Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux...
[linux-2.6.git] / fs / btrfs / super.c
blob8ffaaa8e3df80fff13a75b9d6443373624e70753
1 /*
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>
22 #include <linux/fs.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/ratelimit.h>
44 #include "compat.h"
45 #include "delayed-inode.h"
46 #include "ctree.h"
47 #include "disk-io.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "ioctl.h"
51 #include "print-tree.h"
52 #include "xattr.h"
53 #include "volumes.h"
54 #include "version.h"
55 #include "export.h"
56 #include "compression.h"
58 #define CREATE_TRACE_POINTS
59 #include <trace/events/btrfs.h>
61 static const struct super_operations btrfs_super_ops;
62 static struct file_system_type btrfs_fs_type;
64 static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
65 char nbuf[16])
67 char *errstr = NULL;
69 switch (errno) {
70 case -EIO:
71 errstr = "IO failure";
72 break;
73 case -ENOMEM:
74 errstr = "Out of memory";
75 break;
76 case -EROFS:
77 errstr = "Readonly filesystem";
78 break;
79 default:
80 if (nbuf) {
81 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
82 errstr = nbuf;
84 break;
87 return errstr;
90 static void __save_error_info(struct btrfs_fs_info *fs_info)
93 * today we only save the error info into ram. Long term we'll
94 * also send it down to the disk
96 fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
99 /* NOTE:
100 * We move write_super stuff at umount in order to avoid deadlock
101 * for umount hold all lock.
103 static void save_error_info(struct btrfs_fs_info *fs_info)
105 __save_error_info(fs_info);
108 /* btrfs handle error by forcing the filesystem readonly */
109 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
111 struct super_block *sb = fs_info->sb;
113 if (sb->s_flags & MS_RDONLY)
114 return;
116 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
117 sb->s_flags |= MS_RDONLY;
118 printk(KERN_INFO "btrfs is forced readonly\n");
123 * __btrfs_std_error decodes expected errors from the caller and
124 * invokes the approciate error response.
126 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
127 unsigned int line, int errno)
129 struct super_block *sb = fs_info->sb;
130 char nbuf[16];
131 const char *errstr;
134 * Special case: if the error is EROFS, and we're already
135 * under MS_RDONLY, then it is safe here.
137 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
138 return;
140 errstr = btrfs_decode_error(fs_info, errno, nbuf);
141 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
142 sb->s_id, function, line, errstr);
143 save_error_info(fs_info);
145 btrfs_handle_error(fs_info);
148 static void btrfs_put_super(struct super_block *sb)
150 struct btrfs_root *root = btrfs_sb(sb);
151 int ret;
153 ret = close_ctree(root);
154 sb->s_fs_info = NULL;
156 (void)ret; /* FIXME: need to fix VFS to return error? */
159 enum {
160 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
161 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
162 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
163 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
164 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
165 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
166 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
167 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
168 Opt_check_integrity, Opt_check_integrity_including_extent_data,
169 Opt_check_integrity_print_mask,
170 Opt_err,
173 static match_table_t tokens = {
174 {Opt_degraded, "degraded"},
175 {Opt_subvol, "subvol=%s"},
176 {Opt_subvolid, "subvolid=%d"},
177 {Opt_device, "device=%s"},
178 {Opt_nodatasum, "nodatasum"},
179 {Opt_nodatacow, "nodatacow"},
180 {Opt_nobarrier, "nobarrier"},
181 {Opt_max_inline, "max_inline=%s"},
182 {Opt_alloc_start, "alloc_start=%s"},
183 {Opt_thread_pool, "thread_pool=%d"},
184 {Opt_compress, "compress"},
185 {Opt_compress_type, "compress=%s"},
186 {Opt_compress_force, "compress-force"},
187 {Opt_compress_force_type, "compress-force=%s"},
188 {Opt_ssd, "ssd"},
189 {Opt_ssd_spread, "ssd_spread"},
190 {Opt_nossd, "nossd"},
191 {Opt_noacl, "noacl"},
192 {Opt_notreelog, "notreelog"},
193 {Opt_flushoncommit, "flushoncommit"},
194 {Opt_ratio, "metadata_ratio=%d"},
195 {Opt_discard, "discard"},
196 {Opt_space_cache, "space_cache"},
197 {Opt_clear_cache, "clear_cache"},
198 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
199 {Opt_enospc_debug, "enospc_debug"},
200 {Opt_subvolrootid, "subvolrootid=%d"},
201 {Opt_defrag, "autodefrag"},
202 {Opt_inode_cache, "inode_cache"},
203 {Opt_no_space_cache, "nospace_cache"},
204 {Opt_recovery, "recovery"},
205 {Opt_skip_balance, "skip_balance"},
206 {Opt_check_integrity, "check_int"},
207 {Opt_check_integrity_including_extent_data, "check_int_data"},
208 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
209 {Opt_err, NULL},
213 * Regular mount options parser. Everything that is needed only when
214 * reading in a new superblock is parsed here.
