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,
172 static match_table_t tokens = {
173 {Opt_degraded, "degraded"},
174 {Opt_subvol, "subvol=%s"},
175 {Opt_subvolid, "subvolid=%d"},
176 {Opt_device, "device=%s"},
177 {Opt_nodatasum, "nodatasum"},
178 {Opt_nodatacow, "nodatacow"},
179 {Opt_nobarrier, "nobarrier"},
180 {Opt_max_inline, "max_inline=%s"},
181 {Opt_alloc_start, "alloc_start=%s"},
182 {Opt_thread_pool, "thread_pool=%d"},
183 {Opt_compress, "compress"},
184 {Opt_compress_type, "compress=%s"},
185 {Opt_compress_force, "compress-force"},
186 {Opt_compress_force_type, "compress-force=%s"},
188 {Opt_ssd_spread, "ssd_spread"},
189 {Opt_nossd, "nossd"},
190 {Opt_noacl, "noacl"},
191 {Opt_notreelog, "notreelog"},
192 {Opt_flushoncommit, "flushoncommit"},
193 {Opt_ratio, "metadata_ratio=%d"},
194 {Opt_discard, "discard"},
195 {Opt_space_cache, "space_cache"},
196 {Opt_clear_cache, "clear_cache"},
197 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
198 {Opt_enospc_debug, "enospc_debug"},
199 {Opt_subvolrootid, "subvolrootid=%d"},
200 {Opt_defrag, "autodefrag"},
201 {Opt_inode_cache, "inode_cache"},
202 {Opt_no_space_cache, "nospace_cache"},
203 {Opt_recovery, "recovery"},
204 {Opt_skip_balance, "skip_balance"},
209 * Regular mount options parser. Everything that is needed only when
210 * reading in a new superblock is parsed here.
212 int btrfs_parse_options(struct btrfs_root *root, char *options)
214 struct btrfs_fs_info *info = root->fs_info;
215 substring_t args[MAX_OPT_ARGS];
216 char *p, *num, *orig = NULL;
221 bool compress_force = false;
223 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
225 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
231 * strsep changes the string, duplicate it because parse_options
234 options = kstrdup(options, GFP_NOFS);
240 while ((p = strsep(&options, ",")) != NULL) {
245 token = match_token(p, tokens, args);
248 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
249 btrfs_set_opt(info->mount_opt, DEGRADED);
253 case Opt_subvolrootid:
256 * These are parsed by btrfs_parse_early_options
257 * and can be happily ignored here.
261 printk(KERN_INFO "btrfs: setting nodatasum\n");
262 btrfs_set_opt(info->mount_opt, NODATASUM);
265 printk(KERN_INFO "btrfs: setting nodatacow\n");
266 btrfs_set_opt(info->mount_opt, NODATACOW);
267 btrfs_set_opt(info->mount_opt, NODATASUM);
269 case Opt_compress_force:
270 case Opt_compress_force_type:
271 compress_force = true;
273 case Opt_compress_type:
274 if (token == Opt_compress ||
275 token == Opt_compress_force ||
276 strcmp(args[0].from, "zlib") == 0) {
277 compress_type = "zlib";
278 info->compress_type = BTRFS_COMPRESS_ZLIB;
279 } else if (strcmp(args[0].from, "lzo") == 0) {
280 compress_type = "lzo";
281 info->compress_type = BTRFS_COMPRESS_LZO;
287 btrfs_set_opt(info->mount_opt, COMPRESS);
288 if (compress_force) {
289 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
290 pr_info("btrfs: force %s compression\n",
293 pr_info("btrfs: use %s compression\n",
297 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
298 btrfs_set_opt(info->mount_opt, SSD);
301 printk(KERN_INFO "btrfs: use spread ssd "
302 "allocation scheme\n");
303 btrfs_set_opt(info->mount_opt, SSD);
304 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
307 printk(KERN_INFO "btrfs: not using ssd allocation "
309 btrfs_set_opt(info->mount_opt, NOSSD);
310 btrfs_clear_opt(info->mount_opt, SSD);
311 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
314 printk(KERN_INFO "btrfs: turning off barriers\n");
315 btrfs_set_opt(info->mount_opt, NOBARRIER);
317 case Opt_thread_pool:
319 match_int(&args[0], &intarg);
321 info->thread_pool_size = intarg;
322 printk(KERN_INFO "btrfs: thread pool %d\n",
323 info->thread_pool_size);
327 num = match_strdup(&args[0]);
329 info->max_inline = memparse(num, NULL);
332 if (info->max_inline) {
333 info->max_inline = max_t(u64,
337 printk(KERN_INFO "btrfs: max_inline at %llu\n",
338 (unsigned long long)info->max_inline);
