2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/blkdev.h>
20 #include <linux/module.h>
21 #include <linux/buffer_head.h>
23 #include <linux/pagemap.h>
24 #include <linux/highmem.h>
25 #include <linux/time.h>
26 #include <linux/init.h>
27 #include <linux/seq_file.h>
28 #include <linux/string.h>
29 #include <linux/backing-dev.h>
30 #include <linux/mount.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/parser.h>
37 #include <linux/ctype.h>
38 #include <linux/namei.h>
39 #include <linux/miscdevice.h>
40 #include <linux/magic.h>
41 #include <linux/slab.h>
42 #include <linux/cleancache.h>
43 #include <linux/ratelimit.h>
44 #include <linux/btrfs.h>
45 #include "delayed-inode.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "print-tree.h"
56 #include "compression.h"
57 #include "rcu-string.h"
58 #include "dev-replace.h"
59 #include "free-space-cache.h"
61 #include "tests/btrfs-tests.h"
64 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h>
67 static const struct super_operations btrfs_super_ops;
68 static struct file_system_type btrfs_fs_type;
70 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
72 const char *btrfs_decode_error(int errno)
74 char *errstr = "unknown";
78 errstr = "IO failure";
81 errstr = "Out of memory";
84 errstr = "Readonly filesystem";
87 errstr = "Object already exists";
90 errstr = "No space left";
93 errstr = "No such entry";
100 /* btrfs handle error by forcing the filesystem readonly */
101 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
103 struct super_block *sb = fs_info->sb;
105 if (sb->s_flags & MS_RDONLY)
108 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
109 sb->s_flags |= MS_RDONLY;
110 btrfs_info(fs_info, "forced readonly");
112 * Note that a running device replace operation is not
113 * canceled here although there is no way to update
114 * the progress. It would add the risk of a deadlock,
115 * therefore the canceling is omitted. The only penalty
116 * is that some I/O remains active until the procedure
117 * completes. The next time when the filesystem is
118 * mounted writeable again, the device replace
119 * operation continues.
125 * __btrfs_handle_fs_error decodes expected errors from the caller and
126 * invokes the approciate error response.
129 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
130 unsigned int line, int errno, const char *fmt, ...)
132 struct super_block *sb = fs_info->sb;
138 * Special case: if the error is EROFS, and we're already
139 * under MS_RDONLY, then it is safe here.
141 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
145 errstr = btrfs_decode_error(errno);
147 struct va_format vaf;
154 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
155 sb->s_id, function, line, errno, errstr, &vaf);
158 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
159 sb->s_id, function, line, errno, errstr);
164 * Today we only save the error info to memory. Long term we'll
165 * also send it down to the disk
167 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
169 /* Don't go through full error handling during mount */
170 if (sb->s_flags & MS_BORN)
171 btrfs_handle_error(fs_info);
175 static const char * const logtypes[] = {
188 * Use one ratelimit state per log level so that a flood of less important
189 * messages doesn't cause more important ones to be dropped.
191 static struct ratelimit_state printk_limits[] = {
192 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
193 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
194 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
195 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
196 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
197 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
198 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
199 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
202 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
204 struct super_block *sb = fs_info->sb;
206 struct va_format vaf;
208 const char *type = logtypes[4];
210 struct ratelimit_state *ratelimit;
214 kern_level = printk_get_level(fmt);
216 size_t size = printk_skip_level(fmt) - fmt;
217 memcpy(lvl, fmt, size);
220 type = logtypes[kern_level - '0'];
221 ratelimit = &printk_limits[kern_level - '0'];
224 /* Default to debug output */
225 ratelimit = &printk_limits[7];
231 if (__ratelimit(ratelimit))
232 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
239 * We only mark the transaction aborted and then set the file system read-only.
240 * This will prevent new transactions from starting or trying to join this
243 * This means that error recovery at the call site is limited to freeing
244 * any local memory allocations and passing the error code up without
245 * further cleanup. The transaction should complete as it normally would
246 * in the call path but will return -EIO.
248 * We'll complete the cleanup in btrfs_end_transaction and
249 * btrfs_commit_transaction.
252 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
253 const char *function,
254 unsigned int line, int errno)
256 struct btrfs_fs_info *fs_info = trans->fs_info;
258 trans->aborted = errno;
259 /* Nothing used. The other threads that have joined this
260 * transaction may be able to continue. */
261 if (!trans->dirty && list_empty(&trans->new_bgs)) {
264 errstr = btrfs_decode_error(errno);
266 "%s:%d: Aborting unused transaction(%s).",
267 function, line, errstr);
270 ACCESS_ONCE(trans->transaction->aborted) = errno;
271 /* Wake up anybody who may be waiting on this transaction */
272 wake_up(&fs_info->transaction_wait);
273 wake_up(&fs_info->transaction_blocked_wait);
274 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
277 * __btrfs_panic decodes unexpected, fatal errors from the caller,
278 * issues an alert, and either panics or BUGs, depending on mount options.
281 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
282 unsigned int line, int errno, const char *fmt, ...)
284 char *s_id = "<unknown>";
286 struct va_format vaf = { .fmt = fmt };
290 s_id = fs_info->sb->s_id;
295 errstr = btrfs_decode_error(errno);
296 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
297 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
298 s_id, function, line, &vaf, errno, errstr);
300 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
301 function, line, &vaf, errno, errstr);
303 /* Caller calls BUG() */
306 static void btrfs_put_super(struct super_block *sb)
308 close_ctree(btrfs_sb(sb));
312 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
313 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
314 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
315 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
316 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
317 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
318 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
319 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
320 Opt_skip_balance, Opt_check_integrity,
321 Opt_check_integrity_including_extent_data,
322 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
323 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
324 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
325 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
326 Opt_nologreplay, Opt_norecovery,
327 #ifdef CONFIG_BTRFS_DEBUG
328 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
333 static const match_table_t tokens = {
334 {Opt_degraded, "degraded"},
335 {Opt_subvol, "subvol=%s"},
336 {Opt_subvolid, "subvolid=%s"},
337 {Opt_device, "device=%s"},
338 {Opt_nodatasum, "nodatasum"},
339 {Opt_datasum, "datasum"},
340 {Opt_nodatacow, "nodatacow"},
341 {Opt_datacow, "datacow"},
342 {Opt_nobarrier, "nobarrier"},
343 {Opt_barrier, "barrier"},
344 {Opt_max_inline, "max_inline=%s"},
345 {Opt_alloc_start, "alloc_start=%s"},
346 {Opt_thread_pool, "thread_pool=%d"},
347 {Opt_compress, "compress"},
348 {Opt_compress_type, "compress=%s"},
349 {Opt_compress_force, "compress-force"},
350 {Opt_compress_force_type, "compress-force=%s"},
352 {Opt_ssd_spread, "ssd_spread"},
353 {Opt_nossd, "nossd"},
355 {Opt_noacl, "noacl"},
356 {Opt_notreelog, "notreelog"},
357 {Opt_treelog, "treelog"},
358 {Opt_nologreplay, "nologreplay"},
359 {Opt_norecovery, "norecovery"},
360 {Opt_flushoncommit, "flushoncommit"},
361 {Opt_noflushoncommit, "noflushoncommit"},
362 {Opt_ratio, "metadata_ratio=%d"},
363 {Opt_discard, "discard"},
364 {Opt_nodiscard, "nodiscard"},
365 {Opt_space_cache, "space_cache"},
366 {Opt_space_cache_version, "space_cache=%s"},
367 {Opt_clear_cache, "clear_cache"},
368 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
369 {Opt_enospc_debug, "enospc_debug"},
370 {Opt_noenospc_debug, "noenospc_debug"},
371 {Opt_subvolrootid, "subvolrootid=%d"},
372 {Opt_defrag, "autodefrag"},
373 {Opt_nodefrag, "noautodefrag"},
374 {Opt_inode_cache, "inode_cache"},
375 {Opt_noinode_cache, "noinode_cache"},
376 {Opt_no_space_cache, "nospace_cache"},
377 {Opt_recovery, "recovery"}, /* deprecated */
378 {Opt_usebackuproot, "usebackuproot"},
379 {Opt_skip_balance, "skip_balance"},
380 {Opt_check_integrity, "check_int"},
381 {Opt_check_integrity_including_extent_data, "check_int_data"},
382 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
383 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
384 {Opt_fatal_errors, "fatal_errors=%s"},
385 {Opt_commit_interval, "commit=%d"},
386 #ifdef CONFIG_BTRFS_DEBUG
387 {Opt_fragment_data, "fragment=data"},
388 {Opt_fragment_metadata, "fragment=metadata"},
389 {Opt_fragment_all, "fragment=all"},
395 * Regular mount options parser. Everything that is needed only when
396 * reading in a new superblock is parsed here.
