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 const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
64 errstr = "IO failure";
67 errstr = "Out of memory";
70 errstr = "Readonly filesystem";
74 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
83 static void __save_error_info(struct btrfs_fs_info *fs_info)
86 * today we only save the error info into ram. Long term we'll
87 * also send it down to the disk
89 fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
93 * We move write_super stuff at umount in order to avoid deadlock
94 * for umount hold all lock.
96 static void save_error_info(struct btrfs_fs_info *fs_info)
98 __save_error_info(fs_info);
101 /* btrfs handle error by forcing the filesystem readonly */
102 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
104 struct super_block *sb = fs_info->sb;
106 if (sb->s_flags & MS_RDONLY)
109 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
110 sb->s_flags |= MS_RDONLY;
111 printk(KERN_INFO "btrfs is forced readonly\n");
116 * __btrfs_std_error decodes expected errors from the caller and
117 * invokes the approciate error response.
119 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
120 unsigned int line, int errno)
122 struct super_block *sb = fs_info->sb;
127 * Special case: if the error is EROFS, and we're already
128 * under MS_RDONLY, then it is safe here.
130 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
133 errstr = btrfs_decode_error(fs_info, errno, nbuf);
134 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
135 sb->s_id, function, line, errstr);
136 save_error_info(fs_info);
138 btrfs_handle_error(fs_info);
141 static void btrfs_put_super(struct super_block *sb)
143 struct btrfs_root *root = btrfs_sb(sb);
146 ret = close_ctree(root);
147 sb->s_fs_info = NULL;
149 (void)ret; /* FIXME: need to fix VFS to return error? */
153 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
154 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
155 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
156 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
157 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
158 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed, Opt_err,
161 static match_table_t tokens = {
162 {Opt_degraded, "degraded"},
163 {Opt_subvol, "subvol=%s"},
164 {Opt_subvolid, "subvolid=%d"},
165 {Opt_device, "device=%s"},
166 {Opt_nodatasum, "nodatasum"},
167 {Opt_nodatacow, "nodatacow"},
168 {Opt_nobarrier, "nobarrier"},
169 {Opt_max_inline, "max_inline=%s"},
170 {Opt_alloc_start, "alloc_start=%s"},
171 {Opt_thread_pool, "thread_pool=%d"},
172 {Opt_compress, "compress"},
173 {Opt_compress_type, "compress=%s"},
174 {Opt_compress_force, "compress-force"},
175 {Opt_compress_force_type, "compress-force=%s"},
177 {Opt_ssd_spread, "ssd_spread"},
178 {Opt_nossd, "nossd"},
179 {Opt_noacl, "noacl"},
180 {Opt_notreelog, "notreelog"},
181 {Opt_flushoncommit, "flushoncommit"},
182 {Opt_ratio, "metadata_ratio=%d"},
183 {Opt_discard, "discard"},
184 {Opt_space_cache, "space_cache"},
185 {Opt_clear_cache, "clear_cache"},
186 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
191 * Regular mount options parser. Everything that is needed only when
192 * reading in a new superblock is parsed here.
194 int btrfs_parse_options(struct btrfs_root *root, char *options)
196 struct btrfs_fs_info *info = root->fs_info;
197 substring_t args[MAX_OPT_ARGS];
198 char *p, *num, *orig;
202 bool compress_force = false;
208 * strsep changes the string, duplicate it because parse_options
211 options = kstrdup(options, GFP_NOFS);
217 while ((p = strsep(&options, ",")) != NULL) {
222 token = match_token(p, tokens, args);
225 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
226 btrfs_set_opt(info->mount_opt, DEGRADED);
232 * These are parsed by btrfs_parse_early_options
233 * and can be happily ignored here.
