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,
168 Opt_inode_cache, Opt_no_space_cache, Opt_recovery, Opt_err,
171 static match_table_t tokens = {
172 {Opt_degraded, "degraded"},
173 {Opt_subvol, "subvol=%s"},
174 {Opt_subvolid, "subvolid=%d"},
175 {Opt_device, "device=%s"},
176 {Opt_nodatasum, "nodatasum"},
177 {Opt_nodatacow, "nodatacow"},
178 {Opt_nobarrier, "nobarrier"},
179 {Opt_max_inline, "max_inline=%s"},
180 {Opt_alloc_start, "alloc_start=%s"},
181 {Opt_thread_pool, "thread_pool=%d"},
182 {Opt_compress, "compress"},
183 {Opt_compress_type, "compress=%s"},
184 {Opt_compress_force, "compress-force"},
185 {Opt_compress_force_type, "compress-force=%s"},
187 {Opt_ssd_spread, "ssd_spread"},
188 {Opt_nossd, "nossd"},
189 {Opt_noacl, "noacl"},
190 {Opt_notreelog, "notreelog"},
191 {Opt_flushoncommit, "flushoncommit"},
192 {Opt_ratio, "metadata_ratio=%d"},
193 {Opt_discard, "discard"},
194 {Opt_space_cache, "space_cache"},
195 {Opt_clear_cache, "clear_cache"},
196 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
197 {Opt_enospc_debug, "enospc_debug"},
198 {Opt_subvolrootid, "subvolrootid=%d"},
199 {Opt_defrag, "autodefrag"},
200 {Opt_inode_cache, "inode_cache"},
201 {Opt_no_space_cache, "nospace_cache"},
202 {Opt_recovery, "recovery"},
207 * Regular mount options parser. Everything that is needed only when
208 * reading in a new superblock is parsed here.
210 int btrfs_parse_options(struct btrfs_root *root, char *options)
212 struct btrfs_fs_info *info = root->fs_info;
213 substring_t args[MAX_OPT_ARGS];
214 char *p, *num, *orig = NULL;
219 bool compress_force = false;
221 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
223 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
229 * strsep changes the string, duplicate it because parse_options
232 options = kstrdup(options, GFP_NOFS);
238 while ((p = strsep(&options, ",")) != NULL) {
243 token = match_token(p, tokens, args);
246 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
247 btrfs_set_opt(info->mount_opt, DEGRADED);
251 case Opt_subvolrootid:
254 * These are parsed by btrfs_parse_early_options
255 * and can be happily ignored here.
259 printk(KERN_INFO "btrfs: setting nodatasum\n");
260 btrfs_set_opt(info->mount_opt, NODATASUM);
263 printk(KERN_INFO "btrfs: setting nodatacow\n");
264 btrfs_set_opt(info->mount_opt, NODATACOW);
265 btrfs_set_opt(info->mount_opt, NODATASUM);
267 case Opt_compress_force:
268 case Opt_compress_force_type:
269 compress_force = true;
271 case Opt_compress_type:
272 if (token == Opt_compress ||
273 token == Opt_compress_force ||
274 strcmp(args[0].from, "zlib") == 0) {
275 compress_type = "zlib";
276 info->compress_type = BTRFS_COMPRESS_ZLIB;
277 } else if (strcmp(args[0].from, "lzo") == 0) {
278 compress_type = "lzo";
279 info->compress_type = BTRFS_COMPRESS_LZO;
285 btrfs_set_opt(info->mount_opt, COMPRESS);
286 if (compress_force) {
287 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
288 pr_info("btrfs: force %s compression\n",
291 pr_info("btrfs: use %s compression\n",
295 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
296 btrfs_set_opt(info->mount_opt, SSD);
299 printk(KERN_INFO "btrfs: use spread ssd "
300 "allocation scheme\n");
301 btrfs_set_opt(info->mount_opt, SSD);
302 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
305 printk(KERN_INFO "btrfs: not using ssd allocation "
307 btrfs_set_opt(info->mount_opt, NOSSD);
308 btrfs_clear_opt(info->mount_opt, SSD);
309 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
312 printk(KERN_INFO "btrfs: turning off barriers\n");
313 btrfs_set_opt(info->mount_opt, NOBARRIER);
315 case Opt_thread_pool:
317 match_int(&args[0], &intarg);
319 info->thread_pool_size = intarg;
320 printk(KERN_INFO "btrfs: thread pool %d\n",
321 info->thread_pool_size);
325 num = match_strdup(&args[0]);
327 info->max_inline = memparse(num, NULL);
330 if (info->max_inline) {
331 info->max_inline = max_t(u64,
335 printk(KERN_INFO "btrfs: max_inline at %llu\n",
