/*
-// Copyright (c) 2010-2017 Intel Corporation
+// Copyright (c) 2010-2020 Intel Corporation
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
#include <rte_mbuf.h>
#include <rte_udp.h>
+#include <rte_icmp.h>
#include "task_init.h"
#include "task_base.h"
#include "qinq.h"
#include "gre.h"
#include "prefetch.h"
+#include "defines.h"
#include "igmp.h"
#include "prox_cksum.h"
#include "prox_compat.h"
uint8_t src_dst_mac[12];
uint32_t local_ipv4;
int offload_crc;
+ uint64_t last_echo_req_rcvd_tsc;
+ uint64_t last_echo_rep_rcvd_tsc;
+ uint32_t n_echo_req;
+ uint32_t n_echo_rep;
};
#define NB_IGMP_MBUF 1024
rte_memcpy(hdr, task->src_dst_mac, sizeof(task->src_dst_mac));
} else {
hdr = rte_pktmbuf_mtod(mbuf, prox_rte_ether_hdr *);
- if (likely((task->runtime_flags & TASK_ARG_SRC_MAC_SET) == 0)) {
+ if (unlikely((task->runtime_flags & TASK_ARG_SRC_MAC_SET) == 0)) {
/* dst mac will be used as src mac */
prox_rte_ether_addr_copy(&hdr->d_addr, &mac);
}
else
prox_rte_ether_addr_copy(&hdr->s_addr, &hdr->d_addr);
- if (unlikely(task->runtime_flags & TASK_ARG_SRC_MAC_SET)) {
+ if (likely(task->runtime_flags & TASK_ARG_SRC_MAC_SET)) {
prox_rte_ether_addr_copy((prox_rte_ether_addr *)&task->src_dst_mac[6], &hdr->s_addr);
} else {
prox_rte_ether_addr_copy(&mac, &hdr->s_addr);
memcpy(dst_mac, &mac, sizeof(prox_rte_ether_addr));
}
+static inline void build_icmp_reply_message(struct task_base *tbase, struct rte_mbuf *mbuf)
+{
+ struct task_swap *task = (struct task_swap *)tbase;
+ prox_rte_ether_hdr *hdr = rte_pktmbuf_mtod(mbuf, prox_rte_ether_hdr *);
+ prox_rte_ether_addr dst_mac;
+ prox_rte_ether_addr_copy(&hdr->s_addr, &dst_mac);
+ prox_rte_ether_addr_copy(&hdr->d_addr, &hdr->s_addr);
+ prox_rte_ether_addr_copy(&dst_mac, &hdr->d_addr);
+ prox_rte_ipv4_hdr *ip_hdr = (prox_rte_ipv4_hdr *)(hdr + 1);
+ ip_hdr->dst_addr = ip_hdr->src_addr;
+ ip_hdr->src_addr = task->local_ipv4;
+ prox_rte_icmp_hdr *picmp = (prox_rte_icmp_hdr *)(ip_hdr + 1);
+ picmp->icmp_type = PROX_RTE_IP_ICMP_ECHO_REPLY;
+}
+
static inline void build_igmp_message(struct task_base *tbase, struct rte_mbuf *mbuf, uint32_t ip, uint8_t igmp_message)
{
struct task_swap *task = (struct task_swap *)tbase;
prox_ip_udp_cksum(mbuf, ip_hdr, sizeof(prox_rte_ether_hdr), sizeof(prox_rte_ipv4_hdr), task->offload_crc);
}
+static void stop_swap(struct task_base *tbase)
+{
+ struct task_swap *task = (struct task_swap *)tbase;
+ if (task->igmp_pool) {
+ rte_mempool_free(task->igmp_pool);
+ task->igmp_pool = NULL;
+ }
+}
+
+static void handle_ipv6(struct task_swap *task, struct rte_mbuf *mbufs, prox_rte_ipv6_hdr *ipv6_hdr, uint8_t *out)
+{
+ __m128i ip = _mm_loadu_si128((__m128i*)&(ipv6_hdr->src_addr));
+ uint16_t port;
+ uint16_t payload_len;
+ prox_rte_udp_hdr *udp_hdr;
+
+ rte_mov16((uint8_t *)&(ipv6_hdr->src_addr), (uint8_t *)&(ipv6_hdr->dst_addr)); // Copy dst into src
+ rte_mov16((uint8_t *)&(ipv6_hdr->dst_addr), (uint8_t *)&ip); // Copy src into dst
+ switch(ipv6_hdr->proto) {
+ case IPPROTO_TCP:
+ case IPPROTO_UDP:
+ payload_len = ipv6_hdr->payload_len;
+ udp_hdr = (prox_rte_udp_hdr *)(ipv6_hdr + 1);
+ if (unlikely(udp_hdr->dgram_len < payload_len)) {
+ plog_warn("Unexpected L4 len (%u) versus L3 payload len (%u) in IPv6 packet\n", udp_hdr->dgram_len, payload_len);
+ *out = OUT_DISCARD;
+ break;
+ }
+ port = udp_hdr->dst_port;
+ udp_hdr->dst_port = udp_hdr->src_port;
+ udp_hdr->src_port = port;
+ write_src_and_dst_mac(task, mbufs);
+ *out = 0;
+ break;
+ default:
+ plog_warn("Unsupported next hop %u in IPv6 packet\n", ipv6_hdr->proto);
+ *out = OUT_DISCARD;
+ break;
+ }
+}
+
static int handle_swap_bulk(struct task_base *tbase, struct rte_mbuf **mbufs, uint16_t n_pkts)
{
struct task_swap *task = (struct task_swap *)tbase;
prox_rte_ether_addr mac;
prox_rte_ipv4_hdr *ip_hdr;
prox_rte_udp_hdr *udp_hdr;
+ prox_rte_ipv6_hdr *ipv6_hdr;
struct gre_hdr *pgre;
prox_rte_ipv4_hdr *inner_ip_hdr;
uint32_t ip;
prox_rte_vlan_hdr *vlan;
uint16_t j;
struct igmpv2_hdr *pigmp;
+ prox_rte_icmp_hdr *picmp;
uint8_t type;
for (j = 0; j < n_pkts; ++j) {
}
mpls_len += sizeof(struct mpls_hdr);
ip_hdr = (prox_rte_ipv4_hdr *)(mpls + 1);
+ if (unlikely((ip_hdr->version_ihl >> 4) == 6)) {
+ ipv6_hdr = (prox_rte_ipv6_hdr *)(ip_hdr);
+ handle_ipv6(task, mbufs[j], ipv6_hdr, &out[j]);
+ continue;
+ }
break;
case ETYPE_8021ad:
qinq = (struct qinq_hdr *)hdr;
out[j] = OUT_DISCARD;
continue;
}
- ip_hdr = (prox_rte_ipv4_hdr *)(qinq + 1);
+ if (qinq->ether_type == ETYPE_IPv4) {
+ ip_hdr = (prox_rte_ipv4_hdr *)(qinq + 1);
+ } else if (qinq->ether_type == ETYPE_IPv6) {
+ ipv6_hdr = (prox_rte_ipv6_hdr *)(qinq + 1);
+ handle_ipv6(task, mbufs[j], ipv6_hdr, &out[j]);
+ continue;
+ } else {
+ plog_warn("Unsupported packet type\n");
+ out[j] = OUT_DISCARD;
+ continue;
+ }
break;
case ETYPE_VLAN:
vlan = (prox_rte_vlan_hdr *)(hdr + 1);
if (vlan->eth_proto == ETYPE_IPv4) {
ip_hdr = (prox_rte_ipv4_hdr *)(vlan + 1);
+ } else if (vlan->eth_proto == ETYPE_IPv6) {
+ ipv6_hdr = (prox_rte_ipv6_hdr *)(vlan + 1);
+ handle_ipv6(task, mbufs[j], ipv6_hdr, &out[j]);
+ continue;
} else if (vlan->eth_proto == ETYPE_VLAN) {
vlan = (prox_rte_vlan_hdr *)(vlan + 1);
if (vlan->eth_proto == ETYPE_IPv4) {
ip_hdr = (prox_rte_ipv4_hdr *)(vlan + 1);
}
else if (vlan->eth_proto == ETYPE_IPv6) {
- plog_warn("Unsupported IPv6\n");
- out[j] = OUT_DISCARD;
+ ipv6_hdr = (prox_rte_ipv6_hdr *)(vlan + 1);
+ handle_ipv6(task, mbufs[j], ipv6_hdr, &out[j]);
continue;
}
else {
ip_hdr = (prox_rte_ipv4_hdr *)(hdr + 1);
break;
case ETYPE_IPv6:
- plog_warn("Unsupported IPv6\n");
- out[j] = OUT_DISCARD;
+ ipv6_hdr = (prox_rte_ipv6_hdr *)(hdr + 1);
+ handle_ipv6(task, mbufs[j], ipv6_hdr, &out[j]);
continue;
case ETYPE_LLDP:
out[j] = OUT_DISCARD;
continue;
}
// TODO 2 : check packet is long enough for Ethernet + IP + UDP + extra header (VLAN, MPLS, ...)
+ // IPv4 packet
+
ip = ip_hdr->dst_addr;
+ if (unlikely((ip_hdr->version_ihl >> 4) != 4)) {
+ out[j] = OUT_DISCARD;
+ continue;
+ }
switch (ip_hdr->next_proto_id) {
case IPPROTO_GRE:
break;
case IPPROTO_UDP:
case IPPROTO_TCP:
- if (task->igmp_address && PROX_RTE_IS_IPV4_MCAST(rte_be_to_cpu_32(ip))) {
+ if (unlikely(task->igmp_address && PROX_RTE_IS_IPV4_MCAST(rte_be_to_cpu_32(ip)))) {
out[j] = OUT_DISCARD;
continue;
}
udp_hdr->src_port = port;
write_src_and_dst_mac(task, mbufs[j]);
break;
+ case IPPROTO_ICMP:
+ picmp = (prox_rte_icmp_hdr *)(ip_hdr + 1);
+ type = picmp->icmp_type;
+ if (type == PROX_RTE_IP_ICMP_ECHO_REQUEST) {
+ if (ip_hdr->dst_addr == task->local_ipv4) {
+ task->n_echo_req++;
+ if (rte_rdtsc() - task->last_echo_req_rcvd_tsc > rte_get_tsc_hz()) {
+ plog_info("Received %u Echo Request on IP "IPv4_BYTES_FMT" (last received from IP "IPv4_BYTES_FMT")\n", task->n_echo_req, IPv4_BYTES(((uint8_t*)&ip_hdr->dst_addr)), IPv4_BYTES(((uint8_t*)&ip_hdr->src_addr)));
+ task->n_echo_req = 0;
+ task->last_echo_req_rcvd_tsc = rte_rdtsc();
+ }
+ build_icmp_reply_message(tbase, mbufs[j]);
+ } else {
+ out[j] = OUT_DISCARD;
+ continue;
+ }
+ } else if (type == PROX_RTE_IP_ICMP_ECHO_REPLY) {
+ if (ip_hdr->dst_addr == task->local_ipv4) {
+ task->n_echo_rep++;
+ if (rte_rdtsc() - task->last_echo_rep_rcvd_tsc > rte_get_tsc_hz()) {
+ plog_info("Received %u Echo Reply on IP "IPv4_BYTES_FMT" (last received from IP "IPv4_BYTES_FMT")\n", task->n_echo_rep, IPv4_BYTES(((uint8_t*)&ip_hdr->dst_addr)), IPv4_BYTES(((uint8_t*)&ip_hdr->src_addr)));
+ task->n_echo_rep = 0;
+ task->last_echo_rep_rcvd_tsc = rte_rdtsc();
+ }
+ } else {
+ out[j] = OUT_DISCARD;
+ continue;
+ }
+ } else {
+ out[j] = OUT_DISCARD;
+ continue;
+ }
+ break;
case IPPROTO_IGMP:
pigmp = (struct igmpv2_hdr *)(ip_hdr + 1);
// TODO: check packet len
.handle = handle_swap_bulk,
.flag_features = 0,
.size = sizeof(struct task_swap),
+ .stop_last = stop_swap
};
__attribute__((constructor)) static void reg_task_swap(void)