_garbage_root(reloc_root);
                         }
                 }
 
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
index 5ef72599a581..9351428f30e2 100644
--- a/fs/btrfs/root-tree.c
+++ b/fs/btrfs/root-tree.c
@@ -94,17 +94,23 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
                 goto out;
 
         BUG_ON(ret == 0);
+        if (path->slots[0] == 0) {
+                ret = 1;
+                goto out;
+        }
         l = path->nodes[0];
-        BUG_ON(path->slots[0] == 0);
         slot = path->slots[0] - 1;
         btrfs_item_key_to_cpu(l, &found_key, slot);
-        if (found_key.objectid != objectid) {
+        if (found_key.objectid != objectid ||
+            found_key.type != BTRFS_ROOT_ITEM_KEY) {
                 ret = 1;
                 goto out;
         }
-        read_extent_buffer(l, item, btrfs_item_ptr_offset(l, slot),
-                           sizeof(*item));
-        memcpy(key, &found_key, sizeof(found_key));
+        if (item)
+                read_extent_buffer(l, item, btrfs_item_ptr_offset(l, slot),
+                                   sizeof(*item));
+        if (key)
+                memcpy(key, &found_key, sizeof(found_key));
         ret = 0;
 out:
         btrfs_free_path(path);
@@ -249,6 +255,59 @@ err:
         return ret;
 }
 
+int btrfs_find_orphan_roots(struct btrfs_root *tree_root)
+{
+        struct extent_buffer *leaf;
+        struct btrfs_path *path;
+        struct btrfs_key key;
+        int err = 0;
+        int ret;
+
+        path = btrfs_alloc_path();
+        if (!path)
+                return -ENOMEM;
+
+        key.objectid = BTRFS_ORPHAN_OBJECTID;
+        key.type = BTRFS_ORPHAN_ITEM_KEY;
+        key.offset = 0;
+
+        while (1) {
+                ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0);
+                if (ret < 0) {
+                        err = ret;
+                        break;
+                }
+
+                leaf = path->nodes[0];
+                if (path->slots[0] >= btrfs_header_nritems(leaf)) {
+                        ret = btrfs_next_leaf(tree_root, path);
+                        if (ret < 0)
+                                err = ret;
+                        if (ret != 0)
+                                break;
+                        leaf = path->nodes[0];
+                }
+
+                btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+                btrfs_release_path(tree_root, path);
+
+                if (key.objectid != BTRFS_ORPHAN_OBJECTID ||
+                    key.type != BTRFS_ORPHAN_ITEM_KEY)
+                        break;
+
+                ret = btrfs_find_dead_roots(tree_root, key.offset);
+                if (ret) {
+                        err = ret;
+                        break;
+                }
+
+                key.offset++;
+        }
+
+        btrfs_free_path(path);
+        return err;
+}
+
 /* drop the root item for 'key' from 'root' */
 int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
                    struct btrfs_key *key)
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 6d6d06cb6dfc..0242c8babae9 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -676,6 +676,7 @@ static int btrfs_unfreeze(struct super_block *sb)
 }
 
 static struct super_operations btrfs_super_ops = {
+        .drop_inode     = btrfs_drop_inode,
         .delete_inode   = btrfs_delete_inode,
         .put_super      = btrfs_put_super,
         .sync_fs        = btrfs_sync_fs,
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 915077725fea..88f866f85e7a 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -104,7 +104,6 @@ static noinline int record_root_in_trans(struct btrfs_trans_handle *trans,
 {
         if (root->ref_cows && root->last_trans < trans->transid) {
                 WARN_ON(root == root->fs_info->extent_root);
-                WARN_ON(root->root_item.refs == 0);
                 WARN_ON(root->commit_root != root->node);
 
                 radix_tree_tag_set(&root->fs_info->fs_roots_radix,
@@ -1078,8 +1077,13 @@ int btrfs_clean_old_snapshots(struct btrfs_root *root)
 
         while (!list_empty(&list)) {
                 root = list_entry(list.next, struct btrfs_root, root_list);
-                list_del_init(&root->root_list);
-                btrfs_drop_snapshot(root, 0);
+                list_del(&root->root_list);
+
+                if (btrfs_header_backref_rev(root->node) <
+                    BTRFS_MIXED_BACKREF_REV)
+                        btrfs_drop_snapshot(root, 0);
+                else
+                        btrfs_drop_snapshot(root, 1);
         }
         return 0;
 }
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 6e674d76186d..4d7d9abef42f 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -2841,7 +2841,7 @@ static noinline int check_parent_dirs_for_sync(struct btrfs_trans_handle *trans,
                 if (!parent || !parent->d_inode || sb != parent->d_inode->i_sb)
                         break;
 
-                if (parent == sb->s_root)
+                if (IS_ROOT(parent))
                         break;
 
                 parent = parent->d_parent;
@@ -2880,6 +2880,12 @@ int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
                 goto end_no_trans;
         }
 
+        if (root != BTRFS_I(inode)->root ||
+            btrfs_root_refs(&root->root_item) == 0) {
+                ret = 1;
+                goto end_no_trans;
+        }
+
         ret = check_parent_dirs_for_sync(trans, inode, parent,
                                          sb, last_committed);
         if (ret)
@@ -2907,12 +2913,15 @@ int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
                         break;
 
                 inode = parent->d_inode;
+                if (root != BTRFS_I(inode)->root)
+                        break;
+
                 if (BTRFS_I(inode)->generation >
                     root->fs_info->last_trans_committed) {
                         ret = btrfs_log_inode(trans, root, inode, inode_only);
                         BUG_ON(ret);
                 }
-                if (parent == sb->s_root)
+                if (IS_ROOT(parent))
                         break;
 
                 parent = parent->d_parent;