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 #include <sys/ioctl.h>
  25 #include <sys/mount.h>
  26 #include "ioctl.h"
  27 #include <stdio.h>
  28 #include <stdlib.h>
  29 #include <sys/types.h>
  30 #include <sys/stat.h>
  31 #include <sys/dir.h>
  32 #include <fcntl.h>
  33 #include <unistd.h>
  34 #include <getopt.h>
  35 #include <uuid/uuid.h>
  36 #include <linux/fs.h>
  37 #include <ctype.h>
  38 #include <attr/xattr.h>
  39 #include <blkid/blkid.h>
  40 #include <ftw.h>
  41 #include "ctree.h"
  42 #include "disk-io.h"
  43 #include "volumes.h"
  44 #include "transaction.h"
  45 #include "utils.h"
  46 #include "version.h"
  47
  48 static u64 index_cnt = 2;
  49
  50 struct directory_name_entry {
  51         char *dir_name;
  52         char *path;
  53         ino_t inum;
  54         struct list_head list;
  55 };
  56
  57 static int make_root_dir(struct btrfs_root *root, int mixed)
  58 {
  59         struct btrfs_trans_handle *trans;
  60         struct btrfs_key location;
  61         u64 bytes_used;
  62         u64 chunk_start = 0;
  63         u64 chunk_size = 0;
  64         int ret;
  65
  66         trans = btrfs_start_transaction(root, 1);
  67         bytes_used = btrfs_super_bytes_used(root->fs_info->super_copy);
  68
  69         root->fs_info->system_allocs = 1;
  70         ret = btrfs_make_block_group(trans, root, bytes_used,
  71                                      BTRFS_BLOCK_GROUP_SYSTEM,
  72                                      BTRFS_FIRST_CHUNK_TREE_OBJECTID,
  73                                      0, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
  74         BUG_ON(ret);
  75
  76         if (mixed) {
  77                 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
  78                                         &chunk_start, &chunk_size,
  79                                         BTRFS_BLOCK_GROUP_METADATA |
  80                                         BTRFS_BLOCK_GROUP_DATA);
  81                 BUG_ON(ret);
  82                 ret = btrfs_make_block_group(trans, root, 0,
  83                                              BTRFS_BLOCK_GROUP_METADATA |
  84                                              BTRFS_BLOCK_GROUP_DATA,
  85                                              BTRFS_FIRST_CHUNK_TREE_OBJECTID,
  86                                              chunk_start, chunk_size);
  87                 BUG_ON(ret);
  88                 printf("Created a data/metadata chunk of size %llu\n", chunk_size);
  89         } else {
  90                 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
  91                                         &chunk_start, &chunk_size,
  92                                         BTRFS_BLOCK_GROUP_METADATA);
  93                 BUG_ON(ret);
  94                 ret = btrfs_make_block_group(trans, root, 0,
  95                                              BTRFS_BLOCK_GROUP_METADATA,
  96                                              BTRFS_FIRST_CHUNK_TREE_OBJECTID,
  97                                              chunk_start, chunk_size);
  98                 BUG_ON(ret);
  99         }
 100
 101         root->fs_info->system_allocs = 0;
 102         btrfs_commit_transaction(trans, root);
 103         trans = btrfs_start_transaction(root, 1);
 104         BUG_ON(!trans);
 105
 106         if (!mixed) {
 107                 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
 108                                         &chunk_start, &chunk_size,
 109                                         BTRFS_BLOCK_GROUP_DATA);
 110                 BUG_ON(ret);
 111                 ret = btrfs_make_block_group(trans, root, 0,
 112                                              BTRFS_BLOCK_GROUP_DATA,
 113                                              BTRFS_FIRST_CHUNK_TREE_OBJECTID,
 114                                              chunk_start, chunk_size);
 115                 BUG_ON(ret);
 116         }
 117
 118         ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
 119                               BTRFS_ROOT_TREE_DIR_OBJECTID);
 120         if (ret)
 121                 goto err;
 122         ret = btrfs_make_root_dir(trans, root, BTRFS_FIRST_FREE_OBJECTID);
 123         if (ret)
 124                 goto err;
 125         memcpy(&location, &root->fs_info->fs_root->root_key, sizeof(location));
 126         location.offset = (u64)-1;
 127         ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
 128                         "default", 7,
 129                         btrfs_super_root_dir(root->fs_info->super_copy),
 130                         &location, BTRFS_FT_DIR, 0);
 131         if (ret)
 132                 goto err;
 133
 134         ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
 135                              "default", 7, location.objectid,
 136                              BTRFS_ROOT_TREE_DIR_OBJECTID, 0);
 137         if (ret)
 138                 goto err;
 139
 140         btrfs_commit_transaction(trans, root);
 141 err:
 142         return ret;
 143 }
 144
 145 static void __recow_root(struct btrfs_trans_handle *trans,
 146                          struct btrfs_root *root)
 147 {
 148         int ret;
 149         struct extent_buffer *tmp;
 150
 151         if (trans->transid != btrfs_root_generation(&root->root_item)) {
 152                 ret = __btrfs_cow_block(trans, root, root->node,
 153                                         NULL, 0, &tmp, 0, 0);
 154                 BUG_ON(ret);
 155                 free_extent_buffer(tmp);
 156         }
 157 }
 158
 159 static void recow_roots(struct btrfs_trans_handle *trans,
 160                        struct btrfs_root *root)
 161 {
 162         struct btrfs_fs_info *info = root->fs_info;
 163
 164         __recow_root(trans, info->fs_root);
 165         __recow_root(trans, info->tree_root);
 166         __recow_root(trans, info->extent_root);
 167         __recow_root(trans, info->chunk_root);
 168         __recow_root(trans, info->dev_root);
 169         __recow_root(trans, info->csum_root);
 170 }
 171
 172 static int create_one_raid_group(struct btrfs_trans_handle *trans,
 173                               struct btrfs_root *root, u64 type)
 174 {
 175         u64 chunk_start;
 176         u64 chunk_size;
 177         int ret;
 178
 179         ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
 180                                 &chunk_start, &chunk_size, type);
 181         BUG_ON(ret);
 182         ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
 183                                      type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
 184                                      chunk_start, chunk_size);
 185         BUG_ON(ret);
 186         return ret;
 187 }
 188
 189 static int create_raid_groups(struct btrfs_trans_handle *trans,
 190                               struct btrfs_root *root, u64 data_profile,
 191                               int data_profile_opt, u64 metadata_profile,
 192                               int metadata_profile_opt, int mixed, int ssd)
 193 {
 194         u64 num_devices = btrfs_super_num_devices(root->fs_info->super_copy);
 195         u64 allowed = 0;
 196         u64 devices_for_raid = num_devices;
 197         int ret;
 198
 199         /*
 200          * Set default profiles according to number of added devices.
 201          * For mixed groups defaults are single/single.
 202          */
 203         if (!metadata_profile_opt && !mixed) {
 204                 if (num_devices == 1 && ssd)
 205                         printf("Detected a SSD, turning off metadata "
 206                                "duplication.  Mkfs with -m dup if you want to "
 207                                "force metadata duplication.\n");
 208                 metadata_profile = (num_devices > 1) ?
 209                         BTRFS_BLOCK_GROUP_RAID1 : (ssd) ? 0: BTRFS_BLOCK_GROUP_DUP;
 210         }
 211         if (!data_profile_opt && !mixed) {
 212                 data_profile = (num_devices > 1) ?
