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/mnt_namespace.h>
44 #include <linux/ratelimit.h>
46 #include "delayed-inode.h"
49 #include "transaction.h"
50 #include "btrfs_inode.h"
52 #include "print-tree.h"
57 #include "compression.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";
82 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
91 static void __save_error_info(struct btrfs_fs_info *fs_info)
94 * today we only save the error info into ram. Long term we'll
95 * also send it down to the disk
97 fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
101 * We move write_super stuff at umount in order to avoid deadlock
102 * for umount hold all lock.
104 static void save_error_info(struct btrfs_fs_info *fs_info)
106 __save_error_info(fs_info);
109 /* btrfs handle error by forcing the filesystem readonly */
110 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
112 struct super_block *sb = fs_info->sb;
114 if (sb->s_flags & MS_RDONLY)
117 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
118 sb->s_flags |= MS_RDONLY;
119 printk(KERN_INFO "btrfs is forced readonly\n");
124 * __btrfs_std_error decodes expected errors from the caller and
125 * invokes the approciate error response.
127 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
128 unsigned int line, int errno)
130 struct super_block *sb = fs_info->sb;
135 * Special case: if the error is EROFS, and we're already
136 * under MS_RDONLY, then it is safe here.
138 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
141 errstr = btrfs_decode_error(fs_info, errno, nbuf);
142 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
143 sb->s_id, function, line, errstr);
144 save_error_info(fs_info);
146 btrfs_handle_error(fs_info);
149 static void btrfs_put_super(struct super_block *sb)
151 struct btrfs_root *root = btrfs_sb(sb);
154 ret = close_ctree(root);
155 sb->s_fs_info = NULL;
157 (void)ret; /* FIXME: need to fix VFS to return error? */
161 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
162 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
163 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
164 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
165 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
166 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
167 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
168 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
169 Opt_check_integrity, Opt_check_integrity_including_extent_data,
170 Opt_check_integrity_print_mask,
174 static match_table_t tokens = {
175 {Opt_degraded, "degraded"},
176 {Opt_subvol, "subvol=%s"},
177 {Opt_subvolid, "subvolid=%d"},
178 {Opt_device, "device=%s"},
179 {Opt_nodatasum, "nodatasum"},
180 {Opt_nodatacow, "nodatacow"},
181 {Opt_nobarrier, "nobarrier"},
182 {Opt_max_inline, "max_inline=%s"},
183 {Opt_alloc_start, "alloc_start=%s"},
184 {Opt_thread_pool, "thread_pool=%d"},
185 {Opt_compress, "compress"},
186 {Opt_compress_type, "compress=%s"},
187 {Opt_compress_force, "compress-force"},
188 {Opt_compress_force_type, "compress-force=%s"},
190 {Opt_ssd_spread, "ssd_spread"},
191 {Opt_nossd, "nossd"},
192 {Opt_noacl, "noacl"},
193 {Opt_notreelog, "notreelog"},
194 {Opt_flushoncommit, "flushoncommit"},
195 {Opt_ratio, "metadata_ratio=%d"},
196 {Opt_discard, "discard"},
197 {Opt_space_cache, "space_cache"},
198 {Opt_clear_cache, "clear_cache"},
199 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
200 {Opt_enospc_debug, "enospc_debug"},
201 {Opt_subvolrootid, "subvolrootid=%d"},
202 {Opt_defrag, "autodefrag"},
203 {Opt_inode_cache, "inode_cache"},
204 {Opt_no_space_cache, "nospace_cache"},
205 {Opt_recovery, "recovery"},
206 {Opt_skip_balance, "skip_balance"},
207 {Opt_check_integrity, "check_int"},
208 {Opt_check_integrity_including_extent_data, "check_int_data"},
209 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
214 * Regular mount options parser. Everything that is needed only when
215 * reading in a new superblock is parsed here.
