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"
24 #include <sys/ioctl.h>
25 #include <sys/mount.h>
29 #include <sys/types.h>
35 #include <uuid/uuid.h>
37 #include <attr/xattr.h>
38 #include <blkid/blkid.h>
43 #include "transaction.h"
47 static u64 index_cnt = 2;
49 struct directory_name_entry {
53 struct list_head list;
56 static int make_root_dir(struct btrfs_root *root, int mixed)
58 struct btrfs_trans_handle *trans;
59 struct btrfs_key location;
65 trans = btrfs_start_transaction(root, 1);
66 bytes_used = btrfs_super_bytes_used(root->fs_info->super_copy);
68 root->fs_info->system_allocs = 1;
69 ret = btrfs_make_block_group(trans, root, bytes_used,
70 BTRFS_BLOCK_GROUP_SYSTEM,
71 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
72 0, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
76 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
77 &chunk_start, &chunk_size,
78 BTRFS_BLOCK_GROUP_METADATA |
79 BTRFS_BLOCK_GROUP_DATA);
82 "no space to alloc data/metadata chunk\n");
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 fprintf(stderr, "no space to alloc metadata chunk\n");
102 ret = btrfs_make_block_group(trans, root, 0,
103 BTRFS_BLOCK_GROUP_METADATA,
104 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
105 chunk_start, chunk_size);
109 root->fs_info->system_allocs = 0;
110 btrfs_commit_transaction(trans, root);
111 trans = btrfs_start_transaction(root, 1);
115 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
116 &chunk_start, &chunk_size,
117 BTRFS_BLOCK_GROUP_DATA);
118 if (ret == -ENOSPC) {
119 fprintf(stderr, "no space to alloc data chunk\n");
123 ret = btrfs_make_block_group(trans, root, 0,
124 BTRFS_BLOCK_GROUP_DATA,
125 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
126 chunk_start, chunk_size);
130 ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
131 BTRFS_ROOT_TREE_DIR_OBJECTID);
134 ret = btrfs_make_root_dir(trans, root, BTRFS_FIRST_FREE_OBJECTID);
137 memcpy(&location, &root->fs_info->fs_root->root_key, sizeof(location));
138 location.offset = (u64)-1;
139 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
141 btrfs_super_root_dir(root->fs_info->super_copy),
142 &location, BTRFS_FT_DIR, 0);
146 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
147 "default", 7, location.objectid,
148 BTRFS_ROOT_TREE_DIR_OBJECTID, 0);
152 btrfs_commit_transaction(trans, root);
157 static void __recow_root(struct btrfs_trans_handle *trans,
158 struct btrfs_root *root)
161 struct extent_buffer *tmp;
163 if (trans->transid != btrfs_root_generation(&root->root_item)) {
164 ret = __btrfs_cow_block(trans, root, root->node,
165 NULL, 0, &tmp, 0, 0);
167 free_extent_buffer(tmp);
171 static void recow_roots(struct btrfs_trans_handle *trans,
172 struct btrfs_root *root)
174 struct btrfs_fs_info *info = root->fs_info;
176 __recow_root(trans, info->fs_root);
177 __recow_root(trans, info->tree_root);
178 __recow_root(trans, info->extent_root);
179 __recow_root(trans, info->chunk_root);
180 __recow_root(trans, info->dev_root);
181 __recow_root(trans, info->csum_root);
184 static int create_one_raid_group(struct btrfs_trans_handle *trans,
185 struct btrfs_root *root, u64 type)
191 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
192 &chunk_start, &chunk_size, type);
193 if (ret == -ENOSPC) {
194 fprintf(stderr, "not enough free space\n");
198 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
199 type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
200 chunk_start, chunk_size);
205 static int create_raid_groups(struct btrfs_trans_handle *trans,
206 struct btrfs_root *root, u64 data_profile,
207 int data_profile_opt, u64 metadata_profile,
208 int metadata_profile_opt, int mixed, int ssd)
210 u64 num_devices = btrfs_super_num_devices(root->fs_info->super_copy);
213 if (metadata_profile) {
214 u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
216 ret = create_one_raid_group(trans, root,
217 BTRFS_BLOCK_GROUP_SYSTEM |
222 meta_flags |= BTRFS_BLOCK_GROUP_DATA;
224 ret = create_one_raid_group(trans, root, meta_flags |
229 if (!mixed && num_devices > 1 && data_profile) {
230 ret = create_one_raid_group(trans, root,
231 BTRFS_BLOCK_GROUP_DATA |
235 recow_roots(trans, root);
240 static int create_data_reloc_tree(struct btrfs_trans_handle *trans,
241 struct btrfs_root *root)
243 struct btrfs_key location;
244 struct btrfs_root_item root_item;
245 struct extent_buffer *tmp;
