]> 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
btrfs: better handle btrfs_printk() defaults
[mirror_ubuntu-jammy-kernel.git] / fs / btrfs / super.c
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 <linux/btrfs.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 "print-tree.h"
51 #include "hash.h"
52 #include "props.h"
53 #include "xattr.h"
54 #include "volumes.h"
55 #include "export.h"
56 #include "compression.h"
57 #include "rcu-string.h"
58 #include "dev-replace.h"
59 #include "free-space-cache.h"
60 #include "backref.h"
61 #include "tests/btrfs-tests.h"
62
63 #include "qgroup.h"
64 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h>
66
67 static const struct super_operations btrfs_super_ops;
68 static struct file_system_type btrfs_fs_type;
69
70 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
71
72 const char *btrfs_decode_error(int errno)
73 {
74 char *errstr = "unknown";
75
76 switch (errno) {
77 case -EIO:
78 errstr = "IO failure";
79 break;
80 case -ENOMEM:
81 errstr = "Out of memory";
82 break;
83 case -EROFS:
84 errstr = "Readonly filesystem";
85 break;
86 case -EEXIST:
87 errstr = "Object already exists";
88 break;
89 case -ENOSPC:
90 errstr = "No space left";
91 break;
92 case -ENOENT:
93 errstr = "No such entry";
94 break;
95 }
96
97 return errstr;
98 }
99
100 /* btrfs handle error by forcing the filesystem readonly */
101 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
102 {
103 struct super_block *sb = fs_info->sb;
104
105 if (sb->s_flags & MS_RDONLY)
106 return;
107
108 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
109 sb->s_flags |= MS_RDONLY;
110 btrfs_info(fs_info, "forced readonly");
111 /*
112 * Note that a running device replace operation is not
113 * canceled here although there is no way to update
114 * the progress. It would add the risk of a deadlock,
115 * therefore the canceling is omitted. The only penalty
116 * is that some I/O remains active until the procedure
117 * completes. The next time when the filesystem is
118 * mounted writeable again, the device replace
119 * operation continues.
120 */
121 }
122 }
123
124 /*
125 * __btrfs_handle_fs_error decodes expected errors from the caller and
126 * invokes the approciate error response.
127 */
128 __cold
129 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
130 unsigned int line, int errno, const char *fmt, ...)
131 {
132 struct super_block *sb = fs_info->sb;
133 #ifdef CONFIG_PRINTK
134 const char *errstr;
135 #endif
136
137 /*
138 * Special case: if the error is EROFS, and we're already
139 * under MS_RDONLY, then it is safe here.
140 */
141 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
142 return;
143
144 #ifdef CONFIG_PRINTK
145 errstr = btrfs_decode_error(errno);
146 if (fmt) {
147 struct va_format vaf;
148 va_list args;
149
150 va_start(args, fmt);
151 vaf.fmt = fmt;
152 vaf.va = &args;
153
154 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
155 sb->s_id, function, line, errno, errstr, &vaf);
156 va_end(args);
157 } else {
158 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
159 sb->s_id, function, line, errno, errstr);
160 }
161 #endif
162
163 /*
164 * Today we only save the error info to memory. Long term we'll
165 * also send it down to the disk
166 */
167 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
168
169 /* Don't go through full error handling during mount */
170 if (sb->s_flags & MS_BORN)
171 btrfs_handle_error(fs_info);
172 }
173
174 #ifdef CONFIG_PRINTK
175 static const char * const logtypes[] = {
176 "emergency",
177 "alert",
178 "critical",
179 "error",
180 "warning",
181 "notice",
182 "info",
183 "debug",
184 };
185
186
187 /*
188 * Use one ratelimit state per log level so that a flood of less important
189 * messages doesn't cause more important ones to be dropped.
190 */
191 static struct ratelimit_state printk_limits[] = {
192 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
193 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
194 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
195 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
196 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
197 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
198 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
199 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
200 };
201
202 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
203 {
204 struct super_block *sb = fs_info->sb;
205 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
206 struct va_format vaf;
207 va_list args;
208 int kern_level;
209 const char *type = logtypes[4];
210 struct ratelimit_state *ratelimit = &printk_limits[4];
211
212 va_start(args, fmt);
213
214 while ((kern_level = printk_get_level(fmt)) != 0) {
215 size_t size = printk_skip_level(fmt) - fmt;
216
217 if (kern_level >= '0' && kern_level <= '7') {
218 memcpy(lvl, fmt, size);
219 lvl[size] = '\0';
220 type = logtypes[kern_level - '0'];
221 ratelimit = &printk_limits[kern_level - '0'];
222 }
223 fmt += size;
224 }
225
226 vaf.fmt = fmt;
227 vaf.va = &args;
228
229 if (__ratelimit(ratelimit))
230 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
231
232 va_end(args);
233 }
234 #endif
235
236 /*
237 * We only mark the transaction aborted and then set the file system read-only.
238 * This will prevent new transactions from starting or trying to join this
239 * one.
240 *
241 * This means that error recovery at the call site is limited to freeing
242 * any local memory allocations and passing the error code up without
243 * further cleanup. The transaction should complete as it normally would
244 * in the call path but will return -EIO.
245 *
246 * We'll complete the cleanup in btrfs_end_transaction and
247 * btrfs_commit_transaction.
248 */
249 __cold
250 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
251 const char *function,
252 unsigned int line, int errno)
253 {
254 struct btrfs_fs_info *fs_info = trans->fs_info;
255
256 trans->aborted = errno;
257 /* Nothing used. The other threads that have joined this
258 * transaction may be able to continue. */
259 if (!trans->dirty && list_empty(&trans->new_bgs)) {
260 const char *errstr;
261
262 errstr = btrfs_decode_error(errno);
263 btrfs_warn(fs_info,
264 "%s:%d: Aborting unused transaction(%s).",
265 function, line, errstr);
266 return;
267 }
268 ACCESS_ONCE(trans->transaction->aborted) = errno;
269 /* Wake up anybody who may be waiting on this transaction */
270 wake_up(&fs_info->transaction_wait);
271 wake_up(&fs_info->transaction_blocked_wait);
272 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
273 }
274 /*
275 * __btrfs_panic decodes unexpected, fatal errors from the caller,
276 * issues an alert, and either panics or BUGs, depending on mount options.
277 */
278 __cold
279 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
280 unsigned int line, int errno, const char *fmt, ...)
281 {
282 char *s_id = "<unknown>";
283 const char *errstr;
284 struct va_format vaf = { .fmt = fmt };
285 va_list args;
286
287 if (fs_info)
288 s_id = fs_info->sb->s_id;
289
290 va_start(args, fmt);
291 vaf.va = &args;
292
293 errstr = btrfs_decode_error(errno);
294 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
295 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
296 s_id, function, line, &vaf, errno, errstr);
297
298 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
299 function, line, &vaf, errno, errstr);
300 va_end(args);
301 /* Caller calls BUG() */
302 }
303
304 static void btrfs_put_super(struct super_block *sb)
305 {
306 close_ctree(btrfs_sb(sb)->tree_root);
307 }
308
309 enum {
310 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
311 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
312 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
313 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
314 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
315 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
316 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
317 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
318 Opt_skip_balance, Opt_check_integrity,
319 Opt_check_integrity_including_extent_data,
320 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
321 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
322 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
323 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
324 Opt_nologreplay, Opt_norecovery,
325 #ifdef CONFIG_BTRFS_DEBUG
326 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
327 #endif
328 Opt_err,
329 };
330
331 static const match_table_t tokens = {
332 {Opt_degraded, "degraded"},
333 {Opt_subvol, "subvol=%s"},
334 {Opt_subvolid, "subvolid=%s"},
335 {Opt_device, "device=%s"},
336 {Opt_nodatasum, "nodatasum"},
337 {Opt_datasum, "datasum"},
338 {Opt_nodatacow, "nodatacow"},
339 {Opt_datacow, "datacow"},
340 {Opt_nobarrier, "nobarrier"},
341 {Opt_barrier, "barrier"},
342 {Opt_max_inline, "max_inline=%s"},
343 {Opt_alloc_start, "alloc_start=%s"},
344 {Opt_thread_pool, "thread_pool=%d"},
345 {Opt_compress, "compress"},
346 {Opt_compress_type, "compress=%s"},
347 {Opt_compress_force, "compress-force"},
348 {Opt_compress_force_type, "compress-force=%s"},
349 {Opt_ssd, "ssd"},
350 {Opt_ssd_spread, "ssd_spread"},
351 {Opt_nossd, "nossd"},
352 {Opt_acl, "acl"},
353 {Opt_noacl, "noacl"},
354 {Opt_notreelog, "notreelog"},
355 {Opt_treelog, "treelog"},
356 {Opt_nologreplay, "nologreplay"},
357 {Opt_norecovery, "norecovery"},
358 {Opt_flushoncommit, "flushoncommit"},
359 {Opt_noflushoncommit, "noflushoncommit"},
360 {Opt_ratio, "metadata_ratio=%d"},
361 {Opt_discard, "discard"},
362 {Opt_nodiscard, "nodiscard"},
363 {Opt_space_cache, "space_cache"},
364 {Opt_space_cache_version, "space_cache=%s"},
365 {Opt_clear_cache, "clear_cache"},
366 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
367 {Opt_enospc_debug, "enospc_debug"},
368 {Opt_noenospc_debug, "noenospc_debug"},
369 {Opt_subvolrootid, "subvolrootid=%d"},
370 {Opt_defrag, "autodefrag"},
371 {Opt_nodefrag, "noautodefrag"},
372 {Opt_inode_cache, "inode_cache"},
373 {Opt_noinode_cache, "noinode_cache"},
374 {Opt_no_space_cache, "nospace_cache"},
375 {Opt_recovery, "recovery"}, /* deprecated */
376 {Opt_usebackuproot, "usebackuproot"},
377 {Opt_skip_balance, "skip_balance"},
378 {Opt_check_integrity, "check_int"},
379 {Opt_check_integrity_including_extent_data, "check_int_data"},
380 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
381 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
382 {Opt_fatal_errors, "fatal_errors=%s"},
383 {Opt_commit_interval, "commit=%d"},
384 #ifdef CONFIG_BTRFS_DEBUG
385 {Opt_fragment_data, "fragment=data"},
386 {Opt_fragment_metadata, "fragment=metadata"},
387 {Opt_fragment_all, "fragment=all"},
388 #endif
389 {Opt_err, NULL},
390 };
391
392 /*
393 * Regular mount options parser. Everything that is needed only when
394 * reading in a new superblock is parsed here.
