?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
[cris-mirror.git] / fs / btrfs / super.c
blob4b817947e00f39e327a244f67bbde7ebf2d6b7e0
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
69 /*
70 * Types for mounting the default subvolume and a subvolume explicitly
71 * requested by subvol=/path. That way the callchain is straightforward and we
72 * don't have to play tricks with the mount options and recursive calls to
73 * btrfs_mount.
74 *
75 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
76 */
77 static struct file_system_type btrfs_fs_type;
78 static struct file_system_type btrfs_root_fs_type;
79
80 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
81
82 const char *btrfs_decode_error(int errno)
83 {
84 char *errstr = "unknown";
85
86 switch (errno) {
87 case -EIO:
88 errstr = "IO failure";
89 break;
90 case -ENOMEM:
91 errstr = "Out of memory";
92 break;
93 case -EROFS:
94 errstr = "Readonly filesystem";
95 break;
96 case -EEXIST:
97 errstr = "Object already exists";
98 break;
99 case -ENOSPC:
100 errstr = "No space left";
101 break;
102 case -ENOENT:
103 errstr = "No such entry";
104 break;
105 }
106
107 return errstr;
108 }
109
110 /*
111 * __btrfs_handle_fs_error decodes expected errors from the caller and
112 * invokes the approciate error response.
113 */
114 __cold
115 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
116 unsigned int line, int errno, const char *fmt, ...)
117 {
118 struct super_block *sb = fs_info->sb;
119 #ifdef CONFIG_PRINTK
120 const char *errstr;
121 #endif
122
123 /*
124 * Special case: if the error is EROFS, and we're already
125 * under SB_RDONLY, then it is safe here.
126 */
127 if (errno == -EROFS && sb_rdonly(sb))
128 return;
129
130 #ifdef CONFIG_PRINTK
131 errstr = btrfs_decode_error(errno);
132 if (fmt) {
133 struct va_format vaf;
134 va_list args;
135
136 va_start(args, fmt);
137 vaf.fmt = fmt;
138 vaf.va = &args;
139
140 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
141 sb->s_id, function, line, errno, errstr, &vaf);
142 va_end(args);
143 } else {
144 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
145 sb->s_id, function, line, errno, errstr);
146 }
147 #endif
148
149 /*
150 * Today we only save the error info to memory. Long term we'll
151 * also send it down to the disk
152 */
153 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
154
155 /* Don't go through full error handling during mount */
156 if (!(sb->s_flags & SB_BORN))
157 return;
158
159 if (sb_rdonly(sb))
160 return;
161
162 /* btrfs handle error by forcing the filesystem readonly */
163 sb->s_flags |= SB_RDONLY;
164 btrfs_info(fs_info, "forced readonly");
165 /*
166 * Note that a running device replace operation is not canceled here
167 * although there is no way to update the progress. It would add the
168 * risk of a deadlock, therefore the canceling is omitted. The only
169 * penalty is that some I/O remains active until the procedure
170 * completes. The next time when the filesystem is mounted writeable
171 * again, the device replace operation continues.
172 */
173 }
174
175 #ifdef CONFIG_PRINTK
176 static const char * const logtypes[] = {
177 "emergency",
178 "alert",
179 "critical",
180 "error",
181 "warning",
182 "notice",
183 "info",
184 "debug",
185 };
186
187
188 /*
189 * Use one ratelimit state per log level so that a flood of less important
190 * messages doesn't cause more important ones to be dropped.
191 */
192 static struct ratelimit_state printk_limits[] = {
193 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
194 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
195 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
196 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
197 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
198 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
199 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
200 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
201 };
202
203 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
204 {
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,
231 fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
232
233 va_end(args);
234 }
235 #endif
236
237 /*
238 * We only mark the transaction aborted and then set the file system read-only.
239 * This will prevent new transactions from starting or trying to join this
240 * one.
241 *
242 * This means that error recovery at the call site is limited to freeing
243 * any local memory allocations and passing the error code up without
244 * further cleanup. The transaction should complete as it normally would
245 * in the call path but will return -EIO.
246 *
247 * We'll complete the cleanup in btrfs_end_transaction and
248 * btrfs_commit_transaction.
249 */
250 __cold
251 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
252 const char *function,
253 unsigned int line, int errno)
254 {
255 struct btrfs_fs_info *fs_info = trans->fs_info;
256
257 trans->aborted = errno;
258 /* Nothing used. The other threads that have joined this
259 * transaction may be able to continue. */
260 if (!trans->dirty && list_empty(&trans->new_bgs)) {
261 const char *errstr;
262
263 errstr = btrfs_decode_error(errno);
264 btrfs_warn(fs_info,
265 "%s:%d: Aborting unused transaction(%s).",
266 function, line, errstr);
267 return;
268 }
269 WRITE_ONCE(trans->transaction->aborted, errno);
270 /* Wake up anybody who may be waiting on this transaction */
271 wake_up(&fs_info->transaction_wait);
272 wake_up(&fs_info->transaction_blocked_wait);
273 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
274 }
275 /*
276 * __btrfs_panic decodes unexpected, fatal errors from the caller,
277 * issues an alert, and either panics or BUGs, depending on mount options.
278 */
279 __cold
280 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
281 unsigned int line, int errno, const char *fmt, ...)
282 {
283 char *s_id = "<unknown>";
284 const char *errstr;
285 struct va_format vaf = { .fmt = fmt };
286 va_list args;
287
288 if (fs_info)
289 s_id = fs_info->sb->s_id;
290
291 va_start(args, fmt);
292 vaf.va = &args;
293
294 errstr = btrfs_decode_error(errno);
295 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
296 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
297 s_id, function, line, &vaf, errno, errstr);
298
299 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
300 function, line, &vaf, errno, errstr);
301 va_end(args);
302 /* Caller calls BUG() */
303 }
304
305 static void btrfs_put_super(struct super_block *sb)
306 {
307 close_ctree(btrfs_sb(sb));
308 }
309
310 enum {
311 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
312 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
313 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
314 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
315 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
316 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
317 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
318 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
319 Opt_skip_balance, Opt_check_integrity,
320 Opt_check_integrity_including_extent_data,
321 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
322 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
323 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
324 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
325 Opt_nologreplay, Opt_norecovery,
326 #ifdef CONFIG_BTRFS_DEBUG
327 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
328 #endif
329 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
330 Opt_ref_verify,
331 #endif
332 Opt_err,
333 };
334
335 static const match_table_t tokens = {
336 {Opt_degraded, "degraded"},
337 {Opt_subvol, "subvol=%s"},
338 {Opt_subvolid, "subvolid=%s"},
339 {Opt_device, "device=%s"},
340 {Opt_nodatasum, "nodatasum"},
341 {Opt_datasum, "datasum"},
342 {Opt_nodatacow, "nodatacow"},
343 {Opt_datacow, "datacow"},
344 {Opt_nobarrier, "nobarrier"},
345 {Opt_barrier, "barrier"},
346 {Opt_max_inline, "max_inline=%s"},
347 {Opt_alloc_start, "alloc_start=%s"},
348 {Opt_thread_pool, "thread_pool=%d"},
349 {Opt_compress, "compress"},
350 {Opt_compress_type, "compress=%s"},
351 {Opt_compress_force, "compress-force"},
352 {Opt_compress_force_type, "compress-force=%s"},
353 {Opt_ssd, "ssd"},
354 {Opt_ssd_spread, "ssd_spread"},
355 {Opt_nossd, "nossd"},
356 {Opt_acl, "acl"},
357 {Opt_noacl, "noacl"},
358 {Opt_notreelog, "notreelog"},
359 {Opt_treelog, "treelog"},
360 {Opt_nologreplay, "nologreplay"},
361 {Opt_norecovery, "norecovery"},
362 {Opt_flushoncommit, "flushoncommit"},
363 {Opt_noflushoncommit, "noflushoncommit"},
364 {Opt_ratio, "metadata_ratio=%d"},
365 {Opt_discard, "discard"},
366 {Opt_nodiscard, "nodiscard"},
367 {Opt_space_cache, "space_cache"},
368 {Opt_space_cache_version, "space_cache=%s"},
369 {Opt_clear_cache, "clear_cache"},
370 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
371 {Opt_enospc_debug, "enospc_debug"},
372 {Opt_noenospc_debug, "noenospc_debug"},
373 {Opt_subvolrootid, "subvolrootid=%d"},
374 {Opt_defrag, "autodefrag"},
375 {Opt_nodefrag, "noautodefrag"},
376 {Opt_inode_cache, "inode_cache"},
377 {Opt_noinode_cache, "noinode_cache"},
378 {Opt_no_space_cache, "nospace_cache"},
379 {Opt_recovery, "recovery"}, /* deprecated */
380 {Opt_usebackuproot, "usebackuproot"},
381 {Opt_skip_balance, "skip_balance"},
382 {Opt_check_integrity, "check_int"},
383 {Opt_check_integrity_including_extent_data, "check_int_data"},
384 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
385 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
386 {Opt_fatal_errors, "fatal_errors=%s"},
387 {Opt_commit_interval, "commit=%d"},
388 #ifdef CONFIG_BTRFS_DEBUG
389 {Opt_fragment_data, "fragment=data"},
390 {Opt_fragment_metadata, "fragment=metadata"},
391 {Opt_fragment_all, "fragment=all"},
392 #endif
393 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
394 {Opt_ref_verify, "ref_verify"},
395 #endif
396 {Opt_err, NULL},
397 };
398
399 /*
400 * Regular mount options parser. Everything that is needed only when
401 * reading in a new superblock is parsed here.
