fs/btrfs/super.c

   1 /*
   2  * Copyright (C) 2007 Oracle.  All rights reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public
   6  * License v2 as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public
  14  * License along with this program; if not, write to the
  15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  16  * Boston, MA 021110-1307, USA.
  17  */
  18
  19 #include <linux/blkdev.h>
  20 #include <linux/module.h>
  21 #include <linux/buffer_head.h>
  22 #include <linux/fs.h>
  23 #include <linux/pagemap.h>
  24 #include <linux/highmem.h>
  25 #include <linux/time.h>
  26 #include <linux/init.h>
  27 #include <linux/seq_file.h>
  28 #include <linux/string.h>
  29 #include <linux/backing-dev.h>
  30 #include <linux/mount.h>
  31 #include <linux/mpage.h>
  32 #include <linux/swap.h>
  33 #include <linux/writeback.h>
  34 #include <linux/statfs.h>
  35 #include <linux/compat.h>
  36 #include <linux/parser.h>
  37 #include <linux/ctype.h>
  38 #include <linux/namei.h>
  39 #include <linux/miscdevice.h>
  40 #include <linux/magic.h>
  41 #include <linux/slab.h>
  42 #include <linux/cleancache.h>
  43 #include <linux/ratelimit.h>
  44 #include <linux/btrfs.h>
  45 #include "delayed-inode.h"
  46 #include "ctree.h"
  47 #include "disk-io.h"
  48 #include "transaction.h"
  49 #include "btrfs_inode.h"
  50 #include "print-tree.h"
  51 #include "xattr.h"
  52 #include "volumes.h"
  53 #include "export.h"
  54 #include "compression.h"
  55 #include "rcu-string.h"
  56 #include "dev-replace.h"
  57 #include "free-space-cache.h"
  58 #include "backref.h"
  59 #include "tests/btrfs-tests.h"
  60
  61 #define CREATE_TRACE_POINTS
  62 #include <trace/events/btrfs.h>
  63
  64 static const struct super_operations btrfs_super_ops;
  65 static struct file_system_type btrfs_fs_type;
  66
  67 static const char *btrfs_decode_error(int errno)
  68 {
  69         char *errstr = "unknown";
  70
  71         switch (errno) {
  72         case -EIO:
  73                 errstr = "IO failure";
  74                 break;
  75         case -ENOMEM:
  76                 errstr = "Out of memory";
  77                 break;
  78         case -EROFS:
  79                 errstr = "Readonly filesystem";
  80                 break;
  81         case -EEXIST:
  82                 errstr = "Object already exists";
  83                 break;
  84         case -ENOSPC:
  85                 errstr = "No space left";
  86                 break;
  87         case -ENOENT:
  88                 errstr = "No such entry";
  89                 break;
  90         }
  91
  92         return errstr;
  93 }
  94
  95 static void save_error_info(struct btrfs_fs_info *fs_info)
  96 {
  97         /*
  98          * today we only save the error info into ram.  Long term we'll
  99          * also send it down to the disk
 100          */
 101         set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
 102 }
 103
 104 /* btrfs handle error by forcing the filesystem readonly */
 105 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
 106 {
 107         struct super_block *sb = fs_info->sb;
 108
 109         if (sb->s_flags & MS_RDONLY)
 110                 return;
 111
 112         if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
 113                 sb->s_flags |= MS_RDONLY;
 114                 btrfs_info(fs_info, "forced readonly");
 115                 /*
 116                  * Note that a running device replace operation is not
 117                  * canceled here although there is no way to update
 118                  * the progress. It would add the risk of a deadlock,
 119                  * therefore the canceling is ommited. The only penalty
 120                  * is that some I/O remains active until the procedure
 121                  * completes. The next time when the filesystem is
 122                  * mounted writeable again, the device replace
 123                  * operation continues.
 124                  */
 125         }
 126 }
 127
 128 #ifdef CONFIG_PRINTK
 129 /*
 130  * __btrfs_std_error decodes expected errors from the caller and
 131  * invokes the approciate error response.
 132  */
 133 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 134                        unsigned int line, int errno, const char *fmt, ...)
 135 {
 136         struct super_block *sb = fs_info->sb;
 137         const char *errstr;
 138
 139         /*
 140          * Special case: if the error is EROFS, and we're already
 141          * under MS_RDONLY, then it is safe here.
 142          */
 143         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 144                 return;
 145
 146         errstr = btrfs_decode_error(errno);
 147         if (fmt) {
 148                 struct va_format vaf;
 149                 va_list args;
 150
 151                 va_start(args, fmt);
 152                 vaf.fmt = fmt;
 153                 vaf.va = &args;
 154
 155                 printk(KERN_CRIT
 156                         "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
 157                         sb->s_id, function, line, errno, errstr, &vaf);
 158                 va_end(args);
 159         } else {
 160                 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
 161                         sb->s_id, function, line, errno, errstr);
 162         }
 163
 164         /* Don't go through full error handling during mount */
 165         save_error_info(fs_info);
 166         if (sb->s_flags & MS_BORN)
 167                 btrfs_handle_error(fs_info);
 168 }
 169
 170 static const char * const logtypes[] = {
 171         "emergency",
 172         "alert",
 173         "critical",
 174         "error",
 175         "warning",
 176         "notice",
 177         "info",
 178         "debug",
 179 };
 180
 181 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
 182 {
 183         struct super_block *sb = fs_info->sb;
 184         char lvl[4];
 185         struct va_format vaf;
 186         va_list args;
 187         const char *type = logtypes[4];
 188         int kern_level;
 189
 190         va_start(args, fmt);
 191
 192         kern_level = printk_get_level(fmt);
 193         if (kern_level) {
 194                 size_t size = printk_skip_level(fmt) - fmt;
 195                 memcpy(lvl, fmt,  size);
 196                 lvl[size] = '\0';
 197                 fmt += size;
 198                 type = logtypes[kern_level - '0'];
 199         } else
 200                 *lvl = '\0';
 201
 202         vaf.fmt = fmt;
 203         vaf.va = &args;
 204
 205         printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
 206
 207         va_end(args);
 208 }
 209
 210 #else
 211
 212 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 213                        unsigned int line, int errno, const char *fmt, ...)
 214 {
 215         struct super_block *sb = fs_info->sb;
 216
 217         /*
 218          * Special case: if the error is EROFS, and we're already
 219          * under MS_RDONLY, then it is safe here.
 220          */
 221         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 222                 return;
 223
 224         /* Don't go through full error handling during mount */
 225         if (sb->s_flags & MS_BORN) {
 226                 save_error_info(fs_info);
 227                 btrfs_handle_error(fs_info);
 228         }
 229 }
 230 #endif
 231
 232 /*
 233  * We only mark the transaction aborted and then set the file system read-only.
 234  * This will prevent new transactions from starting or trying to join this
 235  * one.
 236  *
 237  * This means that error recovery at the call site is limited to freeing
 238  * any local memory allocations and passing the error code up without
 239  * further cleanup. The transaction should complete as it normally would
 240  * in the call path but will return -EIO.
 241  *
 242  * We'll complete the cleanup in btrfs_end_transaction and
 243  * btrfs_commit_transaction.
 244  */
 245 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
 246                                struct btrfs_root *root, const char *function,
 247                                unsigned int line, int errno)
 248 {
 249         /*
 250          * Report first abort since mount
 251          */
 252         if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,
 253                                 &root->fs_info->fs_state)) {
 254                 WARN(1, KERN_DEBUG "BTRFS: Transaction aborted (error %d)\n",
 255                                 errno);
 256         }
 257         trans->aborted = errno;
 258         /* Nothing used. The other threads that have joined this
 259          * transaction may be able to continue. */
 260         if (!trans->blocks_used) {
 261                 const char *errstr;
 262
 263                 errstr = btrfs_decode_error(errno);
 264                 btrfs_warn(root->fs_info,
 265                            "%s:%d: Aborting unused transaction(%s).",
 266                            function, line, errstr);
 267                 return;
 268         }
 269         ACCESS_ONCE(trans->transaction->aborted) = errno;
 270         /* Wake up anybody who may be waiting on this transaction */
 271         wake_up(&root->fs_info->transaction_wait);
 272         wake_up(&root->fs_info->transaction_blocked_wait);
 273         __btrfs_std_error(root->fs_info, function, line, errno, NULL);
 274 }
 275 /*
 276  * __btrfs_panic decodes unexpected, fatal errors from the caller,
 277  * issues an alert, and either panics or BUGs, depending on mount options.
