2 * Copyright (C) 2007 Oracle. All rights reserved.
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.
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.
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.
19 #define _XOPEN_SOURCE 500
22 #include "kerncompat.h"
25 #include <sys/ioctl.h>
26 #include <sys/mount.h>
32 #include <sys/types.h>
38 #include <uuid/uuid.h>
41 #include <attr/xattr.h>
42 #include <blkid/blkid.h>
47 #include "transaction.h"
51 static u64 index_cnt = 2;
53 struct directory_name_entry {
57 struct list_head list;
60 static int make_root_dir(struct btrfs_root *root, int mixed)
62 struct btrfs_trans_handle *trans;
63 struct btrfs_key location;
69 trans = btrfs_start_transaction(root, 1);
70 bytes_used = btrfs_super_bytes_used(root->fs_info->super_copy);
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);
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);
85 ret = btrfs_make_block_group(trans, root, 0,
86 BTRFS_BLOCK_GROUP_METADATA |
87 BTRFS_BLOCK_GROUP_DATA,
88 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
89 chunk_start, chunk_size);
91 printf("Created a data/metadata chunk of size %llu\n", chunk_size);
93 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
94 &chunk_start, &chunk_size,
95 BTRFS_BLOCK_GROUP_METADATA);
97 ret = btrfs_make_block_group(trans, root, 0,
98 BTRFS_BLOCK_GROUP_METADATA,
99 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
100 chunk_start, chunk_size);
104 root->fs_info->system_allocs = 0;
105 btrfs_commit_transaction(trans, root);
106 trans = btrfs_start_transaction(root, 1);
110 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
111 &chunk_start, &chunk_size,
112 BTRFS_BLOCK_GROUP_DATA);
114 ret = btrfs_make_block_group(trans, root, 0,
115 BTRFS_BLOCK_GROUP_DATA,
116 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
117 chunk_start, chunk_size);
121 ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
122 BTRFS_ROOT_TREE_DIR_OBJECTID);
125 ret = btrfs_make_root_dir(trans, root, BTRFS_FIRST_FREE_OBJECTID);
128 memcpy(&location, &root->fs_info->fs_root->root_key, sizeof(location));
129 location.offset = (u64)-1;
130 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
132 btrfs_super_root_dir(root->fs_info->super_copy),
133 &location, BTRFS_FT_DIR, 0);
137 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
138 "default", 7, location.objectid,
139 BTRFS_ROOT_TREE_DIR_OBJECTID, 0);
143 btrfs_commit_transaction(trans, root);
148 static int recow_roots(struct btrfs_trans_handle *trans,
149 struct btrfs_root *root)
152 struct extent_buffer *tmp;
153 struct btrfs_fs_info *info = root->fs_info;
155 ret = __btrfs_cow_block(trans, info->fs_root, info->fs_root->node,
156 NULL, 0, &tmp, 0, 0);
158 free_extent_buffer(tmp);
160 ret = __btrfs_cow_block(trans, info->tree_root, info->tree_root->node,
161 NULL, 0, &tmp, 0, 0);
163 free_extent_buffer(tmp);
165 ret = __btrfs_cow_block(trans, info->extent_root,
166 info->extent_root->node, NULL, 0, &tmp, 0, 0);
168 free_extent_buffer(tmp);
170 ret = __btrfs_cow_block(trans, info->chunk_root, info->chunk_root->node,
171 NULL, 0, &tmp, 0, 0);
173 free_extent_buffer(tmp);
176 ret = __btrfs_cow_block(trans, info->dev_root, info->dev_root->node,
177 NULL, 0, &tmp, 0, 0);
179 free_extent_buffer(tmp);
181 ret = __btrfs_cow_block(trans, info->csum_root, info->csum_root->node,
182 NULL, 0, &tmp, 0, 0);
184 free_extent_buffer(tmp);
189 static int create_one_raid_group(struct btrfs_trans_handle *trans,
190 struct btrfs_root *root, u64 type)
196 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
197 &chunk_start, &chunk_size, type);
199 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
200 type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
201 chunk_start, chunk_size);
206 static int create_raid_groups(struct btrfs_trans_handle *trans,
207 struct btrfs_root *root, u64 data_profile,
208 int data_profile_opt, u64 metadata_profile,
209 int metadata_profile_opt, int mixed, int ssd)
211 u64 num_devices = btrfs_super_num_devices(root->fs_info->super_copy);
213 u64 devices_for_raid = num_devices;
217 * Set default profiles according to number of added devices.
218 * For mixed groups defaults are single/single.
220 if (!metadata_profile_opt && !mixed) {
221 if (num_devices == 1 && ssd)
222 printf("Detected a SSD, turning off metadata "
223 "duplication. Mkfs with -m dup if you want to "
224 "force metadata duplication.\n");
225 metadata_profile = (num_devices > 1) ?
226 BTRFS_BLOCK_GROUP_RAID1 : (ssd) ? 0: BTRFS_BLOCK_GROUP_DUP;
228 if (!data_profile_opt && !mixed) {
229 data_profile = (num_devices > 1) ?