402 * XXX JDM: This needs to be cleaned up for remount.
403 */
404 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
405 unsigned long new_flags)
406 {
407 substring_t args[MAX_OPT_ARGS];
408 char *p, *num;
409 u64 cache_gen;
410 int intarg;
411 int ret = 0;
412 char *compress_type;
413 bool compress_force = false;
414 enum btrfs_compression_type saved_compress_type;
415 bool saved_compress_force;
416 int no_compress = 0;
417
418 cache_gen = btrfs_super_cache_generation(info->super_copy);
419 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
420 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
421 else if (cache_gen)
422 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
423
424 /*
425 * Even the options are empty, we still need to do extra check
426 * against new flags
427 */
428 if (!options)
429 goto check;
430
431 while ((p = strsep(&options, ",")) != NULL) {
432 int token;
433 if (!*p)
434 continue;
435
436 token = match_token(p, tokens, args);
437 switch (token) {
438 case Opt_degraded:
439 btrfs_info(info, "allowing degraded mounts");
440 btrfs_set_opt(info->mount_opt, DEGRADED);
441 break;
442 case Opt_subvol:
443 case Opt_subvolid:
444 case Opt_subvolrootid:
445 case Opt_device:
446 /*
447 * These are parsed by btrfs_parse_subvol_options
448 * and btrfs_parse_early_options
449 * and can be happily ignored here.
450 */
451 break;
452 case Opt_nodatasum:
453 btrfs_set_and_info(info, NODATASUM,
454 "setting nodatasum");
455 break;
456 case Opt_datasum:
457 if (btrfs_test_opt(info, NODATASUM)) {
458 if (btrfs_test_opt(info, NODATACOW))
459 btrfs_info(info,
460 "setting datasum, datacow enabled");
461 else
462 btrfs_info(info, "setting datasum");
463 }
464 btrfs_clear_opt(info->mount_opt, NODATACOW);
465 btrfs_clear_opt(info->mount_opt, NODATASUM);
466 break;
467 case Opt_nodatacow:
468 if (!btrfs_test_opt(info, NODATACOW)) {
469 if (!btrfs_test_opt(info, COMPRESS) ||
470 !btrfs_test_opt(info, FORCE_COMPRESS)) {
471 btrfs_info(info,
472 "setting nodatacow, compression disabled");
473 } else {
474 btrfs_info(info, "setting nodatacow");
475 }
476 }
477 btrfs_clear_opt(info->mount_opt, COMPRESS);
478 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
479 btrfs_set_opt(info->mount_opt, NODATACOW);
480 btrfs_set_opt(info->mount_opt, NODATASUM);
481 break;
482 case Opt_datacow:
483 btrfs_clear_and_info(info, NODATACOW,
484 "setting datacow");
485 break;
486 case Opt_compress_force:
487 case Opt_compress_force_type:
488 compress_force = true;
489 /* Fallthrough */
490 case Opt_compress:
491 case Opt_compress_type:
492 saved_compress_type = btrfs_test_opt(info,
493 COMPRESS) ?
494 info->compress_type : BTRFS_COMPRESS_NONE;
495 saved_compress_force =
496 btrfs_test_opt(info, FORCE_COMPRESS);
497 if (token == Opt_compress ||
498 token == Opt_compress_force ||
499 strncmp(args[0].from, "zlib", 4) == 0) {
500 compress_type = "zlib";
501
502 info->compress_type = BTRFS_COMPRESS_ZLIB;
503 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
504 /*
505 * args[0] contains uninitialized data since
506 * for these tokens we don't expect any
507 * parameter.
508 */
509 if (token != Opt_compress &&
510 token != Opt_compress_force)
511 info->compress_level =
512 btrfs_compress_str2level(args[0].from);
513 btrfs_set_opt(info->mount_opt, COMPRESS);
514 btrfs_clear_opt(info->mount_opt, NODATACOW);
515 btrfs_clear_opt(info->mount_opt, NODATASUM);
516 no_compress = 0;
517 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
518 compress_type = "lzo";
519 info->compress_type = BTRFS_COMPRESS_LZO;
520 btrfs_set_opt(info->mount_opt, COMPRESS);
521 btrfs_clear_opt(info->mount_opt, NODATACOW);
522 btrfs_clear_opt(info->mount_opt, NODATASUM);
523 btrfs_set_fs_incompat(info, COMPRESS_LZO);
524 no_compress = 0;
525 } else if (strcmp(args[0].from, "zstd") == 0) {
526 compress_type = "zstd";
527 info->compress_type = BTRFS_COMPRESS_ZSTD;
528 btrfs_set_opt(info->mount_opt, COMPRESS);
529 btrfs_clear_opt(info->mount_opt, NODATACOW);
530 btrfs_clear_opt(info->mount_opt, NODATASUM);
531 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
532 no_compress = 0;
533 } else if (strncmp(args[0].from, "no", 2) == 0) {
534 compress_type = "no";
535 btrfs_clear_opt(info->mount_opt, COMPRESS);
536 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
537 compress_force = false;
538 no_compress++;
539 } else {
540 ret = -EINVAL;
541 goto out;
542 }
543
544 if (compress_force) {
545 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
546 } else {
547 /*
548 * If we remount from compress-force=xxx to
549 * compress=xxx, we need clear FORCE_COMPRESS
550 * flag, otherwise, there is no way for users
551 * to disable forcible compression separately.
552 */
553 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
554 }
555 if ((btrfs_test_opt(info, COMPRESS) &&
556 (info->compress_type != saved_compress_type ||
557 compress_force != saved_compress_force)) ||
558 (!btrfs_test_opt(info, COMPRESS) &&
559 no_compress == 1)) {
560 btrfs_info(info, "%s %s compression, level %d",
561 (compress_force) ? "force" : "use",
562 compress_type, info->compress_level);
563 }
564 compress_force = false;
565 break;
566 case Opt_ssd:
567 btrfs_set_and_info(info, SSD,
568 "enabling ssd optimizations");
569 btrfs_clear_opt(info->mount_opt, NOSSD);
570 break;
571 case Opt_ssd_spread:
572 btrfs_set_and_info(info, SSD,
573 "enabling ssd optimizations");
574 btrfs_set_and_info(info, SSD_SPREAD,
575 "using spread ssd allocation scheme");
576 btrfs_clear_opt(info->mount_opt, NOSSD);
577 break;
578 case Opt_nossd:
579 btrfs_set_opt(info->mount_opt, NOSSD);
580 btrfs_clear_and_info(info, SSD,
581 "not using ssd optimizations");
582 btrfs_clear_and_info(info, SSD_SPREAD,
583 "not using spread ssd allocation scheme");
584 break;
585 case Opt_barrier:
586 btrfs_clear_and_info(info, NOBARRIER,
587 "turning on barriers");
588 break;
589 case Opt_nobarrier:
590 btrfs_set_and_info(info, NOBARRIER,
591 "turning off barriers");
592 break;
593 case Opt_thread_pool:
594 ret = match_int(&args[0], &intarg);
595 if (ret) {
596 goto out;
597 } else if (intarg > 0) {
598 info->thread_pool_size = intarg;
599 } else {
600 ret = -EINVAL;
601 goto out;
602 }
603 break;
604 case Opt_max_inline:
605 num = match_strdup(&args[0]);
606 if (num) {
607 info->max_inline = memparse(num, NULL);
608 kfree(num);
609
610 if (info->max_inline) {
611 info->max_inline = min_t(u64,
612 info->max_inline,
613 info->sectorsize);
614 }
615 btrfs_info(info, "max_inline at %llu",
616 info->max_inline);
617 } else {
618 ret = -ENOMEM;
619 goto out;
620 }
621 break;
622 case Opt_alloc_start:
623 btrfs_info(info,
624 "option alloc_start is obsolete, ignored");
625 break;
626 case Opt_acl:
627 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
628 info->sb->s_flags |= SB_POSIXACL;
629 break;
630 #else
631 btrfs_err(info, "support for ACL not compiled in!");
632 ret = -EINVAL;
633 goto out;
634 #endif
635 case Opt_noacl:
636 info->sb->s_flags &= ~SB_POSIXACL;
637 break;
638 case Opt_notreelog:
639 btrfs_set_and_info(info, NOTREELOG,
640 "disabling tree log");
641 break;
642 case Opt_treelog:
643 btrfs_clear_and_info(info, NOTREELOG,
644 "enabling tree log");
645 break;
646 case Opt_norecovery:
647 case Opt_nologreplay:
648 btrfs_set_and_info(info, NOLOGREPLAY,
649 "disabling log replay at mount time");
650 break;
651 case Opt_flushoncommit:
652 btrfs_set_and_info(info, FLUSHONCOMMIT,
653 "turning on flush-on-commit");
654 break;
655 case Opt_noflushoncommit:
656 btrfs_clear_and_info(info, FLUSHONCOMMIT,
657 "turning off flush-on-commit");
658 break;
659 case Opt_ratio:
660 ret = match_int(&args[0], &intarg);
661 if (ret) {
662 goto out;
663 } else if (intarg >= 0) {
664 info->metadata_ratio = intarg;
665 btrfs_info(info, "metadata ratio %d",
666 info->metadata_ratio);
667 } else {
668 ret = -EINVAL;
669 goto out;
670 }
671 break;
672 case Opt_discard:
673 btrfs_set_and_info(info, DISCARD,
674 "turning on discard");
675 break;
676 case Opt_nodiscard:
677 btrfs_clear_and_info(info, DISCARD,
678 "turning off discard");
679 break;
680 case Opt_space_cache:
681 case Opt_space_cache_version:
682 if (token == Opt_space_cache ||
683 strcmp(args[0].from, "v1") == 0) {
684 btrfs_clear_opt(info->mount_opt,
685 FREE_SPACE_TREE);
686 btrfs_set_and_info(info, SPACE_CACHE,
687 "enabling disk space caching");
688 } else if (strcmp(args[0].from, "v2") == 0) {
689 btrfs_clear_opt(info->mount_opt,
690 SPACE_CACHE);
691 btrfs_set_and_info(info, FREE_SPACE_TREE,
692 "enabling free space tree");
693 } else {
694 ret = -EINVAL;
695 goto out;
696 }
697 break;
698 case Opt_rescan_uuid_tree:
699 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
700 break;
701 case Opt_no_space_cache:
702 if (btrfs_test_opt(info, SPACE_CACHE)) {
703 btrfs_clear_and_info(info, SPACE_CACHE,
704 "disabling disk space caching");
705 }
706 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
707 btrfs_clear_and_info(info, FREE_SPACE_TREE,
708 "disabling free space tree");
709 }
710 break;
711 case Opt_inode_cache:
712 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
713 "enabling inode map caching");
714 break;
715 case Opt_noinode_cache:
716 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
717 "disabling inode map caching");
718 break;
719 case Opt_clear_cache:
720 btrfs_set_and_info(info, CLEAR_CACHE,
721 "force clearing of disk cache");
722 break;
723 case Opt_user_subvol_rm_allowed:
724 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
725 break;
726 case Opt_enospc_debug:
727 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
728 break;
729 case Opt_noenospc_debug:
730 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
731 break;
732 case Opt_defrag:
733 btrfs_set_and_info(info, AUTO_DEFRAG,
734 "enabling auto defrag");
735 break;
736 case Opt_nodefrag:
737 btrfs_clear_and_info(info, AUTO_DEFRAG,
738 "disabling auto defrag");
739 break;
740 case Opt_recovery:
741 btrfs_warn(info,
742 "'recovery' is deprecated, use 'usebackuproot' instead");
743 case Opt_usebackuproot:
744 btrfs_info(info,
745 "trying to use backup root at mount time");
746 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
747 break;
748 case Opt_skip_balance:
749 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
750 break;
751 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
752 case Opt_check_integrity_including_extent_data:
753 btrfs_info(info,
754 "enabling check integrity including extent data");
755 btrfs_set_opt(info->mount_opt,
756 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
757 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
758 break;
759 case Opt_check_integrity:
760 btrfs_info(info, "enabling check integrity");
761 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
762 break;
763 case Opt_check_integrity_print_mask:
764 ret = match_int(&args[0], &intarg);
765 if (ret) {
766 goto out;
767 } else if (intarg >= 0) {
768 info->check_integrity_print_mask = intarg;
769 btrfs_info(info,
770 "check_integrity_print_mask 0x%x",
771 info->check_integrity_print_mask);
772 } else {
773 ret = -EINVAL;
774 goto out;
775 }
776 break;
777 #else
778 case Opt_check_integrity_including_extent_data:
779 case Opt_check_integrity:
780 case Opt_check_integrity_print_mask:
781 btrfs_err(info,
782 "support for check_integrity* not compiled in!");
783 ret = -EINVAL;
784 goto out;
785 #endif
786 case Opt_fatal_errors:
787 if (strcmp(args[0].from, "panic") == 0)
788 btrfs_set_opt(info->mount_opt,
789 PANIC_ON_FATAL_ERROR);
790 else if (strcmp(args[0].from, "bug") == 0)
791 btrfs_clear_opt(info->mount_opt,
792 PANIC_ON_FATAL_ERROR);
793 else {
794 ret = -EINVAL;
795 goto out;
796 }
797 break;
798 case Opt_commit_interval:
799 intarg = 0;
800 ret = match_int(&args[0], &intarg);
801 if (ret < 0) {
802 btrfs_err(info, "invalid commit interval");
803 ret = -EINVAL;
804 goto out;
805 }
806 if (intarg > 0) {
807 if (intarg > 300) {
808 btrfs_warn(info,
809 "excessive commit interval %d",
810 intarg);
811 }
812 info->commit_interval = intarg;
813 } else {
814 btrfs_info(info,
815 "using default commit interval %ds",
816 BTRFS_DEFAULT_COMMIT_INTERVAL);
817 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
818 }
819 break;
820 #ifdef CONFIG_BTRFS_DEBUG
821 case Opt_fragment_all:
822 btrfs_info(info, "fragmenting all space");
823 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
824 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
825 break;
826 case Opt_fragment_metadata:
827 btrfs_info(info, "fragmenting metadata");
828 btrfs_set_opt(info->mount_opt,
829 FRAGMENT_METADATA);
830 break;
831 case Opt_fragment_data:
832 btrfs_info(info, "fragmenting data");
833 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
834 break;
835 #endif
836 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
837 case Opt_ref_verify:
838 btrfs_info(info, "doing ref verification");
839 btrfs_set_opt(info->mount_opt, REF_VERIFY);
840 break;
841 #endif
842 case Opt_err:
843 btrfs_info(info, "unrecognized mount option '%s'", p);
844 ret = -EINVAL;
845 goto out;
846 default:
847 break;
848 }
849 }
850 check:
851 /*
852 * Extra check for current option against current flag
853 */
854 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
855 btrfs_err(info,
856 "nologreplay must be used with ro mount option");
857 ret = -EINVAL;
858 }
859 out:
860 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
861 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
862 !btrfs_test_opt(info, CLEAR_CACHE)) {
863 btrfs_err(info, "cannot disable free space tree");
864 ret = -EINVAL;
865
866 }
867 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
868 btrfs_info(info, "disk space caching is enabled");
869 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
870 btrfs_info(info, "using free space tree");
871 return ret;
872 }
873
874 /*
875 * Parse mount options that are required early in the mount process.
876 *
877 * All other options will be parsed on much later in the mount process and
878 * only when we need to allocate a new super block.
879 */
880 static int btrfs_parse_early_options(const char *options, fmode_t flags,
881 void *holder, struct btrfs_fs_devices **fs_devices)
882 {
883 substring_t args[MAX_OPT_ARGS];
884 char *device_name, *opts, *orig, *p;
885 int error = 0;
886
887 if (!options)
888 return 0;
889
890 /*
891 * strsep changes the string, duplicate it because btrfs_parse_options
892 * gets called later
893 */
894 opts = kstrdup(options, GFP_KERNEL);
895 if (!opts)
896 return -ENOMEM;
897 orig = opts;
898
899 while ((p = strsep(&opts, ",")) != NULL) {
900 int token;
901
902 if (!*p)
903 continue;
904
905 token = match_token(p, tokens, args);
906 if (token == Opt_device) {
907 device_name = match_strdup(&args[0]);
908 if (!device_name) {
909 error = -ENOMEM;
910 goto out;
911 }
912 error = btrfs_scan_one_device(device_name,
913 flags, holder, fs_devices);
914 kfree(device_name);
915 if (error)
916 goto out;
917 }
918 }
919
920 out:
921 kfree(orig);
922 return error;
923 }
924
925 /*
926 * Parse mount options that are related to subvolume id
927 *
928 * The value is later passed to mount_subvol()
929 */
930 static int btrfs_parse_subvol_options(const char *options, fmode_t flags,
931 char **subvol_name, u64 *subvol_objectid)
932 {
933 substring_t args[MAX_OPT_ARGS];
934 char *opts, *orig, *p;
935 char *num = NULL;
936 int error = 0;
937
938 if (!options)
939 return 0;
940
941 /*
942 * strsep changes the string, duplicate it because
943 * btrfs_parse_early_options gets called later
944 */
945 opts = kstrdup(options, GFP_KERNEL);
946 if (!opts)
947 return -ENOMEM;
948 orig = opts;
949
950 while ((p = strsep(&opts, ",")) != NULL) {
951 int token;
952 if (!*p)
953 continue;
954
955 token = match_token(p, tokens, args);
956 switch (token) {
957 case Opt_subvol:
958 kfree(*subvol_name);
959 *subvol_name = match_strdup(&args[0]);
960 if (!*subvol_name) {
961 error = -ENOMEM;
962 goto out;
963 }
964 break;
965 case Opt_subvolid:
966 num = match_strdup(&args[0]);
967 if (num) {
968 *subvol_objectid = memparse(num, NULL);
969 kfree(num);
970 /* we want the original fs_tree */
971 if (!*subvol_objectid)
972 *subvol_objectid =
973 BTRFS_FS_TREE_OBJECTID;
974 } else {
975 error = -EINVAL;
976 goto out;
977 }
978 break;
979 case Opt_subvolrootid:
980 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
981 break;
982 default:
983 break;
984 }
985 }
986
987 out:
988 kfree(orig);
989 return error;
990 }
991
992 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
993 u64 subvol_objectid)
994 {
995 struct btrfs_root *root = fs_info->tree_root;
996 struct btrfs_root *fs_root;
997 struct btrfs_root_ref *root_ref;
998 struct btrfs_inode_ref *inode_ref;
999 struct btrfs_key key;
1000 struct btrfs_path *path = NULL;
1001 char *name = NULL, *ptr;
1002 u64 dirid;
1003 int len;
1004 int ret;
1005
1006 path = btrfs_alloc_path();
1007 if (!path) {
1008 ret = -ENOMEM;
1009 goto err;
1010 }
1011 path->leave_spinning = 1;
1012
1013 name = kmalloc(PATH_MAX, GFP_KERNEL);
1014 if (!name) {
1015 ret = -ENOMEM;
1016 goto err;
1017 }
1018 ptr = name + PATH_MAX - 1;
1019 ptr[0] = '\0';
1020
1021 /*
1022 * Walk up the subvolume trees in the tree of tree roots by root
1023 * backrefs until we hit the top-level subvolume.