217 int btrfs_parse_options(struct btrfs_root *root, char *options)
219 struct btrfs_fs_info *info = root->fs_info;
220 substring_t args[MAX_OPT_ARGS];
221 char *p, *num, *orig = NULL;
226 bool compress_force = false;
228 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
230 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
236 * strsep changes the string, duplicate it because parse_options
239 options = kstrdup(options, GFP_NOFS);
245 while ((p = strsep(&options, ",")) != NULL) {
250 token = match_token(p, tokens, args);
253 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
254 btrfs_set_opt(info->mount_opt, DEGRADED);
258 case Opt_subvolrootid:
261 * These are parsed by btrfs_parse_early_options
262 * and can be happily ignored here.
266 printk(KERN_INFO "btrfs: setting nodatasum\n");
267 btrfs_set_opt(info->mount_opt, NODATASUM);
270 printk(KERN_INFO "btrfs: setting nodatacow\n");
271 btrfs_set_opt(info->mount_opt, NODATACOW);
272 btrfs_set_opt(info->mount_opt, NODATASUM);
274 case Opt_compress_force:
275 case Opt_compress_force_type:
276 compress_force = true;
278 case Opt_compress_type:
279 if (token == Opt_compress ||
280 token == Opt_compress_force ||
281 strcmp(args[0].from, "zlib") == 0) {
282 compress_type = "zlib";
283 info->compress_type = BTRFS_COMPRESS_ZLIB;
284 } else if (strcmp(args[0].from, "lzo") == 0) {
285 compress_type = "lzo";
286 info->compress_type = BTRFS_COMPRESS_LZO;
292 btrfs_set_opt(info->mount_opt, COMPRESS);
293 if (compress_force) {
294 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
295 pr_info("btrfs: force %s compression\n",
298 pr_info("btrfs: use %s compression\n",
302 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
303 btrfs_set_opt(info->mount_opt, SSD);
306 printk(KERN_INFO "btrfs: use spread ssd "
307 "allocation scheme\n");
308 btrfs_set_opt(info->mount_opt, SSD);
309 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
312 printk(KERN_INFO "btrfs: not using ssd allocation "
314 btrfs_set_opt(info->mount_opt, NOSSD);
315 btrfs_clear_opt(info->mount_opt, SSD);
316 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
319 printk(KERN_INFO "btrfs: turning off barriers\n");
320 btrfs_set_opt(info->mount_opt, NOBARRIER);
322 case Opt_thread_pool:
324 match_int(&args[0], &intarg);
326 info->thread_pool_size = intarg;
327 printk(KERN_INFO "btrfs: thread pool %d\n",
328 info->thread_pool_size);
332 num = match_strdup(&args[0]);
334 info->max_inline = memparse(num, NULL);
337 if (info->max_inline) {
338 info->max_inline = max_t(u64,
342 printk(KERN_INFO "btrfs: max_inline at %llu\n",
343 (unsigned long long)info->max_inline);
346 case Opt_alloc_start:
347 num = match_strdup(&args[0]);
349 info->alloc_start = memparse(num, NULL);
352 "btrfs: allocations start at %llu\n",
353 (unsigned long long)info->alloc_start);
357 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
360 printk(KERN_INFO "btrfs: disabling tree log\n");
361 btrfs_set_opt(info->mount_opt, NOTREELOG);
363 case Opt_flushoncommit:
364 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
365 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
369 match_int(&args[0], &intarg);
371 info->metadata_ratio = intarg;
372 printk(KERN_INFO "btrfs: metadata ratio %d\n",
373 info->metadata_ratio);
377 btrfs_set_opt(info->mount_opt, DISCARD);
379 case Opt_space_cache:
380 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
382 case Opt_no_space_cache:
383 printk(KERN_INFO "btrfs: disabling disk space caching\n");
384 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
386 case Opt_inode_cache:
387 printk(KERN_INFO "btrfs: enabling inode map caching\n");
388 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
390 case Opt_clear_cache:
391 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
392 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
394 case Opt_user_subvol_rm_allowed:
395 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
397 case Opt_enospc_debug:
398 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
401 printk(KERN_INFO "btrfs: enabling auto defrag");
402 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
405 printk(KERN_INFO "btrfs: enabling auto recovery");
