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

   1 /*
   2  * Copyright (C) 2007 Oracle.  All rights reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public
   6  * License v2 as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public
  14  * License along with this program; if not, write to the
  15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  16  * Boston, MA 021110-1307, USA.
  17  */
  18
  19 #include <linux/blkdev.h>
  20 #include <linux/module.h>
  21 #include <linux/buffer_head.h>
  22 #include <linux/fs.h>
  23 #include <linux/pagemap.h>
  24 #include <linux/highmem.h>
  25 #include <linux/time.h>
  26 #include <linux/init.h>
  27 #include <linux/seq_file.h>
  28 #include <linux/string.h>
  29 #include <linux/backing-dev.h>
  30 #include <linux/mount.h>
  31 #include <linux/mpage.h>
  32 #include <linux/swap.h>
  33 #include <linux/writeback.h>
  34 #include <linux/statfs.h>
  35 #include <linux/compat.h>
  36 #include <linux/parser.h>
  37 #include <linux/ctype.h>
  38 #include <linux/namei.h>
  39 #include <linux/miscdevice.h>
  40 #include <linux/magic.h>
  41 #include <linux/slab.h>
  42 #include <linux/cleancache.h>
  43 #include <linux/mnt_namespace.h>
  44 #include <linux/ratelimit.h>
  45 #include "compat.h"
  46 #include "delayed-inode.h"
  47 #include "ctree.h"
  48 #include "disk-io.h"
  49 #include "transaction.h"
  50 #include "btrfs_inode.h"
  51 #include "ioctl.h"
  52 #include "print-tree.h"
  53 #include "xattr.h"
  54 #include "volumes.h"
  55 #include "version.h"
  56 #include "export.h"
  57 #include "compression.h"
  58
  59 #define CREATE_TRACE_POINTS
  60 #include <trace/events/btrfs.h>
  61
  62 static const struct super_operations btrfs_super_ops;
  63 static struct file_system_type btrfs_fs_type;
  64
  65 static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
  66                                       char nbuf[16])
  67 {
  68         char *errstr = NULL;
  69
  70         switch (errno) {
  71         case -EIO:
  72                 errstr = "IO failure";
  73                 break;
  74         case -ENOMEM:
  75                 errstr = "Out of memory";
  76                 break;
  77         case -EROFS:
  78                 errstr = "Readonly filesystem";
  79                 break;
  80         default:
  81                 if (nbuf) {
  82                         if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
  83                                 errstr = nbuf;
  84                 }
  85                 break;
  86         }
  87
  88         return errstr;
  89 }
  90
  91 static void __save_error_info(struct btrfs_fs_info *fs_info)
  92 {
  93         /*
  94          * today we only save the error info into ram.  Long term we'll
  95          * also send it down to the disk
  96          */
  97         fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
  98 }
  99
 100 /* NOTE:
 101  *      We move write_super stuff at umount in order to avoid deadlock
 102  *      for umount hold all lock.
 103  */
 104 static void save_error_info(struct btrfs_fs_info *fs_info)
 105 {
 106         __save_error_info(fs_info);
 107 }
 108
 109 /* btrfs handle error by forcing the filesystem readonly */
 110 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
 111 {
 112         struct super_block *sb = fs_info->sb;
 113
 114         if (sb->s_flags & MS_RDONLY)
 115                 return;
 116
 117         if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
 118                 sb->s_flags |= MS_RDONLY;
 119                 printk(KERN_INFO "btrfs is forced readonly\n");
 120         }
 121 }
 122
 123 /*
 124  * __btrfs_std_error decodes expected errors from the caller and
 125  * invokes the approciate error response.
 126  */
 127 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 128                      unsigned int line, int errno)
 129 {
 130         struct super_block *sb = fs_info->sb;
 131         char nbuf[16];
 132         const char *errstr;
 133
 134         /*
 135          * Special case: if the error is EROFS, and we're already
 136          * under MS_RDONLY, then it is safe here.
 137          */
 138         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 139                 return;
 140
 141         errstr = btrfs_decode_error(fs_info, errno, nbuf);
 142         printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
 143                 sb->s_id, function, line, errstr);
 144         save_error_info(fs_info);
 145
 146         btrfs_handle_error(fs_info);
 147 }
 148
 149 static void btrfs_put_super(struct super_block *sb)
 150 {
 151         struct btrfs_root *root = btrfs_sb(sb);
 152         int ret;
 153
 154         ret = close_ctree(root);
 155         sb->s_fs_info = NULL;
 156
 157         (void)ret; /* FIXME: need to fix VFS to return error? */
 158 }
 159
 160 enum {
 161         Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
 162         Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
 163         Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
 164         Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
 165         Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
 166         Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
 167         Opt_enospc_debug, Opt_subvolrootid, Opt_defrag,
 168         Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
 169         Opt_check_integrity, Opt_check_integrity_including_extent_data,
 170         Opt_check_integrity_print_mask,
 171         Opt_err,
 172 };
 173
 174 static match_table_t tokens = {
 175         {Opt_degraded, "degraded"},
 176         {Opt_subvol, "subvol=%s"},
 177         {Opt_subvolid, "subvolid=%d"},
 178         {Opt_device, "device=%s"},
 179         {Opt_nodatasum, "nodatasum"},
 180         {Opt_nodatacow, "nodatacow"},
 181         {Opt_nobarrier, "nobarrier"},
 182         {Opt_max_inline, "max_inline=%s"},
 183         {Opt_alloc_start, "alloc_start=%s"},
 184         {Opt_thread_pool, "thread_pool=%d"},
 185         {Opt_compress, "compress"},
 186         {Opt_compress_type, "compress=%s"},
 187         {Opt_compress_force, "compress-force"},
 188         {Opt_compress_force_type, "compress-force=%s"},
 189         {Opt_ssd, "ssd"},
 190         {Opt_ssd_spread, "ssd_spread"},
 191         {Opt_nossd, "nossd"},
 192         {Opt_noacl, "noacl"},
 193         {Opt_notreelog, "notreelog"},
 194         {Opt_flushoncommit, "flushoncommit"},
 195         {Opt_ratio, "metadata_ratio=%d"},
 196         {Opt_discard, "discard"},
 197         {Opt_space_cache, "space_cache"},
 198         {Opt_clear_cache, "clear_cache"},
 199         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 200         {Opt_enospc_debug, "enospc_debug"},
 201         {Opt_subvolrootid, "subvolrootid=%d"},
 202         {Opt_defrag, "autodefrag"},
 203         {Opt_inode_cache, "inode_cache"},
 204         {Opt_no_space_cache, "nospace_cache"},
 205         {Opt_recovery, "recovery"},
 206         {Opt_check_integrity, "check_int"},
 207         {Opt_check_integrity_including_extent_data, "check_int_data"},
 208         {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
 209         {Opt_err, NULL},
 210 };
 211
 212 /*
 213  * Regular mount options parser.  Everything that is needed only when
 214  * reading in a new superblock is parsed here.
