[RFC] net/gve: add IPv4 checksum offloading capability

[RFC] net/gve: add IPv4 checksum offloading capability

[Rushil Gupta]

Gvnic's DQO format allows offloading IPv4 checksum.
Made changes to Tx and Rx path to translate DPDK flags
to descriptor for offloading (and vice-versa).
Added ptype adminq support to only add this flags for
supported L3/L4 packet-types.
---
 drivers/net/gve/gve_ethdev.c | 29 +++++++++++++++++++++++++--
 drivers/net/gve/gve_ethdev.h |  5 +++++
 drivers/net/gve/gve_rx_dqo.c | 38 ++++++++++++++++++++++++++++++++++--
 drivers/net/gve/gve_tx_dqo.c |  2 +-
 4 files changed, 69 insertions(+), 5 deletions(-)

diff --git a/drivers/net/gve/gve_ethdev.c b/drivers/net/gve/gve_ethdev.c
index 3b8ec5872d..ef0116218e 100644
--- a/drivers/net/gve/gve_ethdev.c
+++ b/drivers/net/gve/gve_ethdev.c
@@ -434,8 +434,14 @@ gve_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 		RTE_ETH_TX_OFFLOAD_SCTP_CKSUM	|
 		RTE_ETH_TX_OFFLOAD_TCP_TSO;
 
-	if (priv->queue_format == GVE_DQO_RDA_FORMAT)
-		dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_TCP_LRO;
+	if (!gve_is_gqi(priv)) {
+		dev_info->tx_offload_capa |= RTE_ETH_TX_OFFLOAD_IPV4_CKSUM;
+		dev_info->rx_offload_capa |=
+				RTE_ETH_RX_OFFLOAD_IPV4_CKSUM   |
+				RTE_ETH_RX_OFFLOAD_UDP_CKSUM	|
+				RTE_ETH_RX_OFFLOAD_TCP_CKSUM	|
+				RTE_ETH_RX_OFFLOAD_TCP_LRO;
+	}
 
 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
 		.rx_free_thresh = GVE_DEFAULT_RX_FREE_THRESH,
@@ -938,6 +944,8 @@ gve_teardown_device_resources(struct gve_priv *priv)
 		if (err)
 			PMD_DRV_LOG(ERR, "Could not deconfigure device resources: err=%d", err);
 	}
+	rte_free(priv->ptype_lut_dqo);
+	priv->ptype_lut_dqo = NULL;
 	gve_free_counter_array(priv);
 	gve_free_irq_db(priv);
 	gve_clear_device_resources_ok(priv);
@@ -997,8 +1005,25 @@ gve_setup_device_resources(struct gve_priv *priv)
 		PMD_DRV_LOG(ERR, "Could not config device resources: err=%d", err);
 		goto free_irq_dbs;
 	}
+
+	priv->ptype_lut_dqo = rte_zmalloc("gve_ptype_lut_dqo",
+			sizeof(struct gve_ptype_lut), 0);
+	if (priv->ptype_lut_dqo == NULL) {
+		PMD_DRV_LOG(ERR, "Failed to alloc ptype lut.");
+		err = -ENOMEM;
+		goto free_irq_dbs;
+	}
+	err = gve_adminq_get_ptype_map_dqo(priv, priv->ptype_lut_dqo);
+	if (unlikely(err)) {
+		PMD_DRV_LOG(ERR, "Failed to get ptype map: err=%d", err);
+		goto free_ptype_lut;
+	}
+
 	return 0;
 
+free_ptype_lut:
+	rte_free(priv->ptype_lut_dqo);
+	priv->ptype_lut_dqo = NULL;
 free_irq_dbs:
 	gve_free_irq_db(priv);
 free_cnt_array:
diff --git a/drivers/net/gve/gve_ethdev.h b/drivers/net/gve/gve_ethdev.h
index d713657d10..9b19fc55e3 100644
--- a/drivers/net/gve/gve_ethdev.h
+++ b/drivers/net/gve/gve_ethdev.h
@@ -36,6 +36,10 @@
 		RTE_MBUF_F_TX_L4_MASK  |	\
 		RTE_MBUF_F_TX_TCP_SEG)
 
