#define _GNU_SOURCE
#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
#include <inttypes.h>
#include <sys/types.h>
#include <string.h>
#include <sys/queue.h>
#include <stdarg.h>
#include <errno.h>
#include <getopt.h>
#include <rte_vect.h>
#include <cmdline_parse.h>
#include <cmdline_parse_etheraddr.h>
#include <lthread_api.h>
#define APP_LOOKUP_EXACT_MATCH 0
#define APP_LOOKUP_LPM 1
#define DO_RFC_1812_CHECKS
#define APP_CPU_LOAD 1
#ifndef APP_LOOKUP_METHOD
#define APP_LOOKUP_METHOD APP_LOOKUP_LPM
#endif
#if ((APP_LOOKUP_METHOD == APP_LOOKUP_LPM) && !defined(__SSE4_1__))
#define ENABLE_MULTI_BUFFER_OPTIMIZE 0
#else
#define ENABLE_MULTI_BUFFER_OPTIMIZE 1
#endif
#if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
#elif (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
#else
#error "APP_LOOKUP_METHOD set to incorrect value"
#endif
#define RTE_LOGTYPE_L3FWD RTE_LOGTYPE_USER1
#define MAX_JUMBO_PKT_LEN 9600
#define IPV6_ADDR_LEN 16
#define MEMPOOL_CACHE_SIZE 256
#define NB_MBUF RTE_MAX(\
(nb_ports*nb_rx_queue*RTE_TEST_RX_DESC_DEFAULT + \
nb_ports*nb_lcores*MAX_PKT_BURST + \
nb_ports*n_tx_queue*RTE_TEST_TX_DESC_DEFAULT + \
nb_lcores*MEMPOOL_CACHE_SIZE), \
(unsigned)8192)
#define MAX_PKT_BURST 32
#define BURST_TX_DRAIN_US 100
#define MAX_TX_BURST (MAX_PKT_BURST / 2)
#define BURST_SIZE MAX_TX_BURST
#define NB_SOCKETS 8
#define PREFETCH_OFFSET 3
#define BAD_PORT ((uint16_t)-1)
#define FWDSTEP 4
#define RTE_TEST_RX_DESC_DEFAULT 128
#define RTE_TEST_TX_DESC_DEFAULT 128
static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
static uint64_t dest_eth_addr[RTE_MAX_ETHPORTS];
static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
static __m128i val_eth[RTE_MAX_ETHPORTS];
#define MASK_ETH 0x3f
static uint32_t enabled_port_mask;
static int promiscuous_on;
static int numa_on = 1;
#if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
static int ipv6;
#endif
#if (APP_CPU_LOAD == 1)
#define MAX_CPU RTE_MAX_LCORE
#define CPU_LOAD_TIMEOUT_US (5 * 1000 * 1000)
#define CPU_PROCESS 0
#define CPU_POLL 1
#define MAX_CPU_COUNTER 2
struct cpu_load {
uint16_t n_cpu;
uint64_t counter;
uint64_t hits[MAX_CPU_COUNTER][MAX_CPU];
static struct cpu_load cpu_load;
static int cpu_load_lcore_id = -1;
#define SET_CPU_BUSY(thread, counter) \
thread->conf.busy[counter] = 1
#define SET_CPU_IDLE(thread, counter) \
thread->conf.busy[counter] = 0
#define IS_CPU_BUSY(thread, counter) \
(thread->conf.busy[counter] > 0)
#else
#define SET_CPU_BUSY(thread, counter)
#define SET_CPU_IDLE(thread, counter)
#define IS_CPU_BUSY(thread, counter) 0
#endif
struct mbuf_table {
uint16_t len;
struct rte_mbuf *m_table[MAX_PKT_BURST];
};
struct lcore_rx_queue {
uint8_t port_id;
uint8_t queue_id;
#define MAX_RX_QUEUE_PER_LCORE 16
#define MAX_TX_QUEUE_PER_PORT RTE_MAX_ETHPORTS
#define MAX_RX_QUEUE_PER_PORT 128
#define MAX_LCORE_PARAMS 1024
struct rx_thread_params {
uint8_t port_id;
uint8_t queue_id;
uint8_t lcore_id;
uint8_t thread_id;
static struct rx_thread_params rx_thread_params_array[MAX_LCORE_PARAMS];
static struct rx_thread_params rx_thread_params_array_default[] = {
{0, 0, 2, 0},
{0, 1, 2, 1},
{0, 2, 2, 2},
{1, 0, 2, 3},
{1, 1, 2, 4},
{1, 2, 2, 5},
{2, 0, 2, 6},
{3, 0, 3, 7},
{3, 1, 3, 8},
};
static struct rx_thread_params *rx_thread_params =
rx_thread_params_array_default;
static uint16_t nb_rx_thread_params =
RTE_DIM(rx_thread_params_array_default);
struct tx_thread_params {
uint8_t lcore_id;
uint8_t thread_id;
static struct tx_thread_params tx_thread_params_array[MAX_LCORE_PARAMS];
static struct tx_thread_params tx_thread_params_array_default[] = {
{4, 0},
{5, 1},
{6, 2},
{7, 3},
{8, 4},
{9, 5},
{10, 6},
{11, 7},
{12, 8},
};
static struct tx_thread_params *tx_thread_params =
tx_thread_params_array_default;
static uint16_t nb_tx_thread_params =
RTE_DIM(tx_thread_params_array_default);
.split_hdr_size = 0,
.header_split = 0,
.hw_ip_checksum = 1,
.hw_vlan_filter = 0,
.jumbo_frame = 0,
.hw_strip_crc = 0,
},
.rx_adv_conf = {
.rss_conf = {
.rss_key = NULL,
.rss_hf = ETH_RSS_TCP,
},
},
.txmode = {
},
};
#if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
#ifdef RTE_MACHINE_CPUFLAG_SSE4_2
#define DEFAULT_HASH_FUNC rte_hash_crc
#else
#define DEFAULT_HASH_FUNC rte_jhash
#endif
struct ipv4_5tuple {
uint32_t ip_dst;
uint32_t ip_src;
uint16_t port_dst;
uint16_t port_src;
uint8_t proto;
} __attribute__((__packed__));
union ipv4_5tuple_host {
struct {
uint8_t pad0;
uint8_t proto;
uint16_t pad1;
uint32_t ip_src;
uint32_t ip_dst;
uint16_t port_src;
uint16_t port_dst;
};
__m128i xmm;
};
#define XMM_NUM_IN_IPV6_5TUPLE 3
struct ipv6_5tuple {
uint8_t ip_dst[IPV6_ADDR_LEN];
uint8_t ip_src[IPV6_ADDR_LEN];
uint16_t port_dst;
uint16_t port_src;
uint8_t proto;
} __attribute__((__packed__));
union ipv6_5tuple_host {
struct {
uint16_t pad0;
uint8_t proto;
uint8_t pad1;
uint8_t ip_src[IPV6_ADDR_LEN];
uint8_t ip_dst[IPV6_ADDR_LEN];
uint16_t port_src;
uint16_t port_dst;
uint64_t reserve;
};
__m128i xmm[XMM_NUM_IN_IPV6_5TUPLE];
};
struct ipv4_l3fwd_route {
struct ipv4_5tuple key;
uint8_t if_out;
};
struct ipv6_l3fwd_route {
struct ipv6_5tuple key;
uint8_t if_out;
};
static struct ipv4_l3fwd_route ipv4_l3fwd_route_array[] = {
{{
IPv4(101, 0, 0, 0),
IPv4(100, 10, 0, 1), 101, 11, IPPROTO_TCP}, 0},
{{
IPv4(201, 0, 0, 0),
IPv4(200, 20, 0, 1), 102, 12, IPPROTO_TCP}, 1},
{{
IPv4(111, 0, 0, 0),
IPv4(100, 30, 0, 1), 101, 11, IPPROTO_TCP}, 2},
{{
IPv4(211, 0, 0, 0),
IPv4(200, 40, 0, 1), 102, 12, IPPROTO_TCP}, 3},
};
static struct ipv6_l3fwd_route ipv6_l3fwd_route_array[] = {
{{
{0xfe, 0x80, 0, 0, 0, 0, 0, 0, 0x02, 0x1e, 0x67, 0xff, 0xfe, 0, 0, 0},
{0xfe, 0x80, 0, 0, 0, 0, 0, 0, 0x02, 0x1b, 0x21, 0xff, 0xfe, 0x91, 0x38,
0x05},
101, 11, IPPROTO_TCP}, 0},