237 printk(KERN_INFO "btrfs: setting nodatasum\n");
238 btrfs_set_opt(info->mount_opt, NODATASUM);
241 printk(KERN_INFO "btrfs: setting nodatacow\n");
242 btrfs_set_opt(info->mount_opt, NODATACOW);
243 btrfs_set_opt(info->mount_opt, NODATASUM);
245 case Opt_compress_force:
246 case Opt_compress_force_type:
247 compress_force = true;
249 case Opt_compress_type:
250 if (token == Opt_compress ||
251 token == Opt_compress_force ||
252 strcmp(args[0].from, "zlib") == 0) {
253 compress_type = "zlib";
254 info->compress_type = BTRFS_COMPRESS_ZLIB;
255 } else if (strcmp(args[0].from, "lzo") == 0) {
256 compress_type = "lzo";
257 info->compress_type = BTRFS_COMPRESS_LZO;
263 btrfs_set_opt(info->mount_opt, COMPRESS);
264 if (compress_force) {
265 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
266 pr_info("btrfs: force %s compression\n",
269 pr_info("btrfs: use %s compression\n",
273 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
274 btrfs_set_opt(info->mount_opt, SSD);
277 printk(KERN_INFO "btrfs: use spread ssd "
278 "allocation scheme\n");
279 btrfs_set_opt(info->mount_opt, SSD);
280 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
283 printk(KERN_INFO "btrfs: not using ssd allocation "
285 btrfs_set_opt(info->mount_opt, NOSSD);
286 btrfs_clear_opt(info->mount_opt, SSD);
287 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
290 printk(KERN_INFO "btrfs: turning off barriers\n");
291 btrfs_set_opt(info->mount_opt, NOBARRIER);
293 case Opt_thread_pool:
295 match_int(&args[0], &intarg);
297 info->thread_pool_size = intarg;
298 printk(KERN_INFO "btrfs: thread pool %d\n",
299 info->thread_pool_size);
303 num = match_strdup(&args[0]);
305 info->max_inline = memparse(num, NULL);
308 if (info->max_inline) {
309 info->max_inline = max_t(u64,
313 printk(KERN_INFO "btrfs: max_inline at %llu\n",
314 (unsigned long long)info->max_inline);
317 case Opt_alloc_start:
318 num = match_strdup(&args[0]);
320 info->alloc_start = memparse(num, NULL);
323 "btrfs: allocations start at %llu\n",
324 (unsigned long long)info->alloc_start);
328 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
331 printk(KERN_INFO "btrfs: disabling tree log\n");
332 btrfs_set_opt(info->mount_opt, NOTREELOG);
334 case Opt_flushoncommit:
335 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
336 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
340 match_int(&args[0], &intarg);
342 info->metadata_ratio = intarg;
343 printk(KERN_INFO "btrfs: metadata ratio %d\n",
344 info->metadata_ratio);
348 btrfs_set_opt(info->mount_opt, DISCARD);
350 case Opt_space_cache:
351 printk(KERN_INFO "btrfs: enabling disk space caching\n");
352 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
354 case Opt_clear_cache:
355 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
356 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
358 case Opt_user_subvol_rm_allowed:
359 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
362 printk(KERN_INFO "btrfs: unrecognized mount option "
376 * Parse mount options that are required early in the mount process.
378 * All other options will be parsed on much later in the mount process and
379 * only when we need to allocate a new super block.
381 static int btrfs_parse_early_options(const char *options, fmode_t flags,
382 void *holder, char **subvol_name, u64 *subvol_objectid,
383 struct btrfs_fs_devices **fs_devices)
385 substring_t args[MAX_OPT_ARGS];
386 char *opts, *orig, *p;
394 * strsep changes the string, duplicate it because parse_options
397 opts = kstrdup(options, GFP_KERNEL);
402 while ((p = strsep(&opts, ",")) != NULL) {
407 token = match_token(p, tokens, args);
410 *subvol_name = match_strdup(&args[0]);
414 error = match_int(&args[0], &intarg);
416 /* we want the original fs_tree */
419 BTRFS_FS_TREE_OBJECTID;
421 *subvol_objectid = intarg;
425 error = btrfs_scan_one_device(match_strdup(&args[0]),
426 flags, holder, fs_devices);
439 * If no subvolume name is specified we use the default one. Allocate
440 * a copy of the string "." here so that code later in the
441 * mount path doesn't care if it's the default volume or another one.
444 *subvol_name = kstrdup(".", GFP_KERNEL);
451 static struct dentry *get_default_root(struct super_block *sb,
454 struct btrfs_root *root = sb->s_fs_info;
455 struct btrfs_root *new_root;
456 struct btrfs_dir_item *di;
457 struct btrfs_path *path;
458 struct btrfs_key location;
460 struct dentry *dentry;
465 * We have a specific subvol we want to mount, just setup location and
466 * go look up the root.
