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 "compat.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 "ioctl.h"
  51 #include "print-tree.h"
  52 #include "xattr.h"
  53 #include "volumes.h"
  54 #include "version.h"
  55 #include "export.h"
  56 #include "compression.h"
  57
  58 #define CREATE_TRACE_POINTS
  59 #include <trace/events/btrfs.h>
  60
  61 static const struct super_operations btrfs_super_ops;
  62 static struct file_system_type btrfs_fs_type;
  63
  64 static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
  65                                       char nbuf[16])
  66 {
  67         char *errstr = NULL;
  68
  69         switch (errno) {
  70         case -EIO:
  71                 errstr = "IO failure";
  72                 break;
  73         case -ENOMEM:
  74                 errstr = "Out of memory";
  75                 break;
  76         case -EROFS:
  77                 errstr = "Readonly filesystem";
  78                 break;
  79         default:
  80                 if (nbuf) {
  81                         if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
  82                                 errstr = nbuf;
  83                 }
  84                 break;
  85         }
  86
  87         return errstr;
  88 }
  89
  90 static void __save_error_info(struct btrfs_fs_info *fs_info)
  91 {
  92         /*
  93          * today we only save the error info into ram.  Long term we'll
  94          * also send it down to the disk
  95          */
  96         fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
  97 }
  98
  99 /* NOTE:
 100  *      We move write_super stuff at umount in order to avoid deadlock
 101  *      for umount hold all lock.
 102  */
 103 static void save_error_info(struct btrfs_fs_info *fs_info)
 104 {
 105         __save_error_info(fs_info);
 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 (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
 117                 sb->s_flags |= MS_RDONLY;
 118                 printk(KERN_INFO "btrfs is forced readonly\n");
 119         }
 120 }
 121
 122 /*
 123  * __btrfs_std_error decodes expected errors from the caller and
 124  * invokes the approciate error response.
 125  */
 126 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 127                      unsigned int line, int errno)
 128 {
 129         struct super_block *sb = fs_info->sb;
 130         char nbuf[16];
 131         const char *errstr;
 132
 133         /*
 134          * Special case: if the error is EROFS, and we're already
 135          * under MS_RDONLY, then it is safe here.
 136          */
 137         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 138                 return;
 139
 140         errstr = btrfs_decode_error(fs_info, errno, nbuf);
 141         printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
 142                 sb->s_id, function, line, errstr);
 143         save_error_info(fs_info);
 144
 145         btrfs_handle_error(fs_info);
 146 }
 147
 148 static void btrfs_put_super(struct super_block *sb)
 149 {
 150         (void)close_ctree(btrfs_sb(sb));
 151         /* FIXME: need to fix VFS to return error? */
 152         /* AV: return it _where_?  ->put_super() can be triggered by any number
 153          * of async events, up to and including delivery of SIGKILL to the
 154          * last process that kept it busy.  Or segfault in the aforementioned
 155          * process...  Whom would you report that to?
 156          */
 157 }
 158
 159 enum {
 160         Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
 161         Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
 162         Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
 163         Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
 164         Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
 165         Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
 166         Opt_enospc_debug, Opt_subvolrootid, Opt_defrag,
 167         Opt_inode_cache, Opt_no_space_cache, Opt_recovery, Opt_err,
 168 };
 169
 170 static match_table_t tokens = {
 171         {Opt_degraded, "degraded"},
 172         {Opt_subvol, "subvol=%s"},
 173         {Opt_subvolid, "subvolid=%d"},
 174         {Opt_device, "device=%s"},
 175         {Opt_nodatasum, "nodatasum"},
 176         {Opt_nodatacow, "nodatacow"},
 177         {Opt_nobarrier, "nobarrier"},
 178         {Opt_max_inline, "max_inline=%s"},
 179         {Opt_alloc_start, "alloc_start=%s"},
 180         {Opt_thread_pool, "thread_pool=%d"},
 181         {Opt_compress, "compress"},
 182         {Opt_compress_type, "compress=%s"},
 183         {Opt_compress_force, "compress-force"},
 184         {Opt_compress_force_type, "compress-force=%s"},
 185         {Opt_ssd, "ssd"},
 186         {Opt_ssd_spread, "ssd_spread"},
 187         {Opt_nossd, "nossd"},
 188         {Opt_noacl, "noacl"},
 189         {Opt_notreelog, "notreelog"},
 190         {Opt_flushoncommit, "flushoncommit"},
 191         {Opt_ratio, "metadata_ratio=%d"},
 192         {Opt_discard, "discard"},
 193         {Opt_space_cache, "space_cache"},
 194         {Opt_clear_cache, "clear_cache"},
 195         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 196         {Opt_enospc_debug, "enospc_debug"},
 197         {Opt_subvolrootid, "subvolrootid=%d"},
 198         {Opt_defrag, "autodefrag"},
 199         {Opt_inode_cache, "inode_cache"},
 200         {Opt_no_space_cache, "nospace_cache"},
 201         {Opt_recovery, "recovery"},
 202         {Opt_err, NULL},
 203 };
 204
 205 /*
 206  * Regular mount options parser.  Everything that is needed only when
 207  * reading in a new superblock is parsed here.
