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
66 static const struct super_operations btrfs_super_ops;
67 static struct file_system_type btrfs_fs_type;
68
69 static 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. */
78 struct 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
91 enum {
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
141 enum {
142 Opt_fatal_errors_panic,
143 Opt_fatal_errors_bug,
144 };
145
146 static const struct constant_table btrfs_parameter_fatal_errors[] = {
147 { "panic", Opt_fatal_errors_panic },
148 { "bug", Opt_fatal_errors_bug },
149 {}
150 };
151
152 enum {
153 Opt_discard_sync,
154 Opt_discard_async,
155 };
156
157 static const struct constant_table btrfs_parameter_discard[] = {
158 { "sync", Opt_discard_sync },
159 { "async", Opt_discard_async },
160 {}
161 };
162
163 enum {
164 Opt_space_cache_v1,
165 Opt_space_cache_v2,
166 };
167
168 static const struct constant_table btrfs_parameter_space_cache[] = {
169 { "v1", Opt_space_cache_v1 },
170 { "v2", Opt_space_cache_v2 },
171 {}
172 };
173
174 enum {
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
184 static 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
200 enum {
201 Opt_fragment_parameter_data,
202 Opt_fragment_parameter_metadata,
203 Opt_fragment_parameter_all,
204 };
205
206 static 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
214 static 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... */
268 static 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
273 static 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 if (opt == Opt_compress || opt == Opt_compress_force) {
343 ctx->compress_type = BTRFS_COMPRESS_ZLIB;
344 ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
345 btrfs_set_opt(ctx->mount_opt, COMPRESS);
346 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
347 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
348 } else if (strncmp(param->string, "zlib", 4) == 0) {
349 ctx->compress_type = BTRFS_COMPRESS_ZLIB;
350 ctx->compress_level =
351 btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB,
352 param->string + 4);
353 btrfs_set_opt(ctx->mount_opt, COMPRESS);
354 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
355 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
356 } else if (strncmp(param->string, "lzo", 3) == 0) {
357 ctx->compress_type = BTRFS_COMPRESS_LZO;
358 ctx->compress_level = 0;
359 btrfs_set_opt(ctx->mount_opt, COMPRESS);
360 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
361 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
362 } else if (strncmp(param->string, "zstd", 4) == 0) {
363 ctx->compress_type = BTRFS_COMPRESS_ZSTD;
364 ctx->compress_level =
365 btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD,
366 param->string + 4);
367 btrfs_set_opt(ctx->mount_opt, COMPRESS);
368 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
369 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
370 } else if (strncmp(param->string, "no", 2) == 0) {
371 ctx->compress_level = 0;
372 ctx->compress_type = 0;
373 btrfs_clear_opt(ctx->mount_opt, COMPRESS);
374 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
375 } else {
376 btrfs_err(NULL, "unrecognized compression value %s",
377 param->string);
378 return -EINVAL;
379 }
380 break;
381 case Opt_ssd:
382 if (result.negated) {
383 btrfs_set_opt(ctx->mount_opt, NOSSD);
384 btrfs_clear_opt(ctx->mount_opt, SSD);
385 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
386 } else {
387 btrfs_set_opt(ctx->mount_opt, SSD);
388 btrfs_clear_opt(ctx->mount_opt, NOSSD);
389 }
390 break;
391 case Opt_ssd_spread:
392 if (result.negated) {
393 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
394 } else {
395 btrfs_set_opt(ctx->mount_opt, SSD);
396 btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
397 btrfs_clear_opt(ctx->mount_opt, NOSSD);
398 }
399 break;
400 case Opt_barrier:
401 if (result.negated)
402 btrfs_set_opt(ctx->mount_opt, NOBARRIER);
403 else
404 btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
405 break;
406 case Opt_thread_pool:
407 if (result.uint_32 == 0) {
408 btrfs_err(NULL, "invalid value 0 for thread_pool");
409 return -EINVAL;
410 }
411 ctx->thread_pool_size = result.uint_32;
412 break;
413 case Opt_max_inline:
414 ctx->max_inline = memparse(param->string, NULL);
415 break;
416 case Opt_acl:
417 if (result.negated) {
418 fc->sb_flags &= ~SB_POSIXACL;
419 } else {
420 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
421 fc->sb_flags |= SB_POSIXACL;
422 #else
423 btrfs_err(NULL, "support for ACL not compiled in");
424 return -EINVAL;
425 #endif
426 }
427 /*
428 * VFS limits the ability to toggle ACL on and off via remount,
429 * despite every file system allowing this. This seems to be
430 * an oversight since we all do, but it'll fail if we're
431 * remounting. So don't set the mask here, we'll check it in
432 * btrfs_reconfigure and do the toggling ourselves.
433 */
434 if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
435 fc->sb_flags_mask |= SB_POSIXACL;
436 break;
437 case Opt_treelog:
438 if (result.negated)
439 btrfs_set_opt(ctx->mount_opt, NOTREELOG);
440 else
441 btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
442 break;
443 case Opt_nologreplay:
444 btrfs_warn(NULL,
445 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
446 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
447 break;
448 case Opt_norecovery:
449 btrfs_info(NULL,
450 "'norecovery' is for compatibility only, recommended to use 'rescue=nologreplay'");
451 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
452 break;
453 case Opt_flushoncommit:
454 if (result.negated)
455 btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
456 else
457 btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
458 break;
459 case Opt_ratio:
460 ctx->metadata_ratio = result.uint_32;
461 break;
462 case Opt_discard:
463 if (result.negated) {
464 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
465 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
466 btrfs_set_opt(ctx->mount_opt, NODISCARD);
467 } else {
468 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
469 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
470 }
471 break;
472 case Opt_discard_mode:
473 switch (result.uint_32) {
474 case Opt_discard_sync:
475 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
476 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
477 break;
478 case Opt_discard_async:
479 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
480 btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
481 break;
482 default:
483 btrfs_err(NULL, "unrecognized discard mode value %s",
484 param->key);
485 return -EINVAL;
486 }
487 btrfs_clear_opt(ctx->mount_opt, NODISCARD);
488 break;
489 case Opt_space_cache:
490 if (result.negated) {
491 btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
492 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
493 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
494 } else {
495 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
496 btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
497 }
498 break;
499 case Opt_space_cache_version:
500 switch (result.uint_32) {
501 case Opt_space_cache_v1:
502 btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
503 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
504 break;
505 case Opt_space_cache_v2:
506 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
507 btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
508 break;
509 default:
510 btrfs_err(NULL, "unrecognized space_cache value %s",
511 param->key);
512 return -EINVAL;
513 }
514 break;
515 case Opt_rescan_uuid_tree:
516 btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
517 break;
518 case Opt_clear_cache:
519 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
520 break;
521 case Opt_user_subvol_rm_allowed:
522 btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
523 break;
524 case Opt_enospc_debug:
525 if (result.negated)
526 btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
527 else
528 btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
529 break;
530 case Opt_defrag:
531 if (result.negated)
532 btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
533 else
534 btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
535 break;
536 case Opt_usebackuproot:
537 btrfs_warn(NULL,
538 "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
539 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
540
541 /* If we're loading the backup roots we can't trust the space cache. */
542 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
543 break;
544 case Opt_skip_balance:
545 btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
546 break;
547 case Opt_fatal_errors:
548 switch (result.uint_32) {
549 case Opt_fatal_errors_panic:
550 btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
551 break;
552 case Opt_fatal_errors_bug:
553 btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
554 break;
555 default:
556 btrfs_err(NULL, "unrecognized fatal_errors value %s",
557 param->key);
558 return -EINVAL;
559 }
560 break;
561 case Opt_commit_interval:
562 ctx->commit_interval = result.uint_32;
563 if (ctx->commit_interval == 0)
564 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
565 break;
566 case Opt_rescue:
567 switch (result.uint_32) {
568 case Opt_rescue_usebackuproot:
569 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
570 break;
571 case Opt_rescue_nologreplay:
572 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
573 break;
574 case Opt_rescue_ignorebadroots:
575 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
576 break;
577 case Opt_rescue_ignoredatacsums:
578 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
579 break;
580 case Opt_rescue_ignoremetacsums:
581 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
582 break;
583 case Opt_rescue_ignoresuperflags:
584 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
585 break;
586 case Opt_rescue_parameter_all:
587 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
588 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
589 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
590 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
591 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
592 break;
593 default:
594 btrfs_info(NULL, "unrecognized rescue option '%s'",
595 param->key);
596 return -EINVAL;
597 }
598 break;
599 #ifdef CONFIG_BTRFS_DEBUG
600 case Opt_fragment:
601 switch (result.uint_32) {
602 case Opt_fragment_parameter_all:
603 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
604 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
605 break;
606 case Opt_fragment_parameter_metadata:
607 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
608 break;
609 case Opt_fragment_parameter_data:
610 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
611 break;
612 default:
613 btrfs_info(NULL, "unrecognized fragment option '%s'",
614 param->key);
615 return -EINVAL;
616 }
617 break;
618 #endif
619 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
620 case Opt_ref_verify:
621 btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
622 break;
623 #endif
624 default:
625 btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
626 return -EINVAL;
627 }
628
629 return 0;
630 }
631
632 /*
633 * Some options only have meaning at mount time and shouldn't persist across
634 * remounts, or be displayed. Clear these at the end of mount and remount code
635 * paths.
