linux/fs/btrfs/uuid-tree.c

345 lines
7.5 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0
Btrfs: introduce a tree for items that map UUIDs to something Mapping UUIDs to subvolume IDs is an operation with a high effort today. Today, the algorithm even has quadratic effort (based on the number of existing subvolumes), which means, that it takes minutes to send/receive a single subvolume if 10,000 subvolumes exist. But even linear effort would be too much since it is a waste. And these data structures to allow mapping UUIDs to subvolume IDs are created every time a btrfs send/receive instance is started. It is much more efficient to maintain a searchable persistent data structure in the filesystem, one that is updated whenever a subvolume/snapshot is created and deleted, and when the received subvolume UUID is set by the btrfs-receive tool. Therefore kernel code is added with this commit that is able to maintain data structures in the filesystem that allow to quickly search for a given UUID and to retrieve data that is assigned to this UUID, like which subvolume ID is related to this UUID. This commit adds a new tree to hold UUID-to-data mapping items. The key of the items is the full UUID plus the key type BTRFS_UUID_KEY. Multiple data blocks can be stored for a given UUID, a type/length/ value scheme is used. Now follows the lengthy justification, why a new tree was added instead of using the existing root tree: The first approach was to not create another tree that holds UUID items. Instead, the items should just go into the top root tree. Unfortunately this confused the algorithm to assign the objectid of subvolumes and snapshots. The reason is that btrfs_find_free_objectid() calls btrfs_find_highest_objectid() for the first created subvol or snapshot after mounting a filesystem, and this function simply searches for the largest used objectid in the root tree keys to pick the next objectid to assign. Of course, the UUID keys have always been the ones with the highest offset value, and the next assigned subvol ID was wastefully huge. To use any other existing tree did not look proper. To apply a workaround such as setting the objectid to zero in the UUID item key and to implement collision handling would either add limitations (in case of a btrfs_extend_item() approach to handle the collisions) or a lot of complexity and source code (in case a key would be looked up that is free of collisions). Adding new code that introduces limitations is not good, and adding code that is complex and lengthy for no good reason is also not good. That's the justification why a completely new tree was introduced. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: Josef Bacik <jbacik@fusionio.com> Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2013-08-15 15:11:17 +00:00
/*
* Copyright (C) STRATO AG 2013. All rights reserved.
*/
Btrfs: introduce a tree for items that map UUIDs to something Mapping UUIDs to subvolume IDs is an operation with a high effort today. Today, the algorithm even has quadratic effort (based on the number of existing subvolumes), which means, that it takes minutes to send/receive a single subvolume if 10,000 subvolumes exist. But even linear effort would be too much since it is a waste. And these data structures to allow mapping UUIDs to subvolume IDs are created every time a btrfs send/receive instance is started. It is much more efficient to maintain a searchable persistent data structure in the filesystem, one that is updated whenever a subvolume/snapshot is created and deleted, and when the received subvolume UUID is set by the btrfs-receive tool. Therefore kernel code is added with this commit that is able to maintain data structures in the filesystem that allow to quickly search for a given UUID and to retrieve data that is assigned to this UUID, like which subvolume ID is related to this UUID. This commit adds a new tree to hold UUID-to-data mapping items. The key of the items is the full UUID plus the key type BTRFS_UUID_KEY. Multiple data blocks can be stored for a given UUID, a type/length/ value scheme is used. Now follows the lengthy justification, why a new tree was added instead of using the existing root tree: The first approach was to not create another tree that holds UUID items. Instead, the items should just go into the top root tree. Unfortunately this confused the algorithm to assign the objectid of subvolumes and snapshots. The reason is that btrfs_find_free_objectid() calls btrfs_find_highest_objectid() for the first created subvol or snapshot after mounting a filesystem, and this function simply searches for the largest used objectid in the root tree keys to pick the next objectid to assign. Of course, the UUID keys have always been the ones with the highest offset value, and the next assigned subvol ID was wastefully huge. To use any other existing tree did not look proper. To apply a workaround such as setting the objectid to zero in the UUID item key and to implement collision handling would either add limitations (in case of a btrfs_extend_item() approach to handle the collisions) or a lot of complexity and source code (in case a key would be looked up that is free of collisions). Adding new code that introduces limitations is not good, and adding code that is complex and lengthy for no good reason is also not good. That's the justification why a completely new tree was introduced. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: Josef Bacik <jbacik@fusionio.com> Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2013-08-15 15:11:17 +00:00
