]> git.ipfire.org Git - thirdparty/kernel/stable.git/blob - fs/btrfs/super.c
KVM: x86/mmu: Remove unnecessary ‘NULL’ values from sptep
[thirdparty/kernel/stable.git] / fs / btrfs / super.c
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 #include "extent-tree.h"
62 #define CREATE_TRACE_POINTS
63 #include <trace/events/btrfs.h>
64
65 static const struct super_operations btrfs_super_ops;
66
67 /*
68 * Types for mounting the default subvolume and a subvolume explicitly
69 * requested by subvol=/path. That way the callchain is straightforward and we
70 * don't have to play tricks with the mount options and recursive calls to
71 * btrfs_mount.
72 *
73 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
74 */
75 static struct file_system_type btrfs_fs_type;
76 static struct file_system_type btrfs_root_fs_type;
77
78 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
79
80 static void btrfs_put_super(struct super_block *sb)
81 {
82 close_ctree(btrfs_sb(sb));
83 }
84
85 enum {
86 Opt_acl, Opt_noacl,
87 Opt_clear_cache,
88 Opt_commit_interval,
89 Opt_compress,
90 Opt_compress_force,
91 Opt_compress_force_type,
92 Opt_compress_type,
93 Opt_degraded,
94 Opt_device,
95 Opt_fatal_errors,
96 Opt_flushoncommit, Opt_noflushoncommit,
97 Opt_max_inline,
98 Opt_barrier, Opt_nobarrier,
99 Opt_datacow, Opt_nodatacow,
100 Opt_datasum, Opt_nodatasum,
101 Opt_defrag, Opt_nodefrag,
102 Opt_discard, Opt_nodiscard,
103 Opt_discard_mode,
104 Opt_norecovery,
105 Opt_ratio,
106 Opt_rescan_uuid_tree,
107 Opt_skip_balance,
108 Opt_space_cache, Opt_no_space_cache,
109 Opt_space_cache_version,
110 Opt_ssd, Opt_nossd,
111 Opt_ssd_spread, Opt_nossd_spread,
112 Opt_subvol,
113 Opt_subvol_empty,
114 Opt_subvolid,
115 Opt_thread_pool,
116 Opt_treelog, Opt_notreelog,
117 Opt_user_subvol_rm_allowed,
118
119 /* Rescue options */
120 Opt_rescue,
121 Opt_usebackuproot,
122 Opt_nologreplay,
123 Opt_ignorebadroots,
124 Opt_ignoredatacsums,
125 Opt_rescue_all,
126
127 /* Deprecated options */
128 Opt_recovery,
129 Opt_inode_cache, Opt_noinode_cache,
130
131 /* Debugging options */
132 Opt_check_integrity,
133 Opt_check_integrity_including_extent_data,
134 Opt_check_integrity_print_mask,
135 Opt_enospc_debug, Opt_noenospc_debug,
136 #ifdef CONFIG_BTRFS_DEBUG
137 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
138 #endif
139 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
140 Opt_ref_verify,
141 #endif
142 Opt_err,
143 };
144
145 static const match_table_t tokens = {
146 {Opt_acl, "acl"},
147 {Opt_noacl, "noacl"},
148 {Opt_clear_cache, "clear_cache"},
149 {Opt_commit_interval, "commit=%u"},
150 {Opt_compress, "compress"},
151 {Opt_compress_type, "compress=%s"},
152 {Opt_compress_force, "compress-force"},
153 {Opt_compress_force_type, "compress-force=%s"},
154 {Opt_degraded, "degraded"},
155 {Opt_device, "device=%s"},
156 {Opt_fatal_errors, "fatal_errors=%s"},
157 {Opt_flushoncommit, "flushoncommit"},
158 {Opt_noflushoncommit, "noflushoncommit"},
159 {Opt_inode_cache, "inode_cache"},
160 {Opt_noinode_cache, "noinode_cache"},
161 {Opt_max_inline, "max_inline=%s"},
162 {Opt_barrier, "barrier"},
163 {Opt_nobarrier, "nobarrier"},
164 {Opt_datacow, "datacow"},
165 {Opt_nodatacow, "nodatacow"},
166 {Opt_datasum, "datasum"},
167 {Opt_nodatasum, "nodatasum"},
168 {Opt_defrag, "autodefrag"},
169 {Opt_nodefrag, "noautodefrag"},
170 {Opt_discard, "discard"},
171 {Opt_discard_mode, "discard=%s"},
172 {Opt_nodiscard, "nodiscard"},
173 {Opt_norecovery, "norecovery"},
174 {Opt_ratio, "metadata_ratio=%u"},
175 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
176 {Opt_skip_balance, "skip_balance"},
177 {Opt_space_cache, "space_cache"},
178 {Opt_no_space_cache, "nospace_cache"},
179 {Opt_space_cache_version, "space_cache=%s"},
180 {Opt_ssd, "ssd"},
181 {Opt_nossd, "nossd"},
182 {Opt_ssd_spread, "ssd_spread"},
183 {Opt_nossd_spread, "nossd_spread"},
184 {Opt_subvol, "subvol=%s"},
185 {Opt_subvol_empty, "subvol="},
186 {Opt_subvolid, "subvolid=%s"},
187 {Opt_thread_pool, "thread_pool=%u"},
188 {Opt_treelog, "treelog"},
189 {Opt_notreelog, "notreelog"},
190 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
191
192 /* Rescue options */
193 {Opt_rescue, "rescue=%s"},
194 /* Deprecated, with alias rescue=nologreplay */
195 {Opt_nologreplay, "nologreplay"},
196 /* Deprecated, with alias rescue=usebackuproot */
197 {Opt_usebackuproot, "usebackuproot"},
198
199 /* Deprecated options */
200 {Opt_recovery, "recovery"},
201
202 /* Debugging options */
203 {Opt_check_integrity, "check_int"},
204 {Opt_check_integrity_including_extent_data, "check_int_data"},
205 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
206 {Opt_enospc_debug, "enospc_debug"},
207 {Opt_noenospc_debug, "noenospc_debug"},
208 #ifdef CONFIG_BTRFS_DEBUG
209 {Opt_fragment_data, "fragment=data"},
210 {Opt_fragment_metadata, "fragment=metadata"},
211 {Opt_fragment_all, "fragment=all"},
212 #endif
213 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
214 {Opt_ref_verify, "ref_verify"},
215 #endif
216 {Opt_err, NULL},
217 };
218
219 static const match_table_t rescue_tokens = {
220 {Opt_usebackuproot, "usebackuproot"},
221 {Opt_nologreplay, "nologreplay"},
222 {Opt_ignorebadroots, "ignorebadroots"},
223 {Opt_ignorebadroots, "ibadroots"},
224 {Opt_ignoredatacsums, "ignoredatacsums"},
225 {Opt_ignoredatacsums, "idatacsums"},
226 {Opt_rescue_all, "all"},
227 {Opt_err, NULL},
228 };
229
230 static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
231 const char *opt_name)
232 {
233 if (fs_info->mount_opt & opt) {
234 btrfs_err(fs_info, "%s must be used with ro mount option",
235 opt_name);
236 return true;
237 }
238 return false;
239 }
240
241 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
242 {
243 char *opts;
244 char *orig;
245 char *p;
246 substring_t args[MAX_OPT_ARGS];
247 int ret = 0;
248
249 opts = kstrdup(options, GFP_KERNEL);
250 if (!opts)
251 return -ENOMEM;
252 orig = opts;
253
254 while ((p = strsep(&opts, ":")) != NULL) {
255 int token;
256
257 if (!*p)
258 continue;
259 token = match_token(p, rescue_tokens, args);
260 switch (token){
261 case Opt_usebackuproot:
262 btrfs_info(info,
263 "trying to use backup root at mount time");
264 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
265 break;
266 case Opt_nologreplay:
267 btrfs_set_and_info(info, NOLOGREPLAY,
268 "disabling log replay at mount time");
269 break;
270 case Opt_ignorebadroots:
271 btrfs_set_and_info(info, IGNOREBADROOTS,
272 "ignoring bad roots");
273 break;
274 case Opt_ignoredatacsums:
275 btrfs_set_and_info(info, IGNOREDATACSUMS,
276 "ignoring data csums");
277 break;
278 case Opt_rescue_all:
279 btrfs_info(info, "enabling all of the rescue options");
280 btrfs_set_and_info(info, IGNOREDATACSUMS,
281 "ignoring data csums");
282 btrfs_set_and_info(info, IGNOREBADROOTS,
283 "ignoring bad roots");
284 btrfs_set_and_info(info, NOLOGREPLAY,
285 "disabling log replay at mount time");
286 break;
287 case Opt_err:
288 btrfs_info(info, "unrecognized rescue option '%s'", p);
289 ret = -EINVAL;
290 goto out;
291 default:
292 break;
293 }
294
295 }
296 out:
297 kfree(orig);
298 return ret;
299 }
300
301 /*
302 * Regular mount options parser. Everything that is needed only when
303 * reading in a new superblock is parsed here.
304 * XXX JDM: This needs to be cleaned up for remount.
305 */
306 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
307 unsigned long new_flags)
308 {
309 substring_t args[MAX_OPT_ARGS];
310 char *p, *num;
311 int intarg;
312 int ret = 0;
313 char *compress_type;
314 bool compress_force = false;
315 enum btrfs_compression_type saved_compress_type;
316 int saved_compress_level;
317 bool saved_compress_force;
318 int no_compress = 0;
319 const bool remounting = test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state);
320
321 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
322 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
323 else if (btrfs_free_space_cache_v1_active(info)) {
324 if (btrfs_is_zoned(info)) {
325 btrfs_info(info,
326 "zoned: clearing existing space cache");
327 btrfs_set_super_cache_generation(info->super_copy, 0);
328 } else {
329 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
330 }
331 }
332
333 /*
334 * Even the options are empty, we still need to do extra check
335 * against new flags
336 */
337 if (!options)
338 goto check;
339
340 while ((p = strsep(&options, ",")) != NULL) {
341 int token;
342 if (!*p)
343 continue;
344
345 token = match_token(p, tokens, args);
346 switch (token) {
347 case Opt_degraded:
348 btrfs_info(info, "allowing degraded mounts");
349 btrfs_set_opt(info->mount_opt, DEGRADED);
350 break;
351 case Opt_subvol:
352 case Opt_subvol_empty:
353 case Opt_subvolid:
354 case Opt_device:
355 /*
356 * These are parsed by btrfs_parse_subvol_options or
357 * btrfs_parse_device_options and can be ignored here.
