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"
63 #define CREATE_TRACE_POINTS
64 #include <trace/events/btrfs.h>
66 static const struct super_operations btrfs_super_ops
;
67 static struct file_system_type btrfs_fs_type
;
69 static const char *btrfs_decode_error(int errno
)
71 char *errstr
= "unknown";
75 errstr
= "IO failure";
78 errstr
= "Out of memory";
81 errstr
= "Readonly filesystem";
84 errstr
= "Object already exists";
87 errstr
= "No space left";
90 errstr
= "No such entry";
97 static void save_error_info(struct btrfs_fs_info
*fs_info
)
100 * today we only save the error info into ram. Long term we'll
101 * also send it down to the disk
103 set_bit(BTRFS_FS_STATE_ERROR
, &fs_info
->fs_state
);
106 /* btrfs handle error by forcing the filesystem readonly */
107 static void btrfs_handle_error(struct btrfs_fs_info
*fs_info
)
109 struct super_block
*sb
= fs_info
->sb
;
111 if (sb
->s_flags
& MS_RDONLY
)
114 if (test_bit(BTRFS_FS_STATE_ERROR
, &fs_info
->fs_state
)) {
115 sb
->s_flags
|= MS_RDONLY
;
116 btrfs_info(fs_info
, "forced readonly");
118 * Note that a running device replace operation is not
119 * canceled here although there is no way to update
120 * the progress. It would add the risk of a deadlock,
121 * therefore the canceling is ommited. The only penalty
122 * is that some I/O remains active until the procedure
123 * completes. The next time when the filesystem is
124 * mounted writeable again, the device replace
125 * operation continues.
132 * __btrfs_std_error decodes expected errors from the caller and
133 * invokes the approciate error response.
135 void __btrfs_std_error(struct btrfs_fs_info
*fs_info
, const char *function
,
136 unsigned int line
, int errno
, const char *fmt
, ...)
138 struct super_block
*sb
= fs_info
->sb
;
142 * Special case: if the error is EROFS, and we're already
143 * under MS_RDONLY, then it is safe here.
145 if (errno
== -EROFS
&& (sb
->s_flags
& MS_RDONLY
))
148 errstr
= btrfs_decode_error(errno
);
150 struct va_format vaf
;
158 "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
159 sb
->s_id
, function
, line
, errno
, errstr
, &vaf
);
162 printk(KERN_CRIT
"BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
163 sb
->s_id
, function
, line
, errno
, errstr
);
166 /* Don't go through full error handling during mount */
167 save_error_info(fs_info
);
168 if (sb
->s_flags
& MS_BORN
)
169 btrfs_handle_error(fs_info
);
172 static const char * const logtypes
[] = {
183 void btrfs_printk(const struct btrfs_fs_info
*fs_info
, const char *fmt
, ...)
185 struct super_block
*sb
= fs_info
->sb
;
187 struct va_format vaf
;
189 const char *type
= logtypes
[4];
194 kern_level
= printk_get_level(fmt
);
196 size_t size
= printk_skip_level(fmt
) - fmt
;
197 memcpy(lvl
, fmt
, size
);
200 type
= logtypes
[kern_level
- '0'];
207 printk("%sBTRFS %s (device %s): %pV\n", lvl
, type
, sb
->s_id
, &vaf
);
214 void __btrfs_std_error(struct btrfs_fs_info
*fs_info
, const char *function
,
215 unsigned int line
, int errno
, const char *fmt
, ...)
217 struct super_block
*sb
= fs_info
->sb
;
220 * Special case: if the error is EROFS, and we're already
221 * under MS_RDONLY, then it is safe here.
223 if (errno
== -EROFS
&& (sb
->s_flags
& MS_RDONLY
))
226 /* Don't go through full error handling during mount */
227 if (sb
->s_flags
& MS_BORN
) {
228 save_error_info(fs_info
);
229 btrfs_handle_error(fs_info
);
235 * We only mark the transaction aborted and then set the file system read-only.
236 * This will prevent new transactions from starting or trying to join this
239 * This means that error recovery at the call site is limited to freeing
240 * any local memory allocations and passing the error code up without
241 * further cleanup. The transaction should complete as it normally would
242 * in the call path but will return -EIO.
244 * We'll complete the cleanup in btrfs_end_transaction and
245 * btrfs_commit_transaction.
247 void __btrfs_abort_transaction(struct btrfs_trans_handle
*trans
,
248 struct btrfs_root
*root
, const char *function
,
249 unsigned int line
, int errno
)
252 * Report first abort since mount
254 if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED
,
255 &root
->fs_info
->fs_state
)) {
256 WARN(1, KERN_DEBUG
"BTRFS: Transaction aborted (error %d)\n",
259 trans
->aborted
= errno
;
260 /* Nothing used. The other threads that have joined this
261 * transaction may be able to continue. */
262 if (!trans
->blocks_used
) {
265 errstr
= btrfs_decode_error(errno
);
266 btrfs_warn(root
->fs_info
,
267 "%s:%d: Aborting unused transaction(%s).",
268 function
, line
, errstr
);
271 ACCESS_ONCE(trans
->transaction
->aborted
) = errno
;
272 /* Wake up anybody who may be waiting on this transaction */
273 wake_up(&root
->fs_info
->transaction_wait
);
274 wake_up(&root
->fs_info
->transaction_blocked_wait
);
275 __btrfs_std_error(root
->fs_info
, function
, line
, errno
, NULL
);
278 * __btrfs_panic decodes unexpected, fatal errors from the caller,
279 * issues an alert, and either panics or BUGs, depending on mount options.
281 void __btrfs_panic(struct btrfs_fs_info
*fs_info
, const char *function
,
282 unsigned int line
, int errno
, const char *fmt
, ...)
284 char *s_id
= "<unknown>";
286 struct va_format vaf
= { .fmt
= fmt
};
290 s_id
= fs_info
->sb
->s_id
;
295 errstr
= btrfs_decode_error(errno
);
296 if (fs_info
&& (fs_info
->mount_opt
& BTRFS_MOUNT_PANIC_ON_FATAL_ERROR
))
297 panic(KERN_CRIT
"BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
298 s_id
, function
, line
, &vaf
, errno
, errstr
);
300 btrfs_crit(fs_info
, "panic in %s:%d: %pV (errno=%d %s)",
301 function
, line
, &vaf
, errno
, errstr
);
303 /* Caller calls BUG() */
306 static void btrfs_put_super(struct super_block
*sb
)
308 (void)close_ctree(btrfs_sb(sb
)->tree_root
);
309 /* FIXME: need to fix VFS to return error? */
310 /* AV: return it _where_? ->put_super() can be triggered by any number
311 * of async events, up to and including delivery of SIGKILL to the
312 * last process that kept it busy. Or segfault in the aforementioned
313 * process... Whom would you report that to?
