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, Opt_barrier, Opt_nodefrag,
 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_barrier, "barrier"},
 339         {Opt_max_inline, "max_inline=%s"},
 340         {Opt_alloc_start, "alloc_start=%s"},
 341         {Opt_thread_pool, "thread_pool=%d"},
 342         {Opt_compress, "compress"},
 343         {Opt_compress_type, "compress=%s"},
 344         {Opt_compress_force, "compress-force"},
 345         {Opt_compress_force_type, "compress-force=%s"},
 346         {Opt_ssd, "ssd"},
 347         {Opt_ssd_spread, "ssd_spread"},
 348         {Opt_nossd, "nossd"},
 349         {Opt_noacl, "noacl"},
 350         {Opt_notreelog, "notreelog"},
 351         {Opt_flushoncommit, "flushoncommit"},
 352         {Opt_ratio, "metadata_ratio=%d"},
 353         {Opt_discard, "discard"},
 354         {Opt_space_cache, "space_cache"},
 355         {Opt_clear_cache, "clear_cache"},
 356         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 357         {Opt_enospc_debug, "enospc_debug"},
 358         {Opt_subvolrootid, "subvolrootid=%d"},
 359         {Opt_defrag, "autodefrag"},
 360         {Opt_nodefrag, "noautodefrag"},
 361         {Opt_inode_cache, "inode_cache"},
 362         {Opt_no_space_cache, "nospace_cache"},
 363         {Opt_recovery, "recovery"},
 364         {Opt_skip_balance, "skip_balance"},
 365         {Opt_check_integrity, "check_int"},
 366         {Opt_check_integrity_including_extent_data, "check_int_data"},
 367         {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
 368         {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 369         {Opt_fatal_errors, "fatal_errors=%s"},
 370         {Opt_commit_interval, "commit=%d"},
 371         {Opt_err, NULL},
 372 };
 373
 374 /*
 375  * Regular mount options parser.  Everything that is needed only when
 376  * reading in a new superblock is parsed here.
 377  * XXX JDM: This needs to be cleaned up for remount.
 378  */
 379 int btrfs_parse_options(struct btrfs_root *root, char *options)
 380 {
 381         struct btrfs_fs_info *info = root->fs_info;
 382         substring_t args[MAX_OPT_ARGS];
 383         char *p, *num, *orig = NULL;
 384         u64 cache_gen;
 385         int intarg;
 386         int ret = 0;
 387         char *compress_type;
 388         bool compress_force = false;
 389
 390         cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
 391         if (cache_gen)
 392                 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 393
 394         if (!options)
 395                 goto out;
 396
 397         /*
 398          * strsep changes the string, duplicate it because parse_options
 399          * gets called twice
 400          */
 401         options = kstrdup(options, GFP_NOFS);
 402         if (!options)
 403                 return -ENOMEM;
 404
 405         orig = options;
 406
 407         while ((p = strsep(&options, ",")) != NULL) {
 408                 int token;
 409                 if (!*p)
 410                         continue;
 411
 412                 token = match_token(p, tokens, args);
 413                 switch (token) {
 414                 case Opt_degraded:
 415                         btrfs_info(root->fs_info, "allowing degraded mounts");
 416                         btrfs_set_opt(info->mount_opt, DEGRADED);
 417                         break;
 418                 case Opt_subvol:
 419                 case Opt_subvolid:
 420                 case Opt_subvolrootid:
 421                 case Opt_device:
 422                         /*
 423                          * These are parsed by btrfs_parse_early_options
 424                          * and can be happily ignored here.
 425                          */
 426                         break;
 427                 case Opt_nodatasum:
 428                         btrfs_info(root->fs_info, "setting nodatasum");
 429                         btrfs_set_opt(info->mount_opt, NODATASUM);
 430                         break;
 431                 case Opt_nodatacow:
 432                         if (!btrfs_test_opt(root, COMPRESS) ||
 433                                 !btrfs_test_opt(root, FORCE_COMPRESS)) {
 434                                         btrfs_info(root->fs_info,
 435                                                 "setting nodatacow, compression disabled");
 436                         } else {
 437                                 btrfs_info(root->fs_info, "setting nodatacow");
 438                         }
 439                         btrfs_clear_opt(info->mount_opt, COMPRESS);
 440                         btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 441                         btrfs_set_opt(info->mount_opt, NODATACOW);
 442                         btrfs_set_opt(info->mount_opt, NODATASUM);
 443                         break;
 444                 case Opt_compress_force:
 445                 case Opt_compress_force_type:
 446                         compress_force = true;
 447                         /* Fallthrough */
 448                 case Opt_compress:
 449                 case Opt_compress_type:
 450                         if (token == Opt_compress ||
 451                             token == Opt_compress_force ||
 452                             strcmp(args[0].from, "zlib") == 0) {
 453                                 compress_type = "zlib";
 454                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 455                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 456                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 457                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 458                         } else if (strcmp(args[0].from, "lzo") == 0) {
 459                                 compress_type = "lzo";
 460                                 info->compress_type = BTRFS_COMPRESS_LZO;
 461                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 462                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 463                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 464                                 btrfs_set_fs_incompat(info, COMPRESS_LZO);
 465                         } else if (strncmp(args[0].from, "no", 2) == 0) {
 466                                 compress_type = "no";
 467                                 btrfs_clear_opt(info->mount_opt, COMPRESS);
 468                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 469                                 compress_force = false;
 470                         } else {
 471                                 ret = -EINVAL;
 472                                 goto out;
 473                         }
 474
 475                         if (compress_force) {
 476                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 477                                 btrfs_info(root->fs_info, "force %s compression",
 478                                         compress_type);
 479                         } else if (btrfs_test_opt(root, COMPRESS)) {
 480                                 pr_info("btrfs: use %s compression\n",
 481                                         compress_type);
 482                         }
 483                         break;
 484                 case Opt_ssd:
 485                         btrfs_info(root->fs_info, "use ssd allocation scheme");
 486                         btrfs_set_opt(info->mount_opt, SSD);
 487                         break;
 488                 case Opt_ssd_spread:
 489                         btrfs_info(root->fs_info, "use spread ssd allocation scheme");
 490                         btrfs_set_opt(info->mount_opt, SSD);
 491                         btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 492                         break;
 493                 case Opt_nossd:
 494                         btrfs_info(root->fs_info, "not using ssd allocation scheme");
 495                         btrfs_set_opt(info->mount_opt, NOSSD);
 496                         btrfs_clear_opt(info->mount_opt, SSD);
 497                         btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 498                         break;
 499                 case Opt_barrier:
 500                         if (btrfs_test_opt(root, NOBARRIER))
 501                                 btrfs_info(root->fs_info, "turning on barriers");
 502                         btrfs_clear_opt(info->mount_opt, NOBARRIER);
 503                         break;
 504                 case Opt_nobarrier:
 505                         btrfs_info(root->fs_info, "turning off barriers");
 506                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 507                         break;
 508                 case Opt_thread_pool:
 509                         ret = match_int(&args[0], &intarg);
 510                         if (ret) {
 511                                 goto out;
 512                         } else if (intarg > 0) {
 513                                 info->thread_pool_size = intarg;
 514                         } else {
 515                                 ret = -EINVAL;
 516                                 goto out;
 517                         }
 518                         break;
 519                 case Opt_max_inline:
 520                         num = match_strdup(&args[0]);
 521                         if (num) {
 522                                 info->max_inline = memparse(num, NULL);
 523                                 kfree(num);
 524
