diff options
author | Christian Ehrhardt <christian.ehrhardt@canonical.com> | 2016-07-06 09:22:35 +0200 |
---|---|---|
committer | Christian Ehrhardt <christian.ehrhardt@canonical.com> | 2016-07-06 16:09:40 +0200 |
commit | 8b25d1ad5d2264bdfc2818c7bda74ee2697df6db (patch) | |
tree | 8c3c769777f7e66a2d1ba7dd7651b563cfde370b /examples/ipsec-secgw/ipsec-secgw.c | |
parent | 97f17497d162afdb82c8704bf097f0fee3724b2e (diff) |
Imported Upstream version 16.07-rc1
Change-Id: I40a523e52f12e8496fdd69e902824b0226c303de
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
Diffstat (limited to 'examples/ipsec-secgw/ipsec-secgw.c')
-rw-r--r-- | examples/ipsec-secgw/ipsec-secgw.c | 338 |
1 files changed, 231 insertions, 107 deletions
diff --git a/examples/ipsec-secgw/ipsec-secgw.c b/examples/ipsec-secgw/ipsec-secgw.c index 00ab2d84..f78743d0 100644 --- a/examples/ipsec-secgw/ipsec-secgw.c +++ b/examples/ipsec-secgw/ipsec-secgw.c @@ -38,6 +38,7 @@ #include <sys/types.h> #include <netinet/in.h> #include <netinet/ip.h> +#include <netinet/ip6.h> #include <string.h> #include <sys/queue.h> #include <stdarg.h> @@ -65,6 +66,7 @@ #include <rte_mbuf.h> #include <rte_acl.h> #include <rte_lpm.h> +#include <rte_lpm6.h> #include <rte_hash.h> #include <rte_jhash.h> #include <rte_cryptodev.h> @@ -192,7 +194,8 @@ struct lcore_conf { struct buffer tx_mbufs[RTE_MAX_ETHPORTS]; struct ipsec_ctx inbound; struct ipsec_ctx outbound; - struct rt_ctx *rt_ctx; + struct rt_ctx *rt4_ctx; + struct rt_ctx *rt6_ctx; } __rte_cache_aligned; static struct lcore_conf lcore_conf[RTE_MAX_LCORE]; @@ -230,27 +233,39 @@ struct traffic_type { }; struct ipsec_traffic { - struct traffic_type ipsec4; - struct traffic_type ipv4; + struct traffic_type ipsec; + struct traffic_type ip4; + struct traffic_type ip6; }; static inline void prepare_one_packet(struct rte_mbuf *pkt, struct ipsec_traffic *t) { uint8_t *nlp; + struct ether_hdr *eth; - if (RTE_ETH_IS_IPV4_HDR(pkt->packet_type)) { - rte_pktmbuf_adj(pkt, ETHER_HDR_LEN); - nlp = rte_pktmbuf_mtod_offset(pkt, uint8_t *, - offsetof(struct ip, ip_p)); + eth = rte_pktmbuf_mtod(pkt, struct ether_hdr *); + if (eth->ether_type == rte_cpu_to_be_16(ETHER_TYPE_IPv4)) { + nlp = (uint8_t *)rte_pktmbuf_adj(pkt, ETHER_HDR_LEN); + nlp = RTE_PTR_ADD(nlp, offsetof(struct ip, ip_p)); if (*nlp == IPPROTO_ESP) - t->ipsec4.pkts[(t->ipsec4.num)++] = pkt; + t->ipsec.pkts[(t->ipsec.num)++] = pkt; else { - t->ipv4.data[t->ipv4.num] = nlp; - t->ipv4.pkts[(t->ipv4.num)++] = pkt; + t->ip4.data[t->ip4.num] = nlp; + t->ip4.pkts[(t->ip4.num)++] = pkt; + } + } else if (eth->ether_type == rte_cpu_to_be_16(ETHER_TYPE_IPv6)) { + nlp = (uint8_t *)rte_pktmbuf_adj(pkt, ETHER_HDR_LEN); + nlp = RTE_PTR_ADD(nlp, offsetof(struct