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;
205 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
206 struct va_format vaf;
209 const char *type = logtypes[4];
210 struct ratelimit_state *ratelimit = &printk_limits[4];
214 while ((kern_level = printk_get_level(fmt)) != 0) {
215 size_t size = printk_skip_level(fmt) - fmt;
217 if (kern_level >= '0' && kern_level <= '7') {
218 memcpy(lvl, fmt, size);
220 type = logtypes[kern_level - '0'];
221 ratelimit = &printk_limits[kern_level - '0'];
229 if (__ratelimit(ratelimit))
230 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
237 * We only mark the transaction aborted and then set the file system read-only.
238 * This will prevent new transactions from starting or trying to join this
241 * This means that error recovery at the call site is limited to freeing
242 * any local memory allocations and passing the error code up without
243 * further cleanup. The transaction should complete as it normally would
244 * in the call path but will return -EIO.
246 * We'll complete the cleanup in btrfs_end_transaction and
247 * btrfs_commit_transaction.
250 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
251 const char *function,
252 unsigned int line, int errno)
254 struct btrfs_fs_info *fs_info = trans->fs_info;
256 trans->aborted = errno;
257 /* Nothing used. The other threads that have joined this
258 * transaction may be able to continue. */
259 if (!trans->dirty && list_empty(&trans->new_bgs)) {
262 errstr = btrfs_decode_error(errno);
264 "%s:%d: Aborting unused transaction(%s).",
265 function, line, errstr);
268 WRITE_ONCE(trans->transaction->aborted, errno);
269 /* Wake up anybody who may be waiting on this transaction */
270 wake_up(&fs_info->transaction_wait);
271 wake_up(&fs_info->transaction_blocked_wait);
272 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
275 * __btrfs_panic decodes unexpected, fatal errors from the caller,
276 * issues an alert, and either panics or BUGs, depending on mount options.
279 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
280 unsigned int line, int errno, const char *fmt, ...)
282 char *s_id = "<unknown>";
284 struct va_format vaf = { .fmt = fmt };
288 s_id = fs_info->sb->s_id;
293 errstr = btrfs_decode_error(errno);
294 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
295 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
296 s_id, function, line, &vaf, errno, errstr);
298 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
299 function, line, &vaf, errno, errstr);
301 /* Caller calls BUG() */
304 static void btrfs_put_super(struct super_block *sb)
306 close_ctree(btrfs_sb(sb));
310 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
311 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
312 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
313 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
314 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
315 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
316 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
317 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
318 Opt_skip_balance, Opt_check_integrity,
319 Opt_check_integrity_including_extent_data,
320 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
321 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
322 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
323 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
324 Opt_nologreplay, Opt_norecovery,
325 #ifdef CONFIG_BTRFS_DEBUG
326 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
331 static const match_table_t tokens = {
332 {Opt_degraded, "degraded"},
333 {Opt_subvol, "subvol=%s"},
334 {Opt_subvolid, "subvolid=%s"},
335 {Opt_device, "device=%s"},
336 {Opt_nodatasum, "nodatasum"},
337 {Opt_datasum, "datasum"},
338 {Opt_nodatacow, "nodatacow"},
339 {Opt_datacow, "datacow"},
340 {Opt_nobarrier, "nobarrier"},
341 {Opt_barrier, "barrier"},
342 {Opt_max_inline, "max_inline=%s"},
343 {Opt_alloc_start, "alloc_start=%s"},
344 {Opt_thread_pool, "thread_pool=%d"},
345 {Opt_compress, "compress"},
346 {Opt_compress_type, "compress=%s"},
347 {Opt_compress_force, "compress-force"},
348 {Opt_compress_force_type, "compress-force=%s"},
350 {Opt_ssd_spread, "ssd_spread"},
351 {Opt_nossd, "nossd"},
353 {Opt_noacl, "noacl"},
354 {Opt_notreelog, "notreelog"},
355 {Opt_treelog, "treelog"},
356 {Opt_nologreplay, "nologreplay"},
357 {Opt_norecovery, "norecovery"},
358 {Opt_flushoncommit, "flushoncommit"},
359 {Opt_noflushoncommit, "noflushoncommit"},
360 {Opt_ratio, "metadata_ratio=%d"},
361 {Opt_discard, "discard"},
362 {Opt_nodiscard, "nodiscard"},
363 {Opt_space_cache, "space_cache"},
364 {Opt_space_cache_version, "space_cache=%s"},
365 {Opt_clear_cache, "clear_cache"},
366 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
367 {Opt_enospc_debug, "enospc_debug"},
368 {Opt_noenospc_debug, "noenospc_debug"},
369 {Opt_subvolrootid, "subvolrootid=%d"},
370 {Opt_defrag, "autodefrag"},
371 {Opt_nodefrag, "noautodefrag"},
372 {Opt_inode_cache, "inode_cache"},
373 {Opt_noinode_cache, "noinode_cache"},
374 {Opt_no_space_cache, "nospace_cache"},
375 {Opt_recovery, "recovery"}, /* deprecated */
376 {Opt_usebackuproot, "usebackuproot"},
377 {Opt_skip_balance, "skip_balance"},
378 {Opt_check_integrity, "check_int"},
379 {Opt_check_integrity_including_extent_data, "check_int_data"},
380 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
381 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
382 {Opt_fatal_errors, "fatal_errors=%s"},
383 {Opt_commit_interval, "commit=%d"},
384 #ifdef CONFIG_BTRFS_DEBUG
385 {Opt_fragment_data, "fragment=data"},
386 {Opt_fragment_metadata, "fragment=metadata"},
387 {Opt_fragment_all, "fragment=all"},
393 * Regular mount options parser. Everything that is needed only when
394 * reading in a new superblock is parsed here.
395 * XXX JDM: This needs to be cleaned up for remount.
397 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
398 unsigned long new_flags)
400 substring_t args[MAX_OPT_ARGS];
401 char *p, *num, *orig = NULL;
406 bool compress_force = false;
407 enum btrfs_compression_type saved_compress_type;
408 bool saved_compress_force;
411 cache_gen = btrfs_super_cache_generation(info->super_copy);
412 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
413 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
415 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
418 * Even the options are empty, we still need to do extra check
425 * strsep changes the string, duplicate it because parse_options
428 options = kstrdup(options, GFP_NOFS);
434 while ((p = strsep(&options, ",")) != NULL) {
439 token = match_token(p, tokens, args);
442 btrfs_info(info, "allowing degraded mounts");
443 btrfs_set_opt(info->mount_opt, DEGRADED);
447 case Opt_subvolrootid:
450 * These are parsed by btrfs_parse_early_options
451 * and can be happily ignored here.