246 u64 objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
249 ret = btrfs_copy_root(trans, root, root->node, &tmp, objectid);
252 memcpy(&root_item, &root->root_item, sizeof(root_item));
253 btrfs_set_root_bytenr(&root_item, tmp->start);
254 btrfs_set_root_level(&root_item, btrfs_header_level(tmp));
255 btrfs_set_root_generation(&root_item, trans->transid);
256 free_extent_buffer(tmp);
258 location.objectid = objectid;
259 location.type = BTRFS_ROOT_ITEM_KEY;
261 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
262 &location, &root_item);
267 static void print_usage(void)
269 fprintf(stderr, "usage: mkfs.btrfs [options] dev [ dev ... ]\n");
270 fprintf(stderr, "options:\n");
271 fprintf(stderr, "\t -A --alloc-start the offset to start the FS\n");
272 fprintf(stderr, "\t -b --byte-count total number of bytes in the FS\n");
273 fprintf(stderr, "\t -d --data data profile, raid0, raid1, raid5, raid6, raid10, dup or single\n");
274 fprintf(stderr, "\t -f --force force overwrite of existing filesystem\n");
275 fprintf(stderr, "\t -l --leafsize size of btree leaves\n");
276 fprintf(stderr, "\t -L --label set a label\n");
277 fprintf(stderr, "\t -m --metadata metadata profile, values like data profile\n");
278 fprintf(stderr, "\t -M --mixed mix metadata and data together\n");
279 fprintf(stderr, "\t -n --nodesize size of btree nodes\n");
280 fprintf(stderr, "\t -s --sectorsize min block allocation (may not mountable by current kernel)\n");
281 fprintf(stderr, "\t -r --rootdir the source directory\n");
282 fprintf(stderr, "\t -K --nodiscard do not perform whole device TRIM\n");
283 fprintf(stderr, "\t -O --features comma separated list of filesystem features\n");
284 fprintf(stderr, "\t -V --version print the mkfs.btrfs version and exit\n");
285 fprintf(stderr, "%s\n", BTRFS_BUILD_VERSION);
289 static void print_version(void)
291 fprintf(stderr, "mkfs.btrfs, part of %s\n", BTRFS_BUILD_VERSION);
295 static u64 parse_profile(char *s)
297 if (strcmp(s, "raid0") == 0) {
298 return BTRFS_BLOCK_GROUP_RAID0;
299 } else if (strcmp(s, "raid1") == 0) {
300 return BTRFS_BLOCK_GROUP_RAID1;
301 } else if (strcmp(s, "raid5") == 0) {
302 return BTRFS_BLOCK_GROUP_RAID5;
303 } else if (strcmp(s, "raid6") == 0) {
304 return BTRFS_BLOCK_GROUP_RAID6;
305 } else if (strcmp(s, "raid10") == 0) {
306 return BTRFS_BLOCK_GROUP_RAID10;
307 } else if (strcmp(s, "dup") == 0) {
308 return BTRFS_BLOCK_GROUP_DUP;
309 } else if (strcmp(s, "single") == 0) {
312 fprintf(stderr, "Unknown profile %s\n", s);
319 static char *parse_label(char *input)
321 int len = strlen(input);
323 if (len >= BTRFS_LABEL_SIZE) {
324 fprintf(stderr, "Label %s is too long (max %d)\n", input,
325 BTRFS_LABEL_SIZE - 1);
328 return strdup(input);
331 static struct option long_options[] = {
332 { "alloc-start", 1, NULL, 'A'},
333 { "byte-count", 1, NULL, 'b' },
334 { "force", 0, NULL, 'f' },
335 { "leafsize", 1, NULL, 'l' },
336 { "label", 1, NULL, 'L'},
337 { "metadata", 1, NULL, 'm' },
338 { "mixed", 0, NULL, 'M' },
339 { "nodesize", 1, NULL, 'n' },
340 { "sectorsize", 1, NULL, 's' },
341 { "data", 1, NULL, 'd' },
342 { "version", 0, NULL, 'V' },
343 { "rootdir", 1, NULL, 'r' },
344 { "nodiscard", 0, NULL, 'K' },
345 { "features", 0, NULL, 'O' },
349 static int add_directory_items(struct btrfs_trans_handle *trans,
350 struct btrfs_root *root, u64 objectid,
351 ino_t parent_inum, const char *name,
352 struct stat *st, int *dir_index_cnt)
356 struct btrfs_key location;
359 name_len = strlen(name);
361 location.objectid = objectid;
363 btrfs_set_key_type(&location, BTRFS_INODE_ITEM_KEY);
365 if (S_ISDIR(st->st_mode))
366 filetype = BTRFS_FT_DIR;
367 if (S_ISREG(st->st_mode))
368 filetype = BTRFS_FT_REG_FILE;
369 if (S_ISLNK(st->st_mode))
370 filetype = BTRFS_FT_SYMLINK;
372 ret = btrfs_insert_dir_item(trans, root, name, name_len,
373 parent_inum, &location,
374 filetype, index_cnt);
376 *dir_index_cnt = index_cnt;
382 static int fill_inode_item(struct btrfs_trans_handle *trans,
383 struct btrfs_root *root,
384 struct btrfs_inode_item *dst, struct stat *src)
387 u64 sectorsize = root->sectorsize;
390 * btrfs_inode_item has some reserved fields
391 * and represents on-disk inode entry, so
392 * zero everything to prevent information leak
394 memset(dst, 0, sizeof (*dst));
396 btrfs_set_stack_inode_generation(dst, trans->transid);
397 btrfs_set_stack_inode_size(dst, src->st_size);
398 btrfs_set_stack_inode_nbytes(dst, 0);
