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: BUG to BUG_ON changes Btrfs: remove dead code Btrfs: remove dead code Btrfs: fix typos in comments Btrfs: remove unused ftrace include Btrfs: fix __ucmpdi2 compile bug on 32 bit builds Btrfs: free inode struct when btrfs_new_inode fails Btrfs: fix race in worker_loop Btrfs: add flushoncommit mount option Btrfs: notreelog mount option Btrfs: introduce btrfs_show_options Btrfs: rework allocation clustering Btrfs: Optimize locking in btrfs_next_leaf() Btrfs: break up btrfs_search_slot into smaller pieces Btrfs: kill the pinned_mutex Btrfs: kill the block group alloc mutex Btrfs: clean up find_free_extent Btrfs: free space cache cleanups Btrfs: unplug in the async bio submission threads Btrfs: keep processing bios for a given bdev if our proc is batching
2026-05-01 15:00:59 -07:00 · 2009-04-03 15:14:44 -07:00
parent 5a3ae27605 c293498be6
commit b983471794
17 changed files with 994 additions and 556 deletions
@@ -20,7 +20,6 @@
 #include <linux/list.h>
 #include <linux/spinlock.h>
 #include <linux/freezer.h>
-#include <linux/ftrace.h>
 #include "async-thread.h"

 #define WORK_QUEUED_BIT 0
@@ -195,6 +194,9 @@ again_locked:
 				if (!list_empty(&worker->pending))
 					continue;

+				if (kthread_should_stop())
+					break;
+
 				/* still no more work?, sleep for real */
 				spin_lock_irq(&worker->lock);
 				set_current_state(TASK_INTERRUPTIBLE);
@@ -208,7 +210,8 @@ again_locked:
 				worker->working = 0;
 				spin_unlock_irq(&worker->lock);

-				schedule();
+				if (!kthread_should_stop())
+					schedule();
 			}
 			__set_current_state(TASK_RUNNING);
 		}
@@ -1244,9 +1244,9 @@ static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 * readahead one full node of leaves, finding things that are close
 * to the block in 'slot', and triggering ra on them.
 */
-static noinline void reada_for_search(struct btrfs_root *root,
-				      struct btrfs_path *path,
-				      int level, int slot, u64 objectid)
+static void reada_for_search(struct btrfs_root *root,
+			     struct btrfs_path *path,
+			     int level, int slot, u64 objectid)
 {
 	struct extent_buffer *node;
 	struct btrfs_disk_key disk_key;
@@ -1446,6 +1446,117 @@ noinline void btrfs_unlock_up_safe(struct btrfs_path *path, int level)
 	}
 }

