]> Pileus Git - ~andy/linux/blob - fs/btrfs/super.c
Btrfs: convert printk to btrfs_ and fix BTRFS prefix
[~andy/linux] / 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 "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 "xattr.h"
52 #include "volumes.h"
53 #include "export.h"
54 #include "compression.h"
55 #include "rcu-string.h"
56 #include "dev-replace.h"
57 #include "free-space-cache.h"
58 #include "backref.h"
59 #include "tests/btrfs-tests.h"
60
61 #define CREATE_TRACE_POINTS
62 #include <trace/events/btrfs.h>
63
64 static const struct super_operations btrfs_super_ops;
65 static struct file_system_type btrfs_fs_type;
66
67 static const char *btrfs_decode_error(int errno)
68 {
69         char *errstr = "unknown";
70
71         switch (errno) {
72         case -EIO:
73                 errstr = "IO failure";
74                 break;
75         case -ENOMEM:
76                 errstr = "Out of memory";
77                 break;
78         case -EROFS:
79                 errstr = "Readonly filesystem";
80                 break;
81         case -EEXIST:
82                 errstr = "Object already exists";
83                 break;
84         case -ENOSPC:
85                 errstr = "No space left";
86                 break;
87         case -ENOENT:
88                 errstr = "No such entry";
89                 break;
90         }
91
92         return errstr;
93 }
94
95 static void save_error_info(struct btrfs_fs_info *fs_info)
96 {
97         /*
98          * today we only save the error info into ram.  Long term we'll
99          * also send it down to the disk
100          */
101         set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
102 }
103
104 /* btrfs handle error by forcing the filesystem readonly */
105 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
106 {
107         struct super_block *sb = fs_info->sb;
108
109         if (sb->s_flags & MS_RDONLY)
110                 return;
111
112         if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
113                 sb->s_flags |= MS_RDONLY;
114                 btrfs_info(fs_info, "forced readonly");
115                 /*
116                  * Note that a running device replace operation is not
117                  * canceled here although there is no way to update
118                  * the progress. It would add the risk of a deadlock,
119                  * therefore the canceling is ommited. The only penalty
120                  * is that some I/O remains active until the procedure
121                  * completes. The next time when the filesystem is
122                  * mounted writeable again, the device replace
123                  * operation continues.
124                  */
125         }
126 }
127
128 #ifdef CONFIG_PRINTK
129 /*
130  * __btrfs_std_error decodes expected errors from the caller and
131  * invokes the approciate error response.
132  */
133 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
134                        unsigned int line, int errno, const char *fmt, ...)
135 {
136         struct super_block *sb = fs_info->sb;
137         const char *errstr;
138
139         /*
140          * Special case: if the error is EROFS, and we're already
141          * under MS_RDONLY, then it is safe here.
142          */
143         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
144                 return;
145
146         errstr = btrfs_decode_error(errno);
147         if (fmt) {
148                 struct va_format vaf;
149                 va_list args;
150
151                 va_start(args, fmt);
152                 vaf.fmt = fmt;
153                 vaf.va = &args;
154
155                 printk(KERN_CRIT
156                         "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
157                         sb->s_id, function, line, errno, errstr, &vaf);
158                 va_end(args);
159         } else {
160                 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
161                         sb->s_id, function, line, errno, errstr);
162         }
163
164         /* Don't go through full error handling during mount */
165         save_error_info(fs_info);
166         if (sb->s_flags & MS_BORN)
167                 btrfs_handle_error(fs_info);
168 }
169
170 static const char * const logtypes[] = {
171         "emergency",
172         "alert",
173         "critical",
174         "error",
175         "warning",
176         "notice",
177         "info",
178         "debug",
179 };
180
181 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
182 {
183         struct super_block *sb = fs_info->sb;
184         char lvl[4];
185         struct va_format vaf;
186         va_list args;
187         const char *type = logtypes[4];
188         int kern_level;
189
190         va_start(args, fmt);
191
192         kern_level = printk_get_level(fmt);
193         if (kern_level) {
194                 size_t size = printk_skip_level(fmt) - fmt;
195                 memcpy(lvl, fmt,  size);
196                 lvl[size] = '\0';
197                 fmt += size;
198                 type = logtypes[kern_level - '0'];
199         } else
200                 *lvl = '\0';
201
202         vaf.fmt = fmt;
203         vaf.va = &args;
204
205         printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
206
207         va_end(args);
208 }
209
210 #else
211
212 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
213                        unsigned int line, int errno, const char *fmt, ...)
214 {
215         struct super_block *sb = fs_info->sb;
216
217         /*
218          * Special case: if the error is EROFS, and we're already
219          * under MS_RDONLY, then it is safe here.
220          */
221         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
222                 return;
223
224         /* Don't go through full error handling during mount */
225         if (sb->s_flags & MS_BORN) {
226                 save_error_info(fs_info);
227                 btrfs_handle_error(fs_info);
228         }
229 }
230 #endif
231
232 /*
233  * We only mark the transaction aborted and then set the file system read-only.
234  * This will prevent new transactions from starting or trying to join this
235  * one.
236  *
237  * This means that error recovery at the call site is limited to freeing
238  * any local memory allocations and passing the error code up without
239  * further cleanup. The transaction should complete as it normally would
240  * in the call path but will return -EIO.
241  *
242  * We'll complete the cleanup in btrfs_end_transaction and
243  * btrfs_commit_transaction.
244  */
245 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
246                                struct btrfs_root *root, const char *function,
247                                unsigned int line, int errno)
248 {
249         /*
250          * Report first abort since mount
251          */
252         if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,
253                                 &root->fs_info->fs_state)) {
254                 WARN(1, KERN_DEBUG "BTRFS: Transaction aborted (error %d)\n",
255                                 errno);
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->blocks_used) {
261                 const char *errstr;
262
263                 errstr = btrfs_decode_error(errno);
264                 btrfs_warn(root->fs_info,
265                            "%s:%d: Aborting unused transaction(%s).",
266                            function, line, errstr);
267                 return;
268         }
269         ACCESS_ONCE(trans->transaction->aborted) = errno;
270         /* Wake up anybody who may be waiting on this transaction */
271         wake_up(&root->fs_info->transaction_wait);
272         wake_up(&root->fs_info->transaction_blocked_wait);
273         __btrfs_std_error(root->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 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
280                    unsigned int line, int errno, const char *fmt, ...)
281 {
282         char *s_id = "<unknown>";
283         const char *errstr;
284         struct va_format vaf = { .fmt = fmt };
285         va_list args;
286
287         if (fs_info)
288                 s_id = fs_info->sb->s_id;
289
290         va_start(args, fmt);
291         vaf.va = &args;
292
293         errstr = btrfs_decode_error(errno);
294         if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
295                 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
296                         s_id, function, line, &vaf, errno, errstr);
297
298         btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
299                    function, line, &vaf, errno, errstr);
300         va_end(args);
301         /* Caller calls BUG() */
302 }
303
304 static void btrfs_put_super(struct super_block *sb)
305 {
306         (void)close_ctree(btrfs_sb(sb)->tree_root);
307         /* FIXME: need to fix VFS to return error? */
308         /* AV: return it _where_?  ->put_super() can be triggered by any number
309          * of async events, up to and including delivery of SIGKILL to the
310          * last process that kept it busy.  Or segfault in the aforementioned
311          * process...  Whom would you report that to?
