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 <linux/swap.h>
32 #include "messages.h"
33 #include "delayed-inode.h"
34 #include "ctree.h"
35 #include "disk-io.h"
36 #include "transaction.h"
37 #include "btrfs_inode.h"
38 #include "direct-io.h"
39 #include "props.h"
40 #include "xattr.h"
41 #include "bio.h"
42 #include "export.h"
43 #include "compression.h"
44 #include "dev-replace.h"
45 #include "free-space-cache.h"
46 #include "backref.h"
47 #include "space-info.h"
48 #include "sysfs.h"
49 #include "zoned.h"
50 #include "tests/btrfs-tests.h"
51 #include "block-group.h"
52 #include "discard.h"
53 #include "qgroup.h"
54 #include "raid56.h"
55 #include "fs.h"
56 #include "accessors.h"
57 #include "defrag.h"
58 #include "dir-item.h"
59 #include "ioctl.h"
60 #include "scrub.h"
61 #include "verity.h"
62 #include "super.h"
63 #include "extent-tree.h"
64 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h>
66
67 static const struct super_operations btrfs_super_ops;
68 static struct file_system_type btrfs_fs_type;
69
70 static void btrfs_put_super(struct super_block *sb)
71 {
72 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
73
74 btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid);
75 close_ctree(fs_info);
76 }
77
78 /* Store the mount options related information. */
79 struct btrfs_fs_context {
80 char *subvol_name;
81 u64 subvol_objectid;
82 u64 max_inline;
83 u32 commit_interval;
84 u32 metadata_ratio;
85 u32 thread_pool_size;
86 unsigned long long mount_opt;
87 unsigned long compress_type:4;
88 unsigned int compress_level;
89 refcount_t refs;
90 };
91
92 enum {
93 Opt_acl,
94 Opt_clear_cache,
95 Opt_commit_interval,
96 Opt_compress,
97 Opt_compress_force,
98 Opt_compress_force_type,
99 Opt_compress_type,
100 Opt_degraded,
101 Opt_device,
102 Opt_fatal_errors,
103 Opt_flushoncommit,
104 Opt_max_inline,
105 Opt_barrier,
106 Opt_datacow,
107 Opt_datasum,
108 Opt_defrag,
109 Opt_discard,
110 Opt_discard_mode,
111 Opt_ratio,
112 Opt_rescan_uuid_tree,
113 Opt_skip_balance,
114 Opt_space_cache,
115 Opt_space_cache_version,
116 Opt_ssd,
117 Opt_ssd_spread,
118 Opt_subvol,
119 Opt_subvol_empty,
120 Opt_subvolid,
121 Opt_thread_pool,
122 Opt_treelog,
123 Opt_user_subvol_rm_allowed,
124 Opt_norecovery,
125
126 /* Rescue options */
127 Opt_rescue,
128 Opt_usebackuproot,
129 Opt_nologreplay,
130
131 /* Debugging options */
132 Opt_enospc_debug,
133 #ifdef CONFIG_BTRFS_DEBUG
134 Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
135 #endif
136 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
137 Opt_ref_verify,
138 #endif
139 Opt_err,
140 };
141
142 enum {
143 Opt_fatal_errors_panic,
144 Opt_fatal_errors_bug,
145 };
146
147 static const struct constant_table btrfs_parameter_fatal_errors[] = {
148 { "panic", Opt_fatal_errors_panic },
149 { "bug", Opt_fatal_errors_bug },
150 {}
151 };
152
153 enum {
154 Opt_discard_sync,
155 Opt_discard_async,
156 };
157
158 static const struct constant_table btrfs_parameter_discard[] = {
159 { "sync", Opt_discard_sync },
160 { "async", Opt_discard_async },
161 {}
162 };
163
164 enum {
165 Opt_space_cache_v1,
166 Opt_space_cache_v2,
167 };
168
169 static const struct constant_table btrfs_parameter_space_cache[] = {
170 { "v1", Opt_space_cache_v1 },
171 { "v2", Opt_space_cache_v2 },
172 {}
173 };
174
175 enum {
176 Opt_rescue_usebackuproot,
177 Opt_rescue_nologreplay,
178 Opt_rescue_ignorebadroots,
179 Opt_rescue_ignoredatacsums,
180 Opt_rescue_ignoremetacsums,
181 Opt_rescue_ignoresuperflags,
182 Opt_rescue_parameter_all,
183 };
184
185 static const struct constant_table btrfs_parameter_rescue[] = {
186 { "usebackuproot", Opt_rescue_usebackuproot },
187 { "nologreplay", Opt_rescue_nologreplay },
188 { "ignorebadroots", Opt_rescue_ignorebadroots },
189 { "ibadroots", Opt_rescue_ignorebadroots },
190 { "ignoredatacsums", Opt_rescue_ignoredatacsums },
191 { "ignoremetacsums", Opt_rescue_ignoremetacsums},
192 { "ignoresuperflags", Opt_rescue_ignoresuperflags},
193 { "idatacsums", Opt_rescue_ignoredatacsums },
194 { "imetacsums", Opt_rescue_ignoremetacsums},
195 { "isuperflags", Opt_rescue_ignoresuperflags},
196 { "all", Opt_rescue_parameter_all },
197 {}
198 };
199
200 #ifdef CONFIG_BTRFS_DEBUG
201 enum {
202 Opt_fragment_parameter_data,
203 Opt_fragment_parameter_metadata,
204 Opt_fragment_parameter_all,
205 };
206
207 static const struct constant_table btrfs_parameter_fragment[] = {
208 { "data", Opt_fragment_parameter_data },
209 { "metadata", Opt_fragment_parameter_metadata },
210 { "all", Opt_fragment_parameter_all },
211 {}
212 };
213 #endif
214
215 static const struct fs_parameter_spec btrfs_fs_parameters[] = {
216 fsparam_flag_no("acl", Opt_acl),
217 fsparam_flag_no("autodefrag", Opt_defrag),
218 fsparam_flag_no("barrier", Opt_barrier),
219 fsparam_flag("clear_cache", Opt_clear_cache),
220 fsparam_u32("commit", Opt_commit_interval),
221 fsparam_flag("compress", Opt_compress),
222 fsparam_string("compress", Opt_compress_type),
223 fsparam_flag("compress-force", Opt_compress_force),
224 fsparam_string("compress-force", Opt_compress_force_type),
225 fsparam_flag_no("datacow", Opt_datacow),
226 fsparam_flag_no("datasum", Opt_datasum),
227 fsparam_flag("degraded", Opt_degraded),
228 fsparam_string("device", Opt_device),
229 fsparam_flag_no("discard", Opt_discard),
230 fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard),
231 fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors),
232 fsparam_flag_no("flushoncommit", Opt_flushoncommit),
233 fsparam_string("max_inline", Opt_max_inline),
234 fsparam_u32("metadata_ratio", Opt_ratio),
235 fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree),
236 fsparam_flag("skip_balance", Opt_skip_balance),
237 fsparam_flag_no("space_cache", Opt_space_cache),
238 fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache),
239 fsparam_flag_no("ssd", Opt_ssd),
240 fsparam_flag_no("ssd_spread", Opt_ssd_spread),
241 fsparam_string("subvol", Opt_subvol),
242 fsparam_flag("subvol=", Opt_subvol_empty),
243 fsparam_u64("subvolid", Opt_subvolid),
244 fsparam_u32("thread_pool", Opt_thread_pool),
245 fsparam_flag_no("treelog", Opt_treelog),
246 fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed),
247
248 /* Rescue options. */
249 fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue),
250 /* Deprecated, with alias rescue=nologreplay */
251 __fsparam(NULL, "nologreplay", Opt_nologreplay, fs_param_deprecated, NULL),
252 /* Deprecated, with alias rescue=usebackuproot */
253 __fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL),
254 /* For compatibility only, alias for "rescue=nologreplay". */
255 fsparam_flag("norecovery", Opt_norecovery),
256
257 /* Debugging options. */
258 fsparam_flag_no("enospc_debug", Opt_enospc_debug),
259 #ifdef CONFIG_BTRFS_DEBUG
260 fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment),
261 #endif
262 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
263 fsparam_flag("ref_verify", Opt_ref_verify),
264 #endif
265 {}
266 };
267
268 /* No support for restricting writes to btrfs devices yet... */
269 static inline blk_mode_t btrfs_open_mode(struct fs_context *fc)
270 {
271 return sb_open_mode(fc->sb_flags) & ~BLK_OPEN_RESTRICT_WRITES;
272 }
273
274 static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
275 {
276 struct btrfs_fs_context *ctx = fc->fs_private;
277 struct fs_parse_result result;
278 int opt;
279
280 opt = fs_parse(fc, btrfs_fs_parameters, param, &result);
281 if (opt < 0)
282 return opt;
283
284 switch (opt) {
285 case Opt_degraded:
286 btrfs_set_opt(ctx->mount_opt, DEGRADED);
287 break;
288 case Opt_subvol_empty:
289 /*
290 * This exists because we used to allow it on accident, so we're
291 * keeping it to maintain ABI. See 37becec95ac3 ("Btrfs: allow
292 * empty subvol= again").
