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"
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);
72 #define STATE_STRING_PREFACE ": state "
73 #define STATE_STRING_BUF_LEN (sizeof(STATE_STRING_PREFACE) + BTRFS_FS_STATE_COUNT)
76 * Characters to print to indicate error conditions or uncommon filesystem state.
79 static const char fs_state_chars[] = {
80 [BTRFS_FS_STATE_ERROR] = 'E',
81 [BTRFS_FS_STATE_REMOUNTING] = 'M',
82 [BTRFS_FS_STATE_RO] = 0,
83 [BTRFS_FS_STATE_TRANS_ABORTED] = 'A',
84 [BTRFS_FS_STATE_DEV_REPLACING] = 'R',
85 [BTRFS_FS_STATE_DUMMY_FS_INFO] = 0,
86 [BTRFS_FS_STATE_NO_CSUMS] = 'C',
87 [BTRFS_FS_STATE_LOG_CLEANUP_ERROR] = 'L',
90 static void btrfs_state_to_string(const struct btrfs_fs_info *info, char *buf)
93 bool states_printed = false;
94 unsigned long fs_state = READ_ONCE(info->fs_state);
97 memcpy(curr, STATE_STRING_PREFACE, sizeof(STATE_STRING_PREFACE));
98 curr += sizeof(STATE_STRING_PREFACE) - 1;
100 for_each_set_bit(bit, &fs_state, sizeof(fs_state)) {
101 WARN_ON_ONCE(bit >= BTRFS_FS_STATE_COUNT);
102 if ((bit < BTRFS_FS_STATE_COUNT) && fs_state_chars[bit]) {
103 *curr++ = fs_state_chars[bit];
104 states_printed = true;
108 /* If no states were printed, reset the buffer */
117 * Generally the error codes correspond to their respective errors, but there
118 * are a few special cases.
120 * EUCLEAN: Any sort of corruption that we encounter. The tree-checker for
121 * instance will return EUCLEAN if any of the blocks are corrupted in
122 * a way that is problematic. We want to reserve EUCLEAN for these
123 * sort of corruptions.
125 * EROFS: If we check BTRFS_FS_STATE_ERROR and fail out with a return error, we
126 * need to use EROFS for this case. We will have no idea of the
127 * original failure, that will have been reported at the time we tripped
128 * over the error. Each subsequent error that doesn't have any context
129 * of the original error should use EROFS when handling BTRFS_FS_STATE_ERROR.
131 const char * __attribute_const__ btrfs_decode_error(int errno)
133 char *errstr = "unknown";
136 case -ENOENT: /* -2 */
137 errstr = "No such entry";
140 errstr = "IO failure";
142 case -ENOMEM: /* -12*/
143 errstr = "Out of memory";
145 case -EEXIST: /* -17 */
146 errstr = "Object already exists";
148 case -ENOSPC: /* -28 */
149 errstr = "No space left";
151 case -EROFS: /* -30 */
152 errstr = "Readonly filesystem";
154 case -EOPNOTSUPP: /* -95 */
155 errstr = "Operation not supported";
157 case -EUCLEAN: /* -117 */
158 errstr = "Filesystem corrupted";
160 case -EDQUOT: /* -122 */
161 errstr = "Quota exceeded";
169 * __btrfs_handle_fs_error decodes expected errors from the caller and
170 * invokes the appropriate error response.
173 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
174 unsigned int line, int errno, const char *fmt, ...)
176 struct super_block *sb = fs_info->sb;
178 char statestr[STATE_STRING_BUF_LEN];
183 * Special case: if the error is EROFS, and we're already
184 * under SB_RDONLY, then it is safe here.
186 if (errno == -EROFS && sb_rdonly(sb))
190 errstr = btrfs_decode_error(errno);
191 btrfs_state_to_string(fs_info, statestr);
193 struct va_format vaf;
200 pr_crit("BTRFS: error (device %s%s) in %s:%d: errno=%d %s (%pV)\n",
201 sb->s_id, statestr, function, line, errno, errstr, &vaf);
204 pr_crit("BTRFS: error (device %s%s) in %s:%d: errno=%d %s\n",
205 sb->s_id, statestr, function, line, errno, errstr);
210 * Today we only save the error info to memory. Long term we'll
211 * also send it down to the disk
213 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
215 /* Don't go through full error handling during mount */
216 if (!(sb->s_flags & SB_BORN))
222 btrfs_discard_stop(fs_info);
224 /* btrfs handle error by forcing the filesystem readonly */
225 btrfs_set_sb_rdonly(sb);
226 btrfs_info(fs_info, "forced readonly");
228 * Note that a running device replace operation is not canceled here
229 * although there is no way to update the progress. It would add the
230 * risk of a deadlock, therefore the canceling is omitted. The only
231 * penalty is that some I/O remains active until the procedure
232 * completes. The next time when the filesystem is mounted writable
233 * again, the device replace operation continues.
238 static const char * const logtypes[] = {
251 * Use one ratelimit state per log level so that a flood of less important
252 * messages doesn't cause more important ones to be dropped.
254 static struct ratelimit_state printk_limits[] = {
255 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
256 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
257 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
258 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
259 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
260 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
261 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
262 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
265 void __cold _btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
267 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
268 struct va_format vaf;
271 const char *type = logtypes[4];
272 struct ratelimit_state *ratelimit = &printk_limits[4];
276 while ((kern_level = printk_get_level(fmt)) != 0) {
277 size_t size = printk_skip_level(fmt) - fmt;
279 if (kern_level >= '0' && kern_level <= '7') {
280 memcpy(lvl, fmt, size);
282 type = logtypes[kern_level - '0'];
283 ratelimit = &printk_limits[kern_level - '0'];
291 if (__ratelimit(ratelimit)) {
293 char statestr[STATE_STRING_BUF_LEN];
295 btrfs_state_to_string(fs_info, statestr);
296 _printk("%sBTRFS %s (device %s%s): %pV\n", lvl, type,
297 fs_info->sb->s_id, statestr, &vaf);
299 _printk("%sBTRFS %s: %pV\n", lvl, type, &vaf);
307 #if BITS_PER_LONG == 32
308 void __cold btrfs_warn_32bit_limit(struct btrfs_fs_info *fs_info)
310 if (!test_and_set_bit(BTRFS_FS_32BIT_WARN, &fs_info->flags)) {
311 btrfs_warn(fs_info, "reaching 32bit limit for logical addresses");
313 "due to page cache limit on 32bit systems, btrfs can't access metadata at or beyond %lluT",
314 BTRFS_32BIT_MAX_FILE_SIZE >> 40);
316 "please consider upgrading to 64bit kernel/hardware");
320 void __cold btrfs_err_32bit_limit(struct btrfs_fs_info *fs_info)
322 if (!test_and_set_bit(BTRFS_FS_32BIT_ERROR, &fs_info->flags)) {
323 btrfs_err(fs_info, "reached 32bit limit for logical addresses");
325 "due to page cache limit on 32bit systems, metadata beyond %lluT can't be accessed",
326 BTRFS_32BIT_MAX_FILE_SIZE >> 40);
328 "please consider upgrading to 64bit kernel/hardware");
334 * We only mark the transaction aborted and then set the file system read-only.
335 * This will prevent new transactions from starting or trying to join this
338 * This means that error recovery at the call site is limited to freeing
339 * any local memory allocations and passing the error code up without
340 * further cleanup. The transaction should complete as it normally would
341 * in the call path but will return -EIO.
343 * We'll complete the cleanup in btrfs_end_transaction and
344 * btrfs_commit_transaction.
347 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
348 const char *function,
349 unsigned int line, int errno)
351 struct btrfs_fs_info *fs_info = trans->fs_info;
353 WRITE_ONCE(trans->aborted, errno);
354 WRITE_ONCE(trans->transaction->aborted, errno);
355 /* Wake up anybody who may be waiting on this transaction */
356 wake_up(&fs_info->transaction_wait);
357 wake_up(&fs_info->transaction_blocked_wait);
358 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
361 * __btrfs_panic decodes unexpected, fatal errors from the caller,
362 * issues an alert, and either panics or BUGs, depending on mount options.
365 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
366 unsigned int line, int errno, const char *fmt, ...)
