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         (void)ret; /* FIXME: need to fix VFS to return error? */
  66 }
  67
  68 enum {
  69         Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
  70         Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
  71         Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
  72         Opt_compress_force, Opt_notreelog, Opt_ratio, Opt_flushoncommit,
  73         Opt_discard, Opt_space_cache, Opt_clear_cache, Opt_err,
  74         Opt_user_subvol_rm_allowed,
  75 };
  76
  77 static match_table_t tokens = {
  78         {Opt_degraded, "degraded"},
  79         {Opt_subvol, "subvol=%s"},
  80         {Opt_subvolid, "subvolid=%d"},
  81         {Opt_device, "device=%s"},
  82         {Opt_nodatasum, "nodatasum"},
  83         {Opt_nodatacow, "nodatacow"},
  84         {Opt_nobarrier, "nobarrier"},
  85         {Opt_max_inline, "max_inline=%s"},
  86         {Opt_alloc_start, "alloc_start=%s"},
  87         {Opt_thread_pool, "thread_pool=%d"},
  88         {Opt_compress, "compress"},
  89         {Opt_compress_force, "compress-force"},
  90         {Opt_ssd, "ssd"},
  91         {Opt_ssd_spread, "ssd_spread"},
  92         {Opt_nossd, "nossd"},
  93         {Opt_noacl, "noacl"},
  94         {Opt_notreelog, "notreelog"},
  95         {Opt_flushoncommit, "flushoncommit"},
  96         {Opt_ratio, "metadata_ratio=%d"},
  97         {Opt_discard, "discard"},
  98         {Opt_space_cache, "space_cache"},
  99         {Opt_clear_cache, "clear_cache"},
 100         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 101         {Opt_err, NULL},
 102 };
 103
 104 /*
 105  * Regular mount options parser.  Everything that is needed only when
 106  * reading in a new superblock is parsed here.
 107  */
 108 int btrfs_parse_options(struct btrfs_root *root, char *options)
 109 {
 110         struct btrfs_fs_info *info = root->fs_info;
 111         substring_t args[MAX_OPT_ARGS];
 112         char *p, *num, *orig;
 113         int intarg;
 114         int ret = 0;
 115
 116         if (!options)
 117                 return 0;
 118
 119         /*
 120          * strsep changes the string, duplicate it because parse_options
 121          * gets called twice
 122          */
 123         options = kstrdup(options, GFP_NOFS);
 124         if (!options)
 125                 return -ENOMEM;
 126
 127         orig = options;
 128
 129         while ((p = strsep(&options, ",")) != NULL) {
 130                 int token;
 131                 if (!*p)
 132                         continue;
 133
 134                 token = match_token(p, tokens, args);
 135                 switch (token) {
 136                 case Opt_degraded:
 137                         printk(KERN_INFO "btrfs: allowing degraded mounts\n");
 138                         btrfs_set_opt(info->mount_opt, DEGRADED);
 139                         break;
 140                 case Opt_subvol:
 141                 case Opt_subvolid:
 142                 case Opt_device:
 143                         /*
 144                          * These are parsed by btrfs_parse_early_options
 145                          * and can be happily ignored here.
