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 <sys/xattr.h>
38 #include <blkid/blkid.h>
43 #include "transaction.h"
47 static u64 index_cnt = 2;
49 #define DEFAULT_MKFS_FEATURES (BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)
51 #define DEFAULT_MKFS_LEAF_SIZE 16384
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);
86 "no space to alloc data/metadata chunk\n");
90 ret = btrfs_make_block_group(trans, root, 0,
91 BTRFS_BLOCK_GROUP_METADATA |
92 BTRFS_BLOCK_GROUP_DATA,
93 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
94 chunk_start, chunk_size);
96 printf("Created a data/metadata chunk of size %llu\n", chunk_size);
98 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
99 &chunk_start, &chunk_size,
100 BTRFS_BLOCK_GROUP_METADATA);
101 if (ret == -ENOSPC) {
102 fprintf(stderr, "no space to alloc metadata chunk\n");
106 ret = btrfs_make_block_group(trans, root, 0,
107 BTRFS_BLOCK_GROUP_METADATA,
108 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
109 chunk_start, chunk_size);
113 root->fs_info->system_allocs = 0;
114 btrfs_commit_transaction(trans, root);
115 trans = btrfs_start_transaction(root, 1);
119 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
120 &chunk_start, &chunk_size,
121 BTRFS_BLOCK_GROUP_DATA);
122 if (ret == -ENOSPC) {
123 fprintf(stderr, "no space to alloc data chunk\n");
127 ret = btrfs_make_block_group(trans, root, 0,
128 BTRFS_BLOCK_GROUP_DATA,
129 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
130 chunk_start, chunk_size);
134 ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
135 BTRFS_ROOT_TREE_DIR_OBJECTID);
138 ret = btrfs_make_root_dir(trans, root, BTRFS_FIRST_FREE_OBJECTID);
141 memcpy(&location, &root->fs_info->fs_root->root_key, sizeof(location));
142 location.offset = (u64)-1;
143 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
145 btrfs_super_root_dir(root->fs_info->super_copy),
146 &location, BTRFS_FT_DIR, 0);
150 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
151 "default", 7, location.objectid,
152 BTRFS_ROOT_TREE_DIR_OBJECTID, 0);
156 btrfs_commit_transaction(trans, root);
161 static void __recow_root(struct btrfs_trans_handle *trans,
162 struct btrfs_root *root)
165 struct extent_buffer *tmp;
167 if (trans->transid != btrfs_root_generation(&root->root_item)) {
168 extent_buffer_get(root->node);
169 ret = __btrfs_cow_block(trans, root, root->node,
170 NULL, 0, &tmp, 0, 0);
172 free_extent_buffer(tmp);
176 static void recow_roots(struct btrfs_trans_handle *trans,
177 struct btrfs_root *root)
179 struct btrfs_fs_info *info = root->fs_info;
181 __recow_root(trans, info->fs_root);
182 __recow_root(trans, info->tree_root);
183 __recow_root(trans, info->extent_root);
184 __recow_root(trans, info->chunk_root);
185 __recow_root(trans, info->dev_root);
186 __recow_root(trans, info->csum_root);
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);
198 if (ret == -ENOSPC) {
199 fprintf(stderr, "not enough free space\n");
203 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
204 type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
205 chunk_start, chunk_size);
210 static int create_raid_groups(struct btrfs_trans_handle *trans,
211 struct btrfs_root *root, u64 data_profile,
212 int data_profile_opt, u64 metadata_profile,
215 u64 num_devices = btrfs_super_num_devices(root->fs_info->super_copy);
218 if (metadata_profile) {
219 u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
221 ret = create_one_raid_group(trans, root,
222 BTRFS_BLOCK_GROUP_SYSTEM |
227 meta_flags |= BTRFS_BLOCK_GROUP_DATA;
229 ret = create_one_raid_group(trans, root, meta_flags |
234 if (!mixed && num_devices > 1 && data_profile) {
235 ret = create_one_raid_group(trans, root,
236 BTRFS_BLOCK_GROUP_DATA |
240 recow_roots(trans, root);
245 static int create_data_reloc_tree(struct btrfs_trans_handle *trans,
246 struct btrfs_root *root)
248 struct btrfs_key location;
249 struct btrfs_root_item root_item;
250 struct extent_buffer *tmp;
251 u64 objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
254 ret = btrfs_copy_root(trans, root, root->node, &tmp, objectid);
257 memcpy(&root_item, &root->root_item, sizeof(root_item));
258 btrfs_set_root_bytenr(&root_item, tmp->start);