#include <linux/uuid.h>
#include <asm/unaligned.h>
#include "ctree.h"
#include "transaction.h"
#include "disk-io.h"
#include "print-tree.h"
static void btrfs_uuid_to_key(u8 *uuid, u8 type, struct btrfs_key *key)
{
key->type = type;
key->objectid = get_unaligned_le64(uuid);
key->offset = get_unaligned_le64(uuid + sizeof(u64));
}
/* return -ENOENT for !found, < 0 for errors, or 0 if an item was found */
static int btrfs_uuid_tree_lookup(struct btrfs_root *uuid_root, u8 *uuid,
u8 type, u64 subid)
{
int ret;
struct btrfs_path *path = NULL;
struct extent_buffer *eb;
int slot;
u32 item_size;
unsigned long offset;
struct btrfs_key key;
if (WARN_ON_ONCE(!uuid_root)) {
ret = -ENOENT;
goto out;
}
path = btrfs_alloc_path();
if (!path) {
ret = -ENOMEM;
goto out;
}
btrfs_uuid_to_key(uuid, type, &key);
ret = btrfs_search_slot(NULL, uuid_root, &key, path, 0, 0);
if (ret < 0) {
goto out;
} else if (ret > 0) {
ret = -ENOENT;
goto out;
}
eb = path->nodes[0];
slot = path->slots[0];
item_size = btrfs_item_size_nr(eb, slot);
offset = btrfs_item_ptr_offset(eb, slot);
ret = -ENOENT;
if (!IS_ALIGNED(item_size, sizeof(u64))) {
btrfs_warn(uuid_root->fs_info,
"uuid item with illegal size %lu!",
(unsigned long)item_size);
Btrfs: introduce a tree for items that map UUIDs to something Mapping UUIDs to subvolume IDs is an operation with a high effort today. Today, the algorithm even has quadratic effort (based on the number of existing subvolumes), which means, that it takes minutes to send/receive a single subvolume if 10,000 subvolumes exist. But even linear effort would be too much since it is a waste. And these data structures to allow mapping UUIDs to subvolume IDs are created every time a btrfs send/receive instance is started. It is much more efficient to maintain a searchable persistent data structure in the filesystem, one that is updated whenever a subvolume/snapshot is created and deleted, and when the received subvolume UUID is set by the btrfs-receive tool. Therefore kernel code is added with this commit that is able to maintain data structures in the filesystem that allow to quickly search for a given UUID and to retrieve data that is assigned to this UUID, like which subvolume ID is related to this UUID. This commit adds a new tree to hold UUID-to-data mapping items. The key of the items is the full UUID plus the key type BTRFS_UUID_KEY. Multiple data blocks can be stored for a given UUID, a type/length/ value scheme is used. Now follows the lengthy justification, why a new tree was added instead of using the existing root tree: The first approach was to not create another tree that holds UUID items. Instead, the items should just go into the top root tree. Unfortunately this confused the algorithm to assign the objectid of subvolumes and snapshots. The reason is that btrfs_find_free_objectid() calls btrfs_find_highest_objectid() for the first created subvol or snapshot after mounting a filesystem, and this function simply searches for the largest used objectid in the root tree keys to pick the next objectid to assign. Of course, the UUID keys have always been the ones with the highest offset value, and the next assigned subvol ID was wastefully huge. To use any other existing tree did not look proper. To apply a workaround such as setting the objectid to zero in the UUID item key and to implement collision handling would either add limitations (in case of a btrfs_extend_item() approach to handle the collisions) or a lot of complexity and source code (in case a key would be looked up that is free of collisions). Adding new code that introduces limitations is not good, and adding code that is complex and lengthy for no good reason is also not good. That's the justification why a completely new tree was introduced. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: Josef Bacik <jbacik@fusionio.com> Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2013-08-15 15:11:17 +00:00
goto out;
}
while (item_size) {
__le64 data;
read_extent_buffer(eb, &data, offset, sizeof(data));
if (le64_to_cpu(data) == subid) {
ret = 0;
break;
}
offset += sizeof(data);
item_size -= sizeof(data);
}
out:
btrfs_free_path(path);
return ret;
}
int btrfs_uuid_tree_add(struct btrfs_trans_handle *trans, u8 *uuid, u8 type,
Btrfs: introduce a tree for items that map UUIDs to something Mapping UUIDs to subvolume IDs is an operation with a high effort today. Today, the algorithm even has quadratic effort (based on the number of existing subvolumes), which means, that it takes minutes to send/receive a single subvolume if 10,000 subvolumes exist. But even linear effort would be too much since it is a waste. And these data structures to allow mapping UUIDs to subvolume IDs are created every time a btrfs send/receive instance is started. It is much more efficient to maintain a searchable persistent data structure in the filesystem, one that is updated whenever a subvolume/snapshot is created and deleted, and when the received subvolume UUID is set by the btrfs-receive tool. Therefore kernel code is added with this commit that is able to maintain data structures in the filesystem that allow to quickly