368 char *s_id = "<unknown>";
370 struct va_format vaf = { .fmt = fmt };
374 s_id = fs_info->sb->s_id;
379 errstr = btrfs_decode_error(errno);
380 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
381 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
382 s_id, function, line, &vaf, errno, errstr);
384 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
385 function, line, &vaf, errno, errstr);
387 /* Caller calls BUG() */
390 static void btrfs_put_super(struct super_block *sb)
392 close_ctree(btrfs_sb(sb));
401 Opt_compress_force_type,
406 Opt_flushoncommit, Opt_noflushoncommit,
408 Opt_barrier, Opt_nobarrier,
409 Opt_datacow, Opt_nodatacow,
410 Opt_datasum, Opt_nodatasum,
411 Opt_defrag, Opt_nodefrag,
412 Opt_discard, Opt_nodiscard,
416 Opt_rescan_uuid_tree,
418 Opt_space_cache, Opt_no_space_cache,
419 Opt_space_cache_version,
421 Opt_ssd_spread, Opt_nossd_spread,
426 Opt_treelog, Opt_notreelog,
427 Opt_user_subvol_rm_allowed,
437 /* Deprecated options */
439 Opt_inode_cache, Opt_noinode_cache,
441 /* Debugging options */
443 Opt_check_integrity_including_extent_data,
444 Opt_check_integrity_print_mask,
445 Opt_enospc_debug, Opt_noenospc_debug,
446 #ifdef CONFIG_BTRFS_DEBUG
447 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
449 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
455 static const match_table_t tokens = {
457 {Opt_noacl, "noacl"},
458 {Opt_clear_cache, "clear_cache"},
459 {Opt_commit_interval, "commit=%u"},
460 {Opt_compress, "compress"},
461 {Opt_compress_type, "compress=%s"},
462 {Opt_compress_force, "compress-force"},
463 {Opt_compress_force_type, "compress-force=%s"},
464 {Opt_degraded, "degraded"},
465 {Opt_device, "device=%s"},
466 {Opt_fatal_errors, "fatal_errors=%s"},
467 {Opt_flushoncommit, "flushoncommit"},
468 {Opt_noflushoncommit, "noflushoncommit"},
469 {Opt_inode_cache, "inode_cache"},
470 {Opt_noinode_cache, "noinode_cache"},
471 {Opt_max_inline, "max_inline=%s"},
472 {Opt_barrier, "barrier"},
473 {Opt_nobarrier, "nobarrier"},
474 {Opt_datacow, "datacow"},
475 {Opt_nodatacow, "nodatacow"},
476 {Opt_datasum, "datasum"},
477 {Opt_nodatasum, "nodatasum"},
478 {Opt_defrag, "autodefrag"},
479 {Opt_nodefrag, "noautodefrag"},
480 {Opt_discard, "discard"},
481 {Opt_discard_mode, "discard=%s"},
482 {Opt_nodiscard, "nodiscard"},
483 {Opt_norecovery, "norecovery"},
484 {Opt_ratio, "metadata_ratio=%u"},
485 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
486 {Opt_skip_balance, "skip_balance"},
487 {Opt_space_cache, "space_cache"},
488 {Opt_no_space_cache, "nospace_cache"},
489 {Opt_space_cache_version, "space_cache=%s"},
491 {Opt_nossd, "nossd"},
492 {Opt_ssd_spread, "ssd_spread"},
493 {Opt_nossd_spread, "nossd_spread"},
494 {Opt_subvol, "subvol=%s"},
495 {Opt_subvol_empty, "subvol="},
496 {Opt_subvolid, "subvolid=%s"},
497 {Opt_thread_pool, "thread_pool=%u"},
498 {Opt_treelog, "treelog"},
499 {Opt_notreelog, "notreelog"},
500 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
503 {Opt_rescue, "rescue=%s"},
504 /* Deprecated, with alias rescue=nologreplay */
505 {Opt_nologreplay, "nologreplay"},
506 /* Deprecated, with alias rescue=usebackuproot */
507 {Opt_usebackuproot, "usebackuproot"},
509 /* Deprecated options */
510 {Opt_recovery, "recovery"},
512 /* Debugging options */
513 {Opt_check_integrity, "check_int"},
514 {Opt_check_integrity_including_extent_data, "check_int_data"},
515 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
516 {Opt_enospc_debug, "enospc_debug"},
517 {Opt_noenospc_debug, "noenospc_debug"},
518 #ifdef CONFIG_BTRFS_DEBUG
519 {Opt_fragment_data, "fragment=data"},
520 {Opt_fragment_metadata, "fragment=metadata"},
521 {Opt_fragment_all, "fragment=all"},
523 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
524 {Opt_ref_verify, "ref_verify"},
529 static const match_table_t rescue_tokens = {
530 {Opt_usebackuproot, "usebackuproot"},
531 {Opt_nologreplay, "nologreplay"},
532 {Opt_ignorebadroots, "ignorebadroots"},
533 {Opt_ignorebadroots, "ibadroots"},
534 {Opt_ignoredatacsums, "ignoredatacsums"},
535 {Opt_ignoredatacsums, "idatacsums"},
536 {Opt_rescue_all, "all"},
540 static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
541 const char *opt_name)
543 if (fs_info->mount_opt & opt) {
544 btrfs_err(fs_info, "%s must be used with ro mount option",
551 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
556 substring_t args[MAX_OPT_ARGS];
559 opts = kstrdup(options, GFP_KERNEL);
564 while ((p = strsep(&opts, ":")) != NULL) {
569 token = match_token(p, rescue_tokens, args);
571 case Opt_usebackuproot:
573 "trying to use backup root at mount time");
574 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
576 case Opt_nologreplay:
577 btrfs_set_and_info(info, NOLOGREPLAY,
578 "disabling log replay at mount time");
580 case Opt_ignorebadroots:
581 btrfs_set_and_info(info, IGNOREBADROOTS,
582 "ignoring bad roots");
584 case Opt_ignoredatacsums:
585 btrfs_set_and_info(info, IGNOREDATACSUMS,
586 "ignoring data csums");
589 btrfs_info(info, "enabling all of the rescue options");
590 btrfs_set_and_info(info, IGNOREDATACSUMS,
591 "ignoring data csums");
592 btrfs_set_and_info(info, IGNOREBADROOTS,
593 "ignoring bad roots");
594 btrfs_set_and_info(info, NOLOGREPLAY,
595 "disabling log replay at mount time");
598 btrfs_info(info, "unrecognized rescue option '%s'", p);
612 * Regular mount options parser. Everything that is needed only when
613 * reading in a new superblock is parsed here.
614 * XXX JDM: This needs to be cleaned up for remount.
616 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
617 unsigned long new_flags)
619 substring_t args[MAX_OPT_ARGS];
624 bool compress_force = false;
625 enum btrfs_compression_type saved_compress_type;
626 int saved_compress_level;
627 bool saved_compress_force;
630 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
631 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
632 else if (btrfs_free_space_cache_v1_active(info)) {
633 if (btrfs_is_zoned(info)) {
635 "zoned: clearing existing space cache");
636 btrfs_set_super_cache_generation(info->super_copy, 0);
638 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
643 * Even the options are empty, we still need to do extra check
649 while ((p = strsep(&options, ",")) != NULL) {
654 token = match_token(p, tokens, args);
657 btrfs_info(info, "allowing degraded mounts");
658 btrfs_set_opt(info->mount_opt, DEGRADED);
661 case Opt_subvol_empty:
665 * These are parsed by btrfs_parse_subvol_options or
666 * btrfs_parse_device_options and can be ignored here.
