2 * Copyright (C) 2007 Oracle. All rights reserved.
3 * Copyright (C) 2008 Morey Roof. All rights reserved.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public
7 * License v2 as published by the Free Software Foundation.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * General Public License for more details.
14 * You should have received a copy of the GNU General Public
15 * License along with this program; if not, write to the
16 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
17 * Boston, MA 021110-1307, USA.
20 #define _XOPEN_SOURCE 700
21 #define __USE_XOPEN2K8
22 #define __XOPEN2K8 /* due to an error in dirent.h, to get dirfd() */
23 #define _GNU_SOURCE /* O_NOATIME */
27 #include <sys/ioctl.h>
28 #include <sys/mount.h>
29 #include <sys/types.h>
31 #include <uuid/uuid.h>
36 #include <linux/loop.h>
37 #include <linux/major.h>
38 #include <linux/kdev_t.h>
40 #include <blkid/blkid.h>
41 #include "kerncompat.h"
42 #include "radix-tree.h"
45 #include "transaction.h"
52 #define BLKDISCARD _IO(0x12,119)
55 static char argv0_buf[ARGV0_BUF_SIZE] = "btrfs";
57 void fixup_argv0(char **argv, const char *token)
59 int len = strlen(argv0_buf);
61 snprintf(argv0_buf + len, sizeof(argv0_buf) - len, " %s", token);
65 void set_argv0(char **argv)
67 sprintf(argv0_buf, "%s", argv[0]);
70 int check_argc_exact(int nargs, int expected)
73 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
75 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
77 return nargs != expected;
80 int check_argc_min(int nargs, int expected)
82 if (nargs < expected) {
83 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
90 int check_argc_max(int nargs, int expected)
92 if (nargs > expected) {
93 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
102 * Discard the given range in one go
104 static int discard_range(int fd, u64 start, u64 len)
106 u64 range[2] = { start, len };
108 if (ioctl(fd, BLKDISCARD, &range) < 0)
114 * Discard blocks in the given range in 1G chunks, the process is interruptible
116 static int discard_blocks(int fd, u64 start, u64 len)
120 u64 chunk_size = min_t(u64, len, 1*1024*1024*1024);
123 ret = discard_range(fd, start, chunk_size);
133 static u64 reference_root_table[] = {
134 [1] = BTRFS_ROOT_TREE_OBJECTID,
135 [2] = BTRFS_EXTENT_TREE_OBJECTID,
136 [3] = BTRFS_CHUNK_TREE_OBJECTID,
137 [4] = BTRFS_DEV_TREE_OBJECTID,
138 [5] = BTRFS_FS_TREE_OBJECTID,
139 [6] = BTRFS_CSUM_TREE_OBJECTID,
142 int test_uuid_unique(char *fs_uuid)
145 blkid_dev_iterate iter = NULL;
146 blkid_dev dev = NULL;
147 blkid_cache cache = NULL;
149 if (blkid_get_cache(&cache, 0) < 0) {
150 printf("ERROR: lblkid cache get failed\n");
153 blkid_probe_all(cache);
154 iter = blkid_dev_iterate_begin(cache);
155 blkid_dev_set_search(iter, "UUID", fs_uuid);
157 while (blkid_dev_next(iter, &dev) == 0) {
158 dev = blkid_verify(cache, dev);
165 blkid_dev_iterate_end(iter);
166 blkid_put_cache(cache);
171 int make_btrfs(int fd, const char *device, const char *label, char *fs_uuid,
172 u64 blocks[7], u64 num_bytes, u32 nodesize,
173 u32 leafsize, u32 sectorsize, u32 stripesize, u64 features)
175 struct btrfs_super_block super;
176 struct extent_buffer *buf = NULL;
177 struct btrfs_root_item root_item;
178 struct btrfs_disk_key disk_key;
179 struct btrfs_extent_item *extent_item;
180 struct btrfs_inode_item *inode_item;
181 struct btrfs_chunk *chunk;
182 struct btrfs_dev_item *dev_item;
183 struct btrfs_dev_extent *dev_extent;
184 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
194 int skinny_metadata = !!(features &
195 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
197 first_free = BTRFS_SUPER_INFO_OFFSET + sectorsize * 2 - 1;
198 first_free &= ~((u64)sectorsize - 1);
200 memset(&super, 0, sizeof(super));
202 num_bytes = (num_bytes / sectorsize) * sectorsize;
204 if (uuid_parse(fs_uuid, super.fsid) != 0) {
205 fprintf(stderr, "could not parse UUID: %s\n", fs_uuid);
209 if (!test_uuid_unique(fs_uuid)) {
210 fprintf(stderr, "non-unique UUID: %s\n", fs_uuid);
215 uuid_generate(super.fsid);
217 uuid_generate(super.dev_item.uuid);
218 uuid_generate(chunk_tree_uuid);
220 btrfs_set_super_bytenr(&super, blocks[0]);
221 btrfs_set_super_num_devices(&super, 1);
222 btrfs_set_super_magic(&super, BTRFS_MAGIC);
223 btrfs_set_super_generation(&super, 1);
224 btrfs_set_super_root(&super, blocks[1]);
225 btrfs_set_super_chunk_root(&super, blocks[3]);
226 btrfs_set_super_total_bytes(&super, num_bytes);
227 btrfs_set_super_bytes_used(&super, 6 * leafsize);
228 btrfs_set_super_sectorsize(&super, sectorsize);
229 btrfs_set_super_leafsize(&super, leafsize);
230 btrfs_set_super_nodesize(&super, nodesize);
231 btrfs_set_super_stripesize(&super, stripesize);
232 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
233 btrfs_set_super_chunk_root_generation(&super, 1);
234 btrfs_set_super_cache_generation(&super, -1);
235 btrfs_set_super_incompat_flags(&super, features);
237 strncpy(super.label, label, BTRFS_LABEL_SIZE - 1);
239 buf = malloc(sizeof(*buf) + max(sectorsize, leafsize));
241 /* create the tree of root objects */
242 memset(buf->data, 0, leafsize);
244 btrfs_set_header_bytenr(buf, blocks[1]);
245 btrfs_set_header_nritems(buf, 4);
246 btrfs_set_header_generation(buf, 1);
247 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
248 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
249 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
252 write_extent_buffer(buf, chunk_tree_uuid,
253 btrfs_header_chunk_tree_uuid(buf),
256 /* create the items for the root tree */
257 memset(&root_item, 0, sizeof(root_item));
258 inode_item = &root_item.inode;
259 btrfs_set_stack_inode_generation(inode_item, 1);
260 btrfs_set_stack_inode_size(inode_item, 3);
261 btrfs_set_stack_inode_nlink(inode_item, 1);
262 btrfs_set_stack_inode_nbytes(inode_item, leafsize);
263 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
264 btrfs_set_root_refs(&root_item, 1);
265 btrfs_set_root_used(&root_item, leafsize);
266 btrfs_set_root_generation(&root_item, 1);
268 memset(&disk_key, 0, sizeof(disk_key));
269 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
270 btrfs_set_disk_key_offset(&disk_key, 0);
273 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - sizeof(root_item);
274 btrfs_set_root_bytenr(&root_item, blocks[2]);
275 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
276 btrfs_set_item_key(buf, &disk_key, nritems);
