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>
43 #include "kerncompat.h"
44 #include "radix-tree.h"
47 #include "transaction.h"
54 #define BLKDISCARD _IO(0x12,119)
57 static int btrfs_scan_done = 0;
59 static char argv0_buf[ARGV0_BUF_SIZE] = "btrfs";
61 void fixup_argv0(char **argv, const char *token)
63 int len = strlen(argv0_buf);
65 snprintf(argv0_buf + len, sizeof(argv0_buf) - len, " %s", token);
69 void set_argv0(char **argv)
71 strncpy(argv0_buf, argv[0], sizeof(argv0_buf));
72 argv0_buf[sizeof(argv0_buf) - 1] = 0;
75 int check_argc_exact(int nargs, int expected)
78 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
80 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
82 return nargs != expected;
85 int check_argc_min(int nargs, int expected)
87 if (nargs < expected) {
88 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
95 int check_argc_max(int nargs, int expected)
97 if (nargs > expected) {
98 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
107 * Discard the given range in one go
109 static int discard_range(int fd, u64 start, u64 len)
111 u64 range[2] = { start, len };
113 if (ioctl(fd, BLKDISCARD, &range) < 0)
119 * Discard blocks in the given range in 1G chunks, the process is interruptible
121 static int discard_blocks(int fd, u64 start, u64 len)
125 u64 chunk_size = min_t(u64, len, 1*1024*1024*1024);
128 ret = discard_range(fd, start, chunk_size);
138 static u64 reference_root_table[] = {
139 [1] = BTRFS_ROOT_TREE_OBJECTID,
140 [2] = BTRFS_EXTENT_TREE_OBJECTID,
141 [3] = BTRFS_CHUNK_TREE_OBJECTID,
142 [4] = BTRFS_DEV_TREE_OBJECTID,
143 [5] = BTRFS_FS_TREE_OBJECTID,
144 [6] = BTRFS_CSUM_TREE_OBJECTID,
147 int test_uuid_unique(char *fs_uuid)
150 blkid_dev_iterate iter = NULL;
151 blkid_dev dev = NULL;
152 blkid_cache cache = NULL;
154 if (blkid_get_cache(&cache, 0) < 0) {
155 printf("ERROR: lblkid cache get failed\n");
158 blkid_probe_all(cache);
159 iter = blkid_dev_iterate_begin(cache);
160 blkid_dev_set_search(iter, "UUID", fs_uuid);
162 while (blkid_dev_next(iter, &dev) == 0) {
163 dev = blkid_verify(cache, dev);
170 blkid_dev_iterate_end(iter);
171 blkid_put_cache(cache);
176 int make_btrfs(int fd, const char *device, const char *label, char *fs_uuid,
177 u64 blocks[7], u64 num_bytes, u32 nodesize,
178 u32 leafsize, u32 sectorsize, u32 stripesize, u64 features)
180 struct btrfs_super_block super;
181 struct extent_buffer *buf = NULL;
182 struct btrfs_root_item root_item;
183 struct btrfs_disk_key disk_key;
184 struct btrfs_extent_item *extent_item;
185 struct btrfs_inode_item *inode_item;
186 struct btrfs_chunk *chunk;
187 struct btrfs_dev_item *dev_item;
188 struct btrfs_dev_extent *dev_extent;
189 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
199 int skinny_metadata = !!(features &
200 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
202 first_free = BTRFS_SUPER_INFO_OFFSET + sectorsize * 2 - 1;
203 first_free &= ~((u64)sectorsize - 1);
205 memset(&super, 0, sizeof(super));
207 num_bytes = (num_bytes / sectorsize) * sectorsize;
209 if (uuid_parse(fs_uuid, super.fsid) != 0) {
210 fprintf(stderr, "could not parse UUID: %s\n", fs_uuid);
214 if (!test_uuid_unique(fs_uuid)) {
215 fprintf(stderr, "non-unique UUID: %s\n", fs_uuid);
220 uuid_generate(super.fsid);
222 uuid_generate(super.dev_item.uuid);
223 uuid_generate(chunk_tree_uuid);
225 btrfs_set_super_bytenr(&super, blocks[0]);
226 btrfs_set_super_num_devices(&super, 1);
227 btrfs_set_super_magic(&super, BTRFS_MAGIC);
228 btrfs_set_super_generation(&super, 1);
229 btrfs_set_super_root(&super, blocks[1]);
230 btrfs_set_super_chunk_root(&super, blocks[3]);
231 btrfs_set_super_total_bytes(&super, num_bytes);
232 btrfs_set_super_bytes_used(&super, 6 * leafsize);
233 btrfs_set_super_sectorsize(&super, sectorsize);
234 btrfs_set_super_leafsize(&super, leafsize);
235 btrfs_set_super_nodesize(&super, nodesize);
236 btrfs_set_super_stripesize(&super, stripesize);
237 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
238 btrfs_set_super_chunk_root_generation(&super, 1);
239 btrfs_set_super_cache_generation(&super, -1);
240 btrfs_set_super_incompat_flags(&super, features);
242 strncpy(super.label, label, BTRFS_LABEL_SIZE - 1);
244 buf = malloc(sizeof(*buf) + max(sectorsize, leafsize));
246 /* create the tree of root objects */
247 memset(buf->data, 0, leafsize);
249 btrfs_set_header_bytenr(buf, blocks[1]);
250 btrfs_set_header_nritems(buf, 4);
251 btrfs_set_header_generation(buf, 1);
252 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
253 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
254 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
257 write_extent_buffer(buf, chunk_tree_uuid,
258 btrfs_header_chunk_tree_uuid(buf),
261 /* create the items for the root tree */
262 memset(&root_item, 0, sizeof(root_item));
263 inode_item = &root_item.inode;
264 btrfs_set_stack_inode_generation(inode_item, 1);
265 btrfs_set_stack_inode_size(inode_item, 3);
266 btrfs_set_stack_inode_nlink(inode_item, 1);
267 btrfs_set_stack_inode_nbytes(inode_item, leafsize);
268 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
269 btrfs_set_root_refs(&root_item, 1);
270 btrfs_set_root_used(&root_item, leafsize);
271 btrfs_set_root_generation(&root_item, 1);
273 memset(&disk_key, 0, sizeof(disk_key));
274 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
275 btrfs_set_disk_key_offset(&disk_key, 0);
278 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - sizeof(root_item);
279 btrfs_set_root_bytenr(&root_item, blocks[2]);
280 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
281 btrfs_set_item_key(buf, &disk_key, nritems);
282 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
283 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
285 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
286 nritems), sizeof(root_item));
289 itemoff = itemoff - sizeof(root_item);
290 btrfs_set_root_bytenr(&root_item, blocks[4]);
291 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
292 btrfs_set_item_key(buf, &disk_key, nritems);
293 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
294 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
296 write_extent_buffer(buf, &root_item,
297 btrfs_item_ptr_offset(buf, nritems),
301 itemoff = itemoff - sizeof(root_item);
302 btrfs_set_root_bytenr(&root_item, blocks[5]);
303 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
304 btrfs_set_item_key(buf, &disk_key, nritems);
305 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
306 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
308 write_extent_buffer(buf, &root_item,
309 btrfs_item_ptr_offset(buf, nritems),
313 itemoff = itemoff - sizeof(root_item);
314 btrfs_set_root_bytenr(&root_item, blocks[6]);
315 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
316 btrfs_set_item_key(buf, &disk_key, nritems);
317 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
318 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
320 write_extent_buffer(buf, &root_item,
321 btrfs_item_ptr_offset(buf, nritems),
326 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
327 ret = pwrite(fd, buf->data, leafsize, blocks[1]);
328 if (ret != leafsize) {
329 ret = (ret < 0 ? -errno : -EIO);
333 /* create the items for the extent tree */
334 memset(buf->data+sizeof(struct btrfs_header), 0,
335 leafsize-sizeof(struct btrfs_header));
337 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize);
338 for (i = 1; i < 7; i++) {