399 btrfs_set_stack_inode_block_group(dst, 0);
400 btrfs_set_stack_inode_nlink(dst, src->st_nlink);
401 btrfs_set_stack_inode_uid(dst, src->st_uid);
402 btrfs_set_stack_inode_gid(dst, src->st_gid);
403 btrfs_set_stack_inode_mode(dst, src->st_mode);
404 btrfs_set_stack_inode_rdev(dst, 0);
405 btrfs_set_stack_inode_flags(dst, 0);
406 btrfs_set_stack_timespec_sec(&dst->atime, src->st_atime);
407 btrfs_set_stack_timespec_nsec(&dst->atime, 0);
408 btrfs_set_stack_timespec_sec(&dst->ctime, src->st_ctime);
409 btrfs_set_stack_timespec_nsec(&dst->ctime, 0);
410 btrfs_set_stack_timespec_sec(&dst->mtime, src->st_mtime);
411 btrfs_set_stack_timespec_nsec(&dst->mtime, 0);
412 btrfs_set_stack_timespec_sec(&dst->otime, 0);
413 btrfs_set_stack_timespec_nsec(&dst->otime, 0);
415 if (S_ISDIR(src->st_mode)) {
416 btrfs_set_stack_inode_size(dst, 0);
417 btrfs_set_stack_inode_nlink(dst, 1);
419 if (S_ISREG(src->st_mode)) {
420 btrfs_set_stack_inode_size(dst, (u64)src->st_size);
421 if (src->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root))
422 btrfs_set_stack_inode_nbytes(dst, src->st_size);
424 blocks = src->st_size / sectorsize;
425 if (src->st_size % sectorsize)
427 blocks *= sectorsize;
428 btrfs_set_stack_inode_nbytes(dst, blocks);
431 if (S_ISLNK(src->st_mode))
432 btrfs_set_stack_inode_nbytes(dst, src->st_size + 1);
437 static int directory_select(const struct direct *entry)
439 if ((strncmp(entry->d_name, ".", entry->d_reclen) == 0) ||
440 (strncmp(entry->d_name, "..", entry->d_reclen) == 0))
446 static void free_namelist(struct direct **files, int count)
453 for (i = 0; i < count; ++i)
458 static u64 calculate_dir_inode_size(char *dirname)
461 struct direct **files, *cur_file;
462 u64 dir_inode_size = 0;
464 count = scandir(dirname, &files, directory_select, NULL);
466 for (i = 0; i < count; i++) {
468 dir_inode_size += strlen(cur_file->d_name);
471 free_namelist(files, count);
474 return dir_inode_size;
477 static int add_inode_items(struct btrfs_trans_handle *trans,
478 struct btrfs_root *root,
479 struct stat *st, char *name,
480 u64 self_objectid, ino_t parent_inum,
481 int dir_index_cnt, struct btrfs_inode_item *inode_ret)
484 struct btrfs_key inode_key;
485 struct btrfs_inode_item btrfs_inode;
490 name_len = strlen(name);
491 fill_inode_item(trans, root, &btrfs_inode, st);
492 objectid = self_objectid;
494 if (S_ISDIR(st->st_mode)) {
495 inode_size = calculate_dir_inode_size(name);
496 btrfs_set_stack_inode_size(&btrfs_inode, inode_size);
499 inode_key.objectid = objectid;
500 inode_key.offset = 0;
501 btrfs_set_key_type(&inode_key, BTRFS_INODE_ITEM_KEY);
503 ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
507 ret = btrfs_insert_inode_ref(trans, root, name, name_len,
508 objectid, parent_inum, dir_index_cnt);
512 *inode_ret = btrfs_inode;
517 static int add_xattr_item(struct btrfs_trans_handle *trans,
518 struct btrfs_root *root, u64 objectid,
519 const char *file_name)
523 char xattr_list[XATTR_LIST_MAX];
525 char cur_value[XATTR_SIZE_MAX];
526 char delimiter = '\0';
527 char *next_location = xattr_list;
529 ret = llistxattr(file_name, xattr_list, XATTR_LIST_MAX);
533 fprintf(stderr, "get a list of xattr failed for %s\n",
540 cur_name = strtok(xattr_list, &delimiter);
541 while (cur_name != NULL) {
542 cur_name_len = strlen(cur_name);
543 next_location += cur_name_len + 1;
545 ret = getxattr(file_name, cur_name, cur_value, XATTR_SIZE_MAX);
549 fprintf(stderr, "get a xattr value failed for %s attr %s\n",
550 file_name, cur_name);
554 ret = btrfs_insert_xattr_item(trans, root, cur_name,
555 cur_name_len, cur_value,
558 fprintf(stderr, "insert a xattr item failed for %s\n",
562 cur_name = strtok(next_location, &delimiter);
567 static int custom_alloc_extent(struct btrfs_root *root, u64 num_bytes,
568 u64 hint_byte, struct btrfs_key *ins)
572 u64 last = hint_byte;
575 struct btrfs_block_group_cache *cache;
578 ret = find_first_extent_bit(&root->fs_info->free_space_cache,
579 last, &start, &end, EXTENT_DIRTY);
581 if (wrapped++ == 0) {
589 start = max(last, start);
591 if (last - start < num_bytes)
594 last = start + num_bytes;
595 if (test_range_bit(&root->fs_info->pinned_extents,
596 start, last - 1, EXTENT_DIRTY, 0))
599 cache = btrfs_lookup_block_group(root->fs_info, start);
601 if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM ||
602 last > cache->key.objectid + cache->key.offset) {
603 last = cache->key.objectid + cache->key.offset;
