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.
23 #include <sys/ioctl.h>
24 #include <sys/mount.h>
25 #include <sys/types.h>
27 #include <uuid/uuid.h>
32 #include <linux/loop.h>
33 #include <linux/major.h>
34 #include <linux/kdev_t.h>
36 #include <blkid/blkid.h>
39 #include "kerncompat.h"
40 #include "radix-tree.h"
43 #include "transaction.h"
50 #define BLKDISCARD _IO(0x12,119)
53 static int btrfs_scan_done = 0;
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 strncpy(argv0_buf, argv[0], sizeof(argv0_buf));
68 argv0_buf[sizeof(argv0_buf) - 1] = 0;
71 int check_argc_exact(int nargs, int expected)
74 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
76 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
78 return nargs != expected;
81 int check_argc_min(int nargs, int expected)
83 if (nargs < expected) {
84 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
91 int check_argc_max(int nargs, int expected)
93 if (nargs > expected) {
94 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
103 * Discard the given range in one go
105 static int discard_range(int fd, u64 start, u64 len)
107 u64 range[2] = { start, len };
109 if (ioctl(fd, BLKDISCARD, &range) < 0)
115 * Discard blocks in the given range in 1G chunks, the process is interruptible
117 static int discard_blocks(int fd, u64 start, u64 len)
121 u64 chunk_size = min_t(u64, len, 1*1024*1024*1024);
124 ret = discard_range(fd, start, chunk_size);
134 static u64 reference_root_table[] = {
135 [1] = BTRFS_ROOT_TREE_OBJECTID,
136 [2] = BTRFS_EXTENT_TREE_OBJECTID,
137 [3] = BTRFS_CHUNK_TREE_OBJECTID,
138 [4] = BTRFS_DEV_TREE_OBJECTID,
139 [5] = BTRFS_FS_TREE_OBJECTID,
140 [6] = BTRFS_CSUM_TREE_OBJECTID,
143 int test_uuid_unique(char *fs_uuid)
146 blkid_dev_iterate iter = NULL;
147 blkid_dev dev = NULL;
148 blkid_cache cache = NULL;
150 if (blkid_get_cache(&cache, 0) < 0) {
151 printf("ERROR: lblkid cache get failed\n");
154 blkid_probe_all(cache);
155 iter = blkid_dev_iterate_begin(cache);
156 blkid_dev_set_search(iter, "UUID", fs_uuid);
158 while (blkid_dev_next(iter, &dev) == 0) {
159 dev = blkid_verify(cache, dev);
166 blkid_dev_iterate_end(iter);
167 blkid_put_cache(cache);
172 int make_btrfs(int fd, const char *device, const char *label, char *fs_uuid,
173 u64 blocks[7], u64 num_bytes, u32 nodesize,
174 u32 leafsize, u32 sectorsize, u32 stripesize, u64 features)
176 struct btrfs_super_block super;
177 struct extent_buffer *buf = NULL;
178 struct btrfs_root_item root_item;
179 struct btrfs_disk_key disk_key;
180 struct btrfs_extent_item *extent_item;
181 struct btrfs_inode_item *inode_item;
182 struct btrfs_chunk *chunk;
183 struct btrfs_dev_item *dev_item;
184 struct btrfs_dev_extent *dev_extent;
185 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
195 int skinny_metadata = !!(features &
196 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
198 first_free = BTRFS_SUPER_INFO_OFFSET + sectorsize * 2 - 1;
199 first_free &= ~((u64)sectorsize - 1);
201 memset(&super, 0, sizeof(super));
203 num_bytes = (num_bytes / sectorsize) * sectorsize;
205 if (uuid_parse(fs_uuid, super.fsid) != 0) {
206 fprintf(stderr, "could not parse UUID: %s\n", fs_uuid);
210 if (!test_uuid_unique(fs_uuid)) {
211 fprintf(stderr, "non-unique UUID: %s\n", fs_uuid);
216 uuid_generate(super.fsid);
218 uuid_generate(super.dev_item.uuid);
219 uuid_generate(chunk_tree_uuid);
221 btrfs_set_super_bytenr(&super, blocks[0]);
222 btrfs_set_super_num_devices(&super, 1);
223 btrfs_set_super_magic(&super, BTRFS_MAGIC);
224 btrfs_set_super_generation(&super, 1);
225 btrfs_set_super_root(&super, blocks[1]);
226 btrfs_set_super_chunk_root(&super, blocks[3]);
227 btrfs_set_super_total_bytes(&super, num_bytes);
228 btrfs_set_super_bytes_used(&super, 6 * leafsize);
229 btrfs_set_super_sectorsize(&super, sectorsize);
230 btrfs_set_super_leafsize(&super, leafsize);
231 btrfs_set_super_nodesize(&super, nodesize);
232 btrfs_set_super_stripesize(&super, stripesize);
233 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
234 btrfs_set_super_chunk_root_generation(&super, 1);
235 btrfs_set_super_cache_generation(&super, -1);
236 btrfs_set_super_incompat_flags(&super, features);
238 strncpy(super.label, label, BTRFS_LABEL_SIZE - 1);
240 buf = malloc(sizeof(*buf) + max(sectorsize, leafsize));
242 /* create the tree of root objects */
243 memset(buf->data, 0, leafsize);
245 btrfs_set_header_bytenr(buf, blocks[1]);
246 btrfs_set_header_nritems(buf, 4);
247 btrfs_set_header_generation(buf, 1);
248 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
249 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
250 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
253 write_extent_buffer(buf, chunk_tree_uuid,
254 btrfs_header_chunk_tree_uuid(buf),
257 /* create the items for the root tree */
258 memset(&root_item, 0, sizeof(root_item));
259 inode_item = &root_item.inode;
260 btrfs_set_stack_inode_generation(inode_item, 1);
261 btrfs_set_stack_inode_size(inode_item, 3);
262 btrfs_set_stack_inode_nlink(inode_item, 1);
263 btrfs_set_stack_inode_nbytes(inode_item, leafsize);
264 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
265 btrfs_set_root_refs(&root_item, 1);
266 btrfs_set_root_used(&root_item, leafsize);
267 btrfs_set_root_generation(&root_item, 1);
269 memset(&disk_key, 0, sizeof(disk_key));
270 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
271 btrfs_set_disk_key_offset(&disk_key, 0);
274 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - sizeof(root_item);
275 btrfs_set_root_bytenr(&root_item, blocks[2]);
276 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
277 btrfs_set_item_key(buf, &disk_key, nritems);
278 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
279 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
281 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
282 nritems), sizeof(root_item));
285 itemoff = itemoff - sizeof(root_item);
286 btrfs_set_root_bytenr(&root_item, blocks[4]);
287 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
288 btrfs_set_item_key(buf, &disk_key, nritems);
289 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
290 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
292 write_extent_buffer(buf, &root_item,
293 btrfs_item_ptr_offset(buf, nritems),
297 itemoff = itemoff - sizeof(root_item);
298 btrfs_set_root_bytenr(&root_item, blocks[5]);
299 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
300 btrfs_set_item_key(buf, &disk_key, nritems);
301 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
302 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
304 write_extent_buffer(buf, &root_item,
305 btrfs_item_ptr_offset(buf, nritems),
309 itemoff = itemoff - sizeof(root_item);
310 btrfs_set_root_bytenr(&root_item, blocks[6]);
311 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
312 btrfs_set_item_key(buf, &disk_key, nritems);
313 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
314 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
316 write_extent_buffer(buf, &root_item,
317 btrfs_item_ptr_offset(buf, nritems),
322 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
323 ret = pwrite(fd, buf->data, leafsize, blocks[1]);
324 if (ret != leafsize) {
325 ret = (ret < 0 ? -errno : -EIO);
329 /* create the items for the extent tree */
330 memset(buf->data+sizeof(struct btrfs_header), 0,
331 leafsize-sizeof(struct btrfs_header));
333 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize);
