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