230 BTRFS_BLOCK_GROUP_RAID0 : 0; /* raid0 or single */
233 if (devices_for_raid > 4)
234 devices_for_raid = 4;
236 switch (devices_for_raid) {
239 allowed |= BTRFS_BLOCK_GROUP_RAID10;
241 allowed |= BTRFS_BLOCK_GROUP_RAID6;
243 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
244 BTRFS_BLOCK_GROUP_RAID5;
247 allowed |= BTRFS_BLOCK_GROUP_DUP;
250 if (metadata_profile & ~allowed) {
251 fprintf(stderr, "unable to create FS with metadata "
252 "profile %llu (have %llu devices)\n", metadata_profile,
256 if (data_profile & ~allowed) {
257 fprintf(stderr, "unable to create FS with data "
258 "profile %llu (have %llu devices)\n", data_profile,
263 /* allow dup'ed data chunks only in mixed mode */
264 if (!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP)) {
265 fprintf(stderr, "dup for data is allowed only in mixed mode\n");
269 if (allowed & metadata_profile) {
270 u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
272 ret = create_one_raid_group(trans, root,
273 BTRFS_BLOCK_GROUP_SYSTEM |
274 (allowed & metadata_profile));
278 meta_flags |= BTRFS_BLOCK_GROUP_DATA;
280 ret = create_one_raid_group(trans, root, meta_flags |
281 (allowed & metadata_profile));
284 ret = recow_roots(trans, root);
287 if (!mixed && num_devices > 1 && (allowed & data_profile)) {
288 ret = create_one_raid_group(trans, root,
289 BTRFS_BLOCK_GROUP_DATA |
290 (allowed & data_profile));
296 static int create_data_reloc_tree(struct btrfs_trans_handle *trans,
297 struct btrfs_root *root)
299 struct btrfs_key location;
300 struct btrfs_root_item root_item;
301 struct extent_buffer *tmp;
302 u64 objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
305 ret = btrfs_copy_root(trans, root, root->node, &tmp, objectid);
308 memcpy(&root_item, &root->root_item, sizeof(root_item));
309 btrfs_set_root_bytenr(&root_item, tmp->start);
310 btrfs_set_root_level(&root_item, btrfs_header_level(tmp));
311 btrfs_set_root_generation(&root_item, trans->transid);
312 free_extent_buffer(tmp);
314 location.objectid = objectid;
315 location.type = BTRFS_ROOT_ITEM_KEY;
317 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
318 &location, &root_item);
323 static void print_usage(void)
325 fprintf(stderr, "usage: mkfs.btrfs [options] dev [ dev ... ]\n");
326 fprintf(stderr, "options:\n");
327 fprintf(stderr, "\t -A --alloc-start the offset to start the FS\n");
328 fprintf(stderr, "\t -b --byte-count total number of bytes in the FS\n");
329 fprintf(stderr, "\t -d --data data profile, raid0, raid1, raid5, raid6, raid10, dup or single\n");
330 fprintf(stderr, "\t -f --force force overwrite of existing filesystem\n");
331 fprintf(stderr, "\t -l --leafsize size of btree leaves\n");
332 fprintf(stderr, "\t -L --label set a label\n");
333 fprintf(stderr, "\t -m --metadata metadata profile, values like data profile\n");
334 fprintf(stderr, "\t -M --mixed mix metadata and data together\n");
335 fprintf(stderr, "\t -n --nodesize size of btree nodes\n");
336 fprintf(stderr, "\t -s --sectorsize min block allocation\n");
337 fprintf(stderr, "\t -r --rootdir the source directory\n");
338 fprintf(stderr, "\t -K --nodiscard do not perform whole device TRIM\n");
339 fprintf(stderr, "\t -V --version print the mkfs.btrfs version and exit\n");
340 fprintf(stderr, "\t -x --skinny-extents use the new skinny extent disk format");
341 fprintf(stderr, "%s\n", BTRFS_BUILD_VERSION);
345 static void print_version(void)
347 fprintf(stderr, "mkfs.btrfs, part of %s\n", BTRFS_BUILD_VERSION);
351 static u64 parse_profile(char *s)
353 if (strcmp(s, "raid0") == 0) {
354 return BTRFS_BLOCK_GROUP_RAID0;
355 } else if (strcmp(s, "raid1") == 0) {
356 return BTRFS_BLOCK_GROUP_RAID1;
357 } else if (strcmp(s, "raid5") == 0) {
358 return BTRFS_BLOCK_GROUP_RAID5;
359 } else if (strcmp(s, "raid6") == 0) {
360 return BTRFS_BLOCK_GROUP_RAID6;
361 } else if (strcmp(s, "raid10") == 0) {
362 return BTRFS_BLOCK_GROUP_RAID10;
363 } else if (strcmp(s, "dup") == 0) {
364 return BTRFS_BLOCK_GROUP_DUP;
365 } else if (strcmp(s, "single") == 0) {
368 fprintf(stderr, "Unknown profile %s\n", s);
375 static char *parse_label(char *input)
377 int len = strlen(input);
379 if (len >= BTRFS_LABEL_SIZE) {
380 fprintf(stderr, "Label %s is too long (max %d)\n", input,
