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 #include <linux/sched.h>
19 #include <linux/bio.h>
20 #include <linux/buffer_head.h>
21 #include <asm/div64.h>
23 #include "extent_map.h"
25 #include "transaction.h"
26 #include "print-tree.h"
30 struct btrfs_device
*dev
;
48 struct stripe stripes
[];
51 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
52 (sizeof(struct stripe) * (n)))
54 static DEFINE_MUTEX(uuid_mutex
);
55 static LIST_HEAD(fs_uuids
);
57 int btrfs_cleanup_fs_uuids(void)
59 struct btrfs_fs_devices
*fs_devices
;
60 struct list_head
*uuid_cur
;
61 struct list_head
*devices_cur
;
62 struct btrfs_device
*dev
;
64 list_for_each(uuid_cur
, &fs_uuids
) {
65 fs_devices
= list_entry(uuid_cur
, struct btrfs_fs_devices
,
67 while(!list_empty(&fs_devices
->devices
)) {
68 devices_cur
= fs_devices
->devices
.next
;
69 dev
= list_entry(devices_cur
, struct btrfs_device
,
71 printk("uuid cleanup finds %s\n", dev
->name
);
74 close_bdev_excl(dev
->bdev
);
76 list_del(&dev
->dev_list
);
83 static struct btrfs_device
*__find_device(struct list_head
*head
, u64 devid
)
85 struct btrfs_device
*dev
;
86 struct list_head
*cur
;
88 list_for_each(cur
, head
) {
89 dev
= list_entry(cur
, struct btrfs_device
, dev_list
);
90 if (dev
->devid
== devid
)
96 static struct btrfs_fs_devices
*find_fsid(u8
*fsid
)
98 struct list_head
*cur
;
99 struct btrfs_fs_devices
*fs_devices
;
101 list_for_each(cur
, &fs_uuids
) {
102 fs_devices
= list_entry(cur
, struct btrfs_fs_devices
, list
);
103 if (memcmp(fsid
, fs_devices
->fsid
, BTRFS_FSID_SIZE
) == 0)
109 static int device_list_add(const char *path
,
110 struct btrfs_super_block
*disk_super
,
111 u64 devid
, struct btrfs_fs_devices
**fs_devices_ret
)
113 struct btrfs_device
*device
;
114 struct btrfs_fs_devices
*fs_devices
;
115 u64 found_transid
= btrfs_super_generation(disk_super
);
117 fs_devices
= find_fsid(disk_super
->fsid
);
119 fs_devices
= kmalloc(sizeof(*fs_devices
), GFP_NOFS
);
122 INIT_LIST_HEAD(&fs_devices
->devices
);
123 list_add(&fs_devices
->list
, &fs_uuids
);
124 memcpy(fs_devices
->fsid
, disk_super
->fsid
, BTRFS_FSID_SIZE
);
125 fs_devices
->latest_devid
= devid
;
126 fs_devices
->latest_trans
= found_transid
;
127 fs_devices
->lowest_devid
= (u64
)-1;
128 fs_devices
->num_devices
= 0;
131 device
= __find_device(&fs_devices
->devices
, devid
);
134 device
= kzalloc(sizeof(*device
), GFP_NOFS
);
136 /* we can safely leave the fs_devices entry around */
139 device
->devid
= devid
;
140 device
->name
= kstrdup(path
, GFP_NOFS
);
145 list_add(&device
->dev_list
, &fs_devices
->devices
);
146 fs_devices
->num_devices
++;
149 if (found_transid
> fs_devices
->latest_trans
) {
150 fs_devices
->latest_devid
= devid
;
151 fs_devices
->latest_trans
= found_transid
;
153 if (fs_devices
->lowest_devid
> devid
) {
154 fs_devices
->lowest_devid
= devid
;
155 printk("lowest devid now %Lu\n", devid
);
157 *fs_devices_ret
= fs_devices
;
161 int btrfs_close_devices(struct btrfs_fs_devices
*fs_devices
)
163 struct list_head
*head
= &fs_devices
->devices
;
164 struct list_head
*cur
;
165 struct btrfs_device
*device
;
167 mutex_lock(&uuid_mutex
);
168 list_for_each(cur
, head
) {
169 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
