[dpdk-dev] [PATCH 11/14] net/sfc: support VXLAN and NVGRE packet types classification

[Andrew Rybchenko <arybchenko@solarflare.com>]

Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Reviewed-by: Ivan Malov <ivan.malov@oktetlabs.ru>
Reviewed-by: Andy Moreton <amoreton@solarflare.com>
---
 doc/guides/nics/sfc_efx.rst   |  11 +++++
 drivers/net/sfc/sfc_dp_rx.h   |   4 +-
 drivers/net/sfc/sfc_ef10_rx.c | 102 ++++++++++++++++++++++++++++++++++++------
 drivers/net/sfc/sfc_ethdev.c  |   8 +++-
 drivers/net/sfc/sfc_rx.c      |   6 ++-
 5 files changed, 114 insertions(+), 17 deletions(-)

diff --git a/doc/guides/nics/sfc_efx.rst b/doc/guides/nics/sfc_efx.rst
index bde3cc8..994e111 100644
--- a/doc/guides/nics/sfc_efx.rst
+++ b/doc/guides/nics/sfc_efx.rst
@@ -124,6 +124,17 @@ with full-feature firmware variant running.
 **sfboot** should be used to configure NIC to run full-feature firmware variant.
 See Solarflare Server Adapter User's Guide for details.
 
+SFN8xxx family adapters provide either inner or outer packet classes.
+If adapter firmware advertises support for tunnels then the PMD
+configures the hardware to report inner classes, and outer classes are
+not reported in received packets.
+However, for VXLAN and GENEVE tunnels the PMD does report UDP as the
+outer layer 4 packet type.
+
+SFN8xxx family adapters report GENEVE packets as VXLAN.
+If UDP ports are configured for only one tunnel type then it is safe to
+treat VXLAN packet type indication as the corresponding UDP tunnel type.
+
 
 Flow API support
 ----------------
diff --git a/drivers/net/sfc/sfc_dp_rx.h b/drivers/net/sfc/sfc_dp_rx.h
index 3f6a604..33e06ac 100644
--- a/drivers/net/sfc/sfc_dp_rx.h
+++ b/drivers/net/sfc/sfc_dp_rx.h
@@ -150,7 +150,8 @@ typedef bool (sfc_dp_rx_qrx_ev_t)(struct sfc_dp_rxq *dp_rxq, unsigned int id);
 typedef void (sfc_dp_rx_qpurge_t)(struct sfc_dp_rxq *dp_rxq);
 
 /** Get packet types recognized/classified */
-typedef const uint32_t * (sfc_dp_rx_supported_ptypes_get_t)(void);
+typedef const uint32_t * (sfc_dp_rx_supported_ptypes_get_t)(
+				uint32_t tunnel_encaps);
 
 /** Get number of pending Rx descriptors */
 typedef unsigned int (sfc_dp_rx_qdesc_npending_t)(struct sfc_dp_rxq *dp_rxq);
@@ -166,6 +167,7 @@ struct sfc_dp_rx {
 	unsigned int				features;
 #define SFC_DP_RX_FEAT_SCATTER			0x1
 #define SFC_DP_RX_FEAT_MULTI_PROCESS		0x2
+#define SFC_DP_RX_FEAT_TUNNELS			0x4
 	sfc_dp_rx_qcreate_t			*qcreate;
 	sfc_dp_rx_qdestroy_t			*qdestroy;
 	sfc_dp_rx_qstart_t			*qstart;
diff --git a/drivers/net/sfc/sfc_ef10_rx.c b/drivers/net/sfc/sfc_ef10_rx.c
index 4c76f74..41c2885 100644
--- a/drivers/net/sfc/sfc_ef10_rx.c
+++ b/drivers/net/sfc/sfc_ef10_rx.c
@@ -251,6 +251,7 @@ static void
 sfc_ef10_rx_ev_to_offloads(struct sfc_ef10_rxq *rxq, const efx_qword_t rx_ev,
 			   struct rte_mbuf *m)
 {
+	uint32_t tun_ptype = 0;
 	uint32_t l2_ptype = 0;
 	uint32_t l3_ptype = 0;
 	uint32_t l4_ptype = 0;
@@ -259,15 +260,40 @@ sfc_ef10_rx_ev_to_offloads(struct sfc_ef10_rxq *rxq, const efx_qword_t rx_ev,
 	if (unlikely(EFX_TEST_QWORD_BIT(rx_ev, ESF_DZ_RX_PARSE_INCOMPLETE_LBN)))
 		goto done;
 