455 btrfs_set_and_info(info, NODATASUM,
456 "setting nodatasum");
459 if (btrfs_test_opt(info, NODATASUM)) {
460 if (btrfs_test_opt(info, NODATACOW))
462 "setting datasum, datacow enabled");
464 btrfs_info(info, "setting datasum");
466 btrfs_clear_opt(info->mount_opt, NODATACOW);
467 btrfs_clear_opt(info->mount_opt, NODATASUM);
470 if (!btrfs_test_opt(info, NODATACOW)) {
471 if (!btrfs_test_opt(info, COMPRESS) ||
472 !btrfs_test_opt(info, FORCE_COMPRESS)) {
474 "setting nodatacow, compression disabled");
476 btrfs_info(info, "setting nodatacow");
479 btrfs_clear_opt(info->mount_opt, COMPRESS);
480 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
481 btrfs_set_opt(info->mount_opt, NODATACOW);
482 btrfs_set_opt(info->mount_opt, NODATASUM);
485 btrfs_clear_and_info(info, NODATACOW,
488 case Opt_compress_force:
489 case Opt_compress_force_type:
490 compress_force = true;
493 case Opt_compress_type:
494 saved_compress_type = btrfs_test_opt(info,
496 info->compress_type : BTRFS_COMPRESS_NONE;
497 saved_compress_force =
498 btrfs_test_opt(info, FORCE_COMPRESS);
499 if (token == Opt_compress ||
500 token == Opt_compress_force ||
501 strcmp(args[0].from, "zlib") == 0) {
502 compress_type = "zlib";
503 info->compress_type = BTRFS_COMPRESS_ZLIB;
504 btrfs_set_opt(info->mount_opt, COMPRESS);
505 btrfs_clear_opt(info->mount_opt, NODATACOW);
506 btrfs_clear_opt(info->mount_opt, NODATASUM);
508 } else if (strcmp(args[0].from, "lzo") == 0) {
509 compress_type = "lzo";
510 info->compress_type = BTRFS_COMPRESS_LZO;
511 btrfs_set_opt(info->mount_opt, COMPRESS);
512 btrfs_clear_opt(info->mount_opt, NODATACOW);
513 btrfs_clear_opt(info->mount_opt, NODATASUM);
514 btrfs_set_fs_incompat(info, COMPRESS_LZO);
516 } else if (strncmp(args[0].from, "no", 2) == 0) {
517 compress_type = "no";
518 btrfs_clear_opt(info->mount_opt, COMPRESS);
519 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
520 compress_force = false;
527 if (compress_force) {
528 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
531 * If we remount from compress-force=xxx to
532 * compress=xxx, we need clear FORCE_COMPRESS
533 * flag, otherwise, there is no way for users
534 * to disable forcible compression separately.
536 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
538 if ((btrfs_test_opt(info, COMPRESS) &&
539 (info->compress_type != saved_compress_type ||
540 compress_force != saved_compress_force)) ||
541 (!btrfs_test_opt(info, COMPRESS) &&
543 btrfs_info(info, "%s %s compression",
544 (compress_force) ? "force" : "use",
547 compress_force = false;
550 btrfs_set_and_info(info, SSD,
551 "use ssd allocation scheme");
552 btrfs_clear_opt(info->mount_opt, NOSSD);
555 btrfs_set_and_info(info, SSD_SPREAD,
556 "use spread ssd allocation scheme");
557 btrfs_set_opt(info->mount_opt, SSD);
558 btrfs_clear_opt(info->mount_opt, NOSSD);
561 btrfs_set_and_info(info, NOSSD,
562 "not using ssd allocation scheme");
563 btrfs_clear_opt(info->mount_opt, SSD);
564 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
567 btrfs_clear_and_info(info, NOBARRIER,
568 "turning on barriers");
571 btrfs_set_and_info(info, NOBARRIER,
572 "turning off barriers");
574 case Opt_thread_pool:
575 ret = match_int(&args[0], &intarg);
578 } else if (intarg > 0) {
579 info->thread_pool_size = intarg;
586 num = match_strdup(&args[0]);
588 info->max_inline = memparse(num, NULL);
591 if (info->max_inline) {
592 info->max_inline = min_t(u64,
596 btrfs_info(info, "max_inline at %llu",
603 case Opt_alloc_start:
604 num = match_strdup(&args[0]);
606 mutex_lock(&info->chunk_mutex);
607 info->alloc_start = memparse(num, NULL);
608 mutex_unlock(&info->chunk_mutex);
610 btrfs_info(info, "allocations start at %llu",
618 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
619 info->sb->s_flags |= MS_POSIXACL;
622 btrfs_err(info, "support for ACL not compiled in!");
627 info->sb->s_flags &= ~MS_POSIXACL;
630 btrfs_set_and_info(info, NOTREELOG,
631 "disabling tree log");
634 btrfs_clear_and_info(info, NOTREELOG,
635 "enabling tree log");
638 case Opt_nologreplay:
639 btrfs_set_and_info(info, NOLOGREPLAY,
640 "disabling log replay at mount time");
642 case Opt_flushoncommit:
643 btrfs_set_and_info(info, FLUSHONCOMMIT,
644 "turning on flush-on-commit");
646 case Opt_noflushoncommit:
647 btrfs_clear_and_info(info, FLUSHONCOMMIT,
648 "turning off flush-on-commit");
651 ret = match_int(&args[0], &intarg);
654 } else if (intarg >= 0) {
655 info->metadata_ratio = intarg;
656 btrfs_info(info, "metadata ratio %d",
657 info->metadata_ratio);
664 btrfs_set_and_info(info, DISCARD,
665 "turning on discard");
668 btrfs_clear_and_info(info, DISCARD,
669 "turning off discard");
671 case Opt_space_cache:
672 case Opt_space_cache_version:
673 if (token == Opt_space_cache ||
674 strcmp(args[0].from, "v1") == 0) {
675 btrfs_clear_opt(info->mount_opt,
677 btrfs_set_and_info(info, SPACE_CACHE,
678 "enabling disk space caching");
679 } else if (strcmp(args[0].from, "v2") == 0) {
680 btrfs_clear_opt(info->mount_opt,
682 btrfs_set_and_info(info, FREE_SPACE_TREE,
683 "enabling free space tree");
689 case Opt_rescan_uuid_tree:
690 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
692 case Opt_no_space_cache:
693 if (btrfs_test_opt(info, SPACE_CACHE)) {
694 btrfs_clear_and_info(info, SPACE_CACHE,
695 "disabling disk space caching");
697 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
698 btrfs_clear_and_info(info, FREE_SPACE_TREE,
699 "disabling free space tree");
702 case Opt_inode_cache:
703 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
704 "enabling inode map caching");
706 case Opt_noinode_cache:
707 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
708 "disabling inode map caching");
710 case Opt_clear_cache:
711 btrfs_set_and_info(info, CLEAR_CACHE,
712 "force clearing of disk cache");
714 case Opt_user_subvol_rm_allowed:
715 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
717 case Opt_enospc_debug:
718 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
720 case Opt_noenospc_debug:
721 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
724 btrfs_set_and_info(info, AUTO_DEFRAG,
725 "enabling auto defrag");
728 btrfs_clear_and_info(info, AUTO_DEFRAG,
729 "disabling auto defrag");
733 "'recovery' is deprecated, use 'usebackuproot' instead");
734 case Opt_usebackuproot:
736 "trying to use backup root at mount time");
737 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
739 case Opt_skip_balance:
740 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
742 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
743 case Opt_check_integrity_including_extent_data:
745 "enabling check integrity including extent data");
