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 "hash.h"
  52 #include "props.h"
  53 #include "xattr.h"
  54 #include "volumes.h"
  55 #include "export.h"
  56 #include "compression.h"
  57 #include "rcu-string.h"
  58 #include "dev-replace.h"
  59 #include "free-space-cache.h"
  60 #include "backref.h"
  61 #include "tests/btrfs-tests.h"
  62
  63 #define CREATE_TRACE_POINTS
  64 #include <trace/events/btrfs.h>
  65
  66 static const struct super_operations btrfs_super_ops;
  67 static struct file_system_type btrfs_fs_type;
  68
  69 static const char *btrfs_decode_error(int errno)
  70 {
  71         char *errstr = "unknown";
  72
  73         switch (errno) {
  74         case -EIO:
  75                 errstr = "IO failure";
  76                 break;
  77         case -ENOMEM:
  78                 errstr = "Out of memory";
  79                 break;
  80         case -EROFS:
  81                 errstr = "Readonly filesystem";
  82                 break;
  83         case -EEXIST:
  84                 errstr = "Object already exists";
  85                 break;
  86         case -ENOSPC:
  87                 errstr = "No space left";
  88                 break;
  89         case -ENOENT:
  90                 errstr = "No such entry";
  91                 break;
  92         }
  93
  94         return errstr;
  95 }
  96
  97 static void save_error_info(struct btrfs_fs_info *fs_info)
  98 {
  99         /*
 100          * today we only save the error info into ram.  Long term we'll
 101          * also send it down to the disk
 102          */
 103         set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
 104 }
 105
 106 /* btrfs handle error by forcing the filesystem readonly */
 107 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
 108 {
 109         struct super_block *sb = fs_info->sb;
 110
 111         if (sb->s_flags & MS_RDONLY)
 112                 return;
 113
 114         if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
 115                 sb->s_flags |= MS_RDONLY;
 116                 btrfs_info(fs_info, "forced readonly");
 117                 /*
 118                  * Note that a running device replace operation is not
 119                  * canceled here although there is no way to update
 120                  * the progress. It would add the risk of a deadlock,
 121                  * therefore the canceling is ommited. The only penalty
 122                  * is that some I/O remains active until the procedure
 123                  * completes. The next time when the filesystem is
 124                  * mounted writeable again, the device replace
 125                  * operation continues.
 126                  */
 127         }
 128 }
 129
 130 #ifdef CONFIG_PRINTK
 131 /*
 132  * __btrfs_std_error decodes expected errors from the caller and
 133  * invokes the approciate error response.
 134  */
 135 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 136                        unsigned int line, int errno, const char *fmt, ...)
 137 {
 138         struct super_block *sb = fs_info->sb;
 139         const char *errstr;
 140
 141         /*
 142          * Special case: if the error is EROFS, and we're already
 143          * under MS_RDONLY, then it is safe here.
 144          */
 145         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 146                 return;
 147
 148         errstr = btrfs_decode_error(errno);
 149         if (fmt) {
 150                 struct va_format vaf;
 151                 va_list args;
 152
 153                 va_start(args, fmt);
 154                 vaf.fmt = fmt;
 155                 vaf.va = &args;
 156
 157                 printk(KERN_CRIT
 158                         "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
 159                         sb->s_id, function, line, errno, errstr, &vaf);
 160                 va_end(args);
 161         } else {
 162                 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
 163                         sb->s_id, function, line, errno, errstr);
 164         }
 165
 166         /* Don't go through full error handling during mount */
 167         save_error_info(fs_info);
 168         if (sb->s_flags & MS_BORN)
 169                 btrfs_handle_error(fs_info);
 170 }
 171
 172 static const char * const logtypes[] = {
 173         "emergency",
 174         "alert",
 175         "critical",
 176         "error",
 177         "warning",
 178         "notice",
 179         "info",
 180         "debug",
 181 };
 182
 183 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
 184 {
 185         struct super_block *sb = fs_info->sb;
 186         char lvl[4];
 187         struct va_format vaf;
 188         va_list args;
 189         const char *type = logtypes[4];
 190         int kern_level;
 191
 192         va_start(args, fmt);
 193
 194         kern_level = printk_get_level(fmt);
 195         if (kern_level) {
 196                 size_t size = printk_skip_level(fmt) - fmt;
 197                 memcpy(lvl, fmt,  size);
 198                 lvl[size] = '\0';
 199                 fmt += size;
 200                 type = logtypes[kern_level - '0'];
 201         } else
 202                 *lvl = '\0';
 203
 204         vaf.fmt = fmt;
 205         vaf.va = &args;
 206
 207         printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
 208
 209         va_end(args);
 210 }
 211
 212 #else
 213
 214 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 215                        unsigned int line, int errno, const char *fmt, ...)
 216 {
 217         struct super_block *sb = fs_info->sb;
 218
 219         /*
 220          * Special case: if the error is EROFS, and we're already
 221          * under MS_RDONLY, then it is safe here.
 222          */
 223         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 224                 return;
 225
 226         /* Don't go through full error handling during mount */
 227         if (sb->s_flags & MS_BORN) {
 228                 save_error_info(fs_info);
 229                 btrfs_handle_error(fs_info);
 230         }
 231 }
 232 #endif
 233
 234 /*
 235  * We only mark the transaction aborted and then set the file system read-only.
 236  * This will prevent new transactions from starting or trying to join this
 237  * one.
 238  *
 239  * This means that error recovery at the call site is limited to freeing
 240  * any local memory allocations and passing the error code up without
 241  * further cleanup. The transaction should complete as it normally would
 242  * in the call path but will return -EIO.
 243  *
 244  * We'll complete the cleanup in btrfs_end_transaction and
 245  * btrfs_commit_transaction.
 246  */
 247 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
 248                                struct btrfs_root *root, const char *function,
 249                                unsigned int line, int errno)
 250 {
 251         /*
 252          * Report first abort since mount
 253          */
 254         if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,
 255                                 &root->fs_info->fs_state)) {
 256                 WARN(1, KERN_DEBUG "BTRFS: Transaction aborted (error %d)\n",
 257                                 errno);
 258         }
 259         trans->aborted = errno;
 260         /* Nothing used. The other threads that have joined this
 261          * transaction may be able to continue. */
 262         if (!trans->blocks_used) {
 263                 const char *errstr;
 264
 265                 errstr = btrfs_decode_error(errno);
 266                 btrfs_warn(root->fs_info,
 267                            "%s:%d: Aborting unused transaction(%s).",
 268                            function, line, errstr);
 269                 return;
 270         }
 271         ACCESS_ONCE(trans->transaction->aborted) = errno;
 272         /* Wake up anybody who may be waiting on this transaction */
 273         wake_up(&root->fs_info->transaction_wait);
 274         wake_up(&root->fs_info->transaction_blocked_wait);
 275         __btrfs_std_error(root->fs_info, function, line, errno, NULL);
 276 }
 277 /*
 278  * __btrfs_panic decodes unexpected, fatal errors from the caller,
 279  * issues an alert, and either panics or BUGs, depending on mount options.
 280  */
 281 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
 282                    unsigned int line, int errno, const char *fmt, ...)
 283 {
 284         char *s_id = "<unknown>";
 285         const char *errstr;
 286         struct va_format vaf = { .fmt = fmt };
 287         va_list args;
 288
 289         if (fs_info)
 290                 s_id = fs_info->sb->s_id;
 291
 292         va_start(args, fmt);
 293         vaf.va = &args;
 294
 295         errstr = btrfs_decode_error(errno);
 296         if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
 297                 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
 298                         s_id, function, line, &vaf, errno, errstr);
 299
 300         btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
 301                    function, line, &vaf, errno, errstr);
 302         va_end(args);
 303         /* Caller calls BUG() */
 304 }
 305
 306 static void btrfs_put_super(struct super_block *sb)
 307 {
 308         (void)close_ctree(btrfs_sb(sb)->tree_root);
 309         /* FIXME: need to fix VFS to return error? */
 310         /* AV: return it _where_?  ->put_super() can be triggered by any number
 311          * of async events, up to and including delivery of SIGKILL to the
 312          * last process that kept it busy.  Or segfault in the aforementioned
 313          * process...  Whom would you report that to?