277 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
278 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
280 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
281 nritems), sizeof(root_item));
284 itemoff = itemoff - sizeof(root_item);
285 btrfs_set_root_bytenr(&root_item, blocks[4]);
286 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
287 btrfs_set_item_key(buf, &disk_key, nritems);
288 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
289 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
291 write_extent_buffer(buf, &root_item,
292 btrfs_item_ptr_offset(buf, nritems),
296 itemoff = itemoff - sizeof(root_item);
297 btrfs_set_root_bytenr(&root_item, blocks[5]);
298 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
299 btrfs_set_item_key(buf, &disk_key, nritems);
300 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
301 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
303 write_extent_buffer(buf, &root_item,
304 btrfs_item_ptr_offset(buf, nritems),
308 itemoff = itemoff - sizeof(root_item);
309 btrfs_set_root_bytenr(&root_item, blocks[6]);
310 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
311 btrfs_set_item_key(buf, &disk_key, nritems);
312 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
313 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
315 write_extent_buffer(buf, &root_item,
316 btrfs_item_ptr_offset(buf, nritems),
321 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
322 ret = pwrite(fd, buf->data, leafsize, blocks[1]);
323 if (ret != leafsize) {
324 ret = (ret < 0 ? -errno : -EIO);
328 /* create the items for the extent tree */
329 memset(buf->data+sizeof(struct btrfs_header), 0,
330 leafsize-sizeof(struct btrfs_header));
332 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize);
333 for (i = 1; i < 7; i++) {
334 item_size = sizeof(struct btrfs_extent_item);
335 if (!skinny_metadata)
336 item_size += sizeof(struct btrfs_tree_block_info);
338 BUG_ON(blocks[i] < first_free);
339 BUG_ON(blocks[i] < blocks[i - 1]);
341 /* create extent item */
342 itemoff -= item_size;
343 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
344 if (skinny_metadata) {
345 btrfs_set_disk_key_type(&disk_key,
346 BTRFS_METADATA_ITEM_KEY);
347 btrfs_set_disk_key_offset(&disk_key, 0);
349 btrfs_set_disk_key_type(&disk_key,
350 BTRFS_EXTENT_ITEM_KEY);
351 btrfs_set_disk_key_offset(&disk_key, leafsize);
353 btrfs_set_item_key(buf, &disk_key, nritems);
354 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
356 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
358 extent_item = btrfs_item_ptr(buf, nritems,
359 struct btrfs_extent_item);
360 btrfs_set_extent_refs(buf, extent_item, 1);
361 btrfs_set_extent_generation(buf, extent_item, 1);
362 btrfs_set_extent_flags(buf, extent_item,
363 BTRFS_EXTENT_FLAG_TREE_BLOCK);
366 /* create extent ref */
367 ref_root = reference_root_table[i];
368 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
369 btrfs_set_disk_key_offset(&disk_key, ref_root);
370 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
371 btrfs_set_item_key(buf, &disk_key, nritems);
372 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
374 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
377 btrfs_set_header_bytenr(buf, blocks[2]);
378 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
379 btrfs_set_header_nritems(buf, nritems);
380 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
381 ret = pwrite(fd, buf->data, leafsize, blocks[2]);
382 if (ret != leafsize) {
383 ret = (ret < 0 ? -errno : -EIO);
387 /* create the chunk tree */
388 memset(buf->data+sizeof(struct btrfs_header), 0,
389 leafsize-sizeof(struct btrfs_header));
391 item_size = sizeof(*dev_item);
392 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - item_size;
394 /* first device 1 (there is no device 0) */
395 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
396 btrfs_set_disk_key_offset(&disk_key, 1);
397 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
398 btrfs_set_item_key(buf, &disk_key, nritems);
399 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
400 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
402 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
403 btrfs_set_device_id(buf, dev_item, 1);
404 btrfs_set_device_generation(buf, dev_item, 0);
405 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
406 btrfs_set_device_bytes_used(buf, dev_item,
407 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
408 btrfs_set_device_io_align(buf, dev_item, sectorsize);
409 btrfs_set_device_io_width(buf, dev_item, sectorsize);
410 btrfs_set_device_sector_size(buf, dev_item, sectorsize);
411 btrfs_set_device_type(buf, dev_item, 0);
413 write_extent_buffer(buf, super.dev_item.uuid,
414 (unsigned long)btrfs_device_uuid(dev_item),
416 write_extent_buffer(buf, super.fsid,
417 (unsigned long)btrfs_device_fsid(dev_item),
419 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
423 item_size = btrfs_chunk_item_size(1);
424 itemoff = itemoff - item_size;
426 /* then we have chunk 0 */
427 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
428 btrfs_set_disk_key_offset(&disk_key, 0);
429 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
430 btrfs_set_item_key(buf, &disk_key, nritems);
431 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
432 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
434 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
435 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
436 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
437 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
438 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
439 btrfs_set_chunk_io_align(buf, chunk, sectorsize);
440 btrfs_set_chunk_io_width(buf, chunk, sectorsize);
441 btrfs_set_chunk_sector_size(buf, chunk, sectorsize);
442 btrfs_set_chunk_num_stripes(buf, chunk, 1);
443 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
444 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
447 write_extent_buffer(buf, super.dev_item.uuid,
448 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
451 /* copy the key for the chunk to the system array */
452 ptr = super.sys_chunk_array;
453 array_size = sizeof(disk_key);
455 memcpy(ptr, &disk_key, sizeof(disk_key));
456 ptr += sizeof(disk_key);
458 /* copy the chunk to the system array */
459 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
460 array_size += item_size;
