projects / deliverable / linux.git / blob
? search:


06537d217419b4b8ac138873a01853bd01f716b9
[deliverable/linux.git] / fs / btrfs / super.c
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19 #include <linux/blkdev.h>
20 #include <linux/module.h>
21 #include <linux/buffer_head.h>
22 #include <linux/fs.h>
23 #include <linux/pagemap.h>
24 #include <linux/highmem.h>
25 #include <linux/time.h>
26 #include <linux/init.h>
27 #include <linux/seq_file.h>
28 #include <linux/string.h>
29 #include <linux/backing-dev.h>
30 #include <linux/mount.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/parser.h>
37 #include <linux/ctype.h>
38 #include <linux/namei.h>
39 #include <linux/miscdevice.h>
40 #include <linux/magic.h>
41 #include <linux/slab.h>
42 #include <linux/cleancache.h>
43 #include <linux/ratelimit.h>
44 #include <linux/btrfs.h>
45 #include "compat.h"
46 #include "delayed-inode.h"
47 #include "ctree.h"
48 #include "disk-io.h"
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
52 #include "xattr.h"
53 #include "volumes.h"
54 #include "version.h"
55 #include "export.h"
56 #include "compression.h"
57 #include "rcu-string.h"
58 #include "dev-replace.h"
59
60 #define CREATE_TRACE_POINTS
61 #include <trace/events/btrfs.h>
62
63 static const struct super_operations btrfs_super_ops;
64 static struct file_system_type btrfs_fs_type;
65
66 static const char *btrfs_decode_error(int errno, char nbuf[16])
67 {
68 char *errstr = NULL;
69
70 switch (errno) {
71 case -EIO:
72 errstr = "IO failure";
73 break;
74 case -ENOMEM:
75 errstr = "Out of memory";
76 break;
77 case -EROFS:
78 errstr = "Readonly filesystem";
79 break;
80 case -EEXIST:
81 errstr = "Object already exists";
82 break;
83 default:
84 if (nbuf) {
85 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
86 errstr = nbuf;
87 }
88 break;
89 }
90
91 return errstr;
92 }
93
94 static void __save_error_info(struct btrfs_fs_info *fs_info)
95 {
96 /*
97 * today we only save the error info into ram. Long term we'll
98 * also send it down to the disk
99 */
100 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
101 }
102
103 static void save_error_info(struct btrfs_fs_info *fs_info)
104 {
105 __save_error_info(fs_info);
106 }
107
108 /* btrfs handle error by forcing the filesystem readonly */
109 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
110 {
111 struct super_block *sb = fs_info->sb;
112
113 if (sb->s_flags & MS_RDONLY)
114 return;
115
116 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
117 sb->s_flags |= MS_RDONLY;
118 printk(KERN_INFO "btrfs is forced readonly\n");
119 /*
120 * Note that a running device replace operation is not
121 * canceled here although there is no way to update
122 * the progress. It would add the risk of a deadlock,
123 * therefore the canceling is ommited. The only penalty
124 * is that some I/O remains active until the procedure
125 * completes. The next time when the filesystem is
126 * mounted writeable again, the device replace
127 * operation continues.
128 */
129 // WARN_ON(1);
130 }
131 }
132
133 #ifdef CONFIG_PRINTK
134 /*
135 * __btrfs_std_error decodes expected errors from the caller and
136 * invokes the approciate error response.
137 */
138 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
139 unsigned int line, int errno, const char *fmt, ...)
140 {
141 struct super_block *sb = fs_info->sb;
142 char nbuf[16];
143 const char *errstr;
144 va_list args;
145 va_start(args, fmt);
146
147 /*
148 * Special case: if the error is EROFS, and we're already
149 * under MS_RDONLY, then it is safe here.
150 */
151 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
152 return;
153
154 errstr = btrfs_decode_error(errno, nbuf);
155 if (fmt) {
156 struct va_format vaf = {
157 .fmt = fmt,
158 .va = &args,
159 };
160
161 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s (%pV)\n",
162 sb->s_id, function, line, errstr, &vaf);
163 } else {
164 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
165 sb->s_id, function, line, errstr);
166 }
167
168 /* Don't go through full error handling during mount */
169 if (sb->s_flags & MS_BORN) {
170 save_error_info(fs_info);
171 btrfs_handle_error(fs_info);
172 }
173 va_end(args);
174 }
175
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 void btrfs_printk(struct btrfs_fs_info *fs_info, const char *fmt, ...)
188 {
189 struct super_block *sb = fs_info->sb;
190 char lvl[4];
191 struct va_format vaf;
192 va_list args;
193 const char *type = logtypes[4];
194 int kern_level;
195
196 va_start(args, fmt);
197
198 kern_level = printk_get_level(fmt);
199 if (kern_level) {
200 size_t size = printk_skip_level(fmt) - fmt;
201 memcpy(lvl, fmt, size);
202 lvl[size] = '\0';
203 fmt += size;
204 type = logtypes[kern_level - '0'];
205 } else
206 *lvl = '\0';
207
208 vaf.fmt = fmt;
209 vaf.va = &args;
210
211 printk("%sBTRFS %s (device %s): %pV", lvl, type, sb->s_id, &vaf);
212
213 va_end(args);
214 }
215
216 #else
217
218 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
219 unsigned int line, int errno, const char *fmt, ...)
220 {
221 struct super_block *sb = fs_info->sb;
222
223 /*
224 * Special case: if the error is EROFS, and we're already
225 * under MS_RDONLY, then it is safe here.
226 */
227 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
228 return;
229
230 /* Don't go through full error handling during mount */
231 if (sb->s_flags & MS_BORN) {
232 save_error_info(fs_info);
233 btrfs_handle_error(fs_info);
234 }
235 }
236 #endif
237
238 /*
239 * We only mark the transaction aborted and then set the file system read-only.
240 * This will prevent new transactions from starting or trying to join this
241 * one.
242 *
243 * This means that error recovery at the call site is limited to freeing
244 * any local memory allocations and passing the error code up without
245 * further cleanup. The transaction should complete as it normally would
246 * in the call path but will return -EIO.
247 *
248 * We'll complete the cleanup in btrfs_end_transaction and
249 * btrfs_commit_transaction.
250 */
251 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
252 struct btrfs_root *root, const char *function,
253 unsigned int line, int errno)
254 {
255 WARN_ONCE(1, KERN_DEBUG "btrfs: Transaction aborted\n");
256 trans->aborted = errno;
257 /* Nothing used. The other threads that have joined this
258 * transaction may be able to continue. */
259 if (!trans->blocks_used) {
260 char nbuf[16];
261 const char *errstr;
262
263 errstr = btrfs_decode_error(errno, nbuf);
264 btrfs_printk(root->fs_info,
265 "%s:%d: Aborting unused transaction(%s).\n",
266 function, line, errstr);
267 return;
268 }
269 ACCESS_ONCE(trans->transaction->aborted) = errno;
270 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
271 }
272 /*
273 * __btrfs_panic decodes unexpected, fatal errors from the caller,
274 * issues an alert, and either panics or BUGs, depending on mount options.
