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"
63 #define CREATE_TRACE_POINTS
64 #include <trace/events/btrfs.h>
66 static const struct super_operations btrfs_super_ops;
67 static struct file_system_type btrfs_fs_type;
69 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
71 static const char *btrfs_decode_error(int errno)
73 char *errstr = "unknown";
77 errstr = "IO failure";
80 errstr = "Out of memory";
83 errstr = "Readonly filesystem";
86 errstr = "Object already exists";
89 errstr = "No space left";
92 errstr = "No such entry";
99 static void save_error_info(struct btrfs_fs_info *fs_info)
102 * today we only save the error info into ram. Long term we'll
103 * also send it down to the disk
105 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
108 /* btrfs handle error by forcing the filesystem readonly */
109 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
111 struct super_block *sb = fs_info->sb;
113 if (sb->s_flags & MS_RDONLY)
116 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
117 sb->s_flags |= MS_RDONLY;
118 btrfs_info(fs_info, "forced readonly");
120 * Note that a running device replace operation is not
121 * canceled here although there is no way to update
122 * the progress. It would add the risk of a deadlock,
123 * therefore the canceling is ommited. The only penalty
124 * is that some I/O remains active until the procedure
125 * completes. The next time when the filesystem is
126 * mounted writeable again, the device replace
127 * operation continues.
134 * __btrfs_std_error decodes expected errors from the caller and
135 * invokes the approciate error response.
137 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
138 unsigned int line, int errno, const char *fmt, ...)
140 struct super_block *sb = fs_info->sb;
144 * Special case: if the error is EROFS, and we're already
145 * under MS_RDONLY, then it is safe here.
147 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
150 errstr = btrfs_decode_error(errno);
152 struct va_format vaf;
160 "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
161 sb->s_id, function, line, errno, errstr, &vaf);
164 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
165 sb->s_id, function, line, errno, errstr);
168 /* Don't go through full error handling during mount */
169 save_error_info(fs_info);
170 if (sb->s_flags & MS_BORN)
171 btrfs_handle_error(fs_info);
174 static const char * const logtypes[] = {
185 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
187 struct super_block *sb = fs_info->sb;
189 struct va_format vaf;
191 const char *type = logtypes[4];
196 kern_level = printk_get_level(fmt);
198 size_t size = printk_skip_level(fmt) - fmt;
199 memcpy(lvl, fmt, size);
202 type = logtypes[kern_level - '0'];
209 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
216 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
217 unsigned int line, int errno, const char *fmt, ...)
219 struct super_block *sb = fs_info->sb;
222 * Special case: if the error is EROFS, and we're already
223 * under MS_RDONLY, then it is safe here.
225 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
228 /* Don't go through full error handling during mount */
229 if (sb->s_flags & MS_BORN) {
230 save_error_info(fs_info);
231 btrfs_handle_error(fs_info);
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.
249 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
250 struct btrfs_root *root, const char *function,
251 unsigned int line, int errno)
254 * Report first abort since mount
256 if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,
257 &root->fs_info->fs_state)) {
258 WARN(1, KERN_DEBUG "BTRFS: Transaction aborted (error %d)\n",
261 trans->aborted = errno;
262 /* Nothing used. The other threads that have joined this
263 * transaction may be able to continue. */
264 if (!trans->blocks_used) {
267 errstr = btrfs_decode_error(errno);
268 btrfs_warn(root->fs_info,
269 "%s:%d: Aborting unused transaction(%s).",
270 function, line, errstr);
273 ACCESS_ONCE(trans->transaction->aborted) = errno;
274 /* Wake up anybody who may be waiting on this transaction */
275 wake_up(&root->fs_info->transaction_wait);
276 wake_up(&root->fs_info->transaction_blocked_wait);
277 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
280 * __btrfs_panic decodes unexpected, fatal errors from the caller,
281 * issues an alert, and either panics or BUGs, depending on mount options.
283 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
284 unsigned int line, int errno, const char *fmt, ...)
286 char *s_id = "<unknown>";
288 struct va_format vaf = { .fmt = fmt };
292 s_id = fs_info->sb->s_id;
297 errstr = btrfs_decode_error(errno);
298 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
299 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
300 s_id, function, line, &vaf, errno, errstr);
302 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
303 function, line, &vaf, errno, errstr);
305 /* Caller calls BUG() */
308 static void btrfs_put_super(struct super_block *sb)
310 (void)close_ctree(btrfs_sb(sb)->tree_root);
311 /* FIXME: need to fix VFS to return error? */
312 /* AV: return it _where_? ->put_super() can be triggered by any number
313 * of async events, up to and including delivery of SIGKILL to the
314 * last process that kept it busy. Or segfault in the aforementioned
315 * process... Whom would you report that to?
