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_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
  68         Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
  69         Opt_compress_force, Opt_notreelog, Opt_ratio, Opt_flushoncommit,
  70         Opt_discard, Opt_err,
  71 };
  72
  73 static match_table_t tokens = {
  74         {Opt_degraded, "degraded"},
  75         {Opt_subvol, "subvol=%s"},
  76         {Opt_subvolid, "subvolid=%d"},
  77         {Opt_device, "device=%s"},
  78         {Opt_nodatasum, "nodatasum"},
  79         {Opt_nodatacow, "nodatacow"},
  80         {Opt_nobarrier, "nobarrier"},
  81         {Opt_max_inline, "max_inline=%s"},
  82         {Opt_alloc_start, "alloc_start=%s"},
  83         {Opt_thread_pool, "thread_pool=%d"},
  84         {Opt_compress, "compress"},
  85         {Opt_compress_force, "compress-force"},
  86         {Opt_ssd, "ssd"},
  87         {Opt_ssd_spread, "ssd_spread"},
  88         {Opt_nossd, "nossd"},
  89         {Opt_noacl, "noacl"},
  90         {Opt_notreelog, "notreelog"},
  91         {Opt_flushoncommit, "flushoncommit"},
  92         {Opt_ratio, "metadata_ratio=%d"},
  93         {Opt_discard, "discard"},
  94         {Opt_err, NULL},
  95 };
  96
  97 /*
  98  * Regular mount options parser.  Everything that is needed only when
  99  * reading in a new superblock is parsed here.
 100  */
 101 int btrfs_parse_options(struct btrfs_root *root, char *options)
 102 {
 103         struct btrfs_fs_info *info = root->fs_info;
 104         substring_t args[MAX_OPT_ARGS];
 105         char *p, *num, *orig;
 106         int intarg;
 107         int ret = 0;
 108
 109         if (!options)
 110                 return 0;
 111
 112         /*
 113          * strsep changes the string, duplicate it because parse_options
 114          * gets called twice
 115          */
 116         options = kstrdup(options, GFP_NOFS);
 117         if (!options)
 118                 return -ENOMEM;
 119
 120         orig = options;
 121
 122         while ((p = strsep(&options, ",")) != NULL) {
 123                 int token;
 124                 if (!*p)
 125                         continue;
 126
 127                 token = match_token(p, tokens, args);
 128                 switch (token) {
 129                 case Opt_degraded:
 130                         printk(KERN_INFO "btrfs: allowing degraded mounts\n");
 131                         btrfs_set_opt(info->mount_opt, DEGRADED);
 132                         break;
 133                 case Opt_subvol:
 134                 case Opt_subvolid:
 135                 case Opt_device:
 136                         /*
 137                          * These are parsed by btrfs_parse_early_options
 138                          * and can be happily ignored here.
 139                          */
 140                         break;
 141                 case Opt_nodatasum:
 142                         printk(KERN_INFO "btrfs: setting nodatasum\n");
 143                         btrfs_set_opt(info->mount_opt, NODATASUM);
 144                         break;
 145                 case Opt_nodatacow:
 146                         printk(KERN_INFO "btrfs: setting nodatacow\n");
 147                         btrfs_set_opt(info->mount_opt, NODATACOW);
 148                         btrfs_set_opt(info->mount_opt, NODATASUM);
 149                         break;
 150                 case Opt_compress:
 151                         printk(KERN_INFO "btrfs: use compression\n");
 152                         btrfs_set_opt(info->mount_opt, COMPRESS);
 153                         break;
 154                 case Opt_compress_force:
 155                         printk(KERN_INFO "btrfs: forcing compression\n");
 156                         btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 157                         btrfs_set_opt(info->mount_opt, COMPRESS);
 158                         break;
 159                 case Opt_ssd:
 160                         printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 161                         btrfs_set_opt(info->mount_opt, SSD);
 162                         break;
 163                 case Opt_ssd_spread:
 164                         printk(KERN_INFO "btrfs: use spread ssd "
 165                                "allocation scheme\n");
 166                         btrfs_set_opt(info->mount_opt, SSD);
 167                         btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 168                         break;
 169                 case Opt_nossd:
 170                         printk(KERN_INFO "btrfs: not using ssd allocation "
 171                                "scheme\n");
 172                         btrfs_set_opt(info->mount_opt, NOSSD);
 173                         btrfs_clear_opt(info->mount_opt, SSD);
 174                         btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 175                         break;
 176                 case Opt_nobarrier:
 177                         printk(KERN_INFO "btrfs: turning off barriers\n");
 178                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 179                         break;
 180                 case Opt_thread_pool:
 181                         intarg = 0;
 182                         match_int(&args[0], &intarg);
 183                         if (intarg) {
 184                                 info->thread_pool_size = intarg;
 185                                 printk(KERN_INFO "btrfs: thread pool %d\n",
 186                                        info->thread_pool_size);
 187                         }
 188                         break;
 189                 case Opt_max_inline:
 190                         num = match_strdup(&args[0]);
 191                         if (num) {
 192                                 info->max_inline = memparse(num, NULL);
 193                                 kfree(num);
 194
 195                                 if (info->max_inline) {
 196                                         info->max_inline = max_t(u64,
 197                                                 info->max_inline,
 198                                                 root->sectorsize);
 199                                 }
 200                                 printk(KERN_INFO "btrfs: max_inline at %llu\n",
 201                                         (unsigned long long)info->max_inline);
 202                         }
 203                         break;
 204                 case Opt_alloc_start:
 205                         num = match_strdup(&args[0]);
 206                         if (num) {
 207                                 info->alloc_start = memparse(num, NULL);
 208                                 kfree(num);
 209                                 printk(KERN_INFO
 210                                         "btrfs: allocations start at %llu\n",
 211                                         (unsigned long long)info->alloc_start);
 212                         }
 213                         break;
 214                 case Opt_noacl:
 215                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 216                         break;
 217                 case Opt_notreelog:
 218                         printk(KERN_INFO "btrfs: disabling tree log\n");
 219                         btrfs_set_opt(info->mount_opt, NOTREELOG);
 220                         break;
 221                 case Opt_flushoncommit:
 222                         printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
 223                         btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
 224                         break;
 225                 case Opt_ratio:
 226                         intarg = 0;
 227                         match_int(&args[0], &intarg);
 228                         if (intarg) {
 229                                 info->metadata_ratio = intarg;
 230                                 printk(KERN_INFO "btrfs: metadata ratio %d\n",
 231                                        info->metadata_ratio);
 232                         }
 233                         break;
 234                 case Opt_discard:
 235                         btrfs_set_opt(info->mount_opt, DISCARD);
 236                         break;
 237                 case Opt_err:
 238                         printk(KERN_INFO "btrfs: unrecognized mount option "
 239                                "'%s'\n", p);
 240                         ret = -EINVAL;
 241                         goto out;
 242                 default:
 243                         break;
 244                 }
 245         }
 246 out:
 247         kfree(orig);
 248         return ret;
 249 }
 250
 251 /*
 252  * Parse mount options that are required early in the mount process.
 253  *
 254  * All other options will be parsed on much later in the mount process and
 255  * only when we need to allocate a new super block.
 256  */
 257 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 258                 void *holder, char **subvol_name, u64 *subvol_objectid,
 259                 struct btrfs_fs_devices **fs_devices)
 260 {
 261         substring_t args[MAX_OPT_ARGS];
 262         char *opts, *p;
 263         int error = 0;
 264         int intarg;
 265
 266         if (!options)
 267                 goto out;
 268
 269         /*
 270          * strsep changes the string, duplicate it because parse_options
 271          * gets called twice
 272          */
 273         opts = kstrdup(options, GFP_KERNEL);
 274         if (!opts)
 275                 return -ENOMEM;
 276
 277         while ((p = strsep(&opts, ",")) != NULL) {
 278                 int token;
 279                 if (!*p)
 280                         continue;
 281
 282                 token = match_token(p, tokens, args);
 283                 switch (token) {
 284                 case Opt_subvol:
 285                         *subvol_name = match_strdup(&args[0]);
 286                         break;
 287                 case Opt_subvolid:
 288                         intarg = 0;
 289                         error = match_int(&args[0], &intarg);
 290                         if (!error) {
 291                                 /* we want the original fs_tree */
 292                                 if (!intarg)
 293                                         *subvol_objectid =
 294                                                 BTRFS_FS_TREE_OBJECTID;
 295                                 else
 296                                         *subvol_objectid = intarg;
 297                         }
 298                         break;
 299                 case Opt_device:
 300                         error = btrfs_scan_one_device(match_strdup(&args[0]),
 301                                         flags, holder, fs_devices);
 302                         if (error)
 303                                 goto out_free_opts;
 304                         break;
 305                 default:
 306                         break;
 307                 }
 308         }
 309
 310  out_free_opts:
 311         kfree(opts);
 312  out:
 313         /*
 314          * If no subvolume name is specified we use the default one.  Allocate
 315          * a copy of the string "." here so that code later in the
 316          * mount path doesn't care if it's the default volume or another one.