334 for (i = 1; i < 7; i++) {
335 item_size = sizeof(struct btrfs_extent_item);
336 if (!skinny_metadata)
337 item_size += sizeof(struct btrfs_tree_block_info);
339 BUG_ON(blocks[i] < first_free);
340 BUG_ON(blocks[i] < blocks[i - 1]);
342 /* create extent item */
343 itemoff -= item_size;
344 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
345 if (skinny_metadata) {
346 btrfs_set_disk_key_type(&disk_key,
347 BTRFS_METADATA_ITEM_KEY);
348 btrfs_set_disk_key_offset(&disk_key, 0);
350 btrfs_set_disk_key_type(&disk_key,
351 BTRFS_EXTENT_ITEM_KEY);
352 btrfs_set_disk_key_offset(&disk_key, leafsize);
354 btrfs_set_item_key(buf, &disk_key, nritems);
355 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
357 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
359 extent_item = btrfs_item_ptr(buf, nritems,
360 struct btrfs_extent_item);
361 btrfs_set_extent_refs(buf, extent_item, 1);
362 btrfs_set_extent_generation(buf, extent_item, 1);
363 btrfs_set_extent_flags(buf, extent_item,
364 BTRFS_EXTENT_FLAG_TREE_BLOCK);
367 /* create extent ref */
368 ref_root = reference_root_table[i];
369 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
370 btrfs_set_disk_key_offset(&disk_key, ref_root);
371 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
372 btrfs_set_item_key(buf, &disk_key, nritems);
373 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
375 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
378 btrfs_set_header_bytenr(buf, blocks[2]);
379 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
380 btrfs_set_header_nritems(buf, nritems);
381 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
382 ret = pwrite(fd, buf->data, leafsize, blocks[2]);
383 if (ret != leafsize) {
384 ret = (ret < 0 ? -errno : -EIO);
388 /* create the chunk tree */
389 memset(buf->data+sizeof(struct btrfs_header), 0,
390 leafsize-sizeof(struct btrfs_header));
392 item_size = sizeof(*dev_item);
393 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - item_size;
395 /* first device 1 (there is no device 0) */
396 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
397 btrfs_set_disk_key_offset(&disk_key, 1);
398 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
399 btrfs_set_item_key(buf, &disk_key, nritems);
400 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
401 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
403 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
404 btrfs_set_device_id(buf, dev_item, 1);
405 btrfs_set_device_generation(buf, dev_item, 0);
406 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
407 btrfs_set_device_bytes_used(buf, dev_item,
408 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
409 btrfs_set_device_io_align(buf, dev_item, sectorsize);
410 btrfs_set_device_io_width(buf, dev_item, sectorsize);
411 btrfs_set_device_sector_size(buf, dev_item, sectorsize);
412 btrfs_set_device_type(buf, dev_item, 0);
414 write_extent_buffer(buf, super.dev_item.uuid,
415 (unsigned long)btrfs_device_uuid(dev_item),
417 write_extent_buffer(buf, super.fsid,
418 (unsigned long)btrfs_device_fsid(dev_item),
420 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
424 item_size = btrfs_chunk_item_size(1);
425 itemoff = itemoff - item_size;
427 /* then we have chunk 0 */
428 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
429 btrfs_set_disk_key_offset(&disk_key, 0);
430 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
431 btrfs_set_item_key(buf, &disk_key, nritems);
432 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
433 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
435 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
436 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
437 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
438 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
439 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
440 btrfs_set_chunk_io_align(buf, chunk, sectorsize);
441 btrfs_set_chunk_io_width(buf, chunk, sectorsize);
442 btrfs_set_chunk_sector_size(buf, chunk, sectorsize);
443 btrfs_set_chunk_num_stripes(buf, chunk, 1);
444 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
445 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
448 write_extent_buffer(buf, super.dev_item.uuid,
449 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
452 /* copy the key for the chunk to the system array */
453 ptr = super.sys_chunk_array;
454 array_size = sizeof(disk_key);
456 memcpy(ptr, &disk_key, sizeof(disk_key));
457 ptr += sizeof(disk_key);
459 /* copy the chunk to the system array */
460 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
461 array_size += item_size;
463 btrfs_set_super_sys_array_size(&super, array_size);
465 btrfs_set_header_bytenr(buf, blocks[3]);
466 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
467 btrfs_set_header_nritems(buf, nritems);
468 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
469 ret = pwrite(fd, buf->data, leafsize, blocks[3]);
470 if (ret != leafsize) {
471 ret = (ret < 0 ? -errno : -EIO);
475 /* create the device tree */
476 memset(buf->data+sizeof(struct btrfs_header), 0,
477 leafsize-sizeof(struct btrfs_header));
479 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) -
480 sizeof(struct btrfs_dev_extent);
482 btrfs_set_disk_key_objectid(&disk_key, 1);
483 btrfs_set_disk_key_offset(&disk_key, 0);
484 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
485 btrfs_set_item_key(buf, &disk_key, nritems);
486 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
487 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
488 sizeof(struct btrfs_dev_extent));
489 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
490 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
491 BTRFS_CHUNK_TREE_OBJECTID);
492 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
493 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
494 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
496 write_extent_buffer(buf, chunk_tree_uuid,
497 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
500 btrfs_set_dev_extent_length(buf, dev_extent,
501 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
504 btrfs_set_header_bytenr(buf, blocks[4]);
505 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
506 btrfs_set_header_nritems(buf, nritems);
507 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
508 ret = pwrite(fd, buf->data, leafsize, blocks[4]);
509 if (ret != leafsize) {
510 ret = (ret < 0 ? -errno : -EIO);
514 /* create the FS root */
515 memset(buf->data+sizeof(struct btrfs_header), 0,
516 leafsize-sizeof(struct btrfs_header));
517 btrfs_set_header_bytenr(buf, blocks[5]);
518 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
519 btrfs_set_header_nritems(buf, 0);
520 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
521 ret = pwrite(fd, buf->data, leafsize, blocks[5]);
522 if (ret != leafsize) {
523 ret = (ret < 0 ? -errno : -EIO);
526 /* finally create the csum root */
527 memset(buf->data+sizeof(struct btrfs_header), 0,
528 leafsize-sizeof(struct btrfs_header));
529 btrfs_set_header_bytenr(buf, blocks[6]);
530 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
531 btrfs_set_header_nritems(buf, 0);
532 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
533 ret = pwrite(fd, buf->data, leafsize, blocks[6]);
534 if (ret != leafsize) {
535 ret = (ret < 0 ? -errno : -EIO);
539 /* and write out the super block */
540 BUG_ON(sizeof(super) > sectorsize);
541 memset(buf->data, 0, sectorsize);