 215  */
 216 int btrfs_parse_options(struct btrfs_root *root, char *options)
 217 {
 218         struct btrfs_fs_info *info = root->fs_info;
 219         substring_t args[MAX_OPT_ARGS];
 220         char *p, *num, *orig = NULL;
 221         u64 cache_gen;
 222         int intarg;
 223         int ret = 0;
 224         char *compress_type;
 225         bool compress_force = false;
 226
 227         cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
 228         if (cache_gen)
 229                 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 230
 231         if (!options)
 232                 goto out;
 233
 234         /*
 235          * strsep changes the string, duplicate it because parse_options
 236          * gets called twice
 237          */
 238         options = kstrdup(options, GFP_NOFS);
 239         if (!options)
 240                 return -ENOMEM;
 241
 242         orig = options;
 243
 244         while ((p = strsep(&options, ",")) != NULL) {
 245                 int token;
 246                 if (!*p)
 247                         continue;
 248
 249                 token = match_token(p, tokens, args);
 250                 switch (token) {
 251                 case Opt_degraded:
 252                         printk(KERN_INFO "btrfs: allowing degraded mounts\n");
 253                         btrfs_set_opt(info->mount_opt, DEGRADED);
 254                         break;
 255                 case Opt_subvol:
 256                 case Opt_subvolid:
 257                 case Opt_subvolrootid:
 258                 case Opt_device:
 259                         /*
 260                          * These are parsed by btrfs_parse_early_options
 261                          * and can be happily ignored here.
 262                          */
 263                         break;
 264                 case Opt_nodatasum:
 265                         printk(KERN_INFO "btrfs: setting nodatasum\n");
 266                         btrfs_set_opt(info->mount_opt, NODATASUM);
 267                         break;
 268                 case Opt_nodatacow:
 269                         printk(KERN_INFO "btrfs: setting nodatacow\n");
 270                         btrfs_set_opt(info->mount_opt, NODATACOW);
 271                         btrfs_set_opt(info->mount_opt, NODATASUM);
 272                         break;
 273                 case Opt_compress_force:
 274                 case Opt_compress_force_type:
 275                         compress_force = true;
 276                 case Opt_compress:
 277                 case Opt_compress_type:
 278                         if (token == Opt_compress ||
 279                             token == Opt_compress_force ||
 280                             strcmp(args[0].from, "zlib") == 0) {
 281                                 compress_type = "zlib";
 282                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 283                         } else if (strcmp(args[0].from, "lzo") == 0) {
 284                                 compress_type = "lzo";
 285                                 info->compress_type = BTRFS_COMPRESS_LZO;
 286                         } else {
 287                                 ret = -EINVAL;
 288                                 goto out;
 289                         }
 290
 291                         btrfs_set_opt(info->mount_opt, COMPRESS);
 292                         if (compress_force) {
 293                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 294                                 pr_info("btrfs: force %s compression\n",
 295                                         compress_type);
 296                         } else
 297                                 pr_info("btrfs: use %s compression\n",
 298                                         compress_type);
 299                         break;
 300                 case Opt_ssd:
 301                         printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 302                         btrfs_set_opt(info->mount_opt, SSD);
 303                         break;
 304                 case Opt_ssd_spread:
 305                         printk(KERN_INFO "btrfs: use spread ssd "
 306                                "allocation scheme\n");
 307                         btrfs_set_opt(info->mount_opt, SSD);
 308                         btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 309                         break;
 310                 case Opt_nossd:
 311                         printk(KERN_INFO "btrfs: not using ssd allocation "
 312                                "scheme\n");
 313                         btrfs_set_opt(info->mount_opt, NOSSD);
 314                         btrfs_clear_opt(info->mount_opt, SSD);
 315                         btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 316                         break;
 317                 case Opt_nobarrier:
 318                         printk(KERN_INFO "btrfs: turning off barriers\n");
 319                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 320                         break;
 321                 case Opt_thread_pool:
 322                         intarg = 0;
 323                         match_int(&args[0], &intarg);
 324                         if (intarg) {
 325                                 info->thread_pool_size = intarg;
 326                                 printk(KERN_INFO "btrfs: thread pool %d\n",
 327                                        info->thread_pool_size);
 328                         }
 329                         break;
 330                 case Opt_max_inline:
 331                         num = match_strdup(&args[0]);
 332                         if (num) {
 333                                 info->max_inline = memparse(num, NULL);
 334                                 kfree(num);
 335
 336                                 if (info->max_inline) {
 337                                         info->max_inline = max_t(u64,
 338                                                 info->max_inline,
 339                                                 root->sectorsize);
 340                                 }
 341                                 printk(KERN_INFO "btrfs: max_inline at %llu\n",
 342                                         (unsigned long long)info->max_inline);
 343                         }
 344                         break;
 345                 case Opt_alloc_start:
 346                         num = match_strdup(&args[0]);
 347                         if (num) {
 348                                 info->alloc_start = memparse(num, NULL);
 349                                 kfree(num);
 350                                 printk(KERN_INFO
 351                                         "btrfs: allocations start at %llu\n",
 352                                         (unsigned long long)info->alloc_start);
 353                         }
 354                         break;
 355                 case Opt_noacl:
 356                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 357                         break;
 358                 case Opt_notreelog:
 359                         printk(KERN_INFO "btrfs: disabling tree log\n");
 360                         btrfs_set_opt(info->mount_opt, NOTREELOG);
 361                         break;
 362                 case Opt_flushoncommit:
 363                         printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
 364                         btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
 365                         break;
 366                 case Opt_ratio:
 367                         intarg = 0;
 368                         match_int(&args[0], &intarg);
 369                         if (intarg) {
 370                                 info->metadata_ratio = intarg;
 371                                 printk(KERN_INFO "btrfs: metadata ratio %d\n",
 372                                        info->metadata_ratio);
 373                         }
 374                         break;
 375                 case Opt_discard:
 376                         btrfs_set_opt(info->mount_opt, DISCARD);
