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"
56 #include "compression.h"
57 #include "rcu-string.h"
58 #include "dev-replace.h"
59 #include "free-space-cache.h"
61 #include "tests/btrfs-tests.h"
64 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h>
67 static const struct super_operations btrfs_super_ops
;
68 static struct file_system_type btrfs_fs_type
;
70 static int btrfs_remount(struct super_block
*sb
, int *flags
, char *data
);
72 static const char *btrfs_decode_error(int errno
)
74 char *errstr
= "unknown";
78 errstr
= "IO failure";
81 errstr
= "Out of memory";
84 errstr
= "Readonly filesystem";
87 errstr
= "Object already exists";
90 errstr
= "No space left";
93 errstr
= "No such entry";
100 static void save_error_info(struct btrfs_fs_info
*fs_info
)
103 * today we only save the error info into ram. Long term we'll
104 * also send it down to the disk
106 set_bit(BTRFS_FS_STATE_ERROR
, &fs_info
->fs_state
);
109 /* btrfs handle error by forcing the filesystem readonly */
110 static void btrfs_handle_error(struct btrfs_fs_info
*fs_info
)
112 struct super_block
*sb
= fs_info
->sb
;
114 if (sb
->s_flags
& MS_RDONLY
)
117 if (test_bit(BTRFS_FS_STATE_ERROR
, &fs_info
->fs_state
)) {
118 sb
->s_flags
|= MS_RDONLY
;
119 btrfs_info(fs_info
, "forced readonly");
121 * Note that a running device replace operation is not
122 * canceled here although there is no way to update
123 * the progress. It would add the risk of a deadlock,
124 * therefore the canceling is ommited. The only penalty
125 * is that some I/O remains active until the procedure
126 * completes. The next time when the filesystem is
127 * mounted writeable again, the device replace
128 * operation continues.
135 * __btrfs_std_error decodes expected errors from the caller and
136 * invokes the approciate error response.
138 void __btrfs_std_error(struct btrfs_fs_info
*fs_info
, const char *function
,
139 unsigned int line
, int errno
, const char *fmt
, ...)
141 struct super_block
*sb
= fs_info
->sb
;
145 * Special case: if the error is EROFS, and we're already
146 * under MS_RDONLY, then it is safe here.
148 if (errno
== -EROFS
&& (sb
->s_flags
& MS_RDONLY
))
151 errstr
= btrfs_decode_error(errno
);
153 struct va_format vaf
;
161 "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
162 sb
->s_id
, function
, line
, errno
, errstr
, &vaf
);
165 printk(KERN_CRIT
"BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
166 sb
->s_id
, function
, line
, errno
, errstr
);
169 /* Don't go through full error handling during mount */
170 save_error_info(fs_info
);
171 if (sb
->s_flags
& MS_BORN
)
172 btrfs_handle_error(fs_info
);
175 static const char * const logtypes
[] = {
186 void btrfs_printk(const struct btrfs_fs_info
*fs_info
, const char *fmt
, ...)
188 struct super_block
*sb
= fs_info
->sb
;
190 struct va_format vaf
;
192 const char *type
= logtypes
[4];
197 kern_level
= printk_get_level(fmt
);
199 size_t size
= printk_skip_level(fmt
) - fmt
;
200 memcpy(lvl
, fmt
, size
);
203 type
= logtypes
[kern_level
- '0'];
210 printk("%sBTRFS %s (device %s): %pV\n", lvl
, type
, sb
->s_id
, &vaf
);
217 void __btrfs_std_error(struct btrfs_fs_info
*fs_info
, const char *function
,
218 unsigned int line
, int errno
, const char *fmt
, ...)
220 struct super_block
*sb
= fs_info
->sb
;
223 * Special case: if the error is EROFS, and we're already
224 * under MS_RDONLY, then it is safe here.
226 if (errno
== -EROFS
&& (sb
->s_flags
& MS_RDONLY
))
229 /* Don't go through full error handling during mount */
230 if (sb
->s_flags
& MS_BORN
) {
231 save_error_info(fs_info
);
232 btrfs_handle_error(fs_info
);
238 * We only mark the transaction aborted and then set the file system read-only.
239 * This will prevent new transactions from starting or trying to join this
242 * This means that error recovery at the call site is limited to freeing
243 * any local memory allocations and passing the error code up without
244 * further cleanup. The transaction should complete as it normally would
245 * in the call path but will return -EIO.
247 * We'll complete the cleanup in btrfs_end_transaction and
248 * btrfs_commit_transaction.
250 void __btrfs_abort_transaction(struct btrfs_trans_handle
*trans
,
251 struct btrfs_root
*root
, const char *function
,
252 unsigned int line
, int errno
)
254 trans
->aborted
= errno
;
255 /* Nothing used. The other threads that have joined this
256 * transaction may be able to continue. */
257 if (!trans
->blocks_used
&& list_empty(&trans
->new_bgs
)) {
260 errstr
= btrfs_decode_error(errno
);
261 btrfs_warn(root
->fs_info
,
262 "%s:%d: Aborting unused transaction(%s).",
263 function
, line
, errstr
);
266 ACCESS_ONCE(trans
->transaction
->aborted
) = errno
;
267 /* Wake up anybody who may be waiting on this transaction */
268 wake_up(&root
->fs_info
->transaction_wait
);
269 wake_up(&root
->fs_info
->transaction_blocked_wait
);
270 __btrfs_std_error(root
->fs_info
, function
, line
, errno
, NULL
);
273 * __btrfs_panic decodes unexpected, fatal errors from the caller,
274 * issues an alert, and either panics or BUGs, depending on mount options.
276 void __btrfs_panic(struct btrfs_fs_info
*fs_info
, const char *function
,
277 unsigned int line
, int errno
, const char *fmt
, ...)
