Btrfs-progs: fall back to the v1 ioctl if the new balance ioctl fails
[btrfs-progs-unstable/devel.git] / mkfs.c
blobe3ced195bb7e96cde10fa9d5e390f991354846c3
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #define _XOPEN_SOURCE 500
20 #define _GNU_SOURCE
22 #ifndef __CHECKER__
23 #include <sys/ioctl.h>
24 #include <sys/mount.h>
25 #include "ioctl.h"
26 #endif
28 #include <stdio.h>
29 #include <stdlib.h>
30 #include <sys/types.h>
31 #include <sys/stat.h>
32 #include <sys/dir.h>
33 #include <fcntl.h>
34 #include <unistd.h>
35 #include <getopt.h>
36 #include <uuid/uuid.h>
37 #include <linux/fs.h>
38 #include <ctype.h>
39 #include <attr/xattr.h>
40 #include "kerncompat.h"
41 #include "ctree.h"
42 #include "disk-io.h"
43 #include "volumes.h"
44 #include "transaction.h"
45 #include "utils.h"
46 #include "version.h"
48 static u64 index_cnt = 2;
50 struct directory_name_entry {
51 char *dir_name;
52 char *path;
53 ino_t inum;
54 struct list_head list;
57 static u64 parse_size(char *s)
59 int len = strlen(s);
60 char c;
61 u64 mult = 1;
63 if (!isdigit(s[len - 1])) {
64 c = tolower(s[len - 1]);
65 switch (c) {
66 case 'g':
67 mult *= 1024;
68 case 'm':
69 mult *= 1024;
70 case 'k':
71 mult *= 1024;
72 case 'b':
73 break;
74 default:
75 fprintf(stderr, "Unknown size descriptor %c\n", c);
76 exit(1);
78 s[len - 1] = '\0';
80 return atol(s) * mult;
83 static int make_root_dir(struct btrfs_root *root, int mixed)
85 struct btrfs_trans_handle *trans;
86 struct btrfs_key location;
87 u64 bytes_used;
88 u64 chunk_start = 0;
89 u64 chunk_size = 0;
90 int ret;
92 trans = btrfs_start_transaction(root, 1);
93 bytes_used = btrfs_super_bytes_used(&root->fs_info->super_copy);
95 root->fs_info->system_allocs = 1;
96 ret = btrfs_make_block_group(trans, root, bytes_used,
97 BTRFS_BLOCK_GROUP_SYSTEM,
98 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
99 0, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
100 BUG_ON(ret);
102 if (mixed) {
103 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
104 &chunk_start, &chunk_size,
105 BTRFS_BLOCK_GROUP_METADATA |
106 BTRFS_BLOCK_GROUP_DATA);
107 BUG_ON(ret);
108 ret = btrfs_make_block_group(trans, root, 0,
109 BTRFS_BLOCK_GROUP_METADATA |
110 BTRFS_BLOCK_GROUP_DATA,
111 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
112 chunk_start, chunk_size);
113 BUG_ON(ret);
114 printf("Created a data/metadata chunk of size %llu\n", chunk_size);
115 } else {
116 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
117 &chunk_start, &chunk_size,
118 BTRFS_BLOCK_GROUP_METADATA);
119 BUG_ON(ret);
120 ret = btrfs_make_block_group(trans, root, 0,
121 BTRFS_BLOCK_GROUP_METADATA,
122 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
123 chunk_start, chunk_size);
124 BUG_ON(ret);
127 root->fs_info->system_allocs = 0;
128 btrfs_commit_transaction(trans, root);
129 trans = btrfs_start_transaction(root, 1);
130 BUG_ON(!trans);
132 if (!mixed) {
133 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
134 &chunk_start, &chunk_size,
135 BTRFS_BLOCK_GROUP_DATA);
136 BUG_ON(ret);
137 ret = btrfs_make_block_group(trans, root, 0,
138 BTRFS_BLOCK_GROUP_DATA,
139 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
140 chunk_start, chunk_size);
141 BUG_ON(ret);
144 ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
145 BTRFS_ROOT_TREE_DIR_OBJECTID);
146 if (ret)
147 goto err;
148 ret = btrfs_make_root_dir(trans, root, BTRFS_FIRST_FREE_OBJECTID);
149 if (ret)
150 goto err;
151 memcpy(&location, &root->fs_info->fs_root->root_key, sizeof(location));
152 location.offset = (u64)-1;
153 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
154 "default", 7,
155 btrfs_super_root_dir(&root->fs_info->super_copy),
156 &location, BTRFS_FT_DIR, 0);
157 if (ret)
158 goto err;
160 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
161 "default", 7, location.objectid,
162 BTRFS_ROOT_TREE_DIR_OBJECTID, 0);
163 if (ret)
164 goto err;
166 btrfs_commit_transaction(trans, root);
167 err:
168 return ret;
171 static int recow_roots(struct btrfs_trans_handle *trans,
172 struct btrfs_root *root)
174 int ret;
175 struct extent_buffer *tmp;
176 struct btrfs_fs_info *info = root->fs_info;
178 ret = __btrfs_cow_block(trans, info->fs_root, info->fs_root->node,
179 NULL, 0, &tmp, 0, 0);
180 BUG_ON(ret);
181 free_extent_buffer(tmp);
183 ret = __btrfs_cow_block(trans, info->tree_root, info->tree_root->node,
184 NULL, 0, &tmp, 0, 0);
185 BUG_ON(ret);
186 free_extent_buffer(tmp);
188 ret = __btrfs_cow_block(trans, info->extent_root,
189 info->extent_root->node, NULL, 0, &tmp, 0, 0);
190 BUG_ON(ret);
191 free_extent_buffer(tmp);
193 ret = __btrfs_cow_block(trans, info->chunk_root, info->chunk_root->node,
194 NULL, 0, &tmp, 0, 0);