 377                         break;
 378                 case Opt_space_cache:
 379                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 380                         break;
 381                 case Opt_no_space_cache:
 382                         printk(KERN_INFO "btrfs: disabling disk space caching\n");
 383                         btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
 384                         break;
 385                 case Opt_inode_cache:
 386                         printk(KERN_INFO "btrfs: enabling inode map caching\n");
 387                         btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
 388                         break;
 389                 case Opt_clear_cache:
 390                         printk(KERN_INFO "btrfs: force clearing of disk cache\n");
 391                         btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
 392                         break;
 393                 case Opt_user_subvol_rm_allowed:
 394                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 395                         break;
 396                 case Opt_enospc_debug:
 397                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 398                         break;
 399                 case Opt_defrag:
 400                         printk(KERN_INFO "btrfs: enabling auto defrag");
 401                         btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
 402                         break;
 403                 case Opt_recovery:
 404                         printk(KERN_INFO "btrfs: enabling auto recovery");
 405                         btrfs_set_opt(info->mount_opt, RECOVERY);
 406                         break;
 407 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 408                 case Opt_check_integrity_including_extent_data:
 409                         printk(KERN_INFO "btrfs: enabling check integrity"
 410                                " including extent data\n");
 411                         btrfs_set_opt(info->mount_opt,
 412                                       CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
 413                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 414                         break;
 415                 case Opt_check_integrity:
 416                         printk(KERN_INFO "btrfs: enabling check integrity\n");
 417                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 418                         break;
 419                 case Opt_check_integrity_print_mask:
 420                         intarg = 0;
 421                         match_int(&args[0], &intarg);
 422                         if (intarg) {
 423                                 info->check_integrity_print_mask = intarg;
 424                                 printk(KERN_INFO "btrfs:"
 425                                        " check_integrity_print_mask 0x%x\n",
 426                                        info->check_integrity_print_mask);
 427                         }
 428                         break;
 429 #else
 430                 case Opt_check_integrity_including_extent_data:
 431                 case Opt_check_integrity:
 432                 case Opt_check_integrity_print_mask:
 433                         printk(KERN_ERR "btrfs: support for check_integrity*"
 434                                " not compiled in!\n");
 435                         ret = -EINVAL;
 436                         goto out;
 437 #endif
 438                 case Opt_err:
 439                         printk(KERN_INFO "btrfs: unrecognized mount option "
 440                                "'%s'\n", p);
 441                         ret = -EINVAL;
 442                         goto out;
 443                 default:
 444                         break;
 445                 }
 446         }
 447 out:
 448         if (!ret && btrfs_test_opt(root, SPACE_CACHE))
 449                 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
 450         kfree(orig);
 451         return ret;
 452 }
 453
 454 /*
 455  * Parse mount options that are required early in the mount process.
 456  *
 457  * All other options will be parsed on much later in the mount process and
 458  * only when we need to allocate a new super block.
 459  */
 460 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 461                 void *holder, char **subvol_name, u64 *subvol_objectid,
 462                 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
 463 {
 464         substring_t args[MAX_OPT_ARGS];
 465         char *device_name, *opts, *orig, *p;
 466         int error = 0;
 467         int intarg;
 468
 469         if (!options)
 470                 return 0;
 471
 472         /*
 473          * strsep changes the string, duplicate it because parse_options
 474          * gets called twice
 475          */
 476         opts = kstrdup(options, GFP_KERNEL);
 477         if (!opts)
 478                 return -ENOMEM;
 479         orig = opts;
 480
 481         while ((p = strsep(&opts, ",")) != NULL) {
 482                 int token;
 483                 if (!*p)
 484                         continue;
 485
 486                 token = match_token(p, tokens, args);
 487                 switch (token) {
 488                 case Opt_subvol:
 489                         kfree(*subvol_name);
 490                         *subvol_name = match_strdup(&args[0]);
 491                         break;
 492                 case Opt_subvolid:
 493                         intarg = 0;
 494                         error = match_int(&args[0], &intarg);
 495                         if (!error) {
 496                                 /* we want the original fs_tree */
 497                                 if (!intarg)
 498                                         *subvol_objectid =
 499                                                 BTRFS_FS_TREE_OBJECTID;
 500                                 else
 501                                         *subvol_objectid = intarg;
 502                         }
 503                         break;
 504                 case Opt_subvolrootid:
 505                         intarg = 0;
 506                         error = match_int(&args[0], &intarg);
 507                         if (!error) {
 508                                 /* we want the original fs_tree */
 509                                 if (!intarg)
 510                                         *subvol_rootid =
 511                                                 BTRFS_FS_TREE_OBJECTID;
 512                                 else
 513                                         *subvol_rootid = intarg;
 514                         }
 515                         break;
 516                 case Opt_device:
 517                         device_name = match_strdup(&args[0]);
 518                         if (!device_name) {
 519                                 error = -ENOMEM;
 520                                 goto out;
 521                         }
 522                         error = btrfs_scan_one_device(device_name,
 523                                         flags, holder, fs_devices);
 524                         kfree(device_name);
 525                         if (error)
 526                                 goto out;
 527                         break;
 528                 default:
 529                         break;
 530                 }
 531         }
 532
 533 out:
 534         kfree(orig);
 535         return error;
 536 }
 537
 538 static struct dentry *get_default_root(struct super_block *sb,
 539                                        u64 subvol_objectid)
 540 {
 541         struct btrfs_root *root = sb->s_fs_info;
 542         struct btrfs_root *new_root;
 543         struct btrfs_dir_item *di;
 544         struct btrfs_path *path;
 545         struct btrfs_key location;
 546         struct inode *inode;
 547         u64 dir_id;
 548         int new = 0;
 549
 550         /*
 551          * We have a specific subvol we want to mount, just setup location and
 552          * go look up the root.