search for a given UUID and to retrieve data that is assigned to this UUID, like which subvolume ID is related to this UUID. This commit adds a new tree to hold UUID-to-data mapping items. The key of the items is the full UUID plus the key type BTRFS_UUID_KEY. Multiple data blocks can be stored for a given UUID, a type/length/ value scheme is used. Now follows the lengthy justification, why a new tree was added instead of using the existing root tree: The first approach was to not create another tree that holds UUID items. Instead, the items should just go into the top root tree. Unfortunately this confused the algorithm to assign the objectid of subvolumes and snapshots. The reason is that btrfs_find_free_objectid() calls btrfs_find_highest_objectid() for the first created subvol or snapshot after mounting a filesystem, and this function simply searches for the largest used objectid in the root tree keys to pick the next objectid to assign. Of course, the UUID keys have always been the ones with the highest offset value, and the next assigned subvol ID was wastefully huge. To use any other existing tree did not look proper. To apply a workaround such as setting the objectid to zero in the UUID item key and to implement collision handling would either add limitations (in case of a btrfs_extend_item() approach to handle the collisions) or a lot of complexity and source code (in case a key would be looked up that is free of collisions). Adding new code that introduces limitations is not good, and adding code that is complex and lengthy for no good reason is also not good. That's the justification why a completely new tree was introduced. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: Josef Bacik <jbacik@fusionio.com> Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2013-08-15 15:11:17 +00:00
u64 subid_cpu)
{
struct btrfs_fs_info *fs_info = trans->fs_info;
struct btrfs_root *uuid_root = fs_info->uuid_root;
Btrfs: introduce a tree for items that map UUIDs to something Mapping UUIDs to subvolume IDs is an operation with a high effort today. Today, the algorithm even has quadratic effort (based on the number of existing subvolumes), which means, that it takes minutes to send/receive a single subvolume if 10,000 subvolumes exist. But even linear effort would be too much since it is a waste. And these data structures to allow mapping UUIDs to subvolume IDs are created every time a btrfs send/receive instance is started. It is much more efficient to maintain a searchable persistent data structure in the filesystem, one that is updated whenever a subvolume/snapshot is created and deleted, and when the received subvolume UUID is set by the btrfs-receive tool. Therefore kernel code is added with this commit that is able to maintain data structures in the filesystem that allow to quickly search for a given UUID and to retrieve data that is assigned to this UUID, like which subvolume ID is related to this UUID. This commit adds a new tree to hold UUID-to-data mapping items. The key of the items is the full UUID plus the key type BTRFS_UUID_KEY. Multiple data blocks can be stored for a given UUID, a type/length/ value scheme is used. Now follows the lengthy justification, why a new tree was added instead of using the existing root tree: The first approach was to not create another tree that holds UUID items. Instead, the items should just go into the top root tree. Unfortunately this confused the algorithm to assign the objectid of subvolumes and snapshots. The reason is that btrfs_find_free_objectid() calls btrfs_find_highest_objectid() for the first created subvol or snapshot after mounting a filesystem, and this function simply searches for the largest used objectid in the root tree keys to pick the next objectid to assign. Of course, the UUID keys have always been the ones with the highest offset value, and the next assigned subvol ID was wastefully huge. To use any other existing tree did not look proper. To apply a workaround such as setting the objectid to zero in the UUID item key and to implement collision handling would either add limitations (in case of a btrfs_extend_item() approach to handle the collisions) or a lot of complexity and source code (in case a key would be looked up that is free of collisions). Adding new code that introduces limitations is not good, and adding code that is complex and lengthy for no good reason is also not good. That's the justification why a completely new tree was introduced. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: Josef Bacik <jbacik@fusionio.com> Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2013-08-15 15:11:17 +00:00
int ret;
struct btrfs_path *path = NULL;
struct btrfs_key key;
struct extent_buffer *eb;
int slot;
unsigned long offset;
__le64 subid_le;
ret = btrfs_uuid_tree_lookup(uuid_root, uuid, type, subid_cpu);
if (ret != -ENOENT)
return ret;
if (WARN_ON_ONCE(!uuid_root)) {
ret = -EINVAL;
goto out;
}
btrfs_uuid_to_key(uuid, type, &key);
path = btrfs_alloc_path();
if (!path) {
ret = -ENOMEM;
goto out;
}
ret = btrfs_insert_empty_item(trans, uuid_root, path, &key,
sizeof(subid_le));
if (ret >= 0) {
/* Add an item for the type for the first time */
eb = path->nodes[0];
slot = path->slots[0];
offset = btrfs_item_ptr_offset(eb, slot);
} else if (ret == -EEXIST) {
/*
* An item with that type already exists.
