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"
58 #define CREATE_TRACE_POINTS
59 #include <trace/events/btrfs.h>
61 static const struct super_operations btrfs_super_ops;
62 static struct file_system_type btrfs_fs_type;
64 static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
71 errstr = "IO failure";
74 errstr = "Out of memory";
77 errstr = "Readonly filesystem";
80 errstr = "Object already exists";
84 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
93 static void __save_error_info(struct btrfs_fs_info *fs_info)
96 * today we only save the error info into ram. Long term we'll
97 * also send it down to the disk
99 fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
103 * We move write_super stuff at umount in order to avoid deadlock
104 * for umount hold all lock.
106 static void save_error_info(struct btrfs_fs_info *fs_info)
108 __save_error_info(fs_info);
111 /* btrfs handle error by forcing the filesystem readonly */
112 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
114 struct super_block *sb = fs_info->sb;
116 if (sb->s_flags & MS_RDONLY)
119 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
120 sb->s_flags |= MS_RDONLY;
121 printk(KERN_INFO "btrfs is forced readonly\n");
122 __btrfs_scrub_cancel(fs_info);
128 * __btrfs_std_error decodes expected errors from the caller and
129 * invokes the approciate error response.
131 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
132 unsigned int line, int errno, const char *fmt, ...)
134 struct super_block *sb = fs_info->sb;
141 * Special case: if the error is EROFS, and we're already
142 * under MS_RDONLY, then it is safe here.
144 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
147 errstr = btrfs_decode_error(fs_info, errno, nbuf);
149 struct va_format vaf = {
154 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s (%pV)\n",
155 sb->s_id, function, line, errstr, &vaf);
157 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
158 sb->s_id, function, line, errstr);
161 /* Don't go through full error handling during mount */
162 if (sb->s_flags & MS_BORN) {
163 save_error_info(fs_info);
164 btrfs_handle_error(fs_info);
169 const char *logtypes[] = {
180 void btrfs_printk(struct btrfs_fs_info *fs_info, const char *fmt, ...)
182 struct super_block *sb = fs_info->sb;
184 struct va_format vaf;
186 const char *type = logtypes[4];
190 if (fmt[0] == '<' && isdigit(fmt[1]) && fmt[2] == '>') {
191 strncpy(lvl, fmt, 3);
193 type = logtypes[fmt[1] - '0'];
199 printk("%sBTRFS %s (device %s): %pV", lvl, type, sb->s_id, &vaf);
203 * We only mark the transaction aborted and then set the file system read-only.
204 * This will prevent new transactions from starting or trying to join this
207 * This means that error recovery at the call site is limited to freeing
208 * any local memory allocations and passing the error code up without
209 * further cleanup. The transaction should complete as it normally would
210 * in the call path but will return -EIO.
212 * We'll complete the cleanup in btrfs_end_transaction and
213 * btrfs_commit_transaction.
215 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
216 struct btrfs_root *root, const char *function,
217 unsigned int line, int errno)
219 WARN_ONCE(1, KERN_DEBUG "btrfs: Transaction aborted");
220 trans->aborted = errno;
221 /* Nothing used. The other threads that have joined this
222 * transaction may be able to continue. */
223 if (!trans->blocks_used) {
224 btrfs_printk(root->fs_info, "Aborting unused transaction.\n");
227 trans->transaction->aborted = errno;
228 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
231 * __btrfs_panic decodes unexpected, fatal errors from the caller,
232 * issues an alert, and either panics or BUGs, depending on mount options.
234 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
235 unsigned int line, int errno, const char *fmt, ...)
238 char *s_id = "<unknown>";
240 struct va_format vaf = { .fmt = fmt };
244 s_id = fs_info->sb->s_id;
249 errstr = btrfs_decode_error(fs_info, errno, nbuf);
250 if (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR)
251 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
252 s_id, function, line, &vaf, errstr);
254 printk(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
255 s_id, function, line, &vaf, errstr);
257 /* Caller calls BUG() */
260 static void btrfs_put_super(struct super_block *sb)
262 (void)close_ctree(btrfs_sb(sb)->tree_root);
263 /* FIXME: need to fix VFS to return error? */
264 /* AV: return it _where_? ->put_super() can be triggered by any number
265 * of async events, up to and including delivery of SIGKILL to the
266 * last process that kept it busy. Or segfault in the aforementioned
267 * process... Whom would you report that to?
