projects/linux·blob master · refs

2775 lines · 80.3 KiB

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/memcontrol.h>
26#include <linux/slab.h>
27#include <linux/ratelimit.h>
28#include <linux/crc32c.h>
29#include <linux/btrfs.h>
30#include <linux/security.h>
31#include <linux/fs_parser.h>
32#include "messages.h"
33#include "delayed-inode.h"
34#include "ctree.h"
35#include "disk-io.h"
36#include "transaction.h"
37#include "btrfs_inode.h"
38#include "direct-io.h"
39#include "props.h"
40#include "xattr.h"
41#include "bio.h"
42#include "export.h"
43#include "compression.h"
44#include "dev-replace.h"
45#include "free-space-cache.h"
46#include "backref.h"
47#include "space-info.h"
48#include "sysfs.h"
49#include "zoned.h"
50#include "tests/btrfs-tests.h"
51#include "block-group.h"
52#include "discard.h"
53#include "qgroup.h"
54#include "raid56.h"
55#include "fs.h"
56#include "accessors.h"
57#include "defrag.h"
58#include "dir-item.h"
59#include "ioctl.h"
60#include "scrub.h"
61#include "verity.h"
62#include "super.h"
63#include "extent-tree.h"
64#include "tree-log.h"
65#define CREATE_TRACE_POINTS
66#include <trace/events/btrfs.h>
67
68static const struct super_operations btrfs_super_ops;
69static struct file_system_type btrfs_fs_type;
70
71static void btrfs_put_super(struct super_block *sb)
72{
73 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
74
75 btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid);
76 close_ctree(fs_info);
77}
78
79/* Store the mount options related information. */
80struct btrfs_fs_context {
81 char *subvol_name;
82 u64 subvol_objectid;
83 u64 max_inline;
84 u32 commit_interval;
85 u32 metadata_ratio;
86 u32 thread_pool_size;
87 unsigned long long mount_opt;
88 unsigned long compress_type:4;
89 int compress_level;
90 int use_backup_slot;
91 refcount_t refs;
92};
93
94static void btrfs_emit_options(struct btrfs_fs_info *info,
95 struct btrfs_fs_context *old);
96
97enum {
98 Opt_acl,
99 Opt_clear_cache,
100 Opt_commit_interval,
101 Opt_compress,
102 Opt_compress_force,
103 Opt_compress_force_type,
104 Opt_compress_type,
105 Opt_degraded,
106 Opt_device,
107 Opt_fatal_errors,
108 Opt_flushoncommit,
109 Opt_max_inline,
110 Opt_barrier,
111 Opt_datacow,
112 Opt_datasum,
113 Opt_defrag,
114 Opt_discard,
115 Opt_discard_mode,
116 Opt_ratio,
117 Opt_rescan_uuid_tree,
118 Opt_skip_balance,
119 Opt_space_cache,
120 Opt_space_cache_version,
121 Opt_ssd,
122 Opt_ssd_spread,
123 Opt_subvol,
124 Opt_subvol_empty,
125 Opt_subvolid,
126 Opt_thread_pool,
127 Opt_treelog,
128 Opt_user_subvol_rm_allowed,
129 Opt_norecovery,
130
131 /* Rescue options */
132 Opt_rescue,
133
134 /* Debugging options */
135 Opt_enospc_debug,
136#ifdef CONFIG_BTRFS_DEBUG
137 Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
138 Opt_ref_verify,
139 Opt_ref_tracker,
140#endif
141 Opt_err,
142};
143
144enum {
145 Opt_fatal_errors_panic,
146 Opt_fatal_errors_bug,
147};
148
149static const struct constant_table btrfs_parameter_fatal_errors[] = {
150 { "panic", Opt_fatal_errors_panic },
151 { "bug", Opt_fatal_errors_bug },
152 {}
153};
154
155enum {
156 Opt_discard_sync,
157 Opt_discard_async,
158};
159
160static const struct constant_table btrfs_parameter_discard[] = {
161 { "sync", Opt_discard_sync },
162 { "async", Opt_discard_async },
163 {}
164};
165
166enum {
167 Opt_space_cache_v1,
168 Opt_space_cache_v2,
169};
170
171static const struct constant_table btrfs_parameter_space_cache[] = {
172 { "v1", Opt_space_cache_v1 },
173 { "v2", Opt_space_cache_v2 },
174 {}
175};
176
177enum {
178 Opt_rescue_usebackuproot,
179 Opt_rescue_usebackup_slot_0,
180 Opt_rescue_usebackup_slot_1,
181 Opt_rescue_usebackup_slot_2,
182 Opt_rescue_usebackup_slot_3,
183 Opt_rescue_nologreplay,
184 Opt_rescue_ignorebadroots,
185 Opt_rescue_ignoredatacsums,
186 Opt_rescue_ignoremetacsums,
187 Opt_rescue_ignoresuperflags,
188 Opt_rescue_parameter_all,
189};
190
191static const struct constant_table btrfs_parameter_rescue[] = {
192 { "usebackuproot", Opt_rescue_usebackuproot },
193 { "usebackuproot_0", Opt_rescue_usebackup_slot_0},
194 { "usebackuproot_1", Opt_rescue_usebackup_slot_1},
195 { "usebackuproot_2", Opt_rescue_usebackup_slot_2},
196 { "usebackuproot_3", Opt_rescue_usebackup_slot_3},
197 { "nologreplay", Opt_rescue_nologreplay },
198 { "ignorebadroots", Opt_rescue_ignorebadroots },
199 { "ibadroots", Opt_rescue_ignorebadroots },
200 { "ignoredatacsums", Opt_rescue_ignoredatacsums },
201 { "ignoremetacsums", Opt_rescue_ignoremetacsums},
202 { "ignoresuperflags", Opt_rescue_ignoresuperflags},
203 { "idatacsums", Opt_rescue_ignoredatacsums },
204 { "imetacsums", Opt_rescue_ignoremetacsums},
205 { "isuperflags", Opt_rescue_ignoresuperflags},
206 { "all", Opt_rescue_parameter_all },
207 {}
208};
209
210#ifdef CONFIG_BTRFS_DEBUG
211enum {
212 Opt_fragment_parameter_data,
213 Opt_fragment_parameter_metadata,
214 Opt_fragment_parameter_all,
215};
216
217static const struct constant_table btrfs_parameter_fragment[] = {
218 { "data", Opt_fragment_parameter_data },
219 { "metadata", Opt_fragment_parameter_metadata },
220 { "all", Opt_fragment_parameter_all },
221 {}
222};
223#endif
224
225static const struct fs_parameter_spec btrfs_fs_parameters[] = {
226 fsparam_flag_no("acl", Opt_acl),
227 fsparam_flag_no("autodefrag", Opt_defrag),
228 fsparam_flag_no("barrier", Opt_barrier),
229 fsparam_flag("clear_cache", Opt_clear_cache),
230 fsparam_u32("commit", Opt_commit_interval),
231 fsparam_flag("compress", Opt_compress),
232 fsparam_string("compress", Opt_compress_type),
233 fsparam_flag("compress-force", Opt_compress_force),
234 fsparam_string("compress-force", Opt_compress_force_type),
235 fsparam_flag_no("datacow", Opt_datacow),
236 fsparam_flag_no("datasum", Opt_datasum),
237 fsparam_flag("degraded", Opt_degraded),
238 fsparam_string("device", Opt_device),
239 fsparam_flag_no("discard", Opt_discard),
240 fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard),
241 fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors),
242 fsparam_flag_no("flushoncommit", Opt_flushoncommit),
243 fsparam_string("max_inline", Opt_max_inline),
244 fsparam_u32("metadata_ratio", Opt_ratio),
245 fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree),
246 fsparam_flag("skip_balance", Opt_skip_balance),
247 fsparam_flag_no("space_cache", Opt_space_cache),
248 fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache),
249 fsparam_flag_no("ssd", Opt_ssd),
250 fsparam_flag_no("ssd_spread", Opt_ssd_spread),
251 fsparam_string("subvol", Opt_subvol),
252 fsparam_flag("subvol=", Opt_subvol_empty),
253 fsparam_u64("subvolid", Opt_subvolid),
254 fsparam_u32("thread_pool", Opt_thread_pool),
255 fsparam_flag_no("treelog", Opt_treelog),
256 fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed),
257
258 /* Rescue options. */
259 fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue),
260 /* For compatibility only, alias for "rescue=nologreplay". */
261 fsparam_flag("norecovery", Opt_norecovery),
262
263 /* Debugging options. */
264 fsparam_flag_no("enospc_debug", Opt_enospc_debug),
265#ifdef CONFIG_BTRFS_DEBUG
266 fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment),
267 fsparam_flag("ref_tracker", Opt_ref_tracker),
268 fsparam_flag("ref_verify", Opt_ref_verify),
269#endif
270 {}
271};
272
273static bool btrfs_match_compress_type(const char *string, const char *type, bool may_have_level)
274{
275 const int len = strlen(type);
276
277 return (strncmp(string, type, len) == 0) &&
278 ((may_have_level && string[len] == ':') || string[len] == '\0');
279}
280
281static int btrfs_parse_compress(struct btrfs_fs_context *ctx,
282 const struct fs_parameter *param, int opt)
283{
284 const char *string = param->string;
285 int ret;
286
287 /*
288 * Provide the same semantics as older kernels that don't use fs
289 * context, specifying the "compress" option clears "force-compress"
290 * without the need to pass "compress-force=[no|none]" before
291 * specifying "compress".
292 */
293 if (opt != Opt_compress_force && opt != Opt_compress_force_type)
294 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
295
296 if (opt == Opt_compress || opt == Opt_compress_force) {
297 ctx->compress_type = BTRFS_COMPRESS_ZLIB;
298 ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
299 btrfs_set_opt(ctx->mount_opt, COMPRESS);
300 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
301 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
302 } else if (btrfs_match_compress_type(string, "zlib", true)) {
303 ctx->compress_type = BTRFS_COMPRESS_ZLIB;
304 ret = btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB, string + 4,
305 &ctx->compress_level);
306 if (ret < 0)
307 goto error;
308 btrfs_set_opt(ctx->mount_opt, COMPRESS);
309 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
310 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
311 } else if (btrfs_match_compress_type(string, "lzo", true)) {
312 ctx->compress_type = BTRFS_COMPRESS_LZO;
313 ret = btrfs_compress_str2level(BTRFS_COMPRESS_LZO, string + 3,
314 &ctx->compress_level);
315 if (ret < 0)
316 goto error;
317 if (string[3] == ':' && string[4])
318 btrfs_warn(NULL, "Compression level ignored for LZO");
319 btrfs_set_opt(ctx->mount_opt, COMPRESS);
320 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
321 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
322 } else if (btrfs_match_compress_type(string, "zstd", true)) {
323 ctx->compress_type = BTRFS_COMPRESS_ZSTD;
324 ret = btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD, string + 4,
325 &ctx->compress_level);
326 if (ret < 0)
327 goto error;
328 btrfs_set_opt(ctx->mount_opt, COMPRESS);
329 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
330 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
331 } else if (btrfs_match_compress_type(string, "no", false) ||
332 btrfs_match_compress_type(string, "none", false)) {
333 ctx->compress_level = 0;
334 ctx->compress_type = 0;
335 btrfs_clear_opt(ctx->mount_opt, COMPRESS);
336 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
337 } else {
338 ret = -EINVAL;
339 goto error;
340 }
341 return 0;
342error:
343 btrfs_err(NULL, "failed to parse compression option '%s'", string);
344 return ret;
345
346}
347
348static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
349{
350 struct btrfs_fs_context *ctx = fc->fs_private;
351 struct fs_parse_result result;
352 int opt;
353
354 opt = fs_parse(fc, btrfs_fs_parameters, param, &result);
355 if (opt < 0)
356 return opt;
357
358 switch (opt) {
359 case Opt_degraded:
360 btrfs_set_opt(ctx->mount_opt, DEGRADED);
361 break;
362 case Opt_subvol_empty:
363 /*
364 * This exists because we used to allow it on accident, so we're
365 * keeping it to maintain ABI. See 37becec95ac3 ("Btrfs: allow
366 * empty subvol= again").
