Pileus Git - ~andy/linux/blob - 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, Opt_noinode_cache,
 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_noinode_cache, "noinode_cache"},
 374         {Opt_no_space_cache, "nospace_cache"},
 375         {Opt_recovery, "recovery"},
 376         {Opt_skip_balance, "skip_balance"},
 377         {Opt_check_integrity, "check_int"},
 378         {Opt_check_integrity_including_extent_data, "check_int_data"},
 379         {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
 380         {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 381         {Opt_fatal_errors, "fatal_errors=%s"},
 382         {Opt_commit_interval, "commit=%d"},
 383         {Opt_err, NULL},
 384 };
 385
 386 /*
 387  * Regular mount options parser.  Everything that is needed only when
 388  * reading in a new superblock is parsed here.
 389  * XXX JDM: This needs to be cleaned up for remount.
 390  */
 391 int btrfs_parse_options(struct btrfs_root *root, char *options)
 392 {
 393         struct btrfs_fs_info *info = root->fs_info;
 394         substring_t args[MAX_OPT_ARGS];
 395         char *p, *num, *orig = NULL;
 396         u64 cache_gen;
 397         int intarg;
 398         int ret = 0;
 399         char *compress_type;
 400         bool compress_force = false;
 401
 402         cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
 403         if (cache_gen)
 404                 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 405
 406         if (!options)
 407                 goto out;
 408
 409         /*
 410          * strsep changes the string, duplicate it because parse_options
 411          * gets called twice
 412          */
 413         options = kstrdup(options, GFP_NOFS);
 414         if (!options)
 415                 return -ENOMEM;
 416
 417         orig = options;
 418
 419         while ((p = strsep(&options, ",")) != NULL) {
 420                 int token;
 421                 if (!*p)
 422                         continue;
 423
 424                 token = match_token(p, tokens, args);
 425                 switch (token) {
 426                 case Opt_degraded:
 427                         btrfs_info(root->fs_info, "allowing degraded mounts");
 428                         btrfs_set_opt(info->mount_opt, DEGRADED);
 429                         break;
 430                 case Opt_subvol:
 431                 case Opt_subvolid:
 432                 case Opt_subvolrootid:
 433                 case Opt_device:
 434                         /*
 435                          * These are parsed by btrfs_parse_early_options
 436                          * and can be happily ignored here.
 437                          */
 438                         break;
 439                 case Opt_nodatasum:
 440                         btrfs_info(root->fs_info, "setting nodatasum");
 441                         btrfs_set_opt(info->mount_opt, NODATASUM);
 442                         break;
 443                 case Opt_datasum:
 444                         if (btrfs_test_opt(root, NODATACOW))
 445                                 btrfs_info(root->fs_info, "setting datasum, datacow enabled");
 446                         else
 447                                 btrfs_info(root->fs_info, "setting datasum");
 448                         btrfs_clear_opt(info->mount_opt, NODATACOW);
 449                         btrfs_clear_opt(info->mount_opt, NODATASUM);
 450                         break;
 451                 case Opt_nodatacow:
 452                         if (!btrfs_test_opt(root, COMPRESS) ||
 453                                 !btrfs_test_opt(root, FORCE_COMPRESS)) {
 454                                         btrfs_info(root->fs_info,
 455                                                 "setting nodatacow, compression disabled");
 456                         } else {
 457                                 btrfs_info(root->fs_info, "setting nodatacow");
 458                         }
 459                         btrfs_clear_opt(info->mount_opt, COMPRESS);
 460                         btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 461                         btrfs_set_opt(info->mount_opt, NODATACOW);
 462                         btrfs_set_opt(info->mount_opt, NODATASUM);
 463                         break;
 464                 case Opt_datacow:
 465                         if (btrfs_test_opt(root, NODATACOW))
 466                                 btrfs_info(root->fs_info, "setting datacow");
 467                         btrfs_clear_opt(info->mount_opt, NODATACOW);
 468                         break;
 469                 case Opt_compress_force:
 470                 case Opt_compress_force_type:
 471                         compress_force = true;
 472                         /* Fallthrough */
 473                 case Opt_compress:
 474                 case Opt_compress_type:
 475                         if (token == Opt_compress ||
 476                             token == Opt_compress_force ||
 477                             strcmp(args[0].from, "zlib") == 0) {
 478                                 compress_type = "zlib";
 479                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 480                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 481                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 482                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 483                         } else if (strcmp(args[0].from, "lzo") == 0) {
 484                                 compress_type = "lzo";
 485                                 info->compress_type = BTRFS_COMPRESS_LZO;
 486                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 487                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 488                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 489                                 btrfs_set_fs_incompat(info, COMPRESS_LZO);
 490                         } else if (strncmp(args[0].from, "no", 2) == 0) {
 491                                 compress_type = "no";
 492                                 btrfs_clear_opt(info->mount_opt, COMPRESS);
 493                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 494                                 compress_force = false;
 495                         } else {
 496                                 ret = -EINVAL;
 497                                 goto out;
 498                         }
 499
 500                         if (compress_force) {
 501                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 502                                 btrfs_info(root->fs_info, "force %s compression",
 503                                         compress_type);
 504                         } else if (btrfs_test_opt(root, COMPRESS)) {
 505                                 pr_info("btrfs: use %s compression\n",
 506                                         compress_type);
 507                         }
 508                         break;
 509                 case Opt_ssd:
 510                         btrfs_info(root->fs_info, "use ssd allocation scheme");
 511                         btrfs_set_opt(info->mount_opt, SSD);
 512                         break;
 513                 case Opt_ssd_spread:
 514                         btrfs_info(root->fs_info, "use spread ssd allocation scheme");
 515                         btrfs_set_opt(info->mount_opt, SSD);
 516                         btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 517                         break;
 518                 case Opt_nossd:
 519                         btrfs_info(root->fs_info, "not using ssd allocation scheme");
 520                         btrfs_set_opt(info->mount_opt, NOSSD);
 521                         btrfs_clear_opt(info->mount_opt, SSD);
 522                         btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 523                         break;
 524                 case Opt_barrier:
 525                         if (btrfs_test_opt(root, NOBARRIER))
 526                                 btrfs_info(root->fs_info, "turning on barriers");
 527                         btrfs_clear_opt(info->mount_opt, NOBARRIER);
 528                         break;
 529                 case Opt_nobarrier:
 530                         btrfs_info(root->fs_info, "turning off barriers");
 531                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 532                         break;
 533                 case Opt_thread_pool:
 534                         ret = match_int(&args[0], &intarg);
 535                         if (ret) {
 536                                 goto out;
 537                         } else if (intarg > 0) {
 538                                 info->thread_pool_size = intarg;
 539                         } else {
 540                                 ret = -EINVAL;
 541                                 goto out;
 542                         }
 543                         break;
 544                 case Opt_max_inline:
 545                         num = match_strdup(&args[0]);
 546                         if (num) {
 547                                 info->max_inline = memparse(num, NULL);
 548                                 kfree(num);
 549
 550                                 if (info->max_inline) {
 551                                         info->max_inline = max_t(u64,
 552                                                 info->max_inline,
 553                                                 root->sectorsize);
 554                                 }
 555                                 btrfs_info(root->fs_info, "max_inline at %llu",
 556                                         info->max_inline);
 557                         } else {
 558                                 ret = -ENOMEM;
 559                                 goto out;
 560                         }
 561                         break;
 562                 case Opt_alloc_start:
 563                         num = match_strdup(&args[0]);
 564                         if (num) {
