1
linux/fs/btrfs/root-tree.c

476 lines
11 KiB
C
Raw Normal View History

/*
* Copyright (C) 2007 Oracle. All rights reserved.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public
* License v2 as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
* You should have received a copy of the GNU General Public
* License along with this program; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 021110-1307, USA.
*/
#include "ctree.h"
#include "transaction.h"
#include "disk-io.h"
#include "print-tree.h"
/*
* search forward for a root, starting with objectid 'search_start'
* if a root key is found, the objectid we find is filled into 'found_objectid'
* and 0 is returned. < 0 is returned on error, 1 if there is nothing
* left in the tree.
*/
int btrfs_search_root(struct btrfs_root *root, u64 search_start,
u64 *found_objectid)
{
struct btrfs_path *path;
struct btrfs_key search_key;
int ret;
root = root->fs_info->tree_root;
search_key.objectid = search_start;
search_key.type = (u8)-1;
search_key.offset = (u64)-1;
path = btrfs_alloc_path();
BUG_ON(!path);
again:
ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
if (ret < 0)
goto out;
if (ret == 0) {
ret = 1;
goto out;
}
if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
ret = btrfs_next_leaf(root, path);
if (ret)
goto out;
}
btrfs_item_key_to_cpu(path->nodes[0], &search_key, path->slots[0]);
if (search_key.type != BTRFS_ROOT_ITEM_KEY) {
search_key.offset++;
btrfs_release_path(root, path);
goto again;
}
ret = 0;
*found_objectid = search_key.objectid;
out:
btrfs_free_path(path);
return ret;
}
/*
* lookup the root with the highest offset for a given objectid. The key we do
* find is copied into 'key'. If we find something return 0, otherwise 1, < 0
* on error.
*/
int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
struct btrfs_root_item *item, struct btrfs_key *key)
{
struct btrfs_path *path;
struct btrfs_key search_key;
struct btrfs_key found_key;
struct extent_buffer *l;
int ret;
int slot;
search_key.objectid = objectid;
search_key.type = BTRFS_ROOT_ITEM_KEY;
search_key.offset = (u64)-1;
path = btrfs_alloc_path();
BUG_ON(!path);
ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
if (ret < 0)
goto out;
BUG_ON(ret == 0);
if (path->slots[0] == 0) {
ret = 1;
goto out;
}
l = path->nodes[0];
slot = path->slots[0] - 1;
btrfs_item_key_to_cpu(l, &found_key, slot);
if (found_key.objectid != objectid ||
found_key.type != BTRFS_ROOT_ITEM_KEY) {
ret = 1;
goto out;
}
if (item)
read_extent_buffer(l, item, btrfs_item_ptr_offset(l, slot),
sizeof(*item));
if (key)
memcpy(key, &found_key, sizeof(found_key));
ret = 0;
out:
btrfs_free_path(path);
return ret;
}
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>
2009-06-10 07:45:14 -07:00
int btrfs_set_root_node(struct btrfs_root_item *item,
struct extent_buffer *node)
{
btrfs_set_root_bytenr(item, node->start);
btrfs_set_root_level(item, btrfs_header_level(node));
btrfs_set_root_generation(item, btrfs_header_generation(node));
return 0;
}
/*
* copy the data in 'item' into the btree
*/
int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root
*root, struct btrfs_key *key, struct btrfs_root_item
*item)
{
struct btrfs_path *path;
struct extent_buffer *l;
int ret;
int slot;
unsigned long ptr;
path = btrfs_alloc_path();
BUG_ON(!path);
ret = btrfs_search_slot(trans, root, key, path, 0, 1);
if (ret < 0)
goto out;
if (ret != 0) {
btrfs_print_leaf(root, path->nodes[0]);
printk(KERN_CRIT "unable to update root key %llu %u %llu\n",
(unsigned long long)key->objectid, key->type,
(unsigned long long)key->offset);
BUG_ON(1);
}
l = path->nodes[0];
slot = path->slots[0];
ptr = btrfs_item_ptr_offset(l, slot);
write_extent_buffer(l, item, ptr, sizeof(*item));
btrfs_mark_buffer_dirty(path->nodes[0]);
out:
btrfs_free_path(path);
return ret;
}
int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root
*root, struct btrfs_key *key, struct btrfs_root_item
*item)
{
int ret;
ret = btrfs_insert_item(trans, root, key, item, sizeof(*item));
return ret;
}
/*
* at mount time we want to find all the old transaction snapshots that were in
* the process of being deleted if we crashed. This is any root item with an
* offset lower than the latest root. They need to be queued for deletion to
* finish what was happening when we crashed.
*/
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>
2009-06-10 07:45:14 -07:00
int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid)
{
struct btrfs_root *dead_root;
struct btrfs_item *item;
struct btrfs_root_item *ri;
struct btrfs_key key;
struct btrfs_key found_key;
struct btrfs_path *path;
int ret;
u32 nritems;
struct extent_buffer *leaf;
int slot;
key.objectid = objectid;
btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
key.offset = 0;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
again:
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
if (ret < 0)
goto err;
while (1) {
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
slot = path->slots[0];
if (slot >= nritems) {
ret = btrfs_next_leaf(root, path);
if (ret)
break;
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
slot = path->slots[0];
}
item = btrfs_item_nr(leaf, slot);
btrfs_item_key_to_cpu(leaf, &key, slot);
if (btrfs_key_type(&key) != BTRFS_ROOT_ITEM_KEY)
goto next;
if (key.objectid < objectid)
goto next;
if (key.objectid > objectid)
break;
ri = btrfs_item_ptr(leaf, slot, struct btrfs_root_item);
if (btrfs_disk_root_refs(leaf, ri) != 0)
goto next;
memcpy(&found_key, &key, sizeof(key));
key.offset++;
btrfs_release_path(root, path);
dead_root =
btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
&found_key);
if (IS_ERR(dead_root)) {
ret = PTR_ERR(dead_root);
goto err;
}
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>
2009-06-10 07:45:14 -07:00
ret = btrfs_add_dead_root(dead_root);
if (ret)
goto err;
goto again;
next:
slot++;
path->slots[0]++;
}
ret = 0;
err:
btrfs_free_path(path);
return ret;
}
int btrfs_find_orphan_roots(struct btrfs_root *tree_root)
{
struct extent_buffer *leaf;
struct btrfs_path *path;
struct btrfs_key key;
struct btrfs_key root_key;
struct btrfs_root *root;
int err = 0;
int ret;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
key.objectid = BTRFS_ORPHAN_OBJECTID;
key.type = BTRFS_ORPHAN_ITEM_KEY;
key.offset = 0;
root_key.type = BTRFS_ROOT_ITEM_KEY;
root_key.offset = (u64)-1;
while (1) {
ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0);
if (ret < 0) {
err = ret;
break;
}
leaf = path->nodes[0];
if (path->slots[0] >= btrfs_header_nritems(leaf)) {
ret = btrfs_next_leaf(tree_root, path);
if (ret < 0)
err = ret;
if (ret != 0)
break;
leaf = path->nodes[0];
}
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
btrfs_release_path(tree_root, path);
if (key.objectid != BTRFS_ORPHAN_OBJECTID ||
key.type != BTRFS_ORPHAN_ITEM_KEY)
break;
root_key.objectid = key.offset;
key.offset++;
root = btrfs_read_fs_root_no_name(tree_root->fs_info,
&root_key);
if (!IS_ERR(root))
continue;
ret = PTR_ERR(root);
if (ret != -ENOENT) {
err = ret;
break;
}
ret = btrfs_find_dead_roots(tree_root, root_key.objectid);
if (ret) {
err = ret;
break;
}
}
btrfs_free_path(path);
return err;
}
/* drop the root item for 'key' from 'root' */
int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
struct btrfs_key *key)
{
struct btrfs_path *path;
int ret;
struct btrfs_root_item *ri;
struct extent_buffer *leaf;
path = btrfs_alloc_path();
BUG_ON(!path);
ret = btrfs_search_slot(trans, root, key, path, -1, 1);
if (ret < 0)
