?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
btrfs: fix mixed block count of available space
[linux-2.6/btrfs-unstable.git] / fs / btrfs / super.c
blobbdca79ce45f1bb1d5097323a6c0136cf3dc58a45
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/ratelimit.h>
44 #include <linux/btrfs.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 "print-tree.h"
51 #include "hash.h"
52 #include "props.h"
53 #include "xattr.h"
54 #include "volumes.h"
55 #include "export.h"
56 #include "compression.h"
57 #include "rcu-string.h"
58 #include "dev-replace.h"
59 #include "free-space-cache.h"
60 #include "backref.h"
61 #include "tests/btrfs-tests.h"
62
63 #include "qgroup.h"
64 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h>
66
67 static const struct super_operations btrfs_super_ops;
68 static struct file_system_type btrfs_fs_type;
69
70 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
71
72 const char *btrfs_decode_error(int errno)
73 {
74 char *errstr = "unknown";
75
76 switch (errno) {
77 case -EIO:
78 errstr = "IO failure";
79 break;
80 case -ENOMEM:
81 errstr = "Out of memory";
82 break;
83 case -EROFS:
84 errstr = "Readonly filesystem";
85 break;
86 case -EEXIST:
87 errstr = "Object already exists";
88 break;
89 case -ENOSPC:
90 errstr = "No space left";
91 break;
92 case -ENOENT:
93 errstr = "No such entry";
94 break;
95 }
96
97 return errstr;
98 }
99
100 static void save_error_info(struct btrfs_fs_info *fs_info)
101 {
102 /*
103 * today we only save the error info into ram. Long term we'll
104 * also send it down to the disk
105 */
106 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
107 }
108
109 /* btrfs handle error by forcing the filesystem readonly */
110 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
111 {
112 struct super_block *sb = fs_info->sb;
113
114 if (sb->s_flags & MS_RDONLY)
115 return;
116
117 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
118 sb->s_flags |= MS_RDONLY;
119 btrfs_info(fs_info, "forced readonly");
120 /*
121 * Note that a running device replace operation is not
122 * canceled here although there is no way to update
123 * the progress. It would add the risk of a deadlock,
124 * therefore the canceling is ommited. The only penalty
125 * is that some I/O remains active until the procedure
126 * completes. The next time when the filesystem is
127 * mounted writeable again, the device replace
128 * operation continues.
129 */
130 }
131 }
132
133 /*
134 * __btrfs_std_error decodes expected errors from the caller and
135 * invokes the approciate error response.
136 */
137 __cold
138 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
139 unsigned int line, int errno, const char *fmt, ...)
140 {
141 struct super_block *sb = fs_info->sb;
142 #ifdef CONFIG_PRINTK
143 const char *errstr;
144 #endif
145
146 /*
147 * Special case: if the error is EROFS, and we're already
148 * under MS_RDONLY, then it is safe here.
149 */
150 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
151 return;
152
153 #ifdef CONFIG_PRINTK
154 errstr = btrfs_decode_error(errno);
155 if (fmt) {
156 struct va_format vaf;
157 va_list args;
158
159 va_start(args, fmt);
160 vaf.fmt = fmt;
161 vaf.va = &args;
162
163 printk(KERN_CRIT
164 "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
165 sb->s_id, function, line, errno, errstr, &vaf);
166 va_end(args);
167 } else {
168 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
169 sb->s_id, function, line, errno, errstr);
170 }
171 #endif
172
173 /* Don't go through full error handling during mount */
174 save_error_info(fs_info);
175 if (sb->s_flags & MS_BORN)
176 btrfs_handle_error(fs_info);
177 }
178
179 #ifdef CONFIG_PRINTK
180 static const char * const logtypes[] = {
181 "emergency",
182 "alert",
183 "critical",
184 "error",
185 "warning",
186 "notice",
187 "info",
188 "debug",
189 };
190
191 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
192 {
193 struct super_block *sb = fs_info->sb;
194 char lvl[4];
195 struct va_format vaf;
196 va_list args;
197 const char *type = logtypes[4];
198 int kern_level;
199
200 va_start(args, fmt);
201
202 kern_level = printk_get_level(fmt);
203 if (kern_level) {
204 size_t size = printk_skip_level(fmt) - fmt;
205 memcpy(lvl, fmt, size);
206 lvl[size] = '\0';
207 fmt += size;
208 type = logtypes[kern_level - '0'];
209 } else
210 *lvl = '\0';
211
212 vaf.fmt = fmt;
213 vaf.va = &args;
214
215 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
216
217 va_end(args);
218 }
219 #endif
220
221 /*
222 * We only mark the transaction aborted and then set the file system read-only.
223 * This will prevent new transactions from starting or trying to join this
224 * one.
225 *
226 * This means that error recovery at the call site is limited to freeing
227 * any local memory allocations and passing the error code up without
228 * further cleanup. The transaction should complete as it normally would
229 * in the call path but will return -EIO.
230 *
231 * We'll complete the cleanup in btrfs_end_transaction and
232 * btrfs_commit_transaction.
233 */
234 __cold
235 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
236 struct btrfs_root *root, const char *function,
237 unsigned int line, int errno)
238 {
239 trans->aborted = errno;
240 /* Nothing used. The other threads that have joined this
241 * transaction may be able to continue. */
242 if (!trans->blocks_used && list_empty(&trans->new_bgs)) {
243 const char *errstr;
244
245 errstr = btrfs_decode_error(errno);
246 btrfs_warn(root->fs_info,
247 "%s:%d: Aborting unused transaction(%s).",
248 function, line, errstr);
249 return;
250 }
251 ACCESS_ONCE(trans->transaction->aborted) = errno;
252 /* Wake up anybody who may be waiting on this transaction */
253 wake_up(&root->fs_info->transaction_wait);
254 wake_up(&root->fs_info->transaction_blocked_wait);
255 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
256 }
257 /*
258 * __btrfs_panic decodes unexpected, fatal errors from the caller,
259 * issues an alert, and either panics or BUGs, depending on mount options.
260 */
261 __cold
262 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
263 unsigned int line, int errno, const char *fmt, ...)
264 {
265 char *s_id = "<unknown>";
266 const char *errstr;
267 struct va_format vaf = { .fmt = fmt };
268 va_list args;
269
270 if (fs_info)
271 s_id = fs_info->sb->s_id;
272
273 va_start(args, fmt);
274 vaf.va = &args;
275
276 errstr = btrfs_decode_error(errno);
277 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
278 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
279 s_id, function, line, &vaf, errno, errstr);
280
281 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
282 function, line, &vaf, errno, errstr);
283 va_end(args);
284 /* Caller calls BUG() */
285 }
286
287 static void btrfs_put_super(struct super_block *sb)
288 {
289 close_ctree(btrfs_sb(sb)->tree_root);
290 }
291
292 enum {
293 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
294 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
295 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
296 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
297 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
298 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
299 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
300 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
301 Opt_skip_balance, Opt_check_integrity,
302 Opt_check_integrity_including_extent_data,
303 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
304 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
305 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
306 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
307 Opt_nologreplay, Opt_norecovery,
308 #ifdef CONFIG_BTRFS_DEBUG
309 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
310 #endif
311 Opt_err,
312 };
313
314 static const match_table_t tokens = {
315 {Opt_degraded, "degraded"},
316 {Opt_subvol, "subvol=%s"},
317 {Opt_subvolid, "subvolid=%s"},
318 {Opt_device, "device=%s"},
319 {Opt_nodatasum, "nodatasum"},
320 {Opt_datasum, "datasum"},
321 {Opt_nodatacow, "nodatacow"},
322 {Opt_datacow, "datacow"},
323 {Opt_nobarrier, "nobarrier"},
324 {Opt_barrier, "barrier"},
325 {Opt_max_inline, "max_inline=%s"},
326 {Opt_alloc_start, "alloc_start=%s"},
327 {Opt_thread_pool, "thread_pool=%d"},
328 {Opt_compress, "compress"},
329 {Opt_compress_type, "compress=%s"},
330 {Opt_compress_force, "compress-force"},
331 {Opt_compress_force_type, "compress-force=%s"},
332 {Opt_ssd, "ssd"},
333 {Opt_ssd_spread, "ssd_spread"},
334 {Opt_nossd, "nossd"},
335 {Opt_acl, "acl"},
336 {Opt_noacl, "noacl"},
337 {Opt_notreelog, "notreelog"},
338 {Opt_treelog, "treelog"},
339 {Opt_nologreplay, "nologreplay"},
340 {Opt_norecovery, "norecovery"},
341 {Opt_flushoncommit, "flushoncommit"},
342 {Opt_noflushoncommit, "noflushoncommit"},
343 {Opt_ratio, "metadata_ratio=%d"},
344 {Opt_discard, "discard"},
345 {Opt_nodiscard, "nodiscard"},
346 {Opt_space_cache, "space_cache"},
347 {Opt_space_cache_version, "space_cache=%s"},
348 {Opt_clear_cache, "clear_cache"},
349 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
350 {Opt_enospc_debug, "enospc_debug"},
351 {Opt_noenospc_debug, "noenospc_debug"},
352 {Opt_subvolrootid, "subvolrootid=%d"},
353 {Opt_defrag, "autodefrag"},
354 {Opt_nodefrag, "noautodefrag"},
355 {Opt_inode_cache, "inode_cache"},
356 {Opt_noinode_cache, "noinode_cache"},
357 {Opt_no_space_cache, "nospace_cache"},
358 {Opt_recovery, "recovery"}, /* deprecated */
359 {Opt_usebackuproot, "usebackuproot"},
360 {Opt_skip_balance, "skip_balance"},
361 {Opt_check_integrity, "check_int"},
362 {Opt_check_integrity_including_extent_data, "check_int_data"},
363 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
364 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
365 {Opt_fatal_errors, "fatal_errors=%s"},
366 {Opt_commit_interval, "commit=%d"},
367 #ifdef CONFIG_BTRFS_DEBUG
368 {Opt_fragment_data, "fragment=data"},
369 {Opt_fragment_metadata, "fragment=metadata"},
370 {Opt_fragment_all, "fragment=all"},
371 #endif
372 {Opt_err, NULL},
373 };
374
375 /*
376 * Regular mount options parser. Everything that is needed only when
377 * reading in a new superblock is parsed here.
