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/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/blkdev.h>
20#include <linux/module.h>
21#include <linux/buffer_head.h>
22#include <linux/fs.h>
23#include <linux/pagemap.h>
24#include <linux/highmem.h>
25#include <linux/time.h>
26#include <linux/init.h>
27#include <linux/seq_file.h>
28#include <linux/string.h>
29#include <linux/backing-dev.h>
30#include <linux/mount.h>
31#include <linux/mpage.h>
32#include <linux/swap.h>
33#include <linux/writeback.h>
34#include <linux/statfs.h>
35#include <linux/compat.h>
36#include <linux/parser.h>
37#include <linux/ctype.h>
38#include <linux/namei.h>
39#include <linux/miscdevice.h>
40#include <linux/magic.h>
41#include <linux/slab.h>
42#include <linux/cleancache.h>
43#include <linux/ratelimit.h>
44#include "compat.h"
45#include "delayed-inode.h"
46#include "ctree.h"
47#include "disk-io.h"
48#include "transaction.h"
49#include "btrfs_inode.h"
50#include "ioctl.h"
51#include "print-tree.h"
52#include "xattr.h"
53#include "volumes.h"
54#include "version.h"
55#include "export.h"
56#include "compression.h"
57#include "rcu-string.h"
58
59#define CREATE_TRACE_POINTS
60#include <trace/events/btrfs.h>
61
62static const struct super_operations btrfs_super_ops;
63static struct file_system_type btrfs_fs_type;
64
65static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
66 char nbuf[16])
67{
68 char *errstr = NULL;
69
70 switch (errno) {
71 case -EIO:
72 errstr = "IO failure";
73 break;
74 case -ENOMEM:
75 errstr = "Out of memory";
76 break;
77 case -EROFS:
78 errstr = "Readonly filesystem";
79 break;
80 case -EEXIST:
81 errstr = "Object already exists";
82 break;
83 default:
84 if (nbuf) {
85 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
86 errstr = nbuf;
87 }
88 break;
89 }
90
91 return errstr;
92}
93
94static void __save_error_info(struct btrfs_fs_info *fs_info)
95{
96 /*
97 * today we only save the error info into ram. Long term we'll
98 * also send it down to the disk
99 */
100 fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
101}
102
103/* NOTE:
104 * We move write_super stuff at umount in order to avoid deadlock
105 * for umount hold all lock.
106 */
107static void save_error_info(struct btrfs_fs_info *fs_info)
108{
109 __save_error_info(fs_info);
110}
111
112/* btrfs handle error by forcing the filesystem readonly */
113static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
114{
115 struct super_block *sb = fs_info->sb;
116
117 if (sb->s_flags & MS_RDONLY)
118 return;
119
120 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
121 sb->s_flags |= MS_RDONLY;
122 printk(KERN_INFO "btrfs is forced readonly\n");
123 __btrfs_scrub_cancel(fs_info);
124// WARN_ON(1);
125 }
126}
127
128/*
129 * __btrfs_std_error decodes expected errors from the caller and
130 * invokes the approciate error response.
131 */
132void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
133 unsigned int line, int errno, const char *fmt, ...)
134{
135 struct super_block *sb = fs_info->sb;
136 char nbuf[16];
137 const char *errstr;
138 va_list args;
139 va_start(args, fmt);
140
141 /*
142 * Special case: if the error is EROFS, and we're already
143 * under MS_RDONLY, then it is safe here.
144 */
145 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
146 return;
147
148 errstr = btrfs_decode_error(fs_info, errno, nbuf);
149 if (fmt) {
150 struct va_format vaf = {
151 .fmt = fmt,
152 .va = &args,
153 };
154
155 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s (%pV)\n",
156 sb->s_id, function, line, errstr, &vaf);
157 } else {
158 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
159 sb->s_id, function, line, errstr);
160 }
161
162 /* Don't go through full error handling during mount */
163 if (sb->s_flags & MS_BORN) {
164 save_error_info(fs_info);
165 btrfs_handle_error(fs_info);
166 }
167 va_end(args);
168}
169
170const char *logtypes[] = {
171 "emergency",
172 "alert",
173 "critical",
174 "error",
175 "warning",
176 "notice",
177 "info",
178 "debug",
179};
180
181void btrfs_printk(struct btrfs_fs_info *fs_info, const char *fmt, ...)
182{
183 struct super_block *sb = fs_info->sb;
184 char lvl[4];
185 struct va_format vaf;
186 va_list args;
187 const char *type = logtypes[4];
188
189 va_start(args, fmt);
190
191 if (fmt[0] == '<' && isdigit(fmt[1]) && fmt[2] == '>') {
192 memcpy(lvl, fmt, 3);
193 lvl[3] = '\0';
194 fmt += 3;
195 type = logtypes[fmt[1] - '0'];
196 } else
197 *lvl = '\0';
198
199 vaf.fmt = fmt;
200 vaf.va = &args;
201 printk("%sBTRFS %s (device %s): %pV", lvl, type, sb->s_id, &vaf);
202}
203
204/*
205 * We only mark the transaction aborted and then set the file system read-only.
206 * This will prevent new transactions from starting or trying to join this
207 * one.
208 *
209 * This means that error recovery at the call site is limited to freeing
210 * any local memory allocations and passing the error code up without
211 * further cleanup. The transaction should complete as it normally would
212 * in the call path but will return -EIO.
213 *
214 * We'll complete the cleanup in btrfs_end_transaction and
215 * btrfs_commit_transaction.
216 */
217void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
218 struct btrfs_root *root, const char *function,
219 unsigned int line, int errno)
220{
221 WARN_ONCE(1, KERN_DEBUG "btrfs: Transaction aborted");
222 trans->aborted = errno;
223 /* Nothing used. The other threads that have joined this
224 * transaction may be able to continue. */
225 if (!trans->blocks_used) {
226 btrfs_printk(root->fs_info, "Aborting unused transaction.\n");
227 return;
228 }
229 trans->transaction->aborted = errno;
230 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
231}
232/*
233 * __btrfs_panic decodes unexpected, fatal errors from the caller,
234 * issues an alert, and either panics or BUGs, depending on mount options.
235 */
236void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
237 unsigned int line, int errno, const char *fmt, ...)
