2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/blkdev.h>
20 #include <linux/module.h>
21 #include <linux/buffer_head.h>
23 #include <linux/pagemap.h>
24 #include <linux/highmem.h>
25 #include <linux/time.h>
26 #include <linux/init.h>
27 #include <linux/seq_file.h>
28 #include <linux/string.h>
29 #include <linux/backing-dev.h>
30 #include <linux/mount.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/parser.h>
37 #include <linux/ctype.h>
38 #include <linux/namei.h>
39 #include <linux/miscdevice.h>
40 #include <linux/magic.h>
41 #include <linux/slab.h>
42 #include <linux/cleancache.h>
43 #include <linux/ratelimit.h>
44 #include <linux/btrfs.h>
45 #include "delayed-inode.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "print-tree.h"
54 #include "compression.h"
55 #include "rcu-string.h"
56 #include "dev-replace.h"
57 #include "free-space-cache.h"
59 #include "tests/btrfs-tests.h"
61 #define CREATE_TRACE_POINTS
62 #include <trace/events/btrfs.h>
64 static const struct super_operations btrfs_super_ops
;
65 static struct file_system_type btrfs_fs_type
;
67 static const char *btrfs_decode_error(int errno
)
69 char *errstr
= "unknown";
73 errstr
= "IO failure";
76 errstr
= "Out of memory";
79 errstr
= "Readonly filesystem";
82 errstr
= "Object already exists";
85 errstr
= "No space left";
88 errstr
= "No such entry";
95 static void save_error_info(struct btrfs_fs_info
*fs_info
)
98 * today we only save the error info into ram. Long term we'll
99 * also send it down to the disk
101 set_bit(BTRFS_FS_STATE_ERROR
, &fs_info
->fs_state
);
104 /* btrfs handle error by forcing the filesystem readonly */
105 static void btrfs_handle_error(struct btrfs_fs_info
*fs_info
)
107 struct super_block
*sb
= fs_info
->sb
;
109 if (sb
->s_flags
& MS_RDONLY
)
112 if (test_bit(BTRFS_FS_STATE_ERROR
, &fs_info
->fs_state
)) {
113 sb
->s_flags
|= MS_RDONLY
;
114 btrfs_info(fs_info
, "forced readonly");
116 * Note that a running device replace operation is not
117 * canceled here although there is no way to update
118 * the progress. It would add the risk of a deadlock,
119 * therefore the canceling is ommited. The only penalty
120 * is that some I/O remains active until the procedure
121 * completes. The next time when the filesystem is
122 * mounted writeable again, the device replace
123 * operation continues.
130 * __btrfs_std_error decodes expected errors from the caller and
131 * invokes the approciate error response.
133 void __btrfs_std_error(struct btrfs_fs_info
*fs_info
, const char *function
,
134 unsigned int line
, int errno
, const char *fmt
, ...)
136 struct super_block
*sb
= fs_info
->sb
;
140 * Special case: if the error is EROFS, and we're already
141 * under MS_RDONLY, then it is safe here.
143 if (errno
== -EROFS
&& (sb
->s_flags
& MS_RDONLY
))
146 errstr
= btrfs_decode_error(errno
);
148 struct va_format vaf
;
155 printk(KERN_CRIT
"BTRFS error (device %s) in %s:%d: errno=%d %s (%pV)\n",
156 sb
->s_id
, function
, line
, errno
, errstr
, &vaf
);
159 printk(KERN_CRIT
"BTRFS error (device %s) in %s:%d: errno=%d %s\n",
160 sb
->s_id
, function
, line
, errno
, errstr
);
163 /* Don't go through full error handling during mount */
164 save_error_info(fs_info
);
165 if (sb
->s_flags
& MS_BORN
)
166 btrfs_handle_error(fs_info
);
169 static const char * const logtypes
[] = {
180 void btrfs_printk(const struct btrfs_fs_info
*fs_info
, const char *fmt
, ...)
182 struct super_block
*sb
= fs_info
->sb
;
184 struct va_format vaf
;
186 const char *type
= logtypes
[4];
191 kern_level
= printk_get_level(fmt
);
193 size_t size
= printk_skip_level(fmt
) - fmt
;
194 memcpy(lvl
, fmt
, size
);
197 type
= logtypes
[kern_level
- '0'];
204 printk("%sBTRFS %s (device %s): %pV\n", lvl
, type
, sb
->s_id
, &vaf
);
211 void __btrfs_std_error(struct btrfs_fs_info
*fs_info
, const char *function
,
212 unsigned int line
, int errno
, const char *fmt
, ...)
214 struct super_block
*sb
= fs_info
->sb
;
217 * Special case: if the error is EROFS, and we're already
218 * under MS_RDONLY, then it is safe here.
220 if (errno
== -EROFS
&& (sb
->s_flags
& MS_RDONLY
))
223 /* Don't go through full error handling during mount */
224 if (sb
->s_flags
& MS_BORN
) {
225 save_error_info(fs_info
);
226 btrfs_handle_error(fs_info
);
232 * We only mark the transaction aborted and then set the file system read-only.
233 * This will prevent new transactions from starting or trying to join this
236 * This means that error recovery at the call site is limited to freeing
237 * any local memory allocations and passing the error code up without
238 * further cleanup. The transaction should complete as it normally would
239 * in the call path but will return -EIO.
241 * We'll complete the cleanup in btrfs_end_transaction and
242 * btrfs_commit_transaction.
244 void __btrfs_abort_transaction(struct btrfs_trans_handle
*trans
,
245 struct btrfs_root
*root
, const char *function
,
246 unsigned int line
, int errno
)
249 * Report first abort since mount
251 if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED
,
252 &root
->fs_info
->fs_state
)) {
253 WARN(1, KERN_DEBUG
"btrfs: Transaction aborted (error %d)\n",
256 trans
->aborted
= errno
;
257 /* Nothing used. The other threads that have joined this
258 * transaction may be able to continue. */
259 if (!trans
->blocks_used
) {
262 errstr
= btrfs_decode_error(errno
);
263 btrfs_warn(root
->fs_info
,
264 "%s:%d: Aborting unused transaction(%s).",
265 function
, line
, errstr
);
268 ACCESS_ONCE(trans
->transaction
->aborted
) = errno
;
269 /* Wake up anybody who may be waiting on this transaction */
270 wake_up(&root
->fs_info
->transaction_wait
);
271 wake_up(&root
->fs_info
->transaction_blocked_wait
);
272 __btrfs_std_error(root
->fs_info
, function
, line
, errno
, NULL
);
275 * __btrfs_panic decodes unexpected, fatal errors from the caller,
276 * issues an alert, and either panics or BUGs, depending on mount options.
278 void __btrfs_panic(struct btrfs_fs_info
*fs_info
, const char *function
,
279 unsigned int line
, int errno
, const char *fmt
, ...)
281 char *s_id
= "<unknown>";
283 struct va_format vaf
= { .fmt
= fmt
};
287 s_id
= fs_info
->sb
->s_id
;
292 errstr
= btrfs_decode_error(errno
);
293 if (fs_info
&& (fs_info
->mount_opt
& BTRFS_MOUNT_PANIC_ON_FATAL_ERROR
))
294 panic(KERN_CRIT
"BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
295 s_id
, function
, line
, &vaf
, errno
, errstr
);
297 printk(KERN_CRIT
"BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
298 s_id
, function
, line
, &vaf
, errno
, errstr
);
300 /* Caller calls BUG() */
303 static void btrfs_put_super(struct super_block
*sb
)
305 (void)close_ctree(btrfs_sb(sb
)->tree_root
);
306 /* FIXME: need to fix VFS to return error? */
307 /* AV: return it _where_? ->put_super() can be triggered by any number
308 * of async events, up to and including delivery of SIGKILL to the
309 * last process that kept it busy. Or segfault in the aforementioned
310 * process... Whom would you report that to?
