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/ratelimit.h>
27#include <linux/crc32c.h>
28#include <linux/btrfs.h>
29#include <linux/security.h>
30#include <linux/fs_parser.h>
31#include "messages.h"
32#include "delayed-inode.h"
33#include "ctree.h"
34#include "disk-io.h"
35#include "transaction.h"
36#include "btrfs_inode.h"
37#include "direct-io.h"
38#include "props.h"
39#include "xattr.h"
40#include "bio.h"
41#include "export.h"
42#include "compression.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;
67static struct file_system_type btrfs_fs_type;
68
69static void btrfs_put_super(struct super_block *sb)
70{
71 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
72
73 btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid);
74 close_ctree(fs_info);
75}
76
77/* Store the mount options related information. */
78struct btrfs_fs_context {
79 char *subvol_name;
80 u64 subvol_objectid;
81 u64 max_inline;
82 u32 commit_interval;
83 u32 metadata_ratio;
84 u32 thread_pool_size;
85 unsigned long long mount_opt;
86 unsigned long compress_type:4;
87 unsigned int compress_level;
88 refcount_t refs;
89};
90
91enum {
92 Opt_acl,
93 Opt_clear_cache,
94 Opt_commit_interval,
95 Opt_compress,
96 Opt_compress_force,
97 Opt_compress_force_type,
98 Opt_compress_type,
99 Opt_degraded,
100 Opt_device,
101 Opt_fatal_errors,
102 Opt_flushoncommit,
103 Opt_max_inline,
104 Opt_barrier,
105 Opt_datacow,
106 Opt_datasum,
107 Opt_defrag,
108 Opt_discard,
109 Opt_discard_mode,
110 Opt_ratio,
111 Opt_rescan_uuid_tree,
112 Opt_skip_balance,
113 Opt_space_cache,
114 Opt_space_cache_version,
115 Opt_ssd,
116 Opt_ssd_spread,
117 Opt_subvol,
118 Opt_subvol_empty,
119 Opt_subvolid,
120 Opt_thread_pool,
121 Opt_treelog,
122 Opt_user_subvol_rm_allowed,
123 Opt_norecovery,
124
125 /* Rescue options */
126 Opt_rescue,
127 Opt_usebackuproot,
128 Opt_nologreplay,
129
130 /* Debugging options */
131 Opt_enospc_debug,
132#ifdef CONFIG_BTRFS_DEBUG
133 Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
134#endif
135#ifdef CONFIG_BTRFS_FS_REF_VERIFY
136 Opt_ref_verify,
137#endif
138 Opt_err,
139};
140
141enum {
142 Opt_fatal_errors_panic,
143 Opt_fatal_errors_bug,
144};
145
146static const struct constant_table btrfs_parameter_fatal_errors[] = {
147 { "panic", Opt_fatal_errors_panic },
148 { "bug", Opt_fatal_errors_bug },
149 {}
150};
151
152enum {
153 Opt_discard_sync,
154 Opt_discard_async,
155};
156
157static const struct constant_table btrfs_parameter_discard[] = {
158 { "sync", Opt_discard_sync },
159 { "async", Opt_discard_async },
160 {}
161};
162
163enum {
164 Opt_space_cache_v1,
165 Opt_space_cache_v2,
166};
167
168static const struct constant_table btrfs_parameter_space_cache[] = {
169 { "v1", Opt_space_cache_v1 },
170 { "v2", Opt_space_cache_v2 },
171 {}
172};
173
174enum {
175 Opt_rescue_usebackuproot,
176 Opt_rescue_nologreplay,
177 Opt_rescue_ignorebadroots,
178 Opt_rescue_ignoredatacsums,
179 Opt_rescue_ignoremetacsums,
180 Opt_rescue_ignoresuperflags,
181 Opt_rescue_parameter_all,
182};
183
184static const struct constant_table btrfs_parameter_rescue[] = {
185 { "usebackuproot", Opt_rescue_usebackuproot },
186 { "nologreplay", Opt_rescue_nologreplay },
187 { "ignorebadroots", Opt_rescue_ignorebadroots },
188 { "ibadroots", Opt_rescue_ignorebadroots },
189 { "ignoredatacsums", Opt_rescue_ignoredatacsums },
190 { "ignoremetacsums", Opt_rescue_ignoremetacsums},
191 { "ignoresuperflags", Opt_rescue_ignoresuperflags},
192 { "idatacsums", Opt_rescue_ignoredatacsums },
193 { "imetacsums", Opt_rescue_ignoremetacsums},
194 { "isuperflags", Opt_rescue_ignoresuperflags},
195 { "all", Opt_rescue_parameter_all },
196 {}
197};
198
199#ifdef CONFIG_BTRFS_DEBUG
200enum {
201 Opt_fragment_parameter_data,
202 Opt_fragment_parameter_metadata,
203 Opt_fragment_parameter_all,
204};
205
206static const struct constant_table btrfs_parameter_fragment[] = {
207 { "data", Opt_fragment_parameter_data },
208 { "metadata", Opt_fragment_parameter_metadata },
209 { "all", Opt_fragment_parameter_all },
210 {}
211};
212#endif
213
214static const struct fs_parameter_spec btrfs_fs_parameters[] = {
215 fsparam_flag_no("acl", Opt_acl),
216 fsparam_flag_no("autodefrag", Opt_defrag),
217 fsparam_flag_no("barrier", Opt_barrier),
218 fsparam_flag("clear_cache", Opt_clear_cache),
219 fsparam_u32("commit", Opt_commit_interval),
220 fsparam_flag("compress", Opt_compress),
221 fsparam_string("compress", Opt_compress_type),
222 fsparam_flag("compress-force", Opt_compress_force),
223 fsparam_string("compress-force", Opt_compress_force_type),
224 fsparam_flag_no("datacow", Opt_datacow),
225 fsparam_flag_no("datasum", Opt_datasum),
226 fsparam_flag("degraded", Opt_degraded),
227 fsparam_string("device", Opt_device),
228 fsparam_flag_no("discard", Opt_discard),
229 fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard),
230 fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors),
231 fsparam_flag_no("flushoncommit", Opt_flushoncommit),
232 fsparam_string("max_inline", Opt_max_inline),
233 fsparam_u32("metadata_ratio", Opt_ratio),
234 fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree),
235 fsparam_flag("skip_balance", Opt_skip_balance),
236 fsparam_flag_no("space_cache", Opt_space_cache),
237 fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache),
238 fsparam_flag_no("ssd", Opt_ssd),
239 fsparam_flag_no("ssd_spread", Opt_ssd_spread),
240 fsparam_string("subvol", Opt_subvol),
241 fsparam_flag("subvol=", Opt_subvol_empty),
242 fsparam_u64("subvolid", Opt_subvolid),
243 fsparam_u32("thread_pool", Opt_thread_pool),
244 fsparam_flag_no("treelog", Opt_treelog),
245 fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed),
246
247 /* Rescue options. */
248 fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue),
249 /* Deprecated, with alias rescue=nologreplay */
250 __fsparam(NULL, "nologreplay", Opt_nologreplay, fs_param_deprecated, NULL),
251 /* Deprecated, with alias rescue=usebackuproot */
252 __fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL),
253 /* For compatibility only, alias for "rescue=nologreplay". */
254 fsparam_flag("norecovery", Opt_norecovery),
255
256 /* Debugging options. */
257 fsparam_flag_no("enospc_debug", Opt_enospc_debug),
258#ifdef CONFIG_BTRFS_DEBUG
259 fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment),
260#endif
261#ifdef CONFIG_BTRFS_FS_REF_VERIFY
262 fsparam_flag("ref_verify", Opt_ref_verify),
263#endif
264 {}
265};
266
267/* No support for restricting writes to btrfs devices yet... */
268static inline blk_mode_t btrfs_open_mode(struct fs_context *fc)
269{
270 return sb_open_mode(fc->sb_flags) & ~BLK_OPEN_RESTRICT_WRITES;
271}
272
273static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
274{
275 struct btrfs_fs_context *ctx = fc->fs_private;
276 struct fs_parse_result result;
277 int opt;
278
279 opt = fs_parse(fc, btrfs_fs_parameters, param, &result);
280 if (opt < 0)
281 return opt;
282
283 switch (opt) {
284 case Opt_degraded:
285 btrfs_set_opt(ctx->mount_opt, DEGRADED);
286 break;
287 case Opt_subvol_empty:
288 /*
289 * This exists because we used to allow it on accident, so we're
290 * keeping it to maintain ABI. See 37becec95ac3 ("Btrfs: allow
291 * empty subvol= again").
292 */
293 break;
294 case Opt_subvol:
295 kfree(ctx->subvol_name);
296 ctx->subvol_name = kstrdup(param->string, GFP_KERNEL);
297 if (!ctx->subvol_name)
298 return -ENOMEM;
299 break;
300 case Opt_subvolid:
301 ctx->subvol_objectid = result.uint_64;
302
303 /* subvolid=0 means give me the original fs_tree. */
304 if (!ctx->subvol_objectid)
305 ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
306 break;
307 case Opt_device: {
308 struct btrfs_device *device;
309 blk_mode_t mode = btrfs_open_mode(fc);
310
311 mutex_lock(&uuid_mutex);
312 device = btrfs_scan_one_device(param->string, mode, false);
313 mutex_unlock(&uuid_mutex);
314 if (IS_ERR(device))
315 return PTR_ERR(device);
316 break;
317 }
318 case Opt_datasum:
319 if (result.negated) {
320 btrfs_set_opt(ctx->mount_opt, NODATASUM);
321 } else {
322 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
323 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
324 }
325 break;
326 case Opt_datacow:
327 if (result.negated) {
328 btrfs_clear_opt(ctx->mount_opt, COMPRESS);
329 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
330 btrfs_set_opt(ctx->mount_opt, NODATACOW);
331 btrfs_set_opt(ctx->mount_opt, NODATASUM);
332 } else {
333 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
334 }
335 break;
336 case Opt_compress_force:
337 case Opt_compress_force_type:
338 btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS);
339 fallthrough;
340 case Opt_compress:
341 case Opt_compress_type:
342 /*
343 * Provide the same semantics as older kernels that don't use fs
344 * context, specifying the "compress" option clears
345 * "force-compress" without the need to pass
346 * "compress-force=[no|none]" before specifying "compress".
