2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
18 #define _XOPEN_SOURCE 600
22 #include <sys/types.h>
24 #include <uuid/uuid.h>
29 #include "transaction.h"
30 #include "print-tree.h"
34 struct btrfs_device *dev;
39 struct cache_extent ce;
47 struct btrfs_bio_stripe stripes[];
50 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
51 (sizeof(struct btrfs_bio_stripe) * (n)))
53 static LIST_HEAD(fs_uuids);
55 static struct btrfs_device *__find_device(struct list_head *head, u64 devid,
58 struct btrfs_device *dev;
59 struct list_head *cur;
61 list_for_each(cur, head) {
62 dev = list_entry(cur, struct btrfs_device, dev_list);
63 if (dev->devid == devid &&
64 !memcmp(dev->uuid, uuid, BTRFS_UUID_SIZE)) {
71 static struct btrfs_fs_devices *find_fsid(u8 *fsid)
73 struct list_head *cur;
74 struct btrfs_fs_devices *fs_devices;
76 list_for_each(cur, &fs_uuids) {
77 fs_devices = list_entry(cur, struct btrfs_fs_devices, list);
78 if (memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE) == 0)
84 static int device_list_add(const char *path,
85 struct btrfs_super_block *disk_super,
86 u64 devid, struct btrfs_fs_devices **fs_devices_ret)
88 struct btrfs_device *device;
89 struct btrfs_fs_devices *fs_devices;
90 u64 found_transid = btrfs_super_generation(disk_super);
92 fs_devices = find_fsid(disk_super->fsid);
94 fs_devices = kmalloc(sizeof(*fs_devices), GFP_NOFS);
97 INIT_LIST_HEAD(&fs_devices->devices);
98 list_add(&fs_devices->list, &fs_uuids);
99 memcpy(fs_devices->fsid, disk_super->fsid, BTRFS_FSID_SIZE);
100 fs_devices->latest_devid = devid;
101 fs_devices->latest_trans = found_transid;
102 fs_devices->lowest_devid = (u64)-1;
105 device = __find_device(&fs_devices->devices, devid,
106 disk_super->dev_item.uuid);
109 device = kzalloc(sizeof(*device), GFP_NOFS);
111 /* we can safely leave the fs_devices entry around */
114 device->devid = devid;
115 memcpy(device->uuid, disk_super->dev_item.uuid,
117 device->name = kstrdup(path, GFP_NOFS);
122 list_add(&device->dev_list, &fs_devices->devices);
125 if (found_transid > fs_devices->latest_trans) {
126 fs_devices->latest_devid = devid;
127 fs_devices->latest_trans = found_transid;
129 if (fs_devices->lowest_devid > devid) {
130 fs_devices->lowest_devid = devid;
132 *fs_devices_ret = fs_devices;
136 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
138 struct list_head *head = &fs_devices->devices;
139 struct list_head *cur;
140 struct btrfs_device *device;
142 list_for_each(cur, head) {
143 device = list_entry(cur, struct btrfs_device, dev_list);
149 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, int flags)
152 struct list_head *head = &fs_devices->devices;
153 struct list_head *cur;
154 struct btrfs_device *device;
157 list_for_each(cur, head) {
158 device = list_entry(cur, struct btrfs_device, dev_list);
159 fd = open(device->name, flags);
160 printk("opening %s devid %llu fd %d\n", device->name,
161 (unsigned long long)device->devid, fd);
166 if (device->devid == fs_devices->latest_devid)
167 fs_devices->latest_bdev = fd;
168 if (device->devid == fs_devices->lowest_devid)
169 fs_devices->lowest_bdev = fd;
174 btrfs_close_devices(fs_devices);
178 int btrfs_scan_one_device(int fd, const char *path,
179 struct btrfs_fs_devices **fs_devices_ret,
180 u64 *total_devs, u64 super_offset)
182 struct btrfs_super_block *disk_super;
193 ret = pread(fd, buf, 4096, super_offset);
198 disk_super = (struct btrfs_super_block *)buf;
199 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
200 sizeof(disk_super->magic))) {
204 devid = le64_to_cpu(disk_super->dev_item.devid);
205 *total_devs = btrfs_super_num_devices(disk_super);
206 uuid_unparse(disk_super->fsid, uuidbuf);
209 if (disk_super->label[0])
210 printf("label %s ", disk_super->label);
212 printf("fsuuid %s ", uuidbuf);
213 printf("devid %llu %s\n", (unsigned long long)devid, path);
214 ret = device_list_add(path, disk_super, devid, fs_devices_ret);
223 * this uses a pretty simple search, the expectation is that it is
