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;
46 struct btrfs_bio_stripe stripes[];
49 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
50 (sizeof(struct btrfs_bio_stripe) * (n)))
52 static LIST_HEAD(fs_uuids);
54 static struct btrfs_device *__find_device(struct list_head *head, u64 devid)
56 struct btrfs_device *dev;
57 struct list_head *cur;
59 list_for_each(cur, head) {
60 dev = list_entry(cur, struct btrfs_device, dev_list);
61 if (dev->devid == devid)
67 static struct btrfs_fs_devices *find_fsid(u8 *fsid)
69 struct list_head *cur;
70 struct btrfs_fs_devices *fs_devices;
72 list_for_each(cur, &fs_uuids) {
73 fs_devices = list_entry(cur, struct btrfs_fs_devices, list);
74 if (memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE) == 0)
80 static int device_list_add(const char *path,
81 struct btrfs_super_block *disk_super,
82 u64 devid, struct btrfs_fs_devices **fs_devices_ret)
84 struct btrfs_device *device;
85 struct btrfs_fs_devices *fs_devices;
86 u64 found_transid = btrfs_super_generation(disk_super);
88 fs_devices = find_fsid(disk_super->fsid);
90 fs_devices = kmalloc(sizeof(*fs_devices), GFP_NOFS);
93 INIT_LIST_HEAD(&fs_devices->devices);
94 list_add(&fs_devices->list, &fs_uuids);
95 memcpy(fs_devices->fsid, disk_super->fsid, BTRFS_FSID_SIZE);
96 fs_devices->latest_devid = devid;
97 fs_devices->latest_trans = found_transid;
98 fs_devices->lowest_devid = (u64)-1;
101 device = __find_device(&fs_devices->devices, devid);
104 device = kzalloc(sizeof(*device), GFP_NOFS);
106 /* we can safely leave the fs_devices entry around */
109 device->devid = devid;
110 device->name = kstrdup(path, GFP_NOFS);
115 list_add(&device->dev_list, &fs_devices->devices);
118 if (found_transid > fs_devices->latest_trans) {
119 fs_devices->latest_devid = devid;
120 fs_devices->latest_trans = found_transid;
122 if (fs_devices->lowest_devid > devid) {
123 fs_devices->lowest_devid = devid;
124 printk("lowest devid now %llu\n", (unsigned long long)devid);
126 *fs_devices_ret = fs_devices;
130 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
132 struct list_head *head = &fs_devices->devices;
133 struct list_head *cur;
134 struct btrfs_device *device;
136 list_for_each(cur, head) {
137 device = list_entry(cur, struct btrfs_device, dev_list);
143 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, int flags)
146 struct list_head *head = &fs_devices->devices;
147 struct list_head *cur;
148 struct btrfs_device *device;
151 list_for_each(cur, head) {
152 device = list_entry(cur, struct btrfs_device, dev_list);
153 fd = open(device->name, flags);
154 printk("opening %s devid %llu fd %d\n", device->name,
155 (unsigned long long)device->devid, fd);
160 if (device->devid == fs_devices->latest_devid)
161 fs_devices->latest_bdev = fd;
162 if (device->devid == fs_devices->lowest_devid)
163 fs_devices->lowest_bdev = fd;
168 btrfs_close_devices(fs_devices);
172 int btrfs_scan_one_device(int fd, const char *path,
173 struct btrfs_fs_devices **fs_devices_ret,
174 u64 *total_devs, u64 super_offset)
176 struct btrfs_super_block *disk_super;
186 ret = pread(fd, buf, 4096, super_offset);
191 disk_super = (struct btrfs_super_block *)buf;
192 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
193 sizeof(disk_super->magic))) {
197 devid = le64_to_cpu(disk_super->dev_item.devid);
198 *total_devs = btrfs_super_num_devices(disk_super);
199 printk("found device %llu on %s\n", (unsigned long long)devid, path);
200 ret = device_list_add(path, disk_super, devid, fs_devices_ret);
209 * this uses a pretty simple search, the expectation is that it is
210 * called very infrequently and that a given device has a small number
213 static int find_free_dev_extent(struct btrfs_trans_handle *trans,