 213                         BTRFS_BLOCK_GROUP_RAID0 : 0; /* raid0 or single */
 214         }
 215
 216         if (devices_for_raid > 4)
 217                 devices_for_raid = 4;
 218
 219         switch (devices_for_raid) {
 220         default:
 221         case 4:
 222                 allowed |= BTRFS_BLOCK_GROUP_RAID10;
 223         case 3:
 224                 allowed |= BTRFS_BLOCK_GROUP_RAID6;
 225         case 2:
 226                 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
 227                         BTRFS_BLOCK_GROUP_RAID5;
 228                 break;
 229         case 1:
 230                 allowed |= BTRFS_BLOCK_GROUP_DUP;
 231         }
 232
 233         if (metadata_profile & ~allowed) {
 234                 fprintf(stderr, "unable to create FS with metadata "
 235                         "profile %llu (have %llu devices)\n", metadata_profile,
 236                         num_devices);
 237                 exit(1);
 238         }
 239         if (data_profile & ~allowed) {
 240                 fprintf(stderr, "unable to create FS with data "
 241                         "profile %llu (have %llu devices)\n", data_profile,
 242                         num_devices);
 243                 exit(1);
 244         }
 245
 246         /* allow dup'ed data chunks only in mixed mode */
 247         if (!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP)) {
 248                 fprintf(stderr, "dup for data is allowed only in mixed mode\n");
 249                 exit(1);
 250         }
 251
 252         if (allowed & metadata_profile) {
 253                 u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
 254
 255                 ret = create_one_raid_group(trans, root,
 256                                             BTRFS_BLOCK_GROUP_SYSTEM |
 257                                             (allowed & metadata_profile));
 258                 BUG_ON(ret);
 259
 260                 if (mixed)
 261                         meta_flags |= BTRFS_BLOCK_GROUP_DATA;
 262
 263                 ret = create_one_raid_group(trans, root, meta_flags |
 264                                             (allowed & metadata_profile));
 265                 BUG_ON(ret);
 266
 267         }
 268         if (!mixed && num_devices > 1 && (allowed & data_profile)) {
 269                 ret = create_one_raid_group(trans, root,
 270                                             BTRFS_BLOCK_GROUP_DATA |
 271                                             (allowed & data_profile));
 272                 BUG_ON(ret);
 273         }
 274         recow_roots(trans, root);
 275
 276         return 0;
 277 }
 278
 279 static int create_data_reloc_tree(struct btrfs_trans_handle *trans,
 280                                   struct btrfs_root *root)
 281 {
 282         struct btrfs_key location;
 283         struct btrfs_root_item root_item;
 284         struct extent_buffer *tmp;
 285         u64 objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
 286         int ret;
 287
 288         ret = btrfs_copy_root(trans, root, root->node, &tmp, objectid);
 289         BUG_ON(ret);
 290
 291         memcpy(&root_item, &root->root_item, sizeof(root_item));
 292         btrfs_set_root_bytenr(&root_item, tmp->start);
 293         btrfs_set_root_level(&root_item, btrfs_header_level(tmp));
 294         btrfs_set_root_generation(&root_item, trans->transid);
 295         free_extent_buffer(tmp);
 296
 297         location.objectid = objectid;
 298         location.type = BTRFS_ROOT_ITEM_KEY;
 299         location.offset = 0;
 300         ret = btrfs_insert_root(trans, root->fs_info->tree_root,
 301                                 &location, &root_item);
 302         BUG_ON(ret);
 303         return 0;
 304 }
 305
 306 static void print_usage(void)
 307 {
 308         fprintf(stderr, "usage: mkfs.btrfs [options] dev [ dev ... ]\n");
 309         fprintf(stderr, "options:\n");
 310         fprintf(stderr, "\t -A --alloc-start the offset to start the FS\n");
 311         fprintf(stderr, "\t -b --byte-count total number of bytes in the FS\n");
 312         fprintf(stderr, "\t -d --data data profile, raid0, raid1, raid5, raid6, raid10, dup or single\n");
 313         fprintf(stderr, "\t -f --force force overwrite of existing filesystem\n");
 314         fprintf(stderr, "\t -l --leafsize size of btree leaves\n");
 315         fprintf(stderr, "\t -L --label set a label\n");
 316         fprintf(stderr, "\t -m --metadata metadata profile, values like data profile\n");
 317         fprintf(stderr, "\t -M --mixed mix metadata and data together\n");
 318         fprintf(stderr, "\t -n --nodesize size of btree nodes\n");
 319         fprintf(stderr, "\t -s --sectorsize min block allocation (may not mountable by current kernel)\n");
 320         fprintf(stderr, "\t -r --rootdir the source directory\n");
 321         fprintf(stderr, "\t -K --nodiscard do not perform whole device TRIM\n");
 322         fprintf(stderr, "\t -O --features comma separated list of filesystem features\n");
 323         fprintf(stderr, "\t -V --version print the mkfs.btrfs version and exit\n");
 324         fprintf(stderr, "%s\n", BTRFS_BUILD_VERSION);
 325         exit(1);
 326 }
 327
 328 static void print_version(void)
 329 {
 330         fprintf(stderr, "mkfs.btrfs, part of %s\n", BTRFS_BUILD_VERSION);
 331         exit(0);
 332 }
 333
 334 static u64 parse_profile(char *s)
 335 {
 336         if (strcmp(s, "raid0") == 0) {
 337                 return BTRFS_BLOCK_GROUP_RAID0;
 338         } else if (strcmp(s, "raid1") == 0) {
 339                 return BTRFS_BLOCK_GROUP_RAID1;
 340         } else if (strcmp(s, "raid5") == 0) {
 341                 return BTRFS_BLOCK_GROUP_RAID5;
 342         } else if (strcmp(s, "raid6") == 0) {
 343                 return BTRFS_BLOCK_GROUP_RAID6;
 344         } else if (strcmp(s, "raid10") == 0) {
 345                 return BTRFS_BLOCK_GROUP_RAID10;
 346         } else if (strcmp(s, "dup") == 0) {
 347                 return BTRFS_BLOCK_GROUP_DUP;
 348         } else if (strcmp(s, "single") == 0) {
 349                 return 0;
 350         } else {
 351                 fprintf(stderr, "Unknown profile %s\n", s);
 352                 print_usage();
 353         }
 354         /* not reached */
 355         return 0;
 356 }
 357
 358 static char *parse_label(char *input)
 359 {
 360         int len = strlen(input);
 361
 362         if (len >= BTRFS_LABEL_SIZE) {
 363                 fprintf(stderr, "Label %s is too long (max %d)\n", input,
 364                         BTRFS_LABEL_SIZE - 1);
 365                 exit(1);
 366         }
 367         return strdup(input);
 368 }
 369
 370 static struct option long_options[] = {
 371         { "alloc-start", 1, NULL, 'A'},
 372         { "byte-count", 1, NULL, 'b' },
 373         { "force", 0, NULL, 'f' },
 374         { "leafsize", 1, NULL, 'l' },
 375         { "label", 1, NULL, 'L'},
 376         { "metadata", 1, NULL, 'm' },
 377         { "mixed", 0, NULL, 'M' },
 378         { "nodesize", 1, NULL, 'n' },
 379         { "sectorsize", 1, NULL, 's' },
 380         { "data", 1, NULL, 'd' },
 381         { "version", 0, NULL, 'V' },
 382         { "rootdir", 1, NULL, 'r' },
 383         { "nodiscard", 0, NULL, 'K' },
 384         { "features", 0, NULL, 'O' },
 385         { 0, 0, 0, 0}
 386 };
 387
 388 static int add_directory_items(struct btrfs_trans_handle *trans,
 389                                struct btrfs_root *root, u64 objectid,
 390                                ino_t parent_inum, const char *name,
 391                                struct stat *st, int *dir_index_cnt)
 392 {
 393         int ret;
 394         int name_len;
 395         struct btrfs_key location;
 396         u8 filetype = 0;
 397
 398         name_len = strlen(name);
 399
 400         location.objectid = objectid;
 401         location.offset = 0;
 402         btrfs_set_key_type(&location, BTRFS_INODE_ITEM_KEY);
 403
 404         if (S_ISDIR(st->st_mode))
 405                 filetype = BTRFS_FT_DIR;
 406         if (S_ISREG(st->st_mode))
 407                 filetype = BTRFS_FT_REG_FILE;
 408         if (S_ISLNK(st->st_mode))
 409                 filetype = BTRFS_FT_SYMLINK;
 410
 411         ret = btrfs_insert_dir_item(trans, root, name, name_len,
 412                                     parent_inum, &location,
 413                                     filetype, index_cnt);
 414
 415         *dir_index_cnt = index_cnt;
 416         index_cnt++;
 417
 418         return ret;
 419 }
 420
 421 static int fill_inode_item(struct btrfs_trans_handle *trans,