 317          */
 318         if (!*subvol_name) {
 319                 *subvol_name = kstrdup(".", GFP_KERNEL);
 320                 if (!*subvol_name)
 321                         return -ENOMEM;
 322         }
 323         return error;
 324 }
 325
 326 static struct dentry *get_default_root(struct super_block *sb,
 327                                        u64 subvol_objectid)
 328 {
 329         struct btrfs_root *root = sb->s_fs_info;
 330         struct btrfs_root *new_root;
 331         struct btrfs_dir_item *di;
 332         struct btrfs_path *path;
 333         struct btrfs_key location;
 334         struct inode *inode;
 335         struct dentry *dentry;
 336         u64 dir_id;
 337         int new = 0;
 338
 339         /*
 340          * We have a specific subvol we want to mount, just setup location and
 341          * go look up the root.
 342          */
 343         if (subvol_objectid) {
 344                 location.objectid = subvol_objectid;
 345                 location.type = BTRFS_ROOT_ITEM_KEY;
 346                 location.offset = (u64)-1;
 347                 goto find_root;
 348         }
 349
 350         path = btrfs_alloc_path();
 351         if (!path)
 352                 return ERR_PTR(-ENOMEM);
 353         path->leave_spinning = 1;
 354
 355         /*
 356          * Find the "default" dir item which points to the root item that we
 357          * will mount by default if we haven't been given a specific subvolume
 358          * to mount.
 359          */
 360         dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
 361         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 362         if (!di) {
 363                 /*
 364                  * Ok the default dir item isn't there.  This is weird since
 365                  * it's always been there, but don't freak out, just try and
 366                  * mount to root most subvolume.
 367                  */
 368                 btrfs_free_path(path);
 369                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 370                 new_root = root->fs_info->fs_root;
 371                 goto setup_root;
 372         }
 373
 374         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 375         btrfs_free_path(path);
 376
 377 find_root:
 378         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
 379         if (IS_ERR(new_root))
 380                 return ERR_PTR(PTR_ERR(new_root));
 381
 382         if (btrfs_root_refs(&new_root->root_item) == 0)
 383                 return ERR_PTR(-ENOENT);
 384
 385         dir_id = btrfs_root_dirid(&new_root->root_item);
 386 setup_root:
 387         location.objectid = dir_id;
 388         location.type = BTRFS_INODE_ITEM_KEY;
 389         location.offset = 0;
 390
 391         inode = btrfs_iget(sb, &location, new_root, &new);
 392         if (!inode)
 393                 return ERR_PTR(-ENOMEM);
 394
 395         /*
 396          * If we're just mounting the root most subvol put the inode and return
 397          * a reference to the dentry.  We will have already gotten a reference
 398          * to the inode in btrfs_fill_super so we're good to go.
 399          */
 400         if (!new && sb->s_root->d_inode == inode) {
 401                 iput(inode);
 402                 return dget(sb->s_root);
 403         }
 404
 405         if (new) {
 406                 const struct qstr name = { .name = "/", .len = 1 };
 407
 408                 /*
 409                  * New inode, we need to make the dentry a sibling of s_root so
 410                  * everything gets cleaned up properly on unmount.
 411                  */
 412                 dentry = d_alloc(sb->s_root, &name);
 413                 if (!dentry) {
 414                         iput(inode);
 415                         return ERR_PTR(-ENOMEM);
 416                 }
 417                 d_splice_alias(inode, dentry);
 418         } else {
 419                 /*
 420                  * We found the inode in cache, just find a dentry for it and
 421                  * put the reference to the inode we just got.
