2007-06-12 15:07:21 +02:00
|
|
|
/*
|
|
|
|
* 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.
|
|
|
|
*/
|
|
|
|
|
2007-08-29 15:11:44 +02:00
|
|
|
#include <linux/blkdev.h>
|
2007-03-21 16:12:56 +01:00
|
|
|
#include <linux/module.h>
|
2007-03-22 17:13:20 +01:00
|
|
|
#include <linux/buffer_head.h>
|
2007-03-21 16:12:56 +01:00
|
|
|
#include <linux/fs.h>
|
|
|
|
#include <linux/pagemap.h>
|
|
|
|
#include <linux/highmem.h>
|
|
|
|
#include <linux/time.h>
|
|
|
|
#include <linux/init.h>
|
2009-04-02 22:46:06 +02:00
|
|
|
#include <linux/seq_file.h>
|
2007-03-21 16:12:56 +01:00
|
|
|
#include <linux/string.h>
|
|
|
|
#include <linux/backing-dev.h>
|
2007-08-29 15:11:44 +02:00
|
|
|
#include <linux/mount.h>
|
2007-03-26 22:00:06 +02:00
|
|
|
#include <linux/mpage.h>
|
2007-03-29 17:56:46 +02:00
|
|
|
#include <linux/swap.h>
|
|
|
|
#include <linux/writeback.h>
|
2007-04-20 03:01:03 +02:00
|
|
|
#include <linux/statfs.h>
|
2007-06-08 21:33:54 +02:00
|
|
|
#include <linux/compat.h>
|
2007-08-29 15:11:44 +02:00
|
|
|
#include <linux/parser.h>
|
2007-12-18 02:14:04 +01:00
|
|
|
#include <linux/ctype.h>
|
2007-12-18 22:15:09 +01:00
|
|
|
#include <linux/namei.h>
|
2008-03-24 20:02:04 +01:00
|
|
|
#include <linux/miscdevice.h>
|
2009-01-15 22:51:03 +01:00
|
|
|
#include <linux/magic.h>
|
2008-11-20 16:22:27 +01:00
|
|
|
#include "compat.h"
|
2007-03-21 16:12:56 +01:00
|
|
|
#include "ctree.h"
|
2007-03-22 17:13:20 +01:00
|
|
|
#include "disk-io.h"
|
2007-03-23 15:01:08 +01:00
|
|
|
#include "transaction.h"
|
2007-04-02 16:50:19 +02:00
|
|
|
#include "btrfs_inode.h"
|
2007-04-10 15:27:04 +02:00
|
|
|
#include "ioctl.h"
|
2007-05-24 19:35:57 +02:00
|
|
|
#include "print-tree.h"
|
2007-11-16 17:45:54 +01:00
|
|
|
#include "xattr.h"
|
2008-03-24 20:02:07 +01:00
|
|
|
#include "volumes.h"
|
2008-07-23 18:12:13 +02:00
|
|
|
#include "version.h"
|
2008-07-20 22:31:56 +02:00
|
|
|
#include "export.h"
|
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>
2008-10-29 19:49:59 +01:00
|
|
|
#include "compression.h"
|
2007-03-21 16:12:56 +01:00
|
|
|
|
2009-09-22 02:01:09 +02:00
|
|
|
static const struct super_operations btrfs_super_ops;
|
2007-03-29 17:56:46 +02:00
|
|
|
|
2009-01-06 03:25:51 +01:00
|
|
|
static void btrfs_put_super(struct super_block *sb)
|
2007-04-17 19:26:50 +02:00
|
|
|
{
|
2007-06-12 12:35:45 +02:00
|
|
|
struct btrfs_root *root = btrfs_sb(sb);
|
2007-04-17 19:26:50 +02:00
|
|
|
int ret;
|
|
|
|
|
2007-06-12 12:35:45 +02:00
|
|
|
ret = close_ctree(root);
|
|
|
|
sb->s_fs_info = NULL;
|
2007-03-29 17:56:46 +02:00
|
|
|
}
|
|
|
|
|
2007-08-29 15:11:44 +02:00
|
|
|
enum {
|
2008-06-10 16:40:46 +02:00
|
|
|
Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow,
|
2008-05-13 19:46:40 +02:00
|
|
|
Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
|
2009-06-10 02:28:34 +02:00
|
|
|
Opt_ssd, Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl,
|
2009-10-14 15:24:59 +02:00
|
|
|
Opt_compress, Opt_notreelog, Opt_ratio, Opt_flushoncommit,
|
|
|
|
Opt_discard, Opt_err,
|
2007-08-29 15:11:44 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
static match_table_t tokens = {
|
2008-05-13 19:46:40 +02:00
|
|
|
{Opt_degraded, "degraded"},
|
2007-08-29 15:11:44 +02:00
|
|
|
{Opt_subvol, "subvol=%s"},
|
2008-06-10 16:40:46 +02:00
|
|
|
{Opt_device, "device=%s"},
|
2007-12-14 21:30:32 +01:00
|
|
|
{Opt_nodatasum, "nodatasum"},
|
2007-12-18 02:14:01 +01:00
|
|
|
{Opt_nodatacow, "nodatacow"},
|
2008-01-09 15:23:21 +01:00
|
|
|
{Opt_nobarrier, "nobarrier"},
|
2007-12-18 02:14:04 +01:00
|
|
|
{Opt_max_extent, "max_extent=%s"},
|
2008-01-29 22:03:38 +01:00
|
|
|
{Opt_max_inline, "max_inline=%s"},
|
2008-01-02 16:01:11 +01:00
|
|
|
{Opt_alloc_start, "alloc_start=%s"},
|
2008-06-12 03:47:56 +02:00
|
|
|
{Opt_thread_pool, "thread_pool=%d"},
|
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>
2008-10-29 19:49:59 +01:00
|
|
|
{Opt_compress, "compress"},
|
2008-01-18 16:54:22 +01:00
|
|
|
{Opt_ssd, "ssd"},
|
2009-06-10 02:28:34 +02:00
|
|
|
{Opt_ssd_spread, "ssd_spread"},
|
2009-06-09 22:42:22 +02:00
|
|
|
{Opt_nossd, "nossd"},
|
2008-07-24 18:16:36 +02:00
|
|
|
{Opt_noacl, "noacl"},
|
2009-04-02 22:49:40 +02:00
|
|
|
{Opt_notreelog, "notreelog"},
|
2009-04-02 22:59:01 +02:00
|
|
|
{Opt_flushoncommit, "flushoncommit"},
|
2009-04-21 23:40:57 +02:00
|
|
|
{Opt_ratio, "metadata_ratio=%d"},
|
2009-10-14 15:24:59 +02:00
|
|
|
{Opt_discard, "discard"},
|
2008-07-24 18:16:36 +02:00
|
|
|
{Opt_err, NULL},
|
2007-08-29 15:11:44 +02:00
|
|
|
};
|
|
|
|
|
2007-12-21 22:27:24 +01:00
|
|
|
u64 btrfs_parse_size(char *str)
|
2007-12-18 02:14:04 +01:00
|
|
|
{
|
2007-12-21 22:27:24 +01:00
|
|
|
u64 res;
|
2007-12-18 02:14:04 +01:00
|
|
|
int mult = 1;
|
|
|
|
char *end;
|
|
|
|
char last;
|
|
|
|
|
|
|
|
res = simple_strtoul(str, &end, 10);
|
|
|
|
|
|
|
|
last = end[0];
|
|
|
|
if (isalpha(last)) {
|
|
|
|
last = tolower(last);
|
|
|
|
switch (last) {
|
|
|
|
case 'g':
|
|
|
|
mult *= 1024;
|
|
|
|
case 'm':
|
|
|
|
mult *= 1024;
|
|
|
|
case 'k':
|
|
|
|
mult *= 1024;
|
|
|
|
}
|
|
|
|
res = res * mult;
|
|
|
|
}
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
2008-06-10 16:40:29 +02:00
|
|
|
/*
|
|
|
|
* Regular mount options parser. Everything that is needed only when
|
|
|
|
* reading in a new superblock is parsed here.
