projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
btrfs: synchronize incompat feature bits with sysfs files
[deliverable/linux.git] / fs / btrfs / super.c
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 "sysfs.h"
62 #include "tests/btrfs-tests.h"
63
64 #include "qgroup.h"
65 #define CREATE_TRACE_POINTS
66 #include <trace/events/btrfs.h>
67
68 static const struct super_operations btrfs_super_ops;
69 static struct file_system_type btrfs_fs_type;
70
71 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
72
73 const char *btrfs_decode_error(int errno)
74 {
75 char *errstr = "unknown";
76
77 switch (errno) {
78 case -EIO:
79 errstr = "IO failure";
80 break;
81 case -ENOMEM:
82 errstr = "Out of memory";
83 break;
84 case -EROFS:
85 errstr = "Readonly filesystem";
86 break;
87 case -EEXIST:
88 errstr = "Object already exists";
89 break;
90 case -ENOSPC:
91 errstr = "No space left";
92 break;
93 case -ENOENT:
94 errstr = "No such entry";
95 break;
96 }
97
98 return errstr;
99 }
100
101 static void save_error_info(struct btrfs_fs_info *fs_info)
102 {
103 /*
104 * today we only save the error info into ram. Long term we'll
105 * also send it down to the disk
106 */
107 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
108 }
109
110 /* btrfs handle error by forcing the filesystem readonly */
111 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
112 {
113 struct super_block *sb = fs_info->sb;
114
115 if (sb->s_flags & MS_RDONLY)
116 return;
117
118 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
119 sb->s_flags |= MS_RDONLY;
120 btrfs_info(fs_info, "forced readonly");
121 /*
122 * Note that a running device replace operation is not
123 * canceled here although there is no way to update
124 * the progress. It would add the risk of a deadlock,
125 * therefore the canceling is ommited. The only penalty
126 * is that some I/O remains active until the procedure
127 * completes. The next time when the filesystem is
128 * mounted writeable again, the device replace
129 * operation continues.
130 */
131 }
132 }
133
134 /*
135 * __btrfs_std_error decodes expected errors from the caller and
136 * invokes the approciate error response.
137 */
138 __cold
139 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
140 unsigned int line, int errno, const char *fmt, ...)
141 {
142 struct super_block *sb = fs_info->sb;
143 #ifdef CONFIG_PRINTK
144 const char *errstr;
145 #endif
146
147 /*
148 * Special case: if the error is EROFS, and we're already
149 * under MS_RDONLY, then it is safe here.
150 */
151 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
152 return;
153
154 #ifdef CONFIG_PRINTK
155 errstr = btrfs_decode_error(errno);
156 if (fmt) {
157 struct va_format vaf;
158 va_list args;
159
160 va_start(args, fmt);
161 vaf.fmt = fmt;
162 vaf.va = &args;
163
164 printk(KERN_CRIT
165 "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
166 sb->s_id, function, line, errno, errstr, &vaf);
167 va_end(args);
168 } else {
169 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
170 sb->s_id, function, line, errno, errstr);
171 }
172 #endif
173
174 /* Don't go through full error handling during mount */
175 save_error_info(fs_info);
176 if (sb->s_flags & MS_BORN)
177 btrfs_handle_error(fs_info);
178 }
179
180 #ifdef CONFIG_PRINTK
181 static const char * const logtypes[] = {
182 "emergency",
183 "alert",
184 "critical",
185 "error",
186 "warning",
187 "notice",
188 "info",
189 "debug",
190 };
191
192 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
193 {
194 struct super_block *sb = fs_info->sb;
195 char lvl[4];
196 struct va_format vaf;
197 va_list args;
198 const char *type = logtypes[4];
199 int kern_level;
200
201 va_start(args, fmt);
202
203 kern_level = printk_get_level(fmt);
204 if (kern_level) {
205 size_t size = printk_skip_level(fmt) - fmt;
206 memcpy(lvl, fmt, size);
207 lvl[size] = '\0';
208 fmt += size;
209 type = logtypes[kern_level - '0'];
210 } else
211 *lvl = '\0';
212
213 vaf.fmt = fmt;
214 vaf.va = &args;
215
216 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
217
218 va_end(args);
219 }
220 #endif
221
222 /*
223 * We only mark the transaction aborted and then set the file system read-only.
224 * This will prevent new transactions from starting or trying to join this
225 * one.
226 *
227 * This means that error recovery at the call site is limited to freeing
228 * any local memory allocations and passing the error code up without
229 * further cleanup. The transaction should complete as it normally would
230 * in the call path but will return -EIO.
231 *
232 * We'll complete the cleanup in btrfs_end_transaction and
233 * btrfs_commit_transaction.
234 */
235 __cold
236 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
237 struct btrfs_root *root, const char *function,
238 unsigned int line, int errno)
239 {
240 trans->aborted = errno;
241 /* Nothing used. The other threads that have joined this
242 * transaction may be able to continue. */
243 if (!trans->blocks_used && list_empty(&trans->new_bgs)) {
244 const char *errstr;
245
246 errstr = btrfs_decode_error(errno);
247 btrfs_warn(root->fs_info,
248 "%s:%d: Aborting unused transaction(%s).",
249 function, line, errstr);
250 return;
251 }
252 ACCESS_ONCE(trans->transaction->aborted) = errno;
253 /* Wake up anybody who may be waiting on this transaction */
254 wake_up(&root->fs_info->transaction_wait);
255 wake_up(&root->fs_info->transaction_blocked_wait);
256 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
257 }
258 /*
259 * __btrfs_panic decodes unexpected, fatal errors from the caller,
260 * issues an alert, and either panics or BUGs, depending on mount options.
261 */
262 __cold
263 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
264 unsigned int line, int errno, const char *fmt, ...)
265 {
266 char *s_id = "<unknown>";
267 const char *errstr;
268 struct va_format vaf = { .fmt = fmt };
269 va_list args;
270
271 if (fs_info)
272 s_id = fs_info->sb->s_id;
273
274 va_start(args, fmt);
275 vaf.va = &args;
276
277 errstr = btrfs_decode_error(errno);
278 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
279 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
280 s_id, function, line, &vaf, errno, errstr);
281
282 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
283 function, line, &vaf, errno, errstr);
284 va_end(args);
285 /* Caller calls BUG() */
286 }
287
288 static void btrfs_put_super(struct super_block *sb)
289 {
290 close_ctree(btrfs_sb(sb)->tree_root);
291 }
292
293 enum {
294 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
295 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
296 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
297 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
298 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
299 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
300 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
301 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
302 Opt_skip_balance, Opt_check_integrity,
303 Opt_check_integrity_including_extent_data,
304 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
305 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
306 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
307 Opt_datasum, Opt_treelog, Opt_noinode_cache,
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_flushoncommit, "flushoncommit"},
340 {Opt_noflushoncommit, "noflushoncommit"},
341 {Opt_ratio, "metadata_ratio=%d"},
342 {Opt_discard, "discard"},
343 {Opt_nodiscard, "nodiscard"},
344 {Opt_space_cache, "space_cache"},
345 {Opt_space_cache_version, "space_cache=%s"},
346 {Opt_clear_cache, "clear_cache"},
347 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
348 {Opt_enospc_debug, "enospc_debug"},
349 {Opt_noenospc_debug, "noenospc_debug"},
350 {Opt_subvolrootid, "subvolrootid=%d"},
351 {Opt_defrag, "autodefrag"},
352 {Opt_nodefrag, "noautodefrag"},
353 {Opt_inode_cache, "inode_cache"},
354 {Opt_noinode_cache, "noinode_cache"},
355 {Opt_no_space_cache, "nospace_cache"},
356 {Opt_recovery, "recovery"},
357 {Opt_skip_balance, "skip_balance"},
358 {Opt_check_integrity, "check_int"},
359 {Opt_check_integrity_including_extent_data, "check_int_data"},
360 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
361 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
362 {Opt_fatal_errors, "fatal_errors=%s"},
363 {Opt_commit_interval, "commit=%d"},
364 #ifdef CONFIG_BTRFS_DEBUG
365 {Opt_fragment_data, "fragment=data"},
366 {Opt_fragment_metadata, "fragment=metadata"},
367 {Opt_fragment_all, "fragment=all"},
368 #endif
369 {Opt_err, NULL},
370 };
371
372 /*
373 * Regular mount options parser. Everything that is needed only when
374 * reading in a new superblock is parsed here.
375 * XXX JDM: This needs to be cleaned up for remount.
376 */
377 int btrfs_parse_options(struct btrfs_root *root, char *options)
378 {
379 struct btrfs_fs_info *info = root->fs_info;
380 substring_t args[MAX_OPT_ARGS];
381 char *p, *num, *orig = NULL;
382 u64 cache_gen;
383 int intarg;
384 int ret = 0;
385 char *compress_type;
386 bool compress_force = false;
387
388 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
389 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE))
390 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
391 else if (cache_gen)
392 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
393
394 if (!options)
395 goto out;
396
397 /*
398 * strsep changes the string, duplicate it because parse_options
399 * gets called twice
400 */
401 options = kstrdup(options, GFP_NOFS);
402 if (!options)
403 return -ENOMEM;
404
405 orig = options;
406
407 while ((p = strsep(&options, ",")) != NULL) {
408 int token;
409 if (!*p)
410 continue;
411
412 token = match_token(p, tokens, args);
413 switch (token) {
414 case Opt_degraded:
415 btrfs_info(root->fs_info, "allowing degraded mounts");
416 btrfs_set_opt(info->mount_opt, DEGRADED);
417 break;
418 case Opt_subvol:
419 case Opt_subvolid:
420 case Opt_subvolrootid:
421 case Opt_device:
422 /*
423 * These are parsed by btrfs_parse_early_options
424 * and can be happily ignored here.