347 */
348 if (opt != Opt_compress_force && opt != Opt_compress_force_type)
349 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
350
351 if (opt == Opt_compress || opt == Opt_compress_force) {
352 ctx->compress_type = BTRFS_COMPRESS_ZLIB;
353 ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
354 btrfs_set_opt(ctx->mount_opt, COMPRESS);
355 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
356 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
357 } else if (strncmp(param->string, "zlib", 4) == 0) {
358 ctx->compress_type = BTRFS_COMPRESS_ZLIB;
359 ctx->compress_level =
360 btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB,
361 param->string + 4);
362 btrfs_set_opt(ctx->mount_opt, COMPRESS);
363 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
364 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
365 } else if (strncmp(param->string, "lzo", 3) == 0) {
366 ctx->compress_type = BTRFS_COMPRESS_LZO;
367 ctx->compress_level = 0;
368 btrfs_set_opt(ctx->mount_opt, COMPRESS);
369 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
370 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
371 } else if (strncmp(param->string, "zstd", 4) == 0) {
372 ctx->compress_type = BTRFS_COMPRESS_ZSTD;
373 ctx->compress_level =
374 btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD,
375 param->string + 4);
376 btrfs_set_opt(ctx->mount_opt, COMPRESS);
377 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
378 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
379 } else if (strncmp(param->string, "no", 2) == 0) {
380 ctx->compress_level = 0;
381 ctx->compress_type = 0;
382 btrfs_clear_opt(ctx->mount_opt, COMPRESS);
383 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
384 } else {
385 btrfs_err(NULL, "unrecognized compression value %s",
386 param->string);
387 return -EINVAL;
388 }
389 break;
390 case Opt_ssd:
391 if (result.negated) {
392 btrfs_set_opt(ctx->mount_opt, NOSSD);
393 btrfs_clear_opt(ctx->mount_opt, SSD);
394 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
395 } else {
396 btrfs_set_opt(ctx->mount_opt, SSD);
397 btrfs_clear_opt(ctx->mount_opt, NOSSD);
398 }
399 break;
400 case Opt_ssd_spread:
401 if (result.negated) {
402 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
403 } else {
404 btrfs_set_opt(ctx->mount_opt, SSD);
405 btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
406 btrfs_clear_opt(ctx->mount_opt, NOSSD);
407 }
408 break;
409 case Opt_barrier:
410 if (result.negated)
411 btrfs_set_opt(ctx->mount_opt, NOBARRIER);
412 else
413 btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
414 break;
415 case Opt_thread_pool:
416 if (result.uint_32 == 0) {
417 btrfs_err(NULL, "invalid value 0 for thread_pool");
418 return -EINVAL;
419 }
420 ctx->thread_pool_size = result.uint_32;
421 break;
422 case Opt_max_inline:
423 ctx->max_inline = memparse(param->string, NULL);
424 break;
425 case Opt_acl:
426 if (result.negated) {
427 fc->sb_flags &= ~SB_POSIXACL;
428 } else {
429#ifdef CONFIG_BTRFS_FS_POSIX_ACL
430 fc->sb_flags |= SB_POSIXACL;
431#else
432 btrfs_err(NULL, "support for ACL not compiled in");
433 return -EINVAL;
434#endif
435 }
436 /*
437 * VFS limits the ability to toggle ACL on and off via remount,
438 * despite every file system allowing this. This seems to be
439 * an oversight since we all do, but it'll fail if we're
440 * remounting. So don't set the mask here, we'll check it in
441 * btrfs_reconfigure and do the toggling ourselves.
442 */
443 if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
444 fc->sb_flags_mask |= SB_POSIXACL;
445 break;
446 case Opt_treelog:
447 if (result.negated)
448 btrfs_set_opt(ctx->mount_opt, NOTREELOG);
449 else
450 btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
451 break;
452 case Opt_nologreplay:
453 btrfs_warn(NULL,
454 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
455 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
456 break;
457 case Opt_norecovery:
458 btrfs_info(NULL,
459"'norecovery' is for compatibility only, recommended to use 'rescue=nologreplay'");
460 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
461 break;
462 case Opt_flushoncommit:
463 if (result.negated)
464 btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
465 else
466 btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
467 break;
468 case Opt_ratio:
469 ctx->metadata_ratio = result.uint_32;
470 break;
471 case Opt_discard:
472 if (result.negated) {
473 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
474 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
475 btrfs_set_opt(ctx->mount_opt, NODISCARD);
476 } else {
477 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
478 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
479 }
480 break;
481 case Opt_discard_mode:
482 switch (result.uint_32) {
483 case Opt_discard_sync:
484 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
485 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
486 break;
487 case Opt_discard_async:
488 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
489 btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
490 break;
491 default:
492 btrfs_err(NULL, "unrecognized discard mode value %s",
493 param->key);
494 return -EINVAL;
495 }
496 btrfs_clear_opt(ctx->mount_opt, NODISCARD);
497 break;
498 case Opt_space_cache:
499 if (result.negated) {
500 btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
501 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
502 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
503 } else {
504 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
505 btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
506 }
507 break;
508 case Opt_space_cache_version:
509 switch (result.uint_32) {
510 case Opt_space_cache_v1:
511 btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
512 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
513 break;
514 case Opt_space_cache_v2:
515 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
516 btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
517 break;
518 default:
519 btrfs_err(NULL, "unrecognized space_cache value %s",
520 param->key);
521 return -EINVAL;
522 }
523 break;
524 case Opt_rescan_uuid_tree:
525 btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
526 break;
527 case Opt_clear_cache:
528 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
529 break;
530 case Opt_user_subvol_rm_allowed:
531 btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
532 break;
533 case Opt_enospc_debug:
534 if (result.negated)
535 btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
536 else
537 btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
538 break;
539 case Opt_defrag:
540 if (result.negated)
541 btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
542 else
543 btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
544 break;
545 case Opt_usebackuproot:
546 btrfs_warn(NULL,
547 "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
548 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
549
550 /* If we're loading the backup roots we can't trust the space cache. */
551 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
552 break;
553 case Opt_skip_balance:
554 btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
555 break;
556 case Opt_fatal_errors:
557 switch (result.uint_32) {
558 case Opt_fatal_errors_panic:
559 btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
560 break;
561 case Opt_fatal_errors_bug:
562 btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
563 break;
564 default:
565 btrfs_err(NULL, "unrecognized fatal_errors value %s",
566 param->key);
567 return -EINVAL;
568 }
569 break;
570 case Opt_commit_interval:
571 ctx->commit_interval = result.uint_32;
572 if (ctx->commit_interval == 0)
573 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
574 break;
575 case Opt_rescue:
576 switch (result.uint_32) {
577 case Opt_rescue_usebackuproot:
578 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
579 break;
580 case Opt_rescue_nologreplay:
581 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
582 break;
583 case Opt_rescue_ignorebadroots:
584 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
585 break;
586 case Opt_rescue_ignoredatacsums:
587 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
588 break;
589 case Opt_rescue_ignoremetacsums:
590 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
591 break;
592 case Opt_rescue_ignoresuperflags:
593 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
594 break;
595 case Opt_rescue_parameter_all:
596 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
597 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
598 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
599 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
600 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
601 break;
602 default:
603 btrfs_info(NULL, "unrecognized rescue option '%s'",
604 param->key);
605 return -EINVAL;
606 }
607 break;
608#ifdef CONFIG_BTRFS_DEBUG
609 case Opt_fragment:
610 switch (result.uint_32) {
611 case Opt_fragment_parameter_all:
612 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
613 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
614 break;
615 case Opt_fragment_parameter_metadata:
616 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
617 break;
618 case Opt_fragment_parameter_data:
619 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
620 break;
621 default:
622 btrfs_info(NULL, "unrecognized fragment option '%s'",
623 param->key);
624 return -EINVAL;
625 }
626 break;
627#endif
628#ifdef CONFIG_BTRFS_FS_REF_VERIFY
629 case Opt_ref_verify:
630 btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
631 break;
632#endif
633 default:
634 btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
635 return -EINVAL;
636 }
637
638 return 0;
639}
640
641/*
642 * Some options only have meaning at mount time and shouldn't persist across
643 * remounts, or be displayed. Clear these at the end of mount and remount code
644 * paths.
645 */
646static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
647{
648 btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
649 btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
650 btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
651}
652
653static bool check_ro_option(const struct btrfs_fs_info *fs_info,
654 unsigned long long mount_opt, unsigned long long opt,
655 const char *opt_name)
656{
657 if (mount_opt & opt) {
658 btrfs_err(fs_info, "%s must be used with ro mount option",
659 opt_name);
660 return true;
661 }
662 return false;
663}
664
665bool btrfs_check_options(const struct btrfs_fs_info *info,
666 unsigned long long *mount_opt,
667 unsigned long flags)
668{
669 bool ret = true;
670
671 if (!(flags & SB_RDONLY) &&
672 (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
673 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
674 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums") ||
675 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREMETACSUMS, "ignoremetacsums") ||
676 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNORESUPERFLAGS, "ignoresuperflags")))
677 ret = false;
678
679 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
680 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
681 !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
682 btrfs_err(info, "cannot disable free-space-tree");
683 ret = false;
684 }
685 if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
686 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
687 btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
688 ret = false;
689 }
690
691 if (btrfs_check_mountopts_zoned(info, mount_opt))
692 ret = false;
693
694 if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
695 if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE)) {
696 btrfs_info(info, "disk space caching is enabled");
697 btrfs_warn(info,
698"space cache v1 is being deprecated and will be removed in a future release, please use -o space_cache=v2");
699 }
700 if (btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE))
701 btrfs_info(info, "using free-space-tree");
702 }
703
704 return ret;
705}
706
707/*
708 * This is subtle, we only call this during open_ctree(). We need to pre-load
709 * the mount options with the on-disk settings. Before the new mount API took
710 * effect we would do this on mount and remount. With the new mount API we'll
711 * only do this on the initial mount.
