mkfs.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 #define _XOPEN_SOURCE 500
  20 #define _GNU_SOURCE
  21
  22 #include "kerncompat.h"
  23
  24 #ifndef __CHECKER__
  25 #include <sys/ioctl.h>
  26 #include <sys/mount.h>
  27 #include "ioctl.h"
  28 #endif
  29
  30 #include <stdio.h>
  31 #include <stdlib.h>
  32 #include <sys/types.h>
  33 #include <sys/stat.h>
  34 #include <sys/dir.h>
  35 #include <fcntl.h>
  36 #include <unistd.h>
  37 #include <getopt.h>
  38 #include <uuid/uuid.h>
  39 #include <linux/fs.h>
  40 #include <ctype.h>
  41 #include <attr/xattr.h>
  42 #include <blkid/blkid.h>
  43 #include <ftw.h>
  44 #include "ctree.h"
  45 #include "disk-io.h"
  46 #include "volumes.h"
  47 #include "transaction.h"
  48 #include "utils.h"
  49 #include "version.h"
  50
  51 static u64 index_cnt = 2;
  52
  53 struct directory_name_entry {
  54         char *dir_name;
  55         char *path;
  56         ino_t inum;
  57         struct list_head list;
  58 };
  59
  60 static int make_root_dir(struct btrfs_root *root, int mixed)
  61 {
  62         struct btrfs_trans_handle *trans;
  63         struct btrfs_key location;
  64         u64 bytes_used;
  65         u64 chunk_start = 0;
  66         u64 chunk_size = 0;
  67         int ret;
  68
  69         trans = btrfs_start_transaction(root, 1);
  70         bytes_used = btrfs_super_bytes_used(root->fs_info->super_copy);
  71
  72         root->fs_info->system_allocs = 1;
  73         ret = btrfs_make_block_group(trans, root, bytes_used,
  74                                      BTRFS_BLOCK_GROUP_SYSTEM,
  75                                      BTRFS_FIRST_CHUNK_TREE_OBJECTID,
  76                                      0, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
  77         BUG_ON(ret);
  78
  79         if (mixed) {
  80                 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
  81                                         &chunk_start, &chunk_size,
  82                                         BTRFS_BLOCK_GROUP_METADATA |
  83                                         BTRFS_BLOCK_GROUP_DATA);
  84                 BUG_ON(ret);
  85                 ret = btrfs_make_block_group(trans, root, 0,
  86                                              BTRFS_BLOCK_GROUP_METADATA |
  87                                              BTRFS_BLOCK_GROUP_DATA,
  88                                              BTRFS_FIRST_CHUNK_TREE_OBJECTID,
  89                                              chunk_start, chunk_size);
  90                 BUG_ON(ret);
  91                 printf("Created a data/metadata chunk of size %llu\n", chunk_size);
  92         } else {
  93                 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
  94                                         &chunk_start, &chunk_size,
  95                                         BTRFS_BLOCK_GROUP_METADATA);
  96                 BUG_ON(ret);
  97                 ret = btrfs_make_block_group(trans, root, 0,
  98                                              BTRFS_BLOCK_GROUP_METADATA,
  99                                              BTRFS_FIRST_CHUNK_TREE_OBJECTID,
 100                                              chunk_start, chunk_size);
 101                 BUG_ON(ret);
 102         }
 103
 104         root->fs_info->system_allocs = 0;
 105         btrfs_commit_transaction(trans, root);
 106         trans = btrfs_start_transaction(root, 1);
 107         BUG_ON(!trans);
 108
 109         if (!mixed) {
 110                 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
 111                                         &chunk_start, &chunk_size,
 112                                         BTRFS_BLOCK_GROUP_DATA);
 113                 BUG_ON(ret);
 114                 ret = btrfs_make_block_group(trans, root, 0,
 115                                              BTRFS_BLOCK_GROUP_DATA,
 116                                              BTRFS_FIRST_CHUNK_TREE_OBJECTID,
 117                                              chunk_start, chunk_size);
 118                 BUG_ON(ret);
 119         }
 120
 121         ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
 122                               BTRFS_ROOT_TREE_DIR_OBJECTID);
 123         if (ret)
 124                 goto err;
 125         ret = btrfs_make_root_dir(trans, root, BTRFS_FIRST_FREE_OBJECTID);
 126         if (ret)
 127                 goto err;
 128         memcpy(&location, &root->fs_info->fs_root->root_key, sizeof(location));
 129         location.offset = (u64)-1;
 130         ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
 131                         "default", 7,
 132                         btrfs_super_root_dir(root->fs_info->super_copy),
 133                         &location, BTRFS_FT_DIR, 0);
 134         if (ret)
 135                 goto err;
 136
 137         ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
 138                              "default", 7, location.objectid,
 139                              BTRFS_ROOT_TREE_DIR_OBJECTID, 0);
 140         if (ret)
 141                 goto err;
 142
 143         btrfs_commit_transaction(trans, root);
 144 err:
 145         return ret;
 146 }
 147
 148 static int recow_roots(struct btrfs_trans_handle *trans,
 149                        struct btrfs_root *root)
 150 {
 151         int ret;
 152         struct extent_buffer *tmp;
 153         struct btrfs_fs_info *info = root->fs_info;
 154
 155         ret = __btrfs_cow_block(trans, info->fs_root, info->fs_root->node,
 156                                 NULL, 0, &tmp, 0, 0);
 157         BUG_ON(ret);
 158         free_extent_buffer(tmp);
 159
 160         ret = __btrfs_cow_block(trans, info->tree_root, info->tree_root->node,
 161                                 NULL, 0, &tmp, 0, 0);
 162         BUG_ON(ret);
 163         free_extent_buffer(tmp);
 164
 165         ret = __btrfs_cow_block(trans, info->extent_root,
 166                                 info->extent_root->node, NULL, 0, &tmp, 0, 0);
 167         BUG_ON(ret);
 168         free_extent_buffer(tmp);
 169
 170         ret = __btrfs_cow_block(trans, info->chunk_root, info->chunk_root->node,
 171                                 NULL, 0, &tmp, 0, 0);
 172         BUG_ON(ret);
 173         free_extent_buffer(tmp);
 174
 175
 176         ret = __btrfs_cow_block(trans, info->dev_root, info->dev_root->node,
 177                                 NULL, 0, &tmp, 0, 0);
 178         BUG_ON(ret);
 179         free_extent_buffer(tmp);
 180
 181         ret = __btrfs_cow_block(trans, info->csum_root, info->csum_root->node,
 182                                 NULL, 0, &tmp, 0, 0);
 183         BUG_ON(ret);
 184         free_extent_buffer(tmp);
 185
 186         return 0;
 187 }
 188
 189 static int create_one_raid_group(struct btrfs_trans_handle *trans,
 190                               struct btrfs_root *root, u64 type)
 191 {
 192         u64 chunk_start;
 193         u64 chunk_size;
 194         int ret;
 195
 196         ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
 197                                 &chunk_start, &chunk_size, type);
 198         BUG_ON(ret);
 199         ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
 200                                      type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
 201                                      chunk_start, chunk_size);
 202         BUG_ON(ret);
 203         return ret;
 204 }
 205
 206 static int create_raid_groups(struct btrfs_trans_handle *trans,
 207                               struct btrfs_root *root, u64 data_profile,
 208                               int data_profile_opt, u64 metadata_profile,
 209                               int metadata_profile_opt, int mixed, int ssd)
 210 {
 211         u64 num_devices = btrfs_super_num_devices(root->fs_info->super_copy);
 212         u64 allowed = 0;
 213         u64 devices_for_raid = num_devices;
 214         int ret;
 215
 216         /*
 217          * Set default profiles according to number of added devices.