272 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
273 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
274 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
275 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
276 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
277 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
278 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
279 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
280 Opt_check_integrity, Opt_check_integrity_including_extent_data,
281 Opt_check_integrity_print_mask, Opt_fatal_errors,
285 static match_table_t tokens = {
286 {Opt_degraded, "degraded"},
287 {Opt_subvol, "subvol=%s"},
288 {Opt_subvolid, "subvolid=%d"},
289 {Opt_device, "device=%s"},
290 {Opt_nodatasum, "nodatasum"},
291 {Opt_nodatacow, "nodatacow"},
292 {Opt_nobarrier, "nobarrier"},
293 {Opt_max_inline, "max_inline=%s"},
294 {Opt_alloc_start, "alloc_start=%s"},
295 {Opt_thread_pool, "thread_pool=%d"},
296 {Opt_compress, "compress"},
297 {Opt_compress_type, "compress=%s"},
298 {Opt_compress_force, "compress-force"},
299 {Opt_compress_force_type, "compress-force=%s"},
301 {Opt_ssd_spread, "ssd_spread"},
302 {Opt_nossd, "nossd"},
303 {Opt_noacl, "noacl"},
304 {Opt_notreelog, "notreelog"},
305 {Opt_flushoncommit, "flushoncommit"},
306 {Opt_ratio, "metadata_ratio=%d"},
307 {Opt_discard, "discard"},
308 {Opt_space_cache, "space_cache"},
309 {Opt_clear_cache, "clear_cache"},
310 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
311 {Opt_enospc_debug, "enospc_debug"},
312 {Opt_subvolrootid, "subvolrootid=%d"},
313 {Opt_defrag, "autodefrag"},
314 {Opt_inode_cache, "inode_cache"},
315 {Opt_no_space_cache, "nospace_cache"},
316 {Opt_recovery, "recovery"},
317 {Opt_skip_balance, "skip_balance"},
318 {Opt_check_integrity, "check_int"},
319 {Opt_check_integrity_including_extent_data, "check_int_data"},
320 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
321 {Opt_fatal_errors, "fatal_errors=%s"},
326 * Regular mount options parser. Everything that is needed only when
327 * reading in a new superblock is parsed here.
328 * XXX JDM: This needs to be cleaned up for remount.
330 int btrfs_parse_options(struct btrfs_root *root, char *options)
332 struct btrfs_fs_info *info = root->fs_info;
333 substring_t args[MAX_OPT_ARGS];
334 char *p, *num, *orig = NULL;
339 bool compress_force = false;
341 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
343 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
349 * strsep changes the string, duplicate it because parse_options
352 options = kstrdup(options, GFP_NOFS);
358 while ((p = strsep(&options, ",")) != NULL) {
363 token = match_token(p, tokens, args);
366 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
367 btrfs_set_opt(info->mount_opt, DEGRADED);
371 case Opt_subvolrootid:
374 * These are parsed by btrfs_parse_early_options
375 * and can be happily ignored here.
379 printk(KERN_INFO "btrfs: setting nodatasum\n");
380 btrfs_set_opt(info->mount_opt, NODATASUM);
383 printk(KERN_INFO "btrfs: setting nodatacow\n");
384 btrfs_set_opt(info->mount_opt, NODATACOW);
385 btrfs_set_opt(info->mount_opt, NODATASUM);
387 case Opt_compress_force:
388 case Opt_compress_force_type:
389 compress_force = true;
391 case Opt_compress_type:
392 if (token == Opt_compress ||
393 token == Opt_compress_force ||
394 strcmp(args[0].from, "zlib") == 0) {
395 compress_type = "zlib";
396 info->compress_type = BTRFS_COMPRESS_ZLIB;
397 } else if (strcmp(args[0].from, "lzo") == 0) {
398 compress_type = "lzo";
399 info->compress_type = BTRFS_COMPRESS_LZO;
405 btrfs_set_opt(info->mount_opt, COMPRESS);
406 if (compress_force) {
407 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
408 pr_info("btrfs: force %s compression\n",
411 pr_info("btrfs: use %s compression\n",
415 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
416 btrfs_set_opt(info->mount_opt, SSD);
419 printk(KERN_INFO "btrfs: use spread ssd "
420 "allocation scheme\n");
421 btrfs_set_opt(info->mount_opt, SSD);
422 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
425 printk(KERN_INFO "btrfs: not using ssd allocation "
427 btrfs_set_opt(info->mount_opt, NOSSD);
428 btrfs_clear_opt(info->mount_opt, SSD);
429 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
432 printk(KERN_INFO "btrfs: turning off barriers\n");
433 btrfs_set_opt(info->mount_opt, NOBARRIER);
435 case Opt_thread_pool:
437 match_int(&args[0], &intarg);
439 info->thread_pool_size = intarg;
440 printk(KERN_INFO "btrfs: thread pool %d\n",
441 info->thread_pool_size);
445 num = match_strdup(&args[0]);
447 info->max_inline = memparse(num, NULL);
450 if (info->max_inline) {
451 info->max_inline = max_t(u64,
