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 /* Nothing used. The other threads that have joined this
303 * transaction may be able to continue. */
304 if (!trans->dirty && list_empty(&trans->new_bgs)) {
307 errstr = btrfs_decode_error(errno);
309 "%s:%d: Aborting unused transaction(%s).",
310 function, line, errstr);
313 WRITE_ONCE(trans->transaction->aborted, errno);
314 /* Wake up anybody who may be waiting on this transaction */
315 wake_up(&fs_info->transaction_wait);
316 wake_up(&fs_info->transaction_blocked_wait);
317 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
320 * __btrfs_panic decodes unexpected, fatal errors from the caller,
321 * issues an alert, and either panics or BUGs, depending on mount options.
324 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
325 unsigned int line, int errno, const char *fmt, ...)
327 char *s_id = "<unknown>";
329 struct va_format vaf = { .fmt = fmt };
333 s_id = fs_info->sb->s_id;
338 errstr = btrfs_decode_error(errno);
339 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
340 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
341 s_id, function, line, &vaf, errno, errstr);
343 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
344 function, line, &vaf, errno, errstr);
346 /* Caller calls BUG() */
349 static void btrfs_put_super(struct super_block *sb)
351 close_ctree(btrfs_sb(sb));
360 Opt_compress_force_type,
365 Opt_flushoncommit, Opt_noflushoncommit,
367 Opt_barrier, Opt_nobarrier,
368 Opt_datacow, Opt_nodatacow,
369 Opt_datasum, Opt_nodatasum,
370 Opt_defrag, Opt_nodefrag,
371 Opt_discard, Opt_nodiscard,
375 Opt_rescan_uuid_tree,
377 Opt_space_cache, Opt_no_space_cache,
378 Opt_space_cache_version,
380 Opt_ssd_spread, Opt_nossd_spread,
385 Opt_treelog, Opt_notreelog,
386 Opt_user_subvol_rm_allowed,
396 /* Deprecated options */
398 Opt_inode_cache, Opt_noinode_cache,
400 /* Debugging options */
402 Opt_check_integrity_including_extent_data,
403 Opt_check_integrity_print_mask,
404 Opt_enospc_debug, Opt_noenospc_debug,
405 #ifdef CONFIG_BTRFS_DEBUG
406 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
408 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
414 static const match_table_t tokens = {
416 {Opt_noacl, "noacl"},
417 {Opt_clear_cache, "clear_cache"},
418 {Opt_commit_interval, "commit=%u"},
419 {Opt_compress, "compress"},
420 {Opt_compress_type, "compress=%s"},
421 {Opt_compress_force, "compress-force"},
422 {Opt_compress_force_type, "compress-force=%s"},
423 {Opt_degraded, "degraded"},
424 {Opt_device, "device=%s"},
425 {Opt_fatal_errors, "fatal_errors=%s"},
426 {Opt_flushoncommit, "flushoncommit"},
427 {Opt_noflushoncommit, "noflushoncommit"},
428 {Opt_inode_cache, "inode_cache"},
429 {Opt_noinode_cache, "noinode_cache"},
430 {Opt_max_inline, "max_inline=%s"},
431 {Opt_barrier, "barrier"},
432 {Opt_nobarrier, "nobarrier"},
433 {Opt_datacow, "datacow"},
434 {Opt_nodatacow, "nodatacow"},
435 {Opt_datasum, "datasum"},
436 {Opt_nodatasum, "nodatasum"},
437 {Opt_defrag, "autodefrag"},
438 {Opt_nodefrag, "noautodefrag"},
439 {Opt_discard, "discard"},
440 {Opt_discard_mode, "discard=%s"},
441 {Opt_nodiscard, "nodiscard"},
442 {Opt_norecovery, "norecovery"},
443 {Opt_ratio, "metadata_ratio=%u"},
444 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
445 {Opt_skip_balance, "skip_balance"},
446 {Opt_space_cache, "space_cache"},
447 {Opt_no_space_cache, "nospace_cache"},
448 {Opt_space_cache_version, "space_cache=%s"},
450 {Opt_nossd, "nossd"},
451 {Opt_ssd_spread, "ssd_spread"},
452 {Opt_nossd_spread, "nossd_spread"},
453 {Opt_subvol, "subvol=%s"},
454 {Opt_subvol_empty, "subvol="},
455 {Opt_subvolid, "subvolid=%s"},
456 {Opt_thread_pool, "thread_pool=%u"},
457 {Opt_treelog, "treelog"},
458 {Opt_notreelog, "notreelog"},
459 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
462 {Opt_rescue, "rescue=%s"},
463 /* Deprecated, with alias rescue=nologreplay */
464 {Opt_nologreplay, "nologreplay"},
465 /* Deprecated, with alias rescue=usebackuproot */
466 {Opt_usebackuproot, "usebackuproot"},
468 /* Deprecated options */
469 {Opt_recovery, "recovery"},
471 /* Debugging options */
472 {Opt_check_integrity, "check_int"},
473 {Opt_check_integrity_including_extent_data, "check_int_data"},
474 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
475 {Opt_enospc_debug, "enospc_debug"},
476 {Opt_noenospc_debug, "noenospc_debug"},
477 #ifdef CONFIG_BTRFS_DEBUG
478 {Opt_fragment_data, "fragment=data"},
479 {Opt_fragment_metadata, "fragment=metadata"},
480 {Opt_fragment_all, "fragment=all"},
482 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
483 {Opt_ref_verify, "ref_verify"},
488 static const match_table_t rescue_tokens = {
489 {Opt_usebackuproot, "usebackuproot"},
490 {Opt_nologreplay, "nologreplay"},
491 {Opt_ignorebadroots, "ignorebadroots"},
492 {Opt_ignorebadroots, "ibadroots"},
493 {Opt_ignoredatacsums, "ignoredatacsums"},
494 {Opt_ignoredatacsums, "idatacsums"},
495 {Opt_rescue_all, "all"},
499 static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
500 const char *opt_name)
502 if (fs_info->mount_opt & opt) {
503 btrfs_err(fs_info, "%s must be used with ro mount option",
510 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
515 substring_t args[MAX_OPT_ARGS];
518 opts = kstrdup(options, GFP_KERNEL);
523 while ((p = strsep(&opts, ":")) != NULL) {
528 token = match_token(p, rescue_tokens, args);
530 case Opt_usebackuproot:
532 "trying to use backup root at mount time");
533 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
535 case Opt_nologreplay:
536 btrfs_set_and_info(info, NOLOGREPLAY,
537 "disabling log replay at mount time");
539 case Opt_ignorebadroots:
540 btrfs_set_and_info(info, IGNOREBADROOTS,
541 "ignoring bad roots");
543 case Opt_ignoredatacsums:
544 btrfs_set_and_info(info, IGNOREDATACSUMS,
545 "ignoring data csums");
548 btrfs_info(info, "enabling all of the rescue options");
549 btrfs_set_and_info(info, IGNOREDATACSUMS,
550 "ignoring data csums");
551 btrfs_set_and_info(info, IGNOREBADROOTS,
552 "ignoring bad roots");
553 btrfs_set_and_info(info, NOLOGREPLAY,
554 "disabling log replay at mount time");
557 btrfs_info(info, "unrecognized rescue option '%s'", p);
571 * Regular mount options parser. Everything that is needed only when
572 * reading in a new superblock is parsed here.
