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;
46 static int verbose = 1;
48 struct directory_name_entry {
52 struct list_head list;
55 struct mkfs_allocation {
62 static int make_root_dir(struct btrfs_root *root, int mixed,
63 struct mkfs_allocation *allocation)
65 struct btrfs_trans_handle *trans;
66 struct btrfs_key location;
72 trans = btrfs_start_transaction(root, 1);
73 bytes_used = btrfs_super_bytes_used(root->fs_info->super_copy);
75 root->fs_info->system_allocs = 1;
76 ret = btrfs_make_block_group(trans, root, bytes_used,
77 BTRFS_BLOCK_GROUP_SYSTEM,
78 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
79 0, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
80 allocation->system += BTRFS_MKFS_SYSTEM_GROUP_SIZE;
84 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
85 &chunk_start, &chunk_size,
86 BTRFS_BLOCK_GROUP_METADATA |
87 BTRFS_BLOCK_GROUP_DATA);
90 "no space to alloc data/metadata chunk\n");
94 ret = btrfs_make_block_group(trans, root, 0,
95 BTRFS_BLOCK_GROUP_METADATA |
96 BTRFS_BLOCK_GROUP_DATA,
97 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
98 chunk_start, chunk_size);
100 allocation->mixed += chunk_size;
102 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
103 &chunk_start, &chunk_size,
104 BTRFS_BLOCK_GROUP_METADATA);
105 if (ret == -ENOSPC) {
106 fprintf(stderr, "no space to alloc metadata chunk\n");
110 ret = btrfs_make_block_group(trans, root, 0,
111 BTRFS_BLOCK_GROUP_METADATA,
112 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
113 chunk_start, chunk_size);
114 allocation->metadata += chunk_size;
118 root->fs_info->system_allocs = 0;
119 btrfs_commit_transaction(trans, root);
120 trans = btrfs_start_transaction(root, 1);
124 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
125 &chunk_start, &chunk_size,
126 BTRFS_BLOCK_GROUP_DATA);
127 if (ret == -ENOSPC) {
128 fprintf(stderr, "no space to alloc data chunk\n");
132 ret = btrfs_make_block_group(trans, root, 0,
133 BTRFS_BLOCK_GROUP_DATA,
134 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
135 chunk_start, chunk_size);
136 allocation->data += chunk_size;
140 ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
141 BTRFS_ROOT_TREE_DIR_OBJECTID);
144 ret = btrfs_make_root_dir(trans, root, BTRFS_FIRST_FREE_OBJECTID);
147 memcpy(&location, &root->fs_info->fs_root->root_key, sizeof(location));
148 location.offset = (u64)-1;
149 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
151 btrfs_super_root_dir(root->fs_info->super_copy),
152 &location, BTRFS_FT_DIR, 0);
156 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
157 "default", 7, location.objectid,
158 BTRFS_ROOT_TREE_DIR_OBJECTID, 0);
162 btrfs_commit_transaction(trans, root);
167 static void __recow_root(struct btrfs_trans_handle *trans,
168 struct btrfs_root *root)
171 struct extent_buffer *tmp;
173 if (trans->transid != btrfs_root_generation(&root->root_item)) {
174 extent_buffer_get(root->node);
175 ret = __btrfs_cow_block(trans, root, root->node,
176 NULL, 0, &tmp, 0, 0);
178 free_extent_buffer(tmp);
182 static void recow_roots(struct btrfs_trans_handle *trans,
183 struct btrfs_root *root)
185 struct btrfs_fs_info *info = root->fs_info;
187 __recow_root(trans, info->fs_root);
188 __recow_root(trans, info->tree_root);
189 __recow_root(trans, info->extent_root);
190 __recow_root(trans, info->chunk_root);
191 __recow_root(trans, info->dev_root);
192 __recow_root(trans, info->csum_root);
195 static int create_one_raid_group(struct btrfs_trans_handle *trans,
196 struct btrfs_root *root, u64 type,
197 struct mkfs_allocation *allocation)
204 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
