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.
20 #include "transaction.h"
21 #include "print-tree.h"
25 struct cache_extent ce;
26 struct btrfs_device *dev;
31 * this uses a pretty simple search, the expectation is that it is
32 * called very infrequently and that a given device has a small number
35 static int find_free_dev_extent(struct btrfs_trans_handle *trans,
36 struct btrfs_device *device,
37 struct btrfs_path *path,
38 u64 num_bytes, u64 *start)
41 struct btrfs_root *root = device->dev_root;
42 struct btrfs_dev_extent *dev_extent = NULL;
46 u64 search_end = device->total_bytes;
50 struct extent_buffer *l;
55 /* FIXME use last free of some kind */
57 key.objectid = device->devid;
58 key.offset = search_start;
59 key.type = BTRFS_DEV_EXTENT_KEY;
60 ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
63 ret = btrfs_previous_item(root, path, 0, key.type);
67 btrfs_item_key_to_cpu(l, &key, path->slots[0]);
70 slot = path->slots[0];
71 if (slot >= btrfs_header_nritems(l)) {
72 ret = btrfs_next_leaf(root, path);
79 if (search_start >= search_end) {
83 *start = search_start;
87 *start = last_byte > search_start ?
88 last_byte : search_start;
89 if (search_end <= *start) {
95 btrfs_item_key_to_cpu(l, &key, slot);
97 if (key.objectid < device->devid)
100 if (key.objectid > device->devid)
103 if (key.offset >= search_start && key.offset > last_byte &&
105 if (last_byte < search_start)
106 last_byte = search_start;
107 hole_size = key.offset - last_byte;
108 if (key.offset > last_byte &&
109 hole_size >= num_bytes) {
114 if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY) {
119 dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent);
120 last_byte = key.offset + btrfs_dev_extent_length(l, dev_extent);
126 /* we have to make sure we didn't find an extent that has already
127 * been allocated by the map tree or the original allocation
129 btrfs_release_path(root, path);
130 BUG_ON(*start < search_start);
132 if (*start + num_bytes > search_end) {
136 /* check for pending inserts here */
140 btrfs_release_path(root, path);
144 int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
145 struct btrfs_device *device,
146 u64 owner, u64 num_bytes, u64 *start)
149 struct btrfs_path *path;
150 struct btrfs_root *root = device->dev_root;
151 struct btrfs_dev_extent *extent;
152 struct extent_buffer *leaf;
153 struct btrfs_key key;
155 path = btrfs_alloc_path();
159 ret = find_free_dev_extent(trans, device, path, num_bytes, start);
164 key.objectid = device->devid;
166 key.type = BTRFS_DEV_EXTENT_KEY;
167 ret = btrfs_insert_empty_item(trans, root, path, &key,
171 leaf = path->nodes[0];
172 extent = btrfs_item_ptr(leaf, path->slots[0],
173 struct btrfs_dev_extent);
174 btrfs_set_dev_extent_owner(leaf, extent, owner);
175 btrfs_set_dev_extent_length(leaf, extent, num_bytes);
176 btrfs_mark_buffer_dirty(leaf);
178 btrfs_free_path(path);
182 static int find_next_chunk(struct btrfs_root *root, u64 *objectid)
184 struct btrfs_path *path;
186 struct btrfs_key key;
187 struct btrfs_key found_key;
189 path = btrfs_alloc_path();
192 key.objectid = (u64)-1;
193 key.offset = (u64)-1;
194 key.type = BTRFS_CHUNK_ITEM_KEY;
196 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
202 ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY);
206 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
208 *objectid = found_key.objectid + found_key.offset;
212 btrfs_free_path(path);
216 static int find_next_devid(struct btrfs_root *root, struct btrfs_path *path,
220 struct btrfs_key key;
221 struct btrfs_key found_key;
223 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
224 key.type = BTRFS_DEV_ITEM_KEY;
225 key.offset = (u64)-1;
227 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
233 ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID,
238 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