320 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
321 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
322 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
323 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
324 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
325 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
326 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
327 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
328 Opt_check_integrity, Opt_check_integrity_including_extent_data,
329 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
330 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
331 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
332 Opt_datasum, Opt_treelog, Opt_noinode_cache,
336 static match_table_t tokens = {
337 {Opt_degraded, "degraded"},
338 {Opt_subvol, "subvol=%s"},
339 {Opt_subvolid, "subvolid=%s"},
340 {Opt_device, "device=%s"},
341 {Opt_nodatasum, "nodatasum"},
342 {Opt_datasum, "datasum"},
343 {Opt_nodatacow, "nodatacow"},
344 {Opt_datacow, "datacow"},
345 {Opt_nobarrier, "nobarrier"},
346 {Opt_barrier, "barrier"},
347 {Opt_max_inline, "max_inline=%s"},
348 {Opt_alloc_start, "alloc_start=%s"},
349 {Opt_thread_pool, "thread_pool=%d"},
350 {Opt_compress, "compress"},
351 {Opt_compress_type, "compress=%s"},
352 {Opt_compress_force, "compress-force"},
353 {Opt_compress_force_type, "compress-force=%s"},
355 {Opt_ssd_spread, "ssd_spread"},
356 {Opt_nossd, "nossd"},
358 {Opt_noacl, "noacl"},
359 {Opt_notreelog, "notreelog"},
360 {Opt_treelog, "treelog"},
361 {Opt_flushoncommit, "flushoncommit"},
362 {Opt_noflushoncommit, "noflushoncommit"},
363 {Opt_ratio, "metadata_ratio=%d"},
364 {Opt_discard, "discard"},
365 {Opt_nodiscard, "nodiscard"},
366 {Opt_space_cache, "space_cache"},
367 {Opt_clear_cache, "clear_cache"},
368 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
369 {Opt_enospc_debug, "enospc_debug"},
370 {Opt_noenospc_debug, "noenospc_debug"},
371 {Opt_subvolrootid, "subvolrootid=%d"},
372 {Opt_defrag, "autodefrag"},
373 {Opt_nodefrag, "noautodefrag"},
374 {Opt_inode_cache, "inode_cache"},
375 {Opt_noinode_cache, "noinode_cache"},
376 {Opt_no_space_cache, "nospace_cache"},
377 {Opt_recovery, "recovery"},
378 {Opt_skip_balance, "skip_balance"},
379 {Opt_check_integrity, "check_int"},
380 {Opt_check_integrity_including_extent_data, "check_int_data"},
381 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
382 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
383 {Opt_fatal_errors, "fatal_errors=%s"},
384 {Opt_commit_interval, "commit=%d"},
389 * Regular mount options parser. Everything that is needed only when
390 * reading in a new superblock is parsed here.
391 * XXX JDM: This needs to be cleaned up for remount.
393 int btrfs_parse_options(struct btrfs_root *root, char *options)
395 struct btrfs_fs_info *info = root->fs_info;
396 substring_t args[MAX_OPT_ARGS];
397 char *p, *num, *orig = NULL;
402 bool compress_force = false;
403 bool compress = false;
405 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
407 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
413 * strsep changes the string, duplicate it because parse_options
416 options = kstrdup(options, GFP_NOFS);
422 while ((p = strsep(&options, ",")) != NULL) {
427 token = match_token(p, tokens, args);
430 btrfs_info(root->fs_info, "allowing degraded mounts");
431 btrfs_set_opt(info->mount_opt, DEGRADED);
435 case Opt_subvolrootid:
438 * These are parsed by btrfs_parse_early_options
439 * and can be happily ignored here.
443 btrfs_set_and_info(root, NODATASUM,
444 "setting nodatasum");
447 if (btrfs_test_opt(root, NODATASUM)) {
448 if (btrfs_test_opt(root, NODATACOW))
449 btrfs_info(root->fs_info, "setting datasum, datacow enabled");
451 btrfs_info(root->fs_info, "setting datasum");
453 btrfs_clear_opt(info->mount_opt, NODATACOW);
454 btrfs_clear_opt(info->mount_opt, NODATASUM);
457 if (!btrfs_test_opt(root, NODATACOW)) {
458 if (!btrfs_test_opt(root, COMPRESS) ||
459 !btrfs_test_opt(root, FORCE_COMPRESS)) {
460 btrfs_info(root->fs_info,
461 "setting nodatacow, compression disabled");
463 btrfs_info(root->fs_info, "setting nodatacow");
466 btrfs_clear_opt(info->mount_opt, COMPRESS);
467 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
468 btrfs_set_opt(info->mount_opt, NODATACOW);
469 btrfs_set_opt(info->mount_opt, NODATASUM);
472 btrfs_clear_and_info(root, NODATACOW,
475 case Opt_compress_force:
476 case Opt_compress_force_type:
477 compress_force = true;
480 case Opt_compress_type:
482 if (token == Opt_compress ||
483 token == Opt_compress_force ||
484 strcmp(args[0].from, "zlib") == 0) {
485 compress_type = "zlib";
486 info->compress_type = BTRFS_COMPRESS_ZLIB;
487 btrfs_set_opt(info->mount_opt, COMPRESS);
488 btrfs_clear_opt(info->mount_opt, NODATACOW);
489 btrfs_clear_opt(info->mount_opt, NODATASUM);
490 } else if (strcmp(args[0].from, "lzo") == 0) {
491 compress_type = "lzo";
492 info->compress_type = BTRFS_COMPRESS_LZO;
493 btrfs_set_opt(info->mount_opt, COMPRESS);
494 btrfs_clear_opt(info->mount_opt, NODATACOW);
495 btrfs_clear_opt(info->mount_opt, NODATASUM);
496 btrfs_set_fs_incompat(info, COMPRESS_LZO);
497 } else if (strncmp(args[0].from, "no", 2) == 0) {
498 compress_type = "no";
499 btrfs_clear_opt(info->mount_opt, COMPRESS);
500 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
501 compress_force = false;
507 if (compress_force) {
508 btrfs_set_and_info(root, FORCE_COMPRESS,
509 "force %s compression",
511 } else if (compress) {
512 if (!btrfs_test_opt(root, COMPRESS))
513 btrfs_info(root->fs_info,
514 "btrfs: use %s compression",
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_and_info(root, CHANGE_INODE_CACHE,
646 "enabling inode map caching");
648 case Opt_noinode_cache:
649 btrfs_clear_and_info(root, CHANGE_INODE_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;
854 struct dentry *dentry;
859 * We have a specific subvol we want to mount, just setup location and
860 * go look up the root.
