@@ -75,6 +75,13 @@ New Features
Added support for firmwares with multiple Ethernet ports per physical port.
+* **Add Generic Receive Offload API support.**
+
+ Generic Receive Offload (GRO) API supports to reassemble TCP/IPv4
+ packets. GRO API assumes all inputted packets are with correct
+ checksums. GRO API doesn't update checksums for merged packets. If
+ inputted packets are IP fragmented, GRO API assumes they are complete
+ packets (i.e. with L4 headers).
Resolved Issues
---------------
@@ -43,6 +43,7 @@ LIBABIVER := 1
# source files
SRCS-$(CONFIG_RTE_LIBRTE_GRO) += rte_gro.c
+SRCS-$(CONFIG_RTE_LIBRTE_GRO) += rte_gro_tcp.c
# install this header file
SYMLINK-$(CONFIG_RTE_LIBRTE_GRO)-include += rte_gro.h
@@ -32,8 +32,11 @@
#include <rte_malloc.h>
#include <rte_mbuf.h>
+#include <rte_cycles.h>
+#include <rte_ethdev.h>
#include "rte_gro.h"
+#include "rte_gro_tcp.h"
typedef void *(*gro_tbl_create_fn)(uint16_t socket_id,
uint16_t max_flow_num,
@@ -41,9 +44,12 @@ typedef void *(*gro_tbl_create_fn)(uint16_t socket_id,
typedef void (*gro_tbl_destroy_fn)(void *tbl);
typedef uint32_t (*gro_tbl_item_num_fn)(void *tbl);
-static gro_tbl_create_fn tbl_create_functions[RTE_GRO_TYPE_MAX_NUM];
-static gro_tbl_destroy_fn tbl_destroy_functions[RTE_GRO_TYPE_MAX_NUM];
-static gro_tbl_item_num_fn tbl_item_num_functions[RTE_GRO_TYPE_MAX_NUM];
+static gro_tbl_create_fn tbl_create_functions[RTE_GRO_TYPE_MAX_NUM] = {
+ gro_tcp_tbl_create, NULL};
+static gro_tbl_destroy_fn tbl_destroy_functions[RTE_GRO_TYPE_MAX_NUM] = {
+ gro_tcp_tbl_destroy, NULL};
+static gro_tbl_item_num_fn tbl_item_num_functions[
+ RTE_GRO_TYPE_MAX_NUM] = {gro_tcp_tbl_item_num, NULL};
/**
* GRO table, which is used to merge packets. It keeps many reassembly
@@ -130,27 +136,120 @@ void rte_gro_tbl_destroy(void *tbl)
}
uint16_t
-rte_gro_reassemble_burst(struct rte_mbuf **pkts __rte_unused,
+rte_gro_reassemble_burst(struct rte_mbuf **pkts,
uint16_t nb_pkts,
- const struct rte_gro_param *param __rte_unused)
+ const struct rte_gro_param *param)
{
- return nb_pkts;
+ uint16_t i;
+ uint16_t nb_after_gro = nb_pkts;
+ uint32_t item_num = RTE_MIN(nb_pkts, param->max_flow_num *
+ param->max_item_per_flow);
+
+ /* allocate a reassembly table for TCP/IPv4 GRO */
+ uint32_t tcp_item_num = RTE_MIN(item_num,
+ RTE_GRO_MAX_BURST_ITEM_NUM);
+ struct gro_tcp_tbl tcp_tbl;
+ struct gro_tcp_key tcp_keys[tcp_item_num];
+ struct gro_tcp_item tcp_items[tcp_item_num];
+
+ struct rte_mbuf *unprocess_pkts[nb_pkts];