397 * XXX JDM: This needs to be cleaned up for remount.
399 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
400 unsigned long new_flags)
402 substring_t args[MAX_OPT_ARGS];
403 char *p, *num, *orig = NULL;
408 bool compress_force = false;
409 enum btrfs_compression_type saved_compress_type;
410 bool saved_compress_force;
413 cache_gen = btrfs_super_cache_generation(info->super_copy);
414 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
415 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
417 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
420 * Even the options are empty, we still need to do extra check
427 * strsep changes the string, duplicate it because parse_options
430 options = kstrdup(options, GFP_NOFS);
436 while ((p = strsep(&options, ",")) != NULL) {
441 token = match_token(p, tokens, args);
444 btrfs_info(info, "allowing degraded mounts");
445 btrfs_set_opt(info->mount_opt, DEGRADED);
449 case Opt_subvolrootid:
452 * These are parsed by btrfs_parse_early_options
453 * and can be happily ignored here.
457 btrfs_set_and_info(info, NODATASUM,
458 "setting nodatasum");
461 if (btrfs_test_opt(info, NODATASUM)) {
462 if (btrfs_test_opt(info, NODATACOW))
464 "setting datasum, datacow enabled");
466 btrfs_info(info, "setting datasum");
468 btrfs_clear_opt(info->mount_opt, NODATACOW);
469 btrfs_clear_opt(info->mount_opt, NODATASUM);
472 if (!btrfs_test_opt(info, NODATACOW)) {
473 if (!btrfs_test_opt(info, COMPRESS) ||
474 !btrfs_test_opt(info, FORCE_COMPRESS)) {
476 "setting nodatacow, compression disabled");
478 btrfs_info(info, "setting nodatacow");
481 btrfs_clear_opt(info->mount_opt, COMPRESS);
482 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
483 btrfs_set_opt(info->mount_opt, NODATACOW);
484 btrfs_set_opt(info->mount_opt, NODATASUM);
487 btrfs_clear_and_info(info, NODATACOW,
490 case Opt_compress_force:
491 case Opt_compress_force_type:
492 compress_force = true;
495 case Opt_compress_type:
496 saved_compress_type = btrfs_test_opt(info,
498 info->compress_type : BTRFS_COMPRESS_NONE;
499 saved_compress_force =
500 btrfs_test_opt(info, FORCE_COMPRESS);
501 if (token == Opt_compress ||
502 token == Opt_compress_force ||
503 strcmp(args[0].from, "zlib") == 0) {
504 compress_type = "zlib";
505 info->compress_type = BTRFS_COMPRESS_ZLIB;
506 btrfs_set_opt(info->mount_opt, COMPRESS);
507 btrfs_clear_opt(info->mount_opt, NODATACOW);
508 btrfs_clear_opt(info->mount_opt, NODATASUM);
510 } else if (strcmp(args[0].from, "lzo") == 0) {
511 compress_type = "lzo";
512 info->compress_type = BTRFS_COMPRESS_LZO;
513 btrfs_set_opt(info->mount_opt, COMPRESS);
514 btrfs_clear_opt(info->mount_opt, NODATACOW);
515 btrfs_clear_opt(info->mount_opt, NODATASUM);
516 btrfs_set_fs_incompat(info, COMPRESS_LZO);
518 } else if (strncmp(args[0].from, "no", 2) == 0) {
519 compress_type = "no";
520 btrfs_clear_opt(info->mount_opt, COMPRESS);
521 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
522 compress_force = false;
529 if (compress_force) {
530 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
533 * If we remount from compress-force=xxx to
534 * compress=xxx, we need clear FORCE_COMPRESS
535 * flag, otherwise, there is no way for users
536 * to disable forcible compression separately.