 525                                 if (info->max_inline) {
 526                                         info->max_inline = max_t(u64,
 527                                                 info->max_inline,
 528                                                 root->sectorsize);
 529                                 }
 530                                 btrfs_info(root->fs_info, "max_inline at %llu",
 531                                         info->max_inline);
 532                         } else {
 533                                 ret = -ENOMEM;
 534                                 goto out;
 535                         }
 536                         break;
 537                 case Opt_alloc_start:
 538                         num = match_strdup(&args[0]);
 539                         if (num) {
 540                                 mutex_lock(&info->chunk_mutex);
 541                                 info->alloc_start = memparse(num, NULL);
 542                                 mutex_unlock(&info->chunk_mutex);
 543                                 kfree(num);
 544                                 btrfs_info(root->fs_info, "allocations start at %llu",
 545                                         info->alloc_start);
 546                         } else {
 547                                 ret = -ENOMEM;
 548                                 goto out;
 549                         }
 550                         break;
 551                 case Opt_noacl:
 552                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 553                         break;
 554                 case Opt_notreelog:
 555                         btrfs_info(root->fs_info, "disabling tree log");
 556                         btrfs_set_opt(info->mount_opt, NOTREELOG);
 557                         break;
 558                 case Opt_flushoncommit:
 559                         btrfs_info(root->fs_info, "turning on flush-on-commit");
 560                         btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
 561                         break;
 562                 case Opt_ratio:
 563                         ret = match_int(&args[0], &intarg);
 564                         if (ret) {
 565                                 goto out;
 566                         } else if (intarg >= 0) {
 567                                 info->metadata_ratio = intarg;
 568                                 btrfs_info(root->fs_info, "metadata ratio %d",
 569                                        info->metadata_ratio);
 570                         } else {
 571                                 ret = -EINVAL;
 572                                 goto out;
 573                         }
 574                         break;
 575                 case Opt_discard:
 576                         btrfs_set_opt(info->mount_opt, DISCARD);
 577                         break;
 578                 case Opt_space_cache:
 579                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 580                         break;
 581                 case Opt_rescan_uuid_tree:
 582                         btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 583                         break;
 584                 case Opt_no_space_cache:
 585                         btrfs_info(root->fs_info, "disabling disk space caching");
 586                         btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
 587                         break;
 588                 case Opt_inode_cache:
 589                         btrfs_info(root->fs_info, "enabling inode map caching");
 590                         btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
 591                         break;
 592                 case Opt_clear_cache:
 593                         btrfs_info(root->fs_info, "force clearing of disk cache");
 594                         btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
 595                         break;
 596                 case Opt_user_subvol_rm_allowed:
 597                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 598                         break;
 599                 case Opt_enospc_debug:
 600                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 601                         break;
 602                 case Opt_defrag:
 603                         btrfs_info(root->fs_info, "enabling auto defrag");
 604                         btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
 605                         break;
 606                 case Opt_nodefrag:
 607                         if (btrfs_test_opt(root, AUTO_DEFRAG))
 608                                 btrfs_info(root->fs_info, "disabling auto defrag");
 609                         btrfs_clear_opt(info->mount_opt, AUTO_DEFRAG);
 610                         break;
 611                 case Opt_recovery:
 612                         btrfs_info(root->fs_info, "enabling auto recovery");
 613                         btrfs_set_opt(info->mount_opt, RECOVERY);
 614                         break;
 615                 case Opt_skip_balance:
 616                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 617                         break;
 618 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 619                 case Opt_check_integrity_including_extent_data:
 620                         btrfs_info(root->fs_info,
 621                                    "enabling check integrity including extent data");
 622                         btrfs_set_opt(info->mount_opt,
 623                                       CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
 624                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 625                         break;
 626                 case Opt_check_integrity:
 627                         btrfs_info(root->fs_info, "enabling check integrity");
 628                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 629                         break;
 630                 case Opt_check_integrity_print_mask:
 631                         ret = match_int(&args[0], &intarg);
 632                         if (ret) {
 633                                 goto out;
 634                         } else if (intarg >= 0) {
 635                                 info->check_integrity_print_mask = intarg;
 636                                 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
 637                                        info->check_integrity_print_mask);
 638                         } else {
 639                                 ret = -EINVAL;
 640                                 goto out;
 641                         }
 642                         break;
 643 #else
 644                 case Opt_check_integrity_including_extent_data:
 645                 case Opt_check_integrity:
 646                 case Opt_check_integrity_print_mask:
 647                         btrfs_err(root->fs_info,
 648                                 "support for check_integrity* not compiled in!");
 649                         ret = -EINVAL;
 650                         goto out;
 651 #endif
 652                 case Opt_fatal_errors:
 653                         if (strcmp(args[0].from, "panic") == 0)
 654                                 btrfs_set_opt(info->mount_opt,
 655                                               PANIC_ON_FATAL_ERROR);
 656                         else if (strcmp(args[0].from, "bug") == 0)
 657                                 btrfs_clear_opt(info->mount_opt,
 658                                               PANIC_ON_FATAL_ERROR);
 659                         else {
 660                                 ret = -EINVAL;
 661                                 goto out;
 662                         }
 663                         break;
 664                 case Opt_commit_interval:
 665                         intarg = 0;
 666                         ret = match_int(&args[0], &intarg);
 667                         if (ret < 0) {
 668                                 btrfs_err(root->fs_info, "invalid commit interval");
 669                                 ret = -EINVAL;
 670                                 goto out;
 671                         }
 672                         if (intarg > 0) {
 673                                 if (intarg > 300) {
 674                                         btrfs_warn(root->fs_info, "excessive commit interval %d",
 675                                                         intarg);
 676                                 }
 677                                 info->commit_interval = intarg;
 678                         } else {
 679                                 btrfs_info(root->fs_info, "using default commit interval %ds",
 680                                     BTRFS_DEFAULT_COMMIT_INTERVAL);
 681                                 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
 682                         }
 683                         break;
 684                 case Opt_err:
 685                         btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
 686                         ret = -EINVAL;
 687                         goto out;
 688                 default:
 689                         break;
 690                 }
 691         }
 692 out:
 693         if (!ret && btrfs_test_opt(root, SPACE_CACHE))
 694                 btrfs_info(root->fs_info, "disk space caching is enabled");
 695         kfree(orig);
 696         return ret;
 697 }
 698
 699 /*
 700  * Parse mount options that are required early in the mount process.