378 * XXX JDM: This needs to be cleaned up for remount.
379 */
380 int btrfs_parse_options(struct btrfs_root *root, char *options,
381 unsigned long new_flags)
382 {
383 struct btrfs_fs_info *info = root->fs_info;
384 substring_t args[MAX_OPT_ARGS];
385 char *p, *num, *orig = NULL;
386 u64 cache_gen;
387 int intarg;
388 int ret = 0;
389 char *compress_type;
390 bool compress_force = false;
391 enum btrfs_compression_type saved_compress_type;
392 bool saved_compress_force;
393 int no_compress = 0;
394
395 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
396 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE))
397 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
398 else if (cache_gen)
399 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
400
401 /*
402 * Even the options are empty, we still need to do extra check
403 * against new flags
404 */
405 if (!options)
406 goto check;
407
408 /*
409 * strsep changes the string, duplicate it because parse_options
410 * gets called twice
411 */
412 options = kstrdup(options, GFP_NOFS);
413 if (!options)
414 return -ENOMEM;
415
416 orig = options;
417
418 while ((p = strsep(&options, ",")) != NULL) {
419 int token;
420 if (!*p)
421 continue;
422
423 token = match_token(p, tokens, args);
424 switch (token) {
425 case Opt_degraded:
426 btrfs_info(root->fs_info, "allowing degraded mounts");
427 btrfs_set_opt(info->mount_opt, DEGRADED);
428 break;
429 case Opt_subvol:
430 case Opt_subvolid:
431 case Opt_subvolrootid:
432 case Opt_device:
433 /*
434 * These are parsed by btrfs_parse_early_options
435 * and can be happily ignored here.
436 */
437 break;
438 case Opt_nodatasum:
439 btrfs_set_and_info(root, NODATASUM,
440 "setting nodatasum");
441 break;
442 case Opt_datasum:
443 if (btrfs_test_opt(root, NODATASUM)) {
444 if (btrfs_test_opt(root, NODATACOW))
445 btrfs_info(root->fs_info, "setting datasum, datacow enabled");
446 else
447 btrfs_info(root->fs_info, "setting datasum");
448 }
449 btrfs_clear_opt(info->mount_opt, NODATACOW);
450 btrfs_clear_opt(info->mount_opt, NODATASUM);
451 break;
452 case Opt_nodatacow:
453 if (!btrfs_test_opt(root, NODATACOW)) {
454 if (!btrfs_test_opt(root, COMPRESS) ||
455 !btrfs_test_opt(root, FORCE_COMPRESS)) {
456 btrfs_info(root->fs_info,
457 "setting nodatacow, compression disabled");
458 } else {
459 btrfs_info(root->fs_info, "setting nodatacow");
460 }
461 }
462 btrfs_clear_opt(info->mount_opt, COMPRESS);
463 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
464 btrfs_set_opt(info->mount_opt, NODATACOW);
465 btrfs_set_opt(info->mount_opt, NODATASUM);
466 break;
467 case Opt_datacow:
468 btrfs_clear_and_info(root, NODATACOW,
469 "setting datacow");
470 break;
471 case Opt_compress_force:
472 case Opt_compress_force_type:
473 compress_force = true;
474 /* Fallthrough */
475 case Opt_compress:
476 case Opt_compress_type:
477 saved_compress_type = btrfs_test_opt(root, COMPRESS) ?
478 info->compress_type : BTRFS_COMPRESS_NONE;
479 saved_compress_force =
480 btrfs_test_opt(root, FORCE_COMPRESS);
481 if (token == Opt_compress ||
482 token == Opt_compress_force ||
483 strcmp(args[0].from, "zlib") == 0) {
484 compress_type = "zlib";
485 info->compress_type = BTRFS_COMPRESS_ZLIB;
486 btrfs_set_opt(info->mount_opt, COMPRESS);
487 btrfs_clear_opt(info->mount_opt, NODATACOW);
488 btrfs_clear_opt(info->mount_opt, NODATASUM);
489 no_compress = 0;
490 } else if (strcmp(args[0].from, "lzo") == 0) {
491 compress_type = "lzo";
492 info->compress_type = BTRFS_COMPRESS_LZO;
493 btrfs_set_opt(info->mount_opt, COMPRESS);
494 btrfs_clear_opt(info->mount_opt, NODATACOW);
495 btrfs_clear_opt(info->mount_opt, NODATASUM);
496 btrfs_set_fs_incompat(info, COMPRESS_LZO);
497 no_compress = 0;
498 } else if (strncmp(args[0].from, "no", 2) == 0) {
499 compress_type = "no";
500 btrfs_clear_opt(info->mount_opt, COMPRESS);
501 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
502 compress_force = false;
503 no_compress++;
504 } else {
505 ret = -EINVAL;
506 goto out;
507 }
508
509 if (compress_force) {
510 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
511 } else {
512 /*
513 * If we remount from compress-force=xxx to
514 * compress=xxx, we need clear FORCE_COMPRESS
515 * flag, otherwise, there is no way for users
516 * to disable forcible compression separately.
517 */
518 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
519 }
520 if ((btrfs_test_opt(root, COMPRESS) &&
521 (info->compress_type != saved_compress_type ||
522 compress_force != saved_compress_force)) ||
523 (!btrfs_test_opt(root, COMPRESS) &&
524 no_compress == 1)) {
525 btrfs_info(root->fs_info,
526 "%s %s compression",
527 (compress_force) ? "force" : "use",
528 compress_type);
529 }
530 compress_force = false;
531 break;
532 case Opt_ssd:
533 btrfs_set_and_info(root, SSD,
534 "use ssd allocation scheme");
535 break;
536 case Opt_ssd_spread:
537 btrfs_set_and_info(root, SSD_SPREAD,
538 "use spread ssd allocation scheme");
539 btrfs_set_opt(info->mount_opt, SSD);
540 break;
541 case Opt_nossd:
542 btrfs_set_and_info(root, NOSSD,
543 "not using ssd allocation scheme");
544 btrfs_clear_opt(info->mount_opt, SSD);
545 break;
546 case Opt_barrier:
547 btrfs_clear_and_info(root, NOBARRIER,
548 "turning on barriers");
549 break;
550 case Opt_nobarrier:
551 btrfs_set_and_info(root, NOBARRIER,
552 "turning off barriers");
553 break;
554 case Opt_thread_pool:
555 ret = match_int(&args[0], &intarg);
556 if (ret) {
557 goto out;
558 } else if (intarg > 0) {
559 info->thread_pool_size = intarg;
560 } else {
561 ret = -EINVAL;
562 goto out;
563 }
564 break;
565 case Opt_max_inline:
566 num = match_strdup(&args[0]);
567 if (num) {
568 info->max_inline = memparse(num, NULL);
569 kfree(num);
570
571 if (info->max_inline) {
572 info->max_inline = min_t(u64,
573 info->max_inline,
574 root->sectorsize);
575 }
576 btrfs_info(root->fs_info, "max_inline at %llu",
577 info->max_inline);
578 } else {
579 ret = -ENOMEM;
580 goto out;
581 }
582 break;
583 case Opt_alloc_start:
584 num = match_strdup(&args[0]);
585 if (num) {
586 mutex_lock(&info->chunk_mutex);
587 info->alloc_start = memparse(num, NULL);
588 mutex_unlock(&info->chunk_mutex);
589 kfree(num);
590 btrfs_info(root->fs_info, "allocations start at %llu",
591 info->alloc_start);
592 } else {
593 ret = -ENOMEM;
594 goto out;
595 }
596 break;
597 case Opt_acl:
598 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
599 root->fs_info->sb->s_flags |= MS_POSIXACL;
600 break;
601 #else
602 btrfs_err(root->fs_info,
603 "support for ACL not compiled in!");
604 ret = -EINVAL;
605 goto out;
606 #endif
607 case Opt_noacl:
608 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
609 break;
610 case Opt_notreelog:
611 btrfs_set_and_info(root, NOTREELOG,
612 "disabling tree log");
613 break;
614 case Opt_treelog:
615 btrfs_clear_and_info(root, NOTREELOG,
616 "enabling tree log");
617 break;
618 case Opt_norecovery:
619 case Opt_nologreplay:
620 btrfs_set_and_info(root, NOLOGREPLAY,
621 "disabling log replay at mount time");
622 break;
623 case Opt_flushoncommit:
624 btrfs_set_and_info(root, FLUSHONCOMMIT,
625 "turning on flush-on-commit");
626 break;
627 case Opt_noflushoncommit:
628 btrfs_clear_and_info(root, FLUSHONCOMMIT,
629 "turning off flush-on-commit");
630 break;
631 case Opt_ratio:
632 ret = match_int(&args[0], &intarg);
633 if (ret) {
634 goto out;
635 } else if (intarg >= 0) {
636 info->metadata_ratio = intarg;
637 btrfs_info(root->fs_info, "metadata ratio %d",
638 info->metadata_ratio);
639 } else {
640 ret = -EINVAL;
641 goto out;
642 }
643 break;
644 case Opt_discard:
645 btrfs_set_and_info(root, DISCARD,
646 "turning on discard");
647 break;
648 case Opt_nodiscard:
649 btrfs_clear_and_info(root, DISCARD,
650 "turning off discard");
651 break;
652 case Opt_space_cache:
653 case Opt_space_cache_version:
654 if (token == Opt_space_cache ||
655 strcmp(args[0].from, "v1") == 0) {
656 btrfs_clear_opt(root->fs_info->mount_opt,
657 FREE_SPACE_TREE);
658 btrfs_set_and_info(root, SPACE_CACHE,
659 "enabling disk space caching");
660 } else if (strcmp(args[0].from, "v2") == 0) {