279 char *s_id
= "<unknown>";
281 struct va_format vaf
= { .fmt
= fmt
};
285 s_id
= fs_info
->sb
->s_id
;
290 errstr
= btrfs_decode_error(errno
);
291 if (fs_info
&& (fs_info
->mount_opt
& BTRFS_MOUNT_PANIC_ON_FATAL_ERROR
))
292 panic(KERN_CRIT
"BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
293 s_id
, function
, line
, &vaf
, errno
, errstr
);
295 btrfs_crit(fs_info
, "panic in %s:%d: %pV (errno=%d %s)",
296 function
, line
, &vaf
, errno
, errstr
);
298 /* Caller calls BUG() */
301 static void btrfs_put_super(struct super_block
*sb
)
303 close_ctree(btrfs_sb(sb
)->tree_root
);
307 Opt_degraded
, Opt_subvol
, Opt_subvolid
, Opt_device
, Opt_nodatasum
,
308 Opt_nodatacow
, Opt_max_inline
, Opt_alloc_start
, Opt_nobarrier
, Opt_ssd
,
309 Opt_nossd
, Opt_ssd_spread
, Opt_thread_pool
, Opt_noacl
, Opt_compress
,
310 Opt_compress_type
, Opt_compress_force
, Opt_compress_force_type
,
311 Opt_notreelog
, Opt_ratio
, Opt_flushoncommit
, Opt_discard
,
312 Opt_space_cache
, Opt_clear_cache
, Opt_user_subvol_rm_allowed
,
313 Opt_enospc_debug
, Opt_subvolrootid
, Opt_defrag
, Opt_inode_cache
,
314 Opt_no_space_cache
, Opt_recovery
, Opt_skip_balance
,
315 Opt_check_integrity
, Opt_check_integrity_including_extent_data
,
316 Opt_check_integrity_print_mask
, Opt_fatal_errors
, Opt_rescan_uuid_tree
,
317 Opt_commit_interval
, Opt_barrier
, Opt_nodefrag
, Opt_nodiscard
,
318 Opt_noenospc_debug
, Opt_noflushoncommit
, Opt_acl
, Opt_datacow
,
319 Opt_datasum
, Opt_treelog
, Opt_noinode_cache
,
323 static match_table_t tokens
= {
324 {Opt_degraded
, "degraded"},
325 {Opt_subvol
, "subvol=%s"},
326 {Opt_subvolid
, "subvolid=%s"},
327 {Opt_device
, "device=%s"},
328 {Opt_nodatasum
, "nodatasum"},
329 {Opt_datasum
, "datasum"},
330 {Opt_nodatacow
, "nodatacow"},
331 {Opt_datacow
, "datacow"},
332 {Opt_nobarrier
, "nobarrier"},
333 {Opt_barrier
, "barrier"},
334 {Opt_max_inline
, "max_inline=%s"},
335 {Opt_alloc_start
, "alloc_start=%s"},
336 {Opt_thread_pool
, "thread_pool=%d"},
337 {Opt_compress
, "compress"},
338 {Opt_compress_type
, "compress=%s"},
339 {Opt_compress_force
, "compress-force"},
340 {Opt_compress_force_type
, "compress-force=%s"},
342 {Opt_ssd_spread
, "ssd_spread"},
343 {Opt_nossd
, "nossd"},
345 {Opt_noacl
, "noacl"},
346 {Opt_notreelog
, "notreelog"},
347 {Opt_treelog
, "treelog"},
348 {Opt_flushoncommit
, "flushoncommit"},
349 {Opt_noflushoncommit
, "noflushoncommit"},
350 {Opt_ratio
, "metadata_ratio=%d"},
351 {Opt_discard
, "discard"},
352 {Opt_nodiscard
, "nodiscard"},
353 {Opt_space_cache
, "space_cache"},
354 {Opt_clear_cache
, "clear_cache"},
355 {Opt_user_subvol_rm_allowed
, "user_subvol_rm_allowed"},
356 {Opt_enospc_debug
, "enospc_debug"},
357 {Opt_noenospc_debug
, "noenospc_debug"},
358 {Opt_subvolrootid
, "subvolrootid=%d"},
359 {Opt_defrag
, "autodefrag"},
360 {Opt_nodefrag
, "noautodefrag"},
361 {Opt_inode_cache
, "inode_cache"},
362 {Opt_noinode_cache
, "noinode_cache"},
363 {Opt_no_space_cache
, "nospace_cache"},
364 {Opt_recovery
, "recovery"},
365 {Opt_skip_balance
, "skip_balance"},
366 {Opt_check_integrity
, "check_int"},
367 {Opt_check_integrity_including_extent_data
, "check_int_data"},
368 {Opt_check_integrity_print_mask
, "check_int_print_mask=%d"},
369 {Opt_rescan_uuid_tree
, "rescan_uuid_tree"},
370 {Opt_fatal_errors
, "fatal_errors=%s"},
371 {Opt_commit_interval
, "commit=%d"},
376 * Regular mount options parser. Everything that is needed only when
377 * reading in a new superblock is parsed here.
378 * XXX JDM: This needs to be cleaned up for remount.
380 int btrfs_parse_options(struct btrfs_root
*root
, char *options
)
382 struct btrfs_fs_info
*info
= root
->fs_info
;
383 substring_t args
[MAX_OPT_ARGS
];
384 char *p
, *num
, *orig
= NULL
;
389 bool compress_force
= false;
391 cache_gen
= btrfs_super_cache_generation(root
->fs_info
->super_copy
);
393 btrfs_set_opt(info
->mount_opt
, SPACE_CACHE
);
399 * strsep changes the string, duplicate it because parse_options
402 options
= kstrdup(options
, GFP_NOFS
);
408 while ((p
= strsep(&options
, ",")) != NULL
) {
413 token
= match_token(p
, tokens
, args
);
416 btrfs_info(root
->fs_info
, "allowing degraded mounts");
417 btrfs_set_opt(info
->mount_opt
, DEGRADED
);
421 case Opt_subvolrootid
:
424 * These are parsed by btrfs_parse_early_options
425 * and can be happily ignored here.
429 btrfs_set_and_info(root
, NODATASUM
,
430 "setting nodatasum");
433 if (btrfs_test_opt(root
, NODATASUM
)) {
434 if (btrfs_test_opt(root
, NODATACOW
))
435 btrfs_info(root
->fs_info
, "setting datasum, datacow enabled");
437 btrfs_info(root
->fs_info
, "setting datasum");
439 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
440 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
443 if (!btrfs_test_opt(root
, NODATACOW
)) {
444 if (!btrfs_test_opt(root
, COMPRESS
) ||
445 !btrfs_test_opt(root
, FORCE_COMPRESS
)) {
446 btrfs_info(root
->fs_info
,
447 "setting nodatacow, compression disabled");
449 btrfs_info(root
->fs_info
, "setting nodatacow");
452 btrfs_clear_opt(info
->mount_opt
, COMPRESS
);
453 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
454 btrfs_set_opt(info
->mount_opt
, NODATACOW
);
455 btrfs_set_opt(info
->mount_opt
, NODATASUM
);
458 btrfs_clear_and_info(root
, NODATACOW
,
461 case Opt_compress_force
:
462 case Opt_compress_force_type
:
463 compress_force
= true;
466 case Opt_compress_type
:
467 if (token
== Opt_compress
||
468 token
== Opt_compress_force
||
469 strcmp(args
[0].from
, "zlib") == 0) {
470 compress_type
= "zlib";
471 info
->compress_type
= BTRFS_COMPRESS_ZLIB
;
472 btrfs_set_opt(info
->mount_opt
, COMPRESS
);
473 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
474 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
475 } else if (strcmp(args
[0].from
, "lzo") == 0) {
476 compress_type
= "lzo";
477 info
->compress_type
= BTRFS_COMPRESS_LZO
;
478 btrfs_set_opt(info
->mount_opt
, COMPRESS
);
479 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
480 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
481 btrfs_set_fs_incompat(info
, COMPRESS_LZO
);
482 } else if (strncmp(args
[0].from
, "no", 2) == 0) {
483 compress_type
= "no";
484 btrfs_clear_opt(info
->mount_opt
, COMPRESS
);
485 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
486 compress_force
= false;
492 if (compress_force
) {
493 btrfs_set_and_info(root
, FORCE_COMPRESS
,
494 "force %s compression",
497 if (!btrfs_test_opt(root
, COMPRESS
))
498 btrfs_info(root
->fs_info
,
499 "btrfs: use %s compression",
502 * If we remount from compress-force=xxx to
503 * compress=xxx, we need clear FORCE_COMPRESS
504 * flag, otherwise, there is no way for users
505 * to disable forcible compression separately.
