X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=examples%2Fperformance-thread%2Fl3fwd-thread%2Fmain.c;h=8237ac65e6e2aaf67dc1336ef34f6e1b4cc561ae;hb=98a7ea332ba3da0f74ec951595d36a616165b255;hp=15c0a4de2d70d9b403edd9ead9b5774131f7250d;hpb=09419f235e099ecb265a590778fe64a685a2a241;p=dpdk.git diff --git a/examples/performance-thread/l3fwd-thread/main.c b/examples/performance-thread/l3fwd-thread/main.c index 15c0a4de2d..8237ac65e6 100644 --- a/examples/performance-thread/l3fwd-thread/main.c +++ b/examples/performance-thread/l3fwd-thread/main.c @@ -1,7 +1,7 @@ /*- * BSD LICENSE * - * Copyright(c) 2010-2015 Intel Corporation. All rights reserved. + * Copyright(c) 2010-2016 Intel Corporation. All rights reserved. * All rights reserved. * * Redistribution and use in source and binary forms, with or without @@ -90,6 +90,68 @@ #define APP_LOOKUP_METHOD APP_LOOKUP_LPM #endif +#ifndef __GLIBC__ /* sched_getcpu() is glibc specific */ +#define sched_getcpu() rte_lcore_id() +#endif + +static int +check_ptype(int portid) +{ + int i, ret; + int ipv4 = 0, ipv6 = 0; + + ret = rte_eth_dev_get_supported_ptypes(portid, RTE_PTYPE_L3_MASK, NULL, + 0); + if (ret <= 0) + return 0; + + uint32_t ptypes[ret]; + + ret = rte_eth_dev_get_supported_ptypes(portid, RTE_PTYPE_L3_MASK, + ptypes, ret); + for (i = 0; i < ret; ++i) { + if (ptypes[i] & RTE_PTYPE_L3_IPV4) + ipv4 = 1; + if (ptypes[i] & RTE_PTYPE_L3_IPV6) + ipv6 = 1; + } + + if (ipv4 && ipv6) + return 1; + + return 0; +} + +static inline void +parse_ptype(struct rte_mbuf *m) +{ + struct ether_hdr *eth_hdr; + uint32_t packet_type = RTE_PTYPE_UNKNOWN; + uint16_t ether_type; + + eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *); + ether_type = eth_hdr->ether_type; + if (ether_type == rte_cpu_to_be_16(ETHER_TYPE_IPv4)) + packet_type |= RTE_PTYPE_L3_IPV4_EXT_UNKNOWN; + else if (ether_type == rte_cpu_to_be_16(ETHER_TYPE_IPv6)) + packet_type |= RTE_PTYPE_L3_IPV6_EXT_UNKNOWN; + + m->packet_type = packet_type; +} + +static uint16_t +cb_parse_ptype(__rte_unused uint8_t port, __rte_unused uint16_t queue, + struct rte_mbuf *pkts[], uint16_t nb_pkts, + __rte_unused uint16_t max_pkts, __rte_unused void *user_param) +{ + unsigned int i; + + for (i = 0; i < nb_pkts; i++) + parse_ptype(pkts[i]); + + return nb_pkts; +} + /* * When set to zero, simple forwaring path is eanbled. * When set to one, optimized forwarding path is enabled. @@ -170,8 +232,9 @@ static __m128i val_eth[RTE_MAX_ETHPORTS]; /* mask of enabled ports */ static uint32_t enabled_port_mask; -static int promiscuous_on; /**< $et in promiscuous mode off by default. */ +static int promiscuous_on; /**< Set in promiscuous mode off by default. */ static int numa_on = 1; /**< NUMA is enabled by default. */ +static int parse_ptype_on; #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH) static int ipv6; /**< ipv6 is false by default. */ @@ -282,7 +345,7 @@ static struct rte_eth_conf port_conf = { .hw_ip_checksum = 1, /**< IP checksum offload enabled */ .hw_vlan_filter = 0, /**< VLAN filtering disabled */ .jumbo_frame = 0, /**< Jumbo Frame Support disabled */ - .hw_strip_crc = 0, /**< CRC stripped by hardware */ + .hw_strip_crc = 1, /**< CRC stripped by hardware */ }, .rx_adv_conf = { .rss_conf = { @@ -657,7 +720,7 @@ send_single_packet(struct rte_mbuf *m, uint8_t port) #if ((APP_LOOKUP_METHOD == APP_LOOKUP_LPM) && \ (ENABLE_MULTI_BUFFER_OPTIMIZE == 1)) -static inline __attribute__((always_inline)) void +static __rte_always_inline void send_packetsx4(uint8_t port, struct rte_mbuf *m[], uint32_t num) { @@ -698,12 +761,15 @@ send_packetsx4(uint8_t port, case 0: qconf->tx_mbufs[port].m_table[len + j] = m[j]; j++; + /* fall-through */ case 3: qconf->tx_mbufs[port].m_table[len + j] = m[j]; j++; + /* fall-through */ case 2: qconf->tx_mbufs[port].m_table[len + j] = m[j]; j++; + /* fall-through */ case 1: qconf->tx_mbufs[port].m_table[len + j] = m[j]; j++; @@ -725,12 +791,15 @@ send_packetsx4(uint8_t port, case 0: qconf->tx_mbufs[port].m_table[j] = m[n + j]; j++; + /* fall-through */ case 3: qconf->tx_mbufs[port].m_table[j] = m[n + j]; j++; + /* fall-through */ case 2: qconf->tx_mbufs[port].m_table[j] = m[n + j]; j++; + /* fall-through */ case 1: qconf->tx_mbufs[port].m_table[j] = m[n + j]; j++; @@ -850,7 +919,7 @@ static inline uint8_t get_ipv6_dst_port(void *ipv6_hdr, uint8_t portid, lookup6_struct_t *ipv6_l3fwd_lookup_struct) { - uint8_t next_hop; + uint32_t next_hop; return (uint8_t) ((rte_lpm6_lookup(ipv6_l3fwd_lookup_struct, ((struct ipv6_hdr *)ipv6_hdr)->dst_addr, &next_hop) == 0) ? @@ -996,7 +1065,7 @@ simple_ipv4_fwd_8pkts(struct rte_mbuf *m[8], uint8_t portid) const void *key_array[8] = {&key[0], &key[1], &key[2], &key[3], &key[4], &key[5], &key[6], &key[7]}; - rte_hash_lookup_multi(RTE_PER_LCORE(lcore_conf)->ipv4_lookup_struct, + rte_hash_lookup_bulk(RTE_PER_LCORE(lcore_conf)->ipv4_lookup_struct, &key_array[0], 8, ret); dst_port[0] = (uint8_t) ((ret[0] < 0) ? portid : ipv4_l3fwd_out_if[ret[0]]); dst_port[1] = (uint8_t) ((ret[1] < 0) ? portid : ipv4_l3fwd_out_if[ret[1]]); @@ -1150,7 +1219,7 @@ simple_ipv6_fwd_8pkts(struct rte_mbuf *m[8], uint8_t portid) const void *key_array[8] = {&key[0], &key[1], &key[2], &key[3], &key[4], &key[5], &key[6], &key[7]}; - rte_hash_lookup_multi(RTE_PER_LCORE(lcore_conf)->ipv6_lookup_struct, + rte_hash_lookup_bulk(RTE_PER_LCORE(lcore_conf)->ipv6_lookup_struct, &key_array[0], 4, ret); dst_port[0] = (uint8_t) ((ret[0] < 0) ? portid : ipv6_l3fwd_out_if[ret[0]]); dst_port[1] = (uint8_t) ((ret[1] < 0) ? portid : ipv6_l3fwd_out_if[ret[1]]); @@ -1218,7 +1287,7 @@ simple_ipv6_fwd_8pkts(struct rte_mbuf *m[8], uint8_t portid) } #endif /* APP_LOOKUP_METHOD */ -static inline __attribute__((always_inline)) void +static __rte_always_inline void l3fwd_simple_forward(struct rte_mbuf *m, uint8_t portid) { struct ether_hdr *eth_hdr; @@ -1306,8 +1375,8 @@ l3fwd_simple_forward(struct rte_mbuf *m, uint8_t portid) * If we encounter invalid IPV4 packet, then set