 314          */
 315 }
 316
 317 enum {
 318         Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
 319         Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
 320         Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
 321         Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
 322         Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
 323         Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
 324         Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
 325         Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
 326         Opt_check_integrity, Opt_check_integrity_including_extent_data,
 327         Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
 328         Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
 329         Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
 330         Opt_datasum, Opt_treelog,
 331         Opt_err,
 332 };
 333
 334 static match_table_t tokens = {
 335         {Opt_degraded, "degraded"},
 336         {Opt_subvol, "subvol=%s"},
 337         {Opt_subvolid, "subvolid=%s"},
 338         {Opt_device, "device=%s"},
 339         {Opt_nodatasum, "nodatasum"},
 340         {Opt_datasum, "datasum"},
 341         {Opt_nodatacow, "nodatacow"},
 342         {Opt_datacow, "datacow"},
 343         {Opt_nobarrier, "nobarrier"},
 344         {Opt_barrier, "barrier"},
 345         {Opt_max_inline, "max_inline=%s"},
 346         {Opt_alloc_start, "alloc_start=%s"},
 347         {Opt_thread_pool, "thread_pool=%d"},
 348         {Opt_compress, "compress"},
 349         {Opt_compress_type, "compress=%s"},
 350         {Opt_compress_force, "compress-force"},
 351         {Opt_compress_force_type, "compress-force=%s"},
 352         {Opt_ssd, "ssd"},
 353         {Opt_ssd_spread, "ssd_spread"},
 354         {Opt_nossd, "nossd"},
 355         {Opt_acl, "acl"},
 356         {Opt_noacl, "noacl"},
 357         {Opt_notreelog, "notreelog"},
 358         {Opt_treelog, "treelog"},
 359         {Opt_flushoncommit, "flushoncommit"},
 360         {Opt_noflushoncommit, "noflushoncommit"},
 361         {Opt_ratio, "metadata_ratio=%d"},
 362         {Opt_discard, "discard"},
 363         {Opt_nodiscard, "nodiscard"},
 364         {Opt_space_cache, "space_cache"},
 365         {Opt_clear_cache, "clear_cache"},
 366         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 367         {Opt_enospc_debug, "enospc_debug"},
 368         {Opt_noenospc_debug, "noenospc_debug"},
 369         {Opt_subvolrootid, "subvolrootid=%d"},
 370         {Opt_defrag, "autodefrag"},
 371         {Opt_nodefrag, "noautodefrag"},
 372         {Opt_inode_cache, "inode_cache"},
 373         {Opt_no_space_cache, "nospace_cache"},
 374         {Opt_recovery, "recovery"},
 375         {Opt_skip_balance, "skip_balance"},
 376         {Opt_check_integrity, "check_int"},
 377         {Opt_check_integrity_including_extent_data, "check_int_data"},
 378         {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
 379         {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 380         {Opt_fatal_errors, "fatal_errors=%s"},
 381         {Opt_commit_interval, "commit=%d"},
 382         {Opt_err, NULL},
 383 };
 384
 385 /*
 386  * Regular mount options parser.  Everything that is needed only when
 387  * reading in a new superblock is parsed here.
 388  * XXX JDM: This needs to be cleaned up for remount.
 389  */
 390 int btrfs_parse_options(struct btrfs_root *root, char *options)
 391 {
 392         struct btrfs_fs_info *info = root->fs_info;
 393         substring_t args[MAX_OPT_ARGS];
 394         char *p, *num, *orig = NULL;
 395         u64 cache_gen;
 396         int intarg;
 397         int ret = 0;
 398         char *compress_type;
 399         bool compress_force = false;
 400
 401         cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
 402         if (cache_gen)
 403                 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 404
 405         if (!options)
 406                 goto out;
 407
 408         /*
 409          * strsep changes the string, duplicate it because parse_options
 410          * gets called twice
 411          */
 412         options = kstrdup(options, GFP_NOFS);
 413         if (!options)
 414                 return -ENOMEM;
 415
 416         orig = options;
 417
 418         while ((p = strsep(&options, ",")) != NULL) {
 419                 int token;
 420                 if (!*p)
 421                         continue;
 422
 423                 token = match_token(p, tokens, args);
 424                 switch (token) {
 425                 case Opt_degraded:
 426                         btrfs_info(root->fs_info, "allowing degraded mounts");
 427                         btrfs_set_opt(info->mount_opt, DEGRADED);
 428                         break;
 429                 case Opt_subvol:
 430                 case Opt_subvolid:
 431                 case Opt_subvolrootid:
 432                 case Opt_device:
 433                         /*
 434                          * These are parsed by btrfs_parse_early_options
 435                          * and can be happily ignored here.
 436                          */
 437                         break;
 438                 case Opt_nodatasum:
 439                         btrfs_info(root->fs_info, "setting nodatasum");
 440                         btrfs_set_opt(info->mount_opt, NODATASUM);
 441                         break;
 442                 case Opt_datasum:
 443                         if (btrfs_test_opt(root, NODATACOW))
 444                                 btrfs_info(root->fs_info, "setting datasum, datacow enabled");
 445                         else
 446                                 btrfs_info(root->fs_info, "setting datasum");
 447                         btrfs_clear_opt(info->mount_opt, NODATACOW);
 448                         btrfs_clear_opt(info->mount_opt, NODATASUM);
 449                         break;
 450                 case Opt_nodatacow:
 451                         if (!btrfs_test_opt(root, COMPRESS) ||
 452                                 !btrfs_test_opt(root, FORCE_COMPRESS)) {
 453                                         btrfs_info(root->fs_info,
 454                                                 "setting nodatacow, compression disabled");
 455                         } else {
 456                                 btrfs_info(root->fs_info, "setting nodatacow");
 457                         }
 458                         btrfs_clear_opt(info->mount_opt, COMPRESS);
 459                         btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 460                         btrfs_set_opt(info->mount_opt, NODATACOW);
 461                         btrfs_set_opt(info->mount_opt, NODATASUM);
 462                         break;
 463                 case Opt_datacow:
 464                         if (btrfs_test_opt(root, NODATACOW))
 465                                 btrfs_info(root->fs_info, "setting datacow");
 466                         btrfs_clear_opt(info->mount_opt, NODATACOW);
 467                         break;
 468                 case Opt_compress_force:
 469                 case Opt_compress_force_type:
 470                         compress_force = true;
 471                         /* Fallthrough */
 472                 case Opt_compress:
 473                 case Opt_compress_type:
 474                         if (token == Opt_compress ||
 475                             token == Opt_compress_force ||
 476                             strcmp(args[0].from, "zlib") == 0) {
 477                                 compress_type = "zlib";
 478                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 479                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 480                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 481                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 482                         } else if (strcmp(args[0].from, "lzo") == 0) {
 483                                 compress_type = "lzo";
 484                                 info->compress_type = BTRFS_COMPRESS_LZO;
 485                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 486                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 487                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 488                                 btrfs_set_fs_incompat(info, COMPRESS_LZO);
 489                         } else if (strncmp(args[0].from, "no", 2) == 0) {
 490                                 compress_type = "no";
 491                                 btrfs_clear_opt(info->mount_opt, COMPRESS);
 492                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 493                                 compress_force = false;
 494                         } else {
 495                                 ret = -EINVAL;
 496                                 goto out;
 497                         }
 498
 499                         if (compress_force) {
 500                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 501                                 btrfs_info(root->fs_info, "force %s compression",
 502                                         compress_type);
 503                         } else if (btrfs_test_opt(root, COMPRESS)) {
 504                                 pr_info("btrfs: use %s compression\n",
 505                                         compress_type);
 506                         }
 507                         break;
 508                 case Opt_ssd:
 509                         btrfs_info(root->fs_info, "use ssd allocation scheme");
 510                         btrfs_set_opt(info->mount_opt, SSD);
 511                         break;
 512                 case Opt_ssd_spread:
 513                         btrfs_info(root->fs_info, "use spread ssd allocation scheme");
 514                         btrfs_set_opt(info->mount_opt, SSD);
 515                         btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 516                         break;
 517                 case Opt_nossd:
 518                         btrfs_info(root->fs_info, "not using ssd allocation scheme");
 519                         btrfs_set_opt(info->mount_opt, NOSSD);
 520                         btrfs_clear_opt(info->mount_opt, SSD);
 521                         btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 522                         break;
 523                 case Opt_barrier:
 524                         if (btrfs_test_opt(root, NOBARRIER))
 525                                 btrfs_info(root->fs_info, "turning on barriers");
 526                         btrfs_clear_opt(info->mount_opt, NOBARRIER);
 527                         break;
 528                 case Opt_nobarrier:
 529                         btrfs_info(root->fs_info, "turning off barriers");
 530                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 531                         break;
 532                 case Opt_thread_pool:
 533                         ret = match_int(&args[0], &intarg);
 534                         if (ret) {
 535                                 goto out;
 536                         } else if (intarg > 0) {