 553          */
 554         if (subvol_objectid) {
 555                 location.objectid = subvol_objectid;
 556                 location.type = BTRFS_ROOT_ITEM_KEY;
 557                 location.offset = (u64)-1;
 558                 goto find_root;
 559         }
 560
 561         path = btrfs_alloc_path();
 562         if (!path)
 563                 return ERR_PTR(-ENOMEM);
 564         path->leave_spinning = 1;
 565
 566         /*
 567          * Find the "default" dir item which points to the root item that we
 568          * will mount by default if we haven't been given a specific subvolume
 569          * to mount.
 570          */
 571         dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
 572         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 573         if (IS_ERR(di)) {
 574                 btrfs_free_path(path);
 575                 return ERR_CAST(di);
 576         }
 577         if (!di) {
 578                 /*
 579                  * Ok the default dir item isn't there.  This is weird since
 580                  * it's always been there, but don't freak out, just try and
 581                  * mount to root most subvolume.
 582                  */
 583                 btrfs_free_path(path);
 584                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 585                 new_root = root->fs_info->fs_root;
 586                 goto setup_root;
 587         }
 588
 589         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 590         btrfs_free_path(path);
 591
 592 find_root:
 593         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
 594         if (IS_ERR(new_root))
 595                 return ERR_CAST(new_root);
 596
 597         if (btrfs_root_refs(&new_root->root_item) == 0)
 598                 return ERR_PTR(-ENOENT);
 599
 600         dir_id = btrfs_root_dirid(&new_root->root_item);
 601 setup_root:
 602         location.objectid = dir_id;
 603         location.type = BTRFS_INODE_ITEM_KEY;
 604         location.offset = 0;
 605
 606         inode = btrfs_iget(sb, &location, new_root, &new);
 607         if (IS_ERR(inode))
 608                 return ERR_CAST(inode);
 609
 610         /*
 611          * If we're just mounting the root most subvol put the inode and return
 612          * a reference to the dentry.  We will have already gotten a reference
 613          * to the inode in btrfs_fill_super so we're good to go.
 614          */
 615         if (!new && sb->s_root->d_inode == inode) {
 616                 iput(inode);
 617                 return dget(sb->s_root);
 618         }
 619
 620         return d_obtain_alias(inode);
 621 }
 622
 623 static int btrfs_fill_super(struct super_block *sb,
 624                             struct btrfs_fs_devices *fs_devices,
 625                             void *data, int silent)
 626 {
 627         struct inode *inode;
 628         struct dentry *root_dentry;
 629         struct btrfs_root *tree_root;
 630         struct btrfs_key key;
 631         int err;
 632
 633         sb->s_maxbytes = MAX_LFS_FILESIZE;
 634         sb->s_magic = BTRFS_SUPER_MAGIC;
 635         sb->s_op = &btrfs_super_ops;
 636         sb->s_d_op = &btrfs_dentry_operations;
 637         sb->s_export_op = &btrfs_export_ops;
 638         sb->s_xattr = btrfs_xattr_handlers;
 639         sb->s_time_gran = 1;
 640 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 641         sb->s_flags |= MS_POSIXACL;
 642 #endif
 643
 644         tree_root = open_ctree(sb, fs_devices, (char *)data);
 645
 646         if (IS_ERR(tree_root)) {
 647                 printk("btrfs: open_ctree failed\n");
 648                 return PTR_ERR(tree_root);
 649         }
 650         sb->s_fs_info = tree_root;
 651
 652         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 653         key.type = BTRFS_INODE_ITEM_KEY;
 654         key.offset = 0;
 655         inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
 656         if (IS_ERR(inode)) {
 657                 err = PTR_ERR(inode);
 658                 goto fail_close;
 659         }
 660
 661         root_dentry = d_alloc_root(inode);
 662         if (!root_dentry) {
 663                 iput(inode);
 664                 err = -ENOMEM;
 665                 goto fail_close;
 666         }
 667
 668         sb->s_root = root_dentry;
 669
 670         save_mount_options(sb, data);
 671         cleancache_init_fs(sb);
 672         return 0;
 673
 674 fail_close:
 675         close_ctree(tree_root);
 676         return err;
 677 }
 678
 679 int btrfs_sync_fs(struct super_block *sb, int wait)
 680 {
 681         struct btrfs_trans_handle *trans;
 682         struct btrfs_root *root = btrfs_sb(sb);
 683         int ret;
 684
 685         trace_btrfs_sync_fs(wait);
 686
 687         if (!wait) {
 688                 filemap_flush(root->fs_info->btree_inode->i_mapping);
 689                 return 0;
 690         }
 691
 692         btrfs_start_delalloc_inodes(root, 0);
 693         btrfs_wait_ordered_extents(root, 0, 0);
 694
 695         trans = btrfs_start_transaction(root, 0);
 696         if (IS_ERR(trans))
 697                 return PTR_ERR(trans);
 698         ret = btrfs_commit_transaction(trans, root);
 699         return ret;
 700 }
 701
 702 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 703 {
 704         struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
 705         struct btrfs_fs_info *info = root->fs_info;
 706         char *compress_type;
 707
 708         if (btrfs_test_opt(root, DEGRADED))
 709                 seq_puts(seq, ",degraded");
 710         if (btrfs_test_opt(root, NODATASUM))
 711                 seq_puts(seq, ",nodatasum");
 712         if (btrfs_test_opt(root, NODATACOW))
 713                 seq_puts(seq, ",nodatacow");
 714         if (btrfs_test_opt(root, NOBARRIER))
 715                 seq_puts(seq, ",nobarrier");
 716         if (info->max_inline != 8192 * 1024)
 717                 seq_printf(seq, ",max_inline=%llu",
 718                            (unsigned long long)info->max_inline);
 719         if (info->alloc_start != 0)
 720                 seq_printf(seq, ",alloc_start=%llu",
 721                            (unsigned long long)info->alloc_start);