607 if (cache->flags & (BTRFS_BLOCK_GROUP_SYSTEM |
608 BTRFS_BLOCK_GROUP_METADATA)) {
609 last = cache->key.objectid + cache->key.offset;
613 clear_extent_dirty(&root->fs_info->free_space_cache,
614 start, start + num_bytes - 1, 0);
616 ins->objectid = start;
617 ins->offset = num_bytes;
618 ins->type = BTRFS_EXTENT_ITEM_KEY;
622 fprintf(stderr, "not enough free space\n");
626 static int record_file_extent(struct btrfs_trans_handle *trans,
627 struct btrfs_root *root, u64 objectid,
628 struct btrfs_inode_item *inode,
629 u64 file_pos, u64 disk_bytenr,
633 struct btrfs_fs_info *info = root->fs_info;
634 struct btrfs_root *extent_root = info->extent_root;
635 struct extent_buffer *leaf;
636 struct btrfs_file_extent_item *fi;
637 struct btrfs_key ins_key;
638 struct btrfs_path path;
639 struct btrfs_extent_item *ei;
641 btrfs_init_path(&path);
643 ins_key.objectid = objectid;
645 btrfs_set_key_type(&ins_key, BTRFS_EXTENT_DATA_KEY);
646 ret = btrfs_insert_empty_item(trans, root, &path, &ins_key,
650 leaf = path.nodes[0];
651 fi = btrfs_item_ptr(leaf, path.slots[0],
652 struct btrfs_file_extent_item);
653 btrfs_set_file_extent_generation(leaf, fi, trans->transid);
654 btrfs_set_file_extent_type(leaf, fi, BTRFS_FILE_EXTENT_REG);
655 btrfs_set_file_extent_disk_bytenr(leaf, fi, disk_bytenr);
656 btrfs_set_file_extent_disk_num_bytes(leaf, fi, num_bytes);
657 btrfs_set_file_extent_offset(leaf, fi, 0);
658 btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
659 btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
660 btrfs_set_file_extent_compression(leaf, fi, 0);
661 btrfs_set_file_extent_encryption(leaf, fi, 0);
662 btrfs_set_file_extent_other_encoding(leaf, fi, 0);
663 btrfs_mark_buffer_dirty(leaf);
665 btrfs_release_path(&path);
667 ins_key.objectid = disk_bytenr;
668 ins_key.offset = num_bytes;
669 ins_key.type = BTRFS_EXTENT_ITEM_KEY;
671 ret = btrfs_insert_empty_item(trans, extent_root, &path,
672 &ins_key, sizeof(*ei));
674 leaf = path.nodes[0];
675 ei = btrfs_item_ptr(leaf, path.slots[0],
676 struct btrfs_extent_item);
678 btrfs_set_extent_refs(leaf, ei, 0);
679 btrfs_set_extent_generation(leaf, ei, trans->transid);
680 btrfs_set_extent_flags(leaf, ei, BTRFS_EXTENT_FLAG_DATA);
682 btrfs_mark_buffer_dirty(leaf);
683 ret = btrfs_update_block_group(trans, root, disk_bytenr,
687 } else if (ret != -EEXIST) {
691 ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, num_bytes, 0,
692 root->root_key.objectid,
695 btrfs_release_path(&path);
699 static int add_symbolic_link(struct btrfs_trans_handle *trans,
700 struct btrfs_root *root,
701 u64 objectid, const char *path_name)
704 u64 sectorsize = root->sectorsize;
705 char *buf = malloc(sectorsize);
707 ret = readlink(path_name, buf, sectorsize);
709 fprintf(stderr, "readlink failed for %s\n", path_name);
712 if (ret >= sectorsize) {
713 fprintf(stderr, "symlink too long for %s", path_name);
718 buf[ret] = '\0'; /* readlink does not do it for us */
719 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
726 static int add_file_items(struct btrfs_trans_handle *trans,
727 struct btrfs_root *root,
728 struct btrfs_inode_item *btrfs_inode, u64 objectid,
729 ino_t parent_inum, struct stat *st,
730 const char *path_name, int out_fd)
736 struct btrfs_key key;
738 u32 sectorsize = root->sectorsize;
743 fd = open(path_name, O_RDONLY);
745 fprintf(stderr, "%s open failed\n", path_name);
749 blocks = st->st_size / sectorsize;
750 if (st->st_size % sectorsize)
753 if (st->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
754 buffer = malloc(st->st_size);
755 ret_read = pread64(fd, buffer, st->st_size, bytes_read);
756 if (ret_read == -1) {
757 fprintf(stderr, "%s read failed\n", path_name);
761 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
762 buffer, st->st_size);
766 ret = custom_alloc_extent(root, blocks * sectorsize, 0, &key);
770 first_block = key.objectid;
772 buffer = malloc(sectorsize);
775 memset(buffer, 0, sectorsize);
776 ret_read = pread64(fd, buffer, sectorsize, bytes_read);
777 if (ret_read == -1) {
778 fprintf(stderr, "%s read failed\n", path_name);
782 ret = pwrite64(out_fd, buffer, sectorsize,
783 first_block + bytes_read);
784 if (ret != sectorsize) {
785 fprintf(stderr, "output file write failed\n");
789 /* checksum for file data */
790 ret = btrfs_csum_file_block(trans, root->fs_info->csum_root,
791 first_block + (blocks * sectorsize),
792 first_block + bytes_read,