395 * XXX JDM: This needs to be cleaned up for remount.
396 */
397 int btrfs_parse_options(struct btrfs_root *root, char *options,
398 unsigned long new_flags)
399 {
400 struct btrfs_fs_info *info = root->fs_info;
401 substring_t args[MAX_OPT_ARGS];
402 char *p, *num, *orig = NULL;
403 u64 cache_gen;
404 int intarg;
405 int ret = 0;
406 char *compress_type;
407 bool compress_force = false;
408 enum btrfs_compression_type saved_compress_type;
409 bool saved_compress_force;
410 int no_compress = 0;
411
412 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
413 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE))
414 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
415 else if (cache_gen)
416 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
417
418 /*
419 * Even the options are empty, we still need to do extra check
420 * against new flags
421 */
422 if (!options)
423 goto check;
424
425 /*
426 * strsep changes the string, duplicate it because parse_options
427 * gets called twice
428 */
429 options = kstrdup(options, GFP_NOFS);
430 if (!options)
431 return -ENOMEM;
432
433 orig = options;
434
435 while ((p = strsep(&options, ",")) != NULL) {
436 int token;
437 if (!*p)
438 continue;
439
440 token = match_token(p, tokens, args);
441 switch (token) {
442 case Opt_degraded:
443 btrfs_info(root->fs_info, "allowing degraded mounts");
444 btrfs_set_opt(info->mount_opt, DEGRADED);
445 break;
446 case Opt_subvol:
447 case Opt_subvolid:
448 case Opt_subvolrootid:
449 case Opt_device:
450 /*
451 * These are parsed by btrfs_parse_early_options
452 * and can be happily ignored here.
453 */
454 break;
455 case Opt_nodatasum:
456 btrfs_set_and_info(info, NODATASUM,
457 "setting nodatasum");
458 break;
459 case Opt_datasum:
460 if (btrfs_test_opt(info, NODATASUM)) {
461 if (btrfs_test_opt(info, NODATACOW))
462 btrfs_info(root->fs_info,
463 "setting datasum, datacow enabled");
464 else
465 btrfs_info(root->fs_info,
466 "setting datasum");
467 }
468 btrfs_clear_opt(info->mount_opt, NODATACOW);
469 btrfs_clear_opt(info->mount_opt, NODATASUM);
470 break;
471 case Opt_nodatacow:
472 if (!btrfs_test_opt(info, NODATACOW)) {
473 if (!btrfs_test_opt(info, COMPRESS) ||
474 !btrfs_test_opt(info, FORCE_COMPRESS)) {
475 btrfs_info(root->fs_info,
476 "setting nodatacow, compression disabled");
477 } else {
478 btrfs_info(root->fs_info,
479 "setting nodatacow");
480 }
481 }
482 btrfs_clear_opt(info->mount_opt, COMPRESS);
483 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
484 btrfs_set_opt(info->mount_opt, NODATACOW);
485 btrfs_set_opt(info->mount_opt, NODATASUM);
486 break;
487 case Opt_datacow:
488 btrfs_clear_and_info(info, NODATACOW,
489 "setting datacow");
490 break;
491 case Opt_compress_force:
492 case Opt_compress_force_type:
493 compress_force = true;
494 /* Fallthrough */
495 case Opt_compress:
496 case Opt_compress_type:
497 saved_compress_type = btrfs_test_opt(info,
498 COMPRESS) ?
499 info->compress_type : BTRFS_COMPRESS_NONE;
500 saved_compress_force =
501 btrfs_test_opt(info, FORCE_COMPRESS);
502 if (token == Opt_compress ||
503 token == Opt_compress_force ||
504 strcmp(args[0].from, "zlib") == 0) {
505 compress_type = "zlib";
506 info->compress_type = BTRFS_COMPRESS_ZLIB;
507 btrfs_set_opt(info->mount_opt, COMPRESS);
508 btrfs_clear_opt(info->mount_opt, NODATACOW);
509 btrfs_clear_opt(info->mount_opt, NODATASUM);
510 no_compress = 0;
511 } else if (strcmp(args[0].from, "lzo") == 0) {
512 compress_type = "lzo";
513 info->compress_type = BTRFS_COMPRESS_LZO;
514 btrfs_set_opt(info->mount_opt, COMPRESS);
515 btrfs_clear_opt(info->mount_opt, NODATACOW);
516 btrfs_clear_opt(info->mount_opt, NODATASUM);
517 btrfs_set_fs_incompat(info, COMPRESS_LZO);
518 no_compress = 0;
519 } else if (strncmp(args[0].from, "no", 2) == 0) {
520 compress_type = "no";
521 btrfs_clear_opt(info->mount_opt, COMPRESS);
522 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
523 compress_force = false;
524 no_compress++;
525 } else {
526 ret = -EINVAL;
527 goto out;
528 }
529
530 if (compress_force) {
531 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
532 } else {
533 /*
534 * If we remount from compress-force=xxx to
535 * compress=xxx, we need clear FORCE_COMPRESS
536 * flag, otherwise, there is no way for users
537 * to disable forcible compression separately.
538 */
539 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
540 }
541 if ((btrfs_test_opt(info, COMPRESS) &&
542 (info->compress_type != saved_compress_type ||
543 compress_force != saved_compress_force)) ||
544 (!btrfs_test_opt(info, COMPRESS) &&
545 no_compress == 1)) {
546 btrfs_info(root->fs_info,
547 "%s %s compression",
548 (compress_force) ? "force" : "use",
549 compress_type);
550 }
551 compress_force = false;
552 break;
553 case Opt_ssd:
554 btrfs_set_and_info(info, SSD,
555 "use ssd allocation scheme");
556 break;
557 case Opt_ssd_spread:
558 btrfs_set_and_info(info, SSD_SPREAD,
559 "use spread ssd allocation scheme");
560 btrfs_set_opt(info->mount_opt, SSD);
561 break;
562 case Opt_nossd:
563 btrfs_set_and_info(info, NOSSD,
564 "not using ssd allocation scheme");
565 btrfs_clear_opt(info->mount_opt, SSD);
566 break;
567 case Opt_barrier:
568 btrfs_clear_and_info(info, NOBARRIER,
569 "turning on barriers");
570 break;
571 case Opt_nobarrier:
572 btrfs_set_and_info(info, NOBARRIER,
573 "turning off barriers");
574 break;
575 case Opt_thread_pool:
576 ret = match_int(&args[0], &intarg);
577 if (ret) {
578 goto out;
579 } else if (intarg > 0) {
580 info->thread_pool_size = intarg;
581 } else {
582 ret = -EINVAL;
583 goto out;
584 }
585 break;
586 case Opt_max_inline:
587 num = match_strdup(&args[0]);
588 if (num) {
589 info->max_inline = memparse(num, NULL);
590 kfree(num);
591
592 if (info->max_inline) {
593 info->max_inline = min_t(u64,
594 info->max_inline,
595 root->sectorsize);
596 }
597 btrfs_info(root->fs_info, "max_inline at %llu",
598 info->max_inline);
599 } else {
600 ret = -ENOMEM;
601 goto out;
602 }
603 break;
604 case Opt_alloc_start:
605 num = match_strdup(&args[0]);
606 if (num) {
607 mutex_lock(&info->chunk_mutex);
608 info->alloc_start = memparse(num, NULL);
609 mutex_unlock(&info->chunk_mutex);
610 kfree(num);
611 btrfs_info(root->fs_info,
612 "allocations start at %llu",
613 info->alloc_start);
614 } else {
615 ret = -ENOMEM;
616 goto out;
617 }
618 break;
619 case Opt_acl:
620 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
621 root->fs_info->sb->s_flags |= MS_POSIXACL;
622 break;
623 #else
624 btrfs_err(root->fs_info,
625 "support for ACL not compiled in!");
626 ret = -EINVAL;
627 goto out;
628 #endif
629 case Opt_noacl:
630 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
631 break;
632 case Opt_notreelog:
633 btrfs_set_and_info(info, NOTREELOG,
634 "disabling tree log");
635 break;
636 case Opt_treelog:
637 btrfs_clear_and_info(info, NOTREELOG,
638 "enabling tree log");
639 break;
640 case Opt_norecovery:
641 case Opt_nologreplay:
642 btrfs_set_and_info(info, NOLOGREPLAY,
643 "disabling log replay at mount time");
644 break;
645 case Opt_flushoncommit:
646 btrfs_set_and_info(info, FLUSHONCOMMIT,
647 "turning on flush-on-commit");
648 break;
649 case Opt_noflushoncommit:
650 btrfs_clear_and_info(info, FLUSHONCOMMIT,
651 "turning off flush-on-commit");
652 break;
653 case Opt_ratio:
654 ret = match_int(&args[0], &intarg);
655 if (ret) {
656 goto out;
657 } else if (intarg >= 0) {
658 info->metadata_ratio = intarg;
659 btrfs_info(root->fs_info, "metadata ratio %d",
660 info->metadata_ratio);
661 } else {
662 ret = -EINVAL;
663 goto out;
664 }
665 break;
666 case Opt_discard:
667 btrfs_set_and_info(info, DISCARD,
668 "turning on discard");
669 break;
670 case Opt_nodiscard:
671 btrfs_clear_and_info(info, DISCARD,
672 "turning off discard");
673 break;
674 case Opt_space_cache:
675 case Opt_space_cache_version:
676 if (token == Opt_space_cache ||
677 strcmp(args[0].from, "v1") == 0) {
678 btrfs_clear_opt(root->fs_info->mount_opt,
679 FREE_SPACE_TREE);
680 btrfs_set_and_info(info, SPACE_CACHE,
681 "enabling disk space caching");
682 } else if (strcmp(args[0].from, "v2") == 0) {