 422                  */
 423                 dentry = d_find_alias(inode);
 424                 iput(inode);
 425         }
 426
 427         return dentry;
 428 }
 429
 430 static int btrfs_fill_super(struct super_block *sb,
 431                             struct btrfs_fs_devices *fs_devices,
 432                             void *data, int silent)
 433 {
 434         struct inode *inode;
 435         struct dentry *root_dentry;
 436         struct btrfs_super_block *disk_super;
 437         struct btrfs_root *tree_root;
 438         struct btrfs_key key;
 439         int err;
 440
 441         sb->s_maxbytes = MAX_LFS_FILESIZE;
 442         sb->s_magic = BTRFS_SUPER_MAGIC;
 443         sb->s_op = &btrfs_super_ops;
 444         sb->s_export_op = &btrfs_export_ops;
 445         sb->s_xattr = btrfs_xattr_handlers;
 446         sb->s_time_gran = 1;
 447 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 448         sb->s_flags |= MS_POSIXACL;
 449 #endif
 450
 451         tree_root = open_ctree(sb, fs_devices, (char *)data);
 452
 453         if (IS_ERR(tree_root)) {
 454                 printk("btrfs: open_ctree failed\n");
 455                 return PTR_ERR(tree_root);
 456         }
 457         sb->s_fs_info = tree_root;
 458         disk_super = &tree_root->fs_info->super_copy;
 459
 460         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 461         key.type = BTRFS_INODE_ITEM_KEY;
 462         key.offset = 0;
 463         inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
 464         if (IS_ERR(inode)) {
 465                 err = PTR_ERR(inode);
 466                 goto fail_close;
 467         }
 468
 469         root_dentry = d_alloc_root(inode);
 470         if (!root_dentry) {
 471                 iput(inode);
 472                 err = -ENOMEM;
 473                 goto fail_close;
 474         }
 475
 476         sb->s_root = root_dentry;
 477
 478         save_mount_options(sb, data);
 479         return 0;
 480
 481 fail_close:
 482         close_ctree(tree_root);
 483         return err;
 484 }
 485
 486 int btrfs_sync_fs(struct super_block *sb, int wait)
 487 {
 488         struct btrfs_trans_handle *trans;
 489         struct btrfs_root *root = btrfs_sb(sb);
 490         int ret;
 491
 492         if (!wait) {
 493                 filemap_flush(root->fs_info->btree_inode->i_mapping);
 494                 return 0;
 495         }
 496
 497         btrfs_start_delalloc_inodes(root, 0);
 498         btrfs_wait_ordered_extents(root, 0, 0);
 499
 500         trans = btrfs_start_transaction(root, 1);
 501         ret = btrfs_commit_transaction(trans, root);
 502         return ret;
 503 }
 504
 505 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 506 {
 507         struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
 508         struct btrfs_fs_info *info = root->fs_info;
 509
 510         if (btrfs_test_opt(root, DEGRADED))
 511                 seq_puts(seq, ",degraded");
 512         if (btrfs_test_opt(root, NODATASUM))
 513                 seq_puts(seq, ",nodatasum");
 514         if (btrfs_test_opt(root, NODATACOW))
 515                 seq_puts(seq, ",nodatacow");
 516         if (btrfs_test_opt(root, NOBARRIER))
 517                 seq_puts(seq, ",nobarrier");
 518         if (info->max_inline != 8192 * 1024)
 519                 seq_printf(seq, ",max_inline=%llu",
 520                            (unsigned long long)info->max_inline);
 521         if (info->alloc_start != 0)
 522                 seq_printf(seq, ",alloc_start=%llu",
 523                            (unsigned long long)info->alloc_start);
 524         if (info->thread_pool_size !=  min_t(unsigned long,
 525                                              num_online_cpus() + 2, 8))
 526                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
 527         if (btrfs_test_opt(root, COMPRESS))
 528                 seq_puts(seq, ",compress");
 529         if (btrfs_test_opt(root, NOSSD))
 530                 seq_puts(seq, ",nossd");
 531         if (btrfs_test_opt(root, SSD_SPREAD))
 532                 seq_puts(seq, ",ssd_spread");
 533         else if (btrfs_test_opt(root, SSD))
 534                 seq_puts(seq, ",ssd");
 535         if (btrfs_test_opt(root, NOTREELOG))
 536                 seq_puts(seq, ",notreelog");
 537         if (btrfs_test_opt(root, FLUSHONCOMMIT))
 538                 seq_puts(seq, ",flushoncommit");
 539         if (btrfs_test_opt(root, DISCARD))
 540                 seq_puts(seq, ",discard");
 541         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
 542                 seq_puts(seq, ",noacl");
 543         return 0;
 544 }
 545
 546 static int btrfs_test_super(struct super_block *s, void *data)
 547 {
 548         struct btrfs_fs_devices *test_fs_devices = data;
 549         struct btrfs_root *root = btrfs_sb(s);
 550
 551         return root->fs_info->fs_devices == test_fs_devices;
 552 }
 553
 554 /*
 555  * Find a superblock for the given device / mount point.