339 item_size = sizeof(struct btrfs_extent_item);
340 if (!skinny_metadata)
341 item_size += sizeof(struct btrfs_tree_block_info);
343 BUG_ON(blocks[i] < first_free);
344 BUG_ON(blocks[i] < blocks[i - 1]);
346 /* create extent item */
347 itemoff -= item_size;
348 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
349 if (skinny_metadata) {
350 btrfs_set_disk_key_type(&disk_key,
351 BTRFS_METADATA_ITEM_KEY);
352 btrfs_set_disk_key_offset(&disk_key, 0);
354 btrfs_set_disk_key_type(&disk_key,
355 BTRFS_EXTENT_ITEM_KEY);
356 btrfs_set_disk_key_offset(&disk_key, leafsize);
358 btrfs_set_item_key(buf, &disk_key, nritems);
359 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
361 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
363 extent_item = btrfs_item_ptr(buf, nritems,
364 struct btrfs_extent_item);
365 btrfs_set_extent_refs(buf, extent_item, 1);
366 btrfs_set_extent_generation(buf, extent_item, 1);
367 btrfs_set_extent_flags(buf, extent_item,
368 BTRFS_EXTENT_FLAG_TREE_BLOCK);
371 /* create extent ref */
372 ref_root = reference_root_table[i];
373 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
374 btrfs_set_disk_key_offset(&disk_key, ref_root);
375 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
376 btrfs_set_item_key(buf, &disk_key, nritems);
377 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
379 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
382 btrfs_set_header_bytenr(buf, blocks[2]);
383 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
384 btrfs_set_header_nritems(buf, nritems);
385 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
386 ret = pwrite(fd, buf->data, leafsize, blocks[2]);
387 if (ret != leafsize) {
388 ret = (ret < 0 ? -errno : -EIO);
392 /* create the chunk tree */
393 memset(buf->data+sizeof(struct btrfs_header), 0,
394 leafsize-sizeof(struct btrfs_header));
396 item_size = sizeof(*dev_item);
397 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - item_size;
399 /* first device 1 (there is no device 0) */
400 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
401 btrfs_set_disk_key_offset(&disk_key, 1);
402 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
403 btrfs_set_item_key(buf, &disk_key, nritems);
404 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
405 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
407 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
408 btrfs_set_device_id(buf, dev_item, 1);
409 btrfs_set_device_generation(buf, dev_item, 0);
410 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
411 btrfs_set_device_bytes_used(buf, dev_item,
412 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
413 btrfs_set_device_io_align(buf, dev_item, sectorsize);
414 btrfs_set_device_io_width(buf, dev_item, sectorsize);
415 btrfs_set_device_sector_size(buf, dev_item, sectorsize);
416 btrfs_set_device_type(buf, dev_item, 0);
418 write_extent_buffer(buf, super.dev_item.uuid,
419 (unsigned long)btrfs_device_uuid(dev_item),
421 write_extent_buffer(buf, super.fsid,
422 (unsigned long)btrfs_device_fsid(dev_item),
424 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
428 item_size = btrfs_chunk_item_size(1);
429 itemoff = itemoff - item_size;
431 /* then we have chunk 0 */
432 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
433 btrfs_set_disk_key_offset(&disk_key, 0);
434 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
435 btrfs_set_item_key(buf, &disk_key, nritems);
436 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
437 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
439 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
440 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
441 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
442 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
443 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
444 btrfs_set_chunk_io_align(buf, chunk, sectorsize);
445 btrfs_set_chunk_io_width(buf, chunk, sectorsize);
446 btrfs_set_chunk_sector_size(buf, chunk, sectorsize);
447 btrfs_set_chunk_num_stripes(buf, chunk, 1);
448 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
449 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
452 write_extent_buffer(buf, super.dev_item.uuid,
453 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
456 /* copy the key for the chunk to the system array */
457 ptr = super.sys_chunk_array;
458 array_size = sizeof(disk_key);
460 memcpy(ptr, &disk_key, sizeof(disk_key));
461 ptr += sizeof(disk_key);
463 /* copy the chunk to the system array */
464 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
465 array_size += item_size;
467 btrfs_set_super_sys_array_size(&super, array_size);
469 btrfs_set_header_bytenr(buf, blocks[3]);
470 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
471 btrfs_set_header_nritems(buf, nritems);
472 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
473 ret = pwrite(fd, buf->data, leafsize, blocks[3]);
474 if (ret != leafsize) {
475 ret = (ret < 0 ? -errno : -EIO);
479 /* create the device tree */
480 memset(buf->data+sizeof(struct btrfs_header), 0,
481 leafsize-sizeof(struct btrfs_header));
483 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) -
484 sizeof(struct btrfs_dev_extent);
486 btrfs_set_disk_key_objectid(&disk_key, 1);
487 btrfs_set_disk_key_offset(&disk_key, 0);
488 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
489 btrfs_set_item_key(buf, &disk_key, nritems);
490 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
491 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
492 sizeof(struct btrfs_dev_extent));
493 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
494 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
495 BTRFS_CHUNK_TREE_OBJECTID);
496 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
497 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
498 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
500 write_extent_buffer(buf, chunk_tree_uuid,
501 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
504 btrfs_set_dev_extent_length(buf, dev_extent,
505 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
508 btrfs_set_header_bytenr(buf, blocks[4]);
509 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
510 btrfs_set_header_nritems(buf, nritems);
511 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
512 ret = pwrite(fd, buf->data, leafsize, blocks[4]);
513 if (ret != leafsize) {
514 ret = (ret < 0 ? -errno : -EIO);
518 /* create the FS root */
519 memset(buf->data+sizeof(struct btrfs_header), 0,
520 leafsize-sizeof(struct btrfs_header));
521 btrfs_set_header_bytenr(buf, blocks[5]);
522 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
523 btrfs_set_header_nritems(buf, 0);
524 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
525 ret = pwrite(fd, buf->data, leafsize, blocks[5]);
526 if (ret != leafsize) {
527 ret = (ret < 0 ? -errno : -EIO);
530 /* finally create the csum root */
531 memset(buf->data+sizeof(struct btrfs_header), 0,
532 leafsize-sizeof(struct btrfs_header));
533 btrfs_set_header_bytenr(buf, blocks[6]);
534 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
535 btrfs_set_header_nritems(buf, 0);
536 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
537 ret = pwrite(fd, buf->data, leafsize, blocks[6]);
538 if (ret != leafsize) {
539 ret = (ret < 0 ? -errno : -EIO);
543 /* and write out the super block */
544 BUG_ON(sizeof(super) > sectorsize);
545 memset(buf->data, 0, sectorsize);