746 btrfs_set_opt(info->mount_opt,
747 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
748 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
750 case Opt_check_integrity:
751 btrfs_info(info, "enabling check integrity");
752 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
754 case Opt_check_integrity_print_mask:
755 ret = match_int(&args[0], &intarg);
758 } else if (intarg >= 0) {
759 info->check_integrity_print_mask = intarg;
761 "check_integrity_print_mask 0x%x",
762 info->check_integrity_print_mask);
769 case Opt_check_integrity_including_extent_data:
770 case Opt_check_integrity:
771 case Opt_check_integrity_print_mask:
773 "support for check_integrity* not compiled in!");
777 case Opt_fatal_errors:
778 if (strcmp(args[0].from, "panic") == 0)
779 btrfs_set_opt(info->mount_opt,
780 PANIC_ON_FATAL_ERROR);
781 else if (strcmp(args[0].from, "bug") == 0)
782 btrfs_clear_opt(info->mount_opt,
783 PANIC_ON_FATAL_ERROR);
789 case Opt_commit_interval:
791 ret = match_int(&args[0], &intarg);
793 btrfs_err(info, "invalid commit interval");
800 "excessive commit interval %d",
803 info->commit_interval = intarg;
806 "using default commit interval %ds",
807 BTRFS_DEFAULT_COMMIT_INTERVAL);
808 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
811 #ifdef CONFIG_BTRFS_DEBUG
812 case Opt_fragment_all:
813 btrfs_info(info, "fragmenting all space");
814 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
815 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
817 case Opt_fragment_metadata:
818 btrfs_info(info, "fragmenting metadata");
819 btrfs_set_opt(info->mount_opt,
822 case Opt_fragment_data:
823 btrfs_info(info, "fragmenting data");
824 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
828 btrfs_info(info, "unrecognized mount option '%s'", p);
837 * Extra check for current option against current flag
839 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
841 "nologreplay must be used with ro mount option");
845 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
846 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
847 !btrfs_test_opt(info, CLEAR_CACHE)) {
848 btrfs_err(info, "cannot disable free space tree");
852 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
853 btrfs_info(info, "disk space caching is enabled");
854 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
855 btrfs_info(info, "using free space tree");
861 * Parse mount options that are required early in the mount process.
863 * All other options will be parsed on much later in the mount process and
864 * only when we need to allocate a new super block.
866 static int btrfs_parse_early_options(const char *options, fmode_t flags,
867 void *holder, char **subvol_name, u64 *subvol_objectid,
868 struct btrfs_fs_devices **fs_devices)
870 substring_t args[MAX_OPT_ARGS];
871 char *device_name, *opts, *orig, *p;
879 * strsep changes the string, duplicate it because parse_options
882 opts = kstrdup(options, GFP_KERNEL);
887 while ((p = strsep(&opts, ",")) != NULL) {
892 token = match_token(p, tokens, args);
896 *subvol_name = match_strdup(&args[0]);
903 num = match_strdup(&args[0]);
905 *subvol_objectid = memparse(num, NULL);
907 /* we want the original fs_tree */
908 if (!*subvol_objectid)
910 BTRFS_FS_TREE_OBJECTID;
916 case Opt_subvolrootid:
917 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
920 device_name = match_strdup(&args[0]);
925 error = btrfs_scan_one_device(device_name,
926 flags, holder, fs_devices);
941 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
944 struct btrfs_root *root = fs_info->tree_root;
945 struct btrfs_root *fs_root;
946 struct btrfs_root_ref *root_ref;
947 struct btrfs_inode_ref *inode_ref;
948 struct btrfs_key key;
949 struct btrfs_path *path = NULL;
950 char *name = NULL, *ptr;
955 path = btrfs_alloc_path();
960 path->leave_spinning = 1;
962 name = kmalloc(PATH_MAX, GFP_NOFS);
967 ptr = name + PATH_MAX - 1;
971 * Walk up the subvolume trees in the tree of tree roots by root
972 * backrefs until we hit the top-level subvolume.
974 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
975 key.objectid = subvol_objectid;
976 key.type = BTRFS_ROOT_BACKREF_KEY;
977 key.offset = (u64)-1;
979 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
982 } else if (ret > 0) {
983 ret = btrfs_previous_item(root, path, subvol_objectid,
984 BTRFS_ROOT_BACKREF_KEY);
987 } else if (ret > 0) {
993 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
994 subvol_objectid = key.offset;
996 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
997 struct btrfs_root_ref);
998 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1001 ret = -ENAMETOOLONG;
1004 read_extent_buffer(path->nodes[0], ptr + 1,
1005 (unsigned long)(root_ref + 1), len);
1007 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1008 btrfs_release_path(path);
1010 key.objectid = subvol_objectid;
1011 key.type = BTRFS_ROOT_ITEM_KEY;
1012 key.offset = (u64)-1;
1013 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1014 if (IS_ERR(fs_root)) {
1015 ret = PTR_ERR(fs_root);
1020 * Walk up the filesystem tree by inode refs until we hit the
1023 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1024 key.objectid = dirid;
1025 key.type = BTRFS_INODE_REF_KEY;
1026 key.offset = (u64)-1;
1028 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1031 } else if (ret > 0) {
1032 ret = btrfs_previous_item(fs_root, path, dirid,
1033 BTRFS_INODE_REF_KEY);
1036 } else if (ret > 0) {
1042 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1045 inode_ref = btrfs_item_ptr(path->nodes[0],
1047 struct btrfs_inode_ref);
1048 len = btrfs_inode_ref_name_len(path->nodes[0],
1052 ret = -ENAMETOOLONG;
1055 read_extent_buffer(path->nodes[0], ptr + 1,
1056 (unsigned long)(inode_ref + 1), len);
1058 btrfs_release_path(path);
1062 btrfs_free_path(path);
1063 if (ptr == name + PATH_MAX - 1) {
1067 memmove(name, ptr, name + PATH_MAX - ptr);
1072 btrfs_free_path(path);
1074 return ERR_PTR(ret);
1077 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1079 struct btrfs_root *root = fs_info->tree_root;
1080 struct btrfs_dir_item *di;
1081 struct btrfs_path *path;
1082 struct btrfs_key location;
1085 path = btrfs_alloc_path();
1088 path->leave_spinning = 1;
1091 * Find the "default" dir item which points to the root item that we
1092 * will mount by default if we haven't been given a specific subvolume
1095 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1096 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1098 btrfs_free_path(path);
1103 * Ok the default dir item isn't there. This is weird since
1104 * it's always been there, but don't freak out, just try and
1105 * mount the top-level subvolume.