573 * XXX JDM: This needs to be cleaned up for remount.
575 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
576 unsigned long new_flags)
578 substring_t args[MAX_OPT_ARGS];
583 bool compress_force = false;
584 enum btrfs_compression_type saved_compress_type;
585 int saved_compress_level;
586 bool saved_compress_force;
589 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
590 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
591 else if (btrfs_free_space_cache_v1_active(info)) {
592 if (btrfs_is_zoned(info)) {
594 "zoned: clearing existing space cache");
595 btrfs_set_super_cache_generation(info->super_copy, 0);
597 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
602 * Even the options are empty, we still need to do extra check
608 while ((p = strsep(&options, ",")) != NULL) {
613 token = match_token(p, tokens, args);
616 btrfs_info(info, "allowing degraded mounts");
617 btrfs_set_opt(info->mount_opt, DEGRADED);
620 case Opt_subvol_empty:
624 * These are parsed by btrfs_parse_subvol_options or
625 * btrfs_parse_device_options and can be ignored here.
629 btrfs_set_and_info(info, NODATASUM,
630 "setting nodatasum");
633 if (btrfs_test_opt(info, NODATASUM)) {
634 if (btrfs_test_opt(info, NODATACOW))
636 "setting datasum, datacow enabled");
638 btrfs_info(info, "setting datasum");
640 btrfs_clear_opt(info->mount_opt, NODATACOW);
641 btrfs_clear_opt(info->mount_opt, NODATASUM);
644 if (!btrfs_test_opt(info, NODATACOW)) {
645 if (!btrfs_test_opt(info, COMPRESS) ||
646 !btrfs_test_opt(info, FORCE_COMPRESS)) {
648 "setting nodatacow, compression disabled");
650 btrfs_info(info, "setting nodatacow");
653 btrfs_clear_opt(info->mount_opt, COMPRESS);
654 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
655 btrfs_set_opt(info->mount_opt, NODATACOW);
656 btrfs_set_opt(info->mount_opt, NODATASUM);
659 btrfs_clear_and_info(info, NODATACOW,
662 case Opt_compress_force:
663 case Opt_compress_force_type:
664 compress_force = true;
667 case Opt_compress_type:
668 saved_compress_type = btrfs_test_opt(info,
670 info->compress_type : BTRFS_COMPRESS_NONE;
671 saved_compress_force =
672 btrfs_test_opt(info, FORCE_COMPRESS);
673 saved_compress_level = info->compress_level;
674 if (token == Opt_compress ||
675 token == Opt_compress_force ||
676 strncmp(args[0].from, "zlib", 4) == 0) {
677 compress_type = "zlib";
679 info->compress_type = BTRFS_COMPRESS_ZLIB;
680 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
682 * args[0] contains uninitialized data since
683 * for these tokens we don't expect any
686 if (token != Opt_compress &&
687 token != Opt_compress_force)
688 info->compress_level =
689 btrfs_compress_str2level(
692 btrfs_set_opt(info->mount_opt, COMPRESS);
693 btrfs_clear_opt(info->mount_opt, NODATACOW);
694 btrfs_clear_opt(info->mount_opt, NODATASUM);
696 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
697 compress_type = "lzo";
698 info->compress_type = BTRFS_COMPRESS_LZO;
699 info->compress_level = 0;
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_LZO);
705 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
706 compress_type = "zstd";
707 info->compress_type = BTRFS_COMPRESS_ZSTD;
708 info->compress_level =
709 btrfs_compress_str2level(
712 btrfs_set_opt(info->mount_opt, COMPRESS);
713 btrfs_clear_opt(info->mount_opt, NODATACOW);
714 btrfs_clear_opt(info->mount_opt, NODATASUM);
715 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
717 } else if (strncmp(args[0].from, "no", 2) == 0) {
718 compress_type = "no";
719 info->compress_level = 0;
720 info->compress_type = 0;
721 btrfs_clear_opt(info->mount_opt, COMPRESS);
722 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
723 compress_force = false;
730 if (compress_force) {
731 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
734 * If we remount from compress-force=xxx to
735 * compress=xxx, we need clear FORCE_COMPRESS
736 * flag, otherwise, there is no way for users
737 * to disable forcible compression separately.
739 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
741 if (no_compress == 1) {
742 btrfs_info(info, "use no compression");
743 } else if ((info->compress_type != saved_compress_type) ||
744 (compress_force != saved_compress_force) ||
745 (info->compress_level != saved_compress_level)) {
746 btrfs_info(info, "%s %s compression, level %d",
747 (compress_force) ? "force" : "use",
748 compress_type, info->compress_level);
750 compress_force = false;
753 btrfs_set_and_info(info, SSD,
754 "enabling ssd optimizations");
755 btrfs_clear_opt(info->mount_opt, NOSSD);
758 btrfs_set_and_info(info, SSD,
759 "enabling ssd optimizations");
760 btrfs_set_and_info(info, SSD_SPREAD,
761 "using spread ssd allocation scheme");
762 btrfs_clear_opt(info->mount_opt, NOSSD);
765 btrfs_set_opt(info->mount_opt, NOSSD);
766 btrfs_clear_and_info(info, SSD,
767 "not using ssd optimizations");
769 case Opt_nossd_spread:
770 btrfs_clear_and_info(info, SSD_SPREAD,
771 "not using spread ssd allocation scheme");
774 btrfs_clear_and_info(info, NOBARRIER,
775 "turning on barriers");
778 btrfs_set_and_info(info, NOBARRIER,
779 "turning off barriers");
781 case Opt_thread_pool:
782 ret = match_int(&args[0], &intarg);
785 } else if (intarg == 0) {
789 info->thread_pool_size = intarg;
792 num = match_strdup(&args[0]);
794 info->max_inline = memparse(num, NULL);
797 if (info->max_inline) {
798 info->max_inline = min_t(u64,
802 btrfs_info(info, "max_inline at %llu",
810 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
811 info->sb->s_flags |= SB_POSIXACL;
814 btrfs_err(info, "support for ACL not compiled in!");
819 info->sb->s_flags &= ~SB_POSIXACL;
822 btrfs_set_and_info(info, NOTREELOG,
823 "disabling tree log");
826 btrfs_clear_and_info(info, NOTREELOG,
827 "enabling tree log");
830 case Opt_nologreplay:
832 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
833 btrfs_set_and_info(info, NOLOGREPLAY,
834 "disabling log replay at mount time");
836 case Opt_flushoncommit:
837 btrfs_set_and_info(info, FLUSHONCOMMIT,
838 "turning on flush-on-commit");
840 case Opt_noflushoncommit:
841 btrfs_clear_and_info(info, FLUSHONCOMMIT,
842 "turning off flush-on-commit");
845 ret = match_int(&args[0], &intarg);
848 info->metadata_ratio = intarg;
849 btrfs_info(info, "metadata ratio %u",
850 info->metadata_ratio);
853 case Opt_discard_mode:
854 if (token == Opt_discard ||
855 strcmp(args[0].from, "sync") == 0) {
856 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
857 btrfs_set_and_info(info, DISCARD_SYNC,
858 "turning on sync discard");
859 } else if (strcmp(args[0].from, "async") == 0) {
860 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
