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                         btrfs_set_opt(info->mount_opt, SSD);
 526                         break;
 527                 case Opt_nossd:
 528                         btrfs_set_and_info(root, NOSSD,
 529                                              "not using ssd allocation scheme");
 530                         btrfs_clear_opt(info->mount_opt, SSD);
 531                         break;
 532                 case Opt_barrier:
 533                         btrfs_clear_and_info(root, NOBARRIER,
 534                                              "turning on barriers");
 535                         break;
 536                 case Opt_nobarrier:
 537                         btrfs_set_and_info(root, NOBARRIER,
 538                                            "turning off barriers");
 539                         break;
 540                 case Opt_thread_pool:
 541                         ret = match_int(&args[0], &intarg);
 542                         if (ret) {
 543                                 goto out;
 544                         } else if (intarg > 0) {
 545                                 info->thread_pool_size = intarg;
 546                         } else {
 547                                 ret = -EINVAL;
 548                                 goto out;
 549                         }
 550                         break;
 551                 case Opt_max_inline:
 552                         num = match_strdup(&args[0]);
 553                         if (num) {
 554                                 info->max_inline = memparse(num, NULL);
 555                                 kfree(num);
 556
 557                                 if (info->max_inline) {
 558                                         info->max_inline = min_t(u64,
 559                                                 info->max_inline,
 560                                                 root->sectorsize);
 561                                 }
 562                                 btrfs_info(root->fs_info, "max_inline at %llu",
 563                                         info->max_inline);
 564                         } else {
 565                                 ret = -ENOMEM;
 566                                 goto out;
 567                         }
 568                         break;
 569                 case Opt_alloc_start:
 570                         num = match_strdup(&args[0]);
 571                         if (num) {
 572                                 mutex_lock(&info->chunk_mutex);
 573                                 info->alloc_start = memparse(num, NULL);
 574                                 mutex_unlock(&info->chunk_mutex);
 575                                 kfree(num);
 576                                 btrfs_info(root->fs_info, "allocations start at %llu",
 577                                         info->alloc_start);
 578                         } else {
 579                                 ret = -ENOMEM;
 580                                 goto out;
 581                         }
 582                         break;
 583                 case Opt_acl:
 584 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 585                         root->fs_info->sb->s_flags |= MS_POSIXACL;
 586                         break;
 587 #else
 588                         btrfs_err(root->fs_info,
 589                                 "support for ACL not compiled in!");
 590                         ret = -EINVAL;
 591                         goto out;
 592 #endif
 593                 case Opt_noacl:
 594                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 595                         break;
 596                 case Opt_notreelog:
 597                         btrfs_set_and_info(root, NOTREELOG,
 598                                            "disabling tree log");
 599                         break;
 600                 case Opt_treelog:
 601                         btrfs_clear_and_info(root, NOTREELOG,
 602                                              "enabling tree log");
 603                         break;
 604                 case Opt_flushoncommit:
 605                         btrfs_set_and_info(root, FLUSHONCOMMIT,
 606                                            "turning on flush-on-commit");
 607                         break;
 608                 case Opt_noflushoncommit:
 609                         btrfs_clear_and_info(root, FLUSHONCOMMIT,
 610                                              "turning off flush-on-commit");
 611                         break;
 612                 case Opt_ratio:
 613                         ret = match_int(&args[0], &intarg);
 614                         if (ret) {
 615                                 goto out;
 616                         } else if (intarg >= 0) {
 617                                 info->metadata_ratio = intarg;
 618                                 btrfs_info(root->fs_info, "metadata ratio %d",
 619                                        info->metadata_ratio);
 620                         } else {
 621                                 ret = -EINVAL;
 622                                 goto out;
 623                         }
 624                         break;
 625                 case Opt_discard:
 626                         btrfs_set_and_info(root, DISCARD,
 627                                            "turning on discard");
 628                         break;
 629                 case Opt_nodiscard:
 630                         btrfs_clear_and_info(root, DISCARD,
 631                                              "turning off discard");
 632                         break;
 633                 case Opt_space_cache:
 634                         btrfs_set_and_info(root, SPACE_CACHE,
 635                                            "enabling disk space caching");
 636                         break;
 637                 case Opt_rescan_uuid_tree:
 638                         btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 639                         break;
 640                 case Opt_no_space_cache:
 641                         btrfs_clear_and_info(root, SPACE_CACHE,
 642                                              "disabling disk space caching");
 643                         break;
 644                 case Opt_inode_cache:
 645                         btrfs_set_and_info(root, CHANGE_INODE_CACHE,
 646                                            "enabling inode map caching");
 647                         break;
 648                 case Opt_noinode_cache:
 649                         btrfs_clear_and_info(root, CHANGE_INODE_CACHE,
 650                                              "disabling inode map caching");
 651                         break;
 652                 case Opt_clear_cache:
 653                         btrfs_set_and_info(root, CLEAR_CACHE,
 654                                            "force clearing of disk cache");
 655                         break;
 656                 case Opt_user_subvol_rm_allowed:
 657                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 658                         break;
 659                 case Opt_enospc_debug:
 660                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 661                         break;
 662                 case Opt_noenospc_debug:
 663                         btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
 664                         break;
 665                 case Opt_defrag:
 666                         btrfs_set_and_info(root, AUTO_DEFRAG,
 667                                            "enabling auto defrag");
 668                         break;
 669                 case Opt_nodefrag:
 670                         btrfs_clear_and_info(root, AUTO_DEFRAG,
 671                                              "disabling auto defrag");
 672                         break;
 673                 case Opt_recovery:
 674                         btrfs_info(root->fs_info, "enabling auto recovery");
 675                         btrfs_set_opt(info->mount_opt, RECOVERY);
 676                         break;
 677                 case Opt_skip_balance:
 678                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 679                         break;
 680 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 681                 case Opt_check_integrity_including_extent_data:
 682                         btrfs_info(root->fs_info,
 683                                    "enabling check integrity including extent data");
 684                         btrfs_set_opt(info->mount_opt,
 685                                       CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
 686                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 687                         break;
 688                 case Opt_check_integrity:
 689                         btrfs_info(root->fs_info, "enabling check integrity");
 690                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 691                         break;
 692                 case Opt_check_integrity_print_mask:
 693                         ret = match_int(&args[0], &intarg);
 694                         if (ret) {
 695                                 goto out;
 696                         } else if (intarg >= 0) {
 697                                 info->check_integrity_print_mask = intarg;
 698                                 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
 699                                        info->check_integrity_print_mask);
 700                         } else {
 701                                 ret = -EINVAL;
 702                                 goto out;
 703                         }
 704                         break;
 705 #else
 706                 case Opt_check_integrity_including_extent_data:
 707                 case Opt_check_integrity:
 708                 case Opt_check_integrity_print_mask:
 709                         btrfs_err(root->fs_info,
 710                                 "support for check_integrity* not compiled in!");
 711                         ret = -EINVAL;
 712                         goto out;
 713 #endif
 714                 case Opt_fatal_errors:
 715                         if (strcmp(args[0].from, "panic") == 0)
 716                                 btrfs_set_opt(info->mount_opt,
 717                                               PANIC_ON_FATAL_ERROR);
 718                         else if (strcmp(args[0].from, "bug") == 0)
 719                                 btrfs_clear_opt(info->mount_opt,
 720                                               PANIC_ON_FATAL_ERROR);
 721                         else {
 722                                 ret = -EINVAL;
 723                                 goto out;
 724                         }
 725                         break;
 726                 case Opt_commit_interval:
 727                         intarg = 0;
 728                         ret = match_int(&args[0], &intarg);
 729                         if (ret < 0) {
 730                                 btrfs_err(root->fs_info, "invalid commit interval");
 731                                 ret = -EINVAL;
 732                                 goto out;
 733                         }
 734                         if (intarg > 0) {
 735                                 if (intarg > 300) {
 736                                         btrfs_warn(root->fs_info, "excessive commit interval %d",
 737                                                         intarg);
 738                                 }
 739                                 info->commit_interval = intarg;
 740                         } else {
 741                                 btrfs_info(root->fs_info, "using default commit interval %ds",
 742                                     BTRFS_DEFAULT_COMMIT_INTERVAL);
 743                                 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
 744                         }
 745                         break;
 746                 case Opt_err:
 747                         btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
 748                         ret = -EINVAL;
 749                         goto out;
 750                 default:
 751                         break;
 752                 }
 753         }
 754 out:
 755         if (!ret && btrfs_test_opt(root, SPACE_CACHE))
 756                 btrfs_info(root->fs_info, "disk space caching is enabled");
 757         kfree(orig);
 758         return ret;
 759 }
 760
 761 /*
 762  * Parse mount options that are required early in the mount process.
 763  *
 764  * All other options will be parsed on much later in the mount process and
 765  * only when we need to allocate a new super block.
 766  */
 767 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 768                 void *holder, char **subvol_name, u64 *subvol_objectid,
 769                 struct btrfs_fs_devices **fs_devices)
 770 {
 771         substring_t args[MAX_OPT_ARGS];
 772         char *device_name, *opts, *orig, *p;
 773         char *num = NULL;
 774         int error = 0;
 775
 776         if (!options)
 777                 return 0;
 778
 779         /*
 780          * strsep changes the string, duplicate it because parse_options
 781          * gets called twice
 782          */
 783         opts = kstrdup(options, GFP_KERNEL);
 784         if (!opts)
 785                 return -ENOMEM;
 786         orig = opts;
 787
 788         while ((p = strsep(&opts, ",")) != NULL) {
 789                 int token;
 790                 if (!*p)
 791                         continue;
 792
 793                 token = match_token(p, tokens, args);
 794                 switch (token) {
 795                 case Opt_subvol:
 796                         kfree(*subvol_name);
 797                         *subvol_name = match_strdup(&args[0]);
 798                         if (!*subvol_name) {
 799                                 error = -ENOMEM;
 800                                 goto out;
 801                         }
 802                         break;
 803                 case Opt_subvolid:
 804                         num = match_strdup(&args[0]);
 805                         if (num) {
 806                                 *subvol_objectid = memparse(num, NULL);
 807                                 kfree(num);
 808                                 /* we want the original fs_tree */
 809                                 if (!*subvol_objectid)
 810                                         *subvol_objectid =
 811                                                 BTRFS_FS_TREE_OBJECTID;
 812                         } else {
 813                                 error = -EINVAL;
 814                                 goto out;
 815                         }
 816                         break;
 817                 case Opt_subvolrootid:
 818                         printk(KERN_WARNING
 819                                 "BTRFS: 'subvolrootid' mount option is deprecated and has "
 820                                 "no effect\n");
 821                         break;
 822                 case Opt_device:
 823                         device_name = match_strdup(&args[0]);
 824                         if (!device_name) {
 825                                 error = -ENOMEM;
 826                                 goto out;
 827                         }
 828                         error = btrfs_scan_one_device(device_name,
 829                                         flags, holder, fs_devices);
 830                         kfree(device_name);
 831                         if (error)
 832                                 goto out;
 833                         break;
 834                 default:
 835                         break;
 836                 }
 837         }
 838
 839 out:
 840         kfree(orig);
 841         return error;
 842 }
 843
 844 static struct dentry *get_default_root(struct super_block *sb,
 845                                        u64 subvol_objectid)
 846 {
 847         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 848         struct btrfs_root *root = fs_info->tree_root;
 849         struct btrfs_root *new_root;
 850         struct btrfs_dir_item *di;
 851         struct btrfs_path *path;
 852         struct btrfs_key location;
 853         struct inode *inode;
 854         struct dentry *dentry;
 855         u64 dir_id;
 856         int new = 0;
 857
 858         /*
 859          * We have a specific subvol we want to mount, just setup location and
 860          * go look up the root.
 861          */
 862         if (subvol_objectid) {
 863                 location.objectid = subvol_objectid;
 864                 location.type = BTRFS_ROOT_ITEM_KEY;
 865                 location.offset = (u64)-1;
 866                 goto find_root;
 867         }
 868
 869         path = btrfs_alloc_path();
 870         if (!path)
 871                 return ERR_PTR(-ENOMEM);
 872         path->leave_spinning = 1;
 873
 874         /*
 875          * Find the "default" dir item which points to the root item that we
 876          * will mount by default if we haven't been given a specific subvolume
 877          * to mount.
 878          */
 879         dir_id = btrfs_super_root_dir(fs_info->super_copy);
 880         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 881         if (IS_ERR(di)) {
 882                 btrfs_free_path(path);
 883                 return ERR_CAST(di);
 884         }
 885         if (!di) {
 886                 /*
 887                  * Ok the default dir item isn't there.  This is weird since
 888                  * it's always been there, but don't freak out, just try and
 889                  * mount to root most subvolume.