 146                          */
 147                         break;
 148                 case Opt_nodatasum:
 149                         printk(KERN_INFO "btrfs: setting nodatasum\n");
 150                         btrfs_set_opt(info->mount_opt, NODATASUM);
 151                         break;
 152                 case Opt_nodatacow:
 153                         printk(KERN_INFO "btrfs: setting nodatacow\n");
 154                         btrfs_set_opt(info->mount_opt, NODATACOW);
 155                         btrfs_set_opt(info->mount_opt, NODATASUM);
 156                         break;
 157                 case Opt_compress:
 158                         printk(KERN_INFO "btrfs: use compression\n");
 159                         btrfs_set_opt(info->mount_opt, COMPRESS);
 160                         break;
 161                 case Opt_compress_force:
 162                         printk(KERN_INFO "btrfs: forcing compression\n");
 163                         btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 164                         btrfs_set_opt(info->mount_opt, COMPRESS);
 165                         break;
 166                 case Opt_ssd:
 167                         printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 168                         btrfs_set_opt(info->mount_opt, SSD);
 169                         break;
 170                 case Opt_ssd_spread:
 171                         printk(KERN_INFO "btrfs: use spread ssd "
 172                                "allocation scheme\n");
 173                         btrfs_set_opt(info->mount_opt, SSD);
 174                         btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 175                         break;
 176                 case Opt_nossd:
 177                         printk(KERN_INFO "btrfs: not using ssd allocation "
 178                                "scheme\n");
 179                         btrfs_set_opt(info->mount_opt, NOSSD);
 180                         btrfs_clear_opt(info->mount_opt, SSD);
 181                         btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 182                         break;
 183                 case Opt_nobarrier:
 184                         printk(KERN_INFO "btrfs: turning off barriers\n");
 185                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 186                         break;
 187                 case Opt_thread_pool:
 188                         intarg = 0;
 189                         match_int(&args[0], &intarg);
 190                         if (intarg) {
 191                                 info->thread_pool_size = intarg;
 192                                 printk(KERN_INFO "btrfs: thread pool %d\n",
 193                                        info->thread_pool_size);
 194                         }
 195                         break;
 196                 case Opt_max_inline:
 197                         num = match_strdup(&args[0]);
 198                         if (num) {
 199                                 info->max_inline = memparse(num, NULL);
 200                                 kfree(num);
 201
 202                                 if (info->max_inline) {
 203                                         info->max_inline = max_t(u64,
 204                                                 info->max_inline,
 205                                                 root->sectorsize);
 206                                 }
 207                                 printk(KERN_INFO "btrfs: max_inline at %llu\n",
 208                                         (unsigned long long)info->max_inline);
 209                         }
 210                         break;
 211                 case Opt_alloc_start:
 212                         num = match_strdup(&args[0]);
 213                         if (num) {
 214                                 info->alloc_start = memparse(num, NULL);
 215                                 kfree(num);
 216                                 printk(KERN_INFO
 217                                         "btrfs: allocations start at %llu\n",
 218                                         (unsigned long long)info->alloc_start);
 219                         }
 220                         break;
 221                 case Opt_noacl:
 222                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 223                         break;
 224                 case Opt_notreelog:
 225                         printk(KERN_INFO "btrfs: disabling tree log\n");
 226                         btrfs_set_opt(info->mount_opt, NOTREELOG);
 227                         break;
 228                 case Opt_flushoncommit:
 229                         printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
 230                         btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
 231                         break;
 232                 case Opt_ratio:
 233                         intarg = 0;
 234                         match_int(&args[0], &intarg);
 235                         if (intarg) {
 236                                 info->metadata_ratio = intarg;
 237                                 printk(KERN_INFO "btrfs: metadata ratio %d\n",
 238                                        info->metadata_ratio);
 239                         }
 240                         break;
 241                 case Opt_discard:
 242                         btrfs_set_opt(info->mount_opt, DISCARD);
 243                         break;
 244                 case Opt_space_cache:
 245                         printk(KERN_INFO "btrfs: enabling disk space caching\n");
 246                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 247                         break;
 248                 case Opt_clear_cache:
 249                         printk(KERN_INFO "btrfs: force clearing of disk cache\n");
 250                         btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
 251                         break;
 252                 case Opt_user_subvol_rm_allowed:
 253                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 254                         break;
 255                 case Opt_err:
 256                         printk(KERN_INFO "btrfs: unrecognized mount option "
 257                                "'%s'\n", p);
 258                         ret = -EINVAL;
 259                         goto out;
 260                 default:
 261                         break;
 262                 }
 263         }
 264 out:
 265         kfree(orig);
 266         return ret;
 267 }
 268
 269 /*
 270  * Parse mount options that are required early in the mount process.