 537                                 info->thread_pool_size = intarg;
 538                         } else {
 539                                 ret = -EINVAL;
 540                                 goto out;
 541                         }
 542                         break;
 543                 case Opt_max_inline:
 544                         num = match_strdup(&args[0]);
 545                         if (num) {
 546                                 info->max_inline = memparse(num, NULL);
 547                                 kfree(num);
 548
 549                                 if (info->max_inline) {
 550                                         info->max_inline = max_t(u64,
 551                                                 info->max_inline,
 552                                                 root->sectorsize);
 553                                 }
 554                                 btrfs_info(root->fs_info, "max_inline at %llu",
 555                                         info->max_inline);
 556                         } else {
 557                                 ret = -ENOMEM;
 558                                 goto out;
 559                         }
 560                         break;
 561                 case Opt_alloc_start:
 562                         num = match_strdup(&args[0]);
 563                         if (num) {
 564                                 mutex_lock(&info->chunk_mutex);
 565                                 info->alloc_start = memparse(num, NULL);
 566                                 mutex_unlock(&info->chunk_mutex);
 567                                 kfree(num);
 568                                 btrfs_info(root->fs_info, "allocations start at %llu",
 569                                         info->alloc_start);
 570                         } else {
 571                                 ret = -ENOMEM;
 572                                 goto out;
 573                         }
 574                         break;
 575                 case Opt_acl:
 576                         root->fs_info->sb->s_flags |= MS_POSIXACL;
 577                         break;
 578                 case Opt_noacl:
 579                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 580                         break;
 581                 case Opt_notreelog:
 582                         btrfs_info(root->fs_info, "disabling tree log");
 583                         btrfs_set_opt(info->mount_opt, NOTREELOG);
 584                         break;
 585                 case Opt_treelog:
 586                         if (btrfs_test_opt(root, NOTREELOG))
 587                                 btrfs_info(root->fs_info, "enabling tree log");
 588                         btrfs_clear_opt(info->mount_opt, NOTREELOG);
 589                         break;
 590                 case Opt_flushoncommit:
 591                         btrfs_info(root->fs_info, "turning on flush-on-commit");
 592                         btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
 593                         break;
 594                 case Opt_noflushoncommit:
 595                         if (btrfs_test_opt(root, FLUSHONCOMMIT))
 596                                 btrfs_info(root->fs_info, "turning off flush-on-commit");
 597                         btrfs_clear_opt(info->mount_opt, FLUSHONCOMMIT);
 598                         break;
 599                 case Opt_ratio:
 600                         ret = match_int(&args[0], &intarg);
 601                         if (ret) {
 602                                 goto out;
 603                         } else if (intarg >= 0) {
 604                                 info->metadata_ratio = intarg;
 605                                 btrfs_info(root->fs_info, "metadata ratio %d",
 606                                        info->metadata_ratio);
 607                         } else {
 608                                 ret = -EINVAL;
 609                                 goto out;
 610                         }
 611                         break;
 612                 case Opt_discard:
 613                         btrfs_set_opt(info->mount_opt, DISCARD);
 614                         break;
 615                 case Opt_nodiscard:
 616                         btrfs_clear_opt(info->mount_opt, DISCARD);
 617                         break;
 618                 case Opt_space_cache:
 619                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 620                         break;
 621                 case Opt_rescan_uuid_tree:
 622                         btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 623                         break;
 624                 case Opt_no_space_cache:
 625                         btrfs_info(root->fs_info, "disabling disk space caching");
 626                         btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
 627                         break;
 628                 case Opt_inode_cache:
 629                         btrfs_info(root->fs_info, "enabling inode map caching");
 630                         btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
 631                         break;
 632                 case Opt_clear_cache:
 633                         btrfs_info(root->fs_info, "force clearing of disk cache");
 634                         btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
 635                         break;
 636                 case Opt_user_subvol_rm_allowed:
 637                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 638                         break;
 639                 case Opt_enospc_debug:
 640                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 641                         break;
 642                 case Opt_noenospc_debug:
 643                         btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
 644                         break;
 645                 case Opt_defrag:
 646                         btrfs_info(root->fs_info, "enabling auto defrag");
 647                         btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
 648                         break;
 649                 case Opt_nodefrag:
 650                         if (btrfs_test_opt(root, AUTO_DEFRAG))
 651                                 btrfs_info(root->fs_info, "disabling auto defrag");
 652                         btrfs_clear_opt(info->mount_opt, AUTO_DEFRAG);
 653                         break;
 654                 case Opt_recovery:
 655                         btrfs_info(root->fs_info, "enabling auto recovery");
 656                         btrfs_set_opt(info->mount_opt, RECOVERY);
 657                         break;
 658                 case Opt_skip_balance:
 659                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 660                         break;
 661 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 662                 case Opt_check_integrity_including_extent_data:
 663                         btrfs_info(root->fs_info,
 664                                    "enabling check integrity including extent data");
 665                         btrfs_set_opt(info->mount_opt,
 666                                       CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
 667                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 668                         break;
 669                 case Opt_check_integrity:
 670                         btrfs_info(root->fs_info, "enabling check integrity");
 671                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 672                         break;
 673                 case Opt_check_integrity_print_mask:
 674                         ret = match_int(&args[0], &intarg);
 675                         if (ret) {
 676                                 goto out;
 677                         } else if (intarg >= 0) {
 678                                 info->check_integrity_print_mask = intarg;
 679                                 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
 680                                        info->check_integrity_print_mask);
 681                         } else {
 682                                 ret = -EINVAL;
 683                                 goto out;
 684                         }
 685                         break;
 686 #else
 687                 case Opt_check_integrity_including_extent_data:
 688                 case Opt_check_integrity:
 689                 case Opt_check_integrity_print_mask:
 690                         btrfs_err(root->fs_info,
 691                                 "support for check_integrity* not compiled in!");
 692                         ret = -EINVAL;
 693                         goto out;
 694 #endif
 695                 case Opt_fatal_errors:
 696                         if (strcmp(args[0].from, "panic") == 0)
 697                                 btrfs_set_opt(info->mount_opt,
 698                                               PANIC_ON_FATAL_ERROR);
 699                         else if (strcmp(args[0].from, "bug") == 0)
 700                                 btrfs_clear_opt(info->mount_opt,
 701                                               PANIC_ON_FATAL_ERROR);
 702                         else {
 703                                 ret = -EINVAL;
 704                                 goto out;
 705                         }
 706                         break;
 707                 case Opt_commit_interval:
 708                         intarg = 0;
 709                         ret = match_int(&args[0], &intarg);
 710                         if (ret < 0) {
 711                                 btrfs_err(root->fs_info, "invalid commit interval");
 712                                 ret = -EINVAL;
 713                                 goto out;
 714                         }
 715                         if (intarg > 0) {
 716                                 if (intarg > 300) {
 717                                         btrfs_warn(root->fs_info, "excessive commit interval %d",
 718                                                         intarg);
 719                                 }
 720                                 info->commit_interval = intarg;
 721                         } else {
 722                                 btrfs_info(root->fs_info, "using default commit interval %ds",
 723                                     BTRFS_DEFAULT_COMMIT_INTERVAL);
 724                                 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
 725                         }
 726                         break;
 727                 case Opt_err:
 728                         btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
 729                         ret = -EINVAL;
 730                         goto out;
 731                 default:
 732                         break;
 733                 }
 734         }
 735 out:
 736         if (!ret && btrfs_test_opt(root, SPACE_CACHE))
 737                 btrfs_info(root->fs_info, "disk space caching is enabled");
 738         kfree(orig);
 739         return ret;
 740 }
 741
 742 /*
 743  * Parse mount options that are required early in the mount process.