367 */
368 break;
369 case Opt_subvol:
370 kfree(ctx->subvol_name);
371 ctx->subvol_name = kstrdup(param->string, GFP_KERNEL);
372 if (!ctx->subvol_name)
373 return -ENOMEM;
374 break;
375 case Opt_subvolid:
376 ctx->subvol_objectid = result.uint_64;
377
378 /* subvolid=0 means give me the original fs_tree. */
379 if (!ctx->subvol_objectid)
380 ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
381 break;
382 case Opt_device: {
383 struct btrfs_device *device;
384
385 mutex_lock(&uuid_mutex);
386 device = btrfs_scan_one_device(param->string, false);
387 mutex_unlock(&uuid_mutex);
388 if (IS_ERR(device))
389 return PTR_ERR(device);
390 break;
391 }
392 case Opt_datasum:
393 if (result.negated) {
394 btrfs_set_opt(ctx->mount_opt, NODATASUM);
395 } else {
396 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
397 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
398 }
399 break;
400 case Opt_datacow:
401 if (result.negated) {
402 btrfs_clear_opt(ctx->mount_opt, COMPRESS);
403 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
404 btrfs_set_opt(ctx->mount_opt, NODATACOW);
405 btrfs_set_opt(ctx->mount_opt, NODATASUM);
406 } else {
407 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
408 }
409 break;
410 case Opt_compress_force:
411 case Opt_compress_force_type:
412 btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS);
413 fallthrough;
414 case Opt_compress:
415 case Opt_compress_type:
416 if (btrfs_parse_compress(ctx, param, opt))
417 return -EINVAL;
418 break;
419 case Opt_ssd:
420 if (result.negated) {
421 btrfs_set_opt(ctx->mount_opt, NOSSD);
422 btrfs_clear_opt(ctx->mount_opt, SSD);
423 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
424 } else {
425 btrfs_set_opt(ctx->mount_opt, SSD);
426 btrfs_clear_opt(ctx->mount_opt, NOSSD);
427 }
428 break;
429 case Opt_ssd_spread:
430 if (result.negated) {
431 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
432 } else {
433 btrfs_set_opt(ctx->mount_opt, SSD);
434 btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
435 btrfs_clear_opt(ctx->mount_opt, NOSSD);
436 }
437 break;
438 case Opt_barrier:
439 if (result.negated)
440 btrfs_set_opt(ctx->mount_opt, NOBARRIER);
441 else
442 btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
443 break;
444 case Opt_thread_pool:
445 if (result.uint_32 == 0) {
446 btrfs_err(NULL, "invalid value 0 for thread_pool");
447 return -EINVAL;
448 }
449 ctx->thread_pool_size = result.uint_32;
450 break;
451 case Opt_max_inline:
452 ctx->max_inline = memparse(param->string, NULL);
453 break;
454 case Opt_acl:
455 if (result.negated) {
456 fc->sb_flags &= ~SB_POSIXACL;
457 } else {
458#ifdef CONFIG_BTRFS_FS_POSIX_ACL
459 fc->sb_flags |= SB_POSIXACL;
460#else
461 btrfs_err(NULL, "support for ACL not compiled in");
462 return -EINVAL;
463#endif
464 }
465 /*
466 * VFS limits the ability to toggle ACL on and off via remount,
467 * despite every file system allowing this. This seems to be
468 * an oversight since we all do, but it'll fail if we're
469 * remounting. So don't set the mask here, we'll check it in
470 * btrfs_reconfigure and do the toggling ourselves.
471 */
472 if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
473 fc->sb_flags_mask |= SB_POSIXACL;
474 break;
475 case Opt_treelog:
476 if (result.negated)
477 btrfs_set_opt(ctx->mount_opt, NOTREELOG);
478 else
479 btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
480 break;
481 case Opt_norecovery:
482 btrfs_info(NULL,
483"'norecovery' is for compatibility only, recommended to use 'rescue=nologreplay'");
484 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
485 break;
486 case Opt_flushoncommit:
487 if (result.negated)
488 btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
489 else
490 btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
491 break;
492 case Opt_ratio:
493 ctx->metadata_ratio = result.uint_32;
494 break;
495 case Opt_discard:
496 if (result.negated) {
497 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
498 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
499 btrfs_set_opt(ctx->mount_opt, NODISCARD);
500 } else {
501 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
502 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
503 }
504 break;
505 case Opt_discard_mode:
506 switch (result.uint_32) {
507 case Opt_discard_sync:
508 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
509 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
510 break;
511 case Opt_discard_async:
512 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
513 btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
514 break;
515 default:
516 btrfs_err(NULL, "unrecognized discard mode value %s",
517 param->key);
518 return -EINVAL;
519 }
520 btrfs_clear_opt(ctx->mount_opt, NODISCARD);
521 break;
522 case Opt_space_cache:
523 if (!result.negated)
524 btrfs_warn(NULL,
525 "v1 space cache is deprecated, falling back to no space cache");
526 btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
527 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
528 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
529 break;
530 case Opt_space_cache_version:
531 switch (result.uint_32) {
532 case Opt_space_cache_v1:
533 btrfs_warn(NULL,
534 "v1 space cache is deprecated, falling back to no space cache");
535 btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
536 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
537 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
538 break;
539 case Opt_space_cache_v2:
540 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
541 btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
542 break;
543 default:
544 btrfs_err(NULL, "unrecognized space_cache value %s",
545 param->key);
546 return -EINVAL;
547 }
548 break;
549 case Opt_rescan_uuid_tree:
550 btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
551 break;
552 case Opt_clear_cache:
553 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
554 break;
555 case Opt_user_subvol_rm_allowed:
556 btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
557 break;
558 case Opt_enospc_debug:
559 if (result.negated)
560 btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
561 else
562 btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
563 break;
564 case Opt_defrag:
565 if (result.negated)
566 btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
567 else
568 btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
569 break;
570 case Opt_skip_balance:
571 btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
572 break;
573 case Opt_fatal_errors:
574 switch (result.uint_32) {
575 case Opt_fatal_errors_panic:
576 btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
577 break;
578 case Opt_fatal_errors_bug:
579 btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
580 break;
581 default:
582 btrfs_err(NULL, "unrecognized fatal_errors value %s",
583 param->key);
584 return -EINVAL;
585 }
586 break;
587 case Opt_commit_interval:
588 ctx->commit_interval = result.uint_32;
589 if (ctx->commit_interval > BTRFS_WARNING_COMMIT_INTERVAL) {
590 btrfs_warn(NULL, "excessive commit interval %u, use with care",
591 ctx->commit_interval);
592 }
593 if (ctx->commit_interval == 0)
594 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
595 break;
596 case Opt_rescue:
597 switch (result.uint_32) {
598 case Opt_rescue_usebackuproot:
599 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
600 ctx->use_backup_slot = 1;
601 break;
602 case Opt_rescue_usebackup_slot_0:
603 case Opt_rescue_usebackup_slot_1:
604 case Opt_rescue_usebackup_slot_2:
605 case Opt_rescue_usebackup_slot_3:
606 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
607 ctx->use_backup_slot = result.uint_32 - Opt_rescue_usebackup_slot_0;
608 break;
609 case Opt_rescue_nologreplay:
610 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
611 break;
612 case Opt_rescue_ignorebadroots:
613 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
614 break;
615 case Opt_rescue_ignoredatacsums:
616 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
617 break;
618 case Opt_rescue_ignoremetacsums:
619 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
620 break;
621 case Opt_rescue_ignoresuperflags:
622 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
623 break;
624 case Opt_rescue_parameter_all:
625 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
626 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
627 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
628 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
629 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
630 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
631 ctx->use_backup_slot = 1;
632 break;
633 default:
634 btrfs_info(NULL, "unrecognized rescue option '%s'",
635 param->key);
636 return -EINVAL;
637 }
638 break;
639#ifdef CONFIG_BTRFS_DEBUG
640 case Opt_fragment:
641 switch (result.uint_32) {
642 case Opt_fragment_parameter_all:
643 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
644 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
645 break;
646 case Opt_fragment_parameter_metadata:
647 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
648 break;
649 case Opt_fragment_parameter_data:
650 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
651 break;
652 default:
653 btrfs_info(NULL, "unrecognized fragment option '%s'",
654 param->key);
655 return -EINVAL;
656 }
657 break;
658 case Opt_ref_verify:
659 btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
660 break;
661 case Opt_ref_tracker:
662 btrfs_set_opt(ctx->mount_opt, REF_TRACKER);
663 break;
664#endif
665 default:
666 btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
667 return -EINVAL;
668 }
669
670 return 0;
671}
672
673/*
674 * Some options only have meaning at mount time and shouldn't persist across
675 * remounts, or be displayed. Clear these at the end of mount and remount code
676 * paths.
677 */
678static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
679{
680 btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
681 btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
682}
683
684static bool check_ro_option(const struct btrfs_fs_info *fs_info,
685 unsigned long long mount_opt, unsigned long long opt,
686 const char *opt_name)
687{
688 if (mount_opt & opt) {
689 btrfs_err(fs_info, "%s must be used with ro mount option",
690 opt_name);
691 return true;
692 }
693 return false;
694}
695
696bool btrfs_check_options(const struct btrfs_fs_info *info,
697 unsigned long long *mount_opt,
698 unsigned long flags)
699{
700 bool ret = true;
701
702 if (!(flags & SB_RDONLY) &&
703 (check_ro_option(info, *mount_opt, BTRFS_MOUNT_USEBACKUPROOT, "usebackuproot") ||
704 check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
705 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
706 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums") ||
707 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREMETACSUMS, "ignoremetacsums") ||
708 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNORESUPERFLAGS, "ignoresuperflags")))
709 ret = false;
710
711 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
712 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
713 !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
714 btrfs_err(info, "cannot disable free-space-tree");
715 ret = false;
716 }
717 if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
718 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
719 btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
720 ret = false;
721 }
722
723 if (btrfs_check_mountopts_zoned(info, mount_opt))
724 ret = false;
725
726 if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
727 if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE)) {
728 btrfs_warn(info,
729"space cache v1 is being deprecated and will be removed in a future release, please use -o space_cache=v2");
730 }
731 }
732
733 return ret;
734}
735
736/*
737 * This is subtle, we only call this during open_ctree(). We need to pre-load
738 * the mount options with the on-disk settings. Before the new mount API took
739 * effect we would do this on mount and remount. With the new mount API we'll
740 * only do this on the initial mount.
741 *
742 * This isn't a change in behavior, because we're using the current state of the
743 * file system to set the current mount options. If you mounted with special
744 * options to disable these features and then remounted we wouldn't revert the
745 * settings, because mounting without these features cleared the on-disk
746 * settings, so this being called on re-mount is not needed.
747 */
748void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
749{
750 if (fs_info->sectorsize != PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) {
751 btrfs_info(fs_info,
752 "forcing free space tree for sector size %u with page size %lu",
753 fs_info->sectorsize, PAGE_SIZE);
754 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
755 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
756 }
757
758 /*
759 * At this point our mount options are populated, so we only mess with
760 * these settings if we don't have any settings already.
761 */
762 if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
763 return;
764
765 if (btrfs_is_zoned(fs_info) &&
766 btrfs_free_space_cache_v1_active(fs_info)) {
767 btrfs_info(fs_info, "zoned: clearing existing space cache");
768 btrfs_set_super_cache_generation(fs_info->super_copy, 0);
769 return;
770 }
771
772 if (btrfs_test_opt(fs_info, SPACE_CACHE))
773 return;
774
775 if (btrfs_test_opt(fs_info, NOSPACECACHE))
776 return;
777
778 /*
779 * At this point we don't have explicit options set by the user, set
780 * them ourselves based on the state of the file system.
781 */
782 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
783 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
784 else if (btrfs_free_space_cache_v1_active(fs_info))
785 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
786}
787
788static void set_device_specific_options(struct btrfs_fs_info *fs_info)
789{
790 if (!btrfs_test_opt(fs_info, NOSSD) &&
791 !fs_info->fs_devices->rotating)
792 btrfs_set_opt(fs_info->mount_opt, SSD);
793
794 /*
795 * For devices supporting discard turn on discard=async automatically,
796 * unless it's already set or disabled. This could be turned off by
797 * nodiscard for the same mount.
798 *
799 * The zoned mode piggy backs on the discard functionality for
800 * resetting a zone. There is no reason to delay the zone reset as it is
801 * fast enough. So, do not enable async discard for zoned mode.
802 */
803 if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
804 btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
805 btrfs_test_opt(fs_info, NODISCARD)) &&
806 fs_info->fs_devices->discardable &&
807 !btrfs_is_zoned(fs_info))
808 btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
809}
810
811char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
812 u64 subvol_objectid)
813{
814 struct btrfs_root *root = fs_info->tree_root;
815 struct btrfs_root *fs_root = NULL;
816 struct btrfs_root_ref *root_ref;
817 struct btrfs_inode_ref *inode_ref;
818 struct btrfs_key key;
819 BTRFS_PATH_AUTO_FREE(path);
820 char *name = NULL, *ptr;
821 u64 dirid;
822 int len;
823 int ret;
824
825 path = btrfs_alloc_path();
826 if (!path)
827 return ERR_PTR(-ENOMEM);
828
829 name = kmalloc(PATH_MAX, GFP_KERNEL);
830 if (!name) {
831 ret = -ENOMEM;
832 goto err;
833 }
834 ptr = name + PATH_MAX - 1;
835 ptr[0] = '\0';
836
837 /*
838 * Walk up the subvolume trees in the tree of tree roots by root
839 * backrefs until we hit the top-level subvolume.
840 */
841 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
842 key.objectid = subvol_objectid;
843 key.type = BTRFS_ROOT_BACKREF_KEY;
844 key.offset = (u64)-1;
845
846 ret = btrfs_search_backwards(root, &key, path);
847 if (ret < 0) {
848 goto err;
849 } else if (ret > 0) {
850 ret = -ENOENT;
851 goto err;
852 }
853
854 subvol_objectid = key.offset;
855
856 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
857 struct btrfs_root_ref);
858 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
859 ptr -= len + 1;
860 if (ptr < name) {
861 ret = -ENAMETOOLONG;
862 goto err;
863 }
864 read_extent_buffer(path->nodes[0], ptr + 1,
865 (unsigned long)(root_ref + 1), len);
866 ptr[0] = '/';
867 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
868 btrfs_release_path(path);
869
870 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
871 if (IS_ERR(fs_root)) {
872 ret = PTR_ERR(fs_root);
873 fs_root = NULL;
874 goto err;
875 }
876
877 /*
878 * Walk up the filesystem tree by inode refs until we hit the
879 * root directory.
880 */
881 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
882 key.objectid = dirid;
883 key.type = BTRFS_INODE_REF_KEY;
884 key.offset = (u64)-1;
885
886 ret = btrfs_search_backwards(fs_root, &key, path);
887 if (ret < 0) {
888 goto err;
889 } else if (ret > 0) {
890 ret = -ENOENT;
891 goto err;
892 }
893
894 dirid = key.offset;
895
896 inode_ref = btrfs_item_ptr(path->nodes[0],
897 path->slots[0],
898 struct btrfs_inode_ref);
899 len = btrfs_inode_ref_name_len(path->nodes[0],
900 inode_ref);
901 ptr -= len + 1;
902 if (ptr < name) {
903 ret = -ENAMETOOLONG;
904 goto err;
905 }
906 read_extent_buffer(path->nodes[0], ptr + 1,
907 (unsigned long)(inode_ref + 1), len);
908 ptr[0] = '/';
909 btrfs_release_path(path);
910 }
911 btrfs_put_root(fs_root);
912 fs_root = NULL;
913 }
914
915 if (ptr == name + PATH_MAX - 1) {
916 name[0] = '/';
917 name[1] = '\0';
918 } else {
919 memmove(name, ptr, name + PATH_MAX - ptr);
920 }
921 return name;
922
923err:
924 btrfs_put_root(fs_root);
925 kfree(name);
926 return ERR_PTR(ret);
927}
928
929static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
930{
931 struct btrfs_root *root = fs_info->tree_root;
932 struct btrfs_dir_item *di;
933 BTRFS_PATH_AUTO_FREE(path);
934 struct btrfs_key location;
935 struct fscrypt_str name = FSTR_INIT("default", 7);
936 u64 dir_id;
937
938 path = btrfs_alloc_path();
939 if (!path)
940 return -ENOMEM;
941
942 /*
943 * Find the "default" dir item which points to the root item that we
944 * will mount by default if we haven't been given a specific subvolume
945 * to mount.
946 */
947 dir_id = btrfs_super_root_dir(fs_info->super_copy);
948 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
949 if (IS_ERR(di)) {
950 return PTR_ERR(di);
951 }
952 if (!di) {
953 /*
954 * Ok the default dir item isn't there. This is weird since
955 * it's always been there, but don't freak out, just try and
956 * mount the top-level subvolume.
957 */
958 *objectid = BTRFS_FS_TREE_OBJECTID;
959 return 0;
960 }
961
962 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
963 *objectid = location.objectid;
964 return 0;
965}
966
967static int btrfs_fill_super(struct super_block *sb,
968 struct btrfs_fs_devices *fs_devices)
969{
970 struct btrfs_inode *inode;
971 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
972 int ret;
973
974 sb->s_maxbytes = MAX_LFS_FILESIZE;
975 sb->s_magic = BTRFS_SUPER_MAGIC;
976 sb->s_op = &btrfs_super_ops;
977 set_default_d_op(sb, &btrfs_dentry_operations);
978 sb->s_export_op = &btrfs_export_ops;
979#ifdef CONFIG_FS_VERITY
980 sb->s_vop = &btrfs_verityops;
981#endif
982 sb->s_xattr = btrfs_xattr_handlers;
983 sb->s_time_gran = 1;
984 sb->s_iflags |= SB_I_CGROUPWB | SB_I_ALLOW_HSM;
985
986 ret = super_setup_bdi(sb);
987 if (ret) {
988 btrfs_err(fs_info, "super_setup_bdi failed");
989 return ret;
990 }
991
992 ret = open_ctree(sb, fs_devices);
993 if (ret) {
994 btrfs_err(fs_info, "open_ctree failed: %pe", ERR_PTR(ret));
995 return ret;
996 }
997
998 btrfs_emit_options(fs_info, NULL);
999
1000 inode = btrfs_iget(BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1001 if (IS_ERR(inode)) {
1002 ret = PTR_ERR(inode);
1003 btrfs_handle_fs_error(fs_info, ret, NULL);
1004 goto fail_close;
1005 }
1006
1007 sb->s_root = d_make_root(&inode->vfs_inode);
1008 if (!sb->s_root) {
1009 ret = -ENOMEM;
1010 goto fail_close;
1011 }
1012
1013 sb->s_flags |= SB_ACTIVE;
1014 return 0;
1015
1016fail_close:
1017 close_ctree(fs_info);
1018 return ret;
1019}
1020
1021int btrfs_sync_fs(struct super_block *sb, int wait)
1022{
1023 struct btrfs_trans_handle *trans;
1024 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1025 struct btrfs_root *root = fs_info->tree_root;
1026
1027 trace_btrfs_sync_fs(fs_info, wait);
1028
1029 if (!wait) {
1030 filemap_flush(fs_info->btree_inode->i_mapping);
1031 return 0;
1032 }
1033
1034 btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL);
1035
1036 trans = btrfs_attach_transaction_barrier(root);
1037 if (IS_ERR(trans)) {
1038 /* no transaction, don't bother */
1039 if (PTR_ERR(trans) == -ENOENT) {
1040 /*
1041 * Exit unless we have some pending changes
1042 * that need to go through commit
1043 */
1044 if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1045 &fs_info->flags))
1046 return 0;
1047 /*
1048 * A non-blocking test if the fs is frozen. We must not
1049 * start a new transaction here otherwise a deadlock
1050 * happens. The pending operations are delayed to the
1051 * next commit after thawing.
1052 */
1053 if (sb_start_write_trylock(sb))
1054 sb_end_write(sb);
1055 else
1056 return 0;
1057 trans = btrfs_start_transaction(root, 0);
1058 }
1059 if (IS_ERR(trans))
1060 return PTR_ERR(trans);
1061 }
1062 return btrfs_commit_transaction(trans);
1063}
1064
1065static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1066{
1067 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1068 *printed = true;
1069}
1070
1071static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1072{
1073 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1074 const char *compress_type;
1075 const char *subvol_name;
1076 bool printed = false;
1077
1078 if (btrfs_test_opt(info, DEGRADED))
1079 seq_puts(seq, ",degraded");
1080 if (btrfs_test_opt(info, NODATASUM))
1081 seq_puts(seq, ",nodatasum");
1082 if (btrfs_test_opt(info, NODATACOW))
1083 seq_puts(seq, ",nodatacow");
1084 if (btrfs_test_opt(info, NOBARRIER))
1085 seq_puts(seq, ",nobarrier");
1086 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1087 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1088 if (info->thread_pool_size != min_t(unsigned long,
1089 num_online_cpus() + 2, 8))
1090 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1091 if (btrfs_test_opt(info, COMPRESS)) {
1092 compress_type = btrfs_compress_type2str(info->compress_type);
1093 if (btrfs_test_opt(info, FORCE_COMPRESS))
1094 seq_printf(seq, ",compress-force=%s", compress_type);
1095 else
1096 seq_printf(seq, ",compress=%s", compress_type);
1097 if (info->compress_level && info->compress_type != BTRFS_COMPRESS_LZO)
1098 seq_printf(seq, ":%d", info->compress_level);
1099 }
1100 if (btrfs_test_opt(info, NOSSD))
1101 seq_puts(seq, ",nossd");
1102 if (btrfs_test_opt(info, SSD_SPREAD))
1103 seq_puts(seq, ",ssd_spread");
1104 else if (btrfs_test_opt(info, SSD))
1105 seq_puts(seq, ",ssd");
1106 if (btrfs_test_opt(info, NOTREELOG))
1107 seq_puts(seq, ",notreelog");
1108 if (btrfs_test_opt(info, NOLOGREPLAY))
1109 print_rescue_option(seq, "nologreplay", &printed);
1110 if (btrfs_test_opt(info, USEBACKUPROOT)) {
1111 seq_printf(seq, "%susebackuproot_%d", printed ? ":" : ",rescue=",
1112 info->use_backup_slot);
1113 printed = true;
1114 }
1115 if (btrfs_test_opt(info, IGNOREBADROOTS))
1116 print_rescue_option(seq, "ignorebadroots", &printed);
1117 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1118 print_rescue_option(seq, "ignoredatacsums", &printed);
1119 if (btrfs_test_opt(info, IGNOREMETACSUMS))
1120 print_rescue_option(seq, "ignoremetacsums", &printed);
1121 if (btrfs_test_opt(info, IGNORESUPERFLAGS))
1122 print_rescue_option(seq, "ignoresuperflags", &printed);
1123 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1124 seq_puts(seq, ",flushoncommit");
1125 if (btrfs_test_opt(info, DISCARD_SYNC))
1126 seq_puts(seq, ",discard");
1127 if (btrfs_test_opt(info, DISCARD_ASYNC))
1128 seq_puts(seq, ",discard=async");
1129 if (!(info->sb->s_flags & SB_POSIXACL))
1130 seq_puts(seq, ",noacl");
1131 if (btrfs_free_space_cache_v1_active(info))
1132 seq_puts(seq, ",space_cache");
1133 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1134 seq_puts(seq, ",space_cache=v2");
1135 else
1136 seq_puts(seq, ",nospace_cache");
1137 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1138 seq_puts(seq, ",rescan_uuid_tree");
1139 if (btrfs_test_opt(info, CLEAR_CACHE))
1140 seq_puts(seq, ",clear_cache");
1141 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1142 seq_puts(seq, ",user_subvol_rm_allowed");
1143 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1144 seq_puts(seq, ",enospc_debug");
1145 if (btrfs_test_opt(info, AUTO_DEFRAG))
1146 seq_puts(seq, ",autodefrag");
1147 if (btrfs_test_opt(info, SKIP_BALANCE))
1148 seq_puts(seq, ",skip_balance");
1149 if (info->metadata_ratio)
1150 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1151 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1152 seq_puts(seq, ",fatal_errors=panic");
1153 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1154 seq_printf(seq, ",commit=%u", info->commit_interval);
1155#ifdef CONFIG_BTRFS_DEBUG
1156 if (btrfs_test_opt(info, FRAGMENT_DATA))
1157 seq_puts(seq, ",fragment=data");
1158 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1159 seq_puts(seq, ",fragment=metadata");
1160#endif
1161 if (btrfs_test_opt(info, REF_VERIFY))
1162 seq_puts(seq, ",ref_verify");
1163 if (btrfs_test_opt(info, REF_TRACKER))
1164 seq_puts(seq, ",ref_tracker");
1165 seq_printf(seq, ",subvolid=%llu", btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1166 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1167 btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1168 if (!IS_ERR(subvol_name)) {
1169 seq_show_option(seq, "subvol", subvol_name);
1170 kfree(subvol_name);
1171 }
1172 return 0;
1173}
1174
1175/*
1176 * subvolumes are identified by ino 256
1177 */
1178static inline bool is_subvolume_inode(struct inode *inode)
1179{
1180 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1181 return true;
1182 return false;
1183}
1184
1185static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1186 struct vfsmount *mnt)
1187{
1188 struct dentry *root;
1189 int ret;
1190
1191 if (!subvol_name) {
1192 if (!subvol_objectid) {
1193 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1194 &subvol_objectid);
1195 if (ret) {
1196 root = ERR_PTR(ret);
1197 goto out;
1198 }
1199 }
1200 subvol_name = btrfs_get_subvol_name_from_objectid(
1201 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1202 if (IS_ERR(subvol_name)) {
1203 root = ERR_CAST(subvol_name);
1204 subvol_name = NULL;
1205 goto out;
1206 }
1207
1208 }
1209
1210 root = mount_subtree(mnt, subvol_name);
1211 /* mount_subtree() drops our reference on the vfsmount. */
1212 mnt = NULL;
1213
1214 if (!IS_ERR(root)) {
1215 struct super_block *s = root->d_sb;
1216 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1217 struct inode *root_inode = d_inode(root);
1218 u64 root_objectid = btrfs_root_id(BTRFS_I(root_inode)->root);
1219
1220 ret = 0;
1221 if (!is_subvolume_inode(root_inode)) {
1222 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1223 subvol_name);
1224 ret = -EINVAL;
1225 }
1226 if (subvol_objectid && root_objectid != subvol_objectid) {
1227 /*
1228 * This will also catch a race condition where a
1229 * subvolume which was passed by ID is renamed and
1230 * another subvolume is renamed over the old location.
1231 */
1232 btrfs_err(fs_info,
1233 "subvol '%s' does not match subvolid %llu",
1234 subvol_name, subvol_objectid);
1235 ret = -EINVAL;
1236 }
1237 if (ret) {
1238 dput(root);
1239 root = ERR_PTR(ret);
1240 deactivate_locked_super(s);
1241 }
1242 }
1243
1244out:
1245 mntput(mnt);
1246 kfree(subvol_name);
1247 return root;
1248}
1249
1250static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1251 u32 new_pool_size, u32 old_pool_size)
1252{
1253 if (new_pool_size == old_pool_size)
1254 return;
1255
1256 fs_info->thread_pool_size = new_pool_size;
1257
1258 btrfs_info(fs_info, "resize thread pool %d -> %d",
1259 old_pool_size, new_pool_size);
1260
1261 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1262 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1263 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1264 workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1265 workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1266 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1267 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1268 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1269}
1270
1271static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1272 unsigned long long old_opts, int flags)
1273{
1274 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1275 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1276 (flags & SB_RDONLY))) {
1277 /* wait for any defraggers to finish */
1278 wait_event(fs_info->transaction_wait,
1279 (atomic_read(&fs_info->defrag_running) == 0));
1280 if (flags & SB_RDONLY)
1281 sync_filesystem(fs_info->sb);
1282 }
1283}
1284
1285static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1286 unsigned long long old_opts)
1287{
1288 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1289
1290 /*
1291 * We need to cleanup all defraggable inodes if the autodefragment is
1292 * close or the filesystem is read only.
1293 */
1294 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1295 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1296 btrfs_cleanup_defrag_inodes(fs_info);
1297 }
1298
1299 /* If we toggled discard async */
1300 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1301 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1302 btrfs_discard_resume(fs_info);
1303 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1304 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1305 btrfs_discard_cleanup(fs_info);
1306
1307 /* If we toggled space cache */
1308 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1309 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1310}
1311
1312static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1313{
1314 int ret;
1315
1316 if (unlikely(BTRFS_FS_ERROR(fs_info))) {
1317 btrfs_err(fs_info,
1318 "remounting read-write after error is not allowed");
1319 return -EINVAL;
1320 }
1321
1322 if (fs_info->fs_devices->rw_devices == 0)
1323 return -EACCES;
1324
1325 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1326 btrfs_warn(fs_info,
1327 "too many missing devices, writable remount is not allowed");
1328 return -EACCES;
1329 }
1330
1331 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1332 btrfs_warn(fs_info,
1333 "mount required to replay tree-log, cannot remount read-write");
1334 return -EINVAL;
1335 }
1336
1337 /*
1338 * NOTE: when remounting with a change that does writes, don't put it
1339 * anywhere above this point, as we are not sure to be safe to write
1340 * until we pass the above checks.
1341 */
1342 ret = btrfs_start_pre_rw_mount(fs_info);
1343 if (ret)
1344 return ret;
1345
1346 btrfs_clear_sb_rdonly(fs_info->sb);
1347
1348 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1349
1350 /*
1351 * If we've gone from readonly -> read-write, we need to get our
1352 * sync/async discard lists in the right state.
1353 */
1354 btrfs_discard_resume(fs_info);
1355
1356 return 0;
1357}
1358
1359static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1360{
1361 /*
1362 * This also happens on 'umount -rf' or on shutdown, when the
1363 * filesystem is busy.
1364 */
1365 cancel_work_sync(&fs_info->async_reclaim_work);
1366 cancel_work_sync(&fs_info->async_data_reclaim_work);
1367
1368 btrfs_discard_cleanup(fs_info);
1369
1370 /* Wait for the uuid_scan task to finish */
1371 down(&fs_info->uuid_tree_rescan_sem);
1372 /* Avoid complains from lockdep et al. */
1373 up(&fs_info->uuid_tree_rescan_sem);
1374
1375 btrfs_set_sb_rdonly(fs_info->sb);
1376
1377 /*
1378 * Setting SB_RDONLY will put the cleaner thread to sleep at the next
1379 * loop if it's already active. If it's already asleep, we'll leave
1380 * unused block groups on disk until we're mounted read-write again
1381 * unless we clean them up here.
1382 */
1383 btrfs_delete_unused_bgs(fs_info);
1384
1385 /*
1386 * The cleaner task could be already running before we set the flag
1387 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock). We must make
1388 * sure that after we finish the remount, i.e. after we call
1389 * btrfs_commit_super(), the cleaner can no longer start a transaction
1390 * - either because it was dropping a dead root, running delayed iputs
1391 * or deleting an unused block group (the cleaner picked a block
1392 * group from the list of unused block groups before we were able to
1393 * in the previous call to btrfs_delete_unused_bgs()).
1394 */
1395 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1396
1397 /*
1398 * We've set the superblock to RO mode, so we might have made the
1399 * cleaner task sleep without running all pending delayed iputs. Go
1400 * through all the delayed iputs here, so that if an unmount happens
1401 * without remounting RW we don't end up at finishing close_ctree()
1402 * with a non-empty list of delayed iputs.
1403 */
1404 btrfs_run_delayed_iputs(fs_info);
1405
1406 btrfs_dev_replace_suspend_for_unmount(fs_info);
1407 btrfs_scrub_cancel(fs_info);
1408 btrfs_pause_balance(fs_info);
1409
1410 /*
1411 * Pause the qgroup rescan worker if it is running. We don't want it to
1412 * be still running after we are in RO mode, as after that, by the time
1413 * we unmount, it might have left a transaction open, so we would leak
1414 * the transaction and/or crash.