341 case Opt_alloc_start:
342 num = match_strdup(&args[0]);
344 info->alloc_start = memparse(num, NULL);
347 "btrfs: allocations start at %llu\n",
348 (unsigned long long)info->alloc_start);
352 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
355 printk(KERN_INFO "btrfs: disabling tree log\n");
356 btrfs_set_opt(info->mount_opt, NOTREELOG);
358 case Opt_flushoncommit:
359 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
360 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
364 match_int(&args[0], &intarg);
366 info->metadata_ratio = intarg;
367 printk(KERN_INFO "btrfs: metadata ratio %d\n",
368 info->metadata_ratio);
372 btrfs_set_opt(info->mount_opt, DISCARD);
374 case Opt_space_cache:
375 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
377 case Opt_no_space_cache:
378 printk(KERN_INFO "btrfs: disabling disk space caching\n");
379 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
381 case Opt_inode_cache:
382 printk(KERN_INFO "btrfs: enabling inode map caching\n");
383 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
385 case Opt_clear_cache:
386 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
387 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
389 case Opt_user_subvol_rm_allowed:
390 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
392 case Opt_enospc_debug:
393 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
396 printk(KERN_INFO "btrfs: enabling auto defrag");
397 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
400 printk(KERN_INFO "btrfs: enabling auto recovery");
401 btrfs_set_opt(info->mount_opt, RECOVERY);
403 case Opt_skip_balance:
404 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
407 printk(KERN_INFO "btrfs: unrecognized mount option "
416 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
417 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
423 * Parse mount options that are required early in the mount process.
425 * All other options will be parsed on much later in the mount process and
426 * only when we need to allocate a new super block.
428 static int btrfs_parse_early_options(const char *options, fmode_t flags,
429 void *holder, char **subvol_name, u64 *subvol_objectid,
430 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
432 substring_t args[MAX_OPT_ARGS];
433 char *device_name, *opts, *orig, *p;
441 * strsep changes the string, duplicate it because parse_options
444 opts = kstrdup(options, GFP_KERNEL);
449 while ((p = strsep(&opts, ",")) != NULL) {
454 token = match_token(p, tokens, args);
458 *subvol_name = match_strdup(&args[0]);
462 error = match_int(&args[0], &intarg);
464 /* we want the original fs_tree */
467 BTRFS_FS_TREE_OBJECTID;
469 *subvol_objectid = intarg;
472 case Opt_subvolrootid:
474 error = match_int(&args[0], &intarg);
476 /* we want the original fs_tree */
479 BTRFS_FS_TREE_OBJECTID;
481 *subvol_rootid = intarg;
485 device_name = match_strdup(&args[0]);
490 error = btrfs_scan_one_device(device_name,
491 flags, holder, fs_devices);
506 static struct dentry *get_default_root(struct super_block *sb,
509 struct btrfs_root *root = sb->s_fs_info;
510 struct btrfs_root *new_root;
511 struct btrfs_dir_item *di;
512 struct btrfs_path *path;
513 struct btrfs_key location;
519 * We have a specific subvol we want to mount, just setup location and
520 * go look up the root.
522 if (subvol_objectid) {
523 location.objectid = subvol_objectid;
524 location.type = BTRFS_ROOT_ITEM_KEY;
525 location.offset = (u64)-1;
529 path = btrfs_alloc_path();
531 return ERR_PTR(-ENOMEM);
532 path->leave_spinning = 1;
535 * Find the "default" dir item which points to the root item that we
536 * will mount by default if we haven't been given a specific subvolume
539 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
540 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
542 btrfs_free_path(path);
547 * Ok the default dir item isn't there. This is weird since
548 * it's always been there, but don't freak out, just try and
549 * mount to root most subvolume.