425 */
426 break;
427 case Opt_nodatasum:
428 btrfs_set_and_info(root, NODATASUM,
429 "setting nodatasum");
430 break;
431 case Opt_datasum:
432 if (btrfs_test_opt(root, NODATASUM)) {
433 if (btrfs_test_opt(root, NODATACOW))
434 btrfs_info(root->fs_info, "setting datasum, datacow enabled");
435 else
436 btrfs_info(root->fs_info, "setting datasum");
437 }
438 btrfs_clear_opt(info->mount_opt, NODATACOW);
439 btrfs_clear_opt(info->mount_opt, NODATASUM);
440 break;
441 case Opt_nodatacow:
442 if (!btrfs_test_opt(root, NODATACOW)) {
443 if (!btrfs_test_opt(root, COMPRESS) ||
444 !btrfs_test_opt(root, FORCE_COMPRESS)) {
445 btrfs_info(root->fs_info,
446 "setting nodatacow, compression disabled");
447 } else {
448 btrfs_info(root->fs_info, "setting nodatacow");
449 }
450 }
451 btrfs_clear_opt(info->mount_opt, COMPRESS);
452 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
453 btrfs_set_opt(info->mount_opt, NODATACOW);
454 btrfs_set_opt(info->mount_opt, NODATASUM);
455 break;
456 case Opt_datacow:
457 btrfs_clear_and_info(root, NODATACOW,
458 "setting datacow");
459 break;
460 case Opt_compress_force:
461 case Opt_compress_force_type:
462 compress_force = true;
463 /* Fallthrough */
464 case Opt_compress:
465 case Opt_compress_type:
466 if (token == Opt_compress ||
467 token == Opt_compress_force ||
468 strcmp(args[0].from, "zlib") == 0) {
469 compress_type = "zlib";
470 info->compress_type = BTRFS_COMPRESS_ZLIB;
471 btrfs_set_opt(info->mount_opt, COMPRESS);
472 btrfs_clear_opt(info->mount_opt, NODATACOW);
473 btrfs_clear_opt(info->mount_opt, NODATASUM);
474 } else if (strcmp(args[0].from, "lzo") == 0) {
475 compress_type = "lzo";
476 info->compress_type = BTRFS_COMPRESS_LZO;
477 btrfs_set_opt(info->mount_opt, COMPRESS);
478 btrfs_clear_opt(info->mount_opt, NODATACOW);
479 btrfs_clear_opt(info->mount_opt, NODATASUM);
480 btrfs_set_fs_incompat(info, COMPRESS_LZO);
481 btrfs_sysfs_feature_update(root->fs_info,
482 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO,
483 FEAT_INCOMPAT);
484 } else if (strncmp(args[0].from, "no", 2) == 0) {
485 compress_type = "no";
486 btrfs_clear_opt(info->mount_opt, COMPRESS);
487 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
488 compress_force = false;
489 } else {
490 ret = -EINVAL;
491 goto out;
492 }
493
494 if (compress_force) {
495 btrfs_set_and_info(root, FORCE_COMPRESS,
496 "force %s compression",
497 compress_type);
498 } else {
499 if (!btrfs_test_opt(root, COMPRESS))
500 btrfs_info(root->fs_info,
501 "btrfs: use %s compression",
502 compress_type);
503 /*
504 * If we remount from compress-force=xxx to
505 * compress=xxx, we need clear FORCE_COMPRESS
506 * flag, otherwise, there is no way for users
507 * to disable forcible compression separately.
508 */
509 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
510 }
511 break;
512 case Opt_ssd:
513 btrfs_set_and_info(root, SSD,
514 "use ssd allocation scheme");
515 break;
516 case Opt_ssd_spread:
517 btrfs_set_and_info(root, SSD_SPREAD,
518 "use spread ssd allocation scheme");
519 btrfs_set_opt(info->mount_opt, SSD);
520 break;
521 case Opt_nossd:
522 btrfs_set_and_info(root, NOSSD,
523 "not using ssd allocation scheme");
524 btrfs_clear_opt(info->mount_opt, SSD);
525 break;
526 case Opt_barrier:
527 btrfs_clear_and_info(root, NOBARRIER,
528 "turning on barriers");
529 break;
530 case Opt_nobarrier:
531 btrfs_set_and_info(root, NOBARRIER,
532 "turning off barriers");
533 break;
534 case Opt_thread_pool:
535 ret = match_int(&args[0], &intarg);
536 if (ret) {
537 goto out;
538 } else if (intarg > 0) {
539 info->thread_pool_size = intarg;
540 } else {
541 ret = -EINVAL;
542 goto out;
543 }
544 break;
545 case Opt_max_inline:
546 num = match_strdup(&args[0]);
547 if (num) {
548 info->max_inline = memparse(num, NULL);
549 kfree(num);
550
551 if (info->max_inline) {
552 info->max_inline = min_t(u64,
553 info->max_inline,
554 root->sectorsize);
555 }
556 btrfs_info(root->fs_info, "max_inline at %llu",
557 info->max_inline);
558 } else {
559 ret = -ENOMEM;
560 goto out;
561 }
562 break;
563 case Opt_alloc_start:
564 num = match_strdup(&args[0]);
565 if (num) {
566 mutex_lock(&info->chunk_mutex);
567 info->alloc_start = memparse(num, NULL);
568 mutex_unlock(&info->chunk_mutex);
569 kfree(num);
570 btrfs_info(root->fs_info, "allocations start at %llu",
571 info->alloc_start);
572 } else {
573 ret = -ENOMEM;
574 goto out;
575 }
576 break;
577 case Opt_acl:
578 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
579 root->fs_info->sb->s_flags |= MS_POSIXACL;
580 break;
581 #else
582 btrfs_err(root->fs_info,
583 "support for ACL not compiled in!");
584 ret = -EINVAL;
585 goto out;
586 #endif
587 case Opt_noacl:
588 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
589 break;
590 case Opt_notreelog:
591 btrfs_set_and_info(root, NOTREELOG,
592 "disabling tree log");
593 break;
594 case Opt_treelog:
595 btrfs_clear_and_info(root, NOTREELOG,
596 "enabling tree log");
597 break;
598 case Opt_flushoncommit:
599 btrfs_set_and_info(root, FLUSHONCOMMIT,
600 "turning on flush-on-commit");
601 break;
602 case Opt_noflushoncommit:
603 btrfs_clear_and_info(root, FLUSHONCOMMIT,
604 "turning off flush-on-commit");
605 break;
606 case Opt_ratio:
607 ret = match_int(&args[0], &intarg);
608 if (ret) {
609 goto out;
610 } else if (intarg >= 0) {
611 info->metadata_ratio = intarg;
612 btrfs_info(root->fs_info, "metadata ratio %d",
613 info->metadata_ratio);
614 } else {
615 ret = -EINVAL;
616 goto out;
617 }
618 break;
619 case Opt_discard:
620 btrfs_set_and_info(root, DISCARD,
621 "turning on discard");
622 break;
623 case Opt_nodiscard:
624 btrfs_clear_and_info(root, DISCARD,
625 "turning off discard");
626 break;
627 case Opt_space_cache:
628 case Opt_space_cache_version:
629 if (token == Opt_space_cache ||
630 strcmp(args[0].from, "v1") == 0) {
631 btrfs_clear_opt(root->fs_info->mount_opt,
632 FREE_SPACE_TREE);
633 btrfs_set_and_info(root, SPACE_CACHE,
634 "enabling disk space caching");
635 } else if (strcmp(args[0].from, "v2") == 0) {
636 btrfs_clear_opt(root->fs_info->mount_opt,
637 SPACE_CACHE);
638 btrfs_set_and_info(root, FREE_SPACE_TREE,
639 "enabling free space tree");
640 } else {
641 ret = -EINVAL;
642 goto out;
643 }
644 break;
645 case Opt_rescan_uuid_tree:
646 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
647 break;
648 case Opt_no_space_cache:
649 if (btrfs_test_opt(root, SPACE_CACHE)) {
650 btrfs_clear_and_info(root, SPACE_CACHE,
651 "disabling disk space caching");
652 }
653 if (btrfs_test_opt(root, FREE_SPACE_TREE)) {
654 btrfs_clear_and_info(root, FREE_SPACE_TREE,
655 "disabling free space tree");
656 }
657 break;
658 case Opt_inode_cache:
659 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
660 "enabling inode map caching");
661 break;
662 case Opt_noinode_cache:
663 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
664 "disabling inode map caching");
665 break;
666 case Opt_clear_cache:
667 btrfs_set_and_info(root, CLEAR_CACHE,
668 "force clearing of disk cache");
669 break;
670 case Opt_user_subvol_rm_allowed:
671 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
672 break;
673 case Opt_enospc_debug:
674 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
675 break;
676 case Opt_noenospc_debug:
677 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
678 break;
679 case Opt_defrag:
680 btrfs_set_and_info(root, AUTO_DEFRAG,
681 "enabling auto defrag");
682 break;
683 case Opt_nodefrag:
684 btrfs_clear_and_info(root, AUTO_DEFRAG,
685 "disabling auto defrag");
