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/ratelimit.h>
27 #include <linux/crc32c.h>
28 #include <linux/btrfs.h>
29 #include "delayed-inode.h"
32 #include "transaction.h"
33 #include "btrfs_inode.h"
34 #include "print-tree.h"
39 #include "compression.h"
40 #include "rcu-string.h"
41 #include "dev-replace.h"
42 #include "free-space-cache.h"
44 #include "space-info.h"
47 #include "tests/btrfs-tests.h"
48 #include "block-group.h"
51 #define CREATE_TRACE_POINTS
52 #include <trace/events/btrfs.h>
54 static const struct super_operations btrfs_super_ops;
57 * Types for mounting the default subvolume and a subvolume explicitly
58 * requested by subvol=/path. That way the callchain is straightforward and we
59 * don't have to play tricks with the mount options and recursive calls to
62 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
64 static struct file_system_type btrfs_fs_type;
65 static struct file_system_type btrfs_root_fs_type;
67 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
70 * Generally the error codes correspond to their respective errors, but there
71 * are a few special cases.
73 * EUCLEAN: Any sort of corruption that we encounter. The tree-checker for
74 * instance will return EUCLEAN if any of the blocks are corrupted in
75 * a way that is problematic. We want to reserve EUCLEAN for these
76 * sort of corruptions.
78 * EROFS: If we check BTRFS_FS_STATE_ERROR and fail out with a return error, we
79 * need to use EROFS for this case. We will have no idea of the
80 * original failure, that will have been reported at the time we tripped
81 * over the error. Each subsequent error that doesn't have any context
82 * of the original error should use EROFS when handling BTRFS_FS_STATE_ERROR.
84 const char * __attribute_const__ btrfs_decode_error(int errno)
86 char *errstr = "unknown";
89 case -ENOENT: /* -2 */
90 errstr = "No such entry";
93 errstr = "IO failure";
95 case -ENOMEM: /* -12*/
96 errstr = "Out of memory";
98 case -EEXIST: /* -17 */
99 errstr = "Object already exists";
101 case -ENOSPC: /* -28 */
102 errstr = "No space left";
104 case -EROFS: /* -30 */
105 errstr = "Readonly filesystem";
107 case -EOPNOTSUPP: /* -95 */
108 errstr = "Operation not supported";
110 case -EUCLEAN: /* -117 */
111 errstr = "Filesystem corrupted";
113 case -EDQUOT: /* -122 */
114 errstr = "Quota exceeded";
122 * __btrfs_handle_fs_error decodes expected errors from the caller and
123 * invokes the appropriate error response.
126 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
127 unsigned int line, int errno, const char *fmt, ...)
129 struct super_block *sb = fs_info->sb;
135 * Special case: if the error is EROFS, and we're already
136 * under SB_RDONLY, then it is safe here.
138 if (errno == -EROFS && sb_rdonly(sb))
142 errstr = btrfs_decode_error(errno);
144 struct va_format vaf;
151 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
152 sb->s_id, function, line, errno, errstr, &vaf);
155 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
156 sb->s_id, function, line, errno, errstr);
161 * Today we only save the error info to memory. Long term we'll
162 * also send it down to the disk
164 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
166 /* Don't go through full error handling during mount */
167 if (!(sb->s_flags & SB_BORN))
173 btrfs_discard_stop(fs_info);
175 /* btrfs handle error by forcing the filesystem readonly */
176 btrfs_set_sb_rdonly(sb);
177 btrfs_info(fs_info, "forced readonly");
179 * Note that a running device replace operation is not canceled here
180 * although there is no way to update the progress. It would add the
181 * risk of a deadlock, therefore the canceling is omitted. The only
182 * penalty is that some I/O remains active until the procedure
183 * completes. The next time when the filesystem is mounted writable
184 * again, the device replace operation continues.
189 static const char * const logtypes[] = {
202 * Use one ratelimit state per log level so that a flood of less important
203 * messages doesn't cause more important ones to be dropped.
205 static struct ratelimit_state printk_limits[] = {
206 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
207 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
208 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
209 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
210 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
211 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
212 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
213 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
216 void __cold btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
218 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
219 struct va_format vaf;
222 const char *type = logtypes[4];
223 struct ratelimit_state *ratelimit = &printk_limits[4];
227 while ((kern_level = printk_get_level(fmt)) != 0) {
228 size_t size = printk_skip_level(fmt) - fmt;
230 if (kern_level >= '0' && kern_level <= '7') {
231 memcpy(lvl, fmt, size);
233 type = logtypes[kern_level - '0'];
234 ratelimit = &printk_limits[kern_level - '0'];
242 if (__ratelimit(ratelimit)) {
244 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
245 fs_info->sb->s_id, &vaf);
247 printk("%sBTRFS %s: %pV\n", lvl, type, &vaf);
254 #if BITS_PER_LONG == 32
255 void __cold btrfs_warn_32bit_limit(struct btrfs_fs_info *fs_info)
257 if (!test_and_set_bit(BTRFS_FS_32BIT_WARN, &fs_info->flags)) {
258 btrfs_warn(fs_info, "reaching 32bit limit for logical addresses");
260 "due to page cache limit on 32bit systems, btrfs can't access metadata at or beyond %lluT",
261 BTRFS_32BIT_MAX_FILE_SIZE >> 40);
263 "please consider upgrading to 64bit kernel/hardware");
267 void __cold btrfs_err_32bit_limit(struct btrfs_fs_info *fs_info)
269 if (!test_and_set_bit(BTRFS_FS_32BIT_ERROR, &fs_info->flags)) {
270 btrfs_err(fs_info, "reached 32bit limit for logical addresses");
272 "due to page cache limit on 32bit systems, metadata beyond %lluT can't be accessed",
273 BTRFS_32BIT_MAX_FILE_SIZE >> 40);
275 "please consider upgrading to 64bit kernel/hardware");
281 * We only mark the transaction aborted and then set the file system read-only.
282 * This will prevent new transactions from starting or trying to join this
285 * This means that error recovery at the call site is limited to freeing
286 * any local memory allocations and passing the error code up without
287 * further cleanup. The transaction should complete as it normally would
288 * in the call path but will return -EIO.
290 * We'll complete the cleanup in btrfs_end_transaction and
291 * btrfs_commit_transaction.
294 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
295 const char *function,
296 unsigned int line, int errno)
298 struct btrfs_fs_info *fs_info = trans->fs_info;
300 WRITE_ONCE(trans->aborted, errno);
301 WRITE_ONCE(trans->transaction->aborted, errno);
302 /* Wake up anybody who may be waiting on this transaction */
303 wake_up(&fs_info->transaction_wait);
304 wake_up(&fs_info->transaction_blocked_wait);
305 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
308 * __btrfs_panic decodes unexpected, fatal errors from the caller,
309 * issues an alert, and either panics or BUGs, depending on mount options.
312 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
313 unsigned int line, int errno, const char *fmt, ...)
