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 <ctype.h>
  37 #include <attr/xattr.h>
  38 #include <blkid/blkid.h>
  39 #include <ftw.h>
  40 #include "ctree.h"
  41 #include "disk-io.h"
  42 #include "volumes.h"
  43 #include "transaction.h"
  44 #include "utils.h"
  45 #include "version.h"
  46
  47 static u64 index_cnt = 2;
  48
  49 #define DEFAULT_MKFS_FEATURES   (BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)
  50
  51 #define DEFAULT_MKFS_LEAF_SIZE 16384
  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                 if (ret == -ENOSPC) {
  85                         fprintf(stderr,
  86                                 "no space to alloc data/metadata chunk\n");
  87                         goto err;
  88                 }
  89                 BUG_ON(ret);
  90                 ret = btrfs_make_block_group(trans, root, 0,
  91                                              BTRFS_BLOCK_GROUP_METADATA |
  92                                              BTRFS_BLOCK_GROUP_DATA,
  93                                              BTRFS_FIRST_CHUNK_TREE_OBJECTID,
  94                                              chunk_start, chunk_size);
  95                 BUG_ON(ret);
  96                 printf("Created a data/metadata chunk of size %llu\n", chunk_size);
  97         } else {
  98                 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
  99                                         &chunk_start, &chunk_size,
 100                                         BTRFS_BLOCK_GROUP_METADATA);
 101                 if (ret == -ENOSPC) {
 102                         fprintf(stderr, "no space to alloc metadata chunk\n");
 103                         goto err;
 104                 }
 105                 BUG_ON(ret);
 106                 ret = btrfs_make_block_group(trans, root, 0,
 107                                              BTRFS_BLOCK_GROUP_METADATA,
 108                                              BTRFS_FIRST_CHUNK_TREE_OBJECTID,
 109                                              chunk_start, chunk_size);
 110                 BUG_ON(ret);
 111         }
 112
 113         root->fs_info->system_allocs = 0;
 114         btrfs_commit_transaction(trans, root);
 115         trans = btrfs_start_transaction(root, 1);
 116         BUG_ON(!trans);
 117
 118         if (!mixed) {
 119                 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
 120                                         &chunk_start, &chunk_size,
 121                                         BTRFS_BLOCK_GROUP_DATA);
 122                 if (ret == -ENOSPC) {
 123                         fprintf(stderr, "no space to alloc data chunk\n");
 124                         goto err;
 125                 }
 126                 BUG_ON(ret);
 127                 ret = btrfs_make_block_group(trans, root, 0,
 128                                              BTRFS_BLOCK_GROUP_DATA,
 129                                              BTRFS_FIRST_CHUNK_TREE_OBJECTID,
 130                                              chunk_start, chunk_size);
 131                 BUG_ON(ret);
 132         }
 133
 134         ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
 135                               BTRFS_ROOT_TREE_DIR_OBJECTID);
 136         if (ret)
 137                 goto err;
 138         ret = btrfs_make_root_dir(trans, root, BTRFS_FIRST_FREE_OBJECTID);
 139         if (ret)
 140                 goto err;
 141         memcpy(&location, &root->fs_info->fs_root->root_key, sizeof(location));
 142         location.offset = (u64)-1;
 143         ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
 144                         "default", 7,
 145                         btrfs_super_root_dir(root->fs_info->super_copy),
 146                         &location, BTRFS_FT_DIR, 0);
 147         if (ret)
 148                 goto err;
 149
 150         ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
 151                              "default", 7, location.objectid,
 152                              BTRFS_ROOT_TREE_DIR_OBJECTID, 0);
 153         if (ret)
 154                 goto err;
 155
 156         btrfs_commit_transaction(trans, root);
 157 err:
 158         return ret;
 159 }
 160
 161 static void __recow_root(struct btrfs_trans_handle *trans,
 162                          struct btrfs_root *root)
 163 {
 164         int ret;
 165         struct extent_buffer *tmp;
 166
 167         if (trans->transid != btrfs_root_generation(&root->root_item)) {
 168                 ret = __btrfs_cow_block(trans, root, root->node,
 169                                         NULL, 0, &tmp, 0, 0);
 170                 BUG_ON(ret);
 171                 free_extent_buffer(tmp);
 172         }
 173 }
 174
 175 static void recow_roots(struct btrfs_trans_handle *trans,
 176                        struct btrfs_root *root)
 177 {
 178         struct btrfs_fs_info *info = root->fs_info;
 179
 180         __recow_root(trans, info->fs_root);
 181         __recow_root(trans, info->tree_root);
 182         __recow_root(trans, info->extent_root);
 183         __recow_root(trans, info->chunk_root);
 184         __recow_root(trans, info->dev_root);
 185         __recow_root(trans, info->csum_root);
 186 }
 187
 188 static int create_one_raid_group(struct btrfs_trans_handle *trans,
 189                               struct btrfs_root *root, u64 type)
 190 {
 191         u64 chunk_start;
 192         u64 chunk_size;
 193         int ret;
 194
 195         ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
 196                                 &chunk_start, &chunk_size, type);
 197         if (ret == -ENOSPC) {
 198                 fprintf(stderr, "not enough free space\n");
 199                 exit(1);
 200         }
 201         BUG_ON(ret);
 202         ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
 203                                      type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
 204                                      chunk_start, chunk_size);
 205         BUG_ON(ret);
 206         return ret;
 207 }
 208
 209 static int create_raid_groups(struct btrfs_trans_handle *trans,
 210                               struct btrfs_root *root, u64 data_profile,
 211                               int data_profile_opt, u64 metadata_profile,
 212                               int metadata_profile_opt, int mixed, int ssd)
 213 {
 214         u64 num_devices = btrfs_super_num_devices(root->fs_info->super_copy);
 215         int ret;
 216
 217         if (metadata_profile) {
 218                 u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
 219
 220                 ret = create_one_raid_group(trans, root,
 221                                             BTRFS_BLOCK_GROUP_SYSTEM |
 222                                             metadata_profile);
 223                 BUG_ON(ret);
 224
 225                 if (mixed)
 226                         meta_flags |= BTRFS_BLOCK_GROUP_DATA;
 227
 228                 ret = create_one_raid_group(trans, root, meta_flags |
 229                                             metadata_profile);
 230                 BUG_ON(ret);
 231
 232         }
 233         if (!mixed && num_devices > 1 && data_profile) {
 234                 ret = create_one_raid_group(trans, root,
 235                                             BTRFS_BLOCK_GROUP_DATA |
 236                                             data_profile);
 237                 BUG_ON(ret);
 238         }
 239         recow_roots(trans, root);
 240
 241         return 0;
 242 }
 243
 244 static int create_data_reloc_tree(struct btrfs_trans_handle *trans,