238{
239 char nbuf[16];
240 char *s_id = "<unknown>";
241 const char *errstr;
242 struct va_format vaf = { .fmt = fmt };
243 va_list args;
244
245 if (fs_info)
246 s_id = fs_info->sb->s_id;
247
248 va_start(args, fmt);
249 vaf.va = &args;
250
251 errstr = btrfs_decode_error(fs_info, errno, nbuf);
252 if (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR)
253 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
254 s_id, function, line, &vaf, errstr);
255
256 printk(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
257 s_id, function, line, &vaf, errstr);
258 va_end(args);
259 /* Caller calls BUG() */
260}
261
262static void btrfs_put_super(struct super_block *sb)
263{
264 (void)close_ctree(btrfs_sb(sb)->tree_root);
265 /* FIXME: need to fix VFS to return error? */
266 /* AV: return it _where_? ->put_super() can be triggered by any number
267 * of async events, up to and including delivery of SIGKILL to the
268 * last process that kept it busy. Or segfault in the aforementioned
269 * process... Whom would you report that to?
270 */
271}
272
273enum {
274 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
275 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
276 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
277 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
278 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
279 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
280 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
281 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
282 Opt_check_integrity, Opt_check_integrity_including_extent_data,
283 Opt_check_integrity_print_mask, Opt_fatal_errors,
284 Opt_err,
285};
286
287static match_table_t tokens = {
288 {Opt_degraded, "degraded"},
289 {Opt_subvol, "subvol=%s"},
290 {Opt_subvolid, "subvolid=%d"},
291 {Opt_device, "device=%s"},
292 {Opt_nodatasum, "nodatasum"},
293 {Opt_nodatacow, "nodatacow"},
294 {Opt_nobarrier, "nobarrier"},
295 {Opt_max_inline, "max_inline=%s"},
296 {Opt_alloc_start, "alloc_start=%s"},
297 {Opt_thread_pool, "thread_pool=%d"},
298 {Opt_compress, "compress"},
299 {Opt_compress_type, "compress=%s"},
300 {Opt_compress_force, "compress-force"},
301 {Opt_compress_force_type, "compress-force=%s"},
302 {Opt_ssd, "ssd"},
303 {Opt_ssd_spread, "ssd_spread"},
304 {Opt_nossd, "nossd"},
305 {Opt_noacl, "noacl"},
306 {Opt_notreelog, "notreelog"},
307 {Opt_flushoncommit, "flushoncommit"},
308 {Opt_ratio, "metadata_ratio=%d"},
309 {Opt_discard, "discard"},
310 {Opt_space_cache, "space_cache"},
311 {Opt_clear_cache, "clear_cache"},
312 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
313 {Opt_enospc_debug, "enospc_debug"},
314 {Opt_subvolrootid, "subvolrootid=%d"},
315 {Opt_defrag, "autodefrag"},
316 {Opt_inode_cache, "inode_cache"},
317 {Opt_no_space_cache, "nospace_cache"},
318 {Opt_recovery, "recovery"},
319 {Opt_skip_balance, "skip_balance"},
320 {Opt_check_integrity, "check_int"},
321 {Opt_check_integrity_including_extent_data, "check_int_data"},
322 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
323 {Opt_fatal_errors, "fatal_errors=%s"},
324 {Opt_err, NULL},
325};
326
327/*
328 * Regular mount options parser. Everything that is needed only when
329 * reading in a new superblock is parsed here.
330 * XXX JDM: This needs to be cleaned up for remount.
331 */
332int btrfs_parse_options(struct btrfs_root *root, char *options)
333{
334 struct btrfs_fs_info *info = root->fs_info;
335 substring_t args[MAX_OPT_ARGS];
336 char *p, *num, *orig = NULL;
337 u64 cache_gen;
338 int intarg;
339 int ret = 0;
340 char *compress_type;
341 bool compress_force = false;
342
343 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
344 if (cache_gen)
345 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
346
347 if (!options)
348 goto out;
349
350 /*
351 * strsep changes the string, duplicate it because parse_options
352 * gets called twice
353 */
354 options = kstrdup(options, GFP_NOFS);
355 if (!options)
356 return -ENOMEM;
357
358 orig = options;
359
360 while ((p = strsep(&options, ",")) != NULL) {
361 int token;
362 if (!*p)
363 continue;
364
365 token = match_token(p, tokens, args);
366 switch (token) {
367 case Opt_degraded:
368 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
369 btrfs_set_opt(info->mount_opt, DEGRADED);
370 break;
371 case Opt_subvol:
372 case Opt_subvolid:
373 case Opt_subvolrootid:
374 case Opt_device:
375 /*
376 * These are parsed by btrfs_parse_early_options
377 * and can be happily ignored here.