358 */
359 break;
360 case Opt_nodatasum:
361 btrfs_set_and_info(info, NODATASUM,
362 "setting nodatasum");
363 break;
364 case Opt_datasum:
365 if (btrfs_test_opt(info, NODATASUM)) {
366 if (btrfs_test_opt(info, NODATACOW))
367 btrfs_info(info,
368 "setting datasum, datacow enabled");
369 else
370 btrfs_info(info, "setting datasum");
371 }
372 btrfs_clear_opt(info->mount_opt, NODATACOW);
373 btrfs_clear_opt(info->mount_opt, NODATASUM);
374 break;
375 case Opt_nodatacow:
376 if (!btrfs_test_opt(info, NODATACOW)) {
377 if (!btrfs_test_opt(info, COMPRESS) ||
378 !btrfs_test_opt(info, FORCE_COMPRESS)) {
379 btrfs_info(info,
380 "setting nodatacow, compression disabled");
381 } else {
382 btrfs_info(info, "setting nodatacow");
383 }
384 }
385 btrfs_clear_opt(info->mount_opt, COMPRESS);
386 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
387 btrfs_set_opt(info->mount_opt, NODATACOW);
388 btrfs_set_opt(info->mount_opt, NODATASUM);
389 break;
390 case Opt_datacow:
391 btrfs_clear_and_info(info, NODATACOW,
392 "setting datacow");
393 break;
394 case Opt_compress_force:
395 case Opt_compress_force_type:
396 compress_force = true;
397 fallthrough;
398 case Opt_compress:
399 case Opt_compress_type:
400 saved_compress_type = btrfs_test_opt(info,
401 COMPRESS) ?
402 info->compress_type : BTRFS_COMPRESS_NONE;
403 saved_compress_force =
404 btrfs_test_opt(info, FORCE_COMPRESS);
405 saved_compress_level = info->compress_level;
406 if (token == Opt_compress ||
407 token == Opt_compress_force ||
408 strncmp(args[0].from, "zlib", 4) == 0) {
409 compress_type = "zlib";
410
411 info->compress_type = BTRFS_COMPRESS_ZLIB;
412 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
413 /*
414 * args[0] contains uninitialized data since
415 * for these tokens we don't expect any
416 * parameter.
417 */
418 if (token != Opt_compress &&
419 token != Opt_compress_force)
420 info->compress_level =
421 btrfs_compress_str2level(
422 BTRFS_COMPRESS_ZLIB,
423 args[0].from + 4);
424 btrfs_set_opt(info->mount_opt, COMPRESS);
425 btrfs_clear_opt(info->mount_opt, NODATACOW);
426 btrfs_clear_opt(info->mount_opt, NODATASUM);
427 no_compress = 0;
428 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
429 compress_type = "lzo";
430 info->compress_type = BTRFS_COMPRESS_LZO;
431 info->compress_level = 0;
432 btrfs_set_opt(info->mount_opt, COMPRESS);
433 btrfs_clear_opt(info->mount_opt, NODATACOW);
434 btrfs_clear_opt(info->mount_opt, NODATASUM);
435 btrfs_set_fs_incompat(info, COMPRESS_LZO);
436 no_compress = 0;
437 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
438 compress_type = "zstd";
439 info->compress_type = BTRFS_COMPRESS_ZSTD;
440 info->compress_level =
441 btrfs_compress_str2level(
442 BTRFS_COMPRESS_ZSTD,
443 args[0].from + 4);
444 btrfs_set_opt(info->mount_opt, COMPRESS);
445 btrfs_clear_opt(info->mount_opt, NODATACOW);
446 btrfs_clear_opt(info->mount_opt, NODATASUM);
447 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
448 no_compress = 0;
449 } else if (strncmp(args[0].from, "no", 2) == 0) {
450 compress_type = "no";
451 info->compress_level = 0;
452 info->compress_type = 0;
453 btrfs_clear_opt(info->mount_opt, COMPRESS);
454 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
455 compress_force = false;
456 no_compress++;
457 } else {
458 btrfs_err(info, "unrecognized compression value %s",
459 args[0].from);
460 ret = -EINVAL;
461 goto out;
462 }
463
464 if (compress_force) {
465 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
466 } else {
467 /*
468 * If we remount from compress-force=xxx to
469 * compress=xxx, we need clear FORCE_COMPRESS
470 * flag, otherwise, there is no way for users
471 * to disable forcible compression separately.
472 */
473 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
474 }
475 if (no_compress == 1) {
476 btrfs_info(info, "use no compression");
477 } else if ((info->compress_type != saved_compress_type) ||
478 (compress_force != saved_compress_force) ||
479 (info->compress_level != saved_compress_level)) {
480 btrfs_info(info, "%s %s compression, level %d",
481 (compress_force) ? "force" : "use",
482 compress_type, info->compress_level);
483 }
484 compress_force = false;
485 break;
486 case Opt_ssd:
487 btrfs_set_and_info(info, SSD,
488 "enabling ssd optimizations");
489 btrfs_clear_opt(info->mount_opt, NOSSD);
490 break;
491 case Opt_ssd_spread:
492 btrfs_set_and_info(info, SSD,
493 "enabling ssd optimizations");
494 btrfs_set_and_info(info, SSD_SPREAD,
495 "using spread ssd allocation scheme");
496 btrfs_clear_opt(info->mount_opt, NOSSD);
497 break;
498 case Opt_nossd:
499 btrfs_set_opt(info->mount_opt, NOSSD);
500 btrfs_clear_and_info(info, SSD,
501 "not using ssd optimizations");
502 fallthrough;
503 case Opt_nossd_spread:
504 btrfs_clear_and_info(info, SSD_SPREAD,
505 "not using spread ssd allocation scheme");
506 break;
507 case Opt_barrier:
508 btrfs_clear_and_info(info, NOBARRIER,
509 "turning on barriers");
510 break;
511 case Opt_nobarrier:
512 btrfs_set_and_info(info, NOBARRIER,
513 "turning off barriers");
514 break;
515 case Opt_thread_pool:
516 ret = match_int(&args[0], &intarg);
517 if (ret) {
518 btrfs_err(info, "unrecognized thread_pool value %s",
519 args[0].from);
520 goto out;
521 } else if (intarg == 0) {
522 btrfs_err(info, "invalid value 0 for thread_pool");
523 ret = -EINVAL;
524 goto out;
525 }
526 info->thread_pool_size = intarg;
527 break;
528 case Opt_max_inline:
529 num = match_strdup(&args[0]);
530 if (num) {
531 info->max_inline = memparse(num, NULL);
532 kfree(num);
533
534 if (info->max_inline) {
535 info->max_inline = min_t(u64,
536 info->max_inline,
537 info->sectorsize);
538 }
539 btrfs_info(info, "max_inline at %llu",
540 info->max_inline);
541 } else {
542 ret = -ENOMEM;
543 goto out;
544 }
545 break;
546 case Opt_acl:
547 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
548 info->sb->s_flags |= SB_POSIXACL;
549 break;
550 #else
551 btrfs_err(info, "support for ACL not compiled in!");
552 ret = -EINVAL;
553 goto out;
554 #endif
555 case Opt_noacl:
556 info->sb->s_flags &= ~SB_POSIXACL;
557 break;
558 case Opt_notreelog:
559 btrfs_set_and_info(info, NOTREELOG,
560 "disabling tree log");
561 break;
562 case Opt_treelog:
563 btrfs_clear_and_info(info, NOTREELOG,
564 "enabling tree log");
565 break;
566 case Opt_norecovery:
567 case Opt_nologreplay:
568 btrfs_warn(info,
569 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
570 btrfs_set_and_info(info, NOLOGREPLAY,
571 "disabling log replay at mount time");
572 break;
573 case Opt_flushoncommit:
574 btrfs_set_and_info(info, FLUSHONCOMMIT,
575 "turning on flush-on-commit");
576 break;
577 case Opt_noflushoncommit:
578 btrfs_clear_and_info(info, FLUSHONCOMMIT,
579 "turning off flush-on-commit");
580 break;
581 case Opt_ratio:
582 ret = match_int(&args[0], &intarg);
583 if (ret) {
584 btrfs_err(info, "unrecognized metadata_ratio value %s",
585 args[0].from);
586 goto out;
587 }
588 info->metadata_ratio = intarg;
589 btrfs_info(info, "metadata ratio %u",
590 info->metadata_ratio);
591 break;
592 case Opt_discard:
593 case Opt_discard_mode:
594 if (token == Opt_discard ||
595 strcmp(args[0].from, "sync") == 0) {
596 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
597 btrfs_set_and_info(info, DISCARD_SYNC,
598 "turning on sync discard");
599 } else if (strcmp(args[0].from, "async") == 0) {
600 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
601 btrfs_set_and_info(info, DISCARD_ASYNC,
602 "turning on async discard");
603 } else {
604 btrfs_err(info, "unrecognized discard mode value %s",
605 args[0].from);
606 ret = -EINVAL;
607 goto out;
608 }
609 btrfs_clear_opt(info->mount_opt, NODISCARD);
610 break;
611 case Opt_nodiscard:
612 btrfs_clear_and_info(info, DISCARD_SYNC,
613 "turning off discard");
614 btrfs_clear_and_info(info, DISCARD_ASYNC,
615 "turning off async discard");
616 btrfs_set_opt(info->mount_opt, NODISCARD);
617 break;
618 case Opt_space_cache:
619 case Opt_space_cache_version:
620 /*
621 * We already set FREE_SPACE_TREE above because we have
622 * compat_ro(FREE_SPACE_TREE) set, and we aren't going
623 * to allow v1 to be set for extent tree v2, simply
624 * ignore this setting if we're extent tree v2.