240 *objectid = found_key.offset + 1;
244 btrfs_release_path(root, path);
249 * the device information is stored in the chunk root
250 * the btrfs_device struct should be fully filled in
252 int btrfs_add_device(struct btrfs_trans_handle *trans,
253 struct btrfs_root *root,
254 struct btrfs_device *device)
257 struct btrfs_path *path;
258 struct btrfs_dev_item *dev_item;
259 struct extent_buffer *leaf;
260 struct btrfs_key key;
264 root = root->fs_info->chunk_root;
266 path = btrfs_alloc_path();
270 ret = find_next_devid(root, path, &free_devid);
274 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
275 key.type = BTRFS_DEV_ITEM_KEY;
276 key.offset = free_devid;
278 ret = btrfs_insert_empty_item(trans, root, path, &key,
283 leaf = path->nodes[0];
284 dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
286 device->devid = free_devid;
287 btrfs_set_device_id(leaf, dev_item, device->devid);
288 btrfs_set_device_type(leaf, dev_item, device->type);
289 btrfs_set_device_io_align(leaf, dev_item, device->io_align);
290 btrfs_set_device_io_width(leaf, dev_item, device->io_width);
291 btrfs_set_device_sector_size(leaf, dev_item, device->sector_size);
292 btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes);
293 btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used);
295 ptr = (unsigned long)btrfs_device_uuid(dev_item);
296 write_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE);
297 btrfs_mark_buffer_dirty(leaf);
301 btrfs_free_path(path);
304 int btrfs_update_device(struct btrfs_trans_handle *trans,
305 struct btrfs_device *device)
308 struct btrfs_path *path;
309 struct btrfs_root *root;
310 struct btrfs_dev_item *dev_item;
311 struct extent_buffer *leaf;
312 struct btrfs_key key;
314 root = device->dev_root->fs_info->chunk_root;
316 path = btrfs_alloc_path();
320 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
321 key.type = BTRFS_DEV_ITEM_KEY;
322 key.offset = device->devid;
324 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
333 leaf = path->nodes[0];
334 dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
336 btrfs_set_device_id(leaf, dev_item, device->devid);
337 btrfs_set_device_type(leaf, dev_item, device->type);
338 btrfs_set_device_io_align(leaf, dev_item, device->io_align);
339 btrfs_set_device_io_width(leaf, dev_item, device->io_width);
340 btrfs_set_device_sector_size(leaf, dev_item, device->sector_size);
341 btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes);
342 btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used);
343 btrfs_mark_buffer_dirty(leaf);
346 btrfs_free_path(path);
350 int btrfs_add_system_chunk(struct btrfs_trans_handle *trans,
351 struct btrfs_root *root,
352 struct btrfs_key *key,
353 struct btrfs_chunk *chunk, int item_size)
355 struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
356 struct btrfs_disk_key disk_key;
360 array_size = btrfs_super_sys_array_size(super_copy);
361 if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE)
364 ptr = super_copy->sys_chunk_array + array_size;
365 btrfs_cpu_key_to_disk(&disk_key, key);
366 memcpy(ptr, &disk_key, sizeof(disk_key));
367 ptr += sizeof(disk_key);
368 memcpy(ptr, chunk, item_size);
369 item_size += sizeof(disk_key);
370 btrfs_set_super_sys_array_size(super_copy, array_size + item_size);
374 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
375 struct btrfs_root *extent_root, u64 *start,
376 u64 *num_bytes, u64 type)
379 struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root;
380 struct btrfs_stripe *stripes;
381 struct btrfs_device *device = NULL;
382 struct btrfs_chunk *chunk;
383 struct list_head private_devs;
384 struct list_head *dev_list = &extent_root->fs_info->devices;
385 struct list_head *cur;
386 struct map_lookup *map;
388 u64 calc_size = 1024 * 1024 * 1024;
395 struct btrfs_key key;