* Extend the item and store the new subid at the end.
*/
btrfs_extend_item(path, sizeof(subid_le));
Btrfs: introduce a tree for items that map UUIDs to something Mapping UUIDs to subvolume IDs is an operation with a high effort today. Today, the algorithm even has quadratic effort (based on the number of existing subvolumes), which means, that it takes minutes to send/receive a single subvolume if 10,000 subvolumes exist. But even linear effort would be too much since it is a waste. And these data structures to allow mapping UUIDs to subvolume IDs are created every time a btrfs send/receive instance is started. It is much more efficient to maintain a searchable persistent data structure in the filesystem, one that is updated whenever a subvolume/snapshot is created and deleted, and when the received subvolume UUID is set by the btrfs-receive tool. Therefore kernel code is added with this commit that is able to maintain data structures in the filesystem that allow to quickly search for a given UUID and to retrieve data that is assigned to this UUID, like which subvolume ID is related to this UUID. This commit adds a new tree to hold UUID-to-data mapping items. The key of the items is the full UUID plus the key type BTRFS_UUID_KEY. Multiple data blocks can be stored for a given UUID, a type/length/ value scheme is used. Now follows the lengthy justification, why a new tree was added instead of using the existing root tree: The first approach was to not create another tree that holds UUID items. Instead, the items should just go into the top root tree. Unfortunately this confused the algorithm to assign the objectid of subvolumes and snapshots. The reason is that btrfs_find_free_objectid() calls btrfs_find_highest_objectid() for the first created subvol or snapshot after mounting a filesystem, and this function simply searches for the largest used objectid in the root tree keys to pick the next objectid to assign. Of course, the UUID keys have always been the ones with the highest offset value, and the next assigned subvol ID was wastefully huge. To use any other existing tree did not look proper. To apply a workaround such as setting the objectid to zero in the UUID item key and to implement collision handling would either add limitations (in case of a btrfs_extend_item() approach to handle the collisions) or a lot of complexity and source code (in case a key would be looked up that is free of collisions). Adding new code that introduces limitations is not good, and adding code that is complex and lengthy for no good reason is also not good. That's the justification why a completely new tree was introduced. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: Josef Bacik <jbacik@fusionio.com> Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2013-08-15 15:11:17 +00:00
eb = path->nodes[0];
slot = path->slots[0];
offset = btrfs_item_ptr_offset(eb, slot);
offset += btrfs_item_size_nr(eb, slot) - sizeof(subid_le);
} else {
btrfs_warn(fs_info,
"insert uuid item failed %d (0x%016llx, 0x%016llx) type %u!",
ret, (unsigned long long)key.objectid,
(unsigned long long)key.offset, type);
Btrfs: introduce a tree for items that map UUIDs to something Mapping UUIDs to subvolume IDs is an operation with a high effort today. Today, the algorithm even has quadratic effort (based on the number of existing subvolumes), which means, that it takes minutes to send/receive a single subvolume if 10,000 subvolumes exist. But even linear effort would be too much since it is a waste. And these data structures to allow mapping UUIDs to subvolume IDs are created every time a btrfs send/receive instance is started. It is much more efficient to maintain a searchable persistent data structure in the filesystem, one that is updated whenever a subvolume/snapshot is created and deleted, and when the received subvolume UUID is set by the btrfs-receive tool. Therefore kernel code is added with this commit that is able to maintain data structures in the filesystem that allow to quickly search for a given UUID and to retrieve data that is assigned to this UUID, like which subvolume ID is related to this UUID. This commit adds a new tree to hold UUID-to-data mapping items. The key of the items is the full UUID plus the key type BTRFS_UUID_KEY. Multiple data blocks can be stored for a given UUID, a type/length/ value scheme is used. Now follows the lengthy justification, why a new tree was added instead of using the existing root tree: The first approach was to not create another tree that holds UUID items. Instead, the items should just go into the top root tree. Unfortunately this confused the algorithm to assign the objectid of subvolumes and snapshots. The reason is that