205 &chunk_start, &chunk_size, type);
206 if (ret == -ENOSPC) {
207 fprintf(stderr, "not enough free space\n");
211 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
212 type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
213 chunk_start, chunk_size);
214 if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == BTRFS_BLOCK_GROUP_DATA)
215 allocation->data += chunk_size;
216 else if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == BTRFS_BLOCK_GROUP_METADATA)
217 allocation->metadata += chunk_size;
218 else if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == BTRFS_BLOCK_GROUP_SYSTEM)
219 allocation->system += chunk_size;
220 else if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) ==
221 (BTRFS_BLOCK_GROUP_METADATA|BTRFS_BLOCK_GROUP_DATA))
222 allocation->mixed += chunk_size;
230 static int create_raid_groups(struct btrfs_trans_handle *trans,
231 struct btrfs_root *root, u64 data_profile,
232 int data_profile_opt, u64 metadata_profile,
234 struct mkfs_allocation *allocation)
236 u64 num_devices = btrfs_super_num_devices(root->fs_info->super_copy);
239 if (metadata_profile) {
240 u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
242 ret = create_one_raid_group(trans, root,
243 BTRFS_BLOCK_GROUP_SYSTEM |
244 metadata_profile, allocation);
248 meta_flags |= BTRFS_BLOCK_GROUP_DATA;
250 ret = create_one_raid_group(trans, root, meta_flags |
251 metadata_profile, allocation);
255 if (!mixed && num_devices > 1 && data_profile) {
256 ret = create_one_raid_group(trans, root,
257 BTRFS_BLOCK_GROUP_DATA |
258 data_profile, allocation);
261 recow_roots(trans, root);
266 static int create_data_reloc_tree(struct btrfs_trans_handle *trans,
267 struct btrfs_root *root)
269 struct btrfs_key location;
270 struct btrfs_root_item root_item;
271 struct extent_buffer *tmp;
272 u64 objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
275 ret = btrfs_copy_root(trans, root, root->node, &tmp, objectid);
278 memcpy(&root_item, &root->root_item, sizeof(root_item));
279 btrfs_set_root_bytenr(&root_item, tmp->start);
280 btrfs_set_root_level(&root_item, btrfs_header_level(tmp));
281 btrfs_set_root_generation(&root_item, trans->transid);
282 free_extent_buffer(tmp);
284 location.objectid = objectid;
285 location.type = BTRFS_ROOT_ITEM_KEY;
287 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
288 &location, &root_item);
293 static void print_usage(int ret)
295 fprintf(stderr, "usage: mkfs.btrfs [options] dev [ dev ... ]\n");
296 fprintf(stderr, "options:\n");
297 fprintf(stderr, "\t-A|--alloc-start START the offset to start the FS\n");
298 fprintf(stderr, "\t-b|--byte-count SIZE total number of bytes in the FS\n");
299 fprintf(stderr, "\t-d|--data PROFILE data profile, raid0, raid1, raid5, raid6, raid10, dup or single\n");
300 fprintf(stderr, "\t-f|--force force overwrite of existing filesystem\n");
301 fprintf(stderr, "\t-l|--leafsize SIZE deprecated, alias for nodesize\n");
302 fprintf(stderr, "\t-L|--label LABEL set a label\n");
303 fprintf(stderr, "\t-m|--metadata PROFILE metadata profile, values like data profile\n");
304 fprintf(stderr, "\t-M|--mixed mix metadata and data together\n");
305 fprintf(stderr, "\t-n|--nodesize SIZE size of btree nodes\n");
306 fprintf(stderr, "\t-s|--sectorsize SIZE min block allocation (may not mountable by current kernel)\n");
307 fprintf(stderr, "\t-r|--rootdir DIR the source directory\n");
308 fprintf(stderr, "\t-K|--nodiscard do not perform whole device TRIM\n");
309 fprintf(stderr, "\t-O|--features LIST comma separated list of filesystem features, use '-O list-all' to list features\n");