destination port for it * to BAD_PORT value. */ -static inline __attribute__((always_inline)) void -rfc1812_process(struct ipv4_hdr *ipv4_hdr, uint32_t *dp, uint32_t ptype) +static __rte_always_inline void +rfc1812_process(struct ipv4_hdr *ipv4_hdr, uint16_t *dp, uint32_t ptype) { uint8_t ihl; @@ -1326,7 +1395,7 @@ rfc1812_process(struct ipv4_hdr *ipv4_hdr, uint32_t *dp, uint32_t ptype) } #else -#define rfc1812_process(mb, dp) do { } while (0) +#define rfc1812_process(mb, dp, ptype) do { } while (0) #endif /* DO_RFC_1812_CHECKS */ #endif /* APP_LOOKUP_LPM && ENABLE_MULTI_BUFFER_OPTIMIZE */ @@ -1334,37 +1403,35 @@ rfc1812_process(struct ipv4_hdr *ipv4_hdr, uint32_t *dp, uint32_t ptype) #if ((APP_LOOKUP_METHOD == APP_LOOKUP_LPM) && \ (ENABLE_MULTI_BUFFER_OPTIMIZE == 1)) -static inline __attribute__((always_inline)) uint16_t +static __rte_always_inline uint16_t get_dst_port(struct rte_mbuf *pkt, uint32_t dst_ipv4, uint8_t portid) { - uint32_t next_hop_ipv4; - uint8_t next_hop_ipv6; + uint32_t next_hop; struct ipv6_hdr *ipv6_hdr; struct ether_hdr *eth_hdr; if (RTE_ETH_IS_IPV4_HDR(pkt->packet_type)) { - if (rte_lpm_lookup(RTE_PER_LCORE(lcore_conf)->ipv4_lookup_struct, - dst_ipv4, &next_hop_ipv4) != 0) { - next_hop_ipv4 = portid; - return next_hop_ipv4; - } + return (uint16_t) ((rte_lpm_lookup( + RTE_PER_LCORE(lcore_conf)->ipv4_lookup_struct, dst_ipv4, + &next_hop) == 0) ? next_hop : portid); + } else if (RTE_ETH_IS_IPV6_HDR(pkt->packet_type)) { + eth_hdr = rte_pktmbuf_mtod(pkt, struct ether_hdr *); ipv6_hdr = (struct ipv6_hdr *)(eth_hdr + 1); - if (rte_lpm6_lookup(RTE_PER_LCORE(lcore_conf)->ipv6_lookup_struct, - ipv6_hdr->dst_addr, &next_hop_ipv6) != 0) { - next_hop_ipv6 = portid; - return next_hop_ipv6; - } - } else { - next_hop_ipv4 = portid; - return next_hop_ipv4; + + return (uint16_t) ((rte_lpm6_lookup( + RTE_PER_LCORE(lcore_conf)->ipv6_lookup_struct, + ipv6_hdr->dst_addr, &next_hop) == 0) ? + next_hop : portid); + } + return portid; } static inline void -process_packet(struct rte_mbuf *pkt, uint32_t *dst_port, uint8_t portid) +process_packet(struct rte_mbuf *pkt, uint16_t *dst_port, uint8_t portid) { struct ether_hdr *eth_hdr; struct ipv4_hdr *ipv4_hdr; @@ -1431,9 +1498,9 @@ processx4_step1(struct rte_mbuf *pkt[FWDSTEP], static inline void processx4_step2(__m128i dip, uint32_t ipv4_flag, - uint32_t portid, + uint8_t portid, struct rte_mbuf *pkt[FWDSTEP], - uint32_t dprt[FWDSTEP]) + uint16_t dprt[FWDSTEP]) { rte_xmm_t dst; const __m128i bswap_mask = _mm_set_epi8(12, 13, 14, 15, 8, 9, 10, 11, @@ -1445,7 +1512,11 @@ processx4_step2(__m128i dip, /* if all 4 packets are IPV4. */ if (likely(ipv4_flag)) { rte_lpm_lookupx4(RTE_PER_LCORE(lcore_conf)->ipv4_lookup_struct, dip, - dprt, portid); + dst.u32, portid); + + /* get rid of unused upper 16 bit for each dport. */ + dst.x = _mm_packs_epi32(dst.x, dst.x); + *(uint64_t *)dprt = dst.u64[0]; } else { dst.x = dip; dprt[0] = get_dst_port(pkt[0], dst.u32[0], portid); @@ -1460,7 +1531,7 @@ processx4_step2(__m128i dip, * Perform RFC1812 checks and updates for IPV4 packets. */ static inline void -processx4_step3(struct rte_mbuf *pkt[FWDSTEP], uint32_t dst_port[FWDSTEP]) +processx4_step3(struct rte_mbuf *pkt[FWDSTEP], uint16_t dst_port[FWDSTEP]) { __m128i te[FWDSTEP]; __m128i ve[FWDSTEP]; @@ -1533,7 +1604,7 @@ processx4_step3(struct rte_mbuf *pkt[FWDSTEP], uint32_t dst_port[FWDSTEP]) * Suppose we have array of destionation ports: * dst_port[] = {a, b, c, d,, e, ... } * dp1 should contain: , dp2: . - * We doing 4 comparisions at once and the result is 4 bit mask. + * We doing 4 comparisons at once and the result is 4 bit mask. * This mask is used as an index into prebuild array of pnum values. */ static inline uint16_t * @@ -1658,9 +1729,9 @@ port_groupx4(uint16_t pn[FWDSTEP + 1], uint16_t *lp, __m128i dp1, __m128i dp2) /* if dest port value has changed. */ if (v != GRPMSK) { - lp = pnum->u16 + gptbl[v].idx; - lp[0] = 1; pnum->u64 = gptbl[v].pnum; + pnum->u16[FWDSTEP] = 1; + lp = pnum->u16 + gptbl[v].idx; } return lp; @@ -1679,7 +1750,7 @@ process_burst(struct rte_mbuf *pkts_burst[MAX_PKT_BURST], int nb_rx, int32_t k; uint16_t dlp; uint16_t *lp; - uint32_t dst_port[MAX_PKT_BURST]; + uint16_t dst_port[MAX_PKT_BURST]; __m128i dip[MAX_PKT_BURST / FWDSTEP]; uint32_t ipv4_flag[MAX_PKT_BURST / FWDSTEP]; uint16_t pnum[MAX_PKT_BURST + 1]; @@ -1795,10 +1866,12 @@ process_burst(struct rte_mbuf *pkts_burst[MAX_PKT_BURST], int nb_rx, process_packet(pkts_burst[j], dst_port + j, portid); GROUP_PORT_STEP(dlp, dst_port, lp, pnum, j); j++; + /* fall-through */ case 2: process_packet(pkts_burst[j], dst_port + j, portid); GROUP_PORT_STEP(dlp, dst_port, lp, pnum, j); j++; + /* fall-through */ case 1: process_packet(pkts_burst[j], dst_port + j, portid); GROUP_PORT_STEP(dlp, dst_port, lp, pnum, j); @@ -2016,7 +2089,7 @@ lthread_tx_per_ring(void *dummy) */ SET_CPU_BUSY(tx_conf, CPU_POLL); nb_rx = rte_ring_sc_dequeue_burst(ring, (void **)pkts_burst, - MAX_PKT_BURST); + MAX_PKT_BURST, NULL); SET_CPU_IDLE(tx_conf, CPU_POLL); if (nb_rx > 0) { @@ -2152,7 +2225,7 @@ lthread_rx(void *dummy) ret = rte_ring_sp_enqueue_burst( rx_conf->ring[worker_id], (void **) pkts_burst, - nb_rx); + nb_rx, NULL); new_len = old_len + ret; @@ -2320,7 +2393,7 @@ pthread_tx(void *dummy) */ SET_CPU_BUSY(tx_conf, CPU_POLL); nb_rx = rte_ring_sc_dequeue_burst(tx_conf->ring, - (void **)pkts_burst, MAX_PKT_BURST); + (void **)pkts_burst, MAX_PKT_BURST, NULL); SET_CPU_IDLE(tx_conf, CPU_POLL); if (unlikely(nb_rx == 0)) { @@ -2392,7 +2465,7 @@ pthread_rx(void *dummy) SET_CPU_BUSY(rx_conf, CPU_PROCESS); worker_id = (worker_id + 1) % rx_conf->n_ring; n = rte_ring_sp_enqueue_burst(rx_conf->ring[worker_id], - (void **)pkts_burst, nb_rx); + (void **)pkts_burst, nb_rx, NULL); if (unlikely(n != nb_rx)) { uint32_t k; @@ -2607,6 +2680,7 @@ print_usage(const char *prgname) " [--rx (port,queue,lcore,thread)[,(port,queue,lcore,thread]]" " [--tx (lcore,thread)[,(lcore,thread]]" " [--enable-jumbo [--max-pkt-len PKTLEN]]\n" + " [--parse-ptype]\n\n" " -p PORTMASK: hexadecimal bitmask of ports to configure\n" " -P : enable promiscuous mode\n" " --rx (port,queue,lcore,thread): rx queues configuration\n" @@ -2618,7 +2692,8 @@ print_usage(const char *prgname) " --enable-jumbo: enable jumbo frame" " which max packet len is PKTLEN in decimal (64-9600)\n" " --hash-entry-num: specify the hash entry number in hexadecimal to be setup\n" - " --no-lthreads: turn off lthread model\n", + " --no-lthreads: turn off lthread model\n" + " --parse-ptype: set to use software to analyze packet type\n\n", prgname); } @@ -2837,6 +2912,7 @@ parse_eth_dest(const char *optarg) #define CMD_LINE_OPT_ENABLE_JUMBO "enable-jumbo" #define CMD_LINE_OPT_HASH_ENTRY_NUM "hash-entry-num" #define CMD_LINE_OPT_NO_LTHREADS "no-lthreads" +#define CMD_LINE_OPT_PARSE_PTYPE "parse-ptype" /* Parse the argument given in the command line of the application */ static int @@ -2856,6 +2932,7 @@ parse_args(int argc, char **argv) {CMD_LINE_OPT_ENABLE_JUMBO, 0, 0, 0}, {CMD_LINE_OPT_HASH_ENTRY_NUM, 1, 0, 0}, {CMD_LINE_OPT_NO_LTHREADS, 0, 0, 0}, + {CMD_LINE_OPT_PARSE_PTYPE, 0, 0, 0}, {NULL, 0, 0, 0} }; @@ -2932,6 +3009,12 @@ parse_args(int argc, char **argv) lthreads_on = 0; } + if (!strncmp(lgopts[option_index].name, CMD_LINE_OPT_PARSE_PTYPE, + sizeof(CMD_LINE_OPT_PARSE_PTYPE))) { + printf("software packet type parsing enabled\n"); + parse_ptype_on = 1; + } + if (!strncmp(lgopts[option_index].name, CMD_LINE_OPT_ENABLE_JUMBO, sizeof(CMD_LINE_OPT_ENABLE_JUMBO))) { struct option lenopts = {"max-pkt-len", required_argument, 0, @@ -2980,7 +3063,7 @@ parse_args(int argc, char **argv) argv[optind-1] = prgname; ret = optind-1; - optind = 0; /* reset getopt lib */ + optind = 1; /* reset getopt lib */ return ret; } @@ -3482,8 +3565,6 @@ main(int argc, char **argv) rte_exit(EXIT_FAILURE, "init_rx_rings failed\n"); nb_ports = rte_eth_dev_count(); - if (nb_ports > RTE_MAX_ETHPORTS) - nb_ports = RTE_MAX_ETHPORTS; if (check_port_config(nb_ports) < 0) rte_exit(EXIT_FAILURE, "check_port_config failed\n"); @@ -3622,6 +3703,31 @@ main(int argc, char **argv) rte_eth_promiscuous_enable(portid); } + for (i = 0; i < n_rx_thread; i++) { + lcore_id = rx_thread[i].conf.lcore_id; + if (rte_lcore_is_enabled(lcore_id) == 0) + continue; + + /* check if hw packet type is supported */ + for (queue = 0; queue < rx_thread[i].n_rx_queue; ++queue) { + portid = rx_thread[i].rx_queue_list[queue].port_id; + queueid = rx_thread[i].rx_queue_list[queue].queue_id; + + if (parse_ptype_on) { + if (!rte_eth_add_rx_callback(portid, queueid, + cb_parse_ptype, NULL)) + rte_exit(EXIT_FAILURE, + "Failed to add rx callback: " + "port=%d\n", portid); + } else if (!check_ptype(portid)) + rte_exit(EXIT_FAILURE, + "Port %d cannot parse packet type.\n\n" + "Please add --parse-ptype to use sw " + "packet type analyzer.\n\n", + portid); + } + } + check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask); if (lthreads_on) {