670 btrfs_set_and_info(info, NODATASUM,
671 "setting nodatasum");
674 if (btrfs_test_opt(info, NODATASUM)) {
675 if (btrfs_test_opt(info, NODATACOW))
677 "setting datasum, datacow enabled");
679 btrfs_info(info, "setting datasum");
681 btrfs_clear_opt(info->mount_opt, NODATACOW);
682 btrfs_clear_opt(info->mount_opt, NODATASUM);
685 if (!btrfs_test_opt(info, NODATACOW)) {
686 if (!btrfs_test_opt(info, COMPRESS) ||
687 !btrfs_test_opt(info, FORCE_COMPRESS)) {
689 "setting nodatacow, compression disabled");
691 btrfs_info(info, "setting nodatacow");
694 btrfs_clear_opt(info->mount_opt, COMPRESS);
695 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
696 btrfs_set_opt(info->mount_opt, NODATACOW);
697 btrfs_set_opt(info->mount_opt, NODATASUM);
700 btrfs_clear_and_info(info, NODATACOW,
703 case Opt_compress_force:
704 case Opt_compress_force_type:
705 compress_force = true;
708 case Opt_compress_type:
709 saved_compress_type = btrfs_test_opt(info,
711 info->compress_type : BTRFS_COMPRESS_NONE;
712 saved_compress_force =
713 btrfs_test_opt(info, FORCE_COMPRESS);
714 saved_compress_level = info->compress_level;
715 if (token == Opt_compress ||
716 token == Opt_compress_force ||
717 strncmp(args[0].from, "zlib", 4) == 0) {
718 compress_type = "zlib";
720 info->compress_type = BTRFS_COMPRESS_ZLIB;
721 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
723 * args[0] contains uninitialized data since
724 * for these tokens we don't expect any
727 if (token != Opt_compress &&
728 token != Opt_compress_force)
729 info->compress_level =
730 btrfs_compress_str2level(
733 btrfs_set_opt(info->mount_opt, COMPRESS);
734 btrfs_clear_opt(info->mount_opt, NODATACOW);
735 btrfs_clear_opt(info->mount_opt, NODATASUM);
737 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
738 compress_type = "lzo";
739 info->compress_type = BTRFS_COMPRESS_LZO;
740 info->compress_level = 0;
741 btrfs_set_opt(info->mount_opt, COMPRESS);
742 btrfs_clear_opt(info->mount_opt, NODATACOW);
743 btrfs_clear_opt(info->mount_opt, NODATASUM);
744 btrfs_set_fs_incompat(info, COMPRESS_LZO);
746 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
747 compress_type = "zstd";
748 info->compress_type = BTRFS_COMPRESS_ZSTD;
749 info->compress_level =
750 btrfs_compress_str2level(
753 btrfs_set_opt(info->mount_opt, COMPRESS);
754 btrfs_clear_opt(info->mount_opt, NODATACOW);
755 btrfs_clear_opt(info->mount_opt, NODATASUM);
756 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
758 } else if (strncmp(args[0].from, "no", 2) == 0) {
759 compress_type = "no";
760 info->compress_level = 0;
761 info->compress_type = 0;
762 btrfs_clear_opt(info->mount_opt, COMPRESS);
763 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
764 compress_force = false;
767 btrfs_err(info, "unrecognized compression value %s",
773 if (compress_force) {
774 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
777 * If we remount from compress-force=xxx to
778 * compress=xxx, we need clear FORCE_COMPRESS
779 * flag, otherwise, there is no way for users
780 * to disable forcible compression separately.
782 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
784 if (no_compress == 1) {
785 btrfs_info(info, "use no compression");
786 } else if ((info->compress_type != saved_compress_type) ||
787 (compress_force != saved_compress_force) ||
788 (info->compress_level != saved_compress_level)) {
789 btrfs_info(info, "%s %s compression, level %d",
790 (compress_force) ? "force" : "use",
791 compress_type, info->compress_level);
793 compress_force = false;
796 btrfs_set_and_info(info, SSD,
797 "enabling ssd optimizations");
798 btrfs_clear_opt(info->mount_opt, NOSSD);
801 btrfs_set_and_info(info, SSD,
802 "enabling ssd optimizations");
803 btrfs_set_and_info(info, SSD_SPREAD,
804 "using spread ssd allocation scheme");
805 btrfs_clear_opt(info->mount_opt, NOSSD);
808 btrfs_set_opt(info->mount_opt, NOSSD);
809 btrfs_clear_and_info(info, SSD,
810 "not using ssd optimizations");
812 case Opt_nossd_spread:
813 btrfs_clear_and_info(info, SSD_SPREAD,
814 "not using spread ssd allocation scheme");
817 btrfs_clear_and_info(info, NOBARRIER,
818 "turning on barriers");
821 btrfs_set_and_info(info, NOBARRIER,
822 "turning off barriers");
824 case Opt_thread_pool:
825 ret = match_int(&args[0], &intarg);
827 btrfs_err(info, "unrecognized thread_pool value %s",
830 } else if (intarg == 0) {
831 btrfs_err(info, "invalid value 0 for thread_pool");
835 info->thread_pool_size = intarg;
838 num = match_strdup(&args[0]);
840 info->max_inline = memparse(num, NULL);
843 if (info->max_inline) {
844 info->max_inline = min_t(u64,
848 btrfs_info(info, "max_inline at %llu",
856 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
857 info->sb->s_flags |= SB_POSIXACL;
860 btrfs_err(info, "support for ACL not compiled in!");
865 info->sb->s_flags &= ~SB_POSIXACL;
868 btrfs_set_and_info(info, NOTREELOG,
869 "disabling tree log");
872 btrfs_clear_and_info(info, NOTREELOG,
873 "enabling tree log");
876 case Opt_nologreplay:
878 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
879 btrfs_set_and_info(info, NOLOGREPLAY,
880 "disabling log replay at mount time");
882 case Opt_flushoncommit:
883 btrfs_set_and_info(info, FLUSHONCOMMIT,
884 "turning on flush-on-commit");
886 case Opt_noflushoncommit:
887 btrfs_clear_and_info(info, FLUSHONCOMMIT,
888 "turning off flush-on-commit");
891 ret = match_int(&args[0], &intarg);
893 btrfs_err(info, "unrecognized metadata_ratio value %s",
897 info->metadata_ratio = intarg;
898 btrfs_info(info, "metadata ratio %u",
899 info->metadata_ratio);
902 case Opt_discard_mode:
903 if (token == Opt_discard ||
904 strcmp(args[0].from, "sync") == 0) {
905 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
906 btrfs_set_and_info(info, DISCARD_SYNC,
907 "turning on sync discard");
908 } else if (strcmp(args[0].from, "async") == 0) {
909 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
910 btrfs_set_and_info(info, DISCARD_ASYNC,
911 "turning on async discard");
913 btrfs_err(info, "unrecognized discard mode value %s",
920 btrfs_clear_and_info(info, DISCARD_SYNC,
921 "turning off discard");
922 btrfs_clear_and_info(info, DISCARD_ASYNC,
923 "turning off async discard");
925 case Opt_space_cache:
926 case Opt_space_cache_version:
928 * We already set FREE_SPACE_TREE above because we have
929 * compat_ro(FREE_SPACE_TREE) set, and we aren't going
930 * to allow v1 to be set for extent tree v2, simply
931 * ignore this setting if we're extent tree v2.
933 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
935 if (token == Opt_space_cache ||
936 strcmp(args[0].from, "v1") == 0) {
937 btrfs_clear_opt(info->mount_opt,
939 btrfs_set_and_info(info, SPACE_CACHE,
940 "enabling disk space caching");
941 } else if (strcmp(args[0].from, "v2") == 0) {
942 btrfs_clear_opt(info->mount_opt,
944 btrfs_set_and_info(info, FREE_SPACE_TREE,
945 "enabling free space tree");
947 btrfs_err(info, "unrecognized space_cache value %s",
953 case Opt_rescan_uuid_tree:
954 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
956 case Opt_no_space_cache:
958 * We cannot operate without the free space tree with
959 * extent tree v2, ignore this option.
961 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
963 if (btrfs_test_opt(info, SPACE_CACHE)) {
964 btrfs_clear_and_info(info, SPACE_CACHE,
965 "disabling disk space caching");
967 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
968 btrfs_clear_and_info(info, FREE_SPACE_TREE,
969 "disabling free space tree");
972 case Opt_inode_cache:
973 case Opt_noinode_cache:
975 "the 'inode_cache' option is deprecated and has no effect since 5.11");
977 case Opt_clear_cache:
979 * We cannot clear the free space tree with extent tree
980 * v2, ignore this option.
