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 /*
 100  * Regular mount options parser.  Everything that is needed only when
 101  * reading in a new superblock is parsed here.
 102  */
 103 int btrfs_parse_options(struct btrfs_root *root, char *options)
 104 {
 105         struct btrfs_fs_info *info = root->fs_info;
 106         substring_t args[MAX_OPT_ARGS];
 107         char *p, *num, *orig;
 108         int intarg;
 109         int ret = 0;
 110
 111         if (!options)
 112                 return 0;
 113
 114         /*
 115          * strsep changes the string, duplicate it because parse_options
 116          * gets called twice
 117          */
 118         options = kstrdup(options, GFP_NOFS);
 119         if (!options)
 120                 return -ENOMEM;
 121
 122         orig = options;
 123
 124         while ((p = strsep(&options, ",")) != NULL) {
 125                 int token;
 126                 if (!*p)
 127                         continue;
 128
 129                 token = match_token(p, tokens, args);
 130                 switch (token) {
 131                 case Opt_degraded:
 132                         printk(KERN_INFO "btrfs: allowing degraded mounts\n");
 133                         btrfs_set_opt(info->mount_opt, DEGRADED);
 134                         break;
 135                 case Opt_subvol:
 136                 case Opt_subvolid:
 137                 case Opt_device:
 138                         /*
 139                          * These are parsed by btrfs_parse_early_options
 140                          * and can be happily ignored here.
 141                          */
 142                         break;
 143                 case Opt_nodatasum:
 144                         printk(KERN_INFO "btrfs: setting nodatasum\n");
 145                         btrfs_set_opt(info->mount_opt, NODATASUM);
 146                         break;
 147                 case Opt_nodatacow:
 148                         printk(KERN_INFO "btrfs: setting nodatacow\n");
 149                         btrfs_set_opt(info->mount_opt, NODATACOW);
 150                         btrfs_set_opt(info->mount_opt, NODATASUM);
 151                         break;
 152                 case Opt_compress:
 153                         printk(KERN_INFO "btrfs: use compression\n");
 154                         btrfs_set_opt(info->mount_opt, COMPRESS);
 155                         break;
 156                 case Opt_compress_force:
 157                         printk(KERN_INFO "btrfs: forcing compression\n");
 158                         btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 159                         btrfs_set_opt(info->mount_opt, COMPRESS);
 160                         break;
 161                 case Opt_ssd:
 162                         printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 163                         btrfs_set_opt(info->mount_opt, SSD);
 164                         break;
 165                 case Opt_ssd_spread:
 166                         printk(KERN_INFO "btrfs: use spread ssd "
 167                                "allocation scheme\n");
 168                         btrfs_set_opt(info->mount_opt, SSD);
 169                         btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 170                         break;
 171                 case Opt_nossd:
 172                         printk(KERN_INFO "btrfs: not using ssd allocation "
 173                                "scheme\n");
 174                         btrfs_set_opt(info->mount_opt, NOSSD);
 175                         btrfs_clear_opt(info->mount_opt, SSD);
 176                         btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 177                         break;
 178                 case Opt_nobarrier:
 179                         printk(KERN_INFO "btrfs: turning off barriers\n");
 180                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 181                         break;
 182                 case Opt_thread_pool:
 183                         intarg = 0;
 184                         match_int(&args[0], &intarg);
 185                         if (intarg) {
 186                                 info->thread_pool_size = intarg;
 187                                 printk(KERN_INFO "btrfs: thread pool %d\n",
 188                                        info->thread_pool_size);
 189                         }
 190                         break;
 191                 case Opt_max_extent:
 192                         num = match_strdup(&args[0]);
 193                         if (num) {
 194                                 info->max_extent = memparse(num, NULL);
 195                                 kfree(num);
 196
 197                                 info->max_extent = max_t(u64,
 198                                         info->max_extent, root->sectorsize);
 199                                 printk(KERN_INFO "btrfs: max_extent at %llu\n",
 200                                        (unsigned long long)info->max_extent);
 201                         }
 202                         break;
 203                 case Opt_max_inline:
 204                         num = match_strdup(&args[0]);
 205                         if (num) {
 206                                 info->max_inline = memparse(num, NULL);
 207                                 kfree(num);
 208
 209                                 if (info->max_inline) {
