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>
45 #include "delayed-inode.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
51 #include "print-tree.h"
56 #include "compression.h"
57 #include "rcu-string.h"
59 #define CREATE_TRACE_POINTS
60 #include <trace/events/btrfs.h>
62 static const struct super_operations btrfs_super_ops;
63 static struct file_system_type btrfs_fs_type;
65 static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
72 errstr = "IO failure";
75 errstr = "Out of memory";
78 errstr = "Readonly filesystem";
81 errstr = "Object already exists";
85 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
94 static void __save_error_info(struct btrfs_fs_info *fs_info)
97 * today we only save the error info into ram. Long term we'll
98 * also send it down to the disk
100 fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
104 * We move write_super stuff at umount in order to avoid deadlock
105 * for umount hold all lock.
107 static void save_error_info(struct btrfs_fs_info *fs_info)
109 __save_error_info(fs_info);
112 /* btrfs handle error by forcing the filesystem readonly */
113 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
115 struct super_block *sb = fs_info->sb;
117 if (sb->s_flags & MS_RDONLY)
120 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
121 sb->s_flags |= MS_RDONLY;
122 printk(KERN_INFO "btrfs is forced readonly\n");
123 __btrfs_scrub_cancel(fs_info);
129 * __btrfs_std_error decodes expected errors from the caller and
130 * invokes the approciate error response.
132 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
133 unsigned int line, int errno, const char *fmt, ...)
135 struct super_block *sb = fs_info->sb;
142 * Special case: if the error is EROFS, and we're already
143 * under MS_RDONLY, then it is safe here.
145 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
148 errstr = btrfs_decode_error(fs_info, errno, nbuf);
150 struct va_format vaf = {
155 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s (%pV)\n",
156 sb->s_id, function, line, errstr, &vaf);
158 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
159 sb->s_id, function, line, errstr);
162 /* Don't go through full error handling during mount */
163 if (sb->s_flags & MS_BORN) {
164 save_error_info(fs_info);
165 btrfs_handle_error(fs_info);
170 const char *logtypes[] = {
181 void btrfs_printk(struct btrfs_fs_info *fs_info, const char *fmt, ...)
183 struct super_block *sb = fs_info->sb;
185 struct va_format vaf;
187 const char *type = logtypes[4];
191 if (fmt[0] == '<' && isdigit(fmt[1]) && fmt[2] == '>') {
195 type = logtypes[fmt[1] - '0'];
201 printk("%sBTRFS %s (device %s): %pV", lvl, type, sb->s_id, &vaf);
205 * We only mark the transaction aborted and then set the file system read-only.
206 * This will prevent new transactions from starting or trying to join this
209 * This means that error recovery at the call site is limited to freeing
210 * any local memory allocations and passing the error code up without
211 * further cleanup. The transaction should complete as it normally would
212 * in the call path but will return -EIO.
214 * We'll complete the cleanup in btrfs_end_transaction and
215 * btrfs_commit_transaction.
217 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
218 struct btrfs_root *root, const char *function,
219 unsigned int line, int errno)
221 WARN_ONCE(1, KERN_DEBUG "btrfs: Transaction aborted");
222 trans->aborted = errno;
223 /* Nothing used. The other threads that have joined this
224 * transaction may be able to continue. */
225 if (!trans->blocks_used) {
226 btrfs_printk(root->fs_info, "Aborting unused transaction.\n");
229 trans->transaction->aborted = errno;
230 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
233 * __btrfs_panic decodes unexpected, fatal errors from the caller,
234 * issues an alert, and either panics or BUGs, depending on mount options.
236 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
237 unsigned int line, int errno, const char *fmt, ...)
240 char *s_id = "<unknown>";
242 struct va_format vaf = { .fmt = fmt };
246 s_id = fs_info->sb->s_id;
251 errstr = btrfs_decode_error(fs_info, errno, nbuf);
252 if (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR)
253 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
254 s_id, function, line, &vaf, errstr);
256 printk(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
257 s_id, function, line, &vaf, errstr);
259 /* Caller calls BUG() */
262 static void btrfs_put_super(struct super_block *sb)
264 (void)close_ctree(btrfs_sb(sb)->tree_root);
265 /* FIXME: need to fix VFS to return error? */
266 /* AV: return it _where_? ->put_super() can be triggered by any number
267 * of async events, up to and including delivery of SIGKILL to the
268 * last process that kept it busy. Or segfault in the aforementioned
269 * process... Whom would you report that to?
