From 61d92c328c16419fc96dc50dd16f8b8c695409ec Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Fri, 2 Oct 2009 19:11:56 -0400
Subject: Btrfs: fix deadlock on async thread startup

The btrfs async worker threads are used for a wide variety of things,
including processing bio end_io functions.  This means that when
the endio threads aren't running, the rest of the FS isn't
able to do the final processing required to clear PageWriteback.

The endio threads also try to exit as they become idle and
start more as the work piles up.  The problem is that starting more
threads means kthreadd may need to allocate ram, and that allocation
may wait until the global number of writeback pages on the system is
below a certain limit.

The result of that throttling is that end IO threads wait on
kthreadd, who is waiting on IO to end, which will never happen.

This commit fixes the deadlock by handing off thread startup to a
dedicated thread.  It also fixes a bug where the on-demand thread
creation was creating far too many threads because it didn't take into
account threads being started by other procs.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/async-thread.c | 81 ++++++++++++++++++++++++++++++++++++++++++++-----
 fs/btrfs/async-thread.h | 10 ++++--
 fs/btrfs/ctree.h        |  1 +
 fs/btrfs/disk-io.c      | 42 +++++++++++++------------
 fs/btrfs/relocation.c   |  4 +--
 5 files changed, 106 insertions(+), 32 deletions(-)

diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c
index 282ca085c2f..c0861e781cd 100644
--- a/fs/btrfs/async-thread.c
+++ b/fs/btrfs/async-thread.c
@@ -63,6 +63,51 @@ struct btrfs_worker_thread {
 	int idle;
 };
 
+/*
+ * 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 fc089b95ec1..5077746cf85 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/ctree.h b/fs/btrfs/ctree.h
index 8184f2feb2f..1b920ffc6a5 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -907,6 +907,7 @@ 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;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 69dce50aabd..9903f042765 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1746,21 +1746,22 @@ 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);
 
 	/* 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 +1775,20 @@ 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 +1800,8 @@ 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);
@@ -2012,6 +2014,7 @@ 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);
@@ -2437,6 +2440,7 @@ 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);
diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index 361ad323faa..cfcc93c93a7 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -3518,7 +3518,7 @@ int btrfs_relocate_block_group(struct btrfs_root *extent_root, u64 group_start)
 	BUG_ON(!rc->block_group);
 
 	btrfs_init_workers(&rc->workers, "relocate",
-			   fs_info->thread_pool_size);
+			   fs_info->thread_pool_size, NULL);
 
 	rc->extent_root = extent_root;
 	btrfs_prepare_block_group_relocation(extent_root, rc->block_group);
@@ -3701,7 +3701,7 @@ int btrfs_recover_relocation(struct btrfs_root *root)
 	mapping_tree_init(&rc->reloc_root_tree);
 	INIT_LIST_HEAD(&rc->reloc_roots);
 	btrfs_init_workers(&rc->workers, "relocate",
-			   root->fs_info->thread_pool_size);
+			   root->fs_info->thread_pool_size, NULL);
 	rc->extent_root = root->fs_info->extent_root;
 
 	set_reloc_control(rc);
-- 
cgit v1.2.3


From 1cdda9b81ac0e6ee986f034fa02f221679e1c11a Mon Sep 17 00:00:00 2001
From: Josef Bacik <josef@redhat.com>
Date: Tue, 6 Oct 2009 10:04:28 -0400
Subject: Btrfs: fix possible softlockup in the allocator

Like the cluster allocating stuff, we can lockup the box with the normal
allocation path.  This happens when we

1) Start to cache a block group that is severely fragmented, but has a decent
amount of free space.
2) Start to commit a transaction
3) Have the commit try and empty out some of the delalloc inodes with extents
that are relatively large.