636 */
637 static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
638 {
639 btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
640 btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
641 btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
642 }
643
644 static bool check_ro_option(const struct btrfs_fs_info *fs_info,
645 unsigned long long mount_opt, unsigned long long opt,
646 const char *opt_name)
647 {
648 if (mount_opt & opt) {
649 btrfs_err(fs_info, "%s must be used with ro mount option",
650 opt_name);
651 return true;
652 }
653 return false;
654 }
655
656 bool btrfs_check_options(const struct btrfs_fs_info *info,
657 unsigned long long *mount_opt,
658 unsigned long flags)
659 {
660 bool ret = true;
661
662 if (!(flags & SB_RDONLY) &&
663 (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
664 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
665 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums") ||
666 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREMETACSUMS, "ignoremetacsums") ||
667 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNORESUPERFLAGS, "ignoresuperflags")))
668 ret = false;
669
670 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
671 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
672 !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
673 btrfs_err(info, "cannot disable free-space-tree");
674 ret = false;
675 }
676 if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
677 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
678 btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
679 ret = false;
680 }
681
682 if (btrfs_check_mountopts_zoned(info, mount_opt))
683 ret = false;
684
685 if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
686 if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE)) {
687 btrfs_info(info, "disk space caching is enabled");
688 btrfs_warn(info,
689 "space cache v1 is being deprecated and will be removed in a future release, please use -o space_cache=v2");
690 }
691 if (btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE))
692 btrfs_info(info, "using free-space-tree");
693 }
694
695 return ret;
696 }
697
698 /*
699 * This is subtle, we only call this during open_ctree(). We need to pre-load
700 * the mount options with the on-disk settings. Before the new mount API took
701 * effect we would do this on mount and remount. With the new mount API we'll
702 * only do this on the initial mount.
703 *
704 * This isn't a change in behavior, because we're using the current state of the
705 * file system to set the current mount options. If you mounted with special
706 * options to disable these features and then remounted we wouldn't revert the
707 * settings, because mounting without these features cleared the on-disk
708 * settings, so this being called on re-mount is not needed.
709 */
710 void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
711 {
712 if (fs_info->sectorsize < PAGE_SIZE) {
713 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
714 if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) {
715 btrfs_info(fs_info,
716 "forcing free space tree for sector size %u with page size %lu",
717 fs_info->sectorsize, PAGE_SIZE);
718 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
719 }
720 }
721
722 /*
723 * At this point our mount options are populated, so we only mess with
724 * these settings if we don't have any settings already.
725 */
726 if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
727 return;
728
729 if (btrfs_is_zoned(fs_info) &&
730 btrfs_free_space_cache_v1_active(fs_info)) {
731 btrfs_info(fs_info, "zoned: clearing existing space cache");
732 btrfs_set_super_cache_generation(fs_info->super_copy, 0);
733 return;
734 }
735
736 if (btrfs_test_opt(fs_info, SPACE_CACHE))
737 return;
738
739 if (btrfs_test_opt(fs_info, NOSPACECACHE))
740 return;
741
742 /*
743 * At this point we don't have explicit options set by the user, set
744 * them ourselves based on the state of the file system.
745 */
746 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
747 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
748 else if (btrfs_free_space_cache_v1_active(fs_info))
749 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
750 }
751
752 static void set_device_specific_options(struct btrfs_fs_info *fs_info)
753 {
754 if (!btrfs_test_opt(fs_info, NOSSD) &&
755 !fs_info->fs_devices->rotating)
756 btrfs_set_opt(fs_info->mount_opt, SSD);
757
758 /*
759 * For devices supporting discard turn on discard=async automatically,
760 * unless it's already set or disabled. This could be turned off by
761 * nodiscard for the same mount.
762 *
763 * The zoned mode piggy backs on the discard functionality for
764 * resetting a zone. There is no reason to delay the zone reset as it is
765 * fast enough. So, do not enable async discard for zoned mode.
766 */
767 if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
768 btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
769 btrfs_test_opt(fs_info, NODISCARD)) &&
770 fs_info->fs_devices->discardable &&
771 !btrfs_is_zoned(fs_info))
772 btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
773 }
774
775 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
776 u64 subvol_objectid)
777 {
778 struct btrfs_root *root = fs_info->tree_root;
779 struct btrfs_root *fs_root = NULL;
780 struct btrfs_root_ref *root_ref;
781 struct btrfs_inode_ref *inode_ref;
782 struct btrfs_key key;
783 struct btrfs_path *path = NULL;
784 char *name = NULL, *ptr;
785 u64 dirid;
786 int len;
787 int ret;
788
789 path = btrfs_alloc_path();
790 if (!path) {
791 ret = -ENOMEM;
792 goto err;
793 }
794
795 name = kmalloc(PATH_MAX, GFP_KERNEL);
796 if (!name) {
797 ret = -ENOMEM;
798 goto err;
799 }
800 ptr = name + PATH_MAX - 1;
801 ptr[0] = '\0';
802
803 /*
804 * Walk up the subvolume trees in the tree of tree roots by root
805 * backrefs until we hit the top-level subvolume.
806 */
807 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
808 key.objectid = subvol_objectid;
809 key.type = BTRFS_ROOT_BACKREF_KEY;
810 key.offset = (u64)-1;
811
812 ret = btrfs_search_backwards(root, &key, path);
813 if (ret < 0) {
814 goto err;
815 } else if (ret > 0) {
816 ret = -ENOENT;
817 goto err;
818 }
819
820 subvol_objectid = key.offset;
821
822 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
823 struct btrfs_root_ref);
824 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
825 ptr -= len + 1;
826 if (ptr < name) {
827 ret = -ENAMETOOLONG;
828 goto err;
829 }
830 read_extent_buffer(path->nodes[0], ptr + 1,
831 (unsigned long)(root_ref + 1), len);
832 ptr[0] = '/';
833 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
834 btrfs_release_path(path);
835
836 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
837 if (IS_ERR(fs_root)) {
838 ret = PTR_ERR(fs_root);
839 fs_root = NULL;
840 goto err;
841 }
842
843 /*
844 * Walk up the filesystem tree by inode refs until we hit the
845 * root directory.
846 */
847 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
848 key.objectid = dirid;
849 key.type = BTRFS_INODE_REF_KEY;
850 key.offset = (u64)-1;
851
852 ret = btrfs_search_backwards(fs_root, &key, path);
853 if (ret < 0) {
854 goto err;
855 } else if (ret > 0) {
856 ret = -ENOENT;
857 goto err;
858 }
859
860 dirid = key.offset;
861
862 inode_ref = btrfs_item_ptr(path->nodes[0],
863 path->slots[0],
864 struct btrfs_inode_ref);
865 len = btrfs_inode_ref_name_len(path->nodes[0],
866 inode_ref);
867 ptr -= len + 1;
868 if (ptr < name) {
869 ret = -ENAMETOOLONG;
870 goto err;
871 }
872 read_extent_buffer(path->nodes[0], ptr + 1,
873 (unsigned long)(inode_ref + 1), len);
874 ptr[0] = '/';
875 btrfs_release_path(path);
876 }
877 btrfs_put_root(fs_root);
878 fs_root = NULL;
879 }
880
881 btrfs_free_path(path);
882 if (ptr == name + PATH_MAX - 1) {
883 name[0] = '/';
884 name[1] = '\0';
885 } else {
886 memmove(name, ptr, name + PATH_MAX - ptr);
887 }
888 return name;
889
890 err:
891 btrfs_put_root(fs_root);
892 btrfs_free_path(path);
893 kfree(name);
894 return ERR_PTR(ret);
895 }
896
897 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
898 {
899 struct btrfs_root *root = fs_info->tree_root;
900 struct btrfs_dir_item *di;
901 struct btrfs_path *path;
902 struct btrfs_key location;
903 struct fscrypt_str name = FSTR_INIT("default", 7);
904 u64 dir_id;
905
906 path = btrfs_alloc_path();
907 if (!path)
908 return -ENOMEM;
909
910 /*
911 * Find the "default" dir item which points to the root item that we
912 * will mount by default if we haven't been given a specific subvolume
913 * to mount.
914 */
915 dir_id = btrfs_super_root_dir(fs_info->super_copy);
916 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
917 if (IS_ERR(di)) {
918 btrfs_free_path(path);
919 return PTR_ERR(di);
920 }
921 if (!di) {
922 /*
923 * Ok the default dir item isn't there. This is weird since
924 * it's always been there, but don't freak out, just try and
925 * mount the top-level subvolume.