|
|
|
|
*/
|
|
|
|
int btrfs_parse_options(struct btrfs_root *root, char *options)
|
2007-08-29 15:11:44 +02:00
|
|
|
{
|
2008-06-10 16:40:29 +02:00
|
|
|
struct btrfs_fs_info *info = root->fs_info;
|
2007-08-29 15:11:44 +02:00
|
|
|
substring_t args[MAX_OPT_ARGS];
|
2008-06-10 16:40:29 +02:00
|
|
|
char *p, *num;
|
2008-06-12 03:47:56 +02:00
|
|
|
int intarg;
|
2007-12-14 21:30:32 +01:00
|
|
|
|
2007-08-29 15:11:44 +02:00
|
|
|
if (!options)
|
2008-06-10 16:40:29 +02:00
|
|
|
return 0;
|
2007-08-29 15:11:44 +02:00
|
|
|
|
2007-12-18 02:14:01 +01:00
|
|
|
/*
|
|
|
|
* strsep changes the string, duplicate it because parse_options
|
|
|
|
* gets called twice
|
|
|
|
*/
|
|
|
|
options = kstrdup(options, GFP_NOFS);
|
|
|
|
if (!options)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
|
2008-06-10 16:40:29 +02:00
|
|
|
while ((p = strsep(&options, ",")) != NULL) {
|
2007-08-29 15:11:44 +02:00
|
|
|
int token;
|
|
|
|
if (!*p)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
token = match_token(p, tokens, args);
|
|
|
|
switch (token) {
|
2008-05-13 19:46:40 +02:00
|
|
|
case Opt_degraded:
|
2008-06-10 16:40:29 +02:00
|
|
|
printk(KERN_INFO "btrfs: allowing degraded mounts\n");
|
|
|
|
btrfs_set_opt(info->mount_opt, DEGRADED);
|
2008-05-13 19:46:40 +02:00
|
|
|
break;
|
2007-08-29 15:11:44 +02:00
|
|
|
case Opt_subvol:
|
2008-06-10 16:40:46 +02:00
|
|
|
case Opt_device:
|
2008-06-10 16:40:29 +02:00
|
|
|
/*
|
2008-06-10 16:40:46 +02:00
|
|
|
* These are parsed by btrfs_parse_early_options
|
2008-06-10 16:40:29 +02:00
|
|
|
* and can be happily ignored here.
|
|
|
|
*/
|
2007-12-14 21:30:32 +01:00
|
|
|
break;
|
|
|
|
case Opt_nodatasum:
|
2009-06-11 15:30:13 +02:00
|
|
|
printk(KERN_INFO "btrfs: setting nodatasum\n");
|
2008-06-10 16:40:29 +02:00
|
|
|
btrfs_set_opt(info->mount_opt, NODATASUM);
|
2007-12-18 02:14:01 +01:00
|
|
|
break;
|
|
|
|
case Opt_nodatacow:
|
2008-06-10 16:40:29 +02:00
|
|
|
printk(KERN_INFO "btrfs: setting nodatacow\n");
|
|
|
|
btrfs_set_opt(info->mount_opt, NODATACOW);
|
|
|
|
btrfs_set_opt(info->mount_opt, NODATASUM);
|
2007-08-29 15:11:44 +02:00
|
|
|
break;
|
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>
2008-10-29 19:49:59 +01:00
|
|
|
case Opt_compress:
|
|
|
|
printk(KERN_INFO "btrfs: use compression\n");
|
|
|
|
btrfs_set_opt(info->mount_opt, COMPRESS);
|
|
|
|
break;
|
2008-01-18 16:54:22 +01:00
|
|
|
case Opt_ssd:
|
2008-06-10 16:40:29 +02:00
|
|
|
printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
|
|
|
|
btrfs_set_opt(info->mount_opt, SSD);
|
2008-01-18 16:54:22 +01:00
|
|
|
break;
|
2009-06-10 02:28:34 +02:00
|
|
|
case Opt_ssd_spread:
|
|
|
|
printk(KERN_INFO "btrfs: use spread ssd "
|
|
|
|
"allocation scheme\n");
|
|
|
|
btrfs_set_opt(info->mount_opt, SSD);
|
|
|
|
btrfs_set_opt(info->mount_opt, SSD_SPREAD);
|
|
|
|
break;
|
2009-06-09 22:42:22 +02:00
|
|
|
case Opt_nossd:
|
2009-06-10 02:28:34 +02:00
|
|
|
printk(KERN_INFO "btrfs: not using ssd allocation "
|
|
|
|
"scheme\n");
|
2009-06-10 15:51:32 +02:00
|
|
|
btrfs_set_opt(info->mount_opt, NOSSD);
|
2009-06-09 22:42:22 +02:00
|
|
|
btrfs_clear_opt(info->mount_opt, SSD);
|
2009-06-10 02:28:34 +02:00
|
|
|
btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
|
2009-06-09 22:42:22 +02:00
|
|
|
break;
|
2008-01-09 15:23:21 +01:00
|
|
|
case Opt_nobarrier:
|
2008-06-10 16:40:29 +02:00
|
|
|
printk(KERN_INFO "btrfs: turning off barriers\n");
|
|
|
|
btrfs_set_opt(info->mount_opt, NOBARRIER);
|
2008-01-09 15:23:21 +01:00
|
|
|
break;
|
2008-06-12 03:47:56 +02:00
|
|
|
case Opt_thread_pool:
|
|
|
|
intarg = 0;
|
|
|
|
match_int(&args[0], &intarg);
|
|
|
|
if (intarg) {
|
|
|
|
info->thread_pool_size = intarg;
|
|
|
|
printk(KERN_INFO "btrfs: thread pool %d\n",
|
|
|
|
info->thread_pool_size);
|
|
|
|
}
|
|
|
|
break;
|
2007-12-18 02:14:04 +01:00
|
|
|
case Opt_max_extent:
|
2008-06-10 16:40:29 +02:00
|
|
|
num = match_strdup(&args[0]);
|
|
|
|
if (num) {
|
|
|
|
info->max_extent = btrfs_parse_size(num);
|
|
|
|
kfree(num);
|
|
|
|
|
|
|
|
info->max_extent = max_t(u64,
|
|
|
|
info->max_extent, root->sectorsize);
|
|
|
|
printk(KERN_INFO "btrfs: max_extent at %llu\n",
|
2009-04-21 21:38:29 +02:00
|
|
|
(unsigned long long)info->max_extent);
|
2007-12-18 02:14:04 +01:00
|
|
|
}
|
|
|
|
break;
|
2008-01-29 22:03:38 +01:00
|
|
|
case Opt_max_inline:
|
2008-06-10 16:40:29 +02:00
|
|
|
num = match_strdup(&args[0]);
|
|
|
|
if (num) {
|
|
|
|
info->max_inline = btrfs_parse_size(num);
|
|
|
|
kfree(num);
|
|
|
|
|
2008-06-11 22:51:38 +02:00
|
|
|
if (info->max_inline) {
|
|
|
|
info->max_inline = max_t(u64,
|
|
|
|
info->max_inline,
|
|
|
|
root->sectorsize);
|
|
|
|
}
|
2008-06-10 16:40:29 +02:00
|
|
|
printk(KERN_INFO "btrfs: max_inline at %llu\n",
|
2009-04-21 21:38:29 +02:00
|
|
|
(unsigned long long)info->max_inline);