 890                  */
 891                 btrfs_free_path(path);
 892                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 893                 new_root = fs_info->fs_root;
 894                 goto setup_root;
 895         }
 896
 897         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 898         btrfs_free_path(path);
 899
 900 find_root:
 901         new_root = btrfs_read_fs_root_no_name(fs_info, &location);
 902         if (IS_ERR(new_root))
 903                 return ERR_CAST(new_root);
 904
 905         dir_id = btrfs_root_dirid(&new_root->root_item);
 906 setup_root:
 907         location.objectid = dir_id;
 908         location.type = BTRFS_INODE_ITEM_KEY;
 909         location.offset = 0;
 910
 911         inode = btrfs_iget(sb, &location, new_root, &new);
 912         if (IS_ERR(inode))
 913                 return ERR_CAST(inode);
 914
 915         /*
 916          * If we're just mounting the root most subvol put the inode and return
 917          * a reference to the dentry.  We will have already gotten a reference
 918          * to the inode in btrfs_fill_super so we're good to go.
 919          */
 920         if (!new && sb->s_root->d_inode == inode) {
 921                 iput(inode);
 922                 return dget(sb->s_root);
 923         }
 924
 925         dentry = d_obtain_alias(inode);
 926         if (!IS_ERR(dentry)) {
 927                 spin_lock(&dentry->d_lock);
 928                 dentry->d_flags &= ~DCACHE_DISCONNECTED;
 929                 spin_unlock(&dentry->d_lock);
 930         }
 931         return dentry;
 932 }
 933
 934 static int btrfs_fill_super(struct super_block *sb,
 935                             struct btrfs_fs_devices *fs_devices,
 936                             void *data, int silent)
 937 {
 938         struct inode *inode;
 939         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 940         struct btrfs_key key;
 941         int err;
 942
 943         sb->s_maxbytes = MAX_LFS_FILESIZE;
 944         sb->s_magic = BTRFS_SUPER_MAGIC;
 945         sb->s_op = &btrfs_super_ops;
 946         sb->s_d_op = &btrfs_dentry_operations;
 947         sb->s_export_op = &btrfs_export_ops;
 948         sb->s_xattr = btrfs_xattr_handlers;
 949         sb->s_time_gran = 1;
 950 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 951         sb->s_flags |= MS_POSIXACL;
 952 #endif
 953         sb->s_flags |= MS_I_VERSION;
 954         err = open_ctree(sb, fs_devices, (char *)data);
 955         if (err) {
 956                 printk(KERN_ERR "BTRFS: open_ctree failed\n");
 957                 return err;
 958         }
 959
 960         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 961         key.type = BTRFS_INODE_ITEM_KEY;
 962         key.offset = 0;
 963         inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
 964         if (IS_ERR(inode)) {
 965                 err = PTR_ERR(inode);
 966                 goto fail_close;
 967         }
 968
 969         sb->s_root = d_make_root(inode);
 970         if (!sb->s_root) {
 971                 err = -ENOMEM;
 972                 goto fail_close;
 973         }
 974
 975         save_mount_options(sb, data);
 976         cleancache_init_fs(sb);
 977         sb->s_flags |= MS_ACTIVE;
 978         return 0;
 979
 980 fail_close:
 981         close_ctree(fs_info->tree_root);
 982         return err;
 983 }
 984
 985 int btrfs_sync_fs(struct super_block *sb, int wait)
 986 {
 987         struct btrfs_trans_handle *trans;
 988         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 989         struct btrfs_root *root = fs_info->tree_root;
 990
 991         trace_btrfs_sync_fs(wait);
 992
 993         if (!wait) {
 994                 filemap_flush(fs_info->btree_inode->i_mapping);
 995                 return 0;
 996         }
 997
 998         btrfs_wait_ordered_roots(fs_info, -1);
 999
1000         trans = btrfs_attach_transaction_barrier(root);
1001         if (IS_ERR(trans)) {
1002                 /* no transaction, don't bother */
1003                 if (PTR_ERR(trans) == -ENOENT)
1004                         return 0;
1005                 return PTR_ERR(trans);
1006         }
1007         return btrfs_commit_transaction(trans, root);
1008 }
1009
1010 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1011 {
1012         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1013         struct btrfs_root *root = info->tree_root;
1014         char *compress_type;
1015
1016         if (btrfs_test_opt(root, DEGRADED))
1017                 seq_puts(seq, ",degraded");
1018         if (btrfs_test_opt(root, NODATASUM))
1019                 seq_puts(seq, ",nodatasum");
1020         if (btrfs_test_opt(root, NODATACOW))
1021                 seq_puts(seq, ",nodatacow");
1022         if (btrfs_test_opt(root, NOBARRIER))
1023                 seq_puts(seq, ",nobarrier");
1024         if (info->max_inline != 8192 * 1024)
1025                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1026         if (info->alloc_start != 0)
1027                 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1028         if (info->thread_pool_size !=  min_t(unsigned long,
1029                                              num_online_cpus() + 2, 8))
1030                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1031         if (btrfs_test_opt(root, COMPRESS)) {
1032                 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1033                         compress_type = "zlib";
1034                 else
1035                         compress_type = "lzo";
1036                 if (btrfs_test_opt(root, FORCE_COMPRESS))
1037                         seq_printf(seq, ",compress-force=%s", compress_type);
1038                 else
1039                         seq_printf(seq, ",compress=%s", compress_type);
1040         }
1041         if (btrfs_test_opt(root, NOSSD))
1042                 seq_puts(seq, ",nossd");
1043         if (btrfs_test_opt(root, SSD_SPREAD))
1044                 seq_puts(seq, ",ssd_spread");
1045         else if (btrfs_test_opt(root, SSD))
1046                 seq_puts(seq, ",ssd");
1047         if (btrfs_test_opt(root, NOTREELOG))
1048                 seq_puts(seq, ",notreelog");
1049         if (btrfs_test_opt(root, FLUSHONCOMMIT))
1050                 seq_puts(seq, ",flushoncommit");
1051         if (btrfs_test_opt(root, DISCARD))
1052                 seq_puts(seq, ",discard");
1053         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1054                 seq_puts(seq, ",noacl");
1055         if (btrfs_test_opt(root, SPACE_CACHE))
1056                 seq_puts(seq, ",space_cache");
