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>
44 #include "kerncompat.h"
48 #include "transaction.h"
52 static u64 index_cnt = 2;
54 struct directory_name_entry {
58 struct list_head list;
61 static int make_root_dir(struct btrfs_root *root, int mixed)
63 struct btrfs_trans_handle *trans;
64 struct btrfs_key location;
70 trans = btrfs_start_transaction(root, 1);
71 bytes_used = btrfs_super_bytes_used(&root->fs_info->super_copy);
73 root->fs_info->system_allocs = 1;
74 ret = btrfs_make_block_group(trans, root, bytes_used,
75 BTRFS_BLOCK_GROUP_SYSTEM,
76 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
77 0, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
81 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
82 &chunk_start, &chunk_size,
83 BTRFS_BLOCK_GROUP_METADATA |
84 BTRFS_BLOCK_GROUP_DATA);
86 ret = btrfs_make_block_group(trans, root, 0,
87 BTRFS_BLOCK_GROUP_METADATA |
88 BTRFS_BLOCK_GROUP_DATA,
89 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
90 chunk_start, chunk_size);
92 printf("Created a data/metadata chunk of size %llu\n", chunk_size);
94 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
95 &chunk_start, &chunk_size,
96 BTRFS_BLOCK_GROUP_METADATA);
98 ret = btrfs_make_block_group(trans, root, 0,
99 BTRFS_BLOCK_GROUP_METADATA,
100 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
101 chunk_start, chunk_size);
105 root->fs_info->system_allocs = 0;
106 btrfs_commit_transaction(trans, root);
107 trans = btrfs_start_transaction(root, 1);
111 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
112 &chunk_start, &chunk_size,
113 BTRFS_BLOCK_GROUP_DATA);
115 ret = btrfs_make_block_group(trans, root, 0,
116 BTRFS_BLOCK_GROUP_DATA,
117 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
118 chunk_start, chunk_size);
122 ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
123 BTRFS_ROOT_TREE_DIR_OBJECTID);
126 ret = btrfs_make_root_dir(trans, root, BTRFS_FIRST_FREE_OBJECTID);
129 memcpy(&location, &root->fs_info->fs_root->root_key, sizeof(location));
130 location.offset = (u64)-1;
131 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
133 btrfs_super_root_dir(&root->fs_info->super_copy),
134 &location, BTRFS_FT_DIR, 0);
138 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
139 "default", 7, location.objectid,
140 BTRFS_ROOT_TREE_DIR_OBJECTID, 0);
144 btrfs_commit_transaction(trans, root);
149 static int recow_roots(struct btrfs_trans_handle *trans,
150 struct btrfs_root *root)
153 struct extent_buffer *tmp;
154 struct btrfs_fs_info *info = root->fs_info;
156 ret = __btrfs_cow_block(trans, info->fs_root, info->fs_root->node,
157 NULL, 0, &tmp, 0, 0);
159 free_extent_buffer(tmp);
161 ret = __btrfs_cow_block(trans, info->tree_root, info->tree_root->node,
162 NULL, 0, &tmp, 0, 0);
164 free_extent_buffer(tmp);
166 ret = __btrfs_cow_block(trans, info->extent_root,
167 info->extent_root->node, NULL, 0, &tmp, 0, 0);
169 free_extent_buffer(tmp);
171 ret = __btrfs_cow_block(trans, info->chunk_root, info->chunk_root->node,
172 NULL, 0, &tmp, 0, 0);
174 free_extent_buffer(tmp);
177 ret = __btrfs_cow_block(trans, info->dev_root, info->dev_root->node,
178 NULL, 0, &tmp, 0, 0);
180 free_extent_buffer(tmp);
182 ret = __btrfs_cow_block(trans, info->csum_root, info->csum_root->node,
183 NULL, 0, &tmp, 0, 0);
185 free_extent_buffer(tmp);
190 static int create_one_raid_group(struct btrfs_trans_handle *trans,
191 struct btrfs_root *root, u64 type)
197 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
198 &chunk_start, &chunk_size, type);
200 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
201 type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
202 chunk_start, chunk_size);
207 static int create_raid_groups(struct btrfs_trans_handle *trans,
208 struct btrfs_root *root, u64 data_profile,
209 int data_profile_opt, u64 metadata_profile,
210 int metadata_profile_opt, int mixed, int ssd)
212 u64 num_devices = btrfs_super_num_devices(&root->fs_info->super_copy);
214 u64 devices_for_raid = num_devices;
218 * Set default profiles according to number of added devices.
219 * For mixed groups defaults are single/single.
221 if (!metadata_profile_opt && !mixed) {
222 if (num_devices == 1 && ssd)
223 printf("Detected a SSD, turning off metadata "
224 "duplication. Mkfs with -m dup if you want to "
225 "force metadata duplication.\n");
226 metadata_profile = (num_devices > 1) ?
227 BTRFS_BLOCK_GROUP_RAID1 : (ssd) ? 0: BTRFS_BLOCK_GROUP_DUP;
229 if (!data_profile_opt && !mixed) {
230 data_profile = (num_devices > 1) ?