712 *
713 * This isn't a change in behavior, because we're using the current state of the
714 * file system to set the current mount options. If you mounted with special
715 * options to disable these features and then remounted we wouldn't revert the
716 * settings, because mounting without these features cleared the on-disk
717 * settings, so this being called on re-mount is not needed.
718 */
719void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
720{
721 if (fs_info->sectorsize < PAGE_SIZE) {
722 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
723 if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) {
724 btrfs_info(fs_info,
725 "forcing free space tree for sector size %u with page size %lu",
726 fs_info->sectorsize, PAGE_SIZE);
727 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
728 }
729 }
730
731 /*
732 * At this point our mount options are populated, so we only mess with
733 * these settings if we don't have any settings already.
734 */
735 if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
736 return;
737
738 if (btrfs_is_zoned(fs_info) &&
739 btrfs_free_space_cache_v1_active(fs_info)) {
740 btrfs_info(fs_info, "zoned: clearing existing space cache");
741 btrfs_set_super_cache_generation(fs_info->super_copy, 0);
742 return;
743 }
744
745 if (btrfs_test_opt(fs_info, SPACE_CACHE))
746 return;
747
748 if (btrfs_test_opt(fs_info, NOSPACECACHE))
749 return;
750
751 /*
752 * At this point we don't have explicit options set by the user, set
753 * them ourselves based on the state of the file system.
754 */
755 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
756 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
757 else if (btrfs_free_space_cache_v1_active(fs_info))
758 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
759}
760
761static void set_device_specific_options(struct btrfs_fs_info *fs_info)
762{
763 if (!btrfs_test_opt(fs_info, NOSSD) &&
764 !fs_info->fs_devices->rotating)
765 btrfs_set_opt(fs_info->mount_opt, SSD);
766
767 /*
768 * For devices supporting discard turn on discard=async automatically,
769 * unless it's already set or disabled. This could be turned off by
770 * nodiscard for the same mount.
771 *
772 * The zoned mode piggy backs on the discard functionality for
773 * resetting a zone. There is no reason to delay the zone reset as it is
774 * fast enough. So, do not enable async discard for zoned mode.
775 */
776 if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
777 btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
778 btrfs_test_opt(fs_info, NODISCARD)) &&
779 fs_info->fs_devices->discardable &&
780 !btrfs_is_zoned(fs_info))
781 btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
782}
783
784char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
785 u64 subvol_objectid)
786{
787 struct btrfs_root *root = fs_info->tree_root;
788 struct btrfs_root *fs_root = NULL;
789 struct btrfs_root_ref *root_ref;
790 struct btrfs_inode_ref *inode_ref;
791 struct btrfs_key key;
792 struct btrfs_path *path = NULL;
793 char *name = NULL, *ptr;
794 u64 dirid;
795 int len;
796 int ret;
797
798 path = btrfs_alloc_path();
799 if (!path) {
800 ret = -ENOMEM;
801 goto err;
802 }
803
804 name = kmalloc(PATH_MAX, GFP_KERNEL);
805 if (!name) {
806 ret = -ENOMEM;
807 goto err;
808 }
809 ptr = name + PATH_MAX - 1;
810 ptr[0] = '\0';
811
812 /*
813 * Walk up the subvolume trees in the tree of tree roots by root
814 * backrefs until we hit the top-level subvolume.
815 */
816 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
817 key.objectid = subvol_objectid;
818 key.type = BTRFS_ROOT_BACKREF_KEY;
819 key.offset = (u64)-1;
820
821 ret = btrfs_search_backwards(root, &key, path);
822 if (ret < 0) {
823 goto err;
824 } else if (ret > 0) {
825 ret = -ENOENT;
826 goto err;
827 }
828
829 subvol_objectid = key.offset;
830
831 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
832 struct btrfs_root_ref);
833 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
834 ptr -= len + 1;
835 if (ptr < name) {
836 ret = -ENAMETOOLONG;
837 goto err;
838 }
839 read_extent_buffer(path->nodes[0], ptr + 1,
840 (unsigned long)(root_ref + 1), len);
841 ptr[0] = '/';
842 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
843 btrfs_release_path(path);
844
845 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
846 if (IS_ERR(fs_root)) {
847 ret = PTR_ERR(fs_root);
848 fs_root = NULL;
849 goto err;
850 }
851
852 /*
853 * Walk up the filesystem tree by inode refs until we hit the
854 * root directory.
855 */
856 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
857 key.objectid = dirid;
858 key.type = BTRFS_INODE_REF_KEY;
859 key.offset = (u64)-1;
860
861 ret = btrfs_search_backwards(fs_root, &key, path);
862 if (ret < 0) {
863 goto err;
864 } else if (ret > 0) {
865 ret = -ENOENT;
866 goto err;
867 }
868
869 dirid = key.offset;
870
871 inode_ref = btrfs_item_ptr(path->nodes[0],
872 path->slots[0],
873 struct btrfs_inode_ref);
874 len = btrfs_inode_ref_name_len(path->nodes[0],
875 inode_ref);
876 ptr -= len + 1;
877 if (ptr < name) {
878 ret = -ENAMETOOLONG;
879 goto err;
880 }
881 read_extent_buffer(path->nodes[0], ptr + 1,
882 (unsigned long)(inode_ref + 1), len);
883 ptr[0] = '/';
884 btrfs_release_path(path);
885 }
886 btrfs_put_root(fs_root);
887 fs_root = NULL;
888 }
889
890 btrfs_free_path(path);
891 if (ptr == name + PATH_MAX - 1) {
892 name[0] = '/';
893 name[1] = '\0';
894 } else {
895 memmove(name, ptr, name + PATH_MAX - ptr);
896 }
897 return name;
898
899err:
900 btrfs_put_root(fs_root);
901 btrfs_free_path(path);
902 kfree(name);
903 return ERR_PTR(ret);
904}
905
906static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
907{
908 struct btrfs_root *root = fs_info->tree_root;
909 struct btrfs_dir_item *di;
910 struct btrfs_path *path;
911 struct btrfs_key location;
912 struct fscrypt_str name = FSTR_INIT("default", 7);
913 u64 dir_id;
914
915 path = btrfs_alloc_path();
916 if (!path)
917 return -ENOMEM;
918
919 /*
920 * Find the "default" dir item which points to the root item that we
921 * will mount by default if we haven't been given a specific subvolume
922 * to mount.
923 */
924 dir_id = btrfs_super_root_dir(fs_info->super_copy);
925 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
926 if (IS_ERR(di)) {
927 btrfs_free_path(path);
928 return PTR_ERR(di);
929 }
930 if (!di) {
931 /*
932 * Ok the default dir item isn't there. This is weird since
933 * it's always been there, but don't freak out, just try and
934 * mount the top-level subvolume.
935 */
936 btrfs_free_path(path);
937 *objectid = BTRFS_FS_TREE_OBJECTID;
938 return 0;
939 }
940
941 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
942 btrfs_free_path(path);
943 *objectid = location.objectid;
944 return 0;
945}
946
947static int btrfs_fill_super(struct super_block *sb,
948 struct btrfs_fs_devices *fs_devices)
949{
950 struct inode *inode;
951 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
952 int err;
953
954 sb->s_maxbytes = MAX_LFS_FILESIZE;
955 sb->s_magic = BTRFS_SUPER_MAGIC;
956 sb->s_op = &btrfs_super_ops;
957 sb->s_d_op = &btrfs_dentry_operations;
958 sb->s_export_op = &btrfs_export_ops;
959#ifdef CONFIG_FS_VERITY
960 sb->s_vop = &btrfs_verityops;
961#endif
962 sb->s_xattr = btrfs_xattr_handlers;
963 sb->s_time_gran = 1;
964 sb->s_iflags |= SB_I_CGROUPWB;
965
966 err = super_setup_bdi(sb);
967 if (err) {
968 btrfs_err(fs_info, "super_setup_bdi failed");
969 return err;
970 }
971
972 err = open_ctree(sb, fs_devices);
973 if (err) {
974 btrfs_err(fs_info, "open_ctree failed: %d", err);
975 return err;
976 }
977
978 inode = btrfs_iget(BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
979 if (IS_ERR(inode)) {
980 err = PTR_ERR(inode);
981 btrfs_handle_fs_error(fs_info, err, NULL);
982 goto fail_close;
983 }
984
985 sb->s_root = d_make_root(inode);
986 if (!sb->s_root) {
987 err = -ENOMEM;
988 goto fail_close;
989 }
990
991 sb->s_flags |= SB_ACTIVE;
992 return 0;
993
994fail_close:
995 close_ctree(fs_info);
996 return err;
997}
998
999int btrfs_sync_fs(struct super_block *sb, int wait)
1000{
1001 struct btrfs_trans_handle *trans;
1002 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1003 struct btrfs_root *root = fs_info->tree_root;
1004
1005 trace_btrfs_sync_fs(fs_info, wait);
1006
1007 if (!wait) {
1008 filemap_flush(fs_info->btree_inode->i_mapping);
1009 return 0;
1010 }
1011
1012 btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL);
1013
1014 trans = btrfs_attach_transaction_barrier(root);
1015 if (IS_ERR(trans)) {
1016 /* no transaction, don't bother */
1017 if (PTR_ERR(trans) == -ENOENT) {
1018 /*
1019 * Exit unless we have some pending changes
1020 * that need to go through commit
1021 */
1022 if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1023 &fs_info->flags))
1024 return 0;
1025 /*
1026 * A non-blocking test if the fs is frozen. We must not
1027 * start a new transaction here otherwise a deadlock
1028 * happens. The pending operations are delayed to the
1029 * next commit after thawing.
1030 */
1031 if (sb_start_write_trylock(sb))
1032 sb_end_write(sb);
1033 else
1034 return 0;
1035 trans = btrfs_start_transaction(root, 0);
1036 }
1037 if (IS_ERR(trans))
1038 return PTR_ERR(trans);
1039 }
1040 return btrfs_commit_transaction(trans);
1041}
1042
1043static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1044{
1045 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1046 *printed = true;
1047}
1048
1049static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1050{
1051 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1052 const char *compress_type;
1053 const char *subvol_name;
1054 bool printed = false;
1055
1056 if (btrfs_test_opt(info, DEGRADED))
1057 seq_puts(seq, ",degraded");
1058 if (btrfs_test_opt(info, NODATASUM))
1059 seq_puts(seq, ",nodatasum");
1060 if (btrfs_test_opt(info, NODATACOW))
1061 seq_puts(seq, ",nodatacow");
1062 if (btrfs_test_opt(info, NOBARRIER))
1063 seq_puts(seq, ",nobarrier");
1064 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1065 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1066 if (info->thread_pool_size != min_t(unsigned long,
1067 num_online_cpus() + 2, 8))
1068 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1069 if (btrfs_test_opt(info, COMPRESS)) {
1070 compress_type = btrfs_compress_type2str(info->compress_type);
1071 if (btrfs_test_opt(info, FORCE_COMPRESS))
1072 seq_printf(seq, ",compress-force=%s", compress_type);
1073 else
1074 seq_printf(seq, ",compress=%s", compress_type);
1075 if (info->compress_level)
1076 seq_printf(seq, ":%d", info->compress_level);
1077 }
1078 if (btrfs_test_opt(info, NOSSD))
1079 seq_puts(seq, ",nossd");
1080 if (btrfs_test_opt(info, SSD_SPREAD))
1081 seq_puts(seq, ",ssd_spread");
1082 else if (btrfs_test_opt(info, SSD))
1083 seq_puts(seq, ",ssd");
1084 if (btrfs_test_opt(info, NOTREELOG))
1085 seq_puts(seq, ",notreelog");
1086 if (btrfs_test_opt(info, NOLOGREPLAY))
1087 print_rescue_option(seq, "nologreplay", &printed);
1088 if (btrfs_test_opt(info, USEBACKUPROOT))
1089 print_rescue_option(seq, "usebackuproot", &printed);
1090 if (btrfs_test_opt(info, IGNOREBADROOTS))
1091 print_rescue_option(seq, "ignorebadroots", &printed);
1092 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1093 print_rescue_option(seq, "ignoredatacsums", &printed);
1094 if (btrfs_test_opt(info, IGNOREMETACSUMS))
1095 print_rescue_option(seq, "ignoremetacsums", &printed);
1096 if (btrfs_test_opt(info, IGNORESUPERFLAGS))
1097 print_rescue_option(seq, "ignoresuperflags", &printed);
1098 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1099 seq_puts(seq, ",flushoncommit");
1100 if (btrfs_test_opt(info, DISCARD_SYNC))
1101 seq_puts(seq, ",discard");
1102 if (btrfs_test_opt(info, DISCARD_ASYNC))
1103 seq_puts(seq, ",discard=async");
1104 if (!(info->sb->s_flags & SB_POSIXACL))
1105 seq_puts(seq, ",noacl");
1106 if (btrfs_free_space_cache_v1_active(info))
1107 seq_puts(seq, ",space_cache");
1108 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1109 seq_puts(seq, ",space_cache=v2");
1110 else
1111 seq_puts(seq, ",nospace_cache");
1112 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1113 seq_puts(seq, ",rescan_uuid_tree");
1114 if (btrfs_test_opt(info, CLEAR_CACHE))
1115 seq_puts(seq, ",clear_cache");
1116 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1117 seq_puts(seq, ",user_subvol_rm_allowed");
1118 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1119 seq_puts(seq, ",enospc_debug");
1120 if (btrfs_test_opt(info, AUTO_DEFRAG))
1121 seq_puts(seq, ",autodefrag");
1122 if (btrfs_test_opt(info, SKIP_BALANCE))
1123 seq_puts(seq, ",skip_balance");
1124 if (info->metadata_ratio)
1125 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1126 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1127 seq_puts(seq, ",fatal_errors=panic");
1128 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1129 seq_printf(seq, ",commit=%u", info->commit_interval);
1130#ifdef CONFIG_BTRFS_DEBUG
1131 if (btrfs_test_opt(info, FRAGMENT_DATA))
1132 seq_puts(seq, ",fragment=data");
1133 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1134 seq_puts(seq, ",fragment=metadata");
1135#endif
1136 if (btrfs_test_opt(info, REF_VERIFY))
1137 seq_puts(seq, ",ref_verify");
1138 seq_printf(seq, ",subvolid=%llu", btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1139 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1140 btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1141 if (!IS_ERR(subvol_name)) {
1142 seq_puts(seq, ",subvol=");
1143 seq_escape(seq, subvol_name, " \t\n\\");
1144 kfree(subvol_name);
1145 }
1146 return 0;
1147}
1148
1149/*
1150 * subvolumes are identified by ino 256
1151 */
1152static inline int is_subvolume_inode(struct inode *inode)
1153{
1154 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1155 return 1;
1156 return 0;
1157}
1158
1159static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1160 struct vfsmount *mnt)
1161{
1162 struct dentry *root;
1163 int ret;
1164
1165 if (!subvol_name) {
1166 if (!subvol_objectid) {
1167 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1168 &subvol_objectid);
1169 if (ret) {
1170 root = ERR_PTR(ret);
1171 goto out;
1172 }
1173 }
1174 subvol_name = btrfs_get_subvol_name_from_objectid(
1175 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1176 if (IS_ERR(subvol_name)) {
1177 root = ERR_CAST(subvol_name);
1178 subvol_name = NULL;
1179 goto out;
1180 }
1181
1182 }
1183
1184 root = mount_subtree(mnt, subvol_name);
1185 /* mount_subtree() drops our reference on the vfsmount. */
1186 mnt = NULL;
1187
1188 if (!IS_ERR(root)) {
1189 struct super_block *s = root->d_sb;
1190 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1191 struct inode *root_inode = d_inode(root);
1192 u64 root_objectid = btrfs_root_id(BTRFS_I(root_inode)->root);
1193
1194 ret = 0;
1195 if (!is_subvolume_inode(root_inode)) {
1196 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1197 subvol_name);
1198 ret = -EINVAL;
1199 }
1200 if (subvol_objectid && root_objectid != subvol_objectid) {
1201 /*
1202 * This will also catch a race condition where a
1203 * subvolume which was passed by ID is renamed and
1204 * another subvolume is renamed over the old location.
1205 */
1206 btrfs_err(fs_info,
1207 "subvol '%s' does not match subvolid %llu",
1208 subvol_name, subvol_objectid);
1209 ret = -EINVAL;
1210 }
1211 if (ret) {
1212 dput(root);
1213 root = ERR_PTR(ret);
1214 deactivate_locked_super(s);
1215 }
1216 }
1217
1218out:
1219 mntput(mnt);
1220 kfree(subvol_name);
1221 return root;
1222}
1223
1224static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1225 u32 new_pool_size, u32 old_pool_size)
1226{
1227 if (new_pool_size == old_pool_size)
1228 return;
1229
1230 fs_info->thread_pool_size = new_pool_size;
1231
1232 btrfs_info(fs_info, "resize thread pool %d -> %d",
1233 old_pool_size, new_pool_size);
1234
1235 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1236 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1237 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1238 workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1239 workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1240 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1241 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1242 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1243}
1244
1245static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1246 unsigned long long old_opts, int flags)
1247{
1248 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1249 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1250 (flags & SB_RDONLY))) {
1251 /* wait for any defraggers to finish */
1252 wait_event(fs_info->transaction_wait,
1253 (atomic_read(&fs_info->defrag_running) == 0));
1254 if (flags & SB_RDONLY)
1255 sync_filesystem(fs_info->sb);
1256 }
1257}
1258
1259static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1260 unsigned long long old_opts)
1261{
1262 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1263
1264 /*
1265 * We need to cleanup all defragable inodes if the autodefragment is
1266 * close or the filesystem is read only.
1267 */
1268 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1269 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1270 btrfs_cleanup_defrag_inodes(fs_info);
1271 }
1272
1273 /* If we toggled discard async */
1274 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1275 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1276 btrfs_discard_resume(fs_info);
1277 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1278 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1279 btrfs_discard_cleanup(fs_info);
1280
1281 /* If we toggled space cache */
1282 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1283 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1284}
1285
1286static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1287{
1288 int ret;
1289
1290 if (BTRFS_FS_ERROR(fs_info)) {
1291 btrfs_err(fs_info,
1292 "remounting read-write after error is not allowed");
1293 return -EINVAL;
1294 }
1295
1296 if (fs_info->fs_devices->rw_devices == 0)
1297 return -EACCES;
1298
1299 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1300 btrfs_warn(fs_info,
1301 "too many missing devices, writable remount is not allowed");
1302 return -EACCES;
1303 }
1304
1305 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1306 btrfs_warn(fs_info,
1307 "mount required to replay tree-log, cannot remount read-write");
1308 return -EINVAL;
1309 }
1310
1311 /*
1312 * NOTE: when remounting with a change that does writes, don't put it
1313 * anywhere above this point, as we are not sure to be safe to write
1314 * until we pass the above checks.
1315 */
1316 ret = btrfs_start_pre_rw_mount(fs_info);
1317 if (ret)
1318 return ret;
1319
1320 btrfs_clear_sb_rdonly(fs_info->sb);
1321
1322 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1323
1324 /*
1325 * If we've gone from readonly -> read-write, we need to get our
1326 * sync/async discard lists in the right state.
1327 */
1328 btrfs_discard_resume(fs_info);
1329
1330 return 0;
1331}
1332
1333static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1334{
1335 /*
1336 * This also happens on 'umount -rf' or on shutdown, when the
1337 * filesystem is busy.
1338 */
1339 cancel_work_sync(&fs_info->async_reclaim_work);
1340 cancel_work_sync(&fs_info->async_data_reclaim_work);
1341
1342 btrfs_discard_cleanup(fs_info);
1343
1344 /* Wait for the uuid_scan task to finish */
1345 down(&fs_info->uuid_tree_rescan_sem);
1346 /* Avoid complains from lockdep et al. */
1347 up(&fs_info->uuid_tree_rescan_sem);
1348
1349 btrfs_set_sb_rdonly(fs_info->sb);
1350
1351 /*
1352 * Setting SB_RDONLY will put the cleaner thread to sleep at the next
1353 * loop if it's already active. If it's already asleep, we'll leave
1354 * unused block groups on disk until we're mounted read-write again
1355 * unless we clean them up here.
