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 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.
134 * __btrfs_std_error decodes expected errors from the caller and
135 * 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
;
147 * Special case: if the error is EROFS, and we're already
148 * under MS_RDONLY, then it is safe here.
150 if (errno
== -EROFS
&& (sb
->s_flags
& MS_RDONLY
))
154 errstr
= btrfs_decode_error(errno
);
156 struct va_format vaf
;
164 "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
165 sb
->s_id
, function
, line
, errno
, errstr
, &vaf
);
168 printk(KERN_CRIT
"BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
169 sb
->s_id
, function
, line
, errno
, errstr
);
173 /* Don't go through full error handling during mount */
174 save_error_info(fs_info
);
175 if (sb
->s_flags
& MS_BORN
)
176 btrfs_handle_error(fs_info
);
180 static const char * const logtypes
[] = {
191 void btrfs_printk(const struct btrfs_fs_info
*fs_info
, const char *fmt
, ...)
193 struct super_block
*sb
= fs_info
->sb
;
195 struct va_format vaf
;
197 const char *type
= logtypes
[4];
202 kern_level
= printk_get_level(fmt
);
204 size_t size
= printk_skip_level(fmt
) - fmt
;
205 memcpy(lvl
, fmt
, size
);
208 type
= logtypes
[kern_level
- '0'];
215 printk("%sBTRFS %s (device %s): %pV\n", lvl
, type
, sb
->s_id
, &vaf
);
222 * We only mark the transaction aborted and then set the file system read-only.
223 * This will prevent new transactions from starting or trying to join this
226 * This means that error recovery at the call site is limited to freeing
227 * any local memory allocations and passing the error code up without
228 * further cleanup. The transaction should complete as it normally would
229 * in the call path but will return -EIO.
231 * We'll complete the cleanup in btrfs_end_transaction and
232 * btrfs_commit_transaction.
235 void __btrfs_abort_transaction(struct btrfs_trans_handle
*trans
,
236 struct btrfs_root
*root
, const char *function
,
237 unsigned int line
, int errno
)
239 trans
->aborted
= errno
;
240 /* Nothing used. The other threads that have joined this
241 * transaction may be able to continue. */
242 if (!trans
->blocks_used
&& list_empty(&trans
->new_bgs
)) {
245 errstr
= btrfs_decode_error(errno
);
246 btrfs_warn(root
->fs_info
,
247 "%s:%d: Aborting unused transaction(%s).",
248 function
, line
, errstr
);
251 ACCESS_ONCE(trans
->transaction
->aborted
) = errno
;
252 /* Wake up anybody who may be waiting on this transaction */
253 wake_up(&root
->fs_info
->transaction_wait
);
254 wake_up(&root
->fs_info
->transaction_blocked_wait
);
255 __btrfs_std_error(root
->fs_info
, function
, line
, errno
, NULL
);
258 * __btrfs_panic decodes unexpected, fatal errors from the caller,
259 * issues an alert, and either panics or BUGs, depending on mount options.
262 void __btrfs_panic(struct btrfs_fs_info
*fs_info
, const char *function
,
263 unsigned int line
, int errno
, const char *fmt
, ...)
265 char *s_id
= "<unknown>";
267 struct va_format vaf
= { .fmt
= fmt
};
271 s_id
= fs_info
->sb
->s_id
;
276 errstr
= btrfs_decode_error(errno
);
277 if (fs_info
&& (fs_info
->mount_opt
& BTRFS_MOUNT_PANIC_ON_FATAL_ERROR
))
278 panic(KERN_CRIT
"BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
279 s_id
, function
, line
, &vaf
, errno
, errstr
);
281 btrfs_crit(fs_info
, "panic in %s:%d: %pV (errno=%d %s)",
282 function
, line
, &vaf
, errno
, errstr
);
284 /* Caller calls BUG() */
287 static void btrfs_put_super(struct super_block
*sb
)
289 close_ctree(btrfs_sb(sb
)->tree_root
);
293 Opt_degraded
, Opt_subvol
, Opt_subvolid
, Opt_device
, Opt_nodatasum
,
294 Opt_nodatacow
, Opt_max_inline
, Opt_alloc_start
, Opt_nobarrier
, Opt_ssd
,
295 Opt_nossd
, Opt_ssd_spread
, Opt_thread_pool
, Opt_noacl
, Opt_compress
,
296 Opt_compress_type
, Opt_compress_force
, Opt_compress_force_type
,
297 Opt_notreelog
, Opt_ratio
, Opt_flushoncommit
, Opt_discard
,
298 Opt_space_cache
, Opt_clear_cache
, Opt_user_subvol_rm_allowed
,
299 Opt_enospc_debug
, Opt_subvolrootid
, Opt_defrag
, Opt_inode_cache
,
300 Opt_no_space_cache
, Opt_recovery
, Opt_skip_balance
,
301 Opt_check_integrity
, Opt_check_integrity_including_extent_data
,
302 Opt_check_integrity_print_mask
, Opt_fatal_errors
, Opt_rescan_uuid_tree
,
303 Opt_commit_interval
, Opt_barrier
, Opt_nodefrag
, Opt_nodiscard
,
304 Opt_noenospc_debug
, Opt_noflushoncommit
, Opt_acl
, Opt_datacow
,
305 Opt_datasum
, Opt_treelog
, Opt_noinode_cache
,
306 #ifdef CONFIG_BTRFS_DEBUG
307 Opt_fragment_data
, Opt_fragment_metadata
, Opt_fragment_all
,
312 static const match_table_t tokens
= {
313 {Opt_degraded
, "degraded"},
314 {Opt_subvol
, "subvol=%s"},
315 {Opt_subvolid
, "subvolid=%s"},
316 {Opt_device
, "device=%s"},
317 {Opt_nodatasum
, "nodatasum"},
318 {Opt_datasum
, "datasum"},
319 {Opt_nodatacow
, "nodatacow"},
320 {Opt_datacow
, "datacow"},
321 {Opt_nobarrier
, "nobarrier"},
322 {Opt_barrier
, "barrier"},
323 {Opt_max_inline
, "max_inline=%s"},
324 {Opt_alloc_start
, "alloc_start=%s"},
325 {Opt_thread_pool
, "thread_pool=%d"},
326 {Opt_compress
, "compress"},
327 {Opt_compress_type
, "compress=%s"},
328 {Opt_compress_force
, "compress-force"},
329 {Opt_compress_force_type
, "compress-force=%s"},
331 {Opt_ssd_spread
, "ssd_spread"},
332 {Opt_nossd
, "nossd"},
334 {Opt_noacl
, "noacl"},
335 {Opt_notreelog
, "notreelog"},
336 {Opt_treelog
, "treelog"},
337 {Opt_flushoncommit
, "flushoncommit"},
338 {Opt_noflushoncommit
, "noflushoncommit"},
339 {Opt_ratio
, "metadata_ratio=%d"},
340 {Opt_discard
, "discard"},
341 {Opt_nodiscard
, "nodiscard"},
342 {Opt_space_cache
, "space_cache"},
343 {Opt_clear_cache
, "clear_cache"},
344 {Opt_user_subvol_rm_allowed
, "user_subvol_rm_allowed"},
345 {Opt_enospc_debug
, "enospc_debug"},
346 {Opt_noenospc_debug
, "noenospc_debug"},
347 {Opt_subvolrootid
, "subvolrootid=%d"},
348 {Opt_defrag
, "autodefrag"},
349 {Opt_nodefrag
, "noautodefrag"},
350 {Opt_inode_cache
, "inode_cache"},
351 {Opt_noinode_cache
, "noinode_cache"},
352 {Opt_no_space_cache
, "nospace_cache"},
353 {Opt_recovery
, "recovery"},
354 {Opt_skip_balance
, "skip_balance"},
355 {Opt_check_integrity
, "check_int"},
356 {Opt_check_integrity_including_extent_data
, "check_int_data"},
357 {Opt_check_integrity_print_mask
, "check_int_print_mask=%d"},
358 {Opt_rescan_uuid_tree
, "rescan_uuid_tree"},
359 {Opt_fatal_errors
, "fatal_errors=%s"},
360 {Opt_commit_interval
, "commit=%d"},
361 #ifdef CONFIG_BTRFS_DEBUG
362 {Opt_fragment_data
, "fragment=data"},
363 {Opt_fragment_metadata
, "fragment=metadata"},
364 {Opt_fragment_all
, "fragment=all"},
370 * Regular mount options parser. Everything that is needed only when
371 * reading in a new superblock is parsed here.
372 * XXX JDM: This needs to be cleaned up for remount.