275 */
276 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
277 unsigned int line, int errno, const char *fmt, ...)
278 {
279 char nbuf[16];
280 char *s_id = "<unknown>";
281 const char *errstr;
282 struct va_format vaf = { .fmt = fmt };
283 va_list args;
284
285 if (fs_info)
286 s_id = fs_info->sb->s_id;
287
288 va_start(args, fmt);
289 vaf.va = &args;
290
291 errstr = btrfs_decode_error(errno, nbuf);
292 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
293 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
294 s_id, function, line, &vaf, errstr);
295
296 printk(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
297 s_id, function, line, &vaf, errstr);
298 va_end(args);
299 /* Caller calls BUG() */
300 }
301
302 static void btrfs_put_super(struct super_block *sb)
303 {
304 (void)close_ctree(btrfs_sb(sb)->tree_root);
305 /* FIXME: need to fix VFS to return error? */
306 /* AV: return it _where_? ->put_super() can be triggered by any number
307 * of async events, up to and including delivery of SIGKILL to the
308 * last process that kept it busy. Or segfault in the aforementioned
309 * process... Whom would you report that to?
310 */
311 }
312
313 enum {
314 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
315 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
316 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
317 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
318 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
319 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
320 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
321 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
322 Opt_check_integrity, Opt_check_integrity_including_extent_data,
323 Opt_check_integrity_print_mask, Opt_fatal_errors,
324 Opt_err,
325 };
326
327 static match_table_t tokens = {
328 {Opt_degraded, "degraded"},
329 {Opt_subvol, "subvol=%s"},
330 {Opt_subvolid, "subvolid=%d"},
331 {Opt_device, "device=%s"},
332 {Opt_nodatasum, "nodatasum"},
333 {Opt_nodatacow, "nodatacow"},
334 {Opt_nobarrier, "nobarrier"},
335 {Opt_max_inline, "max_inline=%s"},
336 {Opt_alloc_start, "alloc_start=%s"},
337 {Opt_thread_pool, "thread_pool=%d"},
338 {Opt_compress, "compress"},
339 {Opt_compress_type, "compress=%s"},
340 {Opt_compress_force, "compress-force"},
341 {Opt_compress_force_type, "compress-force=%s"},
342 {Opt_ssd, "ssd"},
343 {Opt_ssd_spread, "ssd_spread"},
344 {Opt_nossd, "nossd"},
345 {Opt_noacl, "noacl"},
346 {Opt_notreelog, "notreelog"},
347 {Opt_flushoncommit, "flushoncommit"},
348 {Opt_ratio, "metadata_ratio=%d"},
349 {Opt_discard, "discard"},
350 {Opt_space_cache, "space_cache"},
351 {Opt_clear_cache, "clear_cache"},
352 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
353 {Opt_enospc_debug, "enospc_debug"},
354 {Opt_subvolrootid, "subvolrootid=%d"},
355 {Opt_defrag, "autodefrag"},
356 {Opt_inode_cache, "inode_cache"},
357 {Opt_no_space_cache, "nospace_cache"},
358 {Opt_recovery, "recovery"},
359 {Opt_skip_balance, "skip_balance"},
360 {Opt_check_integrity, "check_int"},
361 {Opt_check_integrity_including_extent_data, "check_int_data"},
362 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
363 {Opt_fatal_errors, "fatal_errors=%s"},
364 {Opt_err, NULL},
365 };
366
367 /*
368 * Regular mount options parser. Everything that is needed only when
369 * reading in a new superblock is parsed here.
370 * XXX JDM: This needs to be cleaned up for remount.
371 */
372 int btrfs_parse_options(struct btrfs_root *root, char *options)
373 {
374 struct btrfs_fs_info *info = root->fs_info;
375 substring_t args[MAX_OPT_ARGS];
376 char *p, *num, *orig = NULL;
377 u64 cache_gen;
378 int intarg;
379 int ret = 0;
380 char *compress_type;
381 bool compress_force = false;
382
383 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
384 if (cache_gen)
385 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
386
387 if (!options)
388 goto out;
389
390 /*
391 * strsep changes the string, duplicate it because parse_options
392 * gets called twice
393 */
394 options = kstrdup(options, GFP_NOFS);
395 if (!options)
396 return -ENOMEM;
397
398 orig = options;
399
400 while ((p = strsep(&options, ",")) != NULL) {
401 int token;
402 if (!*p)
403 continue;
404
405 token = match_token(p, tokens, args);
406 switch (token) {
407 case Opt_degraded:
408 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
409 btrfs_set_opt(info->mount_opt, DEGRADED);
410 break;
411 case Opt_subvol:
412 case Opt_subvolid:
413 case Opt_subvolrootid:
414 case Opt_device:
415 /*
416 * These are parsed by btrfs_parse_early_options
417 * and can be happily ignored here.