406 btrfs_set_opt(info->mount_opt, RECOVERY);
408 case Opt_skip_balance:
409 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
411 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
412 case Opt_check_integrity_including_extent_data:
413 printk(KERN_INFO "btrfs: enabling check integrity"
414 " including extent data\n");
415 btrfs_set_opt(info->mount_opt,
416 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
417 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
419 case Opt_check_integrity:
420 printk(KERN_INFO "btrfs: enabling check integrity\n");
421 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
423 case Opt_check_integrity_print_mask:
425 match_int(&args[0], &intarg);
427 info->check_integrity_print_mask = intarg;
428 printk(KERN_INFO "btrfs:"
429 " check_integrity_print_mask 0x%x\n",
430 info->check_integrity_print_mask);
434 case Opt_check_integrity_including_extent_data:
435 case Opt_check_integrity:
436 case Opt_check_integrity_print_mask:
437 printk(KERN_ERR "btrfs: support for check_integrity*"
438 " not compiled in!\n");
443 printk(KERN_INFO "btrfs: unrecognized mount option "
452 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
453 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
459 * Parse mount options that are required early in the mount process.
461 * All other options will be parsed on much later in the mount process and
462 * only when we need to allocate a new super block.
464 static int btrfs_parse_early_options(const char *options, fmode_t flags,
465 void *holder, char **subvol_name, u64 *subvol_objectid,
466 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
468 substring_t args[MAX_OPT_ARGS];
469 char *device_name, *opts, *orig, *p;
477 * strsep changes the string, duplicate it because parse_options
480 opts = kstrdup(options, GFP_KERNEL);
485 while ((p = strsep(&opts, ",")) != NULL) {
490 token = match_token(p, tokens, args);
494 *subvol_name = match_strdup(&args[0]);
498 error = match_int(&args[0], &intarg);
500 /* we want the original fs_tree */
503 BTRFS_FS_TREE_OBJECTID;
505 *subvol_objectid = intarg;
508 case Opt_subvolrootid:
510 error = match_int(&args[0], &intarg);
512 /* we want the original fs_tree */
515 BTRFS_FS_TREE_OBJECTID;
517 *subvol_rootid = intarg;
521 device_name = match_strdup(&args[0]);
526 error = btrfs_scan_one_device(device_name,
527 flags, holder, fs_devices);
542 static struct dentry *get_default_root(struct super_block *sb,
545 struct btrfs_root *root = sb->s_fs_info;
546 struct btrfs_root *new_root;
547 struct btrfs_dir_item *di;
548 struct btrfs_path *path;
549 struct btrfs_key location;
555 * We have a specific subvol we want to mount, just setup location and
556 * go look up the root.
558 if (subvol_objectid) {
559 location.objectid = subvol_objectid;
560 location.type = BTRFS_ROOT_ITEM_KEY;
561 location.offset = (u64)-1;
565 path = btrfs_alloc_path();
567 return ERR_PTR(-ENOMEM);
568 path->leave_spinning = 1;
571 * Find the "default" dir item which points to the root item that we
572 * will mount by default if we haven't been given a specific subvolume
575 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
576 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
578 btrfs_free_path(path);
583 * Ok the default dir item isn't there. This is weird since
584 * it's always been there, but don't freak out, just try and
585 * mount to root most subvolume.
587 btrfs_free_path(path);
588 dir_id = BTRFS_FIRST_FREE_OBJECTID;
589 new_root = root->fs_info->fs_root;
593 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
594 btrfs_free_path(path);
597 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
598 if (IS_ERR(new_root))
599 return ERR_CAST(new_root);
601 if (btrfs_root_refs(&new_root->root_item) == 0)
602 return ERR_PTR(-ENOENT);
604 dir_id = btrfs_root_dirid(&new_root->root_item);
606 location.objectid = dir_id;
607 location.type = BTRFS_INODE_ITEM_KEY;
610 inode = btrfs_iget(sb, &location, new_root, &new);
612 return ERR_CAST(inode);
615 * If we're just mounting the root most subvol put the inode and return
616 * a reference to the dentry. We will have already gotten a reference
617 * to the inode in btrfs_fill_super so we're good to go.