374 int btrfs_parse_options(struct btrfs_root
*root
, char *options
)
376 struct btrfs_fs_info
*info
= root
->fs_info
;
377 substring_t args
[MAX_OPT_ARGS
];
378 char *p
, *num
, *orig
= NULL
;
383 bool compress_force
= false;
384 enum btrfs_compression_type saved_compress_type
;
385 bool saved_compress_force
;
388 cache_gen
= btrfs_super_cache_generation(root
->fs_info
->super_copy
);
390 btrfs_set_opt(info
->mount_opt
, SPACE_CACHE
);
396 * strsep changes the string, duplicate it because parse_options
399 options
= kstrdup(options
, GFP_NOFS
);
405 while ((p
= strsep(&options
, ",")) != NULL
) {
410 token
= match_token(p
, tokens
, args
);
413 btrfs_info(root
->fs_info
, "allowing degraded mounts");
414 btrfs_set_opt(info
->mount_opt
, DEGRADED
);
418 case Opt_subvolrootid
:
421 * These are parsed by btrfs_parse_early_options
422 * and can be happily ignored here.
426 btrfs_set_and_info(root
, NODATASUM
,
427 "setting nodatasum");
430 if (btrfs_test_opt(root
, NODATASUM
)) {
431 if (btrfs_test_opt(root
, NODATACOW
))
432 btrfs_info(root
->fs_info
, "setting datasum, datacow enabled");
434 btrfs_info(root
->fs_info
, "setting datasum");
436 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
437 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
440 if (!btrfs_test_opt(root
, NODATACOW
)) {
441 if (!btrfs_test_opt(root
, COMPRESS
) ||
442 !btrfs_test_opt(root
, FORCE_COMPRESS
)) {
443 btrfs_info(root
->fs_info
,
444 "setting nodatacow, compression disabled");
446 btrfs_info(root
->fs_info
, "setting nodatacow");
449 btrfs_clear_opt(info
->mount_opt
, COMPRESS
);
450 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
451 btrfs_set_opt(info
->mount_opt
, NODATACOW
);
452 btrfs_set_opt(info
->mount_opt
, NODATASUM
);
455 btrfs_clear_and_info(root
, NODATACOW
,
458 case Opt_compress_force
:
459 case Opt_compress_force_type
:
460 compress_force
= true;
463 case Opt_compress_type
:
464 saved_compress_type
= btrfs_test_opt(root
, COMPRESS
) ?
465 info
->compress_type
: BTRFS_COMPRESS_NONE
;
466 saved_compress_force
=
467 btrfs_test_opt(root
, FORCE_COMPRESS
);
468 if (token
== Opt_compress
||
469 token
== Opt_compress_force
||
470 strcmp(args
[0].from
, "zlib") == 0) {
471 compress_type
= "zlib";
472 info
->compress_type
= BTRFS_COMPRESS_ZLIB
;
473 btrfs_set_opt(info
->mount_opt
, COMPRESS
);
474 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
475 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
477 } else if (strcmp(args
[0].from
, "lzo") == 0) {
478 compress_type
= "lzo";
479 info
->compress_type
= BTRFS_COMPRESS_LZO
;
480 btrfs_set_opt(info
->mount_opt
, COMPRESS
);
481 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
482 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
483 btrfs_set_fs_incompat(info
, COMPRESS_LZO
);
485 } else if (strncmp(args
[0].from
, "no", 2) == 0) {
486 compress_type
= "no";
487 btrfs_clear_opt(info
->mount_opt
, COMPRESS
);
488 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
489 compress_force
= false;
496 if (compress_force
) {
497 btrfs_set_opt(info
->mount_opt
, FORCE_COMPRESS
);
500 * If we remount from compress-force=xxx to
501 * compress=xxx, we need clear FORCE_COMPRESS
502 * flag, otherwise, there is no way for users
503 * to disable forcible compression separately.
505 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
507 if ((btrfs_test_opt(root
, COMPRESS
) &&
508 (info
->compress_type
!= saved_compress_type
||
509 compress_force
!= saved_compress_force
)) ||
510 (!btrfs_test_opt(root
, COMPRESS
) &&
512 btrfs_info(root
->fs_info
,
514 (compress_force
) ? "force" : "use",
517 compress_force
= false;
520 btrfs_set_and_info(root
, SSD
,
521 "use ssd allocation scheme");
524 btrfs_set_and_info(root
, SSD_SPREAD
,
525 "use spread ssd allocation scheme");
526 btrfs_set_opt(info
->mount_opt
, SSD
);
529 btrfs_set_and_info(root
, NOSSD
,
530 "not using ssd allocation scheme");
531 btrfs_clear_opt(info
->mount_opt
, SSD
);
534 btrfs_clear_and_info(root
, NOBARRIER
,
535 "turning on barriers");
538 btrfs_set_and_info(root
, NOBARRIER
,
539 "turning off barriers");
541 case Opt_thread_pool
:
542 ret
= match_int(&args
[0], &intarg
);
545 } else if (intarg
> 0) {
546 info
->thread_pool_size
= intarg
;
553 num
= match_strdup(&args
[0]);
555 info
->max_inline
= memparse(num
, NULL
);
558 if (info
->max_inline
) {
559 info
->max_inline
= min_t(u64
,
563 btrfs_info(root
->fs_info
, "max_inline at %llu",
570 case Opt_alloc_start
:
571 num
= match_strdup(&args
[0]);
573 mutex_lock(&info
->chunk_mutex
);
574 info
->alloc_start
= memparse(num
, NULL
);
575 mutex_unlock(&info
->chunk_mutex
);
577 btrfs_info(root
->fs_info
, "allocations start at %llu",
585 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
586 root
->fs_info
->sb
->s_flags
|= MS_POSIXACL
;
589 btrfs_err(root
->fs_info
,
590 "support for ACL not compiled in!");
595 root
->fs_info
->sb
->s_flags
&= ~MS_POSIXACL
;
598 btrfs_set_and_info(root
, NOTREELOG
,
599 "disabling tree log");
602 btrfs_clear_and_info(root
, NOTREELOG
,
603 "enabling tree log");
605 case Opt_flushoncommit
:
606 btrfs_set_and_info(root
, FLUSHONCOMMIT
,
607 "turning on flush-on-commit");
609 case Opt_noflushoncommit
:
610 btrfs_clear_and_info(root
, FLUSHONCOMMIT
,
611 "turning off flush-on-commit");
614 ret
= match_int(&args
[0], &intarg
);
617 } else if (intarg
>= 0) {
618 info
->metadata_ratio
= intarg
;
619 btrfs_info(root
->fs_info
, "metadata ratio %d",
620 info
->metadata_ratio
);
627 btrfs_set_and_info(root
, DISCARD
,
628 "turning on discard");
631 btrfs_clear_and_info(root
, DISCARD
,
632 "turning off discard");
634 case Opt_space_cache
:
635 btrfs_set_and_info(root
, SPACE_CACHE
,
636 "enabling disk space caching");
638 case Opt_rescan_uuid_tree
:
639 btrfs_set_opt(info
->mount_opt
, RESCAN_UUID_TREE
);
641 case Opt_no_space_cache
:
642 btrfs_clear_and_info(root
, SPACE_CACHE
,
643 "disabling disk space caching");
645 case Opt_inode_cache
:
646 btrfs_set_pending_and_info(info
, INODE_MAP_CACHE
,
647 "enabling inode map caching");
649 case Opt_noinode_cache
:
650 btrfs_clear_pending_and_info(info
, INODE_MAP_CACHE
,
651 "disabling inode map caching");
653 case Opt_clear_cache
:
654 btrfs_set_and_info(root
, CLEAR_CACHE
,
655 "force clearing of disk cache");
657 case Opt_user_subvol_rm_allowed
:
658 btrfs_set_opt(info
->mount_opt
, USER_SUBVOL_RM_ALLOWED
);
660 case Opt_enospc_debug
:
661 btrfs_set_opt(info
->mount_opt
, ENOSPC_DEBUG
);
663 case Opt_noenospc_debug
:
664 btrfs_clear_opt(info
->mount_opt
, ENOSPC_DEBUG
);
667 btrfs_set_and_info(root
, AUTO_DEFRAG
,
668 "enabling auto defrag");
671 btrfs_clear_and_info(root
, AUTO_DEFRAG
,
672 "disabling auto defrag");
675 btrfs_info(root
->fs_info
, "enabling auto recovery");
676 btrfs_set_opt(info
->mount_opt
, RECOVERY