171 close_bdev_excl(device
->bdev
);
172 printk("close devices closes %s\n", device
->name
);
176 mutex_unlock(&uuid_mutex
);
180 int btrfs_open_devices(struct btrfs_fs_devices
*fs_devices
,
181 int flags
, void *holder
)
183 struct block_device
*bdev
;
184 struct list_head
*head
= &fs_devices
->devices
;
185 struct list_head
*cur
;
186 struct btrfs_device
*device
;
189 mutex_lock(&uuid_mutex
);
190 list_for_each(cur
, head
) {
191 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
192 bdev
= open_bdev_excl(device
->name
, flags
, holder
);
193 printk("opening %s devid %Lu\n", device
->name
, device
->devid
);
195 printk("open %s failed\n", device
->name
);
199 if (device
->devid
== fs_devices
->latest_devid
)
200 fs_devices
->latest_bdev
= bdev
;
201 if (device
->devid
== fs_devices
->lowest_devid
) {
202 fs_devices
->lowest_bdev
= bdev
;
203 printk("lowest bdev %s\n", device
->name
);
207 mutex_unlock(&uuid_mutex
);
210 mutex_unlock(&uuid_mutex
);
211 btrfs_close_devices(fs_devices
);
215 int btrfs_scan_one_device(const char *path
, int flags
, void *holder
,
216 struct btrfs_fs_devices
**fs_devices_ret
)
218 struct btrfs_super_block
*disk_super
;
219 struct block_device
*bdev
;
220 struct buffer_head
*bh
;
224 mutex_lock(&uuid_mutex
);
226 printk("scan one opens %s\n", path
);
227 bdev
= open_bdev_excl(path
, flags
, holder
);
230 printk("open failed\n");
235 ret
= set_blocksize(bdev
, 4096);
238 bh
= __bread(bdev
, BTRFS_SUPER_INFO_OFFSET
/ 4096, 4096);
243 disk_super
= (struct btrfs_super_block
*)bh
->b_data
;
244 if (strncmp((char *)(&disk_super
->magic
), BTRFS_MAGIC
,
245 sizeof(disk_super
->magic
))) {
246 printk("no btrfs found on %s\n", path
);
250 devid
= le64_to_cpu(disk_super
->dev_item
.devid
);
251 printk("found device %Lu on %s\n", devid
, path
);
252 ret
= device_list_add(path
, disk_super
, devid
, fs_devices_ret
);
257 close_bdev_excl(bdev
);
258 printk("scan one closes bdev %s\n", path
);
260 mutex_unlock(&uuid_mutex
);
265 * this uses a pretty simple search, the expectation is that it is
266 * called very infrequently and that a given device has a small number
269 static int find_free_dev_extent(struct btrfs_trans_handle
*trans
,
270 struct btrfs_device
*device
,
271 struct btrfs_path
*path
,
272 u64 num_bytes
, u64
*start
)
274 struct btrfs_key key
;
275 struct btrfs_root
*root
= device
->dev_root
;
276 struct btrfs_dev_extent
*dev_extent
= NULL
;
279 u64 search_start
= 0;
280 u64 search_end
= device
->total_bytes
;
284 struct extent_buffer
*l
;
289 /* FIXME use last free of some kind */
291 /* we don't want to overwrite the superblock on the drive,
292 * so we make sure to start at an offset of at least 1MB
294 search_start
= max((u64
)1024 * 1024, search_start
);
295 key
.objectid
= device
->devid
;
296 key
.offset
= search_start
;
297 key
.type
= BTRFS_DEV_EXTENT_KEY
;
298 ret
= btrfs_search_slot(trans
, root
, &key
, path
, 0, 0);
301 ret
= btrfs_previous_item(root
, path
, 0, key
.type
);
305 btrfs_item_key_to_cpu(l
, &key
, path
->slots
[0]);
308 slot
= path
->slots
[0];
309 if (slot
>= btrfs_header_nritems(l
)) {
310 ret
= btrfs_next_leaf(root
, path
);
317 if (search_start
>= search_end
) {
321 *start
= search_start
;
325 *start
= last_byte
> search_start
?
326 last_byte
: search_start
;
327 if (search_end
<= *start
) {
333 btrfs_item_key_to_cpu(l
, &key
, slot
);
335 if (key
.objectid
< device
->devid
)
338 if (key
.objectid
> device
->devid
)
341 if (key
.offset
>= search_start
&& key
.offset
> last_byte
&&
343 if (last_byte
< search_start
)
344 last_byte
= search_start
;
345 hole_size
= key
.offset
- last_byte
;
346 if (key
.offset
> last_byte
&&
347 hole_size
>= num_bytes
) {
352 if (btrfs_key_type(&key
) != BTRFS_DEV_EXTENT_KEY
) {
357 dev_extent
= btrfs_item_ptr(l
, slot
, struct btrfs_dev_extent
);
358 last_byte
= key
.offset
+ btrfs_dev_extent_length(l
, dev_extent
);
364 /* we have to make sure we didn't find an extent that has already
365 * been allocated by the map tree or the original allocation
367 btrfs_release_path(root
, path
);
368 BUG_ON(*start
< search_start
);
370 if (*start
+ num_bytes
> search_end
) {
374 /* check for pending inserts here */
378 btrfs_release_path(root
, path
);
382 int btrfs_alloc_dev_extent(struct btrfs_trans_handle
*trans
,
383 struct btrfs_device
*device
,
384 u64 owner
, u64 num_bytes
, u64
*start
)
387 struct btrfs_path
*path
;
388 struct btrfs_root
*root
= device
->dev_root
;
389 struct btrfs_dev_extent
*extent
;
390 struct extent_buffer
*leaf
;
391 struct btrfs_key key
;
393 path
= btrfs_alloc_path();
397 ret
= find_free_dev_extent(trans
, device
, path
, num_bytes
, start
);
402 key
.objectid
= device
->devid
;
404 key
.type
= BTRFS_DEV_EXTENT_KEY
;
405 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
409 leaf
= path
->nodes
[0];
410 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
411 struct btrfs_dev_extent
);
412 btrfs_set_dev_extent_owner(leaf
, extent
, owner
);
413 btrfs_set_dev_extent_length(leaf
, extent
, num_bytes
);
414 btrfs_mark_buffer_dirty(leaf
);
416 btrfs_free_path(path
);
420 static int find_next_chunk(struct btrfs_root
*root
, u64
*objectid
)
422 struct btrfs_path
*path
;
424 struct btrfs_key key
;
425 struct btrfs_key found_key
;
427 path
= btrfs_alloc_path();
430 key
.objectid
= (u64
)-1;
431 key
.offset
= (u64
)-1;
432 key
.type
= BTRFS_CHUNK_ITEM_KEY
;
434 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
440 ret
= btrfs_previous_item(root
, path
, 0, BTRFS_CHUNK_ITEM_KEY
);
444 btrfs_item_key_to_cpu(path
->nodes
[0], &found_key
,
446 *objectid
= found_key
.objectid
+ found_key
.offset
;
450 btrfs_free_path(path
);
454 static int find_next_devid(struct btrfs_root
*root
, struct btrfs_path
*path
,
458 struct btrfs_key key
;
459 struct btrfs_key found_key
;
461 key
.objectid
= BTRFS_DEV_ITEMS_OBJECTID
;
462 key
.type
= BTRFS_DEV_ITEM_KEY
;
463 key
.offset
= (u64
)-1;
465 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
471 ret
= btrfs_previous_item(root
, path
, BTRFS_DEV_ITEMS_OBJECTID
,
476 btrfs_item_key_to_cpu(path
->nodes
[0], &found_key
,
478 *objectid
= found_key
.offset
+ 1;
482 btrfs_release_path(root
, path
);
487 * the device information is stored in the chunk root
488 * the btrfs_device struct should be fully filled in
490 int btrfs_add_device(struct btrfs_trans_handle
*trans
,
491 struct btrfs_root
*root
,
492 struct btrfs_device
*device
)
495 struct btrfs_path
*path
;
496 struct btrfs_dev_item
*dev_item
;
497 struct extent_buffer
*leaf
;
498 struct btrfs_key key
;
502 root
= root
->fs_info
->chunk_root
;
504 path
= btrfs_alloc_path();
508 ret
= find_next_devid(root
, path
, &free_devid
);
512 key
.objectid
= BTRFS_DEV_ITEMS_OBJECTID
;
513 key
.type
= BTRFS_DEV_ITEM_KEY
;
514 key
.offset
= free_devid
;