 208  */
 209 int btrfs_parse_options(struct btrfs_root *root, char *options)
 210 {
 211         struct btrfs_fs_info *info = root->fs_info;
 212         substring_t args[MAX_OPT_ARGS];
 213         char *p, *num, *orig = NULL;
 214         u64 cache_gen;
 215         int intarg;
 216         int ret = 0;
 217         char *compress_type;
 218         bool compress_force = false;
 219
 220         cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
 221         if (cache_gen)
 222                 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 223
 224         if (!options)
 225                 goto out;
 226
 227         /*
 228          * strsep changes the string, duplicate it because parse_options
 229          * gets called twice
 230          */
 231         options = kstrdup(options, GFP_NOFS);
 232         if (!options)
 233                 return -ENOMEM;
 234
 235         orig = options;
 236
 237         while ((p = strsep(&options, ",")) != NULL) {
 238                 int token;
 239                 if (!*p)
 240                         continue;
 241
 242                 token = match_token(p, tokens, args);
 243                 switch (token) {
 244                 case Opt_degraded:
 245                         printk(KERN_INFO "btrfs: allowing degraded mounts\n");
 246                         btrfs_set_opt(info->mount_opt, DEGRADED);
 247                         break;
 248                 case Opt_subvol:
 249                 case Opt_subvolid:
 250                 case Opt_subvolrootid:
 251                 case Opt_device:
 252                         /*
 253                          * These are parsed by btrfs_parse_early_options
 254                          * and can be happily ignored here.
 255                          */
 256                         break;
 257                 case Opt_nodatasum:
 258                         printk(KERN_INFO "btrfs: setting nodatasum\n");
 259                         btrfs_set_opt(info->mount_opt, NODATASUM);
 260                         break;
 261                 case Opt_nodatacow:
 262                         printk(KERN_INFO "btrfs: setting nodatacow\n");
 263                         btrfs_set_opt(info->mount_opt, NODATACOW);
 264                         btrfs_set_opt(info->mount_opt, NODATASUM);
 265                         break;
 266                 case Opt_compress_force:
 267                 case Opt_compress_force_type:
 268                         compress_force = true;
 269                 case Opt_compress:
 270                 case Opt_compress_type:
 271                         if (token == Opt_compress ||
 272                             token == Opt_compress_force ||
 273                             strcmp(args[0].from, "zlib") == 0) {
 274                                 compress_type = "zlib";
 275                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 276                         } else if (strcmp(args[0].from, "lzo") == 0) {
 277                                 compress_type = "lzo";
 278                                 info->compress_type = BTRFS_COMPRESS_LZO;
 279                         } else {
 280                                 ret = -EINVAL;
 281                                 goto out;
 282                         }
 283
 284                         btrfs_set_opt(info->mount_opt, COMPRESS);
 285                         if (compress_force) {
 286                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 287                                 pr_info("btrfs: force %s compression\n",
 288                                         compress_type);
 289                         } else
 290                                 pr_info("btrfs: use %s compression\n",
 291                                         compress_type);
 292                         break;
 293                 case Opt_ssd:
 294                         printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 295                         btrfs_set_opt(info->mount_opt, SSD);
 296                         break;
 297                 case Opt_ssd_spread:
 298                         printk(KERN_INFO "btrfs: use spread ssd "
 299                                "allocation scheme\n");
 300                         btrfs_set_opt(info->mount_opt, SSD);
 301                         btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 302                         break;
 303                 case Opt_nossd:
 304                         printk(KERN_INFO "btrfs: not using ssd allocation "
 305                                "scheme\n");
 306                         btrfs_set_opt(info->mount_opt, NOSSD);
 307                         btrfs_clear_opt(info->mount_opt, SSD);
 308                         btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 309                         break;
 310                 case Opt_nobarrier:
 311                         printk(KERN_INFO "btrfs: turning off barriers\n");
 312                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 313                         break;
 314                 case Opt_thread_pool:
 315                         intarg = 0;
 316                         match_int(&args[0], &intarg);
 317                         if (intarg) {
 318                                 info->thread_pool_size = intarg;
 319                                 printk(KERN_INFO "btrfs: thread pool %d\n",
 320                                        info->thread_pool_size);
 321                         }
 322                         break;
 323                 case Opt_max_inline:
 324                         num = match_strdup(&args[0]);
 325                         if (num) {
 326                                 info->max_inline = memparse(num, NULL);
 327                                 kfree(num);
 328
 329                                 if (info->max_inline) {
 330                                         info->max_inline = max_t(u64,
 331                                                 info->max_inline,
 332                                                 root->sectorsize);
 333                                 }
 334                                 printk(KERN_INFO "btrfs: max_inline at %llu\n",
 335                                         (unsigned long long)info->max_inline);
 336                         }
 337                         break;
 338                 case Opt_alloc_start:
 339                         num = match_strdup(&args[0]);
 340                         if (num) {
 341                                 info->alloc_start = memparse(num, NULL);
 342                                 kfree(num);
 343                                 printk(KERN_INFO
 344                                         "btrfs: allocations start at %llu\n",
 345                                         (unsigned long long)info->alloc_start);
 346                         }
 347                         break;
 348                 case Opt_noacl:
 349                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 350                         break;
 351                 case Opt_notreelog:
 352                         printk(KERN_INFO "btrfs: disabling tree log\n");
 353                         btrfs_set_opt(info->mount_opt, NOTREELOG);
 354                         break;
 355                 case Opt_flushoncommit:
 356                         printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
 357                         btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
 358                         break;
 359                 case Opt_ratio:
 360                         intarg = 0;
 361                         match_int(&args[0], &intarg);
 362                         if (intarg) {
 363                                 info->metadata_ratio = intarg;
 364                                 printk(KERN_INFO "btrfs: metadata ratio %d\n",
 365                                        info->metadata_ratio);
 366                         }
 367                         break;
 368                 case Opt_discard:
 369                         btrfs_set_opt(info->mount_opt, DISCARD);
 370                         break;
 371                 case Opt_space_cache:
 372                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 373                         break;
 374                 case Opt_no_space_cache:
 375                         printk(KERN_INFO "btrfs: disabling disk space caching\n");
 376                         btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
 377                         break;
 378                 case Opt_inode_cache:
 379                         printk(KERN_INFO "btrfs: enabling inode map caching\n");
 380                         btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
 381                         break;
 382                 case Opt_clear_cache:
 383                         printk(KERN_INFO "btrfs: force clearing of disk cache\n");
 384                         btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
 385                         break;
 386                 case Opt_user_subvol_rm_allowed:
 387                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 388                         break;
 389                 case Opt_enospc_debug:
 390                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 391                         break;
 392                 case Opt_defrag:
 393                         printk(KERN_INFO "btrfs: enabling auto defrag");
 394                         btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
 395                         break;
 396                 case Opt_recovery:
 397                         printk(KERN_INFO "btrfs: enabling auto recovery");
 398                         btrfs_set_opt(info->mount_opt, RECOVERY);
 399                         break;
 400                 case Opt_err:
 401                         printk(KERN_INFO "btrfs: unrecognized mount option "
 402                                "'%s'\n", p);
 403                         ret = -EINVAL;
 404                         goto out;
 405                 default:
 406                         break;
 407                 }
 408         }
 409 out:
 410         if (!ret && btrfs_test_opt(root, SPACE_CACHE))
 411                 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
 412         kfree(orig);
 413         return ret;
 414 }
 415
 416 /*
 417  * Parse mount options that are required early in the mount process.
