Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (C) 2007 Oracle. All rights reserved.
4 */
5
6#include <linux/blkdev.h>
7#include <linux/module.h>
8#include <linux/fs.h>
9#include <linux/pagemap.h>
10#include <linux/highmem.h>
11#include <linux/time.h>
12#include <linux/init.h>
13#include <linux/seq_file.h>
14#include <linux/string.h>
15#include <linux/backing-dev.h>
16#include <linux/mount.h>
17#include <linux/writeback.h>
18#include <linux/statfs.h>
19#include <linux/compat.h>
20#include <linux/parser.h>
21#include <linux/ctype.h>
22#include <linux/namei.h>
23#include <linux/miscdevice.h>
24#include <linux/magic.h>
25#include <linux/slab.h>
26#include <linux/ratelimit.h>
27#include <linux/crc32c.h>
28#include <linux/btrfs.h>
29#include "messages.h"
30#include "delayed-inode.h"
31#include "ctree.h"
32#include "disk-io.h"
33#include "transaction.h"
34#include "btrfs_inode.h"
35#include "print-tree.h"
36#include "props.h"
37#include "xattr.h"
38#include "bio.h"
39#include "export.h"
40#include "compression.h"
41#include "rcu-string.h"
42#include "dev-replace.h"
43#include "free-space-cache.h"
44#include "backref.h"
45#include "space-info.h"
46#include "sysfs.h"
47#include "zoned.h"
48#include "tests/btrfs-tests.h"
49#include "block-group.h"
50#include "discard.h"
51#include "qgroup.h"
52#include "raid56.h"
53#include "fs.h"
54#include "accessors.h"
55#include "defrag.h"
56#include "dir-item.h"
57#include "ioctl.h"
58#include "scrub.h"
59#include "verity.h"
60#include "super.h"
61#define CREATE_TRACE_POINTS
62#include <trace/events/btrfs.h>
63
64static const struct super_operations btrfs_super_ops;
65
66/*
67 * Types for mounting the default subvolume and a subvolume explicitly
68 * requested by subvol=/path. That way the callchain is straightforward and we
69 * don't have to play tricks with the mount options and recursive calls to
70 * btrfs_mount.
71 *
72 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
73 */
74static struct file_system_type btrfs_fs_type;
75static struct file_system_type btrfs_root_fs_type;
76
77static int btrfs_remount(struct super_block *sb, int *flags, char *data);
78
79static void btrfs_put_super(struct super_block *sb)
80{
81 close_ctree(btrfs_sb(sb));
82}
83
84enum {
85 Opt_acl, Opt_noacl,
86 Opt_clear_cache,
87 Opt_commit_interval,
88 Opt_compress,
89 Opt_compress_force,
90 Opt_compress_force_type,
91 Opt_compress_type,
92 Opt_degraded,
93 Opt_device,
94 Opt_fatal_errors,
95 Opt_flushoncommit, Opt_noflushoncommit,
96 Opt_max_inline,
97 Opt_barrier, Opt_nobarrier,
98 Opt_datacow, Opt_nodatacow,
99 Opt_datasum, Opt_nodatasum,
100 Opt_defrag, Opt_nodefrag,
101 Opt_discard, Opt_nodiscard,
102 Opt_discard_mode,
103 Opt_norecovery,
104 Opt_ratio,
105 Opt_rescan_uuid_tree,
106 Opt_skip_balance,
107 Opt_space_cache, Opt_no_space_cache,
108 Opt_space_cache_version,
109 Opt_ssd, Opt_nossd,
110 Opt_ssd_spread, Opt_nossd_spread,
111 Opt_subvol,
112 Opt_subvol_empty,
113 Opt_subvolid,
114 Opt_thread_pool,
115 Opt_treelog, Opt_notreelog,
116 Opt_user_subvol_rm_allowed,
117
118 /* Rescue options */
119 Opt_rescue,
120 Opt_usebackuproot,
121 Opt_nologreplay,
122 Opt_ignorebadroots,
123 Opt_ignoredatacsums,
124 Opt_rescue_all,
125
126 /* Deprecated options */
127 Opt_recovery,
128 Opt_inode_cache, Opt_noinode_cache,
129
130 /* Debugging options */
131 Opt_check_integrity,
132 Opt_check_integrity_including_extent_data,
133 Opt_check_integrity_print_mask,
134 Opt_enospc_debug, Opt_noenospc_debug,
135#ifdef CONFIG_BTRFS_DEBUG
136 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
137#endif
138#ifdef CONFIG_BTRFS_FS_REF_VERIFY
139 Opt_ref_verify,
140#endif
141 Opt_err,
142};
143
144static const match_table_t tokens = {
145 {Opt_acl, "acl"},
146 {Opt_noacl, "noacl"},
147 {Opt_clear_cache, "clear_cache"},
148 {Opt_commit_interval, "commit=%u"},
149 {Opt_compress, "compress"},
150 {Opt_compress_type, "compress=%s"},
151 {Opt_compress_force, "compress-force"},
152 {Opt_compress_force_type, "compress-force=%s"},
153 {Opt_degraded, "degraded"},
154 {Opt_device, "device=%s"},
155 {Opt_fatal_errors, "fatal_errors=%s"},
156 {Opt_flushoncommit, "flushoncommit"},
157 {Opt_noflushoncommit, "noflushoncommit"},
158 {Opt_inode_cache, "inode_cache"},
159 {Opt_noinode_cache, "noinode_cache"},
160 {Opt_max_inline, "max_inline=%s"},
161 {Opt_barrier, "barrier"},
162 {Opt_nobarrier, "nobarrier"},
163 {Opt_datacow, "datacow"},
164 {Opt_nodatacow, "nodatacow"},
165 {Opt_datasum, "datasum"},
166 {Opt_nodatasum, "nodatasum"},
167 {Opt_defrag, "autodefrag"},
168 {Opt_nodefrag, "noautodefrag"},
169 {Opt_discard, "discard"},
170 {Opt_discard_mode, "discard=%s"},
171 {Opt_nodiscard, "nodiscard"},
172 {Opt_norecovery, "norecovery"},
173 {Opt_ratio, "metadata_ratio=%u"},
174 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
175 {Opt_skip_balance, "skip_balance"},
176 {Opt_space_cache, "space_cache"},
177 {Opt_no_space_cache, "nospace_cache"},
178 {Opt_space_cache_version, "space_cache=%s"},
179 {Opt_ssd, "ssd"},
180 {Opt_nossd, "nossd"},
181 {Opt_ssd_spread, "ssd_spread"},
182 {Opt_nossd_spread, "nossd_spread"},
183 {Opt_subvol, "subvol=%s"},
184 {Opt_subvol_empty, "subvol="},
185 {Opt_subvolid, "subvolid=%s"},
186 {Opt_thread_pool, "thread_pool=%u"},
187 {Opt_treelog, "treelog"},
188 {Opt_notreelog, "notreelog"},
189 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
190
191 /* Rescue options */
192 {Opt_rescue, "rescue=%s"},
193 /* Deprecated, with alias rescue=nologreplay */
194 {Opt_nologreplay, "nologreplay"},
195 /* Deprecated, with alias rescue=usebackuproot */
196 {Opt_usebackuproot, "usebackuproot"},
197
198 /* Deprecated options */
199 {Opt_recovery, "recovery"},
200
201 /* Debugging options */
202 {Opt_check_integrity, "check_int"},
203 {Opt_check_integrity_including_extent_data, "check_int_data"},
204 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
205 {Opt_enospc_debug, "enospc_debug"},
206 {Opt_noenospc_debug, "noenospc_debug"},
207#ifdef CONFIG_BTRFS_DEBUG
208 {Opt_fragment_data, "fragment=data"},
209 {Opt_fragment_metadata, "fragment=metadata"},
210 {Opt_fragment_all, "fragment=all"},
211#endif
212#ifdef CONFIG_BTRFS_FS_REF_VERIFY
213 {Opt_ref_verify, "ref_verify"},
214#endif
215 {Opt_err, NULL},
216};
217
218static const match_table_t rescue_tokens = {
219 {Opt_usebackuproot, "usebackuproot"},
220 {Opt_nologreplay, "nologreplay"},
221 {Opt_ignorebadroots, "ignorebadroots"},
222 {Opt_ignorebadroots, "ibadroots"},
223 {Opt_ignoredatacsums, "ignoredatacsums"},
224 {Opt_ignoredatacsums, "idatacsums"},
225 {Opt_rescue_all, "all"},
226 {Opt_err, NULL},
227};
228
229static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
230 const char *opt_name)
231{
232 if (fs_info->mount_opt & opt) {
233 btrfs_err(fs_info, "%s must be used with ro mount option",
234 opt_name);
235 return true;
236 }
237 return false;
238}
239
240static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
241{
242 char *opts;
243 char *orig;
244 char *p;
245 substring_t args[MAX_OPT_ARGS];
246 int ret = 0;
247
248 opts = kstrdup(options, GFP_KERNEL);
249 if (!opts)
250 return -ENOMEM;
251 orig = opts;
252
253 while ((p = strsep(&opts, ":")) != NULL) {
254 int token;
255
256 if (!*p)
257 continue;
258 token = match_token(p, rescue_tokens, args);
259 switch (token){
260 case Opt_usebackuproot:
261 btrfs_info(info,
262 "trying to use backup root at mount time");
263 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
264 break;
265 case Opt_nologreplay:
266 btrfs_set_and_info(info, NOLOGREPLAY,
267 "disabling log replay at mount time");
268 break;
269 case Opt_ignorebadroots:
270 btrfs_set_and_info(info, IGNOREBADROOTS,
271 "ignoring bad roots");
272 break;
273 case Opt_ignoredatacsums:
274 btrfs_set_and_info(info, IGNOREDATACSUMS,
275 "ignoring data csums");
276 break;
277 case Opt_rescue_all:
278 btrfs_info(info, "enabling all of the rescue options");
279 btrfs_set_and_info(info, IGNOREDATACSUMS,
280 "ignoring data csums");
281 btrfs_set_and_info(info, IGNOREBADROOTS,
282 "ignoring bad roots");
283 btrfs_set_and_info(info, NOLOGREPLAY,
284 "disabling log replay at mount time");
285 break;
286 case Opt_err:
287 btrfs_info(info, "unrecognized rescue option '%s'", p);
288 ret = -EINVAL;
289 goto out;
290 default:
291 break;
292 }
293
294 }
295out:
296 kfree(orig);
297 return ret;
298}
299
300/*
301 * Regular mount options parser. Everything that is needed only when
302 * reading in a new superblock is parsed here.
303 * XXX JDM: This needs to be cleaned up for remount.
304 */
305int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
306 unsigned long new_flags)
307{
308 substring_t args[MAX_OPT_ARGS];
309 char *p, *num;
310 int intarg;
311 int ret = 0;
312 char *compress_type;
313 bool compress_force = false;
314 enum btrfs_compression_type saved_compress_type;
315 int saved_compress_level;
316 bool saved_compress_force;
317 int no_compress = 0;
318 const bool remounting = test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state);
319
320 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
321 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
322 else if (btrfs_free_space_cache_v1_active(info)) {
323 if (btrfs_is_zoned(info)) {
324 btrfs_info(info,
325 "zoned: clearing existing space cache");
326 btrfs_set_super_cache_generation(info->super_copy, 0);
327 } else {
328 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
329 }
330 }
331
332 /*
333 * Even the options are empty, we still need to do extra check
334 * against new flags
335 */
336 if (!options)
337 goto check;
338
339 while ((p = strsep(&options, ",")) != NULL) {
340 int token;
341 if (!*p)
342 continue;
343
344 token = match_token(p, tokens, args);
345 switch (token) {
346 case Opt_degraded:
347 btrfs_info(info, "allowing degraded mounts");
348 btrfs_set_opt(info->mount_opt, DEGRADED);
349 break;
350 case Opt_subvol:
351 case Opt_subvol_empty:
352 case Opt_subvolid:
353 case Opt_device:
354 /*
355 * These are parsed by btrfs_parse_subvol_options or
356 * btrfs_parse_device_options and can be ignored here.
357 */
358 break;
359 case Opt_nodatasum:
360 btrfs_set_and_info(info, NODATASUM,
361 "setting nodatasum");
362 break;
363 case Opt_datasum:
364 if (btrfs_test_opt(info, NODATASUM)) {
365 if (btrfs_test_opt(info, NODATACOW))
366 btrfs_info(info,
367 "setting datasum, datacow enabled");
368 else
369 btrfs_info(info, "setting datasum");
370 }
371 btrfs_clear_opt(info->mount_opt, NODATACOW);
372 btrfs_clear_opt(info->mount_opt, NODATASUM);
373 break;
374 case Opt_nodatacow:
375 if (!btrfs_test_opt(info, NODATACOW)) {
376 if (!btrfs_test_opt(info, COMPRESS) ||
377 !btrfs_test_opt(info, FORCE_COMPRESS)) {
378 btrfs_info(info,
379 "setting nodatacow, compression disabled");
380 } else {
381 btrfs_info(info, "setting nodatacow");
382 }
383 }
384 btrfs_clear_opt(info->mount_opt, COMPRESS);
385 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
386 btrfs_set_opt(info->mount_opt, NODATACOW);
387 btrfs_set_opt(info->mount_opt, NODATASUM);
388 break;
389 case Opt_datacow:
390 btrfs_clear_and_info(info, NODATACOW,
391 "setting datacow");
392 break;
393 case Opt_compress_force:
394 case Opt_compress_force_type:
395 compress_force = true;
396 fallthrough;
397 case Opt_compress:
398 case Opt_compress_type:
399 saved_compress_type = btrfs_test_opt(info,
400 COMPRESS) ?
401 info->compress_type : BTRFS_COMPRESS_NONE;
402 saved_compress_force =
403 btrfs_test_opt(info, FORCE_COMPRESS);
404 saved_compress_level = info->compress_level;
405 if (token == Opt_compress ||
406 token == Opt_compress_force ||
407 strncmp(args[0].from, "zlib", 4) == 0) {
408 compress_type = "zlib";
409
410 info->compress_type = BTRFS_COMPRESS_ZLIB;
411 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
412 /*
413 * args[0] contains uninitialized data since
414 * for these tokens we don't expect any
415 * parameter.
416 */
417 if (token != Opt_compress &&
418 token != Opt_compress_force)
419 info->compress_level =
420 btrfs_compress_str2level(
421 BTRFS_COMPRESS_ZLIB,
422 args[0].from + 4);
423 btrfs_set_opt(info->mount_opt, COMPRESS);
424 btrfs_clear_opt(info->mount_opt, NODATACOW);
425 btrfs_clear_opt(info->mount_opt, NODATASUM);
426 no_compress = 0;
427 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
428 compress_type = "lzo";
429 info->compress_type = BTRFS_COMPRESS_LZO;
430 info->compress_level = 0;
431 btrfs_set_opt(info->mount_opt, COMPRESS);
432 btrfs_clear_opt(info->mount_opt, NODATACOW);
433 btrfs_clear_opt(info->mount_opt, NODATASUM);
434 btrfs_set_fs_incompat(info, COMPRESS_LZO);
435 no_compress = 0;
436 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
437 compress_type = "zstd";
438 info->compress_type = BTRFS_COMPRESS_ZSTD;
439 info->compress_level =
440 btrfs_compress_str2level(
441 BTRFS_COMPRESS_ZSTD,
442 args[0].from + 4);
443 btrfs_set_opt(info->mount_opt, COMPRESS);
444 btrfs_clear_opt(info->mount_opt, NODATACOW);
445 btrfs_clear_opt(info->mount_opt, NODATASUM);
446 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
447 no_compress = 0;
448 } else if (strncmp(args[0].from, "no", 2) == 0) {
449 compress_type = "no";
450 info->compress_level = 0;
451 info->compress_type = 0;
452 btrfs_clear_opt(info->mount_opt, COMPRESS);
453 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
454 compress_force = false;
455 no_compress++;
456 } else {
457 btrfs_err(info, "unrecognized compression value %s",
458 args[0].from);
459 ret = -EINVAL;
460 goto out;
461 }
462
463 if (compress_force) {
464 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
465 } else {
466 /*
467 * If we remount from compress-force=xxx to
468 * compress=xxx, we need clear FORCE_COMPRESS
469 * flag, otherwise, there is no way for users
470 * to disable forcible compression separately.
471 */
472 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
473 }
474 if (no_compress == 1) {
475 btrfs_info(info, "use no compression");
476 } else if ((info->compress_type != saved_compress_type) ||
477 (compress_force != saved_compress_force) ||
478 (info->compress_level != saved_compress_level)) {
479 btrfs_info(info, "%s %s compression, level %d",
480 (compress_force) ? "force" : "use",
481 compress_type, info->compress_level);
482 }
483 compress_force = false;
484 break;
485 case Opt_ssd:
486 btrfs_set_and_info(info, SSD,
487 "enabling ssd optimizations");
488 btrfs_clear_opt(info->mount_opt, NOSSD);
489 break;
490 case Opt_ssd_spread:
491 btrfs_set_and_info(info, SSD,
492 "enabling ssd optimizations");
493 btrfs_set_and_info(info, SSD_SPREAD,
494 "using spread ssd allocation scheme");
495 btrfs_clear_opt(info->mount_opt, NOSSD);
496 break;
497 case Opt_nossd:
498 btrfs_set_opt(info->mount_opt, NOSSD);
499 btrfs_clear_and_info(info, SSD,
500 "not using ssd optimizations");
501 fallthrough;
502 case Opt_nossd_spread:
503 btrfs_clear_and_info(info, SSD_SPREAD,
504 "not using spread ssd allocation scheme");
505 break;
506 case Opt_barrier:
507 btrfs_clear_and_info(info, NOBARRIER,
508 "turning on barriers");
509 break;
510 case Opt_nobarrier:
511 btrfs_set_and_info(info, NOBARRIER,
512 "turning off barriers");
513 break;
514 case Opt_thread_pool:
515 ret = match_int(&args[0], &intarg);
516 if (ret) {
517 btrfs_err(info, "unrecognized thread_pool value %s",
518 args[0].from);
519 goto out;
520 } else if (intarg == 0) {
521 btrfs_err(info, "invalid value 0 for thread_pool");
522 ret = -EINVAL;
523 goto out;
524 }
525 info->thread_pool_size = intarg;
526 break;
527 case Opt_max_inline:
528 num = match_strdup(&args[0]);
529 if (num) {
530 info->max_inline = memparse(num, NULL);
531 kfree(num);
532
533 if (info->max_inline) {
534 info->max_inline = min_t(u64,
535 info->max_inline,
536 info->sectorsize);
537 }
538 btrfs_info(info, "max_inline at %llu",
539 info->max_inline);
540 } else {
541 ret = -ENOMEM;
542 goto out;
543 }
544 break;
545 case Opt_acl:
546#ifdef CONFIG_BTRFS_FS_POSIX_ACL
547 info->sb->s_flags |= SB_POSIXACL;
548 break;
549#else
550 btrfs_err(info, "support for ACL not compiled in!");
551 ret = -EINVAL;
552 goto out;
553#endif
554 case Opt_noacl:
555 info->sb->s_flags &= ~SB_POSIXACL;
556 break;
557 case Opt_notreelog:
558 btrfs_set_and_info(info, NOTREELOG,
559 "disabling tree log");
560 break;
561 case Opt_treelog:
562 btrfs_clear_and_info(info, NOTREELOG,
563 "enabling tree log");
564 break;
565 case Opt_norecovery:
566 case Opt_nologreplay:
567 btrfs_warn(info,
568 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
569 btrfs_set_and_info(info, NOLOGREPLAY,
570 "disabling log replay at mount time");
571 break;
572 case Opt_flushoncommit:
573 btrfs_set_and_info(info, FLUSHONCOMMIT,
574 "turning on flush-on-commit");
575 break;
576 case Opt_noflushoncommit:
577 btrfs_clear_and_info(info, FLUSHONCOMMIT,
578 "turning off flush-on-commit");
579 break;
580 case Opt_ratio:
581 ret = match_int(&args[0], &intarg);
582 if (ret) {
583 btrfs_err(info, "unrecognized metadata_ratio value %s",
584 args[0].from);
585 goto out;
586 }
587 info->metadata_ratio = intarg;
588 btrfs_info(info, "metadata ratio %u",
589 info->metadata_ratio);
590 break;
591 case Opt_discard:
592 case Opt_discard_mode:
593 if (token == Opt_discard ||
594 strcmp(args[0].from, "sync") == 0) {
595 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
596 btrfs_set_and_info(info, DISCARD_SYNC,
597 "turning on sync discard");
598 } else if (strcmp(args[0].from, "async") == 0) {
599 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
600 btrfs_set_and_info(info, DISCARD_ASYNC,
601 "turning on async discard");
602 } else {
603 btrfs_err(info, "unrecognized discard mode value %s",
604 args[0].from);
605 ret = -EINVAL;
606 goto out;
607 }
608 btrfs_clear_opt(info->mount_opt, NODISCARD);
609 break;
610 case Opt_nodiscard:
611 btrfs_clear_and_info(info, DISCARD_SYNC,
612 "turning off discard");
613 btrfs_clear_and_info(info, DISCARD_ASYNC,
614 "turning off async discard");
615 btrfs_set_opt(info->mount_opt, NODISCARD);
616 break;
617 case Opt_space_cache:
618 case Opt_space_cache_version:
619 /*
620 * We already set FREE_SPACE_TREE above because we have
621 * compat_ro(FREE_SPACE_TREE) set, and we aren't going
622 * to allow v1 to be set for extent tree v2, simply
623 * ignore this setting if we're extent tree v2.