 744  *
 745  * All other options will be parsed on much later in the mount process and
 746  * only when we need to allocate a new super block.
 747  */
 748 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 749                 void *holder, char **subvol_name, u64 *subvol_objectid,
 750                 struct btrfs_fs_devices **fs_devices)
 751 {
 752         substring_t args[MAX_OPT_ARGS];
 753         char *device_name, *opts, *orig, *p;
 754         char *num = NULL;
 755         int error = 0;
 756
 757         if (!options)
 758                 return 0;
 759
 760         /*
 761          * strsep changes the string, duplicate it because parse_options
 762          * gets called twice
 763          */
 764         opts = kstrdup(options, GFP_KERNEL);
 765         if (!opts)
 766                 return -ENOMEM;
 767         orig = opts;
 768
 769         while ((p = strsep(&opts, ",")) != NULL) {
 770                 int token;
 771                 if (!*p)
 772                         continue;
 773
 774                 token = match_token(p, tokens, args);
 775                 switch (token) {
 776                 case Opt_subvol:
 777                         kfree(*subvol_name);
 778                         *subvol_name = match_strdup(&args[0]);
 779                         if (!*subvol_name) {
 780                                 error = -ENOMEM;
 781                                 goto out;
 782                         }
 783                         break;
 784                 case Opt_subvolid:
 785                         num = match_strdup(&args[0]);
 786                         if (num) {
 787                                 *subvol_objectid = memparse(num, NULL);
 788                                 kfree(num);
 789                                 /* we want the original fs_tree */
 790                                 if (!*subvol_objectid)
 791                                         *subvol_objectid =
 792                                                 BTRFS_FS_TREE_OBJECTID;
 793                         } else {
 794                                 error = -EINVAL;
 795                                 goto out;
 796                         }
 797                         break;
 798                 case Opt_subvolrootid:
 799                         printk(KERN_WARNING
 800                                 "BTRFS: 'subvolrootid' mount option is deprecated and has "
 801                                 "no effect\n");
 802                         break;
 803                 case Opt_device:
 804                         device_name = match_strdup(&args[0]);
 805                         if (!device_name) {
 806                                 error = -ENOMEM;
 807                                 goto out;
 808                         }
 809                         error = btrfs_scan_one_device(device_name,
 810                                         flags, holder, fs_devices);
 811                         kfree(device_name);
 812                         if (error)
 813                                 goto out;
 814                         break;
 815                 default:
 816                         break;
 817                 }
 818         }
 819
 820 out:
 821         kfree(orig);
 822         return error;
 823 }
 824
 825 static struct dentry *get_default_root(struct super_block *sb,
 826                                        u64 subvol_objectid)
 827 {
 828         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 829         struct btrfs_root *root = fs_info->tree_root;
 830         struct btrfs_root *new_root;
 831         struct btrfs_dir_item *di;
 832         struct btrfs_path *path;
 833         struct btrfs_key location;
 834         struct inode *inode;
 835         u64 dir_id;
 836         int new = 0;
 837
 838         /*
 839          * We have a specific subvol we want to mount, just setup location and
 840          * go look up the root.
 841          */
 842         if (subvol_objectid) {
 843                 location.objectid = subvol_objectid;
 844                 location.type = BTRFS_ROOT_ITEM_KEY;
 845                 location.offset = (u64)-1;
 846                 goto find_root;
 847         }
 848
 849         path = btrfs_alloc_path();
 850         if (!path)
 851                 return ERR_PTR(-ENOMEM);
 852         path->leave_spinning = 1;
 853
 854         /*
 855          * Find the "default" dir item which points to the root item that we
 856          * will mount by default if we haven't been given a specific subvolume
 857          * to mount.
 858          */
 859         dir_id = btrfs_super_root_dir(fs_info->super_copy);
 860         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 861         if (IS_ERR(di)) {
 862                 btrfs_free_path(path);
 863                 return ERR_CAST(di);
 864         }
 865         if (!di) {
 866                 /*
 867                  * Ok the default dir item isn't there.  This is weird since
 868                  * it's always been there, but don't freak out, just try and
 869                  * mount to root most subvolume.
 870                  */
 871                 btrfs_free_path(path);
 872                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 873                 new_root = fs_info->fs_root;
 874                 goto setup_root;
 875         }
 876
 877         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 878         btrfs_free_path(path);
 879
 880 find_root:
 881         new_root = btrfs_read_fs_root_no_name(fs_info, &location);
 882         if (IS_ERR(new_root))
 883                 return ERR_CAST(new_root);
 884
 885         dir_id = btrfs_root_dirid(&new_root->root_item);
 886 setup_root:
 887         location.objectid = dir_id;
 888         location.type = BTRFS_INODE_ITEM_KEY;
 889         location.offset = 0;
 890
 891         inode = btrfs_iget(sb, &location, new_root, &new);
 892         if (IS_ERR(inode))
 893                 return ERR_CAST(inode);
 894
 895         /*
 896          * If we're just mounting the root most subvol put the inode and return
 897          * a reference to the dentry.  We will have already gotten a reference
 898          * to the inode in btrfs_fill_super so we're good to go.
 899          */
 900         if (!new && sb->s_root->d_inode == inode) {
 901                 iput(inode);
 902                 return dget(sb->s_root);
 903         }
 904
 905         return d_obtain_alias(inode);
 906 }
 907
 908 static int btrfs_fill_super(struct super_block *sb,
 909                             struct btrfs_fs_devices *fs_devices,
 910                             void *data, int silent)
 911 {
 912         struct inode *inode;
 913         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 914         struct btrfs_key key;
 915         int err;
 916
 917         sb->s_maxbytes = MAX_LFS_FILESIZE;
 918         sb->s_magic = BTRFS_SUPER_MAGIC;
 919         sb->s_op = &btrfs_super_ops;
 920         sb->s_d_op = &btrfs_dentry_operations;
 921         sb->s_export_op = &btrfs_export_ops;
 922         sb->s_xattr = btrfs_xattr_handlers;
 923         sb->s_time_gran = 1;
 924 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 925         sb->s_flags |= MS_POSIXACL;
 926 #endif
 927         sb->s_flags |= MS_I_VERSION;
 928         err = open_ctree(sb, fs_devices, (char *)data);
 929         if (err) {
 930                 printk(KERN_ERR "BTRFS: open_ctree failed\n");
 931                 return err;
 932         }
 933
 934         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 935         key.type = BTRFS_INODE_ITEM_KEY;
 936         key.offset = 0;
 937         inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
 938         if (IS_ERR(inode)) {
 939                 err = PTR_ERR(inode);
 940                 goto fail_close;
 941         }
 942
 943         sb->s_root = d_make_root(inode);
 944         if (!sb->s_root) {
 945                 err = -ENOMEM;
 946                 goto fail_close;
 947         }
 948
 949         save_mount_options(sb, data);
 950         cleancache_init_fs(sb);
 951         sb->s_flags |= MS_ACTIVE;
 952         return 0;
 953
 954 fail_close:
 955         close_ctree(fs_info->tree_root);
 956         return err;
 957 }
 958
 959 int btrfs_sync_fs(struct super_block *sb, int wait)
 960 {
 961         struct btrfs_trans_handle *trans;
 962         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 963         struct btrfs_root *root = fs_info->tree_root;
 964
 965         trace_btrfs_sync_fs(wait);
 966
 967         if (!wait) {
 968                 filemap_flush(fs_info->btree_inode->i_mapping);
 969                 return 0;
 970         }
 971
 972         btrfs_wait_ordered_roots(fs_info, -1);
 973
 974         trans = btrfs_attach_transaction_barrier(root);
 975         if (IS_ERR(trans)) {
 976                 /* no transaction, don't bother */
 977                 if (PTR_ERR(trans) == -ENOENT)
 978                         return 0;
 979                 return PTR_ERR(trans);
 980         }
 981         return btrfs_commit_transaction(trans, root);
 982 }
 983
 984 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
 985 {
 986         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
 987         struct btrfs_root *root = info->tree_root;
 988         char *compress_type;
 989