315 char *s_id = "<unknown>";
317 struct va_format vaf = { .fmt = fmt };
321 s_id = fs_info->sb->s_id;
326 errstr = btrfs_decode_error(errno);
327 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
328 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
329 s_id, function, line, &vaf, errno, errstr);
331 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
332 function, line, &vaf, errno, errstr);
334 /* Caller calls BUG() */
337 static void btrfs_put_super(struct super_block *sb)
339 close_ctree(btrfs_sb(sb));
348 Opt_compress_force_type,
353 Opt_flushoncommit, Opt_noflushoncommit,
355 Opt_barrier, Opt_nobarrier,
356 Opt_datacow, Opt_nodatacow,
357 Opt_datasum, Opt_nodatasum,
358 Opt_defrag, Opt_nodefrag,
359 Opt_discard, Opt_nodiscard,
363 Opt_rescan_uuid_tree,
365 Opt_space_cache, Opt_no_space_cache,
366 Opt_space_cache_version,
368 Opt_ssd_spread, Opt_nossd_spread,
373 Opt_treelog, Opt_notreelog,
374 Opt_user_subvol_rm_allowed,
384 /* Deprecated options */
386 Opt_inode_cache, Opt_noinode_cache,
388 /* Debugging options */
390 Opt_check_integrity_including_extent_data,
391 Opt_check_integrity_print_mask,
392 Opt_enospc_debug, Opt_noenospc_debug,
393 #ifdef CONFIG_BTRFS_DEBUG
394 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
396 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
402 static const match_table_t tokens = {
404 {Opt_noacl, "noacl"},
405 {Opt_clear_cache, "clear_cache"},
406 {Opt_commit_interval, "commit=%u"},
407 {Opt_compress, "compress"},
408 {Opt_compress_type, "compress=%s"},
409 {Opt_compress_force, "compress-force"},
410 {Opt_compress_force_type, "compress-force=%s"},
411 {Opt_degraded, "degraded"},
412 {Opt_device, "device=%s"},
413 {Opt_fatal_errors, "fatal_errors=%s"},
414 {Opt_flushoncommit, "flushoncommit"},
415 {Opt_noflushoncommit, "noflushoncommit"},
416 {Opt_inode_cache, "inode_cache"},
417 {Opt_noinode_cache, "noinode_cache"},
418 {Opt_max_inline, "max_inline=%s"},
419 {Opt_barrier, "barrier"},
420 {Opt_nobarrier, "nobarrier"},
421 {Opt_datacow, "datacow"},
422 {Opt_nodatacow, "nodatacow"},
423 {Opt_datasum, "datasum"},
424 {Opt_nodatasum, "nodatasum"},
425 {Opt_defrag, "autodefrag"},
426 {Opt_nodefrag, "noautodefrag"},
427 {Opt_discard, "discard"},
428 {Opt_discard_mode, "discard=%s"},
429 {Opt_nodiscard, "nodiscard"},
430 {Opt_norecovery, "norecovery"},
431 {Opt_ratio, "metadata_ratio=%u"},
432 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
433 {Opt_skip_balance, "skip_balance"},
434 {Opt_space_cache, "space_cache"},
435 {Opt_no_space_cache, "nospace_cache"},
436 {Opt_space_cache_version, "space_cache=%s"},
438 {Opt_nossd, "nossd"},
439 {Opt_ssd_spread, "ssd_spread"},
440 {Opt_nossd_spread, "nossd_spread"},
441 {Opt_subvol, "subvol=%s"},
442 {Opt_subvol_empty, "subvol="},
443 {Opt_subvolid, "subvolid=%s"},
444 {Opt_thread_pool, "thread_pool=%u"},
445 {Opt_treelog, "treelog"},
446 {Opt_notreelog, "notreelog"},
447 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
450 {Opt_rescue, "rescue=%s"},
451 /* Deprecated, with alias rescue=nologreplay */
452 {Opt_nologreplay, "nologreplay"},
453 /* Deprecated, with alias rescue=usebackuproot */
454 {Opt_usebackuproot, "usebackuproot"},
456 /* Deprecated options */
457 {Opt_recovery, "recovery"},
459 /* Debugging options */
460 {Opt_check_integrity, "check_int"},
461 {Opt_check_integrity_including_extent_data, "check_int_data"},
462 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
463 {Opt_enospc_debug, "enospc_debug"},
464 {Opt_noenospc_debug, "noenospc_debug"},
465 #ifdef CONFIG_BTRFS_DEBUG
466 {Opt_fragment_data, "fragment=data"},
467 {Opt_fragment_metadata, "fragment=metadata"},
468 {Opt_fragment_all, "fragment=all"},
470 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
471 {Opt_ref_verify, "ref_verify"},
476 static const match_table_t rescue_tokens = {
477 {Opt_usebackuproot, "usebackuproot"},
478 {Opt_nologreplay, "nologreplay"},
479 {Opt_ignorebadroots, "ignorebadroots"},
480 {Opt_ignorebadroots, "ibadroots"},
481 {Opt_ignoredatacsums, "ignoredatacsums"},
482 {Opt_ignoredatacsums, "idatacsums"},
483 {Opt_rescue_all, "all"},
487 static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
488 const char *opt_name)
490 if (fs_info->mount_opt & opt) {
491 btrfs_err(fs_info, "%s must be used with ro mount option",
498 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
503 substring_t args[MAX_OPT_ARGS];
506 opts = kstrdup(options, GFP_KERNEL);
511 while ((p = strsep(&opts, ":")) != NULL) {
516 token = match_token(p, rescue_tokens, args);
518 case Opt_usebackuproot:
520 "trying to use backup root at mount time");
521 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
523 case Opt_nologreplay:
524 btrfs_set_and_info(info, NOLOGREPLAY,
525 "disabling log replay at mount time");
527 case Opt_ignorebadroots:
528 btrfs_set_and_info(info, IGNOREBADROOTS,
529 "ignoring bad roots");
531 case Opt_ignoredatacsums:
532 btrfs_set_and_info(info, IGNOREDATACSUMS,
533 "ignoring data csums");
536 btrfs_info(info, "enabling all of the rescue options");
537 btrfs_set_and_info(info, IGNOREDATACSUMS,
538 "ignoring data csums");
539 btrfs_set_and_info(info, IGNOREBADROOTS,
540 "ignoring bad roots");
541 btrfs_set_and_info(info, NOLOGREPLAY,
542 "disabling log replay at mount time");
545 btrfs_info(info, "unrecognized rescue option '%s'", p);
559 * Regular mount options parser. Everything that is needed only when
560 * reading in a new superblock is parsed here.
561 * XXX JDM: This needs to be cleaned up for remount.
563 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
564 unsigned long new_flags)
566 substring_t args[MAX_OPT_ARGS];
571 bool compress_force = false;
572 enum btrfs_compression_type saved_compress_type;
573 int saved_compress_level;
574 bool saved_compress_force;
577 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
578 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
579 else if (btrfs_free_space_cache_v1_active(info)) {
580 if (btrfs_is_zoned(info)) {
582 "zoned: clearing existing space cache");
583 btrfs_set_super_cache_generation(info->super_copy, 0);
585 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
590 * Even the options are empty, we still need to do extra check
596 while ((p = strsep(&options, ",")) != NULL) {
601 token = match_token(p, tokens, args);
604 btrfs_info(info, "allowing degraded mounts");
605 btrfs_set_opt(info->mount_opt, DEGRADED);
608 case Opt_subvol_empty:
612 * These are parsed by btrfs_parse_subvol_options or
613 * btrfs_parse_device_options and can be ignored here.