ip6_hdr, ip6_nxt)); + if (*nlp == IPPROTO_ESP) + t->ipsec.pkts[(t->ipsec.num)++] = pkt; + else { + t->ip6.data[t->ip6.num] = nlp; + t->ip6.pkts[(t->ip6.num)++] = pkt; } } else { /* Unknown/Unsupported type, drop the packet */ + RTE_LOG(ERR, IPSEC, "Unsupported packet type\n"); rte_pktmbuf_free(pkt); } } @@ -261,8 +276,9 @@ prepare_traffic(struct rte_mbuf **pkts, struct ipsec_traffic *t, { int32_t i; - t->ipsec4.num = 0; - t->ipv4.num = 0; + t->ipsec.num = 0; + t->ip4.num = 0; + t->ip6.num = 0; for (i = 0; i < (nb_pkts - PREFETCH_OFFSET); i++) { rte_prefetch0(rte_pktmbuf_mtod(pkts[i + PREFETCH_OFFSET], @@ -277,14 +293,27 @@ prepare_traffic(struct rte_mbuf **pkts, struct ipsec_traffic *t, static inline void prepare_tx_pkt(struct rte_mbuf *pkt, uint8_t port) { - pkt->ol_flags |= PKT_TX_IP_CKSUM | PKT_TX_IPV4; - pkt->l3_len = sizeof(struct ip); - pkt->l2_len = ETHER_HDR_LEN; + struct ip *ip; + struct ether_hdr *ethhdr; + + ip = rte_pktmbuf_mtod(pkt, struct ip *); + + ethhdr = (struct ether_hdr *)rte_pktmbuf_prepend(pkt, ETHER_HDR_LEN); - struct ether_hdr *ethhdr = (struct ether_hdr *)rte_pktmbuf_prepend(pkt, - ETHER_HDR_LEN); + if (ip->ip_v == IPVERSION) { + pkt->ol_flags |= PKT_TX_IP_CKSUM | PKT_TX_IPV4; + pkt->l3_len = sizeof(struct ip); + pkt->l2_len = ETHER_HDR_LEN; + + ethhdr->ether_type = rte_cpu_to_be_16(ETHER_TYPE_IPv4); + } else { + pkt->ol_flags |= PKT_TX_IPV6; + pkt->l3_len = sizeof(struct ip6_hdr); + pkt->l2_len = ETHER_HDR_LEN; + + ethhdr->ether_type = rte_cpu_to_be_16(ETHER_TYPE_IPv6); + } - ethhdr->ether_type = rte_cpu_to_be_16(ETHER_TYPE_IPv4); memcpy(ðhdr->s_addr, ðaddr_tbl[port].src, sizeof(struct ether_addr)); memcpy(ðhdr->d_addr, ðaddr_tbl[port].dst, @@ -298,7 +327,7 @@ prepare_tx_burst(struct rte_mbuf *pkts[], uint16_t nb_pkts, uint8_t port) const int32_t prefetch_offset = 2; for (i = 0; i < (nb_pkts - prefetch_offset); i++) { - rte_prefetch0(pkts[i + prefetch_offset]->cacheline1); + rte_mbuf_prefetch_part2(pkts[i + prefetch_offset]); prepare_tx_pkt(pkts[i], port); } /* Process left packets */ @@ -355,94 +384,133 @@ send_single_packet(struct rte_mbuf *m, uint8_t port) } static inline void -process_pkts_inbound(struct ipsec_ctx *ipsec_ctx, - struct ipsec_traffic *traffic) +inbound_sp_sa(struct sp_ctx *sp, struct sa_ctx *sa, struct traffic_type *ip) { struct rte_mbuf *m; - uint16_t idx, nb_pkts_in, i, j; - uint32_t sa_idx, res; - - nb_pkts_in = ipsec_inbound(ipsec_ctx, traffic->ipsec4.pkts, - traffic->ipsec4.num, MAX_PKT_BURST); + uint32_t i, j, res, sa_idx; - /* SP/ACL Inbound check ipsec and ipv4 */ - for (i = 0; i < nb_pkts_in; i++) { - idx = traffic->ipv4.num++; - m = traffic->ipsec4.pkts[i]; - traffic->ipv4.pkts[idx] = m; - traffic->ipv4.data[idx] = rte_pktmbuf_mtod_offset(m, - uint8_t *, offsetof(struct ip, ip_p)); - } + if (ip->num == 0) + return; - rte_acl_classify((struct rte_acl_ctx *)ipsec_ctx->sp_ctx, - traffic->ipv4.data, traffic->ipv4.res, - traffic->ipv4.num, DEFAULT_MAX_CATEGORIES); + rte_acl_classify((struct rte_acl_ctx *)sp, ip->data, ip->res, + ip->num, DEFAULT_MAX_CATEGORIES); j = 0; - for (i = 0; i < traffic->ipv4.num - nb_pkts_in; i++) { - m = traffic->ipv4.pkts[i]; - res = traffic->ipv4.res[i]; - if (res & ~BYPASS) { + for (i = 0; i < ip->num; i++) { + m = ip->pkts[i]; + res = ip->res[i]; + if (res & DISCARD) { rte_pktmbuf_free(m); continue; } - traffic->ipv4.pkts[j++] = m; - } - /* Check return SA SPI matches pkt SPI */ - for ( ; i < traffic->ipv4.num; i++) { - m = traffic->ipv4.pkts[i]; - sa_idx = traffic->ipv4.res[i] & PROTECT_MASK; - if (sa_idx == 0 || !inbound_sa_check(ipsec_ctx->sa_ctx, - m, sa_idx)) { + if (res & BYPASS) { + ip->pkts[j++] = m; + continue; + } + /* Check return SA SPI matches pkt SPI */ + sa_idx = ip->res[i] & PROTECT_MASK; + if (sa_idx == 0 || !inbound_sa_check(sa, m, sa_idx)) { rte_pktmbuf_free(m); continue; } - traffic->ipv4.pkts[j++] = m; + ip->pkts[j++] = m; } - traffic->ipv4.num = j; + ip->num = j; } static inline void -process_pkts_outbound(struct ipsec_ctx *ipsec_ctx, +process_pkts_inbound(struct ipsec_ctx *ipsec_ctx, struct ipsec_traffic *traffic) { struct rte_mbuf *m; - uint16_t idx, nb_pkts_out, i, j; - uint32_t sa_idx, res; + uint16_t idx, nb_pkts_in, i; - rte_acl_classify((struct rte_acl_ctx *)ipsec_ctx->sp_ctx, - traffic->ipv4.data, traffic->ipv4.res, - traffic->ipv4.num, DEFAULT_MAX_CATEGORIES); + nb_pkts_in = ipsec_inbound(ipsec_ctx, traffic->ipsec.pkts, + traffic->ipsec.num, MAX_PKT_BURST); - /* Drop any IPsec traffic from protected ports */ - for (i = 0; i < traffic->ipsec4.num; i++) - rte_pktmbuf_free(traffic->ipsec4.pkts[i]); + /* SP/ACL Inbound check ipsec and ip4 */ + for (i = 0; i < nb_pkts_in; i++) { + m = traffic->ipsec.pkts[i]; + struct ip *ip = rte_pktmbuf_mtod(m, struct ip *); + if (ip->ip_v == IPVERSION) { + idx = traffic->ip4.num++; + traffic->ip4.pkts[idx] = m; + traffic->ip4.data[idx] = rte_pktmbuf_mtod_offset(m, + uint8_t *, offsetof(struct ip, ip_p)); + } else if (ip->ip_v == IP6_VERSION) { + idx = traffic->ip6.num++; + traffic->ip6.pkts[idx] = m; + traffic->ip6.data[idx] = rte_pktmbuf_mtod_offset(m, + uint8_t *, + offsetof(struct ip6_hdr, ip6_nxt)); + } else + rte_pktmbuf_free(m); + } + + inbound_sp_sa(ipsec_ctx->sp4_ctx, ipsec_ctx->sa_ctx, &traffic->ip4); + + inbound_sp_sa(ipsec_ctx->sp6_ctx, ipsec_ctx->sa_ctx, &traffic->ip6); +} + +static inline void +outbound_sp(struct sp_ctx *sp, struct traffic_type *ip, + struct