862 if (subvol_objectid) {
863 location.objectid = subvol_objectid;
864 location.type = BTRFS_ROOT_ITEM_KEY;
865 location.offset = (u64)-1;
869 path = btrfs_alloc_path();
871 return ERR_PTR(-ENOMEM);
872 path->leave_spinning = 1;
875 * Find the "default" dir item which points to the root item that we
876 * will mount by default if we haven't been given a specific subvolume
879 dir_id = btrfs_super_root_dir(fs_info->super_copy);
880 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
882 btrfs_free_path(path);
887 * Ok the default dir item isn't there. This is weird since
888 * it's always been there, but don't freak out, just try and
889 * mount to root most subvolume.
891 btrfs_free_path(path);
892 dir_id = BTRFS_FIRST_FREE_OBJECTID;
893 new_root = fs_info->fs_root;
897 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
898 btrfs_free_path(path);
901 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
902 if (IS_ERR(new_root))
903 return ERR_CAST(new_root);
905 dir_id = btrfs_root_dirid(&new_root->root_item);
907 location.objectid = dir_id;
908 location.type = BTRFS_INODE_ITEM_KEY;
911 inode = btrfs_iget(sb, &location, new_root, &new);
913 return ERR_CAST(inode);
916 * If we're just mounting the root most subvol put the inode and return
917 * a reference to the dentry. We will have already gotten a reference
918 * to the inode in btrfs_fill_super so we're good to go.
920 if (!new && sb->s_root->d_inode == inode) {
922 return dget(sb->s_root);
925 dentry = d_obtain_alias(inode);
926 if (!IS_ERR(dentry)) {
927 spin_lock(&dentry->d_lock);
928 dentry->d_flags &= ~DCACHE_DISCONNECTED;
929 spin_unlock(&dentry->d_lock);
934 static int btrfs_fill_super(struct super_block *sb,
935 struct btrfs_fs_devices *fs_devices,
936 void *data, int silent)
939 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
940 struct btrfs_key key;
943 sb->s_maxbytes = MAX_LFS_FILESIZE;
944 sb->s_magic = BTRFS_SUPER_MAGIC;
945 sb->s_op = &btrfs_super_ops;
946 sb->s_d_op = &btrfs_dentry_operations;
947 sb->s_export_op = &btrfs_export_ops;
948 sb->s_xattr = btrfs_xattr_handlers;
950 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
951 sb->s_flags |= MS_POSIXACL;
953 sb->s_flags |= MS_I_VERSION;
954 err = open_ctree(sb, fs_devices, (char *)data);
956 printk(KERN_ERR "BTRFS: open_ctree failed\n");
960 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
961 key.type = BTRFS_INODE_ITEM_KEY;
963 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
965 err = PTR_ERR(inode);
969 sb->s_root = d_make_root(inode);
975 save_mount_options(sb, data);
976 cleancache_init_fs(sb);
977 sb->s_flags |= MS_ACTIVE;
981 close_ctree(fs_info->tree_root);
985 int btrfs_sync_fs(struct super_block *sb, int wait)
987 struct btrfs_trans_handle *trans;
988 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
989 struct btrfs_root *root = fs_info->tree_root;
991 trace_btrfs_sync_fs(wait);
994 filemap_flush(fs_info->btree_inode->i_mapping);
998 btrfs_wait_ordered_roots(fs_info, -1);
1000 trans = btrfs_attach_transaction_barrier(root);
1001 if (IS_ERR(trans)) {
1002 /* no transaction, don't bother */
1003 if (PTR_ERR(trans) == -ENOENT)
1005 return PTR_ERR(trans);
1007 return btrfs_commit_transaction(trans, root);
1010 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1012 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1013 struct btrfs_root *root = info->tree_root;
1014 char *compress_type;
1016 if (btrfs_test_opt(root, DEGRADED))
1017 seq_puts(seq, ",degraded");
1018 if (btrfs_test_opt(root, NODATASUM))
1019 seq_puts(seq, ",nodatasum");
1020 if (btrfs_test_opt(root, NODATACOW))
1021 seq_puts(seq, ",nodatacow");
1022 if (btrfs_test_opt(root, NOBARRIER))
1023 seq_puts(seq, ",nobarrier");
1024 if (info->max_inline != 8192 * 1024)
1025 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1026 if (info->alloc_start != 0)
1027 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1028 if (info->thread_pool_size != min_t(unsigned long,
