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 static int make_root_dir(struct btrfs_root *root, int mixed)
57 struct btrfs_trans_handle *trans;
58 struct btrfs_key location;
64 trans = btrfs_start_transaction(root, 1);
65 bytes_used = btrfs_super_bytes_used(root->fs_info->super_copy);
67 root->fs_info->system_allocs = 1;
68 ret = btrfs_make_block_group(trans, root, bytes_used,
69 BTRFS_BLOCK_GROUP_SYSTEM,
70 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
71 0, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
75 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
76 &chunk_start, &chunk_size,
77 BTRFS_BLOCK_GROUP_METADATA |
78 BTRFS_BLOCK_GROUP_DATA);
81 "no space to alloc data/metadata chunk\n");
85 ret = btrfs_make_block_group(trans, root, 0,
86 BTRFS_BLOCK_GROUP_METADATA |
87 BTRFS_BLOCK_GROUP_DATA,
88 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
89 chunk_start, chunk_size);
92 printf("Created a data/metadata chunk of size %llu\n",
95 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
96 &chunk_start, &chunk_size,
97 BTRFS_BLOCK_GROUP_METADATA);
99 fprintf(stderr, "no space to alloc metadata chunk\n");
103 ret = btrfs_make_block_group(trans, root, 0,
104 BTRFS_BLOCK_GROUP_METADATA,
105 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
106 chunk_start, chunk_size);
110 root->fs_info->system_allocs = 0;
111 btrfs_commit_transaction(trans, root);
112 trans = btrfs_start_transaction(root, 1);
116 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
117 &chunk_start, &chunk_size,
118 BTRFS_BLOCK_GROUP_DATA);
119 if (ret == -ENOSPC) {
120 fprintf(stderr, "no space to alloc data chunk\n");
124 ret = btrfs_make_block_group(trans, root, 0,
125 BTRFS_BLOCK_GROUP_DATA,
126 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
127 chunk_start, chunk_size);
131 ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
132 BTRFS_ROOT_TREE_DIR_OBJECTID);
135 ret = btrfs_make_root_dir(trans, root, BTRFS_FIRST_FREE_OBJECTID);
138 memcpy(&location, &root->fs_info->fs_root->root_key, sizeof(location));
139 location.offset = (u64)-1;
140 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
142 btrfs_super_root_dir(root->fs_info->super_copy),
143 &location, BTRFS_FT_DIR, 0);
147 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
148 "default", 7, location.objectid,
149 BTRFS_ROOT_TREE_DIR_OBJECTID, 0);
153 btrfs_commit_transaction(trans, root);
158 static void __recow_root(struct btrfs_trans_handle *trans,
159 struct btrfs_root *root)
162 struct extent_buffer *tmp;
164 if (trans->transid != btrfs_root_generation(&root->root_item)) {
165 extent_buffer_get(root->node);
166 ret = __btrfs_cow_block(trans, root, root->node,
167 NULL, 0, &tmp, 0, 0);
169 free_extent_buffer(tmp);
173 static void recow_roots(struct btrfs_trans_handle *trans,
174 struct btrfs_root *root)
176 struct btrfs_fs_info *info = root->fs_info;
178 __recow_root(trans, info->fs_root);
179 __recow_root(trans, info->tree_root);
180 __recow_root(trans, info->extent_root);
181 __recow_root(trans, info->chunk_root);
182 __recow_root(trans, info->dev_root);
183 __recow_root(trans, info->csum_root);
186 static int create_one_raid_group(struct btrfs_trans_handle *trans,
187 struct btrfs_root *root, u64 type)
193 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
194 &chunk_start, &chunk_size, type);
195 if (ret == -ENOSPC) {
196 fprintf(stderr, "not enough free space\n");
200 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
201 type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
202 chunk_start, chunk_size);
207 static int create_raid_groups(struct btrfs_trans_handle *trans,
208 struct btrfs_root *root, u64 data_profile,
209 int data_profile_opt, u64 metadata_profile,
212 u64 num_devices = btrfs_super_num_devices(root->fs_info->super_copy);
215 if (metadata_profile) {
