Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (C) 2007 Oracle. All rights reserved.
4 */
5
6#include <linux/blkdev.h>
7#include <linux/module.h>
8#include <linux/fs.h>
9#include <linux/pagemap.h>
10#include <linux/highmem.h>
11#include <linux/time.h>
12#include <linux/init.h>
13#include <linux/seq_file.h>
14#include <linux/string.h>
15#include <linux/backing-dev.h>
16#include <linux/mount.h>
17#include <linux/writeback.h>
18#include <linux/statfs.h>
19#include <linux/compat.h>
20#include <linux/parser.h>
21#include <linux/ctype.h>
22#include <linux/namei.h>
23#include <linux/miscdevice.h>
24#include <linux/magic.h>
25#include <linux/slab.h>
26#include <linux/ratelimit.h>
27#include <linux/crc32c.h>
28#include <linux/btrfs.h>
29#include <linux/security.h>
30#include <linux/fs_parser.h>
31#include "messages.h"
32#include "delayed-inode.h"
33#include "ctree.h"
34#include "disk-io.h"
35#include "transaction.h"
36#include "btrfs_inode.h"
37#include "direct-io.h"
38#include "props.h"
39#include "xattr.h"
40#include "bio.h"
41#include "export.h"
42#include "compression.h"
43#include "dev-replace.h"
44#include "free-space-cache.h"
45#include "backref.h"
46#include "space-info.h"
47#include "sysfs.h"
48#include "zoned.h"
49#include "tests/btrfs-tests.h"
50#include "block-group.h"
51#include "discard.h"
52#include "qgroup.h"
53#include "raid56.h"
54#include "fs.h"
55#include "accessors.h"
56#include "defrag.h"
57#include "dir-item.h"
58#include "ioctl.h"
59#include "scrub.h"
60#include "verity.h"
61#include "super.h"
62#include "extent-tree.h"
63#define CREATE_TRACE_POINTS
64#include <trace/events/btrfs.h>
65
66static const struct super_operations btrfs_super_ops;
67static struct file_system_type btrfs_fs_type;
68
69static void btrfs_put_super(struct super_block *sb)
70{
71 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
72
73 btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid);
74 close_ctree(fs_info);
75}
76
77/* Store the mount options related information. */
78struct btrfs_fs_context {
79 char *subvol_name;
80 u64 subvol_objectid;
81 u64 max_inline;
82 u32 commit_interval;
83 u32 metadata_ratio;
84 u32 thread_pool_size;
85 unsigned long long mount_opt;
86 unsigned long compress_type:4;
87 unsigned int compress_level;
88 refcount_t refs;
89};
90
91enum {
92 Opt_acl,
93 Opt_clear_cache,
94 Opt_commit_interval,
95 Opt_compress,
96 Opt_compress_force,
97 Opt_compress_force_type,
98 Opt_compress_type,
99 Opt_degraded,
100 Opt_device,
101 Opt_fatal_errors,
102 Opt_flushoncommit,
103 Opt_max_inline,
104 Opt_barrier,
105 Opt_datacow,
106 Opt_datasum,
107 Opt_defrag,
108 Opt_discard,
109 Opt_discard_mode,
110 Opt_ratio,
111 Opt_rescan_uuid_tree,
112 Opt_skip_balance,
113 Opt_space_cache,
114 Opt_space_cache_version,
115 Opt_ssd,
116 Opt_ssd_spread,
117 Opt_subvol,
118 Opt_subvol_empty,
119 Opt_subvolid,
120 Opt_thread_pool,
121 Opt_treelog,
122 Opt_user_subvol_rm_allowed,
123 Opt_norecovery,
124
125 /* Rescue options */
126 Opt_rescue,
127 Opt_usebackuproot,
128 Opt_nologreplay,
129
130 /* Debugging options */
131 Opt_enospc_debug,
132#ifdef CONFIG_BTRFS_DEBUG
133 Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
134#endif
135#ifdef CONFIG_BTRFS_FS_REF_VERIFY
136 Opt_ref_verify,
137#endif
138 Opt_err,
139};
140
141enum {
142 Opt_fatal_errors_panic,
143 Opt_fatal_errors_bug,
144};
145
146static const struct constant_table btrfs_parameter_fatal_errors[] = {
147 { "panic", Opt_fatal_errors_panic },
148 { "bug", Opt_fatal_errors_bug },
149 {}
150};
151
152enum {
153 Opt_discard_sync,
154 Opt_discard_async,
155};
156
157static const struct constant_table btrfs_parameter_discard[] = {
158 { "sync", Opt_discard_sync },
159 { "async", Opt_discard_async },
160 {}
161};
162
163enum {
164 Opt_space_cache_v1,
165 Opt_space_cache_v2,
166};
167
168static const struct constant_table btrfs_parameter_space_cache[] = {
169 { "v1", Opt_space_cache_v1 },
170 { "v2", Opt_space_cache_v2 },
171 {}
172};
173
174enum {
175 Opt_rescue_usebackuproot,
176 Opt_rescue_nologreplay,
177 Opt_rescue_ignorebadroots,
178 Opt_rescue_ignoredatacsums,
179 Opt_rescue_ignoremetacsums,
180 Opt_rescue_ignoresuperflags,
181 Opt_rescue_parameter_all,
182};
183
184static const struct constant_table btrfs_parameter_rescue[] = {
185 { "usebackuproot", Opt_rescue_usebackuproot },
186 { "nologreplay", Opt_rescue_nologreplay },
187 { "ignorebadroots", Opt_rescue_ignorebadroots },
188 { "ibadroots", Opt_rescue_ignorebadroots },
189 { "ignoredatacsums", Opt_rescue_ignoredatacsums },
190 { "ignoremetacsums", Opt_rescue_ignoremetacsums},
191 { "ignoresuperflags", Opt_rescue_ignoresuperflags},
192 { "idatacsums", Opt_rescue_ignoredatacsums },
193 { "imetacsums", Opt_rescue_ignoremetacsums},
194 { "isuperflags", Opt_rescue_ignoresuperflags},
195 { "all", Opt_rescue_parameter_all },
196 {}
197};
198
199#ifdef CONFIG_BTRFS_DEBUG
200enum {
201 Opt_fragment_parameter_data,
202 Opt_fragment_parameter_metadata,
203 Opt_fragment_parameter_all,
204};
205
206static const struct constant_table btrfs_parameter_fragment[] = {
207 { "data", Opt_fragment_parameter_data },
208 { "metadata", Opt_fragment_parameter_metadata },
209 { "all", Opt_fragment_parameter_all },
210 {}
211};
212#endif
213
214static const struct fs_parameter_spec btrfs_fs_parameters[] = {
215 fsparam_flag_no("acl", Opt_acl),
216 fsparam_flag_no("autodefrag", Opt_defrag),
217 fsparam_flag_no("barrier", Opt_barrier),
218 fsparam_flag("clear_cache", Opt_clear_cache),
219 fsparam_u32("commit", Opt_commit_interval),
220 fsparam_flag("compress", Opt_compress),
221 fsparam_string("compress", Opt_compress_type),
222 fsparam_flag("compress-force", Opt_compress_force),
223 fsparam_string("compress-force", Opt_compress_force_type),
224 fsparam_flag_no("datacow", Opt_datacow),
225 fsparam_flag_no("datasum", Opt_datasum),
226 fsparam_flag("degraded", Opt_degraded),
227 fsparam_string("device", Opt_device),
228 fsparam_flag_no("discard", Opt_discard),
229 fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard),
230 fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors),
231 fsparam_flag_no("flushoncommit", Opt_flushoncommit),
232 fsparam_string("max_inline", Opt_max_inline),
233 fsparam_u32("metadata_ratio", Opt_ratio),
234 fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree),
235 fsparam_flag("skip_balance", Opt_skip_balance),
236 fsparam_flag_no("space_cache", Opt_space_cache),
237 fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache),
238 fsparam_flag_no("ssd", Opt_ssd),
239 fsparam_flag_no("ssd_spread", Opt_ssd_spread),
240 fsparam_string("subvol", Opt_subvol),
241 fsparam_flag("subvol=", Opt_subvol_empty),
242 fsparam_u64("subvolid", Opt_subvolid),
243 fsparam_u32("thread_pool", Opt_thread_pool),
244 fsparam_flag_no("treelog", Opt_treelog),
245 fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed),
246
247 /* Rescue options. */
248 fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue),
249 /* Deprecated, with alias rescue=nologreplay */
250 __fsparam(NULL, "nologreplay", Opt_nologreplay, fs_param_deprecated, NULL),
251 /* Deprecated, with alias rescue=usebackuproot */
252 __fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL),
253 /* For compatibility only, alias for "rescue=nologreplay". */
254 fsparam_flag("norecovery", Opt_norecovery),
255
256 /* Debugging options. */
257 fsparam_flag_no("enospc_debug", Opt_enospc_debug),
258#ifdef CONFIG_BTRFS_DEBUG
259 fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment),
260#endif
261#ifdef CONFIG_BTRFS_FS_REF_VERIFY
262 fsparam_flag("ref_verify", Opt_ref_verify),
263#endif
264 {}
265};
266
267/* No support for restricting writes to btrfs devices yet... */
268static inline blk_mode_t btrfs_open_mode(struct fs_context *fc)
269{
270 return sb_open_mode(fc->sb_flags) & ~BLK_OPEN_RESTRICT_WRITES;
271}
272
273static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
274{
275 struct btrfs_fs_context *ctx = fc->fs_private;
276 struct fs_parse_result result;
277 int opt;
278
279 opt = fs_parse(fc, btrfs_fs_parameters, param, &result);
280 if (opt < 0)
281 return opt;
282
283 switch (opt) {
284 case Opt_degraded:
285 btrfs_set_opt(ctx->mount_opt, DEGRADED);
286 break;
287 case Opt_subvol_empty:
288 /*
289 * This exists because we used to allow it on accident, so we're
290 * keeping it to maintain ABI. See 37becec95ac3 ("Btrfs: allow
291 * empty subvol= again").
292 */
293 break;
294 case Opt_subvol:
295 kfree(ctx->subvol_name);
296 ctx->subvol_name = kstrdup(param->string, GFP_KERNEL);
297 if (!ctx->subvol_name)
298 return -ENOMEM;
299 break;
300 case Opt_subvolid:
301 ctx->subvol_objectid = result.uint_64;
302
303 /* subvolid=0 means give me the original fs_tree. */
304 if (!ctx->subvol_objectid)
305 ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
306 break;
307 case Opt_device: {
308 struct btrfs_device *device;
309 blk_mode_t mode = btrfs_open_mode(fc);
310
311 mutex_lock(&uuid_mutex);
312 device = btrfs_scan_one_device(param->string, mode, false);
313 mutex_unlock(&uuid_mutex);
314 if (IS_ERR(device))
315 return PTR_ERR(device);
316 break;
317 }
318 case Opt_datasum:
319 if (result.negated) {
320 btrfs_set_opt(ctx->mount_opt, NODATASUM);
321 } else {
322 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
323 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
324 }
325 break;
326 case Opt_datacow:
327 if (result.negated) {
328 btrfs_clear_opt(ctx->mount_opt, COMPRESS);
329 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
330 btrfs_set_opt(ctx->mount_opt, NODATACOW);
331 btrfs_set_opt(ctx->mount_opt, NODATASUM);
332 } else {
333 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
334 }
335 break;
336 case Opt_compress_force:
337 case Opt_compress_force_type:
338 btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS);
339 fallthrough;
340 case Opt_compress:
341 case Opt_compress_type:
342 /*
343 * Provide the same semantics as older kernels that don't use fs
344 * context, specifying the "compress" option clears
345 * "force-compress" without the need to pass
346 * "compress-force=[no|none]" before specifying "compress".
347 */
348 if (opt != Opt_compress_force && opt != Opt_compress_force_type)
349 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
350
351 if (opt == Opt_compress || opt == Opt_compress_force) {
352 ctx->compress_type = BTRFS_COMPRESS_ZLIB;
353 ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
354 btrfs_set_opt(ctx->mount_opt, COMPRESS);
355 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
356 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
357 } else if (strncmp(param->string, "zlib", 4) == 0) {
358 ctx->compress_type = BTRFS_COMPRESS_ZLIB;
359 ctx->compress_level =
360 btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB,
361 param->string + 4);
362 btrfs_set_opt(ctx->mount_opt, COMPRESS);
363 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
364 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
365 } else if (strncmp(param->string, "lzo", 3) == 0) {
366 ctx->compress_type = BTRFS_COMPRESS_LZO;
367 ctx->compress_level = 0;
368 btrfs_set_opt(ctx->mount_opt, COMPRESS);
369 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
370 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
371 } else if (strncmp(param->string, "zstd", 4) == 0) {
372 ctx->compress_type = BTRFS_COMPRESS_ZSTD;
373 ctx->compress_level =
374 btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD,
375 param->string + 4);
376 btrfs_set_opt(ctx->mount_opt, COMPRESS);
377 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
378 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
379 } else if (strncmp(param->string, "no", 2) == 0) {
380 ctx->compress_level = 0;
381 ctx->compress_type = 0;
382 btrfs_clear_opt(ctx->mount_opt, COMPRESS);
383 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
384 } else {
385 btrfs_err(NULL, "unrecognized compression value %s",
386 param->string);
387 return -EINVAL;
388 }
389 break;
390 case Opt_ssd:
391 if (result.negated) {
392 btrfs_set_opt(ctx->mount_opt, NOSSD);
393 btrfs_clear_opt(ctx->mount_opt, SSD);
394 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
395 } else {
396 btrfs_set_opt(ctx->mount_opt, SSD);
397 btrfs_clear_opt(ctx->mount_opt, NOSSD);
398 }
399 break;
400 case Opt_ssd_spread:
401 if (result.negated) {
402 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
403 } else {
404 btrfs_set_opt(ctx->mount_opt, SSD);
405 btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
406 btrfs_clear_opt(ctx->mount_opt, NOSSD);
407 }
408 break;
409 case Opt_barrier:
410 if (result.negated)
411 btrfs_set_opt(ctx->mount_opt, NOBARRIER);
412 else
413 btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
414 break;
415 case Opt_thread_pool:
416 if (result.uint_32 == 0) {
417 btrfs_err(NULL, "invalid value 0 for thread_pool");
418 return -EINVAL;
419 }
420 ctx->thread_pool_size = result.uint_32;
421 break;
422 case Opt_max_inline:
423 ctx->max_inline = memparse(param->string, NULL);
424 break;
425 case Opt_acl:
426 if (result.negated) {
427 fc->sb_flags &= ~SB_POSIXACL;
428 } else {
429#ifdef CONFIG_BTRFS_FS_POSIX_ACL
430 fc->sb_flags |= SB_POSIXACL;
431#else
432 btrfs_err(NULL, "support for ACL not compiled in");
433 return -EINVAL;
434#endif
435 }
436 /*
437 * VFS limits the ability to toggle ACL on and off via remount,
438 * despite every file system allowing this. This seems to be
439 * an oversight since we all do, but it'll fail if we're
440 * remounting. So don't set the mask here, we'll check it in
441 * btrfs_reconfigure and do the toggling ourselves.
442 */
443 if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
444 fc->sb_flags_mask |= SB_POSIXACL;
445 break;
446 case Opt_treelog:
447 if (result.negated)
448 btrfs_set_opt(ctx->mount_opt, NOTREELOG);
449 else
450 btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
451 break;
452 case Opt_nologreplay:
453 btrfs_warn(NULL,
454 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
455 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
456 break;
457 case Opt_norecovery:
458 btrfs_info(NULL,
459"'norecovery' is for compatibility only, recommended to use 'rescue=nologreplay'");
460 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
461 break;
462 case Opt_flushoncommit:
463 if (result.negated)
464 btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
465 else
466 btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
467 break;
468 case Opt_ratio:
469 ctx->metadata_ratio = result.uint_32;
470 break;
471 case Opt_discard:
472 if (result.negated) {
473 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
474 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
475 btrfs_set_opt(ctx->mount_opt, NODISCARD);
476 } else {
477 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
478 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
479 }
480 break;
481 case Opt_discard_mode:
482 switch (result.uint_32) {
483 case Opt_discard_sync:
484 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
485 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
486 break;
487 case Opt_discard_async:
488 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
489 btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
490 break;
491 default:
492 btrfs_err(NULL, "unrecognized discard mode value %s",
493 param->key);
494 return -EINVAL;
495 }
496 btrfs_clear_opt(ctx->mount_opt, NODISCARD);
497 break;
498 case Opt_space_cache:
499 if (result.negated) {
500 btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
501 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
502 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
503 } else {
504 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
505 btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
506 }
507 break;
508 case Opt_space_cache_version:
509 switch (result.uint_32) {
510 case Opt_space_cache_v1:
511 btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
512 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
513 break;
514 case Opt_space_cache_v2:
515 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
516 btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
517 break;
518 default:
519 btrfs_err(NULL, "unrecognized space_cache value %s",
520 param->key);
521 return -EINVAL;
522 }
523 break;
524 case Opt_rescan_uuid_tree:
525 btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
526 break;
527 case Opt_clear_cache:
528 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
529 break;
530 case Opt_user_subvol_rm_allowed:
531 btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
532 break;
533 case Opt_enospc_debug:
534 if (result.negated)
535 btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
536 else
537 btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
538 break;
539 case Opt_defrag:
540 if (result.negated)
541 btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
542 else
543 btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
544 break;
545 case Opt_usebackuproot:
546 btrfs_warn(NULL,
547 "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
548 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
549
550 /* If we're loading the backup roots we can't trust the space cache. */
551 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
552 break;
553 case Opt_skip_balance:
554 btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
555 break;
556 case Opt_fatal_errors:
557 switch (result.uint_32) {
558 case Opt_fatal_errors_panic:
559 btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
560 break;
561 case Opt_fatal_errors_bug:
562 btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
563 break;
564 default:
565 btrfs_err(NULL, "unrecognized fatal_errors value %s",
566 param->key);
567 return -EINVAL;
568 }
569 break;
570 case Opt_commit_interval:
571 ctx->commit_interval = result.uint_32;
572 if (ctx->commit_interval == 0)
573 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
574 break;
575 case Opt_rescue:
576 switch (result.uint_32) {
577 case Opt_rescue_usebackuproot:
578 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
579 break;
580 case Opt_rescue_nologreplay:
581 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
582 break;
583 case Opt_rescue_ignorebadroots:
584 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
585 break;
586 case Opt_rescue_ignoredatacsums:
587 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
588 break;
589 case Opt_rescue_ignoremetacsums:
590 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
591 break;
592 case Opt_rescue_ignoresuperflags:
593 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
594 break;
595 case Opt_rescue_parameter_all:
596 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
597 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
598 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
599 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
600 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
601 break;
602 default:
603 btrfs_info(NULL, "unrecognized rescue option '%s'",
604 param->key);
605 return -EINVAL;
606 }
607 break;
608#ifdef CONFIG_BTRFS_DEBUG
609 case Opt_fragment:
610 switch (result.uint_32) {
611 case Opt_fragment_parameter_all:
612 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
613 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
614 break;
615 case Opt_fragment_parameter_metadata:
616 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
617 break;
618 case Opt_fragment_parameter_data:
619 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
620 break;
621 default:
622 btrfs_info(NULL, "unrecognized fragment option '%s'",
623 param->key);
624 return -EINVAL;
625 }
626 break;
627#endif
628#ifdef CONFIG_BTRFS_FS_REF_VERIFY
629 case Opt_ref_verify:
630 btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
631 break;
632#endif
633 default:
634 btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
635 return -EINVAL;
636 }
637
638 return 0;
639}
640
641/*
642 * Some options only have meaning at mount time and shouldn't persist across
643 * remounts, or be displayed. Clear these at the end of mount and remount code
644 * paths.
645 */
646static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
647{
648 btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
649 btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
650 btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
651}
652
653static bool check_ro_option(const struct btrfs_fs_info *fs_info,
654 unsigned long long mount_opt, unsigned long long opt,
655 const char *opt_name)
656{
657 if (mount_opt & opt) {
658 btrfs_err(fs_info, "%s must be used with ro mount option",
659 opt_name);
660 return true;
661 }
662 return false;
663}
664
665bool btrfs_check_options(const struct btrfs_fs_info *info,
666 unsigned long long *mount_opt,
667 unsigned long flags)
668{
669 bool ret = true;
670
671 if (!(flags & SB_RDONLY) &&
672 (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
673 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
674 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums") ||
675 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREMETACSUMS, "ignoremetacsums") ||
676 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNORESUPERFLAGS, "ignoresuperflags")))
677 ret = false;
678
679 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
680 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
681 !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
682 btrfs_err(info, "cannot disable free-space-tree");
683 ret = false;
684 }
685 if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
686 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
687 btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
688 ret = false;
689 }
690
691 if (btrfs_check_mountopts_zoned(info, mount_opt))
692 ret = false;
693
694 if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
695 if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE)) {
696 btrfs_info(info, "disk space caching is enabled");
697 btrfs_warn(info,
698"space cache v1 is being deprecated and will be removed in a future release, please use -o space_cache=v2");
699 }
700 if (btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE))
701 btrfs_info(info, "using free-space-tree");
702 }
703
704 return ret;
705}
706
707/*
708 * This is subtle, we only call this during open_ctree(). We need to pre-load
709 * the mount options with the on-disk settings. Before the new mount API took
710 * effect we would do this on mount and remount. With the new mount API we'll
711 * only do this on the initial mount.