231 BTRFS_BLOCK_GROUP_RAID0 : 0; /* raid0 or single */
234 if (devices_for_raid > 4)
235 devices_for_raid = 4;
237 switch (devices_for_raid) {
240 allowed |= BTRFS_BLOCK_GROUP_RAID10;
242 allowed |= BTRFS_BLOCK_GROUP_RAID6;
244 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
245 BTRFS_BLOCK_GROUP_RAID5;
248 allowed |= BTRFS_BLOCK_GROUP_DUP;
251 if (metadata_profile & ~allowed) {
252 fprintf(stderr, "unable to create FS with metadata "
253 "profile %llu (have %llu devices)\n", metadata_profile,
257 if (data_profile & ~allowed) {
258 fprintf(stderr, "unable to create FS with data "
259 "profile %llu (have %llu devices)\n", data_profile,
264 /* allow dup'ed data chunks only in mixed mode */
265 if (!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP)) {
266 fprintf(stderr, "dup for data is allowed only in mixed mode\n");
270 if (allowed & metadata_profile) {
271 u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
273 ret = create_one_raid_group(trans, root,
274 BTRFS_BLOCK_GROUP_SYSTEM |
275 (allowed & metadata_profile));
279 meta_flags |= BTRFS_BLOCK_GROUP_DATA;
281 ret = create_one_raid_group(trans, root, meta_flags |
282 (allowed & metadata_profile));
285 ret = recow_roots(trans, root);
288 if (!mixed && num_devices > 1 && (allowed & data_profile)) {
289 ret = create_one_raid_group(trans, root,
290 BTRFS_BLOCK_GROUP_DATA |
291 (allowed & data_profile));
297 static int create_data_reloc_tree(struct btrfs_trans_handle *trans,
298 struct btrfs_root *root)
300 struct btrfs_key location;
301 struct btrfs_root_item root_item;
302 struct extent_buffer *tmp;
303 u64 objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
306 ret = btrfs_copy_root(trans, root, root->node, &tmp, objectid);
309 memcpy(&root_item, &root->root_item, sizeof(root_item));
310 btrfs_set_root_bytenr(&root_item, tmp->start);
311 btrfs_set_root_level(&root_item, btrfs_header_level(tmp));
312 btrfs_set_root_generation(&root_item, trans->transid);
313 free_extent_buffer(tmp);
315 location.objectid = objectid;
316 location.type = BTRFS_ROOT_ITEM_KEY;
318 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
319 &location, &root_item);
324 static void print_usage(void)
326 fprintf(stderr, "usage: mkfs.btrfs [options] dev [ dev ... ]\n");
327 fprintf(stderr, "options:\n");
328 fprintf(stderr, "\t -A --alloc-start the offset to start the FS\n");
329 fprintf(stderr, "\t -b --byte-count total number of bytes in the FS\n");
330 fprintf(stderr, "\t -d --data data profile, raid0, raid1, raid10, dup or single\n");
331 fprintf(stderr, "\t -l --leafsize size of btree leaves\n");
332 fprintf(stderr, "\t -L --label set a label\n");
333 fprintf(stderr, "\t -m --metadata metadata profile, values like data profile\n");
334 fprintf(stderr, "\t -M --mixed mix metadata and data together\n");
335 fprintf(stderr, "\t -n --nodesize size of btree nodes\n");
336 fprintf(stderr, "\t -s --sectorsize min block allocation\n");
337 fprintf(stderr, "\t -r --rootdir the source directory\n");
338 fprintf(stderr, "\t -K --nodiscard do not perform whole device TRIM\n");
339 fprintf(stderr, "\t -V --version print the mkfs.btrfs version and exit\n");
340 fprintf(stderr, "%s\n", BTRFS_BUILD_VERSION);
344 static void print_version(void)
346 fprintf(stderr, "mkfs.btrfs, part of %s\n", BTRFS_BUILD_VERSION);
350 static u64 parse_profile(char *s)
352 if (strcmp(s, "raid0") == 0) {
353 return BTRFS_BLOCK_GROUP_RAID0;
354 } else if (strcmp(s, "raid1") == 0) {
355 return BTRFS_BLOCK_GROUP_RAID1;
356 } else if (strcmp(s, "raid5") == 0) {
357 return BTRFS_BLOCK_GROUP_RAID5;
358 } else if (strcmp(s, "raid6") == 0) {
359 return BTRFS_BLOCK_GROUP_RAID6;
360 } else if (strcmp(s, "raid10") == 0) {
361 return BTRFS_BLOCK_GROUP_RAID10;
362 } else if (strcmp(s, "dup") == 0) {
363 return BTRFS_BLOCK_GROUP_DUP;
364 } else if (strcmp(s, "single") == 0) {
367 fprintf(stderr, "Unknown profile %s\n", s);
374 static char *parse_label(char *input)
376 int len = strlen(input);
378 if (len >= BTRFS_LABEL_SIZE) {
379 fprintf(stderr, "Label %s is too long (max %d)\n", input,
380 BTRFS_LABEL_SIZE - 1);
383 return strdup(input);
386 static struct option long_options[] = {