{{
{0xfe, 0x90, 0, 0, 0, 0, 0, 0, 0x02, 0x1e, 0x67, 0xff, 0xfe, 0, 0, 0},
{0xfe, 0x90, 0, 0, 0, 0, 0, 0, 0x02, 0x1b, 0x21, 0xff, 0xfe, 0x91, 0x38,
0x05},
102, 12, IPPROTO_TCP}, 1},
{{
{0xfe, 0xa0, 0, 0, 0, 0, 0, 0, 0x02, 0x1e, 0x67, 0xff, 0xfe, 0, 0, 0},
{0xfe, 0xa0, 0, 0, 0, 0, 0, 0, 0x02, 0x1b, 0x21, 0xff, 0xfe, 0x91, 0x38,
0x05},
101, 11, IPPROTO_TCP}, 2},
{{
{0xfe, 0xb0, 0, 0, 0, 0, 0, 0, 0x02, 0x1e, 0x67, 0xff, 0xfe, 0, 0, 0},
{0xfe, 0xb0, 0, 0, 0, 0, 0, 0, 0x02, 0x1b, 0x21, 0xff, 0xfe, 0x91, 0x38,
0x05},
102, 12, IPPROTO_TCP}, 3},
};
typedef struct rte_hash lookup_struct_t;
static lookup_struct_t *ipv4_l3fwd_lookup_struct[NB_SOCKETS];
static lookup_struct_t *ipv6_l3fwd_lookup_struct[NB_SOCKETS];
#ifdef RTE_ARCH_X86_64
#define L3FWD_HASH_ENTRIES (1024*1024*4)
#else
#define L3FWD_HASH_ENTRIES (1024*1024*1)
#endif
#define HASH_ENTRY_NUMBER_DEFAULT 4
static uint32_t hash_entry_number = HASH_ENTRY_NUMBER_DEFAULT;
static inline uint32_t
ipv4_hash_crc(
const void *data,
__rte_unused uint32_t data_len,
uint32_t init_val)
{
const union ipv4_5tuple_host *k;
uint32_t t;
const uint32_t *p;
k = data;
t = k->proto;
p = (const uint32_t *)&k->port_src;
#ifdef RTE_MACHINE_CPUFLAG_SSE4_2
#else
#endif
return init_val;
}
static inline uint32_t
ipv6_hash_crc(
const void *data,
__rte_unused uint32_t data_len,
uint32_t init_val)
{
const union ipv6_5tuple_host *k;
uint32_t t;
const uint32_t *p;
#ifdef RTE_MACHINE_CPUFLAG_SSE4_2
const uint32_t *ip_src0, *ip_src1, *ip_src2, *ip_src3;
const uint32_t *ip_dst0, *ip_dst1, *ip_dst2, *ip_dst3;
#endif
k = data;
t = k->proto;
p = (const uint32_t *)&k->port_src;
#ifdef RTE_MACHINE_CPUFLAG_SSE4_2
ip_src0 = (const uint32_t *) k->ip_src;
ip_src1 = (const uint32_t *)(k->ip_src + 4);
ip_src2 = (const uint32_t *)(k->ip_src + 8);
ip_src3 = (const uint32_t *)(k->ip_src + 12);
ip_dst0 = (const uint32_t *) k->ip_dst;
ip_dst1 = (const uint32_t *)(k->ip_dst + 4);
ip_dst2 = (const uint32_t *)(k->ip_dst + 8);
ip_dst3 = (const uint32_t *)(k->ip_dst + 12);
#else
init_val =
rte_jhash(k->ip_src,
sizeof(uint8_t) * IPV6_ADDR_LEN, init_val);
init_val =
rte_jhash(k->ip_dst,
sizeof(uint8_t) * IPV6_ADDR_LEN, init_val);
#endif
return init_val;
}
#define IPV4_L3FWD_NUM_ROUTES RTE_DIM(ipv4_l3fwd_route_array)
#define IPV6_L3FWD_NUM_ROUTES RTE_DIM(ipv6_l3fwd_route_array)
#endif
#if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
struct ipv4_l3fwd_route {
uint32_t ip;
uint8_t depth;
uint8_t if_out;
};
struct ipv6_l3fwd_route {
uint8_t ip[16];
uint8_t depth;
uint8_t if_out;
};
static struct ipv4_l3fwd_route ipv4_l3fwd_route_array[] = {
{
IPv4(1, 1, 1, 0), 24, 0},
{
IPv4(2, 1, 1, 0), 24, 1},
{
IPv4(3, 1, 1, 0), 24, 2},
{
IPv4(4, 1, 1, 0), 24, 3},
{
IPv4(5, 1, 1, 0), 24, 4},
{
IPv4(6, 1, 1, 0), 24, 5},
{
IPv4(7, 1, 1, 0), 24, 6},
{
IPv4(8, 1, 1, 0), 24, 7},
};
static struct ipv6_l3fwd_route ipv6_l3fwd_route_array[] = {
{{1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}, 48, 0},
{{2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}, 48, 1},
{{3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}, 48, 2},
{{4, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}, 48, 3},
{{5, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}, 48, 4},
{{6, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}, 48, 5},
{{7, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}, 48, 6},
{{8, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}, 48, 7},
};
#define IPV4_L3FWD_NUM_ROUTES RTE_DIM(ipv4_l3fwd_route_array)
#define IPV6_L3FWD_NUM_ROUTES RTE_DIM(ipv6_l3fwd_route_array)
#define IPV4_L3FWD_LPM_MAX_RULES 1024
#define IPV6_L3FWD_LPM_MAX_RULES 1024
#define IPV6_L3FWD_LPM_NUMBER_TBL8S (1 << 16)
typedef struct rte_lpm lookup_struct_t;
typedef struct rte_lpm6 lookup6_struct_t;
static lookup_struct_t *ipv4_l3fwd_lookup_struct[NB_SOCKETS];
static lookup6_struct_t *ipv6_l3fwd_lookup_struct[NB_SOCKETS];
#endif
struct lcore_conf {
lookup_struct_t *ipv4_lookup_struct;
#if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
lookup6_struct_t *ipv6_lookup_struct;
#else
lookup_struct_t *ipv6_lookup_struct;
#endif
void *data;
static struct lcore_conf lcore_conf[RTE_MAX_LCORE];
#define MAX_RX_QUEUE_PER_THREAD 16
#define MAX_TX_PORT_PER_THREAD RTE_MAX_ETHPORTS
#define MAX_TX_QUEUE_PER_PORT RTE_MAX_ETHPORTS
#define MAX_RX_QUEUE_PER_PORT 128
#define MAX_RX_THREAD 1024
#define MAX_TX_THREAD 1024
#define MAX_THREAD (MAX_RX_THREAD + MAX_TX_THREAD)
static int lthreads_on = 1;
struct thread_conf {
uint16_t lcore_id;
uint16_t cpu_id;
uint16_t thread_id;
#if (APP_CPU_LOAD > 0)
int busy[MAX_CPU_COUNTER];
#endif
};
struct thread_rx_conf {
struct thread_conf conf;
uint16_t n_rx_queue;
struct lcore_rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
uint16_t n_ring;
struct lthread_cond *ready[RTE_MAX_LCORE];
#if (APP_CPU_LOAD > 0)
int busy[MAX_CPU_COUNTER];
#endif
uint16_t n_rx_thread;
struct thread_rx_conf rx_thread[MAX_RX_THREAD];
struct thread_tx_conf {
struct thread_conf conf;
uint16_t tx_queue_id[RTE_MAX_LCORE];
struct mbuf_table tx_mbufs[RTE_MAX_LCORE];
struct lthread_cond **ready;
uint16_t n_tx_thread;
struct thread_tx_conf tx_thread[MAX_TX_THREAD];
static inline int
send_burst(
struct thread_tx_conf *qconf, uint16_t n, uint8_t
port)
{
int ret;
uint16_t queueid;
queueid = qconf->tx_queue_id[
port];
m_table = (
struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
do {
} while (++ret < n);
}
return 0;
}
static inline int
send_single_packet(
struct rte_mbuf *m, uint8_t port)
{
uint16_t len;
struct thread_tx_conf *qconf;
if (lthreads_on)
qconf = (struct thread_tx_conf *)lthread_get_data();
else
qconf = (
struct thread_tx_conf *)
RTE_PER_LCORE(lcore_conf)->data;
len = qconf->tx_mbufs[port].len;
