?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'chrome-platform-for-linus-4.13' of git://git.kernel.org/pub/scm/linux...
[linux/fpc-iii.git] / fs / btrfs / super.c
blob74e47794e63ffdda6921f8445bd4f576f22f1434
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19 #include <linux/blkdev.h>
20 #include <linux/module.h>
21 #include <linux/buffer_head.h>
22 #include <linux/fs.h>
23 #include <linux/pagemap.h>
24 #include <linux/highmem.h>
25 #include <linux/time.h>
26 #include <linux/init.h>
27 #include <linux/seq_file.h>
28 #include <linux/string.h>
29 #include <linux/backing-dev.h>
30 #include <linux/mount.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/parser.h>
37 #include <linux/ctype.h>
38 #include <linux/namei.h>
39 #include <linux/miscdevice.h>
40 #include <linux/magic.h>
41 #include <linux/slab.h>
42 #include <linux/cleancache.h>
43 #include <linux/ratelimit.h>
44 #include <linux/btrfs.h>
45 #include "delayed-inode.h"
46 #include "ctree.h"
47 #include "disk-io.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "print-tree.h"
51 #include "hash.h"
52 #include "props.h"
53 #include "xattr.h"
54 #include "volumes.h"
55 #include "export.h"
56 #include "compression.h"
57 #include "rcu-string.h"
58 #include "dev-replace.h"
59 #include "free-space-cache.h"
60 #include "backref.h"
61 #include "tests/btrfs-tests.h"
62
63 #include "qgroup.h"
64 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h>
66
67 static const struct super_operations btrfs_super_ops;
68 static struct file_system_type btrfs_fs_type;
69
70 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
71
72 const char *btrfs_decode_error(int errno)
73 {
74 char *errstr = "unknown";
75
76 switch (errno) {
77 case -EIO:
78 errstr = "IO failure";
79 break;
80 case -ENOMEM:
81 errstr = "Out of memory";
82 break;
83 case -EROFS:
84 errstr = "Readonly filesystem";
85 break;
86 case -EEXIST:
87 errstr = "Object already exists";
88 break;
89 case -ENOSPC:
90 errstr = "No space left";
91 break;
92 case -ENOENT:
93 errstr = "No such entry";
94 break;
95 }
96
97 return errstr;
98 }
99
100 /* btrfs handle error by forcing the filesystem readonly */
101 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
102 {
103 struct super_block *sb = fs_info->sb;
104
105 if (sb->s_flags & MS_RDONLY)
106 return;
107
108 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
109 sb->s_flags |= MS_RDONLY;
110 btrfs_info(fs_info, "forced readonly");
111 /*
112 * Note that a running device replace operation is not
113 * canceled here although there is no way to update
114 * the progress. It would add the risk of a deadlock,
115 * therefore the canceling is omitted. The only penalty
116 * is that some I/O remains active until the procedure
117 * completes. The next time when the filesystem is
118 * mounted writeable again, the device replace
119 * operation continues.
120 */
121 }
122 }
123
124 /*
125 * __btrfs_handle_fs_error decodes expected errors from the caller and
126 * invokes the approciate error response.
127 */
128 __cold
129 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
130 unsigned int line, int errno, const char *fmt, ...)
131 {
132 struct super_block *sb = fs_info->sb;
133 #ifdef CONFIG_PRINTK
134 const char *errstr;
135 #endif
136
137 /*
138 * Special case: if the error is EROFS, and we're already
139 * under MS_RDONLY, then it is safe here.
140 */
141 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
142 return;
143
144 #ifdef CONFIG_PRINTK
145 errstr = btrfs_decode_error(errno);
146 if (fmt) {
147 struct va_format vaf;
148 va_list args;
149
150 va_start(args, fmt);
151 vaf.fmt = fmt;
152 vaf.va = &args;
153
154 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
155 sb->s_id, function, line, errno, errstr, &vaf);
156 va_end(args);
157 } else {
158 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
159 sb->s_id, function, line, errno, errstr);
160 }
161 #endif
162
163 /*
164 * Today we only save the error info to memory. Long term we'll
165 * also send it down to the disk
166 */
167 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
168
169 /* Don't go through full error handling during mount */
170 if (sb->s_flags & MS_BORN)
171 btrfs_handle_error(fs_info);
172 }
173
174 #ifdef CONFIG_PRINTK
175 static const char * const logtypes[] = {
176 "emergency",
177 "alert",
178 "critical",
179 "error",
180 "warning",
181 "notice",
182 "info",
183 "debug",
184 };
185
186
187 /*
188 * Use one ratelimit state per log level so that a flood of less important
189 * messages doesn't cause more important ones to be dropped.
190 */
191 static struct ratelimit_state printk_limits[] = {
192 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
193 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
194 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
195 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
196 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
197 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
198 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
199 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
200 };
201
202 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
203 {
204 struct super_block *sb = fs_info->sb;
205 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
206 struct va_format vaf;
207 va_list args;
208 int kern_level;
209 const char *type = logtypes[4];
210 struct ratelimit_state *ratelimit = &printk_limits[4];
211
212 va_start(args, fmt);
213
214 while ((kern_level = printk_get_level(fmt)) != 0) {
215 size_t size = printk_skip_level(fmt) - fmt;
216
217 if (kern_level >= '0' && kern_level <= '7') {
218 memcpy(lvl, fmt, size);
219 lvl[size] = '\0';
220 type = logtypes[kern_level - '0'];
221 ratelimit = &printk_limits[kern_level - '0'];
222 }
223 fmt += size;
224 }
225
226 vaf.fmt = fmt;
227 vaf.va = &args;
228
229 if (__ratelimit(ratelimit))
230 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
231
232 va_end(args);
233 }
234 #endif
235
236 /*
237 * We only mark the transaction aborted and then set the file system read-only.
238 * This will prevent new transactions from starting or trying to join this
239 * one.
240 *
241 * This means that error recovery at the call site is limited to freeing
242 * any local memory allocations and passing the error code up without
243 * further cleanup. The transaction should complete as it normally would
244 * in the call path but will return -EIO.
245 *
246 * We'll complete the cleanup in btrfs_end_transaction and
247 * btrfs_commit_transaction.
248 */
249 __cold
250 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
251 const char *function,
252 unsigned int line, int errno)
253 {
254 struct btrfs_fs_info *fs_info = trans->fs_info;
255
256 trans->aborted = errno;
257 /* Nothing used. The other threads that have joined this
258 * transaction may be able to continue. */
259 if (!trans->dirty && list_empty(&trans->new_bgs)) {
260 const char *errstr;
261
262 errstr = btrfs_decode_error(errno);
263 btrfs_warn(fs_info,
264 "%s:%d: Aborting unused transaction(%s).",
265 function, line, errstr);
266 return;
267 }
268 WRITE_ONCE(trans->transaction->aborted, errno);
269 /* Wake up anybody who may be waiting on this transaction */
270 wake_up(&fs_info->transaction_wait);
271 wake_up(&fs_info->transaction_blocked_wait);
272 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
273 }
274 /*
275 * __btrfs_panic decodes unexpected, fatal errors from the caller,
276 * issues an alert, and either panics or BUGs, depending on mount options.
277 */
278 __cold
279 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
280 unsigned int line, int errno, const char *fmt, ...)