418 */
419 break;
420 case Opt_nodatasum:
421 printk(KERN_INFO "btrfs: setting nodatasum\n");
422 btrfs_set_opt(info->mount_opt, NODATASUM);
423 break;
424 case Opt_nodatacow:
425 if (!btrfs_test_opt(root, COMPRESS) ||
426 !btrfs_test_opt(root, FORCE_COMPRESS)) {
427 printk(KERN_INFO "btrfs: setting nodatacow, compression disabled\n");
428 } else {
429 printk(KERN_INFO "btrfs: setting nodatacow\n");
430 }
431 info->compress_type = BTRFS_COMPRESS_NONE;
432 btrfs_clear_opt(info->mount_opt, COMPRESS);
433 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
434 btrfs_set_opt(info->mount_opt, NODATACOW);
435 btrfs_set_opt(info->mount_opt, NODATASUM);
436 break;
437 case Opt_compress_force:
438 case Opt_compress_force_type:
439 compress_force = true;
440 case Opt_compress:
441 case Opt_compress_type:
442 if (token == Opt_compress ||
443 token == Opt_compress_force ||
444 strcmp(args[0].from, "zlib") == 0) {
445 compress_type = "zlib";
446 info->compress_type = BTRFS_COMPRESS_ZLIB;
447 btrfs_set_opt(info->mount_opt, COMPRESS);
448 btrfs_clear_opt(info->mount_opt, NODATACOW);
449 btrfs_clear_opt(info->mount_opt, NODATASUM);
450 } else if (strcmp(args[0].from, "lzo") == 0) {
451 compress_type = "lzo";
452 info->compress_type = BTRFS_COMPRESS_LZO;
453 btrfs_set_opt(info->mount_opt, COMPRESS);
454 btrfs_clear_opt(info->mount_opt, NODATACOW);
455 btrfs_clear_opt(info->mount_opt, NODATASUM);
456 btrfs_set_fs_incompat(info, COMPRESS_LZO);
457 } else if (strncmp(args[0].from, "no", 2) == 0) {
458 compress_type = "no";
459 info->compress_type = BTRFS_COMPRESS_NONE;
460 btrfs_clear_opt(info->mount_opt, COMPRESS);
461 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
462 compress_force = false;
463 } else {
464 ret = -EINVAL;
465 goto out;
466 }
467
468 if (compress_force) {
469 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
470 pr_info("btrfs: force %s compression\n",
471 compress_type);
472 } else
473 pr_info("btrfs: use %s compression\n",
474 compress_type);
475 break;
476 case Opt_ssd:
477 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
478 btrfs_set_opt(info->mount_opt, SSD);
479 break;
480 case Opt_ssd_spread:
481 printk(KERN_INFO "btrfs: use spread ssd "
482 "allocation scheme\n");
483 btrfs_set_opt(info->mount_opt, SSD);
484 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
485 break;
486 case Opt_nossd:
487 printk(KERN_INFO "btrfs: not using ssd allocation "
488 "scheme\n");
489 btrfs_set_opt(info->mount_opt, NOSSD);
490 btrfs_clear_opt(info->mount_opt, SSD);
491 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
492 break;
493 case Opt_nobarrier:
494 printk(KERN_INFO "btrfs: turning off barriers\n");
495 btrfs_set_opt(info->mount_opt, NOBARRIER);
496 break;
497 case Opt_thread_pool:
498 intarg = 0;
499 match_int(&args[0], &intarg);
500 if (intarg)
501 info->thread_pool_size = intarg;
502 break;
503 case Opt_max_inline:
504 num = match_strdup(&args[0]);
505 if (num) {
506 info->max_inline = memparse(num, NULL);
507 kfree(num);
508
509 if (info->max_inline) {
510 info->max_inline = max_t(u64,
511 info->max_inline,
512 root->sectorsize);
513 }
514 printk(KERN_INFO "btrfs: max_inline at %llu\n",
515 (unsigned long long)info->max_inline);
516 }
517 break;
518 case Opt_alloc_start:
519 num = match_strdup(&args[0]);
520 if (num) {
521 mutex_lock(&info->chunk_mutex);
522 info->alloc_start = memparse(num, NULL);
523 mutex_unlock(&info->chunk_mutex);
524 kfree(num);
525 printk(KERN_INFO
526 "btrfs: allocations start at %llu\n",
527 (unsigned long long)info->alloc_start);
528 }
529 break;
530 case Opt_noacl:
531 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
532 break;
533 case Opt_notreelog:
534 printk(KERN_INFO "btrfs: disabling tree log\n");
535 btrfs_set_opt(info->mount_opt, NOTREELOG);
536 break;
537 case Opt_flushoncommit:
538 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
539 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
540 break;
541 case Opt_ratio:
542 intarg = 0;
543 match_int(&args[0], &intarg);
544 if (intarg) {
545 info->metadata_ratio = intarg;
546 printk(KERN_INFO "btrfs: metadata ratio %d\n",
547 info->metadata_ratio);
548 }
549 break;
550 case Opt_discard:
551 btrfs_set_opt(info->mount_opt, DISCARD);
552 break;
553 case Opt_space_cache:
554 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
555 break;
556 case Opt_no_space_cache:
557 printk(KERN_INFO "btrfs: disabling disk space caching\n");
558 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
559 break;
560 case Opt_inode_cache:
561 printk(KERN_INFO "btrfs: enabling inode map caching\n");
562 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
563 break;
564 case Opt_clear_cache:
565 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
566 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
567 break;
568 case Opt_user_subvol_rm_allowed:
569 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
570 break;
571 case Opt_enospc_debug:
572 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
573 break;
574 case Opt_defrag:
575 printk(KERN_INFO "btrfs: enabling auto defrag\n");
576 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
577 break;
578 case Opt_recovery:
579 printk(KERN_INFO "btrfs: enabling auto recovery\n");
580 btrfs_set_opt(info->mount_opt, RECOVERY);
581 break;
582 case Opt_skip_balance:
583 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
584 break;
585 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
586 case Opt_check_integrity_including_extent_data:
587 printk(KERN_INFO "btrfs: enabling check integrity"
588 " including extent data\n");
589 btrfs_set_opt(info->mount_opt,
590 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
591 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
592 break;
593 case Opt_check_integrity:
594 printk(KERN_INFO "btrfs: enabling check integrity\n");
595 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
596 break;
597 case Opt_check_integrity_print_mask:
598 intarg = 0;
599 match_int(&args[0], &intarg);
600 if (intarg) {
601 info->check_integrity_print_mask = intarg;
602 printk(KERN_INFO "btrfs:"
603 " check_integrity_print_mask 0x%x\n",
604 info->check_integrity_print_mask);
605 }
606 break;
607 #else
608 case Opt_check_integrity_including_extent_data:
609 case Opt_check_integrity:
610 case Opt_check_integrity_print_mask:
611 printk(KERN_ERR "btrfs: support for check_integrity*"
612 " not compiled in!\n");
613 ret = -EINVAL;
614 goto out;
615 #endif
616 case Opt_fatal_errors:
617 if (strcmp(args[0].from, "panic") == 0)
618 btrfs_set_opt(info->mount_opt,
619 PANIC_ON_FATAL_ERROR);
620 else if (strcmp(args[0].from, "bug") == 0)
621 btrfs_clear_opt(info->mount_opt,
622 PANIC_ON_FATAL_ERROR);
623 else {
624 ret = -EINVAL;
625 goto out;
626 }
627 break;
628 case Opt_err:
629 printk(KERN_INFO "btrfs: unrecognized mount option "
630 "'%s'\n", p);
631 ret = -EINVAL;
632 goto out;
633 default:
634 break;
635 }
636 }
637 out:
638 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
639 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
640 kfree(orig);
641 return ret;
642 }
643
644 /*
645 * Parse mount options that are required early in the mount process.
646 *
647 * All other options will be parsed on much later in the mount process and
648 * only when we need to allocate a new super block.