625 */
626 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
627 break;
628 if (token == Opt_space_cache ||
629 strcmp(args[0].from, "v1") == 0) {
630 btrfs_clear_opt(info->mount_opt,
631 FREE_SPACE_TREE);
632 btrfs_set_and_info(info, SPACE_CACHE,
633 "enabling disk space caching");
634 } else if (strcmp(args[0].from, "v2") == 0) {
635 btrfs_clear_opt(info->mount_opt,
636 SPACE_CACHE);
637 btrfs_set_and_info(info, FREE_SPACE_TREE,
638 "enabling free space tree");
639 } else {
640 btrfs_err(info, "unrecognized space_cache value %s",
641 args[0].from);
642 ret = -EINVAL;
643 goto out;
644 }
645 break;
646 case Opt_rescan_uuid_tree:
647 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
648 break;
649 case Opt_no_space_cache:
650 /*
651 * We cannot operate without the free space tree with
652 * extent tree v2, ignore this option.
653 */
654 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
655 break;
656 if (btrfs_test_opt(info, SPACE_CACHE)) {
657 btrfs_clear_and_info(info, SPACE_CACHE,
658 "disabling disk space caching");
659 }
660 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
661 btrfs_clear_and_info(info, FREE_SPACE_TREE,
662 "disabling free space tree");
663 }
664 break;
665 case Opt_inode_cache:
666 case Opt_noinode_cache:
667 btrfs_warn(info,
668 "the 'inode_cache' option is deprecated and has no effect since 5.11");
669 break;
670 case Opt_clear_cache:
671 /*
672 * We cannot clear the free space tree with extent tree
673 * v2, ignore this option.
674 */
675 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
676 break;
677 btrfs_set_and_info(info, CLEAR_CACHE,
678 "force clearing of disk cache");
679 break;
680 case Opt_user_subvol_rm_allowed:
681 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
682 break;
683 case Opt_enospc_debug:
684 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
685 break;
686 case Opt_noenospc_debug:
687 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
688 break;
689 case Opt_defrag:
690 btrfs_set_and_info(info, AUTO_DEFRAG,
691 "enabling auto defrag");
692 break;
693 case Opt_nodefrag:
694 btrfs_clear_and_info(info, AUTO_DEFRAG,
695 "disabling auto defrag");
696 break;
697 case Opt_recovery:
698 case Opt_usebackuproot:
699 btrfs_warn(info,
700 "'%s' is deprecated, use 'rescue=usebackuproot' instead",
701 token == Opt_recovery ? "recovery" :
702 "usebackuproot");
703 btrfs_info(info,
704 "trying to use backup root at mount time");
705 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
706 break;
707 case Opt_skip_balance:
708 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
709 break;
710 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
711 case Opt_check_integrity_including_extent_data:
712 btrfs_warn(info,
713 "integrity checker is deprecated and will be removed in 6.7");
714 btrfs_info(info,
715 "enabling check integrity including extent data");
716 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
717 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
718 break;
719 case Opt_check_integrity:
720 btrfs_warn(info,
721 "integrity checker is deprecated and will be removed in 6.7");
722 btrfs_info(info, "enabling check integrity");
723 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
724 break;
725 case Opt_check_integrity_print_mask:
726 ret = match_int(&args[0], &intarg);
727 if (ret) {
728 btrfs_err(info,
729 "unrecognized check_integrity_print_mask value %s",
730 args[0].from);
731 goto out;
732 }
733 info->check_integrity_print_mask = intarg;
734 btrfs_warn(info,
735 "integrity checker is deprecated and will be removed in 6.7");
736 btrfs_info(info, "check_integrity_print_mask 0x%x",
737 info->check_integrity_print_mask);
738 break;
739 #else
740 case Opt_check_integrity_including_extent_data:
741 case Opt_check_integrity:
742 case Opt_check_integrity_print_mask:
743 btrfs_err(info,
744 "support for check_integrity* not compiled in!");
745 ret = -EINVAL;
746 goto out;
747 #endif
748 case Opt_fatal_errors:
749 if (strcmp(args[0].from, "panic") == 0) {
750 btrfs_set_opt(info->mount_opt,
751 PANIC_ON_FATAL_ERROR);
752 } else if (strcmp(args[0].from, "bug") == 0) {
753 btrfs_clear_opt(info->mount_opt,
754 PANIC_ON_FATAL_ERROR);
755 } else {
756 btrfs_err(info, "unrecognized fatal_errors value %s",
757 args[0].from);
758 ret = -EINVAL;
759 goto out;
760 }
761 break;
762 case Opt_commit_interval:
763 intarg = 0;
764 ret = match_int(&args[0], &intarg);
765 if (ret) {
766 btrfs_err(info, "unrecognized commit_interval value %s",
767 args[0].from);
768 ret = -EINVAL;
769 goto out;
770 }
771 if (intarg == 0) {
772 btrfs_info(info,
773 "using default commit interval %us",
774 BTRFS_DEFAULT_COMMIT_INTERVAL);
775 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
776 } else if (intarg > 300) {
777 btrfs_warn(info, "excessive commit interval %d",
778 intarg);
779 }
780 info->commit_interval = intarg;
781 break;
782 case Opt_rescue:
783 ret = parse_rescue_options(info, args[0].from);
784 if (ret < 0) {
785 btrfs_err(info, "unrecognized rescue value %s",
786 args[0].from);
787 goto out;
788 }
789 break;
790 #ifdef CONFIG_BTRFS_DEBUG
791 case Opt_fragment_all:
792 btrfs_info(info, "fragmenting all space");
793 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
794 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
795 break;
796 case Opt_fragment_metadata:
797 btrfs_info(info, "fragmenting metadata");
798 btrfs_set_opt(info->mount_opt,
799 FRAGMENT_METADATA);
800 break;
801 case Opt_fragment_data:
802 btrfs_info(info, "fragmenting data");
803 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
804 break;
805 #endif
806 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
807 case Opt_ref_verify:
808 btrfs_info(info, "doing ref verification");
809 btrfs_set_opt(info->mount_opt, REF_VERIFY);
810 break;
811 #endif
812 case Opt_err:
813 btrfs_err(info, "unrecognized mount option '%s'", p);
814 ret = -EINVAL;
815 goto out;
816 default:
817 break;
818 }
819 }
820 check:
821 /* We're read-only, don't have to check. */
822 if (new_flags & SB_RDONLY)
823 goto out;
824
825 if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
826 check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
827 check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
828 ret = -EINVAL;
829 out:
830 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
831 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
832 !btrfs_test_opt(info, CLEAR_CACHE)) {
833 btrfs_err(info, "cannot disable free space tree");
834 ret = -EINVAL;
835 }
836 if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
837 !btrfs_test_opt(info, FREE_SPACE_TREE)) {
838 btrfs_err(info, "cannot disable free space tree with block-group-tree feature");
839 ret = -EINVAL;
840 }
841 if (!ret)
842 ret = btrfs_check_mountopts_zoned(info);
843 if (!ret && !remounting) {
844 if (btrfs_test_opt(info, SPACE_CACHE))
845 btrfs_info(info, "disk space caching is enabled");
846 if (btrfs_test_opt(info, FREE_SPACE_TREE))
847 btrfs_info(info, "using free space tree");
848 }
849 return ret;
850 }
851
852 /*
853 * Parse mount options that are required early in the mount process.
854 *
855 * All other options will be parsed on much later in the mount process and
856 * only when we need to allocate a new super block.
857 */
858 static int btrfs_parse_device_options(const char *options, blk_mode_t flags)
859 {
860 substring_t args[MAX_OPT_ARGS];
861 char *device_name, *opts, *orig, *p;
862 struct btrfs_device *device = NULL;
863 int error = 0;
864
865 lockdep_assert_held(&uuid_mutex);
866
867 if (!options)
868 return 0;
869
870 /*
871 * strsep changes the string, duplicate it because btrfs_parse_options
872 * gets called later
873 */
874 opts = kstrdup(options, GFP_KERNEL);
875 if (!opts)
876 return -ENOMEM;
877 orig = opts;
878
879 while ((p = strsep(&opts, ",")) != NULL) {
880 int token;
881
882 if (!*p)
883 continue;
884
885 token = match_token(p, tokens, args);
886 if (token == Opt_device) {
887 device_name = match_strdup(&args[0]);
888 if (!device_name) {
889 error = -ENOMEM;
890 goto out;
891 }
892 device = btrfs_scan_one_device(device_name, flags);
893 kfree(device_name);
894 if (IS_ERR(device)) {
895 error = PTR_ERR(device);
896 goto out;
897 }
898 }
899 }
900
901 out:
902 kfree(orig);
903 return error;
904 }
905
906 /*
907 * Parse mount options that are related to subvolume id
908 *
909 * The value is later passed to mount_subvol()
910 */
911 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
912 u64 *subvol_objectid)
913 {
914 substring_t args[MAX_OPT_ARGS];
915 char *opts, *orig, *p;
916 int error = 0;
917 u64 subvolid;
918
919 if (!options)
920 return 0;
921
922 /*
923 * strsep changes the string, duplicate it because
924 * btrfs_parse_device_options gets called later
925 */
926 opts = kstrdup(options, GFP_KERNEL);
927 if (!opts)
928 return -ENOMEM;
929 orig = opts;
930
931 while ((p = strsep(&opts, ",")) != NULL) {
932 int token;
933 if (!*p)
934 continue;
935
936 token = match_token(p, tokens, args);
937 switch (token) {
938 case Opt_subvol:
939 kfree(*subvol_name);
940 *subvol_name = match_strdup(&args[0]);
941 if (!*subvol_name) {
942 error = -ENOMEM;
943 goto out;
944 }
945 break;
946 case Opt_subvolid:
947 error = match_u64(&args[0], &subvolid);
948 if (error)
949 goto out;
950
951 /* we want the original fs_tree */
952 if (subvolid == 0)
953 subvolid = BTRFS_FS_TREE_OBJECTID;
954
955 *subvol_objectid = subvolid;
956 break;
957 default:
958 break;
959 }
960 }
961
962 out:
963 kfree(orig);
964 return error;
965 }
966
967 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
968 u64 subvol_objectid)
969 {
970 struct btrfs_root *root = fs_info->tree_root;
971 struct btrfs_root *fs_root = NULL;
972 struct btrfs_root_ref *root_ref;
973 struct btrfs_inode_ref *inode_ref;
974 struct btrfs_key key;
975 struct btrfs_path *path = NULL;
976 char *name = NULL, *ptr;
977 u64 dirid;
978 int len;
979 int ret;
980
981 path = btrfs_alloc_path();
982 if (!path) {
983 ret = -ENOMEM;
984 goto err;
985 }
986
987 name = kmalloc(PATH_MAX, GFP_KERNEL);
988 if (!name) {
989 ret = -ENOMEM;
990 goto err;
991 }
992 ptr = name + PATH_MAX - 1;
993 ptr[0] = '\0';
994
995 /*
996 * Walk up the subvolume trees in the tree of tree roots by root
997 * backrefs until we hit the top-level subvolume.