 556  *
 557  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
 558  *        for multiple device setup.  Make sure to keep it in sync.
 559  */
 560 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 561                 const char *dev_name, void *data, struct vfsmount *mnt)
 562 {
 563         struct block_device *bdev = NULL;
 564         struct super_block *s;
 565         struct dentry *root;
 566         struct btrfs_fs_devices *fs_devices = NULL;
 567         fmode_t mode = FMODE_READ;
 568         char *subvol_name = NULL;
 569         u64 subvol_objectid = 0;
 570         int error = 0;
 571         int found = 0;
 572
 573         if (!(flags & MS_RDONLY))
 574                 mode |= FMODE_WRITE;
 575
 576         error = btrfs_parse_early_options(data, mode, fs_type,
 577                                           &subvol_name, &subvol_objectid,
 578                                           &fs_devices);
 579         if (error)
 580                 return error;
 581
 582         error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
 583         if (error)
 584                 goto error_free_subvol_name;
 585
 586         error = btrfs_open_devices(fs_devices, mode, fs_type);
 587         if (error)
 588                 goto error_free_subvol_name;
 589
 590         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
 591                 error = -EACCES;
 592                 goto error_close_devices;
 593         }
 594
 595         bdev = fs_devices->latest_bdev;
 596         s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
 597         if (IS_ERR(s))
 598                 goto error_s;
 599
 600         if (s->s_root) {
 601                 if ((flags ^ s->s_flags) & MS_RDONLY) {
 602                         deactivate_locked_super(s);
 603                         error = -EBUSY;
 604                         goto error_close_devices;
 605                 }
 606
 607                 found = 1;
 608                 btrfs_close_devices(fs_devices);
 609         } else {
 610                 char b[BDEVNAME_SIZE];
 611
 612                 s->s_flags = flags;
 613                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 614                 error = btrfs_fill_super(s, fs_devices, data,
 615                                          flags & MS_SILENT ? 1 : 0);
 616                 if (error) {
 617                         deactivate_locked_super(s);
 618                         goto error_free_subvol_name;
 619                 }
 620
 621                 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
 622                 s->s_flags |= MS_ACTIVE;
 623         }
 624
 625         root = get_default_root(s, subvol_objectid);
 626         if (IS_ERR(root)) {
 627                 error = PTR_ERR(root);
 628                 deactivate_locked_super(s);
 629                 goto error;
 630         }
 631         /* if they gave us a subvolume name bind mount into that */
 632         if (strcmp(subvol_name, ".")) {
 633                 struct dentry *new_root;
 634                 mutex_lock(&root->d_inode->i_mutex);
 635                 new_root = lookup_one_len(subvol_name, root,
 636                                       strlen(subvol_name));
 637                 mutex_unlock(&root->d_inode->i_mutex);
 638
 639                 if (IS_ERR(new_root)) {
 640                         deactivate_locked_super(s);
 641                         error = PTR_ERR(new_root);
 642                         dput(root);
 643                         goto error_close_devices;
 644                 }
 645                 if (!new_root->d_inode) {
 646                         dput(root);
 647                         dput(new_root);
 648                         deactivate_locked_super(s);
 649                         error = -ENXIO;
 650                         goto error_close_devices;
 651                 }
 652                 dput(root);
 653                 root = new_root;
 654         }
 655
 656         mnt->mnt_sb = s;
 657         mnt->mnt_root = root;
 658
 659         kfree(subvol_name);
 660         return 0;
 661
 662 error_s:
 663         error = PTR_ERR(s);
 664 error_close_devices:
 665         btrfs_close_devices(fs_devices);