617 btrfs_set_and_info(info, NODATASUM,
618 "setting nodatasum");
621 if (btrfs_test_opt(info, NODATASUM)) {
622 if (btrfs_test_opt(info, NODATACOW))
624 "setting datasum, datacow enabled");
626 btrfs_info(info, "setting datasum");
628 btrfs_clear_opt(info->mount_opt, NODATACOW);
629 btrfs_clear_opt(info->mount_opt, NODATASUM);
632 if (!btrfs_test_opt(info, NODATACOW)) {
633 if (!btrfs_test_opt(info, COMPRESS) ||
634 !btrfs_test_opt(info, FORCE_COMPRESS)) {
636 "setting nodatacow, compression disabled");
638 btrfs_info(info, "setting nodatacow");
641 btrfs_clear_opt(info->mount_opt, COMPRESS);
642 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
643 btrfs_set_opt(info->mount_opt, NODATACOW);
644 btrfs_set_opt(info->mount_opt, NODATASUM);
647 btrfs_clear_and_info(info, NODATACOW,
650 case Opt_compress_force:
651 case Opt_compress_force_type:
652 compress_force = true;
655 case Opt_compress_type:
656 saved_compress_type = btrfs_test_opt(info,
658 info->compress_type : BTRFS_COMPRESS_NONE;
659 saved_compress_force =
660 btrfs_test_opt(info, FORCE_COMPRESS);
661 saved_compress_level = info->compress_level;
662 if (token == Opt_compress ||
663 token == Opt_compress_force ||
664 strncmp(args[0].from, "zlib", 4) == 0) {
665 compress_type = "zlib";
667 info->compress_type = BTRFS_COMPRESS_ZLIB;
668 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
670 * args[0] contains uninitialized data since
671 * for these tokens we don't expect any
674 if (token != Opt_compress &&
675 token != Opt_compress_force)
676 info->compress_level =
677 btrfs_compress_str2level(
680 btrfs_set_opt(info->mount_opt, COMPRESS);
681 btrfs_clear_opt(info->mount_opt, NODATACOW);
682 btrfs_clear_opt(info->mount_opt, NODATASUM);
684 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
685 compress_type = "lzo";
686 info->compress_type = BTRFS_COMPRESS_LZO;
687 info->compress_level = 0;
688 btrfs_set_opt(info->mount_opt, COMPRESS);
689 btrfs_clear_opt(info->mount_opt, NODATACOW);
690 btrfs_clear_opt(info->mount_opt, NODATASUM);
691 btrfs_set_fs_incompat(info, COMPRESS_LZO);
693 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
694 compress_type = "zstd";
695 info->compress_type = BTRFS_COMPRESS_ZSTD;
696 info->compress_level =
697 btrfs_compress_str2level(
700 btrfs_set_opt(info->mount_opt, COMPRESS);
701 btrfs_clear_opt(info->mount_opt, NODATACOW);
702 btrfs_clear_opt(info->mount_opt, NODATASUM);
703 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
705 } else if (strncmp(args[0].from, "no", 2) == 0) {
706 compress_type = "no";
707 info->compress_level = 0;
708 info->compress_type = 0;
709 btrfs_clear_opt(info->mount_opt, COMPRESS);
710 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
711 compress_force = false;
718 if (compress_force) {
719 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
722 * If we remount from compress-force=xxx to
723 * compress=xxx, we need clear FORCE_COMPRESS
724 * flag, otherwise, there is no way for users
725 * to disable forcible compression separately.
727 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
729 if (no_compress == 1) {
730 btrfs_info(info, "use no compression");
731 } else if ((info->compress_type != saved_compress_type) ||
732 (compress_force != saved_compress_force) ||
733 (info->compress_level != saved_compress_level)) {
734 btrfs_info(info, "%s %s compression, level %d",
735 (compress_force) ? "force" : "use",
736 compress_type, info->compress_level);
738 compress_force = false;
741 btrfs_set_and_info(info, SSD,
742 "enabling ssd optimizations");
743 btrfs_clear_opt(info->mount_opt, NOSSD);
746 btrfs_set_and_info(info, SSD,
747 "enabling ssd optimizations");
748 btrfs_set_and_info(info, SSD_SPREAD,
749 "using spread ssd allocation scheme");
750 btrfs_clear_opt(info->mount_opt, NOSSD);
753 btrfs_set_opt(info->mount_opt, NOSSD);
754 btrfs_clear_and_info(info, SSD,
755 "not using ssd optimizations");
757 case Opt_nossd_spread:
758 btrfs_clear_and_info(info, SSD_SPREAD,
759 "not using spread ssd allocation scheme");
762 btrfs_clear_and_info(info, NOBARRIER,
763 "turning on barriers");
766 btrfs_set_and_info(info, NOBARRIER,
767 "turning off barriers");
769 case Opt_thread_pool:
770 ret = match_int(&args[0], &intarg);
773 } else if (intarg == 0) {
777 info->thread_pool_size = intarg;
780 num = match_strdup(&args[0]);
782 info->max_inline = memparse(num, NULL);
785 if (info->max_inline) {
786 info->max_inline = min_t(u64,
790 btrfs_info(info, "max_inline at %llu",
798 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
799 info->sb->s_flags |= SB_POSIXACL;
802 btrfs_err(info, "support for ACL not compiled in!");
807 info->sb->s_flags &= ~SB_POSIXACL;
810 btrfs_set_and_info(info, NOTREELOG,
811 "disabling tree log");
814 btrfs_clear_and_info(info, NOTREELOG,
815 "enabling tree log");
818 case Opt_nologreplay:
820 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
821 btrfs_set_and_info(info, NOLOGREPLAY,
822 "disabling log replay at mount time");
824 case Opt_flushoncommit:
825 btrfs_set_and_info(info, FLUSHONCOMMIT,
826 "turning on flush-on-commit");
828 case Opt_noflushoncommit:
829 btrfs_clear_and_info(info, FLUSHONCOMMIT,
830 "turning off flush-on-commit");
833 ret = match_int(&args[0], &intarg);
836 info->metadata_ratio = intarg;
837 btrfs_info(info, "metadata ratio %u",
838 info->metadata_ratio);
841 case Opt_discard_mode:
842 if (token == Opt_discard ||
843 strcmp(args[0].from, "sync") == 0) {
844 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
845 btrfs_set_and_info(info, DISCARD_SYNC,
846 "turning on sync discard");
847 } else if (strcmp(args[0].from, "async") == 0) {
848 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
849 btrfs_set_and_info(info, DISCARD_ASYNC,
850 "turning on async discard");
857 btrfs_clear_and_info(info, DISCARD_SYNC,
858 "turning off discard");
859 btrfs_clear_and_info(info, DISCARD_ASYNC,
860 "turning off async discard");
862 case Opt_space_cache:
863 case Opt_space_cache_version:
864 if (token == Opt_space_cache ||
865 strcmp(args[0].from, "v1") == 0) {
866 btrfs_clear_opt(info->mount_opt,
868 btrfs_set_and_info(info, SPACE_CACHE,
869 "enabling disk space caching");
870 } else if (strcmp(args[0].from, "v2") == 0) {
871 btrfs_clear_opt(info->mount_opt,
873 btrfs_set_and_info(info, FREE_SPACE_TREE,
874 "enabling free space tree");
880 case Opt_rescan_uuid_tree:
881 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
883 case Opt_no_space_cache:
884 if (btrfs_test_opt(info, SPACE_CACHE)) {
885 btrfs_clear_and_info(info, SPACE_CACHE,
886 "disabling disk space caching");
888 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
889 btrfs_clear_and_info(info, FREE_SPACE_TREE,
890 "disabling free space tree");
893 case Opt_inode_cache:
894 case Opt_noinode_cache:
896 "the 'inode_cache' option is deprecated and has no effect since 5.11");
898 case Opt_clear_cache:
899 btrfs_set_and_info(info, CLEAR_CACHE,
900 "force clearing of disk cache");
902 case Opt_user_subvol_rm_allowed:
903 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
905 case Opt_enospc_debug:
906 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
908 case Opt_noenospc_debug:
909 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
912 btrfs_set_and_info(info, AUTO_DEFRAG,
913 "enabling auto defrag");
916 btrfs_clear_and_info(info, AUTO_DEFRAG,
917 "disabling auto defrag");
920 case Opt_usebackuproot:
922 "'%s' is deprecated, use 'rescue=usebackuproot' instead",
923 token == Opt_recovery ? "recovery" :
926 "trying to use backup root at mount time");
927 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
929 case Opt_skip_balance:
930 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
932 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
933 case Opt_check_integrity_including_extent_data:
935 "enabling check integrity including extent data");
936 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
937 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
939 case Opt_check_integrity:
940 btrfs_info(info, "enabling check integrity");
941 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
943 case Opt_check_integrity_print_mask:
944 ret = match_int(&args[0], &intarg);
947 info->check_integrity_print_mask = intarg;
948 btrfs_info(info, "check_integrity_print_mask 0x%x",
949 info->check_integrity_print_mask);
952 case Opt_check_integrity_including_extent_data:
953 case Opt_check_integrity:
954 case Opt_check_integrity_print_mask:
956 "support for check_integrity* not compiled in!");
960 case Opt_fatal_errors:
961 if (strcmp(args[0].from, "panic") == 0)
962 btrfs_set_opt(info->mount_opt,
963 PANIC_ON_FATAL_ERROR);
964 else if (strcmp(args[0].from, "bug") == 0)
965 btrfs_clear_opt(info->mount_opt,
966 PANIC_ON_FATAL_ERROR);
972 case Opt_commit_interval:
974 ret = match_int(&args[0], &intarg);
979 "using default commit interval %us",
980 BTRFS_DEFAULT_COMMIT_INTERVAL);
981 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
982 } else if (intarg > 300) {
983 btrfs_warn(info, "excessive commit interval %d",
986 info->commit_interval = intarg;
989 ret = parse_rescue_options(info, args[0].from);
993 #ifdef CONFIG_BTRFS_DEBUG
994 case Opt_fragment_all:
995 btrfs_info(info, "fragmenting all space");
996 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
997 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
999 case Opt_fragment_metadata:
1000 btrfs_info(info, "fragmenting metadata");
1001 btrfs_set_opt(info->mount_opt,
1004 case Opt_fragment_data:
1005 btrfs_info(info, "fragmenting data");
1006 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1009 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
1010 case Opt_ref_verify:
1011 btrfs_info(info, "doing ref verification");
1012 btrfs_set_opt(info->mount_opt, REF_VERIFY);
1016 btrfs_err(info, "unrecognized mount option '%s'", p);
1024 /* We're read-only, don't have to check. */
1025 if (new_flags & SB_RDONLY)
1028 if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
1029 check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
1030 check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
1033 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
1034 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
1035 !btrfs_test_opt(info, CLEAR_CACHE)) {
1036 btrfs_err(info, "cannot disable free space tree");
1041 ret = btrfs_check_mountopts_zoned(info);
1042 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
1043 btrfs_info(info, "disk space caching is enabled");
1044 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
1045 btrfs_info(info, "using free space tree");
1050 * Parse mount options that are required early in the mount process.
