Btrfs: Leaf reference cache update

This changes the reference cache to make a single cache per root
instead of one cache per transaction, and to key by the byte number
of the disk block instead of the keys inside.

This makes it much less likely to have cache misses if a snapshot
or something has an extra reference on a higher node or a leaf while
the first transaction that added the leaf into the cache is dropping.

Some throttling is added to functions that free blocks heavily so they
wait for old transactions to drop.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 files changed, 61 insertions, 109 deletions

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 34ed23d..4eca0aa 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -594,7 +594,6 @@ struct btrfs_fs_info {
 
 	spinlock_t ref_cache_lock;
 	u64 total_ref_cache_size;
-	u64 running_ref_cache_size;
 
 	u64 avail_data_alloc_bits;
 	u64 avail_metadata_alloc_bits;
@@ -606,10 +605,18 @@ struct btrfs_fs_info {
 	void *bdev_holder;
 };
 
+struct btrfs_leaf_ref_tree {
+	struct rb_root root;
+	struct btrfs_leaf_ref *last;
+	struct list_head list;
+	spinlock_t lock;
+};
+
 /*
  * in ram representation of the tree.  extent_root is used for all allocations
  * and for the extent tree extent_root root.
  */
+struct dirty_root;
 struct btrfs_root {
 	struct extent_buffer *node;
 
@@ -618,6 +625,8 @@ struct btrfs_root {
 
 	struct extent_buffer *commit_root;
 	struct btrfs_leaf_ref_tree *ref_tree;
+	struct btrfs_leaf_ref_tree ref_tree_struct;
+	struct dirty_root *dirty_root;
 
 	struct btrfs_root_item root_item;
 	struct btrfs_key root_key;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 4f0e1d0..eccdf13 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -40,6 +40,7 @@
 #include "print-tree.h"
 #include "async-thread.h"
 #include "locking.h"
+#include "ref-cache.h"
 
 #if 0
 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
@@ -737,6 +738,10 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	spin_lock_init(&root->node_lock);
 	spin_lock_init(&root->orphan_lock);
 	mutex_init(&root->objectid_mutex);
+
+	btrfs_leaf_ref_tree_init(&root->ref_tree_struct);
+	root->ref_tree = &root->ref_tree_struct;
+
 	memset(&root->root_key, 0, sizeof(root->root_key));
 	memset(&root->root_item, 0, sizeof(root->root_item));
 	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
@@ -1176,9 +1181,6 @@ static int transaction_kthread(void *arg)
 			goto sleep;
 		}
 
-		printk("btrfs: running reference cache size %Lu\n",
-			root->fs_info->running_ref_cache_size);
-
 		now = get_seconds();
 		if (now < cur->start_time || now - cur->start_time < 30) {
 			mutex_unlock(&root->fs_info->trans_mutex);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 7b24f15..0e294cf 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1004,8 +1004,6 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			goto out;
 		}
 
-		btrfs_item_key_to_cpu(buf, &ref->key, 0);
-
 		ref->bytenr = buf->start;
 		ref->owner = btrfs_header_owner(buf);
 		ref->generation = btrfs_header_generation(buf);
@@ -2387,19 +2385,15 @@ static void noinline reada_walk_down(struct btrfs_root *root,
 	}
 }
 
-/*
- * we want to avoid as much random IO as we can with the alloc mutex
- * held, so drop the lock and do the lookup, then do it again with the
- * lock held.
- */
 int drop_snap_lookup_refcount(struct btrfs_root *root, u64 start, u64 len,
 			      u32 *refs)
 {
+	int ret;
 	mutex_unlock(&root->fs_info->alloc_mutex);
-	lookup_extent_ref(NULL, root, start, len, refs);
+	ret = lookup_extent_ref(NULL, root, start, len, refs);
 	cond_resched();
 	mutex_lock(&root->fs_info->alloc_mutex);
-	return lookup_extent_ref(NULL, root, start, len, refs);
+	return ret;
 }
 
