Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (C) 2007 Oracle. All rights reserved.
4 */
5
6#include <linux/blkdev.h>
7#include <linux/module.h>
8#include <linux/fs.h>
9#include <linux/pagemap.h>
10#include <linux/highmem.h>
11#include <linux/time.h>
12#include <linux/init.h>
13#include <linux/seq_file.h>
14#include <linux/string.h>
15#include <linux/backing-dev.h>
16#include <linux/mount.h>
17#include <linux/writeback.h>
18#include <linux/statfs.h>
19#include <linux/compat.h>
20#include <linux/parser.h>
21#include <linux/ctype.h>
22#include <linux/namei.h>
23#include <linux/miscdevice.h>
24#include <linux/magic.h>
25#include <linux/slab.h>
26#include <linux/cleancache.h>
27#include <linux/ratelimit.h>
28#include <linux/crc32c.h>
29#include <linux/btrfs.h>
30#include "delayed-inode.h"
31#include "ctree.h"
32#include "disk-io.h"
33#include "transaction.h"
34#include "btrfs_inode.h"
35#include "print-tree.h"
36#include "props.h"
37#include "xattr.h"
38#include "volumes.h"
39#include "export.h"
40#include "compression.h"
41#include "rcu-string.h"
42#include "dev-replace.h"
43#include "free-space-cache.h"
44#include "backref.h"
45#include "space-info.h"
46#include "sysfs.h"
47#include "tests/btrfs-tests.h"
48#include "block-group.h"
49#include "discard.h"
50
51#include "qgroup.h"
52#define CREATE_TRACE_POINTS
53#include <trace/events/btrfs.h>
54
55static const struct super_operations btrfs_super_ops;
56
57/*
58 * Types for mounting the default subvolume and a subvolume explicitly
59 * requested by subvol=/path. That way the callchain is straightforward and we
60 * don't have to play tricks with the mount options and recursive calls to
61 * btrfs_mount.
62 *
63 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
64 */
65static struct file_system_type btrfs_fs_type;
66static struct file_system_type btrfs_root_fs_type;
67
68static int btrfs_remount(struct super_block *sb, int *flags, char *data);
69
70/*
71 * Generally the error codes correspond to their respective errors, but there
72 * are a few special cases.
73 *
74 * EUCLEAN: Any sort of corruption that we encounter. The tree-checker for
75 * instance will return EUCLEAN if any of the blocks are corrupted in
76 * a way that is problematic. We want to reserve EUCLEAN for these
77 * sort of corruptions.
78 *
79 * EROFS: If we check BTRFS_FS_STATE_ERROR and fail out with a return error, we
80 * need to use EROFS for this case. We will have no idea of the
81 * original failure, that will have been reported at the time we tripped
82 * over the error. Each subsequent error that doesn't have any context
83 * of the original error should use EROFS when handling BTRFS_FS_STATE_ERROR.
84 */
85const char * __attribute_const__ btrfs_decode_error(int errno)
86{
87 char *errstr = "unknown";
88
89 switch (errno) {
90 case -ENOENT: /* -2 */
91 errstr = "No such entry";
92 break;
93 case -EIO: /* -5 */
94 errstr = "IO failure";
95 break;
96 case -ENOMEM: /* -12*/
97 errstr = "Out of memory";
98 break;
99 case -EEXIST: /* -17 */
100 errstr = "Object already exists";
101 break;
102 case -ENOSPC: /* -28 */
103 errstr = "No space left";
104 break;
105 case -EROFS: /* -30 */
106 errstr = "Readonly filesystem";
107 break;
108 case -EOPNOTSUPP: /* -95 */
109 errstr = "Operation not supported";
110 break;
111 case -EUCLEAN: /* -117 */
112 errstr = "Filesystem corrupted";
113 break;
114 case -EDQUOT: /* -122 */
115 errstr = "Quota exceeded";
116 break;
117 }
118
119 return errstr;
120}
121
122/*
123 * __btrfs_handle_fs_error decodes expected errors from the caller and
124 * invokes the appropriate error response.
125 */
126__cold
127void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
128 unsigned int line, int errno, const char *fmt, ...)
129{
130 struct super_block *sb = fs_info->sb;
131#ifdef CONFIG_PRINTK
132 const char *errstr;
133#endif
134
135 /*
136 * Special case: if the error is EROFS, and we're already
137 * under SB_RDONLY, then it is safe here.
138 */
139 if (errno == -EROFS && sb_rdonly(sb))
140 return;
141
142#ifdef CONFIG_PRINTK
143 errstr = btrfs_decode_error(errno);
144 if (fmt) {
145 struct va_format vaf;
146 va_list args;
147
148 va_start(args, fmt);
149 vaf.fmt = fmt;
150 vaf.va = &args;
151
152 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
153 sb->s_id, function, line, errno, errstr, &vaf);
154 va_end(args);
155 } else {
156 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
157 sb->s_id, function, line, errno, errstr);
158 }
159#endif
160
161 /*
162 * Today we only save the error info to memory. Long term we'll
163 * also send it down to the disk
164 */
165 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
166
167 /* Don't go through full error handling during mount */
168 if (!(sb->s_flags & SB_BORN))
169 return;
170
171 if (sb_rdonly(sb))
172 return;
173
174 btrfs_discard_stop(fs_info);
175
176 /* btrfs handle error by forcing the filesystem readonly */
177 sb->s_flags |= SB_RDONLY;
178 btrfs_info(fs_info, "forced readonly");
179 /*
180 * Note that a running device replace operation is not canceled here
181 * although there is no way to update the progress. It would add the
182 * risk of a deadlock, therefore the canceling is omitted. The only
183 * penalty is that some I/O remains active until the procedure
184 * completes. The next time when the filesystem is mounted writable
185 * again, the device replace operation continues.
186 */
187}
188
189#ifdef CONFIG_PRINTK
190static const char * const logtypes[] = {
191 "emergency",
192 "alert",
193 "critical",
194 "error",
195 "warning",
196 "notice",
197 "info",
198 "debug",
199};
200
201
202/*
203 * Use one ratelimit state per log level so that a flood of less important
204 * messages doesn't cause more important ones to be dropped.
205 */
206static struct ratelimit_state printk_limits[] = {
207 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
208 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
209 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
210 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
211 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
212 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
213 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
214 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
215};
216
217void __cold btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
218{
219 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
220 struct va_format vaf;
221 va_list args;
222 int kern_level;
223 const char *type = logtypes[4];
224 struct ratelimit_state *ratelimit = &printk_limits[4];
225
226 va_start(args, fmt);
227
228 while ((kern_level = printk_get_level(fmt)) != 0) {
229 size_t size = printk_skip_level(fmt) - fmt;
230
231 if (kern_level >= '0' && kern_level <= '7') {
232 memcpy(lvl, fmt, size);
233 lvl[size] = '\0';
234 type = logtypes[kern_level - '0'];
235 ratelimit = &printk_limits[kern_level - '0'];
236 }
237 fmt += size;
238 }
239
240 vaf.fmt = fmt;
241 vaf.va = &args;
242
243 if (__ratelimit(ratelimit))
244 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
245 fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
246
247 va_end(args);
248}
249#endif
250
251/*
252 * We only mark the transaction aborted and then set the file system read-only.
253 * This will prevent new transactions from starting or trying to join this
254 * one.
255 *
256 * This means that error recovery at the call site is limited to freeing
257 * any local memory allocations and passing the error code up without
258 * further cleanup. The transaction should complete as it normally would
259 * in the call path but will return -EIO.
260 *
261 * We'll complete the cleanup in btrfs_end_transaction and
262 * btrfs_commit_transaction.
263 */
264__cold
265void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
266 const char *function,
267 unsigned int line, int errno)
268{
269 struct btrfs_fs_info *fs_info = trans->fs_info;
270
271 WRITE_ONCE(trans->aborted, errno);
272 /* Nothing used. The other threads that have joined this
273 * transaction may be able to continue. */
274 if (!trans->dirty && list_empty(&trans->new_bgs)) {
275 const char *errstr;
276
277 errstr = btrfs_decode_error(errno);
278 btrfs_warn(fs_info,
279 "%s:%d: Aborting unused transaction(%s).",
280 function, line, errstr);
281 return;
282 }
283 WRITE_ONCE(trans->transaction->aborted, errno);
284 /* Wake up anybody who may be waiting on this transaction */
285 wake_up(&fs_info->transaction_wait);
286 wake_up(&fs_info->transaction_blocked_wait);
287 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
288}
289/*
290 * __btrfs_panic decodes unexpected, fatal errors from the caller,
291 * issues an alert, and either panics or BUGs, depending on mount options.
292 */
293__cold
294void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
295 unsigned int line, int errno, const char *fmt, ...)
296{
297 char *s_id = "<unknown>";
298 const char *errstr;
299 struct va_format vaf = { .fmt = fmt };
300 va_list args;
301
302 if (fs_info)
303 s_id = fs_info->sb->s_id;
304
305 va_start(args, fmt);
306 vaf.va = &args;
307
308 errstr = btrfs_decode_error(errno);
309 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
310 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
311 s_id, function, line, &vaf, errno, errstr);
312
313 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
314 function, line, &vaf, errno, errstr);
315 va_end(args);
316 /* Caller calls BUG() */
317}
318
319static void btrfs_put_super(struct super_block *sb)
320{
321 close_ctree(btrfs_sb(sb));
322}
323
324enum {
325 Opt_acl, Opt_noacl,
326 Opt_clear_cache,
327 Opt_commit_interval,
328 Opt_compress,
329 Opt_compress_force,
330 Opt_compress_force_type,
331 Opt_compress_type,
332 Opt_degraded,
333 Opt_device,
334 Opt_fatal_errors,
335 Opt_flushoncommit, Opt_noflushoncommit,
336 Opt_inode_cache, Opt_noinode_cache,
337 Opt_max_inline,
338 Opt_barrier, Opt_nobarrier,
339 Opt_datacow, Opt_nodatacow,
340 Opt_datasum, Opt_nodatasum,
341 Opt_defrag, Opt_nodefrag,
342 Opt_discard, Opt_nodiscard,
343 Opt_discard_mode,
344 Opt_norecovery,
345 Opt_ratio,
346 Opt_rescan_uuid_tree,
347 Opt_skip_balance,
348 Opt_space_cache, Opt_no_space_cache,
349 Opt_space_cache_version,
350 Opt_ssd, Opt_nossd,
351 Opt_ssd_spread, Opt_nossd_spread,
352 Opt_subvol,
353 Opt_subvol_empty,
354 Opt_subvolid,
355 Opt_thread_pool,
356 Opt_treelog, Opt_notreelog,
357 Opt_user_subvol_rm_allowed,
358
359 /* Rescue options */
360 Opt_rescue,
361 Opt_usebackuproot,
362 Opt_nologreplay,
363
364 /* Deprecated options */
365 Opt_recovery,
366
367 /* Debugging options */
368 Opt_check_integrity,
369 Opt_check_integrity_including_extent_data,
370 Opt_check_integrity_print_mask,
371 Opt_enospc_debug, Opt_noenospc_debug,
372#ifdef CONFIG_BTRFS_DEBUG
373 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
374#endif
375#ifdef CONFIG_BTRFS_FS_REF_VERIFY
376 Opt_ref_verify,
377#endif
378 Opt_err,
379};
380
381static const match_table_t tokens = {
382 {Opt_acl, "acl"},
383 {Opt_noacl, "noacl"},
384 {Opt_clear_cache, "clear_cache"},
385 {Opt_commit_interval, "commit=%u"},
386 {Opt_compress, "compress"},
387 {Opt_compress_type, "compress=%s"},
388 {Opt_compress_force, "compress-force"},
389 {Opt_compress_force_type, "compress-force=%s"},
390 {Opt_degraded, "degraded"},
391 {Opt_device, "device=%s"},
392 {Opt_fatal_errors, "fatal_errors=%s"},
393 {Opt_flushoncommit, "flushoncommit"},
394 {Opt_noflushoncommit, "noflushoncommit"},
395 {Opt_inode_cache, "inode_cache"},
396 {Opt_noinode_cache, "noinode_cache"},
397 {Opt_max_inline, "max_inline=%s"},
398 {Opt_barrier, "barrier"},
399 {Opt_nobarrier, "nobarrier"},
400 {Opt_datacow, "datacow"},
401 {Opt_nodatacow, "nodatacow"},
402 {Opt_datasum, "datasum"},
403 {Opt_nodatasum, "nodatasum"},
404 {Opt_defrag, "autodefrag"},
405 {Opt_nodefrag, "noautodefrag"},
406 {Opt_discard, "discard"},
407 {Opt_discard_mode, "discard=%s"},
408 {Opt_nodiscard, "nodiscard"},
409 {Opt_norecovery, "norecovery"},
410 {Opt_ratio, "metadata_ratio=%u"},
411 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
412 {Opt_skip_balance, "skip_balance"},
413 {Opt_space_cache, "space_cache"},
414 {Opt_no_space_cache, "nospace_cache"},
415 {Opt_space_cache_version, "space_cache=%s"},
416 {Opt_ssd, "ssd"},
417 {Opt_nossd, "nossd"},
418 {Opt_ssd_spread, "ssd_spread"},
419 {Opt_nossd_spread, "nossd_spread"},
420 {Opt_subvol, "subvol=%s"},
421 {Opt_subvol_empty, "subvol="},
422 {Opt_subvolid, "subvolid=%s"},
423 {Opt_thread_pool, "thread_pool=%u"},
424 {Opt_treelog, "treelog"},
425 {Opt_notreelog, "notreelog"},
426 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
427
428 /* Rescue options */
429 {Opt_rescue, "rescue=%s"},
430 /* Deprecated, with alias rescue=nologreplay */
431 {Opt_nologreplay, "nologreplay"},
432 /* Deprecated, with alias rescue=usebackuproot */
433 {Opt_usebackuproot, "usebackuproot"},
434
435 /* Deprecated options */
436 {Opt_recovery, "recovery"},
437
438 /* Debugging options */
439 {Opt_check_integrity, "check_int"},
440 {Opt_check_integrity_including_extent_data, "check_int_data"},
441 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
442 {Opt_enospc_debug, "enospc_debug"},
443 {Opt_noenospc_debug, "noenospc_debug"},
444#ifdef CONFIG_BTRFS_DEBUG
445 {Opt_fragment_data, "fragment=data"},
446 {Opt_fragment_metadata, "fragment=metadata"},
447 {Opt_fragment_all, "fragment=all"},
448#endif
449#ifdef CONFIG_BTRFS_FS_REF_VERIFY
450 {Opt_ref_verify, "ref_verify"},
451#endif
452 {Opt_err, NULL},
453};
454
455static const match_table_t rescue_tokens = {
456 {Opt_usebackuproot, "usebackuproot"},
457 {Opt_nologreplay, "nologreplay"},
458 {Opt_err, NULL},
459};
460
461static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
462{
463 char *opts;
464 char *orig;
465 char *p;
466 substring_t args[MAX_OPT_ARGS];
467 int ret = 0;
468
469 opts = kstrdup(options, GFP_KERNEL);
470 if (!opts)
471 return -ENOMEM;
472 orig = opts;
473
474 while ((p = strsep(&opts, ":")) != NULL) {
475 int token;
476
477 if (!*p)
478 continue;
479 token = match_token(p, rescue_tokens, args);
480 switch (token){
481 case Opt_usebackuproot:
482 btrfs_info(info,
483 "trying to use backup root at mount time");
484 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
485 break;
486 case Opt_nologreplay:
487 btrfs_set_and_info(info, NOLOGREPLAY,
488 "disabling log replay at mount time");
489 break;
490 case Opt_err:
491 btrfs_info(info, "unrecognized rescue option '%s'", p);
492 ret = -EINVAL;
493 goto out;
494 default:
495 break;
496 }
497
498 }
499out:
500 kfree(orig);
501 return ret;
502}
503
504/*
505 * Regular mount options parser. Everything that is needed only when
506 * reading in a new superblock is parsed here.
507 * XXX JDM: This needs to be cleaned up for remount.
508 */
509int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
510 unsigned long new_flags)
511{
512 substring_t args[MAX_OPT_ARGS];
513 char *p, *num;
514 u64 cache_gen;
515 int intarg;
516 int ret = 0;
517 char *compress_type;
518 bool compress_force = false;
519 enum btrfs_compression_type saved_compress_type;
520 int saved_compress_level;
521 bool saved_compress_force;
522 int no_compress = 0;
523
524 cache_gen = btrfs_super_cache_generation(info->super_copy);
525 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
526 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
527 else if (cache_gen)
528 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
529
530 /*
531 * Even the options are empty, we still need to do extra check
532 * against new flags
533 */
534 if (!options)
535 goto check;
536
537 while ((p = strsep(&options, ",")) != NULL) {
538 int token;
539 if (!*p)
540 continue;
541
542 token = match_token(p, tokens, args);
543 switch (token) {
544 case Opt_degraded:
545 btrfs_info(info, "allowing degraded mounts");
546 btrfs_set_opt(info->mount_opt, DEGRADED);
547 break;
548 case Opt_subvol:
549 case Opt_subvol_empty:
550 case Opt_subvolid:
551 case Opt_device:
552 /*
553 * These are parsed by btrfs_parse_subvol_options or
554 * btrfs_parse_device_options and can be ignored here.
