1 files changed, 681 insertions, 306 deletions

diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index b72b068183ec..d58cce77fc6c 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -22,6 +22,7 @@
 #include <linux/writeback.h>
 #include <linux/pagemap.h>
 #include <linux/blkdev.h>
+#include <linux/uuid.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
@@ -29,17 +30,47 @@
 #include "tree-log.h"
 #include "inode-map.h"
 #include "volumes.h"
+#include "dev-replace.h"
 
 #define BTRFS_ROOT_TRANS_TAG 0
 
-void put_transaction(struct btrfs_transaction *transaction)
+static unsigned int btrfs_blocked_trans_types[TRANS_STATE_MAX] = {
+	[TRANS_STATE_RUNNING]		= 0U,
+	[TRANS_STATE_BLOCKED]		= (__TRANS_USERSPACE |
+					   __TRANS_START),
+	[TRANS_STATE_COMMIT_START]	= (__TRANS_USERSPACE |
+					   __TRANS_START |
+					   __TRANS_ATTACH),
+	[TRANS_STATE_COMMIT_DOING]	= (__TRANS_USERSPACE |
+					   __TRANS_START |
+					   __TRANS_ATTACH |
+					   __TRANS_JOIN),
+	[TRANS_STATE_UNBLOCKED]		= (__TRANS_USERSPACE |
+					   __TRANS_START |
+					   __TRANS_ATTACH |
+					   __TRANS_JOIN |
+					   __TRANS_JOIN_NOLOCK),
+	[TRANS_STATE_COMPLETED]		= (__TRANS_USERSPACE |
+					   __TRANS_START |
+					   __TRANS_ATTACH |
+					   __TRANS_JOIN |
+					   __TRANS_JOIN_NOLOCK),
+};
+
+static void put_transaction(struct btrfs_transaction *transaction)
 {
 	WARN_ON(atomic_read(&transaction->use_count) == 0);
 	if (atomic_dec_and_test(&transaction->use_count)) {
 		BUG_ON(!list_empty(&transaction->list));
 		WARN_ON(transaction->delayed_refs.root.rb_node);
-		WARN_ON(!list_empty(&transaction->delayed_refs.seq_head));
-		memset(transaction, 0, sizeof(*transaction));
+		while (!list_empty(&transaction->pending_chunks)) {
+			struct extent_map *em;
+
+			em = list_first_entry(&transaction->pending_chunks,
+					      struct extent_map, list);
+			list_del_init(&em->list);
+			free_extent_map(em);
+		}
 		kmem_cache_free(btrfs_transaction_cachep, transaction);
 	}
 }
@@ -50,10 +81,35 @@ static noinline void switch_commit_root(struct btrfs_root *root)
 	root->commit_root = btrfs_root_node(root);
 }
 
+static inline void extwriter_counter_inc(struct btrfs_transaction *trans,
+					 unsigned int type)
+{
+	if (type & TRANS_EXTWRITERS)
+		atomic_inc(&trans->num_extwriters);
+}
+
+static inline void extwriter_counter_dec(struct btrfs_transaction *trans,
+					 unsigned int type)
+{
+	if (type & TRANS_EXTWRITERS)
+		atomic_dec(&trans->num_extwriters);
+}
+
+static inline void extwriter_counter_init(struct btrfs_transaction *trans,
+					  unsigned int type)
+{
+	atomic_set(&trans->num_extwriters, ((type & TRANS_EXTWRITERS) ? 1 : 0));
+}
+
+static inline int extwriter_counter_read(struct btrfs_transaction *trans)
+{
+	return atomic_read(&trans->num_extwriters);
+}
+
 /*
  * either allocate a new transaction or hop into the existing one
  */
-static noinline int join_transaction(struct btrfs_root *root, int nofail)
+static noinline int join_transaction(struct btrfs_root *root, unsigned int type)
 {
 	struct btrfs_transaction *cur_trans;
 	struct btrfs_fs_info *fs_info = root->fs_info;
@@ -61,32 +117,42 @@ static noinline int join_transaction(struct btrfs_root *root, int nofail)
 	spin_lock(&fs_info->trans_lock);
 loop:
 	/* The file system has been taken offline. No new transactions. */
-	if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
+	if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
 		spin_unlock(&fs_info->trans_lock);
 		return -EROFS;
 	}
 
-	if (fs_info->trans_no_join) {
-		if (!nofail) {
-			spin_unlock(&fs_info->trans_lock);
-			return -EBUSY;
-		}
-	}
-
 	cur_trans = fs_info->running_transaction;
 	if (cur_trans) {
 		if (cur_trans->aborted) {
 			spin_unlock(&fs_info->trans_lock);
 			return cur_trans->aborted;
 		}
+		if (btrfs_blocked_trans_types[cur_trans->state] & type) {
+			spin_unlock(&fs_info->trans_lock);
+			return -EBUSY;
+		}
 		atomic_inc(&cur_trans->use_count);
 		atomic_inc(&cur_trans->num_writers);
-		cur_trans->num_joined++;
+		extwriter_counter_inc(cur_trans, type);
 		spin_unlock(&fs_info->trans_lock);
 		return 0;
 	}
 	spin_unlock(&fs_info->trans_lock);
 
+	/*
+	 * If we are ATTACH, we just want to catch the current transaction,
+	 * and commit it. If there is no transaction, just return ENOENT.
+	 */
+	if (type == TRANS_ATTACH)
+		return -ENOENT;
+
+	/*
+	 * JOIN_NOLOCK only happens during the transaction commit, so
+	 * it is impossible that ->running_transaction is NULL
+	 */
+	BUG_ON(type == TRANS_JOIN_NOLOCK);
+
 	cur_trans = kmem_cache_alloc(btrfs_transaction_cachep, GFP_NOFS);
 	if (!cur_trans)
 		return -ENOMEM;
@@ -95,29 +161,26 @@ loop:
 	if (fs_info->running_transaction) {
 		/*
 		 * someone started a transaction after we unlocked.  Make sure
-		 * to redo the trans_no_join checks above
+		 * to redo the checks above
 		 */
 		kmem_cache_free(btrfs_transaction_cachep, cur_trans);
-		cur_trans = fs_info->running_transaction;
 		goto loop;
-	} else if (root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
-		spin_unlock(&root->fs_info->trans_lock);
+	} else if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
+		spin_unlock(&fs_info->trans_lock);
 		kmem_cache_free(btrfs_transaction_cachep, cur_trans);
 		return -EROFS;
 	}
 
 	atomic_set(&cur_trans->num_writers, 1);
-	cur_trans->num_joined = 0;
+	extwriter_counter_init(cur_trans, type);
 	init_waitqueue_head(&cur_trans->writer_wait);
 	init_waitqueue_head(&cur_trans->commit_wait);
-	cur_trans->in_commit = 0;
-	cur_trans->blocked = 0;
+	cur_trans->state = TRANS_STATE_RUNNING;
 	/*
 	 * One for this trans handle, one so it will live on until we
 	 * commit the transaction.
 	 */
 	atomic_set(&cur_trans->use_count, 2);
-	cur_trans->commit_done = 0;
 	cur_trans->start_time = get_seconds();
 
 	cur_trans->delayed_refs.root = RB_ROOT;
@@ -126,31 +189,28 @@ loop:
 	cur_trans->delayed_refs.num_heads = 0;
 	cur_trans->delayed_refs.flushing = 0;
 	cur_trans->delayed_refs.run_delayed_start = 0;
-	cur_trans->delayed_refs.seq = 1;
 
