From bc4b30bb7c3ce3310d98a43bb5364b327ecd5cf9 Mon Sep 17 00:00:00 2001
From: Dongdong Liu <liudongdong3@huawei.com>
Date: Fri, 21 Apr 2023 17:53:21 +0800
Subject: net/hns3: simplify hardware checksum offloading

[ upstream commit 7fd763a9dd18a3edd98e95c26e96349fd71cbb9b ]

If the NIC support simple BD mode, the hardware will calculate
the checksum from the start position of checksum and fill the
checksum result to the offset position, which simple the
HW operations of calculating the type and header length of
L3/L4.

Add this mode for hns3 PMD when the packet type is L4.

Signed-off-by: Dongdong Liu <liudongdong3@huawei.com>
---
 drivers/net/hns3/hns3_cmd.c    |  3 ++
 drivers/net/hns3/hns3_cmd.h    |  1 +
 drivers/net/hns3/hns3_dump.c   |  1 +
 drivers/net/hns3/hns3_ethdev.h |  1 +
 drivers/net/hns3/hns3_rxtx.c   | 52 +++++++++++++++++++++++++++++++++-
 drivers/net/hns3/hns3_rxtx.h   | 12 +++++++-
 6 files changed, 68 insertions(+), 2 deletions(-)

diff --git a/drivers/net/hns3/hns3_cmd.c b/drivers/net/hns3/hns3_cmd.c
index bdfc85f934..d530650452 100644
--- a/drivers/net/hns3/hns3_cmd.c
+++ b/drivers/net/hns3/hns3_cmd.c
@@ -419,6 +419,7 @@ hns3_get_caps_name(uint32_t caps_id)
 	} dev_caps[] = {
 		{ HNS3_CAPS_FD_QUEUE_REGION_B, "fd_queue_region" },
 		{ HNS3_CAPS_PTP_B,             "ptp"             },
+		{ HNS3_CAPS_SIMPLE_BD_B,       "simple_bd"       },
 		{ HNS3_CAPS_TX_PUSH_B,         "tx_push"         },
 		{ HNS3_CAPS_PHY_IMP_B,         "phy_imp"         },
 		{ HNS3_CAPS_TQP_TXRX_INDEP_B,  "tqp_txrx_indep"  },
@@ -489,6 +490,8 @@ hns3_parse_capability(struct hns3_hw *hw,
 			hns3_warn(hw, "ignore PTP capability due to lack of "
 				  "rxd advanced layout capability.");
 	}
+	if (hns3_get_bit(caps, HNS3_CAPS_SIMPLE_BD_B))
+		hns3_set_bit(hw->capability, HNS3_DEV_SUPPORT_SIMPLE_BD_B, 1);
 	if (hns3_get_bit(caps, HNS3_CAPS_TX_PUSH_B))
 		hns3_set_bit(hw->capability, HNS3_DEV_SUPPORT_TX_PUSH_B, 1);
 	if (hns3_get_bit(caps, HNS3_CAPS_PHY_IMP_B))
diff --git a/drivers/net/hns3/hns3_cmd.h b/drivers/net/hns3/hns3_cmd.h
index eb394c9dec..4abe0f1d13 100644
--- a/drivers/net/hns3/hns3_cmd.h
+++ b/drivers/net/hns3/hns3_cmd.h
@@ -313,6 +313,7 @@ enum HNS3_CAPS_BITS {
 	 */
 	HNS3_CAPS_FD_QUEUE_REGION_B = 2,
 	HNS3_CAPS_PTP_B,
+	HNS3_CAPS_SIMPLE_BD_B = 5,
 	HNS3_CAPS_TX_PUSH_B = 6,
 	HNS3_CAPS_PHY_IMP_B = 7,
 	HNS3_CAPS_TQP_TXRX_INDEP_B,
diff --git a/drivers/net/hns3/hns3_dump.c b/drivers/net/hns3/hns3_dump.c
index 8268506f6f..a793ba64ad 100644
--- a/drivers/net/hns3/hns3_dump.c
+++ b/drivers/net/hns3/hns3_dump.c
@@ -96,6 +96,7 @@ hns3_get_dev_feature_capability(FILE *file, struct hns3_hw *hw)
 		{HNS3_DEV_SUPPORT_TX_PUSH_B, "TX PUSH"},
 		{HNS3_DEV_SUPPORT_INDEP_TXRX_B, "INDEP TXRX"},
 		{HNS3_DEV_SUPPORT_STASH_B, "STASH"},
+		{HNS3_DEV_SUPPORT_SIMPLE_BD_B, "SIMPLE BD"},
 		{HNS3_DEV_SUPPORT_RXD_ADV_LAYOUT_B, "RXD Advanced Layout"},
 		{HNS3_DEV_SUPPORT_OUTER_UDP_CKSUM_B, "OUTER UDP CKSUM"},
 		{HNS3_DEV_SUPPORT_RAS_IMP_B, "RAS IMP"},
diff --git a/drivers/net/hns3/hns3_ethdev.h b/drivers/net/hns3/hns3_ethdev.h
index 9acc5a3d7e..ee4dd18d7b 100644
--- a/drivers/net/hns3/hns3_ethdev.h
+++ b/drivers/net/hns3/hns3_ethdev.h
@@ -886,6 +886,7 @@ enum hns3_dev_cap {
 	HNS3_DEV_SUPPORT_TX_PUSH_B,
 	HNS3_DEV_SUPPORT_INDEP_TXRX_B,
 	HNS3_DEV_SUPPORT_STASH_B,
+	HNS3_DEV_SUPPORT_SIMPLE_BD_B,
 	HNS3_DEV_SUPPORT_RXD_ADV_LAYOUT_B,
 	HNS3_DEV_SUPPORT_OUTER_UDP_CKSUM_B,
 	HNS3_DEV_SUPPORT_RAS_IMP_B,
diff --git a/drivers/net/hns3/hns3_rxtx.c b/drivers/net/hns3/hns3_rxtx.c
index 1f44c0345f..aaf0a06ca6 100644
--- a/drivers/net/hns3/hns3_rxtx.c
+++ b/drivers/net/hns3/hns3_rxtx.c
@@ -3046,6 +3046,10 @@ hns3_tx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t nb_desc,
 					HNS3_PORT_BASE_VLAN_ENABLE;
 	else
 		txq->pvid_sw_shift_en = false;
+
+	if (hns3_dev_get_support(hw, SIMPLE_BD))
+		txq->simple_bd_enable = true;
+
 	txq->max_non_tso_bd_num = hw->max_non_tso_bd_num;
 	txq->configured = true;
 	txq->io_base = (void *)((char *)hw->io_base +
@@ -3162,7 +3166,7 @@ hns3_set_tso(struct hns3_desc *desc, uint32_t paylen, struct rte_mbuf *rxm)
 		return;
 
