Re: [PATCH net-next,v2 04/12] cls_api: add translator to flow_action representation

[Pablo Neira Ayuso <pablo@netfilter.org>]

On Mon, Nov 19, 2018 at 01:12:51PM +0100, Jiri Pirko wrote:
> Mon, Nov 19, 2018 at 01:15:11AM CET, pablo@netfilter.org wrote:
> >@@ -2567,6 +2575,111 @@ int tc_setup_cb_call(struct tcf_block *block, struct tcf_exts *exts,
> > }
> > EXPORT_SYMBOL(tc_setup_cb_call);
> > 
> >+int tc_setup_flow_action(struct flow_action *flow_action,
> >+			 const struct tcf_exts *exts)
> >+{
> >+	const struct tc_action *act;
> >+	int num_acts = 0, i, j, k;
> >+
> >+	if (!exts)
> >+		return 0;
> >+
> >+	tcf_exts_for_each_action(i, act, exts) {
> >+		if (is_tcf_pedit(act))
> >+			num_acts += tcf_pedit_nkeys(act);
> >+		else
> >+			num_acts++;
> >+	}
> >+	if (!num_acts)
> >+		return 0;
> >+
> >+	if (flow_action_init(flow_action, num_acts) < 0)
> 
> This is actually a "alloc" function. And the counterpart is "free".

I can rename it to _alloc() if you prefer.

> How about to allocate the container struct which would have the [0]
> trick for the array of action?

You mean turn *keys into keys[0] stub in struct flow_action? This is
embedded into struct tc_cls_flower_offload, I may need to make a
second look but I think it won't fly.

BTW, side note: I will rename keys to "array" given keys is not
semantically appropriate as you mentioned, BTW.

Thanks!