462 btrfs_set_super_sys_array_size(&super, array_size);
464 btrfs_set_header_bytenr(buf, blocks[3]);
465 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
466 btrfs_set_header_nritems(buf, nritems);
467 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
468 ret = pwrite(fd, buf->data, leafsize, blocks[3]);
469 if (ret != leafsize) {
470 ret = (ret < 0 ? -errno : -EIO);
474 /* create the device tree */
475 memset(buf->data+sizeof(struct btrfs_header), 0,
476 leafsize-sizeof(struct btrfs_header));
478 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) -
479 sizeof(struct btrfs_dev_extent);
481 btrfs_set_disk_key_objectid(&disk_key, 1);
482 btrfs_set_disk_key_offset(&disk_key, 0);
483 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
484 btrfs_set_item_key(buf, &disk_key, nritems);
485 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
486 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
487 sizeof(struct btrfs_dev_extent));
488 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
489 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
490 BTRFS_CHUNK_TREE_OBJECTID);
491 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
492 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
493 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
495 write_extent_buffer(buf, chunk_tree_uuid,
496 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
499 btrfs_set_dev_extent_length(buf, dev_extent,
500 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
503 btrfs_set_header_bytenr(buf, blocks[4]);
504 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
505 btrfs_set_header_nritems(buf, nritems);
506 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
507 ret = pwrite(fd, buf->data, leafsize, blocks[4]);
508 if (ret != leafsize) {
509 ret = (ret < 0 ? -errno : -EIO);
513 /* create the FS root */
514 memset(buf->data+sizeof(struct btrfs_header), 0,
515 leafsize-sizeof(struct btrfs_header));
516 btrfs_set_header_bytenr(buf, blocks[5]);
517 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
518 btrfs_set_header_nritems(buf, 0);
519 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
520 ret = pwrite(fd, buf->data, leafsize, blocks[5]);
521 if (ret != leafsize) {
522 ret = (ret < 0 ? -errno : -EIO);
525 /* finally create the csum root */
526 memset(buf->data+sizeof(struct btrfs_header), 0,
527 leafsize-sizeof(struct btrfs_header));
528 btrfs_set_header_bytenr(buf, blocks[6]);
529 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
530 btrfs_set_header_nritems(buf, 0);
531 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
532 ret = pwrite(fd, buf->data, leafsize, blocks[6]);
533 if (ret != leafsize) {
534 ret = (ret < 0 ? -errno : -EIO);
538 /* and write out the super block */
539 BUG_ON(sizeof(super) > sectorsize);
540 memset(buf->data, 0, sectorsize);
541 memcpy(buf->data, &super, sizeof(super));
542 buf->len = sectorsize;
543 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
544 ret = pwrite(fd, buf->data, sectorsize, blocks[0]);
545 if (ret != sectorsize) {
546 ret = (ret < 0 ? -errno : -EIO);
557 u64 btrfs_device_size(int fd, struct stat *st)
560 if (S_ISREG(st->st_mode)) {
563 if (!S_ISBLK(st->st_mode)) {
566 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
572 static int zero_blocks(int fd, off_t start, size_t len)
574 char *buf = malloc(len);
581 written = pwrite(fd, buf, len, start);
588 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
590 /* don't write outside the device by clamping the region to the device size */
591 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
593 off_t end = max(start, start + len);
596 /* and don't overwrite the disk labels on sparc */
597 start = max(start, 1024);
598 end = max(end, 1024);
601 start = min_t(u64, start, dev_size);
602 end = min_t(u64, end, dev_size);
604 return zero_blocks(fd, start, end - start);
607 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
608 struct btrfs_root *root, int fd, char *path,
609 u64 block_count, u32 io_width, u32 io_align,
612 struct btrfs_super_block *disk_super;
613 struct btrfs_super_block *super = root->fs_info->super_copy;
614 struct btrfs_device *device;
615 struct btrfs_dev_item *dev_item;
621 device = kzalloc(sizeof(*device), GFP_NOFS);
624 buf = kmalloc(sectorsize, GFP_NOFS);
629 BUG_ON(sizeof(*disk_super) > sectorsize);
630 memset(buf, 0, sectorsize);
632 disk_super = (struct btrfs_super_block *)buf;
633 dev_item = &disk_super->dev_item;
635 uuid_generate(device->uuid);
638 device->io_width = io_width;
639 device->io_align = io_align;
640 device->sector_size = sectorsize;
642 device->writeable = 1;
643 device->total_bytes = block_count;
644 device->bytes_used = 0;
645 device->total_ios = 0;
646 device->dev_root = root->fs_info->dev_root;
648 ret = btrfs_add_device(trans, root, device);
651 total_bytes = btrfs_super_total_bytes(super) + block_count;
652 btrfs_set_super_total_bytes(super, total_bytes);
654 num_devs = btrfs_super_num_devices(super) + 1;
655 btrfs_set_super_num_devices(super, num_devs);
657 memcpy(disk_super, super, sizeof(*disk_super));
659 printf("adding device %s id %llu\n", path,
660 (unsigned long long)device->devid);
662 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
663 btrfs_set_stack_device_id(dev_item, device->devid);
664 btrfs_set_stack_device_type(dev_item, device->type);
665 btrfs_set_stack_device_io_align(dev_item, device->io_align);
666 btrfs_set_stack_device_io_width(dev_item, device->io_width);
667 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
668 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
669 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
670 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
672 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
673 BUG_ON(ret != sectorsize);
676 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
677 device->fs_devices = root->fs_info->fs_devices;
681 int btrfs_prepare_device(int fd, char *file, int zero_end, u64 *block_count_ret,
682 u64 max_block_count, int *mixed, int discard)
688 ret = fstat(fd, &st);
690 fprintf(stderr, "unable to stat %s\n", file);
694 block_count = btrfs_device_size(fd, &st);
695 if (block_count == 0) {
696 fprintf(stderr, "unable to find %s size\n", file);
700 block_count = min(block_count, max_block_count);
702 if (block_count < BTRFS_MKFS_SMALL_VOLUME_SIZE && !(*mixed))
707 * We intentionally ignore errors from the discard ioctl. It
708 * is not necessary for the mkfs functionality but just an
