?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 4.18.10
[linux/fpc-iii.git] / fs / btrfs / super.c
blob9b25f29d0e73ffafb4772def7e921e7608395cae
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/buffer_head.h>
9 #include <linux/fs.h>
10 #include <linux/pagemap.h>
11 #include <linux/highmem.h>
12 #include <linux/time.h>
13 #include <linux/init.h>
14 #include <linux/seq_file.h>
15 #include <linux/string.h>
16 #include <linux/backing-dev.h>
17 #include <linux/mount.h>
18 #include <linux/mpage.h>
19 #include <linux/swap.h>
20 #include <linux/writeback.h>
21 #include <linux/statfs.h>
22 #include <linux/compat.h>
23 #include <linux/parser.h>
24 #include <linux/ctype.h>
25 #include <linux/namei.h>
26 #include <linux/miscdevice.h>
27 #include <linux/magic.h>
28 #include <linux/slab.h>
29 #include <linux/cleancache.h>
30 #include <linux/ratelimit.h>
31 #include <linux/crc32c.h>
32 #include <linux/btrfs.h>
33 #include "delayed-inode.h"
34 #include "ctree.h"
35 #include "disk-io.h"
36 #include "transaction.h"
37 #include "btrfs_inode.h"
38 #include "print-tree.h"
39 #include "props.h"
40 #include "xattr.h"
41 #include "volumes.h"
42 #include "export.h"
43 #include "compression.h"
44 #include "rcu-string.h"
45 #include "dev-replace.h"
46 #include "free-space-cache.h"
47 #include "backref.h"
48 #include "tests/btrfs-tests.h"
49
50 #include "qgroup.h"
51 #define CREATE_TRACE_POINTS
52 #include <trace/events/btrfs.h>
53
54 static const struct super_operations btrfs_super_ops;
55
56 /*
57 * Types for mounting the default subvolume and a subvolume explicitly
58 * requested by subvol=/path. That way the callchain is straightforward and we
59 * don't have to play tricks with the mount options and recursive calls to
60 * btrfs_mount.
61 *
62 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
63 */
64 static struct file_system_type btrfs_fs_type;
65 static struct file_system_type btrfs_root_fs_type;
66
67 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
68
69 const char *btrfs_decode_error(int errno)
70 {
71 char *errstr = "unknown";
72
73 switch (errno) {
74 case -EIO:
75 errstr = "IO failure";
76 break;
77 case -ENOMEM:
78 errstr = "Out of memory";
79 break;
80 case -EROFS:
81 errstr = "Readonly filesystem";
82 break;
83 case -EEXIST:
84 errstr = "Object already exists";
85 break;
86 case -ENOSPC:
87 errstr = "No space left";
88 break;
89 case -ENOENT:
90 errstr = "No such entry";
91 break;
92 }
93
94 return errstr;
95 }
96
97 /*
98 * __btrfs_handle_fs_error decodes expected errors from the caller and
99 * invokes the approciate error response.
100 */
101 __cold
102 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
103 unsigned int line, int errno, const char *fmt, ...)
104 {
105 struct super_block *sb = fs_info->sb;
106 #ifdef CONFIG_PRINTK
107 const char *errstr;
108 #endif
109
110 /*
111 * Special case: if the error is EROFS, and we're already
112 * under SB_RDONLY, then it is safe here.
113 */
114 if (errno == -EROFS && sb_rdonly(sb))
115 return;
116
117 #ifdef CONFIG_PRINTK
118 errstr = btrfs_decode_error(errno);
119 if (fmt) {
120 struct va_format vaf;
121 va_list args;
122
123 va_start(args, fmt);
124 vaf.fmt = fmt;
125 vaf.va = &args;
126
127 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
128 sb->s_id, function, line, errno, errstr, &vaf);
129 va_end(args);
130 } else {
131 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
132 sb->s_id, function, line, errno, errstr);
133 }
134 #endif
135
136 /*
137 * Today we only save the error info to memory. Long term we'll
138 * also send it down to the disk
139 */
140 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
141
142 /* Don't go through full error handling during mount */
143 if (!(sb->s_flags & SB_BORN))
144 return;
145
146 if (sb_rdonly(sb))
147 return;
148
149 /* btrfs handle error by forcing the filesystem readonly */
150 sb->s_flags |= SB_RDONLY;
151 btrfs_info(fs_info, "forced readonly");
152 /*
153 * Note that a running device replace operation is not canceled here
154 * although there is no way to update the progress. It would add the
155 * risk of a deadlock, therefore the canceling is omitted. The only
156 * penalty is that some I/O remains active until the procedure
157 * completes. The next time when the filesystem is mounted writeable
158 * again, the device replace operation continues.
159 */
160 }
161
162 #ifdef CONFIG_PRINTK
163 static const char * const logtypes[] = {
164 "emergency",
165 "alert",
166 "critical",
167 "error",
168 "warning",
169 "notice",
170 "info",
171 "debug",
172 };
173
174
175 /*
176 * Use one ratelimit state per log level so that a flood of less important
177 * messages doesn't cause more important ones to be dropped.
178 */
179 static struct ratelimit_state printk_limits[] = {
180 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
181 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
182 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
183 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
184 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
185 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
186 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
187 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
188 };
189
190 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
191 {
192 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
193 struct va_format vaf;
194 va_list args;
195 int kern_level;
196 const char *type = logtypes[4];
197 struct ratelimit_state *ratelimit = &printk_limits[4];
198
199 va_start(args, fmt);
200
201 while ((kern_level = printk_get_level(fmt)) != 0) {
202 size_t size = printk_skip_level(fmt) - fmt;
203
204 if (kern_level >= '0' && kern_level <= '7') {
205 memcpy(lvl, fmt, size);
206 lvl[size] = '\0';
207 type = logtypes[kern_level - '0'];
208 ratelimit = &printk_limits[kern_level - '0'];
209 }
210 fmt += size;
211 }
212
213 vaf.fmt = fmt;
214 vaf.va = &args;
215
216 if (__ratelimit(ratelimit))
217 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
218 fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
219
220 va_end(args);
221 }
222 #endif
223
224 /*
225 * We only mark the transaction aborted and then set the file system read-only.
226 * This will prevent new transactions from starting or trying to join this
227 * one.
228 *
229 * This means that error recovery at the call site is limited to freeing
230 * any local memory allocations and passing the error code up without
231 * further cleanup. The transaction should complete as it normally would
232 * in the call path but will return -EIO.
233 *
234 * We'll complete the cleanup in btrfs_end_transaction and
235 * btrfs_commit_transaction.
236 */
237 __cold
238 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
239 const char *function,
240 unsigned int line, int errno)
241 {
242 struct btrfs_fs_info *fs_info = trans->fs_info;
243
244 trans->aborted = errno;
245 /* Nothing used. The other threads that have joined this
246 * transaction may be able to continue. */
247 if (!trans->dirty && list_empty(&trans->new_bgs)) {
248 const char *errstr;
249
250 errstr = btrfs_decode_error(errno);
251 btrfs_warn(fs_info,
252 "%s:%d: Aborting unused transaction(%s).",
253 function, line, errstr);
254 return;
255 }
256 WRITE_ONCE(trans->transaction->aborted, errno);
257 /* Wake up anybody who may be waiting on this transaction */
258 wake_up(&fs_info->transaction_wait);
259 wake_up(&fs_info->transaction_blocked_wait);
260 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
261 }
262 /*
263 * __btrfs_panic decodes unexpected, fatal errors from the caller,
264 * issues an alert, and either panics or BUGs, depending on mount options.
265 */
266 __cold
267 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
268 unsigned int line, int errno, const char *fmt, ...)