624 */
625 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
626 break;
627 if (token == Opt_space_cache ||
628 strcmp(args[0].from, "v1") == 0) {
629 btrfs_clear_opt(info->mount_opt,
630 FREE_SPACE_TREE);
631 btrfs_set_and_info(info, SPACE_CACHE,
632 "enabling disk space caching");
633 } else if (strcmp(args[0].from, "v2") == 0) {
634 btrfs_clear_opt(info->mount_opt,
635 SPACE_CACHE);
636 btrfs_set_and_info(info, FREE_SPACE_TREE,
637 "enabling free space tree");
638 } else {
639 btrfs_err(info, "unrecognized space_cache value %s",
640 args[0].from);
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 /*
650 * We cannot operate without the free space tree with
651 * extent tree v2, ignore this option.
652 */
653 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
654 break;
655 if (btrfs_test_opt(info, SPACE_CACHE)) {
656 btrfs_clear_and_info(info, SPACE_CACHE,
657 "disabling disk space caching");
658 }
659 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
660 btrfs_clear_and_info(info, FREE_SPACE_TREE,
661 "disabling free space tree");
662 }
663 break;
664 case Opt_inode_cache:
665 case Opt_noinode_cache:
666 btrfs_warn(info,
667 "the 'inode_cache' option is deprecated and has no effect since 5.11");
668 break;
669 case Opt_clear_cache:
670 /*
671 * We cannot clear the free space tree with extent tree
672 * v2, ignore this option.
673 */
674 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
675 break;
676 btrfs_set_and_info(info, CLEAR_CACHE,
677 "force clearing of disk cache");
678 break;
679 case Opt_user_subvol_rm_allowed:
680 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
681 break;
682 case Opt_enospc_debug:
683 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
684 break;
685 case Opt_noenospc_debug:
686 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
687 break;
688 case Opt_defrag:
689 btrfs_set_and_info(info, AUTO_DEFRAG,
690 "enabling auto defrag");
691 break;
692 case Opt_nodefrag:
693 btrfs_clear_and_info(info, AUTO_DEFRAG,
694 "disabling auto defrag");
695 break;
696 case Opt_recovery:
697 case Opt_usebackuproot:
698 btrfs_warn(info,
699 "'%s' is deprecated, use 'rescue=usebackuproot' instead",
700 token == Opt_recovery ? "recovery" :
701 "usebackuproot");
702 btrfs_info(info,
703 "trying to use backup root at mount time");
704 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
705 break;
706 case Opt_skip_balance:
707 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
708 break;
709#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
710 case Opt_check_integrity_including_extent_data:
711 btrfs_info(info,
712 "enabling check integrity including extent data");
713 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
714 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
715 break;
716 case Opt_check_integrity:
717 btrfs_info(info, "enabling check integrity");
718 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
719 break;
720 case Opt_check_integrity_print_mask:
721 ret = match_int(&args[0], &intarg);
722 if (ret) {
723 btrfs_err(info,
724 "unrecognized check_integrity_print_mask value %s",
725 args[0].from);
726 goto out;
727 }
728 info->check_integrity_print_mask = intarg;
729 btrfs_info(info, "check_integrity_print_mask 0x%x",
730 info->check_integrity_print_mask);
731 break;
732#else
733 case Opt_check_integrity_including_extent_data:
734 case Opt_check_integrity:
735 case Opt_check_integrity_print_mask:
736 btrfs_err(info,
737 "support for check_integrity* not compiled in!");
738 ret = -EINVAL;
739 goto out;
740#endif
741 case Opt_fatal_errors:
742 if (strcmp(args[0].from, "panic") == 0) {
743 btrfs_set_opt(info->mount_opt,
744 PANIC_ON_FATAL_ERROR);
745 } else if (strcmp(args[0].from, "bug") == 0) {
746 btrfs_clear_opt(info->mount_opt,
747 PANIC_ON_FATAL_ERROR);
748 } else {
749 btrfs_err(info, "unrecognized fatal_errors value %s",
750 args[0].from);
751 ret = -EINVAL;
752 goto out;
753 }
754 break;
755 case Opt_commit_interval:
756 intarg = 0;
757 ret = match_int(&args[0], &intarg);
758 if (ret) {
759 btrfs_err(info, "unrecognized commit_interval value %s",
760 args[0].from);
761 ret = -EINVAL;
762 goto out;
763 }
764 if (intarg == 0) {
765 btrfs_info(info,
766 "using default commit interval %us",
767 BTRFS_DEFAULT_COMMIT_INTERVAL);
768 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
769 } else if (intarg > 300) {
770 btrfs_warn(info, "excessive commit interval %d",
771 intarg);
772 }
773 info->commit_interval = intarg;
774 break;
775 case Opt_rescue:
776 ret = parse_rescue_options(info, args[0].from);
777 if (ret < 0) {
778 btrfs_err(info, "unrecognized rescue value %s",
779 args[0].from);
780 goto out;
781 }
782 break;
783#ifdef CONFIG_BTRFS_DEBUG
784 case Opt_fragment_all:
785 btrfs_info(info, "fragmenting all space");
786 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
787 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
788 break;
789 case Opt_fragment_metadata:
790 btrfs_info(info, "fragmenting metadata");
791 btrfs_set_opt(info->mount_opt,
792 FRAGMENT_METADATA);
793 break;
794 case Opt_fragment_data:
795 btrfs_info(info, "fragmenting data");
796 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
797 break;
798#endif
799#ifdef CONFIG_BTRFS_FS_REF_VERIFY
800 case Opt_ref_verify:
801 btrfs_info(info, "doing ref verification");
802 btrfs_set_opt(info->mount_opt, REF_VERIFY);
803 break;
804#endif
805 case Opt_err:
806 btrfs_err(info, "unrecognized mount option '%s'", p);
807 ret = -EINVAL;
808 goto out;
809 default:
810 break;
811 }
812 }
813check:
814 /* We're read-only, don't have to check. */
815 if (new_flags & SB_RDONLY)
816 goto out;
817
818 if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
819 check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
820 check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
821 ret = -EINVAL;
822out:
823 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
824 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
825 !btrfs_test_opt(info, CLEAR_CACHE)) {
826 btrfs_err(info, "cannot disable free space tree");
827 ret = -EINVAL;
828
829 }
830 if (!ret)
831 ret = btrfs_check_mountopts_zoned(info);
832 if (!ret && !remounting) {
833 if (btrfs_test_opt(info, SPACE_CACHE))
834 btrfs_info(info, "disk space caching is enabled");
835 if (btrfs_test_opt(info, FREE_SPACE_TREE))
836 btrfs_info(info, "using free space tree");
837 }
838 return ret;
839}
840
841/*
842 * Parse mount options that are required early in the mount process.
843 *
844 * All other options will be parsed on much later in the mount process and
845 * only when we need to allocate a new super block.
846 */
847static int btrfs_parse_device_options(const char *options, fmode_t flags,
848 void *holder)
849{
850 substring_t args[MAX_OPT_ARGS];
851 char *device_name, *opts, *orig, *p;
852 struct btrfs_device *device = NULL;
853 int error = 0;
854
855 lockdep_assert_held(&uuid_mutex);
856
857 if (!options)
858 return 0;
859
860 /*
861 * strsep changes the string, duplicate it because btrfs_parse_options
862 * gets called later
863 */
864 opts = kstrdup(options, GFP_KERNEL);
865 if (!opts)
866 return -ENOMEM;
867 orig = opts;
868
869 while ((p = strsep(&opts, ",")) != NULL) {
870 int token;
871
872 if (!*p)
873 continue;
874
875 token = match_token(p, tokens, args);
876 if (token == Opt_device) {
877 device_name = match_strdup(&args[0]);
878 if (!device_name) {
879 error = -ENOMEM;
880 goto out;
881 }
882 device = btrfs_scan_one_device(device_name, flags,
883 holder);
884 kfree(device_name);
885 if (IS_ERR(device)) {
886 error = PTR_ERR(device);
887 goto out;
888 }
889 }
890 }
891
892out:
893 kfree(orig);
894 return error;
895}
896
897/*
898 * Parse mount options that are related to subvolume id
899 *
900 * The value is later passed to mount_subvol()
901 */
902static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
903 u64 *subvol_objectid)
904{
905 substring_t args[MAX_OPT_ARGS];
906 char *opts, *orig, *p;
907 int error = 0;
908 u64 subvolid;
909
910 if (!options)
911 return 0;
912
913 /*
914 * strsep changes the string, duplicate it because
915 * btrfs_parse_device_options gets called later
916 */
917 opts = kstrdup(options, GFP_KERNEL);
918 if (!opts)
919 return -ENOMEM;
920 orig = opts;
921
922 while ((p = strsep(&opts, ",")) != NULL) {
923 int token;
924 if (!*p)
925 continue;
926
927 token = match_token(p, tokens, args);
928 switch (token) {
929 case Opt_subvol:
930 kfree(*subvol_name);
931 *subvol_name = match_strdup(&args[0]);
932 if (!*subvol_name) {
933 error = -ENOMEM;
934 goto out;
935 }
936 break;
937 case Opt_subvolid:
938 error = match_u64(&args[0], &subvolid);
939 if (error)
940 goto out;
941
942 /* we want the original fs_tree */
943 if (subvolid == 0)
944 subvolid = BTRFS_FS_TREE_OBJECTID;
945
946 *subvol_objectid = subvolid;
947 break;
948 default:
949 break;
950 }
951 }
952
953out:
954 kfree(orig);
955 return error;
956}
957
958char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
959 u64 subvol_objectid)
960{
961 struct btrfs_root *root = fs_info->tree_root;
962 struct btrfs_root *fs_root = NULL;
963 struct btrfs_root_ref *root_ref;
964 struct btrfs_inode_ref *inode_ref;
965 struct btrfs_key key;
966 struct btrfs_path *path = NULL;
967 char *name = NULL, *ptr;
968 u64 dirid;
969 int len;
970 int ret;
971
972 path = btrfs_alloc_path();
973 if (!path) {
974 ret = -ENOMEM;
975 goto err;
976 }
977
978 name = kmalloc(PATH_MAX, GFP_KERNEL);
979 if (!name) {
980 ret = -ENOMEM;
981 goto err;
982 }
983 ptr = name + PATH_MAX - 1;
984 ptr[0] = '\0';
985
986 /*
987 * Walk up the subvolume trees in the tree of tree roots by root
988 * backrefs until we hit the top-level subvolume.
989 */
990 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
991 key.objectid = subvol_objectid;
992 key.type = BTRFS_ROOT_BACKREF_KEY;
993 key.offset = (u64)-1;
994
995 ret = btrfs_search_backwards(root, &key, path);
996 if (ret < 0) {
997 goto err;
998 } else if (ret > 0) {
999 ret = -ENOENT;
1000 goto err;
1001 }
1002
1003 subvol_objectid = key.offset;
1004
1005 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1006 struct btrfs_root_ref);
1007 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1008 ptr -= len + 1;
1009 if (ptr < name) {
1010 ret = -ENAMETOOLONG;
1011 goto err;
1012 }
1013 read_extent_buffer(path->nodes[0], ptr + 1,
1014 (unsigned long)(root_ref + 1), len);
1015 ptr[0] = '/';
1016 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1017 btrfs_release_path(path);
1018
1019 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1020 if (IS_ERR(fs_root)) {
1021 ret = PTR_ERR(fs_root);
1022 fs_root = NULL;
1023 goto err;
1024 }
1025
1026 /*
1027 * Walk up the filesystem tree by inode refs until we hit the
1028 * root directory.
1029 */
1030 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1031 key.objectid = dirid;
1032 key.type = BTRFS_INODE_REF_KEY;
1033 key.offset = (u64)-1;
1034
1035 ret = btrfs_search_backwards(fs_root, &key, path);
1036 if (ret < 0) {
1037 goto err;
1038 } else if (ret > 0) {
1039 ret = -ENOENT;
1040 goto err;
1041 }
1042
1043 dirid = key.offset;
1044
1045 inode_ref = btrfs_item_ptr(path->nodes[0],
1046 path->slots[0],
1047 struct btrfs_inode_ref);
1048 len = btrfs_inode_ref_name_len(path->nodes[0],
1049 inode_ref);
1050 ptr -= len + 1;
1051 if (ptr < name) {
1052 ret = -ENAMETOOLONG;
1053 goto err;
1054 }
1055 read_extent_buffer(path->nodes[0], ptr + 1,
1056 (unsigned long)(inode_ref + 1), len);
1057 ptr[0] = '/';
1058 btrfs_release_path(path);
1059 }
1060 btrfs_put_root(fs_root);
1061 fs_root = NULL;
1062 }
1063
1064 btrfs_free_path(path);
1065 if (ptr == name + PATH_MAX - 1) {
1066 name[0] = '/';
1067 name[1] = '\0';
1068 } else {
1069 memmove(name, ptr, name + PATH_MAX - ptr);
1070 }
1071 return name;
1072
1073err:
1074 btrfs_put_root(fs_root);
1075 btrfs_free_path(path);
1076 kfree(name);
1077 return ERR_PTR(ret);
1078}
1079
1080static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1081{
1082 struct btrfs_root *root = fs_info->tree_root;
1083 struct btrfs_dir_item *di;
1084 struct btrfs_path *path;
1085 struct btrfs_key location;
1086 struct fscrypt_str name = FSTR_INIT("default", 7);
1087 u64 dir_id;
1088
1089 path = btrfs_alloc_path();
1090 if (!path)
1091 return -ENOMEM;
1092
1093 /*
1094 * Find the "default" dir item which points to the root item that we
1095 * will mount by default if we haven't been given a specific subvolume
1096 * to mount.
1097 */
1098 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1099 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
1100 if (IS_ERR(di)) {
1101 btrfs_free_path(path);
1102 return PTR_ERR(di);
1103 }
1104 if (!di) {
1105 /*
1106 * Ok the default dir item isn't there. This is weird since
1107 * it's always been there, but don't freak out, just try and
1108 * mount the top-level subvolume.
1109 */
1110 btrfs_free_path(path);
1111 *objectid = BTRFS_FS_TREE_OBJECTID;
1112 return 0;
1113 }
1114
1115 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1116 btrfs_free_path(path);
1117 *objectid = location.objectid;
1118 return 0;
1119}
1120
1121static int btrfs_fill_super(struct super_block *sb,
1122 struct btrfs_fs_devices *fs_devices,
1123 void *data)
1124{
1125 struct inode *inode;
1126 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1127 int err;
1128
1129 sb->s_maxbytes = MAX_LFS_FILESIZE;
1130 sb->s_magic = BTRFS_SUPER_MAGIC;
1131 sb->s_op = &btrfs_super_ops;
1132 sb->s_d_op = &btrfs_dentry_operations;
1133 sb->s_export_op = &btrfs_export_ops;
1134#ifdef CONFIG_FS_VERITY
1135 sb->s_vop = &btrfs_verityops;
1136#endif
1137 sb->s_xattr = btrfs_xattr_handlers;
1138 sb->s_time_gran = 1;
1139#ifdef CONFIG_BTRFS_FS_POSIX_ACL
1140 sb->s_flags |= SB_POSIXACL;
1141#endif
1142 sb->s_flags |= SB_I_VERSION;
1143 sb->s_iflags |= SB_I_CGROUPWB;
1144
1145 err = super_setup_bdi(sb);
1146 if (err) {
1147 btrfs_err(fs_info, "super_setup_bdi failed");
1148 return err;
1149 }
1150
1151 err = open_ctree(sb, fs_devices, (char *)data);
1152 if (err) {
1153 btrfs_err(fs_info, "open_ctree failed");
1154 return err;
1155 }
1156
1157 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1158 if (IS_ERR(inode)) {
1159 err = PTR_ERR(inode);
1160 goto fail_close;
1161 }
1162
1163 sb->s_root = d_make_root(inode);
1164 if (!sb->s_root) {
1165 err = -ENOMEM;
1166 goto fail_close;
1167 }
1168
1169 sb->s_flags |= SB_ACTIVE;
1170 return 0;
1171
1172fail_close:
1173 close_ctree(fs_info);
1174 return err;
1175}
1176
1177int btrfs_sync_fs(struct super_block *sb, int wait)
1178{
1179 struct btrfs_trans_handle *trans;
1180 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1181 struct btrfs_root *root = fs_info->tree_root;
1182
1183 trace_btrfs_sync_fs(fs_info, wait);
1184
1185 if (!wait) {
1186 filemap_flush(fs_info->btree_inode->i_mapping);
1187 return 0;
1188 }
1189
1190 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1191
1192 trans = btrfs_attach_transaction_barrier(root);
1193 if (IS_ERR(trans)) {
1194 /* no transaction, don't bother */
1195 if (PTR_ERR(trans) == -ENOENT) {
1196 /*
1197 * Exit unless we have some pending changes
1198 * that need to go through commit
1199 */
1200 if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1201 &fs_info->flags))
1202 return 0;
1203 /*
1204 * A non-blocking test if the fs is frozen. We must not
1205 * start a new transaction here otherwise a deadlock
1206 * happens. The pending operations are delayed to the
1207 * next commit after thawing.
