Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs

Pull btrfs fixes from Chris Mason:
 "I've split out the big send/receive update from my last pull request
  and now have just the fixes in my for-linus branch.  The send/recv
  branch will wander over to linux-next shortly though.

  The largest patches in this pull are Josef's patches to fix DIO
  locking problems and his patch to fix a crash during balance.  They
  are both well tested.

  The rest are smaller fixes that we've had queued.  The last rc came
  out while I was hacking new and exciting ways to recover from a
  misplaced rm -rf on my dev box, so these missed rc3."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (25 commits)
  Btrfs: fix that repair code is spuriously executed for transid failures
  Btrfs: fix ordered extent leak when failing to start a transaction
  Btrfs: fix a dio write regression
  Btrfs: fix deadlock with freeze and sync V2
  Btrfs: revert checksum error statistic which can cause a BUG()
  Btrfs: remove superblock writing after fatal error
  Btrfs: allow delayed refs to be merged
  Btrfs: fix enospc problems when deleting a subvol
  Btrfs: fix wrong mtime and ctime when creating snapshots
  Btrfs: fix race in run_clustered_refs
  Btrfs: don't run __tree_mod_log_free_eb on leaves
  Btrfs: increase the size of the free space cache
  Btrfs: barrier before waitqueue_active
  Btrfs: fix deadlock in wait_for_more_refs
  btrfs: fix second lock in btrfs_delete_delayed_items()
  Btrfs: don't allocate a seperate csums array for direct reads
  Btrfs: do not strdup non existent strings
  Btrfs: do not use missing devices when showing devname
  Btrfs: fix that error value is changed by mistake
  Btrfs: lock extents as we map them in DIO
  ...

21 files changed, 418 insertions, 376 deletions

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index a256f3b2a845..ff6475f409d6 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -1438,10 +1438,10 @@ int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
 	ret = extent_from_logical(fs_info, logical, path,
 					&found_key);
 	btrfs_release_path(path);
-	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
-		ret = -EINVAL;
 	if (ret < 0)
 		return ret;
+	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
+		return -EINVAL;
 
 	extent_item_pos = logical - found_key.objectid;
 	ret = iterate_extent_inodes(fs_info, found_key.objectid,
diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c
index 86eff48dab78..43d1c5a3a030 100644
--- a/fs/btrfs/compression.c
+++ b/fs/btrfs/compression.c
@@ -818,6 +818,7 @@ static void free_workspace(int type, struct list_head *workspace)
 	btrfs_compress_op[idx]->free_workspace(workspace);
 	atomic_dec(alloc_workspace);
 wake:
+	smp_mb();
 	if (waitqueue_active(workspace_wait))
 		wake_up(workspace_wait);
 }
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 9d7621f271ff..6d183f60d63a 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -421,12 +421,6 @@ void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info,
 	spin_unlock(&fs_info->tree_mod_seq_lock);
 
 	/*
-	 * we removed the lowest blocker from the blocker list, so there may be
-	 * more processible delayed refs.
-	 */
-	wake_up(&fs_info->tree_mod_seq_wait);
-
-	/*
 	 * anything that's lower than the lowest existing (read: blocked)
 	 * sequence number can be removed from the tree.
 	 */
@@ -631,6 +625,9 @@ __tree_mod_log_free_eb(struct btrfs_fs_info *fs_info, struct extent_buffer *eb)
 	u32 nritems;
 	int ret;
 
+	if (btrfs_header_level(eb) == 0)
+		return;
+
 	nritems = btrfs_header_nritems(eb);
 	for (i = nritems - 1; i >= 0; i--) {
 		ret = tree_mod_log_insert_key_locked(fs_info, eb, i,
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 4bab807227ad..0d195b507660 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1252,7 +1252,6 @@ struct btrfs_fs_info {
 	atomic_t tree_mod_seq;
 	struct list_head tree_mod_seq_list;
 	struct seq_list tree_mod_seq_elem;
-	wait_queue_head_t tree_mod_seq_wait;
 