216 u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
218 ret = create_one_raid_group(trans, root,
219 BTRFS_BLOCK_GROUP_SYSTEM |
224 meta_flags |= BTRFS_BLOCK_GROUP_DATA;
226 ret = create_one_raid_group(trans, root, meta_flags |
231 if (!mixed && num_devices > 1 && data_profile) {
232 ret = create_one_raid_group(trans, root,
233 BTRFS_BLOCK_GROUP_DATA |
237 recow_roots(trans, root);
242 static int create_data_reloc_tree(struct btrfs_trans_handle *trans,
243 struct btrfs_root *root)
245 struct btrfs_key location;
246 struct btrfs_root_item root_item;
247 struct extent_buffer *tmp;
248 u64 objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
251 ret = btrfs_copy_root(trans, root, root->node, &tmp, objectid);
254 memcpy(&root_item, &root->root_item, sizeof(root_item));
255 btrfs_set_root_bytenr(&root_item, tmp->start);
256 btrfs_set_root_level(&root_item, btrfs_header_level(tmp));
257 btrfs_set_root_generation(&root_item, trans->transid);
258 free_extent_buffer(tmp);
260 location.objectid = objectid;
261 location.type = BTRFS_ROOT_ITEM_KEY;
263 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
264 &location, &root_item);
270 static void print_usage(void) __attribute__((noreturn));
271 static void print_usage(void)
273 fprintf(stderr, "usage: mkfs.btrfs [options] dev [ dev ... ]\n");
274 fprintf(stderr, "options:\n");
275 fprintf(stderr, "\t-A|--alloc-start START the offset to start the FS\n");
276 fprintf(stderr, "\t-b|--byte-count SIZE total number of bytes in the FS\n");
277 fprintf(stderr, "\t-d|--data PROFILE data profile, raid0, raid1, raid5, raid6, raid10, dup or single\n");
278 fprintf(stderr, "\t-f|--force force overwrite of existing filesystem\n");
279 fprintf(stderr, "\t-l|--leafsize SIZE deprecated, alias for nodesize\n");
280 fprintf(stderr, "\t-L|--label LABEL set a label\n");
281 fprintf(stderr, "\t-m|--metadata PROFILE metadata profile, values like data profile\n");
282 fprintf(stderr, "\t-M|--mixed mix metadata and data together\n");
283 fprintf(stderr, "\t-n|--nodesize SIZE size of btree nodes\n");
284 fprintf(stderr, "\t-s|--sectorsize SIZE min block allocation (may not mountable by current kernel)\n");
285 fprintf(stderr, "\t-r|--rootdir DIR the source directory\n");
286 fprintf(stderr, "\t-K|--nodiscard do not perform whole device TRIM\n");
287 fprintf(stderr, "\t-O|--features LIST comma separated list of filesystem features, use '-O list-all' to list features\n");
288 fprintf(stderr, "\t-U|--uuid UUID specify the filesystem UUID\n");
289 fprintf(stderr, "\t-q|--quiet no messages except errors\n");
290 fprintf(stderr, "\t-V|--version print the mkfs.btrfs version and exit\n");
291 fprintf(stderr, "%s\n", PACKAGE_STRING);
295 static void print_version(void) __attribute__((noreturn));
296 static void print_version(void)
298 fprintf(stderr, "mkfs.btrfs, part of %s\n", PACKAGE_STRING);
302 static u64 parse_profile(char *s)
304 if (strcmp(s, "raid0") == 0) {
305 return BTRFS_BLOCK_GROUP_RAID0;
306 } else if (strcmp(s, "raid1") == 0) {
307 return BTRFS_BLOCK_GROUP_RAID1;
308 } else if (strcmp(s, "raid5") == 0) {
309 return BTRFS_BLOCK_GROUP_RAID5;
310 } else if (strcmp(s, "raid6") == 0) {
311 return BTRFS_BLOCK_GROUP_RAID6;
312 } else if (strcmp(s, "raid10") == 0) {
313 return BTRFS_BLOCK_GROUP_RAID10;
314 } else if (strcmp(s, "dup") == 0) {
315 return BTRFS_BLOCK_GROUP_DUP;
316 } else if (strcmp(s, "single") == 0) {
319 fprintf(stderr, "Unknown profile %s\n", s);
326 static char *parse_label(char *input)
328 int len = strlen(input);
330 if (len >= BTRFS_LABEL_SIZE) {
331 fprintf(stderr, "Label %s is too long (max %d)\n", input,
332 BTRFS_LABEL_SIZE - 1);
335 return strdup(input);
338 static int add_directory_items(struct btrfs_trans_handle *trans,
339 struct btrfs_root *root, u64 objectid,
340 ino_t parent_inum, const char *name,