 722         if (info->thread_pool_size !=  min_t(unsigned long,
 723                                              num_online_cpus() + 2, 8))
 724                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
 725         if (btrfs_test_opt(root, COMPRESS)) {
 726                 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
 727                         compress_type = "zlib";
 728                 else
 729                         compress_type = "lzo";
 730                 if (btrfs_test_opt(root, FORCE_COMPRESS))
 731                         seq_printf(seq, ",compress-force=%s", compress_type);
 732                 else
 733                         seq_printf(seq, ",compress=%s", compress_type);
 734         }
 735         if (btrfs_test_opt(root, NOSSD))
 736                 seq_puts(seq, ",nossd");
 737         if (btrfs_test_opt(root, SSD_SPREAD))
 738                 seq_puts(seq, ",ssd_spread");
 739         else if (btrfs_test_opt(root, SSD))
 740                 seq_puts(seq, ",ssd");
 741         if (btrfs_test_opt(root, NOTREELOG))
 742                 seq_puts(seq, ",notreelog");
 743         if (btrfs_test_opt(root, FLUSHONCOMMIT))
 744                 seq_puts(seq, ",flushoncommit");
 745         if (btrfs_test_opt(root, DISCARD))
 746                 seq_puts(seq, ",discard");
 747         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
 748                 seq_puts(seq, ",noacl");
 749         if (btrfs_test_opt(root, SPACE_CACHE))
 750                 seq_puts(seq, ",space_cache");
 751         else
 752                 seq_puts(seq, ",nospace_cache");
 753         if (btrfs_test_opt(root, CLEAR_CACHE))
 754                 seq_puts(seq, ",clear_cache");
 755         if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
 756                 seq_puts(seq, ",user_subvol_rm_allowed");
 757         if (btrfs_test_opt(root, ENOSPC_DEBUG))
 758                 seq_puts(seq, ",enospc_debug");
 759         if (btrfs_test_opt(root, AUTO_DEFRAG))
 760                 seq_puts(seq, ",autodefrag");
 761         if (btrfs_test_opt(root, INODE_MAP_CACHE))
 762                 seq_puts(seq, ",inode_cache");
 763         return 0;
 764 }
 765
 766 static int btrfs_test_super(struct super_block *s, void *data)
 767 {
 768         struct btrfs_root *test_root = data;
 769         struct btrfs_root *root = btrfs_sb(s);
 770
 771         /*
 772          * If this super block is going away, return false as it
 773          * can't match as an existing super block.
 774          */
 775         if (!atomic_read(&s->s_active))
 776                 return 0;
 777         return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
 778 }
 779
 780 static int btrfs_set_super(struct super_block *s, void *data)
 781 {
 782         s->s_fs_info = data;
 783
 784         return set_anon_super(s, data);
 785 }
 786
 787 /*
 788  * subvolumes are identified by ino 256
 789  */
 790 static inline int is_subvolume_inode(struct inode *inode)
 791 {
 792         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
 793                 return 1;
 794         return 0;
 795 }
 796
 797 /*
 798  * This will strip out the subvol=%s argument for an argument string and add
 799  * subvolid=0 to make sure we get the actual tree root for path walking to the
 800  * subvol we want.
 801  */
 802 static char *setup_root_args(char *args)
 803 {
 804         unsigned copied = 0;
 805         unsigned len = strlen(args) + 2;
 806         char *pos;
 807         char *ret;
 808
 809         /*
 810          * We need the same args as before, but minus
 811          *
 812          * subvol=a
 813          *
 814          * and add
 815          *
 816          * subvolid=0
 817          *
 818          * which is a difference of 2 characters, so we allocate strlen(args) +
 819          * 2 characters.
 820          */
 821         ret = kzalloc(len * sizeof(char), GFP_NOFS);
 822         if (!ret)
 823                 return NULL;
 824         pos = strstr(args, "subvol=");
 825
 826         /* This shouldn't happen, but just in case.. */
 827         if (!pos) {
 828                 kfree(ret);
 829                 return NULL;
 830         }
 831
 832         /*
 833          * The subvol=<> arg is not at the front of the string, copy everybody
 834          * up to that into ret.
 835          */
 836         if (pos != args) {
 837                 *pos = '\0';
 838                 strcpy(ret, args);
 839                 copied += strlen(args);
 840                 pos++;
 841         }
 842
 843         strncpy(ret + copied, "subvolid=0", len - copied);
 844
 845         /* Length of subvolid=0 */
 846         copied += 10;
 847
 848         /*
 849          * If there is no , after the subvol= option then we know there's no
 850          * other options and we can just return.
 851          */
 852         pos = strchr(pos, ',');
 853         if (!pos)
 854                 return ret;
 855
 856         /* Copy the rest of the arguments into our buffer */
 857         strncpy(ret + copied, pos, len - copied);
 858         copied += strlen(pos);
 859
 860         return ret;
 861 }
 862
 863 static struct dentry *mount_subvol(const char *subvol_name, int flags,
 864                                    const char *device_name, char *data)
 865 {
 866         struct super_block *s;
 867         struct dentry *root;
 868         struct vfsmount *mnt;
 869         struct mnt_namespace *ns_private;
 870         char *newargs;
 871         struct path path;
 872         int error;
 873
 874         newargs = setup_root_args(data);
 875         if (!newargs)
 876                 return ERR_PTR(-ENOMEM);
 877         mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
 878                              newargs);
 879         kfree(newargs);
 880         if (IS_ERR(mnt))
 881                 return ERR_CAST(mnt);
 882
 883         ns_private = create_mnt_ns(mnt);
 884         if (IS_ERR(ns_private)) {
 885                 mntput(mnt);
 886                 return ERR_CAST(ns_private);
 887         }
 888
 889         /*
 890          * This will trigger the automount of the subvol so we can just
 891          * drop the mnt we have here and return the dentry that we
 892          * found.
