Merge git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable

* git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable:
  Btrfs: fix file clone ioctl for bookend extents
  Btrfs: fix uninit compiler warning in cow_file_range_nocow
  Btrfs: constify dentry_operations
  Btrfs: optimize back reference update during btrfs_drop_snapshot
  Btrfs: remove negative dentry when deleting subvolumne
  Btrfs: optimize fsync for the single writer case
  Btrfs: async delalloc flushing under space pressure
  Btrfs: release delalloc reservations on extent item insertion
  Btrfs: delay clearing EXTENT_DELALLOC for compressed extents
  Btrfs: cleanup extent_clear_unlock_delalloc flags
  Btrfs: fix possible softlockup in the allocator
  Btrfs: fix deadlock on async thread startup

14 files changed, 447 insertions, 157 deletions

diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c
index 282ca085c2fb..c0861e781cdb 100644
--- a/fs/btrfs/async-thread.c
+++ b/fs/btrfs/async-thread.c
@@ -64,6 +64,51 @@ struct btrfs_worker_thread {
 };
 
 /*
+ * btrfs_start_workers uses kthread_run, which can block waiting for memory
+ * for a very long time.  It will actually throttle on page writeback,
+ * and so it may not make progress until after our btrfs worker threads
+ * process all of the pending work structs in their queue
+ *
+ * This means we can't use btrfs_start_workers from inside a btrfs worker
+ * thread that is used as part of cleaning dirty memory, which pretty much
+ * involves all of the worker threads.
+ *
+ * Instead we have a helper queue who never has more than one thread
+ * where we scheduler thread start operations.  This worker_start struct
+ * is used to contain the work and hold a pointer to the queue that needs
+ * another worker.
+ */
+struct worker_start {
+	struct btrfs_work work;
+	struct btrfs_workers *queue;
+};
+
+static void start_new_worker_func(struct btrfs_work *work)
+{
+	struct worker_start *start;
+	start = container_of(work, struct worker_start, work);
+	btrfs_start_workers(start->queue, 1);
+	kfree(start);
+}
+
+static int start_new_worker(struct btrfs_workers *queue)
+{
+	struct worker_start *start;
+	int ret;
+
+	start = kzalloc(sizeof(*start), GFP_NOFS);
+	if (!start)
+		return -ENOMEM;
+
+	start->work.func = start_new_worker_func;
+	start->queue = queue;
+	ret = btrfs_queue_worker(queue->atomic_worker_start, &start->work);
+	if (ret)
+		kfree(start);
+	return ret;
+}
+
+/*
  * helper function to move a thread onto the idle list after it
  * has finished some requests.
  */
@@ -118,11 +163,13 @@ static void check_pending_worker_creates(struct btrfs_worker_thread *worker)
 		goto out;
 
 	workers->atomic_start_pending = 0;
-	if (workers->num_workers >= workers->max_workers)
+	if (workers->num_workers + workers->num_workers_starting >=
+	    workers->max_workers)
 		goto out;
 
+	workers->num_workers_starting += 1;
 	spin_unlock_irqrestore(&workers->lock, flags);
-	btrfs_start_workers(workers, 1);
+	start_new_worker(workers);
 	return;
 
 out:
@@ -390,9 +437,11 @@ int btrfs_stop_workers(struct btrfs_workers *workers)
 /*
  * simple init on struct btrfs_workers
  */
-void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max)
+void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max,
+			struct btrfs_workers *async_helper)
 {
 	workers->num_workers = 0;
+	workers->num_workers_starting = 0;
 	INIT_LIST_HEAD(&workers->worker_list);
 	INIT_LIST_HEAD(&workers->idle_list);
 	INIT_LIST_HEAD(&workers->order_list);
@@ -404,14 +453,15 @@ void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max)
 	workers->name = name;
 	workers->ordered = 0;
 	workers->atomic_start_pending = 0;
-	workers->atomic_worker_start = 0;
+	workers->atomic_worker_start = async_helper;
 }
 