274 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
275 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
276 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
277 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
278 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
279 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
280 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
281 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
282 Opt_check_integrity, Opt_check_integrity_including_extent_data,
283 Opt_check_integrity_print_mask, Opt_fatal_errors,
287 static match_table_t tokens = {
288 {Opt_degraded, "degraded"},
289 {Opt_subvol, "subvol=%s"},
290 {Opt_subvolid, "subvolid=%d"},
291 {Opt_device, "device=%s"},
292 {Opt_nodatasum, "nodatasum"},
293 {Opt_nodatacow, "nodatacow"},
294 {Opt_nobarrier, "nobarrier"},
295 {Opt_max_inline, "max_inline=%s"},
296 {Opt_alloc_start, "alloc_start=%s"},
297 {Opt_thread_pool, "thread_pool=%d"},
298 {Opt_compress, "compress"},
299 {Opt_compress_type, "compress=%s"},
300 {Opt_compress_force, "compress-force"},
301 {Opt_compress_force_type, "compress-force=%s"},
303 {Opt_ssd_spread, "ssd_spread"},
304 {Opt_nossd, "nossd"},
305 {Opt_noacl, "noacl"},
306 {Opt_notreelog, "notreelog"},
307 {Opt_flushoncommit, "flushoncommit"},
308 {Opt_ratio, "metadata_ratio=%d"},
309 {Opt_discard, "discard"},
310 {Opt_space_cache, "space_cache"},
311 {Opt_clear_cache, "clear_cache"},
312 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
313 {Opt_enospc_debug, "enospc_debug"},
314 {Opt_subvolrootid, "subvolrootid=%d"},
315 {Opt_defrag, "autodefrag"},
316 {Opt_inode_cache, "inode_cache"},
317 {Opt_no_space_cache, "nospace_cache"},
318 {Opt_recovery, "recovery"},
319 {Opt_skip_balance, "skip_balance"},
320 {Opt_check_integrity, "check_int"},
321 {Opt_check_integrity_including_extent_data, "check_int_data"},
322 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
323 {Opt_fatal_errors, "fatal_errors=%s"},
328 * Regular mount options parser. Everything that is needed only when
329 * reading in a new superblock is parsed here.
330 * XXX JDM: This needs to be cleaned up for remount.
332 int btrfs_parse_options(struct btrfs_root *root, char *options)
334 struct btrfs_fs_info *info = root->fs_info;
335 substring_t args[MAX_OPT_ARGS];
336 char *p, *num, *orig = NULL;
341 bool compress_force = false;
343 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
345 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
351 * strsep changes the string, duplicate it because parse_options
354 options = kstrdup(options, GFP_NOFS);
360 while ((p = strsep(&options, ",")) != NULL) {
365 token = match_token(p, tokens, args);
368 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
369 btrfs_set_opt(info->mount_opt, DEGRADED);
373 case Opt_subvolrootid:
376 * These are parsed by btrfs_parse_early_options
377 * and can be happily ignored here.
381 printk(KERN_INFO "btrfs: setting nodatasum\n");
382 btrfs_set_opt(info->mount_opt, NODATASUM);
385 printk(KERN_INFO "btrfs: setting nodatacow\n");
386 btrfs_set_opt(info->mount_opt, NODATACOW);
387 btrfs_set_opt(info->mount_opt, NODATASUM);
389 case Opt_compress_force:
390 case Opt_compress_force_type:
391 compress_force = true;
393 case Opt_compress_type:
394 if (token == Opt_compress ||
395 token == Opt_compress_force ||
396 strcmp(args[0].from, "zlib") == 0) {
397 compress_type = "zlib";
398 info->compress_type = BTRFS_COMPRESS_ZLIB;
399 } else if (strcmp(args[0].from, "lzo") == 0) {
400 compress_type = "lzo";
401 info->compress_type = BTRFS_COMPRESS_LZO;
407 btrfs_set_opt(info->mount_opt, COMPRESS);
408 if (compress_force) {
409 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
410 pr_info("btrfs: force %s compression\n",
413 pr_info("btrfs: use %s compression\n",
417 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
418 btrfs_set_opt(info->mount_opt, SSD);
421 printk(KERN_INFO "btrfs: use spread ssd "
422 "allocation scheme\n");
423 btrfs_set_opt(info->mount_opt, SSD);
424 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
427 printk(KERN_INFO "btrfs: not using ssd allocation "
429 btrfs_set_opt(info->mount_opt, NOSSD);
430 btrfs_clear_opt(info->mount_opt, SSD);
431 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
434 printk(KERN_INFO "btrfs: turning off barriers\n");
435 btrfs_set_opt(info->mount_opt, NOBARRIER);
437 case Opt_thread_pool:
439 match_int(&args[0], &intarg);
441 info->thread_pool_size = intarg;
444 num = match_strdup(&args[0]);
446 info->max_inline = memparse(num, NULL);
449 if (info->max_inline) {
450 info->max_inline = max_t(u64,
454 printk(KERN_INFO "btrfs: max_inline at %llu\n",