The inodes will not be able to make the allocations because they will ask for
allocations larger than a contiguous area in the free space cache.  So we will
wait for more progress to be made on the block group, but since we're in a
commit the caching kthread won't make any more progress and it already has
enough free space that wait_block_group_cache_progress will just return.  So,
if we wait and fail to make the allocation the next time around, just loop and
go to the next block group.  This keeps us from getting stuck in a softlockup.
Thanks,

Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/extent-tree.c | 23 +++++++++++++++++------
 1 file changed, 17 insertions(+), 6 deletions(-)

diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index d119c0388af..2f82fabd701 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -4028,6 +4028,7 @@ 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);
@@ -4232,14 +4233,23 @@ 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);
@@ -4287,6 +4297,7 @@ checks:
 		break;
 loop:
 		failed_cluster_refill = false;
+		failed_alloc = false;
 		btrfs_put_block_group(block_group);
 	}
 	up_read(&space_info->groups_sem);
-- 
cgit v1.2.3


From a791e35e12ff672e8a0e140abeeaf900c3b2ea77 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Thu, 8 Oct 2009 11:27:10 -0400
Subject: Btrfs: cleanup extent_clear_unlock_delalloc flags

extent_clear_unlock_delalloc has a growing set of ugly parameters
that is very difficult to read and maintain.

This switches to a flag field and well named flag defines.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/extent_io.c | 27 +++++++++++----------------
 fs/btrfs/extent_io.h | 16 ++++++++++------
 fs/btrfs/inode.c     | 45 ++++++++++++++++++++++++++++++---------------
 3 files changed, 51 insertions(+), 37 deletions(-)

diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index de1793ba004..f9708bd0166 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -1401,12 +1401,7 @@ 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 +1411,17 @@ 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;
 
 	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 +1430,20 @@ 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]);
 		}
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index 4794ec891fe..41d2a47ecf3 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -25,6 +25,15 @@
 #define EXTENT_BUFFER_BLOCKING 1
 #define EXTENT_BUFFER_DIRTY 2
 
+/* these are flags for extent_clear_unlock_delalloc */
+#define EXTENT_CLEAR_UNLOCK_PAGE 0x1
+#define EXTENT_CLEAR_UNLOCK	 0x2
+#define EXTENT_CLEAR_DELALLOC	 0x4
+#define EXTENT_CLEAR_DIRTY	 0x8
+#define EXTENT_SET_WRITEBACK	 0x10
+#define EXTENT_END_WRITEBACK	 0x20
+#define EXTENT_SET_PRIVATE2	 0x40
+
 /*
  * page->private values.  Every page that is controlled by the extent
  * map has page->private set to one.
@@ -288,10 +297,5 @@ int extent_range_uptodate(struct extent_io_tree *tree,
 int extent_clear_unlock_delalloc(struct inode *inode,
 				struct extent_io_tree *tree,
 				u64 start, u64 end, struct page *locked_page,
-				int unlock_page,
-				int clear_unlock,
-				int clear_delalloc, int clear_dirty,
-				int set_writeback,
-				int end_writeback,
-				int set_private2);
+				unsigned long op);
 #endif
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 3a6f953337b..a7058fca3aa 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -424,9 +424,10 @@ again:
 			 * and free up our temp pages.
 			 */
 			extent_clear_unlock_delalloc(inode,
-						     &BTRFS_I(inode)->io_tree,
-						     start, end, NULL, 1, 0,
-						     0, 1, 1, 1, 0);
+			     &BTRFS_I(inode)->io_tree,
+			     start, end, NULL,
+			     EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
+			     EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK);
 			ret = 0;
 			goto free_pages_out;
 		}
@@ -637,11 +638,13 @@ static noinline int submit_compressed_extents(struct inode *inode,
 		 * clear dirty, set writeback and unlock the pages.
 		 */
 		extent_clear_unlock_delalloc(inode,
-					     &BTRFS_I(inode)->io_tree,
-					     async_extent->start,
-					     async_extent->start +
-					     async_extent->ram_size - 1,
-					     NULL, 1, 1, 0, 1, 1, 0, 0);
+				&BTRFS_I(inode)->io_tree,
+				async_extent->start,
+				async_extent->start +
+				async_extent->ram_size - 1,
+				NULL, EXTENT_CLEAR_UNLOCK_PAGE |
+				EXTENT_CLEAR_UNLOCK |
+				EXTENT_CLEAR_DIRTY | EXTENT_SET_WRITEBACK);
 