341 struct stat *st, int *dir_index_cnt)
345 struct btrfs_key location;
348 name_len = strlen(name);
350 location.objectid = objectid;
352 btrfs_set_key_type(&location, BTRFS_INODE_ITEM_KEY);
354 if (S_ISDIR(st->st_mode))
355 filetype = BTRFS_FT_DIR;
356 if (S_ISREG(st->st_mode))
357 filetype = BTRFS_FT_REG_FILE;
358 if (S_ISLNK(st->st_mode))
359 filetype = BTRFS_FT_SYMLINK;
361 ret = btrfs_insert_dir_item(trans, root, name, name_len,
362 parent_inum, &location,
363 filetype, index_cnt);
366 ret = btrfs_insert_inode_ref(trans, root, name, name_len,
367 objectid, parent_inum, index_cnt);
368 *dir_index_cnt = index_cnt;
374 static int fill_inode_item(struct btrfs_trans_handle *trans,
375 struct btrfs_root *root,
376 struct btrfs_inode_item *dst, struct stat *src)
379 u64 sectorsize = root->sectorsize;
382 * btrfs_inode_item has some reserved fields
383 * and represents on-disk inode entry, so
384 * zero everything to prevent information leak
386 memset(dst, 0, sizeof (*dst));
388 btrfs_set_stack_inode_generation(dst, trans->transid);
389 btrfs_set_stack_inode_size(dst, src->st_size);
390 btrfs_set_stack_inode_nbytes(dst, 0);
391 btrfs_set_stack_inode_block_group(dst, 0);
392 btrfs_set_stack_inode_nlink(dst, src->st_nlink);
393 btrfs_set_stack_inode_uid(dst, src->st_uid);
394 btrfs_set_stack_inode_gid(dst, src->st_gid);
395 btrfs_set_stack_inode_mode(dst, src->st_mode);
396 btrfs_set_stack_inode_rdev(dst, 0);
397 btrfs_set_stack_inode_flags(dst, 0);
398 btrfs_set_stack_timespec_sec(&dst->atime, src->st_atime);
399 btrfs_set_stack_timespec_nsec(&dst->atime, 0);
400 btrfs_set_stack_timespec_sec(&dst->ctime, src->st_ctime);
401 btrfs_set_stack_timespec_nsec(&dst->ctime, 0);
402 btrfs_set_stack_timespec_sec(&dst->mtime, src->st_mtime);
403 btrfs_set_stack_timespec_nsec(&dst->mtime, 0);
404 btrfs_set_stack_timespec_sec(&dst->otime, 0);
405 btrfs_set_stack_timespec_nsec(&dst->otime, 0);
407 if (S_ISDIR(src->st_mode)) {
408 btrfs_set_stack_inode_size(dst, 0);
409 btrfs_set_stack_inode_nlink(dst, 1);
411 if (S_ISREG(src->st_mode)) {
412 btrfs_set_stack_inode_size(dst, (u64)src->st_size);
413 if (src->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root))
414 btrfs_set_stack_inode_nbytes(dst, src->st_size);
416 blocks = src->st_size / sectorsize;
417 if (src->st_size % sectorsize)
419 blocks *= sectorsize;
420 btrfs_set_stack_inode_nbytes(dst, blocks);
423 if (S_ISLNK(src->st_mode))
424 btrfs_set_stack_inode_nbytes(dst, src->st_size + 1);
429 static int directory_select(const struct direct *entry)
431 if ((strncmp(entry->d_name, ".", entry->d_reclen) == 0) ||
432 (strncmp(entry->d_name, "..", entry->d_reclen) == 0))
438 static void free_namelist(struct direct **files, int count)
445 for (i = 0; i < count; ++i)
450 static u64 calculate_dir_inode_size(char *dirname)
453 struct direct **files, *cur_file;
454 u64 dir_inode_size = 0;
456 count = scandir(dirname, &files, directory_select, NULL);
458 for (i = 0; i < count; i++) {
460 dir_inode_size += strlen(cur_file->d_name);
463 free_namelist(files, count);
466 return dir_inode_size;
469 static int add_inode_items(struct btrfs_trans_handle *trans,
470 struct btrfs_root *root,
471 struct stat *st, char *name,
472 u64 self_objectid, ino_t parent_inum,
473 int dir_index_cnt, struct btrfs_inode_item *inode_ret)
476 struct btrfs_key inode_key;
477 struct btrfs_inode_item btrfs_inode;
481 fill_inode_item(trans, root, &btrfs_inode, st);
482 objectid = self_objectid;
484 if (S_ISDIR(st->st_mode)) {
485 inode_size = calculate_dir_inode_size(name);
486 btrfs_set_stack_inode_size(&btrfs_inode, inode_size);
489 inode_key.objectid = objectid;
490 inode_key.offset = 0;
491 btrfs_set_key_type(&inode_key, BTRFS_INODE_ITEM_KEY);
493 ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
495 *inode_ret = btrfs_inode;