387 { "alloc-start", 1, NULL, 'A'},
388 { "byte-count", 1, NULL, 'b' },
389 { "leafsize", 1, NULL, 'l' },
390 { "label", 1, NULL, 'L'},
391 { "metadata", 1, NULL, 'm' },
392 { "mixed", 0, NULL, 'M' },
393 { "nodesize", 1, NULL, 'n' },
394 { "sectorsize", 1, NULL, 's' },
395 { "data", 1, NULL, 'd' },
396 { "version", 0, NULL, 'V' },
397 { "rootdir", 1, NULL, 'r' },
398 { "nodiscard", 0, NULL, 'K' },
402 static int add_directory_items(struct btrfs_trans_handle *trans,
403 struct btrfs_root *root, u64 objectid,
404 ino_t parent_inum, const char *name,
405 struct stat *st, int *dir_index_cnt)
409 struct btrfs_key location;
412 name_len = strlen(name);
414 location.objectid = objectid;
416 btrfs_set_key_type(&location, BTRFS_INODE_ITEM_KEY);
418 if (S_ISDIR(st->st_mode))
419 filetype = BTRFS_FT_DIR;
420 if (S_ISREG(st->st_mode))
421 filetype = BTRFS_FT_REG_FILE;
422 if (S_ISLNK(st->st_mode))
423 filetype = BTRFS_FT_SYMLINK;
425 ret = btrfs_insert_dir_item(trans, root, name, name_len,
426 parent_inum, &location,
427 filetype, index_cnt);
429 *dir_index_cnt = index_cnt;
435 static int fill_inode_item(struct btrfs_trans_handle *trans,
436 struct btrfs_root *root,
437 struct btrfs_inode_item *dst, struct stat *src)
440 u64 sectorsize = root->sectorsize;
443 * btrfs_inode_item has some reserved fields
444 * and represents on-disk inode entry, so
445 * zero everything to prevent information leak
447 memset(dst, 0, sizeof (*dst));
449 btrfs_set_stack_inode_generation(dst, trans->transid);
450 btrfs_set_stack_inode_size(dst, src->st_size);
451 btrfs_set_stack_inode_nbytes(dst, 0);
452 btrfs_set_stack_inode_block_group(dst, 0);
453 btrfs_set_stack_inode_nlink(dst, src->st_nlink);
454 btrfs_set_stack_inode_uid(dst, src->st_uid);
455 btrfs_set_stack_inode_gid(dst, src->st_gid);
456 btrfs_set_stack_inode_mode(dst, src->st_mode);
457 btrfs_set_stack_inode_rdev(dst, 0);
458 btrfs_set_stack_inode_flags(dst, 0);
459 btrfs_set_stack_timespec_sec(&dst->atime, src->st_atime);
460 btrfs_set_stack_timespec_nsec(&dst->atime, 0);
461 btrfs_set_stack_timespec_sec(&dst->ctime, src->st_ctime);
462 btrfs_set_stack_timespec_nsec(&dst->ctime, 0);
463 btrfs_set_stack_timespec_sec(&dst->mtime, src->st_mtime);
464 btrfs_set_stack_timespec_nsec(&dst->mtime, 0);
465 btrfs_set_stack_timespec_sec(&dst->otime, 0);
466 btrfs_set_stack_timespec_nsec(&dst->otime, 0);
468 if (S_ISDIR(src->st_mode)) {
469 btrfs_set_stack_inode_size(dst, 0);
470 btrfs_set_stack_inode_nlink(dst, 1);
472 if (S_ISREG(src->st_mode)) {
473 btrfs_set_stack_inode_size(dst, (u64)src->st_size);
474 if (src->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root))
475 btrfs_set_stack_inode_nbytes(dst, src->st_size);
477 blocks = src->st_size / sectorsize;
478 if (src->st_size % sectorsize)
480 blocks *= sectorsize;
481 btrfs_set_stack_inode_nbytes(dst, blocks);
484 if (S_ISLNK(src->st_mode))
485 btrfs_set_stack_inode_nbytes(dst, src->st_size + 1);
490 static int directory_select(const struct direct *entry)
492 if ((strncmp(entry->d_name, ".", entry->d_reclen) == 0) ||
493 (strncmp(entry->d_name, "..", entry->d_reclen) == 0))
499 static void free_namelist(struct direct **files, int count)
506 for (i = 0; i < count; ++i)
511 static u64 calculate_dir_inode_size(char *dirname)
514 struct direct **files, *cur_file;
515 u64 dir_inode_size = 0;
517 count = scandir(dirname, &files, directory_select, NULL);
519 for (i = 0; i < count; i++) {
521 dir_inode_size += strlen(cur_file->d_name);
524 free_namelist(files, count);
527 return dir_inode_size;
530 static int add_inode_items(struct btrfs_trans_handle *trans,
531 struct btrfs_root *root,
532 struct stat *st, char *name,
533 u64 self_objectid, ino_t parent_inum,
534 int dir_index_cnt, struct btrfs_inode_item *inode_ret)
537 struct btrfs_key inode_key;
538 struct btrfs_inode_item btrfs_inode;
543 name_len = strlen(name);
544 fill_inode_item(trans, root, &btrfs_inode, st);
545 objectid = self_objectid;
547 if (S_ISDIR(st->st_mode)) {
548 inode_size = calculate_dir_inode_size(name);
549 btrfs_set_stack_inode_size(&btrfs_inode, inode_size);
552 inode_key.objectid = objectid;
553 inode_key.offset = 0;
554 btrfs_set_key_type(&inode_key, BTRFS_INODE_ITEM_KEY);