555 */
556 break;
557 case Opt_nodatasum:
558 btrfs_set_and_info(info, NODATASUM,
559 "setting nodatasum");
560 break;
561 case Opt_datasum:
562 if (btrfs_test_opt(info, NODATASUM)) {
563 if (btrfs_test_opt(info, NODATACOW))
564 btrfs_info(info,
565 "setting datasum, datacow enabled");
566 else
567 btrfs_info(info, "setting datasum");
568 }
569 btrfs_clear_opt(info->mount_opt, NODATACOW);
570 btrfs_clear_opt(info->mount_opt, NODATASUM);
571 break;
572 case Opt_nodatacow:
573 if (!btrfs_test_opt(info, NODATACOW)) {
574 if (!btrfs_test_opt(info, COMPRESS) ||
575 !btrfs_test_opt(info, FORCE_COMPRESS)) {
576 btrfs_info(info,
577 "setting nodatacow, compression disabled");
578 } else {
579 btrfs_info(info, "setting nodatacow");
580 }
581 }
582 btrfs_clear_opt(info->mount_opt, COMPRESS);
583 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
584 btrfs_set_opt(info->mount_opt, NODATACOW);
585 btrfs_set_opt(info->mount_opt, NODATASUM);
586 break;
587 case Opt_datacow:
588 btrfs_clear_and_info(info, NODATACOW,
589 "setting datacow");
590 break;
591 case Opt_compress_force:
592 case Opt_compress_force_type:
593 compress_force = true;
594 fallthrough;
595 case Opt_compress:
596 case Opt_compress_type:
597 saved_compress_type = btrfs_test_opt(info,
598 COMPRESS) ?
599 info->compress_type : BTRFS_COMPRESS_NONE;
600 saved_compress_force =
601 btrfs_test_opt(info, FORCE_COMPRESS);
602 saved_compress_level = info->compress_level;
603 if (token == Opt_compress ||
604 token == Opt_compress_force ||
605 strncmp(args[0].from, "zlib", 4) == 0) {
606 compress_type = "zlib";
607
608 info->compress_type = BTRFS_COMPRESS_ZLIB;
609 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
610 /*
611 * args[0] contains uninitialized data since
612 * for these tokens we don't expect any
613 * parameter.
614 */
615 if (token != Opt_compress &&
616 token != Opt_compress_force)
617 info->compress_level =
618 btrfs_compress_str2level(
619 BTRFS_COMPRESS_ZLIB,
620 args[0].from + 4);
621 btrfs_set_opt(info->mount_opt, COMPRESS);
622 btrfs_clear_opt(info->mount_opt, NODATACOW);
623 btrfs_clear_opt(info->mount_opt, NODATASUM);
624 no_compress = 0;
625 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
626 compress_type = "lzo";
627 info->compress_type = BTRFS_COMPRESS_LZO;
628 info->compress_level = 0;
629 btrfs_set_opt(info->mount_opt, COMPRESS);
630 btrfs_clear_opt(info->mount_opt, NODATACOW);
631 btrfs_clear_opt(info->mount_opt, NODATASUM);
632 btrfs_set_fs_incompat(info, COMPRESS_LZO);
633 no_compress = 0;
634 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
635 compress_type = "zstd";
636 info->compress_type = BTRFS_COMPRESS_ZSTD;
637 info->compress_level =
638 btrfs_compress_str2level(
639 BTRFS_COMPRESS_ZSTD,
640 args[0].from + 4);
641 btrfs_set_opt(info->mount_opt, COMPRESS);
642 btrfs_clear_opt(info->mount_opt, NODATACOW);
643 btrfs_clear_opt(info->mount_opt, NODATASUM);
644 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
645 no_compress = 0;
646 } else if (strncmp(args[0].from, "no", 2) == 0) {
647 compress_type = "no";
648 info->compress_level = 0;
649 info->compress_type = 0;
650 btrfs_clear_opt(info->mount_opt, COMPRESS);
651 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
652 compress_force = false;
653 no_compress++;
654 } else {
655 ret = -EINVAL;
656 goto out;
657 }
658
659 if (compress_force) {
660 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
661 } else {
662 /*
663 * If we remount from compress-force=xxx to
664 * compress=xxx, we need clear FORCE_COMPRESS
665 * flag, otherwise, there is no way for users
666 * to disable forcible compression separately.
667 */
668 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
669 }
670 if (no_compress == 1) {
671 btrfs_info(info, "use no compression");
672 } else if ((info->compress_type != saved_compress_type) ||
673 (compress_force != saved_compress_force) ||
674 (info->compress_level != saved_compress_level)) {
675 btrfs_info(info, "%s %s compression, level %d",
676 (compress_force) ? "force" : "use",
677 compress_type, info->compress_level);
678 }
679 compress_force = false;
680 break;
681 case Opt_ssd:
682 btrfs_set_and_info(info, SSD,
683 "enabling ssd optimizations");
684 btrfs_clear_opt(info->mount_opt, NOSSD);
685 break;
686 case Opt_ssd_spread:
687 btrfs_set_and_info(info, SSD,
688 "enabling ssd optimizations");
689 btrfs_set_and_info(info, SSD_SPREAD,
690 "using spread ssd allocation scheme");
691 btrfs_clear_opt(info->mount_opt, NOSSD);
692 break;
693 case Opt_nossd:
694 btrfs_set_opt(info->mount_opt, NOSSD);
695 btrfs_clear_and_info(info, SSD,
696 "not using ssd optimizations");
697 fallthrough;
698 case Opt_nossd_spread:
699 btrfs_clear_and_info(info, SSD_SPREAD,
700 "not using spread ssd allocation scheme");
701 break;
702 case Opt_barrier:
703 btrfs_clear_and_info(info, NOBARRIER,
704 "turning on barriers");
705 break;
706 case Opt_nobarrier:
707 btrfs_set_and_info(info, NOBARRIER,
708 "turning off barriers");
709 break;
710 case Opt_thread_pool:
711 ret = match_int(&args[0], &intarg);
712 if (ret) {
713 goto out;
714 } else if (intarg == 0) {
715 ret = -EINVAL;
716 goto out;
717 }
718 info->thread_pool_size = intarg;
719 break;
720 case Opt_max_inline:
721 num = match_strdup(&args[0]);
722 if (num) {
723 info->max_inline = memparse(num, NULL);
724 kfree(num);
725
726 if (info->max_inline) {
727 info->max_inline = min_t(u64,
728 info->max_inline,
729 info->sectorsize);
730 }
731 btrfs_info(info, "max_inline at %llu",
732 info->max_inline);
733 } else {
734 ret = -ENOMEM;
735 goto out;
736 }
737 break;
738 case Opt_acl:
739#ifdef CONFIG_BTRFS_FS_POSIX_ACL
740 info->sb->s_flags |= SB_POSIXACL;
741 break;
742#else
743 btrfs_err(info, "support for ACL not compiled in!");
744 ret = -EINVAL;
745 goto out;
746#endif
747 case Opt_noacl:
748 info->sb->s_flags &= ~SB_POSIXACL;
749 break;
750 case Opt_notreelog:
751 btrfs_set_and_info(info, NOTREELOG,
752 "disabling tree log");
753 break;
754 case Opt_treelog:
755 btrfs_clear_and_info(info, NOTREELOG,
756 "enabling tree log");
757 break;
758 case Opt_norecovery:
759 case Opt_nologreplay:
760 btrfs_warn(info,
761 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
762 btrfs_set_and_info(info, NOLOGREPLAY,
763 "disabling log replay at mount time");
764 break;
765 case Opt_flushoncommit:
766 btrfs_set_and_info(info, FLUSHONCOMMIT,
767 "turning on flush-on-commit");
768 break;
769 case Opt_noflushoncommit:
770 btrfs_clear_and_info(info, FLUSHONCOMMIT,
771 "turning off flush-on-commit");
772 break;
773 case Opt_ratio:
774 ret = match_int(&args[0], &intarg);
775 if (ret)
776 goto out;
777 info->metadata_ratio = intarg;
778 btrfs_info(info, "metadata ratio %u",
779 info->metadata_ratio);
780 break;
781 case Opt_discard:
782 case Opt_discard_mode:
783 if (token == Opt_discard ||
784 strcmp(args[0].from, "sync") == 0) {
785 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
786 btrfs_set_and_info(info, DISCARD_SYNC,
787 "turning on sync discard");
788 } else if (strcmp(args[0].from, "async") == 0) {
789 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
790 btrfs_set_and_info(info, DISCARD_ASYNC,
791 "turning on async discard");
792 } else {
793 ret = -EINVAL;
794 goto out;
795 }
796 break;
797 case Opt_nodiscard:
798 btrfs_clear_and_info(info, DISCARD_SYNC,
799 "turning off discard");
800 btrfs_clear_and_info(info, DISCARD_ASYNC,
801 "turning off async discard");
802 break;
803 case Opt_space_cache:
804 case Opt_space_cache_version:
805 if (token == Opt_space_cache ||
806 strcmp(args[0].from, "v1") == 0) {
807 btrfs_clear_opt(info->mount_opt,
808 FREE_SPACE_TREE);
809 btrfs_set_and_info(info, SPACE_CACHE,
810 "enabling disk space caching");
811 } else if (strcmp(args[0].from, "v2") == 0) {
812 btrfs_clear_opt(info->mount_opt,
813 SPACE_CACHE);
814 btrfs_set_and_info(info, FREE_SPACE_TREE,
815 "enabling free space tree");
816 } else {
817 ret = -EINVAL;
818 goto out;
819 }
820 break;
821 case Opt_rescan_uuid_tree:
822 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
823 break;
824 case Opt_no_space_cache:
825 if (btrfs_test_opt(info, SPACE_CACHE)) {
826 btrfs_clear_and_info(info, SPACE_CACHE,
827 "disabling disk space caching");
828 }
829 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
830 btrfs_clear_and_info(info, FREE_SPACE_TREE,
831 "disabling free space tree");
832 }
833 break;
834 case Opt_inode_cache:
835 btrfs_warn(info,
836 "the 'inode_cache' option is deprecated and will have no effect from 5.11");
837 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
838 "enabling inode map caching");
839 break;
840 case Opt_noinode_cache:
841 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
842 "disabling inode map caching");
843 break;
844 case Opt_clear_cache:
845 btrfs_set_and_info(info, CLEAR_CACHE,
846 "force clearing of disk cache");
847 break;
848 case Opt_user_subvol_rm_allowed:
849 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
850 break;
851 case Opt_enospc_debug:
852 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
853 break;
854 case Opt_noenospc_debug:
855 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
856 break;
857 case Opt_defrag:
858 btrfs_set_and_info(info, AUTO_DEFRAG,
859 "enabling auto defrag");
860 break;
861 case Opt_nodefrag:
862 btrfs_clear_and_info(info, AUTO_DEFRAG,
863 "disabling auto defrag");
864 break;
865 case Opt_recovery:
866 case Opt_usebackuproot:
867 btrfs_warn(info,
868 "'%s' is deprecated, use 'rescue=usebackuproot' instead",
869 token == Opt_recovery ? "recovery" :
870 "usebackuproot");
871 btrfs_info(info,
872 "trying to use backup root at mount time");
873 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
874 break;
875 case Opt_skip_balance:
876 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
877 break;
878#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
879 case Opt_check_integrity_including_extent_data:
880 btrfs_info(info,
881 "enabling check integrity including extent data");
882 btrfs_set_opt(info->mount_opt,
883 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
884 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
885 break;
886 case Opt_check_integrity:
887 btrfs_info(info, "enabling check integrity");
888 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
889 break;
890 case Opt_check_integrity_print_mask:
891 ret = match_int(&args[0], &intarg);
892 if (ret)
893 goto out;
894 info->check_integrity_print_mask = intarg;
895 btrfs_info(info, "check_integrity_print_mask 0x%x",
896 info->check_integrity_print_mask);
897 break;
898#else
899 case Opt_check_integrity_including_extent_data:
900 case Opt_check_integrity:
901 case Opt_check_integrity_print_mask:
902 btrfs_err(info,
903 "support for check_integrity* not compiled in!");
904 ret = -EINVAL;
905 goto out;
906#endif
907 case Opt_fatal_errors:
908 if (strcmp(args[0].from, "panic") == 0)
909 btrfs_set_opt(info->mount_opt,
910 PANIC_ON_FATAL_ERROR);
911 else if (strcmp(args[0].from, "bug") == 0)
912 btrfs_clear_opt(info->mount_opt,
913 PANIC_ON_FATAL_ERROR);
914 else {
915 ret = -EINVAL;
916 goto out;
917 }
918 break;
919 case Opt_commit_interval:
920 intarg = 0;
921 ret = match_int(&args[0], &intarg);
922 if (ret)
923 goto out;
924 if (intarg == 0) {
925 btrfs_info(info,
926 "using default commit interval %us",
927 BTRFS_DEFAULT_COMMIT_INTERVAL);
928 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
929 } else if (intarg > 300) {
930 btrfs_warn(info, "excessive commit interval %d",
931 intarg);
932 }
933 info->commit_interval = intarg;
934 break;
935 case Opt_rescue:
936 ret = parse_rescue_options(info, args[0].from);
937 if (ret < 0)
938 goto out;
939 break;
940#ifdef CONFIG_BTRFS_DEBUG
941 case Opt_fragment_all:
942 btrfs_info(info, "fragmenting all space");
943 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
944 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
945 break;
946 case Opt_fragment_metadata:
947 btrfs_info(info, "fragmenting metadata");
948 btrfs_set_opt(info->mount_opt,
949 FRAGMENT_METADATA);
950 break;
951 case Opt_fragment_data:
952 btrfs_info(info, "fragmenting data");
953 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
954 break;
955#endif
956#ifdef CONFIG_BTRFS_FS_REF_VERIFY
957 case Opt_ref_verify:
958 btrfs_info(info, "doing ref verification");
959 btrfs_set_opt(info->mount_opt, REF_VERIFY);
960 break;
961#endif
962 case Opt_err:
963 btrfs_err(info, "unrecognized mount option '%s'", p);
964 ret = -EINVAL;
965 goto out;
966 default:
967 break;
968 }
969 }
970check:
971 /*
972 * Extra check for current option against current flag
973 */
974 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
975 btrfs_err(info,
976 "nologreplay must be used with ro mount option");
977 ret = -EINVAL;
978 }
979out:
980 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
981 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
982 !btrfs_test_opt(info, CLEAR_CACHE)) {
983 btrfs_err(info, "cannot disable free space tree");
984 ret = -EINVAL;
985
986 }
987 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
988 btrfs_info(info, "disk space caching is enabled");
989 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
990 btrfs_info(info, "using free space tree");
991 return ret;
992}
993
994/*
995 * Parse mount options that are required early in the mount process.
996 *
997 * All other options will be parsed on much later in the mount process and
998 * only when we need to allocate a new super block.
999 */
1000static int btrfs_parse_device_options(const char *options, fmode_t flags,
1001 void *holder)
1002{
1003 substring_t args[MAX_OPT_ARGS];
1004 char *device_name, *opts, *orig, *p;
1005 struct btrfs_device *device = NULL;
1006 int error = 0;
1007
1008 lockdep_assert_held(&uuid_mutex);
1009
1010 if (!options)
1011 return 0;
1012
1013 /*
1014 * strsep changes the string, duplicate it because btrfs_parse_options
1015 * gets called later
1016 */
1017 opts = kstrdup(options, GFP_KERNEL);
1018 if (!opts)
1019 return -ENOMEM;
1020 orig = opts;
1021
1022 while ((p = strsep(&opts, ",")) != NULL) {
1023 int token;
1024
1025 if (!*p)
1026 continue;
1027
1028 token = match_token(p, tokens, args);
1029 if (token == Opt_device) {
1030 device_name = match_strdup(&args[0]);
1031 if (!device_name) {
1032 error = -ENOMEM;
1033 goto out;
1034 }
1035 device = btrfs_scan_one_device(device_name, flags,
1036 holder);
1037 kfree(device_name);
1038 if (IS_ERR(device)) {
1039 error = PTR_ERR(device);
1040 goto out;
1041 }
1042 }
1043 }
1044
1045out:
1046 kfree(orig);
1047 return error;
1048}
1049
1050/*
1051 * Parse mount options that are related to subvolume id
1052 *
1053 * The value is later passed to mount_subvol()
1054 */
1055static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1056 u64 *subvol_objectid)
1057{
1058 substring_t args[MAX_OPT_ARGS];
1059 char *opts, *orig, *p;
1060 int error = 0;
1061 u64 subvolid;
1062
1063 if (!options)
1064 return 0;
1065
1066 /*
1067 * strsep changes the string, duplicate it because
1068 * btrfs_parse_device_options gets called later
1069 */
1070 opts = kstrdup(options, GFP_KERNEL);
1071 if (!opts)
1072 return -ENOMEM;
1073 orig = opts;
1074
1075 while ((p = strsep(&opts, ",")) != NULL) {
1076 int token;
1077 if (!*p)
1078 continue;
1079
1080 token = match_token(p, tokens, args);
1081 switch (token) {
1082 case Opt_subvol:
1083 kfree(*subvol_name);
1084 *subvol_name = match_strdup(&args[0]);
1085 if (!*subvol_name) {
1086 error = -ENOMEM;
1087 goto out;
1088 }
1089 break;
1090 case Opt_subvolid:
1091 error = match_u64(&args[0], &subvolid);
1092 if (error)
1093 goto out;
1094
1095 /* we want the original fs_tree */
1096 if (subvolid == 0)
1097 subvolid = BTRFS_FS_TREE_OBJECTID;
1098
1099 *subvol_objectid = subvolid;
1100 break;
1101 default:
1102 break;
1103 }
1104 }
1105
1106out:
1107 kfree(orig);
1108 return error;
1109}
1110
1111char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1112 u64 subvol_objectid)
1113{
1114 struct btrfs_root *root = fs_info->tree_root;
1115 struct btrfs_root *fs_root = NULL;
1116 struct btrfs_root_ref *root_ref;
1117 struct btrfs_inode_ref *inode_ref;
1118 struct btrfs_key key;
1119 struct btrfs_path *path = NULL;
1120 char *name = NULL, *ptr;
1121 u64 dirid;
1122 int len;
1123 int ret;
1124
1125 path = btrfs_alloc_path();
1126 if (!path) {
1127 ret = -ENOMEM;
1128 goto err;
1129 }
1130 path->leave_spinning = 1;
1131
1132 name = kmalloc(PATH_MAX, GFP_KERNEL);
1133 if (!name) {
1134 ret = -ENOMEM;
1135 goto err;
1136 }
1137 ptr = name + PATH_MAX - 1;
1138 ptr[0] = '\0';
1139
1140 /*
1141 * Walk up the subvolume trees in the tree of tree roots by root
1142 * backrefs until we hit the top-level subvolume.