397 if (list_empty(dev_list))
400 INIT_LIST_HEAD(&private_devs);
401 cur = dev_list->next;
403 /* build a private list of devices we will allocate from */
404 while(index < num_stripes) {
405 device = list_entry(cur, struct btrfs_device, dev_list);
406 avail = device->total_bytes - device->bytes_used;
408 if (avail > max_avail)
410 if (avail >= calc_size) {
411 list_move_tail(&device->dev_list, &private_devs);
417 if (index < num_stripes) {
418 list_splice(&private_devs, dev_list);
419 if (!looped && max_avail > 0) {
421 calc_size = max_avail;
427 ret = find_next_chunk(chunk_root, &key.objectid);
431 chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS);
435 stripes = &chunk->stripe;
437 *num_bytes = calc_size;
439 while(index < num_stripes) {
440 BUG_ON(list_empty(&private_devs));
441 cur = private_devs.next;
442 device = list_entry(cur, struct btrfs_device, dev_list);
443 list_move_tail(&device->dev_list, dev_list);
445 ret = btrfs_alloc_dev_extent(trans, device,
447 calc_size, &dev_offset);
450 device->bytes_used += calc_size;
451 ret = btrfs_update_device(trans, device);
454 btrfs_set_stack_stripe_devid(stripes + index, device->devid);
455 btrfs_set_stack_stripe_offset(stripes + index, dev_offset);
456 physical = dev_offset;
459 BUG_ON(!list_empty(&private_devs));
461 /* key.objectid was set above */
462 key.offset = *num_bytes;
463 key.type = BTRFS_CHUNK_ITEM_KEY;
464 btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid);
465 btrfs_set_stack_chunk_stripe_len(chunk, 64 * 1024);
466 btrfs_set_stack_chunk_type(chunk, type);
467 btrfs_set_stack_chunk_num_stripes(chunk, num_stripes);
468 btrfs_set_stack_chunk_io_align(chunk, extent_root->sectorsize);
469 btrfs_set_stack_chunk_io_width(chunk, extent_root->sectorsize);
470 btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize);
472 ret = btrfs_insert_item(trans, chunk_root, &key, chunk,
473 btrfs_chunk_item_size(num_stripes));
475 *start = key.objectid;
477 map = kmalloc(sizeof(*map), GFP_NOFS);
481 map->ce.start = key.objectid;
482 map->ce.size = key.offset;
484 map->physical = physical;
491 ret = insert_existing_cache_extent(
492 &extent_root->fs_info->mapping_tree.cache_tree,
500 void btrfs_mapping_init(struct btrfs_mapping_tree *tree)
502 cache_tree_init(&tree->cache_tree);
505 int btrfs_map_block(struct btrfs_mapping_tree *map_tree,
506 u64 logical, u64 *phys, u64 *length,
507 struct btrfs_device **dev)
509 struct cache_extent *ce;
510 struct map_lookup *map;
513 ce = find_first_cache_extent(&map_tree->cache_tree, logical);
515 BUG_ON(ce->start > logical || ce->start + ce->size < logical);
516 map = container_of(ce, struct map_lookup, ce);
517 offset = logical - ce->start;
518 *phys = map->physical + offset;
519 *length = ce->size - offset;
524 struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid)
526 struct btrfs_device *dev;
527 struct list_head *cur = root->fs_info->devices.next;
528 struct list_head *head = &root->fs_info->devices;
531 dev = list_entry(cur, struct btrfs_device, dev_list);
532 if (dev->devid == devid)
539 static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
540 struct extent_buffer *leaf,
541 struct btrfs_chunk *chunk)
543 struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree;
544 struct map_lookup *map;
545 struct cache_extent *ce;
551 logical = key->objectid;
552 length = key->offset;
553 ce = find_first_cache_extent(&map_tree->cache_tree, logical);
555 /* already mapped? */
556 if (ce && ce->start <= logical && ce->start + ce->size > logical) {
560 map = kmalloc(sizeof(*map), GFP_NOFS);
564 map->ce.start = logical;
565 map->ce.size = length;
567 map->physical = btrfs_stripe_offset_nr(leaf, chunk, 0);
568 devid = btrfs_stripe_devid_nr(leaf, chunk, 0);
569 map->dev = btrfs_find_device(root, devid);
575 ret = insert_existing_cache_extent(&map_tree->cache_tree, &map->ce);
581 static int fill_device_from_item(struct extent_buffer *leaf,
582 struct btrfs_dev_item *dev_item,
583 struct btrfs_device *device)
587 device->devid = btrfs_device_id(leaf, dev_item);
588 device->total_bytes = btrfs_device_total_bytes(leaf, dev_item);
589 device->bytes_used = btrfs_device_bytes_used(leaf, dev_item);
590 device->type = btrfs_device_type(leaf, dev_item);
591 device->io_align = btrfs_device_io_align(leaf, dev_item);
592 device->io_width = btrfs_device_io_width(leaf, dev_item);
593 device->sector_size = btrfs_device_sector_size(leaf, dev_item);
595 ptr = (unsigned long)btrfs_device_uuid(dev_item);
596 read_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE);
601 static int read_one_dev(struct btrfs_root *root,
602 struct extent_buffer *leaf,
603 struct btrfs_dev_item *dev_item)
605 struct btrfs_device *device;
609 devid = btrfs_device_id(leaf, dev_item);
610 device = btrfs_find_device(root, devid);
612 device = kmalloc(sizeof(*device), GFP_NOFS);
615 list_add(&device->dev_list, &root->fs_info->devices);
618 fill_device_from_item(leaf, dev_item, device);
619 device->dev_root = root->fs_info->dev_root;
620 device->fd = root->fs_info->fp;
621 ret = btrfs_open_device(device);
628 int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf)
630 struct btrfs_dev_item *dev_item;
632 dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block,
634 return read_one_dev(root, buf, dev_item);
637 int btrfs_read_sys_array(struct btrfs_root *root)
639 struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
640 struct extent_buffer *sb = root->fs_info->sb_buffer;
641 struct btrfs_disk_key *disk_key;
642 struct btrfs_chunk *chunk;
643 struct btrfs_key key;
648 unsigned long sb_ptr;
652 array_size = btrfs_super_sys_array_size(super_copy);
655 * we do this loop twice, once for the device items and
656 * once for all of the chunks. This way there are device
657 * structs filled in for every chunk
659 ptr = super_copy->sys_chunk_array;
660 sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array);
663 while (cur < array_size) {
664 disk_key = (struct btrfs_disk_key *)ptr;
665 btrfs_disk_key_to_cpu(&key, disk_key);
667 len = sizeof(*disk_key);
672 if (key.type == BTRFS_CHUNK_ITEM_KEY) {
673 chunk = (struct btrfs_chunk *)sb_ptr;
674 ret = read_one_chunk(root, &key, sb, chunk);
676 num_stripes = btrfs_chunk_num_stripes(sb, chunk);
677 len = btrfs_chunk_item_size(num_stripes);
688 int btrfs_read_chunk_tree(struct btrfs_root *root)
690 struct btrfs_path *path;
691 struct extent_buffer *leaf;
692 struct btrfs_key key;
693 struct btrfs_key found_key;
697 root = root->fs_info->chunk_root;
699 path = btrfs_alloc_path();
703 /* first we search for all of the device items, and then we
704 * read in all of the chunk items. This way we can create chunk
705 * mappings that reference all of the devices that are afound
707 key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
711 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
713 leaf = path->nodes[0];
714 slot = path->slots[0];
715 if (slot >= btrfs_header_nritems(leaf)) {
716 ret = btrfs_next_leaf(root, path);
723 btrfs_item_key_to_cpu(leaf, &found_key, slot);
724 if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
725 if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID)
727 if (found_key.type == BTRFS_DEV_ITEM_KEY) {
728 struct btrfs_dev_item *dev_item;
729 dev_item = btrfs_item_ptr(leaf, slot,
730 struct btrfs_dev_item);
731 ret = read_one_dev(root, leaf, dev_item);
734 } else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) {
735 struct btrfs_chunk *chunk;
736 chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
737 ret = read_one_chunk(root, &found_key, leaf, chunk);
741 if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
743 btrfs_release_path(root, path);
747 btrfs_free_path(path);