Pileus Git - ~andy/linux/blob - fs/btrfs/super.c

   1 /*
   2  * Copyright (C) 2007 Oracle.  All rights reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public
   6  * License v2 as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public
  14  * License along with this program; if not, write to the
  15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  16  * Boston, MA 021110-1307, USA.
  17  */
  18
  19 #include <linux/blkdev.h>
  20 #include <linux/module.h>
  21 #include <linux/buffer_head.h>
  22 #include <linux/fs.h>
  23 #include <linux/pagemap.h>
  24 #include <linux/highmem.h>
  25 #include <linux/time.h>
  26 #include <linux/init.h>
  27 #include <linux/seq_file.h>
  28 #include <linux/string.h>
  29 #include <linux/backing-dev.h>
  30 #include <linux/mount.h>
  31 #include <linux/mpage.h>
  32 #include <linux/swap.h>
  33 #include <linux/writeback.h>
  34 #include <linux/statfs.h>
  35 #include <linux/compat.h>
  36 #include <linux/parser.h>
  37 #include <linux/ctype.h>
  38 #include <linux/namei.h>
  39 #include <linux/miscdevice.h>
  40 #include <linux/magic.h>
  41 #include <linux/slab.h>
  42 #include <linux/cleancache.h>
  43 #include <linux/ratelimit.h>
  44 #include <linux/btrfs.h>
  45 #include "delayed-inode.h"
  46 #include "ctree.h"
  47 #include "disk-io.h"
  48 #include "transaction.h"
  49 #include "btrfs_inode.h"
  50 #include "print-tree.h"
  51 #include "hash.h"
  52 #include "props.h"
  53 #include "xattr.h"
  54 #include "volumes.h"
  55 #include "export.h"
  56 #include "compression.h"
  57 #include "rcu-string.h"
  58 #include "dev-replace.h"
  59 #include "free-space-cache.h"
  60 #include "backref.h"
  61 #include "tests/btrfs-tests.h"
  62
  63 #define CREATE_TRACE_POINTS
  64 #include <trace/events/btrfs.h>
  65
  66 static const struct super_operations btrfs_super_ops;
  67 static struct file_system_type btrfs_fs_type;
  68
  69 static const char *btrfs_decode_error(int errno)
  70 {
  71         char *errstr = "unknown";
  72
  73         switch (errno) {
  74         case -EIO:
  75                 errstr = "IO failure";
  76                 break;
  77         case -ENOMEM:
  78                 errstr = "Out of memory";
  79                 break;
  80         case -EROFS:
  81                 errstr = "Readonly filesystem";
  82                 break;
  83         case -EEXIST:
  84                 errstr = "Object already exists";
  85                 break;
  86         case -ENOSPC:
  87                 errstr = "No space left";
  88                 break;
  89         case -ENOENT:
  90                 errstr = "No such entry";
  91                 break;
  92         }
  93
  94         return errstr;
  95 }
  96
  97 static void save_error_info(struct btrfs_fs_info *fs_info)
  98 {
  99         /*
 100          * today we only save the error info into ram.  Long term we'll
 101          * also send it down to the disk
 102          */
 103         set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
 104 }
 105
 106 /* btrfs handle error by forcing the filesystem readonly */
 107 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
 108 {
 109         struct super_block *sb = fs_info->sb;
 110
 111         if (sb->s_flags & MS_RDONLY)
 112                 return;
 113
 114         if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
 115                 sb->s_flags |= MS_RDONLY;
 116                 btrfs_info(fs_info, "forced readonly");
 117                 /*
 118                  * Note that a running device replace operation is not
 119                  * canceled here although there is no way to update
 120                  * the progress. It would add the risk of a deadlock,
 121                  * therefore the canceling is ommited. The only penalty
 122                  * is that some I/O remains active until the procedure
 123                  * completes. The next time when the filesystem is
 124                  * mounted writeable again, the device replace
 125                  * operation continues.
 126                  */
 127         }
 128 }
 129
 130 #ifdef CONFIG_PRINTK
 131 /*
 132  * __btrfs_std_error decodes expected errors from the caller and
 133  * invokes the approciate error response.
 134  */
 135 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 136                        unsigned int line, int errno, const char *fmt, ...)
 137 {
 138         struct super_block *sb = fs_info->sb;
 139         const char *errstr;
 140
 141         /*
 142          * Special case: if the error is EROFS, and we're already
 143          * under MS_RDONLY, then it is safe here.
 144          */
 145         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 146                 return;
 147
 148         errstr = btrfs_decode_error(errno);
 149         if (fmt) {
 150                 struct va_format vaf;
 151                 va_list args;
 152
 153                 va_start(args, fmt);
 154                 vaf.fmt = fmt;
 155                 vaf.va = &args;
 156
 157                 printk(KERN_CRIT
 158                         "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
 159                         sb->s_id, function, line, errno, errstr, &vaf);
 160                 va_end(args);
 161         } else {
 162                 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
 163                         sb->s_id, function, line, errno, errstr);
 164         }
 165
 166         /* Don't go through full error handling during mount */
 167         save_error_info(fs_info);
 168         if (sb->s_flags & MS_BORN)
 169                 btrfs_handle_error(fs_info);
 170 }
 171
 172 static const char * const logtypes[] = {
 173         "emergency",
 174         "alert",
 175         "critical",
 176         "error",
 177         "warning",
 178         "notice",
 179         "info",
 180         "debug",
 181 };
 182
 183 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
 184 {
 185         struct super_block *sb = fs_info->sb;
 186         char lvl[4];
 187         struct va_format vaf;
 188         va_list args;
 189         const char *type = logtypes[4];
 190         int kern_level;
 191
 192         va_start(args, fmt);
 193
 194         kern_level = printk_get_level(fmt);
 195         if (kern_level) {
 196                 size_t size = printk_skip_level(fmt) - fmt;
 197                 memcpy(lvl, fmt,  size);
 198                 lvl[size] = '\0';
 199                 fmt += size;
 200                 type = logtypes[kern_level - '0'];
 201         } else
 202                 *lvl = '\0';
 203
 204         vaf.fmt = fmt;
 205         vaf.va = &args;
 206
 207         printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
 208
 209         va_end(args);
 210 }
 211
 212 #else
 213
 214 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 215                        unsigned int line, int errno, const char *fmt, ...)
 216 {
 217         struct super_block *sb = fs_info->sb;
 218
 219         /*
 220          * Special case: if the error is EROFS, and we're already
 221          * under MS_RDONLY, then it is safe here.
 222          */
 223         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 224                 return;
 225
 226         /* Don't go through full error handling during mount */
 227         if (sb->s_flags & MS_BORN) {
 228                 save_error_info(fs_info);
 229                 btrfs_handle_error(fs_info);
 230         }
 231 }
 232 #endif
 233
 234 /*
 235  * We only mark the transaction aborted and then set the file system read-only.
 236  * This will prevent new transactions from starting or trying to join this
 237  * one.
 238  *
 239  * This means that error recovery at the call site is limited to freeing
 240  * any local memory allocations and passing the error code up without
 241  * further cleanup. The transaction should complete as it normally would
 242  * in the call path but will return -EIO.
 243  *
 244  * We'll complete the cleanup in btrfs_end_transaction and
 245  * btrfs_commit_transaction.
 246  */
 247 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
 248                                struct btrfs_root *root, const char *function,
 249                                unsigned int line, int errno)
 250 {
 251         /*
 252          * Report first abort since mount
 253          */
 254         if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,
 255                                 &root->fs_info->fs_state)) {
 256                 WARN(1, KERN_DEBUG "BTRFS: Transaction aborted (error %d)\n",
 257                                 errno);
 258         }
 259         trans->aborted = errno;
 260         /* Nothing used. The other threads that have joined this
 261          * transaction may be able to continue. */
 262         if (!trans->blocks_used) {
 263                 const char *errstr;
 264
 265                 errstr = btrfs_decode_error(errno);
 266                 btrfs_warn(root->fs_info,
 267                            "%s:%d: Aborting unused transaction(%s).",
 268                            function, line, errstr);
 269                 return;
 270         }
 271         ACCESS_ONCE(trans->transaction->aborted) = errno;
 272         /* Wake up anybody who may be waiting on this transaction */
 273         wake_up(&root->fs_info->transaction_wait);
 274         wake_up(&root->fs_info->transaction_blocked_wait);
 275         __btrfs_std_error(root->fs_info, function, line, errno, NULL);
 276 }
 277 /*
 278  * __btrfs_panic decodes unexpected, fatal errors from the caller,
 279  * issues an alert, and either panics or BUGs, depending on mount options.
 280  */
 281 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
 282                    unsigned int line, int errno, const char *fmt, ...)
 283 {
 284         char *s_id = "<unknown>";
 285         const char *errstr;
 286         struct va_format vaf = { .fmt = fmt };
 287         va_list args;
 288
 289         if (fs_info)
 290                 s_id = fs_info->sb->s_id;
 291
 292         va_start(args, fmt);
 293         vaf.va = &args;
 294
 295         errstr = btrfs_decode_error(errno);
 296         if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
 297                 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
 298                         s_id, function, line, &vaf, errno, errstr);
 299
 300         btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
 301                    function, line, &vaf, errno, errstr);
 302         va_end(args);
 303         /* Caller calls BUG() */
 304 }
 305
 306 static void btrfs_put_super(struct super_block *sb)
 307 {
 308         (void)close_ctree(btrfs_sb(sb)->tree_root);
 309         /* FIXME: need to fix VFS to return error? */
 310         /* AV: return it _where_?  ->put_super() can be triggered by any number
 311          * of async events, up to and including delivery of SIGKILL to the
 312          * last process that kept it busy.  Or segfault in the aforementioned
 313          * process...  Whom would you report that to?
 314          */
 315 }
 316
 317 enum {
 318         Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
 319         Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
 320         Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
 321         Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
 322         Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
 323         Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
 324         Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
 325         Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
 326         Opt_check_integrity, Opt_check_integrity_including_extent_data,
 327         Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
 328         Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
 329         Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
 330         Opt_datasum, Opt_treelog, Opt_noinode_cache,
 331         Opt_err,
 332 };
 333
 334 static match_table_t tokens = {
 335         {Opt_degraded, "degraded"},
 336         {Opt_subvol, "subvol=%s"},
 337         {Opt_subvolid, "subvolid=%s"},
 338         {Opt_device, "device=%s"},
 339         {Opt_nodatasum, "nodatasum"},
 340         {Opt_datasum, "datasum"},
 341         {Opt_nodatacow, "nodatacow"},
 342         {Opt_datacow, "datacow"},
 343         {Opt_nobarrier, "nobarrier"},
 344         {Opt_barrier, "barrier"},
 345         {Opt_max_inline, "max_inline=%s"},
 346         {Opt_alloc_start, "alloc_start=%s"},
 347         {Opt_thread_pool, "thread_pool=%d"},
 348         {Opt_compress, "compress"},
 349         {Opt_compress_type, "compress=%s"},
 350         {Opt_compress_force, "compress-force"},
 351         {Opt_compress_force_type, "compress-force=%s"},
 352         {Opt_ssd, "ssd"},
 353         {Opt_ssd_spread, "ssd_spread"},
 354         {Opt_nossd, "nossd"},
 355         {Opt_acl, "acl"},
 356         {Opt_noacl, "noacl"},
 357         {Opt_notreelog, "notreelog"},
 358         {Opt_treelog, "treelog"},
 359         {Opt_flushoncommit, "flushoncommit"},
 360         {Opt_noflushoncommit, "noflushoncommit"},
 361         {Opt_ratio, "metadata_ratio=%d"},
 362         {Opt_discard, "discard"},
 363         {Opt_nodiscard, "nodiscard"},
 364         {Opt_space_cache, "space_cache"},
 365         {Opt_clear_cache, "clear_cache"},
 366         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 367         {Opt_enospc_debug, "enospc_debug"},
 368         {Opt_noenospc_debug, "noenospc_debug"},
 369         {Opt_subvolrootid, "subvolrootid=%d"},
 370         {Opt_defrag, "autodefrag"},
 371         {Opt_nodefrag, "noautodefrag"},
 372         {Opt_inode_cache, "inode_cache"},
 373         {Opt_noinode_cache, "noinode_cache"},
 374         {Opt_no_space_cache, "nospace_cache"},
 375         {Opt_recovery, "recovery"},
 376         {Opt_skip_balance, "skip_balance"},
 377         {Opt_check_integrity, "check_int"},
 378         {Opt_check_integrity_including_extent_data, "check_int_data"},
 379         {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
 380         {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 381         {Opt_fatal_errors, "fatal_errors=%s"},
 382         {Opt_commit_interval, "commit=%d"},
 383         {Opt_err, NULL},
 384 };
 385
 386 #define btrfs_set_and_info(root, opt, fmt, args...)                     \
 387 {                                                                       \
 388         if (!btrfs_test_opt(root, opt))                                 \
 389                 btrfs_info(root->fs_info, fmt, ##args);                 \
 390         btrfs_set_opt(root->fs_info->mount_opt, opt);                   \
 391 }
 392
 393 #define btrfs_clear_and_info(root, opt, fmt, args...)                   \
 394 {                                                                       \
 395         if (btrfs_test_opt(root, opt))                                  \
 396                 btrfs_info(root->fs_info, fmt, ##args);                 \
 397         btrfs_clear_opt(root->fs_info->mount_opt, opt);                 \
 398 }
 399
 400 /*
 401  * Regular mount options parser.  Everything that is needed only when
 402  * reading in a new superblock is parsed here.
 403  * XXX JDM: This needs to be cleaned up for remount.
 404  */
 405 int btrfs_parse_options(struct btrfs_root *root, char *options)
 406 {
 407         struct btrfs_fs_info *info = root->fs_info;
 408         substring_t args[MAX_OPT_ARGS];
 409         char *p, *num, *orig = NULL;
 410         u64 cache_gen;
 411         int intarg;
 412         int ret = 0;
 413         char *compress_type;
 414         bool compress_force = false;
 415         bool compress = false;
 416
 417         cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
 418         if (cache_gen)
 419                 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 420
 421         if (!options)
 422                 goto out;
 423
 424         /*
 425          * strsep changes the string, duplicate it because parse_options
 426          * gets called twice
 427          */
 428         options = kstrdup(options, GFP_NOFS);
 429         if (!options)
 430                 return -ENOMEM;
 431
 432         orig = options;
 433
 434         while ((p = strsep(&options, ",")) != NULL) {
 435                 int token;
 436                 if (!*p)
 437                         continue;
 438
 439                 token = match_token(p, tokens, args);
 440                 switch (token) {
 441                 case Opt_degraded:
 442                         btrfs_info(root->fs_info, "allowing degraded mounts");
 443                         btrfs_set_opt(info->mount_opt, DEGRADED);
 444                         break;
 445                 case Opt_subvol:
 446                 case Opt_subvolid:
 447                 case Opt_subvolrootid:
 448                 case Opt_device:
 449                         /*
 450                          * These are parsed by btrfs_parse_early_options
 451                          * and can be happily ignored here.
 452                          */
 453                         break;
 454                 case Opt_nodatasum:
 455                         btrfs_set_and_info(root, NODATASUM,
 456                                            "setting nodatasum");
 457                         break;
 458                 case Opt_datasum:
 459                         if (btrfs_test_opt(root, NODATASUM)) {
 460                                 if (btrfs_test_opt(root, NODATACOW))
 461                                         btrfs_info(root->fs_info, "setting datasum, datacow enabled");
 462                                 else
 463                                         btrfs_info(root->fs_info, "setting datasum");
 464                         }
 465                         btrfs_clear_opt(info->mount_opt, NODATACOW);
 466                         btrfs_clear_opt(info->mount_opt, NODATASUM);
 467                         break;
 468                 case Opt_nodatacow:
 469                         if (!btrfs_test_opt(root, NODATACOW)) {
 470                                 if (!btrfs_test_opt(root, COMPRESS) ||
 471                                     !btrfs_test_opt(root, FORCE_COMPRESS)) {
 472                                         btrfs_info(root->fs_info,
 473                                                    "setting nodatacow, compression disabled");
 474                                 } else {
 475                                         btrfs_info(root->fs_info, "setting nodatacow");
 476                                 }
 477                         }
 478                         btrfs_clear_opt(info->mount_opt, COMPRESS);
 479                         btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 480                         btrfs_set_opt(info->mount_opt, NODATACOW);
 481                         btrfs_set_opt(info->mount_opt, NODATASUM);
 482                         break;
 483                 case Opt_datacow:
 484                         btrfs_clear_and_info(root, NODATACOW,
 485                                              "setting datacow");
 486                         break;
 487                 case Opt_compress_force:
 488                 case Opt_compress_force_type:
 489                         compress_force = true;
 490                         /* Fallthrough */
 491                 case Opt_compress:
 492                 case Opt_compress_type:
 493                         compress = true;
 494                         if (token == Opt_compress ||
 495                             token == Opt_compress_force ||
 496                             strcmp(args[0].from, "zlib") == 0) {
 497                                 compress_type = "zlib";
 498                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 499                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 500                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 501                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 502                         } else if (strcmp(args[0].from, "lzo") == 0) {
 503                                 compress_type = "lzo";
 504                                 info->compress_type = BTRFS_COMPRESS_LZO;
 505                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 506                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 507                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 508                                 btrfs_set_fs_incompat(info, COMPRESS_LZO);
 509                         } else if (strncmp(args[0].from, "no", 2) == 0) {
 510                                 compress_type = "no";
 511                                 btrfs_clear_opt(info->mount_opt, COMPRESS);
 512                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 513                                 compress_force = false;
 514                         } else {
 515                                 ret = -EINVAL;
 516                                 goto out;
 517                         }
 518
 519                         if (compress_force) {
 520                                 btrfs_set_and_info(root, FORCE_COMPRESS,
 521                                                    "force %s compression",
 522                                                    compress_type);
 523                         } else if (compress) {
 524                                 if (!btrfs_test_opt(root, COMPRESS))
 525                                         btrfs_info(root->fs_info,
 526                                                    "btrfs: use %s compression\n",
 527                                                    compress_type);
 528                         }
 529                         break;
 530                 case Opt_ssd:
 531                         btrfs_set_and_info(root, SSD,
 532                                            "use ssd allocation scheme");
 533                         break;
 534                 case Opt_ssd_spread:
 535                         btrfs_set_and_info(root, SSD_SPREAD,
 536                                            "use spread ssd allocation scheme");
 537                         break;
 538                 case Opt_nossd:
 539                         btrfs_clear_and_info(root, NOSSD,
 540                                              "not using ssd allocation scheme");
 541                         btrfs_clear_opt(info->mount_opt, SSD);
 542                         break;
 543                 case Opt_barrier:
 544                         btrfs_clear_and_info(root, NOBARRIER,
 545                                              "turning on barriers");
 546                         break;
 547                 case Opt_nobarrier:
 548                         btrfs_set_and_info(root, NOBARRIER,
 549                                            "turning off barriers");
 550                         break;
 551                 case Opt_thread_pool:
 552                         ret = match_int(&args[0], &intarg);
 553                         if (ret) {
 554                                 goto out;
 555                         } else if (intarg > 0) {
 556                                 info->thread_pool_size = intarg;
 557                         } else {
 558                                 ret = -EINVAL;
 559                                 goto out;
 560                         }
 561                         break;
 562                 case Opt_max_inline:
 563                         num = match_strdup(&args[0]);
 564                         if (num) {
 565                                 info->max_inline = memparse(num, NULL);
 566                                 kfree(num);
 567
 568                                 if (info->max_inline) {
 569                                         info->max_inline = min_t(u64,
 570                                                 info->max_inline,
 571                                                 root->sectorsize);
 572                                 }
 573                                 btrfs_info(root->fs_info, "max_inline at %llu",
 574                                         info->max_inline);
 575                         } else {
 576                                 ret = -ENOMEM;
 577                                 goto out;
 578                         }
 579                         break;
 580                 case Opt_alloc_start:
 581                         num = match_strdup(&args[0]);
 582                         if (num) {
 583                                 mutex_lock(&info->chunk_mutex);
 584                                 info->alloc_start = memparse(num, NULL);
 585                                 mutex_unlock(&info->chunk_mutex);
 586                                 kfree(num);
 587                                 btrfs_info(root->fs_info, "allocations start at %llu",
 588                                         info->alloc_start);
 589                         } else {
 590                                 ret = -ENOMEM;
 591                                 goto out;
 592                         }
 593                         break;
 594                 case Opt_acl:
 595                         root->fs_info->sb->s_flags |= MS_POSIXACL;
 596                         break;
 597                 case Opt_noacl:
 598                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 599                         break;
 600                 case Opt_notreelog:
 601                         btrfs_set_and_info(root, NOTREELOG,
 602                                            "disabling tree log");
 603                         break;
 604                 case Opt_treelog:
 605                         btrfs_clear_and_info(root, NOTREELOG,
 606                                              "enabling tree log");
 607                         break;
 608                 case Opt_flushoncommit:
 609                         btrfs_set_and_info(root, FLUSHONCOMMIT,
 610                                            "turning on flush-on-commit");
 611                         break;
 612                 case Opt_noflushoncommit:
 613                         btrfs_clear_and_info(root, FLUSHONCOMMIT,
 614                                              "turning off flush-on-commit");
 615                         break;
 616                 case Opt_ratio:
 617                         ret = match_int(&args[0], &intarg);
 618                         if (ret) {
 619                                 goto out;
 620                         } else if (intarg >= 0) {
 621                                 info->metadata_ratio = intarg;
 622                                 btrfs_info(root->fs_info, "metadata ratio %d",
 623                                        info->metadata_ratio);
 624                         } else {
 625                                 ret = -EINVAL;
 626                                 goto out;
 627                         }
 628                         break;
 629                 case Opt_discard:
 630                         btrfs_set_and_info(root, DISCARD,
 631                                            "turning on discard");
 632                         break;
 633                 case Opt_nodiscard:
 634                         btrfs_clear_and_info(root, DISCARD,
 635                                              "turning off discard");
 636                         break;
 637                 case Opt_space_cache:
 638                         btrfs_set_and_info(root, SPACE_CACHE,
 639                                            "enabling disk space caching");
 640                         break;
 641                 case Opt_rescan_uuid_tree:
 642                         btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 643                         break;
 644                 case Opt_no_space_cache:
 645                         btrfs_clear_and_info(root, SPACE_CACHE,
 646                                              "disabling disk space caching");
 647                         break;
 648                 case Opt_inode_cache:
 649                         btrfs_set_and_info(root, CHANGE_INODE_CACHE,
 650                                            "enabling inode map caching");
 651                         break;
 652                 case Opt_noinode_cache:
 653                         btrfs_clear_and_info(root, CHANGE_INODE_CACHE,
 654                                              "disabling inode map caching");
 655                         break;
 656                 case Opt_clear_cache:
 657                         btrfs_set_and_info(root, CLEAR_CACHE,
 658                                            "force clearing of disk cache");
 659                         break;
 660                 case Opt_user_subvol_rm_allowed:
 661                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 662                         break;
 663                 case Opt_enospc_debug:
 664                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 665                         break;
 666                 case Opt_noenospc_debug:
 667                         btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
 668                         break;
 669                 case Opt_defrag:
 670                         btrfs_set_and_info(root, AUTO_DEFRAG,
 671                                            "enabling auto defrag");
 672                         break;
 673                 case Opt_nodefrag:
 674                         btrfs_clear_and_info(root, AUTO_DEFRAG,
 675                                              "disabling auto defrag");
 676                         break;
 677                 case Opt_recovery:
 678                         btrfs_info(root->fs_info, "enabling auto recovery");
 679                         btrfs_set_opt(info->mount_opt, RECOVERY);
 680                         break;
 681                 case Opt_skip_balance:
 682                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 683                         break;
 684 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 685                 case Opt_check_integrity_including_extent_data:
 686                         btrfs_info(root->fs_info,
 687                                    "enabling check integrity including extent data");
 688                         btrfs_set_opt(info->mount_opt,
 689                                       CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
 690                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 691                         break;
 692                 case Opt_check_integrity:
 693                         btrfs_info(root->fs_info, "enabling check integrity");
 694                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 695                         break;
 696                 case Opt_check_integrity_print_mask:
 697                         ret = match_int(&args[0], &intarg);
 698                         if (ret) {
 699                                 goto out;
 700                         } else if (intarg >= 0) {
 701                                 info->check_integrity_print_mask = intarg;
 702                                 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
 703                                        info->check_integrity_print_mask);
 704                         } else {
 705                                 ret = -EINVAL;
 706                                 goto out;
 707                         }
 708                         break;
 709 #else
 710                 case Opt_check_integrity_including_extent_data:
 711                 case Opt_check_integrity:
 712                 case Opt_check_integrity_print_mask:
 713                         btrfs_err(root->fs_info,
 714                                 "support for check_integrity* not compiled in!");
 715                         ret = -EINVAL;
 716                         goto out;
 717 #endif
 718                 case Opt_fatal_errors:
 719                         if (strcmp(args[0].from, "panic") == 0)
 720                                 btrfs_set_opt(info->mount_opt,
 721                                               PANIC_ON_FATAL_ERROR);
 722                         else if (strcmp(args[0].from, "bug") == 0)
 723                                 btrfs_clear_opt(info->mount_opt,
 724                                               PANIC_ON_FATAL_ERROR);
 725                         else {
 726                                 ret = -EINVAL;
 727                                 goto out;
 728                         }
 729                         break;
 730                 case Opt_commit_interval:
 731                         intarg = 0;
 732                         ret = match_int(&args[0], &intarg);
 733                         if (ret < 0) {
 734                                 btrfs_err(root->fs_info, "invalid commit interval");
 735                                 ret = -EINVAL;
 736                                 goto out;
 737                         }
 738                         if (intarg > 0) {
 739                                 if (intarg > 300) {
 740                                         btrfs_warn(root->fs_info, "excessive commit interval %d",
 741                                                         intarg);
 742                                 }
 743                                 info->commit_interval = intarg;
 744                         } else {
 745                                 btrfs_info(root->fs_info, "using default commit interval %ds",
 746                                     BTRFS_DEFAULT_COMMIT_INTERVAL);
 747                                 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
 748                         }
 749                         break;
 750                 case Opt_err:
 751                         btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
 752                         ret = -EINVAL;
 753                         goto out;
 754                 default:
 755                         break;
 756                 }
 757         }
 758 out:
 759         if (!ret && btrfs_test_opt(root, SPACE_CACHE))
 760                 btrfs_info(root->fs_info, "disk space caching is enabled");
 761         kfree(orig);
 762         return ret;
 763 }
 764
 765 /*
 766  * Parse mount options that are required early in the mount process.
 767  *
 768  * All other options will be parsed on much later in the mount process and
 769  * only when we need to allocate a new super block.
 770  */
 771 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 772                 void *holder, char **subvol_name, u64 *subvol_objectid,
 773                 struct btrfs_fs_devices **fs_devices)
 774 {
 775         substring_t args[MAX_OPT_ARGS];
 776         char *device_name, *opts, *orig, *p;
 777         char *num = NULL;
 778         int error = 0;
 779
 780         if (!options)
 781                 return 0;
 782
 783         /*
 784          * strsep changes the string, duplicate it because parse_options
 785          * gets called twice
 786          */
 787         opts = kstrdup(options, GFP_KERNEL);
 788         if (!opts)
 789                 return -ENOMEM;
 790         orig = opts;
 791
 792         while ((p = strsep(&opts, ",")) != NULL) {
 793                 int token;
 794                 if (!*p)
 795                         continue;
 796
 797                 token = match_token(p, tokens, args);
 798                 switch (token) {
 799                 case Opt_subvol:
 800                         kfree(*subvol_name);
 801                         *subvol_name = match_strdup(&args[0]);
 802                         if (!*subvol_name) {
 803                                 error = -ENOMEM;
 804                                 goto out;
 805                         }
 806                         break;
 807                 case Opt_subvolid:
 808                         num = match_strdup(&args[0]);
 809                         if (num) {
 810                                 *subvol_objectid = memparse(num, NULL);
 811                                 kfree(num);
 812                                 /* we want the original fs_tree */
 813                                 if (!*subvol_objectid)
 814                                         *subvol_objectid =
 815                                                 BTRFS_FS_TREE_OBJECTID;
 816                         } else {
 817                                 error = -EINVAL;
 818                                 goto out;
 819                         }
 820                         break;
 821                 case Opt_subvolrootid:
 822                         printk(KERN_WARNING
 823                                 "BTRFS: 'subvolrootid' mount option is deprecated and has "
 824                                 "no effect\n");
 825                         break;
 826                 case Opt_device:
 827                         device_name = match_strdup(&args[0]);
 828                         if (!device_name) {
 829                                 error = -ENOMEM;
 830                                 goto out;
 831                         }
 832                         error = btrfs_scan_one_device(device_name,
 833                                         flags, holder, fs_devices);
 834                         kfree(device_name);
 835                         if (error)
 836                                 goto out;
 837                         break;
 838                 default:
 839                         break;
 840                 }
 841         }
 842
 843 out:
 844         kfree(orig);
 845         return error;
 846 }
 847
 848 static struct dentry *get_default_root(struct super_block *sb,
 849                                        u64 subvol_objectid)
 850 {
 851         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 852         struct btrfs_root *root = fs_info->tree_root;
 853         struct btrfs_root *new_root;
 854         struct btrfs_dir_item *di;
 855         struct btrfs_path *path;
 856         struct btrfs_key location;
 857         struct inode *inode;
 858         struct dentry *dentry;
 859         u64 dir_id;
 860         int new = 0;
 861
 862         /*
 863          * We have a specific subvol we want to mount, just setup location and
 864          * go look up the root.
 865          */
 866         if (subvol_objectid) {
 867                 location.objectid = subvol_objectid;
 868                 location.type = BTRFS_ROOT_ITEM_KEY;
 869                 location.offset = (u64)-1;
 870                 goto find_root;
 871         }
 872
 873         path = btrfs_alloc_path();
 874         if (!path)
 875                 return ERR_PTR(-ENOMEM);
 876         path->leave_spinning = 1;
 877
 878         /*
 879          * Find the "default" dir item which points to the root item that we
 880          * will mount by default if we haven't been given a specific subvolume
 881          * to mount.
 882          */
 883         dir_id = btrfs_super_root_dir(fs_info->super_copy);
 884         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 885         if (IS_ERR(di)) {
 886                 btrfs_free_path(path);
 887                 return ERR_CAST(di);
 888         }
 889         if (!di) {
 890                 /*
 891                  * Ok the default dir item isn't there.  This is weird since
 892                  * it's always been there, but don't freak out, just try and
 893                  * mount to root most subvolume.
 894                  */
 895                 btrfs_free_path(path);
 896                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 897                 new_root = fs_info->fs_root;
 898                 goto setup_root;
 899         }
 900
 901         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 902         btrfs_free_path(path);
 903
 904 find_root:
 905         new_root = btrfs_read_fs_root_no_name(fs_info, &location);
 906         if (IS_ERR(new_root))
 907                 return ERR_CAST(new_root);
 908
 909         dir_id = btrfs_root_dirid(&new_root->root_item);
 910 setup_root:
 911         location.objectid = dir_id;
 912         location.type = BTRFS_INODE_ITEM_KEY;
 913         location.offset = 0;
 914
 915         inode = btrfs_iget(sb, &location, new_root, &new);
 916         if (IS_ERR(inode))
 917                 return ERR_CAST(inode);
 918
 919         /*
 920          * If we're just mounting the root most subvol put the inode and return
 921          * a reference to the dentry.  We will have already gotten a reference
 922          * to the inode in btrfs_fill_super so we're good to go.
 923          */
 924         if (!new && sb->s_root->d_inode == inode) {
 925                 iput(inode);
 926                 return dget(sb->s_root);
 927         }
 928
 929         dentry = d_obtain_alias(inode);
 930         if (!IS_ERR(dentry)) {
 931                 spin_lock(&dentry->d_lock);
 932                 dentry->d_flags &= ~DCACHE_DISCONNECTED;
 933                 spin_unlock(&dentry->d_lock);
 934         }
 935         return dentry;
 936 }
 937
 938 static int btrfs_fill_super(struct super_block *sb,
 939                             struct btrfs_fs_devices *fs_devices,
 940                             void *data, int silent)
 941 {
 942         struct inode *inode;
 943         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 944         struct btrfs_key key;
 945         int err;
 946
 947         sb->s_maxbytes = MAX_LFS_FILESIZE;
 948         sb->s_magic = BTRFS_SUPER_MAGIC;
 949         sb->s_op = &btrfs_super_ops;
 950         sb->s_d_op = &btrfs_dentry_operations;
 951         sb->s_export_op = &btrfs_export_ops;
 952         sb->s_xattr = btrfs_xattr_handlers;
 953         sb->s_time_gran = 1;
 954 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 955         sb->s_flags |= MS_POSIXACL;
 956 #endif
 957         sb->s_flags |= MS_I_VERSION;
 958         err = open_ctree(sb, fs_devices, (char *)data);
 959         if (err) {
 960                 printk(KERN_ERR "BTRFS: open_ctree failed\n");
 961                 return err;
 962         }
 963
 964         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 965         key.type = BTRFS_INODE_ITEM_KEY;
 966         key.offset = 0;
 967         inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
 968         if (IS_ERR(inode)) {
 969                 err = PTR_ERR(inode);
 970                 goto fail_close;
 971         }
 972
 973         sb->s_root = d_make_root(inode);
 974         if (!sb->s_root) {
 975                 err = -ENOMEM;
 976                 goto fail_close;
 977         }
 978
 979         save_mount_options(sb, data);
 980         cleancache_init_fs(sb);
 981         sb->s_flags |= MS_ACTIVE;
 982         return 0;
 983
 984 fail_close:
 985         close_ctree(fs_info->tree_root);
 986         return err;
 987 }
 988
 989 int btrfs_sync_fs(struct super_block *sb, int wait)
 990 {
 991         struct btrfs_trans_handle *trans;
 992         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 993         struct btrfs_root *root = fs_info->tree_root;
 994
 995         trace_btrfs_sync_fs(wait);
 996
 997         if (!wait) {
 998                 filemap_flush(fs_info->btree_inode->i_mapping);
 999                 return 0;
1000         }
1001
1002         btrfs_wait_ordered_roots(fs_info, -1);
1003
1004         trans = btrfs_attach_transaction_barrier(root);
1005         if (IS_ERR(trans)) {
1006                 /* no transaction, don't bother */
1007                 if (PTR_ERR(trans) == -ENOENT)
1008                         return 0;
1009                 return PTR_ERR(trans);
1010         }
1011         return btrfs_commit_transaction(trans, root);
1012 }
1013
1014 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1015 {
1016         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1017         struct btrfs_root *root = info->tree_root;
1018         char *compress_type;
1019
1020         if (btrfs_test_opt(root, DEGRADED))
1021                 seq_puts(seq, ",degraded");
1022         if (btrfs_test_opt(root, NODATASUM))
1023                 seq_puts(seq, ",nodatasum");
1024         if (btrfs_test_opt(root, NODATACOW))
1025                 seq_puts(seq, ",nodatacow");
1026         if (btrfs_test_opt(root, NOBARRIER))
1027                 seq_puts(seq, ",nobarrier");
1028         if (info->max_inline != 8192 * 1024)
1029                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1030         if (info->alloc_start != 0)
1031                 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1032         if (info->thread_pool_size !=  min_t(unsigned long,
1033                                              num_online_cpus() + 2, 8))
1034                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1035         if (btrfs_test_opt(root, COMPRESS)) {
1036                 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1037                         compress_type = "zlib";
1038                 else
1039                         compress_type = "lzo";
1040                 if (btrfs_test_opt(root, FORCE_COMPRESS))
1041                         seq_printf(seq, ",compress-force=%s", compress_type);
1042                 else
1043                         seq_printf(seq, ",compress=%s", compress_type);
1044         }
1045         if (btrfs_test_opt(root, NOSSD))
1046                 seq_puts(seq, ",nossd");
1047         if (btrfs_test_opt(root, SSD_SPREAD))
1048                 seq_puts(seq, ",ssd_spread");
1049         else if (btrfs_test_opt(root, SSD))
1050                 seq_puts(seq, ",ssd");
1051         if (btrfs_test_opt(root, NOTREELOG))
1052                 seq_puts(seq, ",notreelog");
1053         if (btrfs_test_opt(root, FLUSHONCOMMIT))
1054                 seq_puts(seq, ",flushoncommit");
1055         if (btrfs_test_opt(root, DISCARD))
1056                 seq_puts(seq, ",discard");
1057         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1058                 seq_puts(seq, ",noacl");
1059         if (btrfs_test_opt(root, SPACE_CACHE))
1060                 seq_puts(seq, ",space_cache");
1061         else
1062                 seq_puts(seq, ",nospace_cache");
1063         if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1064                 seq_puts(seq, ",rescan_uuid_tree");
1065         if (btrfs_test_opt(root, CLEAR_CACHE))
1066                 seq_puts(seq, ",clear_cache");
1067         if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1068                 seq_puts(seq, ",user_subvol_rm_allowed");
1069         if (btrfs_test_opt(root, ENOSPC_DEBUG))
1070                 seq_puts(seq, ",enospc_debug");
1071         if (btrfs_test_opt(root, AUTO_DEFRAG))
1072                 seq_puts(seq, ",autodefrag");
1073         if (btrfs_test_opt(root, INODE_MAP_CACHE))
1074                 seq_puts(seq, ",inode_cache");
1075         if (btrfs_test_opt(root, SKIP_BALANCE))
1076                 seq_puts(seq, ",skip_balance");
1077         if (btrfs_test_opt(root, RECOVERY))
1078                 seq_puts(seq, ",recovery");
1079 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1080         if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1081                 seq_puts(seq, ",check_int_data");
1082         else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1083                 seq_puts(seq, ",check_int");
1084         if (info->check_integrity_print_mask)
1085                 seq_printf(seq, ",check_int_print_mask=%d",
1086                                 info->check_integrity_print_mask);
1087 #endif
1088         if (info->metadata_ratio)
1089                 seq_printf(seq, ",metadata_ratio=%d",
1090                                 info->metadata_ratio);
1091         if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1092                 seq_puts(seq, ",fatal_errors=panic");
1093         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1094                 seq_printf(seq, ",commit=%d", info->commit_interval);
1095         return 0;
1096 }
1097
1098 static int btrfs_test_super(struct super_block *s, void *data)
1099 {
1100         struct btrfs_fs_info *p = data;
1101         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1102
1103         return fs_info->fs_devices == p->fs_devices;
1104 }
1105
1106 static int btrfs_set_super(struct super_block *s, void *data)
1107 {
1108         int err = set_anon_super(s, data);
1109         if (!err)
1110                 s->s_fs_info = data;
1111         return err;
1112 }
1113
1114 /*
1115  * subvolumes are identified by ino 256
1116  */
1117 static inline int is_subvolume_inode(struct inode *inode)
1118 {
1119         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1120                 return 1;
1121         return 0;
1122 }
1123
1124 /*
1125  * This will strip out the subvol=%s argument for an argument string and add
1126  * subvolid=0 to make sure we get the actual tree root for path walking to the
1127  * subvol we want.
1128  */
1129 static char *setup_root_args(char *args)
1130 {
1131         unsigned len = strlen(args) + 2 + 1;
1132         char *src, *dst, *buf;
1133
1134         /*
1135          * We need the same args as before, but with this substitution:
1136          * s!subvol=[^,]+!subvolid=0!
1137          *
1138          * Since the replacement string is up to 2 bytes longer than the
1139          * original, allocate strlen(args) + 2 + 1 bytes.
1140          */
1141
1142         src = strstr(args, "subvol=");
1143         /* This shouldn't happen, but just in case.. */
1144         if (!src)
1145                 return NULL;
1146
1147         buf = dst = kmalloc(len, GFP_NOFS);
1148         if (!buf)
1149                 return NULL;
1150
1151         /*
1152          * If the subvol= arg is not at the start of the string,
1153          * copy whatever precedes it into buf.
1154          */
1155         if (src != args) {
1156                 *src++ = '\0';
1157                 strcpy(buf, args);
1158                 dst += strlen(args);
1159         }
1160
1161         strcpy(dst, "subvolid=0");
1162         dst += strlen("subvolid=0");
1163
1164         /*
1165          * If there is a "," after the original subvol=... string,
1166          * copy that suffix into our buffer.  Otherwise, we're done.
1167          */
1168         src = strchr(src, ',');
1169         if (src)
1170                 strcpy(dst, src);
1171
1172         return buf;
1173 }
1174
1175 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1176                                    const char *device_name, char *data)
1177 {
1178         struct dentry *root;
1179         struct vfsmount *mnt;
1180         char *newargs;
1181
1182         newargs = setup_root_args(data);
1183         if (!newargs)
1184                 return ERR_PTR(-ENOMEM);
1185         mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1186                              newargs);
1187         kfree(newargs);
1188         if (IS_ERR(mnt))
1189                 return ERR_CAST(mnt);
1190
1191         root = mount_subtree(mnt, subvol_name);
1192
1193         if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1194                 struct super_block *s = root->d_sb;
1195                 dput(root);
1196                 root = ERR_PTR(-EINVAL);
1197                 deactivate_locked_super(s);
1198                 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1199                                 subvol_name);
1200         }
1201
1202         return root;
1203 }
1204
1205 /*
1206  * Find a superblock for the given device / mount point.
1207  *
1208  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
1209  *        for multiple device setup.  Make sure to keep it in sync.
1210  */
1211 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1212                 const char *device_name, void *data)
1213 {
1214         struct block_device *bdev = NULL;
1215         struct super_block *s;
1216         struct dentry *root;
1217         struct btrfs_fs_devices *fs_devices = NULL;
1218         struct btrfs_fs_info *fs_info = NULL;
1219         fmode_t mode = FMODE_READ;
1220         char *subvol_name = NULL;
1221         u64 subvol_objectid = 0;
1222         int error = 0;
1223
1224         if (!(flags & MS_RDONLY))
1225                 mode |= FMODE_WRITE;
1226
1227         error = btrfs_parse_early_options(data, mode, fs_type,
1228                                           &subvol_name, &subvol_objectid,
1229                                           &fs_devices);
1230         if (error) {
1231                 kfree(subvol_name);
1232                 return ERR_PTR(error);
1233         }
1234
1235         if (subvol_name) {
1236                 root = mount_subvol(subvol_name, flags, device_name, data);
1237                 kfree(subvol_name);
1238                 return root;
1239         }
1240
1241         error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1242         if (error)
1243                 return ERR_PTR(error);
1244
1245         /*
1246          * Setup a dummy root and fs_info for test/set super.  This is because
1247          * we don't actually fill this stuff out until open_ctree, but we need
1248          * it for searching for existing supers, so this lets us do that and
1249          * then open_ctree will properly initialize everything later.
1250          */
1251         fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1252         if (!fs_info)
1253                 return ERR_PTR(-ENOMEM);
1254
1255         fs_info->fs_devices = fs_devices;
1256
1257         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1258         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1259         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1260                 error = -ENOMEM;
1261                 goto error_fs_info;
1262         }
1263
1264         error = btrfs_open_devices(fs_devices, mode, fs_type);
1265         if (error)
1266                 goto error_fs_info;
1267
1268         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1269                 error = -EACCES;
1270                 goto error_close_devices;
1271         }
1272
1273         bdev = fs_devices->latest_bdev;
1274         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1275                  fs_info);
1276         if (IS_ERR(s)) {
1277                 error = PTR_ERR(s);
1278                 goto error_close_devices;
1279         }
1280
1281         if (s->s_root) {
1282                 btrfs_close_devices(fs_devices);
1283                 free_fs_info(fs_info);
1284                 if ((flags ^ s->s_flags) & MS_RDONLY)
1285                         error = -EBUSY;
1286         } else {
1287                 char b[BDEVNAME_SIZE];
1288
1289                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1290                 btrfs_sb(s)->bdev_holder = fs_type;
1291                 error = btrfs_fill_super(s, fs_devices, data,
1292                                          flags & MS_SILENT ? 1 : 0);
1293         }
1294
1295         root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1296         if (IS_ERR(root))
1297                 deactivate_locked_super(s);
1298
1299         return root;
1300
1301 error_close_devices:
1302         btrfs_close_devices(fs_devices);
1303 error_fs_info:
1304         free_fs_info(fs_info);
1305         return ERR_PTR(error);
1306 }
1307
1308 static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1309 {
1310         spin_lock_irq(&workers->lock);
1311         workers->max_workers = new_limit;
1312         spin_unlock_irq(&workers->lock);
1313 }
1314
1315 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1316                                      int new_pool_size, int old_pool_size)
1317 {
1318         if (new_pool_size == old_pool_size)
1319                 return;
1320
1321         fs_info->thread_pool_size = new_pool_size;
1322
1323         btrfs_info(fs_info, "resize thread pool %d -> %d",
1324                old_pool_size, new_pool_size);
1325
1326         btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1327         btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1328         btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1329         btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1330         btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1331         btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1332         btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1333         btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1334         btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1335         btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1336         btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1337         btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1338         btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
1339         btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers,
1340                               new_pool_size);
1341 }
1342
1343 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1344 {
1345         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1346 }
1347
1348 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1349                                        unsigned long old_opts, int flags)
1350 {
1351         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1352             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1353              (flags & MS_RDONLY))) {
1354                 /* wait for any defraggers to finish */
1355                 wait_event(fs_info->transaction_wait,
1356                            (atomic_read(&fs_info->defrag_running) == 0));
1357                 if (flags & MS_RDONLY)
1358                         sync_filesystem(fs_info->sb);
1359         }
1360 }
1361
1362 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1363                                          unsigned long old_opts)
1364 {
1365         /*
1366          * We need cleanup all defragable inodes if the autodefragment is
1367          * close or the fs is R/O.
1368          */
1369         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1370             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1371              (fs_info->sb->s_flags & MS_RDONLY))) {
1372                 btrfs_cleanup_defrag_inodes(fs_info);
1373         }
1374
1375         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1376 }
1377
1378 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1379 {
1380         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1381         struct btrfs_root *root = fs_info->tree_root;
1382         unsigned old_flags = sb->s_flags;
1383         unsigned long old_opts = fs_info->mount_opt;
1384         unsigned long old_compress_type = fs_info->compress_type;
1385         u64 old_max_inline = fs_info->max_inline;
1386         u64 old_alloc_start = fs_info->alloc_start;
1387         int old_thread_pool_size = fs_info->thread_pool_size;
1388         unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1389         int ret;
1390
1391         btrfs_remount_prepare(fs_info);
1392
1393         ret = btrfs_parse_options(root, data);
1394         if (ret) {
1395                 ret = -EINVAL;
1396                 goto restore;
1397         }
1398
1399         btrfs_remount_begin(fs_info, old_opts, *flags);
1400         btrfs_resize_thread_pool(fs_info,
1401                 fs_info->thread_pool_size, old_thread_pool_size);
1402
1403         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1404                 goto out;
1405
1406         if (*flags & MS_RDONLY) {
1407                 /*
1408                  * this also happens on 'umount -rf' or on shutdown, when
1409                  * the filesystem is busy.
1410                  */
1411
1412                 /* wait for the uuid_scan task to finish */
1413                 down(&fs_info->uuid_tree_rescan_sem);
1414                 /* avoid complains from lockdep et al. */
1415                 up(&fs_info->uuid_tree_rescan_sem);
1416
1417                 sb->s_flags |= MS_RDONLY;
1418
1419                 btrfs_dev_replace_suspend_for_unmount(fs_info);
1420                 btrfs_scrub_cancel(fs_info);
1421                 btrfs_pause_balance(fs_info);
1422
1423                 ret = btrfs_commit_super(root);
1424                 if (ret)
1425                         goto restore;
1426         } else {
1427                 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1428                         btrfs_err(fs_info,
1429                                 "Remounting read-write after error is not allowed");
1430                         ret = -EINVAL;
1431                         goto restore;
1432                 }
1433                 if (fs_info->fs_devices->rw_devices == 0) {
1434                         ret = -EACCES;
1435                         goto restore;
1436                 }
1437
1438                 if (fs_info->fs_devices->missing_devices >
1439                      fs_info->num_tolerated_disk_barrier_failures &&
1440                     !(*flags & MS_RDONLY)) {
1441                         btrfs_warn(fs_info,
1442                                 "too many missing devices, writeable remount is not allowed");
1443                         ret = -EACCES;
1444                         goto restore;
1445                 }
1446
1447                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1448                         ret = -EINVAL;
1449                         goto restore;
1450                 }
1451
1452                 ret = btrfs_cleanup_fs_roots(fs_info);
1453                 if (ret)
1454                         goto restore;
1455
1456                 /* recover relocation */
1457                 ret = btrfs_recover_relocation(root);
1458                 if (ret)
1459                         goto restore;
1460
1461                 ret = btrfs_resume_balance_async(fs_info);
1462                 if (ret)
1463                         goto restore;
1464
1465                 ret = btrfs_resume_dev_replace_async(fs_info);
1466                 if (ret) {
1467                         btrfs_warn(fs_info, "failed to resume dev_replace");
1468                         goto restore;
1469                 }
1470
1471                 if (!fs_info->uuid_root) {
1472                         btrfs_info(fs_info, "creating UUID tree");
1473                         ret = btrfs_create_uuid_tree(fs_info);
1474                         if (ret) {
1475                                 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1476                                 goto restore;
1477                         }
1478                 }
1479                 sb->s_flags &= ~MS_RDONLY;
1480         }
1481 out:
1482         btrfs_remount_cleanup(fs_info, old_opts);
1483         return 0;
1484
1485 restore:
1486         /* We've hit an error - don't reset MS_RDONLY */
1487         if (sb->s_flags & MS_RDONLY)
1488                 old_flags |= MS_RDONLY;
1489         sb->s_flags = old_flags;
1490         fs_info->mount_opt = old_opts;
1491         fs_info->compress_type = old_compress_type;
1492         fs_info->max_inline = old_max_inline;
1493         mutex_lock(&fs_info->chunk_mutex);
1494         fs_info->alloc_start = old_alloc_start;
1495         mutex_unlock(&fs_info->chunk_mutex);
1496         btrfs_resize_thread_pool(fs_info,
1497                 old_thread_pool_size, fs_info->thread_pool_size);
1498         fs_info->metadata_ratio = old_metadata_ratio;
1499         btrfs_remount_cleanup(fs_info, old_opts);
1500         return ret;
1501 }
1502
1503 /* Used to sort the devices by max_avail(descending sort) */
1504 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1505                                        const void *dev_info2)
1506 {
1507         if (((struct btrfs_device_info *)dev_info1)->max_avail >
1508             ((struct btrfs_device_info *)dev_info2)->max_avail)
1509                 return -1;
1510         else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1511                  ((struct btrfs_device_info *)dev_info2)->max_avail)
1512                 return 1;
1513         else
1514         return 0;
1515 }
1516
1517 /*
1518  * sort the devices by max_avail, in which max free extent size of each device
1519  * is stored.(Descending Sort)
1520  */
1521 static inline void btrfs_descending_sort_devices(
1522                                         struct btrfs_device_info *devices,
1523                                         size_t nr_devices)
1524 {
1525         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1526              btrfs_cmp_device_free_bytes, NULL);
1527 }
1528
1529 /*
1530  * The helper to calc the free space on the devices that can be used to store
1531  * file data.
1532  */
1533 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1534 {
1535         struct btrfs_fs_info *fs_info = root->fs_info;
1536         struct btrfs_device_info *devices_info;
1537         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1538         struct btrfs_device *device;
1539         u64 skip_space;
1540         u64 type;
1541         u64 avail_space;
1542         u64 used_space;
1543         u64 min_stripe_size;
1544         int min_stripes = 1, num_stripes = 1;
1545         int i = 0, nr_devices;
1546         int ret;
1547
1548         nr_devices = fs_info->fs_devices->open_devices;
1549         BUG_ON(!nr_devices);
1550
1551         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1552                                GFP_NOFS);
1553         if (!devices_info)
1554                 return -ENOMEM;
1555
1556         /* calc min stripe number for data space alloction */
1557         type = btrfs_get_alloc_profile(root, 1);
1558         if (type & BTRFS_BLOCK_GROUP_RAID0) {
1559                 min_stripes = 2;
1560                 num_stripes = nr_devices;
1561         } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1562                 min_stripes = 2;
1563                 num_stripes = 2;
1564         } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1565                 min_stripes = 4;
1566                 num_stripes = 4;
1567         }
1568
1569         if (type & BTRFS_BLOCK_GROUP_DUP)
1570                 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1571         else
1572                 min_stripe_size = BTRFS_STRIPE_LEN;
1573
1574         list_for_each_entry(device, &fs_devices->devices, dev_list) {
1575                 if (!device->in_fs_metadata || !device->bdev ||
1576                     device->is_tgtdev_for_dev_replace)
1577                         continue;
1578
1579                 avail_space = device->total_bytes - device->bytes_used;
1580
1581                 /* align with stripe_len */
1582                 do_div(avail_space, BTRFS_STRIPE_LEN);
1583                 avail_space *= BTRFS_STRIPE_LEN;
1584
1585                 /*
1586                  * In order to avoid overwritting the superblock on the drive,
1587                  * btrfs starts at an offset of at least 1MB when doing chunk
1588                  * allocation.
1589                  */
1590                 skip_space = 1024 * 1024;
1591
1592                 /* user can set the offset in fs_info->alloc_start. */
1593                 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1594                     device->total_bytes)
1595                         skip_space = max(fs_info->alloc_start, skip_space);
1596
1597                 /*
1598                  * btrfs can not use the free space in [0, skip_space - 1],
1599                  * we must subtract it from the total. In order to implement
1600                  * it, we account the used space in this range first.
1601                  */
1602                 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1603                                                      &used_space);
1604                 if (ret) {
1605                         kfree(devices_info);
1606                         return ret;
1607                 }
1608
1609                 /* calc the free space in [0, skip_space - 1] */
1610                 skip_space -= used_space;
1611
1612                 /*
1613                  * we can use the free space in [0, skip_space - 1], subtract
1614                  * it from the total.
1615                  */
1616                 if (avail_space && avail_space >= skip_space)
1617                         avail_space -= skip_space;
1618                 else
1619                         avail_space = 0;
1620
1621                 if (avail_space < min_stripe_size)
1622                         continue;
1623
1624                 devices_info[i].dev = device;
1625                 devices_info[i].max_avail = avail_space;
1626
1627                 i++;
1628         }
1629
1630         nr_devices = i;
1631
1632         btrfs_descending_sort_devices(devices_info, nr_devices);
1633
1634         i = nr_devices - 1;
1635         avail_space = 0;
1636         while (nr_devices >= min_stripes) {
1637                 if (num_stripes > nr_devices)
1638                         num_stripes = nr_devices;
1639
1640                 if (devices_info[i].max_avail >= min_stripe_size) {
1641                         int j;
1642                         u64 alloc_size;
1643
1644                         avail_space += devices_info[i].max_avail * num_stripes;
1645                         alloc_size = devices_info[i].max_avail;
1646                         for (j = i + 1 - num_stripes; j <= i; j++)
1647                                 devices_info[j].max_avail -= alloc_size;
1648                 }
1649                 i--;
1650                 nr_devices--;
1651         }
1652
1653         kfree(devices_info);
1654         *free_bytes = avail_space;
1655         return 0;
1656 }
1657
1658 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1659 {
1660         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1661         struct btrfs_super_block *disk_super = fs_info->super_copy;
1662         struct list_head *head = &fs_info->space_info;
1663         struct btrfs_space_info *found;
1664         u64 total_used = 0;
1665         u64 total_free_data = 0;
1666         int bits = dentry->d_sb->s_blocksize_bits;
1667         __be32 *fsid = (__be32 *)fs_info->fsid;
1668         int ret;
1669
1670         /* holding chunk_muext to avoid allocating new chunks */
1671         mutex_lock(&fs_info->chunk_mutex);
1672         rcu_read_lock();
1673         list_for_each_entry_rcu(found, head, list) {
1674                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1675                         total_free_data += found->disk_total - found->disk_used;
1676                         total_free_data -=
1677                                 btrfs_account_ro_block_groups_free_space(found);
1678                 }
1679
1680                 total_used += found->disk_used;
1681         }
1682         rcu_read_unlock();
1683
1684         buf->f_namelen = BTRFS_NAME_LEN;
1685         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1686         buf->f_bfree = buf->f_blocks - (total_used >> bits);
1687         buf->f_bsize = dentry->d_sb->s_blocksize;
1688         buf->f_type = BTRFS_SUPER_MAGIC;
1689         buf->f_bavail = total_free_data;
1690         ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1691         if (ret) {
1692                 mutex_unlock(&fs_info->chunk_mutex);
1693                 return ret;
1694         }
1695         buf->f_bavail += total_free_data;
1696         buf->f_bavail = buf->f_bavail >> bits;
1697         mutex_unlock(&fs_info->chunk_mutex);
1698
1699         /* We treat it as constant endianness (it doesn't matter _which_)
1700            because we want the fsid to come out the same whether mounted
1701            on a big-endian or little-endian host */
1702         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1703         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1704         /* Mask in the root object ID too, to disambiguate subvols */
1705         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1706         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1707
1708         return 0;
1709 }
1710
1711 static void btrfs_kill_super(struct super_block *sb)
1712 {
1713         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1714         kill_anon_super(sb);
1715         free_fs_info(fs_info);
1716 }
1717
1718 static struct file_system_type btrfs_fs_type = {
1719         .owner          = THIS_MODULE,
1720         .name           = "btrfs",
1721         .mount          = btrfs_mount,
1722         .kill_sb        = btrfs_kill_super,
1723         .fs_flags       = FS_REQUIRES_DEV,
1724 };
1725 MODULE_ALIAS_FS("btrfs");
1726
1727 /*
1728  * used by btrfsctl to scan devices when no FS is mounted
1729  */
1730 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1731                                 unsigned long arg)
1732 {
1733         struct btrfs_ioctl_vol_args *vol;
1734         struct btrfs_fs_devices *fs_devices;
1735         int ret = -ENOTTY;
1736
1737         if (!capable(CAP_SYS_ADMIN))
1738                 return -EPERM;
1739
1740         vol = memdup_user((void __user *)arg, sizeof(*vol));
1741         if (IS_ERR(vol))
1742                 return PTR_ERR(vol);
1743
1744         switch (cmd) {
1745         case BTRFS_IOC_SCAN_DEV:
1746                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1747                                             &btrfs_fs_type, &fs_devices);
1748                 break;
1749         case BTRFS_IOC_DEVICES_READY:
1750                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1751                                             &btrfs_fs_type, &fs_devices);
1752                 if (ret)
1753                         break;
1754                 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1755                 break;
1756         }
1757
1758         kfree(vol);
1759         return ret;
1760 }
1761
1762 static int btrfs_freeze(struct super_block *sb)
1763 {
1764         struct btrfs_trans_handle *trans;
1765         struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1766
1767         trans = btrfs_attach_transaction_barrier(root);
1768         if (IS_ERR(trans)) {
1769                 /* no transaction, don't bother */
1770                 if (PTR_ERR(trans) == -ENOENT)
1771                         return 0;
1772                 return PTR_ERR(trans);
1773         }
1774         return btrfs_commit_transaction(trans, root);
1775 }
1776
1777 static int btrfs_unfreeze(struct super_block *sb)
1778 {
1779         return 0;
1780 }
1781
1782 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1783 {
1784         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1785         struct btrfs_fs_devices *cur_devices;
1786         struct btrfs_device *dev, *first_dev = NULL;
1787         struct list_head *head;
1788         struct rcu_string *name;
1789
1790         mutex_lock(&fs_info->fs_devices->device_list_mutex);
1791         cur_devices = fs_info->fs_devices;
1792         while (cur_devices) {
1793                 head = &cur_devices->devices;
1794                 list_for_each_entry(dev, head, dev_list) {
1795                         if (dev->missing)
1796                                 continue;
1797                         if (!first_dev || dev->devid < first_dev->devid)
1798                                 first_dev = dev;
1799                 }
1800                 cur_devices = cur_devices->seed;
1801         }
1802
1803         if (first_dev) {
1804                 rcu_read_lock();
1805                 name = rcu_dereference(first_dev->name);
1806                 seq_escape(m, name->str, " \t\n\\");
1807                 rcu_read_unlock();
1808         } else {
1809                 WARN_ON(1);
1810         }
1811         mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1812         return 0;
1813 }
1814
1815 static const struct super_operations btrfs_super_ops = {
1816         .drop_inode     = btrfs_drop_inode,
1817         .evict_inode    = btrfs_evict_inode,
1818         .put_super      = btrfs_put_super,
1819         .sync_fs        = btrfs_sync_fs,
1820         .show_options   = btrfs_show_options,
1821         .show_devname   = btrfs_show_devname,
1822         .write_inode    = btrfs_write_inode,
1823         .alloc_inode    = btrfs_alloc_inode,
1824         .destroy_inode  = btrfs_destroy_inode,
1825         .statfs         = btrfs_statfs,
1826         .remount_fs     = btrfs_remount,
1827         .freeze_fs      = btrfs_freeze,
1828         .unfreeze_fs    = btrfs_unfreeze,
1829 };
1830
1831 static const struct file_operations btrfs_ctl_fops = {
1832         .unlocked_ioctl  = btrfs_control_ioctl,
1833         .compat_ioctl = btrfs_control_ioctl,
1834         .owner   = THIS_MODULE,
1835         .llseek = noop_llseek,
1836 };
1837
1838 static struct miscdevice btrfs_misc = {
1839         .minor          = BTRFS_MINOR,
1840         .name           = "btrfs-control",
1841         .fops           = &btrfs_ctl_fops
1842 };
1843
1844 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1845 MODULE_ALIAS("devname:btrfs-control");
1846
1847 static int btrfs_interface_init(void)
1848 {
1849         return misc_register(&btrfs_misc);
1850 }
1851
1852 static void btrfs_interface_exit(void)
1853 {
1854         if (misc_deregister(&btrfs_misc) < 0)
1855                 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
1856 }
1857
1858 static void btrfs_print_info(void)
1859 {
1860         printk(KERN_INFO "Btrfs loaded"
1861 #ifdef CONFIG_BTRFS_DEBUG
1862                         ", debug=on"
1863 #endif
1864 #ifdef CONFIG_BTRFS_ASSERT
1865                         ", assert=on"
1866 #endif
1867 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1868                         ", integrity-checker=on"
1869 #endif
1870                         "\n");
1871 }
1872
1873 static int btrfs_run_sanity_tests(void)
1874 {
1875         int ret;
1876
1877         ret = btrfs_init_test_fs();
1878         if (ret)
1879                 return ret;
1880
1881         ret = btrfs_test_free_space_cache();
1882         if (ret)
1883                 goto out;
1884         ret = btrfs_test_extent_buffer_operations();
1885         if (ret)
1886                 goto out;
1887         ret = btrfs_test_extent_io();
1888         if (ret)
1889                 goto out;
1890         ret = btrfs_test_inodes();
1891 out:
1892         btrfs_destroy_test_fs();
1893         return ret;
1894 }
1895
1896 static int __init init_btrfs_fs(void)
1897 {
1898         int err;
1899
1900         err = btrfs_hash_init();
1901         if (err)
1902                 return err;
1903
1904         btrfs_props_init();
1905
1906         err = btrfs_init_sysfs();
1907         if (err)
1908                 goto free_hash;
1909
1910         btrfs_init_compress();
1911
1912         err = btrfs_init_cachep();
1913         if (err)
1914                 goto free_compress;
1915
1916         err = extent_io_init();
1917         if (err)
1918                 goto free_cachep;
1919
1920         err = extent_map_init();
1921         if (err)
1922                 goto free_extent_io;
1923
1924         err = ordered_data_init();
1925         if (err)
1926                 goto free_extent_map;
1927
1928         err = btrfs_delayed_inode_init();
1929         if (err)
1930                 goto free_ordered_data;
1931
1932         err = btrfs_auto_defrag_init();
1933         if (err)
1934                 goto free_delayed_inode;
1935
1936         err = btrfs_delayed_ref_init();
1937         if (err)
1938                 goto free_auto_defrag;
1939
1940         err = btrfs_prelim_ref_init();
1941         if (err)
1942                 goto free_prelim_ref;
1943
1944         err = btrfs_interface_init();
1945         if (err)
1946                 goto free_delayed_ref;
1947
1948         btrfs_init_lockdep();
1949
1950         btrfs_print_info();
1951
1952         err = btrfs_run_sanity_tests();
1953         if (err)
1954                 goto unregister_ioctl;
1955
1956         err = register_filesystem(&btrfs_fs_type);
1957         if (err)
1958                 goto unregister_ioctl;
1959
1960         return 0;
1961
1962 unregister_ioctl:
1963         btrfs_interface_exit();
1964 free_prelim_ref:
1965         btrfs_prelim_ref_exit();
1966 free_delayed_ref:
1967         btrfs_delayed_ref_exit();
1968 free_auto_defrag:
1969         btrfs_auto_defrag_exit();
1970 free_delayed_inode:
1971         btrfs_delayed_inode_exit();
1972 free_ordered_data:
1973         ordered_data_exit();
1974 free_extent_map:
1975         extent_map_exit();
1976 free_extent_io:
1977         extent_io_exit();
1978 free_cachep:
1979         btrfs_destroy_cachep();
1980 free_compress:
1981         btrfs_exit_compress();
1982         btrfs_exit_sysfs();
1983 free_hash:
1984         btrfs_hash_exit();
1985         return err;
1986 }
1987
1988 static void __exit exit_btrfs_fs(void)
1989 {
1990         btrfs_destroy_cachep();
1991         btrfs_delayed_ref_exit();
1992         btrfs_auto_defrag_exit();
1993         btrfs_delayed_inode_exit();
1994         btrfs_prelim_ref_exit();
1995         ordered_data_exit();
1996         extent_map_exit();
1997         extent_io_exit();
1998         btrfs_interface_exit();
1999         unregister_filesystem(&btrfs_fs_type);
2000         btrfs_exit_sysfs();
2001         btrfs_cleanup_fs_uuids();
2002         btrfs_exit_compress();
2003         btrfs_hash_exit();
2004 }
2005
2006 late_initcall(init_btrfs_fs);
2007 module_exit(exit_btrfs_fs)
2008
2009 MODULE_LICENSE("GPL");