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>
45 #include "transaction.h"
46 #include "btrfs_inode.h"
48 #include "print-tree.h"
53 #include "compression.h"
55 static const struct super_operations btrfs_super_ops;
57 static void btrfs_put_super(struct super_block *sb)
59 struct btrfs_root *root = btrfs_sb(sb);
62 ret = close_ctree(root);
65 (void)ret; /* FIXME: need to fix VFS to return error? */
69 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
70 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
71 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
72 Opt_compress_force, Opt_notreelog, Opt_ratio, Opt_flushoncommit,
73 Opt_discard, Opt_space_cache, Opt_clear_cache, Opt_err,
74 Opt_user_subvol_rm_allowed,
77 static match_table_t tokens = {
78 {Opt_degraded, "degraded"},
79 {Opt_subvol, "subvol=%s"},
80 {Opt_subvolid, "subvolid=%d"},
81 {Opt_device, "device=%s"},
82 {Opt_nodatasum, "nodatasum"},
83 {Opt_nodatacow, "nodatacow"},
84 {Opt_nobarrier, "nobarrier"},
85 {Opt_max_inline, "max_inline=%s"},
86 {Opt_alloc_start, "alloc_start=%s"},
87 {Opt_thread_pool, "thread_pool=%d"},
88 {Opt_compress, "compress"},
89 {Opt_compress_force, "compress-force"},
91 {Opt_ssd_spread, "ssd_spread"},
94 {Opt_notreelog, "notreelog"},
95 {Opt_flushoncommit, "flushoncommit"},
96 {Opt_ratio, "metadata_ratio=%d"},
97 {Opt_discard, "discard"},
98 {Opt_space_cache, "space_cache"},
99 {Opt_clear_cache, "clear_cache"},
100 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
105 * Regular mount options parser. Everything that is needed only when
106 * reading in a new superblock is parsed here.
108 int btrfs_parse_options(struct btrfs_root *root, char *options)
110 struct btrfs_fs_info *info = root->fs_info;
111 substring_t args[MAX_OPT_ARGS];
112 char *p, *num, *orig;
120 * strsep changes the string, duplicate it because parse_options
123 options = kstrdup(options, GFP_NOFS);
129 while ((p = strsep(&options, ",")) != NULL) {
134 token = match_token(p, tokens, args);
137 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
138 btrfs_set_opt(info->mount_opt, DEGRADED);
144 * These are parsed by btrfs_parse_early_options
145 * and can be happily ignored here.
149 printk(KERN_INFO "btrfs: setting nodatasum\n");
150 btrfs_set_opt(info->mount_opt, NODATASUM);
153 printk(KERN_INFO "btrfs: setting nodatacow\n");
154 btrfs_set_opt(info->mount_opt, NODATACOW);
155 btrfs_set_opt(info->mount_opt, NODATASUM);
158 printk(KERN_INFO "btrfs: use compression\n");
159 btrfs_set_opt(info->mount_opt, COMPRESS);
161 case Opt_compress_force:
162 printk(KERN_INFO "btrfs: forcing compression\n");
163 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
164 btrfs_set_opt(info->mount_opt, COMPRESS);
167 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
168 btrfs_set_opt(info->mount_opt, SSD);
171 printk(KERN_INFO "btrfs: use spread ssd "
172 "allocation scheme\n");
173 btrfs_set_opt(info->mount_opt, SSD);
174 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
177 printk(KERN_INFO "btrfs: not using ssd allocation "
179 btrfs_set_opt(info->mount_opt, NOSSD);
180 btrfs_clear_opt(info->mount_opt, SSD);
181 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
184 printk(KERN_INFO "btrfs: turning off barriers\n");
185 btrfs_set_opt(info->mount_opt, NOBARRIER);
187 case Opt_thread_pool:
189 match_int(&args[0], &intarg);
191 info->thread_pool_size = intarg;
192 printk(KERN_INFO "btrfs: thread pool %d\n",
193 info->thread_pool_size);
197 num = match_strdup(&args[0]);
199 info->max_inline = memparse(num, NULL);
202 if (info->max_inline) {
203 info->max_inline = max_t(u64,
207 printk(KERN_INFO "btrfs: max_inline at %llu\n",
208 (unsigned long long)info->max_inline);
211 case Opt_alloc_start:
212 num = match_strdup(&args[0]);
214 info->alloc_start = memparse(num, NULL);
217 "btrfs: allocations start at %llu\n",
218 (unsigned long long)info->alloc_start);
