Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE)

This commit introduces a new kind of back reference for btrfs metadata.
Once a filesystem has been mounted with this commit, IT WILL NO LONGER
BE MOUNTABLE BY OLDER KERNELS.

When a tree block in subvolume tree is cow'd, the reference counts of all
extents it points to are increased by one.  At transaction commit time,
the old root of the subvolume is recorded in a "dead root" data structure,
and the btree it points to is later walked, dropping reference counts
and freeing any blocks where the reference count goes to 0.

The increments done during cow and decrements done after commit cancel out,
and the walk is a very expensive way to go about freeing the blocks that
are no longer referenced by the new btree root.  This commit reduces the
transaction overhead by avoiding the need for dead root records.

When a non-shared tree block is cow'd, we free the old block at once, and the
new block inherits old block's references. When a tree block with reference
count > 1 is cow'd, we increase the reference counts of all extents
the new block points to by one, and decrease the old block's reference count by
one.

This dead tree avoidance code removes the need to modify the reference
counts of lower level extents when a non-shared tree block is cow'd.
But we still need to update back ref for all pointers in the block.
This is because the location of the block is recorded in the back ref
item.

We can solve this by introducing a new type of back ref. The new
back ref provides information about pointer's key, level and in which
tree the pointer lives. This information allow us to find the pointer
by searching the tree. The shortcoming of the new back ref is that it
only works for pointers in tree blocks referenced by their owner trees.

This is mostly a problem for snapshots, where resolving one of these
fuzzy back references would be O(number_of_snapshots) and quite slow.
The solution used here is to use the fuzzy back references in the common
case where a given tree block is only referenced by one root,
and use the full back references when multiple roots have a reference
on a given block.

This commit adds per subvolume red-black tree to keep trace of cached
inodes. The red-black tree helps the balancing code to find cached
inodes whose inode numbers within a given range.

This commit improves the balancing code by introducing several data
structures to keep the state of balancing. The most important one
is the back ref cache. It caches how the upper level tree blocks are
referenced. This greatly reduce the overhead of checking back ref.

The improved balancing code scales significantly better with a large
number of snapshots.

This is a very large commit and was written in a number of
pieces.  But, they depend heavily on the disk format change and were
squashed together to make sure git bisect didn't end up in a
bad state wrt space balancing or the format change.

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

1 files changed, 50 insertions, 45 deletions

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 4b0ea0b..7f5c6e3 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -36,7 +36,6 @@
 #include "print-tree.h"
 #include "async-thread.h"
 #include "locking.h"
-#include "ref-cache.h"
 #include "tree-log.h"
 #include "free-space-cache.h"
 
@@ -884,7 +883,6 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 {
 	root->node = NULL;
 	root->commit_root = NULL;
-	root->ref_tree = NULL;
 	root->sectorsize = sectorsize;
 	root->nodesize = nodesize;
 	root->leafsize = leafsize;
@@ -899,12 +897,14 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	root->last_inode_alloc = 0;
 	root->name = NULL;
 	root->in_sysfs = 0;
+	root->inode_tree.rb_node = NULL;
 
 	INIT_LIST_HEAD(&root->dirty_list);
 	INIT_LIST_HEAD(&root->orphan_list);
-	INIT_LIST_HEAD(&root->dead_list);
+	INIT_LIST_HEAD(&root->root_list);
 	spin_lock_init(&root->node_lock);
 	spin_lock_init(&root->list_lock);
+	spin_lock_init(&root->inode_lock);
 	mutex_init(&root->objectid_mutex);
 	mutex_init(&root->log_mutex);
 	init_waitqueue_head(&root->log_writer_wait);
@@ -918,9 +918,6 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	extent_io_tree_init(&root->dirty_log_pages,
 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
 
-	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));
@@ -959,6 +956,7 @@ static int find_and_setup_root(struct btrfs_root *tree_root,
 	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
 	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
 				     blocksize, generation);
+	root->commit_root = btrfs_root_node(root);
 	BUG_ON(!root->node);
 	return 0;
 }
@@ -1025,20 +1023,19 @@ static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
 	 */
 	root->ref_cows = 0;
 