 210                                         info->max_inline = max_t(u64,
 211                                                 info->max_inline,
 212                                                 root->sectorsize);
 213                                 }
 214                                 printk(KERN_INFO "btrfs: max_inline at %llu\n",
 215                                         (unsigned long long)info->max_inline);
 216                         }
 217                         break;
 218                 case Opt_alloc_start:
 219                         num = match_strdup(&args[0]);
 220                         if (num) {
 221                                 info->alloc_start = memparse(num, NULL);
 222                                 kfree(num);
 223                                 printk(KERN_INFO
 224                                         "btrfs: allocations start at %llu\n",
 225                                         (unsigned long long)info->alloc_start);
 226                         }
 227                         break;
 228                 case Opt_noacl:
 229                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 230                         break;
 231                 case Opt_notreelog:
 232                         printk(KERN_INFO "btrfs: disabling tree log\n");
 233                         btrfs_set_opt(info->mount_opt, NOTREELOG);
 234                         break;
 235                 case Opt_flushoncommit:
 236                         printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
 237                         btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
 238                         break;
 239                 case Opt_ratio:
 240                         intarg = 0;
 241                         match_int(&args[0], &intarg);
 242                         if (intarg) {
 243                                 info->metadata_ratio = intarg;
 244                                 printk(KERN_INFO "btrfs: metadata ratio %d\n",
 245                                        info->metadata_ratio);
 246                         }
 247                         break;
 248                 case Opt_discard:
 249                         btrfs_set_opt(info->mount_opt, DISCARD);
 250                         break;
 251                 case Opt_err:
 252                         printk(KERN_INFO "btrfs: unrecognized mount option "
 253                                "'%s'\n", p);
 254                         ret = -EINVAL;
 255                         goto out;
 256                 default:
 257                         break;
 258                 }
 259         }
 260 out:
 261         kfree(orig);
 262         return ret;
 263 }
 264
 265 /*
 266  * Parse mount options that are required early in the mount process.
 267  *
 268  * All other options will be parsed on much later in the mount process and
 269  * only when we need to allocate a new super block.
 270  */
 271 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 272                 void *holder, char **subvol_name, u64 *subvol_objectid,
 273                 struct btrfs_fs_devices **fs_devices)
 274 {
 275         substring_t args[MAX_OPT_ARGS];
 276         char *opts, *p;
 277         int error = 0;
 278         int intarg;
 279
 280         if (!options)
 281                 goto out;
 282
 283         /*
 284          * strsep changes the string, duplicate it because parse_options
 285          * gets called twice
 286          */
 287         opts = kstrdup(options, GFP_KERNEL);
 288         if (!opts)
 289                 return -ENOMEM;
 290
 291         while ((p = strsep(&opts, ",")) != NULL) {
 292                 int token;
 293                 if (!*p)
 294                         continue;
 295
 296                 token = match_token(p, tokens, args);
 297                 switch (token) {
 298                 case Opt_subvol:
 299                         *subvol_name = match_strdup(&args[0]);
 300                         break;
 301                 case Opt_subvolid:
 302                         intarg = 0;
 303                         error = match_int(&args[0], &intarg);
 304                         if (!error) {
 305                                 /* we want the original fs_tree */
 306                                 if (!intarg)
 307                                         *subvol_objectid =
 308                                                 BTRFS_FS_TREE_OBJECTID;
 309                                 else
 310                                         *subvol_objectid = intarg;
 311                         }
 312                         break;
 313                 case Opt_device:
 314                         error = btrfs_scan_one_device(match_strdup(&args[0]),
 315                                         flags, holder, fs_devices);
 316                         if (error)
 317                                 goto out_free_opts;
 318                         break;
 319                 default:
 320                         break;
 321                 }
 322         }
 323
 324  out_free_opts:
 325         kfree(opts);
 326  out:
 327         /*
 328          * If no subvolume name is specified we use the default one.  Allocate
 329          * a copy of the string "." here so that code later in the
 330          * mount path doesn't care if it's the default volume or another one.