556 ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
560 ret = btrfs_insert_inode_ref(trans, root, name, name_len,
561 objectid, parent_inum, dir_index_cnt);
565 *inode_ret = btrfs_inode;
570 static int add_xattr_item(struct btrfs_trans_handle *trans,
571 struct btrfs_root *root, u64 objectid,
572 const char *file_name)
576 char xattr_list[XATTR_LIST_MAX];
578 char cur_value[XATTR_SIZE_MAX];
579 char delimiter = '\0';
580 char *next_location = xattr_list;
582 ret = llistxattr(file_name, xattr_list, XATTR_LIST_MAX);
586 fprintf(stderr, "get a list of xattr failed for %s\n",
593 cur_name = strtok(xattr_list, &delimiter);
594 while (cur_name != NULL) {
595 cur_name_len = strlen(cur_name);
596 next_location += cur_name_len + 1;
598 ret = getxattr(file_name, cur_name, cur_value, XATTR_SIZE_MAX);
602 fprintf(stderr, "get a xattr value failed for %s attr %s\n",
603 file_name, cur_name);
607 ret = btrfs_insert_xattr_item(trans, root, cur_name,
608 cur_name_len, cur_value,
611 fprintf(stderr, "insert a xattr item failed for %s\n",
615 cur_name = strtok(next_location, &delimiter);
620 static int custom_alloc_extent(struct btrfs_root *root, u64 num_bytes,
621 u64 hint_byte, struct btrfs_key *ins)
625 u64 last = hint_byte;
628 struct btrfs_block_group_cache *cache;
631 ret = find_first_extent_bit(&root->fs_info->free_space_cache,
632 last, &start, &end, EXTENT_DIRTY);
634 if (wrapped++ == 0) {
642 start = max(last, start);
644 if (last - start < num_bytes)
647 last = start + num_bytes;
648 if (test_range_bit(&root->fs_info->pinned_extents,
649 start, last - 1, EXTENT_DIRTY, 0))
652 cache = btrfs_lookup_block_group(root->fs_info, start);
654 if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM ||
655 last > cache->key.objectid + cache->key.offset) {
656 last = cache->key.objectid + cache->key.offset;
660 if (cache->flags & (BTRFS_BLOCK_GROUP_SYSTEM |
661 BTRFS_BLOCK_GROUP_METADATA)) {
662 last = cache->key.objectid + cache->key.offset;
666 clear_extent_dirty(&root->fs_info->free_space_cache,
667 start, start + num_bytes - 1, 0);
669 ins->objectid = start;
670 ins->offset = num_bytes;
671 ins->type = BTRFS_EXTENT_ITEM_KEY;
675 fprintf(stderr, "not enough free space\n");
679 static int record_file_extent(struct btrfs_trans_handle *trans,
680 struct btrfs_root *root, u64 objectid,
681 struct btrfs_inode_item *inode,
682 u64 file_pos, u64 disk_bytenr,
686 struct btrfs_fs_info *info = root->fs_info;
687 struct btrfs_root *extent_root = info->extent_root;
688 struct extent_buffer *leaf;
689 struct btrfs_file_extent_item *fi;
690 struct btrfs_key ins_key;
691 struct btrfs_path path;
692 struct btrfs_extent_item *ei;
694 btrfs_init_path(&path);
696 ins_key.objectid = objectid;
698 btrfs_set_key_type(&ins_key, BTRFS_EXTENT_DATA_KEY);
699 ret = btrfs_insert_empty_item(trans, root, &path, &ins_key,
703 leaf = path.nodes[0];
704 fi = btrfs_item_ptr(leaf, path.slots[0],
705 struct btrfs_file_extent_item);
706 btrfs_set_file_extent_generation(leaf, fi, trans->transid);
707 btrfs_set_file_extent_type(leaf, fi, BTRFS_FILE_EXTENT_REG);
708 btrfs_set_file_extent_disk_bytenr(leaf, fi, disk_bytenr);
709 btrfs_set_file_extent_disk_num_bytes(leaf, fi, num_bytes);
710 btrfs_set_file_extent_offset(leaf, fi, 0);
711 btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
712 btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
713 btrfs_set_file_extent_compression(leaf, fi, 0);
714 btrfs_set_file_extent_encryption(leaf, fi, 0);
715 btrfs_set_file_extent_other_encoding(leaf, fi, 0);
716 btrfs_mark_buffer_dirty(leaf);
718 btrfs_release_path(root, &path);
720 ins_key.objectid = disk_bytenr;
721 ins_key.offset = num_bytes;
722 ins_key.type = BTRFS_EXTENT_ITEM_KEY;
724 ret = btrfs_insert_empty_item(trans, extent_root, &path,
725 &ins_key, sizeof(*ei));
727 leaf = path.nodes[0];
728 ei = btrfs_item_ptr(leaf, path.slots[0],
729 struct btrfs_extent_item);
731 btrfs_set_extent_refs(leaf, ei, 0);
732 btrfs_set_extent_generation(leaf, ei, trans->transid);
733 btrfs_set_extent_flags(leaf, ei, BTRFS_EXTENT_FLAG_DATA);
735 btrfs_mark_buffer_dirty(leaf);
736 ret = btrfs_update_block_group(trans, root, disk_bytenr,
