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 "compat.h"
  42 #include "ctree.h"
  43 #include "disk-io.h"
  44 #include "transaction.h"
  45 #include "btrfs_inode.h"
  46 #include "ioctl.h"
  47 #include "print-tree.h"
  48 #include "xattr.h"
  49 #include "volumes.h"
  50 #include "version.h"
  51 #include "export.h"
  52 #include "compression.h"
  53
  54 static const struct super_operations btrfs_super_ops;
  55
  56 static void btrfs_put_super(struct super_block *sb)
  57 {
  58         struct btrfs_root *root = btrfs_sb(sb);
  59         int ret;
  60
  61         ret = close_ctree(root);
  62         sb->s_fs_info = NULL;
  63 }
  64
  65 enum {
  66         Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
  67         Opt_nodatacow, Opt_max_extent, Opt_max_inline, Opt_alloc_start,
  68         Opt_nobarrier, Opt_ssd, Opt_nossd, Opt_ssd_spread, Opt_thread_pool,
  69         Opt_noacl, Opt_compress, Opt_compress_force, Opt_notreelog, Opt_ratio,
  70         Opt_flushoncommit,
  71         Opt_discard, Opt_err,
  72 };
  73
  74 static match_table_t tokens = {
  75         {Opt_degraded, "degraded"},
  76         {Opt_subvol, "subvol=%s"},
  77         {Opt_subvolid, "subvolid=%d"},
  78         {Opt_device, "device=%s"},
  79         {Opt_nodatasum, "nodatasum"},
  80         {Opt_nodatacow, "nodatacow"},
  81         {Opt_nobarrier, "nobarrier"},
  82         {Opt_max_extent, "max_extent=%s"},
  83         {Opt_max_inline, "max_inline=%s"},
  84         {Opt_alloc_start, "alloc_start=%s"},
  85         {Opt_thread_pool, "thread_pool=%d"},
  86         {Opt_compress, "compress"},
  87         {Opt_compress_force, "compress-force"},
  88         {Opt_ssd, "ssd"},
  89         {Opt_ssd_spread, "ssd_spread"},
  90         {Opt_nossd, "nossd"},
  91         {Opt_noacl, "noacl"},
  92         {Opt_notreelog, "notreelog"},
  93         {Opt_flushoncommit, "flushoncommit"},
  94         {Opt_ratio, "metadata_ratio=%d"},
  95         {Opt_discard, "discard"},
  96         {Opt_err, NULL},
  97 };
  98
  99 u64 btrfs_parse_size(char *str)
 100 {
 101         u64 res;
 102         int mult = 1;
 103         char *end;
 104         char last;
 105
 106         res = simple_strtoul(str, &end, 10);
 107
 108         last = end[0];
 109         if (isalpha(last)) {
 110                 last = tolower(last);
 111                 switch (last) {
 112                 case 'g':
 113                         mult *= 1024;
 114                 case 'm':
 115                         mult *= 1024;
 116                 case 'k':
 117                         mult *= 1024;
 118                 }
 119                 res = res * mult;
 120         }
 121         return res;
 122 }
 123
 124 /*
 125  * Regular mount options parser.  Everything that is needed only when
 126  * reading in a new superblock is parsed here.
 127  */
 128 int btrfs_parse_options(struct btrfs_root *root, char *options)
 129 {
 130         struct btrfs_fs_info *info = root->fs_info;
 131         substring_t args[MAX_OPT_ARGS];
 132         char *p, *num, *orig;
 133         int intarg;
 134         int ret = 0;
 135
 136         if (!options)
 137                 return 0;
 138
 139         /*
 140          * strsep changes the string, duplicate it because parse_options
 141          * gets called twice
 142          */
 143         options = kstrdup(options, GFP_NOFS);
 144         if (!options)
 145                 return -ENOMEM;
 146
 147         orig = options;
 148
 149         while ((p = strsep(&options, ",")) != NULL) {
 150                 int token;
 151                 if (!*p)
 152                         continue;
 153
 154                 token = match_token(p, tokens, args);
 155                 switch (token) {
 156                 case Opt_degraded:
 157                         printk(KERN_INFO "btrfs: allowing degraded mounts\n");
 158                         btrfs_set_opt(info->mount_opt, DEGRADED);
 159                         break;
 160                 case Opt_subvol:
 161                 case Opt_subvolid:
 162                 case Opt_device:
 163                         /*
 164                          * These are parsed by btrfs_parse_early_options
 165                          * and can be happily ignored here.
