Btrfs: Fix per root used space accounting

The bytes_used field in root item was originally planned to
trace the amount of used data and tree blocks. But it never
worked right since we can't trace freeing of data accurately.
This patch changes it to only trace the amount of tree blocks.

Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

5 files changed, 47 insertions, 27 deletions

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 9d4ba3470c1..c4bc570a396 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -456,9 +456,8 @@ static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 		extent_buffer_get(cow);
 		spin_unlock(&root->node_lock);
 
-		btrfs_free_extent(trans, root, buf->start, buf->len,
-				  parent_start, root->root_key.objectid,
-				  level, 0);
+		btrfs_free_tree_block(trans, root, buf->start, buf->len,
+				parent_start, root->root_key.objectid, level);
 		free_extent_buffer(buf);
 		add_root_to_dirty_list(root);
 	} else {
@@ -473,9 +472,8 @@ static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 		btrfs_set_node_ptr_generation(parent, parent_slot,
 					      trans->transid);
 		btrfs_mark_buffer_dirty(parent);
-		btrfs_free_extent(trans, root, buf->start, buf->len,
-				  parent_start, root->root_key.objectid,
-				  level, 0);
+		btrfs_free_tree_block(trans, root, buf->start, buf->len,
+				parent_start, root->root_key.objectid, level);
 	}
 	if (unlock_orig)
 		btrfs_tree_unlock(buf);
@@ -1035,8 +1033,8 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 		btrfs_tree_unlock(mid);
 		/* once for the path */
 		free_extent_buffer(mid);
-		ret = btrfs_free_extent(trans, root, mid->start, mid->len,
-					0, root->root_key.objectid, level, 1);
+		ret = btrfs_free_tree_block(trans, root, mid->start, mid->len,
+					    0, root->root_key.objectid, level);
 		/* once for the root ptr */
 		free_extent_buffer(mid);
 		return ret;
@@ -1100,10 +1098,10 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 				       1);
 			if (wret)
 				ret = wret;
-			wret = btrfs_free_extent(trans, root, bytenr,
-						 blocksize, 0,
-						 root->root_key.objectid,
-						 level, 0);
+			wret = btrfs_free_tree_block(trans, root,
+						     bytenr, blocksize, 0,
+						     root->root_key.objectid,
+						     level);
 			if (wret)
 				ret = wret;
 		} else {
@@ -1148,9 +1146,8 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 		wret = del_ptr(trans, root, path, level + 1, pslot);
 		if (wret)
 			ret = wret;
-		wret = btrfs_free_extent(trans, root, bytenr, blocksize,
-					 0, root->root_key.objectid,
-					 level, 0);
+		wret = btrfs_free_tree_block(trans, root, bytenr, blocksize,
+					 0, root->root_key.objectid, level);
 		if (wret)
 			ret = wret;
 	} else {
@@ -3794,8 +3791,8 @@ static noinline int btrfs_del_leaf(struct btrfs_trans_handle *trans,
 	 */
 	btrfs_unlock_up_safe(path, 0);
 
-	ret = btrfs_free_extent(trans, root, leaf->start, leaf->len,
-				0, root->root_key.objectid, 0, 0);
+	ret = btrfs_free_tree_block(trans, root, leaf->start, leaf->len,
+				    0, root->root_key.objectid, 0);
 	return ret;
 }
 /*
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 1983c889bb1..9f806dd04c2 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1982,6 +1982,10 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 					u64 parent, u64 root_objectid,
 					struct btrfs_disk_key *key, int level,
 					u64 hint, u64 empty_size);
+int btrfs_free_tree_block(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root,
+			  u64 bytenr, u32 blocksize,
+			  u64 parent, u64 root_objectid, int level);
 struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
 					    struct btrfs_root *root,
 					    u64 bytenr, u32 blocksize,
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index fcdccfa4600..44d7a322ec2 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -3454,14 +3454,6 @@ static int update_block_group(struct btrfs_trans_handle *trans,
 	else
 		old_val -= num_bytes;
 	btrfs_set_super_bytes_used(&info->super_copy, old_val);
-
-	/* block accounting for root item */
-	old_val = btrfs_root_used(&root->root_item);
-	if (alloc)
-		old_val += num_bytes;
-	else
-		old_val -= num_bytes;
-	btrfs_set_root_used(&root->root_item, old_val);
 	spin_unlock(&info->delalloc_lock);
 
 	while (total) {
@@ -4049,6 +4041,21 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans,
 	return ret;
 }
 
+int btrfs_free_tree_block(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root,
+			  u64 bytenr, u32 blocksize,
+			  u64 parent, u64 root_objectid, int level)
+{
+	u64 used;
+	spin_lock(&root->node_lock);
+	used = btrfs_root_used(&root->root_item) - blocksize;
+	btrfs_set_root_used(&root->root_item, used);
+	spin_unlock(&root->node_lock);
+
+	return btrfs_free_extent(trans, root, bytenr, blocksize,
+				 parent, root_objectid, level, 0);
+}
+
 static u64 stripe_align(struct btrfs_root *root, u64 val)
 {
 	u64 mask = ((u64)root->stripesize - 1);
@@ -4897,6 +4904,14 @@ static int alloc_tree_block(struct btrfs_trans_handle *trans,
 					extent_op);
 		BUG_ON(ret);
 	}
+
+	if (root_objectid == root->root_key.objectid) {
+		u64 used;
+		spin_lock(&root->node_lock);
+		used = btrfs_root_used(&root->root_item) + num_bytes;
+		btrfs_set_root_used(&root->root_item, used);
+		spin_unlock(&root->node_lock);
+	}
 	return ret;
 }
 
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 3d6b33871af..645a17927a8 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -289,7 +289,7 @@ static noinline int create_subvol(struct btrfs_root *root,
 	btrfs_set_root_generation(&root_item, trans->transid);
 	btrfs_set_root_level(&root_item, 0);
 	btrfs_set_root_refs(&root_item, 1);
-	btrfs_set_root_used(&root_item, 0);
+	btrfs_set_root_used(&root_item, leaf->len);
 	btrfs_set_root_last_snapshot(&root_item, 0);
 
 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 75b31caade2..b2acc79f1b3 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -501,13 +501,16 @@ static int update_cowonly_root(struct btrfs_trans_handle *trans,
 {
 	int ret;
 	u64 old_root_bytenr;
+	u64 old_root_used;
 	struct btrfs_root *tree_root = root->fs_info->tree_root;
 
+	old_root_used = btrfs_root_used(&root->root_item);
 	btrfs_write_dirty_block_groups(trans, root);
 
 	while (1) {
 		old_root_bytenr = btrfs_root_bytenr(&root->root_item);
-		if (old_root_bytenr == root->node->start)
+		if (old_root_bytenr == root->node->start &&
+		    old_root_used == btrfs_root_used(&root->root_item))
 			break;
 
 		btrfs_set_root_node(&root->root_item, root->node);
@@ -516,6 +519,7 @@ static int update_cowonly_root(struct btrfs_trans_handle *trans,
 					&root->root_item);
 		BUG_ON(ret);
 
+		old_root_used = btrfs_root_used(&root->root_item);
 		ret = btrfs_write_dirty_block_groups(trans, root);
 		BUG_ON(ret);
 	}