310 fprintf(stderr, "\t-U|--uuid UUID specify the filesystem UUID\n");
311 fprintf(stderr, "\t-q|--quiet no messages except errors\n");
312 fprintf(stderr, "\t-V|--version print the mkfs.btrfs version and exit\n");
313 fprintf(stderr, "%s\n", PACKAGE_STRING);
317 static void print_version(void) __attribute__((noreturn));
318 static void print_version(void)
320 fprintf(stderr, "mkfs.btrfs, part of %s\n", PACKAGE_STRING);
324 static u64 parse_profile(char *s)
326 if (strcmp(s, "raid0") == 0) {
327 return BTRFS_BLOCK_GROUP_RAID0;
328 } else if (strcasecmp(s, "raid1") == 0) {
329 return BTRFS_BLOCK_GROUP_RAID1;
330 } else if (strcasecmp(s, "raid5") == 0) {
331 return BTRFS_BLOCK_GROUP_RAID5;
332 } else if (strcasecmp(s, "raid6") == 0) {
333 return BTRFS_BLOCK_GROUP_RAID6;
334 } else if (strcasecmp(s, "raid10") == 0) {
335 return BTRFS_BLOCK_GROUP_RAID10;
336 } else if (strcasecmp(s, "dup") == 0) {
337 return BTRFS_BLOCK_GROUP_DUP;
338 } else if (strcasecmp(s, "single") == 0) {
341 fprintf(stderr, "Unknown profile %s\n", s);
347 static char *parse_label(char *input)
349 int len = strlen(input);
351 if (len >= BTRFS_LABEL_SIZE) {
352 fprintf(stderr, "Label %s is too long (max %d)\n", input,
353 BTRFS_LABEL_SIZE - 1);
356 return strdup(input);
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,
920 struct mkfs_allocation *allocation)
924 u64 meta_type = BTRFS_BLOCK_GROUP_METADATA;
925 u64 data_type = BTRFS_BLOCK_GROUP_DATA;
926 u64 minimum_data_chunk_size = 8 * 1024 * 1024;
930 for (i = 0; i < num_of_meta_chunks; i++) {
931 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
932 &chunk_start, &chunk_size, meta_type);
934 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
935 meta_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
936 chunk_start, chunk_size);
937 allocation->metadata += chunk_size;
939 set_extent_dirty(&root->fs_info->free_space_cache,
940 chunk_start, chunk_start + chunk_size - 1, 0);
943 if (size_of_data < minimum_data_chunk_size)
944 size_of_data = minimum_data_chunk_size;
946 ret = btrfs_alloc_data_chunk(trans, root->fs_info->extent_root,
947 &chunk_start, size_of_data, data_type);
949 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
950 data_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
951 chunk_start, size_of_data);
952 allocation->data += size_of_data;
954 set_extent_dirty(&root->fs_info->free_space_cache,
955 chunk_start, chunk_start + size_of_data - 1, 0);
959 static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
962 struct btrfs_trans_handle *trans;
966 struct directory_name_entry dir_head;
968 struct directory_name_entry *dir_entry = NULL;
970 ret = lstat(source_dir, &root_st);
972 fprintf(stderr, "unable to lstat the %s\n", source_dir);
976 INIT_LIST_HEAD(&dir_head.list);
978 trans = btrfs_start_transaction(root, 1);
979 ret = traverse_directory(trans, root, source_dir, &dir_head, out_fd);
981 fprintf(stderr, "unable to traverse_directory\n");
984 btrfs_commit_transaction(trans, root);
987 printf("Making image is completed.\n");
990 while (!list_empty(&dir_head.list)) {
991 dir_entry = list_entry(dir_head.list.next,
992 struct directory_name_entry, list);
993 list_del(&dir_entry->list);
997 fprintf(stderr, "Making image is aborted.\n");
1002 * This ignores symlinks with unreadable targets and subdirs that can't
1003 * be read. It's a best-effort to give a rough estimate of the size of
1004 * a subdir. It doesn't guarantee that prepopulating btrfs from this
1005 * tree won't still run out of space.
1007 * The rounding up to 4096 is questionable. Previous code used du -B 4096.
