Size reduction of btrfs_path

PATCH New David Sterba 1 participant 5 messages Nov 27, 2015 Btrfs file system

Share & tools

David Sterba

author maintainer

2015-11-27 15:33 UTC #0 lore raw

cover Size reduction of btrfs_path

We can reduce size of btrfs_path by 32 bytes, which will lead to more objects packed into one slab page. Performance should not be worse and could even improve in some cases due to less cachelines used.
Targetting 4.5. Thanks.

Can be pulled from

git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux.git dev/path-diet

David Sterba (4):
btrfs: cleanup, use enum values for btrfs_path reada
btrfs: use smaller type for btrfs_path reada
btrfs: use smaller type for btrfs_path lowest_level
btrfs: use smaller type for btrfs_path locks

 fs/btrfs/ctree.c       |  9 ++++-----
 fs/btrfs/ctree.h       |  7 ++++---
 fs/btrfs/extent-tree.c | 12 ++++++------
 fs/btrfs/file-item.c   |  4 ++--
 fs/btrfs/inode-map.c   |  2 +-
 fs/btrfs/inode.c       |  8 ++++----
 fs/btrfs/ioctl.c       |  2 +-
 fs/btrfs/relocation.c  | 12 ++++++------
 fs/btrfs/scrub.c       |  2 +-
 fs/btrfs/volumes.c     |  6 +++---
 fs/btrfs/xattr.c       |  2 +-
 11 files changed, 33 insertions(+), 33 deletions(-)

Signature · 6 lines

2.6.2

--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-

David Sterba

author maintainer

2015-11-27 15:33 UTC #1 lore raw

PATCH 1/4 btrfs: cleanup, use enum values for btrfs_path reada

Replace the integers by enums for better readability. The value 2 does not have any meaning since a717531942f488209dded30f6bc648167bcefa72 "Btrfs: do less aggressive btree readahead" (2009-01-22).

Signed-off-by David Sterba <[email protected]>

---
 fs/btrfs/ctree.c       |  9 ++++-----
 fs/btrfs/ctree.h       |  1 +
 fs/btrfs/extent-tree.c | 12 ++++++------
 fs/btrfs/file-item.c   |  4 ++--
 fs/btrfs/inode-map.c   |  2 +-
 fs/btrfs/inode.c       |  8 ++++----
 fs/btrfs/ioctl.c       |  2 +-
 fs/btrfs/relocation.c  | 12 ++++++------
 fs/btrfs/scrub.c       |  2 +-
 fs/btrfs/volumes.c     |  6 +++---
 fs/btrfs/xattr.c       |  2 +-
 11 files changed, 30 insertions(+), 30 deletions(-)

fs/btrfs/ctree.c +4 -5

index 5b8e235c4b6d..be1be0422ff4 100644
@@ -2248,7 +2248,6 @@ static void reada_for_search(struct btrfs_root *root,
 	u64 target;
 	u64 nread = 0;
 	u64 gen;
-	int direction = path->reada;
 	struct extent_buffer *eb;
 	u32 nr;
 	u32 blocksize;

@@ -2276,16 +2275,16 @@ static void reada_for_search(struct btrfs_root *root,
 	nr = slot;
 
 	while (1) {
-		if (direction < 0) {
+		if (path->reada == READA_BACK) {
 			if (nr == 0)
 				break;
 			nr--;
-		} else if (direction > 0) {
+		} else if (path->reada == READA_FORWARD) {
 			nr++;
 			if (nr >= nritems)
 				break;
 		}
-		if (path->reada < 0 && objectid) {
+		if (path->reada == READA_BACK && objectid) {
 			btrfs_node_key(node, &disk_key, nr);
 			if (btrfs_disk_key_objectid(&disk_key) != objectid)
 				break;