1415 */
1416 btrfs_qgroup_wait_for_completion(fs_info, false);
1417
1418 return btrfs_commit_super(fs_info);
1419}
1420
1421static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1422{
1423 fs_info->max_inline = ctx->max_inline;
1424 fs_info->commit_interval = ctx->commit_interval;
1425 fs_info->metadata_ratio = ctx->metadata_ratio;
1426 fs_info->thread_pool_size = ctx->thread_pool_size;
1427 fs_info->mount_opt = ctx->mount_opt;
1428 fs_info->compress_type = ctx->compress_type;
1429 fs_info->compress_level = ctx->compress_level;
1430 fs_info->use_backup_slot = ctx->use_backup_slot;
1431}
1432
1433static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1434{
1435 ctx->max_inline = fs_info->max_inline;
1436 ctx->commit_interval = fs_info->commit_interval;
1437 ctx->metadata_ratio = fs_info->metadata_ratio;
1438 ctx->thread_pool_size = fs_info->thread_pool_size;
1439 ctx->mount_opt = fs_info->mount_opt;
1440 ctx->compress_type = fs_info->compress_type;
1441 ctx->compress_level = fs_info->compress_level;
1442 ctx->use_backup_slot = fs_info->use_backup_slot;
1443}
1444
1445#define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...) \
1446do { \
1447 if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1448 btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1449 btrfs_info(fs_info, fmt, ##args); \
1450} while (0)
1451
1452#define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...) \
1453do { \
1454 if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1455 !btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1456 btrfs_info(fs_info, fmt, ##args); \
1457} while (0)
1458
1459static void btrfs_emit_options(struct btrfs_fs_info *info,
1460 struct btrfs_fs_context *old)
1461{
1462 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1463 btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1464 btrfs_info_if_set(info, old, NODATACOW, "setting nodatacow");
1465 btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1466 btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1467 btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1468 btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1469 btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1470 btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1471 btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1472 btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1473 btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1474 btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1475 btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1476 btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1477 btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1478 btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1479 btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1480 btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1481 btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1482 btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
1483 btrfs_info_if_set(info, old, IGNOREMETACSUMS, "ignoring meta csums");
1484 btrfs_info_if_set(info, old, IGNORESUPERFLAGS, "ignoring unknown super block flags");
1485
1486 btrfs_info_if_unset(info, old, NODATASUM, "setting datasum");
1487 btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1488 btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1489 btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1490 btrfs_info_if_unset(info, old, NOBARRIER, "turning on barriers");
1491 btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1492 btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1493 btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1494 btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1495 btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1496
1497 /* Did the compression settings change? */
1498 if (btrfs_test_opt(info, COMPRESS) &&
1499 (!old ||
1500 old->compress_type != info->compress_type ||
1501 old->compress_level != info->compress_level ||
1502 (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1503 btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1504 const char *compress_type = btrfs_compress_type2str(info->compress_type);
1505
1506 btrfs_info(info, "%s %s compression, level %d",
1507 btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1508 compress_type, info->compress_level);
1509 }
1510
1511 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1512 btrfs_info(info, "max_inline set to %llu", info->max_inline);
1513}
1514
1515static int btrfs_reconfigure(struct fs_context *fc)
1516{
1517 struct super_block *sb = fc->root->d_sb;
1518 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1519 struct btrfs_fs_context *ctx = fc->fs_private;
1520 struct btrfs_fs_context old_ctx;
1521 int ret = 0;
1522 bool mount_reconfigure = (fc->s_fs_info != NULL);
1523
1524 btrfs_info_to_ctx(fs_info, &old_ctx);
1525
1526 /*
1527 * This is our "bind mount" trick, we don't want to allow the user to do
1528 * anything other than mount a different ro/rw and a different subvol,
1529 * all of the mount options should be maintained.
1530 */
1531 if (mount_reconfigure)
1532 ctx->mount_opt = old_ctx.mount_opt;
1533
1534 sync_filesystem(sb);
1535 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1536
1537 if (!btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags)) {
1538 ret = -EINVAL;
1539 goto restore;
1540 }
1541
1542 ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
1543 if (ret < 0)
1544 goto restore;
1545
1546 btrfs_ctx_to_info(fs_info, ctx);
1547 btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags);
1548 btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size,
1549 old_ctx.thread_pool_size);
1550
1551 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1552 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1553 (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1554 btrfs_warn(fs_info,
1555 "remount supports changing free space tree only from RO to RW");
1556 /* Make sure free space cache options match the state on disk. */
1557 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1558 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1559 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1560 }
1561 if (btrfs_free_space_cache_v1_active(fs_info)) {
1562 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1563 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1564 }
1565 }
1566
1567 ret = 0;
1568 if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1569 ret = btrfs_remount_ro(fs_info);
1570 else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1571 ret = btrfs_remount_rw(fs_info);
1572 if (ret)
1573 goto restore;
1574
1575 /*
1576 * If we set the mask during the parameter parsing VFS would reject the
1577 * remount. Here we can set the mask and the value will be updated
1578 * appropriately.
1579 */
1580 if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1581 fc->sb_flags_mask |= SB_POSIXACL;
1582
1583 btrfs_emit_options(fs_info, &old_ctx);
1584 wake_up_process(fs_info->transaction_kthread);
1585 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1586 btrfs_clear_oneshot_options(fs_info);
1587 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1588
1589 return 0;
1590restore:
1591 btrfs_ctx_to_info(fs_info, &old_ctx);
1592 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1593 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1594 return ret;
1595}
1596
1597/* Used to sort the devices by max_avail(descending sort) */
1598static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1599{
1600 const struct btrfs_device_info *dev_info1 = a;
1601 const struct btrfs_device_info *dev_info2 = b;
1602
1603 if (dev_info1->max_avail > dev_info2->max_avail)
1604 return -1;
1605 else if (dev_info1->max_avail < dev_info2->max_avail)
1606 return 1;
1607 return 0;
1608}
1609
1610/*
1611 * sort the devices by max_avail, in which max free extent size of each device
1612 * is stored.(Descending Sort)
1613 */
1614static inline void btrfs_descending_sort_devices(
1615 struct btrfs_device_info *devices,
1616 size_t nr_devices)
1617{
1618 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1619 btrfs_cmp_device_free_bytes, NULL);
1620}
1621
1622/*
1623 * The helper to calc the free space on the devices that can be used to store
1624 * file data.
1625 */
1626static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1627 u64 *free_bytes)
1628{
1629 struct btrfs_device_info AUTO_KFREE(devices_info);
1630 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1631 struct btrfs_device *device;
1632 u64 type;
1633 u64 avail_space;
1634 u64 min_stripe_size;
1635 int num_stripes = 1;
1636 int i = 0, nr_devices;
1637 const struct btrfs_raid_attr *rattr;
1638
1639 /*
1640 * We aren't under the device list lock, so this is racy-ish, but good
1641 * enough for our purposes.
1642 */
1643 nr_devices = fs_info->fs_devices->open_devices;
1644 if (!nr_devices) {
1645 smp_mb();
1646 nr_devices = fs_info->fs_devices->open_devices;
1647 ASSERT(nr_devices);
1648 if (!nr_devices) {
1649 *free_bytes = 0;
1650 return 0;
1651 }
1652 }
1653
1654 devices_info = kmalloc_objs(*devices_info, nr_devices);
1655 if (!devices_info)
1656 return -ENOMEM;
1657
1658 /* calc min stripe number for data space allocation */
1659 type = btrfs_data_alloc_profile(fs_info);
1660 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1661
1662 if (type & BTRFS_BLOCK_GROUP_RAID0)
1663 num_stripes = nr_devices;
1664 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1665 num_stripes = rattr->ncopies;
1666 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1667 num_stripes = 4;
1668
1669 /* Adjust for more than 1 stripe per device */
1670 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1671
1672 rcu_read_lock();
1673 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1674 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1675 &device->dev_state) ||
1676 !device->bdev ||
1677 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1678 continue;
1679
1680 if (i >= nr_devices)
1681 break;
1682
1683 avail_space = device->total_bytes - device->bytes_used;
1684
1685 /* align with stripe_len */
1686 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1687
1688 /*
1689 * Ensure we have at least min_stripe_size on top of the
1690 * reserved space on the device.
1691 */
1692 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1693 continue;
1694
1695 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1696
1697 devices_info[i].dev = device;
1698 devices_info[i].max_avail = avail_space;
1699
1700 i++;
1701 }
1702 rcu_read_unlock();
1703
1704 nr_devices = i;
1705
1706 btrfs_descending_sort_devices(devices_info, nr_devices);
1707
1708 i = nr_devices - 1;
1709 avail_space = 0;
1710 while (nr_devices >= rattr->devs_min) {
1711 num_stripes = min(num_stripes, nr_devices);
1712
1713 if (devices_info[i].max_avail >= min_stripe_size) {
1714 int j;
1715 u64 alloc_size;
1716
1717 avail_space += devices_info[i].max_avail * num_stripes;
1718 alloc_size = devices_info[i].max_avail;
1719 for (j = i + 1 - num_stripes; j <= i; j++)
1720 devices_info[j].max_avail -= alloc_size;
1721 }
1722 i--;
1723 nr_devices--;
1724 }
1725
1726 *free_bytes = avail_space;
1727 return 0;
1728}
1729
1730/*
1731 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1732 *
1733 * If there's a redundant raid level at DATA block groups, use the respective
1734 * multiplier to scale the sizes.
1735 *
1736 * Unused device space usage is based on simulating the chunk allocator
1737 * algorithm that respects the device sizes and order of allocations. This is
1738 * a close approximation of the actual use but there are other factors that may
1739 * change the result (like a new metadata chunk).
1740 *
1741 * If metadata is exhausted, f_bavail will be 0.
1742 */
1743static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1744{
1745 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1746 struct btrfs_super_block *disk_super = fs_info->super_copy;
1747 struct btrfs_space_info *found;
1748 u64 total_used = 0;
1749 u64 total_free_data = 0;
1750 u64 total_free_meta = 0;
1751 u32 bits = fs_info->sectorsize_bits;
1752 __be32 *fsid;
1753 unsigned factor = 1;
1754 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1755 int ret;
1756 u64 thresh = 0;
1757 bool mixed = false;
1758 __kernel_fsid_t f_fsid;
1759
1760 list_for_each_entry(found, &fs_info->space_info, list) {
1761 if (found->flags & BTRFS_BLOCK_GROUP_DATA &&
1762 found->subgroup_id != BTRFS_SUB_GROUP_DATA_RELOC) {
1763 int i;
1764
1765 total_free_data += found->disk_total - found->disk_used;
1766 total_free_data -=
1767 btrfs_account_ro_block_groups_free_space(found);
1768
1769 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1770 if (!list_empty(&found->block_groups[i]))
1771 factor = btrfs_bg_type_to_factor(
1772 btrfs_raid_array[i].bg_flag);
1773 }
1774 }
1775
1776 /*
1777 * Metadata in mixed block group profiles are accounted in data
1778 */
1779 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1780 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1781 mixed = true;
1782 else
1783 total_free_meta += found->disk_total -
1784 found->disk_used;
1785 }
1786
1787 total_used += found->disk_used;
1788 }
1789
1790 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1791 buf->f_blocks >>= bits;
1792 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1793
1794 /* Account global block reserve as used, it's in logical size already */
1795 spin_lock(&block_rsv->lock);
1796 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
1797 if (buf->f_bfree >= block_rsv->size >> bits)
1798 buf->f_bfree -= block_rsv->size >> bits;
1799 else
1800 buf->f_bfree = 0;
1801 spin_unlock(&block_rsv->lock);
1802
1803 buf->f_bavail = div_u64(total_free_data, factor);
1804 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1805 if (ret)
1806 return ret;
1807 buf->f_bavail += div_u64(total_free_data, factor);
1808 buf->f_bavail = buf->f_bavail >> bits;
1809
1810 /*
1811 * We calculate the remaining metadata space minus global reserve. If
1812 * this is (supposedly) smaller than zero, there's no space. But this
1813 * does not hold in practice, the exhausted state happens where's still
1814 * some positive delta. So we apply some guesswork and compare the
1815 * delta to a 4M threshold. (Practically observed delta was ~2M.)