 	/*
 	 * although the tree mod log is per file system and not per transaction,
 	 * the log must never go across transaction boundaries.
 	 */
 	smp_mb();
-	if (!list_empty(&fs_info->tree_mod_seq_list)) {
-		printk(KERN_ERR "btrfs: tree_mod_seq_list not empty when "
+	if (!list_empty(&fs_info->tree_mod_seq_list))
+		WARN(1, KERN_ERR "btrfs: tree_mod_seq_list not empty when "
 			"creating a fresh transaction\n");
-		WARN_ON(1);
-	}
-	if (!RB_EMPTY_ROOT(&fs_info->tree_mod_log)) {
-		printk(KERN_ERR "btrfs: tree_mod_log rb tree not empty when "
+	if (!RB_EMPTY_ROOT(&fs_info->tree_mod_log))
+		WARN(1, KERN_ERR "btrfs: tree_mod_log rb tree not empty when "
 			"creating a fresh transaction\n");
-		WARN_ON(1);
-	}
-	atomic_set(&fs_info->tree_mod_seq, 0);
+	atomic64_set(&fs_info->tree_mod_seq, 0);
 
-	init_waitqueue_head(&cur_trans->delayed_refs.seq_wait);
-	spin_lock_init(&cur_trans->commit_lock);
 	spin_lock_init(&cur_trans->delayed_refs.lock);
-	INIT_LIST_HEAD(&cur_trans->delayed_refs.seq_head);
+	atomic_set(&cur_trans->delayed_refs.procs_running_refs, 0);
+	atomic_set(&cur_trans->delayed_refs.ref_seq, 0);
+	init_waitqueue_head(&cur_trans->delayed_refs.wait);
 
 	INIT_LIST_HEAD(&cur_trans->pending_snapshots);
+	INIT_LIST_HEAD(&cur_trans->ordered_operations);
+	INIT_LIST_HEAD(&cur_trans->pending_chunks);
 	list_add_tail(&cur_trans->list, &fs_info->trans_list);
 	extent_io_tree_init(&cur_trans->dirty_pages,
 			     fs_info->btree_inode->i_mapping);
@@ -248,6 +308,13 @@ int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
+static inline int is_transaction_blocked(struct btrfs_transaction *trans)
+{
+	return (trans->state >= TRANS_STATE_BLOCKED &&
+		trans->state < TRANS_STATE_UNBLOCKED &&
+		!trans->aborted);
+}
+
 /* wait for commit against the current transaction to become unblocked
  * when this is done, it is safe to start a new transaction, but the current
  * transaction might not be fully on disk.
@@ -258,25 +325,19 @@ static void wait_current_trans(struct btrfs_root *root)
 
 	spin_lock(&root->fs_info->trans_lock);
 	cur_trans = root->fs_info->running_transaction;
-	if (cur_trans && cur_trans->blocked) {
+	if (cur_trans && is_transaction_blocked(cur_trans)) {
 		atomic_inc(&cur_trans->use_count);
 		spin_unlock(&root->fs_info->trans_lock);
 
 		wait_event(root->fs_info->transaction_wait,
-			   !cur_trans->blocked);
+			   cur_trans->state >= TRANS_STATE_UNBLOCKED ||
+			   cur_trans->aborted);
 		put_transaction(cur_trans);
 	} else {
 		spin_unlock(&root->fs_info->trans_lock);
 	}
 }
 
-enum btrfs_trans_type {
-	TRANS_START,
-	TRANS_JOIN,
-	TRANS_USERSPACE,
-	TRANS_JOIN_NOLOCK,
-};
-
 static int may_wait_transaction(struct btrfs_root *root, int type)
 {
 	if (root->fs_info->log_root_recovering)
@@ -292,21 +353,24 @@ static int may_wait_transaction(struct btrfs_root *root, int type)
 	return 0;
 }
 
-static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
-						    u64 num_items, int type)
+static struct btrfs_trans_handle *
+start_transaction(struct btrfs_root *root, u64 num_items, unsigned int type,
+		  enum btrfs_reserve_flush_enum flush)
 {
 	struct btrfs_trans_handle *h;
 	struct btrfs_transaction *cur_trans;
 	u64 num_bytes = 0;
 	int ret;
+	u64 qgroup_reserved = 0;
 
-	if (root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
+	if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state))
 		return ERR_PTR(-EROFS);
 
 	if (current->journal_info) {
-		WARN_ON(type != TRANS_JOIN && type != TRANS_JOIN_NOLOCK);
+		WARN_ON(type & TRANS_EXTWRITERS);
 		h = current->journal_info;
 		h->use_count++;
+		WARN_ON(h->use_count > 2);
 		h->orig_rsv = h->block_rsv;
 		h->block_rsv = NULL;
 		goto got_it;
@@ -317,30 +381,57 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
 	 * the appropriate flushing if need be.
 	 */
 	if (num_items > 0 && root != root->fs_info->chunk_root) {
+		if (root->fs_info->quota_enabled &&
+		    is_fstree(root->root_key.objectid)) {
+			qgroup_reserved = num_items * root->leafsize;
+			ret = btrfs_qgroup_reserve(root, qgroup_reserved);
+			if (ret)
+				return ERR_PTR(ret);
+		}
+
 		num_bytes = btrfs_calc_trans_metadata_size(root, num_items);
 		ret = btrfs_block_rsv_add(root,
 					  &root->fs_info->trans_block_rsv,
-					  num_bytes);
+					  num_bytes, flush);
 		if (ret)
-			return ERR_PTR(ret);
+			goto reserve_fail;
 	}
 again:
 	h = kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
-	if (!h)
-		return ERR_PTR(-ENOMEM);
+	if (!h) {
+		ret = -ENOMEM;
+		goto alloc_fail;
+	}
+
+	/*
+	 * If we are JOIN_NOLOCK we're already committing a transaction and
+	 * waiting on this guy, so we don't need to do the sb_start_intwrite
+	 * because we're already holding a ref.  We need this because we could
+	 * have raced in and did an fsync() on a file which can kick a commit
+	 * and then we deadlock with somebody doing a freeze.
+	 *
+	 * If we are ATTACH, it means we just want to catch the current
+	 * transaction and commit it, so we needn't do sb_start_intwrite(). 
+	 */
+	if (type & __TRANS_FREEZABLE)
+		sb_start_intwrite(root->fs_info->sb);
 
 	if (may_wait_transaction(root, type))
 		wait_current_trans(root);
 
 	do {
-		ret = join_transaction(root, type == TRANS_JOIN_NOLOCK);
-		if (ret == -EBUSY)
+		ret = join_transaction(root, type);
+		if (ret == -EBUSY) {
 			wait_current_trans(root);
+			if (unlikely(type == TRANS_ATTACH))
+				ret = -ENOENT;
+		}
 	} while (ret == -EBUSY);
 
 	if (ret < 0) {
-		kmem_cache_free(btrfs_trans_handle_cachep, h);
-		return ERR_PTR(ret);
+		/* We must get the transaction if we are JOIN_NOLOCK. */
+		BUG_ON(type == TRANS_JOIN_NOLOCK);
+		goto join_fail;
 	}
 
 	cur_trans = root->fs_info->running_transaction;
@@ -349,14 +440,23 @@ again:
 	h->transaction = cur_trans;
 	h->blocks_used = 0;
 	h->bytes_reserved = 0;
+	h->root = root;
 	h->delayed_ref_updates = 0;
 	h->use_count = 1;
+	h->adding_csums = 0;
 	h->block_rsv = NULL;
 	h->orig_rsv = NULL;
 	h->aborted = 0;
+	h->qgroup_reserved = 0;
+	h->delayed_ref_elem.seq = 0;
+	h->type = type;
+	h->allocating_chunk = false;
+	INIT_LIST_HEAD(&h->qgroup_ref_list);
+	INIT_LIST_HEAD(&h->new_bgs);
 
 	smp_mb();
-	if (cur_trans->blocked && may_wait_transaction(root, type)) {
+	if (cur_trans->state >= TRANS_STATE_BLOCKED &&
+	    may_wait_transaction(root, type)) {
 		btrfs_commit_transaction(h, root);
 		goto again;
 	}
@@ -367,6 +467,7 @@ again:
 		h->block_rsv = &root->fs_info->trans_block_rsv;
 		h->bytes_reserved = num_bytes;
 	}
+	h->qgroup_reserved = qgroup_reserved;
 
