btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
/*
|
|
|
|
* Copyright (C) 2011 Fujitsu. All rights reserved.
|
|
|
|
* Written by Miao Xie <miaox@cn.fujitsu.com>
|
|
|
|
*
|
|
|
|
* 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 <linux/slab.h>
|
|
|
|
#include "delayed-inode.h"
|
|
|
|
#include "disk-io.h"
|
|
|
|
#include "transaction.h"
|
|
|
|
|
|
|
|
#define BTRFS_DELAYED_WRITEBACK 400
|
|
|
|
#define BTRFS_DELAYED_BACKGROUND 100
|
|
|
|
|
|
|
|
static struct kmem_cache *delayed_node_cache;
|
|
|
|
|
|
|
|
int __init btrfs_delayed_inode_init(void)
|
|
|
|
{
|
|
|
|
delayed_node_cache = kmem_cache_create("delayed_node",
|
|
|
|
sizeof(struct btrfs_delayed_node),
|
|
|
|
0,
|
|
|
|
SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
|
|
|
|
NULL);
|
|
|
|
if (!delayed_node_cache)
|
|
|
|
return -ENOMEM;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
void btrfs_delayed_inode_exit(void)
|
|
|
|
{
|
|
|
|
if (delayed_node_cache)
|
|
|
|
kmem_cache_destroy(delayed_node_cache);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void btrfs_init_delayed_node(
|
|
|
|
struct btrfs_delayed_node *delayed_node,
|
|
|
|
struct btrfs_root *root, u64 inode_id)
|
|
|
|
{
|
|
|
|
delayed_node->root = root;
|
|
|
|
delayed_node->inode_id = inode_id;
|
|
|
|
atomic_set(&delayed_node->refs, 0);
|
|
|
|
delayed_node->count = 0;
|
|
|
|
delayed_node->in_list = 0;
|
|
|
|
delayed_node->inode_dirty = 0;
|
|
|
|
delayed_node->ins_root = RB_ROOT;
|
|
|
|
delayed_node->del_root = RB_ROOT;
|
|
|
|
mutex_init(&delayed_node->mutex);
|
|
|
|
delayed_node->index_cnt = 0;
|
|
|
|
INIT_LIST_HEAD(&delayed_node->n_list);
|
|
|
|
INIT_LIST_HEAD(&delayed_node->p_list);
|
|
|
|
delayed_node->bytes_reserved = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline int btrfs_is_continuous_delayed_item(
|
|
|
|
struct btrfs_delayed_item *item1,
|
|
|
|
struct btrfs_delayed_item *item2)
|
|
|
|
{
|
|
|
|
if (item1->key.type == BTRFS_DIR_INDEX_KEY &&
|
|
|
|
item1->key.objectid == item2->key.objectid &&
|
|
|
|
item1->key.type == item2->key.type &&
|
|
|
|
item1->key.offset + 1 == item2->key.offset)
|
|
|
|
return 1;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline struct btrfs_delayed_root *btrfs_get_delayed_root(
|
|
|
|
struct btrfs_root *root)
|
|
|
|
{
|
|
|
|
return root->fs_info->delayed_root;
|
|
|
|
}
|
|
|
|
|
2011-06-23 09:27:13 +02:00
|
|
|
static struct btrfs_delayed_node *btrfs_get_delayed_node(struct inode *inode)
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
{
|
|
|
|
struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
|
|
|
|
struct btrfs_root *root = btrfs_inode->root;
|
2011-05-22 13:11:22 +02:00
|
|
|
u64 ino = btrfs_ino(inode);
|
2011-06-23 09:27:13 +02:00
|
|
|
struct btrfs_delayed_node *node;
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
|
|
|
|
node = ACCESS_ONCE(btrfs_inode->delayed_node);
|
|
|
|
if (node) {
|
2011-06-23 09:27:13 +02:00
|
|
|
atomic_inc(&node->refs);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
return node;
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_lock(&root->inode_lock);
|
2011-05-22 13:11:22 +02:00
|
|
|
node = radix_tree_lookup(&root->delayed_nodes_tree, ino);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
if (node) {
|
|
|
|
if (btrfs_inode->delayed_node) {
|
2011-06-23 09:27:13 +02:00
|
|
|
atomic_inc(&node->refs); /* can be accessed */
|
|
|
|
BUG_ON(btrfs_inode->delayed_node != node);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
spin_unlock(&root->inode_lock);
|
2011-06-23 09:27:13 +02:00
|
|
|
return node;
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
}
|
|
|
|
btrfs_inode->delayed_node = node;
|
|
|
|
atomic_inc(&node->refs); /* can be accessed */
|
|
|
|
atomic_inc(&node->refs); /* cached in the inode */
|
|
|
|
spin_unlock(&root->inode_lock);
|
|
|
|
return node;
|
|
|
|
}
|
|
|
|
spin_unlock(&root->inode_lock);
|
|
|
|
|
2011-06-23 09:27:13 +02:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct btrfs_delayed_node *btrfs_get_or_create_delayed_node(
|
|
|
|
struct inode *inode)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_node *node;
|
|
|
|
struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
|
|
|
|
struct btrfs_root *root = btrfs_inode->root;
|
|
|
|
u64 ino = btrfs_ino(inode);
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
again:
|
|
|
|
node = btrfs_get_delayed_node(inode);
|
|
|
|
if (node)
|
|
|
|
return node;
|
|
|
|
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
node = kmem_cache_alloc(delayed_node_cache, GFP_NOFS);
|
|
|
|
if (!node)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
2011-05-22 13:11:22 +02:00
|
|
|
btrfs_init_delayed_node(node, root, ino);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
|
|
|
|
atomic_inc(&node->refs); /* cached in the btrfs inode */
|
|
|
|
atomic_inc(&node->refs); /* can be accessed */
|
|
|
|
|
|
|
|
ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
|
|
|
|
if (ret) {
|
|
|
|
kmem_cache_free(delayed_node_cache, node);
|
|
|
|
return ERR_PTR(ret);
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_lock(&root->inode_lock);
|
2011-05-22 13:11:22 +02:00
|
|
|
ret = radix_tree_insert(&root->delayed_nodes_tree, ino, node);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
if (ret == -EEXIST) {
|
|
|
|
kmem_cache_free(delayed_node_cache, node);
|
|
|
|
spin_unlock(&root->inode_lock);
|
|
|
|
radix_tree_preload_end();
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
btrfs_inode->delayed_node = node;
|
|
|
|
spin_unlock(&root->inode_lock);
|
|
|
|
radix_tree_preload_end();
|
|
|
|
|
|
|
|
return node;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Call it when holding delayed_node->mutex
|
|
|
|
*
|
|
|
|
* If mod = 1, add this node into the prepared list.