qconf->tx_mbufs[port].m_table[len] = m;
len++;
send_burst(qconf, MAX_PKT_BURST, port);
len = 0;
}
qconf->tx_mbufs[port].len = len;
return 0;
}
#if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
static inline __attribute__((always_inline)) void
send_packetsx4(uint8_t port,
{
uint32_t len, j, n;
struct thread_tx_conf *qconf;
if (lthreads_on)
qconf = (struct thread_tx_conf *)lthread_get_data();
else
qconf = (
struct thread_tx_conf *)
RTE_PER_LCORE(lcore_conf)->data;
len = qconf->tx_mbufs[port].len;
if (num >= MAX_TX_BURST && len == 0) {
do {
} while (++n < num);
}
return;
}
n = len + num;
n = (n > MAX_PKT_BURST) ? MAX_PKT_BURST - len : num;
j = 0;
switch (n % FWDSTEP) {
while (j < n) {
case 0:
qconf->tx_mbufs[port].m_table[len + j] = m[j];
j++;
case 3:
qconf->tx_mbufs[port].m_table[len + j] = m[j];
j++;
case 2:
qconf->tx_mbufs[port].m_table[len + j] = m[j];
j++;
case 1:
qconf->tx_mbufs[port].m_table[len + j] = m[j];
j++;
}
}
len += n;
send_burst(qconf, MAX_PKT_BURST, port);
len = num - n;
j = 0;
switch (len % FWDSTEP) {
while (j < len) {
case 0:
qconf->tx_mbufs[port].m_table[j] = m[n + j];
j++;
case 3:
qconf->tx_mbufs[port].m_table[j] = m[n + j];
j++;
case 2:
qconf->tx_mbufs[port].m_table[j] = m[n + j];
j++;
case 1:
qconf->tx_mbufs[port].m_table[j] = m[n + j];
j++;
}
}
}
qconf->tx_mbufs[port].len = len;
}
#endif
#ifdef DO_RFC_1812_CHECKS
static inline int
is_valid_ipv4_pkt(
struct ipv4_hdr *pkt, uint32_t link_len)
{
return -1;
return -3;
return -4;
return -5;
return 0;
}
#endif
#if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
static __m128i mask0;
static __m128i mask1;
static __m128i mask2;
static inline uint8_t
get_ipv4_dst_port(
void *
ipv4_hdr, uint8_t portid,
lookup_struct_t *ipv4_l3fwd_lookup_struct)
{
int ret = 0;
union ipv4_5tuple_host key;
ipv4_hdr = (uint8_t *)ipv4_hdr +
offsetof(
struct ipv4_hdr, time_to_live);
__m128i data = _mm_loadu_si128((__m128i *)(ipv4_hdr));
key.xmm = _mm_and_si128(data, mask0);
return (uint8_t)((ret < 0) ? portid : ipv4_l3fwd_out_if[ret]);
}
static inline uint8_t
get_ipv6_dst_port(
void *
ipv6_hdr, uint8_t portid,
lookup_struct_t *ipv6_l3fwd_lookup_struct)
{
int ret = 0;
union ipv6_5tuple_host key;
ipv6_hdr = (uint8_t *)ipv6_hdr +
offsetof(
struct ipv6_hdr, payload_len);
__m128i data0 = _mm_loadu_si128((__m128i *)(ipv6_hdr));
__m128i data1 = _mm_loadu_si128((__m128i *)(((uint8_t *)ipv6_hdr) +
sizeof(__m128i)));
__m128i data2 = _mm_loadu_si128((__m128i *)(((uint8_t *)ipv6_hdr) +
sizeof(__m128i) + sizeof(__m128i)));
key.xmm[0] = _mm_and_si128(data0, mask1);
key.xmm[1] = data1;
key.xmm[2] = _mm_and_si128(data2, mask2);
return (uint8_t)((ret < 0) ? portid : ipv6_l3fwd_out_if[ret]);
}
#endif
#if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
static inline uint8_t
get_ipv4_dst_port(void *ipv4_hdr, uint8_t portid,
lookup_struct_t *ipv4_l3fwd_lookup_struct)
{
uint8_t next_hop;
&next_hop) == 0) ? next_hop : portid);
}
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;
((struct ipv6_hdr *)ipv6_hdr)->dst_addr, &next_hop) == 0) ?
next_hop : portid);
}
#endif
static inline void l3fwd_simple_forward(
struct rte_mbuf *m, uint8_t portid)
__attribute__((unused));
#if ((APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH) && \
(ENABLE_MULTI_BUFFER_OPTIMIZE == 1))
#define MASK_ALL_PKTS 0xff
#define EXCLUDE_1ST_PKT 0xfe
#define EXCLUDE_2ND_PKT 0xfd
#define EXCLUDE_3RD_PKT 0xfb
#define EXCLUDE_4TH_PKT 0xf7
#define EXCLUDE_5TH_PKT 0xef
#define EXCLUDE_6TH_PKT 0xdf
#define EXCLUDE_7TH_PKT 0xbf
#define EXCLUDE_8TH_PKT 0x7f
static inline void
simple_ipv4_fwd_8pkts(
struct rte_mbuf *m[8], uint8_t portid)
{
struct ipv4_hdr *ipv4_hdr[8];
uint8_t dst_port[8];
int32_t ret[8];
union ipv4_5tuple_host key[8];
__m128i data[8];
#ifdef DO_RFC_1812_CHECKS
uint8_t valid_mask = MASK_ALL_PKTS;
if (is_valid_ipv4_pkt(ipv4_hdr[0], m[0]->pkt_len) < 0) {
valid_mask &= EXCLUDE_1ST_PKT;
}
if (is_valid_ipv4_pkt(ipv4_hdr[1], m[1]->pkt_len) < 0) {
valid_mask &= EXCLUDE_2ND_PKT;
}
if (is_valid_ipv4_pkt(ipv4_hdr[2], m[2]->pkt_len) < 0) {
valid_mask &= EXCLUDE_3RD_PKT;
}
if (is_valid_ipv4_pkt(ipv4_hdr[3], m[3]->pkt_len) < 0) {
valid_mask &= EXCLUDE_4TH_PKT;
}
if (is_valid_ipv4_pkt(ipv4_hdr[4], m[4]->pkt_len) < 0) {
valid_mask &= EXCLUDE_5TH_PKT;
}
if (is_valid_ipv4_pkt(ipv4_hdr[5], m[5]->pkt_len) < 0) {
valid_mask &= EXCLUDE_6TH_PKT;
}
if (is_valid_ipv4_pkt(ipv4_hdr[6], m[6]->pkt_len) < 0) {
valid_mask &= EXCLUDE_7TH_PKT;
}
if (is_valid_ipv4_pkt(ipv4_hdr[7], m[7]->pkt_len) < 0) {
valid_mask &= EXCLUDE_8TH_PKT;
}
if (
unlikely(valid_mask != MASK_ALL_PKTS)) {
if (valid_mask == 0)
return;
uint8_t i = 0;
for (i = 0; i < 8; i++)
if ((0x1 << i) & valid_mask)
l3fwd_simple_forward(m[i], portid);
}
#endif
offsetof(
struct ipv4_hdr, time_to_live)));
offsetof(
struct ipv4_hdr, time_to_live)));
offsetof(
struct ipv4_hdr, time_to_live)));
offsetof(
struct ipv4_hdr, time_to_live)));
offsetof(
struct ipv4_hdr, time_to_live)));
offsetof(
struct ipv4_hdr, time_to_live)));
offsetof(
struct ipv4_hdr, time_to_live)));
offsetof(
struct ipv4_hdr, time_to_live)));
key[0].xmm = _mm_and_si128(data[0], mask0);
key[1].xmm = _mm_and_si128(data[1], mask0);
key[2].xmm = _mm_and_si128(data[2], mask0);
key[3].xmm = _mm_and_si128(data[3], mask0);
key[4].xmm = _mm_and_si128(data[4], mask0);
key[5].xmm = _mm_and_si128(data[5], mask0);
key[6].xmm = _mm_and_si128(data[6], mask0);
key[7].xmm = _mm_and_si128(data[7], mask0);
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,
&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]]);
dst_port[2] = (uint8_t) ((ret[2] < 0) ? portid : ipv4_l3fwd_out_if[ret[2]]);
dst_port[3] = (uint8_t) ((ret[3] < 0) ? portid : ipv4_l3fwd_out_if[ret[3]]);