+/*
+ * helper function for btrfs_search_slot.  The goal is to find a block
+ * in cache without setting the path to blocking.  If we find the block
+ * we return zero and the path is unchanged.
+ *
+ * If we can't find the block, we set the path blocking and do some
+ * reada.  -EAGAIN is returned and the search must be repeated.
+ */
+static int
+read_block_for_search(struct btrfs_trans_handle *trans,
+		       struct btrfs_root *root, struct btrfs_path *p,
+		       struct extent_buffer **eb_ret, int level, int slot,
+		       struct btrfs_key *key)
+{
+	u64 blocknr;
+	u64 gen;
+	u32 blocksize;
+	struct extent_buffer *b = *eb_ret;
+	struct extent_buffer *tmp;
+
+	blocknr = btrfs_node_blockptr(b, slot);
+	gen = btrfs_node_ptr_generation(b, slot);
+	blocksize = btrfs_level_size(root, level - 1);
+
+	tmp = btrfs_find_tree_block(root, blocknr, blocksize);
+	if (tmp && btrfs_buffer_uptodate(tmp, gen)) {
+		*eb_ret = tmp;
+		return 0;
+	}
+
+	/*
+	 * reduce lock contention at high levels
+	 * of the btree by dropping locks before
+	 * we read.
+	 */
+	btrfs_release_path(NULL, p);
+	if (tmp)
+		free_extent_buffer(tmp);
+	if (p->reada)
+		reada_for_search(root, p, level, slot, key->objectid);
+
+	tmp = read_tree_block(root, blocknr, blocksize, gen);
+	if (tmp)
+		free_extent_buffer(tmp);
+	return -EAGAIN;
+}
+
+/*
+ * helper function for btrfs_search_slot.  This does all of the checks
+ * for node-level blocks and does any balancing required based on
+ * the ins_len.
+ *
+ * If no extra work was required, zero is returned.  If we had to
+ * drop the path, -EAGAIN is returned and btrfs_search_slot must
+ * start over
+ */
+static int
+setup_nodes_for_search(struct btrfs_trans_handle *trans,
+		       struct btrfs_root *root, struct btrfs_path *p,
+		       struct extent_buffer *b, int level, int ins_len)
+{
+	int ret;
+	if ((p->search_for_split || ins_len > 0) && btrfs_header_nritems(b) >=
+	    BTRFS_NODEPTRS_PER_BLOCK(root) - 3) {
+		int sret;
+
+		sret = reada_for_balance(root, p, level);
+		if (sret)
+			goto again;
+
+		btrfs_set_path_blocking(p);
+		sret = split_node(trans, root, p, level);
+		btrfs_clear_path_blocking(p, NULL);
+
+		BUG_ON(sret > 0);
+		if (sret) {
+			ret = sret;
+			goto done;
+		}
+		b = p->nodes[level];
+	} else if (ins_len < 0 && btrfs_header_nritems(b) <
+		   BTRFS_NODEPTRS_PER_BLOCK(root) / 4) {
+		int sret;
+
+		sret = reada_for_balance(root, p, level);
+		if (sret)
+			goto again;
+
+		btrfs_set_path_blocking(p);
+		sret = balance_level(trans, root, p, level);
+		btrfs_clear_path_blocking(p, NULL);
+
+		if (sret) {
+			ret = sret;
+			goto done;
+		}
+		b = p->nodes[level];
+		if (!b) {
+			btrfs_release_path(NULL, p);
+			goto again;
+		}
+		BUG_ON(btrfs_header_nritems(b) == 1);
+	}
+	return 0;
+
+again:
+	ret = -EAGAIN;
+done:
+	return ret;
+}
+
 /*
 * look for key in the tree.  path is filled in with nodes along the way
 * if key is found, we return zero and you can find the item in the leaf
@@ -1464,16 +1575,11 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
 		      ins_len, int cow)
 {
 	struct extent_buffer *b;
-	struct extent_buffer *tmp;
 	int slot;
 	int ret;
 	int level;
-	int should_reada = p->reada;
 	int lowest_unlock = 1;
-	int blocksize;
 	u8 lowest_level = 0;
-	u64 blocknr;
-	u64 gen;

 	lowest_level = p->lowest_level;
 	WARN_ON(lowest_level && ins_len > 0);
@@ -1502,7 +1608,11 @@ again:
 		if (cow) {
 			int wret;

-			/* is a cow on this block not required */
+			/*
+			 * if we don't really need to cow this block
+			 * then we don't want to set the path blocking,
+			 * so we test it here
+			 */
 			if (btrfs_header_generation(b) == trans->transid &&
 			    btrfs_header_owner(b) == root->root_key.objectid &&
 			    !btrfs_header_flag(b, BTRFS_HEADER_FLAG_WRITTEN)) {
@@ -1557,51 +1667,15 @@ cow_done:
 			if (ret && slot > 0)
 				slot -= 1;
 			p->slots[level] = slot;
-			if ((p->search_for_split || ins_len > 0) &&
-			    btrfs_header_nritems(b) >=
-			    BTRFS_NODEPTRS_PER_BLOCK(root) - 3) {
-				int sret;
+			ret = setup_nodes_for_search(trans, root, p, b, level,
+						     ins_len);
+			if (ret == -EAGAIN)
+				goto again;
+			else if (ret)
+				goto done;
+			b = p->nodes[level];
+			slot = p->slots[level];

-				sret = reada_for_balance(root, p, level);
-				if (sret)
-					goto again;
-
-				btrfs_set_path_blocking(p);
-				sret = split_node(trans, root, p, level);
-				btrfs_clear_path_blocking(p, NULL);
-
-				BUG_ON(sret > 0);
-				if (sret) {
-					ret = sret;
-					goto done;
-				}
-				b = p->nodes[level];
-				slot = p->slots[level];
-			} else if (ins_len < 0 &&
-				   btrfs_header_nritems(b) <
-				   BTRFS_NODEPTRS_PER_BLOCK(root) / 4) {
-				int sret;
-
-				sret = reada_for_balance(root, p, level);
-				if (sret)
-					goto again;
-
-				btrfs_set_path_blocking(p);
-				sret = balance_level(trans, root, p, level);
-				btrfs_clear_path_blocking(p, NULL);
-
-				if (sret) {
-					ret = sret;
-					goto done;
-				}
-				b = p->nodes[level];
-				if (!b) {
-					btrfs_release_path(NULL, p);
-					goto again;
-				}
-				slot = p->slots[level];
-				BUG_ON(btrfs_header_nritems(b) == 1);
-			}
 			unlock_up(p, level, lowest_unlock);