381 BTRFS_LABEL_SIZE - 1);
384 return strdup(input);
387 static struct option long_options[] = {
388 { "alloc-start", 1, NULL, 'A'},
389 { "byte-count", 1, NULL, 'b' },
390 { "force", 0, NULL, 'f' },
391 { "leafsize", 1, NULL, 'l' },
392 { "label", 1, NULL, 'L'},
393 { "metadata", 1, NULL, 'm' },
394 { "mixed", 0, NULL, 'M' },
395 { "nodesize", 1, NULL, 'n' },
396 { "sectorsize", 1, NULL, 's' },
397 { "data", 1, NULL, 'd' },
398 { "version", 0, NULL, 'V' },
399 { "rootdir", 1, NULL, 'r' },
400 { "nodiscard", 0, NULL, 'K' },
401 { "skinny-extents", 0, NULL, 'x'},
405 static int add_directory_items(struct btrfs_trans_handle *trans,
406 struct btrfs_root *root, u64 objectid,
407 ino_t parent_inum, const char *name,
408 struct stat *st, int *dir_index_cnt)
412 struct btrfs_key location;
415 name_len = strlen(name);
417 location.objectid = objectid;
419 btrfs_set_key_type(&location, BTRFS_INODE_ITEM_KEY);
421 if (S_ISDIR(st->st_mode))
422 filetype = BTRFS_FT_DIR;
423 if (S_ISREG(st->st_mode))
424 filetype = BTRFS_FT_REG_FILE;
425 if (S_ISLNK(st->st_mode))
426 filetype = BTRFS_FT_SYMLINK;
428 ret = btrfs_insert_dir_item(trans, root, name, name_len,
429 parent_inum, &location,
430 filetype, index_cnt);
432 *dir_index_cnt = index_cnt;
438 static int fill_inode_item(struct btrfs_trans_handle *trans,
439 struct btrfs_root *root,
440 struct btrfs_inode_item *dst, struct stat *src)
443 u64 sectorsize = root->sectorsize;
446 * btrfs_inode_item has some reserved fields
447 * and represents on-disk inode entry, so
448 * zero everything to prevent information leak
450 memset(dst, 0, sizeof (*dst));
452 btrfs_set_stack_inode_generation(dst, trans->transid);
453 btrfs_set_stack_inode_size(dst, src->st_size);
454 btrfs_set_stack_inode_nbytes(dst, 0);
455 btrfs_set_stack_inode_block_group(dst, 0);
456 btrfs_set_stack_inode_nlink(dst, src->st_nlink);
457 btrfs_set_stack_inode_uid(dst, src->st_uid);
458 btrfs_set_stack_inode_gid(dst, src->st_gid);
459 btrfs_set_stack_inode_mode(dst, src->st_mode);
460 btrfs_set_stack_inode_rdev(dst, 0);
461 btrfs_set_stack_inode_flags(dst, 0);
462 btrfs_set_stack_timespec_sec(&dst->atime, src->st_atime);
463 btrfs_set_stack_timespec_nsec(&dst->atime, 0);
464 btrfs_set_stack_timespec_sec(&dst->ctime, src->st_ctime);
465 btrfs_set_stack_timespec_nsec(&dst->ctime, 0);
466 btrfs_set_stack_timespec_sec(&dst->mtime, src->st_mtime);
467 btrfs_set_stack_timespec_nsec(&dst->mtime, 0);
468 btrfs_set_stack_timespec_sec(&dst->otime, 0);
469 btrfs_set_stack_timespec_nsec(&dst->otime, 0);
471 if (S_ISDIR(src->st_mode)) {
472 btrfs_set_stack_inode_size(dst, 0);
473 btrfs_set_stack_inode_nlink(dst, 1);
475 if (S_ISREG(src->st_mode)) {
476 btrfs_set_stack_inode_size(dst, (u64)src->st_size);
477 if (src->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root))
478 btrfs_set_stack_inode_nbytes(dst, src->st_size);
480 blocks = src->st_size / sectorsize;
481 if (src->st_size % sectorsize)
483 blocks *= sectorsize;
484 btrfs_set_stack_inode_nbytes(dst, blocks);
487 if (S_ISLNK(src->st_mode))
488 btrfs_set_stack_inode_nbytes(dst, src->st_size + 1);
493 static int directory_select(const struct direct *entry)
495 if ((strncmp(entry->d_name, ".", entry->d_reclen) == 0) ||
496 (strncmp(entry->d_name, "..", entry->d_reclen) == 0))
502 static void free_namelist(struct direct **files, int count)
509 for (i = 0; i < count; ++i)
514 static u64 calculate_dir_inode_size(char *dirname)
517 struct direct **files, *cur_file;
518 u64 dir_inode_size = 0;
520 count = scandir(dirname, &files, directory_select, NULL);
522 for (i = 0; i < count; i++) {
524 dir_inode_size += strlen(cur_file->d_name);
527 free_namelist(files, count);
530 return dir_inode_size;
533 static int add_inode_items(struct btrfs_trans_handle *trans,
534 struct btrfs_root *root,
535 struct stat *st, char *name,
536 u64 self_objectid, ino_t parent_inum,
537 int dir_index_cnt, struct btrfs_inode_item *inode_ret)
540 struct btrfs_key inode_key;
541 struct btrfs_inode_item btrfs_inode;
546 name_len = strlen(name);
547 fill_inode_item(trans, root, &btrfs_inode, st);
548 objectid = self_objectid;
550 if (S_ISDIR(st->st_mode)) {
551 inode_size = calculate_dir_inode_size(name);