661 btrfs_clear_opt(root->fs_info->mount_opt,
662 SPACE_CACHE);
663 btrfs_set_and_info(root, FREE_SPACE_TREE,
664 "enabling free space tree");
665 } else {
666 ret = -EINVAL;
667 goto out;
668 }
669 break;
670 case Opt_rescan_uuid_tree:
671 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
672 break;
673 case Opt_no_space_cache:
674 if (btrfs_test_opt(root, SPACE_CACHE)) {
675 btrfs_clear_and_info(root, SPACE_CACHE,
676 "disabling disk space caching");
677 }
678 if (btrfs_test_opt(root, FREE_SPACE_TREE)) {
679 btrfs_clear_and_info(root, FREE_SPACE_TREE,
680 "disabling free space tree");
681 }
682 break;
683 case Opt_inode_cache:
684 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
685 "enabling inode map caching");
686 break;
687 case Opt_noinode_cache:
688 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
689 "disabling inode map caching");
690 break;
691 case Opt_clear_cache:
692 btrfs_set_and_info(root, CLEAR_CACHE,
693 "force clearing of disk cache");
694 break;
695 case Opt_user_subvol_rm_allowed:
696 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
697 break;
698 case Opt_enospc_debug:
699 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
700 break;
701 case Opt_noenospc_debug:
702 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
703 break;
704 case Opt_defrag:
705 btrfs_set_and_info(root, AUTO_DEFRAG,
706 "enabling auto defrag");
707 break;
708 case Opt_nodefrag:
709 btrfs_clear_and_info(root, AUTO_DEFRAG,
710 "disabling auto defrag");
711 break;
712 case Opt_recovery:
713 btrfs_warn(root->fs_info,
714 "'recovery' is deprecated, use 'usebackuproot' instead");
715 case Opt_usebackuproot:
716 btrfs_info(root->fs_info,
717 "trying to use backup root at mount time");
718 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
719 break;
720 case Opt_skip_balance:
721 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
722 break;
723 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
724 case Opt_check_integrity_including_extent_data:
725 btrfs_info(root->fs_info,
726 "enabling check integrity including extent data");
727 btrfs_set_opt(info->mount_opt,
728 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
729 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
730 break;
731 case Opt_check_integrity:
732 btrfs_info(root->fs_info, "enabling check integrity");
733 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
734 break;
735 case Opt_check_integrity_print_mask:
736 ret = match_int(&args[0], &intarg);
737 if (ret) {
738 goto out;
739 } else if (intarg >= 0) {
740 info->check_integrity_print_mask = intarg;
741 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
742 info->check_integrity_print_mask);
743 } else {
744 ret = -EINVAL;
745 goto out;
746 }
747 break;
748 #else
749 case Opt_check_integrity_including_extent_data:
750 case Opt_check_integrity:
751 case Opt_check_integrity_print_mask:
752 btrfs_err(root->fs_info,
753 "support for check_integrity* not compiled in!");
754 ret = -EINVAL;
755 goto out;
756 #endif
757 case Opt_fatal_errors:
758 if (strcmp(args[0].from, "panic") == 0)
759 btrfs_set_opt(info->mount_opt,
760 PANIC_ON_FATAL_ERROR);
761 else if (strcmp(args[0].from, "bug") == 0)
762 btrfs_clear_opt(info->mount_opt,
763 PANIC_ON_FATAL_ERROR);
764 else {
765 ret = -EINVAL;
766 goto out;
767 }
768 break;
769 case Opt_commit_interval:
770 intarg = 0;
771 ret = match_int(&args[0], &intarg);
772 if (ret < 0) {
773 btrfs_err(root->fs_info, "invalid commit interval");
774 ret = -EINVAL;
775 goto out;
776 }
777 if (intarg > 0) {
778 if (intarg > 300) {
779 btrfs_warn(root->fs_info, "excessive commit interval %d",
780 intarg);
781 }
782 info->commit_interval = intarg;
783 } else {
784 btrfs_info(root->fs_info, "using default commit interval %ds",
785 BTRFS_DEFAULT_COMMIT_INTERVAL);
786 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
787 }
788 break;
789 #ifdef CONFIG_BTRFS_DEBUG
790 case Opt_fragment_all:
791 btrfs_info(root->fs_info, "fragmenting all space");
792 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
793 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
794 break;
795 case Opt_fragment_metadata:
796 btrfs_info(root->fs_info, "fragmenting metadata");
797 btrfs_set_opt(info->mount_opt,
798 FRAGMENT_METADATA);
799 break;
800 case Opt_fragment_data:
801 btrfs_info(root->fs_info, "fragmenting data");
802 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
803 break;
804 #endif
805 case Opt_err:
806 btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
807 ret = -EINVAL;
808 goto out;
809 default:
810 break;
811 }
812 }
813 check:
814 /*
815 * Extra check for current option against current flag
816 */
817 if (btrfs_test_opt(root, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
818 btrfs_err(root->fs_info,
819 "nologreplay must be used with ro mount option");
820 ret = -EINVAL;
821 }
822 out:
823 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE) &&
824 !btrfs_test_opt(root, FREE_SPACE_TREE) &&
825 !btrfs_test_opt(root, CLEAR_CACHE)) {
826 btrfs_err(root->fs_info, "cannot disable free space tree");
827 ret = -EINVAL;
828
829 }
830 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
831 btrfs_info(root->fs_info, "disk space caching is enabled");
832 if (!ret && btrfs_test_opt(root, FREE_SPACE_TREE))
833 btrfs_info(root->fs_info, "using free space tree");
834 kfree(orig);
835 return ret;
836 }
837
838 /*
839 * Parse mount options that are required early in the mount process.
840 *
841 * All other options will be parsed on much later in the mount process and
842 * only when we need to allocate a new super block.
843 */
844 static int btrfs_parse_early_options(const char *options, fmode_t flags,
845 void *holder, char **subvol_name, u64 *subvol_objectid,
846 struct btrfs_fs_devices **fs_devices)
847 {
848 substring_t args[MAX_OPT_ARGS];
849 char *device_name, *opts, *orig, *p;
850 char *num = NULL;
851 int error = 0;
852
853 if (!options)
854 return 0;
855
856 /*
857 * strsep changes the string, duplicate it because parse_options
858 * gets called twice
859 */
860 opts = kstrdup(options, GFP_KERNEL);
861 if (!opts)
862 return -ENOMEM;
863 orig = opts;
864
865 while ((p = strsep(&opts, ",")) != NULL) {
866 int token;
867 if (!*p)
868 continue;
869
870 token = match_token(p, tokens, args);
871 switch (token) {
872 case Opt_subvol:
873 kfree(*subvol_name);
874 *subvol_name = match_strdup(&args[0]);
875 if (!*subvol_name) {
876 error = -ENOMEM;
877 goto out;
878 }
879 break;
880 case Opt_subvolid:
881 num = match_strdup(&args[0]);
882 if (num) {
883 *subvol_objectid = memparse(num, NULL);
884 kfree(num);
885 /* we want the original fs_tree */
886 if (!*subvol_objectid)
887 *subvol_objectid =
888 BTRFS_FS_TREE_OBJECTID;
889 } else {
890 error = -EINVAL;
891 goto out;
892 }
893 break;
894 case Opt_subvolrootid:
895 printk(KERN_WARNING
896 "BTRFS: 'subvolrootid' mount option is deprecated and has "
897 "no effect\n");
898 break;
899 case Opt_device:
900 device_name = match_strdup(&args[0]);
901 if (!device_name) {
902 error = -ENOMEM;
903 goto out;
904 }
905 error = btrfs_scan_one_device(device_name,
906 flags, holder, fs_devices);
907 kfree(device_name);
908 if (error)
909 goto out;
910 break;
911 default:
912 break;
913 }
914 }
915
916 out:
917 kfree(orig);
918 return error;
919 }
920
921 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
922 u64 subvol_objectid)
923 {
924 struct btrfs_root *root = fs_info->tree_root;
925 struct btrfs_root *fs_root;
926 struct btrfs_root_ref *root_ref;
927 struct btrfs_inode_ref *inode_ref;
928 struct btrfs_key key;
929 struct btrfs_path *path = NULL;
930 char *name = NULL, *ptr;
931 u64 dirid;
932 int len;
933 int ret;
934
935 path = btrfs_alloc_path();
936 if (!path) {
937 ret = -ENOMEM;
938 goto err;
939 }
940 path->leave_spinning = 1;
941
942 name = kmalloc(PATH_MAX, GFP_NOFS);
943 if (!name) {
944 ret = -ENOMEM;
945 goto err;
946 }
947 ptr = name + PATH_MAX - 1;
948 ptr[0] = '\0';
949
950 /*
951 * Walk up the subvolume trees in the tree of tree roots by root
952 * backrefs until we hit the top-level subvolume.