195 BUG_ON(ret);
196 free_extent_buffer(tmp);
199 ret = __btrfs_cow_block(trans, info->dev_root, info->dev_root->node,
200 NULL, 0, &tmp, 0, 0);
201 BUG_ON(ret);
202 free_extent_buffer(tmp);
204 ret = __btrfs_cow_block(trans, info->csum_root, info->csum_root->node,
205 NULL, 0, &tmp, 0, 0);
206 BUG_ON(ret);
207 free_extent_buffer(tmp);
209 return 0;
212 static int create_one_raid_group(struct btrfs_trans_handle *trans,
213 struct btrfs_root *root, u64 type)
215 u64 chunk_start;
216 u64 chunk_size;
217 int ret;
219 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
220 &chunk_start, &chunk_size, type);
221 BUG_ON(ret);
222 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
223 type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
224 chunk_start, chunk_size);
225 BUG_ON(ret);
226 return ret;
229 static int create_raid_groups(struct btrfs_trans_handle *trans,
230 struct btrfs_root *root, u64 data_profile,
231 u64 metadata_profile, int mixed)
233 u64 num_devices = btrfs_super_num_devices(&root->fs_info->super_copy);
234 u64 allowed;
235 int ret;
237 if (num_devices == 1)
238 allowed = BTRFS_BLOCK_GROUP_DUP;
239 else if (num_devices >= 4) {
240 allowed = BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
241 BTRFS_BLOCK_GROUP_RAID10;
242 } else
243 allowed = BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1;
245 if (allowed & metadata_profile) {
246 u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
248 ret = create_one_raid_group(trans, root,
249 BTRFS_BLOCK_GROUP_SYSTEM |
250 (allowed & metadata_profile));
251 BUG_ON(ret);
253 if (mixed)
254 meta_flags |= BTRFS_BLOCK_GROUP_DATA;
256 ret = create_one_raid_group(trans, root, meta_flags |
257 (allowed & metadata_profile));
258 BUG_ON(ret);
260 ret = recow_roots(trans, root);
261 BUG_ON(ret);
263 if (!mixed && num_devices > 1 && (allowed & data_profile)) {
264 ret = create_one_raid_group(trans, root,
265 BTRFS_BLOCK_GROUP_DATA |
266 (allowed & data_profile));
267 BUG_ON(ret);
269 return 0;
272 static int create_data_reloc_tree(struct btrfs_trans_handle *trans,
273 struct btrfs_root *root)
275 struct btrfs_key location;
276 struct btrfs_root_item root_item;
277 struct extent_buffer *tmp;
278 u64 objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
279 int ret;
281 ret = btrfs_copy_root(trans, root, root->node, &tmp, objectid);
282 BUG_ON(ret);
284 memcpy(&root_item, &root->root_item, sizeof(root_item));
285 btrfs_set_root_bytenr(&root_item, tmp->start);
286 btrfs_set_root_level(&root_item, btrfs_header_level(tmp));
287 btrfs_set_root_generation(&root_item, trans->transid);
288 free_extent_buffer(tmp);
290 location.objectid = objectid;
291 location.type = BTRFS_ROOT_ITEM_KEY;
292 location.offset = 0;
293 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
294 &location, &root_item);
295 BUG_ON(ret);
296 return 0;
299 static void print_usage(void)
301 fprintf(stderr, "usage: mkfs.btrfs [options] dev [ dev ... ]\n");
302 fprintf(stderr, "options:\n");
303 fprintf(stderr, "\t -A --alloc-start the offset to start the FS\n");
304 fprintf(stderr, "\t -b --byte-count total number of bytes in the FS\n");
305 fprintf(stderr, "\t -d --data data profile, raid0, raid1, raid10 or single\n");
306 fprintf(stderr, "\t -l --leafsize size of btree leaves\n");
307 fprintf(stderr, "\t -L --label set a label\n");
308 fprintf(stderr, "\t -m --metadata metadata profile, values like data profile\n");
309 fprintf(stderr, "\t -M --mixed mix metadata and data together\n");
310 fprintf(stderr, "\t -n --nodesize size of btree nodes\n");
311 fprintf(stderr, "\t -s --sectorsize min block allocation\n");
312 fprintf(stderr, "\t -r --rootdir the source directory\n");
313 fprintf(stderr, "%s\n", BTRFS_BUILD_VERSION);
314 exit(1);
317 static void print_version(void)
319 fprintf(stderr, "mkfs.btrfs, part of %s\n", BTRFS_BUILD_VERSION);
320 exit(0);
323 static u64 parse_profile(char *s)
325 if (strcmp(s, "raid0") == 0) {
326 return BTRFS_BLOCK_GROUP_RAID0;
327 } else if (strcmp(s, "raid1") == 0) {
328 return BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_DUP;
329 } else if (strcmp(s, "raid10") == 0) {
330 return BTRFS_BLOCK_GROUP_RAID10 | BTRFS_BLOCK_GROUP_DUP;
331 } else if (strcmp(s, "single") == 0) {
332 return 0;
333 } else {
334 fprintf(stderr, "Unknown option %s\n", s);
335 print_usage();
337 return 0;
340 static char *parse_label(char *input)
342 int i;
343 int len = strlen(input);
345 if (len >= BTRFS_LABEL_SIZE) {
346 fprintf(stderr, "Label %s is too long (max %d)\n", input,
347 BTRFS_LABEL_SIZE - 1);
348 exit(1);
350 for (i = 0; i < len; i++) {
351 if (input[i] == '/' || input[i] == '\\') {
352 fprintf(stderr, "invalid label %s\n", input);
353 exit(1);