1057         else
1058                 seq_puts(seq, ",nospace_cache");
1059         if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1060                 seq_puts(seq, ",rescan_uuid_tree");
1061         if (btrfs_test_opt(root, CLEAR_CACHE))
1062                 seq_puts(seq, ",clear_cache");
1063         if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1064                 seq_puts(seq, ",user_subvol_rm_allowed");
1065         if (btrfs_test_opt(root, ENOSPC_DEBUG))
1066                 seq_puts(seq, ",enospc_debug");
1067         if (btrfs_test_opt(root, AUTO_DEFRAG))
1068                 seq_puts(seq, ",autodefrag");
1069         if (btrfs_test_opt(root, INODE_MAP_CACHE))
1070                 seq_puts(seq, ",inode_cache");
1071         if (btrfs_test_opt(root, SKIP_BALANCE))
1072                 seq_puts(seq, ",skip_balance");
1073         if (btrfs_test_opt(root, RECOVERY))
1074                 seq_puts(seq, ",recovery");
1075 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1076         if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1077                 seq_puts(seq, ",check_int_data");
1078         else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1079                 seq_puts(seq, ",check_int");
1080         if (info->check_integrity_print_mask)
1081                 seq_printf(seq, ",check_int_print_mask=%d",
1082                                 info->check_integrity_print_mask);
1083 #endif
1084         if (info->metadata_ratio)
1085                 seq_printf(seq, ",metadata_ratio=%d",
1086                                 info->metadata_ratio);
1087         if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1088                 seq_puts(seq, ",fatal_errors=panic");
1089         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1090                 seq_printf(seq, ",commit=%d", info->commit_interval);
1091         return 0;
1092 }
1093
1094 static int btrfs_test_super(struct super_block *s, void *data)
1095 {
1096         struct btrfs_fs_info *p = data;
1097         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1098
1099         return fs_info->fs_devices == p->fs_devices;
1100 }
1101
1102 static int btrfs_set_super(struct super_block *s, void *data)
1103 {
1104         int err = set_anon_super(s, data);
1105         if (!err)
1106                 s->s_fs_info = data;
1107         return err;
1108 }
1109
1110 /*
1111  * subvolumes are identified by ino 256
1112  */
1113 static inline int is_subvolume_inode(struct inode *inode)
1114 {
1115         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1116                 return 1;
1117         return 0;
1118 }
1119
1120 /*
1121  * This will strip out the subvol=%s argument for an argument string and add
1122  * subvolid=0 to make sure we get the actual tree root for path walking to the
1123  * subvol we want.
1124  */
1125 static char *setup_root_args(char *args)
1126 {
1127         unsigned len = strlen(args) + 2 + 1;
1128         char *src, *dst, *buf;
1129
1130         /*
1131          * We need the same args as before, but with this substitution:
1132          * s!subvol=[^,]+!subvolid=0!
1133          *
1134          * Since the replacement string is up to 2 bytes longer than the
1135          * original, allocate strlen(args) + 2 + 1 bytes.
1136          */
1137
1138         src = strstr(args, "subvol=");
1139         /* This shouldn't happen, but just in case.. */
1140         if (!src)
1141                 return NULL;
1142
1143         buf = dst = kmalloc(len, GFP_NOFS);
1144         if (!buf)
1145                 return NULL;
1146
1147         /*
1148          * If the subvol= arg is not at the start of the string,
1149          * copy whatever precedes it into buf.
1150          */
1151         if (src != args) {
1152                 *src++ = '\0';
1153                 strcpy(buf, args);
1154                 dst += strlen(args);
1155         }
1156
1157         strcpy(dst, "subvolid=0");
1158         dst += strlen("subvolid=0");
1159
1160         /*
1161          * If there is a "," after the original subvol=... string,
1162          * copy that suffix into our buffer.  Otherwise, we're done.
1163          */
1164         src = strchr(src, ',');
1165         if (src)
1166                 strcpy(dst, src);
1167
1168         return buf;
1169 }
1170
1171 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1172                                    const char *device_name, char *data)
1173 {
1174         struct dentry *root;
1175         struct vfsmount *mnt;
1176         char *newargs;
1177
1178         newargs = setup_root_args(data);
1179         if (!newargs)
1180                 return ERR_PTR(-ENOMEM);
1181         mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1182                              newargs);
1183
1184         if (PTR_RET(mnt) == -EBUSY) {
1185                 if (flags & MS_RDONLY) {
1186                         mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY, device_name,
1187                                              newargs);
1188                 } else {
1189                         int r;
1190                         mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY, device_name,
1191                                              newargs);
1192                         if (IS_ERR(mnt)) {
1193                                 kfree(newargs);
1194                                 return ERR_CAST(mnt);
1195                         }
1196
1197                         r = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1198                         if (r < 0) {
1199                                 /* FIXME: release vfsmount mnt ??*/
1200                                 kfree(newargs);
1201                                 return ERR_PTR(r);
1202                         }
1203                 }
1204         }
1205
1206         kfree(newargs);
1207
1208         if (IS_ERR(mnt))
1209                 return ERR_CAST(mnt);
1210
1211         root = mount_subtree(mnt, subvol_name);
1212
1213         if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1214                 struct super_block *s = root->d_sb;
1215                 dput(root);
1216                 root = ERR_PTR(-EINVAL);
1217                 deactivate_locked_super(s);
1218                 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1219                                 subvol_name);
1220         }
1221
1222         return root;
1223 }
1224
1225 /*
1226  * Find a superblock for the given device / mount point.
1227  *
1228  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
1229  *        for multiple device setup.  Make sure to keep it in sync.