-	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
-				      0, BTRFS_TREE_LOG_OBJECTID,
-				      trans->transid, 0, 0, 0);
+	leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
+				      BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0);
 	if (IS_ERR(leaf)) {
 		kfree(root);
 		return ERR_CAST(leaf);
 	}
 
+	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
+	btrfs_set_header_bytenr(leaf, leaf->start);
+	btrfs_set_header_generation(leaf, trans->transid);
+	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
+	btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
 	root->node = leaf;
-	btrfs_set_header_nritems(root->node, 0);
-	btrfs_set_header_level(root->node, 0);
-	btrfs_set_header_bytenr(root->node, root->node->start);
-	btrfs_set_header_generation(root->node, trans->transid);
-	btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID);
 
 	write_extent_buffer(root->node, root->fs_info->fsid,
 			    (unsigned long)btrfs_header_fsid(root->node),
@@ -1081,8 +1078,7 @@ int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
 	inode_item->nbytes = cpu_to_le64(root->leafsize);
 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
 
-	btrfs_set_root_bytenr(&log_root->root_item, log_root->node->start);
-	btrfs_set_root_generation(&log_root->root_item, trans->transid);
+	btrfs_set_root_node(&log_root->root_item, log_root->node);
 
 	WARN_ON(root->log_root);
 	root->log_root = log_root;
@@ -1144,6 +1140,7 @@ out:
 	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
 	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
 				     blocksize, generation);
+	root->commit_root = btrfs_root_node(root);
 	BUG_ON(!root->node);
 insert:
 	if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
@@ -1210,7 +1207,7 @@ struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
 	}
 	if (!(fs_info->sb->s_flags & MS_RDONLY)) {
 		ret = btrfs_find_dead_roots(fs_info->tree_root,
-					    root->root_key.objectid, root);
+					    root->root_key.objectid);
 		BUG_ON(ret);
 		btrfs_orphan_cleanup(root);
 	}
@@ -1569,8 +1566,6 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	atomic_set(&fs_info->async_delalloc_pages, 0);
 	atomic_set(&fs_info->async_submit_draining, 0);
 	atomic_set(&fs_info->nr_async_bios, 0);
-	atomic_set(&fs_info->throttles, 0);
-	atomic_set(&fs_info->throttle_gen, 0);
 	fs_info->sb = sb;
 	fs_info->max_extent = (u64)-1;
 	fs_info->max_inline = 8192 * 1024;
@@ -1598,6 +1593,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
 	fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
 
+	RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
 	extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
 			     fs_info->btree_inode->i_mapping,
 			     GFP_NOFS);
@@ -1613,10 +1609,6 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
 	fs_info->do_barriers = 1;
 
-	INIT_LIST_HEAD(&fs_info->dead_reloc_roots);
-	btrfs_leaf_ref_tree_init(&fs_info->reloc_ref_tree);
-	btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree);
-
 	BTRFS_I(fs_info->btree_inode)->root = tree_root;
 	memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
 	       sizeof(struct btrfs_key));
@@ -1674,6 +1666,12 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 		goto fail_iput;
 	}
 
+	features = btrfs_super_incompat_flags(disk_super);
+	if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) {
+		features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
+		btrfs_set_super_incompat_flags(disk_super, features);
+	}
+
 	features = btrfs_super_compat_ro_flags(disk_super) &
 		~BTRFS_FEATURE_COMPAT_RO_SUPP;
 	if (!(sb->s_flags & MS_RDONLY) && features) {
@@ -1771,7 +1769,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	if (ret) {
 		printk(KERN_WARNING "btrfs: failed to read the system "
 		       "array on %s\n", sb->s_id);
-		goto fail_sys_array;
+		goto fail_sb_buffer;
 	}
 
 	blocksize = btrfs_level_size(tree_root,
@@ -1785,6 +1783,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 					   btrfs_super_chunk_root(disk_super),
 					   blocksize, generation);
 	BUG_ON(!chunk_root->node);
+	btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
+	chunk_root->commit_root = btrfs_root_node(chunk_root);
 
 	read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
 	   (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
@@ -1810,7 +1810,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 					  blocksize, generation);
 	if (!tree_root->node)
 		goto fail_chunk_root;
-
+	btrfs_set_root_node(&tree_root->root_item, tree_root->node);
+	tree_root->commit_root = btrfs_root_node(tree_root);
 