312          */
313 }
314
315 enum {
316         Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
317         Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
318         Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
319         Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
320         Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
321         Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
322         Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
323         Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
324         Opt_check_integrity, Opt_check_integrity_including_extent_data,
325         Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
326         Opt_commit_interval,
327         Opt_err,
328 };
329
330 static match_table_t tokens = {
331         {Opt_degraded, "degraded"},
332         {Opt_subvol, "subvol=%s"},
333         {Opt_subvolid, "subvolid=%s"},
334         {Opt_device, "device=%s"},
335         {Opt_nodatasum, "nodatasum"},
336         {Opt_nodatacow, "nodatacow"},
337         {Opt_nobarrier, "nobarrier"},
338         {Opt_max_inline, "max_inline=%s"},
339         {Opt_alloc_start, "alloc_start=%s"},
340         {Opt_thread_pool, "thread_pool=%d"},
341         {Opt_compress, "compress"},
342         {Opt_compress_type, "compress=%s"},
343         {Opt_compress_force, "compress-force"},
344         {Opt_compress_force_type, "compress-force=%s"},
345         {Opt_ssd, "ssd"},
346         {Opt_ssd_spread, "ssd_spread"},
347         {Opt_nossd, "nossd"},
348         {Opt_noacl, "noacl"},
349         {Opt_notreelog, "notreelog"},
350         {Opt_flushoncommit, "flushoncommit"},
351         {Opt_ratio, "metadata_ratio=%d"},
352         {Opt_discard, "discard"},
353         {Opt_space_cache, "space_cache"},
354         {Opt_clear_cache, "clear_cache"},
355         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
356         {Opt_enospc_debug, "enospc_debug"},
357         {Opt_subvolrootid, "subvolrootid=%d"},
358         {Opt_defrag, "autodefrag"},
359         {Opt_inode_cache, "inode_cache"},
360         {Opt_no_space_cache, "nospace_cache"},
361         {Opt_recovery, "recovery"},
362         {Opt_skip_balance, "skip_balance"},
363         {Opt_check_integrity, "check_int"},
364         {Opt_check_integrity_including_extent_data, "check_int_data"},
365         {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
366         {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
367         {Opt_fatal_errors, "fatal_errors=%s"},
368         {Opt_commit_interval, "commit=%d"},
369         {Opt_err, NULL},
370 };
371
372 /*
373  * Regular mount options parser.  Everything that is needed only when
374  * reading in a new superblock is parsed here.
375  * XXX JDM: This needs to be cleaned up for remount.
376  */
377 int btrfs_parse_options(struct btrfs_root *root, char *options)
378 {
379         struct btrfs_fs_info *info = root->fs_info;
380         substring_t args[MAX_OPT_ARGS];
381         char *p, *num, *orig = NULL;
382         u64 cache_gen;
383         int intarg;
384         int ret = 0;
385         char *compress_type;
386         bool compress_force = false;
387
388         cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
389         if (cache_gen)
390                 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
391
392         if (!options)
393                 goto out;
394
395         /*
396          * strsep changes the string, duplicate it because parse_options
397          * gets called twice
398          */
399         options = kstrdup(options, GFP_NOFS);
400         if (!options)
401                 return -ENOMEM;
402
403         orig = options;
404
405         while ((p = strsep(&options, ",")) != NULL) {
406                 int token;
407                 if (!*p)
408                         continue;
409
410                 token = match_token(p, tokens, args);
411                 switch (token) {
412                 case Opt_degraded:
413                         btrfs_info(root->fs_info, "allowing degraded mounts");
414                         btrfs_set_opt(info->mount_opt, DEGRADED);
415                         break;
416                 case Opt_subvol:
417                 case Opt_subvolid:
418                 case Opt_subvolrootid:
419                 case Opt_device:
420                         /*
421                          * These are parsed by btrfs_parse_early_options
422                          * and can be happily ignored here.
423                          */
424                         break;
425                 case Opt_nodatasum:
426                         btrfs_info(root->fs_info, "setting nodatasum");
427                         btrfs_set_opt(info->mount_opt, NODATASUM);
428                         break;
429                 case Opt_nodatacow:
430                         if (!btrfs_test_opt(root, COMPRESS) ||
431                                 !btrfs_test_opt(root, FORCE_COMPRESS)) {
432                                         btrfs_info(root->fs_info,
433                                                 "setting nodatacow, compression disabled");
434                         } else {
435                                 btrfs_info(root->fs_info, "setting nodatacow");
436                         }
437                         btrfs_clear_opt(info->mount_opt, COMPRESS);
438                         btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
439                         btrfs_set_opt(info->mount_opt, NODATACOW);
440                         btrfs_set_opt(info->mount_opt, NODATASUM);
441                         break;
442                 case Opt_compress_force:
443                 case Opt_compress_force_type:
444                         compress_force = true;
445                         /* Fallthrough */
446                 case Opt_compress:
447                 case Opt_compress_type:
448                         if (token == Opt_compress ||
449                             token == Opt_compress_force ||
450                             strcmp(args[0].from, "zlib") == 0) {
451                                 compress_type = "zlib";
452                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
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                         } else if (strcmp(args[0].from, "lzo") == 0) {
457                                 compress_type = "lzo";
458                                 info->compress_type = BTRFS_COMPRESS_LZO;
459                                 btrfs_set_opt(info->mount_opt, COMPRESS);
460                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
461                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
462                                 btrfs_set_fs_incompat(info, COMPRESS_LZO);
463                         } else if (strncmp(args[0].from, "no", 2) == 0) {
464                                 compress_type = "no";
465                                 btrfs_clear_opt(info->mount_opt, COMPRESS);
466                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
467                                 compress_force = false;
468                         } else {
469                                 ret = -EINVAL;
470                                 goto out;
471                         }
472
473                         if (compress_force) {
474                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
475                                 btrfs_info(root->fs_info, "force %s compression",
476                                         compress_type);
477                         } else if (btrfs_test_opt(root, COMPRESS)) {
478                                 pr_info("btrfs: use %s compression\n",
479                                         compress_type);
480                         }
481                         break;
482                 case Opt_ssd:
483                         btrfs_info(root->fs_info, "use ssd allocation scheme");
484                         btrfs_set_opt(info->mount_opt, SSD);
485                         break;
486                 case Opt_ssd_spread:
487                         btrfs_info(root->fs_info, "use spread ssd allocation scheme");
488                         btrfs_set_opt(info->mount_opt, SSD);
489                         btrfs_set_opt(info->mount_opt, SSD_SPREAD);
490                         break;
491                 case Opt_nossd:
492                         btrfs_info(root->fs_info, "not using ssd allocation scheme");
493                         btrfs_set_opt(info->mount_opt, NOSSD);
494                         btrfs_clear_opt(info->mount_opt, SSD);
495                         btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
496                         break;
497                 case Opt_nobarrier:
498                         btrfs_info(root->fs_info, "turning off barriers");
499                         btrfs_set_opt(info->mount_opt, NOBARRIER);
500                         break;
501                 case Opt_thread_pool:
502                         ret = match_int(&args[0], &intarg);
503                         if (ret) {
504                                 goto out;
505                         } else if (intarg > 0) {
506                                 info->thread_pool_size = intarg;
507                         } else {
508                                 ret = -EINVAL;
509                                 goto out;
510                         }
511                         break;
512                 case Opt_max_inline:
513                         num = match_strdup(&args[0]);
514                         if (num) {
515                                 info->max_inline = memparse(num, NULL);
516                                 kfree(num);
517
518                                 if (info->max_inline) {
519                                         info->max_inline = max_t(u64,
520                                                 info->max_inline,
521                                                 root->sectorsize);
522                                 }
523                                 btrfs_info(root->fs_info, "max_inline at %llu",
524                                         info->max_inline);
525                         } else {
526                                 ret = -ENOMEM;
527                                 goto out;
528                         }
529                         break;
530                 case Opt_alloc_start:
531                         num = match_strdup(&args[0]);
532                         if (num) {
533                                 mutex_lock(&info->chunk_mutex);
534                                 info->alloc_start = memparse(num, NULL);
535                                 mutex_unlock(&info->chunk_mutex);
536                                 kfree(num);
537                                 btrfs_info(root->fs_info, "allocations start at %llu",
538                                         info->alloc_start);
539                         } else {
540                                 ret = -ENOMEM;
541                                 goto out;
542                         }
543                         break;
544                 case Opt_noacl:
545                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
546                         break;
547                 case Opt_notreelog:
548                         btrfs_info(root->fs_info, "disabling tree log");
549                         btrfs_set_opt(info->mount_opt, NOTREELOG);
550                         break;
551                 case Opt_flushoncommit:
552                         btrfs_info(root->fs_info, "turning on flush-on-commit");
553                         btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
554                         break;
555                 case Opt_ratio:
556                         ret = match_int(&args[0], &intarg);
557                         if (ret) {
558                                 goto out;
559                         } else if (intarg >= 0) {
560                                 info->metadata_ratio = intarg;
561                                 btrfs_info(root->fs_info, "metadata ratio %d",
562                                        info->metadata_ratio);
563                         } else {
564                                 ret = -EINVAL;
565                                 goto out;
566                         }
567                         break;
568                 case Opt_discard:
569                         btrfs_set_opt(info->mount_opt, DISCARD);
570                         break;
571                 case Opt_space_cache:
572                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
573                         break;
574                 case Opt_rescan_uuid_tree:
575                         btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
576                         break;
577                 case Opt_no_space_cache:
578                         btrfs_info(root->fs_info, "disabling disk space caching");
579                         btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
580                         break;
581                 case Opt_inode_cache:
582                         btrfs_info(root->fs_info, "enabling inode map caching");
583                         btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
584                         break;
585                 case Opt_clear_cache:
586                         btrfs_info(root->fs_info, "force clearing of disk cache");
587                         btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
588                         break;
589                 case Opt_user_subvol_rm_allowed:
590                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
591                         break;
592                 case Opt_enospc_debug:
593                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
594                         break;
595                 case Opt_defrag:
596                         btrfs_info(root->fs_info, "enabling auto defrag");
597                         btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
598                         break;
599                 case Opt_recovery:
600                         btrfs_info(root->fs_info, "enabling auto recovery");
601                         btrfs_set_opt(info->mount_opt, RECOVERY);
602                         break;
603                 case Opt_skip_balance:
604                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
605                         break;
606 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
607                 case Opt_check_integrity_including_extent_data:
608                         btrfs_info(root->fs_info,
609                                    "enabling check integrity including extent data");
610                         btrfs_set_opt(info->mount_opt,
611                                       CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
612                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
613                         break;
614                 case Opt_check_integrity:
615                         btrfs_info(root->fs_info, "enabling check integrity");
616                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
617                         break;
618                 case Opt_check_integrity_print_mask:
619                         ret = match_int(&args[0], &intarg);
620                         if (ret) {
621                                 goto out;
622                         } else if (intarg >= 0) {
623                                 info->check_integrity_print_mask = intarg;
624                                 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
625                                        info->check_integrity_print_mask);
626                         } else {
627                                 ret = -EINVAL;
628                                 goto out;
629                         }
630                         break;
631 #else
632                 case Opt_check_integrity_including_extent_data:
633                 case Opt_check_integrity:
634                 case Opt_check_integrity_print_mask:
635                         btrfs_err(root->fs_info,
636                                 "support for check_integrity* not compiled in!");
637                         ret = -EINVAL;
638                         goto out;
639 #endif
640                 case Opt_fatal_errors:
641                         if (strcmp(args[0].from, "panic") == 0)
642                                 btrfs_set_opt(info->mount_opt,
643                                               PANIC_ON_FATAL_ERROR);
644                         else if (strcmp(args[0].from, "bug") == 0)
645                                 btrfs_clear_opt(info->mount_opt,
646                                               PANIC_ON_FATAL_ERROR);
647                         else {
648                                 ret = -EINVAL;
649                                 goto out;
650                         }
651                         break;
652                 case Opt_commit_interval:
653                         intarg = 0;
654                         ret = match_int(&args[0], &intarg);
655                         if (ret < 0) {
656                                 btrfs_err(root->fs_info, "invalid commit interval");
657                                 ret = -EINVAL;
658                                 goto out;
659                         }
660                         if (intarg > 0) {
661                                 if (intarg > 300) {
662                                         btrfs_warn(root->fs_info, "excessive commit interval %d",
663                                                         intarg);
664                                 }
665                                 info->commit_interval = intarg;
666                         } else {
667                                 btrfs_info(root->fs_info, "using default commit interval %ds",
668                                     BTRFS_DEFAULT_COMMIT_INTERVAL);
669                                 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
670                         }
671                         break;
672                 case Opt_err:
673                         btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
674                         ret = -EINVAL;
675                         goto out;
676                 default:
677                         break;
678                 }
679         }
680 out:
681         if (!ret && btrfs_test_opt(root, SPACE_CACHE))
682                 btrfs_info(root->fs_info, "disk space caching is enabled");
683         kfree(orig);
684         return ret;
685 }
686
687 /*
688  * Parse mount options that are required early in the mount process.
689  *
690  * All other options will be parsed on much later in the mount process and
691  * only when we need to allocate a new super block.
692  */
693 static int btrfs_parse_early_options(const char *options, fmode_t flags,
694                 void *holder, char **subvol_name, u64 *subvol_objectid,
695                 struct btrfs_fs_devices **fs_devices)
696 {
697         substring_t args[MAX_OPT_ARGS];
698         char *device_name, *opts, *orig, *p;
699         char *num = NULL;
700         int error = 0;
701
702         if (!options)
703                 return 0;
704
705         /*
706          * strsep changes the string, duplicate it because parse_options
707          * gets called twice
708          */
709         opts = kstrdup(options, GFP_KERNEL);
710         if (!opts)
711                 return -ENOMEM;
712         orig = opts;
713
714         while ((p = strsep(&opts, ",")) != NULL) {
715                 int token;
716                 if (!*p)
717                         continue;
718
719                 token = match_token(p, tokens, args);
720                 switch (token) {
721                 case Opt_subvol:
722                         kfree(*subvol_name);
723                         *subvol_name = match_strdup(&args[0]);
724                         if (!*subvol_name) {
725                                 error = -ENOMEM;
726                                 goto out;
727                         }
728                         break;
729                 case Opt_subvolid:
730                         num = match_strdup(&args[0]);
731                         if (num) {
732                                 *subvol_objectid = memparse(num, NULL);
733                                 kfree(num);
734                                 /* we want the original fs_tree */
735                                 if (!*subvol_objectid)
736                                         *subvol_objectid =
737                                                 BTRFS_FS_TREE_OBJECTID;
738                         } else {
739                                 error = -EINVAL;
740                                 goto out;
741                         }
742                         break;
743                 case Opt_subvolrootid:
744                         printk(KERN_WARNING
745                                 "BTRFS: 'subvolrootid' mount option is deprecated and has "
746                                 "no effect\n");
747                         break;
748                 case Opt_device:
749                         device_name = match_strdup(&args[0]);
750                         if (!device_name) {
751                                 error = -ENOMEM;
752                                 goto out;
753                         }
754                         error = btrfs_scan_one_device(device_name,
755                                         flags, holder, fs_devices);
756                         kfree(device_name);
757                         if (error)
758                                 goto out;
759                         break;
760                 default:
761                         break;
762                 }
763         }
764
765 out:
766         kfree(orig);
767         return error;
768 }
769
770 static struct dentry *get_default_root(struct super_block *sb,
771                                        u64 subvol_objectid)
772 {
773         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
774         struct btrfs_root *root = fs_info->tree_root;
775         struct btrfs_root *new_root;
776         struct btrfs_dir_item *di;
777         struct btrfs_path *path;
778         struct btrfs_key location;
779         struct inode *inode;
780         u64 dir_id;
781         int new = 0;
782
783         /*
784          * We have a specific subvol we want to mount, just setup location and
785          * go look up the root.