378 */
379 break;
380 case Opt_nodatasum:
381 printk(KERN_INFO "btrfs: setting nodatasum\n");
382 btrfs_set_opt(info->mount_opt, NODATASUM);
383 break;
384 case Opt_nodatacow:
385 printk(KERN_INFO "btrfs: setting nodatacow\n");
386 btrfs_set_opt(info->mount_opt, NODATACOW);
387 btrfs_set_opt(info->mount_opt, NODATASUM);
388 break;
389 case Opt_compress_force:
390 case Opt_compress_force_type:
391 compress_force = true;
392 case Opt_compress:
393 case Opt_compress_type:
394 if (token == Opt_compress ||
395 token == Opt_compress_force ||
396 strcmp(args[0].from, "zlib") == 0) {
397 compress_type = "zlib";
398 info->compress_type = BTRFS_COMPRESS_ZLIB;
399 } else if (strcmp(args[0].from, "lzo") == 0) {
400 compress_type = "lzo";
401 info->compress_type = BTRFS_COMPRESS_LZO;
402 } else {
403 ret = -EINVAL;
404 goto out;
405 }
406
407 btrfs_set_opt(info->mount_opt, COMPRESS);
408 if (compress_force) {
409 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
410 pr_info("btrfs: force %s compression\n",
411 compress_type);
412 } else
413 pr_info("btrfs: use %s compression\n",
414 compress_type);
415 break;
416 case Opt_ssd:
417 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
418 btrfs_set_opt(info->mount_opt, SSD);
419 break;
420 case Opt_ssd_spread:
421 printk(KERN_INFO "btrfs: use spread ssd "
422 "allocation scheme\n");
423 btrfs_set_opt(info->mount_opt, SSD);
424 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
425 break;
426 case Opt_nossd:
427 printk(KERN_INFO "btrfs: not using ssd allocation "
428 "scheme\n");
429 btrfs_set_opt(info->mount_opt, NOSSD);
430 btrfs_clear_opt(info->mount_opt, SSD);
431 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
432 break;
433 case Opt_nobarrier:
434 printk(KERN_INFO "btrfs: turning off barriers\n");
435 btrfs_set_opt(info->mount_opt, NOBARRIER);
436 break;
437 case Opt_thread_pool:
438 intarg = 0;
439 match_int(&args[0], &intarg);
440 if (intarg)
441 info->thread_pool_size = intarg;
442 break;
443 case Opt_max_inline:
444 num = match_strdup(&args[0]);
445 if (num) {
446 info->max_inline = memparse(num, NULL);
447 kfree(num);
448
449 if (info->max_inline) {
450 info->max_inline = max_t(u64,
451 info->max_inline,
452 root->sectorsize);
453 }
454 printk(KERN_INFO "btrfs: max_inline at %llu\n",
455 (unsigned long long)info->max_inline);
456 }
457 break;
458 case Opt_alloc_start:
459 num = match_strdup(&args[0]);
460 if (num) {
461 info->alloc_start = memparse(num, NULL);
462 kfree(num);
463 printk(KERN_INFO
464 "btrfs: allocations start at %llu\n",
465 (unsigned long long)info->alloc_start);
466 }
467 break;
468 case Opt_noacl:
469 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
470 break;
471 case Opt_notreelog:
472 printk(KERN_INFO "btrfs: disabling tree log\n");
473 btrfs_set_opt(info->mount_opt, NOTREELOG);
474 break;
475 case Opt_flushoncommit:
476 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
477 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
478 break;
479 case Opt_ratio:
480 intarg = 0;
481 match_int(&args[0], &intarg);
482 if (intarg) {
483 info->metadata_ratio = intarg;
484 printk(KERN_INFO "btrfs: metadata ratio %d\n",
485 info->metadata_ratio);
486 }
487 break;
488 case Opt_discard:
489 btrfs_set_opt(info->mount_opt, DISCARD);
490 break;
491 case Opt_space_cache:
492 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
493 break;
494 case Opt_no_space_cache:
495 printk(KERN_INFO "btrfs: disabling disk space caching\n");
496 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
497 break;
498 case Opt_inode_cache:
499 printk(KERN_INFO "btrfs: enabling inode map caching\n");
500 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
501 break;
502 case Opt_clear_cache:
503 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
504 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
505 break;
506 case Opt_user_subvol_rm_allowed:
507 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
508 break;
509 case Opt_enospc_debug:
510 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
511 break;
512 case Opt_defrag:
513 printk(KERN_INFO "btrfs: enabling auto defrag");
514 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
515 break;
516 case Opt_recovery:
517 printk(KERN_INFO "btrfs: enabling auto recovery");
518 btrfs_set_opt(info->mount_opt, RECOVERY);
519 break;
520 case Opt_skip_balance:
521 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
522 break;
523#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
524 case Opt_check_integrity_including_extent_data:
525 printk(KERN_INFO "btrfs: enabling check integrity"
526 " including extent data\n");
527 btrfs_set_opt(info->mount_opt,
528 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
529 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
530 break;
531 case Opt_check_integrity:
532 printk(KERN_INFO "btrfs: enabling check integrity\n");
533 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
534 break;
535 case Opt_check_integrity_print_mask:
536 intarg = 0;
537 match_int(&args[0], &intarg);
538 if (intarg) {
539 info->check_integrity_print_mask = intarg;
540 printk(KERN_INFO "btrfs:"
541 " check_integrity_print_mask 0x%x\n",
542 info->check_integrity_print_mask);
543 }
544 break;
545#else
546 case Opt_check_integrity_including_extent_data:
547 case Opt_check_integrity:
548 case Opt_check_integrity_print_mask:
549 printk(KERN_ERR "btrfs: support for check_integrity*"
550 " not compiled in!\n");
551 ret = -EINVAL;
552 goto out;
553#endif
554 case Opt_fatal_errors:
555 if (strcmp(args[0].from, "panic") == 0)
556 btrfs_set_opt(info->mount_opt,
557 PANIC_ON_FATAL_ERROR);
558 else if (strcmp(args[0].from, "bug") == 0)
559 btrfs_clear_opt(info->mount_opt,
560 PANIC_ON_FATAL_ERROR);
561 else {
562 ret = -EINVAL;
563 goto out;
564 }
565 break;
566 case Opt_err:
567 printk(KERN_INFO "btrfs: unrecognized mount option "
568 "'%s'\n", p);
569 ret = -EINVAL;
570 goto out;
571 default:
572 break;
573 }
574 }
575out:
576 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
577 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
578 kfree(orig);
579 return ret;
580}
581
582/*
583 * Parse mount options that are required early in the mount process.
584 *
585 * All other options will be parsed on much later in the mount process and
586 * only when we need to allocate a new super block.