686 break;
687 case Opt_recovery:
688 btrfs_info(root->fs_info, "enabling auto recovery");
689 btrfs_set_opt(info->mount_opt, RECOVERY);
690 break;
691 case Opt_skip_balance:
692 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
693 break;
694 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
695 case Opt_check_integrity_including_extent_data:
696 btrfs_info(root->fs_info,
697 "enabling check integrity including extent data");
698 btrfs_set_opt(info->mount_opt,
699 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
700 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
701 break;
702 case Opt_check_integrity:
703 btrfs_info(root->fs_info, "enabling check integrity");
704 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
705 break;
706 case Opt_check_integrity_print_mask:
707 ret = match_int(&args[0], &intarg);
708 if (ret) {
709 goto out;
710 } else if (intarg >= 0) {
711 info->check_integrity_print_mask = intarg;
712 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
713 info->check_integrity_print_mask);
714 } else {
715 ret = -EINVAL;
716 goto out;
717 }
718 break;
719 #else
720 case Opt_check_integrity_including_extent_data:
721 case Opt_check_integrity:
722 case Opt_check_integrity_print_mask:
723 btrfs_err(root->fs_info,
724 "support for check_integrity* not compiled in!");
725 ret = -EINVAL;
726 goto out;
727 #endif
728 case Opt_fatal_errors:
729 if (strcmp(args[0].from, "panic") == 0)
730 btrfs_set_opt(info->mount_opt,
731 PANIC_ON_FATAL_ERROR);
732 else if (strcmp(args[0].from, "bug") == 0)
733 btrfs_clear_opt(info->mount_opt,
734 PANIC_ON_FATAL_ERROR);
735 else {
736 ret = -EINVAL;
737 goto out;
738 }
739 break;
740 case Opt_commit_interval:
741 intarg = 0;
742 ret = match_int(&args[0], &intarg);
743 if (ret < 0) {
744 btrfs_err(root->fs_info, "invalid commit interval");
745 ret = -EINVAL;
746 goto out;
747 }
748 if (intarg > 0) {
749 if (intarg > 300) {
750 btrfs_warn(root->fs_info, "excessive commit interval %d",
751 intarg);
752 }
753 info->commit_interval = intarg;
754 } else {
755 btrfs_info(root->fs_info, "using default commit interval %ds",
756 BTRFS_DEFAULT_COMMIT_INTERVAL);
757 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
758 }
759 break;
760 #ifdef CONFIG_BTRFS_DEBUG
761 case Opt_fragment_all:
762 btrfs_info(root->fs_info, "fragmenting all space");
763 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
764 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
765 break;
766 case Opt_fragment_metadata:
767 btrfs_info(root->fs_info, "fragmenting metadata");
768 btrfs_set_opt(info->mount_opt,
769 FRAGMENT_METADATA);
770 break;
771 case Opt_fragment_data:
772 btrfs_info(root->fs_info, "fragmenting data");
773 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
774 break;
775 #endif
776 case Opt_err:
777 btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
778 ret = -EINVAL;
779 goto out;
780 default:
781 break;
782 }
783 }
784 out:
785 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE) &&
786 !btrfs_test_opt(root, FREE_SPACE_TREE) &&
787 !btrfs_test_opt(root, CLEAR_CACHE)) {
788 btrfs_err(root->fs_info, "cannot disable free space tree");
789 ret = -EINVAL;
790
791 }
792 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
793 btrfs_info(root->fs_info, "disk space caching is enabled");
794 if (!ret && btrfs_test_opt(root, FREE_SPACE_TREE))
795 btrfs_info(root->fs_info, "using free space tree");
796 kfree(orig);
797 return ret;
798 }
799
800 /*
801 * Parse mount options that are required early in the mount process.
802 *
803 * All other options will be parsed on much later in the mount process and
804 * only when we need to allocate a new super block.
805 */
806 static int btrfs_parse_early_options(const char *options, fmode_t flags,
807 void *holder, char **subvol_name, u64 *subvol_objectid,
808 struct btrfs_fs_devices **fs_devices)
809 {
810 substring_t args[MAX_OPT_ARGS];
811 char *device_name, *opts, *orig, *p;
812 char *num = NULL;
813 int error = 0;
814
815 if (!options)
816 return 0;
817
818 /*
819 * strsep changes the string, duplicate it because parse_options
820 * gets called twice
821 */
822 opts = kstrdup(options, GFP_KERNEL);
823 if (!opts)
824 return -ENOMEM;
825 orig = opts;
826
827 while ((p = strsep(&opts, ",")) != NULL) {
828 int token;
829 if (!*p)
830 continue;
831
832 token = match_token(p, tokens, args);
833 switch (token) {
834 case Opt_subvol:
835 kfree(*subvol_name);
836 *subvol_name = match_strdup(&args[0]);
837 if (!*subvol_name) {
838 error = -ENOMEM;
839 goto out;
840 }
841 break;
842 case Opt_subvolid:
843 num = match_strdup(&args[0]);
844 if (num) {
845 *subvol_objectid = memparse(num, NULL);
846 kfree(num);
847 /* we want the original fs_tree */
848 if (!*subvol_objectid)
849 *subvol_objectid =
850 BTRFS_FS_TREE_OBJECTID;
851 } else {
852 error = -EINVAL;
853 goto out;
854 }
855 break;
856 case Opt_subvolrootid:
857 printk(KERN_WARNING
858 "BTRFS: 'subvolrootid' mount option is deprecated and has "
859 "no effect\n");
860 break;
861 case Opt_device:
862 device_name = match_strdup(&args[0]);
863 if (!device_name) {
864 error = -ENOMEM;
865 goto out;
866 }
867 error = btrfs_scan_one_device(device_name,
868 flags, holder, fs_devices);
869 kfree(device_name);
870 if (error)
871 goto out;
872 break;
873 default:
874 break;
875 }
876 }
877
878 out:
879 kfree(orig);
880 return error;
881 }
882
883 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
884 u64 subvol_objectid)
885 {
886 struct btrfs_root *root = fs_info->tree_root;
887 struct btrfs_root *fs_root;
888 struct btrfs_root_ref *root_ref;
889 struct btrfs_inode_ref *inode_ref;
890 struct btrfs_key key;
891 struct btrfs_path *path = NULL;
892 char *name = NULL, *ptr;
893 u64 dirid;
894 int len;
895 int ret;
896
897 path = btrfs_alloc_path();
898 if (!path) {
899 ret = -ENOMEM;
900 goto err;
901 }
902 path->leave_spinning = 1;
903
904 name = kmalloc(PATH_MAX, GFP_NOFS);
905 if (!name) {
906 ret = -ENOMEM;
907 goto err;
908 }
909 ptr = name + PATH_MAX - 1;
910 ptr[0] = '\0';
911
912 /*
913 * Walk up the subvolume trees in the tree of tree roots by root
914 * backrefs until we hit the top-level subvolume.
915 */
916 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
917 key.objectid = subvol_objectid;
918 key.type = BTRFS_ROOT_BACKREF_KEY;
919 key.offset = (u64)-1;
920
921 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
922 if (ret < 0) {
923 goto err;
924 } else if (ret > 0) {
925 ret = btrfs_previous_item(root, path, subvol_objectid,
926 BTRFS_ROOT_BACKREF_KEY);
927 if (ret < 0) {
928 goto err;
929 } else if (ret > 0) {
930 ret = -ENOENT;
931 goto err;
932 }
933 }
934
935 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
936 subvol_objectid = key.offset;
937
938 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
939 struct btrfs_root_ref);
940 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
941 ptr -= len + 1;
942 if (ptr < name) {
943 ret = -ENAMETOOLONG;
944 goto err;
945 }
946 read_extent_buffer(path->nodes[0], ptr + 1,
947 (unsigned long)(root_ref + 1), len);
948 ptr[0] = '/';
949 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
950 btrfs_release_path(path);
951
952 key.objectid = subvol_objectid;
953 key.type = BTRFS_ROOT_ITEM_KEY;
954 key.offset = (u64)-1;
955 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
956 if (IS_ERR(fs_root)) {
957 ret = PTR_ERR(fs_root);
958 goto err;
959 }
960
961 /*
962 * Walk up the filesystem tree by inode refs until we hit the
963 * root directory.