455 printk(KERN_INFO "btrfs: max_inline at %llu\n",
456 (unsigned long long)info->max_inline);
459 case Opt_alloc_start:
460 num = match_strdup(&args[0]);
462 info->alloc_start = memparse(num, NULL);
465 "btrfs: allocations start at %llu\n",
466 (unsigned long long)info->alloc_start);
470 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
473 printk(KERN_INFO "btrfs: disabling tree log\n");
474 btrfs_set_opt(info->mount_opt, NOTREELOG);
476 case Opt_flushoncommit:
477 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
478 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
482 match_int(&args[0], &intarg);
484 info->metadata_ratio = intarg;
485 printk(KERN_INFO "btrfs: metadata ratio %d\n",
486 info->metadata_ratio);
490 btrfs_set_opt(info->mount_opt, DISCARD);
492 case Opt_space_cache:
493 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
495 case Opt_no_space_cache:
496 printk(KERN_INFO "btrfs: disabling disk space caching\n");
497 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
499 case Opt_inode_cache:
500 printk(KERN_INFO "btrfs: enabling inode map caching\n");
501 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
503 case Opt_clear_cache:
504 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
505 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
507 case Opt_user_subvol_rm_allowed:
508 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
510 case Opt_enospc_debug:
511 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
514 printk(KERN_INFO "btrfs: enabling auto defrag");
515 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
518 printk(KERN_INFO "btrfs: enabling auto recovery");
519 btrfs_set_opt(info->mount_opt, RECOVERY);
521 case Opt_skip_balance:
522 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
524 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
525 case Opt_check_integrity_including_extent_data:
526 printk(KERN_INFO "btrfs: enabling check integrity"
527 " including extent data\n");
528 btrfs_set_opt(info->mount_opt,
529 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
530 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
532 case Opt_check_integrity:
533 printk(KERN_INFO "btrfs: enabling check integrity\n");
534 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
536 case Opt_check_integrity_print_mask:
538 match_int(&args[0], &intarg);
540 info->check_integrity_print_mask = intarg;
541 printk(KERN_INFO "btrfs:"
542 " check_integrity_print_mask 0x%x\n",
543 info->check_integrity_print_mask);
547 case Opt_check_integrity_including_extent_data:
548 case Opt_check_integrity:
549 case Opt_check_integrity_print_mask:
550 printk(KERN_ERR "btrfs: support for check_integrity*"
551 " not compiled in!\n");
555 case Opt_fatal_errors:
556 if (strcmp(args[0].from, "panic") == 0)
557 btrfs_set_opt(info->mount_opt,
558 PANIC_ON_FATAL_ERROR);
559 else if (strcmp(args[0].from, "bug") == 0)
560 btrfs_clear_opt(info->mount_opt,
561 PANIC_ON_FATAL_ERROR);
568 printk(KERN_INFO "btrfs: unrecognized mount option "
577 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
578 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
584 * Parse mount options that are required early in the mount process.
586 * All other options will be parsed on much later in the mount process and
587 * only when we need to allocate a new super block.
589 static int btrfs_parse_early_options(const char *options, fmode_t flags,
590 void *holder, char **subvol_name, u64 *subvol_objectid,
591 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
593 substring_t args[MAX_OPT_ARGS];
594 char *device_name, *opts, *orig, *p;
602 * strsep changes the string, duplicate it because parse_options
605 opts = kstrdup(options, GFP_KERNEL);
610 while ((p = strsep(&opts, ",")) != NULL) {
615 token = match_token(p, tokens, args);
619 *subvol_name = match_strdup(&args[0]);
623 error = match_int(&args[0], &intarg);
625 /* we want the original fs_tree */
628 BTRFS_FS_TREE_OBJECTID;
630 *subvol_objectid = intarg;
633 case Opt_subvolrootid:
635 error = match_int(&args[0], &intarg);
637 /* we want the original fs_tree */
640 BTRFS_FS_TREE_OBJECTID;
642 *subvol_rootid = intarg;
646 device_name = match_strdup(&args[0]);
651 error = btrfs_scan_one_device(device_name,
652 flags, holder, fs_devices);
667 static struct dentry *get_default_root(struct super_block *sb,
670 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
671 struct btrfs_root *root = fs_info->tree_root;
672 struct btrfs_root *new_root;
673 struct btrfs_dir_item *di;
674 struct btrfs_path *path;
675 struct btrfs_key location;
681 * We have a specific subvol we want to mount, just setup location and
682 * go look up the root.