468 if (subvol_objectid) {
469 location.objectid = subvol_objectid;
470 location.type = BTRFS_ROOT_ITEM_KEY;
471 location.offset = (u64)-1;
475 path = btrfs_alloc_path();
477 return ERR_PTR(-ENOMEM);
478 path->leave_spinning = 1;
481 * Find the "default" dir item which points to the root item that we
482 * will mount by default if we haven't been given a specific subvolume
485 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
486 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
491 * Ok the default dir item isn't there. This is weird since
492 * it's always been there, but don't freak out, just try and
493 * mount to root most subvolume.
495 btrfs_free_path(path);
496 dir_id = BTRFS_FIRST_FREE_OBJECTID;
497 new_root = root->fs_info->fs_root;
501 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
502 btrfs_free_path(path);
505 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
506 if (IS_ERR(new_root))
507 return ERR_CAST(new_root);
509 if (btrfs_root_refs(&new_root->root_item) == 0)
510 return ERR_PTR(-ENOENT);
512 dir_id = btrfs_root_dirid(&new_root->root_item);
514 location.objectid = dir_id;
515 location.type = BTRFS_INODE_ITEM_KEY;
518 inode = btrfs_iget(sb, &location, new_root, &new);
520 return ERR_CAST(inode);
523 * If we're just mounting the root most subvol put the inode and return
524 * a reference to the dentry. We will have already gotten a reference
525 * to the inode in btrfs_fill_super so we're good to go.
527 if (!new && sb->s_root->d_inode == inode) {
529 return dget(sb->s_root);
533 const struct qstr name = { .name = "/", .len = 1 };
536 * New inode, we need to make the dentry a sibling of s_root so
537 * everything gets cleaned up properly on unmount.
539 dentry = d_alloc(sb->s_root, &name);
542 return ERR_PTR(-ENOMEM);
544 d_splice_alias(inode, dentry);
547 * We found the inode in cache, just find a dentry for it and
548 * put the reference to the inode we just got.
550 dentry = d_find_alias(inode);
557 static int btrfs_fill_super(struct super_block *sb,
558 struct btrfs_fs_devices *fs_devices,
559 void *data, int silent)
562 struct dentry *root_dentry;
563 struct btrfs_root *tree_root;
564 struct btrfs_key key;
567 sb->s_maxbytes = MAX_LFS_FILESIZE;
568 sb->s_magic = BTRFS_SUPER_MAGIC;
569 sb->s_op = &btrfs_super_ops;
570 sb->s_export_op = &btrfs_export_ops;
571 sb->s_xattr = btrfs_xattr_handlers;
573 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
574 sb->s_flags |= MS_POSIXACL;
577 tree_root = open_ctree(sb, fs_devices, (char *)data);
579 if (IS_ERR(tree_root)) {
580 printk("btrfs: open_ctree failed\n");
581 return PTR_ERR(tree_root);
583 sb->s_fs_info = tree_root;
585 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
586 key.type = BTRFS_INODE_ITEM_KEY;
588 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
590 err = PTR_ERR(inode);
594 root_dentry = d_alloc_root(inode);
601 sb->s_root = root_dentry;
603 save_mount_options(sb, data);
607 close_ctree(tree_root);
611 int btrfs_sync_fs(struct super_block *sb, int wait)
613 struct btrfs_trans_handle *trans;
614 struct btrfs_root *root = btrfs_sb(sb);
618 filemap_flush(root->fs_info->btree_inode->i_mapping);
622 btrfs_start_delalloc_inodes(root, 0);
623 btrfs_wait_ordered_extents(root, 0, 0);
625 trans = btrfs_start_transaction(root, 0);
627 return PTR_ERR(trans);
628 ret = btrfs_commit_transaction(trans, root);
632 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
634 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
635 struct btrfs_fs_info *info = root->fs_info;
637 if (btrfs_test_opt(root, DEGRADED))
638 seq_puts(seq, ",degraded");
639 if (btrfs_test_opt(root, NODATASUM))
640 seq_puts(seq, ",nodatasum");
641 if (btrfs_test_opt(root, NODATACOW))
642 seq_puts(seq, ",nodatacow");
643 if (btrfs_test_opt(root, NOBARRIER))
644 seq_puts(seq, ",nobarrier");
645 if (info->max_inline != 8192 * 1024)
646 seq_printf(seq, ",max_inline=%llu",
647 (unsigned long long)info->max_inline);
648 if (info->alloc_start != 0)