587 */
588static int btrfs_parse_early_options(const char *options, fmode_t flags,
589 void *holder, char **subvol_name, u64 *subvol_objectid,
590 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
591{
592 substring_t args[MAX_OPT_ARGS];
593 char *device_name, *opts, *orig, *p;
594 int error = 0;
595 int intarg;
596
597 if (!options)
598 return 0;
599
600 /*
601 * strsep changes the string, duplicate it because parse_options
602 * gets called twice
603 */
604 opts = kstrdup(options, GFP_KERNEL);
605 if (!opts)
606 return -ENOMEM;
607 orig = opts;
608
609 while ((p = strsep(&opts, ",")) != NULL) {
610 int token;
611 if (!*p)
612 continue;
613
614 token = match_token(p, tokens, args);
615 switch (token) {
616 case Opt_subvol:
617 kfree(*subvol_name);
618 *subvol_name = match_strdup(&args[0]);
619 break;
620 case Opt_subvolid:
621 intarg = 0;
622 error = match_int(&args[0], &intarg);
623 if (!error) {
624 /* we want the original fs_tree */
625 if (!intarg)
626 *subvol_objectid =
627 BTRFS_FS_TREE_OBJECTID;
628 else
629 *subvol_objectid = intarg;
630 }
631 break;
632 case Opt_subvolrootid:
633 intarg = 0;
634 error = match_int(&args[0], &intarg);
635 if (!error) {
636 /* we want the original fs_tree */
637 if (!intarg)
638 *subvol_rootid =
639 BTRFS_FS_TREE_OBJECTID;
640 else
641 *subvol_rootid = intarg;
642 }
643 break;
644 case Opt_device:
645 device_name = match_strdup(&args[0]);
646 if (!device_name) {
647 error = -ENOMEM;
648 goto out;
649 }
650 error = btrfs_scan_one_device(device_name,
651 flags, holder, fs_devices);
652 kfree(device_name);
653 if (error)
654 goto out;
655 break;
656 default:
657 break;
658 }
659 }
660
661out:
662 kfree(orig);
663 return error;
664}
665
666static struct dentry *get_default_root(struct super_block *sb,
667 u64 subvol_objectid)
668{
669 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
670 struct btrfs_root *root = fs_info->tree_root;
671 struct btrfs_root *new_root;
672 struct btrfs_dir_item *di;
673 struct btrfs_path *path;
674 struct btrfs_key location;
675 struct inode *inode;
676 u64 dir_id;
677 int new = 0;
678
679 /*
680 * We have a specific subvol we want to mount, just setup location and
681 * go look up the root.
682 */
683 if (subvol_objectid) {
684 location.objectid = subvol_objectid;
685 location.type = BTRFS_ROOT_ITEM_KEY;
686 location.offset = (u64)-1;
687 goto find_root;
688 }
689
690 path = btrfs_alloc_path();
691 if (!path)
692 return ERR_PTR(-ENOMEM);
693 path->leave_spinning = 1;
694
695 /*
696 * Find the "default" dir item which points to the root item that we
697 * will mount by default if we haven't been given a specific subvolume
698 * to mount.
699 */
700 dir_id = btrfs_super_root_dir(fs_info->super_copy);
701 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
702 if (IS_ERR(di)) {
703 btrfs_free_path(path);
704 return ERR_CAST(di);
705 }
706 if (!di) {
707 /*
708 * Ok the default dir item isn't there. This is weird since
709 * it's always been there, but don't freak out, just try and
710 * mount to root most subvolume.
711 */
712 btrfs_free_path(path);
713 dir_id = BTRFS_FIRST_FREE_OBJECTID;
714 new_root = fs_info->fs_root;
715 goto setup_root;
716 }
717
718 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
719 btrfs_free_path(path);
720
721find_root:
722 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
723 if (IS_ERR(new_root))
724 return ERR_CAST(new_root);
725
726 if (btrfs_root_refs(&new_root->root_item) == 0)
727 return ERR_PTR(-ENOENT);
728
729 dir_id = btrfs_root_dirid(&new_root->root_item);
730setup_root:
731 location.objectid = dir_id;
732 location.type = BTRFS_INODE_ITEM_KEY;
733 location.offset = 0;
734
735 inode = btrfs_iget(sb, &location, new_root, &new);
736 if (IS_ERR(inode))
737 return ERR_CAST(inode);
738
739 /*
740 * If we're just mounting the root most subvol put the inode and return
741 * a reference to the dentry. We will have already gotten a reference
742 * to the inode in btrfs_fill_super so we're good to go.
743 */
744 if (!new && sb->s_root->d_inode == inode) {
745 iput(inode);
746 return dget(sb->s_root);
747 }
748
749 return d_obtain_alias(inode);
750}
751
752static int btrfs_fill_super(struct super_block *sb,
753 struct btrfs_fs_devices *fs_devices,
754 void *data, int silent)
755{
756 struct inode *inode;
757 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
758 struct btrfs_key key;
759 int err;
760
761 sb->s_maxbytes = MAX_LFS_FILESIZE;
762 sb->s_magic = BTRFS_SUPER_MAGIC;
763 sb->s_op = &btrfs_super_ops;
764 sb->s_d_op = &btrfs_dentry_operations;
765 sb->s_export_op = &btrfs_export_ops;
766 sb->s_xattr = btrfs_xattr_handlers;
767 sb->s_time_gran = 1;
768#ifdef CONFIG_BTRFS_FS_POSIX_ACL
769 sb->s_flags |= MS_POSIXACL;
770#endif
771 sb->s_flags |= MS_I_VERSION;
772 err = open_ctree(sb, fs_devices, (char *)data);
773 if (err) {
774 printk("btrfs: open_ctree failed\n");
775 return err;
776 }
777
778 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
779 key.type = BTRFS_INODE_ITEM_KEY;
780 key.offset = 0;
781 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
782 if (IS_ERR(inode)) {
783 err = PTR_ERR(inode);
784 goto fail_close;
785 }
786
787 sb->s_root = d_make_root(inode);
788 if (!sb->s_root) {
789 err = -ENOMEM;
790 goto fail_close;
791 }
792
793 save_mount_options(sb, data);
794 cleancache_init_fs(sb);
795 sb->s_flags |= MS_ACTIVE;
796 return 0;
797
798fail_close:
799 close_ctree(fs_info->tree_root);
800 return err;
801}
802
803int btrfs_sync_fs(struct super_block *sb, int wait)
804{
805 struct btrfs_trans_handle *trans;
806 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
807 struct btrfs_root *root = fs_info->tree_root;
808 int ret;
809
810 trace_btrfs_sync_fs(wait);
811
812 if (!wait) {
813 filemap_flush(fs_info->btree_inode->i_mapping);
814 return 0;
815 }
816
817 btrfs_wait_ordered_extents(root, 0, 0);
818
819 trans = btrfs_start_transaction(root, 0);
820 if (IS_ERR(trans))
821 return PTR_ERR(trans);
822 ret = btrfs_commit_transaction(trans, root);
823 return ret;
824}
825
826static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
827{
828 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
829 struct btrfs_root *root = info->tree_root;
830 char *compress_type;
831
832 if (btrfs_test_opt(root, DEGRADED))
833 seq_puts(seq, ",degraded");