 271  *
 272  * All other options will be parsed on much later in the mount process and
 273  * only when we need to allocate a new super block.
 274  */
 275 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 276                 void *holder, char **subvol_name, u64 *subvol_objectid,
 277                 struct btrfs_fs_devices **fs_devices)
 278 {
 279         substring_t args[MAX_OPT_ARGS];
 280         char *opts, *p;
 281         int error = 0;
 282         int intarg;
 283
 284         if (!options)
 285                 goto out;
 286
 287         /*
 288          * strsep changes the string, duplicate it because parse_options
 289          * gets called twice
 290          */
 291         opts = kstrdup(options, GFP_KERNEL);
 292         if (!opts)
 293                 return -ENOMEM;
 294
 295         while ((p = strsep(&opts, ",")) != NULL) {
 296                 int token;
 297                 if (!*p)
 298                         continue;
 299
 300                 token = match_token(p, tokens, args);
 301                 switch (token) {
 302                 case Opt_subvol:
 303                         *subvol_name = match_strdup(&args[0]);
 304                         break;
 305                 case Opt_subvolid:
 306                         intarg = 0;
 307                         error = match_int(&args[0], &intarg);
 308                         if (!error) {
 309                                 /* we want the original fs_tree */
 310                                 if (!intarg)
 311                                         *subvol_objectid =
 312                                                 BTRFS_FS_TREE_OBJECTID;
 313                                 else
 314                                         *subvol_objectid = intarg;
 315                         }
 316                         break;
 317                 case Opt_device:
 318                         error = btrfs_scan_one_device(match_strdup(&args[0]),
 319                                         flags, holder, fs_devices);
 320                         if (error)
 321                                 goto out_free_opts;
 322                         break;
 323                 default:
 324                         break;
 325                 }
 326         }
 327
 328  out_free_opts:
 329         kfree(opts);
 330  out:
 331         /*
 332          * If no subvolume name is specified we use the default one.  Allocate
 333          * a copy of the string "." here so that code later in the
 334          * mount path doesn't care if it's the default volume or another one.
 335          */
 336         if (!*subvol_name) {
 337                 *subvol_name = kstrdup(".", GFP_KERNEL);
 338                 if (!*subvol_name)
 339                         return -ENOMEM;
 340         }
 341         return error;
 342 }
 343
 344 static struct dentry *get_default_root(struct super_block *sb,
 345                                        u64 subvol_objectid)
 346 {
 347         struct btrfs_root *root = sb->s_fs_info;
 348         struct btrfs_root *new_root;
 349         struct btrfs_dir_item *di;
 350         struct btrfs_path *path;
 351         struct btrfs_key location;
 352         struct inode *inode;
 353         struct dentry *dentry;
 354         u64 dir_id;
 355         int new = 0;
 356
 357         /*
 358          * We have a specific subvol we want to mount, just setup location and
 359          * go look up the root.
 360          */
 361         if (subvol_objectid) {
 362                 location.objectid = subvol_objectid;
 363                 location.type = BTRFS_ROOT_ITEM_KEY;
 364                 location.offset = (u64)-1;
 365                 goto find_root;
 366         }
 367
 368         path = btrfs_alloc_path();
 369         if (!path)
 370                 return ERR_PTR(-ENOMEM);
 371         path->leave_spinning = 1;
 372
 373         /*
 374          * Find the "default" dir item which points to the root item that we
 375          * will mount by default if we haven't been given a specific subvolume
 376          * to mount.
 377          */
 378         dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
 379         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 380         if (IS_ERR(di))
 381                 return ERR_CAST(di);
 382         if (!di) {
 383                 /*
 384                  * Ok the default dir item isn't there.  This is weird since
 385                  * it's always been there, but don't freak out, just try and
 386                  * mount to root most subvolume.