499 static int add_xattr_item(struct btrfs_trans_handle *trans,
500 struct btrfs_root *root, u64 objectid,
501 const char *file_name)
505 char xattr_list[XATTR_LIST_MAX];
507 char cur_value[XATTR_SIZE_MAX];
508 char delimiter = '\0';
509 char *next_location = xattr_list;
511 ret = llistxattr(file_name, xattr_list, XATTR_LIST_MAX);
515 fprintf(stderr, "get a list of xattr failed for %s\n",
522 cur_name = strtok(xattr_list, &delimiter);
523 while (cur_name != NULL) {
524 cur_name_len = strlen(cur_name);
525 next_location += cur_name_len + 1;
527 ret = getxattr(file_name, cur_name, cur_value, XATTR_SIZE_MAX);
531 fprintf(stderr, "get a xattr value failed for %s attr %s\n",
532 file_name, cur_name);
536 ret = btrfs_insert_xattr_item(trans, root, cur_name,
537 cur_name_len, cur_value,
540 fprintf(stderr, "insert a xattr item failed for %s\n",
544 cur_name = strtok(next_location, &delimiter);
550 static int add_symbolic_link(struct btrfs_trans_handle *trans,
551 struct btrfs_root *root,
552 u64 objectid, const char *path_name)
555 u64 sectorsize = root->sectorsize;
556 char *buf = malloc(sectorsize);
558 ret = readlink(path_name, buf, sectorsize);
560 fprintf(stderr, "readlink failed for %s\n", path_name);
563 if (ret >= sectorsize) {
564 fprintf(stderr, "symlink too long for %s", path_name);
569 buf[ret] = '\0'; /* readlink does not do it for us */
570 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
577 static int add_file_items(struct btrfs_trans_handle *trans,
578 struct btrfs_root *root,
579 struct btrfs_inode_item *btrfs_inode, u64 objectid,
580 ino_t parent_inum, struct stat *st,
581 const char *path_name, int out_fd)
586 struct btrfs_key key;
588 u32 sectorsize = root->sectorsize;
593 struct extent_buffer *eb = NULL;
596 if (st->st_size == 0)
599 fd = open(path_name, O_RDONLY);
601 fprintf(stderr, "%s open failed\n", path_name);
605 blocks = st->st_size / sectorsize;
606 if (st->st_size % sectorsize)
609 if (st->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
610 char *buffer = malloc(st->st_size);
611 ret_read = pread64(fd, buffer, st->st_size, bytes_read);
612 if (ret_read == -1) {
613 fprintf(stderr, "%s read failed\n", path_name);
618 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
619 buffer, st->st_size);
624 /* round up our st_size to the FS blocksize */
625 total_bytes = (u64)blocks * sectorsize;
628 * do our IO in extent buffers so it can work
629 * against any raid type
631 eb = malloc(sizeof(*eb) + sectorsize);
636 memset(eb, 0, sizeof(*eb) + sectorsize);
641 * keep our extent size at 1MB max, this makes it easier to work inside
642 * the tiny block groups created during mkfs
644 cur_bytes = min(total_bytes, 1024ULL * 1024);
645 ret = btrfs_reserve_extent(trans, root, cur_bytes, 0, 0, (u64)-1,
650 first_block = key.objectid;
653 while (bytes_read < cur_bytes) {
655 memset(eb->data, 0, sectorsize);
657 ret_read = pread64(fd, eb->data, sectorsize, file_pos + bytes_read);
658 if (ret_read == -1) {
659 fprintf(stderr, "%s read failed\n", path_name);
663 eb->start = first_block + bytes_read;
664 eb->len = sectorsize;
667 * we're doing the csum before we record the extent, but
670 ret = btrfs_csum_file_block(trans, root->fs_info->csum_root,
671 first_block + bytes_read + sectorsize,
672 first_block + bytes_read,
673 eb->data, sectorsize);
677 ret = write_and_map_eb(trans, root, eb);
679 fprintf(stderr, "output file write failed\n");
683 bytes_read += sectorsize;
687 ret = btrfs_record_file_extent(trans, root, objectid, btrfs_inode,
688 file_pos, first_block, cur_bytes);
694 file_pos += cur_bytes;
695 total_bytes -= cur_bytes;
706 static char *make_path(char *dir, char *name)
710 path = malloc(strlen(dir) + strlen(name) + 2);
714 if (dir[strlen(dir) - 1] != '/')
720 static int traverse_directory(struct btrfs_trans_handle *trans,