 	/* this protects tree_mod_log */
 	rwlock_t tree_mod_log_lock;
@@ -3192,7 +3191,7 @@ int btrfs_del_csums(struct btrfs_trans_handle *trans,
 int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
 			  struct bio *bio, u32 *dst);
 int btrfs_lookup_bio_sums_dio(struct btrfs_root *root, struct inode *inode,
-			      struct bio *bio, u64 logical_offset, u32 *dst);
+			      struct bio *bio, u64 logical_offset);
 int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root,
 			     u64 objectid, u64 pos,
diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c
index 335605c8ceab..07d5eeb1e6f1 100644
--- a/fs/btrfs/delayed-inode.c
+++ b/fs/btrfs/delayed-inode.c
@@ -512,8 +512,8 @@ static void __btrfs_remove_delayed_item(struct btrfs_delayed_item *delayed_item)
 
 	rb_erase(&delayed_item->rb_node, root);
 	delayed_item->delayed_node->count--;
-	atomic_dec(&delayed_root->items);
-	if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND &&
+	if (atomic_dec_return(&delayed_root->items) <
+	    BTRFS_DELAYED_BACKGROUND &&
 	    waitqueue_active(&delayed_root->wait))
 		wake_up(&delayed_root->wait);
 }
@@ -1028,9 +1028,10 @@ do_again:
 		btrfs_release_delayed_item(prev);
 		ret = 0;
 		btrfs_release_path(path);
-		if (curr)
+		if (curr) {
+			mutex_unlock(&node->mutex);
 			goto do_again;
-		else
+		} else
 			goto delete_fail;
 	}
 
@@ -1055,8 +1056,7 @@ static void btrfs_release_delayed_inode(struct btrfs_delayed_node *delayed_node)
 		delayed_node->count--;
 
 		delayed_root = delayed_node->root->fs_info->delayed_root;
-		atomic_dec(&delayed_root->items);
-		if (atomic_read(&delayed_root->items) <
+		if (atomic_dec_return(&delayed_root->items) <
 		    BTRFS_DELAYED_BACKGROUND &&
 		    waitqueue_active(&delayed_root->wait))
 			wake_up(&delayed_root->wait);
diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c
index da7419ed01bb..ae9411773397 100644
--- a/fs/btrfs/delayed-ref.c
+++ b/fs/btrfs/delayed-ref.c
@@ -38,17 +38,14 @@
 static int comp_tree_refs(struct btrfs_delayed_tree_ref *ref2,
 			  struct btrfs_delayed_tree_ref *ref1)
 {
-	if (ref1->node.type == BTRFS_TREE_BLOCK_REF_KEY) {
-		if (ref1->root < ref2->root)
-			return -1;
-		if (ref1->root > ref2->root)
-			return 1;
-	} else {
-		if (ref1->parent < ref2->parent)
-			return -1;
-		if (ref1->parent > ref2->parent)
-			return 1;
-	}
+	if (ref1->root < ref2->root)
+		return -1;
+	if (ref1->root > ref2->root)
+		return 1;
+	if (ref1->parent < ref2->parent)
+		return -1;
+	if (ref1->parent > ref2->parent)
+		return 1;
 	return 0;
 }
 
@@ -85,7 +82,8 @@ static int comp_data_refs(struct btrfs_delayed_data_ref *ref2,
  * type of the delayed backrefs and content of delayed backrefs.
  */
 static int comp_entry(struct btrfs_delayed_ref_node *ref2,
-		      struct btrfs_delayed_ref_node *ref1)
+		      struct btrfs_delayed_ref_node *ref1,
+		      bool compare_seq)
 {
 	if (ref1->bytenr < ref2->bytenr)
 		return -1;
@@ -102,10 +100,12 @@ static int comp_entry(struct btrfs_delayed_ref_node *ref2,
 	if (ref1->type > ref2->type)
 		return 1;
 	/* merging of sequenced refs is not allowed */
-	if (ref1->seq < ref2->seq)
-		return -1;
-	if (ref1->seq > ref2->seq)
-		return 1;
+	if (compare_seq) {
+		if (ref1->seq < ref2->seq)
+			return -1;
+		if (ref1->seq > ref2->seq)
+			return 1;
+	}
 	if (ref1->type == BTRFS_TREE_BLOCK_REF_KEY ||
 	    ref1->type == BTRFS_SHARED_BLOCK_REF_KEY) {
 		return comp_tree_refs(btrfs_delayed_node_to_tree_ref(ref2),
@@ -139,7 +139,7 @@ static struct btrfs_delayed_ref_node *tree_insert(struct rb_root *root,
 		entry = rb_entry(parent_node, struct btrfs_delayed_ref_node,
 				 rb_node);
 