1029 num_online_cpus() + 2, 8))
1030 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1031 if (btrfs_test_opt(root, COMPRESS)) {
1032 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1033 compress_type = "zlib";
1035 compress_type = "lzo";
1036 if (btrfs_test_opt(root, FORCE_COMPRESS))
1037 seq_printf(seq, ",compress-force=%s", compress_type);
1039 seq_printf(seq, ",compress=%s", compress_type);
1041 if (btrfs_test_opt(root, NOSSD))
1042 seq_puts(seq, ",nossd");
1043 if (btrfs_test_opt(root, SSD_SPREAD))
1044 seq_puts(seq, ",ssd_spread");
1045 else if (btrfs_test_opt(root, SSD))
1046 seq_puts(seq, ",ssd");
1047 if (btrfs_test_opt(root, NOTREELOG))
1048 seq_puts(seq, ",notreelog");
1049 if (btrfs_test_opt(root, FLUSHONCOMMIT))
1050 seq_puts(seq, ",flushoncommit");
1051 if (btrfs_test_opt(root, DISCARD))
1052 seq_puts(seq, ",discard");
1053 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1054 seq_puts(seq, ",noacl");
1055 if (btrfs_test_opt(root, SPACE_CACHE))
1056 seq_puts(seq, ",space_cache");
1058 seq_puts(seq, ",nospace_cache");
1059 if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1060 seq_puts(seq, ",rescan_uuid_tree");
1061 if (btrfs_test_opt(root, CLEAR_CACHE))
1062 seq_puts(seq, ",clear_cache");
1063 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1064 seq_puts(seq, ",user_subvol_rm_allowed");
1065 if (btrfs_test_opt(root, ENOSPC_DEBUG))
1066 seq_puts(seq, ",enospc_debug");
1067 if (btrfs_test_opt(root, AUTO_DEFRAG))
1068 seq_puts(seq, ",autodefrag");
1069 if (btrfs_test_opt(root, INODE_MAP_CACHE))
1070 seq_puts(seq, ",inode_cache");
1071 if (btrfs_test_opt(root, SKIP_BALANCE))
1072 seq_puts(seq, ",skip_balance");
1073 if (btrfs_test_opt(root, RECOVERY))
1074 seq_puts(seq, ",recovery");
1075 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1076 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1077 seq_puts(seq, ",check_int_data");
1078 else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1079 seq_puts(seq, ",check_int");
1080 if (info->check_integrity_print_mask)
1081 seq_printf(seq, ",check_int_print_mask=%d",
1082 info->check_integrity_print_mask);
1084 if (info->metadata_ratio)
1085 seq_printf(seq, ",metadata_ratio=%d",
1086 info->metadata_ratio);
1087 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1088 seq_puts(seq, ",fatal_errors=panic");
1089 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1090 seq_printf(seq, ",commit=%d", info->commit_interval);
1094 static int btrfs_test_super(struct super_block *s, void *data)
1096 struct btrfs_fs_info *p = data;
1097 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1099 return fs_info->fs_devices == p->fs_devices;
1102 static int btrfs_set_super(struct super_block *s, void *data)
1104 int err = set_anon_super(s, data);
1106 s->s_fs_info = data;
1111 * subvolumes are identified by ino 256
1113 static inline int is_subvolume_inode(struct inode *inode)
1115 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1121 * This will strip out the subvol=%s argument for an argument string and add
1122 * subvolid=0 to make sure we get the actual tree root for path walking to the
1125 static char *setup_root_args(char *args)
1127 unsigned len = strlen(args) + 2 + 1;
1128 char *src, *dst, *buf;
1131 * We need the same args as before, but with this substitution:
1132 * s!subvol=[^,]+!subvolid=0!
1134 * Since the replacement string is up to 2 bytes longer than the
1135 * original, allocate strlen(args) + 2 + 1 bytes.
1138 src = strstr(args, "subvol=");
1139 /* This shouldn't happen, but just in case.. */
1143 buf = dst = kmalloc(len, GFP_NOFS);
1148 * If the subvol= arg is not at the start of the string,
1149 * copy whatever precedes it into buf.
1154 dst += strlen(args);
1157 strcpy(dst, "subvolid=0");
1158 dst += strlen("subvolid=0");
1161 * If there is a "," after the original subvol=... string,
1162 * copy that suffix into our buffer. Otherwise, we're done.
1164 src = strchr(src, ',');
1171 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1172 const char *device_name, char *data)
1174 struct dentry *root;
1175 struct vfsmount *mnt;
1178 newargs = setup_root_args(data);
1180 return ERR_PTR(-ENOMEM);
1181 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1184 if (PTR_RET(mnt) == -EBUSY) {
1185 if (flags & MS_RDONLY) {
1186 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY, device_name,
1190 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY, device_name,
1194 return ERR_CAST(mnt);
1197 r = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1199 /* FIXME: release vfsmount mnt ??*/
1209 return ERR_CAST(mnt);
1211 root = mount_subtree(mnt, subvol_name);
1213 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1214 struct super_block *s = root->d_sb;
1216 root = ERR_PTR(-EINVAL);
1217 deactivate_locked_super(s);
1218 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1226 * Find a superblock for the given device / mount point.
1228 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1229 * for multiple device setup. Make sure to keep it in sync.
1231 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1232 const char *device_name, void *data)
1234 struct block_device *bdev = NULL;
1235 struct super_block *s;
1236 struct dentry *root;
1237 struct btrfs_fs_devices *fs_devices = NULL;
1238 struct btrfs_fs_info *fs_info = NULL;
1239 fmode_t mode = FMODE_READ;
1240 char *subvol_name = NULL;
1241 u64 subvol_objectid = 0;
1244 if (!(flags & MS_RDONLY))
1245 mode |= FMODE_WRITE;
1247 error = btrfs_parse_early_options(data, mode, fs_type,
1248 &subvol_name, &subvol_objectid,
1252 return ERR_PTR(error);
1256 root = mount_subvol(subvol_name, flags, device_name, data);
1261 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1263 return ERR_PTR(error);
1266 * Setup a dummy root and fs_info for test/set super. This is because
1267 * we don't actually fill this stuff out until open_ctree, but we need
1268 * it for searching for existing supers, so this lets us do that and
1269 * then open_ctree will properly initialize everything later.
1271 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1273 return ERR_PTR(-ENOMEM);
1275 fs_info->fs_devices = fs_devices;
1277 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1278 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1279 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1284 error = btrfs_open_devices(fs_devices, mode, fs_type);
1288 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1290 goto error_close_devices;
1293 bdev = fs_devices->latest_bdev;
1294 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1298 goto error_close_devices;
1302 btrfs_close_devices(fs_devices);
1303 free_fs_info(fs_info);
1304 if ((flags ^ s->s_flags) & MS_RDONLY)
1307 char b[BDEVNAME_SIZE];
1309 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1310 btrfs_sb(s)->bdev_holder = fs_type;
1311 error = btrfs_fill_super(s, fs_devices, data,
1312 flags & MS_SILENT ? 1 : 0);
1315 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1317 deactivate_locked_super(s);
1321 error_close_devices:
1322 btrfs_close_devices(fs_devices);
1324 free_fs_info(fs_info);
1325 return ERR_PTR(error);
1328 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1329 int new_pool_size, int old_pool_size)
1331 if (new_pool_size == old_pool_size)
1334 fs_info->thread_pool_size = new_pool_size;
1336 btrfs_info(fs_info, "resize thread pool %d -> %d",
1337 old_pool_size, new_pool_size);
1339 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1340 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1341 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1342 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1343 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1344 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1345 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1347 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1348 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1349 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1350 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1351 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1355 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1357 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1360 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1361 unsigned long old_opts, int flags)
1363 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1364 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1365 (flags & MS_RDONLY))) {
1366 /* wait for any defraggers to finish */
1367 wait_event(fs_info->transaction_wait,
1368 (atomic_read(&fs_info->defrag_running) == 0));
1369 if (flags & MS_RDONLY)
1370 sync_filesystem(fs_info->sb);
1374 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1375 unsigned long old_opts)
1378 * We need cleanup all defragable inodes if the autodefragment is
1379 * close or the fs is R/O.
1381 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1382 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1383 (fs_info->sb->s_flags & MS_RDONLY))) {
1384 btrfs_cleanup_defrag_inodes(fs_info);
1387 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1390 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1392 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1393 struct btrfs_root *root = fs_info->tree_root;
1394 unsigned old_flags = sb->s_flags;
1395 unsigned long old_opts = fs_info->mount_opt;
1396 unsigned long old_compress_type = fs_info->compress_type;
1397 u64 old_max_inline = fs_info->max_inline;
1398 u64 old_alloc_start = fs_info->alloc_start;
1399 int old_thread_pool_size = fs_info->thread_pool_size;
1400 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1403 sync_filesystem(sb);
1404 btrfs_remount_prepare(fs_info);
1406 ret = btrfs_parse_options(root, data);
1412 btrfs_remount_begin(fs_info, old_opts, *flags);
1413 btrfs_resize_thread_pool(fs_info,
1414 fs_info->thread_pool_size, old_thread_pool_size);
1416 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1419 if (*flags & MS_RDONLY) {
1421 * this also happens on 'umount -rf' or on shutdown, when
1422 * the filesystem is busy.
