Re: [PATCH -next] blk-cgroup: move congestion_count to struct blkcg

[Tejun Heo]

Hello,

Sorry about the previous reply. I completely misread the patch.

On Fri, Jul 12, 2024 at 08:51:41AM +0000, Xiu Jianfeng wrote:
...
> only compiling tested

It'd be better if there's a bit more verification.

> diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
> index 37e6cc91d576..01d3408c2fc6 100644
> --- a/block/blk-cgroup.c
> +++ b/block/blk-cgroup.c
> @@ -2183,11 +2183,13 @@ void blk_cgroup_bio_start(struct bio *bio)
>  bool blk_cgroup_congested(void)
>  {
>  	struct cgroup_subsys_state *css;
> +	struct blkcg *blkcg;

It'd be better to define this within the loop.

>  	bool ret = false;
>  
>  	rcu_read_lock();
>  	for (css = blkcg_css(); css; css = css->parent) {

Also, if we're now dealing with blkcg's, there's no reason to go blkcg ->
css -> blkcg again. It'd be better to get the initial blkcg and then walk up
using blkcg_parent().

> @@ -95,6 +95,8 @@ struct blkcg {
>  	struct cgroup_subsys_state	css;
>  	spinlock_t			lock;
>  	refcount_t			online_pin;
> +	/* If there is block congestion on this cgroup. */
> +	atomic_t congestion_count;

Can you please match the indentation?

Thanks.

-- 
tejun