 /*
  * starts new worker threads.  This does not enforce the max worker
  * count in case you need to temporarily go past it.
  */
-int btrfs_start_workers(struct btrfs_workers *workers, int num_workers)
+static int __btrfs_start_workers(struct btrfs_workers *workers,
+				 int num_workers)
 {
 	struct btrfs_worker_thread *worker;
 	int ret = 0;
@@ -444,6 +494,8 @@ int btrfs_start_workers(struct btrfs_workers *workers, int num_workers)
 		list_add_tail(&worker->worker_list, &workers->idle_list);
 		worker->idle = 1;
 		workers->num_workers++;
+		workers->num_workers_starting--;
+		WARN_ON(workers->num_workers_starting < 0);
 		spin_unlock_irq(&workers->lock);
 	}
 	return 0;
@@ -452,6 +504,14 @@ fail:
 	return ret;
 }
 
+int btrfs_start_workers(struct btrfs_workers *workers, int num_workers)
+{
+	spin_lock_irq(&workers->lock);
+	workers->num_workers_starting += num_workers;
+	spin_unlock_irq(&workers->lock);
+	return __btrfs_start_workers(workers, num_workers);
+}
+
 /*
  * run through the list and find a worker thread that doesn't have a lot
  * to do right now.  This can return null if we aren't yet at the thread
@@ -461,7 +521,10 @@ static struct btrfs_worker_thread *next_worker(struct btrfs_workers *workers)
 {
 	struct btrfs_worker_thread *worker;
 	struct list_head *next;
-	int enforce_min = workers->num_workers < workers->max_workers;
+	int enforce_min;
+
+	enforce_min = (workers->num_workers + workers->num_workers_starting) <
+		workers->max_workers;
 
 	/*
 	 * if we find an idle thread, don't move it to the end of the
@@ -509,15 +572,17 @@ again:
 	worker = next_worker(workers);
 
 	if (!worker) {
-		if (workers->num_workers >= workers->max_workers) {
+		if (workers->num_workers + workers->num_workers_starting >=
+		    workers->max_workers) {
 			goto fallback;
 		} else if (workers->atomic_worker_start) {
 			workers->atomic_start_pending = 1;
 			goto fallback;
 		} else {
+			workers->num_workers_starting++;
 			spin_unlock_irqrestore(&workers->lock, flags);
 			/* we're below the limit, start another worker */
-			btrfs_start_workers(workers, 1);
+			__btrfs_start_workers(workers, 1);
 			goto again;
 		}
 	}
diff --git a/fs/btrfs/async-thread.h b/fs/btrfs/async-thread.h
index fc089b95ec14..5077746cf85e 100644
--- a/fs/btrfs/async-thread.h
+++ b/fs/btrfs/async-thread.h
@@ -64,6 +64,8 @@ struct btrfs_workers {
 	/* current number of running workers */
 	int num_workers;
 
+	int num_workers_starting;
+
 	/* max number of workers allowed.  changed by btrfs_start_workers */
 	int max_workers;
 
@@ -78,9 +80,10 @@ struct btrfs_workers {
 
 	/*
 	 * are we allowed to sleep while starting workers or are we required
-	 * to start them at a later time?
+	 * to start them at a later time?  If we can't sleep, this indicates
+	 * which queue we need to use to schedule thread creation.
 	 */
-	int atomic_worker_start;
+	struct btrfs_workers *atomic_worker_start;
 
 	/* list with all the work threads.  The workers on the idle thread
 	 * may be actively servicing jobs, but they haven't yet hit the
@@ -109,7 +112,8 @@ struct btrfs_workers {
 int btrfs_queue_worker(struct btrfs_workers *workers, struct btrfs_work *work);
 int btrfs_start_workers(struct btrfs_workers *workers, int num_workers);
 int btrfs_stop_workers(struct btrfs_workers *workers);
-void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max);
+void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max,
+			struct btrfs_workers *async_starter);
 int btrfs_requeue_work(struct btrfs_work *work);
 void btrfs_set_work_high_prio(struct btrfs_work *work);
 #endif
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index a54d354cefcb..c71abec0ab90 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -128,12 +128,14 @@ struct btrfs_inode {
 	u64 last_unlink_trans;
 
 	/*
-	 * These two counters are for delalloc metadata reservations.  We keep
-	 * track of how many extents we've accounted for vs how many extents we
-	 * have.
+	 * Counters to keep track of the number of extent item's we may use due
+	 * to delalloc and such.  outstanding_extents is the number of extent
+	 * items we think we'll end up using, and reserved_extents is the number
+	 * of extent items we've reserved metadata for.
 	 */
-	int delalloc_reserved_extents;
-	int delalloc_extents;
+	spinlock_t accounting_lock;
+	int reserved_extents;
+	int outstanding_extents;
 