 	desc->tx.type_cs_vlan_tso_len |= rte_cpu_to_le_32(BIT(HNS3_TXD_TSO_B));
-	desc->tx.mss = rte_cpu_to_le_16(rxm->tso_segsz);
+	desc->tx.ckst_mss |= rte_cpu_to_le_16(rxm->tso_segsz);
 }
 
 static inline void
@@ -3901,6 +3905,50 @@ hns3_prep_pkts(__rte_unused void *tx_queue, struct rte_mbuf **tx_pkts,
 	return i;
 }
 
+static inline int
+hns3_handle_simple_bd(struct hns3_tx_queue *txq, struct hns3_desc *desc,
+		      struct rte_mbuf *m)
+{
+#define HNS3_TCP_CSUM_OFFSET	16
+#define HNS3_UDP_CSUM_OFFSET	6
+
+	/*
+	 * In HIP09, NIC HW support Tx simple BD mode that the HW will
+	 * calculate the checksum from the start position of checksum and fill
+	 * the checksum result to the offset position without packet type and
+	 * header length of L3/L4.
+	 * For non-tunneling packet:
+	 * - Tx simple BD support for TCP and UDP checksum.
+	 * For tunneling packet:
+	 * - Tx simple BD support for inner L4 checksum(except sctp checksum).
+	 * - Tx simple BD not support the outer checksum and the inner L3
+	 *   checksum.
+	 * - Besides, Tx simple BD is not support for TSO.
+	 */
+	if (txq->simple_bd_enable && !(m->ol_flags & RTE_MBUF_F_TX_IP_CKSUM) &&
+	    !(m->ol_flags & RTE_MBUF_F_TX_TCP_SEG) &&
+	    !(m->ol_flags & RTE_MBUF_F_TX_OUTER_IP_CKSUM) &&
+	    ((m->ol_flags & RTE_MBUF_F_TX_L4_MASK) == RTE_MBUF_F_TX_TCP_CKSUM ||
+	    (m->ol_flags & RTE_MBUF_F_TX_L4_MASK) == RTE_MBUF_F_TX_UDP_CKSUM)) {
+		/* set checksum start and offset, defined in 2 Bytes */
+		hns3_set_field(desc->tx.type_cs_vlan_tso_len,
+			       HNS3_TXD_L4_START_M, HNS3_TXD_L4_START_S,
+			       (m->l2_len + m->l3_len) >> HNS3_SIMPLE_BD_UNIT);
+		hns3_set_field(desc->tx.ol_type_vlan_len_msec,
+			   HNS3_TXD_L4_CKS_OFFSET_M, HNS3_TXD_L4_CKS_OFFSET_S,
+			   (m->ol_flags & RTE_MBUF_F_TX_L4_MASK) ==
+			   RTE_MBUF_F_TX_TCP_CKSUM ?
+			   HNS3_TCP_CSUM_OFFSET >> HNS3_SIMPLE_BD_UNIT :
+			   HNS3_UDP_CSUM_OFFSET >> HNS3_SIMPLE_BD_UNIT);
+
+		hns3_set_bit(desc->tx.ckst_mss, HNS3_TXD_CKST_B, 1);
+
+		return 0;
+	}
+
+	return -ENOTSUP;
+}
+
 static int
 hns3_parse_cksum(struct hns3_tx_queue *txq, uint16_t tx_desc_id,
 		 struct rte_mbuf *m)
@@ -3910,6 +3958,8 @@ hns3_parse_cksum(struct hns3_tx_queue *txq, uint16_t tx_desc_id,
 