214 struct btrfs_device *device,
215 struct btrfs_path *path,
216 u64 num_bytes, u64 *start)
218 struct btrfs_key key;
219 struct btrfs_root *root = device->dev_root;
220 struct btrfs_dev_extent *dev_extent = NULL;
223 u64 search_start = 0;
224 u64 search_end = device->total_bytes;
228 struct extent_buffer *l;
233 /* FIXME use last free of some kind */
235 /* we don't want to overwrite the superblock on the drive,
236 * so we make sure to start at an offset of at least 1MB
238 search_start = max((u64)1024 * 1024, search_start);
239 key.objectid = device->devid;
240 key.offset = search_start;
241 key.type = BTRFS_DEV_EXTENT_KEY;
242 ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
245 ret = btrfs_previous_item(root, path, 0, key.type);
249 btrfs_item_key_to_cpu(l, &key, path->slots[0]);
252 slot = path->slots[0];
253 if (slot >= btrfs_header_nritems(l)) {
254 ret = btrfs_next_leaf(root, path);
261 if (search_start >= search_end) {
265 *start = search_start;
269 *start = last_byte > search_start ?
270 last_byte : search_start;
271 if (search_end <= *start) {
277 btrfs_item_key_to_cpu(l, &key, slot);
279 if (key.objectid < device->devid)
282 if (key.objectid > device->devid)
285 if (key.offset >= search_start && key.offset > last_byte &&
287 if (last_byte < search_start)
288 last_byte = search_start;
289 hole_size = key.offset - last_byte;
290 if (key.offset > last_byte &&
291 hole_size >= num_bytes) {
296 if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY) {
301 dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent);
302 last_byte = key.offset + btrfs_dev_extent_length(l, dev_extent);
308 /* we have to make sure we didn't find an extent that has already
309 * been allocated by the map tree or the original allocation
311 btrfs_release_path(root, path);
312 BUG_ON(*start < search_start);
314 if (*start + num_bytes > search_end) {
318 /* check for pending inserts here */
322 btrfs_release_path(root, path);
326 int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
327 struct btrfs_device *device,
328 u64 chunk_tree, u64 chunk_objectid,
330 u64 num_bytes, u64 *start)
333 struct btrfs_path *path;
334 struct btrfs_root *root = device->dev_root;
335 struct btrfs_dev_extent *extent;
336 struct extent_buffer *leaf;
337 struct btrfs_key key;
339 path = btrfs_alloc_path();
343 ret = find_free_dev_extent(trans, device, path, num_bytes, start);
348 key.objectid = device->devid;
350 key.type = BTRFS_DEV_EXTENT_KEY;
351 ret = btrfs_insert_empty_item(trans, root, path, &key,
355 leaf = path->nodes[0];
356 extent = btrfs_item_ptr(leaf, path->slots[0],
357 struct btrfs_dev_extent);
358 btrfs_set_dev_extent_chunk_tree(leaf, extent, chunk_tree);
359 btrfs_set_dev_extent_chunk_objectid(leaf, extent, chunk_objectid);
360 btrfs_set_dev_extent_chunk_offset(leaf, extent, chunk_offset);
362 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
363 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(extent),
366 btrfs_set_dev_extent_length(leaf, extent, num_bytes);
367 btrfs_mark_buffer_dirty(leaf);
369 btrfs_free_path(path);
373 static int find_next_chunk(struct btrfs_root *root, u64 objectid, u64 *offset)
375 struct btrfs_path *path;
377 struct btrfs_key key;
378 struct btrfs_chunk *chunk;
379 struct btrfs_key found_key;
381 path = btrfs_alloc_path();
384 key.objectid = objectid;
385 key.offset = (u64)-1;
386 key.type = BTRFS_CHUNK_ITEM_KEY;
388 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
394 ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY);
398 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
400 if (found_key.objectid != objectid)
403 chunk = btrfs_item_ptr(path->nodes[0], path->slots[0],