507 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
511 btrfs_set_and_info(root
, SSD
,
512 "use ssd allocation scheme");
515 btrfs_set_and_info(root
, SSD_SPREAD
,
516 "use spread ssd allocation scheme");
517 btrfs_set_opt(info
->mount_opt
, SSD
);
520 btrfs_set_and_info(root
, NOSSD
,
521 "not using ssd allocation scheme");
522 btrfs_clear_opt(info
->mount_opt
, SSD
);
525 btrfs_clear_and_info(root
, NOBARRIER
,
526 "turning on barriers");
529 btrfs_set_and_info(root
, NOBARRIER
,
530 "turning off barriers");
532 case Opt_thread_pool
:
533 ret
= match_int(&args
[0], &intarg
);
536 } else if (intarg
> 0) {
537 info
->thread_pool_size
= intarg
;
544 num
= match_strdup(&args
[0]);
546 info
->max_inline
= memparse(num
, NULL
);
549 if (info
->max_inline
) {
550 info
->max_inline
= min_t(u64
,
554 btrfs_info(root
->fs_info
, "max_inline at %llu",
561 case Opt_alloc_start
:
562 num
= match_strdup(&args
[0]);
564 mutex_lock(&info
->chunk_mutex
);
565 info
->alloc_start
= memparse(num
, NULL
);
566 mutex_unlock(&info
->chunk_mutex
);
568 btrfs_info(root
->fs_info
, "allocations start at %llu",
576 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
577 root
->fs_info
->sb
->s_flags
|= MS_POSIXACL
;
580 btrfs_err(root
->fs_info
,
581 "support for ACL not compiled in!");
586 root
->fs_info
->sb
->s_flags
&= ~MS_POSIXACL
;
589 btrfs_set_and_info(root
, NOTREELOG
,
590 "disabling tree log");
593 btrfs_clear_and_info(root
, NOTREELOG
,
594 "enabling tree log");
596 case Opt_flushoncommit
:
597 btrfs_set_and_info(root
, FLUSHONCOMMIT
,
598 "turning on flush-on-commit");
600 case Opt_noflushoncommit
:
601 btrfs_clear_and_info(root
, FLUSHONCOMMIT
,
602 "turning off flush-on-commit");
605 ret
= match_int(&args
[0], &intarg
);
608 } else if (intarg
>= 0) {
609 info
->metadata_ratio
= intarg
;
610 btrfs_info(root
->fs_info
, "metadata ratio %d",
611 info
->metadata_ratio
);
618 btrfs_set_and_info(root
, DISCARD
,
619 "turning on discard");
622 btrfs_clear_and_info(root
, DISCARD
,
623 "turning off discard");
625 case Opt_space_cache
:
626 btrfs_set_and_info(root
, SPACE_CACHE
,
627 "enabling disk space caching");
629 case Opt_rescan_uuid_tree
:
630 btrfs_set_opt(info
->mount_opt
, RESCAN_UUID_TREE
);
632 case Opt_no_space_cache
:
633 btrfs_clear_and_info(root
, SPACE_CACHE
,
634 "disabling disk space caching");
636 case Opt_inode_cache
:
637 btrfs_set_pending_and_info(info
, INODE_MAP_CACHE
,
638 "enabling inode map caching");
640 case Opt_noinode_cache
:
641 btrfs_clear_pending_and_info(info
, INODE_MAP_CACHE
,
642 "disabling inode map caching");
644 case Opt_clear_cache
:
645 btrfs_set_and_info(root
, CLEAR_CACHE
,
646 "force clearing of disk cache");
648 case Opt_user_subvol_rm_allowed
:
649 btrfs_set_opt(info
->mount_opt
, USER_SUBVOL_RM_ALLOWED
);
651 case Opt_enospc_debug
:
652 btrfs_set_opt(info
->mount_opt
, ENOSPC_DEBUG
);
654 case Opt_noenospc_debug
:
655 btrfs_clear_opt(info
->mount_opt
, ENOSPC_DEBUG
);
658 btrfs_set_and_info(root
, AUTO_DEFRAG
,
659 "enabling auto defrag");
662 btrfs_clear_and_info(root
, AUTO_DEFRAG
,
663 "disabling auto defrag");
666 btrfs_info(root
->fs_info
, "enabling auto recovery");
667 btrfs_set_opt(info
->mount_opt
, RECOVERY
);
669 case Opt_skip_balance
:
670 btrfs_set_opt(info
->mount_opt
, SKIP_BALANCE
);
672 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
673 case Opt_check_integrity_including_extent_data
:
674 btrfs_info(root
->fs_info
,
675 "enabling check integrity including extent data");
676 btrfs_set_opt(info
->mount_opt
,
677 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
);
678 btrfs_set_opt(info
->mount_opt
, CHECK_INTEGRITY
);
680 case Opt_check_integrity
:
681 btrfs_info(root
->fs_info
, "enabling check integrity");
682 btrfs_set_opt(info
->mount_opt
, CHECK_INTEGRITY
);
684 case Opt_check_integrity_print_mask
:
685 ret
= match_int(&args
[0], &intarg
);
688 } else if (intarg
>= 0) {
689 info
->check_integrity_print_mask
= intarg
;
690 btrfs_info(root
->fs_info
, "check_integrity_print_mask 0x%x",
691 info
->check_integrity_print_mask
);
698 case Opt_check_integrity_including_extent_data
:
699 case Opt_check_integrity
:
700 case Opt_check_integrity_print_mask
:
701 btrfs_err(root
->fs_info
,
702 "support for check_integrity* not compiled in!");
706 case Opt_fatal_errors
:
707 if (strcmp(args
[0].from
, "panic") == 0)
708 btrfs_set_opt(info
->mount_opt
,
709 PANIC_ON_FATAL_ERROR
);
710 else if (strcmp(args
[0].from
, "bug") == 0)
711 btrfs_clear_opt(info
->mount_opt
,
712 PANIC_ON_FATAL_ERROR
);
718 case Opt_commit_interval
:
720 ret
= match_int(&args
[0], &intarg
);
722 btrfs_err(root
->fs_info
, "invalid commit interval");
728 btrfs_warn(root
->fs_info
, "excessive commit interval %d",
731 info
->commit_interval
= intarg
;
733 btrfs_info(root
->fs_info
, "using default commit interval %ds",
734 BTRFS_DEFAULT_COMMIT_INTERVAL
);
735 info
->commit_interval
= BTRFS_DEFAULT_COMMIT_INTERVAL
;
739 btrfs_info(root
->fs_info
, "unrecognized mount option '%s'", p
);
747 if (!ret
&& btrfs_test_opt(root
, SPACE_CACHE
))
748 btrfs_info(root
->fs_info
, "disk space caching is enabled");
754 * Parse mount options that are required early in the mount process.
756 * All other options will be parsed on much later in the mount process and
757 * only when we need to allocate a new super block.
759 static int btrfs_parse_early_options(const char *options
, fmode_t flags
,
760 void *holder
, char **subvol_name
, u64
*subvol_objectid
,
761 struct btrfs_fs_devices
**fs_devices
)
763 substring_t args
[MAX_OPT_ARGS
];
764 char *device_name
, *opts
, *orig
, *p
;
772 * strsep changes the string, duplicate it because parse_options
775 opts
= kstrdup(options
, GFP_KERNEL
);
780 while ((p
= strsep(&opts
, ",")) != NULL
) {
785 token
= match_token(p
, tokens
, args
);
789 *subvol_name
= match_strdup(&args
[0]);
796 num
= match_strdup(&args
[0]);
798 *subvol_objectid
= memparse(num
, NULL
);
800 /* we want the original fs_tree */
801 if (!*subvol_objectid
)
803 BTRFS_FS_TREE_OBJECTID
;
809 case Opt_subvolrootid
:
811 "BTRFS: 'subvolrootid' mount option is deprecated and has "
815 device_name
= match_strdup(&args
[0]);
820 error
= btrfs_scan_one_device(device_name
,
821 flags
, holder
, fs_devices
);
836 static struct dentry
*get_default_root(struct super_block
*sb
,
839 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
840 struct btrfs_root
*root
= fs_info
->tree_root
;
841 struct btrfs_root
*new_root
;
842 struct btrfs_dir_item
*di
;
843 struct btrfs_path
*path
;
844 struct btrfs_key location
;
850 * We have a specific subvol we want to mount, just setup location and
851 * go look up the root.
853 if (subvol_objectid
) {
854 location
.objectid
= subvol_objectid
;
855 location
.type
= BTRFS_ROOT_ITEM_KEY
;
856 location
.offset
= (u64
)-1;
860 path
= btrfs_alloc_path();
862 return ERR_PTR(-ENOMEM
);
863 path
->leave_spinning
= 1;
866 * Find the "default" dir item which points to the root item that we
867 * will mount by default if we haven't been given a specific subvolume
870 dir_id
= btrfs_super_root_dir(fs_info
->super_copy
);
871 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir_id
, "default", 7, 0);
873 btrfs_free_path(path
);
878 * Ok the default dir item isn't there. This is weird since
879 * it's always been there, but don't freak out, just try and
880 * mount to root most subvolume.
882 btrfs_free_path(path
);
883 dir_id
= BTRFS_FIRST_FREE_OBJECTID
;
884 new_root
= fs_info
->fs_root
;
888 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &location
);
889 btrfs_free_path(path
);
892 new_root
= btrfs_read_fs_root_no_name(fs_info
, &location
);
893 if (IS_ERR(new_root
))
894 return ERR_CAST(new_root
);
896 if (!(sb
->s_flags
& MS_RDONLY
)) {
898 down_read(&fs_info
->cleanup_work_sem
);
899 ret
= btrfs_orphan_cleanup(new_root
);
900 up_read(&fs_info
->cleanup_work_sem
);
905 dir_id
= btrfs_root_dirid(&new_root
->root_item
);
907 location
.objectid
= dir_id
;
908 location
.type
= BTRFS_INODE_ITEM_KEY
;
911 inode
= btrfs_iget(sb
, &location
, new_root
, &new);
913 return ERR_CAST(inode
);
916 * If we're just mounting the root most subvol put the inode and return
917 * a reference to the dentry. We will have already gotten a reference
918 * to the inode in btrfs_fill_super so we're good to go.