926 */
927 btrfs_free_path(path);
928 *objectid = BTRFS_FS_TREE_OBJECTID;
929 return 0;
930 }
931
932 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
933 btrfs_free_path(path);
934 *objectid = location.objectid;
935 return 0;
936 }
937
938 static int btrfs_fill_super(struct super_block *sb,
939 struct btrfs_fs_devices *fs_devices,
940 void *data)
941 {
942 struct inode *inode;
943 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
944 int err;
945
946 sb->s_maxbytes = MAX_LFS_FILESIZE;
947 sb->s_magic = BTRFS_SUPER_MAGIC;
948 sb->s_op = &btrfs_super_ops;
949 sb->s_d_op = &btrfs_dentry_operations;
950 sb->s_export_op = &btrfs_export_ops;
951 #ifdef CONFIG_FS_VERITY
952 sb->s_vop = &btrfs_verityops;
953 #endif
954 sb->s_xattr = btrfs_xattr_handlers;
955 sb->s_time_gran = 1;
956 sb->s_iflags |= SB_I_CGROUPWB;
957
958 err = super_setup_bdi(sb);
959 if (err) {
960 btrfs_err(fs_info, "super_setup_bdi failed");
961 return err;
962 }
963
964 err = open_ctree(sb, fs_devices, (char *)data);
965 if (err) {
966 btrfs_err(fs_info, "open_ctree failed");
967 return err;
968 }
969
970 inode = btrfs_iget(BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
971 if (IS_ERR(inode)) {
972 err = PTR_ERR(inode);
973 btrfs_handle_fs_error(fs_info, err, NULL);
974 goto fail_close;
975 }
976
977 sb->s_root = d_make_root(inode);
978 if (!sb->s_root) {
979 err = -ENOMEM;
980 goto fail_close;
981 }
982
983 sb->s_flags |= SB_ACTIVE;
984 return 0;
985
986 fail_close:
987 close_ctree(fs_info);
988 return err;
989 }
990
991 int btrfs_sync_fs(struct super_block *sb, int wait)
992 {
993 struct btrfs_trans_handle *trans;
994 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
995 struct btrfs_root *root = fs_info->tree_root;
996
997 trace_btrfs_sync_fs(fs_info, wait);
998
999 if (!wait) {
1000 filemap_flush(fs_info->btree_inode->i_mapping);
1001 return 0;
1002 }
1003
1004 btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL);
1005
1006 trans = btrfs_attach_transaction_barrier(root);
1007 if (IS_ERR(trans)) {
1008 /* no transaction, don't bother */
1009 if (PTR_ERR(trans) == -ENOENT) {
1010 /*
1011 * Exit unless we have some pending changes
1012 * that need to go through commit
1013 */
1014 if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1015 &fs_info->flags))
1016 return 0;
1017 /*
1018 * A non-blocking test if the fs is frozen. We must not
1019 * start a new transaction here otherwise a deadlock
1020 * happens. The pending operations are delayed to the
1021 * next commit after thawing.
1022 */
1023 if (sb_start_write_trylock(sb))
1024 sb_end_write(sb);
1025 else
1026 return 0;
1027 trans = btrfs_start_transaction(root, 0);
1028 }
1029 if (IS_ERR(trans))
1030 return PTR_ERR(trans);
1031 }
1032 return btrfs_commit_transaction(trans);
1033 }
1034
1035 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1036 {
1037 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1038 *printed = true;
1039 }
1040
1041 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1042 {
1043 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1044 const char *compress_type;
1045 const char *subvol_name;
1046 bool printed = false;
1047
1048 if (btrfs_test_opt(info, DEGRADED))
1049 seq_puts(seq, ",degraded");
1050 if (btrfs_test_opt(info, NODATASUM))
1051 seq_puts(seq, ",nodatasum");
1052 if (btrfs_test_opt(info, NODATACOW))
1053 seq_puts(seq, ",nodatacow");
1054 if (btrfs_test_opt(info, NOBARRIER))
1055 seq_puts(seq, ",nobarrier");
1056 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1057 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1058 if (info->thread_pool_size != min_t(unsigned long,
1059 num_online_cpus() + 2, 8))
1060 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1061 if (btrfs_test_opt(info, COMPRESS)) {
1062 compress_type = btrfs_compress_type2str(info->compress_type);
1063 if (btrfs_test_opt(info, FORCE_COMPRESS))
1064 seq_printf(seq, ",compress-force=%s", compress_type);
1065 else
1066 seq_printf(seq, ",compress=%s", compress_type);
1067 if (info->compress_level)
1068 seq_printf(seq, ":%d", info->compress_level);
1069 }
1070 if (btrfs_test_opt(info, NOSSD))
1071 seq_puts(seq, ",nossd");
1072 if (btrfs_test_opt(info, SSD_SPREAD))
1073 seq_puts(seq, ",ssd_spread");
1074 else if (btrfs_test_opt(info, SSD))
1075 seq_puts(seq, ",ssd");
1076 if (btrfs_test_opt(info, NOTREELOG))
1077 seq_puts(seq, ",notreelog");
1078 if (btrfs_test_opt(info, NOLOGREPLAY))
1079 print_rescue_option(seq, "nologreplay", &printed);
1080 if (btrfs_test_opt(info, USEBACKUPROOT))
1081 print_rescue_option(seq, "usebackuproot", &printed);
1082 if (btrfs_test_opt(info, IGNOREBADROOTS))
1083 print_rescue_option(seq, "ignorebadroots", &printed);
1084 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1085 print_rescue_option(seq, "ignoredatacsums", &printed);
1086 if (btrfs_test_opt(info, IGNOREMETACSUMS))
1087 print_rescue_option(seq, "ignoremetacsums", &printed);
1088 if (btrfs_test_opt(info, IGNORESUPERFLAGS))
1089 print_rescue_option(seq, "ignoresuperflags", &printed);
1090 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1091 seq_puts(seq, ",flushoncommit");
1092 if (btrfs_test_opt(info, DISCARD_SYNC))
1093 seq_puts(seq, ",discard");
1094 if (btrfs_test_opt(info, DISCARD_ASYNC))
1095 seq_puts(seq, ",discard=async");
1096 if (!(info->sb->s_flags & SB_POSIXACL))
1097 seq_puts(seq, ",noacl");
1098 if (btrfs_free_space_cache_v1_active(info))
1099 seq_puts(seq, ",space_cache");
1100 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1101 seq_puts(seq, ",space_cache=v2");
1102 else
1103 seq_puts(seq, ",nospace_cache");
1104 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1105 seq_puts(seq, ",rescan_uuid_tree");
1106 if (btrfs_test_opt(info, CLEAR_CACHE))
1107 seq_puts(seq, ",clear_cache");
1108 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1109 seq_puts(seq, ",user_subvol_rm_allowed");
1110 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1111 seq_puts(seq, ",enospc_debug");
1112 if (btrfs_test_opt(info, AUTO_DEFRAG))
1113 seq_puts(seq, ",autodefrag");
1114 if (btrfs_test_opt(info, SKIP_BALANCE))
1115 seq_puts(seq, ",skip_balance");
1116 if (info->metadata_ratio)
1117 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1118 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1119 seq_puts(seq, ",fatal_errors=panic");
1120 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1121 seq_printf(seq, ",commit=%u", info->commit_interval);
1122 #ifdef CONFIG_BTRFS_DEBUG
1123 if (btrfs_test_opt(info, FRAGMENT_DATA))
1124 seq_puts(seq, ",fragment=data");
1125 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1126 seq_puts(seq, ",fragment=metadata");
1127 #endif
1128 if (btrfs_test_opt(info, REF_VERIFY))
1129 seq_puts(seq, ",ref_verify");
1130 seq_printf(seq, ",subvolid=%llu", btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1131 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1132 btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1133 if (!IS_ERR(subvol_name)) {
1134 seq_puts(seq, ",subvol=");
1135 seq_escape(seq, subvol_name, " \t\n\\");
1136 kfree(subvol_name);
1137 }
1138 return 0;
1139 }
1140
1141 /*
1142 * subvolumes are identified by ino 256
1143 */
1144 static inline int is_subvolume_inode(struct inode *inode)
1145 {
1146 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1147 return 1;
1148 return 0;
1149 }
1150
1151 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1152 struct vfsmount *mnt)
1153 {
1154 struct dentry *root;
1155 int ret;
1156
1157 if (!subvol_name) {
1158 if (!subvol_objectid) {
1159 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1160 &subvol_objectid);
1161 if (ret) {
1162 root = ERR_PTR(ret);
1163 goto out;
1164 }
1165 }
1166 subvol_name = btrfs_get_subvol_name_from_objectid(
1167 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1168 if (IS_ERR(subvol_name)) {
1169 root = ERR_CAST(subvol_name);
1170 subvol_name = NULL;
1171 goto out;
1172 }
1173
1174 }
1175
1176 root = mount_subtree(mnt, subvol_name);
1177 /* mount_subtree() drops our reference on the vfsmount. */
1178 mnt = NULL;
1179
1180 if (!IS_ERR(root)) {
1181 struct super_block *s = root->d_sb;
1182 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1183 struct inode *root_inode = d_inode(root);
1184 u64 root_objectid = btrfs_root_id(BTRFS_I(root_inode)->root);
1185
1186 ret = 0;
1187 if (!is_subvolume_inode(root_inode)) {
1188 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1189 subvol_name);
1190 ret = -EINVAL;
1191 }
1192 if (subvol_objectid && root_objectid != subvol_objectid) {
1193 /*
1194 * This will also catch a race condition where a
1195 * subvolume which was passed by ID is renamed and
1196 * another subvolume is renamed over the old location.