259 btrfs_set_root_level(&root_item, btrfs_header_level(tmp));
260 btrfs_set_root_generation(&root_item, trans->transid);
261 free_extent_buffer(tmp);
263 location.objectid = objectid;
264 location.type = BTRFS_ROOT_ITEM_KEY;
266 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
267 &location, &root_item);
273 static void print_usage(void) __attribute__((noreturn));
274 static void print_usage(void)
276 fprintf(stderr, "usage: mkfs.btrfs [options] dev [ dev ... ]\n");
277 fprintf(stderr, "options:\n");
278 fprintf(stderr, "\t -A --alloc-start the offset to start the FS\n");
279 fprintf(stderr, "\t -b --byte-count total number of bytes in the FS\n");
280 fprintf(stderr, "\t -d --data data profile, raid0, raid1, raid5, raid6, raid10, dup or single\n");
281 fprintf(stderr, "\t -f --force force overwrite of existing filesystem\n");
282 fprintf(stderr, "\t -l --leafsize size of btree leaves\n");
283 fprintf(stderr, "\t -L --label set a label\n");
284 fprintf(stderr, "\t -m --metadata metadata profile, values like data profile\n");
285 fprintf(stderr, "\t -M --mixed mix metadata and data together\n");
286 fprintf(stderr, "\t -n --nodesize size of btree nodes\n");
287 fprintf(stderr, "\t -s --sectorsize min block allocation (may not mountable by current kernel)\n");
288 fprintf(stderr, "\t -r --rootdir the source directory\n");
289 fprintf(stderr, "\t -K --nodiscard do not perform whole device TRIM\n");
290 fprintf(stderr, "\t -O --features comma separated list of filesystem features\n");
291 fprintf(stderr, "\t -U --uuid specify the filesystem UUID\n");
292 fprintf(stderr, "\t -V --version print the mkfs.btrfs version and exit\n");
293 fprintf(stderr, "%s\n", BTRFS_BUILD_VERSION);
297 static void print_version(void) __attribute__((noreturn));
298 static void print_version(void)
300 fprintf(stderr, "mkfs.btrfs, part of %s\n", BTRFS_BUILD_VERSION);
304 static u64 parse_profile(char *s)
306 if (strcmp(s, "raid0") == 0) {
307 return BTRFS_BLOCK_GROUP_RAID0;
308 } else if (strcmp(s, "raid1") == 0) {
309 return BTRFS_BLOCK_GROUP_RAID1;
310 } else if (strcmp(s, "raid5") == 0) {
311 return BTRFS_BLOCK_GROUP_RAID5;
312 } else if (strcmp(s, "raid6") == 0) {
313 return BTRFS_BLOCK_GROUP_RAID6;
314 } else if (strcmp(s, "raid10") == 0) {
315 return BTRFS_BLOCK_GROUP_RAID10;
316 } else if (strcmp(s, "dup") == 0) {
317 return BTRFS_BLOCK_GROUP_DUP;
318 } else if (strcmp(s, "single") == 0) {
321 fprintf(stderr, "Unknown profile %s\n", s);
328 static char *parse_label(char *input)
330 int len = strlen(input);
332 if (len >= BTRFS_LABEL_SIZE) {
333 fprintf(stderr, "Label %s is too long (max %d)\n", input,
334 BTRFS_LABEL_SIZE - 1);
337 return strdup(input);
340 static struct option long_options[] = {
341 { "alloc-start", 1, NULL, 'A'},
342 { "byte-count", 1, NULL, 'b' },
343 { "force", 0, NULL, 'f' },
344 { "leafsize", 1, NULL, 'l' },
345 { "label", 1, NULL, 'L'},
346 { "metadata", 1, NULL, 'm' },
347 { "mixed", 0, NULL, 'M' },
348 { "nodesize", 1, NULL, 'n' },
349 { "sectorsize", 1, NULL, 's' },
350 { "data", 1, NULL, 'd' },
351 { "version", 0, NULL, 'V' },
352 { "rootdir", 1, NULL, 'r' },
353 { "nodiscard", 0, NULL, 'K' },
354 { "features", 1, NULL, 'O' },
355 { "uuid", 0, NULL, 'U' },
359 static int add_directory_items(struct btrfs_trans_handle *trans,
360 struct btrfs_root *root, u64 objectid,
361 ino_t parent_inum, const char *name,
362 struct stat *st, int *dir_index_cnt)
366 struct btrfs_key location;
369 name_len = strlen(name);
371 location.objectid = objectid;
373 btrfs_set_key_type(&location, BTRFS_INODE_ITEM_KEY);
375 if (S_ISDIR(st->st_mode))
376 filetype = BTRFS_FT_DIR;
377 if (S_ISREG(st->st_mode))
378 filetype = BTRFS_FT_REG_FILE;
379 if (S_ISLNK(st->st_mode))
380 filetype = BTRFS_FT_SYMLINK;
382 ret = btrfs_insert_dir_item(trans, root, name, name_len,
383 parent_inum, &location,
384 filetype, index_cnt);
387 ret = btrfs_insert_inode_ref(trans, root, name, name_len,
388 objectid, parent_inum, index_cnt);
389 *dir_index_cnt = index_cnt;
395 static int fill_inode_item(struct btrfs_trans_handle *trans,
396 struct btrfs_root *root,
397 struct btrfs_inode_item *dst, struct stat *src)
400 u64 sectorsize = root->sectorsize;
403 * btrfs_inode_item has some reserved fields