1816 *
1817 * We probably cannot calculate the exact threshold value because this
1818 * depends on the internal reservations requested by various
1819 * operations, so some operations that consume a few metadata will
1820 * succeed even if the Avail is zero. But this is better than the other
1821 * way around.
1822 */
1823 thresh = SZ_4M;
1824
1825 /*
1826 * We only want to claim there's no available space if we can no longer
1827 * allocate chunks for our metadata profile and our global reserve will
1828 * not fit in the free metadata space. If we aren't ->full then we
1829 * still can allocate chunks and thus are fine using the currently
1830 * calculated f_bavail.
1831 */
1832 if (!mixed && block_rsv->space_info->full &&
1833 (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1834 buf->f_bavail = 0;
1835
1836 buf->f_type = BTRFS_SUPER_MAGIC;
1837 buf->f_bsize = fs_info->sectorsize;
1838 buf->f_namelen = BTRFS_NAME_LEN;
1839
1840 /*
1841 * fs_devices->fsid is dynamically generated when temp_fsid is active
1842 * to support cloned filesystems. Use the original on-disk fsid instead,
1843 * as it remains consistent across mount cycles.
1844 */
1845 if (fs_info->fs_devices->temp_fsid)
1846 fsid = (__be32 *)fs_info->super_copy->fsid;
1847 else
1848 fsid = (__be32 *)fs_info->fs_devices->fsid;
1849
1850 /*
1851 * We treat it as constant endianness (it doesn't matter _which_)
1852 * because we want the fsid to come out the same whether mounted
1853 * on a big-endian or little-endian host.
1854 */
1855 f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1856 f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1857
1858 /* Mask in the root object ID too, to disambiguate subvols */
1859 f_fsid.val[0] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root) >> 32;
1860 f_fsid.val[1] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root);
1861
1862 /*
1863 * Hash dev_t to avoid f_fsid collisions with cloned filesystems.
1864 * Only do this when a clone is present so the original filesystem
1865 * (mounted first) maintains backward-compatible f_fsid behavior.
1866 */
1867 if (fs_info->fs_devices->temp_fsid) {
1868 __kernel_fsid_t dev_fsid =
1869 u64_to_fsid(huge_encode_dev(fs_info->fs_devices->latest_dev->bdev->bd_dev));
1870
1871 f_fsid.val[0] ^= dev_fsid.val[1];
1872 f_fsid.val[1] ^= dev_fsid.val[0];
1873 }
1874
1875 memcpy(&buf->f_fsid, &f_fsid, sizeof(f_fsid));
1876
1877 return 0;
1878}
1879
1880static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1881{
1882 struct btrfs_fs_info *p = fc->s_fs_info;
1883 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1884
1885 return fs_info->fs_devices == p->fs_devices;
1886}
1887
1888static int btrfs_get_tree_super(struct fs_context *fc)
1889{
1890 struct btrfs_fs_info *fs_info = fc->s_fs_info;
1891 struct btrfs_fs_context *ctx = fc->fs_private;
1892 struct btrfs_fs_devices *fs_devices = NULL;
1893 struct btrfs_device *device;
1894 struct super_block *sb;
1895 blk_mode_t mode = sb_open_mode(fc->sb_flags);
1896 int ret;
1897
1898 btrfs_ctx_to_info(fs_info, ctx);
1899 mutex_lock(&uuid_mutex);
1900
1901 /*
1902 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1903 * either a valid device or an error.
1904 */
1905 device = btrfs_scan_one_device(fc->source, true);
1906 ASSERT(device != NULL);
1907 if (IS_ERR(device)) {
1908 mutex_unlock(&uuid_mutex);
1909 return PTR_ERR(device);
1910 }
1911 fs_devices = device->fs_devices;
1912 /*
1913 * We cannot hold uuid_mutex calling sget_fc(), it will lead to a
1914 * locking order reversal with s_umount.
1915 *
1916 * So here we increase the holding number of fs_devices, this will ensure
1917 * the fs_devices itself won't be freed.
1918 */
1919 btrfs_fs_devices_inc_holding(fs_devices);
1920 fs_info->fs_devices = fs_devices;
1921 mutex_unlock(&uuid_mutex);
1922
1923 fc->sb_flags |= SB_NOSEC;
1924
1925 sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1926 if (IS_ERR(sb)) {
1927 mutex_lock(&uuid_mutex);
1928 btrfs_fs_devices_dec_holding(fs_devices);
1929 /*
1930 * Since the fs_devices is not opened, it can be freed at any
1931 * time after unlocking uuid_mutex. We need to avoid double
1932 * free through put_fs_context()->btrfs_free_fs_info().
1933 * So here we reset fs_info->fs_devices to NULL, and let the
1934 * regular fs_devices reclaim path to handle it.
1935 *
1936 * This applies to all later branches where no fs_devices is
1937 * opened.
1938 */
1939 fs_info->fs_devices = NULL;
1940 mutex_unlock(&uuid_mutex);
1941 return PTR_ERR(sb);
1942 }
1943
1944 if (sb->s_root) {
1945 /*
1946 * Not the first mount of the fs thus got an existing super block.
1947 * Will reuse the returned super block, fs_info and fs_devices.
1948 *
1949 * fc->s_fs_info is not touched and will be later freed by
1950 * put_fs_context() through btrfs_free_fs_context().
1951 */
1952 ASSERT(fc->s_fs_info == fs_info);
1953
1954 mutex_lock(&uuid_mutex);
1955 btrfs_fs_devices_dec_holding(fs_devices);
1956 fs_info->fs_devices = NULL;
1957 mutex_unlock(&uuid_mutex);
1958 /*
1959 * At this stage we may have RO flag mismatch between
1960 * fc->sb_flags and sb->s_flags. Caller should detect such
1961 * mismatch and reconfigure with sb->s_umount rwsem held if
1962 * needed.
1963 */
1964 } else {
1965 struct block_device *bdev;
1966
1967 /*
1968 * The first mount of the fs thus a new superblock, fc->s_fs_info
1969 * must be NULL, and the ownership of our fs_info and fs_devices is
1970 * transferred to the super block.
1971 */
1972 ASSERT(fc->s_fs_info == NULL);
1973
1974 mutex_lock(&uuid_mutex);
1975 btrfs_fs_devices_dec_holding(fs_devices);
1976 ret = btrfs_open_devices(fs_devices, mode, sb);
1977 if (ret < 0)
1978 fs_info->fs_devices = NULL;
1979 mutex_unlock(&uuid_mutex);
1980 if (ret < 0) {
1981 deactivate_locked_super(sb);
1982 return ret;
1983 }
1984 if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1985 deactivate_locked_super(sb);
1986 return -EACCES;
1987 }
1988 set_device_specific_options(fs_info);
1989 bdev = fs_devices->latest_dev->bdev;
1990 snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
1991 shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id);
1992 ret = btrfs_fill_super(sb, fs_devices);
1993 if (ret) {
1994 deactivate_locked_super(sb);
1995 return ret;
1996 }
1997 }
1998
1999 btrfs_clear_oneshot_options(fs_info);
2000
2001 fc->root = dget(sb->s_root);
2002 return 0;
2003}
2004
2005/*
2006 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
2007 * with different ro/rw options") the following works:
2008 *
2009 * (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
2010 * (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
2011 *
2012 * which looks nice and innocent but is actually pretty intricate and deserves
2013 * a long comment.
2014 *
2015 * On another filesystem a subvolume mount is close to something like:
2016 *
2017 * (iii) # create rw superblock + initial mount
2018 * mount -t xfs /dev/sdb /opt/
2019 *
2020 * # create ro bind mount
2021 * mount --bind -o ro /opt/foo /mnt/foo
2022 *
2023 * # unmount initial mount
2024 * umount /opt
2025 *
2026 * Of course, there's some special subvolume sauce and there's the fact that the
2027 * sb->s_root dentry is really swapped after mount_subtree(). But conceptually
2028 * it's very close and will help us understand the issue.
2029 *
2030 * The old mount API didn't cleanly distinguish between a mount being made ro
2031 * and a superblock being made ro. The only way to change the ro state of
2032 * either object was by passing ms_rdonly. If a new mount was created via
2033 * mount(2) such as:
2034 *
2035 * mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
2036 *
2037 * the MS_RDONLY flag being specified had two effects:
2038 *
2039 * (1) MNT_READONLY was raised -> the resulting mount got
2040 * @mnt->mnt_flags |= MNT_READONLY raised.
2041 *
2042 * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
2043 * made the superblock ro. Note, how SB_RDONLY has the same value as
2044 * ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
2045 *
2046 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
2047 * subtree mounted ro.
2048 *
2049 * But consider the effect on the old mount API on btrfs subvolume mounting
2050 * which combines the distinct step in (iii) into a single step.
2051 *
2052 * By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
2053 * is issued the superblock is ro and thus even if the mount created for (ii) is
2054 * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
2055 * to rw for (ii) which it did using an internal remount call.
2056 *
2057 * IOW, subvolume mounting was inherently complicated due to the ambiguity of
2058 * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
2059 * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
2060 * passed by mount(8) to mount(2).
2061 *
2062 * Enter the new mount API. The new mount API disambiguates making a mount ro
2063 * and making a superblock ro.
2064 *
2065 * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
2066 * fsmount() or mount_setattr() this is a pure VFS level change for a
2067 * specific mount or mount tree that is never seen by the filesystem itself.
2068 *
2069 * (4) To turn a superblock ro the "ro" flag must be used with
2070 * fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
2071 * in fc->sb_flags.
2072 *
2073 * But, currently the util-linux mount command already utilizes the new mount
2074 * API and is still setting fsconfig(FSCONFIG_SET_FLAG, "ro") no matter if it's
2075 * btrfs or not, setting the whole super block RO. To make per-subvolume mounting
2076 * work with different options work we need to keep backward compatibility.
2077 */
2078static int btrfs_reconfigure_for_mount(struct fs_context *fc)
2079{
2080 int ret = 0;
2081
2082 if (!(fc->sb_flags & SB_RDONLY) && (fc->root->d_sb->s_flags & SB_RDONLY))
2083 ret = btrfs_reconfigure(fc);
2084
2085 return ret;
2086}
2087
2088static int btrfs_get_tree_subvol(struct fs_context *fc)
2089{
2090 struct btrfs_fs_info *fs_info = NULL;
2091 struct btrfs_fs_context *ctx = fc->fs_private;
2092 struct fs_context *dup_fc;
2093 struct dentry *dentry;
2094 struct vfsmount *mnt;
2095 int ret = 0;
2096
2097 /*
2098 * Setup a dummy root and fs_info for test/set super. This is because
2099 * we don't actually fill this stuff out until open_ctree, but we need
2100 * then open_ctree will properly initialize the file system specific
2101 * settings later. btrfs_init_fs_info initializes the static elements
2102 * of the fs_info (locks and such) to make cleanup easier if we find a
2103 * superblock with our given fs_devices later on at sget_fc() time.