356 return strdup(input);
359 static struct option long_options[] = {
360 { "alloc-start", 1, NULL, 'A'},
361 { "byte-count", 1, NULL, 'b' },
362 { "leafsize", 1, NULL, 'l' },
363 { "label", 1, NULL, 'L'},
364 { "metadata", 1, NULL, 'm' },
365 { "mixed", 0, NULL, 'M' },
366 { "nodesize", 1, NULL, 'n' },
367 { "sectorsize", 1, NULL, 's' },
368 { "data", 1, NULL, 'd' },
369 { "version", 0, NULL, 'V' },
370 { "rootdir", 1, NULL, 'r' },
371 { 0, 0, 0, 0}
374 static int add_directory_items(struct btrfs_trans_handle *trans,
375 struct btrfs_root *root, u64 objectid,
376 ino_t parent_inum, const char *name,
377 struct stat *st, int *dir_index_cnt)
379 int ret;
380 int name_len;
381 struct btrfs_key location;
382 u8 filetype = 0;
384 name_len = strlen(name);
386 location.objectid = objectid;
387 location.offset = 0;
388 btrfs_set_key_type(&location, BTRFS_INODE_ITEM_KEY);
390 if (S_ISDIR(st->st_mode))
391 filetype = BTRFS_FT_DIR;
392 if (S_ISREG(st->st_mode))
393 filetype = BTRFS_FT_REG_FILE;
394 if (S_ISLNK(st->st_mode))
395 filetype = BTRFS_FT_SYMLINK;
397 ret = btrfs_insert_dir_item(trans, root, name, name_len,
398 parent_inum, &location,
399 filetype, index_cnt);
401 *dir_index_cnt = index_cnt;
402 index_cnt++;
404 return ret;
407 static int fill_inode_item(struct btrfs_trans_handle *trans,
408 struct btrfs_root *root,
409 struct btrfs_inode_item *dst, struct stat *src)
411 u64 blocks = 0;
412 u64 sectorsize = root->sectorsize;
415 * btrfs_inode_item has some reserved fields
416 * and represents on-disk inode entry, so
417 * zero everything to prevent information leak
419 memset(dst, 0, sizeof (*dst));
421 btrfs_set_stack_inode_generation(dst, trans->transid);
422 btrfs_set_stack_inode_size(dst, src->st_size);
423 btrfs_set_stack_inode_nbytes(dst, 0);
424 btrfs_set_stack_inode_block_group(dst, 0);
425 btrfs_set_stack_inode_nlink(dst, src->st_nlink);
426 btrfs_set_stack_inode_uid(dst, src->st_uid);
427 btrfs_set_stack_inode_gid(dst, src->st_gid);
428 btrfs_set_stack_inode_mode(dst, src->st_mode);
429 btrfs_set_stack_inode_rdev(dst, 0);
430 btrfs_set_stack_inode_flags(dst, 0);
431 btrfs_set_stack_timespec_sec(&dst->atime, src->st_atime);
432 btrfs_set_stack_timespec_nsec(&dst->atime, 0);
433 btrfs_set_stack_timespec_sec(&dst->ctime, src->st_ctime);
434 btrfs_set_stack_timespec_nsec(&dst->ctime, 0);
435 btrfs_set_stack_timespec_sec(&dst->mtime, src->st_mtime);
436 btrfs_set_stack_timespec_nsec(&dst->mtime, 0);
437 btrfs_set_stack_timespec_sec(&dst->otime, 0);
438 btrfs_set_stack_timespec_nsec(&dst->otime, 0);
440 if (S_ISDIR(src->st_mode)) {
441 btrfs_set_stack_inode_size(dst, 0);
442 btrfs_set_stack_inode_nlink(dst, 1);
444 if (S_ISREG(src->st_mode)) {
445 btrfs_set_stack_inode_size(dst, (u64)src->st_size);
446 if (src->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root))
447 btrfs_set_stack_inode_nbytes(dst, src->st_size);
448 else {
449 blocks = src->st_size / sectorsize;
450 if (src->st_size % sectorsize)
451 blocks += 1;
452 blocks *= sectorsize;
453 btrfs_set_stack_inode_nbytes(dst, blocks);
456 if (S_ISLNK(src->st_mode))
457 btrfs_set_stack_inode_nbytes(dst, src->st_size + 1);
459 return 0;
462 static int directory_select(const struct direct *entry)
464 if ((strncmp(entry->d_name, ".", entry->d_reclen) == 0) ||
465 (strncmp(entry->d_name, "..", entry->d_reclen) == 0))
466 return 0;
467 else
468 return 1;
471 static void free_namelist(struct direct **files, int count)
473 int i;
475 if (count < 0)
476 return;
478 for (i = 0; i < count; ++i)
479 free(files[i]);
480 free(files);
483 static u64 calculate_dir_inode_size(char *dirname)
485 int count, i;
486 struct direct **files, *cur_file;
487 u64 dir_inode_size = 0;
489 count = scandir(dirname, &files, directory_select, NULL);
491 for (i = 0; i < count; i++) {
492 cur_file = files[i];
493 dir_inode_size += strlen(cur_file->d_name);
496 free_namelist(files, count);
498 dir_inode_size *= 2;
499 return dir_inode_size;
502 static int add_inode_items(struct btrfs_trans_handle *trans,
503 struct btrfs_root *root,
504 struct stat *st, char *name,
505 u64 self_objectid, ino_t parent_inum,
506 int dir_index_cnt, struct btrfs_inode_item *inode_ret)
508 int ret;
509 struct btrfs_key inode_key;
510 struct btrfs_inode_item btrfs_inode;
511 u64 objectid;
512 u64 inode_size = 0;
513 int name_len;
515 name_len = strlen(name);
516 fill_inode_item(trans, root, &btrfs_inode, st);
517 objectid = self_objectid;
519 if (S_ISDIR(st->st_mode)) {
520 inode_size = calculate_dir_inode_size(name);
521 btrfs_set_stack_inode_size(&btrfs_inode, inode_size);
524 inode_key.objectid = objectid;
525 inode_key.offset = 0;
526 btrfs_set_key_type(&inode_key, BTRFS_INODE_ITEM_KEY);