964 */
965 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
966 key.objectid = dirid;
967 key.type = BTRFS_INODE_REF_KEY;
968 key.offset = (u64)-1;
969
970 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
971 if (ret < 0) {
972 goto err;
973 } else if (ret > 0) {
974 ret = btrfs_previous_item(fs_root, path, dirid,
975 BTRFS_INODE_REF_KEY);
976 if (ret < 0) {
977 goto err;
978 } else if (ret > 0) {
979 ret = -ENOENT;
980 goto err;
981 }
982 }
983
984 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
985 dirid = key.offset;
986
987 inode_ref = btrfs_item_ptr(path->nodes[0],
988 path->slots[0],
989 struct btrfs_inode_ref);
990 len = btrfs_inode_ref_name_len(path->nodes[0],
991 inode_ref);
992 ptr -= len + 1;
993 if (ptr < name) {
994 ret = -ENAMETOOLONG;
995 goto err;
996 }
997 read_extent_buffer(path->nodes[0], ptr + 1,
998 (unsigned long)(inode_ref + 1), len);
999 ptr[0] = '/';
1000 btrfs_release_path(path);
1001 }
1002 }
1003
1004 btrfs_free_path(path);
1005 if (ptr == name + PATH_MAX - 1) {
1006 name[0] = '/';
1007 name[1] = '\0';
1008 } else {
1009 memmove(name, ptr, name + PATH_MAX - ptr);
1010 }
1011 return name;
1012
1013 err:
1014 btrfs_free_path(path);
1015 kfree(name);
1016 return ERR_PTR(ret);
1017 }
1018
1019 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1020 {
1021 struct btrfs_root *root = fs_info->tree_root;
1022 struct btrfs_dir_item *di;
1023 struct btrfs_path *path;
1024 struct btrfs_key location;
1025 u64 dir_id;
1026
1027 path = btrfs_alloc_path();
1028 if (!path)
1029 return -ENOMEM;
1030 path->leave_spinning = 1;
1031
1032 /*
1033 * Find the "default" dir item which points to the root item that we
1034 * will mount by default if we haven't been given a specific subvolume
1035 * to mount.
1036 */
1037 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1038 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1039 if (IS_ERR(di)) {
1040 btrfs_free_path(path);
1041 return PTR_ERR(di);
1042 }
1043 if (!di) {
1044 /*
1045 * Ok the default dir item isn't there. This is weird since
1046 * it's always been there, but don't freak out, just try and
1047 * mount the top-level subvolume.
1048 */
1049 btrfs_free_path(path);
1050 *objectid = BTRFS_FS_TREE_OBJECTID;
1051 return 0;
1052 }
1053
1054 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1055 btrfs_free_path(path);
1056 *objectid = location.objectid;
1057 return 0;
1058 }
1059
1060 static int btrfs_fill_super(struct super_block *sb,
1061 struct btrfs_fs_devices *fs_devices,
1062 void *data, int silent)
1063 {
1064 struct inode *inode;
1065 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1066 struct btrfs_key key;
1067 int err;
1068
1069 sb->s_maxbytes = MAX_LFS_FILESIZE;
1070 sb->s_magic = BTRFS_SUPER_MAGIC;
1071 sb->s_op = &btrfs_super_ops;
1072 sb->s_d_op = &btrfs_dentry_operations;
1073 sb->s_export_op = &btrfs_export_ops;
1074 sb->s_xattr = btrfs_xattr_handlers;
1075 sb->s_time_gran = 1;
1076 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1077 sb->s_flags |= MS_POSIXACL;
1078 #endif
1079 sb->s_flags |= MS_I_VERSION;
1080 sb->s_iflags |= SB_I_CGROUPWB;
1081 err = open_ctree(sb, fs_devices, (char *)data);
1082 if (err) {
1083 printk(KERN_ERR "BTRFS: open_ctree failed\n");
1084 return err;
1085 }
1086
1087 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1088 key.type = BTRFS_INODE_ITEM_KEY;
1089 key.offset = 0;
1090 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1091 if (IS_ERR(inode)) {
1092 err = PTR_ERR(inode);
1093 goto fail_close;
1094 }
1095
1096 sb->s_root = d_make_root(inode);
1097 if (!sb->s_root) {
1098 err = -ENOMEM;
1099 goto fail_close;
1100 }
1101
1102 save_mount_options(sb, data);
1103 cleancache_init_fs(sb);
1104 sb->s_flags |= MS_ACTIVE;
1105 return 0;
1106
1107 fail_close:
1108 close_ctree(fs_info->tree_root);
1109 return err;
1110 }
1111
1112 int btrfs_sync_fs(struct super_block *sb, int wait)
1113 {
1114 struct btrfs_trans_handle *trans;
1115 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1116 struct btrfs_root *root = fs_info->tree_root;
1117
1118 trace_btrfs_sync_fs(wait);
1119
1120 if (!wait) {
1121 filemap_flush(fs_info->btree_inode->i_mapping);
1122 return 0;
1123 }
1124
1125 btrfs_wait_ordered_roots(fs_info, -1);
1126
1127 trans = btrfs_attach_transaction_barrier(root);
1128 if (IS_ERR(trans)) {
1129 /* no transaction, don't bother */
1130 if (PTR_ERR(trans) == -ENOENT) {
1131 /*
1132 * Exit unless we have some pending changes
1133 * that need to go through commit
1134 */
1135 if (fs_info->pending_changes == 0)
1136 return 0;
1137 /*
1138 * A non-blocking test if the fs is frozen. We must not
1139 * start a new transaction here otherwise a deadlock
1140 * happens. The pending operations are delayed to the
1141 * next commit after thawing.
1142 */
1143 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1144 __sb_end_write(sb, SB_FREEZE_WRITE);
1145 else
1146 return 0;
1147 trans = btrfs_start_transaction(root, 0);
1148 }
1149 if (IS_ERR(trans))
1150 return PTR_ERR(trans);
1151 }
1152 return btrfs_commit_transaction(trans, root);
1153 }
1154
1155 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1156 {
1157 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1158 struct btrfs_root *root = info->tree_root;
1159 char *compress_type;
1160
1161 if (btrfs_test_opt(root, DEGRADED))
1162 seq_puts(seq, ",degraded");
1163 if (btrfs_test_opt(root, NODATASUM))
1164 seq_puts(seq, ",nodatasum");
1165 if (btrfs_test_opt(root, NODATACOW))
1166 seq_puts(seq, ",nodatacow");
1167 if (btrfs_test_opt(root, NOBARRIER))
1168 seq_puts(seq, ",nobarrier");
1169 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1170 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1171 if (info->alloc_start != 0)
1172 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1173 if (info->thread_pool_size != min_t(unsigned long,
1174 num_online_cpus() + 2, 8))
1175 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1176 if (btrfs_test_opt(root, COMPRESS)) {
1177 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1178 compress_type = "zlib";
1179 else
1180 compress_type = "lzo";
1181 if (btrfs_test_opt(root, FORCE_COMPRESS))
1182 seq_printf(seq, ",compress-force=%s", compress_type);
1183 else
1184 seq_printf(seq, ",compress=%s", compress_type);
1185 }
1186 if (btrfs_test_opt(root, NOSSD))
1187 seq_puts(seq, ",nossd");
1188 if (btrfs_test_opt(root, SSD_SPREAD))
1189 seq_puts(seq, ",ssd_spread");
1190 else if (btrfs_test_opt(root, SSD))
1191 seq_puts(seq, ",ssd");
1192 if (btrfs_test_opt(root, NOTREELOG))
1193 seq_puts(seq, ",notreelog");
1194 if (btrfs_test_opt(root, FLUSHONCOMMIT))
1195 seq_puts(seq, ",flushoncommit");
1196 if (btrfs_test_opt(root, DISCARD))
1197 seq_puts(seq, ",discard");
1198 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1199 seq_puts(seq, ",noacl");
1200 if (btrfs_test_opt(root, SPACE_CACHE))
1201 seq_puts(seq, ",space_cache");
1202 else if (btrfs_test_opt(root, FREE_SPACE_TREE))
1203 seq_puts(seq, ",space_cache=v2");
1204 else
1205 seq_puts(seq, ",nospace_cache");
1206 if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1207 seq_puts(seq, ",rescan_uuid_tree");
1208 if (btrfs_test_opt(root, CLEAR_CACHE))
1209 seq_puts(seq, ",clear_cache");
1210 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1211 seq_puts(seq, ",user_subvol_rm_allowed");
1212 if (btrfs_test_opt(root, ENOSPC_DEBUG))
1213 seq_puts(seq, ",enospc_debug");
1214 if (btrfs_test_opt(root, AUTO_DEFRAG))
1215 seq_puts(seq, ",autodefrag");
1216 if (btrfs_test_opt(root, INODE_MAP_CACHE))
1217 seq_puts(seq, ",inode_cache");
1218 if (btrfs_test_opt(root, SKIP_BALANCE))
1219 seq_puts(seq, ",skip_balance");
1220 if (btrfs_test_opt(root, RECOVERY))
1221 seq_puts(seq, ",recovery");
1222 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1223 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1224 seq_puts(seq, ",check_int_data");
1225 else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1226 seq_puts(seq, ",check_int");
1227 if (info->check_integrity_print_mask)
1228 seq_printf(seq, ",check_int_print_mask=%d",
1229 info->check_integrity_print_mask);
1230 #endif
1231 if (info->metadata_ratio)
1232 seq_printf(seq, ",metadata_ratio=%d",
1233 info->metadata_ratio);
1234 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1235 seq_puts(seq, ",fatal_errors=panic");
1236 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1237 seq_printf(seq, ",commit=%d", info->commit_interval);
1238 #ifdef CONFIG_BTRFS_DEBUG
1239 if (btrfs_test_opt(root, FRAGMENT_DATA))
1240 seq_puts(seq, ",fragment=data");
1241 if (btrfs_test_opt(root, FRAGMENT_METADATA))
1242 seq_puts(seq, ",fragment=metadata");
1243 #endif
1244 seq_printf(seq, ",subvolid=%llu",
1245 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1246 seq_puts(seq, ",subvol=");
1247 seq_dentry(seq, dentry, " \t\n\\");
1248 return 0;
1249 }
1250
1251 static int btrfs_test_super(struct super_block *s, void *data)
1252 {
1253 struct btrfs_fs_info *p = data;
1254 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1255
1256 return fs_info->fs_devices == p->fs_devices;
1257 }
1258
1259 static int btrfs_set_super(struct super_block *s, void *data)
1260 {
1261 int err = set_anon_super(s, data);
1262 if (!err)
1263 s->s_fs_info = data;
1264 return err;
1265 }
1266
1267 /*
1268 * subvolumes are identified by ino 256
1269 */
1270 static inline int is_subvolume_inode(struct inode *inode)
1271 {
1272 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1273 return 1;
1274 return 0;
1275 }
1276
1277 /*
1278 * This will add subvolid=0 to the argument string while removing any subvol=
1279 * and subvolid= arguments to make sure we get the top-level root for path
1280 * walking to the subvol we want.
