Re: [RFC 09/18] net: macsec: add support for getting offloaded stats

[Antoine Tenart <antoine.tenart@bootlin.com>]

Hello,

On Fri, Feb 14, 2020 at 06:02:49PM +0300, Igor Russkikh wrote:
> From: Dmitry Bogdanov <dbogdanov@marvell.com>
> 
> report sw stats only if offloaded stats are not supported

Please improve the changelog.

> -static int copy_tx_sa_stats(struct sk_buff *skb,
> -			    struct macsec_tx_sa_stats __percpu *pstats)
> +static void get_tx_sa_stats(struct net_device *dev, int an,
> +			    struct macsec_tx_sa *tx_sa,
> +			    struct macsec_tx_sa_stats *sum)
>  {
> -	struct macsec_tx_sa_stats sum = {0, };
> +	struct macsec_dev *macsec = macsec_priv(dev);
> +	int err = -EOPNOTSUPP;
>  	int cpu;
>  
> +	/* If h/w offloading is available, propagate to the device */
> +	if (macsec_is_offloaded(macsec)) {
> +		const struct macsec_ops *ops;
> +		struct macsec_context ctx;
> +
> +		ops = macsec_get_ops(macsec, &ctx);
> +		if (ops) {
> +			ctx.sa.assoc_num = an;
> +			ctx.sa.tx_sa = tx_sa;
> +			ctx.stats.tx_sa_stats = sum;
> +			ctx.secy = &macsec_priv(dev)->secy;
> +			err = macsec_offload(ops->mdo_get_tx_sa_stats, &ctx);
> +		}
> +	}
> +
> +	if (err != -EOPNOTSUPP)
> +		return;

If offloading is enabled and ops->mdo_get_tx_sa_stats is returning
-EOPNOTSUPP, do we really want to return the s/w stats as they'll be
wrong and out of sync with the hardware anyway?

(The same applies for the other statistics retrieval in the same patch).

> +
>  	for_each_possible_cpu(cpu) {
> -		const struct macsec_tx_sa_stats *stats = per_cpu_ptr(pstats, cpu);
> +		const struct macsec_tx_sa_stats *stats =
> +			per_cpu_ptr(tx_sa->stats, cpu);
>  
> -		sum.OutPktsProtected += stats->OutPktsProtected;
> -		sum.OutPktsEncrypted += stats->OutPktsEncrypted;
> +		sum->OutPktsProtected += stats->OutPktsProtected;
> +		sum->OutPktsEncrypted += stats->OutPktsEncrypted;
>  	}
> +}

Thanks!
Antoine

-- 
Antoine Ténart, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com