1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2007 Oracle. All rights reserved.
6 #include <linux/blkdev.h>
7 #include <linux/module.h>
9 #include <linux/pagemap.h>
10 #include <linux/highmem.h>
11 #include <linux/time.h>
12 #include <linux/init.h>
13 #include <linux/seq_file.h>
14 #include <linux/string.h>
15 #include <linux/backing-dev.h>
16 #include <linux/mount.h>
17 #include <linux/writeback.h>
18 #include <linux/statfs.h>
19 #include <linux/compat.h>
20 #include <linux/parser.h>
21 #include <linux/ctype.h>
22 #include <linux/namei.h>
23 #include <linux/miscdevice.h>
24 #include <linux/magic.h>
25 #include <linux/slab.h>
26 #include <linux/cleancache.h>
27 #include <linux/ratelimit.h>
28 #include <linux/crc32c.h>
29 #include <linux/btrfs.h>
30 #include "delayed-inode.h"
33 #include "transaction.h"
34 #include "btrfs_inode.h"
35 #include "print-tree.h"
40 #include "compression.h"
41 #include "rcu-string.h"
42 #include "dev-replace.h"
43 #include "free-space-cache.h"
45 #include "space-info.h"
47 #include "tests/btrfs-tests.h"
48 #include "block-group.h"
52 #define CREATE_TRACE_POINTS
53 #include <trace/events/btrfs.h>
55 static const struct super_operations btrfs_super_ops
;
58 * Types for mounting the default subvolume and a subvolume explicitly
59 * requested by subvol=/path. That way the callchain is straightforward and we
60 * don't have to play tricks with the mount options and recursive calls to
63 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
65 static struct file_system_type btrfs_fs_type
;
66 static struct file_system_type btrfs_root_fs_type
;
68 static int btrfs_remount(struct super_block
*sb
, int *flags
, char *data
);
70 const char * __attribute_const__
btrfs_decode_error(int errno
)
72 char *errstr
= "unknown";
76 errstr
= "IO failure";
79 errstr
= "Out of memory";
82 errstr
= "Readonly filesystem";
85 errstr
= "Object already exists";
88 errstr
= "No space left";
91 errstr
= "No such entry";
99 * __btrfs_handle_fs_error decodes expected errors from the caller and
100 * invokes the appropriate error response.
103 void __btrfs_handle_fs_error(struct btrfs_fs_info
*fs_info
, const char *function
,
104 unsigned int line
, int errno
, const char *fmt
, ...)
106 struct super_block
*sb
= fs_info
->sb
;
112 * Special case: if the error is EROFS, and we're already
113 * under SB_RDONLY, then it is safe here.
115 if (errno
== -EROFS
&& sb_rdonly(sb
))
119 errstr
= btrfs_decode_error(errno
);
121 struct va_format vaf
;
128 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
129 sb
->s_id
, function
, line
, errno
, errstr
, &vaf
);
132 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
133 sb
->s_id
, function
, line
, errno
, errstr
);
138 * Today we only save the error info to memory. Long term we'll
139 * also send it down to the disk
141 set_bit(BTRFS_FS_STATE_ERROR
, &fs_info
->fs_state
);
143 /* Don't go through full error handling during mount */
144 if (!(sb
->s_flags
& SB_BORN
))
150 btrfs_discard_stop(fs_info
);
152 /* btrfs handle error by forcing the filesystem readonly */
153 sb
->s_flags
|= SB_RDONLY
;
154 btrfs_info(fs_info
, "forced readonly");
156 * Note that a running device replace operation is not canceled here
157 * although there is no way to update the progress. It would add the
158 * risk of a deadlock, therefore the canceling is omitted. The only
159 * penalty is that some I/O remains active until the procedure
160 * completes. The next time when the filesystem is mounted writable
161 * again, the device replace operation continues.
166 static const char * const logtypes
[] = {
179 * Use one ratelimit state per log level so that a flood of less important
180 * messages doesn't cause more important ones to be dropped.
182 static struct ratelimit_state printk_limits
[] = {
183 RATELIMIT_STATE_INIT(printk_limits
[0], DEFAULT_RATELIMIT_INTERVAL
, 100),
184 RATELIMIT_STATE_INIT(printk_limits
[1], DEFAULT_RATELIMIT_INTERVAL
, 100),
185 RATELIMIT_STATE_INIT(printk_limits
[2], DEFAULT_RATELIMIT_INTERVAL
, 100),
186 RATELIMIT_STATE_INIT(printk_limits
[3], DEFAULT_RATELIMIT_INTERVAL
, 100),
187 RATELIMIT_STATE_INIT(printk_limits
[4], DEFAULT_RATELIMIT_INTERVAL
, 100),
188 RATELIMIT_STATE_INIT(printk_limits
[5], DEFAULT_RATELIMIT_INTERVAL
, 100),
189 RATELIMIT_STATE_INIT(printk_limits
[6], DEFAULT_RATELIMIT_INTERVAL
, 100),
190 RATELIMIT_STATE_INIT(printk_limits
[7], DEFAULT_RATELIMIT_INTERVAL
, 100),
193 void __cold
btrfs_printk(const struct btrfs_fs_info
*fs_info
, const char *fmt
, ...)
195 char lvl
[PRINTK_MAX_SINGLE_HEADER_LEN
+ 1] = "\0";
196 struct va_format vaf
;
199 const char *type
= logtypes
[4];
200 struct ratelimit_state
*ratelimit
= &printk_limits
[4];
204 while ((kern_level
= printk_get_level(fmt
)) != 0) {
205 size_t size
= printk_skip_level(fmt
) - fmt
;
207 if (kern_level
>= '0' && kern_level
<= '7') {
208 memcpy(lvl
, fmt
, size
);
210 type
= logtypes
[kern_level
- '0'];
211 ratelimit
= &printk_limits
[kern_level
- '0'];
219 if (__ratelimit(ratelimit
))
220 printk("%sBTRFS %s (device %s): %pV\n", lvl
, type
,
221 fs_info
? fs_info
->sb
->s_id
: "<unknown>", &vaf
);
228 * We only mark the transaction aborted and then set the file system read-only.
229 * This will prevent new transactions from starting or trying to join this
232 * This means that error recovery at the call site is limited to freeing
233 * any local memory allocations and passing the error code up without
234 * further cleanup. The transaction should complete as it normally would
235 * in the call path but will return -EIO.
237 * We'll complete the cleanup in btrfs_end_transaction and
238 * btrfs_commit_transaction.
241 void __btrfs_abort_transaction(struct btrfs_trans_handle
*trans
,
242 const char *function
,
243 unsigned int line
, int errno
)
245 struct btrfs_fs_info
*fs_info
= trans
->fs_info
;
247 WRITE_ONCE(trans
->aborted
, errno
);
248 /* Nothing used. The other threads that have joined this
249 * transaction may be able to continue. */
250 if (!trans
->dirty
&& list_empty(&trans
->new_bgs
)) {
253 errstr
= btrfs_decode_error(errno
);
255 "%s:%d: Aborting unused transaction(%s).",
256 function
, line
, errstr
);
259 WRITE_ONCE(trans
->transaction
->aborted
, errno
);
260 /* Wake up anybody who may be waiting on this transaction */
261 wake_up(&fs_info
->transaction_wait
);
262 wake_up(&fs_info
->transaction_blocked_wait
);
263 __btrfs_handle_fs_error(fs_info
, function
, line
, errno
, NULL
);
266 * __btrfs_panic decodes unexpected, fatal errors from the caller,
267 * issues an alert, and either panics or BUGs, depending on mount options.
270 void __btrfs_panic(struct btrfs_fs_info
*fs_info
, const char *function
,
271 unsigned int line
, int errno
, const char *fmt
, ...)