|
2008-01-29 22:03:38 +01:00
|
|
|
}
|
|
|
|
break;
|
2008-01-02 16:01:11 +01:00
|
|
|
case Opt_alloc_start:
|
2008-06-10 16:40:29 +02:00
|
|
|
num = match_strdup(&args[0]);
|
|
|
|
if (num) {
|
|
|
|
info->alloc_start = btrfs_parse_size(num);
|
|
|
|
kfree(num);
|
|
|
|
printk(KERN_INFO
|
|
|
|
"btrfs: allocations start at %llu\n",
|
2009-04-21 21:38:29 +02:00
|
|
|
(unsigned long long)info->alloc_start);
|
2008-01-02 16:01:11 +01:00
|
|
|
}
|
|
|
|
break;
|
2008-07-24 18:16:36 +02:00
|
|
|
case Opt_noacl:
|
|
|
|
root->fs_info->sb->s_flags &= ~MS_POSIXACL;
|
|
|
|
break;
|
2009-04-02 22:49:40 +02:00
|
|
|
case Opt_notreelog:
|
|
|
|
printk(KERN_INFO "btrfs: disabling tree log\n");
|
|
|
|
btrfs_set_opt(info->mount_opt, NOTREELOG);
|
|
|
|
break;
|
2009-04-02 22:59:01 +02:00
|
|
|
case Opt_flushoncommit:
|
|
|
|
printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
|
|
|
|
btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
|
|
|
|
break;
|
2009-04-21 23:40:57 +02:00
|
|
|
case Opt_ratio:
|
|
|
|
intarg = 0;
|
|
|
|
match_int(&args[0], &intarg);
|
|
|
|
if (intarg) {
|
|
|
|
info->metadata_ratio = intarg;
|
|
|
|
printk(KERN_INFO "btrfs: metadata ratio %d\n",
|
|
|
|
info->metadata_ratio);
|
|
|
|
}
|
|
|
|
break;
|
2009-10-14 15:24:59 +02:00
|
|
|
case Opt_discard:
|
|
|
|
btrfs_set_opt(info->mount_opt, DISCARD);
|
|
|
|
break;
|
2007-08-29 15:11:44 +02:00
|
|
|
default:
|
2007-12-18 02:14:01 +01:00
|
|
|
break;
|
2007-08-29 15:11:44 +02:00
|
|
|
}
|
|
|
|
}
|
2007-12-18 02:14:01 +01:00
|
|
|
kfree(options);
|
2008-06-10 16:40:29 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Parse mount options that are required early in the mount process.
|
|
|
|
*
|
|
|
|
* All other options will be parsed on much later in the mount process and
|
|
|
|
* only when we need to allocate a new super block.
|
|
|
|
*/
|
2008-12-02 12:36:09 +01:00
|
|
|
static int btrfs_parse_early_options(const char *options, fmode_t flags,
|
2008-06-10 16:40:46 +02:00
|
|
|
void *holder, char **subvol_name,
|
|
|
|
struct btrfs_fs_devices **fs_devices)
|
2008-06-10 16:40:29 +02:00
|
|
|
{
|
|
|
|
substring_t args[MAX_OPT_ARGS];
|
|
|
|
char *opts, *p;
|
|
|
|
int error = 0;
|
|
|
|
|
|
|
|
if (!options)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* strsep changes the string, duplicate it because parse_options
|
|
|
|
* gets called twice
|
|
|
|
*/
|
|
|
|
opts = kstrdup(options, GFP_KERNEL);
|
|
|
|
if (!opts)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
while ((p = strsep(&opts, ",")) != NULL) {
|
|
|
|
int token;
|
|
|
|
if (!*p)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
token = match_token(p, tokens, args);
|
|
|
|
switch (token) {
|
|
|
|
case Opt_subvol:
|
|
|
|
*subvol_name = match_strdup(&args[0]);
|
|
|
|
break;
|
2008-06-10 16:40:46 +02:00
|
|
|
case Opt_device:
|
|
|
|
error = btrfs_scan_one_device(match_strdup(&args[0]),
|
|
|
|
flags, holder, fs_devices);
|
|
|
|
if (error)
|
|
|
|
goto out_free_opts;
|
|
|
|
break;
|
2008-06-10 16:40:29 +02:00
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-06-10 16:40:46 +02:00
|
|
|
out_free_opts:
|
2008-06-10 16:40:29 +02:00
|
|
|
kfree(opts);
|
|
|
|
out:
|
|
|
|
/*
|
|
|
|
* If no subvolume name is specified we use the default one. Allocate
|
2008-11-18 03:02:50 +01:00
|
|
|
* a copy of the string "." here so that code later in the
|
2008-06-10 16:40:29 +02:00
|
|
|
* mount path doesn't care if it's the default volume or another one.
|
|
|
|
*/
|
|
|
|
if (!*subvol_name) {
|
2008-11-18 03:02:50 +01:00
|
|
|
*subvol_name = kstrdup(".", GFP_KERNEL);
|
2008-06-10 16:40:29 +02:00
|
|
|
if (!*subvol_name)
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
return error;
|
2007-08-29 15:11:44 +02:00
|
|
|
}
|
|
|
|
|
2009-01-06 03:25:51 +01:00
|
|
|
static int btrfs_fill_super(struct super_block *sb,
|
2008-03-24 20:02:07 +01:00
|
|
|
struct btrfs_fs_devices *fs_devices,
|
2009-01-06 03:25:51 +01:00
|
|
|
void *data, int silent)
|
2007-03-29 17:56:46 +02:00
|
|
|
{
|
2009-01-06 03:25:51 +01:00
|
|
|
struct inode *inode;
|
|
|
|
struct dentry *root_dentry;
|
2007-06-12 12:35:45 +02:00
|
|
|
struct btrfs_super_block *disk_super;
|
|
|
|
struct btrfs_root *tree_root;
|
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE)
This commit introduces a new kind of back reference for btrfs metadata.
Once a filesystem has been mounted with this commit, IT WILL NO LONGER
BE MOUNTABLE BY OLDER KERNELS.
When a tree block in subvolume tree is cow'd, the reference counts of all
extents it points to are increased by one. At transaction commit time,
the old root of the subvolume is recorded in a "dead root" data structure,
and the btree it points to is later walked, dropping reference counts
and freeing any blocks where the reference count goes to 0.