 331          */
 332         if (!*subvol_name) {
 333                 *subvol_name = kstrdup(".", GFP_KERNEL);
 334                 if (!*subvol_name)
 335                         return -ENOMEM;
 336         }
 337         return error;
 338 }
 339
 340 static struct dentry *get_default_root(struct super_block *sb,
 341                                        u64 subvol_objectid)
 342 {
 343         struct btrfs_root *root = sb->s_fs_info;
 344         struct btrfs_root *new_root;
 345         struct btrfs_dir_item *di;
 346         struct btrfs_path *path;
 347         struct btrfs_key location;
 348         struct inode *inode;
 349         struct dentry *dentry;
 350         u64 dir_id;
 351         int new = 0;
 352
 353         /*
 354          * We have a specific subvol we want to mount, just setup location and
 355          * go look up the root.
 356          */
 357         if (subvol_objectid) {
 358                 location.objectid = subvol_objectid;
 359                 location.type = BTRFS_ROOT_ITEM_KEY;
 360                 location.offset = (u64)-1;
 361                 goto find_root;
 362         }
 363
 364         path = btrfs_alloc_path();
 365         if (!path)
 366                 return ERR_PTR(-ENOMEM);
 367         path->leave_spinning = 1;
 368
 369         /*
 370          * Find the "default" dir item which points to the root item that we
 371          * will mount by default if we haven't been given a specific subvolume
 372          * to mount.
 373          */
 374         dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
 375         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 376         if (!di) {
 377                 /*
 378                  * Ok the default dir item isn't there.  This is weird since
 379                  * it's always been there, but don't freak out, just try and
 380                  * mount to root most subvolume.
 381                  */
 382                 btrfs_free_path(path);
 383                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 384                 new_root = root->fs_info->fs_root;
 385                 goto setup_root;
 386         }
 387
 388         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 389         btrfs_free_path(path);
 390
 391 find_root:
 392         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
 393         if (IS_ERR(new_root))
 394                 return ERR_PTR(PTR_ERR(new_root));
 395
 396         if (btrfs_root_refs(&new_root->root_item) == 0)
 397                 return ERR_PTR(-ENOENT);
 398
 399         dir_id = btrfs_root_dirid(&new_root->root_item);
 400 setup_root:
 401         location.objectid = dir_id;
 402         location.type = BTRFS_INODE_ITEM_KEY;
 403         location.offset = 0;
 404
 405         inode = btrfs_iget(sb, &location, new_root, &new);
 406         if (!inode)
 407                 return ERR_PTR(-ENOMEM);
 408
 409         /*
 410          * If we're just mounting the root most subvol put the inode and return
 411          * a reference to the dentry.  We will have already gotten a reference
 412          * to the inode in btrfs_fill_super so we're good to go.
 413          */
 414         if (!new && sb->s_root->d_inode == inode) {
 415                 iput(inode);
 416                 return dget(sb->s_root);
 417         }
 418
 419         if (new) {
 420                 const struct qstr name = { .name = "/", .len = 1 };
 421
 422                 /*
 423                  * New inode, we need to make the dentry a sibling of s_root so
 424                  * everything gets cleaned up properly on unmount.
 425                  */
 426                 dentry = d_alloc(sb->s_root, &name);
 427                 if (!dentry) {
 428                         iput(inode);
 429                         return ERR_PTR(-ENOMEM);
 430                 }
 431                 d_splice_alias(inode, dentry);
 432         } else {
 433                 /*
 434                  * We found the inode in cache, just find a dentry for it and
 435                  * put the reference to the inode we just got.