1424 cancel_work_sync(&fs_info->async_reclaim_work);
1426 /* wait for the uuid_scan task to finish */
1427 down(&fs_info->uuid_tree_rescan_sem);
1428 /* avoid complains from lockdep et al. */
1429 up(&fs_info->uuid_tree_rescan_sem);
1431 sb->s_flags |= MS_RDONLY;
1433 btrfs_dev_replace_suspend_for_unmount(fs_info);
1434 btrfs_scrub_cancel(fs_info);
1435 btrfs_pause_balance(fs_info);
1437 ret = btrfs_commit_super(root);
1441 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1443 "Remounting read-write after error is not allowed");
1447 if (fs_info->fs_devices->rw_devices == 0) {
1452 if (fs_info->fs_devices->missing_devices >
1453 fs_info->num_tolerated_disk_barrier_failures &&
1454 !(*flags & MS_RDONLY)) {
1456 "too many missing devices, writeable remount is not allowed");
1461 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1466 ret = btrfs_cleanup_fs_roots(fs_info);
1470 /* recover relocation */
1471 mutex_lock(&fs_info->cleaner_mutex);
1472 ret = btrfs_recover_relocation(root);
1473 mutex_unlock(&fs_info->cleaner_mutex);
1477 ret = btrfs_resume_balance_async(fs_info);
1481 ret = btrfs_resume_dev_replace_async(fs_info);
1483 btrfs_warn(fs_info, "failed to resume dev_replace");
1487 if (!fs_info->uuid_root) {
1488 btrfs_info(fs_info, "creating UUID tree");
1489 ret = btrfs_create_uuid_tree(fs_info);
1491 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1495 sb->s_flags &= ~MS_RDONLY;
1498 wake_up_process(fs_info->transaction_kthread);
1499 btrfs_remount_cleanup(fs_info, old_opts);
1503 /* We've hit an error - don't reset MS_RDONLY */
1504 if (sb->s_flags & MS_RDONLY)
1505 old_flags |= MS_RDONLY;
1506 sb->s_flags = old_flags;
1507 fs_info->mount_opt = old_opts;
1508 fs_info->compress_type = old_compress_type;
1509 fs_info->max_inline = old_max_inline;
1510 mutex_lock(&fs_info->chunk_mutex);
1511 fs_info->alloc_start = old_alloc_start;
1512 mutex_unlock(&fs_info->chunk_mutex);
1513 btrfs_resize_thread_pool(fs_info,
1514 old_thread_pool_size, fs_info->thread_pool_size);
1515 fs_info->metadata_ratio = old_metadata_ratio;
1516 btrfs_remount_cleanup(fs_info, old_opts);
1520 /* Used to sort the devices by max_avail(descending sort) */
1521 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1522 const void *dev_info2)
1524 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1525 ((struct btrfs_device_info *)dev_info2)->max_avail)
1527 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1528 ((struct btrfs_device_info *)dev_info2)->max_avail)
1535 * sort the devices by max_avail, in which max free extent size of each device
1536 * is stored.(Descending Sort)
1538 static inline void btrfs_descending_sort_devices(
1539 struct btrfs_device_info *devices,
1542 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1543 btrfs_cmp_device_free_bytes, NULL);
1547 * The helper to calc the free space on the devices that can be used to store
1550 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1552 struct btrfs_fs_info *fs_info = root->fs_info;
1553 struct btrfs_device_info *devices_info;
1554 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1555 struct btrfs_device *device;
1560 u64 min_stripe_size;
1561 int min_stripes = 1, num_stripes = 1;
1562 int i = 0, nr_devices;
1565 nr_devices = fs_info->fs_devices->open_devices;
1566 BUG_ON(!nr_devices);
1568 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1573 /* calc min stripe number for data space alloction */
1574 type = btrfs_get_alloc_profile(root, 1);
1575 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1577 num_stripes = nr_devices;
1578 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1581 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1586 if (type & BTRFS_BLOCK_GROUP_DUP)
1587 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1589 min_stripe_size = BTRFS_STRIPE_LEN;
1591 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1592 if (!device->in_fs_metadata || !device->bdev ||
1593 device->is_tgtdev_for_dev_replace)
1596 avail_space = device->total_bytes - device->bytes_used;
1598 /* align with stripe_len */
1599 do_div(avail_space, BTRFS_STRIPE_LEN);
1600 avail_space *= BTRFS_STRIPE_LEN;
1603 * In order to avoid overwritting the superblock on the drive,
1604 * btrfs starts at an offset of at least 1MB when doing chunk
1607 skip_space = 1024 * 1024;
1609 /* user can set the offset in fs_info->alloc_start. */
1610 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1611 device->total_bytes)
1612 skip_space = max(fs_info->alloc_start, skip_space);
1615 * btrfs can not use the free space in [0, skip_space - 1],
1616 * we must subtract it from the total. In order to implement
1617 * it, we account the used space in this range first.