 418  *
 419  * All other options will be parsed on much later in the mount process and
 420  * only when we need to allocate a new super block.
 421  */
 422 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 423                 void *holder, char **subvol_name, u64 *subvol_objectid,
 424                 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
 425 {
 426         substring_t args[MAX_OPT_ARGS];
 427         char *device_name, *opts, *orig, *p;
 428         int error = 0;
 429         int intarg;
 430
 431         if (!options)
 432                 return 0;
 433
 434         /*
 435          * strsep changes the string, duplicate it because parse_options
 436          * gets called twice
 437          */
 438         opts = kstrdup(options, GFP_KERNEL);
 439         if (!opts)
 440                 return -ENOMEM;
 441         orig = opts;
 442
 443         while ((p = strsep(&opts, ",")) != NULL) {
 444                 int token;
 445                 if (!*p)
 446                         continue;
 447
 448                 token = match_token(p, tokens, args);
 449                 switch (token) {
 450                 case Opt_subvol:
 451                         kfree(*subvol_name);
 452                         *subvol_name = match_strdup(&args[0]);
 453                         break;
 454                 case Opt_subvolid:
 455                         intarg = 0;
 456                         error = match_int(&args[0], &intarg);
 457                         if (!error) {
 458                                 /* we want the original fs_tree */
 459                                 if (!intarg)
 460                                         *subvol_objectid =
 461                                                 BTRFS_FS_TREE_OBJECTID;
 462                                 else
 463                                         *subvol_objectid = intarg;
 464                         }
 465                         break;
 466                 case Opt_subvolrootid:
 467                         intarg = 0;
 468                         error = match_int(&args[0], &intarg);
 469                         if (!error) {
 470                                 /* we want the original fs_tree */
 471                                 if (!intarg)
 472                                         *subvol_rootid =
 473                                                 BTRFS_FS_TREE_OBJECTID;
 474                                 else
 475                                         *subvol_rootid = intarg;
 476                         }
 477                         break;
 478                 case Opt_device:
 479                         device_name = match_strdup(&args[0]);
 480                         if (!device_name) {
 481                                 error = -ENOMEM;
 482                                 goto out;
 483                         }
 484                         error = btrfs_scan_one_device(device_name,
 485                                         flags, holder, fs_devices);
 486                         kfree(device_name);
 487                         if (error)
 488                                 goto out;
 489                         break;
 490                 default:
 491                         break;
 492                 }
 493         }
 494
 495 out:
 496         kfree(orig);
 497         return error;
 498 }
 499
 500 static struct dentry *get_default_root(struct super_block *sb,
 501                                        u64 subvol_objectid)
 502 {
 503         struct btrfs_root *root = sb->s_fs_info;
 504         struct btrfs_root *new_root;
 505         struct btrfs_dir_item *di;
 506         struct btrfs_path *path;
 507         struct btrfs_key location;
 508         struct inode *inode;
 509         u64 dir_id;
 510         int new = 0;
 511
 512         /*
 513          * We have a specific subvol we want to mount, just setup location and
 514          * go look up the root.
 515          */
 516         if (subvol_objectid) {
 517                 location.objectid = subvol_objectid;
 518                 location.type = BTRFS_ROOT_ITEM_KEY;
 519                 location.offset = (u64)-1;
 520                 goto find_root;
 521         }
 522
 523         path = btrfs_alloc_path();
 524         if (!path)
 525                 return ERR_PTR(-ENOMEM);
 526         path->leave_spinning = 1;
 527
 528         /*
 529          * Find the "default" dir item which points to the root item that we
 530          * will mount by default if we haven't been given a specific subvolume
 531          * to mount.
 532          */
 533         dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
 534         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 535         if (IS_ERR(di)) {
 536                 btrfs_free_path(path);
 537                 return ERR_CAST(di);
 538         }
 539         if (!di) {
 540                 /*
 541                  * Ok the default dir item isn't there.  This is weird since
 542                  * it's always been there, but don't freak out, just try and
 543                  * mount to root most subvolume.
