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>
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 1c8b019..917bf10 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -48,7 +48,6 @@
#include "ordered-data.h"
#include "xattr.h"
#include "tree-log.h"
-#include "ref-cache.h"
#include "compression.h"
#include "locking.h"
@@ -944,6 +943,7 @@
u64 cow_start;
u64 cur_offset;
u64 extent_end;
+ u64 extent_offset;
u64 disk_bytenr;
u64 num_bytes;
int extent_type;
@@ -1005,6 +1005,7 @@
if (extent_type == BTRFS_FILE_EXTENT_REG ||
extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+ extent_offset = btrfs_file_extent_offset(leaf, fi);
extent_end = found_key.offset +
btrfs_file_extent_num_bytes(leaf, fi);
if (extent_end <= start) {
@@ -1022,9 +1023,10 @@
if (btrfs_extent_readonly(root, disk_bytenr))
goto out_check;
if (btrfs_cross_ref_exist(trans, root, inode->i_ino,
- disk_bytenr))
+ found_key.offset -
+ extent_offset, disk_bytenr))
goto out_check;
- disk_bytenr += btrfs_file_extent_offset(leaf, fi);
+ disk_bytenr += extent_offset;
disk_bytenr += cur_offset - found_key.offset;
num_bytes = min(end + 1, extent_end) - cur_offset;
/*
@@ -1489,9 +1491,9 @@
ins.objectid = disk_bytenr;
ins.offset = disk_num_bytes;
ins.type = BTRFS_EXTENT_ITEM_KEY;
- ret = btrfs_alloc_reserved_extent(trans, root, leaf->start,
- root->root_key.objectid,
- trans->transid, inode->i_ino, &ins);
+ ret = btrfs_alloc_reserved_file_extent(trans, root,
+ root->root_key.objectid,
+ inode->i_ino, file_pos, &ins);
BUG_ON(ret);
btrfs_free_path(path);
@@ -1956,23 +1958,13 @@
* crossing root thing. we store the inode number in the
* offset of the orphan item.
*/
- inode = btrfs_iget_locked(root->fs_info->sb,
- found_key.offset, root);
- if (!inode)
+ found_key.objectid = found_key.offset;
+ found_key.type = BTRFS_INODE_ITEM_KEY;
+ found_key.offset = 0;
+ inode = btrfs_iget(root->fs_info->sb, &found_key, root);
+ if (IS_ERR(inode))
break;
- if (inode->i_state & I_NEW) {
- BTRFS_I(inode)->root = root;
-
- /* have to set the location manually */
- BTRFS_I(inode)->location.objectid = inode->i_ino;
- BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
- BTRFS_I(inode)->location.offset = 0;
-
- btrfs_read_locked_inode(inode);
- unlock_new_inode(inode);
- }
-
/*
* add this inode to the orphan list so btrfs_orphan_del does
* the proper thing when we hit it
@@ -2069,7 +2061,7 @@
/*
* read an inode from the btree into the in-memory inode
*/
-void btrfs_read_locked_inode(struct inode *inode)
+static void btrfs_read_locked_inode(struct inode *inode)
{
struct btrfs_path *path;
struct extent_buffer *leaf;
@@ -2599,9 +2591,8 @@
struct btrfs_file_extent_item *fi;
u64 extent_start = 0;
u64 extent_num_bytes = 0;
+ u64 extent_offset = 0;
u64 item_end = 0;
- u64 root_gen = 0;
- u64 root_owner = 0;
int found_extent;
int del_item;
int pending_del_nr = 0;
@@ -2716,6 +2707,9 @@
extent_num_bytes =
btrfs_file_extent_disk_num_bytes(leaf,
fi);
+ extent_offset = found_key.offset -
+ btrfs_file_extent_offset(leaf, fi);
+
/* FIXME blocksize != 4096 */
num_dec = btrfs_file_extent_num_bytes(leaf, fi);
if (extent_start != 0) {
@@ -2723,8 +2717,6 @@
if (root->ref_cows)
inode_sub_bytes(inode, num_dec);
}
- root_gen = btrfs_header_generation(leaf);
- root_owner = btrfs_header_owner(leaf);
}
} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
/*
@@ -2768,12 +2760,12 @@
} else {
break;
}
- if (found_extent) {
+ if (found_extent && root->ref_cows) {
btrfs_set_path_blocking(path);
ret = btrfs_free_extent(trans, root, extent_start,
- extent_num_bytes,
- leaf->start, root_owner,
- root_gen, inode->i_ino, 0);
+ extent_num_bytes, 0,
+ btrfs_header_owner(leaf),
+ inode->i_ino, extent_offset);
BUG_ON(ret);
}
next:
@@ -3105,6 +3097,45 @@
