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>
8 #include <linux/buffer_head.h>
10 #include <linux/pagemap.h>
11 #include <linux/highmem.h>
12 #include <linux/time.h>
13 #include <linux/init.h>
14 #include <linux/seq_file.h>
15 #include <linux/string.h>
16 #include <linux/backing-dev.h>
17 #include <linux/mount.h>
18 #include <linux/mpage.h>
19 #include <linux/swap.h>
20 #include <linux/writeback.h>
21 #include <linux/statfs.h>
22 #include <linux/compat.h>
23 #include <linux/parser.h>
24 #include <linux/ctype.h>
25 #include <linux/namei.h>
26 #include <linux/miscdevice.h>
27 #include <linux/magic.h>
28 #include <linux/slab.h>
29 #include <linux/cleancache.h>
30 #include <linux/ratelimit.h>
31 #include <linux/crc32c.h>
32 #include <linux/btrfs.h>
33 #include "delayed-inode.h"
36 #include "transaction.h"
37 #include "btrfs_inode.h"
38 #include "print-tree.h"
43 #include "compression.h"
44 #include "rcu-string.h"
45 #include "dev-replace.h"
46 #include "free-space-cache.h"
48 #include "tests/btrfs-tests.h"
51 #define CREATE_TRACE_POINTS
52 #include <trace/events/btrfs.h>
54 static const struct super_operations btrfs_super_ops;
57 * Types for mounting the default subvolume and a subvolume explicitly
58 * requested by subvol=/path. That way the callchain is straightforward and we
59 * don't have to play tricks with the mount options and recursive calls to
62 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
64 static struct file_system_type btrfs_fs_type;
65 static struct file_system_type btrfs_root_fs_type;
67 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
69 const char *btrfs_decode_error(int errno)
71 char *errstr = "unknown";
75 errstr = "IO failure";
78 errstr = "Out of memory";
81 errstr = "Readonly filesystem";
84 errstr = "Object already exists";
87 errstr = "No space left";
90 errstr = "No such entry";
98 * __btrfs_handle_fs_error decodes expected errors from the caller and
99 * invokes the approciate error response.
102 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
103 unsigned int line, int errno, const char *fmt, ...)
105 struct super_block *sb = fs_info->sb;
111 * Special case: if the error is EROFS, and we're already
112 * under SB_RDONLY, then it is safe here.
114 if (errno == -EROFS && sb_rdonly(sb))
118 errstr = btrfs_decode_error(errno);
120 struct va_format vaf;
127 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
128 sb->s_id, function, line, errno, errstr, &vaf);
131 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
132 sb->s_id, function, line, errno, errstr);
137 * Today we only save the error info to memory. Long term we'll
138 * also send it down to the disk
140 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
142 /* Don't go through full error handling during mount */
143 if (!(sb->s_flags & SB_BORN))
149 /* btrfs handle error by forcing the filesystem readonly */
150 sb->s_flags |= SB_RDONLY;
151 btrfs_info(fs_info, "forced readonly");
153 * Note that a running device replace operation is not canceled here
154 * although there is no way to update the progress. It would add the
155 * risk of a deadlock, therefore the canceling is omitted. The only
156 * penalty is that some I/O remains active until the procedure
157 * completes. The next time when the filesystem is mounted writeable
158 * again, the device replace operation continues.
163 static const char * const logtypes[] = {
176 * Use one ratelimit state per log level so that a flood of less important
177 * messages doesn't cause more important ones to be dropped.
179 static struct ratelimit_state printk_limits[] = {
180 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
181 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
182 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
183 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
184 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
185 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
186 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
187 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
190 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
192 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
193 struct va_format vaf;
196 const char *type = logtypes[4];
197 struct ratelimit_state *ratelimit = &printk_limits[4];
201 while ((kern_level = printk_get_level(fmt)) != 0) {
202 size_t size = printk_skip_level(fmt) - fmt;
204 if (kern_level >= '0' && kern_level <= '7') {
205 memcpy(lvl, fmt, size);
207 type = logtypes[kern_level - '0'];
208 ratelimit = &printk_limits[kern_level - '0'];
216 if (__ratelimit(ratelimit))
217 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
218 fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
225 * We only mark the transaction aborted and then set the file system read-only.
226 * This will prevent new transactions from starting or trying to join this
229 * This means that error recovery at the call site is limited to freeing
230 * any local memory allocations and passing the error code up without
231 * further cleanup. The transaction should complete as it normally would
232 * in the call path but will return -EIO.
234 * We'll complete the cleanup in btrfs_end_transaction and
235 * btrfs_commit_transaction.
238 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
239 const char *function,
240 unsigned int line, int errno)
242 struct btrfs_fs_info *fs_info = trans->fs_info;
244 trans->aborted = errno;
245 /* Nothing used. The other threads that have joined this
246 * transaction may be able to continue. */
247 if (!trans->dirty && list_empty(&trans->new_bgs)) {
250 errstr = btrfs_decode_error(errno);
252 "%s:%d: Aborting unused transaction(%s).",
253 function, line, errstr);
256 WRITE_ONCE(trans->transaction->aborted, errno);
257 /* Wake up anybody who may be waiting on this transaction */
258 wake_up(&fs_info->transaction_wait);
259 wake_up(&fs_info->transaction_blocked_wait);
260 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
263 * __btrfs_panic decodes unexpected, fatal errors from the caller,
264 * issues an alert, and either panics or BUGs, depending on mount options.
267 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
268 unsigned int line, int errno, const char *fmt, ...)