 544                  */
 545                 btrfs_free_path(path);
 546                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 547                 new_root = root->fs_info->fs_root;
 548                 goto setup_root;
 549         }
 550
 551         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 552         btrfs_free_path(path);
 553
 554 find_root:
 555         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
 556         if (IS_ERR(new_root))
 557                 return ERR_CAST(new_root);
 558
 559         if (btrfs_root_refs(&new_root->root_item) == 0)
 560                 return ERR_PTR(-ENOENT);
 561
 562         dir_id = btrfs_root_dirid(&new_root->root_item);
 563 setup_root:
 564         location.objectid = dir_id;
 565         location.type = BTRFS_INODE_ITEM_KEY;
 566         location.offset = 0;
 567
 568         inode = btrfs_iget(sb, &location, new_root, &new);
 569         if (IS_ERR(inode))
 570                 return ERR_CAST(inode);
 571
 572         /*
 573          * If we're just mounting the root most subvol put the inode and return
 574          * a reference to the dentry.  We will have already gotten a reference
 575          * to the inode in btrfs_fill_super so we're good to go.
 576          */
 577         if (!new && sb->s_root->d_inode == inode) {
 578                 iput(inode);
 579                 return dget(sb->s_root);
 580         }
 581
 582         return d_obtain_alias(inode);
 583 }
 584
 585 static int btrfs_fill_super(struct super_block *sb,
 586                             struct btrfs_fs_devices *fs_devices,
 587                             void *data, int silent)
 588 {
 589         struct inode *inode;
 590         struct dentry *root_dentry;
 591         struct btrfs_root *tree_root;
 592         struct btrfs_fs_info *fs_info;
 593         struct btrfs_key key;
 594         int err;
 595
 596         sb->s_maxbytes = MAX_LFS_FILESIZE;
 597         sb->s_magic = BTRFS_SUPER_MAGIC;
 598         sb->s_op = &btrfs_super_ops;
 599         sb->s_d_op = &btrfs_dentry_operations;
 600         sb->s_export_op = &btrfs_export_ops;
 601         sb->s_xattr = btrfs_xattr_handlers;
 602         sb->s_time_gran = 1;
 603 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 604         sb->s_flags |= MS_POSIXACL;
 605 #endif
 606
 607         tree_root = open_ctree(sb, fs_devices, (char *)data);
 608
 609         if (IS_ERR(tree_root)) {
 610                 printk("btrfs: open_ctree failed\n");
 611                 return PTR_ERR(tree_root);
 612         }
 613         fs_info = tree_root->fs_info;
 614         sb->s_fs_info = tree_root;
 615
 616         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 617         key.type = BTRFS_INODE_ITEM_KEY;
 618         key.offset = 0;
 619         inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
 620         if (IS_ERR(inode)) {
 621                 err = PTR_ERR(inode);
 622                 goto fail_close;
 623         }
 624
 625         root_dentry = d_alloc_root(inode);
 626         if (!root_dentry) {
 627                 iput(inode);
 628                 err = -ENOMEM;
 629                 goto fail_close;
 630         }
 631
 632         sb->s_root = root_dentry;
 633
 634         save_mount_options(sb, data);
 635         cleancache_init_fs(sb);
 636         return 0;
 637
 638 fail_close:
 639         close_ctree(tree_root);
 640         free_fs_info(fs_info);
 641         return err;
 642 }
 643
 644 int btrfs_sync_fs(struct super_block *sb, int wait)
 645 {
 646         struct btrfs_trans_handle *trans;
 647         struct btrfs_root *root = btrfs_sb(sb);
 648         int ret;
 649
 650         trace_btrfs_sync_fs(wait);
 651
 652         if (!wait) {
 653                 filemap_flush(root->fs_info->btree_inode->i_mapping);
 654                 return 0;
 655         }
 656
 657         btrfs_start_delalloc_inodes(root, 0);
 658         btrfs_wait_ordered_extents(root, 0, 0);
 659
 660         trans = btrfs_start_transaction(root, 0);
 661         if (IS_ERR(trans))
 662                 return PTR_ERR(trans);
 663         ret = btrfs_commit_transaction(trans, root);
 664         return ret;
 665 }
 666
 667 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
 668 {
 669         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
 670         struct btrfs_fs_info *info = root->fs_info;
 671         char *compress_type;
 672
 673         if (btrfs_test_opt(root, DEGRADED))
 674                 seq_puts(seq, ",degraded");
 675         if (btrfs_test_opt(root, NODATASUM))
 676                 seq_puts(seq, ",nodatasum");
 677         if (btrfs_test_opt(root, NODATACOW))
 678                 seq_puts(seq, ",nodatacow");
 679         if (btrfs_test_opt(root, NOBARRIER))
 680                 seq_puts(seq, ",nobarrier");
 681         if (info->max_inline != 8192 * 1024)
 682                 seq_printf(seq, ",max_inline=%llu",
 683                            (unsigned long long)info->max_inline);
 684         if (info->alloc_start != 0)
 685                 seq_printf(seq, ",alloc_start=%llu",
 686                            (unsigned long long)info->alloc_start);
 687         if (info->thread_pool_size !=  min_t(unsigned long,
 688                                              num_online_cpus() + 2, 8))
 689                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
 690         if (btrfs_test_opt(root, COMPRESS)) {
 691                 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
 692                         compress_type = "zlib";
 693                 else
 694                         compress_type = "lzo";
 695                 if (btrfs_test_opt(root, FORCE_COMPRESS))
 696                         seq_printf(seq, ",compress-force=%s", compress_type);
 697                 else
 698                         seq_printf(seq, ",compress=%s", compress_type);
 699         }
 700         if (btrfs_test_opt(root, NOSSD))
 701                 seq_puts(seq, ",nossd");
 702         if (btrfs_test_opt(root, SSD_SPREAD))
 703                 seq_puts(seq, ",ssd_spread");
 704         else if (btrfs_test_opt(root, SSD))
 705                 seq_puts(seq, ",ssd");
 706         if (btrfs_test_opt(root, NOTREELOG))
 707                 seq_puts(seq, ",notreelog");
 708         if (btrfs_test_opt(root, FLUSHONCOMMIT))
 709                 seq_puts(seq, ",flushoncommit");
 710         if (btrfs_test_opt(root, DISCARD))
 711                 seq_puts(seq, ",discard");
 712         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