649 */
650 static int btrfs_parse_early_options(const char *options, fmode_t flags,
651 void *holder, char **subvol_name, u64 *subvol_objectid,
652 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
653 {
654 substring_t args[MAX_OPT_ARGS];
655 char *device_name, *opts, *orig, *p;
656 int error = 0;
657 int intarg;
658
659 if (!options)
660 return 0;
661
662 /*
663 * strsep changes the string, duplicate it because parse_options
664 * gets called twice
665 */
666 opts = kstrdup(options, GFP_KERNEL);
667 if (!opts)
668 return -ENOMEM;
669 orig = opts;
670
671 while ((p = strsep(&opts, ",")) != NULL) {
672 int token;
673 if (!*p)
674 continue;
675
676 token = match_token(p, tokens, args);
677 switch (token) {
678 case Opt_subvol:
679 kfree(*subvol_name);
680 *subvol_name = match_strdup(&args[0]);
681 break;
682 case Opt_subvolid:
683 intarg = 0;
684 error = match_int(&args[0], &intarg);
685 if (!error) {
686 /* we want the original fs_tree */
687 if (!intarg)
688 *subvol_objectid =
689 BTRFS_FS_TREE_OBJECTID;
690 else
691 *subvol_objectid = intarg;
692 }
693 break;
694 case Opt_subvolrootid:
695 intarg = 0;
696 error = match_int(&args[0], &intarg);
697 if (!error) {
698 /* we want the original fs_tree */
699 if (!intarg)
700 *subvol_rootid =
701 BTRFS_FS_TREE_OBJECTID;
702 else
703 *subvol_rootid = intarg;
704 }
705 break;
706 case Opt_device:
707 device_name = match_strdup(&args[0]);
708 if (!device_name) {
709 error = -ENOMEM;
710 goto out;
711 }
712 error = btrfs_scan_one_device(device_name,
713 flags, holder, fs_devices);
714 kfree(device_name);
715 if (error)
716 goto out;
717 break;
718 default:
719 break;
720 }
721 }
722
723 out:
724 kfree(orig);
725 return error;
726 }
727
728 static struct dentry *get_default_root(struct super_block *sb,
729 u64 subvol_objectid)
730 {
731 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
732 struct btrfs_root *root = fs_info->tree_root;
733 struct btrfs_root *new_root;
734 struct btrfs_dir_item *di;
735 struct btrfs_path *path;
736 struct btrfs_key location;
737 struct inode *inode;
738 u64 dir_id;
739 int new = 0;
740
741 /*
742 * We have a specific subvol we want to mount, just setup location and
743 * go look up the root.
744 */
745 if (subvol_objectid) {
746 location.objectid = subvol_objectid;
747 location.type = BTRFS_ROOT_ITEM_KEY;
748 location.offset = (u64)-1;
749 goto find_root;
750 }
751
752 path = btrfs_alloc_path();
753 if (!path)
754 return ERR_PTR(-ENOMEM);
755 path->leave_spinning = 1;
756
757 /*
758 * Find the "default" dir item which points to the root item that we
759 * will mount by default if we haven't been given a specific subvolume
760 * to mount.
761 */
762 dir_id = btrfs_super_root_dir(fs_info->super_copy);
763 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
764 if (IS_ERR(di)) {
765 btrfs_free_path(path);
766 return ERR_CAST(di);
767 }
768 if (!di) {
769 /*
770 * Ok the default dir item isn't there. This is weird since
771 * it's always been there, but don't freak out, just try and
772 * mount to root most subvolume.
773 */
774 btrfs_free_path(path);
775 dir_id = BTRFS_FIRST_FREE_OBJECTID;
776 new_root = fs_info->fs_root;
777 goto setup_root;
778 }
779
780 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
781 btrfs_free_path(path);
782
783 find_root:
784 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
785 if (IS_ERR(new_root))
786 return ERR_CAST(new_root);
787
788 if (btrfs_root_refs(&new_root->root_item) == 0)
789 return ERR_PTR(-ENOENT);
790
791 dir_id = btrfs_root_dirid(&new_root->root_item);
792 setup_root:
793 location.objectid = dir_id;
794 location.type = BTRFS_INODE_ITEM_KEY;
795 location.offset = 0;
796
797 inode = btrfs_iget(sb, &location, new_root, &new);
798 if (IS_ERR(inode))
799 return ERR_CAST(inode);
800
801 /*
802 * If we're just mounting the root most subvol put the inode and return
803 * a reference to the dentry. We will have already gotten a reference
804 * to the inode in btrfs_fill_super so we're good to go.
805 */
806 if (!new && sb->s_root->d_inode == inode) {
807 iput(inode);
808 return dget(sb->s_root);
809 }
810
811 return d_obtain_alias(inode);
812 }
813
814 static int btrfs_fill_super(struct super_block *sb,
815 struct btrfs_fs_devices *fs_devices,
816 void *data, int silent)
817 {
818 struct inode *inode;
819 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
820 struct btrfs_key key;
821 int err;
822
823 sb->s_maxbytes = MAX_LFS_FILESIZE;
824 sb->s_magic = BTRFS_SUPER_MAGIC;
825 sb->s_op = &btrfs_super_ops;
826 sb->s_d_op = &btrfs_dentry_operations;
827 sb->s_export_op = &btrfs_export_ops;
828 sb->s_xattr = btrfs_xattr_handlers;
829 sb->s_time_gran = 1;
830 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
831 sb->s_flags |= MS_POSIXACL;
832 #endif
833 sb->s_flags |= MS_I_VERSION;
834 err = open_ctree(sb, fs_devices, (char *)data);
835 if (err) {
836 printk("btrfs: open_ctree failed\n");
837 return err;
838 }
839
840 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
841 key.type = BTRFS_INODE_ITEM_KEY;
842 key.offset = 0;
843 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
844 if (IS_ERR(inode)) {
845 err = PTR_ERR(inode);
846 goto fail_close;
847 }
848
849 sb->s_root = d_make_root(inode);
850 if (!sb->s_root) {
851 err = -ENOMEM;
852 goto fail_close;
853 }
854
855 save_mount_options(sb, data);
856 cleancache_init_fs(sb);
857 sb->s_flags |= MS_ACTIVE;
858 return 0;
859
860 fail_close:
861 close_ctree(fs_info->tree_root);
862 return err;
863 }
864
865 int btrfs_sync_fs(struct super_block *sb, int wait)
866 {
867 struct btrfs_trans_handle *trans;
868 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
869 struct btrfs_root *root = fs_info->tree_root;
870
871 trace_btrfs_sync_fs(wait);
872
873 if (!wait) {
874 filemap_flush(fs_info->btree_inode->i_mapping);
875 return 0;
876 }
877
878 btrfs_wait_ordered_extents(root, 0);
879
880 trans = btrfs_attach_transaction(root);
881 if (IS_ERR(trans)) {
882 /* no transaction, don't bother */
883 if (PTR_ERR(trans) == -ENOENT)
884 return 0;
885 return PTR_ERR(trans);
886 }
887 return btrfs_commit_transaction(trans, root);
888 }
889
890 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
891 {
892 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
893 struct btrfs_root *root = info->tree_root;
894 char *compress_type;
895
896 if (btrfs_test_opt(root, DEGRADED))
897 seq_puts(seq, ",degraded");