1024 */
1025 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1026 key.objectid = subvol_objectid;
1027 key.type = BTRFS_ROOT_BACKREF_KEY;
1028 key.offset = (u64)-1;
1029
1030 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1031 if (ret < 0) {
1032 goto err;
1033 } else if (ret > 0) {
1034 ret = btrfs_previous_item(root, path, subvol_objectid,
1035 BTRFS_ROOT_BACKREF_KEY);
1036 if (ret < 0) {
1037 goto err;
1038 } else if (ret > 0) {
1039 ret = -ENOENT;
1040 goto err;
1041 }
1042 }
1043
1044 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1045 subvol_objectid = key.offset;
1046
1047 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1048 struct btrfs_root_ref);
1049 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1050 ptr -= len + 1;
1051 if (ptr < name) {
1052 ret = -ENAMETOOLONG;
1053 goto err;
1054 }
1055 read_extent_buffer(path->nodes[0], ptr + 1,
1056 (unsigned long)(root_ref + 1), len);
1057 ptr[0] = '/';
1058 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1059 btrfs_release_path(path);
1060
1061 key.objectid = subvol_objectid;
1062 key.type = BTRFS_ROOT_ITEM_KEY;
1063 key.offset = (u64)-1;
1064 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1065 if (IS_ERR(fs_root)) {
1066 ret = PTR_ERR(fs_root);
1067 goto err;
1068 }
1069
1070 /*
1071 * Walk up the filesystem tree by inode refs until we hit the
1072 * root directory.
1073 */
1074 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1075 key.objectid = dirid;
1076 key.type = BTRFS_INODE_REF_KEY;
1077 key.offset = (u64)-1;
1078
1079 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1080 if (ret < 0) {
1081 goto err;
1082 } else if (ret > 0) {
1083 ret = btrfs_previous_item(fs_root, path, dirid,
1084 BTRFS_INODE_REF_KEY);
1085 if (ret < 0) {
1086 goto err;
1087 } else if (ret > 0) {
1088 ret = -ENOENT;
1089 goto err;
1090 }
1091 }
1092
1093 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1094 dirid = key.offset;
1095
1096 inode_ref = btrfs_item_ptr(path->nodes[0],
1097 path->slots[0],
1098 struct btrfs_inode_ref);
1099 len = btrfs_inode_ref_name_len(path->nodes[0],
1100 inode_ref);
1101 ptr -= len + 1;
1102 if (ptr < name) {
1103 ret = -ENAMETOOLONG;
1104 goto err;
1105 }
1106 read_extent_buffer(path->nodes[0], ptr + 1,
1107 (unsigned long)(inode_ref + 1), len);
1108 ptr[0] = '/';
1109 btrfs_release_path(path);
1110 }
1111 }
1112
1113 btrfs_free_path(path);
1114 if (ptr == name + PATH_MAX - 1) {
1115 name[0] = '/';
1116 name[1] = '\0';
1117 } else {
1118 memmove(name, ptr, name + PATH_MAX - ptr);
1119 }
1120 return name;
1121
1122 err:
1123 btrfs_free_path(path);
1124 kfree(name);
1125 return ERR_PTR(ret);
1126 }
1127
1128 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1129 {
1130 struct btrfs_root *root = fs_info->tree_root;
1131 struct btrfs_dir_item *di;
1132 struct btrfs_path *path;
1133 struct btrfs_key location;
1134 u64 dir_id;
1135
1136 path = btrfs_alloc_path();
1137 if (!path)
1138 return -ENOMEM;
1139 path->leave_spinning = 1;
1140
1141 /*
1142 * Find the "default" dir item which points to the root item that we
1143 * will mount by default if we haven't been given a specific subvolume
1144 * to mount.
1145 */
1146 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1147 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1148 if (IS_ERR(di)) {
1149 btrfs_free_path(path);
1150 return PTR_ERR(di);
1151 }
1152 if (!di) {
1153 /*
1154 * Ok the default dir item isn't there. This is weird since
1155 * it's always been there, but don't freak out, just try and
1156 * mount the top-level subvolume.
1157 */
1158 btrfs_free_path(path);
1159 *objectid = BTRFS_FS_TREE_OBJECTID;
1160 return 0;
1161 }
1162
1163 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1164 btrfs_free_path(path);
1165 *objectid = location.objectid;
1166 return 0;
1167 }
1168
1169 static int btrfs_fill_super(struct super_block *sb,
1170 struct btrfs_fs_devices *fs_devices,
1171 void *data)
1172 {
1173 struct inode *inode;
1174 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1175 struct btrfs_key key;
1176 int err;
1177
1178 sb->s_maxbytes = MAX_LFS_FILESIZE;
1179 sb->s_magic = BTRFS_SUPER_MAGIC;
1180 sb->s_op = &btrfs_super_ops;
1181 sb->s_d_op = &btrfs_dentry_operations;
1182 sb->s_export_op = &btrfs_export_ops;
1183 sb->s_xattr = btrfs_xattr_handlers;
1184 sb->s_time_gran = 1;
1185 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1186 sb->s_flags |= SB_POSIXACL;
1187 #endif
1188 sb->s_flags |= SB_I_VERSION;
1189 sb->s_iflags |= SB_I_CGROUPWB;
1190
1191 err = super_setup_bdi(sb);
1192 if (err) {
1193 btrfs_err(fs_info, "super_setup_bdi failed");
1194 return err;
1195 }
1196
1197 err = open_ctree(sb, fs_devices, (char *)data);
1198 if (err) {
1199 btrfs_err(fs_info, "open_ctree failed");
1200 return err;
1201 }
1202
1203 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1204 key.type = BTRFS_INODE_ITEM_KEY;
1205 key.offset = 0;
1206 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1207 if (IS_ERR(inode)) {
1208 err = PTR_ERR(inode);
1209 goto fail_close;
1210 }
1211
1212 sb->s_root = d_make_root(inode);
1213 if (!sb->s_root) {
1214 err = -ENOMEM;
1215 goto fail_close;
1216 }
1217
1218 cleancache_init_fs(sb);
1219 sb->s_flags |= SB_ACTIVE;
1220 return 0;
1221
1222 fail_close:
1223 close_ctree(fs_info);
1224 return err;
1225 }
1226
1227 int btrfs_sync_fs(struct super_block *sb, int wait)
1228 {
1229 struct btrfs_trans_handle *trans;
1230 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1231 struct btrfs_root *root = fs_info->tree_root;
1232
1233 trace_btrfs_sync_fs(fs_info, wait);
1234
1235 if (!wait) {
1236 filemap_flush(fs_info->btree_inode->i_mapping);
1237 return 0;
1238 }
1239
1240 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1241
1242 trans = btrfs_attach_transaction_barrier(root);
1243 if (IS_ERR(trans)) {
1244 /* no transaction, don't bother */
1245 if (PTR_ERR(trans) == -ENOENT) {
1246 /*
1247 * Exit unless we have some pending changes
1248 * that need to go through commit
1249 */
1250 if (fs_info->pending_changes == 0)
1251 return 0;
1252 /*
1253 * A non-blocking test if the fs is frozen. We must not
1254 * start a new transaction here otherwise a deadlock
1255 * happens. The pending operations are delayed to the
1256 * next commit after thawing.