455 (unsigned long long)info->max_inline);
458 case Opt_alloc_start:
459 num = match_strdup(&args[0]);
461 info->alloc_start = memparse(num, NULL);
464 "btrfs: allocations start at %llu\n",
465 (unsigned long long)info->alloc_start);
469 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
472 printk(KERN_INFO "btrfs: disabling tree log\n");
473 btrfs_set_opt(info->mount_opt, NOTREELOG);
475 case Opt_flushoncommit:
476 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
477 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
481 match_int(&args[0], &intarg);
483 info->metadata_ratio = intarg;
484 printk(KERN_INFO "btrfs: metadata ratio %d\n",
485 info->metadata_ratio);
489 btrfs_set_opt(info->mount_opt, DISCARD);
491 case Opt_space_cache:
492 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
494 case Opt_no_space_cache:
495 printk(KERN_INFO "btrfs: disabling disk space caching\n");
496 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
498 case Opt_inode_cache:
499 printk(KERN_INFO "btrfs: enabling inode map caching\n");
500 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
502 case Opt_clear_cache:
503 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
504 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
506 case Opt_user_subvol_rm_allowed:
507 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
509 case Opt_enospc_debug:
510 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
513 printk(KERN_INFO "btrfs: enabling auto defrag");
514 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
517 printk(KERN_INFO "btrfs: enabling auto recovery");
518 btrfs_set_opt(info->mount_opt, RECOVERY);
520 case Opt_skip_balance:
521 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
523 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
524 case Opt_check_integrity_including_extent_data:
525 printk(KERN_INFO "btrfs: enabling check integrity"
526 " including extent data\n");
527 btrfs_set_opt(info->mount_opt,
528 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
529 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
531 case Opt_check_integrity:
532 printk(KERN_INFO "btrfs: enabling check integrity\n");
533 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
535 case Opt_check_integrity_print_mask:
537 match_int(&args[0], &intarg);
539 info->check_integrity_print_mask = intarg;
540 printk(KERN_INFO "btrfs:"
541 " check_integrity_print_mask 0x%x\n",
542 info->check_integrity_print_mask);
546 case Opt_check_integrity_including_extent_data:
547 case Opt_check_integrity:
548 case Opt_check_integrity_print_mask:
549 printk(KERN_ERR "btrfs: support for check_integrity*"
550 " not compiled in!\n");
554 case Opt_fatal_errors:
555 if (strcmp(args[0].from, "panic") == 0)
556 btrfs_set_opt(info->mount_opt,
557 PANIC_ON_FATAL_ERROR);
558 else if (strcmp(args[0].from, "bug") == 0)
559 btrfs_clear_opt(info->mount_opt,
560 PANIC_ON_FATAL_ERROR);
567 printk(KERN_INFO "btrfs: unrecognized mount option "
576 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
577 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
583 * Parse mount options that are required early in the mount process.
585 * All other options will be parsed on much later in the mount process and
586 * only when we need to allocate a new super block.
588 static int btrfs_parse_early_options(const char *options, fmode_t flags,
589 void *holder, char **subvol_name, u64 *subvol_objectid,
590 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
592 substring_t args[MAX_OPT_ARGS];
593 char *device_name, *opts, *orig, *p;
601 * strsep changes the string, duplicate it because parse_options
604 opts = kstrdup(options, GFP_KERNEL);
609 while ((p = strsep(&opts, ",")) != NULL) {
614 token = match_token(p, tokens, args);
618 *subvol_name = match_strdup(&args[0]);
622 error = match_int(&args[0], &intarg);
624 /* we want the original fs_tree */
627 BTRFS_FS_TREE_OBJECTID;
629 *subvol_objectid = intarg;
632 case Opt_subvolrootid:
634 error = match_int(&args[0], &intarg);
636 /* we want the original fs_tree */
639 BTRFS_FS_TREE_OBJECTID;
641 *subvol_rootid = intarg;
645 device_name = match_strdup(&args[0]);
650 error = btrfs_scan_one_device(device_name,
651 flags, holder, fs_devices);
666 static struct dentry *get_default_root(struct super_block *sb,
669 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
670 struct btrfs_root *root = fs_info->tree_root;
671 struct btrfs_root *new_root;
672 struct btrfs_dir_item *di;
673 struct btrfs_path *path;
674 struct btrfs_key location;
680 * We have a specific subvol we want to mount, just setup location and
681 * go look up the root.