740 } else if (ret != -EEXIST) {
744 ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, num_bytes, 0,
745 root->root_key.objectid,
748 btrfs_release_path(root, &path);
752 static int add_symbolic_link(struct btrfs_trans_handle *trans,
753 struct btrfs_root *root,
754 u64 objectid, const char *path_name)
757 u64 sectorsize = root->sectorsize;
758 char *buf = malloc(sectorsize);
760 ret = readlink(path_name, buf, sectorsize);
762 fprintf(stderr, "readlink failed for %s\n", path_name);
765 if (ret >= sectorsize) {
766 fprintf(stderr, "symlink too long for %s", path_name);
771 buf[ret] = '\0'; /* readlink does not do it for us */
772 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
779 static int add_file_items(struct btrfs_trans_handle *trans,
780 struct btrfs_root *root,
781 struct btrfs_inode_item *btrfs_inode, u64 objectid,
782 ino_t parent_inum, struct stat *st,
783 const char *path_name, int out_fd)
789 struct btrfs_key key;
791 u32 sectorsize = root->sectorsize;
796 fd = open(path_name, O_RDONLY);
798 fprintf(stderr, "%s open failed\n", path_name);
802 blocks = st->st_size / sectorsize;
803 if (st->st_size % sectorsize)
806 if (st->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
807 buffer = malloc(st->st_size);
808 ret_read = pread64(fd, buffer, st->st_size, bytes_read);
809 if (ret_read == -1) {
810 fprintf(stderr, "%s read failed\n", path_name);
814 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
815 buffer, st->st_size);
819 ret = custom_alloc_extent(root, blocks * sectorsize, 0, &key);
823 first_block = key.objectid;
825 buffer = malloc(sectorsize);
828 memset(buffer, 0, sectorsize);
829 ret_read = pread64(fd, buffer, sectorsize, bytes_read);
830 if (ret_read == -1) {
831 fprintf(stderr, "%s read failed\n", path_name);
835 ret = pwrite64(out_fd, buffer, sectorsize,
836 first_block + bytes_read);
837 if (ret != sectorsize) {
838 fprintf(stderr, "output file write failed\n");
842 /* checksum for file data */
843 ret = btrfs_csum_file_block(trans, root->fs_info->csum_root,
844 first_block + (blocks * sectorsize),
845 first_block + bytes_read,
848 fprintf(stderr, "%s checksum failed\n", path_name);
852 bytes_read += ret_read;
854 } while (ret_read == sectorsize);
856 if (num_blocks > 0) {
857 ret = record_file_extent(trans, root, objectid, btrfs_inode,
858 first_block, first_block,
859 blocks * sectorsize);
871 static char *make_path(char *dir, char *name)
875 path = malloc(strlen(dir) + strlen(name) + 2);
879 if (dir[strlen(dir) - 1] != '/')
885 static int traverse_directory(struct btrfs_trans_handle *trans,
886 struct btrfs_root *root, char *dir_name,
887 struct directory_name_entry *dir_head, int out_fd)
891 struct btrfs_inode_item cur_inode;
892 struct btrfs_inode_item *inode_item;
893 int count, i, dir_index_cnt;
894 struct direct **files;
896 struct directory_name_entry *dir_entry, *parent_dir_entry;
897 struct direct *cur_file;
898 ino_t parent_inum, cur_inum;
899 ino_t highest_inum = 0;
900 char *parent_dir_name;
901 struct btrfs_path path;
902 struct extent_buffer *leaf;
903 struct btrfs_key root_dir_key;
904 u64 root_dir_inode_size = 0;
906 /* Add list for source directory */
907 dir_entry = malloc(sizeof(struct directory_name_entry));
908 dir_entry->dir_name = dir_name;
909 dir_entry->path = strdup(dir_name);
911 parent_inum = highest_inum + BTRFS_FIRST_FREE_OBJECTID;
912 dir_entry->inum = parent_inum;
913 list_add_tail(&dir_entry->list, &dir_head->list);
915 btrfs_init_path(&path);
917 root_dir_key.objectid = btrfs_root_dirid(&root->root_item);
918 root_dir_key.offset = 0;
919 btrfs_set_key_type(&root_dir_key, BTRFS_INODE_ITEM_KEY);
920 ret = btrfs_lookup_inode(trans, root, &path, &root_dir_key, 1);
922 fprintf(stderr, "root dir lookup error\n");
926 leaf = path.nodes[0];
927 inode_item = btrfs_item_ptr(leaf, path.slots[0],
928 struct btrfs_inode_item);
930 root_dir_inode_size = calculate_dir_inode_size(dir_name);
931 btrfs_set_inode_size(leaf, inode_item, root_dir_inode_size);
932 btrfs_mark_buffer_dirty(leaf);
934 btrfs_release_path(root, &path);
937 parent_dir_entry = list_entry(dir_head->list.next,
938 struct directory_name_entry,
940 list_del(&parent_dir_entry->list);