795 fprintf(stderr, "%s checksum failed\n", path_name);
799 bytes_read += ret_read;
801 } while (ret_read == sectorsize);
803 if (num_blocks > 0) {
804 ret = record_file_extent(trans, root, objectid, btrfs_inode,
805 first_block, first_block,
806 blocks * sectorsize);
818 static char *make_path(char *dir, char *name)
822 path = malloc(strlen(dir) + strlen(name) + 2);
826 if (dir[strlen(dir) - 1] != '/')
832 static int traverse_directory(struct btrfs_trans_handle *trans,
833 struct btrfs_root *root, char *dir_name,
834 struct directory_name_entry *dir_head, int out_fd)
838 struct btrfs_inode_item cur_inode;
839 struct btrfs_inode_item *inode_item;
840 int count, i, dir_index_cnt;
841 struct direct **files;
843 struct directory_name_entry *dir_entry, *parent_dir_entry;
844 struct direct *cur_file;
845 ino_t parent_inum, cur_inum;
846 ino_t highest_inum = 0;
847 char *parent_dir_name;
848 struct btrfs_path path;
849 struct extent_buffer *leaf;
850 struct btrfs_key root_dir_key;
851 u64 root_dir_inode_size = 0;
853 /* Add list for source directory */
854 dir_entry = malloc(sizeof(struct directory_name_entry));
855 dir_entry->dir_name = dir_name;
856 dir_entry->path = strdup(dir_name);
858 parent_inum = highest_inum + BTRFS_FIRST_FREE_OBJECTID;
859 dir_entry->inum = parent_inum;
860 list_add_tail(&dir_entry->list, &dir_head->list);
862 btrfs_init_path(&path);
864 root_dir_key.objectid = btrfs_root_dirid(&root->root_item);
865 root_dir_key.offset = 0;
866 btrfs_set_key_type(&root_dir_key, BTRFS_INODE_ITEM_KEY);
867 ret = btrfs_lookup_inode(trans, root, &path, &root_dir_key, 1);
869 fprintf(stderr, "root dir lookup error\n");
873 leaf = path.nodes[0];
874 inode_item = btrfs_item_ptr(leaf, path.slots[0],
875 struct btrfs_inode_item);
877 root_dir_inode_size = calculate_dir_inode_size(dir_name);
878 btrfs_set_inode_size(leaf, inode_item, root_dir_inode_size);
879 btrfs_mark_buffer_dirty(leaf);
881 btrfs_release_path(&path);
884 parent_dir_entry = list_entry(dir_head->list.next,
885 struct directory_name_entry,
887 list_del(&parent_dir_entry->list);
889 parent_inum = parent_dir_entry->inum;
890 parent_dir_name = parent_dir_entry->dir_name;
891 if (chdir(parent_dir_entry->path)) {
892 fprintf(stderr, "chdir error for %s\n",
897 count = scandir(parent_dir_entry->path, &files,
898 directory_select, NULL);
901 fprintf(stderr, "scandir for %s failed: %s\n",
902 parent_dir_name, strerror (errno));
906 for (i = 0; i < count; i++) {
909 if (lstat(cur_file->d_name, &st) == -1) {
910 fprintf(stderr, "lstat failed for file %s\n",
915 cur_inum = ++highest_inum + BTRFS_FIRST_FREE_OBJECTID;
916 ret = add_directory_items(trans, root,
917 cur_inum, parent_inum,
919 &st, &dir_index_cnt);
921 fprintf(stderr, "add_directory_items failed\n");
925 ret = add_inode_items(trans, root, &st,
926 cur_file->d_name, cur_inum,
927 parent_inum, dir_index_cnt,
930 fprintf(stderr, "add_inode_items failed\n");
934 ret = add_xattr_item(trans, root,
935 cur_inum, cur_file->d_name);
937 fprintf(stderr, "add_xattr_item failed\n");
942 if (S_ISDIR(st.st_mode)) {
943 dir_entry = malloc(sizeof(struct directory_name_entry));
944 dir_entry->dir_name = cur_file->d_name;
945 dir_entry->path = make_path(parent_dir_entry->path,
947 dir_entry->inum = cur_inum;
948 list_add_tail(&dir_entry->list, &dir_head->list);
949 } else if (S_ISREG(st.st_mode)) {
950 ret = add_file_items(trans, root, &cur_inode,
951 cur_inum, parent_inum, &st,
952 cur_file->d_name, out_fd);
954 fprintf(stderr, "add_file_items failed\n");
957 } else if (S_ISLNK(st.st_mode)) {
958 ret = add_symbolic_link(trans, root,
959 cur_inum, cur_file->d_name);
961 fprintf(stderr, "add_symbolic_link failed\n");
967 free_namelist(files, count);
968 free(parent_dir_entry->path);
969 free(parent_dir_entry);
973 } while (!list_empty(&dir_head->list));
977 free_namelist(files, count);
979 free(parent_dir_entry->path);
980 free(parent_dir_entry);
984 static int open_target(char *output_name)
987 output_fd = open(output_name, O_CREAT | O_RDWR | O_TRUNC,
988 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH);
993 static int create_chunks(struct btrfs_trans_handle *trans,
994 struct btrfs_root *root, u64 num_of_meta_chunks,
999 u64 meta_type = BTRFS_BLOCK_GROUP_METADATA;
1000 u64 data_type = BTRFS_BLOCK_GROUP_DATA;
1001 u64 minimum_data_chunk_size = 8 * 1024 * 1024;
1005 for (i = 0; i < num_of_meta_chunks; i++) {
1006 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