1281 */
1282 static char *setup_root_args(char *args)
1283 {
1284 char *buf, *dst, *sep;
1285
1286 if (!args)
1287 return kstrdup("subvolid=0", GFP_NOFS);
1288
1289 /* The worst case is that we add ",subvolid=0" to the end. */
1290 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
1291 if (!buf)
1292 return NULL;
1293
1294 while (1) {
1295 sep = strchrnul(args, ',');
1296 if (!strstarts(args, "subvol=") &&
1297 !strstarts(args, "subvolid=")) {
1298 memcpy(dst, args, sep - args);
1299 dst += sep - args;
1300 *dst++ = ',';
1301 }
1302 if (*sep)
1303 args = sep + 1;
1304 else
1305 break;
1306 }
1307 strcpy(dst, "subvolid=0");
1308
1309 return buf;
1310 }
1311
1312 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1313 int flags, const char *device_name,
1314 char *data)
1315 {
1316 struct dentry *root;
1317 struct vfsmount *mnt = NULL;
1318 char *newargs;
1319 int ret;
1320
1321 newargs = setup_root_args(data);
1322 if (!newargs) {
1323 root = ERR_PTR(-ENOMEM);
1324 goto out;
1325 }
1326
1327 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
1328 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
1329 if (flags & MS_RDONLY) {
1330 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
1331 device_name, newargs);
1332 } else {
1333 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
1334 device_name, newargs);
1335 if (IS_ERR(mnt)) {
1336 root = ERR_CAST(mnt);
1337 mnt = NULL;
1338 goto out;
1339 }
1340
1341 down_write(&mnt->mnt_sb->s_umount);
1342 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1343 up_write(&mnt->mnt_sb->s_umount);
1344 if (ret < 0) {
1345 root = ERR_PTR(ret);
1346 goto out;
1347 }
1348 }
1349 }
1350 if (IS_ERR(mnt)) {
1351 root = ERR_CAST(mnt);
1352 mnt = NULL;
1353 goto out;
1354 }
1355
1356 if (!subvol_name) {
1357 if (!subvol_objectid) {
1358 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1359 &subvol_objectid);
1360 if (ret) {
1361 root = ERR_PTR(ret);
1362 goto out;
1363 }
1364 }
1365 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1366 subvol_objectid);
1367 if (IS_ERR(subvol_name)) {
1368 root = ERR_CAST(subvol_name);
1369 subvol_name = NULL;
1370 goto out;
1371 }
1372
1373 }
1374
1375 root = mount_subtree(mnt, subvol_name);
1376 /* mount_subtree() drops our reference on the vfsmount. */
1377 mnt = NULL;
1378
1379 if (!IS_ERR(root)) {
1380 struct super_block *s = root->d_sb;
1381 struct inode *root_inode = d_inode(root);
1382 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1383
1384 ret = 0;
1385 if (!is_subvolume_inode(root_inode)) {
1386 pr_err("BTRFS: '%s' is not a valid subvolume\n",
1387 subvol_name);
1388 ret = -EINVAL;
1389 }
1390 if (subvol_objectid && root_objectid != subvol_objectid) {
1391 /*
1392 * This will also catch a race condition where a
1393 * subvolume which was passed by ID is renamed and
1394 * another subvolume is renamed over the old location.
1395 */
1396 pr_err("BTRFS: subvol '%s' does not match subvolid %llu\n",
1397 subvol_name, subvol_objectid);
1398 ret = -EINVAL;
1399 }
1400 if (ret) {
1401 dput(root);
1402 root = ERR_PTR(ret);
1403 deactivate_locked_super(s);
1404 }
1405 }
1406
1407 out:
1408 mntput(mnt);
1409 kfree(newargs);
1410 kfree(subvol_name);
1411 return root;
1412 }
1413
1414 static int parse_security_options(char *orig_opts,
1415 struct security_mnt_opts *sec_opts)
1416 {
1417 char *secdata = NULL;
1418 int ret = 0;
1419
1420 secdata = alloc_secdata();
1421 if (!secdata)
1422 return -ENOMEM;
1423 ret = security_sb_copy_data(orig_opts, secdata);
1424 if (ret) {
1425 free_secdata(secdata);
1426 return ret;
1427 }
1428 ret = security_sb_parse_opts_str(secdata, sec_opts);
1429 free_secdata(secdata);
1430 return ret;
1431 }
1432
1433 static int setup_security_options(struct btrfs_fs_info *fs_info,
1434 struct super_block *sb,
1435 struct security_mnt_opts *sec_opts)
1436 {
1437 int ret = 0;
1438
1439 /*
1440 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1441 * is valid.
1442 */
1443 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1444 if (ret)
1445 return ret;
1446
1447 #ifdef CONFIG_SECURITY
1448 if (!fs_info->security_opts.num_mnt_opts) {
1449 /* first time security setup, copy sec_opts to fs_info */
1450 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1451 } else {
1452 /*
1453 * Since SELinux(the only one supports security_mnt_opts) does
1454 * NOT support changing context during remount/mount same sb,
1455 * This must be the same or part of the same security options,
1456 * just free it.
1457 */
1458 security_free_mnt_opts(sec_opts);
1459 }
1460 #endif
1461 return ret;
1462 }
1463
1464 /*
1465 * Find a superblock for the given device / mount point.
1466 *
1467 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1468 * for multiple device setup. Make sure to keep it in sync.
1469 */
1470 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1471 const char *device_name, void *data)
1472 {
1473 struct block_device *bdev = NULL;
1474 struct super_block *s;
1475 struct btrfs_fs_devices *fs_devices = NULL;
1476 struct btrfs_fs_info *fs_info = NULL;
1477 struct security_mnt_opts new_sec_opts;
1478 fmode_t mode = FMODE_READ;
1479 char *subvol_name = NULL;
1480 u64 subvol_objectid = 0;
1481 int error = 0;
1482
1483 if (!(flags & MS_RDONLY))
1484 mode |= FMODE_WRITE;
1485
1486 error = btrfs_parse_early_options(data, mode, fs_type,
1487 &subvol_name, &subvol_objectid,
1488 &fs_devices);
1489 if (error) {
1490 kfree(subvol_name);
1491 return ERR_PTR(error);
1492 }
1493
1494 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1495 /* mount_subvol() will free subvol_name. */
1496 return mount_subvol(subvol_name, subvol_objectid, flags,
1497 device_name, data);
1498 }
1499
1500 security_init_mnt_opts(&new_sec_opts);
1501 if (data) {
1502 error = parse_security_options(data, &new_sec_opts);
1503 if (error)
1504 return ERR_PTR(error);
1505 }
1506
1507 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1508 if (error)
1509 goto error_sec_opts;
1510
1511 /*
1512 * Setup a dummy root and fs_info for test/set super. This is because
1513 * we don't actually fill this stuff out until open_ctree, but we need
1514 * it for searching for existing supers, so this lets us do that and
1515 * then open_ctree will properly initialize everything later.