712 *
713 * This isn't a change in behavior, because we're using the current state of the
714 * file system to set the current mount options. If you mounted with special
715 * options to disable these features and then remounted we wouldn't revert the
716 * settings, because mounting without these features cleared the on-disk
717 * settings, so this being called on re-mount is not needed.
718 */
719void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
720{
721 if (fs_info->sectorsize < PAGE_SIZE) {
722 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
723 if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) {
724 btrfs_info(fs_info,
725 "forcing free space tree for sector size %u with page size %lu",
726 fs_info->sectorsize, PAGE_SIZE);
727 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
728 }
729 }
730
731 /*
732 * At this point our mount options are populated, so we only mess with
733 * these settings if we don't have any settings already.
734 */
735 if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
736 return;
737
738 if (btrfs_is_zoned(fs_info) &&
739 btrfs_free_space_cache_v1_active(fs_info)) {
740 btrfs_info(fs_info, "zoned: clearing existing space cache");
741 btrfs_set_super_cache_generation(fs_info->super_copy, 0);
742 return;
743 }
744
745 if (btrfs_test_opt(fs_info, SPACE_CACHE))
746 return;
747
748 if (btrfs_test_opt(fs_info, NOSPACECACHE))
749 return;
750
751 /*
752 * At this point we don't have explicit options set by the user, set
753 * them ourselves based on the state of the file system.
754 */
755 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
756 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
757 else if (btrfs_free_space_cache_v1_active(fs_info))
758 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
759}
760
761static void set_device_specific_options(struct btrfs_fs_info *fs_info)
762{
763 if (!btrfs_test_opt(fs_info, NOSSD) &&
764 !fs_info->fs_devices->rotating)
765 btrfs_set_opt(fs_info->mount_opt, SSD);
766
767 /*
768 * For devices supporting discard turn on discard=async automatically,
769 * unless it's already set or disabled. This could be turned off by
770 * nodiscard for the same mount.
771 *
772 * The zoned mode piggy backs on the discard functionality for
773 * resetting a zone. There is no reason to delay the zone reset as it is
774 * fast enough. So, do not enable async discard for zoned mode.
775 */
776 if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
777 btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
778 btrfs_test_opt(fs_info, NODISCARD)) &&
779 fs_info->fs_devices->discardable &&
780 !btrfs_is_zoned(fs_info))
781 btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
782}
783
784char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
785 u64 subvol_objectid)
786{
787 struct btrfs_root *root = fs_info->tree_root;
788 struct btrfs_root *fs_root = NULL;
789 struct btrfs_root_ref *root_ref;
790 struct btrfs_inode_ref *inode_ref;
791 struct btrfs_key key;
792 struct btrfs_path *path = NULL;
793 char *name = NULL, *ptr;
794 u64 dirid;
795 int len;
796 int ret;
797
798 path = btrfs_alloc_path();
799 if (!path) {
800 ret = -ENOMEM;
801 goto err;
802 }
803
804 name = kmalloc(PATH_MAX, GFP_KERNEL);
805 if (!name) {
806 ret = -ENOMEM;
807 goto err;
808 }
809 ptr = name + PATH_MAX - 1;
810 ptr[0] = '\0';
811
812 /*
813 * Walk up the subvolume trees in the tree of tree roots by root
814 * backrefs until we hit the top-level subvolume.
815 */
816 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
817 key.objectid = subvol_objectid;
818 key.type = BTRFS_ROOT_BACKREF_KEY;
819 key.offset = (u64)-1;
820
821 ret = btrfs_search_backwards(root, &key, path);
822 if (ret < 0) {
823 goto err;
824 } else if (ret > 0) {
825 ret = -ENOENT;
826 goto err;
827 }
828
829 subvol_objectid = key.offset;
830
831 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
832 struct btrfs_root_ref);
833 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
834 ptr -= len + 1;
835 if (ptr < name) {
836 ret = -ENAMETOOLONG;
837 goto err;
838 }
839 read_extent_buffer(path->nodes[0], ptr + 1,
840 (unsigned long)(root_ref + 1), len);
841 ptr[0] = '/';
842 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
843 btrfs_release_path(path);
844
845 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
846 if (IS_ERR(fs_root)) {
847 ret = PTR_ERR(fs_root);
848 fs_root = NULL;
849 goto err;
850 }
851
852 /*
853 * Walk up the filesystem tree by inode refs until we hit the
854 * root directory.
855 */
856 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
857 key.objectid = dirid;
858 key.type = BTRFS_INODE_REF_KEY;
859 key.offset = (u64)-1;
860
861 ret = btrfs_search_backwards(fs_root, &key, path);
862 if (ret < 0) {
863 goto err;
864 } else if (ret > 0) {
865 ret = -ENOENT;
866 goto err;
867 }
868
869 dirid = key.offset;
870
871 inode_ref = btrfs_item_ptr(path->nodes[0],
872 path->slots[0],
873 struct btrfs_inode_ref);
874 len = btrfs_inode_ref_name_len(path->nodes[0],
875 inode_ref);
876 ptr -= len + 1;
877 if (ptr < name) {
878 ret = -ENAMETOOLONG;
879 goto err;
880 }
881 read_extent_buffer(path->nodes[0], ptr + 1,
882 (unsigned long)(inode_ref + 1), len);
883 ptr[0] = '/';
884 btrfs_release_path(path);
885 }
886 btrfs_put_root(fs_root);
887 fs_root = NULL;
888 }
889
890 btrfs_free_path(path);
891 if (ptr == name + PATH_MAX - 1) {
892 name[0] = '/';
893 name[1] = '\0';
894 } else {
895 memmove(name, ptr, name + PATH_MAX - ptr);
896 }
897 return name;
898
899err:
900 btrfs_put_root(fs_root);
901 btrfs_free_path(path);
902 kfree(name);
903 return ERR_PTR(ret);
904}
905
906static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
907{
908 struct btrfs_root *root = fs_info->tree_root;
909 struct btrfs_dir_item *di;
910 struct btrfs_path *path;
911 struct btrfs_key location;
912 struct fscrypt_str name = FSTR_INIT("default", 7);
913 u64 dir_id;
914
915 path = btrfs_alloc_path();
916 if (!path)
917 return -ENOMEM;
918
919 /*
920 * Find the "default" dir item which points to the root item that we
921 * will mount by default if we haven't been given a specific subvolume
922 * to mount.
923 */
924 dir_id = btrfs_super_root_dir(fs_info->super_copy);
925 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
926 if (IS_ERR(di)) {
927 btrfs_free_path(path);
928 return PTR_ERR(di);
929 }
930 if (!di) {
931 /*
932 * Ok the default dir item isn't there. This is weird since
933 * it's always been there, but don't freak out, just try and
934 * mount the top-level subvolume.
935 */
936 btrfs_free_path(path);
937 *objectid = BTRFS_FS_TREE_OBJECTID;
938 return 0;
939 }
940
941 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
942 btrfs_free_path(path);
943 *objectid = location.objectid;
944 return 0;
945}
946
947static int btrfs_fill_super(struct super_block *sb,
948 struct btrfs_fs_devices *fs_devices)
949{
950 struct inode *inode;
951 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
952 int err;
953
954 sb->s_maxbytes = MAX_LFS_FILESIZE;
955 sb->s_magic = BTRFS_SUPER_MAGIC;
956 sb->s_op = &btrfs_super_ops;
957 sb->s_d_op = &btrfs_dentry_operations;
958 sb->s_export_op = &btrfs_export_ops;
959#ifdef CONFIG_FS_VERITY
960 sb->s_vop = &btrfs_verityops;
961#endif
962 sb->s_xattr = btrfs_xattr_handlers;
963 sb->s_time_gran = 1;
964 sb->s_iflags |= SB_I_CGROUPWB;
965
966 err = super_setup_bdi(sb);
967 if (err) {
968 btrfs_err(fs_info, "super_setup_bdi failed");
969 return err;
970 }
971
972 err = open_ctree(sb, fs_devices);
973 if (err) {
974 btrfs_err(fs_info, "open_ctree failed: %d", err);
975 return err;
976 }
977
978 inode = btrfs_iget(BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
979 if (IS_ERR(inode)) {
980 err = PTR_ERR(inode);
981 btrfs_handle_fs_error(fs_info, err, NULL);
982 goto fail_close;
983 }
984
985 sb->s_root = d_make_root(inode);
986 if (!sb->s_root) {
987 err = -ENOMEM;
988 goto fail_close;
989 }
990
991 sb->s_flags |= SB_ACTIVE;
992 return 0;
993
994fail_close:
995 close_ctree(fs_info);
996 return err;
997}
998
999int btrfs_sync_fs(struct super_block *sb, int wait)
1000{
1001 struct btrfs_trans_handle *trans;
1002 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1003 struct btrfs_root *root = fs_info->tree_root;
1004
1005 trace_btrfs_sync_fs(fs_info, wait);
1006
1007 if (!wait) {
1008 filemap_flush(fs_info->btree_inode->i_mapping);
1009 return 0;
1010 }
1011
1012 btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL);
1013
1014 trans = btrfs_attach_transaction_barrier(root);
1015 if (IS_ERR(trans)) {
1016 /* no transaction, don't bother */
1017 if (PTR_ERR(trans) == -ENOENT) {
1018 /*
1019 * Exit unless we have some pending changes
1020 * that need to go through commit
1021 */
1022 if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1023 &fs_info->flags))
1024 return 0;
1025 /*
1026 * A non-blocking test if the fs is frozen. We must not
1027 * start a new transaction here otherwise a deadlock
1028 * happens. The pending operations are delayed to the
1029 * next commit after thawing.
1030 */
1031 if (sb_start_write_trylock(sb))
1032 sb_end_write(sb);
1033 else
1034 return 0;
1035 trans = btrfs_start_transaction(root, 0);
1036 }
1037 if (IS_ERR(trans))
1038 return PTR_ERR(trans);
1039 }
1040 return btrfs_commit_transaction(trans);
1041}
1042
1043static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1044{
1045 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1046 *printed = true;
1047}
1048
1049static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1050{
1051 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1052 const char *compress_type;
1053 const char *subvol_name;
1054 bool printed = false;
1055
1056 if (btrfs_test_opt(info, DEGRADED))
1057 seq_puts(seq, ",degraded");
1058 if (btrfs_test_opt(info, NODATASUM))
1059 seq_puts(seq, ",nodatasum");
1060 if (btrfs_test_opt(info, NODATACOW))
1061 seq_puts(seq, ",nodatacow");
1062 if (btrfs_test_opt(info, NOBARRIER))
1063 seq_puts(seq, ",nobarrier");
1064 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1065 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1066 if (info->thread_pool_size != min_t(unsigned long,
1067 num_online_cpus() + 2, 8))
1068 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1069 if (btrfs_test_opt(info, COMPRESS)) {
1070 compress_type = btrfs_compress_type2str(info->compress_type);
1071 if (btrfs_test_opt(info, FORCE_COMPRESS))
1072 seq_printf(seq, ",compress-force=%s", compress_type);
1073 else
1074 seq_printf(seq, ",compress=%s", compress_type);
1075 if (info->compress_level)
1076 seq_printf(seq, ":%d", info->compress_level);
1077 }
1078 if (btrfs_test_opt(info, NOSSD))
1079 seq_puts(seq, ",nossd");
1080 if (btrfs_test_opt(info, SSD_SPREAD))
1081 seq_puts(seq, ",ssd_spread");
1082 else if (btrfs_test_opt(info, SSD))
1083 seq_puts(seq, ",ssd");
1084 if (btrfs_test_opt(info, NOTREELOG))
1085 seq_puts(seq, ",notreelog");
1086 if (btrfs_test_opt(info, NOLOGREPLAY))
1087 print_rescue_option(seq, "nologreplay", &printed);
1088 if (btrfs_test_opt(info, USEBACKUPROOT))
1089 print_rescue_option(seq, "usebackuproot", &printed);
1090 if (btrfs_test_opt(info, IGNOREBADROOTS))
1091 print_rescue_option(seq, "ignorebadroots", &printed);
1092 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1093 print_rescue_option(seq, "ignoredatacsums", &printed);
1094 if (btrfs_test_opt(info, IGNOREMETACSUMS))
1095 print_rescue_option(seq, "ignoremetacsums", &printed);
1096 if (btrfs_test_opt(info, IGNORESUPERFLAGS))
1097 print_rescue_option(seq, "ignoresuperflags", &printed);
1098 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1099 seq_puts(seq, ",flushoncommit");
1100 if (btrfs_test_opt(info, DISCARD_SYNC))
1101 seq_puts(seq, ",discard");
1102 if (btrfs_test_opt(info, DISCARD_ASYNC))
1103 seq_puts(seq, ",discard=async");
1104 if (!(info->sb->s_flags & SB_POSIXACL))
1105 seq_puts(seq, ",noacl");
1106 if (btrfs_free_space_cache_v1_active(info))
1107 seq_puts(seq, ",space_cache");
1108 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1109 seq_puts(seq, ",space_cache=v2");
1110 else
1111 seq_puts(seq, ",nospace_cache");
1112 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1113 seq_puts(seq, ",rescan_uuid_tree");
1114 if (btrfs_test_opt(info, CLEAR_CACHE))
1115 seq_puts(seq, ",clear_cache");
1116 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1117 seq_puts(seq, ",user_subvol_rm_allowed");
1118 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1119 seq_puts(seq, ",enospc_debug");
1120 if (btrfs_test_opt(info, AUTO_DEFRAG))
1121 seq_puts(seq, ",autodefrag");
1122 if (btrfs_test_opt(info, SKIP_BALANCE))
1123 seq_puts(seq, ",skip_balance");
1124 if (info->metadata_ratio)
1125 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1126 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1127 seq_puts(seq, ",fatal_errors=panic");
1128 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1129 seq_printf(seq, ",commit=%u", info->commit_interval);
1130#ifdef CONFIG_BTRFS_DEBUG
1131 if (btrfs_test_opt(info, FRAGMENT_DATA))
1132 seq_puts(seq, ",fragment=data");
1133 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1134 seq_puts(seq, ",fragment=metadata");
1135#endif
1136 if (btrfs_test_opt(info, REF_VERIFY))
1137 seq_puts(seq, ",ref_verify");
1138 seq_printf(seq, ",subvolid=%llu", btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1139 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1140 btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1141 if (!IS_ERR(subvol_name)) {
1142 seq_puts(seq, ",subvol=");
1143 seq_escape(seq, subvol_name, " \t\n\\");
1144 kfree(subvol_name);
1145 }
1146 return 0;
1147}
1148
1149/*
1150 * subvolumes are identified by ino 256
1151 */
1152static inline int is_subvolume_inode(struct inode *inode)
1153{
1154 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1155 return 1;
1156 return 0;
1157}
1158
1159static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1160 struct vfsmount *mnt)
1161{
1162 struct dentry *root;
1163 int ret;
1164
1165 if (!subvol_name) {
1166 if (!subvol_objectid) {
1167 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1168 &subvol_objectid);
1169 if (ret) {
1170 root = ERR_PTR(ret);
1171 goto out;
1172 }
1173 }
1174 subvol_name = btrfs_get_subvol_name_from_objectid(
1175 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1176 if (IS_ERR(subvol_name)) {
1177 root = ERR_CAST(subvol_name);
1178 subvol_name = NULL;
1179 goto out;
1180 }
1181
1182 }
1183
1184 root = mount_subtree(mnt, subvol_name);
1185 /* mount_subtree() drops our reference on the vfsmount. */
1186 mnt = NULL;
1187
1188 if (!IS_ERR(root)) {
1189 struct super_block *s = root->d_sb;
1190 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1191 struct inode *root_inode = d_inode(root);
1192 u64 root_objectid = btrfs_root_id(BTRFS_I(root_inode)->root);
1193
1194 ret = 0;
1195 if (!is_subvolume_inode(root_inode)) {
1196 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1197 subvol_name);
1198 ret = -EINVAL;
1199 }
1200 if (subvol_objectid && root_objectid != subvol_objectid) {
1201 /*
1202 * This will also catch a race condition where a
1203 * subvolume which was passed by ID is renamed and
1204 * another subvolume is renamed over the old location.
1205 */
1206 btrfs_err(fs_info,
1207 "subvol '%s' does not match subvolid %llu",
1208 subvol_name, subvol_objectid);
1209 ret = -EINVAL;
1210 }
1211 if (ret) {
1212 dput(root);
1213 root = ERR_PTR(ret);
1214 deactivate_locked_super(s);
1215 }
1216 }
1217
1218out:
1219 mntput(mnt);
1220 kfree(subvol_name);
1221 return root;
1222}
1223
1224static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1225 u32 new_pool_size, u32 old_pool_size)
1226{
1227 if (new_pool_size == old_pool_size)
1228 return;
1229
1230 fs_info->thread_pool_size = new_pool_size;
1231
1232 btrfs_info(fs_info, "resize thread pool %d -> %d",
1233 old_pool_size, new_pool_size);
1234
1235 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1236 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1237 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1238 workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1239 workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1240 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1241 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1242 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1243}
1244
1245static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1246 unsigned long long old_opts, int flags)
1247{
1248 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1249 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1250 (flags & SB_RDONLY))) {
1251 /* wait for any defraggers to finish */
1252 wait_event(fs_info->transaction_wait,
1253 (atomic_read(&fs_info->defrag_running) == 0));
1254 if (flags & SB_RDONLY)
1255 sync_filesystem(fs_info->sb);
1256 }
1257}
1258
1259static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1260 unsigned long long old_opts)
1261{
1262 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1263
1264 /*
1265 * We need to cleanup all defragable inodes if the autodefragment is
1266 * close or the filesystem is read only.
1267 */
1268 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1269 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1270 btrfs_cleanup_defrag_inodes(fs_info);
1271 }
1272
1273 /* If we toggled discard async */
1274 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1275 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1276 btrfs_discard_resume(fs_info);
1277 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1278 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1279 btrfs_discard_cleanup(fs_info);
1280
1281 /* If we toggled space cache */
1282 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1283 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1284}
1285
1286static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1287{
1288 int ret;
1289
1290 if (BTRFS_FS_ERROR(fs_info)) {
1291 btrfs_err(fs_info,
1292 "remounting read-write after error is not allowed");
1293 return -EINVAL;
1294 }
1295
1296 if (fs_info->fs_devices->rw_devices == 0)
1297 return -EACCES;
1298
1299 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1300 btrfs_warn(fs_info,
1301 "too many missing devices, writable remount is not allowed");
1302 return -EACCES;
1303 }
1304
1305 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1306 btrfs_warn(fs_info,
1307 "mount required to replay tree-log, cannot remount read-write");
1308 return -EINVAL;
1309 }
1310
1311 /*
1312 * NOTE: when remounting with a change that does writes, don't put it
1313 * anywhere above this point, as we are not sure to be safe to write
1314 * until we pass the above checks.
1315 */
1316 ret = btrfs_start_pre_rw_mount(fs_info);
1317 if (ret)
1318 return ret;
1319
1320 btrfs_clear_sb_rdonly(fs_info->sb);
1321
1322 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1323
1324 /*
1325 * If we've gone from readonly -> read-write, we need to get our
1326 * sync/async discard lists in the right state.
1327 */
1328 btrfs_discard_resume(fs_info);
1329
1330 return 0;
1331}
1332
1333static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1334{
1335 /*
1336 * This also happens on 'umount -rf' or on shutdown, when the
1337 * filesystem is busy.
1338 */
1339 cancel_work_sync(&fs_info->async_reclaim_work);
1340 cancel_work_sync(&fs_info->async_data_reclaim_work);
1341
1342 btrfs_discard_cleanup(fs_info);
1343
1344 /* Wait for the uuid_scan task to finish */
1345 down(&fs_info->uuid_tree_rescan_sem);
1346 /* Avoid complains from lockdep et al. */
1347 up(&fs_info->uuid_tree_rescan_sem);
1348
1349 btrfs_set_sb_rdonly(fs_info->sb);
1350
1351 /*
1352 * Setting SB_RDONLY will put the cleaner thread to sleep at the next
1353 * loop if it's already active. If it's already asleep, we'll leave
1354 * unused block groups on disk until we're mounted read-write again
1355 * unless we clean them up here.