269 {
270 char *s_id = "<unknown>";
271 const char *errstr;
272 struct va_format vaf = { .fmt = fmt };
273 va_list args;
274
275 if (fs_info)
276 s_id = fs_info->sb->s_id;
277
278 va_start(args, fmt);
279 vaf.va = &args;
280
281 errstr = btrfs_decode_error(errno);
282 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
283 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
284 s_id, function, line, &vaf, errno, errstr);
285
286 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
287 function, line, &vaf, errno, errstr);
288 va_end(args);
289 /* Caller calls BUG() */
290 }
291
292 static void btrfs_put_super(struct super_block *sb)
293 {
294 close_ctree(btrfs_sb(sb));
295 }
296
297 enum {
298 Opt_acl, Opt_noacl,
299 Opt_clear_cache,
300 Opt_commit_interval,
301 Opt_compress,
302 Opt_compress_force,
303 Opt_compress_force_type,
304 Opt_compress_type,
305 Opt_degraded,
306 Opt_device,
307 Opt_fatal_errors,
308 Opt_flushoncommit, Opt_noflushoncommit,
309 Opt_inode_cache, Opt_noinode_cache,
310 Opt_max_inline,
311 Opt_barrier, Opt_nobarrier,
312 Opt_datacow, Opt_nodatacow,
313 Opt_datasum, Opt_nodatasum,
314 Opt_defrag, Opt_nodefrag,
315 Opt_discard, Opt_nodiscard,
316 Opt_nologreplay,
317 Opt_norecovery,
318 Opt_ratio,
319 Opt_rescan_uuid_tree,
320 Opt_skip_balance,
321 Opt_space_cache, Opt_no_space_cache,
322 Opt_space_cache_version,
323 Opt_ssd, Opt_nossd,
324 Opt_ssd_spread, Opt_nossd_spread,
325 Opt_subvol,
326 Opt_subvol_empty,
327 Opt_subvolid,
328 Opt_thread_pool,
329 Opt_treelog, Opt_notreelog,
330 Opt_usebackuproot,
331 Opt_user_subvol_rm_allowed,
332
333 /* Deprecated options */
334 Opt_alloc_start,
335 Opt_recovery,
336 Opt_subvolrootid,
337
338 /* Debugging options */
339 Opt_check_integrity,
340 Opt_check_integrity_including_extent_data,
341 Opt_check_integrity_print_mask,
342 Opt_enospc_debug, Opt_noenospc_debug,
343 #ifdef CONFIG_BTRFS_DEBUG
344 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
345 #endif
346 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
347 Opt_ref_verify,
348 #endif
349 Opt_err,
350 };
351
352 static const match_table_t tokens = {
353 {Opt_acl, "acl"},
354 {Opt_noacl, "noacl"},
355 {Opt_clear_cache, "clear_cache"},
356 {Opt_commit_interval, "commit=%u"},
357 {Opt_compress, "compress"},
358 {Opt_compress_type, "compress=%s"},
359 {Opt_compress_force, "compress-force"},
360 {Opt_compress_force_type, "compress-force=%s"},
361 {Opt_degraded, "degraded"},
362 {Opt_device, "device=%s"},
363 {Opt_fatal_errors, "fatal_errors=%s"},
364 {Opt_flushoncommit, "flushoncommit"},
365 {Opt_noflushoncommit, "noflushoncommit"},
366 {Opt_inode_cache, "inode_cache"},
367 {Opt_noinode_cache, "noinode_cache"},
368 {Opt_max_inline, "max_inline=%s"},
369 {Opt_barrier, "barrier"},
370 {Opt_nobarrier, "nobarrier"},
371 {Opt_datacow, "datacow"},
372 {Opt_nodatacow, "nodatacow"},
373 {Opt_datasum, "datasum"},
374 {Opt_nodatasum, "nodatasum"},
375 {Opt_defrag, "autodefrag"},
376 {Opt_nodefrag, "noautodefrag"},
377 {Opt_discard, "discard"},
378 {Opt_nodiscard, "nodiscard"},
379 {Opt_nologreplay, "nologreplay"},
380 {Opt_norecovery, "norecovery"},
381 {Opt_ratio, "metadata_ratio=%u"},
382 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
383 {Opt_skip_balance, "skip_balance"},
384 {Opt_space_cache, "space_cache"},
385 {Opt_no_space_cache, "nospace_cache"},
386 {Opt_space_cache_version, "space_cache=%s"},
387 {Opt_ssd, "ssd"},
388 {Opt_nossd, "nossd"},
389 {Opt_ssd_spread, "ssd_spread"},
390 {Opt_nossd_spread, "nossd_spread"},
391 {Opt_subvol, "subvol=%s"},
392 {Opt_subvol_empty, "subvol="},
393 {Opt_subvolid, "subvolid=%s"},
394 {Opt_thread_pool, "thread_pool=%u"},
395 {Opt_treelog, "treelog"},
396 {Opt_notreelog, "notreelog"},
397 {Opt_usebackuproot, "usebackuproot"},
398 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
399
400 /* Deprecated options */
401 {Opt_alloc_start, "alloc_start=%s"},
402 {Opt_recovery, "recovery"},
403 {Opt_subvolrootid, "subvolrootid=%d"},
404
405 /* Debugging options */
406 {Opt_check_integrity, "check_int"},
407 {Opt_check_integrity_including_extent_data, "check_int_data"},
408 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
409 {Opt_enospc_debug, "enospc_debug"},
410 {Opt_noenospc_debug, "noenospc_debug"},
411 #ifdef CONFIG_BTRFS_DEBUG
412 {Opt_fragment_data, "fragment=data"},
413 {Opt_fragment_metadata, "fragment=metadata"},
414 {Opt_fragment_all, "fragment=all"},
415 #endif
416 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
417 {Opt_ref_verify, "ref_verify"},
418 #endif
419 {Opt_err, NULL},
420 };
421
422 /*
423 * Regular mount options parser. Everything that is needed only when
424 * reading in a new superblock is parsed here.
425 * XXX JDM: This needs to be cleaned up for remount.
426 */
427 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
428 unsigned long new_flags)
429 {
430 substring_t args[MAX_OPT_ARGS];
431 char *p, *num;
432 u64 cache_gen;
433 int intarg;
434 int ret = 0;
435 char *compress_type;
436 bool compress_force = false;
437 enum btrfs_compression_type saved_compress_type;
438 bool saved_compress_force;
439 int no_compress = 0;
440
441 cache_gen = btrfs_super_cache_generation(info->super_copy);
442 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
443 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
444 else if (cache_gen)
445 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
446
447 /*
448 * Even the options are empty, we still need to do extra check
449 * against new flags
450 */
451 if (!options)
452 goto check;
453
454 while ((p = strsep(&options, ",")) != NULL) {
455 int token;
456 if (!*p)
457 continue;
458
459 token = match_token(p, tokens, args);
460 switch (token) {
461 case Opt_degraded:
462 btrfs_info(info, "allowing degraded mounts");
463 btrfs_set_opt(info->mount_opt, DEGRADED);
464 break;
465 case Opt_subvol:
466 case Opt_subvol_empty:
467 case Opt_subvolid:
468 case Opt_subvolrootid:
469 case Opt_device:
470 /*
471 * These are parsed by btrfs_parse_subvol_options
472 * and btrfs_parse_early_options
473 * and can be happily ignored here.
474 */
475 break;
476 case Opt_nodatasum:
477 btrfs_set_and_info(info, NODATASUM,
478 "setting nodatasum");
479 break;
480 case Opt_datasum:
481 if (btrfs_test_opt(info, NODATASUM)) {
482 if (btrfs_test_opt(info, NODATACOW))
483 btrfs_info(info,
484 "setting datasum, datacow enabled");
485 else
486 btrfs_info(info, "setting datasum");
487 }
488 btrfs_clear_opt(info->mount_opt, NODATACOW);
489 btrfs_clear_opt(info->mount_opt, NODATASUM);
490 break;
491 case Opt_nodatacow:
492 if (!btrfs_test_opt(info, NODATACOW)) {
493 if (!btrfs_test_opt(info, COMPRESS) ||
494 !btrfs_test_opt(info, FORCE_COMPRESS)) {
495 btrfs_info(info,
496 "setting nodatacow, compression disabled");
497 } else {
498 btrfs_info(info, "setting nodatacow");
499 }
500 }
501 btrfs_clear_opt(info->mount_opt, COMPRESS);
502 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
503 btrfs_set_opt(info->mount_opt, NODATACOW);
504 btrfs_set_opt(info->mount_opt, NODATASUM);
505 break;
506 case Opt_datacow:
507 btrfs_clear_and_info(info, NODATACOW,
508 "setting datacow");
509 break;
510 case Opt_compress_force:
511 case Opt_compress_force_type:
512 compress_force = true;
513 /* Fallthrough */
514 case Opt_compress:
515 case Opt_compress_type:
516 saved_compress_type = btrfs_test_opt(info,
517 COMPRESS) ?
518 info->compress_type : BTRFS_COMPRESS_NONE;
519 saved_compress_force =
520 btrfs_test_opt(info, FORCE_COMPRESS);
521 if (token == Opt_compress ||
522 token == Opt_compress_force ||
523 strncmp(args[0].from, "zlib", 4) == 0) {
524 compress_type = "zlib";
525
526 info->compress_type = BTRFS_COMPRESS_ZLIB;
527 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
528 /*
529 * args[0] contains uninitialized data since
530 * for these tokens we don't expect any
531 * parameter.
532 */
533 if (token != Opt_compress &&
534 token != Opt_compress_force)
535 info->compress_level =
536 btrfs_compress_str2level(args[0].from);
537 btrfs_set_opt(info->mount_opt, COMPRESS);
538 btrfs_clear_opt(info->mount_opt, NODATACOW);
539 btrfs_clear_opt(info->mount_opt, NODATASUM);
540 no_compress = 0;
541 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
542 compress_type = "lzo";
543 info->compress_type = BTRFS_COMPRESS_LZO;
544 btrfs_set_opt(info->mount_opt, COMPRESS);
545 btrfs_clear_opt(info->mount_opt, NODATACOW);
546 btrfs_clear_opt(info->mount_opt, NODATASUM);
547 btrfs_set_fs_incompat(info, COMPRESS_LZO);
548 no_compress = 0;
549 } else if (strcmp(args[0].from, "zstd") == 0) {
550 compress_type = "zstd";
551 info->compress_type = BTRFS_COMPRESS_ZSTD;
552 btrfs_set_opt(info->mount_opt, COMPRESS);
553 btrfs_clear_opt(info->mount_opt, NODATACOW);
554 btrfs_clear_opt(info->mount_opt, NODATASUM);
555 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
556 no_compress = 0;
557 } else if (strncmp(args[0].from, "no", 2) == 0) {
558 compress_type = "no";
559 btrfs_clear_opt(info->mount_opt, COMPRESS);
560 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
561 compress_force = false;
562 no_compress++;
563 } else {
564 ret = -EINVAL;
565 goto out;
566 }
567
568 if (compress_force) {
569 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
570 } else {
571 /*
572 * If we remount from compress-force=xxx to
573 * compress=xxx, we need clear FORCE_COMPRESS
574 * flag, otherwise, there is no way for users
575 * to disable forcible compression separately.