);
678 case Opt_skip_balance
:
679 btrfs_set_opt(info
->mount_opt
, SKIP_BALANCE
);
681 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
682 case Opt_check_integrity_including_extent_data
:
683 btrfs_info(root
->fs_info
,
684 "enabling check integrity including extent data");
685 btrfs_set_opt(info
->mount_opt
,
686 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
);
687 btrfs_set_opt(info
->mount_opt
, CHECK_INTEGRITY
);
689 case Opt_check_integrity
:
690 btrfs_info(root
->fs_info
, "enabling check integrity");
691 btrfs_set_opt(info
->mount_opt
, CHECK_INTEGRITY
);
693 case Opt_check_integrity_print_mask
:
694 ret
= match_int(&args
[0], &intarg
);
697 } else if (intarg
>= 0) {
698 info
->check_integrity_print_mask
= intarg
;
699 btrfs_info(root
->fs_info
, "check_integrity_print_mask 0x%x",
700 info
->check_integrity_print_mask
);
707 case Opt_check_integrity_including_extent_data
:
708 case Opt_check_integrity
:
709 case Opt_check_integrity_print_mask
:
710 btrfs_err(root
->fs_info
,
711 "support for check_integrity* not compiled in!");
715 case Opt_fatal_errors
:
716 if (strcmp(args
[0].from
, "panic") == 0)
717 btrfs_set_opt(info
->mount_opt
,
718 PANIC_ON_FATAL_ERROR
);
719 else if (strcmp(args
[0].from
, "bug") == 0)
720 btrfs_clear_opt(info
->mount_opt
,
721 PANIC_ON_FATAL_ERROR
);
727 case Opt_commit_interval
:
729 ret
= match_int(&args
[0], &intarg
);
731 btrfs_err(root
->fs_info
, "invalid commit interval");
737 btrfs_warn(root
->fs_info
, "excessive commit interval %d",
740 info
->commit_interval
= intarg
;
742 btrfs_info(root
->fs_info
, "using default commit interval %ds",
743 BTRFS_DEFAULT_COMMIT_INTERVAL
);
744 info
->commit_interval
= BTRFS_DEFAULT_COMMIT_INTERVAL
;
747 #ifdef CONFIG_BTRFS_DEBUG
748 case Opt_fragment_all
:
749 btrfs_info(root
->fs_info
, "fragmenting all space");
750 btrfs_set_opt(info
->mount_opt
, FRAGMENT_DATA
);
751 btrfs_set_opt(info
->mount_opt
, FRAGMENT_METADATA
);
753 case Opt_fragment_metadata
:
754 btrfs_info(root
->fs_info
, "fragmenting metadata");
755 btrfs_set_opt(info
->mount_opt
,
758 case Opt_fragment_data
:
759 btrfs_info(root
->fs_info
, "fragmenting data");
760 btrfs_set_opt(info
->mount_opt
, FRAGMENT_DATA
);
764 btrfs_info(root
->fs_info
, "unrecognized mount option '%s'", p
);
772 if (!ret
&& btrfs_test_opt(root
, SPACE_CACHE
))
773 btrfs_info(root
->fs_info
, "disk space caching is enabled");
779 * Parse mount options that are required early in the mount process.
781 * All other options will be parsed on much later in the mount process and
782 * only when we need to allocate a new super block.
784 static int btrfs_parse_early_options(const char *options
, fmode_t flags
,
785 void *holder
, char **subvol_name
, u64
*subvol_objectid
,
786 struct btrfs_fs_devices
**fs_devices
)
788 substring_t args
[MAX_OPT_ARGS
];
789 char *device_name
, *opts
, *orig
, *p
;
797 * strsep changes the string, duplicate it because parse_options
800 opts
= kstrdup(options
, GFP_KERNEL
);
805 while ((p
= strsep(&opts
, ",")) != NULL
) {
810 token
= match_token(p
, tokens
, args
);
814 *subvol_name
= match_strdup(&args
[0]);
821 num
= match_strdup(&args
[0]);
823 *subvol_objectid
= memparse(num
, NULL
);
825 /* we want the original fs_tree */
826 if (!*subvol_objectid
)
828 BTRFS_FS_TREE_OBJECTID
;
834 case Opt_subvolrootid
:
836 "BTRFS: 'subvolrootid' mount option is deprecated and has "
840 device_name
= match_strdup(&args
[0]);
845 error
= btrfs_scan_one_device(device_name
,
846 flags
, holder
, fs_devices
);
861 static char *get_subvol_name_from_objectid(struct btrfs_fs_info
*fs_info
,
864 struct btrfs_root
*root
= fs_info
->tree_root
;
865 struct btrfs_root
*fs_root
;
866 struct btrfs_root_ref
*root_ref
;
867 struct btrfs_inode_ref
*inode_ref
;
868 struct btrfs_key key
;
869 struct btrfs_path
*path
= NULL
;
870 char *name
= NULL
, *ptr
;
875 path
= btrfs_alloc_path();
880 path
->leave_spinning
= 1;
882 name
= kmalloc(PATH_MAX
, GFP_NOFS
);
887 ptr
= name
+ PATH_MAX
- 1;
891 * Walk up the subvolume trees in the tree of tree roots by root
892 * backrefs until we hit the top-level subvolume.
894 while (subvol_objectid
!= BTRFS_FS_TREE_OBJECTID
) {
895 key
.objectid
= subvol_objectid
;
896 key
.type
= BTRFS_ROOT_BACKREF_KEY
;
897 key
.offset
= (u64
)-1;
899 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
902 } else if (ret
> 0) {
903 ret
= btrfs_previous_item(root
, path
, subvol_objectid
,
904 BTRFS_ROOT_BACKREF_KEY
);
907 } else if (ret
> 0) {
913 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
914 subvol_objectid
= key
.offset
;
916 root_ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
917 struct btrfs_root_ref
);
918 len
= btrfs_root_ref_name_len(path
->nodes
[0], root_ref
);
924 read_extent_buffer(path
->nodes
[0], ptr
+ 1,
925 (unsigned long)(root_ref
+ 1), len
);
927 dirid
= btrfs_root_ref_dirid(path
->nodes
[0], root_ref
);
928 btrfs_release_path(path
);
930 key
.objectid
= subvol_objectid
;
931 key
.type
= BTRFS_ROOT_ITEM_KEY
;
932 key
.offset
= (u64
)-1;
933 fs_root
= btrfs_read_fs_root_no_name(fs_info
, &key
);
934 if (IS_ERR(fs_root
)) {
935 ret
= PTR_ERR(fs_root
);
940 * Walk up the filesystem tree by inode refs until we hit the
943 while (dirid
!= BTRFS_FIRST_FREE_OBJECTID
) {
944 key
.objectid
= dirid
;
945 key
.type
= BTRFS_INODE_REF_KEY
;
946 key
.offset
= (u64
)-1;
948 ret
= btrfs_search_slot(NULL
, fs_root
, &key
, path
, 0, 0);
951 } else if (ret
> 0) {
952 ret
= btrfs_previous_item(fs_root
, path
, dirid
,
953 BTRFS_INODE_REF_KEY
);
956 } else if (ret
> 0) {
962 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
965 inode_ref
= btrfs_item_ptr(path
->nodes
[0],
967 struct btrfs_inode_ref
);
968 len
= btrfs_inode_ref_name_len(path
->nodes
[0],
975 read_extent_buffer(path
->nodes
[0], ptr
+ 1,
976 (unsigned long)(inode_ref
+ 1), len
);
978 btrfs_release_path(path
);
982 btrfs_free_path(path
);
983 if (ptr
== name
+ PATH_MAX
- 1) {
987 memmove(name
, ptr
, name
+ PATH_MAX
- ptr
);
992 btrfs_free_path(path
);
997 static int get_default_subvol_objectid(struct btrfs_fs_info
*fs_info
, u64
*objectid
)
999 struct btrfs_root
*root
= fs_info
->tree_root
;
1000 struct btrfs_dir_item
*di
;
1001 struct btrfs_path
*path
;
1002 struct btrfs_key location
;
1005 path
= btrfs_alloc_path();
1008 path
->leave_spinning
= 1;
1011 * Find the "default" dir item which points to the root item that we
1012 * will mount by default if we haven't been given a specific subvolume
1015 dir_id
= btrfs_super_root_dir(fs_info
->super_copy
);
1016 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir_id
, "default", 7, 0);
1018 btrfs_free_path(path
);
1023 * Ok the default dir item isn't there. This is weird since
1024 * it's always been there, but don't freak out, just try and
1025 * mount the top-level subvolume.