 166                          */
 167                         break;
 168                 case Opt_nodatasum:
 169                         printk(KERN_INFO "btrfs: setting nodatasum\n");
 170                         btrfs_set_opt(info->mount_opt, NODATASUM);
 171                         break;
 172                 case Opt_nodatacow:
 173                         printk(KERN_INFO "btrfs: setting nodatacow\n");
 174                         btrfs_set_opt(info->mount_opt, NODATACOW);
 175                         btrfs_set_opt(info->mount_opt, NODATASUM);
 176                         break;
 177                 case Opt_compress:
 178                         printk(KERN_INFO "btrfs: use compression\n");
 179                         btrfs_set_opt(info->mount_opt, COMPRESS);
 180                         break;
 181                 case Opt_compress_force:
 182                         printk(KERN_INFO "btrfs: forcing compression\n");
 183                         btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 184                         btrfs_set_opt(info->mount_opt, COMPRESS);
 185                         break;
 186                 case Opt_ssd:
 187                         printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 188                         btrfs_set_opt(info->mount_opt, SSD);
 189                         break;
 190                 case Opt_ssd_spread:
 191                         printk(KERN_INFO "btrfs: use spread ssd "
 192                                "allocation scheme\n");
 193                         btrfs_set_opt(info->mount_opt, SSD);
 194                         btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 195                         break;
 196                 case Opt_nossd:
 197                         printk(KERN_INFO "btrfs: not using ssd allocation "
 198                                "scheme\n");
 199                         btrfs_set_opt(info->mount_opt, NOSSD);
 200                         btrfs_clear_opt(info->mount_opt, SSD);
 201                         btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 202                         break;
 203                 case Opt_nobarrier:
 204                         printk(KERN_INFO "btrfs: turning off barriers\n");
 205                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 206                         break;
 207                 case Opt_thread_pool:
 208                         intarg = 0;
 209                         match_int(&args[0], &intarg);
 210                         if (intarg) {
 211                                 info->thread_pool_size = intarg;
 212                                 printk(KERN_INFO "btrfs: thread pool %d\n",
 213                                        info->thread_pool_size);
 214                         }
 215                         break;
 216                 case Opt_max_extent:
 217                         num = match_strdup(&args[0]);
 218                         if (num) {
 219                                 info->max_extent = btrfs_parse_size(num);
 220                                 kfree(num);
 221
 222                                 info->max_extent = max_t(u64,
 223                                         info->max_extent, root->sectorsize);
 224                                 printk(KERN_INFO "btrfs: max_extent at %llu\n",
 225                                        (unsigned long long)info->max_extent);
 226                         }
 227                         break;
 228                 case Opt_max_inline:
 229                         num = match_strdup(&args[0]);
 230                         if (num) {
 231                                 info->max_inline = btrfs_parse_size(num);
 232                                 kfree(num);
 233
 234                                 if (info->max_inline) {
 235                                         info->max_inline = max_t(u64,
 236                                                 info->max_inline,
 237                                                 root->sectorsize);
 238                                 }
 239                                 printk(KERN_INFO "btrfs: max_inline at %llu\n",
 240                                         (unsigned long long)info->max_inline);
 241                         }
 242                         break;
 243                 case Opt_alloc_start:
 244                         num = match_strdup(&args[0]);
 245                         if (num) {
 246                                 info->alloc_start = btrfs_parse_size(num);
 247                                 kfree(num);
 248                                 printk(KERN_INFO
 249                                         "btrfs: allocations start at %llu\n",
 250                                         (unsigned long long)info->alloc_start);
 251                         }
 252                         break;
 253                 case Opt_noacl:
 254                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 255                         break;
 256                 case Opt_notreelog:
 257                         printk(KERN_INFO "btrfs: disabling tree log\n");
 258                         btrfs_set_opt(info->mount_opt, NOTREELOG);
 259                         break;
 260                 case Opt_flushoncommit:
 261                         printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
 262                         btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
 263                         break;
 264                 case Opt_ratio:
 265                         intarg = 0;
 266                         match_int(&args[0], &intarg);
 267                         if (intarg) {
 268                                 info->metadata_ratio = intarg;
 269                                 printk(KERN_INFO "btrfs: metadata ratio %d\n",
 270                                        info->metadata_ratio);
 271                         }
 272                         break;
 273                 case Opt_discard:
 274                         btrfs_set_opt(info->mount_opt, DISCARD);
 275                         break;
 276                 case Opt_err:
 277                         printk(KERN_INFO "btrfs: unrecognized mount option "
 278                                "'%s'\n", p);
 279                         ret = -EINVAL;
 280                         goto out;
 281                 default:
 282                         break;
 283                 }
 284         }
 285 out:
 286         kfree(orig);
 287         return ret;
 288 }
 289
 290 /*
 291  * Parse mount options that are required early in the mount process.