 990         if (btrfs_test_opt(root, DEGRADED))
 991                 seq_puts(seq, ",degraded");
 992         if (btrfs_test_opt(root, NODATASUM))
 993                 seq_puts(seq, ",nodatasum");
 994         if (btrfs_test_opt(root, NODATACOW))
 995                 seq_puts(seq, ",nodatacow");
 996         if (btrfs_test_opt(root, NOBARRIER))
 997                 seq_puts(seq, ",nobarrier");
 998         if (info->max_inline != 8192 * 1024)
 999                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1000         if (info->alloc_start != 0)
1001                 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1002         if (info->thread_pool_size !=  min_t(unsigned long,
1003                                              num_online_cpus() + 2, 8))
1004                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1005         if (btrfs_test_opt(root, COMPRESS)) {
1006                 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1007                         compress_type = "zlib";
1008                 else
1009                         compress_type = "lzo";
1010                 if (btrfs_test_opt(root, FORCE_COMPRESS))
1011                         seq_printf(seq, ",compress-force=%s", compress_type);
1012                 else
1013                         seq_printf(seq, ",compress=%s", compress_type);
1014         }
1015         if (btrfs_test_opt(root, NOSSD))
1016                 seq_puts(seq, ",nossd");
1017         if (btrfs_test_opt(root, SSD_SPREAD))
1018                 seq_puts(seq, ",ssd_spread");
1019         else if (btrfs_test_opt(root, SSD))
1020                 seq_puts(seq, ",ssd");
1021         if (btrfs_test_opt(root, NOTREELOG))
1022                 seq_puts(seq, ",notreelog");
1023         if (btrfs_test_opt(root, FLUSHONCOMMIT))
1024                 seq_puts(seq, ",flushoncommit");
1025         if (btrfs_test_opt(root, DISCARD))
1026                 seq_puts(seq, ",discard");
1027         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1028                 seq_puts(seq, ",noacl");
1029         if (btrfs_test_opt(root, SPACE_CACHE))
1030                 seq_puts(seq, ",space_cache");
1031         else
1032                 seq_puts(seq, ",nospace_cache");
1033         if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1034                 seq_puts(seq, ",rescan_uuid_tree");
1035         if (btrfs_test_opt(root, CLEAR_CACHE))
1036                 seq_puts(seq, ",clear_cache");
1037         if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1038                 seq_puts(seq, ",user_subvol_rm_allowed");
1039         if (btrfs_test_opt(root, ENOSPC_DEBUG))
1040                 seq_puts(seq, ",enospc_debug");
1041         if (btrfs_test_opt(root, AUTO_DEFRAG))
1042                 seq_puts(seq, ",autodefrag");
1043         if (btrfs_test_opt(root, INODE_MAP_CACHE))
1044                 seq_puts(seq, ",inode_cache");
1045         if (btrfs_test_opt(root, SKIP_BALANCE))
1046                 seq_puts(seq, ",skip_balance");
1047         if (btrfs_test_opt(root, RECOVERY))
1048                 seq_puts(seq, ",recovery");
1049 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1050         if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1051                 seq_puts(seq, ",check_int_data");
1052         else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1053                 seq_puts(seq, ",check_int");
1054         if (info->check_integrity_print_mask)
1055                 seq_printf(seq, ",check_int_print_mask=%d",
1056                                 info->check_integrity_print_mask);
1057 #endif
1058         if (info->metadata_ratio)
1059                 seq_printf(seq, ",metadata_ratio=%d",
1060                                 info->metadata_ratio);
1061         if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1062                 seq_puts(seq, ",fatal_errors=panic");
1063         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1064                 seq_printf(seq, ",commit=%d", info->commit_interval);
1065         return 0;
1066 }
1067
1068 static int btrfs_test_super(struct super_block *s, void *data)
1069 {
1070         struct btrfs_fs_info *p = data;
1071         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1072
1073         return fs_info->fs_devices == p->fs_devices;
1074 }
1075
1076 static int btrfs_set_super(struct super_block *s, void *data)
1077 {
1078         int err = set_anon_super(s, data);
1079         if (!err)
1080                 s->s_fs_info = data;
1081         return err;
1082 }
1083
1084 /*
1085  * subvolumes are identified by ino 256
1086  */
1087 static inline int is_subvolume_inode(struct inode *inode)
1088 {
1089         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1090                 return 1;
1091         return 0;
1092 }
1093
1094 /*
1095  * This will strip out the subvol=%s argument for an argument string and add
1096  * subvolid=0 to make sure we get the actual tree root for path walking to the
1097  * subvol we want.
1098  */
1099 static char *setup_root_args(char *args)
1100 {
1101         unsigned len = strlen(args) + 2 + 1;
1102         char *src, *dst, *buf;
1103
1104         /*
1105          * We need the same args as before, but with this substitution:
1106          * s!subvol=[^,]+!subvolid=0!
1107          *
1108          * Since the replacement string is up to 2 bytes longer than the
1109          * original, allocate strlen(args) + 2 + 1 bytes.
1110          */
1111
1112         src = strstr(args, "subvol=");
1113         /* This shouldn't happen, but just in case.. */
1114         if (!src)
1115                 return NULL;
1116
1117         buf = dst = kmalloc(len, GFP_NOFS);
1118         if (!buf)
1119                 return NULL;
1120
1121         /*
1122          * If the subvol= arg is not at the start of the string,
1123          * copy whatever precedes it into buf.
1124          */
1125         if (src != args) {
1126                 *src++ = '\0';
1127                 strcpy(buf, args);
1128                 dst += strlen(args);
1129         }
1130
1131         strcpy(dst, "subvolid=0");
1132         dst += strlen("subvolid=0");
1133
1134         /*
1135          * If there is a "," after the original subvol=... string,
1136          * copy that suffix into our buffer.  Otherwise, we're done.
1137          */
1138         src = strchr(src, ',');
1139         if (src)
1140                 strcpy(dst, src);
1141
1142         return buf;
1143 }
1144
1145 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1146                                    const char *device_name, char *data)
1147 {
1148         struct dentry *root;
1149         struct vfsmount *mnt;
1150         char *newargs;
1151
1152         newargs = setup_root_args(data);
1153         if (!newargs)
1154                 return ERR_PTR(-ENOMEM);
1155         mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1156                              newargs);
1157         kfree(newargs);
1158         if (IS_ERR(mnt))
1159                 return ERR_CAST(mnt);
1160
1161         root = mount_subtree(mnt, subvol_name);
1162
1163         if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1164                 struct super_block *s = root->d_sb;
1165                 dput(root);
1166                 root = ERR_PTR(-EINVAL);
1167                 deactivate_locked_super(s);
1168                 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1169                                 subvol_name);
1170         }
1171
1172         return root;
1173 }
1174
1175 /*
1176  * Find a superblock for the given device / mount point.
1177  *
1178  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
1179  *        for multiple device setup.  Make sure to keep it in sync.
1180  */
1181 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1182                 const char *device_name, void *data)
1183 {
1184         struct block_device *bdev = NULL;
1185         struct super_block *s;
1186         struct dentry *root;
1187         struct btrfs_fs_devices *fs_devices = NULL;
1188         struct btrfs_fs_info *fs_info = NULL;
1189         fmode_t mode = FMODE_READ;
1190         char *subvol_name = NULL;
1191         u64 subvol_objectid = 0;
1192         int error = 0;
1193
1194         if (!(flags & MS_RDONLY))
1195                 mode |= FMODE_WRITE;
1196
1197         error = btrfs_parse_early_options(data, mode, fs_type,
1198                                           &subvol_name, &subvol_objectid,
1199                                           &fs_devices);
1200         if (error) {
1201                 kfree(subvol_name);
1202                 return ERR_PTR(error);
1203         }
1204
1205         if (subvol_name) {
1206                 root = mount_subvol(subvol_name, flags, device_name, data);
1207                 kfree(subvol_name);
1208                 return root;
1209         }
1210
1211         error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1212         if (error)
1213                 return ERR_PTR(error);
1214
1215         /*
1216          * Setup a dummy root and fs_info for test/set super.  This is because
1217          * we don't actually fill this stuff out until open_ctree, but we need
1218          * it for searching for existing supers, so this lets us do that and
1219          * then open_ctree will properly initialize everything later.