576 */
577 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
578 }
579 if ((btrfs_test_opt(info, COMPRESS) &&
580 (info->compress_type != saved_compress_type ||
581 compress_force != saved_compress_force)) ||
582 (!btrfs_test_opt(info, COMPRESS) &&
583 no_compress == 1)) {
584 btrfs_info(info, "%s %s compression, level %d",
585 (compress_force) ? "force" : "use",
586 compress_type, info->compress_level);
587 }
588 compress_force = false;
589 break;
590 case Opt_ssd:
591 btrfs_set_and_info(info, SSD,
592 "enabling ssd optimizations");
593 btrfs_clear_opt(info->mount_opt, NOSSD);
594 break;
595 case Opt_ssd_spread:
596 btrfs_set_and_info(info, SSD,
597 "enabling ssd optimizations");
598 btrfs_set_and_info(info, SSD_SPREAD,
599 "using spread ssd allocation scheme");
600 btrfs_clear_opt(info->mount_opt, NOSSD);
601 break;
602 case Opt_nossd:
603 btrfs_set_opt(info->mount_opt, NOSSD);
604 btrfs_clear_and_info(info, SSD,
605 "not using ssd optimizations");
606 /* Fallthrough */
607 case Opt_nossd_spread:
608 btrfs_clear_and_info(info, SSD_SPREAD,
609 "not using spread ssd allocation scheme");
610 break;
611 case Opt_barrier:
612 btrfs_clear_and_info(info, NOBARRIER,
613 "turning on barriers");
614 break;
615 case Opt_nobarrier:
616 btrfs_set_and_info(info, NOBARRIER,
617 "turning off barriers");
618 break;
619 case Opt_thread_pool:
620 ret = match_int(&args[0], &intarg);
621 if (ret) {
622 goto out;
623 } else if (intarg == 0) {
624 ret = -EINVAL;
625 goto out;
626 }
627 info->thread_pool_size = intarg;
628 break;
629 case Opt_max_inline:
630 num = match_strdup(&args[0]);
631 if (num) {
632 info->max_inline = memparse(num, NULL);
633 kfree(num);
634
635 if (info->max_inline) {
636 info->max_inline = min_t(u64,
637 info->max_inline,
638 info->sectorsize);
639 }
640 btrfs_info(info, "max_inline at %llu",
641 info->max_inline);
642 } else {
643 ret = -ENOMEM;
644 goto out;
645 }
646 break;
647 case Opt_alloc_start:
648 btrfs_info(info,
649 "option alloc_start is obsolete, ignored");
650 break;
651 case Opt_acl:
652 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
653 info->sb->s_flags |= SB_POSIXACL;
654 break;
655 #else
656 btrfs_err(info, "support for ACL not compiled in!");
657 ret = -EINVAL;
658 goto out;
659 #endif
660 case Opt_noacl:
661 info->sb->s_flags &= ~SB_POSIXACL;
662 break;
663 case Opt_notreelog:
664 btrfs_set_and_info(info, NOTREELOG,
665 "disabling tree log");
666 break;
667 case Opt_treelog:
668 btrfs_clear_and_info(info, NOTREELOG,
669 "enabling tree log");
670 break;
671 case Opt_norecovery:
672 case Opt_nologreplay:
673 btrfs_set_and_info(info, NOLOGREPLAY,
674 "disabling log replay at mount time");
675 break;
676 case Opt_flushoncommit:
677 btrfs_set_and_info(info, FLUSHONCOMMIT,
678 "turning on flush-on-commit");
679 break;
680 case Opt_noflushoncommit:
681 btrfs_clear_and_info(info, FLUSHONCOMMIT,
682 "turning off flush-on-commit");
683 break;
684 case Opt_ratio:
685 ret = match_int(&args[0], &intarg);
686 if (ret)
687 goto out;
688 info->metadata_ratio = intarg;
689 btrfs_info(info, "metadata ratio %u",
690 info->metadata_ratio);
691 break;
692 case Opt_discard:
693 btrfs_set_and_info(info, DISCARD,
694 "turning on discard");
695 break;
696 case Opt_nodiscard:
697 btrfs_clear_and_info(info, DISCARD,
698 "turning off discard");
699 break;
700 case Opt_space_cache:
701 case Opt_space_cache_version:
702 if (token == Opt_space_cache ||
703 strcmp(args[0].from, "v1") == 0) {
704 btrfs_clear_opt(info->mount_opt,
705 FREE_SPACE_TREE);
706 btrfs_set_and_info(info, SPACE_CACHE,
707 "enabling disk space caching");
708 } else if (strcmp(args[0].from, "v2") == 0) {
709 btrfs_clear_opt(info->mount_opt,
710 SPACE_CACHE);
711 btrfs_set_and_info(info, FREE_SPACE_TREE,
712 "enabling free space tree");
713 } else {
714 ret = -EINVAL;
715 goto out;
716 }
717 break;
718 case Opt_rescan_uuid_tree:
719 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
720 break;
721 case Opt_no_space_cache:
722 if (btrfs_test_opt(info, SPACE_CACHE)) {
723 btrfs_clear_and_info(info, SPACE_CACHE,
724 "disabling disk space caching");
725 }
726 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
727 btrfs_clear_and_info(info, FREE_SPACE_TREE,
728 "disabling free space tree");
729 }
730 break;
731 case Opt_inode_cache:
732 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
733 "enabling inode map caching");
734 break;
735 case Opt_noinode_cache:
736 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
737 "disabling inode map caching");
738 break;
739 case Opt_clear_cache:
740 btrfs_set_and_info(info, CLEAR_CACHE,
741 "force clearing of disk cache");
742 break;
743 case Opt_user_subvol_rm_allowed:
744 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
745 break;
746 case Opt_enospc_debug:
747 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
748 break;
749 case Opt_noenospc_debug:
750 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
751 break;
752 case Opt_defrag:
753 btrfs_set_and_info(info, AUTO_DEFRAG,
754 "enabling auto defrag");
755 break;
756 case Opt_nodefrag:
757 btrfs_clear_and_info(info, AUTO_DEFRAG,
758 "disabling auto defrag");
759 break;
760 case Opt_recovery:
761 btrfs_warn(info,
762 "'recovery' is deprecated, use 'usebackuproot' instead");
763 case Opt_usebackuproot:
764 btrfs_info(info,
765 "trying to use backup root at mount time");
766 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
767 break;
768 case Opt_skip_balance:
769 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
770 break;
771 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
772 case Opt_check_integrity_including_extent_data:
773 btrfs_info(info,
774 "enabling check integrity including extent data");
775 btrfs_set_opt(info->mount_opt,
776 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
777 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
778 break;
779 case Opt_check_integrity:
780 btrfs_info(info, "enabling check integrity");
781 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
782 break;
783 case Opt_check_integrity_print_mask:
784 ret = match_int(&args[0], &intarg);
785 if (ret)
786 goto out;
787 info->check_integrity_print_mask = intarg;
788 btrfs_info(info, "check_integrity_print_mask 0x%x",
789 info->check_integrity_print_mask);
790 break;
791 #else
792 case Opt_check_integrity_including_extent_data:
793 case Opt_check_integrity:
794 case Opt_check_integrity_print_mask:
795 btrfs_err(info,
796 "support for check_integrity* not compiled in!");
797 ret = -EINVAL;
798 goto out;
799 #endif
800 case Opt_fatal_errors:
801 if (strcmp(args[0].from, "panic") == 0)
802 btrfs_set_opt(info->mount_opt,
803 PANIC_ON_FATAL_ERROR);
804 else if (strcmp(args[0].from, "bug") == 0)
805 btrfs_clear_opt(info->mount_opt,
806 PANIC_ON_FATAL_ERROR);
807 else {
808 ret = -EINVAL;
809 goto out;
810 }
811 break;
812 case Opt_commit_interval:
813 intarg = 0;
814 ret = match_int(&args[0], &intarg);
815 if (ret)
816 goto out;
817 if (intarg == 0) {
818 btrfs_info(info,
819 "using default commit interval %us",
820 BTRFS_DEFAULT_COMMIT_INTERVAL);
821 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
822 } else if (intarg > 300) {
823 btrfs_warn(info, "excessive commit interval %d",
824 intarg);
825 }
826 info->commit_interval = intarg;
827 break;
828 #ifdef CONFIG_BTRFS_DEBUG
829 case Opt_fragment_all:
830 btrfs_info(info, "fragmenting all space");
831 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
832 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
833 break;
834 case Opt_fragment_metadata:
835 btrfs_info(info, "fragmenting metadata");
836 btrfs_set_opt(info->mount_opt,
837 FRAGMENT_METADATA);
838 break;
839 case Opt_fragment_data:
840 btrfs_info(info, "fragmenting data");
841 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
842 break;
843 #endif
844 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
845 case Opt_ref_verify:
846 btrfs_info(info, "doing ref verification");
847 btrfs_set_opt(info->mount_opt, REF_VERIFY);
848 break;
849 #endif
850 case Opt_err:
851 btrfs_info(info, "unrecognized mount option '%s'", p);
852 ret = -EINVAL;
853 goto out;
854 default:
855 break;
856 }
857 }
858 check:
859 /*
860 * Extra check for current option against current flag
861 */
862 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
863 btrfs_err(info,
864 "nologreplay must be used with ro mount option");
865 ret = -EINVAL;
866 }
867 out:
868 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
869 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
870 !btrfs_test_opt(info, CLEAR_CACHE)) {
871 btrfs_err(info, "cannot disable free space tree");
872 ret = -EINVAL;
873
874 }
875 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
876 btrfs_info(info, "disk space caching is enabled");
877 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
878 btrfs_info(info, "using free space tree");
879 return ret;
880 }
881
882 /*
883 * Parse mount options that are required early in the mount process.
884 *
885 * All other options will be parsed on much later in the mount process and
886 * only when we need to allocate a new super block.