318 Opt_degraded
, Opt_subvol
, Opt_subvolid
, Opt_device
, Opt_nodatasum
,
319 Opt_nodatacow
, Opt_max_inline
, Opt_alloc_start
, Opt_nobarrier
, Opt_ssd
,
320 Opt_nossd
, Opt_ssd_spread
, Opt_thread_pool
, Opt_noacl
, Opt_compress
,
321 Opt_compress_type
, Opt_compress_force
, Opt_compress_force_type
,
322 Opt_notreelog
, Opt_ratio
, Opt_flushoncommit
, Opt_discard
,
323 Opt_space_cache
, Opt_clear_cache
, Opt_user_subvol_rm_allowed
,
324 Opt_enospc_debug
, Opt_subvolrootid
, Opt_defrag
, Opt_inode_cache
,
325 Opt_no_space_cache
, Opt_recovery
, Opt_skip_balance
,
326 Opt_check_integrity
, Opt_check_integrity_including_extent_data
,
327 Opt_check_integrity_print_mask
, Opt_fatal_errors
, Opt_rescan_uuid_tree
,
328 Opt_commit_interval
, Opt_barrier
, Opt_nodefrag
, Opt_nodiscard
,
329 Opt_noenospc_debug
, Opt_noflushoncommit
, Opt_acl
, Opt_datacow
,
330 Opt_datasum
, Opt_treelog
, Opt_noinode_cache
,
334 static match_table_t tokens
= {
335 {Opt_degraded
, "degraded"},
336 {Opt_subvol
, "subvol=%s"},
337 {Opt_subvolid
, "subvolid=%s"},
338 {Opt_device
, "device=%s"},
339 {Opt_nodatasum
, "nodatasum"},
340 {Opt_datasum
, "datasum"},
341 {Opt_nodatacow
, "nodatacow"},
342 {Opt_datacow
, "datacow"},
343 {Opt_nobarrier
, "nobarrier"},
344 {Opt_barrier
, "barrier"},
345 {Opt_max_inline
, "max_inline=%s"},
346 {Opt_alloc_start
, "alloc_start=%s"},
347 {Opt_thread_pool
, "thread_pool=%d"},
348 {Opt_compress
, "compress"},
349 {Opt_compress_type
, "compress=%s"},
350 {Opt_compress_force
, "compress-force"},
351 {Opt_compress_force_type
, "compress-force=%s"},
353 {Opt_ssd_spread
, "ssd_spread"},
354 {Opt_nossd
, "nossd"},
356 {Opt_noacl
, "noacl"},
357 {Opt_notreelog
, "notreelog"},
358 {Opt_treelog
, "treelog"},
359 {Opt_flushoncommit
, "flushoncommit"},
360 {Opt_noflushoncommit
, "noflushoncommit"},
361 {Opt_ratio
, "metadata_ratio=%d"},
362 {Opt_discard
, "discard"},
363 {Opt_nodiscard
, "nodiscard"},
364 {Opt_space_cache
, "space_cache"},
365 {Opt_clear_cache
, "clear_cache"},
366 {Opt_user_subvol_rm_allowed
, "user_subvol_rm_allowed"},
367 {Opt_enospc_debug
, "enospc_debug"},
368 {Opt_noenospc_debug
, "noenospc_debug"},
369 {Opt_subvolrootid
, "subvolrootid=%d"},
370 {Opt_defrag
, "autodefrag"},
371 {Opt_nodefrag
, "noautodefrag"},
372 {Opt_inode_cache
, "inode_cache"},
373 {Opt_noinode_cache
, "noinode_cache"},
374 {Opt_no_space_cache
, "nospace_cache"},
375 {Opt_recovery
, "recovery"},
376 {Opt_skip_balance
, "skip_balance"},
377 {Opt_check_integrity
, "check_int"},
378 {Opt_check_integrity_including_extent_data
, "check_int_data"},
379 {Opt_check_integrity_print_mask
, "check_int_print_mask=%d"},
380 {Opt_rescan_uuid_tree
, "rescan_uuid_tree"},
381 {Opt_fatal_errors
, "fatal_errors=%s"},
382 {Opt_commit_interval
, "commit=%d"},
386 #define btrfs_set_and_info(root, opt, fmt, args...) \
388 if (!btrfs_test_opt(root, opt)) \
389 btrfs_info(root->fs_info, fmt, ##args); \
390 btrfs_set_opt(root->fs_info->mount_opt, opt); \
393 #define btrfs_clear_and_info(root, opt, fmt, args...) \
395 if (btrfs_test_opt(root, opt)) \
396 btrfs_info(root->fs_info, fmt, ##args); \
397 btrfs_clear_opt(root->fs_info->mount_opt, opt); \
401 * Regular mount options parser. Everything that is needed only when
402 * reading in a new superblock is parsed here.
403 * XXX JDM: This needs to be cleaned up for remount.
405 int btrfs_parse_options(struct btrfs_root
*root
, char *options
)
407 struct btrfs_fs_info
*info
= root
->fs_info
;
408 substring_t args
[MAX_OPT_ARGS
];
409 char *p
, *num
, *orig
= NULL
;
414 bool compress_force
= false;
415 bool compress
= false;
417 cache_gen
= btrfs_super_cache_generation(root
->fs_info
->super_copy
);
419 btrfs_set_opt(info
->mount_opt
, SPACE_CACHE
);
425 * strsep changes the string, duplicate it because parse_options
428 options
= kstrdup(options
, GFP_NOFS
);
434 while ((p
= strsep(&options
, ",")) != NULL
) {
439 token
= match_token(p
, tokens
, args
);
442 btrfs_info(root
->fs_info
, "allowing degraded mounts");
443 btrfs_set_opt(info
->mount_opt
, DEGRADED
);
447 case Opt_subvolrootid
:
450 * These are parsed by btrfs_parse_early_options
451 * and can be happily ignored here.
455 btrfs_set_and_info(root
, NODATASUM
,
456 "setting nodatasum");
459 if (btrfs_test_opt(root
, NODATASUM
)) {
460 if (btrfs_test_opt(root
, NODATACOW
))
461 btrfs_info(root
->fs_info
, "setting datasum, datacow enabled");
463 btrfs_info(root
->fs_info
, "setting datasum");
465 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
466 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
469 if (!btrfs_test_opt(root
, NODATACOW
)) {
470 if (!btrfs_test_opt(root
, COMPRESS
) ||
471 !btrfs_test_opt(root
, FORCE_COMPRESS
)) {
472 btrfs_info(root
->fs_info
,
473 "setting nodatacow, compression disabled");
475 btrfs_info(root
->fs_info
, "setting nodatacow");
478 btrfs_clear_opt(info
->mount_opt
, COMPRESS
);
479 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
480 btrfs_set_opt(info
->mount_opt
, NODATACOW
);
481 btrfs_set_opt(info
->mount_opt
, NODATASUM
);
484 btrfs_clear_and_info(root
, NODATACOW
,
487 case Opt_compress_force
:
488 case Opt_compress_force_type
:
489 compress_force
= true;
492 case Opt_compress_type
:
494 if (token
== Opt_compress
||
495 token
== Opt_compress_force
||
496 strcmp(args
[0].from
, "zlib") == 0) {
497 compress_type
= "zlib";
498 info
->compress_type
= BTRFS_COMPRESS_ZLIB
;
499 btrfs_set_opt(info
->mount_opt
, COMPRESS
);
500 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
501 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
502 } else if (strcmp(args
[0].from
, "lzo") == 0) {
503 compress_type
= "lzo";
504 info
->compress_type
= BTRFS_COMPRESS_LZO
;
505 btrfs_set_opt(info
->mount_opt
, COMPRESS
);
506 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
507 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
508 btrfs_set_fs_incompat(info
, COMPRESS_LZO
);
509 } else if (strncmp(args
[0].from
, "no", 2) == 0) {
510 compress_type
= "no";
511 btrfs_clear_opt(info
->mount_opt
, COMPRESS
);
512 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
513 compress_force
= false;
519 if (compress_force
) {
520 btrfs_set_and_info(root
, FORCE_COMPRESS
,
521 "force %s compression",
523 } else if (compress
) {
524 if (!btrfs_test_opt(root
, COMPRESS
))
525 btrfs_info(root
->fs_info
,
526 "btrfs: use %s compression\n",
531 btrfs_set_and_info(root
