Pileus Git - ~andy/linux/blob - 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 <linux/cleancache.h>
  43 #include <linux/ratelimit.h>
  44 #include "compat.h"
  45 #include "delayed-inode.h"
  46 #include "ctree.h"
  47 #include "disk-io.h"
  48 #include "transaction.h"
  49 #include "btrfs_inode.h"
  50 #include "ioctl.h"
  51 #include "print-tree.h"
  52 #include "xattr.h"
  53 #include "volumes.h"
  54 #include "version.h"
  55 #include "export.h"
  56 #include "compression.h"
  57
  58 #define CREATE_TRACE_POINTS
  59 #include <trace/events/btrfs.h>
  60
  61 static const struct super_operations btrfs_super_ops;
  62 static struct file_system_type btrfs_fs_type;
  63
  64 static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
  65                                       char nbuf[16])
  66 {
  67         char *errstr = NULL;
  68
  69         switch (errno) {
  70         case -EIO:
  71                 errstr = "IO failure";
  72                 break;
  73         case -ENOMEM:
  74                 errstr = "Out of memory";
  75                 break;
  76         case -EROFS:
  77                 errstr = "Readonly filesystem";
  78                 break;
  79         default:
  80                 if (nbuf) {
  81                         if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
  82                                 errstr = nbuf;
  83                 }
  84                 break;
  85         }
  86
  87         return errstr;
  88 }
  89
  90 static void __save_error_info(struct btrfs_fs_info *fs_info)
  91 {
  92         /*
  93          * today we only save the error info into ram.  Long term we'll
  94          * also send it down to the disk
  95          */
  96         fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
  97 }
  98
  99 /* NOTE:
 100  *      We move write_super stuff at umount in order to avoid deadlock
 101  *      for umount hold all lock.
 102  */
 103 static void save_error_info(struct btrfs_fs_info *fs_info)
 104 {
 105         __save_error_info(fs_info);
 106 }
 107
 108 /* btrfs handle error by forcing the filesystem readonly */
 109 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
 110 {
 111         struct super_block *sb = fs_info->sb;
 112
 113         if (sb->s_flags & MS_RDONLY)
 114                 return;
 115
 116         if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
 117                 sb->s_flags |= MS_RDONLY;
 118                 printk(KERN_INFO "btrfs is forced readonly\n");
 119         }
 120 }
 121
 122 /*
 123  * __btrfs_std_error decodes expected errors from the caller and
 124  * invokes the approciate error response.
 125  */
 126 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 127                      unsigned int line, int errno)
 128 {
 129         struct super_block *sb = fs_info->sb;
 130         char nbuf[16];
 131         const char *errstr;
 132
 133         /*
 134          * Special case: if the error is EROFS, and we're already
 135          * under MS_RDONLY, then it is safe here.
 136          */
 137         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 138                 return;
 139
 140         errstr = btrfs_decode_error(fs_info, errno, nbuf);
 141         printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
 142                 sb->s_id, function, line, errstr);
 143         save_error_info(fs_info);
 144
 145         btrfs_handle_error(fs_info);
 146 }
 147
 148 static void btrfs_put_super(struct super_block *sb)
 149 {
 150         (void)close_ctree(btrfs_sb(sb)->tree_root);
 151         /* FIXME: need to fix VFS to return error? */
 152         /* AV: return it _where_?  ->put_super() can be triggered by any number
 153          * of async events, up to and including delivery of SIGKILL to the
 154          * last process that kept it busy.  Or segfault in the aforementioned
 155          * process...  Whom would you report that to?
 156          */
 157 }
 158
 159 enum {
 160         Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
 161         Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
 162         Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
 163         Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
 164         Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
 165         Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
 166         Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
 167         Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
 168         Opt_check_integrity, Opt_check_integrity_including_extent_data,
 169         Opt_check_integrity_print_mask,
 170         Opt_err,
 171 };
 172
 173 static match_table_t tokens = {
 174         {Opt_degraded, "degraded"},
 175         {Opt_subvol, "subvol=%s"},
 176         {Opt_subvolid, "subvolid=%d"},
 177         {Opt_device, "device=%s"},
 178         {Opt_nodatasum, "nodatasum"},
 179         {Opt_nodatacow, "nodatacow"},
 180         {Opt_nobarrier, "nobarrier"},
 181         {Opt_max_inline, "max_inline=%s"},
 182         {Opt_alloc_start, "alloc_start=%s"},
 183         {Opt_thread_pool, "thread_pool=%d"},
 184         {Opt_compress, "compress"},
 185         {Opt_compress_type, "compress=%s"},
 186         {Opt_compress_force, "compress-force"},
 187         {Opt_compress_force_type, "compress-force=%s"},
 188         {Opt_ssd, "ssd"},
 189         {Opt_ssd_spread, "ssd_spread"},
 190         {Opt_nossd, "nossd"},
 191         {Opt_noacl, "noacl"},
 192         {Opt_notreelog, "notreelog"},
 193         {Opt_flushoncommit, "flushoncommit"},
 194         {Opt_ratio, "metadata_ratio=%d"},
 195         {Opt_discard, "discard"},
 196         {Opt_space_cache, "space_cache"},
 197         {Opt_clear_cache, "clear_cache"},
 198         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 199         {Opt_enospc_debug, "enospc_debug"},
 200         {Opt_subvolrootid, "subvolrootid=%d"},
 201         {Opt_defrag, "autodefrag"},
 202         {Opt_inode_cache, "inode_cache"},
 203         {Opt_no_space_cache, "nospace_cache"},
 204         {Opt_recovery, "recovery"},
 205         {Opt_skip_balance, "skip_balance"},
 206         {Opt_check_integrity, "check_int"},
 207         {Opt_check_integrity_including_extent_data, "check_int_data"},
 208         {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
 209         {Opt_err, NULL},
 210 };
 211
 212 /*
 213  * Regular mount options parser.  Everything that is needed only when
 214  * reading in a new superblock is parsed here.