887 */
888 static int btrfs_parse_early_options(const char *options, fmode_t flags,
889 void *holder, struct btrfs_fs_devices **fs_devices)
890 {
891 substring_t args[MAX_OPT_ARGS];
892 char *device_name, *opts, *orig, *p;
893 int error = 0;
894
895 lockdep_assert_held(&uuid_mutex);
896
897 if (!options)
898 return 0;
899
900 /*
901 * strsep changes the string, duplicate it because btrfs_parse_options
902 * gets called later
903 */
904 opts = kstrdup(options, GFP_KERNEL);
905 if (!opts)
906 return -ENOMEM;
907 orig = opts;
908
909 while ((p = strsep(&opts, ",")) != NULL) {
910 int token;
911
912 if (!*p)
913 continue;
914
915 token = match_token(p, tokens, args);
916 if (token == Opt_device) {
917 device_name = match_strdup(&args[0]);
918 if (!device_name) {
919 error = -ENOMEM;
920 goto out;
921 }
922 error = btrfs_scan_one_device(device_name,
923 flags, holder, fs_devices);
924 kfree(device_name);
925 if (error)
926 goto out;
927 }
928 }
929
930 out:
931 kfree(orig);
932 return error;
933 }
934
935 /*
936 * Parse mount options that are related to subvolume id
937 *
938 * The value is later passed to mount_subvol()
939 */
940 static int btrfs_parse_subvol_options(const char *options, fmode_t flags,
941 char **subvol_name, u64 *subvol_objectid)
942 {
943 substring_t args[MAX_OPT_ARGS];
944 char *opts, *orig, *p;
945 int error = 0;
946 u64 subvolid;
947
948 if (!options)
949 return 0;
950
951 /*
952 * strsep changes the string, duplicate it because
953 * btrfs_parse_early_options gets called later
954 */
955 opts = kstrdup(options, GFP_KERNEL);
956 if (!opts)
957 return -ENOMEM;
958 orig = opts;
959
960 while ((p = strsep(&opts, ",")) != NULL) {
961 int token;
962 if (!*p)
963 continue;
964
965 token = match_token(p, tokens, args);
966 switch (token) {
967 case Opt_subvol:
968 kfree(*subvol_name);
969 *subvol_name = match_strdup(&args[0]);
970 if (!*subvol_name) {
971 error = -ENOMEM;
972 goto out;
973 }
974 break;
975 case Opt_subvolid:
976 error = match_u64(&args[0], &subvolid);
977 if (error)
978 goto out;
979
980 /* we want the original fs_tree */
981 if (subvolid == 0)
982 subvolid = BTRFS_FS_TREE_OBJECTID;
983
984 *subvol_objectid = subvolid;
985 break;
986 case Opt_subvolrootid:
987 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
988 break;
989 default:
990 break;
991 }
992 }
993
994 out:
995 kfree(orig);
996 return error;
997 }
998
999 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1000 u64 subvol_objectid)
1001 {
1002 struct btrfs_root *root = fs_info->tree_root;
1003 struct btrfs_root *fs_root;
1004 struct btrfs_root_ref *root_ref;
1005 struct btrfs_inode_ref *inode_ref;
1006 struct btrfs_key key;
1007 struct btrfs_path *path = NULL;
1008 char *name = NULL, *ptr;
1009 u64 dirid;
1010 int len;
1011 int ret;
1012
1013 path = btrfs_alloc_path();
1014 if (!path) {
1015 ret = -ENOMEM;
1016 goto err;
1017 }
1018 path->leave_spinning = 1;
1019
1020 name = kmalloc(PATH_MAX, GFP_KERNEL);
1021 if (!name) {
1022 ret = -ENOMEM;
1023 goto err;
1024 }
1025 ptr = name + PATH_MAX - 1;
1026 ptr[0] = '\0';
1027
1028 /*
1029 * Walk up the subvolume trees in the tree of tree roots by root
1030 * backrefs until we hit the top-level subvolume.
1031 */
1032 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1033 key.objectid = subvol_objectid;
1034 key.type = BTRFS_ROOT_BACKREF_KEY;
1035 key.offset = (u64)-1;
1036
1037 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1038 if (ret < 0) {
1039 goto err;
1040 } else if (ret > 0) {
1041 ret = btrfs_previous_item(root, path, subvol_objectid,
1042 BTRFS_ROOT_BACKREF_KEY);
1043 if (ret < 0) {
1044 goto err;
1045 } else if (ret > 0) {
1046 ret = -ENOENT;
1047 goto err;
1048 }
1049 }
1050
1051 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1052 subvol_objectid = key.offset;
1053
1054 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1055 struct btrfs_root_ref);
1056 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1057 ptr -= len + 1;
1058 if (ptr < name) {
1059 ret = -ENAMETOOLONG;
1060 goto err;
1061 }
1062 read_extent_buffer(path->nodes[0], ptr + 1,
1063 (unsigned long)(root_ref + 1), len);
1064 ptr[0] = '/';
1065 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1066 btrfs_release_path(path);
1067
1068 key.objectid = subvol_objectid;
1069 key.type = BTRFS_ROOT_ITEM_KEY;
1070 key.offset = (u64)-1;
1071 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1072 if (IS_ERR(fs_root)) {
1073 ret = PTR_ERR(fs_root);
1074 goto err;
1075 }
1076
1077 /*
1078 * Walk up the filesystem tree by inode refs until we hit the
1079 * root directory.
1080 */
1081 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1082 key.objectid = dirid;
1083 key.type = BTRFS_INODE_REF_KEY;
1084 key.offset = (u64)-1;
1085
1086 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1087 if (ret < 0) {
1088 goto err;
1089 } else if (ret > 0) {
1090 ret = btrfs_previous_item(fs_root, path, dirid,
1091 BTRFS_INODE_REF_KEY);
1092 if (ret < 0) {
1093 goto err;
1094 } else if (ret > 0) {
1095 ret = -ENOENT;
1096 goto err;
1097 }
1098 }
1099
1100 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1101 dirid = key.offset;
1102
1103 inode_ref = btrfs_item_ptr(path->nodes[0],
1104 path->slots[0],
1105 struct btrfs_inode_ref);
1106 len = btrfs_inode_ref_name_len(path->nodes[0],
1107 inode_ref);
1108 ptr -= len + 1;
1109 if (ptr < name) {
1110 ret = -ENAMETOOLONG;
1111 goto err;
1112 }
1113 read_extent_buffer(path->nodes[0], ptr + 1,
1114 (unsigned long)(inode_ref + 1), len);
1115 ptr[0] = '/';
1116 btrfs_release_path(path);
1117 }
1118 }
1119
1120 btrfs_free_path(path);
1121 if (ptr == name + PATH_MAX - 1) {
1122 name[0] = '/';
1123 name[1] = '\0';
1124 } else {
1125 memmove(name, ptr, name + PATH_MAX - ptr);
1126 }
1127 return name;
1128
1129 err:
1130 btrfs_free_path(path);
1131 kfree(name);
1132 return ERR_PTR(ret);
1133 }
1134
1135 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1136 {
1137 struct btrfs_root *root = fs_info->tree_root;
1138 struct btrfs_dir_item *di;
1139 struct btrfs_path *path;
1140 struct btrfs_key location;
1141 u64 dir_id;
1142
1143 path = btrfs_alloc_path();
1144 if (!path)
1145 return -ENOMEM;
1146 path->leave_spinning = 1;
1147
1148 /*
1149 * Find the "default" dir item which points to the root item that we
1150 * will mount by default if we haven't been given a specific subvolume
1151 * to mount.
1152 */
1153 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1154 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1155 if (IS_ERR(di)) {
1156 btrfs_free_path(path);
1157 return PTR_ERR(di);
1158 }
1159 if (!di) {
1160 /*
1161 * Ok the default dir item isn't there. This is weird since
1162 * it's always been there, but don't freak out, just try and
1163 * mount the top-level subvolume.
1164 */
1165 btrfs_free_path(path);
1166 *objectid = BTRFS_FS_TREE_OBJECTID;
1167 return 0;
1168 }
1169
1170 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1171 btrfs_free_path(path);
1172 *objectid = location.objectid;
1173 return 0;
1174 }
1175
1176 static int btrfs_fill_super(struct super_block *sb,
1177 struct btrfs_fs_devices *fs_devices,
1178 void *data)
1179 {
1180 struct inode *inode;
1181 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1182 struct btrfs_key key;
1183 int err;
1184
1185 sb->s_maxbytes = MAX_LFS_FILESIZE;
1186 sb->s_magic = BTRFS_SUPER_MAGIC;
1187 sb->s_op = &btrfs_super_ops;
1188 sb->s_d_op = &btrfs_dentry_operations;
1189 sb->s_export_op = &btrfs_export_ops;
1190 sb->s_xattr = btrfs_xattr_handlers;
1191 sb->s_time_gran = 1;
1192 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1193 sb->s_flags |= SB_POSIXACL;
1194 #endif
1195 sb->s_flags |= SB_I_VERSION;
1196 sb->s_iflags |= SB_I_CGROUPWB;
1197
1198 err = super_setup_bdi(sb);
1199 if (err) {
1200 btrfs_err(fs_info, "super_setup_bdi failed");
1201 return err;
1202 }
1203
1204 err = open_ctree(sb, fs_devices, (char *)data);
1205 if (err) {
1206 btrfs_err(fs_info, "open_ctree failed");
1207 return err;
1208 }
1209
1210 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1211 key.type = BTRFS_INODE_ITEM_KEY;
1212 key.offset = 0;
1213 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1214 if (IS_ERR(inode)) {
1215 err = PTR_ERR(inode);
1216 goto fail_close;
1217 }
1218
1219 sb->s_root = d_make_root(inode);
1220 if (!sb->s_root) {
1221 err = -ENOMEM;
1222 goto fail_close;
1223 }
1224
1225 cleancache_init_fs(sb);
1226 sb->s_flags |= SB_ACTIVE;
1227 return 0;
1228
1229 fail_close:
1230 close_ctree(fs_info);
1231 return err;
1232 }
1233
1234 int btrfs_sync_fs(struct super_block *sb, int wait)
1235 {
1236 struct btrfs_trans_handle *trans;
1237 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1238 struct btrfs_root *root = fs_info->tree_root;
1239
1240 trace_btrfs_sync_fs(fs_info, wait);
1241
1242 if (!wait) {
1243 filemap_flush(fs_info->btree_inode->i_mapping);
1244 return 0;
1245 }
1246
1247 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1248
1249 trans = btrfs_attach_transaction_barrier(root);
1250 if (IS_ERR(trans)) {
1251 /* no transaction, don't bother */
1252 if (PTR_ERR(trans) == -ENOENT) {
1253 /*
1254 * Exit unless we have some pending changes
1255 * that need to go through commit
1256 */
1257 if (fs_info->pending_changes == 0)
1258 return 0;
1259 /*
1260 * A non-blocking test if the fs is frozen. We must not
1261 * start a new transaction here otherwise a deadlock
1262 * happens. The pending operations are delayed to the
1263 * next commit after thawing.