@@ -2493,7 +2492,7 @@ read_block_for_search(struct btrfs_trans_handle *trans,
 	btrfs_set_path_blocking(p);
 
 	free_extent_buffer(tmp);
-	if (p->reada)
+	if (p->reada != READA_NONE)
 		reada_for_search(root, p, level, slot, key->objectid);
 
 	btrfs_release_path(p);

fs/btrfs/ctree.h +1 -0

index 8c58191249cc..20579f6aa959 100644
@@ -590,6 +590,7 @@ struct btrfs_node {
  * The slots array records the index of the item or block pointer
  * used while walking the tree.
  */
+enum { READA_NONE = 0, READA_BACK, READA_FORWARD };
 struct btrfs_path {
 	struct extent_buffer *nodes[BTRFS_MAX_LEVEL];
 	int slots[BTRFS_MAX_LEVEL];

fs/btrfs/extent-tree.c +6 -6

index acf3ed11cfb6..45fe1ffcdeac 100644
@@ -438,7 +438,7 @@ static noinline void caching_thread(struct btrfs_work *work)
 	 */
 	path->skip_locking = 1;
 	path->search_commit_root = 1;
-	path->reada = 1;
+	path->reada = READA_FORWARD;
 
 	key.objectid = last;
 	key.offset = 0;

@@ -2115,7 +2115,7 @@ static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 	if (!path)
 		return -ENOMEM;
 
-	path->reada = 1;
+	path->reada = READA_FORWARD;
 	path->leave_spinning = 1;
 	/* this will setup the path even if it fails to insert the back ref */
 	ret = insert_inline_extent_backref(trans, fs_info->extent_root, path,

@@ -2141,7 +2141,7 @@ static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 	btrfs_mark_buffer_dirty(leaf);
 	btrfs_release_path(path);
 
-	path->reada = 1;
+	path->reada = READA_FORWARD;
 	path->leave_spinning = 1;
 	/* now insert the actual backref */
 	ret = insert_extent_backref(trans, root->fs_info->extent_root,

@@ -2254,7 +2254,7 @@ static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
 	}
 
 again:
-	path->reada = 1;
+	path->reada = READA_FORWARD;
 	path->leave_spinning = 1;
 	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key,
 				path, 0, 1);

@@ -6435,7 +6435,7 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 	if (!path)
 		return -ENOMEM;
 
-	path->reada = 1;
+	path->reada = READA_FORWARD;
 	path->leave_spinning = 1;
 
 	is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID;

@@ -9655,7 +9655,7 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
-	path->reada = 1;
+	path->reada = READA_FORWARD;
 
 	cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
 	if (btrfs_test_opt(root, SPACE_CACHE) &&

fs/btrfs/file-item.c +2 -2

index 58ece6558430..a67e1c828d0f 100644
@@ -202,7 +202,7 @@ static int __btrfs_lookup_bio_sums(struct btrfs_root *root,
 	}
 
 	if (bio->bi_iter.bi_size > PAGE_CACHE_SIZE * 8)
-		path->reada = 2;
+		path->reada = READA_FORWARD;
 
 	WARN_ON(bio->bi_vcnt <= 0);

@@ -328,7 +328,7 @@ int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end,
 
 	if (search_commit) {
 		path->skip_locking = 1;
-		path->reada = 2;
+		path->reada = READA_FORWARD;
 		path->search_commit_root = 1;
 	}

fs/btrfs/inode-map.c +1 -1

index 767a6056ac45..f8bea2616ffb 100644
@@ -48,7 +48,7 @@ static int caching_kthread(void *data)
 	/* Since the commit root is read-only, we can safely skip locking. */
 	path->skip_locking = 1;
 	path->search_commit_root = 1;
-	path->reada = 2;
+	path->reada = READA_FORWARD;
 
 	key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 	key.offset = 0;

fs/btrfs/inode.c +4 -4

index 994490d5fa64..e3e56cec5ad7 100644
@@ -3351,7 +3351,7 @@ int btrfs_orphan_cleanup(struct btrfs_root *root)
 		ret = -ENOMEM;
 		goto out;
 	}
-	path->reada = -1;
+	path->reada = READA_BACK;
 
 	key.objectid = BTRFS_ORPHAN_OBJECTID;
 	key.type = BTRFS_ORPHAN_ITEM_KEY;

@@ -4339,7 +4339,7 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
-	path->reada = -1;
+	path->reada = READA_BACK;
 