The increments done during cow and decrements done after commit cancel out,
and the walk is a very expensive way to go about freeing the blocks that
are no longer referenced by the new btree root. This commit reduces the
transaction overhead by avoiding the need for dead root records.
When a non-shared tree block is cow'd, we free the old block at once, and the
new block inherits old block's references. When a tree block with reference
count > 1 is cow'd, we increase the reference counts of all extents
the new block points to by one, and decrease the old block's reference count by
one.
This dead tree avoidance code removes the need to modify the reference
counts of lower level extents when a non-shared tree block is cow'd.
But we still need to update back ref for all pointers in the block.
This is because the location of the block is recorded in the back ref
item.
We can solve this by introducing a new type of back ref. The new
back ref provides information about pointer's key, level and in which
tree the pointer lives. This information allow us to find the pointer
by searching the tree. The shortcoming of the new back ref is that it
only works for pointers in tree blocks referenced by their owner trees.
This is mostly a problem for snapshots, where resolving one of these
fuzzy back references would be O(number_of_snapshots) and quite slow.
The solution used here is to use the fuzzy back references in the common
case where a given tree block is only referenced by one root,
and use the full back references when multiple roots have a reference
on a given block.
This commit adds per subvolume red-black tree to keep trace of cached
inodes. The red-black tree helps the balancing code to find cached
inodes whose inode numbers within a given range.
This commit improves the balancing code by introducing several data
structures to keep the state of balancing. The most important one
is the back ref cache. It caches how the upper level tree blocks are
referenced. This greatly reduce the overhead of checking back ref.
The improved balancing code scales significantly better with a large
number of snapshots.
This is a very large commit and was written in a number of
pieces. But, they depend heavily on the disk format change and were
squashed together to make sure git bisect didn't end up in a
bad state wrt space balancing or the format change.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 16:45:14 +02:00
|
|
|
struct btrfs_key key;
|
2007-06-12 12:35:45 +02:00
|
|
|
int err;
|
2007-04-18 22:15:28 +02:00
|
|
|
|
2007-06-12 12:35:45 +02:00
|
|
|
sb->s_maxbytes = MAX_LFS_FILESIZE;
|
|
|
|
sb->s_magic = BTRFS_SUPER_MAGIC;
|
|
|
|
sb->s_op = &btrfs_super_ops;
|
2008-07-20 22:31:56 +02:00
|
|
|
sb->s_export_op = &btrfs_export_ops;
|
2007-11-16 17:45:54 +01:00
|
|
|
sb->s_xattr = btrfs_xattr_handlers;
|
2007-06-12 12:35:45 +02:00
|
|
|
sb->s_time_gran = 1;
|
2009-10-13 19:50:18 +02:00
|
|
|
#ifdef CONFIG_BTRFS_FS_POSIX_ACL
|
2008-07-24 18:16:36 +02:00
|
|
|
sb->s_flags |= MS_POSIXACL;
|
2009-09-29 19:51:04 +02:00
|
|
|
#endif
|
2007-04-18 22:15:28 +02:00
|
|
|
|
2008-05-13 19:46:40 +02:00
|
|
|
tree_root = open_ctree(sb, fs_devices, (char *)data);
|
2007-04-16 15:22:45 +02:00
|
|
|
|
2008-04-01 17:21:34 +02:00
|
|
|
if (IS_ERR(tree_root)) {
|
2007-06-12 12:35:45 +02:00
|
|
|
printk("btrfs: open_ctree failed\n");
|
2008-04-01 17:21:34 +02:00
|
|
|
return PTR_ERR(tree_root);
|
2007-04-18 22:15:28 +02:00
|
|
|
}
|
2007-06-12 12:35:45 +02:00
|
|
|
sb->s_fs_info = tree_root;
|
2007-10-15 22:14:19 +02:00
|
|
|
disk_super = &tree_root->fs_info->super_copy;
|
2007-04-18 22:15:28 +02:00
|
|
|
|
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE)
This commit introduces a new kind of back reference for btrfs metadata.
Once a filesystem has been mounted with this commit, IT WILL NO LONGER
BE MOUNTABLE BY OLDER KERNELS.
When a tree block in subvolume tree is cow'd, the reference counts of all
extents it points to are increased by one. At transaction commit time,
the old root of the subvolume is recorded in a "dead root" data structure,
and the btree it points to is later walked, dropping reference counts
and freeing any blocks where the reference count goes to 0.
The increments done during cow and decrements done after commit cancel out,
and the walk is a very expensive way to go about freeing the blocks that
are no longer referenced by the new btree root. This commit reduces the
transaction overhead by avoiding the need for dead root records.
When a non-shared tree block is cow'd, we free the old block at once, and the
new block inherits old block's references. When a tree block with reference
count > 1 is cow'd, we increase the reference counts of all extents
the new block points to by one, and decrease the old block's reference count by
one.
This dead tree avoidance code removes the need to modify the reference
counts of lower level extents when a non-shared tree block is cow'd.
But we still need to update back ref for all pointers in the block.
This is because the location of the block is recorded in the back ref
item.
We can solve this by introducing a new type of back ref. The new
back ref provides information about pointer's key, level and in which
tree the pointer lives. This information allow us to find the pointer
by searching the tree. The shortcoming of the new back ref is that it
only works for pointers in tree blocks referenced by their owner trees.
This is mostly a problem for snapshots, where resolving one of these
fuzzy back references would be O(number_of_snapshots) and quite slow.
The solution used here is to use the fuzzy back references in the common
case where a given tree block is only referenced by one root,
and use the full back references when multiple roots have a reference
on a given block.
This commit adds per subvolume red-black tree to keep trace of cached
inodes. The red-black tree helps the balancing code to find cached
inodes whose inode numbers within a given range.
This commit improves the balancing code by introducing several data
structures to keep the state of balancing. The most important one
is the back ref cache. It caches how the upper level tree blocks are
referenced. This greatly reduce the overhead of checking back ref.
The improved balancing code scales significantly better with a large
number of snapshots.