546 memcpy(buf->data, &super, sizeof(super));
547 buf->len = sectorsize;
548 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
549 ret = pwrite(fd, buf->data, sectorsize, blocks[0]);
550 if (ret != sectorsize) {
551 ret = (ret < 0 ? -errno : -EIO);
562 u64 btrfs_device_size(int fd, struct stat *st)
565 if (S_ISREG(st->st_mode)) {
568 if (!S_ISBLK(st->st_mode)) {
571 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
577 static int zero_blocks(int fd, off_t start, size_t len)
579 char *buf = malloc(len);
586 written = pwrite(fd, buf, len, start);
593 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
595 /* don't write outside the device by clamping the region to the device size */
596 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
598 off_t end = max(start, start + len);
601 /* and don't overwrite the disk labels on sparc */
602 start = max(start, 1024);
603 end = max(end, 1024);
606 start = min_t(u64, start, dev_size);
607 end = min_t(u64, end, dev_size);
609 return zero_blocks(fd, start, end - start);
612 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
613 struct btrfs_root *root, int fd, char *path,
614 u64 block_count, u32 io_width, u32 io_align,
617 struct btrfs_super_block *disk_super;
618 struct btrfs_super_block *super = root->fs_info->super_copy;
619 struct btrfs_device *device;
620 struct btrfs_dev_item *dev_item;
626 device = kzalloc(sizeof(*device), GFP_NOFS);
629 buf = kmalloc(sectorsize, GFP_NOFS);
634 BUG_ON(sizeof(*disk_super) > sectorsize);
635 memset(buf, 0, sectorsize);
637 disk_super = (struct btrfs_super_block *)buf;
638 dev_item = &disk_super->dev_item;
640 uuid_generate(device->uuid);
643 device->io_width = io_width;
644 device->io_align = io_align;
645 device->sector_size = sectorsize;
647 device->writeable = 1;
648 device->total_bytes = block_count;
649 device->bytes_used = 0;
650 device->total_ios = 0;
651 device->dev_root = root->fs_info->dev_root;
653 ret = btrfs_add_device(trans, root, device);
656 total_bytes = btrfs_super_total_bytes(super) + block_count;
657 btrfs_set_super_total_bytes(super, total_bytes);
659 num_devs = btrfs_super_num_devices(super) + 1;
660 btrfs_set_super_num_devices(super, num_devs);
662 memcpy(disk_super, super, sizeof(*disk_super));
664 printf("adding device %s id %llu\n", path,
665 (unsigned long long)device->devid);
667 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
668 btrfs_set_stack_device_id(dev_item, device->devid);
669 btrfs_set_stack_device_type(dev_item, device->type);
670 btrfs_set_stack_device_io_align(dev_item, device->io_align);
671 btrfs_set_stack_device_io_width(dev_item, device->io_width);
672 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
673 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
674 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
675 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
677 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
678 BUG_ON(ret != sectorsize);
681 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
682 device->fs_devices = root->fs_info->fs_devices;
686 static void btrfs_wipe_existing_sb(int fd)
688 const char *off = NULL;
693 blkid_probe pr = NULL;
695 pr = blkid_new_probe();
699 if (blkid_probe_set_device(pr, fd, 0, 0))
702 rc = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
704 rc = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
706 if (rc || len == 0 || off == NULL)
709 offset = strtoll(off, NULL, 10);
710 if (len > sizeof(buf))
714 rc = pwrite(fd, buf, len, offset);
718 blkid_free_probe(pr);
722 int btrfs_prepare_device(int fd, char *file, int zero_end, u64 *block_count_ret,
723 u64 max_block_count, int *mixed, int discard)
729 ret = fstat(fd, &st);
731 fprintf(stderr, "unable to stat %s\n", file);
735 block_count = btrfs_device_size(fd, &st);
736 if (block_count == 0) {
737 fprintf(stderr, "unable to find %s size\n", file);
741 block_count = min(block_count, max_block_count);
743 if (block_count < BTRFS_MKFS_SMALL_VOLUME_SIZE && !(*mixed))
748 * We intentionally ignore errors from the discard ioctl. It
749 * is not necessary for the mkfs functionality but just an
752 if (discard_range(fd, 0, 0) == 0) {
753 fprintf(stderr, "Performing full device TRIM (%s) ...\n",
754 pretty_size(block_count));
755 discard_blocks(fd, 0, block_count);
759 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
760 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
761 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
762 BTRFS_SUPER_INFO_SIZE, block_count);
763 if (!ret && zero_end)
764 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
765 ZERO_DEV_BYTES, block_count);
768 fprintf(stderr, "ERROR: failed to zero device '%s' - %s\n",
769 file, strerror(-ret));
773 btrfs_wipe_existing_sb(fd);
775 *block_count_ret = block_count;
779 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
780 struct btrfs_root *root, u64 objectid)
783 struct btrfs_inode_item inode_item;
784 time_t now = time(NULL);
786 memset(&inode_item, 0, sizeof(inode_item));
787 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
788 btrfs_set_stack_inode_size(&inode_item, 0);
789 btrfs_set_stack_inode_nlink(&inode_item, 1);
790 btrfs_set_stack_inode_nbytes(&inode_item, root->leafsize);
791 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
792 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
793 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
794 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
795 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
796 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
797 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
798 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
799 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
801 if (root->fs_info->tree_root == root)
802 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
804 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
808 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
812 btrfs_set_root_dirid(&root->root_item, objectid);
819 * checks if a path is a block device node
820 * Returns negative errno on failure, otherwise
821 * returns 1 for blockdev, 0 for not-blockdev
823 int is_block_device(const char *path)
827 if (stat(path, &statbuf) < 0)
830 return S_ISBLK(statbuf.st_mode);
834 * check if given path is a mount point
835 * return 1 if yes. 0 if no. -1 for error
837 int is_mount_point(const char *path)
843 f = setmntent("/proc/self/mounts", "r");
847 while ((mnt = getmntent(f)) != NULL) {
848 if (strcmp(mnt->mnt_dir, path))
857 static int is_reg_file(const char *path)
861 if (stat(path, &statbuf) < 0)
863 return S_ISREG(statbuf.st_mode);
867 * This function checks if the given input parameter is
869 * return <0 : some error in the given input
870 * return BTRFS_ARG_UNKNOWN: unknown input
871 * return BTRFS_ARG_UUID: given input is uuid
872 * return BTRFS_ARG_MNTPOINT: given input is path
873 * return BTRFS_ARG_REG: given input is regular file
875 int check_arg_type(const char *input)
883 if (realpath(input, path)) {
884 if (is_block_device(path) == 1)
885 return BTRFS_ARG_BLKDEV;
887 if (is_mount_point(path) == 1)
888 return BTRFS_ARG_MNTPOINT;
890 if (is_reg_file(path))
891 return BTRFS_ARG_REG;
893 return BTRFS_ARG_UNKNOWN;
896 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
897 !uuid_parse(input, uuid))
898 return BTRFS_ARG_UUID;
900 return BTRFS_ARG_UNKNOWN;
904 * Find the mount point for a mounted device.
