[PATCH v2 8/8] net: clear outer UDP checksum in Intel prepare helper

[David Marchand <david.marchand@redhat.com>]

If requesting an inner (L3/L4 checksum or L4 segmentation) offload,
when the hardware does not support recomputing outer UDP checksum,
automatically disable it in the common helper.

Signed-off-by: David Marchand <david.marchand@redhat.com>
---
 app/test-pmd/csumonly.c            | 10 ++-----
 doc/guides/prog_guide/mbuf_lib.rst |  8 +++---
 drivers/net/hns3/hns3_rxtx.c       | 44 ------------------------------
 lib/net/rte_net.h                  | 22 +++++++++++----
 4 files changed, 22 insertions(+), 62 deletions(-)

diff --git a/app/test-pmd/csumonly.c b/app/test-pmd/csumonly.c
index 71add6ca47..2246c22e8e 100644
--- a/app/test-pmd/csumonly.c
+++ b/app/test-pmd/csumonly.c
@@ -612,19 +612,13 @@ process_outer_cksums(void *outer_l3_hdr, struct testpmd_offload_info *info,
 		return ol_flags;
 	}
 
-	/* outer UDP checksum is done in software. In the other side, for
-	 * UDP tunneling, like VXLAN or Geneve, outer UDP checksum can be
-	 * set to zero.
+	/* Outer UDP checksum is done in software.
 	 *
 	 * If a packet will be TSOed into small packets by NIC, we cannot
 	 * set/calculate a non-zero checksum, because it will be a wrong
 	 * value after the packet be split into several small packets.
 	 */
-	if (tso_enabled)
-		udp_hdr->dgram_cksum = 0;
-
-	/* do not recalculate udp cksum if it was 0 */
-	if (udp_hdr->dgram_cksum != 0) {
+	if (!tso_enabled && udp_hdr->dgram_cksum != 0) {
 		udp_hdr->dgram_cksum = 0;
 		udp_hdr->dgram_cksum = get_udptcp_checksum(m, outer_l3_hdr,
 					info->outer_l2_len + info->outer_l3_len,
diff --git a/doc/guides/prog_guide/mbuf_lib.rst b/doc/guides/prog_guide/mbuf_lib.rst
index 4e285c0aab..35f7fffbc7 100644
--- a/doc/guides/prog_guide/mbuf_lib.rst
+++ b/doc/guides/prog_guide/mbuf_lib.rst
@@ -158,7 +158,7 @@ a vxlan-encapsulated tcp packet:
 
   This is similar to case 1), but l2_len is different. It is supported
   on hardware advertising RTE_ETH_TX_OFFLOAD_IPV4_CKSUM.
-  Note that it can only work if outer L4 checksum is 0.
+  Note that some driver may set outer L4 checksum to 0.
 
 - calculate checksum of in_ip and in_tcp::
 
@@ -169,7 +169,7 @@ a vxlan-encapsulated tcp packet:
   This is similar to case 2), but l2_len is different. It is supported
   on hardware advertising RTE_ETH_TX_OFFLOAD_IPV4_CKSUM and
   RTE_ETH_TX_OFFLOAD_TCP_CKSUM.
-  Note that it can only work if outer L4 checksum is 0.
+  Note that some driver may set outer L4 checksum to 0.
 
 - segment inner TCP::
 
@@ -180,7 +180,7 @@ a vxlan-encapsulated tcp packet:
       RTE_MBUF_F_TX_TCP_SEG;
 
   This is supported on hardware advertising RTE_ETH_TX_OFFLOAD_TCP_TSO.
-  Note that it can only work if outer L4 checksum is 0.
+  Note that some driver may set outer L4 checksum to 0.
 