dst_port[4] = (uint8_t) ((ret[4] < 0) ? portid : ipv4_l3fwd_out_if[ret[4]]);
dst_port[5] = (uint8_t) ((ret[5] < 0) ? portid : ipv4_l3fwd_out_if[ret[5]]);
dst_port[6] = (uint8_t) ((ret[6] < 0) ? portid : ipv4_l3fwd_out_if[ret[6]]);
dst_port[7] = (uint8_t) ((ret[7] < 0) ? portid : ipv4_l3fwd_out_if[ret[7]]);
if (dst_port[0] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[0]) == 0)
dst_port[0] = portid;
if (dst_port[1] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[1]) == 0)
dst_port[1] = portid;
if (dst_port[2] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[2]) == 0)
dst_port[2] = portid;
if (dst_port[3] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[3]) == 0)
dst_port[3] = portid;
if (dst_port[4] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[4]) == 0)
dst_port[4] = portid;
if (dst_port[5] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[5]) == 0)
dst_port[5] = portid;
if (dst_port[6] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[6]) == 0)
dst_port[6] = portid;
if (dst_port[7] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[7]) == 0)
dst_port[7] = portid;
#ifdef DO_RFC_1812_CHECKS
#endif
*(uint64_t *)ð_hdr[0]->d_addr = dest_eth_addr[dst_port[0]];
*(uint64_t *)ð_hdr[1]->d_addr = dest_eth_addr[dst_port[1]];
*(uint64_t *)ð_hdr[2]->d_addr = dest_eth_addr[dst_port[2]];
*(uint64_t *)ð_hdr[3]->d_addr = dest_eth_addr[dst_port[3]];
*(uint64_t *)ð_hdr[4]->d_addr = dest_eth_addr[dst_port[4]];
*(uint64_t *)ð_hdr[5]->d_addr = dest_eth_addr[dst_port[5]];
*(uint64_t *)ð_hdr[6]->d_addr = dest_eth_addr[dst_port[6]];
*(uint64_t *)ð_hdr[7]->d_addr = dest_eth_addr[dst_port[7]];
send_single_packet(m[0], (uint8_t)dst_port[0]);
send_single_packet(m[1], (uint8_t)dst_port[1]);
send_single_packet(m[2], (uint8_t)dst_port[2]);
send_single_packet(m[3], (uint8_t)dst_port[3]);
send_single_packet(m[4], (uint8_t)dst_port[4]);
send_single_packet(m[5], (uint8_t)dst_port[5]);
send_single_packet(m[6], (uint8_t)dst_port[6]);
send_single_packet(m[7], (uint8_t)dst_port[7]);
}
static inline void get_ipv6_5tuple(
struct rte_mbuf *m0, __m128i mask0,
__m128i mask1, union ipv6_5tuple_host *key)
{
offsetof(
struct ipv6_hdr, payload_len)));
offsetof(
struct ipv6_hdr, payload_len) +
sizeof(__m128i)));
offsetof(
struct ipv6_hdr, payload_len) +
sizeof(__m128i) +
sizeof(__m128i)));
key->xmm[0] = _mm_and_si128(tmpdata0, mask0);
key->xmm[1] = tmpdata1;
key->xmm[2] = _mm_and_si128(tmpdata2, mask1);
}
static inline void
simple_ipv6_fwd_8pkts(
struct rte_mbuf *m[8], uint8_t portid)
{
int32_t ret[8];
uint8_t dst_port[8];
union ipv6_5tuple_host key[8];
__attribute__((unused)) struct ipv6_hdr *ipv6_hdr[8];
sizeof(struct ether_hdr));
sizeof(struct ether_hdr));
sizeof(struct ether_hdr));
sizeof(struct ether_hdr));
sizeof(struct ether_hdr));
sizeof(struct ether_hdr));
sizeof(struct ether_hdr));
sizeof(struct ether_hdr));
get_ipv6_5tuple(m[0], mask1, mask2, &key[0]);
get_ipv6_5tuple(m[1], mask1, mask2, &key[1]);
get_ipv6_5tuple(m[2], mask1, mask2, &key[2]);
get_ipv6_5tuple(m[3], mask1, mask2, &key[3]);
get_ipv6_5tuple(m[4], mask1, mask2, &key[4]);
get_ipv6_5tuple(m[5], mask1, mask2, &key[5]);
get_ipv6_5tuple(m[6], mask1, mask2, &key[6]);
get_ipv6_5tuple(m[7], mask1, mask2, &key[7]);
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,
&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]]);
dst_port[2] = (uint8_t) ((ret[2] < 0) ? portid : ipv6_l3fwd_out_if[ret[2]]);
dst_port[3] = (uint8_t) ((ret[3] < 0) ? portid : ipv6_l3fwd_out_if[ret[3]]);
dst_port[4] = (uint8_t) ((ret[4] < 0) ? portid : ipv6_l3fwd_out_if[ret[4]]);
dst_port[5] = (uint8_t) ((ret[5] < 0) ? portid : ipv6_l3fwd_out_if[ret[5]]);
dst_port[6] = (uint8_t) ((ret[6] < 0) ? portid : ipv6_l3fwd_out_if[ret[6]]);
dst_port[7] = (uint8_t) ((ret[7] < 0) ? portid : ipv6_l3fwd_out_if[ret[7]]);
if (dst_port[0] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[0]) == 0)
dst_port[0] = portid;
if (dst_port[1] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[1]) == 0)
dst_port[1] = portid;
if (dst_port[2] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[2]) == 0)
dst_port[2] = portid;
if (dst_port[3] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[3]) == 0)
dst_port[3] = portid;
if (dst_port[4] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[4]) == 0)
dst_port[4] = portid;
if (dst_port[5] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[5]) == 0)
dst_port[5] = portid;
if (dst_port[6] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[6]) == 0)
dst_port[6] = portid;
if (dst_port[7] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[7]) == 0)
dst_port[7] = portid;
*(uint64_t *)ð_hdr[0]->d_addr = dest_eth_addr[dst_port[0]];
*(uint64_t *)ð_hdr[1]->d_addr = dest_eth_addr[dst_port[1]];
*(uint64_t *)ð_hdr[2]->d_addr = dest_eth_addr[dst_port[2]];
*(uint64_t *)ð_hdr[3]->d_addr = dest_eth_addr[dst_port[3]];
*(uint64_t *)ð_hdr[4]->d_addr = dest_eth_addr[dst_port[4]];
*(uint64_t *)ð_hdr[5]->d_addr = dest_eth_addr[dst_port[5]];
*(uint64_t *)ð_hdr[6]->d_addr = dest_eth_addr[dst_port[6]];
*(uint64_t *)ð_hdr[7]->d_addr = dest_eth_addr[dst_port[7]];
send_single_packet(m[0], (uint8_t)dst_port[0]);
send_single_packet(m[1], (uint8_t)dst_port[1]);
send_single_packet(m[2], (uint8_t)dst_port[2]);
send_single_packet(m[3], (uint8_t)dst_port[3]);
send_single_packet(m[4], (uint8_t)dst_port[4]);
send_single_packet(m[5], (uint8_t)dst_port[5]);
send_single_packet(m[6], (uint8_t)dst_port[6]);
send_single_packet(m[7], (uint8_t)dst_port[7]);
}
#endif
static inline __attribute__((always_inline)) void
l3fwd_simple_forward(struct
rte_mbuf *m, uint8_t portid)
{
struct ipv4_hdr *ipv4_hdr;
uint8_t dst_port;
#ifdef DO_RFC_1812_CHECKS
if (is_valid_ipv4_pkt(ipv4_hdr, m->pkt_len) < 0) {