 218          * For mixed groups defaults are single/single.
 219          */
 220         if (!metadata_profile_opt && !mixed) {
 221                 if (num_devices == 1 && ssd)
 222                         printf("Detected a SSD, turning off metadata "
 223                                "duplication.  Mkfs with -m dup if you want to "
 224                                "force metadata duplication.\n");
 225                 metadata_profile = (num_devices > 1) ?
 226                         BTRFS_BLOCK_GROUP_RAID1 : (ssd) ? 0: BTRFS_BLOCK_GROUP_DUP;
 227         }
 228         if (!data_profile_opt && !mixed) {
 229                 data_profile = (num_devices > 1) ?
 230                         BTRFS_BLOCK_GROUP_RAID0 : 0; /* raid0 or single */
 231         }
 232
 233         if (devices_for_raid > 4)
 234                 devices_for_raid = 4;
 235
 236         switch (devices_for_raid) {
 237         default:
 238         case 4:
 239                 allowed |= BTRFS_BLOCK_GROUP_RAID10;
 240         case 3:
 241                 allowed |= BTRFS_BLOCK_GROUP_RAID6;
 242         case 2:
 243                 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
 244                         BTRFS_BLOCK_GROUP_RAID5;
 245                 break;
 246         case 1:
 247                 allowed |= BTRFS_BLOCK_GROUP_DUP;
 248         }
 249
 250         if (metadata_profile & ~allowed) {
 251                 fprintf(stderr, "unable to create FS with metadata "
 252                         "profile %llu (have %llu devices)\n", metadata_profile,
 253                         num_devices);
 254                 exit(1);
 255         }
 256         if (data_profile & ~allowed) {
 257                 fprintf(stderr, "unable to create FS with data "
 258                         "profile %llu (have %llu devices)\n", data_profile,
 259                         num_devices);
 260                 exit(1);
 261         }
 262
 263         /* allow dup'ed data chunks only in mixed mode */
 264         if (!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP)) {
 265                 fprintf(stderr, "dup for data is allowed only in mixed mode\n");
 266                 exit(1);
 267         }
 268
 269         if (allowed & metadata_profile) {
 270                 u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
 271
 272                 ret = create_one_raid_group(trans, root,
 273                                             BTRFS_BLOCK_GROUP_SYSTEM |
 274                                             (allowed & metadata_profile));
 275                 BUG_ON(ret);
 276
 277                 if (mixed)
 278                         meta_flags |= BTRFS_BLOCK_GROUP_DATA;
 279
 280                 ret = create_one_raid_group(trans, root, meta_flags |
 281                                             (allowed & metadata_profile));
 282                 BUG_ON(ret);
 283
 284                 ret = recow_roots(trans, root);
 285                 BUG_ON(ret);
 286         }
 287         if (!mixed && num_devices > 1 && (allowed & data_profile)) {
 288                 ret = create_one_raid_group(trans, root,
 289                                             BTRFS_BLOCK_GROUP_DATA |
 290                                             (allowed & data_profile));
 291                 BUG_ON(ret);
 292         }
 293         return 0;
 294 }
 295
 296 static int create_data_reloc_tree(struct btrfs_trans_handle *trans,
 297                                   struct btrfs_root *root)
 298 {
 299         struct btrfs_key location;
 300         struct btrfs_root_item root_item;
 301         struct extent_buffer *tmp;
 302         u64 objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
 303         int ret;
 304
 305         ret = btrfs_copy_root(trans, root, root->node, &tmp, objectid);
 306         BUG_ON(ret);
 307
 308         memcpy(&root_item, &root->root_item, sizeof(root_item));
 309         btrfs_set_root_bytenr(&root_item, tmp->start);
 310         btrfs_set_root_level(&root_item, btrfs_header_level(tmp));
 311         btrfs_set_root_generation(&root_item, trans->transid);
 312         free_extent_buffer(tmp);
 313
 314         location.objectid = objectid;
 315         location.type = BTRFS_ROOT_ITEM_KEY;
 316         location.offset = 0;
 317         ret = btrfs_insert_root(trans, root->fs_info->tree_root,
 318                                 &location, &root_item);
 319         BUG_ON(ret);
 320         return 0;
 321 }
 322
 323 static void print_usage(void)
 324 {
 325         fprintf(stderr, "usage: mkfs.btrfs [options] dev [ dev ... ]\n");
 326         fprintf(stderr, "options:\n");
 327         fprintf(stderr, "\t -A --alloc-start the offset to start the FS\n");
 328         fprintf(stderr, "\t -b --byte-count total number of bytes in the FS\n");
 329         fprintf(stderr, "\t -d --data data profile, raid0, raid1, raid5, raid6, raid10, dup or single\n");
 330         fprintf(stderr, "\t -f --force force overwrite of existing filesystem\n");
 331         fprintf(stderr, "\t -l --leafsize size of btree leaves\n");
 332         fprintf(stderr, "\t -L --label set a label\n");
 333         fprintf(stderr, "\t -m --metadata metadata profile, values like data profile\n");
 334         fprintf(stderr, "\t -M --mixed mix metadata and data together\n");
 335         fprintf(stderr, "\t -n --nodesize size of btree nodes\n");
 336         fprintf(stderr, "\t -s --sectorsize min block allocation\n");
 337         fprintf(stderr, "\t -r --rootdir the source directory\n");
 338         fprintf(stderr, "\t -K --nodiscard do not perform whole device TRIM\n");
 339         fprintf(stderr, "\t -V --version print the mkfs.btrfs version and exit\n");
 340         fprintf(stderr, "%s\n", BTRFS_BUILD_VERSION);
 341         exit(1);
 342 }
 343
 344 static void print_version(void)
 345 {
 346         fprintf(stderr, "mkfs.btrfs, part of %s\n", BTRFS_BUILD_VERSION);
 347         exit(0);
 348 }
 349
 350 static u64 parse_profile(char *s)
 351 {
 352         if (strcmp(s, "raid0") == 0) {
 353                 return BTRFS_BLOCK_GROUP_RAID0;
 354         } else if (strcmp(s, "raid1") == 0) {
 355                 return BTRFS_BLOCK_GROUP_RAID1;
 356         } else if (strcmp(s, "raid5") == 0) {
 357                 return BTRFS_BLOCK_GROUP_RAID5;
 358         } else if (strcmp(s, "raid6") == 0) {
 359                 return BTRFS_BLOCK_GROUP_RAID6;
 360         } else if (strcmp(s, "raid10") == 0) {
 361                 return BTRFS_BLOCK_GROUP_RAID10;
 362         } else if (strcmp(s, "dup") == 0) {
 363                 return BTRFS_BLOCK_GROUP_DUP;
 364         } else if (strcmp(s, "single") == 0) {
 365                 return 0;
 366         } else {
 367                 fprintf(stderr, "Unknown profile %s\n", s);
 368                 print_usage();
 369         }
 370         /* not reached */
 371         return 0;
 372 }
 373
 374 static char *parse_label(char *input)
 375 {
 376         int len = strlen(input);
 377
 378         if (len >= BTRFS_LABEL_SIZE) {
 379                 fprintf(stderr, "Label %s is too long (max %d)\n", input,
 380                         BTRFS_LABEL_SIZE - 1);
 381                 exit(1);
 382         }
 383         return strdup(input);
 384 }
 385
 386 static struct option long_options[] = {
 387         { "alloc-start", 1, NULL, 'A'},
 388         { "byte-count", 1, NULL, 'b' },
 389         { "force", 0, NULL, 'f' },
 390         { "leafsize", 1, NULL, 'l' },
 391         { "label", 1, NULL, 'L'},
 392         { "metadata", 1, NULL, 'm' },