1107 btrfs_free_path(path);
1108 *objectid = BTRFS_FS_TREE_OBJECTID;
1112 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1113 btrfs_free_path(path);
1114 *objectid = location.objectid;
1118 static int btrfs_fill_super(struct super_block *sb,
1119 struct btrfs_fs_devices *fs_devices,
1122 struct inode *inode;
1123 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1124 struct btrfs_key key;
1127 sb->s_maxbytes = MAX_LFS_FILESIZE;
1128 sb->s_magic = BTRFS_SUPER_MAGIC;
1129 sb->s_op = &btrfs_super_ops;
1130 sb->s_d_op = &btrfs_dentry_operations;
1131 sb->s_export_op = &btrfs_export_ops;
1132 sb->s_xattr = btrfs_xattr_handlers;
1133 sb->s_time_gran = 1;
1134 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1135 sb->s_flags |= MS_POSIXACL;
1137 sb->s_flags |= MS_I_VERSION;
1138 sb->s_iflags |= SB_I_CGROUPWB;
1140 err = super_setup_bdi(sb);
1142 btrfs_err(fs_info, "super_setup_bdi failed");
1146 err = open_ctree(sb, fs_devices, (char *)data);
1148 btrfs_err(fs_info, "open_ctree failed");
1152 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1153 key.type = BTRFS_INODE_ITEM_KEY;
1155 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1156 if (IS_ERR(inode)) {
1157 err = PTR_ERR(inode);
1161 sb->s_root = d_make_root(inode);
1167 save_mount_options(sb, data);
1168 cleancache_init_fs(sb);
1169 sb->s_flags |= MS_ACTIVE;
1173 close_ctree(fs_info);
1177 int btrfs_sync_fs(struct super_block *sb, int wait)
1179 struct btrfs_trans_handle *trans;
1180 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1181 struct btrfs_root *root = fs_info->tree_root;
1183 trace_btrfs_sync_fs(fs_info, wait);
1186 filemap_flush(fs_info->btree_inode->i_mapping);
1190 btrfs_wait_ordered_roots(fs_info, -1, 0, (u64)-1);
1192 trans = btrfs_attach_transaction_barrier(root);
1193 if (IS_ERR(trans)) {
1194 /* no transaction, don't bother */
1195 if (PTR_ERR(trans) == -ENOENT) {
1197 * Exit unless we have some pending changes
1198 * that need to go through commit
1200 if (fs_info->pending_changes == 0)
1203 * A non-blocking test if the fs is frozen. We must not
1204 * start a new transaction here otherwise a deadlock
1205 * happens. The pending operations are delayed to the
1206 * next commit after thawing.
1208 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1209 __sb_end_write(sb, SB_FREEZE_WRITE);
1212 trans = btrfs_start_transaction(root, 0);
1215 return PTR_ERR(trans);
1217 return btrfs_commit_transaction(trans);
1220 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1222 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1223 char *compress_type;
1225 if (btrfs_test_opt(info, DEGRADED))
1226 seq_puts(seq, ",degraded");
1227 if (btrfs_test_opt(info, NODATASUM))
1228 seq_puts(seq, ",nodatasum");
1229 if (btrfs_test_opt(info, NODATACOW))
1230 seq_puts(seq, ",nodatacow");
1231 if (btrfs_test_opt(info, NOBARRIER))
1232 seq_puts(seq, ",nobarrier");
1233 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1234 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1235 if (info->alloc_start != 0)
1236 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1237 if (info->thread_pool_size != min_t(unsigned long,
1238 num_online_cpus() + 2, 8))
1239 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1240 if (btrfs_test_opt(info, COMPRESS)) {
1241 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1242 compress_type = "zlib";
1244 compress_type = "lzo";
1245 if (btrfs_test_opt(info, FORCE_COMPRESS))
1246 seq_printf(seq, ",compress-force=%s", compress_type);
1248 seq_printf(seq, ",compress=%s", compress_type);
1250 if (btrfs_test_opt(info, NOSSD))
1251 seq_puts(seq, ",nossd");
1252 if (btrfs_test_opt(info, SSD_SPREAD))
1253 seq_puts(seq, ",ssd_spread");
1254 else if (btrfs_test_opt(info, SSD))
1255 seq_puts(seq, ",ssd");
1256 if (btrfs_test_opt(info, NOTREELOG))
1257 seq_puts(seq, ",notreelog");
1258 if (btrfs_test_opt(info, NOLOGREPLAY))
1259 seq_puts(seq, ",nologreplay");
1260 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1261 seq_puts(seq, ",flushoncommit");
1262 if (btrfs_test_opt(info, DISCARD))
1263 seq_puts(seq, ",discard");
1264 if (!(info->sb->s_flags & MS_POSIXACL))
1265 seq_puts(seq, ",noacl");
1266 if (btrfs_test_opt(info, SPACE_CACHE))
1267 seq_puts(seq, ",space_cache");
1268 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1269 seq_puts(seq, ",space_cache=v2");
1271 seq_puts(seq, ",nospace_cache");
1272 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1273 seq_puts(seq, ",rescan_uuid_tree");
1274 if (btrfs_test_opt(info, CLEAR_CACHE))
1275 seq_puts(seq, ",clear_cache");
1276 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1277 seq_puts(seq, ",user_subvol_rm_allowed");
1278 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1279 seq_puts(seq, ",enospc_debug");
1280 if (btrfs_test_opt(info, AUTO_DEFRAG))
1281 seq_puts(seq, ",autodefrag");
1282 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1283 seq_puts(seq, ",inode_cache");
1284 if (btrfs_test_opt(info, SKIP_BALANCE))
1285 seq_puts(seq, ",skip_balance");
1286 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1287 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1288 seq_puts(seq, ",check_int_data");
1289 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1290 seq_puts(seq, ",check_int");
1291 if (info->check_integrity_print_mask)
1292 seq_printf(seq, ",check_int_print_mask=%d",
1293 info->check_integrity_print_mask);
1295 if (info->metadata_ratio)
1296 seq_printf(seq, ",metadata_ratio=%d",
1297 info->metadata_ratio);
1298 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1299 seq_puts(seq, ",fatal_errors=panic");
1300 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1301 seq_printf(seq, ",commit=%d", info->commit_interval);
1302 #ifdef CONFIG_BTRFS_DEBUG
1303 if (btrfs_test_opt(info, FRAGMENT_DATA))
1304 seq_puts(seq, ",fragment=data");
1305 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1306 seq_puts(seq, ",fragment=metadata");
1308 seq_printf(seq, ",subvolid=%llu",
1309 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1310 seq_puts(seq, ",subvol=");
1311 seq_dentry(seq, dentry, " \t\n\\");
1315 static int btrfs_test_super(struct super_block *s, void *data)
1317 struct btrfs_fs_info *p = data;
1318 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1320 return fs_info->fs_devices == p->fs_devices;
1323 static int btrfs_set_super(struct super_block *s, void *data)
1325 int err = set_anon_super(s, data);
1327 s->s_fs_info = data;
1332 * subvolumes are identified by ino 256
1334 static inline int is_subvolume_inode(struct inode *inode)
1336 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1342 * This will add subvolid=0 to the argument string while removing any subvol=
1343 * and subvolid= arguments to make sure we get the top-level root for path
1344 * walking to the subvol we want.