-		cmp = comp_entry(entry, ins);
+		cmp = comp_entry(entry, ins, 1);
 		if (cmp < 0)
 			p = &(*p)->rb_left;
 		else if (cmp > 0)
@@ -233,6 +233,114 @@ int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
+static void inline drop_delayed_ref(struct btrfs_trans_handle *trans,
+				    struct btrfs_delayed_ref_root *delayed_refs,
+				    struct btrfs_delayed_ref_node *ref)
+{
+	rb_erase(&ref->rb_node, &delayed_refs->root);
+	ref->in_tree = 0;
+	btrfs_put_delayed_ref(ref);
+	delayed_refs->num_entries--;
+	if (trans->delayed_ref_updates)
+		trans->delayed_ref_updates--;
+}
+
+static int merge_ref(struct btrfs_trans_handle *trans,
+		     struct btrfs_delayed_ref_root *delayed_refs,
+		     struct btrfs_delayed_ref_node *ref, u64 seq)
+{
+	struct rb_node *node;
+	int merged = 0;
+	int mod = 0;
+	int done = 0;
+
+	node = rb_prev(&ref->rb_node);
+	while (node) {
+		struct btrfs_delayed_ref_node *next;
+
+		next = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
+		node = rb_prev(node);
+		if (next->bytenr != ref->bytenr)
+			break;
+		if (seq && next->seq >= seq)
+			break;
+		if (comp_entry(ref, next, 0))
+			continue;
+
+		if (ref->action == next->action) {
+			mod = next->ref_mod;
+		} else {
+			if (ref->ref_mod < next->ref_mod) {
+				struct btrfs_delayed_ref_node *tmp;
+
+				tmp = ref;
+				ref = next;
+				next = tmp;
+				done = 1;
+			}
+			mod = -next->ref_mod;
+		}
+
+		merged++;
+		drop_delayed_ref(trans, delayed_refs, next);
+		ref->ref_mod += mod;
+		if (ref->ref_mod == 0) {
+			drop_delayed_ref(trans, delayed_refs, ref);
+			break;
+		} else {
+			/*
+			 * You can't have multiples of the same ref on a tree
+			 * block.
+			 */
+			WARN_ON(ref->type == BTRFS_TREE_BLOCK_REF_KEY ||
+				ref->type == BTRFS_SHARED_BLOCK_REF_KEY);
+		}
+
+		if (done)
+			break;
+		node = rb_prev(&ref->rb_node);
+	}
+
+	return merged;
+}
+
+void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
+			      struct btrfs_fs_info *fs_info,
+			      struct btrfs_delayed_ref_root *delayed_refs,
+			      struct btrfs_delayed_ref_head *head)
+{
+	struct rb_node *node;
+	u64 seq = 0;
+
+	spin_lock(&fs_info->tree_mod_seq_lock);
+	if (!list_empty(&fs_info->tree_mod_seq_list)) {
+		struct seq_list *elem;
+
+		elem = list_first_entry(&fs_info->tree_mod_seq_list,
+					struct seq_list, list);
+		seq = elem->seq;
+	}
+	spin_unlock(&fs_info->tree_mod_seq_lock);
+
+	node = rb_prev(&head->node.rb_node);
+	while (node) {
+		struct btrfs_delayed_ref_node *ref;
+
+		ref = rb_entry(node, struct btrfs_delayed_ref_node,
+			       rb_node);
+		if (ref->bytenr != head->node.bytenr)
+			break;
+
+		/* We can't merge refs that are outside of our seq count */
+		if (seq && ref->seq >= seq)
+			break;
+		if (merge_ref(trans, delayed_refs, ref, seq))
+			node = rb_prev(&head->node.rb_node);
+		else
+			node = rb_prev(node);
+	}
+}
+
 int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info,
 			    struct btrfs_delayed_ref_root *delayed_refs,
 			    u64 seq)
@@ -336,18 +444,11 @@ update_existing_ref(struct btrfs_trans_handle *trans,
 		 * every changing the extent allocation tree.
 		 */
 		existing->ref_mod--;
-		if (existing->ref_mod == 0) {
-			rb_erase(&existing->rb_node,
-				 &delayed_refs->root);
-			existing->in_tree = 0;
-			btrfs_put_delayed_ref(existing);
-			delayed_refs->num_entries--;
-			if (trans->delayed_ref_updates)
-				trans->delayed_ref_updates--;
-		} else {
+		if (existing->ref_mod == 0)
+			drop_delayed_ref(trans, delayed_refs, existing);
+		else
 			WARN_ON(existing->type == BTRFS_TREE_BLOCK_REF_KEY ||
 				existing->type == BTRFS_SHARED_BLOCK_REF_KEY);
-		}
 	} else {
 		WARN_ON(existing->type == BTRFS_TREE_BLOCK_REF_KEY ||
 			existing->type == BTRFS_SHARED_BLOCK_REF_KEY);
@@ -662,9 +763,6 @@ int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
 	add_delayed_tree_ref(fs_info, trans, &ref->node, bytenr,
 				   num_bytes, parent, ref_root, level, action,
 				   for_cow);
-	if (!need_ref_seq(for_cow, ref_root) &&
-	    waitqueue_active(&fs_info->tree_mod_seq_wait))
-		wake_up(&fs_info->tree_mod_seq_wait);
 	spin_unlock(&delayed_refs->lock);
 	if (need_ref_seq(for_cow, ref_root))
 		btrfs_qgroup_record_ref(trans, &ref->node, extent_op);
@@ -713,9 +811,6 @@ int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info,
 	add_delayed_data_ref(fs_info, trans, &ref->node, bytenr,
 				   num_bytes, parent, ref_root, owner, offset,
 				   action, for_cow);
-	if (!need_ref_seq(for_cow, ref_root) &&
-	    waitqueue_active(&fs_info->tree_mod_seq_wait))
-		wake_up(&fs_info->tree_mod_seq_wait);
 	spin_unlock(&delayed_refs->lock);
 	if (need_ref_seq(for_cow, ref_root))
 		btrfs_qgroup_record_ref(trans, &ref->node, extent_op);
@@ -744,8 +839,6 @@ int btrfs_add_delayed_extent_op(struct btrfs_fs_info *fs_info,
 				   num_bytes, BTRFS_UPDATE_DELAYED_HEAD,
 				   extent_op->is_data);
 
