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"
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);
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 sb->s_flags |= SB_RDONLY;
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))
244 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
245 fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
252 * We only mark the transaction aborted and then set the file system read-only.
253 * This will prevent new transactions from starting or trying to join this
256 * This means that error recovery at the call site is limited to freeing
257 * any local memory allocations and passing the error code up without
258 * further cleanup. The transaction should complete as it normally would
259 * in the call path but will return -EIO.
261 * We'll complete the cleanup in btrfs_end_transaction and
262 * btrfs_commit_transaction.
265 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
266 const char *function,
267 unsigned int line, int errno)
269 struct btrfs_fs_info *fs_info = trans->fs_info;
271 WRITE_ONCE(trans->aborted, errno);
272 /* Nothing used. The other threads that have joined this
273 * transaction may be able to continue. */
274 if (!trans->dirty && list_empty(&trans->new_bgs)) {
277 errstr = btrfs_decode_error(errno);
279 "%s:%d: Aborting unused transaction(%s).",
280 function, line, errstr);
283 WRITE_ONCE(trans->transaction->aborted, errno);
284 /* Wake up anybody who may be waiting on this transaction */
285 wake_up(&fs_info->transaction_wait);
286 wake_up(&fs_info->transaction_blocked_wait);
287 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
290 * __btrfs_panic decodes unexpected, fatal errors from the caller,
291 * issues an alert, and either panics or BUGs, depending on mount options.
294 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
295 unsigned int line, int errno, const char *fmt, ...)
297 char *s_id = "<unknown>";
299 struct va_format vaf = { .fmt = fmt };
303 s_id = fs_info->sb->s_id;
308 errstr = btrfs_decode_error(errno);
309 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
310 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
311 s_id, function, line, &vaf, errno, errstr);
313 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
314 function, line, &vaf, errno, errstr);
316 /* Caller calls BUG() */
319 static void btrfs_put_super(struct super_block *sb)
321 close_ctree(btrfs_sb(sb));
330 Opt_compress_force_type,
335 Opt_flushoncommit, Opt_noflushoncommit,
336 Opt_inode_cache, Opt_noinode_cache,
338 Opt_barrier, Opt_nobarrier,
339 Opt_datacow, Opt_nodatacow,
340 Opt_datasum, Opt_nodatasum,
341 Opt_defrag, Opt_nodefrag,
342 Opt_discard, Opt_nodiscard,
346 Opt_rescan_uuid_tree,
348 Opt_space_cache, Opt_no_space_cache,
349 Opt_space_cache_version,
351 Opt_ssd_spread, Opt_nossd_spread,
356 Opt_treelog, Opt_notreelog,
357 Opt_user_subvol_rm_allowed,
364 /* Deprecated options */
367 /* Debugging options */
369 Opt_check_integrity_including_extent_data,
370 Opt_check_integrity_print_mask,
371 Opt_enospc_debug, Opt_noenospc_debug,
372 #ifdef CONFIG_BTRFS_DEBUG
373 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
375 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
381 static const match_table_t tokens = {
383 {Opt_noacl, "noacl"},
384 {Opt_clear_cache, "clear_cache"},
385 {Opt_commit_interval, "commit=%u"},
386 {Opt_compress, "compress"},
387 {Opt_compress_type, "compress=%s"},
388 {Opt_compress_force, "compress-force"},
389 {Opt_compress_force_type, "compress-force=%s"},
390 {Opt_degraded, "degraded"},
391 {Opt_device, "device=%s"},
392 {Opt_fatal_errors, "fatal_errors=%s"},
393 {Opt_flushoncommit, "flushoncommit"},
394 {Opt_noflushoncommit, "noflushoncommit"},
395 {Opt_inode_cache, "inode_cache"},
396 {Opt_noinode_cache, "noinode_cache"},
397 {Opt_max_inline, "max_inline=%s"},
398 {Opt_barrier, "barrier"},
399 {Opt_nobarrier, "nobarrier"},
400 {Opt_datacow, "datacow"},
401 {Opt_nodatacow, "nodatacow"},
402 {Opt_datasum, "datasum"},
403 {Opt_nodatasum, "nodatasum"},
404 {Opt_defrag, "autodefrag"},
405 {Opt_nodefrag, "noautodefrag"},
406 {Opt_discard, "discard"},
407 {Opt_discard_mode, "discard=%s"},
408 {Opt_nodiscard, "nodiscard"},
409 {Opt_norecovery, "norecovery"},
410 {Opt_ratio, "metadata_ratio=%u"},
411 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
412 {Opt_skip_balance, "skip_balance"},
413 {Opt_space_cache, "space_cache"},
414 {Opt_no_space_cache, "nospace_cache"},
415 {Opt_space_cache_version, "space_cache=%s"},
417 {Opt_nossd, "nossd"},
418 {Opt_ssd_spread, "ssd_spread"},
419 {Opt_nossd_spread, "nossd_spread"},
420 {Opt_subvol, "subvol=%s"},
421 {Opt_subvol_empty, "subvol="},
422 {Opt_subvolid, "subvolid=%s"},
423 {Opt_thread_pool, "thread_pool=%u"},
424 {Opt_treelog, "treelog"},
425 {Opt_notreelog, "notreelog"},
426 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
429 {Opt_rescue, "rescue=%s"},
430 /* Deprecated, with alias rescue=nologreplay */
431 {Opt_nologreplay, "nologreplay"},
432 /* Deprecated, with alias rescue=usebackuproot */
433 {Opt_usebackuproot, "usebackuproot"},
435 /* Deprecated options */
436 {Opt_recovery, "recovery"},
438 /* Debugging options */
439 {Opt_check_integrity, "check_int"},
440 {Opt_check_integrity_including_extent_data, "check_int_data"},
441 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
442 {Opt_enospc_debug, "enospc_debug"},
443 {Opt_noenospc_debug, "noenospc_debug"},
444 #ifdef CONFIG_BTRFS_DEBUG
445 {Opt_fragment_data, "fragment=data"},
446 {Opt_fragment_metadata, "fragment=metadata"},
447 {Opt_fragment_all, "fragment=all"},
449 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
450 {Opt_ref_verify, "ref_verify"},
455 static const match_table_t rescue_tokens = {
456 {Opt_usebackuproot, "usebackuproot"},
457 {Opt_nologreplay, "nologreplay"},
461 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
466 substring_t args[MAX_OPT_ARGS];
469 opts = kstrdup(options, GFP_KERNEL);
474 while ((p = strsep(&opts, ":")) != NULL) {
479 token = match_token(p, rescue_tokens, args);
481 case Opt_usebackuproot:
483 "trying to use backup root at mount time");
484 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
486 case Opt_nologreplay:
487 btrfs_set_and_info(info, NOLOGREPLAY,
488 "disabling log replay at mount time");
491 btrfs_info(info, "unrecognized rescue option '%s'", p);
505 * Regular mount options parser. Everything that is needed only when
506 * reading in a new superblock is parsed here.
507 * XXX JDM: This needs to be cleaned up for remount.
509 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
510 unsigned long new_flags)
512 substring_t args[MAX_OPT_ARGS];
518 bool compress_force = false;
519 enum btrfs_compression_type saved_compress_type;
520 int saved_compress_level;
521 bool saved_compress_force;
524 cache_gen = btrfs_super_cache_generation(info->super_copy);
525 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
526 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
528 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
531 * Even the options are empty, we still need to do extra check
537 while ((p = strsep(&options, ",")) != NULL) {
542 token = match_token(p, tokens, args);
545 btrfs_info(info, "allowing degraded mounts");
546 btrfs_set_opt(info->mount_opt, DEGRADED);
549 case Opt_subvol_empty:
553 * These are parsed by btrfs_parse_subvol_options or
554 * btrfs_parse_device_options and can be ignored here.
558 btrfs_set_and_info(info, NODATASUM,
559 "setting nodatasum");
562 if (btrfs_test_opt(info, NODATASUM)) {
563 if (btrfs_test_opt(info, NODATACOW))
565 "setting datasum, datacow enabled");
567 btrfs_info(info, "setting datasum");
569 btrfs_clear_opt(info->mount_opt, NODATACOW);
570 btrfs_clear_opt(info->mount_opt, NODATASUM);
573 if (!btrfs_test_opt(info, NODATACOW)) {
574 if (!btrfs_test_opt(info, COMPRESS) ||
575 !btrfs_test_opt(info, FORCE_COMPRESS)) {
577 "setting nodatacow, compression disabled");
579 btrfs_info(info, "setting nodatacow");
582 btrfs_clear_opt(info->mount_opt, COMPRESS);
583 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
584 btrfs_set_opt(info->mount_opt, NODATACOW);
585 btrfs_set_opt(info->mount_opt, NODATASUM);
588 btrfs_clear_and_info(info, NODATACOW,
591 case Opt_compress_force:
592 case Opt_compress_force_type:
593 compress_force = true;
596 case Opt_compress_type:
597 saved_compress_type = btrfs_test_opt(info,
599 info->compress_type : BTRFS_COMPRESS_NONE;
600 saved_compress_force =
601 btrfs_test_opt(info, FORCE_COMPRESS);
602 saved_compress_level = info->compress_level;
603 if (token == Opt_compress ||
604 token == Opt_compress_force ||
605 strncmp(args[0].from, "zlib", 4) == 0) {
606 compress_type = "zlib";
608 info->compress_type = BTRFS_COMPRESS_ZLIB;
609 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
611 * args[0] contains uninitialized data since
612 * for these tokens we don't expect any
615 if (token != Opt_compress &&
616 token != Opt_compress_force)
617 info->compress_level =
618 btrfs_compress_str2level(
621 btrfs_set_opt(info->mount_opt, COMPRESS);
622 btrfs_clear_opt(info->mount_opt, NODATACOW);
623 btrfs_clear_opt(info->mount_opt, NODATASUM);
625 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
626 compress_type = "lzo";
627 info->compress_type = BTRFS_COMPRESS_LZO;
628 info->compress_level = 0;
629 btrfs_set_opt(info->mount_opt, COMPRESS);
630 btrfs_clear_opt(info->mount_opt, NODATACOW);
631 btrfs_clear_opt(info->mount_opt, NODATASUM);
632 btrfs_set_fs_incompat(info, COMPRESS_LZO);
634 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
635 compress_type = "zstd";
636 info->compress_type = BTRFS_COMPRESS_ZSTD;
637 info->compress_level =
638 btrfs_compress_str2level(
641 btrfs_set_opt(info->mount_opt, COMPRESS);
642 btrfs_clear_opt(info->mount_opt, NODATACOW);
643 btrfs_clear_opt(info->mount_opt, NODATASUM);
644 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
646 } else if (strncmp(args[0].from, "no", 2) == 0) {
647 compress_type = "no";
648 info->compress_level = 0;
649 info->compress_type = 0;
650 btrfs_clear_opt(info->mount_opt, COMPRESS);
651 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
652 compress_force = false;
655 btrfs_err(info, "unrecognized compression value %s",
661 if (compress_force) {
662 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
665 * If we remount from compress-force=xxx to
666 * compress=xxx, we need clear FORCE_COMPRESS
667 * flag, otherwise, there is no way for users
668 * to disable forcible compression separately.
670 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
672 if (no_compress == 1) {
673 btrfs_info(info, "use no compression");
674 } else if ((info->compress_type != saved_compress_type) ||
675 (compress_force != saved_compress_force) ||
676 (info->compress_level != saved_compress_level)) {
677 btrfs_info(info, "%s %s compression, level %d",
678 (compress_force) ? "force" : "use",
679 compress_type, info->compress_level);
681 compress_force = false;
684 btrfs_set_and_info(info, SSD,
685 "enabling ssd optimizations");
686 btrfs_clear_opt(info->mount_opt, NOSSD);
689 btrfs_set_and_info(info, SSD,
690 "enabling ssd optimizations");
691 btrfs_set_and_info(info, SSD_SPREAD,
692 "using spread ssd allocation scheme");
693 btrfs_clear_opt(info->mount_opt, NOSSD);
696 btrfs_set_opt(info->mount_opt, NOSSD);
697 btrfs_clear_and_info(info, SSD,
698 "not using ssd optimizations");
700 case Opt_nossd_spread:
701 btrfs_clear_and_info(info, SSD_SPREAD,
702 "not using spread ssd allocation scheme");
705 btrfs_clear_and_info(info, NOBARRIER,
706 "turning on barriers");
709 btrfs_set_and_info(info, NOBARRIER,
710 "turning off barriers");
712 case Opt_thread_pool:
713 ret = match_int(&args[0], &intarg);
715 btrfs_err(info, "unrecognized thread_pool value %s",
718 } else if (intarg == 0) {
719 btrfs_err(info, "invalid value 0 for thread_pool");
723 info->thread_pool_size = intarg;
726 num = match_strdup(&args[0]);
728 info->max_inline = memparse(num, NULL);
731 if (info->max_inline) {
732 info->max_inline = min_t(u64,
736 btrfs_info(info, "max_inline at %llu",
744 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
745 info->sb->s_flags |= SB_POSIXACL;
748 btrfs_err(info, "support for ACL not compiled in!");
753 info->sb->s_flags &= ~SB_POSIXACL;
756 btrfs_set_and_info(info, NOTREELOG,
757 "disabling tree log");
760 btrfs_clear_and_info(info, NOTREELOG,
761 "enabling tree log");
764 case Opt_nologreplay:
766 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
767 btrfs_set_and_info(info, NOLOGREPLAY,
768 "disabling log replay at mount time");
770 case Opt_flushoncommit:
771 btrfs_set_and_info(info, FLUSHONCOMMIT,
772 "turning on flush-on-commit");
774 case Opt_noflushoncommit:
775 btrfs_clear_and_info(info, FLUSHONCOMMIT,
776 "turning off flush-on-commit");
779 ret = match_int(&args[0], &intarg);
781 btrfs_err(info, "unrecognized metadata_ratio value %s",
785 info->metadata_ratio = intarg;
786 btrfs_info(info, "metadata ratio %u",
787 info->metadata_ratio);
790 case Opt_discard_mode:
791 if (token == Opt_discard ||
792 strcmp(args[0].from, "sync") == 0) {
793 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
794 btrfs_set_and_info(info, DISCARD_SYNC,
795 "turning on sync discard");
796 } else if (strcmp(args[0].from, "async") == 0) {
797 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
798 btrfs_set_and_info(info, DISCARD_ASYNC,
799 "turning on async discard");
801 btrfs_err(info, "unrecognized discard mode value %s",
808 btrfs_clear_and_info(info, DISCARD_SYNC,
809 "turning off discard");
810 btrfs_clear_and_info(info, DISCARD_ASYNC,
811 "turning off async discard");
813 case Opt_space_cache:
814 case Opt_space_cache_version:
815 if (token == Opt_space_cache ||
816 strcmp(args[0].from, "v1") == 0) {
817 btrfs_clear_opt(info->mount_opt,
819 btrfs_set_and_info(info, SPACE_CACHE,
820 "enabling disk space caching");
821 } else if (strcmp(args[0].from, "v2") == 0) {
822 btrfs_clear_opt(info->mount_opt,
824 btrfs_set_and_info(info, FREE_SPACE_TREE,
825 "enabling free space tree");
827 btrfs_err(info, "unrecognized space_cache value %s",
833 case Opt_rescan_uuid_tree:
834 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
836 case Opt_no_space_cache:
837 if (btrfs_test_opt(info, SPACE_CACHE)) {
838 btrfs_clear_and_info(info, SPACE_CACHE,
839 "disabling disk space caching");
841 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
842 btrfs_clear_and_info(info, FREE_SPACE_TREE,
843 "disabling free space tree");
846 case Opt_inode_cache:
848 "the 'inode_cache' option is deprecated and will have no effect from 5.11");
849 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
850 "enabling inode map caching");
852 case Opt_noinode_cache:
853 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
854 "disabling inode map caching");
856 case Opt_clear_cache:
857 btrfs_set_and_info(info, CLEAR_CACHE,
858 "force clearing of disk cache");
860 case Opt_user_subvol_rm_allowed:
861 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
863 case Opt_enospc_debug:
864 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
866 case Opt_noenospc_debug:
867 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
870 btrfs_set_and_info(info, AUTO_DEFRAG,
871 "enabling auto defrag");
874 btrfs_clear_and_info(info, AUTO_DEFRAG,
875 "disabling auto defrag");
878 case Opt_usebackuproot:
880 "'%s' is deprecated, use 'rescue=usebackuproot' instead",
881 token == Opt_recovery ? "recovery" :
884 "trying to use backup root at mount time");
885 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
887 case Opt_skip_balance:
888 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
890 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
891 case Opt_check_integrity_including_extent_data:
893 "enabling check integrity including extent data");
894 btrfs_set_opt(info->mount_opt,
895 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
896 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
898 case Opt_check_integrity:
899 btrfs_info(info, "enabling check integrity");
900 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
902 case Opt_check_integrity_print_mask:
903 ret = match_int(&args[0], &intarg);
906 "unrecognized check_integrity_print_mask value %s",
910 info->check_integrity_print_mask = intarg;
911 btrfs_info(info, "check_integrity_print_mask 0x%x",
912 info->check_integrity_print_mask);
915 case Opt_check_integrity_including_extent_data:
916 case Opt_check_integrity:
917 case Opt_check_integrity_print_mask:
919 "support for check_integrity* not compiled in!");
923 case Opt_fatal_errors:
924 if (strcmp(args[0].from, "panic") == 0) {
925 btrfs_set_opt(info->mount_opt,
926 PANIC_ON_FATAL_ERROR);
927 } else if (strcmp(args[0].from, "bug") == 0) {
928 btrfs_clear_opt(info->mount_opt,
929 PANIC_ON_FATAL_ERROR);
931 btrfs_err(info, "unrecognized fatal_errors value %s",
937 case Opt_commit_interval:
939 ret = match_int(&args[0], &intarg);
941 btrfs_err(info, "unrecognized commit_interval value %s",
948 "using default commit interval %us",
949 BTRFS_DEFAULT_COMMIT_INTERVAL);
950 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
951 } else if (intarg > 300) {
952 btrfs_warn(info, "excessive commit interval %d",
955 info->commit_interval = intarg;
958 ret = parse_rescue_options(info, args[0].from);
960 btrfs_err(info, "unrecognized rescue value %s",
965 #ifdef CONFIG_BTRFS_DEBUG
966 case Opt_fragment_all:
967 btrfs_info(info, "fragmenting all space");
968 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
969 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
971 case Opt_fragment_metadata:
972 btrfs_info(info, "fragmenting metadata");
973 btrfs_set_opt(info->mount_opt,
976 case Opt_fragment_data:
977 btrfs_info(info, "fragmenting data");
978 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
981 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
983 btrfs_info(info, "doing ref verification");
984 btrfs_set_opt(info->mount_opt, REF_VERIFY);
988 btrfs_err(info, "unrecognized mount option '%s'", p);
997 * Extra check for current option against current flag
999 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
1001 "nologreplay must be used with ro mount option");
1005 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
1006 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
1007 !btrfs_test_opt(info, CLEAR_CACHE)) {
1008 btrfs_err(info, "cannot disable free space tree");
1012 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
1013 btrfs_info(info, "disk space caching is enabled");
1014 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
1015 btrfs_info(info, "using free space tree");
1020 * Parse mount options that are required early in the mount process.
1022 * All other options will be parsed on much later in the mount process and
1023 * only when we need to allocate a new super block.
1025 static int btrfs_parse_device_options(const char *options, fmode_t flags,
1028 substring_t args[MAX_OPT_ARGS];
1029 char *device_name, *opts, *orig, *p;
1030 struct btrfs_device *device = NULL;
1033 lockdep_assert_held(&uuid_mutex);
1039 * strsep changes the string, duplicate it because btrfs_parse_options
1042 opts = kstrdup(options, GFP_KERNEL);
1047 while ((p = strsep(&opts, ",")) != NULL) {
1053 token = match_token(p, tokens, args);
1054 if (token == Opt_device) {
1055 device_name = match_strdup(&args[0]);
1060 device = btrfs_scan_one_device(device_name, flags,
1063 if (IS_ERR(device)) {
1064 error = PTR_ERR(device);
1076 * Parse mount options that are related to subvolume id
1078 * The value is later passed to mount_subvol()
1080 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1081 u64 *subvol_objectid)
1083 substring_t args[MAX_OPT_ARGS];
1084 char *opts, *orig, *p;
1092 * strsep changes the string, duplicate it because
1093 * btrfs_parse_device_options gets called later
1095 opts = kstrdup(options, GFP_KERNEL);
1100 while ((p = strsep(&opts, ",")) != NULL) {
1105 token = match_token(p, tokens, args);
1108 kfree(*subvol_name);
1109 *subvol_name = match_strdup(&args[0]);
1110 if (!*subvol_name) {
1116 error = match_u64(&args[0], &subvolid);
1120 /* we want the original fs_tree */
1122 subvolid = BTRFS_FS_TREE_OBJECTID;
1124 *subvol_objectid = subvolid;
1136 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1137 u64 subvol_objectid)
1139 struct btrfs_root *root = fs_info->tree_root;
1140 struct btrfs_root *fs_root = NULL;
1141 struct btrfs_root_ref *root_ref;
1142 struct btrfs_inode_ref *inode_ref;
1143 struct btrfs_key key;
1144 struct btrfs_path *path = NULL;
1145 char *name = NULL, *ptr;
1150 path = btrfs_alloc_path();
1155 path->leave_spinning = 1;
1157 name = kmalloc(PATH_MAX, GFP_KERNEL);
1162 ptr = name + PATH_MAX - 1;
1166 * Walk up the subvolume trees in the tree of tree roots by root
1167 * backrefs until we hit the top-level subvolume.
1169 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1170 key.objectid = subvol_objectid;
1171 key.type = BTRFS_ROOT_BACKREF_KEY;
1172 key.offset = (u64)-1;
1174 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1177 } else if (ret > 0) {
1178 ret = btrfs_previous_item(root, path, subvol_objectid,
1179 BTRFS_ROOT_BACKREF_KEY);
1182 } else if (ret > 0) {
1188 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1189 subvol_objectid = key.offset;
1191 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1192 struct btrfs_root_ref);
1193 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1196 ret = -ENAMETOOLONG;
1199 read_extent_buffer(path->nodes[0], ptr + 1,
1200 (unsigned long)(root_ref + 1), len);
1202 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1203 btrfs_release_path(path);
1205 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1206 if (IS_ERR(fs_root)) {
1207 ret = PTR_ERR(fs_root);
1213 * Walk up the filesystem tree by inode refs until we hit the
1216 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1217 key.objectid = dirid;
1218 key.type = BTRFS_INODE_REF_KEY;
1219 key.offset = (u64)-1;
1221 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1224 } else if (ret > 0) {
1225 ret = btrfs_previous_item(fs_root, path, dirid,
1226 BTRFS_INODE_REF_KEY);
1229 } else if (ret > 0) {
1235 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1238 inode_ref = btrfs_item_ptr(path->nodes[0],
1240 struct btrfs_inode_ref);
1241 len = btrfs_inode_ref_name_len(path->nodes[0],
1245 ret = -ENAMETOOLONG;
1248 read_extent_buffer(path->nodes[0], ptr + 1,
1249 (unsigned long)(inode_ref + 1), len);
1251 btrfs_release_path(path);
1253 btrfs_put_root(fs_root);
1257 btrfs_free_path(path);
1258 if (ptr == name + PATH_MAX - 1) {
1262 memmove(name, ptr, name + PATH_MAX - ptr);
1267 btrfs_put_root(fs_root);
1268 btrfs_free_path(path);
1270 return ERR_PTR(ret);
1273 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1275 struct btrfs_root *root = fs_info->tree_root;
1276 struct btrfs_dir_item *di;
1277 struct btrfs_path *path;
1278 struct btrfs_key location;
1281 path = btrfs_alloc_path();
1284 path->leave_spinning = 1;
1287 * Find the "default" dir item which points to the root item that we
1288 * will mount by default if we haven't been given a specific subvolume
1291 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1292 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1294 btrfs_free_path(path);
1299 * Ok the default dir item isn't there. This is weird since
1300 * it's always been there, but don't freak out, just try and
1301 * mount the top-level subvolume.
1303 btrfs_free_path(path);
1304 *objectid = BTRFS_FS_TREE_OBJECTID;
1308 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1309 btrfs_free_path(path);
1310 *objectid = location.objectid;
1314 static int btrfs_fill_super(struct super_block *sb,
1315 struct btrfs_fs_devices *fs_devices,
1318 struct inode *inode;
1319 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1322 sb->s_maxbytes = MAX_LFS_FILESIZE;
1323 sb->s_magic = BTRFS_SUPER_MAGIC;
1324 sb->s_op = &btrfs_super_ops;
1325 sb->s_d_op = &btrfs_dentry_operations;
1326 sb->s_export_op = &btrfs_export_ops;
1327 sb->s_xattr = btrfs_xattr_handlers;
1328 sb->s_time_gran = 1;
1329 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1330 sb->s_flags |= SB_POSIXACL;
1332 sb->s_flags |= SB_I_VERSION;
1333 sb->s_iflags |= SB_I_CGROUPWB;
1335 err = super_setup_bdi(sb);
1337 btrfs_err(fs_info, "super_setup_bdi failed");
1341 err = open_ctree(sb, fs_devices, (char *)data);
1343 btrfs_err(fs_info, "open_ctree failed");
1347 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1348 if (IS_ERR(inode)) {
1349 err = PTR_ERR(inode);
1353 sb->s_root = d_make_root(inode);
1359 cleancache_init_fs(sb);
1360 sb->s_flags |= SB_ACTIVE;
1364 close_ctree(fs_info);
1368 int btrfs_sync_fs(struct super_block *sb, int wait)
1370 struct btrfs_trans_handle *trans;
1371 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1372 struct btrfs_root *root = fs_info->tree_root;
1374 trace_btrfs_sync_fs(fs_info, wait);
1377 filemap_flush(fs_info->btree_inode->i_mapping);
1381 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1383 trans = btrfs_attach_transaction_barrier(root);
1384 if (IS_ERR(trans)) {
1385 /* no transaction, don't bother */
1386 if (PTR_ERR(trans) == -ENOENT) {
1388 * Exit unless we have some pending changes
1389 * that need to go through commit
1391 if (fs_info->pending_changes == 0)
1394 * A non-blocking test if the fs is frozen. We must not
1395 * start a new transaction here otherwise a deadlock
1396 * happens. The pending operations are delayed to the
1397 * next commit after thawing.
1399 if (sb_start_write_trylock(sb))
1403 trans = btrfs_start_transaction(root, 0);
1406 return PTR_ERR(trans);
1408 return btrfs_commit_transaction(trans);
1411 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1413 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1414 const char *compress_type;
1415 const char *subvol_name;
1417 if (btrfs_test_opt(info, DEGRADED))
1418 seq_puts(seq, ",degraded");
1419 if (btrfs_test_opt(info, NODATASUM))
1420 seq_puts(seq, ",nodatasum");
1421 if (btrfs_test_opt(info, NODATACOW))
1422 seq_puts(seq, ",nodatacow");
1423 if (btrfs_test_opt(info, NOBARRIER))
1424 seq_puts(seq, ",nobarrier");
1425 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1426 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1427 if (info->thread_pool_size != min_t(unsigned long,
1428 num_online_cpus() + 2, 8))
1429 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1430 if (btrfs_test_opt(info, COMPRESS)) {
1431 compress_type = btrfs_compress_type2str(info->compress_type);
1432 if (btrfs_test_opt(info, FORCE_COMPRESS))
1433 seq_printf(seq, ",compress-force=%s", compress_type);
1435 seq_printf(seq, ",compress=%s", compress_type);
1436 if (info->compress_level)
1437 seq_printf(seq, ":%d", info->compress_level);
1439 if (btrfs_test_opt(info, NOSSD))
1440 seq_puts(seq, ",nossd");
1441 if (btrfs_test_opt(info, SSD_SPREAD))
1442 seq_puts(seq, ",ssd_spread");
1443 else if (btrfs_test_opt(info, SSD))
1444 seq_puts(seq, ",ssd");
1445 if (btrfs_test_opt(info, NOTREELOG))
1446 seq_puts(seq, ",notreelog");
1447 if (btrfs_test_opt(info, NOLOGREPLAY))
1448 seq_puts(seq, ",rescue=nologreplay");
1449 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1450 seq_puts(seq, ",flushoncommit");
1451 if (btrfs_test_opt(info, DISCARD_SYNC))
1452 seq_puts(seq, ",discard");
1453 if (btrfs_test_opt(info, DISCARD_ASYNC))
1454 seq_puts(seq, ",discard=async");
1455 if (!(info->sb->s_flags & SB_POSIXACL))
1456 seq_puts(seq, ",noacl");
1457 if (btrfs_test_opt(info, SPACE_CACHE))
1458 seq_puts(seq, ",space_cache");
1459 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1460 seq_puts(seq, ",space_cache=v2");
1462 seq_puts(seq, ",nospace_cache");
1463 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1464 seq_puts(seq, ",rescan_uuid_tree");
1465 if (btrfs_test_opt(info, CLEAR_CACHE))
1466 seq_puts(seq, ",clear_cache");
1467 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1468 seq_puts(seq, ",user_subvol_rm_allowed");
1469 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1470 seq_puts(seq, ",enospc_debug");
1471 if (btrfs_test_opt(info, AUTO_DEFRAG))
1472 seq_puts(seq, ",autodefrag");
1473 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1474 seq_puts(seq, ",inode_cache");
1475 if (btrfs_test_opt(info, SKIP_BALANCE))
1476 seq_puts(seq, ",skip_balance");
1477 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1478 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1479 seq_puts(seq, ",check_int_data");
1480 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1481 seq_puts(seq, ",check_int");
1482 if (info->check_integrity_print_mask)
1483 seq_printf(seq, ",check_int_print_mask=%d",
1484 info->check_integrity_print_mask);
1486 if (info->metadata_ratio)
1487 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1488 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1489 seq_puts(seq, ",fatal_errors=panic");
1490 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1491 seq_printf(seq, ",commit=%u", info->commit_interval);
1492 #ifdef CONFIG_BTRFS_DEBUG
1493 if (btrfs_test_opt(info, FRAGMENT_DATA))
1494 seq_puts(seq, ",fragment=data");
1495 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1496 seq_puts(seq, ",fragment=metadata");
1498 if (btrfs_test_opt(info, REF_VERIFY))
1499 seq_puts(seq, ",ref_verify");
1500 seq_printf(seq, ",subvolid=%llu",
1501 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1502 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1503 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1504 if (!IS_ERR(subvol_name)) {
1505 seq_puts(seq, ",subvol=");
1506 seq_escape(seq, subvol_name, " \t\n\\");
1512 static int btrfs_test_super(struct super_block *s, void *data)
1514 struct btrfs_fs_info *p = data;
1515 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1517 return fs_info->fs_devices == p->fs_devices;
1520 static int btrfs_set_super(struct super_block *s, void *data)
1522 int err = set_anon_super(s, data);
1524 s->s_fs_info = data;
1529 * subvolumes are identified by ino 256
1531 static inline int is_subvolume_inode(struct inode *inode)
1533 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1538 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1539 struct vfsmount *mnt)
1541 struct dentry *root;
1545 if (!subvol_objectid) {
1546 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1549 root = ERR_PTR(ret);
1553 subvol_name = btrfs_get_subvol_name_from_objectid(
1554 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1555 if (IS_ERR(subvol_name)) {
1556 root = ERR_CAST(subvol_name);
1563 root = mount_subtree(mnt, subvol_name);
1564 /* mount_subtree() drops our reference on the vfsmount. */
1567 if (!IS_ERR(root)) {
1568 struct super_block *s = root->d_sb;
1569 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1570 struct inode *root_inode = d_inode(root);
1571 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1574 if (!is_subvolume_inode(root_inode)) {
1575 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1579 if (subvol_objectid && root_objectid != subvol_objectid) {
1581 * This will also catch a race condition where a
1582 * subvolume which was passed by ID is renamed and
1583 * another subvolume is renamed over the old location.
1586 "subvol '%s' does not match subvolid %llu",
1587 subvol_name, subvol_objectid);
1592 root = ERR_PTR(ret);
1593 deactivate_locked_super(s);
1604 * Find a superblock for the given device / mount point.
1606 * Note: This is based on mount_bdev from fs/super.c with a few additions
1607 * for multiple device setup. Make sure to keep it in sync.
1609 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1610 int flags, const char *device_name, void *data)
1612 struct block_device *bdev = NULL;
1613 struct super_block *s;
1614 struct btrfs_device *device = NULL;
1615 struct btrfs_fs_devices *fs_devices = NULL;
1616 struct btrfs_fs_info *fs_info = NULL;
1617 void *new_sec_opts = NULL;
1618 fmode_t mode = FMODE_READ;
1621 if (!(flags & SB_RDONLY))
1622 mode |= FMODE_WRITE;
1625 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1627 return ERR_PTR(error);
1631 * Setup a dummy root and fs_info for test/set super. This is because
1632 * we don't actually fill this stuff out until open_ctree, but we need
1633 * then open_ctree will properly initialize the file system specific
1634 * settings later. btrfs_init_fs_info initializes the static elements
1635 * of the fs_info (locks and such) to make cleanup easier if we find a
1636 * superblock with our given fs_devices later on at sget() time.
1638 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1641 goto error_sec_opts;
1643 btrfs_init_fs_info(fs_info);
1645 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1646 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1647 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1652 mutex_lock(&uuid_mutex);
1653 error = btrfs_parse_device_options(data, mode, fs_type);
1655 mutex_unlock(&uuid_mutex);
1659 device = btrfs_scan_one_device(device_name, mode, fs_type);
1660 if (IS_ERR(device)) {
1661 mutex_unlock(&uuid_mutex);
1662 error = PTR_ERR(device);
1666 fs_devices = device->fs_devices;
1667 fs_info->fs_devices = fs_devices;
1669 error = btrfs_open_devices(fs_devices, mode, fs_type);
1670 mutex_unlock(&uuid_mutex);
1674 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1676 goto error_close_devices;
1679 bdev = fs_devices->latest_bdev;
1680 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1684 goto error_close_devices;
1688 btrfs_close_devices(fs_devices);
1689 btrfs_free_fs_info(fs_info);
1690 if ((flags ^ s->s_flags) & SB_RDONLY)
1693 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1694 btrfs_sb(s)->bdev_holder = fs_type;
1695 error = btrfs_fill_super(s, fs_devices, data);
1698 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1699 security_free_mnt_opts(&new_sec_opts);
1701 deactivate_locked_super(s);
1702 return ERR_PTR(error);
1705 return dget(s->s_root);
1707 error_close_devices:
1708 btrfs_close_devices(fs_devices);
1710 btrfs_free_fs_info(fs_info);
1712 security_free_mnt_opts(&new_sec_opts);
1713 return ERR_PTR(error);
1717 * Mount function which is called by VFS layer.
1719 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1720 * which needs vfsmount* of device's root (/). This means device's root has to
1721 * be mounted internally in any case.
1724 * 1. Parse subvol id related options for later use in mount_subvol().
1726 * 2. Mount device's root (/) by calling vfs_kern_mount().
1728 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1729 * first place. In order to avoid calling btrfs_mount() again, we use
1730 * different file_system_type which is not registered to VFS by
1731 * register_filesystem() (btrfs_root_fs_type). As a result,
1732 * btrfs_mount_root() is called. The return value will be used by
1733 * mount_subtree() in mount_subvol().
1735 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1736 * "btrfs subvolume set-default", mount_subvol() is called always.
1738 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1739 const char *device_name, void *data)
1741 struct vfsmount *mnt_root;
1742 struct dentry *root;
1743 char *subvol_name = NULL;
1744 u64 subvol_objectid = 0;
1747 error = btrfs_parse_subvol_options(data, &subvol_name,
1751 return ERR_PTR(error);
1754 /* mount device's root (/) */
1755 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1756 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1757 if (flags & SB_RDONLY) {
1758 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1759 flags & ~SB_RDONLY, device_name, data);
1761 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1762 flags | SB_RDONLY, device_name, data);
1763 if (IS_ERR(mnt_root)) {
1764 root = ERR_CAST(mnt_root);
1769 down_write(&mnt_root->mnt_sb->s_umount);
1770 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1771 up_write(&mnt_root->mnt_sb->s_umount);
1773 root = ERR_PTR(error);
1780 if (IS_ERR(mnt_root)) {
1781 root = ERR_CAST(mnt_root);
1786 /* mount_subvol() will free subvol_name and mnt_root */
1787 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1793 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1794 u32 new_pool_size, u32 old_pool_size)
1796 if (new_pool_size == old_pool_size)
1799 fs_info->thread_pool_size = new_pool_size;
1801 btrfs_info(fs_info, "resize thread pool %d -> %d",
1802 old_pool_size, new_pool_size);
1804 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1805 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1806 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1807 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1808 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1809 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1811 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1812 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1813 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1814 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1815 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1819 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1820 unsigned long old_opts, int flags)
1822 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1823 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1824 (flags & SB_RDONLY))) {
1825 /* wait for any defraggers to finish */
1826 wait_event(fs_info->transaction_wait,
1827 (atomic_read(&fs_info->defrag_running) == 0));
1828 if (flags & SB_RDONLY)
1829 sync_filesystem(fs_info->sb);
1833 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1834 unsigned long old_opts)
1837 * We need to cleanup all defragable inodes if the autodefragment is
1838 * close or the filesystem is read only.
1840 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1841 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1842 btrfs_cleanup_defrag_inodes(fs_info);
1845 /* If we toggled discard async */
1846 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1847 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1848 btrfs_discard_resume(fs_info);
1849 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1850 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1851 btrfs_discard_cleanup(fs_info);
1854 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1856 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1857 struct btrfs_root *root = fs_info->tree_root;
1858 unsigned old_flags = sb->s_flags;
1859 unsigned long old_opts = fs_info->mount_opt;
1860 unsigned long old_compress_type = fs_info->compress_type;
1861 u64 old_max_inline = fs_info->max_inline;
1862 u32 old_thread_pool_size = fs_info->thread_pool_size;
1863 u32 old_metadata_ratio = fs_info->metadata_ratio;
1866 sync_filesystem(sb);
1867 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1870 void *new_sec_opts = NULL;
1872 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1874 ret = security_sb_remount(sb, new_sec_opts);
1875 security_free_mnt_opts(&new_sec_opts);
1880 ret = btrfs_parse_options(fs_info, data, *flags);
1884 btrfs_remount_begin(fs_info, old_opts, *flags);
1885 btrfs_resize_thread_pool(fs_info,
1886 fs_info->thread_pool_size, old_thread_pool_size);
1888 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1891 if (*flags & SB_RDONLY) {
1893 * this also happens on 'umount -rf' or on shutdown, when
1894 * the filesystem is busy.
1896 cancel_work_sync(&fs_info->async_reclaim_work);
1897 cancel_work_sync(&fs_info->async_data_reclaim_work);
1899 btrfs_discard_cleanup(fs_info);
1901 /* wait for the uuid_scan task to finish */
1902 down(&fs_info->uuid_tree_rescan_sem);
1903 /* avoid complains from lockdep et al. */
1904 up(&fs_info->uuid_tree_rescan_sem);
1906 sb->s_flags |= SB_RDONLY;
1909 * Setting SB_RDONLY will put the cleaner thread to
1910 * sleep at the next loop if it's already active.
1911 * If it's already asleep, we'll leave unused block
1912 * groups on disk until we're mounted read-write again
1913 * unless we clean them up here.
1915 btrfs_delete_unused_bgs(fs_info);
1917 btrfs_dev_replace_suspend_for_unmount(fs_info);
1918 btrfs_scrub_cancel(fs_info);
1919 btrfs_pause_balance(fs_info);
1922 * Pause the qgroup rescan worker if it is running. We don't want
1923 * it to be still running after we are in RO mode, as after that,
1924 * by the time we unmount, it might have left a transaction open,
1925 * so we would leak the transaction and/or crash.
1927 btrfs_qgroup_wait_for_completion(fs_info, false);
1929 ret = btrfs_commit_super(fs_info);
1933 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1935 "Remounting read-write after error is not allowed");
1939 if (fs_info->fs_devices->rw_devices == 0) {
1944 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1946 "too many missing devices, writable remount is not allowed");
1951 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1953 "mount required to replay tree-log, cannot remount read-write");
1958 ret = btrfs_cleanup_fs_roots(fs_info);
1962 /* recover relocation */
1963 mutex_lock(&fs_info->cleaner_mutex);
1964 ret = btrfs_recover_relocation(root);
1965 mutex_unlock(&fs_info->cleaner_mutex);
1969 ret = btrfs_resume_balance_async(fs_info);
1973 ret = btrfs_resume_dev_replace_async(fs_info);
1975 btrfs_warn(fs_info, "failed to resume dev_replace");
1979 btrfs_qgroup_rescan_resume(fs_info);
1981 if (!fs_info->uuid_root) {
1982 btrfs_info(fs_info, "creating UUID tree");
1983 ret = btrfs_create_uuid_tree(fs_info);
1986 "failed to create the UUID tree %d",
1991 sb->s_flags &= ~SB_RDONLY;
1993 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1997 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
1998 * since the absence of the flag means it can be toggled off by remount.
2000 *flags |= SB_I_VERSION;
2002 wake_up_process(fs_info->transaction_kthread);
2003 btrfs_remount_cleanup(fs_info, old_opts);
2004 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2009 /* We've hit an error - don't reset SB_RDONLY */
2011 old_flags |= SB_RDONLY;
2012 sb->s_flags = old_flags;
2013 fs_info->mount_opt = old_opts;
2014 fs_info->compress_type = old_compress_type;
2015 fs_info->max_inline = old_max_inline;
2016 btrfs_resize_thread_pool(fs_info,
2017 old_thread_pool_size, fs_info->thread_pool_size);
2018 fs_info->metadata_ratio = old_metadata_ratio;
2019 btrfs_remount_cleanup(fs_info, old_opts);
2020 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2025 /* Used to sort the devices by max_avail(descending sort) */
2026 static inline int btrfs_cmp_device_free_bytes(const void *dev_info1,
2027 const void *dev_info2)
2029 if (((struct btrfs_device_info *)dev_info1)->max_avail >
2030 ((struct btrfs_device_info *)dev_info2)->max_avail)
2032 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
2033 ((struct btrfs_device_info *)dev_info2)->max_avail)
2040 * sort the devices by max_avail, in which max free extent size of each device
2041 * is stored.(Descending Sort)
2043 static inline void btrfs_descending_sort_devices(
2044 struct btrfs_device_info *devices,
2047 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2048 btrfs_cmp_device_free_bytes, NULL);
2052 * The helper to calc the free space on the devices that can be used to store
2055 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2058 struct btrfs_device_info *devices_info;
2059 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2060 struct btrfs_device *device;
2063 u64 min_stripe_size;
2064 int num_stripes = 1;
2065 int i = 0, nr_devices;
2066 const struct btrfs_raid_attr *rattr;
2069 * We aren't under the device list lock, so this is racy-ish, but good
2070 * enough for our purposes.
2072 nr_devices = fs_info->fs_devices->open_devices;
2075 nr_devices = fs_info->fs_devices->open_devices;
2083 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2088 /* calc min stripe number for data space allocation */
2089 type = btrfs_data_alloc_profile(fs_info);
2090 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2092 if (type & BTRFS_BLOCK_GROUP_RAID0)
2093 num_stripes = nr_devices;
2094 else if (type & BTRFS_BLOCK_GROUP_RAID1)
2096 else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
2098 else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
2100 else if (type & BTRFS_BLOCK_GROUP_RAID10)
2103 /* Adjust for more than 1 stripe per device */
2104 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2107 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2108 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2109 &device->dev_state) ||
2111 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2114 if (i >= nr_devices)
2117 avail_space = device->total_bytes - device->bytes_used;
2119 /* align with stripe_len */
2120 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2123 * In order to avoid overwriting the superblock on the drive,
2124 * btrfs starts at an offset of at least 1MB when doing chunk
2127 * This ensures we have at least min_stripe_size free space
2128 * after excluding 1MB.
2130 if (avail_space <= SZ_1M + min_stripe_size)
2133 avail_space -= SZ_1M;
2135 devices_info[i].dev = device;
2136 devices_info[i].max_avail = avail_space;
2144 btrfs_descending_sort_devices(devices_info, nr_devices);
2148 while (nr_devices >= rattr->devs_min) {
2149 num_stripes = min(num_stripes, nr_devices);
2151 if (devices_info[i].max_avail >= min_stripe_size) {
2155 avail_space += devices_info[i].max_avail * num_stripes;
2156 alloc_size = devices_info[i].max_avail;
2157 for (j = i + 1 - num_stripes; j <= i; j++)
2158 devices_info[j].max_avail -= alloc_size;
2164 kfree(devices_info);
2165 *free_bytes = avail_space;
2170 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2172 * If there's a redundant raid level at DATA block groups, use the respective
2173 * multiplier to scale the sizes.
2175 * Unused device space usage is based on simulating the chunk allocator
2176 * algorithm that respects the device sizes and order of allocations. This is
2177 * a close approximation of the actual use but there are other factors that may
2178 * change the result (like a new metadata chunk).
2180 * If metadata is exhausted, f_bavail will be 0.
2182 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2184 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2185 struct btrfs_super_block *disk_super = fs_info->super_copy;
2186 struct btrfs_space_info *found;
2188 u64 total_free_data = 0;
2189 u64 total_free_meta = 0;
2190 int bits = dentry->d_sb->s_blocksize_bits;
2191 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2192 unsigned factor = 1;
2193 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2198 list_for_each_entry(found, &fs_info->space_info, list) {
2199 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2202 total_free_data += found->disk_total - found->disk_used;
2204 btrfs_account_ro_block_groups_free_space(found);
2206 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2207 if (!list_empty(&found->block_groups[i]))
2208 factor = btrfs_bg_type_to_factor(
2209 btrfs_raid_array[i].bg_flag);
2214 * Metadata in mixed block goup profiles are accounted in data
2216 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2217 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2220 total_free_meta += found->disk_total -
2224 total_used += found->disk_used;
2227 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2228 buf->f_blocks >>= bits;
2229 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2231 /* Account global block reserve as used, it's in logical size already */
2232 spin_lock(&block_rsv->lock);
2233 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2234 if (buf->f_bfree >= block_rsv->size >> bits)
2235 buf->f_bfree -= block_rsv->size >> bits;
2238 spin_unlock(&block_rsv->lock);
2240 buf->f_bavail = div_u64(total_free_data, factor);
2241 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2244 buf->f_bavail += div_u64(total_free_data, factor);
2245 buf->f_bavail = buf->f_bavail >> bits;
2248 * We calculate the remaining metadata space minus global reserve. If
2249 * this is (supposedly) smaller than zero, there's no space. But this
2250 * does not hold in practice, the exhausted state happens where's still
2251 * some positive delta. So we apply some guesswork and compare the
2252 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2254 * We probably cannot calculate the exact threshold value because this
2255 * depends on the internal reservations requested by various
2256 * operations, so some operations that consume a few metadata will
2257 * succeed even if the Avail is zero. But this is better than the other
2263 * We only want to claim there's no available space if we can no longer
2264 * allocate chunks for our metadata profile and our global reserve will
2265 * not fit in the free metadata space. If we aren't ->full then we
2266 * still can allocate chunks and thus are fine using the currently
2267 * calculated f_bavail.
2269 if (!mixed && block_rsv->space_info->full &&
2270 (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
2273 buf->f_type = BTRFS_SUPER_MAGIC;
2274 buf->f_bsize = dentry->d_sb->s_blocksize;
2275 buf->f_namelen = BTRFS_NAME_LEN;
2277 /* We treat it as constant endianness (it doesn't matter _which_)
2278 because we want the fsid to come out the same whether mounted
2279 on a big-endian or little-endian host */
2280 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2281 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2282 /* Mask in the root object ID too, to disambiguate subvols */
2283 buf->f_fsid.val[0] ^=
2284 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2285 buf->f_fsid.val[1] ^=
2286 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2291 static void btrfs_kill_super(struct super_block *sb)
2293 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2294 kill_anon_super(sb);
2295 btrfs_free_fs_info(fs_info);
2298 static struct file_system_type btrfs_fs_type = {
2299 .owner = THIS_MODULE,
2301 .mount = btrfs_mount,
2302 .kill_sb = btrfs_kill_super,
2303 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2306 static struct file_system_type btrfs_root_fs_type = {
2307 .owner = THIS_MODULE,
2309 .mount = btrfs_mount_root,
2310 .kill_sb = btrfs_kill_super,
2311 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2314 MODULE_ALIAS_FS("btrfs");
2316 static int btrfs_control_open(struct inode *inode, struct file *file)
2319 * The control file's private_data is used to hold the
2320 * transaction when it is started and is used to keep
2321 * track of whether a transaction is already in progress.
2323 file->private_data = NULL;
2328 * Used by /dev/btrfs-control for devices ioctls.
2330 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2333 struct btrfs_ioctl_vol_args *vol;
2334 struct btrfs_device *device = NULL;
2337 if (!capable(CAP_SYS_ADMIN))
2340 vol = memdup_user((void __user *)arg, sizeof(*vol));
2342 return PTR_ERR(vol);
2343 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2346 case BTRFS_IOC_SCAN_DEV:
2347 mutex_lock(&uuid_mutex);
2348 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2349 &btrfs_root_fs_type);
2350 ret = PTR_ERR_OR_ZERO(device);
2351 mutex_unlock(&uuid_mutex);
2353 case BTRFS_IOC_FORGET_DEV:
2354 ret = btrfs_forget_devices(vol->name);
2356 case BTRFS_IOC_DEVICES_READY:
2357 mutex_lock(&uuid_mutex);
2358 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2359 &btrfs_root_fs_type);
2360 if (IS_ERR(device)) {
2361 mutex_unlock(&uuid_mutex);
2362 ret = PTR_ERR(device);
2365 ret = !(device->fs_devices->num_devices ==
2366 device->fs_devices->total_devices);
2367 mutex_unlock(&uuid_mutex);
2369 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2370 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2378 static int btrfs_freeze(struct super_block *sb)
2380 struct btrfs_trans_handle *trans;
2381 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2382 struct btrfs_root *root = fs_info->tree_root;
2384 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2386 * We don't need a barrier here, we'll wait for any transaction that
2387 * could be in progress on other threads (and do delayed iputs that
2388 * we want to avoid on a frozen filesystem), or do the commit
2391 trans = btrfs_attach_transaction_barrier(root);
2392 if (IS_ERR(trans)) {
2393 /* no transaction, don't bother */
2394 if (PTR_ERR(trans) == -ENOENT)
2396 return PTR_ERR(trans);
2398 return btrfs_commit_transaction(trans);
2401 static int btrfs_unfreeze(struct super_block *sb)
2403 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2405 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2409 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2411 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2412 struct btrfs_device *dev, *first_dev = NULL;
2415 * Lightweight locking of the devices. We should not need
2416 * device_list_mutex here as we only read the device data and the list
2417 * is protected by RCU. Even if a device is deleted during the list
2418 * traversals, we'll get valid data, the freeing callback will wait at
2419 * least until the rcu_read_unlock.
2422 list_for_each_entry_rcu(dev, &fs_info->fs_devices->devices, dev_list) {
2423 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2427 if (!first_dev || dev->devid < first_dev->devid)
2432 seq_escape(m, rcu_str_deref(first_dev->name), " \t\n\\");
2439 static const struct super_operations btrfs_super_ops = {
2440 .drop_inode = btrfs_drop_inode,
2441 .evict_inode = btrfs_evict_inode,
2442 .put_super = btrfs_put_super,
2443 .sync_fs = btrfs_sync_fs,
2444 .show_options = btrfs_show_options,
2445 .show_devname = btrfs_show_devname,
2446 .alloc_inode = btrfs_alloc_inode,
2447 .destroy_inode = btrfs_destroy_inode,
2448 .free_inode = btrfs_free_inode,
2449 .statfs = btrfs_statfs,
2450 .remount_fs = btrfs_remount,
2451 .freeze_fs = btrfs_freeze,
2452 .unfreeze_fs = btrfs_unfreeze,
2455 static const struct file_operations btrfs_ctl_fops = {
2456 .open = btrfs_control_open,
2457 .unlocked_ioctl = btrfs_control_ioctl,
2458 .compat_ioctl = compat_ptr_ioctl,
2459 .owner = THIS_MODULE,
2460 .llseek = noop_llseek,
2463 static struct miscdevice btrfs_misc = {
2464 .minor = BTRFS_MINOR,
2465 .name = "btrfs-control",
2466 .fops = &btrfs_ctl_fops
2469 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2470 MODULE_ALIAS("devname:btrfs-control");
2472 static int __init btrfs_interface_init(void)
2474 return misc_register(&btrfs_misc);
2477 static __cold void btrfs_interface_exit(void)
2479 misc_deregister(&btrfs_misc);
2482 static void __init btrfs_print_mod_info(void)
2484 static const char options[] = ""
2485 #ifdef CONFIG_BTRFS_DEBUG
2488 #ifdef CONFIG_BTRFS_ASSERT
2491 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2492 ", integrity-checker=on"
2494 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2498 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2501 static int __init init_btrfs_fs(void)
2507 err = btrfs_init_sysfs();
2511 btrfs_init_compress();
2513 err = btrfs_init_cachep();
2517 err = extent_io_init();
2521 err = extent_state_cache_init();
2523 goto free_extent_io;
2525 err = extent_map_init();
2527 goto free_extent_state_cache;
2529 err = ordered_data_init();
2531 goto free_extent_map;
2533 err = btrfs_delayed_inode_init();
2535 goto free_ordered_data;
2537 err = btrfs_auto_defrag_init();
2539 goto free_delayed_inode;
2541 err = btrfs_delayed_ref_init();
2543 goto free_auto_defrag;
2545 err = btrfs_prelim_ref_init();
2547 goto free_delayed_ref;
2549 err = btrfs_end_io_wq_init();
2551 goto free_prelim_ref;
2553 err = btrfs_interface_init();
2555 goto free_end_io_wq;
2557 btrfs_init_lockdep();
2559 btrfs_print_mod_info();
2561 err = btrfs_run_sanity_tests();
2563 goto unregister_ioctl;
2565 err = register_filesystem(&btrfs_fs_type);
2567 goto unregister_ioctl;
2572 btrfs_interface_exit();
2574 btrfs_end_io_wq_exit();
2576 btrfs_prelim_ref_exit();
2578 btrfs_delayed_ref_exit();
2580 btrfs_auto_defrag_exit();
2582 btrfs_delayed_inode_exit();
2584 ordered_data_exit();
2587 free_extent_state_cache:
2588 extent_state_cache_exit();
2592 btrfs_destroy_cachep();
2594 btrfs_exit_compress();
2600 static void __exit exit_btrfs_fs(void)
2602 btrfs_destroy_cachep();
2603 btrfs_delayed_ref_exit();
2604 btrfs_auto_defrag_exit();
2605 btrfs_delayed_inode_exit();
2606 btrfs_prelim_ref_exit();
2607 ordered_data_exit();
2609 extent_state_cache_exit();
2611 btrfs_interface_exit();
2612 btrfs_end_io_wq_exit();
2613 unregister_filesystem(&btrfs_fs_type);
2615 btrfs_cleanup_fs_uuids();
2616 btrfs_exit_compress();
2619 late_initcall(init_btrfs_fs);
2620 module_exit(exit_btrfs_fs)
2622 MODULE_LICENSE("GPL");
2623 MODULE_SOFTDEP("pre: crc32c");
2624 MODULE_SOFTDEP("pre: xxhash64");
2625 MODULE_SOFTDEP("pre: sha256");
2626 MODULE_SOFTDEP("pre: blake2b-256");