270 char *s_id = "<unknown>";
272 struct va_format vaf = { .fmt = fmt };
276 s_id = fs_info->sb->s_id;
281 errstr = btrfs_decode_error(errno);
282 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
283 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
284 s_id, function, line, &vaf, errno, errstr);
286 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
287 function, line, &vaf, errno, errstr);
289 /* Caller calls BUG() */
292 static void btrfs_put_super(struct super_block *sb)
294 close_ctree(btrfs_sb(sb));
303 Opt_compress_force_type,
308 Opt_flushoncommit, Opt_noflushoncommit,
309 Opt_inode_cache, Opt_noinode_cache,
311 Opt_barrier, Opt_nobarrier,
312 Opt_datacow, Opt_nodatacow,
313 Opt_datasum, Opt_nodatasum,
314 Opt_defrag, Opt_nodefrag,
315 Opt_discard, Opt_nodiscard,
319 Opt_rescan_uuid_tree,
321 Opt_space_cache, Opt_no_space_cache,
322 Opt_space_cache_version,
324 Opt_ssd_spread, Opt_nossd_spread,
329 Opt_treelog, Opt_notreelog,
331 Opt_user_subvol_rm_allowed,
333 /* Deprecated options */
338 /* Debugging options */
340 Opt_check_integrity_including_extent_data,
341 Opt_check_integrity_print_mask,
342 Opt_enospc_debug, Opt_noenospc_debug,
343 #ifdef CONFIG_BTRFS_DEBUG
344 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
346 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
352 static const match_table_t tokens = {
354 {Opt_noacl, "noacl"},
355 {Opt_clear_cache, "clear_cache"},
356 {Opt_commit_interval, "commit=%u"},
357 {Opt_compress, "compress"},
358 {Opt_compress_type, "compress=%s"},
359 {Opt_compress_force, "compress-force"},
360 {Opt_compress_force_type, "compress-force=%s"},
361 {Opt_degraded, "degraded"},
362 {Opt_device, "device=%s"},
363 {Opt_fatal_errors, "fatal_errors=%s"},
364 {Opt_flushoncommit, "flushoncommit"},
365 {Opt_noflushoncommit, "noflushoncommit"},
366 {Opt_inode_cache, "inode_cache"},
367 {Opt_noinode_cache, "noinode_cache"},
368 {Opt_max_inline, "max_inline=%s"},
369 {Opt_barrier, "barrier"},
370 {Opt_nobarrier, "nobarrier"},
371 {Opt_datacow, "datacow"},
372 {Opt_nodatacow, "nodatacow"},
373 {Opt_datasum, "datasum"},
374 {Opt_nodatasum, "nodatasum"},
375 {Opt_defrag, "autodefrag"},
376 {Opt_nodefrag, "noautodefrag"},
377 {Opt_discard, "discard"},
378 {Opt_nodiscard, "nodiscard"},
379 {Opt_nologreplay, "nologreplay"},
380 {Opt_norecovery, "norecovery"},
381 {Opt_ratio, "metadata_ratio=%u"},
382 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
383 {Opt_skip_balance, "skip_balance"},
384 {Opt_space_cache, "space_cache"},
385 {Opt_no_space_cache, "nospace_cache"},
386 {Opt_space_cache_version, "space_cache=%s"},
388 {Opt_nossd, "nossd"},
389 {Opt_ssd_spread, "ssd_spread"},
390 {Opt_nossd_spread, "nossd_spread"},
391 {Opt_subvol, "subvol=%s"},
392 {Opt_subvol_empty, "subvol="},
393 {Opt_subvolid, "subvolid=%s"},
394 {Opt_thread_pool, "thread_pool=%u"},
395 {Opt_treelog, "treelog"},
396 {Opt_notreelog, "notreelog"},
397 {Opt_usebackuproot, "usebackuproot"},
398 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
400 /* Deprecated options */
401 {Opt_alloc_start, "alloc_start=%s"},
402 {Opt_recovery, "recovery"},
403 {Opt_subvolrootid, "subvolrootid=%d"},
405 /* Debugging options */
406 {Opt_check_integrity, "check_int"},
407 {Opt_check_integrity_including_extent_data, "check_int_data"},
408 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
409 {Opt_enospc_debug, "enospc_debug"},
410 {Opt_noenospc_debug, "noenospc_debug"},
411 #ifdef CONFIG_BTRFS_DEBUG
412 {Opt_fragment_data, "fragment=data"},
413 {Opt_fragment_metadata, "fragment=metadata"},
414 {Opt_fragment_all, "fragment=all"},
416 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
417 {Opt_ref_verify, "ref_verify"},
423 * Regular mount options parser. Everything that is needed only when
424 * reading in a new superblock is parsed here.
425 * XXX JDM: This needs to be cleaned up for remount.
427 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
428 unsigned long new_flags)
430 substring_t args[MAX_OPT_ARGS];
436 bool compress_force = false;
437 enum btrfs_compression_type saved_compress_type;
438 bool saved_compress_force;
441 cache_gen = btrfs_super_cache_generation(info->super_copy);
442 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
443 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
445 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
448 * Even the options are empty, we still need to do extra check
454 while ((p = strsep(&options, ",")) != NULL) {
459 token = match_token(p, tokens, args);
462 btrfs_info(info, "allowing degraded mounts");
463 btrfs_set_opt(info->mount_opt, DEGRADED);
466 case Opt_subvol_empty:
468 case Opt_subvolrootid:
471 * These are parsed by btrfs_parse_subvol_options
472 * and btrfs_parse_early_options
473 * and can be happily ignored here.
477 btrfs_set_and_info(info, NODATASUM,
478 "setting nodatasum");
481 if (btrfs_test_opt(info, NODATASUM)) {
482 if (btrfs_test_opt(info, NODATACOW))
484 "setting datasum, datacow enabled");
486 btrfs_info(info, "setting datasum");
488 btrfs_clear_opt(info->mount_opt, NODATACOW);
489 btrfs_clear_opt(info->mount_opt, NODATASUM);
492 if (!btrfs_test_opt(info, NODATACOW)) {
493 if (!btrfs_test_opt(info, COMPRESS) ||
494 !btrfs_test_opt(info, FORCE_COMPRESS)) {
496 "setting nodatacow, compression disabled");
498 btrfs_info(info, "setting nodatacow");
501 btrfs_clear_opt(info->mount_opt, COMPRESS);
502 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
503 btrfs_set_opt(info->mount_opt, NODATACOW);
504 btrfs_set_opt(info->mount_opt, NODATASUM);
507 btrfs_clear_and_info(info, NODATACOW,
510 case Opt_compress_force:
511 case Opt_compress_force_type:
512 compress_force = true;
515 case Opt_compress_type:
516 saved_compress_type = btrfs_test_opt(info,
518 info->compress_type : BTRFS_COMPRESS_NONE;
519 saved_compress_force =
520 btrfs_test_opt(info, FORCE_COMPRESS);
521 if (token == Opt_compress ||
522 token == Opt_compress_force ||
523 strncmp(args[0].from, "zlib", 4) == 0) {
524 compress_type = "zlib";
526 info->compress_type = BTRFS_COMPRESS_ZLIB;
527 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
529 * args[0] contains uninitialized data since
530 * for these tokens we don't expect any
533 if (token != Opt_compress &&
534 token != Opt_compress_force)
535 info->compress_level =
536 btrfs_compress_str2level(args[0].from);
537 btrfs_set_opt(info->mount_opt, COMPRESS);
538 btrfs_clear_opt(info->mount_opt, NODATACOW);
539 btrfs_clear_opt(info->mount_opt, NODATASUM);
541 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
542 compress_type = "lzo";
543 info->compress_type = BTRFS_COMPRESS_LZO;
544 btrfs_set_opt(info->mount_opt, COMPRESS);
545 btrfs_clear_opt(info->mount_opt, NODATACOW);
546 btrfs_clear_opt(info->mount_opt, NODATASUM);
547 btrfs_set_fs_incompat(info, COMPRESS_LZO);
549 } else if (strcmp(args[0].from, "zstd") == 0) {
550 compress_type = "zstd";
551 info->compress_type = BTRFS_COMPRESS_ZSTD;
552 btrfs_set_opt(info->mount_opt, COMPRESS);
553 btrfs_clear_opt(info->mount_opt, NODATACOW);
554 btrfs_clear_opt(info->mount_opt, NODATASUM);
555 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
557 } else if (strncmp(args[0].from, "no", 2) == 0) {
558 compress_type = "no";
559 btrfs_clear_opt(info->mount_opt, COMPRESS);
560 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
561 compress_force = false;
568 if (compress_force) {
569 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
572 * If we remount from compress-force=xxx to
573 * compress=xxx, we need clear FORCE_COMPRESS
574 * flag, otherwise, there is no way for users
575 * to disable forcible compression separately.
