Re: [PATCH v5 2/6] gso: add TCP/IPv4 GSO support

[Jiayu Hu <jiayu.hu@intel.com>]

Hi Mark,

One comment is inline.

Thanks,
Jiayu

On Thu, Sep 28, 2017 at 11:13:49PM +0100, Mark Kavanagh wrote:
> From: Jiayu Hu <jiayu.hu@intel.com>
> 
> This patch adds GSO support for TCP/IPv4 packets. Supported packets
> may include a single VLAN tag. TCP/IPv4 GSO doesn't check if input
> packets have correct checksums, and doesn't update checksums for
> output packets (the responsibility for this lies with the application).
> Additionally, TCP/IPv4 GSO doesn't process IP fragmented packets.
> 
> TCP/IPv4 GSO uses two chained MBUFs, one direct MBUF and one indrect
> MBUF, to organize an output packet. Note that we refer to these two
> chained MBUFs as a two-segment MBUF. The direct MBUF stores the packet
> header, while the indirect mbuf simply points to a location within the
> original packet's payload. Consequently, use of the GSO library requires
> multi-segment MBUF support in the TX functions of the NIC driver.
> 
> If a packet is GSO'd, TCP/IPv4 GSO reduces its MBUF refcnt by 1. As a
> result, when all of its GSOed segments are freed, the packet is freed
> automatically.
> 
> Signed-off-by: Jiayu Hu <jiayu.hu@intel.com>
> Signed-off-by: Mark Kavanagh <mark.b.kavanagh@intel.com>
> Tested-by: Lei Yao <lei.a.yao@intel.com>
> ---
>  doc/guides/rel_notes/release_17_11.rst  |  12 +++
>  lib/librte_eal/common/include/rte_log.h |   1 +
>  lib/librte_gso/Makefile                 |   2 +
>  lib/librte_gso/gso_common.c             | 153 ++++++++++++++++++++++++++++++++
>  lib/librte_gso/gso_common.h             | 141 +++++++++++++++++++++++++++++
>  lib/librte_gso/gso_tcp4.c               | 106 ++++++++++++++++++++++
>  lib/librte_gso/gso_tcp4.h               |  74 +++++++++++++++
>  lib/librte_gso/rte_gso.c                |  52 ++++++++++-
>  8 files changed, 538 insertions(+), 3 deletions(-)
>  create mode 100644 lib/librte_gso/gso_common.c
>  create mode 100644 lib/librte_gso/gso_common.h
>  create mode 100644 lib/librte_gso/gso_tcp4.c
>  create mode 100644 lib/librte_gso/gso_tcp4.h
> 
> diff --git a/doc/guides/rel_notes/release_17_11.rst b/doc/guides/rel_notes/release_17_11.rst
> index 7508be7..c414f73 100644
> --- a/doc/guides/rel_notes/release_17_11.rst
> +++ b/doc/guides/rel_notes/release_17_11.rst
> @@ -41,6 +41,18 @@ New Features
>       Also, make sure to start the actual text at the margin.
>       =========================================================
>  
> +* **Added the Generic Segmentation Offload Library.**
> +
> +  Added the Generic Segmentation Offload (GSO) library to enable
> +  applications to split large packets (e.g. MTU is 64KB) into small
> +  ones (e.g. MTU is 1500B). Supported packet types are:
> +
> +  * TCP/IPv4 packets, which may include a single VLAN tag.
> +
> +  The GSO library doesn't check if the input packets have correct
> +  checksums, and doesn't update checksums for output packets.
> +  Additionally, the GSO library doesn't process IP fragmented packets.