222 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
225 printk(KERN_INFO "btrfs: disabling tree log\n");
226 btrfs_set_opt(info->mount_opt, NOTREELOG);
228 case Opt_flushoncommit:
229 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
230 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
234 match_int(&args[0], &intarg);
236 info->metadata_ratio = intarg;
237 printk(KERN_INFO "btrfs: metadata ratio %d\n",
238 info->metadata_ratio);
242 btrfs_set_opt(info->mount_opt, DISCARD);
244 case Opt_space_cache:
245 printk(KERN_INFO "btrfs: enabling disk space caching\n");
246 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
247 case Opt_clear_cache:
248 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
249 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
251 case Opt_user_subvol_rm_allowed:
252 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
255 printk(KERN_INFO "btrfs: unrecognized mount option "
269 * Parse mount options that are required early in the mount process.
271 * All other options will be parsed on much later in the mount process and
272 * only when we need to allocate a new super block.
274 static int btrfs_parse_early_options(const char *options, fmode_t flags,
275 void *holder, char **subvol_name, u64 *subvol_objectid,
276 struct btrfs_fs_devices **fs_devices)
278 substring_t args[MAX_OPT_ARGS];
287 * strsep changes the string, duplicate it because parse_options
290 opts = kstrdup(options, GFP_KERNEL);
294 while ((p = strsep(&opts, ",")) != NULL) {
299 token = match_token(p, tokens, args);
302 *subvol_name = match_strdup(&args[0]);
306 error = match_int(&args[0], &intarg);
308 /* we want the original fs_tree */
311 BTRFS_FS_TREE_OBJECTID;
313 *subvol_objectid = intarg;
317 error = btrfs_scan_one_device(match_strdup(&args[0]),
318 flags, holder, fs_devices);
331 * If no subvolume name is specified we use the default one. Allocate
332 * a copy of the string "." here so that code later in the
333 * mount path doesn't care if it's the default volume or another one.
336 *subvol_name = kstrdup(".", GFP_KERNEL);
343 static struct dentry *get_default_root(struct super_block *sb,
346 struct btrfs_root *root = sb->s_fs_info;
347 struct btrfs_root *new_root;
348 struct btrfs_dir_item *di;
349 struct btrfs_path *path;
350 struct btrfs_key location;
352 struct dentry *dentry;
357 * We have a specific subvol we want to mount, just setup location and
358 * go look up the root.
360 if (subvol_objectid) {
361 location.objectid = subvol_objectid;
362 location.type = BTRFS_ROOT_ITEM_KEY;
363 location.offset = (u64)-1;
367 path = btrfs_alloc_path();
369 return ERR_PTR(-ENOMEM);
370 path->leave_spinning = 1;
373 * Find the "default" dir item which points to the root item that we
374 * will mount by default if we haven't been given a specific subvolume
377 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
378 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
383 * Ok the default dir item isn't there. This is weird since
384 * it's always been there, but don't freak out, just try and
385 * mount to root most subvolume.
387 btrfs_free_path(path);
388 dir_id = BTRFS_FIRST_FREE_OBJECTID;
389 new_root = root->fs_info->fs_root;
393 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
394 btrfs_free_path(path);
397 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
398 if (IS_ERR(new_root))
399 return ERR_CAST(new_root);
401 if (btrfs_root_refs(&new_root->root_item) == 0)
402 return ERR_PTR(-ENOENT);
404 dir_id = btrfs_root_dirid(&new_root->root_item);
406 location.objectid = dir_id;
407 location.type = BTRFS_INODE_ITEM_KEY;
410 inode = btrfs_iget(sb, &location, new_root, &new);
412 return ERR_CAST(inode);
415 * If we're just mounting the root most subvol put the inode and return
416 * a reference to the dentry. We will have already gotten a reference
417 * to the inode in btrfs_fill_super so we're good to go.