1257 */
1258 if (sb_start_write_trylock(sb))
1259 sb_end_write(sb);
1260 else
1261 return 0;
1262 trans = btrfs_start_transaction(root, 0);
1263 }
1264 if (IS_ERR(trans))
1265 return PTR_ERR(trans);
1266 }
1267 return btrfs_commit_transaction(trans);
1268 }
1269
1270 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1271 {
1272 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1273 const char *compress_type;
1274
1275 if (btrfs_test_opt(info, DEGRADED))
1276 seq_puts(seq, ",degraded");
1277 if (btrfs_test_opt(info, NODATASUM))
1278 seq_puts(seq, ",nodatasum");
1279 if (btrfs_test_opt(info, NODATACOW))
1280 seq_puts(seq, ",nodatacow");
1281 if (btrfs_test_opt(info, NOBARRIER))
1282 seq_puts(seq, ",nobarrier");
1283 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1284 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1285 if (info->thread_pool_size != min_t(unsigned long,
1286 num_online_cpus() + 2, 8))
1287 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1288 if (btrfs_test_opt(info, COMPRESS)) {
1289 compress_type = btrfs_compress_type2str(info->compress_type);
1290 if (btrfs_test_opt(info, FORCE_COMPRESS))
1291 seq_printf(seq, ",compress-force=%s", compress_type);
1292 else
1293 seq_printf(seq, ",compress=%s", compress_type);
1294 if (info->compress_level)
1295 seq_printf(seq, ":%d", info->compress_level);
1296 }
1297 if (btrfs_test_opt(info, NOSSD))
1298 seq_puts(seq, ",nossd");
1299 if (btrfs_test_opt(info, SSD_SPREAD))
1300 seq_puts(seq, ",ssd_spread");
1301 else if (btrfs_test_opt(info, SSD))
1302 seq_puts(seq, ",ssd");
1303 if (btrfs_test_opt(info, NOTREELOG))
1304 seq_puts(seq, ",notreelog");
1305 if (btrfs_test_opt(info, NOLOGREPLAY))
1306 seq_puts(seq, ",nologreplay");
1307 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1308 seq_puts(seq, ",flushoncommit");
1309 if (btrfs_test_opt(info, DISCARD))
1310 seq_puts(seq, ",discard");
1311 if (!(info->sb->s_flags & SB_POSIXACL))
1312 seq_puts(seq, ",noacl");
1313 if (btrfs_test_opt(info, SPACE_CACHE))
1314 seq_puts(seq, ",space_cache");
1315 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1316 seq_puts(seq, ",space_cache=v2");
1317 else
1318 seq_puts(seq, ",nospace_cache");
1319 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1320 seq_puts(seq, ",rescan_uuid_tree");
1321 if (btrfs_test_opt(info, CLEAR_CACHE))
1322 seq_puts(seq, ",clear_cache");
1323 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1324 seq_puts(seq, ",user_subvol_rm_allowed");
1325 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1326 seq_puts(seq, ",enospc_debug");
1327 if (btrfs_test_opt(info, AUTO_DEFRAG))
1328 seq_puts(seq, ",autodefrag");
1329 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1330 seq_puts(seq, ",inode_cache");
1331 if (btrfs_test_opt(info, SKIP_BALANCE))
1332 seq_puts(seq, ",skip_balance");
1333 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1334 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1335 seq_puts(seq, ",check_int_data");
1336 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1337 seq_puts(seq, ",check_int");
1338 if (info->check_integrity_print_mask)
1339 seq_printf(seq, ",check_int_print_mask=%d",
1340 info->check_integrity_print_mask);
1341 #endif
1342 if (info->metadata_ratio)
1343 seq_printf(seq, ",metadata_ratio=%d",
1344 info->metadata_ratio);
1345 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1346 seq_puts(seq, ",fatal_errors=panic");
1347 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1348 seq_printf(seq, ",commit=%d", info->commit_interval);
1349 #ifdef CONFIG_BTRFS_DEBUG
1350 if (btrfs_test_opt(info, FRAGMENT_DATA))
1351 seq_puts(seq, ",fragment=data");
1352 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1353 seq_puts(seq, ",fragment=metadata");
1354 #endif
1355 if (btrfs_test_opt(info, REF_VERIFY))
1356 seq_puts(seq, ",ref_verify");
1357 seq_printf(seq, ",subvolid=%llu",
1358 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1359 seq_puts(seq, ",subvol=");
1360 seq_dentry(seq, dentry, " \t\n\\");
1361 return 0;
1362 }
1363
1364 static int btrfs_test_super(struct super_block *s, void *data)
1365 {
1366 struct btrfs_fs_info *p = data;
1367 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1368
1369 return fs_info->fs_devices == p->fs_devices;
1370 }
1371
1372 static int btrfs_set_super(struct super_block *s, void *data)
1373 {
1374 int err = set_anon_super(s, data);
1375 if (!err)
1376 s->s_fs_info = data;
1377 return err;
1378 }
1379
1380 /*
1381 * subvolumes are identified by ino 256
1382 */
1383 static inline int is_subvolume_inode(struct inode *inode)
1384 {
1385 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1386 return 1;
1387 return 0;
1388 }
1389
1390 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1391 const char *device_name, struct vfsmount *mnt)
1392 {
1393 struct dentry *root;
1394 int ret;
1395
1396 if (!subvol_name) {
1397 if (!subvol_objectid) {
1398 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1399 &subvol_objectid);
1400 if (ret) {
1401 root = ERR_PTR(ret);
1402 goto out;
1403 }
1404 }
1405 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1406 subvol_objectid);
1407 if (IS_ERR(subvol_name)) {
1408 root = ERR_CAST(subvol_name);
1409 subvol_name = NULL;
1410 goto out;
1411 }
1412
1413 }
1414
1415 root = mount_subtree(mnt, subvol_name);
1416 /* mount_subtree() drops our reference on the vfsmount. */
1417 mnt = NULL;
1418
1419 if (!IS_ERR(root)) {
1420 struct super_block *s = root->d_sb;
1421 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1422 struct inode *root_inode = d_inode(root);
1423 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1424
1425 ret = 0;
1426 if (!is_subvolume_inode(root_inode)) {
1427 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1428 subvol_name);
1429 ret = -EINVAL;
1430 }
1431 if (subvol_objectid && root_objectid != subvol_objectid) {
1432 /*
1433 * This will also catch a race condition where a
1434 * subvolume which was passed by ID is renamed and
1435 * another subvolume is renamed over the old location.
1436 */
1437 btrfs_err(fs_info,
1438 "subvol '%s' does not match subvolid %llu",
1439 subvol_name, subvol_objectid);
1440 ret = -EINVAL;
1441 }
1442 if (ret) {
1443 dput(root);
1444 root = ERR_PTR(ret);
1445 deactivate_locked_super(s);
1446 }
1447 }
1448
1449 out:
1450 mntput(mnt);
1451 kfree(subvol_name);
1452 return root;
1453 }
1454
1455 static int parse_security_options(char *orig_opts,
1456 struct security_mnt_opts *sec_opts)
1457 {
1458 char *secdata = NULL;
1459 int ret = 0;
1460
1461 secdata = alloc_secdata();
1462 if (!secdata)
1463 return -ENOMEM;
1464 ret = security_sb_copy_data(orig_opts, secdata);
1465 if (ret) {
1466 free_secdata(secdata);
1467 return ret;
1468 }
1469 ret = security_sb_parse_opts_str(secdata, sec_opts);
1470 free_secdata(secdata);
1471 return ret;
1472 }
1473
1474 static int setup_security_options(struct btrfs_fs_info *fs_info,
1475 struct super_block *sb,
1476 struct security_mnt_opts *sec_opts)
1477 {
1478 int ret = 0;
1479
1480 /*
1481 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1482 * is valid.
1483 */
1484 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1485 if (ret)
1486 return ret;
1487
1488 #ifdef CONFIG_SECURITY
1489 if (!fs_info->security_opts.num_mnt_opts) {
1490 /* first time security setup, copy sec_opts to fs_info */
1491 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1492 } else {
1493 /*
1494 * Since SELinux (the only one supporting security_mnt_opts)
1495 * does NOT support changing context during remount/mount of
1496 * the same sb, this must be the same or part of the same
1497 * security options, just free it.
1498 */
1499 security_free_mnt_opts(sec_opts);
1500 }
1501 #endif
1502 return ret;
1503 }
1504
1505 /*
1506 * Find a superblock for the given device / mount point.
1507 *
1508 * Note: This is based on mount_bdev from fs/super.c with a few additions
1509 * for multiple device setup. Make sure to keep it in sync.
1510 */
1511 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1512 int flags, const char *device_name, void *data)
1513 {
1514 struct block_device *bdev = NULL;
1515 struct super_block *s;
1516 struct btrfs_fs_devices *fs_devices = NULL;
1517 struct btrfs_fs_info *fs_info = NULL;
1518 struct security_mnt_opts new_sec_opts;
1519 fmode_t mode = FMODE_READ;
1520 int error = 0;
1521
1522 if (!(flags & SB_RDONLY))
1523 mode |= FMODE_WRITE;
1524
1525 error = btrfs_parse_early_options(data, mode, fs_type,
1526 &fs_devices);
1527 if (error) {
1528 return ERR_PTR(error);
1529 }
1530
1531 security_init_mnt_opts(&new_sec_opts);
1532 if (data) {
1533 error = parse_security_options(data, &new_sec_opts);
1534 if (error)
1535 return ERR_PTR(error);
1536 }
1537
1538 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1539 if (error)
1540 goto error_sec_opts;
1541
1542 /*
1543 * Setup a dummy root and fs_info for test/set super. This is because
1544 * we don't actually fill this stuff out until open_ctree, but we need
1545 * it for searching for existing supers, so this lets us do that and
1546 * then open_ctree will properly initialize everything later.
