Re: [PATCH bpf-next] selftests/bpf: use ifname instead of ifindex in XDP compliance test tool

[Daniel Borkmann <daniel@xxxxxxxxxxxxx>]

On 3/3/23 6:21 PM, Lorenzo Bianconi wrote:
> Rely on interface name instead of interface index in error messages or logs
> from XDP compliance test tool.
> Improve XDP compliance test tool error messages.
>
> Signed-off-by: Lorenzo Bianconi <lorenzo@xxxxxxxxxx>
> ---
>   tools/testing/selftests/bpf/xdp_features.c | 92 ++++++++++++++--------
>   1 file changed, 57 insertions(+), 35 deletions(-)
>
> diff --git a/tools/testing/selftests/bpf/xdp_features.c b/tools/testing/selftests/bpf/xdp_features.c
> index fce12165213b..7414801cd7ec 100644
> --- a/tools/testing/selftests/bpf/xdp_features.c
> +++ b/tools/testing/selftests/bpf/xdp_features.c
> @@ -25,6 +25,7 @@
>   
>   static struct env {
>   	bool verbosity;
> +	char ifname[IF_NAMESIZE];
>   	int ifindex;
>   	bool is_tester;
>   	struct {
> @@ -109,25 +110,25 @@ static int get_xdp_feature(const char *arg)
>   	return 0;
>   }
>   
> -static char *get_xdp_feature_str(void)
> +static char *get_xdp_feature_str(bool color)
>   {
>   	switch (env.feature.action) {
>   	case XDP_PASS:
> -		return YELLOW("XDP_PASS");
> +		return color ? YELLOW("XDP_PASS") : "XDP_PASS";
>   	case XDP_DROP:
> -		return YELLOW("XDP_DROP");
> +		return color ? YELLOW("XDP_DROP") : "XDP_DROP";
>   	case XDP_ABORTED:
> -		return YELLOW("XDP_ABORTED");
> +		return color ? YELLOW("XDP_ABORTED") : "XDP_ABORTED";
>   	case XDP_TX:
> -		return YELLOW("XDP_TX");
> +		return color ? YELLOW("XDP_TX") : "XDP_TX";
>   	case XDP_REDIRECT:
> -		return YELLOW("XDP_REDIRECT");
> +		return color ? YELLOW("XDP_REDIRECT") : "XDP_REDIRECT";
>   	default:
>   		break;
>   	}
>   
>   	if (env.feature.drv_feature == NETDEV_XDP_ACT_NDO_XMIT)
> -		return YELLOW("XDP_NDO_XMIT");
> +		return color ? YELLOW("XDP_NDO_XMIT") : "XDP_NDO_XMIT";
>   
>   	return "";
>   }

Please split this into multiple patches, logically separated. This one is changing
multiple things at once and above has not much relation to relying on interface names.

Thanks,
Daniel