224 * called very infrequently and that a given device has a small number
227 static int find_free_dev_extent(struct btrfs_trans_handle *trans,
228 struct btrfs_device *device,
229 struct btrfs_path *path,
230 u64 num_bytes, u64 *start)
232 struct btrfs_key key;
233 struct btrfs_root *root = device->dev_root;
234 struct btrfs_dev_extent *dev_extent = NULL;
237 u64 search_start = 0;
238 u64 search_end = device->total_bytes;
242 struct extent_buffer *l;
247 /* FIXME use last free of some kind */
249 /* we don't want to overwrite the superblock on the drive,
250 * so we make sure to start at an offset of at least 1MB
252 search_start = max((u64)1024 * 1024, search_start);
253 key.objectid = device->devid;
254 key.offset = search_start;
255 key.type = BTRFS_DEV_EXTENT_KEY;
256 ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
259 ret = btrfs_previous_item(root, path, 0, key.type);
263 btrfs_item_key_to_cpu(l, &key, path->slots[0]);
266 slot = path->slots[0];
267 if (slot >= btrfs_header_nritems(l)) {
268 ret = btrfs_next_leaf(root, path);
275 if (search_start >= search_end) {
279 *start = search_start;
283 *start = last_byte > search_start ?
284 last_byte : search_start;
285 if (search_end <= *start) {
291 btrfs_item_key_to_cpu(l, &key, slot);
293 if (key.objectid < device->devid)
296 if (key.objectid > device->devid)
299 if (key.offset >= search_start && key.offset > last_byte &&
301 if (last_byte < search_start)
302 last_byte = search_start;
303 hole_size = key.offset - last_byte;
304 if (key.offset > last_byte &&
305 hole_size >= num_bytes) {
310 if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY) {
315 dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent);
316 last_byte = key.offset + btrfs_dev_extent_length(l, dev_extent);
322 /* we have to make sure we didn't find an extent that has already
323 * been allocated by the map tree or the original allocation
325 btrfs_release_path(root, path);
326 BUG_ON(*start < search_start);
328 if (*start + num_bytes > search_end) {
332 /* check for pending inserts here */
336 btrfs_release_path(root, path);
340 int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
341 struct btrfs_device *device,
342 u64 chunk_tree, u64 chunk_objectid,
344 u64 num_bytes, u64 *start)
347 struct btrfs_path *path;
348 struct btrfs_root *root = device->dev_root;
349 struct btrfs_dev_extent *extent;
350 struct extent_buffer *leaf;
351 struct btrfs_key key;
353 path = btrfs_alloc_path();
357 ret = find_free_dev_extent(trans, device, path, num_bytes, start);
362 key.objectid = device->devid;
364 key.type = BTRFS_DEV_EXTENT_KEY;
365 ret = btrfs_insert_empty_item(trans, root, path, &key,
369 leaf = path->nodes[0];
370 extent = btrfs_item_ptr(leaf, path->slots[0],
371 struct btrfs_dev_extent);
372 btrfs_set_dev_extent_chunk_tree(leaf, extent, chunk_tree);
373 btrfs_set_dev_extent_chunk_objectid(leaf, extent, chunk_objectid);
374 btrfs_set_dev_extent_chunk_offset(leaf, extent, chunk_offset);
376 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
377 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(extent),
380 btrfs_set_dev_extent_length(leaf, extent, num_bytes);
381 btrfs_mark_buffer_dirty(leaf);
383 btrfs_free_path(path);
387 static int find_next_chunk(struct btrfs_root *root, u64 objectid, u64 *offset)
389 struct btrfs_path *path;
391 struct btrfs_key key;
392 struct btrfs_chunk *chunk;
393 struct btrfs_key found_key;
395 path = btrfs_alloc_path();
398 key.objectid = objectid;
399 key.offset = (u64)-1;
400 key.type = BTRFS_CHUNK_ITEM_KEY;
402 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
408 ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY);
412 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
414 if (found_key.objectid != objectid)
417 chunk = btrfs_item_ptr(path->nodes[0], path->slots[0],
419 *offset = found_key.offset +
420 btrfs_chunk_length(path->nodes[0], chunk);
425 btrfs_free_path(path);