1356 */
1357 btrfs_delete_unused_bgs(fs_info);
1358
1359 /*
1360 * The cleaner task could be already running before we set the flag
1361 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock). We must make
1362 * sure that after we finish the remount, i.e. after we call
1363 * btrfs_commit_super(), the cleaner can no longer start a transaction
1364 * - either because it was dropping a dead root, running delayed iputs
1365 * or deleting an unused block group (the cleaner picked a block
1366 * group from the list of unused block groups before we were able to
1367 * in the previous call to btrfs_delete_unused_bgs()).
1368 */
1369 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1370
1371 /*
1372 * We've set the superblock to RO mode, so we might have made the
1373 * cleaner task sleep without running all pending delayed iputs. Go
1374 * through all the delayed iputs here, so that if an unmount happens
1375 * without remounting RW we don't end up at finishing close_ctree()
1376 * with a non-empty list of delayed iputs.
1377 */
1378 btrfs_run_delayed_iputs(fs_info);
1379
1380 btrfs_dev_replace_suspend_for_unmount(fs_info);
1381 btrfs_scrub_cancel(fs_info);
1382 btrfs_pause_balance(fs_info);
1383
1384 /*
1385 * Pause the qgroup rescan worker if it is running. We don't want it to
1386 * be still running after we are in RO mode, as after that, by the time
1387 * we unmount, it might have left a transaction open, so we would leak
1388 * the transaction and/or crash.
1389 */
1390 btrfs_qgroup_wait_for_completion(fs_info, false);
1391
1392 return btrfs_commit_super(fs_info);
1393}
1394
1395static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1396{
1397 fs_info->max_inline = ctx->max_inline;
1398 fs_info->commit_interval = ctx->commit_interval;
1399 fs_info->metadata_ratio = ctx->metadata_ratio;
1400 fs_info->thread_pool_size = ctx->thread_pool_size;
1401 fs_info->mount_opt = ctx->mount_opt;
1402 fs_info->compress_type = ctx->compress_type;
1403 fs_info->compress_level = ctx->compress_level;
1404}
1405
1406static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1407{
1408 ctx->max_inline = fs_info->max_inline;
1409 ctx->commit_interval = fs_info->commit_interval;
1410 ctx->metadata_ratio = fs_info->metadata_ratio;
1411 ctx->thread_pool_size = fs_info->thread_pool_size;
1412 ctx->mount_opt = fs_info->mount_opt;
1413 ctx->compress_type = fs_info->compress_type;
1414 ctx->compress_level = fs_info->compress_level;
1415}
1416
1417#define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...) \
1418do { \
1419 if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1420 btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1421 btrfs_info(fs_info, fmt, ##args); \
1422} while (0)
1423
1424#define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...) \
1425do { \
1426 if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1427 !btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1428 btrfs_info(fs_info, fmt, ##args); \
1429} while (0)
1430
1431static void btrfs_emit_options(struct btrfs_fs_info *info,
1432 struct btrfs_fs_context *old)
1433{
1434 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1435 btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1436 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1437 btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1438 btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1439 btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1440 btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1441 btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1442 btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1443 btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1444 btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1445 btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1446 btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1447 btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1448 btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1449 btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1450 btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1451 btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1452 btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1453 btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1454 btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
1455 btrfs_info_if_set(info, old, IGNOREMETACSUMS, "ignoring meta csums");
1456 btrfs_info_if_set(info, old, IGNORESUPERFLAGS, "ignoring unknown super block flags");
1457
1458 btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1459 btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1460 btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1461 btrfs_info_if_unset(info, old, NOBARRIER, "turning off barriers");
1462 btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1463 btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1464 btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1465 btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1466 btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1467
1468 /* Did the compression settings change? */
1469 if (btrfs_test_opt(info, COMPRESS) &&
1470 (!old ||
1471 old->compress_type != info->compress_type ||
1472 old->compress_level != info->compress_level ||
1473 (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1474 btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1475 const char *compress_type = btrfs_compress_type2str(info->compress_type);
1476
1477 btrfs_info(info, "%s %s compression, level %d",
1478 btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1479 compress_type, info->compress_level);
1480 }
1481
1482 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1483 btrfs_info(info, "max_inline set to %llu", info->max_inline);
1484}
1485
1486static int btrfs_reconfigure(struct fs_context *fc)
1487{
1488 struct super_block *sb = fc->root->d_sb;
1489 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1490 struct btrfs_fs_context *ctx = fc->fs_private;
1491 struct btrfs_fs_context old_ctx;
1492 int ret = 0;
1493 bool mount_reconfigure = (fc->s_fs_info != NULL);
1494
1495 btrfs_info_to_ctx(fs_info, &old_ctx);
1496
1497 /*
1498 * This is our "bind mount" trick, we don't want to allow the user to do
1499 * anything other than mount a different ro/rw and a different subvol,
1500 * all of the mount options should be maintained.
1501 */
1502 if (mount_reconfigure)
1503 ctx->mount_opt = old_ctx.mount_opt;
1504
1505 sync_filesystem(sb);
1506 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1507
1508 if (!btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags))
1509 return -EINVAL;
1510
1511 ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
1512 if (ret < 0)
1513 return ret;
1514
1515 btrfs_ctx_to_info(fs_info, ctx);
1516 btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags);
1517 btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size,
1518 old_ctx.thread_pool_size);
1519
1520 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1521 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1522 (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1523 btrfs_warn(fs_info,
1524 "remount supports changing free space tree only from RO to RW");
1525 /* Make sure free space cache options match the state on disk. */
1526 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1527 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1528 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1529 }
1530 if (btrfs_free_space_cache_v1_active(fs_info)) {
1531 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1532 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1533 }
1534 }
1535
1536 ret = 0;
1537 if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1538 ret = btrfs_remount_ro(fs_info);
1539 else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1540 ret = btrfs_remount_rw(fs_info);
1541 if (ret)
1542 goto restore;
1543
1544 /*
1545 * If we set the mask during the parameter parsing VFS would reject the
1546 * remount. Here we can set the mask and the value will be updated
1547 * appropriately.
1548 */
1549 if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1550 fc->sb_flags_mask |= SB_POSIXACL;
1551
1552 btrfs_emit_options(fs_info, &old_ctx);
1553 wake_up_process(fs_info->transaction_kthread);
1554 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1555 btrfs_clear_oneshot_options(fs_info);
1556 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1557
1558 return 0;
1559restore:
1560 btrfs_ctx_to_info(fs_info, &old_ctx);
1561 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1562 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1563 return ret;
1564}
1565
1566/* Used to sort the devices by max_avail(descending sort) */
1567static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1568{
1569 const struct btrfs_device_info *dev_info1 = a;
1570 const struct btrfs_device_info *dev_info2 = b;
1571
1572 if (dev_info1->max_avail > dev_info2->max_avail)
1573 return -1;
1574 else if (dev_info1->max_avail < dev_info2->max_avail)
1575 return 1;
1576 return 0;
1577}
1578
1579/*
1580 * sort the devices by max_avail, in which max free extent size of each device
1581 * is stored.(Descending Sort)
1582 */
1583static inline void btrfs_descending_sort_devices(
1584 struct btrfs_device_info *devices,
1585 size_t nr_devices)
1586{
1587 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1588 btrfs_cmp_device_free_bytes, NULL);
1589}
1590
1591/*
1592 * The helper to calc the free space on the devices that can be used to store
1593 * file data.
1594 */
1595static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1596 u64 *free_bytes)
1597{
1598 struct btrfs_device_info *devices_info;
1599 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1600 struct btrfs_device *device;
1601 u64 type;
1602 u64 avail_space;
1603 u64 min_stripe_size;
1604 int num_stripes = 1;
1605 int i = 0, nr_devices;
1606 const struct btrfs_raid_attr *rattr;
1607
1608 /*
1609 * We aren't under the device list lock, so this is racy-ish, but good
1610 * enough for our purposes.
1611 */
1612 nr_devices = fs_info->fs_devices->open_devices;
1613 if (!nr_devices) {
1614 smp_mb();
1615 nr_devices = fs_info->fs_devices->open_devices;
1616 ASSERT(nr_devices);
1617 if (!nr_devices) {
1618 *free_bytes = 0;
1619 return 0;
1620 }
1621 }
1622
1623 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1624 GFP_KERNEL);
1625 if (!devices_info)
1626 return -ENOMEM;
1627
1628 /* calc min stripe number for data space allocation */
1629 type = btrfs_data_alloc_profile(fs_info);
1630 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1631
1632 if (type & BTRFS_BLOCK_GROUP_RAID0)
1633 num_stripes = nr_devices;
1634 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1635 num_stripes = rattr->ncopies;
1636 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1637 num_stripes = 4;
1638
1639 /* Adjust for more than 1 stripe per device */
1640 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1641
1642 rcu_read_lock();
1643 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1644 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1645 &device->dev_state) ||
1646 !device->bdev ||
1647 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1648 continue;
1649
1650 if (i >= nr_devices)
1651 break;
1652
1653 avail_space = device->total_bytes - device->bytes_used;
1654
1655 /* align with stripe_len */
1656 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1657
1658 /*
1659 * Ensure we have at least min_stripe_size on top of the
1660 * reserved space on the device.
1661 */
1662 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1663 continue;
1664
1665 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1666
1667 devices_info[i].dev = device;
1668 devices_info[i].max_avail = avail_space;
1669
1670 i++;
1671 }
1672 rcu_read_unlock();
1673
1674 nr_devices = i;
1675
1676 btrfs_descending_sort_devices(devices_info, nr_devices);
1677
1678 i = nr_devices - 1;
1679 avail_space = 0;
1680 while (nr_devices >= rattr->devs_min) {
1681 num_stripes = min(num_stripes, nr_devices);
1682
1683 if (devices_info[i].max_avail >= min_stripe_size) {
1684 int j;
1685 u64 alloc_size;
1686
1687 avail_space += devices_info[i].max_avail * num_stripes;
1688 alloc_size = devices_info[i].max_avail;
1689 for (j = i + 1 - num_stripes; j <= i; j++)
1690 devices_info[j].max_avail -= alloc_size;
1691 }
1692 i--;
1693 nr_devices--;
1694 }
1695
1696 kfree(devices_info);
1697 *free_bytes = avail_space;
1698 return 0;
1699}
1700
1701/*
1702 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1703 *
1704 * If there's a redundant raid level at DATA block groups, use the respective
1705 * multiplier to scale the sizes.