 got_it:
 	btrfs_record_root_in_trans(h, root);
@@ -374,67 +475,129 @@ got_it:
 	if (!current->journal_info && type != TRANS_USERSPACE)
 		current->journal_info = h;
 	return h;
+
+join_fail:
+	if (type & __TRANS_FREEZABLE)
+		sb_end_intwrite(root->fs_info->sb);
+	kmem_cache_free(btrfs_trans_handle_cachep, h);
+alloc_fail:
+	if (num_bytes)
+		btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv,
+					num_bytes);
+reserve_fail:
+	if (qgroup_reserved)
+		btrfs_qgroup_free(root, qgroup_reserved);
+	return ERR_PTR(ret);
 }
 
 struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
 						   int num_items)
 {
-	return start_transaction(root, num_items, TRANS_START);
+	return start_transaction(root, num_items, TRANS_START,
+				 BTRFS_RESERVE_FLUSH_ALL);
 }
+
+struct btrfs_trans_handle *btrfs_start_transaction_lflush(
+					struct btrfs_root *root, int num_items)
+{
+	return start_transaction(root, num_items, TRANS_START,
+				 BTRFS_RESERVE_FLUSH_LIMIT);
+}
+
 struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root)
 {
-	return start_transaction(root, 0, TRANS_JOIN);
+	return start_transaction(root, 0, TRANS_JOIN, 0);
 }
 
 struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root)
 {
-	return start_transaction(root, 0, TRANS_JOIN_NOLOCK);
+	return start_transaction(root, 0, TRANS_JOIN_NOLOCK, 0);
 }
 
 struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *root)
 {
-	return start_transaction(root, 0, TRANS_USERSPACE);
+	return start_transaction(root, 0, TRANS_USERSPACE, 0);
+}
+
+/*
+ * btrfs_attach_transaction() - catch the running transaction
+ *
+ * It is used when we want to commit the current the transaction, but
+ * don't want to start a new one.
+ *
+ * Note: If this function return -ENOENT, it just means there is no
+ * running transaction. But it is possible that the inactive transaction
+ * is still in the memory, not fully on disk. If you hope there is no
+ * inactive transaction in the fs when -ENOENT is returned, you should
+ * invoke
+ *     btrfs_attach_transaction_barrier()
+ */
+struct btrfs_trans_handle *btrfs_attach_transaction(struct btrfs_root *root)
+{
+	return start_transaction(root, 0, TRANS_ATTACH, 0);
+}
+
+/*
+ * btrfs_attach_transaction_barrier() - catch the running transaction
+ *
+ * It is similar to the above function, the differentia is this one
+ * will wait for all the inactive transactions until they fully
+ * complete.
+ */
+struct btrfs_trans_handle *
+btrfs_attach_transaction_barrier(struct btrfs_root *root)
+{
+	struct btrfs_trans_handle *trans;
+
+	trans = start_transaction(root, 0, TRANS_ATTACH, 0);
+	if (IS_ERR(trans) && PTR_ERR(trans) == -ENOENT)
+		btrfs_wait_for_commit(root, 0);
+
+	return trans;
 }
 
 /* wait for a transaction commit to be fully complete */
 static noinline void wait_for_commit(struct btrfs_root *root,
 				    struct btrfs_transaction *commit)
 {
-	wait_event(commit->commit_wait, commit->commit_done);
+	wait_event(commit->commit_wait, commit->state == TRANS_STATE_COMPLETED);
 }
 
 int btrfs_wait_for_commit(struct btrfs_root *root, u64 transid)
 {
 	struct btrfs_transaction *cur_trans = NULL, *t;
-	int ret;
+	int ret = 0;
 
-	ret = 0;
 	if (transid) {
 		if (transid <= root->fs_info->last_trans_committed)
 			goto out;
 
+		ret = -EINVAL;
 		/* find specified transaction */
 		spin_lock(&root->fs_info->trans_lock);
 		list_for_each_entry(t, &root->fs_info->trans_list, list) {
 			if (t->transid == transid) {
 				cur_trans = t;
 				atomic_inc(&cur_trans->use_count);
+				ret = 0;
 				break;
 			}
-			if (t->transid > transid)
+			if (t->transid > transid) {
+				ret = 0;
 				break;
+			}
 		}
 		spin_unlock(&root->fs_info->trans_lock);
-		ret = -EINVAL;
+		/* The specified transaction doesn't exist */
 		if (!cur_trans)
-			goto out;  /* bad transid */
+			goto out;
 	} else {
 		/* find newest transaction that is committing | committed */
 		spin_lock(&root->fs_info->trans_lock);
 		list_for_each_entry_reverse(t, &root->fs_info->trans_list,
 					    list) {
-			if (t->in_commit) {
-				if (t->commit_done)
+			if (t->state >= TRANS_STATE_COMMIT_START) {
+				if (t->state == TRANS_STATE_COMPLETED)
 					break;
 				cur_trans = t;
 				atomic_inc(&cur_trans->use_count);
@@ -447,9 +610,7 @@ int btrfs_wait_for_commit(struct btrfs_root *root, u64 transid)
 	}
 
 	wait_for_commit(root, cur_trans);
-
 	put_transaction(cur_trans);
-	ret = 0;
 out:
 	return ret;
 }
@@ -463,30 +624,25 @@ void btrfs_throttle(struct btrfs_root *root)
 static int should_end_transaction(struct btrfs_trans_handle *trans,
 				  struct btrfs_root *root)
 {
-	int ret;
+	if (root->fs_info->global_block_rsv.space_info->full &&
+	    btrfs_should_throttle_delayed_refs(trans, root))
+		return 1;
 
-	ret = btrfs_block_rsv_check(root, &root->fs_info->global_block_rsv, 5);
-	return ret ? 1 : 0;
+	return !!btrfs_block_rsv_check(root, &root->fs_info->global_block_rsv, 5);
 }
 
 int btrfs_should_end_transaction(struct btrfs_trans_handle *trans,
 				 struct btrfs_root *root)
 {
 	struct btrfs_transaction *cur_trans = trans->transaction;
-	struct btrfs_block_rsv *rsv = trans->block_rsv;
 	int updates;
 	int err;
 
 	smp_mb();
-	if (cur_trans->blocked || cur_trans->delayed_refs.flushing)
+	if (cur_trans->state >= TRANS_STATE_BLOCKED ||
+	    cur_trans->delayed_refs.flushing)
 		return 1;
 
-	/*
-	 * We need to do this in case we're deleting csums so the global block
-	 * rsv get's used instead of the csum block rsv.
-	 */
-	trans->block_rsv = NULL;
-
 	updates = trans->delayed_ref_updates;
 	trans->delayed_ref_updates = 0;
 	if (updates) {
@@ -495,17 +651,16 @@ int btrfs_should_end_transaction(struct btrfs_trans_handle *trans,
 			return err;
 	}
 
-	trans->block_rsv = rsv;
-
 	return should_end_transaction(trans, root);
 }
 
 static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root, int throttle, int lock)
+			  struct btrfs_root *root, int throttle)
 {
 	struct btrfs_transaction *cur_trans = trans->transaction;
 	struct btrfs_fs_info *info = root->fs_info;
-	int count = 0;
+	unsigned long cur = trans->delayed_ref_updates;
+	int lock = (trans->type != TRANS_JOIN_NOLOCK);
 	int err = 0;
 
 	if (--trans->use_count) {
@@ -513,28 +668,49 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
 		return 0;
 	}
 
+	/*
+	 * do the qgroup accounting as early as possible
+	 */
+	err = btrfs_delayed_refs_qgroup_accounting(trans, info);
+
 	btrfs_trans_release_metadata(trans, root);
 	trans->block_rsv = NULL;
-	while (count < 2) {
-		unsigned long cur = trans->delayed_ref_updates;
+
+	if (trans->qgroup_reserved) {
+		/*
+		 * the same root has to be passed here between start_transaction
+		 * and end_transaction. Subvolume quota depends on this.
+		 */
+		btrfs_qgroup_free(trans->root, trans->qgroup_reserved);
+		trans->qgroup_reserved = 0;
+	}
+
+	if (!list_empty(&trans->new_bgs))
+		btrfs_create_pending_block_groups(trans, root);
+
+	trans->delayed_ref_updates = 0;
+	if (btrfs_should_throttle_delayed_refs(trans, root)) {
+		cur = max_t(unsigned long, cur, 1);
 		trans->delayed_ref_updates = 0;
-		if (cur &&
-		    trans->transaction->delayed_refs.num_heads_ready > 64) {
-			trans->delayed_ref_updates = 0;
-			btrfs_run_delayed_refs(trans, root, cur);
-		} else {
-			break;
-		}
-		count++;
+		btrfs_run_delayed_refs(trans, root, cur);
 	}
 