 278  */
 279 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
 280                    unsigned int line, int errno, const char *fmt, ...)
 281 {
 282         char *s_id = "<unknown>";
 283         const char *errstr;
 284         struct va_format vaf = { .fmt = fmt };
 285         va_list args;
 286
 287         if (fs_info)
 288                 s_id = fs_info->sb->s_id;
 289
 290         va_start(args, fmt);
 291         vaf.va = &args;
 292
 293         errstr = btrfs_decode_error(errno);
 294         if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
 295                 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
 296                         s_id, function, line, &vaf, errno, errstr);
 297
 298         btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
 299                    function, line, &vaf, errno, errstr);
 300         va_end(args);
 301         /* Caller calls BUG() */
 302 }
 303
 304 static void btrfs_put_super(struct super_block *sb)
 305 {
 306         (void)close_ctree(btrfs_sb(sb)->tree_root);
 307         /* FIXME: need to fix VFS to return error? */
 308         /* AV: return it _where_?  ->put_super() can be triggered by any number
 309          * of async events, up to and including delivery of SIGKILL to the
 310          * last process that kept it busy.  Or segfault in the aforementioned
 311          * process...  Whom would you report that to?
 312          */
 313 }
 314
 315 enum {
 316         Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
 317         Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
 318         Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
 319         Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
 320         Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
 321         Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
 322         Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
 323         Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
 324         Opt_check_integrity, Opt_check_integrity_including_extent_data,
 325         Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
 326         Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
 327         Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
 328         Opt_datasum, Opt_treelog,
 329         Opt_err,
 330 };
 331
 332 static match_table_t tokens = {
 333         {Opt_degraded, "degraded"},
 334         {Opt_subvol, "subvol=%s"},
 335         {Opt_subvolid, "subvolid=%s"},
 336         {Opt_device, "device=%s"},
 337         {Opt_nodatasum, "nodatasum"},
 338         {Opt_datasum, "datasum"},
 339         {Opt_nodatacow, "nodatacow"},
 340         {Opt_datacow, "datacow"},
 341         {Opt_nobarrier, "nobarrier"},
 342         {Opt_barrier, "barrier"},
 343         {Opt_max_inline, "max_inline=%s"},
 344         {Opt_alloc_start, "alloc_start=%s"},
 345         {Opt_thread_pool, "thread_pool=%d"},
 346         {Opt_compress, "compress"},
 347         {Opt_compress_type, "compress=%s"},
 348         {Opt_compress_force, "compress-force"},
 349         {Opt_compress_force_type, "compress-force=%s"},
 350         {Opt_ssd, "ssd"},
 351         {Opt_ssd_spread, "ssd_spread"},
 352         {Opt_nossd, "nossd"},
 353         {Opt_acl, "acl"},
 354         {Opt_noacl, "noacl"},
 355         {Opt_notreelog, "notreelog"},
 356         {Opt_treelog, "treelog"},
 357         {Opt_flushoncommit, "flushoncommit"},
 358         {Opt_noflushoncommit, "noflushoncommit"},
 359         {Opt_ratio, "metadata_ratio=%d"},
 360         {Opt_discard, "discard"},
 361         {Opt_nodiscard, "nodiscard"},
 362         {Opt_space_cache, "space_cache"},
 363         {Opt_clear_cache, "clear_cache"},
 364         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 365         {Opt_enospc_debug, "enospc_debug"},
 366         {Opt_noenospc_debug, "noenospc_debug"},
 367         {Opt_subvolrootid, "subvolrootid=%d"},
 368         {Opt_defrag, "autodefrag"},
 369         {Opt_nodefrag, "noautodefrag"},
 370         {Opt_inode_cache, "inode_cache"},
 371         {Opt_no_space_cache, "nospace_cache"},
 372         {Opt_recovery, "recovery"},
 373         {Opt_skip_balance, "skip_balance"},
 374         {Opt_check_integrity, "check_int"},
 375         {Opt_check_integrity_including_extent_data, "check_int_data"},
 376         {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
 377         {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 378         {Opt_fatal_errors, "fatal_errors=%s"},
 379         {Opt_commit_interval, "commit=%d"},
 380         {Opt_err, NULL},
 381 };
 382
 383 /*
 384  * Regular mount options parser.  Everything that is needed only when
 385  * reading in a new superblock is parsed here.
 386  * XXX JDM: This needs to be cleaned up for remount.
 387  */
 388 int btrfs_parse_options(struct btrfs_root *root, char *options)
 389 {
 390         struct btrfs_fs_info *info = root->fs_info;
 391         substring_t args[MAX_OPT_ARGS];
 392         char *p, *num, *orig = NULL;
 393         u64 cache_gen;
 394         int intarg;
 395         int ret = 0;
 396         char *compress_type;
 397         bool compress_force = false;
 398
 399         cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
 400         if (cache_gen)
 401                 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 402
 403         if (!options)
 404                 goto out;
 405
 406         /*
 407          * strsep changes the string, duplicate it because parse_options
 408          * gets called twice
 409          */
 410         options = kstrdup(options, GFP_NOFS);
 411         if (!options)
 412                 return -ENOMEM;
 413
 414         orig = options;
 415
 416         while ((p = strsep(&options, ",")) != NULL) {
 417                 int token;
 418                 if (!*p)
 419                         continue;
 420
 421                 token = match_token(p, tokens, args);
 422                 switch (token) {
 423                 case Opt_degraded:
 424                         btrfs_info(root->fs_info, "allowing degraded mounts");
 425                         btrfs_set_opt(info->mount_opt, DEGRADED);
 426                         break;
 427                 case Opt_subvol:
 428                 case Opt_subvolid:
 429                 case Opt_subvolrootid:
 430                 case Opt_device:
 431                         /*
 432                          * These are parsed by btrfs_parse_early_options
 433                          * and can be happily ignored here.
