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 */
28 #include <sys/ioctl.h>
29 #include <sys/mount.h>
31 #include <sys/types.h>
33 #include <uuid/uuid.h>
39 #include <linux/loop.h>
40 #include <linux/major.h>
41 #include <linux/kdev_t.h>
43 #include "kerncompat.h"
44 #include "radix-tree.h"
47 #include "transaction.h"
54 #define BLKGETSIZE64 0
55 static inline int ioctl(int fd, int define, u64 *size) { return 0; }
59 #define BLKDISCARD _IO(0x12,119)
63 discard_blocks(int fd, u64 start, u64 len)
65 u64 range[2] = { start, len };
67 if (ioctl(fd, BLKDISCARD, &range) < 0)
72 static u64 reference_root_table[] = {
73 [1] = BTRFS_ROOT_TREE_OBJECTID,
74 [2] = BTRFS_EXTENT_TREE_OBJECTID,
75 [3] = BTRFS_CHUNK_TREE_OBJECTID,
76 [4] = BTRFS_DEV_TREE_OBJECTID,
77 [5] = BTRFS_FS_TREE_OBJECTID,
78 [6] = BTRFS_CSUM_TREE_OBJECTID,
81 int make_btrfs(int fd, const char *device, const char *label,
82 u64 blocks[7], u64 num_bytes, u32 nodesize,
83 u32 leafsize, u32 sectorsize, u32 stripesize)
85 struct btrfs_super_block super;
86 struct extent_buffer *buf;
87 struct btrfs_root_item root_item;
88 struct btrfs_disk_key disk_key;
89 struct btrfs_extent_item *extent_item;
90 struct btrfs_inode_item *inode_item;
91 struct btrfs_chunk *chunk;
92 struct btrfs_dev_item *dev_item;
93 struct btrfs_dev_extent *dev_extent;
94 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
105 first_free = BTRFS_SUPER_INFO_OFFSET + sectorsize * 2 - 1;
106 first_free &= ~((u64)sectorsize - 1);
108 memset(&super, 0, sizeof(super));
110 num_bytes = (num_bytes / sectorsize) * sectorsize;
111 uuid_generate(super.fsid);
112 uuid_generate(super.dev_item.uuid);
113 uuid_generate(chunk_tree_uuid);
115 btrfs_set_super_bytenr(&super, blocks[0]);
116 btrfs_set_super_num_devices(&super, 1);
117 super.magic = cpu_to_le64(BTRFS_MAGIC);
118 btrfs_set_super_generation(&super, 1);
119 btrfs_set_super_root(&super, blocks[1]);
120 btrfs_set_super_chunk_root(&super, blocks[3]);
121 btrfs_set_super_total_bytes(&super, num_bytes);
122 btrfs_set_super_bytes_used(&super, 6 * leafsize);
123 btrfs_set_super_sectorsize(&super, sectorsize);
124 btrfs_set_super_leafsize(&super, leafsize);
125 btrfs_set_super_nodesize(&super, nodesize);
126 btrfs_set_super_stripesize(&super, stripesize);
127 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
128 btrfs_set_super_chunk_root_generation(&super, 1);
129 btrfs_set_super_cache_generation(&super, -1);
131 strncpy(super.label, label, BTRFS_LABEL_SIZE - 1);
133 buf = malloc(sizeof(*buf) + max(sectorsize, leafsize));
135 /* create the tree of root objects */
136 memset(buf->data, 0, leafsize);
138 btrfs_set_header_bytenr(buf, blocks[1]);
139 btrfs_set_header_nritems(buf, 4);
140 btrfs_set_header_generation(buf, 1);
141 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
142 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
143 write_extent_buffer(buf, super.fsid, (unsigned long)
144 btrfs_header_fsid(buf), BTRFS_FSID_SIZE);
146 write_extent_buffer(buf, chunk_tree_uuid, (unsigned long)
147 btrfs_header_chunk_tree_uuid(buf),
150 /* create the items for the root tree */
151 memset(&root_item, 0, sizeof(root_item));
152 inode_item = &root_item.inode;
153 btrfs_set_stack_inode_generation(inode_item, 1);
154 btrfs_set_stack_inode_size(inode_item, 3);
155 btrfs_set_stack_inode_nlink(inode_item, 1);
156 btrfs_set_stack_inode_nbytes(inode_item, leafsize);
157 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
158 btrfs_set_root_refs(&root_item, 1);
159 btrfs_set_root_used(&root_item, leafsize);
160 btrfs_set_root_generation(&root_item, 1);
162 memset(&disk_key, 0, sizeof(disk_key));
163 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
164 btrfs_set_disk_key_offset(&disk_key, 0);
167 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - sizeof(root_item);
168 btrfs_set_root_bytenr(&root_item, blocks[2]);
169 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
170 btrfs_set_item_key(buf, &disk_key, nritems);
171 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
172 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
174 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
175 nritems), sizeof(root_item));
178 itemoff = itemoff - sizeof(root_item);
179 btrfs_set_root_bytenr(&root_item, blocks[4]);
180 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
181 btrfs_set_item_key(buf, &disk_key, nritems);
182 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
183 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
185 write_extent_buffer(buf, &root_item,
186 btrfs_item_ptr_offset(buf, nritems),
190 itemoff = itemoff - sizeof(root_item);
191 btrfs_set_root_bytenr(&root_item, blocks[5]);
192 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
193 btrfs_set_item_key(buf, &disk_key, nritems);