552 btrfs_set_stack_inode_size(&btrfs_inode, inode_size);
555 inode_key.objectid = objectid;
556 inode_key.offset = 0;
557 btrfs_set_key_type(&inode_key, BTRFS_INODE_ITEM_KEY);
559 ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
563 ret = btrfs_insert_inode_ref(trans, root, name, name_len,
564 objectid, parent_inum, dir_index_cnt);
568 *inode_ret = btrfs_inode;
573 static int add_xattr_item(struct btrfs_trans_handle *trans,
574 struct btrfs_root *root, u64 objectid,
575 const char *file_name)
579 char xattr_list[XATTR_LIST_MAX];
581 char cur_value[XATTR_SIZE_MAX];
582 char delimiter = '\0';
583 char *next_location = xattr_list;
585 ret = llistxattr(file_name, xattr_list, XATTR_LIST_MAX);
589 fprintf(stderr, "get a list of xattr failed for %s\n",
596 cur_name = strtok(xattr_list, &delimiter);
597 while (cur_name != NULL) {
598 cur_name_len = strlen(cur_name);
599 next_location += cur_name_len + 1;
601 ret = getxattr(file_name, cur_name, cur_value, XATTR_SIZE_MAX);
605 fprintf(stderr, "get a xattr value failed for %s attr %s\n",
606 file_name, cur_name);
610 ret = btrfs_insert_xattr_item(trans, root, cur_name,
611 cur_name_len, cur_value,
614 fprintf(stderr, "insert a xattr item failed for %s\n",
618 cur_name = strtok(next_location, &delimiter);
623 static int custom_alloc_extent(struct btrfs_root *root, u64 num_bytes,
624 u64 hint_byte, struct btrfs_key *ins)
628 u64 last = hint_byte;
631 struct btrfs_block_group_cache *cache;
634 ret = find_first_extent_bit(&root->fs_info->free_space_cache,
635 last, &start, &end, EXTENT_DIRTY);
637 if (wrapped++ == 0) {
645 start = max(last, start);
647 if (last - start < num_bytes)
650 last = start + num_bytes;
651 if (test_range_bit(&root->fs_info->pinned_extents,
652 start, last - 1, EXTENT_DIRTY, 0))
655 cache = btrfs_lookup_block_group(root->fs_info, start);
657 if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM ||
658 last > cache->key.objectid + cache->key.offset) {
659 last = cache->key.objectid + cache->key.offset;
663 if (cache->flags & (BTRFS_BLOCK_GROUP_SYSTEM |
664 BTRFS_BLOCK_GROUP_METADATA)) {
665 last = cache->key.objectid + cache->key.offset;
669 clear_extent_dirty(&root->fs_info->free_space_cache,
670 start, start + num_bytes - 1, 0);
672 ins->objectid = start;
673 ins->offset = num_bytes;
674 ins->type = BTRFS_EXTENT_ITEM_KEY;
678 fprintf(stderr, "not enough free space\n");
682 static int record_file_extent(struct btrfs_trans_handle *trans,
683 struct btrfs_root *root, u64 objectid,
684 struct btrfs_inode_item *inode,
685 u64 file_pos, u64 disk_bytenr,
689 struct btrfs_fs_info *info = root->fs_info;
690 struct btrfs_root *extent_root = info->extent_root;
691 struct extent_buffer *leaf;
692 struct btrfs_file_extent_item *fi;
693 struct btrfs_key ins_key;
694 struct btrfs_path path;
695 struct btrfs_extent_item *ei;
697 btrfs_init_path(&path);
699 ins_key.objectid = objectid;
701 btrfs_set_key_type(&ins_key, BTRFS_EXTENT_DATA_KEY);
702 ret = btrfs_insert_empty_item(trans, root, &path, &ins_key,
706 leaf = path.nodes[0];
707 fi = btrfs_item_ptr(leaf, path.slots[0],
708 struct btrfs_file_extent_item);
709 btrfs_set_file_extent_generation(leaf, fi, trans->transid);
710 btrfs_set_file_extent_type(leaf, fi, BTRFS_FILE_EXTENT_REG);
711 btrfs_set_file_extent_disk_bytenr(leaf, fi, disk_bytenr);
712 btrfs_set_file_extent_disk_num_bytes(leaf, fi, num_bytes);
713 btrfs_set_file_extent_offset(leaf, fi, 0);
714 btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
715 btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
716 btrfs_set_file_extent_compression(leaf, fi, 0);
717 btrfs_set_file_extent_encryption(leaf, fi, 0);
718 btrfs_set_file_extent_other_encoding(leaf, fi, 0);
719 btrfs_mark_buffer_dirty(leaf);
721 btrfs_release_path(root, &path);
723 ins_key.objectid = disk_bytenr;
724 ins_key.offset = num_bytes;
725 ins_key.type = BTRFS_EXTENT_ITEM_KEY;
727 ret = btrfs_insert_empty_item(trans, extent_root, &path,
728 &ins_key, sizeof(*ei));
730 leaf = path.nodes[0];
731 ei = btrfs_item_ptr(leaf, path.slots[0],
732 struct btrfs_extent_item);
734 btrfs_set_extent_refs(leaf, ei, 0);
735 btrfs_set_extent_generation(leaf, ei, trans->transid);
736 btrfs_set_extent_flags(leaf, ei, BTRFS_EXTENT_FLAG_DATA);
