?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'linux-next' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad...
[zen-stable.git] / fs / btrfs / super.c
blobe28ad4baf483af6b4c7e5d8450dc4a73909d0247
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
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.
7 *
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.
12 *
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.
17 */
18
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/mnt_namespace.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"
57
58 #define CREATE_TRACE_POINTS
59 #include <trace/events/btrfs.h>
60
61 static const struct super_operations btrfs_super_ops;
62 static struct file_system_type btrfs_fs_type;
63
64 static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
65 char nbuf[16])
66 {
67 char *errstr = NULL;
68
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;
83 }
84 break;
85 }
86
87 return errstr;
88 }
89
90 static void __save_error_info(struct btrfs_fs_info *fs_info)
91 {
92 /*
93 * today we only save the error info into ram. Long term we'll
94 * also send it down to the disk
95 */
96 fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
97 }
98
99 /* NOTE:
100 * We move write_super stuff at umount in order to avoid deadlock
101 * for umount hold all lock.
102 */
103 static void save_error_info(struct btrfs_fs_info *fs_info)
104 {
105 __save_error_info(fs_info);
106 }
107
108 /* btrfs handle error by forcing the filesystem readonly */
109 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
110 {
111 struct super_block *sb = fs_info->sb;
112
113 if (sb->s_flags & MS_RDONLY)
114 return;
115
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");
119 }
120 }
121
122 /*
123 * __btrfs_std_error decodes expected errors from the caller and
124 * invokes the approciate error response.
125 */
126 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
127 unsigned int line, int errno)
128 {
129 struct super_block *sb = fs_info->sb;
130 char nbuf[16];
131 const char *errstr;
132
133 /*
134 * Special case: if the error is EROFS, and we're already
135 * under MS_RDONLY, then it is safe here.
136 */
137 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
138 return;
139
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);
144
145 btrfs_handle_error(fs_info);
146 }
147
148 static void btrfs_put_super(struct super_block *sb)
149 {
150 struct btrfs_root *root = btrfs_sb(sb);
151 int ret;
152
153 ret = close_ctree(root);
154 sb->s_fs_info = NULL;
155
156 (void)ret; /* FIXME: need to fix VFS to return error? */
157 }
158
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,
167 Opt_inode_cache, Opt_no_space_cache, Opt_recovery, Opt_err,
168 };
169
170 static match_table_t tokens = {
171 {Opt_degraded, "degraded"},
172 {Opt_subvol, "subvol=%s"},
173 {Opt_subvolid, "subvolid=%d"},
174 {Opt_device, "device=%s"},
175 {Opt_nodatasum, "nodatasum"},
176 {Opt_nodatacow, "nodatacow"},
177 {Opt_nobarrier, "nobarrier"},
178 {Opt_max_inline, "max_inline=%s"},
179 {Opt_alloc_start, "alloc_start=%s"},
180 {Opt_thread_pool, "thread_pool=%d"},
181 {Opt_compress, "compress"},
182 {Opt_compress_type, "compress=%s"},
183 {Opt_compress_force, "compress-force"},
184 {Opt_compress_force_type, "compress-force=%s"},
185 {Opt_ssd, "ssd"},
186 {Opt_ssd_spread, "ssd_spread"},
187 {Opt_nossd, "nossd"},
188 {Opt_noacl, "noacl"},
189 {Opt_notreelog, "notreelog"},
190 {Opt_flushoncommit, "flushoncommit"},
191 {Opt_ratio, "metadata_ratio=%d"},
192 {Opt_discard, "discard"},
193 {Opt_space_cache, "space_cache"},
194 {Opt_clear_cache, "clear_cache"},
195 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
196 {Opt_enospc_debug, "enospc_debug"},
197 {Opt_subvolrootid, "subvolrootid=%d"},
198 {Opt_defrag, "autodefrag"},
199 {Opt_inode_cache, "inode_cache"},
200 {Opt_no_space_cache, "nospace_cache"},
201 {Opt_recovery, "recovery"},
202 {Opt_err, NULL},
203 };
204
205 /*
206 * Regular mount options parser. Everything that is needed only when
207 * reading in a new superblock is parsed here.
208 */
209 int btrfs_parse_options(struct btrfs_root *root, char *options)
210 {
211 struct btrfs_fs_info *info = root->fs_info;
212 substring_t args[MAX_OPT_ARGS];
213 char *p, *num, *orig = NULL;
214 u64 cache_gen;
215 int intarg;
216 int ret = 0;
217 char *compress_type;
218 bool compress_force = false;
219
220 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
221 if (cache_gen)
222 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
223
224 if (!options)
225 goto out;
226
227 /*
228 * strsep changes the string, duplicate it because parse_options
229 * gets called twice
230 */
231 options = kstrdup(options, GFP_NOFS);
232 if (!options)
233 return -ENOMEM;
234
235 orig = options;
236
237 while ((p = strsep(&options, ",")) != NULL) {
238 int token;
239 if (!*p)
240 continue;
241
242 token = match_token(p, tokens, args);
243 switch (token) {
244 case Opt_degraded:
245 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
246 btrfs_set_opt(info->mount_opt, DEGRADED);
247 break;
248 case Opt_subvol:
249 case Opt_subvolid:
250 case Opt_subvolrootid:
251 case Opt_device:
252 /*
253 * These are parsed by btrfs_parse_early_options
254 * and can be happily ignored here.