920 if (!new && d_inode(sb
->s_root
) == inode
) {
922 return dget(sb
->s_root
);
925 return d_obtain_root(inode
);
928 static int btrfs_fill_super(struct super_block
*sb
,
929 struct btrfs_fs_devices
*fs_devices
,
930 void *data
, int silent
)
933 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
934 struct btrfs_key key
;
937 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
938 sb
->s_magic
= BTRFS_SUPER_MAGIC
;
939 sb
->s_op
= &btrfs_super_ops
;
940 sb
->s_d_op
= &btrfs_dentry_operations
;
941 sb
->s_export_op
= &btrfs_export_ops
;
942 sb
->s_xattr
= btrfs_xattr_handlers
;
944 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
945 sb
->s_flags
|= MS_POSIXACL
;
947 sb
->s_flags
|= MS_I_VERSION
;
948 err
= open_ctree(sb
, fs_devices
, (char *)data
);
950 printk(KERN_ERR
"BTRFS: open_ctree failed\n");
954 key
.objectid
= BTRFS_FIRST_FREE_OBJECTID
;
955 key
.type
= BTRFS_INODE_ITEM_KEY
;
957 inode
= btrfs_iget(sb
, &key
, fs_info
->fs_root
, NULL
);
959 err
= PTR_ERR(inode
);
963 sb
->s_root
= d_make_root(inode
);
969 save_mount_options(sb
, data
);
970 cleancache_init_fs(sb
);
971 sb
->s_flags
|= MS_ACTIVE
;
975 close_ctree(fs_info
->tree_root
);
979 int btrfs_sync_fs(struct super_block
*sb
, int wait
)
981 struct btrfs_trans_handle
*trans
;
982 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
983 struct btrfs_root
*root
= fs_info
->tree_root
;
985 trace_btrfs_sync_fs(wait
);
988 filemap_flush(fs_info
->btree_inode
->i_mapping
);
992 btrfs_wait_ordered_roots(fs_info
, -1);
994 trans
= btrfs_attach_transaction_barrier(root
);
996 /* no transaction, don't bother */
997 if (PTR_ERR(trans
) == -ENOENT
) {
999 * Exit unless we have some pending changes
1000 * that need to go through commit
1002 if (fs_info
->pending_changes
== 0)
1005 * A non-blocking test if the fs is frozen. We must not
1006 * start a new transaction here otherwise a deadlock
1007 * happens. The pending operations are delayed to the
1008 * next commit after thawing.
1010 if (__sb_start_write(sb
, SB_FREEZE_WRITE
, false))
1011 __sb_end_write(sb
, SB_FREEZE_WRITE
);
1014 trans
= btrfs_start_transaction(root
, 0);
1017 return PTR_ERR(trans
);
1019 return btrfs_commit_transaction(trans
, root
);
1022 static int btrfs_show_options(struct seq_file
*seq
, struct dentry
*dentry
)
1024 struct btrfs_fs_info
*info
= btrfs_sb(dentry
->d_sb
);
1025 struct btrfs_root
*root
= info
->tree_root
;
1026 char *compress_type
;
1028 if (btrfs_test_opt(root
, DEGRADED
))
1029 seq_puts(seq
, ",degraded");
1030 if (btrfs_test_opt(root
, NODATASUM
))
1031 seq_puts(seq
, ",nodatasum");
1032 if (btrfs_test_opt(root
, NODATACOW
))
1033 seq_puts(seq
, ",nodatacow");
1034 if (btrfs_test_opt(root
, NOBARRIER
))
1035 seq_puts(seq
, ",nobarrier");
1036 if (info
->max_inline
!= BTRFS_DEFAULT_MAX_INLINE
)
1037 seq_printf(seq
, ",max_inline=%llu", info
->max_inline
);
1038 if (info
->alloc_start
!= 0)
1039 seq_printf(seq
, ",alloc_start=%llu", info
->alloc_start
);
1040 if (info
->thread_pool_size
!= min_t(unsigned long,
1041 num_online_cpus() + 2, 8))
1042 seq_printf(seq
, ",thread_pool=%d", info
->thread_pool_size
);
1043 if (btrfs_test_opt(root
, COMPRESS
)) {
1044 if (info
->compress_type
== BTRFS_COMPRESS_ZLIB
)
1045 compress_type
= "zlib";
1047 compress_type
= "lzo";
1048 if (btrfs_test_opt(root
, FORCE_COMPRESS
))
1049 seq_printf(seq
, ",compress-force=%s", compress_type
);
1051 seq_printf(seq
, ",compress=%s", compress_type
);
1053 if (btrfs_test_opt(root
, NOSSD
))
1054 seq_puts(seq
, ",nossd");
1055 if (btrfs_test_opt(root
, SSD_SPREAD
))
1056 seq_puts(seq
, ",ssd_spread");
1057 else if (btrfs_test_opt(root
, SSD
))
1058 seq_puts(seq
, ",ssd");
1059 if (btrfs_test_opt(root
, NOTREELOG
))
1060 seq_puts(seq
, ",notreelog");
1061 if (btrfs_test_opt(root
, FLUSHONCOMMIT
))
1062 seq_puts(seq
, ",flushoncommit");
1063 if (btrfs_test_opt(root
, DISCARD
))
1064 seq_puts(seq
, ",discard");
1065 if (!(root
->fs_info
->sb
->s_flags
& MS_POSIXACL
))
1066 seq_puts(seq
, ",noacl");
1067 if (btrfs_test_opt(root
, SPACE_CACHE
))
1068 seq_puts(seq
, ",space_cache");
1070 seq_puts(seq
, ",nospace_cache");
1071 if (btrfs_test_opt(root
, RESCAN_UUID_TREE
))
1072 seq_puts(seq
, ",rescan_uuid_tree");
1073 if (btrfs_test_opt(root
, CLEAR_CACHE
))
1074 seq_puts(seq
, ",clear_cache");
1075 if (btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
1076 seq_puts(seq
, ",user_subvol_rm_allowed");
1077 if (btrfs_test_opt(root
, ENOSPC_DEBUG
))
1078 seq_puts(seq
, ",enospc_debug");
1079 if (btrfs_test_opt(root
, AUTO_DEFRAG
))
1080 seq_puts(seq
, ",autodefrag");
1081 if (btrfs_test_opt(root
, INODE_MAP_CACHE
))
1082 seq_puts(seq
, ",inode_cache");
1083 if (btrfs_test_opt(root
, SKIP_BALANCE
))
1084 seq_puts(seq
, ",skip_balance");
1085 if (btrfs_test_opt(root
, RECOVERY
))
1086 seq_puts(seq
, ",recovery");
1087 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1088 if (btrfs_test_opt(root
, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
))
1089 seq_puts(seq
, ",check_int_data");
1090 else if (btrfs_test_opt(root
, CHECK_INTEGRITY
))
1091 seq_puts(seq
, ",check_int");
1092 if (info
->check_integrity_print_mask
)
1093 seq_printf(seq
, ",check_int_print_mask=%d",
1094 info
->check_integrity_print_mask
);
1096 if (info
->metadata_ratio
)
1097 seq_printf(seq
, ",metadata_ratio=%d",
1098 info
->metadata_ratio
);
1099 if (btrfs_test_opt(root
, PANIC_ON_FATAL_ERROR
))
1100 seq_puts(seq
, ",fatal_errors=panic");
1101 if (info
->commit_interval
!= BTRFS_DEFAULT_COMMIT_INTERVAL
)
1102 seq_printf(seq
, ",commit=%d", info
->commit_interval
);
1106 static int btrfs_test_super(struct super_block
*s
, void *data
)
1108 struct btrfs_fs_info
*p
= data
;
1109 struct btrfs_fs_info
*fs_info
= btrfs_sb(s
);
1111 return fs_info
->fs_devices
== p
->fs_devices
;
1114 static int btrfs_set_super(struct super_block
*s
, void *data
)
1116 int err
= set_anon_super(s
, data
);
1118 s
->s_fs_info
= data
;
1123 * subvolumes are identified by ino 256
1125 static inline int is_subvolume_inode(struct inode
*inode
)
1127 if (inode
&& inode
->i_ino
== BTRFS_FIRST_FREE_OBJECTID
)
1133 * This will strip out the subvol=%s argument for an argument string and add
1134 * subvolid=0 to make sure we get the actual tree root for path walking to the
1137 static char *setup_root_args(char *args
)
1139 unsigned len
= strlen(args
) + 2 + 1;
1140 char *src
, *dst
, *buf
;
1143 * We need the same args as before, but with this substitution:
1144 * s!subvol=[^,]+!subvolid=0!
