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 int btrfs_remount(struct super_block *sb, int *flags, char *data);
  70
  71 static const char *btrfs_decode_error(int errno)
  72 {
  73         char *errstr = "unknown";
  74
  75         switch (errno) {
  76         case -EIO:
  77                 errstr = "IO failure";
  78                 break;
  79         case -ENOMEM:
  80                 errstr = "Out of memory";
  81                 break;
  82         case -EROFS:
  83                 errstr = "Readonly filesystem";
  84                 break;
  85         case -EEXIST:
  86                 errstr = "Object already exists";
  87                 break;
  88         case -ENOSPC:
  89                 errstr = "No space left";
  90                 break;
  91         case -ENOENT:
  92                 errstr = "No such entry";
  93                 break;
  94         }
  95
  96         return errstr;
  97 }
  98
  99 static void save_error_info(struct btrfs_fs_info *fs_info)
 100 {
 101         /*
 102          * today we only save the error info into ram.  Long term we'll
 103          * also send it down to the disk
 104          */
 105         set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
 106 }
 107
 108 /* btrfs handle error by forcing the filesystem readonly */
 109 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
 110 {
 111         struct super_block *sb = fs_info->sb;
 112
 113         if (sb->s_flags & MS_RDONLY)
 114                 return;
 115
 116         if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
 117                 sb->s_flags |= MS_RDONLY;
 118                 btrfs_info(fs_info, "forced readonly");
 119                 /*
 120                  * Note that a running device replace operation is not
 121                  * canceled here although there is no way to update
 122                  * the progress. It would add the risk of a deadlock,
 123                  * therefore the canceling is ommited. The only penalty
 124                  * is that some I/O remains active until the procedure
 125                  * completes. The next time when the filesystem is
 126                  * mounted writeable again, the device replace
 127                  * operation continues.
 128                  */
 129         }
 130 }
 131
 132 #ifdef CONFIG_PRINTK
 133 /*
 134  * __btrfs_std_error decodes expected errors from the caller and
 135  * invokes the approciate error response.
 136  */
 137 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 138                        unsigned int line, int errno, const char *fmt, ...)
 139 {
 140         struct super_block *sb = fs_info->sb;
 141         const char *errstr;
 142
 143         /*
 144          * Special case: if the error is EROFS, and we're already
 145          * under MS_RDONLY, then it is safe here.
 146          */
 147         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 148                 return;
 149
 150         errstr = btrfs_decode_error(errno);
 151         if (fmt) {
 152                 struct va_format vaf;
 153                 va_list args;
 154
 155                 va_start(args, fmt);
 156                 vaf.fmt = fmt;
 157                 vaf.va = &args;
 158
 159                 printk(KERN_CRIT
 160                         "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
 161                         sb->s_id, function, line, errno, errstr, &vaf);
 162                 va_end(args);
 163         } else {
 164                 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
 165                         sb->s_id, function, line, errno, errstr);
 166         }
 167
 168         /* Don't go through full error handling during mount */
 169         save_error_info(fs_info);
 170         if (sb->s_flags & MS_BORN)
 171                 btrfs_handle_error(fs_info);
 172 }
 173
 174 static const char * const logtypes[] = {
 175         "emergency",
 176         "alert",
 177         "critical",
 178         "error",
 179         "warning",
 180         "notice",
 181         "info",
 182         "debug",
 183 };
 184
 185 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
 186 {
 187         struct super_block *sb = fs_info->sb;
 188         char lvl[4];
 189         struct va_format vaf;
 190         va_list args;
 191         const char *type = logtypes[4];
 192         int kern_level;
 193
 194         va_start(args, fmt);
 195
 196         kern_level = printk_get_level(fmt);
 197         if (kern_level) {
 198                 size_t size = printk_skip_level(fmt) - fmt;
 199                 memcpy(lvl, fmt,  size);
 200                 lvl[size] = '\0';
 201                 fmt += size;
 202                 type = logtypes[kern_level - '0'];
 203         } else
 204                 *lvl = '\0';
 205
 206         vaf.fmt = fmt;
 207         vaf.va = &args;
 208
 209         printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
 210
 211         va_end(args);
 212 }
 213
 214 #else
 215
 216 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 217                        unsigned int line, int errno, const char *fmt, ...)
 218 {
 219         struct super_block *sb = fs_info->sb;
 220
 221         /*
 222          * Special case: if the error is EROFS, and we're already
 223          * under MS_RDONLY, then it is safe here.
 224          */
 225         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 226                 return;
 227
 228         /* Don't go through full error handling during mount */
 229         if (sb->s_flags & MS_BORN) {
 230                 save_error_info(fs_info);
 231                 btrfs_handle_error(fs_info);
 232         }
 233 }
 234 #endif
 235
 236 /*
 237  * We only mark the transaction aborted and then set the file system read-only.
 238  * This will prevent new transactions from starting or trying to join this
 239  * one.
 240  *
 241  * This means that error recovery at the call site is limited to freeing
 242  * any local memory allocations and passing the error code up without
 243  * further cleanup. The transaction should complete as it normally would
 244  * in the call path but will return -EIO.
 245  *
 246  * We'll complete the cleanup in btrfs_end_transaction and
 247  * btrfs_commit_transaction.
 248  */
 249 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
 250                                struct btrfs_root *root, const char *function,
 251                                unsigned int line, int errno)
 252 {
 253         /*
 254          * Report first abort since mount
 255          */
 256         if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,
 257                                 &root->fs_info->fs_state)) {
 258                 WARN(1, KERN_DEBUG "BTRFS: Transaction aborted (error %d)\n",
 259                                 errno);
 260         }
 261         trans->aborted = errno;
 262         /* Nothing used. The other threads that have joined this
 263          * transaction may be able to continue. */
 264         if (!trans->blocks_used) {
 265                 const char *errstr;
 266
 267                 errstr = btrfs_decode_error(errno);
 268                 btrfs_warn(root->fs_info,
 269                            "%s:%d: Aborting unused transaction(%s).",
 270                            function, line, errstr);
 271                 return;
 272         }
 273         ACCESS_ONCE(trans->transaction->aborted) = errno;
 274         /* Wake up anybody who may be waiting on this transaction */
 275         wake_up(&root->fs_info->transaction_wait);
 276         wake_up(&root->fs_info->transaction_blocked_wait);
 277         __btrfs_std_error(root->fs_info, function, line, errno, NULL);
 278 }
 279 /*
 280  * __btrfs_panic decodes unexpected, fatal errors from the caller,
 281  * issues an alert, and either panics or BUGs, depending on mount options.
 282  */
 283 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
 284                    unsigned int line, int errno, const char *fmt, ...)
 285 {
 286         char *s_id = "<unknown>";
 287         const char *errstr;
 288         struct va_format vaf = { .fmt = fmt };
 289         va_list args;
 290
 291         if (fs_info)
 292                 s_id = fs_info->sb->s_id;
 293
 294         va_start(args, fmt);
 295         vaf.va = &args;
 296
 297         errstr = btrfs_decode_error(errno);
 298         if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
 299                 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
 300                         s_id, function, line, &vaf, errno, errstr);
 301
 302         btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
 303                    function, line, &vaf, errno, errstr);
 304         va_end(args);
 305         /* Caller calls BUG() */
 306 }
 307
 308 static void btrfs_put_super(struct super_block *sb)
 309 {
 310         (void)close_ctree(btrfs_sb(sb)->tree_root);
 311         /* FIXME: need to fix VFS to return error? */
 312         /* AV: return it _where_?  ->put_super() can be triggered by any number
 313          * of async events, up to and including delivery of SIGKILL to the
 314          * last process that kept it busy.  Or segfault in the aforementioned
 315          * process...  Whom would you report that to?
 316          */
 317 }
 318
 319 enum {
 320         Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
 321         Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
 322         Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
 323         Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
 324         Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
 325         Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
 326         Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
 327         Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
 328         Opt_check_integrity, Opt_check_integrity_including_extent_data,
 329         Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
 330         Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
 331         Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
 332         Opt_datasum, Opt_treelog, Opt_noinode_cache,
 333         Opt_err,
 334 };
 335
 336 static match_table_t tokens = {
 337         {Opt_degraded, "degraded"},
 338         {Opt_subvol, "subvol=%s"},
 339         {Opt_subvolid, "subvolid=%s"},
 340         {Opt_device, "device=%s"},
 341         {Opt_nodatasum, "nodatasum"},
 342         {Opt_datasum, "datasum"},
 343         {Opt_nodatacow, "nodatacow"},
 344         {Opt_datacow, "datacow"},
 345         {Opt_nobarrier, "nobarrier"},
 346         {Opt_barrier, "barrier"},
 347         {Opt_max_inline, "max_inline=%s"},
 348         {Opt_alloc_start, "alloc_start=%s"},
 349         {Opt_thread_pool, "thread_pool=%d"},
 350         {Opt_compress, "compress"},
 351         {Opt_compress_type, "compress=%s"},
 352         {Opt_compress_force, "compress-force"},
 353         {Opt_compress_force_type, "compress-force=%s"},
 354         {Opt_ssd, "ssd"},
 355         {Opt_ssd_spread, "ssd_spread"},
 356         {Opt_nossd, "nossd"},
 357         {Opt_acl, "acl"},
 358         {Opt_noacl, "noacl"},
 359         {Opt_notreelog, "notreelog"},
 360         {Opt_treelog, "treelog"},
 361         {Opt_flushoncommit, "flushoncommit"},
 362         {Opt_noflushoncommit, "noflushoncommit"},
 363         {Opt_ratio, "metadata_ratio=%d"},
 364         {Opt_discard, "discard"},
 365         {Opt_nodiscard, "nodiscard"},
 366         {Opt_space_cache, "space_cache"},
 367         {Opt_clear_cache, "clear_cache"},
 368         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 369         {Opt_enospc_debug, "enospc_debug"},
 370         {Opt_noenospc_debug, "noenospc_debug"},
 371         {Opt_subvolrootid, "subvolrootid=%d"},
 372         {Opt_defrag, "autodefrag"},
 373         {Opt_nodefrag, "noautodefrag"},
 374         {Opt_inode_cache, "inode_cache"},
 375         {Opt_noinode_cache, "noinode_cache"},
 376         {Opt_no_space_cache, "nospace_cache"},
 377         {Opt_recovery, "recovery"},
 378         {Opt_skip_balance, "skip_balance"},
 379         {Opt_check_integrity, "check_int"},
 380         {Opt_check_integrity_including_extent_data, "check_int_data"},
 381         {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
 382         {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 383         {Opt_fatal_errors, "fatal_errors=%s"},
 384         {Opt_commit_interval, "commit=%d"},
 385         {Opt_err, NULL},
 386 };
 387
 388 /*
 389  * Regular mount options parser.  Everything that is needed only when
 390  * reading in a new superblock is parsed here.
 391  * XXX JDM: This needs to be cleaned up for remount.
 392  */
 393 int btrfs_parse_options(struct btrfs_root *root, char *options)
 394 {
 395         struct btrfs_fs_info *info = root->fs_info;
 396         substring_t args[MAX_OPT_ARGS];
 397         char *p, *num, *orig = NULL;
 398         u64 cache_gen;
 399         int intarg;
 400         int ret = 0;
 401         char *compress_type;
 402         bool compress_force = false;
 403         bool compress = false;
 404
 405         cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
 406         if (cache_gen)
 407                 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 408
 409         if (!options)
 410                 goto out;
 411
 412         /*
 413          * strsep changes the string, duplicate it because parse_options
 414          * gets called twice
 415          */
 416         options = kstrdup(options, GFP_NOFS);
 417         if (!options)
 418                 return -ENOMEM;
 419
 420         orig = options;
 421
 422         while ((p = strsep(&options, ",")) != NULL) {
 423                 int token;
 424                 if (!*p)
 425                         continue;
 426
 427                 token = match_token(p, tokens, args);
 428                 switch (token) {
 429                 case Opt_degraded:
 430                         btrfs_info(root->fs_info, "allowing degraded mounts");
 431                         btrfs_set_opt(info->mount_opt, DEGRADED);
 432                         break;
 433                 case Opt_subvol:
 434                 case Opt_subvolid:
 435                 case Opt_subvolrootid:
 436                 case Opt_device:
 437                         /*
 438                          * These are parsed by btrfs_parse_early_options
 439                          * and can be happily ignored here.
 440                          */
 441                         break;
 442                 case Opt_nodatasum:
 443                         btrfs_set_and_info(root, NODATASUM,
 444                                            "setting nodatasum");
 445                         break;
 446                 case Opt_datasum:
 447                         if (btrfs_test_opt(root, NODATASUM)) {
 448                                 if (btrfs_test_opt(root, NODATACOW))
 449                                         btrfs_info(root->fs_info, "setting datasum, datacow enabled");
 450                                 else
 451                                         btrfs_info(root->fs_info, "setting datasum");
 452                         }
 453                         btrfs_clear_opt(info->mount_opt, NODATACOW);
 454                         btrfs_clear_opt(info->mount_opt, NODATASUM);
 455                         break;
 456                 case Opt_nodatacow:
 457                         if (!btrfs_test_opt(root, NODATACOW)) {
 458                                 if (!btrfs_test_opt(root, COMPRESS) ||
 459                                     !btrfs_test_opt(root, FORCE_COMPRESS)) {
 460                                         btrfs_info(root->fs_info,
 461                                                    "setting nodatacow, compression disabled");
 462                                 } else {
 463                                         btrfs_info(root->fs_info, "setting nodatacow");
 464                                 }
 465                         }
 466                         btrfs_clear_opt(info->mount_opt, COMPRESS);
 467                         btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 468                         btrfs_set_opt(info->mount_opt, NODATACOW);
 469                         btrfs_set_opt(info->mount_opt, NODATASUM);
 470                         break;
 471                 case Opt_datacow:
 472                         btrfs_clear_and_info(root, NODATACOW,
 473                                              "setting datacow");
 474                         break;
 475                 case Opt_compress_force:
 476                 case Opt_compress_force_type:
 477                         compress_force = true;
 478                         /* Fallthrough */
 479                 case Opt_compress:
 480                 case Opt_compress_type:
 481                         compress = true;
 482                         if (token == Opt_compress ||
 483                             token == Opt_compress_force ||
 484                             strcmp(args[0].from, "zlib") == 0) {
 485                                 compress_type = "zlib";
 486                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 487                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 488                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 489                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 490                         } else if (strcmp(args[0].from, "lzo") == 0) {
 491                                 compress_type = "lzo";
 492                                 info->compress_type = BTRFS_COMPRESS_LZO;
 493                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 494                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 495                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 496                                 btrfs_set_fs_incompat(info, COMPRESS_LZO);
 497                         } else if (strncmp(args[0].from, "no", 2) == 0) {
 498                                 compress_type = "no";
 499                                 btrfs_clear_opt(info->mount_opt, COMPRESS);
 500                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 501                                 compress_force = false;
 502                         } else {
 503                                 ret = -EINVAL;
 504                                 goto out;
 505                         }
 506
 507                         if (compress_force) {
 508                                 btrfs_set_and_info(root, FORCE_COMPRESS,
 509                                                    "force %s compression",
 510                                                    compress_type);
 511                         } else if (compress) {
 512                                 if (!btrfs_test_opt(root, COMPRESS))
 513                                         btrfs_info(root->fs_info,
 514                                                    "btrfs: use %s compression",
 515                                                    compress_type);
 516                         }
 517                         break;
 518                 case Opt_ssd:
 519                         btrfs_set_and_info(root, SSD,
 520                                            "use ssd allocation scheme");
 521                         break;
 522                 case Opt_ssd_spread:
 523                         btrfs_set_and_info(root, SSD_SPREAD,
 524                                            "use spread ssd allocation scheme");
 525                         break;
 526                 case Opt_nossd:
 527                         btrfs_clear_and_info(root, NOSSD,
 528                                              "not using ssd allocation scheme");
 529                         btrfs_clear_opt(info->mount_opt, SSD);
 530                         break;
 531                 case Opt_barrier:
 532                         btrfs_clear_and_info(root, NOBARRIER,
 533                                              "turning on barriers");
 534                         break;
 535                 case Opt_nobarrier:
 536                         btrfs_set_and_info(root, NOBARRIER,
 537                                            "turning off barriers");
 538                         break;
 539                 case Opt_thread_pool:
 540                         ret = match_int(&args[0], &intarg);
 541                         if (ret) {
 542                                 goto out;
 543                         } else if (intarg > 0) {
 544                                 info->thread_pool_size = intarg;
 545                         } else {
 546                                 ret = -EINVAL;
 547                                 goto out;
 548                         }
 549                         break;
 550                 case Opt_max_inline:
 551                         num = match_strdup(&args[0]);
 552                         if (num) {
 553                                 info->max_inline = memparse(num, NULL);
 554                                 kfree(num);
 555
 556                                 if (info->max_inline) {
 557                                         info->max_inline = min_t(u64,
 558                                                 info->max_inline,
 559                                                 root->sectorsize);
 560                                 }
 561                                 btrfs_info(root->fs_info, "max_inline at %llu",
 562                                         info->max_inline);
 563                         } else {
 564                                 ret = -ENOMEM;
 565                                 goto out;
 566                         }
 567                         break;
 568                 case Opt_alloc_start:
 569                         num = match_strdup(&args[0]);
 570                         if (num) {
 571                                 mutex_lock(&info->chunk_mutex);
 572                                 info->alloc_start = memparse(num, NULL);
 573                                 mutex_unlock(&info->chunk_mutex);
 574                                 kfree(num);
 575                                 btrfs_info(root->fs_info, "allocations start at %llu",
 576                                         info->alloc_start);
 577                         } else {
 578                                 ret = -ENOMEM;
 579                                 goto out;
 580                         }
 581                         break;
 582                 case Opt_acl:
 583 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 584                         root->fs_info->sb->s_flags |= MS_POSIXACL;
 585                         break;
 586 #else
 587                         btrfs_err(root->fs_info,
 588                                 "support for ACL not compiled in!");
 589                         ret = -EINVAL;
 590                         goto out;
 591 #endif
 592                 case Opt_noacl:
 593                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 594                         break;
 595                 case Opt_notreelog:
 596                         btrfs_set_and_info(root, NOTREELOG,
 597                                            "disabling tree log");
 598                         break;
 599                 case Opt_treelog:
 600                         btrfs_clear_and_info(root, NOTREELOG,
 601                                              "enabling tree log");
 602                         break;
 603                 case Opt_flushoncommit:
 604                         btrfs_set_and_info(root, FLUSHONCOMMIT,
 605                                            "turning on flush-on-commit");
 606                         break;
 607                 case Opt_noflushoncommit:
 608                         btrfs_clear_and_info(root, FLUSHONCOMMIT,
 609                                              "turning off flush-on-commit");
 610                         break;
 611                 case Opt_ratio:
 612                         ret = match_int(&args[0], &intarg);
 613                         if (ret) {
 614                                 goto out;
 615                         } else if (intarg >= 0) {
 616                                 info->metadata_ratio = intarg;
 617                                 btrfs_info(root->fs_info, "metadata ratio %d",
 618                                        info->metadata_ratio);
 619                         } else {
 620                                 ret = -EINVAL;
 621                                 goto out;
 622                         }
 623                         break;
 624                 case Opt_discard:
 625                         btrfs_set_and_info(root, DISCARD,
 626                                            "turning on discard");
 627                         break;
 628                 case Opt_nodiscard:
 629                         btrfs_clear_and_info(root, DISCARD,
 630                                              "turning off discard");
 631                         break;
 632                 case Opt_space_cache:
 633                         btrfs_set_and_info(root, SPACE_CACHE,
 634                                            "enabling disk space caching");
 635                         break;
 636                 case Opt_rescan_uuid_tree:
 637                         btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 638                         break;
 639                 case Opt_no_space_cache:
 640                         btrfs_clear_and_info(root, SPACE_CACHE,
 641                                              "disabling disk space caching");
 642                         break;
 643                 case Opt_inode_cache:
 644                         btrfs_set_and_info(root, CHANGE_INODE_CACHE,
 645                                            "enabling inode map caching");
 646                         break;
 647                 case Opt_noinode_cache:
 648                         btrfs_clear_and_info(root, CHANGE_INODE_CACHE,
 649                                              "disabling inode map caching");
 650                         break;
 651                 case Opt_clear_cache:
 652                         btrfs_set_and_info(root, CLEAR_CACHE,
 653                                            "force clearing of disk cache");
 654                         break;
 655                 case Opt_user_subvol_rm_allowed:
 656                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 657                         break;
 658                 case Opt_enospc_debug:
 659                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 660                         break;
 661                 case Opt_noenospc_debug:
 662                         btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
 663                         break;
 664                 case Opt_defrag:
 665                         btrfs_set_and_info(root, AUTO_DEFRAG,
 666                                            "enabling auto defrag");
 667                         break;
 668                 case Opt_nodefrag:
 669                         btrfs_clear_and_info(root, AUTO_DEFRAG,
 670                                              "disabling auto defrag");
 671                         break;
 672                 case Opt_recovery:
 673                         btrfs_info(root->fs_info, "enabling auto recovery");
 674                         btrfs_set_opt(info->mount_opt, RECOVERY);
 675                         break;
 676                 case Opt_skip_balance:
 677                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 678                         break;
 679 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 680                 case Opt_check_integrity_including_extent_data:
 681                         btrfs_info(root->fs_info,
 682                                    "enabling check integrity including extent data");
 683                         btrfs_set_opt(info->mount_opt,
 684                                       CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
 685                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 686                         break;
 687                 case Opt_check_integrity:
 688                         btrfs_info(root->fs_info, "enabling check integrity");
 689                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 690                         break;
 691                 case Opt_check_integrity_print_mask:
 692                         ret = match_int(&args[0], &intarg);
 693                         if (ret) {
 694                                 goto out;
 695                         } else if (intarg >= 0) {
 696                                 info->check_integrity_print_mask = intarg;
 697                                 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
 698                                        info->check_integrity_print_mask);
 699                         } else {
 700                                 ret = -EINVAL;
 701                                 goto out;
 702                         }
 703                         break;
 704 #else
 705                 case Opt_check_integrity_including_extent_data:
 706                 case Opt_check_integrity:
 707                 case Opt_check_integrity_print_mask:
 708                         btrfs_err(root->fs_info,
 709                                 "support for check_integrity* not compiled in!");
 710                         ret = -EINVAL;
 711                         goto out;
 712 #endif
 713                 case Opt_fatal_errors:
 714                         if (strcmp(args[0].from, "panic") == 0)
 715                                 btrfs_set_opt(info->mount_opt,
 716                                               PANIC_ON_FATAL_ERROR);
 717                         else if (strcmp(args[0].from, "bug") == 0)
 718                                 btrfs_clear_opt(info->mount_opt,
 719                                               PANIC_ON_FATAL_ERROR);
 720                         else {
 721                                 ret = -EINVAL;
 722                                 goto out;
 723                         }
 724                         break;
 725                 case Opt_commit_interval:
 726                         intarg = 0;
 727                         ret = match_int(&args[0], &intarg);
 728                         if (ret < 0) {
 729                                 btrfs_err(root->fs_info, "invalid commit interval");
 730                                 ret = -EINVAL;
 731                                 goto out;
 732                         }
 733                         if (intarg > 0) {
 734                                 if (intarg > 300) {
 735                                         btrfs_warn(root->fs_info, "excessive commit interval %d",
 736                                                         intarg);
 737                                 }
 738                                 info->commit_interval = intarg;
 739                         } else {
 740                                 btrfs_info(root->fs_info, "using default commit interval %ds",
 741                                     BTRFS_DEFAULT_COMMIT_INTERVAL);
 742                                 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
 743                         }
 744                         break;
 745                 case Opt_err:
 746                         btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
 747                         ret = -EINVAL;
 748                         goto out;
 749                 default:
 750                         break;
 751                 }
 752         }
 753 out:
 754         if (!ret && btrfs_test_opt(root, SPACE_CACHE))
 755                 btrfs_info(root->fs_info, "disk space caching is enabled");
 756         kfree(orig);
 757         return ret;
 758 }
 759
 760 /*
 761  * Parse mount options that are required early in the mount process.
 762  *
 763  * All other options will be parsed on much later in the mount process and
 764  * only when we need to allocate a new super block.
 765  */
 766 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 767                 void *holder, char **subvol_name, u64 *subvol_objectid,
 768                 struct btrfs_fs_devices **fs_devices)
 769 {
 770         substring_t args[MAX_OPT_ARGS];
 771         char *device_name, *opts, *orig, *p;
 772         char *num = NULL;
 773         int error = 0;
 774
 775         if (!options)
 776                 return 0;
 777
 778         /*
 779          * strsep changes the string, duplicate it because parse_options
 780          * gets called twice
 781          */
 782         opts = kstrdup(options, GFP_KERNEL);
 783         if (!opts)
 784                 return -ENOMEM;
 785         orig = opts;
 786
 787         while ((p = strsep(&opts, ",")) != NULL) {
 788                 int token;
 789                 if (!*p)
 790                         continue;
 791
 792                 token = match_token(p, tokens, args);
 793                 switch (token) {
 794                 case Opt_subvol:
 795                         kfree(*subvol_name);
 796                         *subvol_name = match_strdup(&args[0]);
 797                         if (!*subvol_name) {
 798                                 error = -ENOMEM;
 799                                 goto out;
 800                         }
 801                         break;
 802                 case Opt_subvolid:
 803                         num = match_strdup(&args[0]);
 804                         if (num) {
 805                                 *subvol_objectid = memparse(num, NULL);
 806                                 kfree(num);
 807                                 /* we want the original fs_tree */
 808                                 if (!*subvol_objectid)
 809                                         *subvol_objectid =
 810                                                 BTRFS_FS_TREE_OBJECTID;
 811                         } else {
 812                                 error = -EINVAL;
 813                                 goto out;
 814                         }
 815                         break;
 816                 case Opt_subvolrootid:
 817                         printk(KERN_WARNING
 818                                 "BTRFS: 'subvolrootid' mount option is deprecated and has "
 819                                 "no effect\n");
 820                         break;
 821                 case Opt_device:
 822                         device_name = match_strdup(&args[0]);
 823                         if (!device_name) {
 824                                 error = -ENOMEM;
 825                                 goto out;
 826                         }
 827                         error = btrfs_scan_one_device(device_name,
 828                                         flags, holder, fs_devices);
 829                         kfree(device_name);
 830                         if (error)
 831                                 goto out;
 832                         break;
 833                 default:
 834                         break;
 835                 }
 836         }
 837
 838 out:
 839         kfree(orig);
 840         return error;
 841 }
 842
 843 static struct dentry *get_default_root(struct super_block *sb,
 844                                        u64 subvol_objectid)
 845 {
 846         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 847         struct btrfs_root *root = fs_info->tree_root;
 848         struct btrfs_root *new_root;
 849         struct btrfs_dir_item *di;
 850         struct btrfs_path *path;
 851         struct btrfs_key location;
 852         struct inode *inode;
 853         struct dentry *dentry;
 854         u64 dir_id;
 855         int new = 0;
 856
 857         /*
 858          * We have a specific subvol we want to mount, just setup location and
 859          * go look up the root.
 860          */
 861         if (subvol_objectid) {
 862                 location.objectid = subvol_objectid;
 863                 location.type = BTRFS_ROOT_ITEM_KEY;
 864                 location.offset = (u64)-1;
 865                 goto find_root;
 866         }
 867
 868         path = btrfs_alloc_path();
 869         if (!path)
 870                 return ERR_PTR(-ENOMEM);
 871         path->leave_spinning = 1;
 872
 873         /*
 874          * Find the "default" dir item which points to the root item that we
 875          * will mount by default if we haven't been given a specific subvolume
 876          * to mount.
 877          */
 878         dir_id = btrfs_super_root_dir(fs_info->super_copy);
 879         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 880         if (IS_ERR(di)) {
 881                 btrfs_free_path(path);
 882                 return ERR_CAST(di);
 883         }
 884         if (!di) {
 885                 /*
 886                  * Ok the default dir item isn't there.  This is weird since
 887                  * it's always been there, but don't freak out, just try and
 888                  * mount to root most subvolume.
 889                  */
 890                 btrfs_free_path(path);
 891                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 892                 new_root = fs_info->fs_root;
 893                 goto setup_root;
 894         }
 895
 896         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 897         btrfs_free_path(path);
 898
 899 find_root:
 900         new_root = btrfs_read_fs_root_no_name(fs_info, &location);
 901         if (IS_ERR(new_root))
 902                 return ERR_CAST(new_root);
 903
 904         dir_id = btrfs_root_dirid(&new_root->root_item);
 905 setup_root:
 906         location.objectid = dir_id;
 907         location.type = BTRFS_INODE_ITEM_KEY;
 908         location.offset = 0;
 909
 910         inode = btrfs_iget(sb, &location, new_root, &new);
 911         if (IS_ERR(inode))
 912                 return ERR_CAST(inode);
 913
 914         /*
 915          * If we're just mounting the root most subvol put the inode and return
 916          * a reference to the dentry.  We will have already gotten a reference
 917          * to the inode in btrfs_fill_super so we're good to go.
 918          */
 919         if (!new && sb->s_root->d_inode == inode) {
 920                 iput(inode);
 921                 return dget(sb->s_root);
 922         }
 923
 924         dentry = d_obtain_alias(inode);
 925         if (!IS_ERR(dentry)) {
 926                 spin_lock(&dentry->d_lock);
 927                 dentry->d_flags &= ~DCACHE_DISCONNECTED;
 928                 spin_unlock(&dentry->d_lock);
 929         }
 930         return dentry;
 931 }
 932
 933 static int btrfs_fill_super(struct super_block *sb,
 934                             struct btrfs_fs_devices *fs_devices,
 935                             void *data, int silent)
 936 {
 937         struct inode *inode;
 938         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 939         struct btrfs_key key;
 940         int err;
 941
 942         sb->s_maxbytes = MAX_LFS_FILESIZE;
 943         sb->s_magic = BTRFS_SUPER_MAGIC;
 944         sb->s_op = &btrfs_super_ops;
 945         sb->s_d_op = &btrfs_dentry_operations;
 946         sb->s_export_op = &btrfs_export_ops;
 947         sb->s_xattr = btrfs_xattr_handlers;
 948         sb->s_time_gran = 1;
 949 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 950         sb->s_flags |= MS_POSIXACL;
 951 #endif
 952         sb->s_flags |= MS_I_VERSION;
 953         err = open_ctree(sb, fs_devices, (char *)data);
 954         if (err) {
 955                 printk(KERN_ERR "BTRFS: open_ctree failed\n");
 956                 return err;
 957         }
 958
 959         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 960         key.type = BTRFS_INODE_ITEM_KEY;
 961         key.offset = 0;
 962         inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
 963         if (IS_ERR(inode)) {
 964                 err = PTR_ERR(inode);
 965                 goto fail_close;
 966         }
 967
 968         sb->s_root = d_make_root(inode);
 969         if (!sb->s_root) {
 970                 err = -ENOMEM;
 971                 goto fail_close;
 972         }
 973
 974         save_mount_options(sb, data);
 975         cleancache_init_fs(sb);
 976         sb->s_flags |= MS_ACTIVE;
 977         return 0;
 978
 979 fail_close:
 980         close_ctree(fs_info->tree_root);
 981         return err;
 982 }
 983
 984 int btrfs_sync_fs(struct super_block *sb, int wait)
 985 {
 986         struct btrfs_trans_handle *trans;
 987         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 988         struct btrfs_root *root = fs_info->tree_root;
 989
 990         trace_btrfs_sync_fs(wait);
 991
 992         if (!wait) {
 993                 filemap_flush(fs_info->btree_inode->i_mapping);
 994                 return 0;
 995         }
 996
 997         btrfs_wait_ordered_roots(fs_info, -1);
 998
 999         trans = btrfs_attach_transaction_barrier(root);
1000         if (IS_ERR(trans)) {
1001                 /* no transaction, don't bother */
1002                 if (PTR_ERR(trans) == -ENOENT)
1003                         return 0;
1004                 return PTR_ERR(trans);
1005         }
1006         return btrfs_commit_transaction(trans, root);
1007 }
1008
1009 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1010 {
1011         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1012         struct btrfs_root *root = info->tree_root;
1013         char *compress_type;
1014
1015         if (btrfs_test_opt(root, DEGRADED))
1016                 seq_puts(seq, ",degraded");
1017         if (btrfs_test_opt(root, NODATASUM))
1018                 seq_puts(seq, ",nodatasum");
1019         if (btrfs_test_opt(root, NODATACOW))
1020                 seq_puts(seq, ",nodatacow");
1021         if (btrfs_test_opt(root, NOBARRIER))
1022                 seq_puts(seq, ",nobarrier");
1023         if (info->max_inline != 8192 * 1024)
1024                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1025         if (info->alloc_start != 0)
1026                 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1027         if (info->thread_pool_size !=  min_t(unsigned long,
1028                                              num_online_cpus() + 2, 8))
1029                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1030         if (btrfs_test_opt(root, COMPRESS)) {
1031                 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1032                         compress_type = "zlib";
1033                 else
1034                         compress_type = "lzo";
1035                 if (btrfs_test_opt(root, FORCE_COMPRESS))
1036                         seq_printf(seq, ",compress-force=%s", compress_type);
1037                 else
1038                         seq_printf(seq, ",compress=%s", compress_type);
1039         }
1040         if (btrfs_test_opt(root, NOSSD))
1041                 seq_puts(seq, ",nossd");
1042         if (btrfs_test_opt(root, SSD_SPREAD))
1043                 seq_puts(seq, ",ssd_spread");
1044         else if (btrfs_test_opt(root, SSD))
1045                 seq_puts(seq, ",ssd");
1046         if (btrfs_test_opt(root, NOTREELOG))
1047                 seq_puts(seq, ",notreelog");
1048         if (btrfs_test_opt(root, FLUSHONCOMMIT))
1049                 seq_puts(seq, ",flushoncommit");
1050         if (btrfs_test_opt(root, DISCARD))
1051                 seq_puts(seq, ",discard");
1052         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1053                 seq_puts(seq, ",noacl");
1054         if (btrfs_test_opt(root, SPACE_CACHE))
1055                 seq_puts(seq, ",space_cache");
1056         else
1057                 seq_puts(seq, ",nospace_cache");
1058         if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1059                 seq_puts(seq, ",rescan_uuid_tree");
1060         if (btrfs_test_opt(root, CLEAR_CACHE))
1061                 seq_puts(seq, ",clear_cache");
1062         if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1063                 seq_puts(seq, ",user_subvol_rm_allowed");
1064         if (btrfs_test_opt(root, ENOSPC_DEBUG))
1065                 seq_puts(seq, ",enospc_debug");
1066         if (btrfs_test_opt(root, AUTO_DEFRAG))
1067                 seq_puts(seq, ",autodefrag");
1068         if (btrfs_test_opt(root, INODE_MAP_CACHE))
1069                 seq_puts(seq, ",inode_cache");
1070         if (btrfs_test_opt(root, SKIP_BALANCE))
1071                 seq_puts(seq, ",skip_balance");
1072         if (btrfs_test_opt(root, RECOVERY))
1073                 seq_puts(seq, ",recovery");
1074 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1075         if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1076                 seq_puts(seq, ",check_int_data");
1077         else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1078                 seq_puts(seq, ",check_int");
1079         if (info->check_integrity_print_mask)
1080                 seq_printf(seq, ",check_int_print_mask=%d",
1081                                 info->check_integrity_print_mask);
1082 #endif
1083         if (info->metadata_ratio)
1084                 seq_printf(seq, ",metadata_ratio=%d",
1085                                 info->metadata_ratio);
1086         if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1087                 seq_puts(seq, ",fatal_errors=panic");
1088         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1089                 seq_printf(seq, ",commit=%d", info->commit_interval);
1090         return 0;
1091 }
1092
1093 static int btrfs_test_super(struct super_block *s, void *data)
1094 {
1095         struct btrfs_fs_info *p = data;
1096         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1097
1098         return fs_info->fs_devices == p->fs_devices;
1099 }
1100
1101 static int btrfs_set_super(struct super_block *s, void *data)
1102 {
1103         int err = set_anon_super(s, data);
1104         if (!err)
1105                 s->s_fs_info = data;
1106         return err;
1107 }
1108
1109 /*
1110  * subvolumes are identified by ino 256
1111  */
1112 static inline int is_subvolume_inode(struct inode *inode)
1113 {
1114         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1115                 return 1;
1116         return 0;
1117 }
1118
1119 /*
1120  * This will strip out the subvol=%s argument for an argument string and add
1121  * subvolid=0 to make sure we get the actual tree root for path walking to the
1122  * subvol we want.
1123  */
1124 static char *setup_root_args(char *args)
1125 {
1126         unsigned len = strlen(args) + 2 + 1;
1127         char *src, *dst, *buf;
1128
1129         /*
1130          * We need the same args as before, but with this substitution:
1131          * s!subvol=[^,]+!subvolid=0!
1132          *
1133          * Since the replacement string is up to 2 bytes longer than the
1134          * original, allocate strlen(args) + 2 + 1 bytes.
1135          */
1136
1137         src = strstr(args, "subvol=");
1138         /* This shouldn't happen, but just in case.. */
1139         if (!src)
1140                 return NULL;
1141
1142         buf = dst = kmalloc(len, GFP_NOFS);
1143         if (!buf)
1144                 return NULL;
1145
1146         /*
1147          * If the subvol= arg is not at the start of the string,
1148          * copy whatever precedes it into buf.
1149          */
1150         if (src != args) {
1151                 *src++ = '\0';
1152                 strcpy(buf, args);
1153                 dst += strlen(args);
1154         }
1155
1156         strcpy(dst, "subvolid=0");
1157         dst += strlen("subvolid=0");
1158
1159         /*
1160          * If there is a "," after the original subvol=... string,
1161          * copy that suffix into our buffer.  Otherwise, we're done.
1162          */
1163         src = strchr(src, ',');
1164         if (src)
1165                 strcpy(dst, src);
1166
1167         return buf;
1168 }
1169
1170 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1171                                    const char *device_name, char *data)
1172 {
1173         struct dentry *root;
1174         struct vfsmount *mnt;
1175         char *newargs;
1176
1177         newargs = setup_root_args(data);
1178         if (!newargs)
1179                 return ERR_PTR(-ENOMEM);
1180         mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1181                              newargs);
1182
1183         if (PTR_RET(mnt) == -EBUSY) {
1184                 if (flags & MS_RDONLY) {
1185                         mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY, device_name,
1186                                              newargs);
1187                 } else {
1188                         int r;
1189                         mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY, device_name,
1190                                              newargs);
1191                         if (IS_ERR(mnt)) {
1192                                 kfree(newargs);
1193                                 return ERR_CAST(mnt);
1194                         }
1195
1196                         r = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1197                         if (r < 0) {
1198                                 /* FIXME: release vfsmount mnt ??*/
1199                                 kfree(newargs);
1200                                 return ERR_PTR(r);
1201                         }
1202                 }
1203         }
1204
1205         kfree(newargs);
1206
1207         if (IS_ERR(mnt))
1208                 return ERR_CAST(mnt);
1209
1210         root = mount_subtree(mnt, subvol_name);
1211
1212         if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1213                 struct super_block *s = root->d_sb;
1214                 dput(root);
1215                 root = ERR_PTR(-EINVAL);
1216                 deactivate_locked_super(s);
1217                 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1218                                 subvol_name);
1219         }
1220
1221         return root;
1222 }
1223
1224 /*
1225  * Find a superblock for the given device / mount point.
1226  *
1227  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
1228  *        for multiple device setup.  Make sure to keep it in sync.
1229  */
1230 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1231                 const char *device_name, void *data)
1232 {
1233         struct block_device *bdev = NULL;
1234         struct super_block *s;
1235         struct dentry *root;
1236         struct btrfs_fs_devices *fs_devices = NULL;
1237         struct btrfs_fs_info *fs_info = NULL;
1238         fmode_t mode = FMODE_READ;
1239         char *subvol_name = NULL;
1240         u64 subvol_objectid = 0;
1241         int error = 0;
1242
1243         if (!(flags & MS_RDONLY))
1244                 mode |= FMODE_WRITE;
1245
1246         error = btrfs_parse_early_options(data, mode, fs_type,
1247                                           &subvol_name, &subvol_objectid,
1248                                           &fs_devices);
1249         if (error) {
1250                 kfree(subvol_name);
1251                 return ERR_PTR(error);
1252         }
1253
1254         if (subvol_name) {
1255                 root = mount_subvol(subvol_name, flags, device_name, data);
1256                 kfree(subvol_name);
1257                 return root;
1258         }
1259
1260         error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1261         if (error)
1262                 return ERR_PTR(error);
1263
1264         /*
1265          * Setup a dummy root and fs_info for test/set super.  This is because
1266          * we don't actually fill this stuff out until open_ctree, but we need
1267          * it for searching for existing supers, so this lets us do that and
1268          * then open_ctree will properly initialize everything later.
1269          */
1270         fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1271         if (!fs_info)
1272                 return ERR_PTR(-ENOMEM);
1273
1274         fs_info->fs_devices = fs_devices;
1275
1276         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1277         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1278         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1279                 error = -ENOMEM;
1280                 goto error_fs_info;
1281         }
1282
1283         error = btrfs_open_devices(fs_devices, mode, fs_type);
1284         if (error)
1285                 goto error_fs_info;
1286
1287         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1288                 error = -EACCES;
1289                 goto error_close_devices;
1290         }
1291
1292         bdev = fs_devices->latest_bdev;
1293         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1294                  fs_info);
1295         if (IS_ERR(s)) {
1296                 error = PTR_ERR(s);
1297                 goto error_close_devices;
1298         }
1299
1300         if (s->s_root) {
1301                 btrfs_close_devices(fs_devices);
1302                 free_fs_info(fs_info);
1303                 if ((flags ^ s->s_flags) & MS_RDONLY)
1304                         error = -EBUSY;
1305         } else {
1306                 char b[BDEVNAME_SIZE];
1307
1308                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1309                 btrfs_sb(s)->bdev_holder = fs_type;
1310                 error = btrfs_fill_super(s, fs_devices, data,
1311                                          flags & MS_SILENT ? 1 : 0);
1312         }
1313
1314         root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1315         if (IS_ERR(root))
1316                 deactivate_locked_super(s);
1317
1318         return root;
1319
1320 error_close_devices:
1321         btrfs_close_devices(fs_devices);
1322 error_fs_info:
1323         free_fs_info(fs_info);
1324         return ERR_PTR(error);
1325 }
1326
1327 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1328                                      int new_pool_size, int old_pool_size)
1329 {
1330         if (new_pool_size == old_pool_size)
1331                 return;
1332
1333         fs_info->thread_pool_size = new_pool_size;
1334
1335         btrfs_info(fs_info, "resize thread pool %d -> %d",
1336                old_pool_size, new_pool_size);
1337
1338         btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1339         btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1340         btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1341         btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1342         btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1343         btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1344         btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1345                                 new_pool_size);
1346         btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1347         btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1348         btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1349         btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1350         btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1351                                 new_pool_size);
1352 }
1353
1354 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1355 {
1356         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1357 }
1358
1359 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1360                                        unsigned long old_opts, int flags)
1361 {
1362         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1363             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1364              (flags & MS_RDONLY))) {
1365                 /* wait for any defraggers to finish */
1366                 wait_event(fs_info->transaction_wait,
1367                            (atomic_read(&fs_info->defrag_running) == 0));
1368                 if (flags & MS_RDONLY)
1369                         sync_filesystem(fs_info->sb);
1370         }
1371 }
1372
1373 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1374                                          unsigned long old_opts)
1375 {
1376         /*
1377          * We need cleanup all defragable inodes if the autodefragment is
1378          * close or the fs is R/O.
1379          */
1380         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1381             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1382              (fs_info->sb->s_flags & MS_RDONLY))) {
1383                 btrfs_cleanup_defrag_inodes(fs_info);
1384         }
1385
1386         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1387 }
1388
1389 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1390 {
1391         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1392         struct btrfs_root *root = fs_info->tree_root;
1393         unsigned old_flags = sb->s_flags;
1394         unsigned long old_opts = fs_info->mount_opt;
1395         unsigned long old_compress_type = fs_info->compress_type;
1396         u64 old_max_inline = fs_info->max_inline;
1397         u64 old_alloc_start = fs_info->alloc_start;
1398         int old_thread_pool_size = fs_info->thread_pool_size;
1399         unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1400         int ret;
1401
1402         sync_filesystem(sb);
1403         btrfs_remount_prepare(fs_info);
1404
1405         ret = btrfs_parse_options(root, data);
1406         if (ret) {
1407                 ret = -EINVAL;
1408                 goto restore;
1409         }
1410
1411         btrfs_remount_begin(fs_info, old_opts, *flags);
1412         btrfs_resize_thread_pool(fs_info,
1413                 fs_info->thread_pool_size, old_thread_pool_size);
1414
1415         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1416                 goto out;
1417
1418         if (*flags & MS_RDONLY) {
1419                 /*
1420                  * this also happens on 'umount -rf' or on shutdown, when
1421                  * the filesystem is busy.
1422                  */
1423                 cancel_work_sync(&fs_info->async_reclaim_work);
1424
1425                 /* wait for the uuid_scan task to finish */
1426                 down(&fs_info->uuid_tree_rescan_sem);
1427                 /* avoid complains from lockdep et al. */
1428                 up(&fs_info->uuid_tree_rescan_sem);
1429
1430                 sb->s_flags |= MS_RDONLY;
1431
1432                 btrfs_dev_replace_suspend_for_unmount(fs_info);
1433                 btrfs_scrub_cancel(fs_info);
1434                 btrfs_pause_balance(fs_info);
1435
1436                 ret = btrfs_commit_super(root);
1437                 if (ret)
1438                         goto restore;
1439         } else {
1440                 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1441                         btrfs_err(fs_info,
1442                                 "Remounting read-write after error is not allowed");
1443                         ret = -EINVAL;
1444                         goto restore;
1445                 }
1446                 if (fs_info->fs_devices->rw_devices == 0) {
1447                         ret = -EACCES;
1448                         goto restore;
1449                 }
1450
1451                 if (fs_info->fs_devices->missing_devices >
1452                      fs_info->num_tolerated_disk_barrier_failures &&
1453                     !(*flags & MS_RDONLY)) {
1454                         btrfs_warn(fs_info,
1455                                 "too many missing devices, writeable remount is not allowed");
1456                         ret = -EACCES;
1457                         goto restore;
1458                 }
1459
1460                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1461                         ret = -EINVAL;
1462                         goto restore;
1463                 }
1464
1465                 ret = btrfs_cleanup_fs_roots(fs_info);
1466                 if (ret)
1467                         goto restore;
1468
1469                 /* recover relocation */
1470                 ret = btrfs_recover_relocation(root);
1471                 if (ret)
1472                         goto restore;
1473
1474                 ret = btrfs_resume_balance_async(fs_info);
1475                 if (ret)
1476                         goto restore;
1477
1478                 ret = btrfs_resume_dev_replace_async(fs_info);
1479                 if (ret) {
1480                         btrfs_warn(fs_info, "failed to resume dev_replace");
1481                         goto restore;
1482                 }
1483
1484                 if (!fs_info->uuid_root) {
1485                         btrfs_info(fs_info, "creating UUID tree");
1486                         ret = btrfs_create_uuid_tree(fs_info);
1487                         if (ret) {
1488                                 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1489                                 goto restore;
1490                         }
1491                 }
1492                 sb->s_flags &= ~MS_RDONLY;
1493         }
1494 out:
1495         wake_up_process(fs_info->transaction_kthread);
1496         btrfs_remount_cleanup(fs_info, old_opts);
1497         return 0;
1498
1499 restore:
1500         /* We've hit an error - don't reset MS_RDONLY */
1501         if (sb->s_flags & MS_RDONLY)
1502                 old_flags |= MS_RDONLY;
1503         sb->s_flags = old_flags;
1504         fs_info->mount_opt = old_opts;
1505         fs_info->compress_type = old_compress_type;
1506         fs_info->max_inline = old_max_inline;
1507         mutex_lock(&fs_info->chunk_mutex);
1508         fs_info->alloc_start = old_alloc_start;
1509         mutex_unlock(&fs_info->chunk_mutex);
1510         btrfs_resize_thread_pool(fs_info,
1511                 old_thread_pool_size, fs_info->thread_pool_size);
1512         fs_info->metadata_ratio = old_metadata_ratio;
1513         btrfs_remount_cleanup(fs_info, old_opts);
1514         return ret;
1515 }
1516
1517 /* Used to sort the devices by max_avail(descending sort) */
1518 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1519                                        const void *dev_info2)
1520 {
1521         if (((struct btrfs_device_info *)dev_info1)->max_avail >
1522             ((struct btrfs_device_info *)dev_info2)->max_avail)
1523                 return -1;
1524         else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1525                  ((struct btrfs_device_info *)dev_info2)->max_avail)
1526                 return 1;
1527         else
1528         return 0;
1529 }
1530
1531 /*
1532  * sort the devices by max_avail, in which max free extent size of each device
1533  * is stored.(Descending Sort)
1534  */
1535 static inline void btrfs_descending_sort_devices(
1536                                         struct btrfs_device_info *devices,
1537                                         size_t nr_devices)
1538 {
1539         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1540              btrfs_cmp_device_free_bytes, NULL);
1541 }
1542
1543 /*
1544  * The helper to calc the free space on the devices that can be used to store
1545  * file data.
1546  */
1547 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1548 {
1549         struct btrfs_fs_info *fs_info = root->fs_info;
1550         struct btrfs_device_info *devices_info;
1551         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1552         struct btrfs_device *device;
1553         u64 skip_space;
1554         u64 type;
1555         u64 avail_space;
1556         u64 used_space;
1557         u64 min_stripe_size;
1558         int min_stripes = 1, num_stripes = 1;
1559         int i = 0, nr_devices;
1560         int ret;
1561
1562         nr_devices = fs_info->fs_devices->open_devices;
1563         BUG_ON(!nr_devices);
1564
1565         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1566                                GFP_NOFS);
1567         if (!devices_info)
1568                 return -ENOMEM;
1569
1570         /* calc min stripe number for data space alloction */
1571         type = btrfs_get_alloc_profile(root, 1);
1572         if (type & BTRFS_BLOCK_GROUP_RAID0) {
1573                 min_stripes = 2;
1574                 num_stripes = nr_devices;
1575         } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1576                 min_stripes = 2;
1577                 num_stripes = 2;
1578         } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1579                 min_stripes = 4;
1580                 num_stripes = 4;
1581         }
1582
1583         if (type & BTRFS_BLOCK_GROUP_DUP)
1584                 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1585         else
1586                 min_stripe_size = BTRFS_STRIPE_LEN;
1587
1588         list_for_each_entry(device, &fs_devices->devices, dev_list) {
1589                 if (!device->in_fs_metadata || !device->bdev ||
1590                     device->is_tgtdev_for_dev_replace)
1591                         continue;
1592
1593                 avail_space = device->total_bytes - device->bytes_used;
1594
1595                 /* align with stripe_len */
1596                 do_div(avail_space, BTRFS_STRIPE_LEN);
1597                 avail_space *= BTRFS_STRIPE_LEN;
1598
1599                 /*
1600                  * In order to avoid overwritting the superblock on the drive,
1601                  * btrfs starts at an offset of at least 1MB when doing chunk
1602                  * allocation.
1603                  */
1604                 skip_space = 1024 * 1024;
1605
1606                 /* user can set the offset in fs_info->alloc_start. */
1607                 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1608                     device->total_bytes)
1609                         skip_space = max(fs_info->alloc_start, skip_space);
1610
1611                 /*
1612                  * btrfs can not use the free space in [0, skip_space - 1],
1613                  * we must subtract it from the total. In order to implement
1614                  * it, we account the used space in this range first.
1615                  */
1616                 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1617                                                      &used_space);
1618                 if (ret) {
1619                         kfree(devices_info);
1620                         return ret;
1621                 }
1622
1623                 /* calc the free space in [0, skip_space - 1] */
1624                 skip_space -= used_space;
1625
1626                 /*
1627                  * we can use the free space in [0, skip_space - 1], subtract
1628                  * it from the total.
1629                  */
1630                 if (avail_space && avail_space >= skip_space)
1631                         avail_space -= skip_space;
1632                 else
1633                         avail_space = 0;
1634
1635                 if (avail_space < min_stripe_size)
1636                         continue;
1637
1638                 devices_info[i].dev = device;
1639                 devices_info[i].max_avail = avail_space;
1640
1641                 i++;
1642         }
1643
1644         nr_devices = i;
1645
1646         btrfs_descending_sort_devices(devices_info, nr_devices);
1647
1648         i = nr_devices - 1;
1649         avail_space = 0;
1650         while (nr_devices >= min_stripes) {
1651                 if (num_stripes > nr_devices)
1652                         num_stripes = nr_devices;
1653
1654                 if (devices_info[i].max_avail >= min_stripe_size) {
1655                         int j;
1656                         u64 alloc_size;
1657
1658                         avail_space += devices_info[i].max_avail * num_stripes;
1659                         alloc_size = devices_info[i].max_avail;
1660                         for (j = i + 1 - num_stripes; j <= i; j++)
1661                                 devices_info[j].max_avail -= alloc_size;
1662                 }
1663                 i--;
1664                 nr_devices--;
1665         }
1666
1667         kfree(devices_info);
1668         *free_bytes = avail_space;
1669         return 0;
1670 }
1671
1672 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1673 {
1674         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1675         struct btrfs_super_block *disk_super = fs_info->super_copy;
1676         struct list_head *head = &fs_info->space_info;
1677         struct btrfs_space_info *found;
1678         u64 total_used = 0;
1679         u64 total_free_data = 0;
1680         int bits = dentry->d_sb->s_blocksize_bits;
1681         __be32 *fsid = (__be32 *)fs_info->fsid;
1682         int ret;
1683
1684         /* holding chunk_muext to avoid allocating new chunks */
1685         mutex_lock(&fs_info->chunk_mutex);
1686         rcu_read_lock();
1687         list_for_each_entry_rcu(found, head, list) {
1688                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1689                         total_free_data += found->disk_total - found->disk_used;
1690                         total_free_data -=
1691                                 btrfs_account_ro_block_groups_free_space(found);
1692                 }
1693
1694                 total_used += found->disk_used;
1695         }
1696         rcu_read_unlock();
1697
1698         buf->f_namelen = BTRFS_NAME_LEN;
1699         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1700         buf->f_bfree = buf->f_blocks - (total_used >> bits);
1701         buf->f_bsize = dentry->d_sb->s_blocksize;
1702         buf->f_type = BTRFS_SUPER_MAGIC;
1703         buf->f_bavail = total_free_data;
1704         ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1705         if (ret) {
1706                 mutex_unlock(&fs_info->chunk_mutex);
1707                 return ret;
1708         }
1709         buf->f_bavail += total_free_data;
1710         buf->f_bavail = buf->f_bavail >> bits;
1711         mutex_unlock(&fs_info->chunk_mutex);
1712
1713         /* We treat it as constant endianness (it doesn't matter _which_)
1714            because we want the fsid to come out the same whether mounted
1715            on a big-endian or little-endian host */
1716         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1717         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1718         /* Mask in the root object ID too, to disambiguate subvols */
1719         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1720         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1721
1722         return 0;
1723 }
1724
1725 static void btrfs_kill_super(struct super_block *sb)
1726 {
1727         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1728         kill_anon_super(sb);
1729         free_fs_info(fs_info);
1730 }
1731
1732 static struct file_system_type btrfs_fs_type = {
1733         .owner          = THIS_MODULE,
1734         .name           = "btrfs",
1735         .mount          = btrfs_mount,
1736         .kill_sb        = btrfs_kill_super,
1737         .fs_flags       = FS_REQUIRES_DEV,
1738 };
1739 MODULE_ALIAS_FS("btrfs");
1740
1741 /*
1742  * used by btrfsctl to scan devices when no FS is mounted
1743  */
1744 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1745                                 unsigned long arg)
1746 {
1747         struct btrfs_ioctl_vol_args *vol;
1748         struct btrfs_fs_devices *fs_devices;
1749         int ret = -ENOTTY;
1750
1751         if (!capable(CAP_SYS_ADMIN))
1752                 return -EPERM;
1753
1754         vol = memdup_user((void __user *)arg, sizeof(*vol));
1755         if (IS_ERR(vol))
1756                 return PTR_ERR(vol);
1757
1758         switch (cmd) {
1759         case BTRFS_IOC_SCAN_DEV:
1760                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1761                                             &btrfs_fs_type, &fs_devices);
1762                 break;
1763         case BTRFS_IOC_DEVICES_READY:
1764                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1765                                             &btrfs_fs_type, &fs_devices);
1766                 if (ret)
1767                         break;
1768                 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1769                 break;
1770         }
1771
1772         kfree(vol);
1773         return ret;
1774 }
1775
1776 static int btrfs_freeze(struct super_block *sb)
1777 {
1778         struct btrfs_trans_handle *trans;
1779         struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1780
1781         trans = btrfs_attach_transaction_barrier(root);
1782         if (IS_ERR(trans)) {
1783                 /* no transaction, don't bother */
1784                 if (PTR_ERR(trans) == -ENOENT)
1785                         return 0;
1786                 return PTR_ERR(trans);
1787         }
1788         return btrfs_commit_transaction(trans, root);
1789 }
1790
1791 static int btrfs_unfreeze(struct super_block *sb)
1792 {
1793         return 0;
1794 }
1795
1796 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1797 {
1798         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1799         struct btrfs_fs_devices *cur_devices;
1800         struct btrfs_device *dev, *first_dev = NULL;
1801         struct list_head *head;
1802         struct rcu_string *name;
1803
1804         mutex_lock(&fs_info->fs_devices->device_list_mutex);
1805         cur_devices = fs_info->fs_devices;
1806         while (cur_devices) {
1807                 head = &cur_devices->devices;
1808                 list_for_each_entry(dev, head, dev_list) {
1809                         if (dev->missing)
1810                                 continue;
1811                         if (!first_dev || dev->devid < first_dev->devid)
1812                                 first_dev = dev;
1813                 }
1814                 cur_devices = cur_devices->seed;
1815         }
1816
1817         if (first_dev) {
1818                 rcu_read_lock();
1819                 name = rcu_dereference(first_dev->name);
1820                 seq_escape(m, name->str, " \t\n\\");
1821                 rcu_read_unlock();
1822         } else {
1823                 WARN_ON(1);
1824         }
1825         mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1826         return 0;
1827 }
1828
1829 static const struct super_operations btrfs_super_ops = {
1830         .drop_inode     = btrfs_drop_inode,
1831         .evict_inode    = btrfs_evict_inode,
1832         .put_super      = btrfs_put_super,
1833         .sync_fs        = btrfs_sync_fs,
1834         .show_options   = btrfs_show_options,
1835         .show_devname   = btrfs_show_devname,
1836         .write_inode    = btrfs_write_inode,
1837         .alloc_inode    = btrfs_alloc_inode,
1838         .destroy_inode  = btrfs_destroy_inode,
1839         .statfs         = btrfs_statfs,
1840         .remount_fs     = btrfs_remount,
1841         .freeze_fs      = btrfs_freeze,
1842         .unfreeze_fs    = btrfs_unfreeze,
1843 };
1844
1845 static const struct file_operations btrfs_ctl_fops = {
1846         .unlocked_ioctl  = btrfs_control_ioctl,
1847         .compat_ioctl = btrfs_control_ioctl,
1848         .owner   = THIS_MODULE,
1849         .llseek = noop_llseek,
1850 };
1851
1852 static struct miscdevice btrfs_misc = {
1853         .minor          = BTRFS_MINOR,
1854         .name           = "btrfs-control",
1855         .fops           = &btrfs_ctl_fops
1856 };
1857
1858 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1859 MODULE_ALIAS("devname:btrfs-control");
1860
1861 static int btrfs_interface_init(void)
1862 {
1863         return misc_register(&btrfs_misc);
1864 }
1865
1866 static void btrfs_interface_exit(void)
1867 {
1868         if (misc_deregister(&btrfs_misc) < 0)
1869                 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
1870 }
1871
1872 static void btrfs_print_info(void)
1873 {
1874         printk(KERN_INFO "Btrfs loaded"
1875 #ifdef CONFIG_BTRFS_DEBUG
1876                         ", debug=on"
1877 #endif
1878 #ifdef CONFIG_BTRFS_ASSERT
1879                         ", assert=on"
1880 #endif
1881 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1882                         ", integrity-checker=on"
1883 #endif
1884                         "\n");
1885 }
1886
1887 static int btrfs_run_sanity_tests(void)
1888 {
1889         int ret;
1890
1891         ret = btrfs_init_test_fs();
1892         if (ret)
1893                 return ret;
1894
1895         ret = btrfs_test_free_space_cache();
1896         if (ret)
1897                 goto out;
1898         ret = btrfs_test_extent_buffer_operations();
1899         if (ret)
1900                 goto out;
1901         ret = btrfs_test_extent_io();
1902         if (ret)
1903                 goto out;
1904         ret = btrfs_test_inodes();
1905         if (ret)
1906                 goto out;
1907         ret = btrfs_test_qgroups();
1908 out:
1909         btrfs_destroy_test_fs();
1910         return ret;
1911 }
1912
1913 static int __init init_btrfs_fs(void)
1914 {
1915         int err;
1916
1917         err = btrfs_hash_init();
1918         if (err)
1919                 return err;
1920
1921         btrfs_props_init();
1922
1923         err = btrfs_init_sysfs();
1924         if (err)
1925                 goto free_hash;
1926
1927         btrfs_init_compress();
1928
1929         err = btrfs_init_cachep();
1930         if (err)
1931                 goto free_compress;
1932
1933         err = extent_io_init();
1934         if (err)
1935                 goto free_cachep;
1936
1937         err = extent_map_init();
1938         if (err)
1939                 goto free_extent_io;
1940
1941         err = ordered_data_init();
1942         if (err)
1943                 goto free_extent_map;
1944
1945         err = btrfs_delayed_inode_init();
1946         if (err)
1947                 goto free_ordered_data;
1948
1949         err = btrfs_auto_defrag_init();
1950         if (err)
1951                 goto free_delayed_inode;
1952
1953         err = btrfs_delayed_ref_init();
1954         if (err)
1955                 goto free_auto_defrag;
1956
1957         err = btrfs_prelim_ref_init();
1958         if (err)
1959                 goto free_prelim_ref;
1960
1961         err = btrfs_interface_init();
1962         if (err)
1963                 goto free_delayed_ref;
1964
1965         btrfs_init_lockdep();
1966
1967         btrfs_print_info();
1968
1969         err = btrfs_run_sanity_tests();
1970         if (err)
1971                 goto unregister_ioctl;
1972
1973         err = register_filesystem(&btrfs_fs_type);
1974         if (err)
1975                 goto unregister_ioctl;
1976
1977         return 0;
1978
1979 unregister_ioctl:
1980         btrfs_interface_exit();
1981 free_prelim_ref:
1982         btrfs_prelim_ref_exit();
1983 free_delayed_ref:
1984         btrfs_delayed_ref_exit();
1985 free_auto_defrag:
1986         btrfs_auto_defrag_exit();
1987 free_delayed_inode:
1988         btrfs_delayed_inode_exit();
1989 free_ordered_data:
1990         ordered_data_exit();
1991 free_extent_map:
1992         extent_map_exit();
1993 free_extent_io:
1994         extent_io_exit();
1995 free_cachep:
1996         btrfs_destroy_cachep();
1997 free_compress:
1998         btrfs_exit_compress();
1999         btrfs_exit_sysfs();
2000 free_hash:
2001         btrfs_hash_exit();
2002         return err;
2003 }
2004
2005 static void __exit exit_btrfs_fs(void)
2006 {
2007         btrfs_destroy_cachep();
2008         btrfs_delayed_ref_exit();
2009         btrfs_auto_defrag_exit();
2010         btrfs_delayed_inode_exit();
2011         btrfs_prelim_ref_exit();
2012         ordered_data_exit();
2013         extent_map_exit();
2014         extent_io_exit();
2015         btrfs_interface_exit();
2016         unregister_filesystem(&btrfs_fs_type);
2017         btrfs_exit_sysfs();
2018         btrfs_cleanup_fs_uuids();
2019         btrfs_exit_compress();
2020         btrfs_hash_exit();
2021 }
2022
2023 late_initcall(init_btrfs_fs);
2024 module_exit(exit_btrfs_fs)
2025
2026 MODULE_LICENSE("GPL");