194 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
195 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
197 write_extent_buffer(buf, &root_item,
198 btrfs_item_ptr_offset(buf, nritems),
202 itemoff = itemoff - sizeof(root_item);
203 btrfs_set_root_bytenr(&root_item, blocks[6]);
204 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
205 btrfs_set_item_key(buf, &disk_key, nritems);
206 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
207 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
209 write_extent_buffer(buf, &root_item,
210 btrfs_item_ptr_offset(buf, nritems),
215 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
216 ret = pwrite(fd, buf->data, leafsize, blocks[1]);
217 BUG_ON(ret != leafsize);
219 /* create the items for the extent tree */
220 memset(buf->data+sizeof(struct btrfs_header), 0,
221 leafsize-sizeof(struct btrfs_header));
223 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize);
224 for (i = 1; i < 7; i++) {
225 BUG_ON(blocks[i] < first_free);
226 BUG_ON(blocks[i] < blocks[i - 1]);
228 /* create extent item */
229 itemoff -= sizeof(struct btrfs_extent_item) +
230 sizeof(struct btrfs_tree_block_info);
231 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
232 btrfs_set_disk_key_offset(&disk_key, leafsize);
233 btrfs_set_disk_key_type(&disk_key, BTRFS_EXTENT_ITEM_KEY);
234 btrfs_set_item_key(buf, &disk_key, nritems);
235 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems),
237 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
238 sizeof(struct btrfs_extent_item) +
239 sizeof(struct btrfs_tree_block_info));
240 extent_item = btrfs_item_ptr(buf, nritems,
241 struct btrfs_extent_item);
242 btrfs_set_extent_refs(buf, extent_item, 1);
243 btrfs_set_extent_generation(buf, extent_item, 1);
244 btrfs_set_extent_flags(buf, extent_item,
245 BTRFS_EXTENT_FLAG_TREE_BLOCK);
248 /* create extent ref */
249 ref_root = reference_root_table[i];
250 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
251 btrfs_set_disk_key_offset(&disk_key, ref_root);
252 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
253 btrfs_set_item_key(buf, &disk_key, nritems);
254 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems),
256 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems), 0);
259 btrfs_set_header_bytenr(buf, blocks[2]);
260 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
261 btrfs_set_header_nritems(buf, nritems);
262 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
263 ret = pwrite(fd, buf->data, leafsize, blocks[2]);
264 BUG_ON(ret != leafsize);
266 /* create the chunk tree */
267 memset(buf->data+sizeof(struct btrfs_header), 0,
268 leafsize-sizeof(struct btrfs_header));
270 item_size = sizeof(*dev_item);
271 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - item_size;
273 /* first device 1 (there is no device 0) */
274 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
275 btrfs_set_disk_key_offset(&disk_key, 1);
276 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
277 btrfs_set_item_key(buf, &disk_key, nritems);
278 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
279 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems), item_size);
281 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
282 btrfs_set_device_id(buf, dev_item, 1);
283 btrfs_set_device_generation(buf, dev_item, 0);
284 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
285 btrfs_set_device_bytes_used(buf, dev_item,
286 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
287 btrfs_set_device_io_align(buf, dev_item, sectorsize);
288 btrfs_set_device_io_width(buf, dev_item, sectorsize);
289 btrfs_set_device_sector_size(buf, dev_item, sectorsize);
290 btrfs_set_device_type(buf, dev_item, 0);
292 write_extent_buffer(buf, super.dev_item.uuid,
293 (unsigned long)btrfs_device_uuid(dev_item),
295 write_extent_buffer(buf, super.fsid,
296 (unsigned long)btrfs_device_fsid(dev_item),
298 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
302 item_size = btrfs_chunk_item_size(1);
303 itemoff = itemoff - item_size;
305 /* then we have chunk 0 */
306 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
307 btrfs_set_disk_key_offset(&disk_key, 0);
308 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
309 btrfs_set_item_key(buf, &disk_key, nritems);
310 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
311 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems), item_size);
313 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
314 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
315 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
316 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
317 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
318 btrfs_set_chunk_io_align(buf, chunk, sectorsize);
319 btrfs_set_chunk_io_width(buf, chunk, sectorsize);