 422                            struct btrfs_root *root,
 423                            struct btrfs_inode_item *dst, struct stat *src)
 424 {
 425         u64 blocks = 0;
 426         u64 sectorsize = root->sectorsize;
 427
 428         /*
 429          * btrfs_inode_item has some reserved fields
 430          * and represents on-disk inode entry, so
 431          * zero everything to prevent information leak
 432          */
 433         memset(dst, 0, sizeof (*dst));
 434
 435         btrfs_set_stack_inode_generation(dst, trans->transid);
 436         btrfs_set_stack_inode_size(dst, src->st_size);
 437         btrfs_set_stack_inode_nbytes(dst, 0);
 438         btrfs_set_stack_inode_block_group(dst, 0);
 439         btrfs_set_stack_inode_nlink(dst, src->st_nlink);
 440         btrfs_set_stack_inode_uid(dst, src->st_uid);
 441         btrfs_set_stack_inode_gid(dst, src->st_gid);
 442         btrfs_set_stack_inode_mode(dst, src->st_mode);
 443         btrfs_set_stack_inode_rdev(dst, 0);
 444         btrfs_set_stack_inode_flags(dst, 0);
 445         btrfs_set_stack_timespec_sec(&dst->atime, src->st_atime);
 446         btrfs_set_stack_timespec_nsec(&dst->atime, 0);
 447         btrfs_set_stack_timespec_sec(&dst->ctime, src->st_ctime);
 448         btrfs_set_stack_timespec_nsec(&dst->ctime, 0);
 449         btrfs_set_stack_timespec_sec(&dst->mtime, src->st_mtime);
 450         btrfs_set_stack_timespec_nsec(&dst->mtime, 0);
 451         btrfs_set_stack_timespec_sec(&dst->otime, 0);
 452         btrfs_set_stack_timespec_nsec(&dst->otime, 0);
 453
 454         if (S_ISDIR(src->st_mode)) {
 455                 btrfs_set_stack_inode_size(dst, 0);
 456                 btrfs_set_stack_inode_nlink(dst, 1);
 457         }
 458         if (S_ISREG(src->st_mode)) {
 459                 btrfs_set_stack_inode_size(dst, (u64)src->st_size);
 460                 if (src->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root))
 461                         btrfs_set_stack_inode_nbytes(dst, src->st_size);
 462                 else {
 463                         blocks = src->st_size / sectorsize;
 464                         if (src->st_size % sectorsize)
 465                                 blocks += 1;
 466                         blocks *= sectorsize;
 467                         btrfs_set_stack_inode_nbytes(dst, blocks);
 468                 }
 469         }
 470         if (S_ISLNK(src->st_mode))
 471                 btrfs_set_stack_inode_nbytes(dst, src->st_size + 1);
 472
 473         return 0;
 474 }
 475
 476 static int directory_select(const struct direct *entry)
 477 {
 478         if ((strncmp(entry->d_name, ".", entry->d_reclen) == 0) ||
 479                 (strncmp(entry->d_name, "..", entry->d_reclen) == 0))
 480                 return 0;
 481         else
 482                 return 1;
 483 }
 484
 485 static void free_namelist(struct direct **files, int count)
 486 {
 487         int i;
 488
 489         if (count < 0)
 490                 return;
 491
 492         for (i = 0; i < count; ++i)
 493                 free(files[i]);
 494         free(files);
 495 }
 496
 497 static u64 calculate_dir_inode_size(char *dirname)
 498 {
 499         int count, i;
 500         struct direct **files, *cur_file;
 501         u64 dir_inode_size = 0;
 502
 503         count = scandir(dirname, &files, directory_select, NULL);
 504
 505         for (i = 0; i < count; i++) {
 506                 cur_file = files[i];
 507                 dir_inode_size += strlen(cur_file->d_name);
 508         }
 509
 510         free_namelist(files, count);
 511
 512         dir_inode_size *= 2;
 513         return dir_inode_size;
 514 }
 515
 516 static int add_inode_items(struct btrfs_trans_handle *trans,
 517                            struct btrfs_root *root,
 518                            struct stat *st, char *name,
 519                            u64 self_objectid, ino_t parent_inum,
 520                            int dir_index_cnt, struct btrfs_inode_item *inode_ret)
 521 {
 522         int ret;
 523         struct btrfs_key inode_key;
 524         struct btrfs_inode_item btrfs_inode;
 525         u64 objectid;
 526         u64 inode_size = 0;
 527         int name_len;
 528
 529         name_len = strlen(name);
 530         fill_inode_item(trans, root, &btrfs_inode, st);
 531         objectid = self_objectid;
 532
 533         if (S_ISDIR(st->st_mode)) {
 534                 inode_size = calculate_dir_inode_size(name);
 535                 btrfs_set_stack_inode_size(&btrfs_inode, inode_size);
 536         }
 537
 538         inode_key.objectid = objectid;
 539         inode_key.offset = 0;
 540         btrfs_set_key_type(&inode_key, BTRFS_INODE_ITEM_KEY);
 541
 542         ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
 543         if (ret)
 544                 goto fail;
 545
 546         ret = btrfs_insert_inode_ref(trans, root, name, name_len,
 547                                      objectid, parent_inum, dir_index_cnt);
 548         if (ret)
 549                 goto fail;
 550
 551         *inode_ret = btrfs_inode;
 552 fail:
 553         return ret;
 554 }
 555
 556 static int add_xattr_item(struct btrfs_trans_handle *trans,
 557                           struct btrfs_root *root, u64 objectid,
 558                           const char *file_name)
 559 {
 560         int ret;
 561         int cur_name_len;
 562         char xattr_list[XATTR_LIST_MAX];
 563         char *cur_name;
 564         char cur_value[XATTR_SIZE_MAX];
 565         char delimiter = '\0';
 566         char *next_location = xattr_list;
 567
 568         ret = llistxattr(file_name, xattr_list, XATTR_LIST_MAX);
 569         if (ret < 0) {
 570                 if(errno == ENOTSUP)
 571                         return 0;
 572                 fprintf(stderr, "get a list of xattr failed for %s\n",
 573                         file_name);
 574                 return ret;
 575         }
 576         if (ret == 0)
 577                 return ret;
 578
 579         cur_name = strtok(xattr_list, &delimiter);
 580         while (cur_name != NULL) {
 581                 cur_name_len = strlen(cur_name);
 582                 next_location += cur_name_len + 1;
 583
 584                 ret = getxattr(file_name, cur_name, cur_value, XATTR_SIZE_MAX);
 585                 if (ret < 0) {
 586                         if(errno == ENOTSUP)
 587                                 return 0;
 588                         fprintf(stderr, "get a xattr value failed for %s attr %s\n",
 589                                 file_name, cur_name);
 590                         return ret;
 591                 }
 592
 593                 ret = btrfs_insert_xattr_item(trans, root, cur_name,
 594                                               cur_name_len, cur_value,
 595                                               ret, objectid);
 596                 if (ret) {
 597                         fprintf(stderr, "insert a xattr item failed for %s\n",
 598                                 file_name);
 599                 }
 600
 601                 cur_name = strtok(next_location, &delimiter);
 602         }
 603
 604         return ret;
 605 }
 606 static int custom_alloc_extent(struct btrfs_root *root, u64 num_bytes,
 607                                u64 hint_byte, struct btrfs_key *ins)
 608 {
 609         u64 start;
 610         u64 end;
 611         u64 last = hint_byte;
 612         int ret;
 613         int wrapped = 0;
 614         struct btrfs_block_group_cache *cache;
 615
 616         while (1) {
 617                 ret = find_first_extent_bit(&root->fs_info->free_space_cache,
 618                                             last, &start, &end, EXTENT_DIRTY);
 619                 if (ret) {
 620                         if (wrapped++ == 0) {
 621                                 last = 0;
 622                                 continue;
 623                         } else {
 624                                 goto fail;
 625                         }
 626                 }
 627
 628                 start = max(last, start);
 629                 last = end + 1;
 630                 if (last - start < num_bytes)
 631                         continue;
 632
 633                 last = start + num_bytes;
 634                 if (test_range_bit(&root->fs_info->pinned_extents,
 635                                    start, last - 1, EXTENT_DIRTY, 0))
 636                         continue;
 637
 638                 cache = btrfs_lookup_block_group(root->fs_info, start);
 639                 BUG_ON(!cache);
 640                 if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM ||