 292  *
 293  * All other options will be parsed on much later in the mount process and
 294  * only when we need to allocate a new super block.
 295  */
 296 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 297                 void *holder, char **subvol_name, u64 *subvol_objectid,
 298                 struct btrfs_fs_devices **fs_devices)
 299 {
 300         substring_t args[MAX_OPT_ARGS];
 301         char *opts, *p;
 302         int error = 0;
 303         int intarg;
 304
 305         if (!options)
 306                 goto out;
 307
 308         /*
 309          * strsep changes the string, duplicate it because parse_options
 310          * gets called twice
 311          */
 312         opts = kstrdup(options, GFP_KERNEL);
 313         if (!opts)
 314                 return -ENOMEM;
 315
 316         while ((p = strsep(&opts, ",")) != NULL) {
 317                 int token;
 318                 if (!*p)
 319                         continue;
 320
 321                 token = match_token(p, tokens, args);
 322                 switch (token) {
 323                 case Opt_subvol:
 324                         *subvol_name = match_strdup(&args[0]);
 325                         break;
 326                 case Opt_subvolid:
 327                         intarg = 0;
 328                         match_int(&args[0], &intarg);
 329                         if (intarg)
 330                                 *subvol_objectid = intarg;
 331                         break;
 332                 case Opt_device:
 333                         error = btrfs_scan_one_device(match_strdup(&args[0]),
 334                                         flags, holder, fs_devices);
 335                         if (error)
 336                                 goto out_free_opts;
 337                         break;
 338                 default:
 339                         break;
 340                 }
 341         }
 342
 343  out_free_opts:
 344         kfree(opts);
 345  out:
 346         /*
 347          * If no subvolume name is specified we use the default one.  Allocate
 348          * a copy of the string "." here so that code later in the
 349          * mount path doesn't care if it's the default volume or another one.
 350          */
 351         if (!*subvol_name) {
 352                 *subvol_name = kstrdup(".", GFP_KERNEL);
 353                 if (!*subvol_name)
 354                         return -ENOMEM;
 355         }
 356         return error;
 357 }
 358
 359 static struct dentry *get_default_root(struct super_block *sb,
 360                                        u64 subvol_objectid)
 361 {
 362         struct btrfs_root *root = sb->s_fs_info;
 363         struct btrfs_root *new_root;
 364         struct btrfs_dir_item *di;
 365         struct btrfs_path *path;
 366         struct btrfs_key location;
 367         struct inode *inode;
 368         struct dentry *dentry;
 369         u64 dir_id;
 370         int new = 0;
 371
 372         /*
 373          * We have a specific subvol we want to mount, just setup location and
 374          * go look up the root.
 375          */
 376         if (subvol_objectid) {
 377                 location.objectid = subvol_objectid;
 378                 location.type = BTRFS_ROOT_ITEM_KEY;
 379                 location.offset = (u64)-1;
 380                 goto find_root;
 381         }
 382
 383         path = btrfs_alloc_path();
 384         if (!path)
 385                 return ERR_PTR(-ENOMEM);
 386         path->leave_spinning = 1;
 387
 388         /*
 389          * Find the "default" dir item which points to the root item that we
 390          * will mount by default if we haven't been given a specific subvolume
 391          * to mount.
 392          */
 393         dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
 394         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 395         if (!di) {
 396                 /*
 397                  * Ok the default dir item isn't there.  This is weird since
 398                  * it's always been there, but don't freak out, just try and
 399                  * mount to root most subvolume.