1346 static char *setup_root_args(char *args)
1348 char *buf, *dst, *sep;
1351 return kstrdup("subvolid=0", GFP_NOFS);
1353 /* The worst case is that we add ",subvolid=0" to the end. */
1354 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
1359 sep = strchrnul(args, ',');
1360 if (!strstarts(args, "subvol=") &&
1361 !strstarts(args, "subvolid=")) {
1362 memcpy(dst, args, sep - args);
1371 strcpy(dst, "subvolid=0");
1376 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1377 int flags, const char *device_name,
1380 struct dentry *root;
1381 struct vfsmount *mnt = NULL;
1385 newargs = setup_root_args(data);
1387 root = ERR_PTR(-ENOMEM);
1391 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
1392 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
1393 if (flags & MS_RDONLY) {
1394 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
1395 device_name, newargs);
1397 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
1398 device_name, newargs);
1400 root = ERR_CAST(mnt);
1405 down_write(&mnt->mnt_sb->s_umount);
1406 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1407 up_write(&mnt->mnt_sb->s_umount);
1409 root = ERR_PTR(ret);
1415 root = ERR_CAST(mnt);
1421 if (!subvol_objectid) {
1422 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1425 root = ERR_PTR(ret);
1429 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1431 if (IS_ERR(subvol_name)) {
1432 root = ERR_CAST(subvol_name);
1439 root = mount_subtree(mnt, subvol_name);
1440 /* mount_subtree() drops our reference on the vfsmount. */
1443 if (!IS_ERR(root)) {
1444 struct super_block *s = root->d_sb;
1445 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1446 struct inode *root_inode = d_inode(root);
1447 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1450 if (!is_subvolume_inode(root_inode)) {
1451 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1455 if (subvol_objectid && root_objectid != subvol_objectid) {
1457 * This will also catch a race condition where a
1458 * subvolume which was passed by ID is renamed and
1459 * another subvolume is renamed over the old location.
1462 "subvol '%s' does not match subvolid %llu",
1463 subvol_name, subvol_objectid);
1468 root = ERR_PTR(ret);
1469 deactivate_locked_super(s);
1480 static int parse_security_options(char *orig_opts,
1481 struct security_mnt_opts *sec_opts)
1483 char *secdata = NULL;
1486 secdata = alloc_secdata();
1489 ret = security_sb_copy_data(orig_opts, secdata);
1491 free_secdata(secdata);
1494 ret = security_sb_parse_opts_str(secdata, sec_opts);
1495 free_secdata(secdata);
1499 static int setup_security_options(struct btrfs_fs_info *fs_info,
1500 struct super_block *sb,
1501 struct security_mnt_opts *sec_opts)
1506 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1509 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1513 #ifdef CONFIG_SECURITY
1514 if (!fs_info->security_opts.num_mnt_opts) {
1515 /* first time security setup, copy sec_opts to fs_info */
1516 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1519 * Since SELinux (the only one supporting security_mnt_opts)
1520 * does NOT support changing context during remount/mount of
1521 * the same sb, this must be the same or part of the same
1522 * security options, just free it.
1524 security_free_mnt_opts(sec_opts);
1531 * Find a superblock for the given device / mount point.
1533 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1534 * for multiple device setup. Make sure to keep it in sync.
1536 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1537 const char *device_name, void *data)
1539 struct block_device *bdev = NULL;
1540 struct super_block *s;
1541 struct btrfs_fs_devices *fs_devices = NULL;
1542 struct btrfs_fs_info *fs_info = NULL;
1543 struct security_mnt_opts new_sec_opts;
1544 fmode_t mode = FMODE_READ;
1545 char *subvol_name = NULL;
1546 u64 subvol_objectid = 0;
1549 if (!(flags & MS_RDONLY))
1550 mode |= FMODE_WRITE;
1552 error = btrfs_parse_early_options(data, mode, fs_type,
1553 &subvol_name, &subvol_objectid,
1557 return ERR_PTR(error);
1560 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1561 /* mount_subvol() will free subvol_name. */
1562 return mount_subvol(subvol_name, subvol_objectid, flags,
1566 security_init_mnt_opts(&new_sec_opts);
1568 error = parse_security_options(data, &new_sec_opts);
1570 return ERR_PTR(error);
1573 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1575 goto error_sec_opts;
1578 * Setup a dummy root and fs_info for test/set super. This is because
1579 * we don't actually fill this stuff out until open_ctree, but we need
1580 * it for searching for existing supers, so this lets us do that and
1581 * then open_ctree will properly initialize everything later.