281 {
282 char *s_id = "<unknown>";
283 const char *errstr;
284 struct va_format vaf = { .fmt = fmt };
285 va_list args;
286
287 if (fs_info)
288 s_id = fs_info->sb->s_id;
289
290 va_start(args, fmt);
291 vaf.va = &args;
292
293 errstr = btrfs_decode_error(errno);
294 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
295 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
296 s_id, function, line, &vaf, errno, errstr);
297
298 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
299 function, line, &vaf, errno, errstr);
300 va_end(args);
301 /* Caller calls BUG() */
302 }
303
304 static void btrfs_put_super(struct super_block *sb)
305 {
306 close_ctree(btrfs_sb(sb));
307 }
308
309 enum {
310 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
311 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
312 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
313 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
314 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
315 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
316 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
317 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
318 Opt_skip_balance, Opt_check_integrity,
319 Opt_check_integrity_including_extent_data,
320 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
321 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
322 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
323 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
324 Opt_nologreplay, Opt_norecovery,
325 #ifdef CONFIG_BTRFS_DEBUG
326 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
327 #endif
328 Opt_err,
329 };
330
331 static const match_table_t tokens = {
332 {Opt_degraded, "degraded"},
333 {Opt_subvol, "subvol=%s"},
334 {Opt_subvolid, "subvolid=%s"},
335 {Opt_device, "device=%s"},
336 {Opt_nodatasum, "nodatasum"},
337 {Opt_datasum, "datasum"},
338 {Opt_nodatacow, "nodatacow"},
339 {Opt_datacow, "datacow"},
340 {Opt_nobarrier, "nobarrier"},
341 {Opt_barrier, "barrier"},
342 {Opt_max_inline, "max_inline=%s"},
343 {Opt_alloc_start, "alloc_start=%s"},
344 {Opt_thread_pool, "thread_pool=%d"},
345 {Opt_compress, "compress"},
346 {Opt_compress_type, "compress=%s"},
347 {Opt_compress_force, "compress-force"},
348 {Opt_compress_force_type, "compress-force=%s"},
349 {Opt_ssd, "ssd"},
350 {Opt_ssd_spread, "ssd_spread"},
351 {Opt_nossd, "nossd"},
352 {Opt_acl, "acl"},
353 {Opt_noacl, "noacl"},
354 {Opt_notreelog, "notreelog"},
355 {Opt_treelog, "treelog"},
356 {Opt_nologreplay, "nologreplay"},
357 {Opt_norecovery, "norecovery"},
358 {Opt_flushoncommit, "flushoncommit"},
359 {Opt_noflushoncommit, "noflushoncommit"},
360 {Opt_ratio, "metadata_ratio=%d"},
361 {Opt_discard, "discard"},
362 {Opt_nodiscard, "nodiscard"},
363 {Opt_space_cache, "space_cache"},
364 {Opt_space_cache_version, "space_cache=%s"},
365 {Opt_clear_cache, "clear_cache"},
366 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
367 {Opt_enospc_debug, "enospc_debug"},
368 {Opt_noenospc_debug, "noenospc_debug"},
369 {Opt_subvolrootid, "subvolrootid=%d"},
370 {Opt_defrag, "autodefrag"},
371 {Opt_nodefrag, "noautodefrag"},
372 {Opt_inode_cache, "inode_cache"},
373 {Opt_noinode_cache, "noinode_cache"},
374 {Opt_no_space_cache, "nospace_cache"},
375 {Opt_recovery, "recovery"}, /* deprecated */
376 {Opt_usebackuproot, "usebackuproot"},
377 {Opt_skip_balance, "skip_balance"},
378 {Opt_check_integrity, "check_int"},
379 {Opt_check_integrity_including_extent_data, "check_int_data"},
380 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
381 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
382 {Opt_fatal_errors, "fatal_errors=%s"},
383 {Opt_commit_interval, "commit=%d"},
384 #ifdef CONFIG_BTRFS_DEBUG
385 {Opt_fragment_data, "fragment=data"},
386 {Opt_fragment_metadata, "fragment=metadata"},
387 {Opt_fragment_all, "fragment=all"},
388 #endif
389 {Opt_err, NULL},
390 };
391
392 /*
393 * Regular mount options parser. Everything that is needed only when
394 * reading in a new superblock is parsed here.
395 * XXX JDM: This needs to be cleaned up for remount.
396 */
397 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
398 unsigned long new_flags)
399 {
400 substring_t args[MAX_OPT_ARGS];
401 char *p, *num, *orig = NULL;
402 u64 cache_gen;
403 int intarg;
404 int ret = 0;
405 char *compress_type;
406 bool compress_force = false;
407 enum btrfs_compression_type saved_compress_type;
408 bool saved_compress_force;
409 int no_compress = 0;
410
411 cache_gen = btrfs_super_cache_generation(info->super_copy);
412 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
413 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
414 else if (cache_gen)
415 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
416
417 /*
418 * Even the options are empty, we still need to do extra check
419 * against new flags
420 */
421 if (!options)
422 goto check;
423
424 /*
425 * strsep changes the string, duplicate it because parse_options
426 * gets called twice
427 */
428 options = kstrdup(options, GFP_NOFS);
429 if (!options)
430 return -ENOMEM;
431
432 orig = options;
433
434 while ((p = strsep(&options, ",")) != NULL) {
435 int token;
436 if (!*p)
437 continue;
438
439 token = match_token(p, tokens, args);
440 switch (token) {
441 case Opt_degraded:
442 btrfs_info(info, "allowing degraded mounts");
443 btrfs_set_opt(info->mount_opt, DEGRADED);
444 break;
445 case Opt_subvol:
446 case Opt_subvolid:
447 case Opt_subvolrootid:
448 case Opt_device:
449 /*
450 * These are parsed by btrfs_parse_early_options
451 * and can be happily ignored here.
452 */
453 break;
454 case Opt_nodatasum:
455 btrfs_set_and_info(info, NODATASUM,
456 "setting nodatasum");
457 break;
458 case Opt_datasum:
459 if (btrfs_test_opt(info, NODATASUM)) {
460 if (btrfs_test_opt(info, NODATACOW))
461 btrfs_info(info,
462 "setting datasum, datacow enabled");
463 else
464 btrfs_info(info, "setting datasum");
465 }
466 btrfs_clear_opt(info->mount_opt, NODATACOW);
467 btrfs_clear_opt(info->mount_opt, NODATASUM);
468 break;
469 case Opt_nodatacow:
470 if (!btrfs_test_opt(info, NODATACOW)) {
471 if (!btrfs_test_opt(info, COMPRESS) ||
472 !btrfs_test_opt(info, FORCE_COMPRESS)) {
473 btrfs_info(info,
474 "setting nodatacow, compression disabled");
475 } else {
476 btrfs_info(info, "setting nodatacow");
477 }
478 }
479 btrfs_clear_opt(info->mount_opt, COMPRESS);
480 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
481 btrfs_set_opt(info->mount_opt, NODATACOW);
482 btrfs_set_opt(info->mount_opt, NODATASUM);
483 break;
484 case Opt_datacow:
485 btrfs_clear_and_info(info, NODATACOW,
486 "setting datacow");
487 break;
488 case Opt_compress_force:
489 case Opt_compress_force_type:
490 compress_force = true;
491 /* Fallthrough */
492 case Opt_compress:
493 case Opt_compress_type:
494 saved_compress_type = btrfs_test_opt(info,
495 COMPRESS) ?
496 info->compress_type : BTRFS_COMPRESS_NONE;
497 saved_compress_force =
498 btrfs_test_opt(info, FORCE_COMPRESS);
499 if (token == Opt_compress ||
500 token == Opt_compress_force ||
501 strcmp(args[0].from, "zlib") == 0) {
502 compress_type = "zlib";
503 info->compress_type = BTRFS_COMPRESS_ZLIB;
504 btrfs_set_opt(info->mount_opt, COMPRESS);
505 btrfs_clear_opt(info->mount_opt, NODATACOW);
506 btrfs_clear_opt(info->mount_opt, NODATASUM);
507 no_compress = 0;
508 } else if (strcmp(args[0].from, "lzo") == 0) {
509 compress_type = "lzo";
510 info->compress_type = BTRFS_COMPRESS_LZO;
511 btrfs_set_opt(info->mount_opt, COMPRESS);
512 btrfs_clear_opt(info->mount_opt, NODATACOW);
513 btrfs_clear_opt(info->mount_opt, NODATASUM);
514 btrfs_set_fs_incompat(info, COMPRESS_LZO);
515 no_compress = 0;
516 } else if (strncmp(args[0].from, "no", 2) == 0) {
517 compress_type = "no";
518 btrfs_clear_opt(info->mount_opt, COMPRESS);
519 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
520 compress_force = false;
521 no_compress++;
522 } else {
523 ret = -EINVAL;
524 goto out;
525 }
526
527 if (compress_force) {
528 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
529 } else {
530 /*
531 * If we remount from compress-force=xxx to
532 * compress=xxx, we need clear FORCE_COMPRESS
533 * flag, otherwise, there is no way for users
534 * to disable forcible compression separately.