This is a very large commit and was written in a number of
pieces. But, they depend heavily on the disk format change and were
squashed together to make sure git bisect didn't end up in a
bad state wrt space balancing or the format change.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 16:45:14 +02:00
|
|
|
key.objectid = BTRFS_FIRST_FREE_OBJECTID;
|
|
|
|
key.type = BTRFS_INODE_ITEM_KEY;
|
|
|
|
key.offset = 0;
|
|
|
|
inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root);
|
|
|
|
if (IS_ERR(inode)) {
|
|
|
|
err = PTR_ERR(inode);
|
2007-06-12 12:35:45 +02:00
|
|
|
goto fail_close;
|
2007-03-29 21:15:27 +02:00
|
|
|
}
|
|
|
|
|
2007-06-12 12:35:45 +02:00
|
|
|
root_dentry = d_alloc_root(inode);
|
|
|
|
if (!root_dentry) {
|
|
|
|
iput(inode);
|
|
|
|
err = -ENOMEM;
|
|
|
|
goto fail_close;
|
2007-03-29 21:15:27 +02:00
|
|
|
}
|
2008-12-12 16:03:26 +01:00
|
|
|
#if 0
|
2007-08-29 21:47:34 +02:00
|
|
|
/* this does the super kobj at the same time */
|
|
|
|
err = btrfs_sysfs_add_super(tree_root->fs_info);
|
|
|
|
if (err)
|
|
|
|
goto fail_close;
|
2008-12-12 16:03:26 +01:00
|
|
|
#endif
|
2007-08-29 21:47:34 +02:00
|
|
|
|
2007-06-12 12:35:45 +02:00
|
|
|
sb->s_root = root_dentry;
|
2008-02-20 22:11:05 +01:00
|
|
|
|
|
|
|
save_mount_options(sb, data);
|
2007-04-10 22:58:11 +02:00
|
|
|
return 0;
|
2007-06-12 12:35:45 +02:00
|
|
|
|
|
|
|
fail_close:
|
|
|
|
close_ctree(tree_root);
|
|
|
|
return err;
|
2007-04-10 22:58:11 +02:00
|
|
|
}
|
|
|
|
|
2008-06-10 16:07:39 +02:00
|
|
|
int btrfs_sync_fs(struct super_block *sb, int wait)
|
2007-04-10 15:27:04 +02:00
|
|
|
{
|
|
|
|
struct btrfs_trans_handle *trans;
|
2009-04-02 22:59:01 +02:00
|
|
|
struct btrfs_root *root = btrfs_sb(sb);
|
2007-04-10 15:27:04 +02:00
|
|
|
int ret;
|
2007-04-10 22:58:11 +02:00
|
|
|
|
2007-06-12 12:35:45 +02:00
|
|
|
if (!wait) {
|
|
|
|
filemap_flush(root->fs_info->btree_inode->i_mapping);
|
|
|
|
return 0;
|
|
|
|
}
|
2008-11-07 04:02:51 +01:00
|
|
|
|
2009-11-12 10:36:34 +01:00
|
|
|
btrfs_start_delalloc_inodes(root, 0);
|
|
|
|
btrfs_wait_ordered_extents(root, 0, 0);
|
2008-11-07 04:02:51 +01:00
|
|
|
|
2007-04-10 15:27:04 +02:00
|
|
|
trans = btrfs_start_transaction(root, 1);
|
|
|
|
ret = btrfs_commit_transaction(trans, root);
|
2007-06-22 20:16:25 +02:00
|
|
|
return ret;
|
2007-04-02 16:50:19 +02:00
|
|
|
}
|
|
|
|
|
2009-04-02 22:46:06 +02:00
|
|
|
static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
|
|
|
|
{
|
|
|
|
struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
|
|
|
|
struct btrfs_fs_info *info = root->fs_info;
|
|
|
|
|
|
|
|
if (btrfs_test_opt(root, DEGRADED))
|
|
|
|
seq_puts(seq, ",degraded");
|
|
|
|
if (btrfs_test_opt(root, NODATASUM))
|
|
|
|
seq_puts(seq, ",nodatasum");
|
|
|
|
if (btrfs_test_opt(root, NODATACOW))
|
|
|
|
seq_puts(seq, ",nodatacow");
|
|
|
|
if (btrfs_test_opt(root, NOBARRIER))
|
|
|
|
seq_puts(seq, ",nobarrier");
|
|
|
|
if (info->max_extent != (u64)-1)
|
2009-04-21 21:38:29 +02:00
|
|
|
seq_printf(seq, ",max_extent=%llu",
|
|
|
|
(unsigned long long)info->max_extent);
|
2009-04-02 22:46:06 +02:00
|
|
|
if (info->max_inline != 8192 * 1024)
|
2009-04-21 21:38:29 +02:00
|
|
|
seq_printf(seq, ",max_inline=%llu",
|
|
|
|
(unsigned long long)info->max_inline);
|
2009-04-02 22:46:06 +02:00
|
|
|
if (info->alloc_start != 0)
|
2009-04-21 21:38:29 +02:00
|
|
|
seq_printf(seq, ",alloc_start=%llu",
|
|
|
|
(unsigned long long)info->alloc_start);
|
2009-04-02 22:46:06 +02:00
|
|
|
if (info->thread_pool_size != min_t(unsigned long,
|
|
|
|
num_online_cpus() + 2, 8))
|
|
|
|
seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
|
|
|
|
if (btrfs_test_opt(root, COMPRESS))
|
|
|
|
seq_puts(seq, ",compress");
|
2009-06-10 15:51:32 +02:00
|
|
|
if (btrfs_test_opt(root, NOSSD))
|
|
|
|
seq_puts(seq, ",nossd");
|
2009-06-10 02:28:34 +02:00
|
|
|
if (btrfs_test_opt(root, SSD_SPREAD))
|
|
|
|
seq_puts(seq, ",ssd_spread");
|
|
|
|
else if (btrfs_test_opt(root, SSD))
|
2009-04-02 22:46:06 +02:00
|
|
|
seq_puts(seq, ",ssd");
|
2009-04-02 22:49:40 +02:00
|
|
|
if (btrfs_test_opt(root, NOTREELOG))
|
2009-05-14 19:52:21 +02:00
|
|
|
seq_puts(seq, ",notreelog");
|
2009-04-02 22:59:01 +02:00
|
|
|
if (btrfs_test_opt(root, FLUSHONCOMMIT))
|
2009-05-14 19:52:21 +02:00
|
|
|
seq_puts(seq, ",flushoncommit");
|
2009-04-02 22:46:06 +02:00
|
|
|
if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
|
|
|
|
seq_puts(seq, ",noacl");
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-05-07 17:43:44 +02:00
|
|
|
static int btrfs_test_super(struct super_block *s, void *data)
|
2007-08-29 15:11:44 +02:00
|
|
|
{
|
2008-05-07 17:43:44 +02:00
|
|
|
struct btrfs_fs_devices *test_fs_devices = data;
|
|
|
|
struct btrfs_root *root = btrfs_sb(s);
|
2007-08-29 15:11:44 +02:00
|
|
|
|
2008-05-07 17:43:44 +02:00
|
|
|
return root->fs_info->fs_devices == test_fs_devices;
|
2007-08-29 15:11:44 +02:00
|
|
|
}
|
|
|
|
|
2008-06-10 16:40:29 +02:00
|
|
|
/*
|
|
|
|
* Find a superblock for the given device / mount point.
|
|
|
|
*
|
|
|
|
* Note: This is based on get_sb_bdev from fs/super.c with a few additions
|
|
|
|
* for multiple device setup. Make sure to keep it in sync.