516 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
521 leaf
= path
->nodes
[0];
522 dev_item
= btrfs_item_ptr(leaf
, path
->slots
[0], struct btrfs_dev_item
);
524 device
->devid
= free_devid
;
525 btrfs_set_device_id(leaf
, dev_item
, device
->devid
);
526 btrfs_set_device_type(leaf
, dev_item
, device
->type
);
527 btrfs_set_device_io_align(leaf
, dev_item
, device
->io_align
);
528 btrfs_set_device_io_width(leaf
, dev_item
, device
->io_width
);
529 btrfs_set_device_sector_size(leaf
, dev_item
, device
->sector_size
);
530 btrfs_set_device_total_bytes(leaf
, dev_item
, device
->total_bytes
);
531 btrfs_set_device_bytes_used(leaf
, dev_item
, device
->bytes_used
);
533 ptr
= (unsigned long)btrfs_device_uuid(dev_item
);
534 write_extent_buffer(leaf
, device
->uuid
, ptr
, BTRFS_DEV_UUID_SIZE
);
535 btrfs_mark_buffer_dirty(leaf
);
539 btrfs_free_path(path
);
542 int btrfs_update_device(struct btrfs_trans_handle
*trans
,
543 struct btrfs_device
*device
)
546 struct btrfs_path
*path
;
547 struct btrfs_root
*root
;
548 struct btrfs_dev_item
*dev_item
;
549 struct extent_buffer
*leaf
;
550 struct btrfs_key key
;
552 root
= device
->dev_root
->fs_info
->chunk_root
;
554 path
= btrfs_alloc_path();
558 key
.objectid
= BTRFS_DEV_ITEMS_OBJECTID
;
559 key
.type
= BTRFS_DEV_ITEM_KEY
;
560 key
.offset
= device
->devid
;
562 ret
= btrfs_search_slot(trans
, root
, &key
, path
, 0, 1);
571 leaf
= path
->nodes
[0];
572 dev_item
= btrfs_item_ptr(leaf
, path
->slots
[0], struct btrfs_dev_item
);
574 btrfs_set_device_id(leaf
, dev_item
, device
->devid
);
575 btrfs_set_device_type(leaf
, dev_item
, device
->type
);
576 btrfs_set_device_io_align(leaf
, dev_item
, device
->io_align
);
577 btrfs_set_device_io_width(leaf
, dev_item
, device
->io_width
);
578 btrfs_set_device_sector_size(leaf
, dev_item
, device
->sector_size
);
579 btrfs_set_device_total_bytes(leaf
, dev_item
, device
->total_bytes
);
580 btrfs_set_device_bytes_used(leaf
, dev_item
, device
->bytes_used
);
581 btrfs_mark_buffer_dirty(leaf
);
584 btrfs_free_path(path
);
588 int btrfs_add_system_chunk(struct btrfs_trans_handle
*trans
,
589 struct btrfs_root
*root
,
590 struct btrfs_key
*key
,
591 struct btrfs_chunk
*chunk
, int item_size
)
593 struct btrfs_super_block
*super_copy
= &root
->fs_info
->super_copy
;
594 struct btrfs_disk_key disk_key
;
598 array_size
= btrfs_super_sys_array_size(super_copy
);
599 if (array_size
+ item_size
> BTRFS_SYSTEM_CHUNK_ARRAY_SIZE
)
602 ptr
= super_copy
->sys_chunk_array
+ array_size
;
603 btrfs_cpu_key_to_disk(&disk_key
, key
);
604 memcpy(ptr
, &disk_key
, sizeof(disk_key
));
605 ptr
+= sizeof(disk_key
);
606 memcpy(ptr
, chunk
, item_size
);
607 item_size
+= sizeof(disk_key
);
608 btrfs_set_super_sys_array_size(super_copy
, array_size
+ item_size
);
612 int btrfs_alloc_chunk(struct btrfs_trans_handle
*trans
,
613 struct btrfs_root
*extent_root
, u64
*start
,
614 u64
*num_bytes
, u64 type
)
617 struct btrfs_fs_info
*info
= extent_root
->fs_info
;
618 struct btrfs_root
*chunk_root
= extent_root
->fs_info
->chunk_root
;
619 struct btrfs_stripe
*stripes
;
620 struct btrfs_device
*device
= NULL
;
621 struct btrfs_chunk
*chunk
;
622 struct list_head private_devs
;
623 struct list_head
*dev_list
= &extent_root
->fs_info
->fs_devices
->devices
;
624 struct list_head
*cur
;
625 struct extent_map_tree
*em_tree
;
626 struct map_lookup
*map
;
627 struct extent_map
*em
;
629 u64 calc_size
= 1024 * 1024 * 1024;
630 u64 min_free
= calc_size
;
637 int stripe_len
= 64 * 1024;
638 struct btrfs_key key
;
