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 static 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 static void save_error_info(struct btrfs_fs_info *fs_info)
103 * today we only save the error info into ram. Long term we'll
104 * also send it down to the disk
106 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
109 /* btrfs handle error by forcing the filesystem readonly */
110 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
112 struct super_block *sb = fs_info->sb;
114 if (sb->s_flags & MS_RDONLY)
117 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
118 sb->s_flags |= MS_RDONLY;
119 btrfs_info(fs_info, "forced readonly");
121 * Note that a running device replace operation is not
122 * canceled here although there is no way to update
123 * the progress. It would add the risk of a deadlock,
124 * therefore the canceling is ommited. The only penalty
125 * is that some I/O remains active until the procedure
126 * completes. The next time when the filesystem is
127 * mounted writeable again, the device replace
128 * operation continues.
135 * __btrfs_std_error decodes expected errors from the caller and
136 * invokes the approciate error response.
138 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
139 unsigned int line, int errno, const char *fmt, ...)
141 struct super_block *sb = fs_info->sb;
145 * Special case: if the error is EROFS, and we're already
146 * under MS_RDONLY, then it is safe here.
148 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
151 errstr = btrfs_decode_error(errno);
153 struct va_format vaf;
161 "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
162 sb->s_id, function, line, errno, errstr, &vaf);
165 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
166 sb->s_id, function, line, errno, errstr);
169 /* Don't go through full error handling during mount */
170 save_error_info(fs_info);
171 if (sb->s_flags & MS_BORN)
172 btrfs_handle_error(fs_info);
175 static const char * const logtypes[] = {
186 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
188 struct super_block *sb = fs_info->sb;
190 struct va_format vaf;
192 const char *type = logtypes[4];
197 kern_level = printk_get_level(fmt);
199 size_t size = printk_skip_level(fmt) - fmt;
200 memcpy(lvl, fmt, size);
203 type = logtypes[kern_level - '0'];
210 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
217 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
218 unsigned int line, int errno, const char *fmt, ...)
220 struct super_block *sb = fs_info->sb;
223 * Special case: if the error is EROFS, and we're already
224 * under MS_RDONLY, then it is safe here.
226 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
229 /* Don't go through full error handling during mount */
230 if (sb->s_flags & MS_BORN) {
231 save_error_info(fs_info);
232 btrfs_handle_error(fs_info);
238 * We only mark the transaction aborted and then set the file system read-only.
239 * This will prevent new transactions from starting or trying to join this
242 * This means that error recovery at the call site is limited to freeing
243 * any local memory allocations and passing the error code up without
244 * further cleanup. The transaction should complete as it normally would
245 * in the call path but will return -EIO.
247 * We'll complete the cleanup in btrfs_end_transaction and
248 * btrfs_commit_transaction.
250 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
251 struct btrfs_root *root, const char *function,
252 unsigned int line, int errno)
255 * Report first abort since mount
257 if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,
258 &root->fs_info->fs_state)) {
259 WARN(1, KERN_DEBUG "BTRFS: Transaction aborted (error %d)\n",
262 trans->aborted = errno;
263 /* Nothing used. The other threads that have joined this
264 * transaction may be able to continue. */
265 if (!trans->blocks_used && list_empty(&trans->new_bgs)) {
268 errstr = btrfs_decode_error(errno);
269 btrfs_warn(root->fs_info,
270 "%s:%d: Aborting unused transaction(%s).",
271 function, line, errstr);
274 ACCESS_ONCE(trans->transaction->aborted) = errno;
275 /* Wake up anybody who may be waiting on this transaction */
276 wake_up(&root->fs_info->transaction_wait);
277 wake_up(&root->fs_info->transaction_blocked_wait);
278 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
281 * __btrfs_panic decodes unexpected, fatal errors from the caller,
282 * issues an alert, and either panics or BUGs, depending on mount options.
284 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
285 unsigned int line, int errno, const char *fmt, ...)
287 char *s_id = "<unknown>";
289 struct va_format vaf = { .fmt = fmt };
293 s_id = fs_info->sb->s_id;
298 errstr = btrfs_decode_error(errno);
299 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
300 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
301 s_id, function, line, &vaf, errno, errstr);
303 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
304 function, line, &vaf, errno, errstr);
306 /* Caller calls BUG() */
309 static void btrfs_put_super(struct super_block *sb)
311 close_ctree(btrfs_sb(sb)->tree_root);
315 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
316 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
317 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
318 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
319 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
320 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
321 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
322 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
323 Opt_check_integrity, Opt_check_integrity_including_extent_data,
324 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
325 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
326 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
327 Opt_datasum, Opt_treelog, Opt_noinode_cache,
331 static 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_flushoncommit, "flushoncommit"},
357 {Opt_noflushoncommit, "noflushoncommit"},
358 {Opt_ratio, "metadata_ratio=%d"},
359 {Opt_discard, "discard"},
360 {Opt_nodiscard, "nodiscard"},
361 {Opt_space_cache, "space_cache"},
362 {Opt_clear_cache, "clear_cache"},
363 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
364 {Opt_enospc_debug, "enospc_debug"},
365 {Opt_noenospc_debug, "noenospc_debug"},
366 {Opt_subvolrootid, "subvolrootid=%d"},
367 {Opt_defrag, "autodefrag"},
368 {Opt_nodefrag, "noautodefrag"},
369 {Opt_inode_cache, "inode_cache"},
370 {Opt_noinode_cache, "noinode_cache"},
371 {Opt_no_space_cache, "nospace_cache"},
372 {Opt_recovery, "recovery"},
373 {Opt_skip_balance, "skip_balance"},
374 {Opt_check_integrity, "check_int"},
375 {Opt_check_integrity_including_extent_data, "check_int_data"},
376 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
377 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
378 {Opt_fatal_errors, "fatal_errors=%s"},
379 {Opt_commit_interval, "commit=%d"},
384 * Regular mount options parser. Everything that is needed only when
385 * reading in a new superblock is parsed here.