982 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
984 btrfs_set_and_info(info, CLEAR_CACHE,
985 "force clearing of disk cache");
987 case Opt_user_subvol_rm_allowed:
988 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
990 case Opt_enospc_debug:
991 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
993 case Opt_noenospc_debug:
994 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
997 btrfs_set_and_info(info, AUTO_DEFRAG,
998 "enabling auto defrag");
1001 btrfs_clear_and_info(info, AUTO_DEFRAG,
1002 "disabling auto defrag");
1005 case Opt_usebackuproot:
1007 "'%s' is deprecated, use 'rescue=usebackuproot' instead",
1008 token == Opt_recovery ? "recovery" :
1011 "trying to use backup root at mount time");
1012 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
1014 case Opt_skip_balance:
1015 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
1017 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1018 case Opt_check_integrity_including_extent_data:
1020 "enabling check integrity including extent data");
1021 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
1022 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
1024 case Opt_check_integrity:
1025 btrfs_info(info, "enabling check integrity");
1026 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
1028 case Opt_check_integrity_print_mask:
1029 ret = match_int(&args[0], &intarg);
1032 "unrecognized check_integrity_print_mask value %s",
1036 info->check_integrity_print_mask = intarg;
1037 btrfs_info(info, "check_integrity_print_mask 0x%x",
1038 info->check_integrity_print_mask);
1041 case Opt_check_integrity_including_extent_data:
1042 case Opt_check_integrity:
1043 case Opt_check_integrity_print_mask:
1045 "support for check_integrity* not compiled in!");
1049 case Opt_fatal_errors:
1050 if (strcmp(args[0].from, "panic") == 0) {
1051 btrfs_set_opt(info->mount_opt,
1052 PANIC_ON_FATAL_ERROR);
1053 } else if (strcmp(args[0].from, "bug") == 0) {
1054 btrfs_clear_opt(info->mount_opt,
1055 PANIC_ON_FATAL_ERROR);
1057 btrfs_err(info, "unrecognized fatal_errors value %s",
1063 case Opt_commit_interval:
1065 ret = match_int(&args[0], &intarg);
1067 btrfs_err(info, "unrecognized commit_interval value %s",
1074 "using default commit interval %us",
1075 BTRFS_DEFAULT_COMMIT_INTERVAL);
1076 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
1077 } else if (intarg > 300) {
1078 btrfs_warn(info, "excessive commit interval %d",
1081 info->commit_interval = intarg;
1084 ret = parse_rescue_options(info, args[0].from);
1086 btrfs_err(info, "unrecognized rescue value %s",
1091 #ifdef CONFIG_BTRFS_DEBUG
1092 case Opt_fragment_all:
1093 btrfs_info(info, "fragmenting all space");
1094 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1095 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
1097 case Opt_fragment_metadata:
1098 btrfs_info(info, "fragmenting metadata");
1099 btrfs_set_opt(info->mount_opt,
1102 case Opt_fragment_data:
1103 btrfs_info(info, "fragmenting data");
1104 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1107 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
1108 case Opt_ref_verify:
1109 btrfs_info(info, "doing ref verification");
1110 btrfs_set_opt(info->mount_opt, REF_VERIFY);
1114 btrfs_err(info, "unrecognized mount option '%s'", p);
1122 /* We're read-only, don't have to check. */
1123 if (new_flags & SB_RDONLY)
1126 if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
1127 check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
1128 check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
1131 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
1132 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
1133 !btrfs_test_opt(info, CLEAR_CACHE)) {
1134 btrfs_err(info, "cannot disable free space tree");
1139 ret = btrfs_check_mountopts_zoned(info);
1140 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
1141 btrfs_info(info, "disk space caching is enabled");
1142 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
1143 btrfs_info(info, "using free space tree");
1148 * Parse mount options that are required early in the mount process.
1150 * All other options will be parsed on much later in the mount process and
1151 * only when we need to allocate a new super block.
1153 static int btrfs_parse_device_options(const char *options, fmode_t flags,
1156 substring_t args[MAX_OPT_ARGS];
1157 char *device_name, *opts, *orig, *p;
1158 struct btrfs_device *device = NULL;
1161 lockdep_assert_held(&uuid_mutex);
1167 * strsep changes the string, duplicate it because btrfs_parse_options
1170 opts = kstrdup(options, GFP_KERNEL);
1175 while ((p = strsep(&opts, ",")) != NULL) {
1181 token = match_token(p, tokens, args);
1182 if (token == Opt_device) {
1183 device_name = match_strdup(&args[0]);
1188 device = btrfs_scan_one_device(device_name, flags,
1191 if (IS_ERR(device)) {
1192 error = PTR_ERR(device);
1204 * Parse mount options that are related to subvolume id
1206 * The value is later passed to mount_subvol()
1208 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1209 u64 *subvol_objectid)
1211 substring_t args[MAX_OPT_ARGS];
1212 char *opts, *orig, *p;
1220 * strsep changes the string, duplicate it because
1221 * btrfs_parse_device_options gets called later
1223 opts = kstrdup(options, GFP_KERNEL);
1228 while ((p = strsep(&opts, ",")) != NULL) {
1233 token = match_token(p, tokens, args);
1236 kfree(*subvol_name);
1237 *subvol_name = match_strdup(&args[0]);
1238 if (!*subvol_name) {
1244 error = match_u64(&args[0], &subvolid);
1248 /* we want the original fs_tree */
1250 subvolid = BTRFS_FS_TREE_OBJECTID;
1252 *subvol_objectid = subvolid;
1264 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1265 u64 subvol_objectid)
1267 struct btrfs_root *root = fs_info->tree_root;
1268 struct btrfs_root *fs_root = NULL;
1269 struct btrfs_root_ref *root_ref;
1270 struct btrfs_inode_ref *inode_ref;
1271 struct btrfs_key key;
1272 struct btrfs_path *path = NULL;
1273 char *name = NULL, *ptr;
1278 path = btrfs_alloc_path();
1284 name = kmalloc(PATH_MAX, GFP_KERNEL);
1289 ptr = name + PATH_MAX - 1;
1293 * Walk up the subvolume trees in the tree of tree roots by root
1294 * backrefs until we hit the top-level subvolume.
1296 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1297 key.objectid = subvol_objectid;
1298 key.type = BTRFS_ROOT_BACKREF_KEY;
1299 key.offset = (u64)-1;
1301 ret = btrfs_search_backwards(root, &key, path);
1304 } else if (ret > 0) {
1309 subvol_objectid = key.offset;
1311 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1312 struct btrfs_root_ref);
1313 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1316 ret = -ENAMETOOLONG;
1319 read_extent_buffer(path->nodes[0], ptr + 1,
1320 (unsigned long)(root_ref + 1), len);
1322 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1323 btrfs_release_path(path);
1325 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1326 if (IS_ERR(fs_root)) {
1327 ret = PTR_ERR(fs_root);
1333 * Walk up the filesystem tree by inode refs until we hit the
1336 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1337 key.objectid = dirid;
1338 key.type = BTRFS_INODE_REF_KEY;
1339 key.offset = (u64)-1;
1341 ret = btrfs_search_backwards(fs_root, &key, path);
1344 } else if (ret > 0) {
1351 inode_ref = btrfs_item_ptr(path->nodes[0],
1353 struct btrfs_inode_ref);
1354 len = btrfs_inode_ref_name_len(path->nodes[0],
1358 ret = -ENAMETOOLONG;
1361 read_extent_buffer(path->nodes[0], ptr + 1,
1362 (unsigned long)(inode_ref + 1), len);
1364 btrfs_release_path(path);
1366 btrfs_put_root(fs_root);
1370 btrfs_free_path(path);
1371 if (ptr == name + PATH_MAX - 1) {
1375 memmove(name, ptr, name + PATH_MAX - ptr);
1380 btrfs_put_root(fs_root);
1381 btrfs_free_path(path);
1383 return ERR_PTR(ret);
1386 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1388 struct btrfs_root *root = fs_info->tree_root;
1389 struct btrfs_dir_item *di;
1390 struct btrfs_path *path;
1391 struct btrfs_key location;
1394 path = btrfs_alloc_path();
1399 * Find the "default" dir item which points to the root item that we
1400 * will mount by default if we haven't been given a specific subvolume
1403 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1404 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1406 btrfs_free_path(path);
1411 * Ok the default dir item isn't there. This is weird since
1412 * it's always been there, but don't freak out, just try and
1413 * mount the top-level subvolume.