942 parent_inum = parent_dir_entry->inum;
943 parent_dir_name = parent_dir_entry->dir_name;
944 if (chdir(parent_dir_entry->path)) {
945 fprintf(stderr, "chdir error for %s\n",
950 count = scandir(parent_dir_entry->path, &files,
951 directory_select, NULL);
954 fprintf(stderr, "scandir for %s failed: %s\n",
955 parent_dir_name, strerror (errno));
959 for (i = 0; i < count; i++) {
962 if (lstat(cur_file->d_name, &st) == -1) {
963 fprintf(stderr, "lstat failed for file %s\n",
968 cur_inum = ++highest_inum + BTRFS_FIRST_FREE_OBJECTID;
969 ret = add_directory_items(trans, root,
970 cur_inum, parent_inum,
972 &st, &dir_index_cnt);
974 fprintf(stderr, "add_directory_items failed\n");
978 ret = add_inode_items(trans, root, &st,
979 cur_file->d_name, cur_inum,
980 parent_inum, dir_index_cnt,
983 fprintf(stderr, "add_inode_items failed\n");
987 ret = add_xattr_item(trans, root,
988 cur_inum, cur_file->d_name);
990 fprintf(stderr, "add_xattr_item failed\n");
995 if (S_ISDIR(st.st_mode)) {
996 dir_entry = malloc(sizeof(struct directory_name_entry));
997 dir_entry->dir_name = cur_file->d_name;
998 dir_entry->path = make_path(parent_dir_entry->path,
1000 dir_entry->inum = cur_inum;
1001 list_add_tail(&dir_entry->list, &dir_head->list);
1002 } else if (S_ISREG(st.st_mode)) {
1003 ret = add_file_items(trans, root, &cur_inode,
1004 cur_inum, parent_inum, &st,
1005 cur_file->d_name, out_fd);
1007 fprintf(stderr, "add_file_items failed\n");
1010 } else if (S_ISLNK(st.st_mode)) {
1011 ret = add_symbolic_link(trans, root,
1012 cur_inum, cur_file->d_name);
1014 fprintf(stderr, "add_symbolic_link failed\n");
1020 free_namelist(files, count);
1021 free(parent_dir_entry->path);
1022 free(parent_dir_entry);
1026 } while (!list_empty(&dir_head->list));
1030 free_namelist(files, count);
1032 free(parent_dir_entry->path);
1033 free(parent_dir_entry);
1037 static int open_target(char *output_name)
1040 output_fd = open(output_name, O_CREAT | O_RDWR | O_TRUNC,
1041 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH);
1046 static int create_chunks(struct btrfs_trans_handle *trans,
1047 struct btrfs_root *root, u64 num_of_meta_chunks,
1052 u64 meta_type = BTRFS_BLOCK_GROUP_METADATA;
1053 u64 data_type = BTRFS_BLOCK_GROUP_DATA;
1054 u64 minimum_data_chunk_size = 8 * 1024 * 1024;
1058 for (i = 0; i < num_of_meta_chunks; i++) {
1059 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
1060 &chunk_start, &chunk_size, meta_type);
1062 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
1063 meta_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
1064 chunk_start, chunk_size);
1066 set_extent_dirty(&root->fs_info->free_space_cache,
1067 chunk_start, chunk_start + chunk_size - 1, 0);
1070 if (size_of_data < minimum_data_chunk_size)
1071 size_of_data = minimum_data_chunk_size;
1072 ret = btrfs_alloc_data_chunk(trans, root->fs_info->extent_root,
1073 &chunk_start, size_of_data, data_type);
1075 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
1076 data_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
1077 chunk_start, size_of_data);
1079 set_extent_dirty(&root->fs_info->free_space_cache,
1080 chunk_start, chunk_start + size_of_data - 1, 0);
1084 static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
1087 struct btrfs_trans_handle *trans;
1089 struct stat root_st;
1091 struct directory_name_entry dir_head;
1093 ret = lstat(source_dir, &root_st);
1095 fprintf(stderr, "unable to lstat the %s\n", source_dir);
1099 INIT_LIST_HEAD(&dir_head.list);
1101 trans = btrfs_start_transaction(root, 1);
1102 ret = traverse_directory(trans, root, source_dir, &dir_head, out_fd);
1104 fprintf(stderr, "unable to traverse_directory\n");
1107 btrfs_commit_transaction(trans, root);
1109 printf("Making image is completed.\n");
1112 fprintf(stderr, "Making image is aborted.\n");
1117 * This ignores symlinks with unreadable targets and subdirs that can't
1118 * be read. It's a best-effort to give a rough estimate of the size of
1119 * a subdir. It doesn't guarantee that prepopulating btrfs from this
1120 * tree won't still run out of space.
1122 * The rounding up to 4096 is questionable. Previous code used du -B 4096.