542 memcpy(buf->data, &super, sizeof(super));
543 buf->len = sectorsize;
544 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
545 ret = pwrite(fd, buf->data, sectorsize, blocks[0]);
546 if (ret != sectorsize) {
547 ret = (ret < 0 ? -errno : -EIO);
558 u64 btrfs_device_size(int fd, struct stat *st)
561 if (S_ISREG(st->st_mode)) {
564 if (!S_ISBLK(st->st_mode)) {
567 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
573 static int zero_blocks(int fd, off_t start, size_t len)
575 char *buf = malloc(len);
582 written = pwrite(fd, buf, len, start);
589 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
591 /* don't write outside the device by clamping the region to the device size */
592 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
594 off_t end = max(start, start + len);
597 /* and don't overwrite the disk labels on sparc */
598 start = max(start, 1024);
599 end = max(end, 1024);
602 start = min_t(u64, start, dev_size);
603 end = min_t(u64, end, dev_size);
605 return zero_blocks(fd, start, end - start);
608 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
609 struct btrfs_root *root, int fd, char *path,
610 u64 block_count, u32 io_width, u32 io_align,
613 struct btrfs_super_block *disk_super;
614 struct btrfs_super_block *super = root->fs_info->super_copy;
615 struct btrfs_device *device;
616 struct btrfs_dev_item *dev_item;
622 device = kzalloc(sizeof(*device), GFP_NOFS);
625 buf = kmalloc(sectorsize, GFP_NOFS);
630 BUG_ON(sizeof(*disk_super) > sectorsize);
631 memset(buf, 0, sectorsize);
633 disk_super = (struct btrfs_super_block *)buf;
634 dev_item = &disk_super->dev_item;
636 uuid_generate(device->uuid);
639 device->io_width = io_width;
640 device->io_align = io_align;
641 device->sector_size = sectorsize;
643 device->writeable = 1;
644 device->total_bytes = block_count;
645 device->bytes_used = 0;
646 device->total_ios = 0;
647 device->dev_root = root->fs_info->dev_root;
649 ret = btrfs_add_device(trans, root, device);
652 total_bytes = btrfs_super_total_bytes(super) + block_count;
653 btrfs_set_super_total_bytes(super, total_bytes);
655 num_devs = btrfs_super_num_devices(super) + 1;
656 btrfs_set_super_num_devices(super, num_devs);
658 memcpy(disk_super, super, sizeof(*disk_super));
660 printf("adding device %s id %llu\n", path,
661 (unsigned long long)device->devid);
663 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
664 btrfs_set_stack_device_id(dev_item, device->devid);
665 btrfs_set_stack_device_type(dev_item, device->type);
666 btrfs_set_stack_device_io_align(dev_item, device->io_align);
667 btrfs_set_stack_device_io_width(dev_item, device->io_width);
668 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
669 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
670 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
671 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
673 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
674 BUG_ON(ret != sectorsize);
677 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
678 device->fs_devices = root->fs_info->fs_devices;
682 static void btrfs_wipe_existing_sb(int fd)
684 const char *off = NULL;
689 blkid_probe pr = NULL;
691 pr = blkid_new_probe();
695 if (blkid_probe_set_device(pr, fd, 0, 0))
698 rc = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
700 rc = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
702 if (rc || len == 0 || off == NULL)
705 offset = strtoll(off, NULL, 10);
706 if (len > sizeof(buf))
710 rc = pwrite(fd, buf, len, offset);
714 blkid_free_probe(pr);
718 int btrfs_prepare_device(int fd, char *file, int zero_end, u64 *block_count_ret,
719 u64 max_block_count, int *mixed, int discard)
725 ret = fstat(fd, &st);
727 fprintf(stderr, "unable to stat %s\n", file);
731 block_count = btrfs_device_size(fd, &st);
732 if (block_count == 0) {
733 fprintf(stderr, "unable to find %s size\n", file);
737 block_count = min(block_count, max_block_count);
739 if (block_count < BTRFS_MKFS_SMALL_VOLUME_SIZE && !(*mixed))
744 * We intentionally ignore errors from the discard ioctl. It
745 * is not necessary for the mkfs functionality but just an
748 if (discard_range(fd, 0, 0) == 0) {
749 printf("Performing full device TRIM (%s) ...\n",
750 pretty_size(block_count));
751 discard_blocks(fd, 0, block_count);
755 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
756 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
757 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
758 BTRFS_SUPER_INFO_SIZE, block_count);
759 if (!ret && zero_end)
760 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
761 ZERO_DEV_BYTES, block_count);
764 fprintf(stderr, "ERROR: failed to zero device '%s' - %s\n",
765 file, strerror(-ret));
769 btrfs_wipe_existing_sb(fd);
771 *block_count_ret = block_count;
775 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
776 struct btrfs_root *root, u64 objectid)
779 struct btrfs_inode_item inode_item;
780 time_t now = time(NULL);
782 memset(&inode_item, 0, sizeof(inode_item));
783 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
784 btrfs_set_stack_inode_size(&inode_item, 0);
785 btrfs_set_stack_inode_nlink(&inode_item, 1);
786 btrfs_set_stack_inode_nbytes(&inode_item, root->leafsize);
787 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
788 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
789 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
790 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
791 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
792 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
793 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
794 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
795 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
797 if (root->fs_info->tree_root == root)
798 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
800 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
804 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
808 btrfs_set_root_dirid(&root->root_item, objectid);
815 * checks if a path is a block device node
816 * Returns negative errno on failure, otherwise
817 * returns 1 for blockdev, 0 for not-blockdev
819 int is_block_device(const char *path)
823 if (stat(path, &statbuf) < 0)
826 return S_ISBLK(statbuf.st_mode);
830 * check if given path is a mount point
831 * return 1 if yes. 0 if no. -1 for error
833 int is_mount_point(const char *path)
839 f = setmntent("/proc/self/mounts", "r");
843 while ((mnt = getmntent(f)) != NULL) {
844 if (strcmp(mnt->mnt_dir, path))
853 static int is_reg_file(const char *path)
857 if (stat(path, &statbuf) < 0)
859 return S_ISREG(statbuf.st_mode);
863 * This function checks if the given input parameter is
865 * return <0 : some error in the given input
866 * return BTRFS_ARG_UNKNOWN: unknown input
867 * return BTRFS_ARG_UUID: given input is uuid
868 * return BTRFS_ARG_MNTPOINT: given input is path
869 * return BTRFS_ARG_REG: given input is regular file
871 int check_arg_type(const char *input)
879 if (realpath(input, path)) {
880 if (is_block_device(path) == 1)
881 return BTRFS_ARG_BLKDEV;
883 if (is_mount_point(path) == 1)
884 return BTRFS_ARG_MNTPOINT;
886 if (is_reg_file(path))
887 return BTRFS_ARG_REG;
889 return BTRFS_ARG_UNKNOWN;
892 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
893 !uuid_parse(input, uuid))
894 return BTRFS_ARG_UUID;
896 return BTRFS_ARG_UNKNOWN;
900 * Find the mount point for a mounted device.