216 int btrfs_parse_options(struct btrfs_root *root, char *options)
218 struct btrfs_fs_info *info = root->fs_info;
219 substring_t args[MAX_OPT_ARGS];
220 char *p, *num, *orig = NULL;
221 u64 cache_gen;
222 int intarg;
223 int ret = 0;
224 char *compress_type;
225 bool compress_force = false;
227 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
228 if (cache_gen)
229 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
231 if (!options)
232 goto out;
235 * strsep changes the string, duplicate it because parse_options
236 * gets called twice
238 options = kstrdup(options, GFP_NOFS);
239 if (!options)
240 return -ENOMEM;
242 orig = options;
244 while ((p = strsep(&options, ",")) != NULL) {
245 int token;
246 if (!*p)
247 continue;
249 token = match_token(p, tokens, args);
250 switch (token) {
251 case Opt_degraded:
252 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
253 btrfs_set_opt(info->mount_opt, DEGRADED);
254 break;
255 case Opt_subvol:
256 case Opt_subvolid:
257 case Opt_subvolrootid:
258 case Opt_device:
260 * These are parsed by btrfs_parse_early_options
261 * and can be happily ignored here.
263 break;
264 case Opt_nodatasum:
265 printk(KERN_INFO "btrfs: setting nodatasum\n");
266 btrfs_set_opt(info->mount_opt, NODATASUM);
267 break;
268 case Opt_nodatacow:
269 printk(KERN_INFO "btrfs: setting nodatacow\n");
270 btrfs_set_opt(info->mount_opt, NODATACOW);
271 btrfs_set_opt(info->mount_opt, NODATASUM);
272 break;
273 case Opt_compress_force:
274 case Opt_compress_force_type:
275 compress_force = true;
276 case Opt_compress:
277 case Opt_compress_type:
278 if (token == Opt_compress ||
279 token == Opt_compress_force ||
280 strcmp(args[0].from, "zlib") == 0) {
281 compress_type = "zlib";
282 info->compress_type = BTRFS_COMPRESS_ZLIB;
283 } else if (strcmp(args[0].from, "lzo") == 0) {
284 compress_type = "lzo";
285 info->compress_type = BTRFS_COMPRESS_LZO;
286 } else {
287 ret = -EINVAL;
288 goto out;
291 btrfs_set_opt(info->mount_opt, COMPRESS);
292 if (compress_force) {
293 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
294 pr_info("btrfs: force %s compression\n",
295 compress_type);
296 } else
297 pr_info("btrfs: use %s compression\n",
298 compress_type);
299 break;
300 case Opt_ssd:
301 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
302 btrfs_set_opt(info->mount_opt, SSD);
303 break;
304 case Opt_ssd_spread:
305 printk(KERN_INFO "btrfs: use spread ssd "
306 "allocation scheme\n");
307 btrfs_set_opt(info->mount_opt, SSD);
308 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
309 break;
310 case Opt_nossd:
311 printk(KERN_INFO "btrfs: not using ssd allocation "
312 "scheme\n");
313 btrfs_set_opt(info->mount_opt, NOSSD);
314 btrfs_clear_opt(info->mount_opt, SSD);
315 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
316 break;
317 case Opt_nobarrier:
318 printk(KERN_INFO "btrfs: turning off barriers\n");
319 btrfs_set_opt(info->mount_opt, NOBARRIER);
320 break;
321 case Opt_thread_pool:
322 intarg = 0;
323 match_int(&args[0], &intarg);
324 if (intarg) {
325 info->thread_pool_size = intarg;
326 printk(KERN_INFO "btrfs: thread pool %d\n",
327 info->thread_pool_size);
329 break;
330 case Opt_max_inline:
331 num = match_strdup(&args[0]);
332 if (num) {
333 info->max_inline = memparse(num, NULL);
334 kfree(num);
336 if (info->max_inline) {
337 info->max_inline = max_t(u64,
338 info->max_inline,
339 root->sectorsize);
341 printk(KERN_INFO "btrfs: max_inline at %llu\n",
342 (unsigned long long)info->max_inline);
344 break;
345 case Opt_alloc_start:
346 num = match_strdup(&args[0]);
347 if (num) {
348 info->alloc_start = memparse(num, NULL);
349 kfree(num);
350 printk(KERN_INFO
351 "btrfs: allocations start at %llu\n",
352 (unsigned long long)info->alloc_start);
354 break;
355 case Opt_noacl:
356 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
357 break;
358 case Opt_notreelog:
359 printk(KERN_INFO "btrfs: disabling tree log\n");
360 btrfs_set_opt(info->mount_opt, NOTREELOG);
361 break;
362 case Opt_flushoncommit:
363 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
364 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
365 break;
366 case Opt_ratio:
367 intarg = 0;
368 match_int(&args[0], &intarg);
369 if (intarg) {
370 info->metadata_ratio = intarg;
371 printk(KERN_INFO "btrfs: metadata ratio %d\n",
372 info->metadata_ratio);
374 break;
375 case Opt_discard:
376 btrfs_set_opt(info->mount_opt, DISCARD);
377 break;
378 case Opt_space_cache:
379 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
380 break;
381 case Opt_no_space_cache:
382 printk(KERN_INFO "btrfs: disabling disk space caching\n");
383 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
384 break;
385 case Opt_inode_cache:
386 printk(KERN_INFO "btrfs: enabling inode map caching\n");
