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         case -EEXIST:
  80                 errstr = "Object already exists";
  81                 break;
  82         default:
  83                 if (nbuf) {
  84                         if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
  85                                 errstr = nbuf;
  86                 }
  87                 break;
  88         }
  89
  90         return errstr;
  91 }
  92
  93 static void __save_error_info(struct btrfs_fs_info *fs_info)
  94 {
  95         /*
  96          * today we only save the error info into ram.  Long term we'll
  97          * also send it down to the disk
  98          */
  99         fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
 100 }
 101
 102 /* NOTE:
 103  *      We move write_super stuff at umount in order to avoid deadlock
 104  *      for umount hold all lock.
 105  */
 106 static void save_error_info(struct btrfs_fs_info *fs_info)
 107 {
 108         __save_error_info(fs_info);
 109 }
 110
 111 /* btrfs handle error by forcing the filesystem readonly */
 112 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
 113 {
 114         struct super_block *sb = fs_info->sb;
 115
 116         if (sb->s_flags & MS_RDONLY)
 117                 return;
 118
 119         if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
 120                 sb->s_flags |= MS_RDONLY;
 121                 printk(KERN_INFO "btrfs is forced readonly\n");
 122         }
 123 }
 124
 125 /*
 126  * __btrfs_std_error decodes expected errors from the caller and
 127  * invokes the approciate error response.
 128  */
 129 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 130                      unsigned int line, int errno)
 131 {
 132         struct super_block *sb = fs_info->sb;
 133         char nbuf[16];
 134         const char *errstr;
 135
 136         /*
 137          * Special case: if the error is EROFS, and we're already
 138          * under MS_RDONLY, then it is safe here.
 139          */
 140         if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 141                 return;
 142
 143         errstr = btrfs_decode_error(fs_info, errno, nbuf);
 144         printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
 145                 sb->s_id, function, line, errstr);
 146         save_error_info(fs_info);
 147
 148         btrfs_handle_error(fs_info);
 149 }
 150
 151 /*
 152  * __btrfs_panic decodes unexpected, fatal errors from the caller,
 153  * issues an alert, and either panics or BUGs, depending on mount options.
 154  */
 155 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
 156                    unsigned int line, int errno, const char *fmt, ...)
 157 {
 158         char nbuf[16];
 159         char *s_id = "<unknown>";
 160         const char *errstr;
 161         struct va_format vaf = { .fmt = fmt };
 162         va_list args;
 163
 164         if (fs_info)
 165                 s_id = fs_info->sb->s_id;
 166
 167         va_start(args, fmt);
 168         vaf.va = &args;
 169
 170         errstr = btrfs_decode_error(fs_info, errno, nbuf);
 171         if (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR)
 172                 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
 173                         s_id, function, line, &vaf, errstr);
 174
 175         printk(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
 176                s_id, function, line, &vaf, errstr);
 177         va_end(args);
 178         /* Caller calls BUG() */
 179 }
 180
 181 static void btrfs_put_super(struct super_block *sb)
 182 {
 183         (void)close_ctree(btrfs_sb(sb)->tree_root);
 184         /* FIXME: need to fix VFS to return error? */
 185         /* AV: return it _where_?  ->put_super() can be triggered by any number
 186          * of async events, up to and including delivery of SIGKILL to the
 187          * last process that kept it busy.  Or segfault in the aforementioned
 188          * process...  Whom would you report that to?
 189          */
 190 }
 191
 192 enum {
 193         Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
 194         Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
 195         Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
 196         Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
 197         Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
 198         Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
 199         Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
 200         Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
 201         Opt_check_integrity, Opt_check_integrity_including_extent_data,
 202         Opt_check_integrity_print_mask, Opt_fatal_errors,
 203         Opt_err,
 204 };
 205
 206 static match_table_t tokens = {
 207         {Opt_degraded, "degraded"},
 208         {Opt_subvol, "subvol=%s"},
 209         {Opt_subvolid, "subvolid=%d"},
 210         {Opt_device, "device=%s"},
 211         {Opt_nodatasum, "nodatasum"},
 212         {Opt_nodatacow, "nodatacow"},
 213         {Opt_nobarrier, "nobarrier"},
 214         {Opt_max_inline, "max_inline=%s"},
 215         {Opt_alloc_start, "alloc_start=%s"},
 216         {Opt_thread_pool, "thread_pool=%d"},
 217         {Opt_compress, "compress"},
 218         {Opt_compress_type, "compress=%s"},
 219         {Opt_compress_force, "compress-force"},
 220         {Opt_compress_force_type, "compress-force=%s"},
 221         {Opt_ssd, "ssd"},
 222         {Opt_ssd_spread, "ssd_spread"},
 223         {Opt_nossd, "nossd"},
 224         {Opt_noacl, "noacl"},
 225         {Opt_notreelog, "notreelog"},
 226         {Opt_flushoncommit, "flushoncommit"},
 227         {Opt_ratio, "metadata_ratio=%d"},
 228         {Opt_discard, "discard"},
 229         {Opt_space_cache, "space_cache"},
 230         {Opt_clear_cache, "clear_cache"},
 231         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 232         {Opt_enospc_debug, "enospc_debug"},
 233         {Opt_subvolrootid, "subvolrootid=%d"},
 234         {Opt_defrag, "autodefrag"},
 235         {Opt_inode_cache, "inode_cache"},
 236         {Opt_no_space_cache, "nospace_cache"},
 237         {Opt_recovery, "recovery"},
 238         {Opt_skip_balance, "skip_balance"},
 239         {Opt_check_integrity, "check_int"},
 240         {Opt_check_integrity_including_extent_data, "check_int_data"},
 241         {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
 242         {Opt_fatal_errors, "fatal_errors=%s"},
 243         {Opt_err, NULL},
 244 };
 245
 246 /*
 247  * Regular mount options parser.  Everything that is needed only when
 248  * reading in a new superblock is parsed here.