405 *offset = found_key.offset +
406 btrfs_chunk_length(path->nodes[0], chunk);
411 btrfs_free_path(path);
415 static int find_next_devid(struct btrfs_root *root, struct btrfs_path *path,
419 struct btrfs_key key;
420 struct btrfs_key found_key;
422 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
423 key.type = BTRFS_DEV_ITEM_KEY;
424 key.offset = (u64)-1;
426 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
432 ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID,
437 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
439 *objectid = found_key.offset + 1;
443 btrfs_release_path(root, path);
448 * the device information is stored in the chunk root
449 * the btrfs_device struct should be fully filled in
451 int btrfs_add_device(struct btrfs_trans_handle *trans,
452 struct btrfs_root *root,
453 struct btrfs_device *device)
456 struct btrfs_path *path;
457 struct btrfs_dev_item *dev_item;
458 struct extent_buffer *leaf;
459 struct btrfs_key key;
463 root = root->fs_info->chunk_root;
465 path = btrfs_alloc_path();
469 ret = find_next_devid(root, path, &free_devid);
473 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
474 key.type = BTRFS_DEV_ITEM_KEY;
475 key.offset = free_devid;
477 ret = btrfs_insert_empty_item(trans, root, path, &key,
482 leaf = path->nodes[0];
483 dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
485 device->devid = free_devid;
486 btrfs_set_device_id(leaf, dev_item, device->devid);
487 btrfs_set_device_type(leaf, dev_item, device->type);
488 btrfs_set_device_io_align(leaf, dev_item, device->io_align);
489 btrfs_set_device_io_width(leaf, dev_item, device->io_width);
490 btrfs_set_device_sector_size(leaf, dev_item, device->sector_size);
491 btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes);
492 btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used);
493 btrfs_set_device_group(leaf, dev_item, 0);
494 btrfs_set_device_seek_speed(leaf, dev_item, 0);
495 btrfs_set_device_bandwidth(leaf, dev_item, 0);
497 ptr = (unsigned long)btrfs_device_uuid(dev_item);
498 write_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE);
499 btrfs_mark_buffer_dirty(leaf);
503 btrfs_free_path(path);
507 int btrfs_update_device(struct btrfs_trans_handle *trans,
508 struct btrfs_device *device)
511 struct btrfs_path *path;
512 struct btrfs_root *root;
513 struct btrfs_dev_item *dev_item;
514 struct extent_buffer *leaf;
515 struct btrfs_key key;
517 root = device->dev_root->fs_info->chunk_root;
519 path = btrfs_alloc_path();
523 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
524 key.type = BTRFS_DEV_ITEM_KEY;
525 key.offset = device->devid;
527 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
536 leaf = path->nodes[0];
537 dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
539 btrfs_set_device_id(leaf, dev_item, device->devid);
540 btrfs_set_device_type(leaf, dev_item, device->type);
541 btrfs_set_device_io_align(leaf, dev_item, device->io_align);
542 btrfs_set_device_io_width(leaf, dev_item, device->io_width);
543 btrfs_set_device_sector_size(leaf, dev_item, device->sector_size);
544 btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes);
545 btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used);
546 btrfs_mark_buffer_dirty(leaf);
549 btrfs_free_path(path);
553 int btrfs_add_system_chunk(struct btrfs_trans_handle *trans,
554 struct btrfs_root *root,
555 struct btrfs_key *key,
556 struct btrfs_chunk *chunk, int item_size)
558 struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
559 struct btrfs_disk_key disk_key;
563 array_size = btrfs_super_sys_array_size(super_copy);
564 if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE)
567 ptr = super_copy->sys_chunk_array + array_size;