1208 */
1209 if (sb_start_write_trylock(sb))
1210 sb_end_write(sb);
1211 else
1212 return 0;
1213 trans = btrfs_start_transaction(root, 0);
1214 }
1215 if (IS_ERR(trans))
1216 return PTR_ERR(trans);
1217 }
1218 return btrfs_commit_transaction(trans);
1219}
1220
1221static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1222{
1223 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1224 *printed = true;
1225}
1226
1227static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1228{
1229 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1230 const char *compress_type;
1231 const char *subvol_name;
1232 bool printed = false;
1233
1234 if (btrfs_test_opt(info, DEGRADED))
1235 seq_puts(seq, ",degraded");
1236 if (btrfs_test_opt(info, NODATASUM))
1237 seq_puts(seq, ",nodatasum");
1238 if (btrfs_test_opt(info, NODATACOW))
1239 seq_puts(seq, ",nodatacow");
1240 if (btrfs_test_opt(info, NOBARRIER))
1241 seq_puts(seq, ",nobarrier");
1242 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1243 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1244 if (info->thread_pool_size != min_t(unsigned long,
1245 num_online_cpus() + 2, 8))
1246 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1247 if (btrfs_test_opt(info, COMPRESS)) {
1248 compress_type = btrfs_compress_type2str(info->compress_type);
1249 if (btrfs_test_opt(info, FORCE_COMPRESS))
1250 seq_printf(seq, ",compress-force=%s", compress_type);
1251 else
1252 seq_printf(seq, ",compress=%s", compress_type);
1253 if (info->compress_level)
1254 seq_printf(seq, ":%d", info->compress_level);
1255 }
1256 if (btrfs_test_opt(info, NOSSD))
1257 seq_puts(seq, ",nossd");
1258 if (btrfs_test_opt(info, SSD_SPREAD))
1259 seq_puts(seq, ",ssd_spread");
1260 else if (btrfs_test_opt(info, SSD))
1261 seq_puts(seq, ",ssd");
1262 if (btrfs_test_opt(info, NOTREELOG))
1263 seq_puts(seq, ",notreelog");
1264 if (btrfs_test_opt(info, NOLOGREPLAY))
1265 print_rescue_option(seq, "nologreplay", &printed);
1266 if (btrfs_test_opt(info, USEBACKUPROOT))
1267 print_rescue_option(seq, "usebackuproot", &printed);
1268 if (btrfs_test_opt(info, IGNOREBADROOTS))
1269 print_rescue_option(seq, "ignorebadroots", &printed);
1270 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1271 print_rescue_option(seq, "ignoredatacsums", &printed);
1272 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1273 seq_puts(seq, ",flushoncommit");
1274 if (btrfs_test_opt(info, DISCARD_SYNC))
1275 seq_puts(seq, ",discard");
1276 if (btrfs_test_opt(info, DISCARD_ASYNC))
1277 seq_puts(seq, ",discard=async");
1278 if (!(info->sb->s_flags & SB_POSIXACL))
1279 seq_puts(seq, ",noacl");
1280 if (btrfs_free_space_cache_v1_active(info))
1281 seq_puts(seq, ",space_cache");
1282 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1283 seq_puts(seq, ",space_cache=v2");
1284 else
1285 seq_puts(seq, ",nospace_cache");
1286 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1287 seq_puts(seq, ",rescan_uuid_tree");
1288 if (btrfs_test_opt(info, CLEAR_CACHE))
1289 seq_puts(seq, ",clear_cache");
1290 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1291 seq_puts(seq, ",user_subvol_rm_allowed");
1292 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1293 seq_puts(seq, ",enospc_debug");
1294 if (btrfs_test_opt(info, AUTO_DEFRAG))
1295 seq_puts(seq, ",autodefrag");
1296 if (btrfs_test_opt(info, SKIP_BALANCE))
1297 seq_puts(seq, ",skip_balance");
1298#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1299 if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1300 seq_puts(seq, ",check_int_data");
1301 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1302 seq_puts(seq, ",check_int");
1303 if (info->check_integrity_print_mask)
1304 seq_printf(seq, ",check_int_print_mask=%d",
1305 info->check_integrity_print_mask);
1306#endif
1307 if (info->metadata_ratio)
1308 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1309 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1310 seq_puts(seq, ",fatal_errors=panic");
1311 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1312 seq_printf(seq, ",commit=%u", info->commit_interval);
1313#ifdef CONFIG_BTRFS_DEBUG
1314 if (btrfs_test_opt(info, FRAGMENT_DATA))
1315 seq_puts(seq, ",fragment=data");
1316 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1317 seq_puts(seq, ",fragment=metadata");
1318#endif
1319 if (btrfs_test_opt(info, REF_VERIFY))
1320 seq_puts(seq, ",ref_verify");
1321 seq_printf(seq, ",subvolid=%llu",
1322 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1323 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1324 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1325 if (!IS_ERR(subvol_name)) {
1326 seq_puts(seq, ",subvol=");
1327 seq_escape(seq, subvol_name, " \t\n\\");
1328 kfree(subvol_name);
1329 }
1330 return 0;
1331}
1332
1333static int btrfs_test_super(struct super_block *s, void *data)
1334{
1335 struct btrfs_fs_info *p = data;
1336 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1337
1338 return fs_info->fs_devices == p->fs_devices;
1339}
1340
1341static int btrfs_set_super(struct super_block *s, void *data)
1342{
1343 int err = set_anon_super(s, data);
1344 if (!err)
1345 s->s_fs_info = data;
1346 return err;
1347}
1348
1349/*
1350 * subvolumes are identified by ino 256
1351 */
1352static inline int is_subvolume_inode(struct inode *inode)
1353{
1354 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1355 return 1;
1356 return 0;
1357}
1358
1359static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1360 struct vfsmount *mnt)
1361{
1362 struct dentry *root;
1363 int ret;
1364
1365 if (!subvol_name) {
1366 if (!subvol_objectid) {
1367 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1368 &subvol_objectid);
1369 if (ret) {
1370 root = ERR_PTR(ret);
1371 goto out;
1372 }
1373 }
1374 subvol_name = btrfs_get_subvol_name_from_objectid(
1375 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1376 if (IS_ERR(subvol_name)) {
1377 root = ERR_CAST(subvol_name);
1378 subvol_name = NULL;
1379 goto out;
1380 }
1381
1382 }
1383
1384 root = mount_subtree(mnt, subvol_name);
1385 /* mount_subtree() drops our reference on the vfsmount. */
1386 mnt = NULL;
1387
1388 if (!IS_ERR(root)) {
1389 struct super_block *s = root->d_sb;
1390 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1391 struct inode *root_inode = d_inode(root);
1392 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1393
1394 ret = 0;
1395 if (!is_subvolume_inode(root_inode)) {
1396 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1397 subvol_name);
1398 ret = -EINVAL;
1399 }
1400 if (subvol_objectid && root_objectid != subvol_objectid) {
1401 /*
1402 * This will also catch a race condition where a
1403 * subvolume which was passed by ID is renamed and
1404 * another subvolume is renamed over the old location.
1405 */
1406 btrfs_err(fs_info,
1407 "subvol '%s' does not match subvolid %llu",
1408 subvol_name, subvol_objectid);
1409 ret = -EINVAL;
1410 }
1411 if (ret) {
1412 dput(root);
1413 root = ERR_PTR(ret);
1414 deactivate_locked_super(s);
1415 }
1416 }
1417
1418out:
1419 mntput(mnt);
1420 kfree(subvol_name);
1421 return root;
1422}
1423
1424/*
1425 * Find a superblock for the given device / mount point.
1426 *
1427 * Note: This is based on mount_bdev from fs/super.c with a few additions
1428 * for multiple device setup. Make sure to keep it in sync.
1429 */
1430static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1431 int flags, const char *device_name, void *data)
1432{
1433 struct block_device *bdev = NULL;
1434 struct super_block *s;
1435 struct btrfs_device *device = NULL;
1436 struct btrfs_fs_devices *fs_devices = NULL;
1437 struct btrfs_fs_info *fs_info = NULL;
1438 void *new_sec_opts = NULL;
1439 fmode_t mode = FMODE_READ;
1440 int error = 0;
1441
1442 if (!(flags & SB_RDONLY))
1443 mode |= FMODE_WRITE;
1444
1445 if (data) {
1446 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1447 if (error)
1448 return ERR_PTR(error);
1449 }
1450
1451 /*
1452 * Setup a dummy root and fs_info for test/set super. This is because
1453 * we don't actually fill this stuff out until open_ctree, but we need
1454 * then open_ctree will properly initialize the file system specific
1455 * settings later. btrfs_init_fs_info initializes the static elements
1456 * of the fs_info (locks and such) to make cleanup easier if we find a
1457 * superblock with our given fs_devices later on at sget() time.
1458 */
1459 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1460 if (!fs_info) {
1461 error = -ENOMEM;
1462 goto error_sec_opts;
1463 }
1464 btrfs_init_fs_info(fs_info);
1465
1466 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1467 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1468 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1469 error = -ENOMEM;
1470 goto error_fs_info;
1471 }
1472
1473 mutex_lock(&uuid_mutex);
1474 error = btrfs_parse_device_options(data, mode, fs_type);
1475 if (error) {
1476 mutex_unlock(&uuid_mutex);
1477 goto error_fs_info;
1478 }
1479
1480 device = btrfs_scan_one_device(device_name, mode, fs_type);
1481 if (IS_ERR(device)) {
1482 mutex_unlock(&uuid_mutex);
1483 error = PTR_ERR(device);
1484 goto error_fs_info;
1485 }
1486
1487 fs_devices = device->fs_devices;
1488 fs_info->fs_devices = fs_devices;
1489
1490 error = btrfs_open_devices(fs_devices, mode, fs_type);
1491 mutex_unlock(&uuid_mutex);
1492 if (error)
1493 goto error_fs_info;
1494
1495 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1496 error = -EACCES;
1497 goto error_close_devices;
1498 }
1499
1500 bdev = fs_devices->latest_dev->bdev;
1501 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1502 fs_info);
1503 if (IS_ERR(s)) {
1504 error = PTR_ERR(s);
1505 goto error_close_devices;
1506 }
1507
1508 if (s->s_root) {
1509 btrfs_close_devices(fs_devices);
1510 btrfs_free_fs_info(fs_info);
1511 if ((flags ^ s->s_flags) & SB_RDONLY)
1512 error = -EBUSY;
1513 } else {
1514 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1515 shrinker_debugfs_rename(&s->s_shrink, "sb-%s:%s", fs_type->name,
1516 s->s_id);
1517 btrfs_sb(s)->bdev_holder = fs_type;
1518 if (!strstr(crc32c_impl(), "generic"))
1519 set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1520 error = btrfs_fill_super(s, fs_devices, data);
1521 }
1522 if (!error)
1523 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1524 security_free_mnt_opts(&new_sec_opts);
1525 if (error) {
1526 deactivate_locked_super(s);
1527 return ERR_PTR(error);
1528 }
1529
1530 return dget(s->s_root);
1531
1532error_close_devices:
1533 btrfs_close_devices(fs_devices);
1534error_fs_info:
1535 btrfs_free_fs_info(fs_info);
1536error_sec_opts:
1537 security_free_mnt_opts(&new_sec_opts);
1538 return ERR_PTR(error);
1539}
1540
1541/*
1542 * Mount function which is called by VFS layer.
1543 *
1544 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1545 * which needs vfsmount* of device's root (/). This means device's root has to
1546 * be mounted internally in any case.
1547 *
1548 * Operation flow:
1549 * 1. Parse subvol id related options for later use in mount_subvol().
1550 *
1551 * 2. Mount device's root (/) by calling vfs_kern_mount().
1552 *
1553 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1554 * first place. In order to avoid calling btrfs_mount() again, we use
1555 * different file_system_type which is not registered to VFS by
1556 * register_filesystem() (btrfs_root_fs_type). As a result,
1557 * btrfs_mount_root() is called. The return value will be used by
1558 * mount_subtree() in mount_subvol().
1559 *
1560 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1561 * "btrfs subvolume set-default", mount_subvol() is called always.
1562 */
1563static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1564 const char *device_name, void *data)
1565{
1566 struct vfsmount *mnt_root;
1567 struct dentry *root;
1568 char *subvol_name = NULL;
1569 u64 subvol_objectid = 0;
1570 int error = 0;
1571
1572 error = btrfs_parse_subvol_options(data, &subvol_name,
1573 &subvol_objectid);
1574 if (error) {
1575 kfree(subvol_name);
1576 return ERR_PTR(error);
1577 }
1578
1579 /* mount device's root (/) */
1580 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1581 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1582 if (flags & SB_RDONLY) {
1583 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1584 flags & ~SB_RDONLY, device_name, data);
1585 } else {
1586 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1587 flags | SB_RDONLY, device_name, data);
1588 if (IS_ERR(mnt_root)) {
1589 root = ERR_CAST(mnt_root);
1590 kfree(subvol_name);
1591 goto out;
1592 }
1593
1594 down_write(&mnt_root->mnt_sb->s_umount);
1595 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1596 up_write(&mnt_root->mnt_sb->s_umount);
1597 if (error < 0) {
1598 root = ERR_PTR(error);
1599 mntput(mnt_root);
1600 kfree(subvol_name);
1601 goto out;
1602 }
1603 }
1604 }
1605 if (IS_ERR(mnt_root)) {
1606 root = ERR_CAST(mnt_root);
1607 kfree(subvol_name);
1608 goto out;
1609 }
1610
1611 /* mount_subvol() will free subvol_name and mnt_root */
1612 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1613
1614out:
1615 return root;
1616}
1617
1618static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1619 u32 new_pool_size, u32 old_pool_size)
1620{
1621 if (new_pool_size == old_pool_size)
1622 return;
1623
1624 fs_info->thread_pool_size = new_pool_size;
1625
1626 btrfs_info(fs_info, "resize thread pool %d -> %d",
1627 old_pool_size, new_pool_size);
1628
1629 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1630 btrfs_workqueue_set_max(fs_info->hipri_workers, new_pool_size);
1631 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1632 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1633 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1634 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1635 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1636}
1637
1638static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1639 unsigned long old_opts, int flags)
1640{
1641 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1642 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1643 (flags & SB_RDONLY))) {
1644 /* wait for any defraggers to finish */
1645 wait_event(fs_info->transaction_wait,
1646 (atomic_read(&fs_info->defrag_running) == 0));
1647 if (flags & SB_RDONLY)
1648 sync_filesystem(fs_info->sb);
1649 }
1650}
1651
1652static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1653 unsigned long old_opts)
1654{
1655 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1656
1657 /*
1658 * We need to cleanup all defragable inodes if the autodefragment is
1659 * close or the filesystem is read only.
1660 */
1661 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1662 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1663 btrfs_cleanup_defrag_inodes(fs_info);
1664 }
1665
1666 /* If we toggled discard async */
1667 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1668 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1669 btrfs_discard_resume(fs_info);
1670 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1671 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1672 btrfs_discard_cleanup(fs_info);
1673
1674 /* If we toggled space cache */
1675 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1676 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1677}
1678
1679static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1680{
1681 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1682 unsigned old_flags = sb->s_flags;
1683 unsigned long old_opts = fs_info->mount_opt;
1684 unsigned long old_compress_type = fs_info->compress_type;
1685 u64 old_max_inline = fs_info->max_inline;
1686 u32 old_thread_pool_size = fs_info->thread_pool_size;
1687 u32 old_metadata_ratio = fs_info->metadata_ratio;
1688 int ret;
1689
1690 sync_filesystem(sb);
1691 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1692
1693 if (data) {
1694 void *new_sec_opts = NULL;
1695
1696 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1697 if (!ret)
1698 ret = security_sb_remount(sb, new_sec_opts);
1699 security_free_mnt_opts(&new_sec_opts);
1700 if (ret)
1701 goto restore;
1702 }
1703
1704 ret = btrfs_parse_options(fs_info, data, *flags);
1705 if (ret)
1706 goto restore;
1707
1708 ret = btrfs_check_features(fs_info, !(*flags & SB_RDONLY));
1709 if (ret < 0)
1710 goto restore;
1711
1712 btrfs_remount_begin(fs_info, old_opts, *flags);
1713 btrfs_resize_thread_pool(fs_info,
1714 fs_info->thread_pool_size, old_thread_pool_size);
1715
1716 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1717 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1718 (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1719 btrfs_warn(fs_info,
1720 "remount supports changing free space tree only from ro to rw");
1721 /* Make sure free space cache options match the state on disk */
1722 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1723 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1724 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1725 }
1726 if (btrfs_free_space_cache_v1_active(fs_info)) {
1727 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1728 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1729 }
1730 }
1731
1732 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1733 goto out;
1734
1735 if (*flags & SB_RDONLY) {
1736 /*
1737 * this also happens on 'umount -rf' or on shutdown, when
1738 * the filesystem is busy.
1739 */
1740 cancel_work_sync(&fs_info->async_reclaim_work);
1741 cancel_work_sync(&fs_info->async_data_reclaim_work);
1742
1743 btrfs_discard_cleanup(fs_info);
1744
1745 /* wait for the uuid_scan task to finish */
1746 down(&fs_info->uuid_tree_rescan_sem);
1747 /* avoid complains from lockdep et al. */
1748 up(&fs_info->uuid_tree_rescan_sem);
1749
1750 btrfs_set_sb_rdonly(sb);
1751
1752 /*
1753 * Setting SB_RDONLY will put the cleaner thread to
1754 * sleep at the next loop if it's already active.
1755 * If it's already asleep, we'll leave unused block
1756 * groups on disk until we're mounted read-write again
1757 * unless we clean them up here.
1758 */
1759 btrfs_delete_unused_bgs(fs_info);
1760
1761 /*
1762 * The cleaner task could be already running before we set the
1763 * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
1764 * We must make sure that after we finish the remount, i.e. after
1765 * we call btrfs_commit_super(), the cleaner can no longer start
1766 * a transaction - either because it was dropping a dead root,
1767 * running delayed iputs or deleting an unused block group (the
1768 * cleaner picked a block group from the list of unused block
1769 * groups before we were able to in the previous call to
1770 * btrfs_delete_unused_bgs()).
1771 */
1772 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
1773 TASK_UNINTERRUPTIBLE);
1774
1775 /*
1776 * We've set the superblock to RO mode, so we might have made
1777 * the cleaner task sleep without running all pending delayed
1778 * iputs. Go through all the delayed iputs here, so that if an
1779 * unmount happens without remounting RW we don't end up at
1780 * finishing close_ctree() with a non-empty list of delayed
1781 * iputs.
1782 */
1783 btrfs_run_delayed_iputs(fs_info);
1784
1785 btrfs_dev_replace_suspend_for_unmount(fs_info);
1786 btrfs_scrub_cancel(fs_info);
1787 btrfs_pause_balance(fs_info);
1788
1789 /*
1790 * Pause the qgroup rescan worker if it is running. We don't want
1791 * it to be still running after we are in RO mode, as after that,
1792 * by the time we unmount, it might have left a transaction open,
1793 * so we would leak the transaction and/or crash.
1794 */
1795 btrfs_qgroup_wait_for_completion(fs_info, false);
1796
1797 ret = btrfs_commit_super(fs_info);
1798 if (ret)
1799 goto restore;
1800 } else {
1801 if (BTRFS_FS_ERROR(fs_info)) {
1802 btrfs_err(fs_info,
1803 "Remounting read-write after error is not allowed");
1804 ret = -EINVAL;
1805 goto restore;
1806 }
1807 if (fs_info->fs_devices->rw_devices == 0) {
1808 ret = -EACCES;
1809 goto restore;
1810 }
1811
1812 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1813 btrfs_warn(fs_info,
1814 "too many missing devices, writable remount is not allowed");
1815 ret = -EACCES;
1816 goto restore;
1817 }
1818
1819 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1820 btrfs_warn(fs_info,
1821 "mount required to replay tree-log, cannot remount read-write");
1822 ret = -EINVAL;
1823 goto restore;
1824 }
1825
1826 /*
1827 * NOTE: when remounting with a change that does writes, don't
1828 * put it anywhere above this point, as we are not sure to be
1829 * safe to write until we pass the above checks.
1830 */
1831 ret = btrfs_start_pre_rw_mount(fs_info);
1832 if (ret)
1833 goto restore;
1834
1835 btrfs_clear_sb_rdonly(sb);
1836
1837 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1838 }
1839out:
1840 /*
1841 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
1842 * since the absence of the flag means it can be toggled off by remount.
1843 */
1844 *flags |= SB_I_VERSION;
1845
1846 wake_up_process(fs_info->transaction_kthread);
1847 btrfs_remount_cleanup(fs_info, old_opts);
1848 btrfs_clear_oneshot_options(fs_info);
1849 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1850
1851 return 0;
1852
1853restore:
1854 /* We've hit an error - don't reset SB_RDONLY */
1855 if (sb_rdonly(sb))
1856 old_flags |= SB_RDONLY;
1857 if (!(old_flags & SB_RDONLY))
1858 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
1859 sb->s_flags = old_flags;
1860 fs_info->mount_opt = old_opts;
1861 fs_info->compress_type = old_compress_type;
1862 fs_info->max_inline = old_max_inline;
1863 btrfs_resize_thread_pool(fs_info,
1864 old_thread_pool_size, fs_info->thread_pool_size);
1865 fs_info->metadata_ratio = old_metadata_ratio;
1866 btrfs_remount_cleanup(fs_info, old_opts);
1867 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1868
1869 return ret;
1870}
1871
1872/* Used to sort the devices by max_avail(descending sort) */
1873static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1874{
1875 const struct btrfs_device_info *dev_info1 = a;
1876 const struct btrfs_device_info *dev_info2 = b;
1877
1878 if (dev_info1->max_avail > dev_info2->max_avail)
1879 return -1;
1880 else if (dev_info1->max_avail < dev_info2->max_avail)
1881 return 1;
1882 return 0;
1883}
1884
1885/*
1886 * sort the devices by max_avail, in which max free extent size of each device
1887 * is stored.(Descending Sort)
1888 */
1889static inline void btrfs_descending_sort_devices(
1890 struct btrfs_device_info *devices,
1891 size_t nr_devices)
1892{
1893 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1894 btrfs_cmp_device_free_bytes, NULL);
1895}
1896
1897/*
1898 * The helper to calc the free space on the devices that can be used to store
1899 * file data.