273 char *s_id
= "<unknown>";
275 struct va_format vaf
= { .fmt
= fmt
};
279 s_id
= fs_info
->sb
->s_id
;
284 errstr
= btrfs_decode_error(errno
);
285 if (fs_info
&& (btrfs_test_opt(fs_info
, PANIC_ON_FATAL_ERROR
)))
286 panic(KERN_CRIT
"BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
287 s_id
, function
, line
, &vaf
, errno
, errstr
);
289 btrfs_crit(fs_info
, "panic in %s:%d: %pV (errno=%d %s)",
290 function
, line
, &vaf
, errno
, errstr
);
292 /* Caller calls BUG() */
295 static void btrfs_put_super(struct super_block
*sb
)
297 close_ctree(btrfs_sb(sb
));
306 Opt_compress_force_type
,
311 Opt_flushoncommit
, Opt_noflushoncommit
,
312 Opt_inode_cache
, Opt_noinode_cache
,
314 Opt_barrier
, Opt_nobarrier
,
315 Opt_datacow
, Opt_nodatacow
,
316 Opt_datasum
, Opt_nodatasum
,
317 Opt_defrag
, Opt_nodefrag
,
318 Opt_discard
, Opt_nodiscard
,
323 Opt_rescan_uuid_tree
,
325 Opt_space_cache
, Opt_no_space_cache
,
326 Opt_space_cache_version
,
328 Opt_ssd_spread
, Opt_nossd_spread
,
333 Opt_treelog
, Opt_notreelog
,
335 Opt_user_subvol_rm_allowed
,
337 /* Deprecated options */
342 /* Debugging options */
344 Opt_check_integrity_including_extent_data
,
345 Opt_check_integrity_print_mask
,
346 Opt_enospc_debug
, Opt_noenospc_debug
,
347 #ifdef CONFIG_BTRFS_DEBUG
348 Opt_fragment_data
, Opt_fragment_metadata
, Opt_fragment_all
,
350 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
356 static const match_table_t tokens
= {
358 {Opt_noacl
, "noacl"},
359 {Opt_clear_cache
, "clear_cache"},
360 {Opt_commit_interval
, "commit=%u"},
361 {Opt_compress
, "compress"},
362 {Opt_compress_type
, "compress=%s"},
363 {Opt_compress_force
, "compress-force"},
364 {Opt_compress_force_type
, "compress-force=%s"},
365 {Opt_degraded
, "degraded"},
366 {Opt_device
, "device=%s"},
367 {Opt_fatal_errors
, "fatal_errors=%s"},
368 {Opt_flushoncommit
, "flushoncommit"},
369 {Opt_noflushoncommit
, "noflushoncommit"},
370 {Opt_inode_cache
, "inode_cache"},
371 {Opt_noinode_cache
, "noinode_cache"},
372 {Opt_max_inline
, "max_inline=%s"},
373 {Opt_barrier
, "barrier"},
374 {Opt_nobarrier
, "nobarrier"},
375 {Opt_datacow
, "datacow"},
376 {Opt_nodatacow
, "nodatacow"},
377 {Opt_datasum
, "datasum"},
378 {Opt_nodatasum
, "nodatasum"},
379 {Opt_defrag
, "autodefrag"},
380 {Opt_nodefrag
, "noautodefrag"},
381 {Opt_discard
, "discard"},
382 {Opt_discard_mode
, "discard=%s"},
383 {Opt_nodiscard
, "nodiscard"},
384 {Opt_nologreplay
, "nologreplay"},
385 {Opt_norecovery
, "norecovery"},
386 {Opt_ratio
, "metadata_ratio=%u"},
387 {Opt_rescan_uuid_tree
, "rescan_uuid_tree"},
388 {Opt_skip_balance
, "skip_balance"},
389 {Opt_space_cache
, "space_cache"},
390 {Opt_no_space_cache
, "nospace_cache"},
391 {Opt_space_cache_version
, "space_cache=%s"},
393 {Opt_nossd
, "nossd"},
394 {Opt_ssd_spread
, "ssd_spread"},
395 {Opt_nossd_spread
, "nossd_spread"},
396 {Opt_subvol
, "subvol=%s"},
397 {Opt_subvol_empty
, "subvol="},
398 {Opt_subvolid
, "subvolid=%s"},
399 {Opt_thread_pool
, "thread_pool=%u"},
400 {Opt_treelog
, "treelog"},
401 {Opt_notreelog
, "notreelog"},
402 {Opt_usebackuproot
, "usebackuproot"},
403 {Opt_user_subvol_rm_allowed
, "user_subvol_rm_allowed"},
405 /* Deprecated options */
406 {Opt_alloc_start
, "alloc_start=%s"},
407 {Opt_recovery
, "recovery"},
408 {Opt_subvolrootid
, "subvolrootid=%d"},
410 /* Debugging options */
411 {Opt_check_integrity
, "check_int"},
412 {Opt_check_integrity_including_extent_data
, "check_int_data"},
413 {Opt_check_integrity_print_mask
, "check_int_print_mask=%u"},
414 {Opt_enospc_debug
, "enospc_debug"},
415 {Opt_noenospc_debug
, "noenospc_debug"},
416 #ifdef CONFIG_BTRFS_DEBUG
417 {Opt_fragment_data
, "fragment=data"},
418 {Opt_fragment_metadata
, "fragment=metadata"},
419 {Opt_fragment_all
, "fragment=all"},
421 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
422 {Opt_ref_verify
, "ref_verify"},
428 * Regular mount options parser. Everything that is needed only when
429 * reading in a new superblock is parsed here.
430 * XXX JDM: This needs to be cleaned up for remount.
432 int btrfs_parse_options(struct btrfs_fs_info
*info
, char *options
,
433 unsigned long new_flags
)
435 substring_t args
[MAX_OPT_ARGS
];
441 bool compress_force
= false;
442 enum btrfs_compression_type saved_compress_type
;
443 bool saved_compress_force
;
446 cache_gen
= btrfs_super_cache_generation(info
->super_copy
);
447 if (btrfs_fs_compat_ro(info
, FREE_SPACE_TREE
))
448 btrfs_set_opt(info
->mount_opt
, FREE_SPACE_TREE
);
450 btrfs_set_opt(info
->mount_opt
, SPACE_CACHE
);
453 * Even the options are empty, we still need to do extra check
459 while ((p
= strsep(&options
, ",")) != NULL
) {
464 token
= match_token(p
, tokens
, args
);
467 btrfs_info(info
, "allowing degraded mounts");
468 btrfs_set_opt(info
->mount_opt
, DEGRADED
);
471 case Opt_subvol_empty
:
473 case Opt_subvolrootid
:
476 * These are parsed by btrfs_parse_subvol_options or
477 * btrfs_parse_device_options and can be ignored here.
481 btrfs_set_and_info(info
, NODATASUM
,
482 "setting nodatasum");
485 if (btrfs_test_opt(info
, NODATASUM
)) {
486 if (btrfs_test_opt(info
, NODATACOW
))
488 "setting datasum, datacow enabled");
490 btrfs_info(info
, "setting datasum");
492 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
493 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
496 if (!btrfs_test_opt(info
, NODATACOW
)) {
497 if (!btrfs_test_opt(info
, COMPRESS
) ||
498 !btrfs_test_opt(info
, FORCE_COMPRESS
)) {
500 "setting nodatacow, compression disabled");
502 btrfs_info(info
, "setting nodatacow");
505 btrfs_clear_opt(info
->mount_opt
, COMPRESS
);
506 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
507 btrfs_set_opt(info
->mount_opt
, NODATACOW
);
508 btrfs_set_opt(info
->mount_opt
, NODATASUM
);
511 btrfs_clear_and_info(info
, NODATACOW
,
514 case Opt_compress_force
:
515 case Opt_compress_force_type
:
516 compress_force
= true;
519 case Opt_compress_type
:
520 saved_compress_type
= btrfs_test_opt(info
,
522 info
->compress_type
: BTRFS_COMPRESS_NONE
;
523 saved_compress_force
=
524 btrfs_test_opt(info
, FORCE_COMPRESS
);
525 if (token
== Opt_compress
||
526 token
== Opt_compress_force
||
527 strncmp(args
[0].from
, "zlib", 4) == 0) {
528 compress_type
= "zlib";
530 info
->compress_type
= BTRFS_COMPRESS_ZLIB
;
531 info
->compress_level
= BTRFS_ZLIB_DEFAULT_LEVEL
;
533 * args[0] contains uninitialized data since
534 * for these tokens we don't expect any
537 if (token
!= Opt_compress
&&
538 token
!= Opt_compress_force
)
539 info
->compress_level
=
540 btrfs_compress_str2level(
543 btrfs_set_opt(info
->mount_opt
, COMPRESS
);
544 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
545 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
547 } else if (strncmp(args
[0].from
, "lzo", 3) == 0) {
548 compress_type
= "lzo";
549 info
->compress_type
= BTRFS_COMPRESS_LZO
;
550 btrfs_set_opt(info
->mount_opt
, COMPRESS
);
551 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
552 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
553 btrfs_set_fs_incompat(info
, COMPRESS_LZO
);
555 } else if (strncmp(args
[0].from
, "zstd", 4) == 0) {
556 compress_type
= "zstd";
557 info
->compress_type
= BTRFS_COMPRESS_ZSTD
;
558 info
->compress_level
=
559 btrfs_compress_str2level(
562 btrfs_set_opt(info
->mount_opt
, COMPRESS
);
563 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
564 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
565 btrfs_set_fs_incompat(info
, COMPRESS_ZSTD
);
567 } else if (strncmp(args
[0].from
, "no", 2) == 0) {
568 compress_type
= "no";
569 btrfs_clear_opt(info
->mount_opt
, COMPRESS
);
570 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
571 compress_force
= false;
578 if (compress_force
) {
579 btrfs_set_opt(info
->mount_opt
, FORCE_COMPRESS
);
582 * If we remount from compress-force=xxx to
583 * compress=xxx, we need clear FORCE_COMPRESS
584 * flag, otherwise, there is no way for users
585 * to disable forcible compression separately.