 	/*
 	 * ordered_data_close is set by truncate when a file that used
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index dd8ced9814c4..1bb897ecdeeb 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -691,14 +691,17 @@ struct btrfs_space_info {
 
 	struct list_head list;
 
+	/* for controlling how we free up space for allocations */
+	wait_queue_head_t allocate_wait;
+	wait_queue_head_t flush_wait;
+	int allocating_chunk;
+	int flushing;
+
 	/* for block groups in our same type */
 	struct list_head block_groups;
 	spinlock_t lock;
 	struct rw_semaphore groups_sem;
 	atomic_t caching_threads;
-
-	int allocating_chunk;
-	wait_queue_head_t wait;
 };
 
 /*
@@ -907,6 +910,7 @@ struct btrfs_fs_info {
 	 * A third pool does submit_bio to avoid deadlocking with the other
 	 * two
 	 */
+	struct btrfs_workers generic_worker;
 	struct btrfs_workers workers;
 	struct btrfs_workers delalloc_workers;
 	struct btrfs_workers endio_workers;
@@ -914,6 +918,7 @@ struct btrfs_fs_info {
 	struct btrfs_workers endio_meta_write_workers;
 	struct btrfs_workers endio_write_workers;
 	struct btrfs_workers submit_workers;
+	struct btrfs_workers enospc_workers;
 	/*
 	 * fixup workers take dirty pages that didn't properly go through
 	 * the cow mechanism and make them safe to write.  It happens
@@ -1005,6 +1010,8 @@ struct btrfs_root {
 	atomic_t log_commit[2];
 	unsigned long log_transid;
 	unsigned long log_batch;
+	pid_t log_start_pid;
+	bool log_multiple_pids;
 
 	u64 objectid;
 	u64 last_trans;
@@ -2323,7 +2330,7 @@ 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;
+extern const 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 af0435f79fa6..100551a66c46 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1746,21 +1746,25 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 		err = -EINVAL;
 		goto fail_iput;
 	}
-printk("thread pool is %d\n", fs_info->thread_pool_size);
-	/*
-	 * we need to start all the end_io workers up front because the
-	 * queue work function gets called at interrupt time, and so it
-	 * cannot dynamically grow.
-	 */
+
+	btrfs_init_workers(&fs_info->generic_worker,
+			   "genwork", 1, NULL);
+
 	btrfs_init_workers(&fs_info->workers, "worker",
-			   fs_info->thread_pool_size);
+			   fs_info->thread_pool_size,
+			   &fs_info->generic_worker);
 
 	btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
-			   fs_info->thread_pool_size);
+			   fs_info->thread_pool_size,
+			   &fs_info->generic_worker);
 
 	btrfs_init_workers(&fs_info->submit_workers, "submit",
 			   min_t(u64, fs_devices->num_devices,
-			   fs_info->thread_pool_size));
+			   fs_info->thread_pool_size),
+			   &fs_info->generic_worker);
+	btrfs_init_workers(&fs_info->enospc_workers, "enospc",
+			   fs_info->thread_pool_size,
+			   &fs_info->generic_worker);
 
 	/* a higher idle thresh on the submit workers makes it much more
 	 * likely that bios will be send down in a sane order to the
@@ -1774,15 +1778,20 @@ printk("thread pool is %d\n", fs_info->thread_pool_size);
 	fs_info->delalloc_workers.idle_thresh = 2;
 	fs_info->delalloc_workers.ordered = 1;
 
-	btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1);
+	btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
+			   &fs_info->generic_worker);
 	btrfs_init_workers(&fs_info->endio_workers, "endio",
-			   fs_info->thread_pool_size);
+			   fs_info->thread_pool_size,
+			   &fs_info->generic_worker);
 	btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
-			   fs_info->thread_pool_size);
+			   fs_info->thread_pool_size,
+			   &fs_info->generic_worker);
 	btrfs_init_workers(&fs_info->endio_meta_write_workers,
-			   "endio-meta-write", fs_info->thread_pool_size);
+			   "endio-meta-write", fs_info->thread_pool_size,
+			   &fs_info->generic_worker);
 	btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
-			   fs_info->thread_pool_size);
+			   fs_info->thread_pool_size,
+			   &fs_info->generic_worker);
 