683 btrfs_clear_opt(root->fs_info->mount_opt,
684 SPACE_CACHE);
685 btrfs_set_and_info(info,
686 FREE_SPACE_TREE,
687 "enabling free space tree");
688 } else {
689 ret = -EINVAL;
690 goto out;
691 }
692 break;
693 case Opt_rescan_uuid_tree:
694 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
695 break;
696 case Opt_no_space_cache:
697 if (btrfs_test_opt(info, SPACE_CACHE)) {
698 btrfs_clear_and_info(info,
699 SPACE_CACHE,
700 "disabling disk space caching");
701 }
702 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
703 btrfs_clear_and_info(info,
704 FREE_SPACE_TREE,
705 "disabling free space tree");
706 }
707 break;
708 case Opt_inode_cache:
709 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
710 "enabling inode map caching");
711 break;
712 case Opt_noinode_cache:
713 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
714 "disabling inode map caching");
715 break;
716 case Opt_clear_cache:
717 btrfs_set_and_info(info, CLEAR_CACHE,
718 "force clearing of disk cache");
719 break;
720 case Opt_user_subvol_rm_allowed:
721 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
722 break;
723 case Opt_enospc_debug:
724 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
725 break;
726 case Opt_noenospc_debug:
727 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
728 break;
729 case Opt_defrag:
730 btrfs_set_and_info(info, AUTO_DEFRAG,
731 "enabling auto defrag");
732 break;
733 case Opt_nodefrag:
734 btrfs_clear_and_info(info, AUTO_DEFRAG,
735 "disabling auto defrag");
736 break;
737 case Opt_recovery:
738 btrfs_warn(root->fs_info,
739 "'recovery' is deprecated, use 'usebackuproot' instead");
740 case Opt_usebackuproot:
741 btrfs_info(root->fs_info,
742 "trying to use backup root at mount time");
743 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
744 break;
745 case Opt_skip_balance:
746 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
747 break;
748 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
749 case Opt_check_integrity_including_extent_data:
750 btrfs_info(root->fs_info,
751 "enabling check integrity including extent data");
752 btrfs_set_opt(info->mount_opt,
753 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
754 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
755 break;
756 case Opt_check_integrity:
757 btrfs_info(root->fs_info, "enabling check integrity");
758 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
759 break;
760 case Opt_check_integrity_print_mask:
761 ret = match_int(&args[0], &intarg);
762 if (ret) {
763 goto out;
764 } else if (intarg >= 0) {
765 info->check_integrity_print_mask = intarg;
766 btrfs_info(root->fs_info,
767 "check_integrity_print_mask 0x%x",
768 info->check_integrity_print_mask);
769 } else {
770 ret = -EINVAL;
771 goto out;
772 }
773 break;
774 #else
775 case Opt_check_integrity_including_extent_data:
776 case Opt_check_integrity:
777 case Opt_check_integrity_print_mask:
778 btrfs_err(root->fs_info,
779 "support for check_integrity* not compiled in!");
780 ret = -EINVAL;
781 goto out;
782 #endif
783 case Opt_fatal_errors:
784 if (strcmp(args[0].from, "panic") == 0)
785 btrfs_set_opt(info->mount_opt,
786 PANIC_ON_FATAL_ERROR);
787 else if (strcmp(args[0].from, "bug") == 0)
788 btrfs_clear_opt(info->mount_opt,
789 PANIC_ON_FATAL_ERROR);
790 else {
791 ret = -EINVAL;
792 goto out;
793 }
794 break;
795 case Opt_commit_interval:
796 intarg = 0;
797 ret = match_int(&args[0], &intarg);
798 if (ret < 0) {
799 btrfs_err(root->fs_info,
800 "invalid commit interval");
801 ret = -EINVAL;
802 goto out;
803 }
804 if (intarg > 0) {
805 if (intarg > 300) {
806 btrfs_warn(root->fs_info,
807 "excessive commit interval %d",
808 intarg);
809 }
810 info->commit_interval = intarg;
811 } else {
812 btrfs_info(root->fs_info,
813 "using default commit interval %ds",
814 BTRFS_DEFAULT_COMMIT_INTERVAL);
815 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
816 }
817 break;
818 #ifdef CONFIG_BTRFS_DEBUG
819 case Opt_fragment_all:
820 btrfs_info(root->fs_info, "fragmenting all space");
821 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
822 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
823 break;
824 case Opt_fragment_metadata:
825 btrfs_info(root->fs_info, "fragmenting metadata");
826 btrfs_set_opt(info->mount_opt,
827 FRAGMENT_METADATA);
828 break;
829 case Opt_fragment_data:
830 btrfs_info(root->fs_info, "fragmenting data");
831 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
832 break;
833 #endif
834 case Opt_err:
835 btrfs_info(root->fs_info,
836 "unrecognized mount option '%s'", p);
837 ret = -EINVAL;
838 goto out;
839 default:
840 break;
841 }
842 }
843 check:
844 /*
845 * Extra check for current option against current flag
846 */
847 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
848 btrfs_err(root->fs_info,
849 "nologreplay must be used with ro mount option");
850 ret = -EINVAL;
851 }
852 out:
853 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE) &&
854 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
855 !btrfs_test_opt(info, CLEAR_CACHE)) {
856 btrfs_err(root->fs_info, "cannot disable free space tree");
857 ret = -EINVAL;
858
859 }
860 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
861 btrfs_info(root->fs_info, "disk space caching is enabled");
862 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
863 btrfs_info(root->fs_info, "using free space tree");
864 kfree(orig);
865 return ret;
866 }
867
868 /*
869 * Parse mount options that are required early in the mount process.
870 *
871 * All other options will be parsed on much later in the mount process and
872 * only when we need to allocate a new super block.
873 */
874 static int btrfs_parse_early_options(const char *options, fmode_t flags,
875 void *holder, char **subvol_name, u64 *subvol_objectid,
876 struct btrfs_fs_devices **fs_devices)
877 {
878 substring_t args[MAX_OPT_ARGS];
879 char *device_name, *opts, *orig, *p;
880 char *num = NULL;
881 int error = 0;
882
883 if (!options)
884 return 0;
885
886 /*
887 * strsep changes the string, duplicate it because parse_options
888 * gets called twice
889 */
890 opts = kstrdup(options, GFP_KERNEL);
891 if (!opts)
892 return -ENOMEM;
893 orig = opts;
894
895 while ((p = strsep(&opts, ",")) != NULL) {
896 int token;
897 if (!*p)
898 continue;
899
900 token = match_token(p, tokens, args);
901 switch (token) {
902 case Opt_subvol:
903 kfree(*subvol_name);
904 *subvol_name = match_strdup(&args[0]);
905 if (!*subvol_name) {
906 error = -ENOMEM;
907 goto out;
908 }
909 break;
910 case Opt_subvolid:
911 num = match_strdup(&args[0]);
912 if (num) {
913 *subvol_objectid = memparse(num, NULL);
914 kfree(num);
915 /* we want the original fs_tree */
916 if (!*subvol_objectid)
917 *subvol_objectid =
918 BTRFS_FS_TREE_OBJECTID;
919 } else {
920 error = -EINVAL;
921 goto out;
922 }
923 break;
924 case Opt_subvolrootid:
925 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
926 break;
927 case Opt_device:
928 device_name = match_strdup(&args[0]);
929 if (!device_name) {
930 error = -ENOMEM;
931 goto out;
932 }
933 error = btrfs_scan_one_device(device_name,
934 flags, holder, fs_devices);
935 kfree(device_name);
936 if (error)
937 goto out;
938 break;
939 default:
940 break;
941 }
942 }
943
944 out:
945 kfree(orig);
946 return error;
947 }
948
949 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
950 u64 subvol_objectid)
951 {
952 struct btrfs_root *root = fs_info->tree_root;
953 struct btrfs_root *fs_root;
954 struct btrfs_root_ref *root_ref;
955 struct btrfs_inode_ref *inode_ref;
956 struct btrfs_key key;
957 struct btrfs_path *path = NULL;
958 char *name = NULL, *ptr;
959 u64 dirid;
960 int len;
961 int ret;
962
963 path = btrfs_alloc_path();
964 if (!path) {
965 ret = -ENOMEM;
966 goto err;
967 }
968 path->leave_spinning = 1;
969
970 name = kmalloc(PATH_MAX, GFP_NOFS);
971 if (!name) {
972 ret = -ENOMEM;
973 goto err;
974 }
975 ptr = name + PATH_MAX - 1;
976 ptr[0] = '\0';
977
978 /*
979 * Walk up the subvolume trees in the tree of tree roots by root
980 * backrefs until we hit the top-level subvolume.