 387                  */
 388                 btrfs_free_path(path);
 389                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 390                 new_root = root->fs_info->fs_root;
 391                 goto setup_root;
 392         }
 393
 394         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 395         btrfs_free_path(path);
 396
 397 find_root:
 398         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
 399         if (IS_ERR(new_root))
 400                 return ERR_CAST(new_root);
 401
 402         if (btrfs_root_refs(&new_root->root_item) == 0)
 403                 return ERR_PTR(-ENOENT);
 404
 405         dir_id = btrfs_root_dirid(&new_root->root_item);
 406 setup_root:
 407         location.objectid = dir_id;
 408         location.type = BTRFS_INODE_ITEM_KEY;
 409         location.offset = 0;
 410
 411         inode = btrfs_iget(sb, &location, new_root, &new);
 412         if (IS_ERR(inode))
 413                 return ERR_CAST(inode);
 414
 415         /*
 416          * If we're just mounting the root most subvol put the inode and return
 417          * a reference to the dentry.  We will have already gotten a reference
 418          * to the inode in btrfs_fill_super so we're good to go.
 419          */
 420         if (!new && sb->s_root->d_inode == inode) {
 421                 iput(inode);
 422                 return dget(sb->s_root);
 423         }
 424
 425         if (new) {
 426                 const struct qstr name = { .name = "/", .len = 1 };
 427
 428                 /*
 429                  * New inode, we need to make the dentry a sibling of s_root so
 430                  * everything gets cleaned up properly on unmount.
 431                  */
 432                 dentry = d_alloc(sb->s_root, &name);
 433                 if (!dentry) {
 434                         iput(inode);
 435                         return ERR_PTR(-ENOMEM);
 436                 }
 437                 d_splice_alias(inode, dentry);
 438         } else {
 439                 /*
 440                  * We found the inode in cache, just find a dentry for it and
 441                  * put the reference to the inode we just got.
 442                  */
 443                 dentry = d_find_alias(inode);
 444                 iput(inode);
 445         }
 446
 447         return dentry;
 448 }
 449
 450 static int btrfs_fill_super(struct super_block *sb,
 451                             struct btrfs_fs_devices *fs_devices,
 452                             void *data, int silent)
 453 {
 454         struct inode *inode;
 455         struct dentry *root_dentry;
 456         struct btrfs_root *tree_root;
 457         struct btrfs_key key;
 458         int err;
 459
 460         sb->s_maxbytes = MAX_LFS_FILESIZE;
 461         sb->s_magic = BTRFS_SUPER_MAGIC;
 462         sb->s_op = &btrfs_super_ops;
 463         sb->s_d_op = &btrfs_dentry_operations;
 464         sb->s_export_op = &btrfs_export_ops;
 465         sb->s_xattr = btrfs_xattr_handlers;
 466         sb->s_time_gran = 1;
 467 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 468         sb->s_flags |= MS_POSIXACL;
 469 #endif
 470
 471         tree_root = open_ctree(sb, fs_devices, (char *)data);
 472
 473         if (IS_ERR(tree_root)) {
 474                 printk("btrfs: open_ctree failed\n");
 475                 return PTR_ERR(tree_root);
 476         }
 477         sb->s_fs_info = tree_root;
 478
 479         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 480         key.type = BTRFS_INODE_ITEM_KEY;
 481         key.offset = 0;
 482         inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
 483         if (IS_ERR(inode)) {
 484                 err = PTR_ERR(inode);
 485                 goto fail_close;
 486         }
 487
 488         root_dentry = d_alloc_root(inode);
