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 #include "kerncompat.h"
21 #include <sys/ioctl.h>
22 #include <sys/mount.h>
26 #include <sys/types.h>
32 #include <uuid/uuid.h>
34 #include <sys/xattr.h>
36 #include <linux/limits.h>
37 #include <blkid/blkid.h>
42 #include "transaction.h"
45 static u64 index_cnt = 2;
47 #define DEFAULT_MKFS_FEATURES (BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF \
48 | BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
50 #define DEFAULT_MKFS_LEAF_SIZE 16384
52 struct directory_name_entry {
56 struct list_head list;
59 static int make_root_dir(struct btrfs_root *root, int mixed)
61 struct btrfs_trans_handle *trans;
62 struct btrfs_key location;
68 trans = btrfs_start_transaction(root, 1);
69 bytes_used = btrfs_super_bytes_used(root->fs_info->super_copy);
71 root->fs_info->system_allocs = 1;
72 ret = btrfs_make_block_group(trans, root, bytes_used,
73 BTRFS_BLOCK_GROUP_SYSTEM,
74 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
75 0, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
79 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
80 &chunk_start, &chunk_size,
81 BTRFS_BLOCK_GROUP_METADATA |
82 BTRFS_BLOCK_GROUP_DATA);
85 "no space to alloc data/metadata chunk\n");
89 ret = btrfs_make_block_group(trans, root, 0,
90 BTRFS_BLOCK_GROUP_METADATA |
91 BTRFS_BLOCK_GROUP_DATA,
92 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
93 chunk_start, chunk_size);
95 printf("Created a data/metadata chunk of size %llu\n", chunk_size);
97 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
98 &chunk_start, &chunk_size,
99 BTRFS_BLOCK_GROUP_METADATA);
100 if (ret == -ENOSPC) {
101 fprintf(stderr, "no space to alloc metadata chunk\n");
105 ret = btrfs_make_block_group(trans, root, 0,
106 BTRFS_BLOCK_GROUP_METADATA,
107 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
108 chunk_start, chunk_size);
112 root->fs_info->system_allocs = 0;
113 btrfs_commit_transaction(trans, root);
114 trans = btrfs_start_transaction(root, 1);
118 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
119 &chunk_start, &chunk_size,
120 BTRFS_BLOCK_GROUP_DATA);
121 if (ret == -ENOSPC) {
122 fprintf(stderr, "no space to alloc data chunk\n");
126 ret = btrfs_make_block_group(trans, root, 0,
127 BTRFS_BLOCK_GROUP_DATA,
128 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
129 chunk_start, chunk_size);
133 ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
134 BTRFS_ROOT_TREE_DIR_OBJECTID);
137 ret = btrfs_make_root_dir(trans, root, BTRFS_FIRST_FREE_OBJECTID);
140 memcpy(&location, &root->fs_info->fs_root->root_key, sizeof(location));
141 location.offset = (u64)-1;
142 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
144 btrfs_super_root_dir(root->fs_info->super_copy),
145 &location, BTRFS_FT_DIR, 0);
149 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
150 "default", 7, location.objectid,
151 BTRFS_ROOT_TREE_DIR_OBJECTID, 0);
155 btrfs_commit_transaction(trans, root);
160 static void __recow_root(struct btrfs_trans_handle *trans,
161 struct btrfs_root *root)
164 struct extent_buffer *tmp;
166 if (trans->transid != btrfs_root_generation(&root->root_item)) {
167 extent_buffer_get(root->node);
168 ret = __btrfs_cow_block(trans, root, root->node,
169 NULL, 0, &tmp, 0, 0);
171 free_extent_buffer(tmp);
175 static void recow_roots(struct btrfs_trans_handle *trans,
176 struct btrfs_root *root)
178 struct btrfs_fs_info *info = root->fs_info;
180 __recow_root(trans, info->fs_root);
181 __recow_root(trans, info->tree_root);
182 __recow_root(trans, info->extent_root);
183 __recow_root(trans, info->chunk_root);
184 __recow_root(trans, info->dev_root);
185 __recow_root(trans, info->csum_root);
188 static int create_one_raid_group(struct btrfs_trans_handle *trans,
189 struct btrfs_root *root, u64 type)
195 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
196 &chunk_start, &chunk_size, type);
197 if (ret == -ENOSPC) {
198 fprintf(stderr, "not enough free space\n");
202 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
203 type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
204 chunk_start, chunk_size);
209 static int create_raid_groups(struct btrfs_trans_handle *trans,
210 struct btrfs_root *root, u64 data_profile,
211 int data_profile_opt, u64 metadata_profile,
214 u64 num_devices = btrfs_super_num_devices(root->fs_info->super_copy);
217 if (metadata_profile) {
218 u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
220 ret = create_one_raid_group(trans, root,
221 BTRFS_BLOCK_GROUP_SYSTEM |
226 meta_flags |= BTRFS_BLOCK_GROUP_DATA;
228 ret = create_one_raid_group(trans, root, meta_flags |
233 if (!mixed && num_devices > 1 && data_profile) {
234 ret = create_one_raid_group(trans, root,
235 BTRFS_BLOCK_GROUP_DATA |
239 recow_roots(trans, root);
244 static int create_data_reloc_tree(struct btrfs_trans_handle *trans,
245 struct btrfs_root *root)
247 struct btrfs_key location;
248 struct btrfs_root_item root_item;
249 struct extent_buffer *tmp;
250 u64 objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
253 ret = btrfs_copy_root(trans, root, root->node, &tmp, objectid);
256 memcpy(&root_item, &root->root_item, sizeof(root_item));
257 btrfs_set_root_bytenr(&root_item, tmp->start);
258 btrfs_set_root_level(&root_item, btrfs_header_level(tmp));
259 btrfs_set_root_generation(&root_item, trans->transid);
260 free_extent_buffer(tmp);
262 location.objectid = objectid;
263 location.type = BTRFS_ROOT_ITEM_KEY;
265 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
266 &location, &root_item);
272 static void print_usage(void) __attribute__((noreturn));
273 static void print_usage(void)
275 fprintf(stderr, "usage: mkfs.btrfs [options] dev [ dev ... ]\n");
276 fprintf(stderr, "options:\n");
277 fprintf(stderr, "\t -A --alloc-start the offset to start the FS\n");
278 fprintf(stderr, "\t -b --byte-count total number of bytes in the FS\n");
279 fprintf(stderr, "\t -d --data data profile, raid0, raid1, raid5, raid6, raid10, dup or single\n");
280 fprintf(stderr, "\t -f --force force overwrite of existing filesystem\n");
281 fprintf(stderr, "\t -l --leafsize size of btree leaves\n");
282 fprintf(stderr, "\t -L --label set a label\n");
283 fprintf(stderr, "\t -m --metadata metadata profile, values like data profile\n");
284 fprintf(stderr, "\t -M --mixed mix metadata and data together\n");
285 fprintf(stderr, "\t -n --nodesize size of btree nodes\n");
286 fprintf(stderr, "\t -s --sectorsize min block allocation (may not mountable by current kernel)\n");
287 fprintf(stderr, "\t -r --rootdir the source directory\n");
288 fprintf(stderr, "\t -K --nodiscard do not perform whole device TRIM\n");
289 fprintf(stderr, "\t -O --features comma separated list of filesystem features\n");
290 fprintf(stderr, "\t -U --uuid specify the filesystem UUID\n");
291 fprintf(stderr, "\t -V --version print the mkfs.btrfs version and exit\n");
292 fprintf(stderr, "%s\n", PACKAGE_STRING);
296 static void print_version(void) __attribute__((noreturn));
297 static void print_version(void)
299 fprintf(stderr, "mkfs.btrfs, part of %s\n", PACKAGE_STRING);
303 static u64 parse_profile(char *s)
305 if (strcmp(s, "raid0") == 0) {
306 return BTRFS_BLOCK_GROUP_RAID0;
307 } else if (strcmp(s, "raid1") == 0) {
308 return BTRFS_BLOCK_GROUP_RAID1;
309 } else if (strcmp(s, "raid5") == 0) {
310 return BTRFS_BLOCK_GROUP_RAID5;
311 } else if (strcmp(s, "raid6") == 0) {
312 return BTRFS_BLOCK_GROUP_RAID6;
313 } else if (strcmp(s, "raid10") == 0) {
314 return BTRFS_BLOCK_GROUP_RAID10;
315 } else if (strcmp(s, "dup") == 0) {
316 return BTRFS_BLOCK_GROUP_DUP;
317 } else if (strcmp(s, "single") == 0) {
320 fprintf(stderr, "Unknown profile %s\n", s);
327 static char *parse_label(char *input)
329 int len = strlen(input);
331 if (len >= BTRFS_LABEL_SIZE) {
332 fprintf(stderr, "Label %s is too long (max %d)\n", input,
333 BTRFS_LABEL_SIZE - 1);
336 return strdup(input);
339 static int add_directory_items(struct btrfs_trans_handle *trans,
340 struct btrfs_root *root, u64 objectid,
341 ino_t parent_inum, const char *name,
342 struct stat *st, int *dir_index_cnt)
346 struct btrfs_key location;
349 name_len = strlen(name);
351 location.objectid = objectid;
353 btrfs_set_key_type(&location, BTRFS_INODE_ITEM_KEY);
355 if (S_ISDIR(st->st_mode))
356 filetype = BTRFS_FT_DIR;
357 if (S_ISREG(st->st_mode))
358 filetype = BTRFS_FT_REG_FILE;
359 if (S_ISLNK(st->st_mode))
360 filetype = BTRFS_FT_SYMLINK;
362 ret = btrfs_insert_dir_item(trans, root, name, name_len,
363 parent_inum, &location,
364 filetype, index_cnt);
367 ret = btrfs_insert_inode_ref(trans, root, name, name_len,
368 objectid, parent_inum, index_cnt);
369 *dir_index_cnt = index_cnt;
375 static int fill_inode_item(struct btrfs_trans_handle *trans,
376 struct btrfs_root *root,
377 struct btrfs_inode_item *dst, struct stat *src)
380 u64 sectorsize = root->sectorsize;
383 * btrfs_inode_item has some reserved fields
384 * and represents on-disk inode entry, so
385 * zero everything to prevent information leak
387 memset(dst, 0, sizeof (*dst));
389 btrfs_set_stack_inode_generation(dst, trans->transid);
390 btrfs_set_stack_inode_size(dst, src->st_size);
391 btrfs_set_stack_inode_nbytes(dst, 0);
392 btrfs_set_stack_inode_block_group(dst, 0);
393 btrfs_set_stack_inode_nlink(dst, src->st_nlink);
394 btrfs_set_stack_inode_uid(dst, src->st_uid);
395 btrfs_set_stack_inode_gid(dst, src->st_gid);
396 btrfs_set_stack_inode_mode(dst, src->st_mode);
397 btrfs_set_stack_inode_rdev(dst, 0);
398 btrfs_set_stack_inode_flags(dst, 0);
399 btrfs_set_stack_timespec_sec(&dst->atime, src->st_atime);
400 btrfs_set_stack_timespec_nsec(&dst->atime, 0);
401 btrfs_set_stack_timespec_sec(&dst->ctime, src->st_ctime);
402 btrfs_set_stack_timespec_nsec(&dst->ctime, 0);
403 btrfs_set_stack_timespec_sec(&dst->mtime, src->st_mtime);
404 btrfs_set_stack_timespec_nsec(&dst->mtime, 0);
405 btrfs_set_stack_timespec_sec(&dst->otime, 0);
406 btrfs_set_stack_timespec_nsec(&dst->otime, 0);
408 if (S_ISDIR(src->st_mode)) {
409 btrfs_set_stack_inode_size(dst, 0);
410 btrfs_set_stack_inode_nlink(dst, 1);
412 if (S_ISREG(src->st_mode)) {
413 btrfs_set_stack_inode_size(dst, (u64)src->st_size);
414 if (src->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root))
415 btrfs_set_stack_inode_nbytes(dst, src->st_size);
417 blocks = src->st_size / sectorsize;
418 if (src->st_size % sectorsize)
420 blocks *= sectorsize;
421 btrfs_set_stack_inode_nbytes(dst, blocks);
424 if (S_ISLNK(src->st_mode))
425 btrfs_set_stack_inode_nbytes(dst, src->st_size + 1);
430 static int directory_select(const struct direct *entry)
432 if ((strncmp(entry->d_name, ".", entry->d_reclen) == 0) ||
433 (strncmp(entry->d_name, "..", entry->d_reclen) == 0))
439 static void free_namelist(struct direct **files, int count)
446 for (i = 0; i < count; ++i)
451 static u64 calculate_dir_inode_size(char *dirname)
454 struct direct **files, *cur_file;
455 u64 dir_inode_size = 0;
457 count = scandir(dirname, &files, directory_select, NULL);
459 for (i = 0; i < count; i++) {
461 dir_inode_size += strlen(cur_file->d_name);
464 free_namelist(files, count);
467 return dir_inode_size;
470 static int add_inode_items(struct btrfs_trans_handle *trans,
471 struct btrfs_root *root,
472 struct stat *st, char *name,
473 u64 self_objectid, ino_t parent_inum,
474 int dir_index_cnt, struct btrfs_inode_item *inode_ret)
477 struct btrfs_key inode_key;
478 struct btrfs_inode_item btrfs_inode;
482 fill_inode_item(trans, root, &btrfs_inode, st);
483 objectid = self_objectid;
485 if (S_ISDIR(st->st_mode)) {
486 inode_size = calculate_dir_inode_size(name);
487 btrfs_set_stack_inode_size(&btrfs_inode, inode_size);
490 inode_key.objectid = objectid;
491 inode_key.offset = 0;
492 btrfs_set_key_type(&inode_key, BTRFS_INODE_ITEM_KEY);
494 ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
496 *inode_ret = btrfs_inode;
500 static int add_xattr_item(struct btrfs_trans_handle *trans,
501 struct btrfs_root *root, u64 objectid,
502 const char *file_name)
506 char xattr_list[XATTR_LIST_MAX];
508 char cur_value[XATTR_SIZE_MAX];
509 char delimiter = '\0';
510 char *next_location = xattr_list;
512 ret = llistxattr(file_name, xattr_list, XATTR_LIST_MAX);
516 fprintf(stderr, "get a list of xattr failed for %s\n",
523 cur_name = strtok(xattr_list, &delimiter);
524 while (cur_name != NULL) {
525 cur_name_len = strlen(cur_name);
526 next_location += cur_name_len + 1;
528 ret = getxattr(file_name, cur_name, cur_value, XATTR_SIZE_MAX);
532 fprintf(stderr, "get a xattr value failed for %s attr %s\n",
533 file_name, cur_name);
537 ret = btrfs_insert_xattr_item(trans, root, cur_name,
538 cur_name_len, cur_value,
541 fprintf(stderr, "insert a xattr item failed for %s\n",
545 cur_name = strtok(next_location, &delimiter);
551 static int add_symbolic_link(struct btrfs_trans_handle *trans,
552 struct btrfs_root *root,
553 u64 objectid, const char *path_name)
556 u64 sectorsize = root->sectorsize;
557 char *buf = malloc(sectorsize);
559 ret = readlink(path_name, buf, sectorsize);
561 fprintf(stderr, "readlink failed for %s\n", path_name);
564 if (ret >= sectorsize) {
565 fprintf(stderr, "symlink too long for %s", path_name);
570 buf[ret] = '\0'; /* readlink does not do it for us */
571 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
578 static int add_file_items(struct btrfs_trans_handle *trans,
579 struct btrfs_root *root,
580 struct btrfs_inode_item *btrfs_inode, u64 objectid,
581 ino_t parent_inum, struct stat *st,
582 const char *path_name, int out_fd)
587 struct btrfs_key key;
589 u32 sectorsize = root->sectorsize;
594 struct extent_buffer *eb = NULL;
597 if (st->st_size == 0)
600 fd = open(path_name, O_RDONLY);
602 fprintf(stderr, "%s open failed\n", path_name);
606 blocks = st->st_size / sectorsize;
607 if (st->st_size % sectorsize)
610 if (st->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
611 char *buffer = malloc(st->st_size);
612 ret_read = pread64(fd, buffer, st->st_size, bytes_read);
613 if (ret_read == -1) {
614 fprintf(stderr, "%s read failed\n", path_name);
619 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
620 buffer, st->st_size);
625 /* round up our st_size to the FS blocksize */
626 total_bytes = (u64)blocks * sectorsize;
629 * do our IO in extent buffers so it can work
630 * against any raid type
632 eb = malloc(sizeof(*eb) + sectorsize);
637 memset(eb, 0, sizeof(*eb) + sectorsize);
642 * keep our extent size at 1MB max, this makes it easier to work inside
643 * the tiny block groups created during mkfs
645 cur_bytes = min(total_bytes, 1024ULL * 1024);
646 ret = btrfs_reserve_extent(trans, root, cur_bytes, 0, 0, (u64)-1,
651 first_block = key.objectid;
654 while (bytes_read < cur_bytes) {
656 memset(eb->data, 0, sectorsize);
658 ret_read = pread64(fd, eb->data, sectorsize, file_pos + bytes_read);
659 if (ret_read == -1) {
660 fprintf(stderr, "%s read failed\n", path_name);
664 eb->start = first_block + bytes_read;
665 eb->len = sectorsize;
668 * we're doing the csum before we record the extent, but
671 ret = btrfs_csum_file_block(trans, root->fs_info->csum_root,
672 first_block + bytes_read + sectorsize,
673 first_block + bytes_read,
674 eb->data, sectorsize);
678 ret = write_and_map_eb(trans, root, eb);
680 fprintf(stderr, "output file write failed\n");
684 bytes_read += sectorsize;
688 ret = btrfs_record_file_extent(trans, root, objectid, btrfs_inode,
689 file_pos, first_block, cur_bytes);
695 file_pos += cur_bytes;
696 total_bytes -= cur_bytes;
707 static char *make_path(char *dir, char *name)
711 path = malloc(strlen(dir) + strlen(name) + 2);
715 if (dir[strlen(dir) - 1] != '/')
721 static int traverse_directory(struct btrfs_trans_handle *trans,
722 struct btrfs_root *root, char *dir_name,
723 struct directory_name_entry *dir_head, int out_fd)
727 struct btrfs_inode_item cur_inode;
728 struct btrfs_inode_item *inode_item;
729 int count, i, dir_index_cnt;
730 struct direct **files;
732 struct directory_name_entry *dir_entry, *parent_dir_entry;
733 struct direct *cur_file;
734 ino_t parent_inum, cur_inum;
735 ino_t highest_inum = 0;
736 char *parent_dir_name;
737 char real_path[PATH_MAX];
738 struct btrfs_path path;
739 struct extent_buffer *leaf;
740 struct btrfs_key root_dir_key;
741 u64 root_dir_inode_size = 0;
743 /* Add list for source directory */
744 dir_entry = malloc(sizeof(struct directory_name_entry));
745 dir_entry->dir_name = dir_name;
746 dir_entry->path = realpath(dir_name, real_path);
747 if (!dir_entry->path) {
748 fprintf(stderr, "get directory real path error\n");
753 parent_inum = highest_inum + BTRFS_FIRST_FREE_OBJECTID;
754 dir_entry->inum = parent_inum;
755 list_add_tail(&dir_entry->list, &dir_head->list);
757 btrfs_init_path(&path);
759 root_dir_key.objectid = btrfs_root_dirid(&root->root_item);
760 root_dir_key.offset = 0;
761 btrfs_set_key_type(&root_dir_key, BTRFS_INODE_ITEM_KEY);
762 ret = btrfs_lookup_inode(trans, root, &path, &root_dir_key, 1);
764 fprintf(stderr, "root dir lookup error\n");
768 leaf = path.nodes[0];
769 inode_item = btrfs_item_ptr(leaf, path.slots[0],
770 struct btrfs_inode_item);
772 root_dir_inode_size = calculate_dir_inode_size(dir_name);
773 btrfs_set_inode_size(leaf, inode_item, root_dir_inode_size);
774 btrfs_mark_buffer_dirty(leaf);
776 btrfs_release_path(&path);
779 parent_dir_entry = list_entry(dir_head->list.next,
780 struct directory_name_entry,
782 list_del(&parent_dir_entry->list);
784 parent_inum = parent_dir_entry->inum;
785 parent_dir_name = parent_dir_entry->dir_name;
786 if (chdir(parent_dir_entry->path)) {
787 fprintf(stderr, "chdir error for %s\n",
793 count = scandir(parent_dir_entry->path, &files,
794 directory_select, NULL);
797 fprintf(stderr, "scandir for %s failed: %s\n",
798 parent_dir_name, strerror (errno));
803 for (i = 0; i < count; i++) {
806 if (lstat(cur_file->d_name, &st) == -1) {
807 fprintf(stderr, "lstat failed for file %s\n",
813 cur_inum = st.st_ino;
814 ret = add_directory_items(trans, root,
815 cur_inum, parent_inum,
817 &st, &dir_index_cnt);
819 fprintf(stderr, "add_directory_items failed\n");
823 ret = add_inode_items(trans, root, &st,
824 cur_file->d_name, cur_inum,
825 parent_inum, dir_index_cnt,
827 if (ret == -EEXIST) {
828 BUG_ON(st.st_nlink <= 1);
832 fprintf(stderr, "add_inode_items failed\n");
836 ret = add_xattr_item(trans, root,
837 cur_inum, cur_file->d_name);
839 fprintf(stderr, "add_xattr_item failed\n");
844 if (S_ISDIR(st.st_mode)) {
845 dir_entry = malloc(sizeof(struct directory_name_entry));
846 dir_entry->dir_name = cur_file->d_name;
847 dir_entry->path = make_path(parent_dir_entry->path,
849 dir_entry->inum = cur_inum;
850 list_add_tail(&dir_entry->list, &dir_head->list);
851 } else if (S_ISREG(st.st_mode)) {
852 ret = add_file_items(trans, root, &cur_inode,
853 cur_inum, parent_inum, &st,
854 cur_file->d_name, out_fd);
856 fprintf(stderr, "add_file_items failed\n");
859 } else if (S_ISLNK(st.st_mode)) {
860 ret = add_symbolic_link(trans, root,
861 cur_inum, cur_file->d_name);
863 fprintf(stderr, "add_symbolic_link failed\n");
869 free_namelist(files, count);
870 free(parent_dir_entry);
874 } while (!list_empty(&dir_head->list));
879 free_namelist(files, count);
881 free(parent_dir_entry);
888 static int open_target(char *output_name)
891 output_fd = open(output_name, O_CREAT | O_RDWR | O_TRUNC,
892 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH);
897 static int create_chunks(struct btrfs_trans_handle *trans,
898 struct btrfs_root *root, u64 num_of_meta_chunks,
903 u64 meta_type = BTRFS_BLOCK_GROUP_METADATA;
904 u64 data_type = BTRFS_BLOCK_GROUP_DATA;
905 u64 minimum_data_chunk_size = 8 * 1024 * 1024;
909 for (i = 0; i < num_of_meta_chunks; i++) {
910 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
911 &chunk_start, &chunk_size, meta_type);
913 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
914 meta_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
915 chunk_start, chunk_size);
917 set_extent_dirty(&root->fs_info->free_space_cache,
918 chunk_start, chunk_start + chunk_size - 1, 0);
921 if (size_of_data < minimum_data_chunk_size)
922 size_of_data = minimum_data_chunk_size;
924 ret = btrfs_alloc_data_chunk(trans, root->fs_info->extent_root,
925 &chunk_start, size_of_data, data_type);
927 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
928 data_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
929 chunk_start, size_of_data);
931 set_extent_dirty(&root->fs_info->free_space_cache,
932 chunk_start, chunk_start + size_of_data - 1, 0);
936 static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
939 struct btrfs_trans_handle *trans;
943 struct directory_name_entry dir_head;
945 struct directory_name_entry *dir_entry = NULL;
947 ret = lstat(source_dir, &root_st);
949 fprintf(stderr, "unable to lstat the %s\n", source_dir);
953 INIT_LIST_HEAD(&dir_head.list);
955 trans = btrfs_start_transaction(root, 1);
956 ret = traverse_directory(trans, root, source_dir, &dir_head, out_fd);
958 fprintf(stderr, "unable to traverse_directory\n");
961 btrfs_commit_transaction(trans, root);
963 printf("Making image is completed.\n");
966 while (!list_empty(&dir_head.list)) {
967 dir_entry = list_entry(dir_head.list.next,
968 struct directory_name_entry, list);
969 list_del(&dir_entry->list);
973 fprintf(stderr, "Making image is aborted.\n");
978 * This ignores symlinks with unreadable targets and subdirs that can't
979 * be read. It's a best-effort to give a rough estimate of the size of
980 * a subdir. It doesn't guarantee that prepopulating btrfs from this
981 * tree won't still run out of space.
983 * The rounding up to 4096 is questionable. Previous code used du -B 4096.
985 static u64 global_total_size;
986 static int ftw_add_entry_size(const char *fpath, const struct stat *st,
989 if (type == FTW_F || type == FTW_D)
990 global_total_size += round_up(st->st_size, 4096);
995 static u64 size_sourcedir(char *dir_name, u64 sectorsize,
996 u64 *num_of_meta_chunks_ret, u64 *size_of_data_ret)
1001 u64 default_chunk_size = 8 * 1024 * 1024; /* 8MB */
1002 u64 allocated_meta_size = 8 * 1024 * 1024; /* 8MB */
1003 u64 allocated_total_size = 20 * 1024 * 1024; /* 20MB */
1004 u64 num_of_meta_chunks = 0;
1005 u64 num_of_data_chunks = 0;
1006 u64 num_of_allocated_meta_chunks =
1007 allocated_meta_size / default_chunk_size;
1009 global_total_size = 0;
1010 ret = ftw(dir_name, ftw_add_entry_size, 10);
1011 dir_size = global_total_size;
1013 fprintf(stderr, "ftw subdir walk of '%s' failed: %s\n",
1014 dir_name, strerror(errno));
1018 num_of_data_chunks = (dir_size + default_chunk_size - 1) /
1021 num_of_meta_chunks = (dir_size / 2) / default_chunk_size;
1022 if (((dir_size / 2) % default_chunk_size) != 0)
1023 num_of_meta_chunks++;
1024 if (num_of_meta_chunks <= num_of_allocated_meta_chunks)
1025 num_of_meta_chunks = 0;
1027 num_of_meta_chunks -= num_of_allocated_meta_chunks;
1029 total_size = allocated_total_size +
1030 (num_of_data_chunks * default_chunk_size) +
1031 (num_of_meta_chunks * default_chunk_size);
1033 *num_of_meta_chunks_ret = num_of_meta_chunks;
1034 *size_of_data_ret = num_of_data_chunks * default_chunk_size;
1038 static int zero_output_file(int out_fd, u64 size, u32 sectorsize)
1040 int len = sectorsize;
1041 int loop_num = size / sectorsize;
1043 char *buf = malloc(len);
1049 memset(buf, 0, len);
1050 for (i = 0; i < loop_num; i++) {
1051 written = pwrite64(out_fd, buf, len, location);
1054 location += sectorsize;
1060 static int check_leaf_or_node_size(u32 size, u32 sectorsize)
1062 if (size < sectorsize) {
1064 "Illegal leafsize (or nodesize) %u (smaller than %u)\n",
1067 } else if (size > BTRFS_MAX_METADATA_BLOCKSIZE) {
1069 "Illegal leafsize (or nodesize) %u (larger than %u)\n",
1070 size, BTRFS_MAX_METADATA_BLOCKSIZE);
1072 } else if (size & (sectorsize - 1)) {
1074 "Illegal leafsize (or nodesize) %u (not align to %u)\n",
1081 static int is_ssd(const char *file)
1085 char sysfs_path[PATH_MAX];
1091 probe = blkid_new_probe_from_filename(file);
1095 /* Device number of this disk (possibly a partition) */
1096 devno = blkid_probe_get_devno(probe);
1098 blkid_free_probe(probe);
1102 /* Get whole disk name (not full path) for this devno */
1103 ret = blkid_devno_to_wholedisk(devno,
1104 wholedisk, sizeof(wholedisk), NULL);
1106 blkid_free_probe(probe);
1110 snprintf(sysfs_path, PATH_MAX, "/sys/block/%s/queue/rotational",
1113 blkid_free_probe(probe);
1115 fd = open(sysfs_path, O_RDONLY);
1120 if (read(fd, &rotational, sizeof(char)) < sizeof(char)) {
1126 return !atoi((const char *)&rotational);
1129 #define BTRFS_FEATURE_LIST_ALL (1ULL << 63)
1131 static const struct btrfs_fs_feature {
1135 } mkfs_features[] = {
1136 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1137 "mixed data and metadata block groups" },
1138 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1139 "increased hardlink limit per file to 65536" },
1140 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1141 "raid56 extended format" },
1142 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1143 "reduced-size metadata extent refs" },
1144 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1145 "no explicit hole extents for files" },
1146 /* Keep this one last */
1147 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1150 static void list_all_fs_features(void)
1154 fprintf(stderr, "Filesystem features available at mkfs time:\n");
1155 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1156 char *is_default = "";
1158 if (mkfs_features[i].flag & DEFAULT_MKFS_FEATURES)
1159 is_default = ", default";
1160 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
1161 mkfs_features[i].name,
1162 mkfs_features[i].desc,
1163 mkfs_features[i].flag,
1168 static int parse_one_fs_feature(const char *name, u64 *flags)
1173 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1174 if (name[0] == '^' &&
1175 !strcmp(mkfs_features[i].name, name + 1)) {
1176 *flags &= ~ mkfs_features[i].flag;
1178 } else if (!strcmp(mkfs_features[i].name, name)) {
1179 *flags |= mkfs_features[i].flag;
1187 static void process_fs_features(u64 flags)
1191 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1192 if (flags & mkfs_features[i].flag) {
1193 printf("Turning ON incompat feature '%s': %s\n",
1194 mkfs_features[i].name,
1195 mkfs_features[i].desc);
1202 * Return NULL if all features were parsed fine, otherwise return the name of
1203 * the first unparsed.
1205 static char* parse_fs_features(char *namelist, u64 *flags)
1208 char *save_ptr = NULL; /* Satisfy static checkers */
1210 for (this_char = strtok_r(namelist, ",", &save_ptr);
1212 this_char = strtok_r(NULL, ",", &save_ptr)) {
1213 if (parse_one_fs_feature(this_char, flags))
1220 int main(int ac, char **av)
1223 struct btrfs_root *root;
1224 struct btrfs_trans_handle *trans;
1227 u64 block_count = 0;
1228 u64 dev_block_count = 0;
1230 u64 alloc_start = 0;
1231 u64 metadata_profile = 0;
1232 u64 data_profile = 0;
1233 u32 leafsize = max_t(u32, sysconf(_SC_PAGESIZE), DEFAULT_MKFS_LEAF_SIZE);
1234 u32 sectorsize = 4096;
1235 u32 nodesize = leafsize;
1236 u32 stripesize = 4096;
1242 int leaf_forced = 0;
1243 int data_profile_opt = 0;
1244 int metadata_profile_opt = 0;
1247 int force_overwrite = 0;
1249 char *source_dir = NULL;
1250 int source_dir_set = 0;
1251 u64 num_of_meta_chunks = 0;
1252 u64 size_of_data = 0;
1253 u64 source_dir_size = 0;
1257 char *fs_uuid = NULL;
1258 u64 features = DEFAULT_MKFS_FEATURES;
1262 int option_index = 0;
1263 static const struct option long_options[] = {
1264 { "alloc-start", 1, NULL, 'A'},
1265 { "byte-count", 1, NULL, 'b' },
1266 { "force", 0, NULL, 'f' },
1267 { "leafsize", 1, NULL, 'l' },
1268 { "label", 1, NULL, 'L'},
1269 { "metadata", 1, NULL, 'm' },
1270 { "mixed", 0, NULL, 'M' },
1271 { "nodesize", 1, NULL, 'n' },
1272 { "sectorsize", 1, NULL, 's' },
1273 { "data", 1, NULL, 'd' },
1274 { "version", 0, NULL, 'V' },
1275 { "rootdir", 1, NULL, 'r' },
1276 { "nodiscard", 0, NULL, 'K' },
1277 { "features", 1, NULL, 'O' },
1278 { "uuid", required_argument, NULL, 'U' },
1282 c = getopt_long(ac, av, "A:b:fl:n:s:m:d:L:O:r:U:VMK",
1283 long_options, &option_index);
1288 alloc_start = parse_size(optarg);
1291 force_overwrite = 1;
1294 data_profile = parse_profile(optarg);
1295 data_profile_opt = 1;
1299 nodesize = parse_size(optarg);
1300 leafsize = parse_size(optarg);
1304 label = parse_label(optarg);
1307 metadata_profile = parse_profile(optarg);
1308 metadata_profile_opt = 1;
1314 char *orig = strdup(optarg);
1317 tmp = parse_fs_features(tmp, &features);
1320 "Unrecognized filesystem feature '%s'\n",
1326 if (features & BTRFS_FEATURE_LIST_ALL) {
1327 list_all_fs_features();
1333 sectorsize = parse_size(optarg);
1336 block_count = parse_size(optarg);
1337 if (block_count <= BTRFS_MKFS_SMALL_VOLUME_SIZE) {
1339 "SMALL VOLUME: forcing mixed metadata/data groups\n");
1348 source_dir = optarg;
1361 sectorsize = max(sectorsize, (u32)sysconf(_SC_PAGESIZE));
1362 if (check_leaf_or_node_size(leafsize, sectorsize))
1364 if (check_leaf_or_node_size(nodesize, sectorsize))
1366 saved_optind = optind;
1367 dev_cnt = ac - optind;
1371 if (source_dir_set && dev_cnt > 1) {
1373 "The -r option is limited to a single device\n");
1380 if (uuid_parse(fs_uuid, dummy_uuid) != 0) {
1381 fprintf(stderr, "could not parse UUID: %s\n", fs_uuid);
1384 if (!test_uuid_unique(fs_uuid)) {
1385 fprintf(stderr, "non-unique UUID: %s\n", fs_uuid);
1390 while (dev_cnt-- > 0) {
1391 file = av[optind++];
1392 if (is_block_device(file))
1393 if (test_dev_for_mkfs(file, force_overwrite, estr)) {
1394 fprintf(stderr, "Error: %s", estr);
1399 optind = saved_optind;
1400 dev_cnt = ac - optind;
1402 file = av[optind++];
1405 if (is_vol_small(file)) {
1406 printf("SMALL VOLUME: forcing mixed metadata/data groups\n");
1411 * Set default profiles according to number of added devices.
1412 * For mixed groups defaults are single/single.
1415 if (!metadata_profile_opt) {
1416 if (dev_cnt == 1 && ssd)
1417 printf("Detected a SSD, turning off metadata "
1418 "duplication. Mkfs with -m dup if you want to "
1419 "force metadata duplication.\n");
1421 metadata_profile = (dev_cnt > 1) ?
1422 BTRFS_BLOCK_GROUP_RAID1 : (ssd) ?
1423 0: BTRFS_BLOCK_GROUP_DUP;
1425 if (!data_profile_opt) {
1426 data_profile = (dev_cnt > 1) ?
1427 BTRFS_BLOCK_GROUP_RAID0 : 0; /* raid0 or single */
1430 u32 best_leafsize = max_t(u32, sysconf(_SC_PAGESIZE), sectorsize);
1432 if (metadata_profile_opt || data_profile_opt) {
1433 if (metadata_profile != data_profile) {
1435 "ERROR: With mixed block groups data and metadata profiles must be the same\n");
1441 leafsize = best_leafsize;
1442 nodesize = best_leafsize;
1443 if (check_leaf_or_node_size(leafsize, sectorsize))
1446 if (leafsize != sectorsize) {
1447 fprintf(stderr, "Error: mixed metadata/data block groups "
1448 "require metadata blocksizes equal to the sectorsize\n");
1453 /* Check device/block_count after the leafsize is determined */
1454 if (block_count && block_count < btrfs_min_dev_size(leafsize)) {
1456 "Size '%llu' is too small to make a usable filesystem\n",
1459 "Minimum size for btrfs filesystem is %llu\n",
1460 btrfs_min_dev_size(leafsize));
1463 for (i = saved_optind; i < saved_optind + dev_cnt; i++) {
1467 ret = test_minimum_size(path, leafsize);
1469 fprintf(stderr, "Failed to check size for '%s': %s\n",
1470 path, strerror(-ret));
1475 "'%s' is too small to make a usable filesystem\n",
1478 "Minimum size for each btrfs device is %llu.\n",
1479 btrfs_min_dev_size(leafsize));
1483 ret = test_num_disk_vs_raid(metadata_profile, data_profile,
1484 dev_cnt, mixed, estr);
1486 fprintf(stderr, "Error: %s\n", estr);
1490 /* if we are here that means all devs are good to btrfsify */
1491 printf("%s\n", PACKAGE_STRING);
1492 printf("See %s for more information.\n\n", PACKAGE_URL);
1496 if (!source_dir_set) {
1498 * open without O_EXCL so that the problem should not
1499 * occur by the following processing.
1500 * (btrfs_register_one_device() fails if O_EXCL is on)
1502 fd = open(file, O_RDWR);
1504 fprintf(stderr, "unable to open %s: %s\n", file,
1509 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1510 block_count, &mixed, discard);
1515 if (block_count && block_count > dev_block_count) {
1516 fprintf(stderr, "%s is smaller than requested size\n", file);
1520 fd = open_target(file);
1522 fprintf(stderr, "unable to open the %s\n", file);
1527 source_dir_size = size_sourcedir(source_dir, sectorsize,
1528 &num_of_meta_chunks, &size_of_data);
1529 if(block_count < source_dir_size)
1530 block_count = source_dir_size;
1531 ret = zero_output_file(fd, block_count, sectorsize);
1533 fprintf(stderr, "unable to zero the output file\n");
1536 /* our "device" is the new image file */
1537 dev_block_count = block_count;
1540 /* To create the first block group and chunk 0 in make_btrfs */
1541 if (dev_block_count < BTRFS_MKFS_SYSTEM_GROUP_SIZE) {
1542 fprintf(stderr, "device is too small to make filesystem\n");
1546 blocks[0] = BTRFS_SUPER_INFO_OFFSET;
1547 for (i = 1; i < 7; i++) {
1548 blocks[i] = BTRFS_SUPER_INFO_OFFSET + 1024 * 1024 +
1553 * FS features that can be set by other means than -O
1554 * just set the bit here
1557 features |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
1559 if ((data_profile | metadata_profile) &
1560 (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
1561 features |= BTRFS_FEATURE_INCOMPAT_RAID56;
1564 process_fs_features(features);
1566 ret = make_btrfs(fd, file, label, fs_uuid, blocks, dev_block_count,
1568 sectorsize, stripesize, features);
1570 fprintf(stderr, "error during mkfs: %s\n", strerror(-ret));
1574 root = open_ctree(file, 0, OPEN_CTREE_WRITES);
1576 fprintf(stderr, "Open ctree failed\n");
1580 root->fs_info->alloc_start = alloc_start;
1582 ret = make_root_dir(root, mixed);
1584 fprintf(stderr, "failed to setup the root directory\n");
1588 trans = btrfs_start_transaction(root, 1);
1590 btrfs_register_one_device(file);
1595 while (dev_cnt-- > 0) {
1596 int old_mixed = mixed;
1598 file = av[optind++];
1601 * open without O_EXCL so that the problem should not
1602 * occur by the following processing.
1603 * (btrfs_register_one_device() fails if O_EXCL is on)
1605 fd = open(file, O_RDWR);
1607 fprintf(stderr, "unable to open %s: %s\n", file,
1611 ret = btrfs_device_already_in_root(root, fd,
1612 BTRFS_SUPER_INFO_OFFSET);
1614 fprintf(stderr, "skipping duplicate device %s in FS\n",
1619 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1620 block_count, &mixed, discard);
1627 ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
1628 sectorsize, sectorsize, sectorsize);
1630 btrfs_register_one_device(file);
1634 if (!source_dir_set) {
1635 ret = create_raid_groups(trans, root, data_profile,
1636 data_profile_opt, metadata_profile,
1641 ret = create_data_reloc_tree(trans, root);
1644 printf("fs created label %s on %s\n\tnodesize %u leafsize %u "
1645 "sectorsize %u size %s\n",
1646 label, first_file, nodesize, leafsize, sectorsize,
1647 pretty_size(btrfs_super_total_bytes(root->fs_info->super_copy)));
1649 btrfs_commit_transaction(trans, root);
1651 if (source_dir_set) {
1652 trans = btrfs_start_transaction(root, 1);
1653 ret = create_chunks(trans, root,
1654 num_of_meta_chunks, size_of_data);
1656 btrfs_commit_transaction(trans, root);
1658 ret = make_image(source_dir, root, fd);
1662 ret = close_ctree(root);