 249  */
 250 int btrfs_parse_options(struct btrfs_root *root, char *options)
 251 {
 252         struct btrfs_fs_info *info = root->fs_info;
 253         substring_t args[MAX_OPT_ARGS];
 254         char *p, *num, *orig = NULL;
 255         u64 cache_gen;
 256         int intarg;
 257         int ret = 0;
 258         char *compress_type;
 259         bool compress_force = false;
 260
 261         cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
 262         if (cache_gen)
 263                 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 264
 265         if (!options)
 266                 goto out;
 267
 268         /*
 269          * strsep changes the string, duplicate it because parse_options
 270          * gets called twice
 271          */
 272         options = kstrdup(options, GFP_NOFS);
 273         if (!options)
 274                 return -ENOMEM;
 275
 276         orig = options;
 277
 278         while ((p = strsep(&options, ",")) != NULL) {
 279                 int token;
 280                 if (!*p)
 281                         continue;
 282
 283                 token = match_token(p, tokens, args);
 284                 switch (token) {
 285                 case Opt_degraded:
 286                         printk(KERN_INFO "btrfs: allowing degraded mounts\n");
 287                         btrfs_set_opt(info->mount_opt, DEGRADED);
 288                         break;
 289                 case Opt_subvol:
 290                 case Opt_subvolid:
 291                 case Opt_subvolrootid:
 292                 case Opt_device:
 293                         /*
 294                          * These are parsed by btrfs_parse_early_options
 295                          * and can be happily ignored here.
 296                          */
 297                         break;
 298                 case Opt_nodatasum:
 299                         printk(KERN_INFO "btrfs: setting nodatasum\n");
 300                         btrfs_set_opt(info->mount_opt, NODATASUM);
 301                         break;
 302                 case Opt_nodatacow:
 303                         printk(KERN_INFO "btrfs: setting nodatacow\n");
 304                         btrfs_set_opt(info->mount_opt, NODATACOW);
 305                         btrfs_set_opt(info->mount_opt, NODATASUM);
 306                         break;
 307                 case Opt_compress_force:
 308                 case Opt_compress_force_type:
 309                         compress_force = true;
 310                 case Opt_compress:
 311                 case Opt_compress_type:
 312                         if (token == Opt_compress ||
 313                             token == Opt_compress_force ||
 314                             strcmp(args[0].from, "zlib") == 0) {
 315                                 compress_type = "zlib";
 316                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 317                         } else if (strcmp(args[0].from, "lzo") == 0) {
 318                                 compress_type = "lzo";
 319                                 info->compress_type = BTRFS_COMPRESS_LZO;
 320                         } else {
 321                                 ret = -EINVAL;
 322                                 goto out;
 323                         }
 324
 325                         btrfs_set_opt(info->mount_opt, COMPRESS);
 326                         if (compress_force) {
 327                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 328                                 pr_info("btrfs: force %s compression\n",
 329                                         compress_type);
 330                         } else
 331                                 pr_info("btrfs: use %s compression\n",
 332                                         compress_type);
 333                         break;
 334                 case Opt_ssd:
 335                         printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 336                         btrfs_set_opt(info->mount_opt, SSD);
 337                         break;
 338                 case Opt_ssd_spread:
 339                         printk(KERN_INFO "btrfs: use spread ssd "
 340                                "allocation scheme\n");
 341                         btrfs_set_opt(info->mount_opt, SSD);
 342                         btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 343                         break;
 344                 case Opt_nossd:
 345                         printk(KERN_INFO "btrfs: not using ssd allocation "
 346                                "scheme\n");
 347                         btrfs_set_opt(info->mount_opt, NOSSD);
 348                         btrfs_clear_opt(info->mount_opt, SSD);
 349                         btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 350                         break;
 351                 case Opt_nobarrier:
 352                         printk(KERN_INFO "btrfs: turning off barriers\n");
 353                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 354                         break;
 355                 case Opt_thread_pool:
 356                         intarg = 0;
 357                         match_int(&args[0], &intarg);
 358                         if (intarg) {
 359                                 info->thread_pool_size = intarg;
 360                                 printk(KERN_INFO "btrfs: thread pool %d\n",
 361                                        info->thread_pool_size);
 362                         }
 363                         break;
 364                 case Opt_max_inline:
 365                         num = match_strdup(&args[0]);
 366                         if (num) {
 367                                 info->max_inline = memparse(num, NULL);
 368                                 kfree(num);
 369
 370                                 if (info->max_inline) {
 371                                         info->max_inline = max_t(u64,
 372                                                 info->max_inline,
 373                                                 root->sectorsize);
 374                                 }
 375                                 printk(KERN_INFO "btrfs: max_inline at %llu\n",
 376                                         (unsigned long long)info->max_inline);
 377                         }
 378                         break;
 379                 case Opt_alloc_start:
 380                         num = match_strdup(&args[0]);
 381                         if (num) {
 382                                 info->alloc_start = memparse(num, NULL);
 383                                 kfree(num);
 384                                 printk(KERN_INFO
 385                                         "btrfs: allocations start at %llu\n",
 386                                         (unsigned long long)info->alloc_start);
 387                         }
 388                         break;
 389                 case Opt_noacl:
 390                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 391                         break;
 392                 case Opt_notreelog:
 393                         printk(KERN_INFO "btrfs: disabling tree log\n");
 394                         btrfs_set_opt(info->mount_opt, NOTREELOG);
 395                         break;
 396                 case Opt_flushoncommit:
 397                         printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
 398                         btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
 399                         break;
 400                 case Opt_ratio:
 401                         intarg = 0;
 402                         match_int(&args[0], &intarg);
 403                         if (intarg) {
 404                                 info->metadata_ratio = intarg;
 405                                 printk(KERN_INFO "btrfs: metadata ratio %d\n",
 406                                        info->metadata_ratio);
 407                         }
 408                         break;
 409                 case Opt_discard:
 410                         btrfs_set_opt(info->mount_opt, DISCARD);
 411                         break;
 412                 case Opt_space_cache:
 413                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 414                         break;
 415                 case Opt_no_space_cache:
 416                         printk(KERN_INFO "btrfs: disabling disk space caching\n");
 417                         btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
 418                         break;