 			/* this is only true while dropping a snapshot */
@@ -1610,44 +1684,11 @@ cow_done:
 				goto done;
 			}

-			blocknr = btrfs_node_blockptr(b, slot);
-			gen = btrfs_node_ptr_generation(b, slot);
-			blocksize = btrfs_level_size(root, level - 1);
+			ret = read_block_for_search(trans, root, p,
+						    &b, level, slot, key);
+			if (ret == -EAGAIN)
+				goto again;

-			tmp = btrfs_find_tree_block(root, blocknr, blocksize);
-			if (tmp && btrfs_buffer_uptodate(tmp, gen)) {
-				b = tmp;
-			} else {
-				/*
-				 * reduce lock contention at high levels
-				 * of the btree by dropping locks before
-				 * we read.
-				 */
-				if (level > 0) {
-					btrfs_release_path(NULL, p);
-					if (tmp)
-						free_extent_buffer(tmp);
-					if (should_reada)
-						reada_for_search(root, p,
-								 level, slot,
-								 key->objectid);
-
-					tmp = read_tree_block(root, blocknr,
-							 blocksize, gen);
-					if (tmp)
-						free_extent_buffer(tmp);
-					goto again;
-				} else {
-					btrfs_set_path_blocking(p);
-					if (tmp)
-						free_extent_buffer(tmp);
-					if (should_reada)
-						reada_for_search(root, p,
-								 level, slot,
-								 key->objectid);
-					b = read_node_slot(root, b, slot);
-				}
-			}
 			if (!p->skip_locking) {
 				int lret;

@@ -2116,8 +2157,7 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
 	BUG_ON(!path->nodes[level]);
 	lower = path->nodes[level];
 	nritems = btrfs_header_nritems(lower);
-	if (slot > nritems)
-		BUG();
+	BUG_ON(slot > nritems);
 	if (nritems == BTRFS_NODEPTRS_PER_BLOCK(root))
 		BUG();
 	if (slot != nritems) {
@@ -4086,28 +4126,44 @@ next:
 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 {
 	int slot;
-	int level = 1;
+	int level;
 	struct extent_buffer *c;
-	struct extent_buffer *next = NULL;
+	struct extent_buffer *next;
 	struct btrfs_key key;
 	u32 nritems;
 	int ret;
+	int old_spinning = path->leave_spinning;
+	int force_blocking = 0;

 	nritems = btrfs_header_nritems(path->nodes[0]);
 	if (nritems == 0)
 		return 1;

-	btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1);
+	/*
+	 * we take the blocks in an order that upsets lockdep.  Using
+	 * blocking mode is the only way around it.
+	 */
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+	force_blocking = 1;
+#endif

+	btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1);
+again:
+	level = 1;
+	next = NULL;
 	btrfs_release_path(root, path);
+
 	path->keep_locks = 1;
+
+	if (!force_blocking)
+		path->leave_spinning = 1;
+
 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 	path->keep_locks = 0;

 	if (ret < 0)
 		return ret;

-	btrfs_set_path_blocking(path);
 	nritems = btrfs_header_nritems(path->nodes[0]);
 	/*
 	 * by releasing the path above we dropped all our locks.  A balance
@@ -4117,19 +4173,24 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 	 */
 	if (nritems > 0 && path->slots[0] < nritems - 1) {
 		path->slots[0]++;
+		ret = 0;
 		goto done;
 	}

 	while (level < BTRFS_MAX_LEVEL) {
-		if (!path->nodes[level])
-			return 1;
+		if (!path->nodes[level]) {
+			ret = 1;
+			goto done;
+		}

 		slot = path->slots[level] + 1;
 		c = path->nodes[level];
 		if (slot >= btrfs_header_nritems(c)) {
 			level++;
-			if (level == BTRFS_MAX_LEVEL)
-				return 1;
+			if (level == BTRFS_MAX_LEVEL) {
+				ret = 1;
+				goto done;
+			}
 			continue;
 		}

@@ -4138,16 +4199,22 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 			free_extent_buffer(next);
 		}

-		/* the path was set to blocking above */
-		if (level == 1 && (path->locks[1] || path->skip_locking) &&
-		    path->reada)
-			reada_for_search(root, path, level, slot, 0);
+		next = c;
+		ret = read_block_for_search(NULL, root, path, &next, level,
+					    slot, &key);
+		if (ret == -EAGAIN)
+			goto again;