+ uint16_t unprocess_num = 0;
+ int32_t ret;
+ uint64_t current_time;
+
+ if ((param->desired_gro_types & RTE_GRO_TCP_IPV4) == 0)
+ return nb_pkts;
+
+ memset(tcp_keys, 0, sizeof(struct gro_tcp_key) *
+ tcp_item_num);
+ memset(tcp_items, 0, sizeof(struct gro_tcp_item) *
+ tcp_item_num);
+ tcp_tbl.keys = tcp_keys;
+ tcp_tbl.items = tcp_items;
+ tcp_tbl.key_num = 0;
+ tcp_tbl.item_num = 0;
+ tcp_tbl.max_key_num = tcp_item_num;
+ tcp_tbl.max_item_num = tcp_item_num;
+
+ current_time = rte_rdtsc();
+
+ for (i = 0; i < nb_pkts; i++) {
+ if (RTE_ETH_IS_IPV4_HDR(pkts[i]->packet_type) &&
+ (pkts[i]->packet_type & RTE_PTYPE_L4_TCP)) {
+ ret = gro_tcp4_reassemble(pkts[i],
+ &tcp_tbl,
+ param->max_packet_size,
+ current_time);
+ if (ret > 0)
+ /* merge successfully */
+ nb_after_gro--;
+ else if (ret < 0)
+ unprocess_pkts[unprocess_num++] =
+ pkts[i];
+ } else
+ unprocess_pkts[unprocess_num++] =
+ pkts[i];
+ }
+
+ /* re-arrange GROed packets */
+ if (nb_after_gro < nb_pkts) {
+ i = gro_tcp_tbl_timeout_flush(&tcp_tbl, 0,
+ pkts, nb_pkts);
+ if (unprocess_num > 0)
+ memcpy(&pkts[i], unprocess_pkts,
+ sizeof(struct rte_mbuf *) *
+ unprocess_num);
+ }
+ return nb_after_gro;
}
uint16_t
-rte_gro_reassemble(struct rte_mbuf **pkts __rte_unused,
+rte_gro_reassemble(struct rte_mbuf **pkts,
uint16_t nb_pkts,
- void *tbl __rte_unused)
+ void *tbl)
{
- return nb_pkts;
+ uint16_t i, unprocess_num = 0;
+ struct rte_mbuf *unprocess_pkts[nb_pkts];
+ struct rte_gro_tbl *gro_tbl = (struct rte_gro_tbl *)tbl;
+ uint64_t current_time;
+
+ if ((gro_tbl->desired_gro_types & RTE_GRO_TCP_IPV4) == 0)
+ return nb_pkts;
+
+ current_time = rte_rdtsc();
+ for (i = 0; i < nb_pkts; i++) {
+ if (RTE_ETH_IS_IPV4_HDR(pkts[i]->packet_type) &&
+ (pkts[i]->packet_type & RTE_PTYPE_L4_TCP)) {
+ if (gro_tcp4_reassemble(pkts[i],
+ gro_tbl->tbls[RTE_GRO_TCP_IPV4_INDEX],
+ gro_tbl->max_packet_size,
+ current_time) < 0)
+ unprocess_pkts[unprocess_num++] = pkts[i];
+ } else
+ unprocess_pkts[unprocess_num++] = pkts[i];
+ }
+ if (unprocess_num > 0)
+ memcpy(pkts, unprocess_pkts,
+ sizeof(struct rte_mbuf *) * unprocess_num);
+
+ return unprocess_num;
}
uint16_t
-rte_gro_timeout_flush(void *tbl __rte_unused,
- uint64_t desired_gro_types __rte_unused,
- struct rte_mbuf **out __rte_unused,
- uint16_t max_nb_out __rte_unused)
+rte_gro_timeout_flush(void *tbl,
+ uint64_t desired_gro_types,
+ struct rte_mbuf **out,
+ uint16_t max_nb_out)
{
+ struct rte_gro_tbl *gro_tbl = (struct rte_gro_tbl *)tbl;
+
+ desired_gro_types = desired_gro_types &