419 if (!new && sb->s_root->d_inode == inode) {
421 return dget(sb->s_root);
425 const struct qstr name = { .name = "/", .len = 1 };
428 * New inode, we need to make the dentry a sibling of s_root so
429 * everything gets cleaned up properly on unmount.
431 dentry = d_alloc(sb->s_root, &name);
434 return ERR_PTR(-ENOMEM);
436 d_splice_alias(inode, dentry);
439 * We found the inode in cache, just find a dentry for it and
440 * put the reference to the inode we just got.
442 dentry = d_find_alias(inode);
449 static int btrfs_fill_super(struct super_block *sb,
450 struct btrfs_fs_devices *fs_devices,
451 void *data, int silent)
454 struct dentry *root_dentry;
455 struct btrfs_root *tree_root;
456 struct btrfs_key key;
459 sb->s_maxbytes = MAX_LFS_FILESIZE;
460 sb->s_magic = BTRFS_SUPER_MAGIC;
461 sb->s_op = &btrfs_super_ops;
462 sb->s_export_op = &btrfs_export_ops;
463 sb->s_xattr = btrfs_xattr_handlers;
465 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
466 sb->s_flags |= MS_POSIXACL;
469 tree_root = open_ctree(sb, fs_devices, (char *)data);
471 if (IS_ERR(tree_root)) {
472 printk("btrfs: open_ctree failed\n");
473 return PTR_ERR(tree_root);
475 sb->s_fs_info = tree_root;
477 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
478 key.type = BTRFS_INODE_ITEM_KEY;
480 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
482 err = PTR_ERR(inode);
486 root_dentry = d_alloc_root(inode);
493 sb->s_root = root_dentry;
495 save_mount_options(sb, data);
499 close_ctree(tree_root);
503 int btrfs_sync_fs(struct super_block *sb, int wait)
505 struct btrfs_trans_handle *trans;
506 struct btrfs_root *root = btrfs_sb(sb);
510 filemap_flush(root->fs_info->btree_inode->i_mapping);
514 btrfs_start_delalloc_inodes(root, 0);
515 btrfs_wait_ordered_extents(root, 0, 0);
517 trans = btrfs_start_transaction(root, 0);
518 ret = btrfs_commit_transaction(trans, root);
522 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
524 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
525 struct btrfs_fs_info *info = root->fs_info;
527 if (btrfs_test_opt(root, DEGRADED))
528 seq_puts(seq, ",degraded");
529 if (btrfs_test_opt(root, NODATASUM))
530 seq_puts(seq, ",nodatasum");
531 if (btrfs_test_opt(root, NODATACOW))
532 seq_puts(seq, ",nodatacow");
533 if (btrfs_test_opt(root, NOBARRIER))
534 seq_puts(seq, ",nobarrier");
535 if (info->max_inline != 8192 * 1024)
536 seq_printf(seq, ",max_inline=%llu",
537 (unsigned long long)info->max_inline);
538 if (info->alloc_start != 0)
539 seq_printf(seq, ",alloc_start=%llu",
540 (unsigned long long)info->alloc_start);
541 if (info->thread_pool_size != min_t(unsigned long,
542 num_online_cpus() + 2, 8))
543 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
544 if (btrfs_test_opt(root, COMPRESS))
545 seq_puts(seq, ",compress");
546 if (btrfs_test_opt(root, NOSSD))
547 seq_puts(seq, ",nossd");
548 if (btrfs_test_opt(root, SSD_SPREAD))
549 seq_puts(seq, ",ssd_spread");
550 else if (btrfs_test_opt(root, SSD))
551 seq_puts(seq, ",ssd");
552 if (btrfs_test_opt(root, NOTREELOG))
553 seq_puts(seq, ",notreelog");
554 if (btrfs_test_opt(root, FLUSHONCOMMIT))
555 seq_puts(seq, ",flushoncommit");
556 if (btrfs_test_opt(root, DISCARD))
557 seq_puts(seq, ",discard");
558 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
559 seq_puts(seq, ",noacl");
563 static int btrfs_test_super(struct super_block *s, void *data)
565 struct btrfs_fs_devices *test_fs_devices = data;
566 struct btrfs_root *root = btrfs_sb(s);
568 return root->fs_info->fs_devices == test_fs_devices;
572 * Find a superblock for the given device / mount point.
