[PATCH V3] app/testpmd: fix GENEVE parsing in csum forward mode

[Ferruh Yigit]

On 2/20/2022 12:09 PM, Raja Zidane wrote:
> The csum FWD mode parses any received packet to set mbuf offloads for the
> transmitting burst, mainly in the checksum/TSO areas.
> In the case of a tunnel header, the csum FWD tries to detect known tunnels
> by the standard definition using the header'sdata and fallback to check the
> packet type in the mbuf to see if the Rx port driver already sign the
> packet as a tunnel.
> In the fallback case, the csum assumes the tunnel is VXLAN and parses the
> tunnel as VXLAN.
> When the GENEVE tunnel was added to the known tunnels in csum, its parsing
> trial was wrongly located after the pkt type detection, causing the csum to
> parse the GENEVE header as VXLAN when the Rx port set the tunnel packet
> type.
> 
> Remove the fall back case to VxLan.
> Log error of unrecognized tunnel if no tunnel was parsed successfully.
> 
> Fixes: c10a026c3b03 ("app/testpmd: introduce vxlan parsing function in csum fwd engine")
> Cc: stable at dpdk.org
> 
> Signed-off-by: Raja Zidane <rzidane at nvidia.com>
> ---
> V2: Log error when an unrecognized tunnel is found (unknown UDP dst port), instead of parsing it as VxLan by default.
> V3: revert unneeded changes (swapping parse_geneve & parse_vxlan).
>   app/test-pmd/csumonly.c | 15 +++++++++------
>   1 file changed, 9 insertions(+), 6 deletions(-)
> 
> diff --git a/app/test-pmd/csumonly.c b/app/test-pmd/csumonly.c
> index 02bc3929c7..e9ade3b864 100644
> --- a/app/test-pmd/csumonly.c
> +++ b/app/test-pmd/csumonly.c
> @@ -258,8 +258,7 @@ parse_gtp(struct rte_udp_hdr *udp_hdr,
>   /* Parse a vxlan header */
>   static void
>   parse_vxlan(struct rte_udp_hdr *udp_hdr,
> -	    struct testpmd_offload_info *info,
> -	    uint32_t pkt_type)
> +	    struct testpmd_offload_info *info)
>   {
>   	struct rte_ether_hdr *eth_hdr;
>   
> @@ -267,8 +266,7 @@ parse_vxlan(struct rte_udp_hdr *udp_hdr,
>   	 * default vxlan port (rfc7348) or that the rx offload flag is set
>   	 * (i40e only currently)
>   	 */
> -	if (udp_hdr->dst_port != _htons(RTE_VXLAN_DEFAULT_PORT) &&
> -		RTE_ETH_IS_TUNNEL_PKT(pkt_type) == 0)
> +	if (udp_hdr->dst_port != _htons(RTE_VXLAN_DEFAULT_PORT))
>   		return;
>   
>   	update_tunnel_outer(info);
> @@ -922,8 +920,7 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
>   						RTE_MBUF_F_TX_TUNNEL_VXLAN_GPE;
>   					goto tunnel_update;
>   				}
> -				parse_vxlan(udp_hdr, &info,
> -					    m->packet_type);
> +				parse_vxlan(udp_hdr, &info);
>   				if (info.is_tunnel) {
>   					tx_ol_flags |=
>   						RTE_MBUF_F_TX_TUNNEL_VXLAN;
> @@ -935,6 +932,12 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
>   						RTE_MBUF_F_TX_TUNNEL_GENEVE;
>   					goto tunnel_update;
>   				}
> +				/* Always keep last. */
> +				if (unlikely(RTE_ETH_IS_TUNNEL_PKT(
> +							m->packet_type) != 0)) {
> +					TESTPMD_LOG(ERR, "Unknown tunnel packet. UDP dst port: %hu",
> +						udp_hdr->dst_port);

ERR type log can in datapath cause too much log, and not sure
what user can do with this error log.

But I can see tunnel type set by PMD but not parsed goes through
silently also not good.

So what about making log type DEBUG, at least it will be visible
to who wants to get more information about what is going on.

> +				}
>   			} else if (info.l4_proto == IPPROTO_GRE) {
>   				struct simple_gre_hdr *gre_hdr;
>