btrfs_find_free_objectid() calls btrfs_find_highest_objectid() for the first created subvol or snapshot after mounting a filesystem, and this function simply searches for the largest used objectid in the root tree keys to pick the next objectid to assign. Of course, the UUID keys have always been the ones with the highest offset value, and the next assigned subvol ID was wastefully huge. To use any other existing tree did not look proper. To apply a workaround such as setting the objectid to zero in the UUID item key and to implement collision handling would either add limitations (in case of a btrfs_extend_item() approach to handle the collisions) or a lot of complexity and source code (in case a key would be looked up that is free of collisions). Adding new code that introduces limitations is not good, and adding code that is complex and lengthy for no good reason is also not good. That's the justification why a completely new tree was introduced. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: Josef Bacik <jbacik@fusionio.com> Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2013-08-15 15:11:17 +00:00
goto out;
}
ret = 0;
subid_le = cpu_to_le64(subid_cpu);
write_extent_buffer(eb, &subid_le, offset, sizeof(subid_le));
btrfs_mark_buffer_dirty(eb);
out:
btrfs_free_path(path);
return ret;
}
int btrfs_uuid_tree_remove(struct btrfs_trans_handle *trans, u8 *uuid, u8 type,
Btrfs: introduce a tree for items that map UUIDs to something Mapping UUIDs to subvolume IDs is an operation with a high effort today. Today, the algorithm even has quadratic effort (based on the number of existing subvolumes), which means, that it takes minutes to send/receive a single subvolume if 10,000 subvolumes exist. But even linear effort would be too much since it is a waste. And these data structures to allow mapping UUIDs to subvolume IDs are created every time a btrfs send/receive instance is started. It is much more efficient to maintain a searchable persistent data structure in the filesystem, one that is updated whenever a subvolume/snapshot is created and deleted, and when the received subvolume UUID is set by the btrfs-receive tool. Therefore kernel code is added with this commit that is able to maintain data structures in the filesystem that allow to quickly search for a given UUID and to retrieve data that is assigned to this UUID, like which subvolume ID is related to this UUID. This commit adds a new tree to hold UUID-to-data mapping items. The key of the items is the full UUID plus the key type BTRFS_UUID_KEY. Multiple data blocks can be stored for a given UUID, a type/length/ value scheme is used. Now follows the lengthy justification, why a new tree was added instead of using the existing root tree: The first approach was to not create another tree that holds UUID items. Instead, the items should just go into the top root tree. Unfortunately this confused the algorithm to assign the objectid of subvolumes and snapshots. The reason is that btrfs_find_free_objectid() calls btrfs_find_highest_objectid() for the first created subvol or snapshot after mounting a filesystem, and this function simply searches for the largest used objectid in the root tree keys to pick the next objectid to assign. Of course, the UUID keys have always been the ones with the highest offset value, and the next assigned subvol ID was wastefully huge. To use any other existing tree did not look proper. To apply a workaround such as setting the objectid to zero in the UUID item key and to implement collision handling would either add limitations (in case of a btrfs_extend_item() approach to handle the collisions) or a lot of complexity and source code (in case a key would be looked up that is free of collisions). Adding new code that introduces limitations is not good, and adding code that is complex and lengthy for no good reason is also not good. That's the justification why a completely new tree was introduced. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: Josef Bacik <jbacik@fusionio.com> Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2013-08-15 15:11:17 +00:00
u64 subid)
{
struct btrfs_fs_info *fs_info = trans->fs_info;
struct btrfs_root *uuid_root = fs_info->uuid_root;