 701  *
 702  * All other options will be parsed on much later in the mount process and
 703  * only when we need to allocate a new super block.
 704  */
 705 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 706                 void *holder, char **subvol_name, u64 *subvol_objectid,
 707                 struct btrfs_fs_devices **fs_devices)
 708 {
 709         substring_t args[MAX_OPT_ARGS];
 710         char *device_name, *opts, *orig, *p;
 711         char *num = NULL;
 712         int error = 0;
 713
 714         if (!options)
 715                 return 0;
 716
 717         /*
 718          * strsep changes the string, duplicate it because parse_options
 719          * gets called twice
 720          */
 721         opts = kstrdup(options, GFP_KERNEL);
 722         if (!opts)
 723                 return -ENOMEM;
 724         orig = opts;
 725
 726         while ((p = strsep(&opts, ",")) != NULL) {
 727                 int token;
 728                 if (!*p)
 729                         continue;
 730
 731                 token = match_token(p, tokens, args);
 732                 switch (token) {
 733                 case Opt_subvol:
 734                         kfree(*subvol_name);
 735                         *subvol_name = match_strdup(&args[0]);
 736                         if (!*subvol_name) {
 737                                 error = -ENOMEM;
 738                                 goto out;
 739                         }
 740                         break;
 741                 case Opt_subvolid:
 742                         num = match_strdup(&args[0]);
 743                         if (num) {
 744                                 *subvol_objectid = memparse(num, NULL);
 745                                 kfree(num);
 746                                 /* we want the original fs_tree */
 747                                 if (!*subvol_objectid)
 748                                         *subvol_objectid =
 749                                                 BTRFS_FS_TREE_OBJECTID;
 750                         } else {
 751                                 error = -EINVAL;
 752                                 goto out;
 753                         }
 754                         break;
 755                 case Opt_subvolrootid:
 756                         printk(KERN_WARNING
 757                                 "BTRFS: 'subvolrootid' mount option is deprecated and has "
 758                                 "no effect\n");
 759                         break;
 760                 case Opt_device:
 761                         device_name = match_strdup(&args[0]);
 762                         if (!device_name) {
 763                                 error = -ENOMEM;
 764                                 goto out;
 765                         }
 766                         error = btrfs_scan_one_device(device_name,
 767                                         flags, holder, fs_devices);
 768                         kfree(device_name);
 769                         if (error)
 770                                 goto out;
 771                         break;
 772                 default:
 773                         break;
 774                 }
 775         }
 776
 777 out:
 778         kfree(orig);
 779         return error;
 780 }
 781
 782 static struct dentry *get_default_root(struct super_block *sb,
 783                                        u64 subvol_objectid)
 784 {
 785         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 786         struct btrfs_root *root = fs_info->tree_root;
 787         struct btrfs_root *new_root;
 788         struct btrfs_dir_item *di;
 789         struct btrfs_path *path;
 790         struct btrfs_key location;
 791         struct inode *inode;
 792         u64 dir_id;
 793         int new = 0;
 794
 795         /*
 796          * We have a specific subvol we want to mount, just setup location and
 797          * go look up the root.
 798          */
 799         if (subvol_objectid) {
 800                 location.objectid = subvol_objectid;
 801                 location.type = BTRFS_ROOT_ITEM_KEY;
 802                 location.offset = (u64)-1;
 803                 goto find_root;
 804         }
 805
 806         path = btrfs_alloc_path();
 807         if (!path)
 808                 return ERR_PTR(-ENOMEM);
 809         path->leave_spinning = 1;
 810
 811         /*
 812          * Find the "default" dir item which points to the root item that we
 813          * will mount by default if we haven't been given a specific subvolume
 814          * to mount.
 815          */
 816         dir_id = btrfs_super_root_dir(fs_info->super_copy);
 817         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 818         if (IS_ERR(di)) {
 819                 btrfs_free_path(path);
 820                 return ERR_CAST(di);
 821         }
 822         if (!di) {
 823                 /*
 824                  * Ok the default dir item isn't there.  This is weird since
 825                  * it's always been there, but don't freak out, just try and
 826                  * mount to root most subvolume.
 827                  */
 828                 btrfs_free_path(path);
 829                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 830                 new_root = fs_info->fs_root;
 831                 goto setup_root;
 832         }
 833
 834         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 835         btrfs_free_path(path);
 836
 837 find_root:
 838         new_root = btrfs_read_fs_root_no_name(fs_info, &location);
 839         if (IS_ERR(new_root))
 840                 return ERR_CAST(new_root);
 841
 842         dir_id = btrfs_root_dirid(&new_root->root_item);
 843 setup_root:
 844         location.objectid = dir_id;
 845         location.type = BTRFS_INODE_ITEM_KEY;
 846         location.offset = 0;
 847
 848         inode = btrfs_iget(sb, &location, new_root, &new);
 849         if (IS_ERR(inode))
 850                 return ERR_CAST(inode);
 851
 852         /*
 853          * If we're just mounting the root most subvol put the inode and return
 854          * a reference to the dentry.  We will have already gotten a reference
 855          * to the inode in btrfs_fill_super so we're good to go.