568 btrfs_cpu_key_to_disk(&disk_key, key);
569 memcpy(ptr, &disk_key, sizeof(disk_key));
570 ptr += sizeof(disk_key);
571 memcpy(ptr, chunk, item_size);
572 item_size += sizeof(disk_key);
573 btrfs_set_super_sys_array_size(super_copy, array_size + item_size);
577 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
578 struct btrfs_root *extent_root, u64 *start,
579 u64 *num_bytes, u64 type)
582 struct btrfs_fs_info *info = extent_root->fs_info;
583 struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root;
584 struct btrfs_stripe *stripes;
585 struct btrfs_device *device = NULL;
586 struct btrfs_chunk *chunk;
587 struct list_head private_devs;
588 struct list_head *dev_list = &extent_root->fs_info->fs_devices->devices;
589 struct list_head *cur;
590 struct map_lookup *map;
592 u64 calc_size = 8 * 1024 * 1024;
593 u64 min_free = calc_size;
600 int stripe_len = 64 * 1024;
601 struct btrfs_key key;
603 if (list_empty(dev_list)) {
607 if (type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
608 BTRFS_BLOCK_GROUP_DUP)) {
609 if (type & BTRFS_BLOCK_GROUP_SYSTEM)
610 calc_size = 128 * 1024 * 1024;
612 calc_size = 1024 * 1024 * 1024;
614 if (type & BTRFS_BLOCK_GROUP_RAID1) {
615 num_stripes = min_t(u64, 2,
616 btrfs_super_num_devices(&info->super_copy));
618 if (type & BTRFS_BLOCK_GROUP_DUP)
620 if (type & (BTRFS_BLOCK_GROUP_RAID0))
621 num_stripes = btrfs_super_num_devices(&info->super_copy);
623 INIT_LIST_HEAD(&private_devs);
624 cur = dev_list->next;
627 if (type & BTRFS_BLOCK_GROUP_DUP)
628 min_free = calc_size * 2;
630 /* build a private list of devices we will allocate from */
631 while(index < num_stripes) {
632 device = list_entry(cur, struct btrfs_device, dev_list);
633 avail = device->total_bytes - device->bytes_used;
635 if (avail > max_avail)
637 if (avail >= min_free) {
638 list_move_tail(&device->dev_list, &private_devs);
640 if (type & BTRFS_BLOCK_GROUP_DUP)
646 if (index < num_stripes) {
647 list_splice(&private_devs, dev_list);
648 if (!looped && max_avail > 0) {
650 calc_size = max_avail;
656 key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID;
657 key.type = BTRFS_CHUNK_ITEM_KEY;
658 ret = find_next_chunk(chunk_root, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
663 chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS);
667 map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS);
673 stripes = &chunk->stripe;
675 if (type & (BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_DUP))
676 *num_bytes = calc_size;
678 *num_bytes = calc_size * num_stripes;
681 printk("new chunk type %Lu start %Lu size %Lu\n", type, key.offset, *num_bytes);
682 while(index < num_stripes) {
683 struct btrfs_stripe *stripe;
684 BUG_ON(list_empty(&private_devs));
685 cur = private_devs.next;
686 device = list_entry(cur, struct btrfs_device, dev_list);
688 /* loop over this device again if we're doing a dup group */
689 if (!(type & BTRFS_BLOCK_GROUP_DUP) ||
690 (index == num_stripes - 1))
691 list_move_tail(&device->dev_list, dev_list);
693 ret = btrfs_alloc_dev_extent(trans, device,
694 info->chunk_root->root_key.objectid,
695 BTRFS_FIRST_CHUNK_TREE_OBJECTID, key.offset,
696 calc_size, &dev_offset);
698 printk("\talloc chunk size %llu from dev %llu phys %llu\n",
699 (unsigned long long)calc_size,
700 (unsigned long long)device->devid,
701 (unsigned long long)dev_offset);
702 device->bytes_used += calc_size;
703 ret = btrfs_update_device(trans, device);
706 map->stripes[index].dev = device;
707 map->stripes[index].physical = dev_offset;
708 stripe = stripes + index;
709 btrfs_set_stack_stripe_devid(stripe, device->devid);
710 btrfs_set_stack_stripe_offset(stripe, dev_offset);
711 memcpy(stripe->dev_uuid, device->uuid, BTRFS_UUID_SIZE);