738 btrfs_mark_buffer_dirty(leaf);
739 ret = btrfs_update_block_group(trans, root, disk_bytenr,
743 } else if (ret != -EEXIST) {
747 ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, num_bytes, 0,
748 root->root_key.objectid,
751 btrfs_release_path(root, &path);
755 static int add_symbolic_link(struct btrfs_trans_handle *trans,
756 struct btrfs_root *root,
757 u64 objectid, const char *path_name)
760 u64 sectorsize = root->sectorsize;
761 char *buf = malloc(sectorsize);
763 ret = readlink(path_name, buf, sectorsize);
765 fprintf(stderr, "readlink failed for %s\n", path_name);
768 if (ret >= sectorsize) {
769 fprintf(stderr, "symlink too long for %s", path_name);
774 buf[ret] = '\0'; /* readlink does not do it for us */
775 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
782 static int add_file_items(struct btrfs_trans_handle *trans,
783 struct btrfs_root *root,
784 struct btrfs_inode_item *btrfs_inode, u64 objectid,
785 ino_t parent_inum, struct stat *st,
786 const char *path_name, int out_fd)
792 struct btrfs_key key;
794 u32 sectorsize = root->sectorsize;
799 fd = open(path_name, O_RDONLY);
801 fprintf(stderr, "%s open failed\n", path_name);
805 blocks = st->st_size / sectorsize;
806 if (st->st_size % sectorsize)
809 if (st->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
810 buffer = malloc(st->st_size);
811 ret_read = pread64(fd, buffer, st->st_size, bytes_read);
812 if (ret_read == -1) {
813 fprintf(stderr, "%s read failed\n", path_name);
817 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
818 buffer, st->st_size);
822 ret = custom_alloc_extent(root, blocks * sectorsize, 0, &key);
826 first_block = key.objectid;
828 buffer = malloc(sectorsize);
831 memset(buffer, 0, sectorsize);
832 ret_read = pread64(fd, buffer, sectorsize, bytes_read);
833 if (ret_read == -1) {
834 fprintf(stderr, "%s read failed\n", path_name);
838 ret = pwrite64(out_fd, buffer, sectorsize,
839 first_block + bytes_read);
840 if (ret != sectorsize) {
841 fprintf(stderr, "output file write failed\n");
845 /* checksum for file data */
846 ret = btrfs_csum_file_block(trans, root->fs_info->csum_root,
847 first_block + (blocks * sectorsize),
848 first_block + bytes_read,
851 fprintf(stderr, "%s checksum failed\n", path_name);
855 bytes_read += ret_read;
857 } while (ret_read == sectorsize);
859 if (num_blocks > 0) {
860 ret = record_file_extent(trans, root, objectid, btrfs_inode,
861 first_block, first_block,
862 blocks * sectorsize);
874 static char *make_path(char *dir, char *name)
878 path = malloc(strlen(dir) + strlen(name) + 2);
882 if (dir[strlen(dir) - 1] != '/')
888 static int traverse_directory(struct btrfs_trans_handle *trans,
889 struct btrfs_root *root, char *dir_name,
890 struct directory_name_entry *dir_head, int out_fd)
894 struct btrfs_inode_item cur_inode;
895 struct btrfs_inode_item *inode_item;
896 int count, i, dir_index_cnt;
897 struct direct **files;
899 struct directory_name_entry *dir_entry, *parent_dir_entry;
900 struct direct *cur_file;
901 ino_t parent_inum, cur_inum;
902 ino_t highest_inum = 0;
903 char *parent_dir_name;
904 struct btrfs_path path;
905 struct extent_buffer *leaf;
906 struct btrfs_key root_dir_key;
907 u64 root_dir_inode_size = 0;
909 /* Add list for source directory */
910 dir_entry = malloc(sizeof(struct directory_name_entry));
911 dir_entry->dir_name = dir_name;
912 dir_entry->path = strdup(dir_name);
914 parent_inum = highest_inum + BTRFS_FIRST_FREE_OBJECTID;
915 dir_entry->inum = parent_inum;
916 list_add_tail(&dir_entry->list, &dir_head->list);
918 btrfs_init_path(&path);
920 root_dir_key.objectid = btrfs_root_dirid(&root->root_item);
921 root_dir_key.offset = 0;
922 btrfs_set_key_type(&root_dir_key, BTRFS_INODE_ITEM_KEY);
923 ret = btrfs_lookup_inode(trans, root, &path, &root_dir_key, 1);
925 fprintf(stderr, "root dir lookup error\n");
929 leaf = path.nodes[0];
930 inode_item = btrfs_item_ptr(leaf, path.slots[0],
931 struct btrfs_inode_item);
933 root_dir_inode_size = calculate_dir_inode_size(dir_name);
934 btrfs_set_inode_size(leaf, inode_item, root_dir_inode_size);
935 btrfs_mark_buffer_dirty(leaf);
937 btrfs_release_path(root, &path);
940 parent_dir_entry = list_entry(dir_head->list.next,