336 (unsigned long long)info->max_inline);
339 case Opt_alloc_start:
340 num = match_strdup(&args[0]);
342 info->alloc_start = memparse(num, NULL);
345 "btrfs: allocations start at %llu\n",
346 (unsigned long long)info->alloc_start);
350 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
353 printk(KERN_INFO "btrfs: disabling tree log\n");
354 btrfs_set_opt(info->mount_opt, NOTREELOG);
356 case Opt_flushoncommit:
357 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
358 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
362 match_int(&args[0], &intarg);
364 info->metadata_ratio = intarg;
365 printk(KERN_INFO "btrfs: metadata ratio %d\n",
366 info->metadata_ratio);
370 btrfs_set_opt(info->mount_opt, DISCARD);
372 case Opt_space_cache:
373 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
375 case Opt_no_space_cache:
376 printk(KERN_INFO "btrfs: disabling disk space caching\n");
377 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
379 case Opt_inode_cache:
380 printk(KERN_INFO "btrfs: enabling inode map caching\n");
381 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
383 case Opt_clear_cache:
384 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
385 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
387 case Opt_user_subvol_rm_allowed:
388 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
390 case Opt_enospc_debug:
391 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
394 printk(KERN_INFO "btrfs: enabling auto defrag");
395 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
398 printk(KERN_INFO "btrfs: enabling auto recovery");
399 btrfs_set_opt(info->mount_opt, RECOVERY);
402 printk(KERN_INFO "btrfs: unrecognized mount option "
411 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
412 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
418 * Parse mount options that are required early in the mount process.
420 * All other options will be parsed on much later in the mount process and
421 * only when we need to allocate a new super block.
423 static int btrfs_parse_early_options(const char *options, fmode_t flags,
424 void *holder, char **subvol_name, u64 *subvol_objectid,
425 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
427 substring_t args[MAX_OPT_ARGS];
428 char *device_name, *opts, *orig, *p;
436 * strsep changes the string, duplicate it because parse_options
439 opts = kstrdup(options, GFP_KERNEL);
444 while ((p = strsep(&opts, ",")) != NULL) {
449 token = match_token(p, tokens, args);
453 *subvol_name = match_strdup(&args[0]);
457 error = match_int(&args[0], &intarg);
459 /* we want the original fs_tree */
462 BTRFS_FS_TREE_OBJECTID;
464 *subvol_objectid = intarg;
467 case Opt_subvolrootid:
469 error = match_int(&args[0], &intarg);
471 /* we want the original fs_tree */
474 BTRFS_FS_TREE_OBJECTID;
476 *subvol_rootid = intarg;
480 device_name = match_strdup(&args[0]);
485 error = btrfs_scan_one_device(device_name,
486 flags, holder, fs_devices);
501 static struct dentry *get_default_root(struct super_block *sb,
504 struct btrfs_root *root = sb->s_fs_info;
505 struct btrfs_root *new_root;
506 struct btrfs_dir_item *di;
507 struct btrfs_path *path;
508 struct btrfs_key location;
514 * We have a specific subvol we want to mount, just setup location and
515 * go look up the root.
517 if (subvol_objectid) {
518 location.objectid = subvol_objectid;
519 location.type = BTRFS_ROOT_ITEM_KEY;
520 location.offset = (u64)-1;
524 path = btrfs_alloc_path();
526 return ERR_PTR(-ENOMEM);
527 path->leave_spinning = 1;
530 * Find the "default" dir item which points to the root item that we
531 * will mount by default if we haven't been given a specific subvolume
534 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
535 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
537 btrfs_free_path(path);
542 * Ok the default dir item isn't there. This is weird since
543 * it's always been there, but don't freak out, just try and
544 * mount to root most subvolume.
