[dpdk-dev] [PATCH 1/3] pmd: add new flag to indicate TX TSO operation on the packet

[dpdk-dev] [PATCH 1/3] pmd: add new flag to indicate TX TSO operation on the packet

[miroslaw.walukiewicz]

From: Miroslaw Walukiewicz <miroslaw.walukiewicz@intel.com>

Transmission of TCP packets could be accelerated by
HW Transmit Segmentation Offload. With TSO packets up to
64K could be transmismitted.

When this flag is set the PMD drived will enable TCP segmentation.

The new field tso_segsz is added to indicate how long is TCP TSO segment.

Signed-off-by: Mirek Walukiewicz <miroslaw.walukiewicz@intel.com>
---
 lib/librte_mbuf/rte_mbuf.h |    5 +++++
 1 file changed, 5 insertions(+)

diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h
index ddadc21..63cbc36 100644
--- a/lib/librte_mbuf/rte_mbuf.h
+++ b/lib/librte_mbuf/rte_mbuf.h
@@ -117,6 +117,9 @@ extern "C" {
 /* Use final bit of flags to indicate a control mbuf */
 #define CTRL_MBUF_FLAG       (1ULL << 63) /**< Mbuf contains control data */
 
+/* Bit 50 - TSO (TCP Transmit Segmenation Offload) */
+#define PKT_TX_TCP_TSO       (1ULL << 50) /**< Mbuf needs TSO enabling */
+
 /**
  * Bit Mask to indicate what bits required for building TX context
  */
@@ -196,6 +199,8 @@ struct rte_mbuf {
 			uint16_t l2_len:7;      /**< L2 (MAC) Header Length. */
 		};
 	};
+	/* field to support TSO segment size */
+	uint32_t tso_segsz;
 } __rte_cache_aligned;
 
 /**

[dpdk-dev] [PATCH 2/3] pmd: add new header containing TCP offload specific definitions

[miroslaw.walukiewicz]

From: Miroslaw Walukiewicz <miroslaw.walukiewicz@intel.com>

The function for computing initial TCP header checksum.
The file is common for both i40e and ixgbe PMD drivers

Signed-off-by: Mirek Walukiewicz <miroslaw.walukiewicz@intel.com>
---
 lib/librte_net/Makefile      |    3 +
 lib/librte_net/rte_tcp_off.h |  122 ++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 124 insertions(+), 1 deletion(-)
 create mode 100644 lib/librte_net/rte_tcp_off.h

diff --git a/lib/librte_net/Makefile b/lib/librte_net/Makefile
index ad2e482..83e76d1 100644
--- a/lib/librte_net/Makefile
+++ b/lib/librte_net/Makefile
@@ -34,7 +34,8 @@ include $(RTE_SDK)/mk/rte.vars.mk
 CFLAGS += $(WERROR_FLAGS) -I$(SRCDIR) -O3
 
 # install includes
-SYMLINK-$(CONFIG_RTE_LIBRTE_NET)-include := rte_ip.h rte_tcp.h rte_udp.h rte_sctp.h rte_icmp.h rte_arp.h
+SYMLINK-$(CONFIG_RTE_LIBRTE_NET)-include := rte_ip.h rte_tcp.h rte_udp.h rte_sctp.h rte_icmp.h rte_arp.h \
+rte_tcp_off.h
 
 
 include $(RTE_SDK)/mk/rte.install.mk
diff --git a/lib/librte_net/rte_tcp_off.h b/lib/librte_net/rte_tcp_off.h
new file mode 100644
index 0000000..6143396
--- /dev/null
+++ b/lib/librte_net/rte_tcp_off.h
@@ -0,0 +1,122 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/*
+ * Copyright (c) 1982, 1986, 1990, 1993
+ *      The Regents of the University of California.  All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. All advertising materials mentioning features or use of this software
+ *    must display the following acknowledgement:
+ *      This product includes software developed by the University of
+ *      California, Berkeley and its contributors.
+ * 4. Neither the name of the University nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ *      @(#)in.h        8.3 (Berkeley) 1/3/94
+ * $FreeBSD: src/sys/netinet/in.h,v 1.82 2003/10/25 09:37:10 ume Exp $
+ */
+
+#ifndef _RTE_TCP_OFF_H_
+#define _RTE_TCP_OFF_H_
+
+/**
+ * @file
+ *
+ * TCP offload -related defines
+ */
+
+#include <stdint.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/**
+ * function computes the TCP pseudo checksum for TSO
+ * it is a commnon function for i40e/ixgbe drivers
+ * it computes the IP header initial checksum as input 
+ * for computing TCP data checksum made by HW 
+ */
+static inline u_short
+rte_in_pseudo(uint32_t src_ipv4, uint32_t dst_ipv4, uint32_t proto)
+{
+	uint64_t sum;
+	union l_val {
+		uint16_t s[2];
+		uint32_t l;
+	} l_val;
+	union q_val {
+		uint16_t s[4];
+		uint32_t l[2];
+		uint64_t q;
+	} q_val;
+
+	sum = (uint64_t) src_ipv4 + dst_ipv4 + proto;
+
+	q_val.q = sum;	
+	l_val.l = q_val.s[0] + q_val.s[1] + q_val.s[2] + q_val.s[3];
+	sum = l_val.s[0] + l_val.s[1];
+
+	sum > 65535 ? sum -= 65535 : sum;
+
+	return (sum);
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+
+#endif /* _RTE_TCP_OFF_H_ */

[dpdk-dev] [PATCH 3/3] pmd i40e: Enable Transmit Segmentation Offload for TCP traffic

[miroslaw.walukiewicz]

From: Miroslaw Walukiewicz <miroslaw.walukiewicz@intel.com>

The patch enables the TSO HW feature for i40e PMD driver.