-		next = read_node_slot(root, c, slot);
 		if (!path->skip_locking) {
-			btrfs_assert_tree_locked(c);
-			btrfs_tree_lock(next);
-			btrfs_set_lock_blocking(next);
+			ret = btrfs_try_spin_lock(next);
+			if (!ret) {
+				btrfs_set_path_blocking(path);
+				btrfs_tree_lock(next);
+				if (!force_blocking)
+					btrfs_clear_path_blocking(path, next);
+			}
+			if (force_blocking)
+				btrfs_set_lock_blocking(next);
 		}
 		break;
 	}
@@ -4157,27 +4224,42 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 		c = path->nodes[level];
 		if (path->locks[level])
 			btrfs_tree_unlock(c);
+
 		free_extent_buffer(c);
 		path->nodes[level] = next;
 		path->slots[level] = 0;
 		if (!path->skip_locking)
 			path->locks[level] = 1;
+
 		if (!level)
 			break;

-		btrfs_set_path_blocking(path);
-		if (level == 1 && path->locks[1] && path->reada)
-			reada_for_search(root, path, level, slot, 0);
-		next = read_node_slot(root, next, 0);
+		ret = read_block_for_search(NULL, root, path, &next, level,
+					    0, &key);
+		if (ret == -EAGAIN)
+			goto again;
+
 		if (!path->skip_locking) {
 			btrfs_assert_tree_locked(path->nodes[level]);
-			btrfs_tree_lock(next);
-			btrfs_set_lock_blocking(next);
+			ret = btrfs_try_spin_lock(next);
+			if (!ret) {
+				btrfs_set_path_blocking(path);
+				btrfs_tree_lock(next);
+				if (!force_blocking)
+					btrfs_clear_path_blocking(path, next);
+			}
+			if (force_blocking)
+				btrfs_set_lock_blocking(next);
 		}
 	}
+	ret = 0;
 done:
 	unlock_up(path, 0, 1);
-	return 0;
+	path->leave_spinning = old_spinning;
+	if (!old_spinning)
+		btrfs_set_path_blocking(path);
+
+	return ret;
 }

 /*
@@ -143,12 +143,15 @@ static int btrfs_csum_sizes[] = { 4, 0 };
 #define BTRFS_FT_MAX		9

 /*
- * the key defines the order in the tree, and so it also defines (optimal)
- * block layout.  objectid corresonds to the inode number.  The flags
- * tells us things about the object, and is a kind of stream selector.
- * so for a given inode, keys with flags of 1 might refer to the inode
- * data, flags of 2 may point to file data in the btree and flags == 3
- * may point to extents.
+ * The key defines the order in the tree, and so it also defines (optimal)
+ * block layout.
+ *
+ * objectid corresponds to the inode number.
+ *
+ * type tells us things about the object, and is a kind of stream selector.
+ * so for a given inode, keys with type of 1 might refer to the inode data,
+ * type of 2 may point to file data in the btree and type == 3 may point to
+ * extents.
 *
 * offset is the starting byte offset for this key in the stream.
 *
@@ -200,7 +203,7 @@ struct btrfs_dev_item {

 	/*
 	 * starting byte of this partition on the device,
-	 * to allowr for stripe alignment in the future
+	 * to allow for stripe alignment in the future
 	 */
 	__le64 start_offset;

@@ -633,18 +636,35 @@ struct btrfs_space_info {
 	struct rw_semaphore groups_sem;
 };

-struct btrfs_free_space {
-	struct rb_node bytes_index;
-	struct rb_node offset_index;
-	u64 offset;
-	u64 bytes;
+/*
+ * free clusters are used to claim free space in relatively large chunks,
+ * allowing us to do less seeky writes.  They are used for all metadata
+ * allocations and data allocations in ssd mode.
+ */
+struct btrfs_free_cluster {
+	spinlock_t lock;
+	spinlock_t refill_lock;
+	struct rb_root root;
+
+	/* largest extent in this cluster */
+	u64 max_size;
+
+	/* first extent starting offset */
+	u64 window_start;
+
+	struct btrfs_block_group_cache *block_group;
+	/*
+	 * when a cluster is allocated from a block group, we put the
+	 * cluster onto a list in the block group so that it can
+	 * be freed before the block group is freed.
+	 */
+	struct list_head block_group_list;
 };

 struct btrfs_block_group_cache {
 	struct btrfs_key key;
 	struct btrfs_block_group_item item;
 	spinlock_t lock;
-	struct mutex alloc_mutex;
 	struct mutex cache_mutex;
 	u64 pinned;
 	u64 reserved;
@@ -656,6 +676,7 @@ struct btrfs_block_group_cache {
 	struct btrfs_space_info *space_info;