 434                          */
 435                         break;
 436                 case Opt_nodatasum:
 437                         btrfs_info(root->fs_info, "setting nodatasum");
 438                         btrfs_set_opt(info->mount_opt, NODATASUM);
 439                         break;
 440                 case Opt_datasum:
 441                         if (btrfs_test_opt(root, NODATACOW))
 442                                 btrfs_info(root->fs_info, "setting datasum, datacow enabled");
 443                         else
 444                                 btrfs_info(root->fs_info, "setting datasum");
 445                         btrfs_clear_opt(info->mount_opt, NODATACOW);
 446                         btrfs_clear_opt(info->mount_opt, NODATASUM);
 447                         break;
 448                 case Opt_nodatacow:
 449                         if (!btrfs_test_opt(root, COMPRESS) ||
 450                                 !btrfs_test_opt(root, FORCE_COMPRESS)) {
 451                                         btrfs_info(root->fs_info,
 452                                                 "setting nodatacow, compression disabled");
 453                         } else {
 454                                 btrfs_info(root->fs_info, "setting nodatacow");
 455                         }
 456                         btrfs_clear_opt(info->mount_opt, COMPRESS);
 457                         btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 458                         btrfs_set_opt(info->mount_opt, NODATACOW);
 459                         btrfs_set_opt(info->mount_opt, NODATASUM);
 460                         break;
 461                 case Opt_datacow:
 462                         if (btrfs_test_opt(root, NODATACOW))
 463                                 btrfs_info(root->fs_info, "setting datacow");
 464                         btrfs_clear_opt(info->mount_opt, NODATACOW);
 465                         break;
 466                 case Opt_compress_force:
 467                 case Opt_compress_force_type:
 468                         compress_force = true;
 469                         /* Fallthrough */
 470                 case Opt_compress:
 471                 case Opt_compress_type:
 472                         if (token == Opt_compress ||
 473                             token == Opt_compress_force ||
 474                             strcmp(args[0].from, "zlib") == 0) {
 475                                 compress_type = "zlib";
 476                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 477                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 478                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 479                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 480                         } else if (strcmp(args[0].from, "lzo") == 0) {
 481                                 compress_type = "lzo";
 482                                 info->compress_type = BTRFS_COMPRESS_LZO;
 483                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 484                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 485                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 486                                 btrfs_set_fs_incompat(info, COMPRESS_LZO);
 487                         } else if (strncmp(args[0].from, "no", 2) == 0) {
 488                                 compress_type = "no";
 489                                 btrfs_clear_opt(info->mount_opt, COMPRESS);
 490                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 491                                 compress_force = false;
 492                         } else {
 493                                 ret = -EINVAL;
 494                                 goto out;
 495                         }
 496
 497                         if (compress_force) {
 498                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 499                                 btrfs_info(root->fs_info, "force %s compression",
 500                                         compress_type);
 501                         } else if (btrfs_test_opt(root, COMPRESS)) {
 502                                 pr_info("btrfs: use %s compression\n",
 503                                         compress_type);
 504                         }
 505                         break;
 506                 case Opt_ssd:
 507                         btrfs_info(root->fs_info, "use ssd allocation scheme");
 508                         btrfs_set_opt(info->mount_opt, SSD);
 509                         break;
 510                 case Opt_ssd_spread:
 511                         btrfs_info(root->fs_info, "use spread ssd allocation scheme");
 512                         btrfs_set_opt(info->mount_opt, SSD);
 513                         btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 514                         break;
 515                 case Opt_nossd:
 516                         btrfs_info(root->fs_info, "not using ssd allocation scheme");
 517                         btrfs_set_opt(info->mount_opt, NOSSD);
 518                         btrfs_clear_opt(info->mount_opt, SSD);
 519                         btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 520                         break;
 521                 case Opt_barrier:
 522                         if (btrfs_test_opt(root, NOBARRIER))
 523                                 btrfs_info(root->fs_info, "turning on barriers");
 524                         btrfs_clear_opt(info->mount_opt, NOBARRIER);
 525                         break;
 526                 case Opt_nobarrier:
 527                         btrfs_info(root->fs_info, "turning off barriers");
 528                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 529                         break;
 530                 case Opt_thread_pool:
 531                         ret = match_int(&args[0], &intarg);
 532                         if (ret) {
 533                                 goto out;
 534                         } else if (intarg > 0) {
 535                                 info->thread_pool_size = intarg;
 536                         } else {
 537                                 ret = -EINVAL;
 538                                 goto out;
 539                         }
 540                         break;
 541                 case Opt_max_inline:
 542                         num = match_strdup(&args[0]);
 543                         if (num) {
 544                                 info->max_inline = memparse(num, NULL);
 545                                 kfree(num);
 546
 547                                 if (info->max_inline) {
 548                                         info->max_inline = max_t(u64,
 549                                                 info->max_inline,
 550                                                 root->sectorsize);
 551                                 }
 552                                 btrfs_info(root->fs_info, "max_inline at %llu",
 553                                         info->max_inline);
 554                         } else {
 555                                 ret = -ENOMEM;
 556                                 goto out;
 557                         }
 558                         break;
 559                 case Opt_alloc_start:
 560                         num = match_strdup(&args[0]);
 561                         if (num) {
 562                                 mutex_lock(&info->chunk_mutex);
 563                                 info->alloc_start = memparse(num, NULL);
 564                                 mutex_unlock(&info->chunk_mutex);
 565                                 kfree(num);
 566                                 btrfs_info(root->fs_info, "allocations start at %llu",
 567                                         info->alloc_start);
 568                         } else {
 569                                 ret = -ENOMEM;
 570                                 goto out;
 571                         }
 572                         break;
 573                 case Opt_acl:
 574                         root->fs_info->sb->s_flags |= MS_POSIXACL;
 575                         break;
 576                 case Opt_noacl:
 577                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 578                         break;
 579                 case Opt_notreelog:
 580                         btrfs_info(root->fs_info, "disabling tree log");
 581                         btrfs_set_opt(info->mount_opt, NOTREELOG);
 582                         break;
 583                 case Opt_treelog:
 584                         if (btrfs_test_opt(root, NOTREELOG))
 585                                 btrfs_info(root->fs_info, "enabling tree log");
 586                         btrfs_clear_opt(info->mount_opt, NOTREELOG);
 587                         break;
 588                 case Opt_flushoncommit:
 589                         btrfs_info(root->fs_info, "turning on flush-on-commit");
 590                         btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
 591                         break;
 592                 case Opt_noflushoncommit:
 593                         if (btrfs_test_opt(root, FLUSHONCOMMIT))
 594                                 btrfs_info(root->fs_info, "turning off flush-on-commit");
 595                         btrfs_clear_opt(info->mount_opt, FLUSHONCOMMIT);
 596                         break;
 597                 case Opt_ratio:
 598                         ret = match_int(&args[0], &intarg);
 599                         if (ret) {
 600                                 goto out;
 601                         } else if (intarg >= 0) {
 602                                 info->metadata_ratio = intarg;
 603                                 btrfs_info(root->fs_info, "metadata ratio %d",
 604                                        info->metadata_ratio);
 605                         } else {
 606                                 ret = -EINVAL;
 607                                 goto out;
 608                         }
 609                         break;
 610                 case Opt_discard:
 611                         btrfs_set_opt(info->mount_opt, DISCARD);
 612                         break;
 613                 case Opt_nodiscard:
 614                         btrfs_clear_opt(info->mount_opt, DISCARD);
 615                         break;
 616                 case Opt_space_cache:
 617                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 618                         break;
 619                 case Opt_rescan_uuid_tree:
 620                         btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 621                         break;
 622                 case Opt_no_space_cache:
 623                         btrfs_info(root->fs_info, "disabling disk space caching");
 624                         btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
 625                         break;
 626                 case Opt_inode_cache:
 627                         btrfs_info(root->fs_info, "enabling inode map caching");
 628                         btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
 629                         break;
 630                 case Opt_clear_cache:
 631                         btrfs_info(root->fs_info, "force clearing of disk cache");
 632                         btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
 633                         break;
 634                 case Opt_user_subvol_rm_allowed:
 635                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 636                         break;
 637                 case Opt_enospc_debug:
 638                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 639                         break;
 640                 case Opt_noenospc_debug:
 641                         btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
 642                         break;
 643                 case Opt_defrag:
 644                         btrfs_info(root->fs_info, "enabling auto defrag");
 645                         btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
 646                         break;
 647                 case Opt_nodefrag:
 648                         if (btrfs_test_opt(root, AUTO_DEFRAG))
 649                                 btrfs_info(root->fs_info, "disabling auto defrag");
 650                         btrfs_clear_opt(info->mount_opt, AUTO_DEFRAG);
 651                         break;
 652                 case Opt_recovery:
 653                         btrfs_info(root->fs_info, "enabling auto recovery");
 654                         btrfs_set_opt(info->mount_opt, RECOVERY);
 655                         break;
 656                 case Opt_skip_balance:
 657                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 658                         break;
 659 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 660                 case Opt_check_integrity_including_extent_data:
 661                         btrfs_info(root->fs_info,
 662                                    "enabling check integrity including extent data");
 663                         btrfs_set_opt(info->mount_opt,
 664                                       CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
 665                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 666                         break;
 667                 case Opt_check_integrity:
 668                         btrfs_info(root->fs_info, "enabling check integrity");
 669                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 670                         break;
 671                 case Opt_check_integrity_print_mask:
 672                         ret = match_int(&args[0], &intarg);
 673                         if (ret) {
 674                                 goto out;
 675                         } else if (intarg >= 0) {
 676                                 info->check_integrity_print_mask = intarg;
 677                                 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
 678                                        info->check_integrity_print_mask);
 679                         } else {
 680                                 ret = -EINVAL;
 681                                 goto out;
 682                         }
 683                         break;
 684 #else
 685                 case Opt_check_integrity_including_extent_data:
 686                 case Opt_check_integrity:
 687                 case Opt_check_integrity_print_mask:
 688                         btrfs_err(root->fs_info,
 689                                 "support for check_integrity* not compiled in!");
 690                         ret = -EINVAL;
 691                         goto out;
 692 #endif
 693                 case Opt_fatal_errors:
 694                         if (strcmp(args[0].from, "panic") == 0)
 695                                 btrfs_set_opt(info->mount_opt,
 696                                               PANIC_ON_FATAL_ERROR);
 697                         else if (strcmp(args[0].from, "bug") == 0)
 698                                 btrfs_clear_opt(info->mount_opt,
 699                                               PANIC_ON_FATAL_ERROR);
 700                         else {
 701                                 ret = -EINVAL;
 702                                 goto out;
 703                         }
 704                         break;
 705                 case Opt_commit_interval:
 706                         intarg = 0;
 707                         ret = match_int(&args[0], &intarg);
 708                         if (ret < 0) {
 709                                 btrfs_err(root->fs_info, "invalid commit interval");
 710                                 ret = -EINVAL;
 711                                 goto out;
 712                         }
 713                         if (intarg > 0) {
 714                                 if (intarg > 300) {
 715                                         btrfs_warn(root->fs_info, "excessive commit interval %d",
 716                                                         intarg);
 717                                 }
 718                                 info->commit_interval = intarg;
 719                         } else {
 720                                 btrfs_info(root->fs_info, "using default commit interval %ds",
 721                                     BTRFS_DEFAULT_COMMIT_INTERVAL);
 722                                 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
 723                         }
 724                         break;
 725                 case Opt_err:
 726                         btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
 727                         ret = -EINVAL;
 728                         goto out;
 729                 default:
 730                         break;
 731                 }
 732         }
 733 out:
 734         if (!ret && btrfs_test_opt(root, SPACE_CACHE))
 735                 btrfs_info(root->fs_info, "disk space caching is enabled");
 736         kfree(orig);
 737         return ret;
 738 }
 739
 740 /*
 741  * Parse mount options that are required early in the mount process.
 742  *
 743  * All other options will be parsed on much later in the mount process and
 744  * only when we need to allocate a new super block.
 745  */
 746 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 747                 void *holder, char **subvol_name, u64 *subvol_objectid,
 748                 struct btrfs_fs_devices **fs_devices)
 749 {
 750         substring_t args[MAX_OPT_ARGS];
 751         char *device_name, *opts, *orig, *p;
 752         char *num = NULL;
 753         int error = 0;
 754
 755         if (!options)
 756                 return 0;
 757
 758         /*
 759          * strsep changes the string, duplicate it because parse_options
 760          * gets called twice
 761          */
 762         opts = kstrdup(options, GFP_KERNEL);
 763         if (!opts)
 764                 return -ENOMEM;
 765         orig = opts;
 766
 767         while ((p = strsep(&opts, ",")) != NULL) {
 768                 int token;
 769                 if (!*p)
 770                         continue;
 771
 772                 token = match_token(p, tokens, args);
 773                 switch (token) {
 774                 case Opt_subvol:
 775                         kfree(*subvol_name);
 776                         *subvol_name = match_strdup(&args[0]);
 777                         if (!*subvol_name) {
 778                                 error = -ENOMEM;
 779                                 goto out;
 780                         }
 781                         break;
 782                 case Opt_subvolid:
 783                         num = match_strdup(&args[0]);
 784                         if (num) {
 785                                 *subvol_objectid = memparse(num, NULL);
 786                                 kfree(num);
 787                                 /* we want the original fs_tree */
 788                                 if (!*subvol_objectid)
 789                                         *subvol_objectid =
 790                                                 BTRFS_FS_TREE_OBJECTID;
 791                         } else {
 792                                 error = -EINVAL;
 793                                 goto out;
 794                         }
 795                         break;
 796                 case Opt_subvolrootid:
 797                         printk(KERN_WARNING
 798                                 "BTRFS: 'subvolrootid' mount option is deprecated and has "
 799                                 "no effect\n");
 800                         break;
 801                 case Opt_device:
 802                         device_name = match_strdup(&args[0]);
 803                         if (!device_name) {
 804                                 error = -ENOMEM;
 805                                 goto out;
 806                         }
 807                         error = btrfs_scan_one_device(device_name,
 808                                         flags, holder, fs_devices);
 809                         kfree(device_name);
 810                         if (error)
 811                                 goto out;
 812                         break;
 813                 default:
 814                         break;
 815                 }
 816         }
 817
 818 out:
 819         kfree(orig);
 820         return error;
 821 }
 822
 823 static struct dentry *get_default_root(struct super_block *sb,
 824                                        u64 subvol_objectid)
 825 {
 826         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 827         struct btrfs_root *root = fs_info->tree_root;
 828         struct btrfs_root *new_root;
 829         struct btrfs_dir_item *di;
 830         struct btrfs_path *path;
 831         struct btrfs_key location;
 832         struct inode *inode;
 833         u64 dir_id;
 834         int new = 0;
 835
 836         /*
 837          * We have a specific subvol we want to mount, just setup location and
 838          * go look up the root.
 839          */
 840         if (subvol_objectid) {
 841                 location.objectid = subvol_objectid;
 842                 location.type = BTRFS_ROOT_ITEM_KEY;
 843                 location.offset = (u64)-1;
 844                 goto find_root;
 845         }
 846
 847         path = btrfs_alloc_path();
 848         if (!path)
 849                 return ERR_PTR(-ENOMEM);
 850         path->leave_spinning = 1;
 851
 852         /*
 853          * Find the "default" dir item which points to the root item that we
 854          * will mount by default if we haven't been given a specific subvolume
 855          * to mount.
 856          */
 857         dir_id = btrfs_super_root_dir(fs_info->super_copy);
 858         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 859         if (IS_ERR(di)) {
 860                 btrfs_free_path(path);
 861                 return ERR_CAST(di);
 862         }
 863         if (!di) {
 864                 /*
 865                  * Ok the default dir item isn't there.  This is weird since
 866                  * it's always been there, but don't freak out, just try and
 867                  * mount to root most subvolume.
 868                  */
 869                 btrfs_free_path(path);
 870                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 871                 new_root = fs_info->fs_root;
 872                 goto setup_root;
 873         }
 874
 875         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 876         btrfs_free_path(path);
 877
 878 find_root:
 879         new_root = btrfs_read_fs_root_no_name(fs_info, &location);
 880         if (IS_ERR(new_root))
 881                 return ERR_CAST(new_root);
 882
 883         dir_id = btrfs_root_dirid(&new_root->root_item);
 884 setup_root:
 885         location.objectid = dir_id;
 886         location.type = BTRFS_INODE_ITEM_KEY;
 887         location.offset = 0;
 888
 889         inode = btrfs_iget(sb, &location, new_root, &new);
 890         if (IS_ERR(inode))
 891                 return ERR_CAST(inode);
 892
 893         /*
 894          * If we're just mounting the root most subvol put the inode and return
 895          * a reference to the dentry.  We will have already gotten a reference
 896          * to the inode in btrfs_fill_super so we're good to go.