577 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
579 if ((btrfs_test_opt(info, COMPRESS) &&
580 (info->compress_type != saved_compress_type ||
581 compress_force != saved_compress_force)) ||
582 (!btrfs_test_opt(info, COMPRESS) &&
584 btrfs_info(info, "%s %s compression, level %d",
585 (compress_force) ? "force" : "use",
586 compress_type, info->compress_level);
588 compress_force = false;
591 btrfs_set_and_info(info, SSD,
592 "enabling ssd optimizations");
593 btrfs_clear_opt(info->mount_opt, NOSSD);
596 btrfs_set_and_info(info, SSD,
597 "enabling ssd optimizations");
598 btrfs_set_and_info(info, SSD_SPREAD,
599 "using spread ssd allocation scheme");
600 btrfs_clear_opt(info->mount_opt, NOSSD);
603 btrfs_set_opt(info->mount_opt, NOSSD);
604 btrfs_clear_and_info(info, SSD,
605 "not using ssd optimizations");
607 case Opt_nossd_spread:
608 btrfs_clear_and_info(info, SSD_SPREAD,
609 "not using spread ssd allocation scheme");
612 btrfs_clear_and_info(info, NOBARRIER,
613 "turning on barriers");
616 btrfs_set_and_info(info, NOBARRIER,
617 "turning off barriers");
619 case Opt_thread_pool:
620 ret = match_int(&args[0], &intarg);
623 } else if (intarg == 0) {
627 info->thread_pool_size = intarg;
630 num = match_strdup(&args[0]);
632 info->max_inline = memparse(num, NULL);
635 if (info->max_inline) {
636 info->max_inline = min_t(u64,
640 btrfs_info(info, "max_inline at %llu",
647 case Opt_alloc_start:
649 "option alloc_start is obsolete, ignored");
652 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
653 info->sb->s_flags |= SB_POSIXACL;
656 btrfs_err(info, "support for ACL not compiled in!");
661 info->sb->s_flags &= ~SB_POSIXACL;
664 btrfs_set_and_info(info, NOTREELOG,
665 "disabling tree log");
668 btrfs_clear_and_info(info, NOTREELOG,
669 "enabling tree log");
672 case Opt_nologreplay:
673 btrfs_set_and_info(info, NOLOGREPLAY,
674 "disabling log replay at mount time");
676 case Opt_flushoncommit:
677 btrfs_set_and_info(info, FLUSHONCOMMIT,
678 "turning on flush-on-commit");
680 case Opt_noflushoncommit:
681 btrfs_clear_and_info(info, FLUSHONCOMMIT,
682 "turning off flush-on-commit");
685 ret = match_int(&args[0], &intarg);
688 info->metadata_ratio = intarg;
689 btrfs_info(info, "metadata ratio %u",
690 info->metadata_ratio);
693 btrfs_set_and_info(info, DISCARD,
694 "turning on discard");
697 btrfs_clear_and_info(info, DISCARD,
698 "turning off discard");
700 case Opt_space_cache:
701 case Opt_space_cache_version:
702 if (token == Opt_space_cache ||
703 strcmp(args[0].from, "v1") == 0) {
704 btrfs_clear_opt(info->mount_opt,
706 btrfs_set_and_info(info, SPACE_CACHE,
707 "enabling disk space caching");
708 } else if (strcmp(args[0].from, "v2") == 0) {
709 btrfs_clear_opt(info->mount_opt,
711 btrfs_set_and_info(info, FREE_SPACE_TREE,
712 "enabling free space tree");
718 case Opt_rescan_uuid_tree:
719 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
721 case Opt_no_space_cache:
722 if (btrfs_test_opt(info, SPACE_CACHE)) {
723 btrfs_clear_and_info(info, SPACE_CACHE,
724 "disabling disk space caching");
726 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
727 btrfs_clear_and_info(info, FREE_SPACE_TREE,
728 "disabling free space tree");
731 case Opt_inode_cache:
732 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
733 "enabling inode map caching");
735 case Opt_noinode_cache:
736 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
737 "disabling inode map caching");
739 case Opt_clear_cache:
740 btrfs_set_and_info(info, CLEAR_CACHE,
741 "force clearing of disk cache");
743 case Opt_user_subvol_rm_allowed:
744 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
746 case Opt_enospc_debug:
747 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
749 case Opt_noenospc_debug:
750 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
753 btrfs_set_and_info(info, AUTO_DEFRAG,
754 "enabling auto defrag");
757 btrfs_clear_and_info(info, AUTO_DEFRAG,
758 "disabling auto defrag");
762 "'recovery' is deprecated, use 'usebackuproot' instead");
763 case Opt_usebackuproot:
765 "trying to use backup root at mount time");
766 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
768 case Opt_skip_balance:
769 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
771 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
772 case Opt_check_integrity_including_extent_data:
774 "enabling check integrity including extent data");
775 btrfs_set_opt(info->mount_opt,
776 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
777 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
779 case Opt_check_integrity:
780 btrfs_info(info, "enabling check integrity");
781 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
783 case Opt_check_integrity_print_mask:
784 ret = match_int(&args[0], &intarg);
787 info->check_integrity_print_mask = intarg;
788 btrfs_info(info, "check_integrity_print_mask 0x%x",
789 info->check_integrity_print_mask);
792 case Opt_check_integrity_including_extent_data:
793 case Opt_check_integrity:
794 case Opt_check_integrity_print_mask:
796 "support for check_integrity* not compiled in!");
800 case Opt_fatal_errors:
801 if (strcmp(args[0].from, "panic") == 0)
802 btrfs_set_opt(info->mount_opt,
803 PANIC_ON_FATAL_ERROR);
804 else if (strcmp(args[0].from, "bug") == 0)
805 btrfs_clear_opt(info->mount_opt,
806 PANIC_ON_FATAL_ERROR);
812 case Opt_commit_interval:
814 ret = match_int(&args[0], &intarg);
819 "using default commit interval %us",
820 BTRFS_DEFAULT_COMMIT_INTERVAL);
821 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
822 } else if (intarg > 300) {
823 btrfs_warn(info, "excessive commit interval %d",
826 info->commit_interval = intarg;
828 #ifdef CONFIG_BTRFS_DEBUG
829 case Opt_fragment_all:
830 btrfs_info(info, "fragmenting all space");
831 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
832 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
834 case Opt_fragment_metadata:
835 btrfs_info(info, "fragmenting metadata");
836 btrfs_set_opt(info->mount_opt,
839 case Opt_fragment_data:
840 btrfs_info(info, "fragmenting data");
841 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
844 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
846 btrfs_info(info, "doing ref verification");
847 btrfs_set_opt(info->mount_opt, REF_VERIFY);
851 btrfs_info(info, "unrecognized mount option '%s'", p);
860 * Extra check for current option against current flag
862 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
864 "nologreplay must be used with ro mount option");
868 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
869 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
870 !btrfs_test_opt(info, CLEAR_CACHE)) {
871 btrfs_err(info, "cannot disable free space tree");
875 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
876 btrfs_info(info, "disk space caching is enabled");
877 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
878 btrfs_info(info, "using free space tree");
883 * Parse mount options that are required early in the mount process.
885 * All other options will be parsed on much later in the mount process and
886 * only when we need to allocate a new super block.