711 if (discard_range(fd, 0, 0) == 0) {
712 fprintf(stderr, "Performing full device TRIM (%s) ...\n",
713 pretty_size(block_count));
714 discard_blocks(fd, 0, block_count);
718 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
719 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
720 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
721 BTRFS_SUPER_INFO_SIZE, block_count);
722 if (!ret && zero_end)
723 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
724 ZERO_DEV_BYTES, block_count);
727 fprintf(stderr, "ERROR: failed to zero device '%s' - %s\n",
728 file, strerror(-ret));
732 *block_count_ret = block_count;
736 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
737 struct btrfs_root *root, u64 objectid)
740 struct btrfs_inode_item inode_item;
741 time_t now = time(NULL);
743 memset(&inode_item, 0, sizeof(inode_item));
744 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
745 btrfs_set_stack_inode_size(&inode_item, 0);
746 btrfs_set_stack_inode_nlink(&inode_item, 1);
747 btrfs_set_stack_inode_nbytes(&inode_item, root->leafsize);
748 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
749 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
750 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
751 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
752 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
753 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
754 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
755 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
756 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
758 if (root->fs_info->tree_root == root)
759 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
761 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
765 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
769 btrfs_set_root_dirid(&root->root_item, objectid);
776 * checks if a path is a block device node
777 * Returns negative errno on failure, otherwise
778 * returns 1 for blockdev, 0 for not-blockdev
780 int is_block_device(const char *path)
784 if (stat(path, &statbuf) < 0)
787 return S_ISBLK(statbuf.st_mode);
791 * check if given path is a mount point
792 * return 1 if yes. 0 if no. -1 for error
794 int is_mount_point(const char *path)
800 f = setmntent("/proc/self/mounts", "r");
804 while ((mnt = getmntent(f)) != NULL) {
805 if (strcmp(mnt->mnt_dir, path))
815 * Find the mount point for a mounted device.
816 * On success, returns 0 with mountpoint in *mp.
817 * On failure, returns -errno (not mounted yields -EINVAL)
818 * Is noisy on failures, expects to be given a mounted device.
820 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
825 ret = is_block_device(dev);
828 fprintf(stderr, "%s is not a block device\n", dev);
831 fprintf(stderr, "Could not check %s: %s\n",
832 dev, strerror(-ret));
837 fd = open(dev, O_RDONLY);
840 fprintf(stderr, "Could not open %s: %s\n", dev, strerror(errno));
844 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
847 } else { /* mounted, all good */
857 * Given a pathname, return a filehandle to:
858 * the original pathname or,
859 * if the pathname is a mounted btrfs device, to its mountpoint.
861 * On error, return -1, errno should be set.
863 int open_path_or_dev_mnt(const char *path, DIR **dirstream)
865 char mp[BTRFS_PATH_NAME_MAX + 1];
868 if (is_block_device(path)) {
871 ret = get_btrfs_mount(path, mp, sizeof(mp));
873 /* not a mounted btrfs dev */
877 fdmnt = open_file_or_dir(mp, dirstream);
879 fdmnt = open_file_or_dir(path, dirstream);
885 /* checks if a device is a loop device */
886 static int is_loop_device (const char* device) {
889 if(stat(device, &statbuf) < 0)
892 return (S_ISBLK(statbuf.st_mode) &&
893 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
897 /* Takes a loop device path (e.g. /dev/loop0) and returns
898 * the associated file (e.g. /images/my_btrfs.img) */
899 static int resolve_loop_device(const char* loop_dev, char* loop_file,
906 char real_loop_dev[PATH_MAX];
908 if (!realpath(loop_dev, real_loop_dev))
910 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
911 if (!(f = fopen(p, "r")))
914 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
915 ret = fscanf(f, fmt, loop_file);
923 /* Checks whether a and b are identical or device
924 * files associated with the same block device
926 static int is_same_blk_file(const char* a, const char* b)
928 struct stat st_buf_a, st_buf_b;
929 char real_a[PATH_MAX];
930 char real_b[PATH_MAX];
932 if(!realpath(a, real_a))
935 if (!realpath(b, real_b))
938 /* Identical path? */
939 if(strcmp(real_a, real_b) == 0)
942 if(stat(a, &st_buf_a) < 0 ||
943 stat(b, &st_buf_b) < 0)
950 /* Same blockdevice? */
951 if(S_ISBLK(st_buf_a.st_mode) &&
952 S_ISBLK(st_buf_b.st_mode) &&
953 st_buf_a.st_rdev == st_buf_b.st_rdev)
959 if (st_buf_a.st_dev == st_buf_b.st_dev &&
960 st_buf_a.st_ino == st_buf_b.st_ino)
968 /* checks if a and b are identical or device
969 * files associated with the same block device or
970 * if one file is a loop device that uses the other
973 static int is_same_loop_file(const char* a, const char* b)
975 char res_a[PATH_MAX];
976 char res_b[PATH_MAX];
977 const char* final_a = NULL;
978 const char* final_b = NULL;
981 /* Resolve a if it is a loop device */
982 if((ret = is_loop_device(a)) < 0) {
987 ret = resolve_loop_device(a, res_a, sizeof(res_a));
998 /* Resolve b if it is a loop device */
999 if ((ret = is_loop_device(b)) < 0) {
1004 ret = resolve_loop_device(b, res_b, sizeof(res_b));
1015 return is_same_blk_file(final_a, final_b);
1018 /* Checks if a file exists and is a block or regular file*/
1019 static int is_existing_blk_or_reg_file(const char* filename)
1023 if(stat(filename, &st_buf) < 0) {
1030 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
1033 /* Checks if a file is used (directly or indirectly via a loop device)
1034 * by a device in fs_devices
1036 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
1040 struct list_head *head;
1041 struct list_head *cur;
1042 struct btrfs_device *device;
1044 head = &fs_devices->devices;
1045 list_for_each(cur, head) {
1046 device = list_entry(cur, struct btrfs_device, dev_list);
1048 if((ret = is_same_loop_file(device->name, file)))
1056 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1057 * Returns NULL on invalid input or malloc failure; Other failures
1058 * will be handled by the caller using the input pathame.