255 */
256 break;
257 case Opt_nodatasum:
258 printk(KERN_INFO "btrfs: setting nodatasum\n");
259 btrfs_set_opt(info->mount_opt, NODATASUM);
260 break;
261 case Opt_nodatacow:
262 printk(KERN_INFO "btrfs: setting nodatacow\n");
263 btrfs_set_opt(info->mount_opt, NODATACOW);
264 btrfs_set_opt(info->mount_opt, NODATASUM);
265 break;
266 case Opt_compress_force:
267 case Opt_compress_force_type:
268 compress_force = true;
269 case Opt_compress:
270 case Opt_compress_type:
271 if (token == Opt_compress ||
272 token == Opt_compress_force ||
273 strcmp(args[0].from, "zlib") == 0) {
274 compress_type = "zlib";
275 info->compress_type = BTRFS_COMPRESS_ZLIB;
276 } else if (strcmp(args[0].from, "lzo") == 0) {
277 compress_type = "lzo";
278 info->compress_type = BTRFS_COMPRESS_LZO;
279 } else {
280 ret = -EINVAL;
281 goto out;
282 }
283
284 btrfs_set_opt(info->mount_opt, COMPRESS);
285 if (compress_force) {
286 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
287 pr_info("btrfs: force %s compression\n",
288 compress_type);
289 } else
290 pr_info("btrfs: use %s compression\n",
291 compress_type);
292 break;
293 case Opt_ssd:
294 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
295 btrfs_set_opt(info->mount_opt, SSD);
296 break;
297 case Opt_ssd_spread:
298 printk(KERN_INFO "btrfs: use spread ssd "
299 "allocation scheme\n");
300 btrfs_set_opt(info->mount_opt, SSD);
301 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
302 break;
303 case Opt_nossd:
304 printk(KERN_INFO "btrfs: not using ssd allocation "
305 "scheme\n");
306 btrfs_set_opt(info->mount_opt, NOSSD);
307 btrfs_clear_opt(info->mount_opt, SSD);
308 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
309 break;
310 case Opt_nobarrier:
311 printk(KERN_INFO "btrfs: turning off barriers\n");
312 btrfs_set_opt(info->mount_opt, NOBARRIER);
313 break;
314 case Opt_thread_pool:
315 intarg = 0;
316 match_int(&args[0], &intarg);
317 if (intarg) {
318 info->thread_pool_size = intarg;
319 printk(KERN_INFO "btrfs: thread pool %d\n",
320 info->thread_pool_size);
321 }
322 break;
323 case Opt_max_inline:
324 num = match_strdup(&args[0]);
325 if (num) {
326 info->max_inline = memparse(num, NULL);
327 kfree(num);
328
329 if (info->max_inline) {
330 info->max_inline = max_t(u64,
331 info->max_inline,
332 root->sectorsize);
333 }
334 printk(KERN_INFO "btrfs: max_inline at %llu\n",
335 (unsigned long long)info->max_inline);
336 }
337 break;
338 case Opt_alloc_start:
339 num = match_strdup(&args[0]);
340 if (num) {
341 info->alloc_start = memparse(num, NULL);
342 kfree(num);
343 printk(KERN_INFO
344 "btrfs: allocations start at %llu\n",
345 (unsigned long long)info->alloc_start);
346 }
347 break;
348 case Opt_noacl:
349 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
350 break;
351 case Opt_notreelog:
352 printk(KERN_INFO "btrfs: disabling tree log\n");
353 btrfs_set_opt(info->mount_opt, NOTREELOG);
354 break;
355 case Opt_flushoncommit:
356 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
357 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
358 break;
359 case Opt_ratio:
360 intarg = 0;
361 match_int(&args[0], &intarg);
362 if (intarg) {
363 info->metadata_ratio = intarg;
364 printk(KERN_INFO "btrfs: metadata ratio %d\n",
365 info->metadata_ratio);
366 }
367 break;
368 case Opt_discard:
369 btrfs_set_opt(info->mount_opt, DISCARD);
370 break;
371 case Opt_space_cache:
372 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
373 break;
374 case Opt_no_space_cache:
375 printk(KERN_INFO "btrfs: disabling disk space caching\n");
376 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
377 break;
378 case Opt_inode_cache:
379 printk(KERN_INFO "btrfs: enabling inode map caching\n");
380 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
381 break;
382 case Opt_clear_cache:
383 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
384 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
385 break;
386 case Opt_user_subvol_rm_allowed:
387 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
388 break;
389 case Opt_enospc_debug:
390 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
391 break;
392 case Opt_defrag:
393 printk(KERN_INFO "btrfs: enabling auto defrag");
394 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
395 break;
396 case Opt_recovery:
397 printk(KERN_INFO "btrfs: enabling auto recovery");
398 btrfs_set_opt(info->mount_opt, RECOVERY);
399 break;
400 case Opt_err:
401 printk(KERN_INFO "btrfs: unrecognized mount option "
402 "'%s'\n", p);
403 ret = -EINVAL;
404 goto out;
405 default:
406 break;
407 }
408 }
409 out:
410 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
411 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
412 kfree(orig);
413 return ret;
414 }
415
416 /*
417 * Parse mount options that are required early in the mount process.