315 Opt_degraded
, Opt_subvol
, Opt_subvolid
, Opt_device
, Opt_nodatasum
,
316 Opt_nodatacow
, Opt_max_inline
, Opt_alloc_start
, Opt_nobarrier
, Opt_ssd
,
317 Opt_nossd
, Opt_ssd_spread
, Opt_thread_pool
, Opt_noacl
, Opt_compress
,
318 Opt_compress_type
, Opt_compress_force
, Opt_compress_force_type
,
319 Opt_notreelog
, Opt_ratio
, Opt_flushoncommit
, Opt_discard
,
320 Opt_space_cache
, Opt_clear_cache
, Opt_user_subvol_rm_allowed
,
321 Opt_enospc_debug
, Opt_subvolrootid
, Opt_defrag
, Opt_inode_cache
,
322 Opt_no_space_cache
, Opt_recovery
, Opt_skip_balance
,
323 Opt_check_integrity
, Opt_check_integrity_including_extent_data
,
324 Opt_check_integrity_print_mask
, Opt_fatal_errors
, Opt_rescan_uuid_tree
,
329 static match_table_t tokens
= {
330 {Opt_degraded
, "degraded"},
331 {Opt_subvol
, "subvol=%s"},
332 {Opt_subvolid
, "subvolid=%s"},
333 {Opt_device
, "device=%s"},
334 {Opt_nodatasum
, "nodatasum"},
335 {Opt_nodatacow
, "nodatacow"},
336 {Opt_nobarrier
, "nobarrier"},
337 {Opt_max_inline
, "max_inline=%s"},
338 {Opt_alloc_start
, "alloc_start=%s"},
339 {Opt_thread_pool
, "thread_pool=%d"},
340 {Opt_compress
, "compress"},
341 {Opt_compress_type
, "compress=%s"},
342 {Opt_compress_force
, "compress-force"},
343 {Opt_compress_force_type
, "compress-force=%s"},
345 {Opt_ssd_spread
, "ssd_spread"},
346 {Opt_nossd
, "nossd"},
347 {Opt_noacl
, "noacl"},
348 {Opt_notreelog
, "notreelog"},
349 {Opt_flushoncommit
, "flushoncommit"},
350 {Opt_ratio
, "metadata_ratio=%d"},
351 {Opt_discard
, "discard"},
352 {Opt_space_cache
, "space_cache"},
353 {Opt_clear_cache
, "clear_cache"},
354 {Opt_user_subvol_rm_allowed
, "user_subvol_rm_allowed"},
355 {Opt_enospc_debug
, "enospc_debug"},
356 {Opt_subvolrootid
, "subvolrootid=%d"},
357 {Opt_defrag
, "autodefrag"},
358 {Opt_inode_cache
, "inode_cache"},
359 {Opt_no_space_cache
, "nospace_cache"},
360 {Opt_recovery
, "recovery"},
361 {Opt_skip_balance
, "skip_balance"},
362 {Opt_check_integrity
, "check_int"},
363 {Opt_check_integrity_including_extent_data
, "check_int_data"},
364 {Opt_check_integrity_print_mask
, "check_int_print_mask=%d"},
365 {Opt_rescan_uuid_tree
, "rescan_uuid_tree"},
366 {Opt_fatal_errors
, "fatal_errors=%s"},
367 {Opt_commit_interval
, "commit=%d"},
372 * Regular mount options parser. Everything that is needed only when
373 * reading in a new superblock is parsed here.
374 * XXX JDM: This needs to be cleaned up for remount.
376 int btrfs_parse_options(struct btrfs_root
*root
, char *options
)
378 struct btrfs_fs_info
*info
= root
->fs_info
;
379 substring_t args
[MAX_OPT_ARGS
];
380 char *p
, *num
, *orig
= NULL
;
385 bool compress_force
= false;
387 cache_gen
= btrfs_super_cache_generation(root
->fs_info
->super_copy
);
389 btrfs_set_opt(info
->mount_opt
, SPACE_CACHE
);
395 * strsep changes the string, duplicate it because parse_options
398 options
= kstrdup(options
, GFP_NOFS
);
404 while ((p
= strsep(&options
, ",")) != NULL
) {
409 token
= match_token(p
, tokens
, args
);
412 printk(KERN_INFO
"btrfs: allowing degraded mounts\n");
413 btrfs_set_opt(info
->mount_opt
, DEGRADED
);
417 case Opt_subvolrootid
:
420 * These are parsed by btrfs_parse_early_options
421 * and can be happily ignored here.
425 printk(KERN_INFO
"btrfs: setting nodatasum\n");
426 btrfs_set_opt(info
->mount_opt
, NODATASUM
);
429 if (!btrfs_test_opt(root
, COMPRESS
) ||
430 !btrfs_test_opt(root
, FORCE_COMPRESS
)) {
431 printk(KERN_INFO
"btrfs: setting nodatacow, compression disabled\n");
433 printk(KERN_INFO
"btrfs: setting nodatacow\n");
435 info
->compress_type
= BTRFS_COMPRESS_NONE
;
436 btrfs_clear_opt(info
->mount_opt
, COMPRESS
);
437 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
438 btrfs_set_opt(info
->mount_opt
, NODATACOW
);
439 btrfs_set_opt(info
->mount_opt
, NODATASUM
);
441 case Opt_compress_force
:
442 case Opt_compress_force_type
:
443 compress_force
= true;
446 case Opt_compress_type
:
447 if (token
== Opt_compress
||
448 token
== Opt_compress_force
||
449 strcmp(args
[0].from
, "zlib") == 0) {
450 compress_type
= "zlib";
451 info
->compress_type
= BTRFS_COMPRESS_ZLIB
;
452 btrfs_set_opt(info
->mount_opt
, COMPRESS
);
453 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
454 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
455 } else if (strcmp(args
[0].from
, "lzo") == 0) {
456 compress_type
= "lzo";
457 info
->compress_type
= BTRFS_COMPRESS_LZO
;
458 btrfs_set_opt(info
->mount_opt
, COMPRESS
);
459 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
460 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
461 btrfs_set_fs_incompat(info
, COMPRESS_LZO
);
462 } else if (strncmp(args
[0].from
, "no", 2) == 0) {
463 compress_type
= "no";
464 info
->compress_type
= BTRFS_COMPRESS_NONE
;
465 btrfs_clear_opt(info
->mount_opt
, COMPRESS
);
466 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
467 compress_force
= false;
473 if (compress_force
) {
474 btrfs_set_opt(info
->mount_opt
, FORCE_COMPRESS
);
475 pr_info("btrfs: force %s compression\n",
478 pr_info("btrfs: use %s compression\n",
482 printk(KERN_INFO
"btrfs: use ssd allocation scheme\n");
483 btrfs_set_opt(info
->mount_opt
, SSD
);
486 printk(KERN_INFO
"btrfs: use spread ssd "
487 "allocation scheme\n");
488 btrfs_set_opt(info
->mount_opt
, SSD
);
489 btrfs_set_opt(info
->mount_opt
, SSD_SPREAD
);
492 printk(KERN_INFO
"btrfs: not using ssd allocation "
494 btrfs_set_opt(info
->mount_opt
, NOSSD
);
495 btrfs_clear_opt(info
->mount_opt
, SSD
);
496 btrfs_clear_opt(info
->mount_opt
, SSD_SPREAD
);
499 printk(KERN_INFO