640 if (list_empty(dev_list
))
643 if (type
& (BTRFS_BLOCK_GROUP_RAID0
))
644 num_stripes
= btrfs_super_num_devices(&info
->super_copy
);
645 if (type
& (BTRFS_BLOCK_GROUP_DUP
))
647 if (type
& (BTRFS_BLOCK_GROUP_RAID1
)) {
648 num_stripes
= min_t(u64
, 2,
649 btrfs_super_num_devices(&info
->super_copy
));
652 INIT_LIST_HEAD(&private_devs
);
653 cur
= dev_list
->next
;
656 if (type
& BTRFS_BLOCK_GROUP_DUP
)
657 min_free
= calc_size
* 2;
659 /* build a private list of devices we will allocate from */
660 while(index
< num_stripes
) {
661 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
663 avail
= device
->total_bytes
- device
->bytes_used
;
665 if (avail
> max_avail
)
667 if (avail
>= min_free
) {
668 list_move_tail(&device
->dev_list
, &private_devs
);
670 if (type
& BTRFS_BLOCK_GROUP_DUP
)
676 if (index
< num_stripes
) {
677 list_splice(&private_devs
, dev_list
);
678 if (!looped
&& max_avail
> 0) {
680 calc_size
= max_avail
;
686 ret
= find_next_chunk(chunk_root
, &key
.objectid
);
690 chunk
= kmalloc(btrfs_chunk_item_size(num_stripes
), GFP_NOFS
);
694 map
= kmalloc(map_lookup_size(num_stripes
), GFP_NOFS
);
700 stripes
= &chunk
->stripe
;
702 if (type
& (BTRFS_BLOCK_GROUP_RAID1
| BTRFS_BLOCK_GROUP_DUP
))
703 *num_bytes
= calc_size
;
705 *num_bytes
= calc_size
* num_stripes
;
708 printk("new chunk type %Lu start %Lu size %Lu\n", type
, key
.objectid
, *num_bytes
);
709 while(index
< num_stripes
) {
710 BUG_ON(list_empty(&private_devs
));
711 cur
= private_devs
.next
;
712 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
714 /* loop over this device again if we're doing a dup group */
715 if (!(type
& BTRFS_BLOCK_GROUP_DUP
) ||
716 (index
== num_stripes
- 1))
717 list_move_tail(&device
->dev_list
, dev_list
);
719 ret
= btrfs_alloc_dev_extent(trans
, device
,
721 calc_size
, &dev_offset
);
723 printk("alloc chunk start %Lu size %Lu from dev %Lu type %Lu\n", key
.objectid
, calc_size
, device
->devid
, type
);
724 device
->bytes_used
+= calc_size
;
725 ret
= btrfs_update_device(trans
, device
);
728 map
->stripes
[index
].dev
= device
;
729 map
->stripes
[index
].physical
= dev_offset
;
730 btrfs_set_stack_stripe_devid(stripes
+ index
, device
->devid
);
731 btrfs_set_stack_stripe_offset(stripes
+ index
, dev_offset
);
732 physical
= dev_offset
;
735 BUG_ON(!list_empty(&private_devs
));
737 /* key.objectid was set above */
738 key
.offset
= *num_bytes
;
739 key
.type
= BTRFS_CHUNK_ITEM_KEY
;
740 btrfs_set_stack_chunk_owner(chunk
, extent_root
->root_key
.objectid
);
741 btrfs_set_stack_chunk_stripe_len(chunk
, stripe_len
);
742 btrfs_set_stack_chunk_type(chunk
, type
);
743 btrfs_set_stack_chunk_num_stripes(chunk
, num_stripes
);
744 btrfs_set_stack_chunk_io_align(chunk
, stripe_len
);
745 btrfs_set_stack_chunk_io_width(chunk
, stripe_len
);
746 btrfs_set_stack_chunk_sector_size(chunk
, extent_root
->sectorsize
);
747 map
->sector_size
= extent_root
->sectorsize
;
748 map
->stripe_len
= stripe_len
;
749 map
->io_align
= stripe_len
;
750 map
->io_width
= stripe_len
;
752 map
->num_stripes
= num_stripes
;
754 ret
= btrfs_insert_item(trans
, chunk_root
, &key
, chunk
,
755 btrfs_chunk_item_size(num_stripes
));
757 *start
= key
.objectid
;
759 em
= alloc_extent_map(GFP_NOFS
);
762 em
->bdev
= (struct block_device
*)map
;
763 em
->start
= key
.objectid
;
764 em
->len
= key
.offset
;
769 em_tree
= &extent_root
->fs_info
->mapping_tree
.map_tree
;
770 spin_lock(&em_tree
->lock
);
771 ret
= add_extent_mapping(em_tree
, em
);
773 spin_unlock(&em_tree
->lock
);