1547 */
1548 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1549 if (!fs_info) {
1550 error = -ENOMEM;
1551 goto error_sec_opts;
1552 }
1553
1554 fs_info->fs_devices = fs_devices;
1555
1556 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1557 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1558 security_init_mnt_opts(&fs_info->security_opts);
1559 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1560 error = -ENOMEM;
1561 goto error_fs_info;
1562 }
1563
1564 error = btrfs_open_devices(fs_devices, mode, fs_type);
1565 if (error)
1566 goto error_fs_info;
1567
1568 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1569 error = -EACCES;
1570 goto error_close_devices;
1571 }
1572
1573 bdev = fs_devices->latest_bdev;
1574 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1575 fs_info);
1576 if (IS_ERR(s)) {
1577 error = PTR_ERR(s);
1578 goto error_close_devices;
1579 }
1580
1581 if (s->s_root) {
1582 btrfs_close_devices(fs_devices);
1583 free_fs_info(fs_info);
1584 if ((flags ^ s->s_flags) & SB_RDONLY)
1585 error = -EBUSY;
1586 } else {
1587 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1588 btrfs_sb(s)->bdev_holder = fs_type;
1589 error = btrfs_fill_super(s, fs_devices, data);
1590 }
1591 if (error) {
1592 deactivate_locked_super(s);
1593 goto error_sec_opts;
1594 }
1595
1596 fs_info = btrfs_sb(s);
1597 error = setup_security_options(fs_info, s, &new_sec_opts);
1598 if (error) {
1599 deactivate_locked_super(s);
1600 goto error_sec_opts;
1601 }
1602
1603 return dget(s->s_root);
1604
1605 error_close_devices:
1606 btrfs_close_devices(fs_devices);
1607 error_fs_info:
1608 free_fs_info(fs_info);
1609 error_sec_opts:
1610 security_free_mnt_opts(&new_sec_opts);
1611 return ERR_PTR(error);
1612 }
1613
1614 /*
1615 * Mount function which is called by VFS layer.
1616 *
1617 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1618 * which needs vfsmount* of device's root (/). This means device's root has to
1619 * be mounted internally in any case.
1620 *
1621 * Operation flow:
1622 * 1. Parse subvol id related options for later use in mount_subvol().
1623 *
1624 * 2. Mount device's root (/) by calling vfs_kern_mount().
1625 *
1626 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1627 * first place. In order to avoid calling btrfs_mount() again, we use
1628 * different file_system_type which is not registered to VFS by
1629 * register_filesystem() (btrfs_root_fs_type). As a result,
1630 * btrfs_mount_root() is called. The return value will be used by
1631 * mount_subtree() in mount_subvol().
1632 *
1633 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1634 * "btrfs subvolume set-default", mount_subvol() is called always.
1635 */
1636 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1637 const char *device_name, void *data)
1638 {
1639 struct vfsmount *mnt_root;
1640 struct dentry *root;
1641 fmode_t mode = FMODE_READ;
1642 char *subvol_name = NULL;
1643 u64 subvol_objectid = 0;
1644 int error = 0;
1645
1646 if (!(flags & SB_RDONLY))
1647 mode |= FMODE_WRITE;
1648
1649 error = btrfs_parse_subvol_options(data, mode,
1650 &subvol_name, &subvol_objectid);
1651 if (error) {
1652 kfree(subvol_name);
1653 return ERR_PTR(error);
1654 }
1655
1656 /* mount device's root (/) */
1657 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1658 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1659 if (flags & SB_RDONLY) {
1660 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1661 flags & ~SB_RDONLY, device_name, data);
1662 } else {
1663 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1664 flags | SB_RDONLY, device_name, data);
1665 if (IS_ERR(mnt_root)) {
1666 root = ERR_CAST(mnt_root);
1667 goto out;
1668 }
1669
1670 down_write(&mnt_root->mnt_sb->s_umount);
1671 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1672 up_write(&mnt_root->mnt_sb->s_umount);
1673 if (error < 0) {
1674 root = ERR_PTR(error);
1675 mntput(mnt_root);
1676 goto out;
1677 }
1678 }
1679 }
1680 if (IS_ERR(mnt_root)) {
1681 root = ERR_CAST(mnt_root);
1682 goto out;
1683 }
1684
1685 /* mount_subvol() will free subvol_name and mnt_root */
1686 root = mount_subvol(subvol_name, subvol_objectid, device_name, mnt_root);
1687
1688 out:
1689 return root;
1690 }
1691
1692 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1693 int new_pool_size, int old_pool_size)
1694 {
1695 if (new_pool_size == old_pool_size)
1696 return;
1697
1698 fs_info->thread_pool_size = new_pool_size;
1699
1700 btrfs_info(fs_info, "resize thread pool %d -> %d",
1701 old_pool_size, new_pool_size);
1702
1703 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1704 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1705 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1706 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1707 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1708 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1709 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1710 new_pool_size);
1711 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1712 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1713 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1714 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1715 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1716 new_pool_size);
1717 }
1718
1719 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1720 {
1721 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1722 }
1723
1724 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1725 unsigned long old_opts, int flags)
1726 {
1727 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1728 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1729 (flags & SB_RDONLY))) {
1730 /* wait for any defraggers to finish */
1731 wait_event(fs_info->transaction_wait,
1732 (atomic_read(&fs_info->defrag_running) == 0));
1733 if (flags & SB_RDONLY)
1734 sync_filesystem(fs_info->sb);
1735 }
1736 }
1737
1738 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1739 unsigned long old_opts)
1740 {
1741 /*
1742 * We need to cleanup all defragable inodes if the autodefragment is
1743 * close or the filesystem is read only.
1744 */
1745 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1746 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1747 btrfs_cleanup_defrag_inodes(fs_info);
1748 }
1749
1750 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1751 }
1752
1753 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1754 {
1755 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1756 struct btrfs_root *root = fs_info->tree_root;
1757 unsigned old_flags = sb->s_flags;
1758 unsigned long old_opts = fs_info->mount_opt;
1759 unsigned long old_compress_type = fs_info->compress_type;
1760 u64 old_max_inline = fs_info->max_inline;
1761 int old_thread_pool_size = fs_info->thread_pool_size;
1762 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1763 int ret;
1764
1765 sync_filesystem(sb);
1766 btrfs_remount_prepare(fs_info);
1767
1768 if (data) {
1769 struct security_mnt_opts new_sec_opts;
1770
1771 security_init_mnt_opts(&new_sec_opts);
1772 ret = parse_security_options(data, &new_sec_opts);
1773 if (ret)
1774 goto restore;
1775 ret = setup_security_options(fs_info, sb,
1776 &new_sec_opts);
1777 if (ret) {
1778 security_free_mnt_opts(&new_sec_opts);
1779 goto restore;
1780 }
1781 }
1782
1783 ret = btrfs_parse_options(fs_info, data, *flags);
1784 if (ret) {
1785 ret = -EINVAL;
1786 goto restore;
1787 }
1788
1789 btrfs_remount_begin(fs_info, old_opts, *flags);
1790 btrfs_resize_thread_pool(fs_info,
1791 fs_info->thread_pool_size, old_thread_pool_size);
1792
1793 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1794 goto out;
1795
1796 if (*flags & SB_RDONLY) {
1797 /*
1798 * this also happens on 'umount -rf' or on shutdown, when
1799 * the filesystem is busy.
1800 */
1801 cancel_work_sync(&fs_info->async_reclaim_work);
1802
1803 /* wait for the uuid_scan task to finish */
1804 down(&fs_info->uuid_tree_rescan_sem);
1805 /* avoid complains from lockdep et al. */
1806 up(&fs_info->uuid_tree_rescan_sem);
1807
1808 sb->s_flags |= SB_RDONLY;
1809
1810 /*
1811 * Setting SB_RDONLY will put the cleaner thread to
1812 * sleep at the next loop if it's already active.
1813 * If it's already asleep, we'll leave unused block
1814 * groups on disk until we're mounted read-write again
1815 * unless we clean them up here.
1816 */
1817 btrfs_delete_unused_bgs(fs_info);
1818
1819 btrfs_dev_replace_suspend_for_unmount(fs_info);
1820 btrfs_scrub_cancel(fs_info);
1821 btrfs_pause_balance(fs_info);
1822
1823 ret = btrfs_commit_super(fs_info);
1824 if (ret)
1825 goto restore;
1826 } else {
1827 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1828 btrfs_err(fs_info,
1829 "Remounting read-write after error is not allowed");
1830 ret = -EINVAL;
1831 goto restore;
1832 }
1833 if (fs_info->fs_devices->rw_devices == 0) {
1834 ret = -EACCES;
1835 goto restore;
1836 }
1837
1838 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1839 btrfs_warn(fs_info,
1840 "too many missing devices, writeable remount is not allowed");
1841 ret = -EACCES;
1842 goto restore;
1843 }
1844
1845 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1846 ret = -EINVAL;
1847 goto restore;
1848 }
1849
1850 ret = btrfs_cleanup_fs_roots(fs_info);
1851 if (ret)
1852 goto restore;
1853
1854 /* recover relocation */
1855 mutex_lock(&fs_info->cleaner_mutex);
1856 ret = btrfs_recover_relocation(root);
1857 mutex_unlock(&fs_info->cleaner_mutex);
1858 if (ret)
1859 goto restore;
1860
1861 ret = btrfs_resume_balance_async(fs_info);
1862 if (ret)
1863 goto restore;
1864
1865 ret = btrfs_resume_dev_replace_async(fs_info);
1866 if (ret) {
1867 btrfs_warn(fs_info, "failed to resume dev_replace");
1868 goto restore;
1869 }
1870
1871 btrfs_qgroup_rescan_resume(fs_info);
1872
1873 if (!fs_info->uuid_root) {
1874 btrfs_info(fs_info, "creating UUID tree");
1875 ret = btrfs_create_uuid_tree(fs_info);
1876 if (ret) {
1877 btrfs_warn(fs_info,
1878 "failed to create the UUID tree %d",
1879 ret);
1880 goto restore;
1881 }
1882 }
1883 sb->s_flags &= ~SB_RDONLY;
1884
1885 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1886 }
1887 out:
1888 wake_up_process(fs_info->transaction_kthread);
1889 btrfs_remount_cleanup(fs_info, old_opts);
1890 return 0;
1891
1892 restore:
1893 /* We've hit an error - don't reset SB_RDONLY */
1894 if (sb_rdonly(sb))
1895 old_flags |= SB_RDONLY;
1896 sb->s_flags = old_flags;
1897 fs_info->mount_opt = old_opts;
1898 fs_info->compress_type = old_compress_type;
1899 fs_info->max_inline = old_max_inline;
1900 btrfs_resize_thread_pool(fs_info,
1901 old_thread_pool_size, fs_info->thread_pool_size);
1902 fs_info->metadata_ratio = old_metadata_ratio;
1903 btrfs_remount_cleanup(fs_info, old_opts);
1904 return ret;
1905 }
1906
1907 /* Used to sort the devices by max_avail(descending sort) */
1908 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1909 const void *dev_info2)
1910 {
1911 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1912 ((struct btrfs_device_info *)dev_info2)->max_avail)
1913 return -1;
1914 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1915 ((struct btrfs_device_info *)dev_info2)->max_avail)
1916 return 1;
1917 else
1918 return 0;
1919 }
1920
1921 /*
1922 * sort the devices by max_avail, in which max free extent size of each device
1923 * is stored.(Descending Sort)
1924 */
1925 static inline void btrfs_descending_sort_devices(
1926 struct btrfs_device_info *devices,
1927 size_t nr_devices)
1928 {
1929 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1930 btrfs_cmp_device_free_bytes, NULL);
1931 }
1932
1933 /*
1934 * The helper to calc the free space on the devices that can be used to store
1935 * file data.