1007 &chunk_start, &chunk_size, meta_type);
1009 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
1010 meta_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
1011 chunk_start, chunk_size);
1013 set_extent_dirty(&root->fs_info->free_space_cache,
1014 chunk_start, chunk_start + chunk_size - 1, 0);
1017 if (size_of_data < minimum_data_chunk_size)
1018 size_of_data = minimum_data_chunk_size;
1019 ret = btrfs_alloc_data_chunk(trans, root->fs_info->extent_root,
1020 &chunk_start, size_of_data, data_type);
1022 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
1023 data_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
1024 chunk_start, size_of_data);
1026 set_extent_dirty(&root->fs_info->free_space_cache,
1027 chunk_start, chunk_start + size_of_data - 1, 0);
1031 static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
1034 struct btrfs_trans_handle *trans;
1036 struct stat root_st;
1038 struct directory_name_entry dir_head;
1040 struct directory_name_entry *dir_entry = NULL;
1042 ret = lstat(source_dir, &root_st);
1044 fprintf(stderr, "unable to lstat the %s\n", source_dir);
1048 INIT_LIST_HEAD(&dir_head.list);
1050 trans = btrfs_start_transaction(root, 1);
1051 ret = traverse_directory(trans, root, source_dir, &dir_head, out_fd);
1053 fprintf(stderr, "unable to traverse_directory\n");
1056 btrfs_commit_transaction(trans, root);
1058 printf("Making image is completed.\n");
1061 while (!list_empty(&dir_head.list)) {
1062 dir_entry = list_entry(dir_head.list.next,
1063 struct directory_name_entry, list);
1064 list_del(&dir_entry->list);
1067 fprintf(stderr, "Making image is aborted.\n");
1072 * This ignores symlinks with unreadable targets and subdirs that can't
1073 * be read. It's a best-effort to give a rough estimate of the size of
1074 * a subdir. It doesn't guarantee that prepopulating btrfs from this
1075 * tree won't still run out of space.
1077 * The rounding up to 4096 is questionable. Previous code used du -B 4096.
1079 static u64 global_total_size;
1080 static int ftw_add_entry_size(const char *fpath, const struct stat *st,
1083 if (type == FTW_F || type == FTW_D)
1084 global_total_size += round_up(st->st_size, 4096);
1089 static u64 size_sourcedir(char *dir_name, u64 sectorsize,
1090 u64 *num_of_meta_chunks_ret, u64 *size_of_data_ret)
1095 u64 default_chunk_size = 8 * 1024 * 1024; /* 8MB */
1096 u64 allocated_meta_size = 8 * 1024 * 1024; /* 8MB */
1097 u64 allocated_total_size = 20 * 1024 * 1024; /* 20MB */
1098 u64 num_of_meta_chunks = 0;
1099 u64 num_of_allocated_meta_chunks =
1100 allocated_meta_size / default_chunk_size;
1102 global_total_size = 0;
1103 ret = ftw(dir_name, ftw_add_entry_size, 10);
1104 dir_size = global_total_size;
1106 fprintf(stderr, "ftw subdir walk of '%s' failed: %s\n",
1107 dir_name, strerror(errno));
1111 num_of_meta_chunks = (dir_size / 2) / default_chunk_size;
1112 if (((dir_size / 2) % default_chunk_size) != 0)
1113 num_of_meta_chunks++;
1114 if (num_of_meta_chunks <= num_of_allocated_meta_chunks)
1115 num_of_meta_chunks = 0;
1117 num_of_meta_chunks -= num_of_allocated_meta_chunks;
1119 total_size = allocated_total_size + dir_size +
1120 (num_of_meta_chunks * default_chunk_size);
1122 *num_of_meta_chunks_ret = num_of_meta_chunks;
1127 static int zero_output_file(int out_fd, u64 size, u32 sectorsize)
1129 int len = sectorsize;
1130 int loop_num = size / sectorsize;
1132 char *buf = malloc(len);
1138 memset(buf, 0, len);
1139 for (i = 0; i < loop_num; i++) {
1140 written = pwrite64(out_fd, buf, len, location);
1143 location += sectorsize;
1149 static int check_leaf_or_node_size(u32 size, u32 sectorsize)
1151 if (size < sectorsize) {
1153 "Illegal leafsize (or nodesize) %u (smaller than %u)\n",
1156 } else if (size > BTRFS_MAX_METADATA_BLOCKSIZE) {
1158 "Illegal leafsize (or nodesize) %u (larger than %u)\n",
1159 size, BTRFS_MAX_METADATA_BLOCKSIZE);
1161 } else if (size & (sectorsize - 1)) {
1163 "Illegal leafsize (or nodesize) %u (not align to %u)\n",
1170 static int is_ssd(const char *file)
1174 char sysfs_path[PATH_MAX];
1180 probe = blkid_new_probe_from_filename(file);
1184 /* Device number of this disk (possibly a partition) */
1185 devno = blkid_probe_get_devno(probe);
1187 blkid_free_probe(probe);
1191 /* Get whole disk name (not full path) for this devno */
1192 ret = blkid_devno_to_wholedisk(devno,
1193 wholedisk, sizeof(wholedisk), NULL);
1195 blkid_free_probe(probe);