return;
}
#endif
dst_port = get_ipv4_dst_port(ipv4_hdr, portid,
if (dst_port >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port) == 0)
dst_port = portid;
#ifdef DO_RFC_1812_CHECKS
#endif
*(uint64_t *)ð_hdr->
d_addr = dest_eth_addr[dst_port];
send_single_packet(m, dst_port);
struct ipv6_hdr *ipv6_hdr;
dst_port = get_ipv6_dst_port(ipv6_hdr, portid,
if (dst_port >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port) == 0)
dst_port = portid;
*(uint64_t *)ð_hdr->
d_addr = dest_eth_addr[dst_port];
send_single_packet(m, dst_port);
} else
}
#if ((APP_LOOKUP_METHOD == APP_LOOKUP_LPM) && \
(ENABLE_MULTI_BUFFER_OPTIMIZE == 1))
#ifdef DO_RFC_1812_CHECKS
#define IPV4_MIN_VER_IHL 0x45
#define IPV4_MAX_VER_IHL 0x4f
#define IPV4_MAX_VER_IHL_DIFF (IPV4_MAX_VER_IHL - IPV4_MIN_VER_IHL)
#define IPV4_MIN_LEN_BE (sizeof(struct ipv4_hdr) << 8)
static inline __attribute__((always_inline)) void
rfc1812_process(struct ipv4_hdr *ipv4_hdr, uint16_t *dp, uint32_t ptype)
{
uint8_t ihl;
ihl = ipv4_hdr->version_ihl - IPV4_MIN_VER_IHL;
ipv4_hdr->time_to_live--;
ipv4_hdr->hdr_checksum++;
if (ihl > IPV4_MAX_VER_IHL_DIFF ||
((uint8_t)ipv4_hdr->total_length == 0 &&
ipv4_hdr->total_length < IPV4_MIN_LEN_BE)) {
dp[0] = BAD_PORT;
}
}
}
#else
#define rfc1812_process(mb, dp) do { } while (0)
#endif
#endif
#if ((APP_LOOKUP_METHOD == APP_LOOKUP_LPM) && \
(ENABLE_MULTI_BUFFER_OPTIMIZE == 1))
static inline __attribute__((always_inline)) uint16_t
get_dst_port(struct
rte_mbuf *pkt, uint32_t dst_ipv4, uint8_t portid)
{
uint8_t next_hop;
struct ipv6_hdr *ipv6_hdr;
dst_ipv4, &next_hop) != 0)
next_hop = portid;
ipv6_hdr = (struct ipv6_hdr *)(eth_hdr + 1);
next_hop = portid;
} else {
next_hop = portid;
}
return next_hop;
}
static inline void
process_packet(
struct rte_mbuf *pkt, uint16_t *dst_port, uint8_t portid)
{
struct ipv4_hdr *ipv4_hdr;
uint32_t dst_ipv4;
uint16_t dp;
__m128i te, ve;
ipv4_hdr = (struct ipv4_hdr *)(eth_hdr + 1);
dp = get_dst_port(pkt, dst_ipv4, portid);
te = _mm_load_si128((__m128i *)eth_hdr);
ve = val_eth[dp];
dst_port[0] = dp;
te = _mm_blend_epi16(te, ve, MASK_ETH);
_mm_store_si128((__m128i *)eth_hdr, te);
}
static inline void
processx4_step1(
struct rte_mbuf *pkt[FWDSTEP],
__m128i *dip,
uint32_t *ipv4_flag)
{
struct ipv4_hdr *ipv4_hdr;
uint32_t x0, x1, x2, x3;
ipv4_hdr = (struct ipv4_hdr *)(eth_hdr + 1);
ipv4_hdr = (struct ipv4_hdr *)(eth_hdr + 1);
ipv4_hdr = (struct ipv4_hdr *)(eth_hdr + 1);
ipv4_hdr = (struct ipv4_hdr *)(eth_hdr + 1);
dip[0] = _mm_set_epi32(x3, x2, x1, x0);
}
static inline void
processx4_step2(__m128i dip,
uint32_t ipv4_flag,
uint8_t portid,
uint16_t dprt[FWDSTEP])
{
rte_xmm_t dst;
const __m128i bswap_mask = _mm_set_epi8(12, 13, 14, 15, 8, 9, 10, 11,
4, 5, 6, 7, 0, 1, 2, 3);
dip = _mm_shuffle_epi8(dip, bswap_mask);
dprt, portid);
} else {
dst.x = dip;
dprt[0] = get_dst_port(pkt[0], dst.u32[0], portid);
dprt[1] = get_dst_port(pkt[1], dst.u32[1], portid);
dprt[2] = get_dst_port(pkt[2], dst.u32[2], portid);
dprt[3] = get_dst_port(pkt[3], dst.u32[3], portid);
}
}
static inline void
processx4_step3(
struct rte_mbuf *pkt[FWDSTEP], uint16_t dst_port[FWDSTEP])
{
__m128i te[FWDSTEP];
__m128i ve[FWDSTEP];
__m128i *p[FWDSTEP];
ve[0] = val_eth[dst_port[0]];
te[0] = _mm_load_si128(p[0]);
ve[1] = val_eth[dst_port[1]];
te[1] = _mm_load_si128(p[1]);
ve[2] = val_eth[dst_port[2]];
te[2] = _mm_load_si128(p[2]);
ve[3] = val_eth[dst_port[3]];
te[3] = _mm_load_si128(p[3]);
te[0] = _mm_blend_epi16(te[0], ve[0], MASK_ETH);
te[1] = _mm_blend_epi16(te[1], ve[1], MASK_ETH);
te[2] = _mm_blend_epi16(te[2], ve[2], MASK_ETH);
te[3] = _mm_blend_epi16(te[3], ve[3], MASK_ETH);
_mm_store_si128(p[0], te[0]);
_mm_store_si128(p[1], te[1]);
_mm_store_si128(p[2], te[2]);
_mm_store_si128(p[3], te[3]);
rfc1812_process((
struct ipv4_hdr *)((
struct ether_hdr *)p[0] + 1),
&dst_port[0], pkt[0]->packet_type);
rfc1812_process((
struct ipv4_hdr *)((
struct ether_hdr *)p[1] + 1),
&dst_port[1], pkt[1]->packet_type);
rfc1812_process((
struct ipv4_hdr *)((
struct ether_hdr *)p[2] + 1),
&dst_port[2], pkt[2]->packet_type);
rfc1812_process((
struct ipv4_hdr *)((
struct ether_hdr *)p[3] + 1),
&dst_port[3], pkt[3]->packet_type);
}
#define GRPSZ (1 << FWDSTEP)
#define GRPMSK (GRPSZ - 1)
#define GROUP_PORT_STEP(dlp, dcp, lp, pn, idx) do { \
if (likely((dlp) == (dcp)[(idx)])) { \
(lp)[0]++; \
} else { \
(dlp) = (dcp)[idx]; \
(lp) = (pn) + (idx); \
(lp)[0] = 1; \
} \
} while (0)
static inline uint16_t *
port_groupx4(uint16_t pn[FWDSTEP + 1], uint16_t *lp, __m128i dp1, __m128i dp2)
{
static const struct {
uint64_t pnum;
int32_t idx;
uint16_t lpv;
} gptbl[GRPSZ] = {
{
.pnum = UINT64_C(0x0001000100010001),
.idx = 4,
.lpv = 0,
},
{
.pnum = UINT64_C(0x0001000100010002),
.idx = 4,
.lpv = 1,
},
{
.pnum = UINT64_C(0x0001000100020001),
.idx = 4,
.lpv = 0,
},
{
.pnum = UINT64_C(0x0001000100020003),
.idx = 4,
.lpv = 2,
},
{
.pnum = UINT64_C(0x0001000200010001),
.idx = 4,
.lpv = 0,
},
{
.pnum = UINT64_C(0x0001000200010002),
.idx = 4,
.lpv = 1,
},
{
.pnum = UINT64_C(0x0001000200030001),
.idx = 4,
.lpv = 0,
},
{
.pnum = UINT64_C(0x0001000200030004),
.idx = 4,
.lpv = 3,
},
{
.pnum = UINT64_C(0x0002000100010001),
.idx = 3,
.lpv = 0,
},
{
.pnum = UINT64_C(0x0002000100010002),
.idx = 3,
.lpv = 1,
},
{
.pnum = UINT64_C(0x0002000100020001),
.idx = 3,
.lpv = 0,
},
{
.pnum = UINT64_C(0x0002000100020003),
.idx = 3,
.lpv = 2,
},
{
.pnum = UINT64_C(0x0002000300010001),
.idx = 2,
.lpv = 0,
},
{
.pnum = UINT64_C(0x0002000300010002),
.idx = 2,
.lpv = 1,
},
{
.pnum = UINT64_C(0x0002000300040001),
.idx = 1,
.lpv = 0,
},
{
.pnum = UINT64_C(0x0002000300040005),
.idx = 0,
.lpv = 4,
},
};
union {
uint16_t u16[FWDSTEP + 1];
uint64_t u64;