786          */
787         if (subvol_objectid) {
788                 location.objectid = subvol_objectid;
789                 location.type = BTRFS_ROOT_ITEM_KEY;
790                 location.offset = (u64)-1;
791                 goto find_root;
792         }
793
794         path = btrfs_alloc_path();
795         if (!path)
796                 return ERR_PTR(-ENOMEM);
797         path->leave_spinning = 1;
798
799         /*
800          * Find the "default" dir item which points to the root item that we
801          * will mount by default if we haven't been given a specific subvolume
802          * to mount.
803          */
804         dir_id = btrfs_super_root_dir(fs_info->super_copy);
805         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
806         if (IS_ERR(di)) {
807                 btrfs_free_path(path);
808                 return ERR_CAST(di);
809         }
810         if (!di) {
811                 /*
812                  * Ok the default dir item isn't there.  This is weird since
813                  * it's always been there, but don't freak out, just try and
814                  * mount to root most subvolume.
815                  */
816                 btrfs_free_path(path);
817                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
818                 new_root = fs_info->fs_root;
819                 goto setup_root;
820         }
821
822         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
823         btrfs_free_path(path);
824
825 find_root:
826         new_root = btrfs_read_fs_root_no_name(fs_info, &location);
827         if (IS_ERR(new_root))
828                 return ERR_CAST(new_root);
829
830         dir_id = btrfs_root_dirid(&new_root->root_item);
831 setup_root:
832         location.objectid = dir_id;
833         location.type = BTRFS_INODE_ITEM_KEY;
834         location.offset = 0;
835
836         inode = btrfs_iget(sb, &location, new_root, &new);
837         if (IS_ERR(inode))
838                 return ERR_CAST(inode);
839
840         /*
841          * If we're just mounting the root most subvol put the inode and return
842          * a reference to the dentry.  We will have already gotten a reference
843          * to the inode in btrfs_fill_super so we're good to go.
844          */
845         if (!new && sb->s_root->d_inode == inode) {
846                 iput(inode);
847                 return dget(sb->s_root);
848         }
849
850         return d_obtain_alias(inode);
851 }
852
853 static int btrfs_fill_super(struct super_block *sb,
854                             struct btrfs_fs_devices *fs_devices,
855                             void *data, int silent)
856 {
857         struct inode *inode;
858         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
859         struct btrfs_key key;
860         int err;
861
862         sb->s_maxbytes = MAX_LFS_FILESIZE;
863         sb->s_magic = BTRFS_SUPER_MAGIC;
864         sb->s_op = &btrfs_super_ops;
865         sb->s_d_op = &btrfs_dentry_operations;
866         sb->s_export_op = &btrfs_export_ops;
867         sb->s_xattr = btrfs_xattr_handlers;
868         sb->s_time_gran = 1;
869 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
870         sb->s_flags |= MS_POSIXACL;
871 #endif
872         sb->s_flags |= MS_I_VERSION;
873         err = open_ctree(sb, fs_devices, (char *)data);
874         if (err) {
875                 printk(KERN_ERR "BTRFS: open_ctree failed\n");
876                 return err;
877         }
878
879         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
880         key.type = BTRFS_INODE_ITEM_KEY;
881         key.offset = 0;
882         inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
883         if (IS_ERR(inode)) {
884                 err = PTR_ERR(inode);
885                 goto fail_close;
886         }
887
888         sb->s_root = d_make_root(inode);
889         if (!sb->s_root) {
890                 err = -ENOMEM;
891                 goto fail_close;
892         }
893
894         save_mount_options(sb, data);
895         cleancache_init_fs(sb);
896         sb->s_flags |= MS_ACTIVE;
897         return 0;
898
899 fail_close:
900         close_ctree(fs_info->tree_root);
901         return err;
902 }
903
904 int btrfs_sync_fs(struct super_block *sb, int wait)
905 {
906         struct btrfs_trans_handle *trans;
907         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
908         struct btrfs_root *root = fs_info->tree_root;
909
910         trace_btrfs_sync_fs(wait);
911
912         if (!wait) {
913                 filemap_flush(fs_info->btree_inode->i_mapping);
914                 return 0;
915         }
916
917         btrfs_wait_ordered_roots(fs_info, -1);
918
919         trans = btrfs_attach_transaction_barrier(root);
920         if (IS_ERR(trans)) {
921                 /* no transaction, don't bother */
922                 if (PTR_ERR(trans) == -ENOENT)
923                         return 0;
924                 return PTR_ERR(trans);
925         }
926         return btrfs_commit_transaction(trans, root);
927 }
928
929 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
930 {
931         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
932         struct btrfs_root *root = info->tree_root;
933         char *compress_type;
934
935         if (btrfs_test_opt(root, DEGRADED))
936                 seq_puts(seq, ",degraded");
937         if (btrfs_test_opt(root, NODATASUM))
938                 seq_puts(seq, ",nodatasum");
939         if (btrfs_test_opt(root, NODATACOW))
940                 seq_puts(seq, ",nodatacow");
941         if (btrfs_test_opt(root, NOBARRIER))
942                 seq_puts(seq, ",nobarrier");
943         if (info->max_inline != 8192 * 1024)
944                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
945         if (info->alloc_start != 0)
946                 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
947         if (info->thread_pool_size !=  min_t(unsigned long,
948                                              num_online_cpus() + 2, 8))
949                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
950         if (btrfs_test_opt(root, COMPRESS)) {
951                 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
952                         compress_type = "zlib";
953                 else
954                         compress_type = "lzo";
955                 if (btrfs_test_opt(root, FORCE_COMPRESS))
956                         seq_printf(seq, ",compress-force=%s", compress_type);
957                 else
958                         seq_printf(seq, ",compress=%s", compress_type);
959         }
960         if (btrfs_test_opt(root, NOSSD))
961                 seq_puts(seq, ",nossd");
962         if (btrfs_test_opt(root, SSD_SPREAD))
963                 seq_puts(seq, ",ssd_spread");
964         else if (btrfs_test_opt(root, SSD))
965                 seq_puts(seq, ",ssd");
966         if (btrfs_test_opt(root, NOTREELOG))
967                 seq_puts(seq, ",notreelog");
968         if (btrfs_test_opt(root, FLUSHONCOMMIT))
969                 seq_puts(seq, ",flushoncommit");
970         if (btrfs_test_opt(root, DISCARD))
971                 seq_puts(seq, ",discard");
972         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
973                 seq_puts(seq, ",noacl");