834 if (btrfs_test_opt(root, NODATASUM))
835 seq_puts(seq, ",nodatasum");
836 if (btrfs_test_opt(root, NODATACOW))
837 seq_puts(seq, ",nodatacow");
838 if (btrfs_test_opt(root, NOBARRIER))
839 seq_puts(seq, ",nobarrier");
840 if (info->max_inline != 8192 * 1024)
841 seq_printf(seq, ",max_inline=%llu",
842 (unsigned long long)info->max_inline);
843 if (info->alloc_start != 0)
844 seq_printf(seq, ",alloc_start=%llu",
845 (unsigned long long)info->alloc_start);
846 if (info->thread_pool_size != min_t(unsigned long,
847 num_online_cpus() + 2, 8))
848 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
849 if (btrfs_test_opt(root, COMPRESS)) {
850 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
851 compress_type = "zlib";
852 else
853 compress_type = "lzo";
854 if (btrfs_test_opt(root, FORCE_COMPRESS))
855 seq_printf(seq, ",compress-force=%s", compress_type);
856 else
857 seq_printf(seq, ",compress=%s", compress_type);
858 }
859 if (btrfs_test_opt(root, NOSSD))
860 seq_puts(seq, ",nossd");
861 if (btrfs_test_opt(root, SSD_SPREAD))
862 seq_puts(seq, ",ssd_spread");
863 else if (btrfs_test_opt(root, SSD))
864 seq_puts(seq, ",ssd");
865 if (btrfs_test_opt(root, NOTREELOG))
866 seq_puts(seq, ",notreelog");
867 if (btrfs_test_opt(root, FLUSHONCOMMIT))
868 seq_puts(seq, ",flushoncommit");
869 if (btrfs_test_opt(root, DISCARD))
870 seq_puts(seq, ",discard");
871 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
872 seq_puts(seq, ",noacl");
873 if (btrfs_test_opt(root, SPACE_CACHE))
874 seq_puts(seq, ",space_cache");
875 else
876 seq_puts(seq, ",nospace_cache");
877 if (btrfs_test_opt(root, CLEAR_CACHE))
878 seq_puts(seq, ",clear_cache");
879 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
880 seq_puts(seq, ",user_subvol_rm_allowed");
881 if (btrfs_test_opt(root, ENOSPC_DEBUG))
882 seq_puts(seq, ",enospc_debug");
883 if (btrfs_test_opt(root, AUTO_DEFRAG))
884 seq_puts(seq, ",autodefrag");
885 if (btrfs_test_opt(root, INODE_MAP_CACHE))
886 seq_puts(seq, ",inode_cache");
887 if (btrfs_test_opt(root, SKIP_BALANCE))
888 seq_puts(seq, ",skip_balance");
889 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
890 seq_puts(seq, ",fatal_errors=panic");
891 return 0;
892}
893
894static int btrfs_test_super(struct super_block *s, void *data)
895{
896 struct btrfs_fs_info *p = data;
897 struct btrfs_fs_info *fs_info = btrfs_sb(s);
898
899 return fs_info->fs_devices == p->fs_devices;
900}
901
902static int btrfs_set_super(struct super_block *s, void *data)
903{
904 int err = set_anon_super(s, data);
905 if (!err)
906 s->s_fs_info = data;
907 return err;
908}
909
910/*
911 * subvolumes are identified by ino 256
912 */
913static inline int is_subvolume_inode(struct inode *inode)
914{
915 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
916 return 1;
917 return 0;
918}
919
920/*
921 * This will strip out the subvol=%s argument for an argument string and add
922 * subvolid=0 to make sure we get the actual tree root for path walking to the
923 * subvol we want.
924 */
925static char *setup_root_args(char *args)
926{
927 unsigned len = strlen(args) + 2 + 1;
928 char *src, *dst, *buf;
929
930 /*
931 * We need the same args as before, but with this substitution:
932 * s!subvol=[^,]+!subvolid=0!
933 *
934 * Since the replacement string is up to 2 bytes longer than the
935 * original, allocate strlen(args) + 2 + 1 bytes.
936 */
937
938 src = strstr(args, "subvol=");
939 /* This shouldn't happen, but just in case.. */
940 if (!src)
941 return NULL;
942
943 buf = dst = kmalloc(len, GFP_NOFS);
944 if (!buf)
945 return NULL;
946
947 /*
948 * If the subvol= arg is not at the start of the string,
949 * copy whatever precedes it into buf.
950 */
951 if (src != args) {
952 *src++ = '\0';
953 strcpy(buf, args);
954 dst += strlen(args);
955 }
956
957 strcpy(dst, "subvolid=0");
958 dst += strlen("subvolid=0");
959
960 /*
961 * If there is a "," after the original subvol=... string,
962 * copy that suffix into our buffer. Otherwise, we're done.
963 */
964 src = strchr(src, ',');
965 if (src)
966 strcpy(dst, src);
967
968 return buf;
969}
970
971static struct dentry *mount_subvol(const char *subvol_name, int flags,
972 const char *device_name, char *data)
973{
974 struct dentry *root;
975 struct vfsmount *mnt;
976 char *newargs;
977
978 newargs = setup_root_args(data);
979 if (!newargs)
980 return ERR_PTR(-ENOMEM);
981 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
982 newargs);
983 kfree(newargs);
984 if (IS_ERR(mnt))
985 return ERR_CAST(mnt);
986
987 root = mount_subtree(mnt, subvol_name);
988
989 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
990 struct super_block *s = root->d_sb;
991 dput(root);
992 root = ERR_PTR(-EINVAL);
993 deactivate_locked_super(s);
994 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
995 subvol_name);
996 }
997
998 return root;
999}
1000
1001/*
1002 * Find a superblock for the given device / mount point.
1003 *
1004 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1005 * for multiple device setup. Make sure to keep it in sync.
1006 */
1007static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1008 const char *device_name, void *data)
1009{
1010 struct block_device *bdev = NULL;
1011 struct super_block *s;
1012 struct dentry *root;
1013 struct btrfs_fs_devices *fs_devices = NULL;
1014 struct btrfs_fs_info *fs_info = NULL;
1015 fmode_t mode = FMODE_READ;
1016 char *subvol_name = NULL;
1017 u64 subvol_objectid = 0;
1018 u64 subvol_rootid = 0;
1019 int error = 0;
1020
1021 if (!(flags & MS_RDONLY))
1022 mode |= FMODE_WRITE;
1023
1024 error = btrfs_parse_early_options(data, mode, fs_type,
1025 &subvol_name, &subvol_objectid,
1026 &subvol_rootid, &fs_devices);
1027 if (error) {
1028 kfree(subvol_name);
1029 return ERR_PTR(error);
1030 }
1031
1032 if (subvol_name) {
1033 root = mount_subvol(subvol_name, flags, device_name, data);
1034 kfree(subvol_name);
1035 return root;
1036 }
1037
1038 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1039 if (error)
1040 return ERR_PTR(error);
1041
1042 /*
1043 * Setup a dummy root and fs_info for test/set super. This is because
1044 * we don't actually fill this stuff out until open_ctree, but we need
1045 * it for searching for existing supers, so this lets us do that and
1046 * then open_ctree will properly initialize everything later.