953 */
954 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
955 key.objectid = subvol_objectid;
956 key.type = BTRFS_ROOT_BACKREF_KEY;
957 key.offset = (u64)-1;
958
959 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
960 if (ret < 0) {
961 goto err;
962 } else if (ret > 0) {
963 ret = btrfs_previous_item(root, path, subvol_objectid,
964 BTRFS_ROOT_BACKREF_KEY);
965 if (ret < 0) {
966 goto err;
967 } else if (ret > 0) {
968 ret = -ENOENT;
969 goto err;
970 }
971 }
972
973 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
974 subvol_objectid = key.offset;
975
976 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
977 struct btrfs_root_ref);
978 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
979 ptr -= len + 1;
980 if (ptr < name) {
981 ret = -ENAMETOOLONG;
982 goto err;
983 }
984 read_extent_buffer(path->nodes[0], ptr + 1,
985 (unsigned long)(root_ref + 1), len);
986 ptr[0] = '/';
987 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
988 btrfs_release_path(path);
989
990 key.objectid = subvol_objectid;
991 key.type = BTRFS_ROOT_ITEM_KEY;
992 key.offset = (u64)-1;
993 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
994 if (IS_ERR(fs_root)) {
995 ret = PTR_ERR(fs_root);
996 goto err;
997 }
998
999 /*
1000 * Walk up the filesystem tree by inode refs until we hit the
1001 * root directory.
1002 */
1003 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1004 key.objectid = dirid;
1005 key.type = BTRFS_INODE_REF_KEY;
1006 key.offset = (u64)-1;
1007
1008 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1009 if (ret < 0) {
1010 goto err;
1011 } else if (ret > 0) {
1012 ret = btrfs_previous_item(fs_root, path, dirid,
1013 BTRFS_INODE_REF_KEY);
1014 if (ret < 0) {
1015 goto err;
1016 } else if (ret > 0) {
1017 ret = -ENOENT;
1018 goto err;
1019 }
1020 }
1021
1022 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1023 dirid = key.offset;
1024
1025 inode_ref = btrfs_item_ptr(path->nodes[0],
1026 path->slots[0],
1027 struct btrfs_inode_ref);
1028 len = btrfs_inode_ref_name_len(path->nodes[0],
1029 inode_ref);
1030 ptr -= len + 1;
1031 if (ptr < name) {
1032 ret = -ENAMETOOLONG;
1033 goto err;
1034 }
1035 read_extent_buffer(path->nodes[0], ptr + 1,
1036 (unsigned long)(inode_ref + 1), len);
1037 ptr[0] = '/';
1038 btrfs_release_path(path);
1039 }
1040 }
1041
1042 btrfs_free_path(path);
1043 if (ptr == name + PATH_MAX - 1) {
1044 name[0] = '/';
1045 name[1] = '\0';
1046 } else {
1047 memmove(name, ptr, name + PATH_MAX - ptr);
1048 }
1049 return name;
1050
1051 err:
1052 btrfs_free_path(path);
1053 kfree(name);
1054 return ERR_PTR(ret);
1055 }
1056
1057 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1058 {
1059 struct btrfs_root *root = fs_info->tree_root;
1060 struct btrfs_dir_item *di;
1061 struct btrfs_path *path;
1062 struct btrfs_key location;
1063 u64 dir_id;
1064
1065 path = btrfs_alloc_path();
1066 if (!path)
1067 return -ENOMEM;
1068 path->leave_spinning = 1;
1069
1070 /*
1071 * Find the "default" dir item which points to the root item that we
1072 * will mount by default if we haven't been given a specific subvolume
1073 * to mount.
1074 */
1075 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1076 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1077 if (IS_ERR(di)) {
1078 btrfs_free_path(path);
1079 return PTR_ERR(di);
1080 }
1081 if (!di) {
1082 /*
1083 * Ok the default dir item isn't there. This is weird since
1084 * it's always been there, but don't freak out, just try and
1085 * mount the top-level subvolume.
1086 */
1087 btrfs_free_path(path);
1088 *objectid = BTRFS_FS_TREE_OBJECTID;
1089 return 0;
1090 }
1091
1092 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1093 btrfs_free_path(path);
1094 *objectid = location.objectid;
1095 return 0;
1096 }
1097
1098 static int btrfs_fill_super(struct super_block *sb,
1099 struct btrfs_fs_devices *fs_devices,
1100 void *data, int silent)
1101 {
1102 struct inode *inode;
1103 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1104 struct btrfs_key key;
1105 int err;
1106
1107 sb->s_maxbytes = MAX_LFS_FILESIZE;
1108 sb->s_magic = BTRFS_SUPER_MAGIC;
1109 sb->s_op = &btrfs_super_ops;
1110 sb->s_d_op = &btrfs_dentry_operations;
1111 sb->s_export_op = &btrfs_export_ops;
1112 sb->s_xattr = btrfs_xattr_handlers;
1113 sb->s_time_gran = 1;
1114 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1115 sb->s_flags |= MS_POSIXACL;
1116 #endif
1117 sb->s_flags |= MS_I_VERSION;
1118 sb->s_iflags |= SB_I_CGROUPWB;
1119 err = open_ctree(sb, fs_devices, (char *)data);
1120 if (err) {
1121 printk(KERN_ERR "BTRFS: open_ctree failed\n");
1122 return err;
1123 }
1124
1125 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1126 key.type = BTRFS_INODE_ITEM_KEY;
1127 key.offset = 0;
1128 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1129 if (IS_ERR(inode)) {
1130 err = PTR_ERR(inode);
1131 goto fail_close;
1132 }
1133
1134 sb->s_root = d_make_root(inode);
1135 if (!sb->s_root) {
1136 err = -ENOMEM;
1137 goto fail_close;
1138 }
1139
1140 save_mount_options(sb, data);
1141 cleancache_init_fs(sb);
1142 sb->s_flags |= MS_ACTIVE;
1143 return 0;
1144
1145 fail_close:
1146 close_ctree(fs_info->tree_root);
1147 return err;
1148 }
1149
1150 int btrfs_sync_fs(struct super_block *sb, int wait)
1151 {
1152 struct btrfs_trans_handle *trans;
1153 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1154 struct btrfs_root *root = fs_info->tree_root;
1155
1156 trace_btrfs_sync_fs(wait);
1157
1158 if (!wait) {
1159 filemap_flush(fs_info->btree_inode->i_mapping);
1160 return 0;
1161 }
1162
1163 btrfs_wait_ordered_roots(fs_info, -1);
1164
1165 trans = btrfs_attach_transaction_barrier(root);
1166 if (IS_ERR(trans)) {
1167 /* no transaction, don't bother */
1168 if (PTR_ERR(trans) == -ENOENT) {
1169 /*
1170 * Exit unless we have some pending changes
1171 * that need to go through commit
1172 */
1173 if (fs_info->pending_changes == 0)
1174 return 0;
1175 /*
1176 * A non-blocking test if the fs is frozen. We must not
1177 * start a new transaction here otherwise a deadlock
1178 * happens. The pending operations are delayed to the
1179 * next commit after thawing.