574 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
575 * for multiple device setup. Make sure to keep it in sync.
577 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
578 const char *dev_name, void *data, struct vfsmount *mnt)
580 struct block_device *bdev = NULL;
581 struct super_block *s;
583 struct btrfs_fs_devices *fs_devices = NULL;
584 fmode_t mode = FMODE_READ;
585 char *subvol_name = NULL;
586 u64 subvol_objectid = 0;
589 if (!(flags & MS_RDONLY))
592 error = btrfs_parse_early_options(data, mode, fs_type,
593 &subvol_name, &subvol_objectid,
598 error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
600 goto error_free_subvol_name;
602 error = btrfs_open_devices(fs_devices, mode, fs_type);
604 goto error_free_subvol_name;
606 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
608 goto error_close_devices;
611 bdev = fs_devices->latest_bdev;
612 s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
617 if ((flags ^ s->s_flags) & MS_RDONLY) {
618 deactivate_locked_super(s);
620 goto error_close_devices;
623 btrfs_close_devices(fs_devices);
625 char b[BDEVNAME_SIZE];
628 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
629 error = btrfs_fill_super(s, fs_devices, data,
630 flags & MS_SILENT ? 1 : 0);
632 deactivate_locked_super(s);
633 goto error_free_subvol_name;
636 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
637 s->s_flags |= MS_ACTIVE;
640 root = get_default_root(s, subvol_objectid);
642 error = PTR_ERR(root);
643 deactivate_locked_super(s);
644 goto error_free_subvol_name;
646 /* if they gave us a subvolume name bind mount into that */
647 if (strcmp(subvol_name, ".")) {
648 struct dentry *new_root;
649 mutex_lock(&root->d_inode->i_mutex);
650 new_root = lookup_one_len(subvol_name, root,
651 strlen(subvol_name));
652 mutex_unlock(&root->d_inode->i_mutex);
654 if (IS_ERR(new_root)) {
655 deactivate_locked_super(s);
656 error = PTR_ERR(new_root);
658 goto error_free_subvol_name;
660 if (!new_root->d_inode) {
663 deactivate_locked_super(s);
665 goto error_free_subvol_name;
672 mnt->mnt_root = root;
680 btrfs_close_devices(fs_devices);
681 error_free_subvol_name:
686 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
688 struct btrfs_root *root = btrfs_sb(sb);
691 ret = btrfs_parse_options(root, data);
695 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
698 if (*flags & MS_RDONLY) {
699 sb->s_flags |= MS_RDONLY;
701 ret = btrfs_commit_super(root);
704 if (root->fs_info->fs_devices->rw_devices == 0)
707 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
710 ret = btrfs_cleanup_fs_roots(root->fs_info);
713 /* recover relocation */
714 ret = btrfs_recover_relocation(root);
717 sb->s_flags &= ~MS_RDONLY;
723 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
725 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
726 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
727 struct list_head *head = &root->fs_info->space_info;
728 struct btrfs_space_info *found;
730 u64 total_used_data = 0;
731 int bits = dentry->d_sb->s_blocksize_bits;
732 __be32 *fsid = (__be32 *)root->fs_info->fsid;
735 list_for_each_entry_rcu(found, head, list) {
736 if (found->flags & (BTRFS_BLOCK_GROUP_METADATA |
737 BTRFS_BLOCK_GROUP_SYSTEM))
738 total_used_data += found->disk_total;
740 total_used_data += found->disk_used;
741 total_used += found->disk_used;
745 buf->f_namelen = BTRFS_NAME_LEN;
746 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