 856          */
 857         if (!new && sb->s_root->d_inode == inode) {
 858                 iput(inode);
 859                 return dget(sb->s_root);
 860         }
 861
 862         return d_obtain_alias(inode);
 863 }
 864
 865 static int btrfs_fill_super(struct super_block *sb,
 866                             struct btrfs_fs_devices *fs_devices,
 867                             void *data, int silent)
 868 {
 869         struct inode *inode;
 870         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 871         struct btrfs_key key;
 872         int err;
 873
 874         sb->s_maxbytes = MAX_LFS_FILESIZE;
 875         sb->s_magic = BTRFS_SUPER_MAGIC;
 876         sb->s_op = &btrfs_super_ops;
 877         sb->s_d_op = &btrfs_dentry_operations;
 878         sb->s_export_op = &btrfs_export_ops;
 879         sb->s_xattr = btrfs_xattr_handlers;
 880         sb->s_time_gran = 1;
 881 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 882         sb->s_flags |= MS_POSIXACL;
 883 #endif
 884         sb->s_flags |= MS_I_VERSION;
 885         err = open_ctree(sb, fs_devices, (char *)data);
 886         if (err) {
 887                 printk(KERN_ERR "BTRFS: open_ctree failed\n");
 888                 return err;
 889         }
 890
 891         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 892         key.type = BTRFS_INODE_ITEM_KEY;
 893         key.offset = 0;
 894         inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
 895         if (IS_ERR(inode)) {
 896                 err = PTR_ERR(inode);
 897                 goto fail_close;
 898         }
 899
 900         sb->s_root = d_make_root(inode);
 901         if (!sb->s_root) {
 902                 err = -ENOMEM;
 903                 goto fail_close;
 904         }
 905
 906         save_mount_options(sb, data);
 907         cleancache_init_fs(sb);
 908         sb->s_flags |= MS_ACTIVE;
 909         return 0;
 910
 911 fail_close:
 912         close_ctree(fs_info->tree_root);
 913         return err;
 914 }
 915
 916 int btrfs_sync_fs(struct super_block *sb, int wait)
 917 {
 918         struct btrfs_trans_handle *trans;
 919         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 920         struct btrfs_root *root = fs_info->tree_root;
 921
 922         trace_btrfs_sync_fs(wait);
 923
 924         if (!wait) {
 925                 filemap_flush(fs_info->btree_inode->i_mapping);
 926                 return 0;
 927         }
 928
 929         btrfs_wait_ordered_roots(fs_info, -1);
 930
 931         trans = btrfs_attach_transaction_barrier(root);
 932         if (IS_ERR(trans)) {
 933                 /* no transaction, don't bother */
 934                 if (PTR_ERR(trans) == -ENOENT)
 935                         return 0;
 936                 return PTR_ERR(trans);
 937         }
 938         return btrfs_commit_transaction(trans, root);
 939 }
 940
 941 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
 942 {
 943         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
 944         struct btrfs_root *root = info->tree_root;
 945         char *compress_type;
 946
 947         if (btrfs_test_opt(root, DEGRADED))
 948                 seq_puts(seq, ",degraded");
 949         if (btrfs_test_opt(root, NODATASUM))
 950                 seq_puts(seq, ",nodatasum");
 951         if (btrfs_test_opt(root, NODATACOW))
 952                 seq_puts(seq, ",nodatacow");
 953         if (btrfs_test_opt(root, NOBARRIER))
 954                 seq_puts(seq, ",nobarrier");
 955         if (info->max_inline != 8192 * 1024)
 956                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
 957         if (info->alloc_start != 0)
 958                 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
 959         if (info->thread_pool_size !=  min_t(unsigned long,
 960                                              num_online_cpus() + 2, 8))
 961                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
 962         if (btrfs_test_opt(root, COMPRESS)) {
 963                 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
 964                         compress_type = "zlib";
 965                 else
 966                         compress_type = "lzo";
 967                 if (btrfs_test_opt(root, FORCE_COMPRESS))
 968                         seq_printf(seq, ",compress-force=%s", compress_type);
 969                 else
 970                         seq_printf(seq, ",compress=%s", compress_type);
 971         }
 972         if (btrfs_test_opt(root, NOSSD))
 973                 seq_puts(seq, ",nossd");
 974         if (btrfs_test_opt(root, SSD_SPREAD))
 975                 seq_puts(seq, ",ssd_spread");
 976         else if (btrfs_test_opt(root, SSD))
 977                 seq_puts(seq, ",ssd");
 978         if (btrfs_test_opt(root, NOTREELOG))
 979                 seq_puts(seq, ",notreelog");
 980         if (btrfs_test_opt(root, FLUSHONCOMMIT))
 981                 seq_puts(seq, ",flushoncommit");
 982         if (btrfs_test_opt(root, DISCARD))
 983                 seq_puts(seq, ",discard");
 984         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
 985                 seq_puts(seq, ",noacl");
 986         if (btrfs_test_opt(root, SPACE_CACHE))
 987                 seq_puts(seq, ",space_cache");
 988         else
 989                 seq_puts(seq, ",nospace_cache");
 990         if (btrfs_test_opt(root, RESCAN_UUID_TREE))
 991                 seq_puts(seq, ",rescan_uuid_tree");
 992         if (btrfs_test_opt(root, CLEAR_CACHE))
 993                 seq_puts(seq, ",clear_cache");
 994         if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
 995                 seq_puts(seq, ",user_subvol_rm_allowed");
 996         if (btrfs_test_opt(root, ENOSPC_DEBUG))
 997                 seq_puts(seq, ",enospc_debug");
 998         if (btrfs_test_opt(root, AUTO_DEFRAG))
 999                 seq_puts(seq, ",autodefrag");
1000         if (btrfs_test_opt(root, INODE_MAP_CACHE))
1001                 seq_puts(seq, ",inode_cache");
1002         if (btrfs_test_opt(root, SKIP_BALANCE))
1003                 seq_puts(seq, ",skip_balance");
1004         if (btrfs_test_opt(root, RECOVERY))
1005                 seq_puts(seq, ",recovery");
1006 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1007         if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1008                 seq_puts(seq, ",check_int_data");
1009         else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1010                 seq_puts(seq, ",check_int");
1011         if (info->check_integrity_print_mask)
1012                 seq_printf(seq, ",check_int_print_mask=%d",
1013                                 info->check_integrity_print_mask);
1014 #endif
1015         if (info->metadata_ratio)
1016                 seq_printf(seq, ",metadata_ratio=%d",
1017                                 info->metadata_ratio);
1018         if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1019                 seq_puts(seq, ",fatal_errors=panic");
1020         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1021                 seq_printf(seq, ",commit=%d", info->commit_interval);
1022         return 0;
1023 }
1024
1025 static int btrfs_test_super(struct super_block *s, void *data)
1026 {
1027         struct btrfs_fs_info *p = data;
1028         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1029
1030         return fs_info->fs_devices == p->fs_devices;
1031 }
1032
1033 static int btrfs_set_super(struct super_block *s, void *data)
1034 {
1035         int err = set_anon_super(s, data);
1036         if (!err)
1037                 s->s_fs_info = data;
1038         return err;
1039 }
1040
1041 /*
1042  * subvolumes are identified by ino 256
1043  */
1044 static inline int is_subvolume_inode(struct inode *inode)
1045 {
1046         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1047                 return 1;
1048         return 0;
1049 }
1050
1051 /*
1052  * This will strip out the subvol=%s argument for an argument string and add
1053  * subvolid=0 to make sure we get the actual tree root for path walking to the
1054  * subvol we want.
1055  */
1056 static char *setup_root_args(char *args)
1057 {
1058         unsigned len = strlen(args) + 2 + 1;
1059         char *src, *dst, *buf;
1060
1061         /*
1062          * We need the same args as before, but with this substitution:
1063          * s!subvol=[^,]+!subvolid=0!
1064          *
1065          * Since the replacement string is up to 2 bytes longer than the
1066          * original, allocate strlen(args) + 2 + 1 bytes.