 641                     last > cache->key.objectid + cache->key.offset) {
 642                         last = cache->key.objectid + cache->key.offset;
 643                         continue;
 644                 }
 645
 646                 if (cache->flags & (BTRFS_BLOCK_GROUP_SYSTEM |
 647                             BTRFS_BLOCK_GROUP_METADATA)) {
 648                         last = cache->key.objectid + cache->key.offset;
 649                         continue;
 650                 }
 651
 652                 clear_extent_dirty(&root->fs_info->free_space_cache,
 653                                    start, start + num_bytes - 1, 0);
 654
 655                 ins->objectid = start;
 656                 ins->offset = num_bytes;
 657                 ins->type = BTRFS_EXTENT_ITEM_KEY;
 658                 return 0;
 659         }
 660 fail:
 661         fprintf(stderr, "not enough free space\n");
 662         return -ENOSPC;
 663 }
 664
 665 static int record_file_extent(struct btrfs_trans_handle *trans,
 666                               struct btrfs_root *root, u64 objectid,
 667                               struct btrfs_inode_item *inode,
 668                               u64 file_pos, u64 disk_bytenr,
 669                               u64 num_bytes)
 670 {
 671         int ret;
 672         struct btrfs_fs_info *info = root->fs_info;
 673         struct btrfs_root *extent_root = info->extent_root;
 674         struct extent_buffer *leaf;
 675         struct btrfs_file_extent_item *fi;
 676         struct btrfs_key ins_key;
 677         struct btrfs_path path;
 678         struct btrfs_extent_item *ei;
 679
 680         btrfs_init_path(&path);
 681
 682         ins_key.objectid = objectid;
 683         ins_key.offset = 0;
 684         btrfs_set_key_type(&ins_key, BTRFS_EXTENT_DATA_KEY);
 685         ret = btrfs_insert_empty_item(trans, root, &path, &ins_key,
 686                                       sizeof(*fi));
 687         if (ret)
 688                 goto fail;
 689         leaf = path.nodes[0];
 690         fi = btrfs_item_ptr(leaf, path.slots[0],
 691                             struct btrfs_file_extent_item);
 692         btrfs_set_file_extent_generation(leaf, fi, trans->transid);
 693         btrfs_set_file_extent_type(leaf, fi, BTRFS_FILE_EXTENT_REG);
 694         btrfs_set_file_extent_disk_bytenr(leaf, fi, disk_bytenr);
 695         btrfs_set_file_extent_disk_num_bytes(leaf, fi, num_bytes);
 696         btrfs_set_file_extent_offset(leaf, fi, 0);
 697         btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
 698         btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
 699         btrfs_set_file_extent_compression(leaf, fi, 0);
 700         btrfs_set_file_extent_encryption(leaf, fi, 0);
 701         btrfs_set_file_extent_other_encoding(leaf, fi, 0);
 702         btrfs_mark_buffer_dirty(leaf);
 703
 704         btrfs_release_path(&path);
 705
 706         ins_key.objectid = disk_bytenr;
 707         ins_key.offset = num_bytes;
 708         ins_key.type = BTRFS_EXTENT_ITEM_KEY;
 709
 710         ret = btrfs_insert_empty_item(trans, extent_root, &path,
 711                                 &ins_key, sizeof(*ei));
 712         if (ret == 0) {
 713                 leaf = path.nodes[0];
 714                 ei = btrfs_item_ptr(leaf, path.slots[0],
 715                                     struct btrfs_extent_item);
 716
 717                 btrfs_set_extent_refs(leaf, ei, 0);
 718                 btrfs_set_extent_generation(leaf, ei, trans->transid);
 719                 btrfs_set_extent_flags(leaf, ei, BTRFS_EXTENT_FLAG_DATA);
 720
 721                 btrfs_mark_buffer_dirty(leaf);
 722                 ret = btrfs_update_block_group(trans, root, disk_bytenr,
 723                                                num_bytes, 1, 0);
 724                 if (ret)
 725                         goto fail;
 726         } else if (ret != -EEXIST) {
 727                 goto fail;
 728         }
 729
 730         ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, num_bytes, 0,
 731                                    root->root_key.objectid,
 732                                    objectid, 0);
 733 fail:
 734         btrfs_release_path(&path);
 735         return ret;
 736 }
 737
 738 static int add_symbolic_link(struct btrfs_trans_handle *trans,
 739                              struct btrfs_root *root,
 740                              u64 objectid, const char *path_name)
 741 {
 742         int ret;
 743         u64 sectorsize = root->sectorsize;
 744         char *buf = malloc(sectorsize);
 745
 746         ret = readlink(path_name, buf, sectorsize);
 747         if (ret <= 0) {
 748                 fprintf(stderr, "readlink failed for %s\n", path_name);
 749                 goto fail;
 750         }
 751         if (ret >= sectorsize) {
 752                 fprintf(stderr, "symlink too long for %s", path_name);
 753                 ret = -1;
 754                 goto fail;
 755         }
 756
 757         buf[ret] = '\0'; /* readlink does not do it for us */
 758         ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
 759                                          buf, ret + 1);
 760 fail:
 761         free(buf);
 762         return ret;
 763 }
 764
 765 static int add_file_items(struct btrfs_trans_handle *trans,
 766                           struct btrfs_root *root,
 767                           struct btrfs_inode_item *btrfs_inode, u64 objectid,
 768                           ino_t parent_inum, struct stat *st,
 769                           const char *path_name, int out_fd)
 770 {
 771         int ret = -1;
 772         ssize_t ret_read;
 773         u64 bytes_read = 0;
 774         char *buffer = NULL;
 775         struct btrfs_key key;
 776         int blocks;
 777         u32 sectorsize = root->sectorsize;
 778         u64 first_block = 0;
 779         u64 num_blocks = 0;
 780         int fd;
 781
 782         fd = open(path_name, O_RDONLY);
 783         if (fd == -1) {
 784                 fprintf(stderr, "%s open failed\n", path_name);
 785                 return ret;
 786         }
 787
 788         blocks = st->st_size / sectorsize;
 789         if (st->st_size % sectorsize)
 790                 blocks += 1;
 791
 792         if (st->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
 793                 buffer = malloc(st->st_size);
 794                 ret_read = pread64(fd, buffer, st->st_size, bytes_read);
 795                 if (ret_read == -1) {
 796                         fprintf(stderr, "%s read failed\n", path_name);
 797                         goto end;
 798                 }
 799
 800                 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
 801                                                  buffer, st->st_size);
 802                 goto end;
 803         }
 804
 805         ret = custom_alloc_extent(root, blocks * sectorsize, 0, &key);
 806         if (ret)
 807                 goto end;
 808
 809         first_block = key.objectid;
 810         bytes_read = 0;
 811         buffer = malloc(sectorsize);
 812
 813         do {
 814                 memset(buffer, 0, sectorsize);
 815                 ret_read = pread64(fd, buffer, sectorsize, bytes_read);
 816                 if (ret_read == -1) {
 817                         fprintf(stderr, "%s read failed\n", path_name);
 818                         goto end;
 819                 }
 820
 821                 ret = pwrite64(out_fd, buffer, sectorsize,
 822                                first_block + bytes_read);
 823                 if (ret != sectorsize) {
 824                         fprintf(stderr, "output file write failed\n");
 825                         goto end;
 826                 }
 827
 828                 /* checksum for file data */
 829                 ret = btrfs_csum_file_block(trans, root->fs_info->csum_root,
 830                                 first_block + (blocks * sectorsize),
 831                                 first_block + bytes_read,
 832                                 buffer, sectorsize);
 833                 if (ret) {
 834                         fprintf(stderr, "%s checksum failed\n", path_name);
 835                         goto end;
 836                 }
 837
 838                 bytes_read += ret_read;
 839                 num_blocks++;
 840         } while (ret_read == sectorsize);
 841
 842         if (num_blocks > 0) {
 843                 ret = record_file_extent(trans, root, objectid, btrfs_inode,
 844                                          first_block, first_block,