+ gro_tbl->desired_gro_types;
+ if (desired_gro_types & RTE_GRO_TCP_IPV4)
+ return gro_tcp_tbl_timeout_flush(
+ gro_tbl->tbls[RTE_GRO_TCP_IPV4_INDEX],
+ gro_tbl->max_timeout_cycles,
+ out, max_nb_out);
return 0;
}
@@ -45,7 +45,11 @@ extern "C" {
/* max number of supported GRO types */
#define RTE_GRO_TYPE_MAX_NUM 64
-#define RTE_GRO_TYPE_SUPPORT_NUM 0 /**< current supported GRO num */
+#define RTE_GRO_TYPE_SUPPORT_NUM 1 /**< current supported GRO num */
+
+/* TCP/IPv4 GRO flag */
+#define RTE_GRO_TCP_IPV4_INDEX 0
+#define RTE_GRO_TCP_IPV4 (1ULL << RTE_GRO_TCP_IPV4_INDEX)
struct rte_gro_param {
new file mode 100644
@@ -0,0 +1,400 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2017 Intel Corporation. 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 <rte_malloc.h>
+#include <rte_mbuf.h>
+#include <rte_cycles.h>
+
+#include <rte_ethdev.h>
+#include <rte_ip.h>
+#include <rte_tcp.h>
+
+#include "rte_gro_tcp.h"
+
+void *gro_tcp_tbl_create(uint16_t socket_id,
+ uint16_t max_flow_num,
+ uint16_t max_item_per_flow)
+{
+ size_t size;
+ uint32_t entries_num;
+ struct gro_tcp_tbl *tbl;
+
+ entries_num = max_flow_num * max_item_per_flow;
+ entries_num = entries_num > GRO_TCP_TBL_MAX_ITEM_NUM ?
+ GRO_TCP_TBL_MAX_ITEM_NUM : entries_num;
+
+ if (entries_num == 0)
+ return NULL;
+
+ tbl = (struct gro_tcp_tbl *)rte_zmalloc_socket(
+ __func__,
+ sizeof(struct gro_tcp_tbl),
+ RTE_CACHE_LINE_SIZE,
+ socket_id);
+ if (tbl == NULL)
+ return NULL;
+
+ size = sizeof(struct gro_tcp_item) * entries_num;
+ tbl->items = (struct gro_tcp_item *)rte_zmalloc_socket(
+ __func__,
+ size,
+ RTE_CACHE_LINE_SIZE,
+ socket_id);
+ if (tbl->items == NULL) {
+ rte_free(tbl);
+ return NULL;
+ }
+ tbl->max_item_num = entries_num;
+
+ size = sizeof(struct gro_tcp_key) * entries_num;
+ tbl->keys = (struct gro_tcp_key *)rte_zmalloc_socket(
+ __func__,
+ size, RTE_CACHE_LINE_SIZE,
+ socket_id);
+ if (tbl->keys == NULL) {
+ rte_free(tbl->items);
+ rte_free(tbl);
+ return NULL;
+ }
+ tbl->max_key_num = entries_num;
+ return tbl;
+}
+
+void gro_tcp_tbl_destroy(void *tbl)
+{
+ struct gro_tcp_tbl *tcp_tbl = (struct gro_tcp_tbl *)tbl;
+
+ if (tcp_tbl) {
+ rte_free(tcp_tbl->items);
+ rte_free(tcp_tbl->keys);
+ }
+ rte_free(tcp_tbl);
+}
+
+static struct rte_mbuf *get_mbuf_lastseg(struct rte_mbuf *pkt)
+{
+ struct rte_mbuf *lastseg = pkt;
+
+ while (lastseg->next)
+ lastseg = lastseg->next;
+
+ return lastseg;
+}
+
+/**
+ * merge two TCP/IPv4 packets without updating checksums.