"btrfs: turning off barriers\n");
500 btrfs_set_opt(info
->mount_opt
, NOBARRIER
);
502 case Opt_thread_pool
:
503 ret
= match_int(&args
[0], &intarg
);
506 } else if (intarg
> 0) {
507 info
->thread_pool_size
= intarg
;
514 num
= match_strdup(&args
[0]);
516 info
->max_inline
= memparse(num
, NULL
);
519 if (info
->max_inline
) {
520 info
->max_inline
= max_t(u64
,
524 printk(KERN_INFO
"btrfs: max_inline at %llu\n",
531 case Opt_alloc_start
:
532 num
= match_strdup(&args
[0]);
534 mutex_lock(&info
->chunk_mutex
);
535 info
->alloc_start
= memparse(num
, NULL
);
536 mutex_unlock(&info
->chunk_mutex
);
539 "btrfs: allocations start at %llu\n",
547 root
->fs_info
->sb
->s_flags
&= ~MS_POSIXACL
;
550 printk(KERN_INFO
"btrfs: disabling tree log\n");
551 btrfs_set_opt(info
->mount_opt
, NOTREELOG
);
553 case Opt_flushoncommit
:
554 printk(KERN_INFO
"btrfs: turning on flush-on-commit\n");
555 btrfs_set_opt(info
->mount_opt
, FLUSHONCOMMIT
);
558 ret
= match_int(&args
[0], &intarg
);
561 } else if (intarg
>= 0) {
562 info
->metadata_ratio
= intarg
;
563 printk(KERN_INFO
"btrfs: metadata ratio %d\n",
564 info
->metadata_ratio
);
571 btrfs_set_opt(info
->mount_opt
, DISCARD
);
573 case Opt_space_cache
:
574 btrfs_set_opt(info
->mount_opt
, SPACE_CACHE
);
576 case Opt_rescan_uuid_tree
:
577 btrfs_set_opt(info
->mount_opt
, RESCAN_UUID_TREE
);
579 case Opt_no_space_cache
:
580 printk(KERN_INFO
"btrfs: disabling disk space caching\n");
581 btrfs_clear_opt(info
->mount_opt
, SPACE_CACHE
);
583 case Opt_inode_cache
:
584 printk(KERN_INFO
"btrfs: enabling inode map caching\n");
585 btrfs_set_opt(info
->mount_opt
, INODE_MAP_CACHE
);
587 case Opt_clear_cache
:
588 printk(KERN_INFO
"btrfs: force clearing of disk cache\n");
589 btrfs_set_opt(info
->mount_opt
, CLEAR_CACHE
);
591 case Opt_user_subvol_rm_allowed
:
592 btrfs_set_opt(info
->mount_opt
, USER_SUBVOL_RM_ALLOWED
);
594 case Opt_enospc_debug
:
595 btrfs_set_opt(info
->mount_opt
, ENOSPC_DEBUG
);
598 printk(KERN_INFO
"btrfs: enabling auto defrag\n");
599 btrfs_set_opt(info
->mount_opt
, AUTO_DEFRAG
);
602 printk(KERN_INFO
"btrfs: enabling auto recovery\n");
603 btrfs_set_opt(info
->mount_opt
, RECOVERY
);
605 case Opt_skip_balance
:
606 btrfs_set_opt(info
->mount_opt
, SKIP_BALANCE
);
608 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
609 case Opt_check_integrity_including_extent_data
:
610 printk(KERN_INFO
"btrfs: enabling check integrity"
611 " including extent data\n");
612 btrfs_set_opt(info
->mount_opt
,
613 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
);
614 btrfs_set_opt(info
->mount_opt
, CHECK_INTEGRITY
);
616 case Opt_check_integrity
:
617 printk(KERN_INFO
"btrfs: enabling check integrity\n");
618 btrfs_set_opt(info
->mount_opt
, CHECK_INTEGRITY
);
620 case Opt_check_integrity_print_mask
:
621 ret
= match_int(&args
[0], &intarg
);
624 } else if (intarg
>= 0) {
625 info
->check_integrity_print_mask
= intarg
;
626 printk(KERN_INFO
"btrfs:"
627 " check_integrity_print_mask 0x%x\n",
628 info
->check_integrity_print_mask
);
635 case Opt_check_integrity_including_extent_data
:
636 case Opt_check_integrity
:
637 case Opt_check_integrity_print_mask
:
638 printk(KERN_ERR
"btrfs: support for check_integrity*"
639 " not compiled in!\n");
643 case Opt_fatal_errors
:
644 if (strcmp(args
[0].from
, "panic") == 0)
645 btrfs_set_opt(info
->mount_opt
,
646 PANIC_ON_FATAL_ERROR
);
647 else if (strcmp(args
[0].from
, "bug") == 0)
648 btrfs_clear_opt(info
->mount_opt
,
649 PANIC_ON_FATAL_ERROR
);
655 case Opt_commit_interval
:
657 ret
= match_int(&args
[0], &intarg
);
660 "btrfs: invalid commit interval\n");
667 "btrfs: excessive commit interval %d\n",
670 info
->commit_interval
= intarg
;
673 "btrfs: using default commit interval %ds\n",
674 BTRFS_DEFAULT_COMMIT_INTERVAL
);
675 info
->commit_interval
= BTRFS_DEFAULT_COMMIT_INTERVAL
;
679 printk(KERN_INFO
"btrfs: unrecognized mount option "
688 if (!ret
&& btrfs_test_opt(root
, SPACE_CACHE
))
689 printk(KERN_INFO
"btrfs: disk space caching is enabled\n");
695 * Parse mount options that are required early in the mount process.
697 * All other options will be parsed on much later in the mount process and
698 * only when we need to allocate a new super block.
700 static int btrfs_parse_early_options(const char *options
, fmode_t flags
,
701 void *holder
, char **subvol_name
, u64
*subvol_objectid
,
702 struct btrfs_fs_devices
**fs_devices
)
704 substring_t args
[MAX_OPT_ARGS
];
705 char *device_name
, *opts
, *orig
, *p
;
713 * strsep changes the string, duplicate it because parse_options
716 opts
= kstrdup(options
, GFP_KERNEL
);
721 while ((p
= strsep(&opts
, ",")) != NULL
) {
726 token
= match_token(p
, tokens
, args
);
730 *subvol_name
= match_strdup(&args
[0]);
737 num
= match_strdup(&args
[0]);
739 *subvol_objectid
= memparse(num
, NULL
);
741 /* we want the original fs_tree */
742 if (!*subvol_objectid
)
744 BTRFS_FS_TREE_OBJECTID
;
750 case Opt_subvolrootid
:
752 "btrfs: 'subvolrootid' mount option is deprecated and has no effect\n");
755 device_name
= match_strdup(&args
[0]);
760 error
= btrfs_scan_one_device(device_name
,
761 flags
, holder
, fs_devices
);
776 static struct dentry
*get_default_root(struct super_block
*sb
,
779 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
780 struct btrfs_root
*root
= fs_info
->tree_root
;
781 struct btrfs_root
*new_root
;
782 struct btrfs_dir_item
*di
;
783 struct btrfs_path
*path
;
784 struct btrfs_key location
;
790 * We have a specific subvol we want to mount, just setup location and
791 * go look up the root.