 436                  */
 437                 dentry = d_find_alias(inode);
 438                 iput(inode);
 439         }
 440
 441         return dentry;
 442 }
 443
 444 static int btrfs_fill_super(struct super_block *sb,
 445                             struct btrfs_fs_devices *fs_devices,
 446                             void *data, int silent)
 447 {
 448         struct inode *inode;
 449         struct dentry *root_dentry;
 450         struct btrfs_super_block *disk_super;
 451         struct btrfs_root *tree_root;
 452         struct btrfs_key key;
 453         int err;
 454
 455         sb->s_maxbytes = MAX_LFS_FILESIZE;
 456         sb->s_magic = BTRFS_SUPER_MAGIC;
 457         sb->s_op = &btrfs_super_ops;
 458         sb->s_export_op = &btrfs_export_ops;
 459         sb->s_xattr = btrfs_xattr_handlers;
 460         sb->s_time_gran = 1;
 461 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 462         sb->s_flags |= MS_POSIXACL;
 463 #endif
 464
 465         tree_root = open_ctree(sb, fs_devices, (char *)data);
 466
 467         if (IS_ERR(tree_root)) {
 468                 printk("btrfs: open_ctree failed\n");
 469                 return PTR_ERR(tree_root);
 470         }
 471         sb->s_fs_info = tree_root;
 472         disk_super = &tree_root->fs_info->super_copy;
 473
 474         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 475         key.type = BTRFS_INODE_ITEM_KEY;
 476         key.offset = 0;
 477         inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
 478         if (IS_ERR(inode)) {
 479                 err = PTR_ERR(inode);
 480                 goto fail_close;
 481         }
 482
 483         root_dentry = d_alloc_root(inode);
 484         if (!root_dentry) {
 485                 iput(inode);
 486                 err = -ENOMEM;
 487                 goto fail_close;
 488         }
 489
 490         sb->s_root = root_dentry;
 491
 492         save_mount_options(sb, data);
 493         return 0;
 494
 495 fail_close:
 496         close_ctree(tree_root);
 497         return err;
 498 }
 499
 500 int btrfs_sync_fs(struct super_block *sb, int wait)
 501 {
 502         struct btrfs_trans_handle *trans;
 503         struct btrfs_root *root = btrfs_sb(sb);
 504         int ret;
 505
 506         if (!wait) {
 507                 filemap_flush(root->fs_info->btree_inode->i_mapping);
 508                 return 0;
 509         }
 510
 511         btrfs_start_delalloc_inodes(root, 0);
 512         btrfs_wait_ordered_extents(root, 0, 0);
 513
 514         trans = btrfs_start_transaction(root, 1);
 515         ret = btrfs_commit_transaction(trans, root);
 516         return ret;
 517 }
 518
 519 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 520 {
 521         struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
 522         struct btrfs_fs_info *info = root->fs_info;
 523
 524         if (btrfs_test_opt(root, DEGRADED))
 525                 seq_puts(seq, ",degraded");
 526         if (btrfs_test_opt(root, NODATASUM))
 527                 seq_puts(seq, ",nodatasum");
 528         if (btrfs_test_opt(root, NODATACOW))
 529                 seq_puts(seq, ",nodatacow");
 530         if (btrfs_test_opt(root, NOBARRIER))
 531                 seq_puts(seq, ",nobarrier");
 532         if (info->max_extent != (u64)-1)
 533                 seq_printf(seq, ",max_extent=%llu",
 534                            (unsigned long long)info->max_extent);
 535         if (info->max_inline != 8192 * 1024)
 536                 seq_printf(seq, ",max_inline=%llu",
 537                            (unsigned long long)info->max_inline);
 538         if (info->alloc_start != 0)
 539                 seq_printf(seq, ",alloc_start=%llu",
 540                            (unsigned long long)info->alloc_start);
 541         if (info->thread_pool_size !=  min_t(unsigned long,
 542                                              num_online_cpus() + 2, 8))
 543                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
 544         if (btrfs_test_opt(root, COMPRESS))
 545                 seq_puts(seq, ",compress");
 546         if (btrfs_test_opt(root, NOSSD))
 547                 seq_puts(seq, ",nossd");
 548         if (btrfs_test_opt(root, SSD_SPREAD))
 549                 seq_puts(seq, ",ssd_spread");
 550         else if (btrfs_test_opt(root, SSD))
 551                 seq_puts(seq, ",ssd");
 552         if (btrfs_test_opt(root, NOTREELOG))
 553                 seq_puts(seq, ",notreelog");
 554         if (btrfs_test_opt(root, FLUSHONCOMMIT))
 555                 seq_puts(seq, ",flushoncommit");
 556         if (btrfs_test_opt(root, DISCARD))
 557                 seq_puts(seq, ",discard");
 558         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
 559                 seq_puts(seq, ",noacl");
 560         return 0;
 561 }
 562
 563 static int btrfs_test_super(struct super_block *s, void *data)
 564 {
 565         struct btrfs_fs_devices *test_fs_devices = data;
 566         struct btrfs_root *root = btrfs_sb(s);
 567
 568         return root->fs_info->fs_devices == test_fs_devices;
 569 }
 570
 571 /*
 572  * Find a superblock for the given device / mount point.