1900 */
1901static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1902 u64 *free_bytes)
1903{
1904 struct btrfs_device_info *devices_info;
1905 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1906 struct btrfs_device *device;
1907 u64 type;
1908 u64 avail_space;
1909 u64 min_stripe_size;
1910 int num_stripes = 1;
1911 int i = 0, nr_devices;
1912 const struct btrfs_raid_attr *rattr;
1913
1914 /*
1915 * We aren't under the device list lock, so this is racy-ish, but good
1916 * enough for our purposes.
1917 */
1918 nr_devices = fs_info->fs_devices->open_devices;
1919 if (!nr_devices) {
1920 smp_mb();
1921 nr_devices = fs_info->fs_devices->open_devices;
1922 ASSERT(nr_devices);
1923 if (!nr_devices) {
1924 *free_bytes = 0;
1925 return 0;
1926 }
1927 }
1928
1929 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1930 GFP_KERNEL);
1931 if (!devices_info)
1932 return -ENOMEM;
1933
1934 /* calc min stripe number for data space allocation */
1935 type = btrfs_data_alloc_profile(fs_info);
1936 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1937
1938 if (type & BTRFS_BLOCK_GROUP_RAID0)
1939 num_stripes = nr_devices;
1940 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1941 num_stripes = rattr->ncopies;
1942 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1943 num_stripes = 4;
1944
1945 /* Adjust for more than 1 stripe per device */
1946 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1947
1948 rcu_read_lock();
1949 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1950 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1951 &device->dev_state) ||
1952 !device->bdev ||
1953 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1954 continue;
1955
1956 if (i >= nr_devices)
1957 break;
1958
1959 avail_space = device->total_bytes - device->bytes_used;
1960
1961 /* align with stripe_len */
1962 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1963
1964 /*
1965 * Ensure we have at least min_stripe_size on top of the
1966 * reserved space on the device.
1967 */
1968 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1969 continue;
1970
1971 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1972
1973 devices_info[i].dev = device;
1974 devices_info[i].max_avail = avail_space;
1975
1976 i++;
1977 }
1978 rcu_read_unlock();
1979
1980 nr_devices = i;
1981
1982 btrfs_descending_sort_devices(devices_info, nr_devices);
1983
1984 i = nr_devices - 1;
1985 avail_space = 0;
1986 while (nr_devices >= rattr->devs_min) {
1987 num_stripes = min(num_stripes, nr_devices);
1988
1989 if (devices_info[i].max_avail >= min_stripe_size) {
1990 int j;
1991 u64 alloc_size;
1992
1993 avail_space += devices_info[i].max_avail * num_stripes;
1994 alloc_size = devices_info[i].max_avail;
1995 for (j = i + 1 - num_stripes; j <= i; j++)
1996 devices_info[j].max_avail -= alloc_size;
1997 }
1998 i--;
1999 nr_devices--;
2000 }
2001
2002 kfree(devices_info);
2003 *free_bytes = avail_space;
2004 return 0;
2005}
2006
2007/*
2008 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2009 *
2010 * If there's a redundant raid level at DATA block groups, use the respective
2011 * multiplier to scale the sizes.
2012 *
2013 * Unused device space usage is based on simulating the chunk allocator
2014 * algorithm that respects the device sizes and order of allocations. This is
2015 * a close approximation of the actual use but there are other factors that may
2016 * change the result (like a new metadata chunk).
2017 *
2018 * If metadata is exhausted, f_bavail will be 0.
2019 */
2020static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2021{
2022 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2023 struct btrfs_super_block *disk_super = fs_info->super_copy;
2024 struct btrfs_space_info *found;
2025 u64 total_used = 0;
2026 u64 total_free_data = 0;
2027 u64 total_free_meta = 0;
2028 u32 bits = fs_info->sectorsize_bits;
2029 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2030 unsigned factor = 1;
2031 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2032 int ret;
2033 u64 thresh = 0;
2034 int mixed = 0;
2035
2036 list_for_each_entry(found, &fs_info->space_info, list) {
2037 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2038 int i;
2039
2040 total_free_data += found->disk_total - found->disk_used;
2041 total_free_data -=
2042 btrfs_account_ro_block_groups_free_space(found);
2043
2044 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2045 if (!list_empty(&found->block_groups[i]))
2046 factor = btrfs_bg_type_to_factor(
2047 btrfs_raid_array[i].bg_flag);
2048 }
2049 }
2050
2051 /*
2052 * Metadata in mixed block goup profiles are accounted in data
2053 */
2054 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2055 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2056 mixed = 1;
2057 else
2058 total_free_meta += found->disk_total -
2059 found->disk_used;
2060 }
2061
2062 total_used += found->disk_used;
2063 }
2064
2065 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2066 buf->f_blocks >>= bits;
2067 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2068
2069 /* Account global block reserve as used, it's in logical size already */
2070 spin_lock(&block_rsv->lock);
2071 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2072 if (buf->f_bfree >= block_rsv->size >> bits)
2073 buf->f_bfree -= block_rsv->size >> bits;
2074 else
2075 buf->f_bfree = 0;
2076 spin_unlock(&block_rsv->lock);
2077
2078 buf->f_bavail = div_u64(total_free_data, factor);
2079 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2080 if (ret)
2081 return ret;
2082 buf->f_bavail += div_u64(total_free_data, factor);
2083 buf->f_bavail = buf->f_bavail >> bits;
2084
2085 /*
2086 * We calculate the remaining metadata space minus global reserve. If
2087 * this is (supposedly) smaller than zero, there's no space. But this
2088 * does not hold in practice, the exhausted state happens where's still
2089 * some positive delta. So we apply some guesswork and compare the
2090 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2091 *
2092 * We probably cannot calculate the exact threshold value because this
2093 * depends on the internal reservations requested by various
2094 * operations, so some operations that consume a few metadata will
2095 * succeed even if the Avail is zero. But this is better than the other
2096 * way around.
2097 */
2098 thresh = SZ_4M;
2099
2100 /*
2101 * We only want to claim there's no available space if we can no longer
2102 * allocate chunks for our metadata profile and our global reserve will
2103 * not fit in the free metadata space. If we aren't ->full then we
2104 * still can allocate chunks and thus are fine using the currently
2105 * calculated f_bavail.
2106 */
2107 if (!mixed && block_rsv->space_info->full &&
2108 total_free_meta - thresh < block_rsv->size)
2109 buf->f_bavail = 0;
2110
2111 buf->f_type = BTRFS_SUPER_MAGIC;
2112 buf->f_bsize = dentry->d_sb->s_blocksize;
2113 buf->f_namelen = BTRFS_NAME_LEN;
2114
2115 /* We treat it as constant endianness (it doesn't matter _which_)
2116 because we want the fsid to come out the same whether mounted
2117 on a big-endian or little-endian host */
2118 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2119 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2120 /* Mask in the root object ID too, to disambiguate subvols */
2121 buf->f_fsid.val[0] ^=
2122 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2123 buf->f_fsid.val[1] ^=
2124 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2125
2126 return 0;
2127}
2128
2129static void btrfs_kill_super(struct super_block *sb)
2130{
2131 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2132 kill_anon_super(sb);
2133 btrfs_free_fs_info(fs_info);
2134}
2135
2136static struct file_system_type btrfs_fs_type = {
2137 .owner = THIS_MODULE,
2138 .name = "btrfs",
2139 .mount = btrfs_mount,
2140 .kill_sb = btrfs_kill_super,
2141 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2142};
2143
2144static struct file_system_type btrfs_root_fs_type = {
2145 .owner = THIS_MODULE,
2146 .name = "btrfs",
2147 .mount = btrfs_mount_root,
2148 .kill_sb = btrfs_kill_super,
2149 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2150};
2151
2152MODULE_ALIAS_FS("btrfs");
2153
2154static int btrfs_control_open(struct inode *inode, struct file *file)
2155{
2156 /*
2157 * The control file's private_data is used to hold the
2158 * transaction when it is started and is used to keep
2159 * track of whether a transaction is already in progress.
2160 */
2161 file->private_data = NULL;
2162 return 0;
2163}
2164
2165/*
2166 * Used by /dev/btrfs-control for devices ioctls.
2167 */
2168static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2169 unsigned long arg)
2170{
2171 struct btrfs_ioctl_vol_args *vol;
2172 struct btrfs_device *device = NULL;
2173 dev_t devt = 0;
2174 int ret = -ENOTTY;
2175
2176 if (!capable(CAP_SYS_ADMIN))
2177 return -EPERM;
2178
2179 vol = memdup_user((void __user *)arg, sizeof(*vol));
2180 if (IS_ERR(vol))
2181 return PTR_ERR(vol);
2182 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2183
2184 switch (cmd) {
2185 case BTRFS_IOC_SCAN_DEV:
2186 mutex_lock(&uuid_mutex);
2187 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2188 &btrfs_root_fs_type);
2189 ret = PTR_ERR_OR_ZERO(device);
2190 mutex_unlock(&uuid_mutex);
2191 break;
2192 case BTRFS_IOC_FORGET_DEV:
2193 if (vol->name[0] != 0) {
2194 ret = lookup_bdev(vol->name, &devt);
2195 if (ret)
2196 break;
2197 }
2198 ret = btrfs_forget_devices(devt);
2199 break;
2200 case BTRFS_IOC_DEVICES_READY:
2201 mutex_lock(&uuid_mutex);
2202 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2203 &btrfs_root_fs_type);
2204 if (IS_ERR(device)) {
2205 mutex_unlock(&uuid_mutex);
2206 ret = PTR_ERR(device);
2207 break;
2208 }
2209 ret = !(device->fs_devices->num_devices ==
2210 device->fs_devices->total_devices);
2211 mutex_unlock(&uuid_mutex);
2212 break;
2213 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2214 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2215 break;
2216 }
2217
2218 kfree(vol);
2219 return ret;
2220}
2221
2222static int btrfs_freeze(struct super_block *sb)
2223{
2224 struct btrfs_trans_handle *trans;
2225 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2226 struct btrfs_root *root = fs_info->tree_root;
2227
2228 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2229 /*
2230 * We don't need a barrier here, we'll wait for any transaction that
2231 * could be in progress on other threads (and do delayed iputs that
2232 * we want to avoid on a frozen filesystem), or do the commit
2233 * ourselves.
2234 */
2235 trans = btrfs_attach_transaction_barrier(root);
2236 if (IS_ERR(trans)) {
2237 /* no transaction, don't bother */
2238 if (PTR_ERR(trans) == -ENOENT)
2239 return 0;
2240 return PTR_ERR(trans);
2241 }
2242 return btrfs_commit_transaction(trans);
2243}
2244
2245static int check_dev_super(struct btrfs_device *dev)
2246{
2247 struct btrfs_fs_info *fs_info = dev->fs_info;
2248 struct btrfs_super_block *sb;
2249 u16 csum_type;
2250 int ret = 0;
2251
2252 /* This should be called with fs still frozen. */
2253 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2254
2255 /* Missing dev, no need to check. */
2256 if (!dev->bdev)
2257 return 0;
2258
2259 /* Only need to check the primary super block. */
2260 sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2261 if (IS_ERR(sb))
2262 return PTR_ERR(sb);
2263
2264 /* Verify the checksum. */
2265 csum_type = btrfs_super_csum_type(sb);
2266 if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2267 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2268 csum_type, btrfs_super_csum_type(fs_info->super_copy));
2269 ret = -EUCLEAN;
2270 goto out;
2271 }
2272
2273 if (btrfs_check_super_csum(fs_info, sb)) {
2274 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2275 ret = -EUCLEAN;
2276 goto out;
2277 }
2278
2279 /* Btrfs_validate_super() includes fsid check against super->fsid. */
2280 ret = btrfs_validate_super(fs_info, sb, 0);
2281 if (ret < 0)
2282 goto out;
2283
2284 if (btrfs_super_generation(sb) != fs_info->last_trans_committed) {
2285 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2286 btrfs_super_generation(sb),
2287 fs_info->last_trans_committed);
2288 ret = -EUCLEAN;
2289 goto out;
2290 }
2291out:
2292 btrfs_release_disk_super(sb);
2293 return ret;
2294}
2295
2296static int btrfs_unfreeze(struct super_block *sb)
2297{
2298 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2299 struct btrfs_device *device;
2300 int ret = 0;
2301
2302 /*
2303 * Make sure the fs is not changed by accident (like hibernation then
2304 * modified by other OS).
2305 * If we found anything wrong, we mark the fs error immediately.
2306 *
2307 * And since the fs is frozen, no one can modify the fs yet, thus
2308 * we don't need to hold device_list_mutex.
2309 */
2310 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2311 ret = check_dev_super(device);
2312 if (ret < 0) {
2313 btrfs_handle_fs_error(fs_info, ret,
2314 "super block on devid %llu got modified unexpectedly",
2315 device->devid);
2316 break;
2317 }
2318 }
2319 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2320
2321 /*
2322 * We still return 0, to allow VFS layer to unfreeze the fs even the
2323 * above checks failed. Since the fs is either fine or read-only, we're
2324 * safe to continue, without causing further damage.
2325 */
2326 return 0;
2327}
2328
2329static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2330{
2331 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2332
2333 /*
2334 * There should be always a valid pointer in latest_dev, it may be stale
2335 * for a short moment in case it's being deleted but still valid until
2336 * the end of RCU grace period.
2337 */
2338 rcu_read_lock();
2339 seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2340 rcu_read_unlock();
2341
2342 return 0;
2343}
2344
2345static const struct super_operations btrfs_super_ops = {
2346 .drop_inode = btrfs_drop_inode,
2347 .evict_inode = btrfs_evict_inode,
2348 .put_super = btrfs_put_super,
2349 .sync_fs = btrfs_sync_fs,
2350 .show_options = btrfs_show_options,
2351 .show_devname = btrfs_show_devname,
2352 .alloc_inode = btrfs_alloc_inode,
2353 .destroy_inode = btrfs_destroy_inode,
2354 .free_inode = btrfs_free_inode,
2355 .statfs = btrfs_statfs,
2356 .remount_fs = btrfs_remount,
2357 .freeze_fs = btrfs_freeze,
2358 .unfreeze_fs = btrfs_unfreeze,
2359};
2360
2361static const struct file_operations btrfs_ctl_fops = {
2362 .open = btrfs_control_open,
2363 .unlocked_ioctl = btrfs_control_ioctl,
2364 .compat_ioctl = compat_ptr_ioctl,
2365 .owner = THIS_MODULE,
2366 .llseek = noop_llseek,
2367};
2368
2369static struct miscdevice btrfs_misc = {
2370 .minor = BTRFS_MINOR,
2371 .name = "btrfs-control",
2372 .fops = &btrfs_ctl_fops
2373};
2374
2375MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2376MODULE_ALIAS("devname:btrfs-control");
2377
2378static int __init btrfs_interface_init(void)
2379{
2380 return misc_register(&btrfs_misc);
2381}
2382
2383static __cold void btrfs_interface_exit(void)
2384{
2385 misc_deregister(&btrfs_misc);
2386}
2387
2388static int __init btrfs_print_mod_info(void)
2389{
2390 static const char options[] = ""
2391#ifdef CONFIG_BTRFS_DEBUG
2392 ", debug=on"
2393#endif
2394#ifdef CONFIG_BTRFS_ASSERT
2395 ", assert=on"
2396#endif
2397#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2398 ", integrity-checker=on"
2399#endif
2400#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2401 ", ref-verify=on"
2402#endif
2403#ifdef CONFIG_BLK_DEV_ZONED
2404 ", zoned=yes"
2405#else
2406 ", zoned=no"
2407#endif
2408#ifdef CONFIG_FS_VERITY
2409 ", fsverity=yes"
2410#else
2411 ", fsverity=no"
2412#endif
2413 ;
2414 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2415 return 0;
2416}
2417
2418static int register_btrfs(void)
2419{
2420 return register_filesystem(&btrfs_fs_type);
2421}
2422
2423static void unregister_btrfs(void)
2424{
2425 unregister_filesystem(&btrfs_fs_type);
2426}
2427
2428/* Helper structure for long init/exit functions. */
2429struct init_sequence {
2430 int (*init_func)(void);
2431 /* Can be NULL if the init_func doesn't need cleanup. */
2432 void (*exit_func)(void);
2433};
2434
2435static const struct init_sequence mod_init_seq[] = {
2436 {
2437 .init_func = btrfs_props_init,
2438 .exit_func = NULL,
2439 }, {
2440 .init_func = btrfs_init_sysfs,
2441 .exit_func = btrfs_exit_sysfs,
2442 }, {
2443 .init_func = btrfs_init_compress,
2444 .exit_func = btrfs_exit_compress,
2445 }, {
2446 .init_func = btrfs_init_cachep,
2447 .exit_func = btrfs_destroy_cachep,
2448 }, {
2449 .init_func = btrfs_transaction_init,
2450 .exit_func = btrfs_transaction_exit,
2451 }, {
2452 .init_func = btrfs_ctree_init,
2453 .exit_func = btrfs_ctree_exit,
2454 }, {
2455 .init_func = btrfs_free_space_init,
2456 .exit_func = btrfs_free_space_exit,
2457 }, {
2458 .init_func = extent_state_init_cachep,
2459 .exit_func = extent_state_free_cachep,
2460 }, {
2461 .init_func = extent_buffer_init_cachep,
2462 .exit_func = extent_buffer_free_cachep,
2463 }, {
2464 .init_func = btrfs_bioset_init,
2465 .exit_func = btrfs_bioset_exit,
2466 }, {
2467 .init_func = extent_map_init,
2468 .exit_func = extent_map_exit,
2469 }, {
2470 .init_func = ordered_data_init,
2471 .exit_func = ordered_data_exit,
2472 }, {
2473 .init_func = btrfs_delayed_inode_init,
2474 .exit_func = btrfs_delayed_inode_exit,
2475 }, {
2476 .init_func = btrfs_auto_defrag_init,
2477 .exit_func = btrfs_auto_defrag_exit,
2478 }, {
2479 .init_func = btrfs_delayed_ref_init,
2480 .exit_func = btrfs_delayed_ref_exit,
2481 }, {
2482 .init_func = btrfs_prelim_ref_init,
2483 .exit_func = btrfs_prelim_ref_exit,
2484 }, {
2485 .init_func = btrfs_interface_init,
2486 .exit_func = btrfs_interface_exit,
2487 }, {
2488 .init_func = btrfs_print_mod_info,
2489 .exit_func = NULL,
2490 }, {
2491 .init_func = btrfs_run_sanity_tests,
2492 .exit_func = NULL,
2493 }, {
2494 .init_func = register_btrfs,
2495 .exit_func = unregister_btrfs,
2496 }
2497};
2498
2499static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2500
2501static __always_inline void btrfs_exit_btrfs_fs(void)
2502{
2503 int i;
2504
2505 for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2506 if (!mod_init_result[i])
2507 continue;
2508 if (mod_init_seq[i].exit_func)
2509 mod_init_seq[i].exit_func();
2510 mod_init_result[i] = false;
2511 }
2512}
2513
2514static void __exit exit_btrfs_fs(void)
2515{
2516 btrfs_exit_btrfs_fs();
2517 btrfs_cleanup_fs_uuids();
2518}
2519
2520static int __init init_btrfs_fs(void)
2521{
2522 int ret;
2523 int i;
2524
2525 for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2526 ASSERT(!mod_init_result[i]);
2527 ret = mod_init_seq[i].init_func();
2528 if (ret < 0) {
2529 btrfs_exit_btrfs_fs();
2530 return ret;
2531 }
2532 mod_init_result[i] = true;
2533 }
2534 return 0;
2535}
2536
2537late_initcall(init_btrfs_fs);
2538module_exit(exit_btrfs_fs)
2539
2540MODULE_LICENSE("GPL");
2541MODULE_SOFTDEP("pre: crc32c");
2542MODULE_SOFTDEP("pre: xxhash64");
2543MODULE_SOFTDEP("pre: sha256");
2544MODULE_SOFTDEP("pre: blake2b-256");
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (C) 2007 Oracle. All rights reserved.