1415 btrfs_free_path(path);
1416 *objectid = BTRFS_FS_TREE_OBJECTID;
1420 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1421 btrfs_free_path(path);
1422 *objectid = location.objectid;
1426 static int btrfs_fill_super(struct super_block *sb,
1427 struct btrfs_fs_devices *fs_devices,
1430 struct inode *inode;
1431 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1434 sb->s_maxbytes = MAX_LFS_FILESIZE;
1435 sb->s_magic = BTRFS_SUPER_MAGIC;
1436 sb->s_op = &btrfs_super_ops;
1437 sb->s_d_op = &btrfs_dentry_operations;
1438 sb->s_export_op = &btrfs_export_ops;
1439 #ifdef CONFIG_FS_VERITY
1440 sb->s_vop = &btrfs_verityops;
1442 sb->s_xattr = btrfs_xattr_handlers;
1443 sb->s_time_gran = 1;
1444 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1445 sb->s_flags |= SB_POSIXACL;
1447 sb->s_flags |= SB_I_VERSION;
1448 sb->s_iflags |= SB_I_CGROUPWB;
1450 err = super_setup_bdi(sb);
1452 btrfs_err(fs_info, "super_setup_bdi failed");
1456 err = open_ctree(sb, fs_devices, (char *)data);
1458 btrfs_err(fs_info, "open_ctree failed");
1462 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1463 if (IS_ERR(inode)) {
1464 err = PTR_ERR(inode);
1468 sb->s_root = d_make_root(inode);
1474 sb->s_flags |= SB_ACTIVE;
1478 close_ctree(fs_info);
1482 int btrfs_sync_fs(struct super_block *sb, int wait)
1484 struct btrfs_trans_handle *trans;
1485 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1486 struct btrfs_root *root = fs_info->tree_root;
1488 trace_btrfs_sync_fs(fs_info, wait);
1491 filemap_flush(fs_info->btree_inode->i_mapping);
1495 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1497 trans = btrfs_attach_transaction_barrier(root);
1498 if (IS_ERR(trans)) {
1499 /* no transaction, don't bother */
1500 if (PTR_ERR(trans) == -ENOENT) {
1502 * Exit unless we have some pending changes
1503 * that need to go through commit
1505 if (fs_info->pending_changes == 0)
1508 * A non-blocking test if the fs is frozen. We must not
1509 * start a new transaction here otherwise a deadlock
1510 * happens. The pending operations are delayed to the
1511 * next commit after thawing.
1513 if (sb_start_write_trylock(sb))
1517 trans = btrfs_start_transaction(root, 0);
1520 return PTR_ERR(trans);
1522 return btrfs_commit_transaction(trans);
1525 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1527 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1531 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1533 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1534 const char *compress_type;
1535 const char *subvol_name;
1536 bool printed = false;
1538 if (btrfs_test_opt(info, DEGRADED))
1539 seq_puts(seq, ",degraded");
1540 if (btrfs_test_opt(info, NODATASUM))
1541 seq_puts(seq, ",nodatasum");
1542 if (btrfs_test_opt(info, NODATACOW))
1543 seq_puts(seq, ",nodatacow");
1544 if (btrfs_test_opt(info, NOBARRIER))
1545 seq_puts(seq, ",nobarrier");
1546 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1547 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1548 if (info->thread_pool_size != min_t(unsigned long,
1549 num_online_cpus() + 2, 8))
1550 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1551 if (btrfs_test_opt(info, COMPRESS)) {
1552 compress_type = btrfs_compress_type2str(info->compress_type);
1553 if (btrfs_test_opt(info, FORCE_COMPRESS))
1554 seq_printf(seq, ",compress-force=%s", compress_type);
1556 seq_printf(seq, ",compress=%s", compress_type);
1557 if (info->compress_level)
1558 seq_printf(seq, ":%d", info->compress_level);
1560 if (btrfs_test_opt(info, NOSSD))
1561 seq_puts(seq, ",nossd");
1562 if (btrfs_test_opt(info, SSD_SPREAD))
1563 seq_puts(seq, ",ssd_spread");
1564 else if (btrfs_test_opt(info, SSD))
1565 seq_puts(seq, ",ssd");
1566 if (btrfs_test_opt(info, NOTREELOG))
1567 seq_puts(seq, ",notreelog");
1568 if (btrfs_test_opt(info, NOLOGREPLAY))
1569 print_rescue_option(seq, "nologreplay", &printed);
1570 if (btrfs_test_opt(info, USEBACKUPROOT))
1571 print_rescue_option(seq, "usebackuproot", &printed);
1572 if (btrfs_test_opt(info, IGNOREBADROOTS))
1573 print_rescue_option(seq, "ignorebadroots", &printed);
1574 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1575 print_rescue_option(seq, "ignoredatacsums", &printed);
1576 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1577 seq_puts(seq, ",flushoncommit");
1578 if (btrfs_test_opt(info, DISCARD_SYNC))
1579 seq_puts(seq, ",discard");
1580 if (btrfs_test_opt(info, DISCARD_ASYNC))
1581 seq_puts(seq, ",discard=async");
1582 if (!(info->sb->s_flags & SB_POSIXACL))
1583 seq_puts(seq, ",noacl");
1584 if (btrfs_free_space_cache_v1_active(info))
1585 seq_puts(seq, ",space_cache");
1586 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1587 seq_puts(seq, ",space_cache=v2");
1589 seq_puts(seq, ",nospace_cache");
1590 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1591 seq_puts(seq, ",rescan_uuid_tree");
1592 if (btrfs_test_opt(info, CLEAR_CACHE))
1593 seq_puts(seq, ",clear_cache");
1594 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1595 seq_puts(seq, ",user_subvol_rm_allowed");
1596 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1597 seq_puts(seq, ",enospc_debug");
1598 if (btrfs_test_opt(info, AUTO_DEFRAG))
1599 seq_puts(seq, ",autodefrag");
1600 if (btrfs_test_opt(info, SKIP_BALANCE))
1601 seq_puts(seq, ",skip_balance");
1602 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1603 if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1604 seq_puts(seq, ",check_int_data");
1605 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1606 seq_puts(seq, ",check_int");
1607 if (info->check_integrity_print_mask)
1608 seq_printf(seq, ",check_int_print_mask=%d",
1609 info->check_integrity_print_mask);
1611 if (info->metadata_ratio)
1612 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1613 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1614 seq_puts(seq, ",fatal_errors=panic");
1615 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1616 seq_printf(seq, ",commit=%u", info->commit_interval);
1617 #ifdef CONFIG_BTRFS_DEBUG
1618 if (btrfs_test_opt(info, FRAGMENT_DATA))
1619 seq_puts(seq, ",fragment=data");
1620 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1621 seq_puts(seq, ",fragment=metadata");
1623 if (btrfs_test_opt(info, REF_VERIFY))
1624 seq_puts(seq, ",ref_verify");
1625 seq_printf(seq, ",subvolid=%llu",
1626 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1627 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1628 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1629 if (!IS_ERR(subvol_name)) {
1630 seq_puts(seq, ",subvol=");
1631 seq_escape(seq, subvol_name, " \t\n\\");
1637 static int btrfs_test_super(struct super_block *s, void *data)
1639 struct btrfs_fs_info *p = data;
1640 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1642 return fs_info->fs_devices == p->fs_devices;
1645 static int btrfs_set_super(struct super_block *s, void *data)
1647 int err = set_anon_super(s, data);
1649 s->s_fs_info = data;
1654 * subvolumes are identified by ino 256
1656 static inline int is_subvolume_inode(struct inode *inode)
1658 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1663 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1664 struct vfsmount *mnt)
1666 struct dentry *root;
1670 if (!subvol_objectid) {
1671 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1674 root = ERR_PTR(ret);
1678 subvol_name = btrfs_get_subvol_name_from_objectid(
1679 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1680 if (IS_ERR(subvol_name)) {
1681 root = ERR_CAST(subvol_name);
1688 root = mount_subtree(mnt, subvol_name);
1689 /* mount_subtree() drops our reference on the vfsmount. */
1692 if (!IS_ERR(root)) {
1693 struct super_block *s = root->d_sb;
1694 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1695 struct inode *root_inode = d_inode(root);
1696 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1699 if (!is_subvolume_inode(root_inode)) {
1700 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1704 if (subvol_objectid && root_objectid != subvol_objectid) {
1706 * This will also catch a race condition where a
1707 * subvolume which was passed by ID is renamed and
1708 * another subvolume is renamed over the old location.