528 ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
529 if (ret)
530 goto fail;
532 ret = btrfs_insert_inode_ref(trans, root, name, name_len,
533 objectid, parent_inum, dir_index_cnt);
534 if (ret)
535 goto fail;
537 *inode_ret = btrfs_inode;
538 fail:
539 return ret;
542 static int add_xattr_item(struct btrfs_trans_handle *trans,
543 struct btrfs_root *root, u64 objectid,
544 const char *file_name)
546 int ret;
547 int cur_name_len;
548 char xattr_list[XATTR_LIST_MAX];
549 char *cur_name;
550 char cur_value[XATTR_SIZE_MAX];
551 char delimiter = '\0';
552 char *next_location = xattr_list;
554 ret = llistxattr(file_name, xattr_list, XATTR_LIST_MAX);
555 if (ret < 0) {
556 if(errno == ENOTSUP)
557 return 0;
558 fprintf(stderr, "get a list of xattr failed for %s\n",
559 file_name);
560 return ret;
562 if (ret == 0)
563 return ret;
565 cur_name = strtok(xattr_list, &delimiter);
566 while (cur_name != NULL) {
567 cur_name_len = strlen(cur_name);
568 next_location += cur_name_len + 1;
570 ret = getxattr(file_name, cur_name, cur_value, XATTR_SIZE_MAX);
571 if (ret < 0) {
572 if(errno == ENOTSUP)
573 return 0;
574 fprintf(stderr, "get a xattr value failed for %s attr %s\n",
575 file_name, cur_name);
576 return ret;
579 ret = btrfs_insert_xattr_item(trans, root, cur_name,
580 cur_name_len, cur_value,
581 ret, objectid);
582 if (ret) {
583 fprintf(stderr, "insert a xattr item failed for %s\n",
584 file_name);
587 cur_name = strtok(next_location, &delimiter);
590 return ret;
592 static int custom_alloc_extent(struct btrfs_root *root, u64 num_bytes,
593 u64 hint_byte, struct btrfs_key *ins)
595 u64 start;
596 u64 end;
597 u64 last = hint_byte;
598 int ret;
599 int wrapped = 0;
600 struct btrfs_block_group_cache *cache;
602 while (1) {
603 ret = find_first_extent_bit(&root->fs_info->free_space_cache,
604 last, &start, &end, EXTENT_DIRTY);
605 if (ret) {
606 if (wrapped++ == 0) {
607 last = 0;
608 continue;
609 } else {
610 goto fail;
614 start = max(last, start);
615 last = end + 1;
616 if (last - start < num_bytes)
617 continue;
619 last = start + num_bytes;
620 if (test_range_bit(&root->fs_info->pinned_extents,
621 start, last - 1, EXTENT_DIRTY, 0))
622 continue;
624 cache = btrfs_lookup_block_group(root->fs_info, start);
625 BUG_ON(!cache);
626 if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM ||
627 last > cache->key.objectid + cache->key.offset) {
628 last = cache->key.objectid + cache->key.offset;
629 continue;
632 if (cache->flags & (BTRFS_BLOCK_GROUP_SYSTEM |
633 BTRFS_BLOCK_GROUP_METADATA)) {
634 last = cache->key.objectid + cache->key.offset;
635 continue;
638 clear_extent_dirty(&root->fs_info->free_space_cache,
639 start, start + num_bytes - 1, 0);
641 ins->objectid = start;
642 ins->offset = num_bytes;
643 ins->type = BTRFS_EXTENT_ITEM_KEY;
644 return 0;
646 fail:
647 fprintf(stderr, "not enough free space\n");
648 return -ENOSPC;
651 static int record_file_extent(struct btrfs_trans_handle *trans,
652 struct btrfs_root *root, u64 objectid,
653 struct btrfs_inode_item *inode,
654 u64 file_pos, u64 disk_bytenr,
655 u64 num_bytes)
657 int ret;
658 struct btrfs_fs_info *info = root->fs_info;
659 struct btrfs_root *extent_root = info->extent_root;
660 struct extent_buffer *leaf;
661 struct btrfs_file_extent_item *fi;
662 struct btrfs_key ins_key;
663 struct btrfs_path path;
664 struct btrfs_extent_item *ei;
666 btrfs_init_path(&path);
668 ins_key.objectid = objectid;
669 ins_key.offset = 0;
670 btrfs_set_key_type(&ins_key, BTRFS_EXTENT_DATA_KEY);
671 ret = btrfs_insert_empty_item(trans, root, &path, &ins_key,
672 sizeof(*fi));
673 if (ret)
674 goto fail;
675 leaf = path.nodes[0];
676 fi = btrfs_item_ptr(leaf, path.slots[0],
677 struct btrfs_file_extent_item);
678 btrfs_set_file_extent_generation(leaf, fi, trans->transid);
679 btrfs_set_file_extent_type(leaf, fi, BTRFS_FILE_EXTENT_REG);
680 btrfs_set_file_extent_disk_bytenr(leaf, fi, disk_bytenr);
681 btrfs_set_file_extent_disk_num_bytes(leaf, fi, num_bytes);
682 btrfs_set_file_extent_offset(leaf, fi, 0);
683 btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
684 btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
685 btrfs_set_file_extent_compression(leaf, fi, 0);
686 btrfs_set_file_extent_encryption(leaf, fi, 0);
687 btrfs_set_file_extent_other_encoding(leaf, fi, 0);
688 btrfs_mark_buffer_dirty(leaf);
690 btrfs_release_path(root, &path);
692 ins_key.objectid = disk_bytenr;
693 ins_key.offset = num_bytes;
694 ins_key.type = BTRFS_EXTENT_ITEM_KEY;
696 ret = btrfs_insert_empty_item(trans, extent_root, &path,
697 &ins_key, sizeof(*ei));
698 if (ret == 0) {
699 leaf = path.nodes[0];
700 ei = btrfs_item_ptr(leaf, path.slots[0],
701 struct btrfs_extent_item);