 713                 seq_puts(seq, ",noacl");
 714         if (btrfs_test_opt(root, SPACE_CACHE))
 715                 seq_puts(seq, ",space_cache");
 716         else
 717                 seq_puts(seq, ",nospace_cache");
 718         if (btrfs_test_opt(root, CLEAR_CACHE))
 719                 seq_puts(seq, ",clear_cache");
 720         if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
 721                 seq_puts(seq, ",user_subvol_rm_allowed");
 722         if (btrfs_test_opt(root, ENOSPC_DEBUG))
 723                 seq_puts(seq, ",enospc_debug");
 724         if (btrfs_test_opt(root, AUTO_DEFRAG))
 725                 seq_puts(seq, ",autodefrag");
 726         if (btrfs_test_opt(root, INODE_MAP_CACHE))
 727                 seq_puts(seq, ",inode_cache");
 728         return 0;
 729 }
 730
 731 static int btrfs_test_super(struct super_block *s, void *data)
 732 {
 733         struct btrfs_root *test_root = data;
 734         struct btrfs_root *root = btrfs_sb(s);
 735
 736         return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
 737 }
 738
 739 static int btrfs_set_super(struct super_block *s, void *data)
 740 {
 741         int err = set_anon_super(s, data);
 742         if (!err)
 743                 s->s_fs_info = data;
 744         return err;
 745 }
 746
 747 /*
 748  * subvolumes are identified by ino 256
 749  */
 750 static inline int is_subvolume_inode(struct inode *inode)
 751 {
 752         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
 753                 return 1;
 754         return 0;
 755 }
 756
 757 /*
 758  * This will strip out the subvol=%s argument for an argument string and add
 759  * subvolid=0 to make sure we get the actual tree root for path walking to the
 760  * subvol we want.
 761  */
 762 static char *setup_root_args(char *args)
 763 {
 764         unsigned copied = 0;
 765         unsigned len = strlen(args) + 2;
 766         char *pos;
 767         char *ret;
 768
 769         /*
 770          * We need the same args as before, but minus
 771          *
 772          * subvol=a
 773          *
 774          * and add
 775          *
 776          * subvolid=0
 777          *
 778          * which is a difference of 2 characters, so we allocate strlen(args) +
 779          * 2 characters.
 780          */
 781         ret = kzalloc(len * sizeof(char), GFP_NOFS);
 782         if (!ret)
 783                 return NULL;
 784         pos = strstr(args, "subvol=");
 785
 786         /* This shouldn't happen, but just in case.. */
 787         if (!pos) {
 788                 kfree(ret);
 789                 return NULL;
 790         }
 791
 792         /*
 793          * The subvol=<> arg is not at the front of the string, copy everybody
 794          * up to that into ret.
 795          */
 796         if (pos != args) {
 797                 *pos = '\0';
 798                 strcpy(ret, args);
 799                 copied += strlen(args);
 800                 pos++;
 801         }
 802
 803         strncpy(ret + copied, "subvolid=0", len - copied);
 804
 805         /* Length of subvolid=0 */
 806         copied += 10;
 807
 808         /*
 809          * If there is no , after the subvol= option then we know there's no
 810          * other options and we can just return.
 811          */
 812         pos = strchr(pos, ',');
 813         if (!pos)
 814                 return ret;
 815
 816         /* Copy the rest of the arguments into our buffer */
 817         strncpy(ret + copied, pos, len - copied);
 818         copied += strlen(pos);
 819
 820         return ret;
 821 }
 822
 823 static struct dentry *mount_subvol(const char *subvol_name, int flags,
 824                                    const char *device_name, char *data)
 825 {
 826         struct dentry *root;
 827         struct vfsmount *mnt;
 828         char *newargs;
 829
 830         newargs = setup_root_args(data);
 831         if (!newargs)
 832                 return ERR_PTR(-ENOMEM);
 833         mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
 834                              newargs);
 835         kfree(newargs);
 836         if (IS_ERR(mnt))
 837                 return ERR_CAST(mnt);
 838
 839         root = mount_subtree(mnt, subvol_name);
 840
 841         if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
 842                 struct super_block *s = root->d_sb;
 843                 dput(root);
 844                 root = ERR_PTR(-EINVAL);
 845                 deactivate_locked_super(s);
 846                 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
 847                                 subvol_name);
 848         }
 849
 850         return root;
 851 }
 852
 853 /*
 854  * Find a superblock for the given device / mount point.
 855  *
 856  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
 857  *        for multiple device setup.  Make sure to keep it in sync.
 858  */
 859 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
 860                 const char *device_name, void *data)
 861 {
 862         struct block_device *bdev = NULL;
 863         struct super_block *s;
 864         struct dentry *root;
 865         struct btrfs_fs_devices *fs_devices = NULL;
 866         struct btrfs_fs_info *fs_info = NULL;
 867         fmode_t mode = FMODE_READ;
 868         char *subvol_name = NULL;
 869         u64 subvol_objectid = 0;
 870         u64 subvol_rootid = 0;
 871         int error = 0;
 872
 873         if (!(flags & MS_RDONLY))
 874                 mode |= FMODE_WRITE;
 875
 876         error = btrfs_parse_early_options(data, mode, fs_type,
 877                                           &subvol_name, &subvol_objectid,
 878                                           &subvol_rootid, &fs_devices);
 879         if (error) {
 880                 kfree(subvol_name);
 881                 return ERR_PTR(error);
 882         }
 883
 884         if (subvol_name) {
 885                 root = mount_subvol(subvol_name, flags, device_name, data);
 886                 kfree(subvol_name);
 887                 return root;
 888         }
 889
 890         error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
 891         if (error)
 892                 return ERR_PTR(error);
 893
 894         /*
 895          * Setup a dummy root and fs_info for test/set super.  This is because
 896          * we don't actually fill this stuff out until open_ctree, but we need
 897          * it for searching for existing supers, so this lets us do that and
 898          * then open_ctree will properly initialize everything later.