861 btrfs_set_and_info(info, DISCARD_ASYNC,
862 "turning on async discard");
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");
892 case Opt_rescan_uuid_tree:
893 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
895 case Opt_no_space_cache:
896 if (btrfs_test_opt(info, SPACE_CACHE)) {
897 btrfs_clear_and_info(info, SPACE_CACHE,
898 "disabling disk space caching");
900 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
901 btrfs_clear_and_info(info, FREE_SPACE_TREE,
902 "disabling free space tree");
905 case Opt_inode_cache:
906 case Opt_noinode_cache:
908 "the 'inode_cache' option is deprecated and has no effect since 5.11");
910 case Opt_clear_cache:
911 btrfs_set_and_info(info, CLEAR_CACHE,
912 "force clearing of disk cache");
914 case Opt_user_subvol_rm_allowed:
915 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
917 case Opt_enospc_debug:
918 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
920 case Opt_noenospc_debug:
921 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
924 btrfs_set_and_info(info, AUTO_DEFRAG,
925 "enabling auto defrag");
928 btrfs_clear_and_info(info, AUTO_DEFRAG,
929 "disabling auto defrag");
932 case Opt_usebackuproot:
934 "'%s' is deprecated, use 'rescue=usebackuproot' instead",
935 token == Opt_recovery ? "recovery" :
938 "trying to use backup root at mount time");
939 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
941 case Opt_skip_balance:
942 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
944 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
945 case Opt_check_integrity_including_extent_data:
947 "enabling check integrity including extent data");
948 btrfs_set_opt(info->mount_opt,
949 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
950 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
952 case Opt_check_integrity:
953 btrfs_info(info, "enabling check integrity");
954 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
956 case Opt_check_integrity_print_mask:
957 ret = match_int(&args[0], &intarg);
960 info->check_integrity_print_mask = intarg;
961 btrfs_info(info, "check_integrity_print_mask 0x%x",
962 info->check_integrity_print_mask);
965 case Opt_check_integrity_including_extent_data:
966 case Opt_check_integrity:
967 case Opt_check_integrity_print_mask:
969 "support for check_integrity* not compiled in!");
973 case Opt_fatal_errors:
974 if (strcmp(args[0].from, "panic") == 0)
975 btrfs_set_opt(info->mount_opt,
976 PANIC_ON_FATAL_ERROR);
977 else if (strcmp(args[0].from, "bug") == 0)
978 btrfs_clear_opt(info->mount_opt,
979 PANIC_ON_FATAL_ERROR);
985 case Opt_commit_interval:
987 ret = match_int(&args[0], &intarg);
992 "using default commit interval %us",
993 BTRFS_DEFAULT_COMMIT_INTERVAL);
994 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
995 } else if (intarg > 300) {
996 btrfs_warn(info, "excessive commit interval %d",
999 info->commit_interval = intarg;
1002 ret = parse_rescue_options(info, args[0].from);
1006 #ifdef CONFIG_BTRFS_DEBUG
1007 case Opt_fragment_all:
1008 btrfs_info(info, "fragmenting all space");
1009 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1010 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
1012 case Opt_fragment_metadata:
1013 btrfs_info(info, "fragmenting metadata");
1014 btrfs_set_opt(info->mount_opt,
1017 case Opt_fragment_data:
1018 btrfs_info(info, "fragmenting data");
1019 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1022 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
1023 case Opt_ref_verify:
1024 btrfs_info(info, "doing ref verification");
1025 btrfs_set_opt(info->mount_opt, REF_VERIFY);
1029 btrfs_err(info, "unrecognized mount option '%s'", p);
1037 /* We're read-only, don't have to check. */
1038 if (new_flags & SB_RDONLY)
1041 if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
1042 check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
1043 check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
1046 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
1047 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
1048 !btrfs_test_opt(info, CLEAR_CACHE)) {
1049 btrfs_err(info, "cannot disable free space tree");
1054 ret = btrfs_check_mountopts_zoned(info);
1055 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
1056 btrfs_info(info, "disk space caching is enabled");
1057 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
1058 btrfs_info(info, "using free space tree");
1063 * Parse mount options that are required early in the mount process.
1065 * All other options will be parsed on much later in the mount process and
1066 * only when we need to allocate a new super block.
1068 static int btrfs_parse_device_options(const char *options, fmode_t flags,
1071 substring_t args[MAX_OPT_ARGS];
1072 char *device_name, *opts, *orig, *p;
1073 struct btrfs_device *device = NULL;
1076 lockdep_assert_held(&uuid_mutex);
1082 * strsep changes the string, duplicate it because btrfs_parse_options
1085 opts = kstrdup(options, GFP_KERNEL);
1090 while ((p = strsep(&opts, ",")) != NULL) {
1096 token = match_token(p, tokens, args);
1097 if (token == Opt_device) {
1098 device_name = match_strdup(&args[0]);
1103 device = btrfs_scan_one_device(device_name, flags,
1106 if (IS_ERR(device)) {
1107 error = PTR_ERR(device);
1119 * Parse mount options that are related to subvolume id
1121 * The value is later passed to mount_subvol()
1123 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1124 u64 *subvol_objectid)
1126 substring_t args[MAX_OPT_ARGS];
1127 char *opts, *orig, *p;
1135 * strsep changes the string, duplicate it because
1136 * btrfs_parse_device_options gets called later
1138 opts = kstrdup(options, GFP_KERNEL);
1143 while ((p = strsep(&opts, ",")) != NULL) {
1148 token = match_token(p, tokens, args);
1151 kfree(*subvol_name);
1152 *subvol_name = match_strdup(&args[0]);
1153 if (!*subvol_name) {
1159 error = match_u64(&args[0], &subvolid);
1163 /* we want the original fs_tree */
1165 subvolid = BTRFS_FS_TREE_OBJECTID;
1167 *subvol_objectid = subvolid;
1179 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1180 u64 subvol_objectid)
1182 struct btrfs_root *root = fs_info->tree_root;
1183 struct btrfs_root *fs_root = NULL;
1184 struct btrfs_root_ref *root_ref;
1185 struct btrfs_inode_ref *inode_ref;
1186 struct btrfs_key key;
1187 struct btrfs_path *path = NULL;
1188 char *name = NULL, *ptr;
1193 path = btrfs_alloc_path();
1199 name = kmalloc(PATH_MAX, GFP_KERNEL);
1204 ptr = name + PATH_MAX - 1;
1208 * Walk up the subvolume trees in the tree of tree roots by root
1209 * backrefs until we hit the top-level subvolume.