898 if (btrfs_test_opt(root, NODATASUM))
899 seq_puts(seq, ",nodatasum");
900 if (btrfs_test_opt(root, NODATACOW))
901 seq_puts(seq, ",nodatacow");
902 if (btrfs_test_opt(root, NOBARRIER))
903 seq_puts(seq, ",nobarrier");
904 if (info->max_inline != 8192 * 1024)
905 seq_printf(seq, ",max_inline=%llu",
906 (unsigned long long)info->max_inline);
907 if (info->alloc_start != 0)
908 seq_printf(seq, ",alloc_start=%llu",
909 (unsigned long long)info->alloc_start);
910 if (info->thread_pool_size != min_t(unsigned long,
911 num_online_cpus() + 2, 8))
912 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
913 if (btrfs_test_opt(root, COMPRESS)) {
914 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
915 compress_type = "zlib";
916 else
917 compress_type = "lzo";
918 if (btrfs_test_opt(root, FORCE_COMPRESS))
919 seq_printf(seq, ",compress-force=%s", compress_type);
920 else
921 seq_printf(seq, ",compress=%s", compress_type);
922 }
923 if (btrfs_test_opt(root, NOSSD))
924 seq_puts(seq, ",nossd");
925 if (btrfs_test_opt(root, SSD_SPREAD))
926 seq_puts(seq, ",ssd_spread");
927 else if (btrfs_test_opt(root, SSD))
928 seq_puts(seq, ",ssd");
929 if (btrfs_test_opt(root, NOTREELOG))
930 seq_puts(seq, ",notreelog");
931 if (btrfs_test_opt(root, FLUSHONCOMMIT))
932 seq_puts(seq, ",flushoncommit");
933 if (btrfs_test_opt(root, DISCARD))
934 seq_puts(seq, ",discard");
935 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
936 seq_puts(seq, ",noacl");
937 if (btrfs_test_opt(root, SPACE_CACHE))
938 seq_puts(seq, ",space_cache");
939 else
940 seq_puts(seq, ",nospace_cache");
941 if (btrfs_test_opt(root, CLEAR_CACHE))
942 seq_puts(seq, ",clear_cache");
943 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
944 seq_puts(seq, ",user_subvol_rm_allowed");
945 if (btrfs_test_opt(root, ENOSPC_DEBUG))
946 seq_puts(seq, ",enospc_debug");
947 if (btrfs_test_opt(root, AUTO_DEFRAG))
948 seq_puts(seq, ",autodefrag");
949 if (btrfs_test_opt(root, INODE_MAP_CACHE))
950 seq_puts(seq, ",inode_cache");
951 if (btrfs_test_opt(root, SKIP_BALANCE))
952 seq_puts(seq, ",skip_balance");
953 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
954 seq_puts(seq, ",fatal_errors=panic");
955 return 0;
956 }
957
958 static int btrfs_test_super(struct super_block *s, void *data)
959 {
960 struct btrfs_fs_info *p = data;
961 struct btrfs_fs_info *fs_info = btrfs_sb(s);
962
963 return fs_info->fs_devices == p->fs_devices;
964 }
965
966 static int btrfs_set_super(struct super_block *s, void *data)
967 {
968 int err = set_anon_super(s, data);
969 if (!err)
970 s->s_fs_info = data;
971 return err;
972 }
973
974 /*
975 * subvolumes are identified by ino 256
976 */
977 static inline int is_subvolume_inode(struct inode *inode)
978 {
979 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
980 return 1;
981 return 0;
982 }
983
984 /*
985 * This will strip out the subvol=%s argument for an argument string and add
986 * subvolid=0 to make sure we get the actual tree root for path walking to the
987 * subvol we want.
988 */
989 static char *setup_root_args(char *args)
990 {
991 unsigned len = strlen(args) + 2 + 1;
992 char *src, *dst, *buf;
993
994 /*
995 * We need the same args as before, but with this substitution:
996 * s!subvol=[^,]+!subvolid=0!
997 *
998 * Since the replacement string is up to 2 bytes longer than the
999 * original, allocate strlen(args) + 2 + 1 bytes.
1000 */
1001
1002 src = strstr(args, "subvol=");
1003 /* This shouldn't happen, but just in case.. */
1004 if (!src)
1005 return NULL;
1006
1007 buf = dst = kmalloc(len, GFP_NOFS);
1008 if (!buf)
1009 return NULL;
1010
1011 /*
1012 * If the subvol= arg is not at the start of the string,
1013 * copy whatever precedes it into buf.
1014 */
1015 if (src != args) {
1016 *src++ = '\0';
1017 strcpy(buf, args);
1018 dst += strlen(args);
1019 }
1020
1021 strcpy(dst, "subvolid=0");
1022 dst += strlen("subvolid=0");
1023
1024 /*
1025 * If there is a "," after the original subvol=... string,
1026 * copy that suffix into our buffer. Otherwise, we're done.
1027 */
1028 src = strchr(src, ',');
1029 if (src)
1030 strcpy(dst, src);
1031
1032 return buf;
1033 }
1034
1035 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1036 const char *device_name, char *data)
1037 {
1038 struct dentry *root;
1039 struct vfsmount *mnt;
1040 char *newargs;
1041
1042 newargs = setup_root_args(data);
1043 if (!newargs)
1044 return ERR_PTR(-ENOMEM);
1045 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1046 newargs);
1047 kfree(newargs);
1048 if (IS_ERR(mnt))
1049 return ERR_CAST(mnt);
1050
1051 root = mount_subtree(mnt, subvol_name);
1052
1053 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1054 struct super_block *s = root->d_sb;
1055 dput(root);
1056 root = ERR_PTR(-EINVAL);
1057 deactivate_locked_super(s);
1058 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
1059 subvol_name);
1060 }
1061
1062 return root;
1063 }
1064
1065 /*
1066 * Find a superblock for the given device / mount point.
1067 *
1068 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1069 * for multiple device setup. Make sure to keep it in sync.
1070 */
1071 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1072 const char *device_name, void *data)
1073 {
1074 struct block_device *bdev = NULL;
1075 struct super_block *s;
1076 struct dentry *root;
1077 struct btrfs_fs_devices *fs_devices = NULL;
1078 struct btrfs_fs_info *fs_info = NULL;
1079 fmode_t mode = FMODE_READ;
1080 char *subvol_name = NULL;
1081 u64 subvol_objectid = 0;
1082 u64 subvol_rootid = 0;
1083 int error = 0;
1084
1085 if (!(flags & MS_RDONLY))
1086 mode |= FMODE_WRITE;
1087
1088 error = btrfs_parse_early_options(data, mode, fs_type,
1089 &subvol_name, &subvol_objectid,
1090 &subvol_rootid, &fs_devices);
1091 if (error) {
1092 kfree(subvol_name);
1093 return ERR_PTR(error);
1094 }
1095
1096 if (subvol_name) {
1097 root = mount_subvol(subvol_name, flags, device_name, data);
1098 kfree(subvol_name);
1099 return root;
1100 }
1101
1102 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1103 if (error)
1104 return ERR_PTR(error);
1105
1106 /*
1107 * Setup a dummy root and fs_info for test/set super. This is because
1108 * we don't actually fill this stuff out until open_ctree, but we need
1109 * it for searching for existing supers, so this lets us do that and
1110 * then open_ctree will properly initialize everything later.