418 *
419 * All other options will be parsed on much later in the mount process and
420 * only when we need to allocate a new super block.
421 */
422 static int btrfs_parse_early_options(const char *options, fmode_t flags,
423 void *holder, char **subvol_name, u64 *subvol_objectid,
424 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
425 {
426 substring_t args[MAX_OPT_ARGS];
427 char *device_name, *opts, *orig, *p;
428 int error = 0;
429 int intarg;
430
431 if (!options)
432 return 0;
433
434 /*
435 * strsep changes the string, duplicate it because parse_options
436 * gets called twice
437 */
438 opts = kstrdup(options, GFP_KERNEL);
439 if (!opts)
440 return -ENOMEM;
441 orig = opts;
442
443 while ((p = strsep(&opts, ",")) != NULL) {
444 int token;
445 if (!*p)
446 continue;
447
448 token = match_token(p, tokens, args);
449 switch (token) {
450 case Opt_subvol:
451 kfree(*subvol_name);
452 *subvol_name = match_strdup(&args[0]);
453 break;
454 case Opt_subvolid:
455 intarg = 0;
456 error = match_int(&args[0], &intarg);
457 if (!error) {
458 /* we want the original fs_tree */
459 if (!intarg)
460 *subvol_objectid =
461 BTRFS_FS_TREE_OBJECTID;
462 else
463 *subvol_objectid = intarg;
464 }
465 break;
466 case Opt_subvolrootid:
467 intarg = 0;
468 error = match_int(&args[0], &intarg);
469 if (!error) {
470 /* we want the original fs_tree */
471 if (!intarg)
472 *subvol_rootid =
473 BTRFS_FS_TREE_OBJECTID;
474 else
475 *subvol_rootid = intarg;
476 }
477 break;
478 case Opt_device:
479 device_name = match_strdup(&args[0]);
480 if (!device_name) {
481 error = -ENOMEM;
482 goto out;
483 }
484 error = btrfs_scan_one_device(device_name,
485 flags, holder, fs_devices);
486 kfree(device_name);
487 if (error)
488 goto out;
489 break;
490 default:
491 break;
492 }
493 }
494
495 out:
496 kfree(orig);
497 return error;
498 }
499
500 static struct dentry *get_default_root(struct super_block *sb,
501 u64 subvol_objectid)
502 {
503 struct btrfs_root *root = sb->s_fs_info;
504 struct btrfs_root *new_root;
505 struct btrfs_dir_item *di;
506 struct btrfs_path *path;
507 struct btrfs_key location;
508 struct inode *inode;
509 u64 dir_id;
510 int new = 0;
511
512 /*
513 * We have a specific subvol we want to mount, just setup location and
514 * go look up the root.
515 */
516 if (subvol_objectid) {
517 location.objectid = subvol_objectid;
518 location.type = BTRFS_ROOT_ITEM_KEY;
519 location.offset = (u64)-1;
520 goto find_root;
521 }
522
523 path = btrfs_alloc_path();
524 if (!path)
525 return ERR_PTR(-ENOMEM);
526 path->leave_spinning = 1;
527
528 /*
529 * Find the "default" dir item which points to the root item that we
530 * will mount by default if we haven't been given a specific subvolume
531 * to mount.
532 */
533 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
534 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
535 if (IS_ERR(di)) {
536 btrfs_free_path(path);
537 return ERR_CAST(di);
538 }
539 if (!di) {
540 /*
541 * Ok the default dir item isn't there. This is weird since
542 * it's always been there, but don't freak out, just try and
543 * mount to root most subvolume.
544 */
545 btrfs_free_path(path);
546 dir_id = BTRFS_FIRST_FREE_OBJECTID;
547 new_root = root->fs_info->fs_root;
548 goto setup_root;
549 }
550
551 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
552 btrfs_free_path(path);
553
554 find_root:
555 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
556 if (IS_ERR(new_root))
557 return ERR_CAST(new_root);
558
559 if (btrfs_root_refs(&new_root->root_item) == 0)
560 return ERR_PTR(-ENOENT);
561
562 dir_id = btrfs_root_dirid(&new_root->root_item);
563 setup_root:
564 location.objectid = dir_id;
565 location.type = BTRFS_INODE_ITEM_KEY;
566 location.offset = 0;
567
568 inode = btrfs_iget(sb, &location, new_root, &new);
569 if (IS_ERR(inode))
570 return ERR_CAST(inode);
571
572 /*
573 * If we're just mounting the root most subvol put the inode and return
574 * a reference to the dentry. We will have already gotten a reference
575 * to the inode in btrfs_fill_super so we're good to go.