981 */
982 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
983 key.objectid = subvol_objectid;
984 key.type = BTRFS_ROOT_BACKREF_KEY;
985 key.offset = (u64)-1;
986
987 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
988 if (ret < 0) {
989 goto err;
990 } else if (ret > 0) {
991 ret = btrfs_previous_item(root, path, subvol_objectid,
992 BTRFS_ROOT_BACKREF_KEY);
993 if (ret < 0) {
994 goto err;
995 } else if (ret > 0) {
996 ret = -ENOENT;
997 goto err;
998 }
999 }
1000
1001 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1002 subvol_objectid = key.offset;
1003
1004 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1005 struct btrfs_root_ref);
1006 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1007 ptr -= len + 1;
1008 if (ptr < name) {
1009 ret = -ENAMETOOLONG;
1010 goto err;
1011 }
1012 read_extent_buffer(path->nodes[0], ptr + 1,
1013 (unsigned long)(root_ref + 1), len);
1014 ptr[0] = '/';
1015 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1016 btrfs_release_path(path);
1017
1018 key.objectid = subvol_objectid;
1019 key.type = BTRFS_ROOT_ITEM_KEY;
1020 key.offset = (u64)-1;
1021 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1022 if (IS_ERR(fs_root)) {
1023 ret = PTR_ERR(fs_root);
1024 goto err;
1025 }
1026
1027 /*
1028 * Walk up the filesystem tree by inode refs until we hit the
1029 * root directory.
1030 */
1031 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1032 key.objectid = dirid;
1033 key.type = BTRFS_INODE_REF_KEY;
1034 key.offset = (u64)-1;
1035
1036 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1037 if (ret < 0) {
1038 goto err;
1039 } else if (ret > 0) {
1040 ret = btrfs_previous_item(fs_root, path, dirid,
1041 BTRFS_INODE_REF_KEY);
1042 if (ret < 0) {
1043 goto err;
1044 } else if (ret > 0) {
1045 ret = -ENOENT;
1046 goto err;
1047 }
1048 }
1049
1050 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1051 dirid = key.offset;
1052
1053 inode_ref = btrfs_item_ptr(path->nodes[0],
1054 path->slots[0],
1055 struct btrfs_inode_ref);
1056 len = btrfs_inode_ref_name_len(path->nodes[0],
1057 inode_ref);
1058 ptr -= len + 1;
1059 if (ptr < name) {
1060 ret = -ENAMETOOLONG;
1061 goto err;
1062 }
1063 read_extent_buffer(path->nodes[0], ptr + 1,
1064 (unsigned long)(inode_ref + 1), len);
1065 ptr[0] = '/';
1066 btrfs_release_path(path);
1067 }
1068 }
1069
1070 btrfs_free_path(path);
1071 if (ptr == name + PATH_MAX - 1) {
1072 name[0] = '/';
1073 name[1] = '\0';
1074 } else {
1075 memmove(name, ptr, name + PATH_MAX - ptr);
1076 }
1077 return name;
1078
1079 err:
1080 btrfs_free_path(path);
1081 kfree(name);
1082 return ERR_PTR(ret);
1083 }
1084
1085 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1086 {
1087 struct btrfs_root *root = fs_info->tree_root;
1088 struct btrfs_dir_item *di;
1089 struct btrfs_path *path;
1090 struct btrfs_key location;
1091 u64 dir_id;
1092
1093 path = btrfs_alloc_path();
1094 if (!path)
1095 return -ENOMEM;
1096 path->leave_spinning = 1;
1097
1098 /*
1099 * Find the "default" dir item which points to the root item that we
1100 * will mount by default if we haven't been given a specific subvolume
1101 * to mount.
1102 */
1103 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1104 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1105 if (IS_ERR(di)) {
1106 btrfs_free_path(path);
1107 return PTR_ERR(di);
1108 }
1109 if (!di) {
1110 /*
1111 * Ok the default dir item isn't there. This is weird since
1112 * it's always been there, but don't freak out, just try and
1113 * mount the top-level subvolume.
1114 */
1115 btrfs_free_path(path);
1116 *objectid = BTRFS_FS_TREE_OBJECTID;
1117 return 0;
1118 }
1119
1120 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1121 btrfs_free_path(path);
1122 *objectid = location.objectid;
1123 return 0;
1124 }
1125
1126 static int btrfs_fill_super(struct super_block *sb,
1127 struct btrfs_fs_devices *fs_devices,
1128 void *data, int silent)
1129 {
1130 struct inode *inode;
1131 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1132 struct btrfs_key key;
1133 int err;
1134
1135 sb->s_maxbytes = MAX_LFS_FILESIZE;
1136 sb->s_magic = BTRFS_SUPER_MAGIC;
1137 sb->s_op = &btrfs_super_ops;
1138 sb->s_d_op = &btrfs_dentry_operations;
1139 sb->s_export_op = &btrfs_export_ops;
1140 sb->s_xattr = btrfs_xattr_handlers;
1141 sb->s_time_gran = 1;
1142 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1143 sb->s_flags |= MS_POSIXACL;
1144 #endif
1145 sb->s_flags |= MS_I_VERSION;
1146 sb->s_iflags |= SB_I_CGROUPWB;
1147 err = open_ctree(sb, fs_devices, (char *)data);
1148 if (err) {
1149 btrfs_err(fs_info, "open_ctree failed");
1150 return err;
1151 }
1152
1153 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1154 key.type = BTRFS_INODE_ITEM_KEY;
1155 key.offset = 0;
1156 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1157 if (IS_ERR(inode)) {
1158 err = PTR_ERR(inode);
1159 goto fail_close;
1160 }
1161
1162 sb->s_root = d_make_root(inode);
1163 if (!sb->s_root) {
1164 err = -ENOMEM;
1165 goto fail_close;
1166 }
1167
1168 save_mount_options(sb, data);
1169 cleancache_init_fs(sb);
1170 sb->s_flags |= MS_ACTIVE;
1171 return 0;
1172
1173 fail_close:
1174 close_ctree(fs_info->tree_root);
1175 return err;
1176 }
1177
1178 int btrfs_sync_fs(struct super_block *sb, int wait)
1179 {
1180 struct btrfs_trans_handle *trans;
1181 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1182 struct btrfs_root *root = fs_info->tree_root;
1183
1184 trace_btrfs_sync_fs(fs_info, wait);
1185
1186 if (!wait) {
1187 filemap_flush(fs_info->btree_inode->i_mapping);
1188 return 0;
1189 }
1190
1191 btrfs_wait_ordered_roots(fs_info, -1, 0, (u64)-1);
1192
1193 trans = btrfs_attach_transaction_barrier(root);
1194 if (IS_ERR(trans)) {
1195 /* no transaction, don't bother */
1196 if (PTR_ERR(trans) == -ENOENT) {
1197 /*
1198 * Exit unless we have some pending changes
1199 * that need to go through commit
1200 */
1201 if (fs_info->pending_changes == 0)
1202 return 0;
1203 /*
1204 * A non-blocking test if the fs is frozen. We must not
1205 * start a new transaction here otherwise a deadlock
1206 * happens. The pending operations are delayed to the
1207 * next commit after thawing.