1264 */
1265 if (sb_start_write_trylock(sb))
1266 sb_end_write(sb);
1267 else
1268 return 0;
1269 trans = btrfs_start_transaction(root, 0);
1270 }
1271 if (IS_ERR(trans))
1272 return PTR_ERR(trans);
1273 }
1274 return btrfs_commit_transaction(trans);
1275 }
1276
1277 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1278 {
1279 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1280 const char *compress_type;
1281
1282 if (btrfs_test_opt(info, DEGRADED))
1283 seq_puts(seq, ",degraded");
1284 if (btrfs_test_opt(info, NODATASUM))
1285 seq_puts(seq, ",nodatasum");
1286 if (btrfs_test_opt(info, NODATACOW))
1287 seq_puts(seq, ",nodatacow");
1288 if (btrfs_test_opt(info, NOBARRIER))
1289 seq_puts(seq, ",nobarrier");
1290 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1291 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1292 if (info->thread_pool_size != min_t(unsigned long,
1293 num_online_cpus() + 2, 8))
1294 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1295 if (btrfs_test_opt(info, COMPRESS)) {
1296 compress_type = btrfs_compress_type2str(info->compress_type);
1297 if (btrfs_test_opt(info, FORCE_COMPRESS))
1298 seq_printf(seq, ",compress-force=%s", compress_type);
1299 else
1300 seq_printf(seq, ",compress=%s", compress_type);
1301 if (info->compress_level)
1302 seq_printf(seq, ":%d", info->compress_level);
1303 }
1304 if (btrfs_test_opt(info, NOSSD))
1305 seq_puts(seq, ",nossd");
1306 if (btrfs_test_opt(info, SSD_SPREAD))
1307 seq_puts(seq, ",ssd_spread");
1308 else if (btrfs_test_opt(info, SSD))
1309 seq_puts(seq, ",ssd");
1310 if (btrfs_test_opt(info, NOTREELOG))
1311 seq_puts(seq, ",notreelog");
1312 if (btrfs_test_opt(info, NOLOGREPLAY))
1313 seq_puts(seq, ",nologreplay");
1314 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1315 seq_puts(seq, ",flushoncommit");
1316 if (btrfs_test_opt(info, DISCARD))
1317 seq_puts(seq, ",discard");
1318 if (!(info->sb->s_flags & SB_POSIXACL))
1319 seq_puts(seq, ",noacl");
1320 if (btrfs_test_opt(info, SPACE_CACHE))
1321 seq_puts(seq, ",space_cache");
1322 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1323 seq_puts(seq, ",space_cache=v2");
1324 else
1325 seq_puts(seq, ",nospace_cache");
1326 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1327 seq_puts(seq, ",rescan_uuid_tree");
1328 if (btrfs_test_opt(info, CLEAR_CACHE))
1329 seq_puts(seq, ",clear_cache");
1330 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1331 seq_puts(seq, ",user_subvol_rm_allowed");
1332 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1333 seq_puts(seq, ",enospc_debug");
1334 if (btrfs_test_opt(info, AUTO_DEFRAG))
1335 seq_puts(seq, ",autodefrag");
1336 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1337 seq_puts(seq, ",inode_cache");
1338 if (btrfs_test_opt(info, SKIP_BALANCE))
1339 seq_puts(seq, ",skip_balance");
1340 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1341 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1342 seq_puts(seq, ",check_int_data");
1343 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1344 seq_puts(seq, ",check_int");
1345 if (info->check_integrity_print_mask)
1346 seq_printf(seq, ",check_int_print_mask=%d",
1347 info->check_integrity_print_mask);
1348 #endif
1349 if (info->metadata_ratio)
1350 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1351 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1352 seq_puts(seq, ",fatal_errors=panic");
1353 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1354 seq_printf(seq, ",commit=%u", info->commit_interval);
1355 #ifdef CONFIG_BTRFS_DEBUG
1356 if (btrfs_test_opt(info, FRAGMENT_DATA))
1357 seq_puts(seq, ",fragment=data");
1358 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1359 seq_puts(seq, ",fragment=metadata");
1360 #endif
1361 if (btrfs_test_opt(info, REF_VERIFY))
1362 seq_puts(seq, ",ref_verify");
1363 seq_printf(seq, ",subvolid=%llu",
1364 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1365 seq_puts(seq, ",subvol=");
1366 seq_dentry(seq, dentry, " \t\n\\");
1367 return 0;
1368 }
1369
1370 static int btrfs_test_super(struct super_block *s, void *data)
1371 {
1372 struct btrfs_fs_info *p = data;
1373 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1374
1375 return fs_info->fs_devices == p->fs_devices;
1376 }
1377
1378 static int btrfs_set_super(struct super_block *s, void *data)
1379 {
1380 int err = set_anon_super(s, data);
1381 if (!err)
1382 s->s_fs_info = data;
1383 return err;
1384 }
1385
1386 /*
1387 * subvolumes are identified by ino 256
1388 */
1389 static inline int is_subvolume_inode(struct inode *inode)
1390 {
1391 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1392 return 1;
1393 return 0;
1394 }
1395
1396 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1397 const char *device_name, struct vfsmount *mnt)
1398 {
1399 struct dentry *root;
1400 int ret;
1401
1402 if (!subvol_name) {
1403 if (!subvol_objectid) {
1404 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1405 &subvol_objectid);
1406 if (ret) {
1407 root = ERR_PTR(ret);
1408 goto out;
1409 }
1410 }
1411 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1412 subvol_objectid);
1413 if (IS_ERR(subvol_name)) {
1414 root = ERR_CAST(subvol_name);
1415 subvol_name = NULL;
1416 goto out;
1417 }
1418
1419 }
1420
1421 root = mount_subtree(mnt, subvol_name);
1422 /* mount_subtree() drops our reference on the vfsmount. */
1423 mnt = NULL;
1424
1425 if (!IS_ERR(root)) {
1426 struct super_block *s = root->d_sb;
1427 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1428 struct inode *root_inode = d_inode(root);
1429 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1430
1431 ret = 0;
1432 if (!is_subvolume_inode(root_inode)) {
1433 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1434 subvol_name);
1435 ret = -EINVAL;
1436 }
1437 if (subvol_objectid && root_objectid != subvol_objectid) {
1438 /*
1439 * This will also catch a race condition where a
1440 * subvolume which was passed by ID is renamed and
1441 * another subvolume is renamed over the old location.
1442 */
1443 btrfs_err(fs_info,
1444 "subvol '%s' does not match subvolid %llu",
1445 subvol_name, subvol_objectid);
1446 ret = -EINVAL;
1447 }
1448 if (ret) {
1449 dput(root);
1450 root = ERR_PTR(ret);
1451 deactivate_locked_super(s);
1452 }
1453 }
1454
1455 out:
1456 mntput(mnt);
1457 kfree(subvol_name);
1458 return root;
1459 }
1460
1461 static int parse_security_options(char *orig_opts,
1462 struct security_mnt_opts *sec_opts)
1463 {
1464 char *secdata = NULL;
1465 int ret = 0;
1466
1467 secdata = alloc_secdata();
1468 if (!secdata)
1469 return -ENOMEM;
1470 ret = security_sb_copy_data(orig_opts, secdata);
1471 if (ret) {
1472 free_secdata(secdata);
1473 return ret;
1474 }
1475 ret = security_sb_parse_opts_str(secdata, sec_opts);
1476 free_secdata(secdata);
1477 return ret;
1478 }
1479
1480 static int setup_security_options(struct btrfs_fs_info *fs_info,
1481 struct super_block *sb,
1482 struct security_mnt_opts *sec_opts)
1483 {
1484 int ret = 0;
1485
1486 /*
1487 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1488 * is valid.
1489 */
1490 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1491 if (ret)
1492 return ret;
1493
1494 #ifdef CONFIG_SECURITY
1495 if (!fs_info->security_opts.num_mnt_opts) {
1496 /* first time security setup, copy sec_opts to fs_info */
1497 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1498 } else {
1499 /*
1500 * Since SELinux (the only one supporting security_mnt_opts)
1501 * does NOT support changing context during remount/mount of
1502 * the same sb, this must be the same or part of the same
1503 * security options, just free it.
1504 */
1505 security_free_mnt_opts(sec_opts);
1506 }
1507 #endif
1508 return ret;
1509 }
1510
1511 /*
1512 * Find a superblock for the given device / mount point.
1513 *
1514 * Note: This is based on mount_bdev from fs/super.c with a few additions
1515 * for multiple device setup. Make sure to keep it in sync.
1516 */
1517 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1518 int flags, const char *device_name, void *data)
1519 {
1520 struct block_device *bdev = NULL;
1521 struct super_block *s;
1522 struct btrfs_fs_devices *fs_devices = NULL;
1523 struct btrfs_fs_info *fs_info = NULL;
1524 struct security_mnt_opts new_sec_opts;
1525 fmode_t mode = FMODE_READ;
1526 int error = 0;
1527
1528 if (!(flags & SB_RDONLY))
1529 mode |= FMODE_WRITE;
1530
1531 security_init_mnt_opts(&new_sec_opts);
1532 if (data) {
1533 error = parse_security_options(data, &new_sec_opts);
1534 if (error)
1535 return ERR_PTR(error);
1536 }
1537
1538 /*
1539 * Setup a dummy root and fs_info for test/set super. This is because
1540 * we don't actually fill this stuff out until open_ctree, but we need
1541 * it for searching for existing supers, so this lets us do that and
1542 * then open_ctree will properly initialize everything later.