998 */
999 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1000 key.objectid = subvol_objectid;
1001 key.type = BTRFS_ROOT_BACKREF_KEY;
1002 key.offset = (u64)-1;
1003
1004 ret = btrfs_search_backwards(root, &key, path);
1005 if (ret < 0) {
1006 goto err;
1007 } else if (ret > 0) {
1008 ret = -ENOENT;
1009 goto err;
1010 }
1011
1012 subvol_objectid = key.offset;
1013
1014 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1015 struct btrfs_root_ref);
1016 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1017 ptr -= len + 1;
1018 if (ptr < name) {
1019 ret = -ENAMETOOLONG;
1020 goto err;
1021 }
1022 read_extent_buffer(path->nodes[0], ptr + 1,
1023 (unsigned long)(root_ref + 1), len);
1024 ptr[0] = '/';
1025 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1026 btrfs_release_path(path);
1027
1028 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1029 if (IS_ERR(fs_root)) {
1030 ret = PTR_ERR(fs_root);
1031 fs_root = NULL;
1032 goto err;
1033 }
1034
1035 /*
1036 * Walk up the filesystem tree by inode refs until we hit the
1037 * root directory.
1038 */
1039 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1040 key.objectid = dirid;
1041 key.type = BTRFS_INODE_REF_KEY;
1042 key.offset = (u64)-1;
1043
1044 ret = btrfs_search_backwards(fs_root, &key, path);
1045 if (ret < 0) {
1046 goto err;
1047 } else if (ret > 0) {
1048 ret = -ENOENT;
1049 goto err;
1050 }
1051
1052 dirid = key.offset;
1053
1054 inode_ref = btrfs_item_ptr(path->nodes[0],
1055 path->slots[0],
1056 struct btrfs_inode_ref);
1057 len = btrfs_inode_ref_name_len(path->nodes[0],
1058 inode_ref);
1059 ptr -= len + 1;
1060 if (ptr < name) {
1061 ret = -ENAMETOOLONG;
1062 goto err;
1063 }
1064 read_extent_buffer(path->nodes[0], ptr + 1,
1065 (unsigned long)(inode_ref + 1), len);
1066 ptr[0] = '/';
1067 btrfs_release_path(path);
1068 }
1069 btrfs_put_root(fs_root);
1070 fs_root = NULL;
1071 }
1072
1073 btrfs_free_path(path);
1074 if (ptr == name + PATH_MAX - 1) {
1075 name[0] = '/';
1076 name[1] = '\0';
1077 } else {
1078 memmove(name, ptr, name + PATH_MAX - ptr);
1079 }
1080 return name;
1081
1082 err:
1083 btrfs_put_root(fs_root);
1084 btrfs_free_path(path);
1085 kfree(name);
1086 return ERR_PTR(ret);
1087 }
1088
1089 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1090 {
1091 struct btrfs_root *root = fs_info->tree_root;
1092 struct btrfs_dir_item *di;
1093 struct btrfs_path *path;
1094 struct btrfs_key location;
1095 struct fscrypt_str name = FSTR_INIT("default", 7);
1096 u64 dir_id;
1097
1098 path = btrfs_alloc_path();
1099 if (!path)
1100 return -ENOMEM;
1101
1102 /*
1103 * Find the "default" dir item which points to the root item that we
1104 * will mount by default if we haven't been given a specific subvolume
1105 * to mount.
1106 */
1107 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1108 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
1109 if (IS_ERR(di)) {
1110 btrfs_free_path(path);
1111 return PTR_ERR(di);
1112 }
1113 if (!di) {
1114 /*
1115 * Ok the default dir item isn't there. This is weird since
1116 * it's always been there, but don't freak out, just try and
1117 * mount the top-level subvolume.
1118 */
1119 btrfs_free_path(path);
1120 *objectid = BTRFS_FS_TREE_OBJECTID;
1121 return 0;
1122 }
1123
1124 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1125 btrfs_free_path(path);
1126 *objectid = location.objectid;
1127 return 0;
1128 }
1129
1130 static int btrfs_fill_super(struct super_block *sb,
1131 struct btrfs_fs_devices *fs_devices,
1132 void *data)
1133 {
1134 struct inode *inode;
1135 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1136 int err;
1137
1138 sb->s_maxbytes = MAX_LFS_FILESIZE;
1139 sb->s_magic = BTRFS_SUPER_MAGIC;
1140 sb->s_op = &btrfs_super_ops;
1141 sb->s_d_op = &btrfs_dentry_operations;
1142 sb->s_export_op = &btrfs_export_ops;
1143 #ifdef CONFIG_FS_VERITY
1144 sb->s_vop = &btrfs_verityops;
1145 #endif
1146 sb->s_xattr = btrfs_xattr_handlers;
1147 sb->s_time_gran = 1;
1148 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1149 sb->s_flags |= SB_POSIXACL;
1150 #endif
1151 sb->s_flags |= SB_I_VERSION;
1152 sb->s_iflags |= SB_I_CGROUPWB;
1153
1154 err = super_setup_bdi(sb);
1155 if (err) {
1156 btrfs_err(fs_info, "super_setup_bdi failed");
1157 return err;
1158 }
1159
1160 err = open_ctree(sb, fs_devices, (char *)data);
1161 if (err) {
1162 btrfs_err(fs_info, "open_ctree failed");
1163 return err;
1164 }
1165
1166 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1167 if (IS_ERR(inode)) {
1168 err = PTR_ERR(inode);
1169 btrfs_handle_fs_error(fs_info, err, NULL);
1170 goto fail_close;
1171 }
1172
1173 sb->s_root = d_make_root(inode);
1174 if (!sb->s_root) {
1175 err = -ENOMEM;
1176 goto fail_close;
1177 }
1178
1179 sb->s_flags |= SB_ACTIVE;
1180 return 0;
1181
1182 fail_close:
1183 close_ctree(fs_info);
1184 return err;
1185 }
1186
1187 int btrfs_sync_fs(struct super_block *sb, int wait)
1188 {
1189 struct btrfs_trans_handle *trans;
1190 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1191 struct btrfs_root *root = fs_info->tree_root;
1192
1193 trace_btrfs_sync_fs(fs_info, wait);
1194
1195 if (!wait) {
1196 filemap_flush(fs_info->btree_inode->i_mapping);
1197 return 0;
1198 }
1199
1200 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1201
1202 trans = btrfs_attach_transaction_barrier(root);
1203 if (IS_ERR(trans)) {
1204 /* no transaction, don't bother */
1205 if (PTR_ERR(trans) == -ENOENT) {
1206 /*
1207 * Exit unless we have some pending changes
1208 * that need to go through commit
1209 */
1210 if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1211 &fs_info->flags))
1212 return 0;
1213 /*
1214 * A non-blocking test if the fs is frozen. We must not
1215 * start a new transaction here otherwise a deadlock
1216 * happens. The pending operations are delayed to the
1217 * next commit after thawing.