1208 */
1209 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1210 __sb_end_write(sb, SB_FREEZE_WRITE);
1211 else
1212 return 0;
1213 trans = btrfs_start_transaction(root, 0);
1214 }
1215 if (IS_ERR(trans))
1216 return PTR_ERR(trans);
1217 }
1218 return btrfs_commit_transaction(trans, root);
1219 }
1220
1221 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1222 {
1223 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1224 struct btrfs_root *root = info->tree_root;
1225 char *compress_type;
1226
1227 if (btrfs_test_opt(info, DEGRADED))
1228 seq_puts(seq, ",degraded");
1229 if (btrfs_test_opt(info, NODATASUM))
1230 seq_puts(seq, ",nodatasum");
1231 if (btrfs_test_opt(info, NODATACOW))
1232 seq_puts(seq, ",nodatacow");
1233 if (btrfs_test_opt(info, NOBARRIER))
1234 seq_puts(seq, ",nobarrier");
1235 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1236 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1237 if (info->alloc_start != 0)
1238 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1239 if (info->thread_pool_size != min_t(unsigned long,
1240 num_online_cpus() + 2, 8))
1241 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1242 if (btrfs_test_opt(info, COMPRESS)) {
1243 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1244 compress_type = "zlib";
1245 else
1246 compress_type = "lzo";
1247 if (btrfs_test_opt(info, FORCE_COMPRESS))
1248 seq_printf(seq, ",compress-force=%s", compress_type);
1249 else
1250 seq_printf(seq, ",compress=%s", compress_type);
1251 }
1252 if (btrfs_test_opt(info, NOSSD))
1253 seq_puts(seq, ",nossd");
1254 if (btrfs_test_opt(info, SSD_SPREAD))
1255 seq_puts(seq, ",ssd_spread");
1256 else if (btrfs_test_opt(info, SSD))
1257 seq_puts(seq, ",ssd");
1258 if (btrfs_test_opt(info, NOTREELOG))
1259 seq_puts(seq, ",notreelog");
1260 if (btrfs_test_opt(info, NOLOGREPLAY))
1261 seq_puts(seq, ",nologreplay");
1262 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1263 seq_puts(seq, ",flushoncommit");
1264 if (btrfs_test_opt(info, DISCARD))
1265 seq_puts(seq, ",discard");
1266 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1267 seq_puts(seq, ",noacl");
1268 if (btrfs_test_opt(info, SPACE_CACHE))
1269 seq_puts(seq, ",space_cache");
1270 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1271 seq_puts(seq, ",space_cache=v2");
1272 else
1273 seq_puts(seq, ",nospace_cache");
1274 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1275 seq_puts(seq, ",rescan_uuid_tree");
1276 if (btrfs_test_opt(info, CLEAR_CACHE))
1277 seq_puts(seq, ",clear_cache");
1278 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1279 seq_puts(seq, ",user_subvol_rm_allowed");
1280 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1281 seq_puts(seq, ",enospc_debug");
1282 if (btrfs_test_opt(info, AUTO_DEFRAG))
1283 seq_puts(seq, ",autodefrag");
1284 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1285 seq_puts(seq, ",inode_cache");
1286 if (btrfs_test_opt(info, SKIP_BALANCE))
1287 seq_puts(seq, ",skip_balance");
1288 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1289 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1290 seq_puts(seq, ",check_int_data");
1291 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1292 seq_puts(seq, ",check_int");
1293 if (info->check_integrity_print_mask)
1294 seq_printf(seq, ",check_int_print_mask=%d",
1295 info->check_integrity_print_mask);
1296 #endif
1297 if (info->metadata_ratio)
1298 seq_printf(seq, ",metadata_ratio=%d",
1299 info->metadata_ratio);
1300 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1301 seq_puts(seq, ",fatal_errors=panic");
1302 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1303 seq_printf(seq, ",commit=%d", info->commit_interval);
1304 #ifdef CONFIG_BTRFS_DEBUG
1305 if (btrfs_test_opt(info, FRAGMENT_DATA))
1306 seq_puts(seq, ",fragment=data");
1307 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1308 seq_puts(seq, ",fragment=metadata");
1309 #endif
1310 seq_printf(seq, ",subvolid=%llu",
1311 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1312 seq_puts(seq, ",subvol=");
1313 seq_dentry(seq, dentry, " \t\n\\");
1314 return 0;
1315 }
1316
1317 static int btrfs_test_super(struct super_block *s, void *data)
1318 {
1319 struct btrfs_fs_info *p = data;
1320 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1321
1322 return fs_info->fs_devices == p->fs_devices;
1323 }
1324
1325 static int btrfs_set_super(struct super_block *s, void *data)
1326 {
1327 int err = set_anon_super(s, data);
1328 if (!err)
1329 s->s_fs_info = data;
1330 return err;
1331 }
1332
1333 /*
1334 * subvolumes are identified by ino 256
1335 */
1336 static inline int is_subvolume_inode(struct inode *inode)
1337 {
1338 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1339 return 1;
1340 return 0;
1341 }
1342
1343 /*
1344 * This will add subvolid=0 to the argument string while removing any subvol=
1345 * and subvolid= arguments to make sure we get the top-level root for path
1346 * walking to the subvol we want.
1347 */
1348 static char *setup_root_args(char *args)
1349 {
1350 char *buf, *dst, *sep;
1351
1352 if (!args)
1353 return kstrdup("subvolid=0", GFP_NOFS);
1354
1355 /* The worst case is that we add ",subvolid=0" to the end. */
1356 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
1357 if (!buf)
1358 return NULL;
1359
1360 while (1) {
1361 sep = strchrnul(args, ',');
1362 if (!strstarts(args, "subvol=") &&
1363 !strstarts(args, "subvolid=")) {
1364 memcpy(dst, args, sep - args);
1365 dst += sep - args;
1366 *dst++ = ',';
1367 }
1368 if (*sep)
1369 args = sep + 1;
1370 else
1371 break;
1372 }
1373 strcpy(dst, "subvolid=0");
1374
1375 return buf;
1376 }
1377
1378 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1379 int flags, const char *device_name,
1380 char *data)
1381 {
1382 struct dentry *root;
1383 struct vfsmount *mnt = NULL;
1384 char *newargs;
1385 int ret;
1386
1387 newargs = setup_root_args(data);
1388 if (!newargs) {
1389 root = ERR_PTR(-ENOMEM);
1390 goto out;
1391 }
1392
1393 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
1394 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
1395 if (flags & MS_RDONLY) {
1396 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
1397 device_name, newargs);
1398 } else {
1399 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
1400 device_name, newargs);
1401 if (IS_ERR(mnt)) {
1402 root = ERR_CAST(mnt);
1403 mnt = NULL;
1404 goto out;
1405 }
1406
1407 down_write(&mnt->mnt_sb->s_umount);
1408 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1409 up_write(&mnt->mnt_sb->s_umount);
1410 if (ret < 0) {
1411 root = ERR_PTR(ret);
1412 goto out;
1413 }
1414 }
1415 }
1416 if (IS_ERR(mnt)) {
1417 root = ERR_CAST(mnt);
1418 mnt = NULL;
1419 goto out;
1420 }
1421
1422 if (!subvol_name) {
1423 if (!subvol_objectid) {
1424 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1425 &subvol_objectid);
1426 if (ret) {
1427 root = ERR_PTR(ret);
1428 goto out;
1429 }
1430 }
1431 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1432 subvol_objectid);
1433 if (IS_ERR(subvol_name)) {
1434 root = ERR_CAST(subvol_name);
1435 subvol_name = NULL;
1436 goto out;
1437 }
1438
1439 }
1440
1441 root = mount_subtree(mnt, subvol_name);
1442 /* mount_subtree() drops our reference on the vfsmount. */
1443 mnt = NULL;
1444
1445 if (!IS_ERR(root)) {
1446 struct super_block *s = root->d_sb;
1447 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1448 struct inode *root_inode = d_inode(root);
1449 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1450
1451 ret = 0;
1452 if (!is_subvolume_inode(root_inode)) {
1453 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1454 subvol_name);
1455 ret = -EINVAL;
1456 }
1457 if (subvol_objectid && root_objectid != subvol_objectid) {
1458 /*
1459 * This will also catch a race condition where a
1460 * subvolume which was passed by ID is renamed and
1461 * another subvolume is renamed over the old location.
1462 */
1463 btrfs_err(fs_info,
1464 "subvol '%s' does not match subvolid %llu",
1465 subvol_name, subvol_objectid);
1466 ret = -EINVAL;
1467 }
1468 if (ret) {
1469 dput(root);
1470 root = ERR_PTR(ret);
1471 deactivate_locked_super(s);
1472 }
1473 }
1474
1475 out:
1476 mntput(mnt);
1477 kfree(newargs);
1478 kfree(subvol_name);
1479 return root;
1480 }
1481
1482 static int parse_security_options(char *orig_opts,
1483 struct security_mnt_opts *sec_opts)
1484 {
1485 char *secdata = NULL;
1486 int ret = 0;
1487
1488 secdata = alloc_secdata();
1489 if (!secdata)
1490 return -ENOMEM;
1491 ret = security_sb_copy_data(orig_opts, secdata);
1492 if (ret) {
1493 free_secdata(secdata);
1494 return ret;
1495 }
1496 ret = security_sb_parse_opts_str(secdata, sec_opts);
1497 free_secdata(secdata);
1498 return ret;
1499 }
1500
1501 static int setup_security_options(struct btrfs_fs_info *fs_info,
1502 struct super_block *sb,
1503 struct security_mnt_opts *sec_opts)
1504 {
1505 int ret = 0;
1506
1507 /*
1508 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1509 * is valid.
1510 */
1511 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1512 if (ret)
1513 return ret;
1514
1515 #ifdef CONFIG_SECURITY
1516 if (!fs_info->security_opts.num_mnt_opts) {
1517 /* first time security setup, copy sec_opts to fs_info */
1518 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1519 } else {
1520 /*
1521 * Since SELinux (the only one supporting security_mnt_opts)
1522 * does NOT support changing context during remount/mount of
1523 * the same sb, this must be the same or part of the same
1524 * security options, just free it.
1525 */
1526 security_free_mnt_opts(sec_opts);
1527 }
1528 #endif
1529 return ret;
1530 }
1531
1532 /*
1533 * Find a superblock for the given device / mount point.
1534 *
1535 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1536 * for multiple device setup. Make sure to keep it in sync.
1537 */
1538 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1539 const char *device_name, void *data)
1540 {
1541 struct block_device *bdev = NULL;
1542 struct super_block *s;
1543 struct btrfs_fs_devices *fs_devices = NULL;
1544 struct btrfs_fs_info *fs_info = NULL;
1545 struct security_mnt_opts new_sec_opts;
1546 fmode_t mode = FMODE_READ;
1547 char *subvol_name = NULL;
1548 u64 subvol_objectid = 0;
1549 int error = 0;
1550
1551 if (!(flags & MS_RDONLY))
1552 mode |= FMODE_WRITE;
1553
1554 error = btrfs_parse_early_options(data, mode, fs_type,
1555 &subvol_name, &subvol_objectid,
1556 &fs_devices);
1557 if (error) {
1558 kfree(subvol_name);
1559 return ERR_PTR(error);
1560 }
1561
1562 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1563 /* mount_subvol() will free subvol_name. */
1564 return mount_subvol(subvol_name, subvol_objectid, flags,
1565 device_name, data);
1566 }
1567
1568 security_init_mnt_opts(&new_sec_opts);
1569 if (data) {
1570 error = parse_security_options(data, &new_sec_opts);
1571 if (error)
1572 return ERR_PTR(error);
1573 }
1574
1575 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1576 if (error)
1577 goto error_sec_opts;
1578
1579 /*
1580 * Setup a dummy root and fs_info for test/set super. This is because
1581 * we don't actually fill this stuff out until open_ctree, but we need
1582 * it for searching for existing supers, so this lets us do that and
1583 * then open_ctree will properly initialize everything later.