return 0;
}
+static void inode_tree_add(struct inode *inode)
+{
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct btrfs_inode *entry;
+ struct rb_node **p = &root->inode_tree.rb_node;
+ struct rb_node *parent = NULL;
+
+ spin_lock(&root->inode_lock);
+ while (*p) {
+ parent = *p;
+ entry = rb_entry(parent, struct btrfs_inode, rb_node);
+
+ if (inode->i_ino < entry->vfs_inode.i_ino)
+ p = &(*p)->rb_left;
+ else if (inode->i_ino > entry->vfs_inode.i_ino)
+ p = &(*p)->rb_right;
+ else {
+ WARN_ON(!(entry->vfs_inode.i_state &
+ (I_WILL_FREE | I_FREEING | I_CLEAR)));
+ break;
+ }
+ }
+ rb_link_node(&BTRFS_I(inode)->rb_node, parent, p);
+ rb_insert_color(&BTRFS_I(inode)->rb_node, &root->inode_tree);
+ spin_unlock(&root->inode_lock);
+}
+
+static void inode_tree_del(struct inode *inode)
+{
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+
+ if (!RB_EMPTY_NODE(&BTRFS_I(inode)->rb_node)) {
+ spin_lock(&root->inode_lock);
+ rb_erase(&BTRFS_I(inode)->rb_node, &root->inode_tree);
+ spin_unlock(&root->inode_lock);
+ RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node);
+ }
+}
+
static noinline void init_btrfs_i(struct inode *inode)
{
struct btrfs_inode *bi = BTRFS_I(inode);
@@ -3130,6 +3161,7 @@
inode->i_mapping, GFP_NOFS);
INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
INIT_LIST_HEAD(&BTRFS_I(inode)->ordered_operations);
+ RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node);
btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
mutex_init(&BTRFS_I(inode)->extent_mutex);
mutex_init(&BTRFS_I(inode)->log_mutex);
@@ -3152,26 +3184,9 @@
args->root == BTRFS_I(inode)->root;
}
-struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
- struct btrfs_root *root, int wait)
-{
- struct inode *inode;
- struct btrfs_iget_args args;
- args.ino = objectid;
- args.root = root;
-
- if (wait) {
- inode = ilookup5(s, objectid, btrfs_find_actor,
- (void *)&args);
- } else {
- inode = ilookup5_nowait(s, objectid, btrfs_find_actor,
- (void *)&args);
- }
- return inode;
-}
-
-struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
- struct btrfs_root *root)
+static struct inode *btrfs_iget_locked(struct super_block *s,
+ u64 objectid,
+ struct btrfs_root *root)
{
struct inode *inode;
struct btrfs_iget_args args;
@@ -3188,24 +3203,21 @@
* Returns in *is_new if the inode was read from disk
*/
struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location,
- struct btrfs_root *root, int *is_new)
+ struct btrfs_root *root)
{
struct inode *inode;
inode = btrfs_iget_locked(s, location->objectid, root);
if (!inode)
- return ERR_PTR(-EACCES);
+ return ERR_PTR(-ENOMEM);
if (inode->i_state & I_NEW) {
BTRFS_I(inode)->root = root;
memcpy(&BTRFS_I(inode)->location, location, sizeof(*location));
btrfs_read_locked_inode(inode);
+
+ inode_tree_add(inode);
unlock_new_inode(inode);
- if (is_new)
- *is_new = 1;
- } else {
- if (is_new)
- *is_new = 0;
}
return inode;
@@ -3218,7 +3230,7 @@
struct btrfs_root *root = bi->root;
struct btrfs_root *sub_root = root;
struct btrfs_key location;
- int ret, new;
+ int ret;
if (dentry->d_name.len > BTRFS_NAME_LEN)
return ERR_PTR(-ENAMETOOLONG);
@@ -3236,7 +3248,7 @@
return ERR_PTR(ret);
if (ret > 0)
return ERR_PTR(-ENOENT);
- inode = btrfs_iget(dir->i_sb, &location, sub_root, &new);
+ inode = btrfs_iget(dir->i_sb, &location, sub_root);
if (IS_ERR(inode))
return ERR_CAST(inode);
}
@@ -3631,6 +3643,7 @@
btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
insert_inode_hash(inode);
+ inode_tree_add(inode);
return inode;
fail:
if (dir)
@@ -4683,6 +4696,7 @@
btrfs_put_ordered_extent(ordered);
}
}
+ inode_tree_del(inode);
btrfs_drop_extent_cache(inode, 0, (u64)-1, 0);
kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
}