 	/*
 	 * endios are largely parallel and should have a very
@@ -1794,12 +1803,8 @@ printk("thread pool is %d\n", fs_info->thread_pool_size);
 	fs_info->endio_write_workers.idle_thresh = 2;
 	fs_info->endio_meta_write_workers.idle_thresh = 2;
 
-	fs_info->endio_workers.atomic_worker_start = 1;
-	fs_info->endio_meta_workers.atomic_worker_start = 1;
-	fs_info->endio_write_workers.atomic_worker_start = 1;
-	fs_info->endio_meta_write_workers.atomic_worker_start = 1;
-
 	btrfs_start_workers(&fs_info->workers, 1);
+	btrfs_start_workers(&fs_info->generic_worker, 1);
 	btrfs_start_workers(&fs_info->submit_workers, 1);
 	btrfs_start_workers(&fs_info->delalloc_workers, 1);
 	btrfs_start_workers(&fs_info->fixup_workers, 1);
@@ -1807,6 +1812,7 @@ printk("thread pool is %d\n", fs_info->thread_pool_size);
 	btrfs_start_workers(&fs_info->endio_meta_workers, 1);
 	btrfs_start_workers(&fs_info->endio_meta_write_workers, 1);
 	btrfs_start_workers(&fs_info->endio_write_workers, 1);
+	btrfs_start_workers(&fs_info->enospc_workers, 1);
 
 	fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
 	fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
@@ -2012,6 +2018,7 @@ fail_chunk_root:
 	free_extent_buffer(chunk_root->node);
 	free_extent_buffer(chunk_root->commit_root);
 fail_sb_buffer:
+	btrfs_stop_workers(&fs_info->generic_worker);
 	btrfs_stop_workers(&fs_info->fixup_workers);
 	btrfs_stop_workers(&fs_info->delalloc_workers);
 	btrfs_stop_workers(&fs_info->workers);
@@ -2020,6 +2027,7 @@ fail_sb_buffer:
 	btrfs_stop_workers(&fs_info->endio_meta_write_workers);
 	btrfs_stop_workers(&fs_info->endio_write_workers);
 	btrfs_stop_workers(&fs_info->submit_workers);
+	btrfs_stop_workers(&fs_info->enospc_workers);
 fail_iput:
 	invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
 	iput(fs_info->btree_inode);
@@ -2437,6 +2445,7 @@ int close_ctree(struct btrfs_root *root)
 
 	iput(fs_info->btree_inode);
 
+	btrfs_stop_workers(&fs_info->generic_worker);
 	btrfs_stop_workers(&fs_info->fixup_workers);
 	btrfs_stop_workers(&fs_info->delalloc_workers);
 	btrfs_stop_workers(&fs_info->workers);
@@ -2445,6 +2454,7 @@ int close_ctree(struct btrfs_root *root)
 	btrfs_stop_workers(&fs_info->endio_meta_write_workers);
 	btrfs_stop_workers(&fs_info->endio_write_workers);
 	btrfs_stop_workers(&fs_info->submit_workers);
+	btrfs_stop_workers(&fs_info->enospc_workers);
 
 	btrfs_close_devices(fs_info->fs_devices);
 	btrfs_mapping_tree_free(&fs_info->mapping_tree);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 359a754c782c..d0c4d584efad 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -2824,14 +2824,17 @@ int btrfs_unreserve_metadata_for_delalloc(struct btrfs_root *root,
 					   num_items);
 
 	spin_lock(&meta_sinfo->lock);
-	if (BTRFS_I(inode)->delalloc_reserved_extents <=
-	    BTRFS_I(inode)->delalloc_extents) {
+	spin_lock(&BTRFS_I(inode)->accounting_lock);
+	if (BTRFS_I(inode)->reserved_extents <=
+	    BTRFS_I(inode)->outstanding_extents) {
+		spin_unlock(&BTRFS_I(inode)->accounting_lock);
 		spin_unlock(&meta_sinfo->lock);
 		return 0;
 	}
+	spin_unlock(&BTRFS_I(inode)->accounting_lock);
 
-	BTRFS_I(inode)->delalloc_reserved_extents--;
-	BUG_ON(BTRFS_I(inode)->delalloc_reserved_extents < 0);
+	BTRFS_I(inode)->reserved_extents--;
+	BUG_ON(BTRFS_I(inode)->reserved_extents < 0);
 
 	if (meta_sinfo->bytes_delalloc < num_bytes) {
 		bug = true;
@@ -2864,6 +2867,107 @@ static void check_force_delalloc(struct btrfs_space_info *meta_sinfo)
 		meta_sinfo->force_delalloc = 0;
 }
 