1052 * All other options will be parsed on much later in the mount process and
1053 * only when we need to allocate a new super block.
1055 static int btrfs_parse_device_options(const char *options, fmode_t flags,
1058 substring_t args[MAX_OPT_ARGS];
1059 char *device_name, *opts, *orig, *p;
1060 struct btrfs_device *device = NULL;
1063 lockdep_assert_held(&uuid_mutex);
1069 * strsep changes the string, duplicate it because btrfs_parse_options
1072 opts = kstrdup(options, GFP_KERNEL);
1077 while ((p = strsep(&opts, ",")) != NULL) {
1083 token = match_token(p, tokens, args);
1084 if (token == Opt_device) {
1085 device_name = match_strdup(&args[0]);
1090 device = btrfs_scan_one_device(device_name, flags,
1093 if (IS_ERR(device)) {
1094 error = PTR_ERR(device);
1106 * Parse mount options that are related to subvolume id
1108 * The value is later passed to mount_subvol()
1110 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1111 u64 *subvol_objectid)
1113 substring_t args[MAX_OPT_ARGS];
1114 char *opts, *orig, *p;
1122 * strsep changes the string, duplicate it because
1123 * btrfs_parse_device_options gets called later
1125 opts = kstrdup(options, GFP_KERNEL);
1130 while ((p = strsep(&opts, ",")) != NULL) {
1135 token = match_token(p, tokens, args);
1138 kfree(*subvol_name);
1139 *subvol_name = match_strdup(&args[0]);
1140 if (!*subvol_name) {
1146 error = match_u64(&args[0], &subvolid);
1150 /* we want the original fs_tree */
1152 subvolid = BTRFS_FS_TREE_OBJECTID;
1154 *subvol_objectid = subvolid;
1166 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1167 u64 subvol_objectid)
1169 struct btrfs_root *root = fs_info->tree_root;
1170 struct btrfs_root *fs_root = NULL;
1171 struct btrfs_root_ref *root_ref;
1172 struct btrfs_inode_ref *inode_ref;
1173 struct btrfs_key key;
1174 struct btrfs_path *path = NULL;
1175 char *name = NULL, *ptr;
1180 path = btrfs_alloc_path();
1186 name = kmalloc(PATH_MAX, GFP_KERNEL);
1191 ptr = name + PATH_MAX - 1;
1195 * Walk up the subvolume trees in the tree of tree roots by root
1196 * backrefs until we hit the top-level subvolume.
1198 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1199 key.objectid = subvol_objectid;
1200 key.type = BTRFS_ROOT_BACKREF_KEY;
1201 key.offset = (u64)-1;
1203 ret = btrfs_search_backwards(root, &key, path);
1206 } else if (ret > 0) {
1211 subvol_objectid = key.offset;
1213 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1214 struct btrfs_root_ref);
1215 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1218 ret = -ENAMETOOLONG;
1221 read_extent_buffer(path->nodes[0], ptr + 1,
1222 (unsigned long)(root_ref + 1), len);
1224 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1225 btrfs_release_path(path);
1227 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1228 if (IS_ERR(fs_root)) {
1229 ret = PTR_ERR(fs_root);
1235 * Walk up the filesystem tree by inode refs until we hit the
1238 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1239 key.objectid = dirid;
1240 key.type = BTRFS_INODE_REF_KEY;
1241 key.offset = (u64)-1;
1243 ret = btrfs_search_backwards(fs_root, &key, path);
1246 } else if (ret > 0) {
1253 inode_ref = btrfs_item_ptr(path->nodes[0],
1255 struct btrfs_inode_ref);
1256 len = btrfs_inode_ref_name_len(path->nodes[0],
1260 ret = -ENAMETOOLONG;
1263 read_extent_buffer(path->nodes[0], ptr + 1,
1264 (unsigned long)(inode_ref + 1), len);
1266 btrfs_release_path(path);
1268 btrfs_put_root(fs_root);
1272 btrfs_free_path(path);
1273 if (ptr == name + PATH_MAX - 1) {
1277 memmove(name, ptr, name + PATH_MAX - ptr);
1282 btrfs_put_root(fs_root);
1283 btrfs_free_path(path);
1285 return ERR_PTR(ret);
1288 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1290 struct btrfs_root *root = fs_info->tree_root;
1291 struct btrfs_dir_item *di;
1292 struct btrfs_path *path;
1293 struct btrfs_key location;
1296 path = btrfs_alloc_path();
1301 * Find the "default" dir item which points to the root item that we
1302 * will mount by default if we haven't been given a specific subvolume
1305 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1306 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1308 btrfs_free_path(path);
1313 * Ok the default dir item isn't there. This is weird since
1314 * it's always been there, but don't freak out, just try and
1315 * mount the top-level subvolume.
1317 btrfs_free_path(path);
1318 *objectid = BTRFS_FS_TREE_OBJECTID;
1322 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1323 btrfs_free_path(path);
1324 *objectid = location.objectid;
1328 static int btrfs_fill_super(struct super_block *sb,
1329 struct btrfs_fs_devices *fs_devices,
1332 struct inode *inode;
1333 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1336 sb->s_maxbytes = MAX_LFS_FILESIZE;
1337 sb->s_magic = BTRFS_SUPER_MAGIC;
1338 sb->s_op = &btrfs_super_ops;
1339 sb->s_d_op = &btrfs_dentry_operations;
1340 sb->s_export_op = &btrfs_export_ops;
1341 #ifdef CONFIG_FS_VERITY
1342 sb->s_vop = &btrfs_verityops;
1344 sb->s_xattr = btrfs_xattr_handlers;
1345 sb->s_time_gran = 1;
1346 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1347 sb->s_flags |= SB_POSIXACL;
1349 sb->s_flags |= SB_I_VERSION;
1350 sb->s_iflags |= SB_I_CGROUPWB;
1352 err = super_setup_bdi(sb);
1354 btrfs_err(fs_info, "super_setup_bdi failed");
1358 err = open_ctree(sb, fs_devices, (char *)data);
1360 btrfs_err(fs_info, "open_ctree failed");
1364 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1365 if (IS_ERR(inode)) {
1366 err = PTR_ERR(inode);
1370 sb->s_root = d_make_root(inode);
1376 sb->s_flags |= SB_ACTIVE;
1380 close_ctree(fs_info);
1384 int btrfs_sync_fs(struct super_block *sb, int wait)
1386 struct btrfs_trans_handle *trans;
1387 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1388 struct btrfs_root *root = fs_info->tree_root;
1390 trace_btrfs_sync_fs(fs_info, wait);
1393 filemap_flush(fs_info->btree_inode->i_mapping);
1397 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1399 trans = btrfs_attach_transaction_barrier(root);
1400 if (IS_ERR(trans)) {
1401 /* no transaction, don't bother */
1402 if (PTR_ERR(trans) == -ENOENT) {
1404 * Exit unless we have some pending changes
1405 * that need to go through commit
1407 if (fs_info->pending_changes == 0)
1410 * A non-blocking test if the fs is frozen. We must not
1411 * start a new transaction here otherwise a deadlock
1412 * happens. The pending operations are delayed to the
1413 * next commit after thawing.