538 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
540 if ((btrfs_test_opt(info, COMPRESS) &&
541 (info->compress_type != saved_compress_type ||
542 compress_force != saved_compress_force)) ||
543 (!btrfs_test_opt(info, COMPRESS) &&
545 btrfs_info(info, "%s %s compression",
546 (compress_force) ? "force" : "use",
549 compress_force = false;
552 btrfs_set_and_info(info, SSD,
553 "use ssd allocation scheme");
556 btrfs_set_and_info(info, SSD_SPREAD,
557 "use spread ssd allocation scheme");
558 btrfs_set_opt(info->mount_opt, SSD);
561 btrfs_set_and_info(info, NOSSD,
562 "not using ssd allocation scheme");
563 btrfs_clear_opt(info->mount_opt, SSD);
566 btrfs_clear_and_info(info, NOBARRIER,
567 "turning on barriers");
570 btrfs_set_and_info(info, NOBARRIER,
571 "turning off barriers");
573 case Opt_thread_pool:
574 ret = match_int(&args[0], &intarg);
577 } else if (intarg > 0) {
578 info->thread_pool_size = intarg;
585 num = match_strdup(&args[0]);
587 info->max_inline = memparse(num, NULL);
590 if (info->max_inline) {
591 info->max_inline = min_t(u64,
595 btrfs_info(info, "max_inline at %llu",
602 case Opt_alloc_start:
603 num = match_strdup(&args[0]);
605 mutex_lock(&info->chunk_mutex);
606 info->alloc_start = memparse(num, NULL);
607 mutex_unlock(&info->chunk_mutex);
609 btrfs_info(info, "allocations start at %llu",
617 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
618 info->sb->s_flags |= MS_POSIXACL;
621 btrfs_err(info, "support for ACL not compiled in!");
626 info->sb->s_flags &= ~MS_POSIXACL;
629 btrfs_set_and_info(info, NOTREELOG,
630 "disabling tree log");
633 btrfs_clear_and_info(info, NOTREELOG,
634 "enabling tree log");
637 case Opt_nologreplay:
638 btrfs_set_and_info(info, NOLOGREPLAY,
639 "disabling log replay at mount time");
641 case Opt_flushoncommit:
642 btrfs_set_and_info(info, FLUSHONCOMMIT,
643 "turning on flush-on-commit");
645 case Opt_noflushoncommit:
646 btrfs_clear_and_info(info, FLUSHONCOMMIT,
647 "turning off flush-on-commit");
650 ret = match_int(&args[0], &intarg);
653 } else if (intarg >= 0) {
654 info->metadata_ratio = intarg;
655 btrfs_info(info, "metadata ratio %d",
656 info->metadata_ratio);
663 btrfs_set_and_info(info, DISCARD,
664 "turning on discard");
667 btrfs_clear_and_info(info, DISCARD,
668 "turning off discard");
670 case Opt_space_cache:
671 case Opt_space_cache_version:
672 if (token == Opt_space_cache ||
673 strcmp(args[0].from, "v1") == 0) {
674 btrfs_clear_opt(info->mount_opt,
676 btrfs_set_and_info(info, SPACE_CACHE,
677 "enabling disk space caching");
678 } else if (strcmp(args[0].from, "v2") == 0) {
679 btrfs_clear_opt(info->mount_opt,
681 btrfs_set_and_info(info, FREE_SPACE_TREE,
682 "enabling free space tree");
688 case Opt_rescan_uuid_tree:
689 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
691 case Opt_no_space_cache:
692 if (btrfs_test_opt(info, SPACE_CACHE)) {
693 btrfs_clear_and_info(info, SPACE_CACHE,
694 "disabling disk space caching");
696 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
697 btrfs_clear_and_info(info, FREE_SPACE_TREE,
698 "disabling free space tree");
701 case Opt_inode_cache:
702 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
703 "enabling inode map caching");
705 case Opt_noinode_cache:
706 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
707 "disabling inode map caching");
709 case Opt_clear_cache:
710 btrfs_set_and_info(info, CLEAR_CACHE,
711 "force clearing of disk cache");
713 case Opt_user_subvol_rm_allowed:
714 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
716 case Opt_enospc_debug:
717 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
719 case Opt_noenospc_debug:
720 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
723 btrfs_set_and_info(info, AUTO_DEFRAG,
724 "enabling auto defrag");
727 btrfs_clear_and_info(info, AUTO_DEFRAG,
728 "disabling auto defrag");
732 "'recovery' is deprecated, use 'usebackuproot' instead");
733 case Opt_usebackuproot:
735 "trying to use backup root at mount time");
736 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
738 case Opt_skip_balance:
739 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
741 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
742 case Opt_check_integrity_including_extent_data:
744 "enabling check integrity including extent data");
745 btrfs_set_opt(info->mount_opt,
746 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
747 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
749 case Opt_check_integrity:
750 btrfs_info(info, "enabling check integrity");
751 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
753 case Opt_check_integrity_print_mask:
754 ret = match_int(&args[0], &intarg);
757 } else if (intarg >= 0) {
758 info->check_integrity_print_mask = intarg;
760 "check_integrity_print_mask 0x%x",
761 info->check_integrity_print_mask);
768 case Opt_check_integrity_including_extent_data:
769 case Opt_check_integrity:
770 case Opt_check_integrity_print_mask:
772 "support for check_integrity* not compiled in!");
776 case Opt_fatal_errors:
777 if (strcmp(args[0].from, "panic") == 0)
778 btrfs_set_opt(info->mount_opt,
779 PANIC_ON_FATAL_ERROR);
780 else if (strcmp(args[0].from, "bug") == 0)
781 btrfs_clear_opt(info->mount_opt,
782 PANIC_ON_FATAL_ERROR);
788 case Opt_commit_interval:
790 ret = match_int(&args[0], &intarg);
792 btrfs_err(info, "invalid commit interval");
799 "excessive commit interval %d",
802 info->commit_interval = intarg;
805 "using default commit interval %ds",
806 BTRFS_DEFAULT_COMMIT_INTERVAL);
807 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
810 #ifdef CONFIG_BTRFS_DEBUG
811 case Opt_fragment_all:
812 btrfs_info(info, "fragmenting all space");
813 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
814 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
816 case Opt_fragment_metadata:
817 btrfs_info(info, "fragmenting metadata");
818 btrfs_set_opt(info->mount_opt,
821 case Opt_fragment_data:
822 btrfs_info(info, "fragmenting data");
823 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
827 btrfs_info(info, "unrecognized mount option '%s'", p);
836 * Extra check for current option against current flag
838 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
840 "nologreplay must be used with ro mount option");
844 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
845 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
846 !btrfs_test_opt(info, CLEAR_CACHE)) {
847 btrfs_err(info, "cannot disable free space tree");
851 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
852 btrfs_info(info, "disk space caching is enabled");
853 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
854 btrfs_info(info, "using free space tree");
860 * Parse mount options that are required early in the mount process.
862 * All other options will be parsed on much later in the mount process and
863 * only when we need to allocate a new super block.
865 static int btrfs_parse_early_options(const char *options, fmode_t flags,
866 void *holder, char **subvol_name, u64 *subvol_objectid,
867 struct btrfs_fs_devices **fs_devices)
869 substring_t args[MAX_OPT_ARGS];
870 char *device_name, *opts, *orig, *p;
878 * strsep changes the string, duplicate it because parse_options
881 opts = kstrdup(options, GFP_KERNEL);
886 while ((p = strsep(&opts, ",")) != NULL) {
891 token = match_token(p, tokens, args);
895 *subvol_name = match_strdup(&args[0]);
902 num = match_strdup(&args[0]);
904 *subvol_objectid = memparse(num, NULL);
906 /* we want the original fs_tree */
907 if (!*subvol_objectid)
909 BTRFS_FS_TREE_OBJECTID;
915 case Opt_subvolrootid:
916 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
919 device_name = match_strdup(&args[0]);
924 error = btrfs_scan_one_device(device_name,
925 flags, holder, fs_devices);
940 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
943 struct btrfs_root *root = fs_info->tree_root;
944 struct btrfs_root *fs_root;
945 struct btrfs_root_ref *root_ref;
946 struct btrfs_inode_ref *inode_ref;
947 struct btrfs_key key;
948 struct btrfs_path *path = NULL;
949 char *name = NULL, *ptr;
954 path = btrfs_alloc_path();
959 path->leave_spinning = 1;
961 name = kmalloc(PATH_MAX, GFP_NOFS);
966 ptr = name + PATH_MAX - 1;
970 * Walk up the subvolume trees in the tree of tree roots by root
971 * backrefs until we hit the top-level subvolume.