576 */
577 if (!new && sb->s_root->d_inode == inode) {
578 iput(inode);
579 return dget(sb->s_root);
580 }
581
582 return d_obtain_alias(inode);
583 }
584
585 static int btrfs_fill_super(struct super_block *sb,
586 struct btrfs_fs_devices *fs_devices,
587 void *data, int silent)
588 {
589 struct inode *inode;
590 struct dentry *root_dentry;
591 struct btrfs_root *tree_root;
592 struct btrfs_key key;
593 int err;
594
595 sb->s_maxbytes = MAX_LFS_FILESIZE;
596 sb->s_magic = BTRFS_SUPER_MAGIC;
597 sb->s_op = &btrfs_super_ops;
598 sb->s_d_op = &btrfs_dentry_operations;
599 sb->s_export_op = &btrfs_export_ops;
600 sb->s_xattr = btrfs_xattr_handlers;
601 sb->s_time_gran = 1;
602 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
603 sb->s_flags |= MS_POSIXACL;
604 #endif
605
606 tree_root = open_ctree(sb, fs_devices, (char *)data);
607
608 if (IS_ERR(tree_root)) {
609 printk("btrfs: open_ctree failed\n");
610 return PTR_ERR(tree_root);
611 }
612 sb->s_fs_info = tree_root;
613
614 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
615 key.type = BTRFS_INODE_ITEM_KEY;
616 key.offset = 0;
617 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
618 if (IS_ERR(inode)) {
619 err = PTR_ERR(inode);
620 goto fail_close;
621 }
622
623 root_dentry = d_alloc_root(inode);
624 if (!root_dentry) {
625 iput(inode);
626 err = -ENOMEM;
627 goto fail_close;
628 }
629
630 sb->s_root = root_dentry;
631
632 save_mount_options(sb, data);
633 cleancache_init_fs(sb);
634 return 0;
635
636 fail_close:
637 close_ctree(tree_root);
638 return err;
639 }
640
641 int btrfs_sync_fs(struct super_block *sb, int wait)
642 {
643 struct btrfs_trans_handle *trans;
644 struct btrfs_root *root = btrfs_sb(sb);
645 int ret;
646
647 trace_btrfs_sync_fs(wait);
648
649 if (!wait) {
650 filemap_flush(root->fs_info->btree_inode->i_mapping);
651 return 0;
652 }
653
654 btrfs_start_delalloc_inodes(root, 0);
655 btrfs_wait_ordered_extents(root, 0, 0);
656
657 trans = btrfs_start_transaction(root, 0);
658 if (IS_ERR(trans))
659 return PTR_ERR(trans);
660 ret = btrfs_commit_transaction(trans, root);
661 return ret;
662 }
663
664 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
665 {
666 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
667 struct btrfs_fs_info *info = root->fs_info;
668 char *compress_type;
669
670 if (btrfs_test_opt(root, DEGRADED))
671 seq_puts(seq, ",degraded");
672 if (btrfs_test_opt(root, NODATASUM))
673 seq_puts(seq, ",nodatasum");
674 if (btrfs_test_opt(root, NODATACOW))
675 seq_puts(seq, ",nodatacow");
676 if (btrfs_test_opt(root, NOBARRIER))
677 seq_puts(seq, ",nobarrier");
678 if (info->max_inline != 8192 * 1024)
679 seq_printf(seq, ",max_inline=%llu",
680 (unsigned long long)info->max_inline);
681 if (info->alloc_start != 0)
682 seq_printf(seq, ",alloc_start=%llu",
683 (unsigned long long)info->alloc_start);
684 if (info->thread_pool_size != min_t(unsigned long,
685 num_online_cpus() + 2, 8))
686 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
687 if (btrfs_test_opt(root, COMPRESS)) {
688 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
689 compress_type = "zlib";
690 else
691 compress_type = "lzo";
692 if (btrfs_test_opt(root, FORCE_COMPRESS))
693 seq_printf(seq, ",compress-force=%s", compress_type);
694 else
695 seq_printf(seq, ",compress=%s", compress_type);
696 }
697 if (btrfs_test_opt(root, NOSSD))
698 seq_puts(seq, ",nossd");
699 if (btrfs_test_opt(root, SSD_SPREAD))
700 seq_puts(seq, ",ssd_spread");
701 else if (btrfs_test_opt(root, SSD))
702 seq_puts(seq, ",ssd");
703 if (btrfs_test_opt(root, NOTREELOG))
704 seq_puts(seq, ",notreelog");
705 if (btrfs_test_opt(root, FLUSHONCOMMIT))
706 seq_puts(seq, ",flushoncommit");
707 if (btrfs_test_opt(root, DISCARD))
708 seq_puts(seq, ",discard");
709 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
710 seq_puts(seq, ",noacl");
711 if (btrfs_test_opt(root, SPACE_CACHE))
712 seq_puts(seq, ",space_cache");
713 else
714 seq_puts(seq, ",nospace_cache");
715 if (btrfs_test_opt(root, CLEAR_CACHE))
716 seq_puts(seq, ",clear_cache");
717 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
718 seq_puts(seq, ",user_subvol_rm_allowed");
719 if (btrfs_test_opt(root, ENOSPC_DEBUG))
720 seq_puts(seq, ",enospc_debug");
721 if (btrfs_test_opt(root, AUTO_DEFRAG))
722 seq_puts(seq, ",autodefrag");
723 if (btrfs_test_opt(root, INODE_MAP_CACHE))
724 seq_puts(seq, ",inode_cache");
725 return 0;
726 }
727
728 static int btrfs_test_super(struct super_block *s, void *data)
729 {
730 struct btrfs_root *test_root = data;
731 struct btrfs_root *root = btrfs_sb(s);
732
733 /*
734 * If this super block is going away, return false as it
735 * can't match as an existing super block.