|
|
|
|
*/
|
|
|
|
static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
|
|
|
|
const char *dev_name, void *data, struct vfsmount *mnt)
|
2007-08-29 15:11:44 +02:00
|
|
|
{
|
2008-06-10 16:40:29 +02:00
|
|
|
char *subvol_name = NULL;
|
2007-08-29 15:11:44 +02:00
|
|
|
struct block_device *bdev = NULL;
|
|
|
|
struct super_block *s;
|
|
|
|
struct dentry *root;
|
2008-03-24 20:02:07 +01:00
|
|
|
struct btrfs_fs_devices *fs_devices = NULL;
|
2008-12-02 12:36:09 +01:00
|
|
|
fmode_t mode = FMODE_READ;
|
2007-08-29 15:11:44 +02:00
|
|
|
int error = 0;
|
|
|
|
|
2008-12-02 12:36:09 +01:00
|
|
|
if (!(flags & MS_RDONLY))
|
|
|
|
mode |= FMODE_WRITE;
|
|
|
|
|
|
|
|
error = btrfs_parse_early_options(data, mode, fs_type,
|
2008-06-10 16:40:46 +02:00
|
|
|
&subvol_name, &fs_devices);
|
2008-06-10 16:40:29 +02:00
|
|
|
if (error)
|
2009-01-05 21:43:42 +01:00
|
|
|
return error;
|
2008-06-10 16:40:29 +02:00
|
|
|
|
2008-12-02 12:36:09 +01:00
|
|
|
error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
|
2008-03-24 20:02:07 +01:00
|
|
|
if (error)
|
2008-06-10 16:40:29 +02:00
|
|
|
goto error_free_subvol_name;
|
2007-08-29 15:11:44 +02:00
|
|
|
|
2008-12-02 12:36:09 +01:00
|
|
|
error = btrfs_open_devices(fs_devices, mode, fs_type);
|
2008-03-24 20:02:07 +01:00
|
|
|
if (error)
|
2008-06-10 16:40:29 +02:00
|
|
|
goto error_free_subvol_name;
|
2008-03-24 20:02:07 +01:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
|
|
|
|
error = -EACCES;
|
|
|
|
goto error_close_devices;
|
|
|
|
}
|
|
|
|
|
2008-05-13 19:46:40 +02:00
|
|
|
bdev = fs_devices->latest_bdev;
|
2008-05-07 17:43:44 +02:00
|
|
|
s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
|
2007-08-29 15:11:44 +02:00
|
|
|
if (IS_ERR(s))
|
|
|
|
goto error_s;
|
|
|
|
|
|
|
|
if (s->s_root) {
|
|
|
|
if ((flags ^ s->s_flags) & MS_RDONLY) {
|
2009-05-06 07:34:22 +02:00
|
|
|
deactivate_locked_super(s);
|
2007-08-29 15:11:44 +02:00
|
|
|
error = -EBUSY;
|
2008-11-12 20:34:12 +01:00
|
|
|
goto error_close_devices;
|
2007-08-29 15:11:44 +02:00
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
btrfs_close_devices(fs_devices);
|
2007-08-29 15:11:44 +02:00
|
|
|
} else {
|
|
|
|
char b[BDEVNAME_SIZE];
|
|
|
|
|
|
|
|
s->s_flags = flags;
|
|
|
|
strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
|
2008-03-24 20:02:07 +01:00
|
|
|
error = btrfs_fill_super(s, fs_devices, data,
|
|
|
|
flags & MS_SILENT ? 1 : 0);
|
2007-08-29 15:11:44 +02:00
|
|
|
if (error) {
|
2009-05-06 07:34:22 +02:00
|
|
|
deactivate_locked_super(s);
|
2009-01-05 21:43:42 +01:00
|
|
|
goto error_free_subvol_name;
|
2007-08-29 15:11:44 +02:00
|
|
|
}
|
|
|
|
|
2008-04-28 21:29:42 +02:00
|
|
|
btrfs_sb(s)->fs_info->bdev_holder = fs_type;
|
2007-08-29 15:11:44 +02:00
|
|
|
s->s_flags |= MS_ACTIVE;
|
|
|
|
}
|
|
|
|
|
2008-08-19 17:49:35 +02:00
|
|
|
if (!strcmp(subvol_name, "."))
|
|
|
|
root = dget(s->s_root);
|
|
|
|
else {
|
|
|
|
mutex_lock(&s->s_root->d_inode->i_mutex);
|
2009-01-06 03:25:51 +01:00
|
|
|
root = lookup_one_len(subvol_name, s->s_root,
|
|
|
|
strlen(subvol_name));
|
2008-08-19 17:49:35 +02:00
|
|
|
mutex_unlock(&s->s_root->d_inode->i_mutex);
|
2009-01-06 03:25:51 +01:00
|
|
|
|
2008-08-19 17:49:35 +02:00
|
|
|
if (IS_ERR(root)) {
|
2009-05-06 07:34:22 +02:00
|
|
|
deactivate_locked_super(s);
|
2008-08-19 17:49:35 +02:00
|
|
|
error = PTR_ERR(root);
|
2009-01-05 21:43:42 +01:00
|
|
|
goto error_free_subvol_name;
|
2008-08-19 17:49:35 +02:00
|
|
|
}
|
|
|
|
if (!root->d_inode) {
|
|
|
|
dput(root);
|
2009-05-06 07:34:22 +02:00
|
|
|
deactivate_locked_super(s);
|
2008-08-19 17:49:35 +02:00
|
|
|
error = -ENXIO;
|
2009-01-05 21:43:42 +01:00
|
|
|
goto error_free_subvol_name;
|
2008-08-19 17:49:35 +02:00
|
|
|
}
|
2007-08-29 15:11:44 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
mnt->mnt_sb = s;
|
|
|
|
mnt->mnt_root = root;
|
2008-06-10 16:40:29 +02:00
|
|
|
|
|
|
|
kfree(subvol_name);
|
2007-08-29 15:11:44 +02:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
error_s:
|
|
|
|
error = PTR_ERR(s);
|
2008-11-12 20:34:12 +01:00
|
|
|
error_close_devices:
|
2008-03-24 20:02:07 +01:00
|
|
|
btrfs_close_devices(fs_devices);
|
2008-06-10 16:40:29 +02:00
|
|
|
error_free_subvol_name:
|
|
|
|
kfree(subvol_name);
|
2007-08-29 15:11:44 +02:00
|
|
|
return error;
|
|
|
|
}
|
2007-03-21 16:12:56 +01:00
|
|
|
|
2008-11-12 20:34:12 +01:00
|
|
|
static int btrfs_remount(struct super_block *sb, int *flags, char *data)
|
|
|
|
{
|
|
|
|
struct btrfs_root *root = btrfs_sb(sb);
|
|
|
|
int ret;
|
|
|
|
|
2009-02-12 15:37:35 +01:00
|
|
|
ret = btrfs_parse_options(root, data);
|
|
|
|
if (ret)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2008-11-12 20:34:12 +01:00
|
|
|
if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (*flags & MS_RDONLY) {
|
|
|
|
sb->s_flags |= MS_RDONLY;
|
|
|
|
|
|
|
|
ret = btrfs_commit_super(root);
|
|
|
|
WARN_ON(ret);
|
|
|
|
} else {
|
2008-11-18 03:11:30 +01:00
|
|
|
if (root->fs_info->fs_devices->rw_devices == 0)
|
|
|
|
return -EACCES;
|
|
|
|
|
2008-11-12 20:34:12 +01:00
|
|
|
if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
|
|
|
|
return -EINVAL;
|
|
|
|
|
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE)
This commit introduces a new kind of back reference for btrfs metadata.
Once a filesystem has been mounted with this commit, IT WILL NO LONGER
BE MOUNTABLE BY OLDER KERNELS.
When a tree block in subvolume tree is cow'd, the reference counts of all
extents it points to are increased by one. At transaction commit time,
the old root of the subvolume is recorded in a "dead root" data structure,
and the btree it points to is later walked, dropping reference counts
and freeing any blocks where the reference count goes to 0.