973 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
974 key.objectid = subvol_objectid;
975 key.type = BTRFS_ROOT_BACKREF_KEY;
976 key.offset = (u64)-1;
978 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
981 } else if (ret > 0) {
982 ret = btrfs_previous_item(root, path, subvol_objectid,
983 BTRFS_ROOT_BACKREF_KEY);
986 } else if (ret > 0) {
992 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
993 subvol_objectid = key.offset;
995 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
996 struct btrfs_root_ref);
997 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1000 ret = -ENAMETOOLONG;
1003 read_extent_buffer(path->nodes[0], ptr + 1,
1004 (unsigned long)(root_ref + 1), len);
1006 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1007 btrfs_release_path(path);
1009 key.objectid = subvol_objectid;
1010 key.type = BTRFS_ROOT_ITEM_KEY;
1011 key.offset = (u64)-1;
1012 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1013 if (IS_ERR(fs_root)) {
1014 ret = PTR_ERR(fs_root);
1019 * Walk up the filesystem tree by inode refs until we hit the
1022 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1023 key.objectid = dirid;
1024 key.type = BTRFS_INODE_REF_KEY;
1025 key.offset = (u64)-1;
1027 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1030 } else if (ret > 0) {
1031 ret = btrfs_previous_item(fs_root, path, dirid,
1032 BTRFS_INODE_REF_KEY);
1035 } else if (ret > 0) {
1041 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1044 inode_ref = btrfs_item_ptr(path->nodes[0],
1046 struct btrfs_inode_ref);
1047 len = btrfs_inode_ref_name_len(path->nodes[0],
1051 ret = -ENAMETOOLONG;
1054 read_extent_buffer(path->nodes[0], ptr + 1,
1055 (unsigned long)(inode_ref + 1), len);
1057 btrfs_release_path(path);
1061 btrfs_free_path(path);
1062 if (ptr == name + PATH_MAX - 1) {
1066 memmove(name, ptr, name + PATH_MAX - ptr);
1071 btrfs_free_path(path);
1073 return ERR_PTR(ret);
1076 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1078 struct btrfs_root *root = fs_info->tree_root;
1079 struct btrfs_dir_item *di;
1080 struct btrfs_path *path;
1081 struct btrfs_key location;
1084 path = btrfs_alloc_path();
1087 path->leave_spinning = 1;
1090 * Find the "default" dir item which points to the root item that we
1091 * will mount by default if we haven't been given a specific subvolume
1094 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1095 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1097 btrfs_free_path(path);
1102 * Ok the default dir item isn't there. This is weird since
1103 * it's always been there, but don't freak out, just try and
1104 * mount the top-level subvolume.
1106 btrfs_free_path(path);
1107 *objectid = BTRFS_FS_TREE_OBJECTID;
1111 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1112 btrfs_free_path(path);
1113 *objectid = location.objectid;
1117 static int btrfs_fill_super(struct super_block *sb,
1118 struct btrfs_fs_devices *fs_devices,
1119 void *data, int silent)
1121 struct inode *inode;
1122 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1123 struct btrfs_key key;
1126 sb->s_maxbytes = MAX_LFS_FILESIZE;
1127 sb->s_magic = BTRFS_SUPER_MAGIC;
1128 sb->s_op = &btrfs_super_ops;
1129 sb->s_d_op = &btrfs_dentry_operations;
1130 sb->s_export_op = &btrfs_export_ops;
1131 sb->s_xattr = btrfs_xattr_handlers;
1132 sb->s_time_gran = 1;
1133 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1134 sb->s_flags |= MS_POSIXACL;
1136 sb->s_flags |= MS_I_VERSION;
1137 sb->s_iflags |= SB_I_CGROUPWB;
1138 err = open_ctree(sb, fs_devices, (char *)data);
1140 btrfs_err(fs_info, "open_ctree failed");
1144 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1145 key.type = BTRFS_INODE_ITEM_KEY;
1147 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1148 if (IS_ERR(inode)) {
1149 err = PTR_ERR(inode);
1153 sb->s_root = d_make_root(inode);
1159 save_mount_options(sb, data);
1160 cleancache_init_fs(sb);
1161 sb->s_flags |= MS_ACTIVE;
1165 close_ctree(fs_info);
1169 int btrfs_sync_fs(struct super_block *sb, int wait)
1171 struct btrfs_trans_handle *trans;
1172 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1173 struct btrfs_root *root = fs_info->tree_root;
1175 trace_btrfs_sync_fs(fs_info, wait);
1178 filemap_flush(fs_info->btree_inode->i_mapping);
1182 btrfs_wait_ordered_roots(fs_info, -1, 0, (u64)-1);
1184 trans = btrfs_attach_transaction_barrier(root);
1185 if (IS_ERR(trans)) {
1186 /* no transaction, don't bother */
1187 if (PTR_ERR(trans) == -ENOENT) {
1189 * Exit unless we have some pending changes
1190 * that need to go through commit
1192 if (fs_info->pending_changes == 0)
1195 * A non-blocking test if the fs is frozen. We must not
1196 * start a new transaction here otherwise a deadlock
1197 * happens. The pending operations are delayed to the
1198 * next commit after thawing.
