Btrfs: move read only block groups onto their own list V2
Our gluster boxes were spending lots of time in statfs because our fs'es are
huge. The problem is statfs loops through all of the block groups looking for
read only block groups, and when you have several terabytes worth of data that
ends up being a lot of block groups. Move the read only block groups onto a
read only list and only proces that list in
btrfs_account_ro_block_groups_free_space to reduce the amount of churn. Thanks,
Signed-off-by: Josef Bacik <jbacik@fb.com>
Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: Chris Mason <clm@fb.com>
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 47c1ba1..cf3460e 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -3504,6 +3504,7 @@
found->chunk_alloc = 0;
found->flush = 0;
init_waitqueue_head(&found->wait);
+ INIT_LIST_HEAD(&found->ro_bgs);
ret = kobject_init_and_add(&found->kobj, &space_info_ktype,
info->space_info_kobj, "%s",
@@ -8511,6 +8512,7 @@
min_allocable_bytes <= sinfo->total_bytes) {
sinfo->bytes_readonly += num_bytes;
cache->ro = 1;
+ list_add_tail(&cache->ro_list, &sinfo->ro_bgs);
ret = 0;
}
out:
@@ -8565,15 +8567,20 @@
/*
* helper to account the unused space of all the readonly block group in the
- * list. takes mirrors into account.
+ * space_info. takes mirrors into account.
*/
-static u64 __btrfs_get_ro_block_group_free_space(struct list_head *groups_list)
+u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo)
{
struct btrfs_block_group_cache *block_group;
u64 free_bytes = 0;
int factor;
- list_for_each_entry(block_group, groups_list, list) {
+ /* It's df, we don't care if it's racey */
+ if (list_empty(&sinfo->ro_bgs))
+ return 0;
+
+ spin_lock(&sinfo->lock);
+ list_for_each_entry(block_group, &sinfo->ro_bgs, ro_list) {
spin_lock(&block_group->lock);
if (!block_group->ro) {
@@ -8594,26 +8601,6 @@
spin_unlock(&block_group->lock);
}
-
- return free_bytes;
-}
-
-/*
- * helper to account the unused space of all the readonly block group in the
- * space_info. takes mirrors into account.
- */
-u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo)
-{
- int i;
- u64 free_bytes = 0;
-
- spin_lock(&sinfo->lock);
-
- for (i = 0; i < BTRFS_NR_RAID_TYPES; i++)
- if (!list_empty(&sinfo->block_groups[i]))
- free_bytes += __btrfs_get_ro_block_group_free_space(
- &sinfo->block_groups[i]);
-
spin_unlock(&sinfo->lock);
return free_bytes;
@@ -8633,6 +8620,7 @@
cache->bytes_super - btrfs_block_group_used(&cache->item);
sinfo->bytes_readonly -= num_bytes;
cache->ro = 0;
+ list_del_init(&cache->ro_list);
spin_unlock(&cache->lock);
spin_unlock(&sinfo->lock);
}
@@ -9002,6 +8990,7 @@
INIT_LIST_HEAD(&cache->list);
INIT_LIST_HEAD(&cache->cluster_list);
INIT_LIST_HEAD(&cache->bg_list);
+ INIT_LIST_HEAD(&cache->ro_list);
btrfs_init_free_space_ctl(cache);
return cache;
@@ -9411,6 +9400,7 @@
* are still on the list after taking the semaphore
*/
list_del_init(&block_group->list);
+ list_del_init(&block_group->ro_list);
if (list_empty(&block_group->space_info->block_groups[index])) {
kobj = block_group->space_info->block_group_kobjs[index];
block_group->space_info->block_group_kobjs[index] = NULL;