-	if (waitqueue_active(&fs_info->tree_mod_seq_wait))
-		wake_up(&fs_info->tree_mod_seq_wait);
 	spin_unlock(&delayed_refs->lock);
 	return 0;
 }
diff --git a/fs/btrfs/delayed-ref.h b/fs/btrfs/delayed-ref.h
index 0d7c90c366b6..ab5300595847 100644
--- a/fs/btrfs/delayed-ref.h
+++ b/fs/btrfs/delayed-ref.h
@@ -167,6 +167,10 @@ int btrfs_add_delayed_extent_op(struct btrfs_fs_info *fs_info,
 				struct btrfs_trans_handle *trans,
 				u64 bytenr, u64 num_bytes,
 				struct btrfs_delayed_extent_op *extent_op);
+void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
+			      struct btrfs_fs_info *fs_info,
+			      struct btrfs_delayed_ref_root *delayed_refs,
+			      struct btrfs_delayed_ref_head *head);
 
 struct btrfs_delayed_ref_head *
 btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 62e0cafd6e25..22e98e04c2ea 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -377,9 +377,13 @@ static int btree_read_extent_buffer_pages(struct btrfs_root *root,
 		ret = read_extent_buffer_pages(io_tree, eb, start,
 					       WAIT_COMPLETE,
 					       btree_get_extent, mirror_num);
-		if (!ret && !verify_parent_transid(io_tree, eb,
+		if (!ret) {
+			if (!verify_parent_transid(io_tree, eb,
 						   parent_transid, 0))
-			break;
+				break;
+			else
+				ret = -EIO;
+		}
 
 		/*
 		 * This buffer's crc is fine, but its contents are corrupted, so
@@ -754,9 +758,7 @@ static void run_one_async_done(struct btrfs_work *work)
 	limit = btrfs_async_submit_limit(fs_info);
 	limit = limit * 2 / 3;
 