901 * On success, returns 0 with mountpoint in *mp.
902 * On failure, returns -errno (not mounted yields -EINVAL)
903 * Is noisy on failures, expects to be given a mounted device.
905 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
910 ret = is_block_device(dev);
913 fprintf(stderr, "%s is not a block device\n", dev);
916 fprintf(stderr, "Could not check %s: %s\n",
917 dev, strerror(-ret));
922 fd = open(dev, O_RDONLY);
925 fprintf(stderr, "Could not open %s: %s\n", dev, strerror(errno));
929 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
932 } else { /* mounted, all good */
942 * Given a pathname, return a filehandle to:
943 * the original pathname or,
944 * if the pathname is a mounted btrfs device, to its mountpoint.
946 * On error, return -1, errno should be set.
948 int open_path_or_dev_mnt(const char *path, DIR **dirstream)
950 char mp[BTRFS_PATH_NAME_MAX + 1];
953 if (is_block_device(path)) {
956 ret = get_btrfs_mount(path, mp, sizeof(mp));
958 /* not a mounted btrfs dev */
962 fdmnt = open_file_or_dir(mp, dirstream);
964 fdmnt = open_file_or_dir(path, dirstream);
970 /* checks if a device is a loop device */
971 static int is_loop_device (const char* device) {
974 if(stat(device, &statbuf) < 0)
977 return (S_ISBLK(statbuf.st_mode) &&
978 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
982 /* Takes a loop device path (e.g. /dev/loop0) and returns
983 * the associated file (e.g. /images/my_btrfs.img) */
984 static int resolve_loop_device(const char* loop_dev, char* loop_file,
991 char real_loop_dev[PATH_MAX];
993 if (!realpath(loop_dev, real_loop_dev))
995 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
996 if (!(f = fopen(p, "r")))
999 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
1000 ret = fscanf(f, fmt, loop_file);
1009 * Checks whether a and b are identical or device
1010 * files associated with the same block device
1012 static int is_same_blk_file(const char* a, const char* b)
1014 struct stat st_buf_a, st_buf_b;
1015 char real_a[PATH_MAX];
1016 char real_b[PATH_MAX];
1018 if (!realpath(a, real_a))
1019 strncpy_null(real_a, a);
1021 if (!realpath(b, real_b))
1022 strncpy_null(real_b, b);
1024 /* Identical path? */
1025 if (strcmp(real_a, real_b) == 0)
1028 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
1029 if (errno == ENOENT)
1034 /* Same blockdevice? */
1035 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
1036 st_buf_a.st_rdev == st_buf_b.st_rdev) {
1041 if (st_buf_a.st_dev == st_buf_b.st_dev &&
1042 st_buf_a.st_ino == st_buf_b.st_ino) {
1049 /* checks if a and b are identical or device
1050 * files associated with the same block device or
1051 * if one file is a loop device that uses the other
1054 static int is_same_loop_file(const char* a, const char* b)
1056 char res_a[PATH_MAX];
1057 char res_b[PATH_MAX];
1058 const char* final_a = NULL;
1059 const char* final_b = NULL;
1062 /* Resolve a if it is a loop device */
1063 if((ret = is_loop_device(a)) < 0) {
1068 ret = resolve_loop_device(a, res_a, sizeof(res_a));
1079 /* Resolve b if it is a loop device */
1080 if ((ret = is_loop_device(b)) < 0) {
1085 ret = resolve_loop_device(b, res_b, sizeof(res_b));
1096 return is_same_blk_file(final_a, final_b);
1099 /* Checks if a file exists and is a block or regular file*/
1100 static int is_existing_blk_or_reg_file(const char* filename)
1104 if(stat(filename, &st_buf) < 0) {
1111 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
1114 /* Checks if a file is used (directly or indirectly via a loop device)
1115 * by a device in fs_devices
1117 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
1121 struct list_head *head;
1122 struct list_head *cur;
1123 struct btrfs_device *device;
1125 head = &fs_devices->devices;
1126 list_for_each(cur, head) {
1127 device = list_entry(cur, struct btrfs_device, dev_list);
1129 if((ret = is_same_loop_file(device->name, file)))
1137 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1138 * Returns NULL on invalid input or malloc failure; Other failures
1139 * will be handled by the caller using the input pathame.
1141 char *canonicalize_dm_name(const char *ptname)
1145 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
1147 if (!ptname || !*ptname)
1150 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
1151 if (!(f = fopen(path, "r")))
1154 /* read <name>\n from sysfs */
1155 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
1156 name[sz - 1] = '\0';
1157 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
1159 if (access(path, F_OK) == 0)
1167 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
1168 * to a device mapper pathname.
1169 * Returns NULL on invalid input or malloc failure; Other failures
1170 * will be handled by the caller using the input pathame.
1172 char *canonicalize_path(const char *path)
1174 char *canonical, *p;
1176 if (!path || !*path)
1179 canonical = realpath(path, NULL);
1181 return strdup(path);
1182 p = strrchr(canonical, '/');
1183 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
1184 char *dm = canonicalize_dm_name(p + 1);
1195 * returns 1 if the device was mounted, < 0 on error or 0 if everything
1196 * is safe to continue.
1198 int check_mounted(const char* file)
1203 fd = open(file, O_RDONLY);
1205 fprintf (stderr, "check_mounted(): Could not open %s\n", file);
1209 ret = check_mounted_where(fd, file, NULL, 0, NULL);
1215 int check_mounted_where(int fd, const char *file, char *where, int size,
1216 struct btrfs_fs_devices **fs_dev_ret)
1221 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1225 /* scan the initial device */
1226 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
1227 &total_devs, BTRFS_SUPER_INFO_OFFSET, 0);
1228 is_btrfs = (ret >= 0);
1230 /* scan other devices */
1231 if (is_btrfs && total_devs > 1) {
1232 ret = btrfs_scan_lblkid();
1237 /* iterate over the list of currently mountes filesystems */
1238 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
1241 while ((mnt = getmntent (f)) != NULL) {
1243 if(strcmp(mnt->mnt_type, "btrfs") != 0)
1246 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
1248 /* ignore entries in the mount table that are not
1249 associated with a file*/
1250 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
1251 goto out_mntloop_err;
1255 ret = is_same_loop_file(file, mnt->mnt_fsname);
1259 goto out_mntloop_err;
1264 /* Did we find an entry in mnt table? */
1265 if (mnt && size && where) {
1266 strncpy(where, mnt->mnt_dir, size);
1270 *fs_dev_ret = fs_devices_mnt;
1272 ret = (mnt != NULL);
1280 struct pending_dir {
1281 struct list_head list;
1282 char name[PATH_MAX];
1285 int btrfs_register_one_device(const char *fname)
1287 struct btrfs_ioctl_vol_args args;
1292 fd = open("/dev/btrfs-control", O_RDWR);
1294 fprintf(stderr, "failed to open /dev/btrfs-control "
1295 "skipping device registration: %s\n",
1299 strncpy(args.name, fname, BTRFS_PATH_NAME_MAX);
1300 args.name[BTRFS_PATH_NAME_MAX-1] = 0;
1301 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
1304 fprintf(stderr, "ERROR: device scan failed '%s' - %s\n",
1305 fname, strerror(e));
1313 * Register all devices in the fs_uuid list created in the user
1314 * space. Ensure btrfs_scan_lblkid() is called before this func.