+ */
+static int
+merge_two_tcp4_packets(struct gro_tcp_item *item_src,
+ struct rte_mbuf *pkt,
+ uint32_t max_packet_size)
+{
+ struct ipv4_hdr *ipv4_hdr1, *ipv4_hdr2;
+ uint16_t ipv4_ihl1, tcp_hl1, tcp_dl1;
+ struct rte_mbuf *pkt_src = item_src->pkt;
+
+ /* parse the given packet */
+ ipv4_hdr1 = (struct ipv4_hdr *)(rte_pktmbuf_mtod(pkt,
+ char *) + pkt->l2_len);
+ ipv4_ihl1 = pkt->l3_len;
+ tcp_hl1 = pkt->l4_len;
+ tcp_dl1 = rte_be_to_cpu_16(ipv4_hdr1->total_length) -
+ ipv4_ihl1 - tcp_hl1;
+
+ /* parse the original packet */
+ ipv4_hdr2 = (struct ipv4_hdr *)(rte_pktmbuf_mtod(pkt_src,
+ char *) + pkt_src->l2_len);
+
+ if (pkt_src->pkt_len + tcp_dl1 > max_packet_size)
+ return -1;
+
+ /* remove the header of the incoming packet */
+ rte_pktmbuf_adj(pkt, pkt->l2_len + ipv4_ihl1 + tcp_hl1);
+
+ /* chain the two packet together and update lastseg */
+ item_src->lastseg->next = pkt;
+ item_src->lastseg = get_mbuf_lastseg(pkt);
+
+ /* update IP header */
+ ipv4_hdr2->total_length = rte_cpu_to_be_16(
+ rte_be_to_cpu_16(
+ ipv4_hdr2->total_length)
+ + tcp_dl1);
+
+ /* update mbuf metadata for the merged packet */
+ pkt_src->nb_segs += pkt->nb_segs;
+ pkt_src->pkt_len += pkt->pkt_len;
+ return 1;
+}
+
+static int
+check_seq_option(struct rte_mbuf *pkt,
+ struct tcp_hdr *tcp_hdr,
+ uint16_t tcp_hl)
+{
+ struct ipv4_hdr *ipv4_hdr1;
+ struct tcp_hdr *tcp_hdr1;
+ uint16_t ipv4_ihl1, tcp_hl1, tcp_dl1;
+ uint32_t sent_seq1, sent_seq;
+ uint16_t len;
+ int ret = -1;
+
+ ipv4_hdr1 = (struct ipv4_hdr *)(rte_pktmbuf_mtod(pkt,
+ char *) + pkt->l2_len);
+ ipv4_ihl1 = pkt->l3_len;
+ tcp_hdr1 = (struct tcp_hdr *)((char *)ipv4_hdr1 + ipv4_ihl1);
+ tcp_hl1 = pkt->l4_len;
+ tcp_dl1 = rte_be_to_cpu_16(ipv4_hdr1->total_length) - ipv4_ihl1
+ - tcp_hl1;
+ sent_seq1 = rte_be_to_cpu_32(tcp_hdr1->sent_seq) + tcp_dl1;
+ sent_seq = rte_be_to_cpu_32(tcp_hdr->sent_seq);
+
+ /* check if the two packets are neighbor */
+ if (sent_seq == sent_seq1) {
+ ret = 1;
+ len = RTE_MAX(tcp_hl, tcp_hl1) - sizeof(struct tcp_hdr);
+ /* check if TCP option field equals */
+ if ((tcp_hl1 != tcp_hl) || ((len > 0) &&
+ (memcmp(tcp_hdr1 + 1,
+ tcp_hdr + 1,
+ len) != 0)))
+ ret = -1;
+ }
+ return ret;
+}
+
+static uint32_t
+find_an_empty_item(struct gro_tcp_tbl *tbl)
+{
+ uint32_t i;
+
+ for (i = 0; i < tbl->max_item_num; i++)
+ if (tbl->items[i].pkt == NULL)
+ return i;
+ return INVALID_ARRAY_INDEX;
+}
+
+static uint32_t
+find_an_empty_key(struct gro_tcp_tbl *tbl)
+{
+ uint32_t i;
+
+ for (i = 0; i < tbl->max_key_num; i++)
+ if (tbl->keys[i].is_valid == 0)
+ return i;
+ return INVALID_ARRAY_INDEX;
+}
+
+int32_t
+gro_tcp4_reassemble(struct rte_mbuf *pkt,
+ struct gro_tcp_tbl *tbl,
+ uint32_t max_packet_size,
+ uint64_t start_time)