1220          */
1221         fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1222         if (!fs_info)
1223                 return ERR_PTR(-ENOMEM);
1224
1225         fs_info->fs_devices = fs_devices;
1226
1227         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1228         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1229         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1230                 error = -ENOMEM;
1231                 goto error_fs_info;
1232         }
1233
1234         error = btrfs_open_devices(fs_devices, mode, fs_type);
1235         if (error)
1236                 goto error_fs_info;
1237
1238         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1239                 error = -EACCES;
1240                 goto error_close_devices;
1241         }
1242
1243         bdev = fs_devices->latest_bdev;
1244         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1245                  fs_info);
1246         if (IS_ERR(s)) {
1247                 error = PTR_ERR(s);
1248                 goto error_close_devices;
1249         }
1250
1251         if (s->s_root) {
1252                 btrfs_close_devices(fs_devices);
1253                 free_fs_info(fs_info);
1254                 if ((flags ^ s->s_flags) & MS_RDONLY)
1255                         error = -EBUSY;
1256         } else {
1257                 char b[BDEVNAME_SIZE];
1258
1259                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1260                 btrfs_sb(s)->bdev_holder = fs_type;
1261                 error = btrfs_fill_super(s, fs_devices, data,
1262                                          flags & MS_SILENT ? 1 : 0);
1263         }
1264
1265         root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1266         if (IS_ERR(root))
1267                 deactivate_locked_super(s);
1268
1269         return root;
1270
1271 error_close_devices:
1272         btrfs_close_devices(fs_devices);
1273 error_fs_info:
1274         free_fs_info(fs_info);
1275         return ERR_PTR(error);
1276 }
1277
1278 static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1279 {
1280         spin_lock_irq(&workers->lock);
1281         workers->max_workers = new_limit;
1282         spin_unlock_irq(&workers->lock);
1283 }
1284
1285 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1286                                      int new_pool_size, int old_pool_size)
1287 {
1288         if (new_pool_size == old_pool_size)
1289                 return;
1290
1291         fs_info->thread_pool_size = new_pool_size;
1292
1293         btrfs_info(fs_info, "resize thread pool %d -> %d",
1294                old_pool_size, new_pool_size);
1295
1296         btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1297         btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1298         btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1299         btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1300         btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1301         btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1302         btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1303         btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1304         btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1305         btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1306         btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1307         btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1308         btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
1309         btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers,
1310                               new_pool_size);
1311 }
1312
1313 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1314 {
1315         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1316 }
1317
1318 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1319                                        unsigned long old_opts, int flags)
1320 {
1321         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1322             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1323              (flags & MS_RDONLY))) {
1324                 /* wait for any defraggers to finish */
1325                 wait_event(fs_info->transaction_wait,
1326                            (atomic_read(&fs_info->defrag_running) == 0));
1327                 if (flags & MS_RDONLY)
1328                         sync_filesystem(fs_info->sb);
1329         }
1330 }
1331
1332 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1333                                          unsigned long old_opts)
1334 {
1335         /*
1336          * We need cleanup all defragable inodes if the autodefragment is
1337          * close or the fs is R/O.
1338          */
1339         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1340             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1341              (fs_info->sb->s_flags & MS_RDONLY))) {
1342                 btrfs_cleanup_defrag_inodes(fs_info);
1343         }
1344
1345         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1346 }
1347
1348 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1349 {
1350         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1351         struct btrfs_root *root = fs_info->tree_root;
1352         unsigned old_flags = sb->s_flags;
1353         unsigned long old_opts = fs_info->mount_opt;
1354         unsigned long old_compress_type = fs_info->compress_type;
1355         u64 old_max_inline = fs_info->max_inline;
1356         u64 old_alloc_start = fs_info->alloc_start;
1357         int old_thread_pool_size = fs_info->thread_pool_size;
1358         unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1359         int ret;
1360
1361         btrfs_remount_prepare(fs_info);
1362
1363         ret = btrfs_parse_options(root, data);
1364         if (ret) {
1365                 ret = -EINVAL;
1366                 goto restore;
1367         }
1368
1369         btrfs_remount_begin(fs_info, old_opts, *flags);
1370         btrfs_resize_thread_pool(fs_info,
1371                 fs_info->thread_pool_size, old_thread_pool_size);
1372
1373         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1374                 goto out;
1375
1376         if (*flags & MS_RDONLY) {
1377                 /*
1378                  * this also happens on 'umount -rf' or on shutdown, when
1379                  * the filesystem is busy.
1380                  */
1381
1382                 /* wait for the uuid_scan task to finish */
1383                 down(&fs_info->uuid_tree_rescan_sem);
1384                 /* avoid complains from lockdep et al. */
1385                 up(&fs_info->uuid_tree_rescan_sem);
1386
1387                 sb->s_flags |= MS_RDONLY;
1388
1389                 btrfs_dev_replace_suspend_for_unmount(fs_info);
1390                 btrfs_scrub_cancel(fs_info);
1391                 btrfs_pause_balance(fs_info);
1392
1393                 ret = btrfs_commit_super(root);
1394                 if (ret)
1395                         goto restore;
1396         } else {
1397                 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1398                         btrfs_err(fs_info,
1399                                 "Remounting read-write after error is not allowed");
1400                         ret = -EINVAL;
1401                         goto restore;
1402                 }
1403                 if (fs_info->fs_devices->rw_devices == 0) {
1404                         ret = -EACCES;
1405                         goto restore;
1406                 }
1407
1408                 if (fs_info->fs_devices->missing_devices >
1409                      fs_info->num_tolerated_disk_barrier_failures &&
1410                     !(*flags & MS_RDONLY)) {
1411                         btrfs_warn(fs_info,
1412                                 "too many missing devices, writeable remount is not allowed");
1413                         ret = -EACCES;
1414                         goto restore;
1415                 }
1416
1417                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1418                         ret = -EINVAL;
1419                         goto restore;
1420                 }
1421
1422                 ret = btrfs_cleanup_fs_roots(fs_info);
1423                 if (ret)
1424                         goto restore;
1425
1426                 /* recover relocation */
1427                 ret = btrfs_recover_relocation(root);
1428                 if (ret)
1429                         goto restore;
1430
1431                 ret = btrfs_resume_balance_async(fs_info);
1432                 if (ret)
1433                         goto restore;
1434
1435                 ret = btrfs_resume_dev_replace_async(fs_info);
1436                 if (ret) {
1437                         btrfs_warn(fs_info, "failed to resume dev_replace");
1438                         goto restore;
1439                 }
1440
1441                 if (!fs_info->uuid_root) {
1442                         btrfs_info(fs_info, "creating UUID tree");
1443                         ret = btrfs_create_uuid_tree(fs_info);
1444                         if (ret) {
1445                                 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1446                                 goto restore;
1447                         }
1448                 }
1449                 sb->s_flags &= ~MS_RDONLY;
1450         }
1451 out:
1452         btrfs_remount_cleanup(fs_info, old_opts);
1453         return 0;
1454
1455 restore:
1456         /* We've hit an error - don't reset MS_RDONLY */
1457         if (sb->s_flags & MS_RDONLY)
1458                 old_flags |= MS_RDONLY;
1459         sb->s_flags = old_flags;
1460         fs_info->mount_opt = old_opts;
1461         fs_info->compress_type = old_compress_type;
1462         fs_info->max_inline = old_max_inline;
1463         mutex_lock(&fs_info->chunk_mutex);
1464         fs_info->alloc_start = old_alloc_start;
1465         mutex_unlock(&fs_info->chunk_mutex);
1466         btrfs_resize_thread_pool(fs_info,
1467                 old_thread_pool_size, fs_info->thread_pool_size);
1468         fs_info->metadata_ratio = old_metadata_ratio;
1469         btrfs_remount_cleanup(fs_info, old_opts);
1470         return ret;
1471 }
1472
1473 /* Used to sort the devices by max_avail(descending sort) */
1474 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1475                                        const void *dev_info2)
1476 {
1477         if (((struct btrfs_device_info *)dev_info1)->max_avail >
1478             ((struct btrfs_device_info *)dev_info2)->max_avail)
1479                 return -1;
1480         else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1481                  ((struct btrfs_device_info *)dev_info2)->max_avail)
1482                 return 1;
1483         else
1484         return 0;
1485 }
1486
1487 /*
1488  * sort the devices by max_avail, in which max free extent size of each device
1489  * is stored.(Descending Sort)
1490  */
1491 static inline void btrfs_descending_sort_devices(
1492                                         struct btrfs_device_info *devices,
1493                                         size_t nr_devices)
1494 {
1495         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1496              btrfs_cmp_device_free_bytes, NULL);
1497 }
1498
1499 /*
1500  * The helper to calc the free space on the devices that can be used to store
1501  * file data.