721 struct btrfs_root *root, char *dir_name,
722 struct directory_name_entry *dir_head, int out_fd)
726 struct btrfs_inode_item cur_inode;
727 struct btrfs_inode_item *inode_item;
728 int count, i, dir_index_cnt;
729 struct direct **files;
731 struct directory_name_entry *dir_entry, *parent_dir_entry;
732 struct direct *cur_file;
733 ino_t parent_inum, cur_inum;
734 ino_t highest_inum = 0;
735 char *parent_dir_name;
736 char real_path[PATH_MAX];
737 struct btrfs_path path;
738 struct extent_buffer *leaf;
739 struct btrfs_key root_dir_key;
740 u64 root_dir_inode_size = 0;
742 /* Add list for source directory */
743 dir_entry = malloc(sizeof(struct directory_name_entry));
744 dir_entry->dir_name = dir_name;
745 dir_entry->path = realpath(dir_name, real_path);
746 if (!dir_entry->path) {
747 fprintf(stderr, "get directory real path error\n");
752 parent_inum = highest_inum + BTRFS_FIRST_FREE_OBJECTID;
753 dir_entry->inum = parent_inum;
754 list_add_tail(&dir_entry->list, &dir_head->list);
756 btrfs_init_path(&path);
758 root_dir_key.objectid = btrfs_root_dirid(&root->root_item);
759 root_dir_key.offset = 0;
760 btrfs_set_key_type(&root_dir_key, BTRFS_INODE_ITEM_KEY);
761 ret = btrfs_lookup_inode(trans, root, &path, &root_dir_key, 1);
763 fprintf(stderr, "root dir lookup error\n");
767 leaf = path.nodes[0];
768 inode_item = btrfs_item_ptr(leaf, path.slots[0],
769 struct btrfs_inode_item);
771 root_dir_inode_size = calculate_dir_inode_size(dir_name);
772 btrfs_set_inode_size(leaf, inode_item, root_dir_inode_size);
773 btrfs_mark_buffer_dirty(leaf);
775 btrfs_release_path(&path);
778 parent_dir_entry = list_entry(dir_head->list.next,
779 struct directory_name_entry,
781 list_del(&parent_dir_entry->list);
783 parent_inum = parent_dir_entry->inum;
784 parent_dir_name = parent_dir_entry->dir_name;
785 if (chdir(parent_dir_entry->path)) {
786 fprintf(stderr, "chdir error for %s\n",
792 count = scandir(parent_dir_entry->path, &files,
793 directory_select, NULL);
796 fprintf(stderr, "scandir for %s failed: %s\n",
797 parent_dir_name, strerror (errno));
802 for (i = 0; i < count; i++) {
805 if (lstat(cur_file->d_name, &st) == -1) {
806 fprintf(stderr, "lstat failed for file %s\n",
812 cur_inum = st.st_ino;
813 ret = add_directory_items(trans, root,
814 cur_inum, parent_inum,
816 &st, &dir_index_cnt);
818 fprintf(stderr, "add_directory_items failed\n");
822 ret = add_inode_items(trans, root, &st,
823 cur_file->d_name, cur_inum,
824 parent_inum, dir_index_cnt,
826 if (ret == -EEXIST) {
827 BUG_ON(st.st_nlink <= 1);
831 fprintf(stderr, "add_inode_items failed\n");
835 ret = add_xattr_item(trans, root,
836 cur_inum, cur_file->d_name);
838 fprintf(stderr, "add_xattr_item failed\n");
843 if (S_ISDIR(st.st_mode)) {
844 dir_entry = malloc(sizeof(struct directory_name_entry));
845 dir_entry->dir_name = cur_file->d_name;
846 dir_entry->path = make_path(parent_dir_entry->path,
848 dir_entry->inum = cur_inum;
849 list_add_tail(&dir_entry->list, &dir_head->list);
850 } else if (S_ISREG(st.st_mode)) {
851 ret = add_file_items(trans, root, &cur_inode,
852 cur_inum, parent_inum, &st,
853 cur_file->d_name, out_fd);
855 fprintf(stderr, "add_file_items failed\n");
858 } else if (S_ISLNK(st.st_mode)) {
859 ret = add_symbolic_link(trans, root,
860 cur_inum, cur_file->d_name);
862 fprintf(stderr, "add_symbolic_link failed\n");
868 free_namelist(files, count);
869 free(parent_dir_entry);
873 } while (!list_empty(&dir_head->list));
878 free_namelist(files, count);
880 free(parent_dir_entry);
887 static int open_target(char *output_name)
890 output_fd = open(output_name, O_CREAT | O_RDWR | O_TRUNC,
891 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH);
896 static int create_chunks(struct btrfs_trans_handle *trans,
897 struct btrfs_root *root, u64 num_of_meta_chunks,