1218 */
1219 if (sb_start_write_trylock(sb))
1220 sb_end_write(sb);
1221 else
1222 return 0;
1223 trans = btrfs_start_transaction(root, 0);
1224 }
1225 if (IS_ERR(trans))
1226 return PTR_ERR(trans);
1227 }
1228 return btrfs_commit_transaction(trans);
1229 }
1230
1231 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1232 {
1233 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1234 *printed = true;
1235 }
1236
1237 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1238 {
1239 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1240 const char *compress_type;
1241 const char *subvol_name;
1242 bool printed = false;
1243
1244 if (btrfs_test_opt(info, DEGRADED))
1245 seq_puts(seq, ",degraded");
1246 if (btrfs_test_opt(info, NODATASUM))
1247 seq_puts(seq, ",nodatasum");
1248 if (btrfs_test_opt(info, NODATACOW))
1249 seq_puts(seq, ",nodatacow");
1250 if (btrfs_test_opt(info, NOBARRIER))
1251 seq_puts(seq, ",nobarrier");
1252 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1253 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1254 if (info->thread_pool_size != min_t(unsigned long,
1255 num_online_cpus() + 2, 8))
1256 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1257 if (btrfs_test_opt(info, COMPRESS)) {
1258 compress_type = btrfs_compress_type2str(info->compress_type);
1259 if (btrfs_test_opt(info, FORCE_COMPRESS))
1260 seq_printf(seq, ",compress-force=%s", compress_type);
1261 else
1262 seq_printf(seq, ",compress=%s", compress_type);
1263 if (info->compress_level)
1264 seq_printf(seq, ":%d", info->compress_level);
1265 }
1266 if (btrfs_test_opt(info, NOSSD))
1267 seq_puts(seq, ",nossd");
1268 if (btrfs_test_opt(info, SSD_SPREAD))
1269 seq_puts(seq, ",ssd_spread");
1270 else if (btrfs_test_opt(info, SSD))
1271 seq_puts(seq, ",ssd");
1272 if (btrfs_test_opt(info, NOTREELOG))
1273 seq_puts(seq, ",notreelog");
1274 if (btrfs_test_opt(info, NOLOGREPLAY))
1275 print_rescue_option(seq, "nologreplay", &printed);
1276 if (btrfs_test_opt(info, USEBACKUPROOT))
1277 print_rescue_option(seq, "usebackuproot", &printed);
1278 if (btrfs_test_opt(info, IGNOREBADROOTS))
1279 print_rescue_option(seq, "ignorebadroots", &printed);
1280 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1281 print_rescue_option(seq, "ignoredatacsums", &printed);
1282 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1283 seq_puts(seq, ",flushoncommit");
1284 if (btrfs_test_opt(info, DISCARD_SYNC))
1285 seq_puts(seq, ",discard");
1286 if (btrfs_test_opt(info, DISCARD_ASYNC))
1287 seq_puts(seq, ",discard=async");
1288 if (!(info->sb->s_flags & SB_POSIXACL))
1289 seq_puts(seq, ",noacl");
1290 if (btrfs_free_space_cache_v1_active(info))
1291 seq_puts(seq, ",space_cache");
1292 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1293 seq_puts(seq, ",space_cache=v2");
1294 else
1295 seq_puts(seq, ",nospace_cache");
1296 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1297 seq_puts(seq, ",rescan_uuid_tree");
1298 if (btrfs_test_opt(info, CLEAR_CACHE))
1299 seq_puts(seq, ",clear_cache");
1300 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1301 seq_puts(seq, ",user_subvol_rm_allowed");
1302 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1303 seq_puts(seq, ",enospc_debug");
1304 if (btrfs_test_opt(info, AUTO_DEFRAG))
1305 seq_puts(seq, ",autodefrag");
1306 if (btrfs_test_opt(info, SKIP_BALANCE))
1307 seq_puts(seq, ",skip_balance");
1308 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1309 if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1310 seq_puts(seq, ",check_int_data");
1311 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1312 seq_puts(seq, ",check_int");
1313 if (info->check_integrity_print_mask)
1314 seq_printf(seq, ",check_int_print_mask=%d",
1315 info->check_integrity_print_mask);
1316 #endif
1317 if (info->metadata_ratio)
1318 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1319 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1320 seq_puts(seq, ",fatal_errors=panic");
1321 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1322 seq_printf(seq, ",commit=%u", info->commit_interval);
1323 #ifdef CONFIG_BTRFS_DEBUG
1324 if (btrfs_test_opt(info, FRAGMENT_DATA))
1325 seq_puts(seq, ",fragment=data");
1326 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1327 seq_puts(seq, ",fragment=metadata");
1328 #endif
1329 if (btrfs_test_opt(info, REF_VERIFY))
1330 seq_puts(seq, ",ref_verify");
1331 seq_printf(seq, ",subvolid=%llu",
1332 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1333 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1334 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1335 if (!IS_ERR(subvol_name)) {
1336 seq_puts(seq, ",subvol=");
1337 seq_escape(seq, subvol_name, " \t\n\\");
1338 kfree(subvol_name);
1339 }
1340 return 0;
1341 }
1342
1343 static int btrfs_test_super(struct super_block *s, void *data)
1344 {
1345 struct btrfs_fs_info *p = data;
1346 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1347
1348 return fs_info->fs_devices == p->fs_devices;
1349 }
1350
1351 static int btrfs_set_super(struct super_block *s, void *data)
1352 {
1353 int err = set_anon_super(s, data);
1354 if (!err)
1355 s->s_fs_info = data;
1356 return err;
1357 }
1358
1359 /*
1360 * subvolumes are identified by ino 256
1361 */
1362 static inline int is_subvolume_inode(struct inode *inode)
1363 {
1364 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1365 return 1;
1366 return 0;
1367 }
1368
1369 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1370 struct vfsmount *mnt)
1371 {
1372 struct dentry *root;
1373 int ret;
1374
1375 if (!subvol_name) {
1376 if (!subvol_objectid) {
1377 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1378 &subvol_objectid);
1379 if (ret) {
1380 root = ERR_PTR(ret);
1381 goto out;
1382 }
1383 }
1384 subvol_name = btrfs_get_subvol_name_from_objectid(
1385 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1386 if (IS_ERR(subvol_name)) {
1387 root = ERR_CAST(subvol_name);
1388 subvol_name = NULL;
1389 goto out;
1390 }
1391
1392 }
1393
1394 root = mount_subtree(mnt, subvol_name);
1395 /* mount_subtree() drops our reference on the vfsmount. */
1396 mnt = NULL;
1397
1398 if (!IS_ERR(root)) {
1399 struct super_block *s = root->d_sb;
1400 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1401 struct inode *root_inode = d_inode(root);
1402 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1403
1404 ret = 0;
1405 if (!is_subvolume_inode(root_inode)) {
1406 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1407 subvol_name);
1408 ret = -EINVAL;
1409 }
1410 if (subvol_objectid && root_objectid != subvol_objectid) {
1411 /*
1412 * This will also catch a race condition where a
1413 * subvolume which was passed by ID is renamed and
1414 * another subvolume is renamed over the old location.
1415 */
1416 btrfs_err(fs_info,
1417 "subvol '%s' does not match subvolid %llu",
1418 subvol_name, subvol_objectid);
1419 ret = -EINVAL;
1420 }
1421 if (ret) {
1422 dput(root);
1423 root = ERR_PTR(ret);
1424 deactivate_locked_super(s);
1425 }
1426 }
1427
1428 out:
1429 mntput(mnt);
1430 kfree(subvol_name);
1431 return root;
1432 }
1433
1434 /*
1435 * Find a superblock for the given device / mount point.
1436 *
1437 * Note: This is based on mount_bdev from fs/super.c with a few additions
1438 * for multiple device setup. Make sure to keep it in sync.
1439 */
1440 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1441 int flags, const char *device_name, void *data)
1442 {
1443 struct block_device *bdev = NULL;
1444 struct super_block *s;
1445 struct btrfs_device *device = NULL;
1446 struct btrfs_fs_devices *fs_devices = NULL;
1447 struct btrfs_fs_info *fs_info = NULL;
1448 void *new_sec_opts = NULL;
1449 blk_mode_t mode = sb_open_mode(flags);
1450 int error = 0;
1451
1452 if (data) {
1453 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1454 if (error)
1455 return ERR_PTR(error);
1456 }
1457
1458 /*
1459 * Setup a dummy root and fs_info for test/set super. This is because
1460 * we don't actually fill this stuff out until open_ctree, but we need
1461 * then open_ctree will properly initialize the file system specific
1462 * settings later. btrfs_init_fs_info initializes the static elements
1463 * of the fs_info (locks and such) to make cleanup easier if we find a
1464 * superblock with our given fs_devices later on at sget() time.
1465 */
1466 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1467 if (!fs_info) {
1468 error = -ENOMEM;
1469 goto error_sec_opts;
1470 }
1471 btrfs_init_fs_info(fs_info);
1472
1473 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1474 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1475 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1476 error = -ENOMEM;
1477 goto error_fs_info;
1478 }
1479
1480 mutex_lock(&uuid_mutex);
1481 error = btrfs_parse_device_options(data, mode);
1482 if (error) {
1483 mutex_unlock(&uuid_mutex);
1484 goto error_fs_info;
1485 }
1486
1487 device = btrfs_scan_one_device(device_name, mode);
1488 if (IS_ERR(device)) {
1489 mutex_unlock(&uuid_mutex);
1490 error = PTR_ERR(device);
1491 goto error_fs_info;
1492 }
1493
1494 fs_devices = device->fs_devices;
1495 fs_info->fs_devices = fs_devices;
1496
1497 error = btrfs_open_devices(fs_devices, mode, fs_type);
1498 mutex_unlock(&uuid_mutex);
1499 if (error)
1500 goto error_fs_info;
1501
1502 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1503 error = -EACCES;
1504 goto error_close_devices;
1505 }
1506
1507 bdev = fs_devices->latest_dev->bdev;
1508 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1509 fs_info);
1510 if (IS_ERR(s)) {
1511 error = PTR_ERR(s);
1512 goto error_close_devices;
1513 }
1514
1515 if (s->s_root) {
1516 btrfs_close_devices(fs_devices);
1517 btrfs_free_fs_info(fs_info);
1518 if ((flags ^ s->s_flags) & SB_RDONLY)
1519 error = -EBUSY;
1520 } else {
1521 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1522 shrinker_debugfs_rename(&s->s_shrink, "sb-%s:%s", fs_type->name,
1523 s->s_id);
1524 btrfs_sb(s)->bdev_holder = fs_type;
1525 error = btrfs_fill_super(s, fs_devices, data);
1526 }
1527 if (!error)
1528 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1529 security_free_mnt_opts(&new_sec_opts);
1530 if (error) {
1531 deactivate_locked_super(s);
1532 return ERR_PTR(error);
1533 }
1534
1535 return dget(s->s_root);
1536
1537 error_close_devices:
1538 btrfs_close_devices(fs_devices);
1539 error_fs_info:
1540 btrfs_free_fs_info(fs_info);
1541 error_sec_opts:
1542 security_free_mnt_opts(&new_sec_opts);
1543 return ERR_PTR(error);
1544 }
1545
1546 /*
1547 * Mount function which is called by VFS layer.
1548 *
1549 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1550 * which needs vfsmount* of device's root (/). This means device's root has to
1551 * be mounted internally in any case.
1552 *
1553 * Operation flow:
1554 * 1. Parse subvol id related options for later use in mount_subvol().
1555 *
1556 * 2. Mount device's root (/) by calling vfs_kern_mount().
1557 *
1558 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1559 * first place. In order to avoid calling btrfs_mount() again, we use
1560 * different file_system_type which is not registered to VFS by
1561 * register_filesystem() (btrfs_root_fs_type). As a result,
1562 * btrfs_mount_root() is called. The return value will be used by
1563 * mount_subtree() in mount_subvol().
1564 *
1565 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1566 * "btrfs subvolume set-default", mount_subvol() is called always.