649 seq_printf(seq, ",alloc_start=%llu",
650 (unsigned long long)info->alloc_start);
651 if (info->thread_pool_size != min_t(unsigned long,
652 num_online_cpus() + 2, 8))
653 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
654 if (btrfs_test_opt(root, COMPRESS))
655 seq_puts(seq, ",compress");
656 if (btrfs_test_opt(root, NOSSD))
657 seq_puts(seq, ",nossd");
658 if (btrfs_test_opt(root, SSD_SPREAD))
659 seq_puts(seq, ",ssd_spread");
660 else if (btrfs_test_opt(root, SSD))
661 seq_puts(seq, ",ssd");
662 if (btrfs_test_opt(root, NOTREELOG))
663 seq_puts(seq, ",notreelog");
664 if (btrfs_test_opt(root, FLUSHONCOMMIT))
665 seq_puts(seq, ",flushoncommit");
666 if (btrfs_test_opt(root, DISCARD))
667 seq_puts(seq, ",discard");
668 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
669 seq_puts(seq, ",noacl");
673 static int btrfs_test_super(struct super_block *s, void *data)
675 struct btrfs_root *test_root = data;
676 struct btrfs_root *root = btrfs_sb(s);
679 * If this super block is going away, return false as it
680 * can't match as an existing super block.
682 if (!atomic_read(&s->s_active))
684 return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
687 static int btrfs_set_super(struct super_block *s, void *data)
691 return set_anon_super(s, data);
696 * Find a superblock for the given device / mount point.
698 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
699 * for multiple device setup. Make sure to keep it in sync.
701 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
702 const char *dev_name, void *data, struct vfsmount *mnt)
704 struct block_device *bdev = NULL;
705 struct super_block *s;
707 struct btrfs_fs_devices *fs_devices = NULL;
708 struct btrfs_root *tree_root = NULL;
709 struct btrfs_fs_info *fs_info = NULL;
710 fmode_t mode = FMODE_READ;
711 char *subvol_name = NULL;
712 u64 subvol_objectid = 0;
715 if (!(flags & MS_RDONLY))
718 error = btrfs_parse_early_options(data, mode, fs_type,
719 &subvol_name, &subvol_objectid,
724 error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
726 goto error_free_subvol_name;
728 error = btrfs_open_devices(fs_devices, mode, fs_type);
730 goto error_free_subvol_name;
732 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
734 goto error_close_devices;
738 * Setup a dummy root and fs_info for test/set super. This is because
739 * we don't actually fill this stuff out until open_ctree, but we need
740 * it for searching for existing supers, so this lets us do that and
741 * then open_ctree will properly initialize everything later.
743 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
744 tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
745 if (!fs_info || !tree_root) {
747 goto error_close_devices;
749 fs_info->tree_root = tree_root;
750 fs_info->fs_devices = fs_devices;
751 tree_root->fs_info = fs_info;
753 bdev = fs_devices->latest_bdev;
754 s = sget(fs_type, btrfs_test_super, btrfs_set_super, tree_root);
759 if ((flags ^ s->s_flags) & MS_RDONLY) {
760 deactivate_locked_super(s);
762 goto error_close_devices;
765 btrfs_close_devices(fs_devices);
769 char b[BDEVNAME_SIZE];
772 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
773 error = btrfs_fill_super(s, fs_devices, data,
774 flags & MS_SILENT ? 1 : 0);
776 deactivate_locked_super(s);
777 goto error_free_subvol_name;
780 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
781 s->s_flags |= MS_ACTIVE;
784 root = get_default_root(s, subvol_objectid);
786 error = PTR_ERR(root);
787 deactivate_locked_super(s);
788 goto error_free_subvol_name;
790 /* if they gave us a subvolume name bind mount into that */
791 if (strcmp(subvol_name, ".")) {
792 struct dentry *new_root;
793 mutex_lock(&root->d_inode->i_mutex);
794 new_root = lookup_one_len(subvol_name, root,
795 strlen(subvol_name));
796 mutex_unlock(&root->d_inode->i_mutex);
798 if (IS_ERR(new_root)) {
800 deactivate_locked_super(s);
801 error = PTR_ERR(new_root);
802 goto error_free_subvol_name;