 393         { "mixed", 0, NULL, 'M' },
 394         { "nodesize", 1, NULL, 'n' },
 395         { "sectorsize", 1, NULL, 's' },
 396         { "data", 1, NULL, 'd' },
 397         { "version", 0, NULL, 'V' },
 398         { "rootdir", 1, NULL, 'r' },
 399         { "nodiscard", 0, NULL, 'K' },
 400         { 0, 0, 0, 0}
 401 };
 402
 403 static int add_directory_items(struct btrfs_trans_handle *trans,
 404                                struct btrfs_root *root, u64 objectid,
 405                                ino_t parent_inum, const char *name,
 406                                struct stat *st, int *dir_index_cnt)
 407 {
 408         int ret;
 409         int name_len;
 410         struct btrfs_key location;
 411         u8 filetype = 0;
 412
 413         name_len = strlen(name);
 414
 415         location.objectid = objectid;
 416         location.offset = 0;
 417         btrfs_set_key_type(&location, BTRFS_INODE_ITEM_KEY);
 418
 419         if (S_ISDIR(st->st_mode))
 420                 filetype = BTRFS_FT_DIR;
 421         if (S_ISREG(st->st_mode))
 422                 filetype = BTRFS_FT_REG_FILE;
 423         if (S_ISLNK(st->st_mode))
 424                 filetype = BTRFS_FT_SYMLINK;
 425
 426         ret = btrfs_insert_dir_item(trans, root, name, name_len,
 427                                     parent_inum, &location,
 428                                     filetype, index_cnt);
 429
 430         *dir_index_cnt = index_cnt;
 431         index_cnt++;
 432
 433         return ret;
 434 }
 435
 436 static int fill_inode_item(struct btrfs_trans_handle *trans,
 437                            struct btrfs_root *root,
 438                            struct btrfs_inode_item *dst, struct stat *src)
 439 {
 440         u64 blocks = 0;
 441         u64 sectorsize = root->sectorsize;
 442
 443         /*
 444          * btrfs_inode_item has some reserved fields
 445          * and represents on-disk inode entry, so
 446          * zero everything to prevent information leak
 447          */
 448         memset(dst, 0, sizeof (*dst));
 449
 450         btrfs_set_stack_inode_generation(dst, trans->transid);
 451         btrfs_set_stack_inode_size(dst, src->st_size);
 452         btrfs_set_stack_inode_nbytes(dst, 0);
 453         btrfs_set_stack_inode_block_group(dst, 0);
 454         btrfs_set_stack_inode_nlink(dst, src->st_nlink);
 455         btrfs_set_stack_inode_uid(dst, src->st_uid);
 456         btrfs_set_stack_inode_gid(dst, src->st_gid);
 457         btrfs_set_stack_inode_mode(dst, src->st_mode);
 458         btrfs_set_stack_inode_rdev(dst, 0);
 459         btrfs_set_stack_inode_flags(dst, 0);
 460         btrfs_set_stack_timespec_sec(&dst->atime, src->st_atime);
 461         btrfs_set_stack_timespec_nsec(&dst->atime, 0);
 462         btrfs_set_stack_timespec_sec(&dst->ctime, src->st_ctime);
 463         btrfs_set_stack_timespec_nsec(&dst->ctime, 0);
 464         btrfs_set_stack_timespec_sec(&dst->mtime, src->st_mtime);
 465         btrfs_set_stack_timespec_nsec(&dst->mtime, 0);
 466         btrfs_set_stack_timespec_sec(&dst->otime, 0);
 467         btrfs_set_stack_timespec_nsec(&dst->otime, 0);
 468
 469         if (S_ISDIR(src->st_mode)) {
 470                 btrfs_set_stack_inode_size(dst, 0);
 471                 btrfs_set_stack_inode_nlink(dst, 1);
 472         }
 473         if (S_ISREG(src->st_mode)) {
 474                 btrfs_set_stack_inode_size(dst, (u64)src->st_size);
 475                 if (src->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root))
 476                         btrfs_set_stack_inode_nbytes(dst, src->st_size);
 477                 else {
 478                         blocks = src->st_size / sectorsize;
 479                         if (src->st_size % sectorsize)
 480                                 blocks += 1;
 481                         blocks *= sectorsize;
 482                         btrfs_set_stack_inode_nbytes(dst, blocks);
 483                 }
 484         }
 485         if (S_ISLNK(src->st_mode))
 486                 btrfs_set_stack_inode_nbytes(dst, src->st_size + 1);
 487
 488         return 0;
 489 }
 490
 491 static int directory_select(const struct direct *entry)
 492 {
 493         if ((strncmp(entry->d_name, ".", entry->d_reclen) == 0) ||
 494                 (strncmp(entry->d_name, "..", entry->d_reclen) == 0))
 495                 return 0;
 496         else
 497                 return 1;
 498 }
 499
 500 static void free_namelist(struct direct **files, int count)
 501 {
 502         int i;
 503
 504         if (count < 0)
 505                 return;
 506
 507         for (i = 0; i < count; ++i)
 508                 free(files[i]);
 509         free(files);
 510 }
 511
 512 static u64 calculate_dir_inode_size(char *dirname)
 513 {
 514         int count, i;
 515         struct direct **files, *cur_file;
 516         u64 dir_inode_size = 0;
 517
 518         count = scandir(dirname, &files, directory_select, NULL);
 519
 520         for (i = 0; i < count; i++) {
 521                 cur_file = files[i];
 522                 dir_inode_size += strlen(cur_file->d_name);
 523         }
 524
 525         free_namelist(files, count);
 526
 527         dir_inode_size *= 2;
 528         return dir_inode_size;
 529 }
 530
 531 static int add_inode_items(struct btrfs_trans_handle *trans,
 532                            struct btrfs_root *root,
 533                            struct stat *st, char *name,
 534                            u64 self_objectid, ino_t parent_inum,
 535                            int dir_index_cnt, struct btrfs_inode_item *inode_ret)
 536 {
 537         int ret;
 538         struct btrfs_key inode_key;
 539         struct btrfs_inode_item btrfs_inode;
 540         u64 objectid;
 541         u64 inode_size = 0;
 542         int name_len;
 543
 544         name_len = strlen(name);
 545         fill_inode_item(trans, root, &btrfs_inode, st);
 546         objectid = self_objectid;
 547
 548         if (S_ISDIR(st->st_mode)) {
 549                 inode_size = calculate_dir_inode_size(name);
 550                 btrfs_set_stack_inode_size(&btrfs_inode, inode_size);
 551         }
 552
 553         inode_key.objectid = objectid;
 554         inode_key.offset = 0;
 555         btrfs_set_key_type(&inode_key, BTRFS_INODE_ITEM_KEY);
 556
 557         ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
 558         if (ret)
 559                 goto fail;
 560
 561         ret = btrfs_insert_inode_ref(trans, root, name, name_len,
 562                                      objectid, parent_inum, dir_index_cnt);
 563         if (ret)
 564                 goto fail;
 565
 566         *inode_ret = btrfs_inode;
 567 fail:
 568         return ret;
 569 }
 570
 571 static int add_xattr_item(struct btrfs_trans_handle *trans,
 572                           struct btrfs_root *root, u64 objectid,
 573                           const char *file_name)
 574 {
 575         int ret;
 576         int cur_name_len;
 577         char xattr_list[XATTR_LIST_MAX];
 578         char *cur_name;
 579         char cur_value[XATTR_SIZE_MAX];
 580         char delimiter = '\0';
 581         char *next_location = xattr_list;
 582
 583         ret = llistxattr(file_name, xattr_list, XATTR_LIST_MAX);
 584         if (ret < 0) {
 585                 if(errno == ENOTSUP)
 586                         return 0;
 587                 fprintf(stderr, "get a list of xattr failed for %s\n",
 588                         file_name);
 589                 return ret;
 590         }
 591         if (ret == 0)
 592                 return ret;