941 struct directory_name_entry,
943 list_del(&parent_dir_entry->list);
945 parent_inum = parent_dir_entry->inum;
946 parent_dir_name = parent_dir_entry->dir_name;
947 if (chdir(parent_dir_entry->path)) {
948 fprintf(stderr, "chdir error for %s\n",
953 count = scandir(parent_dir_entry->path, &files,
954 directory_select, NULL);
957 fprintf(stderr, "scandir for %s failed: %s\n",
958 parent_dir_name, strerror (errno));
962 for (i = 0; i < count; i++) {
965 if (lstat(cur_file->d_name, &st) == -1) {
966 fprintf(stderr, "lstat failed for file %s\n",
971 cur_inum = ++highest_inum + BTRFS_FIRST_FREE_OBJECTID;
972 ret = add_directory_items(trans, root,
973 cur_inum, parent_inum,
975 &st, &dir_index_cnt);
977 fprintf(stderr, "add_directory_items failed\n");
981 ret = add_inode_items(trans, root, &st,
982 cur_file->d_name, cur_inum,
983 parent_inum, dir_index_cnt,
986 fprintf(stderr, "add_inode_items failed\n");
990 ret = add_xattr_item(trans, root,
991 cur_inum, cur_file->d_name);
993 fprintf(stderr, "add_xattr_item failed\n");
998 if (S_ISDIR(st.st_mode)) {
999 dir_entry = malloc(sizeof(struct directory_name_entry));
1000 dir_entry->dir_name = cur_file->d_name;
1001 dir_entry->path = make_path(parent_dir_entry->path,
1003 dir_entry->inum = cur_inum;
1004 list_add_tail(&dir_entry->list, &dir_head->list);
1005 } else if (S_ISREG(st.st_mode)) {
1006 ret = add_file_items(trans, root, &cur_inode,
1007 cur_inum, parent_inum, &st,
1008 cur_file->d_name, out_fd);
1010 fprintf(stderr, "add_file_items failed\n");
1013 } else if (S_ISLNK(st.st_mode)) {
1014 ret = add_symbolic_link(trans, root,
1015 cur_inum, cur_file->d_name);
1017 fprintf(stderr, "add_symbolic_link failed\n");
1023 free_namelist(files, count);
1024 free(parent_dir_entry->path);
1025 free(parent_dir_entry);
1029 } while (!list_empty(&dir_head->list));
1033 free_namelist(files, count);
1035 free(parent_dir_entry->path);
1036 free(parent_dir_entry);
1040 static int open_target(char *output_name)
1043 output_fd = open(output_name, O_CREAT | O_RDWR | O_TRUNC,
1044 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH);
1049 static int create_chunks(struct btrfs_trans_handle *trans,
1050 struct btrfs_root *root, u64 num_of_meta_chunks,
1055 u64 meta_type = BTRFS_BLOCK_GROUP_METADATA;
1056 u64 data_type = BTRFS_BLOCK_GROUP_DATA;
1057 u64 minimum_data_chunk_size = 8 * 1024 * 1024;
1061 for (i = 0; i < num_of_meta_chunks; i++) {
1062 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
1063 &chunk_start, &chunk_size, meta_type);
1065 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
1066 meta_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
1067 chunk_start, chunk_size);
1069 set_extent_dirty(&root->fs_info->free_space_cache,
1070 chunk_start, chunk_start + chunk_size - 1, 0);
1073 if (size_of_data < minimum_data_chunk_size)
1074 size_of_data = minimum_data_chunk_size;
1075 ret = btrfs_alloc_data_chunk(trans, root->fs_info->extent_root,
1076 &chunk_start, size_of_data, data_type);
1078 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
1079 data_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
1080 chunk_start, size_of_data);
1082 set_extent_dirty(&root->fs_info->free_space_cache,
1083 chunk_start, chunk_start + size_of_data - 1, 0);
1087 static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
1090 struct btrfs_trans_handle *trans;
1092 struct stat root_st;
1094 struct directory_name_entry dir_head;
1096 ret = lstat(source_dir, &root_st);
1098 fprintf(stderr, "unable to lstat the %s\n", source_dir);
1102 INIT_LIST_HEAD(&dir_head.list);
1104 trans = btrfs_start_transaction(root, 1);
1105 ret = traverse_directory(trans, root, source_dir, &dir_head, out_fd);
1107 fprintf(stderr, "unable to traverse_directory\n");
1110 btrfs_commit_transaction(trans, root);
1112 printf("Making image is completed.\n");
1115 fprintf(stderr, "Making image is aborted.\n");
1120 * This ignores symlinks with unreadable targets and subdirs that can't
1121 * be read. It's a best-effort to give a rough estimate of the size of
1122 * a subdir. It doesn't guarantee that prepopulating btrfs from this
1123 * tree won't still run out of space.
1125 * The rounding up to 4096 is questionable. Previous code used du -B 4096.