1415 if (sb_start_write_trylock(sb))
1419 trans = btrfs_start_transaction(root, 0);
1422 return PTR_ERR(trans);
1424 return btrfs_commit_transaction(trans);
1427 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1429 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1433 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1435 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1436 const char *compress_type;
1437 const char *subvol_name;
1438 bool printed = false;
1440 if (btrfs_test_opt(info, DEGRADED))
1441 seq_puts(seq, ",degraded");
1442 if (btrfs_test_opt(info, NODATASUM))
1443 seq_puts(seq, ",nodatasum");
1444 if (btrfs_test_opt(info, NODATACOW))
1445 seq_puts(seq, ",nodatacow");
1446 if (btrfs_test_opt(info, NOBARRIER))
1447 seq_puts(seq, ",nobarrier");
1448 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1449 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1450 if (info->thread_pool_size != min_t(unsigned long,
1451 num_online_cpus() + 2, 8))
1452 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1453 if (btrfs_test_opt(info, COMPRESS)) {
1454 compress_type = btrfs_compress_type2str(info->compress_type);
1455 if (btrfs_test_opt(info, FORCE_COMPRESS))
1456 seq_printf(seq, ",compress-force=%s", compress_type);
1458 seq_printf(seq, ",compress=%s", compress_type);
1459 if (info->compress_level)
1460 seq_printf(seq, ":%d", info->compress_level);
1462 if (btrfs_test_opt(info, NOSSD))
1463 seq_puts(seq, ",nossd");
1464 if (btrfs_test_opt(info, SSD_SPREAD))
1465 seq_puts(seq, ",ssd_spread");
1466 else if (btrfs_test_opt(info, SSD))
1467 seq_puts(seq, ",ssd");
1468 if (btrfs_test_opt(info, NOTREELOG))
1469 seq_puts(seq, ",notreelog");
1470 if (btrfs_test_opt(info, NOLOGREPLAY))
1471 print_rescue_option(seq, "nologreplay", &printed);
1472 if (btrfs_test_opt(info, USEBACKUPROOT))
1473 print_rescue_option(seq, "usebackuproot", &printed);
1474 if (btrfs_test_opt(info, IGNOREBADROOTS))
1475 print_rescue_option(seq, "ignorebadroots", &printed);
1476 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1477 print_rescue_option(seq, "ignoredatacsums", &printed);
1478 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1479 seq_puts(seq, ",flushoncommit");
1480 if (btrfs_test_opt(info, DISCARD_SYNC))
1481 seq_puts(seq, ",discard");
1482 if (btrfs_test_opt(info, DISCARD_ASYNC))
1483 seq_puts(seq, ",discard=async");
1484 if (!(info->sb->s_flags & SB_POSIXACL))
1485 seq_puts(seq, ",noacl");
1486 if (btrfs_free_space_cache_v1_active(info))
1487 seq_puts(seq, ",space_cache");
1488 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1489 seq_puts(seq, ",space_cache=v2");
1491 seq_puts(seq, ",nospace_cache");
1492 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1493 seq_puts(seq, ",rescan_uuid_tree");
1494 if (btrfs_test_opt(info, CLEAR_CACHE))
1495 seq_puts(seq, ",clear_cache");
1496 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1497 seq_puts(seq, ",user_subvol_rm_allowed");
1498 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1499 seq_puts(seq, ",enospc_debug");
1500 if (btrfs_test_opt(info, AUTO_DEFRAG))
1501 seq_puts(seq, ",autodefrag");
1502 if (btrfs_test_opt(info, SKIP_BALANCE))
1503 seq_puts(seq, ",skip_balance");
1504 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1505 if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1506 seq_puts(seq, ",check_int_data");
1507 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1508 seq_puts(seq, ",check_int");
1509 if (info->check_integrity_print_mask)
1510 seq_printf(seq, ",check_int_print_mask=%d",
1511 info->check_integrity_print_mask);
1513 if (info->metadata_ratio)
1514 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1515 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1516 seq_puts(seq, ",fatal_errors=panic");
1517 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1518 seq_printf(seq, ",commit=%u", info->commit_interval);
1519 #ifdef CONFIG_BTRFS_DEBUG
1520 if (btrfs_test_opt(info, FRAGMENT_DATA))
1521 seq_puts(seq, ",fragment=data");
1522 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1523 seq_puts(seq, ",fragment=metadata");
1525 if (btrfs_test_opt(info, REF_VERIFY))
1526 seq_puts(seq, ",ref_verify");
1527 seq_printf(seq, ",subvolid=%llu",
1528 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1529 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1530 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1531 if (!IS_ERR(subvol_name)) {
1532 seq_puts(seq, ",subvol=");
1533 seq_escape(seq, subvol_name, " \t\n\\");
1539 static int btrfs_test_super(struct super_block *s, void *data)
1541 struct btrfs_fs_info *p = data;
1542 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1544 return fs_info->fs_devices == p->fs_devices;
1547 static int btrfs_set_super(struct super_block *s, void *data)
1549 int err = set_anon_super(s, data);
1551 s->s_fs_info = data;
1556 * subvolumes are identified by ino 256
1558 static inline int is_subvolume_inode(struct inode *inode)
1560 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1565 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1566 struct vfsmount *mnt)
1568 struct dentry *root;
1572 if (!subvol_objectid) {
1573 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1576 root = ERR_PTR(ret);
1580 subvol_name = btrfs_get_subvol_name_from_objectid(
1581 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1582 if (IS_ERR(subvol_name)) {
1583 root = ERR_CAST(subvol_name);
1590 root = mount_subtree(mnt, subvol_name);
1591 /* mount_subtree() drops our reference on the vfsmount. */
1594 if (!IS_ERR(root)) {
1595 struct super_block *s = root->d_sb;
1596 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1597 struct inode *root_inode = d_inode(root);
1598 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1601 if (!is_subvolume_inode(root_inode)) {
1602 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1606 if (subvol_objectid && root_objectid != subvol_objectid) {
1608 * This will also catch a race condition where a
1609 * subvolume which was passed by ID is renamed and
1610 * another subvolume is renamed over the old location.
1613 "subvol '%s' does not match subvolid %llu",
1614 subvol_name, subvol_objectid);
1619 root = ERR_PTR(ret);
1620 deactivate_locked_super(s);
1631 * Find a superblock for the given device / mount point.
1633 * Note: This is based on mount_bdev from fs/super.c with a few additions
1634 * for multiple device setup. Make sure to keep it in sync.
1636 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1637 int flags, const char *device_name, void *data)
1639 struct block_device *bdev = NULL;
1640 struct super_block *s;
1641 struct btrfs_device *device = NULL;
1642 struct btrfs_fs_devices *fs_devices = NULL;
1643 struct btrfs_fs_info *fs_info = NULL;
1644 void *new_sec_opts = NULL;
1645 fmode_t mode = FMODE_READ;
1648 if (!(flags & SB_RDONLY))
1649 mode |= FMODE_WRITE;
1652 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1654 return ERR_PTR(error);
1658 * Setup a dummy root and fs_info for test/set super. This is because
1659 * we don't actually fill this stuff out until open_ctree, but we need
1660 * then open_ctree will properly initialize the file system specific
1661 * settings later. btrfs_init_fs_info initializes the static elements
1662 * of the fs_info (locks and such) to make cleanup easier if we find a
1663 * superblock with our given fs_devices later on at sget() time.