 593
 594         cur_name = strtok(xattr_list, &delimiter);
 595         while (cur_name != NULL) {
 596                 cur_name_len = strlen(cur_name);
 597                 next_location += cur_name_len + 1;
 598
 599                 ret = getxattr(file_name, cur_name, cur_value, XATTR_SIZE_MAX);
 600                 if (ret < 0) {
 601                         if(errno == ENOTSUP)
 602                                 return 0;
 603                         fprintf(stderr, "get a xattr value failed for %s attr %s\n",
 604                                 file_name, cur_name);
 605                         return ret;
 606                 }
 607
 608                 ret = btrfs_insert_xattr_item(trans, root, cur_name,
 609                                               cur_name_len, cur_value,
 610                                               ret, objectid);
 611                 if (ret) {
 612                         fprintf(stderr, "insert a xattr item failed for %s\n",
 613                                 file_name);
 614                 }
 615
 616                 cur_name = strtok(next_location, &delimiter);
 617         }
 618
 619         return ret;
 620 }
 621 static int custom_alloc_extent(struct btrfs_root *root, u64 num_bytes,
 622                                u64 hint_byte, struct btrfs_key *ins)
 623 {
 624         u64 start;
 625         u64 end;
 626         u64 last = hint_byte;
 627         int ret;
 628         int wrapped = 0;
 629         struct btrfs_block_group_cache *cache;
 630
 631         while (1) {
 632                 ret = find_first_extent_bit(&root->fs_info->free_space_cache,
 633                                             last, &start, &end, EXTENT_DIRTY);
 634                 if (ret) {
 635                         if (wrapped++ == 0) {
 636                                 last = 0;
 637                                 continue;
 638                         } else {
 639                                 goto fail;
 640                         }
 641                 }
 642
 643                 start = max(last, start);
 644                 last = end + 1;
 645                 if (last - start < num_bytes)
 646                         continue;
 647
 648                 last = start + num_bytes;
 649                 if (test_range_bit(&root->fs_info->pinned_extents,
 650                                    start, last - 1, EXTENT_DIRTY, 0))
 651                         continue;
 652
 653                 cache = btrfs_lookup_block_group(root->fs_info, start);
 654                 BUG_ON(!cache);
 655                 if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM ||
 656                     last > cache->key.objectid + cache->key.offset) {
 657                         last = cache->key.objectid + cache->key.offset;
 658                         continue;
 659                 }
 660
 661                 if (cache->flags & (BTRFS_BLOCK_GROUP_SYSTEM |
 662                             BTRFS_BLOCK_GROUP_METADATA)) {
 663                         last = cache->key.objectid + cache->key.offset;
 664                         continue;
 665                 }
 666
 667                 clear_extent_dirty(&root->fs_info->free_space_cache,
 668                                    start, start + num_bytes - 1, 0);
 669
 670                 ins->objectid = start;
 671                 ins->offset = num_bytes;
 672                 ins->type = BTRFS_EXTENT_ITEM_KEY;
 673                 return 0;
 674         }
 675 fail:
 676         fprintf(stderr, "not enough free space\n");
 677         return -ENOSPC;
 678 }
 679
 680 static int record_file_extent(struct btrfs_trans_handle *trans,
 681                               struct btrfs_root *root, u64 objectid,
 682                               struct btrfs_inode_item *inode,
 683                               u64 file_pos, u64 disk_bytenr,
 684                               u64 num_bytes)
 685 {
 686         int ret;
 687         struct btrfs_fs_info *info = root->fs_info;
 688         struct btrfs_root *extent_root = info->extent_root;
 689         struct extent_buffer *leaf;
 690         struct btrfs_file_extent_item *fi;
 691         struct btrfs_key ins_key;
 692         struct btrfs_path path;
 693         struct btrfs_extent_item *ei;
 694
 695         btrfs_init_path(&path);
 696
 697         ins_key.objectid = objectid;
 698         ins_key.offset = 0;
 699         btrfs_set_key_type(&ins_key, BTRFS_EXTENT_DATA_KEY);
 700         ret = btrfs_insert_empty_item(trans, root, &path, &ins_key,
 701                                       sizeof(*fi));
 702         if (ret)
 703                 goto fail;
 704         leaf = path.nodes[0];
 705         fi = btrfs_item_ptr(leaf, path.slots[0],
 706                             struct btrfs_file_extent_item);
 707         btrfs_set_file_extent_generation(leaf, fi, trans->transid);
 708         btrfs_set_file_extent_type(leaf, fi, BTRFS_FILE_EXTENT_REG);
 709         btrfs_set_file_extent_disk_bytenr(leaf, fi, disk_bytenr);
 710         btrfs_set_file_extent_disk_num_bytes(leaf, fi, num_bytes);
 711         btrfs_set_file_extent_offset(leaf, fi, 0);
 712         btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
 713         btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
 714         btrfs_set_file_extent_compression(leaf, fi, 0);
 715         btrfs_set_file_extent_encryption(leaf, fi, 0);
 716         btrfs_set_file_extent_other_encoding(leaf, fi, 0);
 717         btrfs_mark_buffer_dirty(leaf);
 718
 719         btrfs_release_path(root, &path);
 720
 721         ins_key.objectid = disk_bytenr;
 722         ins_key.offset = num_bytes;
 723         ins_key.type = BTRFS_EXTENT_ITEM_KEY;
 724
 725         ret = btrfs_insert_empty_item(trans, extent_root, &path,
 726                                 &ins_key, sizeof(*ei));
 727         if (ret == 0) {
 728                 leaf = path.nodes[0];
 729                 ei = btrfs_item_ptr(leaf, path.slots[0],
 730                                     struct btrfs_extent_item);
 731
 732                 btrfs_set_extent_refs(leaf, ei, 0);
 733                 btrfs_set_extent_generation(leaf, ei, trans->transid);
 734                 btrfs_set_extent_flags(leaf, ei, BTRFS_EXTENT_FLAG_DATA);
 735
 736                 btrfs_mark_buffer_dirty(leaf);
 737                 ret = btrfs_update_block_group(trans, root, disk_bytenr,
 738                                                num_bytes, 1, 0);
 739                 if (ret)
 740                         goto fail;
 741         } else if (ret != -EEXIST) {
 742                 goto fail;
 743         }
 744
 745         ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, num_bytes, 0,
 746                                    root->root_key.objectid,
 747                                    objectid, 0);
 748 fail:
 749         btrfs_release_path(root, &path);
 750         return ret;
 751 }
 752
 753 static int add_symbolic_link(struct btrfs_trans_handle *trans,
 754                              struct btrfs_root *root,
 755                              u64 objectid, const char *path_name)
 756 {
 757         int ret;
 758         u64 sectorsize = root->sectorsize;
 759         char *buf = malloc(sectorsize);
 760
 761         ret = readlink(path_name, buf, sectorsize);
 762         if (ret <= 0) {
 763                 fprintf(stderr, "readlink failed for %s\n", path_name);
 764                 goto fail;
 765         }
 766         if (ret >= sectorsize) {
 767                 fprintf(stderr, "symlink too long for %s", path_name);