387 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
388 break;
389 case Opt_clear_cache:
390 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
391 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
392 break;
393 case Opt_user_subvol_rm_allowed:
394 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
395 break;
396 case Opt_enospc_debug:
397 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
398 break;
399 case Opt_defrag:
400 printk(KERN_INFO "btrfs: enabling auto defrag");
401 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
402 break;
403 case Opt_recovery:
404 printk(KERN_INFO "btrfs: enabling auto recovery");
405 btrfs_set_opt(info->mount_opt, RECOVERY);
406 break;
407 case Opt_skip_balance:
408 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
409 break;
410 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
411 case Opt_check_integrity_including_extent_data:
412 printk(KERN_INFO "btrfs: enabling check integrity"
413 " including extent data\n");
414 btrfs_set_opt(info->mount_opt,
415 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
416 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
417 break;
418 case Opt_check_integrity:
419 printk(KERN_INFO "btrfs: enabling check integrity\n");
420 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
421 break;
422 case Opt_check_integrity_print_mask:
423 intarg = 0;
424 match_int(&args[0], &intarg);
425 if (intarg) {
426 info->check_integrity_print_mask = intarg;
427 printk(KERN_INFO "btrfs:"
428 " check_integrity_print_mask 0x%x\n",
429 info->check_integrity_print_mask);
431 break;
432 #else
433 case Opt_check_integrity_including_extent_data:
434 case Opt_check_integrity:
435 case Opt_check_integrity_print_mask:
436 printk(KERN_ERR "btrfs: support for check_integrity*"
437 " not compiled in!\n");
438 ret = -EINVAL;
439 goto out;
440 #endif
441 case Opt_err:
442 printk(KERN_INFO "btrfs: unrecognized mount option "
443 "'%s'\n", p);
444 ret = -EINVAL;
445 goto out;
446 default:
447 break;
450 out:
451 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
452 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
453 kfree(orig);
454 return ret;
458 * Parse mount options that are required early in the mount process.
460 * All other options will be parsed on much later in the mount process and
461 * only when we need to allocate a new super block.
463 static int btrfs_parse_early_options(const char *options, fmode_t flags,
464 void *holder, char **subvol_name, u64 *subvol_objectid,
465 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
467 substring_t args[MAX_OPT_ARGS];
468 char *device_name, *opts, *orig, *p;
469 int error = 0;
470 int intarg;
472 if (!options)
473 return 0;
476 * strsep changes the string, duplicate it because parse_options
477 * gets called twice
479 opts = kstrdup(options, GFP_KERNEL);
480 if (!opts)
481 return -ENOMEM;
482 orig = opts;
484 while ((p = strsep(&opts, ",")) != NULL) {
485 int token;
486 if (!*p)
487 continue;
489 token = match_token(p, tokens, args);
490 switch (token) {
491 case Opt_subvol:
492 kfree(*subvol_name);
493 *subvol_name = match_strdup(&args[0]);
494 break;
495 case Opt_subvolid:
496 intarg = 0;
497 error = match_int(&args[0], &intarg);
498 if (!error) {
499 /* we want the original fs_tree */
500 if (!intarg)
501 *subvol_objectid =
502 BTRFS_FS_TREE_OBJECTID;
503 else
504 *subvol_objectid = intarg;
506 break;
507 case Opt_subvolrootid:
508 intarg = 0;
509 error = match_int(&args[0], &intarg);
510 if (!error) {
511 /* we want the original fs_tree */
512 if (!intarg)
513 *subvol_rootid =
514 BTRFS_FS_TREE_OBJECTID;
515 else
516 *subvol_rootid = intarg;
518 break;
519 case Opt_device:
520 device_name = match_strdup(&args[0]);
521 if (!device_name) {
522 error = -ENOMEM;
523 goto out;
525 error = btrfs_scan_one_device(device_name,
526 flags, holder, fs_devices);
527 kfree(device_name);
528 if (error)
529 goto out;
530 break;
531 default:
532 break;
536 out:
537 kfree(orig);
538 return error;
541 static struct dentry *get_default_root(struct super_block *sb,
542 u64 subvol_objectid)
544 struct btrfs_root *root = sb->s_fs_info;
545 struct btrfs_root *new_root;
546 struct btrfs_dir_item *di;
547 struct btrfs_path *path;
548 struct btrfs_key location;
549 struct inode *inode;
550 u64 dir_id;
551 int new = 0;
554 * We have a specific subvol we want to mount, just setup location and
555 * go look up the root.
557 if (subvol_objectid) {
558 location.objectid = subvol_objectid;
559 location.type = BTRFS_ROOT_ITEM_KEY;
560 location.offset = (u64)-1;
561 goto find_root;
564 path = btrfs_alloc_path();
565 if (!path)
566 return ERR_PTR(-ENOMEM);
567 path->leave_spinning = 1;
570 * Find the "default" dir item which points to the root item that we
571 * will mount by default if we haven't been given a specific subvolume
572 * to mount.