1127 static u64 global_total_size;
1128 static int ftw_add_entry_size(const char *fpath, const struct stat *st,
1131 if (type == FTW_F || type == FTW_D)
1132 global_total_size += round_up(st->st_size, 4096);
1137 static u64 size_sourcedir(char *dir_name, u64 sectorsize,
1138 u64 *num_of_meta_chunks_ret, u64 *size_of_data_ret)
1143 u64 default_chunk_size = 8 * 1024 * 1024; /* 8MB */
1144 u64 allocated_meta_size = 8 * 1024 * 1024; /* 8MB */
1145 u64 allocated_total_size = 20 * 1024 * 1024; /* 20MB */
1146 u64 num_of_meta_chunks = 0;
1147 u64 num_of_allocated_meta_chunks =
1148 allocated_meta_size / default_chunk_size;
1150 global_total_size = 0;
1151 ret = ftw(dir_name, ftw_add_entry_size, 10);
1152 dir_size = global_total_size;
1154 fprintf(stderr, "ftw subdir walk of '%s' failed: %s\n",
1155 dir_name, strerror(errno));
1159 num_of_meta_chunks = (dir_size / 2) / default_chunk_size;
1160 if (((dir_size / 2) % default_chunk_size) != 0)
1161 num_of_meta_chunks++;
1162 if (num_of_meta_chunks <= num_of_allocated_meta_chunks)
1163 num_of_meta_chunks = 0;
1165 num_of_meta_chunks -= num_of_allocated_meta_chunks;
1167 total_size = allocated_total_size + dir_size +
1168 (num_of_meta_chunks * default_chunk_size);
1170 *num_of_meta_chunks_ret = num_of_meta_chunks;
1175 static int zero_output_file(int out_fd, u64 size, u32 sectorsize)
1177 int len = sectorsize;
1178 int loop_num = size / sectorsize;
1180 char *buf = malloc(len);
1186 memset(buf, 0, len);
1187 for (i = 0; i < loop_num; i++) {
1188 written = pwrite64(out_fd, buf, len, location);
1191 location += sectorsize;
1197 static int check_leaf_or_node_size(u32 size, u32 sectorsize)
1199 if (size < sectorsize) {
1201 "Illegal leafsize (or nodesize) %u (smaller than %u)\n",
1204 } else if (size > BTRFS_MAX_METADATA_BLOCKSIZE) {
1206 "Illegal leafsize (or nodesize) %u (larger than %u)\n",
1207 size, BTRFS_MAX_METADATA_BLOCKSIZE);
1209 } else if (size & (sectorsize - 1)) {
1211 "Illegal leafsize (or nodesize) %u (not align to %u)\n",
1218 static int is_ssd(const char *file)
1223 char path[PATH_MAX];
1228 probe = blkid_new_probe_from_filename(file);
1233 * We want to use blkid_devno_to_wholedisk() but it's broken for some
1234 * reason on F17 at least so we'll do this trickery
1236 disk = blkid_probe_get_wholedisk_devno(probe);
1240 devname = blkid_devno_to_devname(disk);
1244 dev = strrchr(devname, '/');
1247 snprintf(path, PATH_MAX, "/sys/block/%s/queue/rotational", dev);
1250 blkid_free_probe(probe);
1252 fd = open(path, O_RDONLY);
1257 if (read(fd, &rotational, sizeof(char)) < sizeof(char)) {
1263 return !atoi((const char *)&rotational);
1266 int main(int ac, char **av)
1269 struct btrfs_root *root;
1270 struct btrfs_trans_handle *trans;
1273 u64 block_count = 0;
1274 u64 dev_block_count = 0;
1276 u64 alloc_start = 0;
1277 u64 metadata_profile = 0;
1278 u64 data_profile = 0;
1279 u32 leafsize = sysconf(_SC_PAGESIZE);
1280 u32 sectorsize = 4096;
1281 u32 nodesize = leafsize;
1282 u32 stripesize = 4096;
1284 int option_index = 0;
1289 int data_profile_opt = 0;
1290 int metadata_profile_opt = 0;
1293 int force_overwrite = 0;
1294 int skinny_meta_extents = 0;
1296 char *source_dir = NULL;
1297 int source_dir_set = 0;
1298 u64 num_of_meta_chunks = 0;
1299 u64 size_of_data = 0;
1300 u64 source_dir_size = 0;
1302 struct btrfs_super_block *super;
1310 c = getopt_long(ac, av, "A:b:fl:n:s:m:d:L:r:VMKx",
1311 long_options, &option_index);
1316 alloc_start = parse_size(optarg);
1319 force_overwrite = 1;
1322 data_profile = parse_profile(optarg);
1323 data_profile_opt = 1;
1327 nodesize = parse_size(optarg);
1328 leafsize = parse_size(optarg);
1331 label = parse_label(optarg);
1334 metadata_profile = parse_profile(optarg);
1335 metadata_profile_opt = 1;
1341 sectorsize = parse_size(optarg);
1344 block_count = parse_size(optarg);
1345 if (block_count <= 1024*1024*1024) {
1346 printf("SMALL VOLUME: forcing mixed "
1347 "metadata/data groups\n");
1356 source_dir = optarg;
1363 skinny_meta_extents = 1;
1369 sectorsize = max(sectorsize, (u32)sysconf(_SC_PAGESIZE));
1370 if (check_leaf_or_node_size(leafsize, sectorsize))
1372 if (check_leaf_or_node_size(nodesize, sectorsize))
1374 saved_optind = optind;
1375 dev_cnt = ac - optind;
1379 if (source_dir_set && dev_cnt > 1) {
1381 "The -r option is limited to a single device\n");
1384 while (dev_cnt-- > 0) {
1385 file = av[optind++];
1386 if (is_block_device(file))
1387 if (test_dev_for_mkfs(file, force_overwrite, estr)) {
1388 fprintf(stderr, "Error: %s", estr);
1393 /* if we are here that means all devs are good to btrfsify */
1394 optind = saved_optind;
1395 dev_cnt = ac - optind;
1397 printf("\nWARNING! - %s IS EXPERIMENTAL\n", BTRFS_BUILD_VERSION);
1398 printf("WARNING! - see http://btrfs.wiki.kernel.org before using\n\n");
1400 file = av[optind++];
1403 if (!source_dir_set) {
1405 * open without O_EXCL so that the problem should not
1406 * occur by the following processing.