1060 char *canonicalize_dm_name(const char *ptname)
1064 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
1066 if (!ptname || !*ptname)
1069 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
1070 if (!(f = fopen(path, "r")))
1073 /* read <name>\n from sysfs */
1074 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
1075 name[sz - 1] = '\0';
1076 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
1078 if (access(path, F_OK) == 0)
1086 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
1087 * to a device mapper pathname.
1088 * Returns NULL on invalid input or malloc failure; Other failures
1089 * will be handled by the caller using the input pathame.
1091 char *canonicalize_path(const char *path)
1093 char *canonical, *p;
1095 if (!path || !*path)
1098 canonical = realpath(path, NULL);
1100 return strdup(path);
1101 p = strrchr(canonical, '/');
1102 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
1103 char *dm = canonicalize_dm_name(p + 1);
1114 * returns 1 if the device was mounted, < 0 on error or 0 if everything
1115 * is safe to continue.
1117 int check_mounted(const char* file)
1122 fd = open(file, O_RDONLY);
1124 fprintf (stderr, "check_mounted(): Could not open %s\n", file);
1128 ret = check_mounted_where(fd, file, NULL, 0, NULL);
1134 int check_mounted_where(int fd, const char *file, char *where, int size,
1135 struct btrfs_fs_devices **fs_dev_ret)
1140 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1144 /* scan the initial device */
1145 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
1146 &total_devs, BTRFS_SUPER_INFO_OFFSET, 0);
1147 is_btrfs = (ret >= 0);
1149 /* scan other devices */
1150 if (is_btrfs && total_devs > 1) {
1151 if ((ret = scan_for_btrfs(BTRFS_SCAN_PROC, !BTRFS_UPDATE_KERNEL)))
1155 /* iterate over the list of currently mountes filesystems */
1156 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
1159 while ((mnt = getmntent (f)) != NULL) {
1161 if(strcmp(mnt->mnt_type, "btrfs") != 0)
1164 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
1166 /* ignore entries in the mount table that are not
1167 associated with a file*/
1168 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
1169 goto out_mntloop_err;
1173 ret = is_same_loop_file(file, mnt->mnt_fsname);
1177 goto out_mntloop_err;
1182 /* Did we find an entry in mnt table? */
1183 if (mnt && size && where) {
1184 strncpy(where, mnt->mnt_dir, size);
1188 *fs_dev_ret = fs_devices_mnt;
1190 ret = (mnt != NULL);
1198 struct pending_dir {
1199 struct list_head list;
1200 char name[PATH_MAX];
1203 void btrfs_register_one_device(char *fname)
1205 struct btrfs_ioctl_vol_args args;
1210 fd = open("/dev/btrfs-control", O_RDONLY);
1212 fprintf(stderr, "failed to open /dev/btrfs-control "
1213 "skipping device registration: %s\n",
1217 strncpy(args.name, fname, BTRFS_PATH_NAME_MAX);
1218 args.name[BTRFS_PATH_NAME_MAX-1] = 0;
1219 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
1222 fprintf(stderr, "ERROR: device scan failed '%s' - %s\n",
1223 fname, strerror(e));
1228 int btrfs_scan_one_dir(char *dirname, int run_ioctl)
1231 struct dirent *dirent;
1232 struct pending_dir *pending;
1238 struct list_head pending_list;
1239 struct btrfs_fs_devices *tmp_devices;
1242 INIT_LIST_HEAD(&pending_list);
1244 pending = malloc(sizeof(*pending));
1247 strcpy(pending->name, dirname);
1250 dirname_len = strlen(pending->name);
1251 fullpath = malloc(PATH_MAX);
1252 dirname = pending->name;
1258 dirp = opendir(dirname);
1260 fprintf(stderr, "Unable to open %s for scanning\n", dirname);
1265 dirent = readdir(dirp);
1268 if (dirent->d_name[0] == '.')
1270 if (dirname_len + strlen(dirent->d_name) + 2 > PATH_MAX) {
1274 snprintf(fullpath, PATH_MAX, "%s/%s", dirname, dirent->d_name);
1275 ret = lstat(fullpath, &st);
1277 fprintf(stderr, "failed to stat %s\n", fullpath);
1280 if (S_ISLNK(st.st_mode))
1282 if (S_ISDIR(st.st_mode)) {
1283 struct pending_dir *next = malloc(sizeof(*next));
1288 strcpy(next->name, fullpath);
1289 list_add_tail(&next->list, &pending_list);
1291 if (!S_ISBLK(st.st_mode)) {
1294 fd = open(fullpath, O_RDONLY);
1296 /* ignore the following errors:
1297 ENXIO (device don't exists)
1298 ENOMEDIUM (No medium found ->
1299 like a cd tray empty)
1301 if(errno != ENXIO && errno != ENOMEDIUM)
1302 fprintf(stderr, "failed to read %s: %s\n",
1303 fullpath, strerror(errno));
1306 ret = btrfs_scan_one_device(fd, fullpath, &tmp_devices,
1308 BTRFS_SUPER_INFO_OFFSET, 0);
1309 if (ret == 0 && run_ioctl > 0) {
1310 btrfs_register_one_device(fullpath);
1314 if (!list_empty(&pending_list)) {
1316 pending = list_entry(pending_list.next, struct pending_dir,
1319 list_del(&pending->list);
1328 while (!list_empty(&pending_list)) {
1329 pending = list_entry(pending_list.next, struct pending_dir,
1331 list_del(&pending->list);
1339 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
1342 struct btrfs_super_block *disk_super;
1346 buf = malloc(BTRFS_SUPER_INFO_SIZE);
1351 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
1352 if (ret != BTRFS_SUPER_INFO_SIZE)
1356 disk_super = (struct btrfs_super_block *)buf;
1357 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
1360 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
1369 static char *size_strs[] = { "", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
1370 int pretty_size_snprintf(u64 size, char *str, size_t str_bytes)
1382 u64 last_size = size;
1384 while(size >= 1024){
1390 if (num_divs >= ARRAY_SIZE(size_strs)) {
1394 fraction = (float)last_size / 1024;
1396 return snprintf(str, str_bytes, "%.2f%s", fraction,
1397 size_strs[num_divs]);
1401 * __strncpy__null - strncpy with null termination
1402 * @dest: the target array
1403 * @src: the source string
1404 * @n: maximum bytes to copy (size of *dest)
1406 * Like strncpy, but ensures destination is null-terminated.