1211 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1212 key.objectid = subvol_objectid;
1213 key.type = BTRFS_ROOT_BACKREF_KEY;
1214 key.offset = (u64)-1;
1216 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1219 } else if (ret > 0) {
1220 ret = btrfs_previous_item(root, path, subvol_objectid,
1221 BTRFS_ROOT_BACKREF_KEY);
1224 } else if (ret > 0) {
1230 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1231 subvol_objectid = key.offset;
1233 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1234 struct btrfs_root_ref);
1235 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1238 ret = -ENAMETOOLONG;
1241 read_extent_buffer(path->nodes[0], ptr + 1,
1242 (unsigned long)(root_ref + 1), len);
1244 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1245 btrfs_release_path(path);
1247 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1248 if (IS_ERR(fs_root)) {
1249 ret = PTR_ERR(fs_root);
1255 * Walk up the filesystem tree by inode refs until we hit the
1258 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1259 key.objectid = dirid;
1260 key.type = BTRFS_INODE_REF_KEY;
1261 key.offset = (u64)-1;
1263 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1266 } else if (ret > 0) {
1267 ret = btrfs_previous_item(fs_root, path, dirid,
1268 BTRFS_INODE_REF_KEY);
1271 } else if (ret > 0) {
1277 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1280 inode_ref = btrfs_item_ptr(path->nodes[0],
1282 struct btrfs_inode_ref);
1283 len = btrfs_inode_ref_name_len(path->nodes[0],
1287 ret = -ENAMETOOLONG;
1290 read_extent_buffer(path->nodes[0], ptr + 1,
1291 (unsigned long)(inode_ref + 1), len);
1293 btrfs_release_path(path);
1295 btrfs_put_root(fs_root);
1299 btrfs_free_path(path);
1300 if (ptr == name + PATH_MAX - 1) {
1304 memmove(name, ptr, name + PATH_MAX - ptr);
1309 btrfs_put_root(fs_root);
1310 btrfs_free_path(path);
1312 return ERR_PTR(ret);
1315 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1317 struct btrfs_root *root = fs_info->tree_root;
1318 struct btrfs_dir_item *di;
1319 struct btrfs_path *path;
1320 struct btrfs_key location;
1323 path = btrfs_alloc_path();
1328 * Find the "default" dir item which points to the root item that we
1329 * will mount by default if we haven't been given a specific subvolume
1332 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1333 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1335 btrfs_free_path(path);
1340 * Ok the default dir item isn't there. This is weird since
1341 * it's always been there, but don't freak out, just try and
1342 * mount the top-level subvolume.
1344 btrfs_free_path(path);
1345 *objectid = BTRFS_FS_TREE_OBJECTID;
1349 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1350 btrfs_free_path(path);
1351 *objectid = location.objectid;
1355 static int btrfs_fill_super(struct super_block *sb,
1356 struct btrfs_fs_devices *fs_devices,
1359 struct inode *inode;
1360 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1363 sb->s_maxbytes = MAX_LFS_FILESIZE;
1364 sb->s_magic = BTRFS_SUPER_MAGIC;
1365 sb->s_op = &btrfs_super_ops;
1366 sb->s_d_op = &btrfs_dentry_operations;
1367 sb->s_export_op = &btrfs_export_ops;
1368 sb->s_xattr = btrfs_xattr_handlers;
1369 sb->s_time_gran = 1;
1370 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1371 sb->s_flags |= SB_POSIXACL;
1373 sb->s_flags |= SB_I_VERSION;
1374 sb->s_iflags |= SB_I_CGROUPWB;
1376 err = super_setup_bdi(sb);
1378 btrfs_err(fs_info, "super_setup_bdi failed");
1382 err = open_ctree(sb, fs_devices, (char *)data);
1384 btrfs_err(fs_info, "open_ctree failed");
1388 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1389 if (IS_ERR(inode)) {
1390 err = PTR_ERR(inode);
1394 sb->s_root = d_make_root(inode);
1400 cleancache_init_fs(sb);
1401 sb->s_flags |= SB_ACTIVE;
1405 close_ctree(fs_info);
1409 int btrfs_sync_fs(struct super_block *sb, int wait)
1411 struct btrfs_trans_handle *trans;
1412 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1413 struct btrfs_root *root = fs_info->tree_root;
1415 trace_btrfs_sync_fs(fs_info, wait);
1418 filemap_flush(fs_info->btree_inode->i_mapping);
1422 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1424 trans = btrfs_attach_transaction_barrier(root);
1425 if (IS_ERR(trans)) {
1426 /* no transaction, don't bother */
1427 if (PTR_ERR(trans) == -ENOENT) {
1429 * Exit unless we have some pending changes
1430 * that need to go through commit
1432 if (fs_info->pending_changes == 0)
1435 * A non-blocking test if the fs is frozen. We must not
1436 * start a new transaction here otherwise a deadlock
1437 * happens. The pending operations are delayed to the
1438 * next commit after thawing.
1440 if (sb_start_write_trylock(sb))
1444 trans = btrfs_start_transaction(root, 0);
1447 return PTR_ERR(trans);
1449 return btrfs_commit_transaction(trans);
1452 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1454 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1458 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1460 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1461 const char *compress_type;
1462 const char *subvol_name;
1463 bool printed = false;
1465 if (btrfs_test_opt(info, DEGRADED))
1466 seq_puts(seq, ",degraded");
1467 if (btrfs_test_opt(info, NODATASUM))
1468 seq_puts(seq, ",nodatasum");
1469 if (btrfs_test_opt(info, NODATACOW))
1470 seq_puts(seq, ",nodatacow");
1471 if (btrfs_test_opt(info, NOBARRIER))
1472 seq_puts(seq, ",nobarrier");
1473 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1474 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1475 if (info->thread_pool_size != min_t(unsigned long,
1476 num_online_cpus() + 2, 8))
1477 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1478 if (btrfs_test_opt(info, COMPRESS)) {
1479 compress_type = btrfs_compress_type2str(info->compress_type);
1480 if (btrfs_test_opt(info, FORCE_COMPRESS))
1481 seq_printf(seq, ",compress-force=%s", compress_type);
1483 seq_printf(seq, ",compress=%s", compress_type);
1484 if (info->compress_level)
1485 seq_printf(seq, ":%d", info->compress_level);
1487 if (btrfs_test_opt(info, NOSSD))
1488 seq_puts(seq, ",nossd");
1489 if (btrfs_test_opt(info, SSD_SPREAD))
1490 seq_puts(seq, ",ssd_spread");
1491 else if (btrfs_test_opt(info, SSD))
1492 seq_puts(seq, ",ssd");
1493 if (btrfs_test_opt(info, NOTREELOG))
1494 seq_puts(seq, ",notreelog");
1495 if (btrfs_test_opt(info, NOLOGREPLAY))
1496 print_rescue_option(seq, "nologreplay", &printed);
1497 if (btrfs_test_opt(info, USEBACKUPROOT))
1498 print_rescue_option(seq, "usebackuproot", &printed);
1499 if (btrfs_test_opt(info, IGNOREBADROOTS))
1500 print_rescue_option(seq, "ignorebadroots", &printed);
1501 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1502 print_rescue_option(seq, "ignoredatacsums", &printed);
1503 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1504 seq_puts(seq, ",flushoncommit");
1505 if (btrfs_test_opt(info, DISCARD_SYNC))
1506 seq_puts(seq, ",discard");
1507 if (btrfs_test_opt(info, DISCARD_ASYNC))
1508 seq_puts(seq, ",discard=async");
1509 if (!(info->sb->s_flags & SB_POSIXACL))
1510 seq_puts(seq, ",noacl");
1511 if (btrfs_free_space_cache_v1_active(info))
1512 seq_puts(seq, ",space_cache");
1513 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1514 seq_puts(seq, ",space_cache=v2");
1516 seq_puts(seq, ",nospace_cache");
1517 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1518 seq_puts(seq, ",rescan_uuid_tree");
1519 if (btrfs_test_opt(info, CLEAR_CACHE))
1520 seq_puts(seq, ",clear_cache");
1521 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1522 seq_puts(seq, ",user_subvol_rm_allowed");
1523 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1524 seq_puts(seq, ",enospc_debug");
1525 if (btrfs_test_opt(info, AUTO_DEFRAG))
1526 seq_puts(seq, ",autodefrag");
1527 if (btrfs_test_opt(info, SKIP_BALANCE))
1528 seq_puts(seq, ",skip_balance");
1529 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1530 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1531 seq_puts(seq, ",check_int_data");
1532 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1533 seq_puts(seq, ",check_int");
1534 if (info->check_integrity_print_mask)
1535 seq_printf(seq, ",check_int_print_mask=%d",
1536 info->check_integrity_print_mask);
1538 if (info->metadata_ratio)
1539 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1540 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1541 seq_puts(seq, ",fatal_errors=panic");
1542 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1543 seq_printf(seq, ",commit=%u", info->commit_interval);
1544 #ifdef CONFIG_BTRFS_DEBUG
1545 if (btrfs_test_opt(info, FRAGMENT_DATA))
1546 seq_puts(seq, ",fragment=data");
1547 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1548 seq_puts(seq, ",fragment=metadata");
1550 if (btrfs_test_opt(info, REF_VERIFY))
1551 seq_puts(seq, ",ref_verify");
1552 seq_printf(seq, ",subvolid=%llu",
1553 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1554 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1555 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1556 if (!IS_ERR(subvol_name)) {
1557 seq_puts(seq, ",subvol=");
1558 seq_escape(seq, subvol_name, " \t\n\\");
1564 static int btrfs_test_super(struct super_block *s, void *data)
1566 struct btrfs_fs_info *p = data;
1567 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1569 return fs_info->fs_devices == p->fs_devices;
1572 static int btrfs_set_super(struct super_block *s, void *data)
1574 int err = set_anon_super(s, data);
1576 s->s_fs_info = data;
1581 * subvolumes are identified by ino 256
1583 static inline int is_subvolume_inode(struct inode *inode)
1585 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1590 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1591 struct vfsmount *mnt)
1593 struct dentry *root;
1597 if (!subvol_objectid) {
1598 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1601 root = ERR_PTR(ret);
1605 subvol_name = btrfs_get_subvol_name_from_objectid(
1606 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1607 if (IS_ERR(subvol_name)) {
1608 root = ERR_CAST(subvol_name);
1615 root = mount_subtree(mnt, subvol_name);
1616 /* mount_subtree() drops our reference on the vfsmount. */
1619 if (!IS_ERR(root)) {
1620 struct super_block *s = root->d_sb;
1621 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1622 struct inode *root_inode = d_inode(root);
1623 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1626 if (!is_subvolume_inode(root_inode)) {
1627 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1631 if (subvol_objectid && root_objectid != subvol_objectid) {
1633 * This will also catch a race condition where a
1634 * subvolume which was passed by ID is renamed and
1635 * another subvolume is renamed over the old location.
1638 "subvol '%s' does not match subvolid %llu",
1639 subvol_name, subvol_objectid);
1644 root = ERR_PTR(ret);
1645 deactivate_locked_super(s);
1656 * Find a superblock for the given device / mount point.
1658 * Note: This is based on mount_bdev from fs/super.c with a few additions
1659 * for multiple device setup. Make sure to keep it in sync.
1661 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1662 int flags, const char *device_name, void *data)
1664 struct block_device *bdev = NULL;
1665 struct super_block *s;
1666 struct btrfs_device *device = NULL;
1667 struct btrfs_fs_devices *fs_devices = NULL;
1668 struct btrfs_fs_info *fs_info = NULL;
1669 void *new_sec_opts = NULL;
1670 fmode_t mode = FMODE_READ;
1673 if (!(flags & SB_RDONLY))
1674 mode |= FMODE_WRITE;
1677 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1679 return ERR_PTR(error);
1683 * Setup a dummy root and fs_info for test/set super. This is because
1684 * we don't actually fill this stuff out until open_ctree, but we need
1685 * then open_ctree will properly initialize the file system specific
1686 * settings later. btrfs_init_fs_info initializes the static elements
1687 * of the fs_info (locks and such) to make cleanup easier if we find a
1688 * superblock with our given fs_devices later on at sget() time.
1690 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1693 goto error_sec_opts;
1695 btrfs_init_fs_info(fs_info);
1697 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1698 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1699 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1704 mutex_lock(&uuid_mutex);
1705 error = btrfs_parse_device_options(data, mode, fs_type);
1707 mutex_unlock(&uuid_mutex);
1711 device = btrfs_scan_one_device(device_name, mode, fs_type);
1712 if (IS_ERR(device)) {
1713 mutex_unlock(&uuid_mutex);
1714 error = PTR_ERR(device);
1718 fs_devices = device->fs_devices;
1719 fs_info->fs_devices = fs_devices;
1721 error = btrfs_open_devices(fs_devices, mode, fs_type);
1722 mutex_unlock(&uuid_mutex);
1726 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1728 goto error_close_devices;
1731 bdev = fs_devices->latest_bdev;
1732 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1736 goto error_close_devices;
1740 btrfs_close_devices(fs_devices);
1741 btrfs_free_fs_info(fs_info);
1742 if ((flags ^ s->s_flags) & SB_RDONLY)
1745 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1746 btrfs_sb(s)->bdev_holder = fs_type;
1747 if (!strstr(crc32c_impl(), "generic"))
1748 set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1749 error = btrfs_fill_super(s, fs_devices, data);
1752 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1753 security_free_mnt_opts(&new_sec_opts);
1755 deactivate_locked_super(s);
1756 return ERR_PTR(error);
1759 return dget(s->s_root);
1761 error_close_devices:
1762 btrfs_close_devices(fs_devices);
1764 btrfs_free_fs_info(fs_info);
1766 security_free_mnt_opts(&new_sec_opts);
1767 return ERR_PTR(error);
1771 * Mount function which is called by VFS layer.
1773 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1774 * which needs vfsmount* of device's root (/). This means device's root has to
1775 * be mounted internally in any case.
1778 * 1. Parse subvol id related options for later use in mount_subvol().
1780 * 2. Mount device's root (/) by calling vfs_kern_mount().
1782 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1783 * first place. In order to avoid calling btrfs_mount() again, we use
1784 * different file_system_type which is not registered to VFS by
1785 * register_filesystem() (btrfs_root_fs_type). As a result,
1786 * btrfs_mount_root() is called. The return value will be used by
1787 * mount_subtree() in mount_subvol().
1789 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1790 * "btrfs subvolume set-default", mount_subvol() is called always.