|
|
|
|
*/
|
|
|
|
static void btrfs_queue_delayed_node(struct btrfs_delayed_root *root,
|
|
|
|
struct btrfs_delayed_node *node,
|
|
|
|
int mod)
|
|
|
|
{
|
|
|
|
spin_lock(&root->lock);
|
|
|
|
if (node->in_list) {
|
|
|
|
if (!list_empty(&node->p_list))
|
|
|
|
list_move_tail(&node->p_list, &root->prepare_list);
|
|
|
|
else if (mod)
|
|
|
|
list_add_tail(&node->p_list, &root->prepare_list);
|
|
|
|
} else {
|
|
|
|
list_add_tail(&node->n_list, &root->node_list);
|
|
|
|
list_add_tail(&node->p_list, &root->prepare_list);
|
|
|
|
atomic_inc(&node->refs); /* inserted into list */
|
|
|
|
root->nodes++;
|
|
|
|
node->in_list = 1;
|
|
|
|
}
|
|
|
|
spin_unlock(&root->lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Call it when holding delayed_node->mutex */
|
|
|
|
static void btrfs_dequeue_delayed_node(struct btrfs_delayed_root *root,
|
|
|
|
struct btrfs_delayed_node *node)
|
|
|
|
{
|
|
|
|
spin_lock(&root->lock);
|
|
|
|
if (node->in_list) {
|
|
|
|
root->nodes--;
|
|
|
|
atomic_dec(&node->refs); /* not in the list */
|
|
|
|
list_del_init(&node->n_list);
|
|
|
|
if (!list_empty(&node->p_list))
|
|
|
|
list_del_init(&node->p_list);
|
|
|
|
node->in_list = 0;
|
|
|
|
}
|
|
|
|
spin_unlock(&root->lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
struct btrfs_delayed_node *btrfs_first_delayed_node(
|
|
|
|
struct btrfs_delayed_root *delayed_root)
|
|
|
|
{
|
|
|
|
struct list_head *p;
|
|
|
|
struct btrfs_delayed_node *node = NULL;
|
|
|
|
|
|
|
|
spin_lock(&delayed_root->lock);
|
|
|
|
if (list_empty(&delayed_root->node_list))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
p = delayed_root->node_list.next;
|
|
|
|
node = list_entry(p, struct btrfs_delayed_node, n_list);
|
|
|
|
atomic_inc(&node->refs);
|
|
|
|
out:
|
|
|
|
spin_unlock(&delayed_root->lock);
|
|
|
|
|
|
|
|
return node;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct btrfs_delayed_node *btrfs_next_delayed_node(
|
|
|
|
struct btrfs_delayed_node *node)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_root *delayed_root;
|
|
|
|
struct list_head *p;
|
|
|
|
struct btrfs_delayed_node *next = NULL;
|
|
|
|
|
|
|
|
delayed_root = node->root->fs_info->delayed_root;
|
|
|
|
spin_lock(&delayed_root->lock);
|
|
|
|
if (!node->in_list) { /* not in the list */
|
|
|
|
if (list_empty(&delayed_root->node_list))
|
|
|
|
goto out;
|
|
|
|
p = delayed_root->node_list.next;
|
|
|
|
} else if (list_is_last(&node->n_list, &delayed_root->node_list))
|
|
|
|
goto out;
|
|
|
|
else
|
|
|
|
p = node->n_list.next;
|
|
|
|
|
|
|
|
next = list_entry(p, struct btrfs_delayed_node, n_list);
|
|
|
|
atomic_inc(&next->refs);
|
|
|
|
out:
|
|
|
|
spin_unlock(&delayed_root->lock);
|
|
|
|
|
|
|
|
return next;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __btrfs_release_delayed_node(
|
|
|
|
struct btrfs_delayed_node *delayed_node,
|
|
|
|
int mod)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_root *delayed_root;
|
|
|
|
|
|
|
|
if (!delayed_node)
|
|
|
|
return;
|
|
|
|
|
|
|
|
delayed_root = delayed_node->root->fs_info->delayed_root;
|
|
|
|
|
|
|
|
mutex_lock(&delayed_node->mutex);
|
|
|
|
if (delayed_node->count)
|
|
|
|
btrfs_queue_delayed_node(delayed_root, delayed_node, mod);
|
|
|
|
else
|
|
|
|
btrfs_dequeue_delayed_node(delayed_root, delayed_node);
|
|
|
|
mutex_unlock(&delayed_node->mutex);
|
|
|
|
|
|
|
|
if (atomic_dec_and_test(&delayed_node->refs)) {
|
|
|
|
struct btrfs_root *root = delayed_node->root;
|
|
|
|
spin_lock(&root->inode_lock);
|
|
|
|
if (atomic_read(&delayed_node->refs) == 0) {
|
|
|
|
radix_tree_delete(&root->delayed_nodes_tree,
|
|
|
|
delayed_node->inode_id);
|
|
|
|
kmem_cache_free(delayed_node_cache, delayed_node);
|
|
|
|
}
|
|
|
|
spin_unlock(&root->inode_lock);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void btrfs_release_delayed_node(struct btrfs_delayed_node *node)
|
|
|
|
{
|
|
|
|
__btrfs_release_delayed_node(node, 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
struct btrfs_delayed_node *btrfs_first_prepared_delayed_node(
|
|
|
|
struct btrfs_delayed_root *delayed_root)
|
|
|
|
{
|
|
|
|
struct list_head *p;
|
|
|
|
struct btrfs_delayed_node *node = NULL;
|
|
|
|
|
|
|
|
spin_lock(&delayed_root->lock);
|
|
|
|
if (list_empty(&delayed_root->prepare_list))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
p = delayed_root->prepare_list.next;
|
|
|
|
list_del_init(p);
|
|
|
|
node = list_entry(p, struct btrfs_delayed_node, p_list);
|
|
|
|
atomic_inc(&node->refs);
|
|
|
|
out:
|
|
|
|
spin_unlock(&delayed_root->lock);
|
|
|
|
|
|
|
|
return node;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void btrfs_release_prepared_delayed_node(
|
|
|
|
struct btrfs_delayed_node *node)
|
|
|
|
{
|
|
|
|
__btrfs_release_delayed_node(node, 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
struct btrfs_delayed_item *btrfs_alloc_delayed_item(u32 data_len)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_item *item;
|
|
|
|
item = kmalloc(sizeof(*item) + data_len, GFP_NOFS);
|
|
|
|
if (item) {
|
|
|
|
item->data_len = data_len;
|
|
|
|
item->ins_or_del = 0;
|
|
|
|
item->bytes_reserved = 0;
|
|
|
|
item->delayed_node = NULL;
|
|
|
|
atomic_set(&item->refs, 1);
|
|
|
|
}
|
|
|
|
return item;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* __btrfs_lookup_delayed_item - look up the delayed item by key
|
|
|
|
* @delayed_node: pointer to the delayed node
|
|
|
|
* @key: the key to look up
|
|
|
|
* @prev: used to store the prev item if the right item isn't found
|
|
|
|
* @next: used to store the next item if the right item isn't found
|
|
|
|
*
|
|
|
|
* Note: if we don't find the right item, we will return the prev item and
|
|
|
|
* the next item.
|
|
|
|
*/
|
|
|
|
static struct btrfs_delayed_item *__btrfs_lookup_delayed_item(
|
|
|
|
struct rb_root *root,
|
|
|
|
struct btrfs_key *key,
|
|
|
|
struct btrfs_delayed_item **prev,
|
|
|
|
struct btrfs_delayed_item **next)
|
|
|
|
{
|
|
|
|
struct rb_node *node, *prev_node = NULL;
|
|
|
|
struct btrfs_delayed_item *delayed_item = NULL;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
node = root->rb_node;
|
|
|
|
|
|
|
|
while (node) {
|
|
|
|
delayed_item = rb_entry(node, struct btrfs_delayed_item,
|
|
|
|
rb_node);
|
|
|
|
prev_node = node;
|
|
|
|
ret = btrfs_comp_cpu_keys(&delayed_item->key, key);
|
|
|
|
if (ret < 0)
|
|
|
|
node = node->rb_right;
|
|
|
|
else if (ret > 0)
|
|
|
|
node = node->rb_left;
|
|
|
|
else
|
|
|
|
return delayed_item;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (prev) {
|
|
|
|
if (!prev_node)
|
|
|
|
*prev = NULL;
|
|
|
|
else if (ret < 0)
|
|
|
|
*prev = delayed_item;
|
|
|
|
else if ((node = rb_prev(prev_node)) != NULL) {
|
|
|
|
*prev = rb_entry(node, struct btrfs_delayed_item,
|
|
|
|
rb_node);
|
|
|
|
} else
|
|
|
|
*prev = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (next) {
|
|
|
|
if (!prev_node)
|
|
|
|
*next = NULL;
|
|
|
|
else if (ret > 0)
|
|
|
|
*next = delayed_item;
|
|
|
|
else if ((node = rb_next(prev_node)) != NULL) {
|
|
|
|
*next = rb_entry(node, struct btrfs_delayed_item,
|
|
|
|
rb_node);
|
|
|
|
} else
|
|
|
|
*next = NULL;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct btrfs_delayed_item *__btrfs_lookup_delayed_insertion_item(
|
|
|
|
struct btrfs_delayed_node *delayed_node,
|
|
|
|
struct btrfs_key *key)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_item *item;
|
|
|
|
|
|
|
|
item = __btrfs_lookup_delayed_item(&delayed_node->ins_root, key,
|
|
|
|
NULL, NULL);
|
|
|
|
return item;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct btrfs_delayed_item *__btrfs_lookup_delayed_deletion_item(
|
|
|
|
struct btrfs_delayed_node *delayed_node,
|
|
|
|
struct btrfs_key *key)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_item *item;
|
|
|
|
|
|
|
|
item = __btrfs_lookup_delayed_item(&delayed_node->del_root, key,
|
|
|
|
NULL, NULL);
|
|
|
|
return item;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct btrfs_delayed_item *__btrfs_search_delayed_insertion_item(
|
|
|
|
struct btrfs_delayed_node *delayed_node,
|
|
|
|
struct btrfs_key *key)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_item *item, *next;
|
|
|
|
|
|
|
|
item = __btrfs_lookup_delayed_item(&delayed_node->ins_root, key,
|
|
|
|
NULL, &next);
|
|
|
|
if (!item)
|
|
|
|
item = next;
|
|
|
|
|
|
|
|
return item;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct btrfs_delayed_item *__btrfs_search_delayed_deletion_item(
|
|
|
|
struct btrfs_delayed_node *delayed_node,
|
|
|
|
struct btrfs_key *key)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_item *item, *next;
|
|
|
|
|
|
|
|
item = __btrfs_lookup_delayed_item(&delayed_node->del_root, key,
|
|
|
|
NULL, &next);
|
|
|
|
if (!item)
|
|
|
|
item = next;
|
|
|
|
|
|
|
|
return item;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __btrfs_add_delayed_item(struct btrfs_delayed_node *delayed_node,
|
|
|
|
struct btrfs_delayed_item *ins,
|
|
|
|
int action)
|
|
|
|
{
|
|
|
|
struct rb_node **p, *node;
|
|
|
|
