2008-01-09 04:46:30 +08:00
|
|
|
/*
|
|
|
|
* Copyright (C) 2007 Oracle. All rights reserved.
|
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public
|
|
|
|
* License v2 as published by the Free Software Foundation.
|
|
|
|
*
|
|
|
|
* This program is distributed in the hope that it will be useful,
|
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
|
|
* General Public License for more details.
|
|
|
|
*
|
|
|
|
* You should have received a copy of the GNU General Public
|
|
|
|
* License along with this program; if not, write to the
|
|
|
|
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
|
|
|
|
* Boston, MA 021110-1307, USA.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/slab.h>
|
2008-05-01 01:59:35 +08:00
|
|
|
#include <linux/blkdev.h>
|
2008-07-22 23:18:09 +08:00
|
|
|
#include <linux/writeback.h>
|
|
|
|
#include <linux/pagevec.h>
|
2008-01-09 04:46:30 +08:00
|
|
|
#include "ctree.h"
|
|
|
|
#include "transaction.h"
|
|
|
|
#include "btrfs_inode.h"
|
2008-07-18 00:53:50 +08:00
|
|
|
#include "extent_io.h"
|
2013-05-15 15:48:23 +08:00
|
|
|
#include "disk-io.h"
|
2008-01-09 04:46:30 +08:00
|
|
|
|
2012-09-06 18:01:51 +08:00
|
|
|
static struct kmem_cache *btrfs_ordered_extent_cache;
|
|
|
|
|
2008-07-18 00:53:50 +08:00
|
|
|
static u64 entry_end(struct btrfs_ordered_extent *entry)
|
2008-01-09 04:46:30 +08:00
|
|
|
{
|
2008-07-18 00:53:50 +08:00
|
|
|
if (entry->file_offset + entry->len < entry->file_offset)
|
|
|
|
return (u64)-1;
|
|
|
|
return entry->file_offset + entry->len;
|
2008-01-09 04:46:30 +08:00
|
|
|
}
|
|
|
|
|
2008-09-30 03:18:18 +08:00
|
|
|
/* returns NULL if the insertion worked, or it returns the node it did find
|
|
|
|
* in the tree
|
|
|
|
*/
|
2008-07-18 00:53:50 +08:00
|
|
|
static struct rb_node *tree_insert(struct rb_root *root, u64 file_offset,
|
|
|
|
struct rb_node *node)
|
2008-01-09 04:46:30 +08:00
|
|
|
{
|
2009-01-06 10:25:51 +08:00
|
|
|
struct rb_node **p = &root->rb_node;
|
|
|
|
struct rb_node *parent = NULL;
|
2008-07-18 00:53:50 +08:00
|
|
|
struct btrfs_ordered_extent *entry;
|
2008-01-09 04:46:30 +08:00
|
|
|
|
2009-01-06 10:25:51 +08:00
|
|
|
while (*p) {
|
2008-01-09 04:46:30 +08:00
|
|
|
parent = *p;
|
2008-07-18 00:53:50 +08:00
|
|
|
entry = rb_entry(parent, struct btrfs_ordered_extent, rb_node);
|
2008-01-09 04:46:30 +08:00
|
|
|
|
2008-07-18 00:53:50 +08:00
|
|
|
if (file_offset < entry->file_offset)
|
2008-01-09 04:46:30 +08:00
|
|
|
p = &(*p)->rb_left;
|
2008-07-18 00:53:50 +08:00
|
|
|
else if (file_offset >= entry_end(entry))
|
2008-01-09 04:46:30 +08:00
|
|
|
p = &(*p)->rb_right;
|
|
|
|
else
|
|
|
|
return parent;
|
|
|
|
}
|
|
|
|
|
|
|
|
rb_link_node(node, parent, p);
|
|
|
|
rb_insert_color(node, root);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2011-10-04 11:22:33 +08:00
|
|
|
static void ordered_data_tree_panic(struct inode *inode, int errno,
|
|
|
|
u64 offset)
|
|
|
|
{
|
|
|
|
struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
|
|
|
|
btrfs_panic(fs_info, errno, "Inconsistency in ordered tree at offset "
|
2013-08-20 19:20:07 +08:00
|
|
|
"%llu\n", offset);
|
2011-10-04 11:22:33 +08:00
|
|
|
}
|
|
|
|
|
2008-09-30 03:18:18 +08:00
|
|
|
/*
|
|
|
|
* look for a given offset in the tree, and if it can't be found return the
|
|
|
|
* first lesser offset
|
|
|
|
*/
|
2008-07-18 00:53:50 +08:00
|
|
|
static struct rb_node *__tree_search(struct rb_root *root, u64 file_offset,
|
|
|
|
struct rb_node **prev_ret)
|
2008-01-09 04:46:30 +08:00
|
|
|
{
|
2009-01-06 10:25:51 +08:00
|
|
|
struct rb_node *n = root->rb_node;
|
2008-01-09 04:46:30 +08:00
|
|
|
struct rb_node *prev = NULL;
|
2008-07-18 00:53:50 +08:00
|
|
|
struct rb_node *test;
|
|
|
|
struct btrfs_ordered_extent *entry;
|
|
|
|
struct btrfs_ordered_extent *prev_entry = NULL;
|
2008-01-09 04:46:30 +08:00
|
|
|
|
2009-01-06 10:25:51 +08:00
|
|
|
while (n) {
|
2008-07-18 00:53:50 +08:00
|
|
|
entry = rb_entry(n, struct btrfs_ordered_extent, rb_node);
|
2008-01-09 04:46:30 +08:00
|
|
|
prev = n;
|
|
|
|
prev_entry = entry;
|
|
|
|
|
2008-07-18 00:53:50 +08:00
|
|
|
if (file_offset < entry->file_offset)
|
2008-01-09 04:46:30 +08:00
|
|
|
n = n->rb_left;
|
2008-07-18 00:53:50 +08:00
|
|
|
else if (file_offset >= entry_end(entry))
|
2008-01-09 04:46:30 +08:00
|
|
|
n = n->rb_right;
|
|
|
|
else
|
|
|
|
return n;
|
|
|
|
}
|
|
|
|
if (!prev_ret)
|
|
|
|
return NULL;
|
|
|
|
|
2009-01-06 10:25:51 +08:00
|
|
|
while (prev && file_offset >= entry_end(prev_entry)) {
|
2008-07-18 00:53:50 +08:00
|
|
|
test = rb_next(prev);
|
|
|
|
if (!test)
|
|
|
|
break;
|
|
|
|
prev_entry = rb_entry(test, struct btrfs_ordered_extent,
|
|
|
|
rb_node);
|
|
|
|
if (file_offset < entry_end(prev_entry))
|
|
|
|
break;
|
|
|
|
|
|
|
|
prev = test;
|
|
|
|
}
|
|
|
|
if (prev)
|
|
|
|
prev_entry = rb_entry(prev, struct btrfs_ordered_extent,
|
|
|
|
rb_node);
|
2009-01-06 10:25:51 +08:00
|
|
|
while (prev && file_offset < entry_end(prev_entry)) {
|
2008-07-18 00:53:50 +08:00
|
|
|
test = rb_prev(prev);
|
|
|
|
if (!test)
|
|
|
|
break;
|
|
|
|
prev_entry = rb_entry(test, struct btrfs_ordered_extent,
|
|
|
|
rb_node);
|
|
|
|
prev = test;
|
2008-01-09 04:46:30 +08:00
|
|
|
}
|
|
|
|
*prev_ret = prev;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2008-09-30 03:18:18 +08:00
|
|
|
/*
|
|
|
|
* helper to check if a given offset is inside a given entry
|
|
|
|
*/
|
2008-07-18 00:53:50 +08:00
|
|
|
static int offset_in_entry(struct btrfs_ordered_extent *entry, u64 file_offset)
|
|
|
|
{
|
|
|
|
if (file_offset < entry->file_offset ||
|
|
|
|
entry->file_offset + entry->len <= file_offset)
|
|
|
|
return 0;
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2010-05-23 23:00:55 +08:00
|
|
|
static int range_overlaps(struct btrfs_ordered_extent *entry, u64 file_offset,
|
|
|
|
u64 len)
|
|
|
|
{
|
|
|
|
if (file_offset + len <= entry->file_offset ||
|
|
|
|
entry->file_offset + entry->len <= file_offset)
|
|
|
|
return 0;
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2008-09-30 03:18:18 +08:00
|
|
|
/*
|
|
|
|
* look find the first ordered struct that has this offset, otherwise
|
|
|
|
* the first one less than this offset
|
|
|
|
*/
|
2008-07-18 00:53:50 +08:00
|
|
|
static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree,
|
|
|
|
u64 file_offset)
|
2008-01-09 04:46:30 +08:00
|
|
|
{
|
2008-07-18 00:53:50 +08:00
|
|
|
struct rb_root *root = &tree->tree;
|
2011-02-01 08:54:59 +08:00
|
|
|
struct rb_node *prev = NULL;
|
2008-01-09 04:46:30 +08:00
|
|
|
struct rb_node *ret;
|
2008-07-18 00:53:50 +08:00
|
|
|
struct btrfs_ordered_extent *entry;
|
|
|
|
|
|
|
|
if (tree->last) {
|
|
|
|
entry = rb_entry(tree->last, struct btrfs_ordered_extent,
|
|
|
|
rb_node);
|
|
|
|
if (offset_in_entry(entry, file_offset))
|
|
|
|
return tree->last;
|
|
|
|
}
|
|
|
|
ret = __tree_search(root, file_offset, &prev);
|
2008-01-09 04:46:30 +08:00
|
|
|
if (!ret)
|
2008-07-18 00:53:50 +08:00
|
|
|
ret = prev;
|
|
|
|
if (ret)
|
|
|
|
tree->last = ret;
|
2008-01-09 04:46:30 +08:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-07-18 01:53:27 +08:00
|
|
|
/* allocate and add a new ordered_extent into the per-inode tree.
|
|
|
|
* file_offset is the logical offset in the file
|
|
|
|
*
|
|
|
|
* start is the disk block number of an extent already reserved in the
|
|
|
|
* extent allocation tree
|
|
|
|
*
|
|
|
|
* len is the length of the extent
|
|
|
|
*
|
|
|
|
* The tree is given a single reference on the ordered extent that was
|
|
|
|
* inserted.
