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