4 */
5
6#include <linux/blkdev.h>
7#include <linux/module.h>
8#include <linux/fs.h>
9#include <linux/pagemap.h>
10#include <linux/highmem.h>
11#include <linux/time.h>
12#include <linux/init.h>
13#include <linux/seq_file.h>
14#include <linux/string.h>
15#include <linux/backing-dev.h>
16#include <linux/mount.h>
17#include <linux/writeback.h>
18#include <linux/statfs.h>
19#include <linux/compat.h>
20#include <linux/parser.h>
21#include <linux/ctype.h>
22#include <linux/namei.h>
23#include <linux/miscdevice.h>
24#include <linux/magic.h>
25#include <linux/slab.h>
26#include <linux/cleancache.h>
27#include <linux/ratelimit.h>
28#include <linux/crc32c.h>
29#include <linux/btrfs.h>
30#include "delayed-inode.h"
31#include "ctree.h"
32#include "disk-io.h"
33#include "transaction.h"
34#include "btrfs_inode.h"
35#include "print-tree.h"
36#include "props.h"
37#include "xattr.h"
38#include "volumes.h"
39#include "export.h"
40#include "compression.h"
41#include "rcu-string.h"
42#include "dev-replace.h"
43#include "free-space-cache.h"
44#include "backref.h"
45#include "space-info.h"
46#include "sysfs.h"
47#include "tests/btrfs-tests.h"
48#include "block-group.h"
49#include "discard.h"
50
51#include "qgroup.h"
52#define CREATE_TRACE_POINTS
53#include <trace/events/btrfs.h>
54
55static const struct super_operations btrfs_super_ops;
56
57/*
58 * Types for mounting the default subvolume and a subvolume explicitly
59 * requested by subvol=/path. That way the callchain is straightforward and we
60 * don't have to play tricks with the mount options and recursive calls to
61 * btrfs_mount.
62 *
63 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
64 */
65static struct file_system_type btrfs_fs_type;
66static struct file_system_type btrfs_root_fs_type;
67
68static int btrfs_remount(struct super_block *sb, int *flags, char *data);
69
70/*
71 * Generally the error codes correspond to their respective errors, but there
72 * are a few special cases.
73 *
74 * EUCLEAN: Any sort of corruption that we encounter. The tree-checker for
75 * instance will return EUCLEAN if any of the blocks are corrupted in
76 * a way that is problematic. We want to reserve EUCLEAN for these
77 * sort of corruptions.
78 *
79 * EROFS: If we check BTRFS_FS_STATE_ERROR and fail out with a return error, we
80 * need to use EROFS for this case. We will have no idea of the
81 * original failure, that will have been reported at the time we tripped
82 * over the error. Each subsequent error that doesn't have any context
83 * of the original error should use EROFS when handling BTRFS_FS_STATE_ERROR.
84 */
85const char * __attribute_const__ btrfs_decode_error(int errno)
86{
87 char *errstr = "unknown";
88
89 switch (errno) {
90 case -ENOENT: /* -2 */
91 errstr = "No such entry";
92 break;
93 case -EIO: /* -5 */
94 errstr = "IO failure";
95 break;
96 case -ENOMEM: /* -12*/
97 errstr = "Out of memory";
98 break;
99 case -EEXIST: /* -17 */
100 errstr = "Object already exists";
101 break;
102 case -ENOSPC: /* -28 */
103 errstr = "No space left";
104 break;
105 case -EROFS: /* -30 */
106 errstr = "Readonly filesystem";
107 break;
108 case -EOPNOTSUPP: /* -95 */
109 errstr = "Operation not supported";
110 break;
111 case -EUCLEAN: /* -117 */
112 errstr = "Filesystem corrupted";
113 break;
114 case -EDQUOT: /* -122 */
115 errstr = "Quota exceeded";
116 break;
117 }
118
119 return errstr;
120}
121
122/*
123 * __btrfs_handle_fs_error decodes expected errors from the caller and
124 * invokes the appropriate error response.
125 */
126__cold
127void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
128 unsigned int line, int errno, const char *fmt, ...)
129{
130 struct super_block *sb = fs_info->sb;
131#ifdef CONFIG_PRINTK
132 const char *errstr;
133#endif
134
135 /*
136 * Special case: if the error is EROFS, and we're already
137 * under SB_RDONLY, then it is safe here.
138 */
139 if (errno == -EROFS && sb_rdonly(sb))
140 return;
141
142#ifdef CONFIG_PRINTK
143 errstr = btrfs_decode_error(errno);
144 if (fmt) {
145 struct va_format vaf;
146 va_list args;
147
148 va_start(args, fmt);
149 vaf.fmt = fmt;
150 vaf.va = &args;
151
152 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
153 sb->s_id, function, line, errno, errstr, &vaf);
154 va_end(args);
155 } else {
156 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
157 sb->s_id, function, line, errno, errstr);
158 }
159#endif
160
161 /*
162 * Today we only save the error info to memory. Long term we'll
163 * also send it down to the disk
164 */
165 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
166
167 /* Don't go through full error handling during mount */
168 if (!(sb->s_flags & SB_BORN))
169 return;
170
171 if (sb_rdonly(sb))
172 return;
173
174 btrfs_discard_stop(fs_info);
175
176 /* btrfs handle error by forcing the filesystem readonly */
177 sb->s_flags |= SB_RDONLY;
178 btrfs_info(fs_info, "forced readonly");
179 /*
180 * Note that a running device replace operation is not canceled here
181 * although there is no way to update the progress. It would add the
182 * risk of a deadlock, therefore the canceling is omitted. The only
183 * penalty is that some I/O remains active until the procedure
184 * completes. The next time when the filesystem is mounted writable
185 * again, the device replace operation continues.
186 */
187}
188
189#ifdef CONFIG_PRINTK
190static const char * const logtypes[] = {
191 "emergency",
192 "alert",
193 "critical",
194 "error",
195 "warning",
196 "notice",
197 "info",
198 "debug",
199};
200
201
202/*
203 * Use one ratelimit state per log level so that a flood of less important
204 * messages doesn't cause more important ones to be dropped.
205 */
206static struct ratelimit_state printk_limits[] = {
207 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
208 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
209 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
210 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
211 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
212 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
213 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
214 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
215};
216
217void __cold btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
218{
219 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
220 struct va_format vaf;
221 va_list args;
222 int kern_level;
223 const char *type = logtypes[4];
224 struct ratelimit_state *ratelimit = &printk_limits[4];
225
226 va_start(args, fmt);
227
228 while ((kern_level = printk_get_level(fmt)) != 0) {
229 size_t size = printk_skip_level(fmt) - fmt;
230
231 if (kern_level >= '0' && kern_level <= '7') {
232 memcpy(lvl, fmt, size);
233 lvl[size] = '\0';
234 type = logtypes[kern_level - '0'];
235 ratelimit = &printk_limits[kern_level - '0'];
236 }
237 fmt += size;
238 }
239
240 vaf.fmt = fmt;
241 vaf.va = &args;
242
243 if (__ratelimit(ratelimit))
244 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
245 fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
246
247 va_end(args);
248}
249#endif
250
251/*
252 * We only mark the transaction aborted and then set the file system read-only.
253 * This will prevent new transactions from starting or trying to join this
254 * one.
255 *
256 * This means that error recovery at the call site is limited to freeing
257 * any local memory allocations and passing the error code up without
258 * further cleanup. The transaction should complete as it normally would
259 * in the call path but will return -EIO.
260 *
261 * We'll complete the cleanup in btrfs_end_transaction and
262 * btrfs_commit_transaction.
263 */
264__cold
265void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
266 const char *function,
267 unsigned int line, int errno)
268{
269 struct btrfs_fs_info *fs_info = trans->fs_info;
270
271 WRITE_ONCE(trans->aborted, errno);
272 /* Nothing used. The other threads that have joined this
273 * transaction may be able to continue. */
274 if (!trans->dirty && list_empty(&trans->new_bgs)) {
275 const char *errstr;
276
277 errstr = btrfs_decode_error(errno);
278 btrfs_warn(fs_info,
279 "%s:%d: Aborting unused transaction(%s).",
280 function, line, errstr);
281 return;
282 }
283 WRITE_ONCE(trans->transaction->aborted, errno);
284 /* Wake up anybody who may be waiting on this transaction */
285 wake_up(&fs_info->transaction_wait);
286 wake_up(&fs_info->transaction_blocked_wait);
287 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
288}
289/*
290 * __btrfs_panic decodes unexpected, fatal errors from the caller,
291 * issues an alert, and either panics or BUGs, depending on mount options.
292 */
293__cold
294void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
295 unsigned int line, int errno, const char *fmt, ...)
296{
297 char *s_id = "<unknown>";
298 const char *errstr;
299 struct va_format vaf = { .fmt = fmt };
300 va_list args;
301
302 if (fs_info)
303 s_id = fs_info->sb->s_id;
304
305 va_start(args, fmt);
306 vaf.va = &args;
307
308 errstr = btrfs_decode_error(errno);
309 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
310 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
311 s_id, function, line, &vaf, errno, errstr);
312
313 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
314 function, line, &vaf, errno, errstr);
315 va_end(args);
316 /* Caller calls BUG() */
317}
318
319static void btrfs_put_super(struct super_block *sb)
320{
321 close_ctree(btrfs_sb(sb));
322}
323
324enum {
325 Opt_acl, Opt_noacl,
326 Opt_clear_cache,
327 Opt_commit_interval,
328 Opt_compress,
329 Opt_compress_force,
330 Opt_compress_force_type,
331 Opt_compress_type,
332 Opt_degraded,
333 Opt_device,
334 Opt_fatal_errors,
335 Opt_flushoncommit, Opt_noflushoncommit,
336 Opt_inode_cache, Opt_noinode_cache,
337 Opt_max_inline,
338 Opt_barrier, Opt_nobarrier,
339 Opt_datacow, Opt_nodatacow,
340 Opt_datasum, Opt_nodatasum,
341 Opt_defrag, Opt_nodefrag,
342 Opt_discard, Opt_nodiscard,
343 Opt_discard_mode,
344 Opt_norecovery,
345 Opt_ratio,
346 Opt_rescan_uuid_tree,
347 Opt_skip_balance,
348 Opt_space_cache, Opt_no_space_cache,
349 Opt_space_cache_version,
350 Opt_ssd, Opt_nossd,
351 Opt_ssd_spread, Opt_nossd_spread,
352 Opt_subvol,
353 Opt_subvol_empty,
354 Opt_subvolid,
355 Opt_thread_pool,
356 Opt_treelog, Opt_notreelog,
357 Opt_user_subvol_rm_allowed,
358
359 /* Rescue options */
360 Opt_rescue,
361 Opt_usebackuproot,
362 Opt_nologreplay,
363
364 /* Deprecated options */
365 Opt_recovery,
366
367 /* Debugging options */
368 Opt_check_integrity,
369 Opt_check_integrity_including_extent_data,
370 Opt_check_integrity_print_mask,
371 Opt_enospc_debug, Opt_noenospc_debug,
372#ifdef CONFIG_BTRFS_DEBUG
373 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
374#endif
375#ifdef CONFIG_BTRFS_FS_REF_VERIFY
376 Opt_ref_verify,
377#endif
378 Opt_err,
379};
380
381static const match_table_t tokens = {
382 {Opt_acl, "acl"},
383 {Opt_noacl, "noacl"},
384 {Opt_clear_cache, "clear_cache"},
385 {Opt_commit_interval, "commit=%u"},
386 {Opt_compress, "compress"},
387 {Opt_compress_type, "compress=%s"},
388 {Opt_compress_force, "compress-force"},
389 {Opt_compress_force_type, "compress-force=%s"},
390 {Opt_degraded, "degraded"},
391 {Opt_device, "device=%s"},
392 {Opt_fatal_errors, "fatal_errors=%s"},
393 {Opt_flushoncommit, "flushoncommit"},
394 {Opt_noflushoncommit, "noflushoncommit"},
395 {Opt_inode_cache, "inode_cache"},
396 {Opt_noinode_cache, "noinode_cache"},
397 {Opt_max_inline, "max_inline=%s"},
398 {Opt_barrier, "barrier"},
399 {Opt_nobarrier, "nobarrier"},
400 {Opt_datacow, "datacow"},
401 {Opt_nodatacow, "nodatacow"},
402 {Opt_datasum, "datasum"},
403 {Opt_nodatasum, "nodatasum"},
404 {Opt_defrag, "autodefrag"},
405 {Opt_nodefrag, "noautodefrag"},
406 {Opt_discard, "discard"},
407 {Opt_discard_mode, "discard=%s"},
408 {Opt_nodiscard, "nodiscard"},
409 {Opt_norecovery, "norecovery"},
410 {Opt_ratio, "metadata_ratio=%u"},
411 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
412 {Opt_skip_balance, "skip_balance"},
413 {Opt_space_cache, "space_cache"},
414 {Opt_no_space_cache, "nospace_cache"},
415 {Opt_space_cache_version, "space_cache=%s"},
416 {Opt_ssd, "ssd"},
417 {Opt_nossd, "nossd"},
418 {Opt_ssd_spread, "ssd_spread"},
419 {Opt_nossd_spread, "nossd_spread"},
420 {Opt_subvol, "subvol=%s"},
421 {Opt_subvol_empty, "subvol="},
422 {Opt_subvolid, "subvolid=%s"},
423 {Opt_thread_pool, "thread_pool=%u"},
424 {Opt_treelog, "treelog"},
425 {Opt_notreelog, "notreelog"},
426 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
427
428 /* Rescue options */
429 {Opt_rescue, "rescue=%s"},
430 /* Deprecated, with alias rescue=nologreplay */
431 {Opt_nologreplay, "nologreplay"},
432 /* Deprecated, with alias rescue=usebackuproot */
433 {Opt_usebackuproot, "usebackuproot"},
434
435 /* Deprecated options */
436 {Opt_recovery, "recovery"},
437
438 /* Debugging options */
439 {Opt_check_integrity, "check_int"},
440 {Opt_check_integrity_including_extent_data, "check_int_data"},
441 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
442 {Opt_enospc_debug, "enospc_debug"},
443 {Opt_noenospc_debug, "noenospc_debug"},
444#ifdef CONFIG_BTRFS_DEBUG
445 {Opt_fragment_data, "fragment=data"},
446 {Opt_fragment_metadata, "fragment=metadata"},
447 {Opt_fragment_all, "fragment=all"},
448#endif
449#ifdef CONFIG_BTRFS_FS_REF_VERIFY
450 {Opt_ref_verify, "ref_verify"},
451#endif
452 {Opt_err, NULL},
453};
454
455static const match_table_t rescue_tokens = {
456 {Opt_usebackuproot, "usebackuproot"},
457 {Opt_nologreplay, "nologreplay"},
458 {Opt_err, NULL},
459};
460
461static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
462{
463 char *opts;
464 char *orig;
465 char *p;
466 substring_t args[MAX_OPT_ARGS];
467 int ret = 0;
468
469 opts = kstrdup(options, GFP_KERNEL);
470 if (!opts)
471 return -ENOMEM;
472 orig = opts;
473
474 while ((p = strsep(&opts, ":")) != NULL) {
475 int token;
476
477 if (!*p)
478 continue;
479 token = match_token(p, rescue_tokens, args);
480 switch (token){
481 case Opt_usebackuproot:
482 btrfs_info(info,
483 "trying to use backup root at mount time");
484 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
485 break;
486 case Opt_nologreplay:
487 btrfs_set_and_info(info, NOLOGREPLAY,
488 "disabling log replay at mount time");
489 break;
490 case Opt_err:
491 btrfs_info(info, "unrecognized rescue option '%s'", p);
492 ret = -EINVAL;
493 goto out;
494 default:
495 break;
496 }
497
498 }
499out:
500 kfree(orig);
501 return ret;
502}
503
504/*
505 * Regular mount options parser. Everything that is needed only when
506 * reading in a new superblock is parsed here.
507 * XXX JDM: This needs to be cleaned up for remount.
508 */
509int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
510 unsigned long new_flags)
511{
512 substring_t args[MAX_OPT_ARGS];
513 char *p, *num;
514 u64 cache_gen;
515 int intarg;
516 int ret = 0;
517 char *compress_type;
518 bool compress_force = false;
519 enum btrfs_compression_type saved_compress_type;
520 int saved_compress_level;
521 bool saved_compress_force;
522 int no_compress = 0;
523
524 cache_gen = btrfs_super_cache_generation(info->super_copy);
525 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
526 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
527 else if (cache_gen)
528 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
529
530 /*
531 * Even the options are empty, we still need to do extra check
532 * against new flags
533 */
534 if (!options)
535 goto check;
536
537 while ((p = strsep(&options, ",")) != NULL) {
538 int token;
539 if (!*p)
540 continue;
541
542 token = match_token(p, tokens, args);
543 switch (token) {
544 case Opt_degraded:
545 btrfs_info(info, "allowing degraded mounts");
546 btrfs_set_opt(info->mount_opt, DEGRADED);
547 break;
548 case Opt_subvol:
549 case Opt_subvol_empty:
550 case Opt_subvolid:
551 case Opt_device:
552 /*
553 * These are parsed by btrfs_parse_subvol_options or
554 * btrfs_parse_device_options and can be ignored here.
555 */
556 break;
557 case Opt_nodatasum:
558 btrfs_set_and_info(info, NODATASUM,
559 "setting nodatasum");
560 break;
561 case Opt_datasum:
562 if (btrfs_test_opt(info, NODATASUM)) {
563 if (btrfs_test_opt(info, NODATACOW))
564 btrfs_info(info,
565 "setting datasum, datacow enabled");
566 else
567 btrfs_info(info, "setting datasum");
568 }
569 btrfs_clear_opt(info->mount_opt, NODATACOW);
570 btrfs_clear_opt(info->mount_opt, NODATASUM);
571 break;
572 case Opt_nodatacow:
573 if (!btrfs_test_opt(info, NODATACOW)) {
574 if (!btrfs_test_opt(info, COMPRESS) ||
575 !btrfs_test_opt(info, FORCE_COMPRESS)) {
576 btrfs_info(info,
577 "setting nodatacow, compression disabled");
578 } else {
579 btrfs_info(info, "setting nodatacow");
580 }
581 }
582 btrfs_clear_opt(info->mount_opt, COMPRESS);
583 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
584 btrfs_set_opt(info->mount_opt, NODATACOW);
585 btrfs_set_opt(info->mount_opt, NODATASUM);
586 break;
587 case Opt_datacow:
588 btrfs_clear_and_info(info, NODATACOW,
589 "setting datacow");
590 break;
591 case Opt_compress_force:
592 case Opt_compress_force_type:
593 compress_force = true;
594 fallthrough;
595 case Opt_compress:
596 case Opt_compress_type:
597 saved_compress_type = btrfs_test_opt(info,
598 COMPRESS) ?
599 info->compress_type : BTRFS_COMPRESS_NONE;
600 saved_compress_force =
601 btrfs_test_opt(info, FORCE_COMPRESS);
602 saved_compress_level = info->compress_level;
603 if (token == Opt_compress ||
604 token == Opt_compress_force ||
605 strncmp(args[0].from, "zlib", 4) == 0) {
606 compress_type = "zlib";
607
608 info->compress_type = BTRFS_COMPRESS_ZLIB;
609 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
610 /*
611 * args[0] contains uninitialized data since
612 * for these tokens we don't expect any
613 * parameter.
614 */
615 if (token != Opt_compress &&
616 token != Opt_compress_force)
617 info->compress_level =
618 btrfs_compress_str2level(
619 BTRFS_COMPRESS_ZLIB,
620 args[0].from + 4);
621 btrfs_set_opt(info->mount_opt, COMPRESS);
622 btrfs_clear_opt(info->mount_opt, NODATACOW);
623 btrfs_clear_opt(info->mount_opt, NODATASUM);
624 no_compress = 0;
625 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
626 compress_type = "lzo";
627 info->compress_type = BTRFS_COMPRESS_LZO;
628 info->compress_level = 0;
629 btrfs_set_opt(info->mount_opt, COMPRESS);
630 btrfs_clear_opt(info->mount_opt, NODATACOW);
631 btrfs_clear_opt(info->mount_opt, NODATASUM);
632 btrfs_set_fs_incompat(info, COMPRESS_LZO);
633 no_compress = 0;
634 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
635 compress_type = "zstd";
636 info->compress_type = BTRFS_COMPRESS_ZSTD;
637 info->compress_level =
638 btrfs_compress_str2level(
639 BTRFS_COMPRESS_ZSTD,
640 args[0].from + 4);
641 btrfs_set_opt(info->mount_opt, COMPRESS);
642 btrfs_clear_opt(info->mount_opt, NODATACOW);
643 btrfs_clear_opt(info->mount_opt, NODATASUM);
644 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
645 no_compress = 0;
646 } else if (strncmp(args[0].from, "no", 2) == 0) {
647 compress_type = "no";
648 info->compress_level = 0;
649 info->compress_type = 0;
650 btrfs_clear_opt(info->mount_opt, COMPRESS);
651 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
652 compress_force = false;
653 no_compress++;
654 } else {
655 ret = -EINVAL;
656 goto out;
657 }
658
659 if (compress_force) {
660 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
661 } else {
662 /*
663 * If we remount from compress-force=xxx to
664 * compress=xxx, we need clear FORCE_COMPRESS
665 * flag, otherwise, there is no way for users
666 * to disable forcible compression separately.