320 btrfs_set_chunk_sector_size(buf, chunk, sectorsize);
321 btrfs_set_chunk_num_stripes(buf, chunk, 1);
322 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
323 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
326 write_extent_buffer(buf, super.dev_item.uuid,
327 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
330 /* copy the key for the chunk to the system array */
331 ptr = super.sys_chunk_array;
332 array_size = sizeof(disk_key);
334 memcpy(ptr, &disk_key, sizeof(disk_key));
335 ptr += sizeof(disk_key);
337 /* copy the chunk to the system array */
338 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
339 array_size += item_size;
341 btrfs_set_super_sys_array_size(&super, array_size);
343 btrfs_set_header_bytenr(buf, blocks[3]);
344 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
345 btrfs_set_header_nritems(buf, nritems);
346 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
347 ret = pwrite(fd, buf->data, leafsize, blocks[3]);
349 /* create the device tree */
350 memset(buf->data+sizeof(struct btrfs_header), 0,
351 leafsize-sizeof(struct btrfs_header));
353 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) -
354 sizeof(struct btrfs_dev_extent);
356 btrfs_set_disk_key_objectid(&disk_key, 1);
357 btrfs_set_disk_key_offset(&disk_key, 0);
358 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
359 btrfs_set_item_key(buf, &disk_key, nritems);
360 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
361 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
362 sizeof(struct btrfs_dev_extent));
363 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
364 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
365 BTRFS_CHUNK_TREE_OBJECTID);
366 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
367 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
368 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
370 write_extent_buffer(buf, chunk_tree_uuid,
371 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
374 btrfs_set_dev_extent_length(buf, dev_extent,
375 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
378 btrfs_set_header_bytenr(buf, blocks[4]);
379 btrfs_set_header_owner(buf, BTRFS_DEV_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[4]);
384 /* create the FS root */
385 memset(buf->data+sizeof(struct btrfs_header), 0,
386 leafsize-sizeof(struct btrfs_header));
387 btrfs_set_header_bytenr(buf, blocks[5]);
388 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
389 btrfs_set_header_nritems(buf, 0);
390 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
391 ret = pwrite(fd, buf->data, leafsize, blocks[5]);
392 BUG_ON(ret != leafsize);
394 /* finally create the csum root */
395 memset(buf->data+sizeof(struct btrfs_header), 0,
396 leafsize-sizeof(struct btrfs_header));
397 btrfs_set_header_bytenr(buf, blocks[6]);
398 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
399 btrfs_set_header_nritems(buf, 0);
400 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
401 ret = pwrite(fd, buf->data, leafsize, blocks[6]);
402 BUG_ON(ret != leafsize);
404 /* and write out the super block */
405 BUG_ON(sizeof(super) > sectorsize);
406 memset(buf->data, 0, sectorsize);
407 memcpy(buf->data, &super, sizeof(super));
408 buf->len = sectorsize;
409 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
410 ret = pwrite(fd, buf->data, sectorsize, blocks[0]);
411 BUG_ON(ret != sectorsize);
418 static u64 device_size(int fd, struct stat *st)
421 if (S_ISREG(st->st_mode)) {
424 if (!S_ISBLK(st->st_mode)) {
427 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
433 static int zero_blocks(int fd, off_t start, size_t len)
435 char *buf = malloc(len);
442 written = pwrite(fd, buf, len, start);
449 static int zero_dev_start(int fd)
452 size_t len = 2 * 1024 * 1024;
455 /* don't overwrite the disk labels on sparc */
459 return zero_blocks(fd, start, len);
462 static int zero_dev_end(int fd, u64 dev_size)
464 size_t len = 2 * 1024 * 1024;
465 off_t start = dev_size - len;
467 return zero_blocks(fd, start, len);
470 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
471 struct btrfs_root *root, int fd, char *path,
472 u64 block_count, u32 io_width, u32 io_align,
475 struct btrfs_super_block *disk_super;
476 struct btrfs_super_block *super = root->fs_info->super_copy;
477 struct btrfs_device *device;
478 struct btrfs_dev_item *dev_item;
484 device = kmalloc(sizeof(*device), GFP_NOFS);
487 buf = kmalloc(sectorsize, GFP_NOFS);
492 BUG_ON(sizeof(*disk_super) > sectorsize);
493 memset(buf, 0, sectorsize);
495 disk_super = (struct btrfs_super_block *)buf;
496 dev_item = &disk_super->dev_item;
498 uuid_generate(device->uuid);
501 device->io_width = io_width;
502 device->io_align = io_align;
503 device->sector_size = sectorsize;
505 device->writeable = 1;