1009 static u64 global_total_size;
1010 static int ftw_add_entry_size(const char *fpath, const struct stat *st,
1013 if (type == FTW_F || type == FTW_D)
1014 global_total_size += round_up(st->st_size, 4096);
1019 static u64 size_sourcedir(char *dir_name, u64 sectorsize,
1020 u64 *num_of_meta_chunks_ret, u64 *size_of_data_ret)
1025 u64 default_chunk_size = 8 * 1024 * 1024; /* 8MB */
1026 u64 allocated_meta_size = 8 * 1024 * 1024; /* 8MB */
1027 u64 allocated_total_size = 20 * 1024 * 1024; /* 20MB */
1028 u64 num_of_meta_chunks = 0;
1029 u64 num_of_data_chunks = 0;
1030 u64 num_of_allocated_meta_chunks =
1031 allocated_meta_size / default_chunk_size;
1033 global_total_size = 0;
1034 ret = ftw(dir_name, ftw_add_entry_size, 10);
1035 dir_size = global_total_size;
1037 fprintf(stderr, "ftw subdir walk of '%s' failed: %s\n",
1038 dir_name, strerror(errno));
1042 num_of_data_chunks = (dir_size + default_chunk_size - 1) /
1045 num_of_meta_chunks = (dir_size / 2) / default_chunk_size;
1046 if (((dir_size / 2) % default_chunk_size) != 0)
1047 num_of_meta_chunks++;
1048 if (num_of_meta_chunks <= num_of_allocated_meta_chunks)
1049 num_of_meta_chunks = 0;
1051 num_of_meta_chunks -= num_of_allocated_meta_chunks;
1053 total_size = allocated_total_size +
1054 (num_of_data_chunks * default_chunk_size) +
1055 (num_of_meta_chunks * default_chunk_size);
1057 *num_of_meta_chunks_ret = num_of_meta_chunks;
1058 *size_of_data_ret = num_of_data_chunks * default_chunk_size;
1062 static int zero_output_file(int out_fd, u64 size, u32 sectorsize)
1064 int len = sectorsize;
1065 int loop_num = size / sectorsize;
1067 char *buf = malloc(len);
1073 memset(buf, 0, len);
1074 for (i = 0; i < loop_num; i++) {
1075 written = pwrite64(out_fd, buf, len, location);
1078 location += sectorsize;
1084 static int is_ssd(const char *file)
1088 char sysfs_path[PATH_MAX];
1094 probe = blkid_new_probe_from_filename(file);
1098 /* Device number of this disk (possibly a partition) */
1099 devno = blkid_probe_get_devno(probe);
1101 blkid_free_probe(probe);
1105 /* Get whole disk name (not full path) for this devno */
1106 ret = blkid_devno_to_wholedisk(devno,
1107 wholedisk, sizeof(wholedisk), NULL);
1109 blkid_free_probe(probe);
1113 snprintf(sysfs_path, PATH_MAX, "/sys/block/%s/queue/rotational",
1116 blkid_free_probe(probe);
1118 fd = open(sysfs_path, O_RDONLY);
1123 if (read(fd, &rotational, sizeof(char)) < sizeof(char)) {
1129 return !atoi((const char *)&rotational);
1132 static void list_all_devices(struct btrfs_root *root)
1134 struct btrfs_fs_devices *fs_devices;
1135 struct btrfs_device *device;
1136 int number_of_devices = 0;
1137 u64 total_block_count = 0;
1139 fs_devices = root->fs_info->fs_devices;
1141 list_for_each_entry(device, &fs_devices->devices, dev_list)
1142 number_of_devices++;
1144 printf("Number of devices: %d\n", number_of_devices);
1145 /* printf("Total devices size: %10s\n", */
1146 /* pretty_size(total_block_count)); */
1147 printf("Devices:\n");
1148 printf(" ID SIZE PATH\n");
1149 list_for_each_entry_reverse(device, &fs_devices->devices, dev_list) {
1150 char dev_uuid[BTRFS_UUID_UNPARSED_SIZE];
1152 uuid_unparse(device->uuid, dev_uuid);
1153 printf(" %3llu %10s %s\n",
1155 pretty_size(device->total_bytes),
1157 total_block_count += device->total_bytes;
1163 int main(int ac, char **av)
1166 struct btrfs_root *root;
1167 struct btrfs_trans_handle *trans;
1169 u64 block_count = 0;
1170 u64 dev_block_count = 0;
1172 u64 alloc_start = 0;
1173 u64 metadata_profile = 0;
1174 u64 data_profile = 0;
1175 u32 nodesize = max_t(u32, sysconf(_SC_PAGESIZE),
1176 BTRFS_MKFS_DEFAULT_NODE_SIZE);
1177 u32 sectorsize = 4096;
1178 u32 stripesize = 4096;
1184 int nodesize_forced = 0;
1185 int data_profile_opt = 0;
1186 int metadata_profile_opt = 0;
1189 int force_overwrite = 0;
1191 char *source_dir = NULL;
1192 int source_dir_set = 0;
1193 u64 num_of_meta_chunks = 0;
1194 u64 size_of_data = 0;
1195 u64 source_dir_size = 0;
1199 char fs_uuid[BTRFS_UUID_UNPARSED_SIZE] = { 0 };
1200 u64 features = BTRFS_MKFS_DEFAULT_FEATURES;
1201 struct mkfs_allocation allocation = { 0 };
1205 static const struct option long_options[] = {
1206 { "alloc-start", required_argument, NULL, 'A'},
1207 { "byte-count", required_argument, NULL, 'b' },
1208 { "force", no_argument, NULL, 'f' },
1209 { "leafsize", required_argument, NULL, 'l' },
1210 { "label", required_argument, NULL, 'L'},
1211 { "metadata", required_argument, NULL, 'm' },
1212 { "mixed", no_argument, NULL, 'M' },
1213 { "nodesize", required_argument, NULL, 'n' },
1214 { "sectorsize", required_argument, NULL, 's' },
1215 { "data", required_argument, NULL, 'd' },
1216 { "version", no_argument, NULL, 'V' },
1217 { "rootdir", required_argument, NULL, 'r' },
1218 { "nodiscard", no_argument, NULL, 'K' },
1219 { "features", required_argument, NULL, 'O' },
1220 { "uuid", required_argument, NULL, 'U' },
1221 { "quiet", 0, NULL, 'q' },
1222 { "help", no_argument, NULL, GETOPT_VAL_HELP },
1226 c = getopt_long(ac, av, "A:b:fl:n:s:m:d:L:O:r:U:VMKq",
1227 long_options, NULL);
1232 alloc_start = parse_size(optarg);
1235 force_overwrite = 1;
1238 data_profile = parse_profile(optarg);
1239 data_profile_opt = 1;
1243 "WARNING: --leafsize is deprecated, use --nodesize\n");
1245 nodesize = parse_size(optarg);
1246 nodesize_forced = 1;
1249 label = parse_label(optarg);
1252 metadata_profile = parse_profile(optarg);
1253 metadata_profile_opt = 1;
1259 char *orig = strdup(optarg);
1262 tmp = btrfs_parse_fs_features(tmp, &features);
1265 "Unrecognized filesystem feature '%s'\n",
1271 if (features & BTRFS_FEATURE_LIST_ALL) {
1272 btrfs_list_all_fs_features(0);
1278 sectorsize = parse_size(optarg);
1281 block_count = parse_size(optarg);
1282 if (block_count <= BTRFS_MKFS_SMALL_VOLUME_SIZE)
1290 source_dir = optarg;
1294 strncpy(fs_uuid, optarg,
1295 BTRFS_UUID_UNPARSED_SIZE - 1);
1303 case GETOPT_VAL_HELP:
1305 print_usage(c != GETOPT_VAL_HELP);
1308 sectorsize = max(sectorsize, (u32)sysconf(_SC_PAGESIZE));
1309 if (btrfs_check_nodesize(nodesize, sectorsize))
1311 saved_optind = optind;
1312 dev_cnt = ac - optind;
1316 if (source_dir_set && dev_cnt > 1) {
1318 "The -r option is limited to a single device\n");
1325 if (uuid_parse(fs_uuid, dummy_uuid) != 0) {
1326 fprintf(stderr, "could not parse UUID: %s\n", fs_uuid);
1329 if (!test_uuid_unique(fs_uuid)) {
1330 fprintf(stderr, "non-unique UUID: %s\n", fs_uuid);
1335 while (dev_cnt-- > 0) {
1336 file = av[optind++];
1337 if (is_block_device(file))
1338 if (test_dev_for_mkfs(file, force_overwrite))
1342 optind = saved_optind;
1343 dev_cnt = ac - optind;
1345 file = av[optind++];
1348 if (is_vol_small(file) || mixed) {
1350 printf("SMALL VOLUME: forcing mixed metadata/data groups\n");
1355 * Set default profiles according to number of added devices.