 768                 ret = -1;
 769                 goto fail;
 770         }
 771
 772         buf[ret] = '\0'; /* readlink does not do it for us */
 773         ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
 774                                          buf, ret + 1);
 775 fail:
 776         free(buf);
 777         return ret;
 778 }
 779
 780 static int add_file_items(struct btrfs_trans_handle *trans,
 781                           struct btrfs_root *root,
 782                           struct btrfs_inode_item *btrfs_inode, u64 objectid,
 783                           ino_t parent_inum, struct stat *st,
 784                           const char *path_name, int out_fd)
 785 {
 786         int ret = -1;
 787         ssize_t ret_read;
 788         u64 bytes_read = 0;
 789         char *buffer = NULL;
 790         struct btrfs_key key;
 791         int blocks;
 792         u32 sectorsize = root->sectorsize;
 793         u64 first_block = 0;
 794         u64 num_blocks = 0;
 795         int fd;
 796
 797         fd = open(path_name, O_RDONLY);
 798         if (fd == -1) {
 799                 fprintf(stderr, "%s open failed\n", path_name);
 800                 return ret;
 801         }
 802
 803         blocks = st->st_size / sectorsize;
 804         if (st->st_size % sectorsize)
 805                 blocks += 1;
 806
 807         if (st->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
 808                 buffer = malloc(st->st_size);
 809                 ret_read = pread64(fd, buffer, st->st_size, bytes_read);
 810                 if (ret_read == -1) {
 811                         fprintf(stderr, "%s read failed\n", path_name);
 812                         goto end;
 813                 }
 814
 815                 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
 816                                                  buffer, st->st_size);
 817                 goto end;
 818         }
 819
 820         ret = custom_alloc_extent(root, blocks * sectorsize, 0, &key);
 821         if (ret)
 822                 goto end;
 823
 824         first_block = key.objectid;
 825         bytes_read = 0;
 826         buffer = malloc(sectorsize);
 827
 828         do {
 829                 memset(buffer, 0, sectorsize);
 830                 ret_read = pread64(fd, buffer, sectorsize, bytes_read);
 831                 if (ret_read == -1) {
 832                         fprintf(stderr, "%s read failed\n", path_name);
 833                         goto end;
 834                 }
 835
 836                 ret = pwrite64(out_fd, buffer, sectorsize,
 837                                first_block + bytes_read);
 838                 if (ret != sectorsize) {
 839                         fprintf(stderr, "output file write failed\n");
 840                         goto end;
 841                 }
 842
 843                 /* checksum for file data */
 844                 ret = btrfs_csum_file_block(trans, root->fs_info->csum_root,
 845                                 first_block + (blocks * sectorsize),
 846                                 first_block + bytes_read,
 847                                 buffer, sectorsize);
 848                 if (ret) {
 849                         fprintf(stderr, "%s checksum failed\n", path_name);
 850                         goto end;
 851                 }
 852
 853                 bytes_read += ret_read;
 854                 num_blocks++;
 855         } while (ret_read == sectorsize);
 856
 857         if (num_blocks > 0) {
 858                 ret = record_file_extent(trans, root, objectid, btrfs_inode,
 859                                          first_block, first_block,
 860                                          blocks * sectorsize);
 861                 if (ret)
 862                         goto end;
 863         }
 864
 865 end:
 866         if (buffer)
 867                 free(buffer);
 868         close(fd);
 869         return ret;
 870 }
 871
 872 static char *make_path(char *dir, char *name)
 873 {
 874         char *path;
 875
 876         path = malloc(strlen(dir) + strlen(name) + 2);
 877         if (!path)
 878                 return NULL;
 879         strcpy(path, dir);
 880         if (dir[strlen(dir) - 1] != '/')
 881                 strcat(path, "/");
 882         strcat(path, name);
 883         return path;
 884 }
 885
 886 static int traverse_directory(struct btrfs_trans_handle *trans,
 887                               struct btrfs_root *root, char *dir_name,
 888                               struct directory_name_entry *dir_head, int out_fd)
 889 {
 890         int ret = 0;
 891
 892         struct btrfs_inode_item cur_inode;
 893         struct btrfs_inode_item *inode_item;
 894         int count, i, dir_index_cnt;
 895         struct direct **files;
 896         struct stat st;
 897         struct directory_name_entry *dir_entry, *parent_dir_entry;
 898         struct direct *cur_file;
 899         ino_t parent_inum, cur_inum;
 900         ino_t highest_inum = 0;
 901         char *parent_dir_name;
 902         struct btrfs_path path;
 903         struct extent_buffer *leaf;
 904         struct btrfs_key root_dir_key;
 905         u64 root_dir_inode_size = 0;
 906
 907         /* Add list for source directory */
 908         dir_entry = malloc(sizeof(struct directory_name_entry));
 909         dir_entry->dir_name = dir_name;
 910         dir_entry->path = strdup(dir_name);
 911
 912         parent_inum = highest_inum + BTRFS_FIRST_FREE_OBJECTID;
 913         dir_entry->inum = parent_inum;
 914         list_add_tail(&dir_entry->list, &dir_head->list);
 915
 916         btrfs_init_path(&path);
 917
 918         root_dir_key.objectid = btrfs_root_dirid(&root->root_item);
 919         root_dir_key.offset = 0;
 920         btrfs_set_key_type(&root_dir_key, BTRFS_INODE_ITEM_KEY);
 921         ret = btrfs_lookup_inode(trans, root, &path, &root_dir_key, 1);
 922         if (ret) {
 923                 fprintf(stderr, "root dir lookup error\n");
 924                 return -1;
 925         }
 926
 927         leaf = path.nodes[0];
 928         inode_item = btrfs_item_ptr(leaf, path.slots[0],
 929                                     struct btrfs_inode_item);
 930
 931         root_dir_inode_size = calculate_dir_inode_size(dir_name);
 932         btrfs_set_inode_size(leaf, inode_item, root_dir_inode_size);
 933         btrfs_mark_buffer_dirty(leaf);
 934
 935         btrfs_release_path(root, &path);
 936
 937         do {
 938                 parent_dir_entry = list_entry(dir_head->list.next,
 939                                               struct directory_name_entry,
 940                                               list);
 941                 list_del(&parent_dir_entry->list);
 942
 943                 parent_inum = parent_dir_entry->inum;
 944                 parent_dir_name = parent_dir_entry->dir_name;
 945                 if (chdir(parent_dir_entry->path)) {
 946                         fprintf(stderr, "chdir error for %s\n",
 947                                 parent_dir_name);
 948                         goto fail_no_files;
 949                 }
 950
 951                 count = scandir(parent_dir_entry->path, &files,
 952                                 directory_select, NULL);
 953                 if (count == -1)
 954                 {
 955                         fprintf(stderr, "scandir for %s failed: %s\n",
 956                                 parent_dir_name, strerror (errno));
 957                         goto fail;
 958                 }
 959