 245                                   struct btrfs_root *root)
 246 {
 247         struct btrfs_key location;
 248         struct btrfs_root_item root_item;
 249         struct extent_buffer *tmp;
 250         u64 objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
 251         int ret;
 252
 253         ret = btrfs_copy_root(trans, root, root->node, &tmp, objectid);
 254         BUG_ON(ret);
 255
 256         memcpy(&root_item, &root->root_item, sizeof(root_item));
 257         btrfs_set_root_bytenr(&root_item, tmp->start);
 258         btrfs_set_root_level(&root_item, btrfs_header_level(tmp));
 259         btrfs_set_root_generation(&root_item, trans->transid);
 260         free_extent_buffer(tmp);
 261
 262         location.objectid = objectid;
 263         location.type = BTRFS_ROOT_ITEM_KEY;
 264         location.offset = 0;
 265         ret = btrfs_insert_root(trans, root->fs_info->tree_root,
 266                                 &location, &root_item);
 267         BUG_ON(ret);
 268         return 0;
 269 }
 270
 271
 272 static void print_usage(void) __attribute__((noreturn));
 273 static void print_usage(void)
 274 {
 275         fprintf(stderr, "usage: mkfs.btrfs [options] dev [ dev ... ]\n");
 276         fprintf(stderr, "options:\n");
 277         fprintf(stderr, "\t -A --alloc-start the offset to start the FS\n");
 278         fprintf(stderr, "\t -b --byte-count total number of bytes in the FS\n");
 279         fprintf(stderr, "\t -d --data data profile, raid0, raid1, raid5, raid6, raid10, dup or single\n");
 280         fprintf(stderr, "\t -f --force force overwrite of existing filesystem\n");
 281         fprintf(stderr, "\t -l --leafsize size of btree leaves\n");
 282         fprintf(stderr, "\t -L --label set a label\n");
 283         fprintf(stderr, "\t -m --metadata metadata profile, values like data profile\n");
 284         fprintf(stderr, "\t -M --mixed mix metadata and data together\n");
 285         fprintf(stderr, "\t -n --nodesize size of btree nodes\n");
 286         fprintf(stderr, "\t -s --sectorsize min block allocation (may not mountable by current kernel)\n");
 287         fprintf(stderr, "\t -r --rootdir the source directory\n");
 288         fprintf(stderr, "\t -K --nodiscard do not perform whole device TRIM\n");
 289         fprintf(stderr, "\t -O --features comma separated list of filesystem features\n");
 290         fprintf(stderr, "\t -V --version print the mkfs.btrfs version and exit\n");
 291         fprintf(stderr, "%s\n", BTRFS_BUILD_VERSION);
 292         exit(1);
 293 }
 294
 295 static void print_version(void) __attribute__((noreturn));
 296 static void print_version(void)
 297 {
 298         fprintf(stderr, "mkfs.btrfs, part of %s\n", BTRFS_BUILD_VERSION);
 299         exit(0);
 300 }
 301
 302 static u64 parse_profile(char *s)
 303 {
 304         if (strcmp(s, "raid0") == 0) {
 305                 return BTRFS_BLOCK_GROUP_RAID0;
 306         } else if (strcmp(s, "raid1") == 0) {
 307                 return BTRFS_BLOCK_GROUP_RAID1;
 308         } else if (strcmp(s, "raid5") == 0) {
 309                 return BTRFS_BLOCK_GROUP_RAID5;
 310         } else if (strcmp(s, "raid6") == 0) {
 311                 return BTRFS_BLOCK_GROUP_RAID6;
 312         } else if (strcmp(s, "raid10") == 0) {
 313                 return BTRFS_BLOCK_GROUP_RAID10;
 314         } else if (strcmp(s, "dup") == 0) {
 315                 return BTRFS_BLOCK_GROUP_DUP;
 316         } else if (strcmp(s, "single") == 0) {
 317                 return 0;
 318         } else {
 319                 fprintf(stderr, "Unknown profile %s\n", s);
 320                 print_usage();
 321         }
 322         /* not reached */
 323         return 0;
 324 }
 325
 326 static char *parse_label(char *input)
 327 {
 328         int len = strlen(input);
 329
 330         if (len >= BTRFS_LABEL_SIZE) {
 331                 fprintf(stderr, "Label %s is too long (max %d)\n", input,
 332                         BTRFS_LABEL_SIZE - 1);
 333                 exit(1);
 334         }
 335         return strdup(input);
 336 }
 337
 338 static struct option long_options[] = {
 339         { "alloc-start", 1, NULL, 'A'},
 340         { "byte-count", 1, NULL, 'b' },
 341         { "force", 0, NULL, 'f' },
 342         { "leafsize", 1, NULL, 'l' },
 343         { "label", 1, NULL, 'L'},
 344         { "metadata", 1, NULL, 'm' },
 345         { "mixed", 0, NULL, 'M' },
 346         { "nodesize", 1, NULL, 'n' },
 347         { "sectorsize", 1, NULL, 's' },
 348         { "data", 1, NULL, 'd' },
 349         { "version", 0, NULL, 'V' },
 350         { "rootdir", 1, NULL, 'r' },
 351         { "nodiscard", 0, NULL, 'K' },
 352         { "features", 0, NULL, 'O' },
 353         { NULL, 0, NULL, 0}
 354 };
 355
 356 static int add_directory_items(struct btrfs_trans_handle *trans,
 357                                struct btrfs_root *root, u64 objectid,
 358                                ino_t parent_inum, const char *name,
 359                                struct stat *st, int *dir_index_cnt)
 360 {
 361         int ret;
 362         int name_len;
 363         struct btrfs_key location;
 364         u8 filetype = 0;
 365
 366         name_len = strlen(name);
 367
 368         location.objectid = objectid;
 369         location.offset = 0;
 370         btrfs_set_key_type(&location, BTRFS_INODE_ITEM_KEY);
 371
 372         if (S_ISDIR(st->st_mode))
 373                 filetype = BTRFS_FT_DIR;
 374         if (S_ISREG(st->st_mode))
 375                 filetype = BTRFS_FT_REG_FILE;
 376         if (S_ISLNK(st->st_mode))
 377                 filetype = BTRFS_FT_SYMLINK;
 378
 379         ret = btrfs_insert_dir_item(trans, root, name, name_len,
 380                                     parent_inum, &location,
 381                                     filetype, index_cnt);
 382
 383         *dir_index_cnt = index_cnt;
 384         index_cnt++;
 385
 386         return ret;
 387 }
 388
 389 static int fill_inode_item(struct btrfs_trans_handle *trans,
 390                            struct btrfs_root *root,
 391                            struct btrfs_inode_item *dst, struct stat *src)
 392 {
 393         u64 blocks = 0;
 394         u64 sectorsize = root->sectorsize;
 395
 396         /*
 397          * btrfs_inode_item has some reserved fields
 398          * and represents on-disk inode entry, so
 399          * zero everything to prevent information leak
 400          */
 401         memset(dst, 0, sizeof (*dst));
 402
 403         btrfs_set_stack_inode_generation(dst, trans->transid);
 404         btrfs_set_stack_inode_size(dst, src->st_size);
 405         btrfs_set_stack_inode_nbytes(dst, 0);
 406         btrfs_set_stack_inode_block_group(dst, 0);
 407         btrfs_set_stack_inode_nlink(dst, src->st_nlink);
 408         btrfs_set_stack_inode_uid(dst, src->st_uid);
 409         btrfs_set_stack_inode_gid(dst, src->st_gid);
 410         btrfs_set_stack_inode_mode(dst, src->st_mode);
 411         btrfs_set_stack_inode_rdev(dst, 0);
 412         btrfs_set_stack_inode_flags(dst, 0);
 413         btrfs_set_stack_timespec_sec(&dst->atime, src->st_atime);
 414         btrfs_set_stack_timespec_nsec(&dst->atime, 0);
 415         btrfs_set_stack_timespec_sec(&dst->ctime, src->st_ctime);
 416         btrfs_set_stack_timespec_nsec(&dst->ctime, 0);
 417         btrfs_set_stack_timespec_sec(&dst->mtime, src->st_mtime);
 418         btrfs_set_stack_timespec_nsec(&dst->mtime, 0);
 419         btrfs_set_stack_timespec_sec(&dst->otime, 0);
 420         btrfs_set_stack_timespec_nsec(&dst->otime, 0);