1143 */
1144 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1145 key.objectid = subvol_objectid;
1146 key.type = BTRFS_ROOT_BACKREF_KEY;
1147 key.offset = (u64)-1;
1148
1149 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1150 if (ret < 0) {
1151 goto err;
1152 } else if (ret > 0) {
1153 ret = btrfs_previous_item(root, path, subvol_objectid,
1154 BTRFS_ROOT_BACKREF_KEY);
1155 if (ret < 0) {
1156 goto err;
1157 } else if (ret > 0) {
1158 ret = -ENOENT;
1159 goto err;
1160 }
1161 }
1162
1163 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1164 subvol_objectid = key.offset;
1165
1166 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1167 struct btrfs_root_ref);
1168 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1169 ptr -= len + 1;
1170 if (ptr < name) {
1171 ret = -ENAMETOOLONG;
1172 goto err;
1173 }
1174 read_extent_buffer(path->nodes[0], ptr + 1,
1175 (unsigned long)(root_ref + 1), len);
1176 ptr[0] = '/';
1177 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1178 btrfs_release_path(path);
1179
1180 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1181 if (IS_ERR(fs_root)) {
1182 ret = PTR_ERR(fs_root);
1183 fs_root = NULL;
1184 goto err;
1185 }
1186
1187 /*
1188 * Walk up the filesystem tree by inode refs until we hit the
1189 * root directory.
1190 */
1191 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1192 key.objectid = dirid;
1193 key.type = BTRFS_INODE_REF_KEY;
1194 key.offset = (u64)-1;
1195
1196 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1197 if (ret < 0) {
1198 goto err;
1199 } else if (ret > 0) {
1200 ret = btrfs_previous_item(fs_root, path, dirid,
1201 BTRFS_INODE_REF_KEY);
1202 if (ret < 0) {
1203 goto err;
1204 } else if (ret > 0) {
1205 ret = -ENOENT;
1206 goto err;
1207 }
1208 }
1209
1210 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1211 dirid = key.offset;
1212
1213 inode_ref = btrfs_item_ptr(path->nodes[0],
1214 path->slots[0],
1215 struct btrfs_inode_ref);
1216 len = btrfs_inode_ref_name_len(path->nodes[0],
1217 inode_ref);
1218 ptr -= len + 1;
1219 if (ptr < name) {
1220 ret = -ENAMETOOLONG;
1221 goto err;
1222 }
1223 read_extent_buffer(path->nodes[0], ptr + 1,
1224 (unsigned long)(inode_ref + 1), len);
1225 ptr[0] = '/';
1226 btrfs_release_path(path);
1227 }
1228 btrfs_put_root(fs_root);
1229 fs_root = NULL;
1230 }
1231
1232 btrfs_free_path(path);
1233 if (ptr == name + PATH_MAX - 1) {
1234 name[0] = '/';
1235 name[1] = '\0';
1236 } else {
1237 memmove(name, ptr, name + PATH_MAX - ptr);
1238 }
1239 return name;
1240
1241err:
1242 btrfs_put_root(fs_root);
1243 btrfs_free_path(path);
1244 kfree(name);
1245 return ERR_PTR(ret);
1246}
1247
1248static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1249{
1250 struct btrfs_root *root = fs_info->tree_root;
1251 struct btrfs_dir_item *di;
1252 struct btrfs_path *path;
1253 struct btrfs_key location;
1254 u64 dir_id;
1255
1256 path = btrfs_alloc_path();
1257 if (!path)
1258 return -ENOMEM;
1259 path->leave_spinning = 1;
1260
1261 /*
1262 * Find the "default" dir item which points to the root item that we
1263 * will mount by default if we haven't been given a specific subvolume
1264 * to mount.
1265 */
1266 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1267 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1268 if (IS_ERR(di)) {
1269 btrfs_free_path(path);
1270 return PTR_ERR(di);
1271 }
1272 if (!di) {
1273 /*
1274 * Ok the default dir item isn't there. This is weird since
1275 * it's always been there, but don't freak out, just try and
1276 * mount the top-level subvolume.
1277 */
1278 btrfs_free_path(path);
1279 *objectid = BTRFS_FS_TREE_OBJECTID;
1280 return 0;
1281 }
1282
1283 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1284 btrfs_free_path(path);
1285 *objectid = location.objectid;
1286 return 0;
1287}
1288
1289static int btrfs_fill_super(struct super_block *sb,
1290 struct btrfs_fs_devices *fs_devices,
1291 void *data)
1292{
1293 struct inode *inode;
1294 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1295 int err;
1296
1297 sb->s_maxbytes = MAX_LFS_FILESIZE;
1298 sb->s_magic = BTRFS_SUPER_MAGIC;
1299 sb->s_op = &btrfs_super_ops;
1300 sb->s_d_op = &btrfs_dentry_operations;
1301 sb->s_export_op = &btrfs_export_ops;
1302 sb->s_xattr = btrfs_xattr_handlers;
1303 sb->s_time_gran = 1;
1304#ifdef CONFIG_BTRFS_FS_POSIX_ACL
1305 sb->s_flags |= SB_POSIXACL;
1306#endif
1307 sb->s_flags |= SB_I_VERSION;
1308 sb->s_iflags |= SB_I_CGROUPWB;
1309
1310 err = super_setup_bdi(sb);
1311 if (err) {
1312 btrfs_err(fs_info, "super_setup_bdi failed");
1313 return err;
1314 }
1315
1316 err = open_ctree(sb, fs_devices, (char *)data);
1317 if (err) {
1318 btrfs_err(fs_info, "open_ctree failed");
1319 return err;
1320 }
1321
1322 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1323 if (IS_ERR(inode)) {
1324 err = PTR_ERR(inode);
1325 goto fail_close;
1326 }
1327
1328 sb->s_root = d_make_root(inode);
1329 if (!sb->s_root) {
1330 err = -ENOMEM;
1331 goto fail_close;
1332 }
1333
1334 cleancache_init_fs(sb);
1335 sb->s_flags |= SB_ACTIVE;
1336 return 0;
1337
1338fail_close:
1339 close_ctree(fs_info);
1340 return err;
1341}
1342
1343int btrfs_sync_fs(struct super_block *sb, int wait)
1344{
1345 struct btrfs_trans_handle *trans;
1346 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1347 struct btrfs_root *root = fs_info->tree_root;
1348
1349 trace_btrfs_sync_fs(fs_info, wait);
1350
1351 if (!wait) {
1352 filemap_flush(fs_info->btree_inode->i_mapping);
1353 return 0;
1354 }
1355
1356 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1357
1358 trans = btrfs_attach_transaction_barrier(root);
1359 if (IS_ERR(trans)) {
1360 /* no transaction, don't bother */
1361 if (PTR_ERR(trans) == -ENOENT) {
1362 /*
1363 * Exit unless we have some pending changes
1364 * that need to go through commit
1365 */
1366 if (fs_info->pending_changes == 0)
1367 return 0;
1368 /*
1369 * A non-blocking test if the fs is frozen. We must not
1370 * start a new transaction here otherwise a deadlock
1371 * happens. The pending operations are delayed to the
1372 * next commit after thawing.
1373 */
1374 if (sb_start_write_trylock(sb))
1375 sb_end_write(sb);
1376 else
1377 return 0;
1378 trans = btrfs_start_transaction(root, 0);
1379 }
1380 if (IS_ERR(trans))
1381 return PTR_ERR(trans);
1382 }
1383 return btrfs_commit_transaction(trans);
1384}
1385
1386static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1387{
1388 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1389 const char *compress_type;
1390 const char *subvol_name;
1391
1392 if (btrfs_test_opt(info, DEGRADED))
1393 seq_puts(seq, ",degraded");
1394 if (btrfs_test_opt(info, NODATASUM))
1395 seq_puts(seq, ",nodatasum");
1396 if (btrfs_test_opt(info, NODATACOW))
1397 seq_puts(seq, ",nodatacow");
1398 if (btrfs_test_opt(info, NOBARRIER))
1399 seq_puts(seq, ",nobarrier");
1400 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1401 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1402 if (info->thread_pool_size != min_t(unsigned long,
1403 num_online_cpus() + 2, 8))
1404 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1405 if (btrfs_test_opt(info, COMPRESS)) {
1406 compress_type = btrfs_compress_type2str(info->compress_type);
1407 if (btrfs_test_opt(info, FORCE_COMPRESS))
1408 seq_printf(seq, ",compress-force=%s", compress_type);
1409 else
1410 seq_printf(seq, ",compress=%s", compress_type);
1411 if (info->compress_level)
1412 seq_printf(seq, ":%d", info->compress_level);
1413 }
1414 if (btrfs_test_opt(info, NOSSD))
1415 seq_puts(seq, ",nossd");
1416 if (btrfs_test_opt(info, SSD_SPREAD))
1417 seq_puts(seq, ",ssd_spread");
1418 else if (btrfs_test_opt(info, SSD))
1419 seq_puts(seq, ",ssd");
1420 if (btrfs_test_opt(info, NOTREELOG))
1421 seq_puts(seq, ",notreelog");
1422 if (btrfs_test_opt(info, NOLOGREPLAY))
1423 seq_puts(seq, ",rescue=nologreplay");
1424 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1425 seq_puts(seq, ",flushoncommit");
1426 if (btrfs_test_opt(info, DISCARD_SYNC))
1427 seq_puts(seq, ",discard");
1428 if (btrfs_test_opt(info, DISCARD_ASYNC))
1429 seq_puts(seq, ",discard=async");
1430 if (!(info->sb->s_flags & SB_POSIXACL))
1431 seq_puts(seq, ",noacl");
1432 if (btrfs_test_opt(info, SPACE_CACHE))
1433 seq_puts(seq, ",space_cache");
1434 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1435 seq_puts(seq, ",space_cache=v2");
1436 else
1437 seq_puts(seq, ",nospace_cache");
1438 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1439 seq_puts(seq, ",rescan_uuid_tree");
1440 if (btrfs_test_opt(info, CLEAR_CACHE))
1441 seq_puts(seq, ",clear_cache");
1442 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1443 seq_puts(seq, ",user_subvol_rm_allowed");
1444 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1445 seq_puts(seq, ",enospc_debug");
1446 if (btrfs_test_opt(info, AUTO_DEFRAG))
1447 seq_puts(seq, ",autodefrag");
1448 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1449 seq_puts(seq, ",inode_cache");
1450 if (btrfs_test_opt(info, SKIP_BALANCE))
1451 seq_puts(seq, ",skip_balance");
1452#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1453 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1454 seq_puts(seq, ",check_int_data");
1455 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1456 seq_puts(seq, ",check_int");
1457 if (info->check_integrity_print_mask)
1458 seq_printf(seq, ",check_int_print_mask=%d",
1459 info->check_integrity_print_mask);
1460#endif
1461 if (info->metadata_ratio)
1462 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1463 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1464 seq_puts(seq, ",fatal_errors=panic");
1465 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1466 seq_printf(seq, ",commit=%u", info->commit_interval);
1467#ifdef CONFIG_BTRFS_DEBUG
1468 if (btrfs_test_opt(info, FRAGMENT_DATA))
1469 seq_puts(seq, ",fragment=data");
1470 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1471 seq_puts(seq, ",fragment=metadata");
1472#endif
1473 if (btrfs_test_opt(info, REF_VERIFY))
1474 seq_puts(seq, ",ref_verify");
1475 seq_printf(seq, ",subvolid=%llu",
1476 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1477 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1478 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1479 if (!IS_ERR(subvol_name)) {
1480 seq_puts(seq, ",subvol=");
1481 seq_escape(seq, subvol_name, " \t\n\\");
1482 kfree(subvol_name);
1483 }
1484 return 0;
1485}
1486
1487static int btrfs_test_super(struct super_block *s, void *data)
1488{
1489 struct btrfs_fs_info *p = data;
1490 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1491
1492 return fs_info->fs_devices == p->fs_devices;
1493}
1494
1495static int btrfs_set_super(struct super_block *s, void *data)
1496{
1497 int err = set_anon_super(s, data);
1498 if (!err)
1499 s->s_fs_info = data;
1500 return err;
1501}
1502
1503/*
1504 * subvolumes are identified by ino 256
1505 */
1506static inline int is_subvolume_inode(struct inode *inode)
1507{
1508 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1509 return 1;
1510 return 0;
1511}
1512
1513static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1514 struct vfsmount *mnt)
1515{
1516 struct dentry *root;
1517 int ret;
1518
1519 if (!subvol_name) {
1520 if (!subvol_objectid) {
1521 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1522 &subvol_objectid);
1523 if (ret) {
1524 root = ERR_PTR(ret);
1525 goto out;
1526 }
1527 }
1528 subvol_name = btrfs_get_subvol_name_from_objectid(
1529 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1530 if (IS_ERR(subvol_name)) {
1531 root = ERR_CAST(subvol_name);
1532 subvol_name = NULL;
1533 goto out;
1534 }
1535
1536 }
1537
1538 root = mount_subtree(mnt, subvol_name);
1539 /* mount_subtree() drops our reference on the vfsmount. */
1540 mnt = NULL;
1541
1542 if (!IS_ERR(root)) {
1543 struct super_block *s = root->d_sb;
1544 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1545 struct inode *root_inode = d_inode(root);
1546 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1547
1548 ret = 0;
1549 if (!is_subvolume_inode(root_inode)) {
1550 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1551 subvol_name);
1552 ret = -EINVAL;
1553 }
1554 if (subvol_objectid && root_objectid != subvol_objectid) {
1555 /*
1556 * This will also catch a race condition where a
1557 * subvolume which was passed by ID is renamed and
1558 * another subvolume is renamed over the old location.
1559 */
1560 btrfs_err(fs_info,
1561 "subvol '%s' does not match subvolid %llu",
1562 subvol_name, subvol_objectid);
1563 ret = -EINVAL;
1564 }
1565 if (ret) {
1566 dput(root);
1567 root = ERR_PTR(ret);
1568 deactivate_locked_super(s);
1569 }
1570 }
1571
1572out:
1573 mntput(mnt);
1574 kfree(subvol_name);
1575 return root;
1576}
1577
1578/*
1579 * Find a superblock for the given device / mount point.
1580 *
1581 * Note: This is based on mount_bdev from fs/super.c with a few additions
1582 * for multiple device setup. Make sure to keep it in sync.
1583 */
1584static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1585 int flags, const char *device_name, void *data)
1586{
1587 struct block_device *bdev = NULL;
1588 struct super_block *s;
1589 struct btrfs_device *device = NULL;
1590 struct btrfs_fs_devices *fs_devices = NULL;
1591 struct btrfs_fs_info *fs_info = NULL;
1592 void *new_sec_opts = NULL;
1593 fmode_t mode = FMODE_READ;
1594 int error = 0;
1595
1596 if (!(flags & SB_RDONLY))
1597 mode |= FMODE_WRITE;
1598
1599 if (data) {
1600 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1601 if (error)
1602 return ERR_PTR(error);
1603 }
1604
1605 /*
1606 * Setup a dummy root and fs_info for test/set super. This is because
1607 * we don't actually fill this stuff out until open_ctree, but we need
1608 * then open_ctree will properly initialize the file system specific
1609 * settings later. btrfs_init_fs_info initializes the static elements
1610 * of the fs_info (locks and such) to make cleanup easier if we find a
1611 * superblock with our given fs_devices later on at sget() time.
1612 */
1613 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1614 if (!fs_info) {
1615 error = -ENOMEM;
1616 goto error_sec_opts;
1617 }
1618 btrfs_init_fs_info(fs_info);
1619
1620 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1621 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1622 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1623 error = -ENOMEM;
1624 goto error_fs_info;
1625 }
1626
1627 mutex_lock(&uuid_mutex);
1628 error = btrfs_parse_device_options(data, mode, fs_type);
1629 if (error) {
1630 mutex_unlock(&uuid_mutex);
1631 goto error_fs_info;
1632 }
1633
1634 device = btrfs_scan_one_device(device_name, mode, fs_type);
1635 if (IS_ERR(device)) {
1636 mutex_unlock(&uuid_mutex);
1637 error = PTR_ERR(device);
1638 goto error_fs_info;
1639 }
1640
1641 fs_devices = device->fs_devices;
1642 fs_info->fs_devices = fs_devices;
1643
1644 error = btrfs_open_devices(fs_devices, mode, fs_type);
1645 mutex_unlock(&uuid_mutex);
1646 if (error)
1647 goto error_fs_info;
1648
1649 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1650 error = -EACCES;
1651 goto error_close_devices;
1652 }
1653
1654 bdev = fs_devices->latest_bdev;
1655 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1656 fs_info);
1657 if (IS_ERR(s)) {
1658 error = PTR_ERR(s);
1659 goto error_close_devices;
1660 }
1661
1662 if (s->s_root) {
1663 btrfs_close_devices(fs_devices);
1664 btrfs_free_fs_info(fs_info);
1665 if ((flags ^ s->s_flags) & SB_RDONLY)
1666 error = -EBUSY;
1667 } else {
1668 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1669 btrfs_sb(s)->bdev_holder = fs_type;
1670 if (!strstr(crc32c_impl(), "generic"))
1671 set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1672 error = btrfs_fill_super(s, fs_devices, data);
1673 }
1674 if (!error)
1675 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1676 security_free_mnt_opts(&new_sec_opts);
1677 if (error) {
1678 deactivate_locked_super(s);
1679 return ERR_PTR(error);
1680 }
1681
1682 return dget(s->s_root);
1683
1684error_close_devices:
1685 btrfs_close_devices(fs_devices);
1686error_fs_info:
1687 btrfs_free_fs_info(fs_info);
1688error_sec_opts:
1689 security_free_mnt_opts(&new_sec_opts);
1690 return ERR_PTR(error);
1691}
1692
1693/*
1694 * Mount function which is called by VFS layer.