667 */
668 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
669 }
670 if (no_compress == 1) {
671 btrfs_info(info, "use no compression");
672 } else if ((info->compress_type != saved_compress_type) ||
673 (compress_force != saved_compress_force) ||
674 (info->compress_level != saved_compress_level)) {
675 btrfs_info(info, "%s %s compression, level %d",
676 (compress_force) ? "force" : "use",
677 compress_type, info->compress_level);
678 }
679 compress_force = false;
680 break;
681 case Opt_ssd:
682 btrfs_set_and_info(info, SSD,
683 "enabling ssd optimizations");
684 btrfs_clear_opt(info->mount_opt, NOSSD);
685 break;
686 case Opt_ssd_spread:
687 btrfs_set_and_info(info, SSD,
688 "enabling ssd optimizations");
689 btrfs_set_and_info(info, SSD_SPREAD,
690 "using spread ssd allocation scheme");
691 btrfs_clear_opt(info->mount_opt, NOSSD);
692 break;
693 case Opt_nossd:
694 btrfs_set_opt(info->mount_opt, NOSSD);
695 btrfs_clear_and_info(info, SSD,
696 "not using ssd optimizations");
697 fallthrough;
698 case Opt_nossd_spread:
699 btrfs_clear_and_info(info, SSD_SPREAD,
700 "not using spread ssd allocation scheme");
701 break;
702 case Opt_barrier:
703 btrfs_clear_and_info(info, NOBARRIER,
704 "turning on barriers");
705 break;
706 case Opt_nobarrier:
707 btrfs_set_and_info(info, NOBARRIER,
708 "turning off barriers");
709 break;
710 case Opt_thread_pool:
711 ret = match_int(&args[0], &intarg);
712 if (ret) {
713 goto out;
714 } else if (intarg == 0) {
715 ret = -EINVAL;
716 goto out;
717 }
718 info->thread_pool_size = intarg;
719 break;
720 case Opt_max_inline:
721 num = match_strdup(&args[0]);
722 if (num) {
723 info->max_inline = memparse(num, NULL);
724 kfree(num);
725
726 if (info->max_inline) {
727 info->max_inline = min_t(u64,
728 info->max_inline,
729 info->sectorsize);
730 }
731 btrfs_info(info, "max_inline at %llu",
732 info->max_inline);
733 } else {
734 ret = -ENOMEM;
735 goto out;
736 }
737 break;
738 case Opt_acl:
739#ifdef CONFIG_BTRFS_FS_POSIX_ACL
740 info->sb->s_flags |= SB_POSIXACL;
741 break;
742#else
743 btrfs_err(info, "support for ACL not compiled in!");
744 ret = -EINVAL;
745 goto out;
746#endif
747 case Opt_noacl:
748 info->sb->s_flags &= ~SB_POSIXACL;
749 break;
750 case Opt_notreelog:
751 btrfs_set_and_info(info, NOTREELOG,
752 "disabling tree log");
753 break;
754 case Opt_treelog:
755 btrfs_clear_and_info(info, NOTREELOG,
756 "enabling tree log");
757 break;
758 case Opt_norecovery:
759 case Opt_nologreplay:
760 btrfs_warn(info,
761 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
762 btrfs_set_and_info(info, NOLOGREPLAY,
763 "disabling log replay at mount time");
764 break;
765 case Opt_flushoncommit:
766 btrfs_set_and_info(info, FLUSHONCOMMIT,
767 "turning on flush-on-commit");
768 break;
769 case Opt_noflushoncommit:
770 btrfs_clear_and_info(info, FLUSHONCOMMIT,
771 "turning off flush-on-commit");
772 break;
773 case Opt_ratio:
774 ret = match_int(&args[0], &intarg);
775 if (ret)
776 goto out;
777 info->metadata_ratio = intarg;
778 btrfs_info(info, "metadata ratio %u",
779 info->metadata_ratio);
780 break;
781 case Opt_discard:
782 case Opt_discard_mode:
783 if (token == Opt_discard ||
784 strcmp(args[0].from, "sync") == 0) {
785 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
786 btrfs_set_and_info(info, DISCARD_SYNC,
787 "turning on sync discard");
788 } else if (strcmp(args[0].from, "async") == 0) {
789 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
790 btrfs_set_and_info(info, DISCARD_ASYNC,
791 "turning on async discard");
792 } else {
793 ret = -EINVAL;
794 goto out;
795 }
796 break;
797 case Opt_nodiscard:
798 btrfs_clear_and_info(info, DISCARD_SYNC,
799 "turning off discard");
800 btrfs_clear_and_info(info, DISCARD_ASYNC,
801 "turning off async discard");
802 break;
803 case Opt_space_cache:
804 case Opt_space_cache_version:
805 if (token == Opt_space_cache ||
806 strcmp(args[0].from, "v1") == 0) {
807 btrfs_clear_opt(info->mount_opt,
808 FREE_SPACE_TREE);
809 btrfs_set_and_info(info, SPACE_CACHE,
810 "enabling disk space caching");
811 } else if (strcmp(args[0].from, "v2") == 0) {
812 btrfs_clear_opt(info->mount_opt,
813 SPACE_CACHE);
814 btrfs_set_and_info(info, FREE_SPACE_TREE,
815 "enabling free space tree");
816 } else {
817 ret = -EINVAL;
818 goto out;
819 }
820 break;
821 case Opt_rescan_uuid_tree:
822 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
823 break;
824 case Opt_no_space_cache:
825 if (btrfs_test_opt(info, SPACE_CACHE)) {
826 btrfs_clear_and_info(info, SPACE_CACHE,
827 "disabling disk space caching");
828 }
829 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
830 btrfs_clear_and_info(info, FREE_SPACE_TREE,
831 "disabling free space tree");
832 }
833 break;
834 case Opt_inode_cache:
835 btrfs_warn(info,
836 "the 'inode_cache' option is deprecated and will have no effect from 5.11");
837 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
838 "enabling inode map caching");
839 break;
840 case Opt_noinode_cache:
841 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
842 "disabling inode map caching");
843 break;
844 case Opt_clear_cache:
845 btrfs_set_and_info(info, CLEAR_CACHE,
846 "force clearing of disk cache");
847 break;
848 case Opt_user_subvol_rm_allowed:
849 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
850 break;
851 case Opt_enospc_debug:
852 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
853 break;
854 case Opt_noenospc_debug:
855 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
856 break;
857 case Opt_defrag:
858 btrfs_set_and_info(info, AUTO_DEFRAG,
859 "enabling auto defrag");
860 break;
861 case Opt_nodefrag:
862 btrfs_clear_and_info(info, AUTO_DEFRAG,
863 "disabling auto defrag");
864 break;
865 case Opt_recovery:
866 case Opt_usebackuproot:
867 btrfs_warn(info,
868 "'%s' is deprecated, use 'rescue=usebackuproot' instead",
869 token == Opt_recovery ? "recovery" :
870 "usebackuproot");
871 btrfs_info(info,
872 "trying to use backup root at mount time");
873 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
874 break;
875 case Opt_skip_balance:
876 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
877 break;
878#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
879 case Opt_check_integrity_including_extent_data:
880 btrfs_info(info,
881 "enabling check integrity including extent data");
882 btrfs_set_opt(info->mount_opt,
883 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
884 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
885 break;
886 case Opt_check_integrity:
887 btrfs_info(info, "enabling check integrity");
888 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
889 break;
890 case Opt_check_integrity_print_mask:
891 ret = match_int(&args[0], &intarg);
892 if (ret)
893 goto out;
894 info->check_integrity_print_mask = intarg;
895 btrfs_info(info, "check_integrity_print_mask 0x%x",
896 info->check_integrity_print_mask);
897 break;
898#else
899 case Opt_check_integrity_including_extent_data:
900 case Opt_check_integrity:
901 case Opt_check_integrity_print_mask:
902 btrfs_err(info,
903 "support for check_integrity* not compiled in!");
904 ret = -EINVAL;
905 goto out;
906#endif
907 case Opt_fatal_errors:
908 if (strcmp(args[0].from, "panic") == 0)
909 btrfs_set_opt(info->mount_opt,
910 PANIC_ON_FATAL_ERROR);
911 else if (strcmp(args[0].from, "bug") == 0)
912 btrfs_clear_opt(info->mount_opt,
913 PANIC_ON_FATAL_ERROR);
914 else {
915 ret = -EINVAL;
916 goto out;
917 }
918 break;
919 case Opt_commit_interval:
920 intarg = 0;
921 ret = match_int(&args[0], &intarg);
922 if (ret)
923 goto out;
924 if (intarg == 0) {
925 btrfs_info(info,
926 "using default commit interval %us",
927 BTRFS_DEFAULT_COMMIT_INTERVAL);
928 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
929 } else if (intarg > 300) {
930 btrfs_warn(info, "excessive commit interval %d",
931 intarg);
932 }
933 info->commit_interval = intarg;
934 break;
935 case Opt_rescue:
936 ret = parse_rescue_options(info, args[0].from);
937 if (ret < 0)
938 goto out;
939 break;
940#ifdef CONFIG_BTRFS_DEBUG
941 case Opt_fragment_all:
942 btrfs_info(info, "fragmenting all space");
943 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
944 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
945 break;
946 case Opt_fragment_metadata:
947 btrfs_info(info, "fragmenting metadata");
948 btrfs_set_opt(info->mount_opt,
949 FRAGMENT_METADATA);
950 break;
951 case Opt_fragment_data:
952 btrfs_info(info, "fragmenting data");
953 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
954 break;
955#endif
956#ifdef CONFIG_BTRFS_FS_REF_VERIFY
957 case Opt_ref_verify:
958 btrfs_info(info, "doing ref verification");
959 btrfs_set_opt(info->mount_opt, REF_VERIFY);
960 break;
961#endif
962 case Opt_err:
963 btrfs_err(info, "unrecognized mount option '%s'", p);
964 ret = -EINVAL;
965 goto out;
966 default:
967 break;
968 }
969 }
970check:
971 /*
972 * Extra check for current option against current flag
973 */
974 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
975 btrfs_err(info,
976 "nologreplay must be used with ro mount option");
977 ret = -EINVAL;
978 }
979out:
980 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
981 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
982 !btrfs_test_opt(info, CLEAR_CACHE)) {
983 btrfs_err(info, "cannot disable free space tree");
984 ret = -EINVAL;
985
986 }
987 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
988 btrfs_info(info, "disk space caching is enabled");
989 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
990 btrfs_info(info, "using free space tree");
991 return ret;
992}
993
994/*
995 * Parse mount options that are required early in the mount process.
996 *
997 * All other options will be parsed on much later in the mount process and
998 * only when we need to allocate a new super block.
999 */
1000static int btrfs_parse_device_options(const char *options, fmode_t flags,
1001 void *holder)
1002{
1003 substring_t args[MAX_OPT_ARGS];
1004 char *device_name, *opts, *orig, *p;
1005 struct btrfs_device *device = NULL;
1006 int error = 0;
1007
1008 lockdep_assert_held(&uuid_mutex);
1009
1010 if (!options)
1011 return 0;
1012
1013 /*
1014 * strsep changes the string, duplicate it because btrfs_parse_options
1015 * gets called later
1016 */
1017 opts = kstrdup(options, GFP_KERNEL);
1018 if (!opts)
1019 return -ENOMEM;
1020 orig = opts;
1021
1022 while ((p = strsep(&opts, ",")) != NULL) {
1023 int token;
1024
1025 if (!*p)
1026 continue;
1027
1028 token = match_token(p, tokens, args);
1029 if (token == Opt_device) {
1030 device_name = match_strdup(&args[0]);
1031 if (!device_name) {
1032 error = -ENOMEM;
1033 goto out;
1034 }
1035 device = btrfs_scan_one_device(device_name, flags,
1036 holder);
1037 kfree(device_name);
1038 if (IS_ERR(device)) {
1039 error = PTR_ERR(device);
1040 goto out;
1041 }
1042 }
1043 }
1044
1045out:
1046 kfree(orig);
1047 return error;
1048}
1049
1050/*
1051 * Parse mount options that are related to subvolume id
1052 *
1053 * The value is later passed to mount_subvol()
1054 */
1055static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1056 u64 *subvol_objectid)
1057{
1058 substring_t args[MAX_OPT_ARGS];
1059 char *opts, *orig, *p;
1060 int error = 0;
1061 u64 subvolid;
1062
1063 if (!options)
1064 return 0;
1065
1066 /*
1067 * strsep changes the string, duplicate it because
1068 * btrfs_parse_device_options gets called later
1069 */
1070 opts = kstrdup(options, GFP_KERNEL);
1071 if (!opts)
1072 return -ENOMEM;
1073 orig = opts;
1074
1075 while ((p = strsep(&opts, ",")) != NULL) {
1076 int token;
1077 if (!*p)
1078 continue;
1079
1080 token = match_token(p, tokens, args);
1081 switch (token) {
1082 case Opt_subvol:
1083 kfree(*subvol_name);
1084 *subvol_name = match_strdup(&args[0]);
1085 if (!*subvol_name) {
1086 error = -ENOMEM;
1087 goto out;
1088 }
1089 break;
1090 case Opt_subvolid:
1091 error = match_u64(&args[0], &subvolid);
1092 if (error)
1093 goto out;
1094
1095 /* we want the original fs_tree */
1096 if (subvolid == 0)
1097 subvolid = BTRFS_FS_TREE_OBJECTID;
1098
1099 *subvol_objectid = subvolid;
1100 break;
1101 default:
1102 break;
1103 }
1104 }
1105
1106out:
1107 kfree(orig);
1108 return error;
1109}
1110
1111char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1112 u64 subvol_objectid)
1113{
1114 struct btrfs_root *root = fs_info->tree_root;
1115 struct btrfs_root *fs_root = NULL;
1116 struct btrfs_root_ref *root_ref;
1117 struct btrfs_inode_ref *inode_ref;
1118 struct btrfs_key key;
1119 struct btrfs_path *path = NULL;
1120 char *name = NULL, *ptr;
1121 u64 dirid;
1122 int len;
1123 int ret;
1124
1125 path = btrfs_alloc_path();
1126 if (!path) {
1127 ret = -ENOMEM;
1128 goto err;
1129 }
1130 path->leave_spinning = 1;
1131
1132 name = kmalloc(PATH_MAX, GFP_KERNEL);
1133 if (!name) {
1134 ret = -ENOMEM;
1135 goto err;
1136 }
1137 ptr = name + PATH_MAX - 1;
1138 ptr[0] = '\0';
1139
1140 /*
1141 * Walk up the subvolume trees in the tree of tree roots by root
1142 * backrefs until we hit the top-level subvolume.
1143 */
1144 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1145 key.objectid = subvol_objectid;
1146 key.type = BTRFS_ROOT_BACKREF_KEY;
1147 key.offset = (u64)-1;
1148
1149 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1150 if (ret < 0) {
1151 goto err;
1152 } else if (ret > 0) {
1153 ret = btrfs_previous_item(root, path, subvol_objectid,
1154 BTRFS_ROOT_BACKREF_KEY);
1155 if (ret < 0) {
1156 goto err;
1157 } else if (ret > 0) {
1158 ret = -ENOENT;
1159 goto err;
1160 }
1161 }
1162
1163 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1164 subvol_objectid = key.offset;
1165
1166 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1167 struct btrfs_root_ref);
1168 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1169 ptr -= len + 1;
1170 if (ptr < name) {
1171 ret = -ENAMETOOLONG;
1172 goto err;
1173 }
1174 read_extent_buffer(path->nodes[0], ptr + 1,
1175 (unsigned long)(root_ref + 1), len);
1176 ptr[0] = '/';
1177 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1178 btrfs_release_path(path);
1179
1180 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1181 if (IS_ERR(fs_root)) {
1182 ret = PTR_ERR(fs_root);
1183 fs_root = NULL;
1184 goto err;
1185 }
1186
1187 /*
1188 * Walk up the filesystem tree by inode refs until we hit the
1189 * root directory.
1190 */
1191 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1192 key.objectid = dirid;
1193 key.type = BTRFS_INODE_REF_KEY;
1194 key.offset = (u64)-1;
1195
1196 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1197 if (ret < 0) {
1198 goto err;
1199 } else if (ret > 0) {
1200 ret = btrfs_previous_item(fs_root, path, dirid,
1201 BTRFS_INODE_REF_KEY);
1202 if (ret < 0) {
1203 goto err;
1204 } else if (ret > 0) {
1205 ret = -ENOENT;
1206 goto err;
1207 }
1208 }
1209
1210 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1211 dirid = key.offset;
1212
1213 inode_ref = btrfs_item_ptr(path->nodes[0],
1214 path->slots[0],
1215 struct btrfs_inode_ref);
1216 len = btrfs_inode_ref_name_len(path->nodes[0],
1217 inode_ref);
1218 ptr -= len + 1;
1219 if (ptr < name) {
1220 ret = -ENAMETOOLONG;
1221 goto err;
1222 }
1223 read_extent_buffer(path->nodes[0], ptr + 1,
1224 (unsigned long)(inode_ref + 1), len);
1225 ptr[0] = '/';
1226 btrfs_release_path(path);
1227 }
1228 btrfs_put_root(fs_root);
1229 fs_root = NULL;
1230 }
1231
1232 btrfs_free_path(path);
1233 if (ptr == name + PATH_MAX - 1) {
1234 name[0] = '/';
1235 name[1] = '\0';
1236 } else {
1237 memmove(name, ptr, name + PATH_MAX - ptr);
1238 }
1239 return name;
1240
1241err:
1242 btrfs_put_root(fs_root);
1243 btrfs_free_path(path);
1244 kfree(name);
1245 return ERR_PTR(ret);
1246}
1247
1248static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1249{
1250 struct btrfs_root *root = fs_info->tree_root;
1251 struct btrfs_dir_item *di;
1252 struct btrfs_path *path;
1253 struct btrfs_key location;
1254 u64 dir_id;
1255
1256 path = btrfs_alloc_path();
1257 if (!path)
1258 return -ENOMEM;
1259 path->leave_spinning = 1;
1260
1261 /*
1262 * Find the "default" dir item which points to the root item that we
1263 * will mount by default if we haven't been given a specific subvolume
1264 * to mount.
1265 */
1266 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1267 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1268 if (IS_ERR(di)) {
1269 btrfs_free_path(path);
1270 return PTR_ERR(di);
1271 }
1272 if (!di) {
1273 /*
1274 * Ok the default dir item isn't there. This is weird since
1275 * it's always been there, but don't freak out, just try and
1276 * mount the top-level subvolume.
1277 */
1278 btrfs_free_path(path);
1279 *objectid = BTRFS_FS_TREE_OBJECTID;
1280 return 0;
1281 }
1282
1283 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1284 btrfs_free_path(path);
1285 *objectid = location.objectid;
1286 return 0;
1287}
1288
1289static int btrfs_fill_super(struct super_block *sb,
1290 struct btrfs_fs_devices *fs_devices,
1291 void *data)
1292{
1293 struct inode *inode;
1294 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1295 int err;
1296
1297 sb->s_maxbytes = MAX_LFS_FILESIZE;
1298 sb->s_magic = BTRFS_SUPER_MAGIC;
1299 sb->s_op = &btrfs_super_ops;
1300 sb->s_d_op = &btrfs_dentry_operations;
1301 sb->s_export_op = &btrfs_export_ops;
1302 sb->s_xattr = btrfs_xattr_handlers;
1303 sb->s_time_gran = 1;
1304#ifdef CONFIG_BTRFS_FS_POSIX_ACL
1305 sb->s_flags |= SB_POSIXACL;
1306#endif
1307 sb->s_flags |= SB_I_VERSION;
1308 sb->s_iflags |= SB_I_CGROUPWB;
1309
1310 err = super_setup_bdi(sb);
1311 if (err) {
1312 btrfs_err(fs_info, "super_setup_bdi failed");
1313 return err;
1314 }
1315
1316 err = open_ctree(sb, fs_devices, (char *)data);
1317 if (err) {
1318 btrfs_err(fs_info, "open_ctree failed");
1319 return err;
1320 }
1321
1322 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1323 if (IS_ERR(inode)) {
1324 err = PTR_ERR(inode);
1325 goto fail_close;
1326 }
1327
1328 sb->s_root = d_make_root(inode);
1329 if (!sb->s_root) {
1330 err = -ENOMEM;
1331 goto fail_close;
1332 }
1333
1334 cleancache_init_fs(sb);
1335 sb->s_flags |= SB_ACTIVE;
1336 return 0;
1337
1338fail_close:
1339 close_ctree(fs_info);
1340 return err;
1341}
1342
1343int btrfs_sync_fs(struct super_block *sb, int wait)
1344{
1345 struct btrfs_trans_handle *trans;
1346 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1347 struct btrfs_root *root = fs_info->tree_root;
1348
1349 trace_btrfs_sync_fs(fs_info, wait);
1350
1351 if (!wait) {
1352 filemap_flush(fs_info->btree_inode->i_mapping);
1353 return 0;
1354 }
1355
1356 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1357
1358 trans = btrfs_attach_transaction_barrier(root);
1359 if (IS_ERR(trans)) {
1360 /* no transaction, don't bother */
1361 if (PTR_ERR(trans) == -ENOENT) {
1362 /*
1363 * Exit unless we have some pending changes
1364 * that need to go through commit
1365 */
1366 if (fs_info->pending_changes == 0)
1367 return 0;
1368 /*
1369 * A non-blocking test if the fs is frozen. We must not
1370 * start a new transaction here otherwise a deadlock
1371 * happens. The pending operations are delayed to the
1372 * next commit after thawing.
1373 */
1374 if (sb_start_write_trylock(sb))
1375 sb_end_write(sb);
1376 else
1377 return 0;
1378 trans = btrfs_start_transaction(root, 0);
1379 }
1380 if (IS_ERR(trans))
1381 return PTR_ERR(trans);
1382 }
1383 return btrfs_commit_transaction(trans);
1384}
1385
1386static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1387{
1388 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1389 const char *compress_type;
1390 const char *subvol_name;
1391
1392 if (btrfs_test_opt(info, DEGRADED))
1393 seq_puts(seq, ",degraded");
1394 if (btrfs_test_opt(info, NODATASUM))
1395 seq_puts(seq, ",nodatasum");
1396 if (btrfs_test_opt(info, NODATACOW))
1397 seq_puts(seq, ",nodatacow");
1398 if (btrfs_test_opt(info, NOBARRIER))
1399 seq_puts(seq, ",nobarrier");
1400 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1401 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1402 if (info->thread_pool_size != min_t(unsigned long,
1403 num_online_cpus() + 2, 8))
1404 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1405 if (btrfs_test_opt(info, COMPRESS)) {
1406 compress_type = btrfs_compress_type2str(info->compress_type);
1407 if (btrfs_test_opt(info, FORCE_COMPRESS))
1408 seq_printf(seq, ",compress-force=%s", compress_type);
1409 else
1410 seq_printf(seq, ",compress=%s", compress_type);
1411 if (info->compress_level)
1412 seq_printf(seq, ":%d", info->compress_level);
1413 }
1414 if (btrfs_test_opt(info, NOSSD))
1415 seq_puts(seq, ",nossd");
1416 if (btrfs_test_opt(info, SSD_SPREAD))
1417 seq_puts(seq, ",ssd_spread");
1418 else if (btrfs_test_opt(info, SSD))
1419 seq_puts(seq, ",ssd");
1420 if (btrfs_test_opt(info, NOTREELOG))
1421 seq_puts(seq, ",notreelog");
1422 if (btrfs_test_opt(info, NOLOGREPLAY))
1423 seq_puts(seq, ",rescue=nologreplay");
1424 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1425 seq_puts(seq, ",flushoncommit");
1426 if (btrfs_test_opt(info, DISCARD_SYNC))
1427 seq_puts(seq, ",discard");
1428 if (btrfs_test_opt(info, DISCARD_ASYNC))
1429 seq_puts(seq, ",discard=async");
1430 if (!(info->sb->s_flags & SB_POSIXACL))
1431 seq_puts(seq, ",noacl");
1432 if (btrfs_test_opt(info, SPACE_CACHE))
1433 seq_puts(seq, ",space_cache");
1434 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1435 seq_puts(seq, ",space_cache=v2");
1436 else
1437 seq_puts(seq, ",nospace_cache");
1438 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1439 seq_puts(seq, ",rescan_uuid_tree");
1440 if (btrfs_test_opt(info, CLEAR_CACHE))
1441 seq_puts(seq, ",clear_cache");
1442 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1443 seq_puts(seq, ",user_subvol_rm_allowed");
1444 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1445 seq_puts(seq, ",enospc_debug");
1446 if (btrfs_test_opt(info, AUTO_DEFRAG))
1447 seq_puts(seq, ",autodefrag");
1448 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1449 seq_puts(seq, ",inode_cache");
1450 if (btrfs_test_opt(info, SKIP_BALANCE))
1451 seq_puts(seq, ",skip_balance");
1452#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1453 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1454 seq_puts(seq, ",check_int_data");
1455 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1456 seq_puts(seq, ",check_int");
1457 if (info->check_integrity_print_mask)
1458 seq_printf(seq, ",check_int_print_mask=%d",
1459 info->check_integrity_print_mask);
1460#endif
1461 if (info->metadata_ratio)
1462 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1463 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1464 seq_puts(seq, ",fatal_errors=panic");
1465 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1466 seq_printf(seq, ",commit=%u", info->commit_interval);
1467#ifdef CONFIG_BTRFS_DEBUG
1468 if (btrfs_test_opt(info, FRAGMENT_DATA))
1469 seq_puts(seq, ",fragment=data");
1470 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1471 seq_puts(seq, ",fragment=metadata");
1472#endif
1473 if (btrfs_test_opt(info, REF_VERIFY))
1474 seq_puts(seq, ",ref_verify");
1475 seq_printf(seq, ",subvolid=%llu",
1476 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1477 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1478 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1479 if (!IS_ERR(subvol_name)) {
1480 seq_puts(seq, ",subvol=");
1481 seq_escape(seq, subvol_name, " \t\n\\");
1482 kfree(subvol_name);
1483 }
1484 return 0;
1485}
1486
1487static int btrfs_test_super(struct super_block *s, void *data)
1488{
1489 struct btrfs_fs_info *p = data;
1490 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1491
1492 return fs_info->fs_devices == p->fs_devices;
1493}
1494
1495static int btrfs_set_super(struct super_block *s, void *data)
1496{
1497 int err = set_anon_super(s, data);
1498 if (!err)
1499 s->s_fs_info = data;
1500 return err;
1501}
1502
1503/*
1504 * subvolumes are identified by ino 256
1505 */
1506static inline int is_subvolume_inode(struct inode *inode)
1507{
1508 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1509 return 1;
1510 return 0;
1511}
1512
1513static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1514 struct vfsmount *mnt)
1515{
1516 struct dentry *root;
1517 int ret;
1518
1519 if (!subvol_name) {
1520 if (!subvol_objectid) {
1521 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1522 &subvol_objectid);
1523 if (ret) {
1524 root = ERR_PTR(ret);
1525 goto out;
1526 }
1527 }
1528 subvol_name = btrfs_get_subvol_name_from_objectid(
1529 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1530 if (IS_ERR(subvol_name)) {
1531 root = ERR_CAST(subvol_name);
1532 subvol_name = NULL;
1533 goto out;
1534 }
1535
1536 }
1537
1538 root = mount_subtree(mnt, subvol_name);
1539 /* mount_subtree() drops our reference on the vfsmount. */
1540 mnt = NULL;
1541
1542 if (!IS_ERR(root)) {
1543 struct super_block *s = root->d_sb;
1544 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1545 struct inode *root_inode = d_inode(root);
1546 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1547
1548 ret = 0;
1549 if (!is_subvolume_inode(root_inode)) {
1550 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1551 subvol_name);
1552 ret = -EINVAL;
1553 }
1554 if (subvol_objectid && root_objectid != subvol_objectid) {
1555 /*
1556 * This will also catch a race condition where a
1557 * subvolume which was passed by ID is renamed and
1558 * another subvolume is renamed over the old location.
1559 */
1560 btrfs_err(fs_info,
1561 "subvol '%s' does not match subvolid %llu",
1562 subvol_name, subvol_objectid);
1563 ret = -EINVAL;
1564 }
1565 if (ret) {
1566 dput(root);
1567 root = ERR_PTR(ret);
1568 deactivate_locked_super(s);
1569 }
1570 }
1571
1572out:
1573 mntput(mnt);
1574 kfree(subvol_name);
1575 return root;
1576}
1577
1578/*
1579 * Find a superblock for the given device / mount point.
1580 *
1581 * Note: This is based on mount_bdev from fs/super.c with a few additions
1582 * for multiple device setup. Make sure to keep it in sync.
1583 */
1584static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1585 int flags, const char *device_name, void *data)
1586{
1587 struct block_device *bdev = NULL;
1588 struct super_block *s;
1589 struct btrfs_device *device = NULL;
1590 struct btrfs_fs_devices *fs_devices = NULL;
1591 struct btrfs_fs_info *fs_info = NULL;
1592 void *new_sec_opts = NULL;
1593 fmode_t mode = FMODE_READ;
1594 int error = 0;
1595
1596 if (!(flags & SB_RDONLY))
1597 mode |= FMODE_WRITE;
1598
1599 if (data) {
1600 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1601 if (error)
1602 return ERR_PTR(error);
1603 }
1604
1605 /*
1606 * Setup a dummy root and fs_info for test/set super. This is because
1607 * we don't actually fill this stuff out until open_ctree, but we need
1608 * then open_ctree will properly initialize the file system specific
1609 * settings later. btrfs_init_fs_info initializes the static elements
1610 * of the fs_info (locks and such) to make cleanup easier if we find a
1611 * superblock with our given fs_devices later on at sget() time.
1612 */
1613 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1614 if (!fs_info) {
1615 error = -ENOMEM;
1616 goto error_sec_opts;
1617 }
1618 btrfs_init_fs_info(fs_info);
1619
1620 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1621 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1622 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1623 error = -ENOMEM;
1624 goto error_fs_info;
1625 }
1626
1627 mutex_lock(&uuid_mutex);
1628 error = btrfs_parse_device_options(data, mode, fs_type);
1629 if (error) {
1630 mutex_unlock(&uuid_mutex);
1631 goto error_fs_info;
1632 }
1633
1634 device = btrfs_scan_one_device(device_name, mode, fs_type);
1635 if (IS_ERR(device)) {
1636 mutex_unlock(&uuid_mutex);
1637 error = PTR_ERR(device);
1638 goto error_fs_info;
1639 }
1640
1641 fs_devices = device->fs_devices;
1642 fs_info->fs_devices = fs_devices;
1643
1644 error = btrfs_open_devices(fs_devices, mode, fs_type);
1645 mutex_unlock(&uuid_mutex);
1646 if (error)
1647 goto error_fs_info;
1648
1649 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1650 error = -EACCES;
1651 goto error_close_devices;
1652 }
1653
1654 bdev = fs_devices->latest_bdev;
1655 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1656 fs_info);
1657 if (IS_ERR(s)) {
1658 error = PTR_ERR(s);
1659 goto error_close_devices;
1660 }
1661
1662 if (s->s_root) {
1663 btrfs_close_devices(fs_devices);
1664 btrfs_free_fs_info(fs_info);
1665 if ((flags ^ s->s_flags) & SB_RDONLY)
1666 error = -EBUSY;
1667 } else {
1668 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1669 btrfs_sb(s)->bdev_holder = fs_type;
1670 if (!strstr(crc32c_impl(), "generic"))
1671 set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1672 error = btrfs_fill_super(s, fs_devices, data);
1673 }
1674 if (!error)
1675 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1676 security_free_mnt_opts(&new_sec_opts);
1677 if (error) {
1678 deactivate_locked_super(s);
1679 return ERR_PTR(error);
1680 }
1681
1682 return dget(s->s_root);
1683
1684error_close_devices:
1685 btrfs_close_devices(fs_devices);
1686error_fs_info:
1687 btrfs_free_fs_info(fs_info);
1688error_sec_opts:
1689 security_free_mnt_opts(&new_sec_opts);
1690 return ERR_PTR(error);
1691}
1692
1693/*
1694 * Mount function which is called by VFS layer.
1695 *
1696 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1697 * which needs vfsmount* of device's root (/). This means device's root has to
1698 * be mounted internally in any case.
1699 *
1700 * Operation flow:
1701 * 1. Parse subvol id related options for later use in mount_subvol().
1702 *
1703 * 2. Mount device's root (/) by calling vfs_kern_mount().
1704 *
1705 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1706 * first place. In order to avoid calling btrfs_mount() again, we use
1707 * different file_system_type which is not registered to VFS by
1708 * register_filesystem() (btrfs_root_fs_type). As a result,
1709 * btrfs_mount_root() is called. The return value will be used by
1710 * mount_subtree() in mount_subvol().
1711 *
1712 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1713 * "btrfs subvolume set-default", mount_subvol() is called always.
1714 */
1715static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1716 const char *device_name, void *data)
1717{
1718 struct vfsmount *mnt_root;
1719 struct dentry *root;
1720 char *subvol_name = NULL;
1721 u64 subvol_objectid = 0;
1722 int error = 0;
1723
1724 error = btrfs_parse_subvol_options(data, &subvol_name,
1725 &subvol_objectid);
1726 if (error) {
1727 kfree(subvol_name);
1728 return ERR_PTR(error);
1729 }
1730
1731 /* mount device's root (/) */
1732 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1733 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1734 if (flags & SB_RDONLY) {
1735 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1736 flags & ~SB_RDONLY, device_name, data);
1737 } else {
1738 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1739 flags | SB_RDONLY, device_name, data);
1740 if (IS_ERR(mnt_root)) {
1741 root = ERR_CAST(mnt_root);
1742 kfree(subvol_name);
1743 goto out;
1744 }
1745
1746 down_write(&mnt_root->mnt_sb->s_umount);
1747 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1748 up_write(&mnt_root->mnt_sb->s_umount);
1749 if (error < 0) {
1750 root = ERR_PTR(error);
1751 mntput(mnt_root);
1752 kfree(subvol_name);
1753 goto out;
1754 }
1755 }
1756 }
1757 if (IS_ERR(mnt_root)) {
1758 root = ERR_CAST(mnt_root);
1759 kfree(subvol_name);
1760 goto out;
1761 }
1762
1763 /* mount_subvol() will free subvol_name and mnt_root */
1764 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1765
1766out:
1767 return root;
1768}
1769
1770static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1771 u32 new_pool_size, u32 old_pool_size)
1772{
1773 if (new_pool_size == old_pool_size)
1774 return;
1775
1776 fs_info->thread_pool_size = new_pool_size;
1777
1778 btrfs_info(fs_info, "resize thread pool %d -> %d",
1779 old_pool_size, new_pool_size);
1780
1781 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1782 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1783 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1784 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1785 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1786 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1787 new_pool_size);
1788 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1789 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1790 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1791 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1792 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1793 new_pool_size);
1794}
1795
1796static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1797 unsigned long old_opts, int flags)
1798{
1799 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1800 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1801 (flags & SB_RDONLY))) {
1802 /* wait for any defraggers to finish */
1803 wait_event(fs_info->transaction_wait,
1804 (atomic_read(&fs_info->defrag_running) == 0));
1805 if (flags & SB_RDONLY)
1806 sync_filesystem(fs_info->sb);
1807 }
1808}
1809
1810static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1811 unsigned long old_opts)
1812{
1813 /*
1814 * We need to cleanup all defragable inodes if the autodefragment is
1815 * close or the filesystem is read only.
1816 */
1817 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1818 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1819 btrfs_cleanup_defrag_inodes(fs_info);
1820 }
1821
1822 /* If we toggled discard async */
1823 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1824 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1825 btrfs_discard_resume(fs_info);
1826 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1827 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1828 btrfs_discard_cleanup(fs_info);
1829}
1830
1831static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1832{
1833 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1834 struct btrfs_root *root = fs_info->tree_root;
1835 unsigned old_flags = sb->s_flags;
1836 unsigned long old_opts = fs_info->mount_opt;
1837 unsigned long old_compress_type = fs_info->compress_type;
1838 u64 old_max_inline = fs_info->max_inline;
1839 u32 old_thread_pool_size = fs_info->thread_pool_size;
1840 u32 old_metadata_ratio = fs_info->metadata_ratio;
1841 int ret;
1842
1843 sync_filesystem(sb);
1844 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1845
1846 if (data) {
1847 void *new_sec_opts = NULL;
1848
1849 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1850 if (!ret)
1851 ret = security_sb_remount(sb, new_sec_opts);
1852 security_free_mnt_opts(&new_sec_opts);
1853 if (ret)
1854 goto restore;
1855 }
1856
1857 ret = btrfs_parse_options(fs_info, data, *flags);
1858 if (ret)
1859 goto restore;
1860
1861 btrfs_remount_begin(fs_info, old_opts, *flags);
1862 btrfs_resize_thread_pool(fs_info,
1863 fs_info->thread_pool_size, old_thread_pool_size);
1864
1865 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1866 goto out;
1867
1868 if (*flags & SB_RDONLY) {
1869 /*
1870 * this also happens on 'umount -rf' or on shutdown, when
1871 * the filesystem is busy.
1872 */
1873 cancel_work_sync(&fs_info->async_reclaim_work);
1874
1875 btrfs_discard_cleanup(fs_info);
1876
1877 /* wait for the uuid_scan task to finish */
1878 down(&fs_info->uuid_tree_rescan_sem);
1879 /* avoid complains from lockdep et al. */
1880 up(&fs_info->uuid_tree_rescan_sem);
1881
1882 sb->s_flags |= SB_RDONLY;
1883
1884 /*
1885 * Setting SB_RDONLY will put the cleaner thread to
1886 * sleep at the next loop if it's already active.
1887 * If it's already asleep, we'll leave unused block
1888 * groups on disk until we're mounted read-write again
1889 * unless we clean them up here.
1890 */
1891 btrfs_delete_unused_bgs(fs_info);
1892
1893 btrfs_dev_replace_suspend_for_unmount(fs_info);
1894 btrfs_scrub_cancel(fs_info);
1895 btrfs_pause_balance(fs_info);
1896
1897 ret = btrfs_commit_super(fs_info);
1898 if (ret)
1899 goto restore;
1900 } else {
1901 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1902 btrfs_err(fs_info,
1903 "Remounting read-write after error is not allowed");
1904 ret = -EINVAL;
1905 goto restore;
1906 }
1907 if (fs_info->fs_devices->rw_devices == 0) {
1908 ret = -EACCES;
1909 goto restore;
1910 }
1911
1912 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1913 btrfs_warn(fs_info,
1914 "too many missing devices, writable remount is not allowed");
1915 ret = -EACCES;
1916 goto restore;
1917 }
1918
1919 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1920 btrfs_warn(fs_info,
1921 "mount required to replay tree-log, cannot remount read-write");
1922 ret = -EINVAL;
1923 goto restore;
1924 }
1925
1926 ret = btrfs_cleanup_fs_roots(fs_info);
1927 if (ret)
1928 goto restore;
1929
1930 /* recover relocation */
1931 mutex_lock(&fs_info->cleaner_mutex);
1932 ret = btrfs_recover_relocation(root);
1933 mutex_unlock(&fs_info->cleaner_mutex);
1934 if (ret)
1935 goto restore;
1936
1937 ret = btrfs_resume_balance_async(fs_info);
1938 if (ret)
1939 goto restore;
1940
1941 ret = btrfs_resume_dev_replace_async(fs_info);
1942 if (ret) {
1943 btrfs_warn(fs_info, "failed to resume dev_replace");
1944 goto restore;
1945 }
1946
1947 btrfs_qgroup_rescan_resume(fs_info);
1948
1949 if (!fs_info->uuid_root) {
1950 btrfs_info(fs_info, "creating UUID tree");
1951 ret = btrfs_create_uuid_tree(fs_info);
1952 if (ret) {
1953 btrfs_warn(fs_info,
1954 "failed to create the UUID tree %d",
1955 ret);
1956 goto restore;
1957 }
1958 }
1959 sb->s_flags &= ~SB_RDONLY;
1960
1961 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1962 }
1963out:
1964 /*
1965 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
1966 * since the absence of the flag means it can be toggled off by remount.
1967 */
1968 *flags |= SB_I_VERSION;
1969
1970 wake_up_process(fs_info->transaction_kthread);
1971 btrfs_remount_cleanup(fs_info, old_opts);
1972 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1973
1974 return 0;
1975
1976restore:
1977 /* We've hit an error - don't reset SB_RDONLY */
1978 if (sb_rdonly(sb))
1979 old_flags |= SB_RDONLY;
1980 sb->s_flags = old_flags;
1981 fs_info->mount_opt = old_opts;
1982 fs_info->compress_type = old_compress_type;
1983 fs_info->max_inline = old_max_inline;
1984 btrfs_resize_thread_pool(fs_info,
1985 old_thread_pool_size, fs_info->thread_pool_size);
1986 fs_info->metadata_ratio = old_metadata_ratio;
1987 btrfs_remount_cleanup(fs_info, old_opts);
1988 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1989
1990 return ret;
1991}
1992
1993/* Used to sort the devices by max_avail(descending sort) */
1994static inline int btrfs_cmp_device_free_bytes(const void *dev_info1,
1995 const void *dev_info2)
1996{
1997 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1998 ((struct btrfs_device_info *)dev_info2)->max_avail)
1999 return -1;
2000 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
2001 ((struct btrfs_device_info *)dev_info2)->max_avail)
2002 return 1;
2003 else
2004 return 0;
2005}
2006
2007/*
2008 * sort the devices by max_avail, in which max free extent size of each device
2009 * is stored.(Descending Sort)
2010 */
2011static inline void btrfs_descending_sort_devices(
2012 struct btrfs_device_info *devices,
2013 size_t nr_devices)
2014{
2015 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2016 btrfs_cmp_device_free_bytes, NULL);
2017}
2018
2019/*
2020 * The helper to calc the free space on the devices that can be used to store
2021 * file data.
2022 */
2023static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2024 u64 *free_bytes)
2025{
2026 struct btrfs_device_info *devices_info;
2027 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2028 struct btrfs_device *device;
2029 u64 type;
2030 u64 avail_space;
2031 u64 min_stripe_size;
2032 int num_stripes = 1;
2033 int i = 0, nr_devices;
2034 const struct btrfs_raid_attr *rattr;
2035
2036 /*
2037 * We aren't under the device list lock, so this is racy-ish, but good
2038 * enough for our purposes.
2039 */
2040 nr_devices = fs_info->fs_devices->open_devices;
2041 if (!nr_devices) {
2042 smp_mb();
2043 nr_devices = fs_info->fs_devices->open_devices;
2044 ASSERT(nr_devices);
2045 if (!nr_devices) {
2046 *free_bytes = 0;
2047 return 0;
2048 }
2049 }
2050
2051 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2052 GFP_KERNEL);
2053 if (!devices_info)
2054 return -ENOMEM;
2055
2056 /* calc min stripe number for data space allocation */
2057 type = btrfs_data_alloc_profile(fs_info);
2058 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2059
2060 if (type & BTRFS_BLOCK_GROUP_RAID0)
2061 num_stripes = nr_devices;
2062 else if (type & BTRFS_BLOCK_GROUP_RAID1)
2063 num_stripes = 2;
2064 else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
2065 num_stripes = 3;
2066 else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
2067 num_stripes = 4;
2068 else if (type & BTRFS_BLOCK_GROUP_RAID10)
2069 num_stripes = 4;
2070
2071 /* Adjust for more than 1 stripe per device */
2072 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2073
2074 rcu_read_lock();
2075 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2076 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2077 &device->dev_state) ||
2078 !device->bdev ||
2079 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2080 continue;
2081
2082 if (i >= nr_devices)
2083 break;
2084
2085 avail_space = device->total_bytes - device->bytes_used;
2086
2087 /* align with stripe_len */
2088 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2089
2090 /*
2091 * In order to avoid overwriting the superblock on the drive,
2092 * btrfs starts at an offset of at least 1MB when doing chunk
2093 * allocation.
2094 *
2095 * This ensures we have at least min_stripe_size free space
2096 * after excluding 1MB.
2097 */
2098 if (avail_space <= SZ_1M + min_stripe_size)
2099 continue;
2100
2101 avail_space -= SZ_1M;
2102
2103 devices_info[i].dev = device;
2104 devices_info[i].max_avail = avail_space;
2105
2106 i++;
2107 }
2108 rcu_read_unlock();
2109
2110 nr_devices = i;
2111
2112 btrfs_descending_sort_devices(devices_info, nr_devices);
2113
2114 i = nr_devices - 1;
2115 avail_space = 0;
2116 while (nr_devices >= rattr->devs_min) {
2117 num_stripes = min(num_stripes, nr_devices);
2118
2119 if (devices_info[i].max_avail >= min_stripe_size) {
2120 int j;
2121 u64 alloc_size;
2122
2123 avail_space += devices_info[i].max_avail * num_stripes;
2124 alloc_size = devices_info[i].max_avail;
2125 for (j = i + 1 - num_stripes; j <= i; j++)
2126 devices_info[j].max_avail -= alloc_size;
2127 }
2128 i--;
2129 nr_devices--;
2130 }
2131
2132 kfree(devices_info);
2133 *free_bytes = avail_space;
2134 return 0;
2135}
2136
2137/*
2138 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2139 *
2140 * If there's a redundant raid level at DATA block groups, use the respective
2141 * multiplier to scale the sizes.
2142 *
2143 * Unused device space usage is based on simulating the chunk allocator
2144 * algorithm that respects the device sizes and order of allocations. This is
2145 * a close approximation of the actual use but there are other factors that may
2146 * change the result (like a new metadata chunk).
2147 *
2148 * If metadata is exhausted, f_bavail will be 0.
2149 */
2150static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2151{
2152 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2153 struct btrfs_super_block *disk_super = fs_info->super_copy;
2154 struct btrfs_space_info *found;
2155 u64 total_used = 0;
2156 u64 total_free_data = 0;
2157 u64 total_free_meta = 0;
2158 int bits = dentry->d_sb->s_blocksize_bits;
2159 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2160 unsigned factor = 1;
2161 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2162 int ret;
2163 u64 thresh = 0;
2164 int mixed = 0;
2165
2166 rcu_read_lock();
2167 list_for_each_entry_rcu(found, &fs_info->space_info, list) {
2168 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2169 int i;
2170
2171 total_free_data += found->disk_total - found->disk_used;
2172 total_free_data -=
2173 btrfs_account_ro_block_groups_free_space(found);
2174
2175 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2176 if (!list_empty(&found->block_groups[i]))
2177 factor = btrfs_bg_type_to_factor(
2178 btrfs_raid_array[i].bg_flag);
2179 }
2180 }
2181
2182 /*
2183 * Metadata in mixed block goup profiles are accounted in data
2184 */
2185 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2186 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2187 mixed = 1;
2188 else
2189 total_free_meta += found->disk_total -
2190 found->disk_used;
2191 }
2192
2193 total_used += found->disk_used;
2194 }
2195
2196 rcu_read_unlock();
2197
2198 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2199 buf->f_blocks >>= bits;
2200 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2201
2202 /* Account global block reserve as used, it's in logical size already */
2203 spin_lock(&block_rsv->lock);
2204 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2205 if (buf->f_bfree >= block_rsv->size >> bits)
2206 buf->f_bfree -= block_rsv->size >> bits;
2207 else
2208 buf->f_bfree = 0;
2209 spin_unlock(&block_rsv->lock);
2210
2211 buf->f_bavail = div_u64(total_free_data, factor);
2212 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2213 if (ret)
2214 return ret;
2215 buf->f_bavail += div_u64(total_free_data, factor);
2216 buf->f_bavail = buf->f_bavail >> bits;
2217
2218 /*
2219 * We calculate the remaining metadata space minus global reserve. If
2220 * this is (supposedly) smaller than zero, there's no space. But this
2221 * does not hold in practice, the exhausted state happens where's still
2222 * some positive delta. So we apply some guesswork and compare the
2223 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2224 *
2225 * We probably cannot calculate the exact threshold value because this
2226 * depends on the internal reservations requested by various
2227 * operations, so some operations that consume a few metadata will
2228 * succeed even if the Avail is zero. But this is better than the other
2229 * way around.
2230 */
2231 thresh = SZ_4M;
2232
2233 /*
2234 * We only want to claim there's no available space if we can no longer
2235 * allocate chunks for our metadata profile and our global reserve will
2236 * not fit in the free metadata space. If we aren't ->full then we
2237 * still can allocate chunks and thus are fine using the currently
2238 * calculated f_bavail.
2239 */
2240 if (!mixed && block_rsv->space_info->full &&
2241 total_free_meta - thresh < block_rsv->size)
2242 buf->f_bavail = 0;
2243
2244 buf->f_type = BTRFS_SUPER_MAGIC;
2245 buf->f_bsize = dentry->d_sb->s_blocksize;
2246 buf->f_namelen = BTRFS_NAME_LEN;
2247
2248 /* We treat it as constant endianness (it doesn't matter _which_)
2249 because we want the fsid to come out the same whether mounted
2250 on a big-endian or little-endian host */
2251 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2252 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2253 /* Mask in the root object ID too, to disambiguate subvols */
2254 buf->f_fsid.val[0] ^=
2255 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2256 buf->f_fsid.val[1] ^=
2257 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2258
2259 return 0;
2260}
2261
2262static void btrfs_kill_super(struct super_block *sb)
2263{
2264 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2265 kill_anon_super(sb);
2266 btrfs_free_fs_info(fs_info);
2267}
2268
2269static struct file_system_type btrfs_fs_type = {
2270 .owner = THIS_MODULE,
2271 .name = "btrfs",
2272 .mount = btrfs_mount,
2273 .kill_sb = btrfs_kill_super,
2274 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2275};
2276
2277static struct file_system_type btrfs_root_fs_type = {
2278 .owner = THIS_MODULE,
2279 .name = "btrfs",
2280 .mount = btrfs_mount_root,
2281 .kill_sb = btrfs_kill_super,
2282 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2283};
2284
2285MODULE_ALIAS_FS("btrfs");
2286
2287static int btrfs_control_open(struct inode *inode, struct file *file)
2288{
2289 /*
2290 * The control file's private_data is used to hold the
2291 * transaction when it is started and is used to keep
2292 * track of whether a transaction is already in progress.
2293 */
2294 file->private_data = NULL;
2295 return 0;
2296}
2297
2298/*
2299 * Used by /dev/btrfs-control for devices ioctls.
2300 */
2301static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2302 unsigned long arg)
2303{
2304 struct btrfs_ioctl_vol_args *vol;
2305 struct btrfs_device *device = NULL;
2306 int ret = -ENOTTY;
2307
2308 if (!capable(CAP_SYS_ADMIN))
2309 return -EPERM;
2310
2311 vol = memdup_user((void __user *)arg, sizeof(*vol));
2312 if (IS_ERR(vol))
2313 return PTR_ERR(vol);
2314 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2315
2316 switch (cmd) {
2317 case BTRFS_IOC_SCAN_DEV:
2318 mutex_lock(&uuid_mutex);
2319 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2320 &btrfs_root_fs_type);
2321 ret = PTR_ERR_OR_ZERO(device);
2322 mutex_unlock(&uuid_mutex);
2323 break;
2324 case BTRFS_IOC_FORGET_DEV:
2325 ret = btrfs_forget_devices(vol->name);
2326 break;
2327 case BTRFS_IOC_DEVICES_READY:
2328 mutex_lock(&uuid_mutex);
2329 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2330 &btrfs_root_fs_type);
2331 if (IS_ERR(device)) {
2332 mutex_unlock(&uuid_mutex);
2333 ret = PTR_ERR(device);
2334 break;
2335 }
2336 ret = !(device->fs_devices->num_devices ==
2337 device->fs_devices->total_devices);
2338 mutex_unlock(&uuid_mutex);
2339 break;
2340 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2341 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2342 break;
2343 }
2344
2345 kfree(vol);
2346 return ret;
2347}
2348
2349static int btrfs_freeze(struct super_block *sb)
2350{
2351 struct btrfs_trans_handle *trans;
2352 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2353 struct btrfs_root *root = fs_info->tree_root;
2354
2355 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2356 /*
2357 * We don't need a barrier here, we'll wait for any transaction that
2358 * could be in progress on other threads (and do delayed iputs that
2359 * we want to avoid on a frozen filesystem), or do the commit
2360 * ourselves.
2361 */
2362 trans = btrfs_attach_transaction_barrier(root);
2363 if (IS_ERR(trans)) {
2364 /* no transaction, don't bother */
2365 if (PTR_ERR(trans) == -ENOENT)
2366 return 0;
2367 return PTR_ERR(trans);
2368 }
2369 return btrfs_commit_transaction(trans);
2370}
2371
2372static int btrfs_unfreeze(struct super_block *sb)
2373{
2374 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2375
2376 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2377 return 0;
2378}
2379
2380static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2381{
2382 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2383 struct btrfs_device *dev, *first_dev = NULL;
2384
2385 /*
2386 * Lightweight locking of the devices. We should not need
2387 * device_list_mutex here as we only read the device data and the list
2388 * is protected by RCU. Even if a device is deleted during the list
2389 * traversals, we'll get valid data, the freeing callback will wait at
2390 * least until the rcu_read_unlock.
2391 */
2392 rcu_read_lock();
2393 list_for_each_entry_rcu(dev, &fs_info->fs_devices->devices, dev_list) {
2394 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2395 continue;
2396 if (!dev->name)
2397 continue;
2398 if (!first_dev || dev->devid < first_dev->devid)
2399 first_dev = dev;
2400 }
2401
2402 if (first_dev)
2403 seq_escape(m, rcu_str_deref(first_dev->name), " \t\n\\");
2404 else
2405 WARN_ON(1);
2406 rcu_read_unlock();
2407 return 0;
2408}
2409
2410static const struct super_operations btrfs_super_ops = {
2411 .drop_inode = btrfs_drop_inode,
2412 .evict_inode = btrfs_evict_inode,
2413 .put_super = btrfs_put_super,
2414 .sync_fs = btrfs_sync_fs,
2415 .show_options = btrfs_show_options,
2416 .show_devname = btrfs_show_devname,
2417 .alloc_inode = btrfs_alloc_inode,
2418 .destroy_inode = btrfs_destroy_inode,
2419 .free_inode = btrfs_free_inode,
2420 .statfs = btrfs_statfs,
2421 .remount_fs = btrfs_remount,
2422 .freeze_fs = btrfs_freeze,
2423 .unfreeze_fs = btrfs_unfreeze,
2424};
2425
2426static const struct file_operations btrfs_ctl_fops = {
2427 .open = btrfs_control_open,
2428 .unlocked_ioctl = btrfs_control_ioctl,
2429 .compat_ioctl = compat_ptr_ioctl,
2430 .owner = THIS_MODULE,
2431 .llseek = noop_llseek,
2432};
2433
2434static struct miscdevice btrfs_misc = {
2435 .minor = BTRFS_MINOR,
2436 .name = "btrfs-control",
2437 .fops = &btrfs_ctl_fops
2438};
2439
2440MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2441MODULE_ALIAS("devname:btrfs-control");
2442
2443static int __init btrfs_interface_init(void)
2444{
2445 return misc_register(&btrfs_misc);
2446}
2447
2448static __cold void btrfs_interface_exit(void)
2449{
2450 misc_deregister(&btrfs_misc);
2451}
2452
2453static void __init btrfs_print_mod_info(void)
2454{
2455 static const char options[] = ""
2456#ifdef CONFIG_BTRFS_DEBUG
2457 ", debug=on"
2458#endif
2459#ifdef CONFIG_BTRFS_ASSERT
2460 ", assert=on"
2461#endif
2462#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2463 ", integrity-checker=on"
2464#endif
2465#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2466 ", ref-verify=on"
2467#endif
2468 ;
2469 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2470}
2471
2472static int __init init_btrfs_fs(void)
2473{
2474 int err;
2475
2476 btrfs_props_init();
2477
2478 err = btrfs_init_sysfs();
2479 if (err)
2480 return err;
2481
2482 btrfs_init_compress();
2483
2484 err = btrfs_init_cachep();
2485 if (err)
2486 goto free_compress;
2487
2488 err = extent_io_init();
2489 if (err)
2490 goto free_cachep;
2491
2492 err = extent_state_cache_init();
2493 if (err)
2494 goto free_extent_io;
2495
2496 err = extent_map_init();
2497 if (err)
2498 goto free_extent_state_cache;
2499
2500 err = ordered_data_init();
2501 if (err)
2502 goto free_extent_map;
2503
2504 err = btrfs_delayed_inode_init();
2505 if (err)
2506 goto free_ordered_data;
2507
2508 err = btrfs_auto_defrag_init();
2509 if (err)
2510 goto free_delayed_inode;
2511
2512 err = btrfs_delayed_ref_init();
2513 if (err)
2514 goto free_auto_defrag;
2515
2516 err = btrfs_prelim_ref_init();
2517 if (err)
2518 goto free_delayed_ref;
2519
2520 err = btrfs_end_io_wq_init();
2521 if (err)
2522 goto free_prelim_ref;
2523
2524 err = btrfs_interface_init();
2525 if (err)
2526 goto free_end_io_wq;
2527
2528 btrfs_init_lockdep();
2529
2530 btrfs_print_mod_info();
2531
2532 err = btrfs_run_sanity_tests();
2533 if (err)
2534 goto unregister_ioctl;
2535
2536 err = register_filesystem(&btrfs_fs_type);
2537 if (err)
2538 goto unregister_ioctl;
2539
2540 return 0;
2541
2542unregister_ioctl:
2543 btrfs_interface_exit();
2544free_end_io_wq:
2545 btrfs_end_io_wq_exit();
2546free_prelim_ref:
2547 btrfs_prelim_ref_exit();
2548free_delayed_ref:
2549 btrfs_delayed_ref_exit();
2550free_auto_defrag:
2551 btrfs_auto_defrag_exit();
2552free_delayed_inode:
2553 btrfs_delayed_inode_exit();
2554free_ordered_data:
2555 ordered_data_exit();
2556free_extent_map:
2557 extent_map_exit();
2558free_extent_state_cache:
2559 extent_state_cache_exit();
2560free_extent_io:
2561 extent_io_exit();
2562free_cachep:
2563 btrfs_destroy_cachep();
2564free_compress:
2565 btrfs_exit_compress();
2566 btrfs_exit_sysfs();
2567
2568 return err;
2569}
2570
2571static void __exit exit_btrfs_fs(void)
2572{
2573 btrfs_destroy_cachep();
2574 btrfs_delayed_ref_exit();
2575 btrfs_auto_defrag_exit();
2576 btrfs_delayed_inode_exit();
2577 btrfs_prelim_ref_exit();
2578 ordered_data_exit();
2579 extent_map_exit();
2580 extent_state_cache_exit();
2581 extent_io_exit();
2582 btrfs_interface_exit();
2583 btrfs_end_io_wq_exit();
2584 unregister_filesystem(&btrfs_fs_type);
2585 btrfs_exit_sysfs();
2586 btrfs_cleanup_fs_uuids();
2587 btrfs_exit_compress();
2588}
2589
2590late_initcall(init_btrfs_fs);
2591module_exit(exit_btrfs_fs)
2592
2593MODULE_LICENSE("GPL");
2594MODULE_SOFTDEP("pre: crc32c");
2595MODULE_SOFTDEP("pre: xxhash64");
2596MODULE_SOFTDEP("pre: sha256");
2597MODULE_SOFTDEP("pre: blake2b-256");