404 * and represents on-disk inode entry, so
405 * zero everything to prevent information leak
407 memset(dst, 0, sizeof (*dst));
409 btrfs_set_stack_inode_generation(dst, trans->transid);
410 btrfs_set_stack_inode_size(dst, src->st_size);
411 btrfs_set_stack_inode_nbytes(dst, 0);
412 btrfs_set_stack_inode_block_group(dst, 0);
413 btrfs_set_stack_inode_nlink(dst, src->st_nlink);
414 btrfs_set_stack_inode_uid(dst, src->st_uid);
415 btrfs_set_stack_inode_gid(dst, src->st_gid);
416 btrfs_set_stack_inode_mode(dst, src->st_mode);
417 btrfs_set_stack_inode_rdev(dst, 0);
418 btrfs_set_stack_inode_flags(dst, 0);
419 btrfs_set_stack_timespec_sec(&dst->atime, src->st_atime);
420 btrfs_set_stack_timespec_nsec(&dst->atime, 0);
421 btrfs_set_stack_timespec_sec(&dst->ctime, src->st_ctime);
422 btrfs_set_stack_timespec_nsec(&dst->ctime, 0);
423 btrfs_set_stack_timespec_sec(&dst->mtime, src->st_mtime);
424 btrfs_set_stack_timespec_nsec(&dst->mtime, 0);
425 btrfs_set_stack_timespec_sec(&dst->otime, 0);
426 btrfs_set_stack_timespec_nsec(&dst->otime, 0);
428 if (S_ISDIR(src->st_mode)) {
429 btrfs_set_stack_inode_size(dst, 0);
430 btrfs_set_stack_inode_nlink(dst, 1);
432 if (S_ISREG(src->st_mode)) {
433 btrfs_set_stack_inode_size(dst, (u64)src->st_size);
434 if (src->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root))
435 btrfs_set_stack_inode_nbytes(dst, src->st_size);
437 blocks = src->st_size / sectorsize;
438 if (src->st_size % sectorsize)
440 blocks *= sectorsize;
441 btrfs_set_stack_inode_nbytes(dst, blocks);
444 if (S_ISLNK(src->st_mode))
445 btrfs_set_stack_inode_nbytes(dst, src->st_size + 1);
450 static int directory_select(const struct direct *entry)
452 if ((strncmp(entry->d_name, ".", entry->d_reclen) == 0) ||
453 (strncmp(entry->d_name, "..", entry->d_reclen) == 0))
459 static void free_namelist(struct direct **files, int count)
466 for (i = 0; i < count; ++i)
471 static u64 calculate_dir_inode_size(char *dirname)
474 struct direct **files, *cur_file;
475 u64 dir_inode_size = 0;
477 count = scandir(dirname, &files, directory_select, NULL);
479 for (i = 0; i < count; i++) {
481 dir_inode_size += strlen(cur_file->d_name);
484 free_namelist(files, count);
487 return dir_inode_size;
490 static int add_inode_items(struct btrfs_trans_handle *trans,
491 struct btrfs_root *root,
492 struct stat *st, char *name,
493 u64 self_objectid, ino_t parent_inum,
494 int dir_index_cnt, struct btrfs_inode_item *inode_ret)
497 struct btrfs_key inode_key;
498 struct btrfs_inode_item btrfs_inode;
502 fill_inode_item(trans, root, &btrfs_inode, st);
503 objectid = self_objectid;
505 if (S_ISDIR(st->st_mode)) {
506 inode_size = calculate_dir_inode_size(name);
507 btrfs_set_stack_inode_size(&btrfs_inode, inode_size);
510 inode_key.objectid = objectid;
511 inode_key.offset = 0;
512 btrfs_set_key_type(&inode_key, BTRFS_INODE_ITEM_KEY);
514 ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
516 *inode_ret = btrfs_inode;
520 static int add_xattr_item(struct btrfs_trans_handle *trans,
521 struct btrfs_root *root, u64 objectid,
522 const char *file_name)
526 char xattr_list[XATTR_LIST_MAX];
528 char cur_value[XATTR_SIZE_MAX];
529 char delimiter = '\0';
530 char *next_location = xattr_list;
532 ret = llistxattr(file_name, xattr_list, XATTR_LIST_MAX);
536 fprintf(stderr, "get a list of xattr failed for %s\n",
543 cur_name = strtok(xattr_list, &delimiter);
544 while (cur_name != NULL) {
545 cur_name_len = strlen(cur_name);
546 next_location += cur_name_len + 1;
548 ret = getxattr(file_name, cur_name, cur_value, XATTR_SIZE_MAX);
552 fprintf(stderr, "get a xattr value failed for %s attr %s\n",
553 file_name, cur_name);
557 ret = btrfs_insert_xattr_item(trans, root, cur_name,
558 cur_name_len, cur_value,
561 fprintf(stderr, "insert a xattr item failed for %s\n",
565 cur_name = strtok(next_location, &delimiter);
571 static int add_symbolic_link(struct btrfs_trans_handle *trans,
572 struct btrfs_root *root,
573 u64 objectid, const char *path_name)
576 u64 sectorsize = root->sectorsize;
577 char *buf = malloc(sectorsize);
579 ret = readlink(path_name, buf, sectorsize);
581 fprintf(stderr, "readlink failed for %s\n", path_name);
584 if (ret >= sectorsize) {
585 fprintf(stderr, "symlink too long for %s", path_name);