905 * On success, returns 0 with mountpoint in *mp.
906 * On failure, returns -errno (not mounted yields -EINVAL)
907 * Is noisy on failures, expects to be given a mounted device.
909 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
914 ret = is_block_device(dev);
917 fprintf(stderr, "%s is not a block device\n", dev);
920 fprintf(stderr, "Could not check %s: %s\n",
921 dev, strerror(-ret));
926 fd = open(dev, O_RDONLY);
929 fprintf(stderr, "Could not open %s: %s\n", dev, strerror(errno));
933 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
936 } else { /* mounted, all good */
946 * Given a pathname, return a filehandle to:
947 * the original pathname or,
948 * if the pathname is a mounted btrfs device, to its mountpoint.
950 * On error, return -1, errno should be set.
952 int open_path_or_dev_mnt(const char *path, DIR **dirstream)
954 char mp[BTRFS_PATH_NAME_MAX + 1];
957 if (is_block_device(path)) {
960 ret = get_btrfs_mount(path, mp, sizeof(mp));
962 /* not a mounted btrfs dev */
966 fdmnt = open_file_or_dir(mp, dirstream);
968 fdmnt = open_file_or_dir(path, dirstream);
974 /* checks if a device is a loop device */
975 static int is_loop_device (const char* device) {
978 if(stat(device, &statbuf) < 0)
981 return (S_ISBLK(statbuf.st_mode) &&
982 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
986 /* Takes a loop device path (e.g. /dev/loop0) and returns
987 * the associated file (e.g. /images/my_btrfs.img) */
988 static int resolve_loop_device(const char* loop_dev, char* loop_file,
995 char real_loop_dev[PATH_MAX];
997 if (!realpath(loop_dev, real_loop_dev))
999 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
1000 if (!(f = fopen(p, "r")))
1003 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
1004 ret = fscanf(f, fmt, loop_file);
1013 * Checks whether a and b are identical or device
1014 * files associated with the same block device
1016 static int is_same_blk_file(const char* a, const char* b)
1018 struct stat st_buf_a, st_buf_b;
1019 char real_a[PATH_MAX];
1020 char real_b[PATH_MAX];
1022 if (!realpath(a, real_a))
1023 strncpy_null(real_a, a);
1025 if (!realpath(b, real_b))
1026 strncpy_null(real_b, b);
1028 /* Identical path? */
1029 if (strcmp(real_a, real_b) == 0)
1032 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
1033 if (errno == ENOENT)
1038 /* Same blockdevice? */
1039 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
1040 st_buf_a.st_rdev == st_buf_b.st_rdev) {
1045 if (st_buf_a.st_dev == st_buf_b.st_dev &&
1046 st_buf_a.st_ino == st_buf_b.st_ino) {
1053 /* checks if a and b are identical or device
1054 * files associated with the same block device or
1055 * if one file is a loop device that uses the other
1058 static int is_same_loop_file(const char* a, const char* b)
1060 char res_a[PATH_MAX];
1061 char res_b[PATH_MAX];
1062 const char* final_a = NULL;
1063 const char* final_b = NULL;
1066 /* Resolve a if it is a loop device */
1067 if((ret = is_loop_device(a)) < 0) {
1072 ret = resolve_loop_device(a, res_a, sizeof(res_a));
1083 /* Resolve b if it is a loop device */
1084 if ((ret = is_loop_device(b)) < 0) {
1089 ret = resolve_loop_device(b, res_b, sizeof(res_b));
1100 return is_same_blk_file(final_a, final_b);
1103 /* Checks if a file exists and is a block or regular file*/
1104 static int is_existing_blk_or_reg_file(const char* filename)
1108 if(stat(filename, &st_buf) < 0) {
1115 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
1118 /* Checks if a file is used (directly or indirectly via a loop device)
1119 * by a device in fs_devices
1121 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
1125 struct list_head *head;
1126 struct list_head *cur;
1127 struct btrfs_device *device;
1129 head = &fs_devices->devices;
1130 list_for_each(cur, head) {
1131 device = list_entry(cur, struct btrfs_device, dev_list);
1133 if((ret = is_same_loop_file(device->name, file)))
1141 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1142 * Returns NULL on invalid input or malloc failure; Other failures
1143 * will be handled by the caller using the input pathame.
1145 char *canonicalize_dm_name(const char *ptname)
1149 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
1151 if (!ptname || !*ptname)
1154 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
1155 if (!(f = fopen(path, "r")))
1158 /* read <name>\n from sysfs */
1159 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
1160 name[sz - 1] = '\0';
1161 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
1163 if (access(path, F_OK) == 0)
1171 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
1172 * to a device mapper pathname.
1173 * Returns NULL on invalid input or malloc failure; Other failures
1174 * will be handled by the caller using the input pathame.
1176 char *canonicalize_path(const char *path)
1178 char *canonical, *p;
1180 if (!path || !*path)
1183 canonical = realpath(path, NULL);
1185 return strdup(path);
1186 p = strrchr(canonical, '/');
1187 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
1188 char *dm = canonicalize_dm_name(p + 1);
1199 * returns 1 if the device was mounted, < 0 on error or 0 if everything
1200 * is safe to continue.