 419                 case Opt_inode_cache:
 420                         printk(KERN_INFO "btrfs: enabling inode map caching\n");
 421                         btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
 422                         break;
 423                 case Opt_clear_cache:
 424                         printk(KERN_INFO "btrfs: force clearing of disk cache\n");
 425                         btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
 426                         break;
 427                 case Opt_user_subvol_rm_allowed:
 428                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 429                         break;
 430                 case Opt_enospc_debug:
 431                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 432                         break;
 433                 case Opt_defrag:
 434                         printk(KERN_INFO "btrfs: enabling auto defrag");
 435                         btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
 436                         break;
 437                 case Opt_recovery:
 438                         printk(KERN_INFO "btrfs: enabling auto recovery");
 439                         btrfs_set_opt(info->mount_opt, RECOVERY);
 440                         break;
 441                 case Opt_skip_balance:
 442                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 443                         break;
 444 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 445                 case Opt_check_integrity_including_extent_data:
 446                         printk(KERN_INFO "btrfs: enabling check integrity"
 447                                " including extent data\n");
 448                         btrfs_set_opt(info->mount_opt,
 449                                       CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
 450                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 451                         break;
 452                 case Opt_check_integrity:
 453                         printk(KERN_INFO "btrfs: enabling check integrity\n");
 454                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 455                         break;
 456                 case Opt_check_integrity_print_mask:
 457                         intarg = 0;
 458                         match_int(&args[0], &intarg);
 459                         if (intarg) {
 460                                 info->check_integrity_print_mask = intarg;
 461                                 printk(KERN_INFO "btrfs:"
 462                                        " check_integrity_print_mask 0x%x\n",
 463                                        info->check_integrity_print_mask);
 464                         }
 465                         break;
 466 #else
 467                 case Opt_check_integrity_including_extent_data:
 468                 case Opt_check_integrity:
 469                 case Opt_check_integrity_print_mask:
 470                         printk(KERN_ERR "btrfs: support for check_integrity*"
 471                                " not compiled in!\n");
 472                         ret = -EINVAL;
 473                         goto out;
 474 #endif
 475                 case Opt_fatal_errors:
 476                         if (strcmp(args[0].from, "panic") == 0)
 477                                 btrfs_set_opt(info->mount_opt,
 478                                               PANIC_ON_FATAL_ERROR);
 479                         else if (strcmp(args[0].from, "bug") == 0)
 480                                 btrfs_clear_opt(info->mount_opt,
 481                                               PANIC_ON_FATAL_ERROR);
 482                         else {
 483                                 ret = -EINVAL;
 484                                 goto out;
 485                         }
 486                         break;
 487                 case Opt_err:
 488                         printk(KERN_INFO "btrfs: unrecognized mount option "
 489                                "'%s'\n", p);
 490                         ret = -EINVAL;
 491                         goto out;
 492                 default:
 493                         break;
 494                 }
 495         }
 496 out:
 497         if (!ret && btrfs_test_opt(root, SPACE_CACHE))
 498                 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
 499         kfree(orig);
 500         return ret;
 501 }
 502
 503 /*
 504  * Parse mount options that are required early in the mount process.
 505  *
 506  * All other options will be parsed on much later in the mount process and
 507  * only when we need to allocate a new super block.
 508  */
 509 static int btrfs_parse_early_options(const char *options, fmode_t flags,
 510                 void *holder, char **subvol_name, u64 *subvol_objectid,
 511                 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
 512 {
 513         substring_t args[MAX_OPT_ARGS];
 514         char *device_name, *opts, *orig, *p;
 515         int error = 0;
 516         int intarg;
 517
 518         if (!options)
 519                 return 0;
 520
 521         /*
 522          * strsep changes the string, duplicate it because parse_options
 523          * gets called twice
 524          */
 525         opts = kstrdup(options, GFP_KERNEL);
 526         if (!opts)
 527                 return -ENOMEM;
 528         orig = opts;
 529
 530         while ((p = strsep(&opts, ",")) != NULL) {
 531                 int token;
 532                 if (!*p)
 533                         continue;
 534
 535                 token = match_token(p, tokens, args);
 536                 switch (token) {
 537                 case Opt_subvol:
 538                         kfree(*subvol_name);
 539                         *subvol_name = match_strdup(&args[0]);
 540                         break;
 541                 case Opt_subvolid:
 542                         intarg = 0;
 543                         error = match_int(&args[0], &intarg);
 544                         if (!error) {
 545                                 /* we want the original fs_tree */
 546                                 if (!intarg)
 547                                         *subvol_objectid =
 548                                                 BTRFS_FS_TREE_OBJECTID;
 549                                 else
 550                                         *subvol_objectid = intarg;
 551                         }
 552                         break;
 553                 case Opt_subvolrootid:
 554                         intarg = 0;
 555                         error = match_int(&args[0], &intarg);
 556                         if (!error) {
 557                                 /* we want the original fs_tree */
 558                                 if (!intarg)
 559                                         *subvol_rootid =
 560                                                 BTRFS_FS_TREE_OBJECTID;
 561                                 else
 562                                         *subvol_rootid = intarg;
 563                         }
 564                         break;
 565                 case Opt_device:
 566                         device_name = match_strdup(&args[0]);
 567                         if (!device_name) {
 568                                 error = -ENOMEM;
 569                                 goto out;
 570                         }
 571                         error = btrfs_scan_one_device(device_name,
 572                                         flags, holder, fs_devices);
 573                         kfree(device_name);
 574                         if (error)
 575                                 goto out;
 576                         break;
 577                 default:
 578                         break;
 579                 }
 580         }
 581
 582 out:
 583         kfree(orig);
 584         return error;
 585 }
 586
 587 static struct dentry *get_default_root(struct super_block *sb,
 588                                        u64 subvol_objectid)
 589 {
 590         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 591         struct btrfs_root *root = fs_info->tree_root;
 592         struct btrfs_root *new_root;
 593         struct btrfs_dir_item *di;
 594         struct btrfs_path *path;
 595         struct btrfs_key location;
 596         struct inode *inode;
 597         u64 dir_id;
 598         int new = 0;
 599
 600         /*
 601          * We have a specific subvol we want to mount, just setup location and
 602          * go look up the root.