 899          */
 900         fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
 901         if (!fs_info)
 902                 return ERR_PTR(-ENOMEM);
 903
 904         fs_info->tree_root = btrfs_alloc_root(fs_info);
 905         if (!fs_info->tree_root) {
 906                 error = -ENOMEM;
 907                 goto error_fs_info;
 908         }
 909         fs_info->fs_devices = fs_devices;
 910
 911         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
 912         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
 913         if (!fs_info->super_copy || !fs_info->super_for_commit) {
 914                 error = -ENOMEM;
 915                 goto error_fs_info;
 916         }
 917
 918         error = btrfs_open_devices(fs_devices, mode, fs_type);
 919         if (error)
 920                 goto error_fs_info;
 921
 922         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
 923                 error = -EACCES;
 924                 goto error_close_devices;
 925         }
 926
 927         bdev = fs_devices->latest_bdev;
 928         s = sget(fs_type, btrfs_test_super, btrfs_set_super,
 929                  fs_info->tree_root);
 930         if (IS_ERR(s)) {
 931                 error = PTR_ERR(s);
 932                 goto error_close_devices;
 933         }
 934
 935         if (s->s_root) {
 936                 if ((flags ^ s->s_flags) & MS_RDONLY) {
 937                         deactivate_locked_super(s);
 938                         error = -EBUSY;
 939                         goto error_close_devices;
 940                 }
 941
 942                 btrfs_close_devices(fs_devices);
 943                 free_fs_info(fs_info);
 944         } else {
 945                 char b[BDEVNAME_SIZE];
 946
 947                 s->s_flags = flags | MS_NOSEC;
 948                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 949                 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
 950                 error = btrfs_fill_super(s, fs_devices, data,
 951                                          flags & MS_SILENT ? 1 : 0);
 952                 if (error) {
 953                         deactivate_locked_super(s);
 954                         return ERR_PTR(error);
 955                 }
 956
 957                 s->s_flags |= MS_ACTIVE;
 958         }
 959
 960         root = get_default_root(s, subvol_objectid);
 961         if (IS_ERR(root)) {
 962                 deactivate_locked_super(s);
 963                 return root;
 964         }
 965
 966         return root;
 967
 968 error_close_devices:
 969         btrfs_close_devices(fs_devices);
 970 error_fs_info:
 971         free_fs_info(fs_info);
 972         return ERR_PTR(error);
 973 }
 974
 975 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
 976 {
 977         struct btrfs_root *root = btrfs_sb(sb);
 978         int ret;
 979
 980         ret = btrfs_parse_options(root, data);
 981         if (ret)
 982                 return -EINVAL;
 983
 984         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
 985                 return 0;
 986
 987         if (*flags & MS_RDONLY) {
 988                 sb->s_flags |= MS_RDONLY;
 989
 990                 ret =  btrfs_commit_super(root);
 991                 WARN_ON(ret);
 992         } else {
 993                 if (root->fs_info->fs_devices->rw_devices == 0)
 994                         return -EACCES;
 995
 996                 if (btrfs_super_log_root(root->fs_info->super_copy) != 0)
 997                         return -EINVAL;
 998
 999                 ret = btrfs_cleanup_fs_roots(root->fs_info);
1000                 WARN_ON(ret);
1001
1002                 /* recover relocation */
1003                 ret = btrfs_recover_relocation(root);
1004                 WARN_ON(ret);
1005
1006                 sb->s_flags &= ~MS_RDONLY;
1007         }
1008
1009         return 0;
1010 }
1011
1012 /* Used to sort the devices by max_avail(descending sort) */
1013 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1014                                        const void *dev_info2)
1015 {
1016         if (((struct btrfs_device_info *)dev_info1)->max_avail >
1017             ((struct btrfs_device_info *)dev_info2)->max_avail)
1018                 return -1;
1019         else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1020                  ((struct btrfs_device_info *)dev_info2)->max_avail)
1021                 return 1;
1022         else
1023         return 0;
1024 }
1025
1026 /*
1027  * sort the devices by max_avail, in which max free extent size of each device
1028  * is stored.(Descending Sort)
1029  */
1030 static inline void btrfs_descending_sort_devices(
1031                                         struct btrfs_device_info *devices,
1032                                         size_t nr_devices)
1033 {
1034         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1035              btrfs_cmp_device_free_bytes, NULL);
1036 }
1037
1038 /*
1039  * The helper to calc the free space on the devices that can be used to store
1040  * file data.
1041  */
1042 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1043 {
1044         struct btrfs_fs_info *fs_info = root->fs_info;
1045         struct btrfs_device_info *devices_info;
1046         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1047         struct btrfs_device *device;
1048         u64 skip_space;
1049         u64 type;
1050         u64 avail_space;
1051         u64 used_space;
1052         u64 min_stripe_size;
1053         int min_stripes = 1, num_stripes = 1;
1054         int i = 0, nr_devices;
1055         int ret;
1056
1057         nr_devices = fs_info->fs_devices->open_devices;
1058         BUG_ON(!nr_devices);
1059
1060         devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1061                                GFP_NOFS);
1062         if (!devices_info)
1063                 return -ENOMEM;
1064
1065         /* calc min stripe number for data space alloction */
1066         type = btrfs_get_alloc_profile(root, 1);
1067         if (type & BTRFS_BLOCK_GROUP_RAID0) {
1068                 min_stripes = 2;
1069                 num_stripes = nr_devices;
1070         } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1071                 min_stripes = 2;
1072                 num_stripes = 2;
1073         } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1074                 min_stripes = 4;
1075                 num_stripes = 4;
1076         }
1077
1078         if (type & BTRFS_BLOCK_GROUP_DUP)
1079                 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1080         else
1081                 min_stripe_size = BTRFS_STRIPE_LEN;
1082
1083         list_for_each_entry(device, &fs_devices->devices, dev_list) {
1084                 if (!device->in_fs_metadata || !device->bdev)
1085                         continue;
1086
1087                 avail_space = device->total_bytes - device->bytes_used;
1088
1089                 /* align with stripe_len */
1090                 do_div(avail_space, BTRFS_STRIPE_LEN);
1091                 avail_space *= BTRFS_STRIPE_LEN;
1092
1093                 /*
1094                  * In order to avoid overwritting the superblock on the drive,
1095                  * btrfs starts at an offset of at least 1MB when doing chunk
1096                  * allocation.