} *pnum = (void *)pn;
int32_t v;
dp1 = _mm_cmpeq_epi16(dp1, dp2);
dp1 = _mm_unpacklo_epi16(dp1, dp1);
v = _mm_movemask_ps((__m128)dp1);
lp[0] += gptbl[v].lpv;
if (v != GRPMSK) {
lp = pnum->u16 + gptbl[v].idx;
lp[0] = 1;
pnum->u64 = gptbl[v].pnum;
}
return lp;
}
#endif
static void
process_burst(
struct rte_mbuf *pkts_burst[MAX_PKT_BURST],
int nb_rx,
uint8_t portid) {
int j;
#if ((APP_LOOKUP_METHOD == APP_LOOKUP_LPM) && \
(ENABLE_MULTI_BUFFER_OPTIMIZE == 1))
int32_t k;
uint16_t dlp;
uint16_t *lp;
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];
#endif
#if (ENABLE_MULTI_BUFFER_OPTIMIZE == 1)
#if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
{
for (j = 0; j < n; j += 8) {
uint32_t pkt_type =
pkts_burst[j]->packet_type &
pkts_burst[j+1]->packet_type &
pkts_burst[j+2]->packet_type &
pkts_burst[j+3]->packet_type &
pkts_burst[j+4]->packet_type &
pkts_burst[j+5]->packet_type &
pkts_burst[j+6]->packet_type &
pkts_burst[j+7]->packet_type;
simple_ipv4_fwd_8pkts(&pkts_burst[j], portid);
} else if (pkt_type &
simple_ipv6_fwd_8pkts(&pkts_burst[j], portid);
} else {
l3fwd_simple_forward(pkts_burst[j], portid);
l3fwd_simple_forward(pkts_burst[j+1], portid);
l3fwd_simple_forward(pkts_burst[j+2], portid);
l3fwd_simple_forward(pkts_burst[j+3], portid);
l3fwd_simple_forward(pkts_burst[j+4], portid);
l3fwd_simple_forward(pkts_burst[j+5], portid);
l3fwd_simple_forward(pkts_burst[j+6], portid);
l3fwd_simple_forward(pkts_burst[j+7], portid);
}
}
for (; j < nb_rx ; j++)
l3fwd_simple_forward(pkts_burst[j], portid);
}
#elif (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
for (j = 0; j != k; j += FWDSTEP)
processx4_step1(&pkts_burst[j], &dip[j / FWDSTEP],
&ipv4_flag[j / FWDSTEP]);
for (j = 0; j != k; j += FWDSTEP)
processx4_step2(dip[j / FWDSTEP], ipv4_flag[j / FWDSTEP],
portid, &pkts_burst[j], &dst_port[j]);
if (k != 0) {
__m128i dp1, dp2;
lp = pnum;
lp[0] = 1;
processx4_step3(pkts_burst, dst_port);
dp1 = _mm_loadu_si128((__m128i *)dst_port);
for (j = FWDSTEP; j != k; j += FWDSTEP) {
processx4_step3(&pkts_burst[j], &dst_port[j]);
dp2 = _mm_loadu_si128(
(__m128i *)&dst_port[j - FWDSTEP + 1]);
lp = port_groupx4(&pnum[j - FWDSTEP], lp, dp1, dp2);
dp1 = _mm_srli_si128(dp2, (FWDSTEP - 1) *
sizeof(dst_port[0]));
}
dp2 = _mm_shufflelo_epi16(dp1, 0xf9);
lp = port_groupx4(&pnum[j - FWDSTEP], lp, dp1, dp2);
lp[0]--;
dlp = dst_port[j - 1];
} else {
dlp = BAD_PORT - 1;
lp = pnum + MAX_PKT_BURST;
}
switch (nb_rx % FWDSTEP) {
case 3:
process_packet(pkts_burst[j], dst_port + j, portid);
GROUP_PORT_STEP(dlp, dst_port, lp, pnum, j);
j++;
case 2:
process_packet(pkts_burst[j], dst_port + j, portid);
GROUP_PORT_STEP(dlp, dst_port, lp, pnum, j);
j++;
case 1:
process_packet(pkts_burst[j], dst_port + j, portid);
GROUP_PORT_STEP(dlp, dst_port, lp, pnum, j);
j++;
}
for (j = 0; j < nb_rx; j += k) {
int32_t m;
uint16_t pn;
pn = dst_port[j];
k = pnum[j];
send_packetsx4(pn, pkts_burst + j, k);
else
for (m = j; m != j + k; m++)
}
#endif
#else
for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++)
for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
j + PREFETCH_OFFSET], void *));
l3fwd_simple_forward(pkts_burst[j], portid, qconf);
}
for (; j < nb_rx; j++)
l3fwd_simple_forward(pkts_burst[j], portid, qconf);
#endif
}
#if (APP_CPU_LOAD > 0)
static int
unsigned i, j, k;
uint64_t hits;
uint64_t prev_tsc, diff_tsc, cur_tsc;
uint64_t total[MAX_CPU] = { 0 };
unsigned min_cpu = MAX_CPU;
unsigned max_cpu = 0;
unsigned cpu_id;
int busy_total = 0;
int busy_flag = 0;
unsigned int n_thread_per_cpu[MAX_CPU] = { 0 };
struct thread_conf *thread_per_cpu[MAX_CPU][MAX_THREAD];
struct thread_conf *thread_conf;
US_PER_S * CPU_LOAD_TIMEOUT_US;
prev_tsc = 0;
printf("Waiting for %d rx threads and %d tx threads\n", n_rx_thread,
n_tx_thread);
rte_pause();
rte_pause();
for (i = 0; i < n_rx_thread; i++) {
thread_conf = &rx_thread[i].conf;
cpu_id = thread_conf->cpu_id;
thread_per_cpu[cpu_id][n_thread_per_cpu[cpu_id]++] = thread_conf;
if (cpu_id > max_cpu)
max_cpu = cpu_id;
if (cpu_id < min_cpu)
min_cpu = cpu_id;
}
for (i = 0; i < n_tx_thread; i++) {
thread_conf = &tx_thread[i].conf;
cpu_id = thread_conf->cpu_id;
thread_per_cpu[cpu_id][n_thread_per_cpu[cpu_id]++] = thread_conf;
if (thread_conf->cpu_id > max_cpu)
max_cpu = thread_conf->cpu_id;
if (thread_conf->cpu_id < min_cpu)
min_cpu = thread_conf->cpu_id;
}
while (1) {
cpu_load.counter++;
for (i = min_cpu; i <= max_cpu; i++) {
for (j = 0; j < MAX_CPU_COUNTER; j++) {
for (k = 0; k < n_thread_per_cpu[i]; k++)
if (thread_per_cpu[i][k]->busy[j]) {
busy_flag = 1;
break;
}
if (busy_flag) {
cpu_load.hits[j][i]++;
busy_total = 1;
busy_flag = 0;
}
}
if (busy_total) {
total[i]++;
busy_total = 0;
}
}
cur_tsc = rte_rdtsc();
diff_tsc = cur_tsc - prev_tsc;
if (
unlikely(diff_tsc > interval_tsc)) {
printf("\033c");
printf("Cpu usage for %d rx threads and %d tx threads:\n\n",
n_rx_thread, n_tx_thread);
printf("cpu# proc%% poll%% overhead%%\n\n");
for (i = min_cpu; i <= max_cpu; i++) {
hits = 0;
printf("CPU %d:", i);
for (j = 0; j < MAX_CPU_COUNTER; j++) {
printf("%7" PRIu64 "",
cpu_load.hits[j][i] * 100 / cpu_load.counter);
hits += cpu_load.hits[j][i];
cpu_load.hits[j][i] = 0;
}
printf("%7" PRIu64 "\n",
100 - total[i] * 100 / cpu_load.counter);
total[i] = 0;
}
cpu_load.counter = 0;
prev_tsc = cur_tsc;
}
}
}
#endif
static void
{
RTE_LOG(INFO, L3FWD,
"Starting scheduler on lcore %d.\n", lcore_id);
lthread_exit(NULL);
}
static void
lthread_tx_per_ring(void *dummy)
{
int nb_rx;
uint8_t portid;
struct thread_tx_conf *tx_conf;
struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
struct lthread_cond *ready;
tx_conf = (struct thread_tx_conf *)dummy;
ring = tx_conf->ring;
ready = *tx_conf->ready;