1792 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1793 const char *device_name, void *data)
1795 struct vfsmount *mnt_root;
1796 struct dentry *root;
1797 char *subvol_name = NULL;
1798 u64 subvol_objectid = 0;
1801 error = btrfs_parse_subvol_options(data, &subvol_name,
1805 return ERR_PTR(error);
1808 /* mount device's root (/) */
1809 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1810 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1811 if (flags & SB_RDONLY) {
1812 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1813 flags & ~SB_RDONLY, device_name, data);
1815 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1816 flags | SB_RDONLY, device_name, data);
1817 if (IS_ERR(mnt_root)) {
1818 root = ERR_CAST(mnt_root);
1823 down_write(&mnt_root->mnt_sb->s_umount);
1824 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1825 up_write(&mnt_root->mnt_sb->s_umount);
1827 root = ERR_PTR(error);
1834 if (IS_ERR(mnt_root)) {
1835 root = ERR_CAST(mnt_root);
1840 /* mount_subvol() will free subvol_name and mnt_root */
1841 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1847 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1848 u32 new_pool_size, u32 old_pool_size)
1850 if (new_pool_size == old_pool_size)
1853 fs_info->thread_pool_size = new_pool_size;
1855 btrfs_info(fs_info, "resize thread pool %d -> %d",
1856 old_pool_size, new_pool_size);
1858 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1859 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1860 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1861 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1862 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1863 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1865 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1866 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1867 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1868 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1869 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1873 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1874 unsigned long old_opts, int flags)
1876 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1877 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1878 (flags & SB_RDONLY))) {
1879 /* wait for any defraggers to finish */
1880 wait_event(fs_info->transaction_wait,
1881 (atomic_read(&fs_info->defrag_running) == 0));
1882 if (flags & SB_RDONLY)
1883 sync_filesystem(fs_info->sb);
1887 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1888 unsigned long old_opts)
1890 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1893 * We need to cleanup all defragable inodes if the autodefragment is
1894 * close or the filesystem is read only.
1896 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1897 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1898 btrfs_cleanup_defrag_inodes(fs_info);
1901 /* If we toggled discard async */
1902 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1903 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1904 btrfs_discard_resume(fs_info);
1905 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1906 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1907 btrfs_discard_cleanup(fs_info);
1909 /* If we toggled space cache */
1910 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1911 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1914 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1916 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1917 unsigned old_flags = sb->s_flags;
1918 unsigned long old_opts = fs_info->mount_opt;
1919 unsigned long old_compress_type = fs_info->compress_type;
1920 u64 old_max_inline = fs_info->max_inline;
1921 u32 old_thread_pool_size = fs_info->thread_pool_size;
1922 u32 old_metadata_ratio = fs_info->metadata_ratio;
1925 sync_filesystem(sb);
1926 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1929 void *new_sec_opts = NULL;
1931 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1933 ret = security_sb_remount(sb, new_sec_opts);
1934 security_free_mnt_opts(&new_sec_opts);
1939 ret = btrfs_parse_options(fs_info, data, *flags);
1943 btrfs_remount_begin(fs_info, old_opts, *flags);
1944 btrfs_resize_thread_pool(fs_info,
1945 fs_info->thread_pool_size, old_thread_pool_size);
1947 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1948 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1949 (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1951 "remount supports changing free space tree only from ro to rw");
1952 /* Make sure free space cache options match the state on disk */
1953 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1954 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1955 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1957 if (btrfs_free_space_cache_v1_active(fs_info)) {
1958 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1959 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1963 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1966 if (*flags & SB_RDONLY) {
1968 * this also happens on 'umount -rf' or on shutdown, when
1969 * the filesystem is busy.
1971 cancel_work_sync(&fs_info->async_reclaim_work);
1972 cancel_work_sync(&fs_info->async_data_reclaim_work);
1974 btrfs_discard_cleanup(fs_info);
1976 /* wait for the uuid_scan task to finish */
1977 down(&fs_info->uuid_tree_rescan_sem);
1978 /* avoid complains from lockdep et al. */
1979 up(&fs_info->uuid_tree_rescan_sem);
1981 btrfs_set_sb_rdonly(sb);
1984 * Setting SB_RDONLY will put the cleaner thread to
1985 * sleep at the next loop if it's already active.
1986 * If it's already asleep, we'll leave unused block
1987 * groups on disk until we're mounted read-write again
1988 * unless we clean them up here.
1990 btrfs_delete_unused_bgs(fs_info);
1993 * The cleaner task could be already running before we set the
1994 * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
1995 * We must make sure that after we finish the remount, i.e. after
1996 * we call btrfs_commit_super(), the cleaner can no longer start
1997 * a transaction - either because it was dropping a dead root,
1998 * running delayed iputs or deleting an unused block group (the
1999 * cleaner picked a block group from the list of unused block
2000 * groups before we were able to in the previous call to
2001 * btrfs_delete_unused_bgs()).
2003 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
2004 TASK_UNINTERRUPTIBLE);
2007 * We've set the superblock to RO mode, so we might have made
2008 * the cleaner task sleep without running all pending delayed
2009 * iputs. Go through all the delayed iputs here, so that if an
2010 * unmount happens without remounting RW we don't end up at
2011 * finishing close_ctree() with a non-empty list of delayed
2014 btrfs_run_delayed_iputs(fs_info);
2016 btrfs_dev_replace_suspend_for_unmount(fs_info);
2017 btrfs_scrub_cancel(fs_info);
2018 btrfs_pause_balance(fs_info);
2021 * Pause the qgroup rescan worker if it is running. We don't want
2022 * it to be still running after we are in RO mode, as after that,
2023 * by the time we unmount, it might have left a transaction open,
2024 * so we would leak the transaction and/or crash.
2026 btrfs_qgroup_wait_for_completion(fs_info, false);
2028 ret = btrfs_commit_super(fs_info);
2032 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2034 "Remounting read-write after error is not allowed");
2038 if (fs_info->fs_devices->rw_devices == 0) {
2043 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
2045 "too many missing devices, writable remount is not allowed");
2050 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
2052 "mount required to replay tree-log, cannot remount read-write");
2056 if (fs_info->sectorsize < PAGE_SIZE) {
2058 "read-write mount is not yet allowed for sectorsize %u page size %lu",
2059 fs_info->sectorsize, PAGE_SIZE);
2065 * NOTE: when remounting with a change that does writes, don't
2066 * put it anywhere above this point, as we are not sure to be
2067 * safe to write until we pass the above checks.
2069 ret = btrfs_start_pre_rw_mount(fs_info);
2073 btrfs_clear_sb_rdonly(sb);
2075 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
2079 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
2080 * since the absence of the flag means it can be toggled off by remount.