+struct async_flush {
+	struct btrfs_root *root;
+	struct btrfs_space_info *info;
+	struct btrfs_work work;
+};
+
+static noinline void flush_delalloc_async(struct btrfs_work *work)
+{
+	struct async_flush *async;
+	struct btrfs_root *root;
+	struct btrfs_space_info *info;
+
+	async = container_of(work, struct async_flush, work);
+	root = async->root;
+	info = async->info;
+
+	btrfs_start_delalloc_inodes(root);
+	wake_up(&info->flush_wait);
+	btrfs_wait_ordered_extents(root, 0);
+
+	spin_lock(&info->lock);
+	info->flushing = 0;
+	spin_unlock(&info->lock);
+	wake_up(&info->flush_wait);
+
+	kfree(async);
+}
+
+static void wait_on_flush(struct btrfs_space_info *info)
+{
+	DEFINE_WAIT(wait);
+	u64 used;
+
+	while (1) {
+		prepare_to_wait(&info->flush_wait, &wait,
+				TASK_UNINTERRUPTIBLE);
+		spin_lock(&info->lock);
+		if (!info->flushing) {
+			spin_unlock(&info->lock);
+			break;
+		}
+
+		used = info->bytes_used + info->bytes_reserved +
+			info->bytes_pinned + info->bytes_readonly +
+			info->bytes_super + info->bytes_root +
+			info->bytes_may_use + info->bytes_delalloc;
+		if (used < info->total_bytes) {
+			spin_unlock(&info->lock);
+			break;
+		}
+		spin_unlock(&info->lock);
+		schedule();
+	}
+	finish_wait(&info->flush_wait, &wait);
+}
+
+static void flush_delalloc(struct btrfs_root *root,
+				 struct btrfs_space_info *info)
+{
+	struct async_flush *async;
+	bool wait = false;
+
+	spin_lock(&info->lock);
+
+	if (!info->flushing) {
+		info->flushing = 1;
+		init_waitqueue_head(&info->flush_wait);
+	} else {
+		wait = true;
+	}
+
+	spin_unlock(&info->lock);
+
+	if (wait) {
+		wait_on_flush(info);
+		return;
+	}
+
+	async = kzalloc(sizeof(*async), GFP_NOFS);
+	if (!async)
+		goto flush;
+
+	async->root = root;
+	async->info = info;
+	async->work.func = flush_delalloc_async;
+
+	btrfs_queue_worker(&root->fs_info->enospc_workers,
+			   &async->work);
+	wait_on_flush(info);
+	return;
+
+flush:
+	btrfs_start_delalloc_inodes(root);
+	btrfs_wait_ordered_extents(root, 0);
+
+	spin_lock(&info->lock);
+	info->flushing = 0;
+	spin_unlock(&info->lock);
+	wake_up(&info->flush_wait);
+}
+
 static int maybe_allocate_chunk(struct btrfs_root *root,
 				 struct btrfs_space_info *info)
 {
@@ -2894,7 +2998,7 @@ static int maybe_allocate_chunk(struct btrfs_root *root,
 	if (!info->allocating_chunk) {
 		info->force_alloc = 1;
 		info->allocating_chunk = 1;
-		init_waitqueue_head(&info->wait);
+		init_waitqueue_head(&info->allocate_wait);
 	} else {
 		wait = true;
 	}
@@ -2902,7 +3006,7 @@ static int maybe_allocate_chunk(struct btrfs_root *root,
 	spin_unlock(&info->lock);
 
 	if (wait) {
-		wait_event(info->wait,
+		wait_event(info->allocate_wait,
 			   !info->allocating_chunk);
 		return 1;
 	}
@@ -2923,7 +3027,7 @@ out:
 	spin_lock(&info->lock);
 	info->allocating_chunk = 0;
 	spin_unlock(&info->lock);
-	wake_up(&info->wait);
+	wake_up(&info->allocate_wait);
 