1408 * Copies the string pointed to by src, including the terminating null
1409 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
1410 * of n bytes. Then ensure that dest is null-terminated.
1412 char *__strncpy__null(char *dest, const char *src, size_t n)
1414 strncpy(dest, src, n);
1421 * Checks to make sure that the label matches our requirements.
1423 0 if everything is safe and usable
1424 -1 if the label is too long
1426 static int check_label(const char *input)
1428 int len = strlen(input);
1430 if (len > BTRFS_LABEL_SIZE - 1) {
1431 fprintf(stderr, "ERROR: Label %s is too long (max %d)\n",
1432 input, BTRFS_LABEL_SIZE - 1);
1439 static int set_label_unmounted(const char *dev, const char *label)
1441 struct btrfs_trans_handle *trans;
1442 struct btrfs_root *root;
1445 ret = check_mounted(dev);
1447 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1451 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1456 /* Open the super_block at the default location
1457 * and as read-write.
1459 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
1460 if (!root) /* errors are printed by open_ctree() */
1463 trans = btrfs_start_transaction(root, 1);
1464 snprintf(root->fs_info->super_copy->label, BTRFS_LABEL_SIZE, "%s",
1466 btrfs_commit_transaction(trans, root);
1468 /* Now we close it since we are done. */
1473 static int set_label_mounted(const char *mount_path, const char *label)
1477 fd = open(mount_path, O_RDONLY | O_NOATIME);
1479 fprintf(stderr, "ERROR: unable to access '%s'\n", mount_path);
1483 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
1484 fprintf(stderr, "ERROR: unable to set label %s\n",
1494 static int get_label_unmounted(const char *dev, char *label)
1496 struct btrfs_root *root;
1499 ret = check_mounted(dev);
1501 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1505 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1510 /* Open the super_block at the default location
1513 root = open_ctree(dev, 0, 0);
1517 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
1519 /* Now we close it since we are done. */
1525 * If a partition is mounted, try to get the filesystem label via its
1526 * mounted path rather than device. Return the corresponding error
1527 * the user specified the device path.
1529 int get_label_mounted(const char *mount_path, char *labelp)
1531 char label[BTRFS_LABEL_SIZE];
1534 fd = open(mount_path, O_RDONLY | O_NOATIME);
1536 fprintf(stderr, "ERROR: unable to access '%s'\n", mount_path);
1540 memset(label, '\0', sizeof(label));
1541 if (ioctl(fd, BTRFS_IOC_GET_FSLABEL, label) < 0) {
1542 fprintf(stderr, "ERROR: unable get label %s\n", strerror(errno));
1547 strncpy(labelp, label, sizeof(label));
1552 int get_label(const char *btrfs_dev, char *label)
1556 ret = is_existing_blk_or_reg_file(btrfs_dev);
1558 ret = get_label_mounted(btrfs_dev, label);
1560 ret = get_label_unmounted(btrfs_dev, label);
1565 int set_label(const char *btrfs_dev, const char *label)
1569 if (check_label(label))
1572 ret = is_existing_blk_or_reg_file(btrfs_dev);
1574 ret = set_label_mounted(btrfs_dev, label);
1576 ret = set_label_unmounted(btrfs_dev, label);
1581 int btrfs_scan_block_devices(int run_ioctl)
1587 struct btrfs_fs_devices *tmp_devices;
1589 FILE *proc_partitions;
1597 proc_partitions = fopen("/proc/partitions","r");
1598 if (!proc_partitions) {
1599 fprintf(stderr, "Unable to open '/proc/partitions' for scanning\n");
1602 /* skip the header */
1603 for (i = 0; i < 2; i++)
1604 if (!fgets(buf, 1023, proc_partitions)) {
1606 "Unable to read '/proc/partitions' for scanning\n");
1607 fclose(proc_partitions);
1611 strcpy(fullpath,"/dev/");
1612 while(fgets(buf, 1023, proc_partitions)) {
1613 i = sscanf(buf," %*d %*d %*d %99s", fullpath+5);
1616 * multipath and MD devices may register as a btrfs filesystem
1617 * both through the original block device and through
1618 * the special (/dev/mapper or /dev/mdX) entry.
1619 * This scans the special entries last
1621 special = strncmp(fullpath, "/dev/dm-", strlen("/dev/dm-")) == 0;
1623 special = strncmp(fullpath, "/dev/md", strlen("/dev/md")) == 0;
1625 if (scans == 0 && special)
1627 if (scans > 0 && !special)
1630 ret = lstat(fullpath, &st);
1632 fprintf(stderr, "failed to stat %s\n", fullpath);
1635 if (!S_ISBLK(st.st_mode)) {
1639 fd = open(fullpath, O_RDONLY);
1641 if (errno != ENOMEDIUM)
1642 fprintf(stderr, "failed to open %s: %s\n",
1643 fullpath, strerror(errno));
1646 ret = btrfs_scan_one_device(fd, fullpath, &tmp_devices,
1648 BTRFS_SUPER_INFO_OFFSET, 0);
1649 if (ret == 0 && run_ioctl > 0) {
1650 btrfs_register_one_device(fullpath);
1655 fclose(proc_partitions);
1665 * A not-so-good version fls64. No fascinating optimization since
1666 * no one except parse_size use it
1668 static int fls64(u64 x)
1672 for (i = 0; i <64; i++)
1673 if (x << i & (1ULL << 63))
1678 u64 parse_size(char *s)
1686 fprintf(stderr, "ERROR: Size value is empty\n");
1691 "ERROR: Size value '%s' is less equal than 0\n", s);
1694 ret = strtoull(s, &endptr, 10);
1696 fprintf(stderr, "ERROR: Size value '%s' is invalid\n", s);
1699 if (endptr[0] && endptr[1]) {
1700 fprintf(stderr, "ERROR: Illegal suffix contains character '%c' in wrong position\n",
1705 * strtoll returns LLONG_MAX when overflow, if this happens,
1706 * need to call strtoull to get the real size
1708 if (errno == ERANGE && ret == ULLONG_MAX) {
1710 "ERROR: Size value '%s' is too large for u64\n", s);
1714 c = tolower(endptr[0]);
1737 fprintf(stderr, "ERROR: Unknown size descriptor '%c'\n",
1742 /* Check whether ret * mult overflow */
1743 if (fls64(ret) + fls64(mult) - 1 > 64) {
1745 "ERROR: Size value '%s' is too large for u64\n", s);
1752 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
1758 ret = stat(fname, &st);
1762 if (S_ISDIR(st.st_mode)) {
1763 *dirstream = opendir(fname);
1766 fd = dirfd(*dirstream);
1767 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
1768 fd = open(fname, open_flags);
1771 * we set this on purpose, in case the caller output
1772 * strerror(errno) as success
1780 closedir(*dirstream);
1785 int open_file_or_dir(const char *fname, DIR **dirstream)
1787 return open_file_or_dir3(fname, dirstream, O_RDWR);
1790 void close_file_or_dir(int fd, DIR *dirstream)
1793 closedir(dirstream);
1798 int get_device_info(int fd, u64 devid,
1799 struct btrfs_ioctl_dev_info_args *di_args)
1803 di_args->devid = devid;
1804 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
1806 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
1807 return ret ? -errno : 0;
1811 * For a given path, fill in the ioctl fs_ and info_ args.