goto out;
BUG_ON(ret != 0);
leaf = path->nodes[0];
ri = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_item);
ret = btrfs_del_item(trans, root, path);
out:
btrfs_free_path(path);
return ret;
}
int btrfs_del_root_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *tree_root,
u64 root_id, u64 ref_id, u64 dirid, u64 *sequence,
const char *name, int name_len)
{
struct btrfs_path *path;
struct btrfs_root_ref *ref;
struct extent_buffer *leaf;
struct btrfs_key key;
unsigned long ptr;
int err = 0;
int ret;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
key.objectid = root_id;
key.type = BTRFS_ROOT_BACKREF_KEY;
key.offset = ref_id;
again:
ret = btrfs_search_slot(trans, tree_root, &key, path, -1, 1);
BUG_ON(ret < 0);
if (ret == 0) {
leaf = path->nodes[0];
ref = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_root_ref);
WARN_ON(btrfs_root_ref_dirid(leaf, ref) != dirid);
WARN_ON(btrfs_root_ref_name_len(leaf, ref) != name_len);
ptr = (unsigned long)(ref + 1);
WARN_ON(memcmp_extent_buffer(leaf, name, ptr, name_len));
*sequence = btrfs_root_ref_sequence(leaf, ref);
ret = btrfs_del_item(trans, tree_root, path);
BUG_ON(ret);
} else
err = -ENOENT;
if (key.type == BTRFS_ROOT_BACKREF_KEY) {
btrfs_release_path(tree_root, path);
key.objectid = ref_id;
key.type = BTRFS_ROOT_REF_KEY;
key.offset = root_id;
goto again;
}
btrfs_free_path(path);
return err;
}
int btrfs_find_root_ref(struct btrfs_root *tree_root,
struct btrfs_path *path,
u64 root_id, u64 ref_id)
{
struct btrfs_key key;
int ret;
key.objectid = root_id;
key.type = BTRFS_ROOT_REF_KEY;
key.offset = ref_id;
ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0);
return ret;
}
/*
* add a btrfs_root_ref item. type is either BTRFS_ROOT_REF_KEY
* or BTRFS_ROOT_BACKREF_KEY.
*
* The dirid, sequence, name and name_len refer to the directory entry
* that is referencing the root.
*
* For a forward ref, the root_id is the id of the tree referencing
* the root and ref_id is the id of the subvol or snapshot.
*
* For a back ref the root_id is the id of the subvol or snapshot and
* ref_id is the id of the tree referencing it.
*/
int btrfs_add_root_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *tree_root,
u64 root_id, u64 ref_id, u64 dirid, u64 sequence,
const char *name, int name_len)
{
struct btrfs_key key;
int ret;
struct btrfs_path *path;
struct btrfs_root_ref *ref;
struct extent_buffer *leaf;
unsigned long ptr;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
key.objectid = root_id;
key.type = BTRFS_ROOT_BACKREF_KEY;
key.offset = ref_id;
again:
ret = btrfs_insert_empty_item(trans, tree_root, path, &key,
sizeof(*ref) + name_len);
BUG_ON(ret);
leaf = path->nodes[0];
ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
btrfs_set_root_ref_dirid(leaf, ref, dirid);
btrfs_set_root_ref_sequence(leaf, ref, sequence);
btrfs_set_root_ref_name_len(leaf, ref, name_len);
ptr = (unsigned long)(ref + 1);
write_extent_buffer(leaf, name, ptr, name_len);
btrfs_mark_buffer_dirty(leaf);
if (key.type == BTRFS_ROOT_BACKREF_KEY) {
btrfs_release_path(tree_root, path);
key.objectid = ref_id;
key.type = BTRFS_ROOT_REF_KEY;
key.offset = root_id;
goto again;
}
btrfs_free_path(path);
return 0;
}