 897          */
 898         if (!new && sb->s_root->d_inode == inode) {
 899                 iput(inode);
 900                 return dget(sb->s_root);
 901         }
 902
 903         return d_obtain_alias(inode);
 904 }
 905
 906 static int btrfs_fill_super(struct super_block *sb,
 907                             struct btrfs_fs_devices *fs_devices,
 908                             void *data, int silent)
 909 {
 910         struct inode *inode;
 911         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 912         struct btrfs_key key;
 913         int err;
 914
 915         sb->s_maxbytes = MAX_LFS_FILESIZE;
 916         sb->s_magic = BTRFS_SUPER_MAGIC;
 917         sb->s_op = &btrfs_super_ops;
 918         sb->s_d_op = &btrfs_dentry_operations;
 919         sb->s_export_op = &btrfs_export_ops;
 920         sb->s_xattr = btrfs_xattr_handlers;
 921         sb->s_time_gran = 1;
 922 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 923         sb->s_flags |= MS_POSIXACL;
 924 #endif
 925         sb->s_flags |= MS_I_VERSION;
 926         err = open_ctree(sb, fs_devices, (char *)data);
 927         if (err) {
 928                 printk(KERN_ERR "BTRFS: open_ctree failed\n");
 929                 return err;
 930         }
 931
 932         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 933         key.type = BTRFS_INODE_ITEM_KEY;
 934         key.offset = 0;
 935         inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
 936         if (IS_ERR(inode)) {
 937                 err = PTR_ERR(inode);
 938                 goto fail_close;
 939         }
 940
 941         sb->s_root = d_make_root(inode);
 942         if (!sb->s_root) {
 943                 err = -ENOMEM;
 944                 goto fail_close;
 945         }
 946
 947         save_mount_options(sb, data);
 948         cleancache_init_fs(sb);
 949         sb->s_flags |= MS_ACTIVE;
 950         return 0;
 951
 952 fail_close:
 953         close_ctree(fs_info->tree_root);
 954         return err;
 955 }
 956
 957 int btrfs_sync_fs(struct super_block *sb, int wait)
 958 {
 959         struct btrfs_trans_handle *trans;
 960         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 961         struct btrfs_root *root = fs_info->tree_root;
 962
 963         trace_btrfs_sync_fs(wait);
 964
 965         if (!wait) {
 966                 filemap_flush(fs_info->btree_inode->i_mapping);
 967                 return 0;
 968         }
 969
 970         btrfs_wait_ordered_roots(fs_info, -1);
 971
 972         trans = btrfs_attach_transaction_barrier(root);
 973         if (IS_ERR(trans)) {
 974                 /* no transaction, don't bother */
 975                 if (PTR_ERR(trans) == -ENOENT)
 976                         return 0;
 977                 return PTR_ERR(trans);
 978         }
 979         return btrfs_commit_transaction(trans, root);
 980 }
 981
 982 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
 983 {
 984         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
 985         struct btrfs_root *root = info->tree_root;
 986         char *compress_type;
 987
 988         if (btrfs_test_opt(root, DEGRADED))
 989                 seq_puts(seq, ",degraded");
 990         if (btrfs_test_opt(root, NODATASUM))
 991                 seq_puts(seq, ",nodatasum");
 992         if (btrfs_test_opt(root, NODATACOW))
 993                 seq_puts(seq, ",nodatacow");
 994         if (btrfs_test_opt(root, NOBARRIER))
 995                 seq_puts(seq, ",nobarrier");
 996         if (info->max_inline != 8192 * 1024)
 997                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
 998         if (info->alloc_start != 0)
 999                 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1000         if (info->thread_pool_size !=  min_t(unsigned long,
1001                                              num_online_cpus() + 2, 8))
1002                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1003         if (btrfs_test_opt(root, COMPRESS)) {
1004                 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1005                         compress_type = "zlib";
1006                 else
1007                         compress_type = "lzo";
1008                 if (btrfs_test_opt(root, FORCE_COMPRESS))
1009                         seq_printf(seq, ",compress-force=%s", compress_type);
1010                 else
1011                         seq_printf(seq, ",compress=%s", compress_type);
1012         }
1013         if (btrfs_test_opt(root, NOSSD))
1014                 seq_puts(seq, ",nossd");
1015         if (btrfs_test_opt(root, SSD_SPREAD))
1016                 seq_puts(seq, ",ssd_spread");
1017         else if (btrfs_test_opt(root, SSD))
1018                 seq_puts(seq, ",ssd");
1019         if (btrfs_test_opt(root, NOTREELOG))
1020                 seq_puts(seq, ",notreelog");
1021         if (btrfs_test_opt(root, FLUSHONCOMMIT))
1022                 seq_puts(seq, ",flushoncommit");
1023         if (btrfs_test_opt(root, DISCARD))
1024                 seq_puts(seq, ",discard");
1025         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1026                 seq_puts(seq, ",noacl");
1027         if (btrfs_test_opt(root, SPACE_CACHE))
1028                 seq_puts(seq, ",space_cache");
1029         else
1030                 seq_puts(seq, ",nospace_cache");
1031         if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1032                 seq_puts(seq, ",rescan_uuid_tree");
1033         if (btrfs_test_opt(root, CLEAR_CACHE))
1034                 seq_puts(seq, ",clear_cache");
1035         if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1036                 seq_puts(seq, ",user_subvol_rm_allowed");
1037         if (btrfs_test_opt(root, ENOSPC_DEBUG))
1038                 seq_puts(seq, ",enospc_debug");
1039         if (btrfs_test_opt(root, AUTO_DEFRAG))
1040                 seq_puts(seq, ",autodefrag");
1041         if (btrfs_test_opt(root, INODE_MAP_CACHE))
1042                 seq_puts(seq, ",inode_cache");
1043         if (btrfs_test_opt(root, SKIP_BALANCE))
1044                 seq_puts(seq, ",skip_balance");
1045         if (btrfs_test_opt(root, RECOVERY))
1046                 seq_puts(seq, ",recovery");
1047 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1048         if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1049                 seq_puts(seq, ",check_int_data");
1050         else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1051                 seq_puts(seq, ",check_int");
1052         if (info->check_integrity_print_mask)
1053                 seq_printf(seq, ",check_int_print_mask=%d",
1054                                 info->check_integrity_print_mask);
1055 #endif
1056         if (info->metadata_ratio)
1057                 seq_printf(seq, ",metadata_ratio=%d",
1058                                 info->metadata_ratio);
1059         if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1060                 seq_puts(seq, ",fatal_errors=panic");
1061         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1062                 seq_printf(seq, ",commit=%d", info->commit_interval);
1063         return 0;
1064 }
1065
1066 static int btrfs_test_super(struct super_block *s, void *data)
1067 {
1068         struct btrfs_fs_info *p = data;
1069         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1070
1071         return fs_info->fs_devices == p->fs_devices;
1072 }
1073
1074 static int btrfs_set_super(struct super_block *s, void *data)
1075 {
1076         int err = set_anon_super(s, data);
1077         if (!err)
1078                 s->s_fs_info = data;
1079         return err;
1080 }
1081
1082 /*
1083  * subvolumes are identified by ino 256
1084  */
1085 static inline int is_subvolume_inode(struct inode *inode)
1086 {
1087         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1088                 return 1;
1089         return 0;
1090 }
1091
1092 /*
1093  * This will strip out the subvol=%s argument for an argument string and add
1094  * subvolid=0 to make sure we get the actual tree root for path walking to the
1095  * subvol we want.
1096  */
1097 static char *setup_root_args(char *args)
1098 {
1099         unsigned len = strlen(args) + 2 + 1;
1100         char *src, *dst, *buf;
1101
1102         /*
1103          * We need the same args as before, but with this substitution:
1104          * s!subvol=[^,]+!subvolid=0!
1105          *
1106          * Since the replacement string is up to 2 bytes longer than the
1107          * original, allocate strlen(args) + 2 + 1 bytes.
1108          */
1109
1110         src = strstr(args, "subvol=");
1111         /* This shouldn't happen, but just in case.. */
1112         if (!src)
1113                 return NULL;
1114
1115         buf = dst = kmalloc(len, GFP_NOFS);
1116         if (!buf)
1117                 return NULL;
1118
1119         /*
1120          * If the subvol= arg is not at the start of the string,
1121          * copy whatever precedes it into buf.
1122          */
1123         if (src != args) {
1124                 *src++ = '\0';
1125                 strcpy(buf, args);
1126                 dst += strlen(args);
1127         }
1128
1129         strcpy(dst, "subvolid=0");
1130         dst += strlen("subvolid=0");
1131
1132         /*
1133          * If there is a "," after the original subvol=... string,
1134          * copy that suffix into our buffer.  Otherwise, we're done.
1135          */
1136         src = strchr(src, ',');
1137         if (src)
1138                 strcpy(dst, src);
1139
1140         return buf;
1141 }
1142
1143 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1144                                    const char *device_name, char *data)
1145 {
1146         struct dentry *root;
1147         struct vfsmount *mnt;
1148         char *newargs;
1149
1150         newargs = setup_root_args(data);
1151         if (!newargs)
1152                 return ERR_PTR(-ENOMEM);
1153         mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1154                              newargs);
1155         kfree(newargs);
1156         if (IS_ERR(mnt))
1157                 return ERR_CAST(mnt);
1158
1159         root = mount_subtree(mnt, subvol_name);
1160
1161         if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1162                 struct super_block *s = root->d_sb;
1163                 dput(root);
1164                 root = ERR_PTR(-EINVAL);
1165                 deactivate_locked_super(s);
1166                 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1167                                 subvol_name);
1168         }
1169
1170         return root;
1171 }
1172
1173 /*
1174  * Find a superblock for the given device / mount point.
1175  *
1176  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
1177  *        for multiple device setup.  Make sure to keep it in sync.
1178  */
1179 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1180                 const char *device_name, void *data)
1181 {
1182         struct block_device *bdev = NULL;
1183         struct super_block *s;
1184         struct dentry *root;
1185         struct btrfs_fs_devices *fs_devices = NULL;
1186         struct btrfs_fs_info *fs_info = NULL;
1187         fmode_t mode = FMODE_READ;
1188         char *subvol_name = NULL;
1189         u64 subvol_objectid = 0;
1190         int error = 0;
1191
1192         if (!(flags & MS_RDONLY))
1193                 mode |= FMODE_WRITE;
1194
1195         error = btrfs_parse_early_options(data, mode, fs_type,
1196                                           &subvol_name, &subvol_objectid,
1197                                           &fs_devices);
1198         if (error) {
1199                 kfree(subvol_name);
1200                 return ERR_PTR(error);
1201         }
1202
1203         if (subvol_name) {
1204                 root = mount_subvol(subvol_name, flags, device_name, data);
1205                 kfree(subvol_name);
1206                 return root;
1207         }
1208
1209         error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1210         if (error)
1211                 return ERR_PTR(error);
1212
1213         /*
1214          * Setup a dummy root and fs_info for test/set super.  This is because
1215          * we don't actually fill this stuff out until open_ctree, but we need
1216          * it for searching for existing supers, so this lets us do that and
1217          * then open_ctree will properly initialize everything later.
1218          */
1219         fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1220         if (!fs_info)
1221                 return ERR_PTR(-ENOMEM);
1222
1223         fs_info->fs_devices = fs_devices;
1224
1225         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1226         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1227         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1228                 error = -ENOMEM;
1229                 goto error_fs_info;
1230         }
1231
1232         error = btrfs_open_devices(fs_devices, mode, fs_type);
1233         if (error)
1234                 goto error_fs_info;
1235
1236         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1237                 error = -EACCES;
1238                 goto error_close_devices;
1239         }
1240
1241         bdev = fs_devices->latest_bdev;
1242         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1243                  fs_info);
1244         if (IS_ERR(s)) {
1245                 error = PTR_ERR(s);
1246                 goto error_close_devices;
1247         }
1248
1249         if (s->s_root) {
1250                 btrfs_close_devices(fs_devices);
1251                 free_fs_info(fs_info);
1252                 if ((flags ^ s->s_flags) & MS_RDONLY)
1253                         error = -EBUSY;
1254         } else {
1255                 char b[BDEVNAME_SIZE];
1256
1257                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1258                 btrfs_sb(s)->bdev_holder = fs_type;
1259                 error = btrfs_fill_super(s, fs_devices, data,
1260                                          flags & MS_SILENT ? 1 : 0);
1261         }
1262
1263         root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1264         if (IS_ERR(root))
1265                 deactivate_locked_super(s);
1266
1267         return root;
1268
1269 error_close_devices:
1270         btrfs_close_devices(fs_devices);
1271 error_fs_info:
1272         free_fs_info(fs_info);
1273         return ERR_PTR(error);
1274 }
1275
1276 static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1277 {
1278         spin_lock_irq(&workers->lock);
1279         workers->max_workers = new_limit;
1280         spin_unlock_irq(&workers->lock);
1281 }
1282
1283 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1284                                      int new_pool_size, int old_pool_size)
1285 {
1286         if (new_pool_size == old_pool_size)
1287                 return;
1288
1289         fs_info->thread_pool_size = new_pool_size;
1290
1291         btrfs_info(fs_info, "resize thread pool %d -> %d",
1292                old_pool_size, new_pool_size);
1293
1294         btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1295         btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1296         btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1297         btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1298         btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1299         btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1300         btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1301         btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1302         btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1303         btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1304         btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1305         btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1306         btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
1307         btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers,
1308                               new_pool_size);
1309 }
1310
1311 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1312 {
1313         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1314 }
1315
1316 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1317                                        unsigned long old_opts, int flags)
1318 {
1319         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1320             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1321              (flags & MS_RDONLY))) {
1322                 /* wait for any defraggers to finish */
1323                 wait_event(fs_info->transaction_wait,
1324                            (atomic_read(&fs_info->defrag_running) == 0));
1325                 if (flags & MS_RDONLY)
1326                         sync_filesystem(fs_info->sb);
1327         }
1328 }
1329
1330 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1331                                          unsigned long old_opts)
1332 {
1333         /*
1334          * We need cleanup all defragable inodes if the autodefragment is
1335          * close or the fs is R/O.
1336          */
1337         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1338             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1339              (fs_info->sb->s_flags & MS_RDONLY))) {
1340                 btrfs_cleanup_defrag_inodes(fs_info);
1341         }
1342
1343         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1344 }
1345
1346 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1347 {
1348         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1349         struct btrfs_root *root = fs_info->tree_root;
1350         unsigned old_flags = sb->s_flags;
1351         unsigned long old_opts = fs_info->mount_opt;
1352         unsigned long old_compress_type = fs_info->compress_type;
1353         u64 old_max_inline = fs_info->max_inline;
1354         u64 old_alloc_start = fs_info->alloc_start;
1355         int old_thread_pool_size = fs_info->thread_pool_size;
1356         unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1357         int ret;
1358
1359         btrfs_remount_prepare(fs_info);
1360
1361         ret = btrfs_parse_options(root, data);
1362         if (ret) {
1363                 ret = -EINVAL;
1364                 goto restore;
1365         }
1366
1367         btrfs_remount_begin(fs_info, old_opts, *flags);
1368         btrfs_resize_thread_pool(fs_info,
1369                 fs_info->thread_pool_size, old_thread_pool_size);
1370
1371         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1372                 goto out;
1373
1374         if (*flags & MS_RDONLY) {
1375                 /*
1376                  * this also happens on 'umount -rf' or on shutdown, when
1377                  * the filesystem is busy.
1378                  */
1379
1380                 /* wait for the uuid_scan task to finish */
1381                 down(&fs_info->uuid_tree_rescan_sem);
1382                 /* avoid complains from lockdep et al. */
1383                 up(&fs_info->uuid_tree_rescan_sem);
1384
1385                 sb->s_flags |= MS_RDONLY;
1386
1387                 btrfs_dev_replace_suspend_for_unmount(fs_info);
1388                 btrfs_scrub_cancel(fs_info);
1389                 btrfs_pause_balance(fs_info);
1390
1391                 ret = btrfs_commit_super(root);
1392                 if (ret)
1393                         goto restore;
1394         } else {
1395                 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1396                         btrfs_err(fs_info,
1397                                 "Remounting read-write after error is not allowed");
1398                         ret = -EINVAL;
1399                         goto restore;
1400                 }
1401                 if (fs_info->fs_devices->rw_devices == 0) {
1402                         ret = -EACCES;
1403                         goto restore;
1404                 }
1405
1406                 if (fs_info->fs_devices->missing_devices >
1407                      fs_info->num_tolerated_disk_barrier_failures &&
1408                     !(*flags & MS_RDONLY)) {
1409                         btrfs_warn(fs_info,
1410                                 "too many missing devices, writeable remount is not allowed");
1411                         ret = -EACCES;
1412                         goto restore;
1413                 }
1414
1415                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1416                         ret = -EINVAL;
1417                         goto restore;
1418                 }
1419
1420                 ret = btrfs_cleanup_fs_roots(fs_info);
1421                 if (ret)
1422                         goto restore;
1423
1424                 /* recover relocation */
1425                 ret = btrfs_recover_relocation(root);
1426                 if (ret)
1427                         goto restore;
1428
1429                 ret = btrfs_resume_balance_async(fs_info);
1430                 if (ret)
1431                         goto restore;
1432
1433                 ret = btrfs_resume_dev_replace_async(fs_info);
1434                 if (ret) {
1435                         btrfs_warn(fs_info, "failed to resume dev_replace");
1436                         goto restore;
1437                 }
1438
1439                 if (!fs_info->uuid_root) {
1440                         btrfs_info(fs_info, "creating UUID tree");
1441                         ret = btrfs_create_uuid_tree(fs_info);
1442                         if (ret) {
1443                                 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1444                                 goto restore;
1445                         }
1446                 }
1447                 sb->s_flags &= ~MS_RDONLY;
1448         }
1449 out:
1450         btrfs_remount_cleanup(fs_info, old_opts);
1451         return 0;
1452
1453 restore:
1454         /* We've hit an error - don't reset MS_RDONLY */
1455         if (sb->s_flags & MS_RDONLY)
1456                 old_flags |= MS_RDONLY;
1457         sb->s_flags = old_flags;
1458         fs_info->mount_opt = old_opts;
1459         fs_info->compress_type = old_compress_type;
1460         fs_info->max_inline = old_max_inline;
1461         mutex_lock(&fs_info->chunk_mutex);
1462         fs_info->alloc_start = old_alloc_start;
1463         mutex_unlock(&fs_info->chunk_mutex);
1464         btrfs_resize_thread_pool(fs_info,
1465                 old_thread_pool_size, fs_info->thread_pool_size);
1466         fs_info->metadata_ratio = old_metadata_ratio;
1467         btrfs_remount_cleanup(fs_info, old_opts);
1468         return ret;
1469 }
1470
1471 /* Used to sort the devices by max_avail(descending sort) */
1472 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1473                                        const void *dev_info2)
1474 {
1475         if (((struct btrfs_device_info *)dev_info1)->max_avail >
1476             ((struct btrfs_device_info *)dev_info2)->max_avail)
1477                 return -1;
1478         else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1479                  ((struct btrfs_device_info *)dev_info2)->max_avail)
1480                 return 1;
1481         else
1482         return 0;
1483 }
1484
1485 /*
1486  * sort the devices by max_avail, in which max free extent size of each device
1487  * is stored.(Descending Sort)
1488  */
1489 static inline void btrfs_descending_sort_devices(
1490                                         struct btrfs_device_info *devices,
1491                                         size_t nr_devices)
1492 {
1493         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1494              btrfs_cmp_device_free_bytes, NULL);
1495 }
1496
1497 /*
1498  * The helper to calc the free space on the devices that can be used to store
1499  * file data.
1500  */
1501 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1502 {
1503         struct btrfs_fs_info *fs_info = root->fs_info;
1504         struct btrfs_device_info *devices_info;
1505         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1506         struct btrfs_device *device;
1507         u64 skip_space;
1508         u64 type;
1509         u64 avail_space;
1510         u64 used_space;
1511         u64 min_stripe_size;
1512         int min_stripes = 1, num_stripes = 1;
1513         int i = 0, nr_devices;
1514         int ret;
1515
1516         nr_devices = fs_info->fs_devices->open_devices;
1517         BUG_ON(!nr_devices);
1518
1519         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1520                                GFP_NOFS);
1521         if (!devices_info)
1522                 return -ENOMEM;
1523
1524         /* calc min stripe number for data space alloction */
1525         type = btrfs_get_alloc_profile(root, 1);
1526         if (type & BTRFS_BLOCK_GROUP_RAID0) {
1527                 min_stripes = 2;
1528                 num_stripes = nr_devices;
1529         } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1530                 min_stripes = 2;
1531                 num_stripes = 2;
1532         } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1533                 min_stripes = 4;
1534                 num_stripes = 4;
1535         }
1536
1537         if (type & BTRFS_BLOCK_GROUP_DUP)
1538                 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1539         else
1540                 min_stripe_size = BTRFS_STRIPE_LEN;
1541
1542         list_for_each_entry(device, &fs_devices->devices, dev_list) {
1543                 if (!device->in_fs_metadata || !device->bdev ||
1544                     device->is_tgtdev_for_dev_replace)
1545                         continue;
1546
1547                 avail_space = device->total_bytes - device->bytes_used;
1548
1549                 /* align with stripe_len */
1550                 do_div(avail_space, BTRFS_STRIPE_LEN);
1551                 avail_space *= BTRFS_STRIPE_LEN;
1552
1553                 /*
1554                  * In order to avoid overwritting the superblock on the drive,
1555                  * btrfs starts at an offset of at least 1MB when doing chunk
1556                  * allocation.
1557                  */
1558                 skip_space = 1024 * 1024;
1559
1560                 /* user can set the offset in fs_info->alloc_start. */
1561                 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1562                     device->total_bytes)
1563                         skip_space = max(fs_info->alloc_start, skip_space);
1564
1565                 /*
1566                  * btrfs can not use the free space in [0, skip_space - 1],
1567                  * we must subtract it from the total. In order to implement
1568                  * it, we account the used space in this range first.
1569                  */
1570                 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1571                                                      &used_space);
1572                 if (ret) {
1573                         kfree(devices_info);
1574                         return ret;
1575                 }
1576
1577                 /* calc the free space in [0, skip_space - 1] */
1578                 skip_space -= used_space;
1579
1580                 /*
1581                  * we can use the free space in [0, skip_space - 1], subtract
1582                  * it from the total.
1583                  */
1584                 if (avail_space && avail_space >= skip_space)
1585                         avail_space -= skip_space;
1586                 else
1587                         avail_space = 0;
1588
1589                 if (avail_space < min_stripe_size)
1590                         continue;
1591
1592                 devices_info[i].dev = device;
1593                 devices_info[i].max_avail = avail_space;
1594
1595                 i++;
1596         }
1597
1598         nr_devices = i;
1599
1600         btrfs_descending_sort_devices(devices_info, nr_devices);
1601
1602         i = nr_devices - 1;
1603         avail_space = 0;
1604         while (nr_devices >= min_stripes) {
1605                 if (num_stripes > nr_devices)
1606                         num_stripes = nr_devices;
1607
1608                 if (devices_info[i].max_avail >= min_stripe_size) {
1609                         int j;
1610                         u64 alloc_size;
1611
1612                         avail_space += devices_info[i].max_avail * num_stripes;
1613                         alloc_size = devices_info[i].max_avail;
1614                         for (j = i + 1 - num_stripes; j <= i; j++)
1615                                 devices_info[j].max_avail -= alloc_size;
1616                 }
1617                 i--;
1618                 nr_devices--;
1619         }
1620
1621         kfree(devices_info);
1622         *free_bytes = avail_space;
1623         return 0;
1624 }
1625
1626 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1627 {
1628         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1629         struct btrfs_super_block *disk_super = fs_info->super_copy;
1630         struct list_head *head = &fs_info->space_info;
1631         struct btrfs_space_info *found;
1632         u64 total_used = 0;
1633         u64 total_free_data = 0;
1634         int bits = dentry->d_sb->s_blocksize_bits;
1635         __be32 *fsid = (__be32 *)fs_info->fsid;
1636         int ret;
1637
1638         /* holding chunk_muext to avoid allocating new chunks */
1639         mutex_lock(&fs_info->chunk_mutex);
1640         rcu_read_lock();
1641         list_for_each_entry_rcu(found, head, list) {
1642                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1643                         total_free_data += found->disk_total - found->disk_used;
1644                         total_free_data -=
1645                                 btrfs_account_ro_block_groups_free_space(found);
1646                 }
1647
1648                 total_used += found->disk_used;
1649         }
1650         rcu_read_unlock();
1651
1652         buf->f_namelen = BTRFS_NAME_LEN;
1653         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1654         buf->f_bfree = buf->f_blocks - (total_used >> bits);
1655         buf->f_bsize = dentry->d_sb->s_blocksize;
1656         buf->f_type = BTRFS_SUPER_MAGIC;
1657         buf->f_bavail = total_free_data;
1658         ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1659         if (ret) {
1660                 mutex_unlock(&fs_info->chunk_mutex);
1661                 return ret;
1662         }
1663         buf->f_bavail += total_free_data;
1664         buf->f_bavail = buf->f_bavail >> bits;
1665         mutex_unlock(&fs_info->chunk_mutex);
1666
1667         /* We treat it as constant endianness (it doesn't matter _which_)
1668            because we want the fsid to come out the same whether mounted
1669            on a big-endian or little-endian host */
1670         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1671         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1672         /* Mask in the root object ID too, to disambiguate subvols */
1673         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1674         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1675
1676         return 0;
1677 }
1678
1679 static void btrfs_kill_super(struct super_block *sb)
1680 {
1681         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1682         kill_anon_super(sb);
1683         free_fs_info(fs_info);
1684 }
1685
1686 static struct file_system_type btrfs_fs_type = {
1687         .owner          = THIS_MODULE,
1688         .name           = "btrfs",
1689         .mount          = btrfs_mount,
1690         .kill_sb        = btrfs_kill_super,
1691         .fs_flags       = FS_REQUIRES_DEV,
1692 };
1693 MODULE_ALIAS_FS("btrfs");
1694
1695 /*
1696  * used by btrfsctl to scan devices when no FS is mounted
1697  */
1698 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1699                                 unsigned long arg)
1700 {
1701         struct btrfs_ioctl_vol_args *vol;
1702         struct btrfs_fs_devices *fs_devices;
1703         int ret = -ENOTTY;
1704
1705         if (!capable(CAP_SYS_ADMIN))
1706                 return -EPERM;
1707
1708         vol = memdup_user((void __user *)arg, sizeof(*vol));
1709         if (IS_ERR(vol))
1710                 return PTR_ERR(vol);
1711
1712         switch (cmd) {
1713         case BTRFS_IOC_SCAN_DEV:
1714                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1715                                             &btrfs_fs_type, &fs_devices);
1716                 break;
1717         case BTRFS_IOC_DEVICES_READY:
1718                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1719                                             &btrfs_fs_type, &fs_devices);
1720                 if (ret)
1721                         break;
1722                 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1723                 break;
1724         }
1725
1726         kfree(vol);
1727         return ret;
1728 }
1729
1730 static int btrfs_freeze(struct super_block *sb)
1731 {
1732         struct btrfs_trans_handle *trans;
1733         struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1734
1735         trans = btrfs_attach_transaction_barrier(root);
1736         if (IS_ERR(trans)) {
1737                 /* no transaction, don't bother */
1738                 if (PTR_ERR(trans) == -ENOENT)
1739                         return 0;
1740                 return PTR_ERR(trans);
1741         }
1742         return btrfs_commit_transaction(trans, root);
1743 }
1744
1745 static int btrfs_unfreeze(struct super_block *sb)
1746 {
1747         return 0;
1748 }
1749
1750 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1751 {
1752         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1753         struct btrfs_fs_devices *cur_devices;
1754         struct btrfs_device *dev, *first_dev = NULL;
1755         struct list_head *head;
1756         struct rcu_string *name;
1757
1758         mutex_lock(&fs_info->fs_devices->device_list_mutex);
1759         cur_devices = fs_info->fs_devices;
1760         while (cur_devices) {
1761                 head = &cur_devices->devices;
1762                 list_for_each_entry(dev, head, dev_list) {
1763                         if (dev->missing)
1764                                 continue;
1765                         if (!first_dev || dev->devid < first_dev->devid)
1766                                 first_dev = dev;
1767                 }
1768                 cur_devices = cur_devices->seed;
1769         }
1770
1771         if (first_dev) {
1772                 rcu_read_lock();
1773                 name = rcu_dereference(first_dev->name);
1774                 seq_escape(m, name->str, " \t\n\\");
1775                 rcu_read_unlock();
1776         } else {
1777                 WARN_ON(1);
1778         }
1779         mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1780         return 0;
1781 }
1782
1783 static const struct super_operations btrfs_super_ops = {
1784         .drop_inode     = btrfs_drop_inode,
1785         .evict_inode    = btrfs_evict_inode,
1786         .put_super      = btrfs_put_super,
1787         .sync_fs        = btrfs_sync_fs,
1788         .show_options   = btrfs_show_options,
1789         .show_devname   = btrfs_show_devname,
1790         .write_inode    = btrfs_write_inode,
1791         .alloc_inode    = btrfs_alloc_inode,
1792         .destroy_inode  = btrfs_destroy_inode,
1793         .statfs         = btrfs_statfs,
1794         .remount_fs     = btrfs_remount,
1795         .freeze_fs      = btrfs_freeze,
1796         .unfreeze_fs    = btrfs_unfreeze,
1797 };
1798
1799 static const struct file_operations btrfs_ctl_fops = {
1800         .unlocked_ioctl  = btrfs_control_ioctl,
1801         .compat_ioctl = btrfs_control_ioctl,
1802         .owner   = THIS_MODULE,
1803         .llseek = noop_llseek,
1804 };
1805
1806 static struct miscdevice btrfs_misc = {
1807         .minor          = BTRFS_MINOR,
1808         .name           = "btrfs-control",
1809         .fops           = &btrfs_ctl_fops
1810 };
1811
1812 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1813 MODULE_ALIAS("devname:btrfs-control");
1814
1815 static int btrfs_interface_init(void)
1816 {
1817         return misc_register(&btrfs_misc);
1818 }
1819
1820 static void btrfs_interface_exit(void)
1821 {
1822         if (misc_deregister(&btrfs_misc) < 0)
1823                 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
1824 }
1825
1826 static void btrfs_print_info(void)
1827 {
1828         printk(KERN_INFO "Btrfs loaded"
1829 #ifdef CONFIG_BTRFS_DEBUG
1830                         ", debug=on"
1831 #endif
1832 #ifdef CONFIG_BTRFS_ASSERT
1833                         ", assert=on"
1834 #endif
1835 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1836                         ", integrity-checker=on"
1837 #endif
1838                         "\n");
1839 }
1840
1841 static int btrfs_run_sanity_tests(void)
1842 {
1843         int ret;
1844
1845         ret = btrfs_init_test_fs();
1846         if (ret)
1847                 return ret;
1848
1849         ret = btrfs_test_free_space_cache();
1850         if (ret)
1851                 goto out;
1852         ret = btrfs_test_extent_buffer_operations();
1853         if (ret)
1854                 goto out;
1855         ret = btrfs_test_extent_io();
1856         if (ret)
1857                 goto out;
1858         ret = btrfs_test_inodes();
1859 out:
1860         btrfs_destroy_test_fs();
1861         return ret;
1862 }
1863
1864 static int __init init_btrfs_fs(void)
1865 {
1866         int err;
1867
1868         err = btrfs_init_sysfs();
1869         if (err)
1870                 return err;
1871
1872         btrfs_init_compress();
1873
1874         err = btrfs_init_cachep();
1875         if (err)
1876                 goto free_compress;
1877
1878         err = extent_io_init();
1879         if (err)
1880                 goto free_cachep;
1881
1882         err = extent_map_init();
1883         if (err)
1884                 goto free_extent_io;
1885
1886         err = ordered_data_init();
1887         if (err)
1888                 goto free_extent_map;
1889
1890         err = btrfs_delayed_inode_init();
1891         if (err)
1892                 goto free_ordered_data;
1893
1894         err = btrfs_auto_defrag_init();
1895         if (err)
1896                 goto free_delayed_inode;
1897
1898         err = btrfs_delayed_ref_init();
1899         if (err)
1900                 goto free_auto_defrag;
1901
1902         err = btrfs_prelim_ref_init();
1903         if (err)
1904                 goto free_prelim_ref;
1905
1906         err = btrfs_interface_init();
1907         if (err)
1908                 goto free_delayed_ref;
1909
1910         btrfs_init_lockdep();
1911
1912         btrfs_print_info();
1913
1914         err = btrfs_run_sanity_tests();
1915         if (err)
1916                 goto unregister_ioctl;
1917
1918         err = register_filesystem(&btrfs_fs_type);
1919         if (err)
1920                 goto unregister_ioctl;
1921
1922         return 0;
1923
1924 unregister_ioctl:
1925         btrfs_interface_exit();
1926 free_prelim_ref:
1927         btrfs_prelim_ref_exit();
1928 free_delayed_ref:
1929         btrfs_delayed_ref_exit();
1930 free_auto_defrag:
1931         btrfs_auto_defrag_exit();
1932 free_delayed_inode:
1933         btrfs_delayed_inode_exit();
1934 free_ordered_data:
1935         ordered_data_exit();
1936 free_extent_map:
1937         extent_map_exit();
1938 free_extent_io:
1939         extent_io_exit();
1940 free_cachep:
1941         btrfs_destroy_cachep();
1942 free_compress:
1943         btrfs_exit_compress();
1944         btrfs_exit_sysfs();
1945         return err;
1946 }
1947
1948 static void __exit exit_btrfs_fs(void)
1949 {
1950         btrfs_destroy_cachep();
1951         btrfs_delayed_ref_exit();
1952         btrfs_auto_defrag_exit();
1953         btrfs_delayed_inode_exit();
1954         btrfs_prelim_ref_exit();
1955         ordered_data_exit();
1956         extent_map_exit();
1957         extent_io_exit();
1958         btrfs_interface_exit();
1959         unregister_filesystem(&btrfs_fs_type);
1960         btrfs_exit_sysfs();
1961         btrfs_cleanup_fs_uuids();
1962         btrfs_exit_compress();
1963 }
1964
1965 module_init(init_btrfs_fs)
1966 module_exit(exit_btrfs_fs)
1967
1968 MODULE_LICENSE("GPL");