386 * XXX JDM: This needs to be cleaned up for remount.
388 int btrfs_parse_options(struct btrfs_root *root, char *options)
390 struct btrfs_fs_info *info = root->fs_info;
391 substring_t args[MAX_OPT_ARGS];
392 char *p, *num, *orig = NULL;
397 bool compress_force = false;
399 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
401 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
407 * strsep changes the string, duplicate it because parse_options
410 options = kstrdup(options, GFP_NOFS);
416 while ((p = strsep(&options, ",")) != NULL) {
421 token = match_token(p, tokens, args);
424 btrfs_info(root->fs_info, "allowing degraded mounts");
425 btrfs_set_opt(info->mount_opt, DEGRADED);
429 case Opt_subvolrootid:
432 * These are parsed by btrfs_parse_early_options
433 * and can be happily ignored here.
437 btrfs_set_and_info(root, NODATASUM,
438 "setting nodatasum");
441 if (btrfs_test_opt(root, NODATASUM)) {
442 if (btrfs_test_opt(root, NODATACOW))
443 btrfs_info(root->fs_info, "setting datasum, datacow enabled");
445 btrfs_info(root->fs_info, "setting datasum");
447 btrfs_clear_opt(info->mount_opt, NODATACOW);
448 btrfs_clear_opt(info->mount_opt, NODATASUM);
451 if (!btrfs_test_opt(root, NODATACOW)) {
452 if (!btrfs_test_opt(root, COMPRESS) ||
453 !btrfs_test_opt(root, FORCE_COMPRESS)) {
454 btrfs_info(root->fs_info,
455 "setting nodatacow, compression disabled");
457 btrfs_info(root->fs_info, "setting nodatacow");
460 btrfs_clear_opt(info->mount_opt, COMPRESS);
461 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
462 btrfs_set_opt(info->mount_opt, NODATACOW);
463 btrfs_set_opt(info->mount_opt, NODATASUM);
466 btrfs_clear_and_info(root, NODATACOW,
469 case Opt_compress_force:
470 case Opt_compress_force_type:
471 compress_force = true;
474 case Opt_compress_type:
475 if (token == Opt_compress ||
476 token == Opt_compress_force ||
477 strcmp(args[0].from, "zlib") == 0) {
478 compress_type = "zlib";
479 info->compress_type = BTRFS_COMPRESS_ZLIB;
480 btrfs_set_opt(info->mount_opt, COMPRESS);
481 btrfs_clear_opt(info->mount_opt, NODATACOW);
482 btrfs_clear_opt(info->mount_opt, NODATASUM);
483 } else if (strcmp(args[0].from, "lzo") == 0) {
484 compress_type = "lzo";
485 info->compress_type = BTRFS_COMPRESS_LZO;
486 btrfs_set_opt(info->mount_opt, COMPRESS);
487 btrfs_clear_opt(info->mount_opt, NODATACOW);
488 btrfs_clear_opt(info->mount_opt, NODATASUM);
489 btrfs_set_fs_incompat(info, COMPRESS_LZO);
490 } else if (strncmp(args[0].from, "no", 2) == 0) {
491 compress_type = "no";
492 btrfs_clear_opt(info->mount_opt, COMPRESS);
493 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
494 compress_force = false;
500 if (compress_force) {
501 btrfs_set_and_info(root, FORCE_COMPRESS,
502 "force %s compression",
505 if (!btrfs_test_opt(root, COMPRESS))
506 btrfs_info(root->fs_info,
507 "btrfs: use %s compression",
510 * If we remount from compress-force=xxx to
511 * compress=xxx, we need clear FORCE_COMPRESS
512 * flag, otherwise, there is no way for users
513 * to disable forcible compression separately.
515 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
519 btrfs_set_and_info(root, SSD,
520 "use ssd allocation scheme");
523 btrfs_set_and_info(root, SSD_SPREAD,
524 "use spread ssd allocation scheme");
525 btrfs_set_opt(info->mount_opt, SSD);
528 btrfs_set_and_info(root, NOSSD,
529 "not using ssd allocation scheme");
530 btrfs_clear_opt(info->mount_opt, SSD);
533 btrfs_clear_and_info(root, NOBARRIER,
534 "turning on barriers");
537 btrfs_set_and_info(root, NOBARRIER,
538 "turning off barriers");
540 case Opt_thread_pool:
541 ret = match_int(&args[0], &intarg);
544 } else if (intarg > 0) {
545 info->thread_pool_size = intarg;
552 num = match_strdup(&args[0]);
554 info->max_inline = memparse(num, NULL);
557 if (info->max_inline) {
558 info->max_inline = min_t(u64,
562 btrfs_info(root->fs_info, "max_inline at %llu",
569 case Opt_alloc_start:
570 num = match_strdup(&args[0]);
572 mutex_lock(&info->chunk_mutex);
573 info->alloc_start = memparse(num, NULL);
574 mutex_unlock(&info->chunk_mutex);
576 btrfs_info(root->fs_info, "allocations start at %llu",
584 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
585 root->fs_info->sb->s_flags |= MS_POSIXACL;
588 btrfs_err(root->fs_info,
589 "support for ACL not compiled in!");
594 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
597 btrfs_set_and_info(root, NOTREELOG,
598 "disabling tree log");
601 btrfs_clear_and_info(root, NOTREELOG,
602 "enabling tree log");
604 case Opt_flushoncommit:
605 btrfs_set_and_info(root, FLUSHONCOMMIT,
606 "turning on flush-on-commit");
608 case Opt_noflushoncommit:
609 btrfs_clear_and_info(root, FLUSHONCOMMIT,
610 "turning off flush-on-commit");
613 ret = match_int(&args[0], &intarg);
616 } else if (intarg >= 0) {
617 info->metadata_ratio = intarg;
618 btrfs_info(root->fs_info, "metadata ratio %d",
619 info->metadata_ratio);
626 btrfs_set_and_info(root, DISCARD,
627 "turning on discard");
630 btrfs_clear_and_info(root, DISCARD,
631 "turning off discard");
633 case Opt_space_cache:
634 btrfs_set_and_info(root, SPACE_CACHE,
635 "enabling disk space caching");
637 case Opt_rescan_uuid_tree:
638 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
640 case Opt_no_space_cache:
641 btrfs_clear_and_info(root, SPACE_CACHE,
642 "disabling disk space caching");
644 case Opt_inode_cache:
645 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
646 "enabling inode map caching");