1316 int btrfs_register_all_devices(void)
1319 struct btrfs_fs_devices *fs_devices;
1320 struct btrfs_device *device;
1321 struct list_head *all_uuids;
1323 all_uuids = btrfs_scanned_uuids();
1325 list_for_each_entry(fs_devices, all_uuids, list) {
1326 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1327 if (strlen(device->name) != 0) {
1328 err = btrfs_register_one_device(device->name);
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 const char* unit_suffix_binary[] =
1370 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
1371 static const char* unit_suffix_decimal[] =
1372 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
1374 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
1380 const char** suffix = NULL;
1386 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
1387 snprintf(str, str_size, "%llu", size);
1391 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
1394 suffix = unit_suffix_binary;
1395 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
1398 suffix = unit_suffix_decimal;
1403 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
1411 switch (unit_mode & UNITS_MODE_MASK) {
1412 case UNITS_TBYTES: base *= mult; num_divs++;
1413 case UNITS_GBYTES: base *= mult; num_divs++;
1414 case UNITS_MBYTES: base *= mult; num_divs++;
1415 case UNITS_KBYTES: num_divs++;
1422 while (size >= mult) {
1429 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
1431 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
1436 fraction = (float)last_size / base;
1438 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
1442 * __strncpy__null - strncpy with null termination
1443 * @dest: the target array
1444 * @src: the source string
1445 * @n: maximum bytes to copy (size of *dest)
1447 * Like strncpy, but ensures destination is null-terminated.
1449 * Copies the string pointed to by src, including the terminating null
1450 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
1451 * of n bytes. Then ensure that dest is null-terminated.
1453 char *__strncpy__null(char *dest, const char *src, size_t n)
1455 strncpy(dest, src, n);
1462 * Checks to make sure that the label matches our requirements.
1464 0 if everything is safe and usable
1465 -1 if the label is too long
1467 static int check_label(const char *input)
1469 int len = strlen(input);
1471 if (len > BTRFS_LABEL_SIZE - 1) {
1472 fprintf(stderr, "ERROR: Label %s is too long (max %d)\n",
1473 input, BTRFS_LABEL_SIZE - 1);
1480 static int set_label_unmounted(const char *dev, const char *label)
1482 struct btrfs_trans_handle *trans;
1483 struct btrfs_root *root;
1486 ret = check_mounted(dev);
1488 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1492 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1497 /* Open the super_block at the default location
1498 * and as read-write.
1500 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
1501 if (!root) /* errors are printed by open_ctree() */
1504 trans = btrfs_start_transaction(root, 1);
1505 snprintf(root->fs_info->super_copy->label, BTRFS_LABEL_SIZE, "%s",
1507 btrfs_commit_transaction(trans, root);
1509 /* Now we close it since we are done. */
1514 static int set_label_mounted(const char *mount_path, const char *label)
1518 fd = open(mount_path, O_RDONLY | O_NOATIME);
1520 fprintf(stderr, "ERROR: unable to access '%s'\n", mount_path);
1524 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
1525 fprintf(stderr, "ERROR: unable to set label %s\n",
1535 static int get_label_unmounted(const char *dev, char *label)
1537 struct btrfs_root *root;
1540 ret = check_mounted(dev);
1542 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1546 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1551 /* Open the super_block at the default location
1554 root = open_ctree(dev, 0, 0);
1558 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
1560 /* Now we close it since we are done. */
1566 * If a partition is mounted, try to get the filesystem label via its
1567 * mounted path rather than device. Return the corresponding error
1568 * the user specified the device path.
1570 int get_label_mounted(const char *mount_path, char *labelp)
1572 char label[BTRFS_LABEL_SIZE];
1575 fd = open(mount_path, O_RDONLY | O_NOATIME);
1577 fprintf(stderr, "ERROR: unable to access '%s'\n", mount_path);
1581 memset(label, '\0', sizeof(label));
1582 if (ioctl(fd, BTRFS_IOC_GET_FSLABEL, label) < 0) {
1583 fprintf(stderr, "ERROR: unable get label %s\n", strerror(errno));
1588 strncpy(labelp, label, sizeof(label));
1593 int get_label(const char *btrfs_dev, char *label)
1597 ret = is_existing_blk_or_reg_file(btrfs_dev);
1599 ret = get_label_mounted(btrfs_dev, label);
1601 ret = get_label_unmounted(btrfs_dev, label);
1606 int set_label(const char *btrfs_dev, const char *label)
1610 if (check_label(label))
1613 ret = is_existing_blk_or_reg_file(btrfs_dev);
1615 ret = set_label_mounted(btrfs_dev, label);
1617 ret = set_label_unmounted(btrfs_dev, label);
1622 int btrfs_scan_block_devices(int run_ioctl)
1628 struct btrfs_fs_devices *tmp_devices;
1630 FILE *proc_partitions;
1638 proc_partitions = fopen("/proc/partitions","r");
1639 if (!proc_partitions) {
1640 fprintf(stderr, "Unable to open '/proc/partitions' for scanning\n");
1643 /* skip the header */
1644 for (i = 0; i < 2; i++)
1645 if (!fgets(buf, 1023, proc_partitions)) {
1647 "Unable to read '/proc/partitions' for scanning\n");
1648 fclose(proc_partitions);
1652 strcpy(fullpath,"/dev/");
1653 while(fgets(buf, 1023, proc_partitions)) {
1654 ret = sscanf(buf," %*d %*d %*d %99s", fullpath + 5);
1657 "failed to scan device name from /proc/partitions\n");
1662 * multipath and MD devices may register as a btrfs filesystem
1663 * both through the original block device and through
1664 * the special (/dev/mapper or /dev/mdX) entry.
1665 * This scans the special entries last
1667 special = strncmp(fullpath, "/dev/dm-", strlen("/dev/dm-")) == 0;
1669 special = strncmp(fullpath, "/dev/md", strlen("/dev/md")) == 0;
1671 if (scans == 0 && special)
1673 if (scans > 0 && !special)
1676 ret = lstat(fullpath, &st);
1678 fprintf(stderr, "failed to stat %s\n", fullpath);
1681 if (!S_ISBLK(st.st_mode)) {
1685 fd = open(fullpath, O_RDONLY);
1687 if (errno != ENOMEDIUM)
1688 fprintf(stderr, "failed to open %s: %s\n",
1689 fullpath, strerror(errno));
1692 ret = btrfs_scan_one_device(fd, fullpath, &tmp_devices,
1694 BTRFS_SUPER_INFO_OFFSET, 0);
1695 if (ret == 0 && run_ioctl > 0) {
1696 btrfs_register_one_device(fullpath);
1701 fclose(proc_partitions);
1711 * Unsafe subvolume check.