546 btrfs_free_path(path);
547 dir_id = BTRFS_FIRST_FREE_OBJECTID;
548 new_root = root->fs_info->fs_root;
552 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
553 btrfs_free_path(path);
556 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
557 if (IS_ERR(new_root))
558 return ERR_CAST(new_root);
560 if (btrfs_root_refs(&new_root->root_item) == 0)
561 return ERR_PTR(-ENOENT);
563 dir_id = btrfs_root_dirid(&new_root->root_item);
565 location.objectid = dir_id;
566 location.type = BTRFS_INODE_ITEM_KEY;
569 inode = btrfs_iget(sb, &location, new_root, &new);
571 return ERR_CAST(inode);
574 * If we're just mounting the root most subvol put the inode and return
575 * a reference to the dentry. We will have already gotten a reference
576 * to the inode in btrfs_fill_super so we're good to go.
578 if (!new && sb->s_root->d_inode == inode) {
580 return dget(sb->s_root);
583 return d_obtain_alias(inode);
586 static int btrfs_fill_super(struct super_block *sb,
587 struct btrfs_fs_devices *fs_devices,
588 void *data, int silent)
591 struct dentry *root_dentry;
592 struct btrfs_root *tree_root;
593 struct btrfs_key key;
596 sb->s_maxbytes = MAX_LFS_FILESIZE;
597 sb->s_magic = BTRFS_SUPER_MAGIC;
598 sb->s_op = &btrfs_super_ops;
599 sb->s_d_op = &btrfs_dentry_operations;
600 sb->s_export_op = &btrfs_export_ops;
601 sb->s_xattr = btrfs_xattr_handlers;
603 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
604 sb->s_flags |= MS_POSIXACL;
607 tree_root = open_ctree(sb, fs_devices, (char *)data);
609 if (IS_ERR(tree_root)) {
610 printk("btrfs: open_ctree failed\n");
611 return PTR_ERR(tree_root);
613 sb->s_fs_info = tree_root;
615 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
616 key.type = BTRFS_INODE_ITEM_KEY;
618 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
620 err = PTR_ERR(inode);
624 root_dentry = d_alloc_root(inode);
631 sb->s_root = root_dentry;
633 save_mount_options(sb, data);
634 cleancache_init_fs(sb);
638 close_ctree(tree_root);
642 int btrfs_sync_fs(struct super_block *sb, int wait)
644 struct btrfs_trans_handle *trans;
645 struct btrfs_root *root = btrfs_sb(sb);
648 trace_btrfs_sync_fs(wait);
651 filemap_flush(root->fs_info->btree_inode->i_mapping);
655 btrfs_start_delalloc_inodes(root, 0);
656 btrfs_wait_ordered_extents(root, 0, 0);
658 trans = btrfs_start_transaction(root, 0);
660 return PTR_ERR(trans);
661 ret = btrfs_commit_transaction(trans, root);
665 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
667 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
668 struct btrfs_fs_info *info = root->fs_info;
671 if (btrfs_test_opt(root, DEGRADED))
672 seq_puts(seq, ",degraded");
673 if (btrfs_test_opt(root, NODATASUM))
674 seq_puts(seq, ",nodatasum");
675 if (btrfs_test_opt(root, NODATACOW))
676 seq_puts(seq, ",nodatacow");
677 if (btrfs_test_opt(root, NOBARRIER))
678 seq_puts(seq, ",nobarrier");
679 if (info->max_inline != 8192 * 1024)
680 seq_printf(seq, ",max_inline=%llu",
681 (unsigned long long)info->max_inline);
682 if (info->alloc_start != 0)
683 seq_printf(seq, ",alloc_start=%llu",
684 (unsigned long long)info->alloc_start);
685 if (info->thread_pool_size != min_t(unsigned long,
686 num_online_cpus() + 2, 8))
687 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
688 if (btrfs_test_opt(root, COMPRESS)) {
689 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
690 compress_type = "zlib";
692 compress_type = "lzo";
693 if (btrfs_test_opt(root, FORCE_COMPRESS))
694 seq_printf(seq, ",compress-force=%s", compress_type);
696 seq_printf(seq, ",compress=%s", compress_type);
698 if (btrfs_test_opt(root, NOSSD))
699 seq_puts(seq, ",nossd");
700 if (btrfs_test_opt(root, SSD_SPREAD))