1067          */
1068
1069         src = strstr(args, "subvol=");
1070         /* This shouldn't happen, but just in case.. */
1071         if (!src)
1072                 return NULL;
1073
1074         buf = dst = kmalloc(len, GFP_NOFS);
1075         if (!buf)
1076                 return NULL;
1077
1078         /*
1079          * If the subvol= arg is not at the start of the string,
1080          * copy whatever precedes it into buf.
1081          */
1082         if (src != args) {
1083                 *src++ = '\0';
1084                 strcpy(buf, args);
1085                 dst += strlen(args);
1086         }
1087
1088         strcpy(dst, "subvolid=0");
1089         dst += strlen("subvolid=0");
1090
1091         /*
1092          * If there is a "," after the original subvol=... string,
1093          * copy that suffix into our buffer.  Otherwise, we're done.
1094          */
1095         src = strchr(src, ',');
1096         if (src)
1097                 strcpy(dst, src);
1098
1099         return buf;
1100 }
1101
1102 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1103                                    const char *device_name, char *data)
1104 {
1105         struct dentry *root;
1106         struct vfsmount *mnt;
1107         char *newargs;
1108
1109         newargs = setup_root_args(data);
1110         if (!newargs)
1111                 return ERR_PTR(-ENOMEM);
1112         mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1113                              newargs);
1114         kfree(newargs);
1115         if (IS_ERR(mnt))
1116                 return ERR_CAST(mnt);
1117
1118         root = mount_subtree(mnt, subvol_name);
1119
1120         if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1121                 struct super_block *s = root->d_sb;
1122                 dput(root);
1123                 root = ERR_PTR(-EINVAL);
1124                 deactivate_locked_super(s);
1125                 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1126                                 subvol_name);
1127         }
1128
1129         return root;
1130 }
1131
1132 /*
1133  * Find a superblock for the given device / mount point.
1134  *
1135  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
1136  *        for multiple device setup.  Make sure to keep it in sync.
1137  */
1138 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1139                 const char *device_name, void *data)
1140 {
1141         struct block_device *bdev = NULL;
1142         struct super_block *s;
1143         struct dentry *root;
1144         struct btrfs_fs_devices *fs_devices = NULL;
1145         struct btrfs_fs_info *fs_info = NULL;
1146         fmode_t mode = FMODE_READ;
1147         char *subvol_name = NULL;
1148         u64 subvol_objectid = 0;
1149         int error = 0;
1150
1151         if (!(flags & MS_RDONLY))
1152                 mode |= FMODE_WRITE;
1153
1154         error = btrfs_parse_early_options(data, mode, fs_type,
1155                                           &subvol_name, &subvol_objectid,
1156                                           &fs_devices);
1157         if (error) {
1158                 kfree(subvol_name);
1159                 return ERR_PTR(error);
1160         }
1161
1162         if (subvol_name) {
1163                 root = mount_subvol(subvol_name, flags, device_name, data);
1164                 kfree(subvol_name);
1165                 return root;
1166         }
1167
1168         error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1169         if (error)
1170                 return ERR_PTR(error);
1171
1172         /*
1173          * Setup a dummy root and fs_info for test/set super.  This is because
1174          * we don't actually fill this stuff out until open_ctree, but we need
1175          * it for searching for existing supers, so this lets us do that and
1176          * then open_ctree will properly initialize everything later.
1177          */
1178         fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1179         if (!fs_info)
1180                 return ERR_PTR(-ENOMEM);
1181
1182         fs_info->fs_devices = fs_devices;
1183
1184         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1185         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1186         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1187                 error = -ENOMEM;
1188                 goto error_fs_info;
1189         }
1190
1191         error = btrfs_open_devices(fs_devices, mode, fs_type);
1192         if (error)
1193                 goto error_fs_info;
1194
1195         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1196                 error = -EACCES;
1197                 goto error_close_devices;
1198         }
1199
1200         bdev = fs_devices->latest_bdev;
1201         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1202                  fs_info);
1203         if (IS_ERR(s)) {
1204                 error = PTR_ERR(s);
1205                 goto error_close_devices;
1206         }
1207
1208         if (s->s_root) {
1209                 btrfs_close_devices(fs_devices);
1210                 free_fs_info(fs_info);
1211                 if ((flags ^ s->s_flags) & MS_RDONLY)
1212                         error = -EBUSY;
1213         } else {
1214                 char b[BDEVNAME_SIZE];
1215
1216                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1217                 btrfs_sb(s)->bdev_holder = fs_type;
1218                 error = btrfs_fill_super(s, fs_devices, data,
1219                                          flags & MS_SILENT ? 1 : 0);
1220         }
1221
1222         root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1223         if (IS_ERR(root))
1224                 deactivate_locked_super(s);
1225
1226         return root;
1227
1228 error_close_devices:
1229         btrfs_close_devices(fs_devices);
1230 error_fs_info:
1231         free_fs_info(fs_info);
1232         return ERR_PTR(error);
1233 }
1234
1235 static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1236 {
1237         spin_lock_irq(&workers->lock);
1238         workers->max_workers = new_limit;
1239         spin_unlock_irq(&workers->lock);
1240 }
1241
1242 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1243                                      int new_pool_size, int old_pool_size)
1244 {
1245         if (new_pool_size == old_pool_size)
1246                 return;
1247
1248         fs_info->thread_pool_size = new_pool_size;
1249
1250         btrfs_info(fs_info, "resize thread pool %d -> %d",
1251                old_pool_size, new_pool_size);
1252
1253         btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1254         btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1255         btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1256         btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1257         btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1258         btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1259         btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1260         btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1261         btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1262         btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1263         btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1264         btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1265         btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
1266         btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers,
1267                               new_pool_size);
1268 }
1269
1270 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1271 {
1272         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1273 }
1274
1275 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1276                                        unsigned long old_opts, int flags)
1277 {
1278         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1279             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1280              (flags & MS_RDONLY))) {
1281                 /* wait for any defraggers to finish */
1282                 wait_event(fs_info->transaction_wait,
1283                            (atomic_read(&fs_info->defrag_running) == 0));
1284                 if (flags & MS_RDONLY)
1285                         sync_filesystem(fs_info->sb);
1286         }
1287 }
1288
1289 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1290                                          unsigned long old_opts)
1291 {
1292         /*
1293          * We need cleanup all defragable inodes if the autodefragment is
1294          * close or the fs is R/O.