 215  */
 216 int btrfs_parse_options(struct btrfs_root *root, char *options)
 217 {
 218         struct btrfs_fs_info *info = root->fs_info;
 219         substring_t args[MAX_OPT_ARGS];
 220         char *p, *num, *orig = NULL;
 221         u64 cache_gen;
 222         int intarg;
 223         int ret = 0;
 224         char *compress_type;
 225         bool compress_force = false;
 226
 227         cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
 228         if (cache_gen)
 229                 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 230
 231         if (!options)
 232                 goto out;
 233
 234         /*
 235          * strsep changes the string, duplicate it because parse_options
 236          * gets called twice
 237          */
 238         options = kstrdup(options, GFP_NOFS);
 239         if (!options)
 240                 return -ENOMEM;
 241
 242         orig = options;
 243
 244         while ((p = strsep(&options, ",")) != NULL) {
 245                 int token;
 246                 if (!*p)
 247                         continue;
 248
 249                 token = match_token(p, tokens, args);
 250                 switch (token) {
 251                 case Opt_degraded:
 252                         printk(KERN_INFO "btrfs: allowing degraded mounts\n");
 253                         btrfs_set_opt(info->mount_opt, DEGRADED);
 254                         break;
 255                 case Opt_subvol:
 256                 case Opt_subvolid:
 257                 case Opt_subvolrootid:
 258                 case Opt_device:
 259                         /*
 260                          * These are parsed by btrfs_parse_early_options
 261                          * and can be happily ignored here.
 262                          */
 263                         break;
 264                 case Opt_nodatasum:
 265                         printk(KERN_INFO "btrfs: setting nodatasum\n");
 266                         btrfs_set_opt(info->mount_opt, NODATASUM);
 267                         break;
 268                 case Opt_nodatacow:
 269                         printk(KERN_INFO "btrfs: setting nodatacow\n");
 270                         btrfs_set_opt(info->mount_opt, NODATACOW);
 271                         btrfs_set_opt(info->mount_opt, NODATASUM);
 272                         break;
 273                 case Opt_compress_force:
 274                 case Opt_compress_force_type:
 275                         compress_force = true;
 276                 case Opt_compress:
 277                 case Opt_compress_type:
 278                         if (token == Opt_compress ||
 279                             token == Opt_compress_force ||
 280                             strcmp(args[0].from, "zlib") == 0) {
 281                                 compress_type = "zlib";
 282                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 283                         } else if (strcmp(args[0].from, "lzo") == 0) {
 284                                 compress_type = "lzo";
 285                                 info->compress_type = BTRFS_COMPRESS_LZO;
 286                         } else {
 287                                 ret = -EINVAL;
 288                                 goto out;
 289                         }
 290
 291                         btrfs_set_opt(info->mount_opt, COMPRESS);
 292                         if (compress_force) {
 293                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 294                                 pr_info("btrfs: force %s compression\n",
 295                                         compress_type);
 296                         } else
 297                                 pr_info("btrfs: use %s compression\n",
 298                                         compress_type);
 299                         break;
 300                 case Opt_ssd:
 301                         printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 302                         btrfs_set_opt(info->mount_opt, SSD);
 303                         break;
 304                 case Opt_ssd_spread:
 305                         printk(KERN_INFO "btrfs: use spread ssd "
 306                                "allocation scheme\n");
 307                         btrfs_set_opt(info->mount_opt, SSD);
 308                         btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 309                         break;
 310                 case Opt_nossd:
 311                         printk(KERN_INFO "btrfs: not using ssd allocation "
 312                                "scheme\n");
 313                         btrfs_set_opt(info->mount_opt, NOSSD);
 314                         btrfs_clear_opt(info->mount_opt, SSD);
 315                         btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 316                         break;
 317                 case Opt_nobarrier:
 318                         printk(KERN_INFO "btrfs: turning off barriers\n");
 319                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 320                         break;
 321                 case Opt_thread_pool:
 322                         intarg = 0;
 323                         match_int(&args[0], &intarg);
 324                         if (intarg) {
 325                                 info->thread_pool_size = intarg;
 326                                 printk(KERN_INFO "btrfs: thread pool %d\n",
 327                                        info->thread_pool_size);
 328                         }
 329                         break;
 330                 case Opt_max_inline:
 331                         num = match_strdup(&args[0]);
 332                         if (num) {
 333                                 info->max_inline = memparse(num, NULL);
 334                                 kfree(num);
 335
 336                                 if (info->max_inline) {
 337                                         info->max_inline = max_t(u64,
 338                                                 info->max_inline,
 339                                                 root->sectorsize);
 340                                 }
 341                                 printk(KERN_INFO "btrfs: max_inline at %llu\n",
 342                                         (unsigned long long)info->max_inline);
 343                         }
 344                         break;
 345                 case Opt_alloc_start:
 346                         num = match_strdup(&args[0]);
 347                         if (num) {
 348                                 info->alloc_start = memparse(num, NULL);
 349                                 kfree(num);
 350                                 printk(KERN_INFO
 351                                         "btrfs: allocations start at %llu\n",
 352                                         (unsigned long long)info->alloc_start);
 353                         }
 354                         break;
 355                 case Opt_noacl:
 356                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 357                         break;
 358                 case Opt_notreelog:
 359                         printk(KERN_INFO "btrfs: disabling tree log\n");
 360                         btrfs_set_opt(info->mount_opt, NOTREELOG);
 361                         break;
 362                 case Opt_flushoncommit:
 363                         printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
 364                         btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
 365                         break;
 366                 case Opt_ratio:
 367                         intarg = 0;
 368                         match_int(&args[0], &intarg);
 369                         if (intarg) {
 370                                 info->metadata_ratio = intarg;
 371                                 printk(KERN_INFO "btrfs: metadata ratio %d\n",
 372                                        info->metadata_ratio);
 373                         }
 374                         break;
 375                 case Opt_discard:
 376                         btrfs_set_opt(info->mount_opt, DISCARD);
 377                         break;
 378                 case Opt_space_cache:
 379                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 380                         break;
 381                 case Opt_no_space_cache:
 382                         printk(KERN_INFO "btrfs: disabling disk space caching\n");
 383                         btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
 384                         break;
 385                 case Opt_inode_cache:
 386                         printk(KERN_INFO "btrfs: enabling inode map caching\n");
 387                         btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
 388                         break;
 389                 case Opt_clear_cache:
 390                         printk(KERN_INFO "btrfs: force clearing of disk cache\n");
 391                         btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
 392                         break;
 393                 case Opt_user_subvol_rm_allowed:
 394                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 395                         break;
 396                 case Opt_enospc_debug:
 397                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 398                         break;
 399                 case Opt_defrag:
 400                         printk(KERN_INFO "btrfs: enabling auto defrag");
 401                         btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
 402                         break;
 403                 case Opt_recovery:
 404                         printk(KERN_INFO "btrfs: enabling auto recovery");
 405                         btrfs_set_opt(info->mount_opt, RECOVERY);
 406                         break;
 407                 case Opt_skip_balance:
 408                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 409                         break;
 410 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 411                 case Opt_check_integrity_including_extent_data:
 412                         printk(KERN_INFO "btrfs: enabling check integrity"
 413                                " including extent data\n");
 414                         btrfs_set_opt(info->mount_opt,
 415                                       CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
 416                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 417                         break;
 418                 case Opt_check_integrity:
 419                         printk(KERN_INFO "btrfs: enabling check integrity\n");
 420                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 421                         break;
 422                 case Opt_check_integrity_print_mask:
 423                         intarg = 0;
 424                         match_int(&args[0], &intarg);
 425                         if (intarg) {
 426                                 info->check_integrity_print_mask = intarg;
 427                                 printk(KERN_INFO "btrfs:"
 428                                        " check_integrity_print_mask 0x%x\n",
 429                                        info->check_integrity_print_mask);
 430                         }
 431                         break;
 432 #else
 433                 case Opt_check_integrity_including_extent_data:
 434                 case Opt_check_integrity:
 435                 case Opt_check_integrity_print_mask:
 436                         printk(KERN_ERR "btrfs: support for check_integrity*"
 437                                " not compiled in!\n");
 438                         ret = -EINVAL;
 439                         goto out;
 440 #endif
 441                 case Opt_err:
 442                         printk(KERN_INFO "btrfs: unrecognized mount option "
 443                                "'%s'\n", p);
 444                         ret = -EINVAL;
 445                         goto out;
 446                 default:
 447                         break;
 448                 }
 449         }
 450 out:
 451         if (!ret && btrfs_test_opt(root, SPACE_CACHE))
 452                 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
 453         kfree(orig);
 454         return ret;
 455 }
 456
 457 /*
 458  * Parse mount options that are required early in the mount process.
 459  *
 460  * All other options will be parsed on much later in the mount process and
 461  * only when we need to allocate a new super block.
 462  */
 463 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 464                 void *holder, char **subvol_name, u64 *subvol_objectid,
 465                 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
 466 {
 467         substring_t args[MAX_OPT_ARGS];
 468         char *device_name, *opts, *orig, *p;
 469         int error = 0;
 470         int intarg;
 471
 472         if (!options)
 473                 return 0;
 474
 475         /*
 476          * strsep changes the string, duplicate it because parse_options
 477          * gets called twice
 478          */
 479         opts = kstrdup(options, GFP_KERNEL);
 480         if (!opts)
 481                 return -ENOMEM;
 482         orig = opts;
 483
 484         while ((p = strsep(&opts, ",")) != NULL) {
 485                 int token;
 486                 if (!*p)
 487                         continue;
 488
 489                 token = match_token(p, tokens, args);
 490                 switch (token) {
 491                 case Opt_subvol:
 492                         kfree(*subvol_name);
 493                         *subvol_name = match_strdup(&args[0]);
 494                         break;
 495                 case Opt_subvolid:
 496                         intarg = 0;
 497                         error = match_int(&args[0], &intarg);
 498                         if (!error) {
 499                                 /* we want the original fs_tree */
 500                                 if (!intarg)
 501                                         *subvol_objectid =
 502                                                 BTRFS_FS_TREE_OBJECTID;
 503                                 else
 504                                         *subvol_objectid = intarg;
 505                         }
 506                         break;
 507                 case Opt_subvolrootid:
 508                         intarg = 0;
 509                         error = match_int(&args[0], &intarg);
 510                         if (!error) {
 511                                 /* we want the original fs_tree */
 512                                 if (!intarg)
 513                                         *subvol_rootid =
 514                                                 BTRFS_FS_TREE_OBJECTID;
 515                                 else
 516                                         *subvol_rootid = intarg;
 517                         }
 518                         break;
 519                 case Opt_device:
 520                         device_name = match_strdup(&args[0]);
 521                         if (!device_name) {
 522                                 error = -ENOMEM;
 523                                 goto out;
 524                         }
 525                         error = btrfs_scan_one_device(device_name,
 526                                         flags, holder, fs_devices);
 527                         kfree(device_name);
 528                         if (error)
 529                                 goto out;
 530                         break;
 531                 default:
 532                         break;
 533                 }
 534         }
 535
 536 out:
 537         kfree(orig);
 538         return error;
 539 }
 540
 541 static struct dentry *get_default_root(struct super_block *sb,
 542                                        u64 subvol_objectid)
 543 {
 544         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 545         struct btrfs_root *root = fs_info->tree_root;
 546         struct btrfs_root *new_root;
 547         struct btrfs_dir_item *di;
 548         struct btrfs_path *path;
 549         struct btrfs_key location;
 550         struct inode *inode;
 551         u64 dir_id;
 552         int new = 0;
 553
 554         /*
 555          * We have a specific subvol we want to mount, just setup location and
 556          * go look up the root.