1583 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1586 goto error_sec_opts;
1589 fs_info->fs_devices = fs_devices;
1591 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1592 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1593 security_init_mnt_opts(&fs_info->security_opts);
1594 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1599 error = btrfs_open_devices(fs_devices, mode, fs_type);
1603 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1605 goto error_close_devices;
1608 bdev = fs_devices->latest_bdev;
1609 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1613 goto error_close_devices;
1617 btrfs_close_devices(fs_devices);
1618 free_fs_info(fs_info);
1619 if ((flags ^ s->s_flags) & MS_RDONLY)
1622 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1623 btrfs_sb(s)->bdev_holder = fs_type;
1624 error = btrfs_fill_super(s, fs_devices, data);
1627 deactivate_locked_super(s);
1628 goto error_sec_opts;
1631 fs_info = btrfs_sb(s);
1632 error = setup_security_options(fs_info, s, &new_sec_opts);
1634 deactivate_locked_super(s);
1635 goto error_sec_opts;
1638 return dget(s->s_root);
1640 error_close_devices:
1641 btrfs_close_devices(fs_devices);
1643 free_fs_info(fs_info);
1645 security_free_mnt_opts(&new_sec_opts);
1646 return ERR_PTR(error);
1649 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1650 int new_pool_size, int old_pool_size)
1652 if (new_pool_size == old_pool_size)
1655 fs_info->thread_pool_size = new_pool_size;
1657 btrfs_info(fs_info, "resize thread pool %d -> %d",
1658 old_pool_size, new_pool_size);
1660 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1661 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1662 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1663 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1664 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1665 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1666 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1668 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1669 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1670 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1671 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1672 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1676 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1678 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1681 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1682 unsigned long old_opts, int flags)
1684 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1685 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1686 (flags & MS_RDONLY))) {
1687 /* wait for any defraggers to finish */
1688 wait_event(fs_info->transaction_wait,
1689 (atomic_read(&fs_info->defrag_running) == 0));
1690 if (flags & MS_RDONLY)
1691 sync_filesystem(fs_info->sb);
1695 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1696 unsigned long old_opts)
1699 * We need to cleanup all defragable inodes if the autodefragment is
1700 * close or the filesystem is read only.
1702 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1703 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1704 (fs_info->sb->s_flags & MS_RDONLY))) {
1705 btrfs_cleanup_defrag_inodes(fs_info);
1708 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1711 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1713 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1714 struct btrfs_root *root = fs_info->tree_root;
1715 unsigned old_flags = sb->s_flags;
1716 unsigned long old_opts = fs_info->mount_opt;
1717 unsigned long old_compress_type = fs_info->compress_type;
1718 u64 old_max_inline = fs_info->max_inline;
1719 u64 old_alloc_start = fs_info->alloc_start;
1720 int old_thread_pool_size = fs_info->thread_pool_size;
1721 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1724 sync_filesystem(sb);
1725 btrfs_remount_prepare(fs_info);
1728 struct security_mnt_opts new_sec_opts;
1730 security_init_mnt_opts(&new_sec_opts);
1731 ret = parse_security_options(data, &new_sec_opts);
1734 ret = setup_security_options(fs_info, sb,
1737 security_free_mnt_opts(&new_sec_opts);
1742 ret = btrfs_parse_options(fs_info, data, *flags);
1748 btrfs_remount_begin(fs_info, old_opts, *flags);
1749 btrfs_resize_thread_pool(fs_info,
1750 fs_info->thread_pool_size, old_thread_pool_size);
1752 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1755 if (*flags & MS_RDONLY) {
1757 * this also happens on 'umount -rf' or on shutdown, when
1758 * the filesystem is busy.
1760 cancel_work_sync(&fs_info->async_reclaim_work);
1762 /* wait for the uuid_scan task to finish */
1763 down(&fs_info->uuid_tree_rescan_sem);
1764 /* avoid complains from lockdep et al. */
1765 up(&fs_info->uuid_tree_rescan_sem);
1767 sb->s_flags |= MS_RDONLY;
1770 * Setting MS_RDONLY will put the cleaner thread to
1771 * sleep at the next loop if it's already active.
1772 * If it's already asleep, we'll leave unused block
1773 * groups on disk until we're mounted read-write again
1774 * unless we clean them up here.
1776 btrfs_delete_unused_bgs(fs_info);
1778 btrfs_dev_replace_suspend_for_unmount(fs_info);
1779 btrfs_scrub_cancel(fs_info);
1780 btrfs_pause_balance(fs_info);
1782 ret = btrfs_commit_super(fs_info);
1786 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1788 "Remounting read-write after error is not allowed");
1792 if (fs_info->fs_devices->rw_devices == 0) {
1797 if (fs_info->fs_devices->missing_devices >
1798 fs_info->num_tolerated_disk_barrier_failures) {
1800 "too many missing devices, writeable remount is not allowed");
1805 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1810 ret = btrfs_cleanup_fs_roots(fs_info);
1814 /* recover relocation */
1815 mutex_lock(&fs_info->cleaner_mutex);
1816 ret = btrfs_recover_relocation(root);
1817 mutex_unlock(&fs_info->cleaner_mutex);
1821 ret = btrfs_resume_balance_async(fs_info);
1825 ret = btrfs_resume_dev_replace_async(fs_info);
1827 btrfs_warn(fs_info, "failed to resume dev_replace");
1831 if (!fs_info->uuid_root) {
1832 btrfs_info(fs_info, "creating UUID tree");
1833 ret = btrfs_create_uuid_tree(fs_info);
1836 "failed to create the UUID tree %d",
1841 sb->s_flags &= ~MS_RDONLY;
1843 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1846 wake_up_process(fs_info->transaction_kthread);
1847 btrfs_remount_cleanup(fs_info, old_opts);
1851 /* We've hit an error - don't reset MS_RDONLY */
1852 if (sb->s_flags & MS_RDONLY)
1853 old_flags |= MS_RDONLY;
1854 sb->s_flags = old_flags;
1855 fs_info->mount_opt = old_opts;
1856 fs_info->compress_type = old_compress_type;
1857 fs_info->max_inline = old_max_inline;
1858 mutex_lock(&fs_info->chunk_mutex);
1859 fs_info->alloc_start = old_alloc_start;
1860 mutex_unlock(&fs_info->chunk_mutex);
1861 btrfs_resize_thread_pool(fs_info,
1862 old_thread_pool_size, fs_info->thread_pool_size);
1863 fs_info->metadata_ratio = old_metadata_ratio;
1864 btrfs_remount_cleanup(fs_info, old_opts);
1868 /* Used to sort the devices by max_avail(descending sort) */
1869 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1870 const void *dev_info2)
1872 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1873 ((struct btrfs_device_info *)dev_info2)->max_avail)
1875 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1876 ((struct btrfs_device_info *)dev_info2)->max_avail)
1883 * sort the devices by max_avail, in which max free extent size of each device
1884 * is stored.(Descending Sort)
1886 static inline void btrfs_descending_sort_devices(
1887 struct btrfs_device_info *devices,
1890 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1891 btrfs_cmp_device_free_bytes, NULL);
1895 * The helper to calc the free space on the devices that can be used to store
1898 static int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1901 struct btrfs_root *root = fs_info->tree_root;
1902 struct btrfs_device_info *devices_info;
1903 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1904 struct btrfs_device *device;
1909 u64 min_stripe_size;
1910 int min_stripes = 1, num_stripes = 1;
1911 int i = 0, nr_devices;
1915 * We aren't under the device list lock, so this is racy-ish, but good
1916 * enough for our purposes.