590 buf[ret] = '\0'; /* readlink does not do it for us */
591 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
598 static int add_file_items(struct btrfs_trans_handle *trans,
599 struct btrfs_root *root,
600 struct btrfs_inode_item *btrfs_inode, u64 objectid,
601 ino_t parent_inum, struct stat *st,
602 const char *path_name, int out_fd)
607 struct btrfs_key key;
609 u32 sectorsize = root->sectorsize;
614 struct extent_buffer *eb = NULL;
617 if (st->st_size == 0)
620 fd = open(path_name, O_RDONLY);
622 fprintf(stderr, "%s open failed\n", path_name);
626 blocks = st->st_size / sectorsize;
627 if (st->st_size % sectorsize)
630 if (st->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
631 char *buffer = malloc(st->st_size);
632 ret_read = pread64(fd, buffer, st->st_size, bytes_read);
633 if (ret_read == -1) {
634 fprintf(stderr, "%s read failed\n", path_name);
639 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
640 buffer, st->st_size);
645 /* round up our st_size to the FS blocksize */
646 total_bytes = (u64)blocks * sectorsize;
649 * do our IO in extent buffers so it can work
650 * against any raid type
652 eb = malloc(sizeof(*eb) + sectorsize);
657 memset(eb, 0, sizeof(*eb) + sectorsize);
662 * keep our extent size at 1MB max, this makes it easier to work inside
663 * the tiny block groups created during mkfs
665 cur_bytes = min(total_bytes, 1024ULL * 1024);
666 ret = btrfs_reserve_extent(trans, root, cur_bytes, 0, 0, (u64)-1,
671 first_block = key.objectid;
674 while (bytes_read < cur_bytes) {
676 memset(eb->data, 0, sectorsize);
678 ret_read = pread64(fd, eb->data, sectorsize, file_pos + bytes_read);
679 if (ret_read == -1) {
680 fprintf(stderr, "%s read failed\n", path_name);
684 eb->start = first_block + bytes_read;
685 eb->len = sectorsize;
688 * we're doing the csum before we record the extent, but
691 ret = btrfs_csum_file_block(trans, root->fs_info->csum_root,
692 first_block + bytes_read + sectorsize,
693 first_block + bytes_read,
694 eb->data, sectorsize);
698 ret = write_and_map_eb(trans, root, eb);
700 fprintf(stderr, "output file write failed\n");
704 bytes_read += sectorsize;
708 ret = btrfs_record_file_extent(trans, root, objectid, btrfs_inode,
709 file_pos, first_block, cur_bytes);
715 file_pos += cur_bytes;
716 total_bytes -= cur_bytes;
727 static char *make_path(char *dir, char *name)
731 path = malloc(strlen(dir) + strlen(name) + 2);
735 if (dir[strlen(dir) - 1] != '/')
741 static int traverse_directory(struct btrfs_trans_handle *trans,
742 struct btrfs_root *root, char *dir_name,
743 struct directory_name_entry *dir_head, int out_fd)
747 struct btrfs_inode_item cur_inode;
748 struct btrfs_inode_item *inode_item;
749 int count, i, dir_index_cnt;
750 struct direct **files;
752 struct directory_name_entry *dir_entry, *parent_dir_entry;
753 struct direct *cur_file;
754 ino_t parent_inum, cur_inum;
755 ino_t highest_inum = 0;
756 char *parent_dir_name;
757 char real_path[PATH_MAX];
758 struct btrfs_path path;
759 struct extent_buffer *leaf;
760 struct btrfs_key root_dir_key;
761 u64 root_dir_inode_size = 0;
763 /* Add list for source directory */
764 dir_entry = malloc(sizeof(struct directory_name_entry));
765 dir_entry->dir_name = dir_name;
766 dir_entry->path = realpath(dir_name, real_path);
767 if (!dir_entry->path) {
768 fprintf(stderr, "get directory real path error\n");
773 parent_inum = highest_inum + BTRFS_FIRST_FREE_OBJECTID;
774 dir_entry->inum = parent_inum;
775 list_add_tail(&dir_entry->list, &dir_head->list);
777 btrfs_init_path(&path);
779 root_dir_key.objectid = btrfs_root_dirid(&root->root_item);
780 root_dir_key.offset = 0;
781 btrfs_set_key_type(&root_dir_key, BTRFS_INODE_ITEM_KEY);
782 ret = btrfs_lookup_inode(trans, root, &path, &root_dir_key, 1);
784 fprintf(stderr, "root dir lookup error\n");
788 leaf = path.nodes[0];
789 inode_item = btrfs_item_ptr(leaf, path.slots[0],
790 struct btrfs_inode_item);
792 root_dir_inode_size = calculate_dir_inode_size(dir_name);
793 btrfs_set_inode_size(leaf, inode_item, root_dir_inode_size);
794 btrfs_mark_buffer_dirty(leaf);
796 btrfs_release_path(&path);
799 parent_dir_entry = list_entry(dir_head->list.next,
800 struct directory_name_entry,
802 list_del(&parent_dir_entry->list);
