[PATCH v3 2/2] app/testpmd: support TCP TSO in Tx only mode

[Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>]

Add '--txonly-tso-mss=N' option that enables TSO offload
and generates packets with specified MSS in txonly mode.

Signed-off-by: Georgiy Levashov <georgiy.levashov@oktetlabs.ru>
Signed-off-by: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
---
 app/test-pmd/parameters.c             | 10 ++++++++
 app/test-pmd/testpmd.c                | 12 ++++++++++
 app/test-pmd/testpmd.h                |  1 +
 app/test-pmd/txonly.c                 | 34 ++++++++++++++++++++++++++-
 doc/guides/testpmd_app_ug/run_app.rst |  4 ++++
 5 files changed, 60 insertions(+), 1 deletion(-)

diff --git a/app/test-pmd/parameters.c b/app/test-pmd/parameters.c
index 2ed8afedfd..e71cb3e139 100644
--- a/app/test-pmd/parameters.c
+++ b/app/test-pmd/parameters.c
@@ -156,6 +156,7 @@ usage(char* progname)
 	printf("  --txpkts=X[,Y]*: set TX segment sizes"
 		" or total packet length.\n");
 	printf("  --txonly-multi-flow: generate multiple flows in txonly mode\n");
+	printf("  --txonly-tso-mss=N: enable TSO offload and generate packets with specified MSS in txonly mode\n");
 	printf("  --tx-ip=src,dst: IP addresses in Tx-only mode\n");
 	printf("  --tx-udp=src[,dst]: UDP ports in Tx-only mode\n");
 	printf("  --eth-link-speed: force link speed.\n");
@@ -670,6 +671,7 @@ launch_args_parse(int argc, char** argv)
 		{ "rxhdrs",			1, 0, 0 },
 		{ "txpkts",			1, 0, 0 },
 		{ "txonly-multi-flow",		0, 0, 0 },
+		{ "txonly-tso-mss",		1, 0, 0 },
 		{ "rxq-share",			2, 0, 0 },
 		{ "eth-link-speed",		1, 0, 0 },
 		{ "disable-link-check",		0, 0, 0 },
@@ -1297,6 +1299,14 @@ launch_args_parse(int argc, char** argv)
 			}
 			if (!strcmp(lgopts[opt_idx].name, "txonly-multi-flow"))
 				txonly_multi_flow = 1;
+			if (!strcmp(lgopts[opt_idx].name, "txonly-tso-mss")) {
+				n = atoi(optarg);
+				if (n >= 0 && n <= UINT16_MAX)
+					txonly_tso_segsz = n;
+				else
+					rte_exit(EXIT_FAILURE,
+						 "TSO MSS must be >= 0 and <= UINT16_MAX\n");
+			}
 			if (!strcmp(lgopts[opt_idx].name, "rxq-share")) {
 				if (optarg == NULL) {
 					rxq_share = UINT32_MAX;
diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c
index ef281ccd20..94d37be692 100644
--- a/app/test-pmd/testpmd.c
+++ b/app/test-pmd/testpmd.c
@@ -264,6 +264,9 @@ enum tx_pkt_split tx_pkt_split = TX_PKT_SPLIT_OFF;
 uint8_t txonly_multi_flow;
 /**< Whether multiple flows are generated in TXONLY mode. */
 
+uint16_t txonly_tso_segsz;
+/**< TSO MSS for generated packets in TXONLY mode. */
+
 uint32_t tx_pkt_times_inter;
 /**< Timings for send scheduling in TXONLY mode, time between bursts. */
 
@@ -1615,6 +1618,15 @@ init_config_port_offloads(portid_t pid, uint32_t socket_id)
 		port->dev_conf.txmode.offloads &=
 			~RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
 
+	if (txonly_tso_segsz > 0) {
+		if ((ports[pid].dev_info.tx_offload_capa &
+		    RTE_ETH_TX_OFFLOAD_TCP_TSO) == 0) {
+			rte_exit(EXIT_FAILURE,
+				 "TSO isn't supported for port %d\n", pid);
+		}
+		port->dev_conf.txmode.offloads |= RTE_ETH_TX_OFFLOAD_TCP_TSO;
+	}
+
 	/* Apply Rx offloads configuration */
 	for (i = 0; i < port->dev_info.max_rx_queues; i++)
 		port->rxq[i].conf.offloads = port->dev_conf.rxmode.offloads;
diff --git a/app/test-pmd/testpmd.h b/app/test-pmd/testpmd.h
index 976f4f83dd..fbe1839a8f 100644
--- a/app/test-pmd/testpmd.h
+++ b/app/test-pmd/testpmd.h
@@ -611,6 +611,7 @@ enum tx_pkt_split {
 extern enum tx_pkt_split tx_pkt_split;
 
 extern uint8_t txonly_multi_flow;
+extern uint16_t txonly_tso_segsz;
 
 extern uint32_t rxq_share;
 
diff --git a/app/test-pmd/txonly.c b/app/test-pmd/txonly.c
index b304bd4bf8..59fdb5f953 100644
--- a/app/test-pmd/txonly.c
+++ b/app/test-pmd/txonly.c
@@ -60,6 +60,7 @@ RTE_DEFINE_PER_LCORE(uint8_t, _ip_var); /**< IP address variation */
 