, SSD
,
532 "use ssd allocation scheme");
535 btrfs_set_and_info(root
, SSD_SPREAD
,
536 "use spread ssd allocation scheme");
539 btrfs_clear_and_info(root
, NOSSD
,
540 "not using ssd allocation scheme");
541 btrfs_clear_opt(info
->mount_opt
, SSD
);
544 btrfs_clear_and_info(root
, NOBARRIER
,
545 "turning on barriers");
548 btrfs_set_and_info(root
, NOBARRIER
,
549 "turning off barriers");
551 case Opt_thread_pool
:
552 ret
= match_int(&args
[0], &intarg
);
555 } else if (intarg
> 0) {
556 info
->thread_pool_size
= intarg
;
563 num
= match_strdup(&args
[0]);
565 info
->max_inline
= memparse(num
, NULL
);
568 if (info
->max_inline
) {
569 info
->max_inline
= max_t(u64
,
573 btrfs_info(root
->fs_info
, "max_inline at %llu",
580 case Opt_alloc_start
:
581 num
= match_strdup(&args
[0]);
583 mutex_lock(&info
->chunk_mutex
);
584 info
->alloc_start
= memparse(num
, NULL
);
585 mutex_unlock(&info
->chunk_mutex
);
587 btrfs_info(root
->fs_info
, "allocations start at %llu",
595 root
->fs_info
->sb
->s_flags
|= MS_POSIXACL
;
598 root
->fs_info
->sb
->s_flags
&= ~MS_POSIXACL
;
601 btrfs_set_and_info(root
, NOTREELOG
,
602 "disabling tree log");
605 btrfs_clear_and_info(root
, NOTREELOG
,
606 "enabling tree log");
608 case Opt_flushoncommit
:
609 btrfs_set_and_info(root
, FLUSHONCOMMIT
,
610 "turning on flush-on-commit");
612 case Opt_noflushoncommit
:
613 btrfs_clear_and_info(root
, FLUSHONCOMMIT
,
614 "turning off flush-on-commit");
617 ret
= match_int(&args
[0], &intarg
);
620 } else if (intarg
>= 0) {
621 info
->metadata_ratio
= intarg
;
622 btrfs_info(root
->fs_info
, "metadata ratio %d",
623 info
->metadata_ratio
);
630 btrfs_set_and_info(root
, DISCARD
,
631 "turning on discard");
634 btrfs_clear_and_info(root
, DISCARD
,
635 "turning off discard");
637 case Opt_space_cache
:
638 btrfs_set_and_info(root
, SPACE_CACHE
,
639 "enabling disk space caching");
641 case Opt_rescan_uuid_tree
:
642 btrfs_set_opt(info
->mount_opt
, RESCAN_UUID_TREE
);
644 case Opt_no_space_cache
:
645 btrfs_clear_and_info(root
, SPACE_CACHE
,
646 "disabling disk space caching");
648 case Opt_inode_cache
:
649 btrfs_set_and_info(root
, CHANGE_INODE_CACHE
,
650 "enabling inode map caching");
652 case Opt_noinode_cache
:
653 btrfs_clear_and_info(root
, CHANGE_INODE_CACHE
,
654 "disabling inode map caching");
656 case Opt_clear_cache
:
657 btrfs_set_and_info(root
, CLEAR_CACHE
,
658 "force clearing of disk cache");
660 case Opt_user_subvol_rm_allowed
:
661 btrfs_set_opt(info
->mount_opt
, USER_SUBVOL_RM_ALLOWED
);
663 case Opt_enospc_debug
:
664 btrfs_set_opt(info
->mount_opt
, ENOSPC_DEBUG
);
666 case Opt_noenospc_debug
:
667 btrfs_clear_opt(info
->mount_opt
, ENOSPC_DEBUG
);
670 btrfs_set_and_info(root
, AUTO_DEFRAG
,
671 "enabling auto defrag");
674 btrfs_clear_and_info(root
, AUTO_DEFRAG
,
675 "disabling auto defrag");
678 btrfs_info(root
->fs_info
, "enabling auto recovery");
679 btrfs_set_opt(info
->mount_opt
, RECOVERY
);
681 case Opt_skip_balance
:
682 btrfs_set_opt(info
->mount_opt
, SKIP_BALANCE
);
684 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
685 case Opt_check_integrity_including_extent_data
:
686 btrfs_info(root
->fs_info
,
687 "enabling check integrity including extent data");
688 btrfs_set_opt(info
->mount_opt
,
689 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
);
690 btrfs_set_opt(info
->mount_opt
, CHECK_INTEGRITY
);
692 case Opt_check_integrity
:
693 btrfs_info(root
->fs_info
, "enabling check integrity");
694 btrfs_set_opt(info
->mount_opt
, CHECK_INTEGRITY
);
696 case Opt_check_integrity_print_mask
:
697 ret
= match_int(&args
[0], &intarg
);
700 } else if (intarg
>= 0) {
701 info
->check_integrity_print_mask
= intarg
;
702 btrfs_info(root
->fs_info
, "check_integrity_print_mask 0x%x",
703 info
->check_integrity_print_mask
);
710 case Opt_check_integrity_including_extent_data
:
711 case Opt_check_integrity
:
712 case Opt_check_integrity_print_mask
:
713 btrfs_err(root
->fs_info
,
714 "support for check_integrity* not compiled in!");
718 case Opt_fatal_errors
:
719 if (strcmp(args
[0].from
, "panic") == 0)
720 btrfs_set_opt(info
->mount_opt
,
721 PANIC_ON_FATAL_ERROR
);
722 else if (strcmp(args
[0].from
, "bug") == 0)
723 btrfs_clear_opt(info
->mount_opt
,
724 PANIC_ON_FATAL_ERROR
);
730 case Opt_commit_interval
:
732 ret
= match_int(&args
[0], &intarg
);
734 btrfs_err(root
->fs_info
, "invalid commit interval");
740 btrfs_warn(root
->fs_info
, "excessive commit interval %d",
743 info
->commit_interval
= intarg
;
745 btrfs_info(root
->fs_info
, "using default commit interval %ds",
746 BTRFS_DEFAULT_COMMIT_INTERVAL
);
747 info
->commit_interval
= BTRFS_DEFAULT_COMMIT_INTERVAL
;
751 btrfs_info(root
->fs_info
, "unrecognized mount option '%s'", p
);
759 if (!ret
&& btrfs_test_opt(root
, SPACE_CACHE
))
760 btrfs_info(root
->fs_info
, "disk space caching is enabled");
766 * Parse mount options that are required early in the mount process.
768 * All other options will be parsed on much later in the mount process and
769 * only when we need to allocate a new super block.
771 static int btrfs_parse_early_options(const char *options
, fmode_t flags
,
772 void *holder
, char **subvol_name
, u64
*subvol_objectid
,
773 struct btrfs_fs_devices
**fs_devices
)
775 substring_t args
[MAX_OPT_ARGS
];
776 char *device_name
, *opts
, *orig
, *p
;
784 * strsep changes the string, duplicate it because parse_options
787 opts
= kstrdup(options
, GFP_KERNEL
);
792 while ((p
= strsep(&opts
, ",")) != NULL
) {
797 token
= match_token(p
, tokens
, args
);
801 *subvol_name
= match_strdup(&args
[0]);
808 num
= match_strdup(&args
[0]);
810 *subvol_objectid
= memparse(num
, NULL
);
812 /* we want the original fs_tree */
813 if (!*subvol_objectid
)
815 BTRFS_FS_TREE_OBJECTID
;
821 case Opt_subvolrootid
:
823 "BTRFS: 'subvolrootid' mount option is deprecated and has "
827 device_name
= match_strdup(&args
[0]);
832 error
= btrfs_scan_one_device(device_name
,
833 flags
, holder
, fs_devices
);
848 static struct dentry
*get_default_root(struct super_block
*sb
,
851 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
852 struct btrfs_root
*root
= fs_info
->tree_root
;
853 struct btrfs_root
*new_root
;
854 struct btrfs_dir_item
*di
;
855 struct btrfs_path
*path
;
856 struct btrfs_key location
;
862 * We have a specific subvol we want to mount, just setup location and
863 * go look up the root.
865 if (subvol_objectid
) {
866 location
.objectid
= subvol_objectid
;
867 location
.type
= BTRFS_ROOT_ITEM_KEY
;
868 location
.offset
= (u64
)-1;
872 path
= btrfs_alloc_path();
874 return ERR_PTR(-ENOMEM
);
875 path
->leave_spinning
= 1;
878 * Find the "default" dir item which points to the root item that we
879 * will mount by default if we haven't been given a specific subvolume
882 dir_id
= btrfs_super_root_dir(fs_info
->super_copy
);
883 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir_id
, "default", 7, 0);
885 btrfs_free_path(path
);
890 * Ok the default dir item isn't there. This is weird since
891 * it's always been there, but don't freak out, just try and
892 * mount to root most subvolume.
894 btrfs_free_path(path
);
895 dir_id
= BTRFS_FIRST_FREE_OBJECTID
;
896 new_root
= fs_info
->fs_root
;
900 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &location
);
901 btrfs_free_path(path
);
904 new_root
= btrfs_read_fs_root_no_name(fs_info
, &location
);
905 if (IS_ERR(new_root
))
906 return ERR_CAST(new_root
);
908 dir_id
= btrfs_root_dirid(&new_root
->root_item
);
910 location
.objectid
= dir_id
;
911 location
.type
= BTRFS_INODE_ITEM_KEY
;
914 inode
= btrfs_iget(sb
, &location
, new_root
, &new);
916 return ERR_CAST(inode
);
919 * If we're just mounting the root most subvol put the inode and return
920 * a reference to the dentry. We will have already gotten a reference
921 * to the inode in btrfs_fill_super so we're good to go.
923 if (!new && sb
->s_root
->d_inode
== inode
) {
925 return dget(sb
->s_root
);
928 return d_obtain_alias(inode
);
931 static int btrfs_fill_super(struct super_block
*sb
,
932 struct btrfs_fs_devices
*fs_devices
,
933 void *data
, int silent
)
936 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
937 struct btrfs_key key
;
940 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
941 sb
->s_magic
= BTRFS_SUPER_MAGIC
;
942 sb
->s_op
= &btrfs_super_ops
;
943 sb
->s_d_op
= &btrfs_dentry_operations
;
944 sb
->s_export_op
= &btrfs_export_ops
;
945 sb
->s_xattr
= btrfs_xattr_handlers
;
947 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
948 sb
->s_flags
|= MS_POSIXACL
;
950 sb
->s_flags
|= MS_I_VERSION
;
951 err
= open_ctree(sb
, fs_devices
, (char *)data
);
953 printk(KERN_ERR
"BTRFS: open_ctree failed\n");
957 key
.objectid
= BTRFS_FIRST_FREE_OBJECTID
;
958 key
.type
= BTRFS_INODE_ITEM_KEY
;
960 inode
= btrfs_iget(sb
, &key
, fs_info
->fs_root
, NULL
);
962 err
= PTR_ERR(inode
);
966 sb
->s_root
= d_make_root(inode
);
972 save_mount_options(sb
, data
);
973 cleancache_init_fs(sb
);
974 sb
->s_flags
|= MS_ACTIVE
;
978 close_ctree(fs_info
->tree_root
);
982 int btrfs_sync_fs(struct super_block
*sb
, int wait
)
984 struct btrfs_trans_handle
*trans
;
985 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
986 struct btrfs_root
*root
= fs_info
->tree_root
;
988 trace_btrfs_sync_fs(wait
);
991 filemap_flush(fs_info
->btree_inode
->i_mapping
);
995 btrfs_wait_ordered_roots(fs_info
, -1);
997 trans
= btrfs_attach_transaction_barrier(root
);
999 /* no transaction, don't bother */
1000 if (PTR_ERR(trans
) == -ENOENT
)
1002 return PTR_ERR(trans
);
1004 return btrfs_commit_transaction(trans
, root
);
1007 static int btrfs_show_options(struct seq_file
*seq
, struct dentry
*dentry
)
1009 struct btrfs_fs_info
*info
= btrfs_sb(dentry
->d_sb
);
1010 struct btrfs_root
*root
= info
->tree_root
;
1011 char *compress_type
;
1013 if (btrfs_test_opt(root
, DEGRADED
))
1014 seq_puts(seq
, ",degraded");
1015 if (btrfs_test_opt(root
, NODATASUM
))
1016 seq_puts(seq
, ",nodatasum");
1017 if (btrfs_test_opt(root
, NODATACOW
))
1018 seq_puts(seq
, ",nodatacow");
1019 if (btrfs_test_opt(root
, NOBARRIER
))
1020 seq_puts(seq
, ",nobarrier");
1021 if (info
->max_inline
!= 8192 * 1024)
1022 seq_printf(seq
, ",max_inline=%llu", info
->max_inline
);
1023 if (info
->alloc_start
!= 0)
1024 seq_printf(seq
, ",alloc_start=%llu", info
->alloc_start
);
1025 if (info
->thread_pool_size
!= min_t(unsigned long,
1026 num_online_cpus() + 2, 8))
1027 seq_printf(seq
, ",thread_pool=%d", info
->thread_pool_size
);
1028 if (btrfs_test_opt(root
, COMPRESS
)) {
1029 if (info
->compress_type
== BTRFS_COMPRESS_ZLIB
)
1030 compress_type
= "zlib";
1032 compress_type
= "lzo";
1033 if (btrfs_test_opt(root
, FORCE_COMPRESS
))
1034 seq_printf(seq
, ",compress-force=%s", compress_type
);
1036 seq_printf(seq
, ",compress=%s", compress_type
);
1038 if (btrfs_test_opt(root
, NOSSD
))
1039 seq_puts(seq
, ",nossd");
1040 if (btrfs_test_opt(root
, SSD_SPREAD
))
1041 seq_puts(seq
, ",ssd_spread");
1042 else if (btrfs_test_opt(root
, SSD
))
1043 seq_puts(seq
, ",ssd");
1044 if (btrfs_test_opt(root
, NOTREELOG
))
1045 seq_puts(seq
, ",notreelog");
1046 if (btrfs_test_opt(root
, FLUSHONCOMMIT
))
1047 seq_puts(seq
, ",flushoncommit");
1048 if (btrfs_test_opt(root
, DISCARD
))
1049 seq_puts(seq
, ",discard");
1050 if (!(root
->fs_info
->sb
->s_flags
& MS_POSIXACL
))
1051 seq_puts(seq
, ",noacl");
1052 if (btrfs_test_opt(root
, SPACE_CACHE
))
1053 seq_puts(seq
, ",space_cache");
1055 seq_puts(seq
, ",nospace_cache");
1056 if (btrfs_test_opt(root
, RESCAN_UUID_TREE
))
1057 seq_puts(seq
, ",rescan_uuid_tree");
1058 if (btrfs_test_opt(root
, CLEAR_CACHE
))
1059 seq_puts(seq
, ",clear_cache");
1060 if (btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
1061 seq_puts(seq
, ",user_subvol_rm_allowed");
1062 if (btrfs_test_opt(root
, ENOSPC_DEBUG
))
1063 seq_puts(seq
, ",enospc_debug");
1064 if (btrfs_test_opt(root
, AUTO_DEFRAG
))
1065 seq_puts(seq
, ",autodefrag");
1066 if (btrfs_test_opt(root
, INODE_MAP_CACHE
))
1067 seq_puts(seq
, ",inode_cache");
1068 if (btrfs_test_opt(root
, SKIP_BALANCE
))
1069 seq_puts(seq
, ",skip_balance");
1070 if (btrfs_test_opt(root
, RECOVERY
))
1071 seq_puts(seq
, ",recovery");
1072 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1073 if (btrfs_test_opt(root
, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
))
1074 seq_puts(seq
, ",check_int_data");
1075 else if (btrfs_test_opt(root
, CHECK_INTEGRITY
))
1076 seq_puts(seq
, ",check_int");
1077 if (info
->check_integrity_print_mask
)
1078 seq_printf(seq
, ",check_int_print_mask=%d",
1079 info
->check_integrity_print_mask
);
1081 if (info
->metadata_ratio
)
1082 seq_printf(seq
, ",metadata_ratio=%d",
1083 info
->metadata_ratio
);
1084 if (btrfs_test_opt(root
, PANIC_ON_FATAL_ERROR
))
1085 seq_puts(seq
, ",fatal_errors=panic");
1086 if (info
->commit_interval
!= BTRFS_DEFAULT_COMMIT_INTERVAL
)
1087 seq_printf(seq
, ",commit=%d", info
->commit_interval
);
1091 static int btrfs_test_super(struct super_block
*s
, void *data
)
1093 struct btrfs_fs_info
*p
= data
;
1094 struct btrfs_fs_info
*fs_info
= btrfs_sb(s
);
1096 return fs_info
->fs_devices
== p
->fs_devices
;
1099 static int btrfs_set_super(struct super_block
*s
, void *data
)
1101 int err
= set_anon_super(s
, data
);
1103 s
->s_fs_info
= data
;
1108 * subvolumes are identified by ino 256
1110 static inline int is_subvolume_inode(struct inode
*inode
)
1112 if (inode
&& inode
->i_ino
== BTRFS_FIRST_FREE_OBJECTID
)
1118 * This will strip out the subvol=%s argument for an argument string and add
1119 * subvolid=0 to make sure we get the actual tree root for path walking to the
1122 static char *setup_root_args(char *args
)
1124 unsigned len
= strlen(args
) + 2 + 1;
1125 char *src
, *dst
, *buf
;
1128 * We need the same args as before, but with this substitution:
1129 * s!subvol=[^,]+!subvolid=0!
1131 * Since the replacement string is up to 2 bytes longer than the
1132 * original, allocate strlen(args) + 2 + 1 bytes.
1135 src
= strstr(args
, "subvol=");
1136 /* This shouldn't happen, but just in case.. */
1140 buf
= dst
= kmalloc(len
, GFP_NOFS
);
1145 * If the subvol= arg is not at the start of the string,
1146 * copy whatever precedes it into buf.
1151 dst
+= strlen(args
);
1154 strcpy(dst
, "subvolid=0");
1155 dst
+= strlen("subvolid=0");
1158 * If there is a "," after the original subvol=... string,
1159 * copy that suffix into our buffer. Otherwise, we're done.
1161 src
= strchr(src
, ',');
1168 static struct dentry
*mount_subvol(const char *subvol_name
, int flags
,
1169 const char *device_name
, char *data
)
1171 struct dentry
*root
;
1172 struct vfsmount
*mnt
;
1175 newargs
= setup_root_args(data
);
1177 return ERR_PTR(-ENOMEM
);
1178 mnt
= vfs_kern_mount(&btrfs_fs_type
, flags
, device_name
,
1182 return ERR_CAST(mnt
);
1184 root
= mount_subtree(mnt
, subvol_name
);
1186 if (!IS_ERR(root
) && !is_subvolume_inode(root
->d_inode
)) {
1187 struct super_block
*s
= root
->d_sb
;
1189 root
= ERR_PTR(-EINVAL
);
1190 deactivate_locked_super(s
);
1191 printk(KERN_ERR
"BTRFS: '%s' is not a valid subvolume\n",
1199 * Find a superblock for the given device / mount point.
1201 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1202 * for multiple device setup. Make sure to keep it in sync.
1204 static struct dentry
*btrfs_mount(struct file_system_type
*fs_type
, int flags
,
1205 const char *device_name
, void *data
)
1207 struct block_device
*bdev
= NULL
;
1208 struct super_block
*s
;
1209 struct dentry
*root
;
1210 struct btrfs_fs_devices
*fs_devices
= NULL
;
1211 struct btrfs_fs_info
*fs_info
= NULL
;
1212 fmode_t mode
= FMODE_READ
;
1213 char *subvol_name
= NULL
;
1214 u64 subvol_objectid
= 0;
1217 if (!(flags
& MS_RDONLY
))
1218 mode
|= FMODE_WRITE
;
1220 error
= btrfs_parse_early_options(data
, mode
, fs_type
,
1221 &subvol_name
, &subvol_objectid
,
1225 return ERR_PTR(error
);
1229 root
= mount_subvol(subvol_name
, flags
, device_name
, data
);
1234 error
= btrfs_scan_one_device(device_name
, mode
, fs_type
, &fs_devices
);
1236 return ERR_PTR(error
);
1239 * Setup a dummy root and fs_info for test/set super. This is because
1240 * we don't actually fill this stuff out until open_ctree, but we need
1241 * it for searching for existing supers, so this lets us do that and
1242 * then open_ctree will properly initialize everything later.
1244 fs_info
= kzalloc(sizeof(struct btrfs_fs_info
), GFP_NOFS
);
1246 return ERR_PTR(-ENOMEM
);
1248 fs_info
->fs_devices
= fs_devices
;
1250 fs_info
->super_copy
= kzalloc(BTRFS_SUPER_INFO_SIZE
, GFP_NOFS
);
1251 fs_info
->super_for_commit
= kzalloc(BTRFS_SUPER_INFO_SIZE
, GFP_NOFS
);
1252 if (!fs_info
->super_copy
|| !fs_info
->super_for_commit
) {
1257 error
= btrfs_open_devices(fs_devices
, mode
, fs_type
);
1261 if (!(flags
& MS_RDONLY
) && fs_devices
->rw_devices
== 0) {
1263 goto error_close_devices
;
1266 bdev
= fs_devices
->latest_bdev
;
1267 s
= sget(fs_type
, btrfs_test_super
, btrfs_set_super
, flags
| MS_NOSEC
,
1271 goto error_close_devices
;
1275 btrfs_close_devices(fs_devices
);
1276 free_fs_info(fs_info
);
1277 if ((flags
^ s
->s_flags
) & MS_RDONLY
)
1280 char b
[BDEVNAME_SIZE
];
1282 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
1283 btrfs_sb(s
)->bdev_holder
= fs_type
;
1284 error
= btrfs_fill_super(s
, fs_devices
, data
,
1285 flags
& MS_SILENT
? 1 : 0);
1288 root
= !error
? get_default_root(s
, subvol_objectid
) : ERR_PTR(error
);
1290 deactivate_locked_super(s
);
1294 error_close_devices
:
1295 btrfs_close_devices(fs_devices
);
1297 free_fs_info(fs_info
);
1298 return ERR_PTR(error
);
1301 static void btrfs_set_max_workers(struct btrfs_workers
*workers
, int new_limit
)
1303 spin_lock_irq(&workers
->lock
);
1304 workers
->max_workers
= new_limit
;
1305 spin_unlock_irq(&workers
->lock
);
1308 static void btrfs_resize_thread_pool(struct btrfs_fs_info
*fs_info
,
1309 int new_pool_size
, int old_pool_size
)
1311 if (new_pool_size
== old_pool_size
)
1314 fs_info
->thread_pool_size
= new_pool_size
;
1316 btrfs_info(fs_info
, "resize thread pool %d -> %d",
1317 old_pool_size
, new_pool_size
);
1319 btrfs_set_max_workers(&fs_info
->generic_worker
, new_pool_size
);
1320 btrfs_set_max_workers(&fs_info
->workers
, new_pool_size
);
1321 btrfs_set_max_workers(&fs_info
->delalloc_workers
, new_pool_size
);
1322 btrfs_set_max_workers(&fs_info
->submit_workers
, new_pool_size
);
1323 btrfs_set_max_workers(&fs_info
->caching_workers
, new_pool_size
);
1324 btrfs_set_max_workers(&fs_info
->fixup_workers
, new_pool_size
);
1325 btrfs_set_max_workers(&fs_info
->endio_workers
, new_pool_size
);
1326 btrfs_set_max_workers(&fs_info
->endio_meta_workers
, new_pool_size
);
1327 btrfs_set_max_workers(&fs_info
->endio_meta_write_workers
, new_pool_size
);
1328 btrfs_set_max_workers(&fs_info
->endio_write_workers
, new_pool_size
);
1329 btrfs_set_max_workers(&fs_info
->endio_freespace_worker
, new_pool_size
);
1330 btrfs_set_max_workers(&fs_info
->delayed_workers
, new_pool_size
);
1331 btrfs_set_max_workers(&fs_info
->readahead_workers
, new_pool_size
);
1332 btrfs_set_max_workers(&fs_info
->scrub_wr_completion_workers
,
1336 static inline void btrfs_remount_prepare(struct btrfs_fs_info
*fs_info
)
1338 set_bit(BTRFS_FS_STATE_REMOUNTING
, &fs_info
->fs_state
);
1341 static inline void btrfs_remount_begin(struct btrfs_fs_info
*fs_info
,
1342 unsigned long old_opts
, int flags
)
1344 if (btrfs_raw_test_opt(old_opts
, AUTO_DEFRAG
) &&
1345 (!btrfs_raw_test_opt(fs_info
->mount_opt
, AUTO_DEFRAG
) ||
1346 (flags
& MS_RDONLY
))) {
1347 /* wait for any defraggers to finish */
1348 wait_event(fs_info
->transaction_wait
,
1349 (atomic_read(&fs_info
->defrag_running
) == 0));
1350 if (flags
& MS_RDONLY
)
1351 sync_filesystem(fs_info
->sb
);
1355 static inline void btrfs_remount_cleanup(struct btrfs_fs_info
*fs_info
,
1356 unsigned long old_opts
)
1359 * We need cleanup all defragable inodes if the autodefragment is
1360 * close or the fs is R/O.
1362 if (btrfs_raw_test_opt(old_opts
, AUTO_DEFRAG
) &&
1363 (!btrfs_raw_test_opt(fs_info
->mount_opt
, AUTO_DEFRAG
) ||
1364 (fs_info
->sb
->s_flags
& MS_RDONLY
))) {
1365 btrfs_cleanup_defrag_inodes(fs_info
);
1368 clear_bit(BTRFS_FS_STATE_REMOUNTING
, &fs_info
->fs_state
);
1371 static int btrfs_remount(struct super_block
*sb
, int *flags
, char *data
)
1373 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1374 struct btrfs_root
*root
= fs_info
->tree_root
;
1375 unsigned old_flags
= sb
->s_flags
;
1376 unsigned long old_opts
= fs_info
->mount_opt
;
1377 unsigned long old_compress_type
= fs_info
->compress_type
;
1378 u64 old_max_inline
= fs_info
->max_inline
;
1379 u64 old_alloc_start
= fs_info
->alloc_start
;
1380 int old_thread_pool_size
= fs_info
->thread_pool_size
;
1381 unsigned int old_metadata_ratio
= fs_info
->metadata_ratio
;
1384 btrfs_remount_prepare(fs_info
);
1386 ret
= btrfs_parse_options(root
, data
);
1392 btrfs_remount_begin(fs_info
, old_opts
, *flags
);
1393 btrfs_resize_thread_pool(fs_info
,
1394 fs_info
->thread_pool_size
, old_thread_pool_size
);
1396 if ((*flags
& MS_RDONLY
) == (sb
->s_flags
& MS_RDONLY
))
1399 if (*flags
& MS_RDONLY
) {
1401 * this also happens on 'umount -rf' or on shutdown, when
1402 * the filesystem is busy.
1405 /* wait for the uuid_scan task to finish */
1406 down(&fs_info
->uuid_tree_rescan_sem
);
1407 /* avoid complains from lockdep et al. */
1408 up(&fs_info
->uuid_tree_rescan_sem
);
1410 sb
->s_flags
|= MS_RDONLY
;
1412 btrfs_dev_replace_suspend_for_unmount(fs_info
);
1413 btrfs_scrub_cancel(fs_info
);
1414 btrfs_pause_balance(fs_info
);
1416 ret
= btrfs_commit_super(root
);
1420 if (test_bit(BTRFS_FS_STATE_ERROR
, &root
->fs_info
->fs_state
)) {
1422 "Remounting read-write after error is not allowed");
1426 if (fs_info
->fs_devices
->rw_devices
== 0) {
1431 if (fs_info
->fs_devices
->missing_devices
>
1432 fs_info
->num_tolerated_disk_barrier_failures
&&
1433 !(*flags
& MS_RDONLY
)) {
1435 "too many missing devices, writeable remount is not allowed");
1440 if (btrfs_super_log_root(fs_info
->super_copy
) != 0) {
1445 ret
= btrfs_cleanup_fs_roots(fs_info
);
1449 /* recover relocation */
1450 ret
= btrfs_recover_relocation(root
);
1454 ret
= btrfs_resume_balance_async(fs_info
);
1458 ret
= btrfs_resume_dev_replace_async(fs_info
);
1460 btrfs_warn(fs_info
, "failed to resume dev_replace");
1464 if (!fs_info
->uuid_root
) {
1465 btrfs_info(fs_info
, "creating UUID tree");
1466 ret
= btrfs_create_uuid_tree(fs_info
);
1468 btrfs_warn(fs_info
, "failed to create the UUID tree %d", ret
);
1472 sb
->s_flags
&= ~MS_RDONLY
;
1475 btrfs_remount_cleanup(fs_info
, old_opts
);
1479 /* We've hit an error - don't reset MS_RDONLY */
1480 if (sb
->s_flags
& MS_RDONLY
)
1481 old_flags
|= MS_RDONLY
;
1482 sb
->s_flags
= old_flags
;
1483 fs_info
->mount_opt
= old_opts
;
1484 fs_info
->compress_type
= old_compress_type
;
1485 fs_info
->max_inline
= old_max_inline
;
1486 mutex_lock(&fs_info
->chunk_mutex
);
1487 fs_info
->alloc_start
= old_alloc_start
;
1488 mutex_unlock(&fs_info
->chunk_mutex
);
1489 btrfs_resize_thread_pool(fs_info
,
1490 old_thread_pool_size
, fs_info
->thread_pool_size
);
1491 fs_info
->metadata_ratio
= old_metadata_ratio
;
1492 btrfs_remount_cleanup(fs_info
, old_opts
);
1496 /* Used to sort the devices by max_avail(descending sort) */
1497 static int btrfs_cmp_device_free_bytes(const void *dev_info1
,
1498 const void *dev_info2
)
1500 if (((struct btrfs_device_info
*)dev_info1
)->max_avail
>
1501 ((struct btrfs_device_info
*)dev_info2
)->max_avail
)
1503 else if (((struct btrfs_device_info
*)dev_info1
)->max_avail
<
1504 ((struct btrfs_device_info
*)dev_info2
)->max_avail
)
1511 * sort the devices by max_avail, in which max free extent size of each device
1512 * is stored.(Descending Sort)
1514 static inline void btrfs_descending_sort_devices(
1515 struct btrfs_device_info
*devices
,
1518 sort(devices
, nr_devices
, sizeof(struct btrfs_device_info
),
1519 btrfs_cmp_device_free_bytes
, NULL
);
1523 * The helper to calc the free space on the devices that can be used to store
1526 static int btrfs_calc_avail_data_space(struct btrfs_root
*root
, u64
*free_bytes
)
1528 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1529 struct btrfs_device_info
*devices_info
;
1530 struct btrfs_fs_devices
*fs_devices
= fs_info
->fs_devices
;
1531 struct btrfs_device
*device
;
1536 u64 min_stripe_size
;
1537 int min_stripes
= 1, num_stripes
= 1;
1538 int i
= 0, nr_devices
;
1541 nr_devices
= fs_info
->fs_devices
->open_devices
;
1542 BUG_ON(!nr_devices
);
1544 devices_info
= kmalloc_array(nr_devices
, sizeof(*devices_info
),
1549 /* calc min stripe number for data space alloction */
1550 type
= btrfs_get_alloc_profile(root
, 1);
1551 if (type
& BTRFS_BLOCK_GROUP_RAID0
) {
1553 num_stripes
= nr_devices
;
1554 } else if (type
& BTRFS_BLOCK_GROUP_RAID1
) {
1557 } else if (type
& BTRFS_BLOCK_GROUP_RAID10
) {
1562 if (type
& BTRFS_BLOCK_GROUP_DUP
)
1563 min_stripe_size
= 2 * BTRFS_STRIPE_LEN
;
1565 min_stripe_size
= BTRFS_STRIPE_LEN
;
1567 list_for_each_entry(device
, &fs_devices
->devices
, dev_list
) {
1568 if (!device
->in_fs_metadata
|| !device
->bdev
||
1569 device
->is_tgtdev_for_dev_replace
)
1572 avail_space
= device
->total_bytes
- device
->bytes_used
;
1574 /* align with stripe_len */
1575 do_div(avail_space
, BTRFS_STRIPE_LEN
);
1576 avail_space
*= BTRFS_STRIPE_LEN
;
1579 * In order to avoid overwritting the superblock on the drive,
1580 * btrfs starts at an offset of at least 1MB when doing chunk
1583 skip_space
= 1024 * 1024;
1585 /* user can set the offset in fs_info->alloc_start. */
1586 if (fs_info
->alloc_start
+ BTRFS_STRIPE_LEN
<=
1587 device
->total_bytes
)
1588 skip_space
= max(fs_info
->alloc_start
, skip_space
);
1591 * btrfs can not use the free space in [0, skip_space - 1],
1592 * we must subtract it from the total. In order to implement
1593 * it, we account the used space in this range first.
1595 ret
= btrfs_account_dev_extents_size(device
, 0, skip_space
- 1,
1598 kfree(devices_info
);
1602 /* calc the free space in [0, skip_space - 1] */
1603 skip_space
-= used_space
;
1606 * we can use the free space in [0, skip_space - 1], subtract
1607 * it from the total.
1609 if (avail_space
&& avail_space
>= skip_space
)
1610 avail_space
-= skip_space
;
1614 if (avail_space
< min_stripe_size
)
1617 devices_info
[i
].dev
= device
;
1618 devices_info
[i
].max_avail
= avail_space
;
1625 btrfs_descending_sort_devices(devices_info
, nr_devices
);
1629 while (nr_devices
>= min_stripes
) {
1630 if (num_stripes
> nr_devices
)
1631 num_stripes
= nr_devices
;
1633 if (devices_info
[i
].max_avail
>= min_stripe_size
) {
1637 avail_space
+= devices_info
[i
].max_avail
* num_stripes
;
1638 alloc_size
= devices_info
[i
].max_avail
;
1639 for (j
= i
+ 1 - num_stripes
; j
<= i
; j
++)
1640 devices_info
[j
].max_avail
-= alloc_size
;
1646 kfree(devices_info
);
1647 *free_bytes
= avail_space
;
1651 static int btrfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1653 struct btrfs_fs_info
*fs_info
= btrfs_sb(dentry
->d_sb
);
1654 struct btrfs_super_block
*disk_super
= fs_info
->super_copy
;
1655 struct list_head
*head
= &fs_info
->space_info
;
1656 struct btrfs_space_info
*found
;
1658 u64 total_free_data
= 0;
1659 int bits
= dentry
->d_sb
->s_blocksize_bits
;
1660 __be32
*fsid
= (__be32
*)fs_info
->fsid
;
1663 /* holding chunk_muext to avoid allocating new chunks */
1664 mutex_lock(&fs_info
->chunk_mutex
);
1666 list_for_each_entry_rcu(found
, head
, list
) {
1667 if (found
->flags
& BTRFS_BLOCK_GROUP_DATA
) {
1668 total_free_data
+= found
->disk_total
- found
->disk_used
;
1670 btrfs_account_ro_block_groups_free_space(found
);
1673 total_used
+= found
->disk_used
;
1677 buf
->f_namelen
= BTRFS_NAME_LEN
;
1678 buf
->f_blocks
= btrfs_super_total_bytes(disk_super
) >> bits
;
1679 buf
->f_bfree
= buf
->f_blocks
- (total_used
>> bits
);
1680 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1681 buf
->f_type
= BTRFS_SUPER_MAGIC
;
1682 buf
->f_bavail
= total_free_data
;
1683 ret
= btrfs_calc_avail_data_space(fs_info
->tree_root
, &total_free_data
);
1685 mutex_unlock(&fs_info
->chunk_mutex
);
1688 buf
->f_bavail
+= total_free_data
;
1689 buf
->f_bavail
= buf
->f_bavail
>> bits
;
1690 mutex_unlock(&fs_info
->chunk_mutex
);
1692 /* We treat it as constant endianness (it doesn't matter _which_)