The increments done during cow and decrements done after commit cancel out,
and the walk is a very expensive way to go about freeing the blocks that
are no longer referenced by the new btree root. This commit reduces the
transaction overhead by avoiding the need for dead root records.
When a non-shared tree block is cow'd, we free the old block at once, and the
new block inherits old block's references. When a tree block with reference
count > 1 is cow'd, we increase the reference counts of all extents
the new block points to by one, and decrease the old block's reference count by
one.
This dead tree avoidance code removes the need to modify the reference
counts of lower level extents when a non-shared tree block is cow'd.
But we still need to update back ref for all pointers in the block.
This is because the location of the block is recorded in the back ref
item.
We can solve this by introducing a new type of back ref. The new
back ref provides information about pointer's key, level and in which
tree the pointer lives. This information allow us to find the pointer
by searching the tree. The shortcoming of the new back ref is that it
only works for pointers in tree blocks referenced by their owner trees.
This is mostly a problem for snapshots, where resolving one of these
fuzzy back references would be O(number_of_snapshots) and quite slow.
The solution used here is to use the fuzzy back references in the common
case where a given tree block is only referenced by one root,
and use the full back references when multiple roots have a reference
on a given block.
This commit adds per subvolume red-black tree to keep trace of cached
inodes. The red-black tree helps the balancing code to find cached
inodes whose inode numbers within a given range.
This commit improves the balancing code by introducing several data
structures to keep the state of balancing. The most important one
is the back ref cache. It caches how the upper level tree blocks are
referenced. This greatly reduce the overhead of checking back ref.
The improved balancing code scales significantly better with a large
number of snapshots.
This is a very large commit and was written in a number of
pieces. But, they depend heavily on the disk format change and were
squashed together to make sure git bisect didn't end up in a
bad state wrt space balancing or the format change.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 16:45:14 +02:00
|
|
|
/* recover relocation */
|
|
|
|
ret = btrfs_recover_relocation(root);
|
2008-11-12 20:34:12 +01:00
|
|
|
WARN_ON(ret);
|
|
|
|
|
|
|
|
ret = btrfs_cleanup_fs_roots(root->fs_info);
|
|
|
|
WARN_ON(ret);
|
|
|
|
|
|
|
|
sb->s_flags &= ~MS_RDONLY;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2007-04-20 03:01:03 +02:00
|
|
|
static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
|
|
|
|
{
|
|
|
|
struct btrfs_root *root = btrfs_sb(dentry->d_sb);
|
2007-06-26 16:06:50 +02:00
|
|
|
struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
|
2007-10-15 22:15:53 +02:00
|
|
|
int bits = dentry->d_sb->s_blocksize_bits;
|
2008-08-18 13:01:52 +02:00
|
|
|
__be32 *fsid = (__be32 *)root->fs_info->fsid;
|
2007-04-20 03:01:03 +02:00
|
|
|
|
|
|
|
buf->f_namelen = BTRFS_NAME_LEN;
|
2007-10-15 22:15:53 +02:00
|
|
|
buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
|
|
|
|
buf->f_bfree = buf->f_blocks -
|
|
|
|
(btrfs_super_bytes_used(disk_super) >> bits);
|
2007-04-20 03:01:03 +02:00
|
|
|
buf->f_bavail = buf->f_bfree;
|
|
|
|
buf->f_bsize = dentry->d_sb->s_blocksize;
|
|
|
|
buf->f_type = BTRFS_SUPER_MAGIC;
|
2009-01-06 03:25:51 +01:00
|
|
|
|
2008-08-18 13:01:52 +02:00
|
|
|
/* We treat it as constant endianness (it doesn't matter _which_)
|
2009-01-06 03:25:51 +01:00
|
|
|
because we want the fsid to come out the same whether mounted
|
2008-08-18 13:01:52 +02:00
|
|
|
on a big-endian or little-endian host */
|
|
|
|
buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
|
|
|
|
buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
|
2008-08-18 14:10:20 +02:00
|
|
|
/* Mask in the root object ID too, to disambiguate subvols */
|
|
|
|
buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
|
|
|
|
buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
|
|
|
|
|
2007-04-20 03:01:03 +02:00
|
|
|
return 0;
|
|
|
|
}
|
2007-04-24 17:52:22 +02:00
|
|
|
|
2007-03-21 16:12:56 +01:00
|
|
|
static struct file_system_type btrfs_fs_type = {
|
|
|
|
.owner = THIS_MODULE,
|
|
|
|
.name = "btrfs",
|
|
|
|
.get_sb = btrfs_get_sb,
|
2008-05-07 17:43:44 +02:00
|
|
|
.kill_sb = kill_anon_super,
|
2007-03-21 16:12:56 +01:00
|
|
|
.fs_flags = FS_REQUIRES_DEV,
|
|
|
|
};
|
2008-03-24 20:02:04 +01:00
|
|
|
|
2008-09-29 21:18:18 +02:00
|
|
|
/*
|
|
|
|
* used by btrfsctl to scan devices when no FS is mounted
|
|
|
|
*/
|
2008-03-24 20:02:07 +01:00
|
|
|
static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
|
|
|
|
unsigned long arg)
|
|
|
|
{
|
|
|
|
struct btrfs_ioctl_vol_args *vol;
|
|
|
|
struct btrfs_fs_devices *fs_devices;
|
2009-01-16 17:59:08 +01:00
|
|
|
int ret = -ENOTTY;
|
2008-03-24 20:02:07 +01:00
|
|
|
|
2009-01-05 22:57:23 +01:00
|
|
|
if (!capable(CAP_SYS_ADMIN))
|
|
|
|
return -EPERM;
|
|
|
|
|
2009-04-08 09:06:54 +02:00
|
|
|
vol = memdup_user((void __user *)arg, sizeof(*vol));
|
|
|
|
if (IS_ERR(vol))
|
|
|
|
return PTR_ERR(vol);
|
2009-01-16 17:59:08 +01:00
|
|
|
|
2008-03-24 20:02:07 +01:00
|
|
|
switch (cmd) {
|