lthread_set_data((void *)tx_conf);
lthread_set_affinity(tx_conf->conf.lcore_id);
nb_rx = 0;
while (1) {
SET_CPU_BUSY(tx_conf, CPU_POLL);
MAX_PKT_BURST);
SET_CPU_IDLE(tx_conf, CPU_POLL);
if (nb_rx > 0) {
SET_CPU_BUSY(tx_conf, CPU_PROCESS);
portid = pkts_burst[0]->
port;
process_burst(pkts_burst, nb_rx, portid);
SET_CPU_IDLE(tx_conf, CPU_PROCESS);
lthread_yield();
} else
lthread_cond_wait(ready, 0);
}
}
static void
lthread_tx(void *args)
{
struct lthread *lt;
unsigned lcore_id;
uint8_t portid;
struct thread_tx_conf *tx_conf;
tx_conf = (struct thread_tx_conf *)args;
lthread_set_data((void *)tx_conf);
lthread_set_affinity(tx_conf->conf.lcore_id);
lthread_create(<, tx_conf->conf.lcore_id, lthread_tx_per_ring,
(void *)tx_conf);
RTE_LOG(INFO, L3FWD,
"Entering Tx main loop on lcore %u\n", lcore_id);
tx_conf->conf.cpu_id = sched_getcpu();
while (1) {
lthread_sleep(BURST_TX_DRAIN_US * 1000);
for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
if (tx_conf->tx_mbufs[portid].len == 0)
continue;
SET_CPU_BUSY(tx_conf, CPU_PROCESS);
send_burst(tx_conf, tx_conf->tx_mbufs[portid].len, portid);
SET_CPU_IDLE(tx_conf, CPU_PROCESS);
tx_conf->tx_mbufs[portid].len = 0;
}
}
}
static void
lthread_rx(void *dummy)
{
int ret;
uint16_t nb_rx;
int i;
uint8_t portid, queueid;
int worker_id;
int len[RTE_MAX_LCORE] = { 0 };
int old_len, new_len;
struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
struct thread_rx_conf *rx_conf;
rx_conf = (struct thread_rx_conf *)dummy;
lthread_set_data((void *)rx_conf);
lthread_set_affinity(rx_conf->conf.lcore_id);
if (rx_conf->n_rx_queue == 0) {
return;
}
for (i = 0; i < rx_conf->n_rx_queue; i++) {
portid = rx_conf->rx_queue_list[i].port_id;
queueid = rx_conf->rx_queue_list[i].queue_id;
RTE_LOG(INFO, L3FWD,
" -- lcoreid=%u portid=%hhu rxqueueid=%hhu\n",
}
for (i = 0; i < rx_conf->n_rx_queue; i++)
lthread_cond_init(NULL, &rx_conf->ready[i], NULL);
worker_id = 0;
rx_conf->conf.cpu_id = sched_getcpu();
while (1) {
for (i = 0; i < rx_conf->n_rx_queue; ++i) {
portid = rx_conf->rx_queue_list[i].port_id;
queueid = rx_conf->rx_queue_list[i].queue_id;
SET_CPU_BUSY(rx_conf, CPU_POLL);
MAX_PKT_BURST);
SET_CPU_IDLE(rx_conf, CPU_POLL);
if (nb_rx != 0) {
worker_id = (worker_id + 1) % rx_conf->n_ring;
old_len = len[worker_id];
SET_CPU_BUSY(rx_conf, CPU_PROCESS);
rx_conf->ring[worker_id],
(void **) pkts_burst,
nb_rx);
new_len = old_len + ret;
if (new_len >= BURST_SIZE) {
lthread_cond_signal(rx_conf->ready[worker_id]);
new_len = 0;
}
len[worker_id] = new_len;
uint32_t k;
for (k = ret; k < nb_rx; k++) {
}
}
SET_CPU_IDLE(rx_conf, CPU_PROCESS);
}
lthread_yield();
}
}
}
static void
struct lthread *lt[MAX_THREAD];
int i;
int n_thread = 0;
printf("Entering lthread_spawner\n");
for (i = 0; i < n_rx_thread; i++) {
rx_thread[i].conf.thread_id = i;
lthread_create(<[n_thread], -1, lthread_rx,
(void *)&rx_thread[i]);
n_thread++;
}
lthread_sleep(100000);
for (i = 0; i < n_tx_thread; i++) {
tx_thread[i].conf.thread_id = i;
lthread_create(<[n_thread], -1, lthread_tx,
(void *)&tx_thread[i]);
n_thread++;
}
for (i = 0; i < n_thread; i++)
lthread_join(lt[i], NULL);
}
static int
struct lthread *lt;
lthread_create(<, -1, lthread_spawner, NULL);
lthread_run();
return 0;
}
static int
struct lthread *lt;
#if (APP_CPU_LOAD)
if (lcore_id == cpu_load_lcore_id) {
cpu_load_collector(arg);
return 0;
}
#endif
lthread_create(<, -1, lthread_null, NULL);
lthread_run();
return 0;
}
static int
pthread_tx(void *dummy)
{
struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
uint64_t prev_tsc, diff_tsc, cur_tsc;
int nb_rx;
uint8_t portid;
struct thread_tx_conf *tx_conf;
US_PER_S * BURST_TX_DRAIN_US;
prev_tsc = 0;
tx_conf = (struct thread_tx_conf *)dummy;
tx_conf->conf.cpu_id = sched_getcpu();
while (1) {
cur_tsc = rte_rdtsc();
diff_tsc = cur_tsc - prev_tsc;
SET_CPU_BUSY(tx_conf, CPU_PROCESS);
for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
if (tx_conf->tx_mbufs[portid].len == 0)
continue;
send_burst(tx_conf, tx_conf->tx_mbufs[portid].len, portid);
tx_conf->tx_mbufs[portid].len = 0;
}
SET_CPU_IDLE(tx_conf, CPU_PROCESS);
prev_tsc = cur_tsc;
}
SET_CPU_BUSY(tx_conf, CPU_POLL);
(void **)pkts_burst, MAX_PKT_BURST);
SET_CPU_IDLE(tx_conf, CPU_POLL);
sched_yield();
continue;
}
SET_CPU_BUSY(tx_conf, CPU_PROCESS);
portid = pkts_burst[0]->
port;
process_burst(pkts_burst, nb_rx, portid);
SET_CPU_IDLE(tx_conf, CPU_PROCESS);
}
}
static int
pthread_rx(void *dummy)
{
int i;
int worker_id;
uint32_t n;
uint32_t nb_rx;
unsigned lcore_id;
uint8_t portid, queueid;
struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
struct thread_rx_conf *rx_conf;
rx_conf = (struct thread_rx_conf *)dummy;
if (rx_conf->n_rx_queue == 0) {
RTE_LOG(INFO, L3FWD,
"lcore %u has nothing to do\n", lcore_id);
return 0;
}
RTE_LOG(INFO, L3FWD,
"entering main rx loop on lcore %u\n", lcore_id);
for (i = 0; i < rx_conf->n_rx_queue; i++) {
portid = rx_conf->rx_queue_list[i].port_id;
queueid = rx_conf->rx_queue_list[i].queue_id;
RTE_LOG(INFO, L3FWD,
" -- lcoreid=%u portid=%hhu rxqueueid=%hhu\n",
lcore_id, portid, queueid);
}
worker_id = 0;
rx_conf->conf.cpu_id = sched_getcpu();
while (1) {
for (i = 0; i < rx_conf->n_rx_queue; ++i) {
portid = rx_conf->rx_queue_list[i].port_id;
queueid = rx_conf->rx_queue_list[i].queue_id;
SET_CPU_BUSY(rx_conf, CPU_POLL);
MAX_PKT_BURST);
SET_CPU_IDLE(rx_conf, CPU_POLL);
if (nb_rx == 0) {
sched_yield();
continue;
}
SET_CPU_BUSY(rx_conf, CPU_PROCESS);
worker_id = (worker_id + 1) % rx_conf->n_ring;
(void **)pkts_burst, nb_rx);
uint32_t k;
for (k = n; k < nb_rx; k++) {
}
}
SET_CPU_IDLE(rx_conf, CPU_PROCESS);
}
}
}
static int
int i;
for (i = 0; i < n_rx_thread; i++)
if (rx_thread[i].conf.lcore_id == lcore_id) {