683 if (subvol_objectid) {
684 location.objectid = subvol_objectid;
685 location.type = BTRFS_ROOT_ITEM_KEY;
686 location.offset = (u64)-1;
690 path = btrfs_alloc_path();
692 return ERR_PTR(-ENOMEM);
693 path->leave_spinning = 1;
696 * Find the "default" dir item which points to the root item that we
697 * will mount by default if we haven't been given a specific subvolume
700 dir_id = btrfs_super_root_dir(fs_info->super_copy);
701 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
703 btrfs_free_path(path);
708 * Ok the default dir item isn't there. This is weird since
709 * it's always been there, but don't freak out, just try and
710 * mount to root most subvolume.
712 btrfs_free_path(path);
713 dir_id = BTRFS_FIRST_FREE_OBJECTID;
714 new_root = fs_info->fs_root;
718 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
719 btrfs_free_path(path);
722 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
723 if (IS_ERR(new_root))
724 return ERR_CAST(new_root);
726 if (btrfs_root_refs(&new_root->root_item) == 0)
727 return ERR_PTR(-ENOENT);
729 dir_id = btrfs_root_dirid(&new_root->root_item);
731 location.objectid = dir_id;
732 location.type = BTRFS_INODE_ITEM_KEY;
735 inode = btrfs_iget(sb, &location, new_root, &new);
737 return ERR_CAST(inode);
740 * If we're just mounting the root most subvol put the inode and return
741 * a reference to the dentry. We will have already gotten a reference
742 * to the inode in btrfs_fill_super so we're good to go.
744 if (!new && sb->s_root->d_inode == inode) {
746 return dget(sb->s_root);
749 return d_obtain_alias(inode);
752 static int btrfs_fill_super(struct super_block *sb,
753 struct btrfs_fs_devices *fs_devices,
754 void *data, int silent)
757 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
758 struct btrfs_key key;
761 sb->s_maxbytes = MAX_LFS_FILESIZE;
762 sb->s_magic = BTRFS_SUPER_MAGIC;
763 sb->s_op = &btrfs_super_ops;
764 sb->s_d_op = &btrfs_dentry_operations;
765 sb->s_export_op = &btrfs_export_ops;
766 sb->s_xattr = btrfs_xattr_handlers;
768 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
769 sb->s_flags |= MS_POSIXACL;
771 sb->s_flags |= MS_I_VERSION;
772 err = open_ctree(sb, fs_devices, (char *)data);
774 printk("btrfs: open_ctree failed\n");
778 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
779 key.type = BTRFS_INODE_ITEM_KEY;
781 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
783 err = PTR_ERR(inode);
787 sb->s_root = d_make_root(inode);
793 save_mount_options(sb, data);
794 cleancache_init_fs(sb);
795 sb->s_flags |= MS_ACTIVE;
799 close_ctree(fs_info->tree_root);
803 int btrfs_sync_fs(struct super_block *sb, int wait)
805 struct btrfs_trans_handle *trans;
806 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
807 struct btrfs_root *root = fs_info->tree_root;
810 trace_btrfs_sync_fs(wait);
813 filemap_flush(fs_info->btree_inode->i_mapping);
817 btrfs_wait_ordered_extents(root, 0, 0);
819 trans = btrfs_start_transaction(root, 0);
821 return PTR_ERR(trans);
822 ret = btrfs_commit_transaction(trans, root);
826 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
828 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
829 struct btrfs_root *root = info->tree_root;
832 if (btrfs_test_opt(root, DEGRADED))
833 seq_puts(seq, ",degraded");
834 if (btrfs_test_opt(root, NODATASUM))
835 seq_puts(seq, ",nodatasum");
836 if (btrfs_test_opt(root, NODATACOW))
837 seq_puts(seq, ",nodatacow");
838 if (btrfs_test_opt(root, NOBARRIER))
839 seq_puts(seq, ",nobarrier");
840 if (info->max_inline != 8192 * 1024)
841 seq_printf(seq, ",max_inline=%llu",
842 (unsigned long long)info->max_inline);
843 if (info->alloc_start != 0)
844 seq_printf(seq, ",alloc_start=%llu",
845 (unsigned long long)info->alloc_start);
846 if (info->thread_pool_size != min_t(unsigned long,
847 num_online_cpus() + 2, 8))
848 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
849 if (btrfs_test_opt(root, COMPRESS)) {
850 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
851 compress_type = "zlib";
853 compress_type = "lzo";
854 if (btrfs_test_opt(root, FORCE_COMPRESS))
855 seq_printf(seq, ",compress-force=%s", compress_type);
857 seq_printf(seq, ",compress=%s", compress_type);
859 if (btrfs_test_opt(root, NOSSD))
860 seq_puts(seq, ",nossd");
861 if (btrfs_test_opt(root, SSD_SPREAD))