1295          */
1296         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1297             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1298              (fs_info->sb->s_flags & MS_RDONLY))) {
1299                 btrfs_cleanup_defrag_inodes(fs_info);
1300         }
1301
1302         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1303 }
1304
1305 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1306 {
1307         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1308         struct btrfs_root *root = fs_info->tree_root;
1309         unsigned old_flags = sb->s_flags;
1310         unsigned long old_opts = fs_info->mount_opt;
1311         unsigned long old_compress_type = fs_info->compress_type;
1312         u64 old_max_inline = fs_info->max_inline;
1313         u64 old_alloc_start = fs_info->alloc_start;
1314         int old_thread_pool_size = fs_info->thread_pool_size;
1315         unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1316         int ret;
1317
1318         btrfs_remount_prepare(fs_info);
1319
1320         ret = btrfs_parse_options(root, data);
1321         if (ret) {
1322                 ret = -EINVAL;
1323                 goto restore;
1324         }
1325
1326         btrfs_remount_begin(fs_info, old_opts, *flags);
1327         btrfs_resize_thread_pool(fs_info,
1328                 fs_info->thread_pool_size, old_thread_pool_size);
1329
1330         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1331                 goto out;
1332
1333         if (*flags & MS_RDONLY) {
1334                 /*
1335                  * this also happens on 'umount -rf' or on shutdown, when
1336                  * the filesystem is busy.
1337                  */
1338
1339                 /* wait for the uuid_scan task to finish */
1340                 down(&fs_info->uuid_tree_rescan_sem);
1341                 /* avoid complains from lockdep et al. */
1342                 up(&fs_info->uuid_tree_rescan_sem);
1343
1344                 sb->s_flags |= MS_RDONLY;
1345
1346                 btrfs_dev_replace_suspend_for_unmount(fs_info);
1347                 btrfs_scrub_cancel(fs_info);
1348                 btrfs_pause_balance(fs_info);
1349
1350                 ret = btrfs_commit_super(root);
1351                 if (ret)
1352                         goto restore;
1353         } else {
1354                 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1355                         btrfs_err(fs_info,
1356                                 "Remounting read-write after error is not allowed");
1357                         ret = -EINVAL;
1358                         goto restore;
1359                 }
1360                 if (fs_info->fs_devices->rw_devices == 0) {
1361                         ret = -EACCES;
1362                         goto restore;
1363                 }
1364
1365                 if (fs_info->fs_devices->missing_devices >
1366                      fs_info->num_tolerated_disk_barrier_failures &&
1367                     !(*flags & MS_RDONLY)) {
1368                         btrfs_warn(fs_info,
1369                                 "too many missing devices, writeable remount is not allowed");
1370                         ret = -EACCES;
1371                         goto restore;
1372                 }
1373
1374                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1375                         ret = -EINVAL;
1376                         goto restore;
1377                 }
1378
1379                 ret = btrfs_cleanup_fs_roots(fs_info);
1380                 if (ret)
1381                         goto restore;
1382
1383                 /* recover relocation */
1384                 ret = btrfs_recover_relocation(root);
1385                 if (ret)
1386                         goto restore;
1387
1388                 ret = btrfs_resume_balance_async(fs_info);
1389                 if (ret)
1390                         goto restore;
1391
1392                 ret = btrfs_resume_dev_replace_async(fs_info);
1393                 if (ret) {
1394                         btrfs_warn(fs_info, "failed to resume dev_replace");
1395                         goto restore;
1396                 }
1397
1398                 if (!fs_info->uuid_root) {
1399                         btrfs_info(fs_info, "creating UUID tree");
1400                         ret = btrfs_create_uuid_tree(fs_info);
1401                         if (ret) {
1402                                 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1403                                 goto restore;
1404                         }
1405                 }
1406                 sb->s_flags &= ~MS_RDONLY;
1407         }
1408 out:
1409         btrfs_remount_cleanup(fs_info, old_opts);
1410         return 0;
1411
1412 restore:
1413         /* We've hit an error - don't reset MS_RDONLY */
1414         if (sb->s_flags & MS_RDONLY)
1415                 old_flags |= MS_RDONLY;
1416         sb->s_flags = old_flags;
1417         fs_info->mount_opt = old_opts;
1418         fs_info->compress_type = old_compress_type;
1419         fs_info->max_inline = old_max_inline;
1420         mutex_lock(&fs_info->chunk_mutex);
1421         fs_info->alloc_start = old_alloc_start;
1422         mutex_unlock(&fs_info->chunk_mutex);
1423         btrfs_resize_thread_pool(fs_info,
1424                 old_thread_pool_size, fs_info->thread_pool_size);
1425         fs_info->metadata_ratio = old_metadata_ratio;
1426         btrfs_remount_cleanup(fs_info, old_opts);
1427         return ret;
1428 }
1429
1430 /* Used to sort the devices by max_avail(descending sort) */
1431 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1432                                        const void *dev_info2)
1433 {
1434         if (((struct btrfs_device_info *)dev_info1)->max_avail >
1435             ((struct btrfs_device_info *)dev_info2)->max_avail)
1436                 return -1;
1437         else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1438                  ((struct btrfs_device_info *)dev_info2)->max_avail)
1439                 return 1;
1440         else
1441         return 0;
1442 }
1443
1444 /*
1445  * sort the devices by max_avail, in which max free extent size of each device
1446  * is stored.(Descending Sort)
1447  */
1448 static inline void btrfs_descending_sort_devices(
1449                                         struct btrfs_device_info *devices,
1450                                         size_t nr_devices)
1451 {
1452         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1453              btrfs_cmp_device_free_bytes, NULL);
1454 }
1455
1456 /*
1457  * The helper to calc the free space on the devices that can be used to store
1458  * file data.
1459  */
1460 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1461 {
1462         struct btrfs_fs_info *fs_info = root->fs_info;
1463         struct btrfs_device_info *devices_info;
1464         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1465         struct btrfs_device *device;
1466         u64 skip_space;
1467         u64 type;
1468         u64 avail_space;
1469         u64 used_space;
1470         u64 min_stripe_size;
1471         int min_stripes = 1, num_stripes = 1;
1472         int i = 0, nr_devices;
1473         int ret;
1474
1475         nr_devices = fs_info->fs_devices->open_devices;
1476         BUG_ON(!nr_devices);
1477
1478         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1479                                GFP_NOFS);
1480         if (!devices_info)
1481                 return -ENOMEM;
1482
1483         /* calc min stripe number for data space alloction */
1484         type = btrfs_get_alloc_profile(root, 1);
1485         if (type & BTRFS_BLOCK_GROUP_RAID0) {
1486                 min_stripes = 2;
1487                 num_stripes = nr_devices;
1488         } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1489                 min_stripes = 2;
1490                 num_stripes = 2;
1491         } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1492                 min_stripes = 4;
1493                 num_stripes = 4;
1494         }
1495
1496         if (type & BTRFS_BLOCK_GROUP_DUP)
1497                 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1498         else
1499                 min_stripe_size = BTRFS_STRIPE_LEN;
1500
1501         list_for_each_entry(device, &fs_devices->devices, dev_list) {
1502                 if (!device->in_fs_metadata || !device->bdev ||
1503                     device->is_tgtdev_for_dev_replace)
1504                         continue;
1505
1506                 avail_space = device->total_bytes - device->bytes_used;
1507
1508                 /* align with stripe_len */
1509                 do_div(avail_space, BTRFS_STRIPE_LEN);
1510                 avail_space *= BTRFS_STRIPE_LEN;
1511
1512                 /*
1513                  * In order to avoid overwritting the superblock on the drive,
1514                  * btrfs starts at an offset of at least 1MB when doing chunk
1515                  * allocation.