684 if (subvol_objectid) {
685 location.objectid = subvol_objectid;
686 location.type = BTRFS_ROOT_ITEM_KEY;
687 location.offset = (u64)-1;
691 path = btrfs_alloc_path();
693 return ERR_PTR(-ENOMEM);
694 path->leave_spinning = 1;
697 * Find the "default" dir item which points to the root item that we
698 * will mount by default if we haven't been given a specific subvolume
701 dir_id = btrfs_super_root_dir(fs_info->super_copy);
702 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
704 btrfs_free_path(path);
709 * Ok the default dir item isn't there. This is weird since
710 * it's always been there, but don't freak out, just try and
711 * mount to root most subvolume.
713 btrfs_free_path(path);
714 dir_id = BTRFS_FIRST_FREE_OBJECTID;
715 new_root = fs_info->fs_root;
719 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
720 btrfs_free_path(path);
723 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
724 if (IS_ERR(new_root))
725 return ERR_CAST(new_root);
727 if (btrfs_root_refs(&new_root->root_item) == 0)
728 return ERR_PTR(-ENOENT);
730 dir_id = btrfs_root_dirid(&new_root->root_item);
732 location.objectid = dir_id;
733 location.type = BTRFS_INODE_ITEM_KEY;
736 inode = btrfs_iget(sb, &location, new_root, &new);
738 return ERR_CAST(inode);
741 * If we're just mounting the root most subvol put the inode and return
742 * a reference to the dentry. We will have already gotten a reference
743 * to the inode in btrfs_fill_super so we're good to go.
745 if (!new && sb->s_root->d_inode == inode) {
747 return dget(sb->s_root);
750 return d_obtain_alias(inode);
753 static int btrfs_fill_super(struct super_block *sb,
754 struct btrfs_fs_devices *fs_devices,
755 void *data, int silent)
758 struct dentry *root_dentry;
759 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
760 struct btrfs_key key;
763 sb->s_maxbytes = MAX_LFS_FILESIZE;
764 sb->s_magic = BTRFS_SUPER_MAGIC;
765 sb->s_op = &btrfs_super_ops;
766 sb->s_d_op = &btrfs_dentry_operations;
767 sb->s_export_op = &btrfs_export_ops;
768 sb->s_xattr = btrfs_xattr_handlers;
770 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
771 sb->s_flags |= MS_POSIXACL;
774 err = open_ctree(sb, fs_devices, (char *)data);
776 printk("btrfs: open_ctree failed\n");
780 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
781 key.type = BTRFS_INODE_ITEM_KEY;
783 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
785 err = PTR_ERR(inode);
789 root_dentry = d_alloc_root(inode);
796 sb->s_root = root_dentry;
798 save_mount_options(sb, data);
799 cleancache_init_fs(sb);
800 sb->s_flags |= MS_ACTIVE;
804 close_ctree(fs_info->tree_root);
808 int btrfs_sync_fs(struct super_block *sb, int wait)
810 struct btrfs_trans_handle *trans;
811 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
812 struct btrfs_root *root = fs_info->tree_root;
815 trace_btrfs_sync_fs(wait);
818 filemap_flush(fs_info->btree_inode->i_mapping);
822 btrfs_wait_ordered_extents(root, 0, 0);
824 trans = btrfs_start_transaction(root, 0);
826 return PTR_ERR(trans);
827 ret = btrfs_commit_transaction(trans, root);
831 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
833 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
834 struct btrfs_root *root = info->tree_root;
837 if (btrfs_test_opt(root, DEGRADED))
838 seq_puts(seq, ",degraded");
839 if (btrfs_test_opt(root, NODATASUM))
840 seq_puts(seq, ",nodatasum");
841 if (btrfs_test_opt(root, NODATACOW))
842 seq_puts(seq, ",nodatacow");
843 if (btrfs_test_opt(root, NOBARRIER))
844 seq_puts(seq, ",nobarrier");
845 if (info->max_inline != 8192 * 1024)
846 seq_printf(seq, ",max_inline=%llu",
847 (unsigned long long)info->max_inline);
848 if (info->alloc_start != 0)
849 seq_printf(seq, ",alloc_start=%llu",
850 (unsigned long long)info->alloc_start);