Btrfs: introduce a tree for items that map UUIDs to something Mapping UUIDs to subvolume IDs is an operation with a high effort today. Today, the algorithm even has quadratic effort (based on the number of existing subvolumes), which means, that it takes minutes to send/receive a single subvolume if 10,000 subvolumes exist. But even linear effort would be too much since it is a waste. And these data structures to allow mapping UUIDs to subvolume IDs are created every time a btrfs send/receive instance is started. It is much more efficient to maintain a searchable persistent data structure in the filesystem, one that is updated whenever a subvolume/snapshot is created and deleted, and when the received subvolume UUID is set by the btrfs-receive tool. Therefore kernel code is added with this commit that is able to maintain data structures in the filesystem that allow to quickly search for a given UUID and to retrieve data that is assigned to this UUID, like which subvolume ID is related to this UUID. This commit adds a new tree to hold UUID-to-data mapping items. The key of the items is the full UUID plus the key type BTRFS_UUID_KEY. Multiple data blocks can be stored for a given UUID, a type/length/ value scheme is used. Now follows the lengthy justification, why a new tree was added instead of using the existing root tree: The first approach was to not create another tree that holds UUID items. Instead, the items should just go into the top root tree. Unfortunately this confused the algorithm to assign the objectid of subvolumes and snapshots. The reason is that btrfs_find_free_objectid() calls btrfs_find_highest_objectid() for the first created subvol or snapshot after mounting a filesystem, and this function simply searches for the largest used objectid in the root tree keys to pick the next objectid to assign. Of course, the UUID keys have always been the ones with the highest offset value, and the next assigned subvol ID was wastefully huge. To use any other existing tree did not look proper. To apply a workaround such as setting the objectid to zero in the UUID item key and to implement collision handling would either add limitations (in case of a btrfs_extend_item() approach to handle the collisions) or a lot of complexity and source code (in case a key would be looked up that is free of collisions). Adding new code that introduces limitations is not good, and adding code that is complex and lengthy for no good reason is also not good. That's the justification why a completely new tree was introduced. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: Josef Bacik <jbacik@fusionio.com> Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2013-08-15 15:11:17 +00:00
int ret;
struct btrfs_path *path = NULL;
struct btrfs_key key;
struct extent_buffer *eb;
int slot;
unsigned long offset;
u32 item_size;
unsigned long move_dst;
unsigned long move_src;
unsigned long move_len;
if (WARN_ON_ONCE(!uuid_root)) {
ret = -EINVAL;
goto out;
}
btrfs_uuid_to_key(uuid, type, &key);
path = btrfs_alloc_path();
if (!path) {
ret = -ENOMEM;
goto out;
}
ret = btrfs_search_slot(trans, uuid_root, &key, path, -1, 1);
if (ret < 0) {
btrfs_warn(fs_info, "error %d while searching for uuid item!",
ret);
Btrfs: introduce a tree for items that map UUIDs to something Mapping UUIDs to subvolume IDs is an operation with a high effort today. Today, the algorithm even has quadratic effort (based on the number of existing subvolumes), which means, that it takes minutes to send/receive a single subvolume if 10,000 subvolumes exist. But even linear effort would be too much since it is a waste. And these data structures to allow mapping UUIDs to subvolume IDs are created every time a btrfs send/receive instance is started. It is much more efficient to maintain a searchable persistent data structure in the filesystem, one that is updated whenever a subvolume/snapshot is created and deleted, and when the received subvolume UUID is set by the btrfs-receive tool. Therefore kernel code is added with this commit that is able to maintain data structures in the filesystem that allow to quickly search for a given UUID and to retrieve data that is assigned to this UUID, like which subvolume ID is related to this UUID. This commit adds a new tree to hold UUID-to-data mapping items. The key of the items is the full UUID plus the key type BTRFS_UUID_KEY. Multiple data blocks can be stored for a given UUID, a type/length/ value scheme is used. Now follows the lengthy justification, why a new tree was added instead of using the existing root tree: The first approach was to not create another tree that holds UUID items. Instead, the items should just go into the top root tree. Unfortunately this confused the algorithm to assign the objectid of subvolumes and snapshots. The reason is that btrfs_find_free_objectid() calls btrfs_find_highest_objectid() for the first created subvol or snapshot after mounting a filesystem, and this function simply searches for the largest used objectid in the root tree keys to pick the next objectid to assign. Of course, the UUID keys have always been the ones with the highest offset value, and the next assigned subvol ID was wastefully huge. To use any other existing tree did not look proper. To apply a workaround such as setting the objectid to zero in the UUID item key and to implement collision handling would either add limitations (in case of a btrfs_extend_item() approach to handle the collisions) or a lot of complexity and source code (in case a key would be looked up that is free of collisions). Adding new code that introduces limitations is not good, and adding code that is complex and lengthy for no good reason is also not good. That's the justification why a completely new tree was introduced. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: Josef Bacik <jbacik@fusionio.com> Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2013-08-15 15:11:17 +00:00
goto out;
}
if (ret > 0) {
ret = -ENOENT;
goto out;
}
eb = path->nodes[0];
slot = path->slots[0];
offset = btrfs_item_ptr_offset(eb, slot);
item_size = btrfs_item_size_nr(eb, slot);
if (!IS_ALIGNED(item_size, sizeof(u64))) {
btrfs_warn(fs_info, "uuid item with illegal size %lu!",
(unsigned long)item_size);
Btrfs: introduce a tree for items that map UUIDs to something Mapping UUIDs to subvolume IDs is an operation with a high effort today. Today, the algorithm even has quadratic effort (based on the number of existing subvolumes), which means, that it takes minutes to send/receive a single subvolume if 10,000 subvolumes exist. But even linear effort would be too much since it is a waste. And these data structures to allow mapping UUIDs to subvolume IDs are created every time a btrfs send/receive instance is started. It is much more efficient to maintain a searchable persistent data structure in the filesystem, one that is updated whenever a subvolume/snapshot is created and deleted, and when the received subvolume UUID is set by the btrfs-receive tool. Therefore kernel code is added with this commit that is able to maintain data structures in the filesystem that allow to quickly search for a given UUID and to retrieve data that is assigned to this UUID, like which subvolume ID is related to this UUID. This commit adds a new tree to hold UUID-to-data mapping items. The key of the items is the full UUID plus the key type BTRFS_UUID_KEY. Multiple data blocks can be stored for a given UUID, a type/length/ value scheme is used. Now follows the lengthy justification, why a new tree was added instead of using the existing root tree: The first approach was to not create another tree that holds UUID items. Instead, the items should just go into the top root tree. Unfortunately this confused the algorithm to assign the objectid of subvolumes and snapshots. The reason is that btrfs_find_free_objectid() calls btrfs_find_highest_objectid() for the first created subvol or snapshot after mounting a filesystem, and this function simply searches for the largest used objectid in the root tree keys to pick the next objectid to assign. Of course, the UUID keys have always been the ones with the highest offset value, and the next assigned subvol ID was wastefully huge. To use any other existing tree did not look proper. To apply a workaround such as setting the objectid to zero in the UUID item key and to implement collision handling would either add limitations (in case of a btrfs_extend_item() approach to handle the collisions) or a lot of complexity and source code (in case a key would be looked up that is free of collisions). Adding new code that introduces limitations is not good, and adding code that is complex and lengthy for no good reason is also not good. That's the justification why a completely new tree was introduced. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: Josef Bacik <jbacik@fusionio.com> Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2013-08-15 15:11:17 +00:00
ret = -ENOENT;
goto out;
}
while (item_size) {
__le64 read_subid;
read_extent_buffer(eb, &read_subid, offset, sizeof(read_subid));
if (le64_to_cpu(read_subid) == subid)
break;
offset += sizeof(read_subid);
item_size -= sizeof(read_subid);
}
if (!item_size) {
ret = -ENOENT;
goto out;
}
item_size = btrfs_item_size_nr(eb, slot);
if (item_size == sizeof(subid)) {
ret = btrfs_del_item(trans, uuid_root, path);
goto out;
}
move_dst = offset;
move_src = offset + sizeof(subid);
move_len = item_size - (move_src - btrfs_item_ptr_offset(eb, slot));
memmove_extent_buffer(eb, move_dst, move_src, move_len);
btrfs_truncate_item(path, item_size - sizeof(subid), 1);