1516 */
1517 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1518 if (!fs_info) {
1519 error = -ENOMEM;
1520 goto error_sec_opts;
1521 }
1522
1523 fs_info->fs_devices = fs_devices;
1524
1525 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1526 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1527 security_init_mnt_opts(&fs_info->security_opts);
1528 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1529 error = -ENOMEM;
1530 goto error_fs_info;
1531 }
1532
1533 error = btrfs_open_devices(fs_devices, mode, fs_type);
1534 if (error)
1535 goto error_fs_info;
1536
1537 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1538 error = -EACCES;
1539 goto error_close_devices;
1540 }
1541
1542 bdev = fs_devices->latest_bdev;
1543 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1544 fs_info);
1545 if (IS_ERR(s)) {
1546 error = PTR_ERR(s);
1547 goto error_close_devices;
1548 }
1549
1550 if (s->s_root) {
1551 btrfs_close_devices(fs_devices);
1552 free_fs_info(fs_info);
1553 if ((flags ^ s->s_flags) & MS_RDONLY)
1554 error = -EBUSY;
1555 } else {
1556 char b[BDEVNAME_SIZE];
1557
1558 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1559 btrfs_sb(s)->bdev_holder = fs_type;
1560 error = btrfs_fill_super(s, fs_devices, data,
1561 flags & MS_SILENT ? 1 : 0);
1562 }
1563 if (error) {
1564 deactivate_locked_super(s);
1565 goto error_sec_opts;
1566 }
1567
1568 fs_info = btrfs_sb(s);
1569 error = setup_security_options(fs_info, s, &new_sec_opts);
1570 if (error) {
1571 deactivate_locked_super(s);
1572 goto error_sec_opts;
1573 }
1574
1575 return dget(s->s_root);
1576
1577 error_close_devices:
1578 btrfs_close_devices(fs_devices);
1579 error_fs_info:
1580 free_fs_info(fs_info);
1581 error_sec_opts:
1582 security_free_mnt_opts(&new_sec_opts);
1583 return ERR_PTR(error);
1584 }
1585
1586 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1587 int new_pool_size, int old_pool_size)
1588 {
1589 if (new_pool_size == old_pool_size)
1590 return;
1591
1592 fs_info->thread_pool_size = new_pool_size;
1593
1594 btrfs_info(fs_info, "resize thread pool %d -> %d",
1595 old_pool_size, new_pool_size);
1596
1597 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1598 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1599 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1600 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1601 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1602 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1603 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1604 new_pool_size);
1605 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1606 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1607 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1608 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1609 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1610 new_pool_size);
1611 }
1612
1613 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1614 {
1615 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1616 }
1617
1618 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1619 unsigned long old_opts, int flags)
1620 {
1621 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1622 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1623 (flags & MS_RDONLY))) {
1624 /* wait for any defraggers to finish */
1625 wait_event(fs_info->transaction_wait,
1626 (atomic_read(&fs_info->defrag_running) == 0));
1627 if (flags & MS_RDONLY)
1628 sync_filesystem(fs_info->sb);
1629 }
1630 }
1631
1632 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1633 unsigned long old_opts)
1634 {
1635 /*
1636 * We need cleanup all defragable inodes if the autodefragment is
1637 * close or the fs is R/O.
1638 */
1639 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1640 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1641 (fs_info->sb->s_flags & MS_RDONLY))) {
1642 btrfs_cleanup_defrag_inodes(fs_info);
1643 }
1644
1645 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1646 }
1647
1648 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1649 {
1650 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1651 struct btrfs_root *root = fs_info->tree_root;
1652 unsigned old_flags = sb->s_flags;
1653 unsigned long old_opts = fs_info->mount_opt;
1654 unsigned long old_compress_type = fs_info->compress_type;
1655 u64 old_max_inline = fs_info->max_inline;
1656 u64 old_alloc_start = fs_info->alloc_start;
1657 int old_thread_pool_size = fs_info->thread_pool_size;
1658 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1659 int ret;
1660
1661 sync_filesystem(sb);
1662 btrfs_remount_prepare(fs_info);
1663
1664 if (data) {
1665 struct security_mnt_opts new_sec_opts;
1666
1667 security_init_mnt_opts(&new_sec_opts);
1668 ret = parse_security_options(data, &new_sec_opts);
1669 if (ret)
1670 goto restore;
1671 ret = setup_security_options(fs_info, sb,
1672 &new_sec_opts);
1673 if (ret) {
1674 security_free_mnt_opts(&new_sec_opts);
1675 goto restore;
1676 }
1677 }
1678
1679 ret = btrfs_parse_options(root, data);
1680 if (ret) {
1681 ret = -EINVAL;
1682 goto restore;
1683 }
1684
1685 btrfs_remount_begin(fs_info, old_opts, *flags);
1686 btrfs_resize_thread_pool(fs_info,
1687 fs_info->thread_pool_size, old_thread_pool_size);
1688
1689 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1690 goto out;
1691
1692 if (*flags & MS_RDONLY) {
1693 /*
1694 * this also happens on 'umount -rf' or on shutdown, when
1695 * the filesystem is busy.
1696 */
1697 cancel_work_sync(&fs_info->async_reclaim_work);
1698
1699 /* wait for the uuid_scan task to finish */
1700 down(&fs_info->uuid_tree_rescan_sem);
1701 /* avoid complains from lockdep et al. */
1702 up(&fs_info->uuid_tree_rescan_sem);
1703
1704 sb->s_flags |= MS_RDONLY;
1705
1706 /*
1707 * Setting MS_RDONLY will put the cleaner thread to
1708 * sleep at the next loop if it's already active.
1709 * If it's already asleep, we'll leave unused block
1710 * groups on disk until we're mounted read-write again
1711 * unless we clean them up here.
1712 */
1713 btrfs_delete_unused_bgs(fs_info);
1714
1715 btrfs_dev_replace_suspend_for_unmount(fs_info);
1716 btrfs_scrub_cancel(fs_info);
1717 btrfs_pause_balance(fs_info);
1718
1719 ret = btrfs_commit_super(root);
1720 if (ret)
1721 goto restore;
1722 } else {
1723 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1724 btrfs_err(fs_info,
1725 "Remounting read-write after error is not allowed");
1726 ret = -EINVAL;
1727 goto restore;
1728 }
1729 if (fs_info->fs_devices->rw_devices == 0) {
1730 ret = -EACCES;
1731 goto restore;
1732 }
1733
1734 if (fs_info->fs_devices->missing_devices >
1735 fs_info->num_tolerated_disk_barrier_failures &&
1736 !(*flags & MS_RDONLY)) {
1737 btrfs_warn(fs_info,
1738 "too many missing devices, writeable remount is not allowed");
1739 ret = -EACCES;
1740 goto restore;
1741 }
1742
1743 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1744 ret = -EINVAL;
1745 goto restore;
1746 }
1747
1748 ret = btrfs_cleanup_fs_roots(fs_info);
1749 if (ret)
1750 goto restore;
1751
1752 /* recover relocation */
1753 mutex_lock(&fs_info->cleaner_mutex);
1754 ret = btrfs_recover_relocation(root);
1755 mutex_unlock(&fs_info->cleaner_mutex);
1756 if (ret)
1757 goto restore;
1758
1759 ret = btrfs_resume_balance_async(fs_info);
1760 if (ret)
1761 goto restore;
1762
1763 ret = btrfs_resume_dev_replace_async(fs_info);
1764 if (ret) {
1765 btrfs_warn(fs_info, "failed to resume dev_replace");
1766 goto restore;
1767 }
1768
1769 if (!fs_info->uuid_root) {
1770 btrfs_info(fs_info, "creating UUID tree");
1771 ret = btrfs_create_uuid_tree(fs_info);
1772 if (ret) {
1773 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1774 goto restore;
1775 }
1776 }
1777 sb->s_flags &= ~MS_RDONLY;
1778 }
1779 out:
1780 wake_up_process(fs_info->transaction_kthread);
1781 btrfs_remount_cleanup(fs_info, old_opts);
1782 return 0;
1783
1784 restore:
1785 /* We've hit an error - don't reset MS_RDONLY */
1786 if (sb->s_flags & MS_RDONLY)
1787 old_flags |= MS_RDONLY;
1788 sb->s_flags = old_flags;
1789 fs_info->mount_opt = old_opts;
1790 fs_info->compress_type = old_compress_type;
1791 fs_info->max_inline = old_max_inline;
1792 mutex_lock(&fs_info->chunk_mutex);
1793 fs_info->alloc_start = old_alloc_start;
1794 mutex_unlock(&fs_info->chunk_mutex);
1795 btrfs_resize_thread_pool(fs_info,
1796 old_thread_pool_size, fs_info->thread_pool_size);
1797 fs_info->metadata_ratio = old_metadata_ratio;
1798 btrfs_remount_cleanup(fs_info, old_opts);
1799 return ret;
1800 }
1801
1802 /* Used to sort the devices by max_avail(descending sort) */
1803 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1804 const void *dev_info2)
1805 {
1806 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1807 ((struct btrfs_device_info *)dev_info2)->max_avail)
1808 return -1;
1809 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1810 ((struct btrfs_device_info *)dev_info2)->max_avail)
1811 return 1;
1812 else
1813 return 0;
1814 }
1815
1816 /*
1817 * sort the devices by max_avail, in which max free extent size of each device
1818 * is stored.(Descending Sort)
1819 */
1820 static inline void btrfs_descending_sort_devices(
1821 struct btrfs_device_info *devices,
1822 size_t nr_devices)
1823 {
1824 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1825 btrfs_cmp_device_free_bytes, NULL);
1826 }
1827
1828 /*
1829 * The helper to calc the free space on the devices that can be used to store
1830 * file data.