506 device->total_bytes = block_count;
507 device->bytes_used = 0;
508 device->total_ios = 0;
509 device->dev_root = root->fs_info->dev_root;
511 ret = btrfs_add_device(trans, root, device);
514 total_bytes = btrfs_super_total_bytes(super) + block_count;
515 btrfs_set_super_total_bytes(super, total_bytes);
517 num_devs = btrfs_super_num_devices(super) + 1;
518 btrfs_set_super_num_devices(super, num_devs);
520 memcpy(disk_super, super, sizeof(*disk_super));
522 printf("adding device %s id %llu\n", path,
523 (unsigned long long)device->devid);
525 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
526 btrfs_set_stack_device_id(dev_item, device->devid);
527 btrfs_set_stack_device_type(dev_item, device->type);
528 btrfs_set_stack_device_io_align(dev_item, device->io_align);
529 btrfs_set_stack_device_io_width(dev_item, device->io_width);
530 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
531 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
532 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
533 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
535 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
536 BUG_ON(ret != sectorsize);
539 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
540 device->fs_devices = root->fs_info->fs_devices;
544 int btrfs_prepare_device(int fd, char *file, int zero_end, u64 *block_count_ret,
545 u64 max_block_count, int *mixed, int nodiscard)
552 ret = fstat(fd, &st);
554 fprintf(stderr, "unable to stat %s\n", file);
558 block_count = device_size(fd, &st);
559 if (block_count == 0) {
560 fprintf(stderr, "unable to find %s size\n", file);
564 block_count = min(block_count, max_block_count);
567 if (block_count < 1024 * 1024 * 1024 && !(*mixed)) {
568 printf("SMALL VOLUME: forcing mixed metadata/data groups\n");
574 * We intentionally ignore errors from the discard ioctl. It is
575 * not necessary for the mkfs functionality but just an optimization.
577 discard_blocks(fd, 0, block_count);
580 ret = zero_dev_start(fd);
582 fprintf(stderr, "failed to zero device start %d\n", ret);
586 for (i = 0 ; i < BTRFS_SUPER_MIRROR_MAX; i++) {
587 bytenr = btrfs_sb_offset(i);
588 if (bytenr >= block_count)
590 zero_blocks(fd, bytenr, BTRFS_SUPER_INFO_SIZE);
594 ret = zero_dev_end(fd, block_count);
596 fprintf(stderr, "failed to zero device end %d\n", ret);
600 *block_count_ret = block_count;
604 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
605 struct btrfs_root *root, u64 objectid)
608 struct btrfs_inode_item inode_item;
609 time_t now = time(NULL);
611 memset(&inode_item, 0, sizeof(inode_item));
612 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
613 btrfs_set_stack_inode_size(&inode_item, 0);
614 btrfs_set_stack_inode_nlink(&inode_item, 1);
615 btrfs_set_stack_inode_nbytes(&inode_item, root->leafsize);
616 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
617 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
618 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
619 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
620 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
621 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
622 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
623 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
624 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
626 if (root->fs_info->tree_root == root)
627 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
629 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
633 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
637 btrfs_set_root_dirid(&root->root_item, objectid);
644 * checks if a path is a block device node
645 * Returns negative errno on failure, otherwise
646 * returns 1 for blockdev, 0 for not-blockdev
648 int is_block_device(const char *path) {
651 if (stat(path, &statbuf) < 0)
654 return S_ISBLK(statbuf.st_mode);
658 * Find the mount point for a mounted device.
659 * On success, returns 0 with mountpoint in *mp.
660 * On failure, returns -errno (not mounted yields -EINVAL)
661 * Is noisy on failures, expects to be given a mounted device.
663 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size) {
667 ret = is_block_device(dev);
670 fprintf(stderr, "%s is not a block device\n", dev);
673 fprintf(stderr, "Could not check %s: %s\n",
674 dev, strerror(-ret));
679 fd = open(dev, O_RDONLY);
682 fprintf(stderr, "Could not open %s: %s\n", dev, strerror(errno));
686 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
688 fprintf(stderr, "%s is not a mounted btrfs device\n", dev);
690 } else { /* mounted, all good */
697 fprintf(stderr, "Could not get mountpoint for %s\n", dev);
702 * Given a pathname, return a filehandle to:
703 * the original pathname or,
704 * if the pathname is a mounted btrfs device, to its mountpoint.