 557          */
 558         if (subvol_objectid) {
 559                 location.objectid = subvol_objectid;
 560                 location.type = BTRFS_ROOT_ITEM_KEY;
 561                 location.offset = (u64)-1;
 562                 goto find_root;
 563         }
 564
 565         path = btrfs_alloc_path();
 566         if (!path)
 567                 return ERR_PTR(-ENOMEM);
 568         path->leave_spinning = 1;
 569
 570         /*
 571          * Find the "default" dir item which points to the root item that we
 572          * will mount by default if we haven't been given a specific subvolume
 573          * to mount.
 574          */
 575         dir_id = btrfs_super_root_dir(fs_info->super_copy);
 576         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 577         if (IS_ERR(di)) {
 578                 btrfs_free_path(path);
 579                 return ERR_CAST(di);
 580         }
 581         if (!di) {
 582                 /*
 583                  * Ok the default dir item isn't there.  This is weird since
 584                  * it's always been there, but don't freak out, just try and
 585                  * mount to root most subvolume.
 586                  */
 587                 btrfs_free_path(path);
 588                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 589                 new_root = fs_info->fs_root;
 590                 goto setup_root;
 591         }
 592
 593         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 594         btrfs_free_path(path);
 595
 596 find_root:
 597         new_root = btrfs_read_fs_root_no_name(fs_info, &location);
 598         if (IS_ERR(new_root))
 599                 return ERR_CAST(new_root);
 600
 601         if (btrfs_root_refs(&new_root->root_item) == 0)
 602                 return ERR_PTR(-ENOENT);
 603
 604         dir_id = btrfs_root_dirid(&new_root->root_item);
 605 setup_root:
 606         location.objectid = dir_id;
 607         location.type = BTRFS_INODE_ITEM_KEY;
 608         location.offset = 0;
 609
 610         inode = btrfs_iget(sb, &location, new_root, &new);
 611         if (IS_ERR(inode))
 612                 return ERR_CAST(inode);
 613
 614         /*
 615          * If we're just mounting the root most subvol put the inode and return
 616          * a reference to the dentry.  We will have already gotten a reference
 617          * to the inode in btrfs_fill_super so we're good to go.
 618          */
 619         if (!new && sb->s_root->d_inode == inode) {
 620                 iput(inode);
 621                 return dget(sb->s_root);
 622         }
 623
 624         return d_obtain_alias(inode);
 625 }
 626
 627 static int btrfs_fill_super(struct super_block *sb,
 628                             struct btrfs_fs_devices *fs_devices,
 629                             void *data, int silent)
 630 {
 631         struct inode *inode;
 632         struct dentry *root_dentry;
 633         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 634         struct btrfs_key key;
 635         int err;
 636
 637         sb->s_maxbytes = MAX_LFS_FILESIZE;
 638         sb->s_magic = BTRFS_SUPER_MAGIC;
 639         sb->s_op = &btrfs_super_ops;
 640         sb->s_d_op = &btrfs_dentry_operations;
 641         sb->s_export_op = &btrfs_export_ops;
 642         sb->s_xattr = btrfs_xattr_handlers;
 643         sb->s_time_gran = 1;
 644 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 645         sb->s_flags |= MS_POSIXACL;
 646 #endif
 647
 648         err = open_ctree(sb, fs_devices, (char *)data);
 649         if (err) {
 650                 printk("btrfs: open_ctree failed\n");
 651                 return err;
 652         }
 653
 654         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 655         key.type = BTRFS_INODE_ITEM_KEY;
 656         key.offset = 0;
 657         inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
 658         if (IS_ERR(inode)) {
 659                 err = PTR_ERR(inode);
 660                 goto fail_close;
 661         }
 662
 663         root_dentry = d_alloc_root(inode);
 664         if (!root_dentry) {
 665                 iput(inode);
 666                 err = -ENOMEM;
 667                 goto fail_close;
 668         }
 669
 670         sb->s_root = root_dentry;
 671
 672         save_mount_options(sb, data);
 673         cleancache_init_fs(sb);
 674         sb->s_flags |= MS_ACTIVE;
 675         return 0;
 676
 677 fail_close:
 678         close_ctree(fs_info->tree_root);
 679         return err;
 680 }
 681
 682 int btrfs_sync_fs(struct super_block *sb, int wait)
 683 {
 684         struct btrfs_trans_handle *trans;
 685         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 686         struct btrfs_root *root = fs_info->tree_root;
 687         int ret;
 688
 689         trace_btrfs_sync_fs(wait);
 690
 691         if (!wait) {
 692                 filemap_flush(fs_info->btree_inode->i_mapping);
 693                 return 0;
 694         }
 695
 696         btrfs_start_delalloc_inodes(root, 0);
 697         btrfs_wait_ordered_extents(root, 0, 0);
 698
 699         trans = btrfs_start_transaction(root, 0);
 700         if (IS_ERR(trans))
 701                 return PTR_ERR(trans);
 702         ret = btrfs_commit_transaction(trans, root);
 703         return ret;
 704 }
 705
 706 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
 707 {
 708         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
 709         struct btrfs_root *root = info->tree_root;
 710         char *compress_type;
 711
 712         if (btrfs_test_opt(root, DEGRADED))
 713                 seq_puts(seq, ",degraded");
 714         if (btrfs_test_opt(root, NODATASUM))
 715                 seq_puts(seq, ",nodatasum");
 716         if (btrfs_test_opt(root, NODATACOW))
 717                 seq_puts(seq, ",nodatacow");
 718         if (btrfs_test_opt(root, NOBARRIER))
 719                 seq_puts(seq, ",nobarrier");
 720         if (info->max_inline != 8192 * 1024)
 721                 seq_printf(seq, ",max_inline=%llu",
 722                            (unsigned long long)info->max_inline);
 723         if (info->alloc_start != 0)
 724                 seq_printf(seq, ",alloc_start=%llu",
 725                            (unsigned long long)info->alloc_start);