+	switch (EFX_QWORD_FIELD(rx_ev, ESF_EZ_RX_ENCAP_HDR)) {
+	default:
+		/* Unexpected encapsulation tag class */
+		SFC_ASSERT(false);
+		/* FALLTHROUGH */
+	case ESE_EZ_ENCAP_HDR_NONE:
+		break;
+	case ESE_EZ_ENCAP_HDR_VXLAN:
+		/*
+		 * It is definitely UDP, but we have no information
+		 * about IPv4 vs IPv6 and VLAN tagging.
+		 */
+		tun_ptype = RTE_PTYPE_TUNNEL_VXLAN | RTE_PTYPE_L4_UDP;
+		break;
+	case ESE_EZ_ENCAP_HDR_GRE:
+		/*
+		 * We have no information about IPv4 vs IPv6 and VLAN tagging.
+		 */
+		tun_ptype = RTE_PTYPE_TUNNEL_NVGRE;
+		break;
+	}
+
 	switch (EFX_QWORD_FIELD(rx_ev, ESF_DZ_RX_ETH_TAG_CLASS)) {
 	case ESE_DZ_ETH_TAG_CLASS_NONE:
-		l2_ptype = RTE_PTYPE_L2_ETHER;
+		l2_ptype = (tun_ptype == 0) ? RTE_PTYPE_L2_ETHER :
+			RTE_PTYPE_INNER_L2_ETHER;
 		break;
 	case ESE_DZ_ETH_TAG_CLASS_VLAN1:
-		l2_ptype = RTE_PTYPE_L2_ETHER_VLAN;
+		l2_ptype = (tun_ptype == 0) ? RTE_PTYPE_L2_ETHER_VLAN :
+			RTE_PTYPE_INNER_L2_ETHER_VLAN;
 		break;
 	case ESE_DZ_ETH_TAG_CLASS_VLAN2:
-		l2_ptype = RTE_PTYPE_L2_ETHER_QINQ;
+		l2_ptype = (tun_ptype == 0) ? RTE_PTYPE_L2_ETHER_QINQ :
+			RTE_PTYPE_INNER_L2_ETHER_QINQ;
 		break;
 	default:
 		/* Unexpected Eth tag class */
@@ -276,25 +302,31 @@ sfc_ef10_rx_ev_to_offloads(struct sfc_ef10_rxq *rxq, const efx_qword_t rx_ev,
 
 	switch (EFX_QWORD_FIELD(rx_ev, ESF_DZ_RX_L3_CLASS)) {
 	case ESE_DZ_L3_CLASS_IP4_FRAG:
-		l4_ptype = RTE_PTYPE_L4_FRAG;
+		l4_ptype = (tun_ptype == 0) ? RTE_PTYPE_L4_FRAG :
+			RTE_PTYPE_INNER_L4_FRAG;
 		/* FALLTHROUGH */
 	case ESE_DZ_L3_CLASS_IP4:
-		l3_ptype = RTE_PTYPE_L3_IPV4_EXT_UNKNOWN;
+		l3_ptype = (tun_ptype == 0) ? RTE_PTYPE_L3_IPV4_EXT_UNKNOWN :
+			RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN;
 		ol_flags |= PKT_RX_RSS_HASH |
 			((EFX_TEST_QWORD_BIT(rx_ev,
 					     ESF_DZ_RX_IPCKSUM_ERR_LBN)) ?
 			 PKT_RX_IP_CKSUM_BAD : PKT_RX_IP_CKSUM_GOOD);
 		break;
 	case ESE_DZ_L3_CLASS_IP6_FRAG:
-		l4_ptype = RTE_PTYPE_L4_FRAG;
+		l4_ptype = (tun_ptype == 0) ? RTE_PTYPE_L4_FRAG :
+			RTE_PTYPE_INNER_L4_FRAG;
 		/* FALLTHROUGH */
 	case ESE_DZ_L3_CLASS_IP6:
-		l3_ptype = RTE_PTYPE_L3_IPV6_EXT_UNKNOWN;
+		l3_ptype = (tun_ptype == 0) ? RTE_PTYPE_L3_IPV6_EXT_UNKNOWN :
+			RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN;
 		ol_flags |= PKT_RX_RSS_HASH;
 		break;
 	case ESE_DZ_L3_CLASS_ARP:
 		/* Override Layer 2 packet type */
-		l2_ptype = RTE_PTYPE_L2_ETHER_ARP;
+		/* There is no ARP classification for inner packets */
+		if (tun_ptype == 0)
+			l2_ptype = RTE_PTYPE_L2_ETHER_ARP;
 		break;
 	default:
 		/* Unexpected Layer 3 class */
@@ -303,14 +335,16 @@ sfc_ef10_rx_ev_to_offloads(struct sfc_ef10_rxq *rxq, const efx_qword_t rx_ev,
 