printf("Start rx thread on %d...\n", lcore_id);
pthread_rx((void *)&rx_thread[i]);
return 0;
}
for (i = 0; i < n_tx_thread; i++)
if (tx_thread[i].conf.lcore_id == lcore_id) {
printf("Start tx thread on %d...\n", lcore_id);
pthread_tx((void *)&tx_thread[i]);
return 0;
}
#if (APP_CPU_LOAD)
if (lcore_id == cpu_load_lcore_id)
cpu_load_collector(arg);
#endif
return 0;
}
static int
check_lcore_params(void)
{
uint8_t queue, lcore;
uint16_t i;
int socketid;
for (i = 0; i < nb_rx_thread_params; ++i) {
queue = rx_thread_params[i].queue_id;
if (queue >= MAX_RX_QUEUE_PER_PORT) {
printf("invalid queue number: %hhu\n", queue);
return -1;
}
lcore = rx_thread_params[i].lcore_id;
printf("error: lcore %hhu is not enabled in lcore mask\n", lcore);
return -1;
}
if ((socketid != 0) && (numa_on == 0))
printf("warning: lcore %hhu is on socket %d with numa off\n",
lcore, socketid);
}
return 0;
}
static int
check_port_config(const unsigned nb_ports)
{
unsigned portid;
uint16_t i;
for (i = 0; i < nb_rx_thread_params; ++i) {
portid = rx_thread_params[i].port_id;
if ((enabled_port_mask & (1 << portid)) == 0) {
printf("port %u is not enabled in port mask\n", portid);
return -1;
}
if (portid >= nb_ports) {
printf("port %u is not present on the board\n", portid);
return -1;
}
}
return 0;
}
static uint8_t
get_port_n_rx_queues(const uint8_t port)
{
int queue = -1;
uint16_t i;
for (i = 0; i < nb_rx_thread_params; ++i)
if (rx_thread_params[i].port_id == port &&
rx_thread_params[i].queue_id > queue)
queue = rx_thread_params[i].queue_id;
return (uint8_t)(++queue);
}
static int
init_rx_rings(void)
{
unsigned socket_io;
struct thread_rx_conf *rx_conf;
struct thread_tx_conf *tx_conf;
unsigned rx_thread_id, tx_thread_id;
char name[256];
for (tx_thread_id = 0; tx_thread_id < n_tx_thread; tx_thread_id++) {
tx_conf = &tx_thread[tx_thread_id];
printf("Connecting tx-thread %d with rx-thread %d\n", tx_thread_id,
tx_conf->conf.thread_id);
rx_thread_id = tx_conf->conf.thread_id;
if (rx_thread_id > n_tx_thread) {
printf("connection from tx-thread %u to rx-thread %u fails "
"(rx-thread not defined)\n", tx_thread_id, rx_thread_id);
return -1;
}
rx_conf = &rx_thread[rx_thread_id];
snprintf(name, sizeof(name), "app_ring_s%u_rx%u_tx%u",
socket_io, rx_thread_id, tx_thread_id);
if (ring == NULL) {
rte_panic(
"Cannot create ring to connect rx-thread %u "
"with tx-thread %u\n", rx_thread_id, tx_thread_id);
}
rx_conf->ring[rx_conf->n_ring] = ring;
tx_conf->ring = ring;
tx_conf->ready = &rx_conf->ready[rx_conf->n_ring];
rx_conf->n_ring++;
}
return 0;
}
static int
init_rx_queues(void)
{
uint16_t i, nb_rx_queue;
uint8_t thread;
n_rx_thread = 0;
for (i = 0; i < nb_rx_thread_params; ++i) {
thread = rx_thread_params[i].thread_id;
nb_rx_queue = rx_thread[thread].n_rx_queue;
if (nb_rx_queue >= MAX_RX_QUEUE_PER_LCORE) {
printf("error: too many queues (%u) for thread: %u\n",
(unsigned)nb_rx_queue + 1, (unsigned)thread);
return -1;
}
rx_thread[thread].conf.thread_id = thread;
rx_thread[thread].conf.lcore_id = rx_thread_params[i].lcore_id;
rx_thread[thread].rx_queue_list[nb_rx_queue].port_id =
rx_thread_params[i].port_id;
rx_thread[thread].rx_queue_list[nb_rx_queue].queue_id =
rx_thread_params[i].queue_id;
rx_thread[thread].n_rx_queue++;
if (thread >= n_rx_thread)
n_rx_thread = thread + 1;
}
return 0;
}
static int
init_tx_threads(void)
{
int i;
n_tx_thread = 0;
for (i = 0; i < nb_tx_thread_params; ++i) {
tx_thread[n_tx_thread].conf.thread_id = tx_thread_params[i].thread_id;
tx_thread[n_tx_thread].conf.lcore_id = tx_thread_params[i].lcore_id;
n_tx_thread++;
}
return 0;
}
static void
print_usage(const char *prgname)
{
printf("%s [EAL options] -- -p PORTMASK -P"
" [--rx (port,queue,lcore,thread)[,(port,queue,lcore,thread]]"
" [--tx (lcore,thread)[,(lcore,thread]]"
" [--enable-jumbo [--max-pkt-len PKTLEN]]\n"
" -p PORTMASK: hexadecimal bitmask of ports to configure\n"
" -P : enable promiscuous mode\n"
" --rx (port,queue,lcore,thread): rx queues configuration\n"
" --tx (lcore,thread): tx threads configuration\n"
" --stat-lcore LCORE: use lcore for stat collector\n"
" --eth-dest=X,MM:MM:MM:MM:MM:MM: optional, ethernet destination for port X\n"
" --no-numa: optional, disable numa awareness\n"
" --ipv6: optional, specify it if running ipv6 packets\n"
" --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",
prgname);
}
static int parse_max_pkt_len(const char *pktlen)
{
char *end = NULL;
unsigned long len;
len = strtoul(pktlen, &end, 10);
if ((pktlen[0] == '\0') || (end == NULL) || (*end != '\0'))
return -1;
if (len == 0)
return -1;
return len;
}
static int
parse_portmask(const char *portmask)
{
char *end = NULL;
unsigned long pm;
pm = strtoul(portmask, &end, 16);
if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
return -1;
if (pm == 0)
return -1;
return pm;
}
#if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
static int
parse_hash_entry_number(const char *hash_entry_num)
{
char *end = NULL;
unsigned long hash_en;
hash_en = strtoul(hash_entry_num, &end, 16);
if ((hash_entry_num[0] == '\0') || (end == NULL) || (*end != '\0'))
return -1;
if (hash_en == 0)
return -1;
return hash_en;
}
#endif
static int
parse_rx_config(const char *q_arg)
{
char s[256];
const char *p, *p0 = q_arg;
char *end;
enum fieldnames {
FLD_PORT = 0,
FLD_QUEUE,
FLD_LCORE,
FLD_THREAD,
_NUM_FLD
};
unsigned long int_fld[_NUM_FLD];
char *str_fld[_NUM_FLD];
int i;
unsigned size;
nb_rx_thread_params = 0;
while ((p = strchr(p0, '(')) != NULL) {
++p;
p0 = strchr(p, ')');
if (p0 == NULL)
return -1;
size = p0 - p;
if (size >= sizeof(s))
return -1;
snprintf(s, sizeof(s), "%.*s", size, p);
if (
rte_strsplit(s,
sizeof(s), str_fld, _NUM_FLD,
',') != _NUM_FLD)
return -1;
for (i = 0; i < _NUM_FLD; i++) {
errno