1543 */
1544 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1545 if (!fs_info) {
1546 error = -ENOMEM;
1547 goto error_sec_opts;
1548 }
1549
1550 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1551 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1552 security_init_mnt_opts(&fs_info->security_opts);
1553 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1554 error = -ENOMEM;
1555 goto error_fs_info;
1556 }
1557
1558 mutex_lock(&uuid_mutex);
1559 error = btrfs_parse_early_options(data, mode, fs_type, &fs_devices);
1560 if (error) {
1561 mutex_unlock(&uuid_mutex);
1562 goto error_fs_info;
1563 }
1564
1565 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1566 if (error) {
1567 mutex_unlock(&uuid_mutex);
1568 goto error_fs_info;
1569 }
1570
1571 fs_info->fs_devices = fs_devices;
1572
1573 error = btrfs_open_devices(fs_devices, mode, fs_type);
1574 mutex_unlock(&uuid_mutex);
1575 if (error)
1576 goto error_fs_info;
1577
1578 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1579 error = -EACCES;
1580 goto error_close_devices;
1581 }
1582
1583 bdev = fs_devices->latest_bdev;
1584 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1585 fs_info);
1586 if (IS_ERR(s)) {
1587 error = PTR_ERR(s);
1588 goto error_close_devices;
1589 }
1590
1591 if (s->s_root) {
1592 btrfs_close_devices(fs_devices);
1593 free_fs_info(fs_info);
1594 if ((flags ^ s->s_flags) & SB_RDONLY)
1595 error = -EBUSY;
1596 } else {
1597 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1598 btrfs_sb(s)->bdev_holder = fs_type;
1599 error = btrfs_fill_super(s, fs_devices, data);
1600 }
1601 if (error) {
1602 deactivate_locked_super(s);
1603 goto error_sec_opts;
1604 }
1605
1606 fs_info = btrfs_sb(s);
1607 error = setup_security_options(fs_info, s, &new_sec_opts);
1608 if (error) {
1609 deactivate_locked_super(s);
1610 goto error_sec_opts;
1611 }
1612
1613 return dget(s->s_root);
1614
1615 error_close_devices:
1616 btrfs_close_devices(fs_devices);
1617 error_fs_info:
1618 free_fs_info(fs_info);
1619 error_sec_opts:
1620 security_free_mnt_opts(&new_sec_opts);
1621 return ERR_PTR(error);
1622 }
1623
1624 /*
1625 * Mount function which is called by VFS layer.
1626 *
1627 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1628 * which needs vfsmount* of device's root (/). This means device's root has to
1629 * be mounted internally in any case.
1630 *
1631 * Operation flow:
1632 * 1. Parse subvol id related options for later use in mount_subvol().
1633 *
1634 * 2. Mount device's root (/) by calling vfs_kern_mount().
1635 *
1636 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1637 * first place. In order to avoid calling btrfs_mount() again, we use
1638 * different file_system_type which is not registered to VFS by
1639 * register_filesystem() (btrfs_root_fs_type). As a result,
1640 * btrfs_mount_root() is called. The return value will be used by
1641 * mount_subtree() in mount_subvol().
1642 *
1643 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1644 * "btrfs subvolume set-default", mount_subvol() is called always.
1645 */
1646 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1647 const char *device_name, void *data)
1648 {
1649 struct vfsmount *mnt_root;
1650 struct dentry *root;
1651 fmode_t mode = FMODE_READ;
1652 char *subvol_name = NULL;
1653 u64 subvol_objectid = 0;
1654 int error = 0;
1655
1656 if (!(flags & SB_RDONLY))
1657 mode |= FMODE_WRITE;
1658
1659 error = btrfs_parse_subvol_options(data, mode,
1660 &subvol_name, &subvol_objectid);
1661 if (error) {
1662 kfree(subvol_name);
1663 return ERR_PTR(error);
1664 }
1665
1666 /* mount device's root (/) */
1667 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1668 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1669 if (flags & SB_RDONLY) {
1670 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1671 flags & ~SB_RDONLY, device_name, data);
1672 } else {
1673 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1674 flags | SB_RDONLY, device_name, data);
1675 if (IS_ERR(mnt_root)) {
1676 root = ERR_CAST(mnt_root);
1677 goto out;
1678 }
1679
1680 down_write(&mnt_root->mnt_sb->s_umount);
1681 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1682 up_write(&mnt_root->mnt_sb->s_umount);
1683 if (error < 0) {
1684 root = ERR_PTR(error);
1685 mntput(mnt_root);
1686 goto out;
1687 }
1688 }
1689 }
1690 if (IS_ERR(mnt_root)) {
1691 root = ERR_CAST(mnt_root);
1692 goto out;
1693 }
1694
1695 /* mount_subvol() will free subvol_name and mnt_root */
1696 root = mount_subvol(subvol_name, subvol_objectid, device_name, mnt_root);
1697
1698 out:
1699 return root;
1700 }
1701
1702 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1703 u32 new_pool_size, u32 old_pool_size)
1704 {
1705 if (new_pool_size == old_pool_size)
1706 return;
1707
1708 fs_info->thread_pool_size = new_pool_size;
1709
1710 btrfs_info(fs_info, "resize thread pool %d -> %d",
1711 old_pool_size, new_pool_size);
1712
1713 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1714 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1715 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1716 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1717 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1718 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1719 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1720 new_pool_size);
1721 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1722 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1723 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1724 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1725 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1726 new_pool_size);
1727 }
1728
1729 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1730 {
1731 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1732 }
1733
1734 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1735 unsigned long old_opts, int flags)
1736 {
1737 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1738 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1739 (flags & SB_RDONLY))) {
1740 /* wait for any defraggers to finish */
1741 wait_event(fs_info->transaction_wait,
1742 (atomic_read(&fs_info->defrag_running) == 0));
1743 if (flags & SB_RDONLY)
1744 sync_filesystem(fs_info->sb);
1745 }
1746 }
1747
1748 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1749 unsigned long old_opts)
1750 {
1751 /*
1752 * We need to cleanup all defragable inodes if the autodefragment is
1753 * close or the filesystem is read only.
1754 */
1755 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1756 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1757 btrfs_cleanup_defrag_inodes(fs_info);
1758 }
1759
1760 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1761 }
1762
1763 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1764 {
1765 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1766 struct btrfs_root *root = fs_info->tree_root;
1767 unsigned old_flags = sb->s_flags;
1768 unsigned long old_opts = fs_info->mount_opt;
1769 unsigned long old_compress_type = fs_info->compress_type;
1770 u64 old_max_inline = fs_info->max_inline;
1771 u32 old_thread_pool_size = fs_info->thread_pool_size;
1772 u32 old_metadata_ratio = fs_info->metadata_ratio;
1773 int ret;
1774
1775 sync_filesystem(sb);
1776 btrfs_remount_prepare(fs_info);
1777
1778 if (data) {
1779 struct security_mnt_opts new_sec_opts;
1780
1781 security_init_mnt_opts(&new_sec_opts);
1782 ret = parse_security_options(data, &new_sec_opts);
1783 if (ret)
1784 goto restore;
1785 ret = setup_security_options(fs_info, sb,
1786 &new_sec_opts);
1787 if (ret) {
1788 security_free_mnt_opts(&new_sec_opts);
1789 goto restore;
1790 }
1791 }
1792
1793 ret = btrfs_parse_options(fs_info, data, *flags);
1794 if (ret)
1795 goto restore;
1796
1797 btrfs_remount_begin(fs_info, old_opts, *flags);
1798 btrfs_resize_thread_pool(fs_info,
1799 fs_info->thread_pool_size, old_thread_pool_size);
1800
1801 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1802 goto out;
1803
1804 if (*flags & SB_RDONLY) {
1805 /*
1806 * this also happens on 'umount -rf' or on shutdown, when
1807 * the filesystem is busy.
1808 */
1809 cancel_work_sync(&fs_info->async_reclaim_work);
1810
1811 /* wait for the uuid_scan task to finish */
1812 down(&fs_info->uuid_tree_rescan_sem);
1813 /* avoid complains from lockdep et al. */
1814 up(&fs_info->uuid_tree_rescan_sem);
1815
1816 sb->s_flags |= SB_RDONLY;
1817
1818 /*
1819 * Setting SB_RDONLY will put the cleaner thread to
1820 * sleep at the next loop if it's already active.
1821 * If it's already asleep, we'll leave unused block
1822 * groups on disk until we're mounted read-write again
1823 * unless we clean them up here.