619 if (!new && sb->s_root->d_inode == inode) {
621 return dget(sb->s_root);
624 return d_obtain_alias(inode);
627 static int btrfs_fill_super(struct super_block *sb,
628 struct btrfs_fs_devices *fs_devices,
629 void *data, int silent)
632 struct dentry *root_dentry;
633 struct btrfs_root *tree_root;
634 struct btrfs_key key;
637 sb->s_maxbytes = MAX_LFS_FILESIZE;
638 sb->s_magic = BTRFS_SUPER_MAGIC;
639 sb->s_op = &btrfs_super_ops;
640 sb->s_d_op = &btrfs_dentry_operations;
641 sb->s_export_op = &btrfs_export_ops;
642 sb->s_xattr = btrfs_xattr_handlers;
644 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
645 sb->s_flags |= MS_POSIXACL;
648 tree_root = open_ctree(sb, fs_devices, (char *)data);
650 if (IS_ERR(tree_root)) {
651 printk("btrfs: open_ctree failed\n");
652 return PTR_ERR(tree_root);
654 sb->s_fs_info = tree_root;
656 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
657 key.type = BTRFS_INODE_ITEM_KEY;
659 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
661 err = PTR_ERR(inode);
665 root_dentry = d_alloc_root(inode);
672 sb->s_root = root_dentry;
674 save_mount_options(sb, data);
675 cleancache_init_fs(sb);
679 close_ctree(tree_root);
683 int btrfs_sync_fs(struct super_block *sb, int wait)
685 struct btrfs_trans_handle *trans;
686 struct btrfs_root *root = btrfs_sb(sb);
689 trace_btrfs_sync_fs(wait);
692 filemap_flush(root->fs_info->btree_inode->i_mapping);
696 btrfs_start_delalloc_inodes(root, 0);
697 btrfs_wait_ordered_extents(root, 0, 0);
699 trans = btrfs_start_transaction(root, 0);
701 return PTR_ERR(trans);
702 ret = btrfs_commit_transaction(trans, root);
706 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
708 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
709 struct btrfs_fs_info *info = root->fs_info;
712 if (btrfs_test_opt(root, DEGRADED))
713 seq_puts(seq, ",degraded");
714 if (btrfs_test_opt(root, NODATASUM))
715 seq_puts(seq, ",nodatasum");
716 if (btrfs_test_opt(root, NODATACOW))
717 seq_puts(seq, ",nodatacow");
718 if (btrfs_test_opt(root, NOBARRIER))
719 seq_puts(seq, ",nobarrier");
720 if (info->max_inline != 8192 * 1024)
721 seq_printf(seq, ",max_inline=%llu",
722 (unsigned long long)info->max_inline);
723 if (info->alloc_start != 0)
724 seq_printf(seq, ",alloc_start=%llu",
725 (unsigned long long)info->alloc_start);
726 if (info->thread_pool_size != min_t(unsigned long,
727 num_online_cpus() + 2, 8))
728 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
729 if (btrfs_test_opt(root, COMPRESS)) {
730 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
731 compress_type = "zlib";
733 compress_type = "lzo";
734 if (btrfs_test_opt(root, FORCE_COMPRESS))
735 seq_printf(seq, ",compress-force=%s", compress_type);
737 seq_printf(seq, ",compress=%s", compress_type);
739 if (btrfs_test_opt(root, NOSSD))
740 seq_puts(seq, ",nossd");
741 if (btrfs_test_opt(root, SSD_SPREAD))
742 seq_puts(seq, ",ssd_spread");
743 else if (btrfs_test_opt(root, SSD))
744 seq_puts(seq, ",ssd");
745 if (btrfs_test_opt(root, NOTREELOG))
746 seq_puts(seq, ",notreelog");
747 if (btrfs_test_opt(root, FLUSHONCOMMIT))
748 seq_puts(seq, ",flushoncommit");