 	switch (EFX_QWORD_FIELD(rx_ev, ESF_DZ_RX_L4_CLASS)) {
 	case ESE_DZ_L4_CLASS_TCP:
-		l4_ptype = RTE_PTYPE_L4_TCP;
+		l4_ptype = (tun_ptype == 0) ? RTE_PTYPE_L4_TCP :
+			RTE_PTYPE_INNER_L4_TCP;
 		ol_flags |=
 			(EFX_TEST_QWORD_BIT(rx_ev,
 					    ESF_DZ_RX_TCPUDP_CKSUM_ERR_LBN)) ?
 			PKT_RX_L4_CKSUM_BAD : PKT_RX_L4_CKSUM_GOOD;
 		break;
 	case ESE_DZ_L4_CLASS_UDP:
-		l4_ptype = RTE_PTYPE_L4_UDP;
+		l4_ptype = (tun_ptype == 0) ? RTE_PTYPE_L4_UDP :
+			RTE_PTYPE_INNER_L4_UDP;
 		ol_flags |=
 			(EFX_TEST_QWORD_BIT(rx_ev,
 					    ESF_DZ_RX_TCPUDP_CKSUM_ERR_LBN)) ?
@@ -329,7 +363,7 @@ sfc_ef10_rx_ev_to_offloads(struct sfc_ef10_rxq *rxq, const efx_qword_t rx_ev,
 
 done:
 	m->ol_flags = ol_flags;
-	m->packet_type = l2_ptype | l3_ptype | l4_ptype;
+	m->packet_type = tun_ptype | l2_ptype | l3_ptype | l4_ptype;
 }
 
 static uint16_t
@@ -515,7 +549,7 @@ sfc_ef10_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 }
 
 static const uint32_t *
-sfc_ef10_supported_ptypes_get(void)
+sfc_ef10_supported_ptypes_get(uint32_t tunnel_encaps)
 {
 	static const uint32_t ef10_native_ptypes[] = {
 		RTE_PTYPE_L2_ETHER,
@@ -529,8 +563,47 @@ sfc_ef10_supported_ptypes_get(void)
 		RTE_PTYPE_L4_UDP,
 		RTE_PTYPE_UNKNOWN
 	};
+	static const uint32_t ef10_overlay_ptypes[] = {
+		RTE_PTYPE_L2_ETHER,
+		RTE_PTYPE_L2_ETHER_ARP,
+		RTE_PTYPE_L2_ETHER_VLAN,
+		RTE_PTYPE_L2_ETHER_QINQ,
+		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
+		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
+		RTE_PTYPE_L4_FRAG,
+		RTE_PTYPE_L4_TCP,
+		RTE_PTYPE_L4_UDP,
+		RTE_PTYPE_TUNNEL_VXLAN,
+		RTE_PTYPE_TUNNEL_NVGRE,
+		RTE_PTYPE_INNER_L2_ETHER,
+		RTE_PTYPE_INNER_L2_ETHER_VLAN,
+		RTE_PTYPE_INNER_L2_ETHER_QINQ,
+		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
+		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
+		RTE_PTYPE_INNER_L4_FRAG,
+		RTE_PTYPE_INNER_L4_TCP,
+		RTE_PTYPE_INNER_L4_UDP,
+		RTE_PTYPE_UNKNOWN
+	};
 