1146 * Since the replacement string is up to 2 bytes longer than the
1147 * original, allocate strlen(args) + 2 + 1 bytes.
1150 src
= strstr(args
, "subvol=");
1151 /* This shouldn't happen, but just in case.. */
1155 buf
= dst
= kmalloc(len
, GFP_NOFS
);
1160 * If the subvol= arg is not at the start of the string,
1161 * copy whatever precedes it into buf.
1166 dst
+= strlen(args
);
1169 strcpy(dst
, "subvolid=0");
1170 dst
+= strlen("subvolid=0");
1173 * If there is a "," after the original subvol=... string,
1174 * copy that suffix into our buffer. Otherwise, we're done.
1176 src
= strchr(src
, ',');
1183 static struct dentry
*mount_subvol(const char *subvol_name
, int flags
,
1184 const char *device_name
, char *data
)
1186 struct dentry
*root
;
1187 struct vfsmount
*mnt
;
1190 newargs
= setup_root_args(data
);
1192 return ERR_PTR(-ENOMEM
);
1193 mnt
= vfs_kern_mount(&btrfs_fs_type
, flags
, device_name
,
1196 if (PTR_RET(mnt
) == -EBUSY
) {
1197 if (flags
& MS_RDONLY
) {
1198 mnt
= vfs_kern_mount(&btrfs_fs_type
, flags
& ~MS_RDONLY
, device_name
,
1202 mnt
= vfs_kern_mount(&btrfs_fs_type
, flags
| MS_RDONLY
, device_name
,
1206 return ERR_CAST(mnt
);
1209 r
= btrfs_remount(mnt
->mnt_sb
, &flags
, NULL
);
1211 /* FIXME: release vfsmount mnt ??*/
1221 return ERR_CAST(mnt
);
1223 root
= mount_subtree(mnt
, subvol_name
);
1225 if (!IS_ERR(root
) && !is_subvolume_inode(d_inode(root
))) {
1226 struct super_block
*s
= root
->d_sb
;
1228 root
= ERR_PTR(-EINVAL
);
1229 deactivate_locked_super(s
);
1230 printk(KERN_ERR
"BTRFS: '%s' is not a valid subvolume\n",
1237 static int parse_security_options(char *orig_opts
,
1238 struct security_mnt_opts
*sec_opts
)
1240 char *secdata
= NULL
;
1243 secdata
= alloc_secdata();
1246 ret
= security_sb_copy_data(orig_opts
, secdata
);
1248 free_secdata(secdata
);
1251 ret
= security_sb_parse_opts_str(secdata
, sec_opts
);
1252 free_secdata(secdata
);
1256 static int setup_security_options(struct btrfs_fs_info
*fs_info
,
1257 struct super_block
*sb
,
1258 struct security_mnt_opts
*sec_opts
)
1263 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1266 ret
= security_sb_set_mnt_opts(sb
, sec_opts
, 0, NULL
);
1270 #ifdef CONFIG_SECURITY
1271 if (!fs_info
->security_opts
.num_mnt_opts
) {
1272 /* first time security setup, copy sec_opts to fs_info */
1273 memcpy(&fs_info
->security_opts
, sec_opts
, sizeof(*sec_opts
));
1276 * Since SELinux(the only one supports security_mnt_opts) does
1277 * NOT support changing context during remount/mount same sb,
1278 * This must be the same or part of the same security options,
1281 security_free_mnt_opts(sec_opts
);
1288 * Find a superblock for the given device / mount point.
1290 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1291 * for multiple device setup. Make sure to keep it in sync.
1293 static struct dentry
*btrfs_mount(struct file_system_type
*fs_type
, int flags
,
1294 const char *device_name
, void *data
)
1296 struct block_device
*bdev
= NULL
;
1297 struct super_block
*s
;
1298 struct dentry
*root
;
1299 struct btrfs_fs_devices
*fs_devices
= NULL
;
1300 struct btrfs_fs_info
*fs_info
= NULL
;
1301 struct security_mnt_opts new_sec_opts
;
1302 fmode_t mode
= FMODE_READ
;
1303 char *subvol_name
= NULL
;
1304 u64 subvol_objectid
= 0;
1307 if (!(flags
& MS_RDONLY
))
1308 mode
|= FMODE_WRITE
;
1310 error
= btrfs_parse_early_options(data
, mode
, fs_type
,
1311 &subvol_name
, &subvol_objectid
,
1315 return ERR_PTR(error
);
1319 root
= mount_subvol(subvol_name
, flags
, device_name
, data
);
1324 security_init_mnt_opts(&new_sec_opts
);
1326 error
= parse_security_options(data
, &new_sec_opts
);
1328 return ERR_PTR(error
);
1331 error
= btrfs_scan_one_device(device_name
, mode
, fs_type
, &fs_devices
);
1333 goto error_sec_opts
;
1336 * Setup a dummy root and fs_info for test/set super. This is because
1337 * we don't actually fill this stuff out until open_ctree, but we need
1338 * it for searching for existing supers, so this lets us do that and
1339 * then open_ctree will properly initialize everything later.
1341 fs_info
= kzalloc(sizeof(struct btrfs_fs_info
), GFP_NOFS
);
1344 goto error_sec_opts
;
1347 fs_info
->fs_devices
= fs_devices
;
1349 fs_info
->super_copy
= kzalloc(BTRFS_SUPER_INFO_SIZE
, GFP_NOFS
);
1350 fs_info
->super_for_commit
= kzalloc(BTRFS_SUPER_INFO_SIZE
, GFP_NOFS
);
1351 security_init_mnt_opts(&fs_info
->security_opts
);
1352 if (!fs_info
->super_copy
|| !fs_info
->super_for_commit
) {
1357 error
= btrfs_open_devices(fs_devices
, mode
, fs_type
);
1361 if (!(flags
& MS_RDONLY
) && fs_devices
->rw_devices
== 0) {
1363 goto error_close_devices
;
1366 bdev
= fs_devices
->latest_bdev
;
1367 s
= sget(fs_type
, btrfs_test_super
, btrfs_set_super
, flags
| MS_NOSEC
,
1371 goto error_close_devices
;
1375 btrfs_close_devices(fs_devices
);
1376 free_fs_info(fs_info
);
1377 if ((flags
^ s
->s_flags
) & MS_RDONLY
)
1380 char b
[BDEVNAME_SIZE
];
1382 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
1383 btrfs_sb(s
)->bdev_holder
= fs_type
;
1384 error
= btrfs_fill_super(s
, fs_devices
, data
,
1385 flags
& MS_SILENT
? 1 : 0);
1388 root
= !error
? get_default_root(s
, subvol_objectid
) : ERR_PTR(error
);
1390 deactivate_locked_super(s
);
1391 error
= PTR_ERR(root
);
1392 goto error_sec_opts
;
1395 fs_info
= btrfs_sb(s
);
1396 error
= setup_security_options(fs_info
, s
, &new_sec_opts
);
1399 deactivate_locked_super(s
);
1400 goto error_sec_opts
;
1405 error_close_devices
:
1406 btrfs_close_devices(fs_devices
);
1408 free_fs_info(fs_info
);
1410 security_free_mnt_opts(&new_sec_opts
);
1411 return ERR_PTR(error
);
1414 static void btrfs_resize_thread_pool(struct btrfs_fs_info
*fs_info
,
1415 int new_pool_size
, int old_pool_size
)
1417 if (new_pool_size
== old_pool_size
)
1420 fs_info
->thread_pool_size
= new_pool_size
;
1422 btrfs_info(fs_info
, "resize thread pool %d -> %d",
1423 old_pool_size
, new_pool_size
);
1425 btrfs_workqueue_set_max(fs_info
->workers
, new_pool_size
);
1426 btrfs_workqueue_set_max(fs_info
->delalloc_workers
, new_pool_size
);
1427 btrfs_workqueue_set_max(fs_info
->submit_workers
, new_pool_size
);
1428 btrfs_workqueue_set_max(fs_info
->caching_workers
, new_pool_size
);
1429 btrfs_workqueue_set_max(fs_info
->endio_workers
, new_pool_size
);
1430 btrfs_workqueue_set_max(fs_info
->endio_meta_workers
, new_pool_size
);
1431 btrfs_workqueue_set_max(fs_info
->endio_meta_write_workers
,
1433 btrfs_workqueue_set_max(fs_info
->endio_write_workers
, new_pool_size
);
1434 btrfs_workqueue_set_max(fs_info
->endio_freespace_worker
, new_pool_size
);
1435 btrfs_workqueue_set_max(fs_info
->delayed_workers
, new_pool_size
);
1436 btrfs_workqueue_set_max(fs_info
->readahead_workers
, new_pool_size
);
1437 btrfs_workqueue_set_max(fs_info
->scrub_wr_completion_workers
,
1441 static inline void btrfs_remount_prepare(struct btrfs_fs_info
*fs_info
)
1443 set_bit(BTRFS_FS_STATE_REMOUNTING
, &fs_info
->fs_state
);
1446 static inline void btrfs_remount_begin(struct btrfs_fs_info
*fs_info
,
1447 unsigned long old_opts
, int flags
)
1449 if (btrfs_raw_test_opt(old_opts
, AUTO_DEFRAG
) &&
1450 (!btrfs_raw_test_opt(fs_info
->mount_opt
, AUTO_DEFRAG
) ||
1451 (flags
& MS_RDONLY
))) {
1452 /* wait for any defraggers to finish */
1453 wait_event(fs_info
->transaction_wait
,
1454 (atomic_read(&fs_info
->defrag_running
) == 0));
1455 if (flags
& MS_RDONLY
)
1456 sync_filesystem(fs_info
->sb
);
1460 static inline void btrfs_remount_cleanup(struct btrfs_fs_info
*fs_info
,
1461 unsigned long old_opts
)
1464 * We need cleanup all defragable inodes if the autodefragment is
1465 * close or the fs is R/O.