1199 snprintf(sysfs_path, PATH_MAX, "/sys/block/%s/queue/rotational",
1202 blkid_free_probe(probe);
1204 fd = open(sysfs_path, O_RDONLY);
1209 if (read(fd, &rotational, sizeof(char)) < sizeof(char)) {
1215 return !atoi((const char *)&rotational);
1218 #define BTRFS_FEATURE_LIST_ALL (1ULL << 63)
1220 static const struct btrfs_fs_feature {
1224 } mkfs_features[] = {
1225 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1226 "mixed data and metadata block groups" },
1227 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1228 "increased hardlink limit per file to 65536" },
1229 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1230 "raid56 extended format" },
1231 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1232 "reduced-size metadata extent refs" },
1233 /* Keep this one last */
1234 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1237 static void list_all_fs_features(void)
1241 fprintf(stderr, "Filesystem features available at mkfs time:\n");
1242 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1243 fprintf(stderr, "%-20s- %s (0x%llx)\n",
1244 mkfs_features[i].name,
1245 mkfs_features[i].desc,
1246 mkfs_features[i].flag);
1250 static int parse_one_fs_feature(const char *name, u64 *flags)
1255 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1256 if (!strcmp(mkfs_features[i].name, name)) {
1257 *flags |= mkfs_features[i].flag;
1265 static void process_fs_features(u64 flags)
1269 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1270 if (flags & mkfs_features[i].flag) {
1272 "Turning ON incompat feature '%s': %s\n",
1273 mkfs_features[i].name,
1274 mkfs_features[i].desc);
1281 * Return NULL if all features were parsed fine, otherwise return the name of
1282 * the first unparsed.
1284 static char* parse_fs_features(char *namelist, u64 *flags)
1287 char *save_ptr = NULL; /* Satisfy static checkers */
1289 for (this_char = strtok_r(namelist, ",", &save_ptr);
1291 this_char = strtok_r(NULL, ",", &save_ptr)) {
1292 if (parse_one_fs_feature(this_char, flags))
1299 int main(int ac, char **av)
1302 struct btrfs_root *root;
1303 struct btrfs_trans_handle *trans;
1306 u64 block_count = 0;
1307 u64 dev_block_count = 0;
1309 u64 alloc_start = 0;
1310 u64 metadata_profile = 0;
1311 u64 data_profile = 0;
1312 u32 leafsize = sysconf(_SC_PAGESIZE);
1313 u32 sectorsize = 4096;
1314 u32 nodesize = leafsize;
1315 u32 stripesize = 4096;
1317 int option_index = 0;
1322 int data_profile_opt = 0;
1323 int metadata_profile_opt = 0;
1326 int force_overwrite = 0;
1328 char *source_dir = NULL;
1329 int source_dir_set = 0;
1330 u64 num_of_meta_chunks = 0;
1331 u64 size_of_data = 0;
1332 u64 source_dir_size = 0;
1340 c = getopt_long(ac, av, "A:b:fl:n:s:m:d:L:O:r:VMK",
1341 long_options, &option_index);
1346 alloc_start = parse_size(optarg);
1349 force_overwrite = 1;
1352 data_profile = parse_profile(optarg);
1353 data_profile_opt = 1;
1357 nodesize = parse_size(optarg);
1358 leafsize = parse_size(optarg);
1361 label = parse_label(optarg);
1364 metadata_profile = parse_profile(optarg);
1365 metadata_profile_opt = 1;
1371 char *orig = strdup(optarg);
1374 tmp = parse_fs_features(tmp, &features);
1377 "Unrecognized filesystem feature '%s'\n",
1383 if (features & BTRFS_FEATURE_LIST_ALL) {
1384 list_all_fs_features();
1390 sectorsize = parse_size(optarg);
1393 block_count = parse_size(optarg);
1394 if (block_count <= 1024*1024*1024) {
1395 printf("SMALL VOLUME: forcing mixed "
1396 "metadata/data groups\n");
1405 source_dir = optarg;
1415 sectorsize = max(sectorsize, (u32)sysconf(_SC_PAGESIZE));
1416 if (check_leaf_or_node_size(leafsize, sectorsize))
1418 if (check_leaf_or_node_size(nodesize, sectorsize))
1420 saved_optind = optind;
1421 dev_cnt = ac - optind;
1425 if (source_dir_set && dev_cnt > 1) {
1427 "The -r option is limited to a single device\n");
1430 while (dev_cnt-- > 0) {
1431 file = av[optind++];
1432 if (is_block_device(file))
1433 if (test_dev_for_mkfs(file, force_overwrite, estr)) {
1434 fprintf(stderr, "Error: %s", estr);
1439 optind = saved_optind;
1440 dev_cnt = ac - optind;
1442 file = av[optind++];
1445 if (is_vol_small(file)) {
1446 printf("SMALL VOLUME: forcing mixed metadata/data groups\n");
1448 if (metadata_profile != data_profile) {
1449 if (metadata_profile_opt || data_profile_opt) {
1451 "With mixed block groups data and metadata profiles must be the same\n");
1457 * Set default profiles according to number of added devices.