574 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
575 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
576 if (IS_ERR(di)) {
577 btrfs_free_path(path);
578 return ERR_CAST(di);
580 if (!di) {
582 * Ok the default dir item isn't there. This is weird since
583 * it's always been there, but don't freak out, just try and
584 * mount to root most subvolume.
586 btrfs_free_path(path);
587 dir_id = BTRFS_FIRST_FREE_OBJECTID;
588 new_root = root->fs_info->fs_root;
589 goto setup_root;
592 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
593 btrfs_free_path(path);
595 find_root:
596 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
597 if (IS_ERR(new_root))
598 return ERR_CAST(new_root);
600 if (btrfs_root_refs(&new_root->root_item) == 0)
601 return ERR_PTR(-ENOENT);
603 dir_id = btrfs_root_dirid(&new_root->root_item);
604 setup_root:
605 location.objectid = dir_id;
606 location.type = BTRFS_INODE_ITEM_KEY;
607 location.offset = 0;
609 inode = btrfs_iget(sb, &location, new_root, &new);
610 if (IS_ERR(inode))
611 return ERR_CAST(inode);
614 * If we're just mounting the root most subvol put the inode and return
615 * a reference to the dentry. We will have already gotten a reference
616 * to the inode in btrfs_fill_super so we're good to go.
618 if (!new && sb->s_root->d_inode == inode) {
619 iput(inode);
620 return dget(sb->s_root);
623 return d_obtain_alias(inode);
626 static int btrfs_fill_super(struct super_block *sb,
627 struct btrfs_fs_devices *fs_devices,
628 void *data, int silent)
630 struct inode *inode;
631 struct dentry *root_dentry;
632 struct btrfs_root *tree_root;
633 struct btrfs_key key;
634 int err;
636 sb->s_maxbytes = MAX_LFS_FILESIZE;
637 sb->s_magic = BTRFS_SUPER_MAGIC;
638 sb->s_op = &btrfs_super_ops;
639 sb->s_d_op = &btrfs_dentry_operations;
640 sb->s_export_op = &btrfs_export_ops;
641 sb->s_xattr = btrfs_xattr_handlers;
642 sb->s_time_gran = 1;
643 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
644 sb->s_flags |= MS_POSIXACL;
645 #endif
647 tree_root = open_ctree(sb, fs_devices, (char *)data);
649 if (IS_ERR(tree_root)) {
650 printk("btrfs: open_ctree failed\n");
651 return PTR_ERR(tree_root);
653 sb->s_fs_info = tree_root;
655 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
656 key.type = BTRFS_INODE_ITEM_KEY;
657 key.offset = 0;
658 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
659 if (IS_ERR(inode)) {
660 err = PTR_ERR(inode);
661 goto fail_close;
664 root_dentry = d_alloc_root(inode);
665 if (!root_dentry) {
666 iput(inode);
667 err = -ENOMEM;
668 goto fail_close;
671 sb->s_root = root_dentry;
673 save_mount_options(sb, data);
674 cleancache_init_fs(sb);
675 return 0;
677 fail_close:
678 close_ctree(tree_root);
679 return err;
682 int btrfs_sync_fs(struct super_block *sb, int wait)
684 struct btrfs_trans_handle *trans;
685 struct btrfs_root *root = btrfs_sb(sb);
686 int ret;
688 trace_btrfs_sync_fs(wait);
690 if (!wait) {
691 filemap_flush(root->fs_info->btree_inode->i_mapping);
692 return 0;
695 btrfs_start_delalloc_inodes(root, 0);
696 btrfs_wait_ordered_extents(root, 0, 0);
698 trans = btrfs_start_transaction(root, 0);
699 if (IS_ERR(trans))
700 return PTR_ERR(trans);
701 ret = btrfs_commit_transaction(trans, root);
702 return ret;
705 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
707 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
708 struct btrfs_fs_info *info = root->fs_info;
709 char *compress_type;
711 if (btrfs_test_opt(root, DEGRADED))
712 seq_puts(seq, ",degraded");
713 if (btrfs_test_opt(root, NODATASUM))
714 seq_puts(seq, ",nodatasum");
715 if (btrfs_test_opt(root, NODATACOW))
716 seq_puts(seq, ",nodatacow");
717 if (btrfs_test_opt(root, NOBARRIER))
718 seq_puts(seq, ",nobarrier");
719 if (info->max_inline != 8192 * 1024)
720 seq_printf(seq, ",max_inline=%llu",
721 (unsigned long long)info->max_inline);
722 if (info->alloc_start != 0)
723 seq_printf(seq, ",alloc_start=%llu",
724 (unsigned long long)info->alloc_start);
725 if (info->thread_pool_size != min_t(unsigned long,
726 num_online_cpus() + 2, 8))
727 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
728 if (btrfs_test_opt(root, COMPRESS)) {
729 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
730 compress_type = "zlib";
731 else
732 compress_type = "lzo";