1356 */
1357 btrfs_delete_unused_bgs(fs_info);
1358
1359 /*
1360 * The cleaner task could be already running before we set the flag
1361 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock). We must make
1362 * sure that after we finish the remount, i.e. after we call
1363 * btrfs_commit_super(), the cleaner can no longer start a transaction
1364 * - either because it was dropping a dead root, running delayed iputs
1365 * or deleting an unused block group (the cleaner picked a block
1366 * group from the list of unused block groups before we were able to
1367 * in the previous call to btrfs_delete_unused_bgs()).
1368 */
1369 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1370
1371 /*
1372 * We've set the superblock to RO mode, so we might have made the
1373 * cleaner task sleep without running all pending delayed iputs. Go
1374 * through all the delayed iputs here, so that if an unmount happens
1375 * without remounting RW we don't end up at finishing close_ctree()
1376 * with a non-empty list of delayed iputs.
1377 */
1378 btrfs_run_delayed_iputs(fs_info);
1379
1380 btrfs_dev_replace_suspend_for_unmount(fs_info);
1381 btrfs_scrub_cancel(fs_info);
1382 btrfs_pause_balance(fs_info);
1383
1384 /*
1385 * Pause the qgroup rescan worker if it is running. We don't want it to
1386 * be still running after we are in RO mode, as after that, by the time
1387 * we unmount, it might have left a transaction open, so we would leak
1388 * the transaction and/or crash.
1389 */
1390 btrfs_qgroup_wait_for_completion(fs_info, false);
1391
1392 return btrfs_commit_super(fs_info);
1393}
1394
1395static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1396{
1397 fs_info->max_inline = ctx->max_inline;
1398 fs_info->commit_interval = ctx->commit_interval;
1399 fs_info->metadata_ratio = ctx->metadata_ratio;
1400 fs_info->thread_pool_size = ctx->thread_pool_size;
1401 fs_info->mount_opt = ctx->mount_opt;
1402 fs_info->compress_type = ctx->compress_type;
1403 fs_info->compress_level = ctx->compress_level;
1404}
1405
1406static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1407{
1408 ctx->max_inline = fs_info->max_inline;
1409 ctx->commit_interval = fs_info->commit_interval;
1410 ctx->metadata_ratio = fs_info->metadata_ratio;
1411 ctx->thread_pool_size = fs_info->thread_pool_size;
1412 ctx->mount_opt = fs_info->mount_opt;
1413 ctx->compress_type = fs_info->compress_type;
1414 ctx->compress_level = fs_info->compress_level;
1415}
1416
1417#define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...) \
1418do { \
1419 if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1420 btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1421 btrfs_info(fs_info, fmt, ##args); \
1422} while (0)
1423
1424#define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...) \
1425do { \
1426 if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1427 !btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1428 btrfs_info(fs_info, fmt, ##args); \
1429} while (0)
1430
1431static void btrfs_emit_options(struct btrfs_fs_info *info,
1432 struct btrfs_fs_context *old)
1433{
1434 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1435 btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1436 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1437 btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1438 btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1439 btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1440 btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1441 btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1442 btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1443 btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1444 btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1445 btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1446 btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1447 btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1448 btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1449 btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1450 btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1451 btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1452 btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1453 btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1454 btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
1455 btrfs_info_if_set(info, old, IGNOREMETACSUMS, "ignoring meta csums");
1456 btrfs_info_if_set(info, old, IGNORESUPERFLAGS, "ignoring unknown super block flags");
1457
1458 btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1459 btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1460 btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1461 btrfs_info_if_unset(info, old, NOBARRIER, "turning off barriers");
1462 btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1463 btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1464 btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1465 btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1466 btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1467
1468 /* Did the compression settings change? */
1469 if (btrfs_test_opt(info, COMPRESS) &&
1470 (!old ||
1471 old->compress_type != info->compress_type ||
1472 old->compress_level != info->compress_level ||
1473 (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1474 btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1475 const char *compress_type = btrfs_compress_type2str(info->compress_type);
1476
1477 btrfs_info(info, "%s %s compression, level %d",
1478 btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1479 compress_type, info->compress_level);
1480 }
1481
1482 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1483 btrfs_info(info, "max_inline set to %llu", info->max_inline);
1484}
1485
1486static int btrfs_reconfigure(struct fs_context *fc)
1487{
1488 struct super_block *sb = fc->root->d_sb;
1489 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1490 struct btrfs_fs_context *ctx = fc->fs_private;
1491 struct btrfs_fs_context old_ctx;
1492 int ret = 0;
1493 bool mount_reconfigure = (fc->s_fs_info != NULL);
1494
1495 btrfs_info_to_ctx(fs_info, &old_ctx);
1496
1497 /*
1498 * This is our "bind mount" trick, we don't want to allow the user to do
1499 * anything other than mount a different ro/rw and a different subvol,
1500 * all of the mount options should be maintained.
1501 */
1502 if (mount_reconfigure)
1503 ctx->mount_opt = old_ctx.mount_opt;
1504
1505 sync_filesystem(sb);
1506 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1507
1508 if (!btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags))
1509 return -EINVAL;
1510
1511 ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
1512 if (ret < 0)
1513 return ret;
1514
1515 btrfs_ctx_to_info(fs_info, ctx);
1516 btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags);
1517 btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size,
1518 old_ctx.thread_pool_size);
1519
1520 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1521 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1522 (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1523 btrfs_warn(fs_info,
1524 "remount supports changing free space tree only from RO to RW");
1525 /* Make sure free space cache options match the state on disk. */
1526 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1527 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1528 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1529 }
1530 if (btrfs_free_space_cache_v1_active(fs_info)) {
1531 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1532 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1533 }
1534 }
1535
1536 ret = 0;
1537 if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1538 ret = btrfs_remount_ro(fs_info);
1539 else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1540 ret = btrfs_remount_rw(fs_info);
1541 if (ret)
1542 goto restore;
1543
1544 /*
1545 * If we set the mask during the parameter parsing VFS would reject the
1546 * remount. Here we can set the mask and the value will be updated
1547 * appropriately.
1548 */
1549 if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1550 fc->sb_flags_mask |= SB_POSIXACL;
1551
1552 btrfs_emit_options(fs_info, &old_ctx);
1553 wake_up_process(fs_info->transaction_kthread);
1554 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1555 btrfs_clear_oneshot_options(fs_info);
1556 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1557
1558 return 0;
1559restore:
1560 btrfs_ctx_to_info(fs_info, &old_ctx);
1561 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1562 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1563 return ret;
1564}
1565
1566/* Used to sort the devices by max_avail(descending sort) */
1567static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1568{
1569 const struct btrfs_device_info *dev_info1 = a;
1570 const struct btrfs_device_info *dev_info2 = b;
1571
1572 if (dev_info1->max_avail > dev_info2->max_avail)
1573 return -1;
1574 else if (dev_info1->max_avail < dev_info2->max_avail)
1575 return 1;
1576 return 0;
1577}
1578
1579/*
1580 * sort the devices by max_avail, in which max free extent size of each device
1581 * is stored.(Descending Sort)
1582 */
1583static inline void btrfs_descending_sort_devices(
1584 struct btrfs_device_info *devices,
1585 size_t nr_devices)
1586{
1587 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1588 btrfs_cmp_device_free_bytes, NULL);
1589}
1590
1591/*
1592 * The helper to calc the free space on the devices that can be used to store
1593 * file data.
1594 */
1595static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1596 u64 *free_bytes)
1597{
1598 struct btrfs_device_info *devices_info;
1599 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1600 struct btrfs_device *device;
1601 u64 type;
1602 u64 avail_space;
1603 u64 min_stripe_size;
1604 int num_stripes = 1;
1605 int i = 0, nr_devices;
1606 const struct btrfs_raid_attr *rattr;
1607
1608 /*
1609 * We aren't under the device list lock, so this is racy-ish, but good
1610 * enough for our purposes.
1611 */
1612 nr_devices = fs_info->fs_devices->open_devices;
1613 if (!nr_devices) {
1614 smp_mb();
1615 nr_devices = fs_info->fs_devices->open_devices;
1616 ASSERT(nr_devices);
1617 if (!nr_devices) {
1618 *free_bytes = 0;
1619 return 0;
1620 }
1621 }
1622
1623 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1624 GFP_KERNEL);
1625 if (!devices_info)
1626 return -ENOMEM;
1627
1628 /* calc min stripe number for data space allocation */
1629 type = btrfs_data_alloc_profile(fs_info);
1630 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1631
1632 if (type & BTRFS_BLOCK_GROUP_RAID0)
1633 num_stripes = nr_devices;
1634 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1635 num_stripes = rattr->ncopies;
1636 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1637 num_stripes = 4;
1638
1639 /* Adjust for more than 1 stripe per device */
1640 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1641
1642 rcu_read_lock();
1643 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1644 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1645 &device->dev_state) ||
1646 !device->bdev ||
1647 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1648 continue;
1649
1650 if (i >= nr_devices)
1651 break;
1652
1653 avail_space = device->total_bytes - device->bytes_used;
1654
1655 /* align with stripe_len */
1656 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1657
1658 /*
1659 * Ensure we have at least min_stripe_size on top of the
1660 * reserved space on the device.
1661 */
1662 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1663 continue;
1664
1665 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1666
1667 devices_info[i].dev = device;
1668 devices_info[i].max_avail = avail_space;
1669
1670 i++;
1671 }
1672 rcu_read_unlock();
1673
1674 nr_devices = i;
1675
1676 btrfs_descending_sort_devices(devices_info, nr_devices);
1677
1678 i = nr_devices - 1;
1679 avail_space = 0;
1680 while (nr_devices >= rattr->devs_min) {
1681 num_stripes = min(num_stripes, nr_devices);
1682
1683 if (devices_info[i].max_avail >= min_stripe_size) {
1684 int j;
1685 u64 alloc_size;
1686
1687 avail_space += devices_info[i].max_avail * num_stripes;
1688 alloc_size = devices_info[i].max_avail;
1689 for (j = i + 1 - num_stripes; j <= i; j++)
1690 devices_info[j].max_avail -= alloc_size;
1691 }
1692 i--;
1693 nr_devices--;
1694 }
1695
1696 kfree(devices_info);
1697 *free_bytes = avail_space;
1698 return 0;
1699}
1700
1701/*
1702 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1703 *
1704 * If there's a redundant raid level at DATA block groups, use the respective
1705 * multiplier to scale the sizes.
1706 *
1707 * Unused device space usage is based on simulating the chunk allocator
1708 * algorithm that respects the device sizes and order of allocations. This is
1709 * a close approximation of the actual use but there are other factors that may
1710 * change the result (like a new metadata chunk).
1711 *
1712 * If metadata is exhausted, f_bavail will be 0.
1713 */
1714static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1715{
1716 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1717 struct btrfs_super_block *disk_super = fs_info->super_copy;
1718 struct btrfs_space_info *found;
1719 u64 total_used = 0;
1720 u64 total_free_data = 0;
1721 u64 total_free_meta = 0;
1722 u32 bits = fs_info->sectorsize_bits;
1723 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1724 unsigned factor = 1;
1725 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1726 int ret;
1727 u64 thresh = 0;
1728 int mixed = 0;
1729
1730 list_for_each_entry(found, &fs_info->space_info, list) {
1731 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1732 int i;
1733
1734 total_free_data += found->disk_total - found->disk_used;
1735 total_free_data -=
1736 btrfs_account_ro_block_groups_free_space(found);
1737
1738 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1739 if (!list_empty(&found->block_groups[i]))
1740 factor = btrfs_bg_type_to_factor(
1741 btrfs_raid_array[i].bg_flag);
1742 }
1743 }
1744
1745 /*
1746 * Metadata in mixed block group profiles are accounted in data
1747 */
1748 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1749 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1750 mixed = 1;
1751 else
1752 total_free_meta += found->disk_total -
1753 found->disk_used;
1754 }
1755
1756 total_used += found->disk_used;
1757 }
1758
1759 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1760 buf->f_blocks >>= bits;
1761 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1762
1763 /* Account global block reserve as used, it's in logical size already */
1764 spin_lock(&block_rsv->lock);
1765 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
1766 if (buf->f_bfree >= block_rsv->size >> bits)
1767 buf->f_bfree -= block_rsv->size >> bits;
1768 else
1769 buf->f_bfree = 0;
1770 spin_unlock(&block_rsv->lock);
1771
1772 buf->f_bavail = div_u64(total_free_data, factor);
1773 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1774 if (ret)
1775 return ret;
1776 buf->f_bavail += div_u64(total_free_data, factor);
1777 buf->f_bavail = buf->f_bavail >> bits;
1778
1779 /*
1780 * We calculate the remaining metadata space minus global reserve. If
1781 * this is (supposedly) smaller than zero, there's no space. But this
1782 * does not hold in practice, the exhausted state happens where's still
1783 * some positive delta. So we apply some guesswork and compare the
1784 * delta to a 4M threshold. (Practically observed delta was ~2M.)
1785 *
1786 * We probably cannot calculate the exact threshold value because this
1787 * depends on the internal reservations requested by various
1788 * operations, so some operations that consume a few metadata will
1789 * succeed even if the Avail is zero. But this is better than the other
1790 * way around.
1791 */
1792 thresh = SZ_4M;
1793
1794 /*
1795 * We only want to claim there's no available space if we can no longer
1796 * allocate chunks for our metadata profile and our global reserve will
1797 * not fit in the free metadata space. If we aren't ->full then we
1798 * still can allocate chunks and thus are fine using the currently
1799 * calculated f_bavail.
1800 */
1801 if (!mixed && block_rsv->space_info->full &&
1802 (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1803 buf->f_bavail = 0;
1804
1805 buf->f_type = BTRFS_SUPER_MAGIC;
1806 buf->f_bsize = fs_info->sectorsize;
1807 buf->f_namelen = BTRFS_NAME_LEN;
1808
1809 /* We treat it as constant endianness (it doesn't matter _which_)
1810 because we want the fsid to come out the same whether mounted
1811 on a big-endian or little-endian host */
1812 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1813 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1814 /* Mask in the root object ID too, to disambiguate subvols */
1815 buf->f_fsid.val[0] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root) >> 32;
1816 buf->f_fsid.val[1] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root);
1817
1818 return 0;
1819}
1820
1821static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1822{
1823 struct btrfs_fs_info *p = fc->s_fs_info;
1824 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1825
1826 return fs_info->fs_devices == p->fs_devices;
1827}
1828
1829static int btrfs_get_tree_super(struct fs_context *fc)
1830{
1831 struct btrfs_fs_info *fs_info = fc->s_fs_info;
1832 struct btrfs_fs_context *ctx = fc->fs_private;
1833 struct btrfs_fs_devices *fs_devices = NULL;
1834 struct block_device *bdev;
1835 struct btrfs_device *device;
1836 struct super_block *sb;
1837 blk_mode_t mode = btrfs_open_mode(fc);
1838 int ret;
1839
1840 btrfs_ctx_to_info(fs_info, ctx);
1841 mutex_lock(&uuid_mutex);
1842
1843 /*
1844 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1845 * either a valid device or an error.
1846 */
1847 device = btrfs_scan_one_device(fc->source, mode, true);
1848 ASSERT(device != NULL);
1849 if (IS_ERR(device)) {
1850 mutex_unlock(&uuid_mutex);
1851 return PTR_ERR(device);
1852 }
1853
1854 fs_devices = device->fs_devices;
1855 fs_info->fs_devices = fs_devices;
1856
1857 ret = btrfs_open_devices(fs_devices, mode, &btrfs_fs_type);
1858 mutex_unlock(&uuid_mutex);
1859 if (ret)
1860 return ret;
1861
1862 if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1863 ret = -EACCES;
1864 goto error;
1865 }
1866
1867 bdev = fs_devices->latest_dev->bdev;
1868
1869 /*
1870 * From now on the error handling is not straightforward.
1871 *
1872 * If successful, this will transfer the fs_info into the super block,
1873 * and fc->s_fs_info will be NULL. However if there's an existing
1874 * super, we'll still have fc->s_fs_info populated. If we error
1875 * completely out it'll be cleaned up when we drop the fs_context,
1876 * otherwise it's tied to the lifetime of the super_block.
1877 */
1878 sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1879 if (IS_ERR(sb)) {
1880 ret = PTR_ERR(sb);
1881 goto error;
1882 }
1883
1884 set_device_specific_options(fs_info);
1885
1886 if (sb->s_root) {
1887 btrfs_close_devices(fs_devices);
1888 /*
1889 * At this stage we may have RO flag mismatch between
1890 * fc->sb_flags and sb->s_flags. Caller should detect such
1891 * mismatch and reconfigure with sb->s_umount rwsem held if
1892 * needed.
1893 */
1894 } else {
1895 snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
1896 shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id);
1897 btrfs_sb(sb)->bdev_holder = &btrfs_fs_type;
1898 ret = btrfs_fill_super(sb, fs_devices);
1899 if (ret) {
1900 deactivate_locked_super(sb);
1901 return ret;
1902 }
1903 }
1904
1905 btrfs_clear_oneshot_options(fs_info);
1906
1907 fc->root = dget(sb->s_root);
1908 return 0;
1909
1910error:
1911 btrfs_close_devices(fs_devices);
1912 return ret;
1913}
1914
1915/*
1916 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1917 * with different ro/rw options") the following works:
1918 *
1919 * (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1920 * (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1921 *
1922 * which looks nice and innocent but is actually pretty intricate and deserves
1923 * a long comment.
1924 *
1925 * On another filesystem a subvolume mount is close to something like:
1926 *
1927 * (iii) # create rw superblock + initial mount
1928 * mount -t xfs /dev/sdb /opt/
1929 *
1930 * # create ro bind mount
1931 * mount --bind -o ro /opt/foo /mnt/foo
1932 *
1933 * # unmount initial mount
1934 * umount /opt
1935 *
1936 * Of course, there's some special subvolume sauce and there's the fact that the
1937 * sb->s_root dentry is really swapped after mount_subtree(). But conceptually
1938 * it's very close and will help us understand the issue.
1939 *
1940 * The old mount API didn't cleanly distinguish between a mount being made ro
1941 * and a superblock being made ro. The only way to change the ro state of
1942 * either object was by passing ms_rdonly. If a new mount was created via
1943 * mount(2) such as:
1944 *
1945 * mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1946 *
1947 * the MS_RDONLY flag being specified had two effects:
1948 *
1949 * (1) MNT_READONLY was raised -> the resulting mount got
1950 * @mnt->mnt_flags |= MNT_READONLY raised.
1951 *
1952 * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
1953 * made the superblock ro. Note, how SB_RDONLY has the same value as
1954 * ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
1955 *
1956 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
1957 * subtree mounted ro.
1958 *
1959 * But consider the effect on the old mount API on btrfs subvolume mounting
1960 * which combines the distinct step in (iii) into a single step.
1961 *
1962 * By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
1963 * is issued the superblock is ro and thus even if the mount created for (ii) is
1964 * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
1965 * to rw for (ii) which it did using an internal remount call.
1966 *
1967 * IOW, subvolume mounting was inherently complicated due to the ambiguity of
1968 * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
1969 * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
1970 * passed by mount(8) to mount(2).
1971 *
1972 * Enter the new mount API. The new mount API disambiguates making a mount ro
1973 * and making a superblock ro.
1974 *
1975 * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
1976 * fsmount() or mount_setattr() this is a pure VFS level change for a
1977 * specific mount or mount tree that is never seen by the filesystem itself.
1978 *
1979 * (4) To turn a superblock ro the "ro" flag must be used with
1980 * fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
1981 * in fc->sb_flags.
1982 *
1983 * But, currently the util-linux mount command already utilizes the new mount
1984 * API and is still setting fsconfig(FSCONFIG_SET_FLAG, "ro") no matter if it's
1985 * btrfs or not, setting the whole super block RO. To make per-subvolume mounting
1986 * work with different options work we need to keep backward compatibility.
1987 */
1988static int btrfs_reconfigure_for_mount(struct fs_context *fc, struct vfsmount *mnt)
1989{
1990 int ret = 0;
1991
1992 if (fc->sb_flags & SB_RDONLY)
1993 return ret;
1994
1995 down_write(&mnt->mnt_sb->s_umount);
1996 if (!(fc->sb_flags & SB_RDONLY) && (mnt->mnt_sb->s_flags & SB_RDONLY))
1997 ret = btrfs_reconfigure(fc);
1998 up_write(&mnt->mnt_sb->s_umount);
1999 return ret;
2000}
2001
2002static int btrfs_get_tree_subvol(struct fs_context *fc)
2003{
2004 struct btrfs_fs_info *fs_info = NULL;
2005 struct btrfs_fs_context *ctx = fc->fs_private;
2006 struct fs_context *dup_fc;
2007 struct dentry *dentry;
2008 struct vfsmount *mnt;
2009 int ret = 0;
2010
2011 /*
2012 * Setup a dummy root and fs_info for test/set super. This is because
2013 * we don't actually fill this stuff out until open_ctree, but we need
2014 * then open_ctree will properly initialize the file system specific
2015 * settings later. btrfs_init_fs_info initializes the static elements
2016 * of the fs_info (locks and such) to make cleanup easier if we find a
2017 * superblock with our given fs_devices later on at sget() time.
2018 */
2019 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
2020 if (!fs_info)
2021 return -ENOMEM;
2022
2023 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2024 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2025 if (!fs_info->super_copy || !fs_info->super_for_commit) {
2026 btrfs_free_fs_info(fs_info);
2027 return -ENOMEM;
2028 }
2029 btrfs_init_fs_info(fs_info);
2030
2031 dup_fc = vfs_dup_fs_context(fc);
2032 if (IS_ERR(dup_fc)) {
2033 btrfs_free_fs_info(fs_info);
2034 return PTR_ERR(dup_fc);
2035 }
2036
2037 /*
2038 * When we do the sget_fc this gets transferred to the sb, so we only
2039 * need to set it on the dup_fc as this is what creates the super block.
2040 */
2041 dup_fc->s_fs_info = fs_info;
2042
2043 /*
2044 * We'll do the security settings in our btrfs_get_tree_super() mount
2045 * loop, they were duplicated into dup_fc, we can drop the originals
2046 * here.
2047 */
2048 security_free_mnt_opts(&fc->security);
2049 fc->security = NULL;
2050
2051 mnt = fc_mount(dup_fc);
2052 if (IS_ERR(mnt)) {
2053 put_fs_context(dup_fc);
2054 return PTR_ERR(mnt);
2055 }
2056 ret = btrfs_reconfigure_for_mount(dup_fc, mnt);
2057 put_fs_context(dup_fc);
2058 if (ret) {
2059 mntput(mnt);
2060 return ret;
2061 }
2062
2063 /*
2064 * This free's ->subvol_name, because if it isn't set we have to
2065 * allocate a buffer to hold the subvol_name, so we just drop our
2066 * reference to it here.
2067 */
2068 dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2069 ctx->subvol_name = NULL;
2070 if (IS_ERR(dentry))
2071 return PTR_ERR(dentry);
2072
2073 fc->root = dentry;
2074 return 0;
2075}
2076
2077static int btrfs_get_tree(struct fs_context *fc)
2078{
2079 /*
2080 * Since we use mount_subtree to mount the default/specified subvol, we
2081 * have to do mounts in two steps.
2082 *
2083 * First pass through we call btrfs_get_tree_subvol(), this is just a
2084 * wrapper around fc_mount() to call back into here again, and this time
2085 * we'll call btrfs_get_tree_super(). This will do the open_ctree() and
2086 * everything to open the devices and file system. Then we return back
2087 * with a fully constructed vfsmount in btrfs_get_tree_subvol(), and
2088 * from there we can do our mount_subvol() call, which will lookup
2089 * whichever subvol we're mounting and setup this fc with the
2090 * appropriate dentry for the subvol.
2091 */
2092 if (fc->s_fs_info)
2093 return btrfs_get_tree_super(fc);
2094 return btrfs_get_tree_subvol(fc);
2095}
2096
2097static void btrfs_kill_super(struct super_block *sb)
2098{
2099 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2100 kill_anon_super(sb);
2101 btrfs_free_fs_info(fs_info);
2102}
2103
2104static void btrfs_free_fs_context(struct fs_context *fc)
2105{
2106 struct btrfs_fs_context *ctx = fc->fs_private;
2107 struct btrfs_fs_info *fs_info = fc->s_fs_info;
2108
2109 if (fs_info)
2110 btrfs_free_fs_info(fs_info);
2111
2112 if (ctx && refcount_dec_and_test(&ctx->refs)) {
2113 kfree(ctx->subvol_name);
2114 kfree(ctx);
2115 }
2116}
2117
2118static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2119{
2120 struct btrfs_fs_context *ctx = src_fc->fs_private;
2121
2122 /*
2123 * Give a ref to our ctx to this dup, as we want to keep it around for
2124 * our original fc so we can have the subvolume name or objectid.
2125 *
2126 * We unset ->source in the original fc because the dup needs it for
2127 * mounting, and then once we free the dup it'll free ->source, so we
2128 * need to make sure we're only pointing to it in one fc.
2129 */
2130 refcount_inc(&ctx->refs);
2131 fc->fs_private = ctx;
2132 fc->source = src_fc->source;
2133 src_fc->source = NULL;
2134 return 0;
2135}
2136
2137static const struct fs_context_operations btrfs_fs_context_ops = {
2138 .parse_param = btrfs_parse_param,
2139 .reconfigure = btrfs_reconfigure,
2140 .get_tree = btrfs_get_tree,
2141 .dup = btrfs_dup_fs_context,
2142 .free = btrfs_free_fs_context,
2143};
2144
2145static int btrfs_init_fs_context(struct fs_context *fc)
2146{
2147 struct btrfs_fs_context *ctx;
2148
2149 ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL);
2150 if (!ctx)
2151 return -ENOMEM;
2152
2153 refcount_set(&ctx->refs, 1);
2154 fc->fs_private = ctx;
2155 fc->ops = &btrfs_fs_context_ops;
2156
2157 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2158 btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
2159 } else {
2160 ctx->thread_pool_size =
2161 min_t(unsigned long, num_online_cpus() + 2, 8);
2162 ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2163 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2164 }
2165
2166#ifdef CONFIG_BTRFS_FS_POSIX_ACL
2167 fc->sb_flags |= SB_POSIXACL;
2168#endif
2169 fc->sb_flags |= SB_I_VERSION;
2170
2171 return 0;
2172}
2173
2174static struct file_system_type btrfs_fs_type = {
2175 .owner = THIS_MODULE,
2176 .name = "btrfs",
2177 .init_fs_context = btrfs_init_fs_context,
2178 .parameters = btrfs_fs_parameters,
2179 .kill_sb = btrfs_kill_super,
2180 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA |
2181 FS_ALLOW_IDMAP | FS_MGTIME,
2182 };
2183
2184MODULE_ALIAS_FS("btrfs");
2185
2186static int btrfs_control_open(struct inode *inode, struct file *file)
2187{
2188 /*
2189 * The control file's private_data is used to hold the
2190 * transaction when it is started and is used to keep
2191 * track of whether a transaction is already in progress.
2192 */
2193 file->private_data = NULL;
2194 return 0;
2195}
2196
2197/*
2198 * Used by /dev/btrfs-control for devices ioctls.
2199 */
2200static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2201 unsigned long arg)
2202{
2203 struct btrfs_ioctl_vol_args *vol;
2204 struct btrfs_device *device = NULL;
2205 dev_t devt = 0;
2206 int ret = -ENOTTY;
2207
2208 if (!capable(CAP_SYS_ADMIN))
2209 return -EPERM;
2210
2211 vol = memdup_user((void __user *)arg, sizeof(*vol));
2212 if (IS_ERR(vol))
2213 return PTR_ERR(vol);
2214 ret = btrfs_check_ioctl_vol_args_path(vol);
2215 if (ret < 0)
2216 goto out;
2217
2218 switch (cmd) {
2219 case BTRFS_IOC_SCAN_DEV:
2220 mutex_lock(&uuid_mutex);
2221 /*
2222 * Scanning outside of mount can return NULL which would turn
2223 * into 0 error code.
2224 */
2225 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2226 ret = PTR_ERR_OR_ZERO(device);
2227 mutex_unlock(&uuid_mutex);
2228 break;
2229 case BTRFS_IOC_FORGET_DEV:
2230 if (vol->name[0] != 0) {
2231 ret = lookup_bdev(vol->name, &devt);
2232 if (ret)
2233 break;
2234 }
2235 ret = btrfs_forget_devices(devt);
2236 break;
2237 case BTRFS_IOC_DEVICES_READY:
2238 mutex_lock(&uuid_mutex);
2239 /*
2240 * Scanning outside of mount can return NULL which would turn
2241 * into 0 error code.
2242 */
2243 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2244 if (IS_ERR_OR_NULL(device)) {
2245 mutex_unlock(&uuid_mutex);
2246 if (IS_ERR(device))
2247 ret = PTR_ERR(device);
2248 else
2249 ret = 0;
2250 break;
2251 }
2252 ret = !(device->fs_devices->num_devices ==
2253 device->fs_devices->total_devices);
2254 mutex_unlock(&uuid_mutex);
2255 break;
2256 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2257 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2258 break;
2259 }
2260
2261out:
2262 kfree(vol);
2263 return ret;
2264}
2265
2266static int btrfs_freeze(struct super_block *sb)
2267{
2268 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2269
2270 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2271 /*
2272 * We don't need a barrier here, we'll wait for any transaction that
2273 * could be in progress on other threads (and do delayed iputs that
2274 * we want to avoid on a frozen filesystem), or do the commit
2275 * ourselves.
2276 */
2277 return btrfs_commit_current_transaction(fs_info->tree_root);
2278}
2279
2280static int check_dev_super(struct btrfs_device *dev)
2281{
2282 struct btrfs_fs_info *fs_info = dev->fs_info;
2283 struct btrfs_super_block *sb;
2284 u64 last_trans;
2285 u16 csum_type;
2286 int ret = 0;
2287
2288 /* This should be called with fs still frozen. */
2289 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2290
2291 /* Missing dev, no need to check. */
2292 if (!dev->bdev)
2293 return 0;
2294
2295 /* Only need to check the primary super block. */
2296 sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2297 if (IS_ERR(sb))
2298 return PTR_ERR(sb);
2299
2300 /* Verify the checksum. */
2301 csum_type = btrfs_super_csum_type(sb);
2302 if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2303 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2304 csum_type, btrfs_super_csum_type(fs_info->super_copy));
2305 ret = -EUCLEAN;
2306 goto out;
2307 }
2308
2309 if (btrfs_check_super_csum(fs_info, sb)) {
2310 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2311 ret = -EUCLEAN;
2312 goto out;
2313 }
2314
2315 /* Btrfs_validate_super() includes fsid check against super->fsid. */
2316 ret = btrfs_validate_super(fs_info, sb, 0);
2317 if (ret < 0)
2318 goto out;
2319
2320 last_trans = btrfs_get_last_trans_committed(fs_info);
2321 if (btrfs_super_generation(sb) != last_trans) {
2322 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2323 btrfs_super_generation(sb), last_trans);
2324 ret = -EUCLEAN;
2325 goto out;
2326 }
2327out:
2328 btrfs_release_disk_super(sb);
2329 return ret;
2330}
2331
2332static int btrfs_unfreeze(struct super_block *sb)
2333{
2334 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2335 struct btrfs_device *device;
2336 int ret = 0;
2337
2338 /*
2339 * Make sure the fs is not changed by accident (like hibernation then
2340 * modified by other OS).
2341 * If we found anything wrong, we mark the fs error immediately.
2342 *
2343 * And since the fs is frozen, no one can modify the fs yet, thus
2344 * we don't need to hold device_list_mutex.
2345 */
2346 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2347 ret = check_dev_super(device);
2348 if (ret < 0) {
2349 btrfs_handle_fs_error(fs_info, ret,
2350 "super block on devid %llu got modified unexpectedly",
2351 device->devid);
2352 break;
2353 }
2354 }
2355 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2356
2357 /*
2358 * We still return 0, to allow VFS layer to unfreeze the fs even the
2359 * above checks failed. Since the fs is either fine or read-only, we're
2360 * safe to continue, without causing further damage.
2361 */
2362 return 0;
2363}
2364
2365static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2366{
2367 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2368
2369 /*
2370 * There should be always a valid pointer in latest_dev, it may be stale
2371 * for a short moment in case it's being deleted but still valid until
2372 * the end of RCU grace period.
2373 */
2374 rcu_read_lock();
2375 seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2376 rcu_read_unlock();
2377
2378 return 0;
2379}
2380
2381static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_control *sc)
2382{
2383 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2384 const s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps);
2385
2386 trace_btrfs_extent_map_shrinker_count(fs_info, nr);
2387
2388 return nr;
2389}
2390
2391static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_control *sc)
2392{
2393 const long nr_to_scan = min_t(unsigned long, LONG_MAX, sc->nr_to_scan);
2394 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2395
2396 btrfs_free_extent_maps(fs_info, nr_to_scan);
2397
2398 /* The extent map shrinker runs asynchronously, so always return 0. */
2399 return 0;
2400}
2401
2402static const struct super_operations btrfs_super_ops = {
2403 .drop_inode = btrfs_drop_inode,
2404 .evict_inode = btrfs_evict_inode,
2405 .put_super = btrfs_put_super,
2406 .sync_fs = btrfs_sync_fs,
2407 .show_options = btrfs_show_options,
2408 .show_devname = btrfs_show_devname,
2409 .alloc_inode = btrfs_alloc_inode,
2410 .destroy_inode = btrfs_destroy_inode,
2411 .free_inode = btrfs_free_inode,
2412 .statfs = btrfs_statfs,
2413 .freeze_fs = btrfs_freeze,
2414 .unfreeze_fs = btrfs_unfreeze,
2415 .nr_cached_objects = btrfs_nr_cached_objects,
2416 .free_cached_objects = btrfs_free_cached_objects,
2417};
2418
2419static const struct file_operations btrfs_ctl_fops = {
2420 .open = btrfs_control_open,
2421 .unlocked_ioctl = btrfs_control_ioctl,
2422 .compat_ioctl = compat_ptr_ioctl,
2423 .owner = THIS_MODULE,
2424 .llseek = noop_llseek,
2425};
2426
2427static struct miscdevice btrfs_misc = {
2428 .minor = BTRFS_MINOR,
2429 .name = "btrfs-control",
2430 .fops = &btrfs_ctl_fops
2431};
2432
2433MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2434MODULE_ALIAS("devname:btrfs-control");
2435
2436static int __init btrfs_interface_init(void)
2437{
2438 return misc_register(&btrfs_misc);
2439}
2440
2441static __cold void btrfs_interface_exit(void)
2442{
2443 misc_deregister(&btrfs_misc);
2444}
2445
2446static int __init btrfs_print_mod_info(void)
2447{
2448 static const char options[] = ""
2449#ifdef CONFIG_BTRFS_DEBUG
2450 ", debug=on"
2451#endif
2452#ifdef CONFIG_BTRFS_ASSERT
2453 ", assert=on"
2454#endif
2455#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2456 ", ref-verify=on"
2457#endif
2458#ifdef CONFIG_BLK_DEV_ZONED
2459 ", zoned=yes"
2460#else
2461 ", zoned=no"
2462#endif
2463#ifdef CONFIG_FS_VERITY
2464 ", fsverity=yes"
2465#else
2466 ", fsverity=no"
2467#endif
2468 ;
2469 pr_info("Btrfs loaded%s\n", options);
2470 return 0;
2471}
2472
2473static int register_btrfs(void)
2474{
2475 return register_filesystem(&btrfs_fs_type);
2476}
2477
2478static void unregister_btrfs(void)
2479{
2480 unregister_filesystem(&btrfs_fs_type);
2481}
2482
2483/* Helper structure for long init/exit functions. */
2484struct init_sequence {
2485 int (*init_func)(void);
2486 /* Can be NULL if the init_func doesn't need cleanup. */
2487 void (*exit_func)(void);
2488};
2489
2490static const struct init_sequence mod_init_seq[] = {
2491 {
2492 .init_func = btrfs_props_init,
2493 .exit_func = NULL,
2494 }, {
2495 .init_func = btrfs_init_sysfs,
2496 .exit_func = btrfs_exit_sysfs,
2497 }, {
2498 .init_func = btrfs_init_compress,
2499 .exit_func = btrfs_exit_compress,
2500 }, {
2501 .init_func = btrfs_init_cachep,
2502 .exit_func = btrfs_destroy_cachep,
2503 }, {
2504 .init_func = btrfs_init_dio,
2505 .exit_func = btrfs_destroy_dio,
2506 }, {
2507 .init_func = btrfs_transaction_init,
2508 .exit_func = btrfs_transaction_exit,
2509 }, {
2510 .init_func = btrfs_ctree_init,
2511 .exit_func = btrfs_ctree_exit,
2512 }, {
2513 .init_func = btrfs_free_space_init,
2514 .exit_func = btrfs_free_space_exit,
2515 }, {
2516 .init_func = extent_state_init_cachep,
2517 .exit_func = extent_state_free_cachep,
2518 }, {
2519 .init_func = extent_buffer_init_cachep,
2520 .exit_func = extent_buffer_free_cachep,
2521 }, {
2522 .init_func = btrfs_bioset_init,
2523 .exit_func = btrfs_bioset_exit,
2524 }, {
2525 .init_func = extent_map_init,
2526 .exit_func = extent_map_exit,
2527 }, {
2528 .init_func = ordered_data_init,
2529 .exit_func = ordered_data_exit,
2530 }, {
2531 .init_func = btrfs_delayed_inode_init,
2532 .exit_func = btrfs_delayed_inode_exit,
2533 }, {
2534 .init_func = btrfs_auto_defrag_init,
2535 .exit_func = btrfs_auto_defrag_exit,
2536 }, {
2537 .init_func = btrfs_delayed_ref_init,
2538 .exit_func = btrfs_delayed_ref_exit,
2539 }, {
2540 .init_func = btrfs_prelim_ref_init,
2541 .exit_func = btrfs_prelim_ref_exit,
2542 }, {
2543 .init_func = btrfs_interface_init,
2544 .exit_func = btrfs_interface_exit,
2545 }, {
2546 .init_func = btrfs_print_mod_info,
2547 .exit_func = NULL,
2548 }, {
2549 .init_func = btrfs_run_sanity_tests,
2550 .exit_func = NULL,
2551 }, {
2552 .init_func = register_btrfs,
2553 .exit_func = unregister_btrfs,
2554 }
2555};
2556
2557static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2558
2559static __always_inline void btrfs_exit_btrfs_fs(void)
2560{
2561 int i;
2562
2563 for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2564 if (!mod_init_result[i])
2565 continue;
2566 if (mod_init_seq[i].exit_func)
2567 mod_init_seq[i].exit_func();
2568 mod_init_result[i] = false;
2569 }
2570}
2571
2572static void __exit exit_btrfs_fs(void)
2573{
2574 btrfs_exit_btrfs_fs();
2575 btrfs_cleanup_fs_uuids();
2576}
2577
2578static int __init init_btrfs_fs(void)
2579{
2580 int ret;
2581 int i;
2582
2583 for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2584 ASSERT(!mod_init_result[i]);
2585 ret = mod_init_seq[i].init_func();
2586 if (ret < 0) {
2587 btrfs_exit_btrfs_fs();
2588 return ret;
2589 }
2590 mod_init_result[i] = true;
2591 }
2592 return 0;
2593}
2594
2595late_initcall(init_btrfs_fs);
2596module_exit(exit_btrfs_fs)
2597
2598MODULE_DESCRIPTION("B-Tree File System (BTRFS)");
2599MODULE_LICENSE("GPL");
2600MODULE_SOFTDEP("pre: crc32c");
2601MODULE_SOFTDEP("pre: xxhash64");
2602MODULE_SOFTDEP("pre: sha256");
2603MODULE_SOFTDEP("pre: blake2b-256");
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/cleancache.h>
27#include <linux/ratelimit.h>
28#include <linux/crc32c.h>
29#include <linux/btrfs.h>
30#include "delayed-inode.h"
31#include "ctree.h"
32#include "disk-io.h"
33#include "transaction.h"
34#include "btrfs_inode.h"
35#include "print-tree.h"
36#include "props.h"
37#include "xattr.h"
38#include "volumes.h"
39#include "export.h"
40#include "compression.h"
41#include "rcu-string.h"
42#include "dev-replace.h"
43#include "free-space-cache.h"
44#include "backref.h"
45#include "space-info.h"
46#include "sysfs.h"
47#include "zoned.h"
48#include "tests/btrfs-tests.h"
49#include "block-group.h"
50#include "discard.h"
51#include "qgroup.h"
52#define CREATE_TRACE_POINTS
53#include <trace/events/btrfs.h>
54
55static const struct super_operations btrfs_super_ops;
56
57/*
58 * Types for mounting the default subvolume and a subvolume explicitly
59 * requested by subvol=/path. That way the callchain is straightforward and we
60 * don't have to play tricks with the mount options and recursive calls to
61 * btrfs_mount.
62 *
63 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
64 */
65static struct file_system_type btrfs_fs_type;
66static struct file_system_type btrfs_root_fs_type;
67
68static int btrfs_remount(struct super_block *sb, int *flags, char *data);
69
70/*
71 * Generally the error codes correspond to their respective errors, but there
72 * are a few special cases.
73 *
74 * EUCLEAN: Any sort of corruption that we encounter. The tree-checker for
75 * instance will return EUCLEAN if any of the blocks are corrupted in
76 * a way that is problematic. We want to reserve EUCLEAN for these
77 * sort of corruptions.
78 *
79 * EROFS: If we check BTRFS_FS_STATE_ERROR and fail out with a return error, we
80 * need to use EROFS for this case. We will have no idea of the
81 * original failure, that will have been reported at the time we tripped
82 * over the error. Each subsequent error that doesn't have any context
83 * of the original error should use EROFS when handling BTRFS_FS_STATE_ERROR.
84 */
85const char * __attribute_const__ btrfs_decode_error(int errno)
86{
87 char *errstr = "unknown";
88
89 switch (errno) {
90 case -ENOENT: /* -2 */
91 errstr = "No such entry";
92 break;
93 case -EIO: /* -5 */
94 errstr = "IO failure";
95 break;
96 case -ENOMEM: /* -12*/
97 errstr = "Out of memory";
98 break;
99 case -EEXIST: /* -17 */
100 errstr = "Object already exists";
101 break;
102 case -ENOSPC: /* -28 */
103 errstr = "No space left";
104 break;
105 case -EROFS: /* -30 */
106 errstr = "Readonly filesystem";
107 break;
108 case -EOPNOTSUPP: /* -95 */
109 errstr = "Operation not supported";
110 break;
111 case -EUCLEAN: /* -117 */
112 errstr = "Filesystem corrupted";
113 break;
114 case -EDQUOT: /* -122 */
115 errstr = "Quota exceeded";
116 break;
117 }
118
119 return errstr;
120}
121
122/*
123 * __btrfs_handle_fs_error decodes expected errors from the caller and
124 * invokes the appropriate error response.
125 */
126__cold
127void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
128 unsigned int line, int errno, const char *fmt, ...)
129{
130 struct super_block *sb = fs_info->sb;
131#ifdef CONFIG_PRINTK
132 const char *errstr;
133#endif
134
135 /*
136 * Special case: if the error is EROFS, and we're already
137 * under SB_RDONLY, then it is safe here.
138 */
139 if (errno == -EROFS && sb_rdonly(sb))
140 return;
141
142#ifdef CONFIG_PRINTK
143 errstr = btrfs_decode_error(errno);
144 if (fmt) {
145 struct va_format vaf;
146 va_list args;
147
148 va_start(args, fmt);
149 vaf.fmt = fmt;
150 vaf.va = &args;
151
152 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
153 sb->s_id, function, line, errno, errstr, &vaf);
154 va_end(args);
155 } else {
156 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
157 sb->s_id, function, line, errno, errstr);
158 }
159#endif
160
161 /*
162 * Today we only save the error info to memory. Long term we'll
163 * also send it down to the disk
164 */
165 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
166
167 /* Don't go through full error handling during mount */
168 if (!(sb->s_flags & SB_BORN))
169 return;
170
171 if (sb_rdonly(sb))
172 return;
173
174 btrfs_discard_stop(fs_info);
175
176 /* btrfs handle error by forcing the filesystem readonly */
177 btrfs_set_sb_rdonly(sb);
178 btrfs_info(fs_info, "forced readonly");
179 /*
180 * Note that a running device replace operation is not canceled here
181 * although there is no way to update the progress. It would add the
182 * risk of a deadlock, therefore the canceling is omitted. The only
183 * penalty is that some I/O remains active until the procedure
184 * completes. The next time when the filesystem is mounted writable
185 * again, the device replace operation continues.
186 */
187}
188
189#ifdef CONFIG_PRINTK
190static const char * const logtypes[] = {
191 "emergency",
192 "alert",
193 "critical",
194 "error",
195 "warning",
196 "notice",
197 "info",
198 "debug",
199};
200
201
202/*
203 * Use one ratelimit state per log level so that a flood of less important
204 * messages doesn't cause more important ones to be dropped.
205 */
206static struct ratelimit_state printk_limits[] = {
207 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
208 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
209 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
210 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
211 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
212 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
213 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
214 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
215};
216
217void __cold btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
218{
219 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
220 struct va_format vaf;
221 va_list args;
222 int kern_level;
223 const char *type = logtypes[4];
224 struct ratelimit_state *ratelimit = &printk_limits[4];
225
226 va_start(args, fmt);
227
228 while ((kern_level = printk_get_level(fmt)) != 0) {
229 size_t size = printk_skip_level(fmt) - fmt;
230
231 if (kern_level >= '0' && kern_level <= '7') {
232 memcpy(lvl, fmt, size);
233 lvl[size] = '\0';
234 type = logtypes[kern_level - '0'];
235 ratelimit = &printk_limits[kern_level - '0'];
236 }
237 fmt += size;
238 }
239
240 vaf.fmt = fmt;
241 vaf.va = &args;
242
243 if (__ratelimit(ratelimit)) {
244 if (fs_info)
245 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
246 fs_info->sb->s_id, &vaf);
247 else
248 printk("%sBTRFS %s: %pV\n", lvl, type, &vaf);
249 }
250
251 va_end(args);
252}
253#endif
254
255#if BITS_PER_LONG == 32
256void __cold btrfs_warn_32bit_limit(struct btrfs_fs_info *fs_info)
257{
258 if (!test_and_set_bit(BTRFS_FS_32BIT_WARN, &fs_info->flags)) {
259 btrfs_warn(fs_info, "reaching 32bit limit for logical addresses");
260 btrfs_warn(fs_info,
261"due to page cache limit on 32bit systems, btrfs can't access metadata at or beyond %lluT",
262 BTRFS_32BIT_MAX_FILE_SIZE >> 40);
263 btrfs_warn(fs_info,
264 "please consider upgrading to 64bit kernel/hardware");
265 }
266}
267
268void __cold btrfs_err_32bit_limit(struct btrfs_fs_info *fs_info)
269{
270 if (!test_and_set_bit(BTRFS_FS_32BIT_ERROR, &fs_info->flags)) {
271 btrfs_err(fs_info, "reached 32bit limit for logical addresses");
272 btrfs_err(fs_info,
273"due to page cache limit on 32bit systems, metadata beyond %lluT can't be accessed",
274 BTRFS_32BIT_MAX_FILE_SIZE >> 40);
275 btrfs_err(fs_info,
276 "please consider upgrading to 64bit kernel/hardware");
277 }
278}
279#endif
280
281/*
282 * We only mark the transaction aborted and then set the file system read-only.
283 * This will prevent new transactions from starting or trying to join this
284 * one.
285 *
286 * This means that error recovery at the call site is limited to freeing
287 * any local memory allocations and passing the error code up without
288 * further cleanup. The transaction should complete as it normally would
289 * in the call path but will return -EIO.
290 *
291 * We'll complete the cleanup in btrfs_end_transaction and
292 * btrfs_commit_transaction.
293 */
294__cold
295void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
296 const char *function,
297 unsigned int line, int errno)
298{
299 struct btrfs_fs_info *fs_info = trans->fs_info;
300
301 WRITE_ONCE(trans->aborted, errno);
302 WRITE_ONCE(trans->transaction->aborted, errno);
303 /* Wake up anybody who may be waiting on this transaction */
304 wake_up(&fs_info->transaction_wait);
305 wake_up(&fs_info->transaction_blocked_wait);
306 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
307}
308/*
309 * __btrfs_panic decodes unexpected, fatal errors from the caller,
310 * issues an alert, and either panics or BUGs, depending on mount options.
311 */
312__cold
313void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
314 unsigned int line, int errno, const char *fmt, ...)
315{
316 char *s_id = "<unknown>";
317 const char *errstr;
318 struct va_format vaf = { .fmt = fmt };
319 va_list args;
320
321 if (fs_info)
322 s_id = fs_info->sb->s_id;
323
324 va_start(args, fmt);
325 vaf.va = &args;
326
327 errstr = btrfs_decode_error(errno);
328 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
329 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
330 s_id, function, line, &vaf, errno, errstr);
331
332 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
333 function, line, &vaf, errno, errstr);
334 va_end(args);
335 /* Caller calls BUG() */
336}
337
338static void btrfs_put_super(struct super_block *sb)
339{
340 close_ctree(btrfs_sb(sb));
341}
342
343enum {
344 Opt_acl, Opt_noacl,
345 Opt_clear_cache,
346 Opt_commit_interval,
347 Opt_compress,
348 Opt_compress_force,
349 Opt_compress_force_type,
350 Opt_compress_type,
351 Opt_degraded,
352 Opt_device,
353 Opt_fatal_errors,
354 Opt_flushoncommit, Opt_noflushoncommit,
355 Opt_max_inline,
356 Opt_barrier, Opt_nobarrier,
357 Opt_datacow, Opt_nodatacow,
358 Opt_datasum, Opt_nodatasum,
359 Opt_defrag, Opt_nodefrag,
360 Opt_discard, Opt_nodiscard,
361 Opt_discard_mode,
362 Opt_norecovery,
363 Opt_ratio,
364 Opt_rescan_uuid_tree,
365 Opt_skip_balance,
366 Opt_space_cache, Opt_no_space_cache,
367 Opt_space_cache_version,
368 Opt_ssd, Opt_nossd,
369 Opt_ssd_spread, Opt_nossd_spread,
370 Opt_subvol,
371 Opt_subvol_empty,
372 Opt_subvolid,
373 Opt_thread_pool,
374 Opt_treelog, Opt_notreelog,
375 Opt_user_subvol_rm_allowed,
376
377 /* Rescue options */
378 Opt_rescue,
379 Opt_usebackuproot,
380 Opt_nologreplay,
381 Opt_ignorebadroots,
382 Opt_ignoredatacsums,
383 Opt_rescue_all,
384
385 /* Deprecated options */
386 Opt_recovery,
387 Opt_inode_cache, Opt_noinode_cache,
388
389 /* Debugging options */
390 Opt_check_integrity,
391 Opt_check_integrity_including_extent_data,
392 Opt_check_integrity_print_mask,
393 Opt_enospc_debug, Opt_noenospc_debug,
394#ifdef CONFIG_BTRFS_DEBUG
395 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
396#endif
397#ifdef CONFIG_BTRFS_FS_REF_VERIFY
398 Opt_ref_verify,
399#endif
400 Opt_err,
401};
402
403static const match_table_t tokens = {
404 {Opt_acl, "acl"},
405 {Opt_noacl, "noacl"},
406 {Opt_clear_cache, "clear_cache"},
407 {Opt_commit_interval, "commit=%u"},
408 {Opt_compress, "compress"},
409 {Opt_compress_type, "compress=%s"},
410 {Opt_compress_force, "compress-force"},
411 {Opt_compress_force_type, "compress-force=%s"},
412 {Opt_degraded, "degraded"},
413 {Opt_device, "device=%s"},
414 {Opt_fatal_errors, "fatal_errors=%s"},
415 {Opt_flushoncommit, "flushoncommit"},
416 {Opt_noflushoncommit, "noflushoncommit"},
417 {Opt_inode_cache, "inode_cache"},
418 {Opt_noinode_cache, "noinode_cache"},
419 {Opt_max_inline, "max_inline=%s"},
420 {Opt_barrier, "barrier"},
421 {Opt_nobarrier, "nobarrier"},
422 {Opt_datacow, "datacow"},
423 {Opt_nodatacow, "nodatacow"},
424 {Opt_datasum, "datasum"},
425 {Opt_nodatasum, "nodatasum"},
426 {Opt_defrag, "autodefrag"},
427 {Opt_nodefrag, "noautodefrag"},
428 {Opt_discard, "discard"},
429 {Opt_discard_mode, "discard=%s"},
430 {Opt_nodiscard, "nodiscard"},
431 {Opt_norecovery, "norecovery"},
432 {Opt_ratio, "metadata_ratio=%u"},
433 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
434 {Opt_skip_balance, "skip_balance"},
435 {Opt_space_cache, "space_cache"},
436 {Opt_no_space_cache, "nospace_cache"},
437 {Opt_space_cache_version, "space_cache=%s"},
438 {Opt_ssd, "ssd"},
439 {Opt_nossd, "nossd"},
440 {Opt_ssd_spread, "ssd_spread"},
441 {Opt_nossd_spread, "nossd_spread"},
442 {Opt_subvol, "subvol=%s"},
443 {Opt_subvol_empty, "subvol="},
444 {Opt_subvolid, "subvolid=%s"},
445 {Opt_thread_pool, "thread_pool=%u"},
446 {Opt_treelog, "treelog"},
447 {Opt_notreelog, "notreelog"},
448 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
449
450 /* Rescue options */
451 {Opt_rescue, "rescue=%s"},
452 /* Deprecated, with alias rescue=nologreplay */
453 {Opt_nologreplay, "nologreplay"},
454 /* Deprecated, with alias rescue=usebackuproot */
455 {Opt_usebackuproot, "usebackuproot"},
456
457 /* Deprecated options */
458 {Opt_recovery, "recovery"},
459
460 /* Debugging options */
461 {Opt_check_integrity, "check_int"},
462 {Opt_check_integrity_including_extent_data, "check_int_data"},
463 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
464 {Opt_enospc_debug, "enospc_debug"},
465 {Opt_noenospc_debug, "noenospc_debug"},
466#ifdef CONFIG_BTRFS_DEBUG
467 {Opt_fragment_data, "fragment=data"},
468 {Opt_fragment_metadata, "fragment=metadata"},
469 {Opt_fragment_all, "fragment=all"},
470#endif
471#ifdef CONFIG_BTRFS_FS_REF_VERIFY
472 {Opt_ref_verify, "ref_verify"},
473#endif
474 {Opt_err, NULL},
475};
476
477static const match_table_t rescue_tokens = {
478 {Opt_usebackuproot, "usebackuproot"},
479 {Opt_nologreplay, "nologreplay"},
480 {Opt_ignorebadroots, "ignorebadroots"},
481 {Opt_ignorebadroots, "ibadroots"},
482 {Opt_ignoredatacsums, "ignoredatacsums"},
483 {Opt_ignoredatacsums, "idatacsums"},
484 {Opt_rescue_all, "all"},
485 {Opt_err, NULL},
486};
487
488static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
489 const char *opt_name)
490{
491 if (fs_info->mount_opt & opt) {
492 btrfs_err(fs_info, "%s must be used with ro mount option",
493 opt_name);
494 return true;
495 }
496 return false;
497}
498
499static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
500{
501 char *opts;
502 char *orig;
503 char *p;
504 substring_t args[MAX_OPT_ARGS];
505 int ret = 0;
506
507 opts = kstrdup(options, GFP_KERNEL);
508 if (!opts)
509 return -ENOMEM;
510 orig = opts;
511
512 while ((p = strsep(&opts, ":")) != NULL) {
513 int token;
514
515 if (!*p)
516 continue;
517 token = match_token(p, rescue_tokens, args);
518 switch (token){
519 case Opt_usebackuproot:
520 btrfs_info(info,
521 "trying to use backup root at mount time");
522 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
523 break;
524 case Opt_nologreplay:
525 btrfs_set_and_info(info, NOLOGREPLAY,
526 "disabling log replay at mount time");
527 break;
528 case Opt_ignorebadroots:
529 btrfs_set_and_info(info, IGNOREBADROOTS,
530 "ignoring bad roots");
531 break;
532 case Opt_ignoredatacsums:
533 btrfs_set_and_info(info, IGNOREDATACSUMS,
534 "ignoring data csums");
535 break;
536 case Opt_rescue_all:
537 btrfs_info(info, "enabling all of the rescue options");
538 btrfs_set_and_info(info, IGNOREDATACSUMS,
539 "ignoring data csums");
540 btrfs_set_and_info(info, IGNOREBADROOTS,
541 "ignoring bad roots");
542 btrfs_set_and_info(info, NOLOGREPLAY,
543 "disabling log replay at mount time");
544 break;
545 case Opt_err:
546 btrfs_info(info, "unrecognized rescue option '%s'", p);
547 ret = -EINVAL;
548 goto out;
549 default:
550 break;
551 }
552
553 }
554out:
555 kfree(orig);
556 return ret;
557}
558
559/*
560 * Regular mount options parser. Everything that is needed only when
561 * reading in a new superblock is parsed here.
562 * XXX JDM: This needs to be cleaned up for remount.
563 */
564int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
565 unsigned long new_flags)
566{
567 substring_t args[MAX_OPT_ARGS];
568 char *p, *num;
569 int intarg;
570 int ret = 0;
571 char *compress_type;
572 bool compress_force = false;
573 enum btrfs_compression_type saved_compress_type;
574 int saved_compress_level;
575 bool saved_compress_force;
576 int no_compress = 0;
577
578 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
579 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
580 else if (btrfs_free_space_cache_v1_active(info)) {
581 if (btrfs_is_zoned(info)) {
582 btrfs_info(info,
583 "zoned: clearing existing space cache");
584 btrfs_set_super_cache_generation(info->super_copy, 0);
585 } else {
586 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
587 }
588 }
589
590 /*
591 * Even the options are empty, we still need to do extra check
592 * against new flags
593 */
594 if (!options)
595 goto check;
596
597 while ((p = strsep(&options, ",")) != NULL) {
598 int token;
599 if (!*p)
600 continue;
601
602 token = match_token(p, tokens, args);
603 switch (token) {
604 case Opt_degraded:
605 btrfs_info(info, "allowing degraded mounts");
606 btrfs_set_opt(info->mount_opt, DEGRADED);
607 break;
608 case Opt_subvol:
609 case Opt_subvol_empty:
610 case Opt_subvolid:
611 case Opt_device:
612 /*
613 * These are parsed by btrfs_parse_subvol_options or
614 * btrfs_parse_device_options and can be ignored here.
615 */
616 break;
617 case Opt_nodatasum:
618 btrfs_set_and_info(info, NODATASUM,
619 "setting nodatasum");
620 break;
621 case Opt_datasum:
622 if (btrfs_test_opt(info, NODATASUM)) {
623 if (btrfs_test_opt(info, NODATACOW))
624 btrfs_info(info,
625 "setting datasum, datacow enabled");
626 else
627 btrfs_info(info, "setting datasum");
628 }
629 btrfs_clear_opt(info->mount_opt, NODATACOW);
630 btrfs_clear_opt(info->mount_opt, NODATASUM);
631 break;
632 case Opt_nodatacow:
633 if (!btrfs_test_opt(info, NODATACOW)) {
634 if (!btrfs_test_opt(info, COMPRESS) ||
635 !btrfs_test_opt(info, FORCE_COMPRESS)) {
636 btrfs_info(info,
637 "setting nodatacow, compression disabled");
638 } else {
639 btrfs_info(info, "setting nodatacow");
640 }
641 }
642 btrfs_clear_opt(info->mount_opt, COMPRESS);
643 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
644 btrfs_set_opt(info->mount_opt, NODATACOW);
645 btrfs_set_opt(info->mount_opt, NODATASUM);
646 break;
647 case Opt_datacow:
648 btrfs_clear_and_info(info, NODATACOW,
649 "setting datacow");
650 break;
651 case Opt_compress_force:
652 case Opt_compress_force_type:
653 compress_force = true;
654 fallthrough;
655 case Opt_compress:
656 case Opt_compress_type:
657 saved_compress_type = btrfs_test_opt(info,
658 COMPRESS) ?
659 info->compress_type : BTRFS_COMPRESS_NONE;
660 saved_compress_force =
661 btrfs_test_opt(info, FORCE_COMPRESS);
662 saved_compress_level = info->compress_level;
663 if (token == Opt_compress ||
664 token == Opt_compress_force ||
665 strncmp(args[0].from, "zlib", 4) == 0) {
666 compress_type = "zlib";
667
668 info->compress_type = BTRFS_COMPRESS_ZLIB;
669 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
670 /*
671 * args[0] contains uninitialized data since
672 * for these tokens we don't expect any
673 * parameter.
674 */
675 if (token != Opt_compress &&
676 token != Opt_compress_force)
677 info->compress_level =
678 btrfs_compress_str2level(
679 BTRFS_COMPRESS_ZLIB,
680 args[0].from + 4);
681 btrfs_set_opt(info->mount_opt, COMPRESS);
682 btrfs_clear_opt(info->mount_opt, NODATACOW);
683 btrfs_clear_opt(info->mount_opt, NODATASUM);
684 no_compress = 0;
685 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
686 compress_type = "lzo";
687 info->compress_type = BTRFS_COMPRESS_LZO;
688 info->compress_level = 0;
689 btrfs_set_opt(info->mount_opt, COMPRESS);
690 btrfs_clear_opt(info->mount_opt, NODATACOW);
691 btrfs_clear_opt(info->mount_opt, NODATASUM);
692 btrfs_set_fs_incompat(info, COMPRESS_LZO);
693 no_compress = 0;
694 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
695 compress_type = "zstd";
696 info->compress_type = BTRFS_COMPRESS_ZSTD;
697 info->compress_level =
698 btrfs_compress_str2level(
699 BTRFS_COMPRESS_ZSTD,
700 args[0].from + 4);
701 btrfs_set_opt(info->mount_opt, COMPRESS);
702 btrfs_clear_opt(info->mount_opt, NODATACOW);
703 btrfs_clear_opt(info->mount_opt, NODATASUM);
704 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
705 no_compress = 0;
706 } else if (strncmp(args[0].from, "no", 2) == 0) {
707 compress_type = "no";
708 info->compress_level = 0;
709 info->compress_type = 0;
710 btrfs_clear_opt(info->mount_opt, COMPRESS);
711 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
712 compress_force = false;
713 no_compress++;
714 } else {
715 ret = -EINVAL;
716 goto out;
717 }
718
719 if (compress_force) {
720 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
721 } else {
722 /*
723 * If we remount from compress-force=xxx to
724 * compress=xxx, we need clear FORCE_COMPRESS
725 * flag, otherwise, there is no way for users
726 * to disable forcible compression separately.
727 */
728 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
729 }
730 if (no_compress == 1) {
731 btrfs_info(info, "use no compression");
732 } else if ((info->compress_type != saved_compress_type) ||
733 (compress_force != saved_compress_force) ||
734 (info->compress_level != saved_compress_level)) {
735 btrfs_info(info, "%s %s compression, level %d",
736 (compress_force) ? "force" : "use",
737 compress_type, info->compress_level);
738 }
739 compress_force = false;
740 break;
741 case Opt_ssd:
742 btrfs_set_and_info(info, SSD,
743 "enabling ssd optimizations");
744 btrfs_clear_opt(info->mount_opt, NOSSD);
745 break;
746 case Opt_ssd_spread:
747 btrfs_set_and_info(info, SSD,
748 "enabling ssd optimizations");
749 btrfs_set_and_info(info, SSD_SPREAD,
750 "using spread ssd allocation scheme");
751 btrfs_clear_opt(info->mount_opt, NOSSD);
752 break;
753 case Opt_nossd:
754 btrfs_set_opt(info->mount_opt, NOSSD);
755 btrfs_clear_and_info(info, SSD,
756 "not using ssd optimizations");
757 fallthrough;
758 case Opt_nossd_spread:
759 btrfs_clear_and_info(info, SSD_SPREAD,
760 "not using spread ssd allocation scheme");
761 break;
762 case Opt_barrier:
763 btrfs_clear_and_info(info, NOBARRIER,
764 "turning on barriers");
765 break;
766 case Opt_nobarrier:
767 btrfs_set_and_info(info, NOBARRIER,
768 "turning off barriers");
769 break;
770 case Opt_thread_pool:
771 ret = match_int(&args[0], &intarg);
772 if (ret) {
773 goto out;
774 } else if (intarg == 0) {
775 ret = -EINVAL;
776 goto out;
777 }
778 info->thread_pool_size = intarg;
779 break;
780 case Opt_max_inline:
781 num = match_strdup(&args[0]);
782 if (num) {
783 info->max_inline = memparse(num, NULL);
784 kfree(num);
785
786 if (info->max_inline) {
787 info->max_inline = min_t(u64,
788 info->max_inline,
789 info->sectorsize);
790 }
791 btrfs_info(info, "max_inline at %llu",
792 info->max_inline);
793 } else {
794 ret = -ENOMEM;
795 goto out;
796 }
797 break;
798 case Opt_acl:
799#ifdef CONFIG_BTRFS_FS_POSIX_ACL
800 info->sb->s_flags |= SB_POSIXACL;
801 break;
802#else
803 btrfs_err(info, "support for ACL not compiled in!");
804 ret = -EINVAL;
805 goto out;
806#endif
807 case Opt_noacl:
808 info->sb->s_flags &= ~SB_POSIXACL;
809 break;
810 case Opt_notreelog:
811 btrfs_set_and_info(info, NOTREELOG,
812 "disabling tree log");
813 break;
814 case Opt_treelog:
815 btrfs_clear_and_info(info, NOTREELOG,
816 "enabling tree log");
817 break;
818 case Opt_norecovery:
819 case Opt_nologreplay:
820 btrfs_warn(info,
821 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
822 btrfs_set_and_info(info, NOLOGREPLAY,
823 "disabling log replay at mount time");
824 break;
825 case Opt_flushoncommit:
826 btrfs_set_and_info(info, FLUSHONCOMMIT,
827 "turning on flush-on-commit");
828 break;
829 case Opt_noflushoncommit:
830 btrfs_clear_and_info(info, FLUSHONCOMMIT,
831 "turning off flush-on-commit");
832 break;
833 case Opt_ratio:
834 ret = match_int(&args[0], &intarg);
835 if (ret)
836 goto out;
837 info->metadata_ratio = intarg;
838 btrfs_info(info, "metadata ratio %u",
839 info->metadata_ratio);
840 break;
841 case Opt_discard:
842 case Opt_discard_mode:
843 if (token == Opt_discard ||
844 strcmp(args[0].from, "sync") == 0) {
845 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
846 btrfs_set_and_info(info, DISCARD_SYNC,
847 "turning on sync discard");
848 } else if (strcmp(args[0].from, "async") == 0) {
849 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
850 btrfs_set_and_info(info, DISCARD_ASYNC,
851 "turning on async discard");
852 } else {
853 ret = -EINVAL;
854 goto out;
855 }
856 break;
857 case Opt_nodiscard:
858 btrfs_clear_and_info(info, DISCARD_SYNC,
859 "turning off discard");
860 btrfs_clear_and_info(info, DISCARD_ASYNC,
861 "turning off async discard");
862 break;
863 case Opt_space_cache:
864 case Opt_space_cache_version:
865 if (token == Opt_space_cache ||
866 strcmp(args[0].from, "v1") == 0) {
867 btrfs_clear_opt(info->mount_opt,
868 FREE_SPACE_TREE);
869 btrfs_set_and_info(info, SPACE_CACHE,
870 "enabling disk space caching");
871 } else if (strcmp(args[0].from, "v2") == 0) {
872 btrfs_clear_opt(info->mount_opt,
873 SPACE_CACHE);
874 btrfs_set_and_info(info, FREE_SPACE_TREE,
875 "enabling free space tree");
876 } else {
877 ret = -EINVAL;
878 goto out;
879 }
880 break;
881 case Opt_rescan_uuid_tree:
882 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
883 break;
884 case Opt_no_space_cache:
885 if (btrfs_test_opt(info, SPACE_CACHE)) {
886 btrfs_clear_and_info(info, SPACE_CACHE,
887 "disabling disk space caching");
888 }
889 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
890 btrfs_clear_and_info(info, FREE_SPACE_TREE,
891 "disabling free space tree");
892 }
893 break;
894 case Opt_inode_cache:
895 case Opt_noinode_cache:
896 btrfs_warn(info,
897 "the 'inode_cache' option is deprecated and has no effect since 5.11");
898 break;
899 case Opt_clear_cache:
900 btrfs_set_and_info(info, CLEAR_CACHE,
901 "force clearing of disk cache");
902 break;
903 case Opt_user_subvol_rm_allowed:
904 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
905 break;
906 case Opt_enospc_debug:
907 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
908 break;
909 case Opt_noenospc_debug:
910 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
911 break;
912 case Opt_defrag:
913 btrfs_set_and_info(info, AUTO_DEFRAG,
914 "enabling auto defrag");
915 break;
916 case Opt_nodefrag:
917 btrfs_clear_and_info(info, AUTO_DEFRAG,
918 "disabling auto defrag");
919 break;
920 case Opt_recovery:
921 case Opt_usebackuproot:
922 btrfs_warn(info,
923 "'%s' is deprecated, use 'rescue=usebackuproot' instead",
924 token == Opt_recovery ? "recovery" :
925 "usebackuproot");
926 btrfs_info(info,
927 "trying to use backup root at mount time");
928 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
929 break;
930 case Opt_skip_balance:
931 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
932 break;
933#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
934 case Opt_check_integrity_including_extent_data:
935 btrfs_info(info,
936 "enabling check integrity including extent data");
937 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
938 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
939 break;
940 case Opt_check_integrity:
941 btrfs_info(info, "enabling check integrity");
942 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
943 break;
944 case Opt_check_integrity_print_mask:
945 ret = match_int(&args[0], &intarg);
946 if (ret)
947 goto out;
948 info->check_integrity_print_mask = intarg;
949 btrfs_info(info, "check_integrity_print_mask 0x%x",
950 info->check_integrity_print_mask);
951 break;
952#else
953 case Opt_check_integrity_including_extent_data:
954 case Opt_check_integrity:
955 case Opt_check_integrity_print_mask:
956 btrfs_err(info,
957 "support for check_integrity* not compiled in!");
958 ret = -EINVAL;
959 goto out;
960#endif
961 case Opt_fatal_errors:
962 if (strcmp(args[0].from, "panic") == 0)
963 btrfs_set_opt(info->mount_opt,
964 PANIC_ON_FATAL_ERROR);
965 else if (strcmp(args[0].from, "bug") == 0)
966 btrfs_clear_opt(info->mount_opt,
967 PANIC_ON_FATAL_ERROR);
968 else {
969 ret = -EINVAL;
970 goto out;
971 }
972 break;
973 case Opt_commit_interval:
974 intarg = 0;
975 ret = match_int(&args[0], &intarg);
976 if (ret)
977 goto out;
978 if (intarg == 0) {
979 btrfs_info(info,
980 "using default commit interval %us",
981 BTRFS_DEFAULT_COMMIT_INTERVAL);
982 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
983 } else if (intarg > 300) {
984 btrfs_warn(info, "excessive commit interval %d",
985 intarg);
986 }
987 info->commit_interval = intarg;
988 break;
989 case Opt_rescue:
990 ret = parse_rescue_options(info, args[0].from);
991 if (ret < 0)
992 goto out;
993 break;
994#ifdef CONFIG_BTRFS_DEBUG
995 case Opt_fragment_all:
996 btrfs_info(info, "fragmenting all space");
997 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
998 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
999 break;
1000 case Opt_fragment_metadata:
1001 btrfs_info(info, "fragmenting metadata");
1002 btrfs_set_opt(info->mount_opt,
1003 FRAGMENT_METADATA);
1004 break;
1005 case Opt_fragment_data:
1006 btrfs_info(info, "fragmenting data");
1007 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1008 break;
1009#endif
1010#ifdef CONFIG_BTRFS_FS_REF_VERIFY
1011 case Opt_ref_verify:
1012 btrfs_info(info, "doing ref verification");
1013 btrfs_set_opt(info->mount_opt, REF_VERIFY);
1014 break;
1015#endif
1016 case Opt_err:
1017 btrfs_err(info, "unrecognized mount option '%s'", p);
1018 ret = -EINVAL;
1019 goto out;
1020 default:
1021 break;
1022 }
1023 }
1024check:
1025 /* We're read-only, don't have to check. */
1026 if (new_flags & SB_RDONLY)
1027 goto out;
1028
1029 if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
1030 check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
1031 check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
1032 ret = -EINVAL;
1033out:
1034 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
1035 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
1036 !btrfs_test_opt(info, CLEAR_CACHE)) {
1037 btrfs_err(info, "cannot disable free space tree");
1038 ret = -EINVAL;
1039
1040 }
1041 if (!ret)
1042 ret = btrfs_check_mountopts_zoned(info);
1043 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
1044 btrfs_info(info, "disk space caching is enabled");
1045 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
1046 btrfs_info(info, "using free space tree");
1047 return ret;
1048}
1049
1050/*
1051 * Parse mount options that are required early in the mount process.
1052 *
1053 * All other options will be parsed on much later in the mount process and
1054 * only when we need to allocate a new super block.
1055 */
1056static int btrfs_parse_device_options(const char *options, fmode_t flags,
1057 void *holder)
1058{
1059 substring_t args[MAX_OPT_ARGS];
1060 char *device_name, *opts, *orig, *p;
1061 struct btrfs_device *device = NULL;
1062 int error = 0;
1063
1064 lockdep_assert_held(&uuid_mutex);
1065
1066 if (!options)
1067 return 0;
1068
1069 /*
1070 * strsep changes the string, duplicate it because btrfs_parse_options
1071 * gets called later
1072 */
1073 opts = kstrdup(options, GFP_KERNEL);
1074 if (!opts)
1075 return -ENOMEM;
1076 orig = opts;
1077
1078 while ((p = strsep(&opts, ",")) != NULL) {
1079 int token;
1080
1081 if (!*p)
1082 continue;
1083
1084 token = match_token(p, tokens, args);
1085 if (token == Opt_device) {
1086 device_name = match_strdup(&args[0]);
1087 if (!device_name) {
1088 error = -ENOMEM;
1089 goto out;
1090 }
1091 device = btrfs_scan_one_device(device_name, flags,
1092 holder);
1093 kfree(device_name);
1094 if (IS_ERR(device)) {
1095 error = PTR_ERR(device);
1096 goto out;
1097 }
1098 }
1099 }
1100
1101out:
1102 kfree(orig);
1103 return error;
1104}
1105
1106/*
1107 * Parse mount options that are related to subvolume id
1108 *
1109 * The value is later passed to mount_subvol()
1110 */
1111static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1112 u64 *subvol_objectid)
1113{
1114 substring_t args[MAX_OPT_ARGS];
1115 char *opts, *orig, *p;
1116 int error = 0;
1117 u64 subvolid;
1118
1119 if (!options)
1120 return 0;
1121
1122 /*
1123 * strsep changes the string, duplicate it because
1124 * btrfs_parse_device_options gets called later
1125 */
1126 opts = kstrdup(options, GFP_KERNEL);
1127 if (!opts)
1128 return -ENOMEM;
1129 orig = opts;
1130
1131 while ((p = strsep(&opts, ",")) != NULL) {
1132 int token;
1133 if (!*p)
1134 continue;
1135
1136 token = match_token(p, tokens, args);
1137 switch (token) {
1138 case Opt_subvol:
1139 kfree(*subvol_name);
1140 *subvol_name = match_strdup(&args[0]);
1141 if (!*subvol_name) {
1142 error = -ENOMEM;
1143 goto out;
1144 }
1145 break;
1146 case Opt_subvolid:
1147 error = match_u64(&args[0], &subvolid);
1148 if (error)
1149 goto out;
1150
1151 /* we want the original fs_tree */
1152 if (subvolid == 0)
1153 subvolid = BTRFS_FS_TREE_OBJECTID;
1154
1155 *subvol_objectid = subvolid;
1156 break;
1157 default:
1158 break;
1159 }
1160 }
1161
1162out:
1163 kfree(orig);
1164 return error;
1165}
1166
1167char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1168 u64 subvol_objectid)
1169{
1170 struct btrfs_root *root = fs_info->tree_root;
1171 struct btrfs_root *fs_root = NULL;
1172 struct btrfs_root_ref *root_ref;
1173 struct btrfs_inode_ref *inode_ref;
1174 struct btrfs_key key;
1175 struct btrfs_path *path = NULL;
1176 char *name = NULL, *ptr;
1177 u64 dirid;
1178 int len;
1179 int ret;
1180
1181 path = btrfs_alloc_path();
1182 if (!path) {
1183 ret = -ENOMEM;
1184 goto err;
1185 }
1186
1187 name = kmalloc(PATH_MAX, GFP_KERNEL);
1188 if (!name) {
1189 ret = -ENOMEM;
1190 goto err;
1191 }
1192 ptr = name + PATH_MAX - 1;
1193 ptr[0] = '\0';
1194
1195 /*
1196 * Walk up the subvolume trees in the tree of tree roots by root
1197 * backrefs until we hit the top-level subvolume.
1198 */
1199 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1200 key.objectid = subvol_objectid;
1201 key.type = BTRFS_ROOT_BACKREF_KEY;
1202 key.offset = (u64)-1;
1203
1204 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1205 if (ret < 0) {
1206 goto err;
1207 } else if (ret > 0) {
1208 ret = btrfs_previous_item(root, path, subvol_objectid,
1209 BTRFS_ROOT_BACKREF_KEY);
1210 if (ret < 0) {
1211 goto err;
1212 } else if (ret > 0) {
1213 ret = -ENOENT;
1214 goto err;
1215 }
1216 }
1217
1218 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1219 subvol_objectid = key.offset;
1220
1221 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1222 struct btrfs_root_ref);
1223 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1224 ptr -= len + 1;
1225 if (ptr < name) {
1226 ret = -ENAMETOOLONG;
1227 goto err;
1228 }
1229 read_extent_buffer(path->nodes[0], ptr + 1,
1230 (unsigned long)(root_ref + 1), len);
1231 ptr[0] = '/';
1232 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1233 btrfs_release_path(path);
1234
1235 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1236 if (IS_ERR(fs_root)) {
1237 ret = PTR_ERR(fs_root);
1238 fs_root = NULL;
1239 goto err;
1240 }
1241
1242 /*
1243 * Walk up the filesystem tree by inode refs until we hit the
1244 * root directory.
1245 */
1246 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1247 key.objectid = dirid;
1248 key.type = BTRFS_INODE_REF_KEY;
1249 key.offset = (u64)-1;
1250
1251 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1252 if (ret < 0) {
1253 goto err;
1254 } else if (ret > 0) {
1255 ret = btrfs_previous_item(fs_root, path, dirid,
1256 BTRFS_INODE_REF_KEY);
1257 if (ret < 0) {
1258 goto err;
1259 } else if (ret > 0) {
1260 ret = -ENOENT;
1261 goto err;
1262 }
1263 }
1264
1265 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1266 dirid = key.offset;
1267
1268 inode_ref = btrfs_item_ptr(path->nodes[0],
1269 path->slots[0],
1270 struct btrfs_inode_ref);
1271 len = btrfs_inode_ref_name_len(path->nodes[0],
1272 inode_ref);
1273 ptr -= len + 1;
1274 if (ptr < name) {
1275 ret = -ENAMETOOLONG;
1276 goto err;
1277 }
1278 read_extent_buffer(path->nodes[0], ptr + 1,
1279 (unsigned long)(inode_ref + 1), len);
1280 ptr[0] = '/';
1281 btrfs_release_path(path);
1282 }
1283 btrfs_put_root(fs_root);
1284 fs_root = NULL;
1285 }
1286
1287 btrfs_free_path(path);
1288 if (ptr == name + PATH_MAX - 1) {
1289 name[0] = '/';
1290 name[1] = '\0';
1291 } else {
1292 memmove(name, ptr, name + PATH_MAX - ptr);
1293 }
1294 return name;
1295
1296err:
1297 btrfs_put_root(fs_root);
1298 btrfs_free_path(path);
1299 kfree(name);
1300 return ERR_PTR(ret);
1301}
1302
1303static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1304{
1305 struct btrfs_root *root = fs_info->tree_root;
1306 struct btrfs_dir_item *di;
1307 struct btrfs_path *path;
1308 struct btrfs_key location;
1309 u64 dir_id;
1310
1311 path = btrfs_alloc_path();
1312 if (!path)
1313 return -ENOMEM;
1314
1315 /*
1316 * Find the "default" dir item which points to the root item that we
1317 * will mount by default if we haven't been given a specific subvolume
1318 * to mount.
1319 */
1320 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1321 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1322 if (IS_ERR(di)) {
1323 btrfs_free_path(path);
1324 return PTR_ERR(di);
1325 }
1326 if (!di) {
1327 /*
1328 * Ok the default dir item isn't there. This is weird since
1329 * it's always been there, but don't freak out, just try and
1330 * mount the top-level subvolume.
1331 */
1332 btrfs_free_path(path);
1333 *objectid = BTRFS_FS_TREE_OBJECTID;
1334 return 0;
1335 }
1336
1337 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1338 btrfs_free_path(path);
1339 *objectid = location.objectid;
1340 return 0;
1341}
1342
1343static int btrfs_fill_super(struct super_block *sb,
1344 struct btrfs_fs_devices *fs_devices,
1345 void *data)
1346{
1347 struct inode *inode;
1348 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1349 int err;
1350
1351 sb->s_maxbytes = MAX_LFS_FILESIZE;
1352 sb->s_magic = BTRFS_SUPER_MAGIC;
1353 sb->s_op = &btrfs_super_ops;
1354 sb->s_d_op = &btrfs_dentry_operations;
1355 sb->s_export_op = &btrfs_export_ops;
1356 sb->s_xattr = btrfs_xattr_handlers;
1357 sb->s_time_gran = 1;
1358#ifdef CONFIG_BTRFS_FS_POSIX_ACL
1359 sb->s_flags |= SB_POSIXACL;
1360#endif
1361 sb->s_flags |= SB_I_VERSION;
1362 sb->s_iflags |= SB_I_CGROUPWB;
1363
1364 err = super_setup_bdi(sb);
1365 if (err) {
1366 btrfs_err(fs_info, "super_setup_bdi failed");
1367 return err;
1368 }
1369
1370 err = open_ctree(sb, fs_devices, (char *)data);
1371 if (err) {
1372 btrfs_err(fs_info, "open_ctree failed");
1373 return err;
1374 }
1375
1376 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1377 if (IS_ERR(inode)) {
1378 err = PTR_ERR(inode);
1379 goto fail_close;
1380 }
1381
1382 sb->s_root = d_make_root(inode);
1383 if (!sb->s_root) {
1384 err = -ENOMEM;
1385 goto fail_close;
1386 }
1387
1388 cleancache_init_fs(sb);
1389 sb->s_flags |= SB_ACTIVE;
1390 return 0;
1391
1392fail_close:
1393 close_ctree(fs_info);
1394 return err;
1395}
1396
1397int btrfs_sync_fs(struct super_block *sb, int wait)
1398{
1399 struct btrfs_trans_handle *trans;
1400 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1401 struct btrfs_root *root = fs_info->tree_root;
1402
1403 trace_btrfs_sync_fs(fs_info, wait);
1404
1405 if (!wait) {
1406 filemap_flush(fs_info->btree_inode->i_mapping);
1407 return 0;
1408 }
1409
1410 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1411
1412 trans = btrfs_attach_transaction_barrier(root);
1413 if (IS_ERR(trans)) {
1414 /* no transaction, don't bother */
1415 if (PTR_ERR(trans) == -ENOENT) {
1416 /*
1417 * Exit unless we have some pending changes
1418 * that need to go through commit
1419 */
1420 if (fs_info->pending_changes == 0)
1421 return 0;
1422 /*
1423 * A non-blocking test if the fs is frozen. We must not
1424 * start a new transaction here otherwise a deadlock
1425 * happens. The pending operations are delayed to the
1426 * next commit after thawing.
1427 */
1428 if (sb_start_write_trylock(sb))
1429 sb_end_write(sb);
1430 else
1431 return 0;
1432 trans = btrfs_start_transaction(root, 0);
1433 }
1434 if (IS_ERR(trans))
1435 return PTR_ERR(trans);
1436 }
1437 return btrfs_commit_transaction(trans);
1438}
1439
1440static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1441{
1442 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1443 *printed = true;
1444}
1445
1446static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1447{
1448 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1449 const char *compress_type;
1450 const char *subvol_name;
1451 bool printed = false;
1452
1453 if (btrfs_test_opt(info, DEGRADED))
1454 seq_puts(seq, ",degraded");
1455 if (btrfs_test_opt(info, NODATASUM))
1456 seq_puts(seq, ",nodatasum");
1457 if (btrfs_test_opt(info, NODATACOW))
1458 seq_puts(seq, ",nodatacow");
1459 if (btrfs_test_opt(info, NOBARRIER))
1460 seq_puts(seq, ",nobarrier");
1461 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1462 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1463 if (info->thread_pool_size != min_t(unsigned long,
1464 num_online_cpus() + 2, 8))
1465 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1466 if (btrfs_test_opt(info, COMPRESS)) {
1467 compress_type = btrfs_compress_type2str(info->compress_type);
1468 if (btrfs_test_opt(info, FORCE_COMPRESS))
1469 seq_printf(seq, ",compress-force=%s", compress_type);
1470 else
1471 seq_printf(seq, ",compress=%s", compress_type);
1472 if (info->compress_level)
1473 seq_printf(seq, ":%d", info->compress_level);
1474 }
1475 if (btrfs_test_opt(info, NOSSD))
1476 seq_puts(seq, ",nossd");
1477 if (btrfs_test_opt(info, SSD_SPREAD))
1478 seq_puts(seq, ",ssd_spread");
1479 else if (btrfs_test_opt(info, SSD))
1480 seq_puts(seq, ",ssd");
1481 if (btrfs_test_opt(info, NOTREELOG))
1482 seq_puts(seq, ",notreelog");
1483 if (btrfs_test_opt(info, NOLOGREPLAY))
1484 print_rescue_option(seq, "nologreplay", &printed);
1485 if (btrfs_test_opt(info, USEBACKUPROOT))
1486 print_rescue_option(seq, "usebackuproot", &printed);
1487 if (btrfs_test_opt(info, IGNOREBADROOTS))
1488 print_rescue_option(seq, "ignorebadroots", &printed);
1489 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1490 print_rescue_option(seq, "ignoredatacsums", &printed);
1491 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1492 seq_puts(seq, ",flushoncommit");
1493 if (btrfs_test_opt(info, DISCARD_SYNC))
1494 seq_puts(seq, ",discard");
1495 if (btrfs_test_opt(info, DISCARD_ASYNC))
1496 seq_puts(seq, ",discard=async");
1497 if (!(info->sb->s_flags & SB_POSIXACL))
1498 seq_puts(seq, ",noacl");
1499 if (btrfs_free_space_cache_v1_active(info))
1500 seq_puts(seq, ",space_cache");
1501 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1502 seq_puts(seq, ",space_cache=v2");
1503 else
1504 seq_puts(seq, ",nospace_cache");
1505 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1506 seq_puts(seq, ",rescan_uuid_tree");
1507 if (btrfs_test_opt(info, CLEAR_CACHE))
1508 seq_puts(seq, ",clear_cache");
1509 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1510 seq_puts(seq, ",user_subvol_rm_allowed");
1511 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1512 seq_puts(seq, ",enospc_debug");
1513 if (btrfs_test_opt(info, AUTO_DEFRAG))
1514 seq_puts(seq, ",autodefrag");
1515 if (btrfs_test_opt(info, SKIP_BALANCE))
1516 seq_puts(seq, ",skip_balance");
1517#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1518 if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1519 seq_puts(seq, ",check_int_data");
1520 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1521 seq_puts(seq, ",check_int");
1522 if (info->check_integrity_print_mask)
1523 seq_printf(seq, ",check_int_print_mask=%d",
1524 info->check_integrity_print_mask);
1525#endif
1526 if (info->metadata_ratio)
1527 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1528 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1529 seq_puts(seq, ",fatal_errors=panic");
1530 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1531 seq_printf(seq, ",commit=%u", info->commit_interval);
1532#ifdef CONFIG_BTRFS_DEBUG
1533 if (btrfs_test_opt(info, FRAGMENT_DATA))
1534 seq_puts(seq, ",fragment=data");
1535 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1536 seq_puts(seq, ",fragment=metadata");
1537#endif
1538 if (btrfs_test_opt(info, REF_VERIFY))
1539 seq_puts(seq, ",ref_verify");
1540 seq_printf(seq, ",subvolid=%llu",
1541 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1542 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1543 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1544 if (!IS_ERR(subvol_name)) {
1545 seq_puts(seq, ",subvol=");
1546 seq_escape(seq, subvol_name, " \t\n\\");
1547 kfree(subvol_name);
1548 }
1549 return 0;
1550}
1551
1552static int btrfs_test_super(struct super_block *s, void *data)
1553{
1554 struct btrfs_fs_info *p = data;
1555 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1556
1557 return fs_info->fs_devices == p->fs_devices;
1558}
1559
1560static int btrfs_set_super(struct super_block *s, void *data)
1561{
1562 int err = set_anon_super(s, data);
1563 if (!err)
1564 s->s_fs_info = data;
1565 return err;
1566}
1567
1568/*
1569 * subvolumes are identified by ino 256
1570 */
1571static inline int is_subvolume_inode(struct inode *inode)
1572{
1573 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1574 return 1;
1575 return 0;
1576}
1577
1578static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1579 struct vfsmount *mnt)
1580{
1581 struct dentry *root;
1582 int ret;
1583
1584 if (!subvol_name) {
1585 if (!subvol_objectid) {
1586 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1587 &subvol_objectid);
1588 if (ret) {
1589 root = ERR_PTR(ret);
1590 goto out;
1591 }
1592 }
1593 subvol_name = btrfs_get_subvol_name_from_objectid(
1594 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1595 if (IS_ERR(subvol_name)) {
1596 root = ERR_CAST(subvol_name);
1597 subvol_name = NULL;
1598 goto out;
1599 }
1600
1601 }
1602
1603 root = mount_subtree(mnt, subvol_name);
1604 /* mount_subtree() drops our reference on the vfsmount. */
1605 mnt = NULL;
1606
1607 if (!IS_ERR(root)) {
1608 struct super_block *s = root->d_sb;
1609 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1610 struct inode *root_inode = d_inode(root);
1611 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1612
1613 ret = 0;
1614 if (!is_subvolume_inode(root_inode)) {
1615 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1616 subvol_name);
1617 ret = -EINVAL;
1618 }
1619 if (subvol_objectid && root_objectid != subvol_objectid) {
1620 /*
1621 * This will also catch a race condition where a
1622 * subvolume which was passed by ID is renamed and
1623 * another subvolume is renamed over the old location.
1624 */
1625 btrfs_err(fs_info,
1626 "subvol '%s' does not match subvolid %llu",
1627 subvol_name, subvol_objectid);
1628 ret = -EINVAL;
1629 }
1630 if (ret) {
1631 dput(root);
1632 root = ERR_PTR(ret);
1633 deactivate_locked_super(s);
1634 }
1635 }
1636
1637out:
1638 mntput(mnt);
1639 kfree(subvol_name);
1640 return root;
1641}
1642
1643/*
1644 * Find a superblock for the given device / mount point.
1645 *
1646 * Note: This is based on mount_bdev from fs/super.c with a few additions
1647 * for multiple device setup. Make sure to keep it in sync.
1648 */
1649static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1650 int flags, const char *device_name, void *data)
1651{
1652 struct block_device *bdev = NULL;
1653 struct super_block *s;
1654 struct btrfs_device *device = NULL;
1655 struct btrfs_fs_devices *fs_devices = NULL;
1656 struct btrfs_fs_info *fs_info = NULL;
1657 void *new_sec_opts = NULL;
1658 fmode_t mode = FMODE_READ;
1659 int error = 0;
1660
1661 if (!(flags & SB_RDONLY))
1662 mode |= FMODE_WRITE;
1663
1664 if (data) {
1665 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1666 if (error)
1667 return ERR_PTR(error);
1668 }
1669
1670 /*
1671 * Setup a dummy root and fs_info for test/set super. This is because
1672 * we don't actually fill this stuff out until open_ctree, but we need
1673 * then open_ctree will properly initialize the file system specific
1674 * settings later. btrfs_init_fs_info initializes the static elements
1675 * of the fs_info (locks and such) to make cleanup easier if we find a
1676 * superblock with our given fs_devices later on at sget() time.
1677 */
1678 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1679 if (!fs_info) {
1680 error = -ENOMEM;
1681 goto error_sec_opts;
1682 }
1683 btrfs_init_fs_info(fs_info);
1684
1685 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1686 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1687 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1688 error = -ENOMEM;
1689 goto error_fs_info;
1690 }
1691
1692 mutex_lock(&uuid_mutex);
1693 error = btrfs_parse_device_options(data, mode, fs_type);
1694 if (error) {
1695 mutex_unlock(&uuid_mutex);
1696 goto error_fs_info;
1697 }
1698
1699 device = btrfs_scan_one_device(device_name, mode, fs_type);
1700 if (IS_ERR(device)) {
1701 mutex_unlock(&uuid_mutex);
1702 error = PTR_ERR(device);
1703 goto error_fs_info;
1704 }
1705
1706 fs_devices = device->fs_devices;
1707 fs_info->fs_devices = fs_devices;
1708
1709 error = btrfs_open_devices(fs_devices, mode, fs_type);
1710 mutex_unlock(&uuid_mutex);
1711 if (error)
1712 goto error_fs_info;
1713
1714 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1715 error = -EACCES;
1716 goto error_close_devices;
1717 }
1718
1719 bdev = fs_devices->latest_bdev;
1720 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1721 fs_info);
1722 if (IS_ERR(s)) {
1723 error = PTR_ERR(s);
1724 goto error_close_devices;
1725 }
1726
1727 if (s->s_root) {
1728 btrfs_close_devices(fs_devices);
1729 btrfs_free_fs_info(fs_info);
1730 if ((flags ^ s->s_flags) & SB_RDONLY)
1731 error = -EBUSY;
1732 } else {
1733 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1734 btrfs_sb(s)->bdev_holder = fs_type;
1735 if (!strstr(crc32c_impl(), "generic"))
1736 set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1737 error = btrfs_fill_super(s, fs_devices, data);
1738 }
1739 if (!error)
1740 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1741 security_free_mnt_opts(&new_sec_opts);
1742 if (error) {
1743 deactivate_locked_super(s);
1744 return ERR_PTR(error);
1745 }
1746
1747 return dget(s->s_root);
1748
1749error_close_devices:
1750 btrfs_close_devices(fs_devices);
1751error_fs_info:
1752 btrfs_free_fs_info(fs_info);
1753error_sec_opts:
1754 security_free_mnt_opts(&new_sec_opts);
1755 return ERR_PTR(error);
1756}
1757
1758/*
1759 * Mount function which is called by VFS layer.
1760 *
1761 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1762 * which needs vfsmount* of device's root (/). This means device's root has to
1763 * be mounted internally in any case.
1764 *
1765 * Operation flow:
1766 * 1. Parse subvol id related options for later use in mount_subvol().
1767 *
1768 * 2. Mount device's root (/) by calling vfs_kern_mount().
1769 *
1770 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1771 * first place. In order to avoid calling btrfs_mount() again, we use
1772 * different file_system_type which is not registered to VFS by
1773 * register_filesystem() (btrfs_root_fs_type). As a result,
1774 * btrfs_mount_root() is called. The return value will be used by
1775 * mount_subtree() in mount_subvol().
1776 *
1777 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1778 * "btrfs subvolume set-default", mount_subvol() is called always.
1779 */
1780static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1781 const char *device_name, void *data)
1782{
1783 struct vfsmount *mnt_root;
1784 struct dentry *root;
1785 char *subvol_name = NULL;
1786 u64 subvol_objectid = 0;
1787 int error = 0;
1788
1789 error = btrfs_parse_subvol_options(data, &subvol_name,
1790 &subvol_objectid);
1791 if (error) {
1792 kfree(subvol_name);
1793 return ERR_PTR(error);
1794 }
1795
1796 /* mount device's root (/) */
1797 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1798 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1799 if (flags & SB_RDONLY) {
1800 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1801 flags & ~SB_RDONLY, device_name, data);
1802 } else {
1803 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1804 flags | SB_RDONLY, device_name, data);
1805 if (IS_ERR(mnt_root)) {
1806 root = ERR_CAST(mnt_root);
1807 kfree(subvol_name);
1808 goto out;
1809 }
1810
1811 down_write(&mnt_root->mnt_sb->s_umount);
1812 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1813 up_write(&mnt_root->mnt_sb->s_umount);
1814 if (error < 0) {
1815 root = ERR_PTR(error);
1816 mntput(mnt_root);
1817 kfree(subvol_name);
1818 goto out;
1819 }
1820 }
1821 }
1822 if (IS_ERR(mnt_root)) {
1823 root = ERR_CAST(mnt_root);
1824 kfree(subvol_name);
1825 goto out;
1826 }
1827
1828 /* mount_subvol() will free subvol_name and mnt_root */
1829 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1830
1831out:
1832 return root;
1833}
1834
1835static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1836 u32 new_pool_size, u32 old_pool_size)
1837{
1838 if (new_pool_size == old_pool_size)
1839 return;
1840
1841 fs_info->thread_pool_size = new_pool_size;
1842
1843 btrfs_info(fs_info, "resize thread pool %d -> %d",
1844 old_pool_size, new_pool_size);
1845
1846 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1847 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1848 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1849 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1850 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1851 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1852 new_pool_size);
1853 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1854 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1855 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1856 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1857 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1858 new_pool_size);
1859}
1860
1861static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1862 unsigned long old_opts, int flags)
1863{
1864 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1865 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1866 (flags & SB_RDONLY))) {
1867 /* wait for any defraggers to finish */
1868 wait_event(fs_info->transaction_wait,
1869 (atomic_read(&fs_info->defrag_running) == 0));
1870 if (flags & SB_RDONLY)
1871 sync_filesystem(fs_info->sb);
1872 }
1873}
1874
1875static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1876 unsigned long old_opts)
1877{
1878 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1879
1880 /*
1881 * We need to cleanup all defragable inodes if the autodefragment is
1882 * close or the filesystem is read only.
1883 */
1884 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1885 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1886 btrfs_cleanup_defrag_inodes(fs_info);
1887 }
1888
1889 /* If we toggled discard async */
1890 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1891 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1892 btrfs_discard_resume(fs_info);
1893 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1894 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1895 btrfs_discard_cleanup(fs_info);
1896
1897 /* If we toggled space cache */
1898 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1899 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1900}
1901
1902static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1903{
1904 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1905 unsigned old_flags = sb->s_flags;
1906 unsigned long old_opts = fs_info->mount_opt;
1907 unsigned long old_compress_type = fs_info->compress_type;
1908 u64 old_max_inline = fs_info->max_inline;
1909 u32 old_thread_pool_size = fs_info->thread_pool_size;
1910 u32 old_metadata_ratio = fs_info->metadata_ratio;
1911 int ret;
1912
1913 sync_filesystem(sb);
1914 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1915
1916 if (data) {
1917 void *new_sec_opts = NULL;
1918
1919 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1920 if (!ret)
1921 ret = security_sb_remount(sb, new_sec_opts);
1922 security_free_mnt_opts(&new_sec_opts);
1923 if (ret)
1924 goto restore;
1925 }
1926
1927 ret = btrfs_parse_options(fs_info, data, *flags);
1928 if (ret)
1929 goto restore;
1930
1931 btrfs_remount_begin(fs_info, old_opts, *flags);
1932 btrfs_resize_thread_pool(fs_info,
1933 fs_info->thread_pool_size, old_thread_pool_size);
1934
1935 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1936 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1937 (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1938 btrfs_warn(fs_info,
1939 "remount supports changing free space tree only from ro to rw");
1940 /* Make sure free space cache options match the state on disk */
1941 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1942 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1943 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1944 }
1945 if (btrfs_free_space_cache_v1_active(fs_info)) {
1946 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1947 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1948 }
1949 }
1950
1951 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1952 goto out;
1953
1954 if (*flags & SB_RDONLY) {
1955 /*
1956 * this also happens on 'umount -rf' or on shutdown, when
1957 * the filesystem is busy.
1958 */
1959 cancel_work_sync(&fs_info->async_reclaim_work);
1960 cancel_work_sync(&fs_info->async_data_reclaim_work);
1961
1962 btrfs_discard_cleanup(fs_info);
1963
1964 /* wait for the uuid_scan task to finish */
1965 down(&fs_info->uuid_tree_rescan_sem);
1966 /* avoid complains from lockdep et al. */
1967 up(&fs_info->uuid_tree_rescan_sem);
1968
1969 btrfs_set_sb_rdonly(sb);
1970
1971 /*
1972 * Setting SB_RDONLY will put the cleaner thread to
1973 * sleep at the next loop if it's already active.
1974 * If it's already asleep, we'll leave unused block
1975 * groups on disk until we're mounted read-write again
1976 * unless we clean them up here.
1977 */
1978 btrfs_delete_unused_bgs(fs_info);
1979
1980 /*
1981 * The cleaner task could be already running before we set the
1982 * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
1983 * We must make sure that after we finish the remount, i.e. after
1984 * we call btrfs_commit_super(), the cleaner can no longer start
1985 * a transaction - either because it was dropping a dead root,
1986 * running delayed iputs or deleting an unused block group (the
1987 * cleaner picked a block group from the list of unused block
1988 * groups before we were able to in the previous call to
1989 * btrfs_delete_unused_bgs()).
1990 */
1991 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
1992 TASK_UNINTERRUPTIBLE);
1993
1994 /*
1995 * We've set the superblock to RO mode, so we might have made
1996 * the cleaner task sleep without running all pending delayed
1997 * iputs. Go through all the delayed iputs here, so that if an
1998 * unmount happens without remounting RW we don't end up at
1999 * finishing close_ctree() with a non-empty list of delayed
2000 * iputs.
2001 */
2002 btrfs_run_delayed_iputs(fs_info);
2003
2004 btrfs_dev_replace_suspend_for_unmount(fs_info);
2005 btrfs_scrub_cancel(fs_info);
2006 btrfs_pause_balance(fs_info);
2007
2008 /*
2009 * Pause the qgroup rescan worker if it is running. We don't want
2010 * it to be still running after we are in RO mode, as after that,
2011 * by the time we unmount, it might have left a transaction open,
2012 * so we would leak the transaction and/or crash.
2013 */
2014 btrfs_qgroup_wait_for_completion(fs_info, false);
2015
2016 ret = btrfs_commit_super(fs_info);
2017 if (ret)
2018 goto restore;
2019 } else {
2020 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2021 btrfs_err(fs_info,
2022 "Remounting read-write after error is not allowed");
2023 ret = -EINVAL;
2024 goto restore;
2025 }
2026 if (fs_info->fs_devices->rw_devices == 0) {
2027 ret = -EACCES;
2028 goto restore;
2029 }
2030
2031 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
2032 btrfs_warn(fs_info,
2033 "too many missing devices, writable remount is not allowed");
2034 ret = -EACCES;
2035 goto restore;
2036 }
2037
2038 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
2039 btrfs_warn(fs_info,
2040 "mount required to replay tree-log, cannot remount read-write");
2041 ret = -EINVAL;
2042 goto restore;
2043 }
2044 if (fs_info->sectorsize < PAGE_SIZE) {
2045 btrfs_warn(fs_info,
2046 "read-write mount is not yet allowed for sectorsize %u page size %lu",
2047 fs_info->sectorsize, PAGE_SIZE);
2048 ret = -EINVAL;
2049 goto restore;
2050 }
2051
2052 /*
2053 * NOTE: when remounting with a change that does writes, don't
2054 * put it anywhere above this point, as we are not sure to be
2055 * safe to write until we pass the above checks.
2056 */
2057 ret = btrfs_start_pre_rw_mount(fs_info);
2058 if (ret)
2059 goto restore;
2060
2061 btrfs_clear_sb_rdonly(sb);
2062
2063 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
2064 }
2065out:
2066 /*
2067 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
2068 * since the absence of the flag means it can be toggled off by remount.
2069 */
2070 *flags |= SB_I_VERSION;
2071
2072 wake_up_process(fs_info->transaction_kthread);
2073 btrfs_remount_cleanup(fs_info, old_opts);
2074 btrfs_clear_oneshot_options(fs_info);
2075 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2076
2077 return 0;
2078
2079restore:
2080 /* We've hit an error - don't reset SB_RDONLY */
2081 if (sb_rdonly(sb))
2082 old_flags |= SB_RDONLY;
2083 if (!(old_flags & SB_RDONLY))
2084 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
2085 sb->s_flags = old_flags;
2086 fs_info->mount_opt = old_opts;
2087 fs_info->compress_type = old_compress_type;
2088 fs_info->max_inline = old_max_inline;
2089 btrfs_resize_thread_pool(fs_info,
2090 old_thread_pool_size, fs_info->thread_pool_size);
2091 fs_info->metadata_ratio = old_metadata_ratio;
2092 btrfs_remount_cleanup(fs_info, old_opts);
2093 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2094
2095 return ret;
2096}
2097
2098/* Used to sort the devices by max_avail(descending sort) */
2099static inline int btrfs_cmp_device_free_bytes(const void *dev_info1,
2100 const void *dev_info2)
2101{
2102 if (((struct btrfs_device_info *)dev_info1)->max_avail >
2103 ((struct btrfs_device_info *)dev_info2)->max_avail)
2104 return -1;
2105 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
2106 ((struct btrfs_device_info *)dev_info2)->max_avail)
2107 return 1;
2108 else
2109 return 0;
2110}
2111
2112/*
2113 * sort the devices by max_avail, in which max free extent size of each device
2114 * is stored.(Descending Sort)
2115 */
2116static inline void btrfs_descending_sort_devices(
2117 struct btrfs_device_info *devices,
2118 size_t nr_devices)
2119{
2120 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2121 btrfs_cmp_device_free_bytes, NULL);
2122}
2123
2124/*
2125 * The helper to calc the free space on the devices that can be used to store
2126 * file data.
2127 */
2128static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2129 u64 *free_bytes)
2130{
2131 struct btrfs_device_info *devices_info;
2132 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2133 struct btrfs_device *device;
2134 u64 type;
2135 u64 avail_space;
2136 u64 min_stripe_size;
2137 int num_stripes = 1;
2138 int i = 0, nr_devices;
2139 const struct btrfs_raid_attr *rattr;
2140
2141 /*
2142 * We aren't under the device list lock, so this is racy-ish, but good
2143 * enough for our purposes.
2144 */
2145 nr_devices = fs_info->fs_devices->open_devices;
2146 if (!nr_devices) {
2147 smp_mb();
2148 nr_devices = fs_info->fs_devices->open_devices;
2149 ASSERT(nr_devices);
2150 if (!nr_devices) {
2151 *free_bytes = 0;
2152 return 0;
2153 }
2154 }
2155
2156 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2157 GFP_KERNEL);
2158 if (!devices_info)
2159 return -ENOMEM;
2160
2161 /* calc min stripe number for data space allocation */
2162 type = btrfs_data_alloc_profile(fs_info);
2163 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2164
2165 if (type & BTRFS_BLOCK_GROUP_RAID0)
2166 num_stripes = nr_devices;
2167 else if (type & BTRFS_BLOCK_GROUP_RAID1)
2168 num_stripes = 2;
2169 else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
2170 num_stripes = 3;
2171 else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
2172 num_stripes = 4;
2173 else if (type & BTRFS_BLOCK_GROUP_RAID10)
2174 num_stripes = 4;
2175
2176 /* Adjust for more than 1 stripe per device */
2177 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2178
2179 rcu_read_lock();
2180 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2181 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2182 &device->dev_state) ||
2183 !device->bdev ||
2184 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2185 continue;
2186
2187 if (i >= nr_devices)
2188 break;
2189
2190 avail_space = device->total_bytes - device->bytes_used;
2191
2192 /* align with stripe_len */
2193 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2194
2195 /*
2196 * In order to avoid overwriting the superblock on the drive,
2197 * btrfs starts at an offset of at least 1MB when doing chunk
2198 * allocation.
2199 *
2200 * This ensures we have at least min_stripe_size free space
2201 * after excluding 1MB.
2202 */
2203 if (avail_space <= SZ_1M + min_stripe_size)
2204 continue;
2205
2206 avail_space -= SZ_1M;
2207
2208 devices_info[i].dev = device;
2209 devices_info[i].max_avail = avail_space;
2210
2211 i++;
2212 }
2213 rcu_read_unlock();
2214
2215 nr_devices = i;
2216
2217 btrfs_descending_sort_devices(devices_info, nr_devices);
2218
2219 i = nr_devices - 1;
2220 avail_space = 0;
2221 while (nr_devices >= rattr->devs_min) {
2222 num_stripes = min(num_stripes, nr_devices);
2223
2224 if (devices_info[i].max_avail >= min_stripe_size) {
2225 int j;
2226 u64 alloc_size;
2227
2228 avail_space += devices_info[i].max_avail * num_stripes;
2229 alloc_size = devices_info[i].max_avail;
2230 for (j = i + 1 - num_stripes; j <= i; j++)
2231 devices_info[j].max_avail -= alloc_size;
2232 }
2233 i--;
2234 nr_devices--;
2235 }
2236
2237 kfree(devices_info);
2238 *free_bytes = avail_space;
2239 return 0;
2240}
2241
2242/*
2243 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2244 *
2245 * If there's a redundant raid level at DATA block groups, use the respective
2246 * multiplier to scale the sizes.
2247 *
2248 * Unused device space usage is based on simulating the chunk allocator
2249 * algorithm that respects the device sizes and order of allocations. This is
2250 * a close approximation of the actual use but there are other factors that may
2251 * change the result (like a new metadata chunk).
2252 *
2253 * If metadata is exhausted, f_bavail will be 0.
2254 */
2255static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2256{
2257 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2258 struct btrfs_super_block *disk_super = fs_info->super_copy;
2259 struct btrfs_space_info *found;
2260 u64 total_used = 0;
2261 u64 total_free_data = 0;
2262 u64 total_free_meta = 0;
2263 u32 bits = fs_info->sectorsize_bits;
2264 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2265 unsigned factor = 1;
2266 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2267 int ret;
2268 u64 thresh = 0;
2269 int mixed = 0;
2270
2271 list_for_each_entry(found, &fs_info->space_info, list) {
2272 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2273 int i;
2274
2275 total_free_data += found->disk_total - found->disk_used;
2276 total_free_data -=
2277 btrfs_account_ro_block_groups_free_space(found);
2278
2279 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2280 if (!list_empty(&found->block_groups[i]))
2281 factor = btrfs_bg_type_to_factor(
2282 btrfs_raid_array[i].bg_flag);
2283 }
2284 }
2285
2286 /*
2287 * Metadata in mixed block goup profiles are accounted in data
2288 */
2289 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2290 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2291 mixed = 1;
2292 else
2293 total_free_meta += found->disk_total -
2294 found->disk_used;
2295 }
2296
2297 total_used += found->disk_used;
2298 }
2299
2300 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2301 buf->f_blocks >>= bits;
2302 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2303
2304 /* Account global block reserve as used, it's in logical size already */
2305 spin_lock(&block_rsv->lock);
2306 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2307 if (buf->f_bfree >= block_rsv->size >> bits)
2308 buf->f_bfree -= block_rsv->size >> bits;
2309 else
2310 buf->f_bfree = 0;
2311 spin_unlock(&block_rsv->lock);
2312
2313 buf->f_bavail = div_u64(total_free_data, factor);
2314 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2315 if (ret)
2316 return ret;
2317 buf->f_bavail += div_u64(total_free_data, factor);
2318 buf->f_bavail = buf->f_bavail >> bits;
2319
2320 /*
2321 * We calculate the remaining metadata space minus global reserve. If
2322 * this is (supposedly) smaller than zero, there's no space. But this
2323 * does not hold in practice, the exhausted state happens where's still
2324 * some positive delta. So we apply some guesswork and compare the
2325 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2326 *
2327 * We probably cannot calculate the exact threshold value because this
2328 * depends on the internal reservations requested by various
2329 * operations, so some operations that consume a few metadata will
2330 * succeed even if the Avail is zero. But this is better than the other
2331 * way around.
2332 */
2333 thresh = SZ_4M;
2334
2335 /*
2336 * We only want to claim there's no available space if we can no longer
2337 * allocate chunks for our metadata profile and our global reserve will
2338 * not fit in the free metadata space. If we aren't ->full then we
2339 * still can allocate chunks and thus are fine using the currently
2340 * calculated f_bavail.
2341 */
2342 if (!mixed && block_rsv->space_info->full &&
2343 total_free_meta - thresh < block_rsv->size)
2344 buf->f_bavail = 0;
2345
2346 buf->f_type = BTRFS_SUPER_MAGIC;
2347 buf->f_bsize = dentry->d_sb->s_blocksize;
2348 buf->f_namelen = BTRFS_NAME_LEN;
2349
2350 /* We treat it as constant endianness (it doesn't matter _which_)
2351 because we want the fsid to come out the same whether mounted
2352 on a big-endian or little-endian host */
2353 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2354 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2355 /* Mask in the root object ID too, to disambiguate subvols */
2356 buf->f_fsid.val[0] ^=
2357 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2358 buf->f_fsid.val[1] ^=
2359 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2360
2361 return 0;
2362}
2363
2364static void btrfs_kill_super(struct super_block *sb)
2365{
2366 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2367 kill_anon_super(sb);
2368 btrfs_free_fs_info(fs_info);
2369}
2370
2371static struct file_system_type btrfs_fs_type = {
2372 .owner = THIS_MODULE,
2373 .name = "btrfs",
2374 .mount = btrfs_mount,
2375 .kill_sb = btrfs_kill_super,
2376 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2377};
2378
2379static struct file_system_type btrfs_root_fs_type = {
2380 .owner = THIS_MODULE,
2381 .name = "btrfs",
2382 .mount = btrfs_mount_root,
2383 .kill_sb = btrfs_kill_super,
2384 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2385};
2386
2387MODULE_ALIAS_FS("btrfs");
2388
2389static int btrfs_control_open(struct inode *inode, struct file *file)
2390{
2391 /*
2392 * The control file's private_data is used to hold the
2393 * transaction when it is started and is used to keep
2394 * track of whether a transaction is already in progress.
2395 */
2396 file->private_data = NULL;
2397 return 0;
2398}
2399
2400/*
2401 * Used by /dev/btrfs-control for devices ioctls.
2402 */
2403static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2404 unsigned long arg)
2405{
2406 struct btrfs_ioctl_vol_args *vol;
2407 struct btrfs_device *device = NULL;
2408 int ret = -ENOTTY;
2409
2410 if (!capable(CAP_SYS_ADMIN))
2411 return -EPERM;
2412
2413 vol = memdup_user((void __user *)arg, sizeof(*vol));
2414 if (IS_ERR(vol))
2415 return PTR_ERR(vol);
2416 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2417
2418 switch (cmd) {
2419 case BTRFS_IOC_SCAN_DEV:
2420 mutex_lock(&uuid_mutex);
2421 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2422 &btrfs_root_fs_type);
2423 ret = PTR_ERR_OR_ZERO(device);
2424 mutex_unlock(&uuid_mutex);
2425 break;
2426 case BTRFS_IOC_FORGET_DEV:
2427 ret = btrfs_forget_devices(vol->name);
2428 break;
2429 case BTRFS_IOC_DEVICES_READY:
2430 mutex_lock(&uuid_mutex);
2431 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2432 &btrfs_root_fs_type);
2433 if (IS_ERR(device)) {
2434 mutex_unlock(&uuid_mutex);
2435 ret = PTR_ERR(device);
2436 break;
2437 }
2438 ret = !(device->fs_devices->num_devices ==
2439 device->fs_devices->total_devices);
2440 mutex_unlock(&uuid_mutex);
2441 break;
2442 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2443 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2444 break;
2445 }
2446
2447 kfree(vol);
2448 return ret;
2449}
2450
2451static int btrfs_freeze(struct super_block *sb)
2452{
2453 struct btrfs_trans_handle *trans;
2454 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2455 struct btrfs_root *root = fs_info->tree_root;
2456
2457 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2458 /*
2459 * We don't need a barrier here, we'll wait for any transaction that
2460 * could be in progress on other threads (and do delayed iputs that
2461 * we want to avoid on a frozen filesystem), or do the commit
2462 * ourselves.
2463 */
2464 trans = btrfs_attach_transaction_barrier(root);
2465 if (IS_ERR(trans)) {
2466 /* no transaction, don't bother */
2467 if (PTR_ERR(trans) == -ENOENT)
2468 return 0;
2469 return PTR_ERR(trans);
2470 }
2471 return btrfs_commit_transaction(trans);
2472}
2473
2474static int btrfs_unfreeze(struct super_block *sb)
2475{
2476 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2477
2478 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2479 return 0;
2480}
2481
2482static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2483{
2484 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2485 struct btrfs_device *dev, *first_dev = NULL;
2486
2487 /*
2488 * Lightweight locking of the devices. We should not need
2489 * device_list_mutex here as we only read the device data and the list
2490 * is protected by RCU. Even if a device is deleted during the list
2491 * traversals, we'll get valid data, the freeing callback will wait at
2492 * least until the rcu_read_unlock.
2493 */
2494 rcu_read_lock();
2495 list_for_each_entry_rcu(dev, &fs_info->fs_devices->devices, dev_list) {
2496 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2497 continue;
2498 if (!dev->name)
2499 continue;
2500 if (!first_dev || dev->devid < first_dev->devid)
2501 first_dev = dev;
2502 }
2503
2504 if (first_dev)
2505 seq_escape(m, rcu_str_deref(first_dev->name), " \t\n\\");
2506 else
2507 WARN_ON(1);
2508 rcu_read_unlock();
2509 return 0;
2510}
2511
2512static const struct super_operations btrfs_super_ops = {
2513 .drop_inode = btrfs_drop_inode,
2514 .evict_inode = btrfs_evict_inode,
2515 .put_super = btrfs_put_super,
2516 .sync_fs = btrfs_sync_fs,
2517 .show_options = btrfs_show_options,
2518 .show_devname = btrfs_show_devname,
2519 .alloc_inode = btrfs_alloc_inode,
2520 .destroy_inode = btrfs_destroy_inode,
2521 .free_inode = btrfs_free_inode,
2522 .statfs = btrfs_statfs,
2523 .remount_fs = btrfs_remount,
2524 .freeze_fs = btrfs_freeze,
2525 .unfreeze_fs = btrfs_unfreeze,
2526};
2527
2528static const struct file_operations btrfs_ctl_fops = {
2529 .open = btrfs_control_open,
2530 .unlocked_ioctl = btrfs_control_ioctl,
2531 .compat_ioctl = compat_ptr_ioctl,
2532 .owner = THIS_MODULE,
2533 .llseek = noop_llseek,
2534};
2535
2536static struct miscdevice btrfs_misc = {
2537 .minor = BTRFS_MINOR,
2538 .name = "btrfs-control",
2539 .fops = &btrfs_ctl_fops
2540};
2541
2542MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2543MODULE_ALIAS("devname:btrfs-control");
2544
2545static int __init btrfs_interface_init(void)
2546{
2547 return misc_register(&btrfs_misc);
2548}
2549
2550static __cold void btrfs_interface_exit(void)
2551{
2552 misc_deregister(&btrfs_misc);
2553}
2554
2555static void __init btrfs_print_mod_info(void)
2556{
2557 static const char options[] = ""
2558#ifdef CONFIG_BTRFS_DEBUG
2559 ", debug=on"
2560#endif
2561#ifdef CONFIG_BTRFS_ASSERT
2562 ", assert=on"
2563#endif
2564#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2565 ", integrity-checker=on"
2566#endif
2567#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2568 ", ref-verify=on"
2569#endif
2570#ifdef CONFIG_BLK_DEV_ZONED
2571 ", zoned=yes"
2572#else
2573 ", zoned=no"
2574#endif
2575 ;
2576 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2577}
2578
2579static int __init init_btrfs_fs(void)
2580{
2581 int err;
2582
2583 btrfs_props_init();
2584
2585 err = btrfs_init_sysfs();
2586 if (err)
2587 return err;
2588
2589 btrfs_init_compress();
2590
2591 err = btrfs_init_cachep();
2592 if (err)
2593 goto free_compress;
2594
2595 err = extent_io_init();
2596 if (err)
2597 goto free_cachep;
2598
2599 err = extent_state_cache_init();
2600 if (err)
2601 goto free_extent_io;
2602
2603 err = extent_map_init();
2604 if (err)
2605 goto free_extent_state_cache;
2606
2607 err = ordered_data_init();
2608 if (err)
2609 goto free_extent_map;
2610
2611 err = btrfs_delayed_inode_init();
2612 if (err)
2613 goto free_ordered_data;
2614
2615 err = btrfs_auto_defrag_init();
2616 if (err)
2617 goto free_delayed_inode;
2618
2619 err = btrfs_delayed_ref_init();
2620 if (err)
2621 goto free_auto_defrag;
2622
2623 err = btrfs_prelim_ref_init();
2624 if (err)
2625 goto free_delayed_ref;
2626
2627 err = btrfs_end_io_wq_init();
2628 if (err)
2629 goto free_prelim_ref;
2630
2631 err = btrfs_interface_init();
2632 if (err)
2633 goto free_end_io_wq;
2634
2635 btrfs_print_mod_info();
2636
2637 err = btrfs_run_sanity_tests();
2638 if (err)
2639 goto unregister_ioctl;
2640
2641 err = register_filesystem(&btrfs_fs_type);
2642 if (err)
2643 goto unregister_ioctl;
2644
2645 return 0;
2646
2647unregister_ioctl:
2648 btrfs_interface_exit();
2649free_end_io_wq:
2650 btrfs_end_io_wq_exit();
2651free_prelim_ref:
2652 btrfs_prelim_ref_exit();
2653free_delayed_ref:
2654 btrfs_delayed_ref_exit();
2655free_auto_defrag:
2656 btrfs_auto_defrag_exit();
2657free_delayed_inode:
2658 btrfs_delayed_inode_exit();
2659free_ordered_data:
2660 ordered_data_exit();
2661free_extent_map:
2662 extent_map_exit();
2663free_extent_state_cache:
2664 extent_state_cache_exit();
2665free_extent_io:
2666 extent_io_exit();
2667free_cachep:
2668 btrfs_destroy_cachep();
2669free_compress:
2670 btrfs_exit_compress();
2671 btrfs_exit_sysfs();
2672
2673 return err;
2674}
2675
2676static void __exit exit_btrfs_fs(void)
2677{
2678 btrfs_destroy_cachep();
2679 btrfs_delayed_ref_exit();
2680 btrfs_auto_defrag_exit();
2681 btrfs_delayed_inode_exit();
2682 btrfs_prelim_ref_exit();
2683 ordered_data_exit();
2684 extent_map_exit();
2685 extent_state_cache_exit();
2686 extent_io_exit();
2687 btrfs_interface_exit();
2688 btrfs_end_io_wq_exit();
2689 unregister_filesystem(&btrfs_fs_type);
2690 btrfs_exit_sysfs();
2691 btrfs_cleanup_fs_uuids();
2692 btrfs_exit_compress();
2693}
2694
2695late_initcall(init_btrfs_fs);
2696module_exit(exit_btrfs_fs)
2697
2698MODULE_LICENSE("GPL");
2699MODULE_SOFTDEP("pre: crc32c");
2700MODULE_SOFTDEP("pre: xxhash64");
2701MODULE_SOFTDEP("pre: sha256");
2702MODULE_SOFTDEP("pre: blake2b-256");