The feature is reported by rte_dev_info_get() if enabled.

Signed-off-by: Mirek Walukiewicz <miroslaw.walukiewicz@intel.com>
---
 lib/librte_pmd_i40e/i40e_ethdev.c |    1 +
 lib/librte_pmd_i40e/i40e_rxtx.c   |   56 ++++++++++++++++++++++++++++++++++---
 2 files changed, 53 insertions(+), 4 deletions(-)

diff --git a/lib/librte_pmd_i40e/i40e_ethdev.c b/lib/librte_pmd_i40e/i40e_ethdev.c
index 46c43a7..01b21eb 100644
--- a/lib/librte_pmd_i40e/i40e_ethdev.c
+++ b/lib/librte_pmd_i40e/i40e_ethdev.c
@@ -1399,6 +1399,7 @@ i40e_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 		DEV_TX_OFFLOAD_IPV4_CKSUM |
 		DEV_TX_OFFLOAD_UDP_CKSUM |
 		DEV_TX_OFFLOAD_TCP_CKSUM |
+		DEV_TX_OFFLOAD_TCP_TSO |
 		DEV_TX_OFFLOAD_SCTP_CKSUM;
 
 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
diff --git a/lib/librte_pmd_i40e/i40e_rxtx.c b/lib/librte_pmd_i40e/i40e_rxtx.c
index 2b53677..bc7af2b 100644
--- a/lib/librte_pmd_i40e/i40e_rxtx.c
+++ b/lib/librte_pmd_i40e/i40e_rxtx.c
@@ -50,6 +50,8 @@
 #include <rte_tcp.h>
 #include <rte_sctp.h>
 #include <rte_udp.h>
+#include <rte_ip.h>
+#include <rte_tcp_off.h>
 
 #include "i40e_logs.h"
 #include "i40e/i40e_prototype.h"
@@ -440,6 +442,11 @@ i40e_txd_enable_checksum(uint32_t ol_flags,
 		*td_offset |= (l3_len >> 2) << I40E_TX_DESC_LENGTH_IPLEN_SHIFT;
 	}
 
+	if (ol_flags & PKT_TX_TCP_TSO) {
+		*td_cmd |= I40E_TX_DESC_CMD_L4T_EOFT_TCP;
+		/* td offset will be set next */
+		return;
+	}
 	/* Enable L4 checksum offloads */
 	switch (ol_flags & PKT_TX_L4_MASK) {
 	case PKT_TX_TCP_CKSUM:
@@ -1073,12 +1080,46 @@ i40e_calc_context_desc(uint64_t flags)
 #ifdef RTE_LIBRTE_IEEE1588
 	mask |= PKT_TX_IEEE1588_TMST;
 #endif
+	/* need for context descriptor when TSO enabled */
+	mask |= PKT_TX_TCP_TSO;
 	if (flags & mask)
 		return 1;
 
 	return 0;
 }
 
+/* set i40e TSO context descriptor */
+static inline uint64_t 
+i40e_set_tso_ctx(struct rte_mbuf *mbuf, uint8_t l2_len, uint8_t l3_len, uint32_t *td_offset)
+{
+	uint64_t ctx_desc;
+	struct ipv4_hdr *ip;
+	struct tcp_hdr *th;
+	uint32_t tcp_hlen;
+	uint32_t hdrlen;
+	uint32_t paylen;
+	
+	/* set mss */
+	ip = (struct ipv4_hdr *) (rte_pktmbuf_mtod(mbuf, unsigned char *) + l2_len);
+	ip->hdr_checksum = 0;
+	ip->total_length = 0;
+	th = (struct tcp_hdr *)((caddr_t)ip + l3_len);
+	th->cksum = rte_in_pseudo(ip->src_addr, ip->dst_addr, I40E_HTONS(IPPROTO_TCP));
+	tcp_hlen = (th->data_off >> 4) << 2;
+	*td_offset |= (tcp_hlen >> 2) <<
+				I40E_TX_DESC_LENGTH_L4_FC_LEN_SHIFT;
+	hdrlen = l2_len + l3_len + tcp_hlen;
+	paylen = mbuf->pkt_len - hdrlen;
+
+	ctx_desc = ((uint64_t)mbuf->tso_segsz << 
+						I40E_TXD_CTX_QW1_MSS_SHIFT) |
+		    ((uint64_t)paylen << I40E_TXD_CTX_QW1_TSO_LEN_SHIFT) |
+		    ((uint64_t)I40E_TX_CTX_DESC_TSO <<
+						I40E_TXD_CTX_QW1_CMD_SHIFT);
+
+	return ctx_desc;
+}
+
 uint16_t
 i40e_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 {
@@ -1192,12 +1233,19 @@ i40e_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 				rte_pktmbuf_free_seg(txe->mbuf);
 				txe->mbuf = NULL;
 			}
+			/* TSO enabled means no timestamp */
+			if (ol_flags & PKT_TX_TCP_TSO) {
+				cd_type_cmd_tso_mss |= 
+					i40e_set_tso_ctx(tx_pkt, l2_len, l3_len, &td_offset);
+			}
+			else {	
 #ifdef RTE_LIBRTE_IEEE1588
-			if (ol_flags & PKT_TX_IEEE1588_TMST)
-				cd_type_cmd_tso_mss |=
-					((uint64_t)I40E_TX_CTX_DESC_TSYN <<
-						I40E_TXD_CTX_QW1_CMD_SHIFT);
+				if (ol_flags & PKT_TX_IEEE1588_TMST)
+					cd_type_cmd_tso_mss |=
+						((uint64_t)I40E_TX_CTX_DESC_TSYN <<
+							I40E_TXD_CTX_QW1_CMD_SHIFT);
 #endif
+			}
 			ctx_txd->tunneling_params =
 				rte_cpu_to_le_32(cd_tunneling_params);
 			ctx_txd->l2tag2 = rte_cpu_to_le_16(cd_l2tag2);