 400                  */
 401                 btrfs_free_path(path);
 402                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 403                 new_root = root->fs_info->fs_root;
 404                 goto setup_root;
 405         }
 406
 407         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 408         btrfs_free_path(path);
 409
 410 find_root:
 411         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
 412         if (IS_ERR(new_root))
 413                 return ERR_PTR(PTR_ERR(new_root));
 414
 415         if (btrfs_root_refs(&new_root->root_item) == 0)
 416                 return ERR_PTR(-ENOENT);
 417
 418         dir_id = btrfs_root_dirid(&new_root->root_item);
 419 setup_root:
 420         location.objectid = dir_id;
 421         location.type = BTRFS_INODE_ITEM_KEY;
 422         location.offset = 0;
 423
 424         inode = btrfs_iget(sb, &location, new_root, &new);
 425         if (!inode)
 426                 return ERR_PTR(-ENOMEM);
 427
 428         /*
 429          * If we're just mounting the root most subvol put the inode and return
 430          * a reference to the dentry.  We will have already gotten a reference
 431          * to the inode in btrfs_fill_super so we're good to go.
 432          */
 433         if (!new && sb->s_root->d_inode == inode) {
 434                 iput(inode);
 435                 return dget(sb->s_root);
 436         }
 437
 438         if (new) {
 439                 const struct qstr name = { .name = "/", .len = 1 };
 440
 441                 /*
 442                  * New inode, we need to make the dentry a sibling of s_root so
 443                  * everything gets cleaned up properly on unmount.
 444                  */
 445                 dentry = d_alloc(sb->s_root, &name);
 446                 if (!dentry) {
 447                         iput(inode);
 448                         return ERR_PTR(-ENOMEM);
 449                 }
 450                 d_splice_alias(inode, dentry);
 451         } else {
 452                 /*
 453                  * We found the inode in cache, just find a dentry for it and
 454                  * put the reference to the inode we just got.
 455                  */
 456                 dentry = d_find_alias(inode);
 457                 iput(inode);
 458         }
 459
 460         return dentry;
 461 }
 462
 463 static int btrfs_fill_super(struct super_block *sb,
 464                             struct btrfs_fs_devices *fs_devices,
 465                             void *data, int silent)
 466 {
 467         struct inode *inode;
 468         struct dentry *root_dentry;
 469         struct btrfs_super_block *disk_super;
 470         struct btrfs_root *tree_root;
 471         struct btrfs_key key;
 472         int err;
 473
 474         sb->s_maxbytes = MAX_LFS_FILESIZE;
 475         sb->s_magic = BTRFS_SUPER_MAGIC;
 476         sb->s_op = &btrfs_super_ops;
 477         sb->s_export_op = &btrfs_export_ops;
 478         sb->s_xattr = btrfs_xattr_handlers;
 479         sb->s_time_gran = 1;
 480 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 481         sb->s_flags |= MS_POSIXACL;
 482 #endif
 483
 484         tree_root = open_ctree(sb, fs_devices, (char *)data);
 485
 486         if (IS_ERR(tree_root)) {
 487                 printk("btrfs: open_ctree failed\n");
 488                 return PTR_ERR(tree_root);
 489         }
 490         sb->s_fs_info = tree_root;
 491         disk_super = &tree_root->fs_info->super_copy;
 492
 493         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 494         key.type = BTRFS_INODE_ITEM_KEY;
 495         key.offset = 0;
 496         inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
 497         if (IS_ERR(inode)) {
 498                 err = PTR_ERR(inode);
 499                 goto fail_close;
 500         }
 501
 502         root_dentry = d_alloc_root(inode);
 503         if (!root_dentry) {
 504                 iput(inode);
 505                 err = -ENOMEM;
 506                 goto fail_close;
 507         }
 508
 509         sb->s_root = root_dentry;
 510
 511         save_mount_options(sb, data);
 512         return 0;
 513
 514 fail_close:
 515         close_ctree(tree_root);
 516         return err;
 517 }
 518
 519 int btrfs_sync_fs(struct super_block *sb, int wait)
 520 {
 521         struct btrfs_trans_handle *trans;
 522         struct btrfs_root *root = btrfs_sb(sb);
 523         int ret;
 524
 525         if (!wait) {
 526                 filemap_flush(root->fs_info->btree_inode->i_mapping);
 527                 return 0;
 528         }
 529
 530         btrfs_start_delalloc_inodes(root, 0);
 531         btrfs_wait_ordered_extents(root, 0, 0);
 532