1918 nr_devices = fs_info->fs_devices->open_devices;
1921 nr_devices = fs_info->fs_devices->open_devices;
1929 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1934 /* calc min stripe number for data space allocation */
1935 type = btrfs_get_alloc_profile(root, 1);
1936 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1938 num_stripes = nr_devices;
1939 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1942 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1947 if (type & BTRFS_BLOCK_GROUP_DUP)
1948 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1950 min_stripe_size = BTRFS_STRIPE_LEN;
1952 if (fs_info->alloc_start)
1953 mutex_lock(&fs_devices->device_list_mutex);
1955 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1956 if (!device->in_fs_metadata || !device->bdev ||
1957 device->is_tgtdev_for_dev_replace)
1960 if (i >= nr_devices)
1963 avail_space = device->total_bytes - device->bytes_used;
1965 /* align with stripe_len */
1966 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1967 avail_space *= BTRFS_STRIPE_LEN;
1970 * In order to avoid overwriting the superblock on the drive,
1971 * btrfs starts at an offset of at least 1MB when doing chunk
1976 /* user can set the offset in fs_info->alloc_start. */
1977 if (fs_info->alloc_start &&
1978 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1979 device->total_bytes) {
1981 skip_space = max(fs_info->alloc_start, skip_space);
1984 * btrfs can not use the free space in
1985 * [0, skip_space - 1], we must subtract it from the
1986 * total. In order to implement it, we account the used
1987 * space in this range first.
1989 ret = btrfs_account_dev_extents_size(device, 0,
1993 kfree(devices_info);
1994 mutex_unlock(&fs_devices->device_list_mutex);
2000 /* calc the free space in [0, skip_space - 1] */
2001 skip_space -= used_space;
2005 * we can use the free space in [0, skip_space - 1], subtract
2006 * it from the total.
2008 if (avail_space && avail_space >= skip_space)
2009 avail_space -= skip_space;
2013 if (avail_space < min_stripe_size)
2016 devices_info[i].dev = device;
2017 devices_info[i].max_avail = avail_space;
2022 if (fs_info->alloc_start)
2023 mutex_unlock(&fs_devices->device_list_mutex);
2027 btrfs_descending_sort_devices(devices_info, nr_devices);
2031 while (nr_devices >= min_stripes) {
2032 if (num_stripes > nr_devices)
2033 num_stripes = nr_devices;
2035 if (devices_info[i].max_avail >= min_stripe_size) {
2039 avail_space += devices_info[i].max_avail * num_stripes;
2040 alloc_size = devices_info[i].max_avail;
2041 for (j = i + 1 - num_stripes; j <= i; j++)
2042 devices_info[j].max_avail -= alloc_size;
2048 kfree(devices_info);
2049 *free_bytes = avail_space;
2054 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2056 * If there's a redundant raid level at DATA block groups, use the respective
2057 * multiplier to scale the sizes.
2059 * Unused device space usage is based on simulating the chunk allocator
2060 * algorithm that respects the device sizes, order of allocations and the
2061 * 'alloc_start' value, this is a close approximation of the actual use but
2062 * there are other factors that may change the result (like a new metadata
2065 * If metadata is exhausted, f_bavail will be 0.
2067 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2069 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2070 struct btrfs_super_block *disk_super = fs_info->super_copy;
2071 struct list_head *head = &fs_info->space_info;
2072 struct btrfs_space_info *found;
2074 u64 total_free_data = 0;
2075 u64 total_free_meta = 0;
2076 int bits = dentry->d_sb->s_blocksize_bits;
2077 __be32 *fsid = (__be32 *)fs_info->fsid;
2078 unsigned factor = 1;
2079 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2085 list_for_each_entry_rcu(found, head, list) {
2086 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2089 total_free_data += found->disk_total - found->disk_used;
2091 btrfs_account_ro_block_groups_free_space(found);
2093 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2094 if (!list_empty(&found->block_groups[i])) {
2096 case BTRFS_RAID_DUP:
2097 case BTRFS_RAID_RAID1:
2098 case BTRFS_RAID_RAID10:
2106 * Metadata in mixed block goup profiles are accounted in data
2108 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2109 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2112 total_free_meta += found->disk_total -
2116 total_used += found->disk_used;
2121 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2122 buf->f_blocks >>= bits;
2123 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2125 /* Account global block reserve as used, it's in logical size already */
2126 spin_lock(&block_rsv->lock);
2127 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2128 if (buf->f_bfree >= block_rsv->size >> bits)
2129 buf->f_bfree -= block_rsv->size >> bits;
2132 spin_unlock(&block_rsv->lock);
2134 buf->f_bavail = div_u64(total_free_data, factor);
2135 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2138 buf->f_bavail += div_u64(total_free_data, factor);
2139 buf->f_bavail = buf->f_bavail >> bits;
2142 * We calculate the remaining metadata space minus global reserve. If
2143 * this is (supposedly) smaller than zero, there's no space. But this
2144 * does not hold in practice, the exhausted state happens where's still
2145 * some positive delta. So we apply some guesswork and compare the
2146 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2148 * We probably cannot calculate the exact threshold value because this
2149 * depends on the internal reservations requested by various
2150 * operations, so some operations that consume a few metadata will
2151 * succeed even if the Avail is zero. But this is better than the other
2154 thresh = 4 * 1024 * 1024;
2156 if (!mixed && total_free_meta - thresh < block_rsv->size)
2159 buf->f_type = BTRFS_SUPER_MAGIC;
2160 buf->f_bsize = dentry->d_sb->s_blocksize;
2161 buf->f_namelen = BTRFS_NAME_LEN;
2163 /* We treat it as constant endianness (it doesn't matter _which_)
2164 because we want the fsid to come out the same whether mounted
2165 on a big-endian or little-endian host */
2166 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2167 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2168 /* Mask in the root object ID too, to disambiguate subvols */
2169 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2170 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2175 static void btrfs_kill_super(struct super_block *sb)
2177 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2178 kill_anon_super(sb);
2179 free_fs_info(fs_info);
2182 static struct file_system_type btrfs_fs_type = {
2183 .owner = THIS_MODULE,
2185 .mount = btrfs_mount,
2186 .kill_sb = btrfs_kill_super,
2187 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2189 MODULE_ALIAS_FS("btrfs");
2191 static int btrfs_control_open(struct inode *inode, struct file *file)
2194 * The control file's private_data is used to hold the
2195 * transaction when it is started and is used to keep
2196 * track of whether a transaction is already in progress.