-	atomic_dec(&fs_info->nr_async_submits);
-
-	if (atomic_read(&fs_info->nr_async_submits) < limit &&
+	if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
 	    waitqueue_active(&fs_info->async_submit_wait))
 		wake_up(&fs_info->async_submit_wait);
 
@@ -2032,8 +2034,6 @@ int open_ctree(struct super_block *sb,
 	fs_info->free_chunk_space = 0;
 	fs_info->tree_mod_log = RB_ROOT;
 
-	init_waitqueue_head(&fs_info->tree_mod_seq_wait);
-
 	/* readahead state */
 	INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
 	spin_lock_init(&fs_info->reada_lock);
@@ -2528,8 +2528,7 @@ retry_root_backup:
 		goto fail_trans_kthread;
 
 	/* do not make disk changes in broken FS */
-	if (btrfs_super_log_root(disk_super) != 0 &&
-	    !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) {
+	if (btrfs_super_log_root(disk_super) != 0) {
 		u64 bytenr = btrfs_super_log_root(disk_super);
 
 		if (fs_devices->rw_devices == 0) {
@@ -3189,30 +3188,14 @@ int close_ctree(struct btrfs_root *root)
 	/* clear out the rbtree of defraggable inodes */
 	btrfs_run_defrag_inodes(fs_info);
 
-	/*
-	 * Here come 2 situations when btrfs is broken to flip readonly:
-	 *
-	 * 1. when btrfs flips readonly somewhere else before
-	 * btrfs_commit_super, sb->s_flags has MS_RDONLY flag,
-	 * and btrfs will skip to write sb directly to keep
-	 * ERROR state on disk.
-	 *
-	 * 2. when btrfs flips readonly just in btrfs_commit_super,
-	 * and in such case, btrfs cannot write sb via btrfs_commit_super,
-	 * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag,
-	 * btrfs will cleanup all FS resources first and write sb then.
-	 */
 	if (!(fs_info->sb->s_flags & MS_RDONLY)) {
 		ret = btrfs_commit_super(root);
 		if (ret)
 			printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
 	}
 
-	if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
-		ret = btrfs_error_commit_super(root);
-		if (ret)
-			printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
-	}
+	if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
+		btrfs_error_commit_super(root);
 
 	btrfs_put_block_group_cache(fs_info);
 
@@ -3434,18 +3417,11 @@ static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
 	if (read_only)
 		return 0;
 
-	if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
-		printk(KERN_WARNING "warning: mount fs with errors, "
-		       "running btrfsck is recommended\n");
-	}
-
 	return 0;
 }
 
-int btrfs_error_commit_super(struct btrfs_root *root)
+void btrfs_error_commit_super(struct btrfs_root *root)
 {
-	int ret;
-
 	mutex_lock(&root->fs_info->cleaner_mutex);
 	btrfs_run_delayed_iputs(root);
 	mutex_unlock(&root->fs_info->cleaner_mutex);
@@ -3455,10 +3431,6 @@ int btrfs_error_commit_super(struct btrfs_root *root)
 
 	/* cleanup FS via transaction */
 	btrfs_cleanup_transaction(root);
-
-	ret = write_ctree_super(NULL, root, 0);
-
-	return ret;
 }
 
 static void btrfs_destroy_ordered_operations(struct btrfs_root *root)
@@ -3782,14 +3754,17 @@ int btrfs_cleanup_transaction(struct btrfs_root *root)
 		/* FIXME: cleanup wait for commit */
 		t->in_commit = 1;
 		t->blocked = 1;
+		smp_mb();
 		if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
 			wake_up(&root->fs_info->transaction_blocked_wait);
 
 		t->blocked = 0;
+		smp_mb();
 		if (waitqueue_active(&root->fs_info->transaction_wait))
 			wake_up(&root->fs_info->transaction_wait);
 
 		t->commit_done = 1;
+		smp_mb();
 		if (waitqueue_active(&t->commit_wait))
 			wake_up(&t->commit_wait);
 
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 95e147eea239..c5b00a735fef 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -54,7 +54,7 @@ int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root, int max_mirrors);
 struct buffer_head *btrfs_read_dev_super(struct block_device *bdev);
 int btrfs_commit_super(struct btrfs_root *root);
-int btrfs_error_commit_super(struct btrfs_root *root);
+void btrfs_error_commit_super(struct btrfs_root *root);
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
 					    u64 bytenr, u32 blocksize);
 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 4e1b153b7c47..ba58024d40d3 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -2252,6 +2252,16 @@ static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
 		}
 