535 */
536 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
537 }
538 if ((btrfs_test_opt(info, COMPRESS) &&
539 (info->compress_type != saved_compress_type ||
540 compress_force != saved_compress_force)) ||
541 (!btrfs_test_opt(info, COMPRESS) &&
542 no_compress == 1)) {
543 btrfs_info(info, "%s %s compression",
544 (compress_force) ? "force" : "use",
545 compress_type);
546 }
547 compress_force = false;
548 break;
549 case Opt_ssd:
550 btrfs_set_and_info(info, SSD,
551 "use ssd allocation scheme");
552 btrfs_clear_opt(info->mount_opt, NOSSD);
553 break;
554 case Opt_ssd_spread:
555 btrfs_set_and_info(info, SSD_SPREAD,
556 "use spread ssd allocation scheme");
557 btrfs_set_opt(info->mount_opt, SSD);
558 btrfs_clear_opt(info->mount_opt, NOSSD);
559 break;
560 case Opt_nossd:
561 btrfs_set_and_info(info, NOSSD,
562 "not using ssd allocation scheme");
563 btrfs_clear_opt(info->mount_opt, SSD);
564 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
565 break;
566 case Opt_barrier:
567 btrfs_clear_and_info(info, NOBARRIER,
568 "turning on barriers");
569 break;
570 case Opt_nobarrier:
571 btrfs_set_and_info(info, NOBARRIER,
572 "turning off barriers");
573 break;
574 case Opt_thread_pool:
575 ret = match_int(&args[0], &intarg);
576 if (ret) {
577 goto out;
578 } else if (intarg > 0) {
579 info->thread_pool_size = intarg;
580 } else {
581 ret = -EINVAL;
582 goto out;
583 }
584 break;
585 case Opt_max_inline:
586 num = match_strdup(&args[0]);
587 if (num) {
588 info->max_inline = memparse(num, NULL);
589 kfree(num);
590
591 if (info->max_inline) {
592 info->max_inline = min_t(u64,
593 info->max_inline,
594 info->sectorsize);
595 }
596 btrfs_info(info, "max_inline at %llu",
597 info->max_inline);
598 } else {
599 ret = -ENOMEM;
600 goto out;
601 }
602 break;
603 case Opt_alloc_start:
604 btrfs_info(info,
605 "option alloc_start is obsolete, ignored");
606 break;
607 case Opt_acl:
608 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
609 info->sb->s_flags |= MS_POSIXACL;
610 break;
611 #else
612 btrfs_err(info, "support for ACL not compiled in!");
613 ret = -EINVAL;
614 goto out;
615 #endif
616 case Opt_noacl:
617 info->sb->s_flags &= ~MS_POSIXACL;
618 break;
619 case Opt_notreelog:
620 btrfs_set_and_info(info, NOTREELOG,
621 "disabling tree log");
622 break;
623 case Opt_treelog:
624 btrfs_clear_and_info(info, NOTREELOG,
625 "enabling tree log");
626 break;
627 case Opt_norecovery:
628 case Opt_nologreplay:
629 btrfs_set_and_info(info, NOLOGREPLAY,
630 "disabling log replay at mount time");
631 break;
632 case Opt_flushoncommit:
633 btrfs_set_and_info(info, FLUSHONCOMMIT,
634 "turning on flush-on-commit");
635 break;
636 case Opt_noflushoncommit:
637 btrfs_clear_and_info(info, FLUSHONCOMMIT,
638 "turning off flush-on-commit");
639 break;
640 case Opt_ratio:
641 ret = match_int(&args[0], &intarg);
642 if (ret) {
643 goto out;
644 } else if (intarg >= 0) {
645 info->metadata_ratio = intarg;
646 btrfs_info(info, "metadata ratio %d",
647 info->metadata_ratio);
648 } else {
649 ret = -EINVAL;
650 goto out;
651 }
652 break;
653 case Opt_discard:
654 btrfs_set_and_info(info, DISCARD,
655 "turning on discard");
656 break;
657 case Opt_nodiscard:
658 btrfs_clear_and_info(info, DISCARD,
659 "turning off discard");
660 break;
661 case Opt_space_cache:
662 case Opt_space_cache_version:
663 if (token == Opt_space_cache ||
664 strcmp(args[0].from, "v1") == 0) {
665 btrfs_clear_opt(info->mount_opt,
666 FREE_SPACE_TREE);
667 btrfs_set_and_info(info, SPACE_CACHE,
668 "enabling disk space caching");
669 } else if (strcmp(args[0].from, "v2") == 0) {
670 btrfs_clear_opt(info->mount_opt,
671 SPACE_CACHE);
672 btrfs_set_and_info(info, FREE_SPACE_TREE,
673 "enabling free space tree");
674 } else {
675 ret = -EINVAL;
676 goto out;
677 }
678 break;
679 case Opt_rescan_uuid_tree:
680 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
681 break;
682 case Opt_no_space_cache:
683 if (btrfs_test_opt(info, SPACE_CACHE)) {
684 btrfs_clear_and_info(info, SPACE_CACHE,
685 "disabling disk space caching");
686 }
687 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
688 btrfs_clear_and_info(info, FREE_SPACE_TREE,
689 "disabling free space tree");
690 }
691 break;
692 case Opt_inode_cache:
693 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
694 "enabling inode map caching");
695 break;
696 case Opt_noinode_cache:
697 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
698 "disabling inode map caching");
699 break;
700 case Opt_clear_cache:
701 btrfs_set_and_info(info, CLEAR_CACHE,
702 "force clearing of disk cache");
703 break;
704 case Opt_user_subvol_rm_allowed:
705 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
706 break;
707 case Opt_enospc_debug:
708 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
709 break;
710 case Opt_noenospc_debug:
711 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
712 break;
713 case Opt_defrag:
714 btrfs_set_and_info(info, AUTO_DEFRAG,
715 "enabling auto defrag");
716 break;
717 case Opt_nodefrag:
718 btrfs_clear_and_info(info, AUTO_DEFRAG,
719 "disabling auto defrag");
720 break;
721 case Opt_recovery:
722 btrfs_warn(info,
723 "'recovery' is deprecated, use 'usebackuproot' instead");
724 case Opt_usebackuproot:
725 btrfs_info(info,
726 "trying to use backup root at mount time");
727 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
728 break;
729 case Opt_skip_balance:
730 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
731 break;
732 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
733 case Opt_check_integrity_including_extent_data:
734 btrfs_info(info,
735 "enabling check integrity including extent data");
736 btrfs_set_opt(info->mount_opt,
737 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
738 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
739 break;
740 case Opt_check_integrity:
741 btrfs_info(info, "enabling check integrity");
742 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
743 break;
744 case Opt_check_integrity_print_mask:
745 ret = match_int(&args[0], &intarg);
746 if (ret) {
747 goto out;
748 } else if (intarg >= 0) {
749 info->check_integrity_print_mask = intarg;
750 btrfs_info(info,
751 "check_integrity_print_mask 0x%x",
752 info->check_integrity_print_mask);
753 } else {
754 ret = -EINVAL;
755 goto out;
756 }
757 break;
758 #else
759 case Opt_check_integrity_including_extent_data:
760 case Opt_check_integrity:
761 case Opt_check_integrity_print_mask:
762 btrfs_err(info,
763 "support for check_integrity* not compiled in!");
764 ret = -EINVAL;
765 goto out;
766 #endif
767 case Opt_fatal_errors:
768 if (strcmp(args[0].from, "panic") == 0)
769 btrfs_set_opt(info->mount_opt,
770 PANIC_ON_FATAL_ERROR);
771 else if (strcmp(args[0].from, "bug") == 0)
772 btrfs_clear_opt(info->mount_opt,
773 PANIC_ON_FATAL_ERROR);
774 else {
775 ret = -EINVAL;
776 goto out;
777 }
778 break;
779 case Opt_commit_interval:
780 intarg = 0;
781 ret = match_int(&args[0], &intarg);
782 if (ret < 0) {
783 btrfs_err(info, "invalid commit interval");
784 ret = -EINVAL;
785 goto out;
786 }
787 if (intarg > 0) {
788 if (intarg > 300) {
789 btrfs_warn(info,
790 "excessive commit interval %d",
791 intarg);
792 }
793 info->commit_interval = intarg;
794 } else {
795 btrfs_info(info,
796 "using default commit interval %ds",
797 BTRFS_DEFAULT_COMMIT_INTERVAL);
798 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
799 }
800 break;
801 #ifdef CONFIG_BTRFS_DEBUG
802 case Opt_fragment_all:
803 btrfs_info(info, "fragmenting all space");
804 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
805 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
806 break;
807 case Opt_fragment_metadata:
808 btrfs_info(info, "fragmenting metadata");
809 btrfs_set_opt(info->mount_opt,
810 FRAGMENT_METADATA);
811 break;
812 case Opt_fragment_data:
813 btrfs_info(info, "fragmenting data");
814 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
815 break;
816 #endif
817 case Opt_err:
818 btrfs_info(info, "unrecognized mount option '%s'", p);
819 ret = -EINVAL;
820 goto out;
821 default:
822 break;
823 }
824 }
825 check:
826 /*
827 * Extra check for current option against current flag
828 */
829 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
830 btrfs_err(info,
831 "nologreplay must be used with ro mount option");
832 ret = -EINVAL;
833 }
834 out:
835 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
836 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
837 !btrfs_test_opt(info, CLEAR_CACHE)) {
838 btrfs_err(info, "cannot disable free space tree");
839 ret = -EINVAL;
840
841 }
842 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
843 btrfs_info(info, "disk space caching is enabled");
844 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
845 btrfs_info(info, "using free space tree");
846 kfree(orig);
847 return ret;
848 }
849
850 /*
851 * Parse mount options that are required early in the mount process.
852 *
853 * All other options will be parsed on much later in the mount process and
854 * only when we need to allocate a new super block.
855 */
856 static int btrfs_parse_early_options(const char *options, fmode_t flags,
857 void *holder, char **subvol_name, u64 *subvol_objectid,
858 struct btrfs_fs_devices **fs_devices)
859 {
860 substring_t args[MAX_OPT_ARGS];
861 char *device_name, *opts, *orig, *p;
862 char *num = NULL;
863 int error = 0;
864
865 if (!options)
866 return 0;
867
868 /*
869 * strsep changes the string, duplicate it because parse_options
870 * gets called twice
871 */
872 opts = kstrdup(options, GFP_KERNEL);
873 if (!opts)
874 return -ENOMEM;
875 orig = opts;
876
877 while ((p = strsep(&opts, ",")) != NULL) {
878 int token;
879 if (!*p)
880 continue;
881
882 token = match_token(p, tokens, args);
883 switch (token) {
884 case Opt_subvol:
885 kfree(*subvol_name);
886 *subvol_name = match_strdup(&args[0]);
887 if (!*subvol_name) {
888 error = -ENOMEM;
889 goto out;
890 }
891 break;
892 case Opt_subvolid:
893 num = match_strdup(&args[0]);
894 if (num) {
895 *subvol_objectid = memparse(num, NULL);
896 kfree(num);
897 /* we want the original fs_tree */
898 if (!*subvol_objectid)
899 *subvol_objectid =
900 BTRFS_FS_TREE_OBJECTID;
901 } else {
902 error = -EINVAL;
903 goto out;
904 }
905 break;
906 case Opt_subvolrootid:
907 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
908 break;
909 case Opt_device:
910 device_name = match_strdup(&args[0]);
911 if (!device_name) {
912 error = -ENOMEM;
913 goto out;
914 }
915 error = btrfs_scan_one_device(device_name,
916 flags, holder, fs_devices);
917 kfree(device_name);
918 if (error)
919 goto out;
920 break;
921 default:
922 break;
923 }
924 }
925
926 out:
927 kfree(orig);
928 return error;
929 }
930
931 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
932 u64 subvol_objectid)
933 {
934 struct btrfs_root *root = fs_info->tree_root;
935 struct btrfs_root *fs_root;
936 struct btrfs_root_ref *root_ref;
937 struct btrfs_inode_ref *inode_ref;
938 struct btrfs_key key;
939 struct btrfs_path *path = NULL;
940 char *name = NULL, *ptr;
941 u64 dirid;
942 int len;
943 int ret;
944
945 path = btrfs_alloc_path();
946 if (!path) {
947 ret = -ENOMEM;
948 goto err;
949 }
950 path->leave_spinning = 1;
951
952 name = kmalloc(PATH_MAX, GFP_NOFS);
953 if (!name) {
954 ret = -ENOMEM;
955 goto err;
956 }
957 ptr = name + PATH_MAX - 1;
958 ptr[0] = '\0';
959
960 /*
961 * Walk up the subvolume trees in the tree of tree roots by root
962 * backrefs until we hit the top-level subvolume.