429 static int find_next_devid(struct btrfs_root *root, struct btrfs_path *path,
433 struct btrfs_key key;
434 struct btrfs_key found_key;
436 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
437 key.type = BTRFS_DEV_ITEM_KEY;
438 key.offset = (u64)-1;
440 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
446 ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID,
451 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
453 *objectid = found_key.offset + 1;
457 btrfs_release_path(root, path);
462 * the device information is stored in the chunk root
463 * the btrfs_device struct should be fully filled in
465 int btrfs_add_device(struct btrfs_trans_handle *trans,
466 struct btrfs_root *root,
467 struct btrfs_device *device)
470 struct btrfs_path *path;
471 struct btrfs_dev_item *dev_item;
472 struct extent_buffer *leaf;
473 struct btrfs_key key;
477 root = root->fs_info->chunk_root;
479 path = btrfs_alloc_path();
483 ret = find_next_devid(root, path, &free_devid);
487 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
488 key.type = BTRFS_DEV_ITEM_KEY;
489 key.offset = free_devid;
491 ret = btrfs_insert_empty_item(trans, root, path, &key,
496 leaf = path->nodes[0];
497 dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
499 device->devid = free_devid;
500 btrfs_set_device_id(leaf, dev_item, device->devid);
501 btrfs_set_device_type(leaf, dev_item, device->type);
502 btrfs_set_device_io_align(leaf, dev_item, device->io_align);
503 btrfs_set_device_io_width(leaf, dev_item, device->io_width);
504 btrfs_set_device_sector_size(leaf, dev_item, device->sector_size);
505 btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes);
506 btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used);
507 btrfs_set_device_group(leaf, dev_item, 0);
508 btrfs_set_device_seek_speed(leaf, dev_item, 0);
509 btrfs_set_device_bandwidth(leaf, dev_item, 0);
511 ptr = (unsigned long)btrfs_device_uuid(dev_item);
512 write_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE);
513 btrfs_mark_buffer_dirty(leaf);
517 btrfs_free_path(path);
521 int btrfs_update_device(struct btrfs_trans_handle *trans,
522 struct btrfs_device *device)
525 struct btrfs_path *path;
526 struct btrfs_root *root;
527 struct btrfs_dev_item *dev_item;
528 struct extent_buffer *leaf;
529 struct btrfs_key key;
531 root = device->dev_root->fs_info->chunk_root;
533 path = btrfs_alloc_path();
537 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
538 key.type = BTRFS_DEV_ITEM_KEY;
539 key.offset = device->devid;
541 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
550 leaf = path->nodes[0];
551 dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
553 btrfs_set_device_id(leaf, dev_item, device->devid);
554 btrfs_set_device_type(leaf, dev_item, device->type);
555 btrfs_set_device_io_align(leaf, dev_item, device->io_align);
556 btrfs_set_device_io_width(leaf, dev_item, device->io_width);
557 btrfs_set_device_sector_size(leaf, dev_item, device->sector_size);
558 btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes);
559 btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used);
560 btrfs_mark_buffer_dirty(leaf);
563 btrfs_free_path(path);
567 int btrfs_add_system_chunk(struct btrfs_trans_handle *trans,
568 struct btrfs_root *root,
569 struct btrfs_key *key,
570 struct btrfs_chunk *chunk, int item_size)
572 struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
573 struct btrfs_disk_key disk_key;
577 array_size = btrfs_super_sys_array_size(super_copy);
578 if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE)
581 ptr = super_copy->sys_chunk_array + array_size;
582 btrfs_cpu_key_to_disk(&disk_key, key);
583 memcpy(ptr, &disk_key, sizeof(disk_key));
584 ptr += sizeof(disk_key);
585 memcpy(ptr, chunk, item_size);
586 item_size += sizeof(disk_key);
587 btrfs_set_super_sys_array_size(super_copy, array_size + item_size);
591 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
592 struct btrfs_root *extent_root, u64 *start,