1711 "subvol '%s' does not match subvolid %llu",
1712 subvol_name, subvol_objectid);
1717 root = ERR_PTR(ret);
1718 deactivate_locked_super(s);
1729 * Find a superblock for the given device / mount point.
1731 * Note: This is based on mount_bdev from fs/super.c with a few additions
1732 * for multiple device setup. Make sure to keep it in sync.
1734 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1735 int flags, const char *device_name, void *data)
1737 struct block_device *bdev = NULL;
1738 struct super_block *s;
1739 struct btrfs_device *device = NULL;
1740 struct btrfs_fs_devices *fs_devices = NULL;
1741 struct btrfs_fs_info *fs_info = NULL;
1742 void *new_sec_opts = NULL;
1743 fmode_t mode = FMODE_READ;
1746 if (!(flags & SB_RDONLY))
1747 mode |= FMODE_WRITE;
1750 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1752 return ERR_PTR(error);
1756 * Setup a dummy root and fs_info for test/set super. This is because
1757 * we don't actually fill this stuff out until open_ctree, but we need
1758 * then open_ctree will properly initialize the file system specific
1759 * settings later. btrfs_init_fs_info initializes the static elements
1760 * of the fs_info (locks and such) to make cleanup easier if we find a
1761 * superblock with our given fs_devices later on at sget() time.
1763 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1766 goto error_sec_opts;
1768 btrfs_init_fs_info(fs_info);
1770 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1771 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1772 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1777 mutex_lock(&uuid_mutex);
1778 error = btrfs_parse_device_options(data, mode, fs_type);
1780 mutex_unlock(&uuid_mutex);
1784 device = btrfs_scan_one_device(device_name, mode, fs_type);
1785 if (IS_ERR(device)) {
1786 mutex_unlock(&uuid_mutex);
1787 error = PTR_ERR(device);
1791 fs_devices = device->fs_devices;
1792 fs_info->fs_devices = fs_devices;
1794 error = btrfs_open_devices(fs_devices, mode, fs_type);
1795 mutex_unlock(&uuid_mutex);
1799 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1801 goto error_close_devices;
1804 bdev = fs_devices->latest_dev->bdev;
1805 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1809 goto error_close_devices;
1813 btrfs_close_devices(fs_devices);
1814 btrfs_free_fs_info(fs_info);
1815 if ((flags ^ s->s_flags) & SB_RDONLY)
1818 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1819 shrinker_debugfs_rename(&s->s_shrink, "sb-%s:%s", fs_type->name,
1821 btrfs_sb(s)->bdev_holder = fs_type;
1822 if (!strstr(crc32c_impl(), "generic"))
1823 set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1824 error = btrfs_fill_super(s, fs_devices, data);
1827 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1828 security_free_mnt_opts(&new_sec_opts);
1830 deactivate_locked_super(s);
1831 return ERR_PTR(error);
1834 return dget(s->s_root);
1836 error_close_devices:
1837 btrfs_close_devices(fs_devices);
1839 btrfs_free_fs_info(fs_info);
1841 security_free_mnt_opts(&new_sec_opts);
1842 return ERR_PTR(error);
1846 * Mount function which is called by VFS layer.
1848 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1849 * which needs vfsmount* of device's root (/). This means device's root has to
1850 * be mounted internally in any case.
1853 * 1. Parse subvol id related options for later use in mount_subvol().
1855 * 2. Mount device's root (/) by calling vfs_kern_mount().
1857 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1858 * first place. In order to avoid calling btrfs_mount() again, we use
1859 * different file_system_type which is not registered to VFS by
1860 * register_filesystem() (btrfs_root_fs_type). As a result,
1861 * btrfs_mount_root() is called. The return value will be used by
1862 * mount_subtree() in mount_subvol().
1864 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1865 * "btrfs subvolume set-default", mount_subvol() is called always.
1867 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1868 const char *device_name, void *data)
1870 struct vfsmount *mnt_root;
1871 struct dentry *root;
1872 char *subvol_name = NULL;
1873 u64 subvol_objectid = 0;
1876 error = btrfs_parse_subvol_options(data, &subvol_name,
1880 return ERR_PTR(error);
1883 /* mount device's root (/) */
1884 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1885 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1886 if (flags & SB_RDONLY) {
1887 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1888 flags & ~SB_RDONLY, device_name, data);
1890 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1891 flags | SB_RDONLY, device_name, data);
1892 if (IS_ERR(mnt_root)) {
1893 root = ERR_CAST(mnt_root);
1898 down_write(&mnt_root->mnt_sb->s_umount);
1899 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1900 up_write(&mnt_root->mnt_sb->s_umount);
1902 root = ERR_PTR(error);
1909 if (IS_ERR(mnt_root)) {
1910 root = ERR_CAST(mnt_root);
1915 /* mount_subvol() will free subvol_name and mnt_root */
1916 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1922 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1923 u32 new_pool_size, u32 old_pool_size)
1925 if (new_pool_size == old_pool_size)
1928 fs_info->thread_pool_size = new_pool_size;
1930 btrfs_info(fs_info, "resize thread pool %d -> %d",
1931 old_pool_size, new_pool_size);
1933 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1934 btrfs_workqueue_set_max(fs_info->hipri_workers, new_pool_size);
1935 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1936 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1937 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1938 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1939 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1942 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1943 unsigned long old_opts, int flags)
1945 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1946 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1947 (flags & SB_RDONLY))) {
1948 /* wait for any defraggers to finish */
1949 wait_event(fs_info->transaction_wait,
1950 (atomic_read(&fs_info->defrag_running) == 0));
1951 if (flags & SB_RDONLY)
1952 sync_filesystem(fs_info->sb);
1956 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1957 unsigned long old_opts)
1959 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1962 * We need to cleanup all defragable inodes if the autodefragment is
1963 * close or the filesystem is read only.
1965 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1966 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1967 btrfs_cleanup_defrag_inodes(fs_info);
1970 /* If we toggled discard async */
1971 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1972 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1973 btrfs_discard_resume(fs_info);
1974 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1975 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1976 btrfs_discard_cleanup(fs_info);
1978 /* If we toggled space cache */
1979 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1980 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1983 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1985 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1986 unsigned old_flags = sb->s_flags;
1987 unsigned long old_opts = fs_info->mount_opt;
1988 unsigned long old_compress_type = fs_info->compress_type;
1989 u64 old_max_inline = fs_info->max_inline;
1990 u32 old_thread_pool_size = fs_info->thread_pool_size;
1991 u32 old_metadata_ratio = fs_info->metadata_ratio;
1994 sync_filesystem(sb);
1995 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1998 void *new_sec_opts = NULL;
2000 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
2002 ret = security_sb_remount(sb, new_sec_opts);
2003 security_free_mnt_opts(&new_sec_opts);
2008 ret = btrfs_parse_options(fs_info, data, *flags);
2012 /* V1 cache is not supported for subpage mount. */
2013 if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) {
2015 "v1 space cache is not supported for page size %lu with sectorsize %u",
2016 PAGE_SIZE, fs_info->sectorsize);
2020 btrfs_remount_begin(fs_info, old_opts, *flags);
2021 btrfs_resize_thread_pool(fs_info,
2022 fs_info->thread_pool_size, old_thread_pool_size);
2024 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
2025 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
2026 (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
2028 "remount supports changing free space tree only from ro to rw");
2029 /* Make sure free space cache options match the state on disk */
2030 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
2031 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
2032 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
2034 if (btrfs_free_space_cache_v1_active(fs_info)) {
2035 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
2036 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
2040 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
2043 if (*flags & SB_RDONLY) {
2045 * this also happens on 'umount -rf' or on shutdown, when
2046 * the filesystem is busy.
2048 cancel_work_sync(&fs_info->async_reclaim_work);
2049 cancel_work_sync(&fs_info->async_data_reclaim_work);
2051 btrfs_discard_cleanup(fs_info);
2053 /* wait for the uuid_scan task to finish */
2054 down(&fs_info->uuid_tree_rescan_sem);
2055 /* avoid complains from lockdep et al. */
2056 up(&fs_info->uuid_tree_rescan_sem);
2058 btrfs_set_sb_rdonly(sb);
2061 * Setting SB_RDONLY will put the cleaner thread to
2062 * sleep at the next loop if it's already active.
2063 * If it's already asleep, we'll leave unused block
2064 * groups on disk until we're mounted read-write again
2065 * unless we clean them up here.
