Re: [PATCH v3 2/2] x86/resctrl: Avoid redundant counter read in __mon_event_count()

[Moger, Babu]

On 12/20/22 10:41, Peter Newman wrote:
> __mon_event_count() does the per-RMID, per-domain work for
> user-initiated event count reads and the initialization of new monitor
> groups.
>
> In the initialization case, after resctrl_arch_reset_rmid() calls
> __rmid_read() to record an initial count for a new monitor group, it
> immediately calls resctrl_arch_rmid_read(). This re-read of the hardware
> counter is unnecessary and the following computations are ignored by the
> caller during initialization.
>
> Following return from resctrl_arch_reset_rmid(), just clear the
> mbm_state and return. This involves moving the mbm_state lookup into the
> rr->first case, as it's not needed for regular event count reads: the
> QOS_L3_OCCUP_EVENT_ID case was redundant with the accumulating logic at
> the end of the function.
>
> Signed-off-by: Peter Newman <peternewman@xxxxxxxxxx>
> Reviewed-by: Reinette Chatre <reinette.chatre@xxxxxxxxx>
> ---
> v3:
>  - changelog clarifications suggested by Reinette
> v2:
>  (patch introduced)
>
> v2: https://lore.kernel.org/lkml/20221214160856.2164207-2-peternewman@xxxxxxxxxx/
> ---
>  arch/x86/kernel/cpu/resctrl/monitor.c | 43 ++++++++++++---------------
>  1 file changed, 19 insertions(+), 24 deletions(-)
>
> diff --git a/arch/x86/kernel/cpu/resctrl/monitor.c b/arch/x86/kernel/cpu/resctrl/monitor.c
> index 77538abeb72a..e708df478077 100644
> --- a/arch/x86/kernel/cpu/resctrl/monitor.c
> +++ b/arch/x86/kernel/cpu/resctrl/monitor.c
> @@ -366,41 +366,36 @@ void free_rmid(u32 rmid)
>  		list_add_tail(&entry->list, &rmid_free_lru);
>  }
>  
> +static struct mbm_state *get_mbm_state(struct rdt_domain *d, u32 rmid,
> +				       enum resctrl_event_id evtid)
> +{
> +	switch (evtid) {
> +	case QOS_L3_MBM_TOTAL_EVENT_ID:
> +		return &d->mbm_total[rmid];
> +	case QOS_L3_MBM_LOCAL_EVENT_ID:
> +		return &d->mbm_local[rmid];
> +	default:
> +		return NULL;
> +	}
> +}
> +
>  static int __mon_event_count(u32 rmid, struct rmid_read *rr)
>  {
>  	struct mbm_state *m;
>  	u64 tval = 0;
>  
> -	if (rr->first)
> +	if (rr->first) {
>  		resctrl_arch_reset_rmid(rr->r, rr->d, rmid, rr->evtid);
> +		m = get_mbm_state(rr->d, rmid, rr->evtid);
> +		if (m)
> +			memset(m, 0, sizeof(struct mbm_state));
> +		return 0;
> +	}
>  
>  	rr->err = resctrl_arch_rmid_read(rr->r, rr->d, rmid, rr->evtid, &tval);
>  	if (rr->err)
>  		return rr->err;
>  
> -	switch (rr->evtid) {
> -	case QOS_L3_OCCUP_EVENT_ID:
> -		rr->val += tval;
> -		return 0;
> -	case QOS_L3_MBM_TOTAL_EVENT_ID:
> -		m = &rr->d->mbm_total[rmid];
> -		break;
> -	case QOS_L3_MBM_LOCAL_EVENT_ID:
> -		m = &rr->d->mbm_local[rmid];
> -		break;
> -	default:
> -		/*
> -		 * Code would never reach here because an invalid
> -		 * event id would fail in resctrl_arch_rmid_read().
> -		 */
> -		return -EINVAL;
> -	}
> -
> -	if (rr->first) {
> -		memset(m, 0, sizeof(struct mbm_state));
> -		return 0;
> -	}
> -
>  	rr->val += tval;
>  
>  	return 0;

Tested the patches on AMD systems. Looks good.
Tested-by: Babu Moger <babu.moger@xxxxxxx>

Thanks
Babu Moger