1706 *
1707 * Unused device space usage is based on simulating the chunk allocator
1708 * algorithm that respects the device sizes and order of allocations. This is
1709 * a close approximation of the actual use but there are other factors that may
1710 * change the result (like a new metadata chunk).
1711 *
1712 * If metadata is exhausted, f_bavail will be 0.
1713 */
1714static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1715{
1716 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1717 struct btrfs_super_block *disk_super = fs_info->super_copy;
1718 struct btrfs_space_info *found;
1719 u64 total_used = 0;
1720 u64 total_free_data = 0;
1721 u64 total_free_meta = 0;
1722 u32 bits = fs_info->sectorsize_bits;
1723 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1724 unsigned factor = 1;
1725 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1726 int ret;
1727 u64 thresh = 0;
1728 int mixed = 0;
1729
1730 list_for_each_entry(found, &fs_info->space_info, list) {
1731 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1732 int i;
1733
1734 total_free_data += found->disk_total - found->disk_used;
1735 total_free_data -=
1736 btrfs_account_ro_block_groups_free_space(found);
1737
1738 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1739 if (!list_empty(&found->block_groups[i]))
1740 factor = btrfs_bg_type_to_factor(
1741 btrfs_raid_array[i].bg_flag);
1742 }
1743 }
1744
1745 /*
1746 * Metadata in mixed block group profiles are accounted in data
1747 */
1748 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1749 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1750 mixed = 1;
1751 else
1752 total_free_meta += found->disk_total -
1753 found->disk_used;
1754 }
1755
1756 total_used += found->disk_used;
1757 }
1758
1759 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1760 buf->f_blocks >>= bits;
1761 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1762
1763 /* Account global block reserve as used, it's in logical size already */
1764 spin_lock(&block_rsv->lock);
1765 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
1766 if (buf->f_bfree >= block_rsv->size >> bits)
1767 buf->f_bfree -= block_rsv->size >> bits;
1768 else
1769 buf->f_bfree = 0;
1770 spin_unlock(&block_rsv->lock);
1771
1772 buf->f_bavail = div_u64(total_free_data, factor);
1773 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1774 if (ret)
1775 return ret;
1776 buf->f_bavail += div_u64(total_free_data, factor);
1777 buf->f_bavail = buf->f_bavail >> bits;
1778
1779 /*
1780 * We calculate the remaining metadata space minus global reserve. If
1781 * this is (supposedly) smaller than zero, there's no space. But this
1782 * does not hold in practice, the exhausted state happens where's still
1783 * some positive delta. So we apply some guesswork and compare the
1784 * delta to a 4M threshold. (Practically observed delta was ~2M.)
1785 *
1786 * We probably cannot calculate the exact threshold value because this
1787 * depends on the internal reservations requested by various
1788 * operations, so some operations that consume a few metadata will
1789 * succeed even if the Avail is zero. But this is better than the other
1790 * way around.
1791 */
1792 thresh = SZ_4M;
1793
1794 /*
1795 * We only want to claim there's no available space if we can no longer
1796 * allocate chunks for our metadata profile and our global reserve will
1797 * not fit in the free metadata space. If we aren't ->full then we
1798 * still can allocate chunks and thus are fine using the currently
1799 * calculated f_bavail.
1800 */
1801 if (!mixed && block_rsv->space_info->full &&
1802 (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1803 buf->f_bavail = 0;
1804
1805 buf->f_type = BTRFS_SUPER_MAGIC;
1806 buf->f_bsize = fs_info->sectorsize;
1807 buf->f_namelen = BTRFS_NAME_LEN;
1808
1809 /* We treat it as constant endianness (it doesn't matter _which_)
1810 because we want the fsid to come out the same whether mounted
1811 on a big-endian or little-endian host */
1812 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1813 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1814 /* Mask in the root object ID too, to disambiguate subvols */
1815 buf->f_fsid.val[0] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root) >> 32;
1816 buf->f_fsid.val[1] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root);
1817
1818 return 0;
1819}
1820
1821static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1822{
1823 struct btrfs_fs_info *p = fc->s_fs_info;
1824 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1825
1826 return fs_info->fs_devices == p->fs_devices;
1827}
1828
1829static int btrfs_get_tree_super(struct fs_context *fc)
1830{
1831 struct btrfs_fs_info *fs_info = fc->s_fs_info;
1832 struct btrfs_fs_context *ctx = fc->fs_private;
1833 struct btrfs_fs_devices *fs_devices = NULL;
1834 struct block_device *bdev;
1835 struct btrfs_device *device;
1836 struct super_block *sb;
1837 blk_mode_t mode = btrfs_open_mode(fc);
1838 int ret;
1839
1840 btrfs_ctx_to_info(fs_info, ctx);
1841 mutex_lock(&uuid_mutex);
1842
1843 /*
1844 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1845 * either a valid device or an error.
1846 */
1847 device = btrfs_scan_one_device(fc->source, mode, true);
1848 ASSERT(device != NULL);
1849 if (IS_ERR(device)) {
1850 mutex_unlock(&uuid_mutex);
1851 return PTR_ERR(device);
1852 }
1853
1854 fs_devices = device->fs_devices;
1855 fs_info->fs_devices = fs_devices;
1856
1857 ret = btrfs_open_devices(fs_devices, mode, &btrfs_fs_type);
1858 mutex_unlock(&uuid_mutex);
1859 if (ret)
1860 return ret;
1861
1862 if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1863 ret = -EACCES;
1864 goto error;
1865 }
1866
1867 bdev = fs_devices->latest_dev->bdev;
1868
1869 /*
1870 * From now on the error handling is not straightforward.
1871 *
1872 * If successful, this will transfer the fs_info into the super block,
1873 * and fc->s_fs_info will be NULL. However if there's an existing
1874 * super, we'll still have fc->s_fs_info populated. If we error
1875 * completely out it'll be cleaned up when we drop the fs_context,
1876 * otherwise it's tied to the lifetime of the super_block.
1877 */
1878 sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1879 if (IS_ERR(sb)) {
1880 ret = PTR_ERR(sb);
1881 goto error;
1882 }
1883
1884 set_device_specific_options(fs_info);
1885
1886 if (sb->s_root) {
1887 btrfs_close_devices(fs_devices);
1888 /*
1889 * At this stage we may have RO flag mismatch between
1890 * fc->sb_flags and sb->s_flags. Caller should detect such
1891 * mismatch and reconfigure with sb->s_umount rwsem held if
1892 * needed.
1893 */
1894 } else {
1895 snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
1896 shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id);
1897 btrfs_sb(sb)->bdev_holder = &btrfs_fs_type;
1898 ret = btrfs_fill_super(sb, fs_devices);
1899 if (ret) {
1900 deactivate_locked_super(sb);
1901 return ret;
1902 }
1903 }
1904
1905 btrfs_clear_oneshot_options(fs_info);
1906
1907 fc->root = dget(sb->s_root);
1908 return 0;
1909
1910error:
1911 btrfs_close_devices(fs_devices);
1912 return ret;
1913}
1914
1915/*
1916 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1917 * with different ro/rw options") the following works:
1918 *
1919 * (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1920 * (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1921 *
1922 * which looks nice and innocent but is actually pretty intricate and deserves
1923 * a long comment.
1924 *
1925 * On another filesystem a subvolume mount is close to something like:
1926 *
1927 * (iii) # create rw superblock + initial mount
1928 * mount -t xfs /dev/sdb /opt/
1929 *
1930 * # create ro bind mount
1931 * mount --bind -o ro /opt/foo /mnt/foo
1932 *
1933 * # unmount initial mount
1934 * umount /opt
1935 *
1936 * Of course, there's some special subvolume sauce and there's the fact that the
1937 * sb->s_root dentry is really swapped after mount_subtree(). But conceptually
1938 * it's very close and will help us understand the issue.
1939 *
1940 * The old mount API didn't cleanly distinguish between a mount being made ro
1941 * and a superblock being made ro. The only way to change the ro state of
1942 * either object was by passing ms_rdonly. If a new mount was created via
1943 * mount(2) such as:
1944 *
1945 * mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1946 *
1947 * the MS_RDONLY flag being specified had two effects:
1948 *
1949 * (1) MNT_READONLY was raised -> the resulting mount got
1950 * @mnt->mnt_flags |= MNT_READONLY raised.
1951 *
1952 * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
1953 * made the superblock ro. Note, how SB_RDONLY has the same value as
1954 * ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
1955 *
1956 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
1957 * subtree mounted ro.
1958 *
1959 * But consider the effect on the old mount API on btrfs subvolume mounting
1960 * which combines the distinct step in (iii) into a single step.
1961 *
1962 * By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
1963 * is issued the superblock is ro and thus even if the mount created for (ii) is
1964 * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
1965 * to rw for (ii) which it did using an internal remount call.
1966 *
1967 * IOW, subvolume mounting was inherently complicated due to the ambiguity of
1968 * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
1969 * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
1970 * passed by mount(8) to mount(2).
1971 *
1972 * Enter the new mount API. The new mount API disambiguates making a mount ro
1973 * and making a superblock ro.
1974 *
1975 * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
1976 * fsmount() or mount_setattr() this is a pure VFS level change for a
1977 * specific mount or mount tree that is never seen by the filesystem itself.
1978 *
1979 * (4) To turn a superblock ro the "ro" flag must be used with
1980 * fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
1981 * in fc->sb_flags.
1982 *
1983 * But, currently the util-linux mount command already utilizes the new mount
1984 * API and is still setting fsconfig(FSCONFIG_SET_FLAG, "ro") no matter if it's
1985 * btrfs or not, setting the whole super block RO. To make per-subvolume mounting
1986 * work with different options work we need to keep backward compatibility.