Btrfs: introduce a tree for items that map UUIDs to something Mapping UUIDs to subvolume IDs is an operation with a high effort today. Today, the algorithm even has quadratic effort (based on the number of existing subvolumes), which means, that it takes minutes to send/receive a single subvolume if 10,000 subvolumes exist. But even linear effort would be too much since it is a waste. And these data structures to allow mapping UUIDs to subvolume IDs are created every time a btrfs send/receive instance is started. It is much more efficient to maintain a searchable persistent data structure in the filesystem, one that is updated whenever a subvolume/snapshot is created and deleted, and when the received subvolume UUID is set by the btrfs-receive tool. Therefore kernel code is added with this commit that is able to maintain data structures in the filesystem that allow to quickly search for a given UUID and to retrieve data that is assigned to this UUID, like which subvolume ID is related to this UUID. This commit adds a new tree to hold UUID-to-data mapping items. The key of the items is the full UUID plus the key type BTRFS_UUID_KEY. Multiple data blocks can be stored for a given UUID, a type/length/ value scheme is used. Now follows the lengthy justification, why a new tree was added instead of using the existing root tree: The first approach was to not create another tree that holds UUID items. Instead, the items should just go into the top root tree. Unfortunately this confused the algorithm to assign the objectid of subvolumes and snapshots. The reason is that btrfs_find_free_objectid() calls btrfs_find_highest_objectid() for the first created subvol or snapshot after mounting a filesystem, and this function simply searches for the largest used objectid in the root tree keys to pick the next objectid to assign. Of course, the UUID keys have always been the ones with the highest offset value, and the next assigned subvol ID was wastefully huge. To use any other existing tree did not look proper. To apply a workaround such as setting the objectid to zero in the UUID item key and to implement collision handling would either add limitations (in case of a btrfs_extend_item() approach to handle the collisions) or a lot of complexity and source code (in case a key would be looked up that is free of collisions). Adding new code that introduces limitations is not good, and adding code that is complex and lengthy for no good reason is also not good. That's the justification why a completely new tree was introduced. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: Josef Bacik <jbacik@fusionio.com> Signed-off-by: Chris Mason <chris.mason@fusionio.com>
2013-08-15 15:11:17 +00:00
out:
btrfs_free_path(path);
return ret;
}
static int btrfs_uuid_iter_rem(struct btrfs_root *uuid_root, u8 *uuid, u8 type,
u64 subid)
{
struct btrfs_trans_handle *trans;
int ret;
/* 1 - for the uuid item */
trans = btrfs_start_transaction(uuid_root, 1);
if (IS_ERR(trans)) {
ret = PTR_ERR(trans);
goto out;
}
ret = btrfs_uuid_tree_remove(trans, uuid, type, subid);
btrfs_end_transaction(trans);
out:
return ret;
}
int btrfs_uuid_tree_iterate(struct btrfs_fs_info *fs_info,
int (*check_func)(struct btrfs_fs_info *, u8 *, u8,
u64))
{
struct btrfs_root *root = fs_info->uuid_root;
struct btrfs_key key;
struct btrfs_path *path;
int ret = 0;
struct extent_buffer *leaf;
int slot;
u32 item_size;
unsigned long offset;
path = btrfs_alloc_path();
if (!path) {
ret = -ENOMEM;
goto out;
}
key.objectid = 0;
key.type = 0;
key.offset = 0;
again_search_slot:
ret = btrfs_search_forward(root, &key, path, BTRFS_OLDEST_GENERATION);
if (ret) {
if (ret > 0)
ret = 0;
goto out;
}
while (1) {
cond_resched();
leaf = path->nodes[0];
slot = path->slots[0];
btrfs_item_key_to_cpu(leaf, &key, slot);
if (key.type != BTRFS_UUID_KEY_SUBVOL &&
key.type != BTRFS_UUID_KEY_RECEIVED_SUBVOL)
goto skip;
offset = btrfs_item_ptr_offset(leaf, slot);
item_size = btrfs_item_size_nr(leaf, slot);
if (!IS_ALIGNED(item_size, sizeof(u64))) {
btrfs_warn(fs_info,
"uuid item with illegal size %lu!",
(unsigned long)item_size);
goto skip;
}
while (item_size) {
u8 uuid[BTRFS_UUID_SIZE];
__le64 subid_le;
u64 subid_cpu;
put_unaligned_le64(key.objectid, uuid);
put_unaligned_le64(key.offset, uuid + sizeof(u64));
read_extent_buffer(leaf, &subid_le, offset,
sizeof(subid_le));
subid_cpu = le64_to_cpu(subid_le);
ret = check_func(fs_info, uuid, key.type, subid_cpu);
if (ret < 0)
goto out;
if (ret > 0) {
btrfs_release_path(path);
ret = btrfs_uuid_iter_rem(root, uuid, key.type,
subid_cpu);
if (ret == 0) {
/*
* this might look inefficient, but the
* justification is that it is an
* exception that check_func returns 1,
* and that in the regular case only one
* entry per UUID exists.
*/
goto again_search_slot;
}
if (ret < 0 && ret != -ENOENT)
goto out;
}
item_size -= sizeof(subid_le);
offset += sizeof(subid_le);
}
skip:
ret = btrfs_next_item(root, path);
if (ret == 0)
continue;
else if (ret > 0)
ret = 0;
break;
}
out:
btrfs_free_path(path);
return ret;
}