1111 */
1112 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1113 if (!fs_info)
1114 return ERR_PTR(-ENOMEM);
1115
1116 fs_info->fs_devices = fs_devices;
1117
1118 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1119 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1120 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1121 error = -ENOMEM;
1122 goto error_fs_info;
1123 }
1124
1125 error = btrfs_open_devices(fs_devices, mode, fs_type);
1126 if (error)
1127 goto error_fs_info;
1128
1129 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1130 error = -EACCES;
1131 goto error_close_devices;
1132 }
1133
1134 bdev = fs_devices->latest_bdev;
1135 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1136 fs_info);
1137 if (IS_ERR(s)) {
1138 error = PTR_ERR(s);
1139 goto error_close_devices;
1140 }
1141
1142 if (s->s_root) {
1143 btrfs_close_devices(fs_devices);
1144 free_fs_info(fs_info);
1145 if ((flags ^ s->s_flags) & MS_RDONLY)
1146 error = -EBUSY;
1147 } else {
1148 char b[BDEVNAME_SIZE];
1149
1150 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1151 btrfs_sb(s)->bdev_holder = fs_type;
1152 error = btrfs_fill_super(s, fs_devices, data,
1153 flags & MS_SILENT ? 1 : 0);
1154 }
1155
1156 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1157 if (IS_ERR(root))
1158 deactivate_locked_super(s);
1159
1160 return root;
1161
1162 error_close_devices:
1163 btrfs_close_devices(fs_devices);
1164 error_fs_info:
1165 free_fs_info(fs_info);
1166 return ERR_PTR(error);
1167 }
1168
1169 static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1170 {
1171 spin_lock_irq(&workers->lock);
1172 workers->max_workers = new_limit;
1173 spin_unlock_irq(&workers->lock);
1174 }
1175
1176 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1177 int new_pool_size, int old_pool_size)
1178 {
1179 if (new_pool_size == old_pool_size)
1180 return;
1181
1182 fs_info->thread_pool_size = new_pool_size;
1183
1184 printk(KERN_INFO "btrfs: resize thread pool %d -> %d\n",
1185 old_pool_size, new_pool_size);
1186
1187 btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1188 btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1189 btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1190 btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1191 btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1192 btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1193 btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1194 btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1195 btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1196 btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1197 btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1198 btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1199 btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
1200 btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers,
1201 new_pool_size);
1202 }
1203
1204 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1205 {
1206 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1207 struct btrfs_root *root = fs_info->tree_root;
1208 unsigned old_flags = sb->s_flags;
1209 unsigned long old_opts = fs_info->mount_opt;
1210 unsigned long old_compress_type = fs_info->compress_type;
1211 u64 old_max_inline = fs_info->max_inline;
1212 u64 old_alloc_start = fs_info->alloc_start;
1213 int old_thread_pool_size = fs_info->thread_pool_size;
1214 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1215 int ret;
1216
1217 ret = btrfs_parse_options(root, data);
1218 if (ret) {
1219 ret = -EINVAL;
1220 goto restore;
1221 }
1222
1223 btrfs_resize_thread_pool(fs_info,
1224 fs_info->thread_pool_size, old_thread_pool_size);
1225
1226 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1227 return 0;
1228
1229 if (*flags & MS_RDONLY) {
1230 /*
1231 * this also happens on 'umount -rf' or on shutdown, when
1232 * the filesystem is busy.
1233 */
1234 sb->s_flags |= MS_RDONLY;
1235
1236 btrfs_dev_replace_suspend_for_unmount(fs_info);
1237 btrfs_scrub_cancel(fs_info);
1238
1239 ret = btrfs_commit_super(root);
1240 if (ret)
1241 goto restore;
1242 } else {
1243 if (fs_info->fs_devices->rw_devices == 0) {
1244 ret = -EACCES;
1245 goto restore;
1246 }
1247
1248 if (fs_info->fs_devices->missing_devices >
1249 fs_info->num_tolerated_disk_barrier_failures &&
1250 !(*flags & MS_RDONLY)) {
1251 printk(KERN_WARNING
1252 "Btrfs: too many missing devices, writeable remount is not allowed\n");
1253 ret = -EACCES;
1254 goto restore;
1255 }
1256
1257 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1258 ret = -EINVAL;
1259 goto restore;
1260 }
1261
1262 ret = btrfs_cleanup_fs_roots(fs_info);
1263 if (ret)
1264 goto restore;
1265
1266 /* recover relocation */
1267 ret = btrfs_recover_relocation(root);
1268 if (ret)
1269 goto restore;
1270
1271 ret = btrfs_resume_balance_async(fs_info);
1272 if (ret)
1273 goto restore;
1274
1275 ret = btrfs_resume_dev_replace_async(fs_info);
1276 if (ret) {
1277 pr_warn("btrfs: failed to resume dev_replace\n");
1278 goto restore;
1279 }
1280 sb->s_flags &= ~MS_RDONLY;
1281 }
1282
1283 return 0;
1284
1285 restore:
1286 /* We've hit an error - don't reset MS_RDONLY */
1287 if (sb->s_flags & MS_RDONLY)
1288 old_flags |= MS_RDONLY;
1289 sb->s_flags = old_flags;
1290 fs_info->mount_opt = old_opts;
1291 fs_info->compress_type = old_compress_type;
1292 fs_info->max_inline = old_max_inline;
1293 mutex_lock(&fs_info->chunk_mutex);
1294 fs_info->alloc_start = old_alloc_start;
1295 mutex_unlock(&fs_info->chunk_mutex);
1296 btrfs_resize_thread_pool(fs_info,
1297 old_thread_pool_size, fs_info->thread_pool_size);
1298 fs_info->metadata_ratio = old_metadata_ratio;
1299 return ret;
1300 }
1301
1302 /* Used to sort the devices by max_avail(descending sort) */
1303 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1304 const void *dev_info2)
1305 {
1306 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1307 ((struct btrfs_device_info *)dev_info2)->max_avail)
1308 return -1;
1309 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1310 ((struct btrfs_device_info *)dev_info2)->max_avail)
1311 return 1;
1312 else
1313 return 0;
1314 }
1315
1316 /*
1317 * sort the devices by max_avail, in which max free extent size of each device
1318 * is stored.(Descending Sort)
1319 */
1320 static inline void btrfs_descending_sort_devices(
1321 struct btrfs_device_info *devices,
1322 size_t nr_devices)
1323 {
1324 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1325 btrfs_cmp_device_free_bytes, NULL);
1326 }
1327
1328 /*
1329 * The helper to calc the free space on the devices that can be used to store
1330 * file data.