 		/*
+		 * We need to try and merge add/drops of the same ref since we
+		 * can run into issues with relocate dropping the implicit ref
+		 * and then it being added back again before the drop can
+		 * finish.  If we merged anything we need to re-loop so we can
+		 * get a good ref.
+		 */
+		btrfs_merge_delayed_refs(trans, fs_info, delayed_refs,
+					 locked_ref);
+
+		/*
 		 * locked_ref is the head node, so we have to go one
 		 * node back for any delayed ref updates
 		 */
@@ -2318,12 +2328,23 @@ static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
 		ref->in_tree = 0;
 		rb_erase(&ref->rb_node, &delayed_refs->root);
 		delayed_refs->num_entries--;
-		/*
-		 * we modified num_entries, but as we're currently running
-		 * delayed refs, skip
-		 *     wake_up(&delayed_refs->seq_wait);
-		 * here.
-		 */
+		if (locked_ref) {
+			/*
+			 * when we play the delayed ref, also correct the
+			 * ref_mod on head
+			 */
+			switch (ref->action) {
+			case BTRFS_ADD_DELAYED_REF:
+			case BTRFS_ADD_DELAYED_EXTENT:
+				locked_ref->node.ref_mod -= ref->ref_mod;
+				break;
+			case BTRFS_DROP_DELAYED_REF:
+				locked_ref->node.ref_mod += ref->ref_mod;
+				break;
+			default:
+				WARN_ON(1);
+			}
+		}
 		spin_unlock(&delayed_refs->lock);
 
 		ret = run_one_delayed_ref(trans, root, ref, extent_op,
@@ -2350,22 +2371,6 @@ next:
 	return count;
 }
 
-static void wait_for_more_refs(struct btrfs_fs_info *fs_info,
-			       struct btrfs_delayed_ref_root *delayed_refs,
-			       unsigned long num_refs,
-			       struct list_head *first_seq)
-{
-	spin_unlock(&delayed_refs->lock);
-	pr_debug("waiting for more refs (num %ld, first %p)\n",
-		 num_refs, first_seq);
-	wait_event(fs_info->tree_mod_seq_wait,
-		   num_refs != delayed_refs->num_entries ||
-		   fs_info->tree_mod_seq_list.next != first_seq);
-	pr_debug("done waiting for more refs (num %ld, first %p)\n",
-		 delayed_refs->num_entries, fs_info->tree_mod_seq_list.next);
-	spin_lock(&delayed_refs->lock);
-}
-
 #ifdef SCRAMBLE_DELAYED_REFS
 /*
  * Normally delayed refs get processed in ascending bytenr order. This
@@ -2460,13 +2465,11 @@ int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
 	struct btrfs_delayed_ref_root *delayed_refs;
 	struct btrfs_delayed_ref_node *ref;
 	struct list_head cluster;
-	struct list_head *first_seq = NULL;
 	int ret;
 	u64 delayed_start;
 	int run_all = count == (unsigned long)-1;
 	int run_most = 0;
-	unsigned long num_refs = 0;
-	int consider_waiting;
+	int loops;
 
 	/* We'll clean this up in btrfs_cleanup_transaction */
 	if (trans->aborted)
@@ -2484,7 +2487,7 @@ int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
 	delayed_refs = &trans->transaction->delayed_refs;
 	INIT_LIST_HEAD(&cluster);
 again:
-	consider_waiting = 0;
+	loops = 0;
 	spin_lock(&delayed_refs->lock);
 
 #ifdef SCRAMBLE_DELAYED_REFS
@@ -2512,31 +2515,6 @@ again:
 		if (ret)
 			break;
 