974         if (btrfs_test_opt(root, SPACE_CACHE))
975                 seq_puts(seq, ",space_cache");
976         else
977                 seq_puts(seq, ",nospace_cache");
978         if (btrfs_test_opt(root, RESCAN_UUID_TREE))
979                 seq_puts(seq, ",rescan_uuid_tree");
980         if (btrfs_test_opt(root, CLEAR_CACHE))
981                 seq_puts(seq, ",clear_cache");
982         if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
983                 seq_puts(seq, ",user_subvol_rm_allowed");
984         if (btrfs_test_opt(root, ENOSPC_DEBUG))
985                 seq_puts(seq, ",enospc_debug");
986         if (btrfs_test_opt(root, AUTO_DEFRAG))
987                 seq_puts(seq, ",autodefrag");
988         if (btrfs_test_opt(root, INODE_MAP_CACHE))
989                 seq_puts(seq, ",inode_cache");
990         if (btrfs_test_opt(root, SKIP_BALANCE))
991                 seq_puts(seq, ",skip_balance");
992         if (btrfs_test_opt(root, RECOVERY))
993                 seq_puts(seq, ",recovery");
994 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
995         if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
996                 seq_puts(seq, ",check_int_data");
997         else if (btrfs_test_opt(root, CHECK_INTEGRITY))
998                 seq_puts(seq, ",check_int");
999         if (info->check_integrity_print_mask)
1000                 seq_printf(seq, ",check_int_print_mask=%d",
1001                                 info->check_integrity_print_mask);
1002 #endif
1003         if (info->metadata_ratio)
1004                 seq_printf(seq, ",metadata_ratio=%d",
1005                                 info->metadata_ratio);
1006         if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1007                 seq_puts(seq, ",fatal_errors=panic");
1008         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1009                 seq_printf(seq, ",commit=%d", info->commit_interval);
1010         return 0;
1011 }
1012
1013 static int btrfs_test_super(struct super_block *s, void *data)
1014 {
1015         struct btrfs_fs_info *p = data;
1016         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1017
1018         return fs_info->fs_devices == p->fs_devices;
1019 }
1020
1021 static int btrfs_set_super(struct super_block *s, void *data)
1022 {
1023         int err = set_anon_super(s, data);
1024         if (!err)
1025                 s->s_fs_info = data;
1026         return err;
1027 }
1028
1029 /*
1030  * subvolumes are identified by ino 256
1031  */
1032 static inline int is_subvolume_inode(struct inode *inode)
1033 {
1034         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1035                 return 1;
1036         return 0;
1037 }
1038
1039 /*
1040  * This will strip out the subvol=%s argument for an argument string and add
1041  * subvolid=0 to make sure we get the actual tree root for path walking to the
1042  * subvol we want.
1043  */
1044 static char *setup_root_args(char *args)
1045 {
1046         unsigned len = strlen(args) + 2 + 1;
1047         char *src, *dst, *buf;
1048
1049         /*
1050          * We need the same args as before, but with this substitution:
1051          * s!subvol=[^,]+!subvolid=0!
1052          *
1053          * Since the replacement string is up to 2 bytes longer than the
1054          * original, allocate strlen(args) + 2 + 1 bytes.
1055          */
1056
1057         src = strstr(args, "subvol=");
1058         /* This shouldn't happen, but just in case.. */
1059         if (!src)
1060                 return NULL;
1061
1062         buf = dst = kmalloc(len, GFP_NOFS);
1063         if (!buf)
1064                 return NULL;
1065
1066         /*
1067          * If the subvol= arg is not at the start of the string,
1068          * copy whatever precedes it into buf.
1069          */
1070         if (src != args) {
1071                 *src++ = '\0';
1072                 strcpy(buf, args);
1073                 dst += strlen(args);
1074         }
1075
1076         strcpy(dst, "subvolid=0");
1077         dst += strlen("subvolid=0");
1078
1079         /*
1080          * If there is a "," after the original subvol=... string,
1081          * copy that suffix into our buffer.  Otherwise, we're done.
1082          */
1083         src = strchr(src, ',');
1084         if (src)
1085                 strcpy(dst, src);
1086
1087         return buf;
1088 }
1089
1090 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1091                                    const char *device_name, char *data)
1092 {
1093         struct dentry *root;
1094         struct vfsmount *mnt;
1095         char *newargs;
1096
1097         newargs = setup_root_args(data);
1098         if (!newargs)
1099                 return ERR_PTR(-ENOMEM);
1100         mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1101                              newargs);
1102         kfree(newargs);
1103         if (IS_ERR(mnt))
1104                 return ERR_CAST(mnt);
1105
1106         root = mount_subtree(mnt, subvol_name);
1107
1108         if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1109                 struct super_block *s = root->d_sb;
1110                 dput(root);
1111                 root = ERR_PTR(-EINVAL);
1112                 deactivate_locked_super(s);
1113                 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1114                                 subvol_name);
1115         }
1116
1117         return root;
1118 }
1119
1120 /*
1121  * Find a superblock for the given device / mount point.
1122  *
1123  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
1124  *        for multiple device setup.  Make sure to keep it in sync.
1125  */
1126 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1127                 const char *device_name, void *data)
1128 {
1129         struct block_device *bdev = NULL;
1130         struct super_block *s;
1131         struct dentry *root;
1132         struct btrfs_fs_devices *fs_devices = NULL;
1133         struct btrfs_fs_info *fs_info = NULL;
1134         fmode_t mode = FMODE_READ;
1135         char *subvol_name = NULL;
1136         u64 subvol_objectid = 0;
1137         int error = 0;
1138
1139         if (!(flags & MS_RDONLY))
1140                 mode |= FMODE_WRITE;
1141
1142         error = btrfs_parse_early_options(data, mode, fs_type,
1143                                           &subvol_name, &subvol_objectid,
1144                                           &fs_devices);
1145         if (error) {
1146                 kfree(subvol_name);
1147                 return ERR_PTR(error);
1148         }
1149
1150         if (subvol_name) {
1151                 root = mount_subvol(subvol_name, flags, device_name, data);
1152                 kfree(subvol_name);
1153                 return root;
1154         }
1155
1156         error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1157         if (error)
1158                 return ERR_PTR(error);
1159
1160         /*
1161          * Setup a dummy root and fs_info for test/set super.  This is because
1162          * we don't actually fill this stuff out until open_ctree, but we need
1163          * it for searching for existing supers, so this lets us do that and
1164          * then open_ctree will properly initialize everything later.