1124 static u64 global_total_size;
1125 static int ftw_add_entry_size(const char *fpath, const struct stat *st,
1128 if (type == FTW_F || type == FTW_D)
1129 global_total_size += round_up(st->st_size, 4096);
1134 static u64 size_sourcedir(char *dir_name, u64 sectorsize,
1135 u64 *num_of_meta_chunks_ret, u64 *size_of_data_ret)
1140 u64 default_chunk_size = 8 * 1024 * 1024; /* 8MB */
1141 u64 allocated_meta_size = 8 * 1024 * 1024; /* 8MB */
1142 u64 allocated_total_size = 20 * 1024 * 1024; /* 20MB */
1143 u64 num_of_meta_chunks = 0;
1144 u64 num_of_allocated_meta_chunks =
1145 allocated_meta_size / default_chunk_size;
1147 global_total_size = 0;
1148 ret = ftw(dir_name, ftw_add_entry_size, 10);
1149 dir_size = global_total_size;
1151 fprintf(stderr, "ftw subdir walk of '%s' failed: %s\n",
1152 dir_name, strerror(errno));
1156 num_of_meta_chunks = (dir_size / 2) / default_chunk_size;
1157 if (((dir_size / 2) % default_chunk_size) != 0)
1158 num_of_meta_chunks++;
1159 if (num_of_meta_chunks <= num_of_allocated_meta_chunks)
1160 num_of_meta_chunks = 0;
1162 num_of_meta_chunks -= num_of_allocated_meta_chunks;
1164 total_size = allocated_total_size + dir_size +
1165 (num_of_meta_chunks * default_chunk_size);
1167 *num_of_meta_chunks_ret = num_of_meta_chunks;
1172 static int zero_output_file(int out_fd, u64 size, u32 sectorsize)
1174 int len = sectorsize;
1175 int loop_num = size / sectorsize;
1177 char *buf = malloc(len);
1183 memset(buf, 0, len);
1184 for (i = 0; i < loop_num; i++) {
1185 written = pwrite64(out_fd, buf, len, location);
1188 location += sectorsize;
1194 static int check_leaf_or_node_size(u32 size, u32 sectorsize)
1196 if (size < sectorsize) {
1198 "Illegal leafsize (or nodesize) %u (smaller than %u)\n",
1201 } else if (size > BTRFS_MAX_METADATA_BLOCKSIZE) {
1203 "Illegal leafsize (or nodesize) %u (larger than %u)\n",
1204 size, BTRFS_MAX_METADATA_BLOCKSIZE);
1206 } else if (size & (sectorsize - 1)) {
1208 "Illegal leafsize (or nodesize) %u (not align to %u)\n",
1215 static int is_ssd(const char *file)
1220 char path[PATH_MAX];
1225 probe = blkid_new_probe_from_filename(file);
1230 * We want to use blkid_devno_to_wholedisk() but it's broken for some
1231 * reason on F17 at least so we'll do this trickery
1233 disk = blkid_probe_get_wholedisk_devno(probe);
1237 devname = blkid_devno_to_devname(disk);
1241 dev = strrchr(devname, '/');
1244 snprintf(path, PATH_MAX, "/sys/block/%s/queue/rotational", dev);
1247 blkid_free_probe(probe);
1249 fd = open(path, O_RDONLY);
1254 if (read(fd, &rotational, sizeof(char)) < sizeof(char)) {
1260 return !atoi((const char *)&rotational);
1263 int main(int ac, char **av)
1266 struct btrfs_root *root;
1267 struct btrfs_trans_handle *trans;
1270 u64 block_count = 0;
1271 u64 dev_block_count = 0;
1273 u64 alloc_start = 0;
1274 u64 metadata_profile = 0;
1275 u64 data_profile = 0;
1276 u32 leafsize = sysconf(_SC_PAGESIZE);
1277 u32 sectorsize = 4096;
1278 u32 nodesize = leafsize;
1279 u32 stripesize = 4096;
1281 int option_index = 0;
1286 int data_profile_opt = 0;
1287 int metadata_profile_opt = 0;
1291 char *source_dir = NULL;
1292 int source_dir_set = 0;
1293 u64 num_of_meta_chunks = 0;
1294 u64 size_of_data = 0;
1295 u64 source_dir_size = 0;
1297 struct btrfs_super_block *super;
1302 c = getopt_long(ac, av, "A:b:l:n:s:m:d:L:r:VMK", long_options,
1308 alloc_start = parse_size(optarg);
1311 data_profile = parse_profile(optarg);
1312 data_profile_opt = 1;
1316 nodesize = parse_size(optarg);
1317 leafsize = parse_size(optarg);
1320 label = parse_label(optarg);
1323 metadata_profile = parse_profile(optarg);
1324 metadata_profile_opt = 1;
1330 sectorsize = parse_size(optarg);
1333 block_count = parse_size(optarg);
1334 if (block_count <= 1024*1024*1024) {
1335 printf("SMALL VOLUME: forcing mixed "
1336 "metadata/data groups\n");
1345 source_dir = optarg;
1355 sectorsize = max(sectorsize, (u32)sysconf(_SC_PAGESIZE));
1356 if (check_leaf_or_node_size(leafsize, sectorsize))
1358 if (check_leaf_or_node_size(nodesize, sectorsize))
1364 printf("\nWARNING! - %s IS EXPERIMENTAL\n", BTRFS_BUILD_VERSION);
1365 printf("WARNING! - see http://btrfs.wiki.kernel.org before using\n\n");
1367 if (source_dir == 0) {
1368 file = av[optind++];
1369 ret = is_swap_device(file);
1371 fprintf(stderr, "error checking %s status: %s\n", file,
1376 fprintf(stderr, "%s is a swap device\n", file);
1379 ret = check_mounted(file);
1381 fprintf(stderr, "error checking %s mount status\n", file);
1385 fprintf(stderr, "%s is mounted\n", file);
1389 /* check if the device is busy */
1390 fd = open(file, O_RDWR|O_EXCL);
1392 fprintf(stderr, "unable to open %s: %s\n", file,
1398 * open again without O_EXCL so that the problem should not
1399 * occur by the following processing.