747 buf->f_bfree = buf->f_blocks - (total_used >> bits);
748 buf->f_bavail = buf->f_blocks - (total_used_data >> bits);
749 buf->f_bsize = dentry->d_sb->s_blocksize;
750 buf->f_type = BTRFS_SUPER_MAGIC;
752 /* We treat it as constant endianness (it doesn't matter _which_)
753 because we want the fsid to come out the same whether mounted
754 on a big-endian or little-endian host */
755 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
756 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
757 /* Mask in the root object ID too, to disambiguate subvols */
758 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
759 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
764 static struct file_system_type btrfs_fs_type = {
765 .owner = THIS_MODULE,
767 .get_sb = btrfs_get_sb,
768 .kill_sb = kill_anon_super,
769 .fs_flags = FS_REQUIRES_DEV,
773 * used by btrfsctl to scan devices when no FS is mounted
775 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
778 struct btrfs_ioctl_vol_args *vol;
779 struct btrfs_fs_devices *fs_devices;
782 if (!capable(CAP_SYS_ADMIN))
785 vol = memdup_user((void __user *)arg, sizeof(*vol));
790 case BTRFS_IOC_SCAN_DEV:
791 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
792 &btrfs_fs_type, &fs_devices);
800 static int btrfs_freeze(struct super_block *sb)
802 struct btrfs_root *root = btrfs_sb(sb);
803 mutex_lock(&root->fs_info->transaction_kthread_mutex);
804 mutex_lock(&root->fs_info->cleaner_mutex);
808 static int btrfs_unfreeze(struct super_block *sb)
810 struct btrfs_root *root = btrfs_sb(sb);
811 mutex_unlock(&root->fs_info->cleaner_mutex);
812 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
816 static const struct super_operations btrfs_super_ops = {
817 .drop_inode = btrfs_drop_inode,
818 .evict_inode = btrfs_evict_inode,
819 .put_super = btrfs_put_super,
820 .sync_fs = btrfs_sync_fs,
821 .show_options = btrfs_show_options,
822 .write_inode = btrfs_write_inode,
823 .dirty_inode = btrfs_dirty_inode,
824 .alloc_inode = btrfs_alloc_inode,
825 .destroy_inode = btrfs_destroy_inode,
826 .statfs = btrfs_statfs,
827 .remount_fs = btrfs_remount,
828 .freeze_fs = btrfs_freeze,
829 .unfreeze_fs = btrfs_unfreeze,
832 static const struct file_operations btrfs_ctl_fops = {
833 .unlocked_ioctl = btrfs_control_ioctl,
834 .compat_ioctl = btrfs_control_ioctl,
835 .owner = THIS_MODULE,
838 static struct miscdevice btrfs_misc = {
839 .minor = BTRFS_MINOR,
840 .name = "btrfs-control",
841 .fops = &btrfs_ctl_fops
844 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
845 MODULE_ALIAS("devname:btrfs-control");
847 static int btrfs_interface_init(void)
849 return misc_register(&btrfs_misc);
852 static void btrfs_interface_exit(void)
854 if (misc_deregister(&btrfs_misc) < 0)
855 printk(KERN_INFO "misc_deregister failed for control device");
858 static int __init init_btrfs_fs(void)
862 err = btrfs_init_sysfs();
866 err = btrfs_init_cachep();
870 err = extent_io_init();
874 err = extent_map_init();
878 err = btrfs_interface_init();
880 goto free_extent_map;
882 err = register_filesystem(&btrfs_fs_type);
884 goto unregister_ioctl;
886 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
890 btrfs_interface_exit();
896 btrfs_destroy_cachep();
902 static void __exit exit_btrfs_fs(void)
904 btrfs_destroy_cachep();
907 btrfs_interface_exit();
908 unregister_filesystem(&btrfs_fs_type);
910 btrfs_cleanup_fs_uuids();
914 module_init(init_btrfs_fs)
915 module_exit(exit_btrfs_fs)
917 MODULE_LICENSE("GPL");