Re: [dpdk-dev] [PATCH 1/3] pmd: add new flag to indicate TX TSO operation on the packet

[Liu, Jijiang]

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of
> miroslaw.walukiewicz@intel.com
> Sent: Monday, October 13, 2014 10:39 PM
> To: dev@dpdk.org
> Subject: [dpdk-dev] [PATCH 1/3] pmd: add new flag to indicate TX TSO operation
> on the packet
> 
> From: Miroslaw Walukiewicz <miroslaw.walukiewicz@intel.com>
> 
> Transmission of TCP packets could be accelerated by HW Transmit Segmentation
> Offload. With TSO packets up to 64K could be transmismitted.
> 
> When this flag is set the PMD drived will enable TCP segmentation.
> 
> The new field tso_segsz is added to indicate how long is TCP TSO segment.
> 
> Signed-off-by: Mirek Walukiewicz <miroslaw.walukiewicz@intel.com>
> ---
>  lib/librte_mbuf/rte_mbuf.h |    5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h index
> ddadc21..63cbc36 100644
> --- a/lib/librte_mbuf/rte_mbuf.h
> +++ b/lib/librte_mbuf/rte_mbuf.h
> @@ -117,6 +117,9 @@ extern "C" {
>  /* Use final bit of flags to indicate a control mbuf */
>  #define CTRL_MBUF_FLAG       (1ULL << 63) /**< Mbuf contains control data */
> 
> +/* Bit 50 - TSO (TCP Transmit Segmenation Offload) */
> +#define PKT_TX_TCP_TSO       (1ULL << 50) /**< Mbuf needs TSO enabling */


In VxLAN patch set  [PATCH v5 7/8]i40e:support VxLAN Tx checksum offload, the  bit (1ULL << 50)   in offload flag have already been used.
+#define PKT_TX_VXLAN_CKSUM   (1ULL << 50) /**< TX checksum of VxLAN computed by NIC */

>  /**
>   * Bit Mask to indicate what bits required for building TX context
>   */
> @@ -196,6 +199,8 @@ struct rte_mbuf {
>  			uint16_t l2_len:7;      /**< L2 (MAC) Header Length. */
>  		};
>  	};
> +	/* field to support TSO segment size */
> +	uint32_t tso_segsz;
>  } __rte_cache_aligned;
> 
>  /**

Re: [dpdk-dev] [PATCH 2/3] pmd: add new header containing TCP offload specific definitions

[Thomas Monjalon]

Hi Miroslaw,

2014-10-13 10:38, miroslaw.walukiewicz@intel.com:
> From: Miroslaw Walukiewicz <miroslaw.walukiewicz@intel.com>
> 
> The function for computing initial TCP header checksum.
> The file is common for both i40e and ixgbe PMD drivers
> 
> Signed-off-by: Mirek Walukiewicz <miroslaw.walukiewicz@intel.com>
> ---
>  lib/librte_net/Makefile      |    3 +
>  lib/librte_net/rte_tcp_off.h |  122 ++++++++++++++++++++++++++++++++++++++++++

There is a work in progress about TSO:
	http://thread.gmane.org/gmane.comp.networking.dpdk.devel/2541/focus=2553
This API was discussed and implemented for ixgbe.
Then there was an agreement to rework mbuf in 1.8 version to implement TSO
on top of it. Then it was suggested to rebase his work:
	http://dpdk.org/ml/archives/dev/2014-September/006151.html

How does your patchset match with Olivier's work on TSO and ixgbe driver?
I would like to ensure the API works fine for all TSO-capable driver,
this would be a good start.

Thanks
-- 
Thomas