749 if (btrfs_test_opt(root, DISCARD))
750 seq_puts(seq, ",discard");
751 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
752 seq_puts(seq, ",noacl");
753 if (btrfs_test_opt(root, SPACE_CACHE))
754 seq_puts(seq, ",space_cache");
756 seq_puts(seq, ",nospace_cache");
757 if (btrfs_test_opt(root, CLEAR_CACHE))
758 seq_puts(seq, ",clear_cache");
759 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
760 seq_puts(seq, ",user_subvol_rm_allowed");
761 if (btrfs_test_opt(root, ENOSPC_DEBUG))
762 seq_puts(seq, ",enospc_debug");
763 if (btrfs_test_opt(root, AUTO_DEFRAG))
764 seq_puts(seq, ",autodefrag");
765 if (btrfs_test_opt(root, INODE_MAP_CACHE))
766 seq_puts(seq, ",inode_cache");
767 if (btrfs_test_opt(root, SKIP_BALANCE))
768 seq_puts(seq, ",skip_balance");
772 static int btrfs_test_super(struct super_block *s, void *data)
774 struct btrfs_root *test_root = data;
775 struct btrfs_root *root = btrfs_sb(s);
778 * If this super block is going away, return false as it
779 * can't match as an existing super block.
781 if (!atomic_read(&s->s_active))
783 return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
786 static int btrfs_set_super(struct super_block *s, void *data)
790 return set_anon_super(s, data);
794 * subvolumes are identified by ino 256
796 static inline int is_subvolume_inode(struct inode *inode)
798 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
804 * This will strip out the subvol=%s argument for an argument string and add
805 * subvolid=0 to make sure we get the actual tree root for path walking to the
808 static char *setup_root_args(char *args)
811 unsigned len = strlen(args) + 2;
816 * We need the same args as before, but minus
824 * which is a difference of 2 characters, so we allocate strlen(args) +
827 ret = kzalloc(len * sizeof(char), GFP_NOFS);
830 pos = strstr(args, "subvol=");
832 /* This shouldn't happen, but just in case.. */
839 * The subvol=<> arg is not at the front of the string, copy everybody
840 * up to that into ret.
845 copied += strlen(args);
849 strncpy(ret + copied, "subvolid=0", len - copied);
851 /* Length of subvolid=0 */
855 * If there is no , after the subvol= option then we know there's no
856 * other options and we can just return.
858 pos = strchr(pos, ',');
862 /* Copy the rest of the arguments into our buffer */
863 strncpy(ret + copied, pos, len - copied);
864 copied += strlen(pos);
869 static struct dentry *mount_subvol(const char *subvol_name, int flags,
870 const char *device_name, char *data)
873 struct vfsmount *mnt;
876 newargs = setup_root_args(data);
878 return ERR_PTR(-ENOMEM);
879 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
883 return ERR_CAST(mnt);
885 root = mount_subtree(mnt, subvol_name);
887 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
888 struct super_block *s = root->d_sb;
890 root = ERR_PTR(-EINVAL);
891 deactivate_locked_super(s);
892 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
900 * Find a superblock for the given device / mount point.
902 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
903 * for multiple device setup. Make sure to keep it in sync.