 726         if (info->thread_pool_size !=  min_t(unsigned long,
 727                                              num_online_cpus() + 2, 8))
 728                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
 729         if (btrfs_test_opt(root, COMPRESS)) {
 730                 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
 731                         compress_type = "zlib";
 732                 else
 733                         compress_type = "lzo";
 734                 if (btrfs_test_opt(root, FORCE_COMPRESS))
 735                         seq_printf(seq, ",compress-force=%s", compress_type);
 736                 else
 737                         seq_printf(seq, ",compress=%s", compress_type);
 738         }
 739         if (btrfs_test_opt(root, NOSSD))
 740                 seq_puts(seq, ",nossd");
 741         if (btrfs_test_opt(root, SSD_SPREAD))
 742                 seq_puts(seq, ",ssd_spread");
 743         else if (btrfs_test_opt(root, SSD))
 744                 seq_puts(seq, ",ssd");
 745         if (btrfs_test_opt(root, NOTREELOG))
 746                 seq_puts(seq, ",notreelog");
 747         if (btrfs_test_opt(root, FLUSHONCOMMIT))
 748                 seq_puts(seq, ",flushoncommit");
 749         if (btrfs_test_opt(root, DISCARD))
 750                 seq_puts(seq, ",discard");
 751         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
 752                 seq_puts(seq, ",noacl");
 753         if (btrfs_test_opt(root, SPACE_CACHE))
 754                 seq_puts(seq, ",space_cache");
 755         else
 756                 seq_puts(seq, ",nospace_cache");
 757         if (btrfs_test_opt(root, CLEAR_CACHE))
 758                 seq_puts(seq, ",clear_cache");
 759         if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
 760                 seq_puts(seq, ",user_subvol_rm_allowed");
 761         if (btrfs_test_opt(root, ENOSPC_DEBUG))
 762                 seq_puts(seq, ",enospc_debug");
 763         if (btrfs_test_opt(root, AUTO_DEFRAG))
 764                 seq_puts(seq, ",autodefrag");
 765         if (btrfs_test_opt(root, INODE_MAP_CACHE))
 766                 seq_puts(seq, ",inode_cache");
 767         if (btrfs_test_opt(root, SKIP_BALANCE))
 768                 seq_puts(seq, ",skip_balance");
 769         return 0;
 770 }
 771
 772 static int btrfs_test_super(struct super_block *s, void *data)
 773 {
 774         struct btrfs_fs_info *p = data;
 775         struct btrfs_fs_info *fs_info = btrfs_sb(s);
 776
 777         return fs_info->fs_devices == p->fs_devices;
 778 }
 779
 780 static int btrfs_set_super(struct super_block *s, void *data)
 781 {
 782         int err = set_anon_super(s, data);
 783         if (!err)
 784                 s->s_fs_info = data;
 785         return err;
 786 }
 787
 788 /*
 789  * subvolumes are identified by ino 256
 790  */
 791 static inline int is_subvolume_inode(struct inode *inode)
 792 {
 793         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
 794                 return 1;
 795         return 0;
 796 }
 797
 798 /*
 799  * This will strip out the subvol=%s argument for an argument string and add
 800  * subvolid=0 to make sure we get the actual tree root for path walking to the
 801  * subvol we want.
 802  */
 803 static char *setup_root_args(char *args)
 804 {
 805         unsigned copied = 0;
 806         unsigned len = strlen(args) + 2;
 807         char *pos;
 808         char *ret;
 809
 810         /*
 811          * We need the same args as before, but minus
 812          *
 813          * subvol=a
 814          *
 815          * and add
 816          *
 817          * subvolid=0
 818          *
 819          * which is a difference of 2 characters, so we allocate strlen(args) +
 820          * 2 characters.
 821          */
 822         ret = kzalloc(len * sizeof(char), GFP_NOFS);
 823         if (!ret)
 824                 return NULL;
 825         pos = strstr(args, "subvol=");
 826
 827         /* This shouldn't happen, but just in case.. */
 828         if (!pos) {
 829                 kfree(ret);
 830                 return NULL;
 831         }
 832
 833         /*
 834          * The subvol=<> arg is not at the front of the string, copy everybody
 835          * up to that into ret.
 836          */
 837         if (pos != args) {
 838                 *pos = '\0';
 839                 strcpy(ret, args);
 840                 copied += strlen(args);
 841                 pos++;
 842         }
 843
 844         strncpy(ret + copied, "subvolid=0", len - copied);
 845
 846         /* Length of subvolid=0 */
 847         copied += 10;
 848
 849         /*
 850          * If there is no , after the subvol= option then we know there's no
 851          * other options and we can just return.
 852          */
 853         pos = strchr(pos, ',');
 854         if (!pos)
 855                 return ret;
 856
 857         /* Copy the rest of the arguments into our buffer */
 858         strncpy(ret + copied, pos, len - copied);
 859         copied += strlen(pos);
 860
 861         return ret;
 862 }
 863
 864 static struct dentry *mount_subvol(const char *subvol_name, int flags,
 865                                    const char *device_name, char *data)
 866 {
 867         struct dentry *root;
 868         struct vfsmount *mnt;
 869         char *newargs;
 870
 871         newargs = setup_root_args(data);
 872         if (!newargs)
 873                 return ERR_PTR(-ENOMEM);
 874         mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
 875                              newargs);
 876         kfree(newargs);
 877         if (IS_ERR(mnt))
 878                 return ERR_CAST(mnt);
 879
 880         root = mount_subtree(mnt, subvol_name);
 881
 882         if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
 883                 struct super_block *s = root->d_sb;
 884                 dput(root);
 885                 root = ERR_PTR(-EINVAL);
 886                 deactivate_locked_super(s);
 887                 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
 888                                 subvol_name);
 889         }
 890
 891         return root;
 892 }
 893
 894 /*
 895  * Find a superblock for the given device / mount point.