 421
 422         if (S_ISDIR(src->st_mode)) {
 423                 btrfs_set_stack_inode_size(dst, 0);
 424                 btrfs_set_stack_inode_nlink(dst, 1);
 425         }
 426         if (S_ISREG(src->st_mode)) {
 427                 btrfs_set_stack_inode_size(dst, (u64)src->st_size);
 428                 if (src->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root))
 429                         btrfs_set_stack_inode_nbytes(dst, src->st_size);
 430                 else {
 431                         blocks = src->st_size / sectorsize;
 432                         if (src->st_size % sectorsize)
 433                                 blocks += 1;
 434                         blocks *= sectorsize;
 435                         btrfs_set_stack_inode_nbytes(dst, blocks);
 436                 }
 437         }
 438         if (S_ISLNK(src->st_mode))
 439                 btrfs_set_stack_inode_nbytes(dst, src->st_size + 1);
 440
 441         return 0;
 442 }
 443
 444 static int directory_select(const struct direct *entry)
 445 {
 446         if ((strncmp(entry->d_name, ".", entry->d_reclen) == 0) ||
 447                 (strncmp(entry->d_name, "..", entry->d_reclen) == 0))
 448                 return 0;
 449         else
 450                 return 1;
 451 }
 452
 453 static void free_namelist(struct direct **files, int count)
 454 {
 455         int i;
 456
 457         if (count < 0)
 458                 return;
 459
 460         for (i = 0; i < count; ++i)
 461                 free(files[i]);
 462         free(files);
 463 }
 464
 465 static u64 calculate_dir_inode_size(char *dirname)
 466 {
 467         int count, i;
 468         struct direct **files, *cur_file;
 469         u64 dir_inode_size = 0;
 470
 471         count = scandir(dirname, &files, directory_select, NULL);
 472
 473         for (i = 0; i < count; i++) {
 474                 cur_file = files[i];
 475                 dir_inode_size += strlen(cur_file->d_name);
 476         }
 477
 478         free_namelist(files, count);
 479
 480         dir_inode_size *= 2;
 481         return dir_inode_size;
 482 }
 483
 484 static int add_inode_items(struct btrfs_trans_handle *trans,
 485                            struct btrfs_root *root,
 486                            struct stat *st, char *name,
 487                            u64 self_objectid, ino_t parent_inum,
 488                            int dir_index_cnt, struct btrfs_inode_item *inode_ret)
 489 {
 490         int ret;
 491         struct btrfs_key inode_key;
 492         struct btrfs_inode_item btrfs_inode;
 493         u64 objectid;
 494         u64 inode_size = 0;
 495         int name_len;
 496
 497         name_len = strlen(name);
 498         fill_inode_item(trans, root, &btrfs_inode, st);
 499         objectid = self_objectid;
 500
 501         if (S_ISDIR(st->st_mode)) {
 502                 inode_size = calculate_dir_inode_size(name);
 503                 btrfs_set_stack_inode_size(&btrfs_inode, inode_size);
 504         }
 505
 506         inode_key.objectid = objectid;
 507         inode_key.offset = 0;
 508         btrfs_set_key_type(&inode_key, BTRFS_INODE_ITEM_KEY);
 509
 510         ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
 511         if (ret)
 512                 goto fail;
 513
 514         ret = btrfs_insert_inode_ref(trans, root, name, name_len,
 515                                      objectid, parent_inum, dir_index_cnt);
 516         if (ret)
 517                 goto fail;
 518
 519         *inode_ret = btrfs_inode;
 520 fail:
 521         return ret;
 522 }
 523
 524 static int add_xattr_item(struct btrfs_trans_handle *trans,
 525                           struct btrfs_root *root, u64 objectid,
 526                           const char *file_name)
 527 {
 528         int ret;
 529         int cur_name_len;
 530         char xattr_list[XATTR_LIST_MAX];
 531         char *cur_name;
 532         char cur_value[XATTR_SIZE_MAX];
 533         char delimiter = '\0';
 534         char *next_location = xattr_list;
 535
 536         ret = llistxattr(file_name, xattr_list, XATTR_LIST_MAX);
 537         if (ret < 0) {
 538                 if(errno == ENOTSUP)
 539                         return 0;
 540                 fprintf(stderr, "get a list of xattr failed for %s\n",
 541                         file_name);
 542                 return ret;
 543         }
 544         if (ret == 0)
 545                 return ret;
 546
 547         cur_name = strtok(xattr_list, &delimiter);
 548         while (cur_name != NULL) {
 549                 cur_name_len = strlen(cur_name);
 550                 next_location += cur_name_len + 1;
 551
 552                 ret = getxattr(file_name, cur_name, cur_value, XATTR_SIZE_MAX);
 553                 if (ret < 0) {
 554                         if(errno == ENOTSUP)
 555                                 return 0;
 556                         fprintf(stderr, "get a xattr value failed for %s attr %s\n",
 557                                 file_name, cur_name);
 558                         return ret;
 559                 }
 560
 561                 ret = btrfs_insert_xattr_item(trans, root, cur_name,
 562                                               cur_name_len, cur_value,
 563                                               ret, objectid);
 564                 if (ret) {
 565                         fprintf(stderr, "insert a xattr item failed for %s\n",
 566                                 file_name);
 567                 }
 568
 569                 cur_name = strtok(next_location, &delimiter);
 570         }
 571
 572         return ret;
 573 }
 574
 575 static int add_symbolic_link(struct btrfs_trans_handle *trans,
 576                              struct btrfs_root *root,
 577                              u64 objectid, const char *path_name)
 578 {
 579         int ret;
 580         u64 sectorsize = root->sectorsize;
 581         char *buf = malloc(sectorsize);
 582
 583         ret = readlink(path_name, buf, sectorsize);
 584         if (ret <= 0) {
 585                 fprintf(stderr, "readlink failed for %s\n", path_name);
 586                 goto fail;
 587         }
 588         if (ret >= sectorsize) {
 589                 fprintf(stderr, "symlink too long for %s", path_name);
 590                 ret = -1;
 591                 goto fail;
 592         }
 593
 594         buf[ret] = '\0'; /* readlink does not do it for us */
 595         ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
 596                                          buf, ret + 1);
 597 fail:
 598         free(buf);
 599         return ret;
 600 }
 601
 602 static int add_file_items(struct btrfs_trans_handle *trans,
 603                           struct btrfs_root *root,
 604                           struct btrfs_inode_item *btrfs_inode, u64 objectid,
 605                           ino_t parent_inum, struct stat *st,
 606                           const char *path_name, int out_fd)
 607 {
 608         int ret = -1;
 609         ssize_t ret_read;
 610         u64 bytes_read = 0;
 611         struct btrfs_key key;
 612         int blocks;
 613         u32 sectorsize = root->sectorsize;
 614         u64 first_block = 0;
 615         u64 file_pos = 0;
 616         u64 cur_bytes;
 617         u64 total_bytes;
 618         struct extent_buffer *eb = NULL;
 619         int fd;
 620
 621         fd = open(path_name, O_RDONLY);
 622         if (fd == -1) {
 623                 fprintf(stderr, "%s open failed\n", path_name);
 624                 return ret;
 625         }
 626
 627         blocks = st->st_size / sectorsize;