 	/* free space cache stuff */
+	spinlock_t tree_lock;
 	struct rb_root free_space_bytes;
 	struct rb_root free_space_offset;

@@ -667,6 +688,11 @@ struct btrfs_block_group_cache {

 	/* usage count */
 	atomic_t count;
+
+	/* List of struct btrfs_free_clusters for this block group.
+	 * Today it will only have one thing on it, but that may change
+	 */
+	struct list_head cluster_list;
 };

 struct btrfs_leaf_ref_tree {
@@ -728,7 +754,6 @@ struct btrfs_fs_info {
 	struct mutex tree_log_mutex;
 	struct mutex transaction_kthread_mutex;
 	struct mutex cleaner_mutex;
-	struct mutex pinned_mutex;
 	struct mutex chunk_mutex;
 	struct mutex drop_mutex;
 	struct mutex volume_mutex;
@@ -839,8 +864,12 @@ struct btrfs_fs_info {
 	spinlock_t delalloc_lock;
 	spinlock_t new_trans_lock;
 	u64 delalloc_bytes;
-	u64 last_alloc;
-	u64 last_data_alloc;
+
+	/* data_alloc_cluster is only used in ssd mode */
+	struct btrfs_free_cluster data_alloc_cluster;
+
+	/* all metadata allocations go through this cluster */
+	struct btrfs_free_cluster meta_alloc_cluster;

 	spinlock_t ref_cache_lock;
 	u64 total_ref_cache_size;
@@ -932,7 +961,6 @@ struct btrfs_root {
 };

 /*
-
 * inode items have the data typically returned from stat and store other
 * info about object characteristics.  There is one for every file and dir in
 * the FS
@@ -963,7 +991,7 @@ struct btrfs_root {
 #define BTRFS_EXTENT_CSUM_KEY	128

 /*
- * root items point to tree roots.  There are typically in the root
+ * root items point to tree roots.  They are typically in the root
 * tree used by the super block to find all the other trees
 */
 #define BTRFS_ROOT_ITEM_KEY	132
@@ -1010,6 +1038,8 @@ struct btrfs_root {
 #define BTRFS_MOUNT_SSD			(1 << 3)
 #define BTRFS_MOUNT_DEGRADED		(1 << 4)
 #define BTRFS_MOUNT_COMPRESS		(1 << 5)
+#define BTRFS_MOUNT_NOTREELOG           (1 << 6)
+#define BTRFS_MOUNT_FLUSHONCOMMIT       (1 << 7)

 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
@@ -1748,6 +1778,7 @@ static inline struct dentry *fdentry(struct file *file)
 }

 /* extent-tree.c */
+void btrfs_put_block_group(struct btrfs_block_group_cache *cache);
 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root, unsigned long count);
 int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len);
@@ -2174,21 +2205,4 @@ int btrfs_check_acl(struct inode *inode, int mask);
 int btrfs_init_acl(struct inode *inode, struct inode *dir);
 int btrfs_acl_chmod(struct inode *inode);

-/* free-space-cache.c */
-int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
-			 u64 bytenr, u64 size);
-int btrfs_add_free_space_lock(struct btrfs_block_group_cache *block_group,
-			      u64 offset, u64 bytes);
-int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
-			    u64 bytenr, u64 size);
-int btrfs_remove_free_space_lock(struct btrfs_block_group_cache *block_group,
-				 u64 offset, u64 bytes);
-void btrfs_remove_free_space_cache(struct btrfs_block_group_cache
-				   *block_group);
-struct btrfs_free_space *btrfs_find_free_space(struct btrfs_block_group_cache
-					       *block_group, u64 offset,
-					       u64 bytes);
-void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
-			   u64 bytes);
-u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group);
 #endif
@@ -18,7 +18,6 @@

 #include <linux/sched.h>
 #include <linux/sort.h>
-#include <linux/ftrace.h>
 #include "ctree.h"
 #include "delayed-ref.h"
 #include "transaction.h"
@@ -38,6 +38,7 @@
 #include "locking.h"
 #include "ref-cache.h"
 #include "tree-log.h"
+#include "free-space-cache.h"

 static struct extent_io_ops btree_extent_io_ops;
 static void end_workqueue_fn(struct btrfs_work *work);
@@ -1412,8 +1413,6 @@ static int bio_ready_for_csum(struct bio *bio)

 	ret = extent_range_uptodate(io_tree, start + length,
 				    start + buf_len - 1);
-	if (ret == 1)
-		return ret;
 	return ret;
 }