1584 */
1585 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1586 if (!fs_info) {
1587 error = -ENOMEM;
1588 goto error_sec_opts;
1589 }
1590
1591 fs_info->fs_devices = fs_devices;
1592
1593 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1594 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1595 security_init_mnt_opts(&fs_info->security_opts);
1596 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1597 error = -ENOMEM;
1598 goto error_fs_info;
1599 }
1600
1601 error = btrfs_open_devices(fs_devices, mode, fs_type);
1602 if (error)
1603 goto error_fs_info;
1604
1605 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1606 error = -EACCES;
1607 goto error_close_devices;
1608 }
1609
1610 bdev = fs_devices->latest_bdev;
1611 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1612 fs_info);
1613 if (IS_ERR(s)) {
1614 error = PTR_ERR(s);
1615 goto error_close_devices;
1616 }
1617
1618 if (s->s_root) {
1619 btrfs_close_devices(fs_devices);
1620 free_fs_info(fs_info);
1621 if ((flags ^ s->s_flags) & MS_RDONLY)
1622 error = -EBUSY;
1623 } else {
1624 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1625 btrfs_sb(s)->bdev_holder = fs_type;
1626 error = btrfs_fill_super(s, fs_devices, data,
1627 flags & MS_SILENT ? 1 : 0);
1628 }
1629 if (error) {
1630 deactivate_locked_super(s);
1631 goto error_sec_opts;
1632 }
1633
1634 fs_info = btrfs_sb(s);
1635 error = setup_security_options(fs_info, s, &new_sec_opts);
1636 if (error) {
1637 deactivate_locked_super(s);
1638 goto error_sec_opts;
1639 }
1640
1641 return dget(s->s_root);
1642
1643 error_close_devices:
1644 btrfs_close_devices(fs_devices);
1645 error_fs_info:
1646 free_fs_info(fs_info);
1647 error_sec_opts:
1648 security_free_mnt_opts(&new_sec_opts);
1649 return ERR_PTR(error);
1650 }
1651
1652 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1653 int new_pool_size, int old_pool_size)
1654 {
1655 if (new_pool_size == old_pool_size)
1656 return;
1657
1658 fs_info->thread_pool_size = new_pool_size;
1659
1660 btrfs_info(fs_info, "resize thread pool %d -> %d",
1661 old_pool_size, new_pool_size);
1662
1663 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1664 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1665 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1666 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1667 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1668 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1669 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1670 new_pool_size);
1671 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1672 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1673 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1674 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1675 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1676 new_pool_size);
1677 }
1678
1679 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1680 {
1681 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1682 }
1683
1684 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1685 unsigned long old_opts, int flags)
1686 {
1687 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1688 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1689 (flags & MS_RDONLY))) {
1690 /* wait for any defraggers to finish */
1691 wait_event(fs_info->transaction_wait,
1692 (atomic_read(&fs_info->defrag_running) == 0));
1693 if (flags & MS_RDONLY)
1694 sync_filesystem(fs_info->sb);
1695 }
1696 }
1697
1698 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1699 unsigned long old_opts)
1700 {
1701 /*
1702 * We need to cleanup all defragable inodes if the autodefragment is
1703 * close or the filesystem is read only.
1704 */
1705 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1706 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1707 (fs_info->sb->s_flags & MS_RDONLY))) {
1708 btrfs_cleanup_defrag_inodes(fs_info);
1709 }
1710
1711 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1712 }
1713
1714 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1715 {
1716 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1717 struct btrfs_root *root = fs_info->tree_root;
1718 unsigned old_flags = sb->s_flags;
1719 unsigned long old_opts = fs_info->mount_opt;
1720 unsigned long old_compress_type = fs_info->compress_type;
1721 u64 old_max_inline = fs_info->max_inline;
1722 u64 old_alloc_start = fs_info->alloc_start;
1723 int old_thread_pool_size = fs_info->thread_pool_size;
1724 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1725 int ret;
1726
1727 sync_filesystem(sb);
1728 btrfs_remount_prepare(fs_info);
1729
1730 if (data) {
1731 struct security_mnt_opts new_sec_opts;
1732
1733 security_init_mnt_opts(&new_sec_opts);
1734 ret = parse_security_options(data, &new_sec_opts);
1735 if (ret)
1736 goto restore;
1737 ret = setup_security_options(fs_info, sb,
1738 &new_sec_opts);
1739 if (ret) {
1740 security_free_mnt_opts(&new_sec_opts);
1741 goto restore;
1742 }
1743 }
1744
1745 ret = btrfs_parse_options(root, data, *flags);
1746 if (ret) {
1747 ret = -EINVAL;
1748 goto restore;
1749 }
1750
1751 btrfs_remount_begin(fs_info, old_opts, *flags);
1752 btrfs_resize_thread_pool(fs_info,
1753 fs_info->thread_pool_size, old_thread_pool_size);
1754
1755 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1756 goto out;
1757
1758 if (*flags & MS_RDONLY) {
1759 /*
1760 * this also happens on 'umount -rf' or on shutdown, when
1761 * the filesystem is busy.
1762 */
1763 cancel_work_sync(&fs_info->async_reclaim_work);
1764
1765 /* wait for the uuid_scan task to finish */
1766 down(&fs_info->uuid_tree_rescan_sem);
1767 /* avoid complains from lockdep et al. */
1768 up(&fs_info->uuid_tree_rescan_sem);
1769
1770 sb->s_flags |= MS_RDONLY;
1771
1772 /*
1773 * Setting MS_RDONLY will put the cleaner thread to
1774 * sleep at the next loop if it's already active.
1775 * If it's already asleep, we'll leave unused block
1776 * groups on disk until we're mounted read-write again
1777 * unless we clean them up here.
1778 */
1779 btrfs_delete_unused_bgs(fs_info);
1780
1781 btrfs_dev_replace_suspend_for_unmount(fs_info);
1782 btrfs_scrub_cancel(fs_info);
1783 btrfs_pause_balance(fs_info);
1784
1785 ret = btrfs_commit_super(root);
1786 if (ret)
1787 goto restore;
1788 } else {
1789 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1790 btrfs_err(fs_info,
1791 "Remounting read-write after error is not allowed");
1792 ret = -EINVAL;
1793 goto restore;
1794 }
1795 if (fs_info->fs_devices->rw_devices == 0) {
1796 ret = -EACCES;
1797 goto restore;
1798 }
1799
1800 if (fs_info->fs_devices->missing_devices >
1801 fs_info->num_tolerated_disk_barrier_failures &&
1802 !(*flags & MS_RDONLY)) {
1803 btrfs_warn(fs_info,
1804 "too many missing devices, writeable remount is not allowed");
1805 ret = -EACCES;
1806 goto restore;
1807 }
1808
1809 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1810 ret = -EINVAL;
1811 goto restore;
1812 }
1813
1814 ret = btrfs_cleanup_fs_roots(fs_info);
1815 if (ret)
1816 goto restore;
1817
1818 /* recover relocation */
1819 mutex_lock(&fs_info->cleaner_mutex);
1820 ret = btrfs_recover_relocation(root);
1821 mutex_unlock(&fs_info->cleaner_mutex);
1822 if (ret)
1823 goto restore;
1824
1825 ret = btrfs_resume_balance_async(fs_info);
1826 if (ret)
1827 goto restore;
1828
1829 ret = btrfs_resume_dev_replace_async(fs_info);
1830 if (ret) {
1831 btrfs_warn(fs_info, "failed to resume dev_replace");
1832 goto restore;
1833 }
1834
1835 if (!fs_info->uuid_root) {
1836 btrfs_info(fs_info, "creating UUID tree");
1837 ret = btrfs_create_uuid_tree(fs_info);
1838 if (ret) {
1839 btrfs_warn(fs_info,
1840 "failed to create the UUID tree %d",
1841 ret);
1842 goto restore;
1843 }
1844 }
1845 sb->s_flags &= ~MS_RDONLY;
1846
1847 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1848 }
1849 out:
1850 wake_up_process(fs_info->transaction_kthread);
1851 btrfs_remount_cleanup(fs_info, old_opts);
1852 return 0;
1853
1854 restore:
1855 /* We've hit an error - don't reset MS_RDONLY */
1856 if (sb->s_flags & MS_RDONLY)
1857 old_flags |= MS_RDONLY;
1858 sb->s_flags = old_flags;
1859 fs_info->mount_opt = old_opts;
1860 fs_info->compress_type = old_compress_type;
1861 fs_info->max_inline = old_max_inline;
1862 mutex_lock(&fs_info->chunk_mutex);
1863 fs_info->alloc_start = old_alloc_start;
1864 mutex_unlock(&fs_info->chunk_mutex);
1865 btrfs_resize_thread_pool(fs_info,
1866 old_thread_pool_size, fs_info->thread_pool_size);
1867 fs_info->metadata_ratio = old_metadata_ratio;
1868 btrfs_remount_cleanup(fs_info, old_opts);
1869 return ret;
1870 }
1871
1872 /* Used to sort the devices by max_avail(descending sort) */
1873 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1874 const void *dev_info2)
1875 {
1876 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1877 ((struct btrfs_device_info *)dev_info2)->max_avail)
1878 return -1;
1879 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1880 ((struct btrfs_device_info *)dev_info2)->max_avail)
1881 return 1;
1882 else
1883 return 0;
1884 }
1885
1886 /*
1887 * sort the devices by max_avail, in which max free extent size of each device
1888 * is stored.(Descending Sort)
1889 */
1890 static inline void btrfs_descending_sort_devices(
1891 struct btrfs_device_info *devices,
1892 size_t nr_devices)
1893 {
1894 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1895 btrfs_cmp_device_free_bytes, NULL);
1896 }
1897
1898 /*
1899 * The helper to calc the free space on the devices that can be used to store
1900 * file data.