1027 btrfs_free_path(path
);
1028 *objectid
= BTRFS_FS_TREE_OBJECTID
;
1032 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &location
);
1033 btrfs_free_path(path
);
1034 *objectid
= location
.objectid
;
1038 static int btrfs_fill_super(struct super_block
*sb
,
1039 struct btrfs_fs_devices
*fs_devices
,
1040 void *data
, int silent
)
1042 struct inode
*inode
;
1043 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1044 struct btrfs_key key
;
1047 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
1048 sb
->s_magic
= BTRFS_SUPER_MAGIC
;
1049 sb
->s_op
= &btrfs_super_ops
;
1050 sb
->s_d_op
= &btrfs_dentry_operations
;
1051 sb
->s_export_op
= &btrfs_export_ops
;
1052 sb
->s_xattr
= btrfs_xattr_handlers
;
1053 sb
->s_time_gran
= 1;
1054 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1055 sb
->s_flags
|= MS_POSIXACL
;
1057 sb
->s_flags
|= MS_I_VERSION
;
1058 sb
->s_iflags
|= SB_I_CGROUPWB
;
1059 err
= open_ctree(sb
, fs_devices
, (char *)data
);
1061 printk(KERN_ERR
"BTRFS: open_ctree failed\n");
1065 key
.objectid
= BTRFS_FIRST_FREE_OBJECTID
;
1066 key
.type
= BTRFS_INODE_ITEM_KEY
;
1068 inode
= btrfs_iget(sb
, &key
, fs_info
->fs_root
, NULL
);
1069 if (IS_ERR(inode
)) {
1070 err
= PTR_ERR(inode
);
1074 sb
->s_root
= d_make_root(inode
);
1080 save_mount_options(sb
, data
);
1081 cleancache_init_fs(sb
);
1082 sb
->s_flags
|= MS_ACTIVE
;
1086 close_ctree(fs_info
->tree_root
);
1090 int btrfs_sync_fs(struct super_block
*sb
, int wait
)
1092 struct btrfs_trans_handle
*trans
;
1093 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1094 struct btrfs_root
*root
= fs_info
->tree_root
;
1096 trace_btrfs_sync_fs(wait
);
1099 filemap_flush(fs_info
->btree_inode
->i_mapping
);
1103 btrfs_wait_ordered_roots(fs_info
, -1);
1105 trans
= btrfs_attach_transaction_barrier(root
);
1106 if (IS_ERR(trans
)) {
1107 /* no transaction, don't bother */
1108 if (PTR_ERR(trans
) == -ENOENT
) {
1110 * Exit unless we have some pending changes
1111 * that need to go through commit
1113 if (fs_info
->pending_changes
== 0)
1116 * A non-blocking test if the fs is frozen. We must not
1117 * start a new transaction here otherwise a deadlock
1118 * happens. The pending operations are delayed to the
1119 * next commit after thawing.
1121 if (__sb_start_write(sb
, SB_FREEZE_WRITE
, false))
1122 __sb_end_write(sb
, SB_FREEZE_WRITE
);
1125 trans
= btrfs_start_transaction(root
, 0);
1128 return PTR_ERR(trans
);
1130 return btrfs_commit_transaction(trans
, root
);
1133 static int btrfs_show_options(struct seq_file
*seq
, struct dentry
*dentry
)
1135 struct btrfs_fs_info
*info
= btrfs_sb(dentry
->d_sb
);
1136 struct btrfs_root
*root
= info
->tree_root
;
1137 char *compress_type
;
1139 if (btrfs_test_opt(root
, DEGRADED
))
1140 seq_puts(seq
, ",degraded");
1141 if (btrfs_test_opt(root
, NODATASUM
))
1142 seq_puts(seq
, ",nodatasum");
1143 if (btrfs_test_opt(root
, NODATACOW
))
1144 seq_puts(seq
, ",nodatacow");
1145 if (btrfs_test_opt(root
, NOBARRIER
))
1146 seq_puts(seq
, ",nobarrier");
1147 if (info
->max_inline
!= BTRFS_DEFAULT_MAX_INLINE
)
1148 seq_printf(seq
, ",max_inline=%llu", info
->max_inline
);
1149 if (info
->alloc_start
!= 0)
1150 seq_printf(seq
, ",alloc_start=%llu", info
->alloc_start
);
1151 if (info
->thread_pool_size
!= min_t(unsigned long,
1152 num_online_cpus() + 2, 8))
1153 seq_printf(seq
, ",thread_pool=%d", info
->thread_pool_size
);
1154 if (btrfs_test_opt(root
, COMPRESS
)) {
1155 if (info
->compress_type
== BTRFS_COMPRESS_ZLIB
)
1156 compress_type
= "zlib";
1158 compress_type
= "lzo";
1159 if (btrfs_test_opt(root
, FORCE_COMPRESS
))
1160 seq_printf(seq
, ",compress-force=%s", compress_type
);
1162 seq_printf(seq
, ",compress=%s", compress_type
);
1164 if (btrfs_test_opt(root
, NOSSD
))
1165 seq_puts(seq
, ",nossd");
1166 if (btrfs_test_opt(root
, SSD_SPREAD
))
1167 seq_puts(seq
, ",ssd_spread");
1168 else if (btrfs_test_opt(root
, SSD
))
1169 seq_puts(seq
, ",ssd");
1170 if (btrfs_test_opt(root
, NOTREELOG
))
1171 seq_puts(seq
, ",notreelog");
1172 if (btrfs_test_opt(root
, FLUSHONCOMMIT
))
1173 seq_puts(seq
, ",flushoncommit");
1174 if (btrfs_test_opt(root
, DISCARD
))
1175 seq_puts(seq
, ",discard");
1176 if (!(root
->fs_info
->sb
->s_flags
& MS_POSIXACL
))
1177 seq_puts(seq
, ",noacl");
1178 if (btrfs_test_opt(root
, SPACE_CACHE
))
1179 seq_puts(seq
, ",space_cache");
1181 seq_puts(seq
, ",nospace_cache");
1182 if (btrfs_test_opt(root
, RESCAN_UUID_TREE
))
1183 seq_puts(seq
, ",rescan_uuid_tree");
1184 if (btrfs_test_opt(root
, CLEAR_CACHE
))
1185 seq_puts(seq
, ",clear_cache");
1186 if (btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
1187 seq_puts(seq
, ",user_subvol_rm_allowed");
1188 if (btrfs_test_opt(root
, ENOSPC_DEBUG
))
1189 seq_puts(seq
, ",enospc_debug");
1190 if (btrfs_test_opt(root
, AUTO_DEFRAG
))
1191 seq_puts(seq
, ",autodefrag");
1192 if (btrfs_test_opt(root
, INODE_MAP_CACHE
))
1193 seq_puts(seq
, ",inode_cache");
1194 if (btrfs_test_opt(root
, SKIP_BALANCE
))
1195 seq_puts(seq
, ",skip_balance");
1196 if (btrfs_test_opt(root
, RECOVERY
))
1197 seq_puts(seq
, ",recovery");
1198 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1199 if (btrfs_test_opt(root
, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
))
1200 seq_puts(seq
, ",check_int_data");
1201 else if (btrfs_test_opt(root
, CHECK_INTEGRITY
))
1202 seq_puts(seq
, ",check_int");
1203 if (info
->check_integrity_print_mask
)
1204 seq_printf(seq
, ",check_int_print_mask=%d",
1205 info
->check_integrity_print_mask
);
1207 if (info
->metadata_ratio
)
1208 seq_printf(seq
, ",metadata_ratio=%d",
1209 info
->metadata_ratio
);
1210 if (btrfs_test_opt(root
, PANIC_ON_FATAL_ERROR
))
1211 seq_puts(seq
, ",fatal_errors=panic");
1212 if (info
->commit_interval
!= BTRFS_DEFAULT_COMMIT_INTERVAL
)
1213 seq_printf(seq
, ",commit=%d", info
->commit_interval
);
1214 #ifdef CONFIG_BTRFS_DEBUG
1215 if (btrfs_test_opt(root
, FRAGMENT_DATA
))
1216 seq_puts(seq
, ",fragment=data");
1217 if (btrfs_test_opt(root
, FRAGMENT_METADATA
))
1218 seq_puts(seq
, ",fragment=metadata");
1220 seq_printf(seq
, ",subvolid=%llu",
1221 BTRFS_I(d_inode(dentry
))->root
->root_key
.objectid
);
1222 seq_puts(seq
, ",subvol=");
1223 seq_dentry(seq
, dentry
, " \t\n\\");
1227 static int btrfs_test_super(struct super_block
*s
, void *data
)
1229 struct btrfs_fs_info
*p
= data
;
1230 struct btrfs_fs_info
*fs_info
= btrfs_sb(s
);
1232 return fs_info
->fs_devices
== p
->fs_devices
;
1235 static int btrfs_set_super(struct super_block
*s
, void *data
)
1237 int err
= set_anon_super(s
, data
);
1239 s
->s_fs_info
= data
;
1244 * subvolumes are identified by ino 256
1246 static inline int is_subvolume_inode(struct inode
*inode
)
1248 if (inode
&& inode
->i_ino
== BTRFS_FIRST_FREE_OBJECTID
)
1254 * This will add subvolid=0 to the argument string while removing any subvol=
1255 * and subvolid= arguments to make sure we get the top-level root for path
1256 * walking to the subvol we want.