 	if (ret)
 		return 0;
@@ -2981,21 +3085,20 @@ again:
 			filemap_flush(inode->i_mapping);
 			goto again;
 		} else if (flushed == 3) {
-			btrfs_start_delalloc_inodes(root);
-			btrfs_wait_ordered_extents(root, 0);
+			flush_delalloc(root, meta_sinfo);
 			goto again;
 		}
 		spin_lock(&meta_sinfo->lock);
 		meta_sinfo->bytes_delalloc -= num_bytes;
 		spin_unlock(&meta_sinfo->lock);
 		printk(KERN_ERR "enospc, has %d, reserved %d\n",
-		       BTRFS_I(inode)->delalloc_extents,
-		       BTRFS_I(inode)->delalloc_reserved_extents);
+		       BTRFS_I(inode)->outstanding_extents,
+		       BTRFS_I(inode)->reserved_extents);
 		dump_space_info(meta_sinfo, 0, 0);
 		return -ENOSPC;
 	}
 
-	BTRFS_I(inode)->delalloc_reserved_extents++;
+	BTRFS_I(inode)->reserved_extents++;
 	check_force_delalloc(meta_sinfo);
 	spin_unlock(&meta_sinfo->lock);
 
@@ -3094,8 +3197,7 @@ again:
 		}
 
 		if (retries == 2) {
-			btrfs_start_delalloc_inodes(root);
-			btrfs_wait_ordered_extents(root, 0);
+			flush_delalloc(root, meta_sinfo);
 			goto again;
 		}
 		spin_lock(&meta_sinfo->lock);
@@ -4029,6 +4131,7 @@ static noinline int find_free_extent(struct btrfs_trans_handle *trans,
 	int loop = 0;
 	bool found_uncached_bg = false;
 	bool failed_cluster_refill = false;
+	bool failed_alloc = false;
 
 	WARN_ON(num_bytes < root->sectorsize);
 	btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
@@ -4233,14 +4336,23 @@ refill_cluster:
 
 		offset = btrfs_find_space_for_alloc(block_group, search_start,
 						    num_bytes, empty_size);
-		if (!offset && (cached || (!cached &&
-					   loop == LOOP_CACHING_NOWAIT))) {
-			goto loop;
-		} else if (!offset && (!cached &&
-				       loop > LOOP_CACHING_NOWAIT)) {
+		/*
+		 * If we didn't find a chunk, and we haven't failed on this
+		 * block group before, and this block group is in the middle of
+		 * caching and we are ok with waiting, then go ahead and wait
+		 * for progress to be made, and set failed_alloc to true.
+		 *
+		 * If failed_alloc is true then we've already waited on this
+		 * block group once and should move on to the next block group.
+		 */
+		if (!offset && !failed_alloc && !cached &&
+		    loop > LOOP_CACHING_NOWAIT) {
 			wait_block_group_cache_progress(block_group,
-					num_bytes + empty_size);
+						num_bytes + empty_size);
+			failed_alloc = true;
 			goto have_block_group;
+		} else if (!offset) {
+			goto loop;
 		}
 checks:
 		search_start = stripe_align(root, offset);
@@ -4288,6 +4400,7 @@ checks:
 		break;
 loop:
 		failed_cluster_refill = false;
+		failed_alloc = false;
 		btrfs_put_block_group(block_group);
 	}
 	up_read(&space_info->groups_sem);
@@ -4799,6 +4912,7 @@ static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
 	u64 bytenr;
 	u64 generation;
 	u64 refs;
+	u64 flags;
 	u64 last = 0;
 	u32 nritems;
 	u32 blocksize;
@@ -4836,15 +4950,19 @@ static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
 		    generation <= root->root_key.offset)
 			continue;
 
+		/* We don't lock the tree block, it's OK to be racy here */
+		ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
+					       &refs, &flags);
+		BUG_ON(ret);
+		BUG_ON(refs == 0);
+
 		if (wc->stage == DROP_REFERENCE) {
-			ret = btrfs_lookup_extent_info(trans, root,
-						bytenr, blocksize,
-						&refs, NULL);
-			BUG_ON(ret);
-			BUG_ON(refs == 0);
 			if (refs == 1)
 				goto reada;
 