|
|
|
|
*/
|
2010-05-23 23:00:55 +08:00
|
|
|
static int __btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
|
|
|
|
u64 start, u64 len, u64 disk_len,
|
2010-12-17 14:21:50 +08:00
|
|
|
int type, int dio, int compress_type)
|
2008-01-09 04:46:30 +08:00
|
|
|
{
|
2013-05-15 15:48:23 +08:00
|
|
|
struct btrfs_root *root = BTRFS_I(inode)->root;
|
2008-01-09 04:46:30 +08:00
|
|
|
struct btrfs_ordered_inode_tree *tree;
|
2008-07-18 00:53:50 +08:00
|
|
|
struct rb_node *node;
|
|
|
|
struct btrfs_ordered_extent *entry;
|
2008-01-09 04:46:30 +08:00
|
|
|
|
2008-07-18 00:53:50 +08:00
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
2012-09-06 18:01:51 +08:00
|
|
|
entry = kmem_cache_zalloc(btrfs_ordered_extent_cache, GFP_NOFS);
|
2008-01-09 04:46:30 +08:00
|
|
|
if (!entry)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2008-07-18 00:53:50 +08:00
|
|
|
entry->file_offset = file_offset;
|
|
|
|
entry->start = start;
|
|
|
|
entry->len = len;
|
2012-10-13 03:27:49 +08:00
|
|
|
if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM) &&
|
|
|
|
!(type == BTRFS_ORDERED_NOCOW))
|
|
|
|
entry->csum_bytes_left = disk_len;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 02:49:59 +08:00
|
|
|
entry->disk_len = disk_len;
|
2009-09-03 04:53:46 +08:00
|
|
|
entry->bytes_left = len;
|
2012-05-03 02:00:54 +08:00
|
|
|
entry->inode = igrab(inode);
|
2010-12-17 14:21:50 +08:00
|
|
|
entry->compress_type = compress_type;
|
2013-08-30 01:57:21 +08:00
|
|
|
entry->truncated_len = (u64)-1;
|
2008-10-31 02:25:28 +08:00
|
|
|
if (type != BTRFS_ORDERED_IO_DONE && type != BTRFS_ORDERED_COMPLETE)
|
2008-10-31 02:20:02 +08:00
|
|
|
set_bit(type, &entry->flags);
|
2008-07-24 23:57:52 +08:00
|
|
|
|
2010-05-23 23:00:55 +08:00
|
|
|
if (dio)
|
|
|
|
set_bit(BTRFS_ORDERED_DIRECT, &entry->flags);
|
|
|
|
|
2008-07-18 00:53:50 +08:00
|
|
|
/* one ref for the tree */
|
|
|
|
atomic_set(&entry->refs, 1);
|
|
|
|
init_waitqueue_head(&entry->wait);
|
|
|
|
INIT_LIST_HEAD(&entry->list);
|
2008-07-24 23:57:52 +08:00
|
|
|
INIT_LIST_HEAD(&entry->root_extent_list);
|
2012-10-25 17:41:36 +08:00
|
|
|
INIT_LIST_HEAD(&entry->work_list);
|
|
|
|
init_completion(&entry->completion);
|
2012-10-13 03:27:49 +08:00
|
|
|
INIT_LIST_HEAD(&entry->log_list);
|
2008-01-09 04:46:30 +08:00
|
|
|
|
Btrfs: add initial tracepoint support for btrfs
Tracepoints can provide insight into why btrfs hits bugs and be greatly
helpful for debugging, e.g
dd-7822 [000] 2121.641088: btrfs_inode_request: root = 5(FS_TREE), gen = 4, ino = 256, blocks = 8, disk_i_size = 0, last_trans = 8, logged_trans = 0
dd-7822 [000] 2121.641100: btrfs_inode_new: root = 5(FS_TREE), gen = 8, ino = 257, blocks = 0, disk_i_size = 0, last_trans = 0, logged_trans = 0
btrfs-transacti-7804 [001] 2146.935420: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29368320 (orig_level = 0), cow_buf = 29388800 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.935473: btrfs_cow_block: root = 1(ROOT_TREE), refs = 2, orig_buf = 29364224 (orig_level = 0), cow_buf = 29392896 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.972221: btrfs_transaction_commit: root = 1(ROOT_TREE), gen = 8
flush-btrfs-2-7821 [001] 2155.824210: btrfs_chunk_alloc: root = 3(CHUNK_TREE), offset = 1103101952, size = 1073741824, num_stripes = 1, sub_stripes = 0, type = DATA
flush-btrfs-2-7821 [001] 2155.824241: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29388800 (orig_level = 0), cow_buf = 29396992 (cow_level = 0)
flush-btrfs-2-7821 [001] 2155.824255: btrfs_cow_block: root = 4(DEV_TREE), refs = 2, orig_buf = 29372416 (orig_level = 0), cow_buf = 29401088 (cow_level = 0)
flush-btrfs-2-7821 [000] 2155.824329: btrfs_cow_block: root = 3(CHUNK_TREE), refs = 2, orig_buf = 20971520 (orig_level = 0), cow_buf = 20975616 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898019: btrfs_cow_block: root = 5(FS_TREE), refs = 2, orig_buf = 29384704 (orig_level = 0), cow_buf = 29405184 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898043: btrfs_cow_block: root = 7(CSUM_TREE), refs = 2, orig_buf = 29376512 (orig_level = 0), cow_buf = 29409280 (cow_level = 0)
Here is what I have added:
1) ordere_extent:
btrfs_ordered_extent_add
btrfs_ordered_extent_remove
btrfs_ordered_extent_start
btrfs_ordered_extent_put
These provide critical information to understand how ordered_extents are
updated.
2) extent_map:
btrfs_get_extent
extent_map is used in both read and write cases, and it is useful for tracking
how btrfs specific IO is running.
3) writepage:
__extent_writepage
btrfs_writepage_end_io_hook
Pages are cirtical resourses and produce a lot of corner cases during writeback,
so it is valuable to know how page is written to disk.
4) inode:
btrfs_inode_new
btrfs_inode_request
btrfs_inode_evict
These can show where and when a inode is created, when a inode is evicted.
5) sync:
btrfs_sync_file
btrfs_sync_fs
These show sync arguments.
6) transaction:
btrfs_transaction_commit
In transaction based filesystem, it will be useful to know the generation and
who does commit.
7) back reference and cow:
btrfs_delayed_tree_ref
btrfs_delayed_data_ref
btrfs_delayed_ref_head
btrfs_cow_block
Btrfs natively supports back references, these tracepoints are helpful on
understanding btrfs's COW mechanism.
8) chunk:
btrfs_chunk_alloc
btrfs_chunk_free
Chunk is a link between physical offset and logical offset, and stands for space
infomation in btrfs, and these are helpful on tracing space things.
9) reserved_extent:
btrfs_reserved_extent_alloc
btrfs_reserved_extent_free
These can show how btrfs uses its space.
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-03-24 19:18:59 +08:00
|
|
|
trace_btrfs_ordered_extent_add(inode, entry);
|
|
|
|
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_lock_irq(&tree->lock);
|
2008-07-18 00:53:50 +08:00
|
|
|
node = tree_insert(&tree->tree, file_offset,
|
|
|
|
&entry->rb_node);
|
2011-10-04 11:22:33 +08:00
|
|
|
if (node)
|
|
|
|
ordered_data_tree_panic(inode, -EEXIST, file_offset);
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_unlock_irq(&tree->lock);
|
2009-01-06 10:25:51 +08:00
|
|
|
|
2013-05-15 15:48:23 +08:00
|
|
|
spin_lock(&root->ordered_extent_lock);
|
2008-07-24 23:57:52 +08:00
|
|
|
list_add_tail(&entry->root_extent_list,
|
2013-05-15 15:48:23 +08:00
|
|
|
&root->ordered_extents);
|
|
|
|
root->nr_ordered_extents++;
|
|
|
|
if (root->nr_ordered_extents == 1) {
|
|
|
|
spin_lock(&root->fs_info->ordered_root_lock);
|
|
|
|
BUG_ON(!list_empty(&root->ordered_root));
|
|
|
|
list_add_tail(&root->ordered_root,
|
|
|
|
&root->fs_info->ordered_roots);
|
|
|
|
spin_unlock(&root->fs_info->ordered_root_lock);
|
|
|
|
}
|
|
|
|
spin_unlock(&root->ordered_extent_lock);
|
2008-07-24 23:57:52 +08:00
|
|
|
|
2008-01-09 04:46:30 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-05-23 23:00:55 +08:00
|
|
|
int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
|
|
|
|
u64 start, u64 len, u64 disk_len, int type)
|
|
|
|
{
|
|
|
|
return __btrfs_add_ordered_extent(inode, file_offset, start, len,
|
2010-12-17 14:21:50 +08:00
|
|
|
disk_len, type, 0,
|
|
|
|
BTRFS_COMPRESS_NONE);
|
2010-05-23 23:00:55 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_add_ordered_extent_dio(struct inode *inode, u64 file_offset,
|
|
|
|
u64 start, u64 len, u64 disk_len, int type)
|
|
|
|
{
|
|
|
|
return __btrfs_add_ordered_extent(inode, file_offset, start, len,
|
2010-12-17 14:21:50 +08:00
|
|
|
disk_len, type, 1,
|
|
|
|
BTRFS_COMPRESS_NONE);
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_add_ordered_extent_compress(struct inode *inode, u64 file_offset,
|
|
|
|
u64 start, u64 len, u64 disk_len,
|
|
|
|
int type, int compress_type)
|
|
|
|
{
|
|
|
|
return __btrfs_add_ordered_extent(inode, file_offset, start, len,
|
|
|
|
disk_len, type, 0,
|
|
|
|
compress_type);
|
2010-05-23 23:00:55 +08:00
|
|
|
}
|
|
|
|
|
2008-07-18 01:53:27 +08:00
|
|
|
/*
|
|
|
|
* Add a struct btrfs_ordered_sum into the list of checksums to be inserted
|
2008-07-18 18:17:13 +08:00
|
|
|
* when an ordered extent is finished. If the list covers more than one
|
|
|
|
* ordered extent, it is split across multiples.
|
2008-07-18 01:53:27 +08:00
|
|
|
*/
|
2012-03-01 21:56:26 +08:00
|
|
|
void btrfs_add_ordered_sum(struct inode *inode,
|
|
|
|
struct btrfs_ordered_extent *entry,
|
|
|
|
struct btrfs_ordered_sum *sum)
|
2008-01-09 04:46:30 +08:00
|
|
|
{
|
2008-07-18 00:53:50 +08:00
|
|
|
struct btrfs_ordered_inode_tree *tree;
|
2008-01-09 04:46:30 +08:00
|
|
|
|
2008-07-18 00:53:50 +08:00
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_lock_irq(&tree->lock);
|
2008-07-18 00:53:50 +08:00
|
|
|
list_add_tail(&sum->list, &entry->list);
|
2012-10-13 03:27:49 +08:00
|
|
|
WARN_ON(entry->csum_bytes_left < sum->len);
|
|
|
|
entry->csum_bytes_left -= sum->len;
|
|
|
|
if (entry->csum_bytes_left == 0)
|
|
|
|
wake_up(&entry->wait);
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_unlock_irq(&tree->lock);
|
2008-01-09 04:46:30 +08:00
|
|
|
}
|
|
|
|
|
2010-11-29 08:56:33 +08:00
|
|
|
/*
|
|
|
|
* this is used to account for finished IO across a given range
|
|
|
|
* of the file. The IO may span ordered extents. If
|
|
|
|
* a given ordered_extent is completely done, 1 is returned, otherwise
|
|
|
|
* 0.