1047 */
1048 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1049 if (!fs_info)
1050 return ERR_PTR(-ENOMEM);
1051
1052 fs_info->fs_devices = fs_devices;
1053
1054 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1055 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1056 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1057 error = -ENOMEM;
1058 goto error_fs_info;
1059 }
1060
1061 error = btrfs_open_devices(fs_devices, mode, fs_type);
1062 if (error)
1063 goto error_fs_info;
1064
1065 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1066 error = -EACCES;
1067 goto error_close_devices;
1068 }
1069
1070 bdev = fs_devices->latest_bdev;
1071 s = sget(fs_type, btrfs_test_super, btrfs_set_super, fs_info);
1072 if (IS_ERR(s)) {
1073 error = PTR_ERR(s);
1074 goto error_close_devices;
1075 }
1076
1077 if (s->s_root) {
1078 btrfs_close_devices(fs_devices);
1079 free_fs_info(fs_info);
1080 if ((flags ^ s->s_flags) & MS_RDONLY)
1081 error = -EBUSY;
1082 } else {
1083 char b[BDEVNAME_SIZE];
1084
1085 s->s_flags = flags | MS_NOSEC;
1086 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1087 btrfs_sb(s)->bdev_holder = fs_type;
1088 error = btrfs_fill_super(s, fs_devices, data,
1089 flags & MS_SILENT ? 1 : 0);
1090 }
1091
1092 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1093 if (IS_ERR(root))
1094 deactivate_locked_super(s);
1095
1096 return root;
1097
1098error_close_devices:
1099 btrfs_close_devices(fs_devices);
1100error_fs_info:
1101 free_fs_info(fs_info);
1102 return ERR_PTR(error);
1103}
1104
1105static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1106{
1107 spin_lock_irq(&workers->lock);
1108 workers->max_workers = new_limit;
1109 spin_unlock_irq(&workers->lock);
1110}
1111
1112static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1113 int new_pool_size, int old_pool_size)
1114{
1115 if (new_pool_size == old_pool_size)
1116 return;
1117
1118 fs_info->thread_pool_size = new_pool_size;
1119
1120 printk(KERN_INFO "btrfs: resize thread pool %d -> %d\n",
1121 old_pool_size, new_pool_size);
1122
1123 btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1124 btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1125 btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1126 btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1127 btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1128 btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1129 btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1130 btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1131 btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1132 btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1133 btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1134 btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1135 btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
1136 btrfs_set_max_workers(&fs_info->scrub_workers, new_pool_size);
1137}
1138
1139static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1140{
1141 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1142 struct btrfs_root *root = fs_info->tree_root;
1143 unsigned old_flags = sb->s_flags;
1144 unsigned long old_opts = fs_info->mount_opt;
1145 unsigned long old_compress_type = fs_info->compress_type;
1146 u64 old_max_inline = fs_info->max_inline;
1147 u64 old_alloc_start = fs_info->alloc_start;
1148 int old_thread_pool_size = fs_info->thread_pool_size;
1149 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1150 int ret;
1151
1152 ret = btrfs_parse_options(root, data);
1153 if (ret) {
1154 ret = -EINVAL;
1155 goto restore;
1156 }
1157
1158 btrfs_resize_thread_pool(fs_info,
1159 fs_info->thread_pool_size, old_thread_pool_size);
1160
1161 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1162 return 0;
1163
1164 if (*flags & MS_RDONLY) {
1165 sb->s_flags |= MS_RDONLY;
1166
1167 ret = btrfs_commit_super(root);
1168 if (ret)
1169 goto restore;
1170 } else {
1171 if (fs_info->fs_devices->rw_devices == 0) {
1172 ret = -EACCES;
1173 goto restore;
1174 }
1175
1176 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1177 ret = -EINVAL;
1178 goto restore;
1179 }
1180
1181 ret = btrfs_cleanup_fs_roots(fs_info);
1182 if (ret)
1183 goto restore;
1184
1185 /* recover relocation */
1186 ret = btrfs_recover_relocation(root);
1187 if (ret)
1188 goto restore;
1189
1190 ret = btrfs_resume_balance_async(fs_info);
1191 if (ret)
1192 goto restore;
1193
1194 sb->s_flags &= ~MS_RDONLY;
1195 }
1196
1197 return 0;
1198
1199restore:
1200 /* We've hit an error - don't reset MS_RDONLY */
1201 if (sb->s_flags & MS_RDONLY)
1202 old_flags |= MS_RDONLY;
1203 sb->s_flags = old_flags;
1204 fs_info->mount_opt = old_opts;
1205 fs_info->compress_type = old_compress_type;
1206 fs_info->max_inline = old_max_inline;
1207 fs_info->alloc_start = old_alloc_start;
1208 btrfs_resize_thread_pool(fs_info,
1209 old_thread_pool_size, fs_info->thread_pool_size);
1210 fs_info->metadata_ratio = old_metadata_ratio;
1211 return ret;
1212}
1213
1214/* Used to sort the devices by max_avail(descending sort) */
1215static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1216 const void *dev_info2)
1217{
1218 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1219 ((struct btrfs_device_info *)dev_info2)->max_avail)
1220 return -1;
1221 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1222 ((struct btrfs_device_info *)dev_info2)->max_avail)
1223 return 1;
1224 else
1225 return 0;
1226}
1227
1228/*
1229 * sort the devices by max_avail, in which max free extent size of each device
1230 * is stored.(Descending Sort)
1231 */
1232static inline void btrfs_descending_sort_devices(
1233 struct btrfs_device_info *devices,
1234 size_t nr_devices)
1235{
1236 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1237 btrfs_cmp_device_free_bytes, NULL);
1238}
1239
1240/*
1241 * The helper to calc the free space on the devices that can be used to store
1242 * file data.
