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 int btrfs_remount(struct super_block
*sb
, int *flags
, char *data
);
71 static const char *btrfs_decode_error(int errno
)
73 char *errstr
= "unknown";
77 errstr
= "IO failure";
80 errstr
= "Out of memory";
83 errstr
= "Readonly filesystem";
86 errstr
= "Object already exists";
89 errstr
= "No space left";
92 errstr
= "No such entry";
99 static void save_error_info(struct btrfs_fs_info
*fs_info
)
102 * today we only save the error info into ram. Long term we'll
103 * also send it down to the disk
105 set_bit(BTRFS_FS_STATE_ERROR
, &fs_info
->fs_state
);
108 /* btrfs handle error by forcing the filesystem readonly */
109 static void btrfs_handle_error(struct btrfs_fs_info
*fs_info
)
111 struct super_block
*sb
= fs_info
->sb
;
113 if (sb
->s_flags
& MS_RDONLY
)
116 if (test_bit(BTRFS_FS_STATE_ERROR
, &fs_info
->fs_state
)) {
117 sb
->s_flags
|= MS_RDONLY
;
118 btrfs_info(fs_info
, "forced readonly");
120 * Note that a running device replace operation is not
121 * canceled here although there is no way to update
122 * the progress. It would add the risk of a deadlock,
123 * therefore the canceling is ommited. The only penalty
124 * is that some I/O remains active until the procedure
125 * completes. The next time when the filesystem is
126 * mounted writeable again, the device replace
127 * operation continues.
134 * __btrfs_std_error decodes expected errors from the caller and
135 * invokes the approciate error response.
137 void __btrfs_std_error(struct btrfs_fs_info
*fs_info
, const char *function
,
138 unsigned int line
, int errno
, const char *fmt
, ...)
140 struct super_block
*sb
= fs_info
->sb
;
144 * Special case: if the error is EROFS, and we're already
145 * under MS_RDONLY, then it is safe here.
147 if (errno
== -EROFS
&& (sb
->s_flags
& MS_RDONLY
))
150 errstr
= btrfs_decode_error(errno
);
152 struct va_format vaf
;
160 "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
161 sb
->s_id
, function
, line
, errno
, errstr
, &vaf
);
164 printk(KERN_CRIT
"BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
165 sb
->s_id
, function
, line
, errno
, errstr
);
168 /* Don't go through full error handling during mount */
169 save_error_info(fs_info
);
170 if (sb
->s_flags
& MS_BORN
)
171 btrfs_handle_error(fs_info
);
174 static const char * const logtypes
[] = {
185 void btrfs_printk(const struct btrfs_fs_info
*fs_info
, const char *fmt
, ...)
187 struct super_block
*sb
= fs_info
->sb
;
189 struct va_format vaf
;
191 const char *type
= logtypes
[4];
196 kern_level
= printk_get_level(fmt
);
198 size_t size
= printk_skip_level(fmt
) - fmt
;
199 memcpy(lvl
, fmt
, size
);
202 type
= logtypes
[kern_level
- '0'];
209 printk("%sBTRFS %s (device %s): %pV\n", lvl
, type
, sb
->s_id
, &vaf
);
216 void __btrfs_std_error(struct btrfs_fs_info
*fs_info
, const char *function
,
217 unsigned int line
, int errno
, const char *fmt
, ...)
219 struct super_block
*sb
= fs_info
->sb
;
222 * Special case: if the error is EROFS, and we're already
223 * under MS_RDONLY, then it is safe here.
225 if (errno
== -EROFS
&& (sb
->s_flags
& MS_RDONLY
))
228 /* Don't go through full error handling during mount */
229 if (sb
->s_flags
& MS_BORN
) {
230 save_error_info(fs_info
);
231 btrfs_handle_error(fs_info
);
237 * We only mark the transaction aborted and then set the file system read-only.
238 * This will prevent new transactions from starting or trying to join this
241 * This means that error recovery at the call site is limited to freeing
242 * any local memory allocations and passing the error code up without
243 * further cleanup. The transaction should complete as it normally would
244 * in the call path but will return -EIO.
246 * We'll complete the cleanup in btrfs_end_transaction and
247 * btrfs_commit_transaction.
249 void __btrfs_abort_transaction(struct btrfs_trans_handle
*trans
,
250 struct btrfs_root
*root
, const char *function
,
251 unsigned int line
, int errno
)
254 * Report first abort since mount
256 if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED
,
257 &root
->fs_info
->fs_state
)) {
258 WARN(1, KERN_DEBUG
"BTRFS: Transaction aborted (error %d)\n",
261 trans
->aborted
= errno
;
262 /* Nothing used. The other threads that have joined this
263 * transaction may be able to continue. */
264 if (!trans
->blocks_used
) {
267 errstr
= btrfs_decode_error(errno
);
268 btrfs_warn(root
->fs_info
,
269 "%s:%d: Aborting unused transaction(%s).",
270 function
, line
, errstr
);
273 ACCESS_ONCE(trans
->transaction
->aborted
) = errno
;
274 /* Wake up anybody who may be waiting on this transaction */
275 wake_up(&root
->fs_info
->transaction_wait
);
276 wake_up(&root
->fs_info
->transaction_blocked_wait
);
277 __btrfs_std_error(root
->fs_info
, function
, line
, errno
, NULL
);
280 * __btrfs_panic decodes unexpected, fatal errors from the caller,
281 * issues an alert, and either panics or BUGs, depending on mount options.
283 void __btrfs_panic(struct btrfs_fs_info
*fs_info
, const char *function
,
284 unsigned int line
, int errno
, const char *fmt
, ...)
286 char *s_id
= "<unknown>";
288 struct va_format vaf
= { .fmt
= fmt
};
292 s_id
= fs_info
->sb
->s_id
;
297 errstr
= btrfs_decode_error(errno
);
298 if (fs_info
&& (fs_info
->mount_opt
& BTRFS_MOUNT_PANIC_ON_FATAL_ERROR
))
299 panic(KERN_CRIT
"BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
300 s_id
, function
, line
, &vaf
, errno
, errstr
);
302 btrfs_crit(fs_info
, "panic in %s:%d: %pV (errno=%d %s)",
303 function
, line
, &vaf
, errno
, errstr
);
305 /* Caller calls BUG() */
308 static void btrfs_put_super(struct super_block
*sb
)
310 (void)close_ctree(btrfs_sb(sb
)->tree_root
);
311 /* FIXME: need to fix VFS to return error? */
312 /* AV: return it _where_? ->put_super() can be triggered by any number
313 * of async events, up to and including delivery of SIGKILL to the
314 * last process that kept it busy. Or segfault in the aforementioned
315 * process... Whom would you report that to?