1180 */
1181 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1182 __sb_end_write(sb, SB_FREEZE_WRITE);
1183 else
1184 return 0;
1185 trans = btrfs_start_transaction(root, 0);
1186 }
1187 if (IS_ERR(trans))
1188 return PTR_ERR(trans);
1189 }
1190 return btrfs_commit_transaction(trans, root);
1191 }
1192
1193 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1194 {
1195 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1196 struct btrfs_root *root = info->tree_root;
1197 char *compress_type;
1198
1199 if (btrfs_test_opt(root, DEGRADED))
1200 seq_puts(seq, ",degraded");
1201 if (btrfs_test_opt(root, NODATASUM))
1202 seq_puts(seq, ",nodatasum");
1203 if (btrfs_test_opt(root, NODATACOW))
1204 seq_puts(seq, ",nodatacow");
1205 if (btrfs_test_opt(root, NOBARRIER))
1206 seq_puts(seq, ",nobarrier");
1207 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1208 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1209 if (info->alloc_start != 0)
1210 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1211 if (info->thread_pool_size != min_t(unsigned long,
1212 num_online_cpus() + 2, 8))
1213 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1214 if (btrfs_test_opt(root, COMPRESS)) {
1215 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1216 compress_type = "zlib";
1217 else
1218 compress_type = "lzo";
1219 if (btrfs_test_opt(root, FORCE_COMPRESS))
1220 seq_printf(seq, ",compress-force=%s", compress_type);
1221 else
1222 seq_printf(seq, ",compress=%s", compress_type);
1223 }
1224 if (btrfs_test_opt(root, NOSSD))
1225 seq_puts(seq, ",nossd");
1226 if (btrfs_test_opt(root, SSD_SPREAD))
1227 seq_puts(seq, ",ssd_spread");
1228 else if (btrfs_test_opt(root, SSD))
1229 seq_puts(seq, ",ssd");
1230 if (btrfs_test_opt(root, NOTREELOG))
1231 seq_puts(seq, ",notreelog");
1232 if (btrfs_test_opt(root, NOLOGREPLAY))
1233 seq_puts(seq, ",nologreplay");
1234 if (btrfs_test_opt(root, FLUSHONCOMMIT))
1235 seq_puts(seq, ",flushoncommit");
1236 if (btrfs_test_opt(root, DISCARD))
1237 seq_puts(seq, ",discard");
1238 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1239 seq_puts(seq, ",noacl");
1240 if (btrfs_test_opt(root, SPACE_CACHE))
1241 seq_puts(seq, ",space_cache");
1242 else if (btrfs_test_opt(root, FREE_SPACE_TREE))
1243 seq_puts(seq, ",space_cache=v2");
1244 else
1245 seq_puts(seq, ",nospace_cache");
1246 if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1247 seq_puts(seq, ",rescan_uuid_tree");
1248 if (btrfs_test_opt(root, CLEAR_CACHE))
1249 seq_puts(seq, ",clear_cache");
1250 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1251 seq_puts(seq, ",user_subvol_rm_allowed");
1252 if (btrfs_test_opt(root, ENOSPC_DEBUG))
1253 seq_puts(seq, ",enospc_debug");
1254 if (btrfs_test_opt(root, AUTO_DEFRAG))
1255 seq_puts(seq, ",autodefrag");
1256 if (btrfs_test_opt(root, INODE_MAP_CACHE))
1257 seq_puts(seq, ",inode_cache");
1258 if (btrfs_test_opt(root, SKIP_BALANCE))
1259 seq_puts(seq, ",skip_balance");
1260 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1261 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1262 seq_puts(seq, ",check_int_data");
1263 else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1264 seq_puts(seq, ",check_int");
1265 if (info->check_integrity_print_mask)
1266 seq_printf(seq, ",check_int_print_mask=%d",
1267 info->check_integrity_print_mask);
1268 #endif
1269 if (info->metadata_ratio)
1270 seq_printf(seq, ",metadata_ratio=%d",
1271 info->metadata_ratio);
1272 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1273 seq_puts(seq, ",fatal_errors=panic");
1274 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1275 seq_printf(seq, ",commit=%d", info->commit_interval);
1276 #ifdef CONFIG_BTRFS_DEBUG
1277 if (btrfs_test_opt(root, FRAGMENT_DATA))
1278 seq_puts(seq, ",fragment=data");
1279 if (btrfs_test_opt(root, FRAGMENT_METADATA))
1280 seq_puts(seq, ",fragment=metadata");
1281 #endif
1282 seq_printf(seq, ",subvolid=%llu",
1283 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1284 seq_puts(seq, ",subvol=");
1285 seq_dentry(seq, dentry, " \t\n\\");
1286 return 0;
1287 }
1288
1289 static int btrfs_test_super(struct super_block *s, void *data)
1290 {
1291 struct btrfs_fs_info *p = data;
1292 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1293
1294 return fs_info->fs_devices == p->fs_devices;
1295 }
1296
1297 static int btrfs_set_super(struct super_block *s, void *data)
1298 {
1299 int err = set_anon_super(s, data);
1300 if (!err)
1301 s->s_fs_info = data;
1302 return err;
1303 }
1304
1305 /*
1306 * subvolumes are identified by ino 256
1307 */
1308 static inline int is_subvolume_inode(struct inode *inode)
1309 {
1310 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1311 return 1;
1312 return 0;
1313 }
1314
1315 /*
1316 * This will add subvolid=0 to the argument string while removing any subvol=
1317 * and subvolid= arguments to make sure we get the top-level root for path
1318 * walking to the subvol we want.
1319 */
1320 static char *setup_root_args(char *args)
1321 {
1322 char *buf, *dst, *sep;
1323
1324 if (!args)
1325 return kstrdup("subvolid=0", GFP_NOFS);
1326
1327 /* The worst case is that we add ",subvolid=0" to the end. */
1328 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
1329 if (!buf)
1330 return NULL;
1331
1332 while (1) {
1333 sep = strchrnul(args, ',');
1334 if (!strstarts(args, "subvol=") &&
1335 !strstarts(args, "subvolid=")) {
1336 memcpy(dst, args, sep - args);
1337 dst += sep - args;
1338 *dst++ = ',';
1339 }
1340 if (*sep)
1341 args = sep + 1;
1342 else
1343 break;
1344 }
1345 strcpy(dst, "subvolid=0");
1346
1347 return buf;
1348 }
1349
1350 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1351 int flags, const char *device_name,
1352 char *data)
1353 {
1354 struct dentry *root;
1355 struct vfsmount *mnt = NULL;
1356 char *newargs;
1357 int ret;
1358
1359 newargs = setup_root_args(data);
1360 if (!newargs) {
1361 root = ERR_PTR(-ENOMEM);
1362 goto out;
1363 }
1364
1365 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
1366 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
1367 if (flags & MS_RDONLY) {
1368 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
1369 device_name, newargs);
1370 } else {
1371 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
1372 device_name, newargs);
1373 if (IS_ERR(mnt)) {
1374 root = ERR_CAST(mnt);
1375 mnt = NULL;
1376 goto out;
1377 }
1378
1379 down_write(&mnt->mnt_sb->s_umount);
1380 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1381 up_write(&mnt->mnt_sb->s_umount);
1382 if (ret < 0) {
1383 root = ERR_PTR(ret);
1384 goto out;
1385 }
1386 }
1387 }
1388 if (IS_ERR(mnt)) {
1389 root = ERR_CAST(mnt);
1390 mnt = NULL;
1391 goto out;
1392 }
1393
1394 if (!subvol_name) {
1395 if (!subvol_objectid) {
1396 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1397 &subvol_objectid);
1398 if (ret) {
1399 root = ERR_PTR(ret);
1400 goto out;
1401 }
1402 }
1403 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1404 subvol_objectid);
1405 if (IS_ERR(subvol_name)) {
1406 root = ERR_CAST(subvol_name);
1407 subvol_name = NULL;
1408 goto out;
1409 }
1410
1411 }
1412
1413 root = mount_subtree(mnt, subvol_name);
1414 /* mount_subtree() drops our reference on the vfsmount. */
1415 mnt = NULL;
1416
1417 if (!IS_ERR(root)) {
1418 struct super_block *s = root->d_sb;
1419 struct inode *root_inode = d_inode(root);
1420 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1421
1422 ret = 0;
1423 if (!is_subvolume_inode(root_inode)) {
1424 pr_err("BTRFS: '%s' is not a valid subvolume\n",
1425 subvol_name);
1426 ret = -EINVAL;
1427 }
1428 if (subvol_objectid && root_objectid != subvol_objectid) {
1429 /*
1430 * This will also catch a race condition where a
1431 * subvolume which was passed by ID is renamed and
1432 * another subvolume is renamed over the old location.
1433 */
1434 pr_err("BTRFS: subvol '%s' does not match subvolid %llu\n",
1435 subvol_name, subvol_objectid);
1436 ret = -EINVAL;
1437 }
1438 if (ret) {
1439 dput(root);
1440 root = ERR_PTR(ret);
1441 deactivate_locked_super(s);
1442 }
1443 }
1444
1445 out:
1446 mntput(mnt);
1447 kfree(newargs);
1448 kfree(subvol_name);
1449 return root;
1450 }
1451
1452 static int parse_security_options(char *orig_opts,
1453 struct security_mnt_opts *sec_opts)
1454 {
1455 char *secdata = NULL;
1456 int ret = 0;
1457
1458 secdata = alloc_secdata();
1459 if (!secdata)
1460 return -ENOMEM;
1461 ret = security_sb_copy_data(orig_opts, secdata);
1462 if (ret) {
1463 free_secdata(secdata);
1464 return ret;
1465 }
1466 ret = security_sb_parse_opts_str(secdata, sec_opts);
1467 free_secdata(secdata);
1468 return ret;
1469 }
1470
1471 static int setup_security_options(struct btrfs_fs_info *fs_info,
1472 struct super_block *sb,
1473 struct security_mnt_opts *sec_opts)
1474 {
1475 int ret = 0;
1476
1477 /*
1478 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1479 * is valid.
1480 */
1481 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1482 if (ret)
1483 return ret;
1484
1485 #ifdef CONFIG_SECURITY
1486 if (!fs_info->security_opts.num_mnt_opts) {
1487 /* first time security setup, copy sec_opts to fs_info */
1488 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1489 } else {
1490 /*
1491 * Since SELinux(the only one supports security_mnt_opts) does
1492 * NOT support changing context during remount/mount same sb,
1493 * This must be the same or part of the same security options,
1494 * just free it.
1495 */
1496 security_free_mnt_opts(sec_opts);
1497 }
1498 #endif
1499 return ret;
1500 }
1501
1502 /*
1503 * Find a superblock for the given device / mount point.
1504 *
1505 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1506 * for multiple device setup. Make sure to keep it in sync.
1507 */
1508 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1509 const char *device_name, void *data)
1510 {
1511 struct block_device *bdev = NULL;
1512 struct super_block *s;
1513 struct btrfs_fs_devices *fs_devices = NULL;
1514 struct btrfs_fs_info *fs_info = NULL;
1515 struct security_mnt_opts new_sec_opts;
1516 fmode_t mode = FMODE_READ;
1517 char *subvol_name = NULL;
1518 u64 subvol_objectid = 0;
1519 int error = 0;
1520
1521 if (!(flags & MS_RDONLY))
1522 mode |= FMODE_WRITE;
1523
1524 error = btrfs_parse_early_options(data, mode, fs_type,
1525 &subvol_name, &subvol_objectid,
1526 &fs_devices);
1527 if (error) {
1528 kfree(subvol_name);
1529 return ERR_PTR(error);
1530 }
1531
1532 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1533 /* mount_subvol() will free subvol_name. */
1534 return mount_subvol(subvol_name, subvol_objectid, flags,
1535 device_name, data);
1536 }
1537
1538 security_init_mnt_opts(&new_sec_opts);
1539 if (data) {
1540 error = parse_security_options(data, &new_sec_opts);
1541 if (error)
1542 return ERR_PTR(error);
1543 }
1544
1545 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1546 if (error)
1547 goto error_sec_opts;
1548
1549 /*
1550 * Setup a dummy root and fs_info for test/set super. This is because
1551 * we don't actually fill this stuff out until open_ctree, but we need
1552 * it for searching for existing supers, so this lets us do that and
1553 * then open_ctree will properly initialize everything later.