1713 * This only checks ino == BTRFS_FIRST_FREE_OBJECTID, even it is not in a
1714 * btrfs mount point.
1715 * Must use together with other reliable method like btrfs ioctl.
1717 static int __is_subvol(const char *path)
1722 ret = lstat(path, &st);
1726 return st.st_ino == BTRFS_FIRST_FREE_OBJECTID;
1730 * A not-so-good version fls64. No fascinating optimization since
1731 * no one except parse_size use it
1733 static int fls64(u64 x)
1737 for (i = 0; i <64; i++)
1738 if (x << i & (1ULL << 63))
1743 u64 parse_size(char *s)
1751 fprintf(stderr, "ERROR: Size value is empty\n");
1756 "ERROR: Size value '%s' is less equal than 0\n", s);
1759 ret = strtoull(s, &endptr, 10);
1761 fprintf(stderr, "ERROR: Size value '%s' is invalid\n", s);
1764 if (endptr[0] && endptr[1]) {
1765 fprintf(stderr, "ERROR: Illegal suffix contains character '%c' in wrong position\n",
1770 * strtoll returns LLONG_MAX when overflow, if this happens,
1771 * need to call strtoull to get the real size
1773 if (errno == ERANGE && ret == ULLONG_MAX) {
1775 "ERROR: Size value '%s' is too large for u64\n", s);
1779 c = tolower(endptr[0]);
1802 fprintf(stderr, "ERROR: Unknown size descriptor '%c'\n",
1807 /* Check whether ret * mult overflow */
1808 if (fls64(ret) + fls64(mult) - 1 > 64) {
1810 "ERROR: Size value '%s' is too large for u64\n", s);
1817 u64 parse_qgroupid(const char *p)
1819 char *s = strchr(p, '/');
1820 const char *ptr_src_end = p + strlen(p);
1821 char *ptr_parse_end = NULL;
1830 /* Numeric format like '0/257' is the primary case */
1832 id = strtoull(p, &ptr_parse_end, 10);
1833 if (ptr_parse_end != ptr_src_end)
1837 level = strtoull(p, &ptr_parse_end, 10);
1838 if (ptr_parse_end != s)
1841 id = strtoull(s + 1, &ptr_parse_end, 10);
1842 if (ptr_parse_end != ptr_src_end)
1845 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
1848 /* Path format like subv at 'my_subvol' is the fallback case */
1849 ret = __is_subvol(p);
1850 if (ret < 0 || !ret)
1852 fd = open(p, O_RDONLY);
1855 ret = lookup_ino_rootid(fd, &id);
1862 fprintf(stderr, "ERROR: invalid qgroupid or subvolume path: %s\n", p);
1866 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
1872 ret = stat(fname, &st);
1876 if (S_ISDIR(st.st_mode)) {
1877 *dirstream = opendir(fname);
1880 fd = dirfd(*dirstream);
1881 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
1882 fd = open(fname, open_flags);
1885 * we set this on purpose, in case the caller output
1886 * strerror(errno) as success
1894 closedir(*dirstream);
1899 int open_file_or_dir(const char *fname, DIR **dirstream)
1901 return open_file_or_dir3(fname, dirstream, O_RDWR);
1904 void close_file_or_dir(int fd, DIR *dirstream)
1907 closedir(dirstream);
1912 int get_device_info(int fd, u64 devid,
1913 struct btrfs_ioctl_dev_info_args *di_args)
1917 di_args->devid = devid;
1918 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
1920 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
1921 return ret ? -errno : 0;
1924 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
1927 struct btrfs_dev_item *dev_item;
1928 char *buf = search_args->buf;
1930 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
1931 + sizeof(struct btrfs_dev_item));
1932 buf += sizeof(struct btrfs_ioctl_search_header);
1934 dev_item = (struct btrfs_dev_item *)buf;
1936 return btrfs_stack_device_id(dev_item);
1939 static int search_chunk_tree_for_fs_info(int fd,
1940 struct btrfs_ioctl_fs_info_args *fi_args)
1944 u64 start_devid = 1;
1945 struct btrfs_ioctl_search_args search_args;
1946 struct btrfs_ioctl_search_key *search_key = &search_args.key;
1948 fi_args->num_devices = 0;
1950 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
1951 / (sizeof(struct btrfs_ioctl_search_header)
1952 + sizeof(struct btrfs_dev_item));
1954 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
1955 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
1956 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
1957 search_key->min_type = BTRFS_DEV_ITEM_KEY;
1958 search_key->max_type = BTRFS_DEV_ITEM_KEY;
1959 search_key->min_transid = 0;
1960 search_key->max_transid = (u64)-1;
1961 search_key->nr_items = max_items;
1962 search_key->max_offset = (u64)-1;
1965 search_key->min_offset = start_devid;
1967 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
1971 fi_args->num_devices += (u64)search_key->nr_items;
1973 if (search_key->nr_items == max_items) {
1974 start_devid = find_max_device_id(&search_args,
1975 search_key->nr_items) + 1;
1979 /* get the lastest max_id to stay consistent with the num_devices */
1980 if (search_key->nr_items == 0)
1982 * last tree_search returns an empty buf, use the devid of
1983 * the last dev_item of the previous tree_search
1985 fi_args->max_id = start_devid - 1;
1987 fi_args->max_id = find_max_device_id(&search_args,
1988 search_key->nr_items);
1994 * For a given path, fill in the ioctl fs_ and info_ args.
1995 * If the path is a btrfs mountpoint, fill info for all devices.
1996 * If the path is a btrfs device, fill in only that device.
1998 * The path provided must be either on a mounted btrfs fs,
1999 * or be a mounted btrfs device.
2001 * Returns 0 on success, or a negative errno.
2003 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
2004 struct btrfs_ioctl_dev_info_args **di_ret)
2011 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2012 struct btrfs_ioctl_dev_info_args *di_args;
2013 struct btrfs_ioctl_dev_info_args tmp;
2014 char mp[BTRFS_PATH_NAME_MAX + 1];
2015 DIR *dirstream = NULL;
2017 memset(fi_args, 0, sizeof(*fi_args));
2019 if (is_block_device(path)) {
2020 struct btrfs_super_block *disk_super;
2021 char buf[BTRFS_SUPER_INFO_SIZE];
2024 /* Ensure it's mounted, then set path to the mountpoint */
2025 fd = open(path, O_RDONLY);
2028 fprintf(stderr, "Couldn't open %s: %s\n",
2029 path, strerror(errno));
2032 ret = check_mounted_where(fd, path, mp, sizeof(mp),
2041 /* Only fill in this one device */
2042 fi_args->num_devices = 1;
2044 disk_super = (struct btrfs_super_block *)buf;
2045 ret = btrfs_read_dev_super(fd, disk_super,
2046 BTRFS_SUPER_INFO_OFFSET, 0);
2051 devid = btrfs_stack_device_id(&disk_super->dev_item);
2053 fi_args->max_id = devid;
2056 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
2060 /* at this point path must not be for a block device */
2061 fd = open_file_or_dir(path, &dirstream);
2067 /* fill in fi_args if not just a single device */
2068 if (fi_args->num_devices != 1) {
2069 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
2076 * The fs_args->num_devices does not include seed devices
2078 ret = search_chunk_tree_for_fs_info(fd, fi_args);
2083 * search_chunk_tree_for_fs_info() will lacks the devid 0
2084 * so manual probe for it here.