+	btrfs_trans_release_metadata(trans, root);
+	trans->block_rsv = NULL;
+
+	if (!list_empty(&trans->new_bgs))
+		btrfs_create_pending_block_groups(trans, root);
+
 	if (lock && !atomic_read(&root->fs_info->open_ioctl_trans) &&
-	    should_end_transaction(trans, root)) {
-		trans->transaction->blocked = 1;
-		smp_wmb();
+	    should_end_transaction(trans, root) &&
+	    ACCESS_ONCE(cur_trans->state) == TRANS_STATE_RUNNING) {
+		spin_lock(&info->trans_lock);
+		if (cur_trans->state == TRANS_STATE_RUNNING)
+			cur_trans->state = TRANS_STATE_BLOCKED;
+		spin_unlock(&info->trans_lock);
 	}
 
-	if (lock && cur_trans->blocked && !cur_trans->in_commit) {
+	if (lock && ACCESS_ONCE(cur_trans->state) == TRANS_STATE_BLOCKED) {
 		if (throttle) {
 			/*
 			 * We may race with somebody else here so end up having
@@ -548,9 +724,13 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
 		}
 	}
 
+	if (trans->type & __TRANS_FREEZABLE)
+		sb_end_intwrite(root->fs_info->sb);
+
 	WARN_ON(cur_trans != info->running_transaction);
 	WARN_ON(atomic_read(&cur_trans->num_writers) < 1);
 	atomic_dec(&cur_trans->num_writers);
+	extwriter_counter_dec(cur_trans, trans->type);
 
 	smp_mb();
 	if (waitqueue_active(&cur_trans->writer_wait))
@@ -564,11 +744,10 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
 		btrfs_run_delayed_iputs(root);
 
 	if (trans->aborted ||
-	    root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
+	    test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state))
 		err = -EIO;
-	}
+	assert_qgroups_uptodate(trans);
 
-	memset(trans, 0, sizeof(*trans));
 	kmem_cache_free(btrfs_trans_handle_cachep, trans);
 	return err;
 }
@@ -576,40 +755,19 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
 int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root)
 {
-	int ret;
-
-	ret = __btrfs_end_transaction(trans, root, 0, 1);
-	if (ret)
-		return ret;
-	return 0;
+	return __btrfs_end_transaction(trans, root, 0);
 }
 
 int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
 				   struct btrfs_root *root)
 {
-	int ret;
-
-	ret = __btrfs_end_transaction(trans, root, 1, 1);
-	if (ret)
-		return ret;
-	return 0;
-}
-
-int btrfs_end_transaction_nolock(struct btrfs_trans_handle *trans,
-				 struct btrfs_root *root)
-{
-	int ret;
-
-	ret = __btrfs_end_transaction(trans, root, 0, 0);
-	if (ret)
-		return ret;
-	return 0;
+	return __btrfs_end_transaction(trans, root, 1);
 }
 
 int btrfs_end_transaction_dmeta(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root)
 {
-	return __btrfs_end_transaction(trans, root, 1, 1);
+	return __btrfs_end_transaction(trans, root, 1);
 }
 
 /*
@@ -623,13 +781,15 @@ int btrfs_write_marked_extents(struct btrfs_root *root,
 	int err = 0;
 	int werr = 0;
 	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
+	struct extent_state *cached_state = NULL;
 	u64 start = 0;
 	u64 end;
 
 	while (!find_first_extent_bit(dirty_pages, start, &start, &end,
-				      mark)) {
-		convert_extent_bit(dirty_pages, start, end, EXTENT_NEED_WAIT, mark,
-				   GFP_NOFS);
+				      mark, &cached_state)) {
+		convert_extent_bit(dirty_pages, start, end, EXTENT_NEED_WAIT,
+				   mark, &cached_state, GFP_NOFS);
+		cached_state = NULL;
 		err = filemap_fdatawrite_range(mapping, start, end);
 		if (err)
 			werr = err;
@@ -653,12 +813,14 @@ int btrfs_wait_marked_extents(struct btrfs_root *root,
 	int err = 0;
 	int werr = 0;
 	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
+	struct extent_state *cached_state = NULL;
 	u64 start = 0;
 	u64 end;
 
 	while (!find_first_extent_bit(dirty_pages, start, &start, &end,
-				      EXTENT_NEED_WAIT)) {
-		clear_extent_bits(dirty_pages, start, end, EXTENT_NEED_WAIT, GFP_NOFS);
+				      EXTENT_NEED_WAIT, &cached_state)) {
+		clear_extent_bit(dirty_pages, start, end, EXTENT_NEED_WAIT,
+				 0, 0, &cached_state, GFP_NOFS);
 		err = filemap_fdatawait_range(mapping, start, end);
 		if (err)
 			werr = err;
@@ -680,8 +842,11 @@ int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
 {
 	int ret;
 	int ret2;
+	struct blk_plug plug;
 
+	blk_start_plug(&plug);
 	ret = btrfs_write_marked_extents(root, dirty_pages, mark);
+	blk_finish_plug(&plug);
 	ret2 = btrfs_wait_marked_extents(root, dirty_pages, mark);
 
 	if (ret)
@@ -783,6 +948,15 @@ static noinline int commit_cowonly_roots(struct btrfs_trans_handle *trans,
 		return ret;
 
 	ret = btrfs_run_dev_stats(trans, root->fs_info);
+	WARN_ON(ret);
+	ret = btrfs_run_dev_replace(trans, root->fs_info);
+	WARN_ON(ret);
+
+	ret = btrfs_run_qgroups(trans, root->fs_info);
+	BUG_ON(ret);
+
+	/* run_qgroups might have added some more refs */
+	ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
 	BUG_ON(ret);
 
 	while (!list_empty(&fs_info->dirty_cowonly_roots)) {
@@ -799,6 +973,8 @@ static noinline int commit_cowonly_roots(struct btrfs_trans_handle *trans,
 	switch_commit_root(fs_info->extent_root);
 	up_write(&fs_info->extent_commit_sem);
 
+	btrfs_after_dev_replace_commit(fs_info);
+
 	return 0;
 }
 
@@ -810,7 +986,7 @@ static noinline int commit_cowonly_roots(struct btrfs_trans_handle *trans,
 int btrfs_add_dead_root(struct btrfs_root *root)
 {
 	spin_lock(&root->fs_info->trans_lock);
-	list_add(&root->root_list, &root->fs_info->dead_roots);
+	list_add_tail(&root->root_list, &root->fs_info->dead_roots);
 	spin_unlock(&root->fs_info->trans_lock);
 	return 0;
 }
@@ -875,15 +1051,14 @@ static noinline int commit_fs_roots(struct btrfs_trans_handle *trans,
 }
 
 /*
- * defrag a given btree.  If cacheonly == 1, this won't read from the disk,
- * otherwise every leaf in the btree is read and defragged.
+ * defrag a given btree.
+ * Every leaf in the btree is read and defragged.
  */
-int btrfs_defrag_root(struct btrfs_root *root, int cacheonly)
+int btrfs_defrag_root(struct btrfs_root *root)
 {
 	struct btrfs_fs_info *info = root->fs_info;
 	struct btrfs_trans_handle *trans;
 	int ret;
-	unsigned long nr;
 
 	if (xchg(&root->defrag_running, 1))
 		return 0;
@@ -893,15 +1068,20 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly)
 		if (IS_ERR(trans))
 			return PTR_ERR(trans);
 
-		ret = btrfs_defrag_leaves(trans, root, cacheonly);
+		ret = btrfs_defrag_leaves(trans, root);
 