1831 */
1832 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1833 {
1834 struct btrfs_fs_info *fs_info = root->fs_info;
1835 struct btrfs_device_info *devices_info;
1836 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1837 struct btrfs_device *device;
1838 u64 skip_space;
1839 u64 type;
1840 u64 avail_space;
1841 u64 used_space;
1842 u64 min_stripe_size;
1843 int min_stripes = 1, num_stripes = 1;
1844 int i = 0, nr_devices;
1845 int ret;
1846
1847 /*
1848 * We aren't under the device list lock, so this is racey-ish, but good
1849 * enough for our purposes.
1850 */
1851 nr_devices = fs_info->fs_devices->open_devices;
1852 if (!nr_devices) {
1853 smp_mb();
1854 nr_devices = fs_info->fs_devices->open_devices;
1855 ASSERT(nr_devices);
1856 if (!nr_devices) {
1857 *free_bytes = 0;
1858 return 0;
1859 }
1860 }
1861
1862 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1863 GFP_NOFS);
1864 if (!devices_info)
1865 return -ENOMEM;
1866
1867 /* calc min stripe number for data space alloction */
1868 type = btrfs_get_alloc_profile(root, 1);
1869 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1870 min_stripes = 2;
1871 num_stripes = nr_devices;
1872 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1873 min_stripes = 2;
1874 num_stripes = 2;
1875 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1876 min_stripes = 4;
1877 num_stripes = 4;
1878 }
1879
1880 if (type & BTRFS_BLOCK_GROUP_DUP)
1881 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1882 else
1883 min_stripe_size = BTRFS_STRIPE_LEN;
1884
1885 if (fs_info->alloc_start)
1886 mutex_lock(&fs_devices->device_list_mutex);
1887 rcu_read_lock();
1888 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1889 if (!device->in_fs_metadata || !device->bdev ||
1890 device->is_tgtdev_for_dev_replace)
1891 continue;
1892
1893 if (i >= nr_devices)
1894 break;
1895
1896 avail_space = device->total_bytes - device->bytes_used;
1897
1898 /* align with stripe_len */
1899 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1900 avail_space *= BTRFS_STRIPE_LEN;
1901
1902 /*
1903 * In order to avoid overwritting the superblock on the drive,
1904 * btrfs starts at an offset of at least 1MB when doing chunk
1905 * allocation.
1906 */
1907 skip_space = SZ_1M;
1908
1909 /* user can set the offset in fs_info->alloc_start. */
1910 if (fs_info->alloc_start &&
1911 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1912 device->total_bytes) {
1913 rcu_read_unlock();
1914 skip_space = max(fs_info->alloc_start, skip_space);
1915
1916 /*
1917 * btrfs can not use the free space in
1918 * [0, skip_space - 1], we must subtract it from the
1919 * total. In order to implement it, we account the used
1920 * space in this range first.
1921 */
1922 ret = btrfs_account_dev_extents_size(device, 0,
1923 skip_space - 1,
1924 &used_space);
1925 if (ret) {
1926 kfree(devices_info);
1927 mutex_unlock(&fs_devices->device_list_mutex);
1928 return ret;
1929 }
1930
1931 rcu_read_lock();
1932
1933 /* calc the free space in [0, skip_space - 1] */
1934 skip_space -= used_space;
1935 }
1936
1937 /*
1938 * we can use the free space in [0, skip_space - 1], subtract
1939 * it from the total.
1940 */
1941 if (avail_space && avail_space >= skip_space)
1942 avail_space -= skip_space;
1943 else
1944 avail_space = 0;
1945
1946 if (avail_space < min_stripe_size)
1947 continue;
1948
1949 devices_info[i].dev = device;
1950 devices_info[i].max_avail = avail_space;
1951
1952 i++;
1953 }
1954 rcu_read_unlock();
1955 if (fs_info->alloc_start)
1956 mutex_unlock(&fs_devices->device_list_mutex);
1957
1958 nr_devices = i;
1959
1960 btrfs_descending_sort_devices(devices_info, nr_devices);
1961
1962 i = nr_devices - 1;
1963 avail_space = 0;
1964 while (nr_devices >= min_stripes) {
1965 if (num_stripes > nr_devices)
1966 num_stripes = nr_devices;
1967
1968 if (devices_info[i].max_avail >= min_stripe_size) {
1969 int j;
1970 u64 alloc_size;
1971
1972 avail_space += devices_info[i].max_avail * num_stripes;
1973 alloc_size = devices_info[i].max_avail;
1974 for (j = i + 1 - num_stripes; j <= i; j++)
1975 devices_info[j].max_avail -= alloc_size;
1976 }
1977 i--;
1978 nr_devices--;
1979 }
1980
1981 kfree(devices_info);
1982 *free_bytes = avail_space;
1983 return 0;
1984 }
1985
1986 /*
1987 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1988 *
1989 * If there's a redundant raid level at DATA block groups, use the respective
1990 * multiplier to scale the sizes.
1991 *
1992 * Unused device space usage is based on simulating the chunk allocator
1993 * algorithm that respects the device sizes, order of allocations and the
1994 * 'alloc_start' value, this is a close approximation of the actual use but
1995 * there are other factors that may change the result (like a new metadata
1996 * chunk).
1997 *
1998 * If metadata is exhausted, f_bavail will be 0.
1999 *
2000 * FIXME: not accurate for mixed block groups, total and free/used are ok,
2001 * available appears slightly larger.
2002 */
2003 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2004 {
2005 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2006 struct btrfs_super_block *disk_super = fs_info->super_copy;
2007 struct list_head *head = &fs_info->space_info;
2008 struct btrfs_space_info *found;
2009 u64 total_used = 0;
2010 u64 total_free_data = 0;
2011 u64 total_free_meta = 0;
2012 int bits = dentry->d_sb->s_blocksize_bits;
2013 __be32 *fsid = (__be32 *)fs_info->fsid;
2014 unsigned factor = 1;
2015 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2016 int ret;
2017 u64 thresh = 0;
2018
2019 /*
2020 * holding chunk_muext to avoid allocating new chunks, holding
2021 * device_list_mutex to avoid the device being removed
2022 */
2023 rcu_read_lock();
2024 list_for_each_entry_rcu(found, head, list) {
2025 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2026 int i;
2027
2028 total_free_data += found->disk_total - found->disk_used;
2029 total_free_data -=
2030 btrfs_account_ro_block_groups_free_space(found);
2031
2032 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2033 if (!list_empty(&found->block_groups[i])) {
2034 switch (i) {
2035 case BTRFS_RAID_DUP:
2036 case BTRFS_RAID_RAID1:
2037 case BTRFS_RAID_RAID10:
2038 factor = 2;
2039 }
2040 }
2041 }
2042 }
2043 if (found->flags & BTRFS_BLOCK_GROUP_METADATA)
2044 total_free_meta += found->disk_total - found->disk_used;
2045
2046 total_used += found->disk_used;
2047 }
2048
2049 rcu_read_unlock();
2050
2051 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2052 buf->f_blocks >>= bits;
2053 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2054
2055 /* Account global block reserve as used, it's in logical size already */
2056 spin_lock(&block_rsv->lock);
2057 buf->f_bfree -= block_rsv->size >> bits;
2058 spin_unlock(&block_rsv->lock);
2059
2060 buf->f_bavail = div_u64(total_free_data, factor);
2061 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
2062 if (ret)
2063 return ret;
2064 buf->f_bavail += div_u64(total_free_data, factor);
2065 buf->f_bavail = buf->f_bavail >> bits;
2066
2067 /*
2068 * We calculate the remaining metadata space minus global reserve. If
2069 * this is (supposedly) smaller than zero, there's no space. But this
2070 * does not hold in practice, the exhausted state happens where's still
2071 * some positive delta. So we apply some guesswork and compare the
2072 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2073 *
2074 * We probably cannot calculate the exact threshold value because this
2075 * depends on the internal reservations requested by various
2076 * operations, so some operations that consume a few metadata will
2077 * succeed even if the Avail is zero. But this is better than the other
2078 * way around.
2079 */
2080 thresh = 4 * 1024 * 1024;
2081
2082 if (total_free_meta - thresh < block_rsv->size)
2083 buf->f_bavail = 0;
2084
2085 buf->f_type = BTRFS_SUPER_MAGIC;
2086 buf->f_bsize = dentry->d_sb->s_blocksize;
2087 buf->f_namelen = BTRFS_NAME_LEN;
2088
2089 /* We treat it as constant endianness (it doesn't matter _which_)
2090 because we want the fsid to come out the same whether mounted
2091 on a big-endian or little-endian host */
2092 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2093 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2094 /* Mask in the root object ID too, to disambiguate subvols */
2095 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2096 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2097
2098 return 0;
2099 }
2100
2101 static void btrfs_kill_super(struct super_block *sb)
2102 {
2103 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2104 kill_anon_super(sb);
2105 free_fs_info(fs_info);
2106 }
2107
2108 static struct file_system_type btrfs_fs_type = {
2109 .owner = THIS_MODULE,
2110 .name = "btrfs",
2111 .mount = btrfs_mount,
2112 .kill_sb = btrfs_kill_super,
2113 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2114 };
2115 MODULE_ALIAS_FS("btrfs");
2116
2117 static int btrfs_control_open(struct inode *inode, struct file *file)
2118 {
2119 /*
2120 * The control file's private_data is used to hold the
2121 * transaction when it is started and is used to keep
2122 * track of whether a transaction is already in progress.