@@ -1647,12 +1646,15 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	mutex_init(&fs_info->ordered_operations_mutex);
 	mutex_init(&fs_info->tree_log_mutex);
 	mutex_init(&fs_info->drop_mutex);
-	mutex_init(&fs_info->pinned_mutex);
 	mutex_init(&fs_info->chunk_mutex);
 	mutex_init(&fs_info->transaction_kthread_mutex);
 	mutex_init(&fs_info->cleaner_mutex);
 	mutex_init(&fs_info->volume_mutex);
 	mutex_init(&fs_info->tree_reloc_mutex);
+
+	btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
+	btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
+
 	init_waitqueue_head(&fs_info->transaction_throttle);
 	init_waitqueue_head(&fs_info->transaction_wait);
 	init_waitqueue_head(&fs_info->async_submit_wait);
@@ -2884,25 +2884,19 @@ int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 		disko = 0;
 		flags = 0;

-		switch (em->block_start) {
-		case EXTENT_MAP_LAST_BYTE:
+		if (em->block_start == EXTENT_MAP_LAST_BYTE) {
 			end = 1;
 			flags |= FIEMAP_EXTENT_LAST;
-			break;
-		case EXTENT_MAP_HOLE:
+		} else if (em->block_start == EXTENT_MAP_HOLE) {
 			flags |= FIEMAP_EXTENT_UNWRITTEN;
-			break;
-		case EXTENT_MAP_INLINE:
+		} else if (em->block_start == EXTENT_MAP_INLINE) {
 			flags |= (FIEMAP_EXTENT_DATA_INLINE |
 				  FIEMAP_EXTENT_NOT_ALIGNED);
-			break;
-		case EXTENT_MAP_DELALLOC:
+		} else if (em->block_start == EXTENT_MAP_DELALLOC) {
 			flags |= (FIEMAP_EXTENT_DELALLOC |
 				  FIEMAP_EXTENT_UNKNOWN);
-			break;
-		default:
+		} else {
 			disko = em->block_start;
-			break;
 		}
 		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
 			flags |= FIEMAP_EXTENT_ENCODED;
@@ -234,7 +234,6 @@ int add_extent_mapping(struct extent_map_tree *tree,
 	rb = tree_insert(&tree->map, em->start, &em->rb_node);
 	if (rb) {
 		ret = -EEXIST;
-		free_extent_map(merge);
 		goto out;
 	}
 	atomic_inc(&em->refs);
@@ -0,0 +1,44 @@
+/*
+ * Copyright (C) 2009 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __BTRFS_FREE_SPACE_CACHE
+#define __BTRFS_FREE_SPACE_CACHE
+
+int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
+			 u64 bytenr, u64 size);
+int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
+			    u64 bytenr, u64 size);
+void btrfs_remove_free_space_cache(struct btrfs_block_group_cache
+				   *block_group);
+u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
+			       u64 offset, u64 bytes, u64 empty_size);
+void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
+			   u64 bytes);
+u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group);
+int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
+			     struct btrfs_block_group_cache *block_group,
+			     struct btrfs_free_cluster *cluster,
+			     u64 offset, u64 bytes, u64 empty_size);
+void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster);
+u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
+			     struct btrfs_free_cluster *cluster, u64 bytes,
+			     u64 min_start);
+int btrfs_return_cluster_to_free_space(
+			       struct btrfs_block_group_cache *block_group,
+			       struct btrfs_free_cluster *cluster);
+#endif
@@ -3481,8 +3481,10 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,

 	if (dir) {
 		ret = btrfs_set_inode_index(dir, index);
-		if (ret)
+		if (ret) {
+			iput(inode);
 			return ERR_PTR(ret);
+		}
 	}
 	/*
 	 * index_cnt is ignored for everything but a dir,
@@ -3565,6 +3567,7 @@ fail:
 	if (dir)
 		BTRFS_I(dir)->index_cnt--;
 	btrfs_free_path(path);
+	iput(inode);
 	return ERR_PTR(ret);
 }

@@ -60,8 +60,8 @@ void btrfs_clear_lock_blocking(struct extent_buffer *eb)