1824 */
1825 btrfs_delete_unused_bgs(fs_info);
1826
1827 btrfs_dev_replace_suspend_for_unmount(fs_info);
1828 btrfs_scrub_cancel(fs_info);
1829 btrfs_pause_balance(fs_info);
1830
1831 ret = btrfs_commit_super(fs_info);
1832 if (ret)
1833 goto restore;
1834 } else {
1835 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1836 btrfs_err(fs_info,
1837 "Remounting read-write after error is not allowed");
1838 ret = -EINVAL;
1839 goto restore;
1840 }
1841 if (fs_info->fs_devices->rw_devices == 0) {
1842 ret = -EACCES;
1843 goto restore;
1844 }
1845
1846 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1847 btrfs_warn(fs_info,
1848 "too many missing devices, writeable remount is not allowed");
1849 ret = -EACCES;
1850 goto restore;
1851 }
1852
1853 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1854 ret = -EINVAL;
1855 goto restore;
1856 }
1857
1858 ret = btrfs_cleanup_fs_roots(fs_info);
1859 if (ret)
1860 goto restore;
1861
1862 /* recover relocation */
1863 mutex_lock(&fs_info->cleaner_mutex);
1864 ret = btrfs_recover_relocation(root);
1865 mutex_unlock(&fs_info->cleaner_mutex);
1866 if (ret)
1867 goto restore;
1868
1869 ret = btrfs_resume_balance_async(fs_info);
1870 if (ret)
1871 goto restore;
1872
1873 ret = btrfs_resume_dev_replace_async(fs_info);
1874 if (ret) {
1875 btrfs_warn(fs_info, "failed to resume dev_replace");
1876 goto restore;
1877 }
1878
1879 btrfs_qgroup_rescan_resume(fs_info);
1880
1881 if (!fs_info->uuid_root) {
1882 btrfs_info(fs_info, "creating UUID tree");
1883 ret = btrfs_create_uuid_tree(fs_info);
1884 if (ret) {
1885 btrfs_warn(fs_info,
1886 "failed to create the UUID tree %d",
1887 ret);
1888 goto restore;
1889 }
1890 }
1891 sb->s_flags &= ~SB_RDONLY;
1892
1893 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1894 }
1895 out:
1896 wake_up_process(fs_info->transaction_kthread);
1897 btrfs_remount_cleanup(fs_info, old_opts);
1898 return 0;
1899
1900 restore:
1901 /* We've hit an error - don't reset SB_RDONLY */
1902 if (sb_rdonly(sb))
1903 old_flags |= SB_RDONLY;
1904 sb->s_flags = old_flags;
1905 fs_info->mount_opt = old_opts;
1906 fs_info->compress_type = old_compress_type;
1907 fs_info->max_inline = old_max_inline;
1908 btrfs_resize_thread_pool(fs_info,
1909 old_thread_pool_size, fs_info->thread_pool_size);
1910 fs_info->metadata_ratio = old_metadata_ratio;
1911 btrfs_remount_cleanup(fs_info, old_opts);
1912 return ret;
1913 }
1914
1915 /* Used to sort the devices by max_avail(descending sort) */
1916 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1917 const void *dev_info2)
1918 {
1919 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1920 ((struct btrfs_device_info *)dev_info2)->max_avail)
1921 return -1;
1922 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1923 ((struct btrfs_device_info *)dev_info2)->max_avail)
1924 return 1;
1925 else
1926 return 0;
1927 }
1928
1929 /*
1930 * sort the devices by max_avail, in which max free extent size of each device
1931 * is stored.(Descending Sort)
1932 */
1933 static inline void btrfs_descending_sort_devices(
1934 struct btrfs_device_info *devices,
1935 size_t nr_devices)
1936 {
1937 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1938 btrfs_cmp_device_free_bytes, NULL);
1939 }
1940
1941 /*
1942 * The helper to calc the free space on the devices that can be used to store
1943 * file data.
1944 */
1945 static int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1946 u64 *free_bytes)
1947 {
1948 struct btrfs_device_info *devices_info;
1949 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1950 struct btrfs_device *device;
1951 u64 skip_space;
1952 u64 type;
1953 u64 avail_space;
1954 u64 min_stripe_size;
1955 int min_stripes = 1, num_stripes = 1;
1956 int i = 0, nr_devices;
1957
1958 /*
1959 * We aren't under the device list lock, so this is racy-ish, but good
1960 * enough for our purposes.
1961 */
1962 nr_devices = fs_info->fs_devices->open_devices;
1963 if (!nr_devices) {
1964 smp_mb();
1965 nr_devices = fs_info->fs_devices->open_devices;
1966 ASSERT(nr_devices);
1967 if (!nr_devices) {
1968 *free_bytes = 0;
1969 return 0;
1970 }
1971 }
1972
1973 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1974 GFP_KERNEL);
1975 if (!devices_info)
1976 return -ENOMEM;
1977
1978 /* calc min stripe number for data space allocation */
1979 type = btrfs_data_alloc_profile(fs_info);
1980 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1981 min_stripes = 2;
1982 num_stripes = nr_devices;
1983 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1984 min_stripes = 2;
1985 num_stripes = 2;
1986 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1987 min_stripes = 4;
1988 num_stripes = 4;
1989 }
1990
1991 if (type & BTRFS_BLOCK_GROUP_DUP)
1992 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1993 else
1994 min_stripe_size = BTRFS_STRIPE_LEN;
1995
1996 rcu_read_lock();
1997 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1998 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1999 &device->dev_state) ||
2000 !device->bdev ||
2001 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2002 continue;
2003
2004 if (i >= nr_devices)
2005 break;
2006
2007 avail_space = device->total_bytes - device->bytes_used;
2008
2009 /* align with stripe_len */
2010 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
2011 avail_space *= BTRFS_STRIPE_LEN;
2012
2013 /*
2014 * In order to avoid overwriting the superblock on the drive,
2015 * btrfs starts at an offset of at least 1MB when doing chunk
2016 * allocation.
2017 */
2018 skip_space = SZ_1M;
2019
2020 /*
2021 * we can use the free space in [0, skip_space - 1], subtract
2022 * it from the total.
2023 */
2024 if (avail_space && avail_space >= skip_space)
2025 avail_space -= skip_space;
2026 else
2027 avail_space = 0;
2028
2029 if (avail_space < min_stripe_size)
2030 continue;
2031
2032 devices_info[i].dev = device;
2033 devices_info[i].max_avail = avail_space;
2034
2035 i++;
2036 }
2037 rcu_read_unlock();
2038
2039 nr_devices = i;
2040
2041 btrfs_descending_sort_devices(devices_info, nr_devices);
2042
2043 i = nr_devices - 1;
2044 avail_space = 0;
2045 while (nr_devices >= min_stripes) {
2046 if (num_stripes > nr_devices)
2047 num_stripes = nr_devices;
2048
2049 if (devices_info[i].max_avail >= min_stripe_size) {
2050 int j;
2051 u64 alloc_size;
2052
2053 avail_space += devices_info[i].max_avail * num_stripes;
2054 alloc_size = devices_info[i].max_avail;
2055 for (j = i + 1 - num_stripes; j <= i; j++)
2056 devices_info[j].max_avail -= alloc_size;
2057 }
2058 i--;
2059 nr_devices--;
2060 }
2061
2062 kfree(devices_info);
2063 *free_bytes = avail_space;
2064 return 0;
2065 }
2066
2067 /*
2068 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2069 *
2070 * If there's a redundant raid level at DATA block groups, use the respective
2071 * multiplier to scale the sizes.
2072 *
2073 * Unused device space usage is based on simulating the chunk allocator
2074 * algorithm that respects the device sizes and order of allocations. This is
2075 * a close approximation of the actual use but there are other factors that may
2076 * change the result (like a new metadata chunk).
2077 *
2078 * If metadata is exhausted, f_bavail will be 0.
2079 */
2080 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2081 {
2082 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2083 struct btrfs_super_block *disk_super = fs_info->super_copy;
2084 struct list_head *head = &fs_info->space_info;
2085 struct btrfs_space_info *found;
2086 u64 total_used = 0;
2087 u64 total_free_data = 0;
2088 u64 total_free_meta = 0;
2089 int bits = dentry->d_sb->s_blocksize_bits;
2090 __be32 *fsid = (__be32 *)fs_info->fsid;
2091 unsigned factor = 1;
2092 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2093 int ret;
2094 u64 thresh = 0;
2095 int mixed = 0;
2096
2097 rcu_read_lock();
2098 list_for_each_entry_rcu(found, head, list) {
2099 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2100 int i;
2101
2102 total_free_data += found->disk_total - found->disk_used;
2103 total_free_data -=
2104 btrfs_account_ro_block_groups_free_space(found);
2105
2106 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2107 if (!list_empty(&found->block_groups[i])) {
2108 switch (i) {
2109 case BTRFS_RAID_DUP:
2110 case BTRFS_RAID_RAID1:
2111 case BTRFS_RAID_RAID10:
2112 factor = 2;
2113 }
2114 }
2115 }
2116 }
2117
2118 /*
2119 * Metadata in mixed block goup profiles are accounted in data
2120 */
2121 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2122 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2123 mixed = 1;
2124 else
2125 total_free_meta += found->disk_total -
2126 found->disk_used;
2127 }
2128
2129 total_used += found->disk_used;
2130 }
2131
2132 rcu_read_unlock();
2133
2134 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2135 buf->f_blocks >>= bits;
2136 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2137
2138 /* Account global block reserve as used, it's in logical size already */
2139 spin_lock(&block_rsv->lock);
2140 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2141 if (buf->f_bfree >= block_rsv->size >> bits)
2142 buf->f_bfree -= block_rsv->size >> bits;
2143 else
2144 buf->f_bfree = 0;
2145 spin_unlock(&block_rsv->lock);
2146
2147 buf->f_bavail = div_u64(total_free_data, factor);
2148 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2149 if (ret)
2150 return ret;
2151 buf->f_bavail += div_u64(total_free_data, factor);
2152 buf->f_bavail = buf->f_bavail >> bits;
2153
2154 /*
2155 * We calculate the remaining metadata space minus global reserve. If
2156 * this is (supposedly) smaller than zero, there's no space. But this
2157 * does not hold in practice, the exhausted state happens where's still
2158 * some positive delta. So we apply some guesswork and compare the
2159 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2160 *
2161 * We probably cannot calculate the exact threshold value because this
2162 * depends on the internal reservations requested by various
2163 * operations, so some operations that consume a few metadata will
2164 * succeed even if the Avail is zero. But this is better than the other
2165 * way around.
2166 */
2167 thresh = SZ_4M;
2168
2169 if (!mixed && total_free_meta - thresh < block_rsv->size)
2170 buf->f_bavail = 0;
2171
2172 buf->f_type = BTRFS_SUPER_MAGIC;
2173 buf->f_bsize = dentry->d_sb->s_blocksize;
2174 buf->f_namelen = BTRFS_NAME_LEN;
2175
2176 /* We treat it as constant endianness (it doesn't matter _which_)
2177 because we want the fsid to come out the same whether mounted
2178 on a big-endian or little-endian host */
2179 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2180 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2181 /* Mask in the root object ID too, to disambiguate subvols */
2182 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2183 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2184
2185 return 0;
2186 }
2187
2188 static void btrfs_kill_super(struct super_block *sb)
2189 {
2190 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2191 kill_anon_super(sb);
2192 free_fs_info(fs_info);
2193 }
2194
2195 static struct file_system_type btrfs_fs_type = {
2196 .owner = THIS_MODULE,
2197 .name = "btrfs",
2198 .mount = btrfs_mount,
2199 .kill_sb = btrfs_kill_super,
2200 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2201 };
2202
2203 static struct file_system_type btrfs_root_fs_type = {
2204 .owner = THIS_MODULE,
2205 .name = "btrfs",
2206 .mount = btrfs_mount_root,
2207 .kill_sb = btrfs_kill_super,
2208 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2209 };
2210
2211 MODULE_ALIAS_FS("btrfs");
2212
2213 static int btrfs_control_open(struct inode *inode, struct file *file)
2214 {
2215 /*
2216 * The control file's private_data is used to hold the
2217 * transaction when it is started and is used to keep
2218 * track of whether a transaction is already in progress.