736 */
737 if (!atomic_read(&s->s_active))
738 return 0;
739 return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
740 }
741
742 static int btrfs_set_super(struct super_block *s, void *data)
743 {
744 s->s_fs_info = data;
745
746 return set_anon_super(s, data);
747 }
748
749 /*
750 * subvolumes are identified by ino 256
751 */
752 static inline int is_subvolume_inode(struct inode *inode)
753 {
754 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
755 return 1;
756 return 0;
757 }
758
759 /*
760 * This will strip out the subvol=%s argument for an argument string and add
761 * subvolid=0 to make sure we get the actual tree root for path walking to the
762 * subvol we want.
763 */
764 static char *setup_root_args(char *args)
765 {
766 unsigned copied = 0;
767 unsigned len = strlen(args) + 2;
768 char *pos;
769 char *ret;
770
771 /*
772 * We need the same args as before, but minus
773 *
774 * subvol=a
775 *
776 * and add
777 *
778 * subvolid=0
779 *
780 * which is a difference of 2 characters, so we allocate strlen(args) +
781 * 2 characters.
782 */
783 ret = kzalloc(len * sizeof(char), GFP_NOFS);
784 if (!ret)
785 return NULL;
786 pos = strstr(args, "subvol=");
787
788 /* This shouldn't happen, but just in case.. */
789 if (!pos) {
790 kfree(ret);
791 return NULL;
792 }
793
794 /*
795 * The subvol=<> arg is not at the front of the string, copy everybody
796 * up to that into ret.
797 */
798 if (pos != args) {
799 *pos = '\0';
800 strcpy(ret, args);
801 copied += strlen(args);
802 pos++;
803 }
804
805 strncpy(ret + copied, "subvolid=0", len - copied);
806
807 /* Length of subvolid=0 */
808 copied += 10;
809
810 /*
811 * If there is no , after the subvol= option then we know there's no
812 * other options and we can just return.
813 */
814 pos = strchr(pos, ',');
815 if (!pos)
816 return ret;
817
818 /* Copy the rest of the arguments into our buffer */
819 strncpy(ret + copied, pos, len - copied);
820 copied += strlen(pos);
821
822 return ret;
823 }
824
825 static struct dentry *mount_subvol(const char *subvol_name, int flags,
826 const char *device_name, char *data)
827 {
828 struct dentry *root;
829 struct vfsmount *mnt;
830 char *newargs;
831
832 newargs = setup_root_args(data);
833 if (!newargs)
834 return ERR_PTR(-ENOMEM);
835 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
836 newargs);
837 kfree(newargs);
838 if (IS_ERR(mnt))
839 return ERR_CAST(mnt);
840
841 root = mount_subtree(mnt, subvol_name);
842
843 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
844 struct super_block *s = root->d_sb;
845 dput(root);
846 root = ERR_PTR(-EINVAL);
847 deactivate_locked_super(s);
848 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
849 subvol_name);
850 }
851
852 return root;
853 }
854
855 /*
856 * Find a superblock for the given device / mount point.
857 *
858 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
859 * for multiple device setup. Make sure to keep it in sync.
860 */
861 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
862 const char *device_name, void *data)
863 {
864 struct block_device *bdev = NULL;
865 struct super_block *s;
866 struct dentry *root;
867 struct btrfs_fs_devices *fs_devices = NULL;
868 struct btrfs_fs_info *fs_info = NULL;
869 fmode_t mode = FMODE_READ;
870 char *subvol_name = NULL;
871 u64 subvol_objectid = 0;
872 u64 subvol_rootid = 0;
873 int error = 0;
874
875 if (!(flags & MS_RDONLY))
876 mode |= FMODE_WRITE;
877
878 error = btrfs_parse_early_options(data, mode, fs_type,
879 &subvol_name, &subvol_objectid,
880 &subvol_rootid, &fs_devices);
881 if (error) {
882 kfree(subvol_name);
883 return ERR_PTR(error);
884 }
885
886 if (subvol_name) {
887 root = mount_subvol(subvol_name, flags, device_name, data);
888 kfree(subvol_name);
889 return root;
890 }
891
892 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
893 if (error)
894 return ERR_PTR(error);
895
896 /*
897 * Setup a dummy root and fs_info for test/set super. This is because
898 * we don't actually fill this stuff out until open_ctree, but we need
899 * it for searching for existing supers, so this lets us do that and
900 * then open_ctree will properly initialize everything later.