 		ret = btrfs_submit_compressed_write(inode,
 				    async_extent->start,
@@ -712,9 +715,14 @@ static noinline int cow_file_range(struct inode *inode,
 					    start, end, 0, NULL);
 		if (ret == 0) {
 			extent_clear_unlock_delalloc(inode,
-						     &BTRFS_I(inode)->io_tree,
-						     start, end, NULL, 1, 1,
-						     1, 1, 1, 1, 0);
+				     &BTRFS_I(inode)->io_tree,
+				     start, end, NULL,
+				     EXTENT_CLEAR_UNLOCK_PAGE |
+				     EXTENT_CLEAR_UNLOCK |
+				     EXTENT_CLEAR_DELALLOC |
+				     EXTENT_CLEAR_DIRTY |
+				     EXTENT_SET_WRITEBACK |
+				     EXTENT_END_WRITEBACK);
 			*nr_written = *nr_written +
 			     (end - start + PAGE_CACHE_SIZE) / PAGE_CACHE_SIZE;
 			*page_started = 1;
@@ -738,6 +746,8 @@ static noinline int cow_file_range(struct inode *inode,
 	btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);
 
 	while (disk_num_bytes > 0) {
+		unsigned long op;
+
 		cur_alloc_size = min(disk_num_bytes, root->fs_info->max_extent);
 		ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
 					   root->sectorsize, 0, alloc_hint,
@@ -789,10 +799,13 @@ static noinline int cow_file_range(struct inode *inode,
 		 * Do set the Private2 bit so we know this page was properly
 		 * setup for writepage
 		 */
+		op = unlock ? EXTENT_CLEAR_UNLOCK_PAGE : 0;
+		op |= EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
+			EXTENT_SET_PRIVATE2;
+
 		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
 					     start, start + ram_size - 1,
-					     locked_page, unlock, 1,
-					     1, 0, 0, 0, 1);
+					     locked_page, op);
 		disk_num_bytes -= cur_alloc_size;
 		num_bytes -= cur_alloc_size;
 		alloc_hint = ins.objectid + ins.offset;
@@ -1112,8 +1125,10 @@ out_check:
 		BUG_ON(ret);
 
 		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
-					cur_offset, cur_offset + num_bytes - 1,
-					locked_page, 1, 1, 1, 0, 0, 0, 1);
+				cur_offset, cur_offset + num_bytes - 1,
+				locked_page, EXTENT_CLEAR_UNLOCK_PAGE |
+				EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
+				EXTENT_SET_PRIVATE2);
 		cur_offset = extent_end;
 		if (cur_offset > end)
 			break;
-- 
cgit v1.2.3


From a3429ab70b04363c6190964e82f04f44f3e34cf0 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Thu, 8 Oct 2009 12:30:20 -0400
Subject: Btrfs: delay clearing EXTENT_DELALLOC for compressed extents

When compression is on, the cow_file_range code is farmed off to
worker threads.  This allows us to do significant CPU work in parallel
on SMP machines.

But it is a delicate balance around when we clear flags and how.  In
the past we cleared the delalloc flag immediately, which was safe
because the pages stayed locked.

But this is causing problems with the newest ENOSPC code, and with the
recent extent state cleanups we can now clear the delalloc bit at the
same time the uncompressed code does.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/inode.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index a7058fca3aa..401dfb2a94e 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -427,6 +427,7 @@ again:
 			     &BTRFS_I(inode)->io_tree,
 			     start, end, NULL,
 			     EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
+			     EXTENT_CLEAR_DELALLOC |
 			     EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK);
 			ret = 0;
 			goto free_pages_out;
@@ -644,6 +645,7 @@ static noinline int submit_compressed_extents(struct inode *inode,
 				async_extent->ram_size - 1,
 				NULL, EXTENT_CLEAR_UNLOCK_PAGE |
 				EXTENT_CLEAR_UNLOCK |
+				EXTENT_CLEAR_DELALLOC |
 				EXTENT_CLEAR_DIRTY | EXTENT_SET_WRITEBACK);
 
 		ret = btrfs_submit_compressed_write(inode,
@@ -877,8 +879,8 @@ static int cow_file_range_async(struct inode *inode, struct page *locked_page,
 	u64 cur_end;
 	int limit = 10 * 1024 * 1042;
 
-	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED |
-			 EXTENT_DELALLOC, 1, 0, NULL, GFP_NOFS);
+	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED,
+			 1, 0, NULL, GFP_NOFS);
 	while (start < end) {
 		async_cow = kmalloc(sizeof(*async_cow), GFP_NOFS);
 		async_cow->inode = inode;
-- 
cgit v1.2.3