851 if (info->thread_pool_size != min_t(unsigned long,
852 num_online_cpus() + 2, 8))
853 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
854 if (btrfs_test_opt(root, COMPRESS)) {
855 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
856 compress_type = "zlib";
858 compress_type = "lzo";
859 if (btrfs_test_opt(root, FORCE_COMPRESS))
860 seq_printf(seq, ",compress-force=%s", compress_type);
862 seq_printf(seq, ",compress=%s", compress_type);
864 if (btrfs_test_opt(root, NOSSD))
865 seq_puts(seq, ",nossd");
866 if (btrfs_test_opt(root, SSD_SPREAD))
867 seq_puts(seq, ",ssd_spread");
868 else if (btrfs_test_opt(root, SSD))
869 seq_puts(seq, ",ssd");
870 if (btrfs_test_opt(root, NOTREELOG))
871 seq_puts(seq, ",notreelog");
872 if (btrfs_test_opt(root, FLUSHONCOMMIT))
873 seq_puts(seq, ",flushoncommit");
874 if (btrfs_test_opt(root, DISCARD))
875 seq_puts(seq, ",discard");
876 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
877 seq_puts(seq, ",noacl");
878 if (btrfs_test_opt(root, SPACE_CACHE))
879 seq_puts(seq, ",space_cache");
881 seq_puts(seq, ",nospace_cache");
882 if (btrfs_test_opt(root, CLEAR_CACHE))
883 seq_puts(seq, ",clear_cache");
884 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
885 seq_puts(seq, ",user_subvol_rm_allowed");
886 if (btrfs_test_opt(root, ENOSPC_DEBUG))
887 seq_puts(seq, ",enospc_debug");
888 if (btrfs_test_opt(root, AUTO_DEFRAG))
889 seq_puts(seq, ",autodefrag");
890 if (btrfs_test_opt(root, INODE_MAP_CACHE))
891 seq_puts(seq, ",inode_cache");
892 if (btrfs_test_opt(root, SKIP_BALANCE))
893 seq_puts(seq, ",skip_balance");
894 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
895 seq_puts(seq, ",fatal_errors=panic");
899 static int btrfs_test_super(struct super_block *s, void *data)
901 struct btrfs_fs_info *p = data;
902 struct btrfs_fs_info *fs_info = btrfs_sb(s);
904 return fs_info->fs_devices == p->fs_devices;
907 static int btrfs_set_super(struct super_block *s, void *data)
909 int err = set_anon_super(s, data);
916 * subvolumes are identified by ino 256
918 static inline int is_subvolume_inode(struct inode *inode)
920 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
926 * This will strip out the subvol=%s argument for an argument string and add
927 * subvolid=0 to make sure we get the actual tree root for path walking to the
930 static char *setup_root_args(char *args)
933 unsigned len = strlen(args) + 2;
938 * We need the same args as before, but minus
946 * which is a difference of 2 characters, so we allocate strlen(args) +
949 ret = kzalloc(len * sizeof(char), GFP_NOFS);
952 pos = strstr(args, "subvol=");
954 /* This shouldn't happen, but just in case.. */
961 * The subvol=<> arg is not at the front of the string, copy everybody
962 * up to that into ret.
967 copied += strlen(args);
971 strncpy(ret + copied, "subvolid=0", len - copied);
973 /* Length of subvolid=0 */
977 * If there is no , after the subvol= option then we know there's no
978 * other options and we can just return.
980 pos = strchr(pos, ',');
984 /* Copy the rest of the arguments into our buffer */
985 strncpy(ret + copied, pos, len - copied);
986 copied += strlen(pos);
991 static struct dentry *mount_subvol(const char *subvol_name, int flags,
992 const char *device_name, char *data)
995 struct vfsmount *mnt;
998 newargs = setup_root_args(data);
1000 return ERR_PTR(-ENOMEM);
1001 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1005 return ERR_CAST(mnt);
1007 root = mount_subtree(mnt, subvol_name);
1009 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1010 struct super_block *s = root->d_sb;
1012 root = ERR_PTR(-EINVAL);
1013 deactivate_locked_super(s);
1014 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
1022 * Find a superblock for the given device / mount point.