320 Opt_degraded
, Opt_subvol
, Opt_subvolid
, Opt_device
, Opt_nodatasum
,
321 Opt_nodatacow
, Opt_max_inline
, Opt_alloc_start
, Opt_nobarrier
, Opt_ssd
,
322 Opt_nossd
, Opt_ssd_spread
, Opt_thread_pool
, Opt_noacl
, Opt_compress
,
323 Opt_compress_type
, Opt_compress_force
, Opt_compress_force_type
,
324 Opt_notreelog
, Opt_ratio
, Opt_flushoncommit
, Opt_discard
,
325 Opt_space_cache
, Opt_clear_cache
, Opt_user_subvol_rm_allowed
,
326 Opt_enospc_debug
, Opt_subvolrootid
, Opt_defrag
, Opt_inode_cache
,
327 Opt_no_space_cache
, Opt_recovery
, Opt_skip_balance
,
328 Opt_check_integrity
, Opt_check_integrity_including_extent_data
,
329 Opt_check_integrity_print_mask
, Opt_fatal_errors
, Opt_rescan_uuid_tree
,
330 Opt_commit_interval
, Opt_barrier
, Opt_nodefrag
, Opt_nodiscard
,
331 Opt_noenospc_debug
, Opt_noflushoncommit
, Opt_acl
, Opt_datacow
,
332 Opt_datasum
, Opt_treelog
, Opt_noinode_cache
,
336 static match_table_t tokens
= {
337 {Opt_degraded
, "degraded"},
338 {Opt_subvol
, "subvol=%s"},
339 {Opt_subvolid
, "subvolid=%s"},
340 {Opt_device
, "device=%s"},
341 {Opt_nodatasum
, "nodatasum"},
342 {Opt_datasum
, "datasum"},
343 {Opt_nodatacow
, "nodatacow"},
344 {Opt_datacow
, "datacow"},
345 {Opt_nobarrier
, "nobarrier"},
346 {Opt_barrier
, "barrier"},
347 {Opt_max_inline
, "max_inline=%s"},
348 {Opt_alloc_start
, "alloc_start=%s"},
349 {Opt_thread_pool
, "thread_pool=%d"},
350 {Opt_compress
, "compress"},
351 {Opt_compress_type
, "compress=%s"},
352 {Opt_compress_force
, "compress-force"},
353 {Opt_compress_force_type
, "compress-force=%s"},
355 {Opt_ssd_spread
, "ssd_spread"},
356 {Opt_nossd
, "nossd"},
358 {Opt_noacl
, "noacl"},
359 {Opt_notreelog
, "notreelog"},
360 {Opt_treelog
, "treelog"},
361 {Opt_flushoncommit
, "flushoncommit"},
362 {Opt_noflushoncommit
, "noflushoncommit"},
363 {Opt_ratio
, "metadata_ratio=%d"},
364 {Opt_discard
, "discard"},
365 {Opt_nodiscard
, "nodiscard"},
366 {Opt_space_cache
, "space_cache"},
367 {Opt_clear_cache
, "clear_cache"},
368 {Opt_user_subvol_rm_allowed
, "user_subvol_rm_allowed"},
369 {Opt_enospc_debug
, "enospc_debug"},
370 {Opt_noenospc_debug
, "noenospc_debug"},
371 {Opt_subvolrootid
, "subvolrootid=%d"},
372 {Opt_defrag
, "autodefrag"},
373 {Opt_nodefrag
, "noautodefrag"},
374 {Opt_inode_cache
, "inode_cache"},
375 {Opt_noinode_cache
, "noinode_cache"},
376 {Opt_no_space_cache
, "nospace_cache"},
377 {Opt_recovery
, "recovery"},
378 {Opt_skip_balance
, "skip_balance"},
379 {Opt_check_integrity
, "check_int"},
380 {Opt_check_integrity_including_extent_data
, "check_int_data"},
381 {Opt_check_integrity_print_mask
, "check_int_print_mask=%d"},
382 {Opt_rescan_uuid_tree
, "rescan_uuid_tree"},
383 {Opt_fatal_errors
, "fatal_errors=%s"},
384 {Opt_commit_interval
, "commit=%d"},
389 * Regular mount options parser. Everything that is needed only when
390 * reading in a new superblock is parsed here.
391 * XXX JDM: This needs to be cleaned up for remount.
393 int btrfs_parse_options(struct btrfs_root
*root
, char *options
)
395 struct btrfs_fs_info
*info
= root
->fs_info
;
396 substring_t args
[MAX_OPT_ARGS
];
397 char *p
, *num
, *orig
= NULL
;
402 bool compress_force
= false;
403 bool compress
= false;
405 cache_gen
= btrfs_super_cache_generation(root
->fs_info
->super_copy
);
407 btrfs_set_opt(info
->mount_opt
, SPACE_CACHE
);
413 * strsep changes the string, duplicate it because parse_options
416 options
= kstrdup(options
, GFP_NOFS
);
422 while ((p
= strsep(&options
, ",")) != NULL
) {
427 token
= match_token(p
, tokens
, args
);
430 btrfs_info(root
->fs_info
, "allowing degraded mounts");
431 btrfs_set_opt(info
->mount_opt
, DEGRADED
);
435 case Opt_subvolrootid
:
438 * These are parsed by btrfs_parse_early_options
439 * and can be happily ignored here.
443 btrfs_set_and_info(root
, NODATASUM
,
444 "setting nodatasum");
447 if (btrfs_test_opt(root
, NODATASUM
)) {
448 if (btrfs_test_opt(root
, NODATACOW
))
449 btrfs_info(root
->fs_info
, "setting datasum, datacow enabled");
451 btrfs_info(root
->fs_info
, "setting datasum");
453 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
454 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
457 if (!btrfs_test_opt(root
, NODATACOW
)) {
458 if (!btrfs_test_opt(root
, COMPRESS
) ||
459 !btrfs_test_opt(root
, FORCE_COMPRESS
)) {
460 btrfs_info(root
->fs_info
,
461 "setting nodatacow, compression disabled");
463 btrfs_info(root
->fs_info
, "setting nodatacow");
466 btrfs_clear_opt(info
->mount_opt
, COMPRESS
);
467 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
468 btrfs_set_opt(info
->mount_opt
, NODATACOW
);
469 btrfs_set_opt(info
->mount_opt
, NODATASUM
);
472 btrfs_clear_and_info(root
, NODATACOW
,
475 case Opt_compress_force
:
476 case Opt_compress_force_type
:
477 compress_force
= true;
480 case Opt_compress_type
:
482 if (token
== Opt_compress
||
483 token
== Opt_compress_force
||
484 strcmp(args
[0].from
, "zlib") == 0) {
485 compress_type
= "zlib";
486 info
->compress_type
= BTRFS_COMPRESS_ZLIB
;
487 btrfs_set_opt(info
->mount_opt
, COMPRESS
);
488 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
489 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
490 } else if (strcmp(args
[0].from
, "lzo") == 0) {
491 compress_type
= "lzo";
492 info
->compress_type
= BTRFS_COMPRESS_LZO
;
493 btrfs_set_opt(info
->mount_opt
, COMPRESS
);
494 btrfs_clear_opt(info
->mount_opt
, NODATACOW
);
495 btrfs_clear_opt(info
->mount_opt
, NODATASUM
);
496 btrfs_set_fs_incompat(info
, COMPRESS_LZO
);
497 } else if (strncmp(args
[0].from
, "no", 2) == 0) {
498 compress_type
= "no";
499 btrfs_clear_opt(info
->mount_opt
, COMPRESS
);
500 btrfs_clear_opt(info
->mount_opt
, FORCE_COMPRESS
);
501 compress_force
= false;
507 if (compress_force
) {
508 btrfs_set_and_info(root
, FORCE_COMPRESS
,
509 "force %s compression",
511 } else if (compress
) {
512 if (!btrfs_test_opt(root
, COMPRESS
))
513 btrfs_info(root
->fs_info
,
514 "btrfs: use %s compression\n",
519 btrfs_set_and_info(root
, SSD
,
520 "use ssd allocation scheme");
523 btrfs_set_and_info(root
, SSD_SPREAD
,
524 "use spread ssd allocation scheme");
527 btrfs_clear_and_info(root
, NOSSD
,
528 "not using ssd allocation scheme");