 628         if (st->st_size % sectorsize)
 629                 blocks += 1;
 630
 631         if (st->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
 632                 char *buffer = malloc(st->st_size);
 633                 ret_read = pread64(fd, buffer, st->st_size, bytes_read);
 634                 if (ret_read == -1) {
 635                         fprintf(stderr, "%s read failed\n", path_name);
 636                         free(buffer);
 637                         goto end;
 638                 }
 639
 640                 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
 641                                                  buffer, st->st_size);
 642                 free(buffer);
 643                 goto end;
 644         }
 645
 646         /* round up our st_size to the FS blocksize */
 647         total_bytes = (u64)blocks * sectorsize;
 648
 649         /*
 650          * do our IO in extent buffers so it can work
 651          * against any raid type
 652          */
 653         eb = malloc(sizeof(*eb) + sectorsize);
 654         if (!eb) {
 655                 ret = -ENOMEM;
 656                 goto end;
 657         }
 658         memset(eb, 0, sizeof(*eb) + sectorsize);
 659
 660 again:
 661
 662         /*
 663          * keep our extent size at 1MB max, this makes it easier to work inside
 664          * the tiny block groups created during mkfs
 665          */
 666         cur_bytes = min(total_bytes, 1024ULL * 1024);
 667         ret = btrfs_reserve_extent(trans, root, cur_bytes, 0, 0, (u64)-1,
 668                                    &key, 1);
 669         if (ret)
 670                 goto end;
 671
 672         first_block = key.objectid;
 673         bytes_read = 0;
 674
 675         while (bytes_read < cur_bytes) {
 676
 677                 memset(eb->data, 0, sectorsize);
 678
 679                 ret_read = pread64(fd, eb->data, sectorsize, file_pos + bytes_read);
 680                 if (ret_read == -1) {
 681                         fprintf(stderr, "%s read failed\n", path_name);
 682                         goto end;
 683                 }
 684
 685                 eb->start = first_block + bytes_read;
 686                 eb->len = sectorsize;
 687
 688                 /*
 689                  * we're doing the csum before we record the extent, but
 690                  * that's ok
 691                  */
 692                 ret = btrfs_csum_file_block(trans, root->fs_info->csum_root,
 693                                             first_block + bytes_read + sectorsize,
 694                                             first_block + bytes_read,
 695                                             eb->data, sectorsize);
 696                 if (ret)
 697                         goto end;
 698
 699                 ret = write_and_map_eb(trans, root, eb);
 700                 if (ret) {
 701                         fprintf(stderr, "output file write failed\n");
 702                         goto end;
 703                 }
 704
 705                 bytes_read += sectorsize;
 706         }
 707
 708         if (bytes_read) {
 709                 ret = btrfs_record_file_extent(trans, root, objectid, btrfs_inode,
 710                                                file_pos, first_block, cur_bytes);
 711                 if (ret)
 712                         goto end;
 713
 714         }
 715
 716         file_pos += cur_bytes;
 717         total_bytes -= cur_bytes;
 718
 719         if (total_bytes)
 720                 goto again;
 721
 722 end:
 723         if (eb)
 724                 free(eb);
 725         close(fd);
 726         return ret;
 727 }
 728
 729 static char *make_path(char *dir, char *name)
 730 {
 731         char *path;
 732
 733         path = malloc(strlen(dir) + strlen(name) + 2);
 734         if (!path)
 735                 return NULL;
 736         strcpy(path, dir);
 737         if (dir[strlen(dir) - 1] != '/')
 738                 strcat(path, "/");
 739         strcat(path, name);
 740         return path;
 741 }
 742
 743 static int traverse_directory(struct btrfs_trans_handle *trans,
 744                               struct btrfs_root *root, char *dir_name,
 745                               struct directory_name_entry *dir_head, int out_fd)
 746 {
 747         int ret = 0;
 748
 749         struct btrfs_inode_item cur_inode;
 750         struct btrfs_inode_item *inode_item;
 751         int count, i, dir_index_cnt;
 752         struct direct **files;
 753         struct stat st;
 754         struct directory_name_entry *dir_entry, *parent_dir_entry;
 755         struct direct *cur_file;
 756         ino_t parent_inum, cur_inum;
 757         ino_t highest_inum = 0;
 758         char *parent_dir_name;
 759         struct btrfs_path path;
 760         struct extent_buffer *leaf;
 761         struct btrfs_key root_dir_key;
 762         u64 root_dir_inode_size = 0;
 763
 764         /* Add list for source directory */
 765         dir_entry = malloc(sizeof(struct directory_name_entry));
 766         dir_entry->dir_name = dir_name;
 767         dir_entry->path = strdup(dir_name);
 768
 769         parent_inum = highest_inum + BTRFS_FIRST_FREE_OBJECTID;
 770         dir_entry->inum = parent_inum;
 771         list_add_tail(&dir_entry->list, &dir_head->list);
 772
 773         btrfs_init_path(&path);
 774
 775         root_dir_key.objectid = btrfs_root_dirid(&root->root_item);
 776         root_dir_key.offset = 0;
 777         btrfs_set_key_type(&root_dir_key, BTRFS_INODE_ITEM_KEY);
 778         ret = btrfs_lookup_inode(trans, root, &path, &root_dir_key, 1);
 779         if (ret) {
 780                 fprintf(stderr, "root dir lookup error\n");
 781                 return -1;
 782         }
 783
 784         leaf = path.nodes[0];
 785         inode_item = btrfs_item_ptr(leaf, path.slots[0],
 786                                     struct btrfs_inode_item);
 787
 788         root_dir_inode_size = calculate_dir_inode_size(dir_name);
 789         btrfs_set_inode_size(leaf, inode_item, root_dir_inode_size);
 790         btrfs_mark_buffer_dirty(leaf);
 791
 792         btrfs_release_path(&path);
 793
 794         do {
 795                 parent_dir_entry = list_entry(dir_head->list.next,
 796                                               struct directory_name_entry,
 797                                               list);
 798                 list_del(&parent_dir_entry->list);
 799
 800                 parent_inum = parent_dir_entry->inum;
 801                 parent_dir_name = parent_dir_entry->dir_name;
 802                 if (chdir(parent_dir_entry->path)) {
 803                         fprintf(stderr, "chdir error for %s\n",
 804                                 parent_dir_name);
 805                         goto fail_no_files;
 806                 }
 807
 808                 count = scandir(parent_dir_entry->path, &files,
 809                                 directory_select, NULL);
 810                 if (count == -1)
 811                 {
 812                         fprintf(stderr, "scandir for %s failed: %s\n",
 813                                 parent_dir_name, strerror (errno));
 814                         goto fail;
 815                 }
 816
 817                 for (i = 0; i < count; i++) {
 818                         cur_file = files[i];
 819
 820                         if (lstat(cur_file->d_name, &st) == -1) {
 821                                 fprintf(stderr, "lstat failed for file %s\n",
 822                                         cur_file->d_name);
 823                                 goto fail;
 824                         }
 825
 826                         cur_inum = ++highest_inum + BTRFS_FIRST_FREE_OBJECTID;
 827                         ret = add_directory_items(trans, root,
 828                                                   cur_inum, parent_inum,
 829                                                   cur_file->d_name,
 830                                                   &st, &dir_index_cnt);
 831                         if (ret) {
 832                                 fprintf(stderr, "add_directory_items failed\n");
 833                                 goto fail;
 834                         }
 835
 836                         ret = add_inode_items(trans, root, &st,
 837                                               cur_file->d_name, cur_inum,
 838                                               parent_inum, dir_index_cnt,
 839                                               &cur_inode);
 840                         if (ret) {
 841                                 fprintf(stderr, "add_inode_items failed\n");
 842                                 goto fail;
 843                         }
 844
 845                         ret = add_xattr_item(trans, root,
 846                                              cur_inum, cur_file->d_name);
 847                         if (ret) {
 848                                 fprintf(stderr, "add_xattr_item failed\n");
 849                                 if(ret != -ENOTSUP)
 850                                         goto fail;
 851                         }
 852
 853                         if (S_ISDIR(st.st_mode)) {
 854                                 dir_entry = malloc(sizeof(struct directory_name_entry));
 855                                 dir_entry->dir_name = cur_file->d_name;
 856                                 dir_entry->path = make_path(parent_dir_entry->path,
 857                                                             cur_file->d_name);
 858                                 dir_entry->inum = cur_inum;
 859                                 list_add_tail(&dir_entry->list, &dir_head->list);
 860                         } else if (S_ISREG(st.st_mode)) {
 861                                 ret = add_file_items(trans, root, &cur_inode,
 862                                                      cur_inum, parent_inum, &st,
 863                                                      cur_file->d_name, out_fd);
 864                                 if (ret) {
 865                                         fprintf(stderr, "add_file_items failed\n");
 866                                         goto fail;
 867                                 }
 868                         } else if (S_ISLNK(st.st_mode)) {
 869                                 ret = add_symbolic_link(trans, root,
 870                                                         cur_inum, cur_file->d_name);
 871                                 if (ret) {
 872                                         fprintf(stderr, "add_symbolic_link failed\n");
 873                                         goto fail;
 874                                 }
 875                         }
 876                 }
 877
 878                 free_namelist(files, count);
 879                 free(parent_dir_entry->path);
 880                 free(parent_dir_entry);
 881
 882                 index_cnt = 2;
 883
 884         } while (!list_empty(&dir_head->list));
 885
 886         return 0;
 887 fail:
 888         free_namelist(files, count);
 889 fail_no_files:
 890         free(parent_dir_entry->path);
 891         free(parent_dir_entry);
 892         return -1;
 893 }
 894
 895 static int open_target(char *output_name)
 896 {
 897         int output_fd;
 898         output_fd = open(output_name, O_CREAT | O_RDWR | O_TRUNC,
 899                          S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH);
 900
 901         return output_fd;
 902 }
 903
 904 static int create_chunks(struct btrfs_trans_handle *trans,
 905                          struct btrfs_root *root, u64 num_of_meta_chunks,
 906                          u64 size_of_data)
 907 {
 908         u64 chunk_start;
 909         u64 chunk_size;
 910         u64 meta_type = BTRFS_BLOCK_GROUP_METADATA;
 911         u64 data_type = BTRFS_BLOCK_GROUP_DATA;
 912         u64 minimum_data_chunk_size = 8 * 1024 * 1024;
 913         u64 i;
 914         int ret;
 915
 916         for (i = 0; i < num_of_meta_chunks; i++) {
 917                 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
 918                                         &chunk_start, &chunk_size, meta_type);
 919                 BUG_ON(ret);
 920                 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
 921                                              meta_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
 922                                              chunk_start, chunk_size);
 923                 BUG_ON(ret);
 924                 set_extent_dirty(&root->fs_info->free_space_cache,
 925                                  chunk_start, chunk_start + chunk_size - 1, 0);
 926         }
 927
 928         if (size_of_data < minimum_data_chunk_size)
 929                 size_of_data = minimum_data_chunk_size;
 930
 931         ret = btrfs_alloc_data_chunk(trans, root->fs_info->extent_root,
 932                                      &chunk_start, size_of_data, data_type);
 933         BUG_ON(ret);
 934         ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
 935                                      data_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
 936                                      chunk_start, size_of_data);
 937         BUG_ON(ret);
 938         set_extent_dirty(&root->fs_info->free_space_cache,
 939                          chunk_start, chunk_start + size_of_data - 1, 0);
 940         return ret;
 941 }
 942
 943 static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
 944 {
 945         int ret;
 946         struct btrfs_trans_handle *trans;
 947
 948         struct stat root_st;
 949
 950         struct directory_name_entry dir_head;
 951
 952         struct directory_name_entry *dir_entry = NULL;
 953
 954         ret = lstat(source_dir, &root_st);
 955         if (ret) {
 956                 fprintf(stderr, "unable to lstat the %s\n", source_dir);
 957                 goto fail;
 958         }
 959
 960         INIT_LIST_HEAD(&dir_head.list);
 961
 962         trans = btrfs_start_transaction(root, 1);
 963         ret = traverse_directory(trans, root, source_dir, &dir_head, out_fd);
 964         if (ret) {
 965                 fprintf(stderr, "unable to traverse_directory\n");
 966                 goto fail;
 967         }
 968         btrfs_commit_transaction(trans, root);
 969
 970         printf("Making image is completed.\n");
 971         return 0;
 972 fail:
 973         while (!list_empty(&dir_head.list)) {
 974                 dir_entry = list_entry(dir_head.list.next,
 975                                        struct directory_name_entry, list);
 976                 list_del(&dir_entry->list);
 977                 free(dir_entry);
 978         }
 979         fprintf(stderr, "Making image is aborted.\n");
 980         return -1;
 981 }
 982
 983 /*
 984  * This ignores symlinks with unreadable targets and subdirs that can't
 985  * be read.  It's a best-effort to give a rough estimate of the size of
 986  * a subdir.  It doesn't guarantee that prepopulating btrfs from this
 987  * tree won't still run out of space.
 988  *
 989  * The rounding up to 4096 is questionable.  Previous code used du -B 4096.
 990  */
 991 static u64 global_total_size;
 992 static int ftw_add_entry_size(const char *fpath, const struct stat *st,
 993                               int type)
 994 {
 995         if (type == FTW_F || type == FTW_D)
 996                 global_total_size += round_up(st->st_size, 4096);
 997
 998         return 0;
 999 }
1000
1001 static u64 size_sourcedir(char *dir_name, u64 sectorsize,
1002                           u64 *num_of_meta_chunks_ret, u64 *size_of_data_ret)
1003 {
1004         u64 dir_size = 0;
1005         u64 total_size = 0;
1006         int ret;
1007         u64 default_chunk_size = 8 * 1024 * 1024;       /* 8MB */
1008         u64 allocated_meta_size = 8 * 1024 * 1024;      /* 8MB */
1009         u64 allocated_total_size = 20 * 1024 * 1024;    /* 20MB */
1010         u64 num_of_meta_chunks = 0;
1011         u64 num_of_data_chunks = 0;
1012         u64 num_of_allocated_meta_chunks =
1013                         allocated_meta_size / default_chunk_size;
1014
1015         global_total_size = 0;
1016         ret = ftw(dir_name, ftw_add_entry_size, 10);
1017         dir_size = global_total_size;
1018         if (ret < 0) {
1019                 fprintf(stderr, "ftw subdir walk of '%s' failed: %s\n",
1020                         dir_name, strerror(errno));
1021                 exit(1);
1022         }
1023
1024         num_of_data_chunks = (dir_size + default_chunk_size - 1) /
1025                 default_chunk_size;
1026
1027         num_of_meta_chunks = (dir_size / 2) / default_chunk_size;
1028         if (((dir_size / 2) % default_chunk_size) != 0)
1029                 num_of_meta_chunks++;
1030         if (num_of_meta_chunks <= num_of_allocated_meta_chunks)
1031                 num_of_meta_chunks = 0;
1032         else
1033                 num_of_meta_chunks -= num_of_allocated_meta_chunks;
1034
1035         total_size = allocated_total_size +
1036                      (num_of_data_chunks * default_chunk_size) +
1037                      (num_of_meta_chunks * default_chunk_size);
1038
1039         *num_of_meta_chunks_ret = num_of_meta_chunks;
1040         *size_of_data_ret = num_of_data_chunks * default_chunk_size;
1041         return total_size;
1042 }
1043
1044 static int zero_output_file(int out_fd, u64 size, u32 sectorsize)
1045 {
1046         int len = sectorsize;
1047         int loop_num = size / sectorsize;
1048         u64 location = 0;
1049         char *buf = malloc(len);
1050         int ret = 0, i;
1051         ssize_t written;
1052
1053         if (!buf)
1054                 return -ENOMEM;
1055         memset(buf, 0, len);
1056         for (i = 0; i < loop_num; i++) {
1057                 written = pwrite64(out_fd, buf, len, location);
1058                 if (written != len)
1059                         ret = -EIO;
1060                 location += sectorsize;
1061         }
1062         free(buf);
1063         return ret;
1064 }
1065
1066 static int check_leaf_or_node_size(u32 size, u32 sectorsize)
1067 {
1068         if (size < sectorsize) {
1069                 fprintf(stderr,
1070                         "Illegal leafsize (or nodesize) %u (smaller than %u)\n",
1071                         size, sectorsize);
1072                 return -1;
1073         } else if (size > BTRFS_MAX_METADATA_BLOCKSIZE) {
1074                 fprintf(stderr,
1075                         "Illegal leafsize (or nodesize) %u (larger than %u)\n",
1076                         size, BTRFS_MAX_METADATA_BLOCKSIZE);
1077                 return -1;
1078         } else if (size & (sectorsize - 1)) {
1079                 fprintf(stderr,
1080                         "Illegal leafsize (or nodesize) %u (not align to %u)\n",
1081                         size, sectorsize);
1082                 return -1;
1083         }
1084         return 0;
1085 }
1086
1087 static int is_ssd(const char *file)
1088 {
1089         blkid_probe probe;
1090         char wholedisk[32];
1091         char sysfs_path[PATH_MAX];
1092         dev_t devno;
1093         int fd;
1094         char rotational;
1095         int ret;
1096
1097         probe = blkid_new_probe_from_filename(file);
1098         if (!probe)
1099                 return 0;
1100
1101         /* Device number of this disk (possibly a partition) */
1102         devno = blkid_probe_get_devno(probe);
1103         if (!devno) {
1104                 blkid_free_probe(probe);
1105                 return 0;
1106         }
1107
1108         /* Get whole disk name (not full path) for this devno */
1109         ret = blkid_devno_to_wholedisk(devno,
1110                         wholedisk, sizeof(wholedisk), NULL);
1111         if (ret) {
1112                 blkid_free_probe(probe);
1113                 return 0;
1114         }
1115
1116         snprintf(sysfs_path, PATH_MAX, "/sys/block/%s/queue/rotational",
1117                  wholedisk);
1118
1119         blkid_free_probe(probe);
1120
1121         fd = open(sysfs_path, O_RDONLY);
1122         if (fd < 0) {
1123                 return 0;
1124         }
1125
1126         if (read(fd, &rotational, sizeof(char)) < sizeof(char)) {
1127                 close(fd);
1128                 return 0;
1129         }
1130         close(fd);
1131
1132         return !atoi((const char *)&rotational);
1133 }
1134
1135 #define BTRFS_FEATURE_LIST_ALL          (1ULL << 63)
1136
1137 static const struct btrfs_fs_feature {
1138         const char *name;
1139         u64 flag;
1140         const char *desc;
1141 } mkfs_features[] = {
1142         { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1143                 "mixed data and metadata block groups" },
1144         { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1145                 "increased hardlink limit per file to 65536" },
1146         { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1147                 "raid56 extended format" },
1148         { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1149                 "reduced-size metadata extent refs" },
1150         /* Keep this one last */
1151         { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1152 };
1153
1154 static void list_all_fs_features(void)
1155 {
1156         int i;
1157
1158         fprintf(stderr, "Filesystem features available at mkfs time:\n");
1159         for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1160                 char *is_default = "";
1161
1162                 if (mkfs_features[i].flag & DEFAULT_MKFS_FEATURES)
1163                         is_default = ", default";
1164                 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
1165                                 mkfs_features[i].name,
1166                                 mkfs_features[i].desc,
1167                                 mkfs_features[i].flag,
1168                                 is_default);
1169         }
1170 }
1171
1172 static int parse_one_fs_feature(const char *name, u64 *flags)
1173 {
1174         int i;
1175         int found = 0;
1176
1177         for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1178                 if (name[0] == '^' &&
1179                         !strcmp(mkfs_features[i].name, name + 1)) {
1180                         *flags &= ~ mkfs_features[i].flag;
1181                         found = 1;
1182                 } else if (!strcmp(mkfs_features[i].name, name)) {
1183                         *flags |= mkfs_features[i].flag;
1184                         found = 1;
1185                 }
1186         }
1187
1188         return !found;
1189 }
1190
1191 static void process_fs_features(u64 flags)
1192 {
1193         int i;
1194
1195         for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1196                 if (flags & mkfs_features[i].flag) {
1197                         fprintf(stderr,
1198                                 "Turning ON incompat feature '%s': %s\n",
1199                                 mkfs_features[i].name,
1200                                 mkfs_features[i].desc);
1201                 }
1202         }
1203 }
1204
1205
1206 /*
1207  * Return NULL if all features were parsed fine, otherwise return the name of
1208  * the first unparsed.