From 32c00aff718bb54a214b39146bdd9ac01511cd25 Mon Sep 17 00:00:00 2001
From: Josef Bacik <josef@redhat.com>
Date: Thu, 8 Oct 2009 13:34:05 -0400
Subject: Btrfs: release delalloc reservations on extent item insertion

This patch fixes an issue with the delalloc metadata space reservation
code.  The problem is we used to free the reservation as soon as we
allocated the delalloc region.  The problem with this is if we are not
inserting an inline extent, we don't actually insert the extent item until
after the ordered extent is written out.  This patch does 3 things,

1) It moves the reservation clearing stuff into the ordered code, so when
we remove the ordered extent we remove the reservation.
2) It adds a EXTENT_DO_ACCOUNTING flag that gets passed when we clear
delalloc bits in the cases where we want to clear the metadata reservation
when we clear the delalloc extent, in the case that we do an inline extent
or we invalidate the page.
3) It adds another waitqueue to the space info so that when we start a fs
wide delalloc flush, anybody else who also hits that area will simply wait
for the flush to finish and then try to make their allocation.

This has been tested thoroughly to make sure we did not regress on
performance.

Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/btrfs_inode.h  | 12 ++++++-----
 fs/btrfs/ctree.h        |  3 +++
 fs/btrfs/extent-tree.c  | 54 +++++++++++++++++++++++++++++++++++++++----------
 fs/btrfs/extent_io.c    | 19 +++++++++++------
 fs/btrfs/extent_io.h    |  2 ++
 fs/btrfs/file.c         |  3 ++-
 fs/btrfs/inode.c        | 45 ++++++++++++++++++++++++++++-------------
 fs/btrfs/ordered-data.c |  6 ++++++
 8 files changed, 107 insertions(+), 37 deletions(-)

diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index a54d354cefc..c71abec0ab9 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 1b920ffc6a5..dbdada56950 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -699,6 +699,9 @@ struct btrfs_space_info {
 
 	int allocating_chunk;
 	wait_queue_head_t wait;
+
+	int flushing;
+	wait_queue_head_t flush_wait;
 };
 
 /*
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 2f82fabd701..3d1be0b77f8 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -2823,14 +2823,17 @@ 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;
@@ -2863,6 +2866,37 @@ static void check_force_delalloc(struct btrfs_space_info *meta_sinfo)
 		meta_sinfo->force_delalloc = 0;
 }
 
+static void flush_delalloc(struct btrfs_root *root,
+				 struct btrfs_space_info *info)
+{
+	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_event(info->flush_wait,
+			   !info->flushing);
+		return;
+	}
+
+	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)
 {
@@ -2980,21 +3014,20 @@ 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);
 
@@ -3093,8 +3126,7 @@ 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);
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index f9708bd0166..96577e8bf9f 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);
 }
 
@@ -1419,9 +1421,13 @@ int extent_clear_unlock_delalloc(struct inode *inode,
 	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 (!(op & (EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY | EXTENT_SET_WRITEBACK |
-		    EXTENT_END_WRITEBACK | EXTENT_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) {
@@ -2709,7 +2715,8 @@ int extent_invalidatepage(struct extent_io_tree *tree,
 	lock_extent(tree, start, end, GFP_NOFS);
 	wait_on_page_writeback(page);
 	clear_extent_bit(tree, start, end,
-			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC,
+			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
+			 EXTENT_DO_ACCOUNTING,
 			 1, 1, NULL, GFP_NOFS);
 	return 0;
 }
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index 41d2a47ecf3..36de250a7b2 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -15,6 +15,7 @@
 #define EXTENT_BUFFER_FILLED (1 << 8)
 #define EXTENT_BOUNDARY (1 << 9)
 #define EXTENT_NODATASUM (1 << 10)
+#define EXTENT_DO_ACCOUNTING (1 << 11)
 #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
 
 /* flags for bio submission */
@@ -33,6 +34,7 @@
 #define EXTENT_SET_WRITEBACK	 0x10
 #define EXTENT_END_WRITEBACK	 0x20
 #define EXTENT_SET_PRIVATE2	 0x40
+#define EXTENT_CLEAR_ACCOUNTING  0x80
 