+			if (wc->level == 1 &&
+			    (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
+				continue;
 			if (!wc->update_ref ||
 			    generation <= root->root_key.offset)
 				continue;
@@ -4853,6 +4971,10 @@ static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
 						  &wc->update_progress);
 			if (ret < 0)
 				continue;
+		} else {
+			if (wc->level == 1 &&
+			    (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
+				continue;
 		}
 reada:
 		ret = readahead_tree_block(root, bytenr, blocksize,
@@ -4876,7 +4998,7 @@ reada:
 static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
 				   struct btrfs_root *root,
 				   struct btrfs_path *path,
-				   struct walk_control *wc)
+				   struct walk_control *wc, int lookup_info)
 {
 	int level = wc->level;
 	struct extent_buffer *eb = path->nodes[level];
@@ -4891,8 +5013,9 @@ static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
 	 * when reference count of tree block is 1, it won't increase
 	 * again. once full backref flag is set, we never clear it.
 	 */
-	if ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
-	    (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag))) {
+	if (lookup_info &&
+	    ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
+	     (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
 		BUG_ON(!path->locks[level]);
 		ret = btrfs_lookup_extent_info(trans, root,
 					       eb->start, eb->len,
@@ -4953,7 +5076,7 @@ static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
 static noinline int do_walk_down(struct btrfs_trans_handle *trans,
 				 struct btrfs_root *root,
 				 struct btrfs_path *path,
-				 struct walk_control *wc)
+				 struct walk_control *wc, int *lookup_info)
 {
 	u64 bytenr;
 	u64 generation;
@@ -4973,8 +5096,10 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
 	 * for the subtree
 	 */
 	if (wc->stage == UPDATE_BACKREF &&
-	    generation <= root->root_key.offset)
+	    generation <= root->root_key.offset) {
+		*lookup_info = 1;
 		return 1;
+	}
 
 	bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]);
 	blocksize = btrfs_level_size(root, level - 1);
@@ -4987,14 +5112,19 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
 	btrfs_tree_lock(next);
 	btrfs_set_lock_blocking(next);
 
-	if (wc->stage == DROP_REFERENCE) {
-		ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
-					       &wc->refs[level - 1],
-					       &wc->flags[level - 1]);
-		BUG_ON(ret);
-		BUG_ON(wc->refs[level - 1] == 0);
+	ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
+				       &wc->refs[level - 1],
+				       &wc->flags[level - 1]);
+	BUG_ON(ret);
+	BUG_ON(wc->refs[level - 1] == 0);
+	*lookup_info = 0;
 
+	if (wc->stage == DROP_REFERENCE) {
 		if (wc->refs[level - 1] > 1) {
+			if (level == 1 &&
+			    (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
+				goto skip;
+
 			if (!wc->update_ref ||
 			    generation <= root->root_key.offset)
 				goto skip;
@@ -5008,12 +5138,17 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
 			wc->stage = UPDATE_BACKREF;
 			wc->shared_level = level - 1;
 		}
+	} else {
+		if (level == 1 &&
+		    (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
+			goto skip;
 	}
 
 	if (!btrfs_buffer_uptodate(next, generation)) {
 		btrfs_tree_unlock(next);
 		free_extent_buffer(next);
 		next = NULL;
+		*lookup_info = 1;
 	}
 
 	if (!next) {
@@ -5036,21 +5171,22 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
 skip:
 	wc->refs[level - 1] = 0;
 	wc->flags[level - 1] = 0;
+	if (wc->stage == DROP_REFERENCE) {
+		if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
+			parent = path->nodes[level]->start;
+		} else {
+			BUG_ON(root->root_key.objectid !=
+			       btrfs_header_owner(path->nodes[level]));
+			parent = 0;
+		}
 
-	if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
-		parent = path->nodes[level]->start;
-	} else {
-		BUG_ON(root->root_key.objectid !=
-		       btrfs_header_owner(path->nodes[level]));
-		parent = 0;
+		ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent,
+					root->root_key.objectid, level - 1, 0);
+		BUG_ON(ret);
 	}
-
-	ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent,
-				root->root_key.objectid, level - 1, 0);
-	BUG_ON(ret);
-
 	btrfs_tree_unlock(next);
 	free_extent_buffer(next);
+	*lookup_info = 1;
 	return 1;
 }
 
@@ -5164,6 +5300,7 @@ static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
 				   struct walk_control *wc)
 {
 	int level = wc->level;
+	int lookup_info = 1;
 	int ret;
 
 	while (level >= 0) {
@@ -5171,14 +5308,14 @@ static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
 		    btrfs_header_nritems(path->nodes[level]))
 			break;
 
