1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2007 Oracle. All rights reserved.
6 #include <linux/blkdev.h>
7 #include <linux/module.h>
9 #include <linux/pagemap.h>
10 #include <linux/highmem.h>
11 #include <linux/time.h>
12 #include <linux/init.h>
13 #include <linux/seq_file.h>
14 #include <linux/string.h>
15 #include <linux/backing-dev.h>
16 #include <linux/mount.h>
17 #include <linux/writeback.h>
18 #include <linux/statfs.h>
19 #include <linux/compat.h>
20 #include <linux/parser.h>
21 #include <linux/ctype.h>
22 #include <linux/namei.h>
23 #include <linux/miscdevice.h>
24 #include <linux/magic.h>
25 #include <linux/slab.h>
26 #include <linux/cleancache.h>
27 #include <linux/ratelimit.h>
28 #include <linux/crc32c.h>
29 #include <linux/btrfs.h>
30 #include "delayed-inode.h"
33 #include "transaction.h"
34 #include "btrfs_inode.h"
35 #include "print-tree.h"
40 #include "compression.h"
41 #include "rcu-string.h"
42 #include "dev-replace.h"
43 #include "free-space-cache.h"
45 #include "space-info.h"
48 #include "tests/btrfs-tests.h"
49 #include "block-group.h"
52 #define CREATE_TRACE_POINTS
53 #include <trace/events/btrfs.h>
55 static const struct super_operations btrfs_super_ops;
58 * Types for mounting the default subvolume and a subvolume explicitly
59 * requested by subvol=/path. That way the callchain is straightforward and we
60 * don't have to play tricks with the mount options and recursive calls to
63 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
65 static struct file_system_type btrfs_fs_type;
66 static struct file_system_type btrfs_root_fs_type;
68 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
71 * Generally the error codes correspond to their respective errors, but there
72 * are a few special cases.
74 * EUCLEAN: Any sort of corruption that we encounter. The tree-checker for
75 * instance will return EUCLEAN if any of the blocks are corrupted in
76 * a way that is problematic. We want to reserve EUCLEAN for these
77 * sort of corruptions.
79 * EROFS: If we check BTRFS_FS_STATE_ERROR and fail out with a return error, we
80 * need to use EROFS for this case. We will have no idea of the
81 * original failure, that will have been reported at the time we tripped
82 * over the error. Each subsequent error that doesn't have any context
83 * of the original error should use EROFS when handling BTRFS_FS_STATE_ERROR.
85 const char * __attribute_const__ btrfs_decode_error(int errno)
87 char *errstr = "unknown";
90 case -ENOENT: /* -2 */
91 errstr = "No such entry";
94 errstr = "IO failure";
96 case -ENOMEM: /* -12*/
97 errstr = "Out of memory";
99 case -EEXIST: /* -17 */
100 errstr = "Object already exists";
102 case -ENOSPC: /* -28 */
103 errstr = "No space left";
105 case -EROFS: /* -30 */
106 errstr = "Readonly filesystem";
108 case -EOPNOTSUPP: /* -95 */
109 errstr = "Operation not supported";
111 case -EUCLEAN: /* -117 */
112 errstr = "Filesystem corrupted";
114 case -EDQUOT: /* -122 */
115 errstr = "Quota exceeded";
123 * __btrfs_handle_fs_error decodes expected errors from the caller and
124 * invokes the appropriate error response.
127 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
128 unsigned int line, int errno, const char *fmt, ...)
130 struct super_block *sb = fs_info->sb;
136 * Special case: if the error is EROFS, and we're already
137 * under SB_RDONLY, then it is safe here.
139 if (errno == -EROFS && sb_rdonly(sb))
143 errstr = btrfs_decode_error(errno);
145 struct va_format vaf;
152 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
153 sb->s_id, function, line, errno, errstr, &vaf);
156 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
157 sb->s_id, function, line, errno, errstr);
162 * Today we only save the error info to memory. Long term we'll
163 * also send it down to the disk
165 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
167 /* Don't go through full error handling during mount */
168 if (!(sb->s_flags & SB_BORN))
174 btrfs_discard_stop(fs_info);
176 /* btrfs handle error by forcing the filesystem readonly */
177 btrfs_set_sb_rdonly(sb);
178 btrfs_info(fs_info, "forced readonly");
180 * Note that a running device replace operation is not canceled here
181 * although there is no way to update the progress. It would add the
182 * risk of a deadlock, therefore the canceling is omitted. The only
183 * penalty is that some I/O remains active until the procedure
184 * completes. The next time when the filesystem is mounted writable
185 * again, the device replace operation continues.
190 static const char * const logtypes[] = {
203 * Use one ratelimit state per log level so that a flood of less important
204 * messages doesn't cause more important ones to be dropped.
206 static struct ratelimit_state printk_limits[] = {
207 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
208 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
209 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
210 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
211 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
212 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
213 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
214 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
217 void __cold btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
219 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
220 struct va_format vaf;
223 const char *type = logtypes[4];
224 struct ratelimit_state *ratelimit = &printk_limits[4];
228 while ((kern_level = printk_get_level(fmt)) != 0) {
229 size_t size = printk_skip_level(fmt) - fmt;
231 if (kern_level >= '0' && kern_level <= '7') {
232 memcpy(lvl, fmt, size);
234 type = logtypes[kern_level - '0'];
235 ratelimit = &printk_limits[kern_level - '0'];
243 if (__ratelimit(ratelimit)) {
245 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
246 fs_info->sb->s_id, &vaf);
248 printk("%sBTRFS %s: %pV\n", lvl, type, &vaf);
255 #if BITS_PER_LONG == 32
256 void __cold btrfs_warn_32bit_limit(struct btrfs_fs_info *fs_info)
258 if (!test_and_set_bit(BTRFS_FS_32BIT_WARN, &fs_info->flags)) {
259 btrfs_warn(fs_info, "reaching 32bit limit for logical addresses");
261 "due to page cache limit on 32bit systems, btrfs can't access metadata at or beyond %lluT",
262 BTRFS_32BIT_MAX_FILE_SIZE >> 40);
264 "please consider upgrading to 64bit kernel/hardware");
268 void __cold btrfs_err_32bit_limit(struct btrfs_fs_info *fs_info)
270 if (!test_and_set_bit(BTRFS_FS_32BIT_ERROR, &fs_info->flags)) {
271 btrfs_err(fs_info, "reached 32bit limit for logical addresses");
273 "due to page cache limit on 32bit systems, metadata beyond %lluT can't be accessed",
274 BTRFS_32BIT_MAX_FILE_SIZE >> 40);
276 "please consider upgrading to 64bit kernel/hardware");
282 * We only mark the transaction aborted and then set the file system read-only.
283 * This will prevent new transactions from starting or trying to join this
286 * This means that error recovery at the call site is limited to freeing
287 * any local memory allocations and passing the error code up without
288 * further cleanup. The transaction should complete as it normally would
289 * in the call path but will return -EIO.
291 * We'll complete the cleanup in btrfs_end_transaction and
292 * btrfs_commit_transaction.
295 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
296 const char *function,
297 unsigned int line, int errno)
299 struct btrfs_fs_info *fs_info = trans->fs_info;
301 WRITE_ONCE(trans->aborted, errno);
302 WRITE_ONCE(trans->transaction->aborted, errno);
303 /* Wake up anybody who may be waiting on this transaction */
304 wake_up(&fs_info->transaction_wait);
305 wake_up(&fs_info->transaction_blocked_wait);
306 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
309 * __btrfs_panic decodes unexpected, fatal errors from the caller,
310 * issues an alert, and either panics or BUGs, depending on mount options.
313 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
314 unsigned int line, int errno, const char *fmt, ...)