1200 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1201 __sb_end_write(sb, SB_FREEZE_WRITE);
1204 trans = btrfs_start_transaction(root, 0);
1207 return PTR_ERR(trans);
1209 return btrfs_commit_transaction(trans);
1212 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1214 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1215 char *compress_type;
1217 if (btrfs_test_opt(info, DEGRADED))
1218 seq_puts(seq, ",degraded");
1219 if (btrfs_test_opt(info, NODATASUM))
1220 seq_puts(seq, ",nodatasum");
1221 if (btrfs_test_opt(info, NODATACOW))
1222 seq_puts(seq, ",nodatacow");
1223 if (btrfs_test_opt(info, NOBARRIER))
1224 seq_puts(seq, ",nobarrier");
1225 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1226 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1227 if (info->alloc_start != 0)
1228 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1229 if (info->thread_pool_size != min_t(unsigned long,
1230 num_online_cpus() + 2, 8))
1231 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1232 if (btrfs_test_opt(info, COMPRESS)) {
1233 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1234 compress_type = "zlib";
1236 compress_type = "lzo";
1237 if (btrfs_test_opt(info, FORCE_COMPRESS))
1238 seq_printf(seq, ",compress-force=%s", compress_type);
1240 seq_printf(seq, ",compress=%s", compress_type);
1242 if (btrfs_test_opt(info, NOSSD))
1243 seq_puts(seq, ",nossd");
1244 if (btrfs_test_opt(info, SSD_SPREAD))
1245 seq_puts(seq, ",ssd_spread");
1246 else if (btrfs_test_opt(info, SSD))
1247 seq_puts(seq, ",ssd");
1248 if (btrfs_test_opt(info, NOTREELOG))
1249 seq_puts(seq, ",notreelog");
1250 if (btrfs_test_opt(info, NOLOGREPLAY))
1251 seq_puts(seq, ",nologreplay");
1252 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1253 seq_puts(seq, ",flushoncommit");
1254 if (btrfs_test_opt(info, DISCARD))
1255 seq_puts(seq, ",discard");
1256 if (!(info->sb->s_flags & MS_POSIXACL))
1257 seq_puts(seq, ",noacl");
1258 if (btrfs_test_opt(info, SPACE_CACHE))
1259 seq_puts(seq, ",space_cache");
1260 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1261 seq_puts(seq, ",space_cache=v2");
1263 seq_puts(seq, ",nospace_cache");
1264 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1265 seq_puts(seq, ",rescan_uuid_tree");
1266 if (btrfs_test_opt(info, CLEAR_CACHE))
1267 seq_puts(seq, ",clear_cache");
1268 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1269 seq_puts(seq, ",user_subvol_rm_allowed");
1270 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1271 seq_puts(seq, ",enospc_debug");
1272 if (btrfs_test_opt(info, AUTO_DEFRAG))
1273 seq_puts(seq, ",autodefrag");
1274 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1275 seq_puts(seq, ",inode_cache");
1276 if (btrfs_test_opt(info, SKIP_BALANCE))
1277 seq_puts(seq, ",skip_balance");
1278 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1279 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1280 seq_puts(seq, ",check_int_data");
1281 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1282 seq_puts(seq, ",check_int");
1283 if (info->check_integrity_print_mask)
1284 seq_printf(seq, ",check_int_print_mask=%d",
1285 info->check_integrity_print_mask);
1287 if (info->metadata_ratio)
1288 seq_printf(seq, ",metadata_ratio=%d",
1289 info->metadata_ratio);
1290 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1291 seq_puts(seq, ",fatal_errors=panic");
1292 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1293 seq_printf(seq, ",commit=%d", info->commit_interval);
1294 #ifdef CONFIG_BTRFS_DEBUG
1295 if (btrfs_test_opt(info, FRAGMENT_DATA))
1296 seq_puts(seq, ",fragment=data");
1297 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1298 seq_puts(seq, ",fragment=metadata");
1300 seq_printf(seq, ",subvolid=%llu",
1301 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1302 seq_puts(seq, ",subvol=");
1303 seq_dentry(seq, dentry, " \t\n\\");
1307 static int btrfs_test_super(struct super_block *s, void *data)
1309 struct btrfs_fs_info *p = data;
1310 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1312 return fs_info->fs_devices == p->fs_devices;
1315 static int btrfs_set_super(struct super_block *s, void *data)
1317 int err = set_anon_super(s, data);
1319 s->s_fs_info = data;
1324 * subvolumes are identified by ino 256
1326 static inline int is_subvolume_inode(struct inode *inode)
1328 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1334 * This will add subvolid=0 to the argument string while removing any subvol=
1335 * and subvolid= arguments to make sure we get the top-level root for path
1336 * walking to the subvol we want.
1338 static char *setup_root_args(char *args)
1340 char *buf, *dst, *sep;
1343 return kstrdup("subvolid=0", GFP_NOFS);
1345 /* The worst case is that we add ",subvolid=0" to the end. */
1346 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
1351 sep = strchrnul(args, ',');
1352 if (!strstarts(args, "subvol=") &&
1353 !strstarts(args, "subvolid=")) {
1354 memcpy(dst, args, sep - args);
1363 strcpy(dst, "subvolid=0");
1368 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1369 int flags, const char *device_name,
1372 struct dentry *root;
1373 struct vfsmount *mnt = NULL;
1377 newargs = setup_root_args(data);
1379 root = ERR_PTR(-ENOMEM);
1383 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
1384 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
1385 if (flags & MS_RDONLY) {
1386 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
1387 device_name, newargs);
1389 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
1390 device_name, newargs);
1392 root = ERR_CAST(mnt);
1397 down_write(&mnt->mnt_sb->s_umount);
1398 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1399 up_write(&mnt->mnt_sb->s_umount);
1401 root = ERR_PTR(ret);
1407 root = ERR_CAST(mnt);
1413 if (!subvol_objectid) {
1414 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1417 root = ERR_PTR(ret);
1421 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1423 if (IS_ERR(subvol_name)) {
1424 root = ERR_CAST(subvol_name);
1431 root = mount_subtree(mnt, subvol_name);
1432 /* mount_subtree() drops our reference on the vfsmount. */
1435 if (!IS_ERR(root)) {
1436 struct super_block *s = root->d_sb;
1437 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1438 struct inode *root_inode = d_inode(root);
1439 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1442 if (!is_subvolume_inode(root_inode)) {
1443 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1447 if (subvol_objectid && root_objectid != subvol_objectid) {
1449 * This will also catch a race condition where a
1450 * subvolume which was passed by ID is renamed and
1451 * another subvolume is renamed over the old location.
1454 "subvol '%s' does not match subvolid %llu",
1455 subvol_name, subvol_objectid);
1460 root = ERR_PTR(ret);
1461 deactivate_locked_super(s);
1472 static int parse_security_options(char *orig_opts,
1473 struct security_mnt_opts *sec_opts)
1475 char *secdata = NULL;
1478 secdata = alloc_secdata();
1481 ret = security_sb_copy_data(orig_opts, secdata);
1483 free_secdata(secdata);
1486 ret = security_sb_parse_opts_str(secdata, sec_opts);
1487 free_secdata(secdata);
1491 static int setup_security_options(struct btrfs_fs_info *fs_info,
1492 struct super_block *sb,
1493 struct security_mnt_opts *sec_opts)
1498 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1501 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1505 #ifdef CONFIG_SECURITY
1506 if (!fs_info->security_opts.num_mnt_opts) {
1507 /* first time security setup, copy sec_opts to fs_info */
1508 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1511 * Since SELinux (the only one supporting security_mnt_opts)
1512 * does NOT support changing context during remount/mount of
1513 * the same sb, this must be the same or part of the same
1514 * security options, just free it.
1516 security_free_mnt_opts(sec_opts);
1523 * Find a superblock for the given device / mount point.
1525 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1526 * for multiple device setup. Make sure to keep it in sync.
1528 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1529 const char *device_name, void *data)
1531 struct block_device *bdev = NULL;
1532 struct super_block *s;
1533 struct btrfs_fs_devices *fs_devices = NULL;
1534 struct btrfs_fs_info *fs_info = NULL;
1535 struct security_mnt_opts new_sec_opts;
1536 fmode_t mode = FMODE_READ;
1537 char *subvol_name = NULL;
1538 u64 subvol_objectid = 0;
1541 if (!(flags & MS_RDONLY))
1542 mode |= FMODE_WRITE;
1544 error = btrfs_parse_early_options(data, mode, fs_type,
1545 &subvol_name, &subvol_objectid,
1549 return ERR_PTR(error);
1552 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1553 /* mount_subvol() will free subvol_name. */
1554 return mount_subvol(subvol_name, subvol_objectid, flags,
1558 security_init_mnt_opts(&new_sec_opts);
1560 error = parse_security_options(data, &new_sec_opts);
1562 return ERR_PTR(error);
1565 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1567 goto error_sec_opts;
1570 * Setup a dummy root and fs_info for test/set super. This is because
1571 * we don't actually fill this stuff out until open_ctree, but we need
1572 * it for searching for existing supers, so this lets us do that and
1573 * then open_ctree will properly initialize everything later.