> +
>  
>  Resolved Issues
>  ---------------
> diff --git a/lib/librte_eal/common/include/rte_log.h b/lib/librte_eal/common/include/rte_log.h
> index ec8dba7..2fa1199 100644
> --- a/lib/librte_eal/common/include/rte_log.h
> +++ b/lib/librte_eal/common/include/rte_log.h
> @@ -87,6 +87,7 @@ struct rte_logs {
>  #define RTE_LOGTYPE_CRYPTODEV 17 /**< Log related to cryptodev. */
>  #define RTE_LOGTYPE_EFD       18 /**< Log related to EFD. */
>  #define RTE_LOGTYPE_EVENTDEV  19 /**< Log related to eventdev. */
> +#define RTE_LOGTYPE_GSO       20 /**< Log related to GSO. */
>  
>  /* these log types can be used in an application */
>  #define RTE_LOGTYPE_USER1     24 /**< User-defined log type 1. */
> diff --git a/lib/librte_gso/Makefile b/lib/librte_gso/Makefile
> index aeaacbc..2be64d1 100644
> --- a/lib/librte_gso/Makefile
> +++ b/lib/librte_gso/Makefile
> @@ -42,6 +42,8 @@ LIBABIVER := 1
>  
>  #source files
>  SRCS-$(CONFIG_RTE_LIBRTE_GSO) += rte_gso.c
> +SRCS-$(CONFIG_RTE_LIBRTE_GSO) += gso_common.c
> +SRCS-$(CONFIG_RTE_LIBRTE_GSO) += gso_tcp4.c
>  
>  # install this header file
>  SYMLINK-$(CONFIG_RTE_LIBRTE_GSO)-include += rte_gso.h
> diff --git a/lib/librte_gso/gso_common.c b/lib/librte_gso/gso_common.c
> new file mode 100644
> index 0000000..ee75d4c
> --- /dev/null
> +++ b/lib/librte_gso/gso_common.c
> @@ -0,0 +1,153 @@
> +/*-
> + *   BSD LICENSE
> + *
> + *   Copyright(c) 2017 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.
> + */
> +
> +#include <stdbool.h>
> +#include <errno.h>
> +
> +#include <rte_memcpy.h>
> +#include <rte_mempool.h>
> +
> +#include "gso_common.h"
> +
> +static inline void
> +hdr_segment_init(struct rte_mbuf *hdr_segment, struct rte_mbuf *pkt,
> +		uint16_t pkt_hdr_offset)
> +{
> +	/* Copy MBUF metadata */
> +	hdr_segment->nb_segs = 1;
> +	hdr_segment->port = pkt->port;
> +	hdr_segment->ol_flags = pkt->ol_flags;
> +	hdr_segment->packet_type = pkt->packet_type;
> +	hdr_segment->pkt_len = pkt_hdr_offset;
> +	hdr_segment->data_len = pkt_hdr_offset;
> +	hdr_segment->tx_offload = pkt->tx_offload;
> +
> +	/* Copy the packet header */
> +	rte_memcpy(rte_pktmbuf_mtod(hdr_segment, char *),
> +			rte_pktmbuf_mtod(pkt, char *),
> +			pkt_hdr_offset);
> +}
> +
> +static inline void
> +free_gso_segment(struct rte_mbuf **pkts, uint16_t nb_pkts)
> +{
> +	uint16_t i;
> +
> +	for (i = 0; i < nb_pkts; i++)
> +		rte_pktmbuf_free(pkts[i]);
> +}
> +
> +int
> +gso_do_segment(struct rte_mbuf *pkt,
> +		uint16_t pkt_hdr_offset,
> +		uint16_t pyld_unit_size,
> +		struct rte_mempool *direct_pool,
> +		struct rte_mempool *indirect_pool,
> +		struct rte_mbuf **pkts_out,
> +		uint16_t nb_pkts_out)
> +{
> +	struct rte_mbuf *pkt_in;
> +	struct rte_mbuf *hdr_segment, *pyld_segment, *prev_segment;
> +	uint16_t pkt_in_data_pos, segment_bytes_remaining;
> +	uint16_t pyld_len, nb_segs;
> +	bool more_in_pkt, more_out_segs;
> +
> +	pkt_in = pkt;
> +	nb_segs = 0;
> +	more_in_pkt = 1;
> +	pkt_in_data_pos = pkt_hdr_offset;
> +
> +	while (more_in_pkt) {
> +		if (unlikely(nb_segs >= nb_pkts_out)) {
> +			free_gso_segment(pkts_out, nb_segs);
> +			return -EINVAL;
> +		}