 489         if (!root_dentry) {
 490                 iput(inode);
 491                 err = -ENOMEM;
 492                 goto fail_close;
 493         }
 494
 495         sb->s_root = root_dentry;
 496
 497         save_mount_options(sb, data);
 498         return 0;
 499
 500 fail_close:
 501         close_ctree(tree_root);
 502         return err;
 503 }
 504
 505 int btrfs_sync_fs(struct super_block *sb, int wait)
 506 {
 507         struct btrfs_trans_handle *trans;
 508         struct btrfs_root *root = btrfs_sb(sb);
 509         int ret;
 510
 511         if (!wait) {
 512                 filemap_flush(root->fs_info->btree_inode->i_mapping);
 513                 return 0;
 514         }
 515
 516         btrfs_start_delalloc_inodes(root, 0);
 517         btrfs_wait_ordered_extents(root, 0, 0);
 518
 519         trans = btrfs_start_transaction(root, 0);
 520         ret = btrfs_commit_transaction(trans, root);
 521         return ret;
 522 }
 523
 524 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 525 {
 526         struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
 527         struct btrfs_fs_info *info = root->fs_info;
 528
 529         if (btrfs_test_opt(root, DEGRADED))
 530                 seq_puts(seq, ",degraded");
 531         if (btrfs_test_opt(root, NODATASUM))
 532                 seq_puts(seq, ",nodatasum");
 533         if (btrfs_test_opt(root, NODATACOW))
 534                 seq_puts(seq, ",nodatacow");
 535         if (btrfs_test_opt(root, NOBARRIER))
 536                 seq_puts(seq, ",nobarrier");
 537         if (info->max_inline != 8192 * 1024)
 538                 seq_printf(seq, ",max_inline=%llu",
 539                            (unsigned long long)info->max_inline);
 540         if (info->alloc_start != 0)
 541                 seq_printf(seq, ",alloc_start=%llu",
 542                            (unsigned long long)info->alloc_start);
 543         if (info->thread_pool_size !=  min_t(unsigned long,
 544                                              num_online_cpus() + 2, 8))
 545                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
 546         if (btrfs_test_opt(root, COMPRESS))
 547                 seq_puts(seq, ",compress");
 548         if (btrfs_test_opt(root, NOSSD))
 549                 seq_puts(seq, ",nossd");
 550         if (btrfs_test_opt(root, SSD_SPREAD))
 551                 seq_puts(seq, ",ssd_spread");
 552         else if (btrfs_test_opt(root, SSD))
 553                 seq_puts(seq, ",ssd");
 554         if (btrfs_test_opt(root, NOTREELOG))
 555                 seq_puts(seq, ",notreelog");
 556         if (btrfs_test_opt(root, FLUSHONCOMMIT))
 557                 seq_puts(seq, ",flushoncommit");
 558         if (btrfs_test_opt(root, DISCARD))
 559                 seq_puts(seq, ",discard");
 560         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
 561                 seq_puts(seq, ",noacl");
 562         return 0;
 563 }
 564
 565 static int btrfs_test_super(struct super_block *s, void *data)
 566 {
 567         struct btrfs_root *test_root = data;
 568         struct btrfs_root *root = btrfs_sb(s);
 569
 570         /*
 571          * If this super block is going away, return false as it
 572          * can't match as an existing super block.
 573          */
 574         if (!atomic_read(&s->s_active))
 575                 return 0;
 576         return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
 577 }
 578
 579 static int btrfs_set_super(struct super_block *s, void *data)
 580 {
 581         s->s_fs_info = data;
 582
 583         return set_anon_super(s, data);
 584 }
 585
 586
 587 /*
 588  * Find a superblock for the given device / mount point.
 589  *
 590  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
 591  *        for multiple device setup.  Make sure to keep it in sync.