963 */
964 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
965 key.objectid = subvol_objectid;
966 key.type = BTRFS_ROOT_BACKREF_KEY;
967 key.offset = (u64)-1;
968
969 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
970 if (ret < 0) {
971 goto err;
972 } else if (ret > 0) {
973 ret = btrfs_previous_item(root, path, subvol_objectid,
974 BTRFS_ROOT_BACKREF_KEY);
975 if (ret < 0) {
976 goto err;
977 } else if (ret > 0) {
978 ret = -ENOENT;
979 goto err;
980 }
981 }
982
983 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
984 subvol_objectid = key.offset;
985
986 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
987 struct btrfs_root_ref);
988 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
989 ptr -= len + 1;
990 if (ptr < name) {
991 ret = -ENAMETOOLONG;
992 goto err;
993 }
994 read_extent_buffer(path->nodes[0], ptr + 1,
995 (unsigned long)(root_ref + 1), len);
996 ptr[0] = '/';
997 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
998 btrfs_release_path(path);
999
1000 key.objectid = subvol_objectid;
1001 key.type = BTRFS_ROOT_ITEM_KEY;
1002 key.offset = (u64)-1;
1003 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1004 if (IS_ERR(fs_root)) {
1005 ret = PTR_ERR(fs_root);
1006 goto err;
1007 }
1008
1009 /*
1010 * Walk up the filesystem tree by inode refs until we hit the
1011 * root directory.
1012 */
1013 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1014 key.objectid = dirid;
1015 key.type = BTRFS_INODE_REF_KEY;
1016 key.offset = (u64)-1;
1017
1018 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1019 if (ret < 0) {
1020 goto err;
1021 } else if (ret > 0) {
1022 ret = btrfs_previous_item(fs_root, path, dirid,
1023 BTRFS_INODE_REF_KEY);
1024 if (ret < 0) {
1025 goto err;
1026 } else if (ret > 0) {
1027 ret = -ENOENT;
1028 goto err;
1029 }
1030 }
1031
1032 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1033 dirid = key.offset;
1034
1035 inode_ref = btrfs_item_ptr(path->nodes[0],
1036 path->slots[0],
1037 struct btrfs_inode_ref);
1038 len = btrfs_inode_ref_name_len(path->nodes[0],
1039 inode_ref);
1040 ptr -= len + 1;
1041 if (ptr < name) {
1042 ret = -ENAMETOOLONG;
1043 goto err;
1044 }
1045 read_extent_buffer(path->nodes[0], ptr + 1,
1046 (unsigned long)(inode_ref + 1), len);
1047 ptr[0] = '/';
1048 btrfs_release_path(path);
1049 }
1050 }
1051
1052 btrfs_free_path(path);
1053 if (ptr == name + PATH_MAX - 1) {
1054 name[0] = '/';
1055 name[1] = '\0';
1056 } else {
1057 memmove(name, ptr, name + PATH_MAX - ptr);
1058 }
1059 return name;
1060
1061 err:
1062 btrfs_free_path(path);
1063 kfree(name);
1064 return ERR_PTR(ret);
1065 }
1066
1067 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1068 {
1069 struct btrfs_root *root = fs_info->tree_root;
1070 struct btrfs_dir_item *di;
1071 struct btrfs_path *path;
1072 struct btrfs_key location;
1073 u64 dir_id;
1074
1075 path = btrfs_alloc_path();
1076 if (!path)
1077 return -ENOMEM;
1078 path->leave_spinning = 1;
1079
1080 /*
1081 * Find the "default" dir item which points to the root item that we
1082 * will mount by default if we haven't been given a specific subvolume
1083 * to mount.
1084 */
1085 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1086 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1087 if (IS_ERR(di)) {
1088 btrfs_free_path(path);
1089 return PTR_ERR(di);
1090 }
1091 if (!di) {
1092 /*
1093 * Ok the default dir item isn't there. This is weird since
1094 * it's always been there, but don't freak out, just try and
1095 * mount the top-level subvolume.
1096 */
1097 btrfs_free_path(path);
1098 *objectid = BTRFS_FS_TREE_OBJECTID;
1099 return 0;
1100 }
1101
1102 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1103 btrfs_free_path(path);
1104 *objectid = location.objectid;
1105 return 0;
1106 }
1107
1108 static int btrfs_fill_super(struct super_block *sb,
1109 struct btrfs_fs_devices *fs_devices,
1110 void *data)
1111 {
1112 struct inode *inode;
1113 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1114 struct btrfs_key key;
1115 int err;
1116
1117 sb->s_maxbytes = MAX_LFS_FILESIZE;
1118 sb->s_magic = BTRFS_SUPER_MAGIC;
1119 sb->s_op = &btrfs_super_ops;
1120 sb->s_d_op = &btrfs_dentry_operations;
1121 sb->s_export_op = &btrfs_export_ops;
1122 sb->s_xattr = btrfs_xattr_handlers;
1123 sb->s_time_gran = 1;
1124 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1125 sb->s_flags |= MS_POSIXACL;
1126 #endif
1127 sb->s_flags |= MS_I_VERSION;
1128 sb->s_iflags |= SB_I_CGROUPWB;
1129
1130 err = super_setup_bdi(sb);
1131 if (err) {
1132 btrfs_err(fs_info, "super_setup_bdi failed");
1133 return err;
1134 }
1135
1136 err = open_ctree(sb, fs_devices, (char *)data);
1137 if (err) {
1138 btrfs_err(fs_info, "open_ctree failed");
1139 return err;
1140 }
1141
1142 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1143 key.type = BTRFS_INODE_ITEM_KEY;
1144 key.offset = 0;
1145 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1146 if (IS_ERR(inode)) {
1147 err = PTR_ERR(inode);
1148 goto fail_close;
1149 }
1150
1151 sb->s_root = d_make_root(inode);
1152 if (!sb->s_root) {
1153 err = -ENOMEM;
1154 goto fail_close;
1155 }
1156
1157 save_mount_options(sb, data);
1158 cleancache_init_fs(sb);
1159 sb->s_flags |= MS_ACTIVE;
1160 return 0;
1161
1162 fail_close:
1163 close_ctree(fs_info);
1164 return err;
1165 }
1166
1167 int btrfs_sync_fs(struct super_block *sb, int wait)
1168 {
1169 struct btrfs_trans_handle *trans;
1170 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1171 struct btrfs_root *root = fs_info->tree_root;
1172
1173 trace_btrfs_sync_fs(fs_info, wait);
1174
1175 if (!wait) {
1176 filemap_flush(fs_info->btree_inode->i_mapping);
1177 return 0;
1178 }
1179
1180 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1181
1182 trans = btrfs_attach_transaction_barrier(root);
1183 if (IS_ERR(trans)) {
1184 /* no transaction, don't bother */
1185 if (PTR_ERR(trans) == -ENOENT) {
1186 /*
1187 * Exit unless we have some pending changes
1188 * that need to go through commit
1189 */
1190 if (fs_info->pending_changes == 0)
1191 return 0;
1192 /*
1193 * A non-blocking test if the fs is frozen. We must not
1194 * start a new transaction here otherwise a deadlock
1195 * happens. The pending operations are delayed to the
1196 * next commit after thawing.