2123 */
2124 file->private_data = NULL;
2125 return 0;
2126 }
2127
2128 /*
2129 * used by btrfsctl to scan devices when no FS is mounted
2130 */
2131 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2132 unsigned long arg)
2133 {
2134 struct btrfs_ioctl_vol_args *vol;
2135 struct btrfs_fs_devices *fs_devices;
2136 int ret = -ENOTTY;
2137
2138 if (!capable(CAP_SYS_ADMIN))
2139 return -EPERM;
2140
2141 vol = memdup_user((void __user *)arg, sizeof(*vol));
2142 if (IS_ERR(vol))
2143 return PTR_ERR(vol);
2144
2145 switch (cmd) {
2146 case BTRFS_IOC_SCAN_DEV:
2147 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2148 &btrfs_fs_type, &fs_devices);
2149 break;
2150 case BTRFS_IOC_DEVICES_READY:
2151 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2152 &btrfs_fs_type, &fs_devices);
2153 if (ret)
2154 break;
2155 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2156 break;
2157 }
2158
2159 kfree(vol);
2160 return ret;
2161 }
2162
2163 static int btrfs_freeze(struct super_block *sb)
2164 {
2165 struct btrfs_trans_handle *trans;
2166 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
2167
2168 trans = btrfs_attach_transaction_barrier(root);
2169 if (IS_ERR(trans)) {
2170 /* no transaction, don't bother */
2171 if (PTR_ERR(trans) == -ENOENT)
2172 return 0;
2173 return PTR_ERR(trans);
2174 }
2175 return btrfs_commit_transaction(trans, root);
2176 }
2177
2178 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2179 {
2180 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2181 struct btrfs_fs_devices *cur_devices;
2182 struct btrfs_device *dev, *first_dev = NULL;
2183 struct list_head *head;
2184 struct rcu_string *name;
2185
2186 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2187 cur_devices = fs_info->fs_devices;
2188 while (cur_devices) {
2189 head = &cur_devices->devices;
2190 list_for_each_entry(dev, head, dev_list) {
2191 if (dev->missing)
2192 continue;
2193 if (!dev->name)
2194 continue;
2195 if (!first_dev || dev->devid < first_dev->devid)
2196 first_dev = dev;
2197 }
2198 cur_devices = cur_devices->seed;
2199 }
2200
2201 if (first_dev) {
2202 rcu_read_lock();
2203 name = rcu_dereference(first_dev->name);
2204 seq_escape(m, name->str, " \t\n\\");
2205 rcu_read_unlock();
2206 } else {
2207 WARN_ON(1);
2208 }
2209 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2210 return 0;
2211 }
2212
2213 static const struct super_operations btrfs_super_ops = {
2214 .drop_inode = btrfs_drop_inode,
2215 .evict_inode = btrfs_evict_inode,
2216 .put_super = btrfs_put_super,
2217 .sync_fs = btrfs_sync_fs,
2218 .show_options = btrfs_show_options,
2219 .show_devname = btrfs_show_devname,
2220 .write_inode = btrfs_write_inode,
2221 .alloc_inode = btrfs_alloc_inode,
2222 .destroy_inode = btrfs_destroy_inode,
2223 .statfs = btrfs_statfs,
2224 .remount_fs = btrfs_remount,
2225 .freeze_fs = btrfs_freeze,
2226 };
2227
2228 static const struct file_operations btrfs_ctl_fops = {
2229 .open = btrfs_control_open,
2230 .unlocked_ioctl = btrfs_control_ioctl,
2231 .compat_ioctl = btrfs_control_ioctl,
2232 .owner = THIS_MODULE,
2233 .llseek = noop_llseek,
2234 };
2235
2236 static struct miscdevice btrfs_misc = {
2237 .minor = BTRFS_MINOR,
2238 .name = "btrfs-control",
2239 .fops = &btrfs_ctl_fops
2240 };
2241
2242 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2243 MODULE_ALIAS("devname:btrfs-control");
2244
2245 static int btrfs_interface_init(void)
2246 {
2247 return misc_register(&btrfs_misc);
2248 }
2249
2250 static void btrfs_interface_exit(void)
2251 {
2252 misc_deregister(&btrfs_misc);
2253 }
2254
2255 static void btrfs_print_info(void)
2256 {
2257 printk(KERN_INFO "Btrfs loaded"
2258 #ifdef CONFIG_BTRFS_DEBUG
2259 ", debug=on"
2260 #endif
2261 #ifdef CONFIG_BTRFS_ASSERT
2262 ", assert=on"
2263 #endif
2264 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2265 ", integrity-checker=on"
2266 #endif
2267 "\n");
2268 }
2269
2270 static int btrfs_run_sanity_tests(void)
2271 {
2272 int ret;
2273
2274 ret = btrfs_init_test_fs();
2275 if (ret)
2276 return ret;
2277
2278 ret = btrfs_test_free_space_cache();
2279 if (ret)
2280 goto out;
2281 ret = btrfs_test_extent_buffer_operations();
2282 if (ret)
2283 goto out;
2284 ret = btrfs_test_extent_io();
2285 if (ret)
2286 goto out;
2287 ret = btrfs_test_inodes();
2288 if (ret)
2289 goto out;
2290 ret = btrfs_test_qgroups();
2291 if (ret)
2292 goto out;
2293 ret = btrfs_test_free_space_tree();
2294 out:
2295 btrfs_destroy_test_fs();
2296 return ret;
2297 }
2298
2299 static int __init init_btrfs_fs(void)
2300 {
2301 int err;
2302
2303 err = btrfs_hash_init();
2304 if (err)
2305 return err;
2306
2307 btrfs_props_init();
2308
2309 err = btrfs_init_sysfs();
2310 if (err)
2311 goto free_hash;
2312
2313 btrfs_init_compress();
2314
2315 err = btrfs_init_cachep();
2316 if (err)
2317 goto free_compress;
2318
2319 err = extent_io_init();
2320 if (err)
2321 goto free_cachep;
2322
2323 err = extent_map_init();
2324 if (err)
2325 goto free_extent_io;
2326
2327 err = ordered_data_init();
2328 if (err)
2329 goto free_extent_map;
2330
2331 err = btrfs_delayed_inode_init();
2332 if (err)
2333 goto free_ordered_data;
2334
2335 err = btrfs_auto_defrag_init();
2336 if (err)
2337 goto free_delayed_inode;
2338
2339 err = btrfs_delayed_ref_init();
2340 if (err)
2341 goto free_auto_defrag;
2342
2343 err = btrfs_prelim_ref_init();
2344 if (err)
2345 goto free_delayed_ref;
2346
2347 err = btrfs_end_io_wq_init();
2348 if (err)
2349 goto free_prelim_ref;
2350
2351 err = btrfs_interface_init();
2352 if (err)
2353 goto free_end_io_wq;
2354
2355 btrfs_init_lockdep();
2356
2357 btrfs_print_info();
2358
2359 err = btrfs_run_sanity_tests();
2360 if (err)
2361 goto unregister_ioctl;
2362
2363 err = register_filesystem(&btrfs_fs_type);
2364 if (err)
2365 goto unregister_ioctl;
2366
2367 return 0;
2368
2369 unregister_ioctl:
2370 btrfs_interface_exit();
2371 free_end_io_wq:
2372 btrfs_end_io_wq_exit();
2373 free_prelim_ref:
2374 btrfs_prelim_ref_exit();
2375 free_delayed_ref:
2376 btrfs_delayed_ref_exit();
2377 free_auto_defrag:
2378 btrfs_auto_defrag_exit();
2379 free_delayed_inode:
2380 btrfs_delayed_inode_exit();
2381 free_ordered_data:
2382 ordered_data_exit();
2383 free_extent_map:
2384 extent_map_exit();
2385 free_extent_io:
2386 extent_io_exit();
2387 free_cachep:
2388 btrfs_destroy_cachep();
2389 free_compress:
2390 btrfs_exit_compress();
2391 btrfs_exit_sysfs();
2392 free_hash:
2393 btrfs_hash_exit();
2394 return err;
2395 }
2396
2397 static void __exit exit_btrfs_fs(void)
2398 {
2399 btrfs_destroy_cachep();
2400 btrfs_delayed_ref_exit();
2401 btrfs_auto_defrag_exit();
2402 btrfs_delayed_inode_exit();
2403 btrfs_prelim_ref_exit();
2404 ordered_data_exit();
2405 extent_map_exit();
2406 extent_io_exit();
2407 btrfs_interface_exit();
2408 btrfs_end_io_wq_exit();
2409 unregister_filesystem(&btrfs_fs_type);
2410 btrfs_exit_sysfs();
2411 btrfs_cleanup_fs_uuids();
2412 btrfs_exit_compress();
2413 btrfs_hash_exit();
2414 }
2415
2416 late_initcall(init_btrfs_fs);
2417 module_exit(exit_btrfs_fs)
2418
2419 MODULE_LICENSE("GPL");
Linux kernel - Deliverables
This page took 0.134623 seconds and 5 git commands to generate.