901 */
902 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
903 if (!fs_info)
904 return ERR_PTR(-ENOMEM);
905
906 fs_info->tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
907 if (!fs_info->tree_root) {
908 error = -ENOMEM;
909 goto error_fs_info;
910 }
911 fs_info->tree_root->fs_info = fs_info;
912 fs_info->fs_devices = fs_devices;
913
914 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
915 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
916 if (!fs_info->super_copy || !fs_info->super_for_commit) {
917 error = -ENOMEM;
918 goto error_fs_info;
919 }
920
921 error = btrfs_open_devices(fs_devices, mode, fs_type);
922 if (error)
923 goto error_fs_info;
924
925 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
926 error = -EACCES;
927 goto error_close_devices;
928 }
929
930 bdev = fs_devices->latest_bdev;
931 s = sget(fs_type, btrfs_test_super, btrfs_set_super,
932 fs_info->tree_root);
933 if (IS_ERR(s)) {
934 error = PTR_ERR(s);
935 goto error_close_devices;
936 }
937
938 if (s->s_root) {
939 if ((flags ^ s->s_flags) & MS_RDONLY) {
940 deactivate_locked_super(s);
941 error = -EBUSY;
942 goto error_close_devices;
943 }
944
945 btrfs_close_devices(fs_devices);
946 free_fs_info(fs_info);
947 } else {
948 char b[BDEVNAME_SIZE];
949
950 s->s_flags = flags | MS_NOSEC;
951 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
952 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
953 error = btrfs_fill_super(s, fs_devices, data,
954 flags & MS_SILENT ? 1 : 0);
955 if (error) {
956 deactivate_locked_super(s);
957 return ERR_PTR(error);
958 }
959
960 s->s_flags |= MS_ACTIVE;
961 }
962
963 root = get_default_root(s, subvol_objectid);
964 if (IS_ERR(root)) {
965 deactivate_locked_super(s);
966 return root;
967 }
968
969 return root;
970
971 error_close_devices:
972 btrfs_close_devices(fs_devices);
973 error_fs_info:
974 free_fs_info(fs_info);
975 return ERR_PTR(error);
976 }
977
978 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
979 {
980 struct btrfs_root *root = btrfs_sb(sb);
981 int ret;
982
983 ret = btrfs_parse_options(root, data);
984 if (ret)
985 return -EINVAL;
986
987 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
988 return 0;
989
990 if (*flags & MS_RDONLY) {
991 sb->s_flags |= MS_RDONLY;
992
993 ret = btrfs_commit_super(root);
994 WARN_ON(ret);
995 } else {
996 if (root->fs_info->fs_devices->rw_devices == 0)
997 return -EACCES;
998
999 if (btrfs_super_log_root(root->fs_info->super_copy) != 0)
1000 return -EINVAL;
1001
1002 ret = btrfs_cleanup_fs_roots(root->fs_info);
1003 WARN_ON(ret);
1004
1005 /* recover relocation */
1006 ret = btrfs_recover_relocation(root);
1007 WARN_ON(ret);
1008
1009 sb->s_flags &= ~MS_RDONLY;
1010 }
1011
1012 return 0;
1013 }
1014
1015 /* Used to sort the devices by max_avail(descending sort) */
1016 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1017 const void *dev_info2)
1018 {
1019 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1020 ((struct btrfs_device_info *)dev_info2)->max_avail)
1021 return -1;
1022 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1023 ((struct btrfs_device_info *)dev_info2)->max_avail)
1024 return 1;
1025 else
1026 return 0;
1027 }
1028
1029 /*
1030 * sort the devices by max_avail, in which max free extent size of each device
1031 * is stored.(Descending Sort)
1032 */
1033 static inline void btrfs_descending_sort_devices(
1034 struct btrfs_device_info *devices,
1035 size_t nr_devices)
1036 {
1037 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1038 btrfs_cmp_device_free_bytes, NULL);
1039 }
1040
1041 /*
1042 * The helper to calc the free space on the devices that can be used to store
1043 * file data.
1044 */
1045 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1046 {
1047 struct btrfs_fs_info *fs_info = root->fs_info;
1048 struct btrfs_device_info *devices_info;
1049 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1050 struct btrfs_device *device;
1051 u64 skip_space;
1052 u64 type;
1053 u64 avail_space;
1054 u64 used_space;
1055 u64 min_stripe_size;
1056 int min_stripes = 1;
1057 int i = 0, nr_devices;
1058 int ret;
1059
1060 nr_devices = fs_info->fs_devices->open_devices;
1061 BUG_ON(!nr_devices);
1062
1063 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1064 GFP_NOFS);
1065 if (!devices_info)
1066 return -ENOMEM;
1067
1068 /* calc min stripe number for data space alloction */
1069 type = btrfs_get_alloc_profile(root, 1);
1070 if (type & BTRFS_BLOCK_GROUP_RAID0)
1071 min_stripes = 2;
1072 else if (type & BTRFS_BLOCK_GROUP_RAID1)
1073 min_stripes = 2;
1074 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1075 min_stripes = 4;
1076
1077 if (type & BTRFS_BLOCK_GROUP_DUP)
1078 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1079 else
1080 min_stripe_size = BTRFS_STRIPE_LEN;
1081
1082 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1083 if (!device->in_fs_metadata || !device->bdev)
1084 continue;
1085
1086 avail_space = device->total_bytes - device->bytes_used;
1087
1088 /* align with stripe_len */
1089 do_div(avail_space, BTRFS_STRIPE_LEN);
1090 avail_space *= BTRFS_STRIPE_LEN;
1091
1092 /*
1093 * In order to avoid overwritting the superblock on the drive,
1094 * btrfs starts at an offset of at least 1MB when doing chunk
1095 * allocation.