2067 btrfs_delete_unused_bgs(fs_info);
2070 * The cleaner task could be already running before we set the
2071 * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
2072 * We must make sure that after we finish the remount, i.e. after
2073 * we call btrfs_commit_super(), the cleaner can no longer start
2074 * a transaction - either because it was dropping a dead root,
2075 * running delayed iputs or deleting an unused block group (the
2076 * cleaner picked a block group from the list of unused block
2077 * groups before we were able to in the previous call to
2078 * btrfs_delete_unused_bgs()).
2080 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
2081 TASK_UNINTERRUPTIBLE);
2084 * We've set the superblock to RO mode, so we might have made
2085 * the cleaner task sleep without running all pending delayed
2086 * iputs. Go through all the delayed iputs here, so that if an
2087 * unmount happens without remounting RW we don't end up at
2088 * finishing close_ctree() with a non-empty list of delayed
2091 btrfs_run_delayed_iputs(fs_info);
2093 btrfs_dev_replace_suspend_for_unmount(fs_info);
2094 btrfs_scrub_cancel(fs_info);
2095 btrfs_pause_balance(fs_info);
2098 * Pause the qgroup rescan worker if it is running. We don't want
2099 * it to be still running after we are in RO mode, as after that,
2100 * by the time we unmount, it might have left a transaction open,
2101 * so we would leak the transaction and/or crash.
2103 btrfs_qgroup_wait_for_completion(fs_info, false);
2105 ret = btrfs_commit_super(fs_info);
2109 if (BTRFS_FS_ERROR(fs_info)) {
2111 "Remounting read-write after error is not allowed");
2115 if (fs_info->fs_devices->rw_devices == 0) {
2120 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
2122 "too many missing devices, writable remount is not allowed");
2127 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
2129 "mount required to replay tree-log, cannot remount read-write");
2135 * NOTE: when remounting with a change that does writes, don't
2136 * put it anywhere above this point, as we are not sure to be
2137 * safe to write until we pass the above checks.
2139 ret = btrfs_start_pre_rw_mount(fs_info);
2143 btrfs_clear_sb_rdonly(sb);
2145 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
2149 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
2150 * since the absence of the flag means it can be toggled off by remount.
2152 *flags |= SB_I_VERSION;
2154 wake_up_process(fs_info->transaction_kthread);
2155 btrfs_remount_cleanup(fs_info, old_opts);
2156 btrfs_clear_oneshot_options(fs_info);
2157 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2162 /* We've hit an error - don't reset SB_RDONLY */
2164 old_flags |= SB_RDONLY;
2165 if (!(old_flags & SB_RDONLY))
2166 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
2167 sb->s_flags = old_flags;
2168 fs_info->mount_opt = old_opts;
2169 fs_info->compress_type = old_compress_type;
2170 fs_info->max_inline = old_max_inline;
2171 btrfs_resize_thread_pool(fs_info,
2172 old_thread_pool_size, fs_info->thread_pool_size);
2173 fs_info->metadata_ratio = old_metadata_ratio;
2174 btrfs_remount_cleanup(fs_info, old_opts);
2175 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2180 /* Used to sort the devices by max_avail(descending sort) */
2181 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
2183 const struct btrfs_device_info *dev_info1 = a;
2184 const struct btrfs_device_info *dev_info2 = b;
2186 if (dev_info1->max_avail > dev_info2->max_avail)
2188 else if (dev_info1->max_avail < dev_info2->max_avail)
2194 * sort the devices by max_avail, in which max free extent size of each device
2195 * is stored.(Descending Sort)
2197 static inline void btrfs_descending_sort_devices(
2198 struct btrfs_device_info *devices,
2201 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2202 btrfs_cmp_device_free_bytes, NULL);
2206 * The helper to calc the free space on the devices that can be used to store
2209 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2212 struct btrfs_device_info *devices_info;
2213 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2214 struct btrfs_device *device;
2217 u64 min_stripe_size;
2218 int num_stripes = 1;
2219 int i = 0, nr_devices;
2220 const struct btrfs_raid_attr *rattr;
2223 * We aren't under the device list lock, so this is racy-ish, but good
2224 * enough for our purposes.
2226 nr_devices = fs_info->fs_devices->open_devices;
2229 nr_devices = fs_info->fs_devices->open_devices;
2237 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2242 /* calc min stripe number for data space allocation */
2243 type = btrfs_data_alloc_profile(fs_info);
2244 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2246 if (type & BTRFS_BLOCK_GROUP_RAID0)
2247 num_stripes = nr_devices;
2248 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
2249 num_stripes = rattr->ncopies;
2250 else if (type & BTRFS_BLOCK_GROUP_RAID10)
2253 /* Adjust for more than 1 stripe per device */
2254 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2257 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2258 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2259 &device->dev_state) ||
2261 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2264 if (i >= nr_devices)
2267 avail_space = device->total_bytes - device->bytes_used;
2269 /* align with stripe_len */
2270 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2273 * Ensure we have at least min_stripe_size on top of the
2274 * reserved space on the device.
2276 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
2279 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
2281 devices_info[i].dev = device;
2282 devices_info[i].max_avail = avail_space;
2290 btrfs_descending_sort_devices(devices_info, nr_devices);
2294 while (nr_devices >= rattr->devs_min) {
2295 num_stripes = min(num_stripes, nr_devices);
2297 if (devices_info[i].max_avail >= min_stripe_size) {
2301 avail_space += devices_info[i].max_avail * num_stripes;
2302 alloc_size = devices_info[i].max_avail;
2303 for (j = i + 1 - num_stripes; j <= i; j++)
2304 devices_info[j].max_avail -= alloc_size;
2310 kfree(devices_info);
2311 *free_bytes = avail_space;
2316 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2318 * If there's a redundant raid level at DATA block groups, use the respective
2319 * multiplier to scale the sizes.
2321 * Unused device space usage is based on simulating the chunk allocator
2322 * algorithm that respects the device sizes and order of allocations. This is
2323 * a close approximation of the actual use but there are other factors that may
2324 * change the result (like a new metadata chunk).
2326 * If metadata is exhausted, f_bavail will be 0.
2328 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2330 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2331 struct btrfs_super_block *disk_super = fs_info->super_copy;
2332 struct btrfs_space_info *found;
2334 u64 total_free_data = 0;
2335 u64 total_free_meta = 0;
2336 u32 bits = fs_info->sectorsize_bits;
2337 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2338 unsigned factor = 1;
2339 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2344 list_for_each_entry(found, &fs_info->space_info, list) {
2345 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2348 total_free_data += found->disk_total - found->disk_used;
2350 btrfs_account_ro_block_groups_free_space(found);
2352 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2353 if (!list_empty(&found->block_groups[i]))
2354 factor = btrfs_bg_type_to_factor(
2355 btrfs_raid_array[i].bg_flag);
2360 * Metadata in mixed block goup profiles are accounted in data
2362 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2363 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2366 total_free_meta += found->disk_total -
2370 total_used += found->disk_used;
2373 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2374 buf->f_blocks >>= bits;
2375 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2377 /* Account global block reserve as used, it's in logical size already */
2378 spin_lock(&block_rsv->lock);
2379 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2380 if (buf->f_bfree >= block_rsv->size >> bits)
2381 buf->f_bfree -= block_rsv->size >> bits;
2384 spin_unlock(&block_rsv->lock);
2386 buf->f_bavail = div_u64(total_free_data, factor);
2387 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2390 buf->f_bavail += div_u64(total_free_data, factor);
2391 buf->f_bavail = buf->f_bavail >> bits;
2394 * We calculate the remaining metadata space minus global reserve. If
2395 * this is (supposedly) smaller than zero, there's no space. But this
2396 * does not hold in practice, the exhausted state happens where's still
2397 * some positive delta. So we apply some guesswork and compare the
2398 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2400 * We probably cannot calculate the exact threshold value because this
2401 * depends on the internal reservations requested by various
2402 * operations, so some operations that consume a few metadata will
2403 * succeed even if the Avail is zero. But this is better than the other
2409 * We only want to claim there's no available space if we can no longer
2410 * allocate chunks for our metadata profile and our global reserve will
2411 * not fit in the free metadata space. If we aren't ->full then we
2412 * still can allocate chunks and thus are fine using the currently
2413 * calculated f_bavail.
2415 if (!mixed && block_rsv->space_info->full &&
2416 total_free_meta - thresh < block_rsv->size)
2419 buf->f_type = BTRFS_SUPER_MAGIC;
2420 buf->f_bsize = dentry->d_sb->s_blocksize;
2421 buf->f_namelen = BTRFS_NAME_LEN;
2423 /* We treat it as constant endianness (it doesn't matter _which_)
2424 because we want the fsid to come out the same whether mounted
2425 on a big-endian or little-endian host */
2426 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2427 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2428 /* Mask in the root object ID too, to disambiguate subvols */
2429 buf->f_fsid.val[0] ^=
2430 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2431 buf->f_fsid.val[1] ^=
2432 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2437 static void btrfs_kill_super(struct super_block *sb)
2439 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2440 kill_anon_super(sb);
2441 btrfs_free_fs_info(fs_info);
2444 static struct file_system_type btrfs_fs_type = {
2445 .owner = THIS_MODULE,
2447 .mount = btrfs_mount,
2448 .kill_sb = btrfs_kill_super,
2449 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2452 static struct file_system_type btrfs_root_fs_type = {
2453 .owner = THIS_MODULE,
2455 .mount = btrfs_mount_root,
2456 .kill_sb = btrfs_kill_super,
2457 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2460 MODULE_ALIAS_FS("btrfs");
2462 static int btrfs_control_open(struct inode *inode, struct file *file)
2465 * The control file's private_data is used to hold the
2466 * transaction when it is started and is used to keep
2467 * track of whether a transaction is already in progress.
2469 file->private_data = NULL;
2474 * Used by /dev/btrfs-control for devices ioctls.
2476 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2479 struct btrfs_ioctl_vol_args *vol;
2480 struct btrfs_device *device = NULL;
2484 if (!capable(CAP_SYS_ADMIN))
2487 vol = memdup_user((void __user *)arg, sizeof(*vol));
2489 return PTR_ERR(vol);
2490 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2493 case BTRFS_IOC_SCAN_DEV:
2494 mutex_lock(&uuid_mutex);
2495 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2496 &btrfs_root_fs_type);
2497 ret = PTR_ERR_OR_ZERO(device);
2498 mutex_unlock(&uuid_mutex);
2500 case BTRFS_IOC_FORGET_DEV:
2501 if (vol->name[0] != 0) {
2502 ret = lookup_bdev(vol->name, &devt);
2506 ret = btrfs_forget_devices(devt);
2508 case BTRFS_IOC_DEVICES_READY:
2509 mutex_lock(&uuid_mutex);
2510 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2511 &btrfs_root_fs_type);
2512 if (IS_ERR(device)) {
2513 mutex_unlock(&uuid_mutex);
2514 ret = PTR_ERR(device);
2517 ret = !(device->fs_devices->num_devices ==
2518 device->fs_devices->total_devices);
2519 mutex_unlock(&uuid_mutex);
2521 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2522 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2530 static int btrfs_freeze(struct super_block *sb)
2532 struct btrfs_trans_handle *trans;
2533 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2534 struct btrfs_root *root = fs_info->tree_root;
2536 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2538 * We don't need a barrier here, we'll wait for any transaction that
2539 * could be in progress on other threads (and do delayed iputs that
2540 * we want to avoid on a frozen filesystem), or do the commit
2543 trans = btrfs_attach_transaction_barrier(root);
2544 if (IS_ERR(trans)) {
2545 /* no transaction, don't bother */
2546 if (PTR_ERR(trans) == -ENOENT)
2548 return PTR_ERR(trans);
2550 return btrfs_commit_transaction(trans);
2553 static int btrfs_unfreeze(struct super_block *sb)
2555 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2557 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2561 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2563 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2566 * There should be always a valid pointer in latest_dev, it may be stale
2567 * for a short moment in case it's being deleted but still valid until
2568 * the end of RCU grace period.
2571 seq_escape(m, rcu_str_deref(fs_info->fs_devices->latest_dev->name), " \t\n\\");
2577 static const struct super_operations btrfs_super_ops = {
2578 .drop_inode = btrfs_drop_inode,
2579 .evict_inode = btrfs_evict_inode,
2580 .put_super = btrfs_put_super,
2581 .sync_fs = btrfs_sync_fs,
2582 .show_options = btrfs_show_options,
2583 .show_devname = btrfs_show_devname,
2584 .alloc_inode = btrfs_alloc_inode,
2585 .destroy_inode = btrfs_destroy_inode,
2586 .free_inode = btrfs_free_inode,
2587 .statfs = btrfs_statfs,
2588 .remount_fs = btrfs_remount,
2589 .freeze_fs = btrfs_freeze,
2590 .unfreeze_fs = btrfs_unfreeze,
2593 static const struct file_operations btrfs_ctl_fops = {
2594 .open = btrfs_control_open,
2595 .unlocked_ioctl = btrfs_control_ioctl,
2596 .compat_ioctl = compat_ptr_ioctl,
2597 .owner = THIS_MODULE,
2598 .llseek = noop_llseek,
2601 static struct miscdevice btrfs_misc = {
2602 .minor = BTRFS_MINOR,
2603 .name = "btrfs-control",
2604 .fops = &btrfs_ctl_fops
2607 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2608 MODULE_ALIAS("devname:btrfs-control");
2610 static int __init btrfs_interface_init(void)
2612 return misc_register(&btrfs_misc);
2615 static __cold void btrfs_interface_exit(void)
2617 misc_deregister(&btrfs_misc);
2620 static void __init btrfs_print_mod_info(void)
2622 static const char options[] = ""
2623 #ifdef CONFIG_BTRFS_DEBUG
2626 #ifdef CONFIG_BTRFS_ASSERT
2629 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2630 ", integrity-checker=on"
2632 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2635 #ifdef CONFIG_BLK_DEV_ZONED
2640 #ifdef CONFIG_FS_VERITY
2646 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2649 static int __init init_btrfs_fs(void)
2655 err = btrfs_init_sysfs();
2659 btrfs_init_compress();
2661 err = btrfs_init_cachep();
2665 err = extent_io_init();
2669 err = extent_state_cache_init();
2671 goto free_extent_io;
2673 err = extent_map_init();
2675 goto free_extent_state_cache;
2677 err = ordered_data_init();
2679 goto free_extent_map;
2681 err = btrfs_delayed_inode_init();
2683 goto free_ordered_data;
2685 err = btrfs_auto_defrag_init();
2687 goto free_delayed_inode;
2689 err = btrfs_delayed_ref_init();
2691 goto free_auto_defrag;
2693 err = btrfs_prelim_ref_init();
2695 goto free_delayed_ref;
2697 err = btrfs_interface_init();
2699 goto free_prelim_ref;
2701 btrfs_print_mod_info();
2703 err = btrfs_run_sanity_tests();
2705 goto unregister_ioctl;
2707 err = register_filesystem(&btrfs_fs_type);
2709 goto unregister_ioctl;
2714 btrfs_interface_exit();
2716 btrfs_prelim_ref_exit();
2718 btrfs_delayed_ref_exit();
2720 btrfs_auto_defrag_exit();
2722 btrfs_delayed_inode_exit();
2724 ordered_data_exit();
2727 free_extent_state_cache:
2728 extent_state_cache_exit();
2732 btrfs_destroy_cachep();
2734 btrfs_exit_compress();
2740 static void __exit exit_btrfs_fs(void)
2742 btrfs_destroy_cachep();
2743 btrfs_delayed_ref_exit();
2744 btrfs_auto_defrag_exit();
2745 btrfs_delayed_inode_exit();
2746 btrfs_prelim_ref_exit();
2747 ordered_data_exit();
2749 extent_state_cache_exit();
2751 btrfs_interface_exit();
2752 unregister_filesystem(&btrfs_fs_type);
2754 btrfs_cleanup_fs_uuids();
2755 btrfs_exit_compress();
2758 late_initcall(init_btrfs_fs);
2759 module_exit(exit_btrfs_fs)
2761 MODULE_LICENSE("GPL");
2762 MODULE_SOFTDEP("pre: crc32c");
2763 MODULE_SOFTDEP("pre: xxhash64");
2764 MODULE_SOFTDEP("pre: sha256");
2765 MODULE_SOFTDEP("pre: blake2b-256");