316 char *s_id = "<unknown>";
318 struct va_format vaf = { .fmt = fmt };
322 s_id = fs_info->sb->s_id;
327 errstr = btrfs_decode_error(errno);
328 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
329 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
330 s_id, function, line, &vaf, errno, errstr);
332 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
333 function, line, &vaf, errno, errstr);
335 /* Caller calls BUG() */
338 static void btrfs_put_super(struct super_block *sb)
340 close_ctree(btrfs_sb(sb));
349 Opt_compress_force_type,
354 Opt_flushoncommit, Opt_noflushoncommit,
356 Opt_barrier, Opt_nobarrier,
357 Opt_datacow, Opt_nodatacow,
358 Opt_datasum, Opt_nodatasum,
359 Opt_defrag, Opt_nodefrag,
360 Opt_discard, Opt_nodiscard,
364 Opt_rescan_uuid_tree,
366 Opt_space_cache, Opt_no_space_cache,
367 Opt_space_cache_version,
369 Opt_ssd_spread, Opt_nossd_spread,
374 Opt_treelog, Opt_notreelog,
375 Opt_user_subvol_rm_allowed,
385 /* Deprecated options */
387 Opt_inode_cache, Opt_noinode_cache,
389 /* Debugging options */
391 Opt_check_integrity_including_extent_data,
392 Opt_check_integrity_print_mask,
393 Opt_enospc_debug, Opt_noenospc_debug,
394 #ifdef CONFIG_BTRFS_DEBUG
395 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
397 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
403 static const match_table_t tokens = {
405 {Opt_noacl, "noacl"},
406 {Opt_clear_cache, "clear_cache"},
407 {Opt_commit_interval, "commit=%u"},
408 {Opt_compress, "compress"},
409 {Opt_compress_type, "compress=%s"},
410 {Opt_compress_force, "compress-force"},
411 {Opt_compress_force_type, "compress-force=%s"},
412 {Opt_degraded, "degraded"},
413 {Opt_device, "device=%s"},
414 {Opt_fatal_errors, "fatal_errors=%s"},
415 {Opt_flushoncommit, "flushoncommit"},
416 {Opt_noflushoncommit, "noflushoncommit"},
417 {Opt_inode_cache, "inode_cache"},
418 {Opt_noinode_cache, "noinode_cache"},
419 {Opt_max_inline, "max_inline=%s"},
420 {Opt_barrier, "barrier"},
421 {Opt_nobarrier, "nobarrier"},
422 {Opt_datacow, "datacow"},
423 {Opt_nodatacow, "nodatacow"},
424 {Opt_datasum, "datasum"},
425 {Opt_nodatasum, "nodatasum"},
426 {Opt_defrag, "autodefrag"},
427 {Opt_nodefrag, "noautodefrag"},
428 {Opt_discard, "discard"},
429 {Opt_discard_mode, "discard=%s"},
430 {Opt_nodiscard, "nodiscard"},
431 {Opt_norecovery, "norecovery"},
432 {Opt_ratio, "metadata_ratio=%u"},
433 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
434 {Opt_skip_balance, "skip_balance"},
435 {Opt_space_cache, "space_cache"},
436 {Opt_no_space_cache, "nospace_cache"},
437 {Opt_space_cache_version, "space_cache=%s"},
439 {Opt_nossd, "nossd"},
440 {Opt_ssd_spread, "ssd_spread"},
441 {Opt_nossd_spread, "nossd_spread"},
442 {Opt_subvol, "subvol=%s"},
443 {Opt_subvol_empty, "subvol="},
444 {Opt_subvolid, "subvolid=%s"},
445 {Opt_thread_pool, "thread_pool=%u"},
446 {Opt_treelog, "treelog"},
447 {Opt_notreelog, "notreelog"},
448 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
451 {Opt_rescue, "rescue=%s"},
452 /* Deprecated, with alias rescue=nologreplay */
453 {Opt_nologreplay, "nologreplay"},
454 /* Deprecated, with alias rescue=usebackuproot */
455 {Opt_usebackuproot, "usebackuproot"},
457 /* Deprecated options */
458 {Opt_recovery, "recovery"},
460 /* Debugging options */
461 {Opt_check_integrity, "check_int"},
462 {Opt_check_integrity_including_extent_data, "check_int_data"},
463 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
464 {Opt_enospc_debug, "enospc_debug"},
465 {Opt_noenospc_debug, "noenospc_debug"},
466 #ifdef CONFIG_BTRFS_DEBUG
467 {Opt_fragment_data, "fragment=data"},
468 {Opt_fragment_metadata, "fragment=metadata"},
469 {Opt_fragment_all, "fragment=all"},
471 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
472 {Opt_ref_verify, "ref_verify"},
477 static const match_table_t rescue_tokens = {
478 {Opt_usebackuproot, "usebackuproot"},
479 {Opt_nologreplay, "nologreplay"},
480 {Opt_ignorebadroots, "ignorebadroots"},
481 {Opt_ignorebadroots, "ibadroots"},
482 {Opt_ignoredatacsums, "ignoredatacsums"},
483 {Opt_ignoredatacsums, "idatacsums"},
484 {Opt_rescue_all, "all"},
488 static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
489 const char *opt_name)
491 if (fs_info->mount_opt & opt) {
492 btrfs_err(fs_info, "%s must be used with ro mount option",
499 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
504 substring_t args[MAX_OPT_ARGS];
507 opts = kstrdup(options, GFP_KERNEL);
512 while ((p = strsep(&opts, ":")) != NULL) {
517 token = match_token(p, rescue_tokens, args);
519 case Opt_usebackuproot:
521 "trying to use backup root at mount time");
522 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
524 case Opt_nologreplay:
525 btrfs_set_and_info(info, NOLOGREPLAY,
526 "disabling log replay at mount time");
528 case Opt_ignorebadroots:
529 btrfs_set_and_info(info, IGNOREBADROOTS,
530 "ignoring bad roots");
532 case Opt_ignoredatacsums:
533 btrfs_set_and_info(info, IGNOREDATACSUMS,
534 "ignoring data csums");
537 btrfs_info(info, "enabling all of the rescue options");
538 btrfs_set_and_info(info, IGNOREDATACSUMS,
539 "ignoring data csums");
540 btrfs_set_and_info(info, IGNOREBADROOTS,
541 "ignoring bad roots");
542 btrfs_set_and_info(info, NOLOGREPLAY,
543 "disabling log replay at mount time");
546 btrfs_info(info, "unrecognized rescue option '%s'", p);
560 * Regular mount options parser. Everything that is needed only when
561 * reading in a new superblock is parsed here.
562 * XXX JDM: This needs to be cleaned up for remount.
564 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
565 unsigned long new_flags)
567 substring_t args[MAX_OPT_ARGS];
572 bool compress_force = false;
573 enum btrfs_compression_type saved_compress_type;
574 int saved_compress_level;
575 bool saved_compress_force;
577 const bool remounting = test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state);
579 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
580 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
581 else if (btrfs_free_space_cache_v1_active(info)) {
582 if (btrfs_is_zoned(info)) {
584 "zoned: clearing existing space cache");
585 btrfs_set_super_cache_generation(info->super_copy, 0);
587 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
592 * Even the options are empty, we still need to do extra check
598 while ((p = strsep(&options, ",")) != NULL) {
603 token = match_token(p, tokens, args);
606 btrfs_info(info, "allowing degraded mounts");
607 btrfs_set_opt(info->mount_opt, DEGRADED);
610 case Opt_subvol_empty:
614 * These are parsed by btrfs_parse_subvol_options or
615 * btrfs_parse_device_options and can be ignored here.