293 */
294 break;
295 case Opt_subvol:
296 kfree(ctx->subvol_name);
297 ctx->subvol_name = kstrdup(param->string, GFP_KERNEL);
298 if (!ctx->subvol_name)
299 return -ENOMEM;
300 break;
301 case Opt_subvolid:
302 ctx->subvol_objectid = result.uint_64;
303
304 /* subvolid=0 means give me the original fs_tree. */
305 if (!ctx->subvol_objectid)
306 ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
307 break;
308 case Opt_device: {
309 struct btrfs_device *device;
310 blk_mode_t mode = btrfs_open_mode(fc);
311
312 mutex_lock(&uuid_mutex);
313 device = btrfs_scan_one_device(param->string, mode, false);
314 mutex_unlock(&uuid_mutex);
315 if (IS_ERR(device))
316 return PTR_ERR(device);
317 break;
318 }
319 case Opt_datasum:
320 if (result.negated) {
321 btrfs_set_opt(ctx->mount_opt, NODATASUM);
322 } else {
323 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
324 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
325 }
326 break;
327 case Opt_datacow:
328 if (result.negated) {
329 btrfs_clear_opt(ctx->mount_opt, COMPRESS);
330 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
331 btrfs_set_opt(ctx->mount_opt, NODATACOW);
332 btrfs_set_opt(ctx->mount_opt, NODATASUM);
333 } else {
334 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
335 }
336 break;
337 case Opt_compress_force:
338 case Opt_compress_force_type:
339 btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS);
340 fallthrough;
341 case Opt_compress:
342 case Opt_compress_type:
343 if (opt == Opt_compress || opt == Opt_compress_force) {
344 ctx->compress_type = BTRFS_COMPRESS_ZLIB;
345 ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
346 btrfs_set_opt(ctx->mount_opt, COMPRESS);
347 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
348 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
349 } else if (strncmp(param->string, "zlib", 4) == 0) {
350 ctx->compress_type = BTRFS_COMPRESS_ZLIB;
351 ctx->compress_level =
352 btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB,
353 param->string + 4);
354 btrfs_set_opt(ctx->mount_opt, COMPRESS);
355 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
356 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
357 } else if (strncmp(param->string, "lzo", 3) == 0) {
358 ctx->compress_type = BTRFS_COMPRESS_LZO;
359 ctx->compress_level = 0;
360 btrfs_set_opt(ctx->mount_opt, COMPRESS);
361 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
362 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
363 } else if (strncmp(param->string, "zstd", 4) == 0) {
364 ctx->compress_type = BTRFS_COMPRESS_ZSTD;
365 ctx->compress_level =
366 btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD,
367 param->string + 4);
368 btrfs_set_opt(ctx->mount_opt, COMPRESS);
369 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
370 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
371 } else if (strncmp(param->string, "no", 2) == 0) {
372 ctx->compress_level = 0;
373 ctx->compress_type = 0;
374 btrfs_clear_opt(ctx->mount_opt, COMPRESS);
375 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
376 } else {
377 btrfs_err(NULL, "unrecognized compression value %s",
378 param->string);
379 return -EINVAL;
380 }
381 break;
382 case Opt_ssd:
383 if (result.negated) {
384 btrfs_set_opt(ctx->mount_opt, NOSSD);
385 btrfs_clear_opt(ctx->mount_opt, SSD);
386 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
387 } else {
388 btrfs_set_opt(ctx->mount_opt, SSD);
389 btrfs_clear_opt(ctx->mount_opt, NOSSD);
390 }
391 break;
392 case Opt_ssd_spread:
393 if (result.negated) {
394 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
395 } else {
396 btrfs_set_opt(ctx->mount_opt, SSD);
397 btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
398 btrfs_clear_opt(ctx->mount_opt, NOSSD);
399 }
400 break;
401 case Opt_barrier:
402 if (result.negated)
403 btrfs_set_opt(ctx->mount_opt, NOBARRIER);
404 else
405 btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
406 break;
407 case Opt_thread_pool:
408 if (result.uint_32 == 0) {
409 btrfs_err(NULL, "invalid value 0 for thread_pool");
410 return -EINVAL;
411 }
412 ctx->thread_pool_size = result.uint_32;
413 break;
414 case Opt_max_inline:
415 ctx->max_inline = memparse(param->string, NULL);
416 break;
417 case Opt_acl:
418 if (result.negated) {
419 fc->sb_flags &= ~SB_POSIXACL;
420 } else {
421 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
422 fc->sb_flags |= SB_POSIXACL;
423 #else
424 btrfs_err(NULL, "support for ACL not compiled in");
425 return -EINVAL;
426 #endif
427 }
428 /*
429 * VFS limits the ability to toggle ACL on and off via remount,
430 * despite every file system allowing this. This seems to be
431 * an oversight since we all do, but it'll fail if we're
432 * remounting. So don't set the mask here, we'll check it in
433 * btrfs_reconfigure and do the toggling ourselves.
434 */
435 if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
436 fc->sb_flags_mask |= SB_POSIXACL;
437 break;
438 case Opt_treelog:
439 if (result.negated)
440 btrfs_set_opt(ctx->mount_opt, NOTREELOG);
441 else
442 btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
443 break;
444 case Opt_nologreplay:
445 btrfs_warn(NULL,
446 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
447 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
448 break;
449 case Opt_norecovery:
450 btrfs_info(NULL,
451 "'norecovery' is for compatibility only, recommended to use 'rescue=nologreplay'");
452 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
453 break;
454 case Opt_flushoncommit:
455 if (result.negated)
456 btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
457 else
458 btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
459 break;
460 case Opt_ratio:
461 ctx->metadata_ratio = result.uint_32;
462 break;
463 case Opt_discard:
464 if (result.negated) {
465 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
466 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
467 btrfs_set_opt(ctx->mount_opt, NODISCARD);
468 } else {
469 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
470 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
471 }
472 break;
473 case Opt_discard_mode:
474 switch (result.uint_32) {
475 case Opt_discard_sync:
476 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
477 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
478 break;
479 case Opt_discard_async:
480 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
481 btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
482 break;
483 default:
484 btrfs_err(NULL, "unrecognized discard mode value %s",
485 param->key);
486 return -EINVAL;
487 }
488 btrfs_clear_opt(ctx->mount_opt, NODISCARD);
489 break;
490 case Opt_space_cache:
491 if (result.negated) {
492 btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
493 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
494 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
495 } else {
496 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
497 btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
498 }
499 break;
500 case Opt_space_cache_version:
501 switch (result.uint_32) {
502 case Opt_space_cache_v1:
503 btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
504 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
505 break;
506 case Opt_space_cache_v2:
507 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
508 btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
509 break;
510 default:
511 btrfs_err(NULL, "unrecognized space_cache value %s",
512 param->key);
513 return -EINVAL;
514 }
515 break;
516 case Opt_rescan_uuid_tree:
517 btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
518 break;
519 case Opt_clear_cache:
520 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
521 break;
522 case Opt_user_subvol_rm_allowed:
523 btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
524 break;
525 case Opt_enospc_debug:
526 if (result.negated)
527 btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
528 else
529 btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
530 break;
531 case Opt_defrag:
532 if (result.negated)
533 btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
534 else
535 btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
536 break;
537 case Opt_usebackuproot:
538 btrfs_warn(NULL,
539 "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
540 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
541
542 /* If we're loading the backup roots we can't trust the space cache. */
543 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
544 break;
545 case Opt_skip_balance:
546 btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
547 break;
548 case Opt_fatal_errors:
549 switch (result.uint_32) {
550 case Opt_fatal_errors_panic:
551 btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
552 break;
553 case Opt_fatal_errors_bug:
554 btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
555 break;
556 default:
557 btrfs_err(NULL, "unrecognized fatal_errors value %s",
558 param->key);
559 return -EINVAL;
560 }
561 break;
562 case Opt_commit_interval:
563 ctx->commit_interval = result.uint_32;
564 if (ctx->commit_interval == 0)
565 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
566 break;
567 case Opt_rescue:
568 switch (result.uint_32) {
569 case Opt_rescue_usebackuproot:
570 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
571 break;
572 case Opt_rescue_nologreplay:
573 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
574 break;
575 case Opt_rescue_ignorebadroots:
576 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
577 break;
578 case Opt_rescue_ignoredatacsums:
579 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
580 break;
581 case Opt_rescue_ignoremetacsums:
582 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
583 break;
584 case Opt_rescue_ignoresuperflags:
585 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
586 break;
587 case Opt_rescue_parameter_all:
588 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
589 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
590 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
591 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
592 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
593 break;
594 default:
595 btrfs_info(NULL, "unrecognized rescue option '%s'",
596 param->key);
597 return -EINVAL;
598 }
599 break;
600 #ifdef CONFIG_BTRFS_DEBUG
601 case Opt_fragment:
602 switch (result.uint_32) {
603 case Opt_fragment_parameter_all:
604 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
605 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
606 break;
607 case Opt_fragment_parameter_metadata:
608 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
609 break;
610 case Opt_fragment_parameter_data:
611 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
612 break;
613 default:
614 btrfs_info(NULL, "unrecognized fragment option '%s'",
615 param->key);
616 return -EINVAL;
617 }
618 break;
619 #endif
620 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
621 case Opt_ref_verify:
622 btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
623 break;
624 #endif
625 default:
626 btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
627 return -EINVAL;
628 }
629
630 return 0;
631 }
632
633 /*
634 * Some options only have meaning at mount time and shouldn't persist across
635 * remounts, or be displayed. Clear these at the end of mount and remount code
636 * paths.
637 */
638 static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
639 {
640 btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
641 btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
642 btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
643 }
644
645 static bool check_ro_option(const struct btrfs_fs_info *fs_info,
646 unsigned long long mount_opt, unsigned long long opt,
647 const char *opt_name)
648 {
649 if (mount_opt & opt) {
650 btrfs_err(fs_info, "%s must be used with ro mount option",
651 opt_name);
652 return true;
653 }
654 return false;
655 }
656
657 bool btrfs_check_options(const struct btrfs_fs_info *info,
658 unsigned long long *mount_opt,
659 unsigned long flags)
660 {
661 bool ret = true;
662
663 if (!(flags & SB_RDONLY) &&
664 (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
665 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
666 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums") ||
667 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREMETACSUMS, "ignoremetacsums") ||
668 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNORESUPERFLAGS, "ignoresuperflags")))
669 ret = false;
670
671 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
672 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
673 !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
674 btrfs_err(info, "cannot disable free-space-tree");
675 ret = false;
676 }
677 if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
678 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
679 btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
680 ret = false;
681 }
682
683 if (btrfs_check_mountopts_zoned(info, mount_opt))
684 ret = false;
685
686 if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
687 if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE)) {
688 btrfs_info(info, "disk space caching is enabled");
689 btrfs_warn(info,
690 "space cache v1 is being deprecated and will be removed in a future release, please use -o space_cache=v2");
691 }
692 if (btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE))
693 btrfs_info(info, "using free-space-tree");
694 }
695
696 return ret;
697 }
698
699 /*
700 * This is subtle, we only call this during open_ctree(). We need to pre-load
701 * the mount options with the on-disk settings. Before the new mount API took
702 * effect we would do this on mount and remount. With the new mount API we'll
703 * only do this on the initial mount.
704 *
705 * This isn't a change in behavior, because we're using the current state of the
706 * file system to set the current mount options. If you mounted with special
707 * options to disable these features and then remounted we wouldn't revert the
708 * settings, because mounting without these features cleared the on-disk
709 * settings, so this being called on re-mount is not needed.
