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