 592  */
 593 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
 594                 const char *dev_name, void *data)
 595 {
 596         struct block_device *bdev = NULL;
 597         struct super_block *s;
 598         struct dentry *root;
 599         struct btrfs_fs_devices *fs_devices = NULL;
 600         struct btrfs_root *tree_root = NULL;
 601         struct btrfs_fs_info *fs_info = NULL;
 602         fmode_t mode = FMODE_READ;
 603         char *subvol_name = NULL;
 604         u64 subvol_objectid = 0;
 605         int error = 0;
 606
 607         if (!(flags & MS_RDONLY))
 608                 mode |= FMODE_WRITE;
 609
 610         error = btrfs_parse_early_options(data, mode, fs_type,
 611                                           &subvol_name, &subvol_objectid,
 612                                           &fs_devices);
 613         if (error)
 614                 return ERR_PTR(error);
 615
 616         error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
 617         if (error)
 618                 goto error_free_subvol_name;
 619
 620         error = btrfs_open_devices(fs_devices, mode, fs_type);
 621         if (error)
 622                 goto error_free_subvol_name;
 623
 624         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
 625                 error = -EACCES;
 626                 goto error_close_devices;
 627         }
 628
 629         /*
 630          * Setup a dummy root and fs_info for test/set super.  This is because
 631          * we don't actually fill this stuff out until open_ctree, but we need
 632          * it for searching for existing supers, so this lets us do that and
 633          * then open_ctree will properly initialize everything later.
 634          */
 635         fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
 636         tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
 637         if (!fs_info || !tree_root) {
 638                 error = -ENOMEM;
 639                 goto error_close_devices;
 640         }
 641         fs_info->tree_root = tree_root;
 642         fs_info->fs_devices = fs_devices;
 643         tree_root->fs_info = fs_info;
 644
 645         bdev = fs_devices->latest_bdev;
 646         s = sget(fs_type, btrfs_test_super, btrfs_set_super, tree_root);
 647         if (IS_ERR(s))
 648                 goto error_s;
 649
 650         if (s->s_root) {
 651                 if ((flags ^ s->s_flags) & MS_RDONLY) {
 652                         deactivate_locked_super(s);
 653                         error = -EBUSY;
 654                         goto error_close_devices;
 655                 }
 656
 657                 btrfs_close_devices(fs_devices);
 658         } else {
 659                 char b[BDEVNAME_SIZE];
 660
 661                 s->s_flags = flags;
 662                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 663                 error = btrfs_fill_super(s, fs_devices, data,
 664                                          flags & MS_SILENT ? 1 : 0);
 665                 if (error) {
 666                         deactivate_locked_super(s);
 667                         goto error_free_subvol_name;
 668                 }
 669
 670                 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
 671                 s->s_flags |= MS_ACTIVE;
 672         }
 673
 674         root = get_default_root(s, subvol_objectid);
 675         if (IS_ERR(root)) {
 676                 error = PTR_ERR(root);
 677                 deactivate_locked_super(s);
 678                 goto error_free_subvol_name;
 679         }
 680         /* if they gave us a subvolume name bind mount into that */
 681         if (strcmp(subvol_name, ".")) {
 682                 struct dentry *new_root;
 683                 mutex_lock(&root->d_inode->i_mutex);
 684                 new_root = lookup_one_len(subvol_name, root,
 685                                       strlen(subvol_name));
 686                 mutex_unlock(&root->d_inode->i_mutex);
 687
 688                 if (IS_ERR(new_root)) {
 689                         dput(root);
 690                         deactivate_locked_super(s);
 691                         error = PTR_ERR(new_root);
 692                         goto error_free_subvol_name;
 693                 }
 694                 if (!new_root->d_inode) {