793 if (subvol_objectid
) {
794 location
.objectid
= subvol_objectid
;
795 location
.type
= BTRFS_ROOT_ITEM_KEY
;
796 location
.offset
= (u64
)-1;
800 path
= btrfs_alloc_path();
802 return ERR_PTR(-ENOMEM
);
803 path
->leave_spinning
= 1;
806 * Find the "default" dir item which points to the root item that we
807 * will mount by default if we haven't been given a specific subvolume
810 dir_id
= btrfs_super_root_dir(fs_info
->super_copy
);
811 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir_id
, "default", 7, 0);
813 btrfs_free_path(path
);
818 * Ok the default dir item isn't there. This is weird since
819 * it's always been there, but don't freak out, just try and
820 * mount to root most subvolume.
822 btrfs_free_path(path
);
823 dir_id
= BTRFS_FIRST_FREE_OBJECTID
;
824 new_root
= fs_info
->fs_root
;
828 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &location
);
829 btrfs_free_path(path
);
832 new_root
= btrfs_read_fs_root_no_name(fs_info
, &location
);
833 if (IS_ERR(new_root
))
834 return ERR_CAST(new_root
);
836 dir_id
= btrfs_root_dirid(&new_root
->root_item
);
838 location
.objectid
= dir_id
;
839 location
.type
= BTRFS_INODE_ITEM_KEY
;
842 inode
= btrfs_iget(sb
, &location
, new_root
, &new);
844 return ERR_CAST(inode
);
847 * If we're just mounting the root most subvol put the inode and return
848 * a reference to the dentry. We will have already gotten a reference
849 * to the inode in btrfs_fill_super so we're good to go.
851 if (!new && sb
->s_root
->d_inode
== inode
) {
853 return dget(sb
->s_root
);
856 return d_obtain_alias(inode
);
859 static int btrfs_fill_super(struct super_block
*sb
,
860 struct btrfs_fs_devices
*fs_devices
,
861 void *data
, int silent
)
864 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
865 struct btrfs_key key
;
868 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
869 sb
->s_magic
= BTRFS_SUPER_MAGIC
;
870 sb
->s_op
= &btrfs_super_ops
;
871 sb
->s_d_op
= &btrfs_dentry_operations
;
872 sb
->s_export_op
= &btrfs_export_ops
;
873 sb
->s_xattr
= btrfs_xattr_handlers
;
875 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
876 sb
->s_flags
|= MS_POSIXACL
;
878 sb
->s_flags
|= MS_I_VERSION
;
879 err
= open_ctree(sb
, fs_devices
, (char *)data
);
881 printk("btrfs: open_ctree failed\n");
885 key
.objectid
= BTRFS_FIRST_FREE_OBJECTID
;
886 key
.type
= BTRFS_INODE_ITEM_KEY
;
888 inode
= btrfs_iget(sb
, &key
, fs_info
->fs_root
, NULL
);
890 err
= PTR_ERR(inode
);
894 sb
->s_root
= d_make_root(inode
);
900 save_mount_options(sb
, data
);
901 cleancache_init_fs(sb
);
902 sb
->s_flags
|= MS_ACTIVE
;
906 close_ctree(fs_info
->tree_root
);
910 int btrfs_sync_fs(struct super_block
*sb
, int wait
)
912 struct btrfs_trans_handle
*trans
;
913 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
914 struct btrfs_root
*root
= fs_info
->tree_root
;
916 trace_btrfs_sync_fs(wait
);
919 filemap_flush(fs_info
->btree_inode
->i_mapping
);
923 btrfs_wait_all_ordered_extents(fs_info
);
925 trans
= btrfs_attach_transaction_barrier(root
);
927 /* no transaction, don't bother */
928 if (PTR_ERR(trans
) == -ENOENT
)
930 return PTR_ERR(trans
);
932 return btrfs_commit_transaction(trans
, root
);
935 static int btrfs_show_options(struct seq_file
*seq
, struct dentry
*dentry
)
937 struct btrfs_fs_info
*info
= btrfs_sb(dentry
->d_sb
);
938 struct btrfs_root
*root
= info
->tree_root
;
941 if (btrfs_test_opt(root
, DEGRADED
))
942 seq_puts(seq
, ",degraded");
943 if (btrfs_test_opt(root
, NODATASUM
))
944 seq_puts(seq
, ",nodatasum");
945 if (btrfs_test_opt(root
, NODATACOW
))
946 seq_puts(seq
, ",nodatacow");
947 if (btrfs_test_opt(root
, NOBARRIER
))
948 seq_puts(seq
, ",nobarrier");
949 if (info
->max_inline
!= 8192 * 1024)
950 seq_printf(seq
, ",max_inline=%llu", info
->max_inline
);
951 if (info
->alloc_start
!= 0)
952 seq_printf(seq
, ",alloc_start=%llu", info
->alloc_start
);
953 if (info
->thread_pool_size
!= min_t(unsigned long,
954 num_online_cpus() + 2, 8))
955 seq_printf(seq
, ",thread_pool=%d", info
->thread_pool_size
);
956 if (btrfs_test_opt(root
, COMPRESS
)) {
957 if (info
->compress_type
== BTRFS_COMPRESS_ZLIB
)
958 compress_type
= "zlib";
960 compress_type
= "lzo";
961 if (btrfs_test_opt(root
, FORCE_COMPRESS
))
962 seq_printf(seq
, ",compress-force=%s", compress_type
);
964 seq_printf(seq
, ",compress=%s", compress_type
);
966 if (btrfs_test_opt(root
, NOSSD
))
967 seq_puts(seq
, ",nossd");
968 if (btrfs_test_opt(root
, SSD_SPREAD
))
969 seq_puts(seq
, ",ssd_spread");
970 else if (btrfs_test_opt(root
, SSD
))
971 seq_puts(seq
, ",ssd");
972 if (btrfs_test_opt(root
, NOTREELOG
))
973 seq_puts(seq
, ",notreelog");
974 if (btrfs_test_opt(root
, FLUSHONCOMMIT
))
975 seq_puts(seq
, ",flushoncommit");
976 if (btrfs_test_opt(root
, DISCARD
))
977 seq_puts(seq
, ",discard");
978 if (!(root
->fs_info
->sb
->s_flags
& MS_POSIXACL
))
979 seq_puts(seq
, ",noacl");
980 if (btrfs_test_opt(root