1165          */
1166         fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1167         if (!fs_info)
1168                 return ERR_PTR(-ENOMEM);
1169
1170         fs_info->fs_devices = fs_devices;
1171
1172         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1173         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1174         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1175                 error = -ENOMEM;
1176                 goto error_fs_info;
1177         }
1178
1179         error = btrfs_open_devices(fs_devices, mode, fs_type);
1180         if (error)
1181                 goto error_fs_info;
1182
1183         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1184                 error = -EACCES;
1185                 goto error_close_devices;
1186         }
1187
1188         bdev = fs_devices->latest_bdev;
1189         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1190                  fs_info);
1191         if (IS_ERR(s)) {
1192                 error = PTR_ERR(s);
1193                 goto error_close_devices;
1194         }
1195
1196         if (s->s_root) {
1197                 btrfs_close_devices(fs_devices);
1198                 free_fs_info(fs_info);
1199                 if ((flags ^ s->s_flags) & MS_RDONLY)
1200                         error = -EBUSY;
1201         } else {
1202                 char b[BDEVNAME_SIZE];
1203
1204                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1205                 btrfs_sb(s)->bdev_holder = fs_type;
1206                 error = btrfs_fill_super(s, fs_devices, data,
1207                                          flags & MS_SILENT ? 1 : 0);
1208         }
1209
1210         root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1211         if (IS_ERR(root))
1212                 deactivate_locked_super(s);
1213
1214         return root;
1215
1216 error_close_devices:
1217         btrfs_close_devices(fs_devices);
1218 error_fs_info:
1219         free_fs_info(fs_info);
1220         return ERR_PTR(error);
1221 }
1222
1223 static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1224 {
1225         spin_lock_irq(&workers->lock);
1226         workers->max_workers = new_limit;
1227         spin_unlock_irq(&workers->lock);
1228 }
1229
1230 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1231                                      int new_pool_size, int old_pool_size)
1232 {
1233         if (new_pool_size == old_pool_size)
1234                 return;
1235
1236         fs_info->thread_pool_size = new_pool_size;
1237
1238         btrfs_info(fs_info, "resize thread pool %d -> %d",
1239                old_pool_size, new_pool_size);
1240
1241         btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1242         btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1243         btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1244         btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1245         btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1246         btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1247         btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1248         btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1249         btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1250         btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1251         btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1252         btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1253         btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
1254         btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers,
1255                               new_pool_size);
1256 }
1257
1258 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1259 {
1260         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1261 }
1262
1263 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1264                                        unsigned long old_opts, int flags)
1265 {
1266         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1267             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1268              (flags & MS_RDONLY))) {
1269                 /* wait for any defraggers to finish */
1270                 wait_event(fs_info->transaction_wait,
1271                            (atomic_read(&fs_info->defrag_running) == 0));
1272                 if (flags & MS_RDONLY)
1273                         sync_filesystem(fs_info->sb);
1274         }
1275 }
1276
1277 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1278                                          unsigned long old_opts)
1279 {
1280         /*
1281          * We need cleanup all defragable inodes if the autodefragment is
1282          * close or the fs is R/O.
1283          */
1284         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1285             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1286              (fs_info->sb->s_flags & MS_RDONLY))) {
1287                 btrfs_cleanup_defrag_inodes(fs_info);
1288         }
1289
1290         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1291 }
1292
1293 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1294 {
1295         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1296         struct btrfs_root *root = fs_info->tree_root;
1297         unsigned old_flags = sb->s_flags;
1298         unsigned long old_opts = fs_info->mount_opt;
1299         unsigned long old_compress_type = fs_info->compress_type;
1300         u64 old_max_inline = fs_info->max_inline;
1301         u64 old_alloc_start = fs_info->alloc_start;
1302         int old_thread_pool_size = fs_info->thread_pool_size;
1303         unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1304         int ret;
1305
1306         btrfs_remount_prepare(fs_info);
1307
1308         ret = btrfs_parse_options(root, data);
1309         if (ret) {
1310                 ret = -EINVAL;
1311                 goto restore;
1312         }
1313
1314         btrfs_remount_begin(fs_info, old_opts, *flags);
1315         btrfs_resize_thread_pool(fs_info,
1316                 fs_info->thread_pool_size, old_thread_pool_size);
1317
1318         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1319                 goto out;
1320
1321         if (*flags & MS_RDONLY) {
1322                 /*
1323                  * this also happens on 'umount -rf' or on shutdown, when
1324                  * the filesystem is busy.
1325                  */
1326
1327                 /* wait for the uuid_scan task to finish */
1328                 down(&fs_info->uuid_tree_rescan_sem);
1329                 /* avoid complains from lockdep et al. */
1330                 up(&fs_info->uuid_tree_rescan_sem);
1331
1332                 sb->s_flags |= MS_RDONLY;
1333
1334                 btrfs_dev_replace_suspend_for_unmount(fs_info);
1335                 btrfs_scrub_cancel(fs_info);
1336                 btrfs_pause_balance(fs_info);
1337
1338                 ret = btrfs_commit_super(root);
1339                 if (ret)
1340                         goto restore;
1341         } else {
1342                 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1343                         btrfs_err(fs_info,
1344                                 "Remounting read-write after error is not allowed");
1345                         ret = -EINVAL;
1346                         goto restore;
1347                 }
1348                 if (fs_info->fs_devices->rw_devices == 0) {
1349                         ret = -EACCES;
1350                         goto restore;
1351                 }
1352
1353                 if (fs_info->fs_devices->missing_devices >
1354                      fs_info->num_tolerated_disk_barrier_failures &&
1355                     !(*flags & MS_RDONLY)) {
1356                         btrfs_warn(fs_info,
1357                                 "too many missing devices, writeable remount is not allowed");
1358                         ret = -EACCES;
1359                         goto restore;
1360                 }
1361
1362                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1363                         ret = -EINVAL;
1364                         goto restore;
1365                 }
1366
1367                 ret = btrfs_cleanup_fs_roots(fs_info);
1368                 if (ret)
1369                         goto restore;
1370
1371                 /* recover relocation */
1372                 ret = btrfs_recover_relocation(root);
1373                 if (ret)
1374                         goto restore;
1375
1376                 ret = btrfs_resume_balance_async(fs_info);
1377                 if (ret)
1378                         goto restore;
1379
1380                 ret = btrfs_resume_dev_replace_async(fs_info);
1381                 if (ret) {
1382                         btrfs_warn(fs_info, "failed to resume dev_replace");
1383                         goto restore;
1384                 }
1385
1386                 if (!fs_info->uuid_root) {
1387                         btrfs_info(fs_info, "creating UUID tree");
1388                         ret = btrfs_create_uuid_tree(fs_info);
1389                         if (ret) {
1390                                 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1391                                 goto restore;
1392                         }
1393                 }
1394                 sb->s_flags &= ~MS_RDONLY;
1395         }
1396 out:
1397         btrfs_remount_cleanup(fs_info, old_opts);
1398         return 0;
1399
1400 restore:
1401         /* We've hit an error - don't reset MS_RDONLY */
1402         if (sb->s_flags & MS_RDONLY)
1403                 old_flags |= MS_RDONLY;
1404         sb->s_flags = old_flags;
1405         fs_info->mount_opt = old_opts;
1406         fs_info->compress_type = old_compress_type;
1407         fs_info->max_inline = old_max_inline;
1408         mutex_lock(&fs_info->chunk_mutex);
1409         fs_info->alloc_start = old_alloc_start;
1410         mutex_unlock(&fs_info->chunk_mutex);
1411         btrfs_resize_thread_pool(fs_info,
1412                 old_thread_pool_size, fs_info->thread_pool_size);
1413         fs_info->metadata_ratio = old_metadata_ratio;
1414         btrfs_remount_cleanup(fs_info, old_opts);
1415         return ret;
1416 }
1417
1418 /* Used to sort the devices by max_avail(descending sort) */
1419 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1420                                        const void *dev_info2)
1421 {
1422         if (((struct btrfs_device_info *)dev_info1)->max_avail >
1423             ((struct btrfs_device_info *)dev_info2)->max_avail)
1424                 return -1;
1425         else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1426                  ((struct btrfs_device_info *)dev_info2)->max_avail)
1427                 return 1;
1428         else
1429         return 0;
1430 }
1431
1432 /*
1433  * sort the devices by max_avail, in which max free extent size of each device
1434  * is stored.(Descending Sort)
1435  */
1436 static inline void btrfs_descending_sort_devices(
1437                                         struct btrfs_device_info *devices,
1438                                         size_t nr_devices)
1439 {
1440         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1441              btrfs_cmp_device_free_bytes, NULL);
1442 }
1443
1444 /*
1445  * The helper to calc the free space on the devices that can be used to store
1446  * file data.