1202 int check_mounted(const char* file)
1207 fd = open(file, O_RDONLY);
1209 fprintf (stderr, "check_mounted(): Could not open %s\n", file);
1213 ret = check_mounted_where(fd, file, NULL, 0, NULL);
1219 int check_mounted_where(int fd, const char *file, char *where, int size,
1220 struct btrfs_fs_devices **fs_dev_ret)
1225 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1229 /* scan the initial device */
1230 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
1231 &total_devs, BTRFS_SUPER_INFO_OFFSET, 0);
1232 is_btrfs = (ret >= 0);
1234 /* scan other devices */
1235 if (is_btrfs && total_devs > 1) {
1236 ret = btrfs_scan_lblkid();
1241 /* iterate over the list of currently mountes filesystems */
1242 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
1245 while ((mnt = getmntent (f)) != NULL) {
1247 if(strcmp(mnt->mnt_type, "btrfs") != 0)
1250 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
1252 /* ignore entries in the mount table that are not
1253 associated with a file*/
1254 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
1255 goto out_mntloop_err;
1259 ret = is_same_loop_file(file, mnt->mnt_fsname);
1263 goto out_mntloop_err;
1268 /* Did we find an entry in mnt table? */
1269 if (mnt && size && where) {
1270 strncpy(where, mnt->mnt_dir, size);
1274 *fs_dev_ret = fs_devices_mnt;
1276 ret = (mnt != NULL);
1284 struct pending_dir {
1285 struct list_head list;
1286 char name[PATH_MAX];
1289 int btrfs_register_one_device(const char *fname)
1291 struct btrfs_ioctl_vol_args args;
1296 fd = open("/dev/btrfs-control", O_RDWR);
1298 fprintf(stderr, "failed to open /dev/btrfs-control "
1299 "skipping device registration: %s\n",
1303 strncpy(args.name, fname, BTRFS_PATH_NAME_MAX);
1304 args.name[BTRFS_PATH_NAME_MAX-1] = 0;
1305 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
1308 fprintf(stderr, "ERROR: device scan failed '%s' - %s\n",
1309 fname, strerror(e));
1317 * Register all devices in the fs_uuid list created in the user
1318 * space. Ensure btrfs_scan_lblkid() is called before this func.
1320 int btrfs_register_all_devices(void)
1323 struct btrfs_fs_devices *fs_devices;
1324 struct btrfs_device *device;
1325 struct list_head *all_uuids;
1327 all_uuids = btrfs_scanned_uuids();
1329 list_for_each_entry(fs_devices, all_uuids, list) {
1330 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1331 if (strlen(device->name) != 0) {
1332 err = btrfs_register_one_device(device->name);
1343 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
1346 struct btrfs_super_block *disk_super;
1350 buf = malloc(BTRFS_SUPER_INFO_SIZE);
1355 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
1356 if (ret != BTRFS_SUPER_INFO_SIZE)
1360 disk_super = (struct btrfs_super_block *)buf;
1361 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
1364 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
1373 static const char* unit_suffix_binary[] =
1374 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
1375 static const char* unit_suffix_decimal[] =
1376 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
1378 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
1384 const char** suffix = NULL;
1390 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
1391 snprintf(str, str_size, "%llu", size);
1395 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
1398 suffix = unit_suffix_binary;
1399 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
1402 suffix = unit_suffix_decimal;
1407 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
1415 switch (unit_mode & UNITS_MODE_MASK) {
1416 case UNITS_TBYTES: base *= mult; num_divs++;
1417 case UNITS_GBYTES: base *= mult; num_divs++;
1418 case UNITS_MBYTES: base *= mult; num_divs++;
1419 case UNITS_KBYTES: num_divs++;
1426 while (size >= mult) {
1433 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
1435 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
1440 fraction = (float)last_size / base;
1442 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
1446 * __strncpy__null - strncpy with null termination
1447 * @dest: the target array
1448 * @src: the source string
1449 * @n: maximum bytes to copy (size of *dest)
1451 * Like strncpy, but ensures destination is null-terminated.
1453 * Copies the string pointed to by src, including the terminating null
1454 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
1455 * of n bytes. Then ensure that dest is null-terminated.
1457 char *__strncpy__null(char *dest, const char *src, size_t n)
1459 strncpy(dest, src, n);
1466 * Checks to make sure that the label matches our requirements.
1468 0 if everything is safe and usable
1469 -1 if the label is too long
1471 static int check_label(const char *input)
1473 int len = strlen(input);
1475 if (len > BTRFS_LABEL_SIZE - 1) {
1476 fprintf(stderr, "ERROR: Label %s is too long (max %d)\n",
1477 input, BTRFS_LABEL_SIZE - 1);
1484 static int set_label_unmounted(const char *dev, const char *label)
1486 struct btrfs_trans_handle *trans;
1487 struct btrfs_root *root;
1490 ret = check_mounted(dev);
1492 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1496 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1501 /* Open the super_block at the default location
1502 * and as read-write.
1504 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
1505 if (!root) /* errors are printed by open_ctree() */
1508 trans = btrfs_start_transaction(root, 1);
1509 snprintf(root->fs_info->super_copy->label, BTRFS_LABEL_SIZE, "%s",
1511 btrfs_commit_transaction(trans, root);
1513 /* Now we close it since we are done. */
1518 static int set_label_mounted(const char *mount_path, const char *label)
1522 fd = open(mount_path, O_RDONLY | O_NOATIME);
1524 fprintf(stderr, "ERROR: unable to access '%s'\n", mount_path);
1528 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
1529 fprintf(stderr, "ERROR: unable to set label %s\n",
1539 static int get_label_unmounted(const char *dev, char *label)
1541 struct btrfs_root *root;
1544 ret = check_mounted(dev);
1546 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1550 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1555 /* Open the super_block at the default location
1558 root = open_ctree(dev, 0, 0);
1562 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
1564 /* Now we close it since we are done. */
1570 * If a partition is mounted, try to get the filesystem label via its
1571 * mounted path rather than device. Return the corresponding error
1572 * the user specified the device path.
1574 int get_label_mounted(const char *mount_path, char *labelp)
1576 char label[BTRFS_LABEL_SIZE];
1579 fd = open(mount_path, O_RDONLY | O_NOATIME);
1581 fprintf(stderr, "ERROR: unable to access '%s'\n", mount_path);
1585 memset(label, '\0', sizeof(label));
1586 if (ioctl(fd, BTRFS_IOC_GET_FSLABEL, label) < 0) {
1587 fprintf(stderr, "ERROR: unable get label %s\n", strerror(errno));
1592 strncpy(labelp, label, sizeof(label));
1597 int get_label(const char *btrfs_dev, char *label)
1601 ret = is_existing_blk_or_reg_file(btrfs_dev);
1603 ret = get_label_mounted(btrfs_dev, label);
1605 ret = get_label_unmounted(btrfs_dev, label);
1610 int set_label(const char *btrfs_dev, const char *label)
1614 if (check_label(label))
1617 ret = is_existing_blk_or_reg_file(btrfs_dev);
1619 ret = set_label_mounted(btrfs_dev, label);
1621 ret = set_label_unmounted(btrfs_dev, label);
1626 int btrfs_scan_block_devices(int run_ioctl)
1632 struct btrfs_fs_devices *tmp_devices;
1634 FILE *proc_partitions;
1642 proc_partitions = fopen("/proc/partitions","r");
1643 if (!proc_partitions) {
1644 fprintf(stderr, "Unable to open '/proc/partitions' for scanning\n");
1647 /* skip the header */
1648 for (i = 0; i < 2; i++)
1649 if (!fgets(buf, 1023, proc_partitions)) {
1651 "Unable to read '/proc/partitions' for scanning\n");
1652 fclose(proc_partitions);
1656 strcpy(fullpath,"/dev/");
1657 while(fgets(buf, 1023, proc_partitions)) {
1658 ret = sscanf(buf," %*d %*d %*d %99s", fullpath + 5);
1661 "failed to scan device name from /proc/partitions\n");
1666 * multipath and MD devices may register as a btrfs filesystem
1667 * both through the original block device and through
1668 * the special (/dev/mapper or /dev/mdX) entry.