-		nr = trans->blocks_used;
 		btrfs_end_transaction(trans, root);
-		btrfs_btree_balance_dirty(info->tree_root, nr);
+		btrfs_btree_balance_dirty(info->tree_root);
 		cond_resched();
 
 		if (btrfs_fs_closing(root->fs_info) || ret != -EAGAIN)
 			break;
+
+		if (btrfs_defrag_cancelled(root->fs_info)) {
+			printk(KERN_DEBUG "btrfs: defrag_root cancelled\n");
+			ret = -EAGAIN;
+			break;
+		}
 	}
 	root->defrag_running = 0;
 	return ret;
@@ -909,7 +1089,12 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly)
 
 /*
  * new snapshots need to be created at a very specific time in the
- * transaction commit.  This does the actual creation
+ * transaction commit.  This does the actual creation.
+ *
+ * Note:
+ * If the error which may affect the commitment of the current transaction
+ * happens, we should return the error number. If the error which just affect
+ * the creation of the pending snapshots, just return 0.
  */
 static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
 				   struct btrfs_fs_info *fs_info,
@@ -922,50 +1107,62 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
 	struct btrfs_root *parent_root;
 	struct btrfs_block_rsv *rsv;
 	struct inode *parent_inode;
-	struct dentry *parent;
+	struct btrfs_path *path;
+	struct btrfs_dir_item *dir_item;
 	struct dentry *dentry;
 	struct extent_buffer *tmp;
 	struct extent_buffer *old;
-	int ret;
+	struct timespec cur_time = CURRENT_TIME;
+	int ret = 0;
 	u64 to_reserve = 0;
 	u64 index = 0;
 	u64 objectid;
 	u64 root_flags;
+	uuid_le new_uuid;
 
-	rsv = trans->block_rsv;
+	path = btrfs_alloc_path();
+	if (!path) {
+		pending->error = -ENOMEM;
+		return 0;
+	}
 
 	new_root_item = kmalloc(sizeof(*new_root_item), GFP_NOFS);
 	if (!new_root_item) {
-		ret = pending->error = -ENOMEM;
-		goto fail;
+		pending->error = -ENOMEM;
+		goto root_item_alloc_fail;
 	}
 
-	ret = btrfs_find_free_objectid(tree_root, &objectid);
-	if (ret) {
-		pending->error = ret;
-		goto fail;
-	}
+	pending->error = btrfs_find_free_objectid(tree_root, &objectid);
+	if (pending->error)
+		goto no_free_objectid;
 
 	btrfs_reloc_pre_snapshot(trans, pending, &to_reserve);
 
 	if (to_reserve > 0) {
-		ret = btrfs_block_rsv_add_noflush(root, &pending->block_rsv,
-						  to_reserve);
-		if (ret) {
-			pending->error = ret;
-			goto fail;
-		}
+		pending->error = btrfs_block_rsv_add(root,
+						     &pending->block_rsv,
+						     to_reserve,
+						     BTRFS_RESERVE_NO_FLUSH);
+		if (pending->error)
+			goto no_free_objectid;
 	}
 
+	pending->error = btrfs_qgroup_inherit(trans, fs_info,
+					      root->root_key.objectid,
+					      objectid, pending->inherit);
+	if (pending->error)
+		goto no_free_objectid;
+
 	key.objectid = objectid;
 	key.offset = (u64)-1;
 	key.type = BTRFS_ROOT_ITEM_KEY;
 
+	rsv = trans->block_rsv;
 	trans->block_rsv = &pending->block_rsv;
+	trans->bytes_reserved = trans->block_rsv->reserved;
 
 	dentry = pending->dentry;
-	parent = dget_parent(dentry);
-	parent_inode = parent->d_inode;
+	parent_inode = pending->dir;
 	parent_root = BTRFS_I(parent_inode)->root;
 	record_root_in_trans(trans, parent_root);
 
@@ -974,23 +1171,21 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
 	 */
 	ret = btrfs_set_inode_index(parent_inode, &index);
 	BUG_ON(ret); /* -ENOMEM */
-	ret = btrfs_insert_dir_item(trans, parent_root,
-				dentry->d_name.name, dentry->d_name.len,
-				parent_inode, &key,
-				BTRFS_FT_DIR, index);
-	if (ret == -EEXIST) {
+
+	/* check if there is a file/dir which has the same name. */
+	dir_item = btrfs_lookup_dir_item(NULL, parent_root, path,
+					 btrfs_ino(parent_inode),
+					 dentry->d_name.name,
+					 dentry->d_name.len, 0);
+	if (dir_item != NULL && !IS_ERR(dir_item)) {
 		pending->error = -EEXIST;
-		dput(parent);
+		goto dir_item_existed;
+	} else if (IS_ERR(dir_item)) {
+		ret = PTR_ERR(dir_item);
+		btrfs_abort_transaction(trans, root, ret);
 		goto fail;
-	} else if (ret) {
-		goto abort_trans_dput;
 	}
-
-	btrfs_i_size_write(parent_inode, parent_inode->i_size +
-					 dentry->d_name.len * 2);
-	ret = btrfs_update_inode(trans, parent_root, parent_inode);
-	if (ret)
-		goto abort_trans_dput;
+	btrfs_release_path(path);
 
 	/*
 	 * pull in the delayed directory update
@@ -999,8 +1194,8 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
 	 * snapshot
 	 */
 	ret = btrfs_run_delayed_items(trans, root);
-	if (ret) { /* Transaction aborted */
-		dput(parent);
+	if (ret) {	/* Transaction aborted */
+		btrfs_abort_transaction(trans, root, ret);
 		goto fail;
 	}
 
@@ -1016,12 +1211,31 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
 		root_flags &= ~BTRFS_ROOT_SUBVOL_RDONLY;
 	btrfs_set_root_flags(new_root_item, root_flags);
 
+	btrfs_set_root_generation_v2(new_root_item,
+			trans->transid);
+	uuid_le_gen(&new_uuid);
+	memcpy(new_root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
+	memcpy(new_root_item->parent_uuid, root->root_item.uuid,
+			BTRFS_UUID_SIZE);
+	if (!(root_flags & BTRFS_ROOT_SUBVOL_RDONLY)) {
+		memset(new_root_item->received_uuid, 0,
+		       sizeof(new_root_item->received_uuid));
+		memset(&new_root_item->stime, 0, sizeof(new_root_item->stime));
+		memset(&new_root_item->rtime, 0, sizeof(new_root_item->rtime));
+		btrfs_set_root_stransid(new_root_item, 0);
+		btrfs_set_root_rtransid(new_root_item, 0);
+	}
+	new_root_item->otime.sec = cpu_to_le64(cur_time.tv_sec);
+	new_root_item->otime.nsec = cpu_to_le32(cur_time.tv_nsec);
+	btrfs_set_root_otransid(new_root_item, trans->transid);
+
 	old = btrfs_lock_root_node(root);
 	ret = btrfs_cow_block(trans, root, old, NULL, 0, &old);
 	if (ret) {
 		btrfs_tree_unlock(old);
 		free_extent_buffer(old);
-		goto abort_trans_dput;
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
 	}
 
 	btrfs_set_lock_blocking(old);
@@ -1030,8 +1244,10 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
 	/* clean up in any case */
 	btrfs_tree_unlock(old);
 	free_extent_buffer(old);
-	if (ret)
-		goto abort_trans_dput;
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
 
 	/* see comments in should_cow_block() */
 	root->force_cow = 1;
@@ -1043,8 +1259,10 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
 	ret = btrfs_insert_root(trans, tree_root, &key, new_root_item);
 	btrfs_tree_unlock(tmp);
 	free_extent_buffer(tmp);
-	if (ret)
-		goto abort_trans_dput;
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
 