648 case Opt_noinode_cache:
649 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
650 "disabling inode map caching");
652 case Opt_clear_cache:
653 btrfs_set_and_info(root, CLEAR_CACHE,
654 "force clearing of disk cache");
656 case Opt_user_subvol_rm_allowed:
657 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
659 case Opt_enospc_debug:
660 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
662 case Opt_noenospc_debug:
663 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
666 btrfs_set_and_info(root, AUTO_DEFRAG,
667 "enabling auto defrag");
670 btrfs_clear_and_info(root, AUTO_DEFRAG,
671 "disabling auto defrag");
674 btrfs_info(root->fs_info, "enabling auto recovery");
675 btrfs_set_opt(info->mount_opt, RECOVERY);
677 case Opt_skip_balance:
678 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
680 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
681 case Opt_check_integrity_including_extent_data:
682 btrfs_info(root->fs_info,
683 "enabling check integrity including extent data");
684 btrfs_set_opt(info->mount_opt,
685 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
686 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
688 case Opt_check_integrity:
689 btrfs_info(root->fs_info, "enabling check integrity");
690 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
692 case Opt_check_integrity_print_mask:
693 ret = match_int(&args[0], &intarg);
696 } else if (intarg >= 0) {
697 info->check_integrity_print_mask = intarg;
698 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
699 info->check_integrity_print_mask);
706 case Opt_check_integrity_including_extent_data:
707 case Opt_check_integrity:
708 case Opt_check_integrity_print_mask:
709 btrfs_err(root->fs_info,
710 "support for check_integrity* not compiled in!");
714 case Opt_fatal_errors:
715 if (strcmp(args[0].from, "panic") == 0)
716 btrfs_set_opt(info->mount_opt,
717 PANIC_ON_FATAL_ERROR);
718 else if (strcmp(args[0].from, "bug") == 0)
719 btrfs_clear_opt(info->mount_opt,
720 PANIC_ON_FATAL_ERROR);
726 case Opt_commit_interval:
728 ret = match_int(&args[0], &intarg);
730 btrfs_err(root->fs_info, "invalid commit interval");
736 btrfs_warn(root->fs_info, "excessive commit interval %d",
739 info->commit_interval = intarg;
741 btrfs_info(root->fs_info, "using default commit interval %ds",
742 BTRFS_DEFAULT_COMMIT_INTERVAL);
743 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
747 btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
755 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
756 btrfs_info(root->fs_info, "disk space caching is enabled");
762 * Parse mount options that are required early in the mount process.
764 * All other options will be parsed on much later in the mount process and
765 * only when we need to allocate a new super block.
767 static int btrfs_parse_early_options(const char *options, fmode_t flags,
768 void *holder, char **subvol_name, u64 *subvol_objectid,
769 struct btrfs_fs_devices **fs_devices)
771 substring_t args[MAX_OPT_ARGS];
772 char *device_name, *opts, *orig, *p;
780 * strsep changes the string, duplicate it because parse_options
783 opts = kstrdup(options, GFP_KERNEL);
788 while ((p = strsep(&opts, ",")) != NULL) {
793 token = match_token(p, tokens, args);
797 *subvol_name = match_strdup(&args[0]);
804 num = match_strdup(&args[0]);
806 *subvol_objectid = memparse(num, NULL);
808 /* we want the original fs_tree */
809 if (!*subvol_objectid)
811 BTRFS_FS_TREE_OBJECTID;
817 case Opt_subvolrootid:
819 "BTRFS: 'subvolrootid' mount option is deprecated and has "
823 device_name = match_strdup(&args[0]);
828 error = btrfs_scan_one_device(device_name,
829 flags, holder, fs_devices);
844 static struct dentry *get_default_root(struct super_block *sb,
847 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
848 struct btrfs_root *root = fs_info->tree_root;
849 struct btrfs_root *new_root;
850 struct btrfs_dir_item *di;
851 struct btrfs_path *path;
852 struct btrfs_key location;
858 * We have a specific subvol we want to mount, just setup location and
859 * go look up the root.
861 if (subvol_objectid) {
862 location.objectid = subvol_objectid;
863 location.type = BTRFS_ROOT_ITEM_KEY;
864 location.offset = (u64)-1;
868 path = btrfs_alloc_path();
870 return ERR_PTR(-ENOMEM);
871 path->leave_spinning = 1;
874 * Find the "default" dir item which points to the root item that we
875 * will mount by default if we haven't been given a specific subvolume
878 dir_id = btrfs_super_root_dir(fs_info->super_copy);
879 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
881 btrfs_free_path(path);
886 * Ok the default dir item isn't there. This is weird since
887 * it's always been there, but don't freak out, just try and
888 * mount to root most subvolume.
890 btrfs_free_path(path);
891 dir_id = BTRFS_FIRST_FREE_OBJECTID;
892 new_root = fs_info->fs_root;
896 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
897 btrfs_free_path(path);
900 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
901 if (IS_ERR(new_root))
902 return ERR_CAST(new_root);
904 dir_id = btrfs_root_dirid(&new_root->root_item);
906 location.objectid = dir_id;
907 location.type = BTRFS_INODE_ITEM_KEY;
910 inode = btrfs_iget(sb, &location, new_root, &new);
912 return ERR_CAST(inode);
915 * If we're just mounting the root most subvol put the inode and return
916 * a reference to the dentry. We will have already gotten a reference
917 * to the inode in btrfs_fill_super so we're good to go.