1695 *
1696 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1697 * which needs vfsmount* of device's root (/). This means device's root has to
1698 * be mounted internally in any case.
1699 *
1700 * Operation flow:
1701 * 1. Parse subvol id related options for later use in mount_subvol().
1702 *
1703 * 2. Mount device's root (/) by calling vfs_kern_mount().
1704 *
1705 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1706 * first place. In order to avoid calling btrfs_mount() again, we use
1707 * different file_system_type which is not registered to VFS by
1708 * register_filesystem() (btrfs_root_fs_type). As a result,
1709 * btrfs_mount_root() is called. The return value will be used by
1710 * mount_subtree() in mount_subvol().
1711 *
1712 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1713 * "btrfs subvolume set-default", mount_subvol() is called always.
1714 */
1715static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1716 const char *device_name, void *data)
1717{
1718 struct vfsmount *mnt_root;
1719 struct dentry *root;
1720 char *subvol_name = NULL;
1721 u64 subvol_objectid = 0;
1722 int error = 0;
1723
1724 error = btrfs_parse_subvol_options(data, &subvol_name,
1725 &subvol_objectid);
1726 if (error) {
1727 kfree(subvol_name);
1728 return ERR_PTR(error);
1729 }
1730
1731 /* mount device's root (/) */
1732 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1733 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1734 if (flags & SB_RDONLY) {
1735 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1736 flags & ~SB_RDONLY, device_name, data);
1737 } else {
1738 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1739 flags | SB_RDONLY, device_name, data);
1740 if (IS_ERR(mnt_root)) {
1741 root = ERR_CAST(mnt_root);
1742 kfree(subvol_name);
1743 goto out;
1744 }
1745
1746 down_write(&mnt_root->mnt_sb->s_umount);
1747 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1748 up_write(&mnt_root->mnt_sb->s_umount);
1749 if (error < 0) {
1750 root = ERR_PTR(error);
1751 mntput(mnt_root);
1752 kfree(subvol_name);
1753 goto out;
1754 }
1755 }
1756 }
1757 if (IS_ERR(mnt_root)) {
1758 root = ERR_CAST(mnt_root);
1759 kfree(subvol_name);
1760 goto out;
1761 }
1762
1763 /* mount_subvol() will free subvol_name and mnt_root */
1764 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1765
1766out:
1767 return root;
1768}
1769
1770static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1771 u32 new_pool_size, u32 old_pool_size)
1772{
1773 if (new_pool_size == old_pool_size)
1774 return;
1775
1776 fs_info->thread_pool_size = new_pool_size;
1777
1778 btrfs_info(fs_info, "resize thread pool %d -> %d",
1779 old_pool_size, new_pool_size);
1780
1781 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1782 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1783 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1784 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1785 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1786 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1787 new_pool_size);
1788 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1789 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1790 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1791 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1792 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1793 new_pool_size);
1794}
1795
1796static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1797 unsigned long old_opts, int flags)
1798{
1799 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1800 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1801 (flags & SB_RDONLY))) {
1802 /* wait for any defraggers to finish */
1803 wait_event(fs_info->transaction_wait,
1804 (atomic_read(&fs_info->defrag_running) == 0));
1805 if (flags & SB_RDONLY)
1806 sync_filesystem(fs_info->sb);
1807 }
1808}
1809
1810static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1811 unsigned long old_opts)
1812{
1813 /*
1814 * We need to cleanup all defragable inodes if the autodefragment is
1815 * close or the filesystem is read only.
1816 */
1817 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1818 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1819 btrfs_cleanup_defrag_inodes(fs_info);
1820 }
1821
1822 /* If we toggled discard async */
1823 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1824 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1825 btrfs_discard_resume(fs_info);
1826 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1827 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1828 btrfs_discard_cleanup(fs_info);
1829}
1830
1831static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1832{
1833 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1834 struct btrfs_root *root = fs_info->tree_root;
1835 unsigned old_flags = sb->s_flags;
1836 unsigned long old_opts = fs_info->mount_opt;
1837 unsigned long old_compress_type = fs_info->compress_type;
1838 u64 old_max_inline = fs_info->max_inline;
1839 u32 old_thread_pool_size = fs_info->thread_pool_size;
1840 u32 old_metadata_ratio = fs_info->metadata_ratio;
1841 int ret;
1842
1843 sync_filesystem(sb);
1844 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1845
1846 if (data) {
1847 void *new_sec_opts = NULL;
1848
1849 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1850 if (!ret)
1851 ret = security_sb_remount(sb, new_sec_opts);
1852 security_free_mnt_opts(&new_sec_opts);
1853 if (ret)
1854 goto restore;
1855 }
1856
1857 ret = btrfs_parse_options(fs_info, data, *flags);
1858 if (ret)
1859 goto restore;
1860
1861 btrfs_remount_begin(fs_info, old_opts, *flags);
1862 btrfs_resize_thread_pool(fs_info,
1863 fs_info->thread_pool_size, old_thread_pool_size);
1864
1865 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1866 goto out;
1867
1868 if (*flags & SB_RDONLY) {
1869 /*
1870 * this also happens on 'umount -rf' or on shutdown, when
1871 * the filesystem is busy.
1872 */
1873 cancel_work_sync(&fs_info->async_reclaim_work);
1874
1875 btrfs_discard_cleanup(fs_info);
1876
1877 /* wait for the uuid_scan task to finish */
1878 down(&fs_info->uuid_tree_rescan_sem);
1879 /* avoid complains from lockdep et al. */
1880 up(&fs_info->uuid_tree_rescan_sem);
1881
1882 sb->s_flags |= SB_RDONLY;
1883
1884 /*
1885 * Setting SB_RDONLY will put the cleaner thread to
1886 * sleep at the next loop if it's already active.
1887 * If it's already asleep, we'll leave unused block
1888 * groups on disk until we're mounted read-write again
1889 * unless we clean them up here.
1890 */
1891 btrfs_delete_unused_bgs(fs_info);
1892
1893 btrfs_dev_replace_suspend_for_unmount(fs_info);
1894 btrfs_scrub_cancel(fs_info);
1895 btrfs_pause_balance(fs_info);
1896
1897 ret = btrfs_commit_super(fs_info);
1898 if (ret)
1899 goto restore;
1900 } else {
1901 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1902 btrfs_err(fs_info,
1903 "Remounting read-write after error is not allowed");
1904 ret = -EINVAL;
1905 goto restore;
1906 }
1907 if (fs_info->fs_devices->rw_devices == 0) {
1908 ret = -EACCES;
1909 goto restore;
1910 }
1911
1912 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1913 btrfs_warn(fs_info,
1914 "too many missing devices, writable remount is not allowed");
1915 ret = -EACCES;
1916 goto restore;
1917 }
1918
1919 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1920 btrfs_warn(fs_info,
1921 "mount required to replay tree-log, cannot remount read-write");
1922 ret = -EINVAL;
1923 goto restore;
1924 }
1925
1926 ret = btrfs_cleanup_fs_roots(fs_info);
1927 if (ret)
1928 goto restore;
1929
1930 /* recover relocation */
1931 mutex_lock(&fs_info->cleaner_mutex);
1932 ret = btrfs_recover_relocation(root);
1933 mutex_unlock(&fs_info->cleaner_mutex);
1934 if (ret)
1935 goto restore;
1936
1937 ret = btrfs_resume_balance_async(fs_info);
1938 if (ret)
1939 goto restore;
1940
1941 ret = btrfs_resume_dev_replace_async(fs_info);
1942 if (ret) {
1943 btrfs_warn(fs_info, "failed to resume dev_replace");
1944 goto restore;
1945 }
1946
1947 btrfs_qgroup_rescan_resume(fs_info);
1948
1949 if (!fs_info->uuid_root) {
1950 btrfs_info(fs_info, "creating UUID tree");
1951 ret = btrfs_create_uuid_tree(fs_info);
1952 if (ret) {
1953 btrfs_warn(fs_info,
1954 "failed to create the UUID tree %d",
1955 ret);
1956 goto restore;
1957 }
1958 }
1959 sb->s_flags &= ~SB_RDONLY;
1960
1961 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1962 }
1963out:
1964 /*
1965 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
1966 * since the absence of the flag means it can be toggled off by remount.
1967 */
1968 *flags |= SB_I_VERSION;
1969
1970 wake_up_process(fs_info->transaction_kthread);
1971 btrfs_remount_cleanup(fs_info, old_opts);
1972 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1973
1974 return 0;
1975
1976restore:
1977 /* We've hit an error - don't reset SB_RDONLY */
1978 if (sb_rdonly(sb))
1979 old_flags |= SB_RDONLY;
1980 sb->s_flags = old_flags;
1981 fs_info->mount_opt = old_opts;
1982 fs_info->compress_type = old_compress_type;
1983 fs_info->max_inline = old_max_inline;
1984 btrfs_resize_thread_pool(fs_info,
1985 old_thread_pool_size, fs_info->thread_pool_size);
1986 fs_info->metadata_ratio = old_metadata_ratio;
1987 btrfs_remount_cleanup(fs_info, old_opts);
1988 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1989
1990 return ret;
1991}
1992
1993/* Used to sort the devices by max_avail(descending sort) */
1994static inline int btrfs_cmp_device_free_bytes(const void *dev_info1,
1995 const void *dev_info2)
1996{
1997 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1998 ((struct btrfs_device_info *)dev_info2)->max_avail)
1999 return -1;
2000 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
2001 ((struct btrfs_device_info *)dev_info2)->max_avail)
2002 return 1;
2003 else
2004 return 0;
2005}
2006
2007/*
2008 * sort the devices by max_avail, in which max free extent size of each device
2009 * is stored.(Descending Sort)
2010 */
2011static inline void btrfs_descending_sort_devices(
2012 struct btrfs_device_info *devices,
2013 size_t nr_devices)
2014{
2015 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2016 btrfs_cmp_device_free_bytes, NULL);
2017}
2018
2019/*
2020 * The helper to calc the free space on the devices that can be used to store
2021 * file data.
2022 */
2023static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2024 u64 *free_bytes)
2025{
2026 struct btrfs_device_info *devices_info;
2027 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2028 struct btrfs_device *device;
2029 u64 type;
2030 u64 avail_space;
2031 u64 min_stripe_size;
2032 int num_stripes = 1;
2033 int i = 0, nr_devices;
2034 const struct btrfs_raid_attr *rattr;
2035
2036 /*
2037 * We aren't under the device list lock, so this is racy-ish, but good
2038 * enough for our purposes.
2039 */
2040 nr_devices = fs_info->fs_devices->open_devices;
2041 if (!nr_devices) {
2042 smp_mb();
2043 nr_devices = fs_info->fs_devices->open_devices;
2044 ASSERT(nr_devices);
2045 if (!nr_devices) {
2046 *free_bytes = 0;
2047 return 0;
2048 }
2049 }
2050
2051 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2052 GFP_KERNEL);
2053 if (!devices_info)
2054 return -ENOMEM;
2055
2056 /* calc min stripe number for data space allocation */
2057 type = btrfs_data_alloc_profile(fs_info);
2058 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2059
2060 if (type & BTRFS_BLOCK_GROUP_RAID0)
2061 num_stripes = nr_devices;
2062 else if (type & BTRFS_BLOCK_GROUP_RAID1)
2063 num_stripes = 2;
2064 else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
2065 num_stripes = 3;
2066 else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
2067 num_stripes = 4;
2068 else if (type & BTRFS_BLOCK_GROUP_RAID10)
2069 num_stripes = 4;
2070
2071 /* Adjust for more than 1 stripe per device */
2072 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2073
2074 rcu_read_lock();
2075 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2076 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2077 &device->dev_state) ||
2078 !device->bdev ||
2079 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2080 continue;
2081
2082 if (i >= nr_devices)
2083 break;
2084
2085 avail_space = device->total_bytes - device->bytes_used;
2086
2087 /* align with stripe_len */
2088 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2089
2090 /*
2091 * In order to avoid overwriting the superblock on the drive,
2092 * btrfs starts at an offset of at least 1MB when doing chunk
2093 * allocation.
2094 *
2095 * This ensures we have at least min_stripe_size free space
2096 * after excluding 1MB.
2097 */
2098 if (avail_space <= SZ_1M + min_stripe_size)
2099 continue;
2100
2101 avail_space -= SZ_1M;
2102
2103 devices_info[i].dev = device;
2104 devices_info[i].max_avail = avail_space;
2105
2106 i++;
2107 }
2108 rcu_read_unlock();
2109
2110 nr_devices = i;
2111
2112 btrfs_descending_sort_devices(devices_info, nr_devices);
2113
2114 i = nr_devices - 1;
2115 avail_space = 0;
2116 while (nr_devices >= rattr->devs_min) {
2117 num_stripes = min(num_stripes, nr_devices);
2118
2119 if (devices_info[i].max_avail >= min_stripe_size) {
2120 int j;
2121 u64 alloc_size;
2122
2123 avail_space += devices_info[i].max_avail * num_stripes;
2124 alloc_size = devices_info[i].max_avail;
2125 for (j = i + 1 - num_stripes; j <= i; j++)
2126 devices_info[j].max_avail -= alloc_size;
2127 }
2128 i--;
2129 nr_devices--;
2130 }
2131
2132 kfree(devices_info);
2133 *free_bytes = avail_space;
2134 return 0;
2135}
2136
2137/*
2138 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2139 *
2140 * If there's a redundant raid level at DATA block groups, use the respective
2141 * multiplier to scale the sizes.
2142 *
2143 * Unused device space usage is based on simulating the chunk allocator
2144 * algorithm that respects the device sizes and order of allocations. This is
2145 * a close approximation of the actual use but there are other factors that may
2146 * change the result (like a new metadata chunk).
2147 *
2148 * If metadata is exhausted, f_bavail will be 0.
2149 */
2150static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2151{
2152 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2153 struct btrfs_super_block *disk_super = fs_info->super_copy;
2154 struct btrfs_space_info *found;
2155 u64 total_used = 0;
2156 u64 total_free_data = 0;
2157 u64 total_free_meta = 0;
2158 int bits = dentry->d_sb->s_blocksize_bits;
2159 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2160 unsigned factor = 1;
2161 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2162 int ret;
2163 u64 thresh = 0;
2164 int mixed = 0;
2165
2166 rcu_read_lock();
2167 list_for_each_entry_rcu(found, &fs_info->space_info, list) {
2168 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2169 int i;
2170
2171 total_free_data += found->disk_total - found->disk_used;
2172 total_free_data -=
2173 btrfs_account_ro_block_groups_free_space(found);
2174
2175 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2176 if (!list_empty(&found->block_groups[i]))
2177 factor = btrfs_bg_type_to_factor(
2178 btrfs_raid_array[i].bg_flag);
2179 }
2180 }
2181
2182 /*
2183 * Metadata in mixed block goup profiles are accounted in data
2184 */
2185 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2186 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2187 mixed = 1;
2188 else
2189 total_free_meta += found->disk_total -
2190 found->disk_used;
2191 }
2192
2193 total_used += found->disk_used;
2194 }
2195
2196 rcu_read_unlock();
2197
2198 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2199 buf->f_blocks >>= bits;
2200 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2201
2202 /* Account global block reserve as used, it's in logical size already */
2203 spin_lock(&block_rsv->lock);
2204 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2205 if (buf->f_bfree >= block_rsv->size >> bits)
2206 buf->f_bfree -= block_rsv->size >> bits;
2207 else
2208 buf->f_bfree = 0;
2209 spin_unlock(&block_rsv->lock);
2210
2211 buf->f_bavail = div_u64(total_free_data, factor);
2212 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2213 if (ret)
2214 return ret;
2215 buf->f_bavail += div_u64(total_free_data, factor);
2216 buf->f_bavail = buf->f_bavail >> bits;
2217
2218 /*
2219 * We calculate the remaining metadata space minus global reserve. If
2220 * this is (supposedly) smaller than zero, there's no space. But this
2221 * does not hold in practice, the exhausted state happens where's still
2222 * some positive delta. So we apply some guesswork and compare the
2223 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2224 *
2225 * We probably cannot calculate the exact threshold value because this
2226 * depends on the internal reservations requested by various
2227 * operations, so some operations that consume a few metadata will
2228 * succeed even if the Avail is zero. But this is better than the other
2229 * way around.
2230 */
2231 thresh = SZ_4M;
2232
2233 /*
2234 * We only want to claim there's no available space if we can no longer
2235 * allocate chunks for our metadata profile and our global reserve will
2236 * not fit in the free metadata space. If we aren't ->full then we
2237 * still can allocate chunks and thus are fine using the currently
2238 * calculated f_bavail.