 533         trans = btrfs_start_transaction(root, 1);
 534         ret = btrfs_commit_transaction(trans, root);
 535         return ret;
 536 }
 537
 538 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 539 {
 540         struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
 541         struct btrfs_fs_info *info = root->fs_info;
 542
 543         if (btrfs_test_opt(root, DEGRADED))
 544                 seq_puts(seq, ",degraded");
 545         if (btrfs_test_opt(root, NODATASUM))
 546                 seq_puts(seq, ",nodatasum");
 547         if (btrfs_test_opt(root, NODATACOW))
 548                 seq_puts(seq, ",nodatacow");
 549         if (btrfs_test_opt(root, NOBARRIER))
 550                 seq_puts(seq, ",nobarrier");
 551         if (info->max_extent != (u64)-1)
 552                 seq_printf(seq, ",max_extent=%llu",
 553                            (unsigned long long)info->max_extent);
 554         if (info->max_inline != 8192 * 1024)
 555                 seq_printf(seq, ",max_inline=%llu",
 556                            (unsigned long long)info->max_inline);
 557         if (info->alloc_start != 0)
 558                 seq_printf(seq, ",alloc_start=%llu",
 559                            (unsigned long long)info->alloc_start);
 560         if (info->thread_pool_size !=  min_t(unsigned long,
 561                                              num_online_cpus() + 2, 8))
 562                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
 563         if (btrfs_test_opt(root, COMPRESS))
 564                 seq_puts(seq, ",compress");
 565         if (btrfs_test_opt(root, NOSSD))
 566                 seq_puts(seq, ",nossd");
 567         if (btrfs_test_opt(root, SSD_SPREAD))
 568                 seq_puts(seq, ",ssd_spread");
 569         else if (btrfs_test_opt(root, SSD))
 570                 seq_puts(seq, ",ssd");
 571         if (btrfs_test_opt(root, NOTREELOG))
 572                 seq_puts(seq, ",notreelog");
 573         if (btrfs_test_opt(root, FLUSHONCOMMIT))
 574                 seq_puts(seq, ",flushoncommit");
 575         if (btrfs_test_opt(root, DISCARD))
 576                 seq_puts(seq, ",discard");
 577         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
 578                 seq_puts(seq, ",noacl");
 579         return 0;
 580 }
 581
 582 static int btrfs_test_super(struct super_block *s, void *data)
 583 {
 584         struct btrfs_fs_devices *test_fs_devices = data;
 585         struct btrfs_root *root = btrfs_sb(s);
 586
 587         return root->fs_info->fs_devices == test_fs_devices;
 588 }
 589
 590 /*
 591  * Find a superblock for the given device / mount point.
 592  *
 593  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
 594  *        for multiple device setup.  Make sure to keep it in sync.
 595  */
 596 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 597                 const char *dev_name, void *data, struct vfsmount *mnt)
 598 {
 599         struct block_device *bdev = NULL;
 600         struct super_block *s;
 601         struct dentry *root;
 602         struct btrfs_fs_devices *fs_devices = NULL;
 603         fmode_t mode = FMODE_READ;
 604         char *subvol_name = NULL;
 605         u64 subvol_objectid = 0;
 606         int error = 0;
 607         int found = 0;
 608
 609         if (!(flags & MS_RDONLY))
 610                 mode |= FMODE_WRITE;
 611
 612         error = btrfs_parse_early_options(data, mode, fs_type,
 613                                           &subvol_name, &subvol_objectid,
 614                                           &fs_devices);
 615         if (error)
 616                 return error;
 617
 618         error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
 619         if (error)
 620                 goto error_free_subvol_name;
 621
 622         error = btrfs_open_devices(fs_devices, mode, fs_type);
 623         if (error)
 624                 goto error_free_subvol_name;
 625
 626         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
 627                 error = -EACCES;
 628                 goto error_close_devices;
 629         }
 630
 631         bdev = fs_devices->latest_bdev;
 632         s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
 633         if (IS_ERR(s))
 634                 goto error_s;
 635
 636         if (s->s_root) {
 637                 if ((flags ^ s->s_flags) & MS_RDONLY) {
 638                         deactivate_locked_super(s);
 639                         error = -EBUSY;
 640                         goto error_close_devices;
 641                 }
 642
 643                 found = 1;
 644                 btrfs_close_devices(fs_devices);
 645         } else {
 646                 char b[BDEVNAME_SIZE];
 647