804 parent_inum = parent_dir_entry->inum;
805 parent_dir_name = parent_dir_entry->dir_name;
806 if (chdir(parent_dir_entry->path)) {
807 fprintf(stderr, "chdir error for %s\n",
813 count = scandir(parent_dir_entry->path, &files,
814 directory_select, NULL);
817 fprintf(stderr, "scandir for %s failed: %s\n",
818 parent_dir_name, strerror (errno));
823 for (i = 0; i < count; i++) {
826 if (lstat(cur_file->d_name, &st) == -1) {
827 fprintf(stderr, "lstat failed for file %s\n",
833 cur_inum = st.st_ino;
834 ret = add_directory_items(trans, root,
835 cur_inum, parent_inum,
837 &st, &dir_index_cnt);
839 fprintf(stderr, "add_directory_items failed\n");
843 ret = add_inode_items(trans, root, &st,
844 cur_file->d_name, cur_inum,
845 parent_inum, dir_index_cnt,
847 if (ret == -EEXIST) {
848 BUG_ON(st.st_nlink <= 1);
852 fprintf(stderr, "add_inode_items failed\n");
856 ret = add_xattr_item(trans, root,
857 cur_inum, cur_file->d_name);
859 fprintf(stderr, "add_xattr_item failed\n");
864 if (S_ISDIR(st.st_mode)) {
865 dir_entry = malloc(sizeof(struct directory_name_entry));
866 dir_entry->dir_name = cur_file->d_name;
867 dir_entry->path = make_path(parent_dir_entry->path,
869 dir_entry->inum = cur_inum;
870 list_add_tail(&dir_entry->list, &dir_head->list);
871 } else if (S_ISREG(st.st_mode)) {
872 ret = add_file_items(trans, root, &cur_inode,
873 cur_inum, parent_inum, &st,
874 cur_file->d_name, out_fd);
876 fprintf(stderr, "add_file_items failed\n");
879 } else if (S_ISLNK(st.st_mode)) {
880 ret = add_symbolic_link(trans, root,
881 cur_inum, cur_file->d_name);
883 fprintf(stderr, "add_symbolic_link failed\n");
889 free_namelist(files, count);
890 free(parent_dir_entry);
894 } while (!list_empty(&dir_head->list));
899 free_namelist(files, count);
901 free(parent_dir_entry);
908 static int open_target(char *output_name)
911 output_fd = open(output_name, O_CREAT | O_RDWR | O_TRUNC,
912 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH);
917 static int create_chunks(struct btrfs_trans_handle *trans,
918 struct btrfs_root *root, u64 num_of_meta_chunks,
923 u64 meta_type = BTRFS_BLOCK_GROUP_METADATA;
924 u64 data_type = BTRFS_BLOCK_GROUP_DATA;
925 u64 minimum_data_chunk_size = 8 * 1024 * 1024;
929 for (i = 0; i < num_of_meta_chunks; i++) {
930 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
931 &chunk_start, &chunk_size, meta_type);
933 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
934 meta_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
935 chunk_start, chunk_size);
937 set_extent_dirty(&root->fs_info->free_space_cache,
938 chunk_start, chunk_start + chunk_size - 1, 0);
941 if (size_of_data < minimum_data_chunk_size)
942 size_of_data = minimum_data_chunk_size;
944 ret = btrfs_alloc_data_chunk(trans, root->fs_info->extent_root,
945 &chunk_start, size_of_data, data_type);
947 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
948 data_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
949 chunk_start, size_of_data);
951 set_extent_dirty(&root->fs_info->free_space_cache,
952 chunk_start, chunk_start + size_of_data - 1, 0);
956 static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
959 struct btrfs_trans_handle *trans;
963 struct directory_name_entry dir_head;
965 struct directory_name_entry *dir_entry = NULL;
967 ret = lstat(source_dir, &root_st);
969 fprintf(stderr, "unable to lstat the %s\n", source_dir);
973 INIT_LIST_HEAD(&dir_head.list);
975 trans = btrfs_start_transaction(root, 1);
976 ret = traverse_directory(trans, root, source_dir, &dir_head, out_fd);
978 fprintf(stderr, "unable to traverse_directory\n");
981 btrfs_commit_transaction(trans, root);
983 printf("Making image is completed.\n");
986 while (!list_empty(&dir_head.list)) {
987 dir_entry = list_entry(dir_head.list.next,
988 struct directory_name_entry, list);
989 list_del(&dir_entry->list);
993 fprintf(stderr, "Making image is aborted.\n");
998 * This ignores symlinks with unreadable targets and subdirs that can't
999 * be read. It's a best-effort to give a rough estimate of the size of
1000 * a subdir. It doesn't guarantee that prepopulating btrfs from this
1001 * tree won't still run out of space.
1003 * The rounding up to 4096 is questionable. Previous code used du -B 4096.