712 physical = dev_offset;
715 BUG_ON(!list_empty(&private_devs));
717 /* key was set above */
718 btrfs_set_stack_chunk_length(chunk, *num_bytes);
719 btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid);
720 btrfs_set_stack_chunk_stripe_len(chunk, stripe_len);
721 btrfs_set_stack_chunk_type(chunk, type);
722 btrfs_set_stack_chunk_num_stripes(chunk, num_stripes);
723 btrfs_set_stack_chunk_io_align(chunk, stripe_len);
724 btrfs_set_stack_chunk_io_width(chunk, stripe_len);
725 btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize);
726 map->sector_size = extent_root->sectorsize;
727 map->stripe_len = stripe_len;
728 map->io_align = stripe_len;
729 map->io_width = stripe_len;
731 map->num_stripes = num_stripes;
733 ret = btrfs_insert_item(trans, chunk_root, &key, chunk,
734 btrfs_chunk_item_size(num_stripes));
736 *start = key.offset;;
738 map->ce.start = key.offset;
739 map->ce.size = *num_bytes;
741 ret = insert_existing_cache_extent(
742 &extent_root->fs_info->mapping_tree.cache_tree,
746 if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
747 ret = btrfs_add_system_chunk(trans, chunk_root, &key,
748 chunk, btrfs_chunk_item_size(num_stripes));
756 void btrfs_mapping_init(struct btrfs_mapping_tree *tree)
758 cache_tree_init(&tree->cache_tree);
761 int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len)
763 struct cache_extent *ce;
764 struct map_lookup *map;
768 ce = find_first_cache_extent(&map_tree->cache_tree, logical);
770 BUG_ON(ce->start > logical || ce->start + ce->size < logical);
771 map = container_of(ce, struct map_lookup, ce);
773 offset = logical - ce->start;
774 if (map->type & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1))
775 ret = map->num_stripes;
781 int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
782 u64 logical, u64 *length,
783 struct btrfs_multi_bio **multi_ret, int mirror_num)
785 struct cache_extent *ce;
786 struct map_lookup *map;
790 int stripes_allocated = 8;
793 struct btrfs_multi_bio *multi = NULL;
795 if (multi_ret && rw == READ) {
796 stripes_allocated = 1;
800 multi = kzalloc(btrfs_multi_bio_size(stripes_allocated),
806 ce = find_first_cache_extent(&map_tree->cache_tree, logical);
808 BUG_ON(ce->start > logical || ce->start + ce->size < logical);
809 map = container_of(ce, struct map_lookup, ce);
810 offset = logical - ce->start;
812 /* if our multi bio struct is too small, back off and try again */
813 if (multi_ret && (rw == WRITE) &&
814 stripes_allocated < map->num_stripes &&
815 ((map->type & BTRFS_BLOCK_GROUP_RAID1) ||
816 (map->type & BTRFS_BLOCK_GROUP_DUP))) {
817 stripes_allocated = map->num_stripes;
823 * stripe_nr counts the total number of stripes we have to stride
824 * to get to this block
826 stripe_nr = stripe_nr / map->stripe_len;
828 stripe_offset = stripe_nr * map->stripe_len;
829 BUG_ON(offset < stripe_offset);
831 /* stripe_offset is the offset of this block in its stripe*/
832 stripe_offset = offset - stripe_offset;
834 if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
835 BTRFS_BLOCK_GROUP_DUP)) {
836 /* we limit the length of each bio to what fits in a stripe */
837 *length = min_t(u64, ce->size - offset,
838 map->stripe_len - stripe_offset);
840 *length = ce->size - offset;
846 multi->num_stripes = 1;
848 if (map->type & BTRFS_BLOCK_GROUP_RAID1) {
850 multi->num_stripes = map->num_stripes;
852 stripe_index = mirror_num - 1;
854 stripe_index = stripe_nr % map->num_stripes;
855 } else if (map->type & BTRFS_BLOCK_GROUP_DUP) {
857 multi->num_stripes = map->num_stripes;
859 stripe_index = mirror_num - 1;
862 * after this do_div call, stripe_nr is the number of stripes
863 * on this device we have to walk to find the data, and