, SPACE_CACHE
))
981 seq_puts(seq
, ",space_cache");
983 seq_puts(seq
, ",nospace_cache");
984 if (btrfs_test_opt(root
, RESCAN_UUID_TREE
))
985 seq_puts(seq
, ",rescan_uuid_tree");
986 if (btrfs_test_opt(root
, CLEAR_CACHE
))
987 seq_puts(seq
, ",clear_cache");
988 if (btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
989 seq_puts(seq
, ",user_subvol_rm_allowed");
990 if (btrfs_test_opt(root
, ENOSPC_DEBUG
))
991 seq_puts(seq
, ",enospc_debug");
992 if (btrfs_test_opt(root
, AUTO_DEFRAG
))
993 seq_puts(seq
, ",autodefrag");
994 if (btrfs_test_opt(root
, INODE_MAP_CACHE
))
995 seq_puts(seq
, ",inode_cache");
996 if (btrfs_test_opt(root
, SKIP_BALANCE
))
997 seq_puts(seq
, ",skip_balance");
998 if (btrfs_test_opt(root
, RECOVERY
))
999 seq_puts(seq
, ",recovery");
1000 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1001 if (btrfs_test_opt(root
, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
))
1002 seq_puts(seq
, ",check_int_data");
1003 else if (btrfs_test_opt(root
, CHECK_INTEGRITY
))
1004 seq_puts(seq
, ",check_int");
1005 if (info
->check_integrity_print_mask
)
1006 seq_printf(seq
, ",check_int_print_mask=%d",
1007 info
->check_integrity_print_mask
);
1009 if (info
->metadata_ratio
)
1010 seq_printf(seq
, ",metadata_ratio=%d",
1011 info
->metadata_ratio
);
1012 if (btrfs_test_opt(root
, PANIC_ON_FATAL_ERROR
))
1013 seq_puts(seq
, ",fatal_errors=panic");
1014 if (info
->commit_interval
!= BTRFS_DEFAULT_COMMIT_INTERVAL
)
1015 seq_printf(seq
, ",commit=%d", info
->commit_interval
);
1019 static int btrfs_test_super(struct super_block
*s
, void *data
)
1021 struct btrfs_fs_info
*p
= data
;
1022 struct btrfs_fs_info
*fs_info
= btrfs_sb(s
);
1024 return fs_info
->fs_devices
== p
->fs_devices
;
1027 static int btrfs_set_super(struct super_block
*s
, void *data
)
1029 int err
= set_anon_super(s
, data
);
1031 s
->s_fs_info
= data
;
1036 * subvolumes are identified by ino 256
1038 static inline int is_subvolume_inode(struct inode
*inode
)
1040 if (inode
&& inode
->i_ino
== BTRFS_FIRST_FREE_OBJECTID
)
1046 * This will strip out the subvol=%s argument for an argument string and add
1047 * subvolid=0 to make sure we get the actual tree root for path walking to the
1050 static char *setup_root_args(char *args
)
1052 unsigned len
= strlen(args
) + 2 + 1;
1053 char *src
, *dst
, *buf
;
1056 * We need the same args as before, but with this substitution:
1057 * s!subvol=[^,]+!subvolid=0!
1059 * Since the replacement string is up to 2 bytes longer than the
1060 * original, allocate strlen(args) + 2 + 1 bytes.
1063 src
= strstr(args
, "subvol=");
1064 /* This shouldn't happen, but just in case.. */
1068 buf
= dst
= kmalloc(len
, GFP_NOFS
);
1073 * If the subvol= arg is not at the start of the string,
1074 * copy whatever precedes it into buf.
1079 dst
+= strlen(args
);
1082 strcpy(dst
, "subvolid=0");
1083 dst
+= strlen("subvolid=0");
1086 * If there is a "," after the original subvol=... string,
1087 * copy that suffix into our buffer. Otherwise, we're done.
1089 src
= strchr(src
, ',');
1096 static struct dentry
*mount_subvol(const char *subvol_name
, int flags
,
1097 const char *device_name
, char *data
)
1099 struct dentry
*root
;
1100 struct vfsmount
*mnt
;
1103 newargs
= setup_root_args(data
);
1105 return ERR_PTR(-ENOMEM
);
1106 mnt
= vfs_kern_mount(&btrfs_fs_type
, flags
, device_name
,
1110 return ERR_CAST(mnt
);
1112 root
= mount_subtree(mnt
, subvol_name
);
1114 if (!IS_ERR(root
) && !is_subvolume_inode(root
->d_inode
)) {
1115 struct super_block
*s
= root
->d_sb
;
1117 root
= ERR_PTR(-EINVAL
);
1118 deactivate_locked_super(s
);
1119 printk(KERN_ERR
"btrfs: '%s' is not a valid subvolume\n",
1127 * Find a superblock for the given device / mount point.
1129 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1130 * for multiple device setup. Make sure to keep it in sync.
1132 static struct dentry
*btrfs_mount(struct file_system_type
*fs_type
, int flags
,
1133 const char *device_name
, void *data
)
1135 struct block_device
*bdev
= NULL
;
1136 struct super_block
*s
;
1137 struct dentry
*root
;
1138 struct btrfs_fs_devices
*fs_devices
= NULL
;
1139 struct btrfs_fs_info
*fs_info
= NULL
;
1140 fmode_t mode
= FMODE_READ
;
1141 char *subvol_name
= NULL
;
1142 u64 subvol_objectid
= 0;
1145 if (!(flags
& MS_RDONLY
))
1146 mode
|= FMODE_WRITE
;
1148 error
= btrfs_parse_early_options(data
, mode
, fs_type
,
1149 &subvol_name
, &subvol_objectid
,
1153 return ERR_PTR(error
);
1157 root
= mount_subvol(subvol_name
, flags
, device_name
, data
);
1162 error
= btrfs_scan_one_device(device_name
, mode
, fs_type
, &fs_devices
);
1164 return ERR_PTR(error
);
1167 * Setup a dummy root and fs_info for test/set super. This is because
1168 * we don't actually fill this stuff out until open_ctree, but we need
1169 * it for searching for existing supers, so this lets us do that and
1170 * then open_ctree will properly initialize everything later.