1812 * If the path is a btrfs mountpoint, fill info for all devices.
1813 * If the path is a btrfs device, fill in only that device.
1815 * The path provided must be either on a mounted btrfs fs,
1816 * or be a mounted btrfs device.
1818 * Returns 0 on success, or a negative errno.
1820 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
1821 struct btrfs_ioctl_dev_info_args **di_ret)
1827 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1828 struct btrfs_ioctl_dev_info_args *di_args;
1829 char mp[BTRFS_PATH_NAME_MAX + 1];
1830 DIR *dirstream = NULL;
1832 memset(fi_args, 0, sizeof(*fi_args));
1834 if (is_block_device(path)) {
1835 struct btrfs_super_block *disk_super;
1836 char buf[BTRFS_SUPER_INFO_SIZE];
1839 /* Ensure it's mounted, then set path to the mountpoint */
1840 fd = open(path, O_RDONLY);
1843 fprintf(stderr, "Couldn't open %s: %s\n",
1844 path, strerror(errno));
1847 ret = check_mounted_where(fd, path, mp, sizeof(mp),
1856 /* Only fill in this one device */
1857 fi_args->num_devices = 1;
1859 disk_super = (struct btrfs_super_block *)buf;
1860 ret = btrfs_read_dev_super(fd, disk_super,
1861 BTRFS_SUPER_INFO_OFFSET, 0);
1866 devid = btrfs_stack_device_id(&disk_super->dev_item);
1868 fi_args->max_id = devid;
1871 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
1875 /* at this point path must not be for a block device */
1876 fd = open_file_or_dir(path, &dirstream);
1882 /* fill in fi_args if not just a single device */
1883 if (fi_args->num_devices != 1) {
1884 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
1891 if (!fi_args->num_devices)
1894 di_args = *di_ret = malloc(fi_args->num_devices * sizeof(*di_args));
1900 for (; i <= fi_args->max_id; ++i) {
1901 BUG_ON(ndevs >= fi_args->num_devices);
1902 ret = get_device_info(fd, i, &di_args[ndevs]);
1911 * only when the only dev we wanted to find is not there then
1912 * let any error be returned
1914 if (fi_args->num_devices != 1) {
1920 close_file_or_dir(fd, dirstream);
1924 #define isoctal(c) (((c) & ~7) == '0')
1926 static inline void translate(char *f, char *t)
1928 while (*f != '\0') {
1930 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
1931 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
1941 * Checks if the swap device.
1942 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
1944 static int is_swap_device(const char *file)
1955 if (stat(file, &st_buf) < 0)
1957 if (S_ISBLK(st_buf.st_mode))
1958 dev = st_buf.st_rdev;
1959 else if (S_ISREG(st_buf.st_mode)) {
1960 dev = st_buf.st_dev;
1961 ino = st_buf.st_ino;
1965 if ((f = fopen("/proc/swaps", "r")) == NULL)
1968 /* skip the first line */
1969 if (fgets(tmp, sizeof(tmp), f) == NULL)
1972 while (fgets(tmp, sizeof(tmp), f) != NULL) {
1973 if ((cp = strchr(tmp, ' ')) != NULL)
1975 if ((cp = strchr(tmp, '\t')) != NULL)
1977 translate(tmp, buf);
1978 if (stat(buf, &st_buf) != 0)
1980 if (S_ISBLK(st_buf.st_mode)) {
1981 if (dev == st_buf.st_rdev) {
1985 } else if (S_ISREG(st_buf.st_mode)) {
1986 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
2000 * Check for existing filesystem or partition table on device.
2002 * 1 for existing fs or partition
2003 * 0 for nothing found
2004 * -1 for internal error
2011 blkid_probe pr = NULL;
2015 if (!device || !*device)
2018 ret = -1; /* will reset on success of all setup calls */
2020 pr = blkid_new_probe_from_filename(device);
2024 size = blkid_probe_get_size(pr);
2028 /* nothing to overwrite on a 0-length device */
2034 ret = blkid_probe_enable_partitions(pr, 1);
2038 ret = blkid_do_fullprobe(pr);
2043 * Blkid returns 1 for nothing found and 0 when it finds a signature,
2044 * but we want the exact opposite, so reverse the return value here.