1407 * (btrfs_register_one_device() fails if O_EXCL is on)
1409 fd = open(file, O_RDWR);
1411 fprintf(stderr, "unable to open %s: %s\n", file,
1416 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1417 block_count, &mixed, nodiscard);
1418 if (block_count && block_count > dev_block_count) {
1419 fprintf(stderr, "%s is smaller than requested size\n", file);
1423 fd = open_target(file);
1425 fprintf(stderr, "unable to open the %s\n", file);
1430 source_dir_size = size_sourcedir(source_dir, sectorsize,
1431 &num_of_meta_chunks, &size_of_data);
1432 if(block_count < source_dir_size)
1433 block_count = source_dir_size;
1434 ret = zero_output_file(fd, block_count, sectorsize);
1436 fprintf(stderr, "unable to zero the output file\n");
1439 /* our "device" is the new image file */
1440 dev_block_count = block_count;
1446 if (metadata_profile != data_profile) {
1447 fprintf(stderr, "With mixed block groups data and metadata "
1448 "profiles must be the same\n");
1453 blocks[0] = BTRFS_SUPER_INFO_OFFSET;
1454 for (i = 1; i < 7; i++) {
1455 blocks[i] = BTRFS_SUPER_INFO_OFFSET + 1024 * 1024 +
1459 ret = make_btrfs(fd, file, label, blocks, dev_block_count,
1461 sectorsize, stripesize);
1463 fprintf(stderr, "error during mkfs %d\n", ret);
1467 root = open_ctree(file, 0, O_RDWR);
1469 fprintf(stderr, "Open ctree failed\n");
1473 root->fs_info->alloc_start = alloc_start;
1475 ret = make_root_dir(root, mixed);
1477 fprintf(stderr, "failed to setup the root directory\n");
1481 trans = btrfs_start_transaction(root, 1);
1486 btrfs_register_one_device(file);
1489 while (dev_cnt-- > 0) {
1490 int old_mixed = mixed;
1492 file = av[optind++];
1495 * open without O_EXCL so that the problem should not
1496 * occur by the following processing.
1497 * (btrfs_register_one_device() fails if O_EXCL is on)
1499 fd = open(file, O_RDWR);
1501 fprintf(stderr, "unable to open %s: %s\n", file,
1505 ret = btrfs_device_already_in_root(root, fd,
1506 BTRFS_SUPER_INFO_OFFSET);
1508 fprintf(stderr, "skipping duplicate device %s in FS\n",
1513 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1514 block_count, &mixed, nodiscard);
1518 ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
1519 sectorsize, sectorsize, sectorsize);
1521 btrfs_register_one_device(file);
1525 if (!source_dir_set) {
1526 ret = create_raid_groups(trans, root, data_profile,
1527 data_profile_opt, metadata_profile,
1528 metadata_profile_opt, mixed, ssd);
1532 ret = create_data_reloc_tree(trans, root);
1535 super = root->fs_info->super_copy;
1536 flags = btrfs_super_incompat_flags(super);
1537 flags |= BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF;
1540 flags |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
1542 if (skinny_meta_extents)
1543 flags |= BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA;
1545 btrfs_set_super_incompat_flags(super, flags);
1547 if ((data_profile | metadata_profile) &
1548 (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
1549 struct btrfs_super_block *super = root->fs_info->super_copy;
1550 u64 flags = btrfs_super_incompat_flags(super);
1552 flags |= BTRFS_FEATURE_INCOMPAT_RAID56;
1553 btrfs_set_super_incompat_flags(super, flags);
1554 printf("Setting RAID5/6 feature flag\n");
1557 printf("fs created label %s on %s\n\tnodesize %u leafsize %u "
1558 "sectorsize %u size %s\n",
1559 label, first_file, nodesize, leafsize, sectorsize,
1560 pretty_buf = pretty_sizes(btrfs_super_total_bytes(root->fs_info->super_copy)));
1563 printf("%s\n", BTRFS_BUILD_VERSION);
1564 btrfs_commit_transaction(trans, root);
1566 if (source_dir_set) {
1567 trans = btrfs_start_transaction(root, 1);
1568 ret = create_chunks(trans, root,
1569 num_of_meta_chunks, size_of_data);
1571 btrfs_commit_transaction(trans, root);
1573 ret = make_image(source_dir, root, fd);
1577 ret = close_ctree(root);