 573  *
 574  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
 575  *        for multiple device setup.  Make sure to keep it in sync.
 576  */
 577 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 578                 const char *dev_name, void *data, struct vfsmount *mnt)
 579 {
 580         struct block_device *bdev = NULL;
 581         struct super_block *s;
 582         struct dentry *root;
 583         struct btrfs_fs_devices *fs_devices = NULL;
 584         fmode_t mode = FMODE_READ;
 585         char *subvol_name = NULL;
 586         u64 subvol_objectid = 0;
 587         int error = 0;
 588         int found = 0;
 589
 590         if (!(flags & MS_RDONLY))
 591                 mode |= FMODE_WRITE;
 592
 593         error = btrfs_parse_early_options(data, mode, fs_type,
 594                                           &subvol_name, &subvol_objectid,
 595                                           &fs_devices);
 596         if (error)
 597                 return error;
 598
 599         error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
 600         if (error)
 601                 goto error_free_subvol_name;
 602
 603         error = btrfs_open_devices(fs_devices, mode, fs_type);
 604         if (error)
 605                 goto error_free_subvol_name;
 606
 607         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
 608                 error = -EACCES;
 609                 goto error_close_devices;
 610         }
 611
 612         bdev = fs_devices->latest_bdev;
 613         s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
 614         if (IS_ERR(s))
 615                 goto error_s;
 616
 617         if (s->s_root) {
 618                 if ((flags ^ s->s_flags) & MS_RDONLY) {
 619                         deactivate_locked_super(s);
 620                         error = -EBUSY;
 621                         goto error_close_devices;
 622                 }
 623
 624                 found = 1;
 625                 btrfs_close_devices(fs_devices);
 626         } else {
 627                 char b[BDEVNAME_SIZE];
 628
 629                 s->s_flags = flags;
 630                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 631                 error = btrfs_fill_super(s, fs_devices, data,
 632                                          flags & MS_SILENT ? 1 : 0);
 633                 if (error) {
 634                         deactivate_locked_super(s);
 635                         goto error_free_subvol_name;
 636                 }
 637
 638                 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
 639                 s->s_flags |= MS_ACTIVE;
 640         }
 641
 642         root = get_default_root(s, subvol_objectid);
 643         if (IS_ERR(root)) {
 644                 error = PTR_ERR(root);
 645                 deactivate_locked_super(s);
 646                 goto error;
 647         }
 648         /* if they gave us a subvolume name bind mount into that */
 649         if (strcmp(subvol_name, ".")) {
 650                 struct dentry *new_root;
 651                 mutex_lock(&root->d_inode->i_mutex);
 652                 new_root = lookup_one_len(subvol_name, root,
 653                                       strlen(subvol_name));
 654                 mutex_unlock(&root->d_inode->i_mutex);
 655
 656                 if (IS_ERR(new_root)) {
 657                         deactivate_locked_super(s);
 658                         error = PTR_ERR(new_root);
 659                         dput(root);
 660                         goto error_close_devices;
 661                 }
 662                 if (!new_root->d_inode) {
 663                         dput(root);
 664                         dput(new_root);
 665                         deactivate_locked_super(s);
 666                         error = -ENXIO;
 667                         goto error_close_devices;
 668                 }
 669                 dput(root);
 670                 root = new_root;
 671         }
 672
 673         mnt->mnt_sb = s;
 674         mnt->mnt_root = root;
 675
 676         kfree(subvol_name);
 677         return 0;
 678
 679 error_s:
 680         error = PTR_ERR(s);
 681 error_close_devices:
 682         btrfs_close_devices(fs_devices);