701 seq_puts(seq, ",ssd_spread");
702 else if (btrfs_test_opt(root, SSD))
703 seq_puts(seq, ",ssd");
704 if (btrfs_test_opt(root, NOTREELOG))
705 seq_puts(seq, ",notreelog");
706 if (btrfs_test_opt(root, FLUSHONCOMMIT))
707 seq_puts(seq, ",flushoncommit");
708 if (btrfs_test_opt(root, DISCARD))
709 seq_puts(seq, ",discard");
710 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
711 seq_puts(seq, ",noacl");
712 if (btrfs_test_opt(root, SPACE_CACHE))
713 seq_puts(seq, ",space_cache");
715 seq_puts(seq, ",nospace_cache");
716 if (btrfs_test_opt(root, CLEAR_CACHE))
717 seq_puts(seq, ",clear_cache");
718 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
719 seq_puts(seq, ",user_subvol_rm_allowed");
720 if (btrfs_test_opt(root, ENOSPC_DEBUG))
721 seq_puts(seq, ",enospc_debug");
722 if (btrfs_test_opt(root, AUTO_DEFRAG))
723 seq_puts(seq, ",autodefrag");
724 if (btrfs_test_opt(root, INODE_MAP_CACHE))
725 seq_puts(seq, ",inode_cache");
729 static int btrfs_test_super(struct super_block *s, void *data)
731 struct btrfs_root *test_root = data;
732 struct btrfs_root *root = btrfs_sb(s);
735 * If this super block is going away, return false as it
736 * can't match as an existing super block.
738 if (!atomic_read(&s->s_active))
740 return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
743 static int btrfs_set_super(struct super_block *s, void *data)
747 return set_anon_super(s, data);
751 * subvolumes are identified by ino 256
753 static inline int is_subvolume_inode(struct inode *inode)
755 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
761 * This will strip out the subvol=%s argument for an argument string and add
762 * subvolid=0 to make sure we get the actual tree root for path walking to the
765 static char *setup_root_args(char *args)
768 unsigned len = strlen(args) + 2;
773 * We need the same args as before, but minus
781 * which is a difference of 2 characters, so we allocate strlen(args) +
784 ret = kzalloc(len * sizeof(char), GFP_NOFS);
787 pos = strstr(args, "subvol=");
789 /* This shouldn't happen, but just in case.. */
796 * The subvol=<> arg is not at the front of the string, copy everybody
797 * up to that into ret.
802 copied += strlen(args);
806 strncpy(ret + copied, "subvolid=0", len - copied);
808 /* Length of subvolid=0 */
812 * If there is no , after the subvol= option then we know there's no
813 * other options and we can just return.
815 pos = strchr(pos, ',');
819 /* Copy the rest of the arguments into our buffer */
820 strncpy(ret + copied, pos, len - copied);
821 copied += strlen(pos);
826 static struct dentry *mount_subvol(const char *subvol_name, int flags,
827 const char *device_name, char *data)
829 struct super_block *s;
831 struct vfsmount *mnt;
832 struct mnt_namespace *ns_private;
837 newargs = setup_root_args(data);
839 return ERR_PTR(-ENOMEM);
840 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
844 return ERR_CAST(mnt);
846 ns_private = create_mnt_ns(mnt);
847 if (IS_ERR(ns_private)) {
849 return ERR_CAST(ns_private);
853 * This will trigger the automount of the subvol so we can just
854 * drop the mnt we have here and return the dentry that we
857 error = vfs_path_lookup(mnt->mnt_root, mnt, subvol_name,
858 LOOKUP_FOLLOW, &path);
859 put_mnt_ns(ns_private);
861 return ERR_PTR(error);
863 if (!is_subvolume_inode(path.dentry->d_inode)) {
867 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
869 return ERR_PTR(-EINVAL);
872 /* Get a ref to the sb and the dentry we found and return it */
873 s = path.mnt->mnt_sb;
874 atomic_inc(&s->s_active);
875 root = dget(path.dentry);
877 down_write(&s->s_umount);
883 * Find a superblock for the given device / mount point.
885 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
886 * for multiple device setup. Make sure to keep it in sync.
888 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
889 const char *device_name, void *data)
891 struct block_device *bdev = NULL;
892 struct super_block *s;
894 struct btrfs_fs_devices *fs_devices = NULL;
895 struct btrfs_fs_info *fs_info = NULL;
896 fmode_t mode = FMODE_READ;
897 char *subvol_name = NULL;
898 u64 subvol_objectid = 0;
899 u64 subvol_rootid = 0;
902 if (!(flags & MS_RDONLY))
905 error = btrfs_parse_early_options(data, mode, fs_type,
906 &subvol_name, &subvol_objectid,
907 &subvol_rootid, &fs_devices);
910 return ERR_PTR(error);
914 root = mount_subvol(subvol_name, flags, device_name, data);
919 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
921 return ERR_PTR(error);
924 * Setup a dummy root and fs_info for test/set super. This is because
925 * we don't actually fill this stuff out until open_ctree, but we need
926 * it for searching for existing supers, so this lets us do that and
927 * then open_ctree will properly initialize everything later.
929 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
931 return ERR_PTR(-ENOMEM);
933 fs_info->tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
934 if (!fs_info->tree_root) {
938 fs_info->tree_root->fs_info = fs_info;
939 fs_info->fs_devices = fs_devices;
941 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
942 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
943 if (!fs_info->super_copy || !fs_info->super_for_commit) {
948 error = btrfs_open_devices(fs_devices, mode, fs_type);
952 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
954 goto error_close_devices;
957 bdev = fs_devices->latest_bdev;
958 s = sget(fs_type, btrfs_test_super, btrfs_set_super,
962 goto error_close_devices;
966 if ((flags ^ s->s_flags) & MS_RDONLY) {
967 deactivate_locked_super(s);
969 goto error_close_devices;
972 btrfs_close_devices(fs_devices);
973 free_fs_info(fs_info);
975 char b[BDEVNAME_SIZE];
977 s->s_flags = flags | MS_NOSEC;
978 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
979 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
980 error = btrfs_fill_super(s, fs_devices, data,
981 flags & MS_SILENT ? 1 : 0);
983 deactivate_locked_super(s);
984 return ERR_PTR(error);
987 s->s_flags |= MS_ACTIVE;
990 root = get_default_root(s, subvol_objectid);
992 deactivate_locked_super(s);
999 btrfs_close_devices(fs_devices);
1001 free_fs_info(fs_info);
1002 return ERR_PTR(error);
1005 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1007 struct btrfs_root *root = btrfs_sb(sb);
1010 ret = btrfs_parse_options(root, data);
1014 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1017 if (*flags & MS_RDONLY) {
1018 sb->s_flags |= MS_RDONLY;
1020 ret = btrfs_commit_super(root);
1023 if (root->fs_info->fs_devices->rw_devices == 0)
1026 if (btrfs_super_log_root(root->fs_info->super_copy) != 0)
1029 ret = btrfs_cleanup_fs_roots(root->fs_info);
1032 /* recover relocation */
1033 ret = btrfs_recover_relocation(root);
1036 sb->s_flags &= ~MS_RDONLY;
1042 /* Used to sort the devices by max_avail(descending sort) */
1043 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1044 const void *dev_info2)
1046 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1047 ((struct btrfs_device_info *)dev_info2)->max_avail)
1049 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1050 ((struct btrfs_device_info *)dev_info2)->max_avail)
1057 * sort the devices by max_avail, in which max free extent size of each device
1058 * is stored.(Descending Sort)
1060 static inline void btrfs_descending_sort_devices(
1061 struct btrfs_device_info *devices,
1064 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1065 btrfs_cmp_device_free_bytes, NULL);
1069 * The helper to calc the free space on the devices that can be used to store
1072 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1074 struct btrfs_fs_info *fs_info = root->fs_info;
1075 struct btrfs_device_info *devices_info;
1076 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1077 struct btrfs_device *device;
1082 u64 min_stripe_size;
1083 int min_stripes = 1, num_stripes = 1;
1084 int i = 0, nr_devices;
1087 nr_devices = fs_info->fs_devices->open_devices;
1088 BUG_ON(!nr_devices);
1090 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1095 /* calc min stripe number for data space alloction */
1096 type = btrfs_get_alloc_profile(root, 1);
1097 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1099 num_stripes = nr_devices;
1100 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1103 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1108 if (type & BTRFS_BLOCK_GROUP_DUP)
1109 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1111 min_stripe_size = BTRFS_STRIPE_LEN;
1113 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1114 if (!device->in_fs_metadata || !device->bdev)
1117 avail_space = device->total_bytes - device->bytes_used;
1119 /* align with stripe_len */
1120 do_div(avail_space, BTRFS_STRIPE_LEN);
1121 avail_space *= BTRFS_STRIPE_LEN;
1124 * In order to avoid overwritting the superblock on the drive,
1125 * btrfs starts at an offset of at least 1MB when doing chunk
1128 skip_space = 1024 * 1024;
1130 /* user can set the offset in fs_info->alloc_start. */
1131 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1132 device->total_bytes)
1133 skip_space = max(fs_info->alloc_start, skip_space);
1136 * btrfs can not use the free space in [0, skip_space - 1],
1137 * we must subtract it from the total. In order to implement
1138 * it, we account the used space in this range first.
1140 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1143 kfree(devices_info);
1147 /* calc the free space in [0, skip_space - 1] */
1148 skip_space -= used_space;
1151 * we can use the free space in [0, skip_space - 1], subtract
1152 * it from the total.
1154 if (avail_space && avail_space >= skip_space)
1155 avail_space -= skip_space;
1159 if (avail_space < min_stripe_size)
1162 devices_info[i].dev = device;
1163 devices_info[i].max_avail = avail_space;
1170 btrfs_descending_sort_devices(devices_info, nr_devices);
1174 while (nr_devices >= min_stripes) {
1175 if (num_stripes > nr_devices)
1176 num_stripes = nr_devices;
1178 if (devices_info[i].max_avail >= min_stripe_size) {
1182 avail_space += devices_info[i].max_avail * num_stripes;
1183 alloc_size = devices_info[i].max_avail;
1184 for (j = i + 1 - num_stripes; j <= i; j++)
1185 devices_info[j].max_avail -= alloc_size;
1191 kfree(devices_info);
1192 *free_bytes = avail_space;
1196 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1198 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
1199 struct btrfs_super_block *disk_super = root->fs_info->super_copy;
1200 struct list_head *head = &root->fs_info->space_info;
1201 struct btrfs_space_info *found;
1203 u64 total_free_data = 0;
1204 int bits = dentry->d_sb->s_blocksize_bits;
1205 __be32 *fsid = (__be32 *)root->fs_info->fsid;
1208 /* holding chunk_muext to avoid allocating new chunks */
1209 mutex_lock(&root->fs_info->chunk_mutex);
1211 list_for_each_entry_rcu(found, head, list) {
1212 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1213 total_free_data += found->disk_total - found->disk_used;
1215 btrfs_account_ro_block_groups_free_space(found);
1218 total_used += found->disk_used;
1222 buf->f_namelen = BTRFS_NAME_LEN;
1223 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1224 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1225 buf->f_bsize = dentry->d_sb->s_blocksize;
1226 buf->f_type = BTRFS_SUPER_MAGIC;
1227 buf->f_bavail = total_free_data;
1228 ret = btrfs_calc_avail_data_space(root, &total_free_data);
1230 mutex_unlock(&root->fs_info->chunk_mutex);
1233 buf->f_bavail += total_free_data;
1234 buf->f_bavail = buf->f_bavail >> bits;
1235 mutex_unlock(&root->fs_info->chunk_mutex);
1237 /* We treat it as constant endianness (it doesn't matter _which_)
1238 because we want the fsid to come out the same whether mounted
1239 on a big-endian or little-endian host */
1240 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1241 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1242 /* Mask in the root object ID too, to disambiguate subvols */
1243 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1244 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1249 static struct file_system_type btrfs_fs_type = {
1250 .owner = THIS_MODULE,
1252 .mount = btrfs_mount,
1253 .kill_sb = kill_anon_super,
1254 .fs_flags = FS_REQUIRES_DEV,
1258 * used by btrfsctl to scan devices when no FS is mounted
1260 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1263 struct btrfs_ioctl_vol_args *vol;
1264 struct btrfs_fs_devices *fs_devices;
1267 if (!capable(CAP_SYS_ADMIN))
1270 vol = memdup_user((void __user *)arg, sizeof(*vol));
1272 return PTR_ERR(vol);
1275 case BTRFS_IOC_SCAN_DEV:
1276 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1277 &btrfs_fs_type, &fs_devices);
1285 static int btrfs_freeze(struct super_block *sb)
1287 struct btrfs_root *root = btrfs_sb(sb);
1288 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1289 mutex_lock(&root->fs_info->cleaner_mutex);
1293 static int btrfs_unfreeze(struct super_block *sb)
1295 struct btrfs_root *root = btrfs_sb(sb);
1296 mutex_unlock(&root->fs_info->cleaner_mutex);
1297 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1301 static void btrfs_fs_dirty_inode(struct inode *inode, int flags)
1305 ret = btrfs_dirty_inode(inode);
1307 printk_ratelimited(KERN_ERR "btrfs: fail to dirty inode %Lu "
1308 "error %d\n", btrfs_ino(inode), ret);
1311 static const struct super_operations btrfs_super_ops = {
1312 .drop_inode = btrfs_drop_inode,
1313 .evict_inode = btrfs_evict_inode,
1314 .put_super = btrfs_put_super,
1315 .sync_fs = btrfs_sync_fs,
1316 .show_options = btrfs_show_options,
1317 .write_inode = btrfs_write_inode,
1318 .dirty_inode = btrfs_fs_dirty_inode,
1319 .alloc_inode = btrfs_alloc_inode,
1320 .destroy_inode = btrfs_destroy_inode,
1321 .statfs = btrfs_statfs,
1322 .remount_fs = btrfs_remount,
1323 .freeze_fs = btrfs_freeze,
1324 .unfreeze_fs = btrfs_unfreeze,
1327 static const struct file_operations btrfs_ctl_fops = {
1328 .unlocked_ioctl = btrfs_control_ioctl,
1329 .compat_ioctl = btrfs_control_ioctl,
1330 .owner = THIS_MODULE,
1331 .llseek = noop_llseek,
1334 static struct miscdevice btrfs_misc = {
1335 .minor = BTRFS_MINOR,
1336 .name = "btrfs-control",
1337 .fops = &btrfs_ctl_fops
1340 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1341 MODULE_ALIAS("devname:btrfs-control");
1343 static int btrfs_interface_init(void)
1345 return misc_register(&btrfs_misc);
1348 static void btrfs_interface_exit(void)
1350 if (misc_deregister(&btrfs_misc) < 0)
1351 printk(KERN_INFO "misc_deregister failed for control device");
1354 static int __init init_btrfs_fs(void)
1358 err = btrfs_init_sysfs();
1362 err = btrfs_init_compress();
1366 err = btrfs_init_cachep();
1370 err = extent_io_init();
1374 err = extent_map_init();
1376 goto free_extent_io;
1378 err = btrfs_delayed_inode_init();
1380 goto free_extent_map;
1382 err = btrfs_interface_init();
1384 goto free_delayed_inode;
1386 err = register_filesystem(&btrfs_fs_type);
1388 goto unregister_ioctl;
1390 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1394 btrfs_interface_exit();
1396 btrfs_delayed_inode_exit();
1402 btrfs_destroy_cachep();
1404 btrfs_exit_compress();
1410 static void __exit exit_btrfs_fs(void)
1412 btrfs_destroy_cachep();
1413 btrfs_delayed_inode_exit();
1416 btrfs_interface_exit();
1417 unregister_filesystem(&btrfs_fs_type);
1419 btrfs_cleanup_fs_uuids();
1420 btrfs_exit_compress();
1423 module_init(init_btrfs_fs)
1424 module_exit(exit_btrfs_fs)
1426 MODULE_LICENSE("GPL");