902 u64 meta_type = BTRFS_BLOCK_GROUP_METADATA;
903 u64 data_type = BTRFS_BLOCK_GROUP_DATA;
904 u64 minimum_data_chunk_size = 8 * 1024 * 1024;
908 for (i = 0; i < num_of_meta_chunks; i++) {
909 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
910 &chunk_start, &chunk_size, meta_type);
912 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
913 meta_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
914 chunk_start, chunk_size);
916 set_extent_dirty(&root->fs_info->free_space_cache,
917 chunk_start, chunk_start + chunk_size - 1, 0);
920 if (size_of_data < minimum_data_chunk_size)
921 size_of_data = minimum_data_chunk_size;
923 ret = btrfs_alloc_data_chunk(trans, root->fs_info->extent_root,
924 &chunk_start, size_of_data, data_type);
926 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
927 data_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
928 chunk_start, size_of_data);
930 set_extent_dirty(&root->fs_info->free_space_cache,
931 chunk_start, chunk_start + size_of_data - 1, 0);
935 static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
938 struct btrfs_trans_handle *trans;
942 struct directory_name_entry dir_head;
944 struct directory_name_entry *dir_entry = NULL;
946 ret = lstat(source_dir, &root_st);
948 fprintf(stderr, "unable to lstat the %s\n", source_dir);
952 INIT_LIST_HEAD(&dir_head.list);
954 trans = btrfs_start_transaction(root, 1);
955 ret = traverse_directory(trans, root, source_dir, &dir_head, out_fd);
957 fprintf(stderr, "unable to traverse_directory\n");
960 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 is_ssd(const char *file)
1064 char sysfs_path[PATH_MAX];
1070 probe = blkid_new_probe_from_filename(file);
1074 /* Device number of this disk (possibly a partition) */
1075 devno = blkid_probe_get_devno(probe);
1077 blkid_free_probe(probe);
1081 /* Get whole disk name (not full path) for this devno */
1082 ret = blkid_devno_to_wholedisk(devno,
1083 wholedisk, sizeof(wholedisk), NULL);
1085 blkid_free_probe(probe);
1089 snprintf(sysfs_path, PATH_MAX, "/sys/block/%s/queue/rotational",
1092 blkid_free_probe(probe);
1094 fd = open(sysfs_path, O_RDONLY);
1099 if (read(fd, &rotational, sizeof(char)) < sizeof(char)) {
1105 return !atoi((const char *)&rotational);
1108 int main(int ac, char **av)
1111 struct btrfs_root *root;
1112 struct btrfs_trans_handle *trans;
1115 u64 block_count = 0;
1116 u64 dev_block_count = 0;
1118 u64 alloc_start = 0;
1119 u64 metadata_profile = 0;
1120 u64 data_profile = 0;
1121 u32 nodesize = max_t(u32, sysconf(_SC_PAGESIZE),
1122 BTRFS_MKFS_DEFAULT_NODE_SIZE);
1123 u32 sectorsize = 4096;
1124 u32 stripesize = 4096;
1130 int nodesize_forced = 0;
1131 int data_profile_opt = 0;
1132 int metadata_profile_opt = 0;
1135 int force_overwrite = 0;
1137 char *source_dir = NULL;
1138 int source_dir_set = 0;
1139 u64 num_of_meta_chunks = 0;
1140 u64 size_of_data = 0;
1141 u64 source_dir_size = 0;
1145 char *fs_uuid = NULL;
1146 u64 features = BTRFS_MKFS_DEFAULT_FEATURES;
1150 static const struct option long_options[] = {
1151 { "alloc-start", required_argument, NULL, 'A'},
1152 { "byte-count", required_argument, NULL, 'b' },
1153 { "force", no_argument, NULL, 'f' },
1154 { "leafsize", required_argument, NULL, 'l' },
1155 { "label", required_argument, NULL, 'L'},
1156 { "metadata", required_argument, NULL, 'm' },
1157 { "mixed", no_argument, NULL, 'M' },
1158 { "nodesize", required_argument, NULL, 'n' },
1159 { "sectorsize", required_argument, NULL, 's' },
1160 { "data", required_argument, NULL, 'd' },
1161 { "version", no_argument, NULL, 'V' },
1162 { "rootdir", required_argument, NULL, 'r' },
1163 { "nodiscard", no_argument, NULL, 'K' },
1164 { "features", required_argument, NULL, 'O' },
1165 { "uuid", required_argument, NULL, 'U' },
1166 { "quiet", 0, NULL, 'q' },