 603          */
 604         if (subvol_objectid) {
 605                 location.objectid = subvol_objectid;
 606                 location.type = BTRFS_ROOT_ITEM_KEY;
 607                 location.offset = (u64)-1;
 608                 goto find_root;
 609         }
 610
 611         path = btrfs_alloc_path();
 612         if (!path)
 613                 return ERR_PTR(-ENOMEM);
 614         path->leave_spinning = 1;
 615
 616         /*
 617          * Find the "default" dir item which points to the root item that we
 618          * will mount by default if we haven't been given a specific subvolume
 619          * to mount.
 620          */
 621         dir_id = btrfs_super_root_dir(fs_info->super_copy);
 622         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 623         if (IS_ERR(di)) {
 624                 btrfs_free_path(path);
 625                 return ERR_CAST(di);
 626         }
 627         if (!di) {
 628                 /*
 629                  * Ok the default dir item isn't there.  This is weird since
 630                  * it's always been there, but don't freak out, just try and
 631                  * mount to root most subvolume.
 632                  */
 633                 btrfs_free_path(path);
 634                 dir_id = BTRFS_FIRST_FREE_OBJECTID;
 635                 new_root = fs_info->fs_root;
 636                 goto setup_root;
 637         }
 638
 639         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 640         btrfs_free_path(path);
 641
 642 find_root:
 643         new_root = btrfs_read_fs_root_no_name(fs_info, &location);
 644         if (IS_ERR(new_root))
 645                 return ERR_CAST(new_root);
 646
 647         if (btrfs_root_refs(&new_root->root_item) == 0)
 648                 return ERR_PTR(-ENOENT);
 649
 650         dir_id = btrfs_root_dirid(&new_root->root_item);
 651 setup_root:
 652         location.objectid = dir_id;
 653         location.type = BTRFS_INODE_ITEM_KEY;
 654         location.offset = 0;
 655
 656         inode = btrfs_iget(sb, &location, new_root, &new);
 657         if (IS_ERR(inode))
 658                 return ERR_CAST(inode);
 659
 660         /*
 661          * If we're just mounting the root most subvol put the inode and return
 662          * a reference to the dentry.  We will have already gotten a reference
 663          * to the inode in btrfs_fill_super so we're good to go.
 664          */
 665         if (!new && sb->s_root->d_inode == inode) {
 666                 iput(inode);
 667                 return dget(sb->s_root);
 668         }
 669
 670         return d_obtain_alias(inode);
 671 }
 672
 673 static int btrfs_fill_super(struct super_block *sb,
 674                             struct btrfs_fs_devices *fs_devices,
 675                             void *data, int silent)
 676 {
 677         struct inode *inode;
 678         struct dentry *root_dentry;
 679         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 680         struct btrfs_key key;
 681         int err;
 682
 683         sb->s_maxbytes = MAX_LFS_FILESIZE;
 684         sb->s_magic = BTRFS_SUPER_MAGIC;
 685         sb->s_op = &btrfs_super_ops;
 686         sb->s_d_op = &btrfs_dentry_operations;
 687         sb->s_export_op = &btrfs_export_ops;
 688         sb->s_xattr = btrfs_xattr_handlers;
 689         sb->s_time_gran = 1;
 690 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 691         sb->s_flags |= MS_POSIXACL;
 692 #endif
 693
 694         err = open_ctree(sb, fs_devices, (char *)data);
 695         if (err) {
 696                 printk("btrfs: open_ctree failed\n");
 697                 return err;
 698         }
 699
 700         key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 701         key.type = BTRFS_INODE_ITEM_KEY;
 702         key.offset = 0;
 703         inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
 704         if (IS_ERR(inode)) {
 705                 err = PTR_ERR(inode);
 706                 goto fail_close;
 707         }
 708
 709         root_dentry = d_alloc_root(inode);
 710         if (!root_dentry) {
 711                 iput(inode);
 712                 err = -ENOMEM;
 713                 goto fail_close;
 714         }
 715
 716         sb->s_root = root_dentry;
 717
 718         save_mount_options(sb, data);
 719         cleancache_init_fs(sb);
 720         sb->s_flags |= MS_ACTIVE;
 721         return 0;
 722
 723 fail_close:
 724         close_ctree(fs_info->tree_root);
 725         return err;
 726 }
 727
 728 int btrfs_sync_fs(struct super_block *sb, int wait)
 729 {
 730         struct btrfs_trans_handle *trans;
 731         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 732         struct btrfs_root *root = fs_info->tree_root;
 733         int ret;
 734
 735         trace_btrfs_sync_fs(wait);
 736
 737         if (!wait) {
 738                 filemap_flush(fs_info->btree_inode->i_mapping);
 739                 return 0;
 740         }
 741
 742         btrfs_start_delalloc_inodes(root, 0);
 743         btrfs_wait_ordered_extents(root, 0, 0);
 744
 745         trans = btrfs_start_transaction(root, 0);
 746         if (IS_ERR(trans))
 747                 return PTR_ERR(trans);
 748         ret = btrfs_commit_transaction(trans, root);
 749         return ret;
 750 }
 751
 752 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
 753 {
 754         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
 755         struct btrfs_root *root = info->tree_root;
 756         char *compress_type;
 757
 758         if (btrfs_test_opt(root, DEGRADED))
 759                 seq_puts(seq, ",degraded");
 760         if (btrfs_test_opt(root, NODATASUM))
 761                 seq_puts(seq, ",nodatasum");
 762         if (btrfs_test_opt(root, NODATACOW))
 763                 seq_puts(seq, ",nodatacow");
 764         if (btrfs_test_opt(root, NOBARRIER))
 765                 seq_puts(seq, ",nobarrier");
 766         if (info->max_inline != 8192 * 1024)
 767                 seq_printf(seq, ",max_inline=%llu",
 768                            (unsigned long long)info->max_inline);
 769         if (info->alloc_start != 0)
 770                 seq_printf(seq, ",alloc_start=%llu",
 771                            (unsigned long long)info->alloc_start);
 772         if (info->thread_pool_size !=  min_t(unsigned long,
 773                                              num_online_cpus() + 2, 8))
 774                 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
 775         if (btrfs_test_opt(root, COMPRESS)) {
 776                 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
 777                         compress_type = "zlib";
 778                 else
 779                         compress_type = "lzo";
 780                 if (btrfs_test_opt(root, FORCE_COMPRESS))
 781                         seq_printf(seq, ",compress-force=%s", compress_type);
 782                 else
 783                         seq_printf(seq, ",compress=%s", compress_type);
 784         }
 785         if (btrfs_test_opt(root, NOSSD))
 786                 seq_puts(seq, ",nossd");
 787         if (btrfs_test_opt(root, SSD_SPREAD))
 788                 seq_puts(seq, ",ssd_spread");
 789         else if (btrfs_test_opt(root, SSD))
 790                 seq_puts(seq, ",ssd");
 791         if (btrfs_test_opt(root, NOTREELOG))
 792                 seq_puts(seq, ",notreelog");
 793         if (btrfs_test_opt(root, FLUSHONCOMMIT))
 794                 seq_puts(seq, ",flushoncommit");
 795         if (btrfs_test_opt(root, DISCARD))
 796                 seq_puts(seq, ",discard");
 797         if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
 798                 seq_puts(seq, ",noacl");
 799         if (btrfs_test_opt(root, SPACE_CACHE))
 800                 seq_puts(seq, ",space_cache");
 801         else
 802                 seq_puts(seq, ",nospace_cache");
 803         if (btrfs_test_opt(root, CLEAR_CACHE))
 804                 seq_puts(seq, ",clear_cache");
 805         if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
 806                 seq_puts(seq, ",user_subvol_rm_allowed");
 807         if (btrfs_test_opt(root, ENOSPC_DEBUG))
 808                 seq_puts(seq, ",enospc_debug");
 809         if (btrfs_test_opt(root, AUTO_DEFRAG))
 810                 seq_puts(seq, ",autodefrag");
 811         if (btrfs_test_opt(root, INODE_MAP_CACHE))
 812                 seq_puts(seq, ",inode_cache");
 813         if (btrfs_test_opt(root, SKIP_BALANCE))
 814                 seq_puts(seq, ",skip_balance");
 815         if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
 816                 seq_puts(seq, ",fatal_errors=panic");
 817         return 0;
 818 }
 819
 820 static int btrfs_test_super(struct super_block *s, void *data)
 821 {
 822         struct btrfs_fs_info *p = data;
 823         struct btrfs_fs_info *fs_info = btrfs_sb(s);
 824
 825         return fs_info->fs_devices == p->fs_devices;
 826 }
 827
 828 static int btrfs_set_super(struct super_block *s, void *data)
 829 {
 830         int err = set_anon_super(s, data);
 831         if (!err)
 832                 s->s_fs_info = data;
 833         return err;
 834 }
 835
 836 /*
 837  * subvolumes are identified by ino 256
 838  */
 839 static inline int is_subvolume_inode(struct inode *inode)
 840 {
 841         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
 842                 return 1;
 843         return 0;
 844 }
 845
 846 /*
 847  * This will strip out the subvol=%s argument for an argument string and add
 848  * subvolid=0 to make sure we get the actual tree root for path walking to the
 849  * subvol we want.
 850  */
 851 static char *setup_root_args(char *args)
 852 {
 853         unsigned copied = 0;
 854         unsigned len = strlen(args) + 2;
 855         char *pos;
 856         char *ret;
 857
 858         /*
 859          * We need the same args as before, but minus
 860          *
 861          * subvol=a
 862          *
 863          * and add
 864          *
 865          * subvolid=0
 866          *
 867          * which is a difference of 2 characters, so we allocate strlen(args) +
 868          * 2 characters.
 869          */
 870         ret = kzalloc(len * sizeof(char), GFP_NOFS);
 871         if (!ret)
 872                 return NULL;
 873         pos = strstr(args, "subvol=");
 874
 875         /* This shouldn't happen, but just in case.. */
 876         if (!pos) {
 877                 kfree(ret);
 878                 return NULL;
 879         }
 880
 881         /*
 882          * The subvol=<> arg is not at the front of the string, copy everybody
 883          * up to that into ret.
 884          */
 885         if (pos != args) {
 886                 *pos = '\0';
 887                 strcpy(ret, args);
 888                 copied += strlen(args);
 889                 pos++;
 890         }
 891
 892         strncpy(ret + copied, "subvolid=0", len - copied);
 893
 894         /* Length of subvolid=0 */
 895         copied += 10;
 896
 897         /*
 898          * If there is no , after the subvol= option then we know there's no
 899          * other options and we can just return.
 900          */
 901         pos = strchr(pos, ',');
 902         if (!pos)
 903                 return ret;
 904
 905         /* Copy the rest of the arguments into our buffer */
 906         strncpy(ret + copied, pos, len - copied);
 907         copied += strlen(pos);
 908
 909         return ret;
 910 }
 911
 912 static struct dentry *mount_subvol(const char *subvol_name, int flags,
 913                                    const char *device_name, char *data)
 914 {
 915         struct dentry *root;
 916         struct vfsmount *mnt;
 917         char *newargs;
 918
 919         newargs = setup_root_args(data);
 920         if (!newargs)
 921                 return ERR_PTR(-ENOMEM);
 922         mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
 923                              newargs);
 924         kfree(newargs);
 925         if (IS_ERR(mnt))
 926                 return ERR_CAST(mnt);
 927
 928         root = mount_subtree(mnt, subvol_name);
 929
 930         if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
 931                 struct super_block *s = root->d_sb;
 932                 dput(root);
 933                 root = ERR_PTR(-EINVAL);
 934                 deactivate_locked_super(s);
 935                 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
 936                                 subvol_name);
 937         }
 938
 939         return root;
 940 }
 941
 942 /*
 943  * Find a superblock for the given device / mount point.