1567 */
1568 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1569 const char *device_name, void *data)
1570 {
1571 struct vfsmount *mnt_root;
1572 struct dentry *root;
1573 char *subvol_name = NULL;
1574 u64 subvol_objectid = 0;
1575 int error = 0;
1576
1577 error = btrfs_parse_subvol_options(data, &subvol_name,
1578 &subvol_objectid);
1579 if (error) {
1580 kfree(subvol_name);
1581 return ERR_PTR(error);
1582 }
1583
1584 /* mount device's root (/) */
1585 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1586 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1587 if (flags & SB_RDONLY) {
1588 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1589 flags & ~SB_RDONLY, device_name, data);
1590 } else {
1591 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1592 flags | SB_RDONLY, device_name, data);
1593 if (IS_ERR(mnt_root)) {
1594 root = ERR_CAST(mnt_root);
1595 kfree(subvol_name);
1596 goto out;
1597 }
1598
1599 down_write(&mnt_root->mnt_sb->s_umount);
1600 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1601 up_write(&mnt_root->mnt_sb->s_umount);
1602 if (error < 0) {
1603 root = ERR_PTR(error);
1604 mntput(mnt_root);
1605 kfree(subvol_name);
1606 goto out;
1607 }
1608 }
1609 }
1610 if (IS_ERR(mnt_root)) {
1611 root = ERR_CAST(mnt_root);
1612 kfree(subvol_name);
1613 goto out;
1614 }
1615
1616 /* mount_subvol() will free subvol_name and mnt_root */
1617 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1618
1619 out:
1620 return root;
1621 }
1622
1623 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1624 u32 new_pool_size, u32 old_pool_size)
1625 {
1626 if (new_pool_size == old_pool_size)
1627 return;
1628
1629 fs_info->thread_pool_size = new_pool_size;
1630
1631 btrfs_info(fs_info, "resize thread pool %d -> %d",
1632 old_pool_size, new_pool_size);
1633
1634 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1635 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1636 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1637 workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1638 workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1639 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1640 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1641 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1642 }
1643
1644 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1645 unsigned long old_opts, int flags)
1646 {
1647 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1648 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1649 (flags & SB_RDONLY))) {
1650 /* wait for any defraggers to finish */
1651 wait_event(fs_info->transaction_wait,
1652 (atomic_read(&fs_info->defrag_running) == 0));
1653 if (flags & SB_RDONLY)
1654 sync_filesystem(fs_info->sb);
1655 }
1656 }
1657
1658 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1659 unsigned long old_opts)
1660 {
1661 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1662
1663 /*
1664 * We need to cleanup all defragable inodes if the autodefragment is
1665 * close or the filesystem is read only.
1666 */
1667 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1668 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1669 btrfs_cleanup_defrag_inodes(fs_info);
1670 }
1671
1672 /* If we toggled discard async */
1673 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1674 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1675 btrfs_discard_resume(fs_info);
1676 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1677 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1678 btrfs_discard_cleanup(fs_info);
1679
1680 /* If we toggled space cache */
1681 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1682 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1683 }
1684
1685 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1686 {
1687 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1688 unsigned old_flags = sb->s_flags;
1689 unsigned long old_opts = fs_info->mount_opt;
1690 unsigned long old_compress_type = fs_info->compress_type;
1691 u64 old_max_inline = fs_info->max_inline;
1692 u32 old_thread_pool_size = fs_info->thread_pool_size;
1693 u32 old_metadata_ratio = fs_info->metadata_ratio;
1694 int ret;
1695
1696 sync_filesystem(sb);
1697 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1698
1699 if (data) {
1700 void *new_sec_opts = NULL;
1701
1702 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1703 if (!ret)
1704 ret = security_sb_remount(sb, new_sec_opts);
1705 security_free_mnt_opts(&new_sec_opts);
1706 if (ret)
1707 goto restore;
1708 }
1709
1710 ret = btrfs_parse_options(fs_info, data, *flags);
1711 if (ret)
1712 goto restore;
1713
1714 ret = btrfs_check_features(fs_info, !(*flags & SB_RDONLY));
1715 if (ret < 0)
1716 goto restore;
1717
1718 btrfs_remount_begin(fs_info, old_opts, *flags);
1719 btrfs_resize_thread_pool(fs_info,
1720 fs_info->thread_pool_size, old_thread_pool_size);
1721
1722 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1723 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1724 (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1725 btrfs_warn(fs_info,
1726 "remount supports changing free space tree only from ro to rw");
1727 /* Make sure free space cache options match the state on disk */
1728 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1729 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1730 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1731 }
1732 if (btrfs_free_space_cache_v1_active(fs_info)) {
1733 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1734 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1735 }
1736 }
1737
1738 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1739 goto out;
1740
1741 if (*flags & SB_RDONLY) {
1742 /*
1743 * this also happens on 'umount -rf' or on shutdown, when
1744 * the filesystem is busy.
1745 */
1746 cancel_work_sync(&fs_info->async_reclaim_work);
1747 cancel_work_sync(&fs_info->async_data_reclaim_work);
1748
1749 btrfs_discard_cleanup(fs_info);
1750
1751 /* wait for the uuid_scan task to finish */
1752 down(&fs_info->uuid_tree_rescan_sem);
1753 /* avoid complains from lockdep et al. */
1754 up(&fs_info->uuid_tree_rescan_sem);
1755
1756 btrfs_set_sb_rdonly(sb);
1757
1758 /*
1759 * Setting SB_RDONLY will put the cleaner thread to
1760 * sleep at the next loop if it's already active.
1761 * If it's already asleep, we'll leave unused block
1762 * groups on disk until we're mounted read-write again
1763 * unless we clean them up here.
1764 */
1765 btrfs_delete_unused_bgs(fs_info);
1766
1767 /*
1768 * The cleaner task could be already running before we set the
1769 * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
1770 * We must make sure that after we finish the remount, i.e. after
1771 * we call btrfs_commit_super(), the cleaner can no longer start
1772 * a transaction - either because it was dropping a dead root,
1773 * running delayed iputs or deleting an unused block group (the
1774 * cleaner picked a block group from the list of unused block
1775 * groups before we were able to in the previous call to
1776 * btrfs_delete_unused_bgs()).
1777 */
1778 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
1779 TASK_UNINTERRUPTIBLE);
1780
1781 /*
1782 * We've set the superblock to RO mode, so we might have made
1783 * the cleaner task sleep without running all pending delayed
1784 * iputs. Go through all the delayed iputs here, so that if an
1785 * unmount happens without remounting RW we don't end up at
1786 * finishing close_ctree() with a non-empty list of delayed
1787 * iputs.
1788 */
1789 btrfs_run_delayed_iputs(fs_info);
1790
1791 btrfs_dev_replace_suspend_for_unmount(fs_info);
1792 btrfs_scrub_cancel(fs_info);
1793 btrfs_pause_balance(fs_info);
1794
1795 /*
1796 * Pause the qgroup rescan worker if it is running. We don't want
1797 * it to be still running after we are in RO mode, as after that,
1798 * by the time we unmount, it might have left a transaction open,
1799 * so we would leak the transaction and/or crash.
1800 */
1801 btrfs_qgroup_wait_for_completion(fs_info, false);
1802
1803 ret = btrfs_commit_super(fs_info);
1804 if (ret)
1805 goto restore;
1806 } else {
1807 if (BTRFS_FS_ERROR(fs_info)) {
1808 btrfs_err(fs_info,
1809 "Remounting read-write after error is not allowed");
1810 ret = -EINVAL;
1811 goto restore;
1812 }
1813 if (fs_info->fs_devices->rw_devices == 0) {
1814 ret = -EACCES;
1815 goto restore;
1816 }
1817
1818 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1819 btrfs_warn(fs_info,
1820 "too many missing devices, writable remount is not allowed");
1821 ret = -EACCES;
1822 goto restore;
1823 }
1824
1825 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1826 btrfs_warn(fs_info,
1827 "mount required to replay tree-log, cannot remount read-write");
1828 ret = -EINVAL;
1829 goto restore;
1830 }
1831
1832 /*
1833 * NOTE: when remounting with a change that does writes, don't
1834 * put it anywhere above this point, as we are not sure to be
1835 * safe to write until we pass the above checks.
1836 */
1837 ret = btrfs_start_pre_rw_mount(fs_info);
1838 if (ret)
1839 goto restore;
1840
1841 btrfs_clear_sb_rdonly(sb);
1842
1843 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1844
1845 /*
1846 * If we've gone from readonly -> read/write, we need to get
1847 * our sync/async discard lists in the right state.
1848 */
1849 btrfs_discard_resume(fs_info);
1850 }
1851 out:
1852 /*
1853 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
1854 * since the absence of the flag means it can be toggled off by remount.
1855 */
1856 *flags |= SB_I_VERSION;
1857
1858 wake_up_process(fs_info->transaction_kthread);
1859 btrfs_remount_cleanup(fs_info, old_opts);
1860 btrfs_clear_oneshot_options(fs_info);
1861 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1862
1863 return 0;
1864
1865 restore:
1866 /* We've hit an error - don't reset SB_RDONLY */
1867 if (sb_rdonly(sb))
1868 old_flags |= SB_RDONLY;
1869 if (!(old_flags & SB_RDONLY))
1870 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
1871 sb->s_flags = old_flags;
1872 fs_info->mount_opt = old_opts;
1873 fs_info->compress_type = old_compress_type;
1874 fs_info->max_inline = old_max_inline;
1875 btrfs_resize_thread_pool(fs_info,
1876 old_thread_pool_size, fs_info->thread_pool_size);
1877 fs_info->metadata_ratio = old_metadata_ratio;
1878 btrfs_remount_cleanup(fs_info, old_opts);
1879 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1880
1881 return ret;
1882 }
1883
1884 /* Used to sort the devices by max_avail(descending sort) */
1885 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1886 {
1887 const struct btrfs_device_info *dev_info1 = a;
1888 const struct btrfs_device_info *dev_info2 = b;
1889
1890 if (dev_info1->max_avail > dev_info2->max_avail)
1891 return -1;
1892 else if (dev_info1->max_avail < dev_info2->max_avail)
1893 return 1;
1894 return 0;
1895 }
1896
1897 /*
1898 * sort the devices by max_avail, in which max free extent size of each device
1899 * is stored.(Descending Sort)
1900 */
1901 static inline void btrfs_descending_sort_devices(
1902 struct btrfs_device_info *devices,
1903 size_t nr_devices)
1904 {
1905 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1906 btrfs_cmp_device_free_bytes, NULL);
1907 }
1908
1909 /*
1910 * The helper to calc the free space on the devices that can be used to store
1911 * file data.