1170 c = getopt_long(ac, av, "A:b:fl:n:s:m:d:L:O:r:U:VMKq",
1171 long_options, NULL);
1176 alloc_start = parse_size(optarg);
1179 force_overwrite = 1;
1182 data_profile = parse_profile(optarg);
1183 data_profile_opt = 1;
1187 "WARNING: --leafsize is deprecated, use --nodesize\n");
1189 nodesize = parse_size(optarg);
1190 nodesize_forced = 1;
1193 label = parse_label(optarg);
1196 metadata_profile = parse_profile(optarg);
1197 metadata_profile_opt = 1;
1203 char *orig = strdup(optarg);
1206 tmp = btrfs_parse_fs_features(tmp, &features);
1209 "Unrecognized filesystem feature '%s'\n",
1215 if (features & BTRFS_FEATURE_LIST_ALL) {
1216 btrfs_list_all_fs_features(0);
1222 sectorsize = parse_size(optarg);
1225 block_count = parse_size(optarg);
1226 if (block_count <= BTRFS_MKFS_SMALL_VOLUME_SIZE)
1234 source_dir = optarg;
1250 sectorsize = max(sectorsize, (u32)sysconf(_SC_PAGESIZE));
1251 if (btrfs_check_nodesize(nodesize, sectorsize))
1253 saved_optind = optind;
1254 dev_cnt = ac - optind;
1258 if (source_dir_set && dev_cnt > 1) {
1260 "The -r option is limited to a single device\n");
1267 if (uuid_parse(fs_uuid, dummy_uuid) != 0) {
1268 fprintf(stderr, "could not parse UUID: %s\n", fs_uuid);
1271 if (!test_uuid_unique(fs_uuid)) {
1272 fprintf(stderr, "non-unique UUID: %s\n", fs_uuid);
1277 while (dev_cnt-- > 0) {
1278 file = av[optind++];
1279 if (is_block_device(file))
1280 if (test_dev_for_mkfs(file, force_overwrite, estr)) {
1281 fprintf(stderr, "Error: %s", estr);
1286 optind = saved_optind;
1287 dev_cnt = ac - optind;
1289 file = av[optind++];
1292 if (is_vol_small(file) || mixed) {
1294 printf("SMALL VOLUME: forcing mixed metadata/data groups\n");
1299 * Set default profiles according to number of added devices.
1300 * For mixed groups defaults are single/single.
1303 if (!metadata_profile_opt) {
1304 if (dev_cnt == 1 && ssd && verbose)
1305 printf("Detected a SSD, turning off metadata "
1306 "duplication. Mkfs with -m dup if you want to "
1307 "force metadata duplication.\n");
1309 metadata_profile = (dev_cnt > 1) ?
1310 BTRFS_BLOCK_GROUP_RAID1 : (ssd) ?
1311 0: BTRFS_BLOCK_GROUP_DUP;
1313 if (!data_profile_opt) {
1314 data_profile = (dev_cnt > 1) ?
1315 BTRFS_BLOCK_GROUP_RAID0 : 0; /* raid0 or single */
1318 u32 best_nodesize = max_t(u32, sysconf(_SC_PAGESIZE), sectorsize);
1320 if (metadata_profile_opt || data_profile_opt) {
1321 if (metadata_profile != data_profile) {
1323 "ERROR: With mixed block groups data and metadata profiles must be the same\n");
1328 if (!nodesize_forced) {
1329 nodesize = best_nodesize;
1330 if (btrfs_check_nodesize(nodesize, sectorsize))
1333 if (nodesize != sectorsize) {
1334 fprintf(stderr, "Error: mixed metadata/data block groups "
1335 "require metadata blocksizes equal to the sectorsize\n");
1340 /* Check device/block_count after the nodesize is determined */
1341 if (block_count && block_count < btrfs_min_dev_size(nodesize)) {
1343 "Size '%llu' is too small to make a usable filesystem\n",
1346 "Minimum size for btrfs filesystem is %llu\n",
1347 btrfs_min_dev_size(nodesize));
1350 for (i = saved_optind; i < saved_optind + dev_cnt; i++) {
1354 ret = test_minimum_size(path, nodesize);
1356 fprintf(stderr, "Failed to check size for '%s': %s\n",
1357 path, strerror(-ret));
1362 "'%s' is too small to make a usable filesystem\n",
1365 "Minimum size for each btrfs device is %llu.\n",
1366 btrfs_min_dev_size(nodesize));
1370 ret = test_num_disk_vs_raid(metadata_profile, data_profile,
1371 dev_cnt, mixed, estr);
1373 fprintf(stderr, "Error: %s\n", estr);
1377 /* if we are here that means all devs are good to btrfsify */
1379 printf("%s\n", PACKAGE_STRING);
1380 printf("See %s for more information.\n\n", PACKAGE_URL);
1385 if (!source_dir_set) {
1387 * open without O_EXCL so that the problem should not
1388 * occur by the following processing.