1230  */
1231 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1232                 const char *device_name, void *data)
1233 {
1234         struct block_device *bdev = NULL;
1235         struct super_block *s;
1236         struct dentry *root;
1237         struct btrfs_fs_devices *fs_devices = NULL;
1238         struct btrfs_fs_info *fs_info = NULL;
1239         fmode_t mode = FMODE_READ;
1240         char *subvol_name = NULL;
1241         u64 subvol_objectid = 0;
1242         int error = 0;
1243
1244         if (!(flags & MS_RDONLY))
1245                 mode |= FMODE_WRITE;
1246
1247         error = btrfs_parse_early_options(data, mode, fs_type,
1248                                           &subvol_name, &subvol_objectid,
1249                                           &fs_devices);
1250         if (error) {
1251                 kfree(subvol_name);
1252                 return ERR_PTR(error);
1253         }
1254
1255         if (subvol_name) {
1256                 root = mount_subvol(subvol_name, flags, device_name, data);
1257                 kfree(subvol_name);
1258                 return root;
1259         }
1260
1261         error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1262         if (error)
1263                 return ERR_PTR(error);
1264
1265         /*
1266          * Setup a dummy root and fs_info for test/set super.  This is because
1267          * we don't actually fill this stuff out until open_ctree, but we need
1268          * it for searching for existing supers, so this lets us do that and
1269          * then open_ctree will properly initialize everything later.
1270          */
1271         fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1272         if (!fs_info)
1273                 return ERR_PTR(-ENOMEM);
1274
1275         fs_info->fs_devices = fs_devices;
1276
1277         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1278         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1279         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1280                 error = -ENOMEM;
1281                 goto error_fs_info;
1282         }
1283
1284         error = btrfs_open_devices(fs_devices, mode, fs_type);
1285         if (error)
1286                 goto error_fs_info;
1287
1288         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1289                 error = -EACCES;
1290                 goto error_close_devices;
1291         }
1292
1293         bdev = fs_devices->latest_bdev;
1294         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1295                  fs_info);
1296         if (IS_ERR(s)) {
1297                 error = PTR_ERR(s);
1298                 goto error_close_devices;
1299         }
1300
1301         if (s->s_root) {
1302                 btrfs_close_devices(fs_devices);
1303                 free_fs_info(fs_info);
1304                 if ((flags ^ s->s_flags) & MS_RDONLY)
1305                         error = -EBUSY;
1306         } else {
1307                 char b[BDEVNAME_SIZE];
1308
1309                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1310                 btrfs_sb(s)->bdev_holder = fs_type;
1311                 error = btrfs_fill_super(s, fs_devices, data,
1312                                          flags & MS_SILENT ? 1 : 0);
1313         }
1314
1315         root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1316         if (IS_ERR(root))
1317                 deactivate_locked_super(s);
1318
1319         return root;
1320
1321 error_close_devices:
1322         btrfs_close_devices(fs_devices);
1323 error_fs_info:
1324         free_fs_info(fs_info);
1325         return ERR_PTR(error);
1326 }
1327
1328 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1329                                      int new_pool_size, int old_pool_size)
1330 {
1331         if (new_pool_size == old_pool_size)
1332                 return;
1333
1334         fs_info->thread_pool_size = new_pool_size;
1335
1336         btrfs_info(fs_info, "resize thread pool %d -> %d",
1337                old_pool_size, new_pool_size);
1338
1339         btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1340         btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1341         btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1342         btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1343         btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1344         btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1345         btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1346                                 new_pool_size);
1347         btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1348         btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1349         btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1350         btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1351         btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1352                                 new_pool_size);
1353 }
1354
1355 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1356 {
1357         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1358 }
1359
1360 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1361                                        unsigned long old_opts, int flags)
1362 {
1363         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1364             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1365              (flags & MS_RDONLY))) {
1366                 /* wait for any defraggers to finish */
1367                 wait_event(fs_info->transaction_wait,
1368                            (atomic_read(&fs_info->defrag_running) == 0));
1369                 if (flags & MS_RDONLY)
1370                         sync_filesystem(fs_info->sb);
1371         }
1372 }
1373
1374 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1375                                          unsigned long old_opts)
1376 {
1377         /*
1378          * We need cleanup all defragable inodes if the autodefragment is
1379          * close or the fs is R/O.
1380          */
1381         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1382             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1383              (fs_info->sb->s_flags & MS_RDONLY))) {
1384                 btrfs_cleanup_defrag_inodes(fs_info);
1385         }
1386
1387         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1388 }
1389
1390 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1391 {
1392         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1393         struct btrfs_root *root = fs_info->tree_root;
1394         unsigned old_flags = sb->s_flags;
1395         unsigned long old_opts = fs_info->mount_opt;
1396         unsigned long old_compress_type = fs_info->compress_type;
1397         u64 old_max_inline = fs_info->max_inline;
1398         u64 old_alloc_start = fs_info->alloc_start;
1399         int old_thread_pool_size = fs_info->thread_pool_size;
1400         unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1401         int ret;
1402
1403         sync_filesystem(sb);
1404         btrfs_remount_prepare(fs_info);
1405
1406         ret = btrfs_parse_options(root, data);
1407         if (ret) {
1408                 ret = -EINVAL;
1409                 goto restore;
1410         }
1411
1412         btrfs_remount_begin(fs_info, old_opts, *flags);
1413         btrfs_resize_thread_pool(fs_info,
1414                 fs_info->thread_pool_size, old_thread_pool_size);
1415
1416         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1417                 goto out;
1418
1419         if (*flags & MS_RDONLY) {
1420                 /*
1421                  * this also happens on 'umount -rf' or on shutdown, when
1422                  * the filesystem is busy.