2086 ret = get_device_info(fd, 0, &tmp);
2088 fi_args->num_devices++;
2096 if (!fi_args->num_devices)
2099 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
2106 memcpy(di_args, &tmp, sizeof(tmp));
2107 for (; i <= fi_args->max_id; ++i) {
2108 ret = get_device_info(fd, i, &di_args[ndevs]);
2117 * only when the only dev we wanted to find is not there then
2118 * let any error be returned
2120 if (fi_args->num_devices != 1) {
2126 close_file_or_dir(fd, dirstream);
2130 #define isoctal(c) (((c) & ~7) == '0')
2132 static inline void translate(char *f, char *t)
2134 while (*f != '\0') {
2136 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
2137 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
2147 * Checks if the swap device.
2148 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
2150 static int is_swap_device(const char *file)
2161 if (stat(file, &st_buf) < 0)
2163 if (S_ISBLK(st_buf.st_mode))
2164 dev = st_buf.st_rdev;
2165 else if (S_ISREG(st_buf.st_mode)) {
2166 dev = st_buf.st_dev;
2167 ino = st_buf.st_ino;
2171 if ((f = fopen("/proc/swaps", "r")) == NULL)
2174 /* skip the first line */
2175 if (fgets(tmp, sizeof(tmp), f) == NULL)
2178 while (fgets(tmp, sizeof(tmp), f) != NULL) {
2179 if ((cp = strchr(tmp, ' ')) != NULL)
2181 if ((cp = strchr(tmp, '\t')) != NULL)
2183 translate(tmp, buf);
2184 if (stat(buf, &st_buf) != 0)
2186 if (S_ISBLK(st_buf.st_mode)) {
2187 if (dev == st_buf.st_rdev) {
2191 } else if (S_ISREG(st_buf.st_mode)) {
2192 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
2206 * Check for existing filesystem or partition table on device.
2208 * 1 for existing fs or partition
2209 * 0 for nothing found
2210 * -1 for internal error
2217 blkid_probe pr = NULL;
2221 if (!device || !*device)
2224 ret = -1; /* will reset on success of all setup calls */
2226 pr = blkid_new_probe_from_filename(device);
2230 size = blkid_probe_get_size(pr);
2234 /* nothing to overwrite on a 0-length device */
2240 ret = blkid_probe_enable_partitions(pr, 1);
2244 ret = blkid_do_fullprobe(pr);
2249 * Blkid returns 1 for nothing found and 0 when it finds a signature,
2250 * but we want the exact opposite, so reverse the return value here.
2252 * In addition print some useful diagnostics about what actually is
2260 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
2262 "%s appears to contain an existing "
2263 "filesystem (%s).\n", device, type);
2264 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
2266 "%s appears to contain a partition "
2267 "table (%s).\n", device, type);
2270 "%s appears to contain something weird "
2271 "according to blkid\n", device);
2277 blkid_free_probe(pr);
2280 "probe of %s failed, cannot detect "
2281 "existing filesystem.\n", device);
2285 static int group_profile_devs_min(u64 flag)
2287 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2288 case 0: /* single */
2289 case BTRFS_BLOCK_GROUP_DUP:
2291 case BTRFS_BLOCK_GROUP_RAID0:
2292 case BTRFS_BLOCK_GROUP_RAID1:
2293 case BTRFS_BLOCK_GROUP_RAID5:
2295 case BTRFS_BLOCK_GROUP_RAID6:
2297 case BTRFS_BLOCK_GROUP_RAID10:
2304 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
2305 u64 dev_cnt, int mixed, char *estr)
2313 allowed |= BTRFS_BLOCK_GROUP_RAID10;
2315 allowed |= BTRFS_BLOCK_GROUP_RAID6;
2317 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
2318 BTRFS_BLOCK_GROUP_RAID5;
2321 allowed |= BTRFS_BLOCK_GROUP_DUP;
2325 ((metadata_profile | data_profile) & BTRFS_BLOCK_GROUP_DUP)) {
2327 "DUP is not allowed when FS has multiple devices\n");
2330 if (metadata_profile & ~allowed) {
2332 "unable to create FS with metadata profile %s "
2333 "(have %llu devices but %d devices are required)\n",
2334 btrfs_group_profile_str(metadata_profile), dev_cnt,
2335 group_profile_devs_min(metadata_profile));
2338 if (data_profile & ~allowed) {
2340 "unable to create FS with data profile %s "
2341 "(have %llu devices but %d devices are required)\n",
2342 btrfs_group_profile_str(data_profile), dev_cnt,
2343 group_profile_devs_min(data_profile));
2347 if (!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP)) {
2349 "dup for data is allowed only in mixed mode");
2355 /* Check if disk is suitable for btrfs
2357 * 1: something is wrong, estr provides the error
2360 int test_dev_for_mkfs(char *file, int force_overwrite, char *estr)
2366 ret = is_swap_device(file);
2368 snprintf(estr, sz, "error checking %s status: %s\n", file,
2373 snprintf(estr, sz, "%s is a swap device\n", file);
2376 if (!force_overwrite) {
2377 if (check_overwrite(file)) {
2378 snprintf(estr, sz, "Use the -f option to force overwrite.\n");
2382 ret = check_mounted(file);
2384 snprintf(estr, sz, "error checking %s mount status\n",
2389 snprintf(estr, sz, "%s is mounted\n", file);
2392 /* check if the device is busy */
2393 fd = open(file, O_RDWR|O_EXCL);
2395 snprintf(estr, sz, "unable to open %s: %s\n", file,
2399 if (fstat(fd, &st)) {
2400 snprintf(estr, sz, "unable to stat %s: %s\n", file,
2405 if (!S_ISBLK(st.st_mode)) {
2406 fprintf(stderr, "'%s' is not a block device\n", file);
2414 int btrfs_scan_lblkid()
2419 struct btrfs_fs_devices *tmp_devices;
2420 blkid_dev_iterate iter = NULL;
2421 blkid_dev dev = NULL;
2422 blkid_cache cache = NULL;
2423 char path[PATH_MAX];
2425 if (btrfs_scan_done)
2428 if (blkid_get_cache(&cache, 0) < 0) {
2429 printf("ERROR: lblkid cache get failed\n");
2432 blkid_probe_all(cache);
2433 iter = blkid_dev_iterate_begin(cache);
2434 blkid_dev_set_search(iter, "TYPE", "btrfs");
2435 while (blkid_dev_next(iter, &dev) == 0) {
2436 dev = blkid_verify(cache, dev);
2439 /* if we are here its definitely a btrfs disk*/
2440 strncpy_null(path, blkid_dev_devname(dev));
2442 fd = open(path, O_RDONLY);
2444 printf("ERROR: could not open %s\n", path);
2447 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
2448 &num_devices, BTRFS_SUPER_INFO_OFFSET, 0);
2450 printf("ERROR: could not scan %s\n", path);