1197 */
1198 btrfs_err(fs_info,
1199 "subvol '%s' does not match subvolid %llu",
1200 subvol_name, subvol_objectid);
1201 ret = -EINVAL;
1202 }
1203 if (ret) {
1204 dput(root);
1205 root = ERR_PTR(ret);
1206 deactivate_locked_super(s);
1207 }
1208 }
1209
1210 out:
1211 mntput(mnt);
1212 kfree(subvol_name);
1213 return root;
1214 }
1215
1216 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1217 u32 new_pool_size, u32 old_pool_size)
1218 {
1219 if (new_pool_size == old_pool_size)
1220 return;
1221
1222 fs_info->thread_pool_size = new_pool_size;
1223
1224 btrfs_info(fs_info, "resize thread pool %d -> %d",
1225 old_pool_size, new_pool_size);
1226
1227 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1228 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1229 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1230 workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1231 workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1232 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1233 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1234 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1235 }
1236
1237 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1238 unsigned long long old_opts, int flags)
1239 {
1240 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1241 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1242 (flags & SB_RDONLY))) {
1243 /* wait for any defraggers to finish */
1244 wait_event(fs_info->transaction_wait,
1245 (atomic_read(&fs_info->defrag_running) == 0));
1246 if (flags & SB_RDONLY)
1247 sync_filesystem(fs_info->sb);
1248 }
1249 }
1250
1251 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1252 unsigned long long old_opts)
1253 {
1254 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1255
1256 /*
1257 * We need to cleanup all defragable inodes if the autodefragment is
1258 * close or the filesystem is read only.
1259 */
1260 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1261 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1262 btrfs_cleanup_defrag_inodes(fs_info);
1263 }
1264
1265 /* If we toggled discard async */
1266 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1267 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1268 btrfs_discard_resume(fs_info);
1269 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1270 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1271 btrfs_discard_cleanup(fs_info);
1272
1273 /* If we toggled space cache */
1274 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1275 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1276 }
1277
1278 static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1279 {
1280 int ret;
1281
1282 if (BTRFS_FS_ERROR(fs_info)) {
1283 btrfs_err(fs_info,
1284 "remounting read-write after error is not allowed");
1285 return -EINVAL;
1286 }
1287
1288 if (fs_info->fs_devices->rw_devices == 0)
1289 return -EACCES;
1290
1291 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1292 btrfs_warn(fs_info,
1293 "too many missing devices, writable remount is not allowed");
1294 return -EACCES;
1295 }
1296
1297 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1298 btrfs_warn(fs_info,
1299 "mount required to replay tree-log, cannot remount read-write");
1300 return -EINVAL;
1301 }
1302
1303 /*
1304 * NOTE: when remounting with a change that does writes, don't put it
1305 * anywhere above this point, as we are not sure to be safe to write
1306 * until we pass the above checks.
1307 */
1308 ret = btrfs_start_pre_rw_mount(fs_info);
1309 if (ret)
1310 return ret;
1311
1312 btrfs_clear_sb_rdonly(fs_info->sb);
1313
1314 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1315
1316 /*
1317 * If we've gone from readonly -> read-write, we need to get our
1318 * sync/async discard lists in the right state.
1319 */
1320 btrfs_discard_resume(fs_info);
1321
1322 return 0;
1323 }
1324
1325 static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1326 {
1327 /*
1328 * This also happens on 'umount -rf' or on shutdown, when the
1329 * filesystem is busy.
1330 */
1331 cancel_work_sync(&fs_info->async_reclaim_work);
1332 cancel_work_sync(&fs_info->async_data_reclaim_work);
1333
1334 btrfs_discard_cleanup(fs_info);
1335
1336 /* Wait for the uuid_scan task to finish */
1337 down(&fs_info->uuid_tree_rescan_sem);
1338 /* Avoid complains from lockdep et al. */
1339 up(&fs_info->uuid_tree_rescan_sem);
1340
1341 btrfs_set_sb_rdonly(fs_info->sb);
1342
1343 /*
1344 * Setting SB_RDONLY will put the cleaner thread to sleep at the next
1345 * loop if it's already active. If it's already asleep, we'll leave
1346 * unused block groups on disk until we're mounted read-write again
1347 * unless we clean them up here.
1348 */
1349 btrfs_delete_unused_bgs(fs_info);
1350
1351 /*
1352 * The cleaner task could be already running before we set the flag
1353 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock). We must make
1354 * sure that after we finish the remount, i.e. after we call
1355 * btrfs_commit_super(), the cleaner can no longer start a transaction
1356 * - either because it was dropping a dead root, running delayed iputs
1357 * or deleting an unused block group (the cleaner picked a block
1358 * group from the list of unused block groups before we were able to
1359 * in the previous call to btrfs_delete_unused_bgs()).
1360 */
1361 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1362
1363 /*
1364 * We've set the superblock to RO mode, so we might have made the
1365 * cleaner task sleep without running all pending delayed iputs. Go
1366 * through all the delayed iputs here, so that if an unmount happens
1367 * without remounting RW we don't end up at finishing close_ctree()
1368 * with a non-empty list of delayed iputs.
1369 */
1370 btrfs_run_delayed_iputs(fs_info);
1371
1372 btrfs_dev_replace_suspend_for_unmount(fs_info);
1373 btrfs_scrub_cancel(fs_info);
1374 btrfs_pause_balance(fs_info);
1375
1376 /*
1377 * Pause the qgroup rescan worker if it is running. We don't want it to
1378 * be still running after we are in RO mode, as after that, by the time
1379 * we unmount, it might have left a transaction open, so we would leak
1380 * the transaction and/or crash.
1381 */
1382 btrfs_qgroup_wait_for_completion(fs_info, false);
1383
1384 return btrfs_commit_super(fs_info);
1385 }
1386
1387 static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1388 {
1389 fs_info->max_inline = ctx->max_inline;
1390 fs_info->commit_interval = ctx->commit_interval;
1391 fs_info->metadata_ratio = ctx->metadata_ratio;
1392 fs_info->thread_pool_size = ctx->thread_pool_size;
1393 fs_info->mount_opt = ctx->mount_opt;
1394 fs_info->compress_type = ctx->compress_type;
1395 fs_info->compress_level = ctx->compress_level;
1396 }
1397
1398 static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1399 {
1400 ctx->max_inline = fs_info->max_inline;
1401 ctx->commit_interval = fs_info->commit_interval;
1402 ctx->metadata_ratio = fs_info->metadata_ratio;
1403 ctx->thread_pool_size = fs_info->thread_pool_size;
1404 ctx->mount_opt = fs_info->mount_opt;
1405 ctx->compress_type = fs_info->compress_type;
1406 ctx->compress_level = fs_info->compress_level;
1407 }
1408
1409 #define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...) \
1410 do { \
1411 if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1412 btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1413 btrfs_info(fs_info, fmt, ##args); \
1414 } while (0)
1415
1416 #define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...) \
1417 do { \
1418 if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1419 !btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1420 btrfs_info(fs_info, fmt, ##args); \
1421 } while (0)
1422
1423 static void btrfs_emit_options(struct btrfs_fs_info *info,
1424 struct btrfs_fs_context *old)
1425 {
1426 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1427 btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1428 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1429 btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1430 btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1431 btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1432 btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1433 btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1434 btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1435 btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1436 btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1437 btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1438 btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1439 btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1440 btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1441 btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1442 btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1443 btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1444 btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1445 btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1446 btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
1447 btrfs_info_if_set(info, old, IGNOREMETACSUMS, "ignoring meta csums");
1448 btrfs_info_if_set(info, old, IGNORESUPERFLAGS, "ignoring unknown super block flags");
1449
1450 btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1451 btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1452 btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1453 btrfs_info_if_unset(info, old, NOBARRIER, "turning off barriers");
1454 btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1455 btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1456 btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1457 btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1458 btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1459
1460 /* Did the compression settings change? */
1461 if (btrfs_test_opt(info, COMPRESS) &&
1462 (!old ||
1463 old->compress_type != info->compress_type ||
1464 old->compress_level != info->compress_level ||
1465 (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1466 btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1467 const char *compress_type = btrfs_compress_type2str(info->compress_type);
1468
1469 btrfs_info(info, "%s %s compression, level %d",
1470 btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1471 compress_type, info->compress_level);
1472 }
1473
1474 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1475 btrfs_info(info, "max_inline set to %llu", info->max_inline);
1476 }
1477
1478 static int btrfs_reconfigure(struct fs_context *fc)
1479 {
1480 struct super_block *sb = fc->root->d_sb;
1481 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1482 struct btrfs_fs_context *ctx = fc->fs_private;
1483 struct btrfs_fs_context old_ctx;
1484 int ret = 0;
1485 bool mount_reconfigure = (fc->s_fs_info != NULL);
1486
1487 btrfs_info_to_ctx(fs_info, &old_ctx);
1488
1489 /*
1490 * This is our "bind mount" trick, we don't want to allow the user to do
1491 * anything other than mount a different ro/rw and a different subvol,
1492 * all of the mount options should be maintained.