-		if (delayed_start >= delayed_refs->run_delayed_start) {
-			if (consider_waiting == 0) {
-				/*
-				 * btrfs_find_ref_cluster looped. let's do one
-				 * more cycle. if we don't run any delayed ref
-				 * during that cycle (because we can't because
-				 * all of them are blocked) and if the number of
-				 * refs doesn't change, we avoid busy waiting.
-				 */
-				consider_waiting = 1;
-				num_refs = delayed_refs->num_entries;
-				first_seq = root->fs_info->tree_mod_seq_list.next;
-			} else {
-				wait_for_more_refs(root->fs_info, delayed_refs,
-						   num_refs, first_seq);
-				/*
-				 * after waiting, things have changed. we
-				 * dropped the lock and someone else might have
-				 * run some refs, built new clusters and so on.
-				 * therefore, we restart staleness detection.
-				 */
-				consider_waiting = 0;
-			}
-		}
-
 		ret = run_clustered_refs(trans, root, &cluster);
 		if (ret < 0) {
 			spin_unlock(&delayed_refs->lock);
@@ -2549,9 +2527,26 @@ again:
 		if (count == 0)
 			break;
 
-		if (ret || delayed_refs->run_delayed_start == 0) {
+		if (delayed_start >= delayed_refs->run_delayed_start) {
+			if (loops == 0) {
+				/*
+				 * btrfs_find_ref_cluster looped. let's do one
+				 * more cycle. if we don't run any delayed ref
+				 * during that cycle (because we can't because
+				 * all of them are blocked), bail out.
+				 */
+				loops = 1;
+			} else {
+				/*
+				 * no runnable refs left, stop trying
+				 */
+				BUG_ON(run_all);
+				break;
+			}
+		}
+		if (ret) {
 			/* refs were run, let's reset staleness detection */
-			consider_waiting = 0;
+			loops = 0;
 		}
 	}
 
@@ -3007,17 +3002,16 @@ again:
 	}
 	spin_unlock(&block_group->lock);
 
-	num_pages = (int)div64_u64(block_group->key.offset, 1024 * 1024 * 1024);
+	/*
+	 * Try to preallocate enough space based on how big the block group is.
+	 * Keep in mind this has to include any pinned space which could end up
+	 * taking up quite a bit since it's not folded into the other space
+	 * cache.
+	 */
+	num_pages = (int)div64_u64(block_group->key.offset, 256 * 1024 * 1024);
 	if (!num_pages)
 		num_pages = 1;
 
-	/*
-	 * Just to make absolutely sure we have enough space, we're going to
-	 * preallocate 12 pages worth of space for each block group.  In
-	 * practice we ought to use at most 8, but we need extra space so we can
-	 * add our header and have a terminator between the extents and the
-	 * bitmaps.
-	 */
 	num_pages *= 16;
 	num_pages *= PAGE_CACHE_SIZE;
 
@@ -4571,8 +4565,10 @@ int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes)
 	if (root->fs_info->quota_enabled) {
 		ret = btrfs_qgroup_reserve(root, num_bytes +
 					   nr_extents * root->leafsize);
-		if (ret)
+		if (ret) {
+			mutex_unlock(&BTRFS_I(inode)->delalloc_mutex);
 			return ret;
+		}
 	}
 
 	ret = reserve_metadata_bytes(root, block_rsv, to_reserve, flush);
@@ -5294,9 +5290,6 @@ static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
 	rb_erase(&head->node.rb_node, &delayed_refs->root);
 
 	delayed_refs->num_entries--;
-	smp_mb();
-	if (waitqueue_active(&root->fs_info->tree_mod_seq_wait))
-		wake_up(&root->fs_info->tree_mod_seq_wait);
 
 	/*
 	 * we don't take a ref on the node because we're removing it from the
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 45c81bb4ac82..4c878476bb91 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -2330,23 +2330,10 @@ static void end_bio_extent_readpage(struct bio *bio, int err)
 		if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) {
 			ret = tree->ops->readpage_end_io_hook(page, start, end,
 							      state, mirror);
-			if (ret) {
-				/* no IO indicated but software detected errors
-				 * in the block, either checksum errors or
-				 * issues with the contents */
-				struct btrfs_root *root =
-					BTRFS_I(page->mapping->host)->root;
-				struct btrfs_device *device;
-
+			if (ret)
 				uptodate = 0;
-				device = btrfs_find_device_for_logical(
-						root, start, mirror);
-				if (device)
-					btrfs_dev_stat_inc_and_print(device,
-						BTRFS_DEV_STAT_CORRUPTION_ERRS);
-			} else {
+			else
 				clean_io_failure(start, page);
-			}
 		}
 