619 btrfs_set_and_info(info, NODATASUM,
620 "setting nodatasum");
623 if (btrfs_test_opt(info, NODATASUM)) {
624 if (btrfs_test_opt(info, NODATACOW))
626 "setting datasum, datacow enabled");
628 btrfs_info(info, "setting datasum");
630 btrfs_clear_opt(info->mount_opt, NODATACOW);
631 btrfs_clear_opt(info->mount_opt, NODATASUM);
634 if (!btrfs_test_opt(info, NODATACOW)) {
635 if (!btrfs_test_opt(info, COMPRESS) ||
636 !btrfs_test_opt(info, FORCE_COMPRESS)) {
638 "setting nodatacow, compression disabled");
640 btrfs_info(info, "setting nodatacow");
643 btrfs_clear_opt(info->mount_opt, COMPRESS);
644 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
645 btrfs_set_opt(info->mount_opt, NODATACOW);
646 btrfs_set_opt(info->mount_opt, NODATASUM);
649 btrfs_clear_and_info(info, NODATACOW,
652 case Opt_compress_force:
653 case Opt_compress_force_type:
654 compress_force = true;
657 case Opt_compress_type:
658 saved_compress_type = btrfs_test_opt(info,
660 info->compress_type : BTRFS_COMPRESS_NONE;
661 saved_compress_force =
662 btrfs_test_opt(info, FORCE_COMPRESS);
663 saved_compress_level = info->compress_level;
664 if (token == Opt_compress ||
665 token == Opt_compress_force ||
666 strncmp(args[0].from, "zlib", 4) == 0) {
667 compress_type = "zlib";
669 info->compress_type = BTRFS_COMPRESS_ZLIB;
670 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
672 * args[0] contains uninitialized data since
673 * for these tokens we don't expect any
676 if (token != Opt_compress &&
677 token != Opt_compress_force)
678 info->compress_level =
679 btrfs_compress_str2level(
682 btrfs_set_opt(info->mount_opt, COMPRESS);
683 btrfs_clear_opt(info->mount_opt, NODATACOW);
684 btrfs_clear_opt(info->mount_opt, NODATASUM);
686 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
687 compress_type = "lzo";
688 info->compress_type = BTRFS_COMPRESS_LZO;
689 info->compress_level = 0;
690 btrfs_set_opt(info->mount_opt, COMPRESS);
691 btrfs_clear_opt(info->mount_opt, NODATACOW);
692 btrfs_clear_opt(info->mount_opt, NODATASUM);
693 btrfs_set_fs_incompat(info, COMPRESS_LZO);
695 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
696 compress_type = "zstd";
697 info->compress_type = BTRFS_COMPRESS_ZSTD;
698 info->compress_level =
699 btrfs_compress_str2level(
702 btrfs_set_opt(info->mount_opt, COMPRESS);
703 btrfs_clear_opt(info->mount_opt, NODATACOW);
704 btrfs_clear_opt(info->mount_opt, NODATASUM);
705 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
707 } else if (strncmp(args[0].from, "no", 2) == 0) {
708 compress_type = "no";
709 info->compress_level = 0;
710 info->compress_type = 0;
711 btrfs_clear_opt(info->mount_opt, COMPRESS);
712 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
713 compress_force = false;
716 btrfs_err(info, "unrecognized compression value %s",
722 if (compress_force) {
723 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
726 * If we remount from compress-force=xxx to
727 * compress=xxx, we need clear FORCE_COMPRESS
728 * flag, otherwise, there is no way for users
729 * to disable forcible compression separately.
731 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
733 if (no_compress == 1) {
734 btrfs_info(info, "use no compression");
735 } else if ((info->compress_type != saved_compress_type) ||
736 (compress_force != saved_compress_force) ||
737 (info->compress_level != saved_compress_level)) {
738 btrfs_info(info, "%s %s compression, level %d",
739 (compress_force) ? "force" : "use",
740 compress_type, info->compress_level);
742 compress_force = false;
745 btrfs_set_and_info(info, SSD,
746 "enabling ssd optimizations");
747 btrfs_clear_opt(info->mount_opt, NOSSD);
750 btrfs_set_and_info(info, SSD,
751 "enabling ssd optimizations");
752 btrfs_set_and_info(info, SSD_SPREAD,
753 "using spread ssd allocation scheme");
754 btrfs_clear_opt(info->mount_opt, NOSSD);
757 btrfs_set_opt(info->mount_opt, NOSSD);
758 btrfs_clear_and_info(info, SSD,
759 "not using ssd optimizations");
761 case Opt_nossd_spread:
762 btrfs_clear_and_info(info, SSD_SPREAD,
763 "not using spread ssd allocation scheme");
766 btrfs_clear_and_info(info, NOBARRIER,
767 "turning on barriers");
770 btrfs_set_and_info(info, NOBARRIER,
771 "turning off barriers");
773 case Opt_thread_pool:
774 ret = match_int(&args[0], &intarg);
776 btrfs_err(info, "unrecognized thread_pool value %s",
779 } else if (intarg == 0) {
780 btrfs_err(info, "invalid value 0 for thread_pool");
784 info->thread_pool_size = intarg;
787 num = match_strdup(&args[0]);
789 info->max_inline = memparse(num, NULL);
792 if (info->max_inline) {
793 info->max_inline = min_t(u64,
797 btrfs_info(info, "max_inline at %llu",
805 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
806 info->sb->s_flags |= SB_POSIXACL;
809 btrfs_err(info, "support for ACL not compiled in!");
814 info->sb->s_flags &= ~SB_POSIXACL;
817 btrfs_set_and_info(info, NOTREELOG,
818 "disabling tree log");
821 btrfs_clear_and_info(info, NOTREELOG,
822 "enabling tree log");
825 case Opt_nologreplay:
827 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
828 btrfs_set_and_info(info, NOLOGREPLAY,
829 "disabling log replay at mount time");
831 case Opt_flushoncommit:
832 btrfs_set_and_info(info, FLUSHONCOMMIT,
833 "turning on flush-on-commit");
835 case Opt_noflushoncommit:
836 btrfs_clear_and_info(info, FLUSHONCOMMIT,
837 "turning off flush-on-commit");
840 ret = match_int(&args[0], &intarg);
842 btrfs_err(info, "unrecognized metadata_ratio value %s",
846 info->metadata_ratio = intarg;
847 btrfs_info(info, "metadata ratio %u",
848 info->metadata_ratio);
851 case Opt_discard_mode:
852 if (token == Opt_discard ||
853 strcmp(args[0].from, "sync") == 0) {
854 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
855 btrfs_set_and_info(info, DISCARD_SYNC,
856 "turning on sync discard");
857 } else if (strcmp(args[0].from, "async") == 0) {
858 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
859 btrfs_set_and_info(info, DISCARD_ASYNC,
860 "turning on async discard");
862 btrfs_err(info, "unrecognized discard mode value %s",
869 btrfs_clear_and_info(info, DISCARD_SYNC,
870 "turning off discard");
871 btrfs_clear_and_info(info, DISCARD_ASYNC,
872 "turning off async discard");
874 case Opt_space_cache:
875 case Opt_space_cache_version:
876 if (token == Opt_space_cache ||
877 strcmp(args[0].from, "v1") == 0) {
878 btrfs_clear_opt(info->mount_opt,
880 btrfs_set_and_info(info, SPACE_CACHE,
881 "enabling disk space caching");
882 } else if (strcmp(args[0].from, "v2") == 0) {
883 btrfs_clear_opt(info->mount_opt,
885 btrfs_set_and_info(info, FREE_SPACE_TREE,
886 "enabling free space tree");
888 btrfs_err(info, "unrecognized space_cache value %s",
894 case Opt_rescan_uuid_tree:
895 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
897 case Opt_no_space_cache:
898 if (btrfs_test_opt(info, SPACE_CACHE)) {
899 btrfs_clear_and_info(info, SPACE_CACHE,
900 "disabling disk space caching");
902 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
903 btrfs_clear_and_info(info, FREE_SPACE_TREE,
904 "disabling free space tree");
907 case Opt_inode_cache:
908 case Opt_noinode_cache:
910 "the 'inode_cache' option is deprecated and has no effect since 5.11");
912 case Opt_clear_cache:
913 btrfs_set_and_info(info, CLEAR_CACHE,
914 "force clearing of disk cache");
916 case Opt_user_subvol_rm_allowed:
917 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
919 case Opt_enospc_debug:
920 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
922 case Opt_noenospc_debug:
923 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
926 btrfs_set_and_info(info, AUTO_DEFRAG,
927 "enabling auto defrag");
930 btrfs_clear_and_info(info, AUTO_DEFRAG,
931 "disabling auto defrag");
934 case Opt_usebackuproot:
936 "'%s' is deprecated, use 'rescue=usebackuproot' instead",
937 token == Opt_recovery ? "recovery" :
940 "trying to use backup root at mount time");
941 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
943 case Opt_skip_balance:
944 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
946 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
947 case Opt_check_integrity_including_extent_data:
949 "enabling check integrity including extent data");
950 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
951 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
953 case Opt_check_integrity:
954 btrfs_info(info, "enabling check integrity");
955 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
957 case Opt_check_integrity_print_mask:
958 ret = match_int(&args[0], &intarg);
961 "unrecognized check_integrity_print_mask value %s",
965 info->check_integrity_print_mask = intarg;
966 btrfs_info(info, "check_integrity_print_mask 0x%x",
967 info->check_integrity_print_mask);
970 case Opt_check_integrity_including_extent_data:
971 case Opt_check_integrity:
972 case Opt_check_integrity_print_mask:
974 "support for check_integrity* not compiled in!");
978 case Opt_fatal_errors:
979 if (strcmp(args[0].from, "panic") == 0) {
980 btrfs_set_opt(info->mount_opt,
981 PANIC_ON_FATAL_ERROR);
982 } else if (strcmp(args[0].from, "bug") == 0) {
983 btrfs_clear_opt(info->mount_opt,
984 PANIC_ON_FATAL_ERROR);
986 btrfs_err(info, "unrecognized fatal_errors value %s",
992 case Opt_commit_interval:
994 ret = match_int(&args[0], &intarg);
996 btrfs_err(info, "unrecognized commit_interval value %s",
1003 "using default commit interval %us",
1004 BTRFS_DEFAULT_COMMIT_INTERVAL);
1005 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
1006 } else if (intarg > 300) {
1007 btrfs_warn(info, "excessive commit interval %d",
1010 info->commit_interval = intarg;
1013 ret = parse_rescue_options(info, args[0].from);
1015 btrfs_err(info, "unrecognized rescue value %s",
1020 #ifdef CONFIG_BTRFS_DEBUG
1021 case Opt_fragment_all:
1022 btrfs_info(info, "fragmenting all space");
1023 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1024 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
1026 case Opt_fragment_metadata:
1027 btrfs_info(info, "fragmenting metadata");
1028 btrfs_set_opt(info->mount_opt,
1031 case Opt_fragment_data:
1032 btrfs_info(info, "fragmenting data");
1033 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1036 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
1037 case Opt_ref_verify:
1038 btrfs_info(info, "doing ref verification");
1039 btrfs_set_opt(info->mount_opt, REF_VERIFY);
1043 btrfs_err(info, "unrecognized mount option '%s'", p);
1051 /* We're read-only, don't have to check. */
1052 if (new_flags & SB_RDONLY)
1055 if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
1056 check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
1057 check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
1060 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
1061 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
1062 !btrfs_test_opt(info, CLEAR_CACHE)) {
1063 btrfs_err(info, "cannot disable free space tree");
1068 ret = btrfs_check_mountopts_zoned(info);
1069 if (!ret && !remounting) {
1070 if (btrfs_test_opt(info, SPACE_CACHE))
1071 btrfs_info(info, "disk space caching is enabled");
1072 if (btrfs_test_opt(info, FREE_SPACE_TREE))
1073 btrfs_info(info, "using free space tree");
1079 * Parse mount options that are required early in the mount process.