1243 */
1244static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1245{
1246 struct btrfs_fs_info *fs_info = root->fs_info;
1247 struct btrfs_device_info *devices_info;
1248 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1249 struct btrfs_device *device;
1250 u64 skip_space;
1251 u64 type;
1252 u64 avail_space;
1253 u64 used_space;
1254 u64 min_stripe_size;
1255 int min_stripes = 1, num_stripes = 1;
1256 int i = 0, nr_devices;
1257 int ret;
1258
1259 nr_devices = fs_info->fs_devices->open_devices;
1260 BUG_ON(!nr_devices);
1261
1262 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1263 GFP_NOFS);
1264 if (!devices_info)
1265 return -ENOMEM;
1266
1267 /* calc min stripe number for data space alloction */
1268 type = btrfs_get_alloc_profile(root, 1);
1269 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1270 min_stripes = 2;
1271 num_stripes = nr_devices;
1272 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1273 min_stripes = 2;
1274 num_stripes = 2;
1275 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1276 min_stripes = 4;
1277 num_stripes = 4;
1278 }
1279
1280 if (type & BTRFS_BLOCK_GROUP_DUP)
1281 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1282 else
1283 min_stripe_size = BTRFS_STRIPE_LEN;
1284
1285 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1286 if (!device->in_fs_metadata || !device->bdev)
1287 continue;
1288
1289 avail_space = device->total_bytes - device->bytes_used;
1290
1291 /* align with stripe_len */
1292 do_div(avail_space, BTRFS_STRIPE_LEN);
1293 avail_space *= BTRFS_STRIPE_LEN;
1294
1295 /*
1296 * In order to avoid overwritting the superblock on the drive,
1297 * btrfs starts at an offset of at least 1MB when doing chunk
1298 * allocation.
1299 */
1300 skip_space = 1024 * 1024;
1301
1302 /* user can set the offset in fs_info->alloc_start. */
1303 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1304 device->total_bytes)
1305 skip_space = max(fs_info->alloc_start, skip_space);
1306
1307 /*
1308 * btrfs can not use the free space in [0, skip_space - 1],
1309 * we must subtract it from the total. In order to implement
1310 * it, we account the used space in this range first.
1311 */
1312 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1313 &used_space);
1314 if (ret) {
1315 kfree(devices_info);
1316 return ret;
1317 }
1318
1319 /* calc the free space in [0, skip_space - 1] */
1320 skip_space -= used_space;
1321
1322 /*
1323 * we can use the free space in [0, skip_space - 1], subtract
1324 * it from the total.
1325 */
1326 if (avail_space && avail_space >= skip_space)
1327 avail_space -= skip_space;
1328 else
1329 avail_space = 0;
1330
1331 if (avail_space < min_stripe_size)
1332 continue;
1333
1334 devices_info[i].dev = device;
1335 devices_info[i].max_avail = avail_space;
1336
1337 i++;
1338 }
1339
1340 nr_devices = i;
1341
1342 btrfs_descending_sort_devices(devices_info, nr_devices);
1343
1344 i = nr_devices - 1;
1345 avail_space = 0;
1346 while (nr_devices >= min_stripes) {
1347 if (num_stripes > nr_devices)
1348 num_stripes = nr_devices;
1349
1350 if (devices_info[i].max_avail >= min_stripe_size) {
1351 int j;
1352 u64 alloc_size;
1353
1354 avail_space += devices_info[i].max_avail * num_stripes;
1355 alloc_size = devices_info[i].max_avail;
1356 for (j = i + 1 - num_stripes; j <= i; j++)
1357 devices_info[j].max_avail -= alloc_size;
1358 }
1359 i--;
1360 nr_devices--;
1361 }
1362
1363 kfree(devices_info);
1364 *free_bytes = avail_space;
1365 return 0;
1366}
1367
1368static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1369{
1370 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1371 struct btrfs_super_block *disk_super = fs_info->super_copy;
1372 struct list_head *head = &fs_info->space_info;
1373 struct btrfs_space_info *found;
1374 u64 total_used = 0;
1375 u64 total_free_data = 0;
1376 int bits = dentry->d_sb->s_blocksize_bits;
1377 __be32 *fsid = (__be32 *)fs_info->fsid;
1378 int ret;
1379
1380 /* holding chunk_muext to avoid allocating new chunks */
1381 mutex_lock(&fs_info->chunk_mutex);
1382 rcu_read_lock();
1383 list_for_each_entry_rcu(found, head, list) {
1384 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1385 total_free_data += found->disk_total - found->disk_used;
1386 total_free_data -=
1387 btrfs_account_ro_block_groups_free_space(found);
1388 }
1389
1390 total_used += found->disk_used;
1391 }
1392 rcu_read_unlock();
1393
1394 buf->f_namelen = BTRFS_NAME_LEN;
1395 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1396 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1397 buf->f_bsize = dentry->d_sb->s_blocksize;
1398 buf->f_type = BTRFS_SUPER_MAGIC;
1399 buf->f_bavail = total_free_data;
1400 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1401 if (ret) {
1402 mutex_unlock(&fs_info->chunk_mutex);
1403 return ret;
1404 }
1405 buf->f_bavail += total_free_data;
1406 buf->f_bavail = buf->f_bavail >> bits;
1407 mutex_unlock(&fs_info->chunk_mutex);
1408
1409 /* We treat it as constant endianness (it doesn't matter _which_)
1410 because we want the fsid to come out the same whether mounted
1411 on a big-endian or little-endian host */
1412 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1413 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1414 /* Mask in the root object ID too, to disambiguate subvols */
1415 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1416 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1417
1418 return 0;
1419}
1420
1421static void btrfs_kill_super(struct super_block *sb)