 /*
@@ -2468,11 +2462,11 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
 			BUG_ON(ret);
 			continue;
 		}
-		
+
 		if (*level == 1) {
 			struct btrfs_key key;
 			btrfs_node_key_to_cpu(cur, &key, path->slots[*level]);
-			ref = btrfs_lookup_leaf_ref(root, &key);
+			ref = btrfs_lookup_leaf_ref(root, bytenr);
 			if (ref) {
 				ret = drop_leaf_ref(trans, root, ref);
 				BUG_ON(ret);
@@ -2482,7 +2476,6 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
 				break;
 			}
 		}
-
 		next = btrfs_find_tree_block(root, bytenr, blocksize);
 		if (!next || !btrfs_buffer_uptodate(next, ptr_gen)) {
 			free_extent_buffer(next);
@@ -2672,6 +2665,7 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
 			ret = -EAGAIN;
 			break;
 		}
+		wake_up(&root->fs_info->transaction_throttle);
 	}
 	for (i = 0; i <= orig_level; i++) {
 		if (path->nodes[i]) {
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index e5ffb66..3efec25 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -347,7 +347,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans,
 		btrfs_update_inode(trans, root, inode);
 	}
 failed:
-	err = btrfs_end_transaction_throttle(trans, root);
+	err = btrfs_end_transaction(trans, root);
 out_unlock:
 	unlock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS);
 	return err;
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index cf9534b..4f977ea 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -2482,7 +2482,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry,
 	btrfs_update_inode_block_group(trans, dir);
 out_unlock:
 	nr = trans->blocks_used;
-	btrfs_end_transaction_throttle(trans, root);
+	btrfs_end_transaction(trans, root);
 fail:
 	if (drop_inode) {
 		inode_dec_link_count(inode);
@@ -2535,7 +2535,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
 		drop_inode = 1;
 
 	nr = trans->blocks_used;
-	btrfs_end_transaction_throttle(trans, root);
+	btrfs_end_transaction(trans, root);
 fail:
 	if (drop_inode) {
 		inode_dec_link_count(inode);
@@ -2609,7 +2609,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 
 out_fail:
 	nr = trans->blocks_used;
-	btrfs_end_transaction_throttle(trans, root);
+	btrfs_end_transaction(trans, root);
 
 out_unlock:
 	if (drop_on_err)
@@ -3548,7 +3548,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 
 out_unlock:
 	nr = trans->blocks_used;
-	btrfs_end_transaction_throttle(trans, root);
+	btrfs_end_transaction(trans, root);
 out_fail:
 	if (drop_inode) {
 		inode_dec_link_count(inode);
diff --git a/fs/btrfs/ref-cache.c b/fs/btrfs/ref-cache.c
index 95a9fae..ec95877 100644
--- a/fs/btrfs/ref-cache.c
+++ b/fs/btrfs/ref-cache.c
@@ -29,6 +29,7 @@ struct btrfs_leaf_ref *btrfs_alloc_leaf_ref(int nr_extents)
 	if (ref) {
 		memset(ref, 0, sizeof(*ref));
 		atomic_set(&ref->usage, 1);
+		INIT_LIST_HEAD(&ref->list);
 	}
 	return ref;
 }
@@ -44,40 +45,21 @@ void btrfs_free_leaf_ref(struct btrfs_leaf_ref *ref)
 	}
 }
 
-static int comp_keys(struct btrfs_key *k1, struct btrfs_key *k2)
-{
-	if (k1->objectid > k2->objectid)
-		return 1;
-	if (k1->objectid < k2->objectid)
-		return -1;
-	if (k1->type > k2->type)
-		return 1;
-	if (k1->type < k2->type)
-		return -1;
-	if (k1->offset > k2->offset)
-		return 1;
-	if (k1->offset < k2->offset)
-		return -1;
-	return 0;
-}
-
-static struct rb_node *tree_insert(struct rb_root *root, struct btrfs_key *key,
+static struct rb_node *tree_insert(struct rb_root *root, u64 bytenr,
 				   struct rb_node *node)
 {
 	struct rb_node ** p = &root->rb_node;
 	struct rb_node * parent = NULL;
 	struct btrfs_leaf_ref *entry;
-	int ret;
 
 	while(*p) {
 		parent = *p;
 		entry = rb_entry(parent, struct btrfs_leaf_ref, rb_node);
 		WARN_ON(!entry->in_tree);
 
-		ret = comp_keys(key, &entry->key);
-		if (ret < 0)
+		if (bytenr < entry->bytenr)
 			p = &(*p)->rb_left;
-		else if (ret > 0)
+		else if (bytenr > entry->bytenr)
 			p = &(*p)->rb_right;
 		else
 			return parent;
@@ -90,20 +72,18 @@ static struct rb_node *tree_insert(struct rb_root *root, struct btrfs_key *key,
 	return NULL;
 }
 