1258 static char *setup_root_args(char *args
)
1260 char *buf
, *dst
, *sep
;
1263 return kstrdup("subvolid=0", GFP_NOFS
);
1265 /* The worst case is that we add ",subvolid=0" to the end. */
1266 buf
= dst
= kmalloc(strlen(args
) + strlen(",subvolid=0") + 1, GFP_NOFS
);
1271 sep
= strchrnul(args
, ',');
1272 if (!strstarts(args
, "subvol=") &&
1273 !strstarts(args
, "subvolid=")) {
1274 memcpy(dst
, args
, sep
- args
);
1283 strcpy(dst
, "subvolid=0");
1288 static struct dentry
*mount_subvol(const char *subvol_name
, u64 subvol_objectid
,
1289 int flags
, const char *device_name
,
1292 struct dentry
*root
;
1293 struct vfsmount
*mnt
= NULL
;
1297 newargs
= setup_root_args(data
);
1299 root
= ERR_PTR(-ENOMEM
);
1303 mnt
= vfs_kern_mount(&btrfs_fs_type
, flags
, device_name
, newargs
);
1304 if (PTR_ERR_OR_ZERO(mnt
) == -EBUSY
) {
1305 if (flags
& MS_RDONLY
) {
1306 mnt
= vfs_kern_mount(&btrfs_fs_type
, flags
& ~MS_RDONLY
,
1307 device_name
, newargs
);
1309 mnt
= vfs_kern_mount(&btrfs_fs_type
, flags
| MS_RDONLY
,
1310 device_name
, newargs
);
1312 root
= ERR_CAST(mnt
);
1317 down_write(&mnt
->mnt_sb
->s_umount
);
1318 ret
= btrfs_remount(mnt
->mnt_sb
, &flags
, NULL
);
1319 up_write(&mnt
->mnt_sb
->s_umount
);
1321 root
= ERR_PTR(ret
);
1327 root
= ERR_CAST(mnt
);
1333 if (!subvol_objectid
) {
1334 ret
= get_default_subvol_objectid(btrfs_sb(mnt
->mnt_sb
),
1337 root
= ERR_PTR(ret
);
1341 subvol_name
= get_subvol_name_from_objectid(btrfs_sb(mnt
->mnt_sb
),
1343 if (IS_ERR(subvol_name
)) {
1344 root
= ERR_CAST(subvol_name
);
1351 root
= mount_subtree(mnt
, subvol_name
);
1352 /* mount_subtree() drops our reference on the vfsmount. */
1355 if (!IS_ERR(root
)) {
1356 struct super_block
*s
= root
->d_sb
;
1357 struct inode
*root_inode
= d_inode(root
);
1358 u64 root_objectid
= BTRFS_I(root_inode
)->root
->root_key
.objectid
;
1361 if (!is_subvolume_inode(root_inode
)) {
1362 pr_err("BTRFS: '%s' is not a valid subvolume\n",
1366 if (subvol_objectid
&& root_objectid
!= subvol_objectid
) {
1368 * This will also catch a race condition where a
1369 * subvolume which was passed by ID is renamed and
1370 * another subvolume is renamed over the old location.
1372 pr_err("BTRFS: subvol '%s' does not match subvolid %llu\n",
1373 subvol_name
, subvol_objectid
);
1378 root
= ERR_PTR(ret
);
1379 deactivate_locked_super(s
);
1390 static int parse_security_options(char *orig_opts
,
1391 struct security_mnt_opts
*sec_opts
)
1393 char *secdata
= NULL
;
1396 secdata
= alloc_secdata();
1399 ret
= security_sb_copy_data(orig_opts
, secdata
);
1401 free_secdata(secdata
);
1404 ret
= security_sb_parse_opts_str(secdata
, sec_opts
);
1405 free_secdata(secdata
);
1409 static int setup_security_options(struct btrfs_fs_info
*fs_info
,
1410 struct super_block
*sb
,
1411 struct security_mnt_opts
*sec_opts
)
1416 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1419 ret
= security_sb_set_mnt_opts(sb
, sec_opts
, 0, NULL
);
1423 #ifdef CONFIG_SECURITY
1424 if (!fs_info
->security_opts
.num_mnt_opts
) {
1425 /* first time security setup, copy sec_opts to fs_info */
1426 memcpy(&fs_info
->security_opts
, sec_opts
, sizeof(*sec_opts
));
1429 * Since SELinux(the only one supports security_mnt_opts) does
1430 * NOT support changing context during remount/mount same sb,
1431 * This must be the same or part of the same security options,
1434 security_free_mnt_opts(sec_opts
);
1441 * Find a superblock for the given device / mount point.
1443 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1444 * for multiple device setup. Make sure to keep it in sync.
1446 static struct dentry
*btrfs_mount(struct file_system_type
*fs_type
, int flags
,
1447 const char *device_name
, void *data
)
1449 struct block_device
*bdev
= NULL
;
1450 struct super_block
*s
;
1451 struct btrfs_fs_devices
*fs_devices
= NULL
;
1452 struct btrfs_fs_info
*fs_info
= NULL
;
1453 struct security_mnt_opts new_sec_opts
;
1454 fmode_t mode
= FMODE_READ
;
1455 char *subvol_name
= NULL
;
1456 u64 subvol_objectid
= 0;
1459 if (!(flags
& MS_RDONLY
))
1460 mode
|= FMODE_WRITE
;
1462 error
= btrfs_parse_early_options(data
, mode
, fs_type
,
1463 &subvol_name
, &subvol_objectid
,
1467 return ERR_PTR(error
);
1470 if (subvol_name
|| subvol_objectid
!= BTRFS_FS_TREE_OBJECTID
) {
1471 /* mount_subvol() will free subvol_name. */
1472 return mount_subvol(subvol_name
, subvol_objectid
, flags
,
1476 security_init_mnt_opts(&new_sec_opts
);
1478 error
= parse_security_options(data
, &new_sec_opts
);
1480 return ERR_PTR(error
);
1483 error
= btrfs_scan_one_device(device_name
, mode
, fs_type
, &fs_devices
);
1485 goto error_sec_opts
;
1488 * Setup a dummy root and fs_info for test/set super. This is because
1489 * we don't actually fill this stuff out until open_ctree, but we need
1490 * it for searching for existing supers, so this lets us do that and
1491 * then open_ctree will properly initialize everything later.