587 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
589 if ((btrfs_test_opt(info
, COMPRESS
) &&
590 (info
->compress_type
!= saved_compress_type
||
591 compress_force
!= saved_compress_force
)) ||
592 (!btrfs_test_opt(info
, COMPRESS
) &&
594 btrfs_info(info
, "%s %s compression, level %d",
595 (compress_force
) ? "force" : "use",
596 compress_type
, info
->compress_level
);
598 compress_force
= false;
601 btrfs_set_and_info(info
, SSD
,
602 "enabling ssd optimizations");
603 btrfs_clear_opt(info
->mount_opt
, NOSSD
);
606 btrfs_set_and_info(info
, SSD
,
607 "enabling ssd optimizations");
608 btrfs_set_and_info(info
, SSD_SPREAD
,
609 "using spread ssd allocation scheme");
610 btrfs_clear_opt(info
->mount_opt
, NOSSD
);
613 btrfs_set_opt(info
->mount_opt
, NOSSD
);
614 btrfs_clear_and_info(info
, SSD
,
615 "not using ssd optimizations");
617 case Opt_nossd_spread
:
618 btrfs_clear_and_info(info
, SSD_SPREAD
,
619 "not using spread ssd allocation scheme");
622 btrfs_clear_and_info(info
, NOBARRIER
,
623 "turning on barriers");
626 btrfs_set_and_info(info
, NOBARRIER
,
627 "turning off barriers");
629 case Opt_thread_pool
:
630 ret
= match_int(&args
[0], &intarg
);
633 } else if (intarg
== 0) {
637 info
->thread_pool_size
= intarg
;
640 num
= match_strdup(&args
[0]);
642 info
->max_inline
= memparse(num
, NULL
);
645 if (info
->max_inline
) {
646 info
->max_inline
= min_t(u64
,
650 btrfs_info(info
, "max_inline at %llu",
657 case Opt_alloc_start
:
659 "option alloc_start is obsolete, ignored");
662 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
663 info
->sb
->s_flags
|= SB_POSIXACL
;
666 btrfs_err(info
, "support for ACL not compiled in!");
671 info
->sb
->s_flags
&= ~SB_POSIXACL
;
674 btrfs_set_and_info(info
, NOTREELOG
,
675 "disabling tree log");
678 btrfs_clear_and_info(info
, NOTREELOG
,
679 "enabling tree log");
682 case Opt_nologreplay
:
683 btrfs_set_and_info(info
, NOLOGREPLAY
,
684 "disabling log replay at mount time");
686 case Opt_flushoncommit
:
687 btrfs_set_and_info(info
, FLUSHONCOMMIT
,
688 "turning on flush-on-commit");
690 case Opt_noflushoncommit
:
691 btrfs_clear_and_info(info
, FLUSHONCOMMIT
,
692 "turning off flush-on-commit");
695 ret
= match_int(&args
[0], &intarg
);
698 info
->metadata_ratio
= intarg
;
699 btrfs_info(info
, "metadata ratio %u",
700 info
->metadata_ratio
);
703 case Opt_discard_mode
:
704 if (token
== Opt_discard
||
705 strcmp(args
[0].from
, "sync") == 0) {
706 btrfs_clear_opt(info
->mount_opt
, DISCARD_ASYNC
);
707 btrfs_set_and_info(info
, DISCARD_SYNC
,
708 "turning on sync discard");
709 } else if (strcmp(args
[0].from
, "async") == 0) {
710 btrfs_clear_opt(info
->mount_opt
, DISCARD_SYNC
);
711 btrfs_set_and_info(info
, DISCARD_ASYNC
,
712 "turning on async discard");
719 btrfs_clear_and_info(info
, DISCARD_SYNC
,
720 "turning off discard");
721 btrfs_clear_and_info(info
, DISCARD_ASYNC
,
722 "turning off async discard");
724 case Opt_space_cache
:
725 case Opt_space_cache_version
:
726 if (token
== Opt_space_cache
||
727 strcmp(args
[0].from
, "v1") == 0) {
728 btrfs_clear_opt(info
->mount_opt
,
730 btrfs_set_and_info(info
, SPACE_CACHE
,
731 "enabling disk space caching");
732 } else if (strcmp(args
[0].from
, "v2") == 0) {
733 btrfs_clear_opt(info
->mount_opt
,
735 btrfs_set_and_info(info
, FREE_SPACE_TREE
,
736 "enabling free space tree");
742 case Opt_rescan_uuid_tree
:
743 btrfs_set_opt(info
->mount_opt
, RESCAN_UUID_TREE
);
745 case Opt_no_space_cache
:
746 if (btrfs_test_opt(info
, SPACE_CACHE
)) {
747 btrfs_clear_and_info(info
, SPACE_CACHE
,
748 "disabling disk space caching");
750 if (btrfs_test_opt(info
, FREE_SPACE_TREE
)) {
751 btrfs_clear_and_info(info
, FREE_SPACE_TREE
,
752 "disabling free space tree");
755 case Opt_inode_cache
:
756 btrfs_set_pending_and_info(info
, INODE_MAP_CACHE
,
757 "enabling inode map caching");
759 case Opt_noinode_cache
:
760 btrfs_clear_pending_and_info(info
, INODE_MAP_CACHE
,
761 "disabling inode map caching");
763 case Opt_clear_cache
:
764 btrfs_set_and_info(info
, CLEAR_CACHE
,
765 "force clearing of disk cache");
767 case Opt_user_subvol_rm_allowed
:
768 btrfs_set_opt(info
->mount_opt
, USER_SUBVOL_RM_ALLOWED
);
770 case Opt_enospc_debug
:
771 btrfs_set_opt(info
->mount_opt
, ENOSPC_DEBUG
);
773 case Opt_noenospc_debug
:
774 btrfs_clear_opt(info
->mount_opt
, ENOSPC_DEBUG
);
777 btrfs_set_and_info(info
, AUTO_DEFRAG
,
778 "enabling auto defrag");
781 btrfs_clear_and_info(info
, AUTO_DEFRAG
,
782 "disabling auto defrag");
786 "'recovery' is deprecated, use 'usebackuproot' instead");
788 case Opt_usebackuproot
:
790 "trying to use backup root at mount time");
791 btrfs_set_opt(info
->mount_opt
, USEBACKUPROOT
);
793 case Opt_skip_balance
:
794 btrfs_set_opt(info
->mount_opt
, SKIP_BALANCE
);
796 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
797 case Opt_check_integrity_including_extent_data
:
799 "enabling check integrity including extent data");
800 btrfs_set_opt(info
->mount_opt
,
801 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
);
802 btrfs_set_opt(info
->mount_opt
, CHECK_INTEGRITY
);
804 case Opt_check_integrity
:
805 btrfs_info(info
, "enabling check integrity");
806 btrfs_set_opt(info
->mount_opt
, CHECK_INTEGRITY
);
808 case Opt_check_integrity_print_mask
:
809 ret
= match_int(&args
[0], &intarg
);
812 info
->check_integrity_print_mask
= intarg
;
813 btrfs_info(info
, "check_integrity_print_mask 0x%x",
814 info
->check_integrity_print_mask
);
817 case Opt_check_integrity_including_extent_data
:
818 case Opt_check_integrity
:
819 case Opt_check_integrity_print_mask
:
821 "support for check_integrity* not compiled in!");
825 case Opt_fatal_errors
:
826 if (strcmp(args
[0].from
, "panic") == 0)
827 btrfs_set_opt(info
->mount_opt
,
828 PANIC_ON_FATAL_ERROR
);
829 else if (strcmp(args
[0].from
, "bug") == 0)
830 btrfs_clear_opt(info
->mount_opt
,
831 PANIC_ON_FATAL_ERROR
);
837 case Opt_commit_interval
:
839 ret
= match_int(&args
[0], &intarg
);
844 "using default commit interval %us",
845 BTRFS_DEFAULT_COMMIT_INTERVAL
);
846 intarg
= BTRFS_DEFAULT_COMMIT_INTERVAL
;
847 } else if (intarg
> 300) {
848 btrfs_warn(info
, "excessive commit interval %d",
851 info
->commit_interval
= intarg
;
853 #ifdef CONFIG_BTRFS_DEBUG
854 case Opt_fragment_all
:
855 btrfs_info(info
, "fragmenting all space");
856 btrfs_set_opt(info
->mount_opt
, FRAGMENT_DATA
);
857 btrfs_set_opt(info
->mount_opt
, FRAGMENT_METADATA
);
859 case Opt_fragment_metadata
:
860 btrfs_info(info
, "fragmenting metadata");
861 btrfs_set_opt(info
->mount_opt
,
864 case Opt_fragment_data
:
865 btrfs_info(info
, "fragmenting data");
866 btrfs_set_opt(info
->mount_opt
, FRAGMENT_DATA
);
869 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
871 btrfs_info(info
, "doing ref verification");
872 btrfs_set_opt(info
->mount_opt
, REF_VERIFY
);
876 btrfs_err(info
, "unrecognized mount option '%s'", p
);
885 * Extra check for current option against current flag
887 if (btrfs_test_opt(info
, NOLOGREPLAY
) && !(new_flags
& SB_RDONLY
)) {
889 "nologreplay must be used with ro mount option");
893 if (btrfs_fs_compat_ro(info
, FREE_SPACE_TREE
) &&
894 !btrfs_test_opt(info
, FREE_SPACE_TREE
) &&
895 !btrfs_test_opt(info
, CLEAR_CACHE
)) {
896 btrfs_err(info
, "cannot disable free space tree");
900 if (!ret
&& btrfs_test_opt(info
, SPACE_CACHE
))
901 btrfs_info(info
, "disk space caching is enabled");
902 if (!ret
&& btrfs_test_opt(info
, FREE_SPACE_TREE
))
903 btrfs_info(info
, "using free space tree");
908 * Parse mount options that are required early in the mount process.
910 * All other options will be parsed on much later in the mount process and
911 * only when we need to allocate a new super block.
913 static int btrfs_parse_device_options(const char *options
, fmode_t flags
,
916 substring_t args
[MAX_OPT_ARGS
];
917 char *device_name
, *opts
, *orig
, *p
;
918 struct btrfs_device
*device
= NULL
;
921 lockdep_assert_held(&uuid_mutex
);
927 * strsep changes the string, duplicate it because btrfs_parse_options
930 opts
= kstrdup(options
, GFP_KERNEL
);
935 while ((p
= strsep(&opts
, ",")) != NULL
) {
941 token
= match_token(p
, tokens
, args
);
942 if (token
== Opt_device
) {
943 device_name
= match_strdup(&args
[0]);
948 device
= btrfs_scan_one_device(device_name
, flags
,
951 if (IS_ERR(device
)) {
952 error
= PTR_ERR(device
);
964 * Parse mount options that are related to subvolume id
966 * The value is later passed to mount_subvol()
968 static int btrfs_parse_subvol_options(const char *options
, char **subvol_name
,
969 u64
*subvol_objectid
)
971 substring_t args
[MAX_OPT_ARGS
];
972 char *opts
, *orig
, *p
;
980 * strsep changes the string, duplicate it because
981 * btrfs_parse_device_options gets called later
983 opts
= kstrdup(options
, GFP_KERNEL
);
988 while ((p
= strsep(&opts
, ",")) != NULL
) {
993 token
= match_token(p
, tokens
, args
);
997 *subvol_name
= match_strdup(&args
[0]);
1004 error
= match_u64(&args
[0], &subvolid
);
1008 /* we want the original fs_tree */
1010 subvolid
= BTRFS_FS_TREE_OBJECTID
;
1012 *subvol_objectid
= subvolid
;
1014 case Opt_subvolrootid
:
1015 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
1027 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info
*fs_info
,
1028 u64 subvol_objectid
)
1030 struct btrfs_root
*root
= fs_info
->tree_root
;
1031 struct btrfs_root
*fs_root
= NULL
;
1032 struct btrfs_root_ref
*root_ref
;
1033 struct btrfs_inode_ref
*inode_ref
;
1034 struct btrfs_key key
;
1035 struct btrfs_path
*path
= NULL
;
1036 char *name
= NULL
, *ptr
;
1041 path
= btrfs_alloc_path();
1046 path
->leave_spinning
= 1;
1048 name
= kmalloc(PATH_MAX
, GFP_KERNEL
);
1053 ptr
= name
+ PATH_MAX
- 1;
1057 * Walk up the subvolume trees in the tree of tree roots by root
1058 * backrefs until we hit the top-level subvolume.