1172 fs_info
= kzalloc(sizeof(struct btrfs_fs_info
), GFP_NOFS
);
1174 return ERR_PTR(-ENOMEM
);
1176 fs_info
->fs_devices
= fs_devices
;
1178 fs_info
->super_copy
= kzalloc(BTRFS_SUPER_INFO_SIZE
, GFP_NOFS
);
1179 fs_info
->super_for_commit
= kzalloc(BTRFS_SUPER_INFO_SIZE
, GFP_NOFS
);
1180 if (!fs_info
->super_copy
|| !fs_info
->super_for_commit
) {
1185 error
= btrfs_open_devices(fs_devices
, mode
, fs_type
);
1189 if (!(flags
& MS_RDONLY
) && fs_devices
->rw_devices
== 0) {
1191 goto error_close_devices
;
1194 bdev
= fs_devices
->latest_bdev
;
1195 s
= sget(fs_type
, btrfs_test_super
, btrfs_set_super
, flags
| MS_NOSEC
,
1199 goto error_close_devices
;
1203 btrfs_close_devices(fs_devices
);
1204 free_fs_info(fs_info
);
1205 if ((flags
^ s
->s_flags
) & MS_RDONLY
)
1208 char b
[BDEVNAME_SIZE
];
1210 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
1211 btrfs_sb(s
)->bdev_holder
= fs_type
;
1212 error
= btrfs_fill_super(s
, fs_devices
, data
,
1213 flags
& MS_SILENT
? 1 : 0);
1216 root
= !error
? get_default_root(s
, subvol_objectid
) : ERR_PTR(error
);
1218 deactivate_locked_super(s
);
1222 error_close_devices
:
1223 btrfs_close_devices(fs_devices
);
1225 free_fs_info(fs_info
);
1226 return ERR_PTR(error
);
1229 static void btrfs_set_max_workers(struct btrfs_workers
*workers
, int new_limit
)
1231 spin_lock_irq(&workers
->lock
);
1232 workers
->max_workers
= new_limit
;
1233 spin_unlock_irq(&workers
->lock
);
1236 static void btrfs_resize_thread_pool(struct btrfs_fs_info
*fs_info
,
1237 int new_pool_size
, int old_pool_size
)
1239 if (new_pool_size
== old_pool_size
)
1242 fs_info
->thread_pool_size
= new_pool_size
;
1244 printk(KERN_INFO
"btrfs: resize thread pool %d -> %d\n",
1245 old_pool_size
, new_pool_size
);
1247 btrfs_set_max_workers(&fs_info
->generic_worker
, new_pool_size
);
1248 btrfs_set_max_workers(&fs_info
->workers
, new_pool_size
);
1249 btrfs_set_max_workers(&fs_info
->delalloc_workers
, new_pool_size
);
1250 btrfs_set_max_workers(&fs_info
->submit_workers
, new_pool_size
);
1251 btrfs_set_max_workers(&fs_info
->caching_workers
, new_pool_size
);
1252 btrfs_set_max_workers(&fs_info
->fixup_workers
, new_pool_size
);
1253 btrfs_set_max_workers(&fs_info
->endio_workers
, new_pool_size
);
1254 btrfs_set_max_workers(&fs_info
->endio_meta_workers
, new_pool_size
);
1255 btrfs_set_max_workers(&fs_info
->endio_meta_write_workers
, new_pool_size
);
1256 btrfs_set_max_workers(&fs_info
->endio_write_workers
, new_pool_size
);
1257 btrfs_set_max_workers(&fs_info
->endio_freespace_worker
, new_pool_size
);
1258 btrfs_set_max_workers(&fs_info
->delayed_workers
, new_pool_size
);
1259 btrfs_set_max_workers(&fs_info
->readahead_workers
, new_pool_size
);
1260 btrfs_set_max_workers(&fs_info
->scrub_wr_completion_workers
,
1264 static inline void btrfs_remount_prepare(struct btrfs_fs_info
*fs_info
)
1266 set_bit(BTRFS_FS_STATE_REMOUNTING
, &fs_info
->fs_state
);
1269 static inline void btrfs_remount_begin(struct btrfs_fs_info
*fs_info
,
1270 unsigned long old_opts
, int flags
)
1272 if (btrfs_raw_test_opt(old_opts
, AUTO_DEFRAG
) &&
1273 (!btrfs_raw_test_opt(fs_info
->mount_opt
, AUTO_DEFRAG
) ||
1274 (flags
& MS_RDONLY
))) {
1275 /* wait for any defraggers to finish */
1276 wait_event(fs_info
->transaction_wait
,
1277 (atomic_read(&fs_info
->defrag_running
) == 0));
1278 if (flags
& MS_RDONLY
)
1279 sync_filesystem(fs_info
->sb
);
1283 static inline void btrfs_remount_cleanup(struct btrfs_fs_info
*fs_info
,
1284 unsigned long old_opts
)
1287 * We need cleanup all defragable inodes if the autodefragment is
1288 * close or the fs is R/O.
1290 if (btrfs_raw_test_opt(old_opts
, AUTO_DEFRAG
) &&
1291 (!btrfs_raw_test_opt(fs_info
->mount_opt
, AUTO_DEFRAG
) ||
1292 (fs_info
->sb
->s_flags
& MS_RDONLY
))) {
1293 btrfs_cleanup_defrag_inodes(fs_info
);
1296 clear_bit(BTRFS_FS_STATE_REMOUNTING
, &fs_info
->fs_state
);
1299 static int btrfs_remount(struct super_block
*sb
, int *flags
, char *data
)
1301 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1302 struct btrfs_root
*root
= fs_info
->tree_root
;
1303 unsigned old_flags
= sb
->s_flags
;
1304 unsigned long old_opts
= fs_info
->mount_opt
;
1305 unsigned long old_compress_type
= fs_info
->compress_type
;
1306 u64 old_max_inline
= fs_info
->max_inline
;
1307 u64 old_alloc_start
= fs_info
->alloc_start
;
1308 int old_thread_pool_size
= fs_info
->thread_pool_size
;
1309 unsigned int old_metadata_ratio
= fs_info
->metadata_ratio
;
1312 btrfs_remount_prepare(fs_info
);
1314 ret
= btrfs_parse_options(root
, data
);
1320 btrfs_remount_begin(fs_info
, old_opts
, *flags
);
1321 btrfs_resize_thread_pool(fs_info
,
1322 fs_info
->thread_pool_size
, old_thread_pool_size
);
1324 if ((*flags
& MS_RDONLY
) == (sb
->s_flags
& MS_RDONLY
))
1327 if (*flags
& MS_RDONLY
) {
1329 * this also happens on 'umount -rf' or on shutdown, when
1330 * the filesystem is busy.
1333 /* wait for the uuid_scan task to finish */
1334 down(&fs_info
->uuid_tree_rescan_sem
);
1335 /* avoid complains from lockdep et al. */
1336 up(&fs_info
->uuid_tree_rescan_sem
);
1338 sb
->s_flags
|= MS_RDONLY
;
1340 btrfs_dev_replace_suspend_for_unmount(fs_info
);
1341 btrfs_scrub_cancel(fs_info
);
1342 btrfs_pause_balance(fs_info
);
1344 ret
= btrfs_commit_super(root
);
1348 if (test_bit(BTRFS_FS_STATE_ERROR
, &root
->fs_info
->fs_state
)) {
1350 "Remounting read-write after error is not allowed\n");
1354 if (fs_info
->fs_devices
->rw_devices
== 0) {
1359 if (fs_info
->fs_devices
->missing_devices
>
1360 fs_info
->num_tolerated_disk_barrier_failures
&&
1361 !(*flags
& MS_RDONLY
)) {
1363 "Btrfs: too many missing devices, writeable remount is not allowed\n");
1368 if (btrfs_super_log_root(fs_info
->super_copy
) != 0) {
1373 ret
= btrfs_cleanup_fs_roots(fs_info
);
1377 /* recover relocation */
1378 ret
= btrfs_recover_relocation(root
);
1382 ret
= btrfs_resume_balance_async(fs_info
);
1386 ret
= btrfs_resume_dev_replace_async(fs_info
);