 683 error_free_subvol_name:
 684         kfree(subvol_name);
 685 error:
 686         return error;
 687 }
 688
 689 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
 690 {
 691         struct btrfs_root *root = btrfs_sb(sb);
 692         int ret;
 693
 694         ret = btrfs_parse_options(root, data);
 695         if (ret)
 696                 return -EINVAL;
 697
 698         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
 699                 return 0;
 700
 701         if (*flags & MS_RDONLY) {
 702                 sb->s_flags |= MS_RDONLY;
 703
 704                 ret =  btrfs_commit_super(root);
 705                 WARN_ON(ret);
 706         } else {
 707                 if (root->fs_info->fs_devices->rw_devices == 0)
 708                         return -EACCES;
 709
 710                 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
 711                         return -EINVAL;
 712
 713                 /* recover relocation */
 714                 ret = btrfs_recover_relocation(root);
 715                 WARN_ON(ret);
 716
 717                 ret = btrfs_cleanup_fs_roots(root->fs_info);
 718                 WARN_ON(ret);
 719
 720                 sb->s_flags &= ~MS_RDONLY;
 721         }
 722
 723         return 0;
 724 }
 725
 726 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 727 {
 728         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
 729         struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
 730         struct list_head *head = &root->fs_info->space_info;
 731         struct btrfs_space_info *found;
 732         u64 total_used = 0;
 733         u64 data_used = 0;
 734         int bits = dentry->d_sb->s_blocksize_bits;
 735         __be32 *fsid = (__be32 *)root->fs_info->fsid;
 736
 737         rcu_read_lock();
 738         list_for_each_entry_rcu(found, head, list) {
 739                 if (found->flags & (BTRFS_BLOCK_GROUP_DUP|
 740                                     BTRFS_BLOCK_GROUP_RAID10|
 741                                     BTRFS_BLOCK_GROUP_RAID1)) {
 742                         total_used += found->bytes_used;
 743                         if (found->flags & BTRFS_BLOCK_GROUP_DATA)
 744                                 data_used += found->bytes_used;
 745                         else
 746                                 data_used += found->total_bytes;
 747                 }
 748
 749                 total_used += found->bytes_used;
 750                 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
 751                         data_used += found->bytes_used;
 752                 else
 753                         data_used += found->total_bytes;
 754         }
 755         rcu_read_unlock();
 756
 757         buf->f_namelen = BTRFS_NAME_LEN;
 758         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
 759         buf->f_bfree = buf->f_blocks - (total_used >> bits);
 760         buf->f_bavail = buf->f_blocks - (data_used >> bits);
 761         buf->f_bsize = dentry->d_sb->s_blocksize;
 762         buf->f_type = BTRFS_SUPER_MAGIC;
 763
 764         /* We treat it as constant endianness (it doesn't matter _which_)
 765            because we want the fsid to come out the same whether mounted
 766            on a big-endian or little-endian host */
 767         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
 768         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
 769         /* Mask in the root object ID too, to disambiguate subvols */
 770         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
 771         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
 772
 773         return 0;
 774 }
 775
 776 static struct file_system_type btrfs_fs_type = {
 777         .owner          = THIS_MODULE,
 778         .name           = "btrfs",
 779         .get_sb         = btrfs_get_sb,
 780         .kill_sb        = kill_anon_super,
 781         .fs_flags       = FS_REQUIRES_DEV,
 782 };
 783
 784 /*
 785  * used by btrfsctl to scan devices when no FS is mounted
 786  */
 787 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
 788                                 unsigned long arg)
 789 {
 790         struct btrfs_ioctl_vol_args *vol;