1669 * This scans the special entries last
1671 special = strncmp(fullpath, "/dev/dm-", strlen("/dev/dm-")) == 0;
1673 special = strncmp(fullpath, "/dev/md", strlen("/dev/md")) == 0;
1675 if (scans == 0 && special)
1677 if (scans > 0 && !special)
1680 ret = lstat(fullpath, &st);
1682 fprintf(stderr, "failed to stat %s\n", fullpath);
1685 if (!S_ISBLK(st.st_mode)) {
1689 fd = open(fullpath, O_RDONLY);
1691 if (errno != ENOMEDIUM)
1692 fprintf(stderr, "failed to open %s: %s\n",
1693 fullpath, strerror(errno));
1696 ret = btrfs_scan_one_device(fd, fullpath, &tmp_devices,
1698 BTRFS_SUPER_INFO_OFFSET, 0);
1699 if (ret == 0 && run_ioctl > 0) {
1700 btrfs_register_one_device(fullpath);
1705 fclose(proc_partitions);
1715 * A not-so-good version fls64. No fascinating optimization since
1716 * no one except parse_size use it
1718 static int fls64(u64 x)
1722 for (i = 0; i <64; i++)
1723 if (x << i & (1ULL << 63))
1728 u64 parse_size(char *s)
1736 fprintf(stderr, "ERROR: Size value is empty\n");
1741 "ERROR: Size value '%s' is less equal than 0\n", s);
1744 ret = strtoull(s, &endptr, 10);
1746 fprintf(stderr, "ERROR: Size value '%s' is invalid\n", s);
1749 if (endptr[0] && endptr[1]) {
1750 fprintf(stderr, "ERROR: Illegal suffix contains character '%c' in wrong position\n",
1755 * strtoll returns LLONG_MAX when overflow, if this happens,
1756 * need to call strtoull to get the real size
1758 if (errno == ERANGE && ret == ULLONG_MAX) {
1760 "ERROR: Size value '%s' is too large for u64\n", s);
1764 c = tolower(endptr[0]);
1787 fprintf(stderr, "ERROR: Unknown size descriptor '%c'\n",
1792 /* Check whether ret * mult overflow */
1793 if (fls64(ret) + fls64(mult) - 1 > 64) {
1795 "ERROR: Size value '%s' is too large for u64\n", s);
1802 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
1808 ret = stat(fname, &st);
1812 if (S_ISDIR(st.st_mode)) {
1813 *dirstream = opendir(fname);
1816 fd = dirfd(*dirstream);
1817 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
1818 fd = open(fname, open_flags);
1821 * we set this on purpose, in case the caller output
1822 * strerror(errno) as success
1830 closedir(*dirstream);
1835 int open_file_or_dir(const char *fname, DIR **dirstream)
1837 return open_file_or_dir3(fname, dirstream, O_RDWR);
1840 void close_file_or_dir(int fd, DIR *dirstream)
1843 closedir(dirstream);
1848 int get_device_info(int fd, u64 devid,
1849 struct btrfs_ioctl_dev_info_args *di_args)
1853 di_args->devid = devid;
1854 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
1856 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
1857 return ret ? -errno : 0;
1860 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
1863 struct btrfs_dev_item *dev_item;
1864 char *buf = search_args->buf;
1866 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
1867 + sizeof(struct btrfs_dev_item));
1868 buf += sizeof(struct btrfs_ioctl_search_header);
1870 dev_item = (struct btrfs_dev_item *)buf;
1872 return btrfs_stack_device_id(dev_item);
1875 static int search_chunk_tree_for_fs_info(int fd,
1876 struct btrfs_ioctl_fs_info_args *fi_args)
1880 u64 start_devid = 1;
1881 struct btrfs_ioctl_search_args search_args;
1882 struct btrfs_ioctl_search_key *search_key = &search_args.key;
1884 fi_args->num_devices = 0;
1886 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
1887 / (sizeof(struct btrfs_ioctl_search_header)
1888 + sizeof(struct btrfs_dev_item));
1890 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
1891 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
1892 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
1893 search_key->min_type = BTRFS_DEV_ITEM_KEY;
1894 search_key->max_type = BTRFS_DEV_ITEM_KEY;
1895 search_key->min_transid = 0;
1896 search_key->max_transid = (u64)-1;
1897 search_key->nr_items = max_items;
1898 search_key->max_offset = (u64)-1;
1901 search_key->min_offset = start_devid;
1903 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
1907 fi_args->num_devices += (u64)search_key->nr_items;
1909 if (search_key->nr_items == max_items) {
1910 start_devid = find_max_device_id(&search_args,
1911 search_key->nr_items) + 1;
1915 /* get the lastest max_id to stay consistent with the num_devices */
1916 if (search_key->nr_items == 0)
1918 * last tree_search returns an empty buf, use the devid of
1919 * the last dev_item of the previous tree_search
1921 fi_args->max_id = start_devid - 1;
1923 fi_args->max_id = find_max_device_id(&search_args,
1924 search_key->nr_items);
1930 * For a given path, fill in the ioctl fs_ and info_ args.
1931 * If the path is a btrfs mountpoint, fill info for all devices.
1932 * If the path is a btrfs device, fill in only that device.
1934 * The path provided must be either on a mounted btrfs fs,
1935 * or be a mounted btrfs device.
1937 * Returns 0 on success, or a negative errno.
1939 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
1940 struct btrfs_ioctl_dev_info_args **di_ret)
1946 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1947 struct btrfs_ioctl_dev_info_args *di_args;
1948 char mp[BTRFS_PATH_NAME_MAX + 1];
1949 DIR *dirstream = NULL;
1951 memset(fi_args, 0, sizeof(*fi_args));
1953 if (is_block_device(path)) {
1954 struct btrfs_super_block *disk_super;
1955 char buf[BTRFS_SUPER_INFO_SIZE];
1958 /* Ensure it's mounted, then set path to the mountpoint */
1959 fd = open(path, O_RDONLY);
1962 fprintf(stderr, "Couldn't open %s: %s\n",
1963 path, strerror(errno));
1966 ret = check_mounted_where(fd, path, mp, sizeof(mp),
1975 /* Only fill in this one device */
1976 fi_args->num_devices = 1;
1978 disk_super = (struct btrfs_super_block *)buf;
1979 ret = btrfs_read_dev_super(fd, disk_super,
1980 BTRFS_SUPER_INFO_OFFSET, 0);
1985 devid = btrfs_stack_device_id(&disk_super->dev_item);
1987 fi_args->max_id = devid;
1990 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
1994 /* at this point path must not be for a block device */
1995 fd = open_file_or_dir(path, &dirstream);
2001 /* fill in fi_args if not just a single device */
2002 if (fi_args->num_devices != 1) {
2003 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
2010 * The fs_args->num_devices does not include seed devices
2012 ret = search_chunk_tree_for_fs_info(fd, fi_args);
2017 if (!fi_args->num_devices)
2020 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
2026 for (; i <= fi_args->max_id; ++i) {
2027 ret = get_device_info(fd, i, &di_args[ndevs]);
2036 * only when the only dev we wanted to find is not there then
2037 * let any error be returned
2039 if (fi_args->num_devices != 1) {
2045 close_file_or_dir(fd, dirstream);
2049 #define isoctal(c) (((c) & ~7) == '0')
2051 static inline void translate(char *f, char *t)
2053 while (*f != '\0') {
2055 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
2056 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
2066 * Checks if the swap device.
2067 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
2069 static int is_swap_device(const char *file)
2080 if (stat(file, &st_buf) < 0)
2082 if (S_ISBLK(st_buf.st_mode))
2083 dev = st_buf.st_rdev;
2084 else if (S_ISREG(st_buf.st_mode)) {
2085 dev = st_buf.st_dev;
2086 ino = st_buf.st_ino;
2090 if ((f = fopen("/proc/swaps", "r")) == NULL)
2093 /* skip the first line */
2094 if (fgets(tmp, sizeof(tmp), f) == NULL)
2097 while (fgets(tmp, sizeof(tmp), f) != NULL) {
2098 if ((cp = strchr(tmp, ' ')) != NULL)
2100 if ((cp = strchr(tmp, '\t')) != NULL)
2102 translate(tmp, buf);
2103 if (stat(buf, &st_buf) != 0)
2105 if (S_ISBLK(st_buf.st_mode)) {
2106 if (dev == st_buf.st_rdev) {
2110 } else if (S_ISREG(st_buf.st_mode)) {
2111 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
2125 * Check for existing filesystem or partition table on device.
2127 * 1 for existing fs or partition
2128 * 0 for nothing found
2129 * -1 for internal error
2136 blkid_probe pr = NULL;
2140 if (!device || !*device)
2143 ret = -1; /* will reset on success of all setup calls */
2145 pr = blkid_new_probe_from_filename(device);
2149 size = blkid_probe_get_size(pr);
2153 /* nothing to overwrite on a 0-length device */
2159 ret = blkid_probe_enable_partitions(pr, 1);
2163 ret = blkid_do_fullprobe(pr);
2168 * Blkid returns 1 for nothing found and 0 when it finds a signature,
2169 * but we want the exact opposite, so reverse the return value here.
2171 * In addition print some useful diagnostics about what actually is
2179 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
2181 "%s appears to contain an existing "
2182 "filesystem (%s).\n", device, type);
2183 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
2185 "%s appears to contain a partition "
2186 "table (%s).\n", device, type);
2189 "%s appears to contain something weird "
2190 "according to blkid\n", device);
2196 blkid_free_probe(pr);
2199 "probe of %s failed, cannot detect "
2200 "existing filesystem.\n", device);
2204 static int group_profile_devs_min(u64 flag)
2206 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2207 case 0: /* single */
2208 case BTRFS_BLOCK_GROUP_DUP:
2210 case BTRFS_BLOCK_GROUP_RAID0:
2211 case BTRFS_BLOCK_GROUP_RAID1:
2212 case BTRFS_BLOCK_GROUP_RAID5:
2214 case BTRFS_BLOCK_GROUP_RAID6:
2216 case BTRFS_BLOCK_GROUP_RAID10:
2223 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
2224 u64 dev_cnt, int mixed, char *estr)
2232 allowed |= BTRFS_BLOCK_GROUP_RAID10;
2234 allowed |= BTRFS_BLOCK_GROUP_RAID6;
2236 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
2237 BTRFS_BLOCK_GROUP_RAID5;
2240 allowed |= BTRFS_BLOCK_GROUP_DUP;
2244 ((metadata_profile | data_profile) & BTRFS_BLOCK_GROUP_DUP)) {
2246 "DUP is not allowed when FS has multiple devices\n");
2249 if (metadata_profile & ~allowed) {
2251 "unable to create FS with metadata profile %s "
2252 "(have %llu devices but %d devices are required)\n",
2253 btrfs_group_profile_str(metadata_profile), dev_cnt,
2254 group_profile_devs_min(metadata_profile));
2257 if (data_profile & ~allowed) {
2259 "unable to create FS with data profile %s "
2260 "(have %llu devices but %d devices are required)\n",
2261 btrfs_group_profile_str(data_profile), dev_cnt,
2262 group_profile_devs_min(data_profile));
2266 if (!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP)) {
2268 "dup for data is allowed only in mixed mode");
2274 /* Check if disk is suitable for btrfs
2276 * 1: something is wrong, estr provides the error
2279 int test_dev_for_mkfs(char *file, int force_overwrite, char *estr)
2285 ret = is_swap_device(file);
2287 snprintf(estr, sz, "error checking %s status: %s\n", file,
2292 snprintf(estr, sz, "%s is a swap device\n", file);
2295 if (!force_overwrite) {
2296 if (check_overwrite(file)) {
2297 snprintf(estr, sz, "Use the -f option to force overwrite.\n");
2301 ret = check_mounted(file);
2303 snprintf(estr, sz, "error checking %s mount status\n",
2308 snprintf(estr, sz, "%s is mounted\n", file);
2311 /* check if the device is busy */
2312 fd = open(file, O_RDWR|O_EXCL);
2314 snprintf(estr, sz, "unable to open %s: %s\n", file,
2318 if (fstat(fd, &st)) {
2319 snprintf(estr, sz, "unable to stat %s: %s\n", file,
2324 if (!S_ISBLK(st.st_mode)) {
2325 fprintf(stderr, "'%s' is not a block device\n", file);