706 * On error, return -1, errno should be set.
708 int open_path_or_dev_mnt(const char *path)
710 char mp[BTRFS_PATH_NAME_MAX + 1];
713 if (is_block_device(path)) {
716 ret = get_btrfs_mount(path, mp, sizeof(mp));
718 /* not a mounted btrfs dev */
722 fdmnt = open(mp, O_RDWR);
724 fdmnt = open_file_or_dir(path);
730 /* checks if a device is a loop device */
731 int is_loop_device (const char* device) {
734 if(stat(device, &statbuf) < 0)
737 return (S_ISBLK(statbuf.st_mode) &&
738 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
742 /* Takes a loop device path (e.g. /dev/loop0) and returns
743 * the associated file (e.g. /images/my_btrfs.img) */
744 int resolve_loop_device(const char* loop_dev, char* loop_file, int max_len)
750 char real_loop_dev[PATH_MAX];
752 if (!realpath(loop_dev, real_loop_dev))
754 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
755 if (!(f = fopen(p, "r")))
758 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
759 ret = fscanf(f, fmt, loop_file);
767 /* Checks whether a and b are identical or device
768 * files associated with the same block device
770 int is_same_blk_file(const char* a, const char* b)
772 struct stat st_buf_a, st_buf_b;
773 char real_a[PATH_MAX];
774 char real_b[PATH_MAX];
776 if(!realpath(a, real_a) ||
777 !realpath(b, real_b))
782 /* Identical path? */
783 if(strcmp(real_a, real_b) == 0)
786 if(stat(a, &st_buf_a) < 0 ||
787 stat(b, &st_buf_b) < 0)
794 /* Same blockdevice? */
795 if(S_ISBLK(st_buf_a.st_mode) &&
796 S_ISBLK(st_buf_b.st_mode) &&
797 st_buf_a.st_rdev == st_buf_b.st_rdev)
803 if (st_buf_a.st_dev == st_buf_b.st_dev &&
804 st_buf_a.st_ino == st_buf_b.st_ino)
812 /* checks if a and b are identical or device
813 * files associated with the same block device or
814 * if one file is a loop device that uses the other
817 int is_same_loop_file(const char* a, const char* b)
819 char res_a[PATH_MAX];
820 char res_b[PATH_MAX];
825 /* Resolve a if it is a loop device */
826 if((ret = is_loop_device(a)) < 0) {
831 if ((ret = resolve_loop_device(a, res_a, sizeof(res_a))) < 0)
839 /* Resolve b if it is a loop device */
840 if ((ret = is_loop_device(b)) < 0) {
845 if((ret = resolve_loop_device(b, res_b, sizeof(res_b))) < 0)
853 return is_same_blk_file(final_a, final_b);
856 /* Checks if a file exists and is a block or regular file*/
857 int is_existing_blk_or_reg_file(const char* filename)
861 if(stat(filename, &st_buf) < 0) {
868 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
871 /* Checks if a file is used (directly or indirectly via a loop device)
872 * by a device in fs_devices
874 int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices, const char* file)
877 struct list_head *head;
878 struct list_head *cur;
879 struct btrfs_device *device;
881 head = &fs_devices->devices;
882 list_for_each(cur, head) {
883 device = list_entry(cur, struct btrfs_device, dev_list);
885 if((ret = is_same_loop_file(device->name, file)))
893 * returns 1 if the device was mounted, < 0 on error or 0 if everything
894 * is safe to continue.
896 int check_mounted(const char* file)
901 fd = open(file, O_RDONLY);
903 fprintf (stderr, "check_mounted(): Could not open %s\n", file);
907 ret = check_mounted_where(fd, file, NULL, 0, NULL);
913 int check_mounted_where(int fd, const char *file, char *where, int size,
914 struct btrfs_fs_devices **fs_dev_ret)
919 struct btrfs_fs_devices *fs_devices_mnt = NULL;
923 /* scan the initial device */
924 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
925 &total_devs, BTRFS_SUPER_INFO_OFFSET);
926 is_btrfs = (ret >= 0);
928 /* scan other devices */
929 if (is_btrfs && total_devs > 1) {
930 if((ret = btrfs_scan_for_fsid(fs_devices_mnt, total_devs, 1)))
934 /* iterate over the list of currently mountes filesystems */
935 if ((f = setmntent ("/proc/mounts", "r")) == NULL)
938 while ((mnt = getmntent (f)) != NULL) {
940 if(strcmp(mnt->mnt_type, "btrfs") != 0)
943 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
945 /* ignore entries in the mount table that are not
946 associated with a file*/
947 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
948 goto out_mntloop_err;
952 ret = is_same_loop_file(file, mnt->mnt_fsname);
956 goto out_mntloop_err;
961 /* Did we find an entry in mnt table? */
962 if (mnt && size && where) {
963 strncpy(where, mnt->mnt_dir, size);
967 *fs_dev_ret = fs_devices_mnt;
977 /* Gets the mount point of btrfs filesystem that is using the specified device.