1097                  */
1098                 skip_space = 1024 * 1024;
1099
1100                 /* user can set the offset in fs_info->alloc_start. */
1101                 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1102                     device->total_bytes)
1103                         skip_space = max(fs_info->alloc_start, skip_space);
1104
1105                 /*
1106                  * btrfs can not use the free space in [0, skip_space - 1],
1107                  * we must subtract it from the total. In order to implement
1108                  * it, we account the used space in this range first.
1109                  */
1110                 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1111                                                      &used_space);
1112                 if (ret) {
1113                         kfree(devices_info);
1114                         return ret;
1115                 }
1116
1117                 /* calc the free space in [0, skip_space - 1] */
1118                 skip_space -= used_space;
1119
1120                 /*
1121                  * we can use the free space in [0, skip_space - 1], subtract
1122                  * it from the total.
1123                  */
1124                 if (avail_space && avail_space >= skip_space)
1125                         avail_space -= skip_space;
1126                 else
1127                         avail_space = 0;
1128
1129                 if (avail_space < min_stripe_size)
1130                         continue;
1131
1132                 devices_info[i].dev = device;
1133                 devices_info[i].max_avail = avail_space;
1134
1135                 i++;
1136         }
1137
1138         nr_devices = i;
1139
1140         btrfs_descending_sort_devices(devices_info, nr_devices);
1141
1142         i = nr_devices - 1;
1143         avail_space = 0;
1144         while (nr_devices >= min_stripes) {
1145                 if (num_stripes > nr_devices)
1146                         num_stripes = nr_devices;
1147
1148                 if (devices_info[i].max_avail >= min_stripe_size) {
1149                         int j;
1150                         u64 alloc_size;
1151
1152                         avail_space += devices_info[i].max_avail * num_stripes;
1153                         alloc_size = devices_info[i].max_avail;
1154                         for (j = i + 1 - num_stripes; j <= i; j++)
1155                                 devices_info[j].max_avail -= alloc_size;
1156                 }
1157                 i--;
1158                 nr_devices--;
1159         }
1160
1161         kfree(devices_info);
1162         *free_bytes = avail_space;
1163         return 0;
1164 }
1165
1166 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1167 {
1168         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
1169         struct btrfs_super_block *disk_super = root->fs_info->super_copy;
1170         struct list_head *head = &root->fs_info->space_info;
1171         struct btrfs_space_info *found;
1172         u64 total_used = 0;
1173         u64 total_free_data = 0;
1174         int bits = dentry->d_sb->s_blocksize_bits;
1175         __be32 *fsid = (__be32 *)root->fs_info->fsid;
1176         int ret;
1177
1178         /* holding chunk_muext to avoid allocating new chunks */
1179         mutex_lock(&root->fs_info->chunk_mutex);
1180         rcu_read_lock();
1181         list_for_each_entry_rcu(found, head, list) {
1182                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1183                         total_free_data += found->disk_total - found->disk_used;
1184                         total_free_data -=
1185                                 btrfs_account_ro_block_groups_free_space(found);
1186                 }
1187
1188                 total_used += found->disk_used;
1189         }
1190         rcu_read_unlock();
1191
1192         buf->f_namelen = BTRFS_NAME_LEN;
1193         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1194         buf->f_bfree = buf->f_blocks - (total_used >> bits);
1195         buf->f_bsize = dentry->d_sb->s_blocksize;
1196         buf->f_type = BTRFS_SUPER_MAGIC;
1197         buf->f_bavail = total_free_data;
1198         ret = btrfs_calc_avail_data_space(root, &total_free_data);
1199         if (ret) {
1200                 mutex_unlock(&root->fs_info->chunk_mutex);
1201                 return ret;
1202         }
1203         buf->f_bavail += total_free_data;
1204         buf->f_bavail = buf->f_bavail >> bits;
1205         mutex_unlock(&root->fs_info->chunk_mutex);
1206
1207         /* We treat it as constant endianness (it doesn't matter _which_)
1208            because we want the fsid to come out the same whether mounted
1209            on a big-endian or little-endian host */
1210         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1211         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1212         /* Mask in the root object ID too, to disambiguate subvols */
1213         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1214         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1215
1216         return 0;
1217 }
1218
1219 static void btrfs_kill_super(struct super_block *sb)
1220 {
1221         struct btrfs_fs_info *fs_info = NULL;
1222         if (sb->s_root)
1223                 fs_info = btrfs_sb(sb)->fs_info;
1224         kill_anon_super(sb);
1225         if (fs_info)
1226                 free_fs_info(fs_info);
1227 }
1228
1229 static struct file_system_type btrfs_fs_type = {