 - calculate checksum of out_ip, in_ip, in_tcp::
 
@@ -193,7 +193,7 @@ a vxlan-encapsulated tcp packet:
 
   This is supported on hardware advertising RTE_ETH_TX_OFFLOAD_IPV4_CKSUM,
   RTE_ETH_TX_OFFLOAD_UDP_CKSUM and RTE_ETH_TX_OFFLOAD_OUTER_IPV4_CKSUM.
-  Note that it can only work if outer L4 checksum is 0.
+  Note that some driver may set outer L4 checksum to 0.
 
 The list of flags and their precise meaning is described in the mbuf API
 documentation (rte_mbuf.h). Also refer to the testpmd source code
diff --git a/drivers/net/hns3/hns3_rxtx.c b/drivers/net/hns3/hns3_rxtx.c
index 03fc919fd7..b5436c51e7 100644
--- a/drivers/net/hns3/hns3_rxtx.c
+++ b/drivers/net/hns3/hns3_rxtx.c
@@ -3616,47 +3616,6 @@ hns3_pkt_need_linearized(struct rte_mbuf *tx_pkts, uint32_t bd_num,
 	return false;
 }
 
-static void
-hns3_outer_header_cksum_prepare(struct rte_mbuf *m)
-{
-	uint64_t ol_flags = m->ol_flags;
-	uint32_t paylen, hdr_len, l4_proto;
-	struct rte_udp_hdr *udp_hdr;
-
-	if (!(ol_flags & (RTE_MBUF_F_TX_OUTER_IPV4 | RTE_MBUF_F_TX_OUTER_IPV6)) &&
-			((ol_flags & RTE_MBUF_F_TX_OUTER_UDP_CKSUM) ||
-			!(ol_flags & RTE_MBUF_F_TX_TCP_SEG)))
-		return;
-
-	if (ol_flags & RTE_MBUF_F_TX_OUTER_IPV4) {
-		struct rte_ipv4_hdr *ipv4_hdr;
-
-		ipv4_hdr = rte_pktmbuf_mtod_offset(m, struct rte_ipv4_hdr *,
-			m->outer_l2_len);
-		l4_proto = ipv4_hdr->next_proto_id;
-	} else {
-		struct rte_ipv6_hdr *ipv6_hdr;
-
-		ipv6_hdr = rte_pktmbuf_mtod_offset(m, struct rte_ipv6_hdr *,
-					   m->outer_l2_len);
-		l4_proto = ipv6_hdr->proto;
-	}
-
-	if (l4_proto != IPPROTO_UDP)
-		return;
-
-	/* driver should ensure the outer udp cksum is 0 for TUNNEL TSO */
-	hdr_len = m->l2_len + m->l3_len + m->l4_len;
-	hdr_len += m->outer_l2_len + m->outer_l3_len;
-	paylen = m->pkt_len - hdr_len;
-	if (paylen <= m->tso_segsz)
-		return;
-	udp_hdr = rte_pktmbuf_mtod_offset(m, struct rte_udp_hdr *,
-					  m->outer_l2_len +
-					  m->outer_l3_len);
-	udp_hdr->dgram_cksum = 0;
-}
-
 static int
 hns3_check_tso_pkt_valid(struct rte_mbuf *m)
 {
@@ -3834,7 +3793,6 @@ hns3_prep_pkt_proc(struct hns3_tx_queue *tx_queue, struct rte_mbuf *m)
 			 * checksum of packets that need TSO, so network driver
 			 * software not need to recalculate it.
 			 */
-			hns3_outer_header_cksum_prepare(m);
 			return 0;
 		}
 	}
@@ -3848,8 +3806,6 @@ hns3_prep_pkt_proc(struct hns3_tx_queue *tx_queue, struct rte_mbuf *m)
 	if (!hns3_validate_tunnel_cksum(tx_queue, m))
 		return 0;
 