1554 */
1555 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1556 if (!fs_info) {
1557 error = -ENOMEM;
1558 goto error_sec_opts;
1559 }
1560
1561 fs_info->fs_devices = fs_devices;
1562
1563 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1564 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1565 security_init_mnt_opts(&fs_info->security_opts);
1566 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1567 error = -ENOMEM;
1568 goto error_fs_info;
1569 }
1570
1571 error = btrfs_open_devices(fs_devices, mode, fs_type);
1572 if (error)
1573 goto error_fs_info;
1574
1575 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1576 error = -EACCES;
1577 goto error_close_devices;
1578 }
1579
1580 bdev = fs_devices->latest_bdev;
1581 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1582 fs_info);
1583 if (IS_ERR(s)) {
1584 error = PTR_ERR(s);
1585 goto error_close_devices;
1586 }
1587
1588 if (s->s_root) {
1589 btrfs_close_devices(fs_devices);
1590 free_fs_info(fs_info);
1591 if ((flags ^ s->s_flags) & MS_RDONLY)
1592 error = -EBUSY;
1593 } else {
1594 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1595 btrfs_sb(s)->bdev_holder = fs_type;
1596 error = btrfs_fill_super(s, fs_devices, data,
1597 flags & MS_SILENT ? 1 : 0);
1598 }
1599 if (error) {
1600 deactivate_locked_super(s);
1601 goto error_sec_opts;
1602 }
1603
1604 fs_info = btrfs_sb(s);
1605 error = setup_security_options(fs_info, s, &new_sec_opts);
1606 if (error) {
1607 deactivate_locked_super(s);
1608 goto error_sec_opts;
1609 }
1610
1611 return dget(s->s_root);
1612
1613 error_close_devices:
1614 btrfs_close_devices(fs_devices);
1615 error_fs_info:
1616 free_fs_info(fs_info);
1617 error_sec_opts:
1618 security_free_mnt_opts(&new_sec_opts);
1619 return ERR_PTR(error);
1620 }
1621
1622 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1623 int new_pool_size, int old_pool_size)
1624 {
1625 if (new_pool_size == old_pool_size)
1626 return;
1627
1628 fs_info->thread_pool_size = new_pool_size;
1629
1630 btrfs_info(fs_info, "resize thread pool %d -> %d",
1631 old_pool_size, new_pool_size);
1632
1633 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1634 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1635 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1636 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1637 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1638 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1639 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1640 new_pool_size);
1641 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1642 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1643 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1644 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1645 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1646 new_pool_size);
1647 }
1648
1649 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1650 {
1651 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1652 }
1653
1654 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1655 unsigned long old_opts, int flags)
1656 {
1657 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1658 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1659 (flags & MS_RDONLY))) {
1660 /* wait for any defraggers to finish */
1661 wait_event(fs_info->transaction_wait,
1662 (atomic_read(&fs_info->defrag_running) == 0));
1663 if (flags & MS_RDONLY)
1664 sync_filesystem(fs_info->sb);
1665 }
1666 }
1667
1668 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1669 unsigned long old_opts)
1670 {
1671 /*
1672 * We need cleanup all defragable inodes if the autodefragment is
1673 * close or the fs is R/O.
1674 */
1675 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1676 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1677 (fs_info->sb->s_flags & MS_RDONLY))) {
1678 btrfs_cleanup_defrag_inodes(fs_info);
1679 }
1680
1681 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1682 }
1683
1684 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1685 {
1686 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1687 struct btrfs_root *root = fs_info->tree_root;
1688 unsigned old_flags = sb->s_flags;
1689 unsigned long old_opts = fs_info->mount_opt;
1690 unsigned long old_compress_type = fs_info->compress_type;
1691 u64 old_max_inline = fs_info->max_inline;
1692 u64 old_alloc_start = fs_info->alloc_start;
1693 int old_thread_pool_size = fs_info->thread_pool_size;
1694 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1695 int ret;
1696
1697 sync_filesystem(sb);
1698 btrfs_remount_prepare(fs_info);
1699
1700 if (data) {
1701 struct security_mnt_opts new_sec_opts;
1702
1703 security_init_mnt_opts(&new_sec_opts);
1704 ret = parse_security_options(data, &new_sec_opts);
1705 if (ret)
1706 goto restore;
1707 ret = setup_security_options(fs_info, sb,
1708 &new_sec_opts);
1709 if (ret) {
1710 security_free_mnt_opts(&new_sec_opts);
1711 goto restore;
1712 }
1713 }
1714
1715 ret = btrfs_parse_options(root, data, *flags);
1716 if (ret) {
1717 ret = -EINVAL;
1718 goto restore;
1719 }
1720
1721 btrfs_remount_begin(fs_info, old_opts, *flags);
1722 btrfs_resize_thread_pool(fs_info,
1723 fs_info->thread_pool_size, old_thread_pool_size);
1724
1725 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1726 goto out;
1727
1728 if (*flags & MS_RDONLY) {
1729 /*
1730 * this also happens on 'umount -rf' or on shutdown, when
1731 * the filesystem is busy.
1732 */
1733 cancel_work_sync(&fs_info->async_reclaim_work);
1734
1735 /* wait for the uuid_scan task to finish */
1736 down(&fs_info->uuid_tree_rescan_sem);
1737 /* avoid complains from lockdep et al. */
1738 up(&fs_info->uuid_tree_rescan_sem);
1739
1740 sb->s_flags |= MS_RDONLY;
1741
1742 /*
1743 * Setting MS_RDONLY will put the cleaner thread to
1744 * sleep at the next loop if it's already active.
1745 * If it's already asleep, we'll leave unused block
1746 * groups on disk until we're mounted read-write again
1747 * unless we clean them up here.
1748 */
1749 btrfs_delete_unused_bgs(fs_info);
1750
1751 btrfs_dev_replace_suspend_for_unmount(fs_info);
1752 btrfs_scrub_cancel(fs_info);
1753 btrfs_pause_balance(fs_info);
1754
1755 ret = btrfs_commit_super(root);
1756 if (ret)
1757 goto restore;
1758 } else {
1759 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1760 btrfs_err(fs_info,
1761 "Remounting read-write after error is not allowed");
1762 ret = -EINVAL;
1763 goto restore;
1764 }
1765 if (fs_info->fs_devices->rw_devices == 0) {
1766 ret = -EACCES;
1767 goto restore;
1768 }
1769
1770 if (fs_info->fs_devices->missing_devices >
1771 fs_info->num_tolerated_disk_barrier_failures &&
1772 !(*flags & MS_RDONLY)) {
1773 btrfs_warn(fs_info,
1774 "too many missing devices, writeable remount is not allowed");
1775 ret = -EACCES;
1776 goto restore;
1777 }
1778
1779 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1780 ret = -EINVAL;
1781 goto restore;
1782 }
1783
1784 ret = btrfs_cleanup_fs_roots(fs_info);
1785 if (ret)
1786 goto restore;
1787
1788 /* recover relocation */
1789 mutex_lock(&fs_info->cleaner_mutex);
1790 ret = btrfs_recover_relocation(root);
1791 mutex_unlock(&fs_info->cleaner_mutex);
1792 if (ret)
1793 goto restore;
1794
1795 ret = btrfs_resume_balance_async(fs_info);
1796 if (ret)
1797 goto restore;
1798
1799 ret = btrfs_resume_dev_replace_async(fs_info);
1800 if (ret) {
1801 btrfs_warn(fs_info, "failed to resume dev_replace");
1802 goto restore;
1803 }
1804
1805 if (!fs_info->uuid_root) {
1806 btrfs_info(fs_info, "creating UUID tree");
1807 ret = btrfs_create_uuid_tree(fs_info);
1808 if (ret) {
1809 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1810 goto restore;
1811 }
1812 }
1813 sb->s_flags &= ~MS_RDONLY;
1814 }
1815 out:
1816 wake_up_process(fs_info->transaction_kthread);
1817 btrfs_remount_cleanup(fs_info, old_opts);
1818 return 0;
1819
1820 restore:
1821 /* We've hit an error - don't reset MS_RDONLY */
1822 if (sb->s_flags & MS_RDONLY)
1823 old_flags |= MS_RDONLY;
1824 sb->s_flags = old_flags;
1825 fs_info->mount_opt = old_opts;
1826 fs_info->compress_type = old_compress_type;
1827 fs_info->max_inline = old_max_inline;
1828 mutex_lock(&fs_info->chunk_mutex);
1829 fs_info->alloc_start = old_alloc_start;
1830 mutex_unlock(&fs_info->chunk_mutex);
1831 btrfs_resize_thread_pool(fs_info,
1832 old_thread_pool_size, fs_info->thread_pool_size);
1833 fs_info->metadata_ratio = old_metadata_ratio;
1834 btrfs_remount_cleanup(fs_info, old_opts);
1835 return ret;
1836 }
1837
1838 /* Used to sort the devices by max_avail(descending sort) */
1839 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1840 const void *dev_info2)
1841 {
1842 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1843 ((struct btrfs_device_info *)dev_info2)->max_avail)
1844 return -1;
1845 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1846 ((struct btrfs_device_info *)dev_info2)->max_avail)
1847 return 1;
1848 else
1849 return 0;
1850 }
1851
1852 /*
1853 * sort the devices by max_avail, in which max free extent size of each device
1854 * is stored.(Descending Sort)
1855 */
1856 static inline void btrfs_descending_sort_devices(
1857 struct btrfs_device_info *devices,
1858 size_t nr_devices)
1859 {
1860 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1861 btrfs_cmp_device_free_bytes, NULL);
1862 }
1863
1864 /*
1865 * The helper to calc the free space on the devices that can be used to store
1866 * file data.