864 * stripe_index is the number of our device in the stripe array
866 stripe_index = stripe_nr % map->num_stripes;
867 stripe_nr = stripe_nr / map->num_stripes;
869 BUG_ON(stripe_index >= map->num_stripes);
871 BUG_ON(stripe_index != 0 && multi->num_stripes > 1);
872 for (i = 0; i < multi->num_stripes; i++) {
873 multi->stripes[i].physical =
874 map->stripes[stripe_index].physical + stripe_offset +
875 stripe_nr * map->stripe_len;
876 multi->stripes[i].dev = map->stripes[stripe_index].dev;
884 struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid)
886 struct list_head *head = &root->fs_info->fs_devices->devices;
888 return __find_device(head, devid);
891 int btrfs_bootstrap_super_map(struct btrfs_mapping_tree *map_tree,
892 struct btrfs_fs_devices *fs_devices)
894 struct map_lookup *map;
895 u64 logical = BTRFS_SUPER_INFO_OFFSET;
896 u64 length = BTRFS_SUPER_INFO_SIZE;
900 struct list_head *cur;
902 list_for_each(cur, &fs_devices->devices) {
905 map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS);
909 map->ce.start = logical;
910 map->ce.size = length;
911 map->num_stripes = num_stripes;
912 map->io_width = length;
913 map->io_align = length;
914 map->sector_size = length;
915 map->stripe_len = length;
916 map->type = BTRFS_BLOCK_GROUP_RAID1;
919 list_for_each(cur, &fs_devices->devices) {
920 struct btrfs_device *device = list_entry(cur,
923 map->stripes[i].physical = logical;
924 map->stripes[i].dev = device;
927 ret = insert_existing_cache_extent(&map_tree->cache_tree, &map->ce);
928 if (ret == -EEXIST) {
929 struct cache_extent *old;
930 struct map_lookup *old_map;
931 old = find_cache_extent(&map_tree->cache_tree, logical, length);
932 old_map = container_of(old, struct map_lookup, ce);
933 remove_cache_extent(&map_tree->cache_tree, old);
935 ret = insert_existing_cache_extent(&map_tree->cache_tree,
942 static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
943 struct extent_buffer *leaf,
944 struct btrfs_chunk *chunk)
946 struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree;
947 struct map_lookup *map;
948 struct cache_extent *ce;
952 u64 super_offset_diff = 0;
957 logical = key->offset;
958 length = btrfs_chunk_length(leaf, chunk);
960 if (logical < BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE) {
961 super_offset_diff = BTRFS_SUPER_INFO_OFFSET +
962 BTRFS_SUPER_INFO_SIZE - logical;
963 logical = BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE;
966 ce = find_first_cache_extent(&map_tree->cache_tree, logical);
968 /* already mapped? */
969 if (ce && ce->start <= logical && ce->start + ce->size > logical) {
973 num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
974 map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS);
978 map->ce.start = logical;
979 map->ce.size = length - super_offset_diff;
980 map->num_stripes = num_stripes;
981 map->io_width = btrfs_chunk_io_width(leaf, chunk);
982 map->io_align = btrfs_chunk_io_align(leaf, chunk);
983 map->sector_size = btrfs_chunk_sector_size(leaf, chunk);
984 map->stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
985 map->type = btrfs_chunk_type(leaf, chunk);
987 for (i = 0; i < num_stripes; i++) {
988 map->stripes[i].physical =
989 btrfs_stripe_offset_nr(leaf, chunk, i) +
991 devid = btrfs_stripe_devid_nr(leaf, chunk, i);
992 map->stripes[i].dev = btrfs_find_device(root, devid);
993 if (!map->stripes[i].dev) {
999 ret = insert_existing_cache_extent(&map_tree->cache_tree, &map->ce);
1005 static int fill_device_from_item(struct extent_buffer *leaf,
1006 struct btrfs_dev_item *dev_item,
1007 struct btrfs_device *device)
1011 device->devid = btrfs_device_id(leaf, dev_item);
1012 device->total_bytes = btrfs_device_total_bytes(leaf, dev_item);