1400 * (btrfs_register_one_device() fails if O_EXCL is on)
1402 fd = open(file, O_RDWR);
1404 fprintf(stderr, "unable to open %s: %s\n", file,
1409 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1410 block_count, &mixed, nodiscard);
1411 if (block_count && block_count > dev_block_count) {
1412 fprintf(stderr, "%s is smaller than requested size\n", file);
1417 file = av[optind++];
1418 fd = open_target(file);
1420 fprintf(stderr, "unable to open the %s\n", file);
1425 source_dir_size = size_sourcedir(source_dir, sectorsize,
1426 &num_of_meta_chunks, &size_of_data);
1427 if(block_count < source_dir_size)
1428 block_count = source_dir_size;
1429 ret = zero_output_file(fd, block_count, sectorsize);
1431 fprintf(stderr, "unable to zero the output file\n");
1434 /* our "device" is the new image file */
1435 dev_block_count = block_count;
1441 if (metadata_profile != data_profile) {
1442 fprintf(stderr, "With mixed block groups data and metadata "
1443 "profiles must be the same\n");
1448 blocks[0] = BTRFS_SUPER_INFO_OFFSET;
1449 for (i = 1; i < 7; i++) {
1450 blocks[i] = BTRFS_SUPER_INFO_OFFSET + 1024 * 1024 +
1454 ret = make_btrfs(fd, file, label, blocks, dev_block_count,
1456 sectorsize, stripesize);
1458 fprintf(stderr, "error during mkfs %d\n", ret);
1462 root = open_ctree(file, 0, O_RDWR);
1464 fprintf(stderr, "Open ctree failed\n");
1468 root->fs_info->alloc_start = alloc_start;
1470 ret = make_root_dir(root, mixed);
1472 fprintf(stderr, "failed to setup the root directory\n");
1476 trans = btrfs_start_transaction(root, 1);
1481 btrfs_register_one_device(file);
1485 int old_mixed = mixed;
1487 file = av[optind++];
1488 ret = is_swap_device(file);
1490 fprintf(stderr, "error checking %s status: %s\n", file,
1495 fprintf(stderr, "%s is a swap device\n", file);
1498 ret = check_mounted(file);
1500 fprintf(stderr, "error checking %s mount status\n",
1505 fprintf(stderr, "%s is mounted\n", file);
1508 /* check if the device is busy */
1509 fd = open(file, O_RDWR|O_EXCL);
1511 fprintf(stderr, "unable to open %s: %s\n", file,
1517 * open again without O_EXCL so that the problem should not
1518 * occur by the following processing.
1519 * (btrfs_register_one_device() fails if O_EXCL is on)
1521 fd = open(file, O_RDWR);
1523 fprintf(stderr, "unable to open %s: %s\n", file,
1527 ret = btrfs_device_already_in_root(root, fd,
1528 BTRFS_SUPER_INFO_OFFSET);
1530 fprintf(stderr, "skipping duplicate device %s in FS\n",
1535 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1536 block_count, &mixed, nodiscard);
1540 ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
1541 sectorsize, sectorsize, sectorsize);
1543 btrfs_register_one_device(file);
1547 if (!source_dir_set) {
1548 ret = create_raid_groups(trans, root, data_profile,
1549 data_profile_opt, metadata_profile,
1550 metadata_profile_opt, mixed, ssd);
1554 ret = create_data_reloc_tree(trans, root);
1557 super = &root->fs_info->super_copy;
1558 flags = btrfs_super_incompat_flags(super);
1559 flags |= BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF;
1562 flags |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
1564 btrfs_set_super_incompat_flags(super, flags);
1566 if ((data_profile | metadata_profile) &
1567 (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
1568 struct btrfs_super_block *super = &root->fs_info->super_copy;
1569 u64 flags = btrfs_super_incompat_flags(super);
1571 flags |= BTRFS_FEATURE_INCOMPAT_RAID56;
1572 btrfs_set_super_incompat_flags(super, flags);
1573 printf("Setting RAID5/6 feature flag\n");
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_buf = pretty_sizes(btrfs_super_total_bytes(&root->fs_info->super_copy)));
1582 printf("%s\n", BTRFS_BUILD_VERSION);
1583 btrfs_commit_transaction(trans, root);
1585 if (source_dir_set) {
1586 trans = btrfs_start_transaction(root, 1);
1587 ret = create_chunks(trans, root,
1588 num_of_meta_chunks, size_of_data);
1590 btrfs_commit_transaction(trans, root);
1592 ret = make_image(source_dir, root, fd);
1596 ret = close_ctree(root);