733 if (btrfs_test_opt(root, FORCE_COMPRESS))
734 seq_printf(seq, ",compress-force=%s", compress_type);
735 else
736 seq_printf(seq, ",compress=%s", compress_type);
738 if (btrfs_test_opt(root, NOSSD))
739 seq_puts(seq, ",nossd");
740 if (btrfs_test_opt(root, SSD_SPREAD))
741 seq_puts(seq, ",ssd_spread");
742 else if (btrfs_test_opt(root, SSD))
743 seq_puts(seq, ",ssd");
744 if (btrfs_test_opt(root, NOTREELOG))
745 seq_puts(seq, ",notreelog");
746 if (btrfs_test_opt(root, FLUSHONCOMMIT))
747 seq_puts(seq, ",flushoncommit");
748 if (btrfs_test_opt(root, DISCARD))
749 seq_puts(seq, ",discard");
750 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
751 seq_puts(seq, ",noacl");
752 if (btrfs_test_opt(root, SPACE_CACHE))
753 seq_puts(seq, ",space_cache");
754 else
755 seq_puts(seq, ",nospace_cache");
756 if (btrfs_test_opt(root, CLEAR_CACHE))
757 seq_puts(seq, ",clear_cache");
758 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
759 seq_puts(seq, ",user_subvol_rm_allowed");
760 if (btrfs_test_opt(root, ENOSPC_DEBUG))
761 seq_puts(seq, ",enospc_debug");
762 if (btrfs_test_opt(root, AUTO_DEFRAG))
763 seq_puts(seq, ",autodefrag");
764 if (btrfs_test_opt(root, INODE_MAP_CACHE))
765 seq_puts(seq, ",inode_cache");
766 if (btrfs_test_opt(root, SKIP_BALANCE))
767 seq_puts(seq, ",skip_balance");
768 return 0;
771 static int btrfs_test_super(struct super_block *s, void *data)
773 struct btrfs_root *test_root = data;
774 struct btrfs_root *root = btrfs_sb(s);
777 * If this super block is going away, return false as it
778 * can't match as an existing super block.
780 if (!atomic_read(&s->s_active))
781 return 0;
782 return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
785 static int btrfs_set_super(struct super_block *s, void *data)
787 s->s_fs_info = data;
789 return set_anon_super(s, data);
793 * subvolumes are identified by ino 256
795 static inline int is_subvolume_inode(struct inode *inode)
797 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
798 return 1;
799 return 0;
803 * This will strip out the subvol=%s argument for an argument string and add
804 * subvolid=0 to make sure we get the actual tree root for path walking to the
805 * subvol we want.
807 static char *setup_root_args(char *args)
809 unsigned copied = 0;
810 unsigned len = strlen(args) + 2;
811 char *pos;
812 char *ret;
815 * We need the same args as before, but minus
817 * subvol=a
819 * and add
821 * subvolid=0
823 * which is a difference of 2 characters, so we allocate strlen(args) +
824 * 2 characters.
826 ret = kzalloc(len * sizeof(char), GFP_NOFS);
827 if (!ret)
828 return NULL;
829 pos = strstr(args, "subvol=");
831 /* This shouldn't happen, but just in case.. */
832 if (!pos) {
833 kfree(ret);
834 return NULL;
838 * The subvol=<> arg is not at the front of the string, copy everybody
839 * up to that into ret.
841 if (pos != args) {
842 *pos = '\0';
843 strcpy(ret, args);
844 copied += strlen(args);
845 pos++;
848 strncpy(ret + copied, "subvolid=0", len - copied);
850 /* Length of subvolid=0 */
851 copied += 10;
854 * If there is no , after the subvol= option then we know there's no
855 * other options and we can just return.
857 pos = strchr(pos, ',');
858 if (!pos)
859 return ret;
861 /* Copy the rest of the arguments into our buffer */
862 strncpy(ret + copied, pos, len - copied);
863 copied += strlen(pos);
865 return ret;
868 static struct dentry *mount_subvol(const char *subvol_name, int flags,
869 const char *device_name, char *data)
871 struct dentry *root;
872 struct vfsmount *mnt;
873 char *newargs;
875 newargs = setup_root_args(data);
876 if (!newargs)
877 return ERR_PTR(-ENOMEM);
878 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
879 newargs);
880 kfree(newargs);
881 if (IS_ERR(mnt))
882 return ERR_CAST(mnt);
884 root = mount_subtree(mnt, subvol_name);
886 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
887 struct super_block *s = root->d_sb;
888 dput(root);
889 root = ERR_PTR(-EINVAL);
890 deactivate_locked_super(s);
891 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
892 subvol_name);
895 return root;
899 * Find a superblock for the given device / mount point.
901 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
902 * for multiple device setup. Make sure to keep it in sync.