1005 static u64 global_total_size;
1006 static int ftw_add_entry_size(const char *fpath, const struct stat *st,
1009 if (type == FTW_F || type == FTW_D)
1010 global_total_size += round_up(st->st_size, 4096);
1015 static u64 size_sourcedir(char *dir_name, u64 sectorsize,
1016 u64 *num_of_meta_chunks_ret, u64 *size_of_data_ret)
1021 u64 default_chunk_size = 8 * 1024 * 1024; /* 8MB */
1022 u64 allocated_meta_size = 8 * 1024 * 1024; /* 8MB */
1023 u64 allocated_total_size = 20 * 1024 * 1024; /* 20MB */
1024 u64 num_of_meta_chunks = 0;
1025 u64 num_of_data_chunks = 0;
1026 u64 num_of_allocated_meta_chunks =
1027 allocated_meta_size / default_chunk_size;
1029 global_total_size = 0;
1030 ret = ftw(dir_name, ftw_add_entry_size, 10);
1031 dir_size = global_total_size;
1033 fprintf(stderr, "ftw subdir walk of '%s' failed: %s\n",
1034 dir_name, strerror(errno));
1038 num_of_data_chunks = (dir_size + default_chunk_size - 1) /
1041 num_of_meta_chunks = (dir_size / 2) / default_chunk_size;
1042 if (((dir_size / 2) % default_chunk_size) != 0)
1043 num_of_meta_chunks++;
1044 if (num_of_meta_chunks <= num_of_allocated_meta_chunks)
1045 num_of_meta_chunks = 0;
1047 num_of_meta_chunks -= num_of_allocated_meta_chunks;
1049 total_size = allocated_total_size +
1050 (num_of_data_chunks * default_chunk_size) +
1051 (num_of_meta_chunks * default_chunk_size);
1053 *num_of_meta_chunks_ret = num_of_meta_chunks;
1054 *size_of_data_ret = num_of_data_chunks * default_chunk_size;
1058 static int zero_output_file(int out_fd, u64 size, u32 sectorsize)
1060 int len = sectorsize;
1061 int loop_num = size / sectorsize;
1063 char *buf = malloc(len);
1069 memset(buf, 0, len);
1070 for (i = 0; i < loop_num; i++) {
1071 written = pwrite64(out_fd, buf, len, location);
1074 location += sectorsize;
1080 static int check_leaf_or_node_size(u32 size, u32 sectorsize)
1082 if (size < sectorsize) {
1084 "Illegal leafsize (or nodesize) %u (smaller than %u)\n",
1087 } else if (size > BTRFS_MAX_METADATA_BLOCKSIZE) {
1089 "Illegal leafsize (or nodesize) %u (larger than %u)\n",
1090 size, BTRFS_MAX_METADATA_BLOCKSIZE);
1092 } else if (size & (sectorsize - 1)) {
1094 "Illegal leafsize (or nodesize) %u (not align to %u)\n",
1101 static int is_ssd(const char *file)
1105 char sysfs_path[PATH_MAX];
1111 probe = blkid_new_probe_from_filename(file);
1115 /* Device number of this disk (possibly a partition) */
1116 devno = blkid_probe_get_devno(probe);
1118 blkid_free_probe(probe);
1122 /* Get whole disk name (not full path) for this devno */
1123 ret = blkid_devno_to_wholedisk(devno,
1124 wholedisk, sizeof(wholedisk), NULL);
1126 blkid_free_probe(probe);
1130 snprintf(sysfs_path, PATH_MAX, "/sys/block/%s/queue/rotational",
1133 blkid_free_probe(probe);
1135 fd = open(sysfs_path, O_RDONLY);
1140 if (read(fd, &rotational, sizeof(char)) < sizeof(char)) {
1146 return !atoi((const char *)&rotational);
1149 #define BTRFS_FEATURE_LIST_ALL (1ULL << 63)
1151 static const struct btrfs_fs_feature {
1155 } mkfs_features[] = {
1156 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1157 "mixed data and metadata block groups" },
1158 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1159 "increased hardlink limit per file to 65536" },
1160 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1161 "raid56 extended format" },
1162 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1163 "reduced-size metadata extent refs" },
1164 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1165 "no explicit hole extents for files" },
1166 /* Keep this one last */
1167 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1170 static void list_all_fs_features(void)
1174 fprintf(stderr, "Filesystem features available at mkfs time:\n");
1175 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1176 char *is_default = "";
1178 if (mkfs_features[i].flag & DEFAULT_MKFS_FEATURES)
1179 is_default = ", default";
1180 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
1181 mkfs_features[i].name,
1182 mkfs_features[i].desc,
1183 mkfs_features[i].flag,
1188 static int parse_one_fs_feature(const char *name, u64 *flags)
1193 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1194 if (name[0] == '^' &&
1195 !strcmp(mkfs_features[i].name, name + 1)) {
1196 *flags &= ~ mkfs_features[i].flag;
1198 } else if (!strcmp(mkfs_features[i].name, name)) {
1199 *flags |= mkfs_features[i].flag;
1207 static void process_fs_features(u64 flags)
1211 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1212 if (flags & mkfs_features[i].flag) {
1213 printf("Turning ON incompat feature '%s': %s\n",
1214 mkfs_features[i].name,
1215 mkfs_features[i].desc);
1222 * Return NULL if all features were parsed fine, otherwise return the name of
1223 * the first unparsed.