1230         .owner          = THIS_MODULE,
1231         .name           = "btrfs",
1232         .mount          = btrfs_mount,
1233         .kill_sb        = btrfs_kill_super,
1234         .fs_flags       = FS_REQUIRES_DEV,
1235 };
1236
1237 /*
1238  * used by btrfsctl to scan devices when no FS is mounted
1239  */
1240 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1241                                 unsigned long arg)
1242 {
1243         struct btrfs_ioctl_vol_args *vol;
1244         struct btrfs_fs_devices *fs_devices;
1245         int ret = -ENOTTY;
1246
1247         if (!capable(CAP_SYS_ADMIN))
1248                 return -EPERM;
1249
1250         vol = memdup_user((void __user *)arg, sizeof(*vol));
1251         if (IS_ERR(vol))
1252                 return PTR_ERR(vol);
1253
1254         switch (cmd) {
1255         case BTRFS_IOC_SCAN_DEV:
1256                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1257                                             &btrfs_fs_type, &fs_devices);
1258                 break;
1259         }
1260
1261         kfree(vol);
1262         return ret;
1263 }
1264
1265 static int btrfs_freeze(struct super_block *sb)
1266 {
1267         struct btrfs_root *root = btrfs_sb(sb);
1268         mutex_lock(&root->fs_info->transaction_kthread_mutex);
1269         mutex_lock(&root->fs_info->cleaner_mutex);
1270         return 0;
1271 }
1272
1273 static int btrfs_unfreeze(struct super_block *sb)
1274 {
1275         struct btrfs_root *root = btrfs_sb(sb);
1276         mutex_unlock(&root->fs_info->cleaner_mutex);
1277         mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1278         return 0;
1279 }
1280
1281 static void btrfs_fs_dirty_inode(struct inode *inode, int flags)
1282 {
1283         int ret;
1284
1285         ret = btrfs_dirty_inode(inode);
1286         if (ret)
1287                 printk_ratelimited(KERN_ERR "btrfs: fail to dirty inode %Lu "
1288                                    "error %d\n", btrfs_ino(inode), ret);
1289 }
1290
1291 static const struct super_operations btrfs_super_ops = {
1292         .drop_inode     = btrfs_drop_inode,
1293         .evict_inode    = btrfs_evict_inode,
1294         .put_super      = btrfs_put_super,
1295         .sync_fs        = btrfs_sync_fs,
1296         .show_options   = btrfs_show_options,
1297         .write_inode    = btrfs_write_inode,
1298         .dirty_inode    = btrfs_fs_dirty_inode,
1299         .alloc_inode    = btrfs_alloc_inode,
1300         .destroy_inode  = btrfs_destroy_inode,
1301         .statfs         = btrfs_statfs,
1302         .remount_fs     = btrfs_remount,
1303         .freeze_fs      = btrfs_freeze,
1304         .unfreeze_fs    = btrfs_unfreeze,
1305 };
1306
1307 static const struct file_operations btrfs_ctl_fops = {
1308         .unlocked_ioctl  = btrfs_control_ioctl,
1309         .compat_ioctl = btrfs_control_ioctl,
1310         .owner   = THIS_MODULE,
1311         .llseek = noop_llseek,
1312 };
1313
1314 static struct miscdevice btrfs_misc = {
1315         .minor          = BTRFS_MINOR,
1316         .name           = "btrfs-control",
1317         .fops           = &btrfs_ctl_fops
1318 };
1319
1320 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1321 MODULE_ALIAS("devname:btrfs-control");
1322
1323 static int btrfs_interface_init(void)
1324 {
1325         return misc_register(&btrfs_misc);
1326 }
1327
1328 static void btrfs_interface_exit(void)
1329 {
1330         if (misc_deregister(&btrfs_misc) < 0)
1331                 printk(KERN_INFO "misc_deregister failed for control device");
1332 }
1333
1334 static int __init init_btrfs_fs(void)
1335 {
1336         int err;
1337
1338         err = btrfs_init_sysfs();
1339         if (err)
1340                 return err;
1341
1342         err = btrfs_init_compress();
1343         if (err)
1344                 goto free_sysfs;
1345
1346         err = btrfs_init_cachep();
1347         if (err)
1348                 goto free_compress;
1349
1350         err = extent_io_init();
1351         if (err)
1352                 goto free_cachep;
1353
1354         err = extent_map_init();
1355         if (err)
1356                 goto free_extent_io;
1357
1358         err = btrfs_delayed_inode_init();
1359         if (err)
1360                 goto free_extent_map;
1361
1362         err = btrfs_interface_init();
1363         if (err)
1364                 goto free_delayed_inode;
1365
1366         err = register_filesystem(&btrfs_fs_type);
1367         if (err)
1368                 goto unregister_ioctl;
1369
1370         printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1371         return 0;
1372
1373 unregister_ioctl:
1374         btrfs_interface_exit();
1375 free_delayed_inode:
1376         btrfs_delayed_inode_exit();
1377 free_extent_map:
1378         extent_map_exit();
1379 free_extent_io:
1380         extent_io_exit();
1381 free_cachep:
1382         btrfs_destroy_cachep();
1383 free_compress:
1384         btrfs_exit_compress();
1385 free_sysfs:
1386         btrfs_exit_sysfs();
1387         return err;
1388 }
1389
1390 static void __exit exit_btrfs_fs(void)
1391 {
1392         btrfs_destroy_cachep();
1393         btrfs_delayed_inode_exit();
1394         extent_map_exit();
1395         extent_io_exit();
1396         btrfs_interface_exit();
1397         unregister_filesystem(&btrfs_fs_type);
1398         btrfs_exit_sysfs();
1399         btrfs_cleanup_fs_uuids();
1400         btrfs_exit_compress();
1401 }
1402
1403 module_init(init_btrfs_fs)
1404 module_exit(exit_btrfs_fs)
1405
1406 MODULE_LICENSE("GPL");