1024 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1025 * for multiple device setup. Make sure to keep it in sync.
1027 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1028 const char *device_name, void *data)
1030 struct block_device *bdev = NULL;
1031 struct super_block *s;
1032 struct dentry *root;
1033 struct btrfs_fs_devices *fs_devices = NULL;
1034 struct btrfs_fs_info *fs_info = NULL;
1035 fmode_t mode = FMODE_READ;
1036 char *subvol_name = NULL;
1037 u64 subvol_objectid = 0;
1038 u64 subvol_rootid = 0;
1041 if (!(flags & MS_RDONLY))
1042 mode |= FMODE_WRITE;
1044 error = btrfs_parse_early_options(data, mode, fs_type,
1045 &subvol_name, &subvol_objectid,
1046 &subvol_rootid, &fs_devices);
1049 return ERR_PTR(error);
1053 root = mount_subvol(subvol_name, flags, device_name, data);
1058 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1060 return ERR_PTR(error);
1063 * Setup a dummy root and fs_info for test/set super. This is because
1064 * we don't actually fill this stuff out until open_ctree, but we need
1065 * it for searching for existing supers, so this lets us do that and
1066 * then open_ctree will properly initialize everything later.
1068 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1070 return ERR_PTR(-ENOMEM);
1072 fs_info->fs_devices = fs_devices;
1074 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1075 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1076 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1081 error = btrfs_open_devices(fs_devices, mode, fs_type);
1085 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1087 goto error_close_devices;
1090 bdev = fs_devices->latest_bdev;
1091 s = sget(fs_type, btrfs_test_super, btrfs_set_super, fs_info);
1094 goto error_close_devices;
1098 btrfs_close_devices(fs_devices);
1099 free_fs_info(fs_info);
1100 if ((flags ^ s->s_flags) & MS_RDONLY)
1103 char b[BDEVNAME_SIZE];
1105 s->s_flags = flags | MS_NOSEC;
1106 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1107 btrfs_sb(s)->bdev_holder = fs_type;
1108 error = btrfs_fill_super(s, fs_devices, data,
1109 flags & MS_SILENT ? 1 : 0);
1112 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1114 deactivate_locked_super(s);
1118 error_close_devices:
1119 btrfs_close_devices(fs_devices);
1121 free_fs_info(fs_info);
1122 return ERR_PTR(error);
1125 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1127 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1128 struct btrfs_root *root = fs_info->tree_root;
1129 unsigned old_flags = sb->s_flags;
1130 unsigned long old_opts = fs_info->mount_opt;
1131 unsigned long old_compress_type = fs_info->compress_type;
1132 u64 old_max_inline = fs_info->max_inline;
1133 u64 old_alloc_start = fs_info->alloc_start;
1134 int old_thread_pool_size = fs_info->thread_pool_size;
1135 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1138 ret = btrfs_parse_options(root, data);
1144 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1147 if (*flags & MS_RDONLY) {
1148 sb->s_flags |= MS_RDONLY;
1150 ret = btrfs_commit_super(root);
1154 if (fs_info->fs_devices->rw_devices == 0) {
1159 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1164 ret = btrfs_cleanup_fs_roots(fs_info);
1168 /* recover relocation */
1169 ret = btrfs_recover_relocation(root);
1173 sb->s_flags &= ~MS_RDONLY;
1179 /* We've hit an error - don't reset MS_RDONLY */
1180 if (sb->s_flags & MS_RDONLY)
1181 old_flags |= MS_RDONLY;
1182 sb->s_flags = old_flags;
1183 fs_info->mount_opt = old_opts;
1184 fs_info->compress_type = old_compress_type;
1185 fs_info->max_inline = old_max_inline;
1186 fs_info->alloc_start = old_alloc_start;
1187 fs_info->thread_pool_size = old_thread_pool_size;
1188 fs_info->metadata_ratio = old_metadata_ratio;
1192 /* Used to sort the devices by max_avail(descending sort) */
1193 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1194 const void *dev_info2)
1196 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1197 ((struct btrfs_device_info *)dev_info2)->max_avail)
1199 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1200 ((struct btrfs_device_info *)dev_info2)->max_avail)
1207 * sort the devices by max_avail, in which max free extent size of each device