1867 */
1868 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1869 {
1870 struct btrfs_fs_info *fs_info = root->fs_info;
1871 struct btrfs_device_info *devices_info;
1872 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1873 struct btrfs_device *device;
1874 u64 skip_space;
1875 u64 type;
1876 u64 avail_space;
1877 u64 used_space;
1878 u64 min_stripe_size;
1879 int min_stripes = 1, num_stripes = 1;
1880 int i = 0, nr_devices;
1881 int ret;
1882
1883 /*
1884 * We aren't under the device list lock, so this is racey-ish, but good
1885 * enough for our purposes.
1886 */
1887 nr_devices = fs_info->fs_devices->open_devices;
1888 if (!nr_devices) {
1889 smp_mb();
1890 nr_devices = fs_info->fs_devices->open_devices;
1891 ASSERT(nr_devices);
1892 if (!nr_devices) {
1893 *free_bytes = 0;
1894 return 0;
1895 }
1896 }
1897
1898 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1899 GFP_NOFS);
1900 if (!devices_info)
1901 return -ENOMEM;
1902
1903 /* calc min stripe number for data space alloction */
1904 type = btrfs_get_alloc_profile(root, 1);
1905 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1906 min_stripes = 2;
1907 num_stripes = nr_devices;
1908 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1909 min_stripes = 2;
1910 num_stripes = 2;
1911 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1912 min_stripes = 4;
1913 num_stripes = 4;
1914 }
1915
1916 if (type & BTRFS_BLOCK_GROUP_DUP)
1917 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1918 else
1919 min_stripe_size = BTRFS_STRIPE_LEN;
1920
1921 if (fs_info->alloc_start)
1922 mutex_lock(&fs_devices->device_list_mutex);
1923 rcu_read_lock();
1924 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1925 if (!device->in_fs_metadata || !device->bdev ||
1926 device->is_tgtdev_for_dev_replace)
1927 continue;
1928
1929 if (i >= nr_devices)
1930 break;
1931
1932 avail_space = device->total_bytes - device->bytes_used;
1933
1934 /* align with stripe_len */
1935 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1936 avail_space *= BTRFS_STRIPE_LEN;
1937
1938 /*
1939 * In order to avoid overwritting the superblock on the drive,
1940 * btrfs starts at an offset of at least 1MB when doing chunk
1941 * allocation.
1942 */
1943 skip_space = SZ_1M;
1944
1945 /* user can set the offset in fs_info->alloc_start. */
1946 if (fs_info->alloc_start &&
1947 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1948 device->total_bytes) {
1949 rcu_read_unlock();
1950 skip_space = max(fs_info->alloc_start, skip_space);
1951
1952 /*
1953 * btrfs can not use the free space in
1954 * [0, skip_space - 1], we must subtract it from the
1955 * total. In order to implement it, we account the used
1956 * space in this range first.
1957 */
1958 ret = btrfs_account_dev_extents_size(device, 0,
1959 skip_space - 1,
1960 &used_space);
1961 if (ret) {
1962 kfree(devices_info);
1963 mutex_unlock(&fs_devices->device_list_mutex);
1964 return ret;
1965 }
1966
1967 rcu_read_lock();
1968
1969 /* calc the free space in [0, skip_space - 1] */
1970 skip_space -= used_space;
1971 }
1972
1973 /*
1974 * we can use the free space in [0, skip_space - 1], subtract
1975 * it from the total.
1976 */
1977 if (avail_space && avail_space >= skip_space)
1978 avail_space -= skip_space;
1979 else
1980 avail_space = 0;
1981
1982 if (avail_space < min_stripe_size)
1983 continue;
1984
1985 devices_info[i].dev = device;
1986 devices_info[i].max_avail = avail_space;
1987
1988 i++;
1989 }
1990 rcu_read_unlock();
1991 if (fs_info->alloc_start)
1992 mutex_unlock(&fs_devices->device_list_mutex);
1993
1994 nr_devices = i;
1995
1996 btrfs_descending_sort_devices(devices_info, nr_devices);
1997
1998 i = nr_devices - 1;
1999 avail_space = 0;
2000 while (nr_devices >= min_stripes) {
2001 if (num_stripes > nr_devices)
2002 num_stripes = nr_devices;
2003
2004 if (devices_info[i].max_avail >= min_stripe_size) {
2005 int j;
2006 u64 alloc_size;
2007
2008 avail_space += devices_info[i].max_avail * num_stripes;
2009 alloc_size = devices_info[i].max_avail;
2010 for (j = i + 1 - num_stripes; j <= i; j++)
2011 devices_info[j].max_avail -= alloc_size;
2012 }
2013 i--;
2014 nr_devices--;
2015 }
2016
2017 kfree(devices_info);
2018 *free_bytes = avail_space;
2019 return 0;
2020 }
2021
2022 /*
2023 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2024 *
2025 * If there's a redundant raid level at DATA block groups, use the respective
2026 * multiplier to scale the sizes.
2027 *
2028 * Unused device space usage is based on simulating the chunk allocator
2029 * algorithm that respects the device sizes, order of allocations and the
2030 * 'alloc_start' value, this is a close approximation of the actual use but
2031 * there are other factors that may change the result (like a new metadata
2032 * chunk).
2033 *
2034 * If metadata is exhausted, f_bavail will be 0.
2035 *
2036 * FIXME: not accurate for mixed block groups, total and free/used are ok,
2037 * available appears slightly larger.
2038 */
2039 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2040 {
2041 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2042 struct btrfs_super_block *disk_super = fs_info->super_copy;
2043 struct list_head *head = &fs_info->space_info;
2044 struct btrfs_space_info *found;
2045 u64 total_used = 0;
2046 u64 total_free_data = 0;
2047 u64 total_free_meta = 0;
2048 int bits = dentry->d_sb->s_blocksize_bits;
2049 __be32 *fsid = (__be32 *)fs_info->fsid;
2050 unsigned factor = 1;
2051 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2052 int ret;
2053 u64 thresh = 0;
2054 int mixed = 0;
2055
2056 /*
2057 * holding chunk_muext to avoid allocating new chunks, holding
2058 * device_list_mutex to avoid the device being removed
2059 */
2060 rcu_read_lock();
2061 list_for_each_entry_rcu(found, head, list) {
2062 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2063 int i;
2064
2065 total_free_data += found->disk_total - found->disk_used;
2066 total_free_data -=
2067 btrfs_account_ro_block_groups_free_space(found);
2068
2069 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2070 if (!list_empty(&found->block_groups[i])) {
2071 switch (i) {
2072 case BTRFS_RAID_DUP:
2073 case BTRFS_RAID_RAID1:
2074 case BTRFS_RAID_RAID10:
2075 factor = 2;
2076 }
2077 }
2078 }
2079 }
2080
2081 /*
2082 * Metadata in mixed block goup profiles are accounted in data
2083 */
2084 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2085 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2086 mixed = 1;
2087 else
2088 total_free_meta += found->disk_total -
2089 found->disk_used;
2090 }
2091
2092 total_used += found->disk_used;
2093 }
2094
2095 rcu_read_unlock();
2096
2097 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2098 buf->f_blocks >>= bits;
2099 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2100
2101 /* Account global block reserve as used, it's in logical size already */
2102 spin_lock(&block_rsv->lock);
2103 buf->f_bfree -= block_rsv->size >> bits;
2104 spin_unlock(&block_rsv->lock);
2105
2106 buf->f_bavail = div_u64(total_free_data, factor);
2107 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
2108 if (ret)
2109 return ret;
2110 buf->f_bavail += div_u64(total_free_data, factor);
2111 buf->f_bavail = buf->f_bavail >> bits;
2112
2113 /*
2114 * We calculate the remaining metadata space minus global reserve. If
2115 * this is (supposedly) smaller than zero, there's no space. But this
2116 * does not hold in practice, the exhausted state happens where's still
2117 * some positive delta. So we apply some guesswork and compare the
2118 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2119 *
2120 * We probably cannot calculate the exact threshold value because this
2121 * depends on the internal reservations requested by various
2122 * operations, so some operations that consume a few metadata will
2123 * succeed even if the Avail is zero. But this is better than the other
2124 * way around.
2125 */
2126 thresh = 4 * 1024 * 1024;
2127
2128 if (!mixed && total_free_meta - thresh < block_rsv->size)
2129 buf->f_bavail = 0;
2130
2131 buf->f_type = BTRFS_SUPER_MAGIC;
2132 buf->f_bsize = dentry->d_sb->s_blocksize;
2133 buf->f_namelen = BTRFS_NAME_LEN;
2134
2135 /* We treat it as constant endianness (it doesn't matter _which_)
2136 because we want the fsid to come out the same whether mounted
2137 on a big-endian or little-endian host */
2138 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2139 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2140 /* Mask in the root object ID too, to disambiguate subvols */
2141 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2142 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2143
2144 return 0;
2145 }
2146
2147 static void btrfs_kill_super(struct super_block *sb)
2148 {
2149 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2150 kill_anon_super(sb);
2151 free_fs_info(fs_info);
2152 }
2153
2154 static struct file_system_type btrfs_fs_type = {
2155 .owner = THIS_MODULE,
2156 .name = "btrfs",
2157 .mount = btrfs_mount,
2158 .kill_sb = btrfs_kill_super,
2159 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2160 };
2161 MODULE_ALIAS_FS("btrfs");
2162
2163 static int btrfs_control_open(struct inode *inode, struct file *file)
2164 {
2165 /*
2166 * The control file's private_data is used to hold the
2167 * transaction when it is started and is used to keep
2168 * track of whether a transaction is already in progress.