-static struct rb_node *tree_search(struct rb_root *root, struct btrfs_key *key)
+static struct rb_node *tree_search(struct rb_root *root, u64 bytenr)
 {
 	struct rb_node * n = root->rb_node;
 	struct btrfs_leaf_ref *entry;
-	int ret;
 
 	while(n) {
 		entry = rb_entry(n, struct btrfs_leaf_ref, rb_node);
 		WARN_ON(!entry->in_tree);
 
-		ret = comp_keys(key, &entry->key);
-		if (ret < 0)
+		if (bytenr < entry->bytenr)
 			n = n->rb_left;
-		else if (ret > 0)
+		else if (bytenr > entry->bytenr)
 			n = n->rb_right;
 		else
 			return n;
@@ -122,11 +102,11 @@ int btrfs_remove_leaf_refs(struct btrfs_root *root)
 
 	spin_lock(&tree->lock);
 	while(!btrfs_leaf_ref_tree_empty(tree)) {
-		tree->last = NULL;
 		rb = rb_first(&tree->root);
 		ref = rb_entry(rb, struct btrfs_leaf_ref, rb_node);
 		rb_erase(&ref->rb_node, &tree->root);
 		ref->in_tree = 0;
+		list_del_init(&ref->list);
 
 		spin_unlock(&tree->lock);
 
@@ -140,7 +120,7 @@ int btrfs_remove_leaf_refs(struct btrfs_root *root)
 }
 
 struct btrfs_leaf_ref *btrfs_lookup_leaf_ref(struct btrfs_root *root,
-					     struct btrfs_key *key)
+					     u64 bytenr)
 {
 	struct rb_node *rb;
 	struct btrfs_leaf_ref *ref = NULL;
@@ -150,15 +130,9 @@ struct btrfs_leaf_ref *btrfs_lookup_leaf_ref(struct btrfs_root *root,
 		return NULL;
 
 	spin_lock(&tree->lock);
-	if (tree->last && comp_keys(key, &tree->last->key) == 0) {
-		ref = tree->last;
-	} else {
-		rb = tree_search(&tree->root, key);
-		if (rb) {
-			ref = rb_entry(rb, struct btrfs_leaf_ref, rb_node);
-			tree->last = ref;
-		}
-	}
+	rb = tree_search(&tree->root, bytenr);
+	if (rb)
+		ref = rb_entry(rb, struct btrfs_leaf_ref, rb_node);
 	if (ref)
 		atomic_inc(&ref->usage);
 	spin_unlock(&tree->lock);
@@ -171,21 +145,17 @@ int btrfs_add_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref)
 	struct rb_node *rb;
 	size_t size = btrfs_leaf_ref_size(ref->nritems);
 	struct btrfs_leaf_ref_tree *tree = root->ref_tree;
-	struct btrfs_transaction *trans = root->fs_info->running_transaction;
 
 	spin_lock(&tree->lock);
-	rb = tree_insert(&tree->root, &ref->key, &ref->rb_node);
+	rb = tree_insert(&tree->root, ref->bytenr, &ref->rb_node);
 	if (rb) {
 		ret = -EEXIST;
 	} else {
 		spin_lock(&root->fs_info->ref_cache_lock);
 		root->fs_info->total_ref_cache_size += size;
-		if (trans && tree->generation == trans->transid)
-			root->fs_info->running_ref_cache_size += size;
 		spin_unlock(&root->fs_info->ref_cache_lock);
-
-		tree->last = ref;
 		atomic_inc(&ref->usage);
+		list_add_tail(&ref->list, &tree->list);
 	}
 	spin_unlock(&tree->lock);
 	return ret;
@@ -195,28 +165,17 @@ int btrfs_remove_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref)
 {
 	size_t size = btrfs_leaf_ref_size(ref->nritems);
 	struct btrfs_leaf_ref_tree *tree = root->ref_tree;
-	struct btrfs_transaction *trans = root->fs_info->running_transaction;
 