1060 while (subvol_objectid
!= BTRFS_FS_TREE_OBJECTID
) {
1061 key
.objectid
= subvol_objectid
;
1062 key
.type
= BTRFS_ROOT_BACKREF_KEY
;
1063 key
.offset
= (u64
)-1;
1065 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1068 } else if (ret
> 0) {
1069 ret
= btrfs_previous_item(root
, path
, subvol_objectid
,
1070 BTRFS_ROOT_BACKREF_KEY
);
1073 } else if (ret
> 0) {
1079 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1080 subvol_objectid
= key
.offset
;
1082 root_ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
1083 struct btrfs_root_ref
);
1084 len
= btrfs_root_ref_name_len(path
->nodes
[0], root_ref
);
1087 ret
= -ENAMETOOLONG
;
1090 read_extent_buffer(path
->nodes
[0], ptr
+ 1,
1091 (unsigned long)(root_ref
+ 1), len
);
1093 dirid
= btrfs_root_ref_dirid(path
->nodes
[0], root_ref
);
1094 btrfs_release_path(path
);
1096 key
.objectid
= subvol_objectid
;
1097 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1098 key
.offset
= (u64
)-1;
1099 fs_root
= btrfs_get_fs_root(fs_info
, &key
, true);
1100 if (IS_ERR(fs_root
)) {
1101 ret
= PTR_ERR(fs_root
);
1107 * Walk up the filesystem tree by inode refs until we hit the
1110 while (dirid
!= BTRFS_FIRST_FREE_OBJECTID
) {
1111 key
.objectid
= dirid
;
1112 key
.type
= BTRFS_INODE_REF_KEY
;
1113 key
.offset
= (u64
)-1;
1115 ret
= btrfs_search_slot(NULL
, fs_root
, &key
, path
, 0, 0);
1118 } else if (ret
> 0) {
1119 ret
= btrfs_previous_item(fs_root
, path
, dirid
,
1120 BTRFS_INODE_REF_KEY
);
1123 } else if (ret
> 0) {
1129 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1132 inode_ref
= btrfs_item_ptr(path
->nodes
[0],
1134 struct btrfs_inode_ref
);
1135 len
= btrfs_inode_ref_name_len(path
->nodes
[0],
1139 ret
= -ENAMETOOLONG
;
1142 read_extent_buffer(path
->nodes
[0], ptr
+ 1,
1143 (unsigned long)(inode_ref
+ 1), len
);
1145 btrfs_release_path(path
);
1147 btrfs_put_root(fs_root
);
1151 btrfs_free_path(path
);
1152 if (ptr
== name
+ PATH_MAX
- 1) {
1156 memmove(name
, ptr
, name
+ PATH_MAX
- ptr
);
1161 btrfs_put_root(fs_root
);
1162 btrfs_free_path(path
);
1164 return ERR_PTR(ret
);
1167 static int get_default_subvol_objectid(struct btrfs_fs_info
*fs_info
, u64
*objectid
)
1169 struct btrfs_root
*root
= fs_info
->tree_root
;
1170 struct btrfs_dir_item
*di
;
1171 struct btrfs_path
*path
;
1172 struct btrfs_key location
;
1175 path
= btrfs_alloc_path();
1178 path
->leave_spinning
= 1;
1181 * Find the "default" dir item which points to the root item that we
1182 * will mount by default if we haven't been given a specific subvolume
1185 dir_id
= btrfs_super_root_dir(fs_info
->super_copy
);
1186 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir_id
, "default", 7, 0);
1188 btrfs_free_path(path
);
1193 * Ok the default dir item isn't there. This is weird since
1194 * it's always been there, but don't freak out, just try and
1195 * mount the top-level subvolume.
1197 btrfs_free_path(path
);
1198 *objectid
= BTRFS_FS_TREE_OBJECTID
;
1202 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &location
);
1203 btrfs_free_path(path
);
1204 *objectid
= location
.objectid
;
1208 static int btrfs_fill_super(struct super_block
*sb
,
1209 struct btrfs_fs_devices
*fs_devices
,
1212 struct inode
*inode
;
1213 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1214 struct btrfs_key key
;
1217 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
1218 sb
->s_magic
= BTRFS_SUPER_MAGIC
;
1219 sb
->s_op
= &btrfs_super_ops
;
1220 sb
->s_d_op
= &btrfs_dentry_operations
;
1221 sb
->s_export_op
= &btrfs_export_ops
;
1222 sb
->s_xattr
= btrfs_xattr_handlers
;
1223 sb
->s_time_gran
= 1;
1224 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1225 sb
->s_flags
|= SB_POSIXACL
;
1227 sb
->s_flags
|= SB_I_VERSION
;
1228 sb
->s_iflags
|= SB_I_CGROUPWB
;
1230 err
= super_setup_bdi(sb
);
1232 btrfs_err(fs_info
, "super_setup_bdi failed");
1236 err
= open_ctree(sb
, fs_devices
, (char *)data
);
1238 btrfs_err(fs_info
, "open_ctree failed");
1242 key
.objectid
= BTRFS_FIRST_FREE_OBJECTID
;
1243 key
.type
= BTRFS_INODE_ITEM_KEY
;
1245 inode
= btrfs_iget(sb
, &key
, fs_info
->fs_root
);
1246 if (IS_ERR(inode
)) {
1247 err
= PTR_ERR(inode
);
1251 sb
->s_root
= d_make_root(inode
);
1257 cleancache_init_fs(sb
);
1258 sb
->s_flags
|= SB_ACTIVE
;
1262 close_ctree(fs_info
);
1266 int btrfs_sync_fs(struct super_block
*sb
, int wait
)
1268 struct btrfs_trans_handle
*trans
;
1269 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1270 struct btrfs_root
*root
= fs_info
->tree_root
;
1272 trace_btrfs_sync_fs(fs_info
, wait
);
1275 filemap_flush(fs_info
->btree_inode
->i_mapping
);
1279 btrfs_wait_ordered_roots(fs_info
, U64_MAX
, 0, (u64
)-1);
1281 trans
= btrfs_attach_transaction_barrier(root
);
1282 if (IS_ERR(trans
)) {
1283 /* no transaction, don't bother */
1284 if (PTR_ERR(trans
) == -ENOENT
) {
1286 * Exit unless we have some pending changes
1287 * that need to go through commit
1289 if (fs_info
->pending_changes
== 0)
1292 * A non-blocking test if the fs is frozen. We must not
1293 * start a new transaction here otherwise a deadlock
1294 * happens. The pending operations are delayed to the
1295 * next commit after thawing.
1297 if (sb_start_write_trylock(sb
))
1301 trans
= btrfs_start_transaction(root
, 0);
1304 return PTR_ERR(trans
);
1306 return btrfs_commit_transaction(trans
);
1309 static int btrfs_show_options(struct seq_file
*seq
, struct dentry
*dentry
)
1311 struct btrfs_fs_info
*info
= btrfs_sb(dentry
->d_sb
);
1312 const char *compress_type
;
1314 if (btrfs_test_opt(info
, DEGRADED
))
1315 seq_puts(seq
, ",degraded");
1316 if (btrfs_test_opt(info
, NODATASUM
))
1317 seq_puts(seq
, ",nodatasum");
1318 if (btrfs_test_opt(info
, NODATACOW
))
1319 seq_puts(seq
, ",nodatacow");
1320 if (btrfs_test_opt(info
, NOBARRIER
))
1321 seq_puts(seq
, ",nobarrier");
1322 if (info
->max_inline
!= BTRFS_DEFAULT_MAX_INLINE
)
1323 seq_printf(seq
, ",max_inline=%llu", info
->max_inline
);
1324 if (info
->thread_pool_size
!= min_t(unsigned long,
1325 num_online_cpus() + 2, 8))
1326 seq_printf(seq
, ",thread_pool=%u", info
->thread_pool_size
);
1327 if (btrfs_test_opt(info
, COMPRESS
)) {
1328 compress_type
= btrfs_compress_type2str(info
->compress_type
);
1329 if (btrfs_test_opt(info
, FORCE_COMPRESS
))
1330 seq_printf(seq
, ",compress-force=%s", compress_type
);
1332 seq_printf(seq
, ",compress=%s", compress_type
);
1333 if (info
->compress_level
)
1334 seq_printf(seq
, ":%d", info
->compress_level
);
1336 if (btrfs_test_opt(info
, NOSSD
))
1337 seq_puts(seq
, ",nossd");
1338 if (btrfs_test_opt(info
, SSD_SPREAD
))
1339 seq_puts(seq
, ",ssd_spread");
1340 else if (btrfs_test_opt(info
, SSD
))
1341 seq_puts(seq
, ",ssd");
1342 if (btrfs_test_opt(info
, NOTREELOG
))
1343 seq_puts(seq
, ",notreelog");
1344 if (btrfs_test_opt(info
, NOLOGREPLAY
))
1345 seq_puts(seq
, ",nologreplay");
1346 if (btrfs_test_opt(info
, FLUSHONCOMMIT
))
1347 seq_puts(seq
, ",flushoncommit");
1348 if (btrfs_test_opt(info
, DISCARD_SYNC
))
1349 seq_puts(seq
, ",discard");
1350 if (btrfs_test_opt(info
, DISCARD_ASYNC
))
1351 seq_puts(seq
, ",discard=async");
1352 if (!(info
->sb
->s_flags
& SB_POSIXACL
))
1353 seq_puts(seq
, ",noacl");
1354 if (btrfs_test_opt(info
, SPACE_CACHE
))
1355 seq_puts(seq
, ",space_cache");
1356 else if (btrfs_test_opt(info
, FREE_SPACE_TREE
))
1357 seq_puts(seq
, ",space_cache=v2");
1359 seq_puts(seq
, ",nospace_cache");
1360 if (btrfs_test_opt(info
, RESCAN_UUID_TREE
))
1361 seq_puts(seq
, ",rescan_uuid_tree");
1362 if (btrfs_test_opt(info
, CLEAR_CACHE
))
1363 seq_puts(seq
, ",clear_cache");
1364 if (btrfs_test_opt(info
, USER_SUBVOL_RM_ALLOWED
))
1365 seq_puts(seq
, ",user_subvol_rm_allowed");
1366 if (btrfs_test_opt(info
, ENOSPC_DEBUG
))
1367 seq_puts(seq
, ",enospc_debug");
1368 if (btrfs_test_opt(info
, AUTO_DEFRAG
))
1369 seq_puts(seq
, ",autodefrag");
1370 if (btrfs_test_opt(info
, INODE_MAP_CACHE
))
1371 seq_puts(seq
, ",inode_cache");
1372 if (btrfs_test_opt(info
, SKIP_BALANCE
))
1373 seq_puts(seq
, ",skip_balance");
1374 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1375 if (btrfs_test_opt(info
, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
))
1376 seq_puts(seq
, ",check_int_data");
1377 else if (btrfs_test_opt(info
, CHECK_INTEGRITY
))
1378 seq_puts(seq
, ",check_int");
1379 if (info
->check_integrity_print_mask
)
1380 seq_printf(seq
, ",check_int_print_mask=%d",
1381 info
->check_integrity_print_mask
);
1383 if (info
->metadata_ratio
)
1384 seq_printf(seq
, ",metadata_ratio=%u", info
->metadata_ratio
);
1385 if (btrfs_test_opt(info
, PANIC_ON_FATAL_ERROR
))
1386 seq_puts(seq
, ",fatal_errors=panic");
1387 if (info
->commit_interval
!= BTRFS_DEFAULT_COMMIT_INTERVAL
)
1388 seq_printf(seq
, ",commit=%u", info
->commit_interval
);
1389 #ifdef CONFIG_BTRFS_DEBUG
1390 if (btrfs_test_opt(info
, FRAGMENT_DATA
))
1391 seq_puts(seq
, ",fragment=data");
1392 if (btrfs_test_opt(info
, FRAGMENT_METADATA
))
1393 seq_puts(seq
, ",fragment=metadata");
1395 if (btrfs_test_opt(info
, REF_VERIFY
))
1396 seq_puts(seq
, ",ref_verify");
1397 seq_printf(seq
, ",subvolid=%llu",
1398 BTRFS_I(d_inode(dentry
))->root
->root_key
.objectid
);
1399 seq_puts(seq
, ",subvol=");
1400 seq_dentry(seq
, dentry
, " \t\n\\");
1404 static int btrfs_test_super(struct super_block
*s
, void *data
)
1406 struct btrfs_fs_info
*p
= data
;
1407 struct btrfs_fs_info
*fs_info
= btrfs_sb(s
);
1409 return fs_info
->fs_devices
== p
->fs_devices
;
1412 static int btrfs_set_super(struct super_block
*s
, void *data
)
1414 int err
= set_anon_super(s
, data
);
1416 s
->s_fs_info
= data
;
1421 * subvolumes are identified by ino 256
1423 static inline int is_subvolume_inode(struct inode
*inode
)
1425 if (inode
&& inode
->i_ino
== BTRFS_FIRST_FREE_OBJECTID
)
1430 static struct dentry
*mount_subvol(const char *subvol_name
, u64 subvol_objectid
,
1431 struct vfsmount
*mnt
)
1433 struct dentry
*root
;
1437 if (!subvol_objectid
) {
1438 ret
= get_default_subvol_objectid(btrfs_sb(mnt
->mnt_sb
),
1441 root
= ERR_PTR(ret
);
1445 subvol_name
= btrfs_get_subvol_name_from_objectid(
1446 btrfs_sb(mnt
->mnt_sb
), subvol_objectid
);
1447 if (IS_ERR(subvol_name
)) {
1448 root
= ERR_CAST(subvol_name
);
1455 root
= mount_subtree(mnt
, subvol_name
);
1456 /* mount_subtree() drops our reference on the vfsmount. */
1459 if (!IS_ERR(root
)) {
1460 struct super_block
*s
= root
->d_sb
;
1461 struct btrfs_fs_info
*fs_info
= btrfs_sb(s
);
1462 struct inode
*root_inode
= d_inode(root
);
1463 u64 root_objectid
= BTRFS_I(root_inode
)->root
->root_key
.objectid
;
1466 if (!is_subvolume_inode(root_inode
)) {
1467 btrfs_err(fs_info
, "'%s' is not a valid subvolume",
1471 if (subvol_objectid
&& root_objectid
!= subvol_objectid
) {
1473 * This will also catch a race condition where a
1474 * subvolume which was passed by ID is renamed and
1475 * another subvolume is renamed over the old location.
1478 "subvol '%s' does not match subvolid %llu",
1479 subvol_name
, subvol_objectid
);
1484 root
= ERR_PTR(ret
);
1485 deactivate_locked_super(s
);
1496 * Find a superblock for the given device / mount point.
1498 * Note: This is based on mount_bdev from fs/super.c with a few additions
1499 * for multiple device setup. Make sure to keep it in sync.
1501 static struct dentry
*btrfs_mount_root(struct file_system_type
*fs_type
,
1502 int flags
, const char *device_name
, void *data
)
1504 struct block_device
*bdev
= NULL
;
1505 struct super_block
*s
;
1506 struct btrfs_device
*device
= NULL
;
1507 struct btrfs_fs_devices
*fs_devices
= NULL
;
1508 struct btrfs_fs_info
*fs_info
= NULL
;
1509 void *new_sec_opts
= NULL
;
1510 fmode_t mode
= FMODE_READ
;
1513 if (!(flags
& SB_RDONLY
))
1514 mode
|= FMODE_WRITE
;
1517 error
= security_sb_eat_lsm_opts(data
, &new_sec_opts
);
1519 return ERR_PTR(error
);
1523 * Setup a dummy root and fs_info for test/set super. This is because
1524 * we don't actually fill this stuff out until open_ctree, but we need
1525 * then open_ctree will properly initialize the file system specific
1526 * settings later. btrfs_init_fs_info initializes the static elements
1527 * of the fs_info (locks and such) to make cleanup easier if we find a
1528 * superblock with our given fs_devices later on at sget() time.
1530 fs_info
= kvzalloc(sizeof(struct btrfs_fs_info
), GFP_KERNEL
);
1533 goto error_sec_opts
;
1535 btrfs_init_fs_info(fs_info
);
1537 fs_info
->super_copy
= kzalloc(BTRFS_SUPER_INFO_SIZE
, GFP_KERNEL
);
1538 fs_info
->super_for_commit
= kzalloc(BTRFS_SUPER_INFO_SIZE
, GFP_KERNEL
);
1539 if (!fs_info
->super_copy
|| !fs_info
->super_for_commit
) {
1544 mutex_lock(&uuid_mutex
);
1545 error
= btrfs_parse_device_options(data
, mode
, fs_type
);
1547 mutex_unlock(&uuid_mutex
);
1551 device
= btrfs_scan_one_device(device_name
, mode
, fs_type
);
1552 if (IS_ERR(device
)) {
1553 mutex_unlock(&uuid_mutex
);
1554 error
= PTR_ERR(device
);
1558 fs_devices
= device
->fs_devices
;
1559 fs_info
->fs_devices
= fs_devices
;
1561 error
= btrfs_open_devices(fs_devices
, mode
, fs_type
);
1562 mutex_unlock(&uuid_mutex
);
1566 if (!(flags
& SB_RDONLY
) && fs_devices
->rw_devices
== 0) {
1568 goto error_close_devices
;
1571 bdev
= fs_devices
->latest_bdev
;
1572 s
= sget(fs_type
, btrfs_test_super
, btrfs_set_super
, flags
| SB_NOSEC
,
1576 goto error_close_devices
;
1580 btrfs_close_devices(fs_devices
);
1581 btrfs_free_fs_info(fs_info
);
1582 if ((flags
^ s
->s_flags
) & SB_RDONLY
)
1585 snprintf(s
->s_id
, sizeof(s
->s_id
), "%pg", bdev
);
1586 btrfs_sb(s
)->bdev_holder
= fs_type
;
1587 if (!strstr(crc32c_impl(), "generic"))
1588 set_bit(BTRFS_FS_CSUM_IMPL_FAST
, &fs_info
->flags
);
1589 error
= btrfs_fill_super(s
, fs_devices
, data
);
1592 error
= security_sb_set_mnt_opts(s
, new_sec_opts
, 0, NULL
);
1593 security_free_mnt_opts(&new_sec_opts
);
1595 deactivate_locked_super(s
);
1596 return ERR_PTR(error
);
1599 return dget(s
->s_root
);
1601 error_close_devices
:
1602 btrfs_close_devices(fs_devices
);
1604 btrfs_free_fs_info(fs_info
);
1606 security_free_mnt_opts(&new_sec_opts
);
1607 return ERR_PTR(error
);
1611 * Mount function which is called by VFS layer.
1613 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1614 * which needs vfsmount* of device's root (/). This means device's root has to
1615 * be mounted internally in any case.
1618 * 1. Parse subvol id related options for later use in mount_subvol().
1620 * 2. Mount device's root (/) by calling vfs_kern_mount().
1622 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1623 * first place. In order to avoid calling btrfs_mount() again, we use
1624 * different file_system_type which is not registered to VFS by
1625 * register_filesystem() (btrfs_root_fs_type). As a result,
1626 * btrfs_mount_root() is called. The return value will be used by
1627 * mount_subtree() in mount_subvol().
1629 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1630 * "btrfs subvolume set-default", mount_subvol() is called always.
1632 static struct dentry
*btrfs_mount(struct file_system_type
*fs_type
, int flags
,
1633 const char *device_name
, void *data
)
1635 struct vfsmount
*mnt_root
;
1636 struct dentry
*root
;
1637 char *subvol_name
= NULL
;
1638 u64 subvol_objectid
= 0;
1641 error
= btrfs_parse_subvol_options(data
, &subvol_name
,
1645 return ERR_PTR(error
);
1648 /* mount device's root (/) */
1649 mnt_root
= vfs_kern_mount(&btrfs_root_fs_type
, flags
, device_name
, data
);
1650 if (PTR_ERR_OR_ZERO(mnt_root
) == -EBUSY
) {
1651 if (flags
& SB_RDONLY
) {
1652 mnt_root
= vfs_kern_mount(&btrfs_root_fs_type
,
1653 flags
& ~SB_RDONLY
, device_name
, data
);
1655 mnt_root
= vfs_kern_mount(&btrfs_root_fs_type
,
1656 flags
| SB_RDONLY
, device_name
, data
);
1657 if (IS_ERR(mnt_root
)) {
1658 root
= ERR_CAST(mnt_root
);
1663 down_write(&mnt_root
->mnt_sb
->s_umount
);
1664 error
= btrfs_remount(mnt_root
->mnt_sb
, &flags
, NULL
);
1665 up_write(&mnt_root
->mnt_sb
->s_umount
);
1667 root
= ERR_PTR(error
);
1674 if (IS_ERR(mnt_root
)) {
1675 root
= ERR_CAST(mnt_root
);
1680 /* mount_subvol() will free subvol_name and mnt_root */
1681 root
= mount_subvol(subvol_name
, subvol_objectid
, mnt_root
);
1687 static void btrfs_resize_thread_pool(struct btrfs_fs_info
*fs_info
,
1688 u32 new_pool_size
, u32 old_pool_size
)
1690 if (new_pool_size
== old_pool_size
)
1693 fs_info
->thread_pool_size
= new_pool_size
;
1695 btrfs_info(fs_info
, "resize thread pool %d -> %d",
1696 old_pool_size
, new_pool_size
);
1698 btrfs_workqueue_set_max(fs_info
->workers
, new_pool_size
);
1699 btrfs_workqueue_set_max(fs_info
->delalloc_workers
, new_pool_size
);
1700 btrfs_workqueue_set_max(fs_info
->caching_workers
, new_pool_size
);
1701 btrfs_workqueue_set_max(fs_info
->endio_workers
, new_pool_size
);
1702 btrfs_workqueue_set_max(fs_info
->endio_meta_workers
, new_pool_size
);
1703 btrfs_workqueue_set_max(fs_info
->endio_meta_write_workers
,
1705 btrfs_workqueue_set_max(fs_info
->endio_write_workers
, new_pool_size
);
1706 btrfs_workqueue_set_max(fs_info
->endio_freespace_worker
, new_pool_size
);
1707 btrfs_workqueue_set_max(fs_info
->delayed_workers
, new_pool_size
);
1708 btrfs_workqueue_set_max(fs_info
->readahead_workers
, new_pool_size
);
1709 btrfs_workqueue_set_max(fs_info
->scrub_wr_completion_workers
,
1713 static inline void btrfs_remount_prepare(struct btrfs_fs_info
*fs_info
)
1715 set_bit(BTRFS_FS_STATE_REMOUNTING
, &fs_info
->fs_state
);
1718 static inline void btrfs_remount_begin(struct btrfs_fs_info
*fs_info
,
1719 unsigned long old_opts
, int flags
)
1721 if (btrfs_raw_test_opt(old_opts
, AUTO_DEFRAG
) &&
1722 (!btrfs_raw_test_opt(fs_info
->mount_opt
, AUTO_DEFRAG
) ||
1723 (flags
& SB_RDONLY
))) {
1724 /* wait for any defraggers to finish */
1725 wait_event(fs_info
->transaction_wait
,
1726 (atomic_read(&fs_info
->defrag_running
) == 0));
1727 if (flags
& SB_RDONLY
)
1728 sync_filesystem(fs_info
->sb
);
1732 static inline void btrfs_remount_cleanup(struct btrfs_fs_info
*fs_info
,
1733 unsigned long old_opts
)
1736 * We need to cleanup all defragable inodes if the autodefragment is
1737 * close or the filesystem is read only.
1739 if (btrfs_raw_test_opt(old_opts
, AUTO_DEFRAG
) &&
1740 (!btrfs_raw_test_opt(fs_info
->mount_opt
, AUTO_DEFRAG
) || sb_rdonly(fs_info
->sb
))) {
1741 btrfs_cleanup_defrag_inodes(fs_info
);
1744 /* If we toggled discard async */
1745 if (!btrfs_raw_test_opt(old_opts
, DISCARD_ASYNC
) &&
1746 btrfs_test_opt(fs_info
, DISCARD_ASYNC
))
1747 btrfs_discard_resume(fs_info
);
1748 else if (btrfs_raw_test_opt(old_opts
, DISCARD_ASYNC
) &&
1749 !btrfs_test_opt(fs_info
, DISCARD_ASYNC
))
1750 btrfs_discard_cleanup(fs_info
);
1752 clear_bit(BTRFS_FS_STATE_REMOUNTING
, &fs_info
->fs_state
);
1755 static int btrfs_remount(struct super_block
*sb
, int *flags
, char *data
)
1757 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1758 struct btrfs_root
*root
= fs_info
->tree_root
;
1759 unsigned old_flags
= sb
->s_flags
;
1760 unsigned long old_opts
= fs_info
->mount_opt
;
1761 unsigned long old_compress_type
= fs_info
->compress_type
;
1762 u64 old_max_inline
= fs_info
->max_inline
;
1763 u32 old_thread_pool_size
= fs_info
->thread_pool_size
;
1764 u32 old_metadata_ratio
= fs_info
->metadata_ratio
;
1767 sync_filesystem(sb
);
1768 btrfs_remount_prepare(fs_info
);
1771 void *new_sec_opts
= NULL
;
1773 ret
= security_sb_eat_lsm_opts(data
, &new_sec_opts
);
1775 ret
= security_sb_remount(sb
, new_sec_opts
);
1776 security_free_mnt_opts(&new_sec_opts
);
1781 ret
= btrfs_parse_options(fs_info
, data
, *flags
);
1785 btrfs_remount_begin(fs_info
, old_opts
, *flags
);
1786 btrfs_resize_thread_pool(fs_info
,
1787 fs_info
->thread_pool_size
, old_thread_pool_size
);
1789 if ((bool)(*flags
& SB_RDONLY
) == sb_rdonly(sb
))
1792 if (*flags
& SB_RDONLY
) {
1794 * this also happens on 'umount -rf' or on shutdown, when
1795 * the filesystem is busy.
1797 cancel_work_sync(&fs_info
->async_reclaim_work
);
1799 btrfs_discard_cleanup(fs_info
);
1801 /* wait for the uuid_scan task to finish */
1802 down(&fs_info
->uuid_tree_rescan_sem
);
1803 /* avoid complains from lockdep et al. */
1804 up(&fs_info
->uuid_tree_rescan_sem
);
1806 sb
->s_flags
|= SB_RDONLY
;
1809 * Setting SB_RDONLY will put the cleaner thread to
1810 * sleep at the next loop if it's already active.
1811 * If it's already asleep, we'll leave unused block
1812 * groups on disk until we're mounted read-write again
1813 * unless we clean them up here.
1815 btrfs_delete_unused_bgs(fs_info
);
1817 btrfs_dev_replace_suspend_for_unmount(fs_info
);
1818 btrfs_scrub_cancel(fs_info
);
1819 btrfs_pause_balance(fs_info
);
1821 ret
= btrfs_commit_super(fs_info
);
1825 if (test_bit(BTRFS_FS_STATE_ERROR
, &fs_info
->fs_state
)) {
1827 "Remounting read-write after error is not allowed");
1831 if (fs_info
->fs_devices
->rw_devices
== 0) {
1836 if (!btrfs_check_rw_degradable(fs_info
, NULL
)) {
1838 "too many missing devices, writable remount is not allowed");
1843 if (btrfs_super_log_root(fs_info
->super_copy
) != 0) {
1845 "mount required to replay tree-log, cannot remount read-write");
1850 ret
= btrfs_cleanup_fs_roots(fs_info
);
1854 /* recover relocation */
1855 mutex_lock(&fs_info
->cleaner_mutex
);
1856 ret
= btrfs_recover_relocation(root
);
1857 mutex_unlock(&fs_info
->cleaner_mutex
);
1861 ret
= btrfs_resume_balance_async(fs_info
);
1865 ret
= btrfs_resume_dev_replace_async(fs_info
);
1867 btrfs_warn(fs_info
, "failed to resume dev_replace");
1871 btrfs_qgroup_rescan_resume(fs_info
);
1873 if (!fs_info
->uuid_root
) {
1874 btrfs_info(fs_info
, "creating UUID tree");
1875 ret
= btrfs_create_uuid_tree(fs_info
);
1878 "failed to create the UUID tree %d",
1883 sb
->s_flags
&= ~SB_RDONLY
;
1885 set_bit(BTRFS_FS_OPEN
, &fs_info
->flags
);
1888 wake_up_process(fs_info
->transaction_kthread
);
1889 btrfs_remount_cleanup(fs_info
, old_opts
);
1893 /* We've hit an error - don't reset SB_RDONLY */
1895 old_flags
|= SB_RDONLY
;
1896 sb
->s_flags
= old_flags
;
1897 fs_info
->mount_opt
= old_opts
;
1898 fs_info
->compress_type
= old_compress_type
;
1899 fs_info
->max_inline
= old_max_inline
;
1900 btrfs_resize_thread_pool(fs_info
,
1901 old_thread_pool_size
, fs_info
->thread_pool_size
);
1902 fs_info
->metadata_ratio
= old_metadata_ratio
;
1903 btrfs_remount_cleanup(fs_info
, old_opts
);
1907 /* Used to sort the devices by max_avail(descending sort) */
1908 static inline int btrfs_cmp_device_free_bytes(const void *dev_info1
,
1909 const void *dev_info2
)
1911 if (((struct btrfs_device_info
*)dev_info1
)->max_avail
>
1912 ((struct btrfs_device_info
*)dev_info2
)->max_avail
)
1914 else if (((struct btrfs_device_info
*)dev_info1
)->max_avail
<
1915 ((struct btrfs_device_info
*)dev_info2
)->max_avail
)
1922 * sort the devices by max_avail, in which max free extent size of each device
1923 * is stored.(Descending Sort)
1925 static inline void btrfs_descending_sort_devices(
1926 struct btrfs_device_info
*devices
,
1929 sort(devices
, nr_devices
, sizeof(struct btrfs_device_info
),
1930 btrfs_cmp_device_free_bytes
, NULL
);
1934 * The helper to calc the free space on the devices that can be used to store
1937 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info
*fs_info
,
1940 struct btrfs_device_info
*devices_info
;
1941 struct btrfs_fs_devices
*fs_devices
= fs_info
->fs_devices
;
1942 struct btrfs_device
*device
;
1945 u64 min_stripe_size
;
1946 int num_stripes
= 1;
1947 int i
= 0, nr_devices
;
1948 const struct btrfs_raid_attr
*rattr
;
1951 * We aren't under the device list lock, so this is racy-ish, but good
1952 * enough for our purposes.
1954 nr_devices
= fs_info
->fs_devices
->open_devices
;
1957 nr_devices
= fs_info
->fs_devices
->open_devices
;
1965 devices_info
= kmalloc_array(nr_devices
, sizeof(*devices_info
),
1970 /* calc min stripe number for data space allocation */
1971 type
= btrfs_data_alloc_profile(fs_info
);
1972 rattr
= &btrfs_raid_array
[btrfs_bg_flags_to_raid_index(type
)];
1974 if (type
& BTRFS_BLOCK_GROUP_RAID0
)
1975 num_stripes
= nr_devices
;
1976 else if (type
& BTRFS_BLOCK_GROUP_RAID1
)
1978 else if (type
& BTRFS_BLOCK_GROUP_RAID1C3
)
1980 else if (type
& BTRFS_BLOCK_GROUP_RAID1C4
)
1982 else if (type
& BTRFS_BLOCK_GROUP_RAID10
)
1985 /* Adjust for more than 1 stripe per device */
1986 min_stripe_size
= rattr
->dev_stripes
* BTRFS_STRIPE_LEN
;
1989 list_for_each_entry_rcu(device
, &fs_devices
->devices
, dev_list
) {
1990 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA
,
1991 &device
->dev_state
) ||
1993 test_bit(BTRFS_DEV_STATE_REPLACE_TGT
, &device
->dev_state
))
1996 if (i
>= nr_devices
)
1999 avail_space
= device
->total_bytes
- device
->bytes_used
;
2001 /* align with stripe_len */
2002 avail_space
= rounddown(avail_space
, BTRFS_STRIPE_LEN
);
2005 * In order to avoid overwriting the superblock on the drive,
2006 * btrfs starts at an offset of at least 1MB when doing chunk
2009 * This ensures we have at least min_stripe_size free space
2010 * after excluding 1MB.
2012 if (avail_space
<= SZ_1M
+ min_stripe_size
)
2015 avail_space
-= SZ_1M
;
2017 devices_info
[i
].dev
= device
;
2018 devices_info
[i
].max_avail
= avail_space
;
2026 btrfs_descending_sort_devices(devices_info
, nr_devices
);
2030 while (nr_devices
>= rattr
->devs_min
) {
2031 num_stripes
= min(num_stripes
, nr_devices
);
2033 if (devices_info
[i
].max_avail
>= min_stripe_size
) {
2037 avail_space
+= devices_info
[i
].max_avail
* num_stripes
;
2038 alloc_size
= devices_info
[i
].max_avail
;
2039 for (j
= i
+ 1 - num_stripes
; j
<= i
; j
++)
2040 devices_info
[j
].max_avail
-= alloc_size
;
2046 kfree(devices_info
);
2047 *free_bytes
= avail_space
;
2052 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2054 * If there's a redundant raid level at DATA block groups, use the respective
2055 * multiplier to scale the sizes.
2057 * Unused device space usage is based on simulating the chunk allocator
2058 * algorithm that respects the device sizes and order of allocations. This is
2059 * a close approximation of the actual use but there are other factors that may
2060 * change the result (like a new metadata chunk).
2062 * If metadata is exhausted, f_bavail will be 0.
2064 static int btrfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
2066 struct btrfs_fs_info
*fs_info
= btrfs_sb(dentry
->d_sb
);
2067 struct btrfs_super_block
*disk_super
= fs_info
->super_copy
;
2068 struct btrfs_space_info
*found
;
2070 u64 total_free_data
= 0;
2071 u64 total_free_meta
= 0;
2072 int bits
= dentry
->d_sb
->s_blocksize_bits
;
2073 __be32
*fsid
= (__be32
*)fs_info
->fs_devices
->fsid
;
2074 unsigned factor
= 1;
2075 struct btrfs_block_rsv
*block_rsv
= &fs_info
->global_block_rsv
;
2081 list_for_each_entry_rcu(found
, &fs_info
->space_info
, list
) {
2082 if (found
->flags
& BTRFS_BLOCK_GROUP_DATA
) {
2085 total_free_data
+= found
->disk_total
- found
->disk_used
;
2087 btrfs_account_ro_block_groups_free_space(found
);
2089 for (i
= 0; i
< BTRFS_NR_RAID_TYPES
; i
++) {
2090 if (!list_empty(&found
->block_groups
[i
]))
2091 factor
= btrfs_bg_type_to_factor(
2092 btrfs_raid_array
[i
].bg_flag
);
2097 * Metadata in mixed block goup profiles are accounted in data
2099 if (!mixed
&& found
->flags
& BTRFS_BLOCK_GROUP_METADATA
) {
2100 if (found
->flags
& BTRFS_BLOCK_GROUP_DATA
)
2103 total_free_meta
+= found
->disk_total
-
2107 total_used
+= found
->disk_used
;
2112 buf
->f_blocks
= div_u64(btrfs_super_total_bytes(disk_super
), factor
);
2113 buf
->f_blocks
>>= bits
;
2114 buf
->f_bfree
= buf
->f_blocks
- (div_u64(total_used
, factor
) >> bits
);
2116 /* Account global block reserve as used, it's in logical size already */
2117 spin_lock(&block_rsv
->lock
);
2118 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2119 if (buf
->f_bfree
>= block_rsv
->size
>> bits
)
2120 buf
->f_bfree
-= block_rsv
->size
>> bits
;
2123 spin_unlock(&block_rsv
->lock
);
2125 buf
->f_bavail
= div_u64(total_free_data
, factor
);
2126 ret
= btrfs_calc_avail_data_space(fs_info
, &total_free_data
);
2129 buf
->f_bavail
+= div_u64(total_free_data
, factor
);
2130 buf
->f_bavail
= buf
->f_bavail
>> bits
;
2133 * We calculate the remaining metadata space minus global reserve. If
2134 * this is (supposedly) smaller than zero, there's no space. But this
2135 * does not hold in practice, the exhausted state happens where's still
2136 * some positive delta. So we apply some guesswork and compare the
2137 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2139 * We probably cannot calculate the exact threshold value because this
2140 * depends on the internal reservations requested by various
2141 * operations, so some operations that consume a few metadata will
2142 * succeed even if the Avail is zero. But this is better than the other
2148 * We only want to claim there's no available space if we can no longer
2149 * allocate chunks for our metadata profile and our global reserve will
2150 * not fit in the free metadata space. If we aren't ->full then we
2151 * still can allocate chunks and thus are fine using the currently
2152 * calculated f_bavail.
2154 if (!mixed
&& block_rsv
->space_info
->full
&&
2155 total_free_meta
- thresh
< block_rsv
->size
)
2158 buf
->f_type
= BTRFS_SUPER_MAGIC
;
2159 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
2160 buf
->f_namelen
= BTRFS_NAME_LEN
;
2162 /* We treat it as constant endianness (it doesn't matter _which_)
2163 because we want the fsid to come out the same whether mounted
2164 on a big-endian or little-endian host */
2165 buf
->f_fsid
.val
[0] = be32_to_cpu(fsid
[0]) ^ be32_to_cpu(fsid
[2]);
2166 buf
->f_fsid
.val
[1] = be32_to_cpu(fsid
[1]) ^ be32_to_cpu(fsid
[3]);
2167 /* Mask in the root object ID too, to disambiguate subvols */
2168 buf
->f_fsid
.val
[0] ^=
2169 BTRFS_I(d_inode(dentry
))->root
->root_key
.objectid
>> 32;
2170 buf
->f_fsid
.val
[1] ^=
2171 BTRFS_I(d_inode(dentry
))->root
->root_key
.objectid
;
2176 static void btrfs_kill_super(struct super_block
*sb
)
2178 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
2179 kill_anon_super(sb
);
2180 btrfs_free_fs_info(fs_info
);
2183 static struct file_system_type btrfs_fs_type
= {
2184 .owner
= THIS_MODULE
,
2186 .mount
= btrfs_mount
,
2187 .kill_sb
= btrfs_kill_super
,
2188 .fs_flags
= FS_REQUIRES_DEV
| FS_BINARY_MOUNTDATA
,
2191 static struct file_system_type btrfs_root_fs_type
= {
2192 .owner
= THIS_MODULE
,
2194 .mount
= btrfs_mount_root
,
2195 .kill_sb
= btrfs_kill_super
,
2196 .fs_flags
= FS_REQUIRES_DEV
| FS_BINARY_MOUNTDATA
,
2199 MODULE_ALIAS_FS("btrfs");
2201 static int btrfs_control_open(struct inode
*inode
, struct file
*file
)
2204 * The control file's private_data is used to hold the
2205 * transaction when it is started and is used to keep
2206 * track of whether a transaction is already in progress.
2208 file
->private_data
= NULL
;
2213 * Used by /dev/btrfs-control for devices ioctls.
2215 static long btrfs_control_ioctl(struct file
*file
, unsigned int cmd
,
2218 struct btrfs_ioctl_vol_args
*vol
;
2219 struct btrfs_device
*device
= NULL
;
2222 if (!capable(CAP_SYS_ADMIN
))
2225 vol
= memdup_user((void __user
*)arg
, sizeof(*vol
));
2227 return PTR_ERR(vol
);
2228 vol
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2231 case BTRFS_IOC_SCAN_DEV
:
2232 mutex_lock(&uuid_mutex
);
2233 device
= btrfs_scan_one_device(vol
->name
, FMODE_READ
,
2234 &btrfs_root_fs_type
);
2235 ret
= PTR_ERR_OR_ZERO(device
);
2236 mutex_unlock(&uuid_mutex
);
2238 case BTRFS_IOC_FORGET_DEV
:
2239 ret
= btrfs_forget_devices(vol
->name
);
2241 case BTRFS_IOC_DEVICES_READY
:
2242 mutex_lock(&uuid_mutex
);
2243 device
= btrfs_scan_one_device(vol
->name
, FMODE_READ
,
2244 &btrfs_root_fs_type
);
2245 if (IS_ERR(device
)) {
2246 mutex_unlock(&uuid_mutex
);
2247 ret
= PTR_ERR(device
);
2250 ret
= !(device
->fs_devices
->num_devices
==
2251 device
->fs_devices
->total_devices
);
2252 mutex_unlock(&uuid_mutex
);
2254 case BTRFS_IOC_GET_SUPPORTED_FEATURES
:
2255 ret
= btrfs_ioctl_get_supported_features((void __user
*)arg
);
2263 static int btrfs_freeze(struct super_block
*sb
)
2265 struct btrfs_trans_handle
*trans
;
2266 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
2267 struct btrfs_root
*root
= fs_info
->tree_root
;
2269 set_bit(BTRFS_FS_FROZEN
, &fs_info
->flags
);
2271 * We don't need a barrier here, we'll wait for any transaction that
2272 * could be in progress on other threads (and do delayed iputs that
2273 * we want to avoid on a frozen filesystem), or do the commit
2276 trans
= btrfs_attach_transaction_barrier(root
);
2277 if (IS_ERR(trans
)) {
2278 /* no transaction, don't bother */
2279 if (PTR_ERR(trans
) == -ENOENT
)
2281 return PTR_ERR(trans
);
2283 return btrfs_commit_transaction(trans
);
2286 static int btrfs_unfreeze(struct super_block
*sb
)
2288 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
2290 clear_bit(BTRFS_FS_FROZEN
, &fs_info
->flags
);
2294 static int btrfs_show_devname(struct seq_file
*m
, struct dentry
*root
)
2296 struct btrfs_fs_info
*fs_info
= btrfs_sb(root
->d_sb
);
2297 struct btrfs_fs_devices
*cur_devices
;
2298 struct btrfs_device
*dev
, *first_dev
= NULL
;
2299 struct list_head
*head
;
2302 * Lightweight locking of the devices. We should not need
2303 * device_list_mutex here as we only read the device data and the list
2304 * is protected by RCU. Even if a device is deleted during the list
2305 * traversals, we'll get valid data, the freeing callback will wait at
2306 * least until the rcu_read_unlock.
2309 cur_devices
= fs_info
->fs_devices
;
2310 while (cur_devices
) {
2311 head
= &cur_devices
->devices
;
2312 list_for_each_entry_rcu(dev
, head
, dev_list
) {
2313 if (test_bit(BTRFS_DEV_STATE_MISSING
, &dev
->dev_state
))
2317 if (!first_dev
|| dev
->devid
< first_dev
->devid
)
2320 cur_devices
= cur_devices
->seed
;
2324 seq_escape(m
, rcu_str_deref(first_dev
->name
), " \t\n\\");
2331 static const struct super_operations btrfs_super_ops
= {
2332 .drop_inode
= btrfs_drop_inode
,
2333 .evict_inode
= btrfs_evict_inode
,
2334 .put_super
= btrfs_put_super
,
2335 .sync_fs
= btrfs_sync_fs
,
2336 .show_options
= btrfs_show_options
,
2337 .show_devname
= btrfs_show_devname
,
2338 .alloc_inode
= btrfs_alloc_inode
,
2339 .destroy_inode
= btrfs_destroy_inode
,
2340 .free_inode
= btrfs_free_inode
,
2341 .statfs
= btrfs_statfs
,
2342 .remount_fs
= btrfs_remount
,
2343 .freeze_fs
= btrfs_freeze
,
2344 .unfreeze_fs
= btrfs_unfreeze
,
2347 static const struct file_operations btrfs_ctl_fops
= {
2348 .open
= btrfs_control_open
,
2349 .unlocked_ioctl
= btrfs_control_ioctl
,
2350 .compat_ioctl
= compat_ptr_ioctl
,
2351 .owner
= THIS_MODULE
,
2352 .llseek
= noop_llseek
,
2355 static struct miscdevice btrfs_misc
= {
2356 .minor
= BTRFS_MINOR
,
2357 .name
= "btrfs-control",
2358 .fops
= &btrfs_ctl_fops
2361 MODULE_ALIAS_MISCDEV(BTRFS_MINOR
);
2362 MODULE_ALIAS("devname:btrfs-control");
2364 static int __init
btrfs_interface_init(void)
2366 return misc_register(&btrfs_misc
);
2369 static __cold
void btrfs_interface_exit(void)
2371 misc_deregister(&btrfs_misc
);
2374 static void __init
btrfs_print_mod_info(void)
2376 static const char options
[] = ""
2377 #ifdef CONFIG_BTRFS_DEBUG
2380 #ifdef CONFIG_BTRFS_ASSERT
2383 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2384 ", integrity-checker=on"
2386 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2390 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options
);
2393 static int __init
init_btrfs_fs(void)
2399 err
= btrfs_init_sysfs();
2403 btrfs_init_compress();
2405 err
= btrfs_init_cachep();
2409 err
= extent_io_init();
2413 err
= extent_state_cache_init();
2415 goto free_extent_io
;
2417 err
= extent_map_init();
2419 goto free_extent_state_cache
;
2421 err
= ordered_data_init();
2423 goto free_extent_map
;
2425 err
= btrfs_delayed_inode_init();
2427 goto free_ordered_data
;
2429 err
= btrfs_auto_defrag_init();
2431 goto free_delayed_inode
;
2433 err
= btrfs_delayed_ref_init();
2435 goto free_auto_defrag
;
2437 err
= btrfs_prelim_ref_init();
2439 goto free_delayed_ref
;
2441 err
= btrfs_end_io_wq_init();
2443 goto free_prelim_ref
;
2445 err
= btrfs_interface_init();
2447 goto free_end_io_wq
;
2449 btrfs_init_lockdep();
2451 btrfs_print_mod_info();
2453 err
= btrfs_run_sanity_tests();
2455 goto unregister_ioctl
;
2457 err
= register_filesystem(&btrfs_fs_type
);
2459 goto unregister_ioctl
;
2464 btrfs_interface_exit();
2466 btrfs_end_io_wq_exit();
2468 btrfs_prelim_ref_exit();
2470 btrfs_delayed_ref_exit();
2472 btrfs_auto_defrag_exit();
2474 btrfs_delayed_inode_exit();
2476 ordered_data_exit();
2479 free_extent_state_cache
:
2480 extent_state_cache_exit();
2484 btrfs_destroy_cachep();
2486 btrfs_exit_compress();
2492 static void __exit
exit_btrfs_fs(void)
2494 btrfs_destroy_cachep();
2495 btrfs_delayed_ref_exit();
2496 btrfs_auto_defrag_exit();
2497 btrfs_delayed_inode_exit();
2498 btrfs_prelim_ref_exit();
2499 ordered_data_exit();
2501 extent_state_cache_exit();
2503 btrfs_interface_exit();
2504 btrfs_end_io_wq_exit();
2505 unregister_filesystem(&btrfs_fs_type
);
2507 btrfs_cleanup_fs_uuids();
2508 btrfs_exit_compress();
2511 late_initcall(init_btrfs_fs
);
2512 module_exit(exit_btrfs_fs
)
2514 MODULE_LICENSE("GPL");
2515 MODULE_SOFTDEP("pre: crc32c");
2516 MODULE_SOFTDEP("pre: xxhash64");
2517 MODULE_SOFTDEP("pre: sha256");
2518 MODULE_SOFTDEP("pre: blake2b-256");