 695                         dput(root);
 696                         dput(new_root);
 697                         deactivate_locked_super(s);
 698                         error = -ENXIO;
 699                         goto error_free_subvol_name;
 700                 }
 701                 dput(root);
 702                 root = new_root;
 703         }
 704
 705         kfree(subvol_name);
 706         return root;
 707
 708 error_s:
 709         error = PTR_ERR(s);
 710 error_close_devices:
 711         btrfs_close_devices(fs_devices);
 712         kfree(fs_info);
 713         kfree(tree_root);
 714 error_free_subvol_name:
 715         kfree(subvol_name);
 716         return ERR_PTR(error);
 717 }
 718
 719 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
 720 {
 721         struct btrfs_root *root = btrfs_sb(sb);
 722         int ret;
 723
 724         ret = btrfs_parse_options(root, data);
 725         if (ret)
 726                 return -EINVAL;
 727
 728         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
 729                 return 0;
 730
 731         if (*flags & MS_RDONLY) {
 732                 sb->s_flags |= MS_RDONLY;
 733
 734                 ret =  btrfs_commit_super(root);
 735                 WARN_ON(ret);
 736         } else {
 737                 if (root->fs_info->fs_devices->rw_devices == 0)
 738                         return -EACCES;
 739
 740                 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
 741                         return -EINVAL;
 742
 743                 ret = btrfs_cleanup_fs_roots(root->fs_info);
 744                 WARN_ON(ret);
 745
 746                 /* recover relocation */
 747                 ret = btrfs_recover_relocation(root);
 748                 WARN_ON(ret);
 749
 750                 sb->s_flags &= ~MS_RDONLY;
 751         }
 752
 753         return 0;
 754 }
 755
 756 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 757 {
 758         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
 759         struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
 760         struct list_head *head = &root->fs_info->space_info;
 761         struct btrfs_space_info *found;
 762         u64 total_used = 0;
 763         u64 total_used_data = 0;
 764         int bits = dentry->d_sb->s_blocksize_bits;
 765         __be32 *fsid = (__be32 *)root->fs_info->fsid;
 766
 767         rcu_read_lock();
 768         list_for_each_entry_rcu(found, head, list) {
 769                 if (found->flags & (BTRFS_BLOCK_GROUP_METADATA |
 770                                     BTRFS_BLOCK_GROUP_SYSTEM))
 771                         total_used_data += found->disk_total;
 772                 else
 773                         total_used_data += found->disk_used;
 774                 total_used += found->disk_used;
 775         }
 776         rcu_read_unlock();
 777
 778         buf->f_namelen = BTRFS_NAME_LEN;
 779         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
 780         buf->f_bfree = buf->f_blocks - (total_used >> bits);
 781         buf->f_bavail = buf->f_blocks - (total_used_data >> bits);
 782         buf->f_bsize = dentry->d_sb->s_blocksize;
 783         buf->f_type = BTRFS_SUPER_MAGIC;
 784
 785         /* We treat it as constant endianness (it doesn't matter _which_)
 786            because we want the fsid to come out the same whether mounted
 787            on a big-endian or little-endian host */
 788         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
 789         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
 790         /* Mask in the root object ID too, to disambiguate subvols */
 791         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
 792         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
 793
 794         return 0;
 795 }
 796
 797 static struct file_system_type btrfs_fs_type = {
 798         .owner          = THIS_MODULE,
 799         .name           = "btrfs",
 800         .mount          = btrfs_mount,
 801         .kill_sb        = kill_anon_super,
 802         .fs_flags       = FS_REQUIRES_DEV,
 803 };
 804
 805 /*
 806  * used by btrfsctl to scan devices when no FS is mounted
 807  */
 808 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
 809                                 unsigned long arg)
 810 {
 811         struct btrfs_ioctl_vol_args *vol;
