Btrfs: avoid sleeping in verify_parent_transid while atomic

verify_parent_transid needs to lock the extent range to make sure no IO is underway, and so it can safely clear the uptodate bits if our checks fail. But, a few callers are using it with spinlocks held. Most of the time, the generation numbers are going to match, and we don't want to switch to a blocking lock just for the error case. This adds an atomic flag to verify_parent_transid, and changes it to return EAGAIN if it needs to block to properly verifiy things.

Chris Mason committed 14 years ago

Commit b9fab911 parent ea9947b5 files · +34 −17Browse files

Unified · Split

5 files

fs/btrfs/ctree.c +17 −9
fs/btrfs/disk-io.c +13 −5
fs/btrfs/disk-io.h +2 −1
fs/btrfs/extent-tree.c +1 −1
fs/btrfs/tree-log.c +1 −1

fs/btrfs/ctree.c+17−9

 		cur = btrfs_find_tree_block(root, blocknr, blocksize);
 		if (cur)
-			uptodate = btrfs_buffer_uptodate(cur, gen);
+			uptodate = btrfs_buffer_uptodate(cur, gen, 0);
 		else
 			uptodate = 0;
 		if (!cur || !uptodate) {
 		block1 = btrfs_node_blockptr(parent, slot - 1);
 		gen = btrfs_node_ptr_generation(parent, slot - 1);
 		eb = btrfs_find_tree_block(root, block1, blocksize);
-		if (eb && btrfs_buffer_uptodate(eb, gen))
+		/*
+		 * if we get -eagain from btrfs_buffer_uptodate, we
+		 * don't want to return eagain here.  That will loop
+		 * forever
+		 */
+		if (eb && btrfs_buffer_uptodate(eb, gen, 1) != 0)
 			block1 = 0;
 		free_extent_buffer(eb);
 	}
 		block2 = btrfs_node_blockptr(parent, slot + 1);
 		gen = btrfs_node_ptr_generation(parent, slot + 1);
 		eb = btrfs_find_tree_block(root, block2, blocksize);
-		if (eb && btrfs_buffer_uptodate(eb, gen))
+		if (eb && btrfs_buffer_uptodate(eb, gen, 1) != 0)
 			block2 = 0;
 		free_extent_buffer(eb);
 	}
 	tmp = btrfs_find_tree_block(root, blocknr, blocksize);
 	if (tmp) {
-		if (btrfs_buffer_uptodate(tmp, 0)) {
+		/* first we do an atomic uptodate check */
-			if (btrfs_buffer_uptodate(tmp, gen)) {
+		if (btrfs_buffer_uptodate(tmp, 0, 1) > 0) {
+			if (btrfs_buffer_uptodate(tmp, gen, 1) > 0) {
 				/*
 				 * we found an up to date block without
 				 * sleeping, return
 			free_extent_buffer(tmp);
 			btrfs_set_path_blocking(p);
+			/* now we're allowed to do a blocking uptodate check */
 			tmp = read_tree_block(root, blocknr, blocksize, gen);
-			if (tmp && btrfs_buffer_uptodate(tmp, gen)) {
+			if (tmp && btrfs_buffer_uptodate(tmp, gen, 0) > 0) {
 				*eb_ret = tmp;
 				return 0;
 			}
 		 * and give up so that our caller doesn't loop forever
 		 * on our EAGAINs.
 		 */
-		if (!btrfs_buffer_uptodate(tmp, 0))
+		if (!btrfs_buffer_uptodate(tmp, 0, 0))
 			ret = -EIO;
 		free_extent_buffer(tmp);
 	}
 			tmp = btrfs_find_tree_block(root, blockptr,
 					    btrfs_level_size(root, level - 1));
-			if (tmp && btrfs_buffer_uptodate(tmp, gen)) {
+			if (tmp && btrfs_buffer_uptodate(tmp, gen, 1) > 0) {
 				free_extent_buffer(tmp);
 				break;
 			}
 				struct extent_buffer *cur;
 				cur = btrfs_find_tree_block(root, blockptr,
 					    btrfs_level_size(root, level - 1));
-				if (!cur || !btrfs_buffer_uptodate(cur, gen)) {
+				if (!cur ||
+				    btrfs_buffer_uptodate(cur, gen, 1) <= 0) {
 					slot++;
 					if (cur)
 						free_extent_buffer(cur);

fs/btrfs/disk-io.c+13−5

  * in the wrong place.
  */
 static int verify_parent_transid(struct extent_io_tree *io_tree,
-				 struct extent_buffer *eb, u64 parent_transid)
+				 struct extent_buffer *eb, u64 parent_transid,
+				 int atomic)
 {
 	struct extent_state *cached_state = NULL;
 	int ret;
 	if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
 		return 0;
+	if (atomic)
+		return -EAGAIN;
 	lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
 			 0, &cached_state);
 	if (extent_buffer_uptodate(eb) &&
 		ret = read_extent_buffer_pages(io_tree, eb, start,
 					       WAIT_COMPLETE,
 					       btree_get_extent, mirror_num);
-		if (!ret && !verify_parent_transid(io_tree, eb, parent_transid))
+		if (!ret && !verify_parent_transid(io_tree, eb,
+						   parent_transid, 0))
 			break;
 		/*
 	root->commit_root = NULL;
 	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
 				     blocksize, generation);
-	if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) {
+	if (!root->node || !btrfs_buffer_uptodate(root->node, generation, 0)) {
 		free_extent_buffer(root->node);
 		root->node = NULL;
 		return -EIO;
 	return 0;
 }
-int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
+int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
+			  int atomic)
 {
 	int ret;
 	struct inode *btree_inode = buf->pages[0]->mapping->host;
 		return ret;
 	ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
-				    parent_transid);
+				    parent_transid, atomic);
+	if (ret == -EAGAIN)
+		return ret;
 	return !ret;
 }

fs/btrfs/disk-io.h+2−1

 void __btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr);
 void btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root);
 void btrfs_mark_buffer_dirty(struct extent_buffer *buf);
-int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid);
+int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
+			  int atomic);
 int btrfs_set_buffer_uptodate(struct extent_buffer *buf);
 int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid);
 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len);

fs/btrfs/extent-tree.c+1−1

@@ -6568,7 +6568,7 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans, · 6567 unmodified lines
6568	goto skip;	6568	goto skip;
6569	}	6569	}
6570		6570
6571	if (!btrfs_buffer_uptodate(next, generation)) {	6571	if (!btrfs_buffer_uptodate(next, generation, 0)) {
6572	btrfs_tree_unlock(next);	6572	btrfs_tree_unlock(next);
6573	free_extent_buffer(next);	6573	free_extent_buffer(next);
6574	next = NULL;	6574	next = NULL;

fs/btrfs/tree-log.c+1−1

@@ -279,7 +279,7 @@ static int process_one_buffer(struct btrfs_root *log, · 278 unmodified lines
279	log->fs_info->extent_root,	279	log->fs_info->extent_root,
280	eb->start, eb->len);	280	eb->start, eb->len);
281		281
282	if (btrfs_buffer_uptodate(eb, gen)) {	282	if (btrfs_buffer_uptodate(eb, gen, 0)) {
283	if (wc->write)	283	if (wc->write)
284	btrfs_write_tree_block(eb);	284	btrfs_write_tree_block(eb);
285	if (wc->wait)	285	if (wc->wait)