804 if (!new_root->d_inode) {
807 deactivate_locked_super(s);
809 goto error_free_subvol_name;
816 mnt->mnt_root = root;
824 btrfs_close_devices(fs_devices);
827 error_free_subvol_name:
832 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
834 struct btrfs_root *root = btrfs_sb(sb);
837 ret = btrfs_parse_options(root, data);
841 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
844 if (*flags & MS_RDONLY) {
845 sb->s_flags |= MS_RDONLY;
847 ret = btrfs_commit_super(root);
850 if (root->fs_info->fs_devices->rw_devices == 0)
853 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
856 ret = btrfs_cleanup_fs_roots(root->fs_info);
859 /* recover relocation */
860 ret = btrfs_recover_relocation(root);
863 sb->s_flags &= ~MS_RDONLY;
870 * The helper to calc the free space on the devices that can be used to store
873 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
875 struct btrfs_fs_info *fs_info = root->fs_info;
876 struct btrfs_device_info *devices_info;
877 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
878 struct btrfs_device *device;
885 int i = 0, nr_devices;
888 nr_devices = fs_info->fs_devices->rw_devices;
891 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
896 /* calc min stripe number for data space alloction */
897 type = btrfs_get_alloc_profile(root, 1);
898 if (type & BTRFS_BLOCK_GROUP_RAID0)
900 else if (type & BTRFS_BLOCK_GROUP_RAID1)
902 else if (type & BTRFS_BLOCK_GROUP_RAID10)
905 if (type & BTRFS_BLOCK_GROUP_DUP)
906 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
908 min_stripe_size = BTRFS_STRIPE_LEN;
910 list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) {
911 if (!device->in_fs_metadata)
914 avail_space = device->total_bytes - device->bytes_used;
916 /* align with stripe_len */
917 do_div(avail_space, BTRFS_STRIPE_LEN);
918 avail_space *= BTRFS_STRIPE_LEN;
921 * In order to avoid overwritting the superblock on the drive,
922 * btrfs starts at an offset of at least 1MB when doing chunk
925 skip_space = 1024 * 1024;
927 /* user can set the offset in fs_info->alloc_start. */
928 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
930 skip_space = max(fs_info->alloc_start, skip_space);
933 * btrfs can not use the free space in [0, skip_space - 1],
934 * we must subtract it from the total. In order to implement
935 * it, we account the used space in this range first.
937 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
944 /* calc the free space in [0, skip_space - 1] */
945 skip_space -= used_space;
948 * we can use the free space in [0, skip_space - 1], subtract
951 if (avail_space && avail_space >= skip_space)
952 avail_space -= skip_space;
956 if (avail_space < min_stripe_size)
959 devices_info[i].dev = device;
960 devices_info[i].max_avail = avail_space;
967 btrfs_descending_sort_devices(devices_info, nr_devices);
971 while (nr_devices >= min_stripes) {
972 if (devices_info[i].max_avail >= min_stripe_size) {
976 avail_space += devices_info[i].max_avail * min_stripes;
977 alloc_size = devices_info[i].max_avail;
978 for (j = i + 1 - min_stripes; j <= i; j++)
979 devices_info[j].max_avail -= alloc_size;
986 *free_bytes = avail_space;
990 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
992 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
993 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
994 struct list_head *head = &root->fs_info->space_info;
995 struct btrfs_space_info *found;
997 u64 total_free_data = 0;
998 int bits = dentry->d_sb->s_blocksize_bits;
999 __be32 *fsid = (__be32 *)root->fs_info->fsid;
1002 /* holding chunk_muext to avoid allocating new chunks */
1003 mutex_lock(&root->fs_info->chunk_mutex);
1005 list_for_each_entry_rcu(found, head, list) {
1006 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1007 total_free_data += found->disk_total - found->disk_used;
1009 btrfs_account_ro_block_groups_free_space(found);
1012 total_used += found->disk_used;
1016 buf->f_namelen = BTRFS_NAME_LEN;