 893          */
 894         error = vfs_path_lookup(mnt->mnt_root, mnt, subvol_name,
 895                                 LOOKUP_FOLLOW, &path);
 896         put_mnt_ns(ns_private);
 897         if (error)
 898                 return ERR_PTR(error);
 899
 900         if (!is_subvolume_inode(path.dentry->d_inode)) {
 901                 path_put(&path);
 902                 mntput(mnt);
 903                 error = -EINVAL;
 904                 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
 905                                 subvol_name);
 906                 return ERR_PTR(-EINVAL);
 907         }
 908
 909         /* Get a ref to the sb and the dentry we found and return it */
 910         s = path.mnt->mnt_sb;
 911         atomic_inc(&s->s_active);
 912         root = dget(path.dentry);
 913         path_put(&path);
 914         down_write(&s->s_umount);
 915
 916         return root;
 917 }
 918
 919 /*
 920  * Find a superblock for the given device / mount point.
 921  *
 922  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
 923  *        for multiple device setup.  Make sure to keep it in sync.
 924  */
 925 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
 926                 const char *device_name, void *data)
 927 {
 928         struct block_device *bdev = NULL;
 929         struct super_block *s;
 930         struct dentry *root;
 931         struct btrfs_fs_devices *fs_devices = NULL;
 932         struct btrfs_fs_info *fs_info = NULL;
 933         fmode_t mode = FMODE_READ;
 934         char *subvol_name = NULL;
 935         u64 subvol_objectid = 0;
 936         u64 subvol_rootid = 0;
 937         int error = 0;
 938
 939         if (!(flags & MS_RDONLY))
 940                 mode |= FMODE_WRITE;
 941
 942         error = btrfs_parse_early_options(data, mode, fs_type,
 943                                           &subvol_name, &subvol_objectid,
 944                                           &subvol_rootid, &fs_devices);
 945         if (error) {
 946                 kfree(subvol_name);
 947                 return ERR_PTR(error);
 948         }
 949
 950         if (subvol_name) {
 951                 root = mount_subvol(subvol_name, flags, device_name, data);
 952                 kfree(subvol_name);
 953                 return root;
 954         }
 955
 956         error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
 957         if (error)
 958                 return ERR_PTR(error);
 959
 960         /*
 961          * Setup a dummy root and fs_info for test/set super.  This is because
 962          * we don't actually fill this stuff out until open_ctree, but we need
 963          * it for searching for existing supers, so this lets us do that and
 964          * then open_ctree will properly initialize everything later.
 965          */
 966         fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
 967         if (!fs_info)
 968                 return ERR_PTR(-ENOMEM);
 969
 970         fs_info->tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
 971         if (!fs_info->tree_root) {
 972                 error = -ENOMEM;
 973                 goto error_fs_info;
 974         }
 975         fs_info->tree_root->fs_info = fs_info;
 976         fs_info->fs_devices = fs_devices;
 977
 978         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
 979         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
 980         if (!fs_info->super_copy || !fs_info->super_for_commit) {
 981                 error = -ENOMEM;
 982                 goto error_fs_info;
 983         }
 984
 985         error = btrfs_open_devices(fs_devices, mode, fs_type);
 986         if (error)
 987                 goto error_fs_info;
 988
 989         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
 990                 error = -EACCES;
 991                 goto error_close_devices;
 992         }
 993
 994         bdev = fs_devices->latest_bdev;
 995         s = sget(fs_type, btrfs_test_super, btrfs_set_super,
 996                  fs_info->tree_root);
 997         if (IS_ERR(s)) {
 998                 error = PTR_ERR(s);
 999                 goto error_close_devices;
1000         }
1001
1002         if (s->s_root) {
1003                 if ((flags ^ s->s_flags) & MS_RDONLY) {
1004                         deactivate_locked_super(s);
1005                         error = -EBUSY;
1006                         goto error_close_devices;
1007                 }
1008
1009                 btrfs_close_devices(fs_devices);
1010                 free_fs_info(fs_info);
1011         } else {
1012                 char b[BDEVNAME_SIZE];
1013
1014                 s->s_flags = flags | MS_NOSEC;
1015                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1016                 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
1017                 error = btrfs_fill_super(s, fs_devices, data,
1018                                          flags & MS_SILENT ? 1 : 0);
1019                 if (error) {
1020                         deactivate_locked_super(s);
1021                         return ERR_PTR(error);
1022                 }
1023
1024                 s->s_flags |= MS_ACTIVE;
1025         }
1026
1027         root = get_default_root(s, subvol_objectid);
1028         if (IS_ERR(root)) {
1029                 deactivate_locked_super(s);
1030                 return root;
1031         }
1032
1033         return root;
1034
1035 error_close_devices:
1036         btrfs_close_devices(fs_devices);
1037 error_fs_info:
1038         free_fs_info(fs_info);
1039         return ERR_PTR(error);
1040 }
1041
1042 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1043 {
1044         struct btrfs_root *root = btrfs_sb(sb);
1045         int ret;
1046
1047         ret = btrfs_parse_options(root, data);
1048         if (ret)
1049                 return -EINVAL;
1050
1051         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1052                 return 0;
1053
1054         if (*flags & MS_RDONLY) {
1055                 sb->s_flags |= MS_RDONLY;
1056
1057                 ret =  btrfs_commit_super(root);
1058                 WARN_ON(ret);
1059         } else {
1060                 if (root->fs_info->fs_devices->rw_devices == 0)
1061                         return -EACCES;
1062
1063                 if (btrfs_super_log_root(root->fs_info->super_copy) != 0)
1064                         return -EINVAL;
1065
1066                 ret = btrfs_cleanup_fs_roots(root->fs_info);
1067                 WARN_ON(ret);
1068
1069                 /* recover relocation */
1070                 ret = btrfs_recover_relocation(root);
1071                 WARN_ON(ret);
1072
1073                 sb->s_flags &= ~MS_RDONLY;
1074         }
1075
1076         return 0;
1077 }
1078
1079 /* Used to sort the devices by max_avail(descending sort) */
1080 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1081                                        const void *dev_info2)
1082 {
1083         if (((struct btrfs_device_info *)dev_info1)->max_avail >
1084             ((struct btrfs_device_info *)dev_info2)->max_avail)
1085                 return -1;
1086         else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1087                  ((struct btrfs_device_info *)dev_info2)->max_avail)
1088                 return 1;
1089         else
1090         return 0;
1091 }
1092
1093 /*
1094  * sort the devices by max_avail, in which max free extent size of each device
1095  * is stored.(Descending Sort)
1096  */
1097 static inline void btrfs_descending_sort_devices(
1098                                         struct btrfs_device_info *devices,
1099                                         size_t nr_devices)
1100 {
1101         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1102              btrfs_cmp_device_free_bytes, NULL);
1103 }
1104
1105 /*
1106  * The helper to calc the free space on the devices that can be used to store
1107  * file data.