778 void btrfs_mapping_init(struct btrfs_mapping_tree
*tree
)
780 extent_map_tree_init(&tree
->map_tree
, GFP_NOFS
);
783 void btrfs_mapping_tree_free(struct btrfs_mapping_tree
*tree
)
785 struct extent_map
*em
;
788 spin_lock(&tree
->map_tree
.lock
);
789 em
= lookup_extent_mapping(&tree
->map_tree
, 0, (u64
)-1);
791 remove_extent_mapping(&tree
->map_tree
, em
);
792 spin_unlock(&tree
->map_tree
.lock
);
798 /* once for the tree */
803 int btrfs_map_block(struct btrfs_mapping_tree
*map_tree
, int rw
,
804 int dev_nr
, u64 logical
, u64
*phys
, u64
*length
,
805 struct btrfs_device
**dev
, int *total_devs
)
807 struct extent_map
*em
;
808 struct map_lookup
*map
;
809 struct extent_map_tree
*em_tree
= &map_tree
->map_tree
;
816 spin_lock(&em_tree
->lock
);
817 em
= lookup_extent_mapping(em_tree
, logical
, *length
);
820 BUG_ON(em
->start
> logical
|| em
->start
+ em
->len
< logical
);
821 map
= (struct map_lookup
*)em
->bdev
;
822 offset
= logical
- em
->start
;
826 * stripe_nr counts the total number of stripes we have to stride
827 * to get to this block
829 do_div(stripe_nr
, map
->stripe_len
);
831 stripe_offset
= stripe_nr
* map
->stripe_len
;
832 BUG_ON(offset
< stripe_offset
);
834 /* stripe_offset is the offset of this block in its stripe*/
835 stripe_offset
= offset
- stripe_offset
;
837 if (map
->type
& BTRFS_BLOCK_GROUP_RAID1
) {
838 stripe_index
= dev_nr
;
839 if (rw
& (1 << BIO_RW
))
840 *total_devs
= map
->num_stripes
;
844 struct btrfs_device
*cur
;
846 for (i
= 0; i
< map
->num_stripes
; i
++) {
847 cur
= map
->stripes
[i
].dev
;
848 spin_lock(&cur
->io_lock
);
849 if (cur
->total_ios
< least
) {
850 least
= cur
->total_ios
;
853 spin_unlock(&cur
->io_lock
);
857 } else if (map
->type
& BTRFS_BLOCK_GROUP_DUP
) {
859 *total_devs
= map
->num_stripes
;
860 stripe_index
= dev_nr
;
867 * after this do_div call, stripe_nr is the number of stripes
868 * on this device we have to walk to find the data, and
869 * stripe_index is the number of our device in the stripe array
871 stripe_index
= do_div(stripe_nr
, map
->num_stripes
);
873 BUG_ON(stripe_index
>= map
->num_stripes
);
874 *phys
= map
->stripes
[stripe_index
].physical
+ stripe_offset
+
875 stripe_nr
* map
->stripe_len
;
877 if (map
->type
& (BTRFS_BLOCK_GROUP_RAID0
| BTRFS_BLOCK_GROUP_RAID1
|
878 BTRFS_BLOCK_GROUP_DUP
)) {
879 /* we limit the length of each bio to what fits in a stripe */
880 *length
= min_t(u64
, em
->len
- offset
,
881 map
->stripe_len
- stripe_offset
);
883 *length
= em
->len
- offset
;
885 *dev
= map
->stripes
[stripe_index
].dev
;
887 spin_unlock(&em_tree
->lock
);
891 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
892 static void end_bio_multi_stripe(struct bio
*bio
, int err
)
894 static int end_bio_multi_stripe(struct bio
*bio
,
895 unsigned int bytes_done
, int err
)
898 struct multi_bio
*multi
= bio
->bi_private
;
900 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
907 if (atomic_dec_and_test(&multi
->stripes
)) {
908 bio
->bi_private
= multi
->private;
909 bio
->bi_end_io
= multi
->end_io
;
911 if (!err
&& multi
->error
)
919 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
924 int btrfs_map_bio(struct btrfs_root
*root
, int rw
, struct bio
*bio
)
926 struct btrfs_mapping_tree
*map_tree
;
927 struct btrfs_device
*dev
;
928 struct bio
*first_bio
= bio
;
929 u64 logical
= bio
->bi_sector
<< 9;
933 struct bio_vec
*bvec
;
934 struct multi_bio
*multi
= NULL
;
940 bio_for_each_segment(bvec
, bio
, i
) {
941 length
+= bvec
->bv_len
;
944 map_tree
= &root
->fs_info
->mapping_tree