888 static int btrfs_parse_early_options(const char *options, fmode_t flags,
889 void *holder, struct btrfs_fs_devices **fs_devices)
891 substring_t args[MAX_OPT_ARGS];
892 char *device_name, *opts, *orig, *p;
899 * strsep changes the string, duplicate it because btrfs_parse_options
902 opts = kstrdup(options, GFP_KERNEL);
907 while ((p = strsep(&opts, ",")) != NULL) {
913 token = match_token(p, tokens, args);
914 if (token == Opt_device) {
915 device_name = match_strdup(&args[0]);
920 error = btrfs_scan_one_device(device_name,
921 flags, holder, fs_devices);
934 * Parse mount options that are related to subvolume id
936 * The value is later passed to mount_subvol()
938 static int btrfs_parse_subvol_options(const char *options, fmode_t flags,
939 char **subvol_name, u64 *subvol_objectid)
941 substring_t args[MAX_OPT_ARGS];
942 char *opts, *orig, *p;
950 * strsep changes the string, duplicate it because
951 * btrfs_parse_early_options gets called later
953 opts = kstrdup(options, GFP_KERNEL);
958 while ((p = strsep(&opts, ",")) != NULL) {
963 token = match_token(p, tokens, args);
967 *subvol_name = match_strdup(&args[0]);
974 error = match_u64(&args[0], &subvolid);
978 /* we want the original fs_tree */
980 subvolid = BTRFS_FS_TREE_OBJECTID;
982 *subvol_objectid = subvolid;
984 case Opt_subvolrootid:
985 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
997 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1000 struct btrfs_root *root = fs_info->tree_root;
1001 struct btrfs_root *fs_root;
1002 struct btrfs_root_ref *root_ref;
1003 struct btrfs_inode_ref *inode_ref;
1004 struct btrfs_key key;
1005 struct btrfs_path *path = NULL;
1006 char *name = NULL, *ptr;
1011 path = btrfs_alloc_path();
1016 path->leave_spinning = 1;
1018 name = kmalloc(PATH_MAX, GFP_KERNEL);
1023 ptr = name + PATH_MAX - 1;
1027 * Walk up the subvolume trees in the tree of tree roots by root
1028 * backrefs until we hit the top-level subvolume.
1030 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1031 key.objectid = subvol_objectid;
1032 key.type = BTRFS_ROOT_BACKREF_KEY;
1033 key.offset = (u64)-1;
1035 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1038 } else if (ret > 0) {
1039 ret = btrfs_previous_item(root, path, subvol_objectid,
1040 BTRFS_ROOT_BACKREF_KEY);
1043 } else if (ret > 0) {
1049 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1050 subvol_objectid = key.offset;
1052 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1053 struct btrfs_root_ref);
1054 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1057 ret = -ENAMETOOLONG;
1060 read_extent_buffer(path->nodes[0], ptr + 1,
1061 (unsigned long)(root_ref + 1), len);
1063 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1064 btrfs_release_path(path);
1066 key.objectid = subvol_objectid;
1067 key.type = BTRFS_ROOT_ITEM_KEY;
1068 key.offset = (u64)-1;
1069 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1070 if (IS_ERR(fs_root)) {
1071 ret = PTR_ERR(fs_root);
1076 * Walk up the filesystem tree by inode refs until we hit the
1079 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1080 key.objectid = dirid;
1081 key.type = BTRFS_INODE_REF_KEY;
1082 key.offset = (u64)-1;
1084 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1087 } else if (ret > 0) {
1088 ret = btrfs_previous_item(fs_root, path, dirid,
1089 BTRFS_INODE_REF_KEY);
1092 } else if (ret > 0) {
1098 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1101 inode_ref = btrfs_item_ptr(path->nodes[0],
1103 struct btrfs_inode_ref);
1104 len = btrfs_inode_ref_name_len(path->nodes[0],
1108 ret = -ENAMETOOLONG;
1111 read_extent_buffer(path->nodes[0], ptr + 1,
1112 (unsigned long)(inode_ref + 1), len);
1114 btrfs_release_path(path);
1118 btrfs_free_path(path);
1119 if (ptr == name + PATH_MAX - 1) {
1123 memmove(name, ptr, name + PATH_MAX - ptr);
1128 btrfs_free_path(path);
1130 return ERR_PTR(ret);
1133 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1135 struct btrfs_root *root = fs_info->tree_root;
1136 struct btrfs_dir_item *di;
1137 struct btrfs_path *path;
1138 struct btrfs_key location;
1141 path = btrfs_alloc_path();
1144 path->leave_spinning = 1;
1147 * Find the "default" dir item which points to the root item that we
1148 * will mount by default if we haven't been given a specific subvolume
1151 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1152 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1154 btrfs_free_path(path);
1159 * Ok the default dir item isn't there. This is weird since
1160 * it's always been there, but don't freak out, just try and
1161 * mount the top-level subvolume.
1163 btrfs_free_path(path);
1164 *objectid = BTRFS_FS_TREE_OBJECTID;
1168 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1169 btrfs_free_path(path);
1170 *objectid = location.objectid;
1174 static int btrfs_fill_super(struct super_block *sb,
1175 struct btrfs_fs_devices *fs_devices,
1178 struct inode *inode;
1179 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1180 struct btrfs_key key;
1183 sb->s_maxbytes = MAX_LFS_FILESIZE;
1184 sb->s_magic = BTRFS_SUPER_MAGIC;
1185 sb->s_op = &btrfs_super_ops;
1186 sb->s_d_op = &btrfs_dentry_operations;
1187 sb->s_export_op = &btrfs_export_ops;
1188 sb->s_xattr = btrfs_xattr_handlers;
1189 sb->s_time_gran = 1;
1190 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1191 sb->s_flags |= SB_POSIXACL;
1193 sb->s_flags |= SB_I_VERSION;
1194 sb->s_iflags |= SB_I_CGROUPWB;
1196 err = super_setup_bdi(sb);
1198 btrfs_err(fs_info, "super_setup_bdi failed");
1202 err = open_ctree(sb, fs_devices, (char *)data);
1204 btrfs_err(fs_info, "open_ctree failed");
1208 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1209 key.type = BTRFS_INODE_ITEM_KEY;
1211 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1212 if (IS_ERR(inode)) {
1213 err = PTR_ERR(inode);
1217 sb->s_root = d_make_root(inode);
1223 cleancache_init_fs(sb);
1224 sb->s_flags |= SB_ACTIVE;
1228 close_ctree(fs_info);
1232 int btrfs_sync_fs(struct super_block *sb, int wait)
1234 struct btrfs_trans_handle *trans;
1235 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1236 struct btrfs_root *root = fs_info->tree_root;
1238 trace_btrfs_sync_fs(fs_info, wait);
1241 filemap_flush(fs_info->btree_inode->i_mapping);
1245 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1247 trans = btrfs_attach_transaction_barrier(root);
1248 if (IS_ERR(trans)) {
1249 /* no transaction, don't bother */
1250 if (PTR_ERR(trans) == -ENOENT) {
1252 * Exit unless we have some pending changes
1253 * that need to go through commit
1255 if (fs_info->pending_changes == 0)
1258 * A non-blocking test if the fs is frozen. We must not
1259 * start a new transaction here otherwise a deadlock
1260 * happens. The pending operations are delayed to the
1261 * next commit after thawing.