 565                                 mutex_lock(&info->chunk_mutex);
 566                                 info->alloc_start = memparse(num, NULL);
 567                                 mutex_unlock(&info->chunk_mutex);
 568                                 kfree(num);
 569                                 btrfs_info(root->fs_info, "allocations start at %llu",
 570                                         info->alloc_start);
 571                         } else {
 572                                 ret = -ENOMEM;
 573                                 goto out;
 574                         }
 575                         break;
 576                 case Opt_acl:
 577                         root->fs_info->sb->s_flags |= MS_POSIXACL;
 578                         break;
 579                 case Opt_noacl:
 580                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 581                         break;
 582                 case Opt_notreelog:
 583                         btrfs_info(root->fs_info, "disabling tree log");
 584                         btrfs_set_opt(info->mount_opt, NOTREELOG);
 585                         break;
 586                 case Opt_treelog:
 587                         if (btrfs_test_opt(root, NOTREELOG))
 588                                 btrfs_info(root->fs_info, "enabling tree log");
 589                         btrfs_clear_opt(info->mount_opt, NOTREELOG);
 590                         break;
 591                 case Opt_flushoncommit:
 592                         btrfs_info(root->fs_info, "turning on flush-on-commit");
 593                         btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
 594                         break;
 595                 case Opt_noflushoncommit:
 596                         if (btrfs_test_opt(root, FLUSHONCOMMIT))
 597                                 btrfs_info(root->fs_info, "turning off flush-on-commit");
 598                         btrfs_clear_opt(info->mount_opt, FLUSHONCOMMIT);
 599                         break;
 600                 case Opt_ratio:
 601                         ret = match_int(&args[0], &intarg);
 602                         if (ret) {
 603                                 goto out;
 604                         } else if (intarg >= 0) {
 605                                 info->metadata_ratio = intarg;
 606                                 btrfs_info(root->fs_info, "metadata ratio %d",
 607                                        info->metadata_ratio);
 608                         } else {
 609                                 ret = -EINVAL;
 610                                 goto out;
 611                         }
 612                         break;
 613                 case Opt_discard:
 614                         btrfs_set_opt(info->mount_opt, DISCARD);
 615                         break;
 616                 case Opt_nodiscard:
 617                         btrfs_clear_opt(info->mount_opt, DISCARD);
 618                         break;
 619                 case Opt_space_cache:
 620                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 621                         break;
 622                 case Opt_rescan_uuid_tree:
 623                         btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 624                         break;
 625                 case Opt_no_space_cache:
 626                         btrfs_info(root->fs_info, "disabling disk space caching");
 627                         btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
 628                         break;
 629                 case Opt_inode_cache:
 630                         btrfs_info(root->fs_info, "enabling inode map caching");
 631                         btrfs_set_opt(info->mount_opt, CHANGE_INODE_CACHE);
 632                         break;
 633                 case Opt_noinode_cache:
 634                         if (btrfs_test_opt(root, CHANGE_INODE_CACHE))
 635                                 btrfs_info(root->fs_info, "disabling inode map caching");
 636                         btrfs_clear_opt(info->mount_opt, CHANGE_INODE_CACHE);
 637                         break;
 638                 case Opt_clear_cache:
 639                         btrfs_info(root->fs_info, "force clearing of disk cache");
 640                         btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
 641                         break;
 642                 case Opt_user_subvol_rm_allowed:
 643                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 644                         break;
 645                 case Opt_enospc_debug:
 646                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 647                         break;
 648                 case Opt_noenospc_debug:
 649                         btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
 650                         break;
 651                 case Opt_defrag:
 652                         btrfs_info(root->fs_info, "enabling auto defrag");
 653                         btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
 654                         break;
 655                 case Opt_nodefrag:
 656                         if (btrfs_test_opt(root, AUTO_DEFRAG))
 657                                 btrfs_info(root->fs_info, "disabling auto defrag");
 658                         btrfs_clear_opt(info->mount_opt, AUTO_DEFRAG);
 659                         break;
 660                 case Opt_recovery:
 661                         btrfs_info(root->fs_info, "enabling auto recovery");
 662                         btrfs_set_opt(info->mount_opt, RECOVERY);
 663                         break;
 664                 case Opt_skip_balance:
 665                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 666                         break;
 667 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 668                 case Opt_check_integrity_including_extent_data:
 669                         btrfs_info(root->fs_info,
 670                                    "enabling check integrity including extent data");
 671                         btrfs_set_opt(info->mount_opt,
 672                                       CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
 673                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 674                         break;
 675                 case Opt_check_integrity:
 676                         btrfs_info(root->fs_info, "enabling check integrity");
 677                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 678                         break;
 679                 case Opt_check_integrity_print_mask:
 680                         ret = match_int(&args[0], &intarg);
 681                         if (ret) {
 682                                 goto out;
 683                         } else if (intarg >= 0) {
 684                                 info->check_integrity_print_mask = intarg;
 685                                 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
 686                                        info->check_integrity_print_mask);
 687                         } else {
 688                                 ret = -EINVAL;
 689                                 goto out;
 690                         }
 691                         break;
 692 #else
 693                 case Opt_check_integrity_including_extent_data:
 694                 case Opt_check_integrity:
 695                 case Opt_check_integrity_print_mask:
 696                         btrfs_err(root->fs_info,
 697                                 "support for check_integrity* not compiled in!");
 698                         ret = -EINVAL;
 699                         goto out;
 700 #endif
 701                 case Opt_fatal_errors:
 702                         if (strcmp(args[0].from, "panic") == 0)
 703                                 btrfs_set_opt(info->mount_opt,
 704                                               PANIC_ON_FATAL_ERROR);
 705                         else if (strcmp(args[0].from, "bug") == 0)
 706                                 btrfs_clear_opt(info->mount_opt,
 707                                               PANIC_ON_FATAL_ERROR);
 708                         else {
 709                                 ret = -EINVAL;
 710                                 goto out;
 711                         }
 712                         break;
 713                 case Opt_commit_interval:
 714                         intarg = 0;
 715                         ret = match_int(&args[0], &intarg);
 716                         if (ret < 0) {
 717                                 btrfs_err(root->fs_info, "invalid commit interval");
 718                                 ret = -EINVAL;
 719                                 goto out;
 720                         }
 721                         if (intarg > 0) {
 722                                 if (intarg > 300) {
 723                                         btrfs_warn(root->fs_info, "excessive commit interval %d",
 724                                                         intarg);
 725                                 }
 726                                 info->commit_interval = intarg;
 727                         } else {
 728                                 btrfs_info(root->fs_info, "using default commit interval %ds",
 729                                     BTRFS_DEFAULT_COMMIT_INTERVAL);
 730                                 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
 731                         }
 732                         break;
 733                 case Opt_err:
 734                         btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
 735                         ret = -EINVAL;
 736                         goto out;
 737                 default:
 738                         break;
 739                 }
 740         }
 741 out:
 742         if (!ret && btrfs_test_opt(root, SPACE_CACHE))
 743                 btrfs_info(root->fs_info, "disk space caching is enabled");
 744         kfree(orig);
 745         return ret;
 746 }
 747
 748 /*
 749  * Parse mount options that are required early in the mount process.