1665 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1668 goto error_sec_opts;
1670 btrfs_init_fs_info(fs_info);
1672 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1673 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1674 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1679 mutex_lock(&uuid_mutex);
1680 error = btrfs_parse_device_options(data, mode, fs_type);
1682 mutex_unlock(&uuid_mutex);
1686 device = btrfs_scan_one_device(device_name, mode, fs_type);
1687 if (IS_ERR(device)) {
1688 mutex_unlock(&uuid_mutex);
1689 error = PTR_ERR(device);
1693 fs_devices = device->fs_devices;
1694 fs_info->fs_devices = fs_devices;
1696 error = btrfs_open_devices(fs_devices, mode, fs_type);
1697 mutex_unlock(&uuid_mutex);
1701 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1703 goto error_close_devices;
1706 bdev = fs_devices->latest_dev->bdev;
1707 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1711 goto error_close_devices;
1715 btrfs_close_devices(fs_devices);
1716 btrfs_free_fs_info(fs_info);
1717 if ((flags ^ s->s_flags) & SB_RDONLY)
1720 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1721 btrfs_sb(s)->bdev_holder = fs_type;
1722 if (!strstr(crc32c_impl(), "generic"))
1723 set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1724 error = btrfs_fill_super(s, fs_devices, data);
1727 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1728 security_free_mnt_opts(&new_sec_opts);
1730 deactivate_locked_super(s);
1731 return ERR_PTR(error);
1734 return dget(s->s_root);
1736 error_close_devices:
1737 btrfs_close_devices(fs_devices);
1739 btrfs_free_fs_info(fs_info);
1741 security_free_mnt_opts(&new_sec_opts);
1742 return ERR_PTR(error);
1746 * Mount function which is called by VFS layer.
1748 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1749 * which needs vfsmount* of device's root (/). This means device's root has to
1750 * be mounted internally in any case.
1753 * 1. Parse subvol id related options for later use in mount_subvol().
1755 * 2. Mount device's root (/) by calling vfs_kern_mount().
1757 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1758 * first place. In order to avoid calling btrfs_mount() again, we use
1759 * different file_system_type which is not registered to VFS by
1760 * register_filesystem() (btrfs_root_fs_type). As a result,
1761 * btrfs_mount_root() is called. The return value will be used by
1762 * mount_subtree() in mount_subvol().
1764 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1765 * "btrfs subvolume set-default", mount_subvol() is called always.
1767 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1768 const char *device_name, void *data)
1770 struct vfsmount *mnt_root;
1771 struct dentry *root;
1772 char *subvol_name = NULL;
1773 u64 subvol_objectid = 0;
1776 error = btrfs_parse_subvol_options(data, &subvol_name,
1780 return ERR_PTR(error);
1783 /* mount device's root (/) */
1784 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1785 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1786 if (flags & SB_RDONLY) {
1787 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1788 flags & ~SB_RDONLY, device_name, data);
1790 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1791 flags | SB_RDONLY, device_name, data);
1792 if (IS_ERR(mnt_root)) {
1793 root = ERR_CAST(mnt_root);
1798 down_write(&mnt_root->mnt_sb->s_umount);
1799 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1800 up_write(&mnt_root->mnt_sb->s_umount);
1802 root = ERR_PTR(error);
1809 if (IS_ERR(mnt_root)) {
1810 root = ERR_CAST(mnt_root);
1815 /* mount_subvol() will free subvol_name and mnt_root */
1816 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1822 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1823 u32 new_pool_size, u32 old_pool_size)
1825 if (new_pool_size == old_pool_size)
1828 fs_info->thread_pool_size = new_pool_size;
1830 btrfs_info(fs_info, "resize thread pool %d -> %d",
1831 old_pool_size, new_pool_size);
1833 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1834 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1835 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1836 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1837 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1838 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1840 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1841 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1842 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1843 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1847 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1848 unsigned long old_opts, int flags)
1850 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1851 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1852 (flags & SB_RDONLY))) {
1853 /* wait for any defraggers to finish */
1854 wait_event(fs_info->transaction_wait,
1855 (atomic_read(&fs_info->defrag_running) == 0));
1856 if (flags & SB_RDONLY)
1857 sync_filesystem(fs_info->sb);
1861 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1862 unsigned long old_opts)
1864 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1867 * We need to cleanup all defragable inodes if the autodefragment is
1868 * close or the filesystem is read only.
1870 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1871 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1872 btrfs_cleanup_defrag_inodes(fs_info);
1875 /* If we toggled discard async */
1876 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1877 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1878 btrfs_discard_resume(fs_info);
1879 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1880 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1881 btrfs_discard_cleanup(fs_info);
1883 /* If we toggled space cache */
1884 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1885 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1888 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1890 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1891 unsigned old_flags = sb->s_flags;
1892 unsigned long old_opts = fs_info->mount_opt;
1893 unsigned long old_compress_type = fs_info->compress_type;
1894 u64 old_max_inline = fs_info->max_inline;
1895 u32 old_thread_pool_size = fs_info->thread_pool_size;
1896 u32 old_metadata_ratio = fs_info->metadata_ratio;
1899 sync_filesystem(sb);
1900 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1903 void *new_sec_opts = NULL;
1905 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1907 ret = security_sb_remount(sb, new_sec_opts);
1908 security_free_mnt_opts(&new_sec_opts);
1913 ret = btrfs_parse_options(fs_info, data, *flags);
1917 btrfs_remount_begin(fs_info, old_opts, *flags);
1918 btrfs_resize_thread_pool(fs_info,
1919 fs_info->thread_pool_size, old_thread_pool_size);
1921 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1922 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1923 (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1925 "remount supports changing free space tree only from ro to rw");
1926 /* Make sure free space cache options match the state on disk */
1927 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1928 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1929 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1931 if (btrfs_free_space_cache_v1_active(fs_info)) {
1932 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1933 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1937 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1940 if (*flags & SB_RDONLY) {
1942 * this also happens on 'umount -rf' or on shutdown, when
1943 * the filesystem is busy.
1945 cancel_work_sync(&fs_info->async_reclaim_work);
1946 cancel_work_sync(&fs_info->async_data_reclaim_work);
1948 btrfs_discard_cleanup(fs_info);
1950 /* wait for the uuid_scan task to finish */
1951 down(&fs_info->uuid_tree_rescan_sem);
1952 /* avoid complains from lockdep et al. */
1953 up(&fs_info->uuid_tree_rescan_sem);
1955 btrfs_set_sb_rdonly(sb);
1958 * Setting SB_RDONLY will put the cleaner thread to
1959 * sleep at the next loop if it's already active.
1960 * If it's already asleep, we'll leave unused block
1961 * groups on disk until we're mounted read-write again
1962 * unless we clean them up here.
1964 btrfs_delete_unused_bgs(fs_info);
1967 * The cleaner task could be already running before we set the
1968 * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
1969 * We must make sure that after we finish the remount, i.e. after
1970 * we call btrfs_commit_super(), the cleaner can no longer start
1971 * a transaction - either because it was dropping a dead root,
1972 * running delayed iputs or deleting an unused block group (the
1973 * cleaner picked a block group from the list of unused block
1974 * groups before we were able to in the previous call to
1975 * btrfs_delete_unused_bgs()).
1977 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
1978 TASK_UNINTERRUPTIBLE);
1981 * We've set the superblock to RO mode, so we might have made
1982 * the cleaner task sleep without running all pending delayed
1983 * iputs. Go through all the delayed iputs here, so that if an
1984 * unmount happens without remounting RW we don't end up at
1985 * finishing close_ctree() with a non-empty list of delayed
1988 btrfs_run_delayed_iputs(fs_info);
1990 btrfs_dev_replace_suspend_for_unmount(fs_info);
1991 btrfs_scrub_cancel(fs_info);
1992 btrfs_pause_balance(fs_info);
1995 * Pause the qgroup rescan worker if it is running. We don't want
1996 * it to be still running after we are in RO mode, as after that,
1997 * by the time we unmount, it might have left a transaction open,
1998 * so we would leak the transaction and/or crash.