1693 because we want the fsid to come out the same whether mounted
1694 on a big-endian or little-endian host */
1695 buf
->f_fsid
.val
[0] = be32_to_cpu(fsid
[0]) ^ be32_to_cpu(fsid
[2]);
1696 buf
->f_fsid
.val
[1] = be32_to_cpu(fsid
[1]) ^ be32_to_cpu(fsid
[3]);
1697 /* Mask in the root object ID too, to disambiguate subvols */
1698 buf
->f_fsid
.val
[0] ^= BTRFS_I(dentry
->d_inode
)->root
->objectid
>> 32;
1699 buf
->f_fsid
.val
[1] ^= BTRFS_I(dentry
->d_inode
)->root
->objectid
;
1704 static void btrfs_kill_super(struct super_block
*sb
)
1706 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1707 kill_anon_super(sb
);
1708 free_fs_info(fs_info
);
1711 static struct file_system_type btrfs_fs_type
= {
1712 .owner
= THIS_MODULE
,
1714 .mount
= btrfs_mount
,
1715 .kill_sb
= btrfs_kill_super
,
1716 .fs_flags
= FS_REQUIRES_DEV
,
1718 MODULE_ALIAS_FS("btrfs");
1721 * used by btrfsctl to scan devices when no FS is mounted
1723 static long btrfs_control_ioctl(struct file
*file
, unsigned int cmd
,
1726 struct btrfs_ioctl_vol_args
*vol
;
1727 struct btrfs_fs_devices
*fs_devices
;
1730 if (!capable(CAP_SYS_ADMIN
))
1733 vol
= memdup_user((void __user
*)arg
, sizeof(*vol
));
1735 return PTR_ERR(vol
);
1738 case BTRFS_IOC_SCAN_DEV
:
1739 ret
= btrfs_scan_one_device(vol
->name
, FMODE_READ
,
1740 &btrfs_fs_type
, &fs_devices
);
1742 case BTRFS_IOC_DEVICES_READY
:
1743 ret
= btrfs_scan_one_device(vol
->name
, FMODE_READ
,
1744 &btrfs_fs_type
, &fs_devices
);
1747 ret
= !(fs_devices
->num_devices
== fs_devices
->total_devices
);
1755 static int btrfs_freeze(struct super_block
*sb
)
1757 struct btrfs_trans_handle
*trans
;
1758 struct btrfs_root
*root
= btrfs_sb(sb
)->tree_root
;
1760 trans
= btrfs_attach_transaction_barrier(root
);
1761 if (IS_ERR(trans
)) {
1762 /* no transaction, don't bother */
1763 if (PTR_ERR(trans
) == -ENOENT
)
1765 return PTR_ERR(trans
);
1767 return btrfs_commit_transaction(trans
, root
);
1770 static int btrfs_unfreeze(struct super_block
*sb
)
1775 static int btrfs_show_devname(struct seq_file
*m
, struct dentry
*root
)
1777 struct btrfs_fs_info
*fs_info
= btrfs_sb(root
->d_sb
);
1778 struct btrfs_fs_devices
*cur_devices
;
1779 struct btrfs_device
*dev
, *first_dev
= NULL
;
1780 struct list_head
*head
;
1781 struct rcu_string
*name
;
1783 mutex_lock(&fs_info
->fs_devices
->device_list_mutex
);
1784 cur_devices
= fs_info
->fs_devices
;
1785 while (cur_devices
) {
1786 head
= &cur_devices
->devices
;
1787 list_for_each_entry(dev
, head
, dev_list
) {
1790 if (!first_dev
|| dev
->devid
< first_dev
->devid
)
1793 cur_devices
= cur_devices
->seed
;
1798 name
= rcu_dereference(first_dev
->name
);
1799 seq_escape(m
, name
->str
, " \t\n\\");
1804 mutex_unlock(&fs_info
->fs_devices
->device_list_mutex
);
1808 static const struct super_operations btrfs_super_ops
= {
1809 .drop_inode
= btrfs_drop_inode
,
1810 .evict_inode
= btrfs_evict_inode
,
1811 .put_super
= btrfs_put_super
,
1812 .sync_fs
= btrfs_sync_fs
,
1813 .show_options
= btrfs_show_options
,
1814 .show_devname
= btrfs_show_devname
,
1815 .write_inode
= btrfs_write_inode
,
1816 .alloc_inode
= btrfs_alloc_inode
,
1817 .destroy_inode
= btrfs_destroy_inode
,
1818 .statfs
= btrfs_statfs
,
1819 .remount_fs
= btrfs_remount
,
1820 .freeze_fs
= btrfs_freeze
,
1821 .unfreeze_fs
= btrfs_unfreeze
,
1824 static const struct file_operations btrfs_ctl_fops
= {
1825 .unlocked_ioctl
= btrfs_control_ioctl
,
1826 .compat_ioctl
= btrfs_control_ioctl
,
1827 .owner
= THIS_MODULE
,
1828 .llseek
= noop_llseek
,
1831 static struct miscdevice btrfs_misc
= {
1832 .minor
= BTRFS_MINOR
,
1833 .name
= "btrfs-control",
1834 .fops
= &btrfs_ctl_fops
1837 MODULE_ALIAS_MISCDEV(BTRFS_MINOR
);
1838 MODULE_ALIAS("devname:btrfs-control");
1840 static int btrfs_interface_init(void)
1842 return misc_register(&btrfs_misc
);
1845 static void btrfs_interface_exit(void)
1847 if (misc_deregister(&btrfs_misc
) < 0)
1848 printk(KERN_INFO
"BTRFS: misc_deregister failed for control device\n");
1851 static void btrfs_print_info(void)
1853 printk(KERN_INFO
"Btrfs loaded"
1854 #ifdef CONFIG_BTRFS_DEBUG
1857 #ifdef CONFIG_BTRFS_ASSERT
1860 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1861 ", integrity-checker=on"
1866 static int btrfs_run_sanity_tests(void)
1870 ret
= btrfs_init_test_fs();
1874 ret
= btrfs_test_free_space_cache();
1877 ret
= btrfs_test_extent_buffer_operations();
1880 ret
= btrfs_test_extent_io();
1883 ret
= btrfs_test_inodes();
1885 btrfs_destroy_test_fs();
1889 static int __init
init_btrfs_fs(void)
1893 err
= btrfs_hash_init();
1899 err
= btrfs_init_sysfs();
1903 btrfs_init_compress();
1905 err
= btrfs_init_cachep();
1909 err
= extent_io_init();
1913 err
= extent_map_init();
1915 goto free_extent_io
;
1917 err
= ordered_data_init();
1919 goto free_extent_map
;
1921 err
= btrfs_delayed_inode_init();
1923 goto free_ordered_data
;
1925 err
= btrfs_auto_defrag_init();
1927 goto free_delayed_inode
;
1929 err
= btrfs_delayed_ref_init();
1931 goto free_auto_defrag
;
1933 err
= btrfs_prelim_ref_init();
1935 goto free_prelim_ref
;
1937 err
= btrfs_interface_init();
1939 goto free_delayed_ref
;
1941 btrfs_init_lockdep();
1945 err
= btrfs_run_sanity_tests();
1947 goto unregister_ioctl
;
1949 err
= register_filesystem(&btrfs_fs_type
);
1951 goto unregister_ioctl
;
1956 btrfs_interface_exit();
1958 btrfs_prelim_ref_exit();
1960 btrfs_delayed_ref_exit();
1962 btrfs_auto_defrag_exit();
1964 btrfs_delayed_inode_exit();
1966 ordered_data_exit();
1972 btrfs_destroy_cachep();
1974 btrfs_exit_compress();
1981 static void __exit
exit_btrfs_fs(void)
1983 btrfs_destroy_cachep();
1984 btrfs_delayed_ref_exit();
1985 btrfs_auto_defrag_exit();
1986 btrfs_delayed_inode_exit();
1987 btrfs_prelim_ref_exit();
1988 ordered_data_exit();
1991 btrfs_interface_exit();
1992 unregister_filesystem(&btrfs_fs_type
);
1994 btrfs_cleanup_fs_uuids();
1995 btrfs_exit_compress();
1999 late_initcall(init_btrfs_fs
);
2000 module_exit(exit_btrfs_fs
)
2002 MODULE_LICENSE("GPL");