 /*
 * unfortunately, many of the places that currently set a lock to blocking
- * don't end up blocking for every long, and often they don't block
- * at all.  For a dbench 50 run, if we don't spin one the blocking bit
+ * don't end up blocking for very long, and often they don't block
+ * at all.  For a dbench 50 run, if we don't spin on the blocking bit
 * at all, the context switch rate can jump up to 400,000/sec or more.
 *
 * So, we're still stuck with this crummy spin on the blocking bit,
@@ -24,6 +24,7 @@
 #include <linux/highmem.h>
 #include <linux/time.h>
 #include <linux/init.h>
+#include <linux/seq_file.h>
 #include <linux/string.h>
 #include <linux/smp_lock.h>
 #include <linux/backing-dev.h>
@@ -66,7 +67,8 @@ static void btrfs_put_super(struct super_block *sb)
 enum {
 	Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow,
 	Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
-	Opt_ssd, Opt_thread_pool, Opt_noacl,  Opt_compress, Opt_err,
+	Opt_ssd, Opt_thread_pool, Opt_noacl,  Opt_compress, Opt_notreelog,
+	Opt_flushoncommit, Opt_err,
 };

 static match_table_t tokens = {
@@ -83,6 +85,8 @@ static match_table_t tokens = {
 	{Opt_compress, "compress"},
 	{Opt_ssd, "ssd"},
 	{Opt_noacl, "noacl"},
+	{Opt_notreelog, "notreelog"},
+	{Opt_flushoncommit, "flushoncommit"},
 	{Opt_err, NULL},
 };

@@ -222,6 +226,14 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 		case Opt_noacl:
 			root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 			break;
+		case Opt_notreelog:
+			printk(KERN_INFO "btrfs: disabling tree log\n");
+			btrfs_set_opt(info->mount_opt, NOTREELOG);
+			break;
+		case Opt_flushoncommit:
+			printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
+			btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
+			break;
 		default:
 			break;
 		}
@@ -363,9 +375,8 @@ fail_close:
 int btrfs_sync_fs(struct super_block *sb, int wait)
 {
 	struct btrfs_trans_handle *trans;
-	struct btrfs_root *root;
+	struct btrfs_root *root = btrfs_sb(sb);
 	int ret;
-	root = btrfs_sb(sb);

 	if (sb->s_flags & MS_RDONLY)
 		return 0;
@@ -385,6 +396,41 @@ int btrfs_sync_fs(struct super_block *sb, int wait)
 	return ret;
 }

+static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
+{
+	struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
+	struct btrfs_fs_info *info = root->fs_info;
+
+	if (btrfs_test_opt(root, DEGRADED))
+		seq_puts(seq, ",degraded");
+	if (btrfs_test_opt(root, NODATASUM))
+		seq_puts(seq, ",nodatasum");
+	if (btrfs_test_opt(root, NODATACOW))
+		seq_puts(seq, ",nodatacow");
+	if (btrfs_test_opt(root, NOBARRIER))
+		seq_puts(seq, ",nobarrier");
+	if (info->max_extent != (u64)-1)
+		seq_printf(seq, ",max_extent=%llu", info->max_extent);
+	if (info->max_inline != 8192 * 1024)
+		seq_printf(seq, ",max_inline=%llu", info->max_inline);
+	if (info->alloc_start != 0)
+		seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
+	if (info->thread_pool_size !=  min_t(unsigned long,
+					     num_online_cpus() + 2, 8))
+		seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
+	if (btrfs_test_opt(root, COMPRESS))
+		seq_puts(seq, ",compress");
+	if (btrfs_test_opt(root, SSD))
+		seq_puts(seq, ",ssd");
+	if (btrfs_test_opt(root, NOTREELOG))
+		seq_puts(seq, ",no-treelog");
+	if (btrfs_test_opt(root, FLUSHONCOMMIT))
+		seq_puts(seq, ",flush-on-commit");
+	if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
+		seq_puts(seq, ",noacl");
+	return 0;
+}
+
 static void btrfs_write_super(struct super_block *sb)
 {
 	sb->s_dirt = 0;
@@ -630,7 +676,7 @@ static struct super_operations btrfs_super_ops = {
 	.put_super	= btrfs_put_super,
 	.write_super	= btrfs_write_super,
 	.sync_fs	= btrfs_sync_fs,
-	.show_options	= generic_show_options,
+	.show_options	= btrfs_show_options,
 	.write_inode	= btrfs_write_inode,
 	.dirty_inode	= btrfs_dirty_inode,
 	.alloc_inode	= btrfs_alloc_inode,
@@ -53,8 +53,6 @@ static noinline int join_transaction(struct btrfs_root *root)
 					     GFP_NOFS);
 		BUG_ON(!cur_trans);
 		root->fs_info->generation++;
-		root->fs_info->last_alloc = 0;
-		root->fs_info->last_data_alloc = 0;
 		cur_trans->num_writers = 1;
 		cur_trans->num_joined = 0;
 		cur_trans->transid = root->fs_info->generation;
@@ -974,6 +972,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	int ret;
 	int should_grow = 0;
 	unsigned long now = get_seconds();
+	int flush_on_commit = btrfs_test_opt(root, FLUSHONCOMMIT);

 	btrfs_run_ordered_operations(root, 0);