2333 int btrfs_scan_lblkid()
2338 struct btrfs_fs_devices *tmp_devices;
2339 blkid_dev_iterate iter = NULL;
2340 blkid_dev dev = NULL;
2341 blkid_cache cache = NULL;
2342 char path[PATH_MAX];
2344 if (btrfs_scan_done)
2347 if (blkid_get_cache(&cache, 0) < 0) {
2348 printf("ERROR: lblkid cache get failed\n");
2351 blkid_probe_all(cache);
2352 iter = blkid_dev_iterate_begin(cache);
2353 blkid_dev_set_search(iter, "TYPE", "btrfs");
2354 while (blkid_dev_next(iter, &dev) == 0) {
2355 dev = blkid_verify(cache, dev);
2358 /* if we are here its definitely a btrfs disk*/
2359 strncpy_null(path, blkid_dev_devname(dev));
2361 fd = open(path, O_RDONLY);
2363 printf("ERROR: could not open %s\n", path);
2366 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
2367 &num_devices, BTRFS_SUPER_INFO_OFFSET, 0);
2369 printf("ERROR: could not scan %s\n", path);
2376 blkid_dev_iterate_end(iter);
2377 blkid_put_cache(cache);
2379 btrfs_scan_done = 1;
2384 int is_vol_small(char *file)
2391 fd = open(file, O_RDONLY);
2394 if (fstat(fd, &st) < 0) {
2399 size = btrfs_device_size(fd, &st);
2404 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
2414 * This reads a line from the stdin and only returns non-zero if the
2415 * first whitespace delimited token is a case insensitive match with yes
2418 int ask_user(char *question)
2420 char buf[30] = {0,};
2421 char *saveptr = NULL;
2424 printf("%s [y/N]: ", question);
2426 return fgets(buf, sizeof(buf) - 1, stdin) &&
2427 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
2428 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
2433 * - file or directory return the containing tree root id
2434 * - subvolume return its own tree id
2435 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
2436 * undefined and function returns -1
2438 int lookup_ino_rootid(int fd, u64 *rootid)
2440 struct btrfs_ioctl_ino_lookup_args args;
2444 memset(&args, 0, sizeof(args));
2446 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
2448 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
2451 fprintf(stderr, "ERROR: Failed to lookup root id - %s\n",
2456 *rootid = args.treeid;
2462 * return 0 if a btrfs mount point is found
2463 * return 1 if a mount point is found but not btrfs
2464 * return <0 if something goes wrong
2466 int find_mount_root(const char *path, char **mount_root)
2474 int longest_matchlen = 0;
2475 char *longest_match = NULL;
2477 fd = open(path, O_RDONLY | O_NOATIME);
2482 mnttab = setmntent("/proc/self/mounts", "r");
2486 while ((ent = getmntent(mnttab))) {
2487 len = strlen(ent->mnt_dir);
2488 if (strncmp(ent->mnt_dir, path, len) == 0) {
2489 /* match found and use the latest match */
2490 if (longest_matchlen <= len) {
2491 free(longest_match);
2492 longest_matchlen = len;
2493 longest_match = strdup(ent->mnt_dir);
2494 not_btrfs = strcmp(ent->mnt_type, "btrfs");
2503 free(longest_match);
2508 *mount_root = realpath(longest_match, NULL);
2512 free(longest_match);
2516 int test_minimum_size(const char *file, u32 leafsize)
2519 struct stat statbuf;
2521 fd = open(file, O_RDONLY);
2524 if (stat(file, &statbuf) < 0) {
2528 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(leafsize)) {
2537 * test if name is a correct subvolume name
2538 * this function return
2539 * 0-> name is not a correct subvolume name
2540 * 1-> name is a correct subvolume name
2542 int test_issubvolname(const char *name)
2544 return name[0] != '\0' && !strchr(name, '/') &&
2545 strcmp(name, ".") && strcmp(name, "..");
2549 * test if path is a directory
2550 * this function return
2551 * 0-> path exists but it is not a directory
2552 * 1-> path exists and it is a directory
2553 * -1 -> path is unaccessible
2555 int test_isdir(const char *path)
2560 ret = stat(path, &st);
2564 return S_ISDIR(st.st_mode);
2567 void units_set_mode(unsigned *units, unsigned mode)
2569 unsigned base = *units & UNITS_MODE_MASK;
2571 *units = base | mode;
2574 void units_set_base(unsigned *units, unsigned base)
2576 unsigned mode = *units & ~UNITS_MODE_MASK;
2578 *units = base | mode;
2581 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
2585 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
2586 if (!path->nodes[level])
2588 if (path->slots[level] + 1 >=
2589 btrfs_header_nritems(path->nodes[level]))
2592 btrfs_item_key_to_cpu(path->nodes[level], key,
2593 path->slots[level] + 1);
2595 btrfs_node_key_to_cpu(path->nodes[level], key,
2596 path->slots[level] + 1);
2602 char* btrfs_group_type_str(u64 flag)
2604 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
2605 BTRFS_SPACE_INFO_GLOBAL_RSV;
2607 switch (flag & mask) {
2608 case BTRFS_BLOCK_GROUP_DATA:
2610 case BTRFS_BLOCK_GROUP_SYSTEM:
2612 case BTRFS_BLOCK_GROUP_METADATA:
2614 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
2615 return "Data+Metadata";
2616 case BTRFS_SPACE_INFO_GLOBAL_RSV:
2617 return "GlobalReserve";
2623 char* btrfs_group_profile_str(u64 flag)
2625 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2628 case BTRFS_BLOCK_GROUP_RAID0:
2630 case BTRFS_BLOCK_GROUP_RAID1:
2632 case BTRFS_BLOCK_GROUP_RAID5:
2634 case BTRFS_BLOCK_GROUP_RAID6:
2636 case BTRFS_BLOCK_GROUP_DUP:
2638 case BTRFS_BLOCK_GROUP_RAID10:
2645 u64 disk_size(char *path)
2649 if (statfs(path, &sfs) < 0)
2652 return sfs.f_bsize * sfs.f_blocks;
2655 u64 get_partition_size(char *dev)
2658 int fd = open(dev, O_RDONLY);
2662 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
2671 int btrfs_tree_search2_ioctl_supported(int fd)
2673 struct btrfs_ioctl_search_args_v2 *args2;
2674 struct btrfs_ioctl_search_key *sk;
2675 int args2_size = 1024;
2676 char args2_buf[args2_size];
2678 static int v2_supported = -1;
2680 if (v2_supported != -1)
2681 return v2_supported;
2683 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
2687 * Search for the extent tree item in the root tree.
2689 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
2690 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
2691 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
2692 sk->min_type = BTRFS_ROOT_ITEM_KEY;
2693 sk->max_type = BTRFS_ROOT_ITEM_KEY;
2695 sk->max_offset = (u64)-1;
2696 sk->min_transid = 0;
2697 sk->max_transid = (u64)-1;
2699 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
2700 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
2701 if (ret == -EOPNOTSUPP)
2708 return v2_supported;