703 btrfs_set_extent_refs(leaf, ei, 0);
704 btrfs_set_extent_generation(leaf, ei, trans->transid);
705 btrfs_set_extent_flags(leaf, ei, BTRFS_EXTENT_FLAG_DATA);
707 btrfs_mark_buffer_dirty(leaf);
708 ret = btrfs_update_block_group(trans, root, disk_bytenr,
709 num_bytes, 1, 0);
710 if (ret)
711 goto fail;
712 } else if (ret != -EEXIST) {
713 goto fail;
716 ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, num_bytes, 0,
717 root->root_key.objectid,
718 objectid, 0);
719 fail:
720 btrfs_release_path(root, &path);
721 return ret;
724 static int add_symbolic_link(struct btrfs_trans_handle *trans,
725 struct btrfs_root *root,
726 u64 objectid, const char *path_name)
728 int ret;
729 u64 sectorsize = root->sectorsize;
730 char *buf = malloc(sectorsize);
732 ret = readlink(path_name, buf, sectorsize);
733 if (ret <= 0) {
734 fprintf(stderr, "readlink failed for %s\n", path_name);
735 goto fail;
737 if (ret >= sectorsize) {
738 fprintf(stderr, "symlink too long for %s", path_name);
739 ret = -1;
740 goto fail;
743 buf[ret] = '\0'; /* readlink does not do it for us */
744 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
745 buf, ret + 1);
746 fail:
747 free(buf);
748 return ret;
751 static int add_file_items(struct btrfs_trans_handle *trans,
752 struct btrfs_root *root,
753 struct btrfs_inode_item *btrfs_inode, u64 objectid,
754 ino_t parent_inum, struct stat *st,
755 const char *path_name, int out_fd)
757 int ret = -1;
758 ssize_t ret_read;
759 u64 bytes_read = 0;
760 char *buffer = NULL;
761 struct btrfs_key key;
762 int blocks;
763 u32 sectorsize = root->sectorsize;
764 u64 first_block = 0;
765 u64 num_blocks = 0;
766 int fd;
768 fd = open(path_name, O_RDONLY);
769 if (fd == -1) {
770 fprintf(stderr, "%s open failed\n", path_name);
771 goto end;
774 blocks = st->st_size / sectorsize;
775 if (st->st_size % sectorsize)
776 blocks += 1;
778 if (st->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
779 buffer = malloc(st->st_size);
780 ret_read = pread64(fd, buffer, st->st_size, bytes_read);
781 if (ret_read == -1) {
782 fprintf(stderr, "%s read failed\n", path_name);
783 goto end;
786 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
787 buffer, st->st_size);
788 goto end;
791 ret = custom_alloc_extent(root, blocks * sectorsize, 0, &key);
792 if (ret)
793 goto end;
795 first_block = key.objectid;
796 bytes_read = 0;
797 buffer = malloc(sectorsize);
799 do {
800 memset(buffer, 0, sectorsize);
801 ret_read = pread64(fd, buffer, sectorsize, bytes_read);
802 if (ret_read == -1) {
803 fprintf(stderr, "%s read failed\n", path_name);
804 goto end;
807 ret = pwrite64(out_fd, buffer, sectorsize,
808 first_block + bytes_read);
809 if (ret != sectorsize) {
810 fprintf(stderr, "output file write failed\n");
811 goto end;
814 /* checksum for file data */
815 ret = btrfs_csum_file_block(trans, root->fs_info->csum_root,
816 first_block + (blocks * sectorsize),
817 first_block + bytes_read,
818 buffer, sectorsize);
819 if (ret) {
820 fprintf(stderr, "%s checksum failed\n", path_name);
821 goto end;
824 bytes_read += ret_read;
825 num_blocks++;
826 } while (ret_read == sectorsize);
828 if (num_blocks > 0) {
829 ret = record_file_extent(trans, root, objectid, btrfs_inode,
830 first_block, first_block,
831 blocks * sectorsize);
832 if (ret)
833 goto end;
836 end:
837 if (buffer)
838 free(buffer);
839 close(fd);
840 return ret;
843 static char *make_path(char *dir, char *name)
845 char *path;
847 path = malloc(strlen(dir) + strlen(name) + 2);
848 if (!path)
849 return NULL;
850 strcpy(path, dir);
851 if (dir[strlen(dir) - 1] != '/')
852 strcat(path, "/");
853 strcat(path, name);
854 return path;
857 static int traverse_directory(struct btrfs_trans_handle *trans,
858 struct btrfs_root *root, char *dir_name,
859 struct directory_name_entry *dir_head, int out_fd)
861 int ret = 0;
863 struct btrfs_inode_item cur_inode;
864 struct btrfs_inode_item *inode_item;
865 int count, i, dir_index_cnt;
866 struct direct **files;
867 struct stat st;
868 struct directory_name_entry *dir_entry, *parent_dir_entry;
869 struct direct *cur_file;
870 ino_t parent_inum, cur_inum;
871 ino_t highest_inum = 0;
872 char *parent_dir_name;
873 struct btrfs_path path;
874 struct extent_buffer *leaf;
875 struct btrfs_key root_dir_key;
876 u64 root_dir_inode_size = 0;
878 /* Add list for source directory */
879 dir_entry = malloc(sizeof(struct directory_name_entry));
880 dir_entry->dir_name = dir_name;
881 dir_entry->path = malloc(strlen(dir_name) + 1);
882 strcpy(dir_entry->path, dir_name);
884 parent_inum = highest_inum + BTRFS_FIRST_FREE_OBJECTID;
885 dir_entry->inum = parent_inum;
886 list_add_tail(&dir_entry->list, &dir_head->list);
888 btrfs_init_path(&path);