1447  */
1448 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1449 {
1450         struct btrfs_fs_info *fs_info = root->fs_info;
1451         struct btrfs_device_info *devices_info;
1452         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1453         struct btrfs_device *device;
1454         u64 skip_space;
1455         u64 type;
1456         u64 avail_space;
1457         u64 used_space;
1458         u64 min_stripe_size;
1459         int min_stripes = 1, num_stripes = 1;
1460         int i = 0, nr_devices;
1461         int ret;
1462
1463         nr_devices = fs_info->fs_devices->open_devices;
1464         BUG_ON(!nr_devices);
1465
1466         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1467                                GFP_NOFS);
1468         if (!devices_info)
1469                 return -ENOMEM;
1470
1471         /* calc min stripe number for data space alloction */
1472         type = btrfs_get_alloc_profile(root, 1);
1473         if (type & BTRFS_BLOCK_GROUP_RAID0) {
1474                 min_stripes = 2;
1475                 num_stripes = nr_devices;
1476         } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1477                 min_stripes = 2;
1478                 num_stripes = 2;
1479         } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1480                 min_stripes = 4;
1481                 num_stripes = 4;
1482         }
1483
1484         if (type & BTRFS_BLOCK_GROUP_DUP)
1485                 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1486         else
1487                 min_stripe_size = BTRFS_STRIPE_LEN;
1488
1489         list_for_each_entry(device, &fs_devices->devices, dev_list) {
1490                 if (!device->in_fs_metadata || !device->bdev ||
1491                     device->is_tgtdev_for_dev_replace)
1492                         continue;
1493
1494                 avail_space = device->total_bytes - device->bytes_used;
1495
1496                 /* align with stripe_len */
1497                 do_div(avail_space, BTRFS_STRIPE_LEN);
1498                 avail_space *= BTRFS_STRIPE_LEN;
1499
1500                 /*
1501                  * In order to avoid overwritting the superblock on the drive,
1502                  * btrfs starts at an offset of at least 1MB when doing chunk
1503                  * allocation.
1504                  */
1505                 skip_space = 1024 * 1024;
1506
1507                 /* user can set the offset in fs_info->alloc_start. */
1508                 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1509                     device->total_bytes)
1510                         skip_space = max(fs_info->alloc_start, skip_space);
1511
1512                 /*
1513                  * btrfs can not use the free space in [0, skip_space - 1],
1514                  * we must subtract it from the total. In order to implement
1515                  * it, we account the used space in this range first.
1516                  */
1517                 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1518                                                      &used_space);
1519                 if (ret) {
1520                         kfree(devices_info);
1521                         return ret;
1522                 }
1523
1524                 /* calc the free space in [0, skip_space - 1] */
1525                 skip_space -= used_space;
1526
1527                 /*
1528                  * we can use the free space in [0, skip_space - 1], subtract
1529                  * it from the total.
1530                  */
1531                 if (avail_space && avail_space >= skip_space)
1532                         avail_space -= skip_space;
1533                 else
1534                         avail_space = 0;
1535
1536                 if (avail_space < min_stripe_size)
1537                         continue;
1538
1539                 devices_info[i].dev = device;
1540                 devices_info[i].max_avail = avail_space;
1541
1542                 i++;
1543         }
1544
1545         nr_devices = i;
1546
1547         btrfs_descending_sort_devices(devices_info, nr_devices);
1548
1549         i = nr_devices - 1;
1550         avail_space = 0;
1551         while (nr_devices >= min_stripes) {
1552                 if (num_stripes > nr_devices)
1553                         num_stripes = nr_devices;
1554
1555                 if (devices_info[i].max_avail >= min_stripe_size) {
1556                         int j;
1557                         u64 alloc_size;
1558
1559                         avail_space += devices_info[i].max_avail * num_stripes;
1560                         alloc_size = devices_info[i].max_avail;
1561                         for (j = i + 1 - num_stripes; j <= i; j++)
1562                                 devices_info[j].max_avail -= alloc_size;
1563                 }
1564                 i--;
1565                 nr_devices--;
1566         }
1567
1568         kfree(devices_info);
1569         *free_bytes = avail_space;
1570         return 0;
1571 }
1572
1573 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1574 {
1575         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1576         struct btrfs_super_block *disk_super = fs_info->super_copy;
1577         struct list_head *head = &fs_info->space_info;
1578         struct btrfs_space_info *found;
1579         u64 total_used = 0;
1580         u64 total_free_data = 0;
1581         int bits = dentry->d_sb->s_blocksize_bits;
1582         __be32 *fsid = (__be32 *)fs_info->fsid;
1583         int ret;
1584
1585         /* holding chunk_muext to avoid allocating new chunks */
1586         mutex_lock(&fs_info->chunk_mutex);
1587         rcu_read_lock();
1588         list_for_each_entry_rcu(found, head, list) {
1589                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1590                         total_free_data += found->disk_total - found->disk_used;
1591                         total_free_data -=
1592                                 btrfs_account_ro_block_groups_free_space(found);
1593                 }
1594
1595                 total_used += found->disk_used;
1596         }
1597         rcu_read_unlock();
1598
1599         buf->f_namelen = BTRFS_NAME_LEN;
1600         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1601         buf->f_bfree = buf->f_blocks - (total_used >> bits);
1602         buf->f_bsize = dentry->d_sb->s_blocksize;
1603         buf->f_type = BTRFS_SUPER_MAGIC;
1604         buf->f_bavail = total_free_data;
1605         ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1606         if (ret) {
1607                 mutex_unlock(&fs_info->chunk_mutex);
1608                 return ret;
1609         }
1610         buf->f_bavail += total_free_data;
1611         buf->f_bavail = buf->f_bavail >> bits;
1612         mutex_unlock(&fs_info->chunk_mutex);
1613
1614         /* We treat it as constant endianness (it doesn't matter _which_)
1615            because we want the fsid to come out the same whether mounted
1616            on a big-endian or little-endian host */
1617         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1618         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1619         /* Mask in the root object ID too, to disambiguate subvols */
1620         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1621         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1622
1623         return 0;
1624 }
1625
1626 static void btrfs_kill_super(struct super_block *sb)
1627 {
1628         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1629         kill_anon_super(sb);
1630         free_fs_info(fs_info);
1631 }
1632
1633 static struct file_system_type btrfs_fs_type = {
1634         .owner          = THIS_MODULE,
1635         .name           = "btrfs",
1636         .mount          = btrfs_mount,
1637         .kill_sb        = btrfs_kill_super,
1638         .fs_flags       = FS_REQUIRES_DEV,
1639 };
1640 MODULE_ALIAS_FS("btrfs");
1641
1642 /*
1643  * used by btrfsctl to scan devices when no FS is mounted
1644  */
1645 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1646                                 unsigned long arg)
1647 {
1648         struct btrfs_ioctl_vol_args *vol;
1649         struct btrfs_fs_devices *fs_devices;
1650         int ret = -ENOTTY;
1651