 666 error_free_subvol_name:
 667         kfree(subvol_name);
 668 error:
 669         return error;
 670 }
 671
 672 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
 673 {
 674         struct btrfs_root *root = btrfs_sb(sb);
 675         int ret;
 676
 677         ret = btrfs_parse_options(root, data);
 678         if (ret)
 679                 return -EINVAL;
 680
 681         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
 682                 return 0;
 683
 684         if (*flags & MS_RDONLY) {
 685                 sb->s_flags |= MS_RDONLY;
 686
 687                 ret =  btrfs_commit_super(root);
 688                 WARN_ON(ret);
 689         } else {
 690                 if (root->fs_info->fs_devices->rw_devices == 0)
 691                         return -EACCES;
 692
 693                 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
 694                         return -EINVAL;
 695
 696                 /* recover relocation */
 697                 ret = btrfs_recover_relocation(root);
 698                 WARN_ON(ret);
 699
 700                 ret = btrfs_cleanup_fs_roots(root->fs_info);
 701                 WARN_ON(ret);
 702
 703                 sb->s_flags &= ~MS_RDONLY;
 704         }
 705
 706         return 0;
 707 }
 708
 709 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 710 {
 711         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
 712         struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
 713         struct list_head *head = &root->fs_info->space_info;
 714         struct btrfs_space_info *found;
 715         u64 total_used = 0;
 716         u64 data_used = 0;
 717         int bits = dentry->d_sb->s_blocksize_bits;
 718         __be32 *fsid = (__be32 *)root->fs_info->fsid;
 719
 720         rcu_read_lock();
 721         list_for_each_entry_rcu(found, head, list) {
 722                 if (found->flags & (BTRFS_BLOCK_GROUP_DUP|
 723                                     BTRFS_BLOCK_GROUP_RAID10|
 724                                     BTRFS_BLOCK_GROUP_RAID1)) {
 725                         total_used += found->bytes_used;
 726                         if (found->flags & BTRFS_BLOCK_GROUP_DATA)
 727                                 data_used += found->bytes_used;
 728                         else
 729                                 data_used += found->total_bytes;
 730                 }
 731
 732                 total_used += found->bytes_used;
 733                 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
 734                         data_used += found->bytes_used;
 735                 else
 736                         data_used += found->total_bytes;
 737         }
 738         rcu_read_unlock();
 739
 740         buf->f_namelen = BTRFS_NAME_LEN;
 741         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
 742         buf->f_bfree = buf->f_blocks - (total_used >> bits);
 743         buf->f_bavail = buf->f_blocks - (data_used >> bits);
 744         buf->f_bsize = dentry->d_sb->s_blocksize;
 745         buf->f_type = BTRFS_SUPER_MAGIC;
 746
 747         /* We treat it as constant endianness (it doesn't matter _which_)
 748            because we want the fsid to come out the same whether mounted
 749            on a big-endian or little-endian host */
 750         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
 751         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
 752         /* Mask in the root object ID too, to disambiguate subvols */
 753         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
 754         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
 755
 756         return 0;
 757 }
 758
 759 static struct file_system_type btrfs_fs_type = {
 760         .owner          = THIS_MODULE,
 761         .name           = "btrfs",
 762         .get_sb         = btrfs_get_sb,
 763         .kill_sb        = kill_anon_super,
 764         .fs_flags       = FS_REQUIRES_DEV,
 765 };
 766
 767 /*
 768  * used by btrfsctl to scan devices when no FS is mounted
 769  */
 770 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
 771                                 unsigned long arg)
 772 {
 773         struct btrfs_ioctl_vol_args *vol;
 774         struct btrfs_fs_devices *fs_devices;
 775         int ret = -ENOTTY;
 776
 777         if (!capable(CAP_SYS_ADMIN))
 778                 return -EPERM;
 779
 780         vol = memdup_user((void __user *)arg, sizeof(*vol));
 781         if (IS_ERR(vol))
 782                 return PTR_ERR(vol);
 783
 784         switch (cmd) {
 785         case BTRFS_IOC_SCAN_DEV:
 786                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
 787                                             &btrfs_fs_type, &fs_devices);
 788                 break;
 789         }
 790
 791         kfree(vol);
 792         return ret;
 793 }
 794
 795 static int btrfs_freeze(struct super_block *sb)
 796 {
 797         struct btrfs_root *root = btrfs_sb(sb);
 798         mutex_lock(&root->fs_info->transaction_kthread_mutex);
 799         mutex_lock(&root->fs_info->cleaner_mutex);
 800         return 0;
 801 }
 802
 803 static int btrfs_unfreeze(struct super_block *sb)
 804 {
 805         struct btrfs_root *root = btrfs_sb(sb);
 806         mutex_unlock(&root->fs_info->cleaner_mutex);
 807         mutex_unlock(&root->fs_info->transaction_kthread_mutex);
 808         return 0;
 809 }
 810
 811 static const struct super_operations btrfs_super_ops = {
 812         .drop_inode     = btrfs_drop_inode,
 813         .delete_inode   = btrfs_delete_inode,
 814         .put_super      = btrfs_put_super,
 815         .sync_fs        = btrfs_sync_fs,
 816         .show_options   = btrfs_show_options,
 817         .write_inode    = btrfs_write_inode,
 818         .dirty_inode    = btrfs_dirty_inode,
 819         .alloc_inode    = btrfs_alloc_inode,
 820         .destroy_inode  = btrfs_destroy_inode,
 821         .statfs         = btrfs_statfs,
 822         .remount_fs     = btrfs_remount,
 823         .freeze_fs      = btrfs_freeze,
 824         .unfreeze_fs    = btrfs_unfreeze,
 825 };
 826
 827 static const struct file_operations btrfs_ctl_fops = {
 828         .unlocked_ioctl  = btrfs_control_ioctl,
 829         .compat_ioctl = btrfs_control_ioctl,
 830         .owner   = THIS_MODULE,
 831 };
 832
 833 static struct miscdevice btrfs_misc = {
 834         .minor          = MISC_DYNAMIC_MINOR,
 835         .name           = "btrfs-control",
 836         .fops           = &btrfs_ctl_fops
 837 };
 838
 839 static int btrfs_interface_init(void)
 840 {
 841         return misc_register(&btrfs_misc);
 842 }
 843
 844 static void btrfs_interface_exit(void)
 845 {
 846         if (misc_deregister(&btrfs_misc) < 0)
 847                 printk(KERN_INFO "misc_deregister failed for control device");
 848 }
 849
 850 static int __init init_btrfs_fs(void)
 851 {
 852         int err;
 853
 854         err = btrfs_init_sysfs();
 855         if (err)
 856                 return err;
 857
 858         err = btrfs_init_cachep();
 859         if (err)
 860                 goto free_sysfs;
 861
 862         err = extent_io_init();
 863         if (err)
 864                 goto free_cachep;
 865
 866         err = extent_map_init();
 867         if (err)
 868                 goto free_extent_io;
 869
 870         err = btrfs_interface_init();
 871         if (err)
 872                 goto free_extent_map;
 873
 874         err = register_filesystem(&btrfs_fs_type);
 875         if (err)
 876                 goto unregister_ioctl;
 877
 878         printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
 879         return 0;
 880
 881 unregister_ioctl:
 882         btrfs_interface_exit();
 883 free_extent_map:
 884         extent_map_exit();
 885 free_extent_io:
 886         extent_io_exit();
 887 free_cachep:
 888         btrfs_destroy_cachep();
 889 free_sysfs:
 890         btrfs_exit_sysfs();
 891         return err;
 892 }
 893
 894 static void __exit exit_btrfs_fs(void)
 895 {
 896         btrfs_destroy_cachep();
 897         extent_map_exit();
 898         extent_io_exit();
 899         btrfs_interface_exit();
 900         unregister_filesystem(&btrfs_fs_type);
 901         btrfs_exit_sysfs();
 902         btrfs_cleanup_fs_uuids();
 903         btrfs_zlib_exit();
 904 }
 905
 906 module_init(init_btrfs_fs)
 907 module_exit(exit_btrfs_fs)
 908
 909 MODULE_LICENSE("GPL");