862 seq_puts(seq, ",ssd_spread");
863 else if (btrfs_test_opt(root, SSD))
864 seq_puts(seq, ",ssd");
865 if (btrfs_test_opt(root, NOTREELOG))
866 seq_puts(seq, ",notreelog");
867 if (btrfs_test_opt(root, FLUSHONCOMMIT))
868 seq_puts(seq, ",flushoncommit");
869 if (btrfs_test_opt(root, DISCARD))
870 seq_puts(seq, ",discard");
871 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
872 seq_puts(seq, ",noacl");
873 if (btrfs_test_opt(root, SPACE_CACHE))
874 seq_puts(seq, ",space_cache");
876 seq_puts(seq, ",nospace_cache");
877 if (btrfs_test_opt(root, CLEAR_CACHE))
878 seq_puts(seq, ",clear_cache");
879 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
880 seq_puts(seq, ",user_subvol_rm_allowed");
881 if (btrfs_test_opt(root, ENOSPC_DEBUG))
882 seq_puts(seq, ",enospc_debug");
883 if (btrfs_test_opt(root, AUTO_DEFRAG))
884 seq_puts(seq, ",autodefrag");
885 if (btrfs_test_opt(root, INODE_MAP_CACHE))
886 seq_puts(seq, ",inode_cache");
887 if (btrfs_test_opt(root, SKIP_BALANCE))
888 seq_puts(seq, ",skip_balance");
889 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
890 seq_puts(seq, ",fatal_errors=panic");
894 static int btrfs_test_super(struct super_block *s, void *data)
896 struct btrfs_fs_info *p = data;
897 struct btrfs_fs_info *fs_info = btrfs_sb(s);
899 return fs_info->fs_devices == p->fs_devices;
902 static int btrfs_set_super(struct super_block *s, void *data)
904 int err = set_anon_super(s, data);
911 * subvolumes are identified by ino 256
913 static inline int is_subvolume_inode(struct inode *inode)
915 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
921 * This will strip out the subvol=%s argument for an argument string and add
922 * subvolid=0 to make sure we get the actual tree root for path walking to the
925 static char *setup_root_args(char *args)
927 unsigned len = strlen(args) + 2 + 1;
928 char *src, *dst, *buf;
931 * We need the same args as before, but with this substitution:
932 * s!subvol=[^,]+!subvolid=0!
934 * Since the replacement string is up to 2 bytes longer than the
935 * original, allocate strlen(args) + 2 + 1 bytes.
938 src = strstr(args, "subvol=");
939 /* This shouldn't happen, but just in case.. */
943 buf = dst = kmalloc(len, GFP_NOFS);
948 * If the subvol= arg is not at the start of the string,
949 * copy whatever precedes it into buf.
957 strcpy(dst, "subvolid=0");
958 dst += strlen("subvolid=0");
961 * If there is a "," after the original subvol=... string,
962 * copy that suffix into our buffer. Otherwise, we're done.
964 src = strchr(src, ',');
971 static struct dentry *mount_subvol(const char *subvol_name, int flags,
972 const char *device_name, char *data)
975 struct vfsmount *mnt;
978 newargs = setup_root_args(data);
980 return ERR_PTR(-ENOMEM);
981 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
985 return ERR_CAST(mnt);
987 root = mount_subtree(mnt, subvol_name);
989 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
990 struct super_block *s = root->d_sb;
992 root = ERR_PTR(-EINVAL);
993 deactivate_locked_super(s);
994 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
1002 * Find a superblock for the given device / mount point.
1004 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1005 * for multiple device setup. Make sure to keep it in sync.
1007 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1008 const char *device_name, void *data)
1010 struct block_device *bdev = NULL;
1011 struct super_block *s;
1012 struct dentry *root;
1013 struct btrfs_fs_devices *fs_devices = NULL;
1014 struct btrfs_fs_info *fs_info = NULL;
1015 fmode_t mode = FMODE_READ;
1016 char *subvol_name = NULL;
1017 u64 subvol_objectid = 0;
1018 u64 subvol_rootid = 0;
1021 if (!(flags & MS_RDONLY))
1022 mode |= FMODE_WRITE;
1024 error = btrfs_parse_early_options(data, mode, fs_type,
1025 &subvol_name, &subvol_objectid,
1026 &subvol_rootid, &fs_devices);
1029 return ERR_PTR(error);
1033 root = mount_subvol(subvol_name, flags, device_name, data);
1038 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1040 return ERR_PTR(error);
1043 * Setup a dummy root and fs_info for test/set super. This is because
1044 * we don't actually fill this stuff out until open_ctree, but we need
1045 * it for searching for existing supers, so this lets us do that and
1046 * then open_ctree will properly initialize everything later.