593 u64 *num_bytes, u64 type)
596 struct btrfs_fs_info *info = extent_root->fs_info;
597 struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root;
598 struct btrfs_stripe *stripes;
599 struct btrfs_device *device = NULL;
600 struct btrfs_chunk *chunk;
601 struct list_head private_devs;
602 struct list_head *dev_list = &extent_root->fs_info->fs_devices->devices;
603 struct list_head *cur;
604 struct map_lookup *map;
606 u64 calc_size = 8 * 1024 * 1024;
607 u64 min_free = calc_size;
615 int stripe_len = 64 * 1024;
616 struct btrfs_key key;
618 if (list_empty(dev_list)) {
622 if (type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
623 BTRFS_BLOCK_GROUP_RAID10 |
624 BTRFS_BLOCK_GROUP_DUP)) {
625 if (type & BTRFS_BLOCK_GROUP_SYSTEM)
626 calc_size = 128 * 1024 * 1024;
628 calc_size = 1024 * 1024 * 1024;
630 if (type & BTRFS_BLOCK_GROUP_RAID1) {
631 num_stripes = min_t(u64, 2,
632 btrfs_super_num_devices(&info->super_copy));
634 if (type & BTRFS_BLOCK_GROUP_DUP)
636 if (type & (BTRFS_BLOCK_GROUP_RAID0))
637 num_stripes = btrfs_super_num_devices(&info->super_copy);
638 if (type & (BTRFS_BLOCK_GROUP_RAID10)) {
639 num_stripes = btrfs_super_num_devices(&info->super_copy);
642 num_stripes &= ~(u32)1;
646 INIT_LIST_HEAD(&private_devs);
647 cur = dev_list->next;
650 if (type & BTRFS_BLOCK_GROUP_DUP)
651 min_free = calc_size * 2;
653 /* build a private list of devices we will allocate from */
654 while(index < num_stripes) {
655 device = list_entry(cur, struct btrfs_device, dev_list);
656 avail = device->total_bytes - device->bytes_used;
658 if (avail > max_avail)
660 if (avail >= min_free) {
661 list_move_tail(&device->dev_list, &private_devs);
663 if (type & BTRFS_BLOCK_GROUP_DUP)
669 if (index < num_stripes) {
670 list_splice(&private_devs, dev_list);
671 if (!looped && max_avail > 0) {
673 calc_size = max_avail;
679 key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID;
680 key.type = BTRFS_CHUNK_ITEM_KEY;
681 ret = find_next_chunk(chunk_root, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
686 chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS);
690 map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS);
696 stripes = &chunk->stripe;
698 if (type & (BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_DUP))
699 *num_bytes = calc_size;
700 else if (type & BTRFS_BLOCK_GROUP_RAID10)
701 *num_bytes = calc_size * num_stripes / sub_stripes;
703 *num_bytes = calc_size * num_stripes;
706 printk("new chunk type %Lu start %Lu size %Lu\n", type, key.offset, *num_bytes);
707 while(index < num_stripes) {
708 struct btrfs_stripe *stripe;
709 BUG_ON(list_empty(&private_devs));
710 cur = private_devs.next;
711 device = list_entry(cur, struct btrfs_device, dev_list);
713 /* loop over this device again if we're doing a dup group */
714 if (!(type & BTRFS_BLOCK_GROUP_DUP) ||
715 (index == num_stripes - 1))
716 list_move_tail(&device->dev_list, dev_list);
718 ret = btrfs_alloc_dev_extent(trans, device,
719 info->chunk_root->root_key.objectid,
720 BTRFS_FIRST_CHUNK_TREE_OBJECTID, key.offset,
721 calc_size, &dev_offset);
723 printk("\talloc chunk size %llu from dev %llu phys %llu\n",
724 (unsigned long long)calc_size,
725 (unsigned long long)device->devid,
726 (unsigned long long)dev_offset);
727 device->bytes_used += calc_size;
728 ret = btrfs_update_device(trans, device);
731 map->stripes[index].dev = device;
732 map->stripes[index].physical = dev_offset;
733 stripe = stripes + index;
734 btrfs_set_stack_stripe_devid(stripe, device->devid);
735 btrfs_set_stack_stripe_offset(stripe, dev_offset);
736 memcpy(stripe->dev_uuid, device->uuid, BTRFS_UUID_SIZE);
737 physical = dev_offset;
740 BUG_ON(!list_empty(&private_devs));
742 /* key was set above */
743 btrfs_set_stack_chunk_length(chunk, *num_bytes);
744 btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid);
745 btrfs_set_stack_chunk_stripe_len(chunk, stripe_len);
746 btrfs_set_stack_chunk_type(chunk, type);
747 btrfs_set_stack_chunk_num_stripes(chunk, num_stripes);