1493 fs_info
= kzalloc(sizeof(struct btrfs_fs_info
), GFP_NOFS
);
1496 goto error_sec_opts
;
1499 fs_info
->fs_devices
= fs_devices
;
1501 fs_info
->super_copy
= kzalloc(BTRFS_SUPER_INFO_SIZE
, GFP_NOFS
);
1502 fs_info
->super_for_commit
= kzalloc(BTRFS_SUPER_INFO_SIZE
, GFP_NOFS
);
1503 security_init_mnt_opts(&fs_info
->security_opts
);
1504 if (!fs_info
->super_copy
|| !fs_info
->super_for_commit
) {
1509 error
= btrfs_open_devices(fs_devices
, mode
, fs_type
);
1513 if (!(flags
& MS_RDONLY
) && fs_devices
->rw_devices
== 0) {
1515 goto error_close_devices
;
1518 bdev
= fs_devices
->latest_bdev
;
1519 s
= sget(fs_type
, btrfs_test_super
, btrfs_set_super
, flags
| MS_NOSEC
,
1523 goto error_close_devices
;
1527 btrfs_close_devices(fs_devices
);
1528 free_fs_info(fs_info
);
1529 if ((flags
^ s
->s_flags
) & MS_RDONLY
)
1532 char b
[BDEVNAME_SIZE
];
1534 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
1535 btrfs_sb(s
)->bdev_holder
= fs_type
;
1536 error
= btrfs_fill_super(s
, fs_devices
, data
,
1537 flags
& MS_SILENT
? 1 : 0);
1540 deactivate_locked_super(s
);
1541 goto error_sec_opts
;
1544 fs_info
= btrfs_sb(s
);
1545 error
= setup_security_options(fs_info
, s
, &new_sec_opts
);
1547 deactivate_locked_super(s
);
1548 goto error_sec_opts
;
1551 return dget(s
->s_root
);
1553 error_close_devices
:
1554 btrfs_close_devices(fs_devices
);
1556 free_fs_info(fs_info
);
1558 security_free_mnt_opts(&new_sec_opts
);
1559 return ERR_PTR(error
);
1562 static void btrfs_resize_thread_pool(struct btrfs_fs_info
*fs_info
,
1563 int new_pool_size
, int old_pool_size
)
1565 if (new_pool_size
== old_pool_size
)
1568 fs_info
->thread_pool_size
= new_pool_size
;
1570 btrfs_info(fs_info
, "resize thread pool %d -> %d",
1571 old_pool_size
, new_pool_size
);
1573 btrfs_workqueue_set_max(fs_info
->workers
, new_pool_size
);
1574 btrfs_workqueue_set_max(fs_info
->delalloc_workers
, new_pool_size
);
1575 btrfs_workqueue_set_max(fs_info
->submit_workers
, new_pool_size
);
1576 btrfs_workqueue_set_max(fs_info
->caching_workers
, new_pool_size
);
1577 btrfs_workqueue_set_max(fs_info
->endio_workers
, new_pool_size
);
1578 btrfs_workqueue_set_max(fs_info
->endio_meta_workers
, new_pool_size
);
1579 btrfs_workqueue_set_max(fs_info
->endio_meta_write_workers
,
1581 btrfs_workqueue_set_max(fs_info
->endio_write_workers
, new_pool_size
);
1582 btrfs_workqueue_set_max(fs_info
->endio_freespace_worker
, new_pool_size
);
1583 btrfs_workqueue_set_max(fs_info
->delayed_workers
, new_pool_size
);
1584 btrfs_workqueue_set_max(fs_info
->readahead_workers
, new_pool_size
);
1585 btrfs_workqueue_set_max(fs_info
->scrub_wr_completion_workers
,
1589 static inline void btrfs_remount_prepare(struct btrfs_fs_info
*fs_info
)
1591 set_bit(BTRFS_FS_STATE_REMOUNTING
, &fs_info
->fs_state
);
1594 static inline void btrfs_remount_begin(struct btrfs_fs_info
*fs_info
,
1595 unsigned long old_opts
, int flags
)
1597 if (btrfs_raw_test_opt(old_opts
, AUTO_DEFRAG
) &&
1598 (!btrfs_raw_test_opt(fs_info
->mount_opt
, AUTO_DEFRAG
) ||
1599 (flags
& MS_RDONLY
))) {
1600 /* wait for any defraggers to finish */
1601 wait_event(fs_info
->transaction_wait
,
1602 (atomic_read(&fs_info
->defrag_running
) == 0));
1603 if (flags
& MS_RDONLY
)
1604 sync_filesystem(fs_info
->sb
);
1608 static inline void btrfs_remount_cleanup(struct btrfs_fs_info
*fs_info
,
1609 unsigned long old_opts
)
1612 * We need cleanup all defragable inodes if the autodefragment is
1613 * close or the fs is R/O.
1615 if (btrfs_raw_test_opt(old_opts
, AUTO_DEFRAG
) &&
1616 (!btrfs_raw_test_opt(fs_info
->mount_opt
, AUTO_DEFRAG
) ||
1617 (fs_info
->sb
->s_flags
& MS_RDONLY
))) {
1618 btrfs_cleanup_defrag_inodes(fs_info
);
1621 clear_bit(BTRFS_FS_STATE_REMOUNTING
, &fs_info
->fs_state
);
1624 static int btrfs_remount(struct super_block
*sb
, int *flags
, char *data
)
1626 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1627 struct btrfs_root
*root
= fs_info
->tree_root
;
1628 unsigned old_flags
= sb
->s_flags
;
1629 unsigned long old_opts
= fs_info
->mount_opt
;
1630 unsigned long old_compress_type
= fs_info
->compress_type
;
1631 u64 old_max_inline
= fs_info
->max_inline
;
1632 u64 old_alloc_start
= fs_info
->alloc_start
;
1633 int old_thread_pool_size
= fs_info
->thread_pool_size
;
1634 unsigned int old_metadata_ratio
= fs_info
->metadata_ratio
;
1637 sync_filesystem(sb
);
1638 btrfs_remount_prepare(fs_info
);
1641 struct security_mnt_opts new_sec_opts
;
1643 security_init_mnt_opts(&new_sec_opts
);
1644 ret
= parse_security_options(data
, &new_sec_opts
);
1647 ret
= setup_security_options(fs_info
, sb
,
1650 security_free_mnt_opts(&new_sec_opts
);
1655 ret
= btrfs_parse_options(root
, data
);
1661 btrfs_remount_begin(fs_info
, old_opts
, *flags
);
1662 btrfs_resize_thread_pool(fs_info
,
1663 fs_info
->thread_pool_size
, old_thread_pool_size
);
1665 if ((*flags
& MS_RDONLY
) == (sb
->s_flags
& MS_RDONLY
))
1668 if (*flags
& MS_RDONLY
) {
1670 * this also happens on 'umount -rf' or on shutdown, when
1671 * the filesystem is busy.
1673 cancel_work_sync(&fs_info
->async_reclaim_work
);
1675 /* wait for the uuid_scan task to finish */
1676 down(&fs_info
->uuid_tree_rescan_sem
);
1677 /* avoid complains from lockdep et al. */
1678 up(&fs_info
->uuid_tree_rescan_sem
);
1680 sb
->s_flags
|= MS_RDONLY
;
1683 * Setting MS_RDONLY will put the cleaner thread to
1684 * sleep at the next loop if it's already active.
1685 * If it's already asleep, we'll leave unused block
1686 * groups on disk until we're mounted read-write again
1687 * unless we clean them up here.
1689 btrfs_delete_unused_bgs(fs_info
);
1691 btrfs_dev_replace_suspend_for_unmount(fs_info
);
1692 btrfs_scrub_cancel(fs_info
);
1693 btrfs_pause_balance(fs_info
);
1695 ret
= btrfs_commit_super(root
);
1699 if (test_bit(BTRFS_FS_STATE_ERROR
, &root
->fs_info
->fs_state
)) {
1701 "Remounting read-write after error is not allowed");
1705 if (fs_info
->fs_devices
->rw_devices
== 0) {
1710 if (fs_info
->fs_devices
->missing_devices
>
1711 fs_info
->num_tolerated_disk_barrier_failures
&&
1712 !(*flags
& MS_RDONLY
)) {
1714 "too many missing devices, writeable remount is not allowed");
1719 if (btrfs_super_log_root(fs_info
->super_copy
) != 0) {
1724 ret
= btrfs_cleanup_fs_roots(fs_info
);
1728 /* recover relocation */
1729 mutex_lock(&fs_info
->cleaner_mutex
);
1730 ret
= btrfs_recover_relocation(root
);
1731 mutex_unlock(&fs_info
->cleaner_mutex
);
1735 ret
= btrfs_resume_balance_async(fs_info
);
1739 ret
= btrfs_resume_dev_replace_async(fs_info
);
1741 btrfs_warn(fs_info
, "failed to resume dev_replace");
1745 if (!fs_info
->uuid_root
) {
1746 btrfs_info(fs_info
, "creating UUID tree");
1747 ret
= btrfs_create_uuid_tree(fs_info
);
1749 btrfs_warn(fs_info
, "failed to create the UUID tree %d", ret
);
1753 sb
->s_flags
&= ~MS_RDONLY
;
1756 wake_up_process(fs_info
->transaction_kthread
);
1757 btrfs_remount_cleanup(fs_info
, old_opts
);
1761 /* We've hit an error - don't reset MS_RDONLY */
1762 if (sb
->s_flags
& MS_RDONLY
)
1763 old_flags
|= MS_RDONLY
;
1764 sb
->s_flags
= old_flags
;
1765 fs_info
->mount_opt
= old_opts
;
1766 fs_info
->compress_type
= old_compress_type
;
1767 fs_info
->max_inline
= old_max_inline
;
1768 mutex_lock(&fs_info
->chunk_mutex
);
1769 fs_info
->alloc_start
= old_alloc_start
;
1770 mutex_unlock(&fs_info
->chunk_mutex
);
1771 btrfs_resize_thread_pool(fs_info
,
1772 old_thread_pool_size
, fs_info
->thread_pool_size
);
1773 fs_info
->metadata_ratio
= old_metadata_ratio
;
1774 btrfs_remount_cleanup(fs_info
, old_opts
);
1778 /* Used to sort the devices by max_avail(descending sort) */
1779 static int btrfs_cmp_device_free_bytes(const void *dev_info1
,
1780 const void *dev_info2
)
1782 if (((struct btrfs_device_info
*)dev_info1
)->max_avail
>
1783 ((struct btrfs_device_info
*)dev_info2
)->max_avail
)
1785 else if (((struct btrfs_device_info
*)dev_info1
)->max_avail
<
1786 ((struct btrfs_device_info
*)dev_info2
)->max_avail
)
1793 * sort the devices by max_avail, in which max free extent size of each device
1794 * is stored.(Descending Sort)
1796 static inline void btrfs_descending_sort_devices(
1797 struct btrfs_device_info
*devices
,
1800 sort(devices
, nr_devices
, sizeof(struct btrfs_device_info
),
1801 btrfs_cmp_device_free_bytes
, NULL
);
1805 * The helper to calc the free space on the devices that can be used to store
1808 static int btrfs_calc_avail_data_space(struct btrfs_root
*root
, u64
*free_bytes
)
1810 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1811 struct btrfs_device_info
*devices_info
;
1812 struct btrfs_fs_devices
*fs_devices
= fs_info
->fs_devices
;
1813 struct btrfs_device
*device
;
1818 u64 min_stripe_size
;
1819 int min_stripes
= 1, num_stripes
= 1;
1820 int i
= 0, nr_devices
;
1824 * We aren't under the device list lock, so this is racey-ish, but good
1825 * enough for our purposes.