905 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
906 const char *device_name, void *data)
908 struct block_device *bdev = NULL;
909 struct super_block *s;
911 struct btrfs_fs_devices *fs_devices = NULL;
912 struct btrfs_fs_info *fs_info = NULL;
913 fmode_t mode = FMODE_READ;
914 char *subvol_name = NULL;
915 u64 subvol_objectid = 0;
916 u64 subvol_rootid = 0;
919 if (!(flags & MS_RDONLY))
922 error = btrfs_parse_early_options(data, mode, fs_type,
923 &subvol_name, &subvol_objectid,
924 &subvol_rootid, &fs_devices);
927 return ERR_PTR(error);
931 root = mount_subvol(subvol_name, flags, device_name, data);
936 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
938 return ERR_PTR(error);
941 * Setup a dummy root and fs_info for test/set super. This is because
942 * we don't actually fill this stuff out until open_ctree, but we need
943 * it for searching for existing supers, so this lets us do that and
944 * then open_ctree will properly initialize everything later.
946 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
948 return ERR_PTR(-ENOMEM);
950 fs_info->tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
951 if (!fs_info->tree_root) {
955 fs_info->tree_root->fs_info = fs_info;
956 fs_info->fs_devices = fs_devices;
958 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
959 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
960 if (!fs_info->super_copy || !fs_info->super_for_commit) {
965 error = btrfs_open_devices(fs_devices, mode, fs_type);
969 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
971 goto error_close_devices;
974 bdev = fs_devices->latest_bdev;
975 s = sget(fs_type, btrfs_test_super, btrfs_set_super,
979 goto error_close_devices;
983 if ((flags ^ s->s_flags) & MS_RDONLY) {
984 deactivate_locked_super(s);
986 goto error_close_devices;
989 btrfs_close_devices(fs_devices);
990 free_fs_info(fs_info);
992 char b[BDEVNAME_SIZE];
994 s->s_flags = flags | MS_NOSEC;
995 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
996 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
997 error = btrfs_fill_super(s, fs_devices, data,
998 flags & MS_SILENT ? 1 : 0);
1000 deactivate_locked_super(s);
1001 return ERR_PTR(error);
1004 s->s_flags |= MS_ACTIVE;
1007 root = get_default_root(s, subvol_objectid);
1009 deactivate_locked_super(s);
1015 error_close_devices:
1016 btrfs_close_devices(fs_devices);
1018 free_fs_info(fs_info);
1019 return ERR_PTR(error);
1022 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1024 struct btrfs_root *root = btrfs_sb(sb);
1027 ret = btrfs_parse_options(root, data);
1031 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1034 if (*flags & MS_RDONLY) {
1035 sb->s_flags |= MS_RDONLY;
1037 ret = btrfs_commit_super(root);
1040 if (root->fs_info->fs_devices->rw_devices == 0)
1043 if (btrfs_super_log_root(root->fs_info->super_copy) != 0)
1046 ret = btrfs_cleanup_fs_roots(root->fs_info);
1049 /* recover relocation */
1050 ret = btrfs_recover_relocation(root);
1053 sb->s_flags &= ~MS_RDONLY;
1059 /* Used to sort the devices by max_avail(descending sort) */
1060 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1061 const void *dev_info2)
1063 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1064 ((struct btrfs_device_info *)dev_info2)->max_avail)
1066 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1067 ((struct btrfs_device_info *)dev_info2)->max_avail)
1074 * sort the devices by max_avail, in which max free extent size of each device
1075 * is stored.(Descending Sort)
1077 static inline void btrfs_descending_sort_devices(
1078 struct btrfs_device_info *devices,
1081 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1082 btrfs_cmp_device_free_bytes, NULL);
1086 * The helper to calc the free space on the devices that can be used to store
1089 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1091 struct btrfs_fs_info *fs_info = root->fs_info;
1092 struct btrfs_device_info *devices_info;
1093 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1094 struct btrfs_device *device;
1099 u64 min_stripe_size;
1100 int min_stripes = 1, num_stripes = 1;
1101 int i = 0, nr_devices;
1104 nr_devices = fs_info->fs_devices->open_devices;
1105 BUG_ON(!nr_devices);
1107 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1112 /* calc min stripe number for data space alloction */
1113 type = btrfs_get_alloc_profile(root, 1);
1114 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1116 num_stripes = nr_devices;
1117 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1120 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1125 if (type & BTRFS_BLOCK_GROUP_DUP)
1126 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1128 min_stripe_size = BTRFS_STRIPE_LEN;
1130 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1131 if (!device->in_fs_metadata || !device->bdev)
1134 avail_space = device->total_bytes - device->bytes_used;
1136 /* align with stripe_len */
1137 do_div(avail_space, BTRFS_STRIPE_LEN);
1138 avail_space *= BTRFS_STRIPE_LEN;
1141 * In order to avoid overwritting the superblock on the drive,
1142 * btrfs starts at an offset of at least 1MB when doing chunk
1145 skip_space = 1024 * 1024;
1147 /* user can set the offset in fs_info->alloc_start. */
1148 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1149 device->total_bytes)
1150 skip_space = max(fs_info->alloc_start, skip_space);
1153 * btrfs can not use the free space in [0, skip_space - 1],
1154 * we must subtract it from the total. In order to implement
1155 * it, we account the used space in this range first.