1081 * All other options will be parsed on much later in the mount process and
1082 * only when we need to allocate a new super block.
1084 static int btrfs_parse_device_options(const char *options, fmode_t flags,
1087 substring_t args[MAX_OPT_ARGS];
1088 char *device_name, *opts, *orig, *p;
1089 struct btrfs_device *device = NULL;
1092 lockdep_assert_held(&uuid_mutex);
1098 * strsep changes the string, duplicate it because btrfs_parse_options
1101 opts = kstrdup(options, GFP_KERNEL);
1106 while ((p = strsep(&opts, ",")) != NULL) {
1112 token = match_token(p, tokens, args);
1113 if (token == Opt_device) {
1114 device_name = match_strdup(&args[0]);
1119 device = btrfs_scan_one_device(device_name, flags,
1122 if (IS_ERR(device)) {
1123 error = PTR_ERR(device);
1135 * Parse mount options that are related to subvolume id
1137 * The value is later passed to mount_subvol()
1139 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1140 u64 *subvol_objectid)
1142 substring_t args[MAX_OPT_ARGS];
1143 char *opts, *orig, *p;
1151 * strsep changes the string, duplicate it because
1152 * btrfs_parse_device_options gets called later
1154 opts = kstrdup(options, GFP_KERNEL);
1159 while ((p = strsep(&opts, ",")) != NULL) {
1164 token = match_token(p, tokens, args);
1167 kfree(*subvol_name);
1168 *subvol_name = match_strdup(&args[0]);
1169 if (!*subvol_name) {
1175 error = match_u64(&args[0], &subvolid);
1179 /* we want the original fs_tree */
1181 subvolid = BTRFS_FS_TREE_OBJECTID;
1183 *subvol_objectid = subvolid;
1195 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1196 u64 subvol_objectid)
1198 struct btrfs_root *root = fs_info->tree_root;
1199 struct btrfs_root *fs_root = NULL;
1200 struct btrfs_root_ref *root_ref;
1201 struct btrfs_inode_ref *inode_ref;
1202 struct btrfs_key key;
1203 struct btrfs_path *path = NULL;
1204 char *name = NULL, *ptr;
1209 path = btrfs_alloc_path();
1215 name = kmalloc(PATH_MAX, GFP_KERNEL);
1220 ptr = name + PATH_MAX - 1;
1224 * Walk up the subvolume trees in the tree of tree roots by root
1225 * backrefs until we hit the top-level subvolume.
1227 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1228 key.objectid = subvol_objectid;
1229 key.type = BTRFS_ROOT_BACKREF_KEY;
1230 key.offset = (u64)-1;
1232 ret = btrfs_search_backwards(root, &key, path);
1235 } else if (ret > 0) {
1240 subvol_objectid = key.offset;
1242 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1243 struct btrfs_root_ref);
1244 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1247 ret = -ENAMETOOLONG;
1250 read_extent_buffer(path->nodes[0], ptr + 1,
1251 (unsigned long)(root_ref + 1), len);
1253 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1254 btrfs_release_path(path);
1256 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1257 if (IS_ERR(fs_root)) {
1258 ret = PTR_ERR(fs_root);
1264 * Walk up the filesystem tree by inode refs until we hit the
1267 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1268 key.objectid = dirid;
1269 key.type = BTRFS_INODE_REF_KEY;
1270 key.offset = (u64)-1;
1272 ret = btrfs_search_backwards(fs_root, &key, path);
1275 } else if (ret > 0) {
1282 inode_ref = btrfs_item_ptr(path->nodes[0],
1284 struct btrfs_inode_ref);
1285 len = btrfs_inode_ref_name_len(path->nodes[0],
1289 ret = -ENAMETOOLONG;
1292 read_extent_buffer(path->nodes[0], ptr + 1,
1293 (unsigned long)(inode_ref + 1), len);
1295 btrfs_release_path(path);
1297 btrfs_put_root(fs_root);
1301 btrfs_free_path(path);
1302 if (ptr == name + PATH_MAX - 1) {
1306 memmove(name, ptr, name + PATH_MAX - ptr);
1311 btrfs_put_root(fs_root);
1312 btrfs_free_path(path);
1314 return ERR_PTR(ret);
1317 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1319 struct btrfs_root *root = fs_info->tree_root;
1320 struct btrfs_dir_item *di;
1321 struct btrfs_path *path;
1322 struct btrfs_key location;
1325 path = btrfs_alloc_path();
1330 * Find the "default" dir item which points to the root item that we
1331 * will mount by default if we haven't been given a specific subvolume
1334 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1335 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1337 btrfs_free_path(path);
1342 * Ok the default dir item isn't there. This is weird since
1343 * it's always been there, but don't freak out, just try and
1344 * mount the top-level subvolume.
1346 btrfs_free_path(path);
1347 *objectid = BTRFS_FS_TREE_OBJECTID;
1351 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1352 btrfs_free_path(path);
1353 *objectid = location.objectid;
1357 static int btrfs_fill_super(struct super_block *sb,
1358 struct btrfs_fs_devices *fs_devices,
1361 struct inode *inode;
1362 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1365 sb->s_maxbytes = MAX_LFS_FILESIZE;
1366 sb->s_magic = BTRFS_SUPER_MAGIC;
1367 sb->s_op = &btrfs_super_ops;
1368 sb->s_d_op = &btrfs_dentry_operations;
1369 sb->s_export_op = &btrfs_export_ops;
1370 #ifdef CONFIG_FS_VERITY
1371 sb->s_vop = &btrfs_verityops;
1373 sb->s_xattr = btrfs_xattr_handlers;
1374 sb->s_time_gran = 1;
1375 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1376 sb->s_flags |= SB_POSIXACL;
1378 sb->s_flags |= SB_I_VERSION;
1379 sb->s_iflags |= SB_I_CGROUPWB;
1381 err = super_setup_bdi(sb);
1383 btrfs_err(fs_info, "super_setup_bdi failed");
1387 err = open_ctree(sb, fs_devices, (char *)data);
1389 btrfs_err(fs_info, "open_ctree failed");
1393 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1394 if (IS_ERR(inode)) {
1395 err = PTR_ERR(inode);
1399 sb->s_root = d_make_root(inode);
1405 cleancache_init_fs(sb);
1406 sb->s_flags |= SB_ACTIVE;
1410 close_ctree(fs_info);
1414 int btrfs_sync_fs(struct super_block *sb, int wait)
1416 struct btrfs_trans_handle *trans;
1417 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1418 struct btrfs_root *root = fs_info->tree_root;
1420 trace_btrfs_sync_fs(fs_info, wait);
1423 filemap_flush(fs_info->btree_inode->i_mapping);
1427 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1429 trans = btrfs_attach_transaction_barrier(root);
1430 if (IS_ERR(trans)) {
1431 /* no transaction, don't bother */
1432 if (PTR_ERR(trans) == -ENOENT) {
1434 * Exit unless we have some pending changes
1435 * that need to go through commit
1437 if (fs_info->pending_changes == 0)
1440 * A non-blocking test if the fs is frozen. We must not
1441 * start a new transaction here otherwise a deadlock
1442 * happens. The pending operations are delayed to the
1443 * next commit after thawing.