2104 */
2105 fs_info = kvzalloc_obj(struct btrfs_fs_info);
2106 if (!fs_info)
2107 return -ENOMEM;
2108
2109 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2110 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2111 if (!fs_info->super_copy || !fs_info->super_for_commit) {
2112 /*
2113 * Dont call btrfs_free_fs_info() to free it as it's still
2114 * initialized partially.
2115 */
2116 kfree(fs_info->super_copy);
2117 kfree(fs_info->super_for_commit);
2118 kvfree(fs_info);
2119 return -ENOMEM;
2120 }
2121 btrfs_init_fs_info(fs_info);
2122
2123 dup_fc = vfs_dup_fs_context(fc);
2124 if (IS_ERR(dup_fc)) {
2125 btrfs_free_fs_info(fs_info);
2126 return PTR_ERR(dup_fc);
2127 }
2128
2129 /*
2130 * When we do the sget_fc this gets transferred to the sb, so we only
2131 * need to set it on the dup_fc as this is what creates the super block.
2132 */
2133 dup_fc->s_fs_info = fs_info;
2134
2135 ret = btrfs_get_tree_super(dup_fc);
2136 if (ret)
2137 goto error;
2138
2139 ret = btrfs_reconfigure_for_mount(dup_fc);
2140 up_write(&dup_fc->root->d_sb->s_umount);
2141 if (ret)
2142 goto error;
2143 mnt = vfs_create_mount(dup_fc);
2144 put_fs_context(dup_fc);
2145 if (IS_ERR(mnt))
2146 return PTR_ERR(mnt);
2147
2148 /*
2149 * This free's ->subvol_name, because if it isn't set we have to
2150 * allocate a buffer to hold the subvol_name, so we just drop our
2151 * reference to it here.
2152 */
2153 dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2154 ctx->subvol_name = NULL;
2155 if (IS_ERR(dentry))
2156 return PTR_ERR(dentry);
2157
2158 fc->root = dentry;
2159 return 0;
2160error:
2161 put_fs_context(dup_fc);
2162 return ret;
2163}
2164
2165static int btrfs_get_tree(struct fs_context *fc)
2166{
2167 ASSERT(fc->s_fs_info == NULL);
2168
2169 return btrfs_get_tree_subvol(fc);
2170}
2171
2172static void btrfs_kill_super(struct super_block *sb)
2173{
2174 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2175 kill_anon_super(sb);
2176 btrfs_free_fs_info(fs_info);
2177}
2178
2179static void btrfs_free_fs_context(struct fs_context *fc)
2180{
2181 struct btrfs_fs_context *ctx = fc->fs_private;
2182 struct btrfs_fs_info *fs_info = fc->s_fs_info;
2183
2184 if (fs_info)
2185 btrfs_free_fs_info(fs_info);
2186
2187 if (ctx && refcount_dec_and_test(&ctx->refs)) {
2188 kfree(ctx->subvol_name);
2189 kfree(ctx);
2190 }
2191}
2192
2193static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2194{
2195 struct btrfs_fs_context *ctx = src_fc->fs_private;
2196
2197 /*
2198 * Give a ref to our ctx to this dup, as we want to keep it around for
2199 * our original fc so we can have the subvolume name or objectid.
2200 *
2201 * We unset ->source in the original fc because the dup needs it for
2202 * mounting, and then once we free the dup it'll free ->source, so we
2203 * need to make sure we're only pointing to it in one fc.
2204 */
2205 refcount_inc(&ctx->refs);
2206 fc->fs_private = ctx;
2207 fc->source = src_fc->source;
2208 src_fc->source = NULL;
2209 return 0;
2210}
2211
2212static const struct fs_context_operations btrfs_fs_context_ops = {
2213 .parse_param = btrfs_parse_param,
2214 .reconfigure = btrfs_reconfigure,
2215 .get_tree = btrfs_get_tree,
2216 .dup = btrfs_dup_fs_context,
2217 .free = btrfs_free_fs_context,
2218};
2219
2220static int btrfs_init_fs_context(struct fs_context *fc)
2221{
2222 struct btrfs_fs_context *ctx;
2223
2224 ctx = kzalloc_obj(struct btrfs_fs_context);
2225 if (!ctx)
2226 return -ENOMEM;
2227
2228 refcount_set(&ctx->refs, 1);
2229 fc->fs_private = ctx;
2230 fc->ops = &btrfs_fs_context_ops;
2231
2232 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2233 btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
2234 } else {
2235 ctx->use_backup_slot = -1;
2236 ctx->thread_pool_size =
2237 min_t(unsigned long, num_online_cpus() + 2, 8);
2238 ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2239 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2240 }
2241
2242#ifdef CONFIG_BTRFS_FS_POSIX_ACL
2243 fc->sb_flags |= SB_POSIXACL;
2244#endif
2245 fc->sb_flags |= SB_I_VERSION;
2246
2247 return 0;
2248}
2249
2250static struct file_system_type btrfs_fs_type = {
2251 .owner = THIS_MODULE,
2252 .name = "btrfs",
2253 .init_fs_context = btrfs_init_fs_context,
2254 .parameters = btrfs_fs_parameters,
2255 .kill_sb = btrfs_kill_super,
2256 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA |
2257 FS_ALLOW_IDMAP | FS_MGTIME,
2258 };
2259
2260MODULE_ALIAS_FS("btrfs");
2261
2262static int btrfs_control_open(struct inode *inode, struct file *file)
2263{
2264 /*
2265 * The control file's private_data is used to hold the
2266 * transaction when it is started and is used to keep
2267 * track of whether a transaction is already in progress.
2268 */
2269 file->private_data = NULL;
2270 return 0;
2271}
2272
2273/*
2274 * Used by /dev/btrfs-control for devices ioctls.
2275 */
2276static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2277 unsigned long arg)
2278{
2279 struct btrfs_ioctl_vol_args *vol;
2280 struct btrfs_device *device = NULL;
2281 dev_t devt = 0;
2282 int ret = -ENOTTY;
2283
2284 if (!capable(CAP_SYS_ADMIN))
2285 return -EPERM;
2286
2287 vol = memdup_user((void __user *)arg, sizeof(*vol));
2288 if (IS_ERR(vol))
2289 return PTR_ERR(vol);
2290 ret = btrfs_check_ioctl_vol_args_path(vol);
2291 if (ret < 0)
2292 goto out;
2293
2294 switch (cmd) {
2295 case BTRFS_IOC_SCAN_DEV:
2296 mutex_lock(&uuid_mutex);
2297 /*
2298 * Scanning outside of mount can return NULL which would turn
2299 * into 0 error code.
2300 */
2301 device = btrfs_scan_one_device(vol->name, false);
2302 ret = PTR_ERR_OR_ZERO(device);
2303 mutex_unlock(&uuid_mutex);
2304 break;
2305 case BTRFS_IOC_FORGET_DEV:
2306 if (vol->name[0] != 0) {
2307 ret = lookup_bdev(vol->name, &devt);
2308 if (ret)
2309 break;
2310 }
2311 ret = btrfs_forget_devices(devt);
2312 break;
2313 case BTRFS_IOC_DEVICES_READY:
2314 mutex_lock(&uuid_mutex);
2315 /*
2316 * Scanning outside of mount can return NULL which would turn
2317 * into 0 error code.
2318 */
2319 device = btrfs_scan_one_device(vol->name, false);
2320 if (IS_ERR_OR_NULL(device)) {
2321 mutex_unlock(&uuid_mutex);
2322 ret = PTR_ERR_OR_ZERO(device);
2323 break;
2324 }
2325 ret = !(device->fs_devices->num_devices ==
2326 device->fs_devices->total_devices);
2327 mutex_unlock(&uuid_mutex);
2328 break;
2329 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2330 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2331 break;
2332 default:
2333 ret = -ENOTTY;
2334 break;
2335 }
2336
2337out:
2338 kfree(vol);
2339 return ret;
2340}
2341
2342static int btrfs_freeze(struct super_block *sb)
2343{
2344 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2345
2346 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2347 /*
2348 * We don't need a barrier here, we'll wait for any transaction that
2349 * could be in progress on other threads (and do delayed iputs that
2350 * we want to avoid on a frozen filesystem), or do the commit
2351 * ourselves.
2352 */
2353 return btrfs_commit_current_transaction(fs_info->tree_root);
2354}
2355
2356static int check_dev_super(struct btrfs_device *dev)
2357{
2358 struct btrfs_fs_info *fs_info = dev->fs_info;
2359 struct btrfs_super_block *sb;
2360 u64 last_trans;
2361 u16 csum_type;
2362 int ret = 0;
2363
2364 /* This should be called with fs still frozen. */
2365 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2366
2367 /* Missing dev, no need to check. */
2368 if (!dev->bdev)
2369 return 0;
2370
2371 /* Only need to check the primary super block. */
2372 sb = btrfs_read_disk_super(dev->bdev, 0, true);
2373 if (IS_ERR(sb))
2374 return PTR_ERR(sb);
2375
2376 /* Verify the checksum. */
2377 csum_type = btrfs_super_csum_type(sb);
2378 if (unlikely(csum_type != btrfs_super_csum_type(fs_info->super_copy))) {
2379 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2380 csum_type, btrfs_super_csum_type(fs_info->super_copy));
2381 ret = -EUCLEAN;
2382 goto out;
2383 }
2384
2385 if (unlikely(btrfs_check_super_csum(fs_info, sb))) {
2386 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2387 ret = -EUCLEAN;
2388 goto out;
2389 }
2390
2391 /* Btrfs_validate_super() includes fsid check against super->fsid. */
2392 ret = btrfs_validate_super(fs_info, sb, 0);
2393 if (ret < 0)
2394 goto out;
2395
2396 last_trans = btrfs_get_last_trans_committed(fs_info);
2397 if (unlikely(btrfs_super_generation(sb) != last_trans)) {
2398 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2399 btrfs_super_generation(sb), last_trans);
2400 ret = -EUCLEAN;
2401 goto out;
2402 }
2403out:
2404 btrfs_release_disk_super(sb);
2405 return ret;
2406}
2407
2408static int btrfs_unfreeze(struct super_block *sb)
2409{
2410 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2411 struct btrfs_device *device;
2412 int ret = 0;
2413
2414 /*
2415 * Make sure the fs is not changed by accident (like hibernation then
2416 * modified by other OS).
2417 * If we found anything wrong, we mark the fs error immediately.
2418 *
2419 * And since the fs is frozen, no one can modify the fs yet, thus
2420 * we don't need to hold device_list_mutex.
2421 */
2422 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2423 ret = check_dev_super(device);
2424 if (ret < 0) {
2425 btrfs_handle_fs_error(fs_info, ret,
2426 "super block on devid %llu got modified unexpectedly",
2427 device->devid);
2428 break;
2429 }
2430 }
2431 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2432
2433 /*
2434 * We still return 0, to allow VFS layer to unfreeze the fs even the
2435 * above checks failed. Since the fs is either fine or read-only, we're
2436 * safe to continue, without causing further damage.
2437 */
2438 return 0;
2439}
2440
2441static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2442{
2443 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2444
2445 /*
2446 * There should be always a valid pointer in latest_dev, it may be stale
2447 * for a short moment in case it's being deleted but still valid until
2448 * the end of RCU grace period.
2449 */
2450 rcu_read_lock();
2451 seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2452 rcu_read_unlock();
2453
2454 return 0;
2455}
2456
2457static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_control *sc)
2458{
2459 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2460 const s64 nr = percpu_counter_read_positive(&fs_info->evictable_extent_maps);
2461
2462 /*
2463 * The evictable extent map counter is filesystem-global and does not
2464 * honour sc->memcg, so it is only meaningful on the global (kswapd or
2465 * root direct reclaim) shrink path. Skip the per-memcg iterations of
2466 * shrink_slab_memcg() to avoid queueing duplicate global work.
2467 */
2468 if (!mem_cgroup_shrink_is_root(sc))
2469 return 0;
2470
2471 trace_btrfs_extent_map_shrinker_count(fs_info, nr);
2472
2473 return nr;
2474}
2475
2476static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_control *sc)
2477{
2478 const long nr_to_scan = min_t(unsigned long, LONG_MAX, sc->nr_to_scan);
2479 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2480
2481 btrfs_free_extent_maps(fs_info, nr_to_scan);
2482
2483 /* The extent map shrinker runs asynchronously, so always return 0. */
2484 return 0;
2485}
2486
2487static int btrfs_remove_bdev(struct super_block *sb, struct block_device *bdev)
2488{
2489 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2490 struct btrfs_device *device;
2491 struct btrfs_dev_lookup_args lookup_args = { .devt = bdev->bd_dev };
2492 bool can_rw;
2493
2494 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2495 device = btrfs_find_device(fs_info->fs_devices, &lookup_args);
2496 if (!device) {
2497 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2498 /* Device not found, should not affect the running fs, just give a warning. */
2499 btrfs_warn(fs_info, "unable to find btrfs device for block device '%pg'", bdev);
2500 return 0;
2501 }
2502 /*
2503 * The to-be-removed device is already missing?
2504 *
2505 * That's weird but no special handling needed and can exit right now.
2506 */
2507 if (unlikely(test_and_set_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))) {
2508 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2509 btrfs_warn(fs_info, "btrfs device id %llu is already missing", device->devid);
2510 return 0;
2511 }
2512
2513 device->fs_devices->missing_devices++;
2514 if (test_and_clear_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
2515 list_del_init(&device->dev_alloc_list);
2516 WARN_ON(device->fs_devices->rw_devices < 1);
2517 device->fs_devices->rw_devices--;
2518 }
2519 can_rw = btrfs_check_rw_degradable(fs_info, device);
2520 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2521 /*
2522 * Now device is considered missing, btrfs_device_name() won't give a
2523 * meaningful result anymore, so only output the devid.
2524 */
2525 if (unlikely(!can_rw)) {
2526 btrfs_crit(fs_info,
2527 "btrfs device id %llu has gone missing, can not maintain read-write",
2528 device->devid);
2529 return -EIO;
2530 }
2531 btrfs_warn(fs_info,
2532 "btrfs device id %llu has gone missing, continue as degraded",
2533 device->devid);
2534 btrfs_set_opt(fs_info->mount_opt, DEGRADED);
2535 return 0;
2536}
2537
2538static void btrfs_shutdown(struct super_block *sb)
2539{
2540 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2541
2542 btrfs_force_shutdown(fs_info);
2543}
2544
2545static int btrfs_show_stats(struct seq_file *seq, struct dentry *root)
2546{
2547 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2548
2549 if (btrfs_is_zoned(fs_info)) {
2550 btrfs_show_zoned_stats(fs_info, seq);
2551 return 0;
2552 }
2553
2554 return 0;
2555}
2556
2557static const struct super_operations btrfs_super_ops = {
2558 .drop_inode = btrfs_drop_inode,
2559 .evict_inode = btrfs_evict_inode,
2560 .put_super = btrfs_put_super,
2561 .sync_fs = btrfs_sync_fs,
2562 .show_options = btrfs_show_options,
2563 .show_devname = btrfs_show_devname,
2564 .alloc_inode = btrfs_alloc_inode,
2565 .destroy_inode = btrfs_destroy_inode,
2566 .free_inode = btrfs_free_inode,
2567 .statfs = btrfs_statfs,
2568 .freeze_fs = btrfs_freeze,
2569 .unfreeze_fs = btrfs_unfreeze,
2570 .nr_cached_objects = btrfs_nr_cached_objects,
2571 .free_cached_objects = btrfs_free_cached_objects,
2572 .show_stats = btrfs_show_stats,
2573 .remove_bdev = btrfs_remove_bdev,
2574 .shutdown = btrfs_shutdown,
2575};
2576
2577static const struct file_operations btrfs_ctl_fops = {
2578 .open = btrfs_control_open,
2579 .unlocked_ioctl = btrfs_control_ioctl,
2580 .compat_ioctl = compat_ptr_ioctl,
2581 .owner = THIS_MODULE,
2582 .llseek = noop_llseek,
2583};
2584
2585static struct miscdevice btrfs_misc = {
2586 .minor = BTRFS_MINOR,
2587 .name = "btrfs-control",
2588 .fops = &btrfs_ctl_fops
2589};
2590
2591MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2592MODULE_ALIAS("devname:btrfs-control");
2593
2594static int __init btrfs_interface_init(void)
2595{
2596 return misc_register(&btrfs_misc);
2597}
2598
2599static __cold void btrfs_interface_exit(void)
2600{
2601 misc_deregister(&btrfs_misc);
2602}
2603
2604static int __init btrfs_print_mod_info(void)
2605{
2606 static const char options[] = ""
2607#ifdef CONFIG_BTRFS_EXPERIMENTAL
2608 ", experimental=on"
2609#endif
2610#ifdef CONFIG_BTRFS_DEBUG
2611 ", debug=on"
2612#endif
2613#ifdef CONFIG_BTRFS_ASSERT
2614 ", assert=on"
2615#endif
2616#ifdef CONFIG_BLK_DEV_ZONED
2617 ", zoned=yes"
2618#else
2619 ", zoned=no"
2620#endif
2621#ifdef CONFIG_FS_VERITY
2622 ", fsverity=yes"
2623#else
2624 ", fsverity=no"
2625#endif
2626 ;
2627
2628#ifdef CONFIG_BTRFS_EXPERIMENTAL
2629 if (btrfs_get_mod_read_policy() == NULL)
2630 pr_info("Btrfs loaded%s\n", options);
2631 else
2632 pr_info("Btrfs loaded%s, read_policy=%s\n",
2633 options, btrfs_get_mod_read_policy());
2634#else
2635 pr_info("Btrfs loaded%s\n", options);
2636#endif
2637
2638 return 0;
2639}
2640
2641static int register_btrfs(void)
2642{
2643 return register_filesystem(&btrfs_fs_type);
2644}
2645
2646static void unregister_btrfs(void)
2647{
2648 unregister_filesystem(&btrfs_fs_type);
2649}
2650
2651/* Helper structure for long init/exit functions. */
2652struct init_sequence {
2653 int (*init_func)(void);
2654 /* Can be NULL if the init_func doesn't need cleanup. */
2655 void (*exit_func)(void);
2656};
2657
2658static const struct init_sequence mod_init_seq[] = {
2659 {
2660 .init_func = btrfs_props_init,
2661 .exit_func = NULL,
2662 }, {
2663 .init_func = btrfs_init_sysfs,
2664 .exit_func = btrfs_exit_sysfs,
2665 }, {
2666 .init_func = btrfs_init_compress,
2667 .exit_func = btrfs_exit_compress,
2668 }, {
2669 .init_func = btrfs_init_block_group,
2670 .exit_func = btrfs_exit_block_group,
2671 }, {
2672 .init_func = btrfs_init_cachep,
2673 .exit_func = btrfs_destroy_cachep,
2674 }, {
2675 .init_func = btrfs_init_dio,
2676 .exit_func = btrfs_destroy_dio,
2677 }, {
2678 .init_func = btrfs_transaction_init,
2679 .exit_func = btrfs_transaction_exit,
2680 }, {
2681 .init_func = btrfs_ctree_init,
2682 .exit_func = btrfs_ctree_exit,
2683 }, {
2684 .init_func = btrfs_free_space_init,
2685 .exit_func = btrfs_free_space_exit,
2686 }, {
2687 .init_func = btrfs_extent_state_init_cachep,
2688 .exit_func = btrfs_extent_state_free_cachep,
2689 }, {
2690 .init_func = extent_buffer_init_cachep,
2691 .exit_func = extent_buffer_free_cachep,
2692 }, {
2693 .init_func = btrfs_bioset_init,
2694 .exit_func = btrfs_bioset_exit,
2695 }, {
2696 .init_func = btrfs_extent_map_init,
2697 .exit_func = btrfs_extent_map_exit,
2698#ifdef CONFIG_BTRFS_EXPERIMENTAL
2699 }, {
2700 .init_func = btrfs_read_policy_init,
2701 .exit_func = NULL,
2702#endif
2703 }, {
2704 .init_func = ordered_data_init,
2705 .exit_func = ordered_data_exit,
2706 }, {
2707 .init_func = btrfs_delayed_inode_init,
2708 .exit_func = btrfs_delayed_inode_exit,
2709 }, {
2710 .init_func = btrfs_auto_defrag_init,
2711 .exit_func = btrfs_auto_defrag_exit,
2712 }, {
2713 .init_func = btrfs_delayed_ref_init,
2714 .exit_func = btrfs_delayed_ref_exit,
2715 }, {
2716 .init_func = btrfs_prelim_ref_init,
2717 .exit_func = btrfs_prelim_ref_exit,
2718 }, {
2719 .init_func = btrfs_interface_init,
2720 .exit_func = btrfs_interface_exit,
2721 }, {
2722 .init_func = btrfs_print_mod_info,
2723 .exit_func = NULL,
2724 }, {
2725 .init_func = btrfs_run_sanity_tests,
2726 .exit_func = NULL,
2727 }, {
2728 .init_func = register_btrfs,
2729 .exit_func = unregister_btrfs,
2730 }
2731};
2732
2733static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2734
2735static __always_inline void btrfs_exit_btrfs_fs(void)
2736{
2737 int i;
2738
2739 for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2740 if (!mod_init_result[i])
2741 continue;
2742 if (mod_init_seq[i].exit_func)
2743 mod_init_seq[i].exit_func();
2744 mod_init_result[i] = false;
2745 }
2746}
2747
2748static void __exit exit_btrfs_fs(void)
2749{
2750 btrfs_exit_btrfs_fs();
2751 btrfs_cleanup_fs_uuids();
2752}
2753
2754static int __init init_btrfs_fs(void)
2755{
2756 int ret;
2757 int i;
2758
2759 for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2760 ASSERT(!mod_init_result[i]);
2761 ret = mod_init_seq[i].init_func();
2762 if (ret < 0) {
2763 btrfs_exit_btrfs_fs();
2764 return ret;
2765 }
2766 mod_init_result[i] = true;
2767 }
2768 return 0;
2769}
2770
2771late_initcall(init_btrfs_fs);
2772module_exit(exit_btrfs_fs)
2773
2774MODULE_DESCRIPTION("B-Tree File System (BTRFS)");
2775MODULE_LICENSE("GPL");