890 root_dir_key.objectid = btrfs_root_dirid(&root->root_item);
891 root_dir_key.offset = 0;
892 btrfs_set_key_type(&root_dir_key, BTRFS_INODE_ITEM_KEY);
893 ret = btrfs_lookup_inode(trans, root, &path, &root_dir_key, 1);
894 if (ret) {
895 fprintf(stderr, "root dir lookup error\n");
896 return -1;
899 leaf = path.nodes[0];
900 inode_item = btrfs_item_ptr(leaf, path.slots[0],
901 struct btrfs_inode_item);
903 root_dir_inode_size = calculate_dir_inode_size(dir_name);
904 btrfs_set_inode_size(leaf, inode_item, root_dir_inode_size);
905 btrfs_mark_buffer_dirty(leaf);
907 btrfs_release_path(root, &path);
909 do {
910 parent_dir_entry = list_entry(dir_head->list.next,
911 struct directory_name_entry,
912 list);
913 list_del(&parent_dir_entry->list);
915 parent_inum = parent_dir_entry->inum;
916 parent_dir_name = parent_dir_entry->dir_name;
917 if (chdir(parent_dir_entry->path)) {
918 fprintf(stderr, "chdir error for %s\n",
919 parent_dir_name);
920 goto fail_no_files;
923 count = scandir(parent_dir_entry->path, &files,
924 directory_select, NULL);
925 if (count == -1)
927 fprintf(stderr, "scandir for %s failed: %s\n",
928 parent_dir_name, strerror (errno));
929 goto fail;
932 for (i = 0; i < count; i++) {
933 cur_file = files[i];
935 if (lstat(cur_file->d_name, &st) == -1) {
936 fprintf(stderr, "lstat failed for file %s\n",
937 cur_file->d_name);
938 goto fail;
941 cur_inum = ++highest_inum + BTRFS_FIRST_FREE_OBJECTID;
942 ret = add_directory_items(trans, root,
943 cur_inum, parent_inum,
944 cur_file->d_name,
945 &st, &dir_index_cnt);
946 if (ret) {
947 fprintf(stderr, "add_directory_items failed\n");
948 goto fail;
951 ret = add_inode_items(trans, root, &st,
952 cur_file->d_name, cur_inum,
953 parent_inum, dir_index_cnt,
954 &cur_inode);
955 if (ret) {
956 fprintf(stderr, "add_inode_items failed\n");
957 goto fail;
960 ret = add_xattr_item(trans, root,
961 cur_inum, cur_file->d_name);
962 if (ret) {
963 fprintf(stderr, "add_xattr_item failed\n");
964 if(ret != -ENOTSUP)
965 goto fail;
968 if (S_ISDIR(st.st_mode)) {
969 dir_entry = malloc(sizeof(struct directory_name_entry));
970 dir_entry->dir_name = cur_file->d_name;
971 dir_entry->path = make_path(parent_dir_entry->path,
972 cur_file->d_name);
973 dir_entry->inum = cur_inum;
974 list_add_tail(&dir_entry->list, &dir_head->list);
975 } else if (S_ISREG(st.st_mode)) {
976 ret = add_file_items(trans, root, &cur_inode,
977 cur_inum, parent_inum, &st,
978 cur_file->d_name, out_fd);
979 if (ret) {
980 fprintf(stderr, "add_file_items failed\n");
981 goto fail;
983 } else if (S_ISLNK(st.st_mode)) {
984 ret = add_symbolic_link(trans, root,
985 cur_inum, cur_file->d_name);
986 if (ret) {
987 fprintf(stderr, "add_symbolic_link failed\n");
988 goto fail;
993 free_namelist(files, count);
994 free(parent_dir_entry->path);
995 free(parent_dir_entry);
997 index_cnt = 2;
999 } while (!list_empty(&dir_head->list));
1001 return 0;
1002 fail:
1003 free_namelist(files, count);
1004 fail_no_files:
1005 free(parent_dir_entry->path);
1006 free(parent_dir_entry);
1007 return -1;
1010 static int open_target(char *output_name)
1012 int output_fd;
1013 output_fd = open(output_name, O_CREAT | O_RDWR | O_TRUNC,
1014 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH);
1016 return output_fd;
1019 static int create_chunks(struct btrfs_trans_handle *trans,
1020 struct btrfs_root *root, u64 num_of_meta_chunks,
1021 u64 size_of_data)
1023 u64 chunk_start;
1024 u64 chunk_size;
1025 u64 meta_type = BTRFS_BLOCK_GROUP_METADATA;
1026 u64 data_type = BTRFS_BLOCK_GROUP_DATA;
1027 u64 minimum_data_chunk_size = 8 * 1024 * 1024;
1028 u64 i;
1029 int ret;
1031 for (i = 0; i < num_of_meta_chunks; i++) {
1032 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
1033 &chunk_start, &chunk_size, meta_type);
1034 BUG_ON(ret);
1035 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
1036 meta_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
1037 chunk_start, chunk_size);
1038 BUG_ON(ret);
1039 set_extent_dirty(&root->fs_info->free_space_cache,
1040 chunk_start, chunk_start + chunk_size - 1, 0);
1043 if (size_of_data < minimum_data_chunk_size)
1044 size_of_data = minimum_data_chunk_size;
1045 ret = btrfs_alloc_data_chunk(trans, root->fs_info->extent_root,
1046 &chunk_start, size_of_data, data_type);
1047 BUG_ON(ret);
1048 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
1049 data_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
1050 chunk_start, size_of_data);
1051 BUG_ON(ret);
1052 set_extent_dirty(&root->fs_info->free_space_cache,
1053 chunk_start, chunk_start + size_of_data - 1, 0);
1054 return ret;
1057 static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
1059 int ret;
1060 struct btrfs_trans_handle *trans;