2239 */
2240 if (!mixed && block_rsv->space_info->full &&
2241 total_free_meta - thresh < block_rsv->size)
2242 buf->f_bavail = 0;
2243
2244 buf->f_type = BTRFS_SUPER_MAGIC;
2245 buf->f_bsize = dentry->d_sb->s_blocksize;
2246 buf->f_namelen = BTRFS_NAME_LEN;
2247
2248 /* We treat it as constant endianness (it doesn't matter _which_)
2249 because we want the fsid to come out the same whether mounted
2250 on a big-endian or little-endian host */
2251 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2252 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2253 /* Mask in the root object ID too, to disambiguate subvols */
2254 buf->f_fsid.val[0] ^=
2255 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2256 buf->f_fsid.val[1] ^=
2257 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2258
2259 return 0;
2260}
2261
2262static void btrfs_kill_super(struct super_block *sb)
2263{
2264 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2265 kill_anon_super(sb);
2266 btrfs_free_fs_info(fs_info);
2267}
2268
2269static struct file_system_type btrfs_fs_type = {
2270 .owner = THIS_MODULE,
2271 .name = "btrfs",
2272 .mount = btrfs_mount,
2273 .kill_sb = btrfs_kill_super,
2274 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2275};
2276
2277static struct file_system_type btrfs_root_fs_type = {
2278 .owner = THIS_MODULE,
2279 .name = "btrfs",
2280 .mount = btrfs_mount_root,
2281 .kill_sb = btrfs_kill_super,
2282 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2283};
2284
2285MODULE_ALIAS_FS("btrfs");
2286
2287static int btrfs_control_open(struct inode *inode, struct file *file)
2288{
2289 /*
2290 * The control file's private_data is used to hold the
2291 * transaction when it is started and is used to keep
2292 * track of whether a transaction is already in progress.
2293 */
2294 file->private_data = NULL;
2295 return 0;
2296}
2297
2298/*
2299 * Used by /dev/btrfs-control for devices ioctls.
2300 */
2301static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2302 unsigned long arg)
2303{
2304 struct btrfs_ioctl_vol_args *vol;
2305 struct btrfs_device *device = NULL;
2306 int ret = -ENOTTY;
2307
2308 if (!capable(CAP_SYS_ADMIN))
2309 return -EPERM;
2310
2311 vol = memdup_user((void __user *)arg, sizeof(*vol));
2312 if (IS_ERR(vol))
2313 return PTR_ERR(vol);
2314 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2315
2316 switch (cmd) {
2317 case BTRFS_IOC_SCAN_DEV:
2318 mutex_lock(&uuid_mutex);
2319 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2320 &btrfs_root_fs_type);
2321 ret = PTR_ERR_OR_ZERO(device);
2322 mutex_unlock(&uuid_mutex);
2323 break;
2324 case BTRFS_IOC_FORGET_DEV:
2325 ret = btrfs_forget_devices(vol->name);
2326 break;
2327 case BTRFS_IOC_DEVICES_READY:
2328 mutex_lock(&uuid_mutex);
2329 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2330 &btrfs_root_fs_type);
2331 if (IS_ERR(device)) {
2332 mutex_unlock(&uuid_mutex);
2333 ret = PTR_ERR(device);
2334 break;
2335 }
2336 ret = !(device->fs_devices->num_devices ==
2337 device->fs_devices->total_devices);
2338 mutex_unlock(&uuid_mutex);
2339 break;
2340 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2341 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2342 break;
2343 }
2344
2345 kfree(vol);
2346 return ret;
2347}
2348
2349static int btrfs_freeze(struct super_block *sb)
2350{
2351 struct btrfs_trans_handle *trans;
2352 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2353 struct btrfs_root *root = fs_info->tree_root;
2354
2355 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2356 /*
2357 * We don't need a barrier here, we'll wait for any transaction that
2358 * could be in progress on other threads (and do delayed iputs that
2359 * we want to avoid on a frozen filesystem), or do the commit
2360 * ourselves.
2361 */
2362 trans = btrfs_attach_transaction_barrier(root);
2363 if (IS_ERR(trans)) {
2364 /* no transaction, don't bother */
2365 if (PTR_ERR(trans) == -ENOENT)
2366 return 0;
2367 return PTR_ERR(trans);
2368 }
2369 return btrfs_commit_transaction(trans);
2370}
2371
2372static int btrfs_unfreeze(struct super_block *sb)
2373{
2374 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2375
2376 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2377 return 0;
2378}
2379
2380static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2381{
2382 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2383 struct btrfs_device *dev, *first_dev = NULL;
2384
2385 /*
2386 * Lightweight locking of the devices. We should not need
2387 * device_list_mutex here as we only read the device data and the list
2388 * is protected by RCU. Even if a device is deleted during the list
2389 * traversals, we'll get valid data, the freeing callback will wait at
2390 * least until the rcu_read_unlock.
2391 */
2392 rcu_read_lock();
2393 list_for_each_entry_rcu(dev, &fs_info->fs_devices->devices, dev_list) {
2394 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2395 continue;
2396 if (!dev->name)
2397 continue;
2398 if (!first_dev || dev->devid < first_dev->devid)
2399 first_dev = dev;
2400 }
2401
2402 if (first_dev)
2403 seq_escape(m, rcu_str_deref(first_dev->name), " \t\n\\");
2404 else
2405 WARN_ON(1);
2406 rcu_read_unlock();
2407 return 0;
2408}
2409
2410static const struct super_operations btrfs_super_ops = {
2411 .drop_inode = btrfs_drop_inode,
2412 .evict_inode = btrfs_evict_inode,
2413 .put_super = btrfs_put_super,
2414 .sync_fs = btrfs_sync_fs,
2415 .show_options = btrfs_show_options,
2416 .show_devname = btrfs_show_devname,
2417 .alloc_inode = btrfs_alloc_inode,
2418 .destroy_inode = btrfs_destroy_inode,
2419 .free_inode = btrfs_free_inode,
2420 .statfs = btrfs_statfs,
2421 .remount_fs = btrfs_remount,
2422 .freeze_fs = btrfs_freeze,
2423 .unfreeze_fs = btrfs_unfreeze,
2424};
2425
2426static const struct file_operations btrfs_ctl_fops = {
2427 .open = btrfs_control_open,
2428 .unlocked_ioctl = btrfs_control_ioctl,
2429 .compat_ioctl = compat_ptr_ioctl,
2430 .owner = THIS_MODULE,
2431 .llseek = noop_llseek,
2432};
2433
2434static struct miscdevice btrfs_misc = {
2435 .minor = BTRFS_MINOR,
2436 .name = "btrfs-control",
2437 .fops = &btrfs_ctl_fops
2438};
2439
2440MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2441MODULE_ALIAS("devname:btrfs-control");
2442
2443static int __init btrfs_interface_init(void)
2444{
2445 return misc_register(&btrfs_misc);
2446}
2447
2448static __cold void btrfs_interface_exit(void)
2449{
2450 misc_deregister(&btrfs_misc);
2451}
2452
2453static void __init btrfs_print_mod_info(void)
2454{
2455 static const char options[] = ""
2456#ifdef CONFIG_BTRFS_DEBUG
2457 ", debug=on"
2458#endif
2459#ifdef CONFIG_BTRFS_ASSERT
2460 ", assert=on"
2461#endif
2462#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2463 ", integrity-checker=on"
2464#endif
2465#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2466 ", ref-verify=on"
2467#endif
2468 ;
2469 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2470}
2471
2472static int __init init_btrfs_fs(void)
2473{
2474 int err;
2475
2476 btrfs_props_init();
2477
2478 err = btrfs_init_sysfs();
2479 if (err)
2480 return err;
2481
2482 btrfs_init_compress();
2483
2484 err = btrfs_init_cachep();
2485 if (err)
2486 goto free_compress;
2487
2488 err = extent_io_init();
2489 if (err)
2490 goto free_cachep;
2491
2492 err = extent_state_cache_init();
2493 if (err)
2494 goto free_extent_io;
2495
2496 err = extent_map_init();
2497 if (err)
2498 goto free_extent_state_cache;
2499
2500 err = ordered_data_init();
2501 if (err)
2502 goto free_extent_map;
2503
2504 err = btrfs_delayed_inode_init();
2505 if (err)
2506 goto free_ordered_data;
2507
2508 err = btrfs_auto_defrag_init();
2509 if (err)
2510 goto free_delayed_inode;
2511
2512 err = btrfs_delayed_ref_init();
2513 if (err)
2514 goto free_auto_defrag;
2515
2516 err = btrfs_prelim_ref_init();
2517 if (err)
2518 goto free_delayed_ref;
2519
2520 err = btrfs_end_io_wq_init();
2521 if (err)
2522 goto free_prelim_ref;
2523
2524 err = btrfs_interface_init();
2525 if (err)
2526 goto free_end_io_wq;
2527
2528 btrfs_init_lockdep();
2529
2530 btrfs_print_mod_info();
2531
2532 err = btrfs_run_sanity_tests();
2533 if (err)
2534 goto unregister_ioctl;
2535
2536 err = register_filesystem(&btrfs_fs_type);
2537 if (err)
2538 goto unregister_ioctl;
2539
2540 return 0;
2541
2542unregister_ioctl:
2543 btrfs_interface_exit();
2544free_end_io_wq:
2545 btrfs_end_io_wq_exit();
2546free_prelim_ref:
2547 btrfs_prelim_ref_exit();
2548free_delayed_ref:
2549 btrfs_delayed_ref_exit();
2550free_auto_defrag:
2551 btrfs_auto_defrag_exit();
2552free_delayed_inode:
2553 btrfs_delayed_inode_exit();
2554free_ordered_data:
2555 ordered_data_exit();
2556free_extent_map:
2557 extent_map_exit();
2558free_extent_state_cache:
2559 extent_state_cache_exit();
2560free_extent_io:
2561 extent_io_exit();
2562free_cachep:
2563 btrfs_destroy_cachep();
2564free_compress:
2565 btrfs_exit_compress();
2566 btrfs_exit_sysfs();
2567
2568 return err;
2569}
2570
2571static void __exit exit_btrfs_fs(void)
2572{
2573 btrfs_destroy_cachep();
2574 btrfs_delayed_ref_exit();
2575 btrfs_auto_defrag_exit();
2576 btrfs_delayed_inode_exit();
2577 btrfs_prelim_ref_exit();
2578 ordered_data_exit();
2579 extent_map_exit();
2580 extent_state_cache_exit();
2581 extent_io_exit();
2582 btrfs_interface_exit();
2583 btrfs_end_io_wq_exit();
2584 unregister_filesystem(&btrfs_fs_type);
2585 btrfs_exit_sysfs();
2586 btrfs_cleanup_fs_uuids();
2587 btrfs_exit_compress();
2588}
2589
2590late_initcall(init_btrfs_fs);
2591module_exit(exit_btrfs_fs)
2592
2593MODULE_LICENSE("GPL");
2594MODULE_SOFTDEP("pre: crc32c");
2595MODULE_SOFTDEP("pre: xxhash64");
2596MODULE_SOFTDEP("pre: sha256");
2597MODULE_SOFTDEP("pre: blake2b-256");
1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/blkdev.h>
20#include <linux/module.h>
21#include <linux/buffer_head.h>
22#include <linux/fs.h>
23#include <linux/pagemap.h>
24#include <linux/highmem.h>
25#include <linux/time.h>
26#include <linux/init.h>
27#include <linux/seq_file.h>
28#include <linux/string.h>
29#include <linux/backing-dev.h>
30#include <linux/mount.h>
31#include <linux/mpage.h>
32#include <linux/swap.h>
33#include <linux/writeback.h>
34#include <linux/statfs.h>
35#include <linux/compat.h>
36#include <linux/parser.h>
37#include <linux/ctype.h>
38#include <linux/namei.h>
39#include <linux/miscdevice.h>
40#include <linux/magic.h>
41#include <linux/slab.h>
42#include <linux/cleancache.h>
43#include <linux/ratelimit.h>
44#include <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#define CREATE_TRACE_POINTS
64#include <trace/events/btrfs.h>
65
66static const struct super_operations btrfs_super_ops;
67static struct file_system_type btrfs_fs_type;
68
69static int btrfs_remount(struct super_block *sb, int *flags, char *data);
70
71static const char *btrfs_decode_error(int errno)
72{
73 char *errstr = "unknown";
74
75 switch (errno) {
76 case -EIO:
77 errstr = "IO failure";
78 break;
79 case -ENOMEM:
80 errstr = "Out of memory";
81 break;
82 case -EROFS:
83 errstr = "Readonly filesystem";
84 break;
85 case -EEXIST:
86 errstr = "Object already exists";
87 break;
88 case -ENOSPC:
89 errstr = "No space left";
90 break;
91 case -ENOENT:
92 errstr = "No such entry";
93 break;
94 }
95
96 return errstr;
97}
98
99static void save_error_info(struct btrfs_fs_info *fs_info)
100{
101 /*
102 * today we only save the error info into ram. Long term we'll
103 * also send it down to the disk
104 */
105 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
106}
107
108/* btrfs handle error by forcing the filesystem readonly */
109static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
110{
111 struct super_block *sb = fs_info->sb;
112
113 if (sb->s_flags & MS_RDONLY)
114 return;
115
116 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
117 sb->s_flags |= MS_RDONLY;
118 btrfs_info(fs_info, "forced readonly");
119 /*
120 * Note that a running device replace operation is not
121 * canceled here although there is no way to update
122 * the progress. It would add the risk of a deadlock,
123 * therefore the canceling is ommited. The only penalty
124 * is that some I/O remains active until the procedure
125 * completes. The next time when the filesystem is
126 * mounted writeable again, the device replace
127 * operation continues.
128 */
129 }
130}
131
132#ifdef CONFIG_PRINTK
133/*
134 * __btrfs_std_error decodes expected errors from the caller and
135 * invokes the approciate error response.
136 */
137void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
138 unsigned int line, int errno, const char *fmt, ...)
139{
140 struct super_block *sb = fs_info->sb;
141 const char *errstr;
142
143 /*
144 * Special case: if the error is EROFS, and we're already
145 * under MS_RDONLY, then it is safe here.
146 */
147 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
148 return;
149
150 errstr = btrfs_decode_error(errno);
151 if (fmt) {
152 struct va_format vaf;
153 va_list args;
154
155 va_start(args, fmt);
156 vaf.fmt = fmt;
157 vaf.va = &args;
158
159 printk(KERN_CRIT
160 "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
161 sb->s_id, function, line, errno, errstr, &vaf);
162 va_end(args);
163 } else {
164 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
165 sb->s_id, function, line, errno, errstr);
166 }
167
168 /* Don't go through full error handling during mount */
169 save_error_info(fs_info);
170 if (sb->s_flags & MS_BORN)
171 btrfs_handle_error(fs_info);
172}
173
174static const char * const logtypes[] = {
175 "emergency",
176 "alert",
177 "critical",
178 "error",
179 "warning",
180 "notice",
181 "info",
182 "debug",
183};
184
185void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
186{
187 struct super_block *sb = fs_info->sb;
188 char lvl[4];
189 struct va_format vaf;
190 va_list args;
191 const char *type = logtypes[4];
192 int kern_level;
193
194 va_start(args, fmt);
195
196 kern_level = printk_get_level(fmt);
197 if (kern_level) {
198 size_t size = printk_skip_level(fmt) - fmt;
199 memcpy(lvl, fmt, size);
200 lvl[size] = '\0';
201 fmt += size;
202 type = logtypes[kern_level - '0'];
203 } else
204 *lvl = '\0';
205
206 vaf.fmt = fmt;
207 vaf.va = &args;
208
209 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
210
211 va_end(args);
212}
213
214#else
215
216void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
217 unsigned int line, int errno, const char *fmt, ...)
218{
219 struct super_block *sb = fs_info->sb;
220
221 /*
222 * Special case: if the error is EROFS, and we're already
223 * under MS_RDONLY, then it is safe here.
224 */
225 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
226 return;
227
228 /* Don't go through full error handling during mount */
229 if (sb->s_flags & MS_BORN) {
230 save_error_info(fs_info);
231 btrfs_handle_error(fs_info);
232 }
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 */
249void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
250 struct btrfs_root *root, const char *function,
251 unsigned int line, int errno)
252{
253 /*
254 * Report first abort since mount
255 */
256 if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,
257 &root->fs_info->fs_state)) {
258 WARN(1, KERN_DEBUG "BTRFS: Transaction aborted (error %d)\n",
259 errno);
260 }
261 trans->aborted = errno;
262 /* Nothing used. The other threads that have joined this
263 * transaction may be able to continue. */
264 if (!trans->blocks_used) {
265 const char *errstr;
266
267 errstr = btrfs_decode_error(errno);
268 btrfs_warn(root->fs_info,
269 "%s:%d: Aborting unused transaction(%s).",
270 function, line, errstr);
271 return;
272 }
273 ACCESS_ONCE(trans->transaction->aborted) = errno;
274 /* Wake up anybody who may be waiting on this transaction */
275 wake_up(&root->fs_info->transaction_wait);
276 wake_up(&root->fs_info->transaction_blocked_wait);
277 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
278}
279/*
280 * __btrfs_panic decodes unexpected, fatal errors from the caller,
281 * issues an alert, and either panics or BUGs, depending on mount options.
282 */
283void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
284 unsigned int line, int errno, const char *fmt, ...)
285{
286 char *s_id = "<unknown>";
287 const char *errstr;
288 struct va_format vaf = { .fmt = fmt };
289 va_list args;
290
291 if (fs_info)
292 s_id = fs_info->sb->s_id;
293
294 va_start(args, fmt);
295 vaf.va = &args;
296
297 errstr = btrfs_decode_error(errno);
298 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
299 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
300 s_id, function, line, &vaf, errno, errstr);
301
302 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
303 function, line, &vaf, errno, errstr);
304 va_end(args);
305 /* Caller calls BUG() */
306}
307
308static void btrfs_put_super(struct super_block *sb)
309{
310 (void)close_ctree(btrfs_sb(sb)->tree_root);
311 /* FIXME: need to fix VFS to return error? */
312 /* AV: return it _where_? ->put_super() can be triggered by any number
313 * of async events, up to and including delivery of SIGKILL to the
314 * last process that kept it busy. Or segfault in the aforementioned
315 * process... Whom would you report that to?