 		if (!uptodate && tree->ops && tree->ops->readpage_io_failed_hook) {
diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c
index b45b9de0c21d..857d93cd01dc 100644
--- a/fs/btrfs/file-item.c
+++ b/fs/btrfs/file-item.c
@@ -272,9 +272,9 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
 }
 
 int btrfs_lookup_bio_sums_dio(struct btrfs_root *root, struct inode *inode,
-			      struct bio *bio, u64 offset, u32 *dst)
+			      struct bio *bio, u64 offset)
 {
-	return __btrfs_lookup_bio_sums(root, inode, bio, offset, dst, 1);
+	return __btrfs_lookup_bio_sums(root, inode, bio, offset, NULL, 1);
 }
 
 int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end,
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 6e8f416773d4..ec154f954646 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -1008,9 +1008,7 @@ static noinline void async_cow_submit(struct btrfs_work *work)
 	nr_pages = (async_cow->end - async_cow->start + PAGE_CACHE_SIZE) >>
 		PAGE_CACHE_SHIFT;
 
-	atomic_sub(nr_pages, &root->fs_info->async_delalloc_pages);
-
-	if (atomic_read(&root->fs_info->async_delalloc_pages) <
+	if (atomic_sub_return(nr_pages, &root->fs_info->async_delalloc_pages) <
 	    5 * 1024 * 1024 &&
 	    waitqueue_active(&root->fs_info->async_submit_wait))
 		wake_up(&root->fs_info->async_submit_wait);
@@ -1885,8 +1883,11 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent)
 				trans = btrfs_join_transaction_nolock(root);
 			else
 				trans = btrfs_join_transaction(root);
-			if (IS_ERR(trans))
-				return PTR_ERR(trans);
+			if (IS_ERR(trans)) {
+				ret = PTR_ERR(trans);
+				trans = NULL;
+				goto out;
+			}
 			trans->block_rsv = &root->fs_info->delalloc_block_rsv;
 			ret = btrfs_update_inode_fallback(trans, root, inode);
 			if (ret) /* -ENOMEM or corruption */
@@ -3174,7 +3175,7 @@ int btrfs_unlink_subvol(struct btrfs_trans_handle *trans,
 	btrfs_i_size_write(dir, dir->i_size - name_len * 2);
 	inode_inc_iversion(dir);
 	dir->i_mtime = dir->i_ctime = CURRENT_TIME;
-	ret = btrfs_update_inode(trans, root, dir);
+	ret = btrfs_update_inode_fallback(trans, root, dir);
 	if (ret)
 		btrfs_abort_transaction(trans, root, ret);
 out:
@@ -5774,18 +5775,112 @@ out:
 	return ret;
 }
 
+static int lock_extent_direct(struct inode *inode, u64 lockstart, u64 lockend,
+			      struct extent_state **cached_state, int writing)
+{
+	struct btrfs_ordered_extent *ordered;
+	int ret = 0;
+
+	while (1) {
+		lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend,
+				 0, cached_state);
+		/*
+		 * We're concerned with the entire range that we're going to be
+		 * doing DIO to, so we need to make sure theres no ordered
+		 * extents in this range.
+		 */
+		ordered = btrfs_lookup_ordered_range(inode, lockstart,
+						     lockend - lockstart + 1);
+
+		/*
+		 * We need to make sure there are no buffered pages in this
+		 * range either, we could have raced between the invalidate in
+		 * generic_file_direct_write and locking the extent.  The
+		 * invalidate needs to happen so that reads after a write do not
+		 * get stale data.
+		 */
+		if (!ordered && (!writing ||
+		    !test_range_bit(&BTRFS_I(inode)->io_tree,
+				    lockstart, lockend, EXTENT_UPTODATE, 0,
+				    *cached_state)))
+			break;
+
+		unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
+				     cached_state, GFP_NOFS);
+
+		if (ordered) {
+			btrfs_start_ordered_extent(inode, ordered, 1);
+			btrfs_put_ordered_extent(ordered);
+		} else {
+			/* Screw you mmap */
+			ret = filemap_write_and_wait_range(inode->i_mapping,
+							   lockstart,
+							   lockend);
+			if (ret)
+				break;
+
+			/*
+			 * If we found a page th