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

source code of linux/fs/btrfs/super.c