710 */
711 void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
712 {
713 if (fs_info->sectorsize < PAGE_SIZE) {
714 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
715 if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) {
716 btrfs_info(fs_info,
717 "forcing free space tree for sector size %u with page size %lu",
718 fs_info->sectorsize, PAGE_SIZE);
719 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
720 }
721 }
722
723 /*
724 * At this point our mount options are populated, so we only mess with
725 * these settings if we don't have any settings already.
726 */
727 if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
728 return;
729
730 if (btrfs_is_zoned(fs_info) &&
731 btrfs_free_space_cache_v1_active(fs_info)) {
732 btrfs_info(fs_info, "zoned: clearing existing space cache");
733 btrfs_set_super_cache_generation(fs_info->super_copy, 0);
734 return;
735 }
736
737 if (btrfs_test_opt(fs_info, SPACE_CACHE))
738 return;
739
740 if (btrfs_test_opt(fs_info, NOSPACECACHE))
741 return;
742
743 /*
744 * At this point we don't have explicit options set by the user, set
745 * them ourselves based on the state of the file system.
746 */
747 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
748 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
749 else if (btrfs_free_space_cache_v1_active(fs_info))
750 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
751 }
752
753 static void set_device_specific_options(struct btrfs_fs_info *fs_info)
754 {
755 if (!btrfs_test_opt(fs_info, NOSSD) &&
756 !fs_info->fs_devices->rotating)
757 btrfs_set_opt(fs_info->mount_opt, SSD);
758
759 /*
760 * For devices supporting discard turn on discard=async automatically,
761 * unless it's already set or disabled. This could be turned off by
762 * nodiscard for the same mount.
763 *
764 * The zoned mode piggy backs on the discard functionality for
765 * resetting a zone. There is no reason to delay the zone reset as it is
766 * fast enough. So, do not enable async discard for zoned mode.
767 */
768 if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
769 btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
770 btrfs_test_opt(fs_info, NODISCARD)) &&
771 fs_info->fs_devices->discardable &&
772 !btrfs_is_zoned(fs_info))
773 btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
774 }
775
776 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
777 u64 subvol_objectid)
778 {
779 struct btrfs_root *root = fs_info->tree_root;
780 struct btrfs_root *fs_root = NULL;
781 struct btrfs_root_ref *root_ref;
782 struct btrfs_inode_ref *inode_ref;
783 struct btrfs_key key;
784 struct btrfs_path *path = NULL;
785 char *name = NULL, *ptr;
786 u64 dirid;
787 int len;
788 int ret;
789
790 path = btrfs_alloc_path();
791 if (!path) {
792 ret = -ENOMEM;
793 goto err;
794 }
795
796 name = kmalloc(PATH_MAX, GFP_KERNEL);
797 if (!name) {
798 ret = -ENOMEM;
799 goto err;
800 }
801 ptr = name + PATH_MAX - 1;
802 ptr[0] = '\0';
803
804 /*
805 * Walk up the subvolume trees in the tree of tree roots by root
806 * backrefs until we hit the top-level subvolume.
807 */
808 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
809 key.objectid = subvol_objectid;
810 key.type = BTRFS_ROOT_BACKREF_KEY;
811 key.offset = (u64)-1;
812
813 ret = btrfs_search_backwards(root, &key, path);
814 if (ret < 0) {
815 goto err;
816 } else if (ret > 0) {
817 ret = -ENOENT;
818 goto err;
819 }
820
821 subvol_objectid = key.offset;
822
823 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
824 struct btrfs_root_ref);
825 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
826 ptr -= len + 1;
827 if (ptr < name) {
828 ret = -ENAMETOOLONG;
829 goto err;
830 }
831 read_extent_buffer(path->nodes[0], ptr + 1,
832 (unsigned long)(root_ref + 1), len);
833 ptr[0] = '/';
834 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
835 btrfs_release_path(path);
836
837 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
838 if (IS_ERR(fs_root)) {
839 ret = PTR_ERR(fs_root);
840 fs_root = NULL;
841 goto err;
842 }
843
844 /*
845 * Walk up the filesystem tree by inode refs until we hit the
846 * root directory.
847 */
848 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
849 key.objectid = dirid;
850 key.type = BTRFS_INODE_REF_KEY;
851 key.offset = (u64)-1;
852
853 ret = btrfs_search_backwards(fs_root, &key, path);
854 if (ret < 0) {
855 goto err;
856 } else if (ret > 0) {
857 ret = -ENOENT;
858 goto err;
859 }
860
861 dirid = key.offset;
862
863 inode_ref = btrfs_item_ptr(path->nodes[0],
864 path->slots[0],
865 struct btrfs_inode_ref);
866 len = btrfs_inode_ref_name_len(path->nodes[0],
867 inode_ref);
868 ptr -= len + 1;
869 if (ptr < name) {
870 ret = -ENAMETOOLONG;
871 goto err;
872 }
873 read_extent_buffer(path->nodes[0], ptr + 1,
874 (unsigned long)(inode_ref + 1), len);
875 ptr[0] = '/';
876 btrfs_release_path(path);
877 }
878 btrfs_put_root(fs_root);
879 fs_root = NULL;
880 }
881
882 btrfs_free_path(path);
883 if (ptr == name + PATH_MAX - 1) {
884 name[0] = '/';
885 name[1] = '\0';
886 } else {
887 memmove(name, ptr, name + PATH_MAX - ptr);
888 }
889 return name;
890
891 err:
892 btrfs_put_root(fs_root);
893 btrfs_free_path(path);
894 kfree(name);
895 return ERR_PTR(ret);
896 }
897
898 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
899 {
900 struct btrfs_root *root = fs_info->tree_root;
901 struct btrfs_dir_item *di;
902 struct btrfs_path *path;
903 struct btrfs_key location;
904 struct fscrypt_str name = FSTR_INIT("default", 7);
905 u64 dir_id;
906
907 path = btrfs_alloc_path();
908 if (!path)
909 return -ENOMEM;
910
911 /*
912 * Find the "default" dir item which points to the root item that we
913 * will mount by default if we haven't been given a specific subvolume
914 * to mount.
915 */
916 dir_id = btrfs_super_root_dir(fs_info->super_copy);
917 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
918 if (IS_ERR(di)) {
919 btrfs_free_path(path);
920 return PTR_ERR(di);
921 }
922 if (!di) {
923 /*
924 * Ok the default dir item isn't there. This is weird since
925 * it's always been there, but don't freak out, just try and
926 * mount the top-level subvolume.
927 */
928 btrfs_free_path(path);
929 *objectid = BTRFS_FS_TREE_OBJECTID;
930 return 0;
931 }
932
933 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
934 btrfs_free_path(path);
935 *objectid = location.objectid;
936 return 0;
937 }
938
939 static int btrfs_fill_super(struct super_block *sb,
940 struct btrfs_fs_devices *fs_devices,
941 void *data)
942 {
943 struct inode *inode;
944 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
945 int err;
946
947 sb->s_maxbytes = MAX_LFS_FILESIZE;
948 sb->s_magic = BTRFS_SUPER_MAGIC;
949 sb->s_op = &btrfs_super_ops;
950 sb->s_d_op = &btrfs_dentry_operations;
951 sb->s_export_op = &btrfs_export_ops;
952 #ifdef CONFIG_FS_VERITY
953 sb->s_vop = &btrfs_verityops;
954 #endif
955 sb->s_xattr = btrfs_xattr_handlers;
956 sb->s_time_gran = 1;
957 sb->s_iflags |= SB_I_CGROUPWB;
958
959 err = super_setup_bdi(sb);
960 if (err) {
961 btrfs_err(fs_info, "super_setup_bdi failed");
962 return err;
963 }
964
965 err = open_ctree(sb, fs_devices, (char *)data);
966 if (err) {
967 btrfs_err(fs_info, "open_ctree failed");
968 return err;
969 }
970
971 inode = btrfs_iget(BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
972 if (IS_ERR(inode)) {
973 err = PTR_ERR(inode);
974 btrfs_handle_fs_error(fs_info, err, NULL);
975 goto fail_close;
976 }
977
978 sb->s_root = d_make_root(inode);
979 if (!sb->s_root) {
980 err = -ENOMEM;
981 goto fail_close;
982 }
983
984 sb->s_flags |= SB_ACTIVE;
985 return 0;
986
987 fail_close:
988 close_ctree(fs_info);
989 return err;
990 }
991
992 int btrfs_sync_fs(struct super_block *sb, int wait)
993 {
994 struct btrfs_trans_handle *trans;
995 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
996 struct btrfs_root *root = fs_info->tree_root;
997
998 trace_btrfs_sync_fs(fs_info, wait);
999
1000 if (!wait) {
1001 filemap_flush(fs_info->btree_inode->i_mapping);
1002 return 0;
1003 }
1004
1005 btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL);
1006
1007 trans = btrfs_attach_transaction_barrier(root);
1008 if (IS_ERR(trans)) {
1009 /* no transaction, don't bother */
1010 if (PTR_ERR(trans) == -ENOENT) {
1011 /*
1012 * Exit unless we have some pending changes
1013 * that need to go through commit
1014 */
1015 if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1016 &fs_info->flags))
1017 return 0;
1018 /*
1019 * A non-blocking test if the fs is frozen. We must not
1020 * start a new transaction here otherwise a deadlock
1021 * happens. The pending operations are delayed to the
1022 * next commit after thawing.