919 if (!new && sb->s_root->d_inode == inode) {
921 return dget(sb->s_root);
924 return d_obtain_root(inode);
927 static int btrfs_fill_super(struct super_block *sb,
928 struct btrfs_fs_devices *fs_devices,
929 void *data, int silent)
932 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
933 struct btrfs_key key;
936 sb->s_maxbytes = MAX_LFS_FILESIZE;
937 sb->s_magic = BTRFS_SUPER_MAGIC;
938 sb->s_op = &btrfs_super_ops;
939 sb->s_d_op = &btrfs_dentry_operations;
940 sb->s_export_op = &btrfs_export_ops;
941 sb->s_xattr = btrfs_xattr_handlers;
943 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
944 sb->s_flags |= MS_POSIXACL;
946 sb->s_flags |= MS_I_VERSION;
947 err = open_ctree(sb, fs_devices, (char *)data);
949 printk(KERN_ERR "BTRFS: open_ctree failed\n");
953 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
954 key.type = BTRFS_INODE_ITEM_KEY;
956 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
958 err = PTR_ERR(inode);
962 sb->s_root = d_make_root(inode);
968 save_mount_options(sb, data);
969 cleancache_init_fs(sb);
970 sb->s_flags |= MS_ACTIVE;
974 close_ctree(fs_info->tree_root);
978 int btrfs_sync_fs(struct super_block *sb, int wait)
980 struct btrfs_trans_handle *trans;
981 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
982 struct btrfs_root *root = fs_info->tree_root;
984 trace_btrfs_sync_fs(wait);
987 filemap_flush(fs_info->btree_inode->i_mapping);
991 btrfs_wait_ordered_roots(fs_info, -1);
993 trans = btrfs_attach_transaction_barrier(root);
995 /* no transaction, don't bother */
996 if (PTR_ERR(trans) == -ENOENT) {
998 * Exit unless we have some pending changes
999 * that need to go through commit
1001 if (fs_info->pending_changes == 0)
1004 * A non-blocking test if the fs is frozen. We must not
1005 * start a new transaction here otherwise a deadlock
1006 * happens. The pending operations are delayed to the
1007 * next commit after thawing.
1009 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1010 __sb_end_write(sb, SB_FREEZE_WRITE);
1013 trans = btrfs_start_transaction(root, 0);
1016 return PTR_ERR(trans);
1018 return btrfs_commit_transaction(trans, root);
1021 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1023 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1024 struct btrfs_root *root = info->tree_root;
1025 char *compress_type;
1027 if (btrfs_test_opt(root, DEGRADED))
1028 seq_puts(seq, ",degraded");
1029 if (btrfs_test_opt(root, NODATASUM))
1030 seq_puts(seq, ",nodatasum");
1031 if (btrfs_test_opt(root, NODATACOW))
1032 seq_puts(seq, ",nodatacow");
1033 if (btrfs_test_opt(root, NOBARRIER))
1034 seq_puts(seq, ",nobarrier");
1035 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1036 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1037 if (info->alloc_start != 0)
1038 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1039 if (info->thread_pool_size != min_t(unsigned long,
1040 num_online_cpus() + 2, 8))
1041 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1042 if (btrfs_test_opt(root, COMPRESS)) {
1043 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1044 compress_type = "zlib";
1046 compress_type = "lzo";
1047 if (btrfs_test_opt(root, FORCE_COMPRESS))
1048 seq_printf(seq, ",compress-force=%s", compress_type);
1050 seq_printf(seq, ",compress=%s", compress_type);
1052 if (btrfs_test_opt(root, NOSSD))
1053 seq_puts(seq, ",nossd");
1054 if (btrfs_test_opt(root, SSD_SPREAD))
1055 seq_puts(seq, ",ssd_spread");
1056 else if (btrfs_test_opt(root, SSD))
1057 seq_puts(seq, ",ssd");
1058 if (btrfs_test_opt(root, NOTREELOG))
1059 seq_puts(seq, ",notreelog");
1060 if (btrfs_test_opt(root, FLUSHONCOMMIT))
1061 seq_puts(seq, ",flushoncommit");
1062 if (btrfs_test_opt(root, DISCARD))
1063 seq_puts(seq, ",discard");
1064 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1065 seq_puts(seq, ",noacl");
1066 if (btrfs_test_opt(root, SPACE_CACHE))
1067 seq_puts(seq, ",space_cache");
1069 seq_puts(seq, ",nospace_cache");
1070 if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1071 seq_puts(seq, ",rescan_uuid_tree");
1072 if (btrfs_test_opt(root, CLEAR_CACHE))
1073 seq_puts(seq, ",clear_cache");
1074 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1075 seq_puts(seq, ",user_subvol_rm_allowed");
1076 if (btrfs_test_opt(root, ENOSPC_DEBUG))
1077 seq_puts(seq, ",enospc_debug");
1078 if (btrfs_test_opt(root, AUTO_DEFRAG))
1079 seq_puts(seq, ",autodefrag");
1080 if (btrfs_test_opt(root, INODE_MAP_CACHE))
1081 seq_puts(seq, ",inode_cache");
1082 if (btrfs_test_opt(root, SKIP_BALANCE))
1083 seq_puts(seq, ",skip_balance");
1084 if (btrfs_test_opt(root, RECOVERY))
1085 seq_puts(seq, ",recovery");
1086 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1087 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1088 seq_puts(seq, ",check_int_data");
1089 else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1090 seq_puts(seq, ",check_int");
1091 if (info->check_integrity_print_mask)
1092 seq_printf(seq, ",check_int_print_mask=%d",
1093 info->check_integrity_print_mask);
1095 if (info->metadata_ratio)
1096 seq_printf(seq, ",metadata_ratio=%d",
1097 info->metadata_ratio);
1098 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1099 seq_puts(seq, ",fatal_errors=panic");
1100 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1101 seq_printf(seq, ",commit=%d", info->commit_interval);
1105 static int btrfs_test_super(struct super_block *s, void *data)
1107 struct btrfs_fs_info *p = data;
1108 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1110 return fs_info->fs_devices == p->fs_devices;
1113 static int btrfs_set_super(struct super_block *s, void *data)
1115 int err = set_anon_super(s, data);
1117 s->s_fs_info = data;
1122 * subvolumes are identified by ino 256
1124 static inline int is_subvolume_inode(struct inode *inode)
1126 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1132 * This will strip out the subvol=%s argument for an argument string and add
1133 * subvolid=0 to make sure we get the actual tree root for path walking to the
1136 static char *setup_root_args(char *args)
1138 unsigned len = strlen(args) + 2 + 1;
1139 char *src, *dst, *buf;
1142 * We need the same args as before, but with this substitution:
1143 * s!subvol=[^,]+!subvolid=0!
1145 * Since the replacement string is up to 2 bytes longer than the
1146 * original, allocate strlen(args) + 2 + 1 bytes.
1149 src = strstr(args, "subvol=");
1150 /* This shouldn't happen, but just in case.. */
1154 buf = dst = kmalloc(len, GFP_NOFS);
1159 * If the subvol= arg is not at the start of the string,
1160 * copy whatever precedes it into buf.
1165 dst += strlen(args);
1168 strcpy(dst, "subvolid=0");
1169 dst += strlen("subvolid=0");
1172 * If there is a "," after the original subvol=... string,
1173 * copy that suffix into our buffer. Otherwise, we're done.
1175 src = strchr(src, ',');
1182 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1183 const char *device_name, char *data)
1185 struct dentry *root;
1186 struct vfsmount *mnt;
1189 newargs = setup_root_args(data);
1191 return ERR_PTR(-ENOMEM);
1192 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1195 if (PTR_RET(mnt) == -EBUSY) {
1196 if (flags & MS_RDONLY) {
1197 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY, device_name,
1201 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY, device_name,
1205 return ERR_CAST(mnt);
1208 r = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1210 /* FIXME: release vfsmount mnt ??*/
1220 return ERR_CAST(mnt);
1222 root = mount_subtree(mnt, subvol_name);
1224 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1225 struct super_block *s = root->d_sb;
1227 root = ERR_PTR(-EINVAL);
1228 deactivate_locked_super(s);
1229 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1236 static int parse_security_options(char *orig_opts,
1237 struct security_mnt_opts *sec_opts)
1239 char *secdata = NULL;
1242 secdata = alloc_secdata();
1245 ret = security_sb_copy_data(orig_opts, secdata);
1247 free_secdata(secdata);
1250 ret = security_sb_parse_opts_str(secdata, sec_opts);
1251 free_secdata(secdata);
1255 static int setup_security_options(struct btrfs_fs_info *fs_info,
1256 struct super_block *sb,
1257 struct security_mnt_opts *sec_opts)
1262 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1265 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1269 #ifdef CONFIG_SECURITY
1270 if (!fs_info->security_opts.num_mnt_opts) {
1271 /* first time security setup, copy sec_opts to fs_info */
1272 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1275 * Since SELinux(the only one supports security_mnt_opts) does
1276 * NOT support changing context during remount/mount same sb,
1277 * This must be the same or part of the same security options,
1280 security_free_mnt_opts(sec_opts);
1287 * Find a superblock for the given device / mount point.
1289 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1290 * for multiple device setup. Make sure to keep it in sync.
1292 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1293 const char *device_name, void *data)
1295 struct block_device *bdev = NULL;
1296 struct super_block *s;
1297 struct dentry *root;
1298 struct btrfs_fs_devices *fs_devices = NULL;
1299 struct btrfs_fs_info *fs_info = NULL;
1300 struct security_mnt_opts new_sec_opts;
1301 fmode_t mode = FMODE_READ;
1302 char *subvol_name = NULL;
1303 u64 subvol_objectid = 0;
1306 if (!(flags & MS_RDONLY))
1307 mode |= FMODE_WRITE;
1309 error = btrfs_parse_early_options(data, mode, fs_type,
1310 &subvol_name, &subvol_objectid,
1314 return ERR_PTR(error);
1318 root = mount_subvol(subvol_name, flags, device_name, data);
1323 security_init_mnt_opts(&new_sec_opts);
1325 error = parse_security_options(data, &new_sec_opts);
1327 return ERR_PTR(error);
1330 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1332 goto error_sec_opts;
1335 * Setup a dummy root and fs_info for test/set super. This is because
1336 * we don't actually fill this stuff out until open_ctree, but we need
1337 * it for searching for existing supers, so this lets us do that and
1338 * then open_ctree will properly initialize everything later.
1340 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1343 goto error_sec_opts;
1346 fs_info->fs_devices = fs_devices;
1348 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1349 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1350 security_init_mnt_opts(&fs_info->security_opts);
1351 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1356 error = btrfs_open_devices(fs_devices, mode, fs_type);
1360 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1362 goto error_close_devices;
1365 bdev = fs_devices->latest_bdev;
1366 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1370 goto error_close_devices;
1374 btrfs_close_devices(fs_devices);
1375 free_fs_info(fs_info);
1376 if ((flags ^ s->s_flags) & MS_RDONLY)
1379 char b[BDEVNAME_SIZE];
1381 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1382 btrfs_sb(s)->bdev_holder = fs_type;
1383 error = btrfs_fill_super(s, fs_devices, data,
1384 flags & MS_SILENT ? 1 : 0);
1387 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1389 deactivate_locked_super(s);
1390 error = PTR_ERR(root);
1391 goto error_sec_opts;
1394 fs_info = btrfs_sb(s);
1395 error = setup_security_options(fs_info, s, &new_sec_opts);
1398 deactivate_locked_super(s);
1399 goto error_sec_opts;
1404 error_close_devices:
1405 btrfs_close_devices(fs_devices);
1407 free_fs_info(fs_info);
1409 security_free_mnt_opts(&new_sec_opts);
1410 return ERR_PTR(error);
1413 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1414 int new_pool_size, int old_pool_size)
1416 if (new_pool_size == old_pool_size)
1419 fs_info->thread_pool_size = new_pool_size;
1421 btrfs_info(fs_info, "resize thread pool %d -> %d",
1422 old_pool_size, new_pool_size);
1424 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1425 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1426 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1427 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1428 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1429 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1430 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1432 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1433 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1434 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1435 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1436 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1440 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1442 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1445 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1446 unsigned long old_opts, int flags)
1448 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1449 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1450 (flags & MS_RDONLY))) {
1451 /* wait for any defraggers to finish */
1452 wait_event(fs_info->transaction_wait,
1453 (atomic_read(&fs_info->defrag_running) == 0));
1454 if (flags & MS_RDONLY)
1455 sync_filesystem(fs_info->sb);
1459 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1460 unsigned long old_opts)
1463 * We need cleanup all defragable inodes if the autodefragment is
1464 * close or the fs is R/O.
1466 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1467 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1468 (fs_info->sb->s_flags & MS_RDONLY))) {
1469 btrfs_cleanup_defrag_inodes(fs_info);
1472 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1475 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1477 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1478 struct btrfs_root *root = fs_info->tree_root;
1479 unsigned old_flags = sb->s_flags;
1480 unsigned long old_opts = fs_info->mount_opt;
1481 unsigned long old_compress_type = fs_info->compress_type;
1482 u64 old_max_inline = fs_info->max_inline;
1483 u64 old_alloc_start = fs_info->alloc_start;
1484 int old_thread_pool_size = fs_info->thread_pool_size;
1485 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1488 sync_filesystem(sb);
1489 btrfs_remount_prepare(fs_info);
1492 struct security_mnt_opts new_sec_opts;
1494 security_init_mnt_opts(&new_sec_opts);
1495 ret = parse_security_options(data, &new_sec_opts);
1498 ret = setup_security_options(fs_info, sb,
1501 security_free_mnt_opts(&new_sec_opts);
1506 ret = btrfs_parse_options(root, data);
1512 btrfs_remount_begin(fs_info, old_opts, *flags);
1513 btrfs_resize_thread_pool(fs_info,
1514 fs_info->thread_pool_size, old_thread_pool_size);
1516 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1519 if (*flags & MS_RDONLY) {
1521 * this also happens on 'umount -rf' or on shutdown, when
1522 * the filesystem is busy.
1524 cancel_work_sync(&fs_info->async_reclaim_work);
1526 /* wait for the uuid_scan task to finish */
1527 down(&fs_info->uuid_tree_rescan_sem);
1528 /* avoid complains from lockdep et al. */
1529 up(&fs_info->uuid_tree_rescan_sem);
1531 sb->s_flags |= MS_RDONLY;
1533 btrfs_dev_replace_suspend_for_unmount(fs_info);
1534 btrfs_scrub_cancel(fs_info);
1535 btrfs_pause_balance(fs_info);
1537 ret = btrfs_commit_super(root);
1541 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1543 "Remounting read-write after error is not allowed");
1547 if (fs_info->fs_devices->rw_devices == 0) {
1552 if (fs_info->fs_devices->missing_devices >
1553 fs_info->num_tolerated_disk_barrier_failures &&
1554 !(*flags & MS_RDONLY)) {
1556 "too many missing devices, writeable remount is not allowed");
1561 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1566 ret = btrfs_cleanup_fs_roots(fs_info);
1570 /* recover relocation */
1571 mutex_lock(&fs_info->cleaner_mutex);
1572 ret = btrfs_recover_relocation(root);
1573 mutex_unlock(&fs_info->cleaner_mutex);
1577 ret = btrfs_resume_balance_async(fs_info);
1581 ret = btrfs_resume_dev_replace_async(fs_info);
1583 btrfs_warn(fs_info, "failed to resume dev_replace");
1587 if (!fs_info->uuid_root) {
1588 btrfs_info(fs_info, "creating UUID tree");
1589 ret = btrfs_create_uuid_tree(fs_info);
1591 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1595 sb->s_flags &= ~MS_RDONLY;
1598 wake_up_process(fs_info->transaction_kthread);
1599 btrfs_remount_cleanup(fs_info, old_opts);
1603 /* We've hit an error - don't reset MS_RDONLY */
1604 if (sb->s_flags & MS_RDONLY)
1605 old_flags |= MS_RDONLY;
1606 sb->s_flags = old_flags;
1607 fs_info->mount_opt = old_opts;
1608 fs_info->compress_type = old_compress_type;
1609 fs_info->max_inline = old_max_inline;
1610 mutex_lock(&fs_info->chunk_mutex);
1611 fs_info->alloc_start = old_alloc_start;
1612 mutex_unlock(&fs_info->chunk_mutex);
1613 btrfs_resize_thread_pool(fs_info,
1614 old_thread_pool_size, fs_info->thread_pool_size);
1615 fs_info->metadata_ratio = old_metadata_ratio;
1616 btrfs_remount_cleanup(fs_info, old_opts);
1620 /* Used to sort the devices by max_avail(descending sort) */
1621 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1622 const void *dev_info2)
1624 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1625 ((struct btrfs_device_info *)dev_info2)->max_avail)
1627 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1628 ((struct btrfs_device_info *)dev_info2)->max_avail)
1635 * sort the devices by max_avail, in which max free extent size of each device
1636 * is stored.(Descending Sort)
1638 static inline void btrfs_descending_sort_devices(
1639 struct btrfs_device_info *devices,
1642 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1643 btrfs_cmp_device_free_bytes, NULL);
1647 * The helper to calc the free space on the devices that can be used to store
1650 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1652 struct btrfs_fs_info *fs_info = root->fs_info;
1653 struct btrfs_device_info *devices_info;
1654 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1655 struct btrfs_device *device;
1660 u64 min_stripe_size;
1661 int min_stripes = 1, num_stripes = 1;
1662 int i = 0, nr_devices;
1666 * We aren't under the device list lock, so this is racey-ish, but good
1667 * enough for our purposes.
1669 nr_devices = fs_info->fs_devices->open_devices;
1672 nr_devices = fs_info->fs_devices->open_devices;
1680 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1685 /* calc min stripe number for data space alloction */
1686 type = btrfs_get_alloc_profile(root, 1);
1687 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1689 num_stripes = nr_devices;
1690 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1693 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1698 if (type & BTRFS_BLOCK_GROUP_DUP)
1699 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1701 min_stripe_size = BTRFS_STRIPE_LEN;
1703 if (fs_info->alloc_start)
1704 mutex_lock(&fs_devices->device_list_mutex);
1706 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1707 if (!device->in_fs_metadata || !device->bdev ||
1708 device->is_tgtdev_for_dev_replace)
1711 if (i >= nr_devices)
1714 avail_space = device->total_bytes - device->bytes_used;
1716 /* align with stripe_len */
1717 do_div(avail_space, BTRFS_STRIPE_LEN);
1718 avail_space *= BTRFS_STRIPE_LEN;
1721 * In order to avoid overwritting the superblock on the drive,
1722 * btrfs starts at an offset of at least 1MB when doing chunk
1725 skip_space = 1024 * 1024;
1727 /* user can set the offset in fs_info->alloc_start. */
1728 if (fs_info->alloc_start &&
1729 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1730 device->total_bytes) {
1732 skip_space = max(fs_info->alloc_start, skip_space);
1735 * btrfs can not use the free space in
1736 * [0, skip_space - 1], we must subtract it from the
1737 * total. In order to implement it, we account the used
1738 * space in this range first.
1740 ret = btrfs_account_dev_extents_size(device, 0,
1744 kfree(devices_info);
1745 mutex_unlock(&fs_devices->device_list_mutex);
1751 /* calc the free space in [0, skip_space - 1] */
1752 skip_space -= used_space;
1756 * we can use the free space in [0, skip_space - 1], subtract
1757 * it from the total.
1759 if (avail_space && avail_space >= skip_space)
1760 avail_space -= skip_space;
1764 if (avail_space < min_stripe_size)
1767 devices_info[i].dev = device;
1768 devices_info[i].max_avail = avail_space;
1773 if (fs_info->alloc_start)
1774 mutex_unlock(&fs_devices->device_list_mutex);
1778 btrfs_descending_sort_devices(devices_info, nr_devices);
1782 while (nr_devices >= min_stripes) {
1783 if (num_stripes > nr_devices)
1784 num_stripes = nr_devices;
1786 if (devices_info[i].max_avail >= min_stripe_size) {
1790 avail_space += devices_info[i].max_avail * num_stripes;
1791 alloc_size = devices_info[i].max_avail;
1792 for (j = i + 1 - num_stripes; j <= i; j++)
1793 devices_info[j].max_avail -= alloc_size;
1799 kfree(devices_info);
1800 *free_bytes = avail_space;
1805 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1807 * If there's a redundant raid level at DATA block groups, use the respective
1808 * multiplier to scale the sizes.
1810 * Unused device space usage is based on simulating the chunk allocator
1811 * algorithm that respects the device sizes, order of allocations and the
1812 * 'alloc_start' value, this is a close approximation of the actual use but
1813 * there are other factors that may change the result (like a new metadata
1816 * FIXME: not accurate for mixed block groups, total and free/used are ok,
1817 * available appears slightly larger.
1819 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1821 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1822 struct btrfs_super_block *disk_super = fs_info->super_copy;
1823 struct list_head *head = &fs_info->space_info;
1824 struct btrfs_space_info *found;
1826 u64 total_free_data = 0;
1827 int bits = dentry->d_sb->s_blocksize_bits;
1828 __be32 *fsid = (__be32 *)fs_info->fsid;
1829 unsigned factor = 1;
1830 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1834 * holding chunk_muext to avoid allocating new chunks, holding
1835 * device_list_mutex to avoid the device being removed
1838 list_for_each_entry_rcu(found, head, list) {
1839 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1842 total_free_data += found->disk_total - found->disk_used;
1844 btrfs_account_ro_block_groups_free_space(found);
1846 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1847 if (!list_empty(&found->block_groups[i])) {
1849 case BTRFS_RAID_DUP:
1850 case BTRFS_RAID_RAID1:
1851 case BTRFS_RAID_RAID10:
1858 total_used += found->disk_used;
1863 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1864 buf->f_blocks >>= bits;
1865 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1867 /* Account global block reserve as used, it's in logical size already */
1868 spin_lock(&block_rsv->lock);
1869 buf->f_bfree -= block_rsv->size >> bits;
1870 spin_unlock(&block_rsv->lock);
1872 buf->f_bavail = div_u64(total_free_data, factor);
1873 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1876 buf->f_bavail += div_u64(total_free_data, factor);
1877 buf->f_bavail = buf->f_bavail >> bits;
1879 buf->f_type = BTRFS_SUPER_MAGIC;
1880 buf->f_bsize = dentry->d_sb->s_blocksize;
1881 buf->f_namelen = BTRFS_NAME_LEN;
1883 /* We treat it as constant endianness (it doesn't matter _which_)
1884 because we want the fsid to come out the same whether mounted
1885 on a big-endian or little-endian host */
1886 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1887 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1888 /* Mask in the root object ID too, to disambiguate subvols */
1889 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1890 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1895 static void btrfs_kill_super(struct super_block *sb)
1897 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1898 kill_anon_super(sb);
1899 free_fs_info(fs_info);
1902 static struct file_system_type btrfs_fs_type = {
1903 .owner = THIS_MODULE,
1905 .mount = btrfs_mount,
1906 .kill_sb = btrfs_kill_super,
1907 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
1909 MODULE_ALIAS_FS("btrfs");
1912 * used by btrfsctl to scan devices when no FS is mounted
1914 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1917 struct btrfs_ioctl_vol_args *vol;
1918 struct btrfs_fs_devices *fs_devices;
1921 if (!capable(CAP_SYS_ADMIN))
1924 vol = memdup_user((void __user *)arg, sizeof(*vol));
1926 return PTR_ERR(vol);
1929 case BTRFS_IOC_SCAN_DEV:
1930 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1931 &btrfs_fs_type, &fs_devices);
1933 case BTRFS_IOC_DEVICES_READY:
1934 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1935 &btrfs_fs_type, &fs_devices);
1938 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1946 static int btrfs_freeze(struct super_block *sb)
1948 struct btrfs_trans_handle *trans;
1949 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1951 trans = btrfs_attach_transaction_barrier(root);
1952 if (IS_ERR(trans)) {
1953 /* no transaction, don't bother */
1954 if (PTR_ERR(trans) == -ENOENT)
1956 return PTR_ERR(trans);
1958 return btrfs_commit_transaction(trans, root);
1961 static int btrfs_unfreeze(struct super_block *sb)
1966 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1968 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1969 struct btrfs_fs_devices *cur_devices;
1970 struct btrfs_device *dev, *first_dev = NULL;
1971 struct list_head *head;
1972 struct rcu_string *name;
1974 mutex_lock(&fs_info->fs_devices->device_list_mutex);
1975 cur_devices = fs_info->fs_devices;
1976 while (cur_devices) {
1977 head = &cur_devices->devices;
1978 list_for_each_entry(dev, head, dev_list) {
1983 if (!first_dev || dev->devid < first_dev->devid)
1986 cur_devices = cur_devices->seed;
1991 name = rcu_dereference(first_dev->name);
1992 seq_escape(m, name->str, " \t\n\\");
1997 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2001 static const struct super_operations btrfs_super_ops = {
2002 .drop_inode = btrfs_drop_inode,
2003 .evict_inode = btrfs_evict_inode,
2004 .put_super = btrfs_put_super,
2005 .sync_fs = btrfs_sync_fs,
2006 .show_options = btrfs_show_options,
2007 .show_devname = btrfs_show_devname,
2008 .write_inode = btrfs_write_inode,
2009 .alloc_inode = btrfs_alloc_inode,
2010 .destroy_inode = btrfs_destroy_inode,
2011 .statfs = btrfs_statfs,
2012 .remount_fs = btrfs_remount,
2013 .freeze_fs = btrfs_freeze,
2014 .unfreeze_fs = btrfs_unfreeze,
2017 static const struct file_operations btrfs_ctl_fops = {
2018 .unlocked_ioctl = btrfs_control_ioctl,
2019 .compat_ioctl = btrfs_control_ioctl,
2020 .owner = THIS_MODULE,
2021 .llseek = noop_llseek,
2024 static struct miscdevice btrfs_misc = {
2025 .minor = BTRFS_MINOR,
2026 .name = "btrfs-control",
2027 .fops = &btrfs_ctl_fops
2030 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2031 MODULE_ALIAS("devname:btrfs-control");
2033 static int btrfs_interface_init(void)
2035 return misc_register(&btrfs_misc);
2038 static void btrfs_interface_exit(void)
2040 if (misc_deregister(&btrfs_misc) < 0)
2041 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
2044 static void btrfs_print_info(void)
2046 printk(KERN_INFO "Btrfs loaded"
2047 #ifdef CONFIG_BTRFS_DEBUG
2050 #ifdef CONFIG_BTRFS_ASSERT
2053 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2054 ", integrity-checker=on"
2059 static int btrfs_run_sanity_tests(void)
2063 ret = btrfs_init_test_fs();
2067 ret = btrfs_test_free_space_cache();
2070 ret = btrfs_test_extent_buffer_operations();
2073 ret = btrfs_test_extent_io();
2076 ret = btrfs_test_inodes();
2079 ret = btrfs_test_qgroups();
2081 btrfs_destroy_test_fs();
2085 static int __init init_btrfs_fs(void)
2089 err = btrfs_hash_init();
2095 err = btrfs_init_sysfs();
2099 btrfs_init_compress();
2101 err = btrfs_init_cachep();
2105 err = extent_io_init();
2109 err = extent_map_init();
2111 goto free_extent_io;
2113 err = ordered_data_init();
2115 goto free_extent_map;
2117 err = btrfs_delayed_inode_init();
2119 goto free_ordered_data;
2121 err = btrfs_auto_defrag_init();
2123 goto free_delayed_inode;
2125 err = btrfs_delayed_ref_init();
2127 goto free_auto_defrag;
2129 err = btrfs_prelim_ref_init();
2131 goto free_delayed_ref;
2133 err = btrfs_end_io_wq_init();
2135 goto free_prelim_ref;
2137 err = btrfs_interface_init();
2139 goto free_end_io_wq;
2141 btrfs_init_lockdep();
2145 err = btrfs_run_sanity_tests();
2147 goto unregister_ioctl;
2149 err = register_filesystem(&btrfs_fs_type);
2151 goto unregister_ioctl;
2156 btrfs_interface_exit();
2158 btrfs_end_io_wq_exit();
2160 btrfs_prelim_ref_exit();
2162 btrfs_delayed_ref_exit();
2164 btrfs_auto_defrag_exit();
2166 btrfs_delayed_inode_exit();
2168 ordered_data_exit();
2174 btrfs_destroy_cachep();
2176 btrfs_exit_compress();
2183 static void __exit exit_btrfs_fs(void)
2185 btrfs_destroy_cachep();
2186 btrfs_delayed_ref_exit();
2187 btrfs_auto_defrag_exit();
2188 btrfs_delayed_inode_exit();
2189 btrfs_prelim_ref_exit();
2190 ordered_data_exit();
2193 btrfs_interface_exit();
2194 btrfs_end_io_wq_exit();
2195 unregister_filesystem(&btrfs_fs_type);
2197 btrfs_cleanup_fs_uuids();
2198 btrfs_exit_compress();
2202 late_initcall(init_btrfs_fs);
2203 module_exit(exit_btrfs_fs)
2205 MODULE_LICENSE("GPL");