1209  */
1210 static char* parse_fs_features(char *namelist, u64 *flags)
1211 {
1212         char *this_char;
1213         char *save_ptr = NULL; /* Satisfy static checkers */
1214
1215         for (this_char = strtok_r(namelist, ",", &save_ptr);
1216              this_char != NULL;
1217              this_char = strtok_r(NULL, ",", &save_ptr)) {
1218                 if (parse_one_fs_feature(this_char, flags))
1219                         return this_char;
1220         }
1221
1222         return NULL;
1223 }
1224
1225 int main(int ac, char **av)
1226 {
1227         char *file;
1228         struct btrfs_root *root;
1229         struct btrfs_trans_handle *trans;
1230         char *label = NULL;
1231         char *first_file;
1232         u64 block_count = 0;
1233         u64 dev_block_count = 0;
1234         u64 blocks[7];
1235         u64 alloc_start = 0;
1236         u64 metadata_profile = 0;
1237         u64 data_profile = 0;
1238         u32 leafsize = max_t(u32, sysconf(_SC_PAGESIZE), DEFAULT_MKFS_LEAF_SIZE);
1239         u32 sectorsize = 4096;
1240         u32 nodesize = leafsize;
1241         u32 stripesize = 4096;
1242         int zero_end = 1;
1243         int option_index = 0;
1244         int fd;
1245         int ret;
1246         int i;
1247         int mixed = 0;
1248         int leaf_forced = 0;
1249         int data_profile_opt = 0;
1250         int metadata_profile_opt = 0;
1251         int discard = 1;
1252         int ssd = 0;
1253         int force_overwrite = 0;
1254
1255         char *source_dir = NULL;
1256         int source_dir_set = 0;
1257         u64 num_of_meta_chunks = 0;
1258         u64 size_of_data = 0;
1259         u64 source_dir_size = 0;
1260         int dev_cnt = 0;
1261         int saved_optind;
1262         char estr[100];
1263         u64 features = DEFAULT_MKFS_FEATURES;
1264
1265         while(1) {
1266                 int c;
1267                 c = getopt_long(ac, av, "A:b:fl:n:s:m:d:L:O:r:VMK",
1268                                 long_options, &option_index);
1269                 if (c < 0)
1270                         break;
1271                 switch(c) {
1272                         case 'A':
1273                                 alloc_start = parse_size(optarg);
1274                                 break;
1275                         case 'f':
1276                                 force_overwrite = 1;
1277                                 break;
1278                         case 'd':
1279                                 data_profile = parse_profile(optarg);
1280                                 data_profile_opt = 1;
1281                                 break;
1282                         case 'l':
1283                         case 'n':
1284                                 nodesize = parse_size(optarg);
1285                                 leafsize = parse_size(optarg);
1286                                 leaf_forced = 1;
1287                                 break;
1288                         case 'L':
1289                                 label = parse_label(optarg);
1290                                 break;
1291                         case 'm':
1292                                 metadata_profile = parse_profile(optarg);
1293                                 metadata_profile_opt = 1;
1294                                 break;
1295                         case 'M':
1296                                 mixed = 1;
1297                                 break;
1298                         case 'O': {
1299                                 char *orig = strdup(optarg);
1300                                 char *tmp = orig;
1301
1302                                 tmp = parse_fs_features(tmp, &features);
1303                                 if (tmp) {
1304                                         fprintf(stderr,
1305                                                 "Unrecognized filesystem feature '%s'\n",
1306                                                         tmp);
1307                                         free(orig);
1308                                         exit(1);
1309                                 }
1310                                 free(orig);
1311                                 if (features & BTRFS_FEATURE_LIST_ALL) {
1312                                         list_all_fs_features();
1313                                         exit(0);
1314                                 }
1315                                 break;
1316                                 }
1317                         case 's':
1318                                 sectorsize = parse_size(optarg);
1319                                 break;
1320                         case 'b':
1321                                 block_count = parse_size(optarg);
1322                                 if (block_count <= 1024*1024*1024) {
1323                                         printf("SMALL VOLUME: forcing mixed "
1324                                                "metadata/data groups\n");
1325                                         mixed = 1;
1326                                 }
1327                                 zero_end = 0;
1328                                 break;
1329                         case 'V':
1330                                 print_version();
1331                                 break;
1332                         case 'r':
1333                                 source_dir = optarg;
1334                                 source_dir_set = 1;
1335                                 break;
1336                         case 'K':
1337                                 discard = 0;
1338                                 break;
1339                         default:
1340                                 print_usage();
1341                 }
1342         }
1343         sectorsize = max(sectorsize, (u32)sysconf(_SC_PAGESIZE));
1344         if (check_leaf_or_node_size(leafsize, sectorsize))
1345                 exit(1);
1346         if (check_leaf_or_node_size(nodesize, sectorsize))
1347                 exit(1);
1348         saved_optind = optind;
1349         dev_cnt = ac - optind;
1350         if (dev_cnt == 0)
1351                 print_usage();
1352
1353         if (source_dir_set && dev_cnt > 1) {
1354                 fprintf(stderr,
1355                         "The -r option is limited to a single device\n");
1356                 exit(1);
1357         }
1358         while (dev_cnt-- > 0) {
1359                 file = av[optind++];
1360                 if (is_block_device(file))
1361                         if (test_dev_for_mkfs(file, force_overwrite, estr)) {
1362                                 fprintf(stderr, "Error: %s", estr);
1363                                 exit(1);
1364                         }
1365         }
1366
1367         optind = saved_optind;
1368         dev_cnt = ac - optind;
1369
1370         file = av[optind++];
1371         ssd = is_ssd(file);
1372
1373         if (is_vol_small(file)) {
1374                 printf("SMALL VOLUME: forcing mixed metadata/data groups\n");
1375                 mixed = 1;
1376         }
1377
1378         /*
1379         * Set default profiles according to number of added devices.
1380         * For mixed groups defaults are single/single.
1381         */
1382         if (!mixed) {
1383                 if (!metadata_profile_opt) {
1384                         if (dev_cnt == 1 && ssd)
1385                                 printf("Detected a SSD, turning off metadata "
1386                                 "duplication.  Mkfs with -m dup if you want to "
1387                                 "force metadata duplication.\n");
1388
1389                         metadata_profile = (dev_cnt > 1) ?
1390                                         BTRFS_BLOCK_GROUP_RAID1 : (ssd) ?
1391                                         0: BTRFS_BLOCK_GROUP_DUP;
1392                 }
1393                 if (!data_profile_opt) {
1394                         data_profile = (dev_cnt > 1) ?