1263 if (sb_start_write_trylock(sb))
1267 trans = btrfs_start_transaction(root, 0);
1270 return PTR_ERR(trans);
1272 return btrfs_commit_transaction(trans);
1275 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1277 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1278 const char *compress_type;
1280 if (btrfs_test_opt(info, DEGRADED))
1281 seq_puts(seq, ",degraded");
1282 if (btrfs_test_opt(info, NODATASUM))
1283 seq_puts(seq, ",nodatasum");
1284 if (btrfs_test_opt(info, NODATACOW))
1285 seq_puts(seq, ",nodatacow");
1286 if (btrfs_test_opt(info, NOBARRIER))
1287 seq_puts(seq, ",nobarrier");
1288 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1289 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1290 if (info->thread_pool_size != min_t(unsigned long,
1291 num_online_cpus() + 2, 8))
1292 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1293 if (btrfs_test_opt(info, COMPRESS)) {
1294 compress_type = btrfs_compress_type2str(info->compress_type);
1295 if (btrfs_test_opt(info, FORCE_COMPRESS))
1296 seq_printf(seq, ",compress-force=%s", compress_type);
1298 seq_printf(seq, ",compress=%s", compress_type);
1299 if (info->compress_level)
1300 seq_printf(seq, ":%d", info->compress_level);
1302 if (btrfs_test_opt(info, NOSSD))
1303 seq_puts(seq, ",nossd");
1304 if (btrfs_test_opt(info, SSD_SPREAD))
1305 seq_puts(seq, ",ssd_spread");
1306 else if (btrfs_test_opt(info, SSD))
1307 seq_puts(seq, ",ssd");
1308 if (btrfs_test_opt(info, NOTREELOG))
1309 seq_puts(seq, ",notreelog");
1310 if (btrfs_test_opt(info, NOLOGREPLAY))
1311 seq_puts(seq, ",nologreplay");
1312 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1313 seq_puts(seq, ",flushoncommit");
1314 if (btrfs_test_opt(info, DISCARD))
1315 seq_puts(seq, ",discard");
1316 if (!(info->sb->s_flags & SB_POSIXACL))
1317 seq_puts(seq, ",noacl");
1318 if (btrfs_test_opt(info, SPACE_CACHE))
1319 seq_puts(seq, ",space_cache");
1320 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1321 seq_puts(seq, ",space_cache=v2");
1323 seq_puts(seq, ",nospace_cache");
1324 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1325 seq_puts(seq, ",rescan_uuid_tree");
1326 if (btrfs_test_opt(info, CLEAR_CACHE))
1327 seq_puts(seq, ",clear_cache");
1328 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1329 seq_puts(seq, ",user_subvol_rm_allowed");
1330 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1331 seq_puts(seq, ",enospc_debug");
1332 if (btrfs_test_opt(info, AUTO_DEFRAG))
1333 seq_puts(seq, ",autodefrag");
1334 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1335 seq_puts(seq, ",inode_cache");
1336 if (btrfs_test_opt(info, SKIP_BALANCE))
1337 seq_puts(seq, ",skip_balance");
1338 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1339 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1340 seq_puts(seq, ",check_int_data");
1341 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1342 seq_puts(seq, ",check_int");
1343 if (info->check_integrity_print_mask)
1344 seq_printf(seq, ",check_int_print_mask=%d",
1345 info->check_integrity_print_mask);
1347 if (info->metadata_ratio)
1348 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1349 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1350 seq_puts(seq, ",fatal_errors=panic");
1351 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1352 seq_printf(seq, ",commit=%u", info->commit_interval);
1353 #ifdef CONFIG_BTRFS_DEBUG
1354 if (btrfs_test_opt(info, FRAGMENT_DATA))
1355 seq_puts(seq, ",fragment=data");
1356 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1357 seq_puts(seq, ",fragment=metadata");
1359 if (btrfs_test_opt(info, REF_VERIFY))
1360 seq_puts(seq, ",ref_verify");
1361 seq_printf(seq, ",subvolid=%llu",
1362 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1363 seq_puts(seq, ",subvol=");
1364 seq_dentry(seq, dentry, " \t\n\\");
1368 static int btrfs_test_super(struct super_block *s, void *data)
1370 struct btrfs_fs_info *p = data;
1371 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1373 return fs_info->fs_devices == p->fs_devices;
1376 static int btrfs_set_super(struct super_block *s, void *data)
1378 int err = set_anon_super(s, data);
1380 s->s_fs_info = data;
1385 * subvolumes are identified by ino 256
1387 static inline int is_subvolume_inode(struct inode *inode)
1389 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1394 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1395 const char *device_name, struct vfsmount *mnt)
1397 struct dentry *root;
1401 if (!subvol_objectid) {
1402 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1405 root = ERR_PTR(ret);
1409 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1411 if (IS_ERR(subvol_name)) {
1412 root = ERR_CAST(subvol_name);
1419 root = mount_subtree(mnt, subvol_name);
1420 /* mount_subtree() drops our reference on the vfsmount. */
1423 if (!IS_ERR(root)) {
1424 struct super_block *s = root->d_sb;
1425 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1426 struct inode *root_inode = d_inode(root);
1427 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1430 if (!is_subvolume_inode(root_inode)) {
1431 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1435 if (subvol_objectid && root_objectid != subvol_objectid) {
1437 * This will also catch a race condition where a
1438 * subvolume which was passed by ID is renamed and
1439 * another subvolume is renamed over the old location.
1442 "subvol '%s' does not match subvolid %llu",
1443 subvol_name, subvol_objectid);
1448 root = ERR_PTR(ret);
1449 deactivate_locked_super(s);
1459 static int parse_security_options(char *orig_opts,
1460 struct security_mnt_opts *sec_opts)
1462 char *secdata = NULL;
1465 secdata = alloc_secdata();
1468 ret = security_sb_copy_data(orig_opts, secdata);
1470 free_secdata(secdata);
1473 ret = security_sb_parse_opts_str(secdata, sec_opts);
1474 free_secdata(secdata);
1478 static int setup_security_options(struct btrfs_fs_info *fs_info,
1479 struct super_block *sb,
1480 struct security_mnt_opts *sec_opts)
1485 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1488 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1492 #ifdef CONFIG_SECURITY
1493 if (!fs_info->security_opts.num_mnt_opts) {
1494 /* first time security setup, copy sec_opts to fs_info */
1495 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1498 * Since SELinux (the only one supporting security_mnt_opts)
1499 * does NOT support changing context during remount/mount of
1500 * the same sb, this must be the same or part of the same
1501 * security options, just free it.
1503 security_free_mnt_opts(sec_opts);
1510 * Find a superblock for the given device / mount point.
1512 * Note: This is based on mount_bdev from fs/super.c with a few additions
1513 * for multiple device setup. Make sure to keep it in sync.