316 */
317}
318
319enum {
320 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
321 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
322 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
323 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
324 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
325 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
326 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
327 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
328 Opt_check_integrity, Opt_check_integrity_including_extent_data,
329 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
330 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
331 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
332 Opt_datasum, Opt_treelog, Opt_noinode_cache,
333 Opt_err,
334};
335
336static match_table_t tokens = {
337 {Opt_degraded, "degraded"},
338 {Opt_subvol, "subvol=%s"},
339 {Opt_subvolid, "subvolid=%s"},
340 {Opt_device, "device=%s"},
341 {Opt_nodatasum, "nodatasum"},
342 {Opt_datasum, "datasum"},
343 {Opt_nodatacow, "nodatacow"},
344 {Opt_datacow, "datacow"},
345 {Opt_nobarrier, "nobarrier"},
346 {Opt_barrier, "barrier"},
347 {Opt_max_inline, "max_inline=%s"},
348 {Opt_alloc_start, "alloc_start=%s"},
349 {Opt_thread_pool, "thread_pool=%d"},
350 {Opt_compress, "compress"},
351 {Opt_compress_type, "compress=%s"},
352 {Opt_compress_force, "compress-force"},
353 {Opt_compress_force_type, "compress-force=%s"},
354 {Opt_ssd, "ssd"},
355 {Opt_ssd_spread, "ssd_spread"},
356 {Opt_nossd, "nossd"},
357 {Opt_acl, "acl"},
358 {Opt_noacl, "noacl"},
359 {Opt_notreelog, "notreelog"},
360 {Opt_treelog, "treelog"},
361 {Opt_flushoncommit, "flushoncommit"},
362 {Opt_noflushoncommit, "noflushoncommit"},
363 {Opt_ratio, "metadata_ratio=%d"},
364 {Opt_discard, "discard"},
365 {Opt_nodiscard, "nodiscard"},
366 {Opt_space_cache, "space_cache"},
367 {Opt_clear_cache, "clear_cache"},
368 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
369 {Opt_enospc_debug, "enospc_debug"},
370 {Opt_noenospc_debug, "noenospc_debug"},
371 {Opt_subvolrootid, "subvolrootid=%d"},
372 {Opt_defrag, "autodefrag"},
373 {Opt_nodefrag, "noautodefrag"},
374 {Opt_inode_cache, "inode_cache"},
375 {Opt_noinode_cache, "noinode_cache"},
376 {Opt_no_space_cache, "nospace_cache"},
377 {Opt_recovery, "recovery"},
378 {Opt_skip_balance, "skip_balance"},
379 {Opt_check_integrity, "check_int"},
380 {Opt_check_integrity_including_extent_data, "check_int_data"},
381 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
382 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
383 {Opt_fatal_errors, "fatal_errors=%s"},
384 {Opt_commit_interval, "commit=%d"},
385 {Opt_err, NULL},
386};
387
388/*
389 * Regular mount options parser. Everything that is needed only when
390 * reading in a new superblock is parsed here.
391 * XXX JDM: This needs to be cleaned up for remount.
392 */
393int btrfs_parse_options(struct btrfs_root *root, char *options)
394{
395 struct btrfs_fs_info *info = root->fs_info;
396 substring_t args[MAX_OPT_ARGS];
397 char *p, *num, *orig = NULL;
398 u64 cache_gen;
399 int intarg;
400 int ret = 0;
401 char *compress_type;
402 bool compress_force = false;
403 bool compress = false;
404
405 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
406 if (cache_gen)
407 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
408
409 if (!options)
410 goto out;
411
412 /*
413 * strsep changes the string, duplicate it because parse_options
414 * gets called twice
415 */
416 options = kstrdup(options, GFP_NOFS);
417 if (!options)
418 return -ENOMEM;
419
420 orig = options;
421
422 while ((p = strsep(&options, ",")) != NULL) {
423 int token;
424 if (!*p)
425 continue;
426
427 token = match_token(p, tokens, args);
428 switch (token) {
429 case Opt_degraded:
430 btrfs_info(root->fs_info, "allowing degraded mounts");
431 btrfs_set_opt(info->mount_opt, DEGRADED);
432 break;
433 case Opt_subvol:
434 case Opt_subvolid:
435 case Opt_subvolrootid:
436 case Opt_device:
437 /*
438 * These are parsed by btrfs_parse_early_options
439 * and can be happily ignored here.
440 */
441 break;
442 case Opt_nodatasum:
443 btrfs_set_and_info(root, NODATASUM,
444 "setting nodatasum");
445 break;
446 case Opt_datasum:
447 if (btrfs_test_opt(root, NODATASUM)) {
448 if (btrfs_test_opt(root, NODATACOW))
449 btrfs_info(root->fs_info, "setting datasum, datacow enabled");
450 else
451 btrfs_info(root->fs_info, "setting datasum");
452 }
453 btrfs_clear_opt(info->mount_opt, NODATACOW);
454 btrfs_clear_opt(info->mount_opt, NODATASUM);
455 break;
456 case Opt_nodatacow:
457 if (!btrfs_test_opt(root, NODATACOW)) {
458 if (!btrfs_test_opt(root, COMPRESS) ||
459 !btrfs_test_opt(root, FORCE_COMPRESS)) {
460 btrfs_info(root->fs_info,
461 "setting nodatacow, compression disabled");
462 } else {
463 btrfs_info(root->fs_info, "setting nodatacow");
464 }
465 }
466 btrfs_clear_opt(info->mount_opt, COMPRESS);
467 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
468 btrfs_set_opt(info->mount_opt, NODATACOW);
469 btrfs_set_opt(info->mount_opt, NODATASUM);
470 break;
471 case Opt_datacow:
472 btrfs_clear_and_info(root, NODATACOW,
473 "setting datacow");
474 break;
475 case Opt_compress_force:
476 case Opt_compress_force_type:
477 compress_force = true;
478 /* Fallthrough */
479 case Opt_compress:
480 case Opt_compress_type:
481 compress = true;
482 if (token == Opt_compress ||
483 token == Opt_compress_force ||
484 strcmp(args[0].from, "zlib") == 0) {
485 compress_type = "zlib";
486 info->compress_type = BTRFS_COMPRESS_ZLIB;
487 btrfs_set_opt(info->mount_opt, COMPRESS);
488 btrfs_clear_opt(info->mount_opt, NODATACOW);
489 btrfs_clear_opt(info->mount_opt, NODATASUM);
490 } else if (strcmp(args[0].from, "lzo") == 0) {
491 compress_type = "lzo";
492 info->compress_type = BTRFS_COMPRESS_LZO;
493 btrfs_set_opt(info->mount_opt, COMPRESS);
494 btrfs_clear_opt(info->mount_opt, NODATACOW);
495 btrfs_clear_opt(info->mount_opt, NODATASUM);
496 btrfs_set_fs_incompat(info, COMPRESS_LZO);
497 } else if (strncmp(args[0].from, "no", 2) == 0) {
498 compress_type = "no";
499 btrfs_clear_opt(info->mount_opt, COMPRESS);
500 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
501 compress_force = false;
502 } else {
503 ret = -EINVAL;
504 goto out;
505 }
506
507 if (compress_force) {
508 btrfs_set_and_info(root, FORCE_COMPRESS,
509 "force %s compression",
510 compress_type);
511 } else if (compress) {
512 if (!btrfs_test_opt(root, COMPRESS))
513 btrfs_info(root->fs_info,
514 "btrfs: use %s compression\n",
515 compress_type);
516 }
517 break;
518 case Opt_ssd:
519 btrfs_set_and_info(root, SSD,
520 "use ssd allocation scheme");
521 break;
522 case Opt_ssd_spread:
523 btrfs_set_and_info(root, SSD_SPREAD,
524 "use spread ssd allocation scheme");
525 break;
526 case Opt_nossd:
527 btrfs_clear_and_info(root, NOSSD,
528 "not using ssd allocation scheme");
529 btrfs_clear_opt(info->mount_opt, SSD);
530 break;
531 case Opt_barrier:
532 btrfs_clear_and_info(root, NOBARRIER,
533 "turning on barriers");
534 break;
535 case Opt_nobarrier:
536 btrfs_set_and_info(root, NOBARRIER,
537 "turning off barriers");
538 break;
539 case Opt_thread_pool:
540 ret = match_int(&args[0], &intarg);
541 if (ret) {
542 goto out;
543 } else if (intarg > 0) {
544 info->thread_pool_size = intarg;
545 } else {
546 ret = -EINVAL;
547 goto out;
548 }
549 break;
550 case Opt_max_inline:
551 num = match_strdup(&args[0]);
552 if (num) {
553 info->max_inline = memparse(num, NULL);
554 kfree(num);
555
556 if (info->max_inline) {
557 info->max_inline = min_t(u64,
558 info->max_inline,
559 root->sectorsize);
560 }
561 btrfs_info(root->fs_info, "max_inline at %llu",
562 info->max_inline);
563 } else {
564 ret = -ENOMEM;
565 goto out;
566 }
567 break;
568 case Opt_alloc_start:
569 num = match_strdup(&args[0]);
570 if (num) {
571 mutex_lock(&info->chunk_mutex);
572 info->alloc_start = memparse(num, NULL);
573 mutex_unlock(&info->chunk_mutex);
574 kfree(num);
575 btrfs_info(root->fs_info, "allocations start at %llu",
576 info->alloc_start);
577 } else {
578 ret = -ENOMEM;
579 goto out;
580 }
581 break;
582 case Opt_acl:
583 root->fs_info->sb->s_flags |= MS_POSIXACL;
584 break;
585 case Opt_noacl:
586 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
587 break;
588 case Opt_notreelog:
589 btrfs_set_and_info(root, NOTREELOG,
590 "disabling tree log");
591 break;
592 case Opt_treelog:
593 btrfs_clear_and_info(root, NOTREELOG,
594 "enabling tree log");
595 break;
596 case Opt_flushoncommit:
597 btrfs_set_and_info(root, FLUSHONCOMMIT,
598 "turning on flush-on-commit");
599 break;
600 case Opt_noflushoncommit:
601 btrfs_clear_and_info(root, FLUSHONCOMMIT,
602 "turning off flush-on-commit");
603 break;
604 case Opt_ratio:
605 ret = match_int(&args[0], &intarg);
606 if (ret) {
607 goto out;
608 } else if (intarg >= 0) {
609 info->metadata_ratio = intarg;
610 btrfs_info(root->fs_info, "metadata ratio %d",
611 info->metadata_ratio);
612 } else {
613 ret = -EINVAL;
614 goto out;
615 }
616 break;
617 case Opt_discard:
618 btrfs_set_and_info(root, DISCARD,
619 "turning on discard");
620 break;
621 case Opt_nodiscard:
622 btrfs_clear_and_info(root, DISCARD,
623 "turning off discard");
624 break;
625 case Opt_space_cache:
626 btrfs_set_and_info(root, SPACE_CACHE,
627 "enabling disk space caching");
628 break;
629 case Opt_rescan_uuid_tree:
630 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
631 break;
632 case Opt_no_space_cache:
633 btrfs_clear_and_info(root, SPACE_CACHE,
634 "disabling disk space caching");
635 break;
636 case Opt_inode_cache:
637 btrfs_set_and_info(root, CHANGE_INODE_CACHE,
638 "enabling inode map caching");
639 break;
640 case Opt_noinode_cache:
641 btrfs_clear_and_info(root, CHANGE_INODE_CACHE,
642 "disabling inode map caching");
643 break;
644 case Opt_clear_cache:
645 btrfs_set_and_info(root, CLEAR_CACHE,
646 "force clearing of disk cache");
647 break;
648 case Opt_user_subvol_rm_allowed:
649 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
650 break;
651 case Opt_enospc_debug:
652 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
653 break;
654 case Opt_noenospc_debug:
655 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
656 break;
657 case Opt_defrag:
658 btrfs_set_and_info(root, AUTO_DEFRAG,
659 "enabling auto defrag");
660 break;
661 case Opt_nodefrag:
662 btrfs_clear_and_info(root, AUTO_DEFRAG,
663 "disabling auto defrag");
664 break;
665 case Opt_recovery:
666 btrfs_info(root->fs_info, "enabling auto recovery");
667 btrfs_set_opt(info->mount_opt, RECOVERY);
668 break;
669 case Opt_skip_balance:
670 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
671 break;
672#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
673 case Opt_check_integrity_including_extent_data:
674 btrfs_info(root->fs_info,
675 "enabling check integrity including extent data");
676 btrfs_set_opt(info->mount_opt,
677 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
678 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
679 break;
680 case Opt_check_integrity:
681 btrfs_info(root->fs_info, "enabling check integrity");
682 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
683 break;
684 case Opt_check_integrity_print_mask:
685 ret = match_int(&args[0], &intarg);
686 if (ret) {
687 goto out;
688 } else if (intarg >= 0) {
689 info->check_integrity_print_mask = intarg;
690 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
691 info->check_integrity_print_mask);
692 } else {
693 ret = -EINVAL;
694 goto out;
695 }
696 break;
697#else
698 case Opt_check_integrity_including_extent_data:
699 case Opt_check_integrity:
700 case Opt_check_integrity_print_mask:
701 btrfs_err(root->fs_info,
702 "support for check_integrity* not compiled in!");
703 ret = -EINVAL;
704 goto out;
705#endif
706 case Opt_fatal_errors:
707 if (strcmp(args[0].from, "panic") == 0)
708 btrfs_set_opt(info->mount_opt,
709 PANIC_ON_FATAL_ERROR);
710 else if (strcmp(args[0].from, "bug") == 0)
711 btrfs_clear_opt(info->mount_opt,
712 PANIC_ON_FATAL_ERROR);
713 else {
714 ret = -EINVAL;
715 goto out;
716 }
717 break;
718 case Opt_commit_interval:
719 intarg = 0;
720 ret = match_int(&args[0], &intarg);
721 if (ret < 0) {
722 btrfs_err(root->fs_info, "invalid commit interval");
723 ret = -EINVAL;
724 goto out;
725 }
726 if (intarg > 0) {
727 if (intarg > 300) {
728 btrfs_warn(root->fs_info, "excessive commit interval %d",
729 intarg);
730 }
731 info->commit_interval = intarg;
732 } else {
733 btrfs_info(root->fs_info, "using default commit interval %ds",
734 BTRFS_DEFAULT_COMMIT_INTERVAL);
735 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
736 }
737 break;
738 case Opt_err:
739 btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
740 ret = -EINVAL;
741 goto out;
742 default:
743 break;
744 }
745 }
746out:
747 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
748 btrfs_info(root->fs_info, "disk space caching is enabled");
749 kfree(orig);
750 return ret;
751}
752
753/*
754 * Parse mount options that are required early in the mount process.
755 *
756 * All other options will be parsed on much later in the mount process and
757 * only when we need to allocate a new super block.
758 */
759static int btrfs_parse_early_options(const char *options, fmode_t flags,
760 void *holder, char **subvol_name, u64 *subvol_objectid,
761 struct btrfs_fs_devices **fs_devices)
762{
763 substring_t args[MAX_OPT_ARGS];
764 char *device_name, *opts, *orig, *p;
765 char *num = NULL;
766 int error = 0;
767
768 if (!options)
769 return 0;
770
771 /*
772 * strsep changes the string, duplicate it because parse_options
773 * gets called twice
774 */
775 opts = kstrdup(options, GFP_KERNEL);
776 if (!opts)
777 return -ENOMEM;
778 orig = opts;
779
780 while ((p = strsep(&opts, ",")) != NULL) {
781 int token;
782 if (!*p)
783 continue;
784
785 token = match_token(p, tokens, args);
786 switch (token) {
787 case Opt_subvol:
788 kfree(*subvol_name);
789 *subvol_name = match_strdup(&args[0]);
790 if (!*subvol_name) {
791 error = -ENOMEM;
792 goto out;
793 }
794 break;
795 case Opt_subvolid:
796 num = match_strdup(&args[0]);
797 if (num) {
798 *subvol_objectid = memparse(num, NULL);
799 kfree(num);
800 /* we want the original fs_tree */
801 if (!*subvol_objectid)
802 *subvol_objectid =
803 BTRFS_FS_TREE_OBJECTID;
804 } else {
805 error = -EINVAL;
806 goto out;
807 }
808 break;
809 case Opt_subvolrootid:
810 printk(KERN_WARNING
811 "BTRFS: 'subvolrootid' mount option is deprecated and has "
812 "no effect\n");
813 break;
814 case Opt_device:
815 device_name = match_strdup(&args[0]);
816 if (!device_name) {
817 error = -ENOMEM;
818 goto out;
819 }
820 error = btrfs_scan_one_device(device_name,
821 flags, holder, fs_devices);
822 kfree(device_name);
823 if (error)
824 goto out;
825 break;
826 default:
827 break;
828 }
829 }
830
831out:
832 kfree(orig);
833 return error;
834}
835
836static struct dentry *get_default_root(struct super_block *sb,
837 u64 subvol_objectid)
838{
839 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
840 struct btrfs_root *root = fs_info->tree_root;
841 struct btrfs_root *new_root;
842 struct btrfs_dir_item *di;
843 struct btrfs_path *path;
844 struct btrfs_key location;
845 struct inode *inode;
846 struct dentry *dentry;
847 u64 dir_id;
848 int new = 0;
849
850 /*
851 * We have a specific subvol we want to mount, just setup location and
852 * go look up the root.
853 */
854 if (subvol_objectid) {
855 location.objectid = subvol_objectid;
856 location.type = BTRFS_ROOT_ITEM_KEY;
857 location.offset = (u64)-1;
858 goto find_root;
859 }
860
861 path = btrfs_alloc_path();
862 if (!path)
863 return ERR_PTR(-ENOMEM);
864 path->leave_spinning = 1;
865
866 /*
867 * Find the "default" dir item which points to the root item that we
868 * will mount by default if we haven't been given a specific subvolume
869 * to mount.