 	BUG_ON(!ref->in_tree);
 	spin_lock(&tree->lock);
 	
 	spin_lock(&root->fs_info->ref_cache_lock);
 	root->fs_info->total_ref_cache_size -= size;
-	if (trans && tree->generation == trans->transid)
-		root->fs_info->running_ref_cache_size -= size;
 	spin_unlock(&root->fs_info->ref_cache_lock);
 
-	if (tree->last == ref) {
-		struct rb_node *next = rb_next(&ref->rb_node);
-		if (next) {
-			tree->last = rb_entry(next, struct btrfs_leaf_ref,
-					      rb_node);
-		} else
-			tree->last = NULL;
-	}
-
 	rb_erase(&ref->rb_node, &tree->root);
 	ref->in_tree = 0;
+	list_del_init(&ref->list);
 
 	spin_unlock(&tree->lock);
 
diff --git a/fs/btrfs/ref-cache.h b/fs/btrfs/ref-cache.h
index 79ecc47..823c049 100644
--- a/fs/btrfs/ref-cache.h
+++ b/fs/btrfs/ref-cache.h
@@ -15,6 +15,8 @@
  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  * Boston, MA 021110-1307, USA.
  */
+#ifndef __REFCACHE__
+#define __REFCACHE__
 
 struct btrfs_extent_info {
 	u64 bytenr;
@@ -25,7 +27,6 @@ struct btrfs_extent_info {
 
 struct btrfs_leaf_ref {
 	struct rb_node rb_node;
-	struct btrfs_key key;
 	int in_tree;
 	atomic_t usage;
 
@@ -33,14 +34,9 @@ struct btrfs_leaf_ref {
 	u64 owner;
 	u64 generation;
 	int nritems;
-	struct btrfs_extent_info extents[];
-};
 
-struct btrfs_leaf_ref_tree {
-	struct rb_root root;
-	struct btrfs_leaf_ref *last;
-	u64 generation;
-	spinlock_t lock;
+	struct list_head list;
+	struct btrfs_extent_info extents[];
 };
 
 static inline size_t btrfs_leaf_ref_size(int nr_extents)
@@ -53,7 +49,7 @@ static inline void btrfs_leaf_ref_tree_init(struct btrfs_leaf_ref_tree *tree)
 {
 	tree->root.rb_node = NULL;
 	tree->last = NULL;
-	tree->generation = 0;
+	INIT_LIST_HEAD(&tree->list);
 	spin_lock_init(&tree->lock);
 }
 
@@ -66,7 +62,9 @@ void btrfs_leaf_ref_tree_init(struct btrfs_leaf_ref_tree *tree);
 struct btrfs_leaf_ref *btrfs_alloc_leaf_ref(int nr_extents);
 void btrfs_free_leaf_ref(struct btrfs_leaf_ref *ref);
 struct btrfs_leaf_ref *btrfs_lookup_leaf_ref(struct btrfs_root *root,
-					     struct btrfs_key *key);
+					     u64 bytenr);
 int btrfs_add_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref);
 int btrfs_remove_leaf_refs(struct btrfs_root *root);
 int btrfs_remove_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref);
+
+#endif
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 543e5ee..fcef3ca 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -36,7 +36,6 @@ struct dirty_root {
 	struct list_head list;
 	struct btrfs_root *root;
 	struct btrfs_root *latest_root;
-	struct btrfs_leaf_ref_tree ref_tree;
 };
 
 static noinline void put_transaction(struct btrfs_transaction *transaction)
@@ -108,13 +107,13 @@ static noinline int record_root_in_trans(struct btrfs_root *root)
 
 			dirty->latest_root = root;
 			INIT_LIST_HEAD(&dirty->list);
-			btrfs_leaf_ref_tree_init(&dirty->ref_tree);
-			dirty->ref_tree.generation = running_trans_id;
 
 			root->commit_root = btrfs_root_node(root);
-			root->ref_tree = &dirty->ref_tree;
+			root->dirty_root = dirty;
 
 			memcpy(dirty->root, root, sizeof(*root));
+			dirty->root->ref_tree = &root->ref_tree_struct;
+
 			spin_lock_init(&dirty->root->node_lock);
 			mutex_init(&dirty->root->objectid_mutex);
 			dirty->root->node = root->commit_root;
@@ -217,12 +216,13 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
 	if (waitqueue_active(&cur_trans->writer_wait))
 		wake_up(&cur_trans->writer_wait);
 