 /*
  * page->private values.  Every page that is controlled by the extent
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index f155179877a..53fb1c997f0 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -878,7 +878,8 @@ again:
 			btrfs_put_ordered_extent(ordered);
 
 		clear_extent_bits(&BTRFS_I(inode)->io_tree, start_pos,
-				  last_pos - 1, EXTENT_DIRTY | EXTENT_DELALLOC,
+				  last_pos - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
+				  EXTENT_DO_ACCOUNTING,
 				  GFP_NOFS);
 		unlock_extent(&BTRFS_I(inode)->io_tree,
 			      start_pos, last_pos - 1, GFP_NOFS);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 401dfb2a94e..ccc4f112121 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -428,6 +428,7 @@ again:
 			     start, end, NULL,
 			     EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
 			     EXTENT_CLEAR_DELALLOC |
+			     EXTENT_CLEAR_ACCOUNTING |
 			     EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK);
 			ret = 0;
 			goto free_pages_out;
@@ -722,6 +723,7 @@ static noinline int cow_file_range(struct inode *inode,
 				     EXTENT_CLEAR_UNLOCK_PAGE |
 				     EXTENT_CLEAR_UNLOCK |
 				     EXTENT_CLEAR_DELALLOC |
+				     EXTENT_CLEAR_ACCOUNTING |
 				     EXTENT_CLEAR_DIRTY |
 				     EXTENT_SET_WRITEBACK |
 				     EXTENT_END_WRITEBACK);
@@ -1195,15 +1197,17 @@ static int btrfs_split_extent_hook(struct inode *inode,
 					root->fs_info->max_extent);
 
 		/*
-		 * if we break a large extent up then leave delalloc_extents be,
-		 * since we've already accounted for the large extent.
+		 * if we break a large extent up then leave oustanding_extents
+		 * be, since we've already accounted for the large extent.
 		 */
 		if (div64_u64(new_size + root->fs_info->max_extent - 1,
 			      root->fs_info->max_extent) < num_extents)
 			return 0;
 	}
 
-	BTRFS_I(inode)->delalloc_extents++;
+	spin_lock(&BTRFS_I(inode)->accounting_lock);
+	BTRFS_I(inode)->outstanding_extents++;
+	spin_unlock(&BTRFS_I(inode)->accounting_lock);
 
 	return 0;
 }
@@ -1234,7 +1238,9 @@ static int btrfs_merge_extent_hook(struct inode *inode,
 
 	/* we're not bigger than the max, unreserve the space and go */
 	if (new_size <= root->fs_info->max_extent) {
-		BTRFS_I(inode)->delalloc_extents--;
+		spin_lock(&BTRFS_I(inode)->accounting_lock);
+		BTRFS_I(inode)->outstanding_extents--;
+		spin_unlock(&BTRFS_I(inode)->accounting_lock);
 		return 0;
 	}
 
@@ -1248,7 +1254,9 @@ static int btrfs_merge_extent_hook(struct inode *inode,
 		      root->fs_info->max_extent) > num_extents)
 		return 0;
 
-	BTRFS_I(inode)->delalloc_extents--;
+	spin_lock(&BTRFS_I(inode)->accounting_lock);
+	BTRFS_I(inode)->outstanding_extents--;
+	spin_unlock(&BTRFS_I(inode)->accounting_lock);
 
 	return 0;
 }
@@ -1270,7 +1278,9 @@ static int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
 	if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
 		struct btrfs_root *root = BTRFS_I(inode)->root;
 
-		BTRFS_I(inode)->delalloc_extents++;
+		spin_lock(&BTRFS_I(inode)->accounting_lock);
+		BTRFS_I(inode)->outstanding_extents++;
+		spin_unlock(&BTRFS_I(inode)->accounting_lock);
 		btrfs_delalloc_reserve_space(root, inode, end - start + 1);
 		spin_lock(&root->fs_info->delalloc_lock);
 		BTRFS_I(inode)->delalloc_bytes += end - start + 1;
@@ -1298,8 +1308,12 @@ static int btrfs_clear_bit_hook(struct inode *inode,
 	if ((state->state & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
 		struct btrfs_root *root = BTRFS_I(inode)->root;
 