529 btrfs_clear_opt(info
->mount_opt
, SSD
);
532 btrfs_clear_and_info(root
, NOBARRIER
,
533 "turning on barriers");
536 btrfs_set_and_info(root
, NOBARRIER
,
537 "turning off barriers");
539 case Opt_thread_pool
:
540 ret
= match_int(&args
[0], &intarg
);
543 } else if (intarg
> 0) {
544 info
->thread_pool_size
= intarg
;
551 num
= match_strdup(&args
[0]);
553 info
->max_inline
= memparse(num
, NULL
);
556 if (info
->max_inline
) {
557 info
->max_inline
= min_t(u64
,
561 btrfs_info(root
->fs_info
, "max_inline at %llu",
568 case Opt_alloc_start
:
569 num
= match_strdup(&args
[0]);
571 mutex_lock(&info
->chunk_mutex
);
572 info
->alloc_start
= memparse(num
, NULL
);
573 mutex_unlock(&info
->chunk_mutex
);
575 btrfs_info(root
->fs_info
, "allocations start at %llu",
583 root
->fs_info
->sb
->s_flags
|= MS_POSIXACL
;
586 root
->fs_info
->sb
->s_flags
&= ~MS_POSIXACL
;
589 btrfs_set_and_info(root
, NOTREELOG
,
590 "disabling tree log");
593 btrfs_clear_and_info(root
, NOTREELOG
,
594 "enabling tree log");
596 case Opt_flushoncommit
:
597 btrfs_set_and_info(root
, FLUSHONCOMMIT
,
598 "turning on flush-on-commit");
600 case Opt_noflushoncommit
:
601 btrfs_clear_and_info(root
, FLUSHONCOMMIT
,
602 "turning off flush-on-commit");
605 ret
= match_int(&args
[0], &intarg
);
608 } else if (intarg
>= 0) {
609 info
->metadata_ratio
= intarg
;
610 btrfs_info(root
->fs_info
, "metadata ratio %d",
611 info
->metadata_ratio
);
618 btrfs_set_and_info(root
, DISCARD
,
619 "turning on discard");
622 btrfs_clear_and_info(root
, DISCARD
,
623 "turning off discard");
625 case Opt_space_cache
:
626 btrfs_set_and_info(root
, SPACE_CACHE
,
627 "enabling disk space caching");
629 case Opt_rescan_uuid_tree
:
630 btrfs_set_opt(info
->mount_opt
, RESCAN_UUID_TREE
);
632 case Opt_no_space_cache
:
633 btrfs_clear_and_info(root
, SPACE_CACHE
,
634 "disabling disk space caching");
636 case Opt_inode_cache
:
637 btrfs_set_and_info(root
, CHANGE_INODE_CACHE
,
638 "enabling inode map caching");
640 case Opt_noinode_cache
:
641 btrfs_clear_and_info(root
, CHANGE_INODE_CACHE
,
642 "disabling inode map caching");
644 case Opt_clear_cache
:
645 btrfs_set_and_info(root
, CLEAR_CACHE
,
646 "force clearing of disk cache");
648 case Opt_user_subvol_rm_allowed
:
649 btrfs_set_opt(info
->mount_opt
, USER_SUBVOL_RM_ALLOWED
);
651 case Opt_enospc_debug
:
652 btrfs_set_opt(info
->mount_opt
, ENOSPC_DEBUG
);
654 case Opt_noenospc_debug
:
655 btrfs_clear_opt(info
->mount_opt
, ENOSPC_DEBUG
);
658 btrfs_set_and_info(root
, AUTO_DEFRAG
,
659 "enabling auto defrag");
662 btrfs_clear_and_info(root
, AUTO_DEFRAG
,
663 "disabling auto defrag");
666 btrfs_info(root
->fs_info
, "enabling auto recovery");
667 btrfs_set_opt(info
->mount_opt
, RECOVERY
);
669 case Opt_skip_balance
:
670 btrfs_set_opt(info
->mount_opt
, SKIP_BALANCE
);
672 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
673 case Opt_check_integrity_including_extent_data
:
674 btrfs_info(root
->fs_info
,
675 "enabling check integrity including extent data");
676 btrfs_set_opt(info
->mount_opt
,
677 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
);
678 btrfs_set_opt(info
->mount_opt
, CHECK_INTEGRITY
);
680 case Opt_check_integrity
:
681 btrfs_info(root
->fs_info
, "enabling check integrity");
682 btrfs_set_opt(info
->mount_opt
, CHECK_INTEGRITY
);
684 case Opt_check_integrity_print_mask
:
685 ret
= match_int(&args
[0], &intarg
);
688 } else if (intarg
>= 0) {
689 info
->check_integrity_print_mask
= intarg
;
690 btrfs_info(root
->fs_info
, "check_integrity_print_mask 0x%x",
691 info
->check_integrity_print_mask
);
698 case Opt_check_integrity_including_extent_data
:
699 case Opt_check_integrity
:
700 case Opt_check_integrity_print_mask
:
701 btrfs_err(root
->fs_info
,
702 "support for check_integrity* not compiled in!");
706 case Opt_fatal_errors
:
707 if (strcmp(args
[0].from
, "panic") == 0)
708 btrfs_set_opt(info
->mount_opt
,
709 PANIC_ON_FATAL_ERROR
);
710 else if (strcmp(args
[0].from
, "bug") == 0)
711 btrfs_clear_opt(info
->mount_opt
,
712 PANIC_ON_FATAL_ERROR
);
718 case Opt_commit_interval
:
720 ret
= match_int(&args
[0], &intarg
);
722 btrfs_err(root
->fs_info
, "invalid commit interval");
728 btrfs_warn(root
->fs_info
, "excessive commit interval %d",
731 info
->commit_interval
= intarg
;
733 btrfs_info(root
->fs_info
, "using default commit interval %ds",
734 BTRFS_DEFAULT_COMMIT_INTERVAL
);
735 info
->commit_interval
= BTRFS_DEFAULT_COMMIT_INTERVAL
;
739 btrfs_info(root
->fs_info
, "unrecognized mount option '%s'", p
);
747 if (!ret
&& btrfs_test_opt(root
, SPACE_CACHE
))
748 btrfs_info(root
->fs_info
, "disk space caching is enabled");
754 * Parse mount options that are required early in the mount process.
756 * All other options will be parsed on much later in the mount process and
757 * only when we need to allocate a new super block.
759 static int btrfs_parse_early_options(const char *options
, fmode_t flags
,
760 void *holder
, char **subvol_name
, u64
*subvol_objectid
,
761 struct btrfs_fs_devices
**fs_devices
)
763 substring_t args
[MAX_OPT_ARGS
];
764 char *device_name
, *opts
, *orig
, *p
;
772 * strsep changes the string, duplicate it because parse_options
775 opts
= kstrdup(options
, GFP_KERNEL
);
780 while ((p
= strsep(&opts
, ",")) != NULL
) {
785 token
= match_token(p
, tokens
, args
);
789 *subvol_name
= match_strdup(&args
[0]);
796 num
= match_strdup(&args
[0]);
798 *subvol_objectid
= memparse(num
, NULL
);
800 /* we want the original fs_tree */
801 if (!*subvol_objectid
)
803 BTRFS_FS_TREE_OBJECTID
;
809 case Opt_subvolrootid
:
811 "BTRFS: 'subvolrootid' mount option is deprecated and has "
815 device_name
= match_strdup(&args
[0]);
820 error
= btrfs_scan_one_device(device_name
,
821 flags
, holder
, fs_devices
);
836 static struct dentry
*get_default_root(struct super_block
*sb
,
839 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
840 struct btrfs_root
*root
= fs_info
->tree_root
;
841 struct btrfs_root
*new_root
;
842 struct btrfs_dir_item
*di
;
843 struct btrfs_path
*path
;
844 struct btrfs_key location
;
846 struct dentry
*dentry
;
851 * We have a specific subvol we want to mount, just setup location and
852 * go look up the root.
854 if (subvol_objectid
) {
855 location
.objectid
= subvol_objectid
;
856 location
.type
= BTRFS_ROOT_ITEM_KEY
;
857 location
.offset
= (u64
)-1;
861 path
= btrfs_alloc_path();
863 return ERR_PTR(-ENOMEM
);
864 path
->leave_spinning
= 1;
867 * Find the "default" dir item which points to the root item that we
868 * will mount by default if we haven't been given a specific subvolume
871 dir_id
= btrfs_super_root_dir(fs_info
->super_copy
);
872 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir_id
, "default", 7, 0);
874 btrfs_free_path(path
);
879 * Ok the default dir item isn't there. This is weird since
880 * it's always been there, but don't freak out, just try and
881 * mount to root most subvolume.
883 btrfs_free_path(path
);
884 dir_id
= BTRFS_FIRST_FREE_OBJECTID
;
885 new_root
= fs_info
->fs_root
;
889 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &location
);
890 btrfs_free_path(path
);
893 new_root
= btrfs_read_fs_root_no_name(fs_info
, &location
);
894 if (IS_ERR(new_root
))
895 return ERR_CAST(new_root
);
897 dir_id
= btrfs_root_dirid(&new_root
->root_item
);
899 location
.objectid
= dir_id
;
900 location
.type
= BTRFS_INODE_ITEM_KEY
;
903 inode
= btrfs_iget(sb
, &location
, new_root
, &new);
905 return ERR_CAST(inode
);
908 * If we're just mounting the root most subvol put the inode and return
909 * a reference to the dentry. We will have already gotten a reference
910 * to the inode in btrfs_fill_super so we're good to go.
912 if (!new && sb
->s_root
->d_inode
== inode
) {
914 return dget(sb
->s_root
);
917 dentry
= d_obtain_alias(inode
);
918 if (!IS_ERR(dentry
)) {
919 spin_lock(&dentry
->d_lock
);
920 dentry
->d_flags
&= ~DCACHE_DISCONNECTED
;
921 spin_unlock(&dentry
->d_lock
);
926 static int btrfs_fill_super(struct super_block
*sb
,
927 struct btrfs_fs_devices
*fs_devices
,
928 void *data
, int silent
)
931 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
932 struct btrfs_key key
;
935 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
936 sb
->s_magic
= BTRFS_SUPER_MAGIC
;
937 sb
->s_op
= &btrfs_super_ops
;
938 sb
->s_d_op
= &btrfs_dentry_operations
;
939 sb
->s_export_op
= &btrfs_export_ops
;
940 sb
->s_xattr
= btrfs_xattr_handlers
;
942 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
943 sb
->s_flags
|= MS_POSIXACL
;
945 sb
->s_flags
|= MS_I_VERSION
;
946 err
= open_ctree(sb
, fs_devices
, (char *)data
);
948 printk(KERN_ERR
"BTRFS: open_ctree failed\n");
952 key
.objectid
= BTRFS_FIRST_FREE_OBJECTID
;
953 key
.type
= BTRFS_INODE_ITEM_KEY
;
955 inode
= btrfs_iget(sb
, &key
, fs_info
->fs_root
, NULL
);
957 err
= PTR_ERR(inode
);
961 sb
->s_root
= d_make_root(inode
);
967 save_mount_options(sb
, data
);
968 cleancache_init_fs(sb
);
969 sb
->s_flags
|= MS_ACTIVE
;
973 close_ctree(fs_info
->tree_root
);
977 int btrfs_sync_fs(struct super_block
*sb
, int wait
)
979 struct btrfs_trans_handle
*trans
;
980 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
981 struct btrfs_root
*root
= fs_info
->tree_root
;
983 trace_btrfs_sync_fs(wait
);
986 filemap_flush(fs_info
->btree_inode
->i_mapping
);
990 btrfs_wait_ordered_roots(fs_info
, -1);
992 trans
= btrfs_attach_transaction_barrier(root
);
994 /* no transaction, don't bother */
995 if (PTR_ERR(trans
) == -ENOENT
)
997 return PTR_ERR(trans
);
999 return btrfs_commit_transaction(trans
, root
);
1002 static int btrfs_show_options(struct seq_file
*seq
, struct dentry
*dentry
)
1004 struct btrfs_fs_info
*info
= btrfs_sb(dentry
->d_sb
);
1005 struct btrfs_root
*root
= info
->tree_root
;
1006 char *compress_type
;
1008 if (btrfs_test_opt(root
, DEGRADED
))
1009 seq_puts(seq
, ",degraded");
1010 if (btrfs_test_opt(root
, NODATASUM
))
1011 seq_puts(seq
, ",nodatasum");
1012 if (btrfs_test_opt(root
, NODATACOW
))
1013 seq_puts(seq
, ",nodatacow");
1014 if (btrfs_test_opt(root
, NOBARRIER
))
1015 seq_puts(seq
, ",nobarrier");
1016 if (info
->max_inline
!= 8192 * 1024)
1017 seq_printf(seq
, ",max_inline=%llu", info
->max_inline
);
1018 if (info
->alloc_start
!= 0)
1019 seq_printf(seq
, ",alloc_start=%llu", info
->alloc_start
);
1020 if (info
->thread_pool_size
!= min_t(unsigned long,
1021 num_online_cpus() + 2, 8))
1022 seq_printf(seq
, ",thread_pool=%d", info
->thread_pool_size
);
1023 if (btrfs_test_opt(root
, COMPRESS
)) {
1024 if (info
->compress_type
== BTRFS_COMPRESS_ZLIB
)
1025 compress_type
= "zlib";
1027 compress_type
= "lzo";
1028 if (btrfs_test_opt(root
, FORCE_COMPRESS
))
1029 seq_printf(seq
, ",compress-force=%s", compress_type
);
1031 seq_printf(seq
, ",compress=%s", compress_type
);
1033 if (btrfs_test_opt(root
, NOSSD
))
1034 seq_puts(seq
, ",nossd");
1035 if (btrfs_test_opt(root
, SSD_SPREAD
))
1036 seq_puts(seq
, ",ssd_spread");
1037 else if (btrfs_test_opt(root
, SSD
))
1038 seq_puts(seq
, ",ssd");
1039 if (btrfs_test_opt(root
, NOTREELOG
))
1040 seq_puts(seq
, ",notreelog");
1041 if (btrfs_test_opt(root
, FLUSHONCOMMIT
))
1042 seq_puts(seq
, ",flushoncommit");
1043 if (btrfs_test_opt(root
, DISCARD
))
1044 seq_puts(seq
, ",discard");
1045 if (!(root
->fs_info
->sb
->s_flags
& MS_POSIXACL
))
1046 seq_puts(seq
, ",noacl");
1047 if (btrfs_test_opt(root
, SPACE_CACHE
))
1048 seq_puts(seq
, ",space_cache");
1050 seq_puts(seq