1197 */
1198 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1199 __sb_end_write(sb, SB_FREEZE_WRITE);
1200 else
1201 return 0;
1202 trans = btrfs_start_transaction(root, 0);
1203 }
1204 if (IS_ERR(trans))
1205 return PTR_ERR(trans);
1206 }
1207 return btrfs_commit_transaction(trans);
1208 }
1209
1210 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1211 {
1212 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1213 char *compress_type;
1214
1215 if (btrfs_test_opt(info, DEGRADED))
1216 seq_puts(seq, ",degraded");
1217 if (btrfs_test_opt(info, NODATASUM))
1218 seq_puts(seq, ",nodatasum");
1219 if (btrfs_test_opt(info, NODATACOW))
1220 seq_puts(seq, ",nodatacow");
1221 if (btrfs_test_opt(info, NOBARRIER))
1222 seq_puts(seq, ",nobarrier");
1223 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1224 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1225 if (info->thread_pool_size != min_t(unsigned long,
1226 num_online_cpus() + 2, 8))
1227 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1228 if (btrfs_test_opt(info, COMPRESS)) {
1229 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1230 compress_type = "zlib";
1231 else
1232 compress_type = "lzo";
1233 if (btrfs_test_opt(info, FORCE_COMPRESS))
1234 seq_printf(seq, ",compress-force=%s", compress_type);
1235 else
1236 seq_printf(seq, ",compress=%s", compress_type);
1237 }
1238 if (btrfs_test_opt(info, NOSSD))
1239 seq_puts(seq, ",nossd");
1240 if (btrfs_test_opt(info, SSD_SPREAD))
1241 seq_puts(seq, ",ssd_spread");
1242 else if (btrfs_test_opt(info, SSD))
1243 seq_puts(seq, ",ssd");
1244 if (btrfs_test_opt(info, NOTREELOG))
1245 seq_puts(seq, ",notreelog");
1246 if (btrfs_test_opt(info, NOLOGREPLAY))
1247 seq_puts(seq, ",nologreplay");
1248 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1249 seq_puts(seq, ",flushoncommit");
1250 if (btrfs_test_opt(info, DISCARD))
1251 seq_puts(seq, ",discard");
1252 if (!(info->sb->s_flags & MS_POSIXACL))
1253 seq_puts(seq, ",noacl");
1254 if (btrfs_test_opt(info, SPACE_CACHE))
1255 seq_puts(seq, ",space_cache");
1256 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1257 seq_puts(seq, ",space_cache=v2");
1258 else
1259 seq_puts(seq, ",nospace_cache");
1260 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1261 seq_puts(seq, ",rescan_uuid_tree");
1262 if (btrfs_test_opt(info, CLEAR_CACHE))
1263 seq_puts(seq, ",clear_cache");
1264 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1265 seq_puts(seq, ",user_subvol_rm_allowed");
1266 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1267 seq_puts(seq, ",enospc_debug");
1268 if (btrfs_test_opt(info, AUTO_DEFRAG))
1269 seq_puts(seq, ",autodefrag");
1270 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1271 seq_puts(seq, ",inode_cache");
1272 if (btrfs_test_opt(info, SKIP_BALANCE))
1273 seq_puts(seq, ",skip_balance");
1274 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1275 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1276 seq_puts(seq, ",check_int_data");
1277 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1278 seq_puts(seq, ",check_int");
1279 if (info->check_integrity_print_mask)
1280 seq_printf(seq, ",check_int_print_mask=%d",
1281 info->check_integrity_print_mask);
1282 #endif
1283 if (info->metadata_ratio)
1284 seq_printf(seq, ",metadata_ratio=%d",
1285 info->metadata_ratio);
1286 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1287 seq_puts(seq, ",fatal_errors=panic");
1288 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1289 seq_printf(seq, ",commit=%d", info->commit_interval);
1290 #ifdef CONFIG_BTRFS_DEBUG
1291 if (btrfs_test_opt(info, FRAGMENT_DATA))
1292 seq_puts(seq, ",fragment=data");
1293 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1294 seq_puts(seq, ",fragment=metadata");
1295 #endif
1296 seq_printf(seq, ",subvolid=%llu",
1297 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1298 seq_puts(seq, ",subvol=");
1299 seq_dentry(seq, dentry, " \t\n\\");
1300 return 0;
1301 }
1302
1303 static int btrfs_test_super(struct super_block *s, void *data)
1304 {
1305 struct btrfs_fs_info *p = data;
1306 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1307
1308 return fs_info->fs_devices == p->fs_devices;
1309 }
1310
1311 static int btrfs_set_super(struct super_block *s, void *data)
1312 {
1313 int err = set_anon_super(s, data);
1314 if (!err)
1315 s->s_fs_info = data;
1316 return err;
1317 }
1318
1319 /*
1320 * subvolumes are identified by ino 256
1321 */
1322 static inline int is_subvolume_inode(struct inode *inode)
1323 {
1324 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1325 return 1;
1326 return 0;
1327 }
1328
1329 /*
1330 * This will add subvolid=0 to the argument string while removing any subvol=
1331 * and subvolid= arguments to make sure we get the top-level root for path
1332 * walking to the subvol we want.
1333 */
1334 static char *setup_root_args(char *args)
1335 {
1336 char *buf, *dst, *sep;
1337
1338 if (!args)
1339 return kstrdup("subvolid=0", GFP_NOFS);
1340
1341 /* The worst case is that we add ",subvolid=0" to the end. */
1342 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
1343 if (!buf)
1344 return NULL;
1345
1346 while (1) {
1347 sep = strchrnul(args, ',');
1348 if (!strstarts(args, "subvol=") &&
1349 !strstarts(args, "subvolid=")) {
1350 memcpy(dst, args, sep - args);
1351 dst += sep - args;
1352 *dst++ = ',';
1353 }
1354 if (*sep)
1355 args = sep + 1;
1356 else
1357 break;
1358 }
1359 strcpy(dst, "subvolid=0");
1360
1361 return buf;
1362 }
1363
1364 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1365 int flags, const char *device_name,
1366 char *data)
1367 {
1368 struct dentry *root;
1369 struct vfsmount *mnt = NULL;
1370 char *newargs;
1371 int ret;
1372
1373 newargs = setup_root_args(data);
1374 if (!newargs) {
1375 root = ERR_PTR(-ENOMEM);
1376 goto out;
1377 }
1378
1379 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
1380 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
1381 if (flags & MS_RDONLY) {
1382 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
1383 device_name, newargs);
1384 } else {
1385 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
1386 device_name, newargs);
1387 if (IS_ERR(mnt)) {
1388 root = ERR_CAST(mnt);
1389 mnt = NULL;
1390 goto out;
1391 }
1392
1393 down_write(&mnt->mnt_sb->s_umount);
1394 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1395 up_write(&mnt->mnt_sb->s_umount);
1396 if (ret < 0) {
1397 root = ERR_PTR(ret);
1398 goto out;
1399 }
1400 }
1401 }
1402 if (IS_ERR(mnt)) {
1403 root = ERR_CAST(mnt);
1404 mnt = NULL;
1405 goto out;
1406 }
1407
1408 if (!subvol_name) {
1409 if (!subvol_objectid) {
1410 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1411 &subvol_objectid);
1412 if (ret) {
1413 root = ERR_PTR(ret);
1414 goto out;
1415 }
1416 }
1417 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1418 subvol_objectid);
1419 if (IS_ERR(subvol_name)) {
1420 root = ERR_CAST(subvol_name);
1421 subvol_name = NULL;
1422 goto out;
1423 }
1424
1425 }
1426
1427 root = mount_subtree(mnt, subvol_name);
1428 /* mount_subtree() drops our reference on the vfsmount. */
1429 mnt = NULL;
1430
1431 if (!IS_ERR(root)) {
1432 struct super_block *s = root->d_sb;
1433 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1434 struct inode *root_inode = d_inode(root);
1435 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1436
1437 ret = 0;
1438 if (!is_subvolume_inode(root_inode)) {
1439 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1440 subvol_name);
1441 ret = -EINVAL;
1442 }
1443 if (subvol_objectid && root_objectid != subvol_objectid) {
1444 /*
1445 * This will also catch a race condition where a
1446 * subvolume which was passed by ID is renamed and
1447 * another subvolume is renamed over the old location.
1448 */
1449 btrfs_err(fs_info,
1450 "subvol '%s' does not match subvolid %llu",
1451 subvol_name, subvol_objectid);
1452 ret = -EINVAL;
1453 }
1454 if (ret) {
1455 dput(root);
1456 root = ERR_PTR(ret);
1457 deactivate_locked_super(s);
1458 }
1459 }
1460
1461 out:
1462 mntput(mnt);
1463 kfree(newargs);
1464 kfree(subvol_name);
1465 return root;
1466 }
1467
1468 static int parse_security_options(char *orig_opts,
1469 struct security_mnt_opts *sec_opts)
1470 {
1471 char *secdata = NULL;
1472 int ret = 0;
1473
1474 secdata = alloc_secdata();
1475 if (!secdata)
1476 return -ENOMEM;
1477 ret = security_sb_copy_data(orig_opts, secdata);
1478 if (ret) {
1479 free_secdata(secdata);
1480 return ret;
1481 }
1482 ret = security_sb_parse_opts_str(secdata, sec_opts);
1483 free_secdata(secdata);
1484 return ret;
1485 }
1486
1487 static int setup_security_options(struct btrfs_fs_info *fs_info,
1488 struct super_block *sb,
1489 struct security_mnt_opts *sec_opts)
1490 {
1491 int ret = 0;
1492
1493 /*
1494 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1495 * is valid.
1496 */
1497 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1498 if (ret)
1499 return ret;
1500
1501 #ifdef CONFIG_SECURITY
1502 if (!fs_info->security_opts.num_mnt_opts) {
1503 /* first time security setup, copy sec_opts to fs_info */
1504 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1505 } else {
1506 /*
1507 * Since SELinux (the only one supporting security_mnt_opts)
1508 * does NOT support changing context during remount/mount of
1509 * the same sb, this must be the same or part of the same
1510 * security options, just free it.
1511 */
1512 security_free_mnt_opts(sec_opts);
1513 }
1514 #endif
1515 return ret;
1516 }
1517
1518 /*
1519 * Find a superblock for the given device / mount point.
1520 *
1521 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1522 * for multiple device setup. Make sure to keep it in sync.
1523 */
1524 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1525 const char *device_name, void *data)
1526 {
1527 struct block_device *bdev = NULL;
1528 struct super_block *s;
1529 struct btrfs_fs_devices *fs_devices = NULL;
1530 struct btrfs_fs_info *fs_info = NULL;
1531 struct security_mnt_opts new_sec_opts;
1532 fmode_t mode = FMODE_READ;
1533 char *subvol_name = NULL;
1534 u64 subvol_objectid = 0;
1535 int error = 0;
1536
1537 if (!(flags & MS_RDONLY))
1538 mode |= FMODE_WRITE;
1539
1540 error = btrfs_parse_early_options(data, mode, fs_type,
1541 &subvol_name, &subvol_objectid,
1542 &fs_devices);
1543 if (error) {
1544 kfree(subvol_name);
1545 return ERR_PTR(error);
1546 }
1547
1548 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1549 /* mount_subvol() will free subvol_name. */
1550 return mount_subvol(subvol_name, subvol_objectid, flags,
1551 device_name, data);
1552 }
1553
1554 security_init_mnt_opts(&new_sec_opts);
1555 if (data) {
1556 error = parse_security_options(data, &new_sec_opts);
1557 if (error)
1558 return ERR_PTR(error);
1559 }
1560
1561 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1562 if (error)
1563 goto error_sec_opts;
1564
1565 /*
1566 * Setup a dummy root and fs_info for test/set super. This is because
1567 * we don't actually fill this stuff out until open_ctree, but we need
1568 * it for searching for existing supers, so this lets us do that and
1569 * then open_ctree will properly initialize everything later.
1570 */
1571 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1572 if (!fs_info) {
1573 error = -ENOMEM;
1574 goto error_sec_opts;
1575 }
1576
1577 fs_info->fs_devices = fs_devices;
1578
1579 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1580 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1581 security_init_mnt_opts(&fs_info->security_opts);
1582 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1583 error = -ENOMEM;
1584 goto error_fs_info;
1585 }
1586
1587 error = btrfs_open_devices(fs_devices, mode, fs_type);
1588 if (error)
1589 goto error_fs_info;
1590
1591 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1592 error = -EACCES;
1593 goto error_close_devices;
1594 }
1595
1596 bdev = fs_devices->latest_bdev;
1597 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1598 fs_info);
1599 if (IS_ERR(s)) {
1600 error = PTR_ERR(s);
1601 goto error_close_devices;
1602 }
1603
1604 if (s->s_root) {
1605 btrfs_close_devices(fs_devices);
1606 free_fs_info(fs_info);
1607 if ((flags ^ s->s_flags) & MS_RDONLY)
1608 error = -EBUSY;
1609 } else {
1610 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1611 btrfs_sb(s)->bdev_holder = fs_type;
1612 error = btrfs_fill_super(s, fs_devices, data);
1613 }
1614 if (error) {
1615 deactivate_locked_super(s);
1616 goto error_sec_opts;
1617 }
1618
1619 fs_info = btrfs_sb(s);
1620 error = setup_security_options(fs_info, s, &new_sec_opts);
1621 if (error) {
1622 deactivate_locked_super(s);
1623 goto error_sec_opts;
1624 }
1625
1626 return dget(s->s_root);
1627
1628 error_close_devices:
1629 btrfs_close_devices(fs_devices);
1630 error_fs_info:
1631 free_fs_info(fs_info);
1632 error_sec_opts:
1633 security_free_mnt_opts(&new_sec_opts);
1634 return ERR_PTR(error);
1635 }
1636
1637 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1638 int new_pool_size, int old_pool_size)
1639 {
1640 if (new_pool_size == old_pool_size)
1641 return;
1642
1643 fs_info->thread_pool_size = new_pool_size;
1644
1645 btrfs_info(fs_info, "resize thread pool %d -> %d",
1646 old_pool_size, new_pool_size);
1647
1648 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1649 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1650 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1651 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1652 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1653 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1654 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1655 new_pool_size);
1656 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1657 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1658 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1659 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1660 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1661 new_pool_size);
1662 }
1663
1664 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1665 {
1666 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1667 }
1668
1669 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1670 unsigned long old_opts, int flags)
1671 {
1672 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1673 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1674 (flags & MS_RDONLY))) {
1675 /* wait for any defraggers to finish */
1676 wait_event(fs_info->transaction_wait,
1677 (atomic_read(&fs_info->defrag_running) == 0));
1678 if (flags & MS_RDONLY)
1679 sync_filesystem(fs_info->sb);
1680 }
1681 }
1682
1683 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1684 unsigned long old_opts)
1685 {
1686 /*
1687 * We need to cleanup all defragable inodes if the autodefragment is
1688 * close or the filesystem is read only.
1689 */
1690 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1691 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1692 (fs_info->sb->s_flags & MS_RDONLY))) {
1693 btrfs_cleanup_defrag_inodes(fs_info);
1694 }
1695
1696 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1697 }
1698
1699 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1700 {
1701 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1702 struct btrfs_root *root = fs_info->tree_root;
1703 unsigned old_flags = sb->s_flags;
1704 unsigned long old_opts = fs_info->mount_opt;
1705 unsigned long old_compress_type = fs_info->compress_type;
1706 u64 old_max_inline = fs_info->max_inline;
1707 int old_thread_pool_size = fs_info->thread_pool_size;
1708 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1709 int ret;
1710
1711 sync_filesystem(sb);
1712 btrfs_remount_prepare(fs_info);
1713
1714 if (data) {
1715 struct security_mnt_opts new_sec_opts;
1716
1717 security_init_mnt_opts(&new_sec_opts);
1718 ret = parse_security_options(data, &new_sec_opts);
1719 if (ret)
1720 goto restore;
1721 ret = setup_security_options(fs_info, sb,
1722 &new_sec_opts);
1723 if (ret) {
1724 security_free_mnt_opts(&new_sec_opts);
1725 goto restore;
1726 }
1727 }
1728
1729 ret = btrfs_parse_options(fs_info, data, *flags);
1730 if (ret) {
1731 ret = -EINVAL;
1732 goto restore;
1733 }
1734
1735 btrfs_remount_begin(fs_info, old_opts, *flags);
1736 btrfs_resize_thread_pool(fs_info,
1737 fs_info->thread_pool_size, old_thread_pool_size);
1738
1739 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1740 goto out;
1741
1742 if (*flags & MS_RDONLY) {
1743 /*
1744 * this also happens on 'umount -rf' or on shutdown, when
1745 * the filesystem is busy.
1746 */
1747 cancel_work_sync(&fs_info->async_reclaim_work);
1748
1749 /* wait for the uuid_scan task to finish */
1750 down(&fs_info->uuid_tree_rescan_sem);
1751 /* avoid complains from lockdep et al. */
1752 up(&fs_info->uuid_tree_rescan_sem);
1753
1754 sb->s_flags |= MS_RDONLY;
1755
1756 /*
1757 * Setting MS_RDONLY will put the cleaner thread to
1758 * sleep at the next loop if it's already active.
1759 * If it's already asleep, we'll leave unused block
1760 * groups on disk until we're mounted read-write again
1761 * unless we clean them up here.
1762 */
1763 btrfs_delete_unused_bgs(fs_info);
1764
1765 btrfs_dev_replace_suspend_for_unmount(fs_info);
1766 btrfs_scrub_cancel(fs_info);
1767 btrfs_pause_balance(fs_info);
1768
1769 ret = btrfs_commit_super(fs_info);
1770 if (ret)
1771 goto restore;
1772 } else {
1773 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1774 btrfs_err(fs_info,
1775 "Remounting read-write after error is not allowed");
1776 ret = -EINVAL;
1777 goto restore;
1778 }
1779 if (fs_info->fs_devices->rw_devices == 0) {
1780 ret = -EACCES;
1781 goto restore;
1782 }
1783
1784 if (fs_info->fs_devices->missing_devices >
1785 fs_info->num_tolerated_disk_barrier_failures) {
1786 btrfs_warn(fs_info,
1787 "too many missing devices, writeable remount is not allowed");
1788 ret = -EACCES;
1789 goto restore;
1790 }
1791
1792 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1793 ret = -EINVAL;
1794 goto restore;
1795 }
1796
1797 ret = btrfs_cleanup_fs_roots(fs_info);
1798 if (ret)
1799 goto restore;
1800
1801 /* recover relocation */
1802 mutex_lock(&fs_info->cleaner_mutex);
1803 ret = btrfs_recover_relocation(root);
1804 mutex_unlock(&fs_info->cleaner_mutex);
1805 if (ret)
1806 goto restore;
1807
1808 ret = btrfs_resume_balance_async(fs_info);
1809 if (ret)
1810 goto restore;
1811
1812 ret = btrfs_resume_dev_replace_async(fs_info);
1813 if (ret) {
1814 btrfs_warn(fs_info, "failed to resume dev_replace");
1815 goto restore;
1816 }
1817
1818 if (!fs_info->uuid_root) {
1819 btrfs_info(fs_info, "creating UUID tree");
1820 ret = btrfs_create_uuid_tree(fs_info);
1821 if (ret) {
1822 btrfs_warn(fs_info,
1823 "failed to create the UUID tree %d",
1824 ret);
1825 goto restore;
1826 }
1827 }
1828 sb->s_flags &= ~MS_RDONLY;
1829
1830 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1831 }
1832 out:
1833 wake_up_process(fs_info->transaction_kthread);
1834 btrfs_remount_cleanup(fs_info, old_opts);
1835 return 0;
1836
1837 restore:
1838 /* We've hit an error - don't reset MS_RDONLY */
1839 if (sb->s_flags & MS_RDONLY)
1840 old_flags |= MS_RDONLY;
1841 sb->s_flags = old_flags;
1842 fs_info->mount_opt = old_opts;
1843 fs_info->compress_type = old_compress_type;
1844 fs_info->max_inline = old_max_inline;
1845 btrfs_resize_thread_pool(fs_info,
1846 old_thread_pool_size, fs_info->thread_pool_size);
1847 fs_info->metadata_ratio = old_metadata_ratio;
1848 btrfs_remount_cleanup(fs_info, old_opts);
1849 return ret;
1850 }
1851
1852 /* Used to sort the devices by max_avail(descending sort) */
1853 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1854 const void *dev_info2)
1855 {
1856 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1857 ((struct btrfs_device_info *)dev_info2)->max_avail)
1858 return -1;
1859 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1860 ((struct btrfs_device_info *)dev_info2)->max_avail)
1861 return 1;
1862 else
1863 return 0;
1864 }
1865
1866 /*
1867 * sort the devices by max_avail, in which max free extent size of each device
1868 * is stored.(Descending Sort)
1869 */
1870 static inline void btrfs_descending_sort_devices(
1871 struct btrfs_device_info *devices,
1872 size_t nr_devices)
1873 {
1874 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1875 btrfs_cmp_device_free_bytes, NULL);
1876 }
1877
1878 /*
1879 * The helper to calc the free space on the devices that can be used to store
1880 * file data.