 944  *
 945  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
 946  *        for multiple device setup.  Make sure to keep it in sync.
 947  */
 948 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
 949                 const char *device_name, void *data)
 950 {
 951         struct block_device *bdev = NULL;
 952         struct super_block *s;
 953         struct dentry *root;
 954         struct btrfs_fs_devices *fs_devices = NULL;
 955         struct btrfs_fs_info *fs_info = NULL;
 956         fmode_t mode = FMODE_READ;
 957         char *subvol_name = NULL;
 958         u64 subvol_objectid = 0;
 959         u64 subvol_rootid = 0;
 960         int error = 0;
 961
 962         if (!(flags & MS_RDONLY))
 963                 mode |= FMODE_WRITE;
 964
 965         error = btrfs_parse_early_options(data, mode, fs_type,
 966                                           &subvol_name, &subvol_objectid,
 967                                           &subvol_rootid, &fs_devices);
 968         if (error) {
 969                 kfree(subvol_name);
 970                 return ERR_PTR(error);
 971         }
 972
 973         if (subvol_name) {
 974                 root = mount_subvol(subvol_name, flags, device_name, data);
 975                 kfree(subvol_name);
 976                 return root;
 977         }
 978
 979         error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
 980         if (error)
 981                 return ERR_PTR(error);
 982
 983         /*
 984          * Setup a dummy root and fs_info for test/set super.  This is because
 985          * we don't actually fill this stuff out until open_ctree, but we need
 986          * it for searching for existing supers, so this lets us do that and
 987          * then open_ctree will properly initialize everything later.
 988          */
 989         fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
 990         if (!fs_info)
 991                 return ERR_PTR(-ENOMEM);
 992
 993         fs_info->fs_devices = fs_devices;
 994
 995         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
 996         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
 997         if (!fs_info->super_copy || !fs_info->super_for_commit) {
 998                 error = -ENOMEM;
 999                 goto error_fs_info;
1000         }
1001
1002         error = btrfs_open_devices(fs_devices, mode, fs_type);
1003         if (error)
1004                 goto error_fs_info;
1005
1006         if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1007                 error = -EACCES;
1008                 goto error_close_devices;
1009         }
1010
1011         bdev = fs_devices->latest_bdev;
1012         s = sget(fs_type, btrfs_test_super, btrfs_set_super, fs_info);
1013         if (IS_ERR(s)) {
1014                 error = PTR_ERR(s);
1015                 goto error_close_devices;
1016         }
1017
1018         if (s->s_root) {
1019                 btrfs_close_devices(fs_devices);
1020                 free_fs_info(fs_info);
1021                 if ((flags ^ s->s_flags) & MS_RDONLY)
1022                         error = -EBUSY;
1023         } else {
1024                 char b[BDEVNAME_SIZE];
1025
1026                 s->s_flags = flags | MS_NOSEC;
1027                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1028                 btrfs_sb(s)->bdev_holder = fs_type;
1029                 error = btrfs_fill_super(s, fs_devices, data,
1030                                          flags & MS_SILENT ? 1 : 0);
1031         }
1032
1033         root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1034         if (IS_ERR(root))
1035                 deactivate_locked_super(s);
1036
1037         return root;
1038
1039 error_close_devices:
1040         btrfs_close_devices(fs_devices);
1041 error_fs_info:
1042         free_fs_info(fs_info);
1043         return ERR_PTR(error);
1044 }
1045
1046 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1047 {
1048         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1049         struct btrfs_root *root = fs_info->tree_root;
1050         int ret;
1051
1052         ret = btrfs_parse_options(root, data);
1053         if (ret)
1054                 return -EINVAL;
1055
1056         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1057                 return 0;
1058
1059         if (*flags & MS_RDONLY) {
1060                 sb->s_flags |= MS_RDONLY;
1061
1062                 ret =  btrfs_commit_super(root);
1063                 WARN_ON(ret);
1064         } else {
1065                 if (fs_info->fs_devices->rw_devices == 0)
1066                         return -EACCES;
1067
1068                 if (btrfs_super_log_root(fs_info->super_copy) != 0)
1069                         return -EINVAL;
1070
1071                 ret = btrfs_cleanup_fs_roots(fs_info);
1072                 WARN_ON(ret);
1073
1074                 /* recover relocation */
1075                 ret = btrfs_recover_relocation(root);
1076                 WARN_ON(ret);
1077
1078                 sb->s_flags &= ~MS_RDONLY;
1079         }
1080
1081         return 0;
1082 }
1083
1084 /* Used to sort the devices by max_avail(descending sort) */
1085 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1086                                        const void *dev_info2)
1087 {
1088         if (((struct btrfs_device_info *)dev_info1)->max_avail >
1089             ((struct btrfs_device_info *)dev_info2)->max_avail)
1090                 return -1;
1091         else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1092                  ((struct btrfs_device_info *)dev_info2)->max_avail)
1093                 return 1;
1094         else
1095         return 0;
1096 }
1097
1098 /*
1099  * sort the devices by max_avail, in which max free extent size of each device
1100  * is stored.(Descending Sort)
1101  */
1102 static inline void btrfs_descending_sort_devices(
1103                                         struct btrfs_device_info *devices,
1104                                         size_t nr_devices)
1105 {
1106         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1107              btrfs_cmp_device_free_bytes, NULL);
1108 }
1109
1110 /*
1111  * The helper to calc the free space on the devices that can be used to store
1112  * file data.