1502  */
1503 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1504 {
1505         struct btrfs_fs_info *fs_info = root->fs_info;
1506         struct btrfs_device_info *devices_info;
1507         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1508         struct btrfs_device *device;
1509         u64 skip_space;
1510         u64 type;
1511         u64 avail_space;
1512         u64 used_space;
1513         u64 min_stripe_size;
1514         int min_stripes = 1, num_stripes = 1;
1515         int i = 0, nr_devices;
1516         int ret;
1517
1518         nr_devices = fs_info->fs_devices->open_devices;
1519         BUG_ON(!nr_devices);
1520
1521         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1522                                GFP_NOFS);
1523         if (!devices_info)
1524                 return -ENOMEM;
1525
1526         /* calc min stripe number for data space alloction */
1527         type = btrfs_get_alloc_profile(root, 1);
1528         if (type & BTRFS_BLOCK_GROUP_RAID0) {
1529                 min_stripes = 2;
1530                 num_stripes = nr_devices;
1531         } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1532                 min_stripes = 2;
1533                 num_stripes = 2;
1534         } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1535                 min_stripes = 4;
1536                 num_stripes = 4;
1537         }
1538
1539         if (type & BTRFS_BLOCK_GROUP_DUP)
1540                 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1541         else
1542                 min_stripe_size = BTRFS_STRIPE_LEN;
1543
1544         list_for_each_entry(device, &fs_devices->devices, dev_list) {
1545                 if (!device->in_fs_metadata || !device->bdev ||
1546                     device->is_tgtdev_for_dev_replace)
1547                         continue;
1548
1549                 avail_space = device->total_bytes - device->bytes_used;
1550
1551                 /* align with stripe_len */
1552                 do_div(avail_space, BTRFS_STRIPE_LEN);
1553                 avail_space *= BTRFS_STRIPE_LEN;
1554
1555                 /*
1556                  * In order to avoid overwritting the superblock on the drive,
1557                  * btrfs starts at an offset of at least 1MB when doing chunk
1558                  * allocation.
1559                  */
1560                 skip_space = 1024 * 1024;
1561
1562                 /* user can set the offset in fs_info->alloc_start. */
1563                 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1564                     device->total_bytes)
1565                         skip_space = max(fs_info->alloc_start, skip_space);
1566
1567                 /*
1568                  * btrfs can not use the free space in [0, skip_space - 1],
1569                  * we must subtract it from the total. In order to implement
1570                  * it, we account the used space in this range first.
1571                  */
1572                 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1573                                                      &used_space);
1574                 if (ret) {
1575                         kfree(devices_info);
1576                         return ret;
1577                 }
1578
1579                 /* calc the free space in [0, skip_space - 1] */
1580                 skip_space -= used_space;
1581
1582                 /*
1583                  * we can use the free space in [0, skip_space - 1], subtract
1584                  * it from the total.
1585                  */
1586                 if (avail_space && avail_space >= skip_space)
1587                         avail_space -= skip_space;
1588                 else
1589                         avail_space = 0;
1590
1591                 if (avail_space < min_stripe_size)
1592                         continue;
1593
1594                 devices_info[i].dev = device;
1595                 devices_info[i].max_avail = avail_space;
1596
1597                 i++;
1598         }
1599
1600         nr_devices = i;
1601
1602         btrfs_descending_sort_devices(devices_info, nr_devices);
1603
1604         i = nr_devices - 1;
1605         avail_space = 0;
1606         while (nr_devices >= min_stripes) {
1607                 if (num_stripes > nr_devices)
1608                         num_stripes = nr_devices;
1609
1610                 if (devices_info[i].max_avail >= min_stripe_size) {
1611                         int j;
1612                         u64 alloc_size;
1613
1614                         avail_space += devices_info[i].max_avail * num_stripes;
1615                         alloc_size = devices_info[i].max_avail;
1616                         for (j = i + 1 - num_stripes; j <= i; j++)
1617                                 devices_info[j].max_avail -= alloc_size;
1618                 }
1619                 i--;
1620                 nr_devices--;
1621         }
1622
1623         kfree(devices_info);
1624         *free_bytes = avail_space;
1625         return 0;
1626 }
1627
1628 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1629 {
1630         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1631         struct btrfs_super_block *disk_super = fs_info->super_copy;
1632         struct list_head *head = &fs_info->space_info;
1633         struct btrfs_space_info *found;
1634         u64 total_used = 0;
1635         u64 total_free_data = 0;
1636         int bits = dentry->d_sb->s_blocksize_bits;
1637         __be32 *fsid = (__be32 *)fs_info->fsid;
1638         int ret;
1639
1640         /* holding chunk_muext to avoid allocating new chunks */
1641         mutex_lock(&fs_info->chunk_mutex);
1642         rcu_read_lock();
1643         list_for_each_entry_rcu(found, head, list) {
1644                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1645                         total_free_data += found->disk_total - found->disk_used;
1646                         total_free_data -=
1647                                 btrfs_account_ro_block_groups_free_space(found);
1648                 }
1649
1650                 total_used += found->disk_used;
1651         }
1652         rcu_read_unlock();
1653
1654         buf->f_namelen = BTRFS_NAME_LEN;
1655         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1656         buf->f_bfree = buf->f_blocks - (total_used >> bits);
1657         buf->f_bsize = dentry->d_sb->s_blocksize;
1658         buf->f_type = BTRFS_SUPER_MAGIC;
1659         buf->f_bavail = total_free_data;
1660         ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1661         if (ret) {
1662                 mutex_unlock(&fs_info->chunk_mutex);
1663                 return ret;
1664         }
1665         buf->f_bavail += total_free_data;
1666         buf->f_bavail = buf->f_bavail >> bits;
1667         mutex_unlock(&fs_info->chunk_mutex);
1668
1669         /* We treat it as constant endianness (it doesn't matter _which_)
1670            because we want the fsid to come out the same whether mounted
1671            on a big-endian or little-endian host */
1672         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1673         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1674         /* Mask in the root object ID too, to disambiguate subvols */
1675         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1676         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1677
1678         return 0;
1679 }
1680
1681 static void btrfs_kill_super(struct super_block *sb)
1682 {
1683         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1684         kill_anon_super(sb);
1685         free_fs_info(fs_info);
1686 }
1687
1688 static struct file_system_type btrfs_fs_type = {
1689         .owner          = THIS_MODULE,
1690         .name           = "btrfs",
1691         .mount          = btrfs_mount,
1692         .kill_sb        = btrfs_kill_super,
1693         .fs_flags       = FS_REQUIRES_DEV,
1694 };
1695 MODULE_ALIAS_FS("btrfs");
1696
1697 /*
1698  * used by btrfsctl to scan devices when no FS is mounted
1699  */
1700 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1701                                 unsigned long arg)
1702 {
1703         struct btrfs_ioctl_vol_args *vol;
1704         struct btrfs_fs_devices *fs_devices;
1705         int ret = -ENOTTY;
1706
1707         if (!capable(CAP_SYS_ADMIN))
1708                 return -EPERM;
1709
1710         vol = memdup_user((void __user *)arg, sizeof(*vol));