1048 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1050 return ERR_PTR(-ENOMEM);
1052 fs_info->fs_devices = fs_devices;
1054 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1055 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1056 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1061 error = btrfs_open_devices(fs_devices, mode, fs_type);
1065 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1067 goto error_close_devices;
1070 bdev = fs_devices->latest_bdev;
1071 s = sget(fs_type, btrfs_test_super, btrfs_set_super, fs_info);
1074 goto error_close_devices;
1078 btrfs_close_devices(fs_devices);
1079 free_fs_info(fs_info);
1080 if ((flags ^ s->s_flags) & MS_RDONLY)
1083 char b[BDEVNAME_SIZE];
1085 s->s_flags = flags | MS_NOSEC;
1086 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1087 btrfs_sb(s)->bdev_holder = fs_type;
1088 error = btrfs_fill_super(s, fs_devices, data,
1089 flags & MS_SILENT ? 1 : 0);
1092 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1094 deactivate_locked_super(s);
1098 error_close_devices:
1099 btrfs_close_devices(fs_devices);
1101 free_fs_info(fs_info);
1102 return ERR_PTR(error);
1105 static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1107 spin_lock_irq(&workers->lock);
1108 workers->max_workers = new_limit;
1109 spin_unlock_irq(&workers->lock);
1112 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1113 int new_pool_size, int old_pool_size)
1115 if (new_pool_size == old_pool_size)
1118 fs_info->thread_pool_size = new_pool_size;
1120 printk(KERN_INFO "btrfs: resize thread pool %d -> %d\n",
1121 old_pool_size, new_pool_size);
1123 btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1124 btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1125 btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1126 btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1127 btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1128 btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1129 btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1130 btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1131 btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1132 btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1133 btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1134 btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1135 btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
1136 btrfs_set_max_workers(&fs_info->scrub_workers, new_pool_size);
1139 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1141 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1142 struct btrfs_root *root = fs_info->tree_root;
1143 unsigned old_flags = sb->s_flags;
1144 unsigned long old_opts = fs_info->mount_opt;
1145 unsigned long old_compress_type = fs_info->compress_type;
1146 u64 old_max_inline = fs_info->max_inline;
1147 u64 old_alloc_start = fs_info->alloc_start;
1148 int old_thread_pool_size = fs_info->thread_pool_size;
1149 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1152 ret = btrfs_parse_options(root, data);
1158 btrfs_resize_thread_pool(fs_info,
1159 fs_info->thread_pool_size, old_thread_pool_size);
1161 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1164 if (*flags & MS_RDONLY) {
1165 sb->s_flags |= MS_RDONLY;
1167 ret = btrfs_commit_super(root);
1171 if (fs_info->fs_devices->rw_devices == 0) {
1176 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1181 ret = btrfs_cleanup_fs_roots(fs_info);
1185 /* recover relocation */
1186 ret = btrfs_recover_relocation(root);
1190 sb->s_flags &= ~MS_RDONLY;
1196 /* We've hit an error - don't reset MS_RDONLY */
1197 if (sb->s_flags & MS_RDONLY)
1198 old_flags |= MS_RDONLY;
1199 sb->s_flags = old_flags;
1200 fs_info->mount_opt = old_opts;
1201 fs_info->compress_type = old_compress_type;
1202 fs_info->max_inline = old_max_inline;
1203 fs_info->alloc_start = old_alloc_start;
1204 btrfs_resize_thread_pool(fs_info,
1205 old_thread_pool_size, fs_info->thread_pool_size);
1206 fs_info->metadata_ratio = old_metadata_ratio;
1210 /* Used to sort the devices by max_avail(descending sort) */
1211 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1212 const void *dev_info2)
1214 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1215 ((struct btrfs_device_info *)dev_info2)->max_avail)
1217 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1218 ((struct btrfs_device_info *)dev_info2)->max_avail)
1225 * sort the devices by max_avail, in which max free extent size of each device
1226 * is stored.(Descending Sort)