1356 * For mixed groups defaults are single/single.
1359 if (!metadata_profile_opt) {
1360 if (dev_cnt == 1 && ssd && verbose)
1361 printf("Detected a SSD, turning off metadata "
1362 "duplication. Mkfs with -m dup if you want to "
1363 "force metadata duplication.\n");
1365 metadata_profile = (dev_cnt > 1) ?
1366 BTRFS_BLOCK_GROUP_RAID1 : (ssd) ?
1367 0: BTRFS_BLOCK_GROUP_DUP;
1369 if (!data_profile_opt) {
1370 data_profile = (dev_cnt > 1) ?
1371 BTRFS_BLOCK_GROUP_RAID0 : 0; /* raid0 or single */
1374 u32 best_nodesize = max_t(u32, sysconf(_SC_PAGESIZE), sectorsize);
1376 if (metadata_profile_opt || data_profile_opt) {
1377 if (metadata_profile != data_profile) {
1379 "ERROR: With mixed block groups data and metadata profiles must be the same\n");
1384 if (!nodesize_forced) {
1385 nodesize = best_nodesize;
1386 if (btrfs_check_nodesize(nodesize, sectorsize))
1389 if (nodesize != sectorsize) {
1390 fprintf(stderr, "Error: mixed metadata/data block groups "
1391 "require metadata blocksizes equal to the sectorsize\n");
1396 /* Check device/block_count after the nodesize is determined */
1397 if (block_count && block_count < btrfs_min_dev_size(nodesize)) {
1399 "Size '%llu' is too small to make a usable filesystem\n",
1402 "Minimum size for btrfs filesystem is %llu\n",
1403 btrfs_min_dev_size(nodesize));
1406 for (i = saved_optind; i < saved_optind + dev_cnt; i++) {
1410 ret = test_minimum_size(path, nodesize);
1412 fprintf(stderr, "Failed to check size for '%s': %s\n",
1413 path, strerror(-ret));
1418 "'%s' is too small to make a usable filesystem\n",
1421 "Minimum size for each btrfs device is %llu.\n",
1422 btrfs_min_dev_size(nodesize));
1426 ret = test_num_disk_vs_raid(metadata_profile, data_profile,
1427 dev_cnt, mixed, estr);
1429 fprintf(stderr, "Error: %s\n", estr);
1433 /* if we are here that means all devs are good to btrfsify */
1435 printf("%s\n", PACKAGE_STRING);
1436 printf("See %s for more information.\n\n", PACKAGE_URL);
1441 if (!source_dir_set) {
1443 * open without O_EXCL so that the problem should not
1444 * occur by the following processing.
1445 * (btrfs_register_one_device() fails if O_EXCL is on)
1447 fd = open(file, O_RDWR);
1449 fprintf(stderr, "unable to open %s: %s\n", file,
1453 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1454 block_count, &mixed, discard);
1459 if (block_count && block_count > dev_block_count) {
1460 fprintf(stderr, "%s is smaller than requested size\n", file);
1464 fd = open_target(file);
1466 fprintf(stderr, "unable to open the %s\n", file);
1470 source_dir_size = size_sourcedir(source_dir, sectorsize,
1471 &num_of_meta_chunks, &size_of_data);
1472 if(block_count < source_dir_size)
1473 block_count = source_dir_size;
1474 ret = zero_output_file(fd, block_count, sectorsize);
1476 fprintf(stderr, "unable to zero the output file\n");
1479 /* our "device" is the new image file */
1480 dev_block_count = block_count;
1483 /* To create the first block group and chunk 0 in make_btrfs */
1484 if (dev_block_count < BTRFS_MKFS_SYSTEM_GROUP_SIZE) {
1485 fprintf(stderr, "device is too small to make filesystem\n");
1489 blocks[0] = BTRFS_SUPER_INFO_OFFSET;
1490 for (i = 1; i < 7; i++) {
1491 blocks[i] = BTRFS_SUPER_INFO_OFFSET + 1024 * 1024 +
1495 if (group_profile_max_safe_loss(metadata_profile) <
1496 group_profile_max_safe_loss(data_profile)){
1498 "WARNING: metatdata has lower redundancy than data!\n\n");