;
946 while(dev_nr
< total_devs
) {
947 ret
= btrfs_map_block(map_tree
, rw
, dev_nr
, logical
,
948 &physical
, &map_length
, &dev
,
950 if (map_length
< length
) {
951 printk("mapping failed logical %Lu bio len %Lu physical %Lu "
952 "len %Lu\n", logical
, length
, physical
, map_length
);
955 BUG_ON(map_length
< length
);
956 if (total_devs
> 1) {
958 multi
= kmalloc(sizeof(*multi
), GFP_NOFS
);
959 atomic_set(&multi
->stripes
, 1);
960 multi
->end_io
= bio
->bi_end_io
;
961 multi
->private = first_bio
->bi_private
;
964 atomic_inc(&multi
->stripes
);
966 if (dev_nr
< total_devs
- 1) {
967 bio
= bio_clone(first_bio
, GFP_NOFS
);
972 bio
->bi_private
= multi
;
973 bio
->bi_end_io
= end_bio_multi_stripe
;
975 bio
->bi_sector
= physical
>> 9;
976 bio
->bi_bdev
= dev
->bdev
;
977 spin_lock(&dev
->io_lock
);
979 spin_unlock(&dev
->io_lock
);
986 struct btrfs_device
*btrfs_find_device(struct btrfs_root
*root
, u64 devid
)
988 struct list_head
*head
= &root
->fs_info
->fs_devices
->devices
;
990 return __find_device(head
, devid
);
993 static int read_one_chunk(struct btrfs_root
*root
, struct btrfs_key
*key
,
994 struct extent_buffer
*leaf
,
995 struct btrfs_chunk
*chunk
)
997 struct btrfs_mapping_tree
*map_tree
= &root
->fs_info
->mapping_tree
;
998 struct map_lookup
*map
;
999 struct extent_map
*em
;
1007 logical
= key
->objectid
;
1008 length
= key
->offset
;
1009 spin_lock(&map_tree
->map_tree
.lock
);
1010 em
= lookup_extent_mapping(&map_tree
->map_tree
, logical
, 1);
1012 /* already mapped? */
1013 if (em
&& em
->start
<= logical
&& em
->start
+ em
->len
> logical
) {
1014 free_extent_map(em
);
1015 spin_unlock(&map_tree
->map_tree
.lock
);
1018 free_extent_map(em
);
1020 spin_unlock(&map_tree
->map_tree
.lock
);
1022 map
= kzalloc(sizeof(*map
), GFP_NOFS
);
1026 em
= alloc_extent_map(GFP_NOFS
);
1029 num_stripes
= btrfs_chunk_num_stripes(leaf
, chunk
);
1030 map
= kmalloc(map_lookup_size(num_stripes
), GFP_NOFS
);
1032 free_extent_map(em
);
1036 em
->bdev
= (struct block_device
*)map
;
1037 em
->start
= logical
;
1039 em
->block_start
= 0;
1041 map
->num_stripes
= num_stripes
;
1042 map
->io_width
= btrfs_chunk_io_width(leaf
, chunk
);
1043 map
->io_align
= btrfs_chunk_io_align(leaf
, chunk
);
1044 map
->sector_size
= btrfs_chunk_sector_size(leaf
, chunk
);
1045 map
->stripe_len
= btrfs_chunk_stripe_len(leaf
, chunk
);
1046 map
->type
= btrfs_chunk_type(leaf
, chunk
);
1047 for (i
= 0; i
< num_stripes
; i
++) {
1048 map
->stripes
[i
].physical
=
1049 btrfs_stripe_offset_nr(leaf
, chunk
, i
);
1050 devid
= btrfs_stripe_devid_nr(leaf
, chunk
, i
);
1051 map
->stripes
[i
].dev
= btrfs_find_device(root
, devid
);
1052 if (!map
->stripes
[i
].dev
) {
1054 free_extent_map(em
);
1059 spin_lock(&map_tree
->map_tree
.lock
);
1060 ret
= add_extent_mapping(&map_tree
->map_tree
, em
);
1062 spin_unlock(&map_tree
->map_tree
.lock
);
1063 free_extent_map(em
);
1068 static int fill_device_from_item(struct extent_buffer
*leaf
,
1069 struct btrfs_dev_item
*dev_item
,
1070 struct btrfs_device
*device
)
1074 device
->devid
= btrfs_device_id(leaf
, dev_item
);
1075 device
->total_bytes
= btrfs_device_total_bytes(leaf
, dev_item
);
1076 device
->bytes_used
= btrfs_device_bytes_used(leaf
, dev_item
);
1077 device
->type
= btrfs_device_type(leaf
, dev_item
);
1078 device
->io_align
= btrfs_device_io_align(leaf
, dev_item
);
1079 device
->io_width
= btrfs_device_io_width(leaf
, dev_item
);
1080 device
->sector_size
= btrfs_device_sector_size(leaf
, dev_item
);
1082 ptr