1331 */
1332 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1333 {
1334 struct btrfs_fs_info *fs_info = root->fs_info;
1335 struct btrfs_device_info *devices_info;
1336 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1337 struct btrfs_device *device;
1338 u64 skip_space;
1339 u64 type;
1340 u64 avail_space;
1341 u64 used_space;
1342 u64 min_stripe_size;
1343 int min_stripes = 1, num_stripes = 1;
1344 int i = 0, nr_devices;
1345 int ret;
1346
1347 nr_devices = fs_info->fs_devices->open_devices;
1348 BUG_ON(!nr_devices);
1349
1350 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1351 GFP_NOFS);
1352 if (!devices_info)
1353 return -ENOMEM;
1354
1355 /* calc min stripe number for data space alloction */
1356 type = btrfs_get_alloc_profile(root, 1);
1357 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1358 min_stripes = 2;
1359 num_stripes = nr_devices;
1360 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1361 min_stripes = 2;
1362 num_stripes = 2;
1363 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1364 min_stripes = 4;
1365 num_stripes = 4;
1366 }
1367
1368 if (type & BTRFS_BLOCK_GROUP_DUP)
1369 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1370 else
1371 min_stripe_size = BTRFS_STRIPE_LEN;
1372
1373 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1374 if (!device->in_fs_metadata || !device->bdev ||
1375 device->is_tgtdev_for_dev_replace)
1376 continue;
1377
1378 avail_space = device->total_bytes - device->bytes_used;
1379
1380 /* align with stripe_len */
1381 do_div(avail_space, BTRFS_STRIPE_LEN);
1382 avail_space *= BTRFS_STRIPE_LEN;
1383
1384 /*
1385 * In order to avoid overwritting the superblock on the drive,
1386 * btrfs starts at an offset of at least 1MB when doing chunk
1387 * allocation.
1388 */
1389 skip_space = 1024 * 1024;
1390
1391 /* user can set the offset in fs_info->alloc_start. */
1392 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1393 device->total_bytes)
1394 skip_space = max(fs_info->alloc_start, skip_space);
1395
1396 /*
1397 * btrfs can not use the free space in [0, skip_space - 1],
1398 * we must subtract it from the total. In order to implement
1399 * it, we account the used space in this range first.
1400 */
1401 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1402 &used_space);
1403 if (ret) {
1404 kfree(devices_info);
1405 return ret;
1406 }
1407
1408 /* calc the free space in [0, skip_space - 1] */
1409 skip_space -= used_space;
1410
1411 /*
1412 * we can use the free space in [0, skip_space - 1], subtract
1413 * it from the total.
1414 */
1415 if (avail_space && avail_space >= skip_space)
1416 avail_space -= skip_space;
1417 else
1418 avail_space = 0;
1419
1420 if (avail_space < min_stripe_size)
1421 continue;
1422
1423 devices_info[i].dev = device;
1424 devices_info[i].max_avail = avail_space;
1425
1426 i++;
1427 }
1428
1429 nr_devices = i;
1430
1431 btrfs_descending_sort_devices(devices_info, nr_devices);
1432
1433 i = nr_devices - 1;
1434 avail_space = 0;
1435 while (nr_devices >= min_stripes) {
1436 if (num_stripes > nr_devices)
1437 num_stripes = nr_devices;
1438
1439 if (devices_info[i].max_avail >= min_stripe_size) {
1440 int j;
1441 u64 alloc_size;
1442
1443 avail_space += devices_info[i].max_avail * num_stripes;
1444 alloc_size = devices_info[i].max_avail;
1445 for (j = i + 1 - num_stripes; j <= i; j++)
1446 devices_info[j].max_avail -= alloc_size;
1447 }
1448 i--;
1449 nr_devices--;
1450 }
1451
1452 kfree(devices_info);
1453 *free_bytes = avail_space;
1454 return 0;
1455 }
1456
1457 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1458 {
1459 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1460 struct btrfs_super_block *disk_super = fs_info->super_copy;
1461 struct list_head *head = &fs_info->space_info;
1462 struct btrfs_space_info *found;
1463 u64 total_used = 0;
1464 u64 total_free_data = 0;
1465 int bits = dentry->d_sb->s_blocksize_bits;
1466 __be32 *fsid = (__be32 *)fs_info->fsid;
1467 int ret;
1468
1469 /* holding chunk_muext to avoid allocating new chunks */
1470 mutex_lock(&fs_info->chunk_mutex);
1471 rcu_read_lock();
1472 list_for_each_entry_rcu(found, head, list) {
1473 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1474 total_free_data += found->disk_total - found->disk_used;
1475 total_free_data -=
1476 btrfs_account_ro_block_groups_free_space(found);
1477 }
1478
1479 total_used += found->disk_used;
1480 }
1481 rcu_read_unlock();
1482
1483 buf->f_namelen = BTRFS_NAME_LEN;
1484 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1485 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1486 buf->f_bsize = dentry->d_sb->s_blocksize;
1487 buf->f_type = BTRFS_SUPER_MAGIC;
1488 buf->f_bavail = total_free_data;
1489 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1490 if (ret) {
1491 mutex_unlock(&fs_info->chunk_mutex);
1492 return ret;
1493 }
1494 buf->f_bavail += total_free_data;
1495 buf->f_bavail = buf->f_bavail >> bits;
1496 mutex_unlock(&fs_info->chunk_mutex);
1497
1498 /* We treat it as constant endianness (it doesn't matter _which_)
1499 because we want the fsid to come out the same whether mounted
1500 on a big-endian or little-endian host */
1501 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1502 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1503 /* Mask in the root object ID too, to disambiguate subvols */
1504 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1505 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1506
1507 return 0;
1508 }
1509
1510 static void btrfs_kill_super(struct super_block *sb)
1511 {
1512 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1513 kill_anon_super(sb);
1514 free_fs_info(fs_info);
1515 }
1516
1517 static struct file_system_type btrfs_fs_type = {
1518 .owner = THIS_MODULE,
1519 .name = "btrfs",
1520 .mount = btrfs_mount,