1515 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1516 int flags, const char *device_name, void *data)
1518 struct block_device *bdev = NULL;
1519 struct super_block *s;
1520 struct btrfs_fs_devices *fs_devices = NULL;
1521 struct btrfs_fs_info *fs_info = NULL;
1522 struct security_mnt_opts new_sec_opts;
1523 fmode_t mode = FMODE_READ;
1526 if (!(flags & SB_RDONLY))
1527 mode |= FMODE_WRITE;
1529 error = btrfs_parse_early_options(data, mode, fs_type,
1532 return ERR_PTR(error);
1535 security_init_mnt_opts(&new_sec_opts);
1537 error = parse_security_options(data, &new_sec_opts);
1539 return ERR_PTR(error);
1542 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1544 goto error_sec_opts;
1547 * Setup a dummy root and fs_info for test/set super. This is because
1548 * we don't actually fill this stuff out until open_ctree, but we need
1549 * it for searching for existing supers, so this lets us do that and
1550 * then open_ctree will properly initialize everything later.
1552 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1555 goto error_sec_opts;
1558 fs_info->fs_devices = fs_devices;
1560 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1561 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1562 security_init_mnt_opts(&fs_info->security_opts);
1563 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1568 error = btrfs_open_devices(fs_devices, mode, fs_type);
1572 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1574 goto error_close_devices;
1577 bdev = fs_devices->latest_bdev;
1578 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1582 goto error_close_devices;
1586 btrfs_close_devices(fs_devices);
1587 free_fs_info(fs_info);
1588 if ((flags ^ s->s_flags) & SB_RDONLY)
1591 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1592 btrfs_sb(s)->bdev_holder = fs_type;
1593 error = btrfs_fill_super(s, fs_devices, data);
1596 deactivate_locked_super(s);
1597 goto error_sec_opts;
1600 fs_info = btrfs_sb(s);
1601 error = setup_security_options(fs_info, s, &new_sec_opts);
1603 deactivate_locked_super(s);
1604 goto error_sec_opts;
1607 return dget(s->s_root);
1609 error_close_devices:
1610 btrfs_close_devices(fs_devices);
1612 free_fs_info(fs_info);
1614 security_free_mnt_opts(&new_sec_opts);
1615 return ERR_PTR(error);
1619 * Mount function which is called by VFS layer.
1621 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1622 * which needs vfsmount* of device's root (/). This means device's root has to
1623 * be mounted internally in any case.
1626 * 1. Parse subvol id related options for later use in mount_subvol().
1628 * 2. Mount device's root (/) by calling vfs_kern_mount().
1630 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1631 * first place. In order to avoid calling btrfs_mount() again, we use
1632 * different file_system_type which is not registered to VFS by
1633 * register_filesystem() (btrfs_root_fs_type). As a result,
1634 * btrfs_mount_root() is called. The return value will be used by
1635 * mount_subtree() in mount_subvol().
1637 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1638 * "btrfs subvolume set-default", mount_subvol() is called always.
1640 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1641 const char *device_name, void *data)
1643 struct vfsmount *mnt_root;
1644 struct dentry *root;
1645 fmode_t mode = FMODE_READ;
1646 char *subvol_name = NULL;
1647 u64 subvol_objectid = 0;
1650 if (!(flags & SB_RDONLY))
1651 mode |= FMODE_WRITE;
1653 error = btrfs_parse_subvol_options(data, mode,
1654 &subvol_name, &subvol_objectid);
1657 return ERR_PTR(error);
1660 /* mount device's root (/) */
1661 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1662 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1663 if (flags & SB_RDONLY) {
1664 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1665 flags & ~SB_RDONLY, device_name, data);
1667 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1668 flags | SB_RDONLY, device_name, data);
1669 if (IS_ERR(mnt_root)) {
1670 root = ERR_CAST(mnt_root);
1674 down_write(&mnt_root->mnt_sb->s_umount);
1675 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1676 up_write(&mnt_root->mnt_sb->s_umount);
1678 root = ERR_PTR(error);
1684 if (IS_ERR(mnt_root)) {
1685 root = ERR_CAST(mnt_root);
1689 /* mount_subvol() will free subvol_name and mnt_root */
1690 root = mount_subvol(subvol_name, subvol_objectid, device_name, mnt_root);
1696 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1697 u32 new_pool_size, u32 old_pool_size)
1699 if (new_pool_size == old_pool_size)
1702 fs_info->thread_pool_size = new_pool_size;
1704 btrfs_info(fs_info, "resize thread pool %d -> %d",
1705 old_pool_size, new_pool_size);
1707 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1708 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1709 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1710 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1711 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1712 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1713 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1715 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1716 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1717 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1718 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1719 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1723 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1725 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1728 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1729 unsigned long old_opts, int flags)
1731 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1732 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1733 (flags & SB_RDONLY))) {
1734 /* wait for any defraggers to finish */
1735 wait_event(fs_info->transaction_wait,
1736 (atomic_read(&fs_info->defrag_running) == 0));
1737 if (flags & SB_RDONLY)
1738 sync_filesystem(fs_info->sb);
1742 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1743 unsigned long old_opts)
1746 * We need to cleanup all defragable inodes if the autodefragment is
1747 * close or the filesystem is read only.
1749 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1750 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1751 btrfs_cleanup_defrag_inodes(fs_info);
1754 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1757 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1759 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1760 struct btrfs_root *root = fs_info->tree_root;
1761 unsigned old_flags = sb->s_flags;
1762 unsigned long old_opts = fs_info->mount_opt;
1763 unsigned long old_compress_type = fs_info->compress_type;
1764 u64 old_max_inline = fs_info->max_inline;
1765 u32 old_thread_pool_size = fs_info->thread_pool_size;
1766 u32 old_metadata_ratio = fs_info->metadata_ratio;
1769 sync_filesystem(sb);
1770 btrfs_remount_prepare(fs_info);
1773 struct security_mnt_opts new_sec_opts;
1775 security_init_mnt_opts(&new_sec_opts);
1776 ret = parse_security_options(data, &new_sec_opts);
1779 ret = setup_security_options(fs_info, sb,
1782 security_free_mnt_opts(&new_sec_opts);
1787 ret = btrfs_parse_options(fs_info, data, *flags);
1791 btrfs_remount_begin(fs_info, old_opts, *flags);
1792 btrfs_resize_thread_pool(fs_info,
1793 fs_info->thread_pool_size, old_thread_pool_size);
1795 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1798 if (*flags & SB_RDONLY) {
1800 * this also happens on 'umount -rf' or on shutdown, when
1801 * the filesystem is busy.
1803 cancel_work_sync(&fs_info->async_reclaim_work);
1805 /* wait for the uuid_scan task to finish */
1806 down(&fs_info->uuid_tree_rescan_sem);
1807 /* avoid complains from lockdep et al. */
1808 up(&fs_info->uuid_tree_rescan_sem);
1810 sb->s_flags |= SB_RDONLY;
1813 * Setting SB_RDONLY will put the cleaner thread to
1814 * sleep at the next loop if it's already active.
1815 * If it's already asleep, we'll leave unused block
1816 * groups on disk until we're mounted read-write again
1817 * unless we clean them up here.