 845                                          blocks * sectorsize);
 846                 if (ret)
 847                         goto end;
 848         }
 849
 850 end:
 851         if (buffer)
 852                 free(buffer);
 853         close(fd);
 854         return ret;
 855 }
 856
 857 static char *make_path(char *dir, char *name)
 858 {
 859         char *path;
 860
 861         path = malloc(strlen(dir) + strlen(name) + 2);
 862         if (!path)
 863                 return NULL;
 864         strcpy(path, dir);
 865         if (dir[strlen(dir) - 1] != '/')
 866                 strcat(path, "/");
 867         strcat(path, name);
 868         return path;
 869 }
 870
 871 static int traverse_directory(struct btrfs_trans_handle *trans,
 872                               struct btrfs_root *root, char *dir_name,
 873                               struct directory_name_entry *dir_head, int out_fd)
 874 {
 875         int ret = 0;
 876
 877         struct btrfs_inode_item cur_inode;
 878         struct btrfs_inode_item *inode_item;
 879         int count, i, dir_index_cnt;
 880         struct direct **files;
 881         struct stat st;
 882         struct directory_name_entry *dir_entry, *parent_dir_entry;
 883         struct direct *cur_file;
 884         ino_t parent_inum, cur_inum;
 885         ino_t highest_inum = 0;
 886         char *parent_dir_name;
 887         struct btrfs_path path;
 888         struct extent_buffer *leaf;
 889         struct btrfs_key root_dir_key;
 890         u64 root_dir_inode_size = 0;
 891
 892         /* Add list for source directory */
 893         dir_entry = malloc(sizeof(struct directory_name_entry));
 894         dir_entry->dir_name = dir_name;
 895         dir_entry->path = strdup(dir_name);
 896
 897         parent_inum = highest_inum + BTRFS_FIRST_FREE_OBJECTID;
 898         dir_entry->inum = parent_inum;
 899         list_add_tail(&dir_entry->list, &dir_head->list);
 900
 901         btrfs_init_path(&path);
 902
 903         root_dir_key.objectid = btrfs_root_dirid(&root->root_item);
 904         root_dir_key.offset = 0;
 905         btrfs_set_key_type(&root_dir_key, BTRFS_INODE_ITEM_KEY);
 906         ret = btrfs_lookup_inode(trans, root, &path, &root_dir_key, 1);
 907         if (ret) {
 908                 fprintf(stderr, "root dir lookup error\n");
 909                 return -1;
 910         }
 911
 912         leaf = path.nodes[0];
 913         inode_item = btrfs_item_ptr(leaf, path.slots[0],
 914                                     struct btrfs_inode_item);
 915
 916         root_dir_inode_size = calculate_dir_inode_size(dir_name);
 917         btrfs_set_inode_size(leaf, inode_item, root_dir_inode_size);
 918         btrfs_mark_buffer_dirty(leaf);
 919
 920         btrfs_release_path(&path);
 921
 922         do {
 923                 parent_dir_entry = list_entry(dir_head->list.next,
 924                                               struct directory_name_entry,
 925                                               list);
 926                 list_del(&parent_dir_entry->list);
 927
 928                 parent_inum = parent_dir_entry->inum;
 929                 parent_dir_name = parent_dir_entry->dir_name;
 930                 if (chdir(parent_dir_entry->path)) {
 931                         fprintf(stderr, "chdir error for %s\n",
 932                                 parent_dir_name);
 933                         goto fail_no_files;
 934                 }
 935
 936                 count = scandir(parent_dir_entry->path, &files,
 937                                 directory_select, NULL);
 938                 if (count == -1)
 939                 {
 940                         fprintf(stderr, "scandir for %s failed: %s\n",
 941                                 parent_dir_name, strerror (errno));
 942                         goto fail;
 943                 }
 944
 945                 for (i = 0; i < count; i++) {
 946                         cur_file = files[i];
 947
 948                         if (lstat(cur_file->d_name, &st) == -1) {
 949                                 fprintf(stderr, "lstat failed for file %s\n",
 950                                         cur_file->d_name);
 951                                 goto fail;
 952                         }
 953
 954                         cur_inum = ++highest_inum + BTRFS_FIRST_FREE_OBJECTID;
 955                         ret = add_directory_items(trans, root,
 956                                                   cur_inum, parent_inum,
 957                                                   cur_file->d_name,
 958                                                   &st, &dir_index_cnt);
 959                         if (ret) {
 960                                 fprintf(stderr, "add_directory_items failed\n");
 961                                 goto fail;
 962                         }
 963
 964                         ret = add_inode_items(trans, root, &st,
 965                                               cur_file->d_name, cur_inum,
 966                                               parent_inum, dir_index_cnt,
 967                                               &cur_inode);
 968                         if (ret) {
 969                                 fprintf(stderr, "add_inode_items failed\n");
 970                                 goto fail;
 971                         }
 972
 973                         ret = add_xattr_item(trans, root,
 974                                              cur_inum, cur_file->d_name);
 975                         if (ret) {
 976                                 fprintf(stderr, "add_xattr_item failed\n");
 977                                 if(ret != -ENOTSUP)
 978                                         goto fail;
 979                         }
 980
 981                         if (S_ISDIR(st.st_mode)) {
 982                                 dir_entry = malloc(sizeof(struct directory_name_entry));
 983                                 dir_entry->dir_name = cur_file->d_name;
 984                                 dir_entry->path = make_path(parent_dir_entry->path,
 985                                                             cur_file->d_name);
 986                                 dir_entry->inum = cur_inum;
 987                                 list_add_tail(&dir_entry->list, &dir_head->list);
 988                         } else if (S_ISREG(st.st_mode)) {
 989                                 ret = add_file_items(trans, root, &cur_inode,
 990                                                      cur_inum, parent_inum, &st,
 991                                                      cur_file->d_name, out_fd);
 992                                 if (ret) {
 993                                         fprintf(stderr, "add_file_items failed\n");
 994                                         goto fail;
 995                                 }
 996                         } else if (S_ISLNK(st.st_mode)) {
 997                                 ret = add_symbolic_link(trans, root,
 998                                                         cur_inum, cur_file->d_name);
 999                                 if (ret) {
1000                                         fprintf(stderr, "add_symbolic_link failed\n");
1001                                         goto fail;
1002                                 }
1003                         }
1004                 }
1005
1006                 free_namelist(files, count);
1007                 free(parent_dir_entry->path);
1008                 free(parent_dir_entry);
1009
1010                 index_cnt = 2;
1011
1012         } while (!list_empty(&dir_head->list));
1013
1014         return 0;
1015 fail:
1016         free_namelist(files, count);
1017 fail_no_files:
1018         free(parent_dir_entry->path);
1019         free(parent_dir_entry);
1020         return -1;
1021 }
1022
1023 static int open_target(char *output_name)
1024 {
1025         int output_fd;
1026         output_fd = open(output_name, O_CREAT | O_RDWR | O_TRUNC,
1027                          S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH);
1028
1029         return output_fd;
1030 }
1031
1032 static int create_chunks(struct btrfs_trans_handle *trans,
1033                          struct btrfs_root *root, u64 num_of_meta_chunks,
1034                          u64 size_of_data)
1035 {
1036         u64 chunk_start;
1037         u64 chunk_size;
1038         u64 meta_type = BTRFS_BLOCK_GROUP_METADATA;
1039         u64 data_type = BTRFS_BLOCK_GROUP_DATA;
1040         u64 minimum_data_chunk_size = 8 * 1024 * 1024;
1041         u64 i;
1042         int ret;
1043