 	/*
 	 * insert root back/forward references
@@ -1053,32 +1271,58 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
 				 parent_root->root_key.objectid,
 				 btrfs_ino(parent_inode), index,
 				 dentry->d_name.name, dentry->d_name.len);
-	dput(parent);
-	if (ret)
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
 		goto fail;
+	}
 
 	key.offset = (u64)-1;
 	pending->snap = btrfs_read_fs_root_no_name(root->fs_info, &key);
 	if (IS_ERR(pending->snap)) {
 		ret = PTR_ERR(pending->snap);
-		goto abort_trans;
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
 	}
 
 	ret = btrfs_reloc_post_snapshot(trans, pending);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+
+	ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+
+	ret = btrfs_insert_dir_item(trans, parent_root,
+				    dentry->d_name.name, dentry->d_name.len,
+				    parent_inode, &key,
+				    BTRFS_FT_DIR, index);
+	/* We have check then name at the beginning, so it is impossible. */
+	BUG_ON(ret == -EEXIST || ret == -EOVERFLOW);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+
+	btrfs_i_size_write(parent_inode, parent_inode->i_size +
+					 dentry->d_name.len * 2);
+	parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
+	ret = btrfs_update_inode_fallback(trans, parent_root, parent_inode);
 	if (ret)
-		goto abort_trans;
-	ret = 0;
+		btrfs_abort_transaction(trans, root, ret);
 fail:
-	kfree(new_root_item);
+	pending->error = ret;
+dir_item_existed:
 	trans->block_rsv = rsv;
-	btrfs_block_rsv_release(root, &pending->block_rsv, (u64)-1);
+	trans->bytes_reserved = 0;
+no_free_objectid:
+	kfree(new_root_item);
+root_item_alloc_fail:
+	btrfs_free_path(path);
 	return ret;
-
-abort_trans_dput:
-	dput(parent);
-abort_trans:
-	btrfs_abort_transaction(trans, root, ret);
-	goto fail;
 }
 
 /*
@@ -1087,12 +1331,17 @@ abort_trans:
 static noinline int create_pending_snapshots(struct btrfs_trans_handle *trans,
 					     struct btrfs_fs_info *fs_info)
 {
-	struct btrfs_pending_snapshot *pending;
+	struct btrfs_pending_snapshot *pending, *next;
 	struct list_head *head = &trans->transaction->pending_snapshots;
+	int ret = 0;
 
-	list_for_each_entry(pending, head, list)
-		create_pending_snapshot(trans, fs_info, pending);
-	return 0;
+	list_for_each_entry_safe(pending, next, head, list) {
+		list_del(&pending->list);
+		ret = create_pending_snapshot(trans, fs_info, pending);
+		if (ret)
+			break;
+	}
+	return ret;
 }
 
 static void update_super_roots(struct btrfs_root *root)
@@ -1117,20 +1366,26 @@ static void update_super_roots(struct btrfs_root *root)
 
 int btrfs_transaction_in_commit(struct btrfs_fs_info *info)
 {
+	struct btrfs_transaction *trans;
 	int ret = 0;
+
 	spin_lock(&info->trans_lock);
-	if (info->running_transaction)
-		ret = info->running_transaction->in_commit;
+	trans = info->running_transaction;
+	if (trans)
+		ret = (trans->state >= TRANS_STATE_COMMIT_START);
 	spin_unlock(&info->trans_lock);
 	return ret;
 }
 
 int btrfs_transaction_blocked(struct btrfs_fs_info *info)
 {
+	struct btrfs_transaction *trans;
 	int ret = 0;
+
 	spin_lock(&info->trans_lock);
-	if (info->running_transaction)
-		ret = info->running_transaction->blocked;
+	trans = info->running_transaction;
+	if (trans)
+		ret = is_transaction_blocked(trans);
 	spin_unlock(&info->trans_lock);
 	return ret;
 }
@@ -1142,7 +1397,9 @@ int btrfs_transaction_blocked(struct btrfs_fs_info *info)
 static void wait_current_trans_commit_start(struct btrfs_root *root,
 					    struct btrfs_transaction *trans)
 {
-	wait_event(root->fs_info->transaction_blocked_wait, trans->in_commit);
+	wait_event(root->fs_info->transaction_blocked_wait,
+		   trans->state >= TRANS_STATE_COMMIT_START ||
+		   trans->aborted);
 }
 
 /*
@@ -1153,7 +1410,8 @@ static void wait_current_trans_commit_start_and_unblock(struct btrfs_root *root,
 					 struct btrfs_transaction *trans)
 {
 	wait_event(root->fs_info->transaction_wait,
-		   trans->commit_done || (trans->in_commit && !trans->blocked));
+		   trans->state >= TRANS_STATE_UNBLOCKED ||
+		   trans->aborted);
 }
 
 /*
@@ -1163,13 +1421,24 @@ static void wait_current_trans_commit_start_and_unblock(struct btrfs_root *root,
 struct btrfs_async_commit {
 	struct btrfs_trans_handle *newtrans;
 	struct btrfs_root *root;
-	struct delayed_work work;
+	struct work_struct work;
 };
 
 static void do_async_commit(struct work_struct *work)
 {
 	struct btrfs_async_commit *ac =
-		container_of(work, struct btrfs_async_commit, work.work);
+		container_of(work, struct btrfs_async_commit, work);
+
+	/*
+	 * We've got freeze protection passed with the transaction.
+	 * Tell lockdep about it.
+	 */
+	if (ac->newtrans->type < TRANS_JOIN_NOLOCK)
+		rwsem_acquire_read(
+		     &ac->root->fs_info->sb->s_writers.lock_map[SB_FREEZE_FS-1],
+		     0, 1, _THIS_IP_);
+
+	current->journal_info = ac->newtrans;
 
 	btrfs_commit_transaction(ac->newtrans, ac->root);
 	kfree(ac);
@@ -1186,7 +1455,7 @@ int btrfs_commit_transaction_async(struct btrfs_trans_handle *trans,
 	if (!ac)
 		return -ENOMEM;
 
-	INIT_DELAYED_WORK(&ac->work, do_async_commit);
+	INIT_WORK(&ac->work, do_async_commit);
 	ac->root = root;
 	ac->newtrans = btrfs_join_transaction(root);
 	if (IS_ERR(ac->newtrans)) {
@@ -1200,7 +1469,17 @@ int btrfs_commit_transaction_async(struct btrfs_trans_handle *trans,
 	atomic_inc(&cur_trans->use_count);
 
 	btrfs_end_transaction(trans, root);
-	schedule_delayed_work(&ac->work, 0);
+
+	/*
+	 * Tell lockdep we've released the freeze rwsem, since the
+	 * async commit thread will be the one to unlock it.
+	 */
+	if (trans->type < TRANS_JOIN_NOLOCK)
+		rwsem_release(
+			&root->fs_info->sb->s_writers.lock_map[SB_FREEZE_FS-1],
+			1, _THIS_IP_);
+
+	schedule_work(&ac->work);
 
 	/* wait for transaction to start and unblock */
 	if (wait_for_unblock)
@@ -1220,21 +1499,39 @@ static void cleanup_transaction(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root, int err)
 {
 	struct btrfs_transaction *cur_trans = trans->transaction;
+	DEFINE_WAIT(wait);
 
 	WARN_ON(trans->use_count > 1);
 
 	btrfs_abort_transaction(trans, root, err);
 
 	spin_lock(&root->fs_info->trans_lock);
+
+	/*
+	 * If the transaction is removed from the list, it means this
+	 * transaction has been committed successfully, so it is impossible
+	 * to call the cleanup function.
+	 */
+	BUG_ON(list_empty(&cur_trans->list));
+
 	list_del_init(&cur_trans->list);
 	if (cur_trans == root->fs_info->running_transaction) {
-		root->fs_info->running_transaction = NULL;
-		root->fs_info->trans_no_join = 0;
+		cur_trans->state = TRANS_STATE_COMMIT_DOING;
+		spin_unlock(&root->fs_info->trans_lock);
+		wait_event(cur_trans->writer_wait,
+			   atomic_read(&cur_trans->num_writers) == 1);
+
+		spin_lock(&root->fs_info->trans_lock);
 	}
 	spin_unlock(&root->fs_info->trans_lock);
 
 	btrfs_cleanup_one_transaction(trans->transaction, root);
 