 896  *
 897  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
 898  *        for multiple device setup.  Make sure to keep it in sync.
 899  */
 900 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
 901                 const char *device_name, void *data)
 902 {
 903         struct block_device *bdev = NULL;
 904         struct super_block *s;
 905         struct dentry *root;
 906         struct btrfs_fs_devices *fs_devices = NULL;
 907         struct btrfs_fs_info *fs_info = NULL;
 908         fmode_t mode = FMODE_READ;
 909         char *subvol_name = NULL;
 910         u64 subvol_objectid = 0;
 911         u64 subvol_rootid = 0;
 912         int error = 0;
 913
 914         if (!(flags & MS_RDONLY))
 915                 mode |= FMODE_WRITE;
 916
 917         error = btrfs_parse_early_options(data, mode, fs_type,
 918                                           &subvol_name, &subvol_objectid,
 919                                           &subvol_rootid, &fs_devices);
 920         if (error) {
 921                 kfree(subvol_name);
 922                 return ERR_PTR(error);
 923         }
 924
 925         if (subvol_name) {
 926                 root = mount_subvol(subvol_name, flags, device_name, data);
 927                 kfree(subvol_name);
 928                 return root;
 929         }
 930
 931         error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
 932         if (error)
 933                 return ERR_PTR(error);
 934
 935         /*
 936          * Setup a dummy root and fs_info for test/set super.  This is because
 937          * we don't actually fill this stuff out until open_ctree, but we need
 938          * it for searching for existing supers, so this lets us do that and
 939          * then open_ctree will properly initialize everything later.
 940          */
 941         fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
 942         if (!fs_info)
 943                 return ERR_PTR(-ENOMEM);
 944
 945         fs_info->fs_devices = fs_devices;
 946
 947         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
 948         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
 949         if (!fs_info->super_copy || !fs_info->super_for_commit) {
 950                 error = -ENOMEM;
 951                 goto error_fs_info;
 952         }
 953
 954         error = btrfs_open_devices(fs_devices, mode, fs_type);
 955         if (error)
 956                 goto error_fs_info;
 957
 958         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
 959                 error = -EACCES;
 960                 goto error_close_devices;
 961         }
 962
 963         bdev = fs_devices->latest_bdev;
 964         s = sget(fs_type, btrfs_test_super, btrfs_set_super, fs_info);
 965         if (IS_ERR(s)) {
 966                 error = PTR_ERR(s);
 967                 goto error_close_devices;
 968         }
 969
 970         if (s->s_root) {
 971                 btrfs_close_devices(fs_devices);
 972                 free_fs_info(fs_info);
 973                 if ((flags ^ s->s_flags) & MS_RDONLY)
 974                         error = -EBUSY;
 975         } else {
 976                 char b[BDEVNAME_SIZE];
 977
 978                 s->s_flags = flags | MS_NOSEC;
 979                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 980                 btrfs_sb(s)->bdev_holder = fs_type;
 981                 error = btrfs_fill_super(s, fs_devices, data,
 982                                          flags & MS_SILENT ? 1 : 0);
 983         }
 984
 985         root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
 986         if (IS_ERR(root))
 987                 deactivate_locked_super(s);
 988
 989         return root;
 990
 991 error_close_devices:
 992         btrfs_close_devices(fs_devices);
 993 error_fs_info:
 994         free_fs_info(fs_info);
 995         return ERR_PTR(error);
 996 }
 997
 998 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
 999 {
1000         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1001         struct btrfs_root *root = fs_info->tree_root;
1002         int ret;
1003
1004         ret = btrfs_parse_options(root, data);
1005         if (ret)
1006                 return -EINVAL;
1007
1008         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1009                 return 0;
1010
1011         if (*flags & MS_RDONLY) {
1012                 sb->s_flags |= MS_RDONLY;
1013
1014                 ret =  btrfs_commit_super(root);
1015                 WARN_ON(ret);
1016         } else {
1017                 if (fs_info->fs_devices->rw_devices == 0)
1018                         return -EACCES;
1019
1020                 if (btrfs_super_log_root(fs_info->super_copy) != 0)
1021                         return -EINVAL;
1022
1023                 ret = btrfs_cleanup_fs_roots(fs_info);
1024                 WARN_ON(ret);
1025
1026                 /* recover relocation */
1027                 ret = btrfs_recover_relocation(root);
1028                 WARN_ON(ret);
1029
1030                 sb->s_flags &= ~MS_RDONLY;
1031         }
1032
1033         return 0;
1034 }
1035
1036 /* Used to sort the devices by max_avail(descending sort) */
1037 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1038                                        const void *dev_info2)
1039 {
1040         if (((struct btrfs_device_info *)dev_info1)->max_avail >
1041             ((struct btrfs_device_info *)dev_info2)->max_avail)
1042                 return -1;
1043         else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1044                  ((struct btrfs_device_info *)dev_info2)->max_avail)
1045                 return 1;
1046         else
1047         return 0;
1048 }
1049
1050 /*
1051  * sort the devices by max_avail, in which max free extent size of each device
1052  * is stored.(Descending Sort)
1053  */
1054 static inline void btrfs_descending_sort_devices(
1055                                         struct btrfs_device_info *devices,
1056                                         size_t nr_devices)
1057 {
1058         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1059              btrfs_cmp_device_free_bytes, NULL);
1060 }
1061
1062 /*
1063  * The helper to calc the free space on the devices that can be used to store
1064  * file data.
1065  */
1066 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1067 {
1068         struct btrfs_fs_info *fs_info = root->fs_info;
1069         struct btrfs_device_info *devices_info;
1070         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1071         struct btrfs_device *device;
1072         u64 skip_space;
1073         u64 type;
1074         u64 avail_space;
1075         u64 used_space;
1076         u64 min_stripe_size;
1077         int min_stripes = 1, num_stripes = 1;
1078         int i = 0, nr_devices;
1079         int ret;
1080
1081         nr_devices = fs_info->fs_devices->open_devices;
1082         BUG_ON(!nr_devices);
1083
1084         devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1085                                GFP_NOFS);
1086         if (!devices_info)
1087                 return -ENOMEM;
1088
1089         /* calc min stripe number for data space alloction */
1090         type = btrfs_get_alloc_profile(root, 1);
1091         if (type & BTRFS_BLOCK_GROUP_RAID0) {
1092                 min_stripes = 2;
1093                 num_stripes = nr_devices;
1094         } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1095                 min_stripes = 2;
1096                 num_stripes = 2;
1097         } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1098                 min_stripes = 4;
1099                 num_stripes = 4;
1100         }
1101
1102         if (type & BTRFS_BLOCK_GROUP_DUP)
1103                 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1104         else
1105                 min_stripe_size = BTRFS_STRIPE_LEN;
1106
1107         list_for_each_entry(device, &fs_devices->devices, dev_list) {
1108                 if (!device->in_fs_metadata || !device->bdev)
1109                         continue;
1110
1111                 avail_space = device->total_bytes - device->bytes_used;
1112
1113                 /* align with stripe_len */
1114                 do_div(avail_space, BTRFS_STRIPE_LEN);
1115                 avail_space *= BTRFS_STRIPE_LEN;
1116
1117                 /*
1118                  * In order to avoid overwritting the superblock on the drive,
1119                  * btrfs starts at an offset of at least 1MB when doing chunk
1120                  * allocation.
1121                  */
1122                 skip_space = 1024 * 1024;
1123
1124                 /* user can set the offset in fs_info->alloc_start. */
1125                 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1126                     device->total_bytes)
1127                         skip_space = max(fs_info->alloc_start, skip_space);
1128
1129                 /*
1130                  * btrfs can not use the free space in [0, skip_space - 1],
1131                  * we must subtract it from the total. In order to implement
1132                  * it, we account the used space in this range first.
1133                  */
1134                 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1135                                                      &used_space);
1136                 if (ret) {
1137                         kfree(devices_info);
1138                         return ret;
1139                 }
1140
1141                 /* calc the free space in [0, skip_space - 1] */
1142                 skip_space -= used_space;
1143
1144                 /*
1145                  * we can use the free space in [0, skip_space - 1], subtract
1146                  * it from the total.
1147                  */
1148                 if (avail_space && avail_space >= skip_space)
1149                         avail_space -= skip_space;
1150                 else
1151                         avail_space = 0;
1152
1153                 if (avail_space < min_stripe_size)
1154                         continue;
1155
1156                 devices_info[i].dev = device;
1157                 devices_info[i].max_avail = avail_space;
1158
1159                 i++;
1160         }
1161
1162         nr_devices = i;
1163
1164         btrfs_descending_sort_devices(devices_info, nr_devices);
1165
1166         i = nr_devices - 1;
1167         avail_space = 0;
1168         while (nr_devices >= min_stripes) {
1169                 if (num_stripes > nr_devices)
1170                         num_stripes = nr_devices;
1171
1172                 if (devices_info[i].max_avail >= min_stripe_size) {
1173                         int j;
1174                         u64 alloc_size;
1175
1176                         avail_space += devices_info[i].max_avail * num_stripes;
1177                         alloc_size = devices_info[i].max_avail;
1178                         for (j = i + 1 - num_stripes; j <= i; j++)
1179                                 devices_info[j].max_avail -= alloc_size;
1180                 }
1181                 i--;
1182                 nr_devices--;
1183         }
1184
1185         kfree(devices_info);
1186         *free_bytes = avail_space;
1187         return 0;
1188 }
1189
1190 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1191 {
1192         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1193         struct btrfs_super_block *disk_super = fs_info->super_copy;
1194         struct list_head *head = &fs_info->space_info;
1195         struct btrfs_space_info *found;
1196         u64 total_used = 0;
1197         u64 total_free_data = 0;
1198         int bits = dentry->d_sb->s_blocksize_bits;
1199         __be32 *fsid = (__be32 *)fs_info->fsid;
1200         int ret;
1201
1202         /* holding chunk_muext to avoid allocating new chunks */
1203         mutex_lock(&fs_info->chunk_mutex);
1204         rcu_read_lock();
1205         list_for_each_entry_rcu(found, head, list) {
1206                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1207                         total_free_data += found->disk_total - found->disk_used;
1208                         total_free_data -=
1209                                 btrfs_account_ro_block_groups_free_space(found);
1210                 }
1211
1212                 total_used += found->disk_used;
1213         }
1214         rcu_read_unlock();
1215
1216         buf->f_namelen = BTRFS_NAME_LEN;
1217         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1218         buf->f_bfree = buf->f_blocks - (total_used >> bits);
1219         buf->f_bsize = dentry->d_sb->s_blocksize;
1220         buf->f_type = BTRFS_SUPER_MAGIC;
1221         buf->f_bavail = total_free_data;
1222         ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1223         if (ret) {
1224                 mutex_unlock(&fs_info->chunk_mutex);
1225                 return ret;