struct rb_node *parent_node = NULL;
|
|
|
|
struct rb_root *root;
|
|
|
|
struct btrfs_delayed_item *item;
|
|
|
|
int cmp;
|
|
|
|
|
|
|
|
if (action == BTRFS_DELAYED_INSERTION_ITEM)
|
|
|
|
root = &delayed_node->ins_root;
|
|
|
|
else if (action == BTRFS_DELAYED_DELETION_ITEM)
|
|
|
|
root = &delayed_node->del_root;
|
|
|
|
else
|
|
|
|
BUG();
|
|
|
|
p = &root->rb_node;
|
|
|
|
node = &ins->rb_node;
|
|
|
|
|
|
|
|
while (*p) {
|
|
|
|
parent_node = *p;
|
|
|
|
item = rb_entry(parent_node, struct btrfs_delayed_item,
|
|
|
|
rb_node);
|
|
|
|
|
|
|
|
cmp = btrfs_comp_cpu_keys(&item->key, &ins->key);
|
|
|
|
if (cmp < 0)
|
|
|
|
p = &(*p)->rb_right;
|
|
|
|
else if (cmp > 0)
|
|
|
|
p = &(*p)->rb_left;
|
|
|
|
else
|
|
|
|
return -EEXIST;
|
|
|
|
}
|
|
|
|
|
|
|
|
rb_link_node(node, parent_node, p);
|
|
|
|
rb_insert_color(node, root);
|
|
|
|
ins->delayed_node = delayed_node;
|
|
|
|
ins->ins_or_del = action;
|
|
|
|
|
|
|
|
if (ins->key.type == BTRFS_DIR_INDEX_KEY &&
|
|
|
|
action == BTRFS_DELAYED_INSERTION_ITEM &&
|
|
|
|
ins->key.offset >= delayed_node->index_cnt)
|
|
|
|
delayed_node->index_cnt = ins->key.offset + 1;
|
|
|
|
|
|
|
|
delayed_node->count++;
|
|
|
|
atomic_inc(&delayed_node->root->fs_info->delayed_root->items);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __btrfs_add_delayed_insertion_item(struct btrfs_delayed_node *node,
|
|
|
|
struct btrfs_delayed_item *item)
|
|
|
|
{
|
|
|
|
return __btrfs_add_delayed_item(node, item,
|
|
|
|
BTRFS_DELAYED_INSERTION_ITEM);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __btrfs_add_delayed_deletion_item(struct btrfs_delayed_node *node,
|
|
|
|
struct btrfs_delayed_item *item)
|
|
|
|
{
|
|
|
|
return __btrfs_add_delayed_item(node, item,
|
|
|
|
BTRFS_DELAYED_DELETION_ITEM);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __btrfs_remove_delayed_item(struct btrfs_delayed_item *delayed_item)
|
|
|
|
{
|
|
|
|
struct rb_root *root;
|
|
|
|
struct btrfs_delayed_root *delayed_root;
|
|
|
|
|
|
|
|
delayed_root = delayed_item->delayed_node->root->fs_info->delayed_root;
|
|
|
|
|
|
|
|
BUG_ON(!delayed_root);
|
|
|
|
BUG_ON(delayed_item->ins_or_del != BTRFS_DELAYED_DELETION_ITEM &&
|
|
|
|
delayed_item->ins_or_del != BTRFS_DELAYED_INSERTION_ITEM);
|
|
|
|
|
|
|
|
if (delayed_item->ins_or_del == BTRFS_DELAYED_INSERTION_ITEM)
|
|
|
|
root = &delayed_item->delayed_node->ins_root;
|
|
|
|
else
|
|
|
|
root = &delayed_item->delayed_node->del_root;
|
|
|
|
|
|
|
|
rb_erase(&delayed_item->rb_node, root);
|
|
|
|
delayed_item->delayed_node->count--;
|
|
|
|
atomic_dec(&delayed_root->items);
|
|
|
|
if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND &&
|
|
|
|
waitqueue_active(&delayed_root->wait))
|
|
|
|
wake_up(&delayed_root->wait);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void btrfs_release_delayed_item(struct btrfs_delayed_item *item)
|
|
|
|
{
|
|
|
|
if (item) {
|
|
|
|
__btrfs_remove_delayed_item(item);
|
|
|
|
if (atomic_dec_and_test(&item->refs))
|
|
|
|
kfree(item);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
struct btrfs_delayed_item *__btrfs_first_delayed_insertion_item(
|
|
|
|
struct btrfs_delayed_node *delayed_node)
|
|
|
|
{
|
|
|
|
struct rb_node *p;
|
|
|
|
struct btrfs_delayed_item *item = NULL;
|
|
|
|
|
|
|
|
p = rb_first(&delayed_node->ins_root);
|
|
|
|
if (p)
|
|
|
|
item = rb_entry(p, struct btrfs_delayed_item, rb_node);
|
|
|
|
|
|
|
|
return item;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct btrfs_delayed_item *__btrfs_first_delayed_deletion_item(
|
|
|
|
struct btrfs_delayed_node *delayed_node)
|
|
|
|
{
|
|
|
|
struct rb_node *p;
|
|
|
|
struct btrfs_delayed_item *item = NULL;
|
|
|
|
|
|
|
|
p = rb_first(&delayed_node->del_root);
|
|
|
|
if (p)
|
|
|
|
item = rb_entry(p, struct btrfs_delayed_item, rb_node);
|
|
|
|
|
|
|
|
return item;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct btrfs_delayed_item *__btrfs_next_delayed_item(
|
|
|
|
struct btrfs_delayed_item *item)
|
|
|
|
{
|
|
|
|
struct rb_node *p;
|
|
|
|
struct btrfs_delayed_item *next = NULL;
|
|
|
|
|
|
|
|
p = rb_next(&item->rb_node);
|
|
|
|
if (p)
|
|
|
|
next = rb_entry(p, struct btrfs_delayed_item, rb_node);
|
|
|
|
|
|
|
|
return next;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline struct btrfs_root *btrfs_get_fs_root(struct btrfs_root *root,
|
|
|
|
u64 root_id)
|
|
|
|
{
|
|
|
|
struct btrfs_key root_key;
|
|
|
|
|
|
|
|
if (root->objectid == root_id)
|
|
|
|
return root;
|
|
|
|
|
|
|
|
root_key.objectid = root_id;
|
|
|
|
root_key.type = BTRFS_ROOT_ITEM_KEY;
|
|
|
|
root_key.offset = (u64)-1;
|
|
|
|
return btrfs_read_fs_root_no_name(root->fs_info, &root_key);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int btrfs_delayed_item_reserve_metadata(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root,
|
|
|
|
struct btrfs_delayed_item *item)
|
|
|
|
{
|
|
|
|
struct btrfs_block_rsv *src_rsv;
|
|
|
|
struct btrfs_block_rsv *dst_rsv;
|
|
|
|
u64 num_bytes;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
if (!trans->bytes_reserved)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
src_rsv = trans->block_rsv;
|
|
|
|
dst_rsv = &root->fs_info->global_block_rsv;
|
|
|
|
|
|
|
|
num_bytes = btrfs_calc_trans_metadata_size(root, 1);
|
|
|
|
ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
|
2011-06-15 12:47:30 +02:00
|
|
|
if (!ret)
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
item->bytes_reserved = num_bytes;
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void btrfs_delayed_item_release_metadata(struct btrfs_root *root,
|
|
|
|
struct btrfs_delayed_item *item)
|
|
|
|
{
|
2011-06-15 12:47:30 +02:00
|
|
|
struct btrfs_block_rsv *rsv;
|
|
|
|
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
if (!item->bytes_reserved)
|
|
|
|
return;
|
|
|
|
|
2011-06-15 12:47:30 +02:00
|
|
|
rsv = &root->fs_info->global_block_rsv;
|
|
|
|
btrfs_block_rsv_release(root, rsv,
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
item->bytes_reserved);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int btrfs_delayed_inode_reserve_metadata(
|
|
|
|
struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root,
|
|
|
|
struct btrfs_delayed_node *node)
|
|
|
|
{
|
|
|
|
struct btrfs_block_rsv *src_rsv;
|
|
|
|
struct btrfs_block_rsv *dst_rsv;
|
|
|
|
u64 num_bytes;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
if (!trans->bytes_reserved)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
src_rsv = trans->block_rsv;
|
|
|
|
dst_rsv = &root->fs_info->global_block_rsv;
|
|
|
|
|
|
|
|
num_bytes = btrfs_calc_trans_metadata_size(root, 1);
|
|
|
|
ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
|
|
|
|
if (!ret)
|
|
|
|
node->bytes_reserved = num_bytes;
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void btrfs_delayed_inode_release_metadata(struct btrfs_root *root,
|
|
|
|
struct btrfs_delayed_node *node)
|
|
|
|
{
|
|
|
|
struct btrfs_block_rsv *rsv;
|
|
|
|
|
|
|
|
if (!node->bytes_reserved)
|
|
|
|
return;
|
|
|
|
|
|
|
|
rsv = &root->fs_info->global_block_rsv;
|
|
|
|
btrfs_block_rsv_release(root, rsv,
|
|
|
|
node->bytes_reserved);
|
|
|
|
node->bytes_reserved = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This helper will insert some continuous items into the same leaf according
|
|
|
|
* to the free space of the leaf.
|
|
|
|
*/
|
|
|
|
static int btrfs_batch_insert_items(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root,
|
|
|
|
struct btrfs_path *path,
|
|
|
|
struct btrfs_delayed_item *item)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_item *curr, *next;
|
|
|
|
int free_space;
|
|
|
|
int total_data_size = 0, total_size = 0;
|
|
|
|
struct extent_buffer *leaf;
|
|
|
|
char *data_ptr;
|
|
|
|
struct btrfs_key *keys;
|
|
|
|
u32 *data_size;
|
|
|
|
struct list_head head;
|
|
|
|
int slot;
|
|
|
|
int nitems;
|
|
|
|
int i;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
BUG_ON(!path->nodes[0]);
|
|
|
|
|
|
|
|
leaf = path->nodes[0];
|
|
|
|
free_space = btrfs_leaf_free_space(root, leaf);
|
|
|
|
INIT_LIST_HEAD(&head);
|
|
|
|
|
|
|
|
next = item;
|
2011-06-03 07:13:45 +02:00
|
|
|
nitems = 0;
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* count the number of the continuous items that we can insert in batch
|
|
|
|
*/
|
|
|
|
while (total_size + next->data_len + sizeof(struct btrfs_item) <=
|
|
|
|
free_space) {
|
|
|
|
total_data_size += next->data_len;
|
|
|
|
total_size += next->data_len + sizeof(struct btrfs_item);
|
|
|
|
list_add_tail(&next->tree_list, &head);
|
|
|
|
nitems++;
|
|
|
|
|
|
|
|
curr = next;
|
|
|
|
next = __btrfs_next_delayed_item(curr);
|
|
|
|
if (!next)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (!btrfs_is_continuous_delayed_item(curr, next))
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!nitems) {
|
|
|
|
ret = 0;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* we need allocate some memory space, but it might cause the task
|
|
|
|
* to sleep, so we set all locked nodes in the path to blocking locks
|
|
|
|
* first.
|
|
|
|
*/
|
|
|
|
btrfs_set_path_blocking(path);
|
|
|
|
|
|
|
|
keys = kmalloc(sizeof(struct btrfs_key) * nitems, GFP_NOFS);
|
|
|
|
if (!keys) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
data_size = kmalloc(sizeof(u32) * nitems, GFP_NOFS);
|
|
|
|
if (!data_size) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* get keys of all the delayed items */
|
|
|
|
i = 0;
|
|
|
|
list_for_each_entry(next, &head, tree_list) {
|
|
|
|
keys[i] = next->key;
|
|
|
|
data_size[i] = next->data_len;
|
|
|
|
i++;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* reset all the locked nodes in the patch to spinning locks. */
|
|
|
|
btrfs_clear_path_blocking(path, NULL);
|
|
|
|
|
|
|
|
/* insert the keys of the items */
|
|
|
|
ret = setup_items_for_insert(trans, root, path, keys, data_size,
|
|
|
|
total_data_size, total_size, nitems);
|
|
|
|
if (ret)
|
|
|
|
goto error;
|
|
|
|
|
|
|
|
/* insert the dir index items */
|
|
|
|
slot = path->slots[0];
|
|
|
|
list_for_each_entry_safe(curr, next, &head, tree_list) {
|
|
|
|
data_ptr = btrfs_item_ptr(leaf, slot, char);
|
|
|
|
write_extent_buffer(leaf, &curr->data,
|
|
|
|
(unsigned long)data_ptr,
|
|
|
|
curr->data_len);
|
|
|
|
slot++;
|
|
|
|
|
|
|
|
btrfs_delayed_item_release_metadata(root, curr);
|
|
|
|
|
|
|
|
list_del(&curr->tree_list);
|
|
|
|
btrfs_release_delayed_item(curr);
|
|
|
|
}
|
|
|
|
|
|
|
|
error:
|
|
|
|
kfree(data_size);
|
|
|
|
kfree(keys);
|
|
|
|
out:
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This helper can just do simple insertion that needn't extend item for new
|
|
|
|
* data, such as directory name index insertion, inode insertion.
|
|
|
|
*/
|
|
|
|
static int btrfs_insert_delayed_item(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root,
|
|
|
|
struct btrfs_path *path,
|
|
|
|
struct btrfs_delayed_item *delayed_item)
|
|
|
|
{
|
|
|
|
struct extent_buffer *leaf;
|
|
|
|
struct btrfs_item *item;
|
|
|
|
char *ptr;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = btrfs_insert_empty_item(trans, root, path, &delayed_item->key,
|
|
|
|
delayed_item->data_len);
|
|
|
|
if (ret < 0 && ret != -EEXIST)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
leaf = path->nodes[0];
|
|
|
|
|
|
|
|
item = btrfs_item_nr(leaf, path->slots[0]);
|
|
|
|
ptr = btrfs_item_ptr(leaf, path->slots[0], char);
|
|
|
|
|
|
|
|
write_extent_buffer(leaf, delayed_item->data, (unsigned long)ptr,
|
|
|
|
delayed_item->data_len);
|
|
|
|
btrfs_mark_buffer_dirty(leaf);
|
|
|
|
|
|
|
|
btrfs_delayed_item_release_metadata(root, delayed_item);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* we insert an item first, then if there are some continuous items, we try
|
|
|
|
* to insert those items into the same leaf.
|
|
|
|
*/
|
|
|
|
static int btrfs_insert_delayed_items(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_path *path,
|
|
|
|
struct btrfs_root *root,
|
|
|
|
struct btrfs_delayed_node *node)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_item *curr, *prev;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
do_again:
|
|
|
|
mutex_lock(&node->mutex);
|
|
|
|
curr = __btrfs_first_delayed_insertion_item(node);
|
|
|
|
if (!curr)
|
|
|
|
goto insert_end;
|
|
|
|
|
|
|
|
ret = btrfs_insert_delayed_item(trans, root, path, curr);
|
|
|
|
if (ret < 0) {
|
2011-05-22 18:33:42 +02:00
|
|
|
btrfs_release_path(path);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
goto insert_end;
|
|
|
|
}
|
|
|
|
|
|
|
|
prev = curr;
|
|
|
|
curr = __btrfs_next_delayed_item(prev);
|
|
|
|
if (curr && btrfs_is_continuous_delayed_item(prev, curr)) {
|
|
|
|
/* insert the continuous items into the same leaf */
|
|
|
|
path->slots[0]++;
|
|
|
|
btrfs_batch_insert_items(trans, root, path, curr);
|
|
|
|
}
|
|
|
|
btrfs_release_delayed_item(prev);
|
|
|
|
btrfs_mark_buffer_dirty(path->nodes[0]);
|
|
|
|
|
2011-05-22 18:33:42 +02:00
|
|
|
btrfs_release_path(path);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
mutex_unlock(&node->mutex);
|
|
|
|
goto do_again;
|
|
|
|
|
|
|
|
insert_end:
|
|
|
|
mutex_unlock(&node->mutex);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int btrfs_batch_delete_items(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root,
|
|
|
|
struct btrfs_path *path,
|
|
|
|
struct btrfs_delayed_item *item)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_item *curr, *next;
|
|
|
|
struct extent_buffer *leaf;
|
|
|
|
struct btrfs_key key;
|
|
|
|
struct list_head head;
|
|
|
|
int nitems, i, last_item;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
BUG_ON(!path->nodes[0]);
|
|
|
|
|
|
|
|
leaf = path->nodes[0];
|
|
|
|
|
|
|
|
i = path->slots[0];
|
|
|
|
last_item = btrfs_header_nritems(leaf) - 1;
|
|
|
|
if (i > last_item)
|
|
|
|
return -ENOENT; /* FIXME: Is errno suitable? */
|
|
|
|
|
|
|
|
next = item;
|
|
|
|
INIT_LIST_HEAD(&head);
|
|
|
|
btrfs_item_key_to_cpu(leaf, &key, i);
|
|
|
|
nitems = 0;
|
|
|
|
/*
|
|
|
|
* count the number of the dir index items that we can delete in batch
|
|
|
|
*/
|
|
|
|
while (btrfs_comp_cpu_keys(&next->key, &key) == 0) {
|
|
|
|
list_add_tail(&next->tree_list, &head);
|
|
|
|
nitems++;
|
|
|
|
|
|
|
|
curr = next;
|
|
|
|
next = __btrfs_next_delayed_item(curr);
|
|
|
|
if (!next)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (!btrfs_is_continuous_delayed_item(curr, next))
|
|
|
|
break;
|
|
|
|
|
|
|
|
i++;
|
|
|
|
if (i > last_item)
|
|
|
|
break;
|
|
|
|
btrfs_item_key_to_cpu(leaf, &key, i);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!nitems)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
ret = btrfs_del_items(trans, root, path, path->slots[0], nitems);
|
|
|
|
if (ret)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
list_for_each_entry_safe(curr, next, &head, tree_list) {
|
|
|
|
btrfs_delayed_item_release_metadata(root, curr);
|
|
|
|
list_del(&curr->tree_list);
|
|
|
|
btrfs_release_delayed_item(curr);
|
|
|
|
}
|
|
|
|
|
|
|
|
out:
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int btrfs_delete_delayed_items(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_path *path,
|
|
|
|
struct btrfs_root *root,
|
|
|
|
struct btrfs_delayed_node *node)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_item *curr, *prev;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
do_again:
|
|
|
|
mutex_lock(&node->mutex);
|
|
|
|
curr = __btrfs_first_delayed_deletion_item(node);
|
|
|
|
if (!curr)
|
|
|
|
goto delete_fail;
|
|
|
|
|
|
|
|
ret = btrfs_search_slot(trans, root, &curr->key, path, -1, 1);
|
|
|
|
if (ret < 0)
|
|
|
|
goto delete_fail;
|
|
|
|
else if (ret > 0) {
|
|
|
|
/*
|
|
|
|
* can't find the item which the node points to, so this node
|
|
|
|
* is invalid, just drop it.
|
|
|
|
*/
|
|
|
|
prev = curr;
|
|
|
|
curr = __btrfs_next_delayed_item(prev);
|
|
|
|
btrfs_release_delayed_item(prev);
|
|
|
|
ret = 0;
|
2011-05-22 18:33:42 +02:00
|
|
|
btrfs_release_path(path);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
if (curr)
|
|
|
|
goto do_again;
|
|
|
|
else
|
|
|
|
goto delete_fail;
|
|
|
|
}
|
|
|
|
|
|
|
|
btrfs_batch_delete_items(trans, root, path, curr);
|
2011-05-22 18:33:42 +02:00
|
|
|
btrfs_release_path(path);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
mutex_unlock(&node->mutex);
|
|
|
|
goto do_again;
|
|
|
|
|
|
|
|
delete_fail:
|
2011-05-22 18:33:42 +02:00
|
|
|
btrfs_release_path(path);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
mutex_unlock(&node->mutex);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void btrfs_release_delayed_inode(struct btrfs_delayed_node *delayed_node)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_root *delayed_root;
|
|
|
|
|
|
|
|
if (delayed_node && delayed_node->inode_dirty) {
|
|
|
|
BUG_ON(!delayed_node->root);
|
|
|
|
delayed_node->inode_dirty = 0;
|
|
|
|
delayed_node->count--;
|
|
|
|
|
|
|
|
delayed_root = delayed_node->root->fs_info->delayed_root;
|
|
|
|
atomic_dec(&delayed_root->items);
|
|
|
|
if (atomic_read(&delayed_root->items) <
|
|
|
|
BTRFS_DELAYED_BACKGROUND &&
|
|
|
|
waitqueue_active(&delayed_root->wait))
|
|
|
|
wake_up(&delayed_root->wait);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root,
|
|
|
|
struct btrfs_path *path,
|
|
|
|
struct btrfs_delayed_node *node)
|
|
|
|
{
|
|
|
|
struct btrfs_key key;
|
|
|
|
struct btrfs_inode_item *inode_item;
|
|
|
|
struct extent_buffer *leaf;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
mutex_lock(&node->mutex);
|
|
|
|
if (!node->inode_dirty) {
|
|
|
|
mutex_unlock(&node->mutex);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
key.objectid = node->inode_id;
|
|
|
|
btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
|
|
|
|
key.offset = 0;
|
|
|
|
ret = btrfs_lookup_inode(trans, root, path, &key, 1);
|
|
|
|
if (ret > 0) {
|
2011-05-22 18:33:42 +02:00
|
|
|
btrfs_release_path(path);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
mutex_unlock(&node->mutex);
|
|
|
|
return -ENOENT;
|
|
|
|
} else if (ret < 0) {
|
|
|
|
mutex_unlock(&node->mutex);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
btrfs_unlock_up_safe(path, 1);
|
|
|
|
leaf = path->nodes[0];
|
|
|
|
inode_item = btrfs_item_ptr(leaf, path->slots[0],
|
|
|
|
struct btrfs_inode_item);
|
|
|
|
write_extent_buffer(leaf, &node->inode_item, (unsigned long)inode_item,
|
|
|
|
sizeof(struct btrfs_inode_item));
|
|
|
|
btrfs_mark_buffer_dirty(leaf);
|
2011-05-22 18:33:42 +02:00
|
|
|
btrfs_release_path(path);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
|
|
|
|
btrfs_delayed_inode_release_metadata(root, node);
|
|
|
|
btrfs_release_delayed_inode(node);
|
|
|
|
mutex_unlock(&node->mutex);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Called when committing the transaction. */
|
|
|
|
int btrfs_run_delayed_items(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_root *delayed_root;
|
|
|
|
struct btrfs_delayed_node *curr_node, *prev_node;
|
|
|
|
struct btrfs_path *path;
|
2011-06-15 12:47:30 +02:00
|
|
|
struct btrfs_block_rsv *block_rsv;
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
path->leave_spinning = 1;
|
|
|
|
|
2011-06-15 12:47:30 +02:00
|
|
|
block_rsv = trans->block_rsv;
|
|
|
|
trans->block_rsv = &root->fs_info->global_block_rsv;
|
|
|
|
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
delayed_root = btrfs_get_delayed_root(root);
|
|
|
|
|
|
|
|
curr_node = btrfs_first_delayed_node(delayed_root);
|
|
|
|
while (curr_node) {
|
|
|
|
root = curr_node->root;
|
|
|
|
ret = btrfs_insert_delayed_items(trans, path, root,
|
|
|
|
curr_node);
|
|
|
|
if (!ret)
|
|
|
|
ret = btrfs_delete_delayed_items(trans, path, root,
|
|
|
|
curr_node);
|
|
|
|
if (!ret)
|
|
|
|
ret = btrfs_update_delayed_inode(trans, root, path,
|
|
|
|
curr_node);
|
|
|
|
if (ret) {
|
|
|
|
btrfs_release_delayed_node(curr_node);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
prev_node = curr_node;
|
|
|
|
curr_node = btrfs_next_delayed_node(curr_node);
|
|
|
|
btrfs_release_delayed_node(prev_node);
|
|
|
|
}
|
|
|
|
|
|
|
|
btrfs_free_path(path);
|
2011-06-15 12:47:30 +02:00
|
|
|
trans->block_rsv = block_rsv;
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_delayed_node *node)
|
|
|
|
{
|
|
|
|
struct btrfs_path *path;
|
2011-06-15 12:47:30 +02:00
|
|
|
struct btrfs_block_rsv *block_rsv;
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
path->leave_spinning = 1;
|
|
|
|
|
2011-06-15 12:47:30 +02:00
|
|
|
block_rsv = trans->block_rsv;
|
|
|
|
trans->block_rsv = &node->root->fs_info->global_block_rsv;
|
|
|
|
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
ret = btrfs_insert_delayed_items(trans, path, node->root, node);
|
|
|
|
if (!ret)
|
|
|
|
ret = btrfs_delete_delayed_items(trans, path, node->root, node);
|
|
|
|
if (!ret)
|
|
|
|
ret = btrfs_update_delayed_inode(trans, node->root, path, node);
|
|
|
|
btrfs_free_path(path);
|
|
|
|
|
2011-06-15 12:47:30 +02:00
|
|
|
trans->block_rsv = block_rsv;
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
|
|
|
|
struct inode *inode)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
if (!delayed_node)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
mutex_lock(&delayed_node->mutex);
|
|
|
|
if (!delayed_node->count) {
|
|
|
|
mutex_unlock(&delayed_node->mutex);
|
|
|
|
btrfs_release_delayed_node(delayed_node);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
mutex_unlock(&delayed_node->mutex);
|
|
|
|
|
|
|
|
ret = __btrfs_commit_inode_delayed_items(trans, delayed_node);
|
|
|
|
btrfs_release_delayed_node(delayed_node);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
void btrfs_remove_delayed_node(struct inode *inode)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_node *delayed_node;
|
|
|
|
|
|
|
|
delayed_node = ACCESS_ONCE(BTRFS_I(inode)->delayed_node);
|
|
|
|
if (!delayed_node)
|
|
|
|
return;
|
|
|
|
|
|
|
|
BTRFS_I(inode)->delayed_node = NULL;
|
|
|
|
btrfs_release_delayed_node(delayed_node);
|
|
|
|
}
|
|
|
|
|
|
|
|
struct btrfs_async_delayed_node {
|
|
|
|
struct btrfs_root *root;
|
|
|
|
struct btrfs_delayed_node *delayed_node;
|
|
|
|
struct btrfs_work work;
|
|
|
|
};
|
|
|
|
|
|
|
|
static void btrfs_async_run_delayed_node_done(struct btrfs_work *work)
|
|
|
|
{
|
|
|
|
struct btrfs_async_delayed_node *async_node;
|
|
|
|
struct btrfs_trans_handle *trans;
|
|
|
|
struct btrfs_path *path;
|
|
|
|
struct btrfs_delayed_node *delayed_node = NULL;
|
|
|
|
struct btrfs_root *root;
|
2011-06-15 12:47:30 +02:00
|
|
|
struct btrfs_block_rsv *block_rsv;
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
unsigned long nr = 0;
|
|
|
|
int need_requeue = 0;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
async_node = container_of(work, struct btrfs_async_delayed_node, work);
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
goto out;
|
|
|
|
path->leave_spinning = 1;
|
|
|
|
|
|
|
|
delayed_node = async_node->delayed_node;
|
|
|
|
root = delayed_node->root;
|
|
|
|
|
2011-05-28 13:00:39 +02:00
|
|
|
trans = btrfs_join_transaction(root);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
if (IS_ERR(trans))
|
|
|
|
goto free_path;
|
|
|
|
|
2011-06-15 12:47:30 +02:00
|
|
|
block_rsv = trans->block_rsv;
|
|
|
|
trans->block_rsv = &root->fs_info->global_block_rsv;
|
|
|
|
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
ret = btrfs_insert_delayed_items(trans, path, root, delayed_node);
|
|
|
|
if (!ret)
|
|
|
|
ret = btrfs_delete_delayed_items(trans, path, root,
|
|
|
|
delayed_node);
|
|
|
|
|
|
|
|
if (!ret)
|
|
|
|
btrfs_update_delayed_inode(trans, root, path, delayed_node);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Maybe new delayed items have been inserted, so we need requeue
|
|
|
|
* the work. Besides that, we must dequeue the empty delayed nodes
|
|
|
|
* to avoid the race between delayed items balance and the worker.
|
|
|
|
* The race like this:
|
|
|
|
* Task1 Worker thread
|
|
|
|
* count == 0, needn't requeue
|
|
|
|
* also needn't insert the
|
|
|
|
* delayed node into prepare
|
|
|
|
* list again.
|
|
|
|
* add lots of delayed items
|
|
|
|
* queue the delayed node
|
|
|
|
* already in the list,
|
|
|
|
* and not in the prepare
|
|
|
|
* list, it means the delayed
|
|
|
|
* node is being dealt with
|
|
|
|
* by the worker.
|
|
|
|
* do delayed items balance
|
|
|
|
* the delayed node is being
|
|
|
|
* dealt with by the worker
|
|
|
|
* now, just wait.
|
|
|
|
* the worker goto idle.
|
|
|
|
* Task1 will sleep until the transaction is commited.
|
|
|
|
*/
|
|
|
|
mutex_lock(&delayed_node->mutex);
|
|
|
|
if (delayed_node->count)
|
|
|
|
need_requeue = 1;
|
|
|
|
else
|
|
|
|
btrfs_dequeue_delayed_node(root->fs_info->delayed_root,
|
|
|
|
delayed_node);
|
|
|
|
mutex_unlock(&delayed_node->mutex);
|
|
|
|
|
|
|
|
nr = trans->blocks_used;
|
|
|
|
|
2011-06-15 12:47:30 +02:00
|
|
|
trans->block_rsv = block_rsv;
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
btrfs_end_transaction_dmeta(trans, root);
|
|
|
|
__btrfs_btree_balance_dirty(root, nr);
|
|
|
|
free_path:
|
|
|
|
btrfs_free_path(path);
|
|
|
|
out:
|
|
|
|
if (need_requeue)
|
|
|
|
btrfs_requeue_work(&async_node->work);
|
|
|
|
else {
|
|
|
|
btrfs_release_prepared_delayed_node(delayed_node);
|
|
|
|
kfree(async_node);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int btrfs_wq_run_delayed_node(struct btrfs_delayed_root *delayed_root,
|
|
|
|
struct btrfs_root *root, int all)
|
|
|
|
{
|
|
|
|
struct btrfs_async_delayed_node *async_node;
|
|
|
|
struct btrfs_delayed_node *curr;
|
|
|
|
int count = 0;
|
|
|
|
|
|
|
|
again:
|
|
|
|
curr = btrfs_first_prepared_delayed_node(delayed_root);
|
|
|
|
if (!curr)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
async_node = kmalloc(sizeof(*async_node), GFP_NOFS);
|
|
|
|
if (!async_node) {
|
|
|
|
btrfs_release_prepared_delayed_node(curr);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
async_node->root = root;
|
|
|
|
async_node->delayed_node = curr;
|
|
|
|
|
|
|
|
async_node->work.func = btrfs_async_run_delayed_node_done;
|
|
|
|
async_node->work.flags = 0;
|
|
|
|
|
|
|
|
btrfs_queue_worker(&root->fs_info->delayed_workers, &async_node->work);
|
|
|
|
count++;
|
|
|
|
|
|
|
|
if (all || count < 4)
|
|
|
|
goto again;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2011-06-17 22:14:09 +02:00
|
|
|
void btrfs_assert_delayed_root_empty(struct btrfs_root *root)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_root *delayed_root;
|
|
|
|
delayed_root = btrfs_get_delayed_root(root);
|
|
|
|
WARN_ON(btrfs_first_delayed_node(delayed_root));
|
|
|
|
}
|
|
|
|
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
void btrfs_balance_delayed_items(struct btrfs_root *root)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_root *delayed_root;
|
|
|
|
|
|
|
|
delayed_root = btrfs_get_delayed_root(root);
|
|
|
|
|
|
|
|
if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
|
|
|
|
return;
|
|
|
|
|
|
|
|
if (atomic_read(&delayed_root->items) >= BTRFS_DELAYED_WRITEBACK) {
|
|
|
|
int ret;
|
|
|
|
ret = btrfs_wq_run_delayed_node(delayed_root, root, 1);
|
|
|
|
if (ret)
|
|
|
|
return;
|
|
|
|
|
|
|
|
wait_event_interruptible_timeout(
|
|
|
|
delayed_root->wait,
|
|
|
|
(atomic_read(&delayed_root->items) <
|
|
|
|
BTRFS_DELAYED_BACKGROUND),
|
|
|
|
HZ);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
btrfs_wq_run_delayed_node(delayed_root, root, 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_insert_delayed_dir_index(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root, const char *name,
|
|
|
|
int name_len, struct inode *dir,
|
|
|
|
struct btrfs_disk_key *disk_key, u8 type,
|
|
|
|
u64 index)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_node *delayed_node;
|
|
|
|
struct btrfs_delayed_item *delayed_item;
|
|
|
|
struct btrfs_dir_item *dir_item;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
delayed_node = btrfs_get_or_create_delayed_node(dir);
|
|
|
|
if (IS_ERR(delayed_node))
|
|
|
|
return PTR_ERR(delayed_node);
|
|
|
|
|
|
|
|
delayed_item = btrfs_alloc_delayed_item(sizeof(*dir_item) + name_len);
|
|
|
|
if (!delayed_item) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto release_node;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = btrfs_delayed_item_reserve_metadata(trans, root, delayed_item);
|
|
|
|
/*
|
|
|
|
* we have reserved enough space when we start a new transaction,
|
|
|
|
* so reserving metadata failure is impossible
|
|
|
|
*/
|
|
|
|
BUG_ON(ret);
|
|
|
|
|
2011-05-22 13:11:22 +02:00
|
|
|
delayed_item->key.objectid = btrfs_ino(dir);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
btrfs_set_key_type(&delayed_item->key, BTRFS_DIR_INDEX_KEY);
|
|
|
|
delayed_item->key.offset = index;
|
|
|
|
|
|
|
|
dir_item = (struct btrfs_dir_item *)delayed_item->data;
|
|
|
|
dir_item->location = *disk_key;
|
|
|
|
dir_item->transid = cpu_to_le64(trans->transid);
|
|
|
|
dir_item->data_len = 0;
|
|
|
|
dir_item->name_len = cpu_to_le16(name_len);
|
|
|
|
dir_item->type = type;
|
|
|
|
memcpy((char *)(dir_item + 1), name, name_len);
|
|
|
|
|
|
|
|
mutex_lock(&delayed_node->mutex);
|
|
|
|
ret = __btrfs_add_delayed_insertion_item(delayed_node, delayed_item);
|
|
|
|
if (unlikely(ret)) {
|
|
|
|
printk(KERN_ERR "err add delayed dir index item(name: %s) into "
|
|
|
|
"the insertion tree of the delayed node"
|
|
|
|
"(root id: %llu, inode id: %llu, errno: %d)\n",
|
|
|
|
name,
|
|
|
|
(unsigned long long)delayed_node->root->objectid,
|
|
|
|
(unsigned long long)delayed_node->inode_id,
|
|
|
|
ret);
|
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
mutex_unlock(&delayed_node->mutex);
|
|
|
|
|
|
|
|
release_node:
|
|
|
|
btrfs_release_delayed_node(delayed_node);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int btrfs_delete_delayed_insertion_item(struct btrfs_root *root,
|
|
|
|
struct btrfs_delayed_node *node,
|
|
|
|
struct btrfs_key *key)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_item *item;
|
|
|
|
|
|
|
|
mutex_lock(&node->mutex);
|
|
|
|
item = __btrfs_lookup_delayed_insertion_item(node, key);
|
|
|
|
if (!item) {
|
|
|
|
mutex_unlock(&node->mutex);
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
btrfs_delayed_item_release_metadata(root, item);
|
|
|
|
btrfs_release_delayed_item(item);
|
|
|
|
mutex_unlock(&node->mutex);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_delete_delayed_dir_index(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root, struct inode *dir,
|
|
|
|
u64 index)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_node *node;
|
|
|
|
struct btrfs_delayed_item *item;
|
|
|
|
struct btrfs_key item_key;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
node = btrfs_get_or_create_delayed_node(dir);
|
|
|
|
if (IS_ERR(node))
|
|
|
|
return PTR_ERR(node);
|
|
|
|
|
2011-05-22 13:11:22 +02:00
|
|
|
item_key.objectid = btrfs_ino(dir);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
btrfs_set_key_type(&item_key, BTRFS_DIR_INDEX_KEY);
|
|
|
|
item_key.offset = index;
|
|
|
|
|
|
|
|
ret = btrfs_delete_delayed_insertion_item(root, node, &item_key);
|
|
|
|
if (!ret)
|
|
|
|
goto end;
|
|
|
|
|
|
|
|
item = btrfs_alloc_delayed_item(0);
|
|
|
|
if (!item) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto end;
|
|
|
|
}
|
|
|
|
|
|
|
|
item->key = item_key;
|
|
|
|
|
|
|
|
ret = btrfs_delayed_item_reserve_metadata(trans, root, item);
|
|
|
|
/*
|
|
|
|
* we have reserved enough space when we start a new transaction,
|
|
|
|
* so reserving metadata failure is impossible.
|
|
|
|
*/
|
|
|
|
BUG_ON(ret);
|
|
|
|
|
|
|
|
mutex_lock(&node->mutex);
|
|
|
|
ret = __btrfs_add_delayed_deletion_item(node, item);
|
|
|
|
if (unlikely(ret)) {
|
|
|
|
printk(KERN_ERR "err add delayed dir index item(index: %llu) "
|
|
|
|
"into the deletion tree of the delayed node"
|
|
|
|
"(root id: %llu, inode id: %llu, errno: %d)\n",
|
|
|
|
(unsigned long long)index,
|
|
|
|
(unsigned long long)node->root->objectid,
|
|
|
|
(unsigned long long)node->inode_id,
|
|
|
|
ret);
|
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
mutex_unlock(&node->mutex);
|
|
|
|
end:
|
|
|
|
btrfs_release_delayed_node(node);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_inode_delayed_dir_index_count(struct inode *inode)
|
|
|
|
{
|
2011-06-23 09:27:13 +02:00
|
|
|
struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
|
|
|
|
if (!delayed_node)
|
|
|
|
return -ENOENT;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Since we have held i_mutex of this directory, it is impossible that
|
|
|
|
* a new directory index is added into the delayed node and index_cnt
|
|
|
|
* is updated now. So we needn't lock the delayed node.
|
|
|
|
*/
|
2011-06-23 09:27:13 +02:00
|
|
|
if (!delayed_node->index_cnt) {
|
|
|
|
btrfs_release_delayed_node(delayed_node);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
return -EINVAL;
|
2011-06-23 09:27:13 +02:00
|
|
|
}
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
|
|
|
|
BTRFS_I(inode)->index_cnt = delayed_node->index_cnt;
|
2011-06-23 09:27:13 +02:00
|
|
|
btrfs_release_delayed_node(delayed_node);
|
|
|
|
return 0;
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
void btrfs_get_delayed_items(struct inode *inode, struct list_head *ins_list,
|
|
|
|
struct list_head *del_list)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_node *delayed_node;
|
|
|
|
struct btrfs_delayed_item *item;
|
|
|
|
|
|
|
|
delayed_node = btrfs_get_delayed_node(inode);
|
|
|
|
if (!delayed_node)
|
|
|
|
return;
|
|
|
|
|
|
|
|
mutex_lock(&delayed_node->mutex);
|
|
|
|
item = __btrfs_first_delayed_insertion_item(delayed_node);
|
|
|
|
while (item) {
|
|
|
|
atomic_inc(&item->refs);
|
|
|
|
list_add_tail(&item->readdir_list, ins_list);
|
|
|
|
item = __btrfs_next_delayed_item(item);
|
|
|
|
}
|
|
|
|
|
|
|
|
item = __btrfs_first_delayed_deletion_item(delayed_node);
|
|
|
|
while (item) {
|
|
|
|
atomic_inc(&item->refs);
|
|
|
|
list_add_tail(&item->readdir_list, del_list);
|
|
|
|
item = __btrfs_next_delayed_item(item);
|
|
|
|
}
|
|
|
|
mutex_unlock(&delayed_node->mutex);
|
|
|
|
/*
|
|
|
|
* This delayed node is still cached in the btrfs inode, so refs
|
|
|
|
* must be > 1 now, and we needn't check it is going to be freed
|
|
|
|
* or not.
|
|
|
|
*
|
|
|
|
* Besides that, this function is used to read dir, we do not
|
|
|
|
* insert/delete delayed items in this period. So we also needn't
|
|
|
|
* requeue or dequeue this delayed node.
|
|
|
|
*/
|
|
|
|
atomic_dec(&delayed_node->refs);
|
|
|
|
}
|
|
|
|
|
|
|
|
void btrfs_put_delayed_items(struct list_head *ins_list,
|
|
|
|
struct list_head *del_list)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_item *curr, *next;
|
|
|
|
|
|
|
|
list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
|
|
|
|
list_del(&curr->readdir_list);
|
|
|
|
if (atomic_dec_and_test(&curr->refs))
|
|
|
|
kfree(curr);
|
|
|
|
}
|
|
|
|
|
|
|
|
list_for_each_entry_safe(curr, next, del_list, readdir_list) {
|
|
|
|
list_del(&curr->readdir_list);
|
|
|
|
if (atomic_dec_and_test(&curr->refs))
|
|
|
|
kfree(curr);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_should_delete_dir_index(struct list_head *del_list,
|
|
|
|
u64 index)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_item *curr, *next;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
if (list_empty(del_list))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
list_for_each_entry_safe(curr, next, del_list, readdir_list) {
|
|
|
|
if (curr->key.offset > index)
|
|
|
|
break;
|
|
|
|
|
|
|
|
list_del(&curr->readdir_list);
|
|
|
|
ret = (curr->key.offset == index);
|
|
|
|
|
|
|
|
if (atomic_dec_and_test(&curr->refs))
|
|
|
|
kfree(curr);
|
|
|
|
|
|
|
|
if (ret)
|
|
|
|
return 1;
|
|
|
|
else
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* btrfs_readdir_delayed_dir_index - read dir info stored in the delayed tree
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
int btrfs_readdir_delayed_dir_index(struct file *filp, void *dirent,
|
|
|
|
filldir_t filldir,
|
|
|
|
struct list_head *ins_list)
|
|
|
|
{
|
|
|
|
struct btrfs_dir_item *di;
|
|
|
|
struct btrfs_delayed_item *curr, *next;
|
|
|
|
struct btrfs_key location;
|
|
|
|
char *name;
|
|
|
|
int name_len;
|
|
|
|
int over = 0;
|
|
|
|
unsigned char d_type;
|
|
|
|
|
|
|
|
if (list_empty(ins_list))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Changing the data of the delayed item is impossible. So
|
|
|
|
* we needn't lock them. And we have held i_mutex of the
|
|
|
|
* directory, nobody can delete any directory indexes now.
|
|
|
|
*/
|
|
|
|
list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
|
|
|
|
list_del(&curr->readdir_list);
|
|
|
|
|
|
|
|
if (curr->key.offset < filp->f_pos) {
|
|
|
|
if (atomic_dec_and_test(&curr->refs))
|
|
|
|
kfree(curr);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
filp->f_pos = curr->key.offset;
|
|
|
|
|
|
|
|
di = (struct btrfs_dir_item *)curr->data;
|
|
|
|
name = (char *)(di + 1);
|
|
|
|
name_len = le16_to_cpu(di->name_len);
|
|
|
|
|
|
|
|
d_type = btrfs_filetype_table[di->type];
|
|
|
|
btrfs_disk_key_to_cpu(&location, &di->location);
|
|
|
|
|
|
|
|
over = filldir(dirent, name, name_len, curr->key.offset,
|
|
|
|
location.objectid, d_type);
|
|
|
|
|
|
|
|
if (atomic_dec_and_test(&curr->refs))
|
|
|
|
kfree(curr);
|
|
|
|
|
|
|
|
if (over)
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_inode_generation, struct btrfs_inode_item,
|
|
|
|
generation, 64);
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_inode_sequence, struct btrfs_inode_item,
|
|
|
|
sequence, 64);
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_inode_transid, struct btrfs_inode_item,
|
|
|
|
transid, 64);
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_inode_size, struct btrfs_inode_item, size, 64);
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_inode_nbytes, struct btrfs_inode_item,
|
|
|
|
nbytes, 64);
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_inode_block_group, struct btrfs_inode_item,
|
|
|
|
block_group, 64);
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_inode_nlink, struct btrfs_inode_item, nlink, 32);
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_inode_uid, struct btrfs_inode_item, uid, 32);
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_inode_gid, struct btrfs_inode_item, gid, 32);
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_inode_mode, struct btrfs_inode_item, mode, 32);
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_inode_rdev, struct btrfs_inode_item, rdev, 64);
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_inode_flags, struct btrfs_inode_item, flags, 64);
|
|
|
|
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_timespec_sec, struct btrfs_timespec, sec, 64);
|
|
|
|
BTRFS_SETGET_STACK_FUNCS(stack_timespec_nsec, struct btrfs_timespec, nsec, 32);
|
|
|
|
|
|
|
|
static void fill_stack_inode_item(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_inode_item *inode_item,
|
|
|
|
struct inode *inode)
|
|
|
|
{
|
|
|
|
btrfs_set_stack_inode_uid(inode_item, inode->i_uid);
|
|
|
|
btrfs_set_stack_inode_gid(inode_item, inode->i_gid);
|
|
|
|
btrfs_set_stack_inode_size(inode_item, BTRFS_I(inode)->disk_i_size);
|
|
|
|
btrfs_set_stack_inode_mode(inode_item, inode->i_mode);
|
|
|
|
btrfs_set_stack_inode_nlink(inode_item, inode->i_nlink);
|
|
|
|
btrfs_set_stack_inode_nbytes(inode_item, inode_get_bytes(inode));
|
|
|
|
btrfs_set_stack_inode_generation(inode_item,
|
|
|
|
BTRFS_I(inode)->generation);
|
|
|
|
btrfs_set_stack_inode_sequence(inode_item, BTRFS_I(inode)->sequence);
|
|
|
|
btrfs_set_stack_inode_transid(inode_item, trans->transid);
|
|
|
|
btrfs_set_stack_inode_rdev(inode_item, inode->i_rdev);
|
|
|
|
btrfs_set_stack_inode_flags(inode_item, BTRFS_I(inode)->flags);
|
2011-05-28 13:00:39 +02:00
|
|
|
btrfs_set_stack_inode_block_group(inode_item, 0);
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
|
|
|
|
btrfs_set_stack_timespec_sec(btrfs_inode_atime(inode_item),
|
|
|
|
inode->i_atime.tv_sec);
|
|
|
|
btrfs_set_stack_timespec_nsec(btrfs_inode_atime(inode_item),
|
|
|
|
inode->i_atime.tv_nsec);
|
|
|
|
|
|
|
|
btrfs_set_stack_timespec_sec(btrfs_inode_mtime(inode_item),
|
|
|
|
inode->i_mtime.tv_sec);
|
|
|
|
btrfs_set_stack_timespec_nsec(btrfs_inode_mtime(inode_item),
|
|
|
|
inode->i_mtime.tv_nsec);
|
|
|
|
|
|
|
|
btrfs_set_stack_timespec_sec(btrfs_inode_ctime(inode_item),
|
|
|
|
inode->i_ctime.tv_sec);
|
|
|
|
btrfs_set_stack_timespec_nsec(btrfs_inode_ctime(inode_item),
|
|
|
|
inode->i_ctime.tv_nsec);
|
|
|
|
}
|
|
|
|
|
2011-06-23 09:27:13 +02:00
|
|
|
int btrfs_fill_inode(struct inode *inode, u32 *rdev)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_node *delayed_node;
|
|
|
|
struct btrfs_inode_item *inode_item;
|
|
|
|
struct btrfs_timespec *tspec;
|
|
|
|
|
|
|
|
delayed_node = btrfs_get_delayed_node(inode);
|
|
|
|
if (!delayed_node)
|
|
|
|
return -ENOENT;
|
|
|
|
|
|
|
|
mutex_lock(&delayed_node->mutex);
|
|
|
|
if (!delayed_node->inode_dirty) {
|
|
|
|
mutex_unlock(&delayed_node->mutex);
|
|
|
|
btrfs_release_delayed_node(delayed_node);
|
|
|
|
return -ENOENT;
|
|
|
|
}
|
|
|
|
|
|
|
|
inode_item = &delayed_node->inode_item;
|
|
|
|
|
|
|
|
inode->i_uid = btrfs_stack_inode_uid(inode_item);
|
|
|
|
inode->i_gid = btrfs_stack_inode_gid(inode_item);
|
|
|
|
btrfs_i_size_write(inode, btrfs_stack_inode_size(inode_item));
|
|
|
|
inode->i_mode = btrfs_stack_inode_mode(inode_item);
|
|
|
|
inode->i_nlink = btrfs_stack_inode_nlink(inode_item);
|
|
|
|
inode_set_bytes(inode, btrfs_stack_inode_nbytes(inode_item));
|
|
|
|
BTRFS_I(inode)->generation = btrfs_stack_inode_generation(inode_item);
|
|
|
|
BTRFS_I(inode)->sequence = btrfs_stack_inode_sequence(inode_item);
|
|
|
|
inode->i_rdev = 0;
|
|
|
|
*rdev = btrfs_stack_inode_rdev(inode_item);
|
|
|
|
BTRFS_I(inode)->flags = btrfs_stack_inode_flags(inode_item);
|
|
|
|
|
|
|
|
tspec = btrfs_inode_atime(inode_item);
|
|
|
|
inode->i_atime.tv_sec = btrfs_stack_timespec_sec(tspec);
|
|
|
|
inode->i_atime.tv_nsec = btrfs_stack_timespec_nsec(tspec);
|
|
|
|
|
|
|
|
tspec = btrfs_inode_mtime(inode_item);
|
|
|
|
inode->i_mtime.tv_sec = btrfs_stack_timespec_sec(tspec);
|
|
|
|
inode->i_mtime.tv_nsec = btrfs_stack_timespec_nsec(tspec);
|
|
|
|
|
|
|
|
tspec = btrfs_inode_ctime(inode_item);
|
|
|
|
inode->i_ctime.tv_sec = btrfs_stack_timespec_sec(tspec);
|
|
|
|
inode->i_ctime.tv_nsec = btrfs_stack_timespec_nsec(tspec);
|
|
|
|
|
|
|
|
inode->i_generation = BTRFS_I(inode)->generation;
|
|
|
|
BTRFS_I(inode)->index_cnt = (u64)-1;
|
|
|
|
|
|
|
|
mutex_unlock(&delayed_node->mutex);
|
|
|
|
btrfs_release_delayed_node(delayed_node);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
int btrfs_delayed_update_inode(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root, struct inode *inode)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_node *delayed_node;
|
2011-06-03 16:29:08 +02:00
|
|
|
int ret = 0;
|
btrfs: implement delayed inode items operation
Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
root's radix tree, and letting btrfs inodes go.
Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
Itaru Kitayama.
Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
inode in time.
Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason
Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
which is created for every directory and file, and used to manage the
delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.
Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.
If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.
Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
manage the delayed nodes which are created for every file/directory.
One is used to manage all the delayed nodes that have delayed items. And the
other is used to manage the delayed nodes which is waiting to be dealt with
by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
index which is going to be inserted into b+ tree, and the other is used to
manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
to deal with the works of the delayed directory name index items insertion
and deletion and the delayed inode update.
When the delayed items is beyond the lower limit, we create works for some
delayed nodes and insert them into the work queue of the worker, and then
go back.
When the delayed items is beyond the upper bound, we create works for all
the delayed nodes that haven't been dealt with, and insert them into the work
queue of the worker, and then wait for that the untreated items is below some
threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
information into the delayed inserting rb-tree.
And then we check the number of the delayed items and do delayed items
balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
in the inserting rb-tree at first. If we look it up, just drop it. If not,
add the key of it into the delayed deleting rb-tree.
Similar to the delayed inserting rb-tree, we also check the number of the
delayed items and do delayed items balance.
(The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
inode into the delayed node. the worker will flush it into the b+ tree after
dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
delayed node, By this way, we can cache more delayed items and merge more
inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
and the delayed inode update.
I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.
Before applying this patch:
Create files:
Total files: 50000
Total time: 1.096108
Average time: 0.000022
Delete files:
Total files: 50000
Total time: 1.510403
Average time: 0.000030
After applying this patch:
Create files:
Total files: 50000
Total time: 0.932899
Average time: 0.000019
Delete files:
Total files: 50000
Total time: 1.215732
Average time: 0.000024
[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3
Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-04-22 12:12:22 +02:00
|
|
|
|
|
|
|
delayed_node = btrfs_get_or_create_delayed_node(inode);
|
|
|
|
if (IS_ERR(delayed_node))
|
|
|
|
return PTR_ERR(delayed_node);
|
|
|
|
|
|
|
|
mutex_lock(&delayed_node->mutex);
|
|
|
|
if (delayed_node->inode_dirty) {
|
|
|
|
fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
|
|
|
|
goto release_node;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = btrfs_delayed_inode_reserve_metadata(trans, root, delayed_node);
|
|
|
|
/*
|
|
|
|
* we must reserve enough space when we start a new transaction,
|
|
|
|
* so reserving metadata failure is impossible
|
|
|
|
*/
|
|
|
|
BUG_ON(ret);
|
|
|
|
|
|
|
|
fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
|
|
|
|
delayed_node->inode_dirty = 1;
|
|
|
|
delayed_node->count++;
|
|
|
|
atomic_inc(&root->fs_info->delayed_root->items);
|
|
|
|
release_node:
|
|
|
|
mutex_unlock(&delayed_node->mutex);
|
|
|
|
btrfs_release_delayed_node(delayed_node);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __btrfs_kill_delayed_node(struct btrfs_delayed_node *delayed_node)
|
|
|
|
{
|
|
|
|
struct btrfs_root *root = delayed_node->root;
|
|
|
|
struct btrfs_delayed_item *curr_item, *prev_item;
|
|
|
|
|
|
|
|
mutex_lock(&delayed_node->mutex);
|
|
|
|
curr_item = __btrfs_first_delayed_insertion_item(delayed_node);
|
|
|
|
while (curr_item) {
|
|
|
|
btrfs_delayed_item_release_metadata(root, curr_item);
|
|
|
|
prev_item = curr_item;
|
|
|
|
curr_item = __btrfs_next_delayed_item(prev_item);
|
|
|
|
btrfs_release_delayed_item(prev_item);
|
|
|
|
}
|
|
|
|
|
|
|
|
curr_item = __btrfs_first_delayed_deletion_item(delayed_node);
|
|
|
|
while (curr_item) {
|
|
|
|
btrfs_delayed_item_release_metadata(root, curr_item);
|
|
|
|
prev_item = curr_item;
|
|
|
|
curr_item = __btrfs_next_delayed_item(prev_item);
|
|
|
|
btrfs_release_delayed_item(prev_item);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (delayed_node->inode_dirty) {
|
|
|
|
btrfs_delayed_inode_release_metadata(root, delayed_node);
|
|
|
|
btrfs_release_delayed_inode(delayed_node);
|
|
|
|
}
|
|
|
|
mutex_unlock(&delayed_node->mutex);
|
|
|
|
}
|
|
|
|
|
|
|
|
void btrfs_kill_delayed_inode_items(struct inode *inode)
|
|
|
|
{
|
|
|
|
struct btrfs_delayed_node *delayed_node;
|
|
|
|
|
|
|
|
delayed_node = btrfs_get_delayed_node(inode);
|
|
|
|
if (!delayed_node)
|
|
|
|
return;
|
|
|
|
|
|
|
|
__btrfs_kill_delayed_node(delayed_node);
|
|
|
|
btrfs_release_delayed_node(delayed_node);
|
|
|
|
}
|
|
|
|
|
|
|
|
void btrfs_kill_all_delayed_nodes(struct btrfs_root *root)
|
|
|
|
{
|
|
|
|
u64 inode_id = 0;
|
|
|
|
struct btrfs_delayed_node *delayed_nodes[8];
|
|
|
|
int i, n;
|
|
|
|
|
|
|
|
while (1) {
|
|
|
|
spin_lock(&root->inode_lock);
|
|
|
|
n = radix_tree_gang_lookup(&root->delayed_nodes_tree,
|
|
|
|
(void **)delayed_nodes, inode_id,
|
|
|
|
ARRAY_SIZE(delayed_nodes));
|
|
|
|
if (!n) {
|
|
|
|
spin_unlock(&root->inode_lock);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
inode_id = delayed_nodes[n - 1]->inode_id + 1;
|
|
|
|
|
|
|
|
for (i = 0; i < n; i++)
|
|
|
|
atomic_inc(&delayed_nodes[i]->refs);
|
|
|
|
spin_unlock(&root->inode_lock);
|
|
|
|
|
|
|
|
for (i = 0; i < n; i++) {
|
|
|
|
__btrfs_kill_delayed_node(delayed_nodes[i]);
|
|
|
|
btrfs_release_delayed_node(delayed_nodes[i]);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|