-		ret = walk_down_proc(trans, root, path, wc);
+		ret = walk_down_proc(trans, root, path, wc, lookup_info);
 		if (ret > 0)
 			break;
 
 		if (level == 0)
 			break;
 
-		ret = do_walk_down(trans, root, path, wc);
+		ret = do_walk_down(trans, root, path, wc, &lookup_info);
 		if (ret > 0) {
 			path->slots[level]++;
 			continue;
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index de1793ba004a..96577e8bf9fd 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -460,7 +460,8 @@ static int clear_state_bit(struct extent_io_tree *tree,
 			    struct extent_state *state, int bits, int wake,
 			    int delete)
 {
-	int ret = state->state & bits;
+	int bits_to_clear = bits & ~EXTENT_DO_ACCOUNTING;
+	int ret = state->state & bits_to_clear;
 
 	if ((bits & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
 		u64 range = state->end - state->start + 1;
@@ -468,7 +469,7 @@ static int clear_state_bit(struct extent_io_tree *tree,
 		tree->dirty_bytes -= range;
 	}
 	clear_state_cb(tree, state, bits);
-	state->state &= ~bits;
+	state->state &= ~bits_to_clear;
 	if (wake)
 		wake_up(&state->wq);
 	if (delete || state->state == 0) {
@@ -956,7 +957,8 @@ int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
 		       gfp_t mask)
 {
 	return clear_extent_bit(tree, start, end,
-				EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0,
+				EXTENT_DIRTY | EXTENT_DELALLOC |
+				EXTENT_DO_ACCOUNTING, 0, 0,
 				NULL, mask);
 }
 
@@ -1401,12 +1403,7 @@ out_failed:
 int extent_clear_unlock_delalloc(struct inode *inode,
 				struct extent_io_tree *tree,
 				u64 start, u64 end, struct page *locked_page,
-				int unlock_pages,
-				int clear_unlock,
-				int clear_delalloc, int clear_dirty,
-				int set_writeback,
-				int end_writeback,
-				int set_private2)
+				unsigned long op)
 {
 	int ret;
 	struct page *pages[16];
@@ -1416,17 +1413,21 @@ int extent_clear_unlock_delalloc(struct inode *inode,
 	int i;
 	int clear_bits = 0;
 
-	if (clear_unlock)
+	if (op & EXTENT_CLEAR_UNLOCK)
 		clear_bits |= EXTENT_LOCKED;
-	if (clear_dirty)
+	if (op & EXTENT_CLEAR_DIRTY)
 		clear_bits |= EXTENT_DIRTY;
 
-	if (clear_delalloc)
+	if (op & EXTENT_CLEAR_DELALLOC)
 		clear_bits |= EXTENT_DELALLOC;
 
+	if (op & EXTENT_CLEAR_ACCOUNTING)
+		clear_bits |= EXTENT_DO_ACCOUNTING;
+
 	clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
-	if (!(unlock_pages || clear_dirty || set_writeback || end_writeback ||
-	      set_private2))
+	if (!(op & (EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
+		    EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK |
+		    EXTENT_SET_PRIVATE2)))
 		return 0;
 
 	while (nr_pages > 0) {
@@ -1435,20 +1436,20 @@ int extent_clear_unlock_delalloc(struct inode *inode,
 				     nr_pages, ARRAY_SIZE(pages)), pages);
 		for (i = 0; i < ret; i++) {
 
-			if (set_private2)
+			if (op & EXTENT_SET_PRIVATE2)
 				SetPagePrivate2(pages[i]);
 
 			if (pages[i] == locked_page) {
 				page_cache_release(pages[i]);
 				continue;
 			}
-			if (clear_dirty)
+			if (op & EXTENT_CLEAR_DIRTY)
 				clear_page_dirty_for_io(pages[i]);
-			if (set_writeback)
+			if (op & EXTENT_SET_WRITEBACK)
 				set_page_writeback(pages[i]);
-			if (end_writeback)
+			if (op & EXTENT_END_WRITEBACK)
 				end_page_writeback(pages[i]);
-			if (unlock_pages)
+			if (op & EXTENT_CLEAR_UNLOCK_PAGE)
 				unlock_page(pages[i]);
 			page_cache_release(pages[i]);
 		}
@@ -2714,7 +2715,8 @@ int extent_invalidatepage(struct extent_io_tree *tree,
 	lock_extent(tree, start, end, GFP_NOFS);
 	wait_on_page_writeback(page);