 750  *
 751  * All other options will be parsed on much later in the mount process and
 752  * only when we need to allocate a new super block.
 753  */
 754 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 755                 void *holder, char **subvol_name, u64 *subvol_objectid,
 756                 struct btrfs_fs_devices **fs_devices)
 757 {
 758         substring_t args[MAX_OPT_ARGS];
 759         char *device_name, *opts, *orig, *p;
 760         char *num = NULL;
 761         int error = 0;
 762
 763         if (!options)
 764                 return 0;
 765
 766         /*
 767          * strsep changes the string, duplicate it because parse_options
 768          * gets called twice
 769          */
 770         opts = kstrdup(options, GFP_KERNEL);
 771         if (!opts)
 772                 return -ENOMEM;
 773         orig = opts;
 774
 775         while ((p = strsep(&opts, ",")) != NULL) {
 776                 int token;
 777                 if (!*p)
 778                         continue;
 779
 780                 token = match_token(p, tokens, args);
 781                 switch (token) {
 782                 case Opt_subvol:
 783                         kfree(*subvol_name);
 784                         *subvol_name = match_strdup(&args[0]);
 785                         if (!*subvol_name) {
 786                                 error = -ENOMEM;
 787                                 goto out;
 788                         }
 789                         break;
 790                 case Opt_subvolid:
 791                         num = match_strdup(&args[0]);
 792                         if (num) {
 793                                 *subvol_objectid = memparse(num, NULL);
 794                                 kfree(num);
 795                                 /* we want the original fs_tree */
 796                                 if (!*subvol_objectid)
 797                                         *subvol_objectid =
 798                                                 BTRFS_FS_TREE_OBJECTID;
 799                         } else {
 800                                 error = -EINVAL;
 801                                 goto out;
 802                         }
 803                         break;
 804                 case Opt_subvolrootid:
 805                         printk(KERN_WARNING
 806                                 "BTRFS: 'subvolrootid' mount option is deprecated and has "
 807                                 "no effect\n");
 808                         break;
 809                 case Opt_device:
 810                         device_name = match_strdup(&args[0]);
 811                         if (!device_name) {
 812                                 error = -ENOMEM;
 813                                 goto out;
 814                         }
 815                         error = btrfs_scan_one_device(device_name,
 816                                         flags, holder, fs_devices);
 817                         kfree(device_name);
 818                         if (error)
 819                                 goto out;
 820                         break;
 821                 default:
 822                         break;
 823                 }
 824         }
 825
 826 out:
 827         kfree(orig);
 828         return error;
 829 }
 830
 831 static struct dentry *get_default_root(struct super_block *sb,
 832                                        u64 subvol_objectid)
 833 {
 834         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 835         struct btrfs_root *root = fs_info->tree_root;
 836         struct btrfs_root *new_root;
 837         struct btrfs_dir_item *di;
 838         struct btrfs_path *path;
 839         struct btrfs_key location;
 840         struct inode *inode;
 841         u64 dir_id;
 842         int new = 0;
 843
 844         /*
 845          * We have a specific subvol we want to mount, just setup location and
 846          * go look up the root.
 847          */
 848         if (subvol_objectid) {
 849                 location.objectid = subvol_objectid;
 850                 location.type = BTRFS_ROOT_ITEM_KEY;
 851                 location.offset = (u64)-1;
 852                 goto find_root;
 853         }
 854
 855         path = btrfs_alloc_path();
 856         if (!path)
 857                 return ERR_PTR(-ENOMEM);
 858         path->leave_spinning = 1;
 859
 860         /*
 861          * Find the "default" dir item which points to the root item that we
 862          * will mount by default if we haven't been given a specific subvolume
 863          * to mount.
 864          */
 865         dir_id = btrfs_super_root_dir(fs_info->super_copy);
 866         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 867         if (IS_ERR(di)) {
 868                 btrfs_free_path(path);
 869                 return ERR_CAST(di);
 870         }
 871         if (!di) {
 872                 /*
 873                  * Ok the default dir item isn't there.  This is weird since
 874                  * it's always been there, but don't freak out, just try and
 875                  * mount to root most subvolume.
 876                  */
 877                 btrfs_free_path(path);
 878                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 879                 new_root = fs_info->fs_root;
 880                 goto setup_root;
 881         }
 882
 883         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 884         btrfs_free_path(path);
 885
 886 find_root:
 887         new_root = btrfs_read_fs_root_no_name(fs_info, &location);
 888         if (IS_ERR(new_root))
 889                 return ERR_CAST(new_root);
 890
 891         dir_id = btrfs_root_dirid(&new_root->root_item);
 892 setup_root:
 893         location.objectid = dir_id;
 894         location.type = BTRFS_INODE_ITEM_KEY;
 895         location.offset = 0;
 896
 897         inode = btrfs_iget(sb, &location, new_root, &new);
 898         if (IS_ERR(inode))
 899                 return ERR_CAST(inode);
 900
 901         /*
 902          * If we're just mounting the root most subvol put the inode and return
 903          * a reference to the dentry.  We will have already gotten a reference
 904          * to the inode in btrfs_fill_super so we're good to go.