1096 */
1097 skip_space = 1024 * 1024;
1098
1099 /* user can set the offset in fs_info->alloc_start. */
1100 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1101 device->total_bytes)
1102 skip_space = max(fs_info->alloc_start, skip_space);
1103
1104 /*
1105 * btrfs can not use the free space in [0, skip_space - 1],
1106 * we must subtract it from the total. In order to implement
1107 * it, we account the used space in this range first.
1108 */
1109 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1110 &used_space);
1111 if (ret) {
1112 kfree(devices_info);
1113 return ret;
1114 }
1115
1116 /* calc the free space in [0, skip_space - 1] */
1117 skip_space -= used_space;
1118
1119 /*
1120 * we can use the free space in [0, skip_space - 1], subtract
1121 * it from the total.
1122 */
1123 if (avail_space && avail_space >= skip_space)
1124 avail_space -= skip_space;
1125 else
1126 avail_space = 0;
1127
1128 if (avail_space < min_stripe_size)
1129 continue;
1130
1131 devices_info[i].dev = device;
1132 devices_info[i].max_avail = avail_space;
1133
1134 i++;
1135 }
1136
1137 nr_devices = i;
1138
1139 btrfs_descending_sort_devices(devices_info, nr_devices);
1140
1141 i = nr_devices - 1;
1142 avail_space = 0;
1143 while (nr_devices >= min_stripes) {
1144 if (devices_info[i].max_avail >= min_stripe_size) {
1145 int j;
1146 u64 alloc_size;
1147
1148 avail_space += devices_info[i].max_avail * min_stripes;
1149 alloc_size = devices_info[i].max_avail;
1150 for (j = i + 1 - min_stripes; j <= i; j++)
1151 devices_info[j].max_avail -= alloc_size;
1152 }
1153 i--;
1154 nr_devices--;
1155 }
1156
1157 kfree(devices_info);
1158 *free_bytes = avail_space;
1159 return 0;
1160 }
1161
1162 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1163 {
1164 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
1165 struct btrfs_super_block *disk_super = root->fs_info->super_copy;
1166 struct list_head *head = &root->fs_info->space_info;
1167 struct btrfs_space_info *found;
1168 u64 total_used = 0;
1169 u64 total_free_data = 0;
1170 int bits = dentry->d_sb->s_blocksize_bits;
1171 __be32 *fsid = (__be32 *)root->fs_info->fsid;
1172 int ret;
1173
1174 /* holding chunk_muext to avoid allocating new chunks */
1175 mutex_lock(&root->fs_info->chunk_mutex);
1176 rcu_read_lock();
1177 list_for_each_entry_rcu(found, head, list) {
1178 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1179 total_free_data += found->disk_total - found->disk_used;
1180 total_free_data -=
1181 btrfs_account_ro_block_groups_free_space(found);
1182 }
1183
1184 total_used += found->disk_used;
1185 }
1186 rcu_read_unlock();
1187
1188 buf->f_namelen = BTRFS_NAME_LEN;
1189 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1190 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1191 buf->f_bsize = dentry->d_sb->s_blocksize;
1192 buf->f_type = BTRFS_SUPER_MAGIC;
1193 buf->f_bavail = total_free_data;
1194 ret = btrfs_calc_avail_data_space(root, &total_free_data);
1195 if (ret) {
1196 mutex_unlock(&root->fs_info->chunk_mutex);
1197 return ret;
1198 }
1199 buf->f_bavail += total_free_data;
1200 buf->f_bavail = buf->f_bavail >> bits;
1201 mutex_unlock(&root->fs_info->chunk_mutex);
1202
1203 /* We treat it as constant endianness (it doesn't matter _which_)
1204 because we want the fsid to come out the same whether mounted
1205 on a big-endian or little-endian host */
1206 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1207 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1208 /* Mask in the root object ID too, to disambiguate subvols */
1209 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1210 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1211
1212 return 0;
1213 }
1214
1215 static struct file_system_type btrfs_fs_type = {
1216 .owner = THIS_MODULE,
1217 .name = "btrfs",
1218 .mount = btrfs_mount,
1219 .kill_sb = kill_anon_super,
1220 .fs_flags = FS_REQUIRES_DEV,
1221 };
1222
1223 /*
1224 * used by btrfsctl to scan devices when no FS is mounted
