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 owner, u64 num_bytes, u64 *start)
331 struct btrfs_path *path;
332 struct btrfs_root *root = device->dev_root;
333 struct btrfs_dev_extent *extent;
334 struct extent_buffer *leaf;
335 struct btrfs_key key;
337 path = btrfs_alloc_path();
341 ret = find_free_dev_extent(trans, device, path, num_bytes, start);
346 key.objectid = device->devid;
348 key.type = BTRFS_DEV_EXTENT_KEY;
349 ret = btrfs_insert_empty_item(trans, root, path, &key,
353 leaf = path->nodes[0];
354 extent = btrfs_item_ptr(leaf, path->slots[0],
355 struct btrfs_dev_extent);
356 btrfs_set_dev_extent_owner(leaf, extent, owner);
357 btrfs_set_dev_extent_length(leaf, extent, num_bytes);
358 btrfs_mark_buffer_dirty(leaf);
360 btrfs_free_path(path);
364 static int find_next_chunk(struct btrfs_root *root, u64 *objectid)
366 struct btrfs_path *path;
368 struct btrfs_key key;
369 struct btrfs_key found_key;
371 path = btrfs_alloc_path();
374 key.objectid = (u64)-1;
375 key.offset = (u64)-1;
376 key.type = BTRFS_CHUNK_ITEM_KEY;
378 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
384 ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY);
388 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
390 *objectid = found_key.objectid + found_key.offset;
394 btrfs_free_path(path);
398 static int find_next_devid(struct btrfs_root *root, struct btrfs_path *path,
402 struct btrfs_key key;
403 struct btrfs_key found_key;
405 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
406 key.type = BTRFS_DEV_ITEM_KEY;
407 key.offset = (u64)-1;
409 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
415 ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID,
420 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
422 *objectid = found_key.offset + 1;
426 btrfs_release_path(root, path);
431 * the device information is stored in the chunk root
432 * the btrfs_device struct should be fully filled in
434 int btrfs_add_device(struct btrfs_trans_handle *trans,
435 struct btrfs_root *root,
436 struct btrfs_device *device)
439 struct btrfs_path *path;
440 struct btrfs_dev_item *dev_item;
441 struct extent_buffer *leaf;
442 struct btrfs_key key;
446 root = root->fs_info->chunk_root;
448 path = btrfs_alloc_path();
452 ret = find_next_devid(root, path, &free_devid);
456 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
457 key.type = BTRFS_DEV_ITEM_KEY;
458 key.offset = free_devid;
460 ret = btrfs_insert_empty_item(trans, root, path, &key,
465 leaf = path->nodes[0];
466 dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
468 device->devid = free_devid;
469 btrfs_set_device_id(leaf, dev_item, device->devid);
470 btrfs_set_device_type(leaf, dev_item, device->type);
471 btrfs_set_device_io_align(leaf, dev_item, device->io_align);
472 btrfs_set_device_io_width(leaf, dev_item, device->io_width);
473 btrfs_set_device_sector_size(leaf, dev_item, device->sector_size);
474 btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes);
475 btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used);
477 ptr = (unsigned long)btrfs_device_uuid(dev_item);
478 write_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE);
479 btrfs_mark_buffer_dirty(leaf);
483 btrfs_free_path(path);
487 int btrfs_update_device(struct btrfs_trans_handle *trans,
488 struct btrfs_device *device)
491 struct btrfs_path *path;
492 struct btrfs_root *root;
493 struct btrfs_dev_item *dev_item;
494 struct extent_buffer *leaf;
495 struct btrfs_key key;
497 root = device->dev_root->fs_info->chunk_root;
499 path = btrfs_alloc_path();
503 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
504 key.type = BTRFS_DEV_ITEM_KEY;
505 key.offset = device->devid;
507 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
516 leaf = path->nodes[0];
517 dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
519 btrfs_set_device_id(leaf, dev_item, device->devid);
520 btrfs_set_device_type(leaf, dev_item, device->type);
521 btrfs_set_device_io_align(leaf, dev_item, device->io_align);
522 btrfs_set_device_io_width(leaf, dev_item, device->io_width);
523 btrfs_set_device_sector_size(leaf, dev_item, device->sector_size);
524 btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes);
525 btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used);
526 btrfs_mark_buffer_dirty(leaf);
529 btrfs_free_path(path);
533 int btrfs_add_system_chunk(struct btrfs_trans_handle *trans,
534 struct btrfs_root *root,
535 struct btrfs_key *key,
536 struct btrfs_chunk *chunk, int item_size)
538 struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
539 struct btrfs_disk_key disk_key;
543 array_size = btrfs_super_sys_array_size(super_copy);
544 if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE)
547 ptr = super_copy->sys_chunk_array + array_size;
548 btrfs_cpu_key_to_disk(&disk_key, key);
549 memcpy(ptr, &disk_key, sizeof(disk_key));
550 ptr += sizeof(disk_key);
551 memcpy(ptr, chunk, item_size);
552 item_size += sizeof(disk_key);
553 btrfs_set_super_sys_array_size(super_copy, array_size + item_size);
557 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
558 struct btrfs_root *extent_root, u64 *start,
559 u64 *num_bytes, u64 type)
562 struct btrfs_fs_info *info = extent_root->fs_info;
563 struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root;
564 struct btrfs_stripe *stripes;
565 struct btrfs_device *device = NULL;
566 struct btrfs_chunk *chunk;
567 struct list_head private_devs;
568 struct list_head *dev_list = &extent_root->fs_info->fs_devices->devices;
569 struct list_head *cur;
570 struct map_lookup *map;
572 u64 calc_size = 8 * 1024 * 1024;
573 u64 min_free = calc_size;
580 int stripe_len = 64 * 1024;
581 struct btrfs_key key;
583 if (list_empty(dev_list)) {
587 if (type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
588 BTRFS_BLOCK_GROUP_DUP)) {
589 if (type & BTRFS_BLOCK_GROUP_SYSTEM)
590 calc_size = 128 * 1024 * 1024;
592 calc_size = 1024 * 1024 * 1024;
594 if (type & BTRFS_BLOCK_GROUP_RAID1) {
595 num_stripes = min_t(u64, 2,
596 btrfs_super_num_devices(&info->super_copy));
598 if (type & BTRFS_BLOCK_GROUP_DUP)
600 if (type & (BTRFS_BLOCK_GROUP_RAID0))
601 num_stripes = btrfs_super_num_devices(&info->super_copy);
603 INIT_LIST_HEAD(&private_devs);
604 cur = dev_list->next;
607 if (type & BTRFS_BLOCK_GROUP_DUP)
608 min_free = calc_size * 2;
610 /* build a private list of devices we will allocate from */
611 while(index < num_stripes) {
612 device = list_entry(cur, struct btrfs_device, dev_list);
613 avail = device->total_bytes - device->bytes_used;
615 if (avail > max_avail)
617 if (avail >= min_free) {
618 list_move_tail(&device->dev_list, &private_devs);
620 if (type & BTRFS_BLOCK_GROUP_DUP)
626 if (index < num_stripes) {
627 list_splice(&private_devs, dev_list);
628 if (!looped && max_avail > 0) {
630 calc_size = max_avail;
636 ret = find_next_chunk(chunk_root, &key.objectid);
640 chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS);
644 map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS);
650 stripes = &chunk->stripe;
652 if (type & (BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_DUP))
653 *num_bytes = calc_size;
655 *num_bytes = calc_size * num_stripes;
658 printk("new chunk type %Lu start %Lu size %Lu\n", type, key.objectid, *num_bytes);
659 while(index < num_stripes) {
660 BUG_ON(list_empty(&private_devs));
661 cur = private_devs.next;
662 device = list_entry(cur, struct btrfs_device, dev_list);
664 /* loop over this device again if we're doing a dup group */
665 if (!(type & BTRFS_BLOCK_GROUP_DUP) ||
666 (index == num_stripes - 1))
667 list_move_tail(&device->dev_list, dev_list);
669 ret = btrfs_alloc_dev_extent(trans, device,
671 calc_size, &dev_offset);
673 printk("\talloc chunk size %llu from dev %llu phys %llu\n",
674 (unsigned long long)calc_size,
675 (unsigned long long)device->devid,
676 (unsigned long long)dev_offset);
677 device->bytes_used += calc_size;
678 ret = btrfs_update_device(trans, device);
681 map->stripes[index].dev = device;
682 map->stripes[index].physical = dev_offset;
683 btrfs_set_stack_stripe_devid(stripes + index, device->devid);
684 btrfs_set_stack_stripe_offset(stripes + index, dev_offset);
685 physical = dev_offset;
688 BUG_ON(!list_empty(&private_devs));
690 /* key.objectid was set above */
691 key.offset = *num_bytes;
692 key.type = BTRFS_CHUNK_ITEM_KEY;
693 btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid);
694 btrfs_set_stack_chunk_stripe_len(chunk, stripe_len);
695 btrfs_set_stack_chunk_type(chunk, type);
696 btrfs_set_stack_chunk_num_stripes(chunk, num_stripes);
697 btrfs_set_stack_chunk_io_align(chunk, stripe_len);
698 btrfs_set_stack_chunk_io_width(chunk, stripe_len);
699 btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize);
700 map->sector_size = extent_root->sectorsize;
701 map->stripe_len = stripe_len;
702 map->io_align = stripe_len;
703 map->io_width = stripe_len;
705 map->num_stripes = num_stripes;
707 ret = btrfs_insert_item(trans, chunk_root, &key, chunk,
708 btrfs_chunk_item_size(num_stripes));
710 *start = key.objectid;
712 map->ce.start = key.objectid;
713 map->ce.size = key.offset;
715 ret = insert_existing_cache_extent(
716 &extent_root->fs_info->mapping_tree.cache_tree,
720 if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
721 ret = btrfs_add_system_chunk(trans, chunk_root, &key,
722 chunk, btrfs_chunk_item_size(num_stripes));
730 void btrfs_mapping_init(struct btrfs_mapping_tree *tree)
732 cache_tree_init(&tree->cache_tree);
735 int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len)
737 struct cache_extent *ce;
738 struct map_lookup *map;
742 ce = find_first_cache_extent(&map_tree->cache_tree, logical);
744 BUG_ON(ce->start > logical || ce->start + ce->size < logical);
745 map = container_of(ce, struct map_lookup, ce);
747 offset = logical - ce->start;
748 if (map->type & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1))
749 ret = map->num_stripes;
755 int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
756 u64 logical, u64 *length,
757 struct btrfs_multi_bio **multi_ret, int mirror_num)
759 struct cache_extent *ce;
760 struct map_lookup *map;
764 int stripes_allocated = 8;
767 struct btrfs_multi_bio *multi = NULL;
769 if (multi_ret && rw == READ) {
770 stripes_allocated = 1;
774 multi = kzalloc(btrfs_multi_bio_size(stripes_allocated),
780 ce = find_first_cache_extent(&map_tree->cache_tree, logical);
782 BUG_ON(ce->start > logical || ce->start + ce->size < logical);
783 map = container_of(ce, struct map_lookup, ce);
784 offset = logical - ce->start;
786 /* if our multi bio struct is too small, back off and try again */
787 if (multi_ret && (rw == WRITE) &&
788 stripes_allocated < map->num_stripes &&
789 ((map->type & BTRFS_BLOCK_GROUP_RAID1) ||
790 (map->type & BTRFS_BLOCK_GROUP_DUP))) {
791 stripes_allocated = map->num_stripes;
797 * stripe_nr counts the total number of stripes we have to stride
798 * to get to this block
800 stripe_nr = stripe_nr / map->stripe_len;
802 stripe_offset = stripe_nr * map->stripe_len;
803 BUG_ON(offset < stripe_offset);
805 /* stripe_offset is the offset of this block in its stripe*/
806 stripe_offset = offset - stripe_offset;
808 if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
809 BTRFS_BLOCK_GROUP_DUP)) {
810 /* we limit the length of each bio to what fits in a stripe */
811 *length = min_t(u64, ce->size - offset,
812 map->stripe_len - stripe_offset);
814 *length = ce->size - offset;
820 multi->num_stripes = 1;
822 if (map->type & BTRFS_BLOCK_GROUP_RAID1) {
824 multi->num_stripes = map->num_stripes;
826 stripe_index = mirror_num - 1;
828 stripe_index = stripe_nr % map->num_stripes;
829 } else if (map->type & BTRFS_BLOCK_GROUP_DUP) {
831 multi->num_stripes = map->num_stripes;
833 stripe_index = mirror_num - 1;
836 * after this do_div call, stripe_nr is the number of stripes
837 * on this device we have to walk to find the data, and
838 * stripe_index is the number of our device in the stripe array
840 stripe_index = stripe_nr % map->num_stripes;
841 stripe_nr = stripe_nr / map->num_stripes;
843 BUG_ON(stripe_index >= map->num_stripes);
845 BUG_ON(stripe_index != 0 && multi->num_stripes > 1);
846 for (i = 0; i < multi->num_stripes; i++) {
847 multi->stripes[i].physical =
848 map->stripes[stripe_index].physical + stripe_offset +
849 stripe_nr * map->stripe_len;
850 multi->stripes[i].dev = map->stripes[stripe_index].dev;
858 struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid)
860 struct list_head *head = &root->fs_info->fs_devices->devices;
862 return __find_device(head, devid);
865 int btrfs_bootstrap_super_map(struct btrfs_mapping_tree *map_tree,
866 struct btrfs_fs_devices *fs_devices)
868 struct map_lookup *map;
869 u64 logical = BTRFS_SUPER_INFO_OFFSET;
870 u64 length = BTRFS_SUPER_INFO_SIZE;
874 struct list_head *cur;
876 list_for_each(cur, &fs_devices->devices) {
879 map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS);
883 map->ce.start = logical;
884 map->ce.size = length;
885 map->num_stripes = num_stripes;
886 map->io_width = length;
887 map->io_align = length;
888 map->sector_size = length;
889 map->stripe_len = length;
890 map->type = BTRFS_BLOCK_GROUP_RAID1;
893 list_for_each(cur, &fs_devices->devices) {
894 struct btrfs_device *device = list_entry(cur,
897 map->stripes[i].physical = logical;
898 map->stripes[i].dev = device;
901 ret = insert_existing_cache_extent(&map_tree->cache_tree, &map->ce);
902 if (ret == -EEXIST) {
903 struct cache_extent *old;
904 struct map_lookup *old_map;
905 old = find_cache_extent(&map_tree->cache_tree, logical, length);
906 old_map = container_of(old, struct map_lookup, ce);
907 remove_cache_extent(&map_tree->cache_tree, old);
909 ret = insert_existing_cache_extent(&map_tree->cache_tree,
916 static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
917 struct extent_buffer *leaf,
918 struct btrfs_chunk *chunk)
920 struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree;
921 struct map_lookup *map;
922 struct cache_extent *ce;
926 u64 super_offset_diff = 0;
931 logical = key->objectid;
932 length = key->offset;
934 if (logical < BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE) {
935 super_offset_diff = BTRFS_SUPER_INFO_OFFSET +
936 BTRFS_SUPER_INFO_SIZE - logical;
937 logical = BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE;
940 ce = find_first_cache_extent(&map_tree->cache_tree, logical);
942 /* already mapped? */
943 if (ce && ce->start <= logical && ce->start + ce->size > logical) {
947 num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
948 map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS);
952 map->ce.start = logical;
953 map->ce.size = length - super_offset_diff;
954 map->num_stripes = num_stripes;
955 map->io_width = btrfs_chunk_io_width(leaf, chunk);
956 map->io_align = btrfs_chunk_io_align(leaf, chunk);
957 map->sector_size = btrfs_chunk_sector_size(leaf, chunk);
958 map->stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
959 map->type = btrfs_chunk_type(leaf, chunk);
961 for (i = 0; i < num_stripes; i++) {
962 map->stripes[i].physical =
963 btrfs_stripe_offset_nr(leaf, chunk, i) +
965 devid = btrfs_stripe_devid_nr(leaf, chunk, i);
966 map->stripes[i].dev = btrfs_find_device(root, devid);
967 if (!map->stripes[i].dev) {
973 ret = insert_existing_cache_extent(&map_tree->cache_tree, &map->ce);
979 static int fill_device_from_item(struct extent_buffer *leaf,
980 struct btrfs_dev_item *dev_item,
