Btrfs: Add RAID10 support
Signed-off-by: Chris Mason <chris.mason@oracle.com>
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 82d67c3..a22edcf 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -197,6 +197,9 @@
* item in the btree
*/
__le16 num_stripes;
+
+ /* sub stripes only matter for raid10 */
+ __le16 sub_stripes;
struct btrfs_stripe stripe;
/* additional stripes go here */
} __attribute__ ((__packed__));
@@ -444,6 +447,7 @@
#define BTRFS_BLOCK_GROUP_RAID0 (1 << 3)
#define BTRFS_BLOCK_GROUP_RAID1 (1 << 4)
#define BTRFS_BLOCK_GROUP_DUP (1 << 5)
+#define BTRFS_BLOCK_GROUP_RAID10 (1 << 6)
struct btrfs_block_group_item {
@@ -757,6 +761,7 @@
BTRFS_SETGET_FUNCS(chunk_sector_size, struct btrfs_chunk, sector_size, 32);
BTRFS_SETGET_FUNCS(chunk_type, struct btrfs_chunk, type, 64);
BTRFS_SETGET_FUNCS(chunk_num_stripes, struct btrfs_chunk, num_stripes, 16);
+BTRFS_SETGET_FUNCS(chunk_sub_stripes, struct btrfs_chunk, sub_stripes, 16);
BTRFS_SETGET_FUNCS(stripe_devid, struct btrfs_stripe, devid, 64);
BTRFS_SETGET_FUNCS(stripe_offset, struct btrfs_stripe, offset, 64);
@@ -778,6 +783,8 @@
BTRFS_SETGET_STACK_FUNCS(stack_chunk_type, struct btrfs_chunk, type, 64);
BTRFS_SETGET_STACK_FUNCS(stack_chunk_num_stripes, struct btrfs_chunk,
num_stripes, 16);
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_sub_stripes, struct btrfs_chunk,
+ sub_stripes, 16);
BTRFS_SETGET_STACK_FUNCS(stack_stripe_devid, struct btrfs_stripe, devid, 64);
BTRFS_SETGET_STACK_FUNCS(stack_stripe_offset, struct btrfs_stripe, offset, 64);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 71f045c..4e5bd62 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1042,6 +1042,7 @@
{
u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
BTRFS_BLOCK_GROUP_RAID1 |
+ BTRFS_BLOCK_GROUP_RAID10 |
BTRFS_BLOCK_GROUP_DUP);
if (extra_flags) {
if (flags & BTRFS_BLOCK_GROUP_DATA)
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 23ebd95..e6417a5 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -33,6 +33,7 @@
int stripe_len;
int sector_size;
int num_stripes;
+ int sub_stripes;
struct btrfs_bio_stripe stripes[];
};
@@ -641,6 +642,7 @@
u64 avail;
u64 max_avail = 0;
int num_stripes = 1;
+ int sub_stripes = 0;
int looped = 0;
int ret;
int index;
@@ -658,6 +660,13 @@
num_stripes = min_t(u64, 2,
btrfs_super_num_devices(&info->super_copy));
}
+ if (type & (BTRFS_BLOCK_GROUP_RAID10)) {
+ num_stripes = btrfs_super_num_devices(&info->super_copy);
+ if (num_stripes < 4)
+ return -ENOSPC;
+ num_stripes &= ~(u32)1;
+ sub_stripes = 2;
+ }
again:
INIT_LIST_HEAD(&private_devs);
cur = dev_list->next;
@@ -714,6 +723,8 @@
if (type & (BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_DUP))
*num_bytes = calc_size;
+ else if (type & BTRFS_BLOCK_GROUP_RAID10)
+ *num_bytes = calc_size * num_stripes / sub_stripes;
else
*num_bytes = calc_size * num_stripes;
@@ -760,12 +771,14 @@
btrfs_set_stack_chunk_io_align(chunk, stripe_len);
btrfs_set_stack_chunk_io_width(chunk, stripe_len);
btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize);
+ btrfs_set_stack_chunk_sub_stripes(chunk, sub_stripes);
map->sector_size = extent_root->sectorsize;
map->stripe_len = stripe_len;
map->io_align = stripe_len;
map->io_width = stripe_len;
map->type = type;
map->num_stripes = num_stripes;
+ map->sub_stripes = sub_stripes;
ret = btrfs_insert_item(trans, chunk_root, &key, chunk,
btrfs_chunk_item_size(num_stripes));
@@ -832,6 +845,8 @@
map = (struct map_lookup *)em->bdev;
if (map->type & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1))
ret = map->num_stripes;
+ else if (map->type & BTRFS_BLOCK_GROUP_RAID10)
+ ret = map->sub_stripes;
else
ret = 1;
free_extent_map(em);
@@ -849,6 +864,7 @@
u64 stripe_offset;
u64 stripe_nr;
int stripes_allocated = 8;
+ int stripes_required = 1;
int stripe_index;
int i;
struct btrfs_multi_bio *multi = NULL;
@@ -877,10 +893,16 @@
mirror_num = 0;
/* if our multi bio struct is too small, back off and try again */
- if (multi_ret && (rw & (1 << BIO_RW)) &&
- stripes_allocated < map->num_stripes &&
- ((map->type & BTRFS_BLOCK_GROUP_RAID1) ||
- (map->type & BTRFS_BLOCK_GROUP_DUP))) {
+ if (rw & (1 << BIO_RW)) {
+ if (map->type & (BTRFS_BLOCK_GROUP_RAID1 |
+ BTRFS_BLOCK_GROUP_DUP)) {
+ stripes_required = map->num_stripes;
+ } else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
+ stripes_required = map->sub_stripes;
+ }
+ }
+ if (multi_ret && rw == WRITE &&
+ stripes_allocated < stripes_required) {
stripes_allocated = map->num_stripes;
free_extent_map(em);
kfree(multi);
@@ -900,6 +922,7 @@
stripe_offset = offset - stripe_offset;
if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
+ BTRFS_BLOCK_GROUP_RAID10 |
BTRFS_BLOCK_GROUP_DUP)) {
/* we limit the length of each bio to what fits in a stripe */
*length = min_t(u64, em->len - offset,
@@ -937,6 +960,19 @@
multi->num_stripes = map->num_stripes;
else if (mirror_num)
stripe_index = mirror_num - 1;
+ } else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
+ int factor = map->num_stripes / map->sub_stripes;
+ int orig_stripe_nr = stripe_nr;
+
+ stripe_index = do_div(stripe_nr, factor);
+ stripe_index *= map->sub_stripes;
+
+ if (rw & (1 << BIO_RW))
+ multi->num_stripes = map->sub_stripes;
+ else if (mirror_num)
+ stripe_index += mirror_num - 1;
+ else
+ stripe_index += orig_stripe_nr % map->sub_stripes;
} else {
/*
* after this do_div call, stripe_nr is the number of stripes
@@ -946,7 +982,6 @@
stripe_index = do_div(stripe_nr, map->num_stripes);
}
BUG_ON(stripe_index >= map->num_stripes);
- BUG_ON(stripe_index != 0 && multi->num_stripes > 1);
for (i = 0; i < multi->num_stripes; i++) {
multi->stripes[i].physical =
@@ -1120,6 +1155,7 @@
map->sector_size = btrfs_chunk_sector_size(leaf, chunk);
map->stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
map->type = btrfs_chunk_type(leaf, chunk);
+ map->sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk);
for (i = 0; i < num_stripes; i++) {
map->stripes[i].physical =
btrfs_stripe_offset_nr(leaf, chunk, i);