-	if (0 && cur_trans->in_commit && throttle) {
+	if (throttle && atomic_read(&root->fs_info->throttles)) {
 		DEFINE_WAIT(wait);
 		mutex_unlock(&root->fs_info->trans_mutex);
 		prepare_to_wait(&root->fs_info->transaction_throttle, &wait,
 				TASK_UNINTERRUPTIBLE);
-		schedule();
+		if (atomic_read(&root->fs_info->throttles))
+			schedule();
 		finish_wait(&root->fs_info->transaction_throttle, &wait);
 		mutex_lock(&root->fs_info->trans_mutex);
 	}
@@ -333,6 +333,8 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
 		list_del_init(next);
 		root = list_entry(next, struct btrfs_root, dirty_list);
 		update_cowonly_root(trans, root);
+		if (root->fs_info->closing)
+			btrfs_remove_leaf_refs(root);
 	}
 	return 0;
 }
@@ -346,10 +348,8 @@ int btrfs_add_dead_root(struct btrfs_root *root,
 	dirty = kmalloc(sizeof(*dirty), GFP_NOFS);
 	if (!dirty)
 		return -ENOMEM;
-	btrfs_leaf_ref_tree_init(&dirty->ref_tree);
 	dirty->root = root;
 	dirty->latest_root = latest;
-	root->ref_tree = NULL;
 	list_add(&dirty->list, dead_list);
 	return 0;
 }
@@ -379,18 +379,14 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans,
 				     BTRFS_ROOT_TRANS_TAG);
 
 			BUG_ON(!root->ref_tree);
-			dirty = container_of(root->ref_tree, struct dirty_root,
-					     ref_tree);
+			dirty = root->dirty_root;
 
 			if (root->commit_root == root->node) {
 				WARN_ON(root->node->start !=
 					btrfs_root_bytenr(&root->root_item));
 
-				BUG_ON(!btrfs_leaf_ref_tree_empty(
-							root->ref_tree));
 				free_extent_buffer(root->commit_root);
 				root->commit_root = NULL;
-				root->ref_tree = NULL;
 				
 				kfree(dirty->root);
 				kfree(dirty);
@@ -410,7 +406,6 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans,
 			       sizeof(struct btrfs_disk_key));
 			root->root_item.drop_level = 0;
 			root->commit_root = NULL;
-			root->ref_tree = NULL;
 			root->root_key.offset = root->fs_info->generation;
 			btrfs_set_root_bytenr(&root->root_item,
 					      root->node->start);
@@ -485,7 +480,7 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root,
 	while(!list_empty(list)) {
 		struct btrfs_root *root;
 
-		dirty = list_entry(list->next, struct dirty_root, list);
+		dirty = list_entry(list->prev, struct dirty_root, list);
 		list_del_init(&dirty->list);
 
 		num_bytes = btrfs_root_used(&dirty->root->root_item);
@@ -507,7 +502,7 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root,
 			if (err)
 				ret = err;
 			nr = trans->blocks_used;
-			ret = btrfs_end_transaction_throttle(trans, tree_root);
+			ret = btrfs_end_transaction(trans, tree_root);
 			BUG_ON(ret);
 
 			mutex_unlock(&root->fs_info->drop_mutex);
@@ -517,6 +512,7 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root,
 		}
 		BUG_ON(ret);
 		atomic_dec(&root->fs_info->throttles);
+		wake_up(&root->fs_info->transaction_throttle);
 
 		mutex_lock(&root->fs_info->alloc_mutex);
 		num_bytes -= btrfs_root_used(&dirty->root->root_item);
@@ -539,8 +535,6 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root,
 		ret = btrfs_end_transaction(trans, tree_root);
 		BUG_ON(ret);
 
-		btrfs_remove_leaf_refs(dirty->root);
-
 		free_extent_buffer(dirty->root->node);
 		kfree(dirty->root);
 		kfree(dirty);
@@ -725,10 +719,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 			      &dirty_fs_roots);
 	BUG_ON(ret);
 
-	spin_lock(&root->fs_info->ref_cache_lock);
-	root->fs_info->running_ref_cache_size = 0;
-	spin_unlock(&root->fs_info->ref_cache_lock);
-
 	ret = btrfs_commit_tree_roots(trans, root);
 	BUG_ON(ret);