904 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
905 const char *device_name, void *data)
907 struct block_device *bdev = NULL;
908 struct super_block *s;
909 struct dentry *root;
910 struct btrfs_fs_devices *fs_devices = NULL;
911 struct btrfs_fs_info *fs_info = NULL;
912 fmode_t mode = FMODE_READ;
913 char *subvol_name = NULL;
914 u64 subvol_objectid = 0;
915 u64 subvol_rootid = 0;
916 int error = 0;
918 if (!(flags & MS_RDONLY))
919 mode |= FMODE_WRITE;
921 error = btrfs_parse_early_options(data, mode, fs_type,
922 &subvol_name, &subvol_objectid,
923 &subvol_rootid, &fs_devices);
924 if (error) {
925 kfree(subvol_name);
926 return ERR_PTR(error);
929 if (subvol_name) {
930 root = mount_subvol(subvol_name, flags, device_name, data);
931 kfree(subvol_name);
932 return root;
935 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
936 if (error)
937 return ERR_PTR(error);
940 * Setup a dummy root and fs_info for test/set super. This is because
941 * we don't actually fill this stuff out until open_ctree, but we need
942 * it for searching for existing supers, so this lets us do that and
943 * then open_ctree will properly initialize everything later.
945 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
946 if (!fs_info)
947 return ERR_PTR(-ENOMEM);
949 fs_info->tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
950 if (!fs_info->tree_root) {
951 error = -ENOMEM;
952 goto error_fs_info;
954 fs_info->tree_root->fs_info = fs_info;
955 fs_info->fs_devices = fs_devices;
957 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
958 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
959 if (!fs_info->super_copy || !fs_info->super_for_commit) {
960 error = -ENOMEM;
961 goto error_fs_info;
964 error = btrfs_open_devices(fs_devices, mode, fs_type);
965 if (error)
966 goto error_fs_info;
968 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
969 error = -EACCES;
970 goto error_close_devices;
973 bdev = fs_devices->latest_bdev;
974 s = sget(fs_type, btrfs_test_super, btrfs_set_super,
975 fs_info->tree_root);
976 if (IS_ERR(s)) {
977 error = PTR_ERR(s);
978 goto error_close_devices;
981 if (s->s_root) {
982 if ((flags ^ s->s_flags) & MS_RDONLY) {
983 deactivate_locked_super(s);
984 error = -EBUSY;
985 goto error_close_devices;
988 btrfs_close_devices(fs_devices);
989 free_fs_info(fs_info);
990 } else {
991 char b[BDEVNAME_SIZE];
993 s->s_flags = flags | MS_NOSEC;
994 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
995 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
996 error = btrfs_fill_super(s, fs_devices, data,
997 flags & MS_SILENT ? 1 : 0);
998 if (error) {
999 deactivate_locked_super(s);
1000 return ERR_PTR(error);
1003 s->s_flags |= MS_ACTIVE;
1006 root = get_default_root(s, subvol_objectid);
1007 if (IS_ERR(root)) {
1008 deactivate_locked_super(s);
1009 return root;
1012 return root;
1014 error_close_devices:
1015 btrfs_close_devices(fs_devices);
1016 error_fs_info:
1017 free_fs_info(fs_info);
1018 return ERR_PTR(error);
1021 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1023 struct btrfs_root *root = btrfs_sb(sb);
1024 int ret;
1026 ret = btrfs_parse_options(root, data);
1027 if (ret)
1028 return -EINVAL;
1030 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1031 return 0;
1033 if (*flags & MS_RDONLY) {
1034 sb->s_flags |= MS_RDONLY;
1036 ret = btrfs_commit_super(root);
1037 WARN_ON(ret);
1038 } else {
1039 if (root->fs_info->fs_devices->rw_devices == 0)
1040 return -EACCES;
1042 if (btrfs_super_log_root(root->fs_info->super_copy) != 0)
1043 return -EINVAL;
1045 ret = btrfs_cleanup_fs_roots(root->fs_info);
1046 WARN_ON(ret);
1048 /* recover relocation */
1049 ret = btrfs_recover_relocation(root);
1050 WARN_ON(ret);
1052 sb->s_flags &= ~MS_RDONLY;
1055 return 0;
1058 /* Used to sort the devices by max_avail(descending sort) */
1059 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1060 const void *dev_info2)
1062 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1063 ((struct btrfs_device_info *)dev_info2)->max_avail)
1064 return -1;
1065 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1066 ((struct btrfs_device_info *)dev_info2)->max_avail)
1067 return 1;
1068 else
1069 return 0;
1073 * sort the devices by max_avail, in which max free extent size of each device
1074 * is stored.(Descending Sort)
1076 static inline void btrfs_descending_sort_devices(
1077 struct btrfs_device_info *devices,
1078 size_t nr_devices)
1080 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1081 btrfs_cmp_device_free_bytes, NULL);
1085 * The helper to calc the free space on the devices that can be used to store
1086 * file data.
1088 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1090 struct btrfs_fs_info *fs_info = root->fs_info;
1091 struct btrfs_device_info *devices_info;
1092 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1093 struct btrfs_device *device;
1094 u64 skip_space;
1095 u64 type;
1096 u64 avail_space;
1097 u64 used_space;
1098 u64 min_stripe_size;
1099 int min_stripes = 1, num_stripes = 1;
1100 int i = 0, nr_devices;
1101 int ret;
1103 nr_devices = fs_info->fs_devices->open_devices;
1104 BUG_ON(!nr_devices);
1106 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1107 GFP_NOFS);
1108 if (!devices_info)
1109 return -ENOMEM;
1111 /* calc min stripe number for data space alloction */
1112 type = btrfs_get_alloc_profile(root, 1);
1113 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1114 min_stripes = 2;
1115 num_stripes = nr_devices;
1116 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1117 min_stripes = 2;
1118 num_stripes = 2;
1119 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1120 min_stripes = 4;
1121 num_stripes = 4;
1124 if (type & BTRFS_BLOCK_GROUP_DUP)
1125 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1126 else
1127 min_stripe_size = BTRFS_STRIPE_LEN;
1129 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1130 if (!device->in_fs_metadata || !device->bdev)
1131 continue;
1133 avail_space = device->total_bytes - device->bytes_used;
1135 /* align with stripe_len */
1136 do_div(avail_space, BTRFS_STRIPE_LEN);
1137 avail_space *= BTRFS_STRIPE_LEN;
1140 * In order to avoid overwritting the superblock on the drive,
1141 * btrfs starts at an offset of at least 1MB when doing chunk
1142 * allocation.
1144 skip_space = 1024 * 1024;
1146 /* user can set the offset in fs_info->alloc_start. */
1147 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1148 device->total_bytes)
1149 skip_space = max(fs_info->alloc_start, skip_space);
1152 * btrfs can not use the free space in [0, skip_space - 1],
1153 * we must subtract it from the total. In order to implement
1154 * it, we account the used space in this range first.
1156 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1157 &used_space);
1158 if (ret) {
1159 kfree(devices_info);
1160 return ret;
1163 /* calc the free space in [0, skip_space - 1] */
1164 skip_space -= used_space;
1167 * we can use the free space in [0, skip_space - 1], subtract
1168 * it from the total.
1170 if (avail_space && avail_space >= skip_space)
1171 avail_space -= skip_space;
1172 else
1173 avail_space = 0;
1175 if (avail_space < min_stripe_size)
1176 continue;
1178 devices_info[i].dev = device;
1179 devices_info[i].max_avail = avail_space;
1181 i++;
1184 nr_devices = i;
1186 btrfs_descending_sort_devices(devices_info, nr_devices);
1188 i = nr_devices - 1;
1189 avail_space = 0;
1190 while (nr_devices >= min_stripes) {
1191 if (num_stripes > nr_devices)
1192 num_stripes = nr_devices;
1194 if (devices_info[i].max_avail >= min_stripe_size) {
1195 int j;
1196 u64 alloc_size;
1198 avail_space += devices_info[i].max_avail * num_stripes;
1199 alloc_size = devices_info[i].max_avail;
1200 for (j = i + 1 - num_stripes; j <= i; j++)
1201 devices_info[j].max_avail -= alloc_size;
1203 i--;
1204 nr_devices--;
1207 kfree(devices_info);
1208 *free_bytes = avail_space;
1209 return 0;
1212 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1214 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
1215 struct btrfs_super_block *disk_super = root->fs_info->super_copy;
1216 struct list_head *head = &root->fs_info->space_info;
1217 struct btrfs_space_info *found;
1218 u64 total_used = 0;
1219 u64 total_free_data = 0;
1220 int bits = dentry->d_sb->s_blocksize_bits;
1221 __be32 *fsid = (__be32 *)root->fs_info->fsid;
1222 int ret;
1224 /* holding chunk_muext to avoid allocating new chunks */
1225 mutex_lock(&root->fs_info->chunk_mutex);
1226 rcu_read_lock();
1227 list_for_each_entry_rcu(found, head, list) {
1228 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1229 total_free_data += found->disk_total - found->disk_used;
1230 total_free_data -=
1231 btrfs_account_ro_block_groups_free_space(found);
1234 total_used += found->disk_used;
1236 rcu_read_unlock();
1238 buf->f_namelen = BTRFS_NAME_LEN;
1239 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1240 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1241 buf->f_bsize = dentry->d_sb->s_blocksize;
1242 buf->f_type = BTRFS_SUPER_MAGIC;
1243 buf->f_bavail = total_free_data;
1244 ret = btrfs_calc_avail_data_space(root, &total_free_data);
1245 if (ret) {
1246 mutex_unlock(&root->fs_info->chunk_mutex);
1247 return ret;
1249 buf->f_bavail += total_free_data;
1250 buf->f_bavail = buf->f_bavail >> bits;
1251 mutex_unlock(&root->fs_info->chunk_mutex);
1253 /* We treat it as constant endianness (it doesn't matter _which_)
1254 because we want the fsid to come out the same whether mounted
1255 on a big-endian or little-endian host */
1256 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1257 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1258 /* Mask in the root object ID too, to disambiguate subvols */
1259 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1260 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1262 return 0;
1265 static struct file_system_type btrfs_fs_type = {
1266 .owner = THIS_MODULE,
1267 .name = "btrfs",
1268 .mount = btrfs_mount,
1269 .kill_sb = kill_anon_super,
1270 .fs_flags = FS_REQUIRES_DEV,
1274 * used by btrfsctl to scan devices when no FS is mounted
1276 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1277 unsigned long arg)
1279 struct btrfs_ioctl_vol_args *vol;
1280 struct btrfs_fs_devices *fs_devices;
1281 int ret = -ENOTTY;
1283 if (!capable(CAP_SYS_ADMIN))
1284 return -EPERM;
1286 vol = memdup_user((void __user *)arg, sizeof(*vol));
1287 if (IS_ERR(vol))
1288 return PTR_ERR(vol);
1290 switch (cmd) {
1291 case BTRFS_IOC_SCAN_DEV:
1292 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1293 &btrfs_fs_type, &fs_devices);
1294 break;
1297 kfree(vol);
1298 return ret;
1301 static int btrfs_freeze(struct super_block *sb)
1303 struct btrfs_root *root = btrfs_sb(sb);
1304 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1305 mutex_lock(&root->fs_info->cleaner_mutex);
1306 return 0;
1309 static int btrfs_unfreeze(struct super_block *sb)
1311 struct btrfs_root *root = btrfs_sb(sb);
1312 mutex_unlock(&root->fs_info->cleaner_mutex);
1313 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1314 return 0;
1317 static void btrfs_fs_dirty_inode(struct inode *inode, int flags)
1319 int ret;
1321 ret = btrfs_dirty_inode(inode);
1322 if (ret)
1323 printk_ratelimited(KERN_ERR "btrfs: fail to dirty inode %Lu "
1324 "error %d\n", btrfs_ino(inode), ret);
1327 static const struct super_operations btrfs_super_ops = {
1328 .drop_inode = btrfs_drop_inode,
1329 .evict_inode = btrfs_evict_inode,
1330 .put_super = btrfs_put_super,
1331 .sync_fs = btrfs_sync_fs,
1332 .show_options = btrfs_show_options,
1333 .write_inode = btrfs_write_inode,
1334 .dirty_inode = btrfs_fs_dirty_inode,
1335 .alloc_inode = btrfs_alloc_inode,
1336 .destroy_inode = btrfs_destroy_inode,
1337 .statfs = btrfs_statfs,
1338 .remount_fs = btrfs_remount,
1339 .freeze_fs = btrfs_freeze,
1340 .unfreeze_fs = btrfs_unfreeze,
1343 static const struct file_operations btrfs_ctl_fops = {
1344 .unlocked_ioctl = btrfs_control_ioctl,
1345 .compat_ioctl = btrfs_control_ioctl,
1346 .owner = THIS_MODULE,
1347 .llseek = noop_llseek,
1350 static struct miscdevice btrfs_misc = {
1351 .minor = BTRFS_MINOR,
1352 .name = "btrfs-control",
1353 .fops = &btrfs_ctl_fops
1356 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1357 MODULE_ALIAS("devname:btrfs-control");
1359 static int btrfs_interface_init(void)
1361 return misc_register(&btrfs_misc);
1364 static void btrfs_interface_exit(void)
1366 if (misc_deregister(&btrfs_misc) < 0)
1367 printk(KERN_INFO "misc_deregister failed for control device");
1370 static int __init init_btrfs_fs(void)
1372 int err;
1374 err = btrfs_init_sysfs();
1375 if (err)
1376 return err;
1378 err = btrfs_init_compress();
1379 if (err)
1380 goto free_sysfs;
1382 err = btrfs_init_cachep();
1383 if (err)
1384 goto free_compress;
1386 err = extent_io_init();
1387 if (err)
1388 goto free_cachep;
1390 err = extent_map_init();
1391 if (err)
1392 goto free_extent_io;
1394 err = btrfs_delayed_inode_init();
1395 if (err)
1396 goto free_extent_map;
1398 err = btrfs_interface_init();
1399 if (err)
1400 goto free_delayed_inode;
1402 err = register_filesystem(&btrfs_fs_type);
1403 if (err)
1404 goto unregister_ioctl;
1406 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1407 return 0;
1409 unregister_ioctl:
1410 btrfs_interface_exit();
1411 free_delayed_inode:
1412 btrfs_delayed_inode_exit();
1413 free_extent_map:
1414 extent_map_exit();
1415 free_extent_io:
1416 extent_io_exit();
1417 free_cachep:
1418 btrfs_destroy_cachep();
1419 free_compress:
1420 btrfs_exit_compress();
1421 free_sysfs:
1422 btrfs_exit_sysfs();
1423 return err;
1426 static void __exit exit_btrfs_fs(void)
1428 btrfs_destroy_cachep();
1429 btrfs_delayed_inode_exit();
1430 extent_map_exit();
1431 extent_io_exit();
1432 btrfs_interface_exit();
1433 unregister_filesystem(&btrfs_fs_type);
1434 btrfs_exit_sysfs();
1435 btrfs_cleanup_fs_uuids();
1436 btrfs_exit_compress();
1439 module_init(init_btrfs_fs)
1440 module_exit(exit_btrfs_fs)
1442 MODULE_LICENSE("GPL");