 960                 for (i = 0; i < count; i++) {
 961                         cur_file = files[i];
 962
 963                         if (lstat(cur_file->d_name, &st) == -1) {
 964                                 fprintf(stderr, "lstat failed for file %s\n",
 965                                         cur_file->d_name);
 966                                 goto fail;
 967                         }
 968
 969                         cur_inum = ++highest_inum + BTRFS_FIRST_FREE_OBJECTID;
 970                         ret = add_directory_items(trans, root,
 971                                                   cur_inum, parent_inum,
 972                                                   cur_file->d_name,
 973                                                   &st, &dir_index_cnt);
 974                         if (ret) {
 975                                 fprintf(stderr, "add_directory_items failed\n");
 976                                 goto fail;
 977                         }
 978
 979                         ret = add_inode_items(trans, root, &st,
 980                                               cur_file->d_name, cur_inum,
 981                                               parent_inum, dir_index_cnt,
 982                                               &cur_inode);
 983                         if (ret) {
 984                                 fprintf(stderr, "add_inode_items failed\n");
 985                                 goto fail;
 986                         }
 987
 988                         ret = add_xattr_item(trans, root,
 989                                              cur_inum, cur_file->d_name);
 990                         if (ret) {
 991                                 fprintf(stderr, "add_xattr_item failed\n");
 992                                 if(ret != -ENOTSUP)
 993                                         goto fail;
 994                         }
 995
 996                         if (S_ISDIR(st.st_mode)) {
 997                                 dir_entry = malloc(sizeof(struct directory_name_entry));
 998                                 dir_entry->dir_name = cur_file->d_name;
 999                                 dir_entry->path = make_path(parent_dir_entry->path,
1000                                                             cur_file->d_name);
1001                                 dir_entry->inum = cur_inum;
1002                                 list_add_tail(&dir_entry->list, &dir_head->list);
1003                         } else if (S_ISREG(st.st_mode)) {
1004                                 ret = add_file_items(trans, root, &cur_inode,
1005                                                      cur_inum, parent_inum, &st,
1006                                                      cur_file->d_name, out_fd);
1007                                 if (ret) {
1008                                         fprintf(stderr, "add_file_items failed\n");
1009                                         goto fail;
1010                                 }
1011                         } else if (S_ISLNK(st.st_mode)) {
1012                                 ret = add_symbolic_link(trans, root,
1013                                                         cur_inum, cur_file->d_name);
1014                                 if (ret) {
1015                                         fprintf(stderr, "add_symbolic_link failed\n");
1016                                         goto fail;
1017                                 }
1018                         }
1019                 }
1020
1021                 free_namelist(files, count);
1022                 free(parent_dir_entry->path);
1023                 free(parent_dir_entry);
1024
1025                 index_cnt = 2;
1026
1027         } while (!list_empty(&dir_head->list));
1028
1029         return 0;
1030 fail:
1031         free_namelist(files, count);
1032 fail_no_files:
1033         free(parent_dir_entry->path);
1034         free(parent_dir_entry);
1035         return -1;
1036 }
1037
1038 static int open_target(char *output_name)
1039 {
1040         int output_fd;
1041         output_fd = open(output_name, O_CREAT | O_RDWR | O_TRUNC,
1042                          S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH);
1043
1044         return output_fd;
1045 }
1046
1047 static int create_chunks(struct btrfs_trans_handle *trans,
1048                          struct btrfs_root *root, u64 num_of_meta_chunks,
1049                          u64 size_of_data)
1050 {
1051         u64 chunk_start;
1052         u64 chunk_size;
1053         u64 meta_type = BTRFS_BLOCK_GROUP_METADATA;
1054         u64 data_type = BTRFS_BLOCK_GROUP_DATA;
1055         u64 minimum_data_chunk_size = 8 * 1024 * 1024;
1056         u64 i;
1057         int ret;
1058
1059         for (i = 0; i < num_of_meta_chunks; i++) {
1060                 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
1061                                         &chunk_start, &chunk_size, meta_type);
1062                 BUG_ON(ret);
1063                 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
1064                                              meta_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
1065                                              chunk_start, chunk_size);
1066                 BUG_ON(ret);
1067                 set_extent_dirty(&root->fs_info->free_space_cache,
1068                                  chunk_start, chunk_start + chunk_size - 1, 0);
1069         }
1070
1071         if (size_of_data < minimum_data_chunk_size)
1072                 size_of_data = minimum_data_chunk_size;
1073         ret = btrfs_alloc_data_chunk(trans, root->fs_info->extent_root,
1074                                      &chunk_start, size_of_data, data_type);
1075         BUG_ON(ret);
1076         ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
1077                                      data_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
1078                                      chunk_start, size_of_data);
1079         BUG_ON(ret);
1080         set_extent_dirty(&root->fs_info->free_space_cache,
1081                          chunk_start, chunk_start + size_of_data - 1, 0);
1082         return ret;
1083 }
1084
1085 static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
1086 {
1087         int ret;
1088         struct btrfs_trans_handle *trans;
1089
1090         struct stat root_st;
1091
1092         struct directory_name_entry dir_head;
1093
1094         ret = lstat(source_dir, &root_st);
1095         if (ret) {
1096                 fprintf(stderr, "unable to lstat the %s\n", source_dir);
1097                 goto fail;
1098         }
1099
1100         INIT_LIST_HEAD(&dir_head.list);
1101
1102         trans = btrfs_start_transaction(root, 1);
1103         ret = traverse_directory(trans, root, source_dir, &dir_head, out_fd);
1104         if (ret) {
1105                 fprintf(stderr, "unable to traverse_directory\n");
1106                 goto fail;
1107         }
1108         btrfs_commit_transaction(trans, root);
1109
1110         printf("Making image is completed.\n");
1111         return 0;
1112 fail:
1113         fprintf(stderr, "Making image is aborted.\n");
1114         return -1;
1115 }
1116
1117 /*
1118  * This ignores symlinks with unreadable targets and subdirs that can't
1119  * be read.  It's a best-effort to give a rough estimate of the size of
1120  * a subdir.  It doesn't guarantee that prepopulating btrfs from this
1121  * tree won't still run out of space.
1122  *
1123  * The rounding up to 4096 is questionable.  Previous code used du -B 4096.
1124  */
1125 static u64 global_total_size;
1126 static int ftw_add_entry_size(const char *fpath, const struct stat *st,
1127                               int type)
1128 {
1129         if (type == FTW_F || type == FTW_D)
1130                 global_total_size += round_up(st->st_size, 4096);
1131
1132         return 0;
1133 }
1134
1135 static u64 size_sourcedir(char *dir_name, u64 sectorsize,
1136                           u64 *num_of_meta_chunks_ret, u64 *size_of_data_ret)
1137 {
1138         u64 dir_size = 0;
1139         u64 total_size = 0;
1140         int ret;
1141         u64 default_chunk_size = 8 * 1024 * 1024;       /* 8MB */
1142         u64 allocated_meta_size = 8 * 1024 * 1024;      /* 8MB */
1143         u64 allocated_total_size = 20 * 1024 * 1024;    /* 20MB */
1144         u64 num_of_meta_chunks = 0;
1145         u64 num_of_allocated_meta_chunks =
1146                         allocated_meta_size / default_chunk_size;
1147
1148         global_total_size = 0;
1149         ret = ftw(dir_name, ftw_add_entry_size, 10);
1150         dir_size = global_total_size;
1151         if (ret < 0) {
1152                 fprintf(stderr, "ftw subdir walk of '%s' failed: %s\n",
1153                         dir_name, strerror(errno));
1154                 exit(1);
1155         }
1156
1157         num_of_meta_chunks = (dir_size / 2) / default_chunk_size;
1158         if (((dir_size / 2) % default_chunk_size) != 0)
1159                 num_of_meta_chunks++;
1160         if (num_of_meta_chunks <= num_of_allocated_meta_chunks)
1161                 num_of_meta_chunks = 0;
1162         else
1163                 num_of_meta_chunks -= num_of_allocated_meta_chunks;
1164
1165         total_size = allocated_total_size + dir_size +
1166                      (num_of_meta_chunks * default_chunk_size);
1167
1168         *num_of_meta_chunks_ret = num_of_meta_chunks;
1169
1170         return total_size;
1171 }
1172
1173 static int zero_output_file(int out_fd, u64 size, u32 sectorsize)
1174 {
1175         int len = sectorsize;
1176         int loop_num = size / sectorsize;
1177         u64 location = 0;
1178         char *buf = malloc(len);
1179         int ret = 0, i;
1180         ssize_t written;
1181
1182         if (!buf)
1183                 return -ENOMEM;
1184         memset(buf, 0, len);
1185         for (i = 0; i < loop_num; i++) {
1186                 written = pwrite64(out_fd, buf, len, location);
1187                 if (written != len)
1188                         ret = -EIO;
1189                 location += sectorsize;
1190         }
1191         free(buf);
1192         return ret;
1193 }
1194
1195 static int check_leaf_or_node_size(u32 size, u32 sectorsize)
1196 {
1197         if (size < sectorsize) {
1198                 fprintf(stderr,
1199                         "Illegal leafsize (or nodesize) %u (smaller than %u)\n",
1200                         size, sectorsize);
1201                 return -1;
1202         } else if (size > BTRFS_MAX_METADATA_BLOCKSIZE) {
1203                 fprintf(stderr,
1204                         "Illegal leafsize (or nodesize) %u (larger than %u)\n",
1205                         size, BTRFS_MAX_METADATA_BLOCKSIZE);
1206                 return -1;
1207         } else if (size & (sectorsize - 1)) {
1208                 fprintf(stderr,
1209                         "Illegal leafsize (or nodesize) %u (not align to %u)\n",
1210                         size, sectorsize);
1211                 return -1;
1212         }
1213         return 0;
1214 }
1215
1216 static int is_ssd(const char *file)
1217 {
1218         char *devname;
1219         blkid_probe probe;
1220         char *dev;
1221         char path[PATH_MAX];
1222         dev_t disk;
1223         int fd;
1224         char rotational;
1225
1226         probe = blkid_new_probe_from_filename(file);
1227         if (!probe)
1228                 return 0;
1229
1230         /*
1231          * We want to use blkid_devno_to_wholedisk() but it's broken for some
1232          * reason on F17 at least so we'll do this trickery
1233          */
1234         disk = blkid_probe_get_wholedisk_devno(probe);
1235         if (!disk)
1236                 return 0;
1237
1238         devname = blkid_devno_to_devname(disk);
1239         if (!devname)
1240                 return 0;
1241
1242         dev = strrchr(devname, '/');
1243         dev++;
1244
1245         snprintf(path, PATH_MAX, "/sys/block/%s/queue/rotational", dev);
1246
1247         free(devname);
1248         blkid_free_probe(probe);
1249
1250         fd = open(path, O_RDONLY);
1251         if (fd < 0) {
1252                 return 0;
1253         }
1254
1255         if (read(fd, &rotational, sizeof(char)) < sizeof(char)) {
1256                 close(fd);
1257                 return 0;
1258         }
1259         close(fd);
1260
1261         return !atoi((const char *)&rotational);
1262 }
1263
1264 int main(int ac, char **av)
1265 {
1266         char *file;
1267         struct btrfs_root *root;
1268         struct btrfs_trans_handle *trans;
1269         char *label = NULL;
1270         char *first_file;
1271         u64 block_count = 0;
1272         u64 dev_block_count = 0;
1273         u64 blocks[7];
1274         u64 alloc_start = 0;
1275         u64 metadata_profile = 0;
1276         u64 data_profile = 0;
1277         u32 leafsize = sysconf(_SC_PAGESIZE);
1278         u32 sectorsize = 4096;
1279         u32 nodesize = leafsize;
1280         u32 stripesize = 4096;
1281         int zero_end = 1;
1282         int option_index = 0;
1283         int fd;
1284         int ret;
1285         int i;
1286         int mixed = 0;
1287         int data_profile_opt = 0;
1288         int metadata_profile_opt = 0;
1289         int nodiscard = 0;
1290         int ssd = 0;
1291         int force_overwrite = 0;
1292
1293         char *source_dir = NULL;
1294         int source_dir_set = 0;
1295         u64 num_of_meta_chunks = 0;
1296         u64 size_of_data = 0;
1297         u64 source_dir_size = 0;
1298         char *pretty_buf;
1299         struct btrfs_super_block *super;
1300         u64 flags;
1301         int dev_cnt = 0;
1302         int saved_optind;
1303         char estr[100];
1304
1305         while(1) {
1306                 int c;
1307                 c = getopt_long(ac, av, "A:b:fl:n:s:m:d:L:r:VMK", long_options,
1308                                 &option_index);
1309                 if (c < 0)
1310                         break;
1311                 switch(c) {
1312                         case 'A':
1313                                 alloc_start = parse_size(optarg);
1314                                 break;
1315                         case 'f':
1316                                 force_overwrite = 1;
1317                                 break;
1318                         case 'd':
1319                                 data_profile = parse_profile(optarg);
1320                                 data_profile_opt = 1;
1321                                 break;
1322                         case 'l':
1323                         case 'n':
1324                                 nodesize = parse_size(optarg);
1325                                 leafsize = parse_size(optarg);
1326                                 break;
1327                         case 'L':
1328                                 label = parse_label(optarg);
1329                                 break;
1330                         case 'm':
1331                                 metadata_profile = parse_profile(optarg);
1332                                 metadata_profile_opt = 1;
1333                                 break;
1334                         case 'M':
1335                                 mixed = 1;
1336                                 break;
1337                         case 's':
1338                                 sectorsize = parse_size(optarg);
1339                                 break;
1340                         case 'b':
1341                                 block_count = parse_size(optarg);
1342                                 if (block_count <= 1024*1024*1024) {
1343                                         printf("SMALL VOLUME: forcing mixed "
1344                                                "metadata/data groups\n");
1345                                         mixed = 1;
1346                                 }
1347                                 zero_end = 0;
1348                                 break;
1349                         case 'V':
1350                                 print_version();
1351                                 break;
1352                         case 'r':
1353                                 source_dir = optarg;
1354                                 source_dir_set = 1;
1355                                 break;
1356                         case 'K':
1357                                 nodiscard=1;
1358                                 break;
1359                         default:
1360                                 print_usage();
1361                 }
1362         }
1363         sectorsize = max(sectorsize, (u32)sysconf(_SC_PAGESIZE));
1364         if (check_leaf_or_node_size(leafsize, sectorsize))
1365                 exit(1);
1366         if (check_leaf_or_node_size(nodesize, sectorsize))
1367                 exit(1);
1368         saved_optind = optind;
1369         dev_cnt = ac - optind;
1370         if (dev_cnt == 0)
1371                 print_usage();
1372
1373         if (source_dir_set && dev_cnt > 1) {
1374                 fprintf(stderr,
1375                         "The -r option is limited to a single device\n");
1376                 exit(1);
1377         }
1378         while (dev_cnt-- > 0) {
1379                 file = av[optind++];
1380                 if (is_block_device(file))
1381                         if (test_dev_for_mkfs(file, force_overwrite, estr)) {
1382                                 fprintf(stderr, "Error: %s", estr);
1383                                 exit(1);
1384                         }
1385         }
1386
1387         /* if we are here that means all devs are good to btrfsify */
1388         optind = saved_optind;
1389         dev_cnt = ac - optind;
1390
1391         printf("\nWARNING! - %s IS EXPERIMENTAL\n", BTRFS_BUILD_VERSION);
1392         printf("WARNING! - see http://btrfs.wiki.kernel.org before using\n\n");
1393
1394         file = av[optind++];
1395         dev_cnt--;
1396
1397         if (!source_dir_set) {
1398                 /*
1399                  * open without O_EXCL so that the problem should not
1400                  * occur by the following processing.