1044         for (i = 0; i < num_of_meta_chunks; i++) {
1045                 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
1046                                         &chunk_start, &chunk_size, meta_type);
1047                 BUG_ON(ret);
1048                 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
1049                                              meta_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
1050                                              chunk_start, chunk_size);
1051                 BUG_ON(ret);
1052                 set_extent_dirty(&root->fs_info->free_space_cache,
1053                                  chunk_start, chunk_start + chunk_size - 1, 0);
1054         }
1055
1056         if (size_of_data < minimum_data_chunk_size)
1057                 size_of_data = minimum_data_chunk_size;
1058         ret = btrfs_alloc_data_chunk(trans, root->fs_info->extent_root,
1059                                      &chunk_start, size_of_data, data_type);
1060         BUG_ON(ret);
1061         ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
1062                                      data_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
1063                                      chunk_start, size_of_data);
1064         BUG_ON(ret);
1065         set_extent_dirty(&root->fs_info->free_space_cache,
1066                          chunk_start, chunk_start + size_of_data - 1, 0);
1067         return ret;
1068 }
1069
1070 static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
1071 {
1072         int ret;
1073         struct btrfs_trans_handle *trans;
1074
1075         struct stat root_st;
1076
1077         struct directory_name_entry dir_head;
1078
1079         ret = lstat(source_dir, &root_st);
1080         if (ret) {
1081                 fprintf(stderr, "unable to lstat the %s\n", source_dir);
1082                 goto fail;
1083         }
1084
1085         INIT_LIST_HEAD(&dir_head.list);
1086
1087         trans = btrfs_start_transaction(root, 1);
1088         ret = traverse_directory(trans, root, source_dir, &dir_head, out_fd);
1089         if (ret) {
1090                 fprintf(stderr, "unable to traverse_directory\n");
1091                 goto fail;
1092         }
1093         btrfs_commit_transaction(trans, root);
1094
1095         printf("Making image is completed.\n");
1096         return 0;
1097 fail:
1098         fprintf(stderr, "Making image is aborted.\n");
1099         return -1;
1100 }
1101
1102 /*
1103  * This ignores symlinks with unreadable targets and subdirs that can't
1104  * be read.  It's a best-effort to give a rough estimate of the size of
1105  * a subdir.  It doesn't guarantee that prepopulating btrfs from this
1106  * tree won't still run out of space.
1107  *
1108  * The rounding up to 4096 is questionable.  Previous code used du -B 4096.
1109  */
1110 static u64 global_total_size;
1111 static int ftw_add_entry_size(const char *fpath, const struct stat *st,
1112                               int type)
1113 {
1114         if (type == FTW_F || type == FTW_D)
1115                 global_total_size += round_up(st->st_size, 4096);
1116
1117         return 0;
1118 }
1119
1120 static u64 size_sourcedir(char *dir_name, u64 sectorsize,
1121                           u64 *num_of_meta_chunks_ret, u64 *size_of_data_ret)
1122 {
1123         u64 dir_size = 0;
1124         u64 total_size = 0;
1125         int ret;
1126         u64 default_chunk_size = 8 * 1024 * 1024;       /* 8MB */
1127         u64 allocated_meta_size = 8 * 1024 * 1024;      /* 8MB */
1128         u64 allocated_total_size = 20 * 1024 * 1024;    /* 20MB */
1129         u64 num_of_meta_chunks = 0;
1130         u64 num_of_allocated_meta_chunks =
1131                         allocated_meta_size / default_chunk_size;
1132
1133         global_total_size = 0;
1134         ret = ftw(dir_name, ftw_add_entry_size, 10);
1135         dir_size = global_total_size;
1136         if (ret < 0) {
1137                 fprintf(stderr, "ftw subdir walk of '%s' failed: %s\n",
1138                         dir_name, strerror(errno));
1139                 exit(1);
1140         }
1141
1142         num_of_meta_chunks = (dir_size / 2) / default_chunk_size;
1143         if (((dir_size / 2) % default_chunk_size) != 0)
1144                 num_of_meta_chunks++;
1145         if (num_of_meta_chunks <= num_of_allocated_meta_chunks)
1146                 num_of_meta_chunks = 0;
1147         else
1148                 num_of_meta_chunks -= num_of_allocated_meta_chunks;
1149
1150         total_size = allocated_total_size + dir_size +
1151                      (num_of_meta_chunks * default_chunk_size);
1152
1153         *num_of_meta_chunks_ret = num_of_meta_chunks;
1154
1155         return total_size;
1156 }
1157
1158 static int zero_output_file(int out_fd, u64 size, u32 sectorsize)
1159 {
1160         int len = sectorsize;
1161         int loop_num = size / sectorsize;
1162         u64 location = 0;
1163         char *buf = malloc(len);
1164         int ret = 0, i;
1165         ssize_t written;
1166
1167         if (!buf)
1168                 return -ENOMEM;
1169         memset(buf, 0, len);
1170         for (i = 0; i < loop_num; i++) {
1171                 written = pwrite64(out_fd, buf, len, location);
1172                 if (written != len)
1173                         ret = -EIO;
1174                 location += sectorsize;
1175         }
1176         free(buf);
1177         return ret;
1178 }
1179
1180 static int check_leaf_or_node_size(u32 size, u32 sectorsize)
1181 {
1182         if (size < sectorsize) {
1183                 fprintf(stderr,
1184                         "Illegal leafsize (or nodesize) %u (smaller than %u)\n",
1185                         size, sectorsize);
1186                 return -1;
1187         } else if (size > BTRFS_MAX_METADATA_BLOCKSIZE) {
1188                 fprintf(stderr,
1189                         "Illegal leafsize (or nodesize) %u (larger than %u)\n",
1190                         size, BTRFS_MAX_METADATA_BLOCKSIZE);
1191                 return -1;
1192         } else if (size & (sectorsize - 1)) {
1193                 fprintf(stderr,
1194                         "Illegal leafsize (or nodesize) %u (not align to %u)\n",
1195                         size, sectorsize);
1196                 return -1;
1197         }
1198         return 0;
1199 }
1200
1201 static int is_ssd(const char *file)
1202 {
1203         blkid_probe probe;
1204         char wholedisk[32];
1205         char sysfs_path[PATH_MAX];
1206         dev_t devno;
1207         int fd;
1208         char rotational;
1209
1210         probe = blkid_new_probe_from_filename(file);
1211         if (!probe)
1212                 return 0;
1213
1214         /* Device number of this disk (possibly a partition) */
1215         devno = blkid_probe_get_devno(probe);
1216         if (!devno)
1217                 return 0;
1218
1219         /* Get whole disk name (not full path) for this devno */
1220         blkid_devno_to_wholedisk(devno, wholedisk, sizeof(wholedisk), NULL);
1221
1222         snprintf(sysfs_path, PATH_MAX, "/sys/block/%s/queue/rotational",
1223                  wholedisk);
1224
1225         blkid_free_probe(probe);
1226
1227         fd = open(sysfs_path, O_RDONLY);
1228         if (fd < 0) {
1229                 return 0;
1230         }
1231
1232         if (read(fd, &rotational, sizeof(char)) < sizeof(char)) {
1233                 close(fd);
1234                 return 0;
1235         }
1236         close(fd);
1237
1238         return !atoi((const char *)&rotational);
1239 }
1240
1241 #define BTRFS_FEATURE_LIST_ALL          (1ULL << 63)
1242
1243 static const struct btrfs_fs_feature {
1244         const char *name;
1245         u64 flag;
1246         const char *desc;
1247 } mkfs_features[] = {
1248         { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1249                 "mixed data and metadata block groups" },
1250         { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1251                 "increased hardlink limit per file to 65536" },
1252         { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1253                 "raid56 extended format" },
1254         { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1255                 "reduced-size metadata extent refs" },
1256         /* Keep this one last */
1257         { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1258 };
1259
1260 static void list_all_fs_features(void)
1261 {
1262         int i;
1263
1264         fprintf(stderr, "Filesystem features available at mkfs time:\n");