-	return ef10_native_ptypes;
+	/*
+	 * The function returns static set of supported packet types,
+	 * so we can't build it dynamically based on supported tunnel
+	 * encapsulations and should limit to known sets.
+	 */
+	switch (tunnel_encaps) {
+	case (1u << EFX_TUNNEL_PROTOCOL_VXLAN |
+	      1u << EFX_TUNNEL_PROTOCOL_GENEVE |
+	      1u << EFX_TUNNEL_PROTOCOL_NVGRE):
+		return ef10_overlay_ptypes;
+	default:
+		RTE_LOG(ERR, PMD,
+			"Unexpected set of supported tunnel encapsulations: %#x\n",
+			tunnel_encaps);
+		/* FALLTHROUGH */
+	case 0:
+		return ef10_native_ptypes;
+	}
 }
 
 static sfc_dp_rx_qdesc_npending_t sfc_ef10_rx_qdesc_npending;
@@ -707,7 +780,8 @@ struct sfc_dp_rx sfc_ef10_rx = {
 		.type		= SFC_DP_RX,
 		.hw_fw_caps	= SFC_DP_HW_FW_CAP_EF10,
 	},
-	.features		= SFC_DP_RX_FEAT_MULTI_PROCESS,
+	.features		= SFC_DP_RX_FEAT_MULTI_PROCESS |
+				  SFC_DP_RX_FEAT_TUNNELS,
 	.qcreate		= sfc_ef10_rx_qcreate,
 	.qdestroy		= sfc_ef10_rx_qdestroy,
 	.qstart			= sfc_ef10_rx_qstart,
diff --git a/drivers/net/sfc/sfc_ethdev.c b/drivers/net/sfc/sfc_ethdev.c
index 837fd55..0fea997 100644
--- a/drivers/net/sfc/sfc_ethdev.c
+++ b/drivers/net/sfc/sfc_ethdev.c
@@ -131,6 +131,10 @@ sfc_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 		DEV_RX_OFFLOAD_UDP_CKSUM |
 		DEV_RX_OFFLOAD_TCP_CKSUM;
 
+	if ((encp->enc_tunnel_encapsulations_supported != 0) &&
+	    (sa->dp_rx->features & SFC_DP_RX_FEAT_TUNNELS))
+		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM;
+
 	dev_info->tx_offload_capa =
 		DEV_TX_OFFLOAD_IPV4_CKSUM |
 		DEV_TX_OFFLOAD_UDP_CKSUM |
@@ -183,8 +187,10 @@ static const uint32_t *
 sfc_dev_supported_ptypes_get(struct rte_eth_dev *dev)
 {
 	struct sfc_adapter *sa = dev->data->dev_private;
+	const efx_nic_cfg_t *encp = efx_nic_cfg_get(sa->nic);
+	uint32_t tunnel_encaps = encp->enc_tunnel_encapsulations_supported;
 
-	return sa->dp_rx->supported_ptypes_get();
+	return sa->dp_rx->supported_ptypes_get(tunnel_encaps);
 }
 
 static int
diff --git a/drivers/net/sfc/sfc_rx.c b/drivers/net/sfc/sfc_rx.c
index 22bf372..70a72b3 100644
--- a/drivers/net/sfc/sfc_rx.c
+++ b/drivers/net/sfc/sfc_rx.c
@@ -193,7 +193,7 @@ sfc_efx_rx_desc_flags_to_packet_type(const unsigned int desc_flags)
 }
 
 static const uint32_t *
-sfc_efx_supported_ptypes_get(void)
+sfc_efx_supported_ptypes_get(__rte_unused uint32_t tunnel_encaps)
 {
 	static const uint32_t ptypes[] = {
 		RTE_PTYPE_L2_ETHER,
@@ -947,6 +947,10 @@ sfc_rx_qinit(struct sfc_adapter *sa, unsigned int sw_index,
 		sa->eth_dev->data->dev_conf.rxmode.enable_scatter ?
 		EFX_RXQ_FLAG_SCATTER : EFX_RXQ_FLAG_NONE;
 
+	if ((encp->enc_tunnel_encapsulations_supported != 0) &&
+	    (sa->dp_rx->features & SFC_DP_RX_FEAT_TUNNELS))
+		rxq_info->type_flags |= EFX_RXQ_FLAG_INNER_CLASSES;
+
 	rc = sfc_ev_qinit(sa, SFC_EVQ_TYPE_RX, sw_index,
 			  rxq_info->entries, socket_id, &evq);
 	if (rc != 0)
-- 
2.7.4