1108  */
1109 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1110 {
1111         struct btrfs_fs_info *fs_info = root->fs_info;
1112         struct btrfs_device_info *devices_info;
1113         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1114         struct btrfs_device *device;
1115         u64 skip_space;
1116         u64 type;
1117         u64 avail_space;
1118         u64 used_space;
1119         u64 min_stripe_size;
1120         int min_stripes = 1, num_stripes = 1;
1121         int i = 0, nr_devices;
1122         int ret;
1123
1124         nr_devices = fs_info->fs_devices->open_devices;
1125         BUG_ON(!nr_devices);
1126
1127         devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1128                                GFP_NOFS);
1129         if (!devices_info)
1130                 return -ENOMEM;
1131
1132         /* calc min stripe number for data space alloction */
1133         type = btrfs_get_alloc_profile(root, 1);
1134         if (type & BTRFS_BLOCK_GROUP_RAID0) {
1135                 min_stripes = 2;
1136                 num_stripes = nr_devices;
1137         } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1138                 min_stripes = 2;
1139                 num_stripes = 2;
1140         } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1141                 min_stripes = 4;
1142                 num_stripes = 4;
1143         }
1144
1145         if (type & BTRFS_BLOCK_GROUP_DUP)
1146                 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1147         else
1148                 min_stripe_size = BTRFS_STRIPE_LEN;
1149
1150         list_for_each_entry(device, &fs_devices->devices, dev_list) {
1151                 if (!device->in_fs_metadata || !device->bdev)
1152                         continue;
1153
1154                 avail_space = device->total_bytes - device->bytes_used;
1155
1156                 /* align with stripe_len */
1157                 do_div(avail_space, BTRFS_STRIPE_LEN);
1158                 avail_space *= BTRFS_STRIPE_LEN;
1159
1160                 /*
1161                  * In order to avoid overwritting the superblock on the drive,
1162                  * btrfs starts at an offset of at least 1MB when doing chunk
1163                  * allocation.
1164                  */
1165                 skip_space = 1024 * 1024;
1166
1167                 /* user can set the offset in fs_info->alloc_start. */
1168                 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1169                     device->total_bytes)
1170                         skip_space = max(fs_info->alloc_start, skip_space);
1171
1172                 /*
1173                  * btrfs can not use the free space in [0, skip_space - 1],
1174                  * we must subtract it from the total. In order to implement
1175                  * it, we account the used space in this range first.
1176                  */
1177                 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1178                                                      &used_space);
1179                 if (ret) {
1180                         kfree(devices_info);
1181                         return ret;
1182                 }
1183
1184                 /* calc the free space in [0, skip_space - 1] */
1185                 skip_space -= used_space;
1186
1187                 /*
1188                  * we can use the free space in [0, skip_space - 1], subtract
1189                  * it from the total.
1190                  */
1191                 if (avail_space && avail_space >= skip_space)
1192                         avail_space -= skip_space;
1193                 else
1194                         avail_space = 0;
1195
1196                 if (avail_space < min_stripe_size)
1197                         continue;
1198
1199                 devices_info[i].dev = device;
1200                 devices_info[i].max_avail = avail_space;
1201
1202                 i++;
1203         }
1204
1205         nr_devices = i;
1206
1207         btrfs_descending_sort_devices(devices_info, nr_devices);
1208
1209         i = nr_devices - 1;
1210         avail_space = 0;
1211         while (nr_devices >= min_stripes) {
1212                 if (num_stripes > nr_devices)
1213                         num_stripes = nr_devices;
1214
1215                 if (devices_info[i].max_avail >= min_stripe_size) {
1216                         int j;
1217                         u64 alloc_size;
1218
1219                         avail_space += devices_info[i].max_avail * num_stripes;
1220                         alloc_size = devices_info[i].max_avail;
1221                         for (j = i + 1 - num_stripes; j <= i; j++)
1222                                 devices_info[j].max_avail -= alloc_size;
1223                 }
1224                 i--;
1225                 nr_devices--;
1226         }
1227
1228         kfree(devices_info);
1229         *free_bytes = avail_space;
1230         return 0;
1231 }
1232
1233 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1234 {
1235         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
1236         struct btrfs_super_block *disk_super = root->fs_info->super_copy;
1237         struct list_head *head = &root->fs_info->space_info;
1238         struct btrfs_space_info *found;
1239         u64 total_used = 0;
1240         u64 total_free_data = 0;
1241         int bits = dentry->d_sb->s_blocksize_bits;
1242         __be32 *fsid = (__be32 *)root->fs_info->fsid;
1243         int ret;
1244
1245         /* holding chunk_muext to avoid allocating new chunks */
1246         mutex_lock(&root->fs_info->chunk_mutex);
1247         rcu_read_lock();
1248         list_for_each_entry_rcu(found, head, list) {
1249                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1250                         total_free_data += found->disk_total - found->disk_used;
1251                         total_free_data -=
1252                                 btrfs_account_ro_block_groups_free_space(found);
1253                 }
1254
1255                 total_used += found->disk_used;
1256         }
1257         rcu_read_unlock();
1258
1259         buf->f_namelen = BTRFS_NAME_LEN;
1260         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1261         buf->f_bfree = buf->f_blocks - (total_used >> bits);
1262         buf->f_bsize = dentry->d_sb->s_blocksize;
1263         buf->f_type = BTRFS_SUPER_MAGIC;
1264         buf->f_bavail = total_free_data;
1265         ret = btrfs_calc_avail_data_space(root, &total_free_data);