551 btrfs_free_path(path);
552 dir_id = BTRFS_FIRST_FREE_OBJECTID;
553 new_root = root->fs_info->fs_root;
557 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
558 btrfs_free_path(path);
561 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
562 if (IS_ERR(new_root))
563 return ERR_CAST(new_root);
565 if (btrfs_root_refs(&new_root->root_item) == 0)
566 return ERR_PTR(-ENOENT);
568 dir_id = btrfs_root_dirid(&new_root->root_item);
570 location.objectid = dir_id;
571 location.type = BTRFS_INODE_ITEM_KEY;
574 inode = btrfs_iget(sb, &location, new_root, &new);
576 return ERR_CAST(inode);
579 * If we're just mounting the root most subvol put the inode and return
580 * a reference to the dentry. We will have already gotten a reference
581 * to the inode in btrfs_fill_super so we're good to go.
583 if (!new && sb->s_root->d_inode == inode) {
585 return dget(sb->s_root);
588 return d_obtain_alias(inode);
591 static int btrfs_fill_super(struct super_block *sb,
592 struct btrfs_fs_devices *fs_devices,
593 void *data, int silent)
596 struct dentry *root_dentry;
597 struct btrfs_root *tree_root;
598 struct btrfs_key key;
601 sb->s_maxbytes = MAX_LFS_FILESIZE;
602 sb->s_magic = BTRFS_SUPER_MAGIC;
603 sb->s_op = &btrfs_super_ops;
604 sb->s_d_op = &btrfs_dentry_operations;
605 sb->s_export_op = &btrfs_export_ops;
606 sb->s_xattr = btrfs_xattr_handlers;
608 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
609 sb->s_flags |= MS_POSIXACL;
612 tree_root = open_ctree(sb, fs_devices, (char *)data);
614 if (IS_ERR(tree_root)) {
615 printk("btrfs: open_ctree failed\n");
616 return PTR_ERR(tree_root);
618 sb->s_fs_info = tree_root;
620 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
621 key.type = BTRFS_INODE_ITEM_KEY;
623 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
625 err = PTR_ERR(inode);
629 root_dentry = d_alloc_root(inode);
636 sb->s_root = root_dentry;
638 save_mount_options(sb, data);
639 cleancache_init_fs(sb);
643 close_ctree(tree_root);
647 int btrfs_sync_fs(struct super_block *sb, int wait)
649 struct btrfs_trans_handle *trans;
650 struct btrfs_root *root = btrfs_sb(sb);
653 trace_btrfs_sync_fs(wait);
656 filemap_flush(root->fs_info->btree_inode->i_mapping);
660 btrfs_start_delalloc_inodes(root, 0);
661 btrfs_wait_ordered_extents(root, 0, 0);
663 trans = btrfs_start_transaction(root, 0);
665 return PTR_ERR(trans);
666 ret = btrfs_commit_transaction(trans, root);
670 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
672 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
673 struct btrfs_fs_info *info = root->fs_info;
676 if (btrfs_test_opt(root, DEGRADED))
677 seq_puts(seq, ",degraded");
678 if (btrfs_test_opt(root, NODATASUM))
679 seq_puts(seq, ",nodatasum");
680 if (btrfs_test_opt(root, NODATACOW))
681 seq_puts(seq, ",nodatacow");
682 if (btrfs_test_opt(root, NOBARRIER))
683 seq_puts(seq, ",nobarrier");
684 if (info->max_inline != 8192 * 1024)
685 seq_printf(seq, ",max_inline=%llu",
686 (unsigned long long)info->max_inline);
687 if (info->alloc_start != 0)
688 seq_printf(seq, ",alloc_start=%llu",
689 (unsigned long long)info->alloc_start);
690 if (info->thread_pool_size != min_t(unsigned long,
691 num_online_cpus() + 2, 8))
692 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
693 if (btrfs_test_opt(root, COMPRESS)) {
694 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
695 compress_type = "zlib";
697 compress_type = "lzo";
698 if (btrfs_test_opt(root, FORCE_COMPRESS))
699 seq_printf(seq, ",compress-force=%s", compress_type);