 static union pkt_l4_hdr_t {
 	struct rte_udp_hdr udp;	/**< UDP header of tx packets. */
+	struct rte_tcp_hdr tcp; /**< TCP header of tx packets. */
 } pkt_l4_hdr; /**< Layer 4 header of tx packets. */
 
 static uint64_t timestamp_mask; /**< Timestamp dynamic flag mask */
@@ -112,8 +113,19 @@ setup_pkt_l4_ip_headers(uint8_t ip_proto, struct rte_ipv4_hdr *ip_hdr,
 	uint32_t ip_cksum;
 	uint16_t pkt_len;
 	struct rte_udp_hdr *udp_hdr;
+	struct rte_tcp_hdr *tcp_hdr;
 
 	switch (ip_proto) {
+	case IPPROTO_TCP:
+		/*
+		 * Initialize TCP header.
+		 */
+		pkt_len = (uint16_t)(pkt_data_len + sizeof(struct rte_tcp_hdr));
+		tcp_hdr = &l4_hdr->tcp;
+		tcp_hdr->src_port = rte_cpu_to_be_16(tx_l4_src_port);
+		tcp_hdr->dst_port = rte_cpu_to_be_16(tx_l4_dst_port);
+		tcp_hdr->data_off = (sizeof(struct rte_tcp_hdr) << 2) & 0xF0;
+		break;
 	case IPPROTO_UDP:
 		/*
 		 * Initialize UDP header.
@@ -189,6 +201,8 @@ update_pkt_header(struct rte_mbuf *pkt, uint32_t total_pkt_len)
 	ip_hdr->hdr_checksum = rte_ipv4_cksum(ip_hdr);
 
 	switch (ip_hdr->next_proto_id) {
+	case IPPROTO_TCP:
+		break;
 	case IPPROTO_UDP:
 		/* update UDP packet length */
 		udp_hdr = rte_pktmbuf_mtod_offset(pkt, struct rte_udp_hdr *,
@@ -232,6 +246,12 @@ pkt_burst_prepare(struct rte_mbuf *pkt, struct rte_mempool *mbp,
 	pkt->l2_len = sizeof(struct rte_ether_hdr);
 	pkt->l3_len = sizeof(struct rte_ipv4_hdr);
 
+	if (txonly_tso_segsz > 0) {
+		pkt->tso_segsz = txonly_tso_segsz;
+		pkt->ol_flags |= RTE_MBUF_F_TX_TCP_SEG | RTE_MBUF_F_TX_IPV4 |
+				 RTE_MBUF_F_TX_IP_CKSUM;
+	}
+
 	pkt_len = pkt->data_len;
 	pkt_seg = pkt;
 	for (i = 1; i < nb_segs; i++) {
@@ -267,6 +287,12 @@ pkt_burst_prepare(struct rte_mbuf *pkt, struct rte_mempool *mbp,
 		RTE_PER_LCORE(_ip_var) = ip_var;
 	}
 	switch (ip_hdr->next_proto_id) {
+	case IPPROTO_TCP:
+		copy_buf_to_pkt(&pkt_l4_hdr.tcp, sizeof(pkt_l4_hdr.tcp), pkt,
+				sizeof(struct rte_ether_hdr) +
+				sizeof(struct rte_ipv4_hdr));
+		l4_hdr_size = sizeof(pkt_l4_hdr.tcp);
+		break;
 	case IPPROTO_UDP:
 		copy_buf_to_pkt(&pkt_l4_hdr.udp, sizeof(pkt_l4_hdr.udp), pkt,
 				sizeof(struct rte_ether_hdr) +
@@ -277,6 +303,7 @@ pkt_burst_prepare(struct rte_mbuf *pkt, struct rte_mempool *mbp,
 		l4_hdr_size = 0;
 		break;
 	}
+	pkt->l4_len = l4_hdr_size;
 
 	if (unlikely(tx_pkt_split == TX_PKT_SPLIT_RND) || txonly_multi_flow)
 		update_pkt_header(pkt, pkt_len);
@@ -459,11 +486,16 @@ tx_only_begin(portid_t pi)
 {
 	uint16_t pkt_hdr_len, pkt_data_len;
 	int dynf;
-	uint8_t ip_proto = IPPROTO_UDP;
+	uint8_t ip_proto;
 
 	pkt_hdr_len = (uint16_t)(sizeof(struct rte_ether_hdr) +
 				 sizeof(struct rte_ipv4_hdr));
+
+	ip_proto = txonly_tso_segsz > 0 ? IPPROTO_TCP : IPPROTO_UDP;
 	switch (ip_proto) {
+	case IPPROTO_TCP:
+		pkt_hdr_len += sizeof(struct rte_tcp_hdr);
+		break;
 	case IPPROTO_UDP:
 		pkt_hdr_len += sizeof(struct rte_udp_hdr);
 		break;
diff --git a/doc/guides/testpmd_app_ug/run_app.rst b/doc/guides/testpmd_app_ug/run_app.rst
index 610e442924..01d6852fd3 100644
--- a/doc/guides/testpmd_app_ug/run_app.rst
+++ b/doc/guides/testpmd_app_ug/run_app.rst
@@ -369,6 +369,10 @@ The command line options are:
 
     Generate multiple flows in txonly mode.
 
+*   ``--txonly-tso-mss=N```
+
+    Enable TSO offload and generate TCP packets with specified MSS in txonly mode.
+
 *   ``--rxq-share=[X]``
 
     Create queues in shared Rx queue mode if device supports.
-- 
2.30.2