1516                  */
1517                 skip_space = 1024 * 1024;
1518
1519                 /* user can set the offset in fs_info->alloc_start. */
1520                 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1521                     device->total_bytes)
1522                         skip_space = max(fs_info->alloc_start, skip_space);
1523
1524                 /*
1525                  * btrfs can not use the free space in [0, skip_space - 1],
1526                  * we must subtract it from the total. In order to implement
1527                  * it, we account the used space in this range first.
1528                  */
1529                 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1530                                                      &used_space);
1531                 if (ret) {
1532                         kfree(devices_info);
1533                         return ret;
1534                 }
1535
1536                 /* calc the free space in [0, skip_space - 1] */
1537                 skip_space -= used_space;
1538
1539                 /*
1540                  * we can use the free space in [0, skip_space - 1], subtract
1541                  * it from the total.
1542                  */
1543                 if (avail_space && avail_space >= skip_space)
1544                         avail_space -= skip_space;
1545                 else
1546                         avail_space = 0;
1547
1548                 if (avail_space < min_stripe_size)
1549                         continue;
1550
1551                 devices_info[i].dev = device;
1552                 devices_info[i].max_avail = avail_space;
1553
1554                 i++;
1555         }
1556
1557         nr_devices = i;
1558
1559         btrfs_descending_sort_devices(devices_info, nr_devices);
1560
1561         i = nr_devices - 1;
1562         avail_space = 0;
1563         while (nr_devices >= min_stripes) {
1564                 if (num_stripes > nr_devices)
1565                         num_stripes = nr_devices;
1566
1567                 if (devices_info[i].max_avail >= min_stripe_size) {
1568                         int j;
1569                         u64 alloc_size;
1570
1571                         avail_space += devices_info[i].max_avail * num_stripes;
1572                         alloc_size = devices_info[i].max_avail;
1573                         for (j = i + 1 - num_stripes; j <= i; j++)
1574                                 devices_info[j].max_avail -= alloc_size;
1575                 }
1576                 i--;
1577                 nr_devices--;
1578         }
1579
1580         kfree(devices_info);
1581         *free_bytes = avail_space;
1582         return 0;
1583 }
1584
1585 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1586 {
1587         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1588         struct btrfs_super_block *disk_super = fs_info->super_copy;
1589         struct list_head *head = &fs_info->space_info;
1590         struct btrfs_space_info *found;
1591         u64 total_used = 0;
1592         u64 total_free_data = 0;
1593         int bits = dentry->d_sb->s_blocksize_bits;
1594         __be32 *fsid = (__be32 *)fs_info->fsid;
1595         int ret;
1596
1597         /* holding chunk_muext to avoid allocating new chunks */
1598         mutex_lock(&fs_info->chunk_mutex);
1599         rcu_read_lock();
1600         list_for_each_entry_rcu(found, head, list) {
1601                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1602                         total_free_data += found->disk_total - found->disk_used;
1603                         total_free_data -=
1604                                 btrfs_account_ro_block_groups_free_space(found);
1605                 }
1606
1607                 total_used += found->disk_used;
1608         }
1609         rcu_read_unlock();
1610
1611         buf->f_namelen = BTRFS_NAME_LEN;
1612         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1613         buf->f_bfree = buf->f_blocks - (total_used >> bits);
1614         buf->f_bsize = dentry->d_sb->s_blocksize;
1615         buf->f_type = BTRFS_SUPER_MAGIC;
1616         buf->f_bavail = total_free_data;
1617         ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1618         if (ret) {
1619                 mutex_unlock(&fs_info->chunk_mutex);
1620                 return ret;
1621         }
1622         buf->f_bavail += total_free_data;
1623         buf->f_bavail = buf->f_bavail >> bits;
1624         mutex_unlock(&fs_info->chunk_mutex);
1625
1626         /* We treat it as constant endianness (it doesn't matter _which_)
1627            because we want the fsid to come out the same whether mounted
1628            on a big-endian or little-endian host */
1629         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1630         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1631         /* Mask in the root object ID too, to disambiguate subvols */
1632         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1633         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1634
1635         return 0;
1636 }
1637
1638 static void btrfs_kill_super(struct super_block *sb)
1639 {
1640         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1641         kill_anon_super(sb);
1642         free_fs_info(fs_info);
1643 }
1644
1645 static struct file_system_type btrfs_fs_type = {
1646         .owner          = THIS_MODULE,
1647         .name           = "btrfs",
1648         .mount          = btrfs_mount,
1649         .kill_sb        = btrfs_kill_super,
1650         .fs_flags       = FS_REQUIRES_DEV,
1651 };
1652 MODULE_ALIAS_FS("btrfs");
1653
1654 /*
1655  * used by btrfsctl to scan devices when no FS is mounted
1656  */
1657 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1658                                 unsigned long arg)
1659 {
1660         struct btrfs_ioctl_vol_args *vol;
1661         struct btrfs_fs_devices *fs_devices;
1662         int ret = -ENOTTY;
1663
1664         if (!capable(CAP_SYS_ADMIN))
1665                 return -EPERM;
1666
1667         vol = memdup_user((void __user *)arg, sizeof(*vol));
1668         if (IS_ERR(vol))
1669                 return PTR_ERR(vol);
1670
1671         switch (cmd) {
1672         case BTRFS_IOC_SCAN_DEV:
1673                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1674                                             &btrfs_fs_type, &fs_devices);
1675                 break;
1676         case BTRFS_IOC_DEVICES_READY:
1677                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1678                                             &btrfs_fs_type, &fs_devices);
1679                 if (ret)
1680                         break;
1681                 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1682                 break;
1683         }
1684
1685         kfree(vol);
1686         return ret;
1687 }
1688
1689 static int btrfs_freeze(struct super_block *sb)
1690 {
1691         struct btrfs_trans_handle *trans;
1692         struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1693
1694         trans = btrfs_attach_transaction_barrier(root);
1695         if (IS_ERR(trans)) {