748 btrfs_set_stack_chunk_io_align(chunk, stripe_len);
749 btrfs_set_stack_chunk_io_width(chunk, stripe_len);
750 btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize);
751 btrfs_set_stack_chunk_sub_stripes(chunk, sub_stripes);
752 map->sector_size = extent_root->sectorsize;
753 map->stripe_len = stripe_len;
754 map->io_align = stripe_len;
755 map->io_width = stripe_len;
757 map->num_stripes = num_stripes;
758 map->sub_stripes = sub_stripes;
760 ret = btrfs_insert_item(trans, chunk_root, &key, chunk,
761 btrfs_chunk_item_size(num_stripes));
763 *start = key.offset;;
765 map->ce.start = key.offset;
766 map->ce.size = *num_bytes;
768 ret = insert_existing_cache_extent(
769 &extent_root->fs_info->mapping_tree.cache_tree,
773 if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
774 ret = btrfs_add_system_chunk(trans, chunk_root, &key,
775 chunk, btrfs_chunk_item_size(num_stripes));
783 void btrfs_mapping_init(struct btrfs_mapping_tree *tree)
785 cache_tree_init(&tree->cache_tree);
788 int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len)
790 struct cache_extent *ce;
791 struct map_lookup *map;
795 ce = find_first_cache_extent(&map_tree->cache_tree, logical);
797 BUG_ON(ce->start > logical || ce->start + ce->size < logical);
798 map = container_of(ce, struct map_lookup, ce);
800 offset = logical - ce->start;
801 if (map->type & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1))
802 ret = map->num_stripes;
803 else if (map->type & BTRFS_BLOCK_GROUP_RAID10)
804 ret = map->sub_stripes;
810 int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
811 u64 logical, u64 *length,
812 struct btrfs_multi_bio **multi_ret, int mirror_num)
814 struct cache_extent *ce;
815 struct map_lookup *map;
819 int stripes_allocated = 8;
820 int stripes_required = 1;
823 struct btrfs_multi_bio *multi = NULL;
825 if (multi_ret && rw == READ) {
826 stripes_allocated = 1;
830 multi = kzalloc(btrfs_multi_bio_size(stripes_allocated),
836 ce = find_first_cache_extent(&map_tree->cache_tree, logical);
838 BUG_ON(ce->start > logical || ce->start + ce->size < logical);
839 map = container_of(ce, struct map_lookup, ce);
840 offset = logical - ce->start;
843 if (map->type & (BTRFS_BLOCK_GROUP_RAID1 |
844 BTRFS_BLOCK_GROUP_DUP)) {
845 stripes_required = map->num_stripes;
846 } else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
847 stripes_required = map->sub_stripes;
850 /* if our multi bio struct is too small, back off and try again */
851 if (multi_ret && rw == WRITE &&
852 stripes_allocated < stripes_required) {
853 stripes_allocated = map->num_stripes;
859 * stripe_nr counts the total number of stripes we have to stride
860 * to get to this block
862 stripe_nr = stripe_nr / map->stripe_len;
864 stripe_offset = stripe_nr * map->stripe_len;
865 BUG_ON(offset < stripe_offset);
867 /* stripe_offset is the offset of this block in its stripe*/
868 stripe_offset = offset - stripe_offset;
870 if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
871 BTRFS_BLOCK_GROUP_RAID10 |
872 BTRFS_BLOCK_GROUP_DUP)) {
873 /* we limit the length of each bio to what fits in a stripe */
874 *length = min_t(u64, ce->size - offset,
875 map->stripe_len - stripe_offset);
877 *length = ce->size - offset;
883 multi->num_stripes = 1;
885 if (map->type & BTRFS_BLOCK_GROUP_RAID1) {
887 multi->num_stripes = map->num_stripes;
889 stripe_index = mirror_num - 1;
891 stripe_index = stripe_nr % map->num_stripes;
892 } else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
893 int factor = map->num_stripes / map->sub_stripes;
895 stripe_index = stripe_nr % factor;
896 stripe_index *= map->sub_stripes;
899 multi->num_stripes = map->sub_stripes;
901 stripe_index += mirror_num - 1;
903 stripe_index = stripe_nr % map->sub_stripes;
905 stripe_nr = stripe_nr / factor;
906 } else if (map->type & BTRFS_BLOCK_GROUP_DUP) {
908 multi->num_stripes = map->num_stripes;
910 stripe_index = mirror_num - 1;