1445 if (sb_start_write_trylock(sb))
1449 trans = btrfs_start_transaction(root, 0);
1452 return PTR_ERR(trans);
1454 return btrfs_commit_transaction(trans);
1457 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1459 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1463 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1465 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1466 const char *compress_type;
1467 const char *subvol_name;
1468 bool printed = false;
1470 if (btrfs_test_opt(info, DEGRADED))
1471 seq_puts(seq, ",degraded");
1472 if (btrfs_test_opt(info, NODATASUM))
1473 seq_puts(seq, ",nodatasum");
1474 if (btrfs_test_opt(info, NODATACOW))
1475 seq_puts(seq, ",nodatacow");
1476 if (btrfs_test_opt(info, NOBARRIER))
1477 seq_puts(seq, ",nobarrier");
1478 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1479 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1480 if (info->thread_pool_size != min_t(unsigned long,
1481 num_online_cpus() + 2, 8))
1482 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1483 if (btrfs_test_opt(info, COMPRESS)) {
1484 compress_type = btrfs_compress_type2str(info->compress_type);
1485 if (btrfs_test_opt(info, FORCE_COMPRESS))
1486 seq_printf(seq, ",compress-force=%s", compress_type);
1488 seq_printf(seq, ",compress=%s", compress_type);
1489 if (info->compress_level)
1490 seq_printf(seq, ":%d", info->compress_level);
1492 if (btrfs_test_opt(info, NOSSD))
1493 seq_puts(seq, ",nossd");
1494 if (btrfs_test_opt(info, SSD_SPREAD))
1495 seq_puts(seq, ",ssd_spread");
1496 else if (btrfs_test_opt(info, SSD))
1497 seq_puts(seq, ",ssd");
1498 if (btrfs_test_opt(info, NOTREELOG))
1499 seq_puts(seq, ",notreelog");
1500 if (btrfs_test_opt(info, NOLOGREPLAY))
1501 print_rescue_option(seq, "nologreplay", &printed);
1502 if (btrfs_test_opt(info, USEBACKUPROOT))
1503 print_rescue_option(seq, "usebackuproot", &printed);
1504 if (btrfs_test_opt(info, IGNOREBADROOTS))
1505 print_rescue_option(seq, "ignorebadroots", &printed);
1506 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1507 print_rescue_option(seq, "ignoredatacsums", &printed);
1508 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1509 seq_puts(seq, ",flushoncommit");
1510 if (btrfs_test_opt(info, DISCARD_SYNC))
1511 seq_puts(seq, ",discard");
1512 if (btrfs_test_opt(info, DISCARD_ASYNC))
1513 seq_puts(seq, ",discard=async");
1514 if (!(info->sb->s_flags & SB_POSIXACL))
1515 seq_puts(seq, ",noacl");
1516 if (btrfs_free_space_cache_v1_active(info))
1517 seq_puts(seq, ",space_cache");
1518 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1519 seq_puts(seq, ",space_cache=v2");
1521 seq_puts(seq, ",nospace_cache");
1522 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1523 seq_puts(seq, ",rescan_uuid_tree");
1524 if (btrfs_test_opt(info, CLEAR_CACHE))
1525 seq_puts(seq, ",clear_cache");
1526 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1527 seq_puts(seq, ",user_subvol_rm_allowed");
1528 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1529 seq_puts(seq, ",enospc_debug");
1530 if (btrfs_test_opt(info, AUTO_DEFRAG))
1531 seq_puts(seq, ",autodefrag");
1532 if (btrfs_test_opt(info, SKIP_BALANCE))
1533 seq_puts(seq, ",skip_balance");
1534 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1535 if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1536 seq_puts(seq, ",check_int_data");
1537 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1538 seq_puts(seq, ",check_int");
1539 if (info->check_integrity_print_mask)
1540 seq_printf(seq, ",check_int_print_mask=%d",
1541 info->check_integrity_print_mask);
1543 if (info->metadata_ratio)
1544 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1545 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1546 seq_puts(seq, ",fatal_errors=panic");
1547 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1548 seq_printf(seq, ",commit=%u", info->commit_interval);
1549 #ifdef CONFIG_BTRFS_DEBUG
1550 if (btrfs_test_opt(info, FRAGMENT_DATA))
1551 seq_puts(seq, ",fragment=data");
1552 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1553 seq_puts(seq, ",fragment=metadata");
1555 if (btrfs_test_opt(info, REF_VERIFY))
1556 seq_puts(seq, ",ref_verify");
1557 seq_printf(seq, ",subvolid=%llu",
1558 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1559 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1560 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1561 if (!IS_ERR(subvol_name)) {
1562 seq_puts(seq, ",subvol=");
1563 seq_escape(seq, subvol_name, " \t\n\\");
1569 static int btrfs_test_super(struct super_block *s, void *data)
1571 struct btrfs_fs_info *p = data;
1572 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1574 return fs_info->fs_devices == p->fs_devices;
1577 static int btrfs_set_super(struct super_block *s, void *data)
1579 int err = set_anon_super(s, data);
1581 s->s_fs_info = data;
1586 * subvolumes are identified by ino 256
1588 static inline int is_subvolume_inode(struct inode *inode)
1590 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1595 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1596 struct vfsmount *mnt)
1598 struct dentry *root;
1602 if (!subvol_objectid) {
1603 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1606 root = ERR_PTR(ret);
1610 subvol_name = btrfs_get_subvol_name_from_objectid(
1611 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1612 if (IS_ERR(subvol_name)) {
1613 root = ERR_CAST(subvol_name);
1620 root = mount_subtree(mnt, subvol_name);
1621 /* mount_subtree() drops our reference on the vfsmount. */
1624 if (!IS_ERR(root)) {
1625 struct super_block *s = root->d_sb;
1626 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1627 struct inode *root_inode = d_inode(root);
1628 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1631 if (!is_subvolume_inode(root_inode)) {
1632 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1636 if (subvol_objectid && root_objectid != subvol_objectid) {
1638 * This will also catch a race condition where a
1639 * subvolume which was passed by ID is renamed and
1640 * another subvolume is renamed over the old location.
1643 "subvol '%s' does not match subvolid %llu",
1644 subvol_name, subvol_objectid);
1649 root = ERR_PTR(ret);
1650 deactivate_locked_super(s);
1661 * Find a superblock for the given device / mount point.
1663 * Note: This is based on mount_bdev from fs/super.c with a few additions
1664 * for multiple device setup. Make sure to keep it in sync.
1666 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1667 int flags, const char *device_name, void *data)
1669 struct block_device *bdev = NULL;
1670 struct super_block *s;
1671 struct btrfs_device *device = NULL;
1672 struct btrfs_fs_devices *fs_devices = NULL;
1673 struct btrfs_fs_info *fs_info = NULL;
1674 void *new_sec_opts = NULL;
1675 fmode_t mode = FMODE_READ;
1678 if (!(flags & SB_RDONLY))
1679 mode |= FMODE_WRITE;
1682 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1684 return ERR_PTR(error);
1688 * Setup a dummy root and fs_info for test/set super. This is because
1689 * we don't actually fill this stuff out until open_ctree, but we need
1690 * then open_ctree will properly initialize the file system specific
1691 * settings later. btrfs_init_fs_info initializes the static elements
1692 * of the fs_info (locks and such) to make cleanup easier if we find a
1693 * superblock with our given fs_devices later on at sget() time.