> +
> +		/* Allocate a direct MBUF */
> +		hdr_segment = rte_pktmbuf_alloc(direct_pool);
> +		if (unlikely(hdr_segment == NULL)) {
> +			free_gso_segment(pkts_out, nb_segs);
> +			return -ENOMEM;
> +		}
> +		/* Fill the packet header */
> +		hdr_segment_init(hdr_segment, pkt, pkt_hdr_offset);
> +
> +		prev_segment = hdr_segment;
> +		segment_bytes_remaining = pyld_unit_size;
> +		more_out_segs = 1;
> +
> +		while (more_out_segs && more_in_pkt) {
> +			/* Allocate an indirect MBUF */
> +			pyld_segment = rte_pktmbuf_alloc(indirect_pool);
> +			if (unlikely(pyld_segment == NULL)) {
> +				rte_pktmbuf_free(hdr_segment);
> +				free_gso_segment(pkts_out, nb_segs);
> +				return -ENOMEM;
> +			}
> +			/* Attach to current MBUF segment of pkt */
> +			rte_pktmbuf_attach(pyld_segment, pkt_in);
> +
> +			prev_segment->next = pyld_segment;
> +			prev_segment = pyld_segment;
> +
> +			pyld_len = segment_bytes_remaining;
> +			if (pyld_len + pkt_in_data_pos > pkt_in->data_len)
> +				pyld_len = pkt_in->data_len - pkt_in_data_pos;
> +
> +			pyld_segment->data_off = pkt_in_data_pos +
> +				pkt_in->data_off;
> +			pyld_segment->data_len = pyld_len;
> +
> +			/* Update header segment */
> +			hdr_segment->pkt_len += pyld_len;
> +			hdr_segment->nb_segs++;
> +
> +			pkt_in_data_pos += pyld_len;
> +			segment_bytes_remaining -= pyld_len;
> +
> +			/* Finish processing a MBUF segment of pkt */
> +			if (pkt_in_data_pos == pkt_in->data_len) {
> +				pkt_in = pkt_in->next;
> +				pkt_in_data_pos = 0;
> +				if (pkt_in == NULL)
> +					more_in_pkt = 0;
> +			}
> +
> +			/* Finish generating a GSO segment */
> +			if (segment_bytes_remaining == 0)
> +				more_out_segs = 0;
> +		}
> +		pkts_out[nb_segs++] = hdr_segment;
> +	}
> +	return nb_segs;
> +}
> diff --git a/lib/librte_gso/gso_common.h b/lib/librte_gso/gso_common.h
> new file mode 100644
> index 0000000..8d9b94e
> --- /dev/null
> +++ b/lib/librte_gso/gso_common.h
> @@ -0,0 +1,141 @@
> +/*-
> + *   BSD LICENSE
> + *
> + *   Copyright(c) 2017 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.
> + */
> +
> +#ifndef _GSO_COMMON_H_
> +#define _GSO_COMMON_H_
> +
> +#include <stdint.h>
> +
> +#include <rte_mbuf.h>
> +#include <rte_ip.h>
> +#include <rte_tcp.h>
> +
> +#define IS_FRAGMENTED(frag_off) (((frag_off) & IPV4_HDR_OFFSET_MASK) != 0 \
> +		|| ((frag_off) & IPV4_HDR_MF_FLAG) == IPV4_HDR_MF_FLAG)
> +
> +#define TCP_HDR_PSH_MASK ((uint8_t)0x08)
> +#define TCP_HDR_FIN_MASK ((uint8_t)0x01)
> +
> +#define IS_IPV4_TCP(flag) (((flag) & (PKT_TX_TCP_SEG | PKT_TX_IPV4)) == \
> +		(PKT_TX_TCP_SEG | PKT_TX_IPV4))
> +
> +/**
> + * Internal function which updates the TCP header of a packet, following
> + * segmentation. This is required to update the header's 'sent' sequence
> + * number, and also to clear 'PSH' and 'FIN' flags for non-tail segments.
> + *
> + * @param pkt
> + *  The packet containing the TCP header.
> + * @param l4_offset
> + *  The offset of the TCP header from the start of the packet.
> + * @param sent_seq
> + *  The sent sequence number.
> + * @param non-tail
> + *  Indicates whether or not this is a tail segment.