1575 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1578 goto error_sec_opts;
1581 fs_info->fs_devices = fs_devices;
1583 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1584 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1585 security_init_mnt_opts(&fs_info->security_opts);
1586 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1591 error = btrfs_open_devices(fs_devices, mode, fs_type);
1595 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1597 goto error_close_devices;
1600 bdev = fs_devices->latest_bdev;
1601 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1605 goto error_close_devices;
1609 btrfs_close_devices(fs_devices);
1610 free_fs_info(fs_info);
1611 if ((flags ^ s->s_flags) & MS_RDONLY)
1614 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1615 btrfs_sb(s)->bdev_holder = fs_type;
1616 error = btrfs_fill_super(s, fs_devices, data,
1617 flags & MS_SILENT ? 1 : 0);
1620 deactivate_locked_super(s);
1621 goto error_sec_opts;
1624 fs_info = btrfs_sb(s);
1625 error = setup_security_options(fs_info, s, &new_sec_opts);
1627 deactivate_locked_super(s);
1628 goto error_sec_opts;
1631 return dget(s->s_root);
1633 error_close_devices:
1634 btrfs_close_devices(fs_devices);
1636 free_fs_info(fs_info);
1638 security_free_mnt_opts(&new_sec_opts);
1639 return ERR_PTR(error);
1642 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1643 int new_pool_size, int old_pool_size)
1645 if (new_pool_size == old_pool_size)
1648 fs_info->thread_pool_size = new_pool_size;
1650 btrfs_info(fs_info, "resize thread pool %d -> %d",
1651 old_pool_size, new_pool_size);
1653 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1654 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1655 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1656 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1657 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1658 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1659 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1661 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1662 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1663 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1664 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1665 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1669 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1671 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1674 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1675 unsigned long old_opts, int flags)
1677 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1678 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1679 (flags & MS_RDONLY))) {
1680 /* wait for any defraggers to finish */
1681 wait_event(fs_info->transaction_wait,
1682 (atomic_read(&fs_info->defrag_running) == 0));
1683 if (flags & MS_RDONLY)
1684 sync_filesystem(fs_info->sb);
1688 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1689 unsigned long old_opts)
1692 * We need to cleanup all defragable inodes if the autodefragment is
1693 * close or the filesystem is read only.
1695 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1696 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1697 (fs_info->sb->s_flags & MS_RDONLY))) {
1698 btrfs_cleanup_defrag_inodes(fs_info);
1701 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1704 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1706 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1707 struct btrfs_root *root = fs_info->tree_root;
1708 unsigned old_flags = sb->s_flags;
1709 unsigned long old_opts = fs_info->mount_opt;
1710 unsigned long old_compress_type = fs_info->compress_type;
1711 u64 old_max_inline = fs_info->max_inline;
1712 u64 old_alloc_start = fs_info->alloc_start;
1713 int old_thread_pool_size = fs_info->thread_pool_size;
1714 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1717 sync_filesystem(sb);
1718 btrfs_remount_prepare(fs_info);
1721 struct security_mnt_opts new_sec_opts;
1723 security_init_mnt_opts(&new_sec_opts);
1724 ret = parse_security_options(data, &new_sec_opts);
1727 ret = setup_security_options(fs_info, sb,
1730 security_free_mnt_opts(&new_sec_opts);
1735 ret = btrfs_parse_options(fs_info, data, *flags);
1741 btrfs_remount_begin(fs_info, old_opts, *flags);
1742 btrfs_resize_thread_pool(fs_info,
1743 fs_info->thread_pool_size, old_thread_pool_size);
1745 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1748 if (*flags & MS_RDONLY) {
1750 * this also happens on 'umount -rf' or on shutdown, when
1751 * the filesystem is busy.
1753 cancel_work_sync(&fs_info->async_reclaim_work);
1755 /* wait for the uuid_scan task to finish */
1756 down(&fs_info->uuid_tree_rescan_sem);
1757 /* avoid complains from lockdep et al. */
1758 up(&fs_info->uuid_tree_rescan_sem);
1760 sb->s_flags |= MS_RDONLY;
1763 * Setting MS_RDONLY will put the cleaner thread to
1764 * sleep at the next loop if it's already active.
1765 * If it's already asleep, we'll leave unused block
1766 * groups on disk until we're mounted read-write again
1767 * unless we clean them up here.
1769 btrfs_delete_unused_bgs(fs_info);
1771 btrfs_dev_replace_suspend_for_unmount(fs_info);
1772 btrfs_scrub_cancel(fs_info);
1773 btrfs_pause_balance(fs_info);
1775 ret = btrfs_commit_super(fs_info);
1779 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1781 "Remounting read-write after error is not allowed");
1785 if (fs_info->fs_devices->rw_devices == 0) {
1790 if (fs_info->fs_devices->missing_devices >
1791 fs_info->num_tolerated_disk_barrier_failures &&
1792 !(*flags & MS_RDONLY)) {
1794 "too many missing devices, writeable remount is not allowed");
1799 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1804 ret = btrfs_cleanup_fs_roots(fs_info);
1808 /* recover relocation */
1809 mutex_lock(&fs_info->cleaner_mutex);
1810 ret = btrfs_recover_relocation(root);
1811 mutex_unlock(&fs_info->cleaner_mutex);
1815 ret = btrfs_resume_balance_async(fs_info);
1819 ret = btrfs_resume_dev_replace_async(fs_info);
1821 btrfs_warn(fs_info, "failed to resume dev_replace");
1825 if (!fs_info->uuid_root) {
1826 btrfs_info(fs_info, "creating UUID tree");
1827 ret = btrfs_create_uuid_tree(fs_info);
1830 "failed to create the UUID tree %d",
1835 sb->s_flags &= ~MS_RDONLY;
1837 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1840 wake_up_process(fs_info->transaction_kthread);
1841 btrfs_remount_cleanup(fs_info, old_opts);
1845 /* We've hit an error - don't reset MS_RDONLY */
1846 if (sb->s_flags & MS_RDONLY)
1847 old_flags |= MS_RDONLY;
1848 sb->s_flags = old_flags;
1849 fs_info->mount_opt = old_opts;
1850 fs_info->compress_type = old_compress_type;
1851 fs_info->max_inline = old_max_inline;
1852 mutex_lock(&fs_info->chunk_mutex);
1853 fs_info->alloc_start = old_alloc_start;
1854 mutex_unlock(&fs_info->chunk_mutex);
1855 btrfs_resize_thread_pool(fs_info,
1856 old_thread_pool_size, fs_info->thread_pool_size);
1857 fs_info->metadata_ratio = old_metadata_ratio;
1858 btrfs_remount_cleanup(fs_info, old_opts);
1862 /* Used to sort the devices by max_avail(descending sort) */
1863 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1864 const void *dev_info2)
1866 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1867 ((struct btrfs_device_info *)dev_info2)->max_avail)
1869 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1870 ((struct btrfs_device_info *)dev_info2)->max_avail)
1877 * sort the devices by max_avail, in which max free extent size of each device
1878 * is stored.(Descending Sort)
1880 static inline void btrfs_descending_sort_devices(
1881 struct btrfs_device_info *devices,
1884 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1885 btrfs_cmp_device_free_bytes, NULL);
1889 * The helper to calc the free space on the devices that can be used to store
1892 static int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1895 struct btrfs_root *root = fs_info->tree_root;
1896 struct btrfs_device_info *devices_info;
1897 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1898 struct btrfs_device *device;
1903 u64 min_stripe_size;
1904 int min_stripes = 1, num_stripes = 1;
1905 int i = 0, nr_devices;
1909 * We aren't under the device list lock, so this is racy-ish, but good
1910 * enough for our purposes.