1208 * is stored.(Descending Sort)
1210 static inline void btrfs_descending_sort_devices(
1211 struct btrfs_device_info *devices,
1214 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1215 btrfs_cmp_device_free_bytes, NULL);
1219 * The helper to calc the free space on the devices that can be used to store
1222 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1224 struct btrfs_fs_info *fs_info = root->fs_info;
1225 struct btrfs_device_info *devices_info;
1226 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1227 struct btrfs_device *device;
1232 u64 min_stripe_size;
1233 int min_stripes = 1, num_stripes = 1;
1234 int i = 0, nr_devices;
1237 nr_devices = fs_info->fs_devices->open_devices;
1238 BUG_ON(!nr_devices);
1240 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1245 /* calc min stripe number for data space alloction */
1246 type = btrfs_get_alloc_profile(root, 1);
1247 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1249 num_stripes = nr_devices;
1250 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1253 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1258 if (type & BTRFS_BLOCK_GROUP_DUP)
1259 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1261 min_stripe_size = BTRFS_STRIPE_LEN;
1263 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1264 if (!device->in_fs_metadata || !device->bdev)
1267 avail_space = device->total_bytes - device->bytes_used;
1269 /* align with stripe_len */
1270 do_div(avail_space, BTRFS_STRIPE_LEN);
1271 avail_space *= BTRFS_STRIPE_LEN;
1274 * In order to avoid overwritting the superblock on the drive,
1275 * btrfs starts at an offset of at least 1MB when doing chunk
1278 skip_space = 1024 * 1024;
1280 /* user can set the offset in fs_info->alloc_start. */
1281 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1282 device->total_bytes)
1283 skip_space = max(fs_info->alloc_start, skip_space);
1286 * btrfs can not use the free space in [0, skip_space - 1],
1287 * we must subtract it from the total. In order to implement
1288 * it, we account the used space in this range first.
1290 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1293 kfree(devices_info);
1297 /* calc the free space in [0, skip_space - 1] */
1298 skip_space -= used_space;
1301 * we can use the free space in [0, skip_space - 1], subtract
1302 * it from the total.
1304 if (avail_space && avail_space >= skip_space)
1305 avail_space -= skip_space;
1309 if (avail_space < min_stripe_size)
1312 devices_info[i].dev = device;
1313 devices_info[i].max_avail = avail_space;
1320 btrfs_descending_sort_devices(devices_info, nr_devices);
1324 while (nr_devices >= min_stripes) {
1325 if (num_stripes > nr_devices)
1326 num_stripes = nr_devices;
1328 if (devices_info[i].max_avail >= min_stripe_size) {
1332 avail_space += devices_info[i].max_avail * num_stripes;
1333 alloc_size = devices_info[i].max_avail;
1334 for (j = i + 1 - num_stripes; j <= i; j++)
1335 devices_info[j].max_avail -= alloc_size;
1341 kfree(devices_info);
1342 *free_bytes = avail_space;
1346 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1348 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1349 struct btrfs_super_block *disk_super = fs_info->super_copy;
1350 struct list_head *head = &fs_info->space_info;
1351 struct btrfs_space_info *found;
1353 u64 total_free_data = 0;
1354 int bits = dentry->d_sb->s_blocksize_bits;
1355 __be32 *fsid = (__be32 *)fs_info->fsid;
1358 /* holding chunk_muext to avoid allocating new chunks */
1359 mutex_lock(&fs_info->chunk_mutex);
1361 list_for_each_entry_rcu(found, head, list) {
1362 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1363 total_free_data += found->disk_total - found->disk_used;
1365 btrfs_account_ro_block_groups_free_space(found);
1368 total_used += found->disk_used;
1372 buf->f_namelen = BTRFS_NAME_LEN;
1373 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1374 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1375 buf->f_bsize = dentry->d_sb->s_blocksize;
1376 buf->f_type = BTRFS_SUPER_MAGIC;
1377 buf->f_bavail = total_free_data;
1378 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1380 mutex_unlock(&fs_info->chunk_mutex);