+{
+ struct ether_hdr *eth_hdr;
+ struct ipv4_hdr *ipv4_hdr;
+ struct tcp_hdr *tcp_hdr;
+ uint16_t ipv4_ihl, tcp_hl, tcp_dl;
+
+ struct tcp_key key;
+ uint32_t cur_idx, prev_idx, item_idx;
+ uint32_t i, key_idx;
+
+ eth_hdr = rte_pktmbuf_mtod(pkt, struct ether_hdr *);
+ ipv4_hdr = (struct ipv4_hdr *)((char *)eth_hdr + pkt->l2_len);
+ ipv4_ihl = pkt->l3_len;
+
+ /* check if the packet should be processed */
+ if (ipv4_ihl < sizeof(struct ipv4_hdr))
+ return -1;
+ tcp_hdr = (struct tcp_hdr *)((char *)ipv4_hdr + ipv4_ihl);
+ tcp_hl = pkt->l4_len;
+ tcp_dl = rte_be_to_cpu_16(ipv4_hdr->total_length) - ipv4_ihl
+ - tcp_hl;
+ if (tcp_dl == 0)
+ return -1;
+
+ /* find a key and traverse all packets in its item group */
+ key.eth_saddr = eth_hdr->s_addr;
+ key.eth_daddr = eth_hdr->d_addr;
+ key.ip_src_addr[0] = rte_be_to_cpu_32(ipv4_hdr->src_addr);
+ key.ip_dst_addr[0] = rte_be_to_cpu_32(ipv4_hdr->dst_addr);
+ key.src_port = rte_be_to_cpu_16(tcp_hdr->src_port);
+ key.dst_port = rte_be_to_cpu_16(tcp_hdr->dst_port);
+ key.recv_ack = rte_be_to_cpu_32(tcp_hdr->recv_ack);
+ key.tcp_flags = tcp_hdr->tcp_flags;
+
+ for (i = 0; i < tbl->max_key_num; i++) {
+ /* search for a key */
+ if ((tbl->keys[i].is_valid == 0) ||
+ (memcmp(&(tbl->keys[i].key), &key,
+ sizeof(struct tcp_key)) != 0))
+ continue;
+
+ cur_idx = tbl->keys[i].start_index;
+ prev_idx = cur_idx;
+ while (cur_idx != INVALID_ARRAY_INDEX) {
+ if (check_seq_option(tbl->items[cur_idx].pkt,
+ tcp_hdr,
+ tcp_hl) > 0) {
+ if (merge_two_tcp4_packets(
+ &(tbl->items[cur_idx]),
+ pkt,
+ max_packet_size) > 0)
+ return 1;
+ /**
+ * fail to merge two packets since
+ * it's beyond the max packet length.
+ * Insert it into the item group.
+ */
+ item_idx = find_an_empty_item(tbl);
+ if (item_idx == INVALID_ARRAY_INDEX)
+ return -1;
+ tbl->items[prev_idx].next_pkt_idx = item_idx;
+ tbl->items[item_idx].pkt = pkt;
+ tbl->items[item_idx].lastseg =
+ get_mbuf_lastseg(pkt);
+ tbl->items[item_idx].next_pkt_idx =
+ INVALID_ARRAY_INDEX;
+ tbl->items[item_idx].start_time = start_time;
+ tbl->item_num++;
+ return 0;
+ }
+ prev_idx = cur_idx;
+ cur_idx = tbl->items[cur_idx].next_pkt_idx;
+ }
+ /**
+ * find a corresponding item group but fails to find
+ * one packet to merge. Insert it into this item group.
+ */
+ item_idx = find_an_empty_item(tbl);
+ if (item_idx == INVALID_ARRAY_INDEX)
+ return -1;
+ tbl->items[prev_idx].next_pkt_idx = item_idx;
+ tbl->items[item_idx].pkt = pkt;
+ tbl->items[item_idx].lastseg =
+ get_mbuf_lastseg(pkt);
+ tbl->items[item_idx].next_pkt_idx = INVALID_ARRAY_INDEX;
+ tbl->items[item_idx].start_time = start_time;
+ tbl->item_num++;
+ return 0;
+ }
+
+ /**
+ * merge fail as the given packet has
+ * a new key. So insert a new key.