2082 *flags |= SB_I_VERSION;
2084 wake_up_process(fs_info->transaction_kthread);
2085 btrfs_remount_cleanup(fs_info, old_opts);
2086 btrfs_clear_oneshot_options(fs_info);
2087 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2092 /* We've hit an error - don't reset SB_RDONLY */
2094 old_flags |= SB_RDONLY;
2095 if (!(old_flags & SB_RDONLY))
2096 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
2097 sb->s_flags = old_flags;
2098 fs_info->mount_opt = old_opts;
2099 fs_info->compress_type = old_compress_type;
2100 fs_info->max_inline = old_max_inline;
2101 btrfs_resize_thread_pool(fs_info,
2102 old_thread_pool_size, fs_info->thread_pool_size);
2103 fs_info->metadata_ratio = old_metadata_ratio;
2104 btrfs_remount_cleanup(fs_info, old_opts);
2105 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2110 /* Used to sort the devices by max_avail(descending sort) */
2111 static inline int btrfs_cmp_device_free_bytes(const void *dev_info1,
2112 const void *dev_info2)
2114 if (((struct btrfs_device_info *)dev_info1)->max_avail >
2115 ((struct btrfs_device_info *)dev_info2)->max_avail)
2117 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
2118 ((struct btrfs_device_info *)dev_info2)->max_avail)
2125 * sort the devices by max_avail, in which max free extent size of each device
2126 * is stored.(Descending Sort)
2128 static inline void btrfs_descending_sort_devices(
2129 struct btrfs_device_info *devices,
2132 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2133 btrfs_cmp_device_free_bytes, NULL);
2137 * The helper to calc the free space on the devices that can be used to store
2140 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2143 struct btrfs_device_info *devices_info;
2144 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2145 struct btrfs_device *device;
2148 u64 min_stripe_size;
2149 int num_stripes = 1;
2150 int i = 0, nr_devices;
2151 const struct btrfs_raid_attr *rattr;
2154 * We aren't under the device list lock, so this is racy-ish, but good
2155 * enough for our purposes.
2157 nr_devices = fs_info->fs_devices->open_devices;
2160 nr_devices = fs_info->fs_devices->open_devices;
2168 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2173 /* calc min stripe number for data space allocation */
2174 type = btrfs_data_alloc_profile(fs_info);
2175 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2177 if (type & BTRFS_BLOCK_GROUP_RAID0)
2178 num_stripes = nr_devices;
2179 else if (type & BTRFS_BLOCK_GROUP_RAID1)
2181 else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
2183 else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
2185 else if (type & BTRFS_BLOCK_GROUP_RAID10)
2188 /* Adjust for more than 1 stripe per device */
2189 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2192 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2193 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2194 &device->dev_state) ||
2196 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2199 if (i >= nr_devices)
2202 avail_space = device->total_bytes - device->bytes_used;
2204 /* align with stripe_len */
2205 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2208 * In order to avoid overwriting the superblock on the drive,
2209 * btrfs starts at an offset of at least 1MB when doing chunk
2212 * This ensures we have at least min_stripe_size free space
2213 * after excluding 1MB.
2215 if (avail_space <= SZ_1M + min_stripe_size)
2218 avail_space -= SZ_1M;
2220 devices_info[i].dev = device;
2221 devices_info[i].max_avail = avail_space;
2229 btrfs_descending_sort_devices(devices_info, nr_devices);
2233 while (nr_devices >= rattr->devs_min) {
2234 num_stripes = min(num_stripes, nr_devices);
2236 if (devices_info[i].max_avail >= min_stripe_size) {
2240 avail_space += devices_info[i].max_avail * num_stripes;
2241 alloc_size = devices_info[i].max_avail;
2242 for (j = i + 1 - num_stripes; j <= i; j++)
2243 devices_info[j].max_avail -= alloc_size;
2249 kfree(devices_info);
2250 *free_bytes = avail_space;
2255 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2257 * If there's a redundant raid level at DATA block groups, use the respective
2258 * multiplier to scale the sizes.
2260 * Unused device space usage is based on simulating the chunk allocator
2261 * algorithm that respects the device sizes and order of allocations. This is
2262 * a close approximation of the actual use but there are other factors that may
2263 * change the result (like a new metadata chunk).
2265 * If metadata is exhausted, f_bavail will be 0.
2267 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2269 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2270 struct btrfs_super_block *disk_super = fs_info->super_copy;
2271 struct btrfs_space_info *found;
2273 u64 total_free_data = 0;
2274 u64 total_free_meta = 0;
2275 u32 bits = fs_info->sectorsize_bits;
2276 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2277 unsigned factor = 1;
2278 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2283 list_for_each_entry(found, &fs_info->space_info, list) {
2284 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2287 total_free_data += found->disk_total - found->disk_used;
2289 btrfs_account_ro_block_groups_free_space(found);
2291 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2292 if (!list_empty(&found->block_groups[i]))
2293 factor = btrfs_bg_type_to_factor(
2294 btrfs_raid_array[i].bg_flag);
2299 * Metadata in mixed block goup profiles are accounted in data
2301 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2302 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2305 total_free_meta += found->disk_total -
2309 total_used += found->disk_used;
2312 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2313 buf->f_blocks >>= bits;
2314 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2316 /* Account global block reserve as used, it's in logical size already */
2317 spin_lock(&block_rsv->lock);
2318 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2319 if (buf->f_bfree >= block_rsv->size >> bits)
2320 buf->f_bfree -= block_rsv->size >> bits;
2323 spin_unlock(&block_rsv->lock);
2325 buf->f_bavail = div_u64(total_free_data, factor);
2326 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2329 buf->f_bavail += div_u64(total_free_data, factor);
2330 buf->f_bavail = buf->f_bavail >> bits;
2333 * We calculate the remaining metadata space minus global reserve. If
2334 * this is (supposedly) smaller than zero, there's no space. But this
2335 * does not hold in practice, the exhausted state happens where's still
2336 * some positive delta. So we apply some guesswork and compare the
2337 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2339 * We probably cannot calculate the exact threshold value because this
2340 * depends on the internal reservations requested by various
2341 * operations, so some operations that consume a few metadata will
2342 * succeed even if the Avail is zero. But this is better than the other
2348 * We only want to claim there's no available space if we can no longer
2349 * allocate chunks for our metadata profile and our global reserve will
2350 * not fit in the free metadata space. If we aren't ->full then we
2351 * still can allocate chunks and thus are fine using the currently
2352 * calculated f_bavail.
2354 if (!mixed && block_rsv->space_info->full &&
2355 total_free_meta - thresh < block_rsv->size)
2358 buf->f_type = BTRFS_SUPER_MAGIC;
2359 buf->f_bsize = dentry->d_sb->s_blocksize;
2360 buf->f_namelen = BTRFS_NAME_LEN;
2362 /* We treat it as constant endianness (it doesn't matter _which_)
2363 because we want the fsid to come out the same whether mounted
2364 on a big-endian or little-endian host */
2365 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2366 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2367 /* Mask in the root object ID too, to disambiguate subvols */
2368 buf->f_fsid.val[0] ^=
2369 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2370 buf->f_fsid.val[1] ^=
2371 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2376 static void btrfs_kill_super(struct super_block *sb)
2378 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2379 kill_anon_super(sb);
2380 btrfs_free_fs_info(fs_info);
2383 static struct file_system_type btrfs_fs_type = {
2384 .owner = THIS_MODULE,
2386 .mount = btrfs_mount,
2387 .kill_sb = btrfs_kill_super,
2388 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2391 static struct file_system_type btrfs_root_fs_type = {
2392 .owner = THIS_MODULE,
2394 .mount = btrfs_mount_root,
2395 .kill_sb = btrfs_kill_super,
2396 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2399 MODULE_ALIAS_FS("btrfs");
2401 static int btrfs_control_open(struct inode *inode, struct file *file)
2404 * The control file's private_data is used to hold the
2405 * transaction when it is started and is used to keep
2406 * track of whether a transaction is already in progress.