1228 static inline void btrfs_descending_sort_devices(
1229 struct btrfs_device_info *devices,
1232 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1233 btrfs_cmp_device_free_bytes, NULL);
1237 * The helper to calc the free space on the devices that can be used to store
1240 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1242 struct btrfs_fs_info *fs_info = root->fs_info;
1243 struct btrfs_device_info *devices_info;
1244 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1245 struct btrfs_device *device;
1250 u64 min_stripe_size;
1251 int min_stripes = 1, num_stripes = 1;
1252 int i = 0, nr_devices;
1255 nr_devices = fs_info->fs_devices->open_devices;
1256 BUG_ON(!nr_devices);
1258 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1263 /* calc min stripe number for data space alloction */
1264 type = btrfs_get_alloc_profile(root, 1);
1265 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1267 num_stripes = nr_devices;
1268 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1271 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1276 if (type & BTRFS_BLOCK_GROUP_DUP)
1277 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1279 min_stripe_size = BTRFS_STRIPE_LEN;
1281 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1282 if (!device->in_fs_metadata || !device->bdev)
1285 avail_space = device->total_bytes - device->bytes_used;
1287 /* align with stripe_len */
1288 do_div(avail_space, BTRFS_STRIPE_LEN);
1289 avail_space *= BTRFS_STRIPE_LEN;
1292 * In order to avoid overwritting the superblock on the drive,
1293 * btrfs starts at an offset of at least 1MB when doing chunk
1296 skip_space = 1024 * 1024;
1298 /* user can set the offset in fs_info->alloc_start. */
1299 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1300 device->total_bytes)
1301 skip_space = max(fs_info->alloc_start, skip_space);
1304 * btrfs can not use the free space in [0, skip_space - 1],
1305 * we must subtract it from the total. In order to implement
1306 * it, we account the used space in this range first.
1308 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1311 kfree(devices_info);
1315 /* calc the free space in [0, skip_space - 1] */
1316 skip_space -= used_space;
1319 * we can use the free space in [0, skip_space - 1], subtract
1320 * it from the total.
1322 if (avail_space && avail_space >= skip_space)
1323 avail_space -= skip_space;
1327 if (avail_space < min_stripe_size)
1330 devices_info[i].dev = device;
1331 devices_info[i].max_avail = avail_space;
1338 btrfs_descending_sort_devices(devices_info, nr_devices);
1342 while (nr_devices >= min_stripes) {
1343 if (num_stripes > nr_devices)
1344 num_stripes = nr_devices;
1346 if (devices_info[i].max_avail >= min_stripe_size) {
1350 avail_space += devices_info[i].max_avail * num_stripes;
1351 alloc_size = devices_info[i].max_avail;
1352 for (j = i + 1 - num_stripes; j <= i; j++)
1353 devices_info[j].max_avail -= alloc_size;
1359 kfree(devices_info);
1360 *free_bytes = avail_space;
1364 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1366 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1367 struct btrfs_super_block *disk_super = fs_info->super_copy;
1368 struct list_head *head = &fs_info->space_info;
1369 struct btrfs_space_info *found;
1371 u64 total_free_data = 0;
1372 int bits = dentry->d_sb->s_blocksize_bits;
1373 __be32 *fsid = (__be32 *)fs_info->fsid;
1376 /* holding chunk_muext to avoid allocating new chunks */
1377 mutex_lock(&fs_info->chunk_mutex);
1379 list_for_each_entry_rcu(found, head, list) {
1380 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1381 total_free_data += found->disk_total - found->disk_used;
1383 btrfs_account_ro_block_groups_free_space(found);
1386 total_used += found->disk_used;
1390 buf->f_namelen = BTRFS_NAME_LEN;
1391 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1392 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1393 buf->f_bsize = dentry->d_sb->s_blocksize;
1394 buf->f_type = BTRFS_SUPER_MAGIC;
1395 buf->f_bavail = total_free_data;
1396 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1398 mutex_unlock(&fs_info->chunk_mutex);
1401 buf->f_bavail += total_free_data;
1402 buf->f_bavail = buf->f_bavail >> bits;
1403 mutex_unlock(&fs_info->chunk_mutex);
1405 /* We treat it as constant endianness (it doesn't matter _which_)
1406 because we want the fsid to come out the same whether mounted
1407 on a big-endian or little-endian host */
1408 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1409 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1410 /* Mask in the root object ID too, to disambiguate subvols */