1388 pr_warn("btrfs: failed to resume dev_replace\n");
1392 if (!fs_info
->uuid_root
) {
1393 pr_info("btrfs: creating UUID tree\n");
1394 ret
= btrfs_create_uuid_tree(fs_info
);
1396 pr_warn("btrfs: failed to create the uuid tree"
1401 sb
->s_flags
&= ~MS_RDONLY
;
1404 btrfs_remount_cleanup(fs_info
, old_opts
);
1408 /* We've hit an error - don't reset MS_RDONLY */
1409 if (sb
->s_flags
& MS_RDONLY
)
1410 old_flags
|= MS_RDONLY
;
1411 sb
->s_flags
= old_flags
;
1412 fs_info
->mount_opt
= old_opts
;
1413 fs_info
->compress_type
= old_compress_type
;
1414 fs_info
->max_inline
= old_max_inline
;
1415 mutex_lock(&fs_info
->chunk_mutex
);
1416 fs_info
->alloc_start
= old_alloc_start
;
1417 mutex_unlock(&fs_info
->chunk_mutex
);
1418 btrfs_resize_thread_pool(fs_info
,
1419 old_thread_pool_size
, fs_info
->thread_pool_size
);
1420 fs_info
->metadata_ratio
= old_metadata_ratio
;
1421 btrfs_remount_cleanup(fs_info
, old_opts
);
1425 /* Used to sort the devices by max_avail(descending sort) */
1426 static int btrfs_cmp_device_free_bytes(const void *dev_info1
,
1427 const void *dev_info2
)
1429 if (((struct btrfs_device_info
*)dev_info1
)->max_avail
>
1430 ((struct btrfs_device_info
*)dev_info2
)->max_avail
)
1432 else if (((struct btrfs_device_info
*)dev_info1
)->max_avail
<
1433 ((struct btrfs_device_info
*)dev_info2
)->max_avail
)
1440 * sort the devices by max_avail, in which max free extent size of each device
1441 * is stored.(Descending Sort)
1443 static inline void btrfs_descending_sort_devices(
1444 struct btrfs_device_info
*devices
,
1447 sort(devices
, nr_devices
, sizeof(struct btrfs_device_info
),
1448 btrfs_cmp_device_free_bytes
, NULL
);
1452 * The helper to calc the free space on the devices that can be used to store
1455 static int btrfs_calc_avail_data_space(struct btrfs_root
*root
, u64
*free_bytes
)
1457 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1458 struct btrfs_device_info
*devices_info
;
1459 struct btrfs_fs_devices
*fs_devices
= fs_info
->fs_devices
;
1460 struct btrfs_device
*device
;
1465 u64 min_stripe_size
;
1466 int min_stripes
= 1, num_stripes
= 1;
1467 int i
= 0, nr_devices
;
1470 nr_devices
= fs_info
->fs_devices
->open_devices
;
1471 BUG_ON(!nr_devices
);
1473 devices_info
= kmalloc_array(nr_devices
, sizeof(*devices_info
),
1478 /* calc min stripe number for data space alloction */
1479 type
= btrfs_get_alloc_profile(root
, 1);
1480 if (type
& BTRFS_BLOCK_GROUP_RAID0
) {
1482 num_stripes
= nr_devices
;
1483 } else if (type
& BTRFS_BLOCK_GROUP_RAID1
) {
1486 } else if (type
& BTRFS_BLOCK_GROUP_RAID10
) {
1491 if (type
& BTRFS_BLOCK_GROUP_DUP
)
1492 min_stripe_size
= 2 * BTRFS_STRIPE_LEN
;
1494 min_stripe_size
= BTRFS_STRIPE_LEN
;
1496 list_for_each_entry(device
, &fs_devices
->devices
, dev_list
) {
1497 if (!device
->in_fs_metadata
|| !device
->bdev
||
1498 device
->is_tgtdev_for_dev_replace
)
1501 avail_space
= device
->total_bytes
- device
->bytes_used
;
1503 /* align with stripe_len */
1504 do_div(avail_space
, BTRFS_STRIPE_LEN
);
1505 avail_space
*= BTRFS_STRIPE_LEN
;
1508 * In order to avoid overwritting the superblock on the drive,
1509 * btrfs starts at an offset of at least 1MB when doing chunk
1512 skip_space
= 1024 * 1024;
1514 /* user can set the offset in fs_info->alloc_start. */
1515 if (fs_info
->alloc_start
+ BTRFS_STRIPE_LEN
<=
1516 device
->total_bytes
)
1517 skip_space
= max(fs_info
->alloc_start
, skip_space
);
1520 * btrfs can not use the free space in [0, skip_space - 1],
1521 * we must subtract it from the total. In order to implement
1522 * it, we account the used space in this range first.
1524 ret
= btrfs_account_dev_extents_size(device
, 0, skip_space
- 1,
1527 kfree(devices_info
);
1531 /* calc the free space in [0, skip_space - 1] */
1532 skip_space
-= used_space
;
1535 * we can use the free space in [0, skip_space - 1], subtract
1536 * it from the total.
1538 if (avail_space
&& avail_space
>= skip_space
)
1539 avail_space
-= skip_space
;
1543 if (avail_space
< min_stripe_size
)
1546 devices_info
[i
].dev
= device
;
1547 devices_info
[i
].max_avail
= avail_space
;
1554 btrfs_descending_sort_devices(devices_info
, nr_devices
);
1558 while (nr_devices
>= min_stripes
) {
1559 if (num_stripes
> nr_devices
)
1560 num_stripes
= nr_devices
;
1562 if (devices_info
[i
].max_avail
>= min_stripe_size
) {
1566 avail_space
+= devices_info
[i
].max_avail
* num_stripes
;
1567 alloc_size
= devices_info
[i
].max_avail
;
1568 for (j
= i
+ 1 - num_stripes
; j
<= i
; j
++)
1569 devices_info
[j
].max_avail
-= alloc_size
;
1575 kfree(devices_info
);
1576 *free_bytes
= avail_space
;
1580 static int btrfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1582 struct btrfs_fs_info
*fs_info
= btrfs_sb(dentry
->d_sb
);
1583 struct btrfs_super_block
*disk_super
= fs_info
->super_copy
;
1584 struct list_head
*head
= &fs_info
->space_info
;
1585 struct btrfs_space_info
*found
;
1587 u64 total_free_data
= 0;
1588 int bits
= dentry
->d_sb
->s_blocksize_bits
;
1589 __be32
*fsid
= (__be32
*)fs_info
->fsid
;
1592 /* holding chunk_muext to avoid allocating new chunks */
1593 mutex_lock(&fs_info
->chunk_mutex
);
1595 list_for_each_entry_rcu(found
, head
, list
) {
1596 if (found
->flags
& BTRFS_BLOCK_GROUP_DATA
) {
1597 total_free_data
+= found
->disk_total
- found
->disk_used
;
1599 btrfs_account_ro_block_groups_free_space(found
);
1602 total_used
+= found
->disk_used
;
1606 buf
->f_namelen
= BTRFS_NAME_LEN
;
1607 buf
->f_blocks
= btrfs_super_total_bytes(disk_super
) >> bits
;
1608 buf
->f_bfree
= buf
->f_blocks
- (total_used
>> bits
);
1609 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1610 buf
->f_type
= BTRFS_SUPER_MAGIC
;
1611 buf
->f_bavail
= total_free_data
;
1612 ret
= btrfs_calc_avail_data_space(fs_info
->tree_root
, &total_free_data
);
1614 mutex_unlock(&fs_info
->chunk_mutex
);
1617 buf
->f_bavail
+= total_free_data
;
1618 buf
->f_bavail
= buf
->f_bavail
>> bits
;
1619 mutex_unlock(&fs_info
->chunk_mutex
);
1621 /* We treat it as constant endianness (it doesn't matter _which_)