+#define GVE_TX_CKSUM_OFFLOAD_MASK_DQO (		\
+		GVE_TX_CKSUM_OFFLOAD_MASK |	\
+		RTE_MBUF_F_TX_IP_CKSUM)
+
 #define GVE_RTE_RSS_OFFLOAD_ALL (	\
 	RTE_ETH_RSS_IPV4 |		\
 	RTE_ETH_RSS_NONFRAG_IPV4_TCP |	\
@@ -295,6 +299,7 @@ struct gve_priv {
 	uint16_t stats_end_idx; /* end index of array of stats written by NIC */
 
 	struct gve_rss_config rss_config;
+	struct gve_ptype_lut *ptype_lut_dqo;
 };
 
 static inline bool
diff --git a/drivers/net/gve/gve_rx_dqo.c b/drivers/net/gve/gve_rx_dqo.c
index 7c7a8c48d0..1c37c54cb7 100644
--- a/drivers/net/gve/gve_rx_dqo.c
+++ b/drivers/net/gve/gve_rx_dqo.c
@@ -75,6 +75,40 @@ gve_rx_refill_dqo(struct gve_rx_queue *rxq)
 	rxq->bufq_tail = next_avail;
 }
 
+static inline uint16_t
+gve_parse_csum_ol_flags(volatile struct gve_rx_compl_desc_dqo *rx_desc,
+		struct gve_priv *priv) {
+	uint64_t ol_flags = 0;
+	struct gve_ptype ptype =
+		priv->ptype_lut_dqo->ptypes[rx_desc->packet_type];
+
+	if(!rx_desc->l3_l4_processed)
+		return ol_flags;
+
+	if (ptype.l3_type == GVE_L3_TYPE_IPV4) {
+		if (rx_desc->csum_ip_err)
+			ol_flags |= RTE_MBUF_F_RX_IP_CKSUM_BAD;
+		else
+			ol_flags |= RTE_MBUF_F_RX_IP_CKSUM_GOOD;
+	}
+
+	if (rx_desc->csum_l4_err) {
+		ol_flags |= RTE_MBUF_F_RX_L4_CKSUM_BAD;
+		return ol_flags;
+	}
+	switch (ptype.l4_type) {
+	case GVE_L4_TYPE_TCP:
+	case GVE_L4_TYPE_UDP:
+	case GVE_L4_TYPE_ICMP:
+	case GVE_L4_TYPE_SCTP:
+		ol_flags |= RTE_MBUF_F_RX_L4_CKSUM_GOOD;
+		break;
+	default:
+		break;
+	}
+	return ol_flags;
+}
+
 uint16_t
 gve_rx_burst_dqo(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 {
@@ -125,8 +159,8 @@ gve_rx_burst_dqo(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 		rxm->data_len = pkt_len;
 		rxm->port = rxq->port_id;
 		rxm->ol_flags = 0;
-
-		rxm->ol_flags |= RTE_MBUF_F_RX_RSS_HASH;
+		rxm->ol_flags |= RTE_MBUF_F_RX_RSS_HASH |
+				gve_parse_csum_ol_flags(rx_desc, rxq->hw);
 		rxm->hash.rss = rte_be_to_cpu_32(rx_desc->hash);
 
 		rx_pkts[nb_rx++] = rxm;
diff --git a/drivers/net/gve/gve_tx_dqo.c b/drivers/net/gve/gve_tx_dqo.c
index 30a1455b20..a65e6aaf00 100644
--- a/drivers/net/gve/gve_tx_dqo.c
+++ b/drivers/net/gve/gve_tx_dqo.c
@@ -138,7 +138,7 @@ gve_tx_burst_dqo(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 		/* fill the last descriptor with End of Packet (EOP) bit */
 		txd->pkt.end_of_packet = 1;
 
-		if (ol_flags & GVE_TX_CKSUM_OFFLOAD_MASK)
+		if (ol_flags & GVE_TX_CKSUM_OFFLOAD_MASK_DQO)
 			txd->pkt.checksum_offload_enable = 1;
 
 		txq->nb_free -= nb_used;
-- 
2.34.1