1467 if (btrfs_raw_test_opt(old_opts
, AUTO_DEFRAG
) &&
1468 (!btrfs_raw_test_opt(fs_info
->mount_opt
, AUTO_DEFRAG
) ||
1469 (fs_info
->sb
->s_flags
& MS_RDONLY
))) {
1470 btrfs_cleanup_defrag_inodes(fs_info
);
1473 clear_bit(BTRFS_FS_STATE_REMOUNTING
, &fs_info
->fs_state
);
1476 static int btrfs_remount(struct super_block
*sb
, int *flags
, char *data
)
1478 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1479 struct btrfs_root
*root
= fs_info
->tree_root
;
1480 unsigned old_flags
= sb
->s_flags
;
1481 unsigned long old_opts
= fs_info
->mount_opt
;
1482 unsigned long old_compress_type
= fs_info
->compress_type
;
1483 u64 old_max_inline
= fs_info
->max_inline
;
1484 u64 old_alloc_start
= fs_info
->alloc_start
;
1485 int old_thread_pool_size
= fs_info
->thread_pool_size
;
1486 unsigned int old_metadata_ratio
= fs_info
->metadata_ratio
;
1489 sync_filesystem(sb
);
1490 btrfs_remount_prepare(fs_info
);
1493 struct security_mnt_opts new_sec_opts
;
1495 security_init_mnt_opts(&new_sec_opts
);
1496 ret
= parse_security_options(data
, &new_sec_opts
);
1499 ret
= setup_security_options(fs_info
, sb
,
1502 security_free_mnt_opts(&new_sec_opts
);
1507 ret
= btrfs_parse_options(root
, data
);
1513 btrfs_remount_begin(fs_info
, old_opts
, *flags
);
1514 btrfs_resize_thread_pool(fs_info
,
1515 fs_info
->thread_pool_size
, old_thread_pool_size
);
1517 if ((*flags
& MS_RDONLY
) == (sb
->s_flags
& MS_RDONLY
))
1520 if (*flags
& MS_RDONLY
) {
1522 * this also happens on 'umount -rf' or on shutdown, when
1523 * the filesystem is busy.
1525 cancel_work_sync(&fs_info
->async_reclaim_work
);
1527 /* wait for the uuid_scan task to finish */
1528 down(&fs_info
->uuid_tree_rescan_sem
);
1529 /* avoid complains from lockdep et al. */
1530 up(&fs_info
->uuid_tree_rescan_sem
);
1532 sb
->s_flags
|= MS_RDONLY
;
1534 btrfs_dev_replace_suspend_for_unmount(fs_info
);
1535 btrfs_scrub_cancel(fs_info
);
1536 btrfs_pause_balance(fs_info
);
1538 ret
= btrfs_commit_super(root
);
1542 if (test_bit(BTRFS_FS_STATE_ERROR
, &root
->fs_info
->fs_state
)) {
1544 "Remounting read-write after error is not allowed");
1548 if (fs_info
->fs_devices
->rw_devices
== 0) {
1553 if (fs_info
->fs_devices
->missing_devices
>
1554 fs_info
->num_tolerated_disk_barrier_failures
&&
1555 !(*flags
& MS_RDONLY
)) {
1557 "too many missing devices, writeable remount is not allowed");
1562 if (btrfs_super_log_root(fs_info
->super_copy
) != 0) {
1567 ret
= btrfs_cleanup_fs_roots(fs_info
);
1571 /* recover relocation */
1572 mutex_lock(&fs_info
->cleaner_mutex
);
1573 ret
= btrfs_recover_relocation(root
);
1574 mutex_unlock(&fs_info
->cleaner_mutex
);
1578 ret
= btrfs_resume_balance_async(fs_info
);
1582 ret
= btrfs_resume_dev_replace_async(fs_info
);
1584 btrfs_warn(fs_info
, "failed to resume dev_replace");
1588 if (!fs_info
->uuid_root
) {
1589 btrfs_info(fs_info
, "creating UUID tree");
1590 ret
= btrfs_create_uuid_tree(fs_info
);
1592 btrfs_warn(fs_info
, "failed to create the UUID tree %d", ret
);
1596 sb
->s_flags
&= ~MS_RDONLY
;
1599 wake_up_process(fs_info
->transaction_kthread
);
1600 btrfs_remount_cleanup(fs_info
, old_opts
);
1604 /* We've hit an error - don't reset MS_RDONLY */
1605 if (sb
->s_flags
& MS_RDONLY
)
1606 old_flags
|= MS_RDONLY
;
1607 sb
->s_flags
= old_flags
;
1608 fs_info
->mount_opt
= old_opts
;
1609 fs_info
->compress_type
= old_compress_type
;
1610 fs_info
->max_inline
= old_max_inline
;
1611 mutex_lock(&fs_info
->chunk_mutex
);
1612 fs_info
->alloc_start
= old_alloc_start
;
1613 mutex_unlock(&fs_info
->chunk_mutex
);
1614 btrfs_resize_thread_pool(fs_info
,
1615 old_thread_pool_size
, fs_info
->thread_pool_size
);
1616 fs_info
->metadata_ratio
= old_metadata_ratio
;
1617 btrfs_remount_cleanup(fs_info
, old_opts
);
1621 /* Used to sort the devices by max_avail(descending sort) */
1622 static int btrfs_cmp_device_free_bytes(const void *dev_info1
,
1623 const void *dev_info2
)
1625 if (((struct btrfs_device_info
*)dev_info1
)->max_avail
>
1626 ((struct btrfs_device_info
*)dev_info2
)->max_avail
)
1628 else if (((struct btrfs_device_info
*)dev_info1
)->max_avail
<
1629 ((struct btrfs_device_info
*)dev_info2
)->max_avail
)
1636 * sort the devices by max_avail, in which max free extent size of each device
1637 * is stored.(Descending Sort)
1639 static inline void btrfs_descending_sort_devices(
1640 struct btrfs_device_info
*devices
,
1643 sort(devices
, nr_devices
, sizeof(struct btrfs_device_info
),
1644 btrfs_cmp_device_free_bytes
, NULL
);
1648 * The helper to calc the free space on the devices that can be used to store
1651 static int btrfs_calc_avail_data_space(struct btrfs_root
*root
, u64
*free_bytes
)
1653 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1654 struct btrfs_device_info
*devices_info
;
1655 struct btrfs_fs_devices
*fs_devices
= fs_info
->fs_devices
;
1656 struct btrfs_device
*device
;
1661 u64 min_stripe_size
;
1662 int min_stripes
= 1, num_stripes
= 1;
1663 int i
= 0, nr_devices
;
1667 * We aren't under the device list lock, so this is racey-ish, but good
1668 * enough for our purposes.