1423                  */
1424                 cancel_work_sync(&fs_info->async_reclaim_work);
1425
1426                 /* wait for the uuid_scan task to finish */
1427                 down(&fs_info->uuid_tree_rescan_sem);
1428                 /* avoid complains from lockdep et al. */
1429                 up(&fs_info->uuid_tree_rescan_sem);
1430
1431                 sb->s_flags |= MS_RDONLY;
1432
1433                 btrfs_dev_replace_suspend_for_unmount(fs_info);
1434                 btrfs_scrub_cancel(fs_info);
1435                 btrfs_pause_balance(fs_info);
1436
1437                 ret = btrfs_commit_super(root);
1438                 if (ret)
1439                         goto restore;
1440         } else {
1441                 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1442                         btrfs_err(fs_info,
1443                                 "Remounting read-write after error is not allowed");
1444                         ret = -EINVAL;
1445                         goto restore;
1446                 }
1447                 if (fs_info->fs_devices->rw_devices == 0) {
1448                         ret = -EACCES;
1449                         goto restore;
1450                 }
1451
1452                 if (fs_info->fs_devices->missing_devices >
1453                      fs_info->num_tolerated_disk_barrier_failures &&
1454                     !(*flags & MS_RDONLY)) {
1455                         btrfs_warn(fs_info,
1456                                 "too many missing devices, writeable remount is not allowed");
1457                         ret = -EACCES;
1458                         goto restore;
1459                 }
1460
1461                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1462                         ret = -EINVAL;
1463                         goto restore;
1464                 }
1465
1466                 ret = btrfs_cleanup_fs_roots(fs_info);
1467                 if (ret)
1468                         goto restore;
1469
1470                 /* recover relocation */
1471                 mutex_lock(&fs_info->cleaner_mutex);
1472                 ret = btrfs_recover_relocation(root);
1473                 mutex_unlock(&fs_info->cleaner_mutex);
1474                 if (ret)
1475                         goto restore;
1476
1477                 ret = btrfs_resume_balance_async(fs_info);
1478                 if (ret)
1479                         goto restore;
1480
1481                 ret = btrfs_resume_dev_replace_async(fs_info);
1482                 if (ret) {
1483                         btrfs_warn(fs_info, "failed to resume dev_replace");
1484                         goto restore;
1485                 }
1486
1487                 if (!fs_info->uuid_root) {
1488                         btrfs_info(fs_info, "creating UUID tree");
1489                         ret = btrfs_create_uuid_tree(fs_info);
1490                         if (ret) {
1491                                 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1492                                 goto restore;
1493                         }
1494                 }
1495                 sb->s_flags &= ~MS_RDONLY;
1496         }
1497 out:
1498         wake_up_process(fs_info->transaction_kthread);
1499         btrfs_remount_cleanup(fs_info, old_opts);
1500         return 0;
1501
1502 restore:
1503         /* We've hit an error - don't reset MS_RDONLY */
1504         if (sb->s_flags & MS_RDONLY)
1505                 old_flags |= MS_RDONLY;
1506         sb->s_flags = old_flags;
1507         fs_info->mount_opt = old_opts;
1508         fs_info->compress_type = old_compress_type;
1509         fs_info->max_inline = old_max_inline;
1510         mutex_lock(&fs_info->chunk_mutex);
1511         fs_info->alloc_start = old_alloc_start;
1512         mutex_unlock(&fs_info->chunk_mutex);
1513         btrfs_resize_thread_pool(fs_info,
1514                 old_thread_pool_size, fs_info->thread_pool_size);
1515         fs_info->metadata_ratio = old_metadata_ratio;
1516         btrfs_remount_cleanup(fs_info, old_opts);
1517         return ret;
1518 }
1519
1520 /* Used to sort the devices by max_avail(descending sort) */
1521 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1522                                        const void *dev_info2)
1523 {
1524         if (((struct btrfs_device_info *)dev_info1)->max_avail >
1525             ((struct btrfs_device_info *)dev_info2)->max_avail)
1526                 return -1;
1527         else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1528                  ((struct btrfs_device_info *)dev_info2)->max_avail)
1529                 return 1;
1530         else
1531         return 0;
1532 }
1533
1534 /*
1535  * sort the devices by max_avail, in which max free extent size of each device
1536  * is stored.(Descending Sort)
1537  */
1538 static inline void btrfs_descending_sort_devices(
1539                                         struct btrfs_device_info *devices,
1540                                         size_t nr_devices)
1541 {
1542         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1543              btrfs_cmp_device_free_bytes, NULL);
1544 }
1545
1546 /*
1547  * The helper to calc the free space on the devices that can be used to store
1548  * file data.
1549  */
1550 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1551 {
1552         struct btrfs_fs_info *fs_info = root->fs_info;
1553         struct btrfs_device_info *devices_info;
1554         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1555         struct btrfs_device *device;
1556         u64 skip_space;
1557         u64 type;
1558         u64 avail_space;
1559         u64 used_space;
1560         u64 min_stripe_size;
1561         int min_stripes = 1, num_stripes = 1;
1562         int i = 0, nr_devices;
1563         int ret;
1564
1565         nr_devices = fs_info->fs_devices->open_devices;
1566         BUG_ON(!nr_devices);
1567
1568         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1569                                GFP_NOFS);
1570         if (!devices_info)
1571                 return -ENOMEM;
1572
1573         /* calc min stripe number for data space alloction */
1574         type = btrfs_get_alloc_profile(root, 1);
1575         if (type & BTRFS_BLOCK_GROUP_RAID0) {
1576                 min_stripes = 2;
1577                 num_stripes = nr_devices;
1578         } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1579                 min_stripes = 2;
1580                 num_stripes = 2;
1581         } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1582                 min_stripes = 4;
1583                 num_stripes = 4;
1584         }
1585
1586         if (type & BTRFS_BLOCK_GROUP_DUP)
1587                 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1588         else
1589                 min_stripe_size = BTRFS_STRIPE_LEN;
1590
1591         list_for_each_entry(device, &fs_devices->devices, dev_list) {
1592                 if (!device->in_fs_metadata || !device->bdev ||
1593                     device->is_tgtdev_for_dev_replace)
1594                         continue;
1595
1596                 avail_space = device->total_bytes - device->bytes_used;
1597
1598                 /* align with stripe_len */
1599                 do_div(avail_space, BTRFS_STRIPE_LEN);
1600                 avail_space *= BTRFS_STRIPE_LEN;
1601
1602                 /*
1603                  * In order to avoid overwritting the superblock on the drive,
1604                  * btrfs starts at an offset of at least 1MB when doing chunk
1605                  * allocation.
1606                  */
1607                 skip_space = 1024 * 1024;
1608
1609                 /* user can set the offset in fs_info->alloc_start. */
1610                 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1611                     device->total_bytes)
1612                         skip_space = max(fs_info->alloc_start, skip_space);
1613
1614                 /*
1615                  * btrfs can not use the free space in [0, skip_space - 1],
1616                  * we must subtract it from the total. In order to implement
1617                  * it, we account the used space in this range first.
1618                  */
1619                 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1620                                                      &used_space);
1621                 if (ret) {
1622                         kfree(devices_info);
1623                         return ret;
1624                 }
1625
1626                 /* calc the free space in [0, skip_space - 1] */
1627                 skip_space -= used_space;
1628
1629                 /*
1630                  * we can use the free space in [0, skip_space - 1], subtract
1631                  * it from the total.