913 * after this do_div call, stripe_nr is the number of stripes
914 * on this device we have to walk to find the data, and
915 * stripe_index is the number of our device in the stripe array
917 stripe_index = stripe_nr % map->num_stripes;
918 stripe_nr = stripe_nr / map->num_stripes;
920 BUG_ON(stripe_index >= map->num_stripes);
922 BUG_ON(stripe_index != 0 && multi->num_stripes > 1);
923 for (i = 0; i < multi->num_stripes; i++) {
924 multi->stripes[i].physical =
925 map->stripes[stripe_index].physical + stripe_offset +
926 stripe_nr * map->stripe_len;
927 multi->stripes[i].dev = map->stripes[stripe_index].dev;
935 struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid,
938 struct list_head *head = &root->fs_info->fs_devices->devices;
940 return __find_device(head, devid, uuid);
943 int btrfs_bootstrap_super_map(struct btrfs_mapping_tree *map_tree,
944 struct btrfs_fs_devices *fs_devices)
946 struct map_lookup *map;
947 u64 logical = BTRFS_SUPER_INFO_OFFSET;
948 u64 length = BTRFS_SUPER_INFO_SIZE;
953 struct list_head *cur;
955 list_for_each(cur, &fs_devices->devices) {
958 map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS);
962 map->ce.start = logical;
963 map->ce.size = length;
964 map->num_stripes = num_stripes;
965 map->sub_stripes = sub_stripes;
966 map->io_width = length;
967 map->io_align = length;
968 map->sector_size = length;
969 map->stripe_len = length;
970 map->type = BTRFS_BLOCK_GROUP_RAID1;
973 list_for_each(cur, &fs_devices->devices) {
974 struct btrfs_device *device = list_entry(cur,
977 map->stripes[i].physical = logical;
978 map->stripes[i].dev = device;
981 ret = insert_existing_cache_extent(&map_tree->cache_tree, &map->ce);
982 if (ret == -EEXIST) {
983 struct cache_extent *old;
984 struct map_lookup *old_map;
985 old = find_cache_extent(&map_tree->cache_tree, logical, length);
986 old_map = container_of(old, struct map_lookup, ce);
987 remove_cache_extent(&map_tree->cache_tree, old);
989 ret = insert_existing_cache_extent(&map_tree->cache_tree,
996 static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
997 struct extent_buffer *leaf,
998 struct btrfs_chunk *chunk)
1000 struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree;
1001 struct map_lookup *map;
1002 struct cache_extent *ce;
1006 u64 super_offset_diff = 0;
1007 u8 uuid[BTRFS_UUID_SIZE];
1012 logical = key->offset;
1013 length = btrfs_chunk_length(leaf, chunk);
1015 if (logical < BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE) {
1016 super_offset_diff = BTRFS_SUPER_INFO_OFFSET +
1017 BTRFS_SUPER_INFO_SIZE - logical;
1018 logical = BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE;
1021 ce = find_first_cache_extent(&map_tree->cache_tree, logical);
1023 /* already mapped? */
1024 if (ce && ce->start <= logical && ce->start + ce->size > logical) {
1028 num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
1029 map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS);
1033 map->ce.start = logical;
1034 map->ce.size = length - super_offset_diff;
1035 map->num_stripes = num_stripes;
1036 map->io_width = btrfs_chunk_io_width(leaf, chunk);
1037 map->io_align = btrfs_chunk_io_align(leaf, chunk);
1038 map->sector_size = btrfs_chunk_sector_size(leaf, chunk);
1039 map->stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
1040 map->type = btrfs_chunk_type(leaf, chunk);
1041 map->sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk);
1043 for (i = 0; i < num_stripes; i++) {
1044 map->stripes[i].physical =
1045 btrfs_stripe_offset_nr(leaf, chunk, i) +
1047 devid = btrfs_stripe_devid_nr(leaf, chunk, i);
1048 read_extent_buffer(leaf, uuid, (unsigned long)
1049 btrfs_stripe_dev_uuid_nr(chunk, i),
1051 map->stripes[i].dev = btrfs_find_device(root, devid, uuid);
1052 if (!map->stripes[i].dev) {
1058 ret = insert_existing_cache_extent(&map_tree->cache_tree, &map->ce);
1064 static int fill_device_from_item(struct extent_buffer *leaf,
1065 struct btrfs_dev_item *dev_item,
1066 struct btrfs_device *device)