1113  */
1114 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1115 {
1116         struct btrfs_fs_info *fs_info = root->fs_info;
1117         struct btrfs_device_info *devices_info;
1118         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1119         struct btrfs_device *device;
1120         u64 skip_space;
1121         u64 type;
1122         u64 avail_space;
1123         u64 used_space;
1124         u64 min_stripe_size;
1125         int min_stripes = 1, num_stripes = 1;
1126         int i = 0, nr_devices;
1127         int ret;
1128
1129         nr_devices = fs_info->fs_devices->open_devices;
1130         BUG_ON(!nr_devices);
1131
1132         devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1133                                GFP_NOFS);
1134         if (!devices_info)
1135                 return -ENOMEM;
1136
1137         /* calc min stripe number for data space alloction */
1138         type = btrfs_get_alloc_profile(root, 1);
1139         if (type & BTRFS_BLOCK_GROUP_RAID0) {
1140                 min_stripes = 2;
1141                 num_stripes = nr_devices;
1142         } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1143                 min_stripes = 2;
1144                 num_stripes = 2;
1145         } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1146                 min_stripes = 4;
1147                 num_stripes = 4;
1148         }
1149
1150         if (type & BTRFS_BLOCK_GROUP_DUP)
1151                 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1152         else
1153                 min_stripe_size = BTRFS_STRIPE_LEN;
1154
1155         list_for_each_entry(device, &fs_devices->devices, dev_list) {
1156                 if (!device->in_fs_metadata || !device->bdev)
1157                         continue;
1158
1159                 avail_space = device->total_bytes - device->bytes_used;
1160
1161                 /* align with stripe_len */
1162                 do_div(avail_space, BTRFS_STRIPE_LEN);
1163                 avail_space *= BTRFS_STRIPE_LEN;
1164
1165                 /*
1166                  * In order to avoid overwritting the superblock on the drive,
1167                  * btrfs starts at an offset of at least 1MB when doing chunk
1168                  * allocation.
1169                  */
1170                 skip_space = 1024 * 1024;
1171
1172                 /* user can set the offset in fs_info->alloc_start. */
1173                 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1174                     device->total_bytes)
1175                         skip_space = max(fs_info->alloc_start, skip_space);
1176
1177                 /*
1178                  * btrfs can not use the free space in [0, skip_space - 1],
1179                  * we must subtract it from the total. In order to implement
1180                  * it, we account the used space in this range first.
1181                  */
1182                 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1183                                                      &used_space);
1184                 if (ret) {
1185                         kfree(devices_info);
1186                         return ret;
1187                 }
1188
1189                 /* calc the free space in [0, skip_space - 1] */
1190                 skip_space -= used_space;
1191
1192                 /*
1193                  * we can use the free space in [0, skip_space - 1], subtract
1194                  * it from the total.
1195                  */
1196                 if (avail_space && avail_space >= skip_space)
1197                         avail_space -= skip_space;
1198                 else
1199                         avail_space = 0;
1200
1201                 if (avail_space < min_stripe_size)
1202                         continue;
1203
1204                 devices_info[i].dev = device;
1205                 devices_info[i].max_avail = avail_space;
1206
1207                 i++;
1208         }
1209
1210         nr_devices = i;
1211
1212         btrfs_descending_sort_devices(devices_info, nr_devices);
1213
1214         i = nr_devices - 1;
1215         avail_space = 0;
1216         while (nr_devices >= min_stripes) {
1217                 if (num_stripes > nr_devices)
1218                         num_stripes = nr_devices;
1219
1220                 if (devices_info[i].max_avail >= min_stripe_size) {
1221                         int j;
1222                         u64 alloc_size;
1223
1224                         avail_space += devices_info[i].max_avail * num_stripes;
1225                         alloc_size = devices_info[i].max_avail;
1226                         for (j = i + 1 - num_stripes; j <= i; j++)
1227                                 devices_info[j].max_avail -= alloc_size;
1228                 }
1229                 i--;
1230                 nr_devices--;
1231         }
1232
1233         kfree(devices_info);
1234         *free_bytes = avail_space;
1235         return 0;
1236 }
1237
1238 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1239 {
1240         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1241         struct btrfs_super_block *disk_super = fs_info->super_copy;
1242         struct list_head *head = &fs_info->space_info;
1243         struct btrfs_space_info *found;
1244         u64 total_used = 0;
1245         u64 total_free_data = 0;
1246         int bits = dentry->d_sb->s_blocksize_bits;
1247         __be32 *fsid = (__be32 *)fs_info->fsid;
1248         int ret;
1249
1250         /* holding chunk_muext to avoid allocating new chunks */
1251         mutex_lock(&fs_info->chunk_mutex);
1252         rcu_read_lock();
1253         list_for_each_entry_rcu(found, head, list) {
1254                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1255                         total_free_data += found->disk_total - found->disk_used;
1256                         total_free_data -=
1257                                 btrfs_account_ro_block_groups_free_space(found);
1258                 }
1259
1260                 total_used += found->disk_used;
1261         }
1262         rcu_read_unlock();
1263
1264         buf->f_namelen = BTRFS_NAME_LEN;
1265         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1266         buf->f_bfree = buf->f_blocks - (total_used >> bits);
1267         buf->f_bsize = dentry->d_sb->s_blocksize;
1268         buf->f_type = BTRFS_SUPER_MAGIC;
1269         buf->f_bavail = total_free_data;
1270         ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1271         if (ret) {
1272                 mutex_unlock(&fs_info->chunk_mutex);
1273                 return ret;