2046 * In addition print some useful diagnostics about what actually is
2054 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
2056 "%s appears to contain an existing "
2057 "filesystem (%s).\n", device, type);
2058 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
2060 "%s appears to contain a partition "
2061 "table (%s).\n", device, type);
2064 "%s appears to contain something weird "
2065 "according to blkid\n", device);
2071 blkid_free_probe(pr);
2074 "probe of %s failed, cannot detect "
2075 "existing filesystem.\n", device);
2079 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
2080 u64 dev_cnt, int mixed, char *estr)
2088 allowed |= BTRFS_BLOCK_GROUP_RAID10;
2090 allowed |= BTRFS_BLOCK_GROUP_RAID6;
2092 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
2093 BTRFS_BLOCK_GROUP_RAID5;
2096 allowed |= BTRFS_BLOCK_GROUP_DUP;
2099 if (metadata_profile & ~allowed) {
2100 snprintf(estr, sz, "unable to create FS with metadata "
2101 "profile %llu (have %llu devices)\n",
2102 metadata_profile, dev_cnt);
2105 if (data_profile & ~allowed) {
2106 snprintf(estr, sz, "unable to create FS with data "
2107 "profile %llu (have %llu devices)\n",
2108 metadata_profile, dev_cnt);
2112 if (!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP)) {
2114 "dup for data is allowed only in mixed mode");
2120 /* Check if disk is suitable for btrfs
2122 * 1: something is wrong, estr provides the error
2125 int test_dev_for_mkfs(char *file, int force_overwrite, char *estr)
2131 ret = is_swap_device(file);
2133 snprintf(estr, sz, "error checking %s status: %s\n", file,
2138 snprintf(estr, sz, "%s is a swap device\n", file);
2141 if (!force_overwrite) {
2142 if (check_overwrite(file)) {
2143 snprintf(estr, sz, "Use the -f option to force overwrite.\n");
2147 ret = check_mounted(file);
2149 snprintf(estr, sz, "error checking %s mount status\n",
2154 snprintf(estr, sz, "%s is mounted\n", file);
2157 /* check if the device is busy */
2158 fd = open(file, O_RDWR|O_EXCL);
2160 snprintf(estr, sz, "unable to open %s: %s\n", file,
2164 if (fstat(fd, &st)) {
2165 snprintf(estr, sz, "unable to stat %s: %s\n", file,
2170 if (!S_ISBLK(st.st_mode)) {
2171 fprintf(stderr, "'%s' is not a block device\n", file);
2179 int btrfs_scan_lblkid(int update_kernel)
2184 struct btrfs_fs_devices *tmp_devices;
2185 blkid_dev_iterate iter = NULL;
2186 blkid_dev dev = NULL;
2187 blkid_cache cache = NULL;
2188 char path[PATH_MAX];
2190 if (blkid_get_cache(&cache, 0) < 0) {
2191 printf("ERROR: lblkid cache get failed\n");
2194 blkid_probe_all(cache);
2195 iter = blkid_dev_iterate_begin(cache);
2196 blkid_dev_set_search(iter, "TYPE", "btrfs");
2197 while (blkid_dev_next(iter, &dev) == 0) {
2198 dev = blkid_verify(cache, dev);
2201 /* if we are here its definitely a btrfs disk*/
2202 strncpy(path, blkid_dev_devname(dev), PATH_MAX);
2204 fd = open(path, O_RDONLY);
2206 printf("ERROR: could not open %s\n", path);
2209 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
2210 &num_devices, BTRFS_SUPER_INFO_OFFSET, 0);
2212 printf("ERROR: could not scan %s\n", path);
2219 btrfs_register_one_device(path);
2221 blkid_dev_iterate_end(iter);
2222 blkid_put_cache(cache);
2227 * scans devs for the btrfs
2229 int scan_for_btrfs(int where, int update_kernel)
2234 case BTRFS_SCAN_PROC:
2235 ret = btrfs_scan_block_devices(update_kernel);
2237 case BTRFS_SCAN_DEV:
2238 ret = btrfs_scan_one_dir("/dev", update_kernel);
2240 case BTRFS_SCAN_LBLKID:
2241 ret = btrfs_scan_lblkid(update_kernel);
2247 int is_vol_small(char *file)
2254 fd = open(file, O_RDONLY);
2257 if (fstat(fd, &st) < 0) {
2262 size = btrfs_device_size(fd, &st);
2267 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
2277 * This reads a line from the stdin and only returns non-zero if the
2278 * first whitespace delimited token is a case insensitive match with yes
2281 int ask_user(char *question)
2283 char buf[30] = {0,};
2284 char *saveptr = NULL;
2287 printf("%s [y/N]: ", question);
2289 return fgets(buf, sizeof(buf) - 1, stdin) &&
2290 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
2291 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
2296 * - file or directory return the containing tree root id
2297 * - subvolume return its own tree id
2298 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
2299 * undefined and function returns -1
2301 int lookup_ino_rootid(int fd, u64 *rootid)
2303 struct btrfs_ioctl_ino_lookup_args args;
2307 memset(&args, 0, sizeof(args));
2309 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
2311 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
2314 fprintf(stderr, "ERROR: Failed to lookup root id - %s\n",
2319 *rootid = args.treeid;
2325 * return 0 if a btrfs mount point is found
2326 * return 1 if a mount point is found but not btrfs
2327 * return <0 if something goes wrong
2329 int find_mount_root(const char *path, char **mount_root)
2337 int longest_matchlen = 0;
2338 char *longest_match = NULL;
2340 fd = open(path, O_RDONLY | O_NOATIME);
2345 mnttab = setmntent("/proc/self/mounts", "r");
2349 while ((ent = getmntent(mnttab))) {
2350 len = strlen(ent->mnt_dir);
2351 if (strncmp(ent->mnt_dir, path, len) == 0) {
2352 /* match found and use the latest match */
2353 if (longest_matchlen <= len) {
2354 free(longest_match);
2355 longest_matchlen = len;
2356 longest_match = strdup(ent->mnt_dir);
2357 not_btrfs = strcmp(ent->mnt_type, "btrfs");
2366 free(longest_match);
2371 *mount_root = realpath(longest_match, NULL);
2375 free(longest_match);
2379 int test_minimum_size(const char *file, u32 leafsize)
2382 struct stat statbuf;
2384 fd = open(file, O_RDONLY);
2387 if (stat(file, &statbuf) < 0) {
2391 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(leafsize)) {
2400 * test if name is a correct subvolume name
2401 * this function return
2402 * 0-> name is not a correct subvolume name
2403 * 1-> name is a correct subvolume name
2405 int test_issubvolname(const char *name)
2407 return name[0] != '\0' && !strchr(name, '/') &&
2408 strcmp(name, ".") && strcmp(name, "..");
2412 * test if path is a directory
2413 * this function return
2414 * 0-> path exists but it is not a directory
2415 * 1-> path exists and it is a directory
2416 * -1 -> path is unaccessible
2418 int test_isdir(const char *path)
2423 ret = stat(path, &st);
2427 return S_ISDIR(st.st_mode);