traffic_type *ipsec) +{ + struct rte_mbuf *m; + uint32_t i, j, sa_idx; + + if (ip->num == 0) + return; - traffic->ipsec4.num = 0; + rte_acl_classify((struct rte_acl_ctx *)sp, ip->data, ip->res, + ip->num, DEFAULT_MAX_CATEGORIES); j = 0; - for (i = 0; i < traffic->ipv4.num; i++) { - m = traffic->ipv4.pkts[i]; - res = traffic->ipv4.res[i]; - sa_idx = res & PROTECT_MASK; - if ((res == 0) || (res & DISCARD)) + for (i = 0; i < ip->num; i++) { + m = ip->pkts[i]; + sa_idx = ip->res[i] & PROTECT_MASK; + if ((ip->res[i] == 0) || (ip->res[i] & DISCARD)) rte_pktmbuf_free(m); else if (sa_idx != 0) { - traffic->ipsec4.res[traffic->ipsec4.num] = sa_idx; - traffic->ipsec4.pkts[traffic->ipsec4.num++] = m; + ipsec->res[ipsec->num] = sa_idx; + ipsec->pkts[ipsec->num++] = m; } else /* BYPASS */ - traffic->ipv4.pkts[j++] = m; + ip->pkts[j++] = m; } - traffic->ipv4.num = j; + ip->num = j; +} + +static inline void +process_pkts_outbound(struct ipsec_ctx *ipsec_ctx, + struct ipsec_traffic *traffic) +{ + struct rte_mbuf *m; + uint16_t idx, nb_pkts_out, i; + + /* Drop any IPsec traffic from protected ports */ + for (i = 0; i < traffic->ipsec.num; i++) + rte_pktmbuf_free(traffic->ipsec.pkts[i]); - nb_pkts_out = ipsec_outbound(ipsec_ctx, traffic->ipsec4.pkts, - traffic->ipsec4.res, traffic->ipsec4.num, + traffic->ipsec.num = 0; + + outbound_sp(ipsec_ctx->sp4_ctx, &traffic->ip4, &traffic->ipsec); + + outbound_sp(ipsec_ctx->sp6_ctx, &traffic->ip6, &traffic->ipsec); + + nb_pkts_out = ipsec_outbound(ipsec_ctx, traffic->ipsec.pkts, + traffic->ipsec.res, traffic->ipsec.num, MAX_PKT_BURST); for (i = 0; i < nb_pkts_out; i++) { - idx = traffic->ipv4.num++; - m = traffic->ipsec4.pkts[i]; - traffic->ipv4.pkts[idx] = m; + m = traffic->ipsec.pkts[i]; + struct ip *ip = rte_pktmbuf_mtod(m, struct ip *); + if (ip->ip_v == IPVERSION) { + idx = traffic->ip4.num++; + traffic->ip4.pkts[idx] = m; + } else { + idx = traffic->ip6.num++; + traffic->ip6.pkts[idx] = m; + } } } @@ -450,47 +518,72 @@ static inline void process_pkts_inbound_nosp(struct ipsec_ctx *ipsec_ctx, struct ipsec_traffic *traffic) { - uint16_t nb_pkts_in, i; + struct rte_mbuf *m; + uint32_t nb_pkts_in, i, idx; /* Drop any IPv4 traffic from unprotected ports */ - for (i = 0; i < traffic->ipv4.num; i++) - rte_pktmbuf_free(traffic->ipv4.pkts[i]); + for (i = 0; i < traffic->ip4.num; i++) + rte_pktmbuf_free(traffic->ip4.pkts[i]); - traffic->ipv4.num = 0; + traffic->ip4.num = 0; - nb_pkts_in = ipsec_inbound(ipsec_ctx, traffic->ipsec4.pkts, - traffic->ipsec4.num, MAX_PKT_BURST); + /* Drop any IPv6 traffic from unprotected ports */ + for (i = 0; i < traffic->ip6.num; i++) + rte_pktmbuf_free(traffic->ip6.pkts[i]); - for (i = 0; i < nb_pkts_in; i++) - traffic->ipv4.pkts[i] = traffic->ipsec4.pkts[i]; + traffic->ip6.num = 0; - traffic->ipv4.num = nb_pkts_in; + nb_pkts_in = ipsec_inbound(ipsec_ctx, traffic->ipsec.pkts, + traffic->ipsec.num, MAX_PKT_BURST); + + for (i = 0; i < nb_pkts_in; i++) { + m = traffic->ipsec.pkts[i]; + struct ip *ip = rte_pktmbuf_mtod(m, struct ip *); + if (ip->ip_v == IPVERSION) { + idx = traffic->ip4.num++; + traffic->ip4.pkts[idx] = m; + } else { + idx = traffic->ip6.num++; + traffic->ip6.pkts[idx] = m; + } + } } static inline void process_pkts_outbound_nosp(struct ipsec_ctx *ipsec_ctx, struct ipsec_traffic *traffic) { - uint16_t nb_pkts_out, i; + struct rte_mbuf *m; + uint32_t nb_pkts_out, i; + struct ip *ip; /* Drop any IPsec traffic from protected ports */ - for (i = 0; i < traffic->ipsec4.num; i++) - rte_pktmbuf_free(traffic->ipsec4.pkts[i]); + for (i = 0; i < traffic->ipsec.num; i++) + rte_pktmbuf_free(traffic->ipsec.pkts[i]); - traffic->ipsec4.num = 0; + traffic->ipsec.num = 0; - for (i = 0; i < traffic->ipv4.num; i++) - traffic->ipv4.res[i] = single_sa_idx; + for (i = 0; i < traffic->ip4.num; i++) + traffic->ip4.res[i] = single_sa_idx; - nb_pkts_out = ipsec_outbound(ipsec_ctx, traffic->ipv4.pkts, - traffic->ipv4.res, traffic->ipv4.num, + for (i = 0; i < traffic->ip6.num; i++) + traffic->ip6.res[i] = single_sa_idx; + + nb_pkts_out = ipsec_outbound(ipsec_ctx, traffic->ip4.pkts, + traffic->ip4.res, traffic->ip4.num, MAX_PKT_BURST); - traffic->ipv4.num = nb_pkts_out; + /* They all sue the same SA (ip4 or ip6 tunnel) */ + m = traffic->ipsec.pkts[i]; + ip = rte_pktmbuf_mtod(m, struct ip *); + if (ip->ip_v == IPVERSION) + traffic->ip4.num = nb_pkts_out; + else + traffic->ip6.num = nb_pkts_out; } static inline void -route_pkts(struct rt_ctx *rt_ctx, struct rte_mbuf *pkts[], uint8_t nb_pkts) +route4_pkts(struct rt_ctx *rt_ctx, struct rte_mbuf *pkts[], uint8_t nb_pkts) { uint32_t hop[MAX_PKT_BURST * 2]; uint32_t dst_ip[MAX_PKT_BURST * 2]; @@ -518,6 +611,35 @@ route_pkts(struct rt_ctx *rt_ctx, struct rte_mbuf *pkts[], uint8_t nb_pkts) } static inline void +route6_pkts(struct rt_ctx *rt_ctx, struct rte_mbuf *pkts[], uint8_t nb_pkts) +{ + int16_t hop[MAX_PKT_BURST * 2]; + uint8_t dst_ip[MAX_PKT_BURST * 2][16]; + uint8_t *ip6_dst; + uint16_t i, offset; + + if (nb_pkts == 0) + return; + + for (i = 0; i < nb_pkts; i++) { + offset = offsetof(struct ip6_hdr, ip6_dst); + ip6_dst = rte_pktmbuf_mtod_offset(pkts[i], uint8_t *, offset); + memcpy(&dst_ip[i][0], ip6_dst, 16); + } + + rte_lpm6_lookup_bulk_func((struct rte_lpm6 *)rt_ctx, dst_ip, + hop, nb_pkts); + + for (i = 0; i < nb_pkts; i++) { + if (hop[i] == -1) { + rte_pktmbuf_free(pkts[i]); + continue; + } + send_single_packet(pkts[i], hop[i] & 