1912 */
1913 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1914 u64 *free_bytes)
1915 {
1916 struct btrfs_device_info *devices_info;
1917 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1918 struct btrfs_device *device;
1919 u64 type;
1920 u64 avail_space;
1921 u64 min_stripe_size;
1922 int num_stripes = 1;
1923 int i = 0, nr_devices;
1924 const struct btrfs_raid_attr *rattr;
1925
1926 /*
1927 * We aren't under the device list lock, so this is racy-ish, but good
1928 * enough for our purposes.
1929 */
1930 nr_devices = fs_info->fs_devices->open_devices;
1931 if (!nr_devices) {
1932 smp_mb();
1933 nr_devices = fs_info->fs_devices->open_devices;
1934 ASSERT(nr_devices);
1935 if (!nr_devices) {
1936 *free_bytes = 0;
1937 return 0;
1938 }
1939 }
1940
1941 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1942 GFP_KERNEL);
1943 if (!devices_info)
1944 return -ENOMEM;
1945
1946 /* calc min stripe number for data space allocation */
1947 type = btrfs_data_alloc_profile(fs_info);
1948 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1949
1950 if (type & BTRFS_BLOCK_GROUP_RAID0)
1951 num_stripes = nr_devices;
1952 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1953 num_stripes = rattr->ncopies;
1954 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1955 num_stripes = 4;
1956
1957 /* Adjust for more than 1 stripe per device */
1958 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1959
1960 rcu_read_lock();
1961 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1962 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1963 &device->dev_state) ||
1964 !device->bdev ||
1965 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1966 continue;
1967
1968 if (i >= nr_devices)
1969 break;
1970
1971 avail_space = device->total_bytes - device->bytes_used;
1972
1973 /* align with stripe_len */
1974 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1975
1976 /*
1977 * Ensure we have at least min_stripe_size on top of the
1978 * reserved space on the device.
1979 */
1980 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1981 continue;
1982
1983 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1984
1985 devices_info[i].dev = device;
1986 devices_info[i].max_avail = avail_space;
1987
1988 i++;
1989 }
1990 rcu_read_unlock();
1991
1992 nr_devices = i;
1993
1994 btrfs_descending_sort_devices(devices_info, nr_devices);
1995
1996 i = nr_devices - 1;
1997 avail_space = 0;
1998 while (nr_devices >= rattr->devs_min) {
1999 num_stripes = min(num_stripes, nr_devices);
2000
2001 if (devices_info[i].max_avail >= min_stripe_size) {
2002 int j;
2003 u64 alloc_size;
2004
2005 avail_space += devices_info[i].max_avail * num_stripes;
2006 alloc_size = devices_info[i].max_avail;
2007 for (j = i + 1 - num_stripes; j <= i; j++)
2008 devices_info[j].max_avail -= alloc_size;
2009 }
2010 i--;
2011 nr_devices--;
2012 }
2013
2014 kfree(devices_info);
2015 *free_bytes = avail_space;
2016 return 0;
2017 }
2018
2019 /*
2020 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2021 *
2022 * If there's a redundant raid level at DATA block groups, use the respective
2023 * multiplier to scale the sizes.
2024 *
2025 * Unused device space usage is based on simulating the chunk allocator
2026 * algorithm that respects the device sizes and order of allocations. This is
2027 * a close approximation of the actual use but there are other factors that may
2028 * change the result (like a new metadata chunk).
2029 *
2030 * If metadata is exhausted, f_bavail will be 0.
2031 */
2032 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2033 {
2034 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2035 struct btrfs_super_block *disk_super = fs_info->super_copy;
2036 struct btrfs_space_info *found;
2037 u64 total_used = 0;
2038 u64 total_free_data = 0;
2039 u64 total_free_meta = 0;
2040 u32 bits = fs_info->sectorsize_bits;
2041 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2042 unsigned factor = 1;
2043 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2044 int ret;
2045 u64 thresh = 0;
2046 int mixed = 0;
2047
2048 list_for_each_entry(found, &fs_info->space_info, list) {
2049 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2050 int i;
2051
2052 total_free_data += found->disk_total - found->disk_used;
2053 total_free_data -=
2054 btrfs_account_ro_block_groups_free_space(found);
2055
2056 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2057 if (!list_empty(&found->block_groups[i]))
2058 factor = btrfs_bg_type_to_factor(
2059 btrfs_raid_array[i].bg_flag);
2060 }
2061 }
2062
2063 /*
2064 * Metadata in mixed block group profiles are accounted in data
2065 */
2066 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2067 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2068 mixed = 1;
2069 else
2070 total_free_meta += found->disk_total -
2071 found->disk_used;
2072 }
2073
2074 total_used += found->disk_used;
2075 }
2076
2077 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2078 buf->f_blocks >>= bits;
2079 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2080
2081 /* Account global block reserve as used, it's in logical size already */
2082 spin_lock(&block_rsv->lock);
2083 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2084 if (buf->f_bfree >= block_rsv->size >> bits)
2085 buf->f_bfree -= block_rsv->size >> bits;
2086 else
2087 buf->f_bfree = 0;
2088 spin_unlock(&block_rsv->lock);
2089
2090 buf->f_bavail = div_u64(total_free_data, factor);
2091 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2092 if (ret)
2093 return ret;
2094 buf->f_bavail += div_u64(total_free_data, factor);
2095 buf->f_bavail = buf->f_bavail >> bits;
2096
2097 /*
2098 * We calculate the remaining metadata space minus global reserve. If
2099 * this is (supposedly) smaller than zero, there's no space. But this
2100 * does not hold in practice, the exhausted state happens where's still
2101 * some positive delta. So we apply some guesswork and compare the
2102 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2103 *
2104 * We probably cannot calculate the exact threshold value because this
2105 * depends on the internal reservations requested by various
2106 * operations, so some operations that consume a few metadata will
2107 * succeed even if the Avail is zero. But this is better than the other
2108 * way around.
2109 */
2110 thresh = SZ_4M;
2111
2112 /*
2113 * We only want to claim there's no available space if we can no longer
2114 * allocate chunks for our metadata profile and our global reserve will
2115 * not fit in the free metadata space. If we aren't ->full then we
2116 * still can allocate chunks and thus are fine using the currently
2117 * calculated f_bavail.
2118 */
2119 if (!mixed && block_rsv->space_info->full &&
2120 total_free_meta - thresh < block_rsv->size)
2121 buf->f_bavail = 0;
2122
2123 buf->f_type = BTRFS_SUPER_MAGIC;
2124 buf->f_bsize = dentry->d_sb->s_blocksize;
2125 buf->f_namelen = BTRFS_NAME_LEN;
2126
2127 /* We treat it as constant endianness (it doesn't matter _which_)
2128 because we want the fsid to come out the same whether mounted
2129 on a big-endian or little-endian host */
2130 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2131 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2132 /* Mask in the root object ID too, to disambiguate subvols */
2133 buf->f_fsid.val[0] ^=
2134 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2135 buf->f_fsid.val[1] ^=
2136 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2137
2138 return 0;
2139 }
2140
2141 static void btrfs_kill_super(struct super_block *sb)
2142 {
2143 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2144 kill_anon_super(sb);
2145 btrfs_free_fs_info(fs_info);
2146 }
2147
2148 static struct file_system_type btrfs_fs_type = {
2149 .owner = THIS_MODULE,
2150 .name = "btrfs",
2151 .mount = btrfs_mount,
2152 .kill_sb = btrfs_kill_super,
2153 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_MGTIME,
2154 };
2155
2156 static struct file_system_type btrfs_root_fs_type = {
2157 .owner = THIS_MODULE,
2158 .name = "btrfs",
2159 .mount = btrfs_mount_root,
2160 .kill_sb = btrfs_kill_super,
2161 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA |
2162 FS_ALLOW_IDMAP | FS_MGTIME,
2163 };
2164
2165 MODULE_ALIAS_FS("btrfs");
2166
2167 static int btrfs_control_open(struct inode *inode, struct file *file)
2168 {
2169 /*
2170 * The control file's private_data is used to hold the
2171 * transaction when it is started and is used to keep
2172 * track of whether a transaction is already in progress.
2173 */
2174 file->private_data = NULL;
2175 return 0;
2176 }
2177
2178 /*
2179 * Used by /dev/btrfs-control for devices ioctls.
2180 */
2181 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2182 unsigned long arg)
2183 {
2184 struct btrfs_ioctl_vol_args *vol;
2185 struct btrfs_device *device = NULL;
2186 dev_t devt = 0;
2187 int ret = -ENOTTY;
2188
2189 if (!capable(CAP_SYS_ADMIN))
2190 return -EPERM;
2191
2192 vol = memdup_user((void __user *)arg, sizeof(*vol));
2193 if (IS_ERR(vol))
2194 return PTR_ERR(vol);
2195 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2196
2197 switch (cmd) {
2198 case BTRFS_IOC_SCAN_DEV:
2199 mutex_lock(&uuid_mutex);
2200 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ);
2201 ret = PTR_ERR_OR_ZERO(device);
2202 mutex_unlock(&uuid_mutex);
2203 break;
2204 case BTRFS_IOC_FORGET_DEV:
2205 if (vol->name[0] != 0) {
2206 ret = lookup_bdev(vol->name, &devt);
2207 if (ret)
2208 break;
2209 }
2210 ret = btrfs_forget_devices(devt);
2211 break;
2212 case BTRFS_IOC_DEVICES_READY:
2213 mutex_lock(&uuid_mutex);
2214 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ);
2215 if (IS_ERR(device)) {
2216 mutex_unlock(&uuid_mutex);
2217 ret = PTR_ERR(device);
2218 break;
2219 }
2220 ret = !(device->fs_devices->num_devices ==
2221 device->fs_devices->total_devices);
2222 mutex_unlock(&uuid_mutex);
2223 break;
2224 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2225 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2226 break;
2227 }
2228
2229 kfree(vol);
2230 return ret;
2231 }
2232
2233 static int btrfs_freeze(struct super_block *sb)
2234 {
2235 struct btrfs_trans_handle *trans;
2236 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2237 struct btrfs_root *root = fs_info->tree_root;
2238
2239 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2240 /*
2241 * We don't need a barrier here, we'll wait for any transaction that
2242 * could be in progress on other threads (and do delayed iputs that
2243 * we want to avoid on a frozen filesystem), or do the commit
2244 * ourselves.