1695 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1698 goto error_sec_opts;
1700 btrfs_init_fs_info(fs_info);
1702 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1703 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1704 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1709 mutex_lock(&uuid_mutex);
1710 error = btrfs_parse_device_options(data, mode, fs_type);
1712 mutex_unlock(&uuid_mutex);
1716 device = btrfs_scan_one_device(device_name, mode, fs_type);
1717 if (IS_ERR(device)) {
1718 mutex_unlock(&uuid_mutex);
1719 error = PTR_ERR(device);
1723 fs_devices = device->fs_devices;
1724 fs_info->fs_devices = fs_devices;
1726 error = btrfs_open_devices(fs_devices, mode, fs_type);
1727 mutex_unlock(&uuid_mutex);
1731 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1733 goto error_close_devices;
1736 bdev = fs_devices->latest_dev->bdev;
1737 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1741 goto error_close_devices;
1745 btrfs_close_devices(fs_devices);
1746 btrfs_free_fs_info(fs_info);
1747 if ((flags ^ s->s_flags) & SB_RDONLY)
1750 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1751 btrfs_sb(s)->bdev_holder = fs_type;
1752 if (!strstr(crc32c_impl(), "generic"))
1753 set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1754 error = btrfs_fill_super(s, fs_devices, data);
1757 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1758 security_free_mnt_opts(&new_sec_opts);
1760 deactivate_locked_super(s);
1761 return ERR_PTR(error);
1764 return dget(s->s_root);
1766 error_close_devices:
1767 btrfs_close_devices(fs_devices);
1769 btrfs_free_fs_info(fs_info);
1771 security_free_mnt_opts(&new_sec_opts);
1772 return ERR_PTR(error);
1776 * Mount function which is called by VFS layer.
1778 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1779 * which needs vfsmount* of device's root (/). This means device's root has to
1780 * be mounted internally in any case.
1783 * 1. Parse subvol id related options for later use in mount_subvol().
1785 * 2. Mount device's root (/) by calling vfs_kern_mount().
1787 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1788 * first place. In order to avoid calling btrfs_mount() again, we use
1789 * different file_system_type which is not registered to VFS by
1790 * register_filesystem() (btrfs_root_fs_type). As a result,
1791 * btrfs_mount_root() is called. The return value will be used by
1792 * mount_subtree() in mount_subvol().
1794 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1795 * "btrfs subvolume set-default", mount_subvol() is called always.
1797 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1798 const char *device_name, void *data)
1800 struct vfsmount *mnt_root;
1801 struct dentry *root;
1802 char *subvol_name = NULL;
1803 u64 subvol_objectid = 0;
1806 error = btrfs_parse_subvol_options(data, &subvol_name,
1810 return ERR_PTR(error);
1813 /* mount device's root (/) */
1814 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1815 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1816 if (flags & SB_RDONLY) {
1817 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1818 flags & ~SB_RDONLY, device_name, data);
1820 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1821 flags | SB_RDONLY, device_name, data);
1822 if (IS_ERR(mnt_root)) {
1823 root = ERR_CAST(mnt_root);
1828 down_write(&mnt_root->mnt_sb->s_umount);
1829 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1830 up_write(&mnt_root->mnt_sb->s_umount);
1832 root = ERR_PTR(error);
1839 if (IS_ERR(mnt_root)) {
1840 root = ERR_CAST(mnt_root);
1845 /* mount_subvol() will free subvol_name and mnt_root */
1846 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1852 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1853 u32 new_pool_size, u32 old_pool_size)
1855 if (new_pool_size == old_pool_size)
1858 fs_info->thread_pool_size = new_pool_size;
1860 btrfs_info(fs_info, "resize thread pool %d -> %d",
1861 old_pool_size, new_pool_size);
1863 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1864 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1865 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1866 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1867 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1868 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1870 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1871 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1872 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1873 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1874 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1878 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1879 unsigned long old_opts, int flags)
1881 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1882 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1883 (flags & SB_RDONLY))) {
1884 /* wait for any defraggers to finish */
1885 wait_event(fs_info->transaction_wait,
1886 (atomic_read(&fs_info->defrag_running) == 0));
1887 if (flags & SB_RDONLY)
1888 sync_filesystem(fs_info->sb);
1892 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1893 unsigned long old_opts)
1895 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1898 * We need to cleanup all defragable inodes if the autodefragment is
1899 * close or the filesystem is read only.
1901 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1902 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1903 btrfs_cleanup_defrag_inodes(fs_info);
1906 /* If we toggled discard async */
1907 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1908 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1909 btrfs_discard_resume(fs_info);
1910 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1911 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1912 btrfs_discard_cleanup(fs_info);
1914 /* If we toggled space cache */
1915 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1916 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1919 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1921 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1922 unsigned old_flags = sb->s_flags;
1923 unsigned long old_opts = fs_info->mount_opt;
1924 unsigned long old_compress_type = fs_info->compress_type;
1925 u64 old_max_inline = fs_info->max_inline;
1926 u32 old_thread_pool_size = fs_info->thread_pool_size;
1927 u32 old_metadata_ratio = fs_info->metadata_ratio;
1930 sync_filesystem(sb);
1931 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1934 void *new_sec_opts = NULL;
1936 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1938 ret = security_sb_remount(sb, new_sec_opts);
1939 security_free_mnt_opts(&new_sec_opts);
1944 ret = btrfs_parse_options(fs_info, data, *flags);
1948 /* V1 cache is not supported for subpage mount. */
1949 if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) {
1951 "v1 space cache is not supported for page size %lu with sectorsize %u",
1952 PAGE_SIZE, fs_info->sectorsize);
1956 btrfs_remount_begin(fs_info, old_opts, *flags);
1957 btrfs_resize_thread_pool(fs_info,
1958 fs_info->thread_pool_size, old_thread_pool_size);
1960 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1961 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1962 (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1964 "remount supports changing free space tree only from ro to rw");
1965 /* Make sure free space cache options match the state on disk */
1966 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1967 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1968 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1970 if (btrfs_free_space_cache_v1_active(fs_info)) {
1971 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1972 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1976 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1979 if (*flags & SB_RDONLY) {
1981 * this also happens on 'umount -rf' or on shutdown, when
1982 * the filesystem is busy.
1984 cancel_work_sync(&fs_info->async_reclaim_work);
1985 cancel_work_sync(&fs_info->async_data_reclaim_work);
1987 btrfs_discard_cleanup(fs_info);
1989 /* wait for the uuid_scan task to finish */
1990 down(&fs_info->uuid_tree_rescan_sem);
1991 /* avoid complains from lockdep et al. */
1992 up(&fs_info->uuid_tree_rescan_sem);
1994 btrfs_set_sb_rdonly(sb);
1997 * Setting SB_RDONLY will put the cleaner thread to
1998 * sleep at the next loop if it's already active.
1999 * If it's already asleep, we'll leave unused block
2000 * groups on disk until we're mounted read-write again
2001 * unless we clean them up here.
2003 btrfs_delete_unused_bgs(fs_info);
2006 * The cleaner task could be already running before we set the
2007 * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
2008 * We must make sure that after we finish the remount, i.e. after
2009 * we call btrfs_commit_super(), the cleaner can no longer start
2010 * a transaction - either because it was dropping a dead root,
2011 * running delayed iputs or deleting an unused block group (the
2012 * cleaner picked a block group from the list of unused block
2013 * groups before we were able to in the previous call to
2014 * btrfs_delete_unused_bgs()).
2016 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
2017 TASK_UNINTERRUPTIBLE);
2020 * We've set the superblock to RO mode, so we might have made
2021 * the cleaner task sleep without running all pending delayed
2022 * iputs. Go through all the delayed iputs here, so that if an
2023 * unmount happens without remounting RW we don't end up at
2024 * finishing close_ctree() with a non-empty list of delayed
2027 btrfs_run_delayed_iputs(fs_info);
2029 btrfs_dev_replace_suspend_for_unmount(fs_info);
2030 btrfs_scrub_cancel(fs_info);
2031 btrfs_pause_balance(fs_info);
2034 * Pause the qgroup rescan worker if it is running. We don't want
2035 * it to be still running after we are in RO mode, as after that,
2036 * by the time we unmount, it might have left a transaction open,
2037 * so we would leak the transaction and/or crash.
2039 btrfs_qgroup_wait_for_completion(fs_info, false);
2041 ret = btrfs_commit_super(fs_info);
2045 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2047 "Remounting read-write after error is not allowed");
2051 if (btrfs_super_compat_ro_flags(fs_info->super_copy) &
2052 ~BTRFS_FEATURE_COMPAT_RO_SUPP) {
2054 "can not remount read-write due to unsupported optional flags 0x%llx",
2055 btrfs_super_compat_ro_flags(fs_info->super_copy) &
2056 ~BTRFS_FEATURE_COMPAT_RO_SUPP);
2060 if (fs_info->fs_devices->rw_devices == 0) {
2065 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
2067 "too many missing devices, writable remount is not allowed");
2072 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
2074 "mount required to replay tree-log, cannot remount read-write");
2080 * NOTE: when remounting with a change that does writes, don't
2081 * put it anywhere above this point, as we are not sure to be
2082 * safe to write until we pass the above checks.