1062 struct stat root_st;
1064 struct directory_name_entry dir_head;
1066 ret = lstat(source_dir, &root_st);
1067 if (ret) {
1068 fprintf(stderr, "unable to lstat the %s\n", source_dir);
1069 goto fail;
1072 INIT_LIST_HEAD(&dir_head.list);
1074 trans = btrfs_start_transaction(root, 1);
1075 ret = traverse_directory(trans, root, source_dir, &dir_head, out_fd);
1076 if (ret) {
1077 fprintf(stderr, "unable to traverse_directory\n");
1078 goto fail;
1080 btrfs_commit_transaction(trans, root);
1082 printf("Making image is completed.\n");
1083 return 0;
1084 fail:
1085 fprintf(stderr, "Making image is aborted.\n");
1086 return -1;
1089 static u64 size_sourcedir(char *dir_name, u64 sectorsize,
1090 u64 *num_of_meta_chunks_ret, u64 *size_of_data_ret)
1092 u64 dir_size = 0;
1093 u64 total_size = 0;
1094 int ret;
1095 char command[1024];
1096 char path[512];
1097 char *file_name = "temp_file";
1098 FILE *file;
1099 u64 default_chunk_size = 8 * 1024 * 1024; /* 8MB */
1100 u64 allocated_meta_size = 8 * 1024 * 1024; /* 8MB */
1101 u64 allocated_total_size = 20 * 1024 * 1024; /* 20MB */
1102 u64 num_of_meta_chunks = 0;
1103 u64 num_of_allocated_meta_chunks =
1104 allocated_meta_size / default_chunk_size;
1106 ret = sprintf(command, "du -B 4096 -s ");
1107 if (ret < 0) {
1108 fprintf(stderr, "error executing sprintf for du command\n");
1109 return -1;
1111 strcat(command, dir_name);
1112 strcat(command, " > ");
1113 strcat(command, file_name);
1114 ret = system(command);
1116 file = fopen(file_name, "r");
1117 ret = fscanf(file, "%lld %s\n", &dir_size, path);
1118 fclose(file);
1119 remove(file_name);
1121 dir_size *= sectorsize;
1122 *size_of_data_ret = dir_size;
1124 num_of_meta_chunks = (dir_size / 2) / default_chunk_size;
1125 if (((dir_size / 2) % default_chunk_size) != 0)
1126 num_of_meta_chunks++;
1127 if (num_of_meta_chunks <= num_of_allocated_meta_chunks)
1128 num_of_meta_chunks = 0;
1129 else
1130 num_of_meta_chunks -= num_of_allocated_meta_chunks;
1132 total_size = allocated_total_size + dir_size +
1133 (num_of_meta_chunks * default_chunk_size);
1135 *num_of_meta_chunks_ret = num_of_meta_chunks;
1137 return total_size;
1140 static int zero_output_file(int out_fd, u64 size, u32 sectorsize)
1142 int len = sectorsize;
1143 int loop_num = size / sectorsize;
1144 u64 location = 0;
1145 char *buf = malloc(len);
1146 int ret = 0, i;
1147 ssize_t written;
1149 if (!buf)
1150 return -ENOMEM;
1151 memset(buf, 0, len);
1152 for (i = 0; i < loop_num; i++) {
1153 written = pwrite64(out_fd, buf, len, location);
1154 if (written != len)
1155 ret = -EIO;
1156 location += sectorsize;
1158 free(buf);
1159 return ret;
1162 int main(int ac, char **av)
1164 char *file;
1165 struct btrfs_root *root;
1166 struct btrfs_trans_handle *trans;
1167 char *label = NULL;
1168 char *first_file;
1169 u64 block_count = 0;
1170 u64 dev_block_count = 0;
1171 u64 blocks[7];
1172 u64 alloc_start = 0;
1173 u64 metadata_profile = BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_DUP;
1174 u64 data_profile = BTRFS_BLOCK_GROUP_RAID0;
1175 u32 leafsize = getpagesize();
1176 u32 sectorsize = 4096;
1177 u32 nodesize = leafsize;
1178 u32 stripesize = 4096;
1179 int zero_end = 1;
1180 int option_index = 0;
1181 int fd;
1182 int ret;
1183 int i;
1184 int mixed = 0;
1185 int data_profile_opt = 0;
1186 int metadata_profile_opt = 0;
1188 char *source_dir = NULL;
1189 int source_dir_set = 0;
1190 u64 num_of_meta_chunks = 0;
1191 u64 size_of_data = 0;
1192 u64 source_dir_size = 0;
1193 char *pretty_buf;
1195 while(1) {
1196 int c;
1197 c = getopt_long(ac, av, "A:b:l:n:s:m:d:L:r:VM", long_options,
1198 &option_index);
1199 if (c < 0)
1200 break;
1201 switch(c) {
1202 case 'A':
1203 alloc_start = parse_size(optarg);
1204 break;
1205 case 'd':
1206 data_profile = parse_profile(optarg);
1207 data_profile_opt = 1;
1208 break;
1209 case 'l':
1210 leafsize = parse_size(optarg);
1211 break;
1212 case 'L':
1213 label = parse_label(optarg);
1214 break;
1215 case 'm':
1216 metadata_profile = parse_profile(optarg);
1217 metadata_profile_opt = 1;
1218 break;
1219 case 'M':
1220 mixed = 1;
1221 break;
1222 case 'n':
1223 nodesize = parse_size(optarg);
1224 break;
1225 case 's':
1226 sectorsize = parse_size(optarg);
1227 break;
1228 case 'b':
1229 block_count = parse_size(optarg);
1230 if (block_count <= 1024*1024*1024) {
1231 printf("SMALL VOLUME: forcing mixed "
1232 "metadata/data groups\n");
1233 mixed = 1;
1235 zero_end = 0;
1236 break;
1237 case 'V':
1238 print_version();
1239 break;
1240 case 'r':
1241 source_dir = optarg;
1242 source_dir_set = 1;
1243 break;
1244 default:
1245 print_usage();
1248 sectorsize = max(sectorsize, (u32)getpagesize());