2245 */
2246 trans = btrfs_attach_transaction_barrier(root);
2247 if (IS_ERR(trans)) {
2248 /* no transaction, don't bother */
2249 if (PTR_ERR(trans) == -ENOENT)
2250 return 0;
2251 return PTR_ERR(trans);
2252 }
2253 return btrfs_commit_transaction(trans);
2254 }
2255
2256 static int check_dev_super(struct btrfs_device *dev)
2257 {
2258 struct btrfs_fs_info *fs_info = dev->fs_info;
2259 struct btrfs_super_block *sb;
2260 u16 csum_type;
2261 int ret = 0;
2262
2263 /* This should be called with fs still frozen. */
2264 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2265
2266 /* Missing dev, no need to check. */
2267 if (!dev->bdev)
2268 return 0;
2269
2270 /* Only need to check the primary super block. */
2271 sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2272 if (IS_ERR(sb))
2273 return PTR_ERR(sb);
2274
2275 /* Verify the checksum. */
2276 csum_type = btrfs_super_csum_type(sb);
2277 if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2278 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2279 csum_type, btrfs_super_csum_type(fs_info->super_copy));
2280 ret = -EUCLEAN;
2281 goto out;
2282 }
2283
2284 if (btrfs_check_super_csum(fs_info, sb)) {
2285 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2286 ret = -EUCLEAN;
2287 goto out;
2288 }
2289
2290 /* Btrfs_validate_super() includes fsid check against super->fsid. */
2291 ret = btrfs_validate_super(fs_info, sb, 0);
2292 if (ret < 0)
2293 goto out;
2294
2295 if (btrfs_super_generation(sb) != fs_info->last_trans_committed) {
2296 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2297 btrfs_super_generation(sb),
2298 fs_info->last_trans_committed);
2299 ret = -EUCLEAN;
2300 goto out;
2301 }
2302 out:
2303 btrfs_release_disk_super(sb);
2304 return ret;
2305 }
2306
2307 static int btrfs_unfreeze(struct super_block *sb)
2308 {
2309 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2310 struct btrfs_device *device;
2311 int ret = 0;
2312
2313 /*
2314 * Make sure the fs is not changed by accident (like hibernation then
2315 * modified by other OS).
2316 * If we found anything wrong, we mark the fs error immediately.
2317 *
2318 * And since the fs is frozen, no one can modify the fs yet, thus
2319 * we don't need to hold device_list_mutex.
2320 */
2321 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2322 ret = check_dev_super(device);
2323 if (ret < 0) {
2324 btrfs_handle_fs_error(fs_info, ret,
2325 "super block on devid %llu got modified unexpectedly",
2326 device->devid);
2327 break;
2328 }
2329 }
2330 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2331
2332 /*
2333 * We still return 0, to allow VFS layer to unfreeze the fs even the
2334 * above checks failed. Since the fs is either fine or read-only, we're
2335 * safe to continue, without causing further damage.
2336 */
2337 return 0;
2338 }
2339
2340 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2341 {
2342 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2343
2344 /*
2345 * There should be always a valid pointer in latest_dev, it may be stale
2346 * for a short moment in case it's being deleted but still valid until
2347 * the end of RCU grace period.
2348 */
2349 rcu_read_lock();
2350 seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2351 rcu_read_unlock();
2352
2353 return 0;
2354 }
2355
2356 static const struct super_operations btrfs_super_ops = {
2357 .drop_inode = btrfs_drop_inode,
2358 .evict_inode = btrfs_evict_inode,
2359 .put_super = btrfs_put_super,
2360 .sync_fs = btrfs_sync_fs,
2361 .show_options = btrfs_show_options,
2362 .show_devname = btrfs_show_devname,
2363 .alloc_inode = btrfs_alloc_inode,
2364 .destroy_inode = btrfs_destroy_inode,
2365 .free_inode = btrfs_free_inode,
2366 .statfs = btrfs_statfs,
2367 .remount_fs = btrfs_remount,
2368 .freeze_fs = btrfs_freeze,
2369 .unfreeze_fs = btrfs_unfreeze,
2370 };
2371
2372 static const struct file_operations btrfs_ctl_fops = {
2373 .open = btrfs_control_open,
2374 .unlocked_ioctl = btrfs_control_ioctl,
2375 .compat_ioctl = compat_ptr_ioctl,
2376 .owner = THIS_MODULE,
2377 .llseek = noop_llseek,
2378 };
2379
2380 static struct miscdevice btrfs_misc = {
2381 .minor = BTRFS_MINOR,
2382 .name = "btrfs-control",
2383 .fops = &btrfs_ctl_fops
2384 };
2385
2386 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2387 MODULE_ALIAS("devname:btrfs-control");
2388
2389 static int __init btrfs_interface_init(void)
2390 {
2391 return misc_register(&btrfs_misc);
2392 }
2393
2394 static __cold void btrfs_interface_exit(void)
2395 {
2396 misc_deregister(&btrfs_misc);
2397 }
2398
2399 static int __init btrfs_print_mod_info(void)
2400 {
2401 static const char options[] = ""
2402 #ifdef CONFIG_BTRFS_DEBUG
2403 ", debug=on"
2404 #endif
2405 #ifdef CONFIG_BTRFS_ASSERT
2406 ", assert=on"
2407 #endif
2408 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2409 ", integrity-checker=on"
2410 #endif
2411 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2412 ", ref-verify=on"
2413 #endif
2414 #ifdef CONFIG_BLK_DEV_ZONED
2415 ", zoned=yes"
2416 #else
2417 ", zoned=no"
2418 #endif
2419 #ifdef CONFIG_FS_VERITY
2420 ", fsverity=yes"
2421 #else
2422 ", fsverity=no"
2423 #endif
2424 ;
2425 pr_info("Btrfs loaded%s\n", options);
2426 return 0;
2427 }
2428
2429 static int register_btrfs(void)
2430 {
2431 return register_filesystem(&btrfs_fs_type);
2432 }
2433
2434 static void unregister_btrfs(void)
2435 {
2436 unregister_filesystem(&btrfs_fs_type);
2437 }
2438
2439 /* Helper structure for long init/exit functions. */
2440 struct init_sequence {
2441 int (*init_func)(void);
2442 /* Can be NULL if the init_func doesn't need cleanup. */
2443 void (*exit_func)(void);
2444 };
2445
2446 static const struct init_sequence mod_init_seq[] = {
2447 {
2448 .init_func = btrfs_props_init,
2449 .exit_func = NULL,
2450 }, {
2451 .init_func = btrfs_init_sysfs,
2452 .exit_func = btrfs_exit_sysfs,
2453 }, {
2454 .init_func = btrfs_init_compress,
2455 .exit_func = btrfs_exit_compress,
2456 }, {
2457 .init_func = btrfs_init_cachep,
2458 .exit_func = btrfs_destroy_cachep,
2459 }, {
2460 .init_func = btrfs_transaction_init,
2461 .exit_func = btrfs_transaction_exit,
2462 }, {
2463 .init_func = btrfs_ctree_init,
2464 .exit_func = btrfs_ctree_exit,
2465 }, {
2466 .init_func = btrfs_free_space_init,
2467 .exit_func = btrfs_free_space_exit,
2468 }, {
2469 .init_func = extent_state_init_cachep,
2470 .exit_func = extent_state_free_cachep,
2471 }, {
2472 .init_func = extent_buffer_init_cachep,
2473 .exit_func = extent_buffer_free_cachep,
2474 }, {
2475 .init_func = btrfs_bioset_init,
2476 .exit_func = btrfs_bioset_exit,
2477 }, {
2478 .init_func = extent_map_init,
2479 .exit_func = extent_map_exit,
2480 }, {
2481 .init_func = ordered_data_init,
2482 .exit_func = ordered_data_exit,
2483 }, {
2484 .init_func = btrfs_delayed_inode_init,
2485 .exit_func = btrfs_delayed_inode_exit,
2486 }, {
2487 .init_func = btrfs_auto_defrag_init,
2488 .exit_func = btrfs_auto_defrag_exit,
2489 }, {
2490 .init_func = btrfs_delayed_ref_init,
2491 .exit_func = btrfs_delayed_ref_exit,
2492 }, {
2493 .init_func = btrfs_prelim_ref_init,
2494 .exit_func = btrfs_prelim_ref_exit,
2495 }, {
2496 .init_func = btrfs_interface_init,
2497 .exit_func = btrfs_interface_exit,
2498 }, {
2499 .init_func = btrfs_print_mod_info,
2500 .exit_func = NULL,
2501 }, {
2502 .init_func = btrfs_run_sanity_tests,
2503 .exit_func = NULL,
2504 }, {
2505 .init_func = register_btrfs,
2506 .exit_func = unregister_btrfs,
2507 }
2508 };
2509
2510 static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2511
2512 static __always_inline void btrfs_exit_btrfs_fs(void)
2513 {
2514 int i;
2515
2516 for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2517 if (!mod_init_result[i])
2518 continue;
2519 if (mod_init_seq[i].exit_func)
2520 mod_init_seq[i].exit_func();
2521 mod_init_result[i] = false;
2522 }
2523 }
2524
2525 static void __exit exit_btrfs_fs(void)
2526 {
2527 btrfs_exit_btrfs_fs();
2528 btrfs_cleanup_fs_uuids();
2529 }
2530
2531 static int __init init_btrfs_fs(void)
2532 {
2533 int ret;
2534 int i;
2535
2536 for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2537 ASSERT(!mod_init_result[i]);
2538 ret = mod_init_seq[i].init_func();
2539 if (ret < 0) {
2540 btrfs_exit_btrfs_fs();
2541 return ret;
2542 }
2543 mod_init_result[i] = true;
2544 }
2545 return 0;
2546 }
2547
2548 late_initcall(init_btrfs_fs);
2549 module_exit(exit_btrfs_fs)
2550
2551 MODULE_LICENSE("GPL");
2552 MODULE_SOFTDEP("pre: crc32c");
2553 MODULE_SOFTDEP("pre: xxhash64");
2554 MODULE_SOFTDEP("pre: sha256");
2555 MODULE_SOFTDEP("pre: blake2b-256");