1622 because we want the fsid to come out the same whether mounted
1623 on a big-endian or little-endian host */
1624 buf
->f_fsid
.val
[0] = be32_to_cpu(fsid
[0]) ^ be32_to_cpu(fsid
[2]);
1625 buf
->f_fsid
.val
[1] = be32_to_cpu(fsid
[1]) ^ be32_to_cpu(fsid
[3]);
1626 /* Mask in the root object ID too, to disambiguate subvols */
1627 buf
->f_fsid
.val
[0] ^= BTRFS_I(dentry
->d_inode
)->root
->objectid
>> 32;
1628 buf
->f_fsid
.val
[1] ^= BTRFS_I(dentry
->d_inode
)->root
->objectid
;
1633 static void btrfs_kill_super(struct super_block
*sb
)
1635 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1636 kill_anon_super(sb
);
1637 free_fs_info(fs_info
);
1640 static struct file_system_type btrfs_fs_type
= {
1641 .owner
= THIS_MODULE
,
1643 .mount
= btrfs_mount
,
1644 .kill_sb
= btrfs_kill_super
,
1645 .fs_flags
= FS_REQUIRES_DEV
,
1647 MODULE_ALIAS_FS("btrfs");
1650 * used by btrfsctl to scan devices when no FS is mounted
1652 static long btrfs_control_ioctl(struct file
*file
, unsigned int cmd
,
1655 struct btrfs_ioctl_vol_args
*vol
;
1656 struct btrfs_fs_devices
*fs_devices
;
1659 if (!capable(CAP_SYS_ADMIN
))
1662 vol
= memdup_user((void __user
*)arg
, sizeof(*vol
));
1664 return PTR_ERR(vol
);
1667 case BTRFS_IOC_SCAN_DEV
:
1668 ret
= btrfs_scan_one_device(vol
->name
, FMODE_READ
,
1669 &btrfs_fs_type
, &fs_devices
);
1671 case BTRFS_IOC_DEVICES_READY
:
1672 ret
= btrfs_scan_one_device(vol
->name
, FMODE_READ
,
1673 &btrfs_fs_type
, &fs_devices
);
1676 ret
= !(fs_devices
->num_devices
== fs_devices
->total_devices
);
1684 static int btrfs_freeze(struct super_block
*sb
)
1686 struct btrfs_trans_handle
*trans
;
1687 struct btrfs_root
*root
= btrfs_sb(sb
)->tree_root
;
1689 trans
= btrfs_attach_transaction_barrier(root
);
1690 if (IS_ERR(trans
)) {
1691 /* no transaction, don't bother */
1692 if (PTR_ERR(trans
) == -ENOENT
)
1694 return PTR_ERR(trans
);
1696 return btrfs_commit_transaction(trans
, root
);
1699 static int btrfs_unfreeze(struct super_block
*sb
)
1704 static int btrfs_show_devname(struct seq_file
*m
, struct dentry
*root
)
1706 struct btrfs_fs_info
*fs_info
= btrfs_sb(root
->d_sb
);
1707 struct btrfs_fs_devices
*cur_devices
;
1708 struct btrfs_device
*dev
, *first_dev
= NULL
;
1709 struct list_head
*head
;
1710 struct rcu_string
*name
;
1712 mutex_lock(&fs_info
->fs_devices
->device_list_mutex
);
1713 cur_devices
= fs_info
->fs_devices
;
1714 while (cur_devices
) {
1715 head
= &cur_devices
->devices
;
1716 list_for_each_entry(dev
, head
, dev_list
) {
1719 if (!first_dev
|| dev
->devid
< first_dev
->devid
)
1722 cur_devices
= cur_devices
->seed
;
1727 name
= rcu_dereference(first_dev
->name
);
1728 seq_escape(m
, name
->str
, " \t\n\\");
1733 mutex_unlock(&fs_info
->fs_devices
->device_list_mutex
);
1737 static const struct super_operations btrfs_super_ops
= {
1738 .drop_inode
= btrfs_drop_inode
,
1739 .evict_inode
= btrfs_evict_inode
,
1740 .put_super
= btrfs_put_super
,
1741 .sync_fs
= btrfs_sync_fs
,
1742 .show_options
= btrfs_show_options
,
1743 .show_devname
= btrfs_show_devname
,
1744 .write_inode
= btrfs_write_inode
,
1745 .alloc_inode
= btrfs_alloc_inode
,
1746 .destroy_inode
= btrfs_destroy_inode
,
1747 .statfs
= btrfs_statfs
,
1748 .remount_fs
= btrfs_remount
,
1749 .freeze_fs
= btrfs_freeze
,
1750 .unfreeze_fs
= btrfs_unfreeze
,
1753 static const struct file_operations btrfs_ctl_fops
= {
1754 .unlocked_ioctl
= btrfs_control_ioctl
,
1755 .compat_ioctl
= btrfs_control_ioctl
,
1756 .owner
= THIS_MODULE
,
1757 .llseek
= noop_llseek
,
1760 static struct miscdevice btrfs_misc
= {
1761 .minor
= BTRFS_MINOR
,
1762 .name
= "btrfs-control",
1763 .fops
= &btrfs_ctl_fops
1766 MODULE_ALIAS_MISCDEV(BTRFS_MINOR
);
1767 MODULE_ALIAS("devname:btrfs-control");
1769 static int btrfs_interface_init(void)
1771 return misc_register(&btrfs_misc
);
1774 static void btrfs_interface_exit(void)
1776 if (misc_deregister(&btrfs_misc
) < 0)
1777 printk(KERN_INFO
"btrfs: misc_deregister failed for control device\n");
1780 static void btrfs_print_info(void)
1782 printk(KERN_INFO
"Btrfs loaded"
1783 #ifdef CONFIG_BTRFS_DEBUG
1786 #ifdef CONFIG_BTRFS_ASSERT
1789 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1790 ", integrity-checker=on"
1795 static int btrfs_run_sanity_tests(void)
1799 ret
= btrfs_init_test_fs();
1803 ret
= btrfs_test_free_space_cache();
1806 ret
= btrfs_test_extent_buffer_operations();
1809 ret
= btrfs_test_extent_io();
1812 ret
= btrfs_test_inodes();
1814 btrfs_destroy_test_fs();
1818 static int __init
init_btrfs_fs(void)
1822 err
= btrfs_init_sysfs();
1826 btrfs_init_compress();
1828 err
= btrfs_init_cachep();
1832 err
= extent_io_init();
1836 err
= extent_map_init();
1838 goto free_extent_io
;
1840 err
= ordered_data_init();
1842 goto free_extent_map
;
1844 err
= btrfs_delayed_inode_init();
1846 goto free_ordered_data
;
1848 err
= btrfs_auto_defrag_init();
1850 goto free_delayed_inode
;
1852 err
= btrfs_delayed_ref_init();
1854 goto free_auto_defrag
;
1856 err
= btrfs_prelim_ref_init();
1858 goto free_prelim_ref
;
1860 err
= btrfs_interface_init();
1862 goto free_delayed_ref
;
1864 btrfs_init_lockdep();
1868 err
= btrfs_run_sanity_tests();
1870 goto unregister_ioctl
;
1872 err
= register_filesystem(&btrfs_fs_type
);
1874 goto unregister_ioctl
;
1879 btrfs_interface_exit();
1881 btrfs_prelim_ref_exit();
1883 btrfs_delayed_ref_exit();
1885 btrfs_auto_defrag_exit();
1887 btrfs_delayed_inode_exit();
1889 ordered_data_exit();
1895 btrfs_destroy_cachep();
1897 btrfs_exit_compress();
1902 static void __exit
exit_btrfs_fs(void)
1904 btrfs_destroy_cachep();
1905 btrfs_delayed_ref_exit();
1906 btrfs_auto_defrag_exit();
1907 btrfs_delayed_inode_exit();
1908 btrfs_prelim_ref_exit();
1909 ordered_data_exit();
1912 btrfs_interface_exit();
1913 unregister_filesystem(&btrfs_fs_type
);
1915 btrfs_cleanup_fs_uuids();
1916 btrfs_exit_compress();
1919 module_init(init_btrfs_fs
)
1920 module_exit(exit_btrfs_fs
)
1922 MODULE_LICENSE("GPL");