1401                  * (btrfs_register_one_device() fails if O_EXCL is on)
1402                  */
1403                 fd = open(file, O_RDWR);
1404                 if (fd < 0) {
1405                         fprintf(stderr, "unable to open %s: %s\n", file,
1406                                 strerror(errno));
1407                         exit(1);
1408                 }
1409                 first_file = file;
1410                 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1411                                            block_count, &mixed, nodiscard);
1412                 if (block_count && block_count > dev_block_count) {
1413                         fprintf(stderr, "%s is smaller than requested size\n", file);
1414                         exit(1);
1415                 }
1416         } else {
1417                 fd = open_target(file);
1418                 if (fd < 0) {
1419                         fprintf(stderr, "unable to open the %s\n", file);
1420                         exit(1);
1421                 }
1422
1423                 first_file = file;
1424                 source_dir_size = size_sourcedir(source_dir, sectorsize,
1425                                              &num_of_meta_chunks, &size_of_data);
1426                 if(block_count < source_dir_size)
1427                         block_count = source_dir_size;
1428                 ret = zero_output_file(fd, block_count, sectorsize);
1429                 if (ret) {
1430                         fprintf(stderr, "unable to zero the output file\n");
1431                         exit(1);
1432                 }
1433                 /* our "device" is the new image file */
1434                 dev_block_count = block_count;
1435         }
1436
1437         ssd = is_ssd(file);
1438
1439         if (mixed) {
1440                 if (metadata_profile != data_profile) {
1441                         fprintf(stderr, "With mixed block groups data and metadata "
1442                                 "profiles must be the same\n");
1443                         exit(1);
1444                 }
1445         }
1446
1447         blocks[0] = BTRFS_SUPER_INFO_OFFSET;
1448         for (i = 1; i < 7; i++) {
1449                 blocks[i] = BTRFS_SUPER_INFO_OFFSET + 1024 * 1024 +
1450                         leafsize * i;
1451         }
1452
1453         ret = make_btrfs(fd, file, label, blocks, dev_block_count,
1454                          nodesize, leafsize,
1455                          sectorsize, stripesize);
1456         if (ret) {
1457                 fprintf(stderr, "error during mkfs %d\n", ret);
1458                 exit(1);
1459         }
1460
1461         root = open_ctree(file, 0, O_RDWR);
1462         if (!root) {
1463                 fprintf(stderr, "Open ctree failed\n");
1464                 close(fd);
1465                 exit(1);
1466         }
1467         root->fs_info->alloc_start = alloc_start;
1468
1469         ret = make_root_dir(root, mixed);
1470         if (ret) {
1471                 fprintf(stderr, "failed to setup the root directory\n");
1472                 exit(1);
1473         }
1474
1475         trans = btrfs_start_transaction(root, 1);
1476
1477         if (dev_cnt == 0)
1478                 goto raid_groups;
1479
1480         btrfs_register_one_device(file);
1481
1482         zero_end = 1;
1483         while (dev_cnt-- > 0) {
1484                 int old_mixed = mixed;
1485
1486                 file = av[optind++];
1487
1488                 /*
1489                  * open without O_EXCL so that the problem should not
1490                  * occur by the following processing.
1491                  * (btrfs_register_one_device() fails if O_EXCL is on)
1492                  */
1493                 fd = open(file, O_RDWR);
1494                 if (fd < 0) {
1495                         fprintf(stderr, "unable to open %s: %s\n", file,
1496                                 strerror(errno));
1497                         exit(1);
1498                 }
1499                 ret = btrfs_device_already_in_root(root, fd,
1500                                                    BTRFS_SUPER_INFO_OFFSET);
1501                 if (ret) {
1502                         fprintf(stderr, "skipping duplicate device %s in FS\n",
1503                                 file);
1504                         close(fd);
1505                         continue;
1506                 }
1507                 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1508                                            block_count, &mixed, nodiscard);
1509                 mixed = old_mixed;
1510                 BUG_ON(ret);
1511
1512                 ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
1513                                         sectorsize, sectorsize, sectorsize);
1514                 BUG_ON(ret);
1515                 btrfs_register_one_device(file);
1516         }
1517
1518 raid_groups:
1519         if (!source_dir_set) {
1520                 ret = create_raid_groups(trans, root, data_profile,
1521                                  data_profile_opt, metadata_profile,
1522                                  metadata_profile_opt, mixed, ssd);
1523                 BUG_ON(ret);
1524         }
1525
1526         ret = create_data_reloc_tree(trans, root);
1527         BUG_ON(ret);
1528
1529         super = root->fs_info->super_copy;
1530         flags = btrfs_super_incompat_flags(super);
1531         flags |= BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF;
1532
1533         if (mixed)
1534                 flags |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
1535
1536         btrfs_set_super_incompat_flags(super, flags);
1537
1538         if ((data_profile | metadata_profile) &
1539             (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
1540                 struct btrfs_super_block *super = root->fs_info->super_copy;
1541                 u64 flags = btrfs_super_incompat_flags(super);
1542
1543                 flags |= BTRFS_FEATURE_INCOMPAT_RAID56;
1544                 btrfs_set_super_incompat_flags(super, flags);
1545                 printf("Setting RAID5/6 feature flag\n");
1546         }
1547
1548         printf("fs created label %s on %s\n\tnodesize %u leafsize %u "
1549             "sectorsize %u size %s\n",
1550             label, first_file, nodesize, leafsize, sectorsize,
1551             pretty_buf = pretty_sizes(btrfs_super_total_bytes(root->fs_info->super_copy)));
1552         free(pretty_buf);
1553
1554         printf("%s\n", BTRFS_BUILD_VERSION);
1555         btrfs_commit_transaction(trans, root);
1556
1557         if (source_dir_set) {
1558                 trans = btrfs_start_transaction(root, 1);
1559                 ret = create_chunks(trans, root,
1560                                     num_of_meta_chunks, size_of_data);
1561                 BUG_ON(ret);
1562                 btrfs_commit_transaction(trans, root);
1563
1564                 ret = make_image(source_dir, root, fd);
1565                 BUG_ON(ret);
1566         }
1567
1568         ret = close_ctree(root);
1569         BUG_ON(ret);
1570         free(label);
1571         return 0;
1572 }