= (unsigned long)btrfs_device_uuid(dev_item
);
1083 read_extent_buffer(leaf
, device
->uuid
, ptr
, BTRFS_DEV_UUID_SIZE
);
1088 static int read_one_dev(struct btrfs_root
*root
,
1089 struct extent_buffer
*leaf
,
1090 struct btrfs_dev_item
*dev_item
)
1092 struct btrfs_device
*device
;
1096 devid
= btrfs_device_id(leaf
, dev_item
);
1097 device
= btrfs_find_device(root
, devid
);
1099 printk("warning devid %Lu not found already\n", devid
);
1100 device
= kmalloc(sizeof(*device
), GFP_NOFS
);
1103 list_add(&device
->dev_list
,
1104 &root
->fs_info
->fs_devices
->devices
);
1105 device
->total_ios
= 0;
1106 spin_lock_init(&device
->io_lock
);
1109 fill_device_from_item(leaf
, dev_item
, device
);
1110 device
->dev_root
= root
->fs_info
->dev_root
;
1113 ret
= btrfs_open_device(device
);
1121 int btrfs_read_super_device(struct btrfs_root
*root
, struct extent_buffer
*buf
)
1123 struct btrfs_dev_item
*dev_item
;
1125 dev_item
= (struct btrfs_dev_item
*)offsetof(struct btrfs_super_block
,
1127 return read_one_dev(root
, buf
, dev_item
);
1130 int btrfs_read_sys_array(struct btrfs_root
*root
)
1132 struct btrfs_super_block
*super_copy
= &root
->fs_info
->super_copy
;
1133 struct extent_buffer
*sb
= root
->fs_info
->sb_buffer
;
1134 struct btrfs_disk_key
*disk_key
;
1135 struct btrfs_chunk
*chunk
;
1136 struct btrfs_key key
;
1141 unsigned long sb_ptr
;
1145 array_size
= btrfs_super_sys_array_size(super_copy
);
1148 * we do this loop twice, once for the device items and
1149 * once for all of the chunks. This way there are device
1150 * structs filled in for every chunk
1152 ptr
= super_copy
->sys_chunk_array
;
1153 sb_ptr
= offsetof(struct btrfs_super_block
, sys_chunk_array
);
1156 while (cur
< array_size
) {
1157 disk_key
= (struct btrfs_disk_key
*)ptr
;
1158 btrfs_disk_key_to_cpu(&key
, disk_key
);
1160 len
= sizeof(*disk_key
);
1165 if (key
.type
== BTRFS_CHUNK_ITEM_KEY
) {
1166 chunk
= (struct btrfs_chunk
*)sb_ptr
;
1167 ret
= read_one_chunk(root
, &key
, sb
, chunk
);
1169 num_stripes
= btrfs_chunk_num_stripes(sb
, chunk
);
1170 len
= btrfs_chunk_item_size(num_stripes
);
1181 int btrfs_read_chunk_tree(struct btrfs_root
*root
)
1183 struct btrfs_path
*path
;
1184 struct extent_buffer
*leaf
;
1185 struct btrfs_key key
;
1186 struct btrfs_key found_key
;
1190 root
= root
->fs_info
->chunk_root
;
1192 path
= btrfs_alloc_path();
1196 /* first we search for all of the device items, and then we
1197 * read in all of the chunk items. This way we can create chunk
1198 * mappings that reference all of the devices that are afound
1200 key
.objectid
= BTRFS_DEV_ITEMS_OBJECTID
;
1204 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1206 leaf
= path
->nodes
[0];
1207 slot
= path
->slots
[0];
1208 if (slot
>= btrfs_header_nritems(leaf
)) {
1209 ret
= btrfs_next_leaf(root
, path
);
1216 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1217 if (key
.objectid
== BTRFS_DEV_ITEMS_OBJECTID
) {
1218 if (found_key
.objectid
!= BTRFS_DEV_ITEMS_OBJECTID
)
1220 if (found_key
.type
== BTRFS_DEV_ITEM_KEY
) {
1221 struct btrfs_dev_item
*dev_item
;
1222 dev_item
= btrfs_item_ptr(leaf
, slot
,
1223 struct btrfs_dev_item
);
1224 ret
= read_one_dev(root
, leaf
, dev_item
);
1227 } else if (found_key
.type
== BTRFS_CHUNK_ITEM_KEY
) {
1228 struct btrfs_chunk
*chunk
;
1229 chunk
= btrfs_item_ptr(leaf
, slot
, struct btrfs_chunk
);
1230 ret
= read_one_chunk(root
, &found_key
, leaf
, chunk
);
1234 if (key
.objectid
== BTRFS_DEV_ITEMS_OBJECTID
) {
1236 btrfs_release_path(root
, path
);
1240 btrfs_free_path(path
);