2198 file->private_data = NULL;
2203 * used by btrfsctl to scan devices when no FS is mounted
2205 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2208 struct btrfs_ioctl_vol_args *vol;
2209 struct btrfs_fs_devices *fs_devices;
2212 if (!capable(CAP_SYS_ADMIN))
2215 vol = memdup_user((void __user *)arg, sizeof(*vol));
2217 return PTR_ERR(vol);
2220 case BTRFS_IOC_SCAN_DEV:
2221 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2222 &btrfs_fs_type, &fs_devices);
2224 case BTRFS_IOC_DEVICES_READY:
2225 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2226 &btrfs_fs_type, &fs_devices);
2229 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2231 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2232 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2240 static int btrfs_freeze(struct super_block *sb)
2242 struct btrfs_trans_handle *trans;
2243 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2244 struct btrfs_root *root = fs_info->tree_root;
2246 fs_info->fs_frozen = 1;
2248 * We don't need a barrier here, we'll wait for any transaction that
2249 * could be in progress on other threads (and do delayed iputs that
2250 * we want to avoid on a frozen filesystem), or do the commit
2253 trans = btrfs_attach_transaction_barrier(root);
2254 if (IS_ERR(trans)) {
2255 /* no transaction, don't bother */
2256 if (PTR_ERR(trans) == -ENOENT)
2258 return PTR_ERR(trans);
2260 return btrfs_commit_transaction(trans);
2263 static int btrfs_unfreeze(struct super_block *sb)
2265 btrfs_sb(sb)->fs_frozen = 0;
2269 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2271 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2272 struct btrfs_fs_devices *cur_devices;
2273 struct btrfs_device *dev, *first_dev = NULL;
2274 struct list_head *head;
2275 struct rcu_string *name;
2277 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2278 cur_devices = fs_info->fs_devices;
2279 while (cur_devices) {
2280 head = &cur_devices->devices;
2281 list_for_each_entry(dev, head, dev_list) {
2286 if (!first_dev || dev->devid < first_dev->devid)
2289 cur_devices = cur_devices->seed;
2294 name = rcu_dereference(first_dev->name);
2295 seq_escape(m, name->str, " \t\n\\");
2300 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2304 static const struct super_operations btrfs_super_ops = {
2305 .drop_inode = btrfs_drop_inode,
2306 .evict_inode = btrfs_evict_inode,
2307 .put_super = btrfs_put_super,
2308 .sync_fs = btrfs_sync_fs,
2309 .show_options = btrfs_show_options,
2310 .show_devname = btrfs_show_devname,
2311 .write_inode = btrfs_write_inode,
2312 .alloc_inode = btrfs_alloc_inode,
2313 .destroy_inode = btrfs_destroy_inode,
2314 .statfs = btrfs_statfs,
2315 .remount_fs = btrfs_remount,
2316 .freeze_fs = btrfs_freeze,
2317 .unfreeze_fs = btrfs_unfreeze,
2320 static const struct file_operations btrfs_ctl_fops = {
2321 .open = btrfs_control_open,
2322 .unlocked_ioctl = btrfs_control_ioctl,
2323 .compat_ioctl = btrfs_control_ioctl,
2324 .owner = THIS_MODULE,
2325 .llseek = noop_llseek,
2328 static struct miscdevice btrfs_misc = {
2329 .minor = BTRFS_MINOR,
2330 .name = "btrfs-control",
2331 .fops = &btrfs_ctl_fops
2334 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2335 MODULE_ALIAS("devname:btrfs-control");
2337 static int btrfs_interface_init(void)
2339 return misc_register(&btrfs_misc);
2342 static void btrfs_interface_exit(void)
2344 misc_deregister(&btrfs_misc);
2347 static void btrfs_print_mod_info(void)
2349 pr_info("Btrfs loaded, crc32c=%s"
2350 #ifdef CONFIG_BTRFS_DEBUG
2353 #ifdef CONFIG_BTRFS_ASSERT
2356 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2357 ", integrity-checker=on"
2360 btrfs_crc32c_impl());
2363 static int __init init_btrfs_fs(void)
2367 err = btrfs_hash_init();
2373 err = btrfs_init_sysfs();
2377 btrfs_init_compress();
2379 err = btrfs_init_cachep();
2383 err = extent_io_init();
2387 err = extent_map_init();
2389 goto free_extent_io;
2391 err = ordered_data_init();
2393 goto free_extent_map;
2395 err = btrfs_delayed_inode_init();
2397 goto free_ordered_data;
2399 err = btrfs_auto_defrag_init();
2401 goto free_delayed_inode;
2403 err = btrfs_delayed_ref_init();
2405 goto free_auto_defrag;
2407 err = btrfs_prelim_ref_init();
2409 goto free_delayed_ref;
2411 err = btrfs_end_io_wq_init();
2413 goto free_prelim_ref;
2415 err = btrfs_interface_init();
2417 goto free_end_io_wq;
2419 btrfs_init_lockdep();
2421 btrfs_print_mod_info();
2423 err = btrfs_run_sanity_tests();
2425 goto unregister_ioctl;
2427 err = register_filesystem(&btrfs_fs_type);
2429 goto unregister_ioctl;
2434 btrfs_interface_exit();
2436 btrfs_end_io_wq_exit();
2438 btrfs_prelim_ref_exit();
2440 btrfs_delayed_ref_exit();
2442 btrfs_auto_defrag_exit();
2444 btrfs_delayed_inode_exit();
2446 ordered_data_exit();
2452 btrfs_destroy_cachep();
2454 btrfs_exit_compress();
2461 static void __exit exit_btrfs_fs(void)
2463 btrfs_destroy_cachep();
2464 btrfs_delayed_ref_exit();
2465 btrfs_auto_defrag_exit();
2466 btrfs_delayed_inode_exit();
2467 btrfs_prelim_ref_exit();
2468 ordered_data_exit();
2471 btrfs_interface_exit();
2472 btrfs_end_io_wq_exit();
2473 unregister_filesystem(&btrfs_fs_type);
2475 btrfs_cleanup_fs_uuids();
2476 btrfs_exit_compress();
2480 late_initcall(init_btrfs_fs);
2481 module_exit(exit_btrfs_fs)
2483 MODULE_LICENSE("GPL");