1670 nr_devices
= fs_info
->fs_devices
->open_devices
;
1673 nr_devices
= fs_info
->fs_devices
->open_devices
;
1681 devices_info
= kmalloc_array(nr_devices
, sizeof(*devices_info
),
1686 /* calc min stripe number for data space alloction */
1687 type
= btrfs_get_alloc_profile(root
, 1);
1688 if (type
& BTRFS_BLOCK_GROUP_RAID0
) {
1690 num_stripes
= nr_devices
;
1691 } else if (type
& BTRFS_BLOCK_GROUP_RAID1
) {
1694 } else if (type
& BTRFS_BLOCK_GROUP_RAID10
) {
1699 if (type
& BTRFS_BLOCK_GROUP_DUP
)
1700 min_stripe_size
= 2 * BTRFS_STRIPE_LEN
;
1702 min_stripe_size
= BTRFS_STRIPE_LEN
;
1704 if (fs_info
->alloc_start
)
1705 mutex_lock(&fs_devices
->device_list_mutex
);
1707 list_for_each_entry_rcu(device
, &fs_devices
->devices
, dev_list
) {
1708 if (!device
->in_fs_metadata
|| !device
->bdev
||
1709 device
->is_tgtdev_for_dev_replace
)
1712 if (i
>= nr_devices
)
1715 avail_space
= device
->total_bytes
- device
->bytes_used
;
1717 /* align with stripe_len */
1718 avail_space
= div_u64(avail_space
, BTRFS_STRIPE_LEN
);
1719 avail_space
*= BTRFS_STRIPE_LEN
;
1722 * In order to avoid overwritting the superblock on the drive,
1723 * btrfs starts at an offset of at least 1MB when doing chunk
1726 skip_space
= 1024 * 1024;
1728 /* user can set the offset in fs_info->alloc_start. */
1729 if (fs_info
->alloc_start
&&
1730 fs_info
->alloc_start
+ BTRFS_STRIPE_LEN
<=
1731 device
->total_bytes
) {
1733 skip_space
= max(fs_info
->alloc_start
, skip_space
);
1736 * btrfs can not use the free space in
1737 * [0, skip_space - 1], we must subtract it from the
1738 * total. In order to implement it, we account the used
1739 * space in this range first.
1741 ret
= btrfs_account_dev_extents_size(device
, 0,
1745 kfree(devices_info
);
1746 mutex_unlock(&fs_devices
->device_list_mutex
);
1752 /* calc the free space in [0, skip_space - 1] */
1753 skip_space
-= used_space
;
1757 * we can use the free space in [0, skip_space - 1], subtract
1758 * it from the total.
1760 if (avail_space
&& avail_space
>= skip_space
)
1761 avail_space
-= skip_space
;
1765 if (avail_space
< min_stripe_size
)
1768 devices_info
[i
].dev
= device
;
1769 devices_info
[i
].max_avail
= avail_space
;
1774 if (fs_info
->alloc_start
)
1775 mutex_unlock(&fs_devices
->device_list_mutex
);
1779 btrfs_descending_sort_devices(devices_info
, nr_devices
);
1783 while (nr_devices
>= min_stripes
) {
1784 if (num_stripes
> nr_devices
)
1785 num_stripes
= nr_devices
;
1787 if (devices_info
[i
].max_avail
>= min_stripe_size
) {
1791 avail_space
+= devices_info
[i
].max_avail
* num_stripes
;
1792 alloc_size
= devices_info
[i
].max_avail
;
1793 for (j
= i
+ 1 - num_stripes
; j
<= i
; j
++)
1794 devices_info
[j
].max_avail
-= alloc_size
;
1800 kfree(devices_info
);
1801 *free_bytes
= avail_space
;
1806 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1808 * If there's a redundant raid level at DATA block groups, use the respective
1809 * multiplier to scale the sizes.
1811 * Unused device space usage is based on simulating the chunk allocator
1812 * algorithm that respects the device sizes, order of allocations and the
1813 * 'alloc_start' value, this is a close approximation of the actual use but
1814 * there are other factors that may change the result (like a new metadata
1817 * FIXME: not accurate for mixed block groups, total and free/used are ok,
1818 * available appears slightly larger.
1820 static int btrfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1822 struct btrfs_fs_info
*fs_info
= btrfs_sb(dentry
->d_sb
);
1823 struct btrfs_super_block
*disk_super
= fs_info
->super_copy
;
1824 struct list_head
*head
= &fs_info
->space_info
;
1825 struct btrfs_space_info
*found
;
1827 u64 total_free_data
= 0;
1828 int bits
= dentry
->d_sb
->s_blocksize_bits
;
1829 __be32
*fsid
= (__be32
*)fs_info
->fsid
;
1830 unsigned factor
= 1;
1831 struct btrfs_block_rsv
*block_rsv
= &fs_info
->global_block_rsv
;
1835 * holding chunk_muext to avoid allocating new chunks, holding
1836 * device_list_mutex to avoid the device being removed
1839 list_for_each_entry_rcu(found
, head
, list
) {
1840 if (found
->flags
& BTRFS_BLOCK_GROUP_DATA
) {
1843 total_free_data
+= found
->disk_total
- found
->disk_used
;
1845 btrfs_account_ro_block_groups_free_space(found
);
1847 for (i
= 0; i
< BTRFS_NR_RAID_TYPES
; i
++) {
1848 if (!list_empty(&found
->block_groups
[i
])) {
1850 case BTRFS_RAID_DUP
:
1851 case BTRFS_RAID_RAID1
:
1852 case BTRFS_RAID_RAID10
:
1859 total_used
+= found
->disk_used
;
1864 buf
->f_blocks
= div_u64(btrfs_super_total_bytes(disk_super
), factor
);
1865 buf
->f_blocks
>>= bits
;
1866 buf
->f_bfree
= buf
->f_blocks
- (div_u64(total_used
, factor
) >> bits
);
1868 /* Account global block reserve as used, it's in logical size already */
1869 spin_lock(&block_rsv
->lock
);
1870 buf
->f_bfree
-= block_rsv
->size
>> bits
;
1871 spin_unlock(&block_rsv
->lock
);
1873 buf
->f_bavail
= div_u64(total_free_data
, factor
);
1874 ret
= btrfs_calc_avail_data_space(fs_info
->tree_root
, &total_free_data
);
1877 buf
->f_bavail
+= div_u64(total_free_data
, factor
);
1878 buf
->f_bavail
= buf
->f_bavail
>> bits
;
1880 buf
->f_type
= BTRFS_SUPER_MAGIC
;
1881 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1882 buf
->f_namelen
= BTRFS_NAME_LEN
;
1884 /* We treat it as constant endianness (it doesn't matter _which_)
1885 because we want the fsid to come out the same whether mounted
1886 on a big-endian or little-endian host */
1887 buf
->f_fsid
.val
[0] = be32_to_cpu(fsid
[0]) ^ be32_to_cpu(fsid
[2]);
1888 buf
->f_fsid
.val
[1] = be32_to_cpu(fsid
[1]) ^ be32_to_cpu(fsid
[3]);
1889 /* Mask in the root object ID too, to disambiguate subvols */
1890 buf
->f_fsid
.val
[0] ^= BTRFS_I(d_inode(dentry
))->root
->objectid
>> 32;
1891 buf
->f_fsid
.val
[1] ^= BTRFS_I(d_inode(dentry
))->root
->objectid
;
1896 static void btrfs_kill_super(struct super_block
*sb
)
1898 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1899 kill_anon_super(sb
);
1900 free_fs_info(fs_info
);
1903 static struct file_system_type btrfs_fs_type
= {
1904 .owner
= THIS_MODULE
,
1906 .mount
= btrfs_mount
,
1907 .kill_sb
= btrfs_kill_super
,
1908 .fs_flags
= FS_REQUIRES_DEV
| FS_BINARY_MOUNTDATA
,
1910 MODULE_ALIAS_FS("btrfs");
1912 static int btrfs_control_open(struct inode
*inode
, struct file
*file
)
1915 * The control file's private_data is used to hold the
1916 * transaction when it is started and is used to keep
1917 * track of whether a transaction is already in progress.