-		BTRFS_I(inode)->delalloc_extents--;
-		btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
+		if (bits & EXTENT_DO_ACCOUNTING) {
+			spin_lock(&BTRFS_I(inode)->accounting_lock);
+			BTRFS_I(inode)->outstanding_extents--;
+			spin_unlock(&BTRFS_I(inode)->accounting_lock);
+			btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
+		}
 
 		spin_lock(&root->fs_info->delalloc_lock);
 		if (state->end - state->start + 1 >
@@ -4825,7 +4839,8 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset)
 		 */
 		clear_extent_bit(tree, page_start, page_end,
 				 EXTENT_DIRTY | EXTENT_DELALLOC |
-				 EXTENT_LOCKED, 1, 0, NULL, GFP_NOFS);
+				 EXTENT_LOCKED | EXTENT_DO_ACCOUNTING, 1, 0,
+				 NULL, GFP_NOFS);
 		/*
 		 * whoever cleared the private bit is responsible
 		 * for the finish_ordered_io
@@ -4838,8 +4853,8 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset)
 		lock_extent(tree, page_start, page_end, GFP_NOFS);
 	}
 	clear_extent_bit(tree, page_start, page_end,
-		 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC,
-		 1, 1, NULL, GFP_NOFS);
+		 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
+		 EXTENT_DO_ACCOUNTING, 1, 1, NULL, GFP_NOFS);
 	__btrfs_releasepage(page, GFP_NOFS);
 
 	ClearPageChecked(page);
@@ -4934,7 +4949,8 @@ again:
 	 * prepare_pages in the normal write path.
 	 */
 	clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start, page_end,
-			  EXTENT_DIRTY | EXTENT_DELALLOC, GFP_NOFS);
+			  EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING,
+			  GFP_NOFS);
 
 	ret = btrfs_set_extent_delalloc(inode, page_start, page_end);
 	if (ret) {
@@ -5082,8 +5098,9 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
 		return NULL;
 	ei->last_trans = 0;
 	ei->logged_trans = 0;
-	ei->delalloc_extents = 0;
-	ei->delalloc_reserved_extents = 0;
+	ei->outstanding_extents = 0;
+	ei->reserved_extents = 0;
+	spin_lock_init(&ei->accounting_lock);
 	btrfs_ordered_inode_tree_init(&ei->ordered_tree);
 	INIT_LIST_HEAD(&ei->i_orphan);
 	INIT_LIST_HEAD(&ei->ordered_operations);
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
index 4a9c8c4cec2..ab21c29f224 100644
--- a/fs/btrfs/ordered-data.c
+++ b/fs/btrfs/ordered-data.c
@@ -306,6 +306,12 @@ int btrfs_remove_ordered_extent(struct inode *inode,
 	tree->last = NULL;
 	set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags);
 
+	spin_lock(&BTRFS_I(inode)->accounting_lock);
+	BTRFS_I(inode)->outstanding_extents--;
+	spin_unlock(&BTRFS_I(inode)->accounting_lock);
+	btrfs_unreserve_metadata_for_delalloc(BTRFS_I(inode)->root,
+					      inode, 1);
+
 	spin_lock(&BTRFS_I(inode)->root->fs_info->ordered_extent_lock);
 	list_del_init(&entry->root_extent_list);
 
-- 
cgit v1.2.3


From e3ccfa989752c083ceb23c823a84f7ce3a081e61 Mon Sep 17 00:00:00 2001
From: Josef Bacik <josef@redhat.com>
Date: Wed, 7 Oct 2009 20:44:34 -0400
Subject: Btrfs: async delalloc flushing under space pressure

This patch moves the delalloc flushing that occurs when we are under space
pressure off to a async thread pool.  This helps since we only free up
metadata space when we actually insert the extent item, which means it takes
quite a while for space to be free'ed up if we wait on all ordered extents.
However, if space is freed up due to inline extents being inserted, we can
wake people who are waiting up early, and they can finish their work.

Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.h       | 13 ++++----
 fs/btrfs/disk-io.c     |  6 ++++
 fs/btrfs/extent-tree.c | 80 ++++++++++++++++++++++++++++++++++++++++++++++----
 3 files changed, 88 insertions(+), 11 deletions(-)

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index dbdada56950..a362dd617e9 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -691,17 +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;
-
-	int flushing;
-	wait_queue_head_t flush_wait;
 };
 
 /*
@@ -918,6 +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
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 9903f042765..ac8927bdc33 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1762,6 +1762,9 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 			   min_t(u64, fs_devices->num_devices,
 			   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
@@ -1809,6 +1812,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	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,
@@ -2023,6 +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);
@@ -2449,6 +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 3d1be0b77f8..53026806ae9 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -2866,9 +2866,66 @@ 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);
@@ -2883,11 +2940,24 @@ static void flush_delalloc(struct btrfs_root *root,
 	spin_unlock(&info->lock);
 
 	if (wait) {
-		wait_event(info->flush_wait,
-			   !info->flushing);
+		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);
 