1274         }
1275         buf->f_bavail += total_free_data;
1276         buf->f_bavail = buf->f_bavail >> bits;
1277         mutex_unlock(&fs_info->chunk_mutex);
1278
1279         /* We treat it as constant endianness (it doesn't matter _which_)
1280            because we want the fsid to come out the same whether mounted
1281            on a big-endian or little-endian host */
1282         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1283         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1284         /* Mask in the root object ID too, to disambiguate subvols */
1285         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1286         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1287
1288         return 0;
1289 }
1290
1291 static void btrfs_kill_super(struct super_block *sb)
1292 {
1293         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1294         kill_anon_super(sb);
1295         free_fs_info(fs_info);
1296 }
1297
1298 static struct file_system_type btrfs_fs_type = {
1299         .owner          = THIS_MODULE,
1300         .name           = "btrfs",
1301         .mount          = btrfs_mount,
1302         .kill_sb        = btrfs_kill_super,
1303         .fs_flags       = FS_REQUIRES_DEV,
1304 };
1305
1306 /*
1307  * used by btrfsctl to scan devices when no FS is mounted
1308  */
1309 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1310                                 unsigned long arg)
1311 {
1312         struct btrfs_ioctl_vol_args *vol;
1313         struct btrfs_fs_devices *fs_devices;
1314         int ret = -ENOTTY;
1315
1316         if (!capable(CAP_SYS_ADMIN))
1317                 return -EPERM;
1318
1319         vol = memdup_user((void __user *)arg, sizeof(*vol));
1320         if (IS_ERR(vol))
1321                 return PTR_ERR(vol);
1322
1323         switch (cmd) {
1324         case BTRFS_IOC_SCAN_DEV:
1325                 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1326                                             &btrfs_fs_type, &fs_devices);
1327                 break;
1328         }
1329
1330         kfree(vol);
1331         return ret;
1332 }
1333
1334 static int btrfs_freeze(struct super_block *sb)
1335 {
1336         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1337         mutex_lock(&fs_info->transaction_kthread_mutex);
1338         mutex_lock(&fs_info->cleaner_mutex);
1339         return 0;
1340 }
1341
1342 static int btrfs_unfreeze(struct super_block *sb)
1343 {
1344         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1345         mutex_unlock(&fs_info->cleaner_mutex);
1346         mutex_unlock(&fs_info->transaction_kthread_mutex);
1347         return 0;
1348 }
1349
1350 static void btrfs_fs_dirty_inode(struct inode *inode, int flags)
1351 {
1352         int ret;
1353
1354         ret = btrfs_dirty_inode(inode);
1355         if (ret)
1356                 printk_ratelimited(KERN_ERR "btrfs: fail to dirty inode %Lu "
1357                                    "error %d\n", btrfs_ino(inode), ret);
1358 }
1359
1360 static const struct super_operations btrfs_super_ops = {
1361         .drop_inode     = btrfs_drop_inode,
1362         .evict_inode    = btrfs_evict_inode,
1363         .put_super      = btrfs_put_super,
1364         .sync_fs        = btrfs_sync_fs,
1365         .show_options   = btrfs_show_options,
1366         .write_inode    = btrfs_write_inode,
1367         .dirty_inode    = btrfs_fs_dirty_inode,
1368         .alloc_inode    = btrfs_alloc_inode,
1369         .destroy_inode  = btrfs_destroy_inode,
1370         .statfs         = btrfs_statfs,
1371         .remount_fs     = btrfs_remount,
1372         .freeze_fs      = btrfs_freeze,
1373         .unfreeze_fs    = btrfs_unfreeze,
1374 };
1375
1376 static const struct file_operations btrfs_ctl_fops = {
1377         .unlocked_ioctl  = btrfs_control_ioctl,
1378         .compat_ioctl = btrfs_control_ioctl,
1379         .owner   = THIS_MODULE,
1380         .llseek = noop_llseek,
1381 };
1382
1383 static struct miscdevice btrfs_misc = {
1384         .minor          = BTRFS_MINOR,
1385         .name           = "btrfs-control",
1386         .fops           = &btrfs_ctl_fops
1387 };
1388
1389 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1390 MODULE_ALIAS("devname:btrfs-control");
1391
1392 static int btrfs_interface_init(void)
1393 {
1394         return misc_register(&btrfs_misc);
1395 }
1396
1397 static void btrfs_interface_exit(void)
1398 {
1399         if (misc_deregister(&btrfs_misc) < 0)
1400                 printk(KERN_INFO "misc_deregister failed for control device");
1401 }
1402
1403 static int __init init_btrfs_fs(void)
1404 {
1405         int err;
1406
1407         err = btrfs_init_sysfs();
1408         if (err)
1409                 return err;
1410
1411         err = btrfs_init_compress();
1412         if (err)
1413                 goto free_sysfs;
1414
1415         err = btrfs_init_cachep();
1416         if (err)
1417                 goto free_compress;
1418
1419         err = extent_io_init();
1420         if (err)
1421                 goto free_cachep;
1422
1423         err = extent_map_init();
1424         if (err)
1425                 goto free_extent_io;
1426
1427         err = btrfs_delayed_inode_init();
1428         if (err)
1429                 goto free_extent_map;
1430
1431         err = btrfs_interface_init();
1432         if (err)
1433                 goto free_delayed_inode;
1434
1435         err = register_filesystem(&btrfs_fs_type);
1436         if (err)
1437                 goto unregister_ioctl;
1438
1439         printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1440         return 0;
1441
1442 unregister_ioctl:
1443         btrfs_interface_exit();
1444 free_delayed_inode:
1445         btrfs_delayed_inode_exit();
1446 free_extent_map:
1447         extent_map_exit();
1448 free_extent_io:
1449         extent_io_exit();
1450 free_cachep:
1451         btrfs_destroy_cachep();
1452 free_compress:
1453         btrfs_exit_compress();
1454 free_sysfs:
1455         btrfs_exit_sysfs();
1456         return err;
1457 }
1458
1459 static void __exit exit_btrfs_fs(void)
1460 {
1461         btrfs_destroy_cachep();
1462         btrfs_delayed_inode_exit();
1463         extent_map_exit();
1464         extent_io_exit();
1465         btrfs_interface_exit();
1466         unregister_filesystem(&btrfs_fs_type);
1467         btrfs_exit_sysfs();
1468         btrfs_cleanup_fs_uuids();
1469         btrfs_exit_compress();
1470 }
1471
1472 module_init(init_btrfs_fs)
1473 module_exit(exit_btrfs_fs)
1474
1475 MODULE_LICENSE("GPL");