1225 */
1226 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1227 unsigned long arg)
1228 {
1229 struct btrfs_ioctl_vol_args *vol;
1230 struct btrfs_fs_devices *fs_devices;
1231 int ret = -ENOTTY;
1232
1233 if (!capable(CAP_SYS_ADMIN))
1234 return -EPERM;
1235
1236 vol = memdup_user((void __user *)arg, sizeof(*vol));
1237 if (IS_ERR(vol))
1238 return PTR_ERR(vol);
1239
1240 switch (cmd) {
1241 case BTRFS_IOC_SCAN_DEV:
1242 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1243 &btrfs_fs_type, &fs_devices);
1244 break;
1245 }
1246
1247 kfree(vol);
1248 return ret;
1249 }
1250
1251 static int btrfs_freeze(struct super_block *sb)
1252 {
1253 struct btrfs_root *root = btrfs_sb(sb);
1254 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1255 mutex_lock(&root->fs_info->cleaner_mutex);
1256 return 0;
1257 }
1258
1259 static int btrfs_unfreeze(struct super_block *sb)
1260 {
1261 struct btrfs_root *root = btrfs_sb(sb);
1262 mutex_unlock(&root->fs_info->cleaner_mutex);
1263 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1264 return 0;
1265 }
1266
1267 static const struct super_operations btrfs_super_ops = {
1268 .drop_inode = btrfs_drop_inode,
1269 .evict_inode = btrfs_evict_inode,
1270 .put_super = btrfs_put_super,
1271 .sync_fs = btrfs_sync_fs,
1272 .show_options = btrfs_show_options,
1273 .write_inode = btrfs_write_inode,
1274 .dirty_inode = btrfs_dirty_inode,
1275 .alloc_inode = btrfs_alloc_inode,
1276 .destroy_inode = btrfs_destroy_inode,
1277 .statfs = btrfs_statfs,
1278 .remount_fs = btrfs_remount,
1279 .freeze_fs = btrfs_freeze,
1280 .unfreeze_fs = btrfs_unfreeze,
1281 };
1282
1283 static const struct file_operations btrfs_ctl_fops = {
1284 .unlocked_ioctl = btrfs_control_ioctl,
1285 .compat_ioctl = btrfs_control_ioctl,
1286 .owner = THIS_MODULE,
1287 .llseek = noop_llseek,
1288 };
1289
1290 static struct miscdevice btrfs_misc = {
1291 .minor = BTRFS_MINOR,
1292 .name = "btrfs-control",
1293 .fops = &btrfs_ctl_fops
1294 };
1295
1296 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1297 MODULE_ALIAS("devname:btrfs-control");
1298
1299 static int btrfs_interface_init(void)
1300 {
1301 return misc_register(&btrfs_misc);
1302 }
1303
1304 static void btrfs_interface_exit(void)
1305 {
1306 if (misc_deregister(&btrfs_misc) < 0)
1307 printk(KERN_INFO "misc_deregister failed for control device");
1308 }
1309
1310 static int __init init_btrfs_fs(void)
1311 {
1312 int err;
1313
1314 err = btrfs_init_sysfs();
1315 if (err)
1316 return err;
1317
1318 err = btrfs_init_compress();
1319 if (err)
1320 goto free_sysfs;
1321
1322 err = btrfs_init_cachep();
1323 if (err)
1324 goto free_compress;
1325
1326 err = extent_io_init();
1327 if (err)
1328 goto free_cachep;
1329
1330 err = extent_map_init();
1331 if (err)
1332 goto free_extent_io;
1333
1334 err = btrfs_delayed_inode_init();
1335 if (err)
1336 goto free_extent_map;
1337
1338 err = btrfs_interface_init();
1339 if (err)
1340 goto free_delayed_inode;
1341
1342 err = register_filesystem(&btrfs_fs_type);
1343 if (err)
1344 goto unregister_ioctl;
1345
1346 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1347 return 0;
1348
1349 unregister_ioctl:
1350 btrfs_interface_exit();
1351 free_delayed_inode:
1352 btrfs_delayed_inode_exit();
1353 free_extent_map:
1354 extent_map_exit();
1355 free_extent_io:
1356 extent_io_exit();
1357 free_cachep:
1358 btrfs_destroy_cachep();
1359 free_compress:
1360 btrfs_exit_compress();
1361 free_sysfs:
1362 btrfs_exit_sysfs();
1363 return err;
1364 }
1365
1366 static void __exit exit_btrfs_fs(void)
1367 {
1368 btrfs_destroy_cachep();
1369 btrfs_delayed_inode_exit();
1370 extent_map_exit();
1371 extent_io_exit();
1372 btrfs_interface_exit();
1373 unregister_filesystem(&btrfs_fs_type);
1374 btrfs_exit_sysfs();
1375 btrfs_cleanup_fs_uuids();
1376 btrfs_exit_compress();
1377 }
1378
1379 module_init(init_btrfs_fs)
1380 module_exit(exit_btrfs_fs)
1381
1382 MODULE_LICENSE("GPL");