1901 */
1902 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1903 {
1904 struct btrfs_fs_info *fs_info = root->fs_info;
1905 struct btrfs_device_info *devices_info;
1906 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1907 struct btrfs_device *device;
1908 u64 skip_space;
1909 u64 type;
1910 u64 avail_space;
1911 u64 used_space;
1912 u64 min_stripe_size;
1913 int min_stripes = 1, num_stripes = 1;
1914 int i = 0, nr_devices;
1915 int ret;
1916
1917 /*
1918 * We aren't under the device list lock, so this is racy-ish, but good
1919 * enough for our purposes.
1920 */
1921 nr_devices = fs_info->fs_devices->open_devices;
1922 if (!nr_devices) {
1923 smp_mb();
1924 nr_devices = fs_info->fs_devices->open_devices;
1925 ASSERT(nr_devices);
1926 if (!nr_devices) {
1927 *free_bytes = 0;
1928 return 0;
1929 }
1930 }
1931
1932 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1933 GFP_NOFS);
1934 if (!devices_info)
1935 return -ENOMEM;
1936
1937 /* calc min stripe number for data space allocation */
1938 type = btrfs_get_alloc_profile(root, 1);
1939 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1940 min_stripes = 2;
1941 num_stripes = nr_devices;
1942 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1943 min_stripes = 2;
1944 num_stripes = 2;
1945 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1946 min_stripes = 4;
1947 num_stripes = 4;
1948 }
1949
1950 if (type & BTRFS_BLOCK_GROUP_DUP)
1951 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1952 else
1953 min_stripe_size = BTRFS_STRIPE_LEN;
1954
1955 if (fs_info->alloc_start)
1956 mutex_lock(&fs_devices->device_list_mutex);
1957 rcu_read_lock();
1958 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1959 if (!device->in_fs_metadata || !device->bdev ||
1960 device->is_tgtdev_for_dev_replace)
1961 continue;
1962
1963 if (i >= nr_devices)
1964 break;
1965
1966 avail_space = device->total_bytes - device->bytes_used;
1967
1968 /* align with stripe_len */
1969 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1970 avail_space *= BTRFS_STRIPE_LEN;
1971
1972 /*
1973 * In order to avoid overwriting the superblock on the drive,
1974 * btrfs starts at an offset of at least 1MB when doing chunk
1975 * allocation.
1976 */
1977 skip_space = SZ_1M;
1978
1979 /* user can set the offset in fs_info->alloc_start. */
1980 if (fs_info->alloc_start &&
1981 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1982 device->total_bytes) {
1983 rcu_read_unlock();
1984 skip_space = max(fs_info->alloc_start, skip_space);
1985
1986 /*
1987 * btrfs can not use the free space in
1988 * [0, skip_space - 1], we must subtract it from the
1989 * total. In order to implement it, we account the used
1990 * space in this range first.
1991 */
1992 ret = btrfs_account_dev_extents_size(device, 0,
1993 skip_space - 1,
1994 &used_space);
1995 if (ret) {
1996 kfree(devices_info);
1997 mutex_unlock(&fs_devices->device_list_mutex);
1998 return ret;
1999 }
2000
2001 rcu_read_lock();
2002
2003 /* calc the free space in [0, skip_space - 1] */
2004 skip_space -= used_space;
2005 }
2006
2007 /*
2008 * we can use the free space in [0, skip_space - 1], subtract
2009 * it from the total.
2010 */
2011 if (avail_space && avail_space >= skip_space)
2012 avail_space -= skip_space;
2013 else
2014 avail_space = 0;
2015
2016 if (avail_space < min_stripe_size)
2017 continue;
2018
2019 devices_info[i].dev = device;
2020 devices_info[i].max_avail = avail_space;
2021
2022 i++;
2023 }
2024 rcu_read_unlock();
2025 if (fs_info->alloc_start)
2026 mutex_unlock(&fs_devices->device_list_mutex);
2027
2028 nr_devices = i;
2029
2030 btrfs_descending_sort_devices(devices_info, nr_devices);
2031
2032 i = nr_devices - 1;
2033 avail_space = 0;
2034 while (nr_devices >= min_stripes) {
2035 if (num_stripes > nr_devices)
2036 num_stripes = nr_devices;
2037
2038 if (devices_info[i].max_avail >= min_stripe_size) {
2039 int j;
2040 u64 alloc_size;
2041
2042 avail_space += devices_info[i].max_avail * num_stripes;
2043 alloc_size = devices_info[i].max_avail;
2044 for (j = i + 1 - num_stripes; j <= i; j++)
2045 devices_info[j].max_avail -= alloc_size;
2046 }
2047 i--;
2048 nr_devices--;
2049 }
2050
2051 kfree(devices_info);
2052 *free_bytes = avail_space;
2053 return 0;
2054 }
2055
2056 /*
2057 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2058 *
2059 * If there's a redundant raid level at DATA block groups, use the respective
2060 * multiplier to scale the sizes.
2061 *
2062 * Unused device space usage is based on simulating the chunk allocator
2063 * algorithm that respects the device sizes, order of allocations and the
2064 * 'alloc_start' value, this is a close approximation of the actual use but
2065 * there are other factors that may change the result (like a new metadata
2066 * chunk).
2067 *
2068 * If metadata is exhausted, f_bavail will be 0.
2069 */
2070 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2071 {
2072 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2073 struct btrfs_super_block *disk_super = fs_info->super_copy;
2074 struct list_head *head = &fs_info->space_info;
2075 struct btrfs_space_info *found;
2076 u64 total_used = 0;
2077 u64 total_free_data = 0;
2078 u64 total_free_meta = 0;
2079 int bits = dentry->d_sb->s_blocksize_bits;
2080 __be32 *fsid = (__be32 *)fs_info->fsid;
2081 unsigned factor = 1;
2082 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2083 int ret;
2084 u64 thresh = 0;
2085 int mixed = 0;
2086
2087 /*
2088 * holding chunk_mutex to avoid allocating new chunks, holding
2089 * device_list_mutex to avoid the device being removed
2090 */
2091 rcu_read_lock();
2092 list_for_each_entry_rcu(found, head, list) {
2093 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2094 int i;
2095
2096 total_free_data += found->disk_total - found->disk_used;
2097 total_free_data -=
2098 btrfs_account_ro_block_groups_free_space(found);
2099
2100 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2101 if (!list_empty(&found->block_groups[i])) {
2102 switch (i) {
2103 case BTRFS_RAID_DUP:
2104 case BTRFS_RAID_RAID1:
2105 case BTRFS_RAID_RAID10:
2106 factor = 2;
2107 }
2108 }
2109 }
2110 }
2111
2112 /*
2113 * Metadata in mixed block goup profiles are accounted in data
2114 */
2115 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2116 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2117 mixed = 1;
2118 else
2119 total_free_meta += found->disk_total -
2120 found->disk_used;
2121 }
2122
2123 total_used += found->disk_used;
2124 }
2125
2126 rcu_read_unlock();
2127
2128 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2129 buf->f_blocks >>= bits;
2130 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2131
2132 /* Account global block reserve as used, it's in logical size already */
2133 spin_lock(&block_rsv->lock);
2134 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2135 if (buf->f_bfree >= block_rsv->size >> bits)
2136 buf->f_bfree -= block_rsv->size >> bits;
2137 else
2138 buf->f_bfree = 0;
2139 spin_unlock(&block_rsv->lock);
2140
2141 buf->f_bavail = div_u64(total_free_data, factor);
2142 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
2143 if (ret)
2144 return ret;
2145 buf->f_bavail += div_u64(total_free_data, factor);
2146 buf->f_bavail = buf->f_bavail >> bits;
2147
2148 /*
2149 * We calculate the remaining metadata space minus global reserve. If
2150 * this is (supposedly) smaller than zero, there's no space. But this
2151 * does not hold in practice, the exhausted state happens where's still
2152 * some positive delta. So we apply some guesswork and compare the
2153 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2154 *
2155 * We probably cannot calculate the exact threshold value because this
2156 * depends on the internal reservations requested by various
2157 * operations, so some operations that consume a few metadata will
2158 * succeed even if the Avail is zero. But this is better than the other
2159 * way around.