1696                 /* no transaction, don't bother */
1697                 if (PTR_ERR(trans) == -ENOENT)
1698                         return 0;
1699                 return PTR_ERR(trans);
1700         }
1701         return btrfs_commit_transaction(trans, root);
1702 }
1703
1704 static int btrfs_unfreeze(struct super_block *sb)
1705 {
1706         return 0;
1707 }
1708
1709 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1710 {
1711         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1712         struct btrfs_fs_devices *cur_devices;
1713         struct btrfs_device *dev, *first_dev = NULL;
1714         struct list_head *head;
1715         struct rcu_string *name;
1716
1717         mutex_lock(&fs_info->fs_devices->device_list_mutex);
1718         cur_devices = fs_info->fs_devices;
1719         while (cur_devices) {
1720                 head = &cur_devices->devices;
1721                 list_for_each_entry(dev, head, dev_list) {
1722                         if (dev->missing)
1723                                 continue;
1724                         if (!first_dev || dev->devid < first_dev->devid)
1725                                 first_dev = dev;
1726                 }
1727                 cur_devices = cur_devices->seed;
1728         }
1729
1730         if (first_dev) {
1731                 rcu_read_lock();
1732                 name = rcu_dereference(first_dev->name);
1733                 seq_escape(m, name->str, " \t\n\\");
1734                 rcu_read_unlock();
1735         } else {
1736                 WARN_ON(1);
1737         }
1738         mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1739         return 0;
1740 }
1741
1742 static const struct super_operations btrfs_super_ops = {
1743         .drop_inode     = btrfs_drop_inode,
1744         .evict_inode    = btrfs_evict_inode,
1745         .put_super      = btrfs_put_super,
1746         .sync_fs        = btrfs_sync_fs,
1747         .show_options   = btrfs_show_options,
1748         .show_devname   = btrfs_show_devname,
1749         .write_inode    = btrfs_write_inode,
1750         .alloc_inode    = btrfs_alloc_inode,
1751         .destroy_inode  = btrfs_destroy_inode,
1752         .statfs         = btrfs_statfs,
1753         .remount_fs     = btrfs_remount,
1754         .freeze_fs      = btrfs_freeze,
1755         .unfreeze_fs    = btrfs_unfreeze,
1756 };
1757
1758 static const struct file_operations btrfs_ctl_fops = {
1759         .unlocked_ioctl  = btrfs_control_ioctl,
1760         .compat_ioctl = btrfs_control_ioctl,
1761         .owner   = THIS_MODULE,
1762         .llseek = noop_llseek,
1763 };
1764
1765 static struct miscdevice btrfs_misc = {
1766         .minor          = BTRFS_MINOR,
1767         .name           = "btrfs-control",
1768         .fops           = &btrfs_ctl_fops
1769 };
1770
1771 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1772 MODULE_ALIAS("devname:btrfs-control");
1773
1774 static int btrfs_interface_init(void)
1775 {
1776         return misc_register(&btrfs_misc);
1777 }
1778
1779 static void btrfs_interface_exit(void)
1780 {
1781         if (misc_deregister(&btrfs_misc) < 0)
1782                 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
1783 }
1784
1785 static void btrfs_print_info(void)
1786 {
1787         printk(KERN_INFO "Btrfs loaded"
1788 #ifdef CONFIG_BTRFS_DEBUG
1789                         ", debug=on"
1790 #endif
1791 #ifdef CONFIG_BTRFS_ASSERT
1792                         ", assert=on"
1793 #endif
1794 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1795                         ", integrity-checker=on"
1796 #endif
1797                         "\n");
1798 }
1799
1800 static int btrfs_run_sanity_tests(void)
1801 {
1802         int ret;
1803
1804         ret = btrfs_init_test_fs();
1805         if (ret)
1806                 return ret;
1807
1808         ret = btrfs_test_free_space_cache();
1809         if (ret)
1810                 goto out;
1811         ret = btrfs_test_extent_buffer_operations();
1812         if (ret)
1813                 goto out;
1814         ret = btrfs_test_extent_io();
1815         if (ret)
1816                 goto out;
1817         ret = btrfs_test_inodes();
1818 out:
1819         btrfs_destroy_test_fs();
1820         return ret;
1821 }
1822
1823 static int __init init_btrfs_fs(void)
1824 {
1825         int err;
1826
1827         err = btrfs_init_sysfs();
1828         if (err)
1829                 return err;
1830
1831         btrfs_init_compress();
1832
1833         err = btrfs_init_cachep();
1834         if (err)
1835                 goto free_compress;
1836
1837         err = extent_io_init();
1838         if (err)
1839                 goto free_cachep;
1840
1841         err = extent_map_init();
1842         if (err)
1843                 goto free_extent_io;
1844
1845         err = ordered_data_init();
1846         if (err)
1847                 goto free_extent_map;
1848
1849         err = btrfs_delayed_inode_init();
1850         if (err)
1851                 goto free_ordered_data;
1852
1853         err = btrfs_auto_defrag_init();
1854         if (err)
1855                 goto free_delayed_inode;
1856
1857         err = btrfs_delayed_ref_init();
1858         if (err)
1859                 goto free_auto_defrag;
1860
1861         err = btrfs_prelim_ref_init();
1862         if (err)
1863                 goto free_prelim_ref;
1864
1865         err = btrfs_interface_init();
1866         if (err)
1867                 goto free_delayed_ref;
1868
1869         btrfs_init_lockdep();
1870
1871         btrfs_print_info();
1872
1873         err = btrfs_run_sanity_tests();
1874         if (err)
1875                 goto unregister_ioctl;
1876
1877         err = register_filesystem(&btrfs_fs_type);
1878         if (err)
1879                 goto unregister_ioctl;
1880
1881         return 0;
1882
1883 unregister_ioctl:
1884         btrfs_interface_exit();
1885 free_prelim_ref:
1886         btrfs_prelim_ref_exit();
1887 free_delayed_ref:
1888         btrfs_delayed_ref_exit();
1889 free_auto_defrag:
1890         btrfs_auto_defrag_exit();
1891 free_delayed_inode:
1892         btrfs_delayed_inode_exit();
1893 free_ordered_data:
1894         ordered_data_exit();
1895 free_extent_map:
1896         extent_map_exit();
1897 free_extent_io:
1898         extent_io_exit();
1899 free_cachep:
1900         btrfs_destroy_cachep();
1901 free_compress:
1902         btrfs_exit_compress();
1903         btrfs_exit_sysfs();
1904         return err;
1905 }
1906
1907 static void __exit exit_btrfs_fs(void)
1908 {
1909         btrfs_destroy_cachep();
1910         btrfs_delayed_ref_exit();
1911         btrfs_auto_defrag_exit();
1912         btrfs_delayed_inode_exit();
1913         btrfs_prelim_ref_exit();
1914         ordered_data_exit();
1915         extent_map_exit();
1916         extent_io_exit();
1917         btrfs_interface_exit();
1918         unregister_filesystem(&btrfs_fs_type);
1919         btrfs_exit_sysfs();
1920         btrfs_cleanup_fs_uuids();
1921         btrfs_exit_compress();
1922 }
1923
1924 module_init(init_btrfs_fs)
1925 module_exit(exit_btrfs_fs)
1926
1927 MODULE_LICENSE("GPL");