1157 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1160 kfree(devices_info);
1164 /* calc the free space in [0, skip_space - 1] */
1165 skip_space -= used_space;
1168 * we can use the free space in [0, skip_space - 1], subtract
1169 * it from the total.
1171 if (avail_space && avail_space >= skip_space)
1172 avail_space -= skip_space;
1176 if (avail_space < min_stripe_size)
1179 devices_info[i].dev = device;
1180 devices_info[i].max_avail = avail_space;
1187 btrfs_descending_sort_devices(devices_info, nr_devices);
1191 while (nr_devices >= min_stripes) {
1192 if (num_stripes > nr_devices)
1193 num_stripes = nr_devices;
1195 if (devices_info[i].max_avail >= min_stripe_size) {
1199 avail_space += devices_info[i].max_avail * num_stripes;
1200 alloc_size = devices_info[i].max_avail;
1201 for (j = i + 1 - num_stripes; j <= i; j++)
1202 devices_info[j].max_avail -= alloc_size;
1208 kfree(devices_info);
1209 *free_bytes = avail_space;
1213 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1215 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
1216 struct btrfs_super_block *disk_super = root->fs_info->super_copy;
1217 struct list_head *head = &root->fs_info->space_info;
1218 struct btrfs_space_info *found;
1220 u64 total_free_data = 0;
1221 int bits = dentry->d_sb->s_blocksize_bits;
1222 __be32 *fsid = (__be32 *)root->fs_info->fsid;
1225 /* holding chunk_muext to avoid allocating new chunks */
1226 mutex_lock(&root->fs_info->chunk_mutex);
1228 list_for_each_entry_rcu(found, head, list) {
1229 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1230 total_free_data += found->disk_total - found->disk_used;
1232 btrfs_account_ro_block_groups_free_space(found);
1235 total_used += found->disk_used;
1239 buf->f_namelen = BTRFS_NAME_LEN;
1240 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1241 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1242 buf->f_bsize = dentry->d_sb->s_blocksize;
1243 buf->f_type = BTRFS_SUPER_MAGIC;
1244 buf->f_bavail = total_free_data;
1245 ret = btrfs_calc_avail_data_space(root, &total_free_data);
1247 mutex_unlock(&root->fs_info->chunk_mutex);
1250 buf->f_bavail += total_free_data;
1251 buf->f_bavail = buf->f_bavail >> bits;
1252 mutex_unlock(&root->fs_info->chunk_mutex);
1254 /* We treat it as constant endianness (it doesn't matter _which_)
1255 because we want the fsid to come out the same whether mounted
1256 on a big-endian or little-endian host */
1257 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1258 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1259 /* Mask in the root object ID too, to disambiguate subvols */
1260 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1261 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1266 static struct file_system_type btrfs_fs_type = {
1267 .owner = THIS_MODULE,
1269 .mount = btrfs_mount,
1270 .kill_sb = kill_anon_super,
1271 .fs_flags = FS_REQUIRES_DEV,
1275 * used by btrfsctl to scan devices when no FS is mounted
1277 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1280 struct btrfs_ioctl_vol_args *vol;
1281 struct btrfs_fs_devices *fs_devices;
1284 if (!capable(CAP_SYS_ADMIN))
1287 vol = memdup_user((void __user *)arg, sizeof(*vol));
1289 return PTR_ERR(vol);
1292 case BTRFS_IOC_SCAN_DEV:
1293 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1294 &btrfs_fs_type, &fs_devices);
1302 static int btrfs_freeze(struct super_block *sb)
1304 struct btrfs_root *root = btrfs_sb(sb);
1305 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1306 mutex_lock(&root->fs_info->cleaner_mutex);
1310 static int btrfs_unfreeze(struct super_block *sb)
1312 struct btrfs_root *root = btrfs_sb(sb);
1313 mutex_unlock(&root->fs_info->cleaner_mutex);
1314 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1318 static void btrfs_fs_dirty_inode(struct inode *inode, int flags)
1322 ret = btrfs_dirty_inode(inode);
1324 printk_ratelimited(KERN_ERR "btrfs: fail to dirty inode %Lu "
1325 "error %d\n", btrfs_ino(inode), ret);
1328 static const struct super_operations btrfs_super_ops = {
1329 .drop_inode = btrfs_drop_inode,
1330 .evict_inode = btrfs_evict_inode,
1331 .put_super = btrfs_put_super,
1332 .sync_fs = btrfs_sync_fs,
1333 .show_options = btrfs_show_options,
1334 .write_inode = btrfs_write_inode,
1335 .dirty_inode = btrfs_fs_dirty_inode,
1336 .alloc_inode = btrfs_alloc_inode,
1337 .destroy_inode = btrfs_destroy_inode,
1338 .statfs = btrfs_statfs,
1339 .remount_fs = btrfs_remount,
1340 .freeze_fs = btrfs_freeze,
1341 .unfreeze_fs = btrfs_unfreeze,
1344 static const struct file_operations btrfs_ctl_fops = {
1345 .unlocked_ioctl = btrfs_control_ioctl,
1346 .compat_ioctl = btrfs_control_ioctl,
1347 .owner = THIS_MODULE,
1348 .llseek = noop_llseek,
1351 static struct miscdevice btrfs_misc = {
1352 .minor = BTRFS_MINOR,
1353 .name = "btrfs-control",
1354 .fops = &btrfs_ctl_fops
1357 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1358 MODULE_ALIAS("devname:btrfs-control");
1360 static int btrfs_interface_init(void)
1362 return misc_register(&btrfs_misc);
1365 static void btrfs_interface_exit(void)
1367 if (misc_deregister(&btrfs_misc) < 0)
1368 printk(KERN_INFO "misc_deregister failed for control device");
1371 static int __init init_btrfs_fs(void)
1375 err = btrfs_init_sysfs();
1379 err = btrfs_init_compress();
1383 err = btrfs_init_cachep();
1387 err = extent_io_init();
1391 err = extent_map_init();
1393 goto free_extent_io;
1395 err = btrfs_delayed_inode_init();
1397 goto free_extent_map;
1399 err = btrfs_interface_init();
1401 goto free_delayed_inode;
1403 err = register_filesystem(&btrfs_fs_type);
1405 goto unregister_ioctl;
1407 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1411 btrfs_interface_exit();
1413 btrfs_delayed_inode_exit();
1419 btrfs_destroy_cachep();
1421 btrfs_exit_compress();
1427 static void __exit exit_btrfs_fs(void)
1429 btrfs_destroy_cachep();
1430 btrfs_delayed_inode_exit();
1433 btrfs_interface_exit();
1434 unregister_filesystem(&btrfs_fs_type);
1436 btrfs_cleanup_fs_uuids();
1437 btrfs_exit_compress();
1440 module_init(init_btrfs_fs)
1441 module_exit(exit_btrfs_fs)
1443 MODULE_LICENSE("GPL");