1881 */
1882 static int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1883 u64 *free_bytes)
1884 {
1885 struct btrfs_device_info *devices_info;
1886 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1887 struct btrfs_device *device;
1888 u64 skip_space;
1889 u64 type;
1890 u64 avail_space;
1891 u64 min_stripe_size;
1892 int min_stripes = 1, num_stripes = 1;
1893 int i = 0, nr_devices;
1894
1895 /*
1896 * We aren't under the device list lock, so this is racy-ish, but good
1897 * enough for our purposes.
1898 */
1899 nr_devices = fs_info->fs_devices->open_devices;
1900 if (!nr_devices) {
1901 smp_mb();
1902 nr_devices = fs_info->fs_devices->open_devices;
1903 ASSERT(nr_devices);
1904 if (!nr_devices) {
1905 *free_bytes = 0;
1906 return 0;
1907 }
1908 }
1909
1910 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1911 GFP_KERNEL);
1912 if (!devices_info)
1913 return -ENOMEM;
1914
1915 /* calc min stripe number for data space allocation */
1916 type = btrfs_data_alloc_profile(fs_info);
1917 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1918 min_stripes = 2;
1919 num_stripes = nr_devices;
1920 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1921 min_stripes = 2;
1922 num_stripes = 2;
1923 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1924 min_stripes = 4;
1925 num_stripes = 4;
1926 }
1927
1928 if (type & BTRFS_BLOCK_GROUP_DUP)
1929 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1930 else
1931 min_stripe_size = BTRFS_STRIPE_LEN;
1932
1933 rcu_read_lock();
1934 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1935 if (!device->in_fs_metadata || !device->bdev ||
1936 device->is_tgtdev_for_dev_replace)
1937 continue;
1938
1939 if (i >= nr_devices)
1940 break;
1941
1942 avail_space = device->total_bytes - device->bytes_used;
1943
1944 /* align with stripe_len */
1945 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1946 avail_space *= BTRFS_STRIPE_LEN;
1947
1948 /*
1949 * In order to avoid overwriting the superblock on the drive,
1950 * btrfs starts at an offset of at least 1MB when doing chunk
1951 * allocation.
1952 */
1953 skip_space = SZ_1M;
1954
1955 /*
1956 * we can use the free space in [0, skip_space - 1], subtract
1957 * it from the total.
1958 */
1959 if (avail_space && avail_space >= skip_space)
1960 avail_space -= skip_space;
1961 else
1962 avail_space = 0;
1963
1964 if (avail_space < min_stripe_size)
1965 continue;
1966
1967 devices_info[i].dev = device;
1968 devices_info[i].max_avail = avail_space;
1969
1970 i++;
1971 }
1972 rcu_read_unlock();
1973
1974 nr_devices = i;
1975
1976 btrfs_descending_sort_devices(devices_info, nr_devices);
1977
1978 i = nr_devices - 1;
1979 avail_space = 0;
1980 while (nr_devices >= min_stripes) {
1981 if (num_stripes > nr_devices)
1982 num_stripes = nr_devices;
1983
1984 if (devices_info[i].max_avail >= min_stripe_size) {
1985 int j;
1986 u64 alloc_size;
1987
1988 avail_space += devices_info[i].max_avail * num_stripes;
1989 alloc_size = devices_info[i].max_avail;
1990 for (j = i + 1 - num_stripes; j <= i; j++)
1991 devices_info[j].max_avail -= alloc_size;
1992 }
1993 i--;
1994 nr_devices--;
1995 }
1996
1997 kfree(devices_info);
1998 *free_bytes = avail_space;
1999 return 0;
2000 }
2001
2002 /*
2003 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2004 *
2005 * If there's a redundant raid level at DATA block groups, use the respective
2006 * multiplier to scale the sizes.
2007 *
2008 * Unused device space usage is based on simulating the chunk allocator
2009 * algorithm that respects the device sizes and order of allocations. This is
2010 * a close approximation of the actual use but there are other factors that may
2011 * change the result (like a new metadata chunk).
2012 *
2013 * If metadata is exhausted, f_bavail will be 0.
2014 */
2015 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2016 {
2017 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2018 struct btrfs_super_block *disk_super = fs_info->super_copy;
2019 struct list_head *head = &fs_info->space_info;
2020 struct btrfs_space_info *found;
2021 u64 total_used = 0;
2022 u64 total_free_data = 0;
2023 u64 total_free_meta = 0;
2024 int bits = dentry->d_sb->s_blocksize_bits;
2025 __be32 *fsid = (__be32 *)fs_info->fsid;
2026 unsigned factor = 1;
2027 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2028 int ret;
2029 u64 thresh = 0;
2030 int mixed = 0;
2031
2032 rcu_read_lock();
2033 list_for_each_entry_rcu(found, head, list) {
2034 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2035 int i;
2036
2037 total_free_data += found->disk_total - found->disk_used;
2038 total_free_data -=
2039 btrfs_account_ro_block_groups_free_space(found);
2040
2041 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2042 if (!list_empty(&found->block_groups[i])) {
2043 switch (i) {
2044 case BTRFS_RAID_DUP:
2045 case BTRFS_RAID_RAID1:
2046 case BTRFS_RAID_RAID10:
2047 factor = 2;
2048 }
2049 }
2050 }
2051 }
2052
2053 /*
2054 * Metadata in mixed block goup profiles are accounted in data
2055 */
2056 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2057 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2058 mixed = 1;
2059 else
2060 total_free_meta += found->disk_total -
2061 found->disk_used;
2062 }
2063
2064 total_used += found->disk_used;
2065 }
2066
2067 rcu_read_unlock();
2068
2069 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2070 buf->f_blocks >>= bits;
2071 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2072
2073 /* Account global block reserve as used, it's in logical size already */
2074 spin_lock(&block_rsv->lock);
2075 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2076 if (buf->f_bfree >= block_rsv->size >> bits)
2077 buf->f_bfree -= block_rsv->size >> bits;
2078 else
2079 buf->f_bfree = 0;
2080 spin_unlock(&block_rsv->lock);
2081
2082 buf->f_bavail = div_u64(total_free_data, factor);
2083 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2084 if (ret)
2085 return ret;
2086 buf->f_bavail += div_u64(total_free_data, factor);
2087 buf->f_bavail = buf->f_bavail >> bits;
2088
2089 /*
2090 * We calculate the remaining metadata space minus global reserve. If
2091 * this is (supposedly) smaller than zero, there's no space. But this
2092 * does not hold in practice, the exhausted state happens where's still
2093 * some positive delta. So we apply some guesswork and compare the
2094 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2095 *
2096 * We probably cannot calculate the exact threshold value because this
2097 * depends on the internal reservations requested by various
2098 * operations, so some operations that consume a few metadata will
2099 * succeed even if the Avail is zero. But this is better than the other
2100 * way around.
2101 */
2102 thresh = 4 * 1024 * 1024;
2103
2104 if (!mixed && total_free_meta - thresh < block_rsv->size)
2105 buf->f_bavail = 0;
2106
2107 buf->f_type = BTRFS_SUPER_MAGIC;
2108 buf->f_bsize = dentry->d_sb->s_blocksize;
2109 buf->f_namelen = BTRFS_NAME_LEN;
2110
2111 /* We treat it as constant endianness (it doesn't matter _which_)
2112 because we want the fsid to come out the same whether mounted
2113 on a big-endian or little-endian host */
2114 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2115 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2116 /* Mask in the root object ID too, to disambiguate subvols */
2117 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2118 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2119
2120 return 0;
2121 }
2122
2123 static void btrfs_kill_super(struct super_block *sb)
2124 {
2125 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2126 kill_anon_super(sb);
2127 free_fs_info(fs_info);
2128 }
2129
2130 static struct file_system_type btrfs_fs_type = {
2131 .owner = THIS_MODULE,
2132 .name = "btrfs",
2133 .mount = btrfs_mount,
2134 .kill_sb = btrfs_kill_super,
2135 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2136 };
2137 MODULE_ALIAS_FS("btrfs");
2138
2139 static int btrfs_control_open(struct inode *inode, struct file *file)
2140 {
2141 /*
2142 * The control file's private_data is used to hold the
2143 * transaction when it is started and is used to keep
2144 * track of whether a transaction is already in progress.