|
|
|
case BTRFS_IOC_SCAN_DEV:
|
2008-12-02 12:36:09 +01:00
|
|
|
ret = btrfs_scan_one_device(vol->name, FMODE_READ,
|
2008-03-24 20:02:07 +01:00
|
|
|
&btrfs_fs_type, &fs_devices);
|
|
|
|
break;
|
|
|
|
}
|
2009-04-08 09:06:54 +02:00
|
|
|
|
2008-03-24 20:02:07 +01:00
|
|
|
kfree(vol);
|
2008-06-10 04:17:11 +02:00
|
|
|
return ret;
|
2008-03-24 20:02:07 +01:00
|
|
|
}
|
|
|
|
|
2009-01-10 15:09:52 +01:00
|
|
|
static int btrfs_freeze(struct super_block *sb)
|
2008-01-22 18:46:56 +01:00
|
|
|
{
|
|
|
|
struct btrfs_root *root = btrfs_sb(sb);
|
2008-06-25 22:01:31 +02:00
|
|
|
mutex_lock(&root->fs_info->transaction_kthread_mutex);
|
|
|
|
mutex_lock(&root->fs_info->cleaner_mutex);
|
2009-01-10 15:09:52 +01:00
|
|
|
return 0;
|
2008-01-22 18:46:56 +01:00
|
|
|
}
|
|
|
|
|
2009-01-10 15:09:52 +01:00
|
|
|
static int btrfs_unfreeze(struct super_block *sb)
|
2008-01-22 18:46:56 +01:00
|
|
|
{
|
|
|
|
struct btrfs_root *root = btrfs_sb(sb);
|
2008-06-25 22:01:31 +02:00
|
|
|
mutex_unlock(&root->fs_info->cleaner_mutex);
|
|
|
|
mutex_unlock(&root->fs_info->transaction_kthread_mutex);
|
2009-01-10 15:09:52 +01:00
|
|
|
return 0;
|
2008-01-22 18:46:56 +01:00
|
|
|
}
|
2007-03-21 16:12:56 +01:00
|
|
|
|
2009-09-22 02:01:09 +02:00
|
|
|
static const struct super_operations btrfs_super_ops = {
|
2009-09-21 22:00:26 +02:00
|
|
|
.drop_inode = btrfs_drop_inode,
|
2007-03-25 19:44:56 +02:00
|
|
|
.delete_inode = btrfs_delete_inode,
|
2007-03-22 17:13:20 +01:00
|
|
|
.put_super = btrfs_put_super,
|
2007-03-23 15:01:08 +01:00
|
|
|
.sync_fs = btrfs_sync_fs,
|
2009-04-02 22:46:06 +02:00
|
|
|
.show_options = btrfs_show_options,
|
2007-03-26 18:00:39 +02:00
|
|
|
.write_inode = btrfs_write_inode,
|
2007-04-24 17:52:22 +02:00
|
|
|
.dirty_inode = btrfs_dirty_inode,
|
2007-04-02 16:50:19 +02:00
|
|
|
.alloc_inode = btrfs_alloc_inode,
|
|
|
|
.destroy_inode = btrfs_destroy_inode,
|
2007-04-20 03:01:03 +02:00
|
|
|
.statfs = btrfs_statfs,
|
2008-11-12 20:34:12 +01:00
|
|
|
.remount_fs = btrfs_remount,
|
2009-01-10 15:09:52 +01:00
|
|
|
.freeze_fs = btrfs_freeze,
|
|
|
|
.unfreeze_fs = btrfs_unfreeze,
|
2007-03-22 17:13:20 +01:00
|
|
|
};
|
2008-03-24 20:02:04 +01:00
|
|
|
|
|
|
|
static const struct file_operations btrfs_ctl_fops = {
|
|
|
|
.unlocked_ioctl = btrfs_control_ioctl,
|
|
|
|
.compat_ioctl = btrfs_control_ioctl,
|
|
|
|
.owner = THIS_MODULE,
|
|
|
|
};
|
|
|
|
|
|
|
|
static struct miscdevice btrfs_misc = {
|
|
|
|
.minor = MISC_DYNAMIC_MINOR,
|
|
|
|
.name = "btrfs-control",
|
|
|
|
.fops = &btrfs_ctl_fops
|
|
|
|
};
|
|
|
|
|
|
|
|
static int btrfs_interface_init(void)
|
|
|
|
{
|
|
|
|
return misc_register(&btrfs_misc);
|
|
|
|
}
|
|
|
|
|
2008-12-02 15:54:17 +01:00
|
|
|
static void btrfs_interface_exit(void)
|
2008-03-24 20:02:04 +01:00
|
|
|
{
|
|
|
|
if (misc_deregister(&btrfs_misc) < 0)
|
2009-01-06 03:25:51 +01:00
|
|
|
printk(KERN_INFO "misc_deregister failed for control device");
|
2008-03-24 20:02:04 +01:00
|
|
|
}
|
|
|
|
|
2007-03-21 16:12:56 +01:00
|
|
|
static int __init init_btrfs_fs(void)
|
|
|
|
{
|
2007-04-02 16:50:19 +02:00
|
|
|
int err;
|
2007-08-29 21:47:34 +02:00
|
|
|
|
|
|
|
err = btrfs_init_sysfs();
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
2007-06-12 12:35:45 +02:00
|
|
|
err = btrfs_init_cachep();
|
2007-04-02 16:50:19 +02:00
|
|
|
if (err)
|
2008-06-25 22:01:31 +02:00
|
|
|
goto free_sysfs;
|
2008-01-24 22:13:08 +01:00
|
|
|
|
|
|
|
err = extent_io_init();
|
2007-11-19 16:22:33 +01:00
|
|
|
if (err)
|
|
|
|
goto free_cachep;
|
|
|
|
|
2008-01-24 22:13:08 +01:00
|
|
|
err = extent_map_init();
|
|
|
|
if (err)
|
|
|
|
goto free_extent_io;
|
|
|
|
|
2008-03-24 20:02:04 +01:00
|
|
|
err = btrfs_interface_init();
|
2007-11-19 16:22:33 +01:00
|
|
|
if (err)
|
|
|
|
goto free_extent_map;
|
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>
2008-10-29 19:49:59 +01:00
|
|
|
|
2008-03-24 20:02:04 +01:00
|
|
|
err = register_filesystem(&btrfs_fs_type);
|
|
|
|
if (err)
|
|
|
|
goto unregister_ioctl;
|
2008-07-23 18:12:13 +02:00
|
|
|
|
|
|
|
printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
|
2007-11-19 16:22:33 +01:00
|
|
|
return 0;
|
|
|
|
|
2008-03-24 20:02:04 +01:00
|
|
|
unregister_ioctl:
|
|
|
|
btrfs_interface_exit();
|
2007-11-19 16:22:33 +01:00
|
|
|
free_extent_map:
|
|
|
|
extent_map_exit();
|
2008-01-24 22:13:08 +01:00
|
|
|
free_extent_io:
|
|
|
|
extent_io_exit();
|
2007-11-19 16:22:33 +01:00
|
|
|
free_cachep:
|
|
|
|
btrfs_destroy_cachep();
|
2008-06-25 22:01:31 +02:00
|
|
|
free_sysfs:
|
2007-11-19 16:22:33 +01:00
|
|
|
btrfs_exit_sysfs();
|
|
|
|
return err;
|
2007-03-21 16:12:56 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static void __exit exit_btrfs_fs(void)
|
|
|
|
{
|
2007-06-12 12:35:45 +02:00
|
|
|
btrfs_destroy_cachep();
|
2007-08-27 22:49:44 +02:00
|
|
|
extent_map_exit();
|
2008-01-24 22:13:08 +01:00
|
|
|
extent_io_exit();
|
2008-03-24 20:02:04 +01:00
|
|
|
btrfs_interface_exit();
|
2007-03-21 16:12:56 +01:00
|
|
|
unregister_filesystem(&btrfs_fs_type);
|
2007-08-29 21:47:34 +02:00
|
|
|
btrfs_exit_sysfs();
|
2008-03-24 20:02:07 +01:00
|
|
|
btrfs_cleanup_fs_uuids();
|
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>
2008-10-29 19:49:59 +01:00
|
|
|
btrfs_zlib_exit();
|
2007-03-21 16:12:56 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
module_init(init_btrfs_fs)
|
|
|
|
module_exit(exit_btrfs_fs)
|
|
|
|
|
|
|
|
MODULE_LICENSE("GPL");
|