1422{
1423 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1424 kill_anon_super(sb);
1425 free_fs_info(fs_info);
1426}
1427
1428static struct file_system_type btrfs_fs_type = {
1429 .owner = THIS_MODULE,
1430 .name = "btrfs",
1431 .mount = btrfs_mount,
1432 .kill_sb = btrfs_kill_super,
1433 .fs_flags = FS_REQUIRES_DEV,
1434};
1435
1436/*
1437 * used by btrfsctl to scan devices when no FS is mounted
1438 */
1439static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1440 unsigned long arg)
1441{
1442 struct btrfs_ioctl_vol_args *vol;
1443 struct btrfs_fs_devices *fs_devices;
1444 int ret = -ENOTTY;
1445
1446 if (!capable(CAP_SYS_ADMIN))
1447 return -EPERM;
1448
1449 vol = memdup_user((void __user *)arg, sizeof(*vol));
1450 if (IS_ERR(vol))
1451 return PTR_ERR(vol);
1452
1453 switch (cmd) {
1454 case BTRFS_IOC_SCAN_DEV:
1455 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1456 &btrfs_fs_type, &fs_devices);
1457 break;
1458 }
1459
1460 kfree(vol);
1461 return ret;
1462}
1463
1464static int btrfs_freeze(struct super_block *sb)
1465{
1466 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1467 mutex_lock(&fs_info->transaction_kthread_mutex);
1468 mutex_lock(&fs_info->cleaner_mutex);
1469 return 0;
1470}
1471
1472static int btrfs_unfreeze(struct super_block *sb)
1473{
1474 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1475 mutex_unlock(&fs_info->cleaner_mutex);
1476 mutex_unlock(&fs_info->transaction_kthread_mutex);
1477 return 0;
1478}
1479
1480static void btrfs_fs_dirty_inode(struct inode *inode, int flags)
1481{
1482 int ret;
1483
1484 ret = btrfs_dirty_inode(inode);
1485 if (ret)
1486 printk_ratelimited(KERN_ERR "btrfs: fail to dirty inode %Lu "
1487 "error %d\n", btrfs_ino(inode), ret);
1488}
1489
1490static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1491{
1492 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1493 struct btrfs_fs_devices *cur_devices;
1494 struct btrfs_device *dev, *first_dev = NULL;
1495 struct list_head *head;
1496 struct rcu_string *name;
1497
1498 mutex_lock(&fs_info->fs_devices->device_list_mutex);
1499 cur_devices = fs_info->fs_devices;
1500 while (cur_devices) {
1501 head = &cur_devices->devices;
1502 list_for_each_entry(dev, head, dev_list) {
1503 if (!first_dev || dev->devid < first_dev->devid)
1504 first_dev = dev;
1505 }
1506 cur_devices = cur_devices->seed;
1507 }
1508
1509 if (first_dev) {
1510 rcu_read_lock();
1511 name = rcu_dereference(first_dev->name);
1512 seq_escape(m, name->str, " \t\n\\");
1513 rcu_read_unlock();
1514 } else {
1515 WARN_ON(1);
1516 }
1517 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1518 return 0;
1519}
1520
1521static const struct super_operations btrfs_super_ops = {
1522 .drop_inode = btrfs_drop_inode,
1523 .evict_inode = btrfs_evict_inode,
1524 .put_super = btrfs_put_super,
1525 .sync_fs = btrfs_sync_fs,
1526 .show_options = btrfs_show_options,
1527 .show_devname = btrfs_show_devname,
1528 .write_inode = btrfs_write_inode,
1529 .dirty_inode = btrfs_fs_dirty_inode,
1530 .alloc_inode = btrfs_alloc_inode,
1531 .destroy_inode = btrfs_destroy_inode,
1532 .statfs = btrfs_statfs,
1533 .remount_fs = btrfs_remount,
1534 .freeze_fs = btrfs_freeze,
1535 .unfreeze_fs = btrfs_unfreeze,
1536};
1537
1538static const struct file_operations btrfs_ctl_fops = {
1539 .unlocked_ioctl = btrfs_control_ioctl,
1540 .compat_ioctl = btrfs_control_ioctl,
1541 .owner = THIS_MODULE,
1542 .llseek = noop_llseek,
1543};
1544
1545static struct miscdevice btrfs_misc = {
1546 .minor = BTRFS_MINOR,
1547 .name = "btrfs-control",
1548 .fops = &btrfs_ctl_fops
1549};
1550
1551MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1552MODULE_ALIAS("devname:btrfs-control");
1553
1554static int btrfs_interface_init(void)
1555{
1556 return misc_register(&btrfs_misc);
1557}
1558
1559static void btrfs_interface_exit(void)
1560{
1561 if (misc_deregister(&btrfs_misc) < 0)
1562 printk(KERN_INFO "misc_deregister failed for control device");
1563}
1564
1565static int __init init_btrfs_fs(void)
1566{
1567 int err;
1568
1569 err = btrfs_init_sysfs();
1570 if (err)
1571 return err;
1572
1573 btrfs_init_compress();
1574
1575 err = btrfs_init_cachep();
1576 if (err)
1577 goto free_compress;
1578
1579 err = extent_io_init();
1580 if (err)
1581 goto free_cachep;
1582
1583 err = extent_map_init();
1584 if (err)
1585 goto free_extent_io;
1586
1587 err = btrfs_delayed_inode_init();
1588 if (err)
1589 goto free_extent_map;
1590
1591 err = btrfs_interface_init();
1592 if (err)
1593 goto free_delayed_inode;
1594
1595 err = register_filesystem(&btrfs_fs_type);
1596 if (err)
1597 goto unregister_ioctl;
1598
1599 btrfs_init_lockdep();
1600
1601 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1602 return 0;
1603
1604unregister_ioctl:
1605 btrfs_interface_exit();
1606free_delayed_inode:
1607 btrfs_delayed_inode_exit();
1608free_extent_map:
1609 extent_map_exit();
1610free_extent_io:
1611 extent_io_exit();
1612free_cachep:
1613 btrfs_destroy_cachep();
1614free_compress:
1615 btrfs_exit_compress();
1616 btrfs_exit_sysfs();
1617 return err;
1618}
1619
1620static void __exit exit_btrfs_fs(void)
1621{
1622 btrfs_destroy_cachep();
1623 btrfs_delayed_inode_exit();
1624 extent_map_exit();
1625 extent_io_exit();
1626 btrfs_interface_exit();
1627 unregister_filesystem(&btrfs_fs_type);
1628 btrfs_exit_sysfs();
1629 btrfs_cleanup_fs_uuids();
1630 btrfs_exit_compress();
1631}
1632
1633module_init(init_btrfs_fs)
1634module_exit(exit_btrfs_fs)
1635
1636MODULE_LICENSE("GPL");