+ */
+ item_idx = find_an_empty_item(tbl);
+ key_idx = find_an_empty_key(tbl);
+ /**
+ * if current key or item number is greater than the max
+ * value, don't insert the packet into the table and return
+ * immediately.
+ */
+ if (item_idx == INVALID_ARRAY_INDEX ||
+ key_idx == INVALID_ARRAY_INDEX)
+ return -1;
+ tbl->items[item_idx].pkt = pkt;
+ tbl->items[item_idx].lastseg = get_mbuf_lastseg(pkt);
+ tbl->items[item_idx].next_pkt_idx = INVALID_ARRAY_INDEX;
+ tbl->items[item_idx].start_time = start_time;
+ tbl->item_num++;
+
+ memcpy(&(tbl->keys[key_idx].key),
+ &key, sizeof(struct tcp_key));
+ tbl->keys[key_idx].start_index = item_idx;
+ tbl->keys[key_idx].is_valid = 1;
+ tbl->key_num++;
+
+ return 0;
+}
+
+uint16_t
+gro_tcp_tbl_timeout_flush(struct gro_tcp_tbl *tbl,
+ uint64_t timeout_cycles,
+ struct rte_mbuf **out,
+ uint16_t nb_out)
+{
+ uint16_t k = 0;
+ uint32_t i, j;
+ uint64_t current_time;
+
+ current_time = rte_rdtsc();
+ for (i = 0; i < tbl->max_key_num; i++) {
+ /* all keys have been checked, return immediately */
+ if (tbl->key_num == 0)
+ return k;
+
+ if (tbl->keys[i].is_valid == 0)
+ continue;
+
+ j = tbl->keys[i].start_index;
+ while (j != INVALID_ARRAY_INDEX) {
+ if (current_time - tbl->items[j].start_time >=
+ timeout_cycles) {
+ out[k++] = tbl->items[j].pkt;
+ tbl->items[j].pkt = NULL;
+ tbl->item_num--;
+ j = tbl->items[j].next_pkt_idx;
+
+ /**
+ * delete the key as all of
+ * its packets are flushed.
+ */
+ if (j == INVALID_ARRAY_INDEX) {
+ tbl->keys[i].is_valid = 0;
+ tbl->key_num--;
+ } else
+ /* update start_index of the key */
+ tbl->keys[i].start_index = j;
+
+ if (k == nb_out)
+ return k;
+ } else
+ /**
+ * left packets of this key won't be
+ * timeout, so go to check other keys.
+ */
+ break;
+ }
+ }
+ return k;
+}
+
+uint32_t gro_tcp_tbl_item_num(void *tbl)
+{
+ struct gro_tcp_tbl *gro_tbl = (struct gro_tcp_tbl *)tbl;
+
+ if (gro_tbl)
+ return gro_tbl->item_num;
+ return 0;
+}
new file mode 100644
@@ -0,0 +1,172 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2017 Intel Corporation. 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 _RTE_GRO_TCP_H_
+#define _RTE_GRO_TCP_H_
+
+#define INVALID_ARRAY_INDEX 0xffffffffUL
+#define GRO_TCP_TBL_MAX_ITEM_NUM (UINT32_MAX - 1)
+
+/* criteria of mergeing packets */
+struct tcp_key {
+ struct ether_addr eth_saddr;
+ struct ether_addr eth_daddr;
+ uint32_t ip_src_addr[4]; /**< IPv4 uses the first 4B */
+ uint32_t ip_dst_addr[4];
+
+ uint32_t recv_ack; /**< acknowledgment sequence number. */
+ uint16_t src_port;
+ uint16_t dst_port;
+ uint8_t tcp_flags; /**< TCP flags. */
+};
+
+struct gro_tcp_key {
+ struct tcp_key key;
+ uint32_t start_index; /**< the first packet index of the flow */
+ uint8_t is_valid;
+};
+
+struct gro_tcp_item {
+ struct rte_mbuf *pkt; /**< packet address. */
+ struct rte_mbuf *lastseg; /**< last segment of the packet */
+ /* the time when the packet in added into the table */
+ uint64_t start_time;
+ uint32_t next_pkt_idx; /**< next packet index. */
+};
+
+/**
+ * TCP reassembly table. Both TCP/IPv4 and TCP/IPv6 use the same table
+ * structure.