1249 if (leafsize < sectorsize || (leafsize & (sectorsize - 1))) {
1250 fprintf(stderr, "Illegal leafsize %u\n", leafsize);
1251 exit(1);
1253 if (nodesize < sectorsize || (nodesize & (sectorsize - 1))) {
1254 fprintf(stderr, "Illegal nodesize %u\n", nodesize);
1255 exit(1);
1257 ac = ac - optind;
1258 if (ac == 0)
1259 print_usage();
1261 printf("\nWARNING! - %s IS EXPERIMENTAL\n", BTRFS_BUILD_VERSION);
1262 printf("WARNING! - see http://btrfs.wiki.kernel.org before using\n\n");
1264 if (source_dir == 0) {
1265 file = av[optind++];
1266 ret = check_mounted(file);
1267 if (ret < 0) {
1268 fprintf(stderr, "error checking %s mount status\n", file);
1269 exit(1);
1271 if (ret == 1) {
1272 fprintf(stderr, "%s is mounted\n", file);
1273 exit(1);
1275 ac--;
1276 fd = open(file, O_RDWR);
1277 if (fd < 0) {
1278 fprintf(stderr, "unable to open %s\n", file);
1279 exit(1);
1281 first_file = file;
1282 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count, &mixed);
1283 if (block_count == 0)
1284 block_count = dev_block_count;
1285 } else {
1286 ac = 0;
1287 file = av[optind++];
1288 fd = open_target(file);
1289 if (fd < 0) {
1290 fprintf(stderr, "unable to open the %s\n", file);
1291 exit(1);
1294 first_file = file;
1295 source_dir_size = size_sourcedir(source_dir, sectorsize,
1296 &num_of_meta_chunks, &size_of_data);
1297 if(block_count < source_dir_size)
1298 block_count = source_dir_size;
1299 ret = zero_output_file(fd, block_count, sectorsize);
1300 if (ret) {
1301 fprintf(stderr, "unable to zero the output file\n");
1302 exit(1);
1305 if (mixed) {
1306 if (!metadata_profile_opt)
1307 metadata_profile = 0;
1308 if (!data_profile_opt)
1309 data_profile = 0;
1311 if (metadata_profile != data_profile) {
1312 fprintf(stderr, "With mixed block groups data and metadata "
1313 "profiles must be the same\n");
1314 exit(1);
1318 blocks[0] = BTRFS_SUPER_INFO_OFFSET;
1319 for (i = 1; i < 7; i++) {
1320 blocks[i] = BTRFS_SUPER_INFO_OFFSET + 1024 * 1024 +
1321 leafsize * i;
1324 ret = make_btrfs(fd, file, label, blocks, block_count,
1325 nodesize, leafsize,
1326 sectorsize, stripesize);
1327 if (ret) {
1328 fprintf(stderr, "error during mkfs %d\n", ret);
1329 exit(1);
1331 root = open_ctree(file, 0, O_RDWR);
1332 root->fs_info->alloc_start = alloc_start;
1334 ret = make_root_dir(root, mixed);
1335 if (ret) {
1336 fprintf(stderr, "failed to setup the root directory\n");
1337 exit(1);
1340 trans = btrfs_start_transaction(root, 1);
1342 if (ac == 0)
1343 goto raid_groups;
1345 btrfs_register_one_device(file);
1346 if (!root) {
1347 fprintf(stderr, "ctree init failed\n");
1348 return -1;
1351 zero_end = 1;
1352 while(ac-- > 0) {
1353 int old_mixed = mixed;
1355 file = av[optind++];
1356 ret = check_mounted(file);
1357 if (ret < 0) {
1358 fprintf(stderr, "error checking %s mount status\n",
1359 file);
1360 exit(1);
1362 if (ret == 1) {
1363 fprintf(stderr, "%s is mounted\n", file);
1364 exit(1);
1366 fd = open(file, O_RDWR);
1367 if (fd < 0) {
1368 fprintf(stderr, "unable to open %s\n", file);
1369 exit(1);
1371 ret = btrfs_device_already_in_root(root, fd,
1372 BTRFS_SUPER_INFO_OFFSET);
1373 if (ret) {
1374 fprintf(stderr, "skipping duplicate device %s in FS\n",
1375 file);
1376 close(fd);
1377 continue;
1379 ret = btrfs_prepare_device(fd, file, zero_end,
1380 &dev_block_count, &mixed);
1381 mixed = old_mixed;
1382 BUG_ON(ret);
1384 ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
1385 sectorsize, sectorsize, sectorsize);
1386 BUG_ON(ret);
1387 btrfs_register_one_device(file);
1390 raid_groups:
1391 if (!source_dir_set) {
1392 ret = create_raid_groups(trans, root, data_profile,
1393 metadata_profile, mixed);
1394 BUG_ON(ret);
1397 ret = create_data_reloc_tree(trans, root);
1398 BUG_ON(ret);
1400 if (mixed) {
1401 struct btrfs_super_block *super = &root->fs_info->super_copy;
1402 u64 flags = btrfs_super_incompat_flags(super);
1404 flags |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
1405 btrfs_set_super_incompat_flags(super, flags);
1408 printf("fs created label %s on %s\n\tnodesize %u leafsize %u "
1409 "sectorsize %u size %s\n",
1410 label, first_file, nodesize, leafsize, sectorsize,
1411 pretty_buf = pretty_sizes(btrfs_super_total_bytes(&root->fs_info->super_copy)));
1412 free(pretty_buf);
1414 printf("%s\n", BTRFS_BUILD_VERSION);
1415 btrfs_commit_transaction(trans, root);
1417 if (source_dir_set) {
1418 trans = btrfs_start_transaction(root, 1);
1419 ret = create_chunks(trans, root,
1420 num_of_meta_chunks, size_of_data);
1421 BUG_ON(ret);
1422 btrfs_commit_transaction(trans, root);
1424 ret = make_image(source_dir, root, fd);
1425 BUG_ON(ret);
1428 ret = close_ctree(root);
1429 BUG_ON(ret);
1431 free(label);
1432 return 0;