1619 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1622 kfree(devices_info);
1626 /* calc the free space in [0, skip_space - 1] */
1627 skip_space -= used_space;
1630 * we can use the free space in [0, skip_space - 1], subtract
1631 * it from the total.
1633 if (avail_space && avail_space >= skip_space)
1634 avail_space -= skip_space;
1638 if (avail_space < min_stripe_size)
1641 devices_info[i].dev = device;
1642 devices_info[i].max_avail = avail_space;
1649 btrfs_descending_sort_devices(devices_info, nr_devices);
1653 while (nr_devices >= min_stripes) {
1654 if (num_stripes > nr_devices)
1655 num_stripes = nr_devices;
1657 if (devices_info[i].max_avail >= min_stripe_size) {
1661 avail_space += devices_info[i].max_avail * num_stripes;
1662 alloc_size = devices_info[i].max_avail;
1663 for (j = i + 1 - num_stripes; j <= i; j++)
1664 devices_info[j].max_avail -= alloc_size;
1670 kfree(devices_info);
1671 *free_bytes = avail_space;
1676 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1678 * If there's a redundant raid level at DATA block groups, use the respective
1679 * multiplier to scale the sizes.
1681 * Unused device space usage is based on simulating the chunk allocator
1682 * algorithm that respects the device sizes, order of allocations and the
1683 * 'alloc_start' value, this is a close approximation of the actual use but
1684 * there are other factors that may change the result (like a new metadata
1687 * FIXME: not accurate for mixed block groups, total and free/used are ok,
1688 * available appears slightly larger.
1690 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1692 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1693 struct btrfs_super_block *disk_super = fs_info->super_copy;
1694 struct list_head *head = &fs_info->space_info;
1695 struct btrfs_space_info *found;
1697 u64 total_free_data = 0;
1698 int bits = dentry->d_sb->s_blocksize_bits;
1699 __be32 *fsid = (__be32 *)fs_info->fsid;
1700 unsigned factor = 1;
1701 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1704 /* holding chunk_muext to avoid allocating new chunks */
1705 mutex_lock(&fs_info->chunk_mutex);
1707 list_for_each_entry_rcu(found, head, list) {
1708 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1711 total_free_data += found->disk_total - found->disk_used;
1713 btrfs_account_ro_block_groups_free_space(found);
1715 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1716 if (!list_empty(&found->block_groups[i])) {
1718 case BTRFS_RAID_DUP:
1719 case BTRFS_RAID_RAID1:
1720 case BTRFS_RAID_RAID10:
1727 total_used += found->disk_used;
1732 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1733 buf->f_blocks >>= bits;
1734 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1736 /* Account global block reserve as used, it's in logical size already */
1737 spin_lock(&block_rsv->lock);
1738 buf->f_bfree -= block_rsv->size >> bits;
1739 spin_unlock(&block_rsv->lock);
1741 buf->f_bavail = total_free_data;
1742 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1744 mutex_unlock(&fs_info->chunk_mutex);
1747 buf->f_bavail += div_u64(total_free_data, factor);
1748 buf->f_bavail = buf->f_bavail >> bits;
1749 mutex_unlock(&fs_info->chunk_mutex);
1751 buf->f_type = BTRFS_SUPER_MAGIC;
1752 buf->f_bsize = dentry->d_sb->s_blocksize;
1753 buf->f_namelen = BTRFS_NAME_LEN;
1755 /* We treat it as constant endianness (it doesn't matter _which_)
1756 because we want the fsid to come out the same whether mounted
1757 on a big-endian or little-endian host */
1758 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1759 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1760 /* Mask in the root object ID too, to disambiguate subvols */
1761 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1762 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1767 static void btrfs_kill_super(struct super_block *sb)
1769 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1770 kill_anon_super(sb);
1771 free_fs_info(fs_info);
1774 static struct file_system_type btrfs_fs_type = {
1775 .owner = THIS_MODULE,
1777 .mount = btrfs_mount,
1778 .kill_sb = btrfs_kill_super,
1779 .fs_flags = FS_REQUIRES_DEV,
1781 MODULE_ALIAS_FS("btrfs");
1784 * used by btrfsctl to scan devices when no FS is mounted
1786 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1789 struct btrfs_ioctl_vol_args *vol;
1790 struct btrfs_fs_devices *fs_devices;
1793 if (!capable(CAP_SYS_ADMIN))
1796 vol = memdup_user((void __user *)arg, sizeof(*vol));
1798 return PTR_ERR(vol);
1801 case BTRFS_IOC_SCAN_DEV:
1802 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1803 &btrfs_fs_type, &fs_devices);