1225 static char* parse_fs_features(char *namelist, u64 *flags)
1228 char *save_ptr = NULL; /* Satisfy static checkers */
1230 for (this_char = strtok_r(namelist, ",", &save_ptr);
1232 this_char = strtok_r(NULL, ",", &save_ptr)) {
1233 if (parse_one_fs_feature(this_char, flags))
1240 int main(int ac, char **av)
1243 struct btrfs_root *root;
1244 struct btrfs_trans_handle *trans;
1247 u64 block_count = 0;
1248 u64 dev_block_count = 0;
1250 u64 alloc_start = 0;
1251 u64 metadata_profile = 0;
1252 u64 data_profile = 0;
1253 u32 leafsize = max_t(u32, sysconf(_SC_PAGESIZE), DEFAULT_MKFS_LEAF_SIZE);
1254 u32 sectorsize = 4096;
1255 u32 nodesize = leafsize;
1256 u32 stripesize = 4096;
1258 int option_index = 0;
1263 int leaf_forced = 0;
1264 int data_profile_opt = 0;
1265 int metadata_profile_opt = 0;
1268 int force_overwrite = 0;
1270 char *source_dir = NULL;
1271 int source_dir_set = 0;
1272 u64 num_of_meta_chunks = 0;
1273 u64 size_of_data = 0;
1274 u64 source_dir_size = 0;
1278 char *fs_uuid = NULL;
1279 u64 features = DEFAULT_MKFS_FEATURES;
1283 c = getopt_long(ac, av, "A:b:fl:n:s:m:d:L:O:r:U:VMK",
1284 long_options, &option_index);
1289 alloc_start = parse_size(optarg);
1292 force_overwrite = 1;
1295 data_profile = parse_profile(optarg);
1296 data_profile_opt = 1;
1300 nodesize = parse_size(optarg);
1301 leafsize = parse_size(optarg);
1305 label = parse_label(optarg);
1308 metadata_profile = parse_profile(optarg);
1309 metadata_profile_opt = 1;
1315 char *orig = strdup(optarg);
1318 tmp = parse_fs_features(tmp, &features);
1321 "Unrecognized filesystem feature '%s'\n",
1327 if (features & BTRFS_FEATURE_LIST_ALL) {
1328 list_all_fs_features();
1334 sectorsize = parse_size(optarg);
1337 block_count = parse_size(optarg);
1338 if (block_count <= 1024*1024*1024) {
1339 printf("SMALL VOLUME: forcing mixed "
1340 "metadata/data groups\n");
1349 source_dir = optarg;
1362 sectorsize = max(sectorsize, (u32)sysconf(_SC_PAGESIZE));
1363 if (check_leaf_or_node_size(leafsize, sectorsize))
1365 if (check_leaf_or_node_size(nodesize, sectorsize))
1367 saved_optind = optind;
1368 dev_cnt = ac - optind;
1372 if (source_dir_set && dev_cnt > 1) {
1374 "The -r option is limited to a single device\n");
1381 if (uuid_parse(fs_uuid, dummy_uuid) != 0) {
1382 fprintf(stderr, "could not parse UUID: %s\n", fs_uuid);
1385 if (!test_uuid_unique(fs_uuid)) {
1386 fprintf(stderr, "non-unique UUID: %s\n", fs_uuid);
1391 while (dev_cnt-- > 0) {
1392 file = av[optind++];
1393 if (is_block_device(file))
1394 if (test_dev_for_mkfs(file, force_overwrite, estr)) {
1395 fprintf(stderr, "Error: %s", estr);
1400 optind = saved_optind;
1401 dev_cnt = ac - optind;
1403 file = av[optind++];
1406 if (is_vol_small(file)) {
1407 printf("SMALL VOLUME: forcing mixed metadata/data groups\n");
1412 * Set default profiles according to number of added devices.
1413 * For mixed groups defaults are single/single.
1416 if (!metadata_profile_opt) {
1417 if (dev_cnt == 1 && ssd)
1418 printf("Detected a SSD, turning off metadata "
1419 "duplication. Mkfs with -m dup if you want to "
1420 "force metadata duplication.\n");
1422 metadata_profile = (dev_cnt > 1) ?
1423 BTRFS_BLOCK_GROUP_RAID1 : (ssd) ?
1424 0: BTRFS_BLOCK_GROUP_DUP;
1426 if (!data_profile_opt) {
1427 data_profile = (dev_cnt > 1) ?
1428 BTRFS_BLOCK_GROUP_RAID0 : 0; /* raid0 or single */
1431 u32 best_leafsize = max_t(u32, sysconf(_SC_PAGESIZE), sectorsize);
1433 if (metadata_profile_opt || data_profile_opt) {
1434 if (metadata_profile != data_profile) {
1436 "ERROR: With mixed block groups data and metadata profiles must be the same\n");
1442 leafsize = best_leafsize;
1443 nodesize = best_leafsize;
1444 if (check_leaf_or_node_size(leafsize, sectorsize))
1447 if (leafsize != sectorsize) {
1448 fprintf(stderr, "Error: mixed metadata/data block groups "
1449 "require metadata blocksizes equal to the sectorsize\n");
1454 /* Check device/block_count after the leafsize is determined */
1455 if (block_count && block_count < btrfs_min_dev_size(leafsize)) {
1457 "Size '%llu' is too small to make a usable filesystem\n",
1460 "Minimum size for btrfs filesystem is %llu\n",
1461 btrfs_min_dev_size(leafsize));
1464 for (i = saved_optind; i < saved_optind + dev_cnt; i++) {
1468 ret = test_minimum_size(path, leafsize);
1470 fprintf(stderr, "Failed to check size for '%s': %s\n",
1471 path, strerror(-ret));
1476 "'%s' is too small to make a usable filesystem\n",
1479 "Minimum size for each btrfs device is %llu.\n",
1480 btrfs_min_dev_size(leafsize));
1484 ret = test_num_disk_vs_raid(metadata_profile, data_profile,
1485 dev_cnt, mixed, estr);
1487 fprintf(stderr, "Error: %s\n", estr);
1491 /* if we are here that means all devs are good to btrfsify */
1492 printf("\nWARNING! - %s IS EXPERIMENTAL\n", BTRFS_BUILD_VERSION);
1493 printf("WARNING! - see http://btrfs.wiki.kernel.org before using\n\n");
1497 if (!source_dir_set) {
1499 * open without O_EXCL so that the problem should not
1500 * occur by the following processing.
1501 * (btrfs_register_one_device() fails if O_EXCL is on)
1503 fd = open(file, O_RDWR);
1505 fprintf(stderr, "unable to open %s: %s\n", file,
1510 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1511 block_count, &mixed, discard);
1516 if (block_count && block_count > dev_block_count) {
1517 fprintf(stderr, "%s is smaller than requested size\n", file);
1521 fd = open_target(file);
1523 fprintf(stderr, "unable to open the %s\n", file);
1528 source_dir_size = size_sourcedir(source_dir, sectorsize,
1529 &num_of_meta_chunks, &size_of_data);
1530 if(block_count < source_dir_size)
1531 block_count = source_dir_size;
1532 ret = zero_output_file(fd, block_count, sectorsize);
1534 fprintf(stderr, "unable to zero the output file\n");
1537 /* our "device" is the new image file */
1538 dev_block_count = block_count;
1541 /* To create the first block group and chunk 0 in make_btrfs */
1542 if (dev_block_count < BTRFS_MKFS_SYSTEM_GROUP_SIZE) {
1543 fprintf(stderr, "device is too small to make filesystem\n");
1547 blocks[0] = BTRFS_SUPER_INFO_OFFSET;
1548 for (i = 1; i < 7; i++) {
1549 blocks[i] = BTRFS_SUPER_INFO_OFFSET + 1024 * 1024 +
1554 * FS features that can be set by other means than -O
1555 * just set the bit here
1558 features |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
1560 if ((data_profile | metadata_profile) &
1561 (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
1562 features |= BTRFS_FEATURE_INCOMPAT_RAID56;
1565 process_fs_features(features);
1567 ret = make_btrfs(fd, file, label, fs_uuid, blocks, dev_block_count,
1569 sectorsize, stripesize, features);
1571 fprintf(stderr, "error during mkfs: %s\n", strerror(-ret));
1575 root = open_ctree(file, 0, OPEN_CTREE_WRITES);
1577 fprintf(stderr, "Open ctree failed\n");
1581 root->fs_info->alloc_start = alloc_start;
1583 ret = make_root_dir(root, mixed);
1585 fprintf(stderr, "failed to setup the root directory\n");
1589 trans = btrfs_start_transaction(root, 1);
1594 btrfs_register_one_device(file);
1596 while (dev_cnt-- > 0) {
1597 int old_mixed = mixed;
1599 file = av[optind++];
1602 * open without O_EXCL so that the problem should not
1603 * occur by the following processing.
1604 * (btrfs_register_one_device() fails if O_EXCL is on)
1606 fd = open(file, O_RDWR);
1608 fprintf(stderr, "unable to open %s: %s\n", file,
1612 ret = btrfs_device_already_in_root(root, fd,
1613 BTRFS_SUPER_INFO_OFFSET);
1615 fprintf(stderr, "skipping duplicate device %s in FS\n",
1620 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1621 block_count, &mixed, discard);
1628 ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
1629 sectorsize, sectorsize, sectorsize);
1631 btrfs_register_one_device(file);
1635 if (!source_dir_set) {
1636 ret = create_raid_groups(trans, root, data_profile,
1637 data_profile_opt, metadata_profile,
1642 ret = create_data_reloc_tree(trans, root);
1645 printf("fs created label %s on %s\n\tnodesize %u leafsize %u "
1646 "sectorsize %u size %s\n",
1647 label, first_file, nodesize, leafsize, sectorsize,
1648 pretty_size(btrfs_super_total_bytes(root->fs_info->super_copy)));
1650 printf("%s\n", BTRFS_BUILD_VERSION);
1651 btrfs_commit_transaction(trans, root);
1653 if (source_dir_set) {
1654 trans = btrfs_start_transaction(root, 1);
1655 ret = create_chunks(trans, root,
1656 num_of_meta_chunks, size_of_data);
1658 btrfs_commit_transaction(trans, root);
1660 ret = make_image(source_dir, root, fd);
1664 ret = close_ctree(root);