1632                  */
1633                 if (avail_space && avail_space >= skip_space)
1634                         avail_space -= skip_space;
1635                 else
1636                         avail_space = 0;
1637
1638                 if (avail_space < min_stripe_size)
1639                         continue;
1640
1641                 devices_info[i].dev = device;
1642                 devices_info[i].max_avail = avail_space;
1643
1644                 i++;
1645         }
1646
1647         nr_devices = i;
1648
1649         btrfs_descending_sort_devices(devices_info, nr_devices);
1650
1651         i = nr_devices - 1;
1652         avail_space = 0;
1653         while (nr_devices >= min_stripes) {
1654                 if (num_stripes > nr_devices)
1655                         num_stripes = nr_devices;
1656
1657                 if (devices_info[i].max_avail >= min_stripe_size) {
1658                         int j;
1659                         u64 alloc_size;
1660
1661                         avail_space += devices_info[i].max_avail * num_stripes;
1662                         alloc_size = devices_info[i].max_avail;
1663                         for (j = i + 1 - num_stripes; j <= i; j++)
1664                                 devices_info[j].max_avail -= alloc_size;
1665                 }
1666                 i--;
1667                 nr_devices--;
1668         }
1669
1670         kfree(devices_info);
1671         *free_bytes = avail_space;
1672         return 0;
1673 }
1674
1675 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1676 {
1677         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1678         struct btrfs_super_block *disk_super = fs_info->super_copy;
1679         struct list_head *head = &fs_info->space_info;
1680         struct btrfs_space_info *found;
1681         u64 total_used = 0;
1682         u64 total_free_data = 0;
1683         int bits = dentry->d_sb->s_blocksize_bits;
1684         __be32 *fsid = (__be32 *)fs_info->fsid;
1685         int ret;
1686
1687         /* holding chunk_muext to avoid allocating new chunks */
1688         mutex_lock(&fs_info->chunk_mutex);
1689         rcu_read_lock();
1690         list_for_each_entry_rcu(found, head, list) {
1691                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1692                         total_free_data += found->disk_total - found->disk_used;
1693                         total_free_data -=
1694                                 btrfs_account_ro_block_groups_free_space(found);
1695                 }
1696
1697                 total_used += found->disk_used;
1698         }
1699         rcu_read_unlock();
1700
1701         buf->f_namelen = BTRFS_NAME_LEN;
1702         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1703         buf->f_bfree = buf->f_blocks - (total_used >> bits);
1704         buf->f_bsize = dentry->d_sb->s_blocksize;
1705         buf->f_type = BTRFS_SUPER_MAGIC;
1706         buf->f_bavail = total_free_data;
1707         ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1708         if (ret) {
1709                 mutex_unlock(&fs_info->chunk_mutex);
1710                 return ret;
1711         }
1712         buf->f_bavail += total_free_data;
1713         buf->f_bavail = buf->f_bavail >> bits;
1714         mutex_unlock(&fs_info->chunk_mutex);
1715
1716         /* We treat it as constant endianness (it doesn't matter _which_)
1717            because we want the fsid to come out the same whether mounted
1718            on a big-endian or little-endian host */
1719         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1720         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1721         /* Mask in the root object ID too, to disambiguate subvols */
1722         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1723         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1724
1725         return 0;
1726 }
1727
1728 static void btrfs_kill_super(struct super_block *sb)
1729 {
1730         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1731         kill_anon_super(sb);
1732         free_fs_info(fs_info);
1733 }
1734
1735 static struct file_system_type btrfs_fs_type = {
1736         .owner          = THIS_MODULE,
1737         .name           = "btrfs",
1738         .mount          = btrfs_mount,
1739         .kill_sb        = btrfs_kill_super,
1740         .fs_flags       = FS_REQUIRES_DEV,
1741 };
1742 MODULE_ALIAS_FS("btrfs");
1743
1744 /*
1745  * used by btrfsctl to scan devices when no FS is mounted
1746  */
1747 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1748                                 unsigned long arg)
1749 {
1750         struct btrfs_ioctl_vol_args *vol;
1751         struct btrfs_fs_devices *fs_devices;
1752         int ret = -ENOTTY;
1753
1754         if (!capable(CAP_SYS_ADMIN))
1755                 return -EPERM;
1756
1757         vol = memdup_user((void __user *)arg, sizeof(*vol));
1758         if (IS_ERR(vol))
1759                 return PTR_ERR(vol);
1760
1761         switch (cmd) {
1762         case BTRFS_IOC_SCAN_DEV:
1763                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1764                                             &btrfs_fs_type, &fs_devices);
1765                 break;
1766         case BTRFS_IOC_DEVICES_READY:
1767                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1768                                             &btrfs_fs_type, &fs_devices);
1769                 if (ret)
1770                         break;
1771                 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1772                 break;
1773         }
1774
1775         kfree(vol);
1776         return ret;
1777 }
1778
1779 static int btrfs_freeze(struct super_block *sb)
1780 {
1781         struct btrfs_trans_handle *trans;
1782         struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1783
1784         trans = btrfs_attach_transaction_barrier(root);
1785         if (IS_ERR(trans)) {
1786                 /* no transaction, don't bother */
1787                 if (PTR_ERR(trans) == -ENOENT)
1788                         return 0;
1789                 return PTR_ERR(trans);
1790         }
1791         return btrfs_commit_transaction(trans, root);
1792 }
1793
1794 static int btrfs_unfreeze(struct super_block *sb)
1795 {
1796         return 0;
1797 }
1798
1799 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1800 {
1801         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1802         struct btrfs_fs_devices *cur_devices;
1803         struct btrfs_device *dev, *first_dev = NULL;
1804         struct list_head *head;
1805         struct rcu_string *name;
1806
1807         mutex_lock(&fs_info->fs_devices->device_list_mutex);
1808         cur_devices = fs_info->fs_devices;
1809         while (cur_devices) {
1810                 head = &cur_devices->devices;
1811                 list_for_each_entry(dev, head, dev_list) {
1812                         if (dev->missing)
1813                                 continue;
1814                         if (!dev->name)
1815                                 continue;
1816                         if (!first_dev || dev->devid < first_dev->devid)
1817                                 first_dev = dev;
1818                 }
1819                 cur_devices = cur_devices->seed;
1820         }
1821
1822         if (first_dev) {
1823                 rcu_read_lock();
1824                 name = rcu_dereference(first_dev->name);
1825                 seq_escape(m, name->str, " \t\n\\");
1826                 rcu_read_unlock();
1827         } else {
1828                 WARN_ON(1);
1829         }
1830         mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1831         return 0;
1832 }
1833
1834 static const struct super_operations btrfs_super_ops = {
1835         .drop_inode     = btrfs_drop_inode,
1836         .evict_inode    = btrfs_evict_inode,
1837         .put_super      = btrfs_put_super,
1838         .sync_fs        = btrfs_sync_fs,
1839         .show_options   = btrfs_show_options,
1840         .show_devname   = btrfs_show_devname,
1841         .write_inode    = btrfs_write_inode,
1842         .alloc_inode    = btrfs_alloc_inode,
1843         .destroy_inode  = btrfs_destroy_inode,
1844         .statfs         = btrfs_statfs,
1845         .remount_fs     = btrfs_remount,
1846         .freeze_fs      = btrfs_freeze,
1847         .unfreeze_fs    = btrfs_unfreeze,
1848 };
1849
1850 static const struct file_operations btrfs_ctl_fops = {
1851         .unlocked_ioctl  = btrfs_control_ioctl,
1852         .compat_ioctl = btrfs_control_ioctl,
1853         .owner   = THIS_MODULE,
1854         .llseek = noop_llseek,
1855 };
1856
1857 static struct miscdevice btrfs_misc = {
1858         .minor          = BTRFS_MINOR,
1859         .name           = "btrfs-control",
1860         .fops           = &btrfs_ctl_fops
1861 };
1862
1863 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1864 MODULE_ALIAS("devname:btrfs-control");
1865
1866 static int btrfs_interface_init(void)
1867 {
1868         return misc_register(&btrfs_misc);
1869 }
1870
1871 static void btrfs_interface_exit(void)
1872 {
1873         if (misc_deregister(&btrfs_misc) < 0)
1874                 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
1875 }
1876
1877 static void btrfs_print_info(void)
1878 {
1879         printk(KERN_INFO "Btrfs loaded"
1880 #ifdef CONFIG_BTRFS_DEBUG
1881                         ", debug=on"
1882 #endif
1883 #ifdef CONFIG_BTRFS_ASSERT
1884                         ", assert=on"
1885 #endif
1886 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1887                         ", integrity-checker=on"
1888 #endif
1889                         "\n");
1890 }
1891
1892 static int btrfs_run_sanity_tests(void)
1893 {
1894         int ret;
1895
1896         ret = btrfs_init_test_fs();
1897         if (ret)
1898                 return ret;
1899
1900         ret = btrfs_test_free_space_cache();
1901         if (ret)
1902                 goto out;
1903         ret = btrfs_test_extent_buffer_operations();
1904         if (ret)
1905                 goto out;
1906         ret = btrfs_test_extent_io();
1907         if (ret)
1908                 goto out;
1909         ret = btrfs_test_inodes();
1910         if (ret)
1911                 goto out;
1912         ret = btrfs_test_qgroups();
1913 out:
1914         btrfs_destroy_test_fs();
1915         return ret;
1916 }
1917
1918 static int __init init_btrfs_fs(void)
1919 {
1920         int err;
1921
1922         err = btrfs_hash_init();
1923         if (err)
1924                 return err;
1925
1926         btrfs_props_init();
1927
1928         err = btrfs_init_sysfs();
1929         if (err)
1930                 goto free_hash;
1931
1932         btrfs_init_compress();
1933
1934         err = btrfs_init_cachep();
1935         if (err)
1936                 goto free_compress;
1937
1938         err = extent_io_init();
1939         if (err)
1940                 goto free_cachep;
1941
1942         err = extent_map_init();
1943         if (err)
1944                 goto free_extent_io;
1945
1946         err = ordered_data_init();
1947         if (err)
1948                 goto free_extent_map;
1949
1950         err = btrfs_delayed_inode_init();
1951         if (err)
1952                 goto free_ordered_data;
1953
1954         err = btrfs_auto_defrag_init();
1955         if (err)
1956                 goto free_delayed_inode;
1957
1958         err = btrfs_delayed_ref_init();
1959         if (err)
1960                 goto free_auto_defrag;
1961
1962         err = btrfs_prelim_ref_init();
1963         if (err)
1964                 goto free_prelim_ref;
1965
1966         err = btrfs_interface_init();
1967         if (err)
1968                 goto free_delayed_ref;
1969
1970         btrfs_init_lockdep();
1971
1972         btrfs_print_info();
1973
1974         err = btrfs_run_sanity_tests();
1975         if (err)
1976                 goto unregister_ioctl;
1977
1978         err = register_filesystem(&btrfs_fs_type);
1979         if (err)
1980                 goto unregister_ioctl;
1981
1982         return 0;
1983
1984 unregister_ioctl:
1985         btrfs_interface_exit();
1986 free_prelim_ref:
1987         btrfs_prelim_ref_exit();
1988 free_delayed_ref:
1989         btrfs_delayed_ref_exit();
1990 free_auto_defrag:
1991         btrfs_auto_defrag_exit();
1992 free_delayed_inode:
1993         btrfs_delayed_inode_exit();
1994 free_ordered_data:
1995         ordered_data_exit();
1996 free_extent_map:
1997         extent_map_exit();
1998 free_extent_io:
1999         extent_io_exit();
2000 free_cachep:
2001         btrfs_destroy_cachep();
2002 free_compress:
2003         btrfs_exit_compress();
2004         btrfs_exit_sysfs();
2005 free_hash:
2006         btrfs_hash_exit();
2007         return err;
2008 }
2009
2010 static void __exit exit_btrfs_fs(void)
2011 {
2012         btrfs_destroy_cachep();
2013         btrfs_delayed_ref_exit();
2014         btrfs_auto_defrag_exit();
2015         btrfs_delayed_inode_exit();
2016         btrfs_prelim_ref_exit();
2017         ordered_data_exit();
2018         extent_map_exit();
2019         extent_io_exit();
2020         btrfs_interface_exit();
2021         unregister_filesystem(&btrfs_fs_type);
2022         btrfs_exit_sysfs();
2023         btrfs_cleanup_fs_uuids();
2024         btrfs_exit_compress();
2025         btrfs_hash_exit();
2026 }
2027
2028 late_initcall(init_btrfs_fs);
2029 module_exit(exit_btrfs_fs)
2030
2031 MODULE_LICENSE("GPL");