701 seq_printf(seq, ",compress=%s", compress_type);
703 if (btrfs_test_opt(root, NOSSD))
704 seq_puts(seq, ",nossd");
705 if (btrfs_test_opt(root, SSD_SPREAD))
706 seq_puts(seq, ",ssd_spread");
707 else if (btrfs_test_opt(root, SSD))
708 seq_puts(seq, ",ssd");
709 if (btrfs_test_opt(root, NOTREELOG))
710 seq_puts(seq, ",notreelog");
711 if (btrfs_test_opt(root, FLUSHONCOMMIT))
712 seq_puts(seq, ",flushoncommit");
713 if (btrfs_test_opt(root, DISCARD))
714 seq_puts(seq, ",discard");
715 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
716 seq_puts(seq, ",noacl");
717 if (btrfs_test_opt(root, SPACE_CACHE))
718 seq_puts(seq, ",space_cache");
720 seq_puts(seq, ",nospace_cache");
721 if (btrfs_test_opt(root, CLEAR_CACHE))
722 seq_puts(seq, ",clear_cache");
723 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
724 seq_puts(seq, ",user_subvol_rm_allowed");
725 if (btrfs_test_opt(root, ENOSPC_DEBUG))
726 seq_puts(seq, ",enospc_debug");
727 if (btrfs_test_opt(root, AUTO_DEFRAG))
728 seq_puts(seq, ",autodefrag");
729 if (btrfs_test_opt(root, INODE_MAP_CACHE))
730 seq_puts(seq, ",inode_cache");
731 if (btrfs_test_opt(root, SKIP_BALANCE))
732 seq_puts(seq, ",skip_balance");
736 static int btrfs_test_super(struct super_block *s, void *data)
738 struct btrfs_root *test_root = data;
739 struct btrfs_root *root = btrfs_sb(s);
742 * If this super block is going away, return false as it
743 * can't match as an existing super block.
745 if (!atomic_read(&s->s_active))
747 return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
750 static int btrfs_set_super(struct super_block *s, void *data)
754 return set_anon_super(s, data);
758 * subvolumes are identified by ino 256
760 static inline int is_subvolume_inode(struct inode *inode)
762 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
768 * This will strip out the subvol=%s argument for an argument string and add
769 * subvolid=0 to make sure we get the actual tree root for path walking to the
772 static char *setup_root_args(char *args)
775 unsigned len = strlen(args) + 2;
780 * We need the same args as before, but minus
788 * which is a difference of 2 characters, so we allocate strlen(args) +
791 ret = kzalloc(len * sizeof(char), GFP_NOFS);
794 pos = strstr(args, "subvol=");
796 /* This shouldn't happen, but just in case.. */
803 * The subvol=<> arg is not at the front of the string, copy everybody
804 * up to that into ret.
809 copied += strlen(args);
813 strncpy(ret + copied, "subvolid=0", len - copied);
815 /* Length of subvolid=0 */
819 * If there is no , after the subvol= option then we know there's no
820 * other options and we can just return.
822 pos = strchr(pos, ',');
826 /* Copy the rest of the arguments into our buffer */
827 strncpy(ret + copied, pos, len - copied);
828 copied += strlen(pos);
833 static struct dentry *mount_subvol(const char *subvol_name, int flags,
834 const char *device_name, char *data)
837 struct vfsmount *mnt;
840 newargs = setup_root_args(data);
842 return ERR_PTR(-ENOMEM);
843 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
847 return ERR_CAST(mnt);
849 root = mount_subtree(mnt, subvol_name);
851 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
852 struct super_block *s = root->d_sb;
854 root = ERR_PTR(-EINVAL);
855 deactivate_locked_super(s);
856 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
864 * Find a superblock for the given device / mount point.
866 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
867 * for multiple device setup. Make sure to keep it in sync.
869 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
870 const char *device_name, void *data)
872 struct block_device *bdev = NULL;
873 struct super_block *s;
875 struct btrfs_fs_devices *fs_devices = NULL;
876 struct btrfs_fs_info *fs_info = NULL;
877 fmode_t mode = FMODE_READ;
878 char *subvol_name = NULL;
879 u64 subvol_objectid = 0;
880 u64 subvol_rootid = 0;
883 if (!(flags & MS_RDONLY))
886 error = btrfs_parse_early_options(data, mode, fs_type,
887 &subvol_name, &subvol_objectid,
888 &subvol_rootid, &fs_devices);
891 return ERR_PTR(error);
895 root = mount_subvol(subvol_name, flags, device_name, data);
900 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
902 return ERR_PTR(error);
905 * Setup a dummy root and fs_info for test/set super. This is because
906 * we don't actually fill this stuff out until open_ctree, but we need
907 * it for searching for existing supers, so this lets us do that and
908 * then open_ctree will properly initialize everything later.
910 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
912 return ERR_PTR(-ENOMEM);
914 fs_info->tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
915 if (!fs_info->tree_root) {
919 fs_info->tree_root->fs_info = fs_info;
920 fs_info->fs_devices = fs_devices;
922 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
923 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
924 if (!fs_info->super_copy || !fs_info->super_for_commit) {
929 error = btrfs_open_devices(fs_devices, mode, fs_type);
933 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
935 goto error_close_devices;
938 bdev = fs_devices->latest_bdev;
939 s = sget(fs_type, btrfs_test_super, btrfs_set_super,
943 goto error_close_devices;
947 if ((flags ^ s->s_flags) & MS_RDONLY) {
948 deactivate_locked_super(s);
950 goto error_close_devices;
953 btrfs_close_devices(fs_devices);
954 free_fs_info(fs_info);
956 char b[BDEVNAME_SIZE];
958 s->s_flags = flags | MS_NOSEC;
959 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
960 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
961 error = btrfs_fill_super(s, fs_devices, data,
962 flags & MS_SILENT ? 1 : 0);
964 deactivate_locked_super(s);
965 return ERR_PTR(error);
968 s->s_flags |= MS_ACTIVE;
971 root = get_default_root(s, subvol_objectid);
973 deactivate_locked_super(s);
980 btrfs_close_devices(fs_devices);
982 free_fs_info(fs_info);
983 return ERR_PTR(error);
986 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
988 struct btrfs_root *root = btrfs_sb(sb);
991 ret = btrfs_parse_options(root, data);
995 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
998 if (*flags & MS_RDONLY) {
999 sb->s_flags |= MS_RDONLY;
1001 ret = btrfs_commit_super(root);
1004 if (root->fs_info->fs_devices->rw_devices == 0)
1007 if (btrfs_super_log_root(root->fs_info->super_copy) != 0)
1010 ret = btrfs_cleanup_fs_roots(root->fs_info);
1013 /* recover relocation */
1014 ret = btrfs_recover_relocation(root);
1017 sb->s_flags &= ~MS_RDONLY;
1023 /* Used to sort the devices by max_avail(descending sort) */
1024 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1025 const void *dev_info2)
1027 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1028 ((struct btrfs_device_info *)dev_info2)->max_avail)
1030 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1031 ((struct btrfs_device_info *)dev_info2)->max_avail)
1038 * sort the devices by max_avail, in which max free extent size of each device
1039 * is stored.(Descending Sort)
1041 static inline void btrfs_descending_sort_devices(
1042 struct btrfs_device_info *devices,
1045 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1046 btrfs_cmp_device_free_bytes, NULL);
1050 * The helper to calc the free space on the devices that can be used to store
1053 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1055 struct btrfs_fs_info *fs_info = root->fs_info;
1056 struct btrfs_device_info *devices_info;
1057 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1058 struct btrfs_device *device;
1063 u64 min_stripe_size;
1064 int min_stripes = 1, num_stripes = 1;
1065 int i = 0, nr_devices;
1068 nr_devices = fs_info->fs_devices->open_devices;
1069 BUG_ON(!nr_devices);
1071 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1076 /* calc min stripe number for data space alloction */
1077 type = btrfs_get_alloc_profile(root, 1);
1078 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1080 num_stripes = nr_devices;
1081 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1084 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1089 if (type & BTRFS_BLOCK_GROUP_DUP)
1090 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1092 min_stripe_size = BTRFS_STRIPE_LEN;
1094 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1095 if (!device->in_fs_metadata || !device->bdev)
1098 avail_space = device->total_bytes - device->bytes_used;
1100 /* align with stripe_len */
1101 do_div(avail_space, BTRFS_STRIPE_LEN);
1102 avail_space *= BTRFS_STRIPE_LEN;
1105 * In order to avoid overwritting the superblock on the drive,
1106 * btrfs starts at an offset of at least 1MB when doing chunk
1109 skip_space = 1024 * 1024;
1111 /* user can set the offset in fs_info->alloc_start. */
1112 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1113 device->total_bytes)
1114 skip_space = max(fs_info->alloc_start, skip_space);
1117 * btrfs can not use the free space in [0, skip_space - 1],
1118 * we must subtract it from the total. In order to implement
1119 * it, we account the used space in this range first.
1121 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1124 kfree(devices_info);
1128 /* calc the free space in [0, skip_space - 1] */
1129 skip_space -= used_space;
1132 * we can use the free space in [0, skip_space - 1], subtract
1133 * it from the total.
1135 if (avail_space && avail_space >= skip_space)
1136 avail_space -= skip_space;
1140 if (avail_space < min_stripe_size)
1143 devices_info[i].dev = device;
1144 devices_info[i].max_avail = avail_space;
1151 btrfs_descending_sort_devices(devices_info, nr_devices);
1155 while (nr_devices >= min_stripes) {
1156 if (num_stripes > nr_devices)
1157 num_stripes = nr_devices;
1159 if (devices_info[i].max_avail >= min_stripe_size) {
1163 avail_space += devices_info[i].max_avail * num_stripes;
1164 alloc_size = devices_info[i].max_avail;
1165 for (j = i + 1 - num_stripes; j <= i; j++)
1166 devices_info[j].max_avail -= alloc_size;
1172 kfree(devices_info);
1173 *free_bytes = avail_space;
1177 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1179 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
1180 struct btrfs_super_block *disk_super = root->fs_info->super_copy;
1181 struct list_head *head = &root->fs_info->space_info;
1182 struct btrfs_space_info *found;
1184 u64 total_free_data = 0;
1185 int bits = dentry->d_sb->s_blocksize_bits;
1186 __be32 *fsid = (__be32 *)root->fs_info->fsid;
1189 /* holding chunk_muext to avoid allocating new chunks */
1190 mutex_lock(&root->fs_info->chunk_mutex);
1192 list_for_each_entry_rcu(found, head, list) {
1193 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1194 total_free_data += found->disk_total - found->disk_used;
1196 btrfs_account_ro_block_groups_free_space(found);
1199 total_used += found->disk_used;
1203 buf->f_namelen = BTRFS_NAME_LEN;
1204 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1205 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1206 buf->f_bsize = dentry->d_sb->s_blocksize;
1207 buf->f_type = BTRFS_SUPER_MAGIC;
1208 buf->f_bavail = total_free_data;
1209 ret = btrfs_calc_avail_data_space(root, &total_free_data);
1211 mutex_unlock(&root->fs_info->chunk_mutex);
1214 buf->f_bavail += total_free_data;
1215 buf->f_bavail = buf->f_bavail >> bits;
1216 mutex_unlock(&root->fs_info->chunk_mutex);
1218 /* We treat it as constant endianness (it doesn't matter _which_)
1219 because we want the fsid to come out the same whether mounted
1220 on a big-endian or little-endian host */
1221 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1222 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1223 /* Mask in the root object ID too, to disambiguate subvols */
1224 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1225 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1230 static struct file_system_type btrfs_fs_type = {
1231 .owner = THIS_MODULE,
1233 .mount = btrfs_mount,
1234 .kill_sb = kill_anon_super,
1235 .fs_flags = FS_REQUIRES_DEV,
1239 * used by btrfsctl to scan devices when no FS is mounted
1241 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1244 struct btrfs_ioctl_vol_args *vol;
1245 struct btrfs_fs_devices *fs_devices;
1248 if (!capable(CAP_SYS_ADMIN))
1251 vol = memdup_user((void __user *)arg, sizeof(*vol));
1253 return PTR_ERR(vol);
1256 case BTRFS_IOC_SCAN_DEV:
1257 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1258 &btrfs_fs_type, &fs_devices);
1266 static int btrfs_freeze(struct super_block *sb)
1268 struct btrfs_root *root = btrfs_sb(sb);
1269 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1270 mutex_lock(&root->fs_info->cleaner_mutex);
1274 static int btrfs_unfreeze(struct super_block *sb)
1276 struct btrfs_root *root = btrfs_sb(sb);
1277 mutex_unlock(&root->fs_info->cleaner_mutex);
1278 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1282 static void btrfs_fs_dirty_inode(struct inode *inode, int flags)
1286 ret = btrfs_dirty_inode(inode);
1288 printk_ratelimited(KERN_ERR "btrfs: fail to dirty inode %Lu "
1289 "error %d\n", btrfs_ino(inode), ret);
1292 static const struct super_operations btrfs_super_ops = {
1293 .drop_inode = btrfs_drop_inode,
1294 .evict_inode = btrfs_evict_inode,
1295 .put_super = btrfs_put_super,
1296 .sync_fs = btrfs_sync_fs,
1297 .show_options = btrfs_show_options,
1298 .write_inode = btrfs_write_inode,
1299 .dirty_inode = btrfs_fs_dirty_inode,
1300 .alloc_inode = btrfs_alloc_inode,
1301 .destroy_inode = btrfs_destroy_inode,
1302 .statfs = btrfs_statfs,
1303 .remount_fs = btrfs_remount,
1304 .freeze_fs = btrfs_freeze,
1305 .unfreeze_fs = btrfs_unfreeze,
1308 static const struct file_operations btrfs_ctl_fops = {
1309 .unlocked_ioctl = btrfs_control_ioctl,
1310 .compat_ioctl = btrfs_control_ioctl,
1311 .owner = THIS_MODULE,
1312 .llseek = noop_llseek,
1315 static struct miscdevice btrfs_misc = {
1316 .minor = BTRFS_MINOR,
1317 .name = "btrfs-control",
1318 .fops = &btrfs_ctl_fops
1321 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1322 MODULE_ALIAS("devname:btrfs-control");
1324 static int btrfs_interface_init(void)
1326 return misc_register(&btrfs_misc);
1329 static void btrfs_interface_exit(void)
1331 if (misc_deregister(&btrfs_misc) < 0)
1332 printk(KERN_INFO "misc_deregister failed for control device");
1335 static int __init init_btrfs_fs(void)
1339 err = btrfs_init_sysfs();
1343 err = btrfs_init_compress();
1347 err = btrfs_init_cachep();
1351 err = extent_io_init();
1355 err = extent_map_init();
1357 goto free_extent_io;
1359 err = btrfs_delayed_inode_init();
1361 goto free_extent_map;
1363 err = btrfs_interface_init();
1365 goto free_delayed_inode;
1367 err = register_filesystem(&btrfs_fs_type);
1369 goto unregister_ioctl;
1371 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1375 btrfs_interface_exit();
1377 btrfs_delayed_inode_exit();
1383 btrfs_destroy_cachep();
1385 btrfs_exit_compress();
1391 static void __exit exit_btrfs_fs(void)
1393 btrfs_destroy_cachep();
1394 btrfs_delayed_inode_exit();
1397 btrfs_interface_exit();
1398 unregister_filesystem(&btrfs_fs_type);
1400 btrfs_cleanup_fs_uuids();
1401 btrfs_exit_compress();
1404 module_init(init_btrfs_fs)
1405 module_exit(exit_btrfs_fs)
1407 MODULE_LICENSE("GPL");