1819 btrfs_delete_unused_bgs(fs_info);
1821 btrfs_dev_replace_suspend_for_unmount(fs_info);
1822 btrfs_scrub_cancel(fs_info);
1823 btrfs_pause_balance(fs_info);
1825 ret = btrfs_commit_super(fs_info);
1829 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1831 "Remounting read-write after error is not allowed");
1835 if (fs_info->fs_devices->rw_devices == 0) {
1840 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1842 "too many missing devices, writeable remount is not allowed");
1847 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1852 ret = btrfs_cleanup_fs_roots(fs_info);
1856 /* recover relocation */
1857 mutex_lock(&fs_info->cleaner_mutex);
1858 ret = btrfs_recover_relocation(root);
1859 mutex_unlock(&fs_info->cleaner_mutex);
1863 ret = btrfs_resume_balance_async(fs_info);
1867 ret = btrfs_resume_dev_replace_async(fs_info);
1869 btrfs_warn(fs_info, "failed to resume dev_replace");
1873 btrfs_qgroup_rescan_resume(fs_info);
1875 if (!fs_info->uuid_root) {
1876 btrfs_info(fs_info, "creating UUID tree");
1877 ret = btrfs_create_uuid_tree(fs_info);
1880 "failed to create the UUID tree %d",
1885 sb->s_flags &= ~SB_RDONLY;
1887 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1890 wake_up_process(fs_info->transaction_kthread);
1891 btrfs_remount_cleanup(fs_info, old_opts);
1895 /* We've hit an error - don't reset SB_RDONLY */
1897 old_flags |= SB_RDONLY;
1898 sb->s_flags = old_flags;
1899 fs_info->mount_opt = old_opts;
1900 fs_info->compress_type = old_compress_type;
1901 fs_info->max_inline = old_max_inline;
1902 btrfs_resize_thread_pool(fs_info,
1903 old_thread_pool_size, fs_info->thread_pool_size);
1904 fs_info->metadata_ratio = old_metadata_ratio;
1905 btrfs_remount_cleanup(fs_info, old_opts);
1909 /* Used to sort the devices by max_avail(descending sort) */
1910 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1911 const void *dev_info2)
1913 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1914 ((struct btrfs_device_info *)dev_info2)->max_avail)
1916 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1917 ((struct btrfs_device_info *)dev_info2)->max_avail)
1924 * sort the devices by max_avail, in which max free extent size of each device
1925 * is stored.(Descending Sort)
1927 static inline void btrfs_descending_sort_devices(
1928 struct btrfs_device_info *devices,
1931 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1932 btrfs_cmp_device_free_bytes, NULL);
1936 * The helper to calc the free space on the devices that can be used to store
1939 static int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1942 struct btrfs_device_info *devices_info;
1943 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1944 struct btrfs_device *device;
1948 u64 min_stripe_size;
1949 int min_stripes = 1, num_stripes = 1;
1950 int i = 0, nr_devices;
1953 * We aren't under the device list lock, so this is racy-ish, but good
1954 * enough for our purposes.
1956 nr_devices = fs_info->fs_devices->open_devices;
1959 nr_devices = fs_info->fs_devices->open_devices;
1967 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1972 /* calc min stripe number for data space allocation */
1973 type = btrfs_data_alloc_profile(fs_info);
1974 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1976 num_stripes = nr_devices;
1977 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1980 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1985 if (type & BTRFS_BLOCK_GROUP_DUP)
1986 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1988 min_stripe_size = BTRFS_STRIPE_LEN;
1991 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1992 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1993 &device->dev_state) ||
1995 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1998 if (i >= nr_devices)
2001 avail_space = device->total_bytes - device->bytes_used;
2003 /* align with stripe_len */
2004 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
2005 avail_space *= BTRFS_STRIPE_LEN;
2008 * In order to avoid overwriting the superblock on the drive,
2009 * btrfs starts at an offset of at least 1MB when doing chunk
2015 * we can use the free space in [0, skip_space - 1], subtract
2016 * it from the total.
2018 if (avail_space && avail_space >= skip_space)
2019 avail_space -= skip_space;
2023 if (avail_space < min_stripe_size)
2026 devices_info[i].dev = device;
2027 devices_info[i].max_avail = avail_space;
2035 btrfs_descending_sort_devices(devices_info, nr_devices);
2039 while (nr_devices >= min_stripes) {
2040 if (num_stripes > nr_devices)
2041 num_stripes = nr_devices;
2043 if (devices_info[i].max_avail >= min_stripe_size) {
2047 avail_space += devices_info[i].max_avail * num_stripes;
2048 alloc_size = devices_info[i].max_avail;
2049 for (j = i + 1 - num_stripes; j <= i; j++)
2050 devices_info[j].max_avail -= alloc_size;
2056 kfree(devices_info);
2057 *free_bytes = avail_space;
2062 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2064 * If there's a redundant raid level at DATA block groups, use the respective
2065 * multiplier to scale the sizes.
2067 * Unused device space usage is based on simulating the chunk allocator
2068 * algorithm that respects the device sizes and order of allocations. This is
2069 * a close approximation of the actual use but there are other factors that may
2070 * change the result (like a new metadata chunk).
2072 * If metadata is exhausted, f_bavail will be 0.
2074 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2076 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2077 struct btrfs_super_block *disk_super = fs_info->super_copy;
2078 struct list_head *head = &fs_info->space_info;
2079 struct btrfs_space_info *found;
2081 u64 total_free_data = 0;
2082 u64 total_free_meta = 0;
2083 int bits = dentry->d_sb->s_blocksize_bits;
2084 __be32 *fsid = (__be32 *)fs_info->fsid;
2085 unsigned factor = 1;
2086 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2092 list_for_each_entry_rcu(found, head, list) {
2093 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2096 total_free_data += found->disk_total - found->disk_used;
2098 btrfs_account_ro_block_groups_free_space(found);
2100 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2101 if (!list_empty(&found->block_groups[i])) {
2103 case BTRFS_RAID_DUP:
2104 case BTRFS_RAID_RAID1:
2105 case BTRFS_RAID_RAID10:
2113 * Metadata in mixed block goup profiles are accounted in data
2115 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2116 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2119 total_free_meta += found->disk_total -
2123 total_used += found->disk_used;
2128 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2129 buf->f_blocks >>= bits;
2130 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2132 /* Account global block reserve as used, it's in logical size already */
2133 spin_lock(&block_rsv->lock);
2134 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2135 if (buf->f_bfree >= block_rsv->size >> bits)
2136 buf->f_bfree -= block_rsv->size >> bits;
2139 spin_unlock(&block_rsv->lock);
2141 buf->f_bavail = div_u64(total_free_data, factor);
2142 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2145 buf->f_bavail += div_u64(total_free_data, factor);
2146 buf->f_bavail = buf->f_bavail >> bits;
2149 * We calculate the remaining metadata space minus global reserve. If
2150 * this is (supposedly) smaller than zero, there's no space. But this
2151 * does not hold in practice, the exhausted state happens where's still
2152 * some positive delta. So we apply some guesswork and compare the
2153 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2155 * We probably cannot calculate the exact threshold value because this
2156 * depends on the internal reservations requested by various
2157 * operations, so some operations that consume a few metadata will
2158 * succeed even if the Avail is zero. But this is better than the other
2163 if (!mixed && total_free_meta - thresh < block_rsv->size)
2166 buf->f_type = BTRFS_SUPER_MAGIC;
2167 buf->f_bsize = dentry->d_sb->s_blocksize;
2168 buf->f_namelen = BTRFS_NAME_LEN;
2170 /* We treat it as constant endianness (it doesn't matter _which_)
2171 because we want the fsid to come out the same whether mounted
2172 on a big-endian or little-endian host */
2173 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2174 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2175 /* Mask in the root object ID too, to disambiguate subvols */
2176 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2177 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2182 static void btrfs_kill_super(struct super_block *sb)
2184 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2185 kill_anon_super(sb);
2186 free_fs_info(fs_info);
2189 static struct file_system_type btrfs_fs_type = {
2190 .owner = THIS_MODULE,
2192 .mount = btrfs_mount,
2193 .kill_sb = btrfs_kill_super,
2194 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2197 static struct file_system_type btrfs_root_fs_type = {
2198 .owner = THIS_MODULE,
2200 .mount = btrfs_mount_root,
2201 .kill_sb = btrfs_kill_super,
2202 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2205 MODULE_ALIAS_FS("btrfs");
2207 static int btrfs_control_open(struct inode *inode, struct file *file)
2210 * The control file's private_data is used to hold the
2211 * transaction when it is started and is used to keep
2212 * track of whether a transaction is already in progress.