1023 */
1024 if (sb_start_write_trylock(sb))
1025 sb_end_write(sb);
1026 else
1027 return 0;
1028 trans = btrfs_start_transaction(root, 0);
1029 }
1030 if (IS_ERR(trans))
1031 return PTR_ERR(trans);
1032 }
1033 return btrfs_commit_transaction(trans);
1034 }
1035
1036 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1037 {
1038 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1039 *printed = true;
1040 }
1041
1042 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1043 {
1044 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1045 const char *compress_type;
1046 const char *subvol_name;
1047 bool printed = false;
1048
1049 if (btrfs_test_opt(info, DEGRADED))
1050 seq_puts(seq, ",degraded");
1051 if (btrfs_test_opt(info, NODATASUM))
1052 seq_puts(seq, ",nodatasum");
1053 if (btrfs_test_opt(info, NODATACOW))
1054 seq_puts(seq, ",nodatacow");
1055 if (btrfs_test_opt(info, NOBARRIER))
1056 seq_puts(seq, ",nobarrier");
1057 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1058 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1059 if (info->thread_pool_size != min_t(unsigned long,
1060 num_online_cpus() + 2, 8))
1061 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1062 if (btrfs_test_opt(info, COMPRESS)) {
1063 compress_type = btrfs_compress_type2str(info->compress_type);
1064 if (btrfs_test_opt(info, FORCE_COMPRESS))
1065 seq_printf(seq, ",compress-force=%s", compress_type);
1066 else
1067 seq_printf(seq, ",compress=%s", compress_type);
1068 if (info->compress_level)
1069 seq_printf(seq, ":%d", info->compress_level);
1070 }
1071 if (btrfs_test_opt(info, NOSSD))
1072 seq_puts(seq, ",nossd");
1073 if (btrfs_test_opt(info, SSD_SPREAD))
1074 seq_puts(seq, ",ssd_spread");
1075 else if (btrfs_test_opt(info, SSD))
1076 seq_puts(seq, ",ssd");
1077 if (btrfs_test_opt(info, NOTREELOG))
1078 seq_puts(seq, ",notreelog");
1079 if (btrfs_test_opt(info, NOLOGREPLAY))
1080 print_rescue_option(seq, "nologreplay", &printed);
1081 if (btrfs_test_opt(info, USEBACKUPROOT))
1082 print_rescue_option(seq, "usebackuproot", &printed);
1083 if (btrfs_test_opt(info, IGNOREBADROOTS))
1084 print_rescue_option(seq, "ignorebadroots", &printed);
1085 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1086 print_rescue_option(seq, "ignoredatacsums", &printed);
1087 if (btrfs_test_opt(info, IGNOREMETACSUMS))
1088 print_rescue_option(seq, "ignoremetacsums", &printed);
1089 if (btrfs_test_opt(info, IGNORESUPERFLAGS))
1090 print_rescue_option(seq, "ignoresuperflags", &printed);
1091 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1092 seq_puts(seq, ",flushoncommit");
1093 if (btrfs_test_opt(info, DISCARD_SYNC))
1094 seq_puts(seq, ",discard");
1095 if (btrfs_test_opt(info, DISCARD_ASYNC))
1096 seq_puts(seq, ",discard=async");
1097 if (!(info->sb->s_flags & SB_POSIXACL))
1098 seq_puts(seq, ",noacl");
1099 if (btrfs_free_space_cache_v1_active(info))
1100 seq_puts(seq, ",space_cache");
1101 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1102 seq_puts(seq, ",space_cache=v2");
1103 else
1104 seq_puts(seq, ",nospace_cache");
1105 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1106 seq_puts(seq, ",rescan_uuid_tree");
1107 if (btrfs_test_opt(info, CLEAR_CACHE))
1108 seq_puts(seq, ",clear_cache");
1109 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1110 seq_puts(seq, ",user_subvol_rm_allowed");
1111 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1112 seq_puts(seq, ",enospc_debug");
1113 if (btrfs_test_opt(info, AUTO_DEFRAG))
1114 seq_puts(seq, ",autodefrag");
1115 if (btrfs_test_opt(info, SKIP_BALANCE))
1116 seq_puts(seq, ",skip_balance");
1117 if (info->metadata_ratio)
1118 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1119 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1120 seq_puts(seq, ",fatal_errors=panic");
1121 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1122 seq_printf(seq, ",commit=%u", info->commit_interval);
1123 #ifdef CONFIG_BTRFS_DEBUG
1124 if (btrfs_test_opt(info, FRAGMENT_DATA))
1125 seq_puts(seq, ",fragment=data");
1126 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1127 seq_puts(seq, ",fragment=metadata");
1128 #endif
1129 if (btrfs_test_opt(info, REF_VERIFY))
1130 seq_puts(seq, ",ref_verify");
1131 seq_printf(seq, ",subvolid=%llu", btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1132 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1133 btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1134 if (!IS_ERR(subvol_name)) {
1135 seq_puts(seq, ",subvol=");
1136 seq_escape(seq, subvol_name, " \t\n\\");
1137 kfree(subvol_name);
1138 }
1139 return 0;
1140 }
1141
1142 /*
1143 * subvolumes are identified by ino 256
1144 */
1145 static inline int is_subvolume_inode(struct inode *inode)
1146 {
1147 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1148 return 1;
1149 return 0;
1150 }
1151
1152 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1153 struct vfsmount *mnt)
1154 {
1155 struct dentry *root;
1156 int ret;
1157
1158 if (!subvol_name) {
1159 if (!subvol_objectid) {
1160 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1161 &subvol_objectid);
1162 if (ret) {
1163 root = ERR_PTR(ret);
1164 goto out;
1165 }
1166 }
1167 subvol_name = btrfs_get_subvol_name_from_objectid(
1168 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1169 if (IS_ERR(subvol_name)) {
1170 root = ERR_CAST(subvol_name);
1171 subvol_name = NULL;
1172 goto out;
1173 }
1174
1175 }
1176
1177 root = mount_subtree(mnt, subvol_name);
1178 /* mount_subtree() drops our reference on the vfsmount. */
1179 mnt = NULL;
1180
1181 if (!IS_ERR(root)) {
1182 struct super_block *s = root->d_sb;
1183 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1184 struct inode *root_inode = d_inode(root);
1185 u64 root_objectid = btrfs_root_id(BTRFS_I(root_inode)->root);
1186
1187 ret = 0;
1188 if (!is_subvolume_inode(root_inode)) {
1189 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1190 subvol_name);
1191 ret = -EINVAL;
1192 }
1193 if (subvol_objectid && root_objectid != subvol_objectid) {
1194 /*
1195 * This will also catch a race condition where a
1196 * subvolume which was passed by ID is renamed and
1197 * another subvolume is renamed over the old location.
1198 */
1199 btrfs_err(fs_info,
1200 "subvol '%s' does not match subvolid %llu",
1201 subvol_name, subvol_objectid);
1202 ret = -EINVAL;
1203 }
1204 if (ret) {
1205 dput(root);
1206 root = ERR_PTR(ret);
1207 deactivate_locked_super(s);
1208 }
1209 }
1210
1211 out:
1212 mntput(mnt);
1213 kfree(subvol_name);
1214 return root;
1215 }
1216
1217 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1218 u32 new_pool_size, u32 old_pool_size)
1219 {
1220 if (new_pool_size == old_pool_size)
1221 return;
1222
1223 fs_info->thread_pool_size = new_pool_size;
1224
1225 btrfs_info(fs_info, "resize thread pool %d -> %d",
1226 old_pool_size, new_pool_size);
1227
1228 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1229 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1230 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1231 workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1232 workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1233 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1234 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1235 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1236 }
1237
1238 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1239 unsigned long long old_opts, int flags)
1240 {
1241 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1242 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1243 (flags & SB_RDONLY))) {
1244 /* wait for any defraggers to finish */
1245 wait_event(fs_info->transaction_wait,
1246 (atomic_read(&fs_info->defrag_running) == 0));
1247 if (flags & SB_RDONLY)
1248 sync_filesystem(fs_info->sb);
1249 }
1250 }
1251
1252 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1253 unsigned long long old_opts)
1254 {
1255 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1256
1257 /*
1258 * We need to cleanup all defragable inodes if the autodefragment is
1259 * close or the filesystem is read only.
1260 */
1261 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1262 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1263 btrfs_cleanup_defrag_inodes(fs_info);
1264 }
1265
1266 /* If we toggled discard async */
1267 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1268 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1269 btrfs_discard_resume(fs_info);
1270 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1271 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1272 btrfs_discard_cleanup(fs_info);
1273
1274 /* If we toggled space cache */
1275 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1276 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1277 }
1278
1279 static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1280 {
1281 int ret;
1282
1283 if (BTRFS_FS_ERROR(fs_info)) {
1284 btrfs_err(fs_info,
1285 "remounting read-write after error is not allowed");
1286 return -EINVAL;
1287 }
1288
1289 if (fs_info->fs_devices->rw_devices == 0)
1290 return -EACCES;
1291
1292 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1293 btrfs_warn(fs_info,
1294 "too many missing devices, writable remount is not allowed");
1295 return -EACCES;
1296 }
1297
1298 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1299 btrfs_warn(fs_info,
1300 "mount required to replay tree-log, cannot remount read-write");
1301 return -EINVAL;
1302 }
1303
1304 /*
1305 * NOTE: when remounting with a change that does writes, don't put it
1306 * anywhere above this point, as we are not sure to be safe to write
1307 * until we pass the above checks.
1308 */
1309 ret = btrfs_start_pre_rw_mount(fs_info);
1310 if (ret)
1311 return ret;
1312
1313 btrfs_clear_sb_rdonly(fs_info->sb);
1314
1315 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1316
1317 /*
1318 * If we've gone from readonly -> read-write, we need to get our
1319 * sync/async discard lists in the right state.
1320 */
1321 btrfs_discard_resume(fs_info);
1322
1323 return 0;
1324 }
1325
1326 static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1327 {
1328 /*
1329 * This also happens on 'umount -rf' or on shutdown, when the
1330 * filesystem is busy.
1331 */
1332 cancel_work_sync(&fs_info->async_reclaim_work);
1333 cancel_work_sync(&fs_info->async_data_reclaim_work);
1334
1335 btrfs_discard_cleanup(fs_info);
1336
1337 /* Wait for the uuid_scan task to finish */
1338 down(&fs_info->uuid_tree_rescan_sem);
1339 /* Avoid complains from lockdep et al. */
1340 up(&fs_info->uuid_tree_rescan_sem);
1341
1342 btrfs_set_sb_rdonly(fs_info->sb);
1343
1344 /*
1345 * Setting SB_RDONLY will put the cleaner thread to sleep at the next
1346 * loop if it's already active. If it's already asleep, we'll leave
1347 * unused block groups on disk until we're mounted read-write again
1348 * unless we clean them up here.
1349 */
1350 btrfs_delete_unused_bgs(fs_info);
1351
1352 /*
1353 * The cleaner task could be already running before we set the flag
1354 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock). We must make
1355 * sure that after we finish the remount, i.e. after we call
1356 * btrfs_commit_super(), the cleaner can no longer start a transaction
1357 * - either because it was dropping a dead root, running delayed iputs
1358 * or deleting an unused block group (the cleaner picked a block
1359 * group from the list of unused block groups before we were able to
1360 * in the previous call to btrfs_delete_unused_bgs()).