2169 */
2170 file->private_data = NULL;
2171 return 0;
2172 }
2173
2174 /*
2175 * used by btrfsctl to scan devices when no FS is mounted
2176 */
2177 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2178 unsigned long arg)
2179 {
2180 struct btrfs_ioctl_vol_args *vol;
2181 struct btrfs_fs_devices *fs_devices;
2182 int ret = -ENOTTY;
2183
2184 if (!capable(CAP_SYS_ADMIN))
2185 return -EPERM;
2186
2187 vol = memdup_user((void __user *)arg, sizeof(*vol));
2188 if (IS_ERR(vol))
2189 return PTR_ERR(vol);
2190
2191 switch (cmd) {
2192 case BTRFS_IOC_SCAN_DEV:
2193 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2194 &btrfs_fs_type, &fs_devices);
2195 break;
2196 case BTRFS_IOC_DEVICES_READY:
2197 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2198 &btrfs_fs_type, &fs_devices);
2199 if (ret)
2200 break;
2201 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2202 break;
2203 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2204 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2205 break;
2206 }
2207
2208 kfree(vol);
2209 return ret;
2210 }
2211
2212 static int btrfs_freeze(struct super_block *sb)
2213 {
2214 struct btrfs_trans_handle *trans;
2215 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
2216
2217 trans = btrfs_attach_transaction_barrier(root);
2218 if (IS_ERR(trans)) {
2219 /* no transaction, don't bother */
2220 if (PTR_ERR(trans) == -ENOENT)
2221 return 0;
2222 return PTR_ERR(trans);
2223 }
2224 return btrfs_commit_transaction(trans, root);
2225 }
2226
2227 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2228 {
2229 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2230 struct btrfs_fs_devices *cur_devices;
2231 struct btrfs_device *dev, *first_dev = NULL;
2232 struct list_head *head;
2233 struct rcu_string *name;
2234
2235 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2236 cur_devices = fs_info->fs_devices;
2237 while (cur_devices) {
2238 head = &cur_devices->devices;
2239 list_for_each_entry(dev, head, dev_list) {
2240 if (dev->missing)
2241 continue;
2242 if (!dev->name)
2243 continue;
2244 if (!first_dev || dev->devid < first_dev->devid)
2245 first_dev = dev;
2246 }
2247 cur_devices = cur_devices->seed;
2248 }
2249
2250 if (first_dev) {
2251 rcu_read_lock();
2252 name = rcu_dereference(first_dev->name);
2253 seq_escape(m, name->str, " \t\n\\");
2254 rcu_read_unlock();
2255 } else {
2256 WARN_ON(1);
2257 }
2258 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2259 return 0;
2260 }
2261
2262 static const struct super_operations btrfs_super_ops = {
2263 .drop_inode = btrfs_drop_inode,
2264 .evict_inode = btrfs_evict_inode,
2265 .put_super = btrfs_put_super,
2266 .sync_fs = btrfs_sync_fs,
2267 .show_options = btrfs_show_options,
2268 .show_devname = btrfs_show_devname,
2269 .write_inode = btrfs_write_inode,
2270 .alloc_inode = btrfs_alloc_inode,
2271 .destroy_inode = btrfs_destroy_inode,
2272 .statfs = btrfs_statfs,
2273 .remount_fs = btrfs_remount,
2274 .freeze_fs = btrfs_freeze,
2275 };
2276
2277 static const struct file_operations btrfs_ctl_fops = {
2278 .open = btrfs_control_open,
2279 .unlocked_ioctl = btrfs_control_ioctl,
2280 .compat_ioctl = btrfs_control_ioctl,
2281 .owner = THIS_MODULE,
2282 .llseek = noop_llseek,
2283 };
2284
2285 static struct miscdevice btrfs_misc = {
2286 .minor = BTRFS_MINOR,
2287 .name = "btrfs-control",
2288 .fops = &btrfs_ctl_fops
2289 };
2290
2291 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2292 MODULE_ALIAS("devname:btrfs-control");
2293
2294 static int btrfs_interface_init(void)
2295 {
2296 return misc_register(&btrfs_misc);
2297 }
2298
2299 static void btrfs_interface_exit(void)
2300 {
2301 misc_deregister(&btrfs_misc);
2302 }
2303
2304 static void btrfs_print_mod_info(void)
2305 {
2306 printk(KERN_INFO "Btrfs loaded"
2307 #ifdef CONFIG_BTRFS_DEBUG
2308 ", debug=on"
2309 #endif
2310 #ifdef CONFIG_BTRFS_ASSERT
2311 ", assert=on"
2312 #endif
2313 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2314 ", integrity-checker=on"
2315 #endif
2316 "\n");
2317 }
2318
2319 static int btrfs_run_sanity_tests(void)
2320 {
2321 int ret;
2322
2323 ret = btrfs_init_test_fs();
2324 if (ret)
2325 return ret;
2326
2327 ret = btrfs_test_free_space_cache();
2328 if (ret)
2329 goto out;
2330 ret = btrfs_test_extent_buffer_operations();
2331 if (ret)
2332 goto out;
2333 ret = btrfs_test_extent_io();
2334 if (ret)
2335 goto out;
2336 ret = btrfs_test_inodes();
2337 if (ret)
2338 goto out;
2339 ret = btrfs_test_qgroups();
2340 if (ret)
2341 goto out;
2342 ret = btrfs_test_free_space_tree();
2343 out:
2344 btrfs_destroy_test_fs();
2345 return ret;
2346 }
2347
2348 static int __init init_btrfs_fs(void)
2349 {
2350 int err;
2351
2352 err = btrfs_hash_init();
2353 if (err)
2354 return err;
2355
2356 btrfs_props_init();
2357
2358 err = btrfs_init_sysfs();
2359 if (err)
2360 goto free_hash;
2361
2362 btrfs_init_compress();
2363
2364 err = btrfs_init_cachep();
2365 if (err)
2366 goto free_compress;
2367
2368 err = extent_io_init();
2369 if (err)
2370 goto free_cachep;
2371
2372 err = extent_map_init();
2373 if (err)
2374 goto free_extent_io;
2375
2376 err = ordered_data_init();
2377 if (err)
2378 goto free_extent_map;
2379
2380 err = btrfs_delayed_inode_init();
2381 if (err)
2382 goto free_ordered_data;
2383
2384 err = btrfs_auto_defrag_init();
2385 if (err)
2386 goto free_delayed_inode;
2387
2388 err = btrfs_delayed_ref_init();
2389 if (err)
2390 goto free_auto_defrag;
2391
2392 err = btrfs_prelim_ref_init();
2393 if (err)
2394 goto free_delayed_ref;
2395
2396 err = btrfs_end_io_wq_init();
2397 if (err)
2398 goto free_prelim_ref;
2399
2400 err = btrfs_interface_init();
2401 if (err)
2402 goto free_end_io_wq;
2403
2404 btrfs_init_lockdep();
2405
2406 btrfs_print_mod_info();
2407
2408 err = btrfs_run_sanity_tests();
2409 if (err)
2410 goto unregister_ioctl;
2411
2412 err = register_filesystem(&btrfs_fs_type);
2413 if (err)
2414 goto unregister_ioctl;
2415
2416 return 0;
2417
2418 unregister_ioctl:
2419 btrfs_interface_exit();
2420 free_end_io_wq:
2421 btrfs_end_io_wq_exit();
2422 free_prelim_ref:
2423 btrfs_prelim_ref_exit();
2424 free_delayed_ref:
2425 btrfs_delayed_ref_exit();
2426 free_auto_defrag:
2427 btrfs_auto_defrag_exit();
2428 free_delayed_inode:
2429 btrfs_delayed_inode_exit();
2430 free_ordered_data:
2431 ordered_data_exit();
2432 free_extent_map:
2433 extent_map_exit();
2434 free_extent_io:
2435 extent_io_exit();
2436 free_cachep:
2437 btrfs_destroy_cachep();
2438 free_compress:
2439 btrfs_exit_compress();
2440 btrfs_exit_sysfs();
2441 free_hash:
2442 btrfs_hash_exit();
2443 return err;
2444 }
2445
2446 static void __exit exit_btrfs_fs(void)
2447 {
2448 btrfs_destroy_cachep();
2449 btrfs_delayed_ref_exit();
2450 btrfs_auto_defrag_exit();
2451 btrfs_delayed_inode_exit();
2452 btrfs_prelim_ref_exit();
2453 ordered_data_exit();
2454 extent_map_exit();
2455 extent_io_exit();
2456 btrfs_interface_exit();
2457 btrfs_end_io_wq_exit();
2458 unregister_filesystem(&btrfs_fs_type);
2459 btrfs_exit_sysfs();
2460 btrfs_cleanup_fs_uuids();
2461 btrfs_exit_compress();
2462 btrfs_hash_exit();
2463 }
2464
2465 late_initcall(init_btrfs_fs);
2466 module_exit(exit_btrfs_fs)
2467
2468 MODULE_LICENSE("GPL");
(deprecated) Btrfs devel tree