> + */
> +static inline void
> +update_tcp_header(struct rte_mbuf *pkt, uint16_t l4_offset, uint32_t sent_seq,
> +		uint8_t non_tail)
> +{
> +	struct tcp_hdr *tcp_hdr;
> +
> +	tcp_hdr = (struct tcp_hdr *)(rte_pktmbuf_mtod(pkt, char *) +
> +			l4_offset);
> +	tcp_hdr->sent_seq = rte_cpu_to_be_32(sent_seq);
> +	if (likely(non_tail))
> +		tcp_hdr->tcp_flags &= (~(TCP_HDR_PSH_MASK |
> +					TCP_HDR_FIN_MASK));
> +}
> +
> +/**
> + * Internal function which updates the IPv4 header of a packet, following
> + * segmentation. This is required to update the header's 'total_length' field,
> + * to reflect the reduced length of the now-segmented packet. Furthermore, the
> + * header's 'packet_id' field must be updated to reflect the new ID of the
> + * now-segmented packet.
> + *
> + * @param pkt
> + *  The packet containing the IPv4 header.
> + * @param l3_offset
> + *  The offset of the IPv4 header from the start of the packet.
> + * @param id
> + *  The new ID of the packet.
> +  */
> +static inline void
> +update_ipv4_header(struct rte_mbuf *pkt, uint16_t l3_offset, uint16_t id)
> +{
> +	struct ipv4_hdr *ipv4_hdr;
> +
> +	ipv4_hdr = (struct ipv4_hdr *)(rte_pktmbuf_mtod(pkt, char *) +
> +			l3_offset);
> +	ipv4_hdr->total_length = rte_cpu_to_be_16(pkt->pkt_len - l3_offset);
> +	ipv4_hdr->packet_id = rte_cpu_to_be_16(id);
> +}
> +
> +/**
> + * Internal function which divides the input packet into small segments.
> + * Each of the newly-created segments is organized as a two-segment MBUF,
> + * where the first segment is a standard mbuf, which stores a copy of
> + * packet header, and the second is an indirect mbuf which points to a
> + * section of data in the input packet.
> + *
> + * @param pkt
> + *  Packet to segment.
> + * @param pkt_hdr_offset
> + *  Packet header offset, measured in bytes.
> + * @param pyld_unit_size
> + *  The max payload length of a GSO segment.
> + * @param direct_pool
> + *  MBUF pool used for allocating direct buffers for output segments.
> + * @param indirect_pool
> + *  MBUF pool used for allocating indirect buffers for output segments.
> + * @param pkts_out
> + *  Pointer array used to keep the mbuf addresses of output segments. If
> + *  the memory space in pkts_out is insufficient, gso_do_segment() fails
> + *  and returns -EINVAL.
> + * @param nb_pkts_out
> + *  The max number of items that pkts_out can keep.
> + *
> + * @return
> + *  - The number of segments created in the event of success.
> + *  - Return -ENOMEM if run out of memory in MBUF pools.
> + *  - Return -EINVAL for invalid parameters.
> + */
> +int gso_do_segment(struct rte_mbuf *pkt,
> +		uint16_t pkt_hdr_offset,
> +		uint16_t pyld_unit_size,
> +		struct rte_mempool *direct_pool,
> +		struct rte_mempool *indirect_pool,
> +		struct rte_mbuf **pkts_out,
> +		uint16_t nb_pkts_out);
> +#endif
> diff --git a/lib/librte_gso/gso_tcp4.c b/lib/librte_gso/gso_tcp4.c
> new file mode 100644
> index 0000000..584a77d
> --- /dev/null
> +++ b/lib/librte_gso/gso_tcp4.c
> @@ -0,0 +1,106 @@
> +/*-
> + *   BSD LICENSE
> + *
> + *   Copyright(c) 2017 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.
> + */
> +
> +#include "gso_common.h"
> +#include "gso_tcp4.h"
> +
> +static void
> +update_ipv4_tcp_headers(struct rte_mbuf *pkt, uint64_t l3_offset,
> +		uint8_t ipid_delta, struct rte_mbuf **segs, uint16_t nb_segs)
> +{

No need to add "uint64_t l3_offset" as one parameter. We can directly get
l3_offset inside update_ipv4_tcp_headers().

Additionally, the type of l3_offset should be "uint16_t" instead of "uint64_t".