@@ -2927,7 +2997,7 @@ 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;
 	}
@@ -2935,7 +3005,7 @@ 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;
 	}
@@ -2956,7 +3026,7 @@ 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;
-- 
cgit v1.2.3


From ff782e0a131c7f669445c07fe5c7ba91e043b7ed Mon Sep 17 00:00:00 2001
From: Josef Bacik <josef@redhat.com>
Date: Thu, 8 Oct 2009 15:30:04 -0400
Subject: Btrfs: optimize fsync for the single writer case

This patch optimizes the tree logging stuff so it doesn't always wait 1 jiffie
for new people to join the logging transaction if there is only ever 1 writer.
This helps a little bit with latency where we have something like RPM where it
will fdatasync every file it writes, and so waiting the 1 jiffie for every
fdatasync really starts to add up.

Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.h    |  2 ++
 fs/btrfs/tree-log.c | 12 +++++++++++-
 2 files changed, 13 insertions(+), 1 deletion(-)

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index a362dd617e9..16ddb19dda8 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1010,6 +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;
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 4d7d9abef42..78f6254ac2d 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -137,11 +137,20 @@ static int start_log_trans(struct btrfs_trans_handle *trans,
 
 	mutex_lock(&root->log_mutex);
 	if (root->log_root) {
+		if (!root->log_start_pid) {
+			root->log_start_pid = current->pid;
+			root->log_multiple_pids = false;
+		} else if (root->log_start_pid != current->pid) {
+			root->log_multiple_pids = true;
+		}
+
 		root->log_batch++;
 		atomic_inc(&root->log_writers);
 		mutex_unlock(&root->log_mutex);
 		return 0;
 	}
+	root->log_multiple_pids = false;
+	root->log_start_pid = current->pid;
 	mutex_lock(&root->fs_info->tree_log_mutex);
 	if (!root->fs_info->log_root_tree) {
 		ret = btrfs_init_log_root_tree(trans, root->fs_info);
@@ -1985,7 +1994,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	if (atomic_read(&root->log_commit[(index1 + 1) % 2]))
 		wait_log_commit(trans, root, root->log_transid - 1);
 
-	while (1) {
+	while (root->log_multiple_pids) {
 		unsigned long batch = root->log_batch;
 		mutex_unlock(&root->log_mutex);
 		schedule_timeout_uninterruptible(1);
@@ -2011,6 +2020,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	root->log_batch = 0;
 	root->log_transid++;
 	log->log_transid = root->log_transid;
+	root->log_start_pid = 0;
 	smp_mb();
 	/*
 	 * log tree has been flushed to disk, new modifications of
-- 
cgit v1.2.3


From efefb1438be269897585934fc6c05deb4dfa01ce Mon Sep 17 00:00:00 2001
From: "Yan, Zheng" <zheng.yan@oracle.com>
Date: Fri, 9 Oct 2009 09:25:16 -0400
Subject: Btrfs: remove negative dentry when deleting subvolumne

The use of btrfs_dentry_delete is removing dentries from the
dcache when deleting subvolumne. btrfs_dentry_delete ignores
negative dentries. This is incorrect since if we don't remove
the negative dentry, its parent dentry can't be removed.

Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/inode.c | 12 +++++++-----
 fs/btrfs/ioctl.c |  1 +
 2 files changed, 8 insertions(+), 5 deletions(-)

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index ccc4f112121..d3585cb6483 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -3629,12 +3629,14 @@ static int btrfs_dentry_delete(struct dentry *dentry)
 {
 	struct btrfs_root *root;
 
-	if (!dentry->d_inode)
-		return 0;
+	if (!dentry->d_inode && !IS_ROOT(dentry))
+		dentry = dentry->d_parent;
 
-	root = BTRFS_I(dentry->d_inode)->root;
-	if (btrfs_root_refs(&root->root_item) == 0)
-		return 1;
+	if (dentry->d_inode) {
+		root = BTRFS_I(dentry->d_inode)->root;
+		if (btrfs_root_refs(&root->root_item) == 0)
+			return 1;
+	}
 	return 0;
 }
 