1936 */
1937 static int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1938 u64 *free_bytes)
1939 {
1940 struct btrfs_device_info *devices_info;
1941 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1942 struct btrfs_device *device;
1943 u64 skip_space;
1944 u64 type;
1945 u64 avail_space;
1946 u64 min_stripe_size;
1947 int min_stripes = 1, num_stripes = 1;
1948 int i = 0, nr_devices;
1949
1950 /*
1951 * We aren't under the device list lock, so this is racy-ish, but good
1952 * enough for our purposes.
1953 */
1954 nr_devices = fs_info->fs_devices->open_devices;
1955 if (!nr_devices) {
1956 smp_mb();
1957 nr_devices = fs_info->fs_devices->open_devices;
1958 ASSERT(nr_devices);
1959 if (!nr_devices) {
1960 *free_bytes = 0;
1961 return 0;
1962 }
1963 }
1964
1965 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1966 GFP_KERNEL);
1967 if (!devices_info)
1968 return -ENOMEM;
1969
1970 /* calc min stripe number for data space allocation */
1971 type = btrfs_data_alloc_profile(fs_info);
1972 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1973 min_stripes = 2;
1974 num_stripes = nr_devices;
1975 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1976 min_stripes = 2;
1977 num_stripes = 2;
1978 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1979 min_stripes = 4;
1980 num_stripes = 4;
1981 }
1982
1983 if (type & BTRFS_BLOCK_GROUP_DUP)
1984 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1985 else
1986 min_stripe_size = BTRFS_STRIPE_LEN;
1987
1988 rcu_read_lock();
1989 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1990 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1991 &device->dev_state) ||
1992 !device->bdev ||
1993 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1994 continue;
1995
1996 if (i >= nr_devices)
1997 break;
1998
1999 avail_space = device->total_bytes - device->bytes_used;
2000
2001 /* align with stripe_len */
2002 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
2003 avail_space *= BTRFS_STRIPE_LEN;
2004
2005 /*
2006 * In order to avoid overwriting the superblock on the drive,
2007 * btrfs starts at an offset of at least 1MB when doing chunk
2008 * allocation.
2009 */
2010 skip_space = SZ_1M;
2011
2012 /*
2013 * we can use the free space in [0, skip_space - 1], subtract
2014 * it from the total.
2015 */
2016 if (avail_space && avail_space >= skip_space)
2017 avail_space -= skip_space;
2018 else
2019 avail_space = 0;
2020
2021 if (avail_space < min_stripe_size)
2022 continue;
2023
2024 devices_info[i].dev = device;
2025 devices_info[i].max_avail = avail_space;
2026
2027 i++;
2028 }
2029 rcu_read_unlock();
2030
2031 nr_devices = i;
2032
2033 btrfs_descending_sort_devices(devices_info, nr_devices);
2034
2035 i = nr_devices - 1;
2036 avail_space = 0;
2037 while (nr_devices >= min_stripes) {
2038 if (num_stripes > nr_devices)
2039 num_stripes = nr_devices;
2040
2041 if (devices_info[i].max_avail >= min_stripe_size) {
2042 int j;
2043 u64 alloc_size;
2044
2045 avail_space += devices_info[i].max_avail * num_stripes;
2046 alloc_size = devices_info[i].max_avail;
2047 for (j = i + 1 - num_stripes; j <= i; j++)
2048 devices_info[j].max_avail -= alloc_size;
2049 }
2050 i--;
2051 nr_devices--;
2052 }
2053
2054 kfree(devices_info);
2055 *free_bytes = avail_space;
2056 return 0;
2057 }
2058
2059 /*
2060 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2061 *
2062 * If there's a redundant raid level at DATA block groups, use the respective
2063 * multiplier to scale the sizes.
2064 *
2065 * Unused device space usage is based on simulating the chunk allocator
2066 * algorithm that respects the device sizes and order of allocations. This is
2067 * a close approximation of the actual use but there are other factors that may
2068 * change the result (like a new metadata chunk).
2069 *
2070 * If metadata is exhausted, f_bavail will be 0.
2071 */
2072 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2073 {
2074 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2075 struct btrfs_super_block *disk_super = fs_info->super_copy;
2076 struct list_head *head = &fs_info->space_info;
2077 struct btrfs_space_info *found;
2078 u64 total_used = 0;
2079 u64 total_free_data = 0;
2080 u64 total_free_meta = 0;
2081 int bits = dentry->d_sb->s_blocksize_bits;
2082 __be32 *fsid = (__be32 *)fs_info->fsid;
2083 unsigned factor = 1;
2084 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2085 int ret;
2086 u64 thresh = 0;
2087 int mixed = 0;
2088
2089 rcu_read_lock();
2090 list_for_each_entry_rcu(found, head, list) {
2091 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2092 int i;
2093
2094 total_free_data += found->disk_total - found->disk_used;
2095 total_free_data -=
2096 btrfs_account_ro_block_groups_free_space(found);
2097
2098 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2099 if (!list_empty(&found->block_groups[i])) {
2100 switch (i) {
2101 case BTRFS_RAID_DUP:
2102 case BTRFS_RAID_RAID1:
2103 case BTRFS_RAID_RAID10:
2104 factor = 2;
2105 }
2106 }
2107 }
2108 }
2109
2110 /*
2111 * Metadata in mixed block goup profiles are accounted in data
2112 */
2113 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2114 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2115 mixed = 1;
2116 else
2117 total_free_meta += found->disk_total -
2118 found->disk_used;
2119 }
2120
2121 total_used += found->disk_used;
2122 }
2123
2124 rcu_read_unlock();
2125
2126 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2127 buf->f_blocks >>= bits;
2128 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2129
2130 /* Account global block reserve as used, it's in logical size already */
2131 spin_lock(&block_rsv->lock);
2132 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2133 if (buf->f_bfree >= block_rsv->size >> bits)
2134 buf->f_bfree -= block_rsv->size >> bits;
2135 else
2136 buf->f_bfree = 0;
2137 spin_unlock(&block_rsv->lock);
2138
2139 buf->f_bavail = div_u64(total_free_data, factor);
2140 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2141 if (ret)
2142 return ret;
2143 buf->f_bavail += div_u64(total_free_data, factor);
2144 buf->f_bavail = buf->f_bavail >> bits;
2145
2146 /*
2147 * We calculate the remaining metadata space minus global reserve. If
2148 * this is (supposedly) smaller than zero, there's no space. But this
2149 * does not hold in practice, the exhausted state happens where's still
2150 * some positive delta. So we apply some guesswork and compare the
2151 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2152 *
2153 * We probably cannot calculate the exact threshold value because this
2154 * depends on the internal reservations requested by various
2155 * operations, so some operations that consume a few metadata will
2156 * succeed even if the Avail is zero. But this is better than the other
2157 * way around.
2158 */
2159 thresh = SZ_4M;
2160
2161 if (!mixed && total_free_meta - thresh < block_rsv->size)
2162 buf->f_bavail = 0;
2163
2164 buf->f_type = BTRFS_SUPER_MAGIC;
2165 buf->f_bsize = dentry->d_sb->s_blocksize;
2166 buf->f_namelen = BTRFS_NAME_LEN;
2167
2168 /* We treat it as constant endianness (it doesn't matter _which_)
2169 because we want the fsid to come out the same whether mounted
2170 on a big-endian or little-endian host */
2171 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2172 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2173 /* Mask in the root object ID too, to disambiguate subvols */
2174 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2175 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2176
2177 return 0;
2178 }
2179
2180 static void btrfs_kill_super(struct super_block *sb)
2181 {
2182 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2183 kill_anon_super(sb);
2184 free_fs_info(fs_info);
2185 }
2186
2187 static struct file_system_type btrfs_fs_type = {
2188 .owner = THIS_MODULE,
2189 .name = "btrfs",
2190 .mount = btrfs_mount,
2191 .kill_sb = btrfs_kill_super,
2192 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2193 };
2194
2195 static struct file_system_type btrfs_root_fs_type = {
2196 .owner = THIS_MODULE,
2197 .name = "btrfs",
2198 .mount = btrfs_mount_root,
2199 .kill_sb = btrfs_kill_super,
2200 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2201 };
2202
2203 MODULE_ALIAS_FS("btrfs");
2204
2205 static int btrfs_control_open(struct inode *inode, struct file *file)
2206 {
2207 /*
2208 * The control file's private_data is used to hold the
2209 * transaction when it is started and is used to keep
2210 * track of whether a transaction is already in progress.