870 */
871 dir_id = btrfs_super_root_dir(fs_info->super_copy);
872 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
873 if (IS_ERR(di)) {
874 btrfs_free_path(path);
875 return ERR_CAST(di);
876 }
877 if (!di) {
878 /*
879 * Ok the default dir item isn't there. This is weird since
880 * it's always been there, but don't freak out, just try and
881 * mount to root most subvolume.
882 */
883 btrfs_free_path(path);
884 dir_id = BTRFS_FIRST_FREE_OBJECTID;
885 new_root = fs_info->fs_root;
886 goto setup_root;
887 }
888
889 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
890 btrfs_free_path(path);
891
892find_root:
893 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
894 if (IS_ERR(new_root))
895 return ERR_CAST(new_root);
896
897 dir_id = btrfs_root_dirid(&new_root->root_item);
898setup_root:
899 location.objectid = dir_id;
900 location.type = BTRFS_INODE_ITEM_KEY;
901 location.offset = 0;
902
903 inode = btrfs_iget(sb, &location, new_root, &new);
904 if (IS_ERR(inode))
905 return ERR_CAST(inode);
906
907 /*
908 * If we're just mounting the root most subvol put the inode and return
909 * a reference to the dentry. We will have already gotten a reference
910 * to the inode in btrfs_fill_super so we're good to go.
911 */
912 if (!new && sb->s_root->d_inode == inode) {
913 iput(inode);
914 return dget(sb->s_root);
915 }
916
917 dentry = d_obtain_alias(inode);
918 if (!IS_ERR(dentry)) {
919 spin_lock(&dentry->d_lock);
920 dentry->d_flags &= ~DCACHE_DISCONNECTED;
921 spin_unlock(&dentry->d_lock);
922 }
923 return dentry;
924}
925
926static int btrfs_fill_super(struct super_block *sb,
927 struct btrfs_fs_devices *fs_devices,
928 void *data, int silent)
929{
930 struct inode *inode;
931 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
932 struct btrfs_key key;
933 int err;
934
935 sb->s_maxbytes = MAX_LFS_FILESIZE;
936 sb->s_magic = BTRFS_SUPER_MAGIC;
937 sb->s_op = &btrfs_super_ops;
938 sb->s_d_op = &btrfs_dentry_operations;
939 sb->s_export_op = &btrfs_export_ops;
940 sb->s_xattr = btrfs_xattr_handlers;
941 sb->s_time_gran = 1;
942#ifdef CONFIG_BTRFS_FS_POSIX_ACL
943 sb->s_flags |= MS_POSIXACL;
944#endif
945 sb->s_flags |= MS_I_VERSION;
946 err = open_ctree(sb, fs_devices, (char *)data);
947 if (err) {
948 printk(KERN_ERR "BTRFS: open_ctree failed\n");
949 return err;
950 }
951
952 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
953 key.type = BTRFS_INODE_ITEM_KEY;
954 key.offset = 0;
955 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
956 if (IS_ERR(inode)) {
957 err = PTR_ERR(inode);
958 goto fail_close;
959 }
960
961 sb->s_root = d_make_root(inode);
962 if (!sb->s_root) {
963 err = -ENOMEM;
964 goto fail_close;
965 }
966
967 save_mount_options(sb, data);
968 cleancache_init_fs(sb);
969 sb->s_flags |= MS_ACTIVE;
970 return 0;
971
972fail_close:
973 close_ctree(fs_info->tree_root);
974 return err;
975}
976
977int btrfs_sync_fs(struct super_block *sb, int wait)
978{
979 struct btrfs_trans_handle *trans;
980 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
981 struct btrfs_root *root = fs_info->tree_root;
982
983 trace_btrfs_sync_fs(wait);
984
985 if (!wait) {
986 filemap_flush(fs_info->btree_inode->i_mapping);
987 return 0;
988 }
989
990 btrfs_wait_ordered_roots(fs_info, -1);
991
992 trans = btrfs_attach_transaction_barrier(root);
993 if (IS_ERR(trans)) {
994 /* no transaction, don't bother */
995 if (PTR_ERR(trans) == -ENOENT)
996 return 0;
997 return PTR_ERR(trans);
998 }
999 return btrfs_commit_transaction(trans, root);
1000}
1001
1002static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1003{
1004 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1005 struct btrfs_root *root = info->tree_root;
1006 char *compress_type;
1007
1008 if (btrfs_test_opt(root, DEGRADED))
1009 seq_puts(seq, ",degraded");
1010 if (btrfs_test_opt(root, NODATASUM))
1011 seq_puts(seq, ",nodatasum");
1012 if (btrfs_test_opt(root, NODATACOW))
1013 seq_puts(seq, ",nodatacow");
1014 if (btrfs_test_opt(root, NOBARRIER))
1015 seq_puts(seq, ",nobarrier");
1016 if (info->max_inline != 8192 * 1024)
1017 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1018 if (info->alloc_start != 0)
1019 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1020 if (info->thread_pool_size != min_t(unsigned long,
1021 num_online_cpus() + 2, 8))
1022 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1023 if (btrfs_test_opt(root, COMPRESS)) {
1024 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1025 compress_type = "zlib";
1026 else
1027 compress_type = "lzo";
1028 if (btrfs_test_opt(root, FORCE_COMPRESS))
1029 seq_printf(seq, ",compress-force=%s", compress_type);
1030 else
1031 seq_printf(seq, ",compress=%s", compress_type);
1032 }
1033 if (btrfs_test_opt(root, NOSSD))
1034 seq_puts(seq, ",nossd");
1035 if (btrfs_test_opt(root, SSD_SPREAD))
1036 seq_puts(seq, ",ssd_spread");
1037 else if (btrfs_test_opt(root, SSD))
1038 seq_puts(seq, ",ssd");
1039 if (btrfs_test_opt(root, NOTREELOG))
1040 seq_puts(seq, ",notreelog");
1041 if (btrfs_test_opt(root, FLUSHONCOMMIT))
1042 seq_puts(seq, ",flushoncommit");
1043 if (btrfs_test_opt(root, DISCARD))
1044 seq_puts(seq, ",discard");
1045 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1046 seq_puts(seq, ",noacl");
1047 if (btrfs_test_opt(root, SPACE_CACHE))
1048 seq_puts(seq, ",space_cache");
1049 else
1050 seq_puts(seq, ",nospace_cache");
1051 if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1052 seq_puts(seq, ",rescan_uuid_tree");
1053 if (btrfs_test_opt(root, CLEAR_CACHE))
1054 seq_puts(seq, ",clear_cache");
1055 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1056 seq_puts(seq, ",user_subvol_rm_allowed");
1057 if (btrfs_test_opt(root, ENOSPC_DEBUG))
1058 seq_puts(seq, ",enospc_debug");
1059 if (btrfs_test_opt(root, AUTO_DEFRAG))
1060 seq_puts(seq, ",autodefrag");
1061 if (btrfs_test_opt(root, INODE_MAP_CACHE))
1062 seq_puts(seq, ",inode_cache");
1063 if (btrfs_test_opt(root, SKIP_BALANCE))
1064 seq_puts(seq, ",skip_balance");
1065 if (btrfs_test_opt(root, RECOVERY))
1066 seq_puts(seq, ",recovery");
1067#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1068 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1069 seq_puts(seq, ",check_int_data");
1070 else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1071 seq_puts(seq, ",check_int");
1072 if (info->check_integrity_print_mask)
1073 seq_printf(seq, ",check_int_print_mask=%d",
1074 info->check_integrity_print_mask);
1075#endif
1076 if (info->metadata_ratio)
1077 seq_printf(seq, ",metadata_ratio=%d",
1078 info->metadata_ratio);
1079 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1080 seq_puts(seq, ",fatal_errors=panic");
1081 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1082 seq_printf(seq, ",commit=%d", info->commit_interval);
1083 return 0;
1084}
1085
1086static int btrfs_test_super(struct super_block *s, void *data)
1087{
1088 struct btrfs_fs_info *p = data;
1089 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1090
1091 return fs_info->fs_devices == p->fs_devices;
1092}
1093
1094static int btrfs_set_super(struct super_block *s, void *data)
1095{
1096 int err = set_anon_super(s, data);
1097 if (!err)
1098 s->s_fs_info = data;
1099 return err;
1100}
1101
1102/*
1103 * subvolumes are identified by ino 256
1104 */
1105static inline int is_subvolume_inode(struct inode *inode)
1106{
1107 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1108 return 1;
1109 return 0;
1110}
1111
1112/*
1113 * This will strip out the subvol=%s argument for an argument string and add
1114 * subvolid=0 to make sure we get the actual tree root for path walking to the
1115 * subvol we want.
1116 */
1117static char *setup_root_args(char *args)
1118{
1119 unsigned len = strlen(args) + 2 + 1;
1120 char *src, *dst, *buf;
1121
1122 /*
1123 * We need the same args as before, but with this substitution:
1124 * s!subvol=[^,]+!subvolid=0!
1125 *
1126 * Since the replacement string is up to 2 bytes longer than the
1127 * original, allocate strlen(args) + 2 + 1 bytes.
1128 */
1129
1130 src = strstr(args, "subvol=");
1131 /* This shouldn't happen, but just in case.. */
1132 if (!src)
1133 return NULL;
1134
1135 buf = dst = kmalloc(len, GFP_NOFS);
1136 if (!buf)
1137 return NULL;
1138
1139 /*
1140 * If the subvol= arg is not at the start of the string,
1141 * copy whatever precedes it into buf.
1142 */
1143 if (src != args) {
1144 *src++ = '\0';
1145 strcpy(buf, args);
1146 dst += strlen(args);
1147 }
1148
1149 strcpy(dst, "subvolid=0");
1150 dst += strlen("subvolid=0");
1151
1152 /*
1153 * If there is a "," after the original subvol=... string,
1154 * copy that suffix into our buffer. Otherwise, we're done.
1155 */
1156 src = strchr(src, ',');
1157 if (src)
1158 strcpy(dst, src);
1159
1160 return buf;
1161}
1162
1163static struct dentry *mount_subvol(const char *subvol_name, int flags,
1164 const char *device_name, char *data)
1165{
1166 struct dentry *root;
1167 struct vfsmount *mnt;
1168 char *newargs;
1169
1170 newargs = setup_root_args(data);
1171 if (!newargs)
1172 return ERR_PTR(-ENOMEM);
1173 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1174 newargs);
1175
1176 if (PTR_RET(mnt) == -EBUSY) {
1177 if (flags & MS_RDONLY) {
1178 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY, device_name,
1179 newargs);
1180 } else {
1181 int r;
1182 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY, device_name,
1183 newargs);
1184 if (IS_ERR(mnt)) {
1185 kfree(newargs);
1186 return ERR_CAST(mnt);
1187 }
1188
1189 r = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1190 if (r < 0) {
1191 /* FIXME: release vfsmount mnt ??*/
1192 kfree(newargs);
1193 return ERR_PTR(r);
1194 }
1195 }
1196 }
1197
1198 kfree(newargs);
1199
1200 if (IS_ERR(mnt))
1201 return ERR_CAST(mnt);
1202
1203 root = mount_subtree(mnt, subvol_name);
1204
1205 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1206 struct super_block *s = root->d_sb;
1207 dput(root);
1208 root = ERR_PTR(-EINVAL);
1209 deactivate_locked_super(s);
1210 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1211 subvol_name);
1212 }
1213
1214 return root;
1215}
1216
1217/*
1218 * Find a superblock for the given device / mount point.
1219 *
1220 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1221 * for multiple device setup. Make sure to keep it in sync.
1222 */
1223static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1224 const char *device_name, void *data)
1225{
1226 struct block_device *bdev = NULL;
1227 struct super_block *s;
1228 struct dentry *root;
1229 struct btrfs_fs_devices *fs_devices = NULL;
1230 struct btrfs_fs_info *fs_info = NULL;
1231 fmode_t mode = FMODE_READ;
1232 char *subvol_name = NULL;
1233 u64 subvol_objectid = 0;
1234 int error = 0;
1235
1236 if (!(flags & MS_RDONLY))
1237 mode |= FMODE_WRITE;
1238
1239 error = btrfs_parse_early_options(data, mode, fs_type,
1240 &subvol_name, &subvol_objectid,
1241 &fs_devices);
1242 if (error) {
1243 kfree(subvol_name);
1244 return ERR_PTR(error);
1245 }
1246
1247 if (subvol_name) {
1248 root = mount_subvol(subvol_name, flags, device_name, data);
1249 kfree(subvol_name);
1250 return root;
1251 }
1252
1253 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1254 if (error)
1255 return ERR_PTR(error);
1256
1257 /*
1258 * Setup a dummy root and fs_info for test/set super. This is because
1259 * we don't actually fill this stuff out until open_ctree, but we need
1260 * it for searching for existing supers, so this lets us do that and
1261 * then open_ctree will properly initialize everything later.
1262 */
1263 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1264 if (!fs_info)
1265 return ERR_PTR(-ENOMEM);
1266
1267 fs_info->fs_devices = fs_devices;
1268
1269 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1270 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1271 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1272 error = -ENOMEM;
1273 goto error_fs_info;
1274 }
1275
1276 error = btrfs_open_devices(fs_devices, mode, fs_type);
1277 if (error)
1278 goto error_fs_info;
1279
1280 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1281 error = -EACCES;
1282 goto error_close_devices;
1283 }
1284
1285 bdev = fs_devices->latest_bdev;
1286 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1287 fs_info);
1288 if (IS_ERR(s)) {
1289 error = PTR_ERR(s);
1290 goto error_close_devices;
1291 }
1292
1293 if (s->s_root) {
1294 btrfs_close_devices(fs_devices);
1295 free_fs_info(fs_info);
1296 if ((flags ^ s->s_flags) & MS_RDONLY)
1297 error = -EBUSY;
1298 } else {
1299 char b[BDEVNAME_SIZE];
1300
1301 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1302 btrfs_sb(s)->bdev_holder = fs_type;
1303 error = btrfs_fill_super(s, fs_devices, data,
1304 flags & MS_SILENT ? 1 : 0);
1305 }
1306
1307 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1308 if (IS_ERR(root))
1309 deactivate_locked_super(s);
1310
1311 return root;
1312
1313error_close_devices:
1314 btrfs_close_devices(fs_devices);
1315error_fs_info:
1316 free_fs_info(fs_info);
1317 return ERR_PTR(error);
1318}
1319
1320static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1321 int new_pool_size, int old_pool_size)
1322{
1323 if (new_pool_size == old_pool_size)
1324 return;
1325
1326 fs_info->thread_pool_size = new_pool_size;
1327
1328 btrfs_info(fs_info, "resize thread pool %d -> %d",
1329 old_pool_size, new_pool_size);
1330
1331 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1332 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1333 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1334 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1335 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1336 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1337 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1338 new_pool_size);
1339 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1340 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1341 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1342 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1343 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1344 new_pool_size);
1345}
1346
1347static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1348{
1349 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1350}
1351
1352static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1353 unsigned long old_opts, int flags)
1354{
1355 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1356 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1357 (flags & MS_RDONLY))) {
1358 /* wait for any defraggers to finish */
1359 wait_event(fs_info->transaction_wait,
1360 (atomic_read(&fs_info->defrag_running) == 0));
1361 if (flags & MS_RDONLY)
1362 sync_filesystem(fs_info->sb);
1363 }
1364}
1365
1366static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1367 unsigned long old_opts)
1368{
1369 /*
1370 * We need cleanup all defragable inodes if the autodefragment is
1371 * close or the fs is R/O.
1372 */
1373 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1374 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1375 (fs_info->sb->s_flags & MS_RDONLY))) {
1376 btrfs_cleanup_defrag_inodes(fs_info);
1377 }
1378
1379 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1380}
1381
1382static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1383{
1384 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1385 struct btrfs_root *root = fs_info->tree_root;
1386 unsigned old_flags = sb->s_flags;
1387 unsigned long old_opts = fs_info->mount_opt;
1388 unsigned long old_compress_type = fs_info->compress_type;
1389 u64 old_max_inline = fs_info->max_inline;
1390 u64 old_alloc_start = fs_info->alloc_start;
1391 int old_thread_pool_size = fs_info->thread_pool_size;
1392 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1393 int ret;
1394
1395 sync_filesystem(sb);
1396 btrfs_remount_prepare(fs_info);
1397
1398 ret = btrfs_parse_options(root, data);
1399 if (ret) {
1400 ret = -EINVAL;
1401 goto restore;
1402 }
1403
1404 btrfs_remount_begin(fs_info, old_opts, *flags);
1405 btrfs_resize_thread_pool(fs_info,
1406 fs_info->thread_pool_size, old_thread_pool_size);
1407
1408 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1409 goto out;
1410
1411 if (*flags & MS_RDONLY) {
1412 /*
1413 * this also happens on 'umount -rf' or on shutdown, when
1414 * the filesystem is busy.