1493 */
1494 if (mount_reconfigure)
1495 ctx->mount_opt = old_ctx.mount_opt;
1496
1497 sync_filesystem(sb);
1498 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1499
1500 if (!mount_reconfigure &&
1501 !btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags))
1502 return -EINVAL;
1503
1504 ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
1505 if (ret < 0)
1506 return ret;
1507
1508 btrfs_ctx_to_info(fs_info, ctx);
1509 btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags);
1510 btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size,
1511 old_ctx.thread_pool_size);
1512
1513 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1514 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1515 (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1516 btrfs_warn(fs_info,
1517 "remount supports changing free space tree only from RO to RW");
1518 /* Make sure free space cache options match the state on disk. */
1519 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1520 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1521 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1522 }
1523 if (btrfs_free_space_cache_v1_active(fs_info)) {
1524 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1525 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1526 }
1527 }
1528
1529 ret = 0;
1530 if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1531 ret = btrfs_remount_ro(fs_info);
1532 else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1533 ret = btrfs_remount_rw(fs_info);
1534 if (ret)
1535 goto restore;
1536
1537 /*
1538 * If we set the mask during the parameter parsing VFS would reject the
1539 * remount. Here we can set the mask and the value will be updated
1540 * appropriately.
1541 */
1542 if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1543 fc->sb_flags_mask |= SB_POSIXACL;
1544
1545 btrfs_emit_options(fs_info, &old_ctx);
1546 wake_up_process(fs_info->transaction_kthread);
1547 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1548 btrfs_clear_oneshot_options(fs_info);
1549 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1550
1551 return 0;
1552 restore:
1553 btrfs_ctx_to_info(fs_info, &old_ctx);
1554 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1555 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1556 return ret;
1557 }
1558
1559 /* Used to sort the devices by max_avail(descending sort) */
1560 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1561 {
1562 const struct btrfs_device_info *dev_info1 = a;
1563 const struct btrfs_device_info *dev_info2 = b;
1564
1565 if (dev_info1->max_avail > dev_info2->max_avail)
1566 return -1;
1567 else if (dev_info1->max_avail < dev_info2->max_avail)
1568 return 1;
1569 return 0;
1570 }
1571
1572 /*
1573 * sort the devices by max_avail, in which max free extent size of each device
1574 * is stored.(Descending Sort)
1575 */
1576 static inline void btrfs_descending_sort_devices(
1577 struct btrfs_device_info *devices,
1578 size_t nr_devices)
1579 {
1580 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1581 btrfs_cmp_device_free_bytes, NULL);
1582 }
1583
1584 /*
1585 * The helper to calc the free space on the devices that can be used to store
1586 * file data.
1587 */
1588 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1589 u64 *free_bytes)
1590 {
1591 struct btrfs_device_info *devices_info;
1592 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1593 struct btrfs_device *device;
1594 u64 type;
1595 u64 avail_space;
1596 u64 min_stripe_size;
1597 int num_stripes = 1;
1598 int i = 0, nr_devices;
1599 const struct btrfs_raid_attr *rattr;
1600
1601 /*
1602 * We aren't under the device list lock, so this is racy-ish, but good
1603 * enough for our purposes.
1604 */
1605 nr_devices = fs_info->fs_devices->open_devices;
1606 if (!nr_devices) {
1607 smp_mb();
1608 nr_devices = fs_info->fs_devices->open_devices;
1609 ASSERT(nr_devices);
1610 if (!nr_devices) {
1611 *free_bytes = 0;
1612 return 0;
1613 }
1614 }
1615
1616 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1617 GFP_KERNEL);
1618 if (!devices_info)
1619 return -ENOMEM;
1620
1621 /* calc min stripe number for data space allocation */
1622 type = btrfs_data_alloc_profile(fs_info);
1623 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1624
1625 if (type & BTRFS_BLOCK_GROUP_RAID0)
1626 num_stripes = nr_devices;
1627 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1628 num_stripes = rattr->ncopies;
1629 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1630 num_stripes = 4;
1631
1632 /* Adjust for more than 1 stripe per device */
1633 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1634
1635 rcu_read_lock();
1636 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1637 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1638 &device->dev_state) ||
1639 !device->bdev ||
1640 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1641 continue;
1642
1643 if (i >= nr_devices)
1644 break;
1645
1646 avail_space = device->total_bytes - device->bytes_used;
1647
1648 /* align with stripe_len */
1649 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1650
1651 /*
1652 * Ensure we have at least min_stripe_size on top of the
1653 * reserved space on the device.
1654 */
1655 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1656 continue;
1657
1658 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1659
1660 devices_info[i].dev = device;
1661 devices_info[i].max_avail = avail_space;
1662
1663 i++;
1664 }
1665 rcu_read_unlock();
1666
1667 nr_devices = i;
1668
1669 btrfs_descending_sort_devices(devices_info, nr_devices);
1670
1671 i = nr_devices - 1;
1672 avail_space = 0;
1673 while (nr_devices >= rattr->devs_min) {
1674 num_stripes = min(num_stripes, nr_devices);
1675
1676 if (devices_info[i].max_avail >= min_stripe_size) {
1677 int j;
1678 u64 alloc_size;
1679
1680 avail_space += devices_info[i].max_avail * num_stripes;
1681 alloc_size = devices_info[i].max_avail;
1682 for (j = i + 1 - num_stripes; j <= i; j++)
1683 devices_info[j].max_avail -= alloc_size;
1684 }
1685 i--;
1686 nr_devices--;
1687 }
1688
1689 kfree(devices_info);
1690 *free_bytes = avail_space;
1691 return 0;
1692 }
1693
1694 /*
1695 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1696 *
1697 * If there's a redundant raid level at DATA block groups, use the respective
1698 * multiplier to scale the sizes.
1699 *
1700 * Unused device space usage is based on simulating the chunk allocator
1701 * algorithm that respects the device sizes and order of allocations. This is
1702 * a close approximation of the actual use but there are other factors that may
1703 * change the result (like a new metadata chunk).
1704 *
1705 * If metadata is exhausted, f_bavail will be 0.
1706 */
1707 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1708 {
1709 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1710 struct btrfs_super_block *disk_super = fs_info->super_copy;
1711 struct btrfs_space_info *found;
1712 u64 total_used = 0;
1713 u64 total_free_data = 0;
1714 u64 total_free_meta = 0;
1715 u32 bits = fs_info->sectorsize_bits;
1716 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1717 unsigned factor = 1;
1718 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1719 int ret;
1720 u64 thresh = 0;
1721 int mixed = 0;
1722
1723 list_for_each_entry(found, &fs_info->space_info, list) {
1724 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1725 int i;
1726
1727 total_free_data += found->disk_total - found->disk_used;
1728 total_free_data -=
1729 btrfs_account_ro_block_groups_free_space(found);
1730
1731 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1732 if (!list_empty(&found->block_groups[i]))
1733 factor = btrfs_bg_type_to_factor(
1734 btrfs_raid_array[i].bg_flag);
1735 }
1736 }
1737
1738 /*
1739 * Metadata in mixed block group profiles are accounted in data
1740 */
1741 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1742 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1743 mixed = 1;
1744 else
1745 total_free_meta += found->disk_total -
1746 found->disk_used;
1747 }
1748
1749 total_used += found->disk_used;
1750 }
1751
1752 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1753 buf->f_blocks >>= bits;
1754 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1755
1756 /* Account global block reserve as used, it's in logical size already */
1757 spin_lock(&block_rsv->lock);
1758 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
1759 if (buf->f_bfree >= block_rsv->size >> bits)
1760 buf->f_bfree -= block_rsv->size >> bits;
1761 else
1762 buf->f_bfree = 0;
1763 spin_unlock(&block_rsv->lock);
1764
1765 buf->f_bavail = div_u64(total_free_data, factor);
1766 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1767 if (ret)
1768 return ret;
1769 buf->f_bavail += div_u64(total_free_data, factor);
1770 buf->f_bavail = buf->f_bavail >> bits;
1771
1772 /*
1773 * We calculate the remaining metadata space minus global reserve. If
1774 * this is (supposedly) smaller than zero, there's no space. But this
1775 * does not hold in practice, the exhausted state happens where's still
1776 * some positive delta. So we apply some guesswork and compare the
1777 * delta to a 4M threshold. (Practically observed delta was ~2M.)
1778 *
1779 * We probably cannot calculate the exact threshold value because this
1780 * depends on the internal reservations requested by various
1781 * operations, so some operations that consume a few metadata will
1782 * succeed even if the Avail is zero. But this is better than the other
1783 * way around.
1784 */
1785 thresh = SZ_4M;
1786
1787 /*
1788 * We only want to claim there's no available space if we can no longer
1789 * allocate chunks for our metadata profile and our global reserve will
1790 * not fit in the free metadata space. If we aren't ->full then we
1791 * still can allocate chunks and thus are fine using the currently
1792 * calculated f_bavail.