2000 btrfs_qgroup_wait_for_completion(fs_info, false);
2002 ret = btrfs_commit_super(fs_info);
2006 if (BTRFS_FS_ERROR(fs_info)) {
2008 "Remounting read-write after error is not allowed");
2012 if (fs_info->fs_devices->rw_devices == 0) {
2017 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
2019 "too many missing devices, writable remount is not allowed");
2024 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
2026 "mount required to replay tree-log, cannot remount read-write");
2032 * NOTE: when remounting with a change that does writes, don't
2033 * put it anywhere above this point, as we are not sure to be
2034 * safe to write until we pass the above checks.
2036 ret = btrfs_start_pre_rw_mount(fs_info);
2040 btrfs_clear_sb_rdonly(sb);
2042 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
2046 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
2047 * since the absence of the flag means it can be toggled off by remount.
2049 *flags |= SB_I_VERSION;
2051 wake_up_process(fs_info->transaction_kthread);
2052 btrfs_remount_cleanup(fs_info, old_opts);
2053 btrfs_clear_oneshot_options(fs_info);
2054 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2059 /* We've hit an error - don't reset SB_RDONLY */
2061 old_flags |= SB_RDONLY;
2062 if (!(old_flags & SB_RDONLY))
2063 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
2064 sb->s_flags = old_flags;
2065 fs_info->mount_opt = old_opts;
2066 fs_info->compress_type = old_compress_type;
2067 fs_info->max_inline = old_max_inline;
2068 btrfs_resize_thread_pool(fs_info,
2069 old_thread_pool_size, fs_info->thread_pool_size);
2070 fs_info->metadata_ratio = old_metadata_ratio;
2071 btrfs_remount_cleanup(fs_info, old_opts);
2072 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2077 /* Used to sort the devices by max_avail(descending sort) */
2078 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
2080 const struct btrfs_device_info *dev_info1 = a;
2081 const struct btrfs_device_info *dev_info2 = b;
2083 if (dev_info1->max_avail > dev_info2->max_avail)
2085 else if (dev_info1->max_avail < dev_info2->max_avail)
2091 * sort the devices by max_avail, in which max free extent size of each device
2092 * is stored.(Descending Sort)
2094 static inline void btrfs_descending_sort_devices(
2095 struct btrfs_device_info *devices,
2098 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2099 btrfs_cmp_device_free_bytes, NULL);
2103 * The helper to calc the free space on the devices that can be used to store
2106 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2109 struct btrfs_device_info *devices_info;
2110 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2111 struct btrfs_device *device;
2114 u64 min_stripe_size;
2115 int num_stripes = 1;
2116 int i = 0, nr_devices;
2117 const struct btrfs_raid_attr *rattr;
2120 * We aren't under the device list lock, so this is racy-ish, but good
2121 * enough for our purposes.
2123 nr_devices = fs_info->fs_devices->open_devices;
2126 nr_devices = fs_info->fs_devices->open_devices;
2134 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2139 /* calc min stripe number for data space allocation */
2140 type = btrfs_data_alloc_profile(fs_info);
2141 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2143 if (type & BTRFS_BLOCK_GROUP_RAID0)
2144 num_stripes = nr_devices;
2145 else if (type & BTRFS_BLOCK_GROUP_RAID1)
2147 else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
2149 else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
2151 else if (type & BTRFS_BLOCK_GROUP_RAID10)
2154 /* Adjust for more than 1 stripe per device */
2155 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2158 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2159 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2160 &device->dev_state) ||
2162 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2165 if (i >= nr_devices)
2168 avail_space = device->total_bytes - device->bytes_used;
2170 /* align with stripe_len */
2171 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2174 * In order to avoid overwriting the superblock on the drive,
2175 * btrfs starts at an offset of at least 1MB when doing chunk
2178 * This ensures we have at least min_stripe_size free space
2179 * after excluding 1MB.
2181 if (avail_space <= SZ_1M + min_stripe_size)
2184 avail_space -= SZ_1M;
2186 devices_info[i].dev = device;
2187 devices_info[i].max_avail = avail_space;
2195 btrfs_descending_sort_devices(devices_info, nr_devices);
2199 while (nr_devices >= rattr->devs_min) {
2200 num_stripes = min(num_stripes, nr_devices);
2202 if (devices_info[i].max_avail >= min_stripe_size) {
2206 avail_space += devices_info[i].max_avail * num_stripes;
2207 alloc_size = devices_info[i].max_avail;
2208 for (j = i + 1 - num_stripes; j <= i; j++)
2209 devices_info[j].max_avail -= alloc_size;
2215 kfree(devices_info);
2216 *free_bytes = avail_space;
2221 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2223 * If there's a redundant raid level at DATA block groups, use the respective
2224 * multiplier to scale the sizes.
2226 * Unused device space usage is based on simulating the chunk allocator
2227 * algorithm that respects the device sizes and order of allocations. This is
2228 * a close approximation of the actual use but there are other factors that may
2229 * change the result (like a new metadata chunk).
2231 * If metadata is exhausted, f_bavail will be 0.
2233 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2235 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2236 struct btrfs_super_block *disk_super = fs_info->super_copy;
2237 struct btrfs_space_info *found;
2239 u64 total_free_data = 0;
2240 u64 total_free_meta = 0;
2241 u32 bits = fs_info->sectorsize_bits;
2242 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2243 unsigned factor = 1;
2244 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2249 list_for_each_entry(found, &fs_info->space_info, list) {
2250 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2253 total_free_data += found->disk_total - found->disk_used;
2255 btrfs_account_ro_block_groups_free_space(found);
2257 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2258 if (!list_empty(&found->block_groups[i]))
2259 factor = btrfs_bg_type_to_factor(
2260 btrfs_raid_array[i].bg_flag);
2265 * Metadata in mixed block goup profiles are accounted in data
2267 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2268 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2271 total_free_meta += found->disk_total -
2275 total_used += found->disk_used;
2278 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2279 buf->f_blocks >>= bits;
2280 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2282 /* Account global block reserve as used, it's in logical size already */
2283 spin_lock(&block_rsv->lock);
2284 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2285 if (buf->f_bfree >= block_rsv->size >> bits)
2286 buf->f_bfree -= block_rsv->size >> bits;
2289 spin_unlock(&block_rsv->lock);
2291 buf->f_bavail = div_u64(total_free_data, factor);
2292 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2295 buf->f_bavail += div_u64(total_free_data, factor);
2296 buf->f_bavail = buf->f_bavail >> bits;
2299 * We calculate the remaining metadata space minus global reserve. If
2300 * this is (supposedly) smaller than zero, there's no space. But this
2301 * does not hold in practice, the exhausted state happens where's still
2302 * some positive delta. So we apply some guesswork and compare the
2303 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2305 * We probably cannot calculate the exact threshold value because this
2306 * depends on the internal reservations requested by various
2307 * operations, so some operations that consume a few metadata will
2308 * succeed even if the Avail is zero. But this is better than the other
2314 * We only want to claim there's no available space if we can no longer
2315 * allocate chunks for our metadata profile and our global reserve will
2316 * not fit in the free metadata space. If we aren't ->full then we
2317 * still can allocate chunks and thus are fine using the currently
2318 * calculated f_bavail.