1827 nr_devices
= fs_info
->fs_devices
->open_devices
;
1830 nr_devices
= fs_info
->fs_devices
->open_devices
;
1838 devices_info
= kmalloc_array(nr_devices
, sizeof(*devices_info
),
1843 /* calc min stripe number for data space alloction */
1844 type
= btrfs_get_alloc_profile(root
, 1);
1845 if (type
& BTRFS_BLOCK_GROUP_RAID0
) {
1847 num_stripes
= nr_devices
;
1848 } else if (type
& BTRFS_BLOCK_GROUP_RAID1
) {
1851 } else if (type
& BTRFS_BLOCK_GROUP_RAID10
) {
1856 if (type
& BTRFS_BLOCK_GROUP_DUP
)
1857 min_stripe_size
= 2 * BTRFS_STRIPE_LEN
;
1859 min_stripe_size
= BTRFS_STRIPE_LEN
;
1861 if (fs_info
->alloc_start
)
1862 mutex_lock(&fs_devices
->device_list_mutex
);
1864 list_for_each_entry_rcu(device
, &fs_devices
->devices
, dev_list
) {
1865 if (!device
->in_fs_metadata
|| !device
->bdev
||
1866 device
->is_tgtdev_for_dev_replace
)
1869 if (i
>= nr_devices
)
1872 avail_space
= device
->total_bytes
- device
->bytes_used
;
1874 /* align with stripe_len */
1875 avail_space
= div_u64(avail_space
, BTRFS_STRIPE_LEN
);
1876 avail_space
*= BTRFS_STRIPE_LEN
;
1879 * In order to avoid overwritting the superblock on the drive,
1880 * btrfs starts at an offset of at least 1MB when doing chunk
1883 skip_space
= 1024 * 1024;
1885 /* user can set the offset in fs_info->alloc_start. */
1886 if (fs_info
->alloc_start
&&
1887 fs_info
->alloc_start
+ BTRFS_STRIPE_LEN
<=
1888 device
->total_bytes
) {
1890 skip_space
= max(fs_info
->alloc_start
, skip_space
);
1893 * btrfs can not use the free space in
1894 * [0, skip_space - 1], we must subtract it from the
1895 * total. In order to implement it, we account the used
1896 * space in this range first.
1898 ret
= btrfs_account_dev_extents_size(device
, 0,
1902 kfree(devices_info
);
1903 mutex_unlock(&fs_devices
->device_list_mutex
);
1909 /* calc the free space in [0, skip_space - 1] */
1910 skip_space
-= used_space
;
1914 * we can use the free space in [0, skip_space - 1], subtract
1915 * it from the total.
1917 if (avail_space
&& avail_space
>= skip_space
)
1918 avail_space
-= skip_space
;
1922 if (avail_space
< min_stripe_size
)
1925 devices_info
[i
].dev
= device
;
1926 devices_info
[i
].max_avail
= avail_space
;
1931 if (fs_info
->alloc_start
)
1932 mutex_unlock(&fs_devices
->device_list_mutex
);
1936 btrfs_descending_sort_devices(devices_info
, nr_devices
);
1940 while (nr_devices
>= min_stripes
) {
1941 if (num_stripes
> nr_devices
)
1942 num_stripes
= nr_devices
;
1944 if (devices_info
[i
].max_avail
>= min_stripe_size
) {
1948 avail_space
+= devices_info
[i
].max_avail
* num_stripes
;
1949 alloc_size
= devices_info
[i
].max_avail
;
1950 for (j
= i
+ 1 - num_stripes
; j
<= i
; j
++)
1951 devices_info
[j
].max_avail
-= alloc_size
;
1957 kfree(devices_info
);
1958 *free_bytes
= avail_space
;
1963 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1965 * If there's a redundant raid level at DATA block groups, use the respective
1966 * multiplier to scale the sizes.
1968 * Unused device space usage is based on simulating the chunk allocator
1969 * algorithm that respects the device sizes, order of allocations and the
1970 * 'alloc_start' value, this is a close approximation of the actual use but
1971 * there are other factors that may change the result (like a new metadata
1974 * If metadata is exhausted, f_bavail will be 0.
1976 * FIXME: not accurate for mixed block groups, total and free/used are ok,
1977 * available appears slightly larger.
1979 static int btrfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1981 struct btrfs_fs_info
*fs_info
= btrfs_sb(dentry
->d_sb
);
1982 struct btrfs_super_block
*disk_super
= fs_info
->super_copy
;
1983 struct list_head
*head
= &fs_info
->space_info
;
1984 struct btrfs_space_info
*found
;
1986 u64 total_free_data
= 0;
1987 u64 total_free_meta
= 0;
1988 int bits
= dentry
->d_sb
->s_blocksize_bits
;
1989 __be32
*fsid
= (__be32
*)fs_info
->fsid
;
1990 unsigned factor
= 1;
1991 struct btrfs_block_rsv
*block_rsv
= &fs_info
->global_block_rsv
;
1996 * holding chunk_muext to avoid allocating new chunks, holding
1997 * device_list_mutex to avoid the device being removed
2000 list_for_each_entry_rcu(found
, head
, list
) {
2001 if (found
->flags
& BTRFS_BLOCK_GROUP_DATA
) {
2004 total_free_data
+= found
->disk_total
- found
->disk_used
;
2006 btrfs_account_ro_block_groups_free_space(found
);
2008 for (i
= 0; i
< BTRFS_NR_RAID_TYPES
; i
++) {
2009 if (!list_empty(&found
->block_groups
[i
])) {
2011 case BTRFS_RAID_DUP
:
2012 case BTRFS_RAID_RAID1
:
2013 case BTRFS_RAID_RAID10
:
2019 if (found
->flags
& BTRFS_BLOCK_GROUP_METADATA
)
2020 total_free_meta
+= found
->disk_total
- found
->disk_used
;
2022 total_used
+= found
->disk_used
;
2027 buf
->f_blocks
= div_u64(btrfs_super_total_bytes(disk_super
), factor
);
2028 buf
->f_blocks
>>= bits
;
2029 buf
->f_bfree
= buf
->f_blocks
- (div_u64(total_used
, factor
) >> bits
);
2031 /* Account global block reserve as used, it's in logical size already */
2032 spin_lock(&block_rsv
->lock
);
2033 buf
->f_bfree
-= block_rsv
->size
>> bits
;
2034 spin_unlock(&block_rsv
->lock
);
2036 buf
->f_bavail
= div_u64(total_free_data
, factor
);
2037 ret
= btrfs_calc_avail_data_space(fs_info
->tree_root
, &total_free_data
);
2040 buf
->f_bavail
+= div_u64(total_free_data
, factor
);
2041 buf
->f_bavail
= buf
->f_bavail
>> bits
;
2044 * We calculate the remaining metadata space minus global reserve. If
2045 * this is (supposedly) smaller than zero, there's no space. But this
2046 * does not hold in practice, the exhausted state happens where's still
2047 * some positive delta. So we apply some guesswork and compare the
2048 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2050 * We probably cannot calculate the exact threshold value because this
2051 * depends on the internal reservations requested by various
2052 * operations, so some operations that consume a few metadata will
2053 * succeed even if the Avail is zero. But this is better than the other
2056 thresh
= 4 * 1024 * 1024;
2058 if (total_free_meta
- thresh
< block_rsv
->size
)
2061 buf
->f_type
= BTRFS_SUPER_MAGIC
;
2062 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
2063 buf
->f_namelen
= BTRFS_NAME_LEN
;
2065 /* We treat it as constant endianness (it doesn't matter _which_)
2066 because we want the fsid to come out the same whether mounted
2067 on a big-endian or little-endian host */
2068 buf
->f_fsid
.val
[0] = be32_to_cpu(fsid
[0]) ^ be32_to_cpu(fsid
[2]);
2069 buf
->f_fsid
.val
[1] = be32_to_cpu(fsid
[1]) ^ be32_to_cpu(fsid
[3]);
2070 /* Mask in the root object ID too, to disambiguate subvols */
2071 buf
->f_fsid
.val
[0] ^= BTRFS_I(d_inode(dentry
))->root
->objectid
>> 32;
2072 buf
->f_fsid
.val
[1] ^= BTRFS_I(d_inode(dentry
))->root
->objectid
;
2077 static void btrfs_kill_super(struct super_block
*sb
)
2079 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
2080 kill_anon_super(sb
);
2081 free_fs_info(fs_info
);
2084 static struct file_system_type btrfs_fs_type
= {
2085 .owner
= THIS_MODULE
,
2087 .mount
= btrfs_mount
,
2088 .kill_sb
= btrfs_kill_super
,
2089 .fs_flags
= FS_REQUIRES_DEV
| FS_BINARY_MOUNTDATA
,
2091 MODULE_ALIAS_FS("btrfs");
2093 static int btrfs_control_open(struct inode
*inode
, struct file
*file
)
2096 * The control file's private_data is used to hold the
2097 * transaction when it is started and is used to keep
2098 * track of whether a transaction is already in progress.