1652         if (!capable(CAP_SYS_ADMIN))
1653                 return -EPERM;
1654
1655         vol = memdup_user((void __user *)arg, sizeof(*vol));
1656         if (IS_ERR(vol))
1657                 return PTR_ERR(vol);
1658
1659         switch (cmd) {
1660         case BTRFS_IOC_SCAN_DEV:
1661                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1662                                             &btrfs_fs_type, &fs_devices);
1663                 break;
1664         case BTRFS_IOC_DEVICES_READY:
1665                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1666                                             &btrfs_fs_type, &fs_devices);
1667                 if (ret)
1668                         break;
1669                 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1670                 break;
1671         }
1672
1673         kfree(vol);
1674         return ret;
1675 }
1676
1677 static int btrfs_freeze(struct super_block *sb)
1678 {
1679         struct btrfs_trans_handle *trans;
1680         struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1681
1682         trans = btrfs_attach_transaction_barrier(root);
1683         if (IS_ERR(trans)) {
1684                 /* no transaction, don't bother */
1685                 if (PTR_ERR(trans) == -ENOENT)
1686                         return 0;
1687                 return PTR_ERR(trans);
1688         }
1689         return btrfs_commit_transaction(trans, root);
1690 }
1691
1692 static int btrfs_unfreeze(struct super_block *sb)
1693 {
1694         return 0;
1695 }
1696
1697 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1698 {
1699         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1700         struct btrfs_fs_devices *cur_devices;
1701         struct btrfs_device *dev, *first_dev = NULL;
1702         struct list_head *head;
1703         struct rcu_string *name;
1704
1705         mutex_lock(&fs_info->fs_devices->device_list_mutex);
1706         cur_devices = fs_info->fs_devices;
1707         while (cur_devices) {
1708                 head = &cur_devices->devices;
1709                 list_for_each_entry(dev, head, dev_list) {
1710                         if (dev->missing)
1711                                 continue;
1712                         if (!first_dev || dev->devid < first_dev->devid)
1713                                 first_dev = dev;
1714                 }
1715                 cur_devices = cur_devices->seed;
1716         }
1717
1718         if (first_dev) {
1719                 rcu_read_lock();
1720                 name = rcu_dereference(first_dev->name);
1721                 seq_escape(m, name->str, " \t\n\\");
1722                 rcu_read_unlock();
1723         } else {
1724                 WARN_ON(1);
1725         }
1726         mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1727         return 0;
1728 }
1729
1730 static const struct super_operations btrfs_super_ops = {
1731         .drop_inode     = btrfs_drop_inode,
1732         .evict_inode    = btrfs_evict_inode,
1733         .put_super      = btrfs_put_super,
1734         .sync_fs        = btrfs_sync_fs,
1735         .show_options   = btrfs_show_options,
1736         .show_devname   = btrfs_show_devname,
1737         .write_inode    = btrfs_write_inode,
1738         .alloc_inode    = btrfs_alloc_inode,
1739         .destroy_inode  = btrfs_destroy_inode,
1740         .statfs         = btrfs_statfs,
1741         .remount_fs     = btrfs_remount,
1742         .freeze_fs      = btrfs_freeze,
1743         .unfreeze_fs    = btrfs_unfreeze,
1744 };
1745
1746 static const struct file_operations btrfs_ctl_fops = {
1747         .unlocked_ioctl  = btrfs_control_ioctl,
1748         .compat_ioctl = btrfs_control_ioctl,
1749         .owner   = THIS_MODULE,
1750         .llseek = noop_llseek,
1751 };
1752
1753 static struct miscdevice btrfs_misc = {
1754         .minor          = BTRFS_MINOR,
1755         .name           = "btrfs-control",
1756         .fops           = &btrfs_ctl_fops
1757 };
1758
1759 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1760 MODULE_ALIAS("devname:btrfs-control");
1761
1762 static int btrfs_interface_init(void)
1763 {
1764         return misc_register(&btrfs_misc);
1765 }
1766
1767 static void btrfs_interface_exit(void)
1768 {
1769         if (misc_deregister(&btrfs_misc) < 0)
1770                 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
1771 }
1772
1773 static void btrfs_print_info(void)
1774 {
1775         printk(KERN_INFO "Btrfs loaded"
1776 #ifdef CONFIG_BTRFS_DEBUG
1777                         ", debug=on"
1778 #endif
1779 #ifdef CONFIG_BTRFS_ASSERT
1780                         ", assert=on"
1781 #endif
1782 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1783                         ", integrity-checker=on"
1784 #endif
1785                         "\n");
1786 }
1787
1788 static int btrfs_run_sanity_tests(void)
1789 {
1790         int ret;
1791
1792         ret = btrfs_init_test_fs();
1793         if (ret)
1794                 return ret;
1795
1796         ret = btrfs_test_free_space_cache();
1797         if (ret)
1798                 goto out;
1799         ret = btrfs_test_extent_buffer_operations();
1800         if (ret)
1801                 goto out;
1802         ret = btrfs_test_extent_io();
1803         if (ret)
1804                 goto out;
1805         ret = btrfs_test_inodes();
1806 out:
1807         btrfs_destroy_test_fs();
1808         return ret;
1809 }
1810
1811 static int __init init_btrfs_fs(void)
1812 {
1813         int err;
1814
1815         err = btrfs_init_sysfs();
1816         if (err)
1817                 return err;
1818
1819         btrfs_init_compress();
1820
1821         err = btrfs_init_cachep();
1822         if (err)
1823                 goto free_compress;
1824
1825         err = extent_io_init();
1826         if (err)
1827                 goto free_cachep;
1828
1829         err = extent_map_init();
1830         if (err)
1831                 goto free_extent_io;
1832
1833         err = ordered_data_init();
1834         if (err)
1835                 goto free_extent_map;
1836
1837         err = btrfs_delayed_inode_init();
1838         if (err)
1839                 goto free_ordered_data;
1840
1841         err = btrfs_auto_defrag_init();
1842         if (err)
1843                 goto free_delayed_inode;
1844
1845         err = btrfs_delayed_ref_init();
1846         if (err)
1847                 goto free_auto_defrag;
1848
1849         err = btrfs_prelim_ref_init();
1850         if (err)
1851                 goto free_prelim_ref;
1852
1853         err = btrfs_interface_init();
1854         if (err)
1855                 goto free_delayed_ref;
1856
1857         btrfs_init_lockdep();
1858
1859         btrfs_print_info();
1860
1861         err = btrfs_run_sanity_tests();
1862         if (err)
1863                 goto unregister_ioctl;
1864
1865         err = register_filesystem(&btrfs_fs_type);
1866         if (err)
1867                 goto unregister_ioctl;
1868
1869         return 0;
1870
1871 unregister_ioctl:
1872         btrfs_interface_exit();
1873 free_prelim_ref:
1874         btrfs_prelim_ref_exit();
1875 free_delayed_ref:
1876         btrfs_delayed_ref_exit();
1877 free_auto_defrag:
1878         btrfs_auto_defrag_exit();
1879 free_delayed_inode:
1880         btrfs_delayed_inode_exit();
1881 free_ordered_data:
1882         ordered_data_exit();
1883 free_extent_map:
1884         extent_map_exit();
1885 free_extent_io:
1886         extent_io_exit();
1887 free_cachep:
1888         btrfs_destroy_cachep();
1889 free_compress:
1890         btrfs_exit_compress();
1891         btrfs_exit_sysfs();
1892         return err;
1893 }
1894
1895 static void __exit exit_btrfs_fs(void)
1896 {
1897         btrfs_destroy_cachep();
1898         btrfs_delayed_ref_exit();
1899         btrfs_auto_defrag_exit();
1900         btrfs_delayed_inode_exit();
1901         btrfs_prelim_ref_exit();
1902         ordered_data_exit();
1903         extent_map_exit();
1904         extent_io_exit();
1905         btrfs_interface_exit();
1906         unregister_filesystem(&btrfs_fs_type);
1907         btrfs_exit_sysfs();
1908         btrfs_cleanup_fs_uuids();
1909         btrfs_exit_compress();
1910 }
1911
1912 module_init(init_btrfs_fs)
1913 module_exit(exit_btrfs_fs)
1914
1915 MODULE_LICENSE("GPL");