1070 device->devid = btrfs_device_id(leaf, dev_item);
1071 device->total_bytes = btrfs_device_total_bytes(leaf, dev_item);
1072 device->bytes_used = btrfs_device_bytes_used(leaf, dev_item);
1073 device->type = btrfs_device_type(leaf, dev_item);
1074 device->io_align = btrfs_device_io_align(leaf, dev_item);
1075 device->io_width = btrfs_device_io_width(leaf, dev_item);
1076 device->sector_size = btrfs_device_sector_size(leaf, dev_item);
1078 ptr = (unsigned long)btrfs_device_uuid(dev_item);
1079 read_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE);
1084 static int read_one_dev(struct btrfs_root *root,
1085 struct extent_buffer *leaf,
1086 struct btrfs_dev_item *dev_item)
1088 struct btrfs_device *device;
1091 u8 dev_uuid[BTRFS_UUID_SIZE];
1093 devid = btrfs_device_id(leaf, dev_item);
1094 read_extent_buffer(leaf, dev_uuid,
1095 (unsigned long)btrfs_device_uuid(dev_item),
1097 device = btrfs_find_device(root, devid, dev_uuid);
1099 printk("warning devid %llu not found already\n",
1100 (unsigned long long)devid);
1101 device = kmalloc(sizeof(*device), GFP_NOFS);
1104 device->total_ios = 0;
1105 list_add(&device->dev_list,
1106 &root->fs_info->fs_devices->devices);
1109 fill_device_from_item(leaf, dev_item, device);
1110 device->dev_root = root->fs_info->dev_root;
1114 int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf)
1116 struct btrfs_dev_item *dev_item;
1118 dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block,
1120 return read_one_dev(root, buf, dev_item);
1123 int btrfs_read_sys_array(struct btrfs_root *root)
1125 struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
1126 struct extent_buffer *sb = root->fs_info->sb_buffer;
1127 struct btrfs_disk_key *disk_key;
1128 struct btrfs_chunk *chunk;
1129 struct btrfs_key key;
1134 unsigned long sb_ptr;
1138 array_size = btrfs_super_sys_array_size(super_copy);
1141 * we do this loop twice, once for the device items and
1142 * once for all of the chunks. This way there are device
1143 * structs filled in for every chunk
1145 ptr = super_copy->sys_chunk_array;
1146 sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array);
1149 while (cur < array_size) {
1150 disk_key = (struct btrfs_disk_key *)ptr;
1151 btrfs_disk_key_to_cpu(&key, disk_key);
1153 len = sizeof(*disk_key);
1158 if (key.type == BTRFS_CHUNK_ITEM_KEY) {
1159 chunk = (struct btrfs_chunk *)sb_ptr;
1160 ret = read_one_chunk(root, &key, sb, chunk);
1162 num_stripes = btrfs_chunk_num_stripes(sb, chunk);
1163 len = btrfs_chunk_item_size(num_stripes);
1174 int btrfs_read_chunk_tree(struct btrfs_root *root)
1176 struct btrfs_path *path;
1177 struct extent_buffer *leaf;
1178 struct btrfs_key key;
1179 struct btrfs_key found_key;
1183 root = root->fs_info->chunk_root;
1185 path = btrfs_alloc_path();
1189 /* first we search for all of the device items, and then we
1190 * read in all of the chunk items. This way we can create chunk
1191 * mappings that reference all of the devices that are afound
1193 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
1197 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1199 leaf = path->nodes[0];
1200 slot = path->slots[0];
1201 if (slot >= btrfs_header_nritems(leaf)) {
1202 ret = btrfs_next_leaf(root, path);
1209 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1210 if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
1211 if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID)
1213 if (found_key.type == BTRFS_DEV_ITEM_KEY) {
1214 struct btrfs_dev_item *dev_item;
1215 dev_item = btrfs_item_ptr(leaf, slot,
1216 struct btrfs_dev_item);
1217 ret = read_one_dev(root, leaf, dev_item);
1220 } else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) {
1221 struct btrfs_chunk *chunk;
1222 chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
1223 ret = read_one_chunk(root, &found_key, leaf, chunk);
1227 if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
1229 btrfs_release_path(root, path);
1233 btrfs_free_path(path);