1805 case BTRFS_IOC_DEVICES_READY:
1806 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1807 &btrfs_fs_type, &fs_devices);
1810 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1818 static int btrfs_freeze(struct super_block *sb)
1820 struct btrfs_trans_handle *trans;
1821 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1823 trans = btrfs_attach_transaction_barrier(root);
1824 if (IS_ERR(trans)) {
1825 /* no transaction, don't bother */
1826 if (PTR_ERR(trans) == -ENOENT)
1828 return PTR_ERR(trans);
1830 return btrfs_commit_transaction(trans, root);
1833 static int btrfs_unfreeze(struct super_block *sb)
1838 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1840 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1841 struct btrfs_fs_devices *cur_devices;
1842 struct btrfs_device *dev, *first_dev = NULL;
1843 struct list_head *head;
1844 struct rcu_string *name;
1846 mutex_lock(&fs_info->fs_devices->device_list_mutex);
1847 cur_devices = fs_info->fs_devices;
1848 while (cur_devices) {
1849 head = &cur_devices->devices;
1850 list_for_each_entry(dev, head, dev_list) {
1855 if (!first_dev || dev->devid < first_dev->devid)
1858 cur_devices = cur_devices->seed;
1863 name = rcu_dereference(first_dev->name);
1864 seq_escape(m, name->str, " \t\n\\");
1869 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1873 static const struct super_operations btrfs_super_ops = {
1874 .drop_inode = btrfs_drop_inode,
1875 .evict_inode = btrfs_evict_inode,
1876 .put_super = btrfs_put_super,
1877 .sync_fs = btrfs_sync_fs,
1878 .show_options = btrfs_show_options,
1879 .show_devname = btrfs_show_devname,
1880 .write_inode = btrfs_write_inode,
1881 .alloc_inode = btrfs_alloc_inode,
1882 .destroy_inode = btrfs_destroy_inode,
1883 .statfs = btrfs_statfs,
1884 .remount_fs = btrfs_remount,
1885 .freeze_fs = btrfs_freeze,
1886 .unfreeze_fs = btrfs_unfreeze,
1889 static const struct file_operations btrfs_ctl_fops = {
1890 .unlocked_ioctl = btrfs_control_ioctl,
1891 .compat_ioctl = btrfs_control_ioctl,
1892 .owner = THIS_MODULE,
1893 .llseek = noop_llseek,
1896 static struct miscdevice btrfs_misc = {
1897 .minor = BTRFS_MINOR,
1898 .name = "btrfs-control",
1899 .fops = &btrfs_ctl_fops
1902 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1903 MODULE_ALIAS("devname:btrfs-control");
1905 static int btrfs_interface_init(void)
1907 return misc_register(&btrfs_misc);
1910 static void btrfs_interface_exit(void)
1912 if (misc_deregister(&btrfs_misc) < 0)
1913 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
1916 static void btrfs_print_info(void)
1918 printk(KERN_INFO "Btrfs loaded"
1919 #ifdef CONFIG_BTRFS_DEBUG
1922 #ifdef CONFIG_BTRFS_ASSERT
1925 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1926 ", integrity-checker=on"
1931 static int btrfs_run_sanity_tests(void)
1935 ret = btrfs_init_test_fs();
1939 ret = btrfs_test_free_space_cache();
1942 ret = btrfs_test_extent_buffer_operations();
1945 ret = btrfs_test_extent_io();
1948 ret = btrfs_test_inodes();
1951 ret = btrfs_test_qgroups();
1953 btrfs_destroy_test_fs();
1957 static int __init init_btrfs_fs(void)
1961 err = btrfs_hash_init();
1967 err = btrfs_init_sysfs();
1971 btrfs_init_compress();
1973 err = btrfs_init_cachep();
1977 err = extent_io_init();
1981 err = extent_map_init();
1983 goto free_extent_io;
1985 err = ordered_data_init();
1987 goto free_extent_map;
1989 err = btrfs_delayed_inode_init();
1991 goto free_ordered_data;
1993 err = btrfs_auto_defrag_init();
1995 goto free_delayed_inode;
1997 err = btrfs_delayed_ref_init();
1999 goto free_auto_defrag;
2001 err = btrfs_prelim_ref_init();
2003 goto free_prelim_ref;
2005 err = btrfs_interface_init();
2007 goto free_delayed_ref;
2009 btrfs_init_lockdep();
2013 err = btrfs_run_sanity_tests();
2015 goto unregister_ioctl;
2017 err = register_filesystem(&btrfs_fs_type);
2019 goto unregister_ioctl;
2024 btrfs_interface_exit();
2026 btrfs_prelim_ref_exit();
2028 btrfs_delayed_ref_exit();
2030 btrfs_auto_defrag_exit();
2032 btrfs_delayed_inode_exit();
2034 ordered_data_exit();
2040 btrfs_destroy_cachep();
2042 btrfs_exit_compress();
2049 static void __exit exit_btrfs_fs(void)
2051 btrfs_destroy_cachep();
2052 btrfs_delayed_ref_exit();
2053 btrfs_auto_defrag_exit();
2054 btrfs_delayed_inode_exit();
2055 btrfs_prelim_ref_exit();
2056 ordered_data_exit();
2059 btrfs_interface_exit();
2060 unregister_filesystem(&btrfs_fs_type);
2062 btrfs_cleanup_fs_uuids();
2063 btrfs_exit_compress();
2067 late_initcall(init_btrfs_fs);
2068 module_exit(exit_btrfs_fs)
2070 MODULE_LICENSE("GPL");