Re: [ovs-dev] [PATCH net-next v7 05/10] net: openvswitch: add psample action

[Simon Horman <horms@kernel.org>]

On Tue, Jul 02, 2024 at 11:53:01AM +0200, Ilya Maximets wrote:
> On 7/2/24 11:37, Simon Horman wrote:
> > On Tue, Jul 02, 2024 at 03:05:02AM -0400, Adrián Moreno wrote:
> >> On Mon, Jul 01, 2024 at 02:23:12PM GMT, Aaron Conole wrote:
> >>> Adrian Moreno <amorenoz@redhat.com> writes:
> > 
> > ...
> > 
> >>>> diff --git a/net/openvswitch/actions.c b/net/openvswitch/actions.c
> > 
> > ...
> > 
> >>>> @@ -1299,6 +1304,39 @@ static int execute_dec_ttl(struct sk_buff *skb, struct sw_flow_key *key)
> >>>>  	return 0;
> >>>>  }
> >>>>
> >>>> +#if IS_ENABLED(CONFIG_PSAMPLE)
> >>>> +static void execute_psample(struct datapath *dp, struct sk_buff *skb,
> >>>> +			    const struct nlattr *attr)
> >>>> +{
> >>>> +	struct psample_group psample_group = {};
> >>>> +	struct psample_metadata md = {};
> >>>> +	const struct nlattr *a;
> >>>> +	int rem;
> >>>> +
> >>>> +	nla_for_each_attr(a, nla_data(attr), nla_len(attr), rem) {
> >>>> +		switch (nla_type(a)) {
> >>>> +		case OVS_PSAMPLE_ATTR_GROUP:
> >>>> +			psample_group.group_num = nla_get_u32(a);
> >>>> +			break;
> >>>> +
> >>>> +		case OVS_PSAMPLE_ATTR_COOKIE:
> >>>> +			md.user_cookie = nla_data(a);
> >>>> +			md.user_cookie_len = nla_len(a);
> >>>> +			break;
> >>>> +		}
> >>>> +	}
> >>>> +
> >>>> +	psample_group.net = ovs_dp_get_net(dp);
> >>>> +	md.in_ifindex = OVS_CB(skb)->input_vport->dev->ifindex;
> >>>> +	md.trunc_size = skb->len - OVS_CB(skb)->cutlen;
> >>>> +
> >>>> +	psample_sample_packet(&psample_group, skb, 0, &md);
> >>>> +}
> >>>> +#else
> >>>> +static inline void execute_psample(struct datapath *dp, struct sk_buff *skb,
> >>>> +				   const struct nlattr *attr) {}
> >>>
> >>> I noticed that this got flagged in patchwork since it is 'static inline'
> >>> while being part of a complete translation unit - but I also see some
> >>> other places where that has been done.  I guess it should be just
> >>> 'static' though.  I don't feel very strongly about it.
> >>>
> >>
> >> We had a bit of discussion about this with Ilya. It seems "static
> >> inline" is a common pattern around the kernel. The coding style
> >> documentation says:
> >> "Generally, inline functions are preferable to macros resembling functions."
> >>
> >> So I think this "inline" is correct but I might be missing something.
> > 
> > Hi Adrián,
> > 
> > TL;DR: Please remove this inline keyword
> > 
> > For Kernel networking code at least it is strongly preferred not
> > to use inline in .c files unless there is a demonstrable - usually
> > performance - reason to do so. Rather, it is preferred to let the
> > compiler decide when to inline such functions. OTOH, the inline
> > keyword in .h files is fine.
> 
> FWIW, the main reason for 'inline' here is not performance, but silencing
> compiler's potential 'maybe unused' warnings:
> 
>  Function-like macros with unused parameters should be replaced by static
>  inline functions to avoid the issue of unused variables
> 
> I think, the rule for static inline functions in .c files is at odds with
> the 'Conditional Compilation' section of coding style.  The section does
> recommend to avoid conditional function declaration in .c files, but I'm not
> sure it is reasonable to export internal static functions for that reason.
> 
> In this particular case we can either define a macro, which is discouraged
> by the coding style:
> 
>  Generally, inline functions are preferable to macros resembling functions.
> 
> Or create a static inline function, that is against rule of no static
> inline functions in .c files.
> 
> Or create a simple static function and mark all the arguments as unused,
> which kind of compliant to the coding style, but the least pretty.

Hi Ilya,

I guess I would lean towards the last option.
But in any case, thanks for pointing out that this is complex:
I had not realised that when I wrote my previous email.