1919 file
->private_data
= NULL
;
1924 * used by btrfsctl to scan devices when no FS is mounted
1926 static long btrfs_control_ioctl(struct file
*file
, unsigned int cmd
,
1929 struct btrfs_ioctl_vol_args
*vol
;
1930 struct btrfs_fs_devices
*fs_devices
;
1933 if (!capable(CAP_SYS_ADMIN
))
1936 vol
= memdup_user((void __user
*)arg
, sizeof(*vol
));
1938 return PTR_ERR(vol
);
1941 case BTRFS_IOC_SCAN_DEV
:
1942 ret
= btrfs_scan_one_device(vol
->name
, FMODE_READ
,
1943 &btrfs_fs_type
, &fs_devices
);
1945 case BTRFS_IOC_DEVICES_READY
:
1946 ret
= btrfs_scan_one_device(vol
->name
, FMODE_READ
,
1947 &btrfs_fs_type
, &fs_devices
);
1950 ret
= !(fs_devices
->num_devices
== fs_devices
->total_devices
);
1958 static int btrfs_freeze(struct super_block
*sb
)
1960 struct btrfs_trans_handle
*trans
;
1961 struct btrfs_root
*root
= btrfs_sb(sb
)->tree_root
;
1963 trans
= btrfs_attach_transaction_barrier(root
);
1964 if (IS_ERR(trans
)) {
1965 /* no transaction, don't bother */
1966 if (PTR_ERR(trans
) == -ENOENT
)
1968 return PTR_ERR(trans
);
1970 return btrfs_commit_transaction(trans
, root
);
1973 static int btrfs_show_devname(struct seq_file
*m
, struct dentry
*root
)
1975 struct btrfs_fs_info
*fs_info
= btrfs_sb(root
->d_sb
);
1976 struct btrfs_fs_devices
*cur_devices
;
1977 struct btrfs_device
*dev
, *first_dev
= NULL
;
1978 struct list_head
*head
;
1979 struct rcu_string
*name
;
1981 mutex_lock(&fs_info
->fs_devices
->device_list_mutex
);
1982 cur_devices
= fs_info
->fs_devices
;
1983 while (cur_devices
) {
1984 head
= &cur_devices
->devices
;
1985 list_for_each_entry(dev
, head
, dev_list
) {
1990 if (!first_dev
|| dev
->devid
< first_dev
->devid
)
1993 cur_devices
= cur_devices
->seed
;
1998 name
= rcu_dereference(first_dev
->name
);
1999 seq_escape(m
, name
->str
, " \t\n\\");
2004 mutex_unlock(&fs_info
->fs_devices
->device_list_mutex
);
2008 static const struct super_operations btrfs_super_ops
= {
2009 .drop_inode
= btrfs_drop_inode
,
2010 .evict_inode
= btrfs_evict_inode
,
2011 .put_super
= btrfs_put_super
,
2012 .sync_fs
= btrfs_sync_fs
,
2013 .show_options
= btrfs_show_options
,
2014 .show_devname
= btrfs_show_devname
,
2015 .write_inode
= btrfs_write_inode
,
2016 .alloc_inode
= btrfs_alloc_inode
,
2017 .destroy_inode
= btrfs_destroy_inode
,
2018 .statfs
= btrfs_statfs
,
2019 .remount_fs
= btrfs_remount
,
2020 .freeze_fs
= btrfs_freeze
,
2023 static const struct file_operations btrfs_ctl_fops
= {
2024 .open
= btrfs_control_open
,
2025 .unlocked_ioctl
= btrfs_control_ioctl
,
2026 .compat_ioctl
= btrfs_control_ioctl
,
2027 .owner
= THIS_MODULE
,
2028 .llseek
= noop_llseek
,
2031 static struct miscdevice btrfs_misc
= {
2032 .minor
= BTRFS_MINOR
,
2033 .name
= "btrfs-control",
2034 .fops
= &btrfs_ctl_fops
2037 MODULE_ALIAS_MISCDEV(BTRFS_MINOR
);
2038 MODULE_ALIAS("devname:btrfs-control");
2040 static int btrfs_interface_init(void)
2042 return misc_register(&btrfs_misc
);
2045 static void btrfs_interface_exit(void)
2047 if (misc_deregister(&btrfs_misc
) < 0)
2048 printk(KERN_INFO
"BTRFS: misc_deregister failed for control device\n");
2051 static void btrfs_print_info(void)
2053 printk(KERN_INFO
"Btrfs loaded"
2054 #ifdef CONFIG_BTRFS_DEBUG
2057 #ifdef CONFIG_BTRFS_ASSERT
2060 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2061 ", integrity-checker=on"
2066 static int btrfs_run_sanity_tests(void)
2070 ret
= btrfs_init_test_fs();
2074 ret
= btrfs_test_free_space_cache();
2077 ret
= btrfs_test_extent_buffer_operations();
2080 ret
= btrfs_test_extent_io();
2083 ret
= btrfs_test_inodes();
2086 ret
= btrfs_test_qgroups();
2088 btrfs_destroy_test_fs();
2092 static int __init
init_btrfs_fs(void)
2096 err
= btrfs_hash_init();
2102 err
= btrfs_init_sysfs();
2106 btrfs_init_compress();
2108 err
= btrfs_init_cachep();
2112 err
= extent_io_init();
2116 err
= extent_map_init();
2118 goto free_extent_io
;
2120 err
= ordered_data_init();
2122 goto free_extent_map
;
2124 err
= btrfs_delayed_inode_init();
2126 goto free_ordered_data
;
2128 err
= btrfs_auto_defrag_init();
2130 goto free_delayed_inode
;
2132 err
= btrfs_delayed_ref_init();
2134 goto free_auto_defrag
;
2136 err
= btrfs_prelim_ref_init();
2138 goto free_delayed_ref
;
2140 err
= btrfs_end_io_wq_init();
2142 goto free_prelim_ref
;
2144 err
= btrfs_interface_init();
2146 goto free_end_io_wq
;
2148 btrfs_init_lockdep();
2152 err
= btrfs_run_sanity_tests();
2154 goto unregister_ioctl
;
2156 err
= register_filesystem(&btrfs_fs_type
);
2158 goto unregister_ioctl
;
2163 btrfs_interface_exit();
2165 btrfs_end_io_wq_exit();
2167 btrfs_prelim_ref_exit();
2169 btrfs_delayed_ref_exit();
2171 btrfs_auto_defrag_exit();
2173 btrfs_delayed_inode_exit();
2175 ordered_data_exit();
2181 btrfs_destroy_cachep();
2183 btrfs_exit_compress();
2190 static void __exit
exit_btrfs_fs(void)
2192 btrfs_destroy_cachep();
2193 btrfs_delayed_ref_exit();
2194 btrfs_auto_defrag_exit();
2195 btrfs_delayed_inode_exit();
2196 btrfs_prelim_ref_exit();
2197 ordered_data_exit();
2200 btrfs_interface_exit();
2201 btrfs_end_io_wq_exit();
2202 unregister_filesystem(&btrfs_fs_type
);
2204 btrfs_cleanup_fs_uuids();
2205 btrfs_exit_compress();
2209 late_initcall(init_btrfs_fs
);
2210 module_exit(exit_btrfs_fs
)
2212 MODULE_LICENSE("GPL");