 	/* Enable checksum offloading */
 	if (m->ol_flags & HNS3_TX_CKSUM_OFFLOAD_MASK) {
+		if (hns3_handle_simple_bd(txq, desc, m) == 0)
+			return 0;
 		/* Fill in tunneling parameters if necessary */
 		if (hns3_parse_tunneling_params(txq, m, tx_desc_id)) {
 			txq->dfx_stats.unsupported_tunnel_pkt_cnt++;
diff --git a/drivers/net/hns3/hns3_rxtx.h b/drivers/net/hns3/hns3_rxtx.h
index fa39f6481a..7685ac2ea3 100644
--- a/drivers/net/hns3/hns3_rxtx.h
+++ b/drivers/net/hns3/hns3_rxtx.h
@@ -134,6 +134,9 @@
 #define HNS3_TXD_L4LEN_S			24
 #define HNS3_TXD_L4LEN_M			(0xffUL << HNS3_TXD_L4LEN_S)
 
+#define HNS3_TXD_L4_START_S			8
+#define HNS3_TXD_L4_START_M			(0xffff << HNS3_TXD_L4_START_S)
+
 #define HNS3_TXD_OL3T_S				0
 #define HNS3_TXD_OL3T_M				(0x3 << HNS3_TXD_OL3T_S)
 #define HNS3_TXD_OVLAN_B			2
@@ -141,6 +144,9 @@
 #define HNS3_TXD_TUNTYPE_S			4
 #define HNS3_TXD_TUNTYPE_M			(0xf << HNS3_TXD_TUNTYPE_S)
 
+#define HNS3_TXD_L4_CKS_OFFSET_S		8
+#define HNS3_TXD_L4_CKS_OFFSET_M	(0xffff << HNS3_TXD_L4_CKS_OFFSET_S)
+
 #define HNS3_TXD_BDTYPE_S			0
 #define HNS3_TXD_BDTYPE_M			(0xf << HNS3_TXD_BDTYPE_S)
 #define HNS3_TXD_FE_B				4
@@ -157,10 +163,13 @@
 #define HNS3_TXD_MSS_S				0
 #define HNS3_TXD_MSS_M				(0x3fff << HNS3_TXD_MSS_S)
 
+#define HNS3_TXD_CKST_B				14
+
 #define HNS3_TXD_OL4CS_B			22
 #define HNS3_L2_LEN_UNIT			1UL
 #define HNS3_L3_LEN_UNIT			2UL
 #define HNS3_L4_LEN_UNIT			2UL
+#define HNS3_SIMPLE_BD_UNIT			1UL
 
 #define HNS3_TXD_DEFAULT_BDTYPE		0
 #define HNS3_TXD_VLD_CMD		(0x1 << HNS3_TXD_VLD_B)
@@ -247,7 +256,7 @@ struct hns3_desc {
 
 			uint32_t paylen_fd_dop_ol4cs;
 			uint16_t tp_fe_sc_vld_ra_ri;
-			uint16_t mss;
+			uint16_t ckst_mss;
 		} tx;
 
 		struct {
@@ -488,6 +497,7 @@ struct hns3_tx_queue {
 	 */
 	uint16_t udp_cksum_mode:1;
 
+	/* check whether the simple BD mode is supported */
 	uint16_t simple_bd_enable:1;
 	uint16_t tx_push_enable:1;    /* check whether the tx push is enabled */
 	/*
-- 
2.23.0