 648                 s->s_flags = flags;
 649                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 650                 error = btrfs_fill_super(s, fs_devices, data,
 651                                          flags & MS_SILENT ? 1 : 0);
 652                 if (error) {
 653                         deactivate_locked_super(s);
 654                         goto error_free_subvol_name;
 655                 }
 656
 657                 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
 658                 s->s_flags |= MS_ACTIVE;
 659         }
 660
 661         root = get_default_root(s, subvol_objectid);
 662         if (IS_ERR(root)) {
 663                 error = PTR_ERR(root);
 664                 deactivate_locked_super(s);
 665                 goto error;
 666         }
 667         /* if they gave us a subvolume name bind mount into that */
 668         if (strcmp(subvol_name, ".")) {
 669                 struct dentry *new_root;
 670                 mutex_lock(&root->d_inode->i_mutex);
 671                 new_root = lookup_one_len(subvol_name, root,
 672                                       strlen(subvol_name));
 673                 mutex_unlock(&root->d_inode->i_mutex);
 674
 675                 if (IS_ERR(new_root)) {
 676                         deactivate_locked_super(s);
 677                         error = PTR_ERR(new_root);
 678                         dput(root);
 679                         goto error_close_devices;
 680                 }
 681                 if (!new_root->d_inode) {
 682                         dput(root);
 683                         dput(new_root);
 684                         deactivate_locked_super(s);
 685                         error = -ENXIO;
 686                         goto error_close_devices;
 687                 }
 688                 dput(root);
 689                 root = new_root;
 690         }
 691
 692         mnt->mnt_sb = s;
 693         mnt->mnt_root = root;
 694
 695         kfree(subvol_name);
 696         return 0;
 697
 698 error_s:
 699         error = PTR_ERR(s);
 700 error_close_devices:
 701         btrfs_close_devices(fs_devices);
 702 error_free_subvol_name:
 703         kfree(subvol_name);
 704 error:
 705         return error;
 706 }
 707
 708 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
 709 {
 710         struct btrfs_root *root = btrfs_sb(sb);
 711         int ret;
 712
 713         ret = btrfs_parse_options(root, data);
 714         if (ret)
 715                 return -EINVAL;
 716
 717         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
 718                 return 0;
 719
 720         if (*flags & MS_RDONLY) {
 721                 sb->s_flags |= MS_RDONLY;
 722
 723                 ret =  btrfs_commit_super(root);
 724                 WARN_ON(ret);
 725         } else {
 726                 if (root->fs_info->fs_devices->rw_devices == 0)
 727                         return -EACCES;
 728
 729                 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
 730                         return -EINVAL;
 731
 732                 /* recover relocation */
 733                 ret = btrfs_recover_relocation(root);
 734                 WARN_ON(ret);
 735
 736                 ret = btrfs_cleanup_fs_roots(root->fs_info);
 737                 WARN_ON(ret);
 738
 739                 sb->s_flags &= ~MS_RDONLY;
 740         }
 741
 742         return 0;
 743 }
 744
 745 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 746 {
 747         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
 748         struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
 749         int bits = dentry->d_sb->s_blocksize_bits;
 750         __be32 *fsid = (__be32 *)root->fs_info->fsid;
 751
 752         buf->f_namelen = BTRFS_NAME_LEN;
 753         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
 754         buf->f_bfree = buf->f_blocks -
 755                 (btrfs_super_bytes_used(disk_super) >> bits);
 756         buf->f_bavail = buf->f_bfree;
 757         buf->f_bsize = dentry->d_sb->s_blocksize;
 758         buf->f_type = BTRFS_SUPER_MAGIC;
 759
 760         /* We treat it as constant endianness (it doesn't matter _which_)
 761            because we want the fsid to come out the same whether mounted
 762            on a big-endian or little-endian host */
 763         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
 764         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
 765         /* Mask in the root object ID too, to disambiguate subvols */
 766         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
 767         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
 768
 769         return 0;
 770 }
 771
