Re: [RFC] psi: Add additional PSI counters for each type of memory pressure

[Peter Zijlstra]

On Wed, Nov 10, 2021 at 07:36:37AM -0800, Georgi Djakov wrote:
> @@ -21,7 +19,18 @@ enum psi_task_count {
>  	 * don't have to special case any state tracking for it.
>  	 */
>  	NR_ONCPU,
> -	NR_PSI_TASK_COUNTS = 4,
> +	NR_BLK_CGROUP_THROTTLE,
> +	NR_BIO,
> +	NR_COMPACTION,
> +	NR_THRASHING,
> +	NR_CGROUP_RECLAIM_HIGH,
> +	NR_CGROUP_RECLAIM_HIGH_SLEEP,
> +	NR_CGROUP_TRY_CHARGE,
> +	NR_DIRECT_COMPACTION,
> +	NR_DIRECT_RECLAIM,
> +	NR_READ_SWAPPAGE,
> +	NR_KSWAPD,
> +	NR_PSI_TASK_COUNTS = 16,
>  };
>  

> @@ -51,9 +80,20 @@ enum psi_states {
>  	PSI_MEM_FULL,
>  	PSI_CPU_SOME,
>  	PSI_CPU_FULL,
> +	PSI_BLK_CGROUP_THROTTLE,
> +	PSI_BIO,
> +	PSI_COMPACTION,
> +	PSI_THRASHING,
> +	PSI_CGROUP_RECLAIM_HIGH,
> +	PSI_CGROUP_RECLAIM_HIGH_SLEEP,
> +	PSI_CGROUP_TRY_CHARGE,
> +	PSI_DIRECT_COMPACTION,
> +	PSI_DIRECT_RECLAIM,
> +	PSI_READ_SWAPPAGE,
> +	PSI_KSWAPD,
>  	/* Only per-CPU, to weigh the CPU in the global average: */
>  	PSI_NONIDLE,
> -	NR_PSI_STATES = 7,
> +	NR_PSI_STATES = 18,
>  };

Have you considered what this does to psi_group_cpu's size and layout
and the impact thereof on performance?