2214 file->private_data = NULL;
2219 * used by btrfsctl to scan devices when no FS is mounted
2221 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2224 struct btrfs_ioctl_vol_args *vol;
2225 struct btrfs_fs_devices *fs_devices;
2228 if (!capable(CAP_SYS_ADMIN))
2231 vol = memdup_user((void __user *)arg, sizeof(*vol));
2233 return PTR_ERR(vol);
2236 case BTRFS_IOC_SCAN_DEV:
2237 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2238 &btrfs_root_fs_type, &fs_devices);
2240 case BTRFS_IOC_DEVICES_READY:
2241 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2242 &btrfs_root_fs_type, &fs_devices);
2245 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2247 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2248 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2256 static int btrfs_freeze(struct super_block *sb)
2258 struct btrfs_trans_handle *trans;
2259 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2260 struct btrfs_root *root = fs_info->tree_root;
2262 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2264 * We don't need a barrier here, we'll wait for any transaction that
2265 * could be in progress on other threads (and do delayed iputs that
2266 * we want to avoid on a frozen filesystem), or do the commit
2269 trans = btrfs_attach_transaction_barrier(root);
2270 if (IS_ERR(trans)) {
2271 /* no transaction, don't bother */
2272 if (PTR_ERR(trans) == -ENOENT)
2274 return PTR_ERR(trans);
2276 return btrfs_commit_transaction(trans);
2279 static int btrfs_unfreeze(struct super_block *sb)
2281 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2283 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2287 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2289 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2290 struct btrfs_fs_devices *cur_devices;
2291 struct btrfs_device *dev, *first_dev = NULL;
2292 struct list_head *head;
2293 struct rcu_string *name;
2296 * Lightweight locking of the devices. We should not need
2297 * device_list_mutex here as we only read the device data and the list
2298 * is protected by RCU. Even if a device is deleted during the list
2299 * traversals, we'll get valid data, the freeing callback will wait at
2300 * least until until the rcu_read_unlock.
2303 cur_devices = fs_info->fs_devices;
2304 while (cur_devices) {
2305 head = &cur_devices->devices;
2306 list_for_each_entry_rcu(dev, head, dev_list) {
2307 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2311 if (!first_dev || dev->devid < first_dev->devid)
2314 cur_devices = cur_devices->seed;
2318 name = rcu_dereference(first_dev->name);
2319 seq_escape(m, name->str, " \t\n\\");
2327 static const struct super_operations btrfs_super_ops = {
2328 .drop_inode = btrfs_drop_inode,
2329 .evict_inode = btrfs_evict_inode,
2330 .put_super = btrfs_put_super,
2331 .sync_fs = btrfs_sync_fs,
2332 .show_options = btrfs_show_options,
2333 .show_devname = btrfs_show_devname,
2334 .write_inode = btrfs_write_inode,
2335 .alloc_inode = btrfs_alloc_inode,
2336 .destroy_inode = btrfs_destroy_inode,
2337 .statfs = btrfs_statfs,
2338 .remount_fs = btrfs_remount,
2339 .freeze_fs = btrfs_freeze,
2340 .unfreeze_fs = btrfs_unfreeze,
2343 static const struct file_operations btrfs_ctl_fops = {
2344 .open = btrfs_control_open,
2345 .unlocked_ioctl = btrfs_control_ioctl,
2346 .compat_ioctl = btrfs_control_ioctl,
2347 .owner = THIS_MODULE,
2348 .llseek = noop_llseek,
2351 static struct miscdevice btrfs_misc = {
2352 .minor = BTRFS_MINOR,
2353 .name = "btrfs-control",
2354 .fops = &btrfs_ctl_fops
2357 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2358 MODULE_ALIAS("devname:btrfs-control");
2360 static int __init btrfs_interface_init(void)
2362 return misc_register(&btrfs_misc);
2365 static __cold void btrfs_interface_exit(void)
2367 misc_deregister(&btrfs_misc);
2370 static void __init btrfs_print_mod_info(void)
2372 pr_info("Btrfs loaded, crc32c=%s"
2373 #ifdef CONFIG_BTRFS_DEBUG
2376 #ifdef CONFIG_BTRFS_ASSERT
2379 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2380 ", integrity-checker=on"
2382 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2389 static int __init init_btrfs_fs(void)
2395 err = btrfs_init_sysfs();
2399 btrfs_init_compress();
2401 err = btrfs_init_cachep();
2405 err = extent_io_init();
2409 err = extent_map_init();
2411 goto free_extent_io;
2413 err = ordered_data_init();
2415 goto free_extent_map;
2417 err = btrfs_delayed_inode_init();
2419 goto free_ordered_data;
2421 err = btrfs_auto_defrag_init();
2423 goto free_delayed_inode;
2425 err = btrfs_delayed_ref_init();
2427 goto free_auto_defrag;
2429 err = btrfs_prelim_ref_init();
2431 goto free_delayed_ref;
2433 err = btrfs_end_io_wq_init();
2435 goto free_prelim_ref;
2437 err = btrfs_interface_init();
2439 goto free_end_io_wq;
2441 btrfs_init_lockdep();
2443 btrfs_print_mod_info();
2445 err = btrfs_run_sanity_tests();
2447 goto unregister_ioctl;
2449 err = register_filesystem(&btrfs_fs_type);
2451 goto unregister_ioctl;
2456 btrfs_interface_exit();
2458 btrfs_end_io_wq_exit();
2460 btrfs_prelim_ref_exit();
2462 btrfs_delayed_ref_exit();
2464 btrfs_auto_defrag_exit();
2466 btrfs_delayed_inode_exit();
2468 ordered_data_exit();
2474 btrfs_destroy_cachep();
2476 btrfs_exit_compress();
2482 static void __exit exit_btrfs_fs(void)
2484 btrfs_destroy_cachep();
2485 btrfs_delayed_ref_exit();
2486 btrfs_auto_defrag_exit();
2487 btrfs_delayed_inode_exit();
2488 btrfs_prelim_ref_exit();
2489 ordered_data_exit();
2492 btrfs_interface_exit();
2493 btrfs_end_io_wq_exit();
2494 unregister_filesystem(&btrfs_fs_type);
2496 btrfs_cleanup_fs_uuids();
2497 btrfs_exit_compress();
2500 late_initcall(init_btrfs_fs);
2501 module_exit(exit_btrfs_fs)
2503 MODULE_LICENSE("GPL");