1361 */
1362 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1363
1364 /*
1365 * We've set the superblock to RO mode, so we might have made the
1366 * cleaner task sleep without running all pending delayed iputs. Go
1367 * through all the delayed iputs here, so that if an unmount happens
1368 * without remounting RW we don't end up at finishing close_ctree()
1369 * with a non-empty list of delayed iputs.
1370 */
1371 btrfs_run_delayed_iputs(fs_info);
1372
1373 btrfs_dev_replace_suspend_for_unmount(fs_info);
1374 btrfs_scrub_cancel(fs_info);
1375 btrfs_pause_balance(fs_info);
1376
1377 /*
1378 * Pause the qgroup rescan worker if it is running. We don't want it to
1379 * be still running after we are in RO mode, as after that, by the time
1380 * we unmount, it might have left a transaction open, so we would leak
1381 * the transaction and/or crash.
1382 */
1383 btrfs_qgroup_wait_for_completion(fs_info, false);
1384
1385 return btrfs_commit_super(fs_info);
1386 }
1387
1388 static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1389 {
1390 fs_info->max_inline = ctx->max_inline;
1391 fs_info->commit_interval = ctx->commit_interval;
1392 fs_info->metadata_ratio = ctx->metadata_ratio;
1393 fs_info->thread_pool_size = ctx->thread_pool_size;
1394 fs_info->mount_opt = ctx->mount_opt;
1395 fs_info->compress_type = ctx->compress_type;
1396 fs_info->compress_level = ctx->compress_level;
1397 }
1398
1399 static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1400 {
1401 ctx->max_inline = fs_info->max_inline;
1402 ctx->commit_interval = fs_info->commit_interval;
1403 ctx->metadata_ratio = fs_info->metadata_ratio;
1404 ctx->thread_pool_size = fs_info->thread_pool_size;
1405 ctx->mount_opt = fs_info->mount_opt;
1406 ctx->compress_type = fs_info->compress_type;
1407 ctx->compress_level = fs_info->compress_level;
1408 }
1409
1410 #define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...) \
1411 do { \
1412 if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1413 btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1414 btrfs_info(fs_info, fmt, ##args); \
1415 } while (0)
1416
1417 #define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...) \
1418 do { \
1419 if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1420 !btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1421 btrfs_info(fs_info, fmt, ##args); \
1422 } while (0)
1423
1424 static void btrfs_emit_options(struct btrfs_fs_info *info,
1425 struct btrfs_fs_context *old)
1426 {
1427 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1428 btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1429 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1430 btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1431 btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1432 btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1433 btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1434 btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1435 btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1436 btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1437 btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1438 btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1439 btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1440 btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1441 btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1442 btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1443 btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1444 btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1445 btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1446 btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1447 btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
1448 btrfs_info_if_set(info, old, IGNOREMETACSUMS, "ignoring meta csums");
1449 btrfs_info_if_set(info, old, IGNORESUPERFLAGS, "ignoring unknown super block flags");
1450
1451 btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1452 btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1453 btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1454 btrfs_info_if_unset(info, old, NOBARRIER, "turning off barriers");
1455 btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1456 btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1457 btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1458 btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1459 btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1460
1461 /* Did the compression settings change? */
1462 if (btrfs_test_opt(info, COMPRESS) &&
1463 (!old ||
1464 old->compress_type != info->compress_type ||
1465 old->compress_level != info->compress_level ||
1466 (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1467 btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1468 const char *compress_type = btrfs_compress_type2str(info->compress_type);
1469
1470 btrfs_info(info, "%s %s compression, level %d",
1471 btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1472 compress_type, info->compress_level);
1473 }
1474
1475 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1476 btrfs_info(info, "max_inline set to %llu", info->max_inline);
1477 }
1478
1479 static int btrfs_reconfigure(struct fs_context *fc)
1480 {
1481 struct super_block *sb = fc->root->d_sb;
1482 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1483 struct btrfs_fs_context *ctx = fc->fs_private;
1484 struct btrfs_fs_context old_ctx;
1485 int ret = 0;
1486 bool mount_reconfigure = (fc->s_fs_info != NULL);
1487
1488 btrfs_info_to_ctx(fs_info, &old_ctx);
1489
1490 /*
1491 * This is our "bind mount" trick, we don't want to allow the user to do
1492 * anything other than mount a different ro/rw and a different subvol,
1493 * all of the mount options should be maintained.
1494 */
1495 if (mount_reconfigure)
1496 ctx->mount_opt = old_ctx.mount_opt;
1497
1498 sync_filesystem(sb);
1499 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1500
1501 if (!mount_reconfigure &&
1502 !btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags))
1503 return -EINVAL;
1504
1505 ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
1506 if (ret < 0)
1507 return ret;
1508
1509 btrfs_ctx_to_info(fs_info, ctx);
1510 btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags);
1511 btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size,
1512 old_ctx.thread_pool_size);
1513
1514 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1515 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1516 (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1517 btrfs_warn(fs_info,
1518 "remount supports changing free space tree only from RO to RW");
1519 /* Make sure free space cache options match the state on disk. */
1520 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1521 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1522 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1523 }
1524 if (btrfs_free_space_cache_v1_active(fs_info)) {
1525 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1526 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1527 }
1528 }
1529
1530 ret = 0;
1531 if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1532 ret = btrfs_remount_ro(fs_info);
1533 else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1534 ret = btrfs_remount_rw(fs_info);
1535 if (ret)
1536 goto restore;
1537
1538 /*
1539 * If we set the mask during the parameter parsing VFS would reject the
1540 * remount. Here we can set the mask and the value will be updated
1541 * appropriately.
1542 */
1543 if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1544 fc->sb_flags_mask |= SB_POSIXACL;
1545
1546 btrfs_emit_options(fs_info, &old_ctx);
1547 wake_up_process(fs_info->transaction_kthread);
1548 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1549 btrfs_clear_oneshot_options(fs_info);
1550 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1551
1552 return 0;
1553 restore:
1554 btrfs_ctx_to_info(fs_info, &old_ctx);
1555 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1556 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1557 return ret;
1558 }
1559
1560 /* Used to sort the devices by max_avail(descending sort) */
1561 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1562 {
1563 const struct btrfs_device_info *dev_info1 = a;
1564 const struct btrfs_device_info *dev_info2 = b;
1565
1566 if (dev_info1->max_avail > dev_info2->max_avail)
1567 return -1;
1568 else if (dev_info1->max_avail < dev_info2->max_avail)
1569 return 1;
1570 return 0;
1571 }
1572
1573 /*
1574 * sort the devices by max_avail, in which max free extent size of each device
1575 * is stored.(Descending Sort)
1576 */
1577 static inline void btrfs_descending_sort_devices(
1578 struct btrfs_device_info *devices,
1579 size_t nr_devices)
1580 {
1581 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1582 btrfs_cmp_device_free_bytes, NULL);
1583 }
1584
1585 /*
1586 * The helper to calc the free space on the devices that can be used to store
1587 * file data.
1588 */
1589 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1590 u64 *free_bytes)
1591 {
1592 struct btrfs_device_info *devices_info;
1593 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1594 struct btrfs_device *device;
1595 u64 type;
1596 u64 avail_space;
1597 u64 min_stripe_size;
1598 int num_stripes = 1;
1599 int i = 0, nr_devices;
1600 const struct btrfs_raid_attr *rattr;
1601
1602 /*
1603 * We aren't under the device list lock, so this is racy-ish, but good
1604 * enough for our purposes.
1605 */
1606 nr_devices = fs_info->fs_devices->open_devices;
1607 if (!nr_devices) {
1608 smp_mb();
1609 nr_devices = fs_info->fs_devices->open_devices;
1610 ASSERT(nr_devices);
1611 if (!nr_devices) {
1612 *free_bytes = 0;
1613 return 0;
1614 }
1615 }
1616
1617 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1618 GFP_KERNEL);
1619 if (!devices_info)
1620 return -ENOMEM;
1621
1622 /* calc min stripe number for data space allocation */
1623 type = btrfs_data_alloc_profile(fs_info);
1624 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1625
1626 if (type & BTRFS_BLOCK_GROUP_RAID0)
1627 num_stripes = nr_devices;
1628 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1629 num_stripes = rattr->ncopies;
1630 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1631 num_stripes = 4;
1632
1633 /* Adjust for more than 1 stripe per device */
1634 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1635
1636 rcu_read_lock();
1637 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1638 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1639 &device->dev_state) ||
1640 !device->bdev ||
1641 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1642 continue;
1643
1644 if (i >= nr_devices)
1645 break;
1646
1647 avail_space = device->total_bytes - device->bytes_used;
1648
1649 /* align with stripe_len */
1650 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1651
1652 /*
1653 * Ensure we have at least min_stripe_size on top of the
1654 * reserved space on the device.
1655 */
1656 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1657 continue;
1658
1659 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1660
1661 devices_info[i].dev = device;
1662 devices_info[i].max_avail = avail_space;
1663
1664 i++;
1665 }
1666 rcu_read_unlock();
1667
1668 nr_devices = i;
1669
1670 btrfs_descending_sort_devices(devices_info, nr_devices);
1671
1672 i = nr_devices - 1;
1673 avail_space = 0;
1674 while (nr_devices >= rattr->devs_min) {
1675 num_stripes = min(num_stripes, nr_devices);
1676
1677 if (devices_info[i].max_avail >= min_stripe_size) {
1678 int j;
1679 u64 alloc_size;
1680
1681 avail_space += devices_info[i].max_avail * num_stripes;
1682 alloc_size = devices_info[i].max_avail;
1683 for (j = i + 1 - num_stripes; j <= i; j++)
1684 devices_info[j].max_avail -= alloc_size;
1685 }
1686 i--;
1687 nr_devices--;
1688 }
1689
1690 kfree(devices_info);
1691 *free_bytes = avail_space;
1692 return 0;
1693 }
1694
1695 /*
1696 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1697 *
1698 * If there's a redundant raid level at DATA block groups, use the respective
1699 * multiplier to scale the sizes.
1700 *
1701 * Unused device space usage is based on simulating the chunk allocator
1702 * algorithm that respects the device sizes and order of allocations. This is
1703 * a close approximation of the actual use but there are other factors that may
1704 * change the result (like a new metadata chunk).
1705 *
1706 * If metadata is exhausted, f_bavail will be 0.