1912 nr_devices = fs_info->fs_devices->open_devices;
1915 nr_devices = fs_info->fs_devices->open_devices;
1923 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1928 /* calc min stripe number for data space allocation */
1929 type = btrfs_get_alloc_profile(root, 1);
1930 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1932 num_stripes = nr_devices;
1933 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1936 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1941 if (type & BTRFS_BLOCK_GROUP_DUP)
1942 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1944 min_stripe_size = BTRFS_STRIPE_LEN;
1946 if (fs_info->alloc_start)
1947 mutex_lock(&fs_devices->device_list_mutex);
1949 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1950 if (!device->in_fs_metadata || !device->bdev ||
1951 device->is_tgtdev_for_dev_replace)
1954 if (i >= nr_devices)
1957 avail_space = device->total_bytes - device->bytes_used;
1959 /* align with stripe_len */
1960 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1961 avail_space *= BTRFS_STRIPE_LEN;
1964 * In order to avoid overwriting the superblock on the drive,
1965 * btrfs starts at an offset of at least 1MB when doing chunk
1970 /* user can set the offset in fs_info->alloc_start. */
1971 if (fs_info->alloc_start &&
1972 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1973 device->total_bytes) {
1975 skip_space = max(fs_info->alloc_start, skip_space);
1978 * btrfs can not use the free space in
1979 * [0, skip_space - 1], we must subtract it from the
1980 * total. In order to implement it, we account the used
1981 * space in this range first.
1983 ret = btrfs_account_dev_extents_size(device, 0,
1987 kfree(devices_info);
1988 mutex_unlock(&fs_devices->device_list_mutex);
1994 /* calc the free space in [0, skip_space - 1] */
1995 skip_space -= used_space;
1999 * we can use the free space in [0, skip_space - 1], subtract
2000 * it from the total.
2002 if (avail_space && avail_space >= skip_space)
2003 avail_space -= skip_space;
2007 if (avail_space < min_stripe_size)
2010 devices_info[i].dev = device;
2011 devices_info[i].max_avail = avail_space;
2016 if (fs_info->alloc_start)
2017 mutex_unlock(&fs_devices->device_list_mutex);
2021 btrfs_descending_sort_devices(devices_info, nr_devices);
2025 while (nr_devices >= min_stripes) {
2026 if (num_stripes > nr_devices)
2027 num_stripes = nr_devices;
2029 if (devices_info[i].max_avail >= min_stripe_size) {
2033 avail_space += devices_info[i].max_avail * num_stripes;
2034 alloc_size = devices_info[i].max_avail;
2035 for (j = i + 1 - num_stripes; j <= i; j++)
2036 devices_info[j].max_avail -= alloc_size;
2042 kfree(devices_info);
2043 *free_bytes = avail_space;
2048 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2050 * If there's a redundant raid level at DATA block groups, use the respective
2051 * multiplier to scale the sizes.
2053 * Unused device space usage is based on simulating the chunk allocator
2054 * algorithm that respects the device sizes, order of allocations and the
2055 * 'alloc_start' value, this is a close approximation of the actual use but
2056 * there are other factors that may change the result (like a new metadata
2059 * If metadata is exhausted, f_bavail will be 0.
2061 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2063 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2064 struct btrfs_super_block *disk_super = fs_info->super_copy;
2065 struct list_head *head = &fs_info->space_info;
2066 struct btrfs_space_info *found;
2068 u64 total_free_data = 0;
2069 u64 total_free_meta = 0;
2070 int bits = dentry->d_sb->s_blocksize_bits;
2071 __be32 *fsid = (__be32 *)fs_info->fsid;
2072 unsigned factor = 1;
2073 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2079 list_for_each_entry_rcu(found, head, list) {
2080 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2083 total_free_data += found->disk_total - found->disk_used;
2085 btrfs_account_ro_block_groups_free_space(found);
2087 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2088 if (!list_empty(&found->block_groups[i])) {
2090 case BTRFS_RAID_DUP:
2091 case BTRFS_RAID_RAID1:
2092 case BTRFS_RAID_RAID10:
2100 * Metadata in mixed block goup profiles are accounted in data
2102 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2103 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2106 total_free_meta += found->disk_total -
2110 total_used += found->disk_used;
2115 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2116 buf->f_blocks >>= bits;
2117 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2119 /* Account global block reserve as used, it's in logical size already */
2120 spin_lock(&block_rsv->lock);
2121 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2122 if (buf->f_bfree >= block_rsv->size >> bits)
2123 buf->f_bfree -= block_rsv->size >> bits;
2126 spin_unlock(&block_rsv->lock);
2128 buf->f_bavail = div_u64(total_free_data, factor);
2129 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2132 buf->f_bavail += div_u64(total_free_data, factor);
2133 buf->f_bavail = buf->f_bavail >> bits;
2136 * We calculate the remaining metadata space minus global reserve. If
2137 * this is (supposedly) smaller than zero, there's no space. But this
2138 * does not hold in practice, the exhausted state happens where's still
2139 * some positive delta. So we apply some guesswork and compare the
2140 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2142 * We probably cannot calculate the exact threshold value because this
2143 * depends on the internal reservations requested by various
2144 * operations, so some operations that consume a few metadata will
2145 * succeed even if the Avail is zero. But this is better than the other
2148 thresh = 4 * 1024 * 1024;
2150 if (!mixed && total_free_meta - thresh < block_rsv->size)
2153 buf->f_type = BTRFS_SUPER_MAGIC;
2154 buf->f_bsize = dentry->d_sb->s_blocksize;
2155 buf->f_namelen = BTRFS_NAME_LEN;
2157 /* We treat it as constant endianness (it doesn't matter _which_)
2158 because we want the fsid to come out the same whether mounted
2159 on a big-endian or little-endian host */
2160 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2161 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2162 /* Mask in the root object ID too, to disambiguate subvols */
2163 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2164 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2169 static void btrfs_kill_super(struct super_block *sb)
2171 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2172 kill_anon_super(sb);
2173 free_fs_info(fs_info);
2176 static struct file_system_type btrfs_fs_type = {
2177 .owner = THIS_MODULE,
2179 .mount = btrfs_mount,
2180 .kill_sb = btrfs_kill_super,
2181 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2183 MODULE_ALIAS_FS("btrfs");
2185 static int btrfs_control_open(struct inode *inode, struct file *file)
2188 * The control file's private_data is used to hold the
2189 * transaction when it is started and is used to keep
2190 * track of whether a transaction is already in progress.