1458 * For mixed groups defaults are single/single.
1461 if (!metadata_profile_opt) {
1462 if (dev_cnt == 1 && ssd)
1463 printf("Detected a SSD, turning off metadata "
1464 "duplication. Mkfs with -m dup if you want to "
1465 "force metadata duplication.\n");
1467 metadata_profile = (dev_cnt > 1) ?
1468 BTRFS_BLOCK_GROUP_RAID1 : (ssd) ?
1469 0: BTRFS_BLOCK_GROUP_DUP;
1471 if (!data_profile_opt) {
1472 data_profile = (dev_cnt > 1) ?
1473 BTRFS_BLOCK_GROUP_RAID0 : 0; /* raid0 or single */
1476 metadata_profile = 0;
1480 ret = test_num_disk_vs_raid(metadata_profile, data_profile,
1481 dev_cnt, mixed, estr);
1483 fprintf(stderr, "Error: %s\n", estr);
1487 /* if we are here that means all devs are good to btrfsify */
1488 printf("\nWARNING! - %s IS EXPERIMENTAL\n", BTRFS_BUILD_VERSION);
1489 printf("WARNING! - see http://btrfs.wiki.kernel.org before using\n\n");
1493 if (!source_dir_set) {
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,
1506 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1507 block_count, &mixed, nodiscard);
1508 if (block_count && block_count > dev_block_count) {
1509 fprintf(stderr, "%s is smaller than requested size\n", file);
1513 fd = open_target(file);
1515 fprintf(stderr, "unable to open the %s\n", file);
1520 source_dir_size = size_sourcedir(source_dir, sectorsize,
1521 &num_of_meta_chunks, &size_of_data);
1522 if(block_count < source_dir_size)
1523 block_count = source_dir_size;
1524 ret = zero_output_file(fd, block_count, sectorsize);
1526 fprintf(stderr, "unable to zero the output file\n");
1529 /* our "device" is the new image file */
1530 dev_block_count = block_count;
1533 /* To create the first block group and chunk 0 in make_btrfs */
1534 if (dev_block_count < BTRFS_MKFS_SYSTEM_GROUP_SIZE) {
1535 fprintf(stderr, "device is too small to make filesystem\n");
1539 blocks[0] = BTRFS_SUPER_INFO_OFFSET;
1540 for (i = 1; i < 7; i++) {
1541 blocks[i] = BTRFS_SUPER_INFO_OFFSET + 1024 * 1024 +
1546 * FS features that can be set by other means than -O
1547 * just set the bit here
1550 features |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
1552 if ((data_profile | metadata_profile) &
1553 (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
1554 features |= BTRFS_FEATURE_INCOMPAT_RAID56;
1557 process_fs_features(features);
1559 ret = make_btrfs(fd, file, label, blocks, dev_block_count,
1561 sectorsize, stripesize, features);
1563 fprintf(stderr, "error during mkfs: %s\n", strerror(-ret));
1567 root = open_ctree(file, 0, O_RDWR);
1569 fprintf(stderr, "Open ctree failed\n");
1573 root->fs_info->alloc_start = alloc_start;
1575 ret = make_root_dir(root, mixed);
1577 fprintf(stderr, "failed to setup the root directory\n");
1581 trans = btrfs_start_transaction(root, 1);
1586 btrfs_register_one_device(file);
1589 while (dev_cnt-- > 0) {
1590 int old_mixed = mixed;
1592 file = av[optind++];
1595 * open without O_EXCL so that the problem should not
1596 * occur by the following processing.
1597 * (btrfs_register_one_device() fails if O_EXCL is on)
1599 fd = open(file, O_RDWR);
1601 fprintf(stderr, "unable to open %s: %s\n", file,
1605 ret = btrfs_device_already_in_root(root, fd,
1606 BTRFS_SUPER_INFO_OFFSET);
1608 fprintf(stderr, "skipping duplicate device %s in FS\n",
1613 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1614 block_count, &mixed, nodiscard);
1618 ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
1619 sectorsize, sectorsize, sectorsize);
1621 btrfs_register_one_device(file);
1625 if (!source_dir_set) {
1626 ret = create_raid_groups(trans, root, data_profile,
1627 data_profile_opt, metadata_profile,
1628 metadata_profile_opt, mixed, ssd);
1632 ret = create_data_reloc_tree(trans, root);
1635 printf("fs created label %s on %s\n\tnodesize %u leafsize %u "
1636 "sectorsize %u size %s\n",
1637 label, first_file, nodesize, leafsize, sectorsize,
1638 pretty_size(btrfs_super_total_bytes(root->fs_info->super_copy)));
1640 printf("%s\n", BTRFS_BUILD_VERSION);
1641 btrfs_commit_transaction(trans, root);
1643 if (source_dir_set) {
1644 trans = btrfs_start_transaction(root, 1);
1645 ret = create_chunks(trans, root,
1646 num_of_meta_chunks, size_of_data);
1648 btrfs_commit_transaction(trans, root);
1650 ret = make_image(source_dir, root, fd);
1654 ret = close_ctree(root);