|
|
|
|
*
|
|
|
|
* test_and_set_bit on a flag in the struct btrfs_ordered_extent is used
|
|
|
|
* to make sure this function only returns 1 once for a given ordered extent.
|
|
|
|
*
|
|
|
|
* file_offset is updated to one byte past the range that is recorded as
|
|
|
|
* complete. This allows you to walk forward in the file.
|
|
|
|
*/
|
|
|
|
int btrfs_dec_test_first_ordered_pending(struct inode *inode,
|
|
|
|
struct btrfs_ordered_extent **cached,
|
2012-05-03 02:00:54 +08:00
|
|
|
u64 *file_offset, u64 io_size, int uptodate)
|
2010-11-29 08:56:33 +08:00
|
|
|
{
|
|
|
|
struct btrfs_ordered_inode_tree *tree;
|
|
|
|
struct rb_node *node;
|
|
|
|
struct btrfs_ordered_extent *entry = NULL;
|
|
|
|
int ret;
|
2012-05-03 02:00:54 +08:00
|
|
|
unsigned long flags;
|
2010-11-29 08:56:33 +08:00
|
|
|
u64 dec_end;
|
|
|
|
u64 dec_start;
|
|
|
|
u64 to_dec;
|
|
|
|
|
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_lock_irqsave(&tree->lock, flags);
|
2010-11-29 08:56:33 +08:00
|
|
|
node = tree_search(tree, *file_offset);
|
|
|
|
if (!node) {
|
|
|
|
ret = 1;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
|
|
|
|
if (!offset_in_entry(entry, *file_offset)) {
|
|
|
|
ret = 1;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
dec_start = max(*file_offset, entry->file_offset);
|
|
|
|
dec_end = min(*file_offset + io_size, entry->file_offset +
|
|
|
|
entry->len);
|
|
|
|
*file_offset = dec_end;
|
|
|
|
if (dec_start > dec_end) {
|
|
|
|
printk(KERN_CRIT "bad ordering dec_start %llu end %llu\n",
|
2013-08-20 19:20:07 +08:00
|
|
|
dec_start, dec_end);
|
2010-11-29 08:56:33 +08:00
|
|
|
}
|
|
|
|
to_dec = dec_end - dec_start;
|
|
|
|
if (to_dec > entry->bytes_left) {
|
|
|
|
printk(KERN_CRIT "bad ordered accounting left %llu size %llu\n",
|
2013-08-20 19:20:07 +08:00
|
|
|
entry->bytes_left, to_dec);
|
2010-11-29 08:56:33 +08:00
|
|
|
}
|
|
|
|
entry->bytes_left -= to_dec;
|
2012-05-03 02:00:54 +08:00
|
|
|
if (!uptodate)
|
|
|
|
set_bit(BTRFS_ORDERED_IOERR, &entry->flags);
|
|
|
|
|
2010-11-29 08:56:33 +08:00
|
|
|
if (entry->bytes_left == 0)
|
|
|
|
ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags);
|
|
|
|
else
|
|
|
|
ret = 1;
|
|
|
|
out:
|
|
|
|
if (!ret && cached && entry) {
|
|
|
|
*cached = entry;
|
|
|
|
atomic_inc(&entry->refs);
|
|
|
|
}
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_unlock_irqrestore(&tree->lock, flags);
|
2010-11-29 08:56:33 +08:00
|
|
|
return ret == 0;
|
|
|
|
}
|
|
|
|
|
2008-07-18 01:53:27 +08:00
|
|
|
/*
|
|
|
|
* this is used to account for finished IO across a given range
|
|
|
|
* of the file. The IO should not span ordered extents. If
|
|
|
|
* a given ordered_extent is completely done, 1 is returned, otherwise
|
|
|
|
* 0.
|
|
|
|
*
|
|
|
|
* test_and_set_bit on a flag in the struct btrfs_ordered_extent is used
|
|
|
|
* to make sure this function only returns 1 once for a given ordered extent.
|
|
|
|
*/
|
2008-07-18 00:53:50 +08:00
|
|
|
int btrfs_dec_test_ordered_pending(struct inode *inode,
|
2010-02-03 04:51:14 +08:00
|
|
|
struct btrfs_ordered_extent **cached,
|
2012-05-03 02:00:54 +08:00
|
|
|
u64 file_offset, u64 io_size, int uptodate)
|
2008-01-09 04:46:30 +08:00
|
|
|
{
|
2008-07-18 00:53:50 +08:00
|
|
|
struct btrfs_ordered_inode_tree *tree;
|
2008-01-09 04:46:30 +08:00
|
|
|
struct rb_node *node;
|
2010-02-03 04:51:14 +08:00
|
|
|
struct btrfs_ordered_extent *entry = NULL;
|
2012-05-03 02:00:54 +08:00
|
|
|
unsigned long flags;
|
2008-07-18 00:53:50 +08:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_lock_irqsave(&tree->lock, flags);
|
|
|
|
if (cached && *cached) {
|
|
|
|
entry = *cached;
|
|
|
|
goto have_entry;
|
|
|
|
}
|
|
|
|
|
2008-07-18 00:53:50 +08:00
|
|
|
node = tree_search(tree, file_offset);
|
2008-01-09 04:46:30 +08:00
|
|
|
if (!node) {
|
2008-07-18 00:53:50 +08:00
|
|
|
ret = 1;
|
|
|
|
goto out;
|
2008-01-09 04:46:30 +08:00
|
|
|
}
|
|
|
|
|
2008-07-18 00:53:50 +08:00
|
|
|
entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
|
2012-05-03 02:00:54 +08:00
|
|
|
have_entry:
|
2008-07-18 00:53:50 +08:00
|
|
|
if (!offset_in_entry(entry, file_offset)) {
|
|
|
|
ret = 1;
|
|
|
|
goto out;
|
2008-01-09 04:46:30 +08:00
|
|
|
}
|
2008-07-18 00:53:50 +08:00
|
|
|
|
2009-09-03 04:53:46 +08:00
|
|
|
if (io_size > entry->bytes_left) {
|
|
|
|
printk(KERN_CRIT "bad ordered accounting left %llu size %llu\n",
|
2013-08-20 19:20:07 +08:00
|
|
|
entry->bytes_left, io_size);
|
2009-09-03 04:53:46 +08:00
|
|
|
}
|
|
|
|
entry->bytes_left -= io_size;
|
2012-05-03 02:00:54 +08:00
|
|
|
if (!uptodate)
|
|
|
|
set_bit(BTRFS_ORDERED_IOERR, &entry->flags);
|
|
|
|
|
2009-09-03 04:53:46 +08:00
|
|
|
if (entry->bytes_left == 0)
|
2008-07-18 00:53:50 +08:00
|
|
|
ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags);
|
2009-09-03 04:53:46 +08:00
|
|
|
else
|
|
|
|
ret = 1;
|
2008-07-18 00:53:50 +08:00
|
|
|
out:
|
2010-02-03 04:51:14 +08:00
|
|
|
if (!ret && cached && entry) {
|
|
|
|
*cached = entry;
|
|
|
|
atomic_inc(&entry->refs);
|
|
|
|
}
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_unlock_irqrestore(&tree->lock, flags);
|
2008-07-18 00:53:50 +08:00
|
|
|
return ret == 0;
|
|
|
|
}
|
2008-01-09 04:46:30 +08:00
|
|
|
|
2012-10-13 03:27:49 +08:00
|
|
|
/* Needs to either be called under a log transaction or the log_mutex */
|
|
|
|
void btrfs_get_logged_extents(struct btrfs_root *log, struct inode *inode)
|
|
|
|
{
|
|
|
|
struct btrfs_ordered_inode_tree *tree;
|
|
|
|
struct btrfs_ordered_extent *ordered;
|
|
|
|
struct rb_node *n;
|
|
|
|
int index = log->log_transid % 2;
|
|
|
|
|
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
|
|
|
spin_lock_irq(&tree->lock);
|
|
|
|
for (n = rb_first(&tree->tree); n; n = rb_next(n)) {
|
|
|
|
ordered = rb_entry(n, struct btrfs_ordered_extent, rb_node);
|
|
|
|
spin_lock(&log->log_extents_lock[index]);
|
|
|
|
if (list_empty(&ordered->log_list)) {
|
|
|
|
list_add_tail(&ordered->log_list, &log->logged_list[index]);
|
|
|
|
atomic_inc(&ordered->refs);
|
|
|
|
}
|
|
|
|
spin_unlock(&log->log_extents_lock[index]);
|
|
|
|
}
|
|
|
|
spin_unlock_irq(&tree->lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
void btrfs_wait_logged_extents(struct btrfs_root *log, u64 transid)
|
|
|
|
{
|
|
|
|
struct btrfs_ordered_extent *ordered;
|
|
|
|
int index = transid % 2;
|
|
|
|
|
|
|
|
spin_lock_irq(&log->log_extents_lock[index]);
|
|
|
|
while (!list_empty(&log->logged_list[index])) {
|
|
|
|
ordered = list_first_entry(&log->logged_list[index],
|
|
|
|
struct btrfs_ordered_extent,
|
|
|
|
log_list);
|
|
|
|
list_del_init(&ordered->log_list);
|
|
|
|
spin_unlock_irq(&log->log_extents_lock[index]);
|
|
|
|
wait_event(ordered->wait, test_bit(BTRFS_ORDERED_IO_DONE,
|
|
|
|
&ordered->flags));
|
|
|
|
btrfs_put_ordered_extent(ordered);
|
|
|
|
spin_lock_irq(&log->log_extents_lock[index]);
|
|
|
|
}
|
|
|
|
spin_unlock_irq(&log->log_extents_lock[index]);
|
|
|
|
}
|
|
|
|
|
|
|
|
void btrfs_free_logged_extents(struct btrfs_root *log, u64 transid)
|
|
|
|
{
|
|
|
|
struct btrfs_ordered_extent *ordered;
|
|
|
|
int index = transid % 2;
|
|
|
|
|
|
|
|
spin_lock_irq(&log->log_extents_lock[index]);
|
|
|
|
while (!list_empty(&log->logged_list[index])) {
|
|
|
|
ordered = list_first_entry(&log->logged_list[index],
|
|
|
|
struct btrfs_ordered_extent,
|
|
|
|
log_list);
|
|
|
|
list_del_init(&ordered->log_list);
|
|
|
|
spin_unlock_irq(&log->log_extents_lock[index]);
|
|
|
|
btrfs_put_ordered_extent(ordered);
|
|
|
|
spin_lock_irq(&log->log_extents_lock[index]);
|
|
|
|
}
|
|
|
|
spin_unlock_irq(&log->log_extents_lock[index]);
|
|
|
|
}
|
|
|
|
|
2008-07-18 01:53:27 +08:00
|
|
|
/*
|
|
|
|
* used to drop a reference on an ordered extent. This will free
|
|
|
|
* the extent if the last reference is dropped
|
|
|
|
*/
|
2012-03-01 21:56:26 +08:00
|
|
|
void btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry)
|
2008-07-18 00:53:50 +08:00
|
|
|
{
|
2008-07-18 00:54:15 +08:00
|
|
|
struct list_head *cur;
|
|
|
|
struct btrfs_ordered_sum *sum;
|
|
|
|
|
Btrfs: add initial tracepoint support for btrfs
Tracepoints can provide insight into why btrfs hits bugs and be greatly
helpful for debugging, e.g