-	hns3_outer_header_cksum_prepare(m);
-
 	return 0;
 }
 
diff --git a/lib/net/rte_net.h b/lib/net/rte_net.h
index efd9d5f5ee..79e969464b 100644
--- a/lib/net/rte_net.h
+++ b/lib/net/rte_net.h
@@ -109,6 +109,10 @@ static inline int
 rte_net_intel_cksum_flags_prepare(struct rte_mbuf *m, uint64_t ol_flags)
 {
 	/* Initialise ipv4_hdr to avoid false positive compiler warnings. */
+	const uint64_t inner_requests = RTE_MBUF_F_TX_IP_CKSUM | RTE_MBUF_F_TX_L4_MASK |
+		RTE_MBUF_F_TX_TCP_SEG | RTE_MBUF_F_TX_UDP_SEG;
+	const uint64_t outer_requests = RTE_MBUF_F_TX_OUTER_IP_CKSUM |
+		RTE_MBUF_F_TX_OUTER_UDP_CKSUM;
 	struct rte_ipv4_hdr *ipv4_hdr = NULL;
 	struct rte_ipv6_hdr *ipv6_hdr;
 	struct rte_tcp_hdr *tcp_hdr;
@@ -120,9 +124,7 @@ rte_net_intel_cksum_flags_prepare(struct rte_mbuf *m, uint64_t ol_flags)
 	 * Mainly it is required to avoid fragmented headers check if
 	 * no offloads are requested.
 	 */
-	if (!(ol_flags & (RTE_MBUF_F_TX_IP_CKSUM | RTE_MBUF_F_TX_L4_MASK | RTE_MBUF_F_TX_TCP_SEG |
-					RTE_MBUF_F_TX_UDP_SEG | RTE_MBUF_F_TX_OUTER_IP_CKSUM |
-					RTE_MBUF_F_TX_OUTER_UDP_CKSUM)))
+	if (!(ol_flags & (inner_requests | outer_requests)))
 		return 0;
 
 	if (ol_flags & (RTE_MBUF_F_TX_OUTER_IPV4 | RTE_MBUF_F_TX_OUTER_IPV6)) {
@@ -136,19 +138,27 @@ rte_net_intel_cksum_flags_prepare(struct rte_mbuf *m, uint64_t ol_flags)
 					struct rte_ipv4_hdr *, m->outer_l2_len);
 			ipv4_hdr->hdr_checksum = 0;
 		}
-		if (ol_flags & RTE_MBUF_F_TX_OUTER_UDP_CKSUM) {
+		if (ol_flags & RTE_MBUF_F_TX_OUTER_UDP_CKSUM || ol_flags & inner_requests) {
 			if (ol_flags & RTE_MBUF_F_TX_OUTER_IPV4) {
 				ipv4_hdr = rte_pktmbuf_mtod_offset(m, struct rte_ipv4_hdr *,
 					m->outer_l2_len);
 				udp_hdr = (struct rte_udp_hdr *)((char *)ipv4_hdr +
 					m->outer_l3_len);
-				udp_hdr->dgram_cksum = rte_ipv4_phdr_cksum(ipv4_hdr, m->ol_flags);
+				if (ol_flags & RTE_MBUF_F_TX_OUTER_UDP_CKSUM)
+					udp_hdr->dgram_cksum = rte_ipv4_phdr_cksum(ipv4_hdr,
+						m->ol_flags);
+				else
+					udp_hdr->dgram_cksum = 0;
 			} else {
 				ipv6_hdr = rte_pktmbuf_mtod_offset(m, struct rte_ipv6_hdr *,
 					m->outer_l2_len);
 				udp_hdr = rte_pktmbuf_mtod_offset(m, struct rte_udp_hdr *,
 					 m->outer_l2_len + m->outer_l3_len);
-				udp_hdr->dgram_cksum = rte_ipv6_phdr_cksum(ipv6_hdr, m->ol_flags);
+				if (ol_flags & RTE_MBUF_F_TX_OUTER_UDP_CKSUM)
+					udp_hdr->dgram_cksum = rte_ipv6_phdr_cksum(ipv6_hdr,
+						m->ol_flags);
+				else
+					udp_hdr->dgram_cksum = 0;
 			}
 		}
 	}
-- 
2.44.0