Re: [dpdk-dev] [PATCH v6 08/10] examples/ipsec-secgw: make data-path to use ipsec library

["Ananyev, Konstantin" <konstantin.ananyev@intel.com>]

> 
> On 1/4/2019 1:55 AM, Konstantin Ananyev wrote:
> >   	/* process ipv4 packets */
> > -	inbound_sp_sa(ctx->sp4_ctx, ctx->sa_ctx, &trf.ip4, 0);
> > -	route4_pkts(qconf->rt4_ctx, trf.ip4.pkts, trf.ip4.num);
> > +	if (trf.ip4.num != 0) {
> > +		inbound_sp_sa(ctx->sp4_ctx, ctx->sa_ctx, &trf.ip4, 0);
> > +		route4_pkts(qconf->rt4_ctx, trf.ip4.pkts, trf.ip4.num);
> > +	}
> >
> >   	/* process ipv6 packets */
> > -	inbound_sp_sa(ctx->sp6_ctx, ctx->sa_ctx, &trf.ip6, 0);
> > -	route6_pkts(qconf->rt6_ctx, trf.ip6.pkts, trf.ip6.num);
> > +	if (trf.ip6.num != 0) {
> > +		inbound_sp_sa(ctx->sp6_ctx, ctx->sa_ctx, &trf.ip6, 0);
> > +		route6_pkts(qconf->rt6_ctx, trf.ip6.pkts, trf.ip6.num);
> > +	}
> One more thing.
> Why do you need these changes. There are multiple places where this
> change is done.
> 
> route6_pkts and route4_pkts already checks that internally for 0 pkts.
> inbound_sp_sa also does that.
> These are getting called from multiple places. So it is better we add this check inside the function itself instead of outside it.

You right, these checks are not necessary here.
I put these extra checks to avoid unnecessary calls -
there is no guarantee that both route_pkts() and inbound_sp_sa()
will always be inlined by the compiler (they are not that small).
Konstantin