But if we change the prototype of update_ipv4_tcp_headers(), this issue doesn't
need to fix.

> +	struct ipv4_hdr *ipv4_hdr;
> +	struct tcp_hdr *tcp_hdr;
> +	uint32_t sent_seq;
> +	uint16_t id, tail_idx, i;
> +	uint16_t l4_offset;
> +
> +	ipv4_hdr = (struct ipv4_hdr *)(rte_pktmbuf_mtod(pkt, char*) +
> +			l3_offset);
> +	tcp_hdr = (struct tcp_hdr *)((char *)ipv4_hdr + pkt->l3_len);
> +	id = rte_be_to_cpu_16(ipv4_hdr->packet_id);
> +	sent_seq = rte_be_to_cpu_32(tcp_hdr->sent_seq);
> +	tail_idx = nb_segs - 1;
> +	l4_offset = l3_offset + pkt->l3_len;
> +
> +	for (i = 0; i < nb_segs; i++) {
> +		update_ipv4_header(segs[i], l3_offset, id);
> +		update_tcp_header(segs[i], l4_offset, sent_seq, i < tail_idx);
> +		id += ipid_delta;
> +		sent_seq += (segs[i]->pkt_len - segs[i]->data_len);
> +	}
> +}
> +
> +int
> +gso_tcp4_segment(struct rte_mbuf *pkt,
> +		uint16_t gso_size,
> +		uint8_t ipid_delta,
> +		struct rte_mempool *direct_pool,
> +		struct rte_mempool *indirect_pool,
> +		struct rte_mbuf **pkts_out,
> +		uint16_t nb_pkts_out)
> +{
> +	struct ipv4_hdr *ipv4_hdr;
> +	uint16_t tcp_dl;
> +	uint16_t pyld_unit_size, hdr_offset;
> +	uint16_t frag_off;
> +	int ret;
> +
> +	/* Don't process the fragmented packet */
> +	ipv4_hdr = (struct ipv4_hdr *)(rte_pktmbuf_mtod(pkt, char *) +
> +			pkt->l2_len);
> +	frag_off = rte_be_to_cpu_16(ipv4_hdr->fragment_offset);
> +	if (unlikely(IS_FRAGMENTED(frag_off))) {
> +		pkts_out[0] = pkt;
> +		return 1;
> +	}
> +
> +	/* Don't process the packet without data */
> +	tcp_dl = pkt->pkt_len - pkt->l2_len - pkt->l3_len - pkt->l4_len;
> +	if (unlikely(tcp_dl == 0)) {
> +		pkts_out[0] = pkt;
> +		return 1;
> +	}
> +
> +	hdr_offset = pkt->l2_len + pkt->l3_len + pkt->l4_len;
> +	pyld_unit_size = gso_size - hdr_offset;
> +
> +	/* Segment the payload */
> +	ret = gso_do_segment(pkt, hdr_offset, pyld_unit_size, direct_pool,
> +			indirect_pool, pkts_out, nb_pkts_out);
> +	if (ret > 1) {
> +		update_ipv4_tcp_headers(pkt, pkt->l2_len, ipid_delta,
> +				pkts_out, ret);
> +	}
> +
> +	return ret;
> +}
> diff --git a/lib/librte_gso/gso_tcp4.h b/lib/librte_gso/gso_tcp4.h
> new file mode 100644
> index 0000000..1c57441
> --- /dev/null
> +++ b/lib/librte_gso/gso_tcp4.h
> @@ -0,0 +1,74 @@
> +/*-
> + *   BSD LICENSE
> + *
> + *   Copyright(c) 2017 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.
> + */
> +
> +#ifndef _GSO_TCP4_H_
> +#define _GSO_TCP4_H_
> +
> +#include <stdint.h>
> +#include <rte_mbuf.h>
> +
> +/**
> + * Segment an IPv4/TCP packet. This function doesn't check if the input
> + * packet has correct checksums, and doesn't update checksums for output
> + * GSO segments. Furthermore, it doesn't process IP fragment packets.
> + *
> + * @param pkt
> + *  The packet mbuf to segment.
> + * @param gso_size
> + *  The max length of a GSO segment, measured in bytes.
> + * @param ipid_delta
> + *  The increasing unit of IP ids.