1987 */
1988static int btrfs_reconfigure_for_mount(struct fs_context *fc, struct vfsmount *mnt)
1989{
1990 int ret = 0;
1991
1992 if (fc->sb_flags & SB_RDONLY)
1993 return ret;
1994
1995 down_write(&mnt->mnt_sb->s_umount);
1996 if (!(fc->sb_flags & SB_RDONLY) && (mnt->mnt_sb->s_flags & SB_RDONLY))
1997 ret = btrfs_reconfigure(fc);
1998 up_write(&mnt->mnt_sb->s_umount);
1999 return ret;
2000}
2001
2002static int btrfs_get_tree_subvol(struct fs_context *fc)
2003{
2004 struct btrfs_fs_info *fs_info = NULL;
2005 struct btrfs_fs_context *ctx = fc->fs_private;
2006 struct fs_context *dup_fc;
2007 struct dentry *dentry;
2008 struct vfsmount *mnt;
2009 int ret = 0;
2010
2011 /*
2012 * Setup a dummy root and fs_info for test/set super. This is because
2013 * we don't actually fill this stuff out until open_ctree, but we need
2014 * then open_ctree will properly initialize the file system specific
2015 * settings later. btrfs_init_fs_info initializes the static elements
2016 * of the fs_info (locks and such) to make cleanup easier if we find a
2017 * superblock with our given fs_devices later on at sget() time.
2018 */
2019 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
2020 if (!fs_info)
2021 return -ENOMEM;
2022
2023 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2024 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2025 if (!fs_info->super_copy || !fs_info->super_for_commit) {
2026 btrfs_free_fs_info(fs_info);
2027 return -ENOMEM;
2028 }
2029 btrfs_init_fs_info(fs_info);
2030
2031 dup_fc = vfs_dup_fs_context(fc);
2032 if (IS_ERR(dup_fc)) {
2033 btrfs_free_fs_info(fs_info);
2034 return PTR_ERR(dup_fc);
2035 }
2036
2037 /*
2038 * When we do the sget_fc this gets transferred to the sb, so we only
2039 * need to set it on the dup_fc as this is what creates the super block.
2040 */
2041 dup_fc->s_fs_info = fs_info;
2042
2043 /*
2044 * We'll do the security settings in our btrfs_get_tree_super() mount
2045 * loop, they were duplicated into dup_fc, we can drop the originals
2046 * here.
2047 */
2048 security_free_mnt_opts(&fc->security);
2049 fc->security = NULL;
2050
2051 mnt = fc_mount(dup_fc);
2052 if (IS_ERR(mnt)) {
2053 put_fs_context(dup_fc);
2054 return PTR_ERR(mnt);
2055 }
2056 ret = btrfs_reconfigure_for_mount(dup_fc, mnt);
2057 put_fs_context(dup_fc);
2058 if (ret) {
2059 mntput(mnt);
2060 return ret;
2061 }
2062
2063 /*
2064 * This free's ->subvol_name, because if it isn't set we have to
2065 * allocate a buffer to hold the subvol_name, so we just drop our
2066 * reference to it here.
2067 */
2068 dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2069 ctx->subvol_name = NULL;
2070 if (IS_ERR(dentry))
2071 return PTR_ERR(dentry);
2072
2073 fc->root = dentry;
2074 return 0;
2075}
2076
2077static int btrfs_get_tree(struct fs_context *fc)
2078{
2079 /*
2080 * Since we use mount_subtree to mount the default/specified subvol, we
2081 * have to do mounts in two steps.
2082 *
2083 * First pass through we call btrfs_get_tree_subvol(), this is just a
2084 * wrapper around fc_mount() to call back into here again, and this time
2085 * we'll call btrfs_get_tree_super(). This will do the open_ctree() and
2086 * everything to open the devices and file system. Then we return back
2087 * with a fully constructed vfsmount in btrfs_get_tree_subvol(), and
2088 * from there we can do our mount_subvol() call, which will lookup
2089 * whichever subvol we're mounting and setup this fc with the
2090 * appropriate dentry for the subvol.
2091 */
2092 if (fc->s_fs_info)
2093 return btrfs_get_tree_super(fc);
2094 return btrfs_get_tree_subvol(fc);
2095}
2096
2097static void btrfs_kill_super(struct super_block *sb)
2098{
2099 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2100 kill_anon_super(sb);
2101 btrfs_free_fs_info(fs_info);
2102}
2103
2104static void btrfs_free_fs_context(struct fs_context *fc)
2105{
2106 struct btrfs_fs_context *ctx = fc->fs_private;
2107 struct btrfs_fs_info *fs_info = fc->s_fs_info;
2108
2109 if (fs_info)
2110 btrfs_free_fs_info(fs_info);
2111
2112 if (ctx && refcount_dec_and_test(&ctx->refs)) {
2113 kfree(ctx->subvol_name);
2114 kfree(ctx);
2115 }
2116}
2117
2118static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2119{
2120 struct btrfs_fs_context *ctx = src_fc->fs_private;
2121
2122 /*
2123 * Give a ref to our ctx to this dup, as we want to keep it around for
2124 * our original fc so we can have the subvolume name or objectid.
2125 *
2126 * We unset ->source in the original fc because the dup needs it for
2127 * mounting, and then once we free the dup it'll free ->source, so we
2128 * need to make sure we're only pointing to it in one fc.
2129 */
2130 refcount_inc(&ctx->refs);
2131 fc->fs_private = ctx;
2132 fc->source = src_fc->source;
2133 src_fc->source = NULL;
2134 return 0;
2135}
2136
2137static const struct fs_context_operations btrfs_fs_context_ops = {
2138 .parse_param = btrfs_parse_param,
2139 .reconfigure = btrfs_reconfigure,
2140 .get_tree = btrfs_get_tree,
2141 .dup = btrfs_dup_fs_context,
2142 .free = btrfs_free_fs_context,
2143};
2144
2145static int btrfs_init_fs_context(struct fs_context *fc)
2146{
2147 struct btrfs_fs_context *ctx;
2148
2149 ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL);
2150 if (!ctx)
2151 return -ENOMEM;
2152
2153 refcount_set(&ctx->refs, 1);
2154 fc->fs_private = ctx;
2155 fc->ops = &btrfs_fs_context_ops;
2156
2157 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2158 btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
2159 } else {
2160 ctx->thread_pool_size =
2161 min_t(unsigned long, num_online_cpus() + 2, 8);
2162 ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2163 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2164 }
2165
2166#ifdef CONFIG_BTRFS_FS_POSIX_ACL
2167 fc->sb_flags |= SB_POSIXACL;
2168#endif
2169 fc->sb_flags |= SB_I_VERSION;
2170
2171 return 0;
2172}
2173
2174static struct file_system_type btrfs_fs_type = {
2175 .owner = THIS_MODULE,
2176 .name = "btrfs",
2177 .init_fs_context = btrfs_init_fs_context,
2178 .parameters = btrfs_fs_parameters,
2179 .kill_sb = btrfs_kill_super,
2180 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA |
2181 FS_ALLOW_IDMAP | FS_MGTIME,
2182 };
2183
2184MODULE_ALIAS_FS("btrfs");
2185
2186static int btrfs_control_open(struct inode *inode, struct file *file)
2187{
2188 /*
2189 * The control file's private_data is used to hold the
2190 * transaction when it is started and is used to keep
2191 * track of whether a transaction is already in progress.
2192 */
2193 file->private_data = NULL;
2194 return 0;
2195}
2196
2197/*
2198 * Used by /dev/btrfs-control for devices ioctls.
2199 */
2200static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2201 unsigned long arg)
2202{
2203 struct btrfs_ioctl_vol_args *vol;
2204 struct btrfs_device *device = NULL;
2205 dev_t devt = 0;
2206 int ret = -ENOTTY;
2207
2208 if (!capable(CAP_SYS_ADMIN))
2209 return -EPERM;
2210
2211 vol = memdup_user((void __user *)arg, sizeof(*vol));
2212 if (IS_ERR(vol))
2213 return PTR_ERR(vol);
2214 ret = btrfs_check_ioctl_vol_args_path(vol);
2215 if (ret < 0)
2216 goto out;
2217
2218 switch (cmd) {
2219 case BTRFS_IOC_SCAN_DEV:
2220 mutex_lock(&uuid_mutex);
2221 /*
2222 * Scanning outside of mount can return NULL which would turn
2223 * into 0 error code.
2224 */
2225 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2226 ret = PTR_ERR_OR_ZERO(device);
2227 mutex_unlock(&uuid_mutex);
2228 break;
2229 case BTRFS_IOC_FORGET_DEV:
2230 if (vol->name[0] != 0) {
2231 ret = lookup_bdev(vol->name, &devt);
2232 if (ret)
2233 break;
2234 }
2235 ret = btrfs_forget_devices(devt);
2236 break;
2237 case BTRFS_IOC_DEVICES_READY:
2238 mutex_lock(&uuid_mutex);
2239 /*
2240 * Scanning outside of mount can return NULL which would turn
2241 * into 0 error code.
2242 */
2243 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2244 if (IS_ERR_OR_NULL(device)) {
2245 mutex_unlock(&uuid_mutex);
2246 if (IS_ERR(device))
2247 ret = PTR_ERR(device);
2248 else
2249 ret = 0;
2250 break;
2251 }
2252 ret = !(device->fs_devices->num_devices ==
2253 device->fs_devices->total_devices);
2254 mutex_unlock(&uuid_mutex);
2255 break;
2256 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2257 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2258 break;
2259 }
2260
2261out:
2262 kfree(vol);
2263 return ret;
2264}
2265
2266static int btrfs_freeze(struct super_block *sb)
2267{
2268 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2269
2270 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2271 /*
2272 * We don't need a barrier here, we'll wait for any transaction that
2273 * could be in progress on other threads (and do delayed iputs that
2274 * we want to avoid on a frozen filesystem), or do the commit
2275 * ourselves.