1707 */
1708 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1709 {
1710 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1711 struct btrfs_super_block *disk_super = fs_info->super_copy;
1712 struct btrfs_space_info *found;
1713 u64 total_used = 0;
1714 u64 total_free_data = 0;
1715 u64 total_free_meta = 0;
1716 u32 bits = fs_info->sectorsize_bits;
1717 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1718 unsigned factor = 1;
1719 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1720 int ret;
1721 u64 thresh = 0;
1722 int mixed = 0;
1723
1724 list_for_each_entry(found, &fs_info->space_info, list) {
1725 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1726 int i;
1727
1728 total_free_data += found->disk_total - found->disk_used;
1729 total_free_data -=
1730 btrfs_account_ro_block_groups_free_space(found);
1731
1732 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1733 if (!list_empty(&found->block_groups[i]))
1734 factor = btrfs_bg_type_to_factor(
1735 btrfs_raid_array[i].bg_flag);
1736 }
1737 }
1738
1739 /*
1740 * Metadata in mixed block group profiles are accounted in data
1741 */
1742 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1743 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1744 mixed = 1;
1745 else
1746 total_free_meta += found->disk_total -
1747 found->disk_used;
1748 }
1749
1750 total_used += found->disk_used;
1751 }
1752
1753 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1754 buf->f_blocks >>= bits;
1755 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1756
1757 /* Account global block reserve as used, it's in logical size already */
1758 spin_lock(&block_rsv->lock);
1759 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
1760 if (buf->f_bfree >= block_rsv->size >> bits)
1761 buf->f_bfree -= block_rsv->size >> bits;
1762 else
1763 buf->f_bfree = 0;
1764 spin_unlock(&block_rsv->lock);
1765
1766 buf->f_bavail = div_u64(total_free_data, factor);
1767 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1768 if (ret)
1769 return ret;
1770 buf->f_bavail += div_u64(total_free_data, factor);
1771 buf->f_bavail = buf->f_bavail >> bits;
1772
1773 /*
1774 * We calculate the remaining metadata space minus global reserve. If
1775 * this is (supposedly) smaller than zero, there's no space. But this
1776 * does not hold in practice, the exhausted state happens where's still
1777 * some positive delta. So we apply some guesswork and compare the
1778 * delta to a 4M threshold. (Practically observed delta was ~2M.)
1779 *
1780 * We probably cannot calculate the exact threshold value because this
1781 * depends on the internal reservations requested by various
1782 * operations, so some operations that consume a few metadata will
1783 * succeed even if the Avail is zero. But this is better than the other
1784 * way around.
1785 */
1786 thresh = SZ_4M;
1787
1788 /*
1789 * We only want to claim there's no available space if we can no longer
1790 * allocate chunks for our metadata profile and our global reserve will
1791 * not fit in the free metadata space. If we aren't ->full then we
1792 * still can allocate chunks and thus are fine using the currently
1793 * calculated f_bavail.
1794 */
1795 if (!mixed && block_rsv->space_info->full &&
1796 (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1797 buf->f_bavail = 0;
1798
1799 buf->f_type = BTRFS_SUPER_MAGIC;
1800 buf->f_bsize = fs_info->sectorsize;
1801 buf->f_namelen = BTRFS_NAME_LEN;
1802
1803 /* We treat it as constant endianness (it doesn't matter _which_)
1804 because we want the fsid to come out the same whether mounted
1805 on a big-endian or little-endian host */
1806 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1807 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1808 /* Mask in the root object ID too, to disambiguate subvols */
1809 buf->f_fsid.val[0] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root) >> 32;
1810 buf->f_fsid.val[1] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root);
1811
1812 return 0;
1813 }
1814
1815 static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1816 {
1817 struct btrfs_fs_info *p = fc->s_fs_info;
1818 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1819
1820 return fs_info->fs_devices == p->fs_devices;
1821 }
1822
1823 static int btrfs_get_tree_super(struct fs_context *fc)
1824 {
1825 struct btrfs_fs_info *fs_info = fc->s_fs_info;
1826 struct btrfs_fs_context *ctx = fc->fs_private;
1827 struct btrfs_fs_devices *fs_devices = NULL;
1828 struct block_device *bdev;
1829 struct btrfs_device *device;
1830 struct super_block *sb;
1831 blk_mode_t mode = btrfs_open_mode(fc);
1832 int ret;
1833
1834 btrfs_ctx_to_info(fs_info, ctx);
1835 mutex_lock(&uuid_mutex);
1836
1837 /*
1838 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1839 * either a valid device or an error.
1840 */
1841 device = btrfs_scan_one_device(fc->source, mode, true);
1842 ASSERT(device != NULL);
1843 if (IS_ERR(device)) {
1844 mutex_unlock(&uuid_mutex);
1845 return PTR_ERR(device);
1846 }
1847
1848 fs_devices = device->fs_devices;
1849 fs_info->fs_devices = fs_devices;
1850
1851 ret = btrfs_open_devices(fs_devices, mode, &btrfs_fs_type);
1852 mutex_unlock(&uuid_mutex);
1853 if (ret)
1854 return ret;
1855
1856 if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1857 ret = -EACCES;
1858 goto error;
1859 }
1860
1861 bdev = fs_devices->latest_dev->bdev;
1862
1863 /*
1864 * From now on the error handling is not straightforward.
1865 *
1866 * If successful, this will transfer the fs_info into the super block,
1867 * and fc->s_fs_info will be NULL. However if there's an existing
1868 * super, we'll still have fc->s_fs_info populated. If we error
1869 * completely out it'll be cleaned up when we drop the fs_context,
1870 * otherwise it's tied to the lifetime of the super_block.
1871 */
1872 sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1873 if (IS_ERR(sb)) {
1874 ret = PTR_ERR(sb);
1875 goto error;
1876 }
1877
1878 set_device_specific_options(fs_info);
1879
1880 if (sb->s_root) {
1881 btrfs_close_devices(fs_devices);
1882 if ((fc->sb_flags ^ sb->s_flags) & SB_RDONLY)
1883 ret = -EBUSY;
1884 } else {
1885 snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
1886 shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id);
1887 btrfs_sb(sb)->bdev_holder = &btrfs_fs_type;
1888 ret = btrfs_fill_super(sb, fs_devices, NULL);
1889 }
1890
1891 if (ret) {
1892 deactivate_locked_super(sb);
1893 return ret;
1894 }
1895
1896 btrfs_clear_oneshot_options(fs_info);
1897
1898 fc->root = dget(sb->s_root);
1899 return 0;
1900
1901 error:
1902 btrfs_close_devices(fs_devices);
1903 return ret;
1904 }
1905
1906 /*
1907 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1908 * with different ro/rw options") the following works:
1909 *
1910 * (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1911 * (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1912 *
1913 * which looks nice and innocent but is actually pretty intricate and deserves
1914 * a long comment.
1915 *
1916 * On another filesystem a subvolume mount is close to something like:
1917 *
1918 * (iii) # create rw superblock + initial mount
1919 * mount -t xfs /dev/sdb /opt/
1920 *
1921 * # create ro bind mount
1922 * mount --bind -o ro /opt/foo /mnt/foo
1923 *
1924 * # unmount initial mount
1925 * umount /opt
1926 *
1927 * Of course, there's some special subvolume sauce and there's the fact that the
1928 * sb->s_root dentry is really swapped after mount_subtree(). But conceptually
1929 * it's very close and will help us understand the issue.
1930 *
1931 * The old mount API didn't cleanly distinguish between a mount being made ro
1932 * and a superblock being made ro. The only way to change the ro state of
1933 * either object was by passing ms_rdonly. If a new mount was created via
1934 * mount(2) such as:
1935 *
1936 * mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1937 *
1938 * the MS_RDONLY flag being specified had two effects:
1939 *
1940 * (1) MNT_READONLY was raised -> the resulting mount got
1941 * @mnt->mnt_flags |= MNT_READONLY raised.
1942 *
1943 * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
1944 * made the superblock ro. Note, how SB_RDONLY has the same value as
1945 * ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
1946 *
1947 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
1948 * subtree mounted ro.
1949 *
1950 * But consider the effect on the old mount API on btrfs subvolume mounting
1951 * which combines the distinct step in (iii) into a single step.
1952 *
1953 * By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
1954 * is issued the superblock is ro and thus even if the mount created for (ii) is
1955 * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
1956 * to rw for (ii) which it did using an internal remount call.
1957 *
1958 * IOW, subvolume mounting was inherently complicated due to the ambiguity of
1959 * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
1960 * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
1961 * passed by mount(8) to mount(2).
1962 *
1963 * Enter the new mount API. The new mount API disambiguates making a mount ro
1964 * and making a superblock ro.
1965 *
1966 * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
1967 * fsmount() or mount_setattr() this is a pure VFS level change for a
1968 * specific mount or mount tree that is never seen by the filesystem itself.
1969 *
1970 * (4) To turn a superblock ro the "ro" flag must be used with
1971 * fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
1972 * in fc->sb_flags.
1973 *
1974 * This disambiguation has rather positive consequences. Mounting a subvolume
1975 * ro will not also turn the superblock ro. Only the mount for the subvolume
1976 * will become ro.
1977 *
1978 * So, if the superblock creation request comes from the new mount API the
1979 * caller must have explicitly done:
1980 *
1981 * fsconfig(FSCONFIG_SET_FLAG, "ro")
1982 * fsmount/mount_setattr(MOUNT_ATTR_RDONLY)
1983 *
1984 * IOW, at some point the caller must have explicitly turned the whole
1985 * superblock ro and we shouldn't just undo it like we did for the old mount
1986 * API. In any case, it lets us avoid the hack in the new mount API.
1987 *
1988 * Consequently, the remounting hack must only be used for requests originating
1989 * from the old mount API and should be marked for full deprecation so it can be
1990 * turned off in a couple of years.
1991 *
1992 * The new mount API has no reason to support this hack.
1993 */
1994 static struct vfsmount *btrfs_reconfigure_for_mount(struct fs_context *fc)
1995 {
1996 struct vfsmount *mnt;
1997 int ret;
1998 const bool ro2rw = !(fc->sb_flags & SB_RDONLY);
1999
2000 /*
2001 * We got an EBUSY because our SB_RDONLY flag didn't match the existing
2002 * super block, so invert our setting here and retry the mount so we
2003 * can get our vfsmount.