2100 file
->private_data
= NULL
;
2105 * used by btrfsctl to scan devices when no FS is mounted
2107 static long btrfs_control_ioctl(struct file
*file
, unsigned int cmd
,
2110 struct btrfs_ioctl_vol_args
*vol
;
2111 struct btrfs_fs_devices
*fs_devices
;
2114 if (!capable(CAP_SYS_ADMIN
))
2117 vol
= memdup_user((void __user
*)arg
, sizeof(*vol
));
2119 return PTR_ERR(vol
);
2122 case BTRFS_IOC_SCAN_DEV
:
2123 ret
= btrfs_scan_one_device(vol
->name
, FMODE_READ
,
2124 &btrfs_fs_type
, &fs_devices
);
2126 case BTRFS_IOC_DEVICES_READY
:
2127 ret
= btrfs_scan_one_device(vol
->name
, FMODE_READ
,
2128 &btrfs_fs_type
, &fs_devices
);
2131 ret
= !(fs_devices
->num_devices
== fs_devices
->total_devices
);
2139 static int btrfs_freeze(struct super_block
*sb
)
2141 struct btrfs_trans_handle
*trans
;
2142 struct btrfs_root
*root
= btrfs_sb(sb
)->tree_root
;
2144 trans
= btrfs_attach_transaction_barrier(root
);
2145 if (IS_ERR(trans
)) {
2146 /* no transaction, don't bother */
2147 if (PTR_ERR(trans
) == -ENOENT
)
2149 return PTR_ERR(trans
);
2151 return btrfs_commit_transaction(trans
, root
);
2154 static int btrfs_show_devname(struct seq_file
*m
, struct dentry
*root
)
2156 struct btrfs_fs_info
*fs_info
= btrfs_sb(root
->d_sb
);
2157 struct btrfs_fs_devices
*cur_devices
;
2158 struct btrfs_device
*dev
, *first_dev
= NULL
;
2159 struct list_head
*head
;
2160 struct rcu_string
*name
;
2162 mutex_lock(&fs_info
->fs_devices
->device_list_mutex
);
2163 cur_devices
= fs_info
->fs_devices
;
2164 while (cur_devices
) {
2165 head
= &cur_devices
->devices
;
2166 list_for_each_entry(dev
, head
, dev_list
) {
2171 if (!first_dev
|| dev
->devid
< first_dev
->devid
)
2174 cur_devices
= cur_devices
->seed
;
2179 name
= rcu_dereference(first_dev
->name
);
2180 seq_escape(m
, name
->str
, " \t\n\\");
2185 mutex_unlock(&fs_info
->fs_devices
->device_list_mutex
);
2189 static const struct super_operations btrfs_super_ops
= {
2190 .drop_inode
= btrfs_drop_inode
,
2191 .evict_inode
= btrfs_evict_inode
,
2192 .put_super
= btrfs_put_super
,
2193 .sync_fs
= btrfs_sync_fs
,
2194 .show_options
= btrfs_show_options
,
2195 .show_devname
= btrfs_show_devname
,
2196 .write_inode
= btrfs_write_inode
,
2197 .alloc_inode
= btrfs_alloc_inode
,
2198 .destroy_inode
= btrfs_destroy_inode
,
2199 .statfs
= btrfs_statfs
,
2200 .remount_fs
= btrfs_remount
,
2201 .freeze_fs
= btrfs_freeze
,
2204 static const struct file_operations btrfs_ctl_fops
= {
2205 .open
= btrfs_control_open
,
2206 .unlocked_ioctl
= btrfs_control_ioctl
,
2207 .compat_ioctl
= btrfs_control_ioctl
,
2208 .owner
= THIS_MODULE
,
2209 .llseek
= noop_llseek
,
2212 static struct miscdevice btrfs_misc
= {
2213 .minor
= BTRFS_MINOR
,
2214 .name
= "btrfs-control",
2215 .fops
= &btrfs_ctl_fops
2218 MODULE_ALIAS_MISCDEV(BTRFS_MINOR
);
2219 MODULE_ALIAS("devname:btrfs-control");
2221 static int btrfs_interface_init(void)
2223 return misc_register(&btrfs_misc
);
2226 static void btrfs_interface_exit(void)
2228 misc_deregister(&btrfs_misc
);
2231 static void btrfs_print_info(void)
2233 printk(KERN_INFO
"Btrfs loaded"
2234 #ifdef CONFIG_BTRFS_DEBUG
2237 #ifdef CONFIG_BTRFS_ASSERT
2240 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2241 ", integrity-checker=on"
2246 static int btrfs_run_sanity_tests(void)
2250 ret
= btrfs_init_test_fs();
2254 ret
= btrfs_test_free_space_cache();
2257 ret
= btrfs_test_extent_buffer_operations();
2260 ret
= btrfs_test_extent_io();
2263 ret
= btrfs_test_inodes();
2266 ret
= btrfs_test_qgroups();
2268 btrfs_destroy_test_fs();
2272 static int __init
init_btrfs_fs(void)
2276 err
= btrfs_hash_init();
2282 err
= btrfs_init_sysfs();
2286 btrfs_init_compress();
2288 err
= btrfs_init_cachep();
2292 err
= extent_io_init();
2296 err
= extent_map_init();
2298 goto free_extent_io
;
2300 err
= ordered_data_init();
2302 goto free_extent_map
;
2304 err
= btrfs_delayed_inode_init();
2306 goto free_ordered_data
;
2308 err
= btrfs_auto_defrag_init();
2310 goto free_delayed_inode
;
2312 err
= btrfs_delayed_ref_init();
2314 goto free_auto_defrag
;
2316 err
= btrfs_prelim_ref_init();
2318 goto free_delayed_ref
;
2320 err
= btrfs_end_io_wq_init();
2322 goto free_prelim_ref
;
2324 err
= btrfs_interface_init();
2326 goto free_end_io_wq
;
2328 btrfs_init_lockdep();
2332 err
= btrfs_run_sanity_tests();
2334 goto unregister_ioctl
;
2336 err
= register_filesystem(&btrfs_fs_type
);
2338 goto unregister_ioctl
;
2343 btrfs_interface_exit();
2345 btrfs_end_io_wq_exit();
2347 btrfs_prelim_ref_exit();
2349 btrfs_delayed_ref_exit();
2351 btrfs_auto_defrag_exit();
2353 btrfs_delayed_inode_exit();
2355 ordered_data_exit();
2361 btrfs_destroy_cachep();
2363 btrfs_exit_compress();
2370 static void __exit
exit_btrfs_fs(void)
2372 btrfs_destroy_cachep();
2373 btrfs_delayed_ref_exit();
2374 btrfs_auto_defrag_exit();
2375 btrfs_delayed_inode_exit();
2376 btrfs_prelim_ref_exit();
2377 ordered_data_exit();
2380 btrfs_interface_exit();
2381 btrfs_end_io_wq_exit();
2382 unregister_filesystem(&btrfs_fs_type
);
2384 btrfs_cleanup_fs_uuids();
2385 btrfs_exit_compress();
2389 late_initcall(init_btrfs_fs
);
2390 module_exit(exit_btrfs_fs
)
2392 MODULE_LICENSE("GPL");