2276 */
2277 return btrfs_commit_current_transaction(fs_info->tree_root);
2278}
2279
2280static int check_dev_super(struct btrfs_device *dev)
2281{
2282 struct btrfs_fs_info *fs_info = dev->fs_info;
2283 struct btrfs_super_block *sb;
2284 u64 last_trans;
2285 u16 csum_type;
2286 int ret = 0;
2287
2288 /* This should be called with fs still frozen. */
2289 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2290
2291 /* Missing dev, no need to check. */
2292 if (!dev->bdev)
2293 return 0;
2294
2295 /* Only need to check the primary super block. */
2296 sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2297 if (IS_ERR(sb))
2298 return PTR_ERR(sb);
2299
2300 /* Verify the checksum. */
2301 csum_type = btrfs_super_csum_type(sb);
2302 if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2303 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2304 csum_type, btrfs_super_csum_type(fs_info->super_copy));
2305 ret = -EUCLEAN;
2306 goto out;
2307 }
2308
2309 if (btrfs_check_super_csum(fs_info, sb)) {
2310 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2311 ret = -EUCLEAN;
2312 goto out;
2313 }
2314
2315 /* Btrfs_validate_super() includes fsid check against super->fsid. */
2316 ret = btrfs_validate_super(fs_info, sb, 0);
2317 if (ret < 0)
2318 goto out;
2319
2320 last_trans = btrfs_get_last_trans_committed(fs_info);
2321 if (btrfs_super_generation(sb) != last_trans) {
2322 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2323 btrfs_super_generation(sb), last_trans);
2324 ret = -EUCLEAN;
2325 goto out;
2326 }
2327out:
2328 btrfs_release_disk_super(sb);
2329 return ret;
2330}
2331
2332static int btrfs_unfreeze(struct super_block *sb)
2333{
2334 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2335 struct btrfs_device *device;
2336 int ret = 0;
2337
2338 /*
2339 * Make sure the fs is not changed by accident (like hibernation then
2340 * modified by other OS).
2341 * If we found anything wrong, we mark the fs error immediately.
2342 *
2343 * And since the fs is frozen, no one can modify the fs yet, thus
2344 * we don't need to hold device_list_mutex.
2345 */
2346 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2347 ret = check_dev_super(device);
2348 if (ret < 0) {
2349 btrfs_handle_fs_error(fs_info, ret,
2350 "super block on devid %llu got modified unexpectedly",
2351 device->devid);
2352 break;
2353 }
2354 }
2355 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2356
2357 /*
2358 * We still return 0, to allow VFS layer to unfreeze the fs even the
2359 * above checks failed. Since the fs is either fine or read-only, we're
2360 * safe to continue, without causing further damage.
2361 */
2362 return 0;
2363}
2364
2365static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2366{
2367 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2368
2369 /*
2370 * There should be always a valid pointer in latest_dev, it may be stale
2371 * for a short moment in case it's being deleted but still valid until
2372 * the end of RCU grace period.
2373 */
2374 rcu_read_lock();
2375 seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2376 rcu_read_unlock();
2377
2378 return 0;
2379}
2380
2381static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_control *sc)
2382{
2383 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2384 const s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps);
2385
2386 trace_btrfs_extent_map_shrinker_count(fs_info, nr);
2387
2388 return nr;
2389}
2390
2391static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_control *sc)
2392{
2393 const long nr_to_scan = min_t(unsigned long, LONG_MAX, sc->nr_to_scan);
2394 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2395
2396 btrfs_free_extent_maps(fs_info, nr_to_scan);
2397
2398 /* The extent map shrinker runs asynchronously, so always return 0. */
2399 return 0;
2400}
2401
2402static const struct super_operations btrfs_super_ops = {
2403 .drop_inode = btrfs_drop_inode,
2404 .evict_inode = btrfs_evict_inode,
2405 .put_super = btrfs_put_super,
2406 .sync_fs = btrfs_sync_fs,
2407 .show_options = btrfs_show_options,
2408 .show_devname = btrfs_show_devname,
2409 .alloc_inode = btrfs_alloc_inode,
2410 .destroy_inode = btrfs_destroy_inode,
2411 .free_inode = btrfs_free_inode,
2412 .statfs = btrfs_statfs,
2413 .freeze_fs = btrfs_freeze,
2414 .unfreeze_fs = btrfs_unfreeze,
2415 .nr_cached_objects = btrfs_nr_cached_objects,
2416 .free_cached_objects = btrfs_free_cached_objects,
2417};
2418
2419static const struct file_operations btrfs_ctl_fops = {
2420 .open = btrfs_control_open,
2421 .unlocked_ioctl = btrfs_control_ioctl,
2422 .compat_ioctl = compat_ptr_ioctl,
2423 .owner = THIS_MODULE,
2424 .llseek = noop_llseek,
2425};
2426
2427static struct miscdevice btrfs_misc = {
2428 .minor = BTRFS_MINOR,
2429 .name = "btrfs-control",
2430 .fops = &btrfs_ctl_fops
2431};
2432
2433MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2434MODULE_ALIAS("devname:btrfs-control");
2435
2436static int __init btrfs_interface_init(void)
2437{
2438 return misc_register(&btrfs_misc);
2439}
2440
2441static __cold void btrfs_interface_exit(void)
2442{
2443 misc_deregister(&btrfs_misc);
2444}
2445
2446static int __init btrfs_print_mod_info(void)
2447{
2448 static const char options[] = ""
2449#ifdef CONFIG_BTRFS_DEBUG
2450 ", debug=on"
2451#endif
2452#ifdef CONFIG_BTRFS_ASSERT
2453 ", assert=on"
2454#endif
2455#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2456 ", ref-verify=on"
2457#endif
2458#ifdef CONFIG_BLK_DEV_ZONED
2459 ", zoned=yes"
2460#else
2461 ", zoned=no"
2462#endif
2463#ifdef CONFIG_FS_VERITY
2464 ", fsverity=yes"
2465#else
2466 ", fsverity=no"
2467#endif
2468 ;
2469 pr_info("Btrfs loaded%s\n", options);
2470 return 0;
2471}
2472
2473static int register_btrfs(void)
2474{
2475 return register_filesystem(&btrfs_fs_type);
2476}
2477
2478static void unregister_btrfs(void)
2479{
2480 unregister_filesystem(&btrfs_fs_type);
2481}
2482
2483/* Helper structure for long init/exit functions. */
2484struct init_sequence {
2485 int (*init_func)(void);
2486 /* Can be NULL if the init_func doesn't need cleanup. */
2487 void (*exit_func)(void);
2488};
2489
2490static const struct init_sequence mod_init_seq[] = {
2491 {
2492 .init_func = btrfs_props_init,
2493 .exit_func = NULL,
2494 }, {
2495 .init_func = btrfs_init_sysfs,
2496 .exit_func = btrfs_exit_sysfs,
2497 }, {
2498 .init_func = btrfs_init_compress,
2499 .exit_func = btrfs_exit_compress,
2500 }, {
2501 .init_func = btrfs_init_cachep,
2502 .exit_func = btrfs_destroy_cachep,
2503 }, {
2504 .init_func = btrfs_init_dio,
2505 .exit_func = btrfs_destroy_dio,
2506 }, {
2507 .init_func = btrfs_transaction_init,
2508 .exit_func = btrfs_transaction_exit,
2509 }, {
2510 .init_func = btrfs_ctree_init,
2511 .exit_func = btrfs_ctree_exit,
2512 }, {
2513 .init_func = btrfs_free_space_init,
2514 .exit_func = btrfs_free_space_exit,
2515 }, {
2516 .init_func = extent_state_init_cachep,
2517 .exit_func = extent_state_free_cachep,
2518 }, {
2519 .init_func = extent_buffer_init_cachep,
2520 .exit_func = extent_buffer_free_cachep,
2521 }, {
2522 .init_func = btrfs_bioset_init,
2523 .exit_func = btrfs_bioset_exit,
2524 }, {
2525 .init_func = extent_map_init,
2526 .exit_func = extent_map_exit,
2527 }, {
2528 .init_func = ordered_data_init,
2529 .exit_func = ordered_data_exit,
2530 }, {
2531 .init_func = btrfs_delayed_inode_init,
2532 .exit_func = btrfs_delayed_inode_exit,
2533 }, {
2534 .init_func = btrfs_auto_defrag_init,
2535 .exit_func = btrfs_auto_defrag_exit,
2536 }, {
2537 .init_func = btrfs_delayed_ref_init,
2538 .exit_func = btrfs_delayed_ref_exit,
2539 }, {
2540 .init_func = btrfs_prelim_ref_init,
2541 .exit_func = btrfs_prelim_ref_exit,
2542 }, {
2543 .init_func = btrfs_interface_init,
2544 .exit_func = btrfs_interface_exit,
2545 }, {
2546 .init_func = btrfs_print_mod_info,
2547 .exit_func = NULL,
2548 }, {
2549 .init_func = btrfs_run_sanity_tests,
2550 .exit_func = NULL,
2551 }, {
2552 .init_func = register_btrfs,
2553 .exit_func = unregister_btrfs,
2554 }
2555};
2556
2557static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2558
2559static __always_inline void btrfs_exit_btrfs_fs(void)
2560{
2561 int i;
2562
2563 for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2564 if (!mod_init_result[i])
2565 continue;
2566 if (mod_init_seq[i].exit_func)
2567 mod_init_seq[i].exit_func();
2568 mod_init_result[i] = false;
2569 }
2570}
2571
2572static void __exit exit_btrfs_fs(void)
2573{
2574 btrfs_exit_btrfs_fs();
2575 btrfs_cleanup_fs_uuids();
2576}
2577
2578static int __init init_btrfs_fs(void)
2579{
2580 int ret;
2581 int i;
2582
2583 for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2584 ASSERT(!mod_init_result[i]);
2585 ret = mod_init_seq[i].init_func();
2586 if (ret < 0) {
2587 btrfs_exit_btrfs_fs();
2588 return ret;
2589 }
2590 mod_init_result[i] = true;
2591 }
2592 return 0;
2593}
2594
2595late_initcall(init_btrfs_fs);
2596module_exit(exit_btrfs_fs)
2597
2598MODULE_DESCRIPTION("B-Tree File System (BTRFS)");
2599MODULE_LICENSE("GPL");
2600MODULE_SOFTDEP("pre: crc32c");
2601MODULE_SOFTDEP("pre: xxhash64");
2602MODULE_SOFTDEP("pre: sha256");
2603MODULE_SOFTDEP("pre: blake2b-256");