1017 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1018 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1019 buf->f_bsize = dentry->d_sb->s_blocksize;
1020 buf->f_type = BTRFS_SUPER_MAGIC;
1021 buf->f_bavail = total_free_data;
1022 ret = btrfs_calc_avail_data_space(root, &total_free_data);
1024 mutex_unlock(&root->fs_info->chunk_mutex);
1027 buf->f_bavail += total_free_data;
1028 buf->f_bavail = buf->f_bavail >> bits;
1029 mutex_unlock(&root->fs_info->chunk_mutex);
1031 /* We treat it as constant endianness (it doesn't matter _which_)
1032 because we want the fsid to come out the same whether mounted
1033 on a big-endian or little-endian host */
1034 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1035 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1036 /* Mask in the root object ID too, to disambiguate subvols */
1037 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1038 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1043 static struct file_system_type btrfs_fs_type = {
1044 .owner = THIS_MODULE,
1046 .get_sb = btrfs_get_sb,
1047 .kill_sb = kill_anon_super,
1048 .fs_flags = FS_REQUIRES_DEV,
1052 * used by btrfsctl to scan devices when no FS is mounted
1054 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1057 struct btrfs_ioctl_vol_args *vol;
1058 struct btrfs_fs_devices *fs_devices;
1061 if (!capable(CAP_SYS_ADMIN))
1064 vol = memdup_user((void __user *)arg, sizeof(*vol));
1066 return PTR_ERR(vol);
1069 case BTRFS_IOC_SCAN_DEV:
1070 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1071 &btrfs_fs_type, &fs_devices);
1079 static int btrfs_freeze(struct super_block *sb)
1081 struct btrfs_root *root = btrfs_sb(sb);
1082 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1083 mutex_lock(&root->fs_info->cleaner_mutex);
1087 static int btrfs_unfreeze(struct super_block *sb)
1089 struct btrfs_root *root = btrfs_sb(sb);
1090 mutex_unlock(&root->fs_info->cleaner_mutex);
1091 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1095 static const struct super_operations btrfs_super_ops = {
1096 .drop_inode = btrfs_drop_inode,
1097 .evict_inode = btrfs_evict_inode,
1098 .put_super = btrfs_put_super,
1099 .sync_fs = btrfs_sync_fs,
1100 .show_options = btrfs_show_options,
1101 .write_inode = btrfs_write_inode,
1102 .dirty_inode = btrfs_dirty_inode,
1103 .alloc_inode = btrfs_alloc_inode,
1104 .destroy_inode = btrfs_destroy_inode,
1105 .statfs = btrfs_statfs,
1106 .remount_fs = btrfs_remount,
1107 .freeze_fs = btrfs_freeze,
1108 .unfreeze_fs = btrfs_unfreeze,
1111 static const struct file_operations btrfs_ctl_fops = {
1112 .unlocked_ioctl = btrfs_control_ioctl,
1113 .compat_ioctl = btrfs_control_ioctl,
1114 .owner = THIS_MODULE,
1117 static struct miscdevice btrfs_misc = {
1118 .minor = BTRFS_MINOR,
1119 .name = "btrfs-control",
1120 .fops = &btrfs_ctl_fops
1123 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1124 MODULE_ALIAS("devname:btrfs-control");
1126 static int btrfs_interface_init(void)
1128 return misc_register(&btrfs_misc);
1131 static void btrfs_interface_exit(void)
1133 if (misc_deregister(&btrfs_misc) < 0)
1134 printk(KERN_INFO "misc_deregister failed for control device");
1137 static int __init init_btrfs_fs(void)
1141 err = btrfs_init_sysfs();
1145 err = btrfs_init_compress();
1149 err = btrfs_init_cachep();
1153 err = extent_io_init();
1157 err = extent_map_init();
1159 goto free_extent_io;
1161 err = btrfs_interface_init();
1163 goto free_extent_map;
1165 err = register_filesystem(&btrfs_fs_type);
1167 goto unregister_ioctl;
1169 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1173 btrfs_interface_exit();
1179 btrfs_destroy_cachep();
1181 btrfs_exit_compress();
1187 static void __exit exit_btrfs_fs(void)
1189 btrfs_destroy_cachep();
1192 btrfs_interface_exit();
1193 unregister_filesystem(&btrfs_fs_type);
1195 btrfs_cleanup_fs_uuids();
1196 btrfs_exit_compress();
1199 module_init(init_btrfs_fs)
1200 module_exit(exit_btrfs_fs)
1202 MODULE_LICENSE("GPL");