2160 */
2161 thresh = 4 * 1024 * 1024;
2162
2163 if (!mixed && total_free_meta - thresh < block_rsv->size)
2164 buf->f_bavail = 0;
2165
2166 buf->f_type = BTRFS_SUPER_MAGIC;
2167 buf->f_bsize = dentry->d_sb->s_blocksize;
2168 buf->f_namelen = BTRFS_NAME_LEN;
2169
2170 /* We treat it as constant endianness (it doesn't matter _which_)
2171 because we want the fsid to come out the same whether mounted
2172 on a big-endian or little-endian host */
2173 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2174 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2175 /* Mask in the root object ID too, to disambiguate subvols */
2176 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2177 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2178
2179 return 0;
2180 }
2181
2182 static void btrfs_kill_super(struct super_block *sb)
2183 {
2184 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2185 kill_anon_super(sb);
2186 free_fs_info(fs_info);
2187 }
2188
2189 static struct file_system_type btrfs_fs_type = {
2190 .owner = THIS_MODULE,
2191 .name = "btrfs",
2192 .mount = btrfs_mount,
2193 .kill_sb = btrfs_kill_super,
2194 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2195 };
2196 MODULE_ALIAS_FS("btrfs");
2197
2198 static int btrfs_control_open(struct inode *inode, struct file *file)
2199 {
2200 /*
2201 * The control file's private_data is used to hold the
2202 * transaction when it is started and is used to keep
2203 * track of whether a transaction is already in progress.
2204 */
2205 file->private_data = NULL;
2206 return 0;
2207 }
2208
2209 /*
2210 * used by btrfsctl to scan devices when no FS is mounted
2211 */
2212 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2213 unsigned long arg)
2214 {
2215 struct btrfs_ioctl_vol_args *vol;
2216 struct btrfs_fs_devices *fs_devices;
2217 int ret = -ENOTTY;
2218
2219 if (!capable(CAP_SYS_ADMIN))
2220 return -EPERM;
2221
2222 vol = memdup_user((void __user *)arg, sizeof(*vol));
2223 if (IS_ERR(vol))
2224 return PTR_ERR(vol);
2225
2226 switch (cmd) {
2227 case BTRFS_IOC_SCAN_DEV:
2228 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2229 &btrfs_fs_type, &fs_devices);
2230 break;
2231 case BTRFS_IOC_DEVICES_READY:
2232 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2233 &btrfs_fs_type, &fs_devices);
2234 if (ret)
2235 break;
2236 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2237 break;
2238 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2239 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2240 break;
2241 }
2242
2243 kfree(vol);
2244 return ret;
2245 }
2246
2247 static int btrfs_freeze(struct super_block *sb)
2248 {
2249 struct btrfs_trans_handle *trans;
2250 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
2251
2252 root->fs_info->fs_frozen = 1;
2253 /*
2254 * We don't need a barrier here, we'll wait for any transaction that
2255 * could be in progress on other threads (and do delayed iputs that
2256 * we want to avoid on a frozen filesystem), or do the commit
2257 * ourselves.
2258 */
2259 trans = btrfs_attach_transaction_barrier(root);
2260 if (IS_ERR(trans)) {
2261 /* no transaction, don't bother */
2262 if (PTR_ERR(trans) == -ENOENT)
2263 return 0;
2264 return PTR_ERR(trans);
2265 }
2266 return btrfs_commit_transaction(trans, root);
2267 }
2268
2269 static int btrfs_unfreeze(struct super_block *sb)
2270 {
2271 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
2272
2273 root->fs_info->fs_frozen = 0;
2274 return 0;
2275 }
2276
2277 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2278 {
2279 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2280 struct btrfs_fs_devices *cur_devices;
2281 struct btrfs_device *dev, *first_dev = NULL;
2282 struct list_head *head;
2283 struct rcu_string *name;
2284
2285 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2286 cur_devices = fs_info->fs_devices;
2287 while (cur_devices) {
2288 head = &cur_devices->devices;
2289 list_for_each_entry(dev, head, dev_list) {
2290 if (dev->missing)
2291 continue;
2292 if (!dev->name)
2293 continue;
2294 if (!first_dev || dev->devid < first_dev->devid)
2295 first_dev = dev;
2296 }
2297 cur_devices = cur_devices->seed;
2298 }
2299
2300 if (first_dev) {
2301 rcu_read_lock();
2302 name = rcu_dereference(first_dev->name);
2303 seq_escape(m, name->str, " \t\n\\");
2304 rcu_read_unlock();
2305 } else {
2306 WARN_ON(1);
2307 }
2308 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2309 return 0;
2310 }
2311
2312 static const struct super_operations btrfs_super_ops = {
2313 .drop_inode = btrfs_drop_inode,
2314 .evict_inode = btrfs_evict_inode,
2315 .put_super = btrfs_put_super,
2316 .sync_fs = btrfs_sync_fs,
2317 .show_options = btrfs_show_options,
2318 .show_devname = btrfs_show_devname,
2319 .write_inode = btrfs_write_inode,
2320 .alloc_inode = btrfs_alloc_inode,
2321 .destroy_inode = btrfs_destroy_inode,
2322 .statfs = btrfs_statfs,
2323 .remount_fs = btrfs_remount,
2324 .freeze_fs = btrfs_freeze,
2325 .unfreeze_fs = btrfs_unfreeze,
2326 };
2327
2328 static const struct file_operations btrfs_ctl_fops = {
2329 .open = btrfs_control_open,
2330 .unlocked_ioctl = btrfs_control_ioctl,
2331 .compat_ioctl = btrfs_control_ioctl,
2332 .owner = THIS_MODULE,
2333 .llseek = noop_llseek,
2334 };
2335
2336 static struct miscdevice btrfs_misc = {
2337 .minor = BTRFS_MINOR,
2338 .name = "btrfs-control",
2339 .fops = &btrfs_ctl_fops
2340 };
2341
2342 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2343 MODULE_ALIAS("devname:btrfs-control");
2344
2345 static int btrfs_interface_init(void)
2346 {
2347 return misc_register(&btrfs_misc);
2348 }
2349
2350 static void btrfs_interface_exit(void)
2351 {
2352 misc_deregister(&btrfs_misc);
2353 }
2354
2355 static void btrfs_print_mod_info(void)
2356 {
2357 pr_info("Btrfs loaded, crc32c=%s"
2358 #ifdef CONFIG_BTRFS_DEBUG
2359 ", debug=on"
2360 #endif
2361 #ifdef CONFIG_BTRFS_ASSERT
2362 ", assert=on"
2363 #endif
2364 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2365 ", integrity-checker=on"
2366 #endif
2367 "\n",
2368 btrfs_crc32c_impl());
2369 }
2370
2371 static int __init init_btrfs_fs(void)
2372 {
2373 int err;
2374
2375 err = btrfs_hash_init();
2376 if (err)
2377 return err;
2378
2379 btrfs_props_init();
2380
2381 err = btrfs_init_sysfs();
2382 if (err)
2383 goto free_hash;
2384
2385 btrfs_init_compress();
2386
2387 err = btrfs_init_cachep();
2388 if (err)
2389 goto free_compress;
2390
2391 err = extent_io_init();
2392 if (err)
2393 goto free_cachep;
2394
2395 err = extent_map_init();
2396 if (err)
2397 goto free_extent_io;
2398
2399 err = ordered_data_init();
2400 if (err)
2401 goto free_extent_map;
2402
2403 err = btrfs_delayed_inode_init();
2404 if (err)
2405 goto free_ordered_data;
2406
2407 err = btrfs_auto_defrag_init();
2408 if (err)
2409 goto free_delayed_inode;
2410
2411 err = btrfs_delayed_ref_init();
2412 if (err)
2413 goto free_auto_defrag;
2414
2415 err = btrfs_prelim_ref_init();
2416 if (err)
2417 goto free_delayed_ref;
2418
2419 err = btrfs_end_io_wq_init();
2420 if (err)
2421 goto free_prelim_ref;
2422
2423 err = btrfs_interface_init();
2424 if (err)
2425 goto free_end_io_wq;
2426
2427 btrfs_init_lockdep();
2428
2429 btrfs_print_mod_info();
2430
2431 err = btrfs_run_sanity_tests();
2432 if (err)
2433 goto unregister_ioctl;
2434
2435 err = register_filesystem(&btrfs_fs_type);
2436 if (err)
2437 goto unregister_ioctl;
2438
2439 return 0;
2440
2441 unregister_ioctl:
2442 btrfs_interface_exit();
2443 free_end_io_wq:
2444 btrfs_end_io_wq_exit();
2445 free_prelim_ref:
2446 btrfs_prelim_ref_exit();
2447 free_delayed_ref:
2448 btrfs_delayed_ref_exit();
2449 free_auto_defrag:
2450 btrfs_auto_defrag_exit();
2451 free_delayed_inode:
2452 btrfs_delayed_inode_exit();
2453 free_ordered_data:
2454 ordered_data_exit();
2455 free_extent_map:
2456 extent_map_exit();
2457 free_extent_io:
2458 extent_io_exit();
2459 free_cachep:
2460 btrfs_destroy_cachep();
2461 free_compress:
2462 btrfs_exit_compress();
2463 btrfs_exit_sysfs();
2464 free_hash:
2465 btrfs_hash_exit();
2466 return err;
2467 }
2468
2469 static void __exit exit_btrfs_fs(void)
2470 {
2471 btrfs_destroy_cachep();
2472 btrfs_delayed_ref_exit();
2473 btrfs_auto_defrag_exit();
2474 btrfs_delayed_inode_exit();
2475 btrfs_prelim_ref_exit();
2476 ordered_data_exit();
2477 extent_map_exit();
2478 extent_io_exit();
2479 btrfs_interface_exit();
2480 btrfs_end_io_wq_exit();
2481 unregister_filesystem(&btrfs_fs_type);
2482 btrfs_exit_sysfs();
2483 btrfs_cleanup_fs_uuids();
2484 btrfs_exit_compress();
2485 btrfs_hash_exit();
2486 }
2487
2488 late_initcall(init_btrfs_fs);
2489 module_exit(exit_btrfs_fs)
2490
2491 MODULE_LICENSE("GPL");
Ubuntu Jammy Linux kernel mirror