981 struct btrfs_device *device)
985 device->devid = btrfs_device_id(leaf, dev_item);
986 device->total_bytes = btrfs_device_total_bytes(leaf, dev_item);
987 device->bytes_used = btrfs_device_bytes_used(leaf, dev_item);
988 device->type = btrfs_device_type(leaf, dev_item);
989 device->io_align = btrfs_device_io_align(leaf, dev_item);
990 device->io_width = btrfs_device_io_width(leaf, dev_item);
991 device->sector_size = btrfs_device_sector_size(leaf, dev_item);
993 ptr = (unsigned long)btrfs_device_uuid(dev_item);
994 read_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE);
999 static int read_one_dev(struct btrfs_root *root,
1000 struct extent_buffer *leaf,
1001 struct btrfs_dev_item *dev_item)
1003 struct btrfs_device *device;
1007 devid = btrfs_device_id(leaf, dev_item);
1008 device = btrfs_find_device(root, devid);
1010 printk("warning devid %llu not found already\n",
1011 (unsigned long long)devid);
1012 device = kmalloc(sizeof(*device), GFP_NOFS);
1015 device->total_ios = 0;
1016 list_add(&device->dev_list,
1017 &root->fs_info->fs_devices->devices);
1020 fill_device_from_item(leaf, dev_item, device);
1021 device->dev_root = root->fs_info->dev_root;
1025 int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf)
1027 struct btrfs_dev_item *dev_item;
1029 dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block,
1031 return read_one_dev(root, buf, dev_item);
1034 int btrfs_read_sys_array(struct btrfs_root *root)
1036 struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
1037 struct extent_buffer *sb = root->fs_info->sb_buffer;
1038 struct btrfs_disk_key *disk_key;
1039 struct btrfs_chunk *chunk;
1040 struct btrfs_key key;
1045 unsigned long sb_ptr;
1049 array_size = btrfs_super_sys_array_size(super_copy);
1052 * we do this loop twice, once for the device items and
1053 * once for all of the chunks. This way there are device
1054 * structs filled in for every chunk
1056 ptr = super_copy->sys_chunk_array;
1057 sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array);
1060 while (cur < array_size) {
1061 disk_key = (struct btrfs_disk_key *)ptr;
1062 btrfs_disk_key_to_cpu(&key, disk_key);
1064 len = sizeof(*disk_key);
1069 if (key.type == BTRFS_CHUNK_ITEM_KEY) {
1070 chunk = (struct btrfs_chunk *)sb_ptr;
1071 ret = read_one_chunk(root, &key, sb, chunk);
1073 num_stripes = btrfs_chunk_num_stripes(sb, chunk);
1074 len = btrfs_chunk_item_size(num_stripes);
1085 int btrfs_read_chunk_tree(struct btrfs_root *root)
1087 struct btrfs_path *path;
1088 struct extent_buffer *leaf;
1089 struct btrfs_key key;
1090 struct btrfs_key found_key;
1094 root = root->fs_info->chunk_root;
1096 path = btrfs_alloc_path();
1100 /* first we search for all of the device items, and then we
1101 * read in all of the chunk items. This way we can create chunk
1102 * mappings that reference all of the devices that are afound
1104 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
1108 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1110 leaf = path->nodes[0];
1111 slot = path->slots[0];
1112 if (slot >= btrfs_header_nritems(leaf)) {
1113 ret = btrfs_next_leaf(root, path);
1120 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1121 if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
1122 if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID)
1124 if (found_key.type == BTRFS_DEV_ITEM_KEY) {
1125 struct btrfs_dev_item *dev_item;
1126 dev_item = btrfs_item_ptr(leaf, slot,
1127 struct btrfs_dev_item);
1128 ret = read_one_dev(root, leaf, dev_item);
1131 } else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) {
1132 struct btrfs_chunk *chunk;
1133 chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
1134 ret = read_one_chunk(root, &found_key, leaf, chunk);
1138 if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
1140 btrfs_release_path(root, path);
1144 btrfs_free_path(path);