dd-7822 [000] 2121.641088: btrfs_inode_request: root = 5(FS_TREE), gen = 4, ino = 256, blocks = 8, disk_i_size = 0, last_trans = 8, logged_trans = 0
dd-7822 [000] 2121.641100: btrfs_inode_new: root = 5(FS_TREE), gen = 8, ino = 257, blocks = 0, disk_i_size = 0, last_trans = 0, logged_trans = 0
btrfs-transacti-7804 [001] 2146.935420: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29368320 (orig_level = 0), cow_buf = 29388800 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.935473: btrfs_cow_block: root = 1(ROOT_TREE), refs = 2, orig_buf = 29364224 (orig_level = 0), cow_buf = 29392896 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.972221: btrfs_transaction_commit: root = 1(ROOT_TREE), gen = 8
flush-btrfs-2-7821 [001] 2155.824210: btrfs_chunk_alloc: root = 3(CHUNK_TREE), offset = 1103101952, size = 1073741824, num_stripes = 1, sub_stripes = 0, type = DATA
flush-btrfs-2-7821 [001] 2155.824241: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29388800 (orig_level = 0), cow_buf = 29396992 (cow_level = 0)
flush-btrfs-2-7821 [001] 2155.824255: btrfs_cow_block: root = 4(DEV_TREE), refs = 2, orig_buf = 29372416 (orig_level = 0), cow_buf = 29401088 (cow_level = 0)
flush-btrfs-2-7821 [000] 2155.824329: btrfs_cow_block: root = 3(CHUNK_TREE), refs = 2, orig_buf = 20971520 (orig_level = 0), cow_buf = 20975616 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898019: btrfs_cow_block: root = 5(FS_TREE), refs = 2, orig_buf = 29384704 (orig_level = 0), cow_buf = 29405184 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898043: btrfs_cow_block: root = 7(CSUM_TREE), refs = 2, orig_buf = 29376512 (orig_level = 0), cow_buf = 29409280 (cow_level = 0)
Here is what I have added:
1) ordere_extent:
btrfs_ordered_extent_add
btrfs_ordered_extent_remove
btrfs_ordered_extent_start
btrfs_ordered_extent_put
These provide critical information to understand how ordered_extents are
updated.
2) extent_map:
btrfs_get_extent
extent_map is used in both read and write cases, and it is useful for tracking
how btrfs specific IO is running.
3) writepage:
__extent_writepage
btrfs_writepage_end_io_hook
Pages are cirtical resourses and produce a lot of corner cases during writeback,
so it is valuable to know how page is written to disk.
4) inode:
btrfs_inode_new
btrfs_inode_request
btrfs_inode_evict
These can show where and when a inode is created, when a inode is evicted.
5) sync:
btrfs_sync_file
btrfs_sync_fs
These show sync arguments.
6) transaction:
btrfs_transaction_commit
In transaction based filesystem, it will be useful to know the generation and
who does commit.
7) back reference and cow:
btrfs_delayed_tree_ref
btrfs_delayed_data_ref
btrfs_delayed_ref_head
btrfs_cow_block
Btrfs natively supports back references, these tracepoints are helpful on
understanding btrfs's COW mechanism.
8) chunk:
btrfs_chunk_alloc
btrfs_chunk_free
Chunk is a link between physical offset and logical offset, and stands for space
infomation in btrfs, and these are helpful on tracing space things.
9) reserved_extent:
btrfs_reserved_extent_alloc
btrfs_reserved_extent_free
These can show how btrfs uses its space.
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-03-24 19:18:59 +08:00
|
|
|
trace_btrfs_ordered_extent_put(entry->inode, entry);
|
|
|
|
|
2008-07-18 00:54:15 +08:00
|
|
|
if (atomic_dec_and_test(&entry->refs)) {
|
2012-05-03 02:00:54 +08:00
|
|
|
if (entry->inode)
|
|
|
|
btrfs_add_delayed_iput(entry->inode);
|
2009-01-06 10:25:51 +08:00
|
|
|
while (!list_empty(&entry->list)) {
|
2008-07-18 00:54:15 +08:00
|
|
|
cur = entry->list.next;
|
|
|
|
sum = list_entry(cur, struct btrfs_ordered_sum, list);
|
|
|
|
list_del(&sum->list);
|
|
|
|
kfree(sum);
|
|
|
|
}
|
2012-09-06 18:01:51 +08:00
|
|
|
kmem_cache_free(btrfs_ordered_extent_cache, entry);
|
2008-07-18 00:54:15 +08:00
|
|
|
}
|
2008-01-09 04:46:30 +08:00
|
|
|
}
|
2008-01-15 21:40:48 +08:00
|
|
|
|
2008-07-18 01:53:27 +08:00
|
|
|
/*
|
|
|
|
* remove an ordered extent from the tree. No references are dropped
|
2012-05-03 02:00:54 +08:00
|
|
|
* and waiters are woken up.
|
2008-07-18 01:53:27 +08:00
|
|
|
*/
|
2012-05-03 02:00:54 +08:00
|
|
|
void btrfs_remove_ordered_extent(struct inode *inode,
|
|
|
|
struct btrfs_ordered_extent *entry)
|
2008-01-15 21:40:48 +08:00
|
|
|
{
|
2008-07-18 00:53:50 +08:00
|
|
|
struct btrfs_ordered_inode_tree *tree;
|
2010-03-20 02:07:23 +08:00
|
|
|
struct btrfs_root *root = BTRFS_I(inode)->root;
|
2008-01-15 21:40:48 +08:00
|
|
|
struct rb_node *node;
|
|
|
|
|
2008-07-18 00:53:50 +08:00
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_lock_irq(&tree->lock);
|
2008-07-18 00:53:50 +08:00
|
|
|
node = &entry->rb_node;
|
2008-01-15 21:40:48 +08:00
|
|
|
rb_erase(node, &tree->tree);
|
2008-07-18 00:53:50 +08:00
|
|
|
tree->last = NULL;
|
|
|
|
set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags);
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_unlock_irq(&tree->lock);
|
2008-07-24 23:57:52 +08:00
|
|
|
|
2013-05-15 15:48:23 +08:00
|
|
|
spin_lock(&root->ordered_extent_lock);
|
2008-07-24 23:57:52 +08:00
|
|
|
list_del_init(&entry->root_extent_list);
|
2013-05-15 15:48:23 +08:00
|
|
|
root->nr_ordered_extents--;
|
2009-04-01 01:27:11 +08:00
|
|
|
|
Btrfs: add initial tracepoint support for btrfs
Tracepoints can provide insight into why btrfs hits bugs and be greatly
helpful for debugging, e.g
dd-7822 [000] 2121.641088: btrfs_inode_request: root = 5(FS_TREE), gen = 4, ino = 256, blocks = 8, disk_i_size = 0, last_trans = 8, logged_trans = 0
dd-7822 [000] 2121.641100: btrfs_inode_new: root = 5(FS_TREE), gen = 8, ino = 257, blocks = 0, disk_i_size = 0, last_trans = 0, logged_trans = 0
btrfs-transacti-7804 [001] 2146.935420: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29368320 (orig_level = 0), cow_buf = 29388800 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.935473: btrfs_cow_block: root = 1(ROOT_TREE), refs = 2, orig_buf = 29364224 (orig_level = 0), cow_buf = 29392896 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.972221: btrfs_transaction_commit: root = 1(ROOT_TREE), gen = 8
flush-btrfs-2-7821 [001] 2155.824210: btrfs_chunk_alloc: root = 3(CHUNK_TREE), offset = 1103101952, size = 1073741824, num_stripes = 1, sub_stripes = 0, type = DATA
flush-btrfs-2-7821 [001] 2155.824241: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29388800 (orig_level = 0), cow_buf = 29396992 (cow_level = 0)
flush-btrfs-2-7821 [001] 2155.824255: btrfs_cow_block: root = 4(DEV_TREE), refs = 2, orig_buf = 29372416 (orig_level = 0), cow_buf = 29401088 (cow_level = 0)
flush-btrfs-2-7821 [000] 2155.824329: btrfs_cow_block: root = 3(CHUNK_TREE), refs = 2, orig_buf = 20971520 (orig_level = 0), cow_buf = 20975616 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898019: btrfs_cow_block: root = 5(FS_TREE), refs = 2, orig_buf = 29384704 (orig_level = 0), cow_buf = 29405184 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898043: btrfs_cow_block: root = 7(CSUM_TREE), refs = 2, orig_buf = 29376512 (orig_level = 0), cow_buf = 29409280 (cow_level = 0)
Here is what I have added:
1) ordere_extent:
btrfs_ordered_extent_add
btrfs_ordered_extent_remove
btrfs_ordered_extent_start
btrfs_ordered_extent_put
These provide critical information to understand how ordered_extents are
updated.
2) extent_map:
btrfs_get_extent
extent_map is used in both read and write cases, and it is useful for tracking
how btrfs specific IO is running.
3) writepage:
__extent_writepage
btrfs_writepage_end_io_hook
Pages are cirtical resourses and produce a lot of corner cases during writeback,
so it is valuable to know how page is written to disk.
4) inode:
btrfs_inode_new
btrfs_inode_request
btrfs_inode_evict
These can show where and when a inode is created, when a inode is evicted.
5) sync:
btrfs_sync_file
btrfs_sync_fs
These show sync arguments.
6) transaction:
btrfs_transaction_commit
In transaction based filesystem, it will be useful to know the generation and
who does commit.
7) back reference and cow:
btrfs_delayed_tree_ref
btrfs_delayed_data_ref
btrfs_delayed_ref_head
btrfs_cow_block
Btrfs natively supports back references, these tracepoints are helpful on
understanding btrfs's COW mechanism.
8) chunk:
btrfs_chunk_alloc
btrfs_chunk_free
Chunk is a link between physical offset and logical offset, and stands for space
infomation in btrfs, and these are helpful on tracing space things.
9) reserved_extent:
btrfs_reserved_extent_alloc
btrfs_reserved_extent_free
These can show how btrfs uses its space.
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-03-24 19:18:59 +08:00
|
|
|
trace_btrfs_ordered_extent_remove(inode, entry);
|
|
|
|
|
2009-04-01 01:27:11 +08:00
|
|
|
/*
|
|
|
|
* we have no more ordered extents for this inode and
|
|
|
|
* no dirty pages. We can safely remove it from the
|
|
|
|
* list of ordered extents
|
|
|
|
*/
|
|
|
|
if (RB_EMPTY_ROOT(&tree->tree) &&
|
|
|
|
!mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY)) {
|
|
|
|
list_del_init(&BTRFS_I(inode)->ordered_operations);
|
|
|
|
}
|
2013-05-15 15:48:23 +08:00
|
|
|
|
|
|
|
if (!root->nr_ordered_extents) {
|
|
|
|
spin_lock(&root->fs_info->ordered_root_lock);
|
|
|
|
BUG_ON(list_empty(&root->ordered_root));
|
|
|
|
list_del_init(&root->ordered_root);
|
|
|
|
spin_unlock(&root->fs_info->ordered_root_lock);
|
|
|
|
}
|
|
|
|
spin_unlock(&root->ordered_extent_lock);
|
2008-07-18 00:53:50 +08:00
|
|
|
wake_up(&entry->wait);
|
2008-01-15 21:40:48 +08:00
|
|
|
}
|
|
|
|
|
2012-10-25 17:41:36 +08:00
|
|
|
static void btrfs_run_ordered_extent_work(struct btrfs_work *work)
|
|
|
|
{
|
|
|
|
struct btrfs_ordered_extent *ordered;
|
|
|
|
|
|
|
|
ordered = container_of(work, struct btrfs_ordered_extent, flush_work);
|
|
|
|
btrfs_start_ordered_extent(ordered->inode, ordered, 1);
|
|
|
|
complete(&ordered->completion);
|
|
|
|
}
|
|
|
|
|
2008-09-30 03:18:18 +08:00
|
|
|
/*
|
|
|
|
* wait for all the ordered extents in a root. This is done when balancing
|
|
|
|
* space between drives.
|
|
|
|
*/
|
2012-09-14 16:58:07 +08:00
|
|
|
void btrfs_wait_ordered_extents(struct btrfs_root *root, int delay_iput)
|
2008-07-24 23:57:52 +08:00
|
|
|
{
|
2012-10-25 17:41:36 +08:00
|
|
|
struct list_head splice, works;
|
|
|
|
struct btrfs_ordered_extent *ordered, *next;
|
2008-07-24 23:57:52 +08:00
|
|
|
struct inode *inode;
|
|
|
|
|
|
|
|
INIT_LIST_HEAD(&splice);
|
2012-10-25 17:41:36 +08:00
|
|
|
INIT_LIST_HEAD(&works);
|
2008-07-24 23:57:52 +08:00
|
|
|
|
2013-03-27 03:29:11 +08:00
|
|
|
mutex_lock(&root->fs_info->ordered_operations_mutex);
|
2013-05-15 15:48:23 +08:00
|
|
|
spin_lock(&root->ordered_extent_lock);
|
|
|
|
list_splice_init(&root->ordered_extents, &splice);
|
2008-09-26 22:05:38 +08:00
|
|
|
while (!list_empty(&splice)) {
|
2013-05-15 15:48:23 +08:00
|
|
|
ordered = list_first_entry(&splice, struct btrfs_ordered_extent,
|
|
|
|
root_extent_list);
|
|
|
|
list_move_tail(&ordered->root_extent_list,
|
|
|
|
&root->ordered_extents);
|
2008-07-24 23:57:52 +08:00
|
|
|
/*
|
2008-09-26 22:05:38 +08:00
|
|
|
* the inode may be getting freed (in sys_unlink path).
|
2008-07-24 23:57:52 +08:00
|
|
|
*/
|
2008-09-26 22:05:38 +08:00
|
|
|
inode = igrab(ordered->inode);
|
2013-05-15 15:48:23 +08:00
|
|
|
if (!inode) {
|
|
|
|
cond_resched_lock(&root->ordered_extent_lock);
|
|
|
|
continue;
|
|
|
|
}
|
2008-09-26 22:05:38 +08:00
|
|
|
|
2013-05-15 15:48:23 +08:00
|
|
|
atomic_inc(&ordered->refs);
|
|
|
|
spin_unlock(&root->ordered_extent_lock);
|
2008-07-24 23:57:52 +08:00
|
|
|
|
2013-05-15 15:48:23 +08:00
|
|
|
ordered->flush_work.func = btrfs_run_ordered_extent_work;
|
|
|
|
list_add_tail(&ordered->work_list, &works);
|
|
|
|
btrfs_queue_worker(&root->fs_info->flush_workers,
|
|
|
|
&ordered->flush_work);
|
2008-07-24 23:57:52 +08:00
|
|
|
|
2012-10-25 17:41:36 +08:00
|
|
|
cond_resched();
|
2013-05-15 15:48:23 +08:00
|
|
|
spin_lock(&root->ordered_extent_lock);
|
2008-07-24 23:57:52 +08:00
|
|
|
}
|
2013-05-15 15:48:23 +08:00
|
|
|
spin_unlock(&root->ordered_extent_lock);
|
2012-10-25 17:41:36 +08:00
|
|
|
|
|
|
|
list_for_each_entry_safe(ordered, next, &works, work_list) {
|
|
|
|
list_del_init(&ordered->work_list);
|
|
|
|
wait_for_completion(&ordered->completion);
|
|
|
|
|
|
|
|
inode = ordered->inode;
|
|
|
|
btrfs_put_ordered_extent(ordered);
|
|
|
|
if (delay_iput)
|
|
|
|
btrfs_add_delayed_iput(inode);
|
|
|
|
else
|
|
|
|
iput(inode);
|
|
|
|
|
|
|
|
cond_resched();
|
|
|
|
}
|
2013-03-27 03:29:11 +08:00
|
|
|
mutex_unlock(&root->fs_info->ordered_operations_mutex);
|
2008-07-24 23:57:52 +08:00
|
|
|
}
|
|
|
|
|
2013-05-15 15:48:23 +08:00
|
|
|
void btrfs_wait_all_ordered_extents(struct btrfs_fs_info *fs_info,
|
|
|
|
int delay_iput)
|
|
|
|
{
|
|
|
|
struct btrfs_root *root;
|
|
|
|
struct list_head splice;
|
|
|
|
|
|
|
|
INIT_LIST_HEAD(&splice);
|
|
|
|
|
|
|
|
spin_lock(&fs_info->ordered_root_lock);
|
|
|
|
list_splice_init(&fs_info->ordered_roots, &splice);
|
|
|
|
while (!list_empty(&splice)) {
|
|
|
|
root = list_first_entry(&splice, struct btrfs_root,
|
|
|
|
ordered_root);
|
|
|
|
root = btrfs_grab_fs_root(root);
|
|
|
|
BUG_ON(!root);
|
|
|
|
list_move_tail(&root->ordered_root,
|
|
|
|
&fs_info->ordered_roots);
|
|
|
|
spin_unlock(&fs_info->ordered_root_lock);
|
|
|
|
|
|
|
|
btrfs_wait_ordered_extents(root, delay_iput);
|
|
|
|
btrfs_put_fs_root(root);
|
|
|
|
|
|
|
|
spin_lock(&fs_info->ordered_root_lock);
|
|
|
|
}
|
|
|
|
spin_unlock(&fs_info->ordered_root_lock);
|
|
|
|
}
|
|
|
|
|
2009-04-01 01:27:11 +08:00
|
|
|
/*
|
|
|
|
* this is used during transaction commit to write all the inodes
|
|
|
|
* added to the ordered operation list. These files must be fully on
|
|
|
|
* disk before the transaction commits.
|
|
|
|
*
|
|
|
|
* we have two modes here, one is to just start the IO via filemap_flush
|
|
|
|
* and the other is to wait for all the io. When we wait, we have an
|
|
|
|
* extra check to make sure the ordered operation list really is empty
|
|
|
|
* before we return
|
|
|
|
*/
|
2013-02-14 00:09:14 +08:00
|
|
|
int btrfs_run_ordered_operations(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root, int wait)
|
2009-04-01 01:27:11 +08:00
|
|
|
{
|
|
|
|
struct btrfs_inode *btrfs_inode;
|
|
|
|
struct inode *inode;
|
2013-02-14 00:09:14 +08:00
|
|
|
struct btrfs_transaction *cur_trans = trans->transaction;
|
2009-04-01 01:27:11 +08:00
|
|
|
struct list_head splice;
|
2012-10-25 17:31:03 +08:00
|
|
|
struct list_head works;
|
|
|
|
struct btrfs_delalloc_work *work, *next;
|
|
|
|
int ret = 0;
|
2009-04-01 01:27:11 +08:00
|
|
|
|
|
|
|
INIT_LIST_HEAD(&splice);
|
2012-10-25 17:31:03 +08:00
|
|
|
INIT_LIST_HEAD(&works);
|
2009-04-01 01:27:11 +08:00
|
|
|
|
2013-08-14 23:33:56 +08:00
|
|
|
mutex_lock(&root->fs_info->ordered_extent_flush_mutex);
|
2013-05-15 15:48:23 +08:00
|
|
|
spin_lock(&root->fs_info->ordered_root_lock);
|
2013-02-14 00:09:14 +08:00
|
|
|
list_splice_init(&cur_trans->ordered_operations, &splice);
|
2009-04-01 01:27:11 +08:00
|
|
|
while (!list_empty(&splice)) {
|
|
|
|
btrfs_inode = list_entry(splice.next, struct btrfs_inode,
|
|
|
|
ordered_operations);
|
|
|
|
inode = &btrfs_inode->vfs_inode;
|
|
|
|
|
|
|
|
list_del_init(&btrfs_inode->ordered_operations);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* the inode may be getting freed (in sys_unlink path).
|
|
|
|
*/
|
|
|
|
inode = igrab(inode);
|
2012-10-25 17:31:03 +08:00
|
|
|
if (!inode)
|
|
|
|
continue;
|
2013-01-22 18:52:04 +08:00
|
|
|
|
|
|
|
if (!wait)
|
|
|
|
list_add_tail(&BTRFS_I(inode)->ordered_operations,
|
2013-02-14 00:09:14 +08:00
|
|
|
&cur_trans->ordered_operations);
|
2013-05-15 15:48:23 +08:00
|
|
|
spin_unlock(&root->fs_info->ordered_root_lock);
|
2009-04-01 01:27:11 +08:00
|
|
|
|
2012-10-25 17:31:03 +08:00
|
|
|
work = btrfs_alloc_delalloc_work(inode, wait, 1);
|
|
|
|
if (!work) {
|
2013-05-15 15:48:23 +08:00
|
|
|
spin_lock(&root->fs_info->ordered_root_lock);
|
2012-10-25 17:31:03 +08:00
|
|
|
if (list_empty(&BTRFS_I(inode)->ordered_operations))
|
|
|
|
list_add_tail(&btrfs_inode->ordered_operations,
|
|
|
|
&splice);
|
|
|
|
list_splice_tail(&splice,
|
2013-02-14 00:09:14 +08:00
|
|
|
&cur_trans->ordered_operations);
|
2013-05-15 15:48:23 +08:00
|
|
|
spin_unlock(&root->fs_info->ordered_root_lock);
|
2012-10-25 17:31:03 +08:00
|
|
|
ret = -ENOMEM;
|
|
|
|
goto out;
|
2009-04-01 01:27:11 +08:00
|
|
|
}
|
2012-10-25 17:31:03 +08:00
|
|
|
list_add_tail(&work->list, &works);
|
|
|
|
btrfs_queue_worker(&root->fs_info->flush_workers,
|
|
|
|
&work->work);
|
2009-04-01 01:27:11 +08:00
|
|
|
|
|
|
|
cond_resched();
|
2013-05-15 15:48:23 +08:00
|
|
|
spin_lock(&root->fs_info->ordered_root_lock);
|
2009-04-01 01:27:11 +08:00
|
|
|
}
|
2013-05-15 15:48:23 +08:00
|
|
|
spin_unlock(&root->fs_info->ordered_root_lock);
|
2012-10-25 17:31:03 +08:00
|
|
|
out:
|
|
|
|
list_for_each_entry_safe(work, next, &works, list) {
|
|
|
|
list_del_init(&work->list);
|
|
|
|
btrfs_wait_and_free_delalloc_work(work);
|
|
|
|
}
|
2013-08-14 23:33:56 +08:00
|
|
|
mutex_unlock(&root->fs_info->ordered_extent_flush_mutex);
|
2012-10-25 17:31:03 +08:00
|
|
|
return ret;
|
2009-04-01 01:27:11 +08:00
|
|
|
}
|
|
|
|
|
2008-07-18 01:53:27 +08:00
|
|
|
/*
|
|
|
|
* Used to start IO or wait for a given ordered extent to finish.
|
|
|
|
*
|
|
|
|
* If wait is one, this effectively waits on page writeback for all the pages
|
|
|
|
* in the extent, and it waits on the io completion code to insert
|
|
|
|
* metadata into the btree corresponding to the extent
|
|
|
|
*/
|
|
|
|
void btrfs_start_ordered_extent(struct inode *inode,
|
|
|
|
struct btrfs_ordered_extent *entry,
|
|
|
|
int wait)
|
2008-07-18 00:53:50 +08:00
|
|
|
{
|
|
|
|
u64 start = entry->file_offset;
|
|
|
|
u64 end = start + entry->len - 1;
|
2008-05-27 22:55:43 +08:00
|
|
|
|
Btrfs: add initial tracepoint support for btrfs
Tracepoints can provide insight into why btrfs hits bugs and be greatly
helpful for debugging, e.g
dd-7822 [000] 2121.641088: btrfs_inode_request: root = 5(FS_TREE), gen = 4, ino = 256, blocks = 8, disk_i_size = 0, last_trans = 8, logged_trans = 0
dd-7822 [000] 2121.641100: btrfs_inode_new: root = 5(FS_TREE), gen = 8, ino = 257, blocks = 0, disk_i_size = 0, last_trans = 0, logged_trans = 0
btrfs-transacti-7804 [001] 2146.935420: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29368320 (orig_level = 0), cow_buf = 29388800 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.935473: btrfs_cow_block: root = 1(ROOT_TREE), refs = 2, orig_buf = 29364224 (orig_level = 0), cow_buf = 29392896 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.972221: btrfs_transaction_commit: root = 1(ROOT_TREE), gen = 8
flush-btrfs-2-7821 [001] 2155.824210: btrfs_chunk_alloc: root = 3(CHUNK_TREE), offset = 1103101952, size = 1073741824, num_stripes = 1, sub_stripes = 0, type = DATA
flush-btrfs-2-7821 [001] 2155.824241: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29388800 (orig_level = 0), cow_buf = 29396992 (cow_level = 0)
flush-btrfs-2-7821 [001] 2155.824255: btrfs_cow_block: root = 4(DEV_TREE), refs = 2, orig_buf = 29372416 (orig_level = 0), cow_buf = 29401088 (cow_level = 0)
flush-btrfs-2-7821 [000] 2155.824329: btrfs_cow_block: root = 3(CHUNK_TREE), refs = 2, orig_buf = 20971520 (orig_level = 0), cow_buf = 20975616 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898019: btrfs_cow_block: root = 5(FS_TREE), refs = 2, orig_buf = 29384704 (orig_level = 0), cow_buf = 29405184 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898043: btrfs_cow_block: root = 7(CSUM_TREE), refs = 2, orig_buf = 29376512 (orig_level = 0), cow_buf = 29409280 (cow_level = 0)
Here is what I have added:
1) ordere_extent:
btrfs_ordered_extent_add
btrfs_ordered_extent_remove
btrfs_ordered_extent_start
btrfs_ordered_extent_put
These provide critical information to understand how ordered_extents are
updated.
2) extent_map:
btrfs_get_extent
extent_map is used in both read and write cases, and it is useful for tracking
how btrfs specific IO is running.
3) writepage:
__extent_writepage
btrfs_writepage_end_io_hook
Pages are cirtical resourses and produce a lot of corner cases during writeback,
so it is valuable to know how page is written to disk.
4) inode:
btrfs_inode_new
btrfs_inode_request
btrfs_inode_evict
These can show where and when a inode is created, when a inode is evicted.
5) sync:
btrfs_sync_file
btrfs_sync_fs
These show sync arguments.
6) transaction:
btrfs_transaction_commit
In transaction based filesystem, it will be useful to know the generation and
who does commit.
7) back reference and cow:
btrfs_delayed_tree_ref
btrfs_delayed_data_ref
btrfs_delayed_ref_head
btrfs_cow_block
Btrfs natively supports back references, these tracepoints are helpful on
understanding btrfs's COW mechanism.
8) chunk:
btrfs_chunk_alloc
btrfs_chunk_free
Chunk is a link between physical offset and logical offset, and stands for space
infomation in btrfs, and these are helpful on tracing space things.
9) reserved_extent:
btrfs_reserved_extent_alloc
btrfs_reserved_extent_free
These can show how btrfs uses its space.
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-03-24 19:18:59 +08:00
|
|
|
trace_btrfs_ordered_extent_start(inode, entry);
|
|
|
|
|
2008-07-18 01:53:27 +08:00
|
|
|
/*
|
|
|
|
* pages in the range can be dirty, clean or writeback. We
|
|
|
|
* start IO on any dirty ones so the wait doesn't stall waiting
|
2012-07-25 23:12:06 +08:00
|
|
|
* for the flusher thread to find them
|
2008-07-18 01:53:27 +08:00
|
|
|
*/
|
2010-05-23 23:00:55 +08:00
|
|
|
if (!test_bit(BTRFS_ORDERED_DIRECT, &entry->flags))
|
|
|
|
filemap_fdatawrite_range(inode->i_mapping, start, end);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 02:49:59 +08:00
|
|
|
if (wait) {
|
2008-07-18 00:53:50 +08:00
|
|
|
wait_event(entry->wait, test_bit(BTRFS_ORDERED_COMPLETE,
|
|
|
|
&entry->flags));
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 02:49:59 +08:00
|
|
|
}
|
2008-07-18 00:53:50 +08:00
|
|
|
}
|
2008-01-15 21:40:48 +08:00
|
|
|
|
2008-07-18 01:53:27 +08:00
|
|
|
/*
|
|
|
|
* Used to wait on ordered extents across a large range of bytes.
|
|
|
|
*/
|
2012-03-01 21:56:26 +08:00
|
|
|
void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len)
|
2008-07-18 00:53:50 +08:00
|
|
|
{
|
|
|
|
u64 end;
|
2008-07-19 08:42:20 +08:00
|
|
|
u64 orig_end;
|
2008-07-18 00:53:50 +08:00
|
|
|
struct btrfs_ordered_extent *ordered;
|
2008-07-19 08:42:20 +08:00
|
|
|
|
|
|
|
if (start + len < start) {
|
2008-07-22 23:18:09 +08:00
|
|
|
orig_end = INT_LIMIT(loff_t);
|
2008-07-19 08:42:20 +08:00
|
|
|
} else {
|
|
|
|
orig_end = start + len - 1;
|
2008-07-22 23:18:09 +08:00
|
|
|
if (orig_end > INT_LIMIT(loff_t))
|
|
|
|
orig_end = INT_LIMIT(loff_t);
|
2008-07-19 08:42:20 +08:00
|
|
|
}
|
2012-04-24 02:41:09 +08:00
|
|
|
|
2008-07-19 08:42:20 +08:00
|
|
|
/* start IO across the range first to instantiate any delalloc
|
|
|
|
* extents
|
|
|
|
*/
|
2012-06-09 03:26:47 +08:00
|
|
|
filemap_fdatawrite_range(inode->i_mapping, start, orig_end);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* So with compression we will find and lock a dirty page and clear the
|
|
|
|
* first one as dirty, setup an async extent, and immediately return
|
|
|
|
* with the entire range locked but with nobody actually marked with
|
|
|
|
* writeback. So we can't just filemap_write_and_wait_range() and
|
|
|
|
* expect it to work since it will just kick off a thread to do the
|
|
|
|
* actual work. So we need to call filemap_fdatawrite_range _again_
|
|
|
|
* since it will wait on the page lock, which won't be unlocked until
|
|
|
|
* after the pages have been marked as writeback and so we're good to go
|
|
|
|
* from there. We have to do this otherwise we'll miss the ordered
|
|
|
|
* extents and that results in badness. Please Josef, do not think you
|
|
|
|
* know better and pull this out at some point in the future, it is
|
|
|
|
* right and you are wrong.
|
|
|
|
*/
|
|
|
|
if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
|
|
|
|
&BTRFS_I(inode)->runtime_flags))
|
|
|
|
filemap_fdatawrite_range(inode->i_mapping, start, orig_end);
|
|
|
|
|
|
|
|
filemap_fdatawait_range(inode->i_mapping, start, orig_end);
|
2008-07-19 08:42:20 +08:00
|
|
|
|
2008-07-22 23:18:09 +08:00
|
|
|
end = orig_end;
|
2009-01-06 10:25:51 +08:00
|
|
|
while (1) {
|
2008-07-18 00:53:50 +08:00
|
|
|
ordered = btrfs_lookup_first_ordered_extent(inode, end);
|
2009-01-06 10:25:51 +08:00
|
|
|
if (!ordered)
|
2008-07-18 00:53:50 +08:00
|
|
|
break;
|
2008-07-19 08:42:20 +08:00
|
|
|
if (ordered->file_offset > orig_end) {
|
2008-07-18 00:53:50 +08:00
|
|
|
btrfs_put_ordered_extent(ordered);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (ordered->file_offset + ordered->len < start) {
|
|
|
|
btrfs_put_ordered_extent(ordered);
|
|
|
|
break;
|
|
|
|
}
|
2008-07-19 08:42:20 +08:00
|
|
|
btrfs_start_ordered_extent(inode, ordered, 1);
|
2008-07-18 00:53:50 +08:00
|
|
|
end = ordered->file_offset;
|
|
|
|
btrfs_put_ordered_extent(ordered);
|
2008-07-19 08:42:20 +08:00
|
|
|
if (end == 0 || end == start)
|
2008-07-18 00:53:50 +08:00
|
|
|
break;
|
|
|
|
end--;
|
|
|
|
}
|
2008-01-15 21:40:48 +08:00
|
|
|
}
|
|
|
|
|
2008-07-18 01:53:27 +08:00
|
|
|
/*
|
|
|
|
* find an ordered extent corresponding to file_offset. return NULL if
|
|
|
|
* nothing is found, otherwise take a reference on the extent and return it
|
|
|
|
*/
|
2008-07-18 00:53:50 +08:00
|
|
|
struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode,
|
|
|
|
u64 file_offset)
|
|
|
|
{
|
|
|
|
struct btrfs_ordered_inode_tree *tree;
|
|
|
|
struct rb_node *node;
|
|
|
|
struct btrfs_ordered_extent *entry = NULL;
|
|
|
|
|
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_lock_irq(&tree->lock);
|
2008-07-18 00:53:50 +08:00
|
|
|
node = tree_search(tree, file_offset);
|
|
|
|
if (!node)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
|
|
|
|
if (!offset_in_entry(entry, file_offset))
|
|
|
|
entry = NULL;
|
|
|
|
if (entry)
|
|
|
|
atomic_inc(&entry->refs);
|
|
|
|
out:
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_unlock_irq(&tree->lock);
|
2008-07-18 00:53:50 +08:00
|
|
|
return entry;
|
|
|
|
}
|
|
|
|
|
2010-05-23 23:00:55 +08:00
|
|
|
/* Since the DIO code tries to lock a wide area we need to look for any ordered
|
|
|
|
* extents that exist in the range, rather than just the start of the range.
|
|
|
|
*/
|
|
|
|
struct btrfs_ordered_extent *btrfs_lookup_ordered_range(struct inode *inode,
|
|
|
|
u64 file_offset,
|
|
|
|
u64 len)
|
|
|
|
{
|
|
|
|
struct btrfs_ordered_inode_tree *tree;
|
|
|
|
struct rb_node *node;
|
|
|
|
struct btrfs_ordered_extent *entry = NULL;
|
|
|
|
|
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_lock_irq(&tree->lock);
|
2010-05-23 23:00:55 +08:00
|
|
|
node = tree_search(tree, file_offset);
|
|
|
|
if (!node) {
|
|
|
|
node = tree_search(tree, file_offset + len);
|
|
|
|
if (!node)
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
while (1) {
|
|
|
|
entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
|
|
|
|
if (range_overlaps(entry, file_offset, len))
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (entry->file_offset >= file_offset + len) {
|
|
|
|
entry = NULL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
entry = NULL;
|
|
|
|
node = rb_next(node);
|
|
|
|
if (!node)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
out:
|
|
|
|
if (entry)
|
|
|
|
atomic_inc(&entry->refs);
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_unlock_irq(&tree->lock);
|
2010-05-23 23:00:55 +08:00
|
|
|
return entry;
|
|
|
|
}
|
|
|
|
|
2008-07-18 01:53:27 +08:00
|
|
|
/*
|
|
|
|
* lookup and return any extent before 'file_offset'. NULL is returned
|
|
|
|
* if none is found
|
|
|
|
*/
|
2008-07-18 00:53:50 +08:00
|
|
|
struct btrfs_ordered_extent *
|
2009-01-06 10:25:51 +08:00
|
|
|
btrfs_lookup_first_ordered_extent(struct inode *inode, u64 file_offset)
|
2008-07-18 00:53:50 +08:00
|
|
|
{
|
|
|
|
struct btrfs_ordered_inode_tree *tree;
|
|
|
|
struct rb_node *node;
|
|
|
|
struct btrfs_ordered_extent *entry = NULL;
|
|
|
|
|
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_lock_irq(&tree->lock);
|
2008-07-18 00:53:50 +08:00
|
|
|
node = tree_search(tree, file_offset);
|
|
|
|
if (!node)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
|
|
|
|
atomic_inc(&entry->refs);
|
|
|
|
out:
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_unlock_irq(&tree->lock);
|
2008-07-18 00:53:50 +08:00
|
|
|
return entry;
|
2008-04-25 20:51:48 +08:00
|
|
|
}
|
2008-07-18 00:54:05 +08:00
|
|
|
|
2008-07-18 01:53:27 +08:00
|
|
|
/*
|
|
|
|
* After an extent is done, call this to conditionally update the on disk
|
|
|
|
* i_size. i_size is updated to cover any fully written part of the file.
|
|
|
|
*/
|
2009-11-12 17:34:21 +08:00
|
|
|
int btrfs_ordered_update_i_size(struct inode *inode, u64 offset,
|
2008-07-18 00:54:05 +08:00
|
|
|
struct btrfs_ordered_extent *ordered)
|
|
|
|
{
|
|
|
|
struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree;
|
|
|
|
u64 disk_i_size;
|
|
|
|
u64 new_i_size;
|
2009-11-12 17:34:21 +08:00
|
|
|
u64 i_size = i_size_read(inode);
|
2008-07-18 00:54:05 +08:00
|
|
|
struct rb_node *node;
|
2009-11-12 17:34:21 +08:00
|
|
|
struct rb_node *prev = NULL;
|
2008-07-18 00:54:05 +08:00
|
|
|
struct btrfs_ordered_extent *test;
|
2009-11-12 17:34:21 +08:00
|
|
|
int ret = 1;
|
|
|
|
|
2013-08-30 01:57:21 +08:00
|
|
|
spin_lock_irq(&tree->lock);
|
|
|
|
if (ordered) {
|
2009-11-12 17:34:21 +08:00
|
|
|
offset = entry_end(ordered);
|
2013-08-30 01:57:21 +08:00
|
|
|
if (test_bit(BTRFS_ORDERED_TRUNCATED, &ordered->flags))
|
|
|
|
offset = min(offset,
|
|
|
|
ordered->file_offset +
|
|
|
|
ordered->truncated_len);
|
|
|
|
} else {
|
2009-12-28 13:01:58 +08:00
|
|
|
offset = ALIGN(offset, BTRFS_I(inode)->root->sectorsize);
|
2013-08-30 01:57:21 +08:00
|
|
|
}
|
2008-07-18 00:54:05 +08:00
|
|
|
disk_i_size = BTRFS_I(inode)->disk_i_size;
|
|
|
|
|
2009-11-12 17:34:21 +08:00
|
|
|
/* truncate file */
|
|
|
|
if (disk_i_size > i_size) {
|
|
|
|
BTRFS_I(inode)->disk_i_size = i_size;
|
|
|
|
ret = 0;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2008-07-18 00:54:05 +08:00
|
|
|
/*
|
|
|
|
* if the disk i_size is already at the inode->i_size, or
|
|
|
|
* this ordered extent is inside the disk i_size, we're done
|
|
|
|
*/
|
2013-01-31 03:17:31 +08:00
|
|
|
if (disk_i_size == i_size)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We still need to update disk_i_size if outstanding_isize is greater
|
|
|
|
* than disk_i_size.
|
|
|
|
*/
|
|
|
|
if (offset <= disk_i_size &&
|
|
|
|
(!ordered || ordered->outstanding_isize <= disk_i_size))
|
2008-07-18 00:54:05 +08:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* walk backward from this ordered extent to disk_i_size.
|
|
|
|
* if we find an ordered extent then we can't update disk i_size
|
|
|
|
* yet
|
|
|
|
*/
|
2009-11-12 17:34:21 +08:00
|
|
|
if (ordered) {
|
|
|
|
node = rb_prev(&ordered->rb_node);
|
|
|
|
} else {
|
|
|
|
prev = tree_search(tree, offset);
|
|
|
|
/*
|
|
|
|
* we insert file extents without involving ordered struct,
|
|
|
|
* so there should be no ordered struct cover this offset
|
|
|
|
*/
|
|
|
|
if (prev) {
|
|
|
|
test = rb_entry(prev, struct btrfs_ordered_extent,
|
|
|
|
rb_node);
|
|
|
|
BUG_ON(offset_in_entry(test, offset));
|
|
|
|
}
|
|
|
|
node = prev;
|
|
|
|
}
|
2012-05-03 02:00:54 +08:00
|
|
|
for (; node; node = rb_prev(node)) {
|
2008-07-18 00:54:05 +08:00
|
|
|
test = rb_entry(node, struct btrfs_ordered_extent, rb_node);
|
2012-05-03 02:00:54 +08:00
|
|
|
|
|
|
|
/* We treat this entry as if it doesnt exist */
|
|
|
|
if (test_bit(BTRFS_ORDERED_UPDATED_ISIZE, &test->flags))
|
|
|
|
continue;
|
2008-07-18 00:54:05 +08:00
|
|
|
if (test->file_offset + test->len <= disk_i_size)
|
|
|
|
break;
|
2009-11-12 17:34:21 +08:00
|
|
|
if (test->file_offset >= i_size)
|
2008-07-18 00:54:05 +08:00
|
|
|
break;
|
2013-01-31 03:31:31 +08:00
|
|
|
if (entry_end(test) > disk_i_size) {
|
2012-09-06 18:01:21 +08:00
|
|
|
/*
|
|
|
|
* we don't update disk_i_size now, so record this
|
|
|
|
* undealt i_size. Or we will not know the real
|
|
|
|
* i_size.