, ",nospace_cache");
1051 if (btrfs_test_opt(root
, RESCAN_UUID_TREE
))
1052 seq_puts(seq
, ",rescan_uuid_tree");
1053 if (btrfs_test_opt(root
, CLEAR_CACHE
))
1054 seq_puts(seq
, ",clear_cache");
1055 if (btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
1056 seq_puts(seq
, ",user_subvol_rm_allowed");
1057 if (btrfs_test_opt(root
, ENOSPC_DEBUG
))
1058 seq_puts(seq
, ",enospc_debug");
1059 if (btrfs_test_opt(root
, AUTO_DEFRAG
))
1060 seq_puts(seq
, ",autodefrag");
1061 if (btrfs_test_opt(root
, INODE_MAP_CACHE
))
1062 seq_puts(seq
, ",inode_cache");
1063 if (btrfs_test_opt(root
, SKIP_BALANCE
))
1064 seq_puts(seq
, ",skip_balance");
1065 if (btrfs_test_opt(root
, RECOVERY
))
1066 seq_puts(seq
, ",recovery");
1067 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1068 if (btrfs_test_opt(root
, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
))
1069 seq_puts(seq
, ",check_int_data");
1070 else if (btrfs_test_opt(root
, CHECK_INTEGRITY
))
1071 seq_puts(seq
, ",check_int");
1072 if (info
->check_integrity_print_mask
)
1073 seq_printf(seq
, ",check_int_print_mask=%d",
1074 info
->check_integrity_print_mask
);
1076 if (info
->metadata_ratio
)
1077 seq_printf(seq
, ",metadata_ratio=%d",
1078 info
->metadata_ratio
);
1079 if (btrfs_test_opt(root
, PANIC_ON_FATAL_ERROR
))
1080 seq_puts(seq
, ",fatal_errors=panic");
1081 if (info
->commit_interval
!= BTRFS_DEFAULT_COMMIT_INTERVAL
)
1082 seq_printf(seq
, ",commit=%d", info
->commit_interval
);
1086 static int btrfs_test_super(struct super_block
*s
, void *data
)
1088 struct btrfs_fs_info
*p
= data
;
1089 struct btrfs_fs_info
*fs_info
= btrfs_sb(s
);
1091 return fs_info
->fs_devices
== p
->fs_devices
;
1094 static int btrfs_set_super(struct super_block
*s
, void *data
)
1096 int err
= set_anon_super(s
, data
);
1098 s
->s_fs_info
= data
;
1103 * subvolumes are identified by ino 256
1105 static inline int is_subvolume_inode(struct inode
*inode
)
1107 if (inode
&& inode
->i_ino
== BTRFS_FIRST_FREE_OBJECTID
)
1113 * This will strip out the subvol=%s argument for an argument string and add
1114 * subvolid=0 to make sure we get the actual tree root for path walking to the
1117 static char *setup_root_args(char *args
)
1119 unsigned len
= strlen(args
) + 2 + 1;
1120 char *src
, *dst
, *buf
;
1123 * We need the same args as before, but with this substitution:
1124 * s!subvol=[^,]+!subvolid=0!
1126 * Since the replacement string is up to 2 bytes longer than the
1127 * original, allocate strlen(args) + 2 + 1 bytes.
1130 src
= strstr(args
, "subvol=");
1131 /* This shouldn't happen, but just in case.. */
1135 buf
= dst
= kmalloc(len
, GFP_NOFS
);
1140 * If the subvol= arg is not at the start of the string,
1141 * copy whatever precedes it into buf.
1146 dst
+= strlen(args
);
1149 strcpy(dst
, "subvolid=0");
1150 dst
+= strlen("subvolid=0");
1153 * If there is a "," after the original subvol=... string,
1154 * copy that suffix into our buffer. Otherwise, we're done.
1156 src
= strchr(src
, ',');
1163 static struct dentry
*mount_subvol(const char *subvol_name
, int flags
,
1164 const char *device_name
, char *data
)
1166 struct dentry
*root
;
1167 struct vfsmount
*mnt
;
1170 newargs
= setup_root_args(data
);
1172 return ERR_PTR(-ENOMEM
);
1173 mnt
= vfs_kern_mount(&btrfs_fs_type
, flags
, device_name
,
1176 if (PTR_RET(mnt
) == -EBUSY
) {
1177 if (flags
& MS_RDONLY
) {
1178 mnt
= vfs_kern_mount(&btrfs_fs_type
, flags
& ~MS_RDONLY
, device_name
,
1182 mnt
= vfs_kern_mount(&btrfs_fs_type
, flags
| MS_RDONLY
, device_name
,
1186 return ERR_CAST(mnt
);
1189 r
= btrfs_remount(mnt
->mnt_sb
, &flags
, NULL
);
1191 /* FIXME: release vfsmount mnt ??*/
1201 return ERR_CAST(mnt
);
1203 root
= mount_subtree(mnt
, subvol_name
);
1205 if (!IS_ERR(root
) && !is_subvolume_inode(root
->d_inode
)) {
1206 struct super_block
*s
= root
->d_sb
;
1208 root
= ERR_PTR(-EINVAL
);
1209 deactivate_locked_super(s
);
1210 printk(KERN_ERR
"BTRFS: '%s' is not a valid subvolume\n",
1218 * Find a superblock for the given device / mount point.
1220 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1221 * for multiple device setup. Make sure to keep it in sync.
1223 static struct dentry
*btrfs_mount(struct file_system_type
*fs_type
, int flags
,
1224 const char *device_name
, void *data
)
1226 struct block_device
*bdev
= NULL
;
1227 struct super_block
*s
;
1228 struct dentry
*root
;
1229 struct btrfs_fs_devices
*fs_devices
= NULL
;
1230 struct btrfs_fs_info
*fs_info
= NULL
;
1231 fmode_t mode
= FMODE_READ
;
1232 char *subvol_name
= NULL
;
1233 u64 subvol_objectid
= 0;
1236 if (!(flags
& MS_RDONLY
))
1237 mode
|= FMODE_WRITE
;
1239 error
= btrfs_parse_early_options(data
, mode
, fs_type
,
1240 &subvol_name
, &subvol_objectid
,
1244 return ERR_PTR(error
);
1248 root
= mount_subvol(subvol_name
, flags
, device_name
, data
);
1253 error
= btrfs_scan_one_device(device_name
, mode
, fs_type
, &fs_devices
);
1255 return ERR_PTR(error
);
1258 * Setup a dummy root and fs_info for test/set super. This is because
1259 * we don't actually fill this stuff out until open_ctree, but we need
1260 * it for searching for existing supers, so this lets us do that and
1261 * then open_ctree will properly initialize everything later.
1263 fs_info
= kzalloc(sizeof(struct btrfs_fs_info
), GFP_NOFS
);
1265 return ERR_PTR(-ENOMEM
);
1267 fs_info
->fs_devices
= fs_devices
;
1269 fs_info
->super_copy
= kzalloc(BTRFS_SUPER_INFO_SIZE
, GFP_NOFS
);
1270 fs_info
->super_for_commit
= kzalloc(BTRFS_SUPER_INFO_SIZE
, GFP_NOFS
);
1271 if (!fs_info
->super_copy
|| !fs_info
->super_for_commit
) {
1276 error
= btrfs_open_devices(fs_devices
, mode
, fs_type
);
1280 if (!(flags
& MS_RDONLY
) && fs_devices
->rw_devices
== 0) {
1282 goto error_close_devices
;
1285 bdev
= fs_devices
->latest_bdev
;
1286 s
= sget(fs_type
, btrfs_test_super
, btrfs_set_super
, flags
| MS_NOSEC
,
1290 goto error_close_devices
;
1294 btrfs_close_devices(fs_devices
);
1295 free_fs_info(fs_info
);
1296 if ((flags
^ s
->s_flags
) & MS_RDONLY
)
1299 char b
[BDEVNAME_SIZE
];
1301 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
1302 btrfs_sb(s
)->bdev_holder
= fs_type
;
1303 error
= btrfs_fill_super(s
, fs_devices
, data
,
1304 flags
& MS_SILENT
? 1 : 0);
1307 root
= !error
? get_default_root(s
, subvol_objectid
) : ERR_PTR(error
);
1309 deactivate_locked_super(s
);
1313 error_close_devices
:
1314 btrfs_close_devices(fs_devices
);
1316 free_fs_info(fs_info
);
1317 return ERR_PTR(error
);
1320 static void btrfs_resize_thread_pool(struct btrfs_fs_info
*fs_info
,
1321 int new_pool_size
, int old_pool_size
)
1323 if (new_pool_size
== old_pool_size
)
1326 fs_info
->thread_pool_size
= new_pool_size
;
1328 btrfs_info(fs_info
, "resize thread pool %d -> %d",
1329 old_pool_size
, new_pool_size
);
1331 btrfs_workqueue_set_max(fs_info
->workers
, new_pool_size
);
1332 btrfs_workqueue_set_max(fs_info
->delalloc_workers
, new_pool_size
);
1333 btrfs_workqueue_set_max(fs_info
->submit_workers
, new_pool_size
);
1334 btrfs_workqueue_set_max(fs_info
->caching_workers
, new_pool_size
);
1335 btrfs_workqueue_set_max(fs_info
->endio_workers
, new_pool_size
);
1336 btrfs_workqueue_set_max(fs_info
->endio_meta_workers
, new_pool_size
);
1337 btrfs_workqueue_set_max(fs_info
->endio_meta_write_workers
,
1339 btrfs_workqueue_set_max(fs_info
->endio_write_workers
, new_pool_size
);
1340 btrfs_workqueue_set_max(fs_info
->endio_freespace_worker
, new_pool_size
);
1341 btrfs_workqueue_set_max(fs_info
->delayed_workers
, new_pool_size
);
1342 btrfs_workqueue_set_max(fs_info
->readahead_workers
, new_pool_size
);
1343 btrfs_workqueue_set_max(fs_info
->scrub_wr_completion_workers
,
1347 static inline void btrfs_remount_prepare(struct btrfs_fs_info
*fs_info
)
1349 set_bit(BTRFS_FS_STATE_REMOUNTING
, &fs_info
->fs_state
);
1352 static inline void btrfs_remount_begin(struct btrfs_fs_info
*fs_info
,
1353 unsigned long old_opts
, int flags
)
1355 if (btrfs_raw_test_opt(old_opts
, AUTO_DEFRAG
) &&
1356 (!btrfs_raw_test_opt(fs_info
->mount_opt
, AUTO_DEFRAG
) ||
1357 (flags
& MS_RDONLY
))) {
1358 /* wait for any defraggers to finish */
1359 wait_event(fs_info
->transaction_wait
,
1360 (atomic_read(&fs_info
->defrag_running
) == 0));
1361 if (flags
& MS_RDONLY
)
1362 sync_filesystem(fs_info
->sb
);
1366 static inline void btrfs_remount_cleanup(struct btrfs_fs_info
*fs_info
,
1367 unsigned long old_opts
)
1370 * We need cleanup all defragable inodes if the autodefragment is
1371 * close or the fs is R/O.
1373 if (btrfs_raw_test_opt(old_opts
, AUTO_DEFRAG
) &&
1374 (!btrfs_raw_test_opt(fs_info
->mount_opt
, AUTO_DEFRAG
) ||
1375 (fs_info
->sb
->s_flags
& MS_RDONLY
))) {
1376 btrfs_cleanup_defrag_inodes(fs_info
);
1379 clear_bit(BTRFS_FS_STATE_REMOUNTING
, &fs_info
->fs_state
);
1382 static int btrfs_remount(struct super_block
*sb
, int *flags
, char *data
)
1384 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1385 struct btrfs_root
*root
= fs_info
->tree_root
;
1386 unsigned old_flags
= sb
->s_flags
;
1387 unsigned long old_opts
= fs_info
->mount_opt
;
1388 unsigned long old_compress_type
= fs_info
->compress_type
;
1389 u64 old_max_inline
= fs_info
->max_inline
;
1390 u64 old_alloc_start
= fs_info
->alloc_start
;
1391 int old_thread_pool_size
= fs_info
->thread_pool_size
;
1392 unsigned int old_metadata_ratio
= fs_info
->metadata_ratio
;
1395 sync_filesystem(sb
);
1396 btrfs_remount_prepare(fs_info
);
1398 ret
= btrfs_parse_options(root
, data
);
1404 btrfs_remount_begin(fs_info
, old_opts
, *flags
);
1405 btrfs_resize_thread_pool(fs_info
,
1406 fs_info
->thread_pool_size
, old_thread_pool_size
);
1408 if ((*flags
& MS_RDONLY
) == (sb
->s_flags
& MS_RDONLY
))
1411 if (*flags
& MS_RDONLY
) {
1413 * this also happens on 'umount -rf' or on shutdown, when
1414 * the filesystem is busy.
1417 /* wait for the uuid_scan task to finish */
1418 down(&fs_info
->uuid_tree_rescan_sem
);
1419 /* avoid complains from lockdep et al. */
1420 up(&fs_info
->uuid_tree_rescan_sem
);
1422 sb
->s_flags
|= MS_RDONLY
;
1424 btrfs_dev_replace_suspend_for_unmount(fs_info
);
1425 btrfs_scrub_cancel(fs_info
);
1426 btrfs_pause_balance(fs_info
);
1428 ret
= btrfs_commit_super(root
);
1432 if (test_bit(BTRFS_FS_STATE_ERROR
, &root
->fs_info
->fs_state
)) {
1434 "Remounting read-write after error is not allowed");
1438 if (fs_info
->fs_devices
->rw_devices
== 0) {
1443 if (fs_info
->fs_devices
->missing_devices
>
1444 fs_info
->num_tolerated_disk_barrier_failures
&&
1445 !(*flags
& MS_RDONLY
)) {
1447 "too many missing devices, writeable remount is not allowed");
1452 if (btrfs_super_log_root(fs_info
->super_copy
) != 0) {
1457 ret
= btrfs_cleanup_fs_roots(fs_info
);
1461 /* recover relocation */
1462 ret
= btrfs_recover_relocation(root
);
1466 ret
= btrfs_resume_balance_async(fs_info
);
1470 ret
= btrfs_resume_dev_replace_async(fs_info
);
1472 btrfs_warn(fs_info
, "failed to resume dev_replace");
1476 if (!fs_info
->uuid_root
) {
1477 btrfs_info(fs_info
, "creating UUID tree");
1478 ret
= btrfs_create_uuid_tree(fs_info
);
1480 btrfs_warn(fs_info
, "failed to create the UUID tree %d", ret
);
1484 sb
->s_flags
&= ~MS_RDONLY
;
1487 wake_up_process(fs_info
->transaction_kthread
);
1488 btrfs_remount_cleanup(fs_info
, old_opts
);
1492 /* We've hit an error - don't reset MS_RDONLY */
1493 if (sb
->s_flags
& MS_RDONLY
)
1494 old_flags
|= MS_RDONLY
;
1495 sb
->s_flags
= old_flags
;
1496 fs_info
->mount_opt
= old_opts
;
1497 fs_info
->compress_type
= old_compress_type
;
1498 fs_info
->max_inline
= old_max_inline
;
1499 mutex_lock(&fs_info
->chunk_mutex
);
1500 fs_info
->alloc_start
= old_alloc_start
;
1501 mutex_unlock(&fs_info
->chunk_mutex
);
1502 btrfs_resize_thread_pool(fs_info
,
1503 old_thread_pool_size
, fs_info
->thread_pool_size
);
1504 fs_info
->metadata_ratio
= old_metadata_ratio
;
1505 btrfs_remount_cleanup(fs_info
, old_opts
);
1509 /* Used to sort the devices by max_avail(descending sort) */
1510 static int btrfs_cmp_device_free_bytes(const void *dev_info1
,
1511 const void *dev_info2
)
1513 if (((struct btrfs_device_info
*)dev_info1
)->max_avail
>
1514 ((struct btrfs_device_info
*)dev_info2
)->max_avail
)
1516 else if (((struct btrfs_device_info
*)dev_info1
)->max_avail
<
1517 ((struct btrfs_device_info
*)dev_info2
)->max_avail
)
1524 * sort the devices by max_avail, in which max free extent size of each device
1525 * is stored.(Descending Sort)
1527 static inline void btrfs_descending_sort_devices(
1528 struct btrfs_device_info
*devices
,
1531 sort(devices
, nr_devices
, sizeof(struct btrfs_device_info
),
1532 btrfs_cmp_device_free_bytes
, NULL
);
1536 * The helper to calc the free space on the devices that can be used to store
1539 static int btrfs_calc_avail_data_space(struct btrfs_root
*root
, u64
*free_bytes
)
1541 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1542 struct btrfs_device_info
*devices_info
;
1543 struct btrfs_fs_devices
*fs_devices
= fs_info
->fs_devices
;
1544 struct btrfs_device
*device
;
1549 u64 min_stripe_size
;
1550 int min_stripes
= 1, num_stripes
= 1;
1551 int i
= 0, nr_devices
;
1554 nr_devices
= fs_info
->fs_devices
->open_devices
;
1555 BUG_ON(!nr_devices
);
1557 devices_info
= kmalloc_array(nr_devices
, sizeof(*devices_info
),
1562 /* calc min stripe number for data space alloction */
1563 type
= btrfs_get_alloc_profile(root
, 1);
1564 if (type
& BTRFS_BLOCK_GROUP_RAID0
) {
1566 num_stripes
= nr_devices
;
1567 } else if (type
& BTRFS_BLOCK_GROUP_RAID1
) {
1570 } else if (type
& BTRFS_BLOCK_GROUP_RAID10
) {
1575 if (type
& BTRFS_BLOCK_GROUP_DUP
)
1576 min_stripe_size
= 2 * BTRFS_STRIPE_LEN
;
1578 min_stripe_size
= BTRFS_STRIPE_LEN
;
1580 list_for_each_entry(device
, &fs_devices
->devices
, dev_list
) {
1581 if (!device
->in_fs_metadata
|| !device
->bdev
||
1582 device
->is_tgtdev_for_dev_replace
)
1585 avail_space
= device
->total_bytes
- device
->bytes_used
;
1587 /* align with stripe_len */
1588 do_div(avail_space
, BTRFS_STRIPE_LEN
);
1589 avail_space
*= BTRFS_STRIPE_LEN
;
1592 * In order to avoid overwritting the superblock on the drive,
1593 * btrfs starts at an offset of at least 1MB when doing chunk
1596 skip_space
= 1024 * 1024;
1598 /* user can set the offset in fs_info->alloc_start. */
1599 if (fs_info
->alloc_start
+ BTRFS_STRIPE_LEN
<=
1600 device
->total_bytes
)
1601 skip_space
= max(fs_info
->alloc_start
, skip_space
);
1604 * btrfs can not use the free space in [0, skip_space - 1],
1605 * we must subtract it from the total. In order to implement
1606 * it, we account the used space in this range first.
1608 ret
= btrfs_account_dev_extents_size(device
, 0, skip_space
- 1,
1611 kfree(devices_info
);
1615 /* calc the free space in [0, skip_space - 1] */
1616 skip_space
-= used_space
;
1619 * we can use the free space in [0, skip_space - 1], subtract
1620 * it from the total.
1622 if (avail_space
&& avail_space
>= skip_space
)
1623 avail_space
-= skip_space
;
1627 if (avail_space
< min_stripe_size
)
1630 devices_info
[i
].dev
= device
;
1631 devices_info
[i
].max_avail
= avail_space
;
1638 btrfs_descending_sort_devices(devices_info
, nr_devices
);
1642 while (nr_devices
>= min_stripes
) {
1643 if (num_stripes
> nr_devices
)
1644 num_stripes
= nr_devices
;
1646 if (devices_info
[i
].max_avail
>= min_stripe_size
) {
1650 avail_space
+= devices_info
[i
].max_avail
* num_stripes
;
1651 alloc_size
= devices_info
[i
].max_avail
;
1652 for (j
= i
+ 1 - num_stripes
; j
<= i
; j
++)
1653 devices_info
[j
].max_avail
-= alloc_size
;
1659 kfree(devices_info
);
1660 *free_bytes
= avail_space
;
1664 static int btrfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1666 struct btrfs_fs_info
*fs_info
= btrfs_sb(dentry
->d_sb
);
1667 struct btrfs_super_block
*disk_super
= fs_info
->super_copy
;
1668 struct list_head
*head
= &fs_info
->space_info
;
1669 struct btrfs_space_info
*found
;
1671 u64 total_free_data
= 0;
1672 int bits
= dentry
->d_sb
->s_blocksize_bits
;
1673 __be32
*fsid
= (__be32
*)fs_info
->fsid
;
1676 /* holding chunk_muext to avoid allocating new chunks */
1677 mutex_lock(&fs_info
->chunk_mutex
);
1679 list_for_each_entry_rcu(found
, head
, list
) {
1680 if (found
->flags
& BTRFS_BLOCK_GROUP_DATA
) {
1681 total_free_data
+= found
->disk_total
- found
->disk_used
;
1683 btrfs_account_ro_block_groups_free_space(found
);
1686 total_used
+= found
->disk_used
;
1690 buf
->f_namelen
= BTRFS_NAME_LEN
;
1691 buf
->f_blocks
= btrfs_super_total_bytes(disk_super
) >> bits
;
1692 buf
->f_bfree
= buf
->f_blocks
- (total_used
>> bits
);
1693 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1694 buf
->f_type
= BTRFS_SUPER_MAGIC
;
1695 buf
->f_bavail
= total_free_data
;
1696 ret
= btrfs_calc_avail_data_space(fs_info
->tree_root
, &total_free_data
);
1698 mutex_unlock(&fs_info
->chunk_mutex
);
1701 buf
->f_bavail
+= total_free_data
;
1702 buf
->f_bavail
= buf
->f_bavail
>> bits
;
1703 mutex_unlock(&fs_info
->chunk_mutex
);
1705 /* We treat it as constant endianness (it doesn't matter _which_)
1706 because we want the fsid to come out the same whether mounted
1707 on a big-endian or little-endian host */
1708 buf
->f_fsid
.val
[0] = be32_to_cpu(fsid
[0]) ^ be32_to_cpu(fsid
[2]);
1709 buf
->f_fsid
.val
[1] = be32_to_cpu(fsid
[1]) ^ be32_to_cpu(fsid
[3]);
1710 /* Mask in the root object ID too, to disambiguate subvols */
1711 buf
->f_fsid
.val
[0] ^= BTRFS_I(dentry
->d_inode
)->root
->objectid
>> 32;
1712 buf
->f_fsid
.val
[1] ^= BTRFS_I(dentry
->d_inode
)->root
->objectid
;
1717 static void btrfs_kill_super(struct super_block
*sb
)
1719 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1720 kill_anon_super(sb
);
1721 free_fs_info(fs_info
);
1724 static struct file_system_type btrfs_fs_type
= {
1725 .owner
= THIS_MODULE
,
1727 .mount
= btrfs_mount
,
1728 .kill_sb
= btrfs_kill_super
,
1729 .fs_flags
= FS_REQUIRES_DEV
,
1731 MODULE_ALIAS_FS("btrfs");
1734 * used by btrfsctl to scan devices when no FS is mounted
1736 static long btrfs_control_ioctl(struct file
*file
, unsigned int cmd
,
1739 struct btrfs_ioctl_vol_args
*vol
;
1740 struct btrfs_fs_devices
*fs_devices
;
1743 if (!capable(CAP_SYS_ADMIN
))
1746 vol
= memdup_user((void __user
*)arg
, sizeof(*vol
));
1748 return PTR_ERR(vol
);
1751 case BTRFS_IOC_SCAN_DEV
:
1752 ret
= btrfs_scan_one_device(vol
->name
, FMODE_READ
,
1753 &btrfs_fs_type
, &fs_devices
);
1755 case BTRFS_IOC_DEVICES_READY
:
1756 ret
= btrfs_scan_one_device(vol
->name
, FMODE_READ
,
1757 &btrfs_fs_type
, &fs_devices
);
1760 ret
= !(fs_devices
->num_devices
== fs_devices
->total_devices
);
1768 static int btrfs_freeze(struct super_block
*sb
)
1770 struct btrfs_trans_handle
*trans
;
1771 struct btrfs_root
*root
= btrfs_sb(sb
)->tree_root
;
1773 trans
= btrfs_attach_transaction_barrier(root
);
1774 if (IS_ERR(trans
)) {
1775 /* no transaction, don't bother */
1776 if (PTR_ERR(trans
) == -ENOENT
)
1778 return PTR_ERR(trans
);
1780 return btrfs_commit_transaction(trans
, root
);
1783 static int btrfs_unfreeze(struct super_block
*sb
)
1788 static int btrfs_show_devname(struct seq_file
*m
, struct dentry
*root
)
1790 struct btrfs_fs_info
*fs_info
= btrfs_sb(root
->d_sb
);
1791 struct btrfs_fs_devices
*cur_devices
;
1792 struct btrfs_device
*dev
, *first_dev
= NULL
;
1793 struct list_head
*head
;
1794 struct rcu_string
*name
;
1796 mutex_lock(&fs_info
->fs_devices
->device_list_mutex
);
1797 cur_devices
= fs_info
->fs_devices
;
1798 while (cur_devices
) {
1799 head
= &cur_devices
->devices
;
1800 list_for_each_entry(dev
, head
, dev_list
) {
1803 if (!first_dev
|| dev
->devid
< first_dev
->devid
)
1806 cur_devices
= cur_devices
->seed
;
1811 name
= rcu_dereference(first_dev
->name
);
1812 seq_escape(m
, name
->str
, " \t\n\\");
1817 mutex_unlock(&fs_info
->fs_devices
->device_list_mutex
);
1821 static const struct super_operations btrfs_super_ops
= {
1822 .drop_inode
= btrfs_drop_inode
,
1823 .evict_inode
= btrfs_evict_inode
,
1824 .put_super
= btrfs_put_super
,
1825 .sync_fs
= btrfs_sync_fs
,
1826 .show_options
= btrfs_show_options
,
1827 .show_devname
= btrfs_show_devname
,
1828 .write_inode
= btrfs_write_inode
,
1829 .alloc_inode
= btrfs_alloc_inode
,
1830 .destroy_inode
= btrfs_destroy_inode
,
1831 .statfs
= btrfs_statfs
,
1832 .remount_fs
= btrfs_remount
,
1833 .freeze_fs
= btrfs_freeze
,
1834 .unfreeze_fs
= btrfs_unfreeze
,
1837 static const struct file_operations btrfs_ctl_fops
= {
1838 .unlocked_ioctl
= btrfs_control_ioctl
,
1839 .compat_ioctl
= btrfs_control_ioctl
,
1840 .owner
= THIS_MODULE
,
1841 .llseek
= noop_llseek
,
1844 static struct miscdevice btrfs_misc
= {
1845 .minor
= BTRFS_MINOR
,
1846 .name
= "btrfs-control",
1847 .fops
= &btrfs_ctl_fops
1850 MODULE_ALIAS_MISCDEV(BTRFS_MINOR
);
1851 MODULE_ALIAS("devname:btrfs-control");
1853 static int btrfs_interface_init(void)
1855 return misc_register(&btrfs_misc
);
1858 static void btrfs_interface_exit(void)
1860 if (misc_deregister(&btrfs_misc
) < 0)
1861 printk(KERN_INFO
"BTRFS: misc_deregister failed for control device\n");
1864 static void btrfs_print_info(void)
1866 printk(KERN_INFO
"Btrfs loaded"
1867 #ifdef CONFIG_BTRFS_DEBUG
1870 #ifdef CONFIG_BTRFS_ASSERT
1873 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1874 ", integrity-checker=on"
1879 static int btrfs_run_sanity_tests(void)
1883 ret
= btrfs_init_test_fs();
1887 ret
= btrfs_test_free_space_cache();
1890 ret
= btrfs_test_extent_buffer_operations();
1893 ret
= btrfs_test_extent_io();
1896 ret
= btrfs_test_inodes();
1898 btrfs_destroy_test_fs();
1902 static int __init
init_btrfs_fs(void)
1906 err
= btrfs_hash_init();
1912 err
= btrfs_init_sysfs();
1916 btrfs_init_compress();
1918 err
= btrfs_init_cachep();
1922 err
= extent_io_init();
1926 err
= extent_map_init();
1928 goto free_extent_io
;
1930 err
= ordered_data_init();
1932 goto free_extent_map
;
1934 err
= btrfs_delayed_inode_init();
1936 goto free_ordered_data
;
1938 err
= btrfs_auto_defrag_init();
1940 goto free_delayed_inode
;
1942 err
= btrfs_delayed_ref_init();
1944 goto free_auto_defrag
;
1946 err
= btrfs_prelim_ref_init();
1948 goto free_prelim_ref
;
1950 err
= btrfs_interface_init();
1952 goto free_delayed_ref
;
1954 btrfs_init_lockdep();
1958 err
= btrfs_run_sanity_tests();
1960 goto unregister_ioctl
;
1962 err
= register_filesystem(&btrfs_fs_type
);
1964 goto unregister_ioctl
;
1969 btrfs_interface_exit();
1971 btrfs_prelim_ref_exit();
1973 btrfs_delayed_ref_exit();
1975 btrfs_auto_defrag_exit();
1977 btrfs_delayed_inode_exit();
1979 ordered_data_exit();
1985 btrfs_destroy_cachep();
1987 btrfs_exit_compress();
1994 static void __exit
exit_btrfs_fs(void)
1996 btrfs_destroy_cachep();
1997 btrfs_delayed_ref_exit();
1998 btrfs_auto_defrag_exit();
1999 btrfs_delayed_inode_exit();
2000 btrfs_prelim_ref_exit();
2001 ordered_data_exit();
2004 btrfs_interface_exit();
2005 unregister_filesystem(&btrfs_fs_type
);
2007 btrfs_cleanup_fs_uuids();
2008 btrfs_exit_compress();
2012 late_initcall(init_btrfs_fs
);
2013 module_exit(exit_btrfs_fs
)
2015 MODULE_LICENSE("GPL");