1793 */
1794 if (!mixed && block_rsv->space_info->full &&
1795 (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1796 buf->f_bavail = 0;
1797
1798 buf->f_type = BTRFS_SUPER_MAGIC;
1799 buf->f_bsize = fs_info->sectorsize;
1800 buf->f_namelen = BTRFS_NAME_LEN;
1801
1802 /* We treat it as constant endianness (it doesn't matter _which_)
1803 because we want the fsid to come out the same whether mounted
1804 on a big-endian or little-endian host */
1805 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1806 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1807 /* Mask in the root object ID too, to disambiguate subvols */
1808 buf->f_fsid.val[0] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root) >> 32;
1809 buf->f_fsid.val[1] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root);
1810
1811 return 0;
1812 }
1813
1814 static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1815 {
1816 struct btrfs_fs_info *p = fc->s_fs_info;
1817 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1818
1819 return fs_info->fs_devices == p->fs_devices;
1820 }
1821
1822 static int btrfs_get_tree_super(struct fs_context *fc)
1823 {
1824 struct btrfs_fs_info *fs_info = fc->s_fs_info;
1825 struct btrfs_fs_context *ctx = fc->fs_private;
1826 struct btrfs_fs_devices *fs_devices = NULL;
1827 struct block_device *bdev;
1828 struct btrfs_device *device;
1829 struct super_block *sb;
1830 blk_mode_t mode = btrfs_open_mode(fc);
1831 int ret;
1832
1833 btrfs_ctx_to_info(fs_info, ctx);
1834 mutex_lock(&uuid_mutex);
1835
1836 /*
1837 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1838 * either a valid device or an error.
1839 */
1840 device = btrfs_scan_one_device(fc->source, mode, true);
1841 ASSERT(device != NULL);
1842 if (IS_ERR(device)) {
1843 mutex_unlock(&uuid_mutex);
1844 return PTR_ERR(device);
1845 }
1846
1847 fs_devices = device->fs_devices;
1848 fs_info->fs_devices = fs_devices;
1849
1850 ret = btrfs_open_devices(fs_devices, mode, &btrfs_fs_type);
1851 mutex_unlock(&uuid_mutex);
1852 if (ret)
1853 return ret;
1854
1855 if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1856 ret = -EACCES;
1857 goto error;
1858 }
1859
1860 bdev = fs_devices->latest_dev->bdev;
1861
1862 /*
1863 * From now on the error handling is not straightforward.
1864 *
1865 * If successful, this will transfer the fs_info into the super block,
1866 * and fc->s_fs_info will be NULL. However if there's an existing
1867 * super, we'll still have fc->s_fs_info populated. If we error
1868 * completely out it'll be cleaned up when we drop the fs_context,
1869 * otherwise it's tied to the lifetime of the super_block.
1870 */
1871 sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1872 if (IS_ERR(sb)) {
1873 ret = PTR_ERR(sb);
1874 goto error;
1875 }
1876
1877 set_device_specific_options(fs_info);
1878
1879 if (sb->s_root) {
1880 btrfs_close_devices(fs_devices);
1881 if ((fc->sb_flags ^ sb->s_flags) & SB_RDONLY)
1882 ret = -EBUSY;
1883 } else {
1884 snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
1885 shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id);
1886 btrfs_sb(sb)->bdev_holder = &btrfs_fs_type;
1887 ret = btrfs_fill_super(sb, fs_devices, NULL);
1888 }
1889
1890 if (ret) {
1891 deactivate_locked_super(sb);
1892 return ret;
1893 }
1894
1895 btrfs_clear_oneshot_options(fs_info);
1896
1897 fc->root = dget(sb->s_root);
1898 return 0;
1899
1900 error:
1901 btrfs_close_devices(fs_devices);
1902 return ret;
1903 }
1904
1905 /*
1906 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1907 * with different ro/rw options") the following works:
1908 *
1909 * (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1910 * (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1911 *
1912 * which looks nice and innocent but is actually pretty intricate and deserves
1913 * a long comment.
1914 *
1915 * On another filesystem a subvolume mount is close to something like:
1916 *
1917 * (iii) # create rw superblock + initial mount
1918 * mount -t xfs /dev/sdb /opt/
1919 *
1920 * # create ro bind mount
1921 * mount --bind -o ro /opt/foo /mnt/foo
1922 *
1923 * # unmount initial mount
1924 * umount /opt
1925 *
1926 * Of course, there's some special subvolume sauce and there's the fact that the
1927 * sb->s_root dentry is really swapped after mount_subtree(). But conceptually
1928 * it's very close and will help us understand the issue.
1929 *
1930 * The old mount API didn't cleanly distinguish between a mount being made ro
1931 * and a superblock being made ro. The only way to change the ro state of
1932 * either object was by passing ms_rdonly. If a new mount was created via
1933 * mount(2) such as:
1934 *
1935 * mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1936 *
1937 * the MS_RDONLY flag being specified had two effects:
1938 *
1939 * (1) MNT_READONLY was raised -> the resulting mount got
1940 * @mnt->mnt_flags |= MNT_READONLY raised.
1941 *
1942 * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
1943 * made the superblock ro. Note, how SB_RDONLY has the same value as
1944 * ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
1945 *
1946 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
1947 * subtree mounted ro.
1948 *
1949 * But consider the effect on the old mount API on btrfs subvolume mounting
1950 * which combines the distinct step in (iii) into a single step.
1951 *
1952 * By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
1953 * is issued the superblock is ro and thus even if the mount created for (ii) is
1954 * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
1955 * to rw for (ii) which it did using an internal remount call.
1956 *
1957 * IOW, subvolume mounting was inherently complicated due to the ambiguity of
1958 * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
1959 * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
1960 * passed by mount(8) to mount(2).
1961 *
1962 * Enter the new mount API. The new mount API disambiguates making a mount ro
1963 * and making a superblock ro.
1964 *
1965 * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
1966 * fsmount() or mount_setattr() this is a pure VFS level change for a
1967 * specific mount or mount tree that is never seen by the filesystem itself.
1968 *
1969 * (4) To turn a superblock ro the "ro" flag must be used with
1970 * fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
1971 * in fc->sb_flags.
1972 *
1973 * This disambiguation has rather positive consequences. Mounting a subvolume
1974 * ro will not also turn the superblock ro. Only the mount for the subvolume
1975 * will become ro.
1976 *
1977 * So, if the superblock creation request comes from the new mount API the
1978 * caller must have explicitly done:
1979 *
1980 * fsconfig(FSCONFIG_SET_FLAG, "ro")
1981 * fsmount/mount_setattr(MOUNT_ATTR_RDONLY)
1982 *
1983 * IOW, at some point the caller must have explicitly turned the whole
1984 * superblock ro and we shouldn't just undo it like we did for the old mount
1985 * API. In any case, it lets us avoid the hack in the new mount API.
1986 *
1987 * Consequently, the remounting hack must only be used for requests originating
1988 * from the old mount API and should be marked for full deprecation so it can be
1989 * turned off in a couple of years.
1990 *
1991 * The new mount API has no reason to support this hack.
1992 */
1993 static struct vfsmount *btrfs_reconfigure_for_mount(struct fs_context *fc)
1994 {
1995 struct vfsmount *mnt;
1996 int ret;
1997 const bool ro2rw = !(fc->sb_flags & SB_RDONLY);
1998
1999 /*
2000 * We got an EBUSY because our SB_RDONLY flag didn't match the existing
2001 * super block, so invert our setting here and retry the mount so we
2002 * can get our vfsmount.
2003 */
2004 if (ro2rw)
2005 fc->sb_flags |= SB_RDONLY;
2006 else
2007 fc->sb_flags &= ~SB_RDONLY;
2008
2009 mnt = fc_mount(fc);
2010 if (IS_ERR(mnt))
2011 return mnt;
2012
2013 if (!fc->oldapi || !ro2rw)
2014 return mnt;
2015
2016 /* We need to convert to rw, call reconfigure. */
2017 fc->sb_flags &= ~SB_RDONLY;
2018 down_write(&mnt->mnt_sb->s_umount);
2019 ret = btrfs_reconfigure(fc);
2020 up_write(&mnt->mnt_sb->s_umount);
2021 if (ret) {
2022 mntput(mnt);
2023 return ERR_PTR(ret);
2024 }
2025 return mnt;
2026 }
2027
2028 static int btrfs_get_tree_subvol(struct fs_context *fc)
2029 {
2030 struct btrfs_fs_info *fs_info = NULL;
2031 struct btrfs_fs_context *ctx = fc->fs_private;
2032 struct fs_context *dup_fc;
2033 struct dentry *dentry;
2034 struct vfsmount *mnt;
2035
2036 /*
2037 * Setup a dummy root and fs_info for test/set super. This is because
2038 * we don't actually fill this stuff out until open_ctree, but we need
2039 * then open_ctree will properly initialize the file system specific
2040 * settings later. btrfs_init_fs_info initializes the static elements
2041 * of the fs_info (locks and such) to make cleanup easier if we find a
2042 * superblock with our given fs_devices later on at sget() time.
2043 */
2044 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
2045 if (!fs_info)
2046 return -ENOMEM;
2047
2048 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2049 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2050 if (!fs_info->super_copy || !fs_info->super_for_commit) {
2051 btrfs_free_fs_info(fs_info);
2052 return -ENOMEM;
2053 }
2054 btrfs_init_fs_info(fs_info);
2055
2056 dup_fc = vfs_dup_fs_context(fc);
2057 if (IS_ERR(dup_fc)) {
2058 btrfs_free_fs_info(fs_info);
2059 return PTR_ERR(dup_fc);
2060 }
2061
2062 /*
2063 * When we do the sget_fc this gets transferred to the sb, so we only
2064 * need to set it on the dup_fc as this is what creates the super block.