2145 */
2146 file->private_data = NULL;
2147 return 0;
2148 }
2149
2150 /*
2151 * used by btrfsctl to scan devices when no FS is mounted
2152 */
2153 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2154 unsigned long arg)
2155 {
2156 struct btrfs_ioctl_vol_args *vol;
2157 struct btrfs_fs_devices *fs_devices;
2158 int ret = -ENOTTY;
2159
2160 if (!capable(CAP_SYS_ADMIN))
2161 return -EPERM;
2162
2163 vol = memdup_user((void __user *)arg, sizeof(*vol));
2164 if (IS_ERR(vol))
2165 return PTR_ERR(vol);
2166
2167 switch (cmd) {
2168 case BTRFS_IOC_SCAN_DEV:
2169 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2170 &btrfs_fs_type, &fs_devices);
2171 break;
2172 case BTRFS_IOC_DEVICES_READY:
2173 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2174 &btrfs_fs_type, &fs_devices);
2175 if (ret)
2176 break;
2177 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2178 break;
2179 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2180 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2181 break;
2182 }
2183
2184 kfree(vol);
2185 return ret;
2186 }
2187
2188 static int btrfs_freeze(struct super_block *sb)
2189 {
2190 struct btrfs_trans_handle *trans;
2191 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2192 struct btrfs_root *root = fs_info->tree_root;
2193
2194 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2195 /*
2196 * We don't need a barrier here, we'll wait for any transaction that
2197 * could be in progress on other threads (and do delayed iputs that
2198 * we want to avoid on a frozen filesystem), or do the commit
2199 * ourselves.
2200 */
2201 trans = btrfs_attach_transaction_barrier(root);
2202 if (IS_ERR(trans)) {
2203 /* no transaction, don't bother */
2204 if (PTR_ERR(trans) == -ENOENT)
2205 return 0;
2206 return PTR_ERR(trans);
2207 }
2208 return btrfs_commit_transaction(trans);
2209 }
2210
2211 static int btrfs_unfreeze(struct super_block *sb)
2212 {
2213 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2214
2215 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2216 return 0;
2217 }
2218
2219 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2220 {
2221 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2222 struct btrfs_fs_devices *cur_devices;
2223 struct btrfs_device *dev, *first_dev = NULL;
2224 struct list_head *head;
2225 struct rcu_string *name;
2226
2227 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2228 cur_devices = fs_info->fs_devices;
2229 while (cur_devices) {
2230 head = &cur_devices->devices;
2231 list_for_each_entry(dev, head, dev_list) {
2232 if (dev->missing)
2233 continue;
2234 if (!dev->name)
2235 continue;
2236 if (!first_dev || dev->devid < first_dev->devid)
2237 first_dev = dev;
2238 }
2239 cur_devices = cur_devices->seed;
2240 }
2241
2242 if (first_dev) {
2243 rcu_read_lock();
2244 name = rcu_dereference(first_dev->name);
2245 seq_escape(m, name->str, " \t\n\\");
2246 rcu_read_unlock();
2247 } else {
2248 WARN_ON(1);
2249 }
2250 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2251 return 0;
2252 }
2253
2254 static const struct super_operations btrfs_super_ops = {
2255 .drop_inode = btrfs_drop_inode,
2256 .evict_inode = btrfs_evict_inode,
2257 .put_super = btrfs_put_super,
2258 .sync_fs = btrfs_sync_fs,
2259 .show_options = btrfs_show_options,
2260 .show_devname = btrfs_show_devname,
2261 .write_inode = btrfs_write_inode,
2262 .alloc_inode = btrfs_alloc_inode,
2263 .destroy_inode = btrfs_destroy_inode,
2264 .statfs = btrfs_statfs,
2265 .remount_fs = btrfs_remount,
2266 .freeze_fs = btrfs_freeze,
2267 .unfreeze_fs = btrfs_unfreeze,
2268 };
2269
2270 static const struct file_operations btrfs_ctl_fops = {
2271 .open = btrfs_control_open,
2272 .unlocked_ioctl = btrfs_control_ioctl,
2273 .compat_ioctl = btrfs_control_ioctl,
2274 .owner = THIS_MODULE,
2275 .llseek = noop_llseek,
2276 };
2277
2278 static struct miscdevice btrfs_misc = {
2279 .minor = BTRFS_MINOR,
2280 .name = "btrfs-control",
2281 .fops = &btrfs_ctl_fops
2282 };
2283
2284 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2285 MODULE_ALIAS("devname:btrfs-control");
2286
2287 static int btrfs_interface_init(void)
2288 {
2289 return misc_register(&btrfs_misc);
2290 }
2291
2292 static void btrfs_interface_exit(void)
2293 {
2294 misc_deregister(&btrfs_misc);
2295 }
2296
2297 static void btrfs_print_mod_info(void)
2298 {
2299 pr_info("Btrfs loaded, crc32c=%s"
2300 #ifdef CONFIG_BTRFS_DEBUG
2301 ", debug=on"
2302 #endif
2303 #ifdef CONFIG_BTRFS_ASSERT
2304 ", assert=on"
2305 #endif
2306 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2307 ", integrity-checker=on"
2308 #endif
2309 "\n",
2310 btrfs_crc32c_impl());
2311 }
2312
2313 static int __init init_btrfs_fs(void)
2314 {
2315 int err;
2316
2317 err = btrfs_hash_init();
2318 if (err)
2319 return err;
2320
2321 btrfs_props_init();
2322
2323 err = btrfs_init_sysfs();
2324 if (err)
2325 goto free_hash;
2326
2327 btrfs_init_compress();
2328
2329 err = btrfs_init_cachep();
2330 if (err)
2331 goto free_compress;
2332
2333 err = extent_io_init();
2334 if (err)
2335 goto free_cachep;
2336
2337 err = extent_map_init();
2338 if (err)
2339 goto free_extent_io;
2340
2341 err = ordered_data_init();
2342 if (err)
2343 goto free_extent_map;
2344
2345 err = btrfs_delayed_inode_init();
2346 if (err)
2347 goto free_ordered_data;
2348
2349 err = btrfs_auto_defrag_init();
2350 if (err)
2351 goto free_delayed_inode;
2352
2353 err = btrfs_delayed_ref_init();
2354 if (err)
2355 goto free_auto_defrag;
2356
2357 err = btrfs_prelim_ref_init();
2358 if (err)
2359 goto free_delayed_ref;
2360
2361 err = btrfs_end_io_wq_init();
2362 if (err)
2363 goto free_prelim_ref;
2364
2365 err = btrfs_interface_init();
2366 if (err)
2367 goto free_end_io_wq;
2368
2369 btrfs_init_lockdep();
2370
2371 btrfs_print_mod_info();
2372
2373 err = btrfs_run_sanity_tests();
2374 if (err)
2375 goto unregister_ioctl;
2376
2377 err = register_filesystem(&btrfs_fs_type);
2378 if (err)
2379 goto unregister_ioctl;
2380
2381 return 0;
2382
2383 unregister_ioctl:
2384 btrfs_interface_exit();
2385 free_end_io_wq:
2386 btrfs_end_io_wq_exit();
2387 free_prelim_ref:
2388 btrfs_prelim_ref_exit();
2389 free_delayed_ref:
2390 btrfs_delayed_ref_exit();
2391 free_auto_defrag:
2392 btrfs_auto_defrag_exit();
2393 free_delayed_inode:
2394 btrfs_delayed_inode_exit();
2395 free_ordered_data:
2396 ordered_data_exit();
2397 free_extent_map:
2398 extent_map_exit();
2399 free_extent_io:
2400 extent_io_exit();
2401 free_cachep:
2402 btrfs_destroy_cachep();
2403 free_compress:
2404 btrfs_exit_compress();
2405 btrfs_exit_sysfs();
2406 free_hash:
2407 btrfs_hash_exit();
2408 return err;
2409 }
2410
2411 static void __exit exit_btrfs_fs(void)
2412 {
2413 btrfs_destroy_cachep();
2414 btrfs_delayed_ref_exit();
2415 btrfs_auto_defrag_exit();
2416 btrfs_delayed_inode_exit();
2417 btrfs_prelim_ref_exit();
2418 ordered_data_exit();
2419 extent_map_exit();
2420 extent_io_exit();
2421 btrfs_interface_exit();
2422 btrfs_end_io_wq_exit();
2423 unregister_filesystem(&btrfs_fs_type);
2424 btrfs_exit_sysfs();
2425 btrfs_cleanup_fs_uuids();
2426 btrfs_exit_compress();
2427 btrfs_hash_exit();
2428 }
2429
2430 late_initcall(init_btrfs_fs);
2431 module_exit(exit_btrfs_fs)
2432
2433 MODULE_LICENSE("GPL");
Linux with added FPC-III support