+ */
+struct gro_tcp_tbl {
+ struct gro_tcp_item *items; /**< item array */
+ struct gro_tcp_key *keys; /**< key array */
+ uint32_t item_num; /**< current item number */
+ uint32_t key_num; /**< current key num */
+ uint32_t max_item_num; /**< item array size */
+ uint32_t max_key_num; /**< key array size */
+};
+
+/**
+ * This function creates a TCP reassembly table.
+ *
+ * @param socket_id
+ * socket index where the Ethernet port connects to.
+ * @param max_flow_num
+ * the maximum number of flows in the TCP GRO table
+ * @param max_item_per_flow
+ * the maximum packet number per flow.
+ * @return
+ * if create successfully, return a pointer which points to the
+ * created TCP GRO table. Otherwise, return NULL.
+ */
+void *gro_tcp_tbl_create(uint16_t socket_id,
+ uint16_t max_flow_num,
+ uint16_t max_item_per_flow);
+
+/**
+ * This function destroys a TCP reassembly table.
+ * @param tbl
+ * a pointer points to the TCP reassembly table.
+ */
+void gro_tcp_tbl_destroy(void *tbl);
+
+/**
+ * This function searches for a packet in the TCP reassembly table to
+ * merge with the inputted one. To merge two packets is to chain them
+ * together and update packet headers. If the packet is without data
+ * (e.g. SYN, SYN-ACK packet), this function returns immediately.
+ * Otherwise, the packet is either merged, or inserted into the table.
+ * Besides, if there is no available space to insert the packet, this
+ * function returns immediately too.
+ *
+ * This function assumes the inputted packet is with correct IPv4 and
+ * TCP checksums. And if two packets are merged, it won't re-calculate
+ * IPv4 and TCP checksums. Besides, if the inputted packet is IP
+ * fragmented, it assumes the packet is complete (with TCP header).
+ *
+ * @param pkt
+ * packet to reassemble.
+ * @param tbl
+ * a pointer that points to a TCP reassembly table.
+ * @param max_packet_size
+ * max packet length after merged
+ * @start_time
+ * the start time that the packet is inserted into the table
+ * @return
+ * if the packet doesn't have data, or there is no available space
+ * in the table to insert a new item or a new key, return a negative
+ * value. If the packet is merged successfully, return an positive
+ * value. If the packet is inserted into the table, return 0.
+ */
+int32_t
+gro_tcp4_reassemble(struct rte_mbuf *pkt,
+ struct gro_tcp_tbl *tbl,
+ uint32_t max_packet_size,
+ uint64_t start_time);
+
+/**
+ * This function flushes timeout packets in a TCP reassembly table to
+ * applications, and without updating checksums for merged packets.
+ * The max number of flushed timeout packets is the element number of
+ * the array which is used to keep flushed packets.
+ *
+ * @param tbl
+ * a pointer that points to a TCP GRO table.
+ * @param timeout_cycles
+ * the maximum time that packets can stay in the table.
+ * @param out
+ * pointer array which is used to keep flushed packets.
+ * @param nb_out
+ * the element number of out. It's also the max number of timeout
+ * packets that can be flushed finally.
+ * @return
+ * the number of packets that are returned.
+ */
+uint16_t
+gro_tcp_tbl_timeout_flush(struct gro_tcp_tbl *tbl,
+ uint64_t timeout_cycles,
+ struct rte_mbuf **out,
+ uint16_t nb_out);
+
+/**
+ * This function returns the number of the packets in the TCP
+ * reassembly table.
+ *
+ * @param tbl
+ * pointer points to a TCP reassembly table.
+ * @return
+ * the number of packets in the table
+ */
+uint32_t
+gro_tcp_tbl_item_num(void *tbl);
+#endif