+	spin_lock(&root->fs_info->trans_lock);
+	if (cur_trans == root->fs_info->running_transaction)
+		root->fs_info->running_transaction = NULL;
+	spin_unlock(&root->fs_info->trans_lock);
+
 	put_transaction(cur_trans);
 	put_transaction(cur_trans);
 
@@ -1248,39 +1545,83 @@ static void cleanup_transaction(struct btrfs_trans_handle *trans,
 	kmem_cache_free(btrfs_trans_handle_cachep, trans);
 }
 
-/*
- * btrfs_transaction state sequence:
- *    in_commit = 0, blocked = 0  (initial)
- *    in_commit = 1, blocked = 1
- *    blocked = 0
- *    commit_done = 1
- */
+static int btrfs_flush_all_pending_stuffs(struct btrfs_trans_handle *trans,
+					  struct btrfs_root *root)
+{
+	int ret;
+
+	ret = btrfs_run_delayed_items(trans, root);
+	if (ret)
+		return ret;
+
+	/*
+	 * running the delayed items may have added new refs. account
+	 * them now so that they hinder processing of more delayed refs
+	 * as little as possible.
+	 */
+	btrfs_delayed_refs_qgroup_accounting(trans, root->fs_info);
+
+	/*
+	 * rename don't use btrfs_join_transaction, so, once we
+	 * set the transaction to blocked above, we aren't going
+	 * to get any new ordered operations.  We can safely run
+	 * it here and no for sure that nothing new will be added
+	 * to the list
+	 */
+	ret = btrfs_run_ordered_operations(trans, root, 1);
+
+	return ret;
+}
+
+static inline int btrfs_start_delalloc_flush(struct btrfs_fs_info *fs_info)
+{
+	if (btrfs_test_opt(fs_info->tree_root, FLUSHONCOMMIT))
+		return btrfs_start_all_delalloc_inodes(fs_info, 1);
+	return 0;
+}
+
+static inline void btrfs_wait_delalloc_flush(struct btrfs_fs_info *fs_info)
+{
+	if (btrfs_test_opt(fs_info->tree_root, FLUSHONCOMMIT))
+		btrfs_wait_all_ordered_extents(fs_info, 1);
+}
+
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root)
 {
-	unsigned long joined = 0;
 	struct btrfs_transaction *cur_trans = trans->transaction;
 	struct btrfs_transaction *prev_trans = NULL;
-	DEFINE_WAIT(wait);
-	int ret = -EIO;
-	int should_grow = 0;
-	unsigned long now = get_seconds();
-	int flush_on_commit = btrfs_test_opt(root, FLUSHONCOMMIT);
-
-	btrfs_run_ordered_operations(root, 0);
+	int ret;
 
-	btrfs_trans_release_metadata(trans, root);
-	trans->block_rsv = NULL;
+	ret = btrfs_run_ordered_operations(trans, root, 0);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		btrfs_end_transaction(trans, root);
+		return ret;
+	}
 
-	if (cur_trans->aborted)
-		goto cleanup_transaction;
+	/* Stop the commit early if ->aborted is set */
+	if (unlikely(ACCESS_ONCE(cur_trans->aborted))) {
+		ret = cur_trans->aborted;
+		btrfs_end_transaction(trans, root);
+		return ret;
+	}
 
 	/* make a pass through all the delayed refs we have so far
 	 * any runnings procs may add more while we are here
 	 */
 	ret = btrfs_run_delayed_refs(trans, root, 0);
-	if (ret)
-		goto cleanup_transaction;
+	if (ret) {
+		btrfs_end_transaction(trans, root);
+		return ret;
+	}
+
+	btrfs_trans_release_metadata(trans, root);
+	trans->block_rsv = NULL;
+	if (trans->qgroup_reserved) {
+		btrfs_qgroup_free(root, trans->qgroup_reserved);
+		trans->qgroup_reserved = 0;
+	}
 
 	cur_trans = trans->transaction;
 
@@ -1289,14 +1630,20 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	 * start sending their work down.
 	 */
 	cur_trans->delayed_refs.flushing = 1;
+	smp_wmb();
+
+	if (!list_empty(&trans->new_bgs))
+		btrfs_create_pending_block_groups(trans, root);
 
 	ret = btrfs_run_delayed_refs(trans, root, 0);
-	if (ret)
-		goto cleanup_transaction;
+	if (ret) {
+		btrfs_end_transaction(trans, root);
+		return ret;
+	}
 
-	spin_lock(&cur_trans->commit_lock);
-	if (cur_trans->in_commit) {
-		spin_unlock(&cur_trans->commit_lock);
+	spin_lock(&root->fs_info->trans_lock);
+	if (cur_trans->state >= TRANS_STATE_COMMIT_START) {
+		spin_unlock(&root->fs_info->trans_lock);
 		atomic_inc(&cur_trans->use_count);
 		ret = btrfs_end_transaction(trans, root);
 
@@ -1307,16 +1654,13 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 		return ret;
 	}
 
-	trans->transaction->in_commit = 1;
-	trans->transaction->blocked = 1;
-	spin_unlock(&cur_trans->commit_lock);
+	cur_trans->state = TRANS_STATE_COMMIT_START;
 	wake_up(&root->fs_info->transaction_blocked_wait);
 
-	spin_lock(&root->fs_info->trans_lock);
 	if (cur_trans->list.prev != &root->fs_info->trans_list) {
 		prev_trans = list_entry(cur_trans->list.prev,
 					struct btrfs_transaction, list);
-		if (!prev_trans->commit_done) {
+		if (prev_trans->state != TRANS_STATE_COMPLETED) {
 			atomic_inc(&prev_trans->use_count);
 			spin_unlock(&root->fs_info->trans_lock);
 
@@ -1330,59 +1674,41 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 		spin_unlock(&root->fs_info->trans_lock);
 	}
 
-	if (now < cur_trans->start_time || now - cur_trans->start_time < 1)
-		should_grow = 1;
-
-	do {
-		int snap_pending = 0;
-
-		joined = cur_trans->num_joined;
-		if (!list_empty(&trans->transaction->pending_snapshots))
-			snap_pending = 1;
-
-		WARN_ON(cur_trans != trans->transaction);
+	extwriter_counter_dec(cur_trans, trans->type);
 
-		if (flush_on_commit || snap_pending) {
-			btrfs_start_delalloc_inodes(root, 1);
-			btrfs_wait_ordered_extents(root, 0, 1);
-		}
-
-		ret = btrfs_run_delayed_items(trans, root);
-		if (ret)
-			goto cleanup_transaction;
-
-		/*
-		 * rename don't use btrfs_join_transaction, so, once we
-		 * set the transaction to blocked above, we aren't going
-		 * to get any new ordered operations.  We can safely run
-		 * it here and no for sure that nothing new will be added
-		 * to the list
-		 */
-		btrfs_run_ordered_operations(root, 1);
+	ret = btrfs_start_delalloc_flush(root->fs_info);
+	if (ret)
+		goto cleanup_transaction;
 
-		prepare_to_wait(&cur_trans->writer_wait, &wait,
-				TASK_UNINTERRUPTIBLE);
+	ret = btrfs_flush_all_pending_stuffs(trans, root);
+	if (ret)
+		goto cleanup_transaction;
 
-		if (atomic_read(&cur_trans->num_writers) > 1)
-			schedule_timeout(MAX_SCHEDULE_TIMEOUT);
-		else if (should_grow)
-			schedule_timeout(1);
+	wait_event(cur_trans->writer_wait,
+		   extwriter_counter_read(cur_trans) == 0);
 
-		finish_wait(&cur_trans->writer_wait, &wait);
-	} while (atomic_read(&cur_trans->num_writers) > 1 ||
-		 (should_grow && cur_trans->num_joined != joined));
+	/* some pending stuffs might be added after the previous flush. */
+	ret = btrfs_flush_all_pending_stuffs(trans, root);
+	if (ret)
+		goto cleanup_transaction;
 