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 9a780c8d0ac..e8795becad4 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -830,6 +830,7 @@ out_up_write:
 out_unlock:
 	mutex_unlock(&inode->i_mutex);
 	if (!err) {
+		shrink_dcache_sb(root->fs_info->sb);
 		btrfs_invalidate_inodes(dest);
 		d_delete(dentry);
 	}
-- 
cgit v1.2.3


From 94fcca9f8999e7828d5f4dc181daa39cad2af38a Mon Sep 17 00:00:00 2001
From: "Yan, Zheng" <zheng.yan@oracle.com>
Date: Fri, 9 Oct 2009 09:25:16 -0400
Subject: Btrfs: optimize back reference update during btrfs_drop_snapshot

This patch reading level 0 tree blocks that already use full backrefs.

Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/extent-tree.c | 82 ++++++++++++++++++++++++++++++++------------------
 1 file changed, 53 insertions(+), 29 deletions(-)

diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 53026806ae9..4aedbff36b8 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -4911,6 +4911,7 @@ 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;
@@ -4948,15 +4949,19 @@ 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;
@@ -4965,6 +4970,10 @@ 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,
@@ -4988,7 +4997,7 @@ 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];
@@ -5003,8 +5012,9 @@ 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,
@@ -5065,7 +5075,7 @@ 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;
@@ -5085,8 +5095,10 @@ 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);
@@ -5099,14 +5111,19 @@ 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;
@@ -5120,12 +5137,17 @@ 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) {
@@ -5148,21 +5170,22 @@ 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;
 }
 
@@ -5276,6 +5299,7 @@ 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) {
@@ -5283,14 +5307,14 @@ 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;
-- 
cgit v1.2.3


From 82d339d9b3a6395f17d3253887653250b693b74b Mon Sep 17 00:00:00 2001
From: Alexey Dobriyan <adobriyan@gmail.com>
Date: Fri, 9 Oct 2009 09:54:36 -0400
Subject: Btrfs: constify dentry_operations

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.h | 2 +-
 fs/btrfs/inode.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 16ddb19dda8..36a19cd43e0 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -2330,7 +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/inode.c b/fs/btrfs/inode.c
index d3585cb6483..d960492d3d8 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -5840,6 +5840,6 @@ static struct inode_operations btrfs_symlink_inode_operations = {
 	.removexattr	= btrfs_removexattr,
 };
 
-struct dentry_operations btrfs_dentry_operations = {
+const struct dentry_operations btrfs_dentry_operations = {
 	.d_delete	= btrfs_dentry_delete,
 };
-- 
cgit v1.2.3


From e9061e214810c9534381a705a1b46533e09f2676 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Fri, 9 Oct 2009 09:57:45 -0400
Subject: Btrfs: fix uninit compiler warning in cow_file_range_nocow

The extent_type variable was exposed uninit via a goto.  It should be
impossible to trigger because it is protected by a check on another
variable, but this makes sure.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/inode.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index d960492d3d8..ef399a7794f 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -1023,6 +1023,7 @@ next_slot:
 
 		if (found_key.offset > cur_offset) {
 			extent_end = found_key.offset;
+			extent_type = 0;
 			goto out_check;
 		}
 
-- 
cgit v1.2.3


From ac6889cbb254be1ffea376bea4a96ce9be0e0ed0 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Fri, 9 Oct 2009 11:29:53 -0400
Subject: Btrfs: fix file clone ioctl for bookend extents

The file clone ioctl was incorrectly taking the offset into the
extent on disk into account when calculating the length of the
cloned extent.

The length never changes based on the offset into the physical extent.

Test case:

fallocate -l 1g image
mke2fs image
bcp image image2
e2fsck -f image2

(errors on image2)

The math bug ends up wrapping the length of the extent, and things
go wrong from there.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ioctl.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index e8795becad4..cdbb054102b 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -1123,8 +1123,10 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
 					datao += off - key.offset;
 					datal -= off - key.offset;
 				}
-				if (key.offset + datao + datal > off + len)
-					datal = off + len - key.offset - datao;
+
+				if (key.offset + datal > off + len)
+					datal = off + len - key.offset;
+
 				/* disko == 0 means it's a hole */
 				if (!disko)
 					datao = 0;
-- 
cgit v1.2.3