978 * Returns 0 is everything is good, <0 if we have an error.
979 * TODO: Fix this fucntion and check_mounted to work with multiple drive BTRFS
982 int get_mountpt(char *dev, char *mntpt, size_t size)
988 f = setmntent("/proc/mounts", "r");
992 while ((mnt = getmntent(f)) != NULL )
994 if (strcmp(dev, mnt->mnt_fsname) == 0)
996 strncpy(mntpt, mnt->mnt_dir, size);
1005 /* We didn't find an entry so lets report an error */
1012 struct pending_dir {
1013 struct list_head list;
1014 char name[PATH_MAX];
1017 void btrfs_register_one_device(char *fname)
1019 struct btrfs_ioctl_vol_args args;
1024 fd = open("/dev/btrfs-control", O_RDONLY);
1026 fprintf(stderr, "failed to open /dev/btrfs-control "
1027 "skipping device registration\n");
1030 strncpy(args.name, fname, BTRFS_PATH_NAME_MAX);
1031 args.name[BTRFS_PATH_NAME_MAX-1] = 0;
1032 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
1035 fprintf(stderr, "ERROR: device scan failed '%s' - %s\n",
1036 fname, strerror(e));
1041 int btrfs_scan_one_dir(char *dirname, int run_ioctl)
1044 struct dirent *dirent;
1045 struct pending_dir *pending;
1051 struct list_head pending_list;
1052 struct btrfs_fs_devices *tmp_devices;
1055 INIT_LIST_HEAD(&pending_list);
1057 pending = malloc(sizeof(*pending));
1060 strcpy(pending->name, dirname);
1063 dirname_len = strlen(pending->name);
1064 fullpath = malloc(PATH_MAX);
1065 dirname = pending->name;
1071 dirp = opendir(dirname);
1073 fprintf(stderr, "Unable to open %s for scanning\n", dirname);
1078 dirent = readdir(dirp);
1081 if (dirent->d_name[0] == '.')
1083 if (dirname_len + strlen(dirent->d_name) + 2 > PATH_MAX) {
1087 snprintf(fullpath, PATH_MAX, "%s/%s", dirname, dirent->d_name);
1088 ret = lstat(fullpath, &st);
1090 fprintf(stderr, "failed to stat %s\n", fullpath);
1093 if (S_ISLNK(st.st_mode))
1095 if (S_ISDIR(st.st_mode)) {
1096 struct pending_dir *next = malloc(sizeof(*next));
1101 strcpy(next->name, fullpath);
1102 list_add_tail(&next->list, &pending_list);
1104 if (!S_ISBLK(st.st_mode)) {
1107 fd = open(fullpath, O_RDONLY);
1109 /* ignore the following errors:
1110 ENXIO (device don't exists)
1111 ENOMEDIUM (No medium found ->
1112 like a cd tray empty)
1114 if(errno != ENXIO && errno != ENOMEDIUM)
1115 fprintf(stderr, "failed to read %s: %s\n",
1116 fullpath, strerror(errno));
1119 ret = btrfs_scan_one_device(fd, fullpath, &tmp_devices,
1121 BTRFS_SUPER_INFO_OFFSET);
1122 if (ret == 0 && run_ioctl > 0) {
1123 btrfs_register_one_device(fullpath);
1127 if (!list_empty(&pending_list)) {
1129 pending = list_entry(pending_list.next, struct pending_dir,
1132 list_del(&pending->list);
1146 int btrfs_scan_for_fsid(struct btrfs_fs_devices *fs_devices, u64 total_devs,
1151 ret = btrfs_scan_block_devices(run_ioctls);
1153 ret = btrfs_scan_one_dir("/dev", run_ioctls);
1157 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
1160 struct btrfs_super_block *disk_super;
1164 buf = malloc(BTRFS_SUPER_INFO_SIZE);
1169 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
1170 if (ret != BTRFS_SUPER_INFO_SIZE)
1174 disk_super = (struct btrfs_super_block *)buf;
1175 if (disk_super->magic != cpu_to_le64(BTRFS_MAGIC))
1178 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
1187 static char *size_strs[] = { "", "KB", "MB", "GB", "TB",
1188 "PB", "EB", "ZB", "YB"};
1189 char *pretty_sizes(u64 size)
1192 int pretty_len = 16;
1200 u64 last_size = size;
1202 while(size >= 1024){
1208 if (num_divs >= ARRAY_SIZE(size_strs))
1210 fraction = (float)last_size / 1024;
1212 pretty = malloc(pretty_len);
1213 snprintf(pretty, pretty_len, "%.2f%s", fraction, size_strs[num_divs]);
1218 * __strncpy__null - strncpy with null termination
1219 * @dest: the target array
1220 * @src: the source string
1221 * @n: maximum bytes to copy (size of *dest)
1223 * Like strncpy, but ensures destination is null-terminated.