 905          */
 906         if (!new && sb->s_root->d_inode == inode) {
 907                 iput(inode);
 908                 return dget(sb->s_root);
 909         }
 910
 911         return d_obtain_alias(inode);
 912 }
 913
 914 static int btrfs_fill_super(struct super_block *sb,
 915                             struct btrfs_fs_devices *fs_devices,
 916                             void *data, int silent)
 917 {
 918         struct inode *inode;
 919         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 920         struct btrfs_key key;
 921         int err;
 922
 923         sb->s_maxbytes = MAX_LFS_FILESIZE;
 924         sb->s_magic = BTRFS_SUPER_MAGIC;
 925         sb->s_op = &btrfs_super_ops;
 926         sb->s_d_op = &btrfs_dentry_operations;
 927         sb->s_export_op = &btrfs_export_ops;
 928         sb->s_xattr = btrfs_xattr_handlers;
 929         sb->s_time_gran = 1;
 930 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 931         sb->s_flags |= MS_POSIXACL;
 932 #endif
 933         sb->s_flags |= MS_I_VERSION;
 934         err = open_ctree(sb, fs_devices, (char *)data);
 935         if (err) {
 936                 printk(KERN_ERR "BTRFS: open_ctree failed\n");
 937                 return err;
 938         }
 939
 940         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 941         key.type = BTRFS_INODE_ITEM_KEY;
 942         key.offset = 0;
 943         inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
 944         if (IS_ERR(inode)) {
 945                 err = PTR_ERR(inode);
 946                 goto fail_close;
 947         }
 948
 949         sb->s_root = d_make_root(inode);
 950         if (!sb->s_root) {
 951                 err = -ENOMEM;
 952                 goto fail_close;
 953         }
 954
 955         save_mount_options(sb, data);
 956         cleancache_init_fs(sb);
 957         sb->s_flags |= MS_ACTIVE;
 958         return 0;
 959
 960 fail_close:
 961         close_ctree(fs_info->tree_root);
 962         return err;
 963 }
 964
 965 int btrfs_sync_fs(struct super_block *sb, int wait)
 966 {
 967         struct btrfs_trans_handle *trans;
 968         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 969         struct btrfs_root *root = fs_info->tree_root;
 970
 971         trace_btrfs_sync_fs(wait);
 972
 973         if (!wait) {
 974                 filemap_flush(fs_info->btree_inode->i_mapping);
 975                 return 0;
 976         }
 977
 978         btrfs_wait_ordered_roots(fs_info, -1);
 979
 980         trans = btrfs_attach_transaction_barrier(root);
 981         if (IS_ERR(trans)) {
 982                 /* no transaction, don't bother */
 983                 if (PTR_ERR(trans) == -ENOENT)
 984                         return 0;
 985                 return PTR_ERR(trans);
 986         }
 987         return btrfs_commit_transaction(trans, root);
 988 }
 989
 990 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
 991 {
 992         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
 993         struct btrfs_root *root = info->tree_root;
 994         char *compress_type;
 995
 996         if (btrfs_test_opt(root, DEGRADED))
 997                 seq_puts(seq, ",degraded");
 998         if (btrfs_test_opt(root, NODATASUM))
 999                 seq_puts(seq, ",nodatasum");
1000         if (btrfs_test_opt(root, NODATACOW))
1001                 seq_puts(seq, ",nodatacow");
1002         if (btrfs_test_opt(root, NOBARRIER))
1003                 seq_puts(seq, ",nobarrier");
1004         if (info->max_inline != 8192 * 1024)
1005                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1006         if (info->alloc_start != 0)
1007                 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1008         if (info->thread_pool_size !=  min_t(unsigned long,
1009                                              num_online_cpus() + 2, 8))
1010                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1011         if (btrfs_test_opt(root, COMPRESS)) {
1012                 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1013                         compress_type = "zlib";
1014                 else
1015                         compress_type = "lzo";
1016                 if (btrfs_test_opt(root, FORCE_COMPRESS))
1017                         seq_printf(seq, ",compress-force=%s", compress_type);
1018                 else
1019                         seq_printf(seq, ",compress=%s", compress_type);
1020         }
1021         if (btrfs_test_opt(root, NOSSD))
1022                 seq_puts(seq, ",nossd");
1023         if (btrfs_test_opt(root, SSD_SPREAD))
1024                 seq_puts(seq, ",ssd_spread");
1025         else if (btrfs_test_opt(root, SSD))
1026                 seq_puts(seq, ",ssd");
1027         if (btrfs_test_opt(root, NOTREELOG))
1028                 seq_puts(seq, ",notreelog");
1029         if (btrfs_test_opt(root, FLUSHONCOMMIT))
1030                 seq_puts(seq, ",flushoncommit");
1031         if (btrfs_test_opt(root, DISCARD))
1032                 seq_puts(seq, ",discard");
1033         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1034                 seq_puts(seq, ",noacl");
1035         if (btrfs_test_opt(root, SPACE_CACHE))
1036                 seq_puts(seq, ",space_cache");
1037         else
1038                 seq_puts(seq, ",nospace_cache");
1039         if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1040                 seq_puts(seq, ",rescan_uuid_tree");
1041         if (btrfs_test_opt(root, CLEAR_CACHE))
1042                 seq_puts(seq, ",clear_cache");
1043         if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1044                 seq_puts(seq, ",user_subvol_rm_allowed");
1045         if (btrfs_test_opt(root, ENOSPC_DEBUG))
1046                 seq_puts(seq, ",enospc_debug");
1047         if (btrfs_test_opt(root, AUTO_DEFRAG))
1048                 seq_puts(seq, ",autodefrag");
1049         if (btrfs_test_opt(root, INODE_MAP_CACHE))
1050                 seq_puts(seq, ",inode_cache");
1051         if (btrfs_test_opt(root, SKIP_BALANCE))
1052                 seq_puts(seq, ",skip_balance");
1053         if (btrfs_test_opt(root, RECOVERY))
1054                 seq_puts(seq, ",recovery");
1055 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1056         if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1057                 seq_puts(seq, ",check_int_data");
1058         else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1059                 seq_puts(seq, ",check_int");
1060         if (info->check_integrity_print_mask)
1061                 seq_printf(seq, ",check_int_print_mask=%d",
1062                                 info->check_integrity_print_mask);
1063 #endif
1064         if (info->metadata_ratio)
1065                 seq_printf(seq, ",metadata_ratio=%d",
1066                                 info->metadata_ratio);
1067         if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1068                 seq_puts(seq, ",fatal_errors=panic");
1069         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1070                 seq_printf(seq, ",commit=%d", info->commit_interval);
1071         return 0;
1072 }
1073
1074 static int btrfs_test_super(struct super_block *s, void *data)
1075 {
1076         struct btrfs_fs_info *p = data;
1077         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1078
1079         return fs_info->fs_devices == p->fs_devices;
1080 }
1081
1082 static int btrfs_set_super(struct super_block *s, void *data)
1083 {
1084         int err = set_anon_super(s, data);
1085         if (!err)
1086                 s->s_fs_info = data;
1087         return err;
1088 }
1089
1090 /*
1091  * subvolumes are identified by ino 256
1092  */
1093 static inline int is_subvolume_inode(struct inode *inode)
1094 {
1095         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1096                 return 1;
1097         return 0;
1098 }
1099
1100 /*
1101  * This will strip out the subvol=%s argument for an argument string and add
1102  * subvolid=0 to make sure we get the actual tree root for path walking to the
1103  * subvol we want.
1104  */
1105 static char *setup_root_args(char *args)
1106 {
1107         unsigned len = strlen(args) + 2 + 1;
1108         char *src, *dst, *buf;
1109
1110         /*
1111          * We need the same args as before, but with this substitution:
1112          * s!subvol=[^,]+!subvolid=0!
1113          *
1114          * Since the replacement string is up to 2 bytes longer than the
1115          * original, allocate strlen(args) + 2 + 1 bytes.
1116          */
1117
1118         src = strstr(args, "subvol=");
1119         /* This shouldn't happen, but just in case.. */
1120         if (!src)
1121                 return NULL;
1122
1123         buf = dst = kmalloc(len, GFP_NOFS);
1124         if (!buf)
1125                 return NULL;
1126
1127         /*
1128          * If the subvol= arg is not at the start of the string,
1129          * copy whatever precedes it into buf.
1130          */
1131         if (src != args) {
1132                 *src++ = '\0';
1133                 strcpy(buf, args);
1134                 dst += strlen(args);
1135         }
1136
1137         strcpy(dst, "subvolid=0");
1138         dst += strlen("subvolid=0");
1139
1140         /*
1141          * If there is a "," after the original subvol=... string,
1142          * copy that suffix into our buffer.  Otherwise, we're done.
1143          */
1144         src = strchr(src, ',');
1145         if (src)
1146                 strcpy(dst, src);
1147
1148         return buf;
1149 }
1150
1151 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1152                                    const char *device_name, char *data)
1153 {
1154         struct dentry *root;
1155         struct vfsmount *mnt;
1156         char *newargs;
1157
1158         newargs = setup_root_args(data);
1159         if (!newargs)
1160                 return ERR_PTR(-ENOMEM);
1161         mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1162                              newargs);
1163         kfree(newargs);
1164         if (IS_ERR(mnt))
1165                 return ERR_CAST(mnt);
1166
1167         root = mount_subtree(mnt, subvol_name);
1168
1169         if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1170                 struct super_block *s = root->d_sb;
1171                 dput(root);
1172                 root = ERR_PTR(-EINVAL);
1173                 deactivate_locked_super(s);
1174                 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1175                                 subvol_name);
1176         }
1177
1178         return root;
1179 }
1180
1181 /*
1182  * Find a superblock for the given device / mount point.
1183  *
1184  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
1185  *        for multiple device setup.  Make sure to keep it in sync.
1186  */
1187 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1188                 const char *device_name, void *data)
1189 {
1190         struct block_device *bdev = NULL;
1191         struct super_block *s;
1192         struct dentry *root;
1193         struct btrfs_fs_devices *fs_devices = NULL;
1194         struct btrfs_fs_info *fs_info = NULL;
1195         fmode_t mode = FMODE_READ;
1196         char *subvol_name = NULL;
1197         u64 subvol_objectid = 0;
1198         int error = 0;
1199
1200         if (!(flags & MS_RDONLY))
1201                 mode |= FMODE_WRITE;
1202
1203         error = btrfs_parse_early_options(data, mode, fs_type,
1204                                           &subvol_name, &subvol_objectid,
1205                                           &fs_devices);
1206         if (error) {
1207                 kfree(subvol_name);
1208                 return ERR_PTR(error);
1209         }
1210
1211         if (subvol_name) {
1212                 root = mount_subvol(subvol_name, flags, device_name, data);
1213                 kfree(subvol_name);
1214                 return root;
1215         }
1216
1217         error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1218         if (error)
1219                 return ERR_PTR(error);
1220
1221         /*
1222          * Setup a dummy root and fs_info for test/set super.  This is because
1223          * we don't actually fill this stuff out until open_ctree, but we need
1224          * it for searching for existing supers, so this lets us do that and
1225          * then open_ctree will properly initialize everything later.
1226          */
1227         fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1228         if (!fs_info)
1229                 return ERR_PTR(-ENOMEM);
1230
1231         fs_info->fs_devices = fs_devices;
1232
1233         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1234         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1235         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1236                 error = -ENOMEM;
1237                 goto error_fs_info;
1238         }
1239
1240         error = btrfs_open_devices(fs_devices, mode, fs_type);
1241         if (error)
1242                 goto error_fs_info;
1243
1244         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1245                 error = -EACCES;
1246                 goto error_close_devices;
1247         }
1248
1249         bdev = fs_devices->latest_bdev;
1250         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1251                  fs_info);
1252         if (IS_ERR(s)) {
1253                 error = PTR_ERR(s);
1254                 goto error_close_devices;
1255         }
1256
1257         if (s->s_root) {
1258                 btrfs_close_devices(fs_devices);
1259                 free_fs_info(fs_info);
1260                 if ((flags ^ s->s_flags) & MS_RDONLY)
1261                         error = -EBUSY;
1262         } else {
1263                 char b[BDEVNAME_SIZE];
1264
1265                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1266                 btrfs_sb(s)->bdev_holder = fs_type;
1267                 error = btrfs_fill_super(s, fs_devices, data,
1268                                          flags & MS_SILENT ? 1 : 0);
1269         }
1270
1271         root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1272         if (IS_ERR(root))
1273                 deactivate_locked_super(s);
1274
1275         return root;
1276
1277 error_close_devices:
1278         btrfs_close_devices(fs_devices);
1279 error_fs_info:
1280         free_fs_info(fs_info);
1281         return ERR_PTR(error);
1282 }
1283
1284 static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1285 {
1286         spin_lock_irq(&workers->lock);
1287         workers->max_workers = new_limit;
1288         spin_unlock_irq(&workers->lock);
1289 }
1290
1291 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1292                                      int new_pool_size, int old_pool_size)
1293 {
1294         if (new_pool_size == old_pool_size)
1295                 return;
1296
1297         fs_info->thread_pool_size = new_pool_size;
1298
1299         btrfs_info(fs_info, "resize thread pool %d -> %d",
1300                old_pool_size, new_pool_size);
1301
1302         btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1303         btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1304         btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1305         btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1306         btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1307         btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1308         btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1309         btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1310         btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1311         btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1312         btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1313         btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1314         btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
1315         btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers,
1316                               new_pool_size);
1317 }
1318
1319 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1320 {
1321         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1322 }
1323
1324 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1325                                        unsigned long old_opts, int flags)
1326 {
1327         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1328             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1329              (flags & MS_RDONLY))) {
1330                 /* wait for any defraggers to finish */
1331                 wait_event(fs_info->transaction_wait,
1332                            (atomic_read(&fs_info->defrag_running) == 0));
1333                 if (flags & MS_RDONLY)
1334                         sync_filesystem(fs_info->sb);
1335         }
1336 }
1337
1338 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1339                                          unsigned long old_opts)
1340 {
1341         /*
1342          * We need cleanup all defragable inodes if the autodefragment is
1343          * close or the fs is R/O.
1344          */
1345         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1346             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1347              (fs_info->sb->s_flags & MS_RDONLY))) {
1348                 btrfs_cleanup_defrag_inodes(fs_info);
1349         }
1350
1351         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1352 }
1353
1354 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1355 {
1356         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1357         struct btrfs_root *root = fs_info->tree_root;
1358         unsigned old_flags = sb->s_flags;
1359         unsigned long old_opts = fs_info->mount_opt;
1360         unsigned long old_compress_type = fs_info->compress_type;
1361         u64 old_max_inline = fs_info->max_inline;
1362         u64 old_alloc_start = fs_info->alloc_start;
1363         int old_thread_pool_size = fs_info->thread_pool_size;
1364         unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1365         int ret;
1366
1367         btrfs_remount_prepare(fs_info);
1368
1369         ret = btrfs_parse_options(root, data);
1370         if (ret) {
1371                 ret = -EINVAL;
1372                 goto restore;
1373         }
1374
1375         btrfs_remount_begin(fs_info, old_opts, *flags);
1376         btrfs_resize_thread_pool(fs_info,
1377                 fs_info->thread_pool_size, old_thread_pool_size);
1378
1379         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1380                 goto out;
1381
1382         if (*flags & MS_RDONLY) {
1383                 /*
1384                  * this also happens on 'umount -rf' or on shutdown, when
1385                  * the filesystem is busy.
1386                  */
1387
1388                 /* wait for the uuid_scan task to finish */
1389                 down(&fs_info->uuid_tree_rescan_sem);
1390                 /* avoid complains from lockdep et al. */
1391                 up(&fs_info->uuid_tree_rescan_sem);
1392
1393                 sb->s_flags |= MS_RDONLY;
1394
1395                 btrfs_dev_replace_suspend_for_unmount(fs_info);
1396                 btrfs_scrub_cancel(fs_info);
1397                 btrfs_pause_balance(fs_info);
1398
1399                 ret = btrfs_commit_super(root);
1400                 if (ret)
1401                         goto restore;
1402         } else {
1403                 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1404                         btrfs_err(fs_info,
1405                                 "Remounting read-write after error is not allowed");
1406                         ret = -EINVAL;
1407                         goto restore;
1408                 }
1409                 if (fs_info->fs_devices->rw_devices == 0) {
1410                         ret = -EACCES;
1411                         goto restore;
1412                 }
1413
1414                 if (fs_info->fs_devices->missing_devices >
1415                      fs_info->num_tolerated_disk_barrier_failures &&
1416                     !(*flags & MS_RDONLY)) {
1417                         btrfs_warn(fs_info,
1418                                 "too many missing devices, writeable remount is not allowed");
1419                         ret = -EACCES;
1420                         goto restore;
1421                 }
1422
1423                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1424                         ret = -EINVAL;
1425                         goto restore;
1426                 }
1427
1428                 ret = btrfs_cleanup_fs_roots(fs_info);
1429                 if (ret)
1430                         goto restore;
1431
1432                 /* recover relocation */
1433                 ret = btrfs_recover_relocation(root);
1434                 if (ret)
1435                         goto restore;
1436
1437                 ret = btrfs_resume_balance_async(fs_info);
1438                 if (ret)
1439                         goto restore;
1440
1441                 ret = btrfs_resume_dev_replace_async(fs_info);
1442                 if (ret) {
1443                         btrfs_warn(fs_info, "failed to resume dev_replace");
1444                         goto restore;
1445                 }
1446
1447                 if (!fs_info->uuid_root) {
1448                         btrfs_info(fs_info, "creating UUID tree");
1449                         ret = btrfs_create_uuid_tree(fs_info);
1450                         if (ret) {
1451                                 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1452                                 goto restore;
1453                         }
1454                 }
1455                 sb->s_flags &= ~MS_RDONLY;
1456         }
1457 out:
1458         btrfs_remount_cleanup(fs_info, old_opts);
1459         return 0;
1460
1461 restore:
1462         /* We've hit an error - don't reset MS_RDONLY */
1463         if (sb->s_flags & MS_RDONLY)
1464                 old_flags |= MS_RDONLY;
1465         sb->s_flags = old_flags;
1466         fs_info->mount_opt = old_opts;
1467         fs_info->compress_type = old_compress_type;
1468         fs_info->max_inline = old_max_inline;
1469         mutex_lock(&fs_info->chunk_mutex);
1470         fs_info->alloc_start = old_alloc_start;
1471         mutex_unlock(&fs_info->chunk_mutex);
1472         btrfs_resize_thread_pool(fs_info,
1473                 old_thread_pool_size, fs_info->thread_pool_size);
1474         fs_info->metadata_ratio = old_metadata_ratio;
1475         btrfs_remount_cleanup(fs_info, old_opts);
1476         return ret;
1477 }
1478
1479 /* Used to sort the devices by max_avail(descending sort) */
1480 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1481                                        const void *dev_info2)
1482 {
1483         if (((struct btrfs_device_info *)dev_info1)->max_avail >
1484             ((struct btrfs_device_info *)dev_info2)->max_avail)
1485                 return -1;
1486         else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1487                  ((struct btrfs_device_info *)dev_info2)->max_avail)
1488                 return 1;
1489         else
1490         return 0;
1491 }
1492
1493 /*
1494  * sort the devices by max_avail, in which max free extent size of each device
1495  * is stored.(Descending Sort)
1496  */
1497 static inline void btrfs_descending_sort_devices(
1498                                         struct btrfs_device_info *devices,
1499                                         size_t nr_devices)
1500 {
1501         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1502              btrfs_cmp_device_free_bytes, NULL);
1503 }
1504
1505 /*
1506  * The helper to calc the free space on the devices that can be used to store
1507  * file data.
1508  */
1509 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1510 {
1511         struct btrfs_fs_info *fs_info = root->fs_info;
1512         struct btrfs_device_info *devices_info;
1513         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1514         struct btrfs_device *device;
1515         u64 skip_space;
1516         u64 type;
1517         u64 avail_space;
1518         u64 used_space;
1519         u64 min_stripe_size;
1520         int min_stripes = 1, num_stripes = 1;
1521         int i = 0, nr_devices;
1522         int ret;
1523
1524         nr_devices = fs_info->fs_devices->open_devices;
1525         BUG_ON(!nr_devices);
1526
1527         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1528                                GFP_NOFS);
1529         if (!devices_info)
1530                 return -ENOMEM;
1531
1532         /* calc min stripe number for data space alloction */
1533         type = btrfs_get_alloc_profile(root, 1);
1534         if (type & BTRFS_BLOCK_GROUP_RAID0) {
1535                 min_stripes = 2;
1536                 num_stripes = nr_devices;
1537         } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1538                 min_stripes = 2;
1539                 num_stripes = 2;
1540         } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1541                 min_stripes = 4;
1542                 num_stripes = 4;
1543         }
1544
1545         if (type & BTRFS_BLOCK_GROUP_DUP)
1546                 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1547         else
1548                 min_stripe_size = BTRFS_STRIPE_LEN;
1549
1550         list_for_each_entry(device, &fs_devices->devices, dev_list) {
1551                 if (!device->in_fs_metadata || !device->bdev ||
1552                     device->is_tgtdev_for_dev_replace)
1553                         continue;
1554
1555                 avail_space = device->total_bytes - device->bytes_used;
1556
1557                 /* align with stripe_len */
1558                 do_div(avail_space, BTRFS_STRIPE_LEN);
1559                 avail_space *= BTRFS_STRIPE_LEN;
1560
1561                 /*
1562                  * In order to avoid overwritting the superblock on the drive,
1563                  * btrfs starts at an offset of at least 1MB when doing chunk
1564                  * allocation.
1565                  */
1566                 skip_space = 1024 * 1024;
1567
1568                 /* user can set the offset in fs_info->alloc_start. */
1569                 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1570                     device->total_bytes)
1571                         skip_space = max(fs_info->alloc_start, skip_space);
1572
1573                 /*
1574                  * btrfs can not use the free space in [0, skip_space - 1],
1575                  * we must subtract it from the total. In order to implement
1576                  * it, we account the used space in this range first.
1577                  */
1578                 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1579                                                      &used_space);
1580                 if (ret) {
1581                         kfree(devices_info);
1582                         return ret;
1583                 }
1584
1585                 /* calc the free space in [0, skip_space - 1] */
1586                 skip_space -= used_space;
1587
1588                 /*
1589                  * we can use the free space in [0, skip_space - 1], subtract
1590                  * it from the total.
1591                  */
1592                 if (avail_space && avail_space >= skip_space)
1593                         avail_space -= skip_space;
1594                 else
1595                         avail_space = 0;
1596
1597                 if (avail_space < min_stripe_size)
1598                         continue;
1599
1600                 devices_info[i].dev = device;
1601                 devices_info[i].max_avail = avail_space;
1602
1603                 i++;
1604         }
1605
1606         nr_devices = i;
1607
1608         btrfs_descending_sort_devices(devices_info, nr_devices);
1609
1610         i = nr_devices - 1;
1611         avail_space = 0;
1612         while (nr_devices >= min_stripes) {
1613                 if (num_stripes > nr_devices)
1614                         num_stripes = nr_devices;
1615
1616                 if (devices_info[i].max_avail >= min_stripe_size) {
1617                         int j;
1618                         u64 alloc_size;
1619
1620                         avail_space += devices_info[i].max_avail * num_stripes;
1621                         alloc_size = devices_info[i].max_avail;
1622                         for (j = i + 1 - num_stripes; j <= i; j++)
1623                                 devices_info[j].max_avail -= alloc_size;
1624                 }
1625                 i--;
1626                 nr_devices--;
1627         }
1628
1629         kfree(devices_info);
1630         *free_bytes = avail_space;
1631         return 0;
1632 }
1633
1634 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1635 {
1636         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1637         struct btrfs_super_block *disk_super = fs_info->super_copy;
1638         struct list_head *head = &fs_info->space_info;
1639         struct btrfs_space_info *found;
1640         u64 total_used = 0;
1641         u64 total_free_data = 0;
1642         int bits = dentry->d_sb->s_blocksize_bits;
1643         __be32 *fsid = (__be32 *)fs_info->fsid;
1644         int ret;
1645
1646         /* holding chunk_muext to avoid allocating new chunks */
1647         mutex_lock(&fs_info->chunk_mutex);
1648         rcu_read_lock();
1649         list_for_each_entry_rcu(found, head, list) {
1650                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1651                         total_free_data += found->disk_total - found->disk_used;
1652                         total_free_data -=
1653                                 btrfs_account_ro_block_groups_free_space(found);
1654                 }
1655
1656                 total_used += found->disk_used;
1657         }
1658         rcu_read_unlock();
1659
1660         buf->f_namelen = BTRFS_NAME_LEN;
1661         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1662         buf->f_bfree = buf->f_blocks - (total_used >> bits);
1663         buf->f_bsize = dentry->d_sb->s_blocksize;
1664         buf->f_type = BTRFS_SUPER_MAGIC;
1665         buf->f_bavail = total_free_data;
1666         ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1667         if (ret) {
1668                 mutex_unlock(&fs_info->chunk_mutex);
1669                 return ret;
1670         }
1671         buf->f_bavail += total_free_data;
1672         buf->f_bavail = buf->f_bavail >> bits;
1673         mutex_unlock(&fs_info->chunk_mutex);
1674
1675         /* We treat it as constant endianness (it doesn't matter _which_)
1676            because we want the fsid to come out the same whether mounted
1677            on a big-endian or little-endian host */
1678         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1679         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1680         /* Mask in the root object ID too, to disambiguate subvols */
1681         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1682         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1683
1684         return 0;
1685 }
1686
1687 static void btrfs_kill_super(struct super_block *sb)
1688 {
1689         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1690         kill_anon_super(sb);
1691         free_fs_info(fs_info);
1692 }
1693
1694 static struct file_system_type btrfs_fs_type = {
1695         .owner          = THIS_MODULE,
1696         .name           = "btrfs",
1697         .mount          = btrfs_mount,
1698         .kill_sb        = btrfs_kill_super,
1699         .fs_flags       = FS_REQUIRES_DEV,
1700 };
1701 MODULE_ALIAS_FS("btrfs");
1702
1703 /*
1704  * used by btrfsctl to scan devices when no FS is mounted
1705  */
1706 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1707                                 unsigned long arg)
1708 {
1709         struct btrfs_ioctl_vol_args *vol;
1710         struct btrfs_fs_devices *fs_devices;
1711         int ret = -ENOTTY;
1712
1713         if (!capable(CAP_SYS_ADMIN))
1714                 return -EPERM;
1715
1716         vol = memdup_user((void __user *)arg, sizeof(*vol));
1717         if (IS_ERR(vol))
1718                 return PTR_ERR(vol);
1719
1720         switch (cmd) {
1721         case BTRFS_IOC_SCAN_DEV:
1722                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1723                                             &btrfs_fs_type, &fs_devices);
1724                 break;
1725         case BTRFS_IOC_DEVICES_READY:
1726                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1727                                             &btrfs_fs_type, &fs_devices);
1728                 if (ret)
1729                         break;
1730                 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1731                 break;
1732         }
1733
1734         kfree(vol);
1735         return ret;
1736 }
1737
1738 static int btrfs_freeze(struct super_block *sb)
1739 {
1740         struct btrfs_trans_handle *trans;
1741         struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1742
1743         trans = btrfs_attach_transaction_barrier(root);
1744         if (IS_ERR(trans)) {
1745                 /* no transaction, don't bother */
1746                 if (PTR_ERR(trans) == -ENOENT)
1747                         return 0;
1748                 return PTR_ERR(trans);
1749         }
1750         return btrfs_commit_transaction(trans, root);
1751 }
1752
1753 static int btrfs_unfreeze(struct super_block *sb)
1754 {
1755         return 0;
1756 }
1757
1758 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1759 {
1760         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1761         struct btrfs_fs_devices *cur_devices;
1762         struct btrfs_device *dev, *first_dev = NULL;
1763         struct list_head *head;
1764         struct rcu_string *name;
1765
1766         mutex_lock(&fs_info->fs_devices->device_list_mutex);
1767         cur_devices = fs_info->fs_devices;
1768         while (cur_devices) {
1769                 head = &cur_devices->devices;
1770                 list_for_each_entry(dev, head, dev_list) {
1771                         if (dev->missing)
1772                                 continue;
1773                         if (!first_dev || dev->devid < first_dev->devid)
1774                                 first_dev = dev;
1775                 }
1776                 cur_devices = cur_devices->seed;
1777         }
1778
1779         if (first_dev) {
1780                 rcu_read_lock();
1781                 name = rcu_dereference(first_dev->name);
1782                 seq_escape(m, name->str, " \t\n\\");
1783                 rcu_read_unlock();
1784         } else {
1785                 WARN_ON(1);
1786         }
1787         mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1788         return 0;
1789 }
1790
1791 static const struct super_operations btrfs_super_ops = {
1792         .drop_inode     = btrfs_drop_inode,
1793         .evict_inode    = btrfs_evict_inode,
1794         .put_super      = btrfs_put_super,
1795         .sync_fs        = btrfs_sync_fs,
1796         .show_options   = btrfs_show_options,
1797         .show_devname   = btrfs_show_devname,
1798         .write_inode    = btrfs_write_inode,
1799         .alloc_inode    = btrfs_alloc_inode,
1800         .destroy_inode  = btrfs_destroy_inode,
1801         .statfs         = btrfs_statfs,
1802         .remount_fs     = btrfs_remount,
1803         .freeze_fs      = btrfs_freeze,
1804         .unfreeze_fs    = btrfs_unfreeze,
1805 };
1806
1807 static const struct file_operations btrfs_ctl_fops = {
1808         .unlocked_ioctl  = btrfs_control_ioctl,
1809         .compat_ioctl = btrfs_control_ioctl,
1810         .owner   = THIS_MODULE,
1811         .llseek = noop_llseek,
1812 };
1813
1814 static struct miscdevice btrfs_misc = {
1815         .minor          = BTRFS_MINOR,
1816         .name           = "btrfs-control",
1817         .fops           = &btrfs_ctl_fops
1818 };
1819
1820 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1821 MODULE_ALIAS("devname:btrfs-control");
1822
1823 static int btrfs_interface_init(void)
1824 {
1825         return misc_register(&btrfs_misc);
1826 }
1827
1828 static void btrfs_interface_exit(void)
1829 {
1830         if (misc_deregister(&btrfs_misc) < 0)
1831                 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
1832 }
1833
1834 static void btrfs_print_info(void)
1835 {
1836         printk(KERN_INFO "Btrfs loaded"
1837 #ifdef CONFIG_BTRFS_DEBUG
1838                         ", debug=on"
1839 #endif
1840 #ifdef CONFIG_BTRFS_ASSERT
1841                         ", assert=on"
1842 #endif
1843 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1844                         ", integrity-checker=on"
1845 #endif
1846                         "\n");
1847 }
1848
1849 static int btrfs_run_sanity_tests(void)
1850 {
1851         int ret;
1852
1853         ret = btrfs_init_test_fs();
1854         if (ret)
1855                 return ret;
1856
1857         ret = btrfs_test_free_space_cache();
1858         if (ret)
1859                 goto out;
1860         ret = btrfs_test_extent_buffer_operations();
1861         if (ret)
1862                 goto out;
1863         ret = btrfs_test_extent_io();
1864         if (ret)
1865                 goto out;
1866         ret = btrfs_test_inodes();
1867 out:
1868         btrfs_destroy_test_fs();
1869         return ret;
1870 }
1871
1872 static int __init init_btrfs_fs(void)
1873 {
1874         int err;
1875
1876         err = btrfs_hash_init();
1877         if (err)
1878                 return err;
1879
1880         btrfs_props_init();
1881
1882         err = btrfs_init_sysfs();
1883         if (err)
1884                 goto free_hash;
1885
1886         btrfs_init_compress();
1887
1888         err = btrfs_init_cachep();
1889         if (err)
1890                 goto free_compress;
1891
1892         err = extent_io_init();
1893         if (err)
1894                 goto free_cachep;
1895
1896         err = extent_map_init();
1897         if (err)
1898                 goto free_extent_io;
1899
1900         err = ordered_data_init();
1901         if (err)
1902                 goto free_extent_map;
1903
1904         err = btrfs_delayed_inode_init();
1905         if (err)
1906                 goto free_ordered_data;
1907
1908         err = btrfs_auto_defrag_init();
1909         if (err)
1910                 goto free_delayed_inode;
1911
1912         err = btrfs_delayed_ref_init();
1913         if (err)
1914                 goto free_auto_defrag;
1915
1916         err = btrfs_prelim_ref_init();
1917         if (err)
1918                 goto free_prelim_ref;
1919
1920         err = btrfs_interface_init();
1921         if (err)
1922                 goto free_delayed_ref;
1923
1924         btrfs_init_lockdep();
1925
1926         btrfs_print_info();
1927
1928         err = btrfs_run_sanity_tests();
1929         if (err)
1930                 goto unregister_ioctl;
1931
1932         err = register_filesystem(&btrfs_fs_type);
1933         if (err)
1934                 goto unregister_ioctl;
1935
1936         return 0;
1937
1938 unregister_ioctl:
1939         btrfs_interface_exit();
1940 free_prelim_ref:
1941         btrfs_prelim_ref_exit();
1942 free_delayed_ref:
1943         btrfs_delayed_ref_exit();
1944 free_auto_defrag:
1945         btrfs_auto_defrag_exit();
1946 free_delayed_inode:
1947         btrfs_delayed_inode_exit();
1948 free_ordered_data:
1949         ordered_data_exit();
1950 free_extent_map:
1951         extent_map_exit();
1952 free_extent_io:
1953         extent_io_exit();
1954 free_cachep:
1955         btrfs_destroy_cachep();
1956 free_compress:
1957         btrfs_exit_compress();
1958         btrfs_exit_sysfs();
1959 free_hash:
1960         btrfs_hash_exit();
1961         return err;
1962 }
1963
1964 static void __exit exit_btrfs_fs(void)
1965 {
1966         btrfs_destroy_cachep();
1967         btrfs_delayed_ref_exit();
1968         btrfs_auto_defrag_exit();
1969         btrfs_delayed_inode_exit();
1970         btrfs_prelim_ref_exit();
1971         ordered_data_exit();
1972         extent_map_exit();
1973         extent_io_exit();
1974         btrfs_interface_exit();
1975         unregister_filesystem(&btrfs_fs_type);
1976         btrfs_exit_sysfs();
1977         btrfs_cleanup_fs_uuids();
1978         btrfs_exit_compress();
1979         btrfs_hash_exit();
1980 }
1981
1982 module_init(init_btrfs_fs)
1983 module_exit(exit_btrfs_fs)
1984
1985 MODULE_LICENSE("GPL");