2192 file->private_data = NULL;
2197 * used by btrfsctl to scan devices when no FS is mounted
2199 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2202 struct btrfs_ioctl_vol_args *vol;
2203 struct btrfs_fs_devices *fs_devices;
2206 if (!capable(CAP_SYS_ADMIN))
2209 vol = memdup_user((void __user *)arg, sizeof(*vol));
2211 return PTR_ERR(vol);
2214 case BTRFS_IOC_SCAN_DEV:
2215 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2216 &btrfs_fs_type, &fs_devices);
2218 case BTRFS_IOC_DEVICES_READY:
2219 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2220 &btrfs_fs_type, &fs_devices);
2223 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2225 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2226 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2234 static int btrfs_freeze(struct super_block *sb)
2236 struct btrfs_trans_handle *trans;
2237 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2238 struct btrfs_root *root = fs_info->tree_root;
2240 fs_info->fs_frozen = 1;
2242 * We don't need a barrier here, we'll wait for any transaction that
2243 * could be in progress on other threads (and do delayed iputs that
2244 * we want to avoid on a frozen filesystem), or do the commit
2247 trans = btrfs_attach_transaction_barrier(root);
2248 if (IS_ERR(trans)) {
2249 /* no transaction, don't bother */
2250 if (PTR_ERR(trans) == -ENOENT)
2252 return PTR_ERR(trans);
2254 return btrfs_commit_transaction(trans);
2257 static int btrfs_unfreeze(struct super_block *sb)
2259 btrfs_sb(sb)->fs_frozen = 0;
2263 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2265 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2266 struct btrfs_fs_devices *cur_devices;
2267 struct btrfs_device *dev, *first_dev = NULL;
2268 struct list_head *head;
2269 struct rcu_string *name;
2271 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2272 cur_devices = fs_info->fs_devices;
2273 while (cur_devices) {
2274 head = &cur_devices->devices;
2275 list_for_each_entry(dev, head, dev_list) {
2280 if (!first_dev || dev->devid < first_dev->devid)
2283 cur_devices = cur_devices->seed;
2288 name = rcu_dereference(first_dev->name);
2289 seq_escape(m, name->str, " \t\n\\");
2294 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2298 static const struct super_operations btrfs_super_ops = {
2299 .drop_inode = btrfs_drop_inode,
2300 .evict_inode = btrfs_evict_inode,
2301 .put_super = btrfs_put_super,
2302 .sync_fs = btrfs_sync_fs,
2303 .show_options = btrfs_show_options,
2304 .show_devname = btrfs_show_devname,
2305 .write_inode = btrfs_write_inode,
2306 .alloc_inode = btrfs_alloc_inode,
2307 .destroy_inode = btrfs_destroy_inode,
2308 .statfs = btrfs_statfs,
2309 .remount_fs = btrfs_remount,
2310 .freeze_fs = btrfs_freeze,
2311 .unfreeze_fs = btrfs_unfreeze,
2314 static const struct file_operations btrfs_ctl_fops = {
2315 .open = btrfs_control_open,
2316 .unlocked_ioctl = btrfs_control_ioctl,
2317 .compat_ioctl = btrfs_control_ioctl,
2318 .owner = THIS_MODULE,
2319 .llseek = noop_llseek,
2322 static struct miscdevice btrfs_misc = {
2323 .minor = BTRFS_MINOR,
2324 .name = "btrfs-control",
2325 .fops = &btrfs_ctl_fops
2328 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2329 MODULE_ALIAS("devname:btrfs-control");
2331 static int btrfs_interface_init(void)
2333 return misc_register(&btrfs_misc);
2336 static void btrfs_interface_exit(void)
2338 misc_deregister(&btrfs_misc);
2341 static void btrfs_print_mod_info(void)
2343 pr_info("Btrfs loaded, crc32c=%s"
2344 #ifdef CONFIG_BTRFS_DEBUG
2347 #ifdef CONFIG_BTRFS_ASSERT
2350 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2351 ", integrity-checker=on"
2354 btrfs_crc32c_impl());
2357 static int __init init_btrfs_fs(void)
2361 err = btrfs_hash_init();
2367 err = btrfs_init_sysfs();
2371 btrfs_init_compress();
2373 err = btrfs_init_cachep();
2377 err = extent_io_init();
2381 err = extent_map_init();
2383 goto free_extent_io;
2385 err = ordered_data_init();
2387 goto free_extent_map;
2389 err = btrfs_delayed_inode_init();
2391 goto free_ordered_data;
2393 err = btrfs_auto_defrag_init();
2395 goto free_delayed_inode;
2397 err = btrfs_delayed_ref_init();
2399 goto free_auto_defrag;
2401 err = btrfs_prelim_ref_init();
2403 goto free_delayed_ref;
2405 err = btrfs_end_io_wq_init();
2407 goto free_prelim_ref;
2409 err = btrfs_interface_init();
2411 goto free_end_io_wq;
2413 btrfs_init_lockdep();
2415 btrfs_print_mod_info();
2417 err = btrfs_run_sanity_tests();
2419 goto unregister_ioctl;
2421 err = register_filesystem(&btrfs_fs_type);
2423 goto unregister_ioctl;
2428 btrfs_interface_exit();
2430 btrfs_end_io_wq_exit();
2432 btrfs_prelim_ref_exit();
2434 btrfs_delayed_ref_exit();
2436 btrfs_auto_defrag_exit();
2438 btrfs_delayed_inode_exit();
2440 ordered_data_exit();
2446 btrfs_destroy_cachep();
2448 btrfs_exit_compress();
2455 static void __exit exit_btrfs_fs(void)
2457 btrfs_destroy_cachep();
2458 btrfs_delayed_ref_exit();
2459 btrfs_auto_defrag_exit();
2460 btrfs_delayed_inode_exit();
2461 btrfs_prelim_ref_exit();
2462 ordered_data_exit();
2465 btrfs_interface_exit();
2466 btrfs_end_io_wq_exit();
2467 unregister_filesystem(&btrfs_fs_type);
2469 btrfs_cleanup_fs_uuids();
2470 btrfs_exit_compress();
2474 late_initcall(init_btrfs_fs);
2475 module_exit(exit_btrfs_fs)
2477 MODULE_LICENSE("GPL");