 	/*
 	 * We want to drop from the next block forward in case this new size is

@@ -5775,7 +5775,7 @@ static int btrfs_real_readdir(struct file *file, struct dir_context *ctx)
 	if (!path)
 		return -ENOMEM;
 
-	path->reada = 1;
+	path->reada = READA_FORWARD;
 
 	if (key_type == BTRFS_DIR_INDEX_KEY) {
 		INIT_LIST_HEAD(&ins_list);

@@ -6806,7 +6806,7 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
 		 * Chances are we'll be called again, so go ahead and do
 		 * readahead
 		 */
-		path->reada = 1;
+		path->reada = READA_FORWARD;
 	}
 
 	ret = btrfs_lookup_file_extent(trans, root, path,

fs/btrfs/ioctl.c +1 -1

index da94138eb85e..b8e3ed4013c5 100644
@@ -3478,7 +3478,7 @@ static int btrfs_clone(struct inode *src, struct inode *inode,
 		return ret;
 	}
 
-	path->reada = 2;
+	path->reada = READA_FORWARD;
 	/* clone data */
 	key.objectid = btrfs_ino(src);
 	key.type = BTRFS_EXTENT_DATA_KEY;

fs/btrfs/relocation.c +6 -6

index b4ca5454ef1a..ef6d8fc85853 100644
@@ -708,8 +708,8 @@ struct backref_node *build_backref_tree(struct reloc_control *rc,
 		err = -ENOMEM;
 		goto out;
 	}
-	path1->reada = 1;
-	path2->reada = 2;
+	path1->reada = READA_FORWARD;
+	path2->reada = READA_FORWARD;
 
 	node = alloc_backref_node(cache);
 	if (!node) {

@@ -2130,7 +2130,7 @@ static noinline_for_stack int merge_reloc_root(struct reloc_control *rc,
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
-	path->reada = 1;
+	path->reada = READA_FORWARD;
 
 	reloc_root = root->reloc_root;
 	root_item = &reloc_root->root_item;

@@ -3527,7 +3527,7 @@ static int find_data_references(struct reloc_control *rc,
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
-	path->reada = 1;
+	path->reada = READA_FORWARD;
 
 	root = read_fs_root(rc->extent_root->fs_info, ref_root);
 	if (IS_ERR(root)) {

@@ -3917,7 +3917,7 @@ static noinline_for_stack int relocate_block_group(struct reloc_control *rc)
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
-	path->reada = 1;
+	path->reada = READA_FORWARD;
 
 	ret = prepare_to_relocate(rc);
 	if (ret) {

@@ -4343,7 +4343,7 @@ int btrfs_recover_relocation(struct btrfs_root *root)
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
-	path->reada = -1;
+	path->reada = READA_BACK;
 
 	key.objectid = BTRFS_TREE_RELOC_OBJECTID;
 	key.type = BTRFS_ROOT_ITEM_KEY;

fs/btrfs/scrub.c +1 -1

index 2907a77fb1f6..df8dc98737f4 100644
@@ -3494,7 +3494,7 @@ int scrub_enumerate_chunks(struct scrub_ctx *sctx,
 	if (!path)
 		return -ENOMEM;
 
-	path->reada = 2;
+	path->reada = READA_FORWARD;
 	path->search_commit_root = 1;
 	path->skip_locking = 1;

fs/btrfs/volumes.c +3 -3

index a6df8fdc1312..750c43df7b79 100644
@@ -1102,7 +1102,7 @@ int btrfs_account_dev_extents_size(struct btrfs_device *device, u64 start,
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
-	path->reada = 2;
+	path->reada = READA_FORWARD;
 
 	key.objectid = device->devid;
 	key.offset = start;

@@ -1271,7 +1271,7 @@ int find_free_dev_extent_start(struct btrfs_transaction *transaction,
 		goto out;
 	}
 
-	path->reada = 2;
+	path->reada = READA_FORWARD;
 	path->search_commit_root = 1;
 	path->skip_locking = 1;

@@ -4270,7 +4270,7 @@ int btrfs_shrink_device(struct btrfs_device *device, u64 new_size)
 	if (!path)
 		return -ENOMEM;
 