1395                                 BTRFS_BLOCK_GROUP_RAID0 : 0; /* raid0 or single */
1396                 }
1397         } else {
1398                 u32 best_leafsize = max_t(u32, sysconf(_SC_PAGESIZE), sectorsize);
1399
1400                 if (metadata_profile_opt || data_profile_opt) {
1401                         if (metadata_profile != data_profile) {
1402                                 fprintf(stderr,
1403         "ERROR: With mixed block groups data and metadata profiles must be the same\n");
1404                                 exit(1);
1405                         }
1406                 }
1407
1408                 if (!leaf_forced) {
1409                         leafsize = best_leafsize;
1410                         nodesize = best_leafsize;
1411                         if (check_leaf_or_node_size(leafsize, sectorsize))
1412                                 exit(1);
1413                 }
1414                 if (leafsize != sectorsize) {
1415                         fprintf(stderr, "Error: mixed metadata/data block groups "
1416                                 "require metadata blocksizes equal to the sectorsize\n");
1417                         exit(1);
1418                 }
1419         }
1420
1421         ret = test_num_disk_vs_raid(metadata_profile, data_profile,
1422                         dev_cnt, mixed, estr);
1423         if (ret) {
1424                 fprintf(stderr, "Error: %s\n", estr);
1425                 exit(1);
1426         }
1427
1428         /* if we are here that means all devs are good to btrfsify */
1429         printf("\nWARNING! - %s IS EXPERIMENTAL\n", BTRFS_BUILD_VERSION);
1430         printf("WARNING! - see http://btrfs.wiki.kernel.org before using\n\n");
1431
1432         dev_cnt--;
1433
1434         if (!source_dir_set) {
1435                 /*
1436                  * open without O_EXCL so that the problem should not
1437                  * occur by the following processing.
1438                  * (btrfs_register_one_device() fails if O_EXCL is on)
1439                  */
1440                 fd = open(file, O_RDWR);
1441                 if (fd < 0) {
1442                         fprintf(stderr, "unable to open %s: %s\n", file,
1443                                 strerror(errno));
1444                         exit(1);
1445                 }
1446                 first_file = file;
1447                 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1448                                            block_count, &mixed, discard);
1449                 if (block_count && block_count > dev_block_count) {
1450                         fprintf(stderr, "%s is smaller than requested size\n", file);
1451                         exit(1);
1452                 }
1453         } else {
1454                 fd = open_target(file);
1455                 if (fd < 0) {
1456                         fprintf(stderr, "unable to open the %s\n", file);
1457                         exit(1);
1458                 }
1459
1460                 first_file = file;
1461                 source_dir_size = size_sourcedir(source_dir, sectorsize,
1462                                              &num_of_meta_chunks, &size_of_data);
1463                 if(block_count < source_dir_size)
1464                         block_count = source_dir_size;
1465                 ret = zero_output_file(fd, block_count, sectorsize);
1466                 if (ret) {
1467                         fprintf(stderr, "unable to zero the output file\n");
1468                         exit(1);
1469                 }
1470                 /* our "device" is the new image file */
1471                 dev_block_count = block_count;
1472         }
1473
1474         /* To create the first block group and chunk 0 in make_btrfs */
1475         if (dev_block_count < BTRFS_MKFS_SYSTEM_GROUP_SIZE) {
1476                 fprintf(stderr, "device is too small to make filesystem\n");
1477                 exit(1);
1478         }
1479
1480         blocks[0] = BTRFS_SUPER_INFO_OFFSET;
1481         for (i = 1; i < 7; i++) {
1482                 blocks[i] = BTRFS_SUPER_INFO_OFFSET + 1024 * 1024 +
1483                         leafsize * i;
1484         }
1485
1486         /*
1487          * FS features that can be set by other means than -O
1488          * just set the bit here
1489          */
1490         if (mixed)
1491                 features |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
1492
1493         if ((data_profile | metadata_profile) &
1494             (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
1495                 features |= BTRFS_FEATURE_INCOMPAT_RAID56;
1496         }
1497
1498         process_fs_features(features);
1499
1500         ret = make_btrfs(fd, file, label, blocks, dev_block_count,
1501                          nodesize, leafsize,
1502                          sectorsize, stripesize, features);
1503         if (ret) {
1504                 fprintf(stderr, "error during mkfs: %s\n", strerror(-ret));
1505                 exit(1);
1506         }
1507
1508         root = open_ctree(file, 0, OPEN_CTREE_WRITES);
1509         if (!root) {
1510                 fprintf(stderr, "Open ctree failed\n");
1511                 close(fd);
1512                 exit(1);
1513         }
1514         root->fs_info->alloc_start = alloc_start;
1515
1516         ret = make_root_dir(root, mixed);
1517         if (ret) {
1518                 fprintf(stderr, "failed to setup the root directory\n");
1519                 exit(1);
1520         }
1521
1522         trans = btrfs_start_transaction(root, 1);
1523
1524         if (dev_cnt == 0)
1525                 goto raid_groups;
1526
1527         btrfs_register_one_device(file);
1528
1529         zero_end = 1;
1530         while (dev_cnt-- > 0) {
1531                 int old_mixed = mixed;
1532
1533                 file = av[optind++];
1534
1535                 /*
1536                  * open without O_EXCL so that the problem should not
1537                  * occur by the following processing.
1538                  * (btrfs_register_one_device() fails if O_EXCL is on)
1539                  */
1540                 fd = open(file, O_RDWR);
1541                 if (fd < 0) {
1542                         fprintf(stderr, "unable to open %s: %s\n", file,
1543                                 strerror(errno));
1544                         exit(1);
1545                 }
1546                 ret = btrfs_device_already_in_root(root, fd,
1547                                                    BTRFS_SUPER_INFO_OFFSET);
1548                 if (ret) {
1549                         fprintf(stderr, "skipping duplicate device %s in FS\n",
1550                                 file);
1551                         close(fd);
1552                         continue;
1553                 }
1554                 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1555                                            block_count, &mixed, discard);
1556                 mixed = old_mixed;
1557                 BUG_ON(ret);
1558
1559                 ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
1560                                         sectorsize, sectorsize, sectorsize);
1561                 BUG_ON(ret);
1562                 btrfs_register_one_device(file);
1563         }
1564
1565 raid_groups:
1566         if (!source_dir_set) {
1567                 ret = create_raid_groups(trans, root, data_profile,
1568                                  data_profile_opt, metadata_profile,
1569                                  metadata_profile_opt, mixed, ssd);
1570                 BUG_ON(ret);
1571         }
1572
1573         ret = create_data_reloc_tree(trans, root);
1574         BUG_ON(ret);
1575
1576         printf("fs created label %s on %s\n\tnodesize %u leafsize %u "
1577             "sectorsize %u size %s\n",
1578             label, first_file, nodesize, leafsize, sectorsize,
1579             pretty_size(btrfs_super_total_bytes(root->fs_info->super_copy)));
1580
1581         printf("%s\n", BTRFS_BUILD_VERSION);
1582         btrfs_commit_transaction(trans, root);
1583
1584         if (source_dir_set) {
1585                 trans = btrfs_start_transaction(root, 1);
1586                 ret = create_chunks(trans, root,
1587                                     num_of_meta_chunks, size_of_data);
1588                 BUG_ON(ret);
1589                 btrfs_commit_transaction(trans, root);
1590
1591                 ret = make_image(source_dir, root, fd);
1592                 BUG_ON(ret);
1593         }
1594
1595         ret = close_ctree(root);
1596         BUG_ON(ret);
1597         free(label);
1598         return 0;
1599 }