 791         struct btrfs_fs_devices *fs_devices;
 792         int ret = -ENOTTY;
 793
 794         if (!capable(CAP_SYS_ADMIN))
 795                 return -EPERM;
 796
 797         vol = memdup_user((void __user *)arg, sizeof(*vol));
 798         if (IS_ERR(vol))
 799                 return PTR_ERR(vol);
 800
 801         switch (cmd) {
 802         case BTRFS_IOC_SCAN_DEV:
 803                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
 804                                             &btrfs_fs_type, &fs_devices);
 805                 break;
 806         }
 807
 808         kfree(vol);
 809         return ret;
 810 }
 811
 812 static int btrfs_freeze(struct super_block *sb)
 813 {
 814         struct btrfs_root *root = btrfs_sb(sb);
 815         mutex_lock(&root->fs_info->transaction_kthread_mutex);
 816         mutex_lock(&root->fs_info->cleaner_mutex);
 817         return 0;
 818 }
 819
 820 static int btrfs_unfreeze(struct super_block *sb)
 821 {
 822         struct btrfs_root *root = btrfs_sb(sb);
 823         mutex_unlock(&root->fs_info->cleaner_mutex);
 824         mutex_unlock(&root->fs_info->transaction_kthread_mutex);
 825         return 0;
 826 }
 827
 828 static const struct super_operations btrfs_super_ops = {
 829         .drop_inode     = btrfs_drop_inode,
 830         .delete_inode   = btrfs_delete_inode,
 831         .put_super      = btrfs_put_super,
 832         .sync_fs        = btrfs_sync_fs,
 833         .show_options   = btrfs_show_options,
 834         .write_inode    = btrfs_write_inode,
 835         .dirty_inode    = btrfs_dirty_inode,
 836         .alloc_inode    = btrfs_alloc_inode,
 837         .destroy_inode  = btrfs_destroy_inode,
 838         .statfs         = btrfs_statfs,
 839         .remount_fs     = btrfs_remount,
 840         .freeze_fs      = btrfs_freeze,
 841         .unfreeze_fs    = btrfs_unfreeze,
 842 };
 843
 844 static const struct file_operations btrfs_ctl_fops = {
 845         .unlocked_ioctl  = btrfs_control_ioctl,
 846         .compat_ioctl = btrfs_control_ioctl,
 847         .owner   = THIS_MODULE,
 848 };
 849
 850 static struct miscdevice btrfs_misc = {
 851         .minor          = MISC_DYNAMIC_MINOR,
 852         .name           = "btrfs-control",
 853         .fops           = &btrfs_ctl_fops
 854 };
 855
 856 static int btrfs_interface_init(void)
 857 {
 858         return misc_register(&btrfs_misc);
 859 }
 860
 861 static void btrfs_interface_exit(void)
 862 {
 863         if (misc_deregister(&btrfs_misc) < 0)
 864                 printk(KERN_INFO "misc_deregister failed for control device");
 865 }
 866
 867 static int __init init_btrfs_fs(void)
 868 {
 869         int err;
 870
 871         err = btrfs_init_sysfs();
 872         if (err)
 873                 return err;
 874
 875         err = btrfs_init_cachep();
 876         if (err)
 877                 goto free_sysfs;
 878
 879         err = extent_io_init();
 880         if (err)
 881                 goto free_cachep;
 882
 883         err = extent_map_init();
 884         if (err)
 885                 goto free_extent_io;
 886
 887         err = btrfs_interface_init();
 888         if (err)
 889                 goto free_extent_map;
 890
 891         err = register_filesystem(&btrfs_fs_type);
 892         if (err)
 893                 goto unregister_ioctl;
 894
 895         printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
 896         return 0;
 897
 898 unregister_ioctl:
 899         btrfs_interface_exit();
 900 free_extent_map:
 901         extent_map_exit();
 902 free_extent_io:
 903         extent_io_exit();
 904 free_cachep:
 905         btrfs_destroy_cachep();
 906 free_sysfs:
 907         btrfs_exit_sysfs();
 908         return err;
 909 }
 910
 911 static void __exit exit_btrfs_fs(void)
 912 {
 913         btrfs_destroy_cachep();
 914         extent_map_exit();
 915         extent_io_exit();
 916         btrfs_interface_exit();
 917         unregister_filesystem(&btrfs_fs_type);
 918         btrfs_exit_sysfs();
 919         btrfs_cleanup_fs_uuids();
 920         btrfs_zlib_exit();
 921 }
 922
 923 module_init(init_btrfs_fs)
 924 module_exit(exit_btrfs_fs)
 925
 926 MODULE_LICENSE("GPL");