2219 */
2220 file->private_data = NULL;
2221 return 0;
2222 }
2223
2224 /*
2225 * used by btrfsctl to scan devices when no FS is mounted
2226 */
2227 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2228 unsigned long arg)
2229 {
2230 struct btrfs_ioctl_vol_args *vol;
2231 struct btrfs_fs_devices *fs_devices;
2232 int ret = -ENOTTY;
2233
2234 if (!capable(CAP_SYS_ADMIN))
2235 return -EPERM;
2236
2237 vol = memdup_user((void __user *)arg, sizeof(*vol));
2238 if (IS_ERR(vol))
2239 return PTR_ERR(vol);
2240
2241 switch (cmd) {
2242 case BTRFS_IOC_SCAN_DEV:
2243 mutex_lock(&uuid_mutex);
2244 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2245 &btrfs_root_fs_type, &fs_devices);
2246 mutex_unlock(&uuid_mutex);
2247 break;
2248 case BTRFS_IOC_DEVICES_READY:
2249 mutex_lock(&uuid_mutex);
2250 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2251 &btrfs_root_fs_type, &fs_devices);
2252 if (ret) {
2253 mutex_unlock(&uuid_mutex);
2254 break;
2255 }
2256 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2257 mutex_unlock(&uuid_mutex);
2258 break;
2259 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2260 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2261 break;
2262 }
2263
2264 kfree(vol);
2265 return ret;
2266 }
2267
2268 static int btrfs_freeze(struct super_block *sb)
2269 {
2270 struct btrfs_trans_handle *trans;
2271 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2272 struct btrfs_root *root = fs_info->tree_root;
2273
2274 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2275 /*
2276 * We don't need a barrier here, we'll wait for any transaction that
2277 * could be in progress on other threads (and do delayed iputs that
2278 * we want to avoid on a frozen filesystem), or do the commit
2279 * ourselves.
2280 */
2281 trans = btrfs_attach_transaction_barrier(root);
2282 if (IS_ERR(trans)) {
2283 /* no transaction, don't bother */
2284 if (PTR_ERR(trans) == -ENOENT)
2285 return 0;
2286 return PTR_ERR(trans);
2287 }
2288 return btrfs_commit_transaction(trans);
2289 }
2290
2291 static int btrfs_unfreeze(struct super_block *sb)
2292 {
2293 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2294
2295 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2296 return 0;
2297 }
2298
2299 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2300 {
2301 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2302 struct btrfs_fs_devices *cur_devices;
2303 struct btrfs_device *dev, *first_dev = NULL;
2304 struct list_head *head;
2305 struct rcu_string *name;
2306
2307 /*
2308 * Lightweight locking of the devices. We should not need
2309 * device_list_mutex here as we only read the device data and the list
2310 * is protected by RCU. Even if a device is deleted during the list
2311 * traversals, we'll get valid data, the freeing callback will wait at
2312 * least until until the rcu_read_unlock.
2313 */
2314 rcu_read_lock();
2315 cur_devices = fs_info->fs_devices;
2316 while (cur_devices) {
2317 head = &cur_devices->devices;
2318 list_for_each_entry_rcu(dev, head, dev_list) {
2319 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2320 continue;
2321 if (!dev->name)
2322 continue;
2323 if (!first_dev || dev->devid < first_dev->devid)
2324 first_dev = dev;
2325 }
2326 cur_devices = cur_devices->seed;
2327 }
2328
2329 if (first_dev) {
2330 name = rcu_dereference(first_dev->name);
2331 seq_escape(m, name->str, " \t\n\\");
2332 } else {
2333 WARN_ON(1);
2334 }
2335 rcu_read_unlock();
2336 return 0;
2337 }
2338
2339 static const struct super_operations btrfs_super_ops = {
2340 .drop_inode = btrfs_drop_inode,
2341 .evict_inode = btrfs_evict_inode,
2342 .put_super = btrfs_put_super,
2343 .sync_fs = btrfs_sync_fs,
2344 .show_options = btrfs_show_options,
2345 .show_devname = btrfs_show_devname,
2346 .alloc_inode = btrfs_alloc_inode,
2347 .destroy_inode = btrfs_destroy_inode,
2348 .statfs = btrfs_statfs,
2349 .remount_fs = btrfs_remount,
2350 .freeze_fs = btrfs_freeze,
2351 .unfreeze_fs = btrfs_unfreeze,
2352 };
2353
2354 static const struct file_operations btrfs_ctl_fops = {
2355 .open = btrfs_control_open,
2356 .unlocked_ioctl = btrfs_control_ioctl,
2357 .compat_ioctl = btrfs_control_ioctl,
2358 .owner = THIS_MODULE,
2359 .llseek = noop_llseek,
2360 };
2361
2362 static struct miscdevice btrfs_misc = {
2363 .minor = BTRFS_MINOR,
2364 .name = "btrfs-control",
2365 .fops = &btrfs_ctl_fops
2366 };
2367
2368 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2369 MODULE_ALIAS("devname:btrfs-control");
2370
2371 static int __init btrfs_interface_init(void)
2372 {
2373 return misc_register(&btrfs_misc);
2374 }
2375
2376 static __cold void btrfs_interface_exit(void)
2377 {
2378 misc_deregister(&btrfs_misc);
2379 }
2380
2381 static void __init btrfs_print_mod_info(void)
2382 {
2383 static const char options[] = ""
2384 #ifdef CONFIG_BTRFS_DEBUG
2385 ", debug=on"
2386 #endif
2387 #ifdef CONFIG_BTRFS_ASSERT
2388 ", assert=on"
2389 #endif
2390 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2391 ", integrity-checker=on"
2392 #endif
2393 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2394 ", ref-verify=on"
2395 #endif
2396 ;
2397 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2398 }
2399
2400 static int __init init_btrfs_fs(void)
2401 {
2402 int err;
2403
2404 btrfs_props_init();
2405
2406 err = btrfs_init_sysfs();
2407 if (err)
2408 return err;
2409
2410 btrfs_init_compress();
2411
2412 err = btrfs_init_cachep();
2413 if (err)
2414 goto free_compress;
2415
2416 err = extent_io_init();
2417 if (err)
2418 goto free_cachep;
2419
2420 err = extent_map_init();
2421 if (err)
2422 goto free_extent_io;
2423
2424 err = ordered_data_init();
2425 if (err)
2426 goto free_extent_map;
2427
2428 err = btrfs_delayed_inode_init();
2429 if (err)
2430 goto free_ordered_data;
2431
2432 err = btrfs_auto_defrag_init();
2433 if (err)
2434 goto free_delayed_inode;
2435
2436 err = btrfs_delayed_ref_init();
2437 if (err)
2438 goto free_auto_defrag;
2439
2440 err = btrfs_prelim_ref_init();
2441 if (err)
2442 goto free_delayed_ref;
2443
2444 err = btrfs_end_io_wq_init();
2445 if (err)
2446 goto free_prelim_ref;
2447
2448 err = btrfs_interface_init();
2449 if (err)
2450 goto free_end_io_wq;
2451
2452 btrfs_init_lockdep();
2453
2454 btrfs_print_mod_info();
2455
2456 err = btrfs_run_sanity_tests();
2457 if (err)
2458 goto unregister_ioctl;
2459
2460 err = register_filesystem(&btrfs_fs_type);
2461 if (err)
2462 goto unregister_ioctl;
2463
2464 return 0;
2465
2466 unregister_ioctl:
2467 btrfs_interface_exit();
2468 free_end_io_wq:
2469 btrfs_end_io_wq_exit();
2470 free_prelim_ref:
2471 btrfs_prelim_ref_exit();
2472 free_delayed_ref:
2473 btrfs_delayed_ref_exit();
2474 free_auto_defrag:
2475 btrfs_auto_defrag_exit();
2476 free_delayed_inode:
2477 btrfs_delayed_inode_exit();
2478 free_ordered_data:
2479 ordered_data_exit();
2480 free_extent_map:
2481 extent_map_exit();
2482 free_extent_io:
2483 extent_io_exit();
2484 free_cachep:
2485 btrfs_destroy_cachep();
2486 free_compress:
2487 btrfs_exit_compress();
2488 btrfs_exit_sysfs();
2489
2490 return err;
2491 }
2492
2493 static void __exit exit_btrfs_fs(void)
2494 {
2495 btrfs_destroy_cachep();
2496 btrfs_delayed_ref_exit();
2497 btrfs_auto_defrag_exit();
2498 btrfs_delayed_inode_exit();
2499 btrfs_prelim_ref_exit();
2500 ordered_data_exit();
2501 extent_map_exit();
2502 extent_io_exit();
2503 btrfs_interface_exit();
2504 btrfs_end_io_wq_exit();
2505 unregister_filesystem(&btrfs_fs_type);
2506 btrfs_exit_sysfs();
2507 btrfs_cleanup_fs_uuids();
2508 btrfs_exit_compress();
2509 }
2510
2511 late_initcall(init_btrfs_fs);
2512 module_exit(exit_btrfs_fs)
2513
2514 MODULE_LICENSE("GPL");
Linux with added FPC-III support