2408 file->private_data = NULL;
2413 * Used by /dev/btrfs-control for devices ioctls.
2415 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2418 struct btrfs_ioctl_vol_args *vol;
2419 struct btrfs_device *device = NULL;
2422 if (!capable(CAP_SYS_ADMIN))
2425 vol = memdup_user((void __user *)arg, sizeof(*vol));
2427 return PTR_ERR(vol);
2428 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2431 case BTRFS_IOC_SCAN_DEV:
2432 mutex_lock(&uuid_mutex);
2433 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2434 &btrfs_root_fs_type);
2435 ret = PTR_ERR_OR_ZERO(device);
2436 mutex_unlock(&uuid_mutex);
2438 case BTRFS_IOC_FORGET_DEV:
2439 ret = btrfs_forget_devices(vol->name);
2441 case BTRFS_IOC_DEVICES_READY:
2442 mutex_lock(&uuid_mutex);
2443 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2444 &btrfs_root_fs_type);
2445 if (IS_ERR(device)) {
2446 mutex_unlock(&uuid_mutex);
2447 ret = PTR_ERR(device);
2450 ret = !(device->fs_devices->num_devices ==
2451 device->fs_devices->total_devices);
2452 mutex_unlock(&uuid_mutex);
2454 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2455 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2463 static int btrfs_freeze(struct super_block *sb)
2465 struct btrfs_trans_handle *trans;
2466 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2467 struct btrfs_root *root = fs_info->tree_root;
2469 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2471 * We don't need a barrier here, we'll wait for any transaction that
2472 * could be in progress on other threads (and do delayed iputs that
2473 * we want to avoid on a frozen filesystem), or do the commit
2476 trans = btrfs_attach_transaction_barrier(root);
2477 if (IS_ERR(trans)) {
2478 /* no transaction, don't bother */
2479 if (PTR_ERR(trans) == -ENOENT)
2481 return PTR_ERR(trans);
2483 return btrfs_commit_transaction(trans);
2486 static int btrfs_unfreeze(struct super_block *sb)
2488 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2490 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2494 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2496 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2497 struct btrfs_device *dev, *first_dev = NULL;
2500 * Lightweight locking of the devices. We should not need
2501 * device_list_mutex here as we only read the device data and the list
2502 * is protected by RCU. Even if a device is deleted during the list
2503 * traversals, we'll get valid data, the freeing callback will wait at
2504 * least until the rcu_read_unlock.
2507 list_for_each_entry_rcu(dev, &fs_info->fs_devices->devices, dev_list) {
2508 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2512 if (!first_dev || dev->devid < first_dev->devid)
2517 seq_escape(m, rcu_str_deref(first_dev->name), " \t\n\\");
2524 static const struct super_operations btrfs_super_ops = {
2525 .drop_inode = btrfs_drop_inode,
2526 .evict_inode = btrfs_evict_inode,
2527 .put_super = btrfs_put_super,
2528 .sync_fs = btrfs_sync_fs,
2529 .show_options = btrfs_show_options,
2530 .show_devname = btrfs_show_devname,
2531 .alloc_inode = btrfs_alloc_inode,
2532 .destroy_inode = btrfs_destroy_inode,
2533 .free_inode = btrfs_free_inode,
2534 .statfs = btrfs_statfs,
2535 .remount_fs = btrfs_remount,
2536 .freeze_fs = btrfs_freeze,
2537 .unfreeze_fs = btrfs_unfreeze,
2540 static const struct file_operations btrfs_ctl_fops = {
2541 .open = btrfs_control_open,
2542 .unlocked_ioctl = btrfs_control_ioctl,
2543 .compat_ioctl = compat_ptr_ioctl,
2544 .owner = THIS_MODULE,
2545 .llseek = noop_llseek,
2548 static struct miscdevice btrfs_misc = {
2549 .minor = BTRFS_MINOR,
2550 .name = "btrfs-control",
2551 .fops = &btrfs_ctl_fops
2554 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2555 MODULE_ALIAS("devname:btrfs-control");
2557 static int __init btrfs_interface_init(void)
2559 return misc_register(&btrfs_misc);
2562 static __cold void btrfs_interface_exit(void)
2564 misc_deregister(&btrfs_misc);
2567 static void __init btrfs_print_mod_info(void)
2569 static const char options[] = ""
2570 #ifdef CONFIG_BTRFS_DEBUG
2573 #ifdef CONFIG_BTRFS_ASSERT
2576 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2577 ", integrity-checker=on"
2579 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2582 #ifdef CONFIG_BLK_DEV_ZONED
2588 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2591 static int __init init_btrfs_fs(void)
2597 err = btrfs_init_sysfs();
2601 btrfs_init_compress();
2603 err = btrfs_init_cachep();
2607 err = extent_io_init();
2611 err = extent_state_cache_init();
2613 goto free_extent_io;
2615 err = extent_map_init();
2617 goto free_extent_state_cache;
2619 err = ordered_data_init();
2621 goto free_extent_map;
2623 err = btrfs_delayed_inode_init();
2625 goto free_ordered_data;
2627 err = btrfs_auto_defrag_init();
2629 goto free_delayed_inode;
2631 err = btrfs_delayed_ref_init();
2633 goto free_auto_defrag;
2635 err = btrfs_prelim_ref_init();
2637 goto free_delayed_ref;
2639 err = btrfs_end_io_wq_init();
2641 goto free_prelim_ref;
2643 err = btrfs_interface_init();
2645 goto free_end_io_wq;
2647 btrfs_print_mod_info();
2649 err = btrfs_run_sanity_tests();
2651 goto unregister_ioctl;
2653 err = register_filesystem(&btrfs_fs_type);
2655 goto unregister_ioctl;
2660 btrfs_interface_exit();
2662 btrfs_end_io_wq_exit();
2664 btrfs_prelim_ref_exit();
2666 btrfs_delayed_ref_exit();
2668 btrfs_auto_defrag_exit();
2670 btrfs_delayed_inode_exit();
2672 ordered_data_exit();
2675 free_extent_state_cache:
2676 extent_state_cache_exit();
2680 btrfs_destroy_cachep();
2682 btrfs_exit_compress();
2688 static void __exit exit_btrfs_fs(void)
2690 btrfs_destroy_cachep();
2691 btrfs_delayed_ref_exit();
2692 btrfs_auto_defrag_exit();
2693 btrfs_delayed_inode_exit();
2694 btrfs_prelim_ref_exit();
2695 ordered_data_exit();
2697 extent_state_cache_exit();
2699 btrfs_interface_exit();
2700 btrfs_end_io_wq_exit();
2701 unregister_filesystem(&btrfs_fs_type);
2703 btrfs_cleanup_fs_uuids();
2704 btrfs_exit_compress();
2707 late_initcall(init_btrfs_fs);
2708 module_exit(exit_btrfs_fs)
2710 MODULE_LICENSE("GPL");
2711 MODULE_SOFTDEP("pre: crc32c");
2712 MODULE_SOFTDEP("pre: xxhash64");
2713 MODULE_SOFTDEP("pre: sha256");
2714 MODULE_SOFTDEP("pre: blake2b-256");