2457 blkid_dev_iterate_end(iter);
2458 blkid_put_cache(cache);
2460 btrfs_scan_done = 1;
2465 int is_vol_small(char *file)
2472 fd = open(file, O_RDONLY);
2475 if (fstat(fd, &st) < 0) {
2480 size = btrfs_device_size(fd, &st);
2485 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
2495 * This reads a line from the stdin and only returns non-zero if the
2496 * first whitespace delimited token is a case insensitive match with yes
2499 int ask_user(char *question)
2501 char buf[30] = {0,};
2502 char *saveptr = NULL;
2505 printf("%s [y/N]: ", question);
2507 return fgets(buf, sizeof(buf) - 1, stdin) &&
2508 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
2509 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
2514 * - file or directory return the containing tree root id
2515 * - subvolume return its own tree id
2516 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
2517 * undefined and function returns -1
2519 int lookup_ino_rootid(int fd, u64 *rootid)
2521 struct btrfs_ioctl_ino_lookup_args args;
2525 memset(&args, 0, sizeof(args));
2527 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
2529 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
2532 fprintf(stderr, "ERROR: Failed to lookup root id - %s\n",
2537 *rootid = args.treeid;
2543 * return 0 if a btrfs mount point is found
2544 * return 1 if a mount point is found but not btrfs
2545 * return <0 if something goes wrong
2547 int find_mount_root(const char *path, char **mount_root)
2555 int longest_matchlen = 0;
2556 char *longest_match = NULL;
2558 fd = open(path, O_RDONLY | O_NOATIME);
2563 mnttab = setmntent("/proc/self/mounts", "r");
2567 while ((ent = getmntent(mnttab))) {
2568 len = strlen(ent->mnt_dir);
2569 if (strncmp(ent->mnt_dir, path, len) == 0) {
2570 /* match found and use the latest match */
2571 if (longest_matchlen <= len) {
2572 free(longest_match);
2573 longest_matchlen = len;
2574 longest_match = strdup(ent->mnt_dir);
2575 not_btrfs = strcmp(ent->mnt_type, "btrfs");
2584 free(longest_match);
2589 *mount_root = realpath(longest_match, NULL);
2593 free(longest_match);
2597 int test_minimum_size(const char *file, u32 leafsize)
2600 struct stat statbuf;
2602 fd = open(file, O_RDONLY);
2605 if (stat(file, &statbuf) < 0) {
2609 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(leafsize)) {
2618 * test if name is a correct subvolume name
2619 * this function return
2620 * 0-> name is not a correct subvolume name
2621 * 1-> name is a correct subvolume name
2623 int test_issubvolname(const char *name)
2625 return name[0] != '\0' && !strchr(name, '/') &&
2626 strcmp(name, ".") && strcmp(name, "..");
2630 * test if path is a directory
2631 * this function return
2632 * 0-> path exists but it is not a directory
2633 * 1-> path exists and it is a directory
2634 * -1 -> path is unaccessible
2636 int test_isdir(const char *path)
2641 ret = stat(path, &st);
2645 return S_ISDIR(st.st_mode);
2648 void units_set_mode(unsigned *units, unsigned mode)
2650 unsigned base = *units & UNITS_MODE_MASK;
2652 *units = base | mode;
2655 void units_set_base(unsigned *units, unsigned base)
2657 unsigned mode = *units & ~UNITS_MODE_MASK;
2659 *units = base | mode;
2662 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
2666 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
2667 if (!path->nodes[level])
2669 if (path->slots[level] + 1 >=
2670 btrfs_header_nritems(path->nodes[level]))
2673 btrfs_item_key_to_cpu(path->nodes[level], key,
2674 path->slots[level] + 1);
2676 btrfs_node_key_to_cpu(path->nodes[level], key,
2677 path->slots[level] + 1);
2683 char* btrfs_group_type_str(u64 flag)
2685 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
2686 BTRFS_SPACE_INFO_GLOBAL_RSV;
2688 switch (flag & mask) {
2689 case BTRFS_BLOCK_GROUP_DATA:
2691 case BTRFS_BLOCK_GROUP_SYSTEM:
2693 case BTRFS_BLOCK_GROUP_METADATA:
2695 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
2696 return "Data+Metadata";
2697 case BTRFS_SPACE_INFO_GLOBAL_RSV:
2698 return "GlobalReserve";
2704 char* btrfs_group_profile_str(u64 flag)
2706 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2709 case BTRFS_BLOCK_GROUP_RAID0:
2711 case BTRFS_BLOCK_GROUP_RAID1:
2713 case BTRFS_BLOCK_GROUP_RAID5:
2715 case BTRFS_BLOCK_GROUP_RAID6:
2717 case BTRFS_BLOCK_GROUP_DUP:
2719 case BTRFS_BLOCK_GROUP_RAID10:
2726 u64 disk_size(char *path)
2730 if (statfs(path, &sfs) < 0)
2733 return sfs.f_bsize * sfs.f_blocks;
2736 u64 get_partition_size(char *dev)
2739 int fd = open(dev, O_RDONLY);
2743 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
2752 int btrfs_tree_search2_ioctl_supported(int fd)
2754 struct btrfs_ioctl_search_args_v2 *args2;
2755 struct btrfs_ioctl_search_key *sk;
2756 int args2_size = 1024;
2757 char args2_buf[args2_size];
2759 static int v2_supported = -1;
2761 if (v2_supported != -1)
2762 return v2_supported;
2764 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
2768 * Search for the extent tree item in the root tree.
2770 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
2771 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
2772 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
2773 sk->min_type = BTRFS_ROOT_ITEM_KEY;
2774 sk->max_type = BTRFS_ROOT_ITEM_KEY;
2776 sk->max_offset = (u64)-1;
2777 sk->min_transid = 0;
2778 sk->max_transid = (u64)-1;
2780 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
2781 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
2782 if (ret == -EOPNOTSUPP)
2789 return v2_supported;
2792 int btrfs_check_node_or_leaf_size(u32 size, u32 sectorsize)
2794 if (size < sectorsize) {
2796 "ERROR: Illegal nodesize (or leafsize) %u (smaller than %u)\n",
2799 } else if (size > BTRFS_MAX_METADATA_BLOCKSIZE) {
2801 "ERROR: Illegal nodesize (or leafsize) %u (larger than %u)\n",
2802 size, BTRFS_MAX_METADATA_BLOCKSIZE);
2804 } else if (size & (sectorsize - 1)) {
2806 "ERROR: Illegal nodesize (or leafsize) %u (not aligned to %u)\n",