|
|
|
|
*/
|
|
|
|
if (test->outstanding_isize < offset)
|
|
|
|
test->outstanding_isize = offset;
|
|
|
|
if (ordered &&
|
|
|
|
ordered->outstanding_isize >
|
|
|
|
test->outstanding_isize)
|
|
|
|
test->outstanding_isize =
|
|
|
|
ordered->outstanding_isize;
|
2008-07-18 00:54:05 +08:00
|
|
|
goto out;
|
2012-05-03 02:00:54 +08:00
|
|
|
}
|
2008-07-18 00:54:05 +08:00
|
|
|
}
|
2012-09-06 18:01:21 +08:00
|
|
|
new_i_size = min_t(u64, offset, i_size);
|
2008-07-18 00:54:05 +08:00
|
|
|
|
|
|
|
/*
|
2012-09-06 18:01:21 +08:00
|
|
|
* Some ordered extents may completed before the current one, and
|
|
|
|
* we hold the real i_size in ->outstanding_isize.
|
2008-07-18 00:54:05 +08:00
|
|
|
*/
|
2012-09-06 18:01:21 +08:00
|
|
|
if (ordered && ordered->outstanding_isize > new_i_size)
|
|
|
|
new_i_size = min_t(u64, ordered->outstanding_isize, i_size);
|
2008-07-18 00:54:05 +08:00
|
|
|
BTRFS_I(inode)->disk_i_size = new_i_size;
|
2009-11-12 17:34:21 +08:00
|
|
|
ret = 0;
|
2008-07-18 00:54:05 +08:00
|
|
|
out:
|
2009-11-12 17:34:21 +08:00
|
|
|
/*
|
2012-05-03 02:00:54 +08:00
|
|
|
* We need to do this because we can't remove ordered extents until
|
|
|
|
* after the i_disk_size has been updated and then the inode has been
|
|
|
|
* updated to reflect the change, so we need to tell anybody who finds
|
|
|
|
* this ordered extent that we've already done all the real work, we
|
|
|
|
* just haven't completed all the other work.
|
2009-11-12 17:34:21 +08:00
|
|
|
*/
|
|
|
|
if (ordered)
|
2012-05-03 02:00:54 +08:00
|
|
|
set_bit(BTRFS_ORDERED_UPDATED_ISIZE, &ordered->flags);
|
|
|
|
spin_unlock_irq(&tree->lock);
|
2009-11-12 17:34:21 +08:00
|
|
|
return ret;
|
2008-07-18 00:54:05 +08:00
|
|
|
}
|
2008-07-18 00:54:15 +08:00
|
|
|
|
2008-07-18 01:53:27 +08:00
|
|
|
/*
|
|
|
|
* search the ordered extents for one corresponding to 'offset' and
|
|
|
|
* try to find a checksum. This is used because we allow pages to
|
|
|
|
* be reclaimed before their checksum is actually put into the btree
|
|
|
|
*/
|
Btrfs: move data checksumming into a dedicated tree
Btrfs stores checksums for each data block. Until now, they have
been stored in the subvolume trees, indexed by the inode that is
referencing the data block. This means that when we read the inode,
we've probably read in at least some checksums as well.
But, this has a few problems:
* The checksums are indexed by logical offset in the file. When
compression is on, this means we have to do the expensive checksumming
on the uncompressed data. It would be faster if we could checksum
the compressed data instead.
* If we implement encryption, we'll be checksumming the plain text and
storing that on disk. This is significantly less secure.
* For either compression or encryption, we have to get the plain text
back before we can verify the checksum as correct. This makes the raid
layer balancing and extent moving much more expensive.
* It makes the front end caching code more complex, as we have touch
the subvolume and inodes as we cache extents.
* There is potentitally one copy of the checksum in each subvolume
referencing an extent.
The solution used here is to store the extent checksums in a dedicated
tree. This allows us to index the checksums by phyiscal extent
start and length. It means:
* The checksum is against the data stored on disk, after any compression
or encryption is done.
* The checksum is stored in a central location, and can be verified without
following back references, or reading inodes.
This makes compression significantly faster by reducing the amount of
data that needs to be checksummed. It will also allow much faster
raid management code in general.
The checksums are indexed by a key with a fixed objectid (a magic value
in ctree.h) and offset set to the starting byte of the extent. This
allows us to copy the checksum items into the fsync log tree directly (or
any other tree), without having to invent a second format for them.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-12-09 05:58:54 +08:00
|
|
|
int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr,
|
2013-04-05 15:20:56 +08:00
|
|
|
u32 *sum, int len)
|
2008-07-18 00:54:15 +08:00
|
|
|
{
|
|
|
|
struct btrfs_ordered_sum *ordered_sum;
|
|
|
|
struct btrfs_ordered_extent *ordered;
|
|
|
|
struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree;
|
2008-07-18 18:17:13 +08:00
|
|
|
unsigned long num_sectors;
|
|
|
|
unsigned long i;
|
|
|
|
u32 sectorsize = BTRFS_I(inode)->root->sectorsize;
|
2013-04-05 15:20:56 +08:00
|
|
|
int index = 0;
|
2008-07-18 00:54:15 +08:00
|
|
|
|
|
|
|
ordered = btrfs_lookup_ordered_extent(inode, offset);
|
|
|
|
if (!ordered)
|
2013-04-05 15:20:56 +08:00
|
|
|
return 0;
|
2008-07-18 00:54:15 +08:00
|
|
|
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_lock_irq(&tree->lock);
|
2009-01-21 23:59:08 +08:00
|
|
|
list_for_each_entry_reverse(ordered_sum, &ordered->list, list) {
|
2013-04-05 15:20:56 +08:00
|
|
|
if (disk_bytenr >= ordered_sum->bytenr &&
|
|
|
|
disk_bytenr < ordered_sum->bytenr + ordered_sum->len) {
|
|
|
|
i = (disk_bytenr - ordered_sum->bytenr) >>
|
|
|
|
inode->i_sb->s_blocksize_bits;
|
|
|
|
num_sectors = ordered_sum->len >>
|
|
|
|
inode->i_sb->s_blocksize_bits;
|
Btrfs: remove btrfs_sector_sum structure
Using the structure btrfs_sector_sum to keep the checksum value is
unnecessary, because the extents that btrfs_sector_sum points to are
continuous, we can find out the expected checksums by btrfs_ordered_sum's
bytenr and the offset, so we can remove btrfs_sector_sum's bytenr. After
removing bytenr, there is only one member in the structure, so it makes
no sense to keep the structure, just remove it, and use a u32 array to
store the checksum value.
By this change, we don't use the while loop to get the checksums one by
one. Now, we can get several checksum value at one time, it improved the
performance by ~74% on my SSD (31MB/s -> 54MB/s).
test command:
# dd if=/dev/zero of=/mnt/btrfs/file0 bs=1M count=1024 oflag=sync
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
2013-06-19 10:36:09 +08:00
|
|
|
num_sectors = min_t(int, len - index, num_sectors - i);
|
|
|
|
memcpy(sum + index, ordered_sum->sums + i,
|
|
|
|
num_sectors);
|
|
|
|
|
|
|
|
index += (int)num_sectors;
|
|
|
|
if (index == len)
|
|
|
|
goto out;
|
|
|
|
disk_bytenr += num_sectors * sectorsize;
|
2008-07-18 00:54:15 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
out:
|
2012-05-03 02:00:54 +08:00
|
|
|
spin_unlock_irq(&tree->lock);
|
2008-07-24 21:41:53 +08:00
|
|
|
btrfs_put_ordered_extent(ordered);
|
2013-04-05 15:20:56 +08:00
|
|
|
return index;
|
2008-07-18 00:54:15 +08:00
|
|
|
}
|
|
|
|
|
2008-07-22 23:18:09 +08:00
|
|
|
|
2009-04-01 01:27:11 +08:00
|
|
|
/*
|
|
|
|
* add a given inode to the list of inodes that must be fully on
|
|
|
|
* disk before a transaction commit finishes.
|
|
|
|
*
|
|
|
|
* This basically gives us the ext3 style data=ordered mode, and it is mostly
|
|
|
|
* used to make sure renamed files are fully on disk.
|
|
|
|
*
|
|
|
|
* It is a noop if the inode is already fully on disk.
|
|
|
|
*
|
|
|
|
* If trans is not null, we'll do a friendly check for a transaction that
|
|
|
|
* is already flushing things and force the IO down ourselves.
|
|
|
|
*/
|
2012-03-01 21:56:26 +08:00
|
|
|
void btrfs_add_ordered_operation(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root, struct inode *inode)
|
2009-04-01 01:27:11 +08:00
|
|
|
{
|
2013-02-14 00:09:14 +08:00
|
|
|
struct btrfs_transaction *cur_trans = trans->transaction;
|
2009-04-01 01:27:11 +08:00
|
|
|
u64 last_mod;
|
|
|
|
|
|
|
|
last_mod = max(BTRFS_I(inode)->generation, BTRFS_I(inode)->last_trans);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* if this file hasn't been changed since the last transaction
|
|
|
|
* commit, we can safely return without doing anything
|
|
|
|
*/
|
|
|
|
if (last_mod < root->fs_info->last_trans_committed)
|
2012-03-01 21:56:26 +08:00
|
|
|
return;
|
2009-04-01 01:27:11 +08:00
|
|
|
|
2013-05-15 15:48:23 +08:00
|
|
|
spin_lock(&root->fs_info->ordered_root_lock);
|
2009-04-01 01:27:11 +08:00
|
|
|
if (list_empty(&BTRFS_I(inode)->ordered_operations)) {
|
|
|
|
list_add_tail(&BTRFS_I(inode)->ordered_operations,
|
2013-02-14 00:09:14 +08:00
|
|
|
&cur_trans->ordered_operations);
|
2009-04-01 01:27:11 +08:00
|
|
|
}
|
2013-05-15 15:48:23 +08:00
|
|
|
spin_unlock(&root->fs_info->ordered_root_lock);
|
2009-04-01 01:27:11 +08:00
|
|
|
}
|
2012-09-06 18:01:51 +08:00
|
|
|
|
|
|
|
int __init ordered_data_init(void)
|
|
|
|
{
|
|
|
|
btrfs_ordered_extent_cache = kmem_cache_create("btrfs_ordered_extent",
|
|
|
|
sizeof(struct btrfs_ordered_extent), 0,
|
|
|
|
SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
|
|
|
|
NULL);
|
|
|
|
if (!btrfs_ordered_extent_cache)
|
|
|
|
return -ENOMEM;
|
2012-10-25 17:31:03 +08:00
|
|
|
|
2012-09-06 18:01:51 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
void ordered_data_exit(void)
|
|
|
|
{
|
|
|
|
if (btrfs_ordered_extent_cache)
|
|
|
|
kmem_cache_destroy(btrfs_ordered_extent_cache);
|
|
|
|
}
|