2211 */
2212 file->private_data = NULL;
2213 return 0;
2214 }
2215
2216 /*
2217 * used by btrfsctl to scan devices when no FS is mounted
2218 */
2219 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2220 unsigned long arg)
2221 {
2222 struct btrfs_ioctl_vol_args *vol;
2223 struct btrfs_fs_devices *fs_devices;
2224 int ret = -ENOTTY;
2225
2226 if (!capable(CAP_SYS_ADMIN))
2227 return -EPERM;
2228
2229 vol = memdup_user((void __user *)arg, sizeof(*vol));
2230 if (IS_ERR(vol))
2231 return PTR_ERR(vol);
2232
2233 switch (cmd) {
2234 case BTRFS_IOC_SCAN_DEV:
2235 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2236 &btrfs_root_fs_type, &fs_devices);
2237 break;
2238 case BTRFS_IOC_DEVICES_READY:
2239 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2240 &btrfs_root_fs_type, &fs_devices);
2241 if (ret)
2242 break;
2243 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2244 break;
2245 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2246 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2247 break;
2248 }
2249
2250 kfree(vol);
2251 return ret;
2252 }
2253
2254 static int btrfs_freeze(struct super_block *sb)
2255 {
2256 struct btrfs_trans_handle *trans;
2257 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2258 struct btrfs_root *root = fs_info->tree_root;
2259
2260 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2261 /*
2262 * We don't need a barrier here, we'll wait for any transaction that
2263 * could be in progress on other threads (and do delayed iputs that
2264 * we want to avoid on a frozen filesystem), or do the commit
2265 * ourselves.
2266 */
2267 trans = btrfs_attach_transaction_barrier(root);
2268 if (IS_ERR(trans)) {
2269 /* no transaction, don't bother */
2270 if (PTR_ERR(trans) == -ENOENT)
2271 return 0;
2272 return PTR_ERR(trans);
2273 }
2274 return btrfs_commit_transaction(trans);
2275 }
2276
2277 static int btrfs_unfreeze(struct super_block *sb)
2278 {
2279 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2280
2281 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2282 return 0;
2283 }
2284
2285 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2286 {
2287 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2288 struct btrfs_fs_devices *cur_devices;
2289 struct btrfs_device *dev, *first_dev = NULL;
2290 struct list_head *head;
2291 struct rcu_string *name;
2292
2293 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2294 cur_devices = fs_info->fs_devices;
2295 while (cur_devices) {
2296 head = &cur_devices->devices;
2297 list_for_each_entry(dev, head, dev_list) {
2298 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2299 continue;
2300 if (!dev->name)
2301 continue;
2302 if (!first_dev || dev->devid < first_dev->devid)
2303 first_dev = dev;
2304 }
2305 cur_devices = cur_devices->seed;
2306 }
2307
2308 if (first_dev) {
2309 rcu_read_lock();
2310 name = rcu_dereference(first_dev->name);
2311 seq_escape(m, name->str, " \t\n\\");
2312 rcu_read_unlock();
2313 } else {
2314 WARN_ON(1);
2315 }
2316 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2317 return 0;
2318 }
2319
2320 static const struct super_operations btrfs_super_ops = {
2321 .drop_inode = btrfs_drop_inode,
2322 .evict_inode = btrfs_evict_inode,
2323 .put_super = btrfs_put_super,
2324 .sync_fs = btrfs_sync_fs,
2325 .show_options = btrfs_show_options,
2326 .show_devname = btrfs_show_devname,
2327 .write_inode = btrfs_write_inode,
2328 .alloc_inode = btrfs_alloc_inode,
2329 .destroy_inode = btrfs_destroy_inode,
2330 .statfs = btrfs_statfs,
2331 .remount_fs = btrfs_remount,
2332 .freeze_fs = btrfs_freeze,
2333 .unfreeze_fs = btrfs_unfreeze,
2334 };
2335
2336 static const struct file_operations btrfs_ctl_fops = {
2337 .open = btrfs_control_open,
2338 .unlocked_ioctl = btrfs_control_ioctl,
2339 .compat_ioctl = btrfs_control_ioctl,
2340 .owner = THIS_MODULE,
2341 .llseek = noop_llseek,
2342 };
2343
2344 static struct miscdevice btrfs_misc = {
2345 .minor = BTRFS_MINOR,
2346 .name = "btrfs-control",
2347 .fops = &btrfs_ctl_fops
2348 };
2349
2350 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2351 MODULE_ALIAS("devname:btrfs-control");
2352
2353 static int __init btrfs_interface_init(void)
2354 {
2355 return misc_register(&btrfs_misc);
2356 }
2357
2358 static void btrfs_interface_exit(void)
2359 {
2360 misc_deregister(&btrfs_misc);
2361 }
2362
2363 static void __init btrfs_print_mod_info(void)
2364 {
2365 pr_info("Btrfs loaded, crc32c=%s"
2366 #ifdef CONFIG_BTRFS_DEBUG
2367 ", debug=on"
2368 #endif
2369 #ifdef CONFIG_BTRFS_ASSERT
2370 ", assert=on"
2371 #endif
2372 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2373 ", integrity-checker=on"
2374 #endif
2375 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2376 ", ref-verify=on"
2377 #endif
2378 "\n",
2379 btrfs_crc32c_impl());
2380 }
2381
2382 static int __init init_btrfs_fs(void)
2383 {
2384 int err;
2385
2386 err = btrfs_hash_init();
2387 if (err)
2388 return err;
2389
2390 btrfs_props_init();
2391
2392 err = btrfs_init_sysfs();
2393 if (err)
2394 goto free_hash;
2395
2396 btrfs_init_compress();
2397
2398 err = btrfs_init_cachep();
2399 if (err)
2400 goto free_compress;
2401
2402 err = extent_io_init();
2403 if (err)
2404 goto free_cachep;
2405
2406 err = extent_map_init();
2407 if (err)
2408 goto free_extent_io;
2409
2410 err = ordered_data_init();
2411 if (err)
2412 goto free_extent_map;
2413
2414 err = btrfs_delayed_inode_init();
2415 if (err)
2416 goto free_ordered_data;
2417
2418 err = btrfs_auto_defrag_init();
2419 if (err)
2420 goto free_delayed_inode;
2421
2422 err = btrfs_delayed_ref_init();
2423 if (err)
2424 goto free_auto_defrag;
2425
2426 err = btrfs_prelim_ref_init();
2427 if (err)
2428 goto free_delayed_ref;
2429
2430 err = btrfs_end_io_wq_init();
2431 if (err)
2432 goto free_prelim_ref;
2433
2434 err = btrfs_interface_init();
2435 if (err)
2436 goto free_end_io_wq;
2437
2438 btrfs_init_lockdep();
2439
2440 btrfs_print_mod_info();
2441
2442 err = btrfs_run_sanity_tests();
2443 if (err)
2444 goto unregister_ioctl;
2445
2446 err = register_filesystem(&btrfs_fs_type);
2447 if (err)
2448 goto unregister_ioctl;
2449
2450 return 0;
2451
2452 unregister_ioctl:
2453 btrfs_interface_exit();
2454 free_end_io_wq:
2455 btrfs_end_io_wq_exit();
2456 free_prelim_ref:
2457 btrfs_prelim_ref_exit();
2458 free_delayed_ref:
2459 btrfs_delayed_ref_exit();
2460 free_auto_defrag:
2461 btrfs_auto_defrag_exit();
2462 free_delayed_inode:
2463 btrfs_delayed_inode_exit();
2464 free_ordered_data:
2465 ordered_data_exit();
2466 free_extent_map:
2467 extent_map_exit();
2468 free_extent_io:
2469 extent_io_exit();
2470 free_cachep:
2471 btrfs_destroy_cachep();
2472 free_compress:
2473 btrfs_exit_compress();
2474 btrfs_exit_sysfs();
2475 free_hash:
2476 btrfs_hash_exit();
2477 return err;
2478 }
2479
2480 static void __exit exit_btrfs_fs(void)
2481 {
2482 btrfs_destroy_cachep();
2483 btrfs_delayed_ref_exit();
2484 btrfs_auto_defrag_exit();
2485 btrfs_delayed_inode_exit();
2486 btrfs_prelim_ref_exit();
2487 ordered_data_exit();
2488 extent_map_exit();
2489 extent_io_exit();
2490 btrfs_interface_exit();
2491 btrfs_end_io_wq_exit();
2492 unregister_filesystem(&btrfs_fs_type);
2493 btrfs_exit_sysfs();
2494 btrfs_cleanup_fs_uuids();
2495 btrfs_exit_compress();
2496 btrfs_hash_exit();
2497 }
2498
2499 late_initcall(init_btrfs_fs);
2500 module_exit(exit_btrfs_fs)
2501
2502 MODULE_LICENSE("GPL");
CRIS linux kernel tree