1411 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1412 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1417 static void btrfs_kill_super(struct super_block *sb)
1419 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1420 kill_anon_super(sb);
1421 free_fs_info(fs_info);
1424 static struct file_system_type btrfs_fs_type = {
1425 .owner = THIS_MODULE,
1427 .mount = btrfs_mount,
1428 .kill_sb = btrfs_kill_super,
1429 .fs_flags = FS_REQUIRES_DEV,
1433 * used by btrfsctl to scan devices when no FS is mounted
1435 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1438 struct btrfs_ioctl_vol_args *vol;
1439 struct btrfs_fs_devices *fs_devices;
1442 if (!capable(CAP_SYS_ADMIN))
1445 vol = memdup_user((void __user *)arg, sizeof(*vol));
1447 return PTR_ERR(vol);
1450 case BTRFS_IOC_SCAN_DEV:
1451 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1452 &btrfs_fs_type, &fs_devices);
1460 static int btrfs_freeze(struct super_block *sb)
1462 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1463 mutex_lock(&fs_info->transaction_kthread_mutex);
1464 mutex_lock(&fs_info->cleaner_mutex);
1468 static int btrfs_unfreeze(struct super_block *sb)
1470 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1471 mutex_unlock(&fs_info->cleaner_mutex);
1472 mutex_unlock(&fs_info->transaction_kthread_mutex);
1476 static void btrfs_fs_dirty_inode(struct inode *inode, int flags)
1480 ret = btrfs_dirty_inode(inode);
1482 printk_ratelimited(KERN_ERR "btrfs: fail to dirty inode %Lu "
1483 "error %d\n", btrfs_ino(inode), ret);
1486 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1488 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1489 struct btrfs_fs_devices *cur_devices;
1490 struct btrfs_device *dev, *first_dev = NULL;
1491 struct list_head *head;
1492 struct rcu_string *name;
1494 mutex_lock(&fs_info->fs_devices->device_list_mutex);
1495 cur_devices = fs_info->fs_devices;
1496 while (cur_devices) {
1497 head = &cur_devices->devices;
1498 list_for_each_entry(dev, head, dev_list) {
1499 if (!first_dev || dev->devid < first_dev->devid)
1502 cur_devices = cur_devices->seed;
1507 name = rcu_dereference(first_dev->name);
1508 seq_escape(m, name->str, " \t\n\\");
1513 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1517 static const struct super_operations btrfs_super_ops = {
1518 .drop_inode = btrfs_drop_inode,
1519 .evict_inode = btrfs_evict_inode,
1520 .put_super = btrfs_put_super,
1521 .sync_fs = btrfs_sync_fs,
1522 .show_options = btrfs_show_options,
1523 .show_devname = btrfs_show_devname,
1524 .write_inode = btrfs_write_inode,
1525 .dirty_inode = btrfs_fs_dirty_inode,
1526 .alloc_inode = btrfs_alloc_inode,
1527 .destroy_inode = btrfs_destroy_inode,
1528 .statfs = btrfs_statfs,
1529 .remount_fs = btrfs_remount,
1530 .freeze_fs = btrfs_freeze,
1531 .unfreeze_fs = btrfs_unfreeze,
1534 static const struct file_operations btrfs_ctl_fops = {
1535 .unlocked_ioctl = btrfs_control_ioctl,
1536 .compat_ioctl = btrfs_control_ioctl,
1537 .owner = THIS_MODULE,
1538 .llseek = noop_llseek,
1541 static struct miscdevice btrfs_misc = {
1542 .minor = BTRFS_MINOR,
1543 .name = "btrfs-control",
1544 .fops = &btrfs_ctl_fops
1547 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1548 MODULE_ALIAS("devname:btrfs-control");
1550 static int btrfs_interface_init(void)
1552 return misc_register(&btrfs_misc);
1555 static void btrfs_interface_exit(void)
1557 if (misc_deregister(&btrfs_misc) < 0)
1558 printk(KERN_INFO "misc_deregister failed for control device");
1561 static int __init init_btrfs_fs(void)
1565 err = btrfs_init_sysfs();
1569 btrfs_init_compress();
1571 err = btrfs_init_cachep();
1575 err = extent_io_init();
1579 err = extent_map_init();
1581 goto free_extent_io;
1583 err = btrfs_delayed_inode_init();
1585 goto free_extent_map;
1587 err = btrfs_interface_init();
1589 goto free_delayed_inode;
1591 err = register_filesystem(&btrfs_fs_type);
1593 goto unregister_ioctl;
1595 btrfs_init_lockdep();
1597 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1601 btrfs_interface_exit();
1603 btrfs_delayed_inode_exit();
1609 btrfs_destroy_cachep();
1611 btrfs_exit_compress();
1616 static void __exit exit_btrfs_fs(void)
1618 btrfs_destroy_cachep();
1619 btrfs_delayed_inode_exit();
1622 btrfs_interface_exit();
1623 unregister_filesystem(&btrfs_fs_type);
1625 btrfs_cleanup_fs_uuids();
1626 btrfs_exit_compress();
1629 module_init(init_btrfs_fs)
1630 module_exit(exit_btrfs_fs)
1632 MODULE_LICENSE("GPL");