2065 */
2066 dup_fc->s_fs_info = fs_info;
2067
2068 /*
2069 * We'll do the security settings in our btrfs_get_tree_super() mount
2070 * loop, they were duplicated into dup_fc, we can drop the originals
2071 * here.
2072 */
2073 security_free_mnt_opts(&fc->security);
2074 fc->security = NULL;
2075
2076 mnt = fc_mount(dup_fc);
2077 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY)
2078 mnt = btrfs_reconfigure_for_mount(dup_fc);
2079 put_fs_context(dup_fc);
2080 if (IS_ERR(mnt))
2081 return PTR_ERR(mnt);
2082
2083 /*
2084 * This free's ->subvol_name, because if it isn't set we have to
2085 * allocate a buffer to hold the subvol_name, so we just drop our
2086 * reference to it here.
2087 */
2088 dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2089 ctx->subvol_name = NULL;
2090 if (IS_ERR(dentry))
2091 return PTR_ERR(dentry);
2092
2093 fc->root = dentry;
2094 return 0;
2095 }
2096
2097 static int btrfs_get_tree(struct fs_context *fc)
2098 {
2099 /*
2100 * Since we use mount_subtree to mount the default/specified subvol, we
2101 * have to do mounts in two steps.
2102 *
2103 * First pass through we call btrfs_get_tree_subvol(), this is just a
2104 * wrapper around fc_mount() to call back into here again, and this time
2105 * we'll call btrfs_get_tree_super(). This will do the open_ctree() and
2106 * everything to open the devices and file system. Then we return back
2107 * with a fully constructed vfsmount in btrfs_get_tree_subvol(), and
2108 * from there we can do our mount_subvol() call, which will lookup
2109 * whichever subvol we're mounting and setup this fc with the
2110 * appropriate dentry for the subvol.
2111 */
2112 if (fc->s_fs_info)
2113 return btrfs_get_tree_super(fc);
2114 return btrfs_get_tree_subvol(fc);
2115 }
2116
2117 static void btrfs_kill_super(struct super_block *sb)
2118 {
2119 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2120 kill_anon_super(sb);
2121 btrfs_free_fs_info(fs_info);
2122 }
2123
2124 static void btrfs_free_fs_context(struct fs_context *fc)
2125 {
2126 struct btrfs_fs_context *ctx = fc->fs_private;
2127 struct btrfs_fs_info *fs_info = fc->s_fs_info;
2128
2129 if (fs_info)
2130 btrfs_free_fs_info(fs_info);
2131
2132 if (ctx && refcount_dec_and_test(&ctx->refs)) {
2133 kfree(ctx->subvol_name);
2134 kfree(ctx);
2135 }
2136 }
2137
2138 static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2139 {
2140 struct btrfs_fs_context *ctx = src_fc->fs_private;
2141
2142 /*
2143 * Give a ref to our ctx to this dup, as we want to keep it around for
2144 * our original fc so we can have the subvolume name or objectid.
2145 *
2146 * We unset ->source in the original fc because the dup needs it for
2147 * mounting, and then once we free the dup it'll free ->source, so we
2148 * need to make sure we're only pointing to it in one fc.
2149 */
2150 refcount_inc(&ctx->refs);
2151 fc->fs_private = ctx;
2152 fc->source = src_fc->source;
2153 src_fc->source = NULL;
2154 return 0;
2155 }
2156
2157 static const struct fs_context_operations btrfs_fs_context_ops = {
2158 .parse_param = btrfs_parse_param,
2159 .reconfigure = btrfs_reconfigure,
2160 .get_tree = btrfs_get_tree,
2161 .dup = btrfs_dup_fs_context,
2162 .free = btrfs_free_fs_context,
2163 };
2164
2165 static int btrfs_init_fs_context(struct fs_context *fc)
2166 {
2167 struct btrfs_fs_context *ctx;
2168
2169 ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL);
2170 if (!ctx)
2171 return -ENOMEM;
2172
2173 refcount_set(&ctx->refs, 1);
2174 fc->fs_private = ctx;
2175 fc->ops = &btrfs_fs_context_ops;
2176
2177 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2178 btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
2179 } else {
2180 ctx->thread_pool_size =
2181 min_t(unsigned long, num_online_cpus() + 2, 8);
2182 ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2183 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2184 }
2185
2186 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
2187 fc->sb_flags |= SB_POSIXACL;
2188 #endif
2189 fc->sb_flags |= SB_I_VERSION;
2190
2191 return 0;
2192 }
2193
2194 static struct file_system_type btrfs_fs_type = {
2195 .owner = THIS_MODULE,
2196 .name = "btrfs",
2197 .init_fs_context = btrfs_init_fs_context,
2198 .parameters = btrfs_fs_parameters,
2199 .kill_sb = btrfs_kill_super,
2200 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2201 };
2202
2203 MODULE_ALIAS_FS("btrfs");
2204
2205 static int btrfs_control_open(struct inode *inode, struct file *file)
2206 {
2207 /*
2208 * The control file's private_data is used to hold the
2209 * transaction when it is started and is used to keep
2210 * track of whether a transaction is already in progress.
2211 */
2212 file->private_data = NULL;
2213 return 0;
2214 }
2215
2216 /*
2217 * Used by /dev/btrfs-control for devices ioctls.
2218 */
2219 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2220 unsigned long arg)
2221 {
2222 struct btrfs_ioctl_vol_args *vol;
2223 struct btrfs_device *device = NULL;
2224 dev_t devt = 0;
2225 int ret = -ENOTTY;
2226
2227 if (!capable(CAP_SYS_ADMIN))
2228 return -EPERM;
2229
2230 vol = memdup_user((void __user *)arg, sizeof(*vol));
2231 if (IS_ERR(vol))
2232 return PTR_ERR(vol);
2233 ret = btrfs_check_ioctl_vol_args_path(vol);
2234 if (ret < 0)
2235 goto out;
2236
2237 switch (cmd) {
2238 case BTRFS_IOC_SCAN_DEV:
2239 mutex_lock(&uuid_mutex);
2240 /*
2241 * Scanning outside of mount can return NULL which would turn
2242 * into 0 error code.
2243 */
2244 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2245 ret = PTR_ERR_OR_ZERO(device);
2246 mutex_unlock(&uuid_mutex);
2247 break;
2248 case BTRFS_IOC_FORGET_DEV:
2249 if (vol->name[0] != 0) {
2250 ret = lookup_bdev(vol->name, &devt);
2251 if (ret)
2252 break;
2253 }
2254 ret = btrfs_forget_devices(devt);
2255 break;
2256 case BTRFS_IOC_DEVICES_READY:
2257 mutex_lock(&uuid_mutex);
2258 /*
2259 * Scanning outside of mount can return NULL which would turn
2260 * into 0 error code.
2261 */
2262 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2263 if (IS_ERR_OR_NULL(device)) {
2264 mutex_unlock(&uuid_mutex);
2265 ret = PTR_ERR(device);
2266 break;
2267 }
2268 ret = !(device->fs_devices->num_devices ==
2269 device->fs_devices->total_devices);
2270 mutex_unlock(&uuid_mutex);
2271 break;
2272 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2273 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2274 break;
2275 }
2276
2277 out:
2278 kfree(vol);
2279 return ret;
2280 }
2281
2282 static int btrfs_freeze(struct super_block *sb)
2283 {
2284 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2285
2286 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2287 /*
2288 * We don't need a barrier here, we'll wait for any transaction that
2289 * could be in progress on other threads (and do delayed iputs that
2290 * we want to avoid on a frozen filesystem), or do the commit
2291 * ourselves.
2292 */
2293 return btrfs_commit_current_transaction(fs_info->tree_root);
2294 }
2295
2296 static int check_dev_super(struct btrfs_device *dev)
2297 {
2298 struct btrfs_fs_info *fs_info = dev->fs_info;
2299 struct btrfs_super_block *sb;
2300 u64 last_trans;
2301 u16 csum_type;
2302 int ret = 0;
2303
2304 /* This should be called with fs still frozen. */
2305 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2306
2307 /* Missing dev, no need to check. */
2308 if (!dev->bdev)
2309 return 0;
2310
2311 /* Only need to check the primary super block. */
2312 sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2313 if (IS_ERR(sb))
2314 return PTR_ERR(sb);
2315
2316 /* Verify the checksum. */
2317 csum_type = btrfs_super_csum_type(sb);
2318 if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2319 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2320 csum_type, btrfs_super_csum_type(fs_info->super_copy));
2321 ret = -EUCLEAN;
2322 goto out;
2323 }
2324
2325 if (btrfs_check_super_csum(fs_info, sb)) {
2326 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2327 ret = -EUCLEAN;
2328 goto out;
2329 }
2330
2331 /* Btrfs_validate_super() includes fsid check against super->fsid. */
2332 ret = btrfs_validate_super(fs_info, sb, 0);
2333 if (ret < 0)
2334 goto out;
2335
2336 last_trans = btrfs_get_last_trans_committed(fs_info);
2337 if (btrfs_super_generation(sb) != last_trans) {
2338 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2339 btrfs_super_generation(sb), last_trans);
2340 ret = -EUCLEAN;
2341 goto out;
2342 }
2343 out:
2344 btrfs_release_disk_super(sb);
2345 return ret;
2346 }
2347
2348 static int btrfs_unfreeze(struct super_block *sb)
2349 {
2350 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2351 struct btrfs_device *device;
2352 int ret = 0;
2353
2354 /*
2355 * Make sure the fs is not changed by accident (like hibernation then
2356 * modified by other OS).
2357 * If we found anything wrong, we mark the fs error immediately.
2358 *
2359 * And since the fs is frozen, no one can modify the fs yet, thus
2360 * we don't need to hold device_list_mutex.
2361 */
2362 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2363 ret = check_dev_super(device);
2364 if (ret < 0) {
2365 btrfs_handle_fs_error(fs_info, ret,
2366 "super block on devid %llu got modified unexpectedly",
2367 device->devid);
2368 break;
2369 }
2370 }
2371 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2372
2373 /*
2374 * We still return 0, to allow VFS layer to unfreeze the fs even the
2375 * above checks failed. Since the fs is either fine or read-only, we're
2376 * safe to continue, without causing further damage.
2377 */
2378 return 0;
2379 }
2380
2381 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2382 {
2383 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2384
2385 /*
2386 * There should be always a valid pointer in latest_dev, it may be stale
2387 * for a short moment in case it's being deleted but still valid until
2388 * the end of RCU grace period.
2389 */
2390 rcu_read_lock();
2391 seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2392 rcu_read_unlock();
2393
2394 return 0;
2395 }
2396
2397 static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_control *sc)
2398 {
2399 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2400 const s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps);
2401
2402 trace_btrfs_extent_map_shrinker_count(fs_info, nr);
2403
2404 return nr;
2405 }
2406
2407 static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_control *sc)
2408 {
2409 const long nr_to_scan = min_t(unsigned long, LONG_MAX, sc->nr_to_scan);
2410 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2411
2412 return btrfs_free_extent_maps(fs_info, nr_to_scan);
2413 }
2414
2415 static const struct super_operations btrfs_super_ops = {
2416 .drop_inode = btrfs_drop_inode,
2417 .evict_inode = btrfs_evict_inode,
2418 .put_super = btrfs_put_super,
2419 .sync_fs = btrfs_sync_fs,
2420 .show_options = btrfs_show_options,
2421 .show_devname = btrfs_show_devname,
2422 .alloc_inode = btrfs_alloc_inode,
2423 .destroy_inode = btrfs_destroy_inode,
2424 .free_inode = btrfs_free_inode,
2425 .statfs = btrfs_statfs,
2426 .freeze_fs = btrfs_freeze,
2427 .unfreeze_fs = btrfs_unfreeze,
2428 .nr_cached_objects = btrfs_nr_cached_objects,
2429 .free_cached_objects = btrfs_free_cached_objects,
2430 };
2431
2432 static const struct file_operations btrfs_ctl_fops = {
2433 .open = btrfs_control_open,
2434 .unlocked_ioctl = btrfs_control_ioctl,
2435 .compat_ioctl = compat_ptr_ioctl,
2436 .owner = THIS_MODULE,
2437 .llseek = noop_llseek,
2438 };
2439
2440 static struct miscdevice btrfs_misc = {
2441 .minor = BTRFS_MINOR,
2442 .name = "btrfs-control",
2443 .fops = &btrfs_ctl_fops
2444 };
2445
2446 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2447 MODULE_ALIAS("devname:btrfs-control");
2448
2449 static int __init btrfs_interface_init(void)
2450 {
2451 return misc_register(&btrfs_misc);
2452 }
2453
2454 static __cold void btrfs_interface_exit(void)
2455 {
2456 misc_deregister(&btrfs_misc);
2457 }
2458
2459 static int __init btrfs_print_mod_info(void)
2460 {
2461 static const char options[] = ""
2462 #ifdef CONFIG_BTRFS_DEBUG
2463 ", debug=on"
2464 #endif
2465 #ifdef CONFIG_BTRFS_ASSERT
2466 ", assert=on"
2467 #endif
2468 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2469 ", ref-verify=on"
2470 #endif
2471 #ifdef CONFIG_BLK_DEV_ZONED
2472 ", zoned=yes"
2473 #else
2474 ", zoned=no"
2475 #endif
2476 #ifdef CONFIG_FS_VERITY
2477 ", fsverity=yes"
2478 #else
2479 ", fsverity=no"
2480 #endif
2481 ;
2482 pr_info("Btrfs loaded%s\n", options);
2483 return 0;
2484 }
2485
2486 static int register_btrfs(void)
2487 {
2488 return register_filesystem(&btrfs_fs_type);
2489 }
2490
2491 static void unregister_btrfs(void)
2492 {
2493 unregister_filesystem(&btrfs_fs_type);
2494 }
2495
2496 /* Helper structure for long init/exit functions. */
2497 struct init_sequence {
2498 int (*init_func)(void);
2499 /* Can be NULL if the init_func doesn't need cleanup. */
2500 void (*exit_func)(void);
2501 };
2502
2503 static const struct init_sequence mod_init_seq[] = {
2504 {
2505 .init_func = btrfs_props_init,
2506 .exit_func = NULL,
2507 }, {
2508 .init_func = btrfs_init_sysfs,
2509 .exit_func = btrfs_exit_sysfs,
2510 }, {
2511 .init_func = btrfs_init_compress,
2512 .exit_func = btrfs_exit_compress,
2513 }, {
2514 .init_func = btrfs_init_cachep,
2515 .exit_func = btrfs_destroy_cachep,
2516 }, {
2517 .init_func = btrfs_init_dio,
2518 .exit_func = btrfs_destroy_dio,
2519 }, {
2520 .init_func = btrfs_transaction_init,
2521 .exit_func = btrfs_transaction_exit,
2522 }, {
2523 .init_func = btrfs_ctree_init,
2524 .exit_func = btrfs_ctree_exit,
2525 }, {
2526 .init_func = btrfs_free_space_init,
2527 .exit_func = btrfs_free_space_exit,
2528 }, {
2529 .init_func = extent_state_init_cachep,
2530 .exit_func = extent_state_free_cachep,
2531 }, {
2532 .init_func = extent_buffer_init_cachep,
2533 .exit_func = extent_buffer_free_cachep,
2534 }, {
2535 .init_func = btrfs_bioset_init,
2536 .exit_func = btrfs_bioset_exit,
2537 }, {
2538 .init_func = extent_map_init,
2539 .exit_func = extent_map_exit,
2540 }, {
2541 .init_func = ordered_data_init,
2542 .exit_func = ordered_data_exit,
2543 }, {
2544 .init_func = btrfs_delayed_inode_init,
2545 .exit_func = btrfs_delayed_inode_exit,
2546 }, {
2547 .init_func = btrfs_auto_defrag_init,
2548 .exit_func = btrfs_auto_defrag_exit,
2549 }, {
2550 .init_func = btrfs_delayed_ref_init,
2551 .exit_func = btrfs_delayed_ref_exit,
2552 }, {
2553 .init_func = btrfs_prelim_ref_init,
2554 .exit_func = btrfs_prelim_ref_exit,
2555 }, {
2556 .init_func = btrfs_interface_init,
2557 .exit_func = btrfs_interface_exit,
2558 }, {
2559 .init_func = btrfs_print_mod_info,
2560 .exit_func = NULL,
2561 }, {
2562 .init_func = btrfs_run_sanity_tests,
2563 .exit_func = NULL,
2564 }, {
2565 .init_func = register_btrfs,
2566 .exit_func = unregister_btrfs,
2567 }
2568 };
2569
2570 static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2571
2572 static __always_inline void btrfs_exit_btrfs_fs(void)
2573 {
2574 int i;
2575
2576 for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2577 if (!mod_init_result[i])
2578 continue;
2579 if (mod_init_seq[i].exit_func)
2580 mod_init_seq[i].exit_func();
2581 mod_init_result[i] = false;
2582 }
2583 }
2584
2585 static void __exit exit_btrfs_fs(void)
2586 {
2587 btrfs_exit_btrfs_fs();
2588 btrfs_cleanup_fs_uuids();
2589 }
2590
2591 static int __init init_btrfs_fs(void)
2592 {
2593 int ret;
2594 int i;
2595
2596 for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2597 ASSERT(!mod_init_result[i]);
2598 ret = mod_init_seq[i].init_func();
2599 if (ret < 0) {
2600 btrfs_exit_btrfs_fs();
2601 return ret;
2602 }
2603 mod_init_result[i] = true;
2604 }
2605 return 0;
2606 }
2607
2608 late_initcall(init_btrfs_fs);
2609 module_exit(exit_btrfs_fs)
2610
2611 MODULE_DESCRIPTION("B-Tree File System (BTRFS)");
2612 MODULE_LICENSE("GPL");
2613 MODULE_SOFTDEP("pre: crc32c");
2614 MODULE_SOFTDEP("pre: xxhash64");
2615 MODULE_SOFTDEP("pre: sha256");
2616 MODULE_SOFTDEP("pre: blake2b-256");
2617
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.