1711         if (IS_ERR(vol))
1712                 return PTR_ERR(vol);
1713
1714         switch (cmd) {
1715         case BTRFS_IOC_SCAN_DEV:
1716                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1717                                             &btrfs_fs_type, &fs_devices);
1718                 break;
1719         case BTRFS_IOC_DEVICES_READY:
1720                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1721                                             &btrfs_fs_type, &fs_devices);
1722                 if (ret)
1723                         break;
1724                 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1725                 break;
1726         }
1727
1728         kfree(vol);
1729         return ret;
1730 }
1731
1732 static int btrfs_freeze(struct super_block *sb)
1733 {
1734         struct btrfs_trans_handle *trans;
1735         struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1736
1737         trans = btrfs_attach_transaction_barrier(root);
1738         if (IS_ERR(trans)) {
1739                 /* no transaction, don't bother */
1740                 if (PTR_ERR(trans) == -ENOENT)
1741                         return 0;
1742                 return PTR_ERR(trans);
1743         }
1744         return btrfs_commit_transaction(trans, root);
1745 }
1746
1747 static int btrfs_unfreeze(struct super_block *sb)
1748 {
1749         return 0;
1750 }
1751
1752 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1753 {
1754         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1755         struct btrfs_fs_devices *cur_devices;
1756         struct btrfs_device *dev, *first_dev = NULL;
1757         struct list_head *head;
1758         struct rcu_string *name;
1759
1760         mutex_lock(&fs_info->fs_devices->device_list_mutex);
1761         cur_devices = fs_info->fs_devices;
1762         while (cur_devices) {
1763                 head = &cur_devices->devices;
1764                 list_for_each_entry(dev, head, dev_list) {
1765                         if (dev->missing)
1766                                 continue;
1767                         if (!first_dev || dev->devid < first_dev->devid)
1768                                 first_dev = dev;
1769                 }
1770                 cur_devices = cur_devices->seed;
1771         }
1772
1773         if (first_dev) {
1774                 rcu_read_lock();
1775                 name = rcu_dereference(first_dev->name);
1776                 seq_escape(m, name->str, " \t\n\\");
1777                 rcu_read_unlock();
1778         } else {
1779                 WARN_ON(1);
1780         }
1781         mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1782         return 0;
1783 }
1784
1785 static const struct super_operations btrfs_super_ops = {
1786         .drop_inode     = btrfs_drop_inode,
1787         .evict_inode    = btrfs_evict_inode,
1788         .put_super      = btrfs_put_super,
1789         .sync_fs        = btrfs_sync_fs,
1790         .show_options   = btrfs_show_options,
1791         .show_devname   = btrfs_show_devname,
1792         .write_inode    = btrfs_write_inode,
1793         .alloc_inode    = btrfs_alloc_inode,
1794         .destroy_inode  = btrfs_destroy_inode,
1795         .statfs         = btrfs_statfs,
1796         .remount_fs     = btrfs_remount,
1797         .freeze_fs      = btrfs_freeze,
1798         .unfreeze_fs    = btrfs_unfreeze,
1799 };
1800
1801 static const struct file_operations btrfs_ctl_fops = {
1802         .unlocked_ioctl  = btrfs_control_ioctl,
1803         .compat_ioctl = btrfs_control_ioctl,
1804         .owner   = THIS_MODULE,
1805         .llseek = noop_llseek,
1806 };
1807
1808 static struct miscdevice btrfs_misc = {
1809         .minor          = BTRFS_MINOR,
1810         .name           = "btrfs-control",
1811         .fops           = &btrfs_ctl_fops
1812 };
1813
1814 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1815 MODULE_ALIAS("devname:btrfs-control");
1816
1817 static int btrfs_interface_init(void)
1818 {
1819         return misc_register(&btrfs_misc);
1820 }
1821
1822 static void btrfs_interface_exit(void)
1823 {
1824         if (misc_deregister(&btrfs_misc) < 0)
1825                 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
1826 }
1827
1828 static void btrfs_print_info(void)
1829 {
1830         printk(KERN_INFO "Btrfs loaded"
1831 #ifdef CONFIG_BTRFS_DEBUG
1832                         ", debug=on"
1833 #endif
1834 #ifdef CONFIG_BTRFS_ASSERT
1835                         ", assert=on"
1836 #endif
1837 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1838                         ", integrity-checker=on"
1839 #endif
1840                         "\n");
1841 }
1842
1843 static int btrfs_run_sanity_tests(void)
1844 {
1845         int ret;
1846
1847         ret = btrfs_init_test_fs();
1848         if (ret)
1849                 return ret;
1850
1851         ret = btrfs_test_free_space_cache();
1852         if (ret)
1853                 goto out;
1854         ret = btrfs_test_extent_buffer_operations();
1855         if (ret)
1856                 goto out;
1857         ret = btrfs_test_extent_io();
1858         if (ret)
1859                 goto out;
1860         ret = btrfs_test_inodes();
1861 out:
1862         btrfs_destroy_test_fs();
1863         return ret;
1864 }
1865
1866 static int __init init_btrfs_fs(void)
1867 {
1868         int err;
1869
1870         err = btrfs_hash_init();
1871         if (err)
1872                 return err;
1873
1874         btrfs_props_init();
1875
1876         err = btrfs_init_sysfs();
1877         if (err)
1878                 goto free_hash;
1879
1880         btrfs_init_compress();
1881
1882         err = btrfs_init_cachep();
1883         if (err)
1884                 goto free_compress;
1885
1886         err = extent_io_init();
1887         if (err)
1888                 goto free_cachep;
1889
1890         err = extent_map_init();
1891         if (err)
1892                 goto free_extent_io;
1893
1894         err = ordered_data_init();
1895         if (err)
1896                 goto free_extent_map;
1897
1898         err = btrfs_delayed_inode_init();
1899         if (err)
1900                 goto free_ordered_data;
1901
1902         err = btrfs_auto_defrag_init();
1903         if (err)
1904                 goto free_delayed_inode;
1905
1906         err = btrfs_delayed_ref_init();
1907         if (err)
1908                 goto free_auto_defrag;
1909
1910         err = btrfs_prelim_ref_init();
1911         if (err)
1912                 goto free_prelim_ref;
1913
1914         err = btrfs_interface_init();
1915         if (err)
1916                 goto free_delayed_ref;
1917
1918         btrfs_init_lockdep();
1919
1920         btrfs_print_info();
1921
1922         err = btrfs_run_sanity_tests();
1923         if (err)
1924                 goto unregister_ioctl;
1925
1926         err = register_filesystem(&btrfs_fs_type);
1927         if (err)
1928                 goto unregister_ioctl;
1929
1930         return 0;
1931
1932 unregister_ioctl:
1933         btrfs_interface_exit();
1934 free_prelim_ref:
1935         btrfs_prelim_ref_exit();
1936 free_delayed_ref:
1937         btrfs_delayed_ref_exit();
1938 free_auto_defrag:
1939         btrfs_auto_defrag_exit();
1940 free_delayed_inode:
1941         btrfs_delayed_inode_exit();
1942 free_ordered_data:
1943         ordered_data_exit();
1944 free_extent_map:
1945         extent_map_exit();
1946 free_extent_io:
1947         extent_io_exit();
1948 free_cachep:
1949         btrfs_destroy_cachep();
1950 free_compress:
1951         btrfs_exit_compress();
1952         btrfs_exit_sysfs();
1953 free_hash:
1954         btrfs_hash_exit();
1955         return err;
1956 }
1957
1958 static void __exit exit_btrfs_fs(void)
1959 {
1960         btrfs_destroy_cachep();
1961         btrfs_delayed_ref_exit();
1962         btrfs_auto_defrag_exit();
1963         btrfs_delayed_inode_exit();
1964         btrfs_prelim_ref_exit();
1965         ordered_data_exit();
1966         extent_map_exit();
1967         extent_io_exit();
1968         btrfs_interface_exit();
1969         unregister_filesystem(&btrfs_fs_type);
1970         btrfs_exit_sysfs();
1971         btrfs_cleanup_fs_uuids();
1972         btrfs_exit_compress();
1973         btrfs_hash_exit();
1974 }
1975
1976 module_init(init_btrfs_fs)
1977 module_exit(exit_btrfs_fs)
1978
1979 MODULE_LICENSE("GPL");