@@ -1053,7 +1052,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,

 		mutex_unlock(&root->fs_info->trans_mutex);

-		if (snap_pending) {
+		if (flush_on_commit || snap_pending) {
+			if (flush_on_commit)
+				btrfs_start_delalloc_inodes(root);
 			ret = btrfs_wait_ordered_extents(root, 1);
 			BUG_ON(ret);
 		}
@@ -262,11 +262,9 @@ static int process_one_buffer(struct btrfs_root *log,
 			      struct extent_buffer *eb,
 			      struct walk_control *wc, u64 gen)
 {
-	if (wc->pin) {
-		mutex_lock(&log->fs_info->pinned_mutex);
+	if (wc->pin)
 		btrfs_update_pinned_extents(log->fs_info->extent_root,
 					    eb->start, eb->len, 1);
-	}

 	if (btrfs_buffer_uptodate(eb, gen)) {
 		if (wc->write)
@@ -1224,8 +1222,7 @@ insert:
 	ret = insert_one_name(trans, root, path, key->objectid, key->offset,
 			      name, name_len, log_type, &log_key);

-	if (ret && ret != -ENOENT)
-		BUG();
+	BUG_ON(ret && ret != -ENOENT);
 	goto out;
 }

@@ -2900,6 +2897,11 @@ int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,

 	sb = inode->i_sb;

+	if (btrfs_test_opt(root, NOTREELOG)) {
+		ret = 1;
+		goto end_no_trans;
+	}
+
 	if (root->fs_info->last_trans_log_full_commit >
 	    root->fs_info->last_trans_committed) {
 		ret = 1;
@@ -20,6 +20,7 @@
 #include <linux/buffer_head.h>
 #include <linux/blkdev.h>
 #include <linux/random.h>
+#include <linux/iocontext.h>
 #include <asm/div64.h>
 #include "compat.h"
 #include "ctree.h"
@@ -145,8 +146,9 @@ static noinline int run_scheduled_bios(struct btrfs_device *device)
 	int again = 0;
 	unsigned long num_run = 0;
 	unsigned long limit;
+	unsigned long last_waited = 0;

-	bdi = device->bdev->bd_inode->i_mapping->backing_dev_info;
+	bdi = blk_get_backing_dev_info(device->bdev);
 	fs_info = device->dev_root->fs_info;
 	limit = btrfs_async_submit_limit(fs_info);
 	limit = limit * 2 / 3;
@@ -207,7 +209,32 @@ loop_lock:
 		if (pending && bdi_write_congested(bdi) && num_run > 16 &&
 		    fs_info->fs_devices->open_devices > 1) {
 			struct bio *old_head;
+			struct io_context *ioc;

+			ioc = current->io_context;
+
+			/*
+			 * the main goal here is that we don't want to
+			 * block if we're going to be able to submit
+			 * more requests without blocking.
+			 *
+			 * This code does two great things, it pokes into
+			 * the elevator code from a filesystem _and_
+			 * it makes assumptions about how batching works.
+			 */
+			if (ioc && ioc->nr_batch_requests > 0 &&
+			    time_before(jiffies, ioc->last_waited + HZ/50UL) &&
+			    (last_waited == 0 ||
+			     ioc->last_waited == last_waited)) {
+				/*
+				 * we want to go through our batch of
+				 * requests and stop.  So, we copy out
+				 * the ioc->last_waited time and test
+				 * against it before looping
+				 */
+				last_waited = ioc->last_waited;
+				continue;
+			}
 			spin_lock(&device->io_lock);

 			old_head = device->pending_bios;
@@ -231,6 +258,18 @@ loop_lock:
 	if (device->pending_bios)
 		goto loop_lock;
 	spin_unlock(&device->io_lock);
+
+	/*
+	 * IO has already been through a long path to get here.  Checksumming,
+	 * async helper threads, perhaps compression.  We've done a pretty
+	 * good job of collecting a batch of IO and should just unplug
+	 * the device right away.
+	 *
+	 * This will help anyone who is waiting on the IO, they might have
+	 * already unplugged, but managed to do so before the bio they
+	 * cared about found its way down here.
+	 */
+	blk_run_backing_dev(bdi, NULL);
 done:
 	return 0;
 }
@@ -76,7 +76,7 @@ struct btrfs_device {
 struct btrfs_fs_devices {
 	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */

-	/* the device with this id has the most recent coyp of the super */
+	/* the device with this id has the most recent copy of the super */
 	u64 latest_devid;
 	u64 latest_trans;
 	u64 num_devices;