1265         for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1266                 fprintf(stderr, "%-20s- %s (0x%llx)\n",
1267                                 mkfs_features[i].name,
1268                                 mkfs_features[i].desc,
1269                                 mkfs_features[i].flag);
1270         }
1271 }
1272
1273 static int parse_one_fs_feature(const char *name, u64 *flags)
1274 {
1275         int i;
1276         int found = 0;
1277
1278         for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1279                 if (!strcmp(mkfs_features[i].name, name)) {
1280                         *flags |= mkfs_features[i].flag;
1281                         found = 1;
1282                 }
1283         }
1284
1285         return !found;
1286 }
1287
1288 static void process_fs_features(u64 flags)
1289 {
1290         int i;
1291
1292         for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1293                 if (flags & mkfs_features[i].flag) {
1294                         fprintf(stderr,
1295                                 "Turning ON incompat feature '%s': %s\n",
1296                                 mkfs_features[i].name,
1297                                 mkfs_features[i].desc);
1298                 }
1299         }
1300 }
1301
1302
1303 /*
1304  * Return NULL if all features were parsed fine, otherwise return the name of
1305  * the first unparsed.
1306  */
1307 static char* parse_fs_features(char *namelist, u64 *flags)
1308 {
1309         char *this_char;
1310         char *save_ptr = NULL; /* Satisfy static checkers */
1311
1312         for (this_char = strtok_r(namelist, ",", &save_ptr);
1313              this_char != NULL;
1314              this_char = strtok_r(NULL, ",", &save_ptr)) {
1315                 if (parse_one_fs_feature(this_char, flags))
1316                         return this_char;
1317         }
1318
1319         return NULL;
1320 }
1321
1322 int main(int ac, char **av)
1323 {
1324         char *file;
1325         struct btrfs_root *root;
1326         struct btrfs_trans_handle *trans;
1327         char *label = NULL;
1328         char *first_file;
1329         u64 block_count = 0;
1330         u64 dev_block_count = 0;
1331         u64 blocks[7];
1332         u64 alloc_start = 0;
1333         u64 metadata_profile = 0;
1334         u64 data_profile = 0;
1335         u32 leafsize = sysconf(_SC_PAGESIZE);
1336         u32 sectorsize = 4096;
1337         u32 nodesize = leafsize;
1338         u32 stripesize = 4096;
1339         int zero_end = 1;
1340         int option_index = 0;
1341         int fd;
1342         int ret;
1343         int i;
1344         int mixed = 0;
1345         int data_profile_opt = 0;
1346         int metadata_profile_opt = 0;
1347         int nodiscard = 0;
1348         int ssd = 0;
1349         int force_overwrite = 0;
1350
1351         char *source_dir = NULL;
1352         int source_dir_set = 0;
1353         u64 num_of_meta_chunks = 0;
1354         u64 size_of_data = 0;
1355         u64 source_dir_size = 0;
1356         struct btrfs_super_block *super;
1357         u64 flags;
1358         int dev_cnt = 0;
1359         int saved_optind;
1360         char estr[100];
1361         u64 features = 0;
1362
1363         while(1) {
1364                 int c;
1365                 c = getopt_long(ac, av, "A:b:fl:n:s:m:d:L:O:r:VMK",
1366                                 long_options, &option_index);
1367                 if (c < 0)
1368                         break;
1369                 switch(c) {
1370                         case 'A':
1371                                 alloc_start = parse_size(optarg);
1372                                 break;
1373                         case 'f':
1374                                 force_overwrite = 1;
1375                                 break;
1376                         case 'd':
1377                                 data_profile = parse_profile(optarg);
1378                                 data_profile_opt = 1;
1379                                 break;
1380                         case 'l':
1381                         case 'n':
1382                                 nodesize = parse_size(optarg);
1383                                 leafsize = parse_size(optarg);
1384                                 break;
1385                         case 'L':
1386                                 label = parse_label(optarg);
1387                                 break;
1388                         case 'm':
1389                                 metadata_profile = parse_profile(optarg);
1390                                 metadata_profile_opt = 1;
1391                                 break;
1392                         case 'M':
1393                                 mixed = 1;
1394                                 break;
1395                         case 'O': {
1396                                 char *orig = strdup(optarg);
1397                                 char *tmp = orig;
1398
1399                                 tmp = parse_fs_features(tmp, &features);
1400                                 if (tmp) {
1401                                         fprintf(stderr,
1402                                                 "Unrecognized filesystem feature '%s'\n",
1403                                                         tmp);
1404                                         free(orig);
1405                                         exit(1);
1406                                 }
1407                                 free(orig);
1408                                 if (features & BTRFS_FEATURE_LIST_ALL) {
1409                                         list_all_fs_features();
1410                                         exit(0);
1411                                 }
1412                                 break;
1413                                 }
1414                         case 's':
1415                                 sectorsize = parse_size(optarg);
1416                                 break;
1417                         case 'b':
1418                                 block_count = parse_size(optarg);
1419                                 if (block_count <= 1024*1024*1024) {
1420                                         printf("SMALL VOLUME: forcing mixed "
1421                                                "metadata/data groups\n");
1422                                         mixed = 1;
1423                                 }
1424                                 zero_end = 0;
1425                                 break;
1426                         case 'V':
1427                                 print_version();
1428                                 break;
1429                         case 'r':
1430                                 source_dir = optarg;
1431                                 source_dir_set = 1;
1432                                 break;
1433                         case 'K':
1434                                 nodiscard=1;
1435                                 break;
1436                         default:
1437                                 print_usage();
1438                 }
1439         }
1440         sectorsize = max(sectorsize, (u32)sysconf(_SC_PAGESIZE));
1441         if (check_leaf_or_node_size(leafsize, sectorsize))
1442                 exit(1);
1443         if (check_leaf_or_node_size(nodesize, sectorsize))
1444                 exit(1);
1445         saved_optind = optind;
1446         dev_cnt = ac - optind;
1447         if (dev_cnt == 0)
1448                 print_usage();
1449
1450         if (source_dir_set && dev_cnt > 1) {
1451                 fprintf(stderr,
1452                         "The -r option is limited to a single device\n");
1453                 exit(1);
1454         }
1455         while (dev_cnt-- > 0) {
1456                 file = av[optind++];
1457                 if (is_block_device(file))
1458                         if (test_dev_for_mkfs(file, force_overwrite, estr)) {
1459                                 fprintf(stderr, "Error: %s", estr);
1460                                 exit(1);
1461                         }
1462         }
1463
1464         /* if we are here that means all devs are good to btrfsify */
1465         optind = saved_optind;
1466         dev_cnt = ac - optind;
1467
1468         printf("\nWARNING! - %s IS EXPERIMENTAL\n", BTRFS_BUILD_VERSION);
1469         printf("WARNING! - see http://btrfs.wiki.kernel.org before using\n\n");
1470
1471         file = av[optind++];
1472         dev_cnt--;
1473
1474         if (!source_dir_set) {
1475                 /*
1476                  * open without O_EXCL so that the problem should not
1477                  * occur by the following processing.