0xff); + } +} + +static inline void process_pkts(struct lcore_conf *qconf, struct rte_mbuf **pkts, uint8_t nb_pkts, uint8_t portid) { @@ -525,7 +647,7 @@ process_pkts(struct lcore_conf *qconf, struct rte_mbuf **pkts, prepare_traffic(pkts, &traffic, nb_pkts); - if (single_sa) { + if (unlikely(single_sa)) { if (UNPROTECTED_PORT(portid)) process_pkts_inbound_nosp(&qconf->inbound, &traffic); else @@ -537,7 +659,8 @@ process_pkts(struct lcore_conf *qconf, struct rte_mbuf **pkts, process_pkts_outbound(&qconf->outbound, &traffic); } - route_pkts(qconf->rt_ctx, traffic.ipv4.pkts, traffic.ipv4.num); + route4_pkts(qconf->rt4_ctx, traffic.ip4.pkts, traffic.ip4.num); + route6_pkts(qconf->rt6_ctx, traffic.ip6.pkts, traffic.ip6.num); } static inline void @@ -576,12 +699,15 @@ main_loop(__attribute__((unused)) void *dummy) rxql = qconf->rx_queue_list; socket_id = rte_lcore_to_socket_id(lcore_id); - qconf->rt_ctx = socket_ctx[socket_id].rt_ipv4; - qconf->inbound.sp_ctx = socket_ctx[socket_id].sp_ipv4_in; - qconf->inbound.sa_ctx = socket_ctx[socket_id].sa_ipv4_in; + qconf->rt4_ctx = socket_ctx[socket_id].rt_ip4; + qconf->rt6_ctx = socket_ctx[socket_id].rt_ip6; + qconf->inbound.sp4_ctx = socket_ctx[socket_id].sp_ip4_in; + qconf->inbound.sp6_ctx = socket_ctx[socket_id].sp_ip6_in; + qconf->inbound.sa_ctx = socket_ctx[socket_id].sa_in; qconf->inbound.cdev_map = cdev_map_in; - qconf->outbound.sp_ctx = socket_ctx[socket_id].sp_ipv4_out; - qconf->outbound.sa_ctx = socket_ctx[socket_id].sa_ipv4_out; + qconf->outbound.sp4_ctx = socket_ctx[socket_id].sp_ip4_out; + qconf->outbound.sp6_ctx = socket_ctx[socket_id].sp_ip6_out; + qconf->outbound.sa_ctx = socket_ctx[socket_id].sa_out; qconf->outbound.cdev_map = cdev_map_out; if (qconf->nb_rx_queue == 0) { @@ -636,8 +762,6 @@ check_params(void) } nb_ports = rte_eth_dev_count(); - if (nb_ports > RTE_MAX_ETHPORTS) - nb_ports = RTE_MAX_ETHPORTS; for (i = 0; i < nb_lcore_params; ++i) { lcore = lcore_params[i].lcore_id; @@ -762,7 +886,7 @@ parse_config(const char *q_arg) FLD_LCORE, _NUM_FLD }; - int long int_fld[_NUM_FLD]; + unsigned long int_fld[_NUM_FLD]; char *str_fld[_NUM_FLD]; int32_t i; uint32_t size; @@ -1286,8 +1410,6 @@ main(int32_t argc, char **argv) unprotected_port_mask); nb_ports = rte_eth_dev_count(); - if (nb_ports > RTE_MAX_ETHPORTS) - nb_ports = RTE_MAX_ETHPORTS; if (check_params() < 0) rte_exit(EXIT_FAILURE, "check_params failed\n"); @@ -1313,7 +1435,9 @@ main(int32_t argc, char **argv) sa_init(&socket_ctx[socket_id], socket_id, ep); - sp_init(&socket_ctx[socket_id], socket_id, ep); + sp4_init(&socket_ctx[socket_id], socket_id, ep); + + sp6_init(&socket_ctx[socket_id], socket_id, ep); rt_init(&socket_ctx[socket_id], socket_id, ep); |