1225 * Copies the string pointed to by src, including the terminating null
1226 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
1227 * of n bytes. Then ensure that dest is null-terminated.
1229 char *__strncpy__null(char *dest, const char *src, size_t n)
1231 strncpy(dest, src, n);
1238 * Checks to make sure that the label matches our requirements.
1240 0 if everything is safe and usable
1241 -1 if the label is too long
1243 static int check_label(const char *input)
1245 int len = strlen(input);
1247 if (len > BTRFS_LABEL_SIZE - 1) {
1248 fprintf(stderr, "ERROR: Label %s is too long (max %d)\n",
1249 input, BTRFS_LABEL_SIZE - 1);
1256 static int set_label_unmounted(const char *dev, const char *label)
1258 struct btrfs_trans_handle *trans;
1259 struct btrfs_root *root;
1262 ret = check_mounted(dev);
1264 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1268 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1273 /* Open the super_block at the default location
1274 * and as read-write.
1276 root = open_ctree(dev, 0, 1);
1277 if (!root) /* errors are printed by open_ctree() */
1280 trans = btrfs_start_transaction(root, 1);
1281 snprintf(root->fs_info->super_copy->label, BTRFS_LABEL_SIZE, "%s",
1283 btrfs_commit_transaction(trans, root);
1285 /* Now we close it since we are done. */
1290 static int set_label_mounted(const char *mount_path, const char *label)
1294 fd = open(mount_path, O_RDONLY | O_NOATIME);
1296 fprintf(stderr, "ERROR: unable access to '%s'\n", mount_path);
1300 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
1301 fprintf(stderr, "ERROR: unable to set label %s\n",
1311 static int get_label_unmounted(const char *dev)
1313 struct btrfs_root *root;
1316 ret = check_mounted(dev);
1318 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1322 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1327 /* Open the super_block at the default location
1330 root = open_ctree(dev, 0, 0);
1334 fprintf(stdout, "%s\n", root->fs_info->super_copy->label);
1336 /* Now we close it since we are done. */
1342 * If a partition is mounted, try to get the filesystem label via its
1343 * mounted path rather than device. Return the corresponding error
1344 * the user specified the device path.
1346 static int get_label_mounted(const char *mount_path)
1348 char label[BTRFS_LABEL_SIZE];
1351 fd = open(mount_path, O_RDONLY | O_NOATIME);
1353 fprintf(stderr, "ERROR: unable access to '%s'\n", mount_path);
1357 memset(label, '\0', sizeof(label));
1358 if (ioctl(fd, BTRFS_IOC_GET_FSLABEL, label) < 0) {
1359 fprintf(stderr, "ERROR: unable get label %s\n", strerror(errno));
1364 fprintf(stdout, "%s\n", label);
1369 int get_label(const char *btrfs_dev)
1371 return is_existing_blk_or_reg_file(btrfs_dev) ?
1372 get_label_unmounted(btrfs_dev) :
1373 get_label_mounted(btrfs_dev);
1376 int set_label(const char *btrfs_dev, const char *label)
1378 if (check_label(label))
1381 return is_existing_blk_or_reg_file(btrfs_dev) ?
1382 set_label_unmounted(btrfs_dev, label) :
1383 set_label_mounted(btrfs_dev, label);
1386 int btrfs_scan_block_devices(int run_ioctl)
1392 struct btrfs_fs_devices *tmp_devices;
1394 FILE *proc_partitions;
1402 proc_partitions = fopen("/proc/partitions","r");
1403 if (!proc_partitions) {
1404 fprintf(stderr, "Unable to open '/proc/partitions' for scanning\n");
1407 /* skip the header */
1408 for (i = 0; i < 2; i++)
1409 if (!fgets(buf, 1023, proc_partitions)) {
1411 "Unable to read '/proc/partitions' for scanning\n");
1412 fclose(proc_partitions);
1416 strcpy(fullpath,"/dev/");
1417 while(fgets(buf, 1023, proc_partitions)) {
1418 i = sscanf(buf," %*d %*d %*d %99s", fullpath+5);
1421 * multipath and MD devices may register as a btrfs filesystem
1422 * both through the original block device and through
1423 * the special (/dev/mapper or /dev/mdX) entry.