1478                  * (btrfs_register_one_device() fails if O_EXCL is on)
1479                  */
1480                 fd = open(file, O_RDWR);
1481                 if (fd < 0) {
1482                         fprintf(stderr, "unable to open %s: %s\n", file,
1483                                 strerror(errno));
1484                         exit(1);
1485                 }
1486                 first_file = file;
1487                 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1488                                            block_count, &mixed, nodiscard);
1489                 if (block_count && block_count > dev_block_count) {
1490                         fprintf(stderr, "%s is smaller than requested size\n", file);
1491                         exit(1);
1492                 }
1493         } else {
1494                 fd = open_target(file);
1495                 if (fd < 0) {
1496                         fprintf(stderr, "unable to open the %s\n", file);
1497                         exit(1);
1498                 }
1499
1500                 first_file = file;
1501                 source_dir_size = size_sourcedir(source_dir, sectorsize,
1502                                              &num_of_meta_chunks, &size_of_data);
1503                 if(block_count < source_dir_size)
1504                         block_count = source_dir_size;
1505                 ret = zero_output_file(fd, block_count, sectorsize);
1506                 if (ret) {
1507                         fprintf(stderr, "unable to zero the output file\n");
1508                         exit(1);
1509                 }
1510                 /* our "device" is the new image file */
1511                 dev_block_count = block_count;
1512         }
1513
1514         ssd = is_ssd(file);
1515
1516         if (mixed) {
1517                 if (metadata_profile != data_profile) {
1518                         fprintf(stderr, "With mixed block groups data and metadata "
1519                                 "profiles must be the same\n");
1520                         exit(1);
1521                 }
1522         }
1523
1524         blocks[0] = BTRFS_SUPER_INFO_OFFSET;
1525         for (i = 1; i < 7; i++) {
1526                 blocks[i] = BTRFS_SUPER_INFO_OFFSET + 1024 * 1024 +
1527                         leafsize * i;
1528         }
1529
1530         ret = make_btrfs(fd, file, label, blocks, dev_block_count,
1531                          nodesize, leafsize,
1532                          sectorsize, stripesize);
1533         if (ret) {
1534                 fprintf(stderr, "error during mkfs: %s\n", strerror(-ret));
1535                 exit(1);
1536         }
1537
1538         root = open_ctree(file, 0, O_RDWR);
1539         if (!root) {
1540                 fprintf(stderr, "Open ctree failed\n");
1541                 close(fd);
1542                 exit(1);
1543         }
1544         root->fs_info->alloc_start = alloc_start;
1545
1546         ret = make_root_dir(root, mixed);
1547         if (ret) {
1548                 fprintf(stderr, "failed to setup the root directory\n");
1549                 exit(1);
1550         }
1551
1552         trans = btrfs_start_transaction(root, 1);
1553
1554         if (dev_cnt == 0)
1555                 goto raid_groups;
1556
1557         btrfs_register_one_device(file);
1558
1559         zero_end = 1;
1560         while (dev_cnt-- > 0) {
1561                 int old_mixed = mixed;
1562
1563                 file = av[optind++];
1564
1565                 /*
1566                  * open without O_EXCL so that the problem should not
1567                  * occur by the following processing.
1568                  * (btrfs_register_one_device() fails if O_EXCL is on)
1569                  */
1570                 fd = open(file, O_RDWR);
1571                 if (fd < 0) {
1572                         fprintf(stderr, "unable to open %s: %s\n", file,
1573                                 strerror(errno));
1574                         exit(1);
1575                 }
1576                 ret = btrfs_device_already_in_root(root, fd,
1577                                                    BTRFS_SUPER_INFO_OFFSET);
1578                 if (ret) {
1579                         fprintf(stderr, "skipping duplicate device %s in FS\n",
1580                                 file);
1581                         close(fd);
1582                         continue;
1583                 }
1584                 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1585                                            block_count, &mixed, nodiscard);
1586                 mixed = old_mixed;
1587                 BUG_ON(ret);
1588
1589                 ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
1590                                         sectorsize, sectorsize, sectorsize);
1591                 BUG_ON(ret);
1592                 btrfs_register_one_device(file);
1593         }
1594
1595 raid_groups:
1596         if (!source_dir_set) {
1597                 ret = create_raid_groups(trans, root, data_profile,
1598                                  data_profile_opt, metadata_profile,
1599                                  metadata_profile_opt, mixed, ssd);
1600                 BUG_ON(ret);
1601         }
1602
1603         ret = create_data_reloc_tree(trans, root);
1604         BUG_ON(ret);
1605
1606         super = root->fs_info->super_copy;
1607         flags = btrfs_super_incompat_flags(super);
1608
1609         /*
1610          * FS features that can be set by other means than -O
1611          * just set the bit here
1612          */
1613         if (mixed)
1614                 features |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
1615
1616         if ((data_profile | metadata_profile) &
1617             (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
1618                 features |= BTRFS_FEATURE_INCOMPAT_RAID56;
1619         }
1620
1621         process_fs_features(features);
1622         flags |= features;
1623         btrfs_set_super_incompat_flags(super, flags);
1624
1625         printf("fs created label %s on %s\n\tnodesize %u leafsize %u "
1626             "sectorsize %u size %s\n",
1627             label, first_file, nodesize, leafsize, sectorsize,
1628             pretty_size(btrfs_super_total_bytes(root->fs_info->super_copy)));
1629
1630         printf("%s\n", BTRFS_BUILD_VERSION);
1631         btrfs_commit_transaction(trans, root);
1632
1633         if (source_dir_set) {
1634                 trans = btrfs_start_transaction(root, 1);
1635                 ret = create_chunks(trans, root,
1636                                     num_of_meta_chunks, size_of_data);
1637                 BUG_ON(ret);
1638                 btrfs_commit_transaction(trans, root);
1639
1640                 ret = make_image(source_dir, root, fd);
1641                 BUG_ON(ret);
1642         }
1643
1644         ret = close_ctree(root);
1645         BUG_ON(ret);
1646         free(label);
1647         return 0;
1648 }