1521 .kill_sb = btrfs_kill_super,
1522 .fs_flags = FS_REQUIRES_DEV,
1523 };
1524
1525 /*
1526 * used by btrfsctl to scan devices when no FS is mounted
1527 */
1528 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1529 unsigned long arg)
1530 {
1531 struct btrfs_ioctl_vol_args *vol;
1532 struct btrfs_fs_devices *fs_devices;
1533 int ret = -ENOTTY;
1534
1535 if (!capable(CAP_SYS_ADMIN))
1536 return -EPERM;
1537
1538 vol = memdup_user((void __user *)arg, sizeof(*vol));
1539 if (IS_ERR(vol))
1540 return PTR_ERR(vol);
1541
1542 switch (cmd) {
1543 case BTRFS_IOC_SCAN_DEV:
1544 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1545 &btrfs_fs_type, &fs_devices);
1546 break;
1547 case BTRFS_IOC_DEVICES_READY:
1548 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1549 &btrfs_fs_type, &fs_devices);
1550 if (ret)
1551 break;
1552 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1553 break;
1554 }
1555
1556 kfree(vol);
1557 return ret;
1558 }
1559
1560 static int btrfs_freeze(struct super_block *sb)
1561 {
1562 struct btrfs_trans_handle *trans;
1563 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1564
1565 trans = btrfs_attach_transaction(root);
1566 if (IS_ERR(trans)) {
1567 /* no transaction, don't bother */
1568 if (PTR_ERR(trans) == -ENOENT)
1569 return 0;
1570 return PTR_ERR(trans);
1571 }
1572 return btrfs_commit_transaction(trans, root);
1573 }
1574
1575 static int btrfs_unfreeze(struct super_block *sb)
1576 {
1577 return 0;
1578 }
1579
1580 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1581 {
1582 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1583 struct btrfs_fs_devices *cur_devices;
1584 struct btrfs_device *dev, *first_dev = NULL;
1585 struct list_head *head;
1586 struct rcu_string *name;
1587
1588 mutex_lock(&fs_info->fs_devices->device_list_mutex);
1589 cur_devices = fs_info->fs_devices;
1590 while (cur_devices) {
1591 head = &cur_devices->devices;
1592 list_for_each_entry(dev, head, dev_list) {
1593 if (dev->missing)
1594 continue;
1595 if (!first_dev || dev->devid < first_dev->devid)
1596 first_dev = dev;
1597 }
1598 cur_devices = cur_devices->seed;
1599 }
1600
1601 if (first_dev) {
1602 rcu_read_lock();
1603 name = rcu_dereference(first_dev->name);
1604 seq_escape(m, name->str, " \t\n\\");
1605 rcu_read_unlock();
1606 } else {
1607 WARN_ON(1);
1608 }
1609 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1610 return 0;
1611 }
1612
1613 static const struct super_operations btrfs_super_ops = {
1614 .drop_inode = btrfs_drop_inode,
1615 .evict_inode = btrfs_evict_inode,
1616 .put_super = btrfs_put_super,
1617 .sync_fs = btrfs_sync_fs,
1618 .show_options = btrfs_show_options,
1619 .show_devname = btrfs_show_devname,
1620 .write_inode = btrfs_write_inode,
1621 .alloc_inode = btrfs_alloc_inode,
1622 .destroy_inode = btrfs_destroy_inode,
1623 .statfs = btrfs_statfs,
1624 .remount_fs = btrfs_remount,
1625 .freeze_fs = btrfs_freeze,
1626 .unfreeze_fs = btrfs_unfreeze,
1627 };
1628
1629 static const struct file_operations btrfs_ctl_fops = {
1630 .unlocked_ioctl = btrfs_control_ioctl,
1631 .compat_ioctl = btrfs_control_ioctl,
1632 .owner = THIS_MODULE,
1633 .llseek = noop_llseek,
1634 };
1635
1636 static struct miscdevice btrfs_misc = {
1637 .minor = BTRFS_MINOR,
1638 .name = "btrfs-control",
1639 .fops = &btrfs_ctl_fops
1640 };
1641
1642 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1643 MODULE_ALIAS("devname:btrfs-control");
1644
1645 static int btrfs_interface_init(void)
1646 {
1647 return misc_register(&btrfs_misc);
1648 }
1649
1650 static void btrfs_interface_exit(void)
1651 {
1652 if (misc_deregister(&btrfs_misc) < 0)
1653 printk(KERN_INFO "btrfs: misc_deregister failed for control device\n");
1654 }
1655
1656 static int __init init_btrfs_fs(void)
1657 {
1658 int err;
1659
1660 err = btrfs_init_sysfs();
1661 if (err)
1662 return err;
1663
1664 btrfs_init_compress();
1665
1666 err = btrfs_init_cachep();
1667 if (err)
1668 goto free_compress;
1669
1670 err = extent_io_init();
1671 if (err)
1672 goto free_cachep;
1673
1674 err = extent_map_init();
1675 if (err)
1676 goto free_extent_io;
1677
1678 err = ordered_data_init();
1679 if (err)
1680 goto free_extent_map;
1681
1682 err = btrfs_delayed_inode_init();
1683 if (err)
1684 goto free_ordered_data;
1685
1686 err = btrfs_auto_defrag_init();
1687 if (err)
1688 goto free_delayed_inode;
1689
1690 err = btrfs_delayed_ref_init();
1691 if (err)
1692 goto free_auto_defrag;
1693
1694 err = btrfs_interface_init();
1695 if (err)
1696 goto free_delayed_ref;
1697
1698 err = register_filesystem(&btrfs_fs_type);
1699 if (err)
1700 goto unregister_ioctl;
1701
1702 btrfs_init_lockdep();
1703
1704 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1705 return 0;
1706
1707 unregister_ioctl:
1708 btrfs_interface_exit();
1709 free_delayed_ref:
1710 btrfs_delayed_ref_exit();
1711 free_auto_defrag:
1712 btrfs_auto_defrag_exit();
1713 free_delayed_inode:
1714 btrfs_delayed_inode_exit();
1715 free_ordered_data:
1716 ordered_data_exit();
1717 free_extent_map:
1718 extent_map_exit();
1719 free_extent_io:
1720 extent_io_exit();
1721 free_cachep:
1722 btrfs_destroy_cachep();
1723 free_compress:
1724 btrfs_exit_compress();
1725 btrfs_exit_sysfs();
1726 return err;
1727 }
1728
1729 static void __exit exit_btrfs_fs(void)
1730 {
1731 btrfs_destroy_cachep();
1732 btrfs_delayed_ref_exit();
1733 btrfs_auto_defrag_exit();
1734 btrfs_delayed_inode_exit();
1735 ordered_data_exit();
1736 extent_map_exit();
1737 extent_io_exit();
1738 btrfs_interface_exit();
1739 unregister_filesystem(&btrfs_fs_type);
1740 btrfs_exit_sysfs();
1741 btrfs_cleanup_fs_uuids();
1742 btrfs_exit_compress();
1743 }
1744
1745 module_init(init_btrfs_fs)
1746 module_exit(exit_btrfs_fs)
1747
1748 MODULE_LICENSE("GPL");
Linux kernel - Deliverables
This page took 0.062944 seconds and 4 git commands to generate.