1383 buf->f_bavail += total_free_data;
1384 buf->f_bavail = buf->f_bavail >> bits;
1385 mutex_unlock(&fs_info->chunk_mutex);
1387 /* We treat it as constant endianness (it doesn't matter _which_)
1388 because we want the fsid to come out the same whether mounted
1389 on a big-endian or little-endian host */
1390 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1391 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1392 /* Mask in the root object ID too, to disambiguate subvols */
1393 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1394 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1399 static void btrfs_kill_super(struct super_block *sb)
1401 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1402 kill_anon_super(sb);
1403 free_fs_info(fs_info);
1406 static struct file_system_type btrfs_fs_type = {
1407 .owner = THIS_MODULE,
1409 .mount = btrfs_mount,
1410 .kill_sb = btrfs_kill_super,
1411 .fs_flags = FS_REQUIRES_DEV,
1415 * used by btrfsctl to scan devices when no FS is mounted
1417 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1420 struct btrfs_ioctl_vol_args *vol;
1421 struct btrfs_fs_devices *fs_devices;
1424 if (!capable(CAP_SYS_ADMIN))
1427 vol = memdup_user((void __user *)arg, sizeof(*vol));
1429 return PTR_ERR(vol);
1432 case BTRFS_IOC_SCAN_DEV:
1433 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1434 &btrfs_fs_type, &fs_devices);
1442 static int btrfs_freeze(struct super_block *sb)
1444 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1445 mutex_lock(&fs_info->transaction_kthread_mutex);
1446 mutex_lock(&fs_info->cleaner_mutex);
1450 static int btrfs_unfreeze(struct super_block *sb)
1452 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1453 mutex_unlock(&fs_info->cleaner_mutex);
1454 mutex_unlock(&fs_info->transaction_kthread_mutex);
1458 static void btrfs_fs_dirty_inode(struct inode *inode, int flags)
1462 ret = btrfs_dirty_inode(inode);
1464 printk_ratelimited(KERN_ERR "btrfs: fail to dirty inode %Lu "
1465 "error %d\n", btrfs_ino(inode), ret);
1468 static const struct super_operations btrfs_super_ops = {
1469 .drop_inode = btrfs_drop_inode,
1470 .evict_inode = btrfs_evict_inode,
1471 .put_super = btrfs_put_super,
1472 .sync_fs = btrfs_sync_fs,
1473 .show_options = btrfs_show_options,
1474 .write_inode = btrfs_write_inode,
1475 .dirty_inode = btrfs_fs_dirty_inode,
1476 .alloc_inode = btrfs_alloc_inode,
1477 .destroy_inode = btrfs_destroy_inode,
1478 .statfs = btrfs_statfs,
1479 .remount_fs = btrfs_remount,
1480 .freeze_fs = btrfs_freeze,
1481 .unfreeze_fs = btrfs_unfreeze,
1484 static const struct file_operations btrfs_ctl_fops = {
1485 .unlocked_ioctl = btrfs_control_ioctl,
1486 .compat_ioctl = btrfs_control_ioctl,
1487 .owner = THIS_MODULE,
1488 .llseek = noop_llseek,
1491 static struct miscdevice btrfs_misc = {
1492 .minor = BTRFS_MINOR,
1493 .name = "btrfs-control",
1494 .fops = &btrfs_ctl_fops
1497 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1498 MODULE_ALIAS("devname:btrfs-control");
1500 static int btrfs_interface_init(void)
1502 return misc_register(&btrfs_misc);
1505 static void btrfs_interface_exit(void)
1507 if (misc_deregister(&btrfs_misc) < 0)
1508 printk(KERN_INFO "misc_deregister failed for control device");
1511 static int __init init_btrfs_fs(void)
1515 err = btrfs_init_sysfs();
1519 btrfs_init_compress();
1521 err = btrfs_init_cachep();
1525 err = extent_io_init();
1529 err = extent_map_init();
1531 goto free_extent_io;
1533 err = btrfs_delayed_inode_init();
1535 goto free_extent_map;
1537 err = btrfs_interface_init();
1539 goto free_delayed_inode;
1541 err = register_filesystem(&btrfs_fs_type);
1543 goto unregister_ioctl;
1545 btrfs_init_lockdep();
1547 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1551 btrfs_interface_exit();
1553 btrfs_delayed_inode_exit();
1559 btrfs_destroy_cachep();
1561 btrfs_exit_compress();
1566 static void __exit exit_btrfs_fs(void)
1568 btrfs_destroy_cachep();
1569 btrfs_delayed_inode_exit();
1572 btrfs_interface_exit();
1573 unregister_filesystem(&btrfs_fs_type);
1575 btrfs_cleanup_fs_uuids();
1576 btrfs_exit_compress();
1579 module_init(init_btrfs_fs)
1580 module_exit(exit_btrfs_fs)
1582 MODULE_LICENSE("GPL");