1226         }
1227         buf->f_bavail += total_free_data;
1228         buf->f_bavail = buf->f_bavail >> bits;
1229         mutex_unlock(&fs_info->chunk_mutex);
1230
1231         /* We treat it as constant endianness (it doesn't matter _which_)
1232            because we want the fsid to come out the same whether mounted
1233            on a big-endian or little-endian host */
1234         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1235         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1236         /* Mask in the root object ID too, to disambiguate subvols */
1237         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1238         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1239
1240         return 0;
1241 }
1242
1243 static void btrfs_kill_super(struct super_block *sb)
1244 {
1245         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1246         kill_anon_super(sb);
1247         free_fs_info(fs_info);
1248 }
1249
1250 static struct file_system_type btrfs_fs_type = {
1251         .owner          = THIS_MODULE,
1252         .name           = "btrfs",
1253         .mount          = btrfs_mount,
1254         .kill_sb        = btrfs_kill_super,
1255         .fs_flags       = FS_REQUIRES_DEV,
1256 };
1257
1258 /*
1259  * used by btrfsctl to scan devices when no FS is mounted
1260  */
1261 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1262                                 unsigned long arg)
1263 {
1264         struct btrfs_ioctl_vol_args *vol;
1265         struct btrfs_fs_devices *fs_devices;
1266         int ret = -ENOTTY;
1267
1268         if (!capable(CAP_SYS_ADMIN))
1269                 return -EPERM;
1270
1271         vol = memdup_user((void __user *)arg, sizeof(*vol));
1272         if (IS_ERR(vol))
1273                 return PTR_ERR(vol);
1274
1275         switch (cmd) {
1276         case BTRFS_IOC_SCAN_DEV:
1277                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1278                                             &btrfs_fs_type, &fs_devices);
1279                 break;
1280         }
1281
1282         kfree(vol);
1283         return ret;
1284 }
1285
1286 static int btrfs_freeze(struct super_block *sb)
1287 {
1288         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1289         mutex_lock(&fs_info->transaction_kthread_mutex);
1290         mutex_lock(&fs_info->cleaner_mutex);
1291         return 0;
1292 }
1293
1294 static int btrfs_unfreeze(struct super_block *sb)
1295 {
1296         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1297         mutex_unlock(&fs_info->cleaner_mutex);
1298         mutex_unlock(&fs_info->transaction_kthread_mutex);
1299         return 0;
1300 }
1301
1302 static void btrfs_fs_dirty_inode(struct inode *inode, int flags)
1303 {
1304         int ret;
1305
1306         ret = btrfs_dirty_inode(inode);
1307         if (ret)
1308                 printk_ratelimited(KERN_ERR "btrfs: fail to dirty inode %Lu "
1309                                    "error %d\n", btrfs_ino(inode), ret);
1310 }
1311
1312 static const struct super_operations btrfs_super_ops = {
1313         .drop_inode     = btrfs_drop_inode,
1314         .evict_inode    = btrfs_evict_inode,
1315         .put_super      = btrfs_put_super,
1316         .sync_fs        = btrfs_sync_fs,
1317         .show_options   = btrfs_show_options,
1318         .write_inode    = btrfs_write_inode,
1319         .dirty_inode    = btrfs_fs_dirty_inode,
1320         .alloc_inode    = btrfs_alloc_inode,
1321         .destroy_inode  = btrfs_destroy_inode,
1322         .statfs         = btrfs_statfs,
1323         .remount_fs     = btrfs_remount,
1324         .freeze_fs      = btrfs_freeze,
1325         .unfreeze_fs    = btrfs_unfreeze,
1326 };
1327
1328 static const struct file_operations btrfs_ctl_fops = {
1329         .unlocked_ioctl  = btrfs_control_ioctl,
1330         .compat_ioctl = btrfs_control_ioctl,
1331         .owner   = THIS_MODULE,
1332         .llseek = noop_llseek,
1333 };
1334
1335 static struct miscdevice btrfs_misc = {
1336         .minor          = BTRFS_MINOR,
1337         .name           = "btrfs-control",
1338         .fops           = &btrfs_ctl_fops
1339 };
1340
1341 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1342 MODULE_ALIAS("devname:btrfs-control");
1343
1344 static int btrfs_interface_init(void)
1345 {
1346         return misc_register(&btrfs_misc);
1347 }
1348
1349 static void btrfs_interface_exit(void)
1350 {
1351         if (misc_deregister(&btrfs_misc) < 0)
1352                 printk(KERN_INFO "misc_deregister failed for control device");
1353 }
1354
1355 static int __init init_btrfs_fs(void)
1356 {
1357         int err;
1358
1359         err = btrfs_init_sysfs();
1360         if (err)
1361                 return err;
1362
1363         err = btrfs_init_compress();
1364         if (err)
1365                 goto free_sysfs;
1366
1367         err = btrfs_init_cachep();
1368         if (err)
1369                 goto free_compress;
1370
1371         err = extent_io_init();
1372         if (err)
1373                 goto free_cachep;
1374
1375         err = extent_map_init();
1376         if (err)
1377                 goto free_extent_io;
1378
1379         err = btrfs_delayed_inode_init();
1380         if (err)
1381                 goto free_extent_map;
1382
1383         err = btrfs_interface_init();
1384         if (err)
1385                 goto free_delayed_inode;
1386
1387         err = register_filesystem(&btrfs_fs_type);
1388         if (err)
1389                 goto unregister_ioctl;
1390
1391         printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1392         return 0;
1393
1394 unregister_ioctl:
1395         btrfs_interface_exit();
1396 free_delayed_inode:
1397         btrfs_delayed_inode_exit();
1398 free_extent_map:
1399         extent_map_exit();
1400 free_extent_io:
1401         extent_io_exit();
1402 free_cachep:
1403         btrfs_destroy_cachep();
1404 free_compress:
1405         btrfs_exit_compress();
1406 free_sysfs:
1407         btrfs_exit_sysfs();
1408         return err;
1409 }
1410
1411 static void __exit exit_btrfs_fs(void)
1412 {
1413         btrfs_destroy_cachep();
1414         btrfs_delayed_inode_exit();
1415         extent_map_exit();
1416         extent_io_exit();
1417         btrfs_interface_exit();
1418         unregister_filesystem(&btrfs_fs_type);
1419         btrfs_exit_sysfs();
1420         btrfs_cleanup_fs_uuids();
1421         btrfs_exit_compress();
1422 }
1423
1424 module_init(init_btrfs_fs)
1425 module_exit(exit_btrfs_fs)
1426
1427 MODULE_LICENSE("GPL");