> + * @param direct_pool
> + *  MBUF pool used for allocating direct buffers for output segments.
> + * @param indirect_pool
> + *  MBUF pool used for allocating indirect buffers for output segments.
> + * @param pkts_out
> + *  Pointer array used to store the MBUF addresses of output GSO
> + *  segments, when the function succeeds. If the memory space in
> + *  pkts_out is insufficient, it fails and returns -EINVAL.
> + * @param nb_pkts_out
> + *  The max number of items that 'pkts_out' can keep.
> + *
> + * @return
> + *   - The number of GSO segments filled in pkts_out on success.
> + *   - Return -ENOMEM if run out of memory in MBUF pools.
> + *   - Return -EINVAL for invalid parameters.
> + */
> +int gso_tcp4_segment(struct rte_mbuf *pkt,
> +		uint16_t gso_size,
> +		uint8_t ip_delta,
> +		struct rte_mempool *direct_pool,
> +		struct rte_mempool *indirect_pool,
> +		struct rte_mbuf **pkts_out,
> +		uint16_t nb_pkts_out);
> +#endif
> diff --git a/lib/librte_gso/rte_gso.c b/lib/librte_gso/rte_gso.c
> index b773636..a4fce50 100644
> --- a/lib/librte_gso/rte_gso.c
> +++ b/lib/librte_gso/rte_gso.c
> @@ -33,7 +33,12 @@
>  
>  #include <errno.h>
>  
> +#include <rte_log.h>
> +#include <rte_ethdev.h>
> +
>  #include "rte_gso.h"
> +#include "gso_common.h"
> +#include "gso_tcp4.h"
>  
>  int
>  rte_gso_segment(struct rte_mbuf *pkt,
> @@ -41,12 +46,53 @@
>  		struct rte_mbuf **pkts_out,
>  		uint16_t nb_pkts_out)
>  {
> +	struct rte_mempool *direct_pool, *indirect_pool;
> +	struct rte_mbuf *pkt_seg;
> +	uint64_t ol_flags;
> +	uint16_t gso_size;
> +	uint8_t ipid_delta;
> +	int ret = 1;
> +
>  	if (pkt == NULL || pkts_out == NULL || gso_ctx == NULL ||
>  			nb_pkts_out < 1)
>  		return -EINVAL;
>  
> -	pkt->ol_flags &= (~PKT_TX_TCP_SEG);
> -	pkts_out[0] = pkt;
> +	if ((gso_ctx->gso_size >= pkt->pkt_len) || (gso_ctx->gso_types &
> +				DEV_TX_OFFLOAD_TCP_TSO) !=
> +			gso_ctx->gso_types) {
> +		pkt->ol_flags &= (~PKT_TX_TCP_SEG);
> +		pkts_out[0] = pkt;
> +		return 1;
> +	}
> +
> +	direct_pool = gso_ctx->direct_pool;
> +	indirect_pool = gso_ctx->indirect_pool;
> +	gso_size = gso_ctx->gso_size;
> +	ipid_delta = (gso_ctx->ipid_flag != RTE_GSO_IPID_FIXED);
> +	ol_flags = pkt->ol_flags;
> +
> +	if (IS_IPV4_TCP(pkt->ol_flags)) {
> +		pkt->ol_flags &= (~PKT_TX_TCP_SEG);
> +		ret = gso_tcp4_segment(pkt, gso_size, ipid_delta,
> +				direct_pool, indirect_pool,
> +				pkts_out, nb_pkts_out);
> +	} else {
> +		pkt->ol_flags &= (~PKT_TX_TCP_SEG);
> +		pkts_out[0] = pkt;
> +		RTE_LOG(WARNING, GSO, "Unsupported packet type\n");
> +		return 1;
> +	}
> +
> +	if (ret > 1) {
> +		pkt_seg = pkt;
> +		while (pkt_seg) {
> +			rte_mbuf_refcnt_update(pkt_seg, -1);
> +			pkt_seg = pkt_seg->next;
> +		}
> +	} else if (ret < 0) {
> +		/* Revert the ol_flags in the event of failure. */
> +		pkt->ol_flags = ol_flags;
> +	}
>  
> -	return 1;
> +	return ret;
>  }
> -- 
> 1.9.3