 812         struct btrfs_fs_devices *fs_devices;
 813         int ret = -ENOTTY;
 814
 815         if (!capable(CAP_SYS_ADMIN))
 816                 return -EPERM;
 817
 818         vol = memdup_user((void __user *)arg, sizeof(*vol));
 819         if (IS_ERR(vol))
 820                 return PTR_ERR(vol);
 821
 822         switch (cmd) {
 823         case BTRFS_IOC_SCAN_DEV:
 824                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
 825                                             &btrfs_fs_type, &fs_devices);
 826                 break;
 827         }
 828
 829         kfree(vol);
 830         return ret;
 831 }
 832
 833 static int btrfs_freeze(struct super_block *sb)
 834 {
 835         struct btrfs_root *root = btrfs_sb(sb);
 836         mutex_lock(&root->fs_info->transaction_kthread_mutex);
 837         mutex_lock(&root->fs_info->cleaner_mutex);
 838         return 0;
 839 }
 840
 841 static int btrfs_unfreeze(struct super_block *sb)
 842 {
 843         struct btrfs_root *root = btrfs_sb(sb);
 844         mutex_unlock(&root->fs_info->cleaner_mutex);
 845         mutex_unlock(&root->fs_info->transaction_kthread_mutex);
 846         return 0;
 847 }
 848
 849 static const struct super_operations btrfs_super_ops = {
 850         .drop_inode     = btrfs_drop_inode,
 851         .evict_inode    = btrfs_evict_inode,
 852         .put_super      = btrfs_put_super,
 853         .sync_fs        = btrfs_sync_fs,
 854         .show_options   = btrfs_show_options,
 855         .write_inode    = btrfs_write_inode,
 856         .dirty_inode    = btrfs_dirty_inode,
 857         .alloc_inode    = btrfs_alloc_inode,
 858         .destroy_inode  = btrfs_destroy_inode,
 859         .statfs         = btrfs_statfs,
 860         .remount_fs     = btrfs_remount,
 861         .freeze_fs      = btrfs_freeze,
 862         .unfreeze_fs    = btrfs_unfreeze,
 863 };
 864
 865 static const struct file_operations btrfs_ctl_fops = {
 866         .unlocked_ioctl  = btrfs_control_ioctl,
 867         .compat_ioctl = btrfs_control_ioctl,
 868         .owner   = THIS_MODULE,
 869         .llseek = noop_llseek,
 870 };
 871
 872 static struct miscdevice btrfs_misc = {
 873         .minor          = BTRFS_MINOR,
 874         .name           = "btrfs-control",
 875         .fops           = &btrfs_ctl_fops
 876 };
 877
 878 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
 879 MODULE_ALIAS("devname:btrfs-control");
 880
 881 static int btrfs_interface_init(void)
 882 {
 883         return misc_register(&btrfs_misc);
 884 }
 885
 886 static void btrfs_interface_exit(void)
 887 {
 888         if (misc_deregister(&btrfs_misc) < 0)
 889                 printk(KERN_INFO "misc_deregister failed for control device");
 890 }
 891
 892 static int __init init_btrfs_fs(void)
 893 {
 894         int err;
 895
 896         err = btrfs_init_sysfs();
 897         if (err)
 898                 return err;
 899
 900         err = btrfs_init_cachep();
 901         if (err)
 902                 goto free_sysfs;
 903
 904         err = extent_io_init();
 905         if (err)
 906                 goto free_cachep;
 907
 908         err = extent_map_init();
 909         if (err)
 910                 goto free_extent_io;
 911
 912         err = btrfs_interface_init();
 913         if (err)
 914                 goto free_extent_map;
 915
 916         err = register_filesystem(&btrfs_fs_type);
 917         if (err)
 918                 goto unregister_ioctl;
 919
 920         printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
 921         return 0;
 922
 923 unregister_ioctl:
 924         btrfs_interface_exit();
 925 free_extent_map:
 926         extent_map_exit();
 927 free_extent_io:
 928         extent_io_exit();
 929 free_cachep:
 930         btrfs_destroy_cachep();
 931 free_sysfs:
 932         btrfs_exit_sysfs();
 933         return err;
 934 }
 935
 936 static void __exit exit_btrfs_fs(void)
 937 {
 938         btrfs_destroy_cachep();
 939         extent_map_exit();
 940         extent_io_exit();
 941         btrfs_interface_exit();
 942         unregister_filesystem(&btrfs_fs_type);
 943         btrfs_exit_sysfs();
 944         btrfs_cleanup_fs_uuids();
 945         btrfs_zlib_exit();
 946 }
 947
 948 module_init(init_btrfs_fs)
 949 module_exit(exit_btrfs_fs)
 950
 951 MODULE_LICENSE("GPL");