1502 * FS features that can be set by other means than -O
1503 * just set the bit here
1506 features |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
1508 if ((data_profile | metadata_profile) &
1509 (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
1510 features |= BTRFS_FEATURE_INCOMPAT_RAID56;
1513 ret = make_btrfs(fd, file, label, fs_uuid, blocks, dev_block_count,
1514 nodesize, sectorsize, stripesize, features);
1516 fprintf(stderr, "error during mkfs: %s\n", strerror(-ret));
1520 root = open_ctree(file, 0, OPEN_CTREE_WRITES);
1522 fprintf(stderr, "Open ctree failed\n");
1526 root->fs_info->alloc_start = alloc_start;
1528 ret = make_root_dir(root, mixed, &allocation);
1530 fprintf(stderr, "failed to setup the root directory\n");
1534 trans = btrfs_start_transaction(root, 1);
1536 if (is_block_device(file))
1537 btrfs_register_one_device(file);
1542 while (dev_cnt-- > 0) {
1543 int old_mixed = mixed;
1545 file = av[optind++];
1548 * open without O_EXCL so that the problem should not
1549 * occur by the following processing.
1550 * (btrfs_register_one_device() fails if O_EXCL is on)
1552 fd = open(file, O_RDWR);
1554 fprintf(stderr, "unable to open %s: %s\n", file,
1558 ret = btrfs_device_already_in_root(root, fd,
1559 BTRFS_SUPER_INFO_OFFSET);
1561 fprintf(stderr, "skipping duplicate device %s in FS\n",
1566 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1567 block_count, &mixed, discard);
1574 ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
1575 sectorsize, sectorsize, sectorsize);
1578 struct btrfs_device *device;
1580 device = container_of(root->fs_info->fs_devices->devices.next,
1581 struct btrfs_device, dev_list);
1582 printf("adding device %s id %llu\n", file,
1583 (unsigned long long)device->devid);
1586 if (is_block_device(file))
1587 btrfs_register_one_device(file);
1591 if (!source_dir_set) {
1592 ret = create_raid_groups(trans, root, data_profile,
1593 data_profile_opt, metadata_profile,
1594 mixed, &allocation);
1598 ret = create_data_reloc_tree(trans, root);
1601 btrfs_commit_transaction(trans, root);
1603 if (source_dir_set) {
1604 trans = btrfs_start_transaction(root, 1);
1605 ret = create_chunks(trans, root,
1606 num_of_meta_chunks, size_of_data,
1609 btrfs_commit_transaction(trans, root);
1611 ret = make_image(source_dir, root, fd);
1616 char features_buf[64];
1618 printf("Label: %s\n", label);
1619 printf("UUID: %s\n", fs_uuid);
1620 printf("Node size: %u\n", nodesize);
1621 printf("Sector size: %u\n", sectorsize);
1622 printf("Filesystem size: %s\n",
1623 pretty_size(btrfs_super_total_bytes(root->fs_info->super_copy)));
1624 printf("Block group profiles:\n");
1625 if (allocation.data)
1626 printf(" Data: %-8s %16s\n",
1627 btrfs_group_profile_str(data_profile),
1628 pretty_size(allocation.data));
1629 if (allocation.metadata)
1630 printf(" Metadata: %-8s %16s\n",
1631 btrfs_group_profile_str(metadata_profile),
1632 pretty_size(allocation.metadata));
1633 if (allocation.mixed)
1634 printf(" Data+Metadata: %-8s %16s\n",
1635 btrfs_group_profile_str(data_profile),
1636 pretty_size(allocation.mixed));
1637 printf(" System: %-8s %16s\n",
1638 btrfs_group_profile_str(metadata_profile),
1639 pretty_size(allocation.system));
1640 printf("SSD detected: %s\n", ssd ? "yes" : "no");
1641 btrfs_parse_features_to_string(features_buf, features);
1642 printf("Incompat features: %s", features_buf);
1645 list_all_devices(root);
1648 ret = close_ctree(root);