1266         if (ret) {
1267                 mutex_unlock(&root->fs_info->chunk_mutex);
1268                 return ret;
1269         }
1270         buf->f_bavail += total_free_data;
1271         buf->f_bavail = buf->f_bavail >> bits;
1272         mutex_unlock(&root->fs_info->chunk_mutex);
1273
1274         /* We treat it as constant endianness (it doesn't matter _which_)
1275            because we want the fsid to come out the same whether mounted
1276            on a big-endian or little-endian host */
1277         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1278         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1279         /* Mask in the root object ID too, to disambiguate subvols */
1280         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1281         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1282
1283         return 0;
1284 }
1285
1286 static struct file_system_type btrfs_fs_type = {
1287         .owner          = THIS_MODULE,
1288         .name           = "btrfs",
1289         .mount          = btrfs_mount,
1290         .kill_sb        = kill_anon_super,
1291         .fs_flags       = FS_REQUIRES_DEV,
1292 };
1293
1294 /*
1295  * used by btrfsctl to scan devices when no FS is mounted
1296  */
1297 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1298                                 unsigned long arg)
1299 {
1300         struct btrfs_ioctl_vol_args *vol;
1301         struct btrfs_fs_devices *fs_devices;
1302         int ret = -ENOTTY;
1303
1304         if (!capable(CAP_SYS_ADMIN))
1305                 return -EPERM;
1306
1307         vol = memdup_user((void __user *)arg, sizeof(*vol));
1308         if (IS_ERR(vol))
1309                 return PTR_ERR(vol);
1310
1311         switch (cmd) {
1312         case BTRFS_IOC_SCAN_DEV:
1313                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1314                                             &btrfs_fs_type, &fs_devices);
1315                 break;
1316         }
1317
1318         kfree(vol);
1319         return ret;
1320 }
1321
1322 static int btrfs_freeze(struct super_block *sb)
1323 {
1324         struct btrfs_root *root = btrfs_sb(sb);
1325         mutex_lock(&root->fs_info->transaction_kthread_mutex);
1326         mutex_lock(&root->fs_info->cleaner_mutex);
1327         return 0;
1328 }
1329
1330 static int btrfs_unfreeze(struct super_block *sb)
1331 {
1332         struct btrfs_root *root = btrfs_sb(sb);
1333         mutex_unlock(&root->fs_info->cleaner_mutex);
1334         mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1335         return 0;
1336 }
1337
1338 static void btrfs_fs_dirty_inode(struct inode *inode, int flags)
1339 {
1340         int ret;
1341
1342         ret = btrfs_dirty_inode(inode);
1343         if (ret)
1344                 printk_ratelimited(KERN_ERR "btrfs: fail to dirty inode %Lu "
1345                                    "error %d\n", btrfs_ino(inode), ret);
1346 }
1347
1348 static const struct super_operations btrfs_super_ops = {
1349         .drop_inode     = btrfs_drop_inode,
1350         .evict_inode    = btrfs_evict_inode,
1351         .put_super      = btrfs_put_super,
1352         .sync_fs        = btrfs_sync_fs,
1353         .show_options   = btrfs_show_options,
1354         .write_inode    = btrfs_write_inode,
1355         .dirty_inode    = btrfs_fs_dirty_inode,
1356         .alloc_inode    = btrfs_alloc_inode,
1357         .destroy_inode  = btrfs_destroy_inode,
1358         .statfs         = btrfs_statfs,
1359         .remount_fs     = btrfs_remount,
1360         .freeze_fs      = btrfs_freeze,
1361         .unfreeze_fs    = btrfs_unfreeze,
1362 };
1363
1364 static const struct file_operations btrfs_ctl_fops = {
1365         .unlocked_ioctl  = btrfs_control_ioctl,
1366         .compat_ioctl = btrfs_control_ioctl,
1367         .owner   = THIS_MODULE,
1368         .llseek = noop_llseek,
1369 };
1370
1371 static struct miscdevice btrfs_misc = {
1372         .minor          = BTRFS_MINOR,
1373         .name           = "btrfs-control",
1374         .fops           = &btrfs_ctl_fops
1375 };
1376
1377 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1378 MODULE_ALIAS("devname:btrfs-control");
1379
1380 static int btrfs_interface_init(void)
1381 {
1382         return misc_register(&btrfs_misc);
1383 }
1384
1385 static void btrfs_interface_exit(void)
1386 {
1387         if (misc_deregister(&btrfs_misc) < 0)
1388                 printk(KERN_INFO "misc_deregister failed for control device");
1389 }
1390
1391 static int __init init_btrfs_fs(void)
1392 {
1393         int err;
1394
1395         err = btrfs_init_sysfs();
1396         if (err)
1397                 return err;
1398
1399         err = btrfs_init_compress();
1400         if (err)
1401                 goto free_sysfs;
1402
1403         err = btrfs_init_cachep();
1404         if (err)
1405                 goto free_compress;
1406
1407         err = extent_io_init();
1408         if (err)
1409                 goto free_cachep;
1410
1411         err = extent_map_init();
1412         if (err)
1413                 goto free_extent_io;
1414
1415         err = btrfs_delayed_inode_init();
1416         if (err)
1417                 goto free_extent_map;
1418
1419         err = btrfs_interface_init();
1420         if (err)
1421                 goto free_delayed_inode;
1422
1423         err = register_filesystem(&btrfs_fs_type);
1424         if (err)
1425                 goto unregister_ioctl;
1426
1427         printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1428         return 0;
1429
1430 unregister_ioctl:
1431         btrfs_interface_exit();
1432 free_delayed_inode:
1433         btrfs_delayed_inode_exit();
1434 free_extent_map:
1435         extent_map_exit();
1436 free_extent_io:
1437         extent_io_exit();
1438 free_cachep:
1439         btrfs_destroy_cachep();
1440 free_compress:
1441         btrfs_exit_compress();
1442 free_sysfs:
1443         btrfs_exit_sysfs();
1444         return err;
1445 }
1446
1447 static void __exit exit_btrfs_fs(void)
1448 {
1449         btrfs_destroy_cachep();
1450         btrfs_delayed_inode_exit();
1451         extent_map_exit();
1452         extent_io_exit();
1453         btrfs_interface_exit();
1454         unregister_filesystem(&btrfs_fs_type);
1455         btrfs_exit_sysfs();
1456         btrfs_cleanup_fs_uuids();
1457         btrfs_exit_compress();
1458 }
1459
1460 module_init(init_btrfs_fs)
1461 module_exit(exit_btrfs_fs)
1462
1463 MODULE_LICENSE("GPL");