2084 ret = btrfs_start_pre_rw_mount(fs_info);
2088 btrfs_clear_sb_rdonly(sb);
2090 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
2094 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
2095 * since the absence of the flag means it can be toggled off by remount.
2097 *flags |= SB_I_VERSION;
2099 wake_up_process(fs_info->transaction_kthread);
2100 btrfs_remount_cleanup(fs_info, old_opts);
2101 btrfs_clear_oneshot_options(fs_info);
2102 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2107 /* We've hit an error - don't reset SB_RDONLY */
2109 old_flags |= SB_RDONLY;
2110 if (!(old_flags & SB_RDONLY))
2111 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
2112 sb->s_flags = old_flags;
2113 fs_info->mount_opt = old_opts;
2114 fs_info->compress_type = old_compress_type;
2115 fs_info->max_inline = old_max_inline;
2116 btrfs_resize_thread_pool(fs_info,
2117 old_thread_pool_size, fs_info->thread_pool_size);
2118 fs_info->metadata_ratio = old_metadata_ratio;
2119 btrfs_remount_cleanup(fs_info, old_opts);
2120 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2125 /* Used to sort the devices by max_avail(descending sort) */
2126 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
2128 const struct btrfs_device_info *dev_info1 = a;
2129 const struct btrfs_device_info *dev_info2 = b;
2131 if (dev_info1->max_avail > dev_info2->max_avail)
2133 else if (dev_info1->max_avail < dev_info2->max_avail)
2139 * sort the devices by max_avail, in which max free extent size of each device
2140 * is stored.(Descending Sort)
2142 static inline void btrfs_descending_sort_devices(
2143 struct btrfs_device_info *devices,
2146 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2147 btrfs_cmp_device_free_bytes, NULL);
2151 * The helper to calc the free space on the devices that can be used to store
2154 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2157 struct btrfs_device_info *devices_info;
2158 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2159 struct btrfs_device *device;
2162 u64 min_stripe_size;
2163 int num_stripes = 1;
2164 int i = 0, nr_devices;
2165 const struct btrfs_raid_attr *rattr;
2168 * We aren't under the device list lock, so this is racy-ish, but good
2169 * enough for our purposes.
2171 nr_devices = fs_info->fs_devices->open_devices;
2174 nr_devices = fs_info->fs_devices->open_devices;
2182 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2187 /* calc min stripe number for data space allocation */
2188 type = btrfs_data_alloc_profile(fs_info);
2189 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2191 if (type & BTRFS_BLOCK_GROUP_RAID0)
2192 num_stripes = nr_devices;
2193 else if (type & BTRFS_BLOCK_GROUP_RAID1)
2195 else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
2197 else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
2199 else if (type & BTRFS_BLOCK_GROUP_RAID10)
2202 /* Adjust for more than 1 stripe per device */
2203 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2206 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2207 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2208 &device->dev_state) ||
2210 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2213 if (i >= nr_devices)
2216 avail_space = device->total_bytes - device->bytes_used;
2218 /* align with stripe_len */
2219 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2222 * In order to avoid overwriting the superblock on the drive,
2223 * btrfs starts at an offset of at least 1MB when doing chunk
2226 * This ensures we have at least min_stripe_size free space
2227 * after excluding 1MB.
2229 if (avail_space <= SZ_1M + min_stripe_size)
2232 avail_space -= SZ_1M;
2234 devices_info[i].dev = device;
2235 devices_info[i].max_avail = avail_space;
2243 btrfs_descending_sort_devices(devices_info, nr_devices);
2247 while (nr_devices >= rattr->devs_min) {
2248 num_stripes = min(num_stripes, nr_devices);
2250 if (devices_info[i].max_avail >= min_stripe_size) {
2254 avail_space += devices_info[i].max_avail * num_stripes;
2255 alloc_size = devices_info[i].max_avail;
2256 for (j = i + 1 - num_stripes; j <= i; j++)
2257 devices_info[j].max_avail -= alloc_size;
2263 kfree(devices_info);
2264 *free_bytes = avail_space;
2269 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2271 * If there's a redundant raid level at DATA block groups, use the respective
2272 * multiplier to scale the sizes.
2274 * Unused device space usage is based on simulating the chunk allocator
2275 * algorithm that respects the device sizes and order of allocations. This is
2276 * a close approximation of the actual use but there are other factors that may
2277 * change the result (like a new metadata chunk).
2279 * If metadata is exhausted, f_bavail will be 0.
2281 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2283 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2284 struct btrfs_super_block *disk_super = fs_info->super_copy;
2285 struct btrfs_space_info *found;
2287 u64 total_free_data = 0;
2288 u64 total_free_meta = 0;
2289 u32 bits = fs_info->sectorsize_bits;
2290 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2291 unsigned factor = 1;
2292 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2297 list_for_each_entry(found, &fs_info->space_info, list) {
2298 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2301 total_free_data += found->disk_total - found->disk_used;
2303 btrfs_account_ro_block_groups_free_space(found);
2305 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2306 if (!list_empty(&found->block_groups[i]))
2307 factor = btrfs_bg_type_to_factor(
2308 btrfs_raid_array[i].bg_flag);
2313 * Metadata in mixed block goup profiles are accounted in data
2315 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2316 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2319 total_free_meta += found->disk_total -
2323 total_used += found->disk_used;
2326 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2327 buf->f_blocks >>= bits;
2328 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2330 /* Account global block reserve as used, it's in logical size already */
2331 spin_lock(&block_rsv->lock);
2332 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2333 if (buf->f_bfree >= block_rsv->size >> bits)
2334 buf->f_bfree -= block_rsv->size >> bits;
2337 spin_unlock(&block_rsv->lock);
2339 buf->f_bavail = div_u64(total_free_data, factor);
2340 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2343 buf->f_bavail += div_u64(total_free_data, factor);
2344 buf->f_bavail = buf->f_bavail >> bits;
2347 * We calculate the remaining metadata space minus global reserve. If
2348 * this is (supposedly) smaller than zero, there's no space. But this
2349 * does not hold in practice, the exhausted state happens where's still
2350 * some positive delta. So we apply some guesswork and compare the
2351 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2353 * We probably cannot calculate the exact threshold value because this
2354 * depends on the internal reservations requested by various
2355 * operations, so some operations that consume a few metadata will
2356 * succeed even if the Avail is zero. But this is better than the other
2362 * We only want to claim there's no available space if we can no longer
2363 * allocate chunks for our metadata profile and our global reserve will
2364 * not fit in the free metadata space. If we aren't ->full then we
2365 * still can allocate chunks and thus are fine using the currently
2366 * calculated f_bavail.
2368 if (!mixed && block_rsv->space_info->full &&
2369 total_free_meta - thresh < block_rsv->size)
2372 buf->f_type = BTRFS_SUPER_MAGIC;
2373 buf->f_bsize = dentry->d_sb->s_blocksize;
2374 buf->f_namelen = BTRFS_NAME_LEN;
2376 /* We treat it as constant endianness (it doesn't matter _which_)
2377 because we want the fsid to come out the same whether mounted
2378 on a big-endian or little-endian host */
2379 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2380 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2381 /* Mask in the root object ID too, to disambiguate subvols */
2382 buf->f_fsid.val[0] ^=
2383 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2384 buf->f_fsid.val[1] ^=
2385 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2390 static void btrfs_kill_super(struct super_block *sb)
2392 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2393 kill_anon_super(sb);
2394 btrfs_free_fs_info(fs_info);
2397 static struct file_system_type btrfs_fs_type = {
2398 .owner = THIS_MODULE,
2400 .mount = btrfs_mount,
2401 .kill_sb = btrfs_kill_super,
2402 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2405 static struct file_system_type btrfs_root_fs_type = {
2406 .owner = THIS_MODULE,
2408 .mount = btrfs_mount_root,
2409 .kill_sb = btrfs_kill_super,
2410 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2413 MODULE_ALIAS_FS("btrfs");
2415 static int btrfs_control_open(struct inode *inode, struct file *file)
2418 * The control file's private_data is used to hold the
2419 * transaction when it is started and is used to keep
2420 * track of whether a transaction is already in progress.