2320 if (!mixed && block_rsv->space_info->full &&
2321 total_free_meta - thresh < block_rsv->size)
2324 buf->f_type = BTRFS_SUPER_MAGIC;
2325 buf->f_bsize = dentry->d_sb->s_blocksize;
2326 buf->f_namelen = BTRFS_NAME_LEN;
2328 /* We treat it as constant endianness (it doesn't matter _which_)
2329 because we want the fsid to come out the same whether mounted
2330 on a big-endian or little-endian host */
2331 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2332 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2333 /* Mask in the root object ID too, to disambiguate subvols */
2334 buf->f_fsid.val[0] ^=
2335 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2336 buf->f_fsid.val[1] ^=
2337 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2342 static void btrfs_kill_super(struct super_block *sb)
2344 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2345 kill_anon_super(sb);
2346 btrfs_free_fs_info(fs_info);
2349 static struct file_system_type btrfs_fs_type = {
2350 .owner = THIS_MODULE,
2352 .mount = btrfs_mount,
2353 .kill_sb = btrfs_kill_super,
2354 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2357 static struct file_system_type btrfs_root_fs_type = {
2358 .owner = THIS_MODULE,
2360 .mount = btrfs_mount_root,
2361 .kill_sb = btrfs_kill_super,
2362 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2365 MODULE_ALIAS_FS("btrfs");
2367 static int btrfs_control_open(struct inode *inode, struct file *file)
2370 * The control file's private_data is used to hold the
2371 * transaction when it is started and is used to keep
2372 * track of whether a transaction is already in progress.
2374 file->private_data = NULL;
2379 * Used by /dev/btrfs-control for devices ioctls.
2381 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2384 struct btrfs_ioctl_vol_args *vol;
2385 struct btrfs_device *device = NULL;
2388 if (!capable(CAP_SYS_ADMIN))
2391 vol = memdup_user((void __user *)arg, sizeof(*vol));
2393 return PTR_ERR(vol);
2394 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2397 case BTRFS_IOC_SCAN_DEV:
2398 mutex_lock(&uuid_mutex);
2399 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2400 &btrfs_root_fs_type);
2401 ret = PTR_ERR_OR_ZERO(device);
2402 mutex_unlock(&uuid_mutex);
2404 case BTRFS_IOC_FORGET_DEV:
2405 ret = btrfs_forget_devices(vol->name);
2407 case BTRFS_IOC_DEVICES_READY:
2408 mutex_lock(&uuid_mutex);
2409 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2410 &btrfs_root_fs_type);
2411 if (IS_ERR(device)) {
2412 mutex_unlock(&uuid_mutex);
2413 ret = PTR_ERR(device);
2416 ret = !(device->fs_devices->num_devices ==
2417 device->fs_devices->total_devices);
2418 mutex_unlock(&uuid_mutex);
2420 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2421 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2429 static int btrfs_freeze(struct super_block *sb)
2431 struct btrfs_trans_handle *trans;
2432 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2433 struct btrfs_root *root = fs_info->tree_root;
2435 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2437 * We don't need a barrier here, we'll wait for any transaction that
2438 * could be in progress on other threads (and do delayed iputs that
2439 * we want to avoid on a frozen filesystem), or do the commit
2442 trans = btrfs_attach_transaction_barrier(root);
2443 if (IS_ERR(trans)) {
2444 /* no transaction, don't bother */
2445 if (PTR_ERR(trans) == -ENOENT)
2447 return PTR_ERR(trans);
2449 return btrfs_commit_transaction(trans);
2452 static int btrfs_unfreeze(struct super_block *sb)
2454 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2456 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2460 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2462 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2465 * There should be always a valid pointer in latest_dev, it may be stale
2466 * for a short moment in case it's being deleted but still valid until
2467 * the end of RCU grace period.
2470 seq_escape(m, rcu_str_deref(fs_info->fs_devices->latest_dev->name), " \t\n\\");
2476 static const struct super_operations btrfs_super_ops = {
2477 .drop_inode = btrfs_drop_inode,
2478 .evict_inode = btrfs_evict_inode,
2479 .put_super = btrfs_put_super,
2480 .sync_fs = btrfs_sync_fs,
2481 .show_options = btrfs_show_options,
2482 .show_devname = btrfs_show_devname,
2483 .alloc_inode = btrfs_alloc_inode,
2484 .destroy_inode = btrfs_destroy_inode,
2485 .free_inode = btrfs_free_inode,
2486 .statfs = btrfs_statfs,
2487 .remount_fs = btrfs_remount,
2488 .freeze_fs = btrfs_freeze,
2489 .unfreeze_fs = btrfs_unfreeze,
2492 static const struct file_operations btrfs_ctl_fops = {
2493 .open = btrfs_control_open,
2494 .unlocked_ioctl = btrfs_control_ioctl,
2495 .compat_ioctl = compat_ptr_ioctl,
2496 .owner = THIS_MODULE,
2497 .llseek = noop_llseek,
2500 static struct miscdevice btrfs_misc = {
2501 .minor = BTRFS_MINOR,
2502 .name = "btrfs-control",
2503 .fops = &btrfs_ctl_fops
2506 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2507 MODULE_ALIAS("devname:btrfs-control");
2509 static int __init btrfs_interface_init(void)
2511 return misc_register(&btrfs_misc);
2514 static __cold void btrfs_interface_exit(void)
2516 misc_deregister(&btrfs_misc);
2519 static void __init btrfs_print_mod_info(void)
2521 static const char options[] = ""
2522 #ifdef CONFIG_BTRFS_DEBUG
2525 #ifdef CONFIG_BTRFS_ASSERT
2528 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2529 ", integrity-checker=on"
2531 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2534 #ifdef CONFIG_BLK_DEV_ZONED
2539 #ifdef CONFIG_FS_VERITY
2545 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2548 static int __init init_btrfs_fs(void)
2554 err = btrfs_init_sysfs();
2558 btrfs_init_compress();
2560 err = btrfs_init_cachep();
2564 err = extent_io_init();
2568 err = extent_state_cache_init();
2570 goto free_extent_io;
2572 err = extent_map_init();
2574 goto free_extent_state_cache;
2576 err = ordered_data_init();
2578 goto free_extent_map;
2580 err = btrfs_delayed_inode_init();
2582 goto free_ordered_data;
2584 err = btrfs_auto_defrag_init();
2586 goto free_delayed_inode;
2588 err = btrfs_delayed_ref_init();
2590 goto free_auto_defrag;
2592 err = btrfs_prelim_ref_init();
2594 goto free_delayed_ref;
2596 err = btrfs_end_io_wq_init();
2598 goto free_prelim_ref;
2600 err = btrfs_interface_init();
2602 goto free_end_io_wq;
2604 btrfs_print_mod_info();
2606 err = btrfs_run_sanity_tests();
2608 goto unregister_ioctl;
2610 err = register_filesystem(&btrfs_fs_type);
2612 goto unregister_ioctl;
2617 btrfs_interface_exit();
2619 btrfs_end_io_wq_exit();
2621 btrfs_prelim_ref_exit();
2623 btrfs_delayed_ref_exit();
2625 btrfs_auto_defrag_exit();
2627 btrfs_delayed_inode_exit();
2629 ordered_data_exit();
2632 free_extent_state_cache:
2633 extent_state_cache_exit();
2637 btrfs_destroy_cachep();
2639 btrfs_exit_compress();
2645 static void __exit exit_btrfs_fs(void)
2647 btrfs_destroy_cachep();
2648 btrfs_delayed_ref_exit();
2649 btrfs_auto_defrag_exit();
2650 btrfs_delayed_inode_exit();
2651 btrfs_prelim_ref_exit();
2652 ordered_data_exit();
2654 extent_state_cache_exit();
2656 btrfs_interface_exit();
2657 btrfs_end_io_wq_exit();
2658 unregister_filesystem(&btrfs_fs_type);
2660 btrfs_cleanup_fs_uuids();
2661 btrfs_exit_compress();
2664 late_initcall(init_btrfs_fs);
2665 module_exit(exit_btrfs_fs)
2667 MODULE_LICENSE("GPL");
2668 MODULE_SOFTDEP("pre: crc32c");
2669 MODULE_SOFTDEP("pre: xxhash64");
2670 MODULE_SOFTDEP("pre: sha256");
2671 MODULE_SOFTDEP("pre: blake2b-256");