2004 */
2005 if (ro2rw)
2006 fc->sb_flags |= SB_RDONLY;
2007 else
2008 fc->sb_flags &= ~SB_RDONLY;
2009
2010 mnt = fc_mount(fc);
2011 if (IS_ERR(mnt))
2012 return mnt;
2013
2014 if (!fc->oldapi || !ro2rw)
2015 return mnt;
2016
2017 /* We need to convert to rw, call reconfigure. */
2018 fc->sb_flags &= ~SB_RDONLY;
2019 down_write(&mnt->mnt_sb->s_umount);
2020 ret = btrfs_reconfigure(fc);
2021 up_write(&mnt->mnt_sb->s_umount);
2022 if (ret) {
2023 mntput(mnt);
2024 return ERR_PTR(ret);
2025 }
2026 return mnt;
2027 }
2028
2029 static int btrfs_get_tree_subvol(struct fs_context *fc)
2030 {
2031 struct btrfs_fs_info *fs_info = NULL;
2032 struct btrfs_fs_context *ctx = fc->fs_private;
2033 struct fs_context *dup_fc;
2034 struct dentry *dentry;
2035 struct vfsmount *mnt;
2036
2037 /*
2038 * Setup a dummy root and fs_info for test/set super. This is because
2039 * we don't actually fill this stuff out until open_ctree, but we need
2040 * then open_ctree will properly initialize the file system specific
2041 * settings later. btrfs_init_fs_info initializes the static elements
2042 * of the fs_info (locks and such) to make cleanup easier if we find a
2043 * superblock with our given fs_devices later on at sget() time.
2044 */
2045 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
2046 if (!fs_info)
2047 return -ENOMEM;
2048
2049 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2050 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2051 if (!fs_info->super_copy || !fs_info->super_for_commit) {
2052 btrfs_free_fs_info(fs_info);
2053 return -ENOMEM;
2054 }
2055 btrfs_init_fs_info(fs_info);
2056
2057 dup_fc = vfs_dup_fs_context(fc);
2058 if (IS_ERR(dup_fc)) {
2059 btrfs_free_fs_info(fs_info);
2060 return PTR_ERR(dup_fc);
2061 }
2062
2063 /*
2064 * When we do the sget_fc this gets transferred to the sb, so we only
2065 * need to set it on the dup_fc as this is what creates the super block.
2066 */
2067 dup_fc->s_fs_info = fs_info;
2068
2069 /*
2070 * We'll do the security settings in our btrfs_get_tree_super() mount
2071 * loop, they were duplicated into dup_fc, we can drop the originals
2072 * here.
2073 */
2074 security_free_mnt_opts(&fc->security);
2075 fc->security = NULL;
2076
2077 mnt = fc_mount(dup_fc);
2078 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY)
2079 mnt = btrfs_reconfigure_for_mount(dup_fc);
2080 put_fs_context(dup_fc);
2081 if (IS_ERR(mnt))
2082 return PTR_ERR(mnt);
2083
2084 /*
2085 * This free's ->subvol_name, because if it isn't set we have to
2086 * allocate a buffer to hold the subvol_name, so we just drop our
2087 * reference to it here.
2088 */
2089 dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2090 ctx->subvol_name = NULL;
2091 if (IS_ERR(dentry))
2092 return PTR_ERR(dentry);
2093
2094 fc->root = dentry;
2095 return 0;
2096 }
2097
2098 static int btrfs_get_tree(struct fs_context *fc)
2099 {
2100 /*
2101 * Since we use mount_subtree to mount the default/specified subvol, we
2102 * have to do mounts in two steps.
2103 *
2104 * First pass through we call btrfs_get_tree_subvol(), this is just a
2105 * wrapper around fc_mount() to call back into here again, and this time
2106 * we'll call btrfs_get_tree_super(). This will do the open_ctree() and
2107 * everything to open the devices and file system. Then we return back
2108 * with a fully constructed vfsmount in btrfs_get_tree_subvol(), and
2109 * from there we can do our mount_subvol() call, which will lookup
2110 * whichever subvol we're mounting and setup this fc with the
2111 * appropriate dentry for the subvol.
2112 */
2113 if (fc->s_fs_info)
2114 return btrfs_get_tree_super(fc);
2115 return btrfs_get_tree_subvol(fc);
2116 }
2117
2118 static void btrfs_kill_super(struct super_block *sb)
2119 {
2120 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2121 kill_anon_super(sb);
2122 btrfs_free_fs_info(fs_info);
2123 }
2124
2125 static void btrfs_free_fs_context(struct fs_context *fc)
2126 {
2127 struct btrfs_fs_context *ctx = fc->fs_private;
2128 struct btrfs_fs_info *fs_info = fc->s_fs_info;
2129
2130 if (fs_info)
2131 btrfs_free_fs_info(fs_info);
2132
2133 if (ctx && refcount_dec_and_test(&ctx->refs)) {
2134 kfree(ctx->subvol_name);
2135 kfree(ctx);
2136 }
2137 }
2138
2139 static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2140 {
2141 struct btrfs_fs_context *ctx = src_fc->fs_private;
2142
2143 /*
2144 * Give a ref to our ctx to this dup, as we want to keep it around for
2145 * our original fc so we can have the subvolume name or objectid.
2146 *
2147 * We unset ->source in the original fc because the dup needs it for
2148 * mounting, and then once we free the dup it'll free ->source, so we
2149 * need to make sure we're only pointing to it in one fc.
2150 */
2151 refcount_inc(&ctx->refs);
2152 fc->fs_private = ctx;
2153 fc->source = src_fc->source;
2154 src_fc->source = NULL;
2155 return 0;
2156 }
2157
2158 static const struct fs_context_operations btrfs_fs_context_ops = {
2159 .parse_param = btrfs_parse_param,
2160 .reconfigure = btrfs_reconfigure,
2161 .get_tree = btrfs_get_tree,
2162 .dup = btrfs_dup_fs_context,
2163 .free = btrfs_free_fs_context,
2164 };
2165
2166 static int btrfs_init_fs_context(struct fs_context *fc)
2167 {
2168 struct btrfs_fs_context *ctx;
2169
2170 ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL);
2171 if (!ctx)
2172 return -ENOMEM;
2173
2174 refcount_set(&ctx->refs, 1);
2175 fc->fs_private = ctx;
2176 fc->ops = &btrfs_fs_context_ops;
2177
2178 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2179 btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
2180 } else {
2181 ctx->thread_pool_size =
2182 min_t(unsigned long, num_online_cpus() + 2, 8);
2183 ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2184 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2185 }
2186
2187 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
2188 fc->sb_flags |= SB_POSIXACL;
2189 #endif
2190 fc->sb_flags |= SB_I_VERSION;
2191
2192 return 0;
2193 }
2194
2195 static struct file_system_type btrfs_fs_type = {
2196 .owner = THIS_MODULE,
2197 .name = "btrfs",
2198 .init_fs_context = btrfs_init_fs_context,
2199 .parameters = btrfs_fs_parameters,
2200 .kill_sb = btrfs_kill_super,
2201 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2202 };
2203
2204 MODULE_ALIAS_FS("btrfs");
2205
2206 static int btrfs_control_open(struct inode *inode, struct file *file)
2207 {
2208 /*
2209 * The control file's private_data is used to hold the
2210 * transaction when it is started and is used to keep
2211 * track of whether a transaction is already in progress.
2212 */
2213 file->private_data = NULL;
2214 return 0;
2215 }
2216
2217 /*
2218 * Used by /dev/btrfs-control for devices ioctls.
2219 */
2220 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2221 unsigned long arg)
2222 {
2223 struct btrfs_ioctl_vol_args *vol;
2224 struct btrfs_device *device = NULL;
2225 dev_t devt = 0;
2226 int ret = -ENOTTY;
2227
2228 if (!capable(CAP_SYS_ADMIN))
2229 return -EPERM;
2230
2231 vol = memdup_user((void __user *)arg, sizeof(*vol));
2232 if (IS_ERR(vol))
2233 return PTR_ERR(vol);
2234 ret = btrfs_check_ioctl_vol_args_path(vol);
2235 if (ret < 0)
2236 goto out;
2237
2238 switch (cmd) {
2239 case BTRFS_IOC_SCAN_DEV:
2240 mutex_lock(&uuid_mutex);
2241 /*
2242 * Scanning outside of mount can return NULL which would turn
2243 * into 0 error code.
2244 */
2245 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2246 ret = PTR_ERR_OR_ZERO(device);
2247 mutex_unlock(&uuid_mutex);
2248 break;
2249 case BTRFS_IOC_FORGET_DEV:
2250 if (vol->name[0] != 0) {
2251 ret = lookup_bdev(vol->name, &devt);
2252 if (ret)
2253 break;
2254 }
2255 ret = btrfs_forget_devices(devt);
2256 break;
2257 case BTRFS_IOC_DEVICES_READY:
2258 mutex_lock(&uuid_mutex);
2259 /*
2260 * Scanning outside of mount can return NULL which would turn
2261 * into 0 error code.
2262 */
2263 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2264 if (IS_ERR_OR_NULL(device)) {
2265 mutex_unlock(&uuid_mutex);
2266 ret = PTR_ERR(device);
2267 break;
2268 }
2269 ret = !(device->fs_devices->num_devices ==
2270 device->fs_devices->total_devices);
2271 mutex_unlock(&uuid_mutex);
2272 break;
2273 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2274 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2275 break;
2276 }
2277
2278 out:
2279 kfree(vol);
2280 return ret;
2281 }
2282
2283 static int btrfs_freeze(struct super_block *sb)
2284 {
2285 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2286
2287 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2288 /*
2289 * We don't need a barrier here, we'll wait for any transaction that
2290 * could be in progress on other threads (and do delayed iputs that
2291 * we want to avoid on a frozen filesystem), or do the commit
2292 * ourselves.
2293 */
2294 return btrfs_commit_current_transaction(fs_info->tree_root);
2295 }
2296
2297 static int check_dev_super(struct btrfs_device *dev)
2298 {
2299 struct btrfs_fs_info *fs_info = dev->fs_info;
2300 struct btrfs_super_block *sb;
2301 u64 last_trans;
2302 u16 csum_type;
2303 int ret = 0;
2304
2305 /* This should be called with fs still frozen. */
2306 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2307
2308 /* Missing dev, no need to check. */
2309 if (!dev->bdev)
2310 return 0;
2311
2312 /* Only need to check the primary super block. */
2313 sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2314 if (IS_ERR(sb))
2315 return PTR_ERR(sb);
2316
2317 /* Verify the checksum. */
2318 csum_type = btrfs_super_csum_type(sb);
2319 if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2320 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2321 csum_type, btrfs_super_csum_type(fs_info->super_copy));
2322 ret = -EUCLEAN;
2323 goto out;
2324 }
2325
2326 if (btrfs_check_super_csum(fs_info, sb)) {
2327 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2328 ret = -EUCLEAN;
2329 goto out;
2330 }
2331
2332 /* Btrfs_validate_super() includes fsid check against super->fsid. */
2333 ret = btrfs_validate_super(fs_info, sb, 0);
2334 if (ret < 0)
2335 goto out;
2336
2337 last_trans = btrfs_get_last_trans_committed(fs_info);
2338 if (btrfs_super_generation(sb) != last_trans) {
2339 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2340 btrfs_super_generation(sb), last_trans);
2341 ret = -EUCLEAN;
2342 goto out;
2343 }
2344 out:
2345 btrfs_release_disk_super(sb);
2346 return ret;
2347 }
2348
2349 static int btrfs_unfreeze(struct super_block *sb)
2350 {
2351 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2352 struct btrfs_device *device;
2353 int ret = 0;
2354
2355 /*
2356 * Make sure the fs is not changed by accident (like hibernation then
2357 * modified by other OS).
2358 * If we found anything wrong, we mark the fs error immediately.
2359 *
2360 * And since the fs is frozen, no one can modify the fs yet, thus
2361 * we don't need to hold device_list_mutex.
2362 */
2363 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2364 ret = check_dev_super(device);
2365 if (ret < 0) {
2366 btrfs_handle_fs_error(fs_info, ret,
2367 "super block on devid %llu got modified unexpectedly",
2368 device->devid);
2369 break;
2370 }
2371 }
2372 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2373
2374 /*
2375 * We still return 0, to allow VFS layer to unfreeze the fs even the
2376 * above checks failed. Since the fs is either fine or read-only, we're
2377 * safe to continue, without causing further damage.
2378 */
2379 return 0;
2380 }
2381
2382 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2383 {
2384 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2385
2386 /*
2387 * There should be always a valid pointer in latest_dev, it may be stale
2388 * for a short moment in case it's being deleted but still valid until
2389 * the end of RCU grace period.
2390 */
2391 rcu_read_lock();
2392 seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2393 rcu_read_unlock();
2394
2395 return 0;
2396 }
2397
2398 static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_control *sc)
2399 {
2400 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2401 const s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps);
2402
2403 trace_btrfs_extent_map_shrinker_count(fs_info, nr);
2404
2405 /*
2406 * Only report the real number for DEBUG builds, as there are reports of
2407 * serious performance degradation caused by too frequent shrinks.
2408 */
2409 if (IS_ENABLED(CONFIG_BTRFS_DEBUG))
2410 return nr;
2411 return 0;
2412 }
2413
2414 static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_control *sc)
2415 {
2416 const long nr_to_scan = min_t(unsigned long, LONG_MAX, sc->nr_to_scan);
2417 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2418
2419 /*
2420 * We may be called from any task trying to allocate memory and we don't
2421 * want to slow it down with scanning and dropping extent maps. It would
2422 * also cause heavy lock contention if many tasks concurrently enter
2423 * here. Therefore only allow kswapd tasks to scan and drop extent maps.
2424 */
2425 if (!current_is_kswapd())
2426 return 0;
2427
2428 return btrfs_free_extent_maps(fs_info, nr_to_scan);
2429 }
2430
2431 static const struct super_operations btrfs_super_ops = {
2432 .drop_inode = btrfs_drop_inode,
2433 .evict_inode = btrfs_evict_inode,
2434 .put_super = btrfs_put_super,
2435 .sync_fs = btrfs_sync_fs,
2436 .show_options = btrfs_show_options,
2437 .show_devname = btrfs_show_devname,
2438 .alloc_inode = btrfs_alloc_inode,
2439 .destroy_inode = btrfs_destroy_inode,
2440 .free_inode = btrfs_free_inode,
2441 .statfs = btrfs_statfs,
2442 .freeze_fs = btrfs_freeze,
2443 .unfreeze_fs = btrfs_unfreeze,
2444 .nr_cached_objects = btrfs_nr_cached_objects,
2445 .free_cached_objects = btrfs_free_cached_objects,
2446 };
2447
2448 static const struct file_operations btrfs_ctl_fops = {
2449 .open = btrfs_control_open,
2450 .unlocked_ioctl = btrfs_control_ioctl,
2451 .compat_ioctl = compat_ptr_ioctl,
2452 .owner = THIS_MODULE,
2453 .llseek = noop_llseek,
2454 };
2455
2456 static struct miscdevice btrfs_misc = {
2457 .minor = BTRFS_MINOR,
2458 .name = "btrfs-control",
2459 .fops = &btrfs_ctl_fops
2460 };
2461
2462 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2463 MODULE_ALIAS("devname:btrfs-control");
2464
2465 static int __init btrfs_interface_init(void)
2466 {
2467 return misc_register(&btrfs_misc);
2468 }
2469
2470 static __cold void btrfs_interface_exit(void)
2471 {
2472 misc_deregister(&btrfs_misc);
2473 }
2474
2475 static int __init btrfs_print_mod_info(void)
2476 {
2477 static const char options[] = ""
2478 #ifdef CONFIG_BTRFS_DEBUG
2479 ", debug=on"
2480 #endif
2481 #ifdef CONFIG_BTRFS_ASSERT
2482 ", assert=on"
2483 #endif
2484 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2485 ", ref-verify=on"
2486 #endif
2487 #ifdef CONFIG_BLK_DEV_ZONED
2488 ", zoned=yes"
2489 #else
2490 ", zoned=no"
2491 #endif
2492 #ifdef CONFIG_FS_VERITY
2493 ", fsverity=yes"
2494 #else
2495 ", fsverity=no"
2496 #endif
2497 ;
2498 pr_info("Btrfs loaded%s\n", options);
2499 return 0;
2500 }
2501
2502 static int register_btrfs(void)
2503 {
2504 return register_filesystem(&btrfs_fs_type);
2505 }
2506
2507 static void unregister_btrfs(void)
2508 {
2509 unregister_filesystem(&btrfs_fs_type);
2510 }
2511
2512 /* Helper structure for long init/exit functions. */
2513 struct init_sequence {
2514 int (*init_func)(void);
2515 /* Can be NULL if the init_func doesn't need cleanup. */
2516 void (*exit_func)(void);
2517 };
2518
2519 static const struct init_sequence mod_init_seq[] = {
2520 {
2521 .init_func = btrfs_props_init,
2522 .exit_func = NULL,
2523 }, {
2524 .init_func = btrfs_init_sysfs,
2525 .exit_func = btrfs_exit_sysfs,
2526 }, {
2527 .init_func = btrfs_init_compress,
2528 .exit_func = btrfs_exit_compress,
2529 }, {
2530 .init_func = btrfs_init_cachep,
2531 .exit_func = btrfs_destroy_cachep,
2532 }, {
2533 .init_func = btrfs_init_dio,
2534 .exit_func = btrfs_destroy_dio,
2535 }, {
2536 .init_func = btrfs_transaction_init,
2537 .exit_func = btrfs_transaction_exit,
2538 }, {
2539 .init_func = btrfs_ctree_init,
2540 .exit_func = btrfs_ctree_exit,
2541 }, {
2542 .init_func = btrfs_free_space_init,
2543 .exit_func = btrfs_free_space_exit,
2544 }, {
2545 .init_func = extent_state_init_cachep,
2546 .exit_func = extent_state_free_cachep,
2547 }, {
2548 .init_func = extent_buffer_init_cachep,
2549 .exit_func = extent_buffer_free_cachep,
2550 }, {
2551 .init_func = btrfs_bioset_init,
2552 .exit_func = btrfs_bioset_exit,
2553 }, {
2554 .init_func = extent_map_init,
2555 .exit_func = extent_map_exit,
2556 }, {
2557 .init_func = ordered_data_init,
2558 .exit_func = ordered_data_exit,
2559 }, {
2560 .init_func = btrfs_delayed_inode_init,
2561 .exit_func = btrfs_delayed_inode_exit,
2562 }, {
2563 .init_func = btrfs_auto_defrag_init,
2564 .exit_func = btrfs_auto_defrag_exit,
2565 }, {
2566 .init_func = btrfs_delayed_ref_init,
2567 .exit_func = btrfs_delayed_ref_exit,
2568 }, {
2569 .init_func = btrfs_prelim_ref_init,
2570 .exit_func = btrfs_prelim_ref_exit,
2571 }, {
2572 .init_func = btrfs_interface_init,
2573 .exit_func = btrfs_interface_exit,
2574 }, {
2575 .init_func = btrfs_print_mod_info,
2576 .exit_func = NULL,
2577 }, {
2578 .init_func = btrfs_run_sanity_tests,
2579 .exit_func = NULL,
2580 }, {
2581 .init_func = register_btrfs,
2582 .exit_func = unregister_btrfs,
2583 }
2584 };
2585
2586 static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2587
2588 static __always_inline void btrfs_exit_btrfs_fs(void)
2589 {
2590 int i;
2591
2592 for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2593 if (!mod_init_result[i])
2594 continue;
2595 if (mod_init_seq[i].exit_func)
2596 mod_init_seq[i].exit_func();
2597 mod_init_result[i] = false;
2598 }
2599 }
2600
2601 static void __exit exit_btrfs_fs(void)
2602 {
2603 btrfs_exit_btrfs_fs();
2604 btrfs_cleanup_fs_uuids();
2605 }
2606
2607 static int __init init_btrfs_fs(void)
2608 {
2609 int ret;
2610 int i;
2611
2612 for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2613 ASSERT(!mod_init_result[i]);
2614 ret = mod_init_seq[i].init_func();
2615 if (ret < 0) {
2616 btrfs_exit_btrfs_fs();
2617 return ret;
2618 }
2619 mod_init_result[i] = true;
2620 }
2621 return 0;
2622 }
2623
2624 late_initcall(init_btrfs_fs);
2625 module_exit(exit_btrfs_fs)
2626
2627 MODULE_DESCRIPTION("B-Tree File System (BTRFS)");
2628 MODULE_LICENSE("GPL");
2629 MODULE_SOFTDEP("pre: crc32c");
2630 MODULE_SOFTDEP("pre: xxhash64");
2631 MODULE_SOFTDEP("pre: sha256");
2632 MODULE_SOFTDEP("pre: blake2b-256");
2633
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.