2422 file->private_data = NULL;
2427 * Used by /dev/btrfs-control for devices ioctls.
2429 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2432 struct btrfs_ioctl_vol_args *vol;
2433 struct btrfs_device *device = NULL;
2436 if (!capable(CAP_SYS_ADMIN))
2439 vol = memdup_user((void __user *)arg, sizeof(*vol));
2441 return PTR_ERR(vol);
2442 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2445 case BTRFS_IOC_SCAN_DEV:
2446 mutex_lock(&uuid_mutex);
2447 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2448 &btrfs_root_fs_type);
2449 ret = PTR_ERR_OR_ZERO(device);
2450 mutex_unlock(&uuid_mutex);
2452 case BTRFS_IOC_FORGET_DEV:
2453 ret = btrfs_forget_devices(vol->name);
2455 case BTRFS_IOC_DEVICES_READY:
2456 mutex_lock(&uuid_mutex);
2457 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2458 &btrfs_root_fs_type);
2459 if (IS_ERR(device)) {
2460 mutex_unlock(&uuid_mutex);
2461 ret = PTR_ERR(device);
2464 ret = !(device->fs_devices->num_devices ==
2465 device->fs_devices->total_devices);
2466 mutex_unlock(&uuid_mutex);
2468 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2469 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2477 static int btrfs_freeze(struct super_block *sb)
2479 struct btrfs_trans_handle *trans;
2480 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2481 struct btrfs_root *root = fs_info->tree_root;
2483 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2485 * We don't need a barrier here, we'll wait for any transaction that
2486 * could be in progress on other threads (and do delayed iputs that
2487 * we want to avoid on a frozen filesystem), or do the commit
2490 trans = btrfs_attach_transaction_barrier(root);
2491 if (IS_ERR(trans)) {
2492 /* no transaction, don't bother */
2493 if (PTR_ERR(trans) == -ENOENT)
2495 return PTR_ERR(trans);
2497 return btrfs_commit_transaction(trans);
2500 static int check_dev_super(struct btrfs_device *dev)
2502 struct btrfs_fs_info *fs_info = dev->fs_info;
2503 struct btrfs_super_block *sb;
2507 /* This should be called with fs still frozen. */
2508 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2510 /* Missing dev, no need to check. */
2514 /* Only need to check the primary super block. */
2515 sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2519 /* Verify the checksum. */
2520 csum_type = btrfs_super_csum_type(sb);
2521 if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2522 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2523 csum_type, btrfs_super_csum_type(fs_info->super_copy));
2528 if (btrfs_check_super_csum(fs_info, sb)) {
2529 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2534 /* Btrfs_validate_super() includes fsid check against super->fsid. */
2535 ret = btrfs_validate_super(fs_info, sb, 0);
2539 if (btrfs_super_generation(sb) != fs_info->last_trans_committed) {
2540 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2541 btrfs_super_generation(sb),
2542 fs_info->last_trans_committed);
2547 btrfs_release_disk_super(sb);
2551 static int btrfs_unfreeze(struct super_block *sb)
2553 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2554 struct btrfs_device *device;
2558 * Make sure the fs is not changed by accident (like hibernation then
2559 * modified by other OS).
2560 * If we found anything wrong, we mark the fs error immediately.
2562 * And since the fs is frozen, no one can modify the fs yet, thus
2563 * we don't need to hold device_list_mutex.
2565 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2566 ret = check_dev_super(device);
2568 btrfs_handle_fs_error(fs_info, ret,
2569 "super block on devid %llu got modified unexpectedly",
2574 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2577 * We still return 0, to allow VFS layer to unfreeze the fs even the
2578 * above checks failed. Since the fs is either fine or read-only, we're
2579 * safe to continue, without causing further damage.
2584 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2586 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2589 * There should be always a valid pointer in latest_dev, it may be stale
2590 * for a short moment in case it's being deleted but still valid until
2591 * the end of RCU grace period.
2594 seq_escape(m, rcu_str_deref(fs_info->fs_devices->latest_dev->name), " \t\n\\");
2600 static const struct super_operations btrfs_super_ops = {
2601 .drop_inode = btrfs_drop_inode,
2602 .evict_inode = btrfs_evict_inode,
2603 .put_super = btrfs_put_super,
2604 .sync_fs = btrfs_sync_fs,
2605 .show_options = btrfs_show_options,
2606 .show_devname = btrfs_show_devname,
2607 .alloc_inode = btrfs_alloc_inode,
2608 .destroy_inode = btrfs_destroy_inode,
2609 .free_inode = btrfs_free_inode,
2610 .statfs = btrfs_statfs,
2611 .remount_fs = btrfs_remount,
2612 .freeze_fs = btrfs_freeze,
2613 .unfreeze_fs = btrfs_unfreeze,
2616 static const struct file_operations btrfs_ctl_fops = {
2617 .open = btrfs_control_open,
2618 .unlocked_ioctl = btrfs_control_ioctl,
2619 .compat_ioctl = compat_ptr_ioctl,
2620 .owner = THIS_MODULE,
2621 .llseek = noop_llseek,
2624 static struct miscdevice btrfs_misc = {
2625 .minor = BTRFS_MINOR,
2626 .name = "btrfs-control",
2627 .fops = &btrfs_ctl_fops
2630 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2631 MODULE_ALIAS("devname:btrfs-control");
2633 static int __init btrfs_interface_init(void)
2635 return misc_register(&btrfs_misc);
2638 static __cold void btrfs_interface_exit(void)
2640 misc_deregister(&btrfs_misc);
2643 static void __init btrfs_print_mod_info(void)
2645 static const char options[] = ""
2646 #ifdef CONFIG_BTRFS_DEBUG
2649 #ifdef CONFIG_BTRFS_ASSERT
2652 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2653 ", integrity-checker=on"
2655 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2658 #ifdef CONFIG_BLK_DEV_ZONED
2663 #ifdef CONFIG_FS_VERITY
2669 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2672 static int __init init_btrfs_fs(void)
2678 err = btrfs_init_sysfs();
2682 btrfs_init_compress();
2684 err = btrfs_init_cachep();
2688 err = extent_io_init();
2692 err = extent_state_cache_init();
2694 goto free_extent_io;
2696 err = extent_map_init();
2698 goto free_extent_state_cache;
2700 err = ordered_data_init();
2702 goto free_extent_map;
2704 err = btrfs_delayed_inode_init();
2706 goto free_ordered_data;
2708 err = btrfs_auto_defrag_init();
2710 goto free_delayed_inode;
2712 err = btrfs_delayed_ref_init();
2714 goto free_auto_defrag;
2716 err = btrfs_prelim_ref_init();
2718 goto free_delayed_ref;
2720 err = btrfs_end_io_wq_init();
2722 goto free_prelim_ref;
2724 err = btrfs_interface_init();
2726 goto free_end_io_wq;
2728 btrfs_print_mod_info();
2730 err = btrfs_run_sanity_tests();
2732 goto unregister_ioctl;
2734 err = register_filesystem(&btrfs_fs_type);
2736 goto unregister_ioctl;
2741 btrfs_interface_exit();
2743 btrfs_end_io_wq_exit();
2745 btrfs_prelim_ref_exit();
2747 btrfs_delayed_ref_exit();
2749 btrfs_auto_defrag_exit();
2751 btrfs_delayed_inode_exit();
2753 ordered_data_exit();
2756 free_extent_state_cache:
2757 extent_state_cache_exit();
2761 btrfs_destroy_cachep();
2763 btrfs_exit_compress();
2769 static void __exit exit_btrfs_fs(void)
2771 btrfs_destroy_cachep();
2772 btrfs_delayed_ref_exit();
2773 btrfs_auto_defrag_exit();
2774 btrfs_delayed_inode_exit();
2775 btrfs_prelim_ref_exit();
2776 ordered_data_exit();
2778 extent_state_cache_exit();
2780 btrfs_interface_exit();
2781 btrfs_end_io_wq_exit();
2782 unregister_filesystem(&btrfs_fs_type);
2784 btrfs_cleanup_fs_uuids();
2785 btrfs_exit_compress();
2788 late_initcall(init_btrfs_fs);
2789 module_exit(exit_btrfs_fs)
2791 MODULE_LICENSE("GPL");
2792 MODULE_SOFTDEP("pre: crc32c");
2793 MODULE_SOFTDEP("pre: xxhash64");
2794 MODULE_SOFTDEP("pre: sha256");
2795 MODULE_SOFTDEP("pre: blake2b-256");