-	path->reada = 2;
+	path->reada = READA_FORWARD;
 
 	lock_chunks(root);

fs/btrfs/xattr.c +1 -1

index 1fcd7b6e7564..28b4e7e71e61 100644
@@ -283,7 +283,7 @@ ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size)
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
-	path->reada = 2;
+	path->reada = READA_FORWARD;
 
 	/* search for our xattrs */
 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);

Signature · 6 lines

2.6.2

--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo

David Sterba

author maintainer

2015-11-27 15:33 UTC #2 lore raw

PATCH 2/4 btrfs: use smaller type for btrfs_path reada

In reply to David Sterba

The possible values for reada are all positive and bounded, we can later save some bytes by storing it in u8.

Signed-off-by David Sterba <[email protected]>

---
 fs/btrfs/ctree.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

fs/btrfs/ctree.h +1 -1

index 20579f6aa959..64c1d96d9bf9 100644
@@ -596,7 +596,7 @@ struct btrfs_path {
 	int slots[BTRFS_MAX_LEVEL];
 	/* if there is real range locking, this locks field will change */
 	int locks[BTRFS_MAX_LEVEL];
-	int reada;
+	u8 reada;
 	/* keep some upper locks as we walk down */
 	int lowest_level;

Signature · 6 lines

2.6.2

--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-

David Sterba

author maintainer

2015-11-27 15:33 UTC #3 lore raw

PATCH 3/4 btrfs: use smaller type for btrfs_path lowest_level

In reply to David Sterba

The level is 0..7, we can use smaller type. The size of btrfs_path is now 136 bytes from 144, which is +2 objects that fit into a 4k slab.

Signed-off-by David Sterba <[email protected]>

---
 fs/btrfs/ctree.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

fs/btrfs/ctree.h +1 -1

index 64c1d96d9bf9..4e00412dc5ff 100644
@@ -598,7 +598,7 @@ struct btrfs_path {
 	int locks[BTRFS_MAX_LEVEL];
 	u8 reada;
 	/* keep some upper locks as we walk down */
-	int lowest_level;
+	u8 lowest_level;
 
 	/*
 	 * set by btrfs_split_item, tells search_slot to keep all locks

Signature · 6 lines

2.6.2

--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-

David Sterba

author maintainer

2015-11-27 15:33 UTC #4 lore raw

PATCH 4/4 btrfs: use smaller type for btrfs_path locks

In reply to David Sterba

The values of btrfs_path::locks are 0 to 4, fit into a u8. Let's see:

* overall size of btrfs_path drops down from 136 to 112 (-24 bytes),
* better packing in a slab page +6 objects
* the whole structure now fits to 2 cachelines
* slight decrease in code size:

text data bss dec hex filename

 938731   43670   23144 1005545   f57e9 fs/btrfs/btrfs.ko.before
 938203   43670   23144 1005017   f55d9 fs/btrfs/btrfs.ko.after

(and the generated assembly does not change much)

The main purpose is to decrease the size of the structure without affecting performance. The byte access is usually well behaving accross arches, the locks are not accessed frequently and sometimes just compared to zero.

Note for further size reduction attempts: the slots could be made u16 but this might generate worse code on some arches (non-byte and non-int access). Also the range of operations on slots is wider compared to locks and the potential performance drop should be evaluated first.

Signed-off-by David Sterba <[email protected]>

---
 fs/btrfs/ctree.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

fs/btrfs/ctree.h +1 -1

index 4e00412dc5ff..caa8d53912a7 100644
@@ -595,7 +595,7 @@ struct btrfs_path {
 	struct extent_buffer *nodes[BTRFS_MAX_LEVEL];
 	int slots[BTRFS_MAX_LEVEL];
 	/* if there is real range locking, this locks field will change */
-	int locks[BTRFS_MAX_LEVEL];
+	u8 locks[BTRFS_MAX_LEVEL];
 	u8 reada;
 	/* keep some upper locks as we walk down */
 	u8 lowest_level;

Signature · 6 lines

2.6.2

--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-