 772 static struct file_system_type btrfs_fs_type = {
 773         .owner          = THIS_MODULE,
 774         .name           = "btrfs",
 775         .get_sb         = btrfs_get_sb,
 776         .kill_sb        = kill_anon_super,
 777         .fs_flags       = FS_REQUIRES_DEV,
 778 };
 779
 780 /*
 781  * used by btrfsctl to scan devices when no FS is mounted
 782  */
 783 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
 784                                 unsigned long arg)
 785 {
 786         struct btrfs_ioctl_vol_args *vol;
 787         struct btrfs_fs_devices *fs_devices;
 788         int ret = -ENOTTY;
 789
 790         if (!capable(CAP_SYS_ADMIN))
 791                 return -EPERM;
 792
 793         vol = memdup_user((void __user *)arg, sizeof(*vol));
 794         if (IS_ERR(vol))
 795                 return PTR_ERR(vol);
 796
 797         switch (cmd) {
 798         case BTRFS_IOC_SCAN_DEV:
 799                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
 800                                             &btrfs_fs_type, &fs_devices);
 801                 break;
 802         }
 803
 804         kfree(vol);
 805         return ret;
 806 }
 807
 808 static int btrfs_freeze(struct super_block *sb)
 809 {
 810         struct btrfs_root *root = btrfs_sb(sb);
 811         mutex_lock(&root->fs_info->transaction_kthread_mutex);
 812         mutex_lock(&root->fs_info->cleaner_mutex);
 813         return 0;
 814 }
 815
 816 static int btrfs_unfreeze(struct super_block *sb)
 817 {
 818         struct btrfs_root *root = btrfs_sb(sb);
 819         mutex_unlock(&root->fs_info->cleaner_mutex);
 820         mutex_unlock(&root->fs_info->transaction_kthread_mutex);
 821         return 0;
 822 }
 823
 824 static const struct super_operations btrfs_super_ops = {
 825         .drop_inode     = btrfs_drop_inode,
 826         .delete_inode   = btrfs_delete_inode,
 827         .put_super      = btrfs_put_super,
 828         .sync_fs        = btrfs_sync_fs,
 829         .show_options   = btrfs_show_options,
 830         .write_inode    = btrfs_write_inode,
 831         .dirty_inode    = btrfs_dirty_inode,
 832         .alloc_inode    = btrfs_alloc_inode,
 833         .destroy_inode  = btrfs_destroy_inode,
 834         .statfs         = btrfs_statfs,
 835         .remount_fs     = btrfs_remount,
 836         .freeze_fs      = btrfs_freeze,
 837         .unfreeze_fs    = btrfs_unfreeze,
 838 };
 839
 840 static const struct file_operations btrfs_ctl_fops = {
 841         .unlocked_ioctl  = btrfs_control_ioctl,
 842         .compat_ioctl = btrfs_control_ioctl,
 843         .owner   = THIS_MODULE,
 844 };
 845
 846 static struct miscdevice btrfs_misc = {
 847         .minor          = MISC_DYNAMIC_MINOR,
 848         .name           = "btrfs-control",
 849         .fops           = &btrfs_ctl_fops
 850 };
 851
 852 static int btrfs_interface_init(void)
 853 {
 854         return misc_register(&btrfs_misc);
 855 }
 856
 857 static void btrfs_interface_exit(void)
 858 {
 859         if (misc_deregister(&btrfs_misc) < 0)
 860                 printk(KERN_INFO "misc_deregister failed for control device");
 861 }
 862
 863 static int __init init_btrfs_fs(void)
 864 {
 865         int err;
 866
 867         err = btrfs_init_sysfs();
 868         if (err)
 869                 return err;
 870
 871         err = btrfs_init_cachep();
 872         if (err)
 873                 goto free_sysfs;
 874
 875         err = extent_io_init();
 876         if (err)
 877                 goto free_cachep;
 878
 879         err = extent_map_init();
 880         if (err)
 881                 goto free_extent_io;
 882
 883         err = btrfs_interface_init();
 884         if (err)
 885                 goto free_extent_map;
 886
 887         err = register_filesystem(&btrfs_fs_type);
 888         if (err)
 889                 goto unregister_ioctl;
 890
 891         printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
 892         return 0;
 893
 894 unregister_ioctl:
 895         btrfs_interface_exit();
 896 free_extent_map:
 897         extent_map_exit();
 898 free_extent_io:
 899         extent_io_exit();
 900 free_cachep:
 901         btrfs_destroy_cachep();
 902 free_sysfs:
 903         btrfs_exit_sysfs();
 904         return err;
 905 }
 906
 907 static void __exit exit_btrfs_fs(void)
 908 {
 909         btrfs_destroy_cachep();
 910         extent_map_exit();
 911         extent_io_exit();
 912         btrfs_interface_exit();
 913         unregister_filesystem(&btrfs_fs_type);
 914         btrfs_exit_sysfs();
 915         btrfs_cleanup_fs_uuids();
 916         btrfs_zlib_exit();
 917 }
 918
 919 module_init(init_btrfs_fs)
 920 module_exit(exit_btrfs_fs)
 921
 922 MODULE_LICENSE("GPL");