 	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
@@ -1820,14 +1821,14 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 
 	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_DEV_TREE_OBJECTID, dev_root);
-	dev_root->track_dirty = 1;
 	if (ret)
 		goto fail_extent_root;
+	dev_root->track_dirty = 1;
 
 	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_CSUM_TREE_OBJECTID, csum_root);
 	if (ret)
-		goto fail_extent_root;
+		goto fail_dev_root;
 
 	csum_root->track_dirty = 1;
 
@@ -1881,7 +1882,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	}
 
 	if (!(sb->s_flags & MS_RDONLY)) {
-		ret = btrfs_cleanup_reloc_trees(tree_root);
+		ret = btrfs_recover_relocation(tree_root);
 		BUG_ON(ret);
 	}
 
@@ -1908,14 +1909,19 @@ fail_cleaner:
 
 fail_csum_root:
 	free_extent_buffer(csum_root->node);
+	free_extent_buffer(csum_root->commit_root);
+fail_dev_root:
+	free_extent_buffer(dev_root->node);
+	free_extent_buffer(dev_root->commit_root);
 fail_extent_root:
 	free_extent_buffer(extent_root->node);
+	free_extent_buffer(extent_root->commit_root);
 fail_tree_root:
 	free_extent_buffer(tree_root->node);
+	free_extent_buffer(tree_root->commit_root);
 fail_chunk_root:
 	free_extent_buffer(chunk_root->node);
-fail_sys_array:
-	free_extent_buffer(dev_root->node);
+	free_extent_buffer(chunk_root->commit_root);
 fail_sb_buffer:
 	btrfs_stop_workers(&fs_info->fixup_workers);
 	btrfs_stop_workers(&fs_info->delalloc_workers);
@@ -2173,6 +2179,7 @@ int write_ctree_super(struct btrfs_trans_handle *trans,
 
 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
 {
+	WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
 	radix_tree_delete(&fs_info->fs_roots_radix,
 			  (unsigned long)root->root_key.objectid);
 	if (root->anon_super.s_dev) {
@@ -2219,10 +2226,12 @@ int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
 					     ARRAY_SIZE(gang));
 		if (!ret)
 			break;
+
+		root_objectid = gang[ret - 1]->root_key.objectid + 1;
 		for (i = 0; i < ret; i++) {
 			root_objectid = gang[i]->root_key.objectid;
 			ret = btrfs_find_dead_roots(fs_info->tree_root,
-						    root_objectid, gang[i]);
+						    root_objectid);
 			BUG_ON(ret);
 			btrfs_orphan_cleanup(gang[i]);
 		}
@@ -2278,20 +2287,16 @@ int close_ctree(struct btrfs_root *root)
 		       (unsigned long long)fs_info->total_ref_cache_size);
 	}
 
-	if (fs_info->extent_root->node)
-		free_extent_buffer(fs_info->extent_root->node);
-
-	if (fs_info->tree_root->node)
-		free_extent_buffer(fs_info->tree_root->node);
-
-	if (root->fs_info->chunk_root->node)
-		free_extent_buffer(root->fs_info->chunk_root->node);
-
-	if (root->fs_info->dev_root->node)
-		free_extent_buffer(root->fs_info->dev_root->node);
-
-	if (root->fs_info->csum_root->node)
-		free_extent_buffer(root->fs_info->csum_root->node);
+	free_extent_buffer(fs_info->extent_root->node);
+	free_extent_buffer(fs_info->extent_root->commit_root);
+	free_extent_buffer(fs_info->tree_root->node);
+	free_extent_buffer(fs_info->tree_root->commit_root);
+	free_extent_buffer(root->fs_info->chunk_root->node);
+	free_extent_buffer(root->fs_info->chunk_root->commit_root);
+	free_extent_buffer(root->fs_info->dev_root->node);
+	free_extent_buffer(root->fs_info->dev_root->commit_root);
+	free_extent_buffer(root->fs_info->csum_root->node);
+	free_extent_buffer(root->fs_info->csum_root->commit_root);
 
 	btrfs_free_block_groups(root->fs_info);