1013 device->bytes_used = btrfs_device_bytes_used(leaf, dev_item);
1014 device->type = btrfs_device_type(leaf, dev_item);
1015 device->io_align = btrfs_device_io_align(leaf, dev_item);
1016 device->io_width = btrfs_device_io_width(leaf, dev_item);
1017 device->sector_size = btrfs_device_sector_size(leaf, dev_item);
1019 ptr = (unsigned long)btrfs_device_uuid(dev_item);
1020 read_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE);
1025 static int read_one_dev(struct btrfs_root *root,
1026 struct extent_buffer *leaf,
1027 struct btrfs_dev_item *dev_item)
1029 struct btrfs_device *device;
1033 devid = btrfs_device_id(leaf, dev_item);
1034 device = btrfs_find_device(root, devid);
1036 printk("warning devid %llu not found already\n",
1037 (unsigned long long)devid);
1038 device = kmalloc(sizeof(*device), GFP_NOFS);
1041 device->total_ios = 0;
1042 list_add(&device->dev_list,
1043 &root->fs_info->fs_devices->devices);
1046 fill_device_from_item(leaf, dev_item, device);
1047 device->dev_root = root->fs_info->dev_root;
1051 int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf)
1053 struct btrfs_dev_item *dev_item;
1055 dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block,
1057 return read_one_dev(root, buf, dev_item);
1060 int btrfs_read_sys_array(struct btrfs_root *root)
1062 struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
1063 struct extent_buffer *sb = root->fs_info->sb_buffer;
1064 struct btrfs_disk_key *disk_key;
1065 struct btrfs_chunk *chunk;
1066 struct btrfs_key key;
1071 unsigned long sb_ptr;
1075 array_size = btrfs_super_sys_array_size(super_copy);
1078 * we do this loop twice, once for the device items and
1079 * once for all of the chunks. This way there are device
1080 * structs filled in for every chunk
1082 ptr = super_copy->sys_chunk_array;
1083 sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array);
1086 while (cur < array_size) {
1087 disk_key = (struct btrfs_disk_key *)ptr;
1088 btrfs_disk_key_to_cpu(&key, disk_key);
1090 len = sizeof(*disk_key);
1095 if (key.type == BTRFS_CHUNK_ITEM_KEY) {
1096 chunk = (struct btrfs_chunk *)sb_ptr;
1097 ret = read_one_chunk(root, &key, sb, chunk);
1099 num_stripes = btrfs_chunk_num_stripes(sb, chunk);
1100 len = btrfs_chunk_item_size(num_stripes);
1111 int btrfs_read_chunk_tree(struct btrfs_root *root)
1113 struct btrfs_path *path;
1114 struct extent_buffer *leaf;
1115 struct btrfs_key key;
1116 struct btrfs_key found_key;
1120 root = root->fs_info->chunk_root;
1122 path = btrfs_alloc_path();
1126 /* first we search for all of the device items, and then we
1127 * read in all of the chunk items. This way we can create chunk
1128 * mappings that reference all of the devices that are afound
1130 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
1134 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1136 leaf = path->nodes[0];
1137 slot = path->slots[0];
1138 if (slot >= btrfs_header_nritems(leaf)) {
1139 ret = btrfs_next_leaf(root, path);
1146 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1147 if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
1148 if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID)
1150 if (found_key.type == BTRFS_DEV_ITEM_KEY) {
1151 struct btrfs_dev_item *dev_item;
1152 dev_item = btrfs_item_ptr(leaf, slot,
1153 struct btrfs_dev_item);
1154 ret = read_one_dev(root, leaf, dev_item);
1157 } else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) {
1158 struct btrfs_chunk *chunk;
1159 chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
1160 ret = read_one_chunk(root, &found_key, leaf, chunk);
1164 if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
1166 btrfs_release_path(root, path);
1170 btrfs_free_path(path);