1389 * (btrfs_register_one_device() fails if O_EXCL is on)
1391 fd = open(file, O_RDWR);
1393 fprintf(stderr, "unable to open %s: %s\n", file,
1398 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1399 block_count, &mixed, discard);
1404 if (block_count && block_count > dev_block_count) {
1405 fprintf(stderr, "%s is smaller than requested size\n", file);
1409 fd = open_target(file);
1411 fprintf(stderr, "unable to open the %s\n", file);
1416 source_dir_size = size_sourcedir(source_dir, sectorsize,
1417 &num_of_meta_chunks, &size_of_data);
1418 if(block_count < source_dir_size)
1419 block_count = source_dir_size;
1420 ret = zero_output_file(fd, block_count, sectorsize);
1422 fprintf(stderr, "unable to zero the output file\n");
1425 /* our "device" is the new image file */
1426 dev_block_count = block_count;
1429 /* To create the first block group and chunk 0 in make_btrfs */
1430 if (dev_block_count < BTRFS_MKFS_SYSTEM_GROUP_SIZE) {
1431 fprintf(stderr, "device is too small to make filesystem\n");
1435 blocks[0] = BTRFS_SUPER_INFO_OFFSET;
1436 for (i = 1; i < 7; i++) {
1437 blocks[i] = BTRFS_SUPER_INFO_OFFSET + 1024 * 1024 +
1441 if (group_profile_max_safe_loss(metadata_profile) <
1442 group_profile_max_safe_loss(data_profile)){
1444 "WARNING: metatdata has lower redundancy than data!\n\n");
1448 * FS features that can be set by other means than -O
1449 * just set the bit here
1452 features |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
1454 if ((data_profile | metadata_profile) &
1455 (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
1456 features |= BTRFS_FEATURE_INCOMPAT_RAID56;
1460 btrfs_process_fs_features(features);
1462 ret = make_btrfs(fd, file, label, fs_uuid, blocks, dev_block_count,
1463 nodesize, sectorsize, stripesize, features);
1465 fprintf(stderr, "error during mkfs: %s\n", strerror(-ret));
1469 root = open_ctree(file, 0, OPEN_CTREE_WRITES);
1471 fprintf(stderr, "Open ctree failed\n");
1475 root->fs_info->alloc_start = alloc_start;
1477 ret = make_root_dir(root, mixed);
1479 fprintf(stderr, "failed to setup the root directory\n");
1483 trans = btrfs_start_transaction(root, 1);
1485 btrfs_register_one_device(file);
1490 while (dev_cnt-- > 0) {
1491 int old_mixed = mixed;
1493 file = av[optind++];
1496 * open without O_EXCL so that the problem should not
1497 * occur by the following processing.
1498 * (btrfs_register_one_device() fails if O_EXCL is on)
1500 fd = open(file, O_RDWR);
1502 fprintf(stderr, "unable to open %s: %s\n", file,
1506 ret = btrfs_device_already_in_root(root, fd,
1507 BTRFS_SUPER_INFO_OFFSET);
1509 fprintf(stderr, "skipping duplicate device %s in FS\n",
1514 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1515 block_count, &mixed, discard);
1522 ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
1523 sectorsize, sectorsize, sectorsize);
1525 btrfs_register_one_device(file);
1529 if (!source_dir_set) {
1530 ret = create_raid_groups(trans, root, data_profile,
1531 data_profile_opt, metadata_profile,
1536 ret = create_data_reloc_tree(trans, root);
1541 "fs created label %s on %s\n\tnodesize %u leafsize %u sectorsize %u size %s\n",
1542 label, first_file, nodesize, nodesize, sectorsize,
1543 pretty_size(btrfs_super_total_bytes(root->fs_info->super_copy)));
1546 btrfs_commit_transaction(trans, root);
1548 if (source_dir_set) {
1549 trans = btrfs_start_transaction(root, 1);
1550 ret = create_chunks(trans, root,
1551 num_of_meta_chunks, size_of_data);
1553 btrfs_commit_transaction(trans, root);
1555 ret = make_image(source_dir, root, fd);
1559 ret = close_ctree(root);