1424 * This scans the special entries last
1426 special = strncmp(fullpath, "/dev/dm-", strlen("/dev/dm-")) == 0;
1428 special = strncmp(fullpath, "/dev/md", strlen("/dev/md")) == 0;
1430 if (scans == 0 && special)
1432 if (scans > 0 && !special)
1435 ret = lstat(fullpath, &st);
1437 fprintf(stderr, "failed to stat %s\n", fullpath);
1440 if (!S_ISBLK(st.st_mode)) {
1444 fd = open(fullpath, O_RDONLY);
1446 fprintf(stderr, "failed to open %s: %s\n",
1447 fullpath, strerror(errno));
1450 ret = btrfs_scan_one_device(fd, fullpath, &tmp_devices,
1452 BTRFS_SUPER_INFO_OFFSET);
1453 if (ret == 0 && run_ioctl > 0) {
1454 btrfs_register_one_device(fullpath);
1459 fclose(proc_partitions);
1468 u64 parse_size(char *s)
1474 for (i = 0; s && s[i] && isdigit(s[i]); i++) ;
1476 fprintf(stderr, "ERROR: size value is empty\n");
1498 fprintf(stderr, "ERROR: Unknown size descriptor "
1503 if (s[i] && s[i+1]) {
1504 fprintf(stderr, "ERROR: Illegal suffix contains "
1505 "character '%c' in wrong position\n",
1509 return strtoull(s, NULL, 10) * mult;
1512 int open_file_or_dir(const char *fname)
1519 ret = stat(fname, &st);
1523 if (S_ISDIR(st.st_mode)) {
1524 dirstream = opendir(fname);
1528 fd = dirfd(dirstream);
1530 fd = open(fname, O_RDWR);
1538 int get_device_info(int fd, u64 devid,
1539 struct btrfs_ioctl_dev_info_args *di_args)
1543 di_args->devid = devid;
1544 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
1546 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
1547 return ret ? -errno : 0;
1550 int get_fs_info(int fd, char *path, struct btrfs_ioctl_fs_info_args *fi_args,
1551 struct btrfs_ioctl_dev_info_args **di_ret)
1556 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1557 struct btrfs_ioctl_dev_info_args *di_args;
1558 char mp[BTRFS_PATH_NAME_MAX + 1];
1560 memset(fi_args, 0, sizeof(*fi_args));
1562 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
1563 if (ret && (errno == EINVAL || errno == ENOTTY)) {
1564 /* path is not a mounted btrfs. Try if it's a device */
1565 ret = check_mounted_where(fd, path, mp, sizeof(mp),
1571 fi_args->num_devices = 1;
1572 fi_args->max_id = fs_devices_mnt->latest_devid;
1573 i = fs_devices_mnt->latest_devid;
1574 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
1576 fd = open_file_or_dir(mp);
1583 if (!fi_args->num_devices)
1586 di_args = *di_ret = malloc(fi_args->num_devices * sizeof(*di_args));
1590 for (; i <= fi_args->max_id; ++i) {
1591 BUG_ON(ndevs >= fi_args->num_devices);
1592 ret = get_device_info(fd, i, &di_args[ndevs]);
1605 #define isoctal(c) (((c) & ~7) == '0')
1607 static inline void translate(char *f, char *t)
1609 while (*f != '\0') {
1611 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
1612 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
1622 * Checks if the swap device.
1623 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
1625 int is_swap_device(const char *file)
1636 if (stat(file, &st_buf) < 0)
1638 if (S_ISBLK(st_buf.st_mode))
1639 dev = st_buf.st_rdev;
1640 else if (S_ISREG(st_buf.st_mode)) {
1641 dev = st_buf.st_dev;
1642 ino = st_buf.st_ino;
1646 if ((f = fopen("/proc/swaps", "r")) == NULL)
1649 /* skip the first line */
1650 if (fgets(tmp, sizeof(tmp), f) == NULL)
1653 while (fgets(tmp, sizeof(tmp), f) != NULL) {
1654 if ((cp = strchr(tmp, ' ')) != NULL)
1656 if ((cp = strchr(tmp, '\t')) != NULL)
1658 translate(tmp, buf);
1659 if (stat(buf, &st_buf) != 0)
1661 if (S_ISBLK(st_buf.st_mode)) {
1662 if (dev == st_buf.st_rdev) {
1666 } else if (S_ISREG(st_buf.st_mode)) {
1667 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {