2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/blkdev.h>
20 #include <linux/module.h>
21 #include <linux/buffer_head.h>
23 #include <linux/pagemap.h>
24 #include <linux/highmem.h>
25 #include <linux/time.h>
26 #include <linux/init.h>
27 #include <linux/seq_file.h>
28 #include <linux/string.h>
29 #include <linux/backing-dev.h>
30 #include <linux/mount.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/parser.h>
37 #include <linux/ctype.h>
38 #include <linux/namei.h>
39 #include <linux/miscdevice.h>
40 #include <linux/magic.h>
41 #include <linux/slab.h>
42 #include <linux/cleancache.h>
43 #include <linux/ratelimit.h>
44 #include <linux/btrfs.h>
45 #include "delayed-inode.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "print-tree.h"
56 #include "compression.h"
57 #include "rcu-string.h"
58 #include "dev-replace.h"
59 #include "free-space-cache.h"
61 #include "tests/btrfs-tests.h"
64 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h>
67 static const struct super_operations btrfs_super_ops;
70 * Types for mounting the default subvolume and a subvolume explicitly
71 * requested by subvol=/path. That way the callchain is straightforward and we
72 * don't have to play tricks with the mount options and recursive calls to
75 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
77 static struct file_system_type btrfs_fs_type;
78 static struct file_system_type btrfs_root_fs_type;
80 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
82 const char *btrfs_decode_error(int errno)
84 char *errstr = "unknown";
88 errstr = "IO failure";
91 errstr = "Out of memory";
94 errstr = "Readonly filesystem";
97 errstr = "Object already exists";
100 errstr = "No space left";
103 errstr = "No such entry";
111 * __btrfs_handle_fs_error decodes expected errors from the caller and
112 * invokes the approciate error response.
115 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
116 unsigned int line, int errno, const char *fmt, ...)
118 struct super_block *sb = fs_info->sb;
124 * Special case: if the error is EROFS, and we're already
125 * under SB_RDONLY, then it is safe here.
127 if (errno == -EROFS && sb_rdonly(sb))
131 errstr = btrfs_decode_error(errno);
133 struct va_format vaf;
140 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
141 sb->s_id, function, line, errno, errstr, &vaf);
144 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
145 sb->s_id, function, line, errno, errstr);
150 * Today we only save the error info to memory. Long term we'll
151 * also send it down to the disk
153 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
155 /* Don't go through full error handling during mount */
156 if (!(sb->s_flags & SB_BORN))
162 /* btrfs handle error by forcing the filesystem readonly */
163 sb->s_flags |= SB_RDONLY;
164 btrfs_info(fs_info, "forced readonly");
166 * Note that a running device replace operation is not canceled here
167 * although there is no way to update the progress. It would add the
168 * risk of a deadlock, therefore the canceling is omitted. The only
169 * penalty is that some I/O remains active until the procedure
170 * completes. The next time when the filesystem is mounted writeable
171 * again, the device replace operation continues.
176 static const char * const logtypes[] = {
189 * Use one ratelimit state per log level so that a flood of less important
190 * messages doesn't cause more important ones to be dropped.
192 static struct ratelimit_state printk_limits[] = {
193 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
194 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
195 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
196 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
197 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
198 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
199 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
200 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
203 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
205 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
206 struct va_format vaf;
209 const char *type = logtypes[4];
210 struct ratelimit_state *ratelimit = &printk_limits[4];
214 while ((kern_level = printk_get_level(fmt)) != 0) {
215 size_t size = printk_skip_level(fmt) - fmt;
217 if (kern_level >= '0' && kern_level <= '7') {
218 memcpy(lvl, fmt, size);
220 type = logtypes[kern_level - '0'];
221 ratelimit = &printk_limits[kern_level - '0'];
229 if (__ratelimit(ratelimit))
230 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
231 fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
238 * We only mark the transaction aborted and then set the file system read-only.
239 * This will prevent new transactions from starting or trying to join this
242 * This means that error recovery at the call site is limited to freeing
243 * any local memory allocations and passing the error code up without
244 * further cleanup. The transaction should complete as it normally would
245 * in the call path but will return -EIO.
247 * We'll complete the cleanup in btrfs_end_transaction and
248 * btrfs_commit_transaction.
251 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
252 const char *function,
253 unsigned int line, int errno)
255 struct btrfs_fs_info *fs_info = trans->fs_info;
257 trans->aborted = errno;
258 /* Nothing used. The other threads that have joined this
259 * transaction may be able to continue. */
260 if (!trans->dirty && list_empty(&trans->new_bgs)) {
263 errstr = btrfs_decode_error(errno);
265 "%s:%d: Aborting unused transaction(%s).",
266 function, line, errstr);
269 WRITE_ONCE(trans->transaction->aborted, errno);
270 /* Wake up anybody who may be waiting on this transaction */
271 wake_up(&fs_info->transaction_wait);
272 wake_up(&fs_info->transaction_blocked_wait);
273 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
276 * __btrfs_panic decodes unexpected, fatal errors from the caller,
277 * issues an alert, and either panics or BUGs, depending on mount options.
280 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
281 unsigned int line, int errno, const char *fmt, ...)
283 char *s_id = "<unknown>";
285 struct va_format vaf = { .fmt = fmt };
289 s_id = fs_info->sb->s_id;
294 errstr = btrfs_decode_error(errno);
295 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
296 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
297 s_id, function, line, &vaf, errno, errstr);
299 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
300 function, line, &vaf, errno, errstr);
302 /* Caller calls BUG() */
305 static void btrfs_put_super(struct super_block *sb)
307 close_ctree(btrfs_sb(sb));
311 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
312 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
313 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
314 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
315 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
316 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
317 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
318 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
319 Opt_skip_balance, Opt_check_integrity,
320 Opt_check_integrity_including_extent_data,
321 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
322 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
323 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
324 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
325 Opt_nologreplay, Opt_norecovery,
326 #ifdef CONFIG_BTRFS_DEBUG
327 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
329 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
335 static const match_table_t tokens = {
336 {Opt_degraded, "degraded"},
337 {Opt_subvol, "subvol=%s"},
338 {Opt_subvolid, "subvolid=%s"},
339 {Opt_device, "device=%s"},
340 {Opt_nodatasum, "nodatasum"},
341 {Opt_datasum, "datasum"},
342 {Opt_nodatacow, "nodatacow"},
343 {Opt_datacow, "datacow"},
344 {Opt_nobarrier, "nobarrier"},
345 {Opt_barrier, "barrier"},
346 {Opt_max_inline, "max_inline=%s"},
347 {Opt_alloc_start, "alloc_start=%s"},
348 {Opt_thread_pool, "thread_pool=%u"},
349 {Opt_compress, "compress"},
350 {Opt_compress_type, "compress=%s"},
351 {Opt_compress_force, "compress-force"},
352 {Opt_compress_force_type, "compress-force=%s"},
354 {Opt_ssd_spread, "ssd_spread"},
355 {Opt_nossd, "nossd"},
357 {Opt_noacl, "noacl"},
358 {Opt_notreelog, "notreelog"},
359 {Opt_treelog, "treelog"},
360 {Opt_nologreplay, "nologreplay"},
361 {Opt_norecovery, "norecovery"},
362 {Opt_flushoncommit, "flushoncommit"},
363 {Opt_noflushoncommit, "noflushoncommit"},
364 {Opt_ratio, "metadata_ratio=%u"},
365 {Opt_discard, "discard"},
366 {Opt_nodiscard, "nodiscard"},
367 {Opt_space_cache, "space_cache"},
368 {Opt_space_cache_version, "space_cache=%s"},
369 {Opt_clear_cache, "clear_cache"},
370 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
371 {Opt_enospc_debug, "enospc_debug"},
372 {Opt_noenospc_debug, "noenospc_debug"},
373 {Opt_subvolrootid, "subvolrootid=%d"},
374 {Opt_defrag, "autodefrag"},
375 {Opt_nodefrag, "noautodefrag"},
376 {Opt_inode_cache, "inode_cache"},
377 {Opt_noinode_cache, "noinode_cache"},
378 {Opt_no_space_cache, "nospace_cache"},
379 {Opt_recovery, "recovery"}, /* deprecated */
380 {Opt_usebackuproot, "usebackuproot"},
381 {Opt_skip_balance, "skip_balance"},
382 {Opt_check_integrity, "check_int"},
383 {Opt_check_integrity_including_extent_data, "check_int_data"},
384 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
385 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
386 {Opt_fatal_errors, "fatal_errors=%s"},
387 {Opt_commit_interval, "commit=%d"},
388 #ifdef CONFIG_BTRFS_DEBUG
389 {Opt_fragment_data, "fragment=data"},
390 {Opt_fragment_metadata, "fragment=metadata"},
391 {Opt_fragment_all, "fragment=all"},
393 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
394 {Opt_ref_verify, "ref_verify"},
400 * Regular mount options parser. Everything that is needed only when
401 * reading in a new superblock is parsed here.
402 * XXX JDM: This needs to be cleaned up for remount.
404 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
405 unsigned long new_flags)
407 substring_t args[MAX_OPT_ARGS];
413 bool compress_force = false;
414 enum btrfs_compression_type saved_compress_type;
415 bool saved_compress_force;
418 cache_gen = btrfs_super_cache_generation(info->super_copy);
419 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
420 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
422 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
425 * Even the options are empty, we still need to do extra check
431 while ((p = strsep(&options, ",")) != NULL) {
436 token = match_token(p, tokens, args);
439 btrfs_info(info, "allowing degraded mounts");
440 btrfs_set_opt(info->mount_opt, DEGRADED);
444 case Opt_subvolrootid:
447 * These are parsed by btrfs_parse_subvol_options
448 * and btrfs_parse_early_options
449 * and can be happily ignored here.
453 btrfs_set_and_info(info, NODATASUM,
454 "setting nodatasum");
457 if (btrfs_test_opt(info, NODATASUM)) {
458 if (btrfs_test_opt(info, NODATACOW))
460 "setting datasum, datacow enabled");
462 btrfs_info(info, "setting datasum");
464 btrfs_clear_opt(info->mount_opt, NODATACOW);
465 btrfs_clear_opt(info->mount_opt, NODATASUM);
468 if (!btrfs_test_opt(info, NODATACOW)) {
469 if (!btrfs_test_opt(info, COMPRESS) ||
470 !btrfs_test_opt(info, FORCE_COMPRESS)) {
472 "setting nodatacow, compression disabled");
474 btrfs_info(info, "setting nodatacow");
477 btrfs_clear_opt(info->mount_opt, COMPRESS);
478 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
479 btrfs_set_opt(info->mount_opt, NODATACOW);
480 btrfs_set_opt(info->mount_opt, NODATASUM);
483 btrfs_clear_and_info(info, NODATACOW,
486 case Opt_compress_force:
487 case Opt_compress_force_type:
488 compress_force = true;
491 case Opt_compress_type:
492 saved_compress_type = btrfs_test_opt(info,
494 info->compress_type : BTRFS_COMPRESS_NONE;
495 saved_compress_force =
496 btrfs_test_opt(info, FORCE_COMPRESS);
497 if (token == Opt_compress ||
498 token == Opt_compress_force ||
499 strncmp(args[0].from, "zlib", 4) == 0) {
500 compress_type = "zlib";
502 info->compress_type = BTRFS_COMPRESS_ZLIB;
503 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
505 * args[0] contains uninitialized data since
506 * for these tokens we don't expect any
509 if (token != Opt_compress &&
510 token != Opt_compress_force)
511 info->compress_level =
512 btrfs_compress_str2level(args[0].from);
513 btrfs_set_opt(info->mount_opt, COMPRESS);
514 btrfs_clear_opt(info->mount_opt, NODATACOW);
515 btrfs_clear_opt(info->mount_opt, NODATASUM);
517 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
518 compress_type = "lzo";
519 info->compress_type = BTRFS_COMPRESS_LZO;
520 btrfs_set_opt(info->mount_opt, COMPRESS);
521 btrfs_clear_opt(info->mount_opt, NODATACOW);
522 btrfs_clear_opt(info->mount_opt, NODATASUM);
523 btrfs_set_fs_incompat(info, COMPRESS_LZO);
525 } else if (strcmp(args[0].from, "zstd") == 0) {
526 compress_type = "zstd";
527 info->compress_type = BTRFS_COMPRESS_ZSTD;
528 btrfs_set_opt(info->mount_opt, COMPRESS);
529 btrfs_clear_opt(info->mount_opt, NODATACOW);
530 btrfs_clear_opt(info->mount_opt, NODATASUM);
531 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
533 } else if (strncmp(args[0].from, "no", 2) == 0) {
534 compress_type = "no";
535 btrfs_clear_opt(info->mount_opt, COMPRESS);
536 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
537 compress_force = false;
544 if (compress_force) {
545 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
548 * If we remount from compress-force=xxx to
549 * compress=xxx, we need clear FORCE_COMPRESS
550 * flag, otherwise, there is no way for users
551 * to disable forcible compression separately.
553 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
555 if ((btrfs_test_opt(info, COMPRESS) &&
556 (info->compress_type != saved_compress_type ||
557 compress_force != saved_compress_force)) ||
558 (!btrfs_test_opt(info, COMPRESS) &&
560 btrfs_info(info, "%s %s compression, level %d",
561 (compress_force) ? "force" : "use",
562 compress_type, info->compress_level);
564 compress_force = false;
567 btrfs_set_and_info(info, SSD,
568 "enabling ssd optimizations");
569 btrfs_clear_opt(info->mount_opt, NOSSD);
572 btrfs_set_and_info(info, SSD,
573 "enabling ssd optimizations");
574 btrfs_set_and_info(info, SSD_SPREAD,
575 "using spread ssd allocation scheme");
576 btrfs_clear_opt(info->mount_opt, NOSSD);
579 btrfs_set_opt(info->mount_opt, NOSSD);
580 btrfs_clear_and_info(info, SSD,
581 "not using ssd optimizations");
582 btrfs_clear_and_info(info, SSD_SPREAD,
583 "not using spread ssd allocation scheme");
586 btrfs_clear_and_info(info, NOBARRIER,
587 "turning on barriers");
590 btrfs_set_and_info(info, NOBARRIER,
591 "turning off barriers");
593 case Opt_thread_pool:
594 ret = match_int(&args[0], &intarg);
597 } else if (intarg == 0) {
601 info->thread_pool_size = intarg;
604 num = match_strdup(&args[0]);
606 info->max_inline = memparse(num, NULL);
609 if (info->max_inline) {
610 info->max_inline = min_t(u64,
614 btrfs_info(info, "max_inline at %llu",
621 case Opt_alloc_start:
623 "option alloc_start is obsolete, ignored");
626 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
627 info->sb->s_flags |= SB_POSIXACL;
630 btrfs_err(info, "support for ACL not compiled in!");
635 info->sb->s_flags &= ~SB_POSIXACL;
638 btrfs_set_and_info(info, NOTREELOG,
639 "disabling tree log");
642 btrfs_clear_and_info(info, NOTREELOG,
643 "enabling tree log");
646 case Opt_nologreplay:
647 btrfs_set_and_info(info, NOLOGREPLAY,
648 "disabling log replay at mount time");
650 case Opt_flushoncommit:
651 btrfs_set_and_info(info, FLUSHONCOMMIT,
652 "turning on flush-on-commit");
654 case Opt_noflushoncommit:
655 btrfs_clear_and_info(info, FLUSHONCOMMIT,
656 "turning off flush-on-commit");
659 ret = match_int(&args[0], &intarg);
662 info->metadata_ratio = intarg;
663 btrfs_info(info, "metadata ratio %u",
664 info->metadata_ratio);
667 btrfs_set_and_info(info, DISCARD,
668 "turning on discard");
671 btrfs_clear_and_info(info, DISCARD,
672 "turning off discard");
674 case Opt_space_cache:
675 case Opt_space_cache_version:
676 if (token == Opt_space_cache ||
677 strcmp(args[0].from, "v1") == 0) {
678 btrfs_clear_opt(info->mount_opt,
680 btrfs_set_and_info(info, SPACE_CACHE,
681 "enabling disk space caching");
682 } else if (strcmp(args[0].from, "v2") == 0) {
683 btrfs_clear_opt(info->mount_opt,
685 btrfs_set_and_info(info, FREE_SPACE_TREE,
686 "enabling free space tree");
692 case Opt_rescan_uuid_tree:
693 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
695 case Opt_no_space_cache:
696 if (btrfs_test_opt(info, SPACE_CACHE)) {
697 btrfs_clear_and_info(info, SPACE_CACHE,
698 "disabling disk space caching");
700 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
701 btrfs_clear_and_info(info, FREE_SPACE_TREE,
702 "disabling free space tree");
705 case Opt_inode_cache:
706 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
707 "enabling inode map caching");
709 case Opt_noinode_cache:
710 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
711 "disabling inode map caching");
713 case Opt_clear_cache:
714 btrfs_set_and_info(info, CLEAR_CACHE,
715 "force clearing of disk cache");
717 case Opt_user_subvol_rm_allowed:
718 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
720 case Opt_enospc_debug:
721 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
723 case Opt_noenospc_debug:
724 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
727 btrfs_set_and_info(info, AUTO_DEFRAG,
728 "enabling auto defrag");
731 btrfs_clear_and_info(info, AUTO_DEFRAG,
732 "disabling auto defrag");
736 "'recovery' is deprecated, use 'usebackuproot' instead");
737 case Opt_usebackuproot:
739 "trying to use backup root at mount time");
740 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
742 case Opt_skip_balance:
743 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
745 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
746 case Opt_check_integrity_including_extent_data:
748 "enabling check integrity including extent data");
749 btrfs_set_opt(info->mount_opt,
750 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
751 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
753 case Opt_check_integrity:
754 btrfs_info(info, "enabling check integrity");
755 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
757 case Opt_check_integrity_print_mask:
758 ret = match_int(&args[0], &intarg);
761 info->check_integrity_print_mask = intarg;
762 btrfs_info(info, "check_integrity_print_mask 0x%x",
763 info->check_integrity_print_mask);
766 case Opt_check_integrity_including_extent_data:
767 case Opt_check_integrity:
768 case Opt_check_integrity_print_mask:
770 "support for check_integrity* not compiled in!");
774 case Opt_fatal_errors:
775 if (strcmp(args[0].from, "panic") == 0)
776 btrfs_set_opt(info->mount_opt,
777 PANIC_ON_FATAL_ERROR);
778 else if (strcmp(args[0].from, "bug") == 0)
779 btrfs_clear_opt(info->mount_opt,
780 PANIC_ON_FATAL_ERROR);
786 case Opt_commit_interval:
788 ret = match_int(&args[0], &intarg);
790 btrfs_err(info, "invalid commit interval");
797 "excessive commit interval %d",
800 info->commit_interval = intarg;
803 "using default commit interval %ds",
804 BTRFS_DEFAULT_COMMIT_INTERVAL);
805 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
808 #ifdef CONFIG_BTRFS_DEBUG
809 case Opt_fragment_all:
810 btrfs_info(info, "fragmenting all space");
811 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
812 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
814 case Opt_fragment_metadata:
815 btrfs_info(info, "fragmenting metadata");
816 btrfs_set_opt(info->mount_opt,
819 case Opt_fragment_data:
820 btrfs_info(info, "fragmenting data");
821 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
824 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
826 btrfs_info(info, "doing ref verification");
827 btrfs_set_opt(info->mount_opt, REF_VERIFY);
831 btrfs_info(info, "unrecognized mount option '%s'", p);
840 * Extra check for current option against current flag
842 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
844 "nologreplay must be used with ro mount option");
848 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
849 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
850 !btrfs_test_opt(info, CLEAR_CACHE)) {
851 btrfs_err(info, "cannot disable free space tree");
855 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
856 btrfs_info(info, "disk space caching is enabled");
857 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
858 btrfs_info(info, "using free space tree");
863 * Parse mount options that are required early in the mount process.
865 * All other options will be parsed on much later in the mount process and
866 * only when we need to allocate a new super block.
868 static int btrfs_parse_early_options(const char *options, fmode_t flags,
869 void *holder, struct btrfs_fs_devices **fs_devices)
871 substring_t args[MAX_OPT_ARGS];
872 char *device_name, *opts, *orig, *p;
879 * strsep changes the string, duplicate it because btrfs_parse_options
882 opts = kstrdup(options, GFP_KERNEL);
887 while ((p = strsep(&opts, ",")) != NULL) {
893 token = match_token(p, tokens, args);
894 if (token == Opt_device) {
895 device_name = match_strdup(&args[0]);
900 error = btrfs_scan_one_device(device_name,
901 flags, holder, fs_devices);
914 * Parse mount options that are related to subvolume id
916 * The value is later passed to mount_subvol()
918 static int btrfs_parse_subvol_options(const char *options, fmode_t flags,
919 char **subvol_name, u64 *subvol_objectid)
921 substring_t args[MAX_OPT_ARGS];
922 char *opts, *orig, *p;
930 * strsep changes the string, duplicate it because
931 * btrfs_parse_early_options gets called later
933 opts = kstrdup(options, GFP_KERNEL);
938 while ((p = strsep(&opts, ",")) != NULL) {
943 token = match_token(p, tokens, args);
947 *subvol_name = match_strdup(&args[0]);
954 num = match_strdup(&args[0]);
956 *subvol_objectid = memparse(num, NULL);
958 /* we want the original fs_tree */
959 if (!*subvol_objectid)
961 BTRFS_FS_TREE_OBJECTID;
967 case Opt_subvolrootid:
968 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
980 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
983 struct btrfs_root *root = fs_info->tree_root;
984 struct btrfs_root *fs_root;
985 struct btrfs_root_ref *root_ref;
986 struct btrfs_inode_ref *inode_ref;
987 struct btrfs_key key;
988 struct btrfs_path *path = NULL;
989 char *name = NULL, *ptr;
994 path = btrfs_alloc_path();
999 path->leave_spinning = 1;
1001 name = kmalloc(PATH_MAX, GFP_KERNEL);
1006 ptr = name + PATH_MAX - 1;
1010 * Walk up the subvolume trees in the tree of tree roots by root
1011 * backrefs until we hit the top-level subvolume.
1013 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1014 key.objectid = subvol_objectid;
1015 key.type = BTRFS_ROOT_BACKREF_KEY;
1016 key.offset = (u64)-1;
1018 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1021 } else if (ret > 0) {
1022 ret = btrfs_previous_item(root, path, subvol_objectid,
1023 BTRFS_ROOT_BACKREF_KEY);
1026 } else if (ret > 0) {
1032 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1033 subvol_objectid = key.offset;
1035 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1036 struct btrfs_root_ref);
1037 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1040 ret = -ENAMETOOLONG;
1043 read_extent_buffer(path->nodes[0], ptr + 1,
1044 (unsigned long)(root_ref + 1), len);
1046 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1047 btrfs_release_path(path);
1049 key.objectid = subvol_objectid;
1050 key.type = BTRFS_ROOT_ITEM_KEY;
1051 key.offset = (u64)-1;
1052 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1053 if (IS_ERR(fs_root)) {
1054 ret = PTR_ERR(fs_root);
1059 * Walk up the filesystem tree by inode refs until we hit the
1062 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1063 key.objectid = dirid;
1064 key.type = BTRFS_INODE_REF_KEY;
1065 key.offset = (u64)-1;
1067 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1070 } else if (ret > 0) {
1071 ret = btrfs_previous_item(fs_root, path, dirid,
1072 BTRFS_INODE_REF_KEY);
1075 } else if (ret > 0) {
1081 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1084 inode_ref = btrfs_item_ptr(path->nodes[0],
1086 struct btrfs_inode_ref);
1087 len = btrfs_inode_ref_name_len(path->nodes[0],
1091 ret = -ENAMETOOLONG;
1094 read_extent_buffer(path->nodes[0], ptr + 1,
1095 (unsigned long)(inode_ref + 1), len);
1097 btrfs_release_path(path);
1101 btrfs_free_path(path);
1102 if (ptr == name + PATH_MAX - 1) {
1106 memmove(name, ptr, name + PATH_MAX - ptr);
1111 btrfs_free_path(path);
1113 return ERR_PTR(ret);
1116 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1118 struct btrfs_root *root = fs_info->tree_root;
1119 struct btrfs_dir_item *di;
1120 struct btrfs_path *path;
1121 struct btrfs_key location;
1124 path = btrfs_alloc_path();
1127 path->leave_spinning = 1;
1130 * Find the "default" dir item which points to the root item that we
1131 * will mount by default if we haven't been given a specific subvolume
1134 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1135 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1137 btrfs_free_path(path);
1142 * Ok the default dir item isn't there. This is weird since
1143 * it's always been there, but don't freak out, just try and
1144 * mount the top-level subvolume.
1146 btrfs_free_path(path);
1147 *objectid = BTRFS_FS_TREE_OBJECTID;
1151 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1152 btrfs_free_path(path);
1153 *objectid = location.objectid;
1157 static int btrfs_fill_super(struct super_block *sb,
1158 struct btrfs_fs_devices *fs_devices,
1161 struct inode *inode;
1162 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1163 struct btrfs_key key;
1166 sb->s_maxbytes = MAX_LFS_FILESIZE;
1167 sb->s_magic = BTRFS_SUPER_MAGIC;
1168 sb->s_op = &btrfs_super_ops;
1169 sb->s_d_op = &btrfs_dentry_operations;
1170 sb->s_export_op = &btrfs_export_ops;
1171 sb->s_xattr = btrfs_xattr_handlers;
1172 sb->s_time_gran = 1;
1173 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1174 sb->s_flags |= SB_POSIXACL;
1176 sb->s_flags |= SB_I_VERSION;
1177 sb->s_iflags |= SB_I_CGROUPWB;
1179 err = super_setup_bdi(sb);
1181 btrfs_err(fs_info, "super_setup_bdi failed");
1185 err = open_ctree(sb, fs_devices, (char *)data);
1187 btrfs_err(fs_info, "open_ctree failed");
1191 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1192 key.type = BTRFS_INODE_ITEM_KEY;
1194 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1195 if (IS_ERR(inode)) {
1196 err = PTR_ERR(inode);
1200 sb->s_root = d_make_root(inode);
1206 cleancache_init_fs(sb);
1207 sb->s_flags |= SB_ACTIVE;
1211 close_ctree(fs_info);
1215 int btrfs_sync_fs(struct super_block *sb, int wait)
1217 struct btrfs_trans_handle *trans;
1218 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1219 struct btrfs_root *root = fs_info->tree_root;
1221 trace_btrfs_sync_fs(fs_info, wait);
1224 filemap_flush(fs_info->btree_inode->i_mapping);
1228 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1230 trans = btrfs_attach_transaction_barrier(root);
1231 if (IS_ERR(trans)) {
1232 /* no transaction, don't bother */
1233 if (PTR_ERR(trans) == -ENOENT) {
1235 * Exit unless we have some pending changes
1236 * that need to go through commit
1238 if (fs_info->pending_changes == 0)
1241 * A non-blocking test if the fs is frozen. We must not
1242 * start a new transaction here otherwise a deadlock
1243 * happens. The pending operations are delayed to the
1244 * next commit after thawing.
1246 if (sb_start_write_trylock(sb))
1250 trans = btrfs_start_transaction(root, 0);
1253 return PTR_ERR(trans);
1255 return btrfs_commit_transaction(trans);
1258 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1260 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1261 const char *compress_type;
1263 if (btrfs_test_opt(info, DEGRADED))
1264 seq_puts(seq, ",degraded");
1265 if (btrfs_test_opt(info, NODATASUM))
1266 seq_puts(seq, ",nodatasum");
1267 if (btrfs_test_opt(info, NODATACOW))
1268 seq_puts(seq, ",nodatacow");
1269 if (btrfs_test_opt(info, NOBARRIER))
1270 seq_puts(seq, ",nobarrier");
1271 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1272 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1273 if (info->thread_pool_size != min_t(unsigned long,
1274 num_online_cpus() + 2, 8))
1275 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1276 if (btrfs_test_opt(info, COMPRESS)) {
1277 compress_type = btrfs_compress_type2str(info->compress_type);
1278 if (btrfs_test_opt(info, FORCE_COMPRESS))
1279 seq_printf(seq, ",compress-force=%s", compress_type);
1281 seq_printf(seq, ",compress=%s", compress_type);
1282 if (info->compress_level)
1283 seq_printf(seq, ":%d", info->compress_level);
1285 if (btrfs_test_opt(info, NOSSD))
1286 seq_puts(seq, ",nossd");
1287 if (btrfs_test_opt(info, SSD_SPREAD))
1288 seq_puts(seq, ",ssd_spread");
1289 else if (btrfs_test_opt(info, SSD))
1290 seq_puts(seq, ",ssd");
1291 if (btrfs_test_opt(info, NOTREELOG))
1292 seq_puts(seq, ",notreelog");
1293 if (btrfs_test_opt(info, NOLOGREPLAY))
1294 seq_puts(seq, ",nologreplay");
1295 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1296 seq_puts(seq, ",flushoncommit");
1297 if (btrfs_test_opt(info, DISCARD))
1298 seq_puts(seq, ",discard");
1299 if (!(info->sb->s_flags & SB_POSIXACL))
1300 seq_puts(seq, ",noacl");
1301 if (btrfs_test_opt(info, SPACE_CACHE))
1302 seq_puts(seq, ",space_cache");
1303 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1304 seq_puts(seq, ",space_cache=v2");
1306 seq_puts(seq, ",nospace_cache");
1307 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1308 seq_puts(seq, ",rescan_uuid_tree");
1309 if (btrfs_test_opt(info, CLEAR_CACHE))
1310 seq_puts(seq, ",clear_cache");
1311 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1312 seq_puts(seq, ",user_subvol_rm_allowed");
1313 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1314 seq_puts(seq, ",enospc_debug");
1315 if (btrfs_test_opt(info, AUTO_DEFRAG))
1316 seq_puts(seq, ",autodefrag");
1317 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1318 seq_puts(seq, ",inode_cache");
1319 if (btrfs_test_opt(info, SKIP_BALANCE))
1320 seq_puts(seq, ",skip_balance");
1321 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1322 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1323 seq_puts(seq, ",check_int_data");
1324 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1325 seq_puts(seq, ",check_int");
1326 if (info->check_integrity_print_mask)
1327 seq_printf(seq, ",check_int_print_mask=%d",
1328 info->check_integrity_print_mask);
1330 if (info->metadata_ratio)
1331 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1332 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1333 seq_puts(seq, ",fatal_errors=panic");
1334 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1335 seq_printf(seq, ",commit=%d", info->commit_interval);
1336 #ifdef CONFIG_BTRFS_DEBUG
1337 if (btrfs_test_opt(info, FRAGMENT_DATA))
1338 seq_puts(seq, ",fragment=data");
1339 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1340 seq_puts(seq, ",fragment=metadata");
1342 if (btrfs_test_opt(info, REF_VERIFY))
1343 seq_puts(seq, ",ref_verify");
1344 seq_printf(seq, ",subvolid=%llu",
1345 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1346 seq_puts(seq, ",subvol=");
1347 seq_dentry(seq, dentry, " \t\n\\");
1351 static int btrfs_test_super(struct super_block *s, void *data)
1353 struct btrfs_fs_info *p = data;
1354 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1356 return fs_info->fs_devices == p->fs_devices;
1359 static int btrfs_set_super(struct super_block *s, void *data)
1361 int err = set_anon_super(s, data);
1363 s->s_fs_info = data;
1368 * subvolumes are identified by ino 256
1370 static inline int is_subvolume_inode(struct inode *inode)
1372 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1377 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1378 const char *device_name, struct vfsmount *mnt)
1380 struct dentry *root;
1384 if (!subvol_objectid) {
1385 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1388 root = ERR_PTR(ret);
1392 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1394 if (IS_ERR(subvol_name)) {
1395 root = ERR_CAST(subvol_name);
1402 root = mount_subtree(mnt, subvol_name);
1403 /* mount_subtree() drops our reference on the vfsmount. */
1406 if (!IS_ERR(root)) {
1407 struct super_block *s = root->d_sb;
1408 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1409 struct inode *root_inode = d_inode(root);
1410 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1413 if (!is_subvolume_inode(root_inode)) {
1414 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1418 if (subvol_objectid && root_objectid != subvol_objectid) {
1420 * This will also catch a race condition where a
1421 * subvolume which was passed by ID is renamed and
1422 * another subvolume is renamed over the old location.
1425 "subvol '%s' does not match subvolid %llu",
1426 subvol_name, subvol_objectid);
1431 root = ERR_PTR(ret);
1432 deactivate_locked_super(s);
1442 static int parse_security_options(char *orig_opts,
1443 struct security_mnt_opts *sec_opts)
1445 char *secdata = NULL;
1448 secdata = alloc_secdata();
1451 ret = security_sb_copy_data(orig_opts, secdata);
1453 free_secdata(secdata);
1456 ret = security_sb_parse_opts_str(secdata, sec_opts);
1457 free_secdata(secdata);
1461 static int setup_security_options(struct btrfs_fs_info *fs_info,
1462 struct super_block *sb,
1463 struct security_mnt_opts *sec_opts)
1468 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1471 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1475 #ifdef CONFIG_SECURITY
1476 if (!fs_info->security_opts.num_mnt_opts) {
1477 /* first time security setup, copy sec_opts to fs_info */
1478 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1481 * Since SELinux (the only one supporting security_mnt_opts)
1482 * does NOT support changing context during remount/mount of
1483 * the same sb, this must be the same or part of the same
1484 * security options, just free it.
1486 security_free_mnt_opts(sec_opts);
1493 * Find a superblock for the given device / mount point.
1495 * Note: This is based on mount_bdev from fs/super.c with a few additions
1496 * for multiple device setup. Make sure to keep it in sync.
1498 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1499 int flags, const char *device_name, void *data)
1501 struct block_device *bdev = NULL;
1502 struct super_block *s;
1503 struct btrfs_fs_devices *fs_devices = NULL;
1504 struct btrfs_fs_info *fs_info = NULL;
1505 struct security_mnt_opts new_sec_opts;
1506 fmode_t mode = FMODE_READ;
1509 if (!(flags & SB_RDONLY))
1510 mode |= FMODE_WRITE;
1512 error = btrfs_parse_early_options(data, mode, fs_type,
1515 return ERR_PTR(error);
1518 security_init_mnt_opts(&new_sec_opts);
1520 error = parse_security_options(data, &new_sec_opts);
1522 return ERR_PTR(error);
1525 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1527 goto error_sec_opts;
1530 * Setup a dummy root and fs_info for test/set super. This is because
1531 * we don't actually fill this stuff out until open_ctree, but we need
1532 * it for searching for existing supers, so this lets us do that and
1533 * then open_ctree will properly initialize everything later.
1535 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1538 goto error_sec_opts;
1541 fs_info->fs_devices = fs_devices;
1543 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1544 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1545 security_init_mnt_opts(&fs_info->security_opts);
1546 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1551 error = btrfs_open_devices(fs_devices, mode, fs_type);
1555 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1557 goto error_close_devices;
1560 bdev = fs_devices->latest_bdev;
1561 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1565 goto error_close_devices;
1569 btrfs_close_devices(fs_devices);
1570 free_fs_info(fs_info);
1571 if ((flags ^ s->s_flags) & SB_RDONLY)
1574 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1575 btrfs_sb(s)->bdev_holder = fs_type;
1576 error = btrfs_fill_super(s, fs_devices, data);
1579 deactivate_locked_super(s);
1580 goto error_sec_opts;
1583 fs_info = btrfs_sb(s);
1584 error = setup_security_options(fs_info, s, &new_sec_opts);
1586 deactivate_locked_super(s);
1587 goto error_sec_opts;
1590 return dget(s->s_root);
1592 error_close_devices:
1593 btrfs_close_devices(fs_devices);
1595 free_fs_info(fs_info);
1597 security_free_mnt_opts(&new_sec_opts);
1598 return ERR_PTR(error);
1602 * Mount function which is called by VFS layer.
1604 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1605 * which needs vfsmount* of device's root (/). This means device's root has to
1606 * be mounted internally in any case.
1609 * 1. Parse subvol id related options for later use in mount_subvol().
1611 * 2. Mount device's root (/) by calling vfs_kern_mount().
1613 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1614 * first place. In order to avoid calling btrfs_mount() again, we use
1615 * different file_system_type which is not registered to VFS by
1616 * register_filesystem() (btrfs_root_fs_type). As a result,
1617 * btrfs_mount_root() is called. The return value will be used by
1618 * mount_subtree() in mount_subvol().
1620 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1621 * "btrfs subvolume set-default", mount_subvol() is called always.
1623 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1624 const char *device_name, void *data)
1626 struct vfsmount *mnt_root;
1627 struct dentry *root;
1628 fmode_t mode = FMODE_READ;
1629 char *subvol_name = NULL;
1630 u64 subvol_objectid = 0;
1633 if (!(flags & SB_RDONLY))
1634 mode |= FMODE_WRITE;
1636 error = btrfs_parse_subvol_options(data, mode,
1637 &subvol_name, &subvol_objectid);
1640 return ERR_PTR(error);
1643 /* mount device's root (/) */
1644 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1645 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1646 if (flags & SB_RDONLY) {
1647 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1648 flags & ~SB_RDONLY, device_name, data);
1650 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1651 flags | SB_RDONLY, device_name, data);
1652 if (IS_ERR(mnt_root)) {
1653 root = ERR_CAST(mnt_root);
1657 down_write(&mnt_root->mnt_sb->s_umount);
1658 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1659 up_write(&mnt_root->mnt_sb->s_umount);
1661 root = ERR_PTR(error);
1667 if (IS_ERR(mnt_root)) {
1668 root = ERR_CAST(mnt_root);
1672 /* mount_subvol() will free subvol_name and mnt_root */
1673 root = mount_subvol(subvol_name, subvol_objectid, device_name, mnt_root);
1679 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1680 u32 new_pool_size, u32 old_pool_size)
1682 if (new_pool_size == old_pool_size)
1685 fs_info->thread_pool_size = new_pool_size;
1687 btrfs_info(fs_info, "resize thread pool %d -> %d",
1688 old_pool_size, new_pool_size);
1690 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1691 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1692 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1693 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1694 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1695 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1696 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1698 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1699 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1700 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1701 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1702 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1706 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1708 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1711 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1712 unsigned long old_opts, int flags)
1714 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1715 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1716 (flags & SB_RDONLY))) {
1717 /* wait for any defraggers to finish */
1718 wait_event(fs_info->transaction_wait,
1719 (atomic_read(&fs_info->defrag_running) == 0));
1720 if (flags & SB_RDONLY)
1721 sync_filesystem(fs_info->sb);
1725 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1726 unsigned long old_opts)
1729 * We need to cleanup all defragable inodes if the autodefragment is
1730 * close or the filesystem is read only.
1732 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1733 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1734 btrfs_cleanup_defrag_inodes(fs_info);
1737 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1740 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1742 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1743 struct btrfs_root *root = fs_info->tree_root;
1744 unsigned old_flags = sb->s_flags;
1745 unsigned long old_opts = fs_info->mount_opt;
1746 unsigned long old_compress_type = fs_info->compress_type;
1747 u64 old_max_inline = fs_info->max_inline;
1748 u32 old_thread_pool_size = fs_info->thread_pool_size;
1749 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1752 sync_filesystem(sb);
1753 btrfs_remount_prepare(fs_info);
1756 struct security_mnt_opts new_sec_opts;
1758 security_init_mnt_opts(&new_sec_opts);
1759 ret = parse_security_options(data, &new_sec_opts);
1762 ret = setup_security_options(fs_info, sb,
1765 security_free_mnt_opts(&new_sec_opts);
1770 ret = btrfs_parse_options(fs_info, data, *flags);
1776 btrfs_remount_begin(fs_info, old_opts, *flags);
1777 btrfs_resize_thread_pool(fs_info,
1778 fs_info->thread_pool_size, old_thread_pool_size);
1780 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1783 if (*flags & SB_RDONLY) {
1785 * this also happens on 'umount -rf' or on shutdown, when
1786 * the filesystem is busy.
1788 cancel_work_sync(&fs_info->async_reclaim_work);
1790 /* wait for the uuid_scan task to finish */
1791 down(&fs_info->uuid_tree_rescan_sem);
1792 /* avoid complains from lockdep et al. */
1793 up(&fs_info->uuid_tree_rescan_sem);
1795 sb->s_flags |= SB_RDONLY;
1798 * Setting SB_RDONLY will put the cleaner thread to
1799 * sleep at the next loop if it's already active.
1800 * If it's already asleep, we'll leave unused block
1801 * groups on disk until we're mounted read-write again
1802 * unless we clean them up here.
1804 btrfs_delete_unused_bgs(fs_info);
1806 btrfs_dev_replace_suspend_for_unmount(fs_info);
1807 btrfs_scrub_cancel(fs_info);
1808 btrfs_pause_balance(fs_info);
1810 ret = btrfs_commit_super(fs_info);
1814 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1816 "Remounting read-write after error is not allowed");
1820 if (fs_info->fs_devices->rw_devices == 0) {
1825 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1827 "too many missing devices, writeable remount is not allowed");
1832 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1837 ret = btrfs_cleanup_fs_roots(fs_info);
1841 /* recover relocation */
1842 mutex_lock(&fs_info->cleaner_mutex);
1843 ret = btrfs_recover_relocation(root);
1844 mutex_unlock(&fs_info->cleaner_mutex);
1848 ret = btrfs_resume_balance_async(fs_info);
1852 ret = btrfs_resume_dev_replace_async(fs_info);
1854 btrfs_warn(fs_info, "failed to resume dev_replace");
1858 btrfs_qgroup_rescan_resume(fs_info);
1860 if (!fs_info->uuid_root) {
1861 btrfs_info(fs_info, "creating UUID tree");
1862 ret = btrfs_create_uuid_tree(fs_info);
1865 "failed to create the UUID tree %d",
1870 sb->s_flags &= ~SB_RDONLY;
1872 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1875 wake_up_process(fs_info->transaction_kthread);
1876 btrfs_remount_cleanup(fs_info, old_opts);
1880 /* We've hit an error - don't reset SB_RDONLY */
1882 old_flags |= SB_RDONLY;
1883 sb->s_flags = old_flags;
1884 fs_info->mount_opt = old_opts;
1885 fs_info->compress_type = old_compress_type;
1886 fs_info->max_inline = old_max_inline;
1887 btrfs_resize_thread_pool(fs_info,
1888 old_thread_pool_size, fs_info->thread_pool_size);
1889 fs_info->metadata_ratio = old_metadata_ratio;
1890 btrfs_remount_cleanup(fs_info, old_opts);
1894 /* Used to sort the devices by max_avail(descending sort) */
1895 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1896 const void *dev_info2)
1898 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1899 ((struct btrfs_device_info *)dev_info2)->max_avail)
1901 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1902 ((struct btrfs_device_info *)dev_info2)->max_avail)
1909 * sort the devices by max_avail, in which max free extent size of each device
1910 * is stored.(Descending Sort)
1912 static inline void btrfs_descending_sort_devices(
1913 struct btrfs_device_info *devices,
1916 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1917 btrfs_cmp_device_free_bytes, NULL);
1921 * The helper to calc the free space on the devices that can be used to store
1924 static int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1927 struct btrfs_device_info *devices_info;
1928 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1929 struct btrfs_device *device;
1933 u64 min_stripe_size;
1934 int min_stripes = 1, num_stripes = 1;
1935 int i = 0, nr_devices;
1938 * We aren't under the device list lock, so this is racy-ish, but good
1939 * enough for our purposes.
1941 nr_devices = fs_info->fs_devices->open_devices;
1944 nr_devices = fs_info->fs_devices->open_devices;
1952 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1957 /* calc min stripe number for data space allocation */
1958 type = btrfs_data_alloc_profile(fs_info);
1959 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1961 num_stripes = nr_devices;
1962 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1965 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1970 if (type & BTRFS_BLOCK_GROUP_DUP)
1971 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1973 min_stripe_size = BTRFS_STRIPE_LEN;
1976 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1977 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1978 &device->dev_state) ||
1980 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1983 if (i >= nr_devices)
1986 avail_space = device->total_bytes - device->bytes_used;
1988 /* align with stripe_len */
1989 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1990 avail_space *= BTRFS_STRIPE_LEN;
1993 * In order to avoid overwriting the superblock on the drive,
1994 * btrfs starts at an offset of at least 1MB when doing chunk
2000 * we can use the free space in [0, skip_space - 1], subtract
2001 * it from the total.
2003 if (avail_space && avail_space >= skip_space)
2004 avail_space -= skip_space;
2008 if (avail_space < min_stripe_size)
2011 devices_info[i].dev = device;
2012 devices_info[i].max_avail = avail_space;
2020 btrfs_descending_sort_devices(devices_info, nr_devices);
2024 while (nr_devices >= min_stripes) {
2025 if (num_stripes > nr_devices)
2026 num_stripes = nr_devices;
2028 if (devices_info[i].max_avail >= min_stripe_size) {
2032 avail_space += devices_info[i].max_avail * num_stripes;
2033 alloc_size = devices_info[i].max_avail;
2034 for (j = i + 1 - num_stripes; j <= i; j++)
2035 devices_info[j].max_avail -= alloc_size;
2041 kfree(devices_info);
2042 *free_bytes = avail_space;
2047 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2049 * If there's a redundant raid level at DATA block groups, use the respective
2050 * multiplier to scale the sizes.
2052 * Unused device space usage is based on simulating the chunk allocator
2053 * algorithm that respects the device sizes and order of allocations. This is
2054 * a close approximation of the actual use but there are other factors that may
2055 * change the result (like a new metadata chunk).
2057 * If metadata is exhausted, f_bavail will be 0.
2059 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2061 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2062 struct btrfs_super_block *disk_super = fs_info->super_copy;
2063 struct list_head *head = &fs_info->space_info;
2064 struct btrfs_space_info *found;
2066 u64 total_free_data = 0;
2067 u64 total_free_meta = 0;
2068 int bits = dentry->d_sb->s_blocksize_bits;
2069 __be32 *fsid = (__be32 *)fs_info->fsid;
2070 unsigned factor = 1;
2071 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2077 list_for_each_entry_rcu(found, head, list) {
2078 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2081 total_free_data += found->disk_total - found->disk_used;
2083 btrfs_account_ro_block_groups_free_space(found);
2085 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2086 if (!list_empty(&found->block_groups[i])) {
2088 case BTRFS_RAID_DUP:
2089 case BTRFS_RAID_RAID1:
2090 case BTRFS_RAID_RAID10:
2098 * Metadata in mixed block goup profiles are accounted in data
2100 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2101 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2104 total_free_meta += found->disk_total -
2108 total_used += found->disk_used;
2113 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2114 buf->f_blocks >>= bits;
2115 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2117 /* Account global block reserve as used, it's in logical size already */
2118 spin_lock(&block_rsv->lock);
2119 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2120 if (buf->f_bfree >= block_rsv->size >> bits)
2121 buf->f_bfree -= block_rsv->size >> bits;
2124 spin_unlock(&block_rsv->lock);
2126 buf->f_bavail = div_u64(total_free_data, factor);
2127 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2130 buf->f_bavail += div_u64(total_free_data, factor);
2131 buf->f_bavail = buf->f_bavail >> bits;
2134 * We calculate the remaining metadata space minus global reserve. If
2135 * this is (supposedly) smaller than zero, there's no space. But this
2136 * does not hold in practice, the exhausted state happens where's still
2137 * some positive delta. So we apply some guesswork and compare the
2138 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2140 * We probably cannot calculate the exact threshold value because this
2141 * depends on the internal reservations requested by various
2142 * operations, so some operations that consume a few metadata will
2143 * succeed even if the Avail is zero. But this is better than the other
2148 if (!mixed && total_free_meta - thresh < block_rsv->size)
2151 buf->f_type = BTRFS_SUPER_MAGIC;
2152 buf->f_bsize = dentry->d_sb->s_blocksize;
2153 buf->f_namelen = BTRFS_NAME_LEN;
2155 /* We treat it as constant endianness (it doesn't matter _which_)
2156 because we want the fsid to come out the same whether mounted
2157 on a big-endian or little-endian host */
2158 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2159 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2160 /* Mask in the root object ID too, to disambiguate subvols */
2161 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2162 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2167 static void btrfs_kill_super(struct super_block *sb)
2169 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2170 kill_anon_super(sb);
2171 free_fs_info(fs_info);
2174 static struct file_system_type btrfs_fs_type = {
2175 .owner = THIS_MODULE,
2177 .mount = btrfs_mount,
2178 .kill_sb = btrfs_kill_super,
2179 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2182 static struct file_system_type btrfs_root_fs_type = {
2183 .owner = THIS_MODULE,
2185 .mount = btrfs_mount_root,
2186 .kill_sb = btrfs_kill_super,
2187 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2190 MODULE_ALIAS_FS("btrfs");
2192 static int btrfs_control_open(struct inode *inode, struct file *file)
2195 * The control file's private_data is used to hold the
2196 * transaction when it is started and is used to keep
2197 * track of whether a transaction is already in progress.
2199 file->private_data = NULL;
2204 * used by btrfsctl to scan devices when no FS is mounted
2206 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2209 struct btrfs_ioctl_vol_args *vol;
2210 struct btrfs_fs_devices *fs_devices;
2213 if (!capable(CAP_SYS_ADMIN))
2216 vol = memdup_user((void __user *)arg, sizeof(*vol));
2218 return PTR_ERR(vol);
2221 case BTRFS_IOC_SCAN_DEV:
2222 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2223 &btrfs_root_fs_type, &fs_devices);
2225 case BTRFS_IOC_DEVICES_READY:
2226 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2227 &btrfs_root_fs_type, &fs_devices);
2230 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2232 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2233 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2241 static int btrfs_freeze(struct super_block *sb)
2243 struct btrfs_trans_handle *trans;
2244 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2245 struct btrfs_root *root = fs_info->tree_root;
2247 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2249 * We don't need a barrier here, we'll wait for any transaction that
2250 * could be in progress on other threads (and do delayed iputs that
2251 * we want to avoid on a frozen filesystem), or do the commit
2254 trans = btrfs_attach_transaction_barrier(root);
2255 if (IS_ERR(trans)) {
2256 /* no transaction, don't bother */
2257 if (PTR_ERR(trans) == -ENOENT)
2259 return PTR_ERR(trans);
2261 return btrfs_commit_transaction(trans);
2264 static int btrfs_unfreeze(struct super_block *sb)
2266 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2268 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2272 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2274 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2275 struct btrfs_fs_devices *cur_devices;
2276 struct btrfs_device *dev, *first_dev = NULL;
2277 struct list_head *head;
2278 struct rcu_string *name;
2280 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2281 cur_devices = fs_info->fs_devices;
2282 while (cur_devices) {
2283 head = &cur_devices->devices;
2284 list_for_each_entry(dev, head, dev_list) {
2285 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2289 if (!first_dev || dev->devid < first_dev->devid)
2292 cur_devices = cur_devices->seed;
2297 name = rcu_dereference(first_dev->name);
2298 seq_escape(m, name->str, " \t\n\\");
2303 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2307 static const struct super_operations btrfs_super_ops = {
2308 .drop_inode = btrfs_drop_inode,
2309 .evict_inode = btrfs_evict_inode,
2310 .put_super = btrfs_put_super,
2311 .sync_fs = btrfs_sync_fs,
2312 .show_options = btrfs_show_options,
2313 .show_devname = btrfs_show_devname,
2314 .write_inode = btrfs_write_inode,
2315 .alloc_inode = btrfs_alloc_inode,
2316 .destroy_inode = btrfs_destroy_inode,
2317 .statfs = btrfs_statfs,
2318 .remount_fs = btrfs_remount,
2319 .freeze_fs = btrfs_freeze,
2320 .unfreeze_fs = btrfs_unfreeze,
2323 static const struct file_operations btrfs_ctl_fops = {
2324 .open = btrfs_control_open,
2325 .unlocked_ioctl = btrfs_control_ioctl,
2326 .compat_ioctl = btrfs_control_ioctl,
2327 .owner = THIS_MODULE,
2328 .llseek = noop_llseek,
2331 static struct miscdevice btrfs_misc = {
2332 .minor = BTRFS_MINOR,
2333 .name = "btrfs-control",
2334 .fops = &btrfs_ctl_fops
2337 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2338 MODULE_ALIAS("devname:btrfs-control");
2340 static int __init btrfs_interface_init(void)
2342 return misc_register(&btrfs_misc);
2345 static void btrfs_interface_exit(void)
2347 misc_deregister(&btrfs_misc);
2350 static void __init btrfs_print_mod_info(void)
2352 pr_info("Btrfs loaded, crc32c=%s"
2353 #ifdef CONFIG_BTRFS_DEBUG
2356 #ifdef CONFIG_BTRFS_ASSERT
2359 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2360 ", integrity-checker=on"
2362 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2366 btrfs_crc32c_impl());
2369 static int __init init_btrfs_fs(void)
2373 err = btrfs_hash_init();
2379 err = btrfs_init_sysfs();
2383 btrfs_init_compress();
2385 err = btrfs_init_cachep();
2389 err = extent_io_init();
2393 err = extent_map_init();
2395 goto free_extent_io;
2397 err = ordered_data_init();
2399 goto free_extent_map;
2401 err = btrfs_delayed_inode_init();
2403 goto free_ordered_data;
2405 err = btrfs_auto_defrag_init();
2407 goto free_delayed_inode;
2409 err = btrfs_delayed_ref_init();
2411 goto free_auto_defrag;
2413 err = btrfs_prelim_ref_init();
2415 goto free_delayed_ref;
2417 err = btrfs_end_io_wq_init();
2419 goto free_prelim_ref;
2421 err = btrfs_interface_init();
2423 goto free_end_io_wq;
2425 btrfs_init_lockdep();
2427 btrfs_print_mod_info();
2429 err = btrfs_run_sanity_tests();
2431 goto unregister_ioctl;
2433 err = register_filesystem(&btrfs_fs_type);
2435 goto unregister_ioctl;
2440 btrfs_interface_exit();
2442 btrfs_end_io_wq_exit();
2444 btrfs_prelim_ref_exit();
2446 btrfs_delayed_ref_exit();
2448 btrfs_auto_defrag_exit();
2450 btrfs_delayed_inode_exit();
2452 ordered_data_exit();
2458 btrfs_destroy_cachep();
2460 btrfs_exit_compress();
2467 static void __exit exit_btrfs_fs(void)
2469 btrfs_destroy_cachep();
2470 btrfs_delayed_ref_exit();
2471 btrfs_auto_defrag_exit();
2472 btrfs_delayed_inode_exit();
2473 btrfs_prelim_ref_exit();
2474 ordered_data_exit();
2477 btrfs_interface_exit();
2478 btrfs_end_io_wq_exit();
2479 unregister_filesystem(&btrfs_fs_type);
2481 btrfs_cleanup_fs_uuids();
2482 btrfs_exit_compress();
2486 late_initcall(init_btrfs_fs);
2487 module_exit(exit_btrfs_fs)
2489 MODULE_LICENSE("GPL");