1415 */
1416
1417 /* wait for the uuid_scan task to finish */
1418 down(&fs_info->uuid_tree_rescan_sem);
1419 /* avoid complains from lockdep et al. */
1420 up(&fs_info->uuid_tree_rescan_sem);
1421
1422 sb->s_flags |= MS_RDONLY;
1423
1424 btrfs_dev_replace_suspend_for_unmount(fs_info);
1425 btrfs_scrub_cancel(fs_info);
1426 btrfs_pause_balance(fs_info);
1427
1428 ret = btrfs_commit_super(root);
1429 if (ret)
1430 goto restore;
1431 } else {
1432 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1433 btrfs_err(fs_info,
1434 "Remounting read-write after error is not allowed");
1435 ret = -EINVAL;
1436 goto restore;
1437 }
1438 if (fs_info->fs_devices->rw_devices == 0) {
1439 ret = -EACCES;
1440 goto restore;
1441 }
1442
1443 if (fs_info->fs_devices->missing_devices >
1444 fs_info->num_tolerated_disk_barrier_failures &&
1445 !(*flags & MS_RDONLY)) {
1446 btrfs_warn(fs_info,
1447 "too many missing devices, writeable remount is not allowed");
1448 ret = -EACCES;
1449 goto restore;
1450 }
1451
1452 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1453 ret = -EINVAL;
1454 goto restore;
1455 }
1456
1457 ret = btrfs_cleanup_fs_roots(fs_info);
1458 if (ret)
1459 goto restore;
1460
1461 /* recover relocation */
1462 ret = btrfs_recover_relocation(root);
1463 if (ret)
1464 goto restore;
1465
1466 ret = btrfs_resume_balance_async(fs_info);
1467 if (ret)
1468 goto restore;
1469
1470 ret = btrfs_resume_dev_replace_async(fs_info);
1471 if (ret) {
1472 btrfs_warn(fs_info, "failed to resume dev_replace");
1473 goto restore;
1474 }
1475
1476 if (!fs_info->uuid_root) {
1477 btrfs_info(fs_info, "creating UUID tree");
1478 ret = btrfs_create_uuid_tree(fs_info);
1479 if (ret) {
1480 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1481 goto restore;
1482 }
1483 }
1484 sb->s_flags &= ~MS_RDONLY;
1485 }
1486out:
1487 wake_up_process(fs_info->transaction_kthread);
1488 btrfs_remount_cleanup(fs_info, old_opts);
1489 return 0;
1490
1491restore:
1492 /* We've hit an error - don't reset MS_RDONLY */
1493 if (sb->s_flags & MS_RDONLY)
1494 old_flags |= MS_RDONLY;
1495 sb->s_flags = old_flags;
1496 fs_info->mount_opt = old_opts;
1497 fs_info->compress_type = old_compress_type;
1498 fs_info->max_inline = old_max_inline;
1499 mutex_lock(&fs_info->chunk_mutex);
1500 fs_info->alloc_start = old_alloc_start;
1501 mutex_unlock(&fs_info->chunk_mutex);
1502 btrfs_resize_thread_pool(fs_info,
1503 old_thread_pool_size, fs_info->thread_pool_size);
1504 fs_info->metadata_ratio = old_metadata_ratio;
1505 btrfs_remount_cleanup(fs_info, old_opts);
1506 return ret;
1507}
1508
1509/* Used to sort the devices by max_avail(descending sort) */
1510static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1511 const void *dev_info2)
1512{
1513 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1514 ((struct btrfs_device_info *)dev_info2)->max_avail)
1515 return -1;
1516 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1517 ((struct btrfs_device_info *)dev_info2)->max_avail)
1518 return 1;
1519 else
1520 return 0;
1521}
1522
1523/*
1524 * sort the devices by max_avail, in which max free extent size of each device
1525 * is stored.(Descending Sort)
1526 */
1527static inline void btrfs_descending_sort_devices(
1528 struct btrfs_device_info *devices,
1529 size_t nr_devices)
1530{
1531 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1532 btrfs_cmp_device_free_bytes, NULL);
1533}
1534
1535/*
1536 * The helper to calc the free space on the devices that can be used to store
1537 * file data.
1538 */
1539static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1540{
1541 struct btrfs_fs_info *fs_info = root->fs_info;
1542 struct btrfs_device_info *devices_info;
1543 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1544 struct btrfs_device *device;
1545 u64 skip_space;
1546 u64 type;
1547 u64 avail_space;
1548 u64 used_space;
1549 u64 min_stripe_size;
1550 int min_stripes = 1, num_stripes = 1;
1551 int i = 0, nr_devices;
1552 int ret;
1553
1554 nr_devices = fs_info->fs_devices->open_devices;
1555 BUG_ON(!nr_devices);
1556
1557 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1558 GFP_NOFS);
1559 if (!devices_info)
1560 return -ENOMEM;
1561
1562 /* calc min stripe number for data space alloction */
1563 type = btrfs_get_alloc_profile(root, 1);
1564 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1565 min_stripes = 2;
1566 num_stripes = nr_devices;
1567 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1568 min_stripes = 2;
1569 num_stripes = 2;
1570 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1571 min_stripes = 4;
1572 num_stripes = 4;
1573 }
1574
1575 if (type & BTRFS_BLOCK_GROUP_DUP)
1576 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1577 else
1578 min_stripe_size = BTRFS_STRIPE_LEN;
1579
1580 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1581 if (!device->in_fs_metadata || !device->bdev ||
1582 device->is_tgtdev_for_dev_replace)
1583 continue;
1584
1585 avail_space = device->total_bytes - device->bytes_used;
1586
1587 /* align with stripe_len */
1588 do_div(avail_space, BTRFS_STRIPE_LEN);
1589 avail_space *= BTRFS_STRIPE_LEN;
1590
1591 /*
1592 * In order to avoid overwritting the superblock on the drive,
1593 * btrfs starts at an offset of at least 1MB when doing chunk
1594 * allocation.
1595 */
1596 skip_space = 1024 * 1024;
1597
1598 /* user can set the offset in fs_info->alloc_start. */
1599 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1600 device->total_bytes)
1601 skip_space = max(fs_info->alloc_start, skip_space);
1602
1603 /*
1604 * btrfs can not use the free space in [0, skip_space - 1],
1605 * we must subtract it from the total. In order to implement
1606 * it, we account the used space in this range first.
1607 */
1608 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1609 &used_space);
1610 if (ret) {
1611 kfree(devices_info);
1612 return ret;
1613 }
1614
1615 /* calc the free space in [0, skip_space - 1] */
1616 skip_space -= used_space;
1617
1618 /*
1619 * we can use the free space in [0, skip_space - 1], subtract
1620 * it from the total.
1621 */
1622 if (avail_space && avail_space >= skip_space)
1623 avail_space -= skip_space;
1624 else
1625 avail_space = 0;
1626
1627 if (avail_space < min_stripe_size)
1628 continue;
1629
1630 devices_info[i].dev = device;
1631 devices_info[i].max_avail = avail_space;
1632
1633 i++;
1634 }
1635
1636 nr_devices = i;
1637
1638 btrfs_descending_sort_devices(devices_info, nr_devices);
1639
1640 i = nr_devices - 1;
1641 avail_space = 0;
1642 while (nr_devices >= min_stripes) {
1643 if (num_stripes > nr_devices)
1644 num_stripes = nr_devices;
1645
1646 if (devices_info[i].max_avail >= min_stripe_size) {
1647 int j;
1648 u64 alloc_size;
1649
1650 avail_space += devices_info[i].max_avail * num_stripes;
1651 alloc_size = devices_info[i].max_avail;
1652 for (j = i + 1 - num_stripes; j <= i; j++)
1653 devices_info[j].max_avail -= alloc_size;
1654 }
1655 i--;
1656 nr_devices--;
1657 }
1658
1659 kfree(devices_info);
1660 *free_bytes = avail_space;
1661 return 0;
1662}
1663
1664static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1665{
1666 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1667 struct btrfs_super_block *disk_super = fs_info->super_copy;
1668 struct list_head *head = &fs_info->space_info;
1669 struct btrfs_space_info *found;
1670 u64 total_used = 0;
1671 u64 total_free_data = 0;
1672 int bits = dentry->d_sb->s_blocksize_bits;
1673 __be32 *fsid = (__be32 *)fs_info->fsid;
1674 int ret;
1675
1676 /* holding chunk_muext to avoid allocating new chunks */
1677 mutex_lock(&fs_info->chunk_mutex);
1678 rcu_read_lock();
1679 list_for_each_entry_rcu(found, head, list) {
1680 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1681 total_free_data += found->disk_total - found->disk_used;
1682 total_free_data -=
1683 btrfs_account_ro_block_groups_free_space(found);
1684 }
1685
1686 total_used += found->disk_used;
1687 }
1688 rcu_read_unlock();
1689
1690 buf->f_namelen = BTRFS_NAME_LEN;
1691 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1692 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1693 buf->f_bsize = dentry->d_sb->s_blocksize;
1694 buf->f_type = BTRFS_SUPER_MAGIC;
1695 buf->f_bavail = total_free_data;
1696 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1697 if (ret) {
1698 mutex_unlock(&fs_info->chunk_mutex);
1699 return ret;
1700 }
1701 buf->f_bavail += total_free_data;
1702 buf->f_bavail = buf->f_bavail >> bits;
1703 mutex_unlock(&fs_info->chunk_mutex);
1704
1705 /* We treat it as constant endianness (it doesn't matter _which_)
1706 because we want the fsid to come out the same whether mounted
1707 on a big-endian or little-endian host */
1708 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1709 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1710 /* Mask in the root object ID too, to disambiguate subvols */
1711 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1712 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1713
1714 return 0;
1715}
1716
1717static void btrfs_kill_super(struct super_block *sb)
1718{
1719 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1720 kill_anon_super(sb);
1721 free_fs_info(fs_info);
1722}
1723
1724static struct file_system_type btrfs_fs_type = {
1725 .owner = THIS_MODULE,
1726 .name = "btrfs",
1727 .mount = btrfs_mount,
1728 .kill_sb = btrfs_kill_super,
1729 .fs_flags = FS_REQUIRES_DEV,
1730};
1731MODULE_ALIAS_FS("btrfs");
1732
1733/*
1734 * used by btrfsctl to scan devices when no FS is mounted
1735 */
1736static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1737 unsigned long arg)
1738{
1739 struct btrfs_ioctl_vol_args *vol;
1740 struct btrfs_fs_devices *fs_devices;
1741 int ret = -ENOTTY;
1742
1743 if (!capable(CAP_SYS_ADMIN))
1744 return -EPERM;
1745
1746 vol = memdup_user((void __user *)arg, sizeof(*vol));
1747 if (IS_ERR(vol))
1748 return PTR_ERR(vol);
1749
1750 switch (cmd) {
1751 case BTRFS_IOC_SCAN_DEV:
1752 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1753 &btrfs_fs_type, &fs_devices);
1754 break;
1755 case BTRFS_IOC_DEVICES_READY:
1756 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1757 &btrfs_fs_type, &fs_devices);
1758 if (ret)
1759 break;
1760 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1761 break;
1762 }
1763
1764 kfree(vol);
1765 return ret;
1766}
1767
1768static int btrfs_freeze(struct super_block *sb)
1769{
1770 struct btrfs_trans_handle *trans;
1771 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1772
1773 trans = btrfs_attach_transaction_barrier(root);
1774 if (IS_ERR(trans)) {
1775 /* no transaction, don't bother */
1776 if (PTR_ERR(trans) == -ENOENT)
1777 return 0;
1778 return PTR_ERR(trans);
1779 }
1780 return btrfs_commit_transaction(trans, root);
1781}
1782
1783static int btrfs_unfreeze(struct super_block *sb)
1784{
1785 return 0;
1786}
1787
1788static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1789{
1790 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1791 struct btrfs_fs_devices *cur_devices;
1792 struct btrfs_device *dev, *first_dev = NULL;
1793 struct list_head *head;
1794 struct rcu_string *name;
1795
1796 mutex_lock(&fs_info->fs_devices->device_list_mutex);
1797 cur_devices = fs_info->fs_devices;
1798 while (cur_devices) {
1799 head = &cur_devices->devices;
1800 list_for_each_entry(dev, head, dev_list) {
1801 if (dev->missing)
1802 continue;
1803 if (!first_dev || dev->devid < first_dev->devid)
1804 first_dev = dev;
1805 }
1806 cur_devices = cur_devices->seed;
1807 }
1808
1809 if (first_dev) {
1810 rcu_read_lock();
1811 name = rcu_dereference(first_dev->name);
1812 seq_escape(m, name->str, " \t\n\\");
1813 rcu_read_unlock();
1814 } else {
1815 WARN_ON(1);
1816 }
1817 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1818 return 0;
1819}
1820
1821static const struct super_operations btrfs_super_ops = {
1822 .drop_inode = btrfs_drop_inode,
1823 .evict_inode = btrfs_evict_inode,
1824 .put_super = btrfs_put_super,
1825 .sync_fs = btrfs_sync_fs,
1826 .show_options = btrfs_show_options,
1827 .show_devname = btrfs_show_devname,
1828 .write_inode = btrfs_write_inode,
1829 .alloc_inode = btrfs_alloc_inode,
1830 .destroy_inode = btrfs_destroy_inode,
1831 .statfs = btrfs_statfs,
1832 .remount_fs = btrfs_remount,
1833 .freeze_fs = btrfs_freeze,
1834 .unfreeze_fs = btrfs_unfreeze,
1835};
1836
1837static const struct file_operations btrfs_ctl_fops = {
1838 .unlocked_ioctl = btrfs_control_ioctl,
1839 .compat_ioctl = btrfs_control_ioctl,
1840 .owner = THIS_MODULE,
1841 .llseek = noop_llseek,
1842};
1843
1844static struct miscdevice btrfs_misc = {
1845 .minor = BTRFS_MINOR,
1846 .name = "btrfs-control",
1847 .fops = &btrfs_ctl_fops
1848};
1849
1850MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1851MODULE_ALIAS("devname:btrfs-control");
1852
1853static int btrfs_interface_init(void)
1854{
1855 return misc_register(&btrfs_misc);
1856}
1857
1858static void btrfs_interface_exit(void)
1859{
1860 if (misc_deregister(&btrfs_misc) < 0)
1861 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
1862}
1863
1864static void btrfs_print_info(void)
1865{
1866 printk(KERN_INFO "Btrfs loaded"
1867#ifdef CONFIG_BTRFS_DEBUG
1868 ", debug=on"
1869#endif
1870#ifdef CONFIG_BTRFS_ASSERT
1871 ", assert=on"
1872#endif
1873#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1874 ", integrity-checker=on"
1875#endif
1876 "\n");
1877}
1878
1879static int btrfs_run_sanity_tests(void)
1880{
1881 int ret;
1882
1883 ret = btrfs_init_test_fs();
1884 if (ret)
1885 return ret;
1886
1887 ret = btrfs_test_free_space_cache();
1888 if (ret)
1889 goto out;
1890 ret = btrfs_test_extent_buffer_operations();
1891 if (ret)
1892 goto out;
1893 ret = btrfs_test_extent_io();
1894 if (ret)
1895 goto out;
1896 ret = btrfs_test_inodes();
1897out:
1898 btrfs_destroy_test_fs();
1899 return ret;
1900}
1901
1902static int __init init_btrfs_fs(void)
1903{
1904 int err;
1905
1906 err = btrfs_hash_init();
1907 if (err)
1908 return err;
1909
1910 btrfs_props_init();
1911
1912 err = btrfs_init_sysfs();
1913 if (err)
1914 goto free_hash;
1915
1916 btrfs_init_compress();
1917
1918 err = btrfs_init_cachep();
1919 if (err)
1920 goto free_compress;
1921
1922 err = extent_io_init();
1923 if (err)
1924 goto free_cachep;
1925
1926 err = extent_map_init();
1927 if (err)
1928 goto free_extent_io;
1929
1930 err = ordered_data_init();
1931 if (err)
1932 goto free_extent_map;
1933
1934 err = btrfs_delayed_inode_init();
1935 if (err)
1936 goto free_ordered_data;
1937
1938 err = btrfs_auto_defrag_init();
1939 if (err)
1940 goto free_delayed_inode;
1941
1942 err = btrfs_delayed_ref_init();
1943 if (err)
1944 goto free_auto_defrag;
1945
1946 err = btrfs_prelim_ref_init();
1947 if (err)
1948 goto free_prelim_ref;
1949
1950 err = btrfs_interface_init();
1951 if (err)
1952 goto free_delayed_ref;
1953
1954 btrfs_init_lockdep();
1955
1956 btrfs_print_info();
1957
1958 err = btrfs_run_sanity_tests();
1959 if (err)
1960 goto unregister_ioctl;
1961
1962 err = register_filesystem(&btrfs_fs_type);
1963 if (err)
1964 goto unregister_ioctl;
1965
1966 return 0;
1967
1968unregister_ioctl:
1969 btrfs_interface_exit();
1970free_prelim_ref:
1971 btrfs_prelim_ref_exit();
1972free_delayed_ref:
1973 btrfs_delayed_ref_exit();
1974free_auto_defrag:
1975 btrfs_auto_defrag_exit();
1976free_delayed_inode:
1977 btrfs_delayed_inode_exit();
1978free_ordered_data:
1979 ordered_data_exit();
1980free_extent_map:
1981 extent_map_exit();
1982free_extent_io:
1983 extent_io_exit();
1984free_cachep:
1985 btrfs_destroy_cachep();
1986free_compress:
1987 btrfs_exit_compress();
1988 btrfs_exit_sysfs();
1989free_hash:
1990 btrfs_hash_exit();
1991 return err;
1992}
1993
1994static void __exit exit_btrfs_fs(void)
1995{
1996 btrfs_destroy_cachep();
1997 btrfs_delayed_ref_exit();
1998 btrfs_auto_defrag_exit();
1999 btrfs_delayed_inode_exit();
2000 btrfs_prelim_ref_exit();
2001 ordered_data_exit();
2002 extent_map_exit();
2003 extent_io_exit();
2004 btrfs_interface_exit();
2005 unregister_filesystem(&btrfs_fs_type);
2006 btrfs_exit_sysfs();
2007 btrfs_cleanup_fs_uuids();
2008 btrfs_exit_compress();
2009 btrfs_hash_exit();
2010}
2011
2012late_initcall(init_btrfs_fs);
2013module_exit(exit_btrfs_fs)
2014
2015MODULE_LICENSE("GPL");