+	btrfs_wait_delalloc_flush(root->fs_info);
 	/*
 	 * Ok now we need to make sure to block out any other joins while we
 	 * commit the transaction.  We could have started a join before setting
-	 * no_join so make sure to wait for num_writers to == 1 again.
+	 * COMMIT_DOING so make sure to wait for num_writers to == 1 again.
 	 */
 	spin_lock(&root->fs_info->trans_lock);
-	root->fs_info->trans_no_join = 1;
+	cur_trans->state = TRANS_STATE_COMMIT_DOING;
 	spin_unlock(&root->fs_info->trans_lock);
 	wait_event(cur_trans->writer_wait,
 		   atomic_read(&cur_trans->num_writers) == 1);
 
+	/* ->aborted might be set after the previous check, so check it */
+	if (unlikely(ACCESS_ONCE(cur_trans->aborted))) {
+		ret = cur_trans->aborted;
+		goto cleanup_transaction;
+	}
 	/*
 	 * the reloc mutex makes sure that we stop
 	 * the balancing code from coming in and moving
@@ -1390,13 +1716,28 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	 */
 	mutex_lock(&root->fs_info->reloc_mutex);
 
-	ret = btrfs_run_delayed_items(trans, root);
+	/*
+	 * We needn't worry about the delayed items because we will
+	 * deal with them in create_pending_snapshot(), which is the
+	 * core function of the snapshot creation.
+	 */
+	ret = create_pending_snapshots(trans, root->fs_info);
 	if (ret) {
 		mutex_unlock(&root->fs_info->reloc_mutex);
 		goto cleanup_transaction;
 	}
 
-	ret = create_pending_snapshots(trans, root->fs_info);
+	/*
+	 * We insert the dir indexes of the snapshots and update the inode
+	 * of the snapshots' parents after the snapshot creation, so there
+	 * are some delayed items which are not dealt with. Now deal with
+	 * them.
+	 *
+	 * We needn't worry that this operation will corrupt the snapshots,
+	 * because all the tree which are snapshoted will be forced to COW
+	 * the nodes and leaves.
+	 */
+	ret = btrfs_run_delayed_items(trans, root);
 	if (ret) {
 		mutex_unlock(&root->fs_info->reloc_mutex);
 		goto cleanup_transaction;
@@ -1451,6 +1792,17 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 		goto cleanup_transaction;
 	}
 
+	/*
+	 * The tasks which save the space cache and inode cache may also
+	 * update ->aborted, check it.
+	 */
+	if (unlikely(ACCESS_ONCE(cur_trans->aborted))) {
+		ret = cur_trans->aborted;
+		mutex_unlock(&root->fs_info->tree_log_mutex);
+		mutex_unlock(&root->fs_info->reloc_mutex);
+		goto cleanup_transaction;
+	}
+
 	btrfs_prepare_extent_commit(trans, root);
 
 	cur_trans = root->fs_info->running_transaction;
@@ -1463,6 +1815,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 			    root->fs_info->chunk_root->node);
 	switch_commit_root(root->fs_info->chunk_root);
 
+	assert_qgroups_uptodate(trans);
 	update_super_roots(root);
 
 	if (!root->fs_info->log_root_recovering) {
@@ -1473,10 +1826,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	memcpy(root->fs_info->super_for_commit, root->fs_info->super_copy,
 	       sizeof(*root->fs_info->super_copy));
 
-	trans->transaction->blocked = 0;
 	spin_lock(&root->fs_info->trans_lock);
+	cur_trans->state = TRANS_STATE_UNBLOCKED;
 	root->fs_info->running_transaction = NULL;
-	root->fs_info->trans_no_join = 0;
 	spin_unlock(&root->fs_info->trans_lock);
 	mutex_unlock(&root->fs_info->reloc_mutex);
 
@@ -1485,7 +1837,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	ret = btrfs_write_and_wait_transaction(trans, root);
 	if (ret) {
 		btrfs_error(root->fs_info, ret,
-			    "Error while writing out transaction.");
+			    "Error while writing out transaction");
 		mutex_unlock(&root->fs_info->tree_log_mutex);
 		goto cleanup_transaction;
 	}
@@ -1504,10 +1856,12 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 
 	btrfs_finish_extent_commit(trans, root);
 
-	cur_trans->commit_done = 1;
-
 	root->fs_info->last_trans_committed = cur_trans->transid;
-
+	/*
+	 * We needn't acquire the lock here because there is no other task
+	 * which can change it.
+	 */
+	cur_trans->state = TRANS_STATE_COMPLETED;
 	wake_up(&cur_trans->commit_wait);
 
 	spin_lock(&root->fs_info->trans_lock);
@@ -1517,6 +1871,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	put_transaction(cur_trans);
 	put_transaction(cur_trans);
 
+	if (trans->type & __TRANS_FREEZABLE)
+		sb_end_intwrite(root->fs_info->sb);
+
 	trace_btrfs_transaction_commit(root);
 
 	btrfs_scrub_continue(root);
@@ -1532,8 +1889,13 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	return ret;
 
 cleanup_transaction:
-	btrfs_printk(root->fs_info, "Skipping commit of aborted transaction.\n");
-//	WARN_ON(1);
+	btrfs_trans_release_metadata(trans, root);
+	trans->block_rsv = NULL;
+	if (trans->qgroup_reserved) {
+		btrfs_qgroup_free(root, trans->qgroup_reserved);
+		trans->qgroup_reserved = 0;
+	}
+	btrfs_warn(root->fs_info, "Skipping commit of aborted transaction.");
 	if (current->journal_info == trans)
 		current->journal_info = NULL;
 	cleanup_transaction(trans, root, ret);
@@ -1542,31 +1904,44 @@ cleanup_transaction:
 }
 
 /*
- * interface function to delete all the snapshots we have scheduled for deletion
+ * return < 0 if error
+ * 0 if there are no more dead_roots at the time of call
+ * 1 there are more to be processed, call me again
+ *
+ * The return value indicates there are certainly more snapshots to delete, but
+ * if there comes a new one during processing, it may return 0. We don't mind,
+ * because btrfs_commit_super will poke cleaner thread and it will process it a
+ * few seconds later.
  */
-int btrfs_clean_old_snapshots(struct btrfs_root *root)
+int btrfs_clean_one_deleted_snapshot(struct btrfs_root *root)
 {
-	LIST_HEAD(list);
+	int ret;
 	struct btrfs_fs_info *fs_info = root->fs_info;
 
 	spin_lock(&fs_info->trans_lock);
-	list_splice_init(&fs_info->dead_roots, &list);
+	if (list_empty(&fs_info->dead_roots)) {
+		spin_unlock(&fs_info->trans_lock);
+		return 0;
+	}
+	root = list_first_entry(&fs_info->dead_roots,
+			struct btrfs_root, root_list);
+	list_del(&root->root_list);
 	spin_unlock(&fs_info->trans_lock);
 
-	while (!list_empty(&list)) {
-		int ret;
-
-		root = list_entry(list.next, struct btrfs_root, root_list);
-		list_del(&root->root_list);
+	pr_debug("btrfs: cleaner removing %llu\n",
+			(unsigned long long)root->objectid);
 
-		btrfs_kill_all_delayed_nodes(root);
+	btrfs_kill_all_delayed_nodes(root);
 
-		if (btrfs_header_backref_rev(root->node) <
-		    BTRFS_MIXED_BACKREF_REV)
-			ret = btrfs_drop_snapshot(root, NULL, 0, 0);
-		else
-			ret =btrfs_drop_snapshot(root, NULL, 1, 0);
-		BUG_ON(ret < 0);
-	}
-	return 0;
+	if (btrfs_header_backref_rev(root->node) <
+			BTRFS_MIXED_BACKREF_REV)
+		ret = btrfs_drop_snapshot(root, NULL, 0, 0);
+	else
+		ret = btrfs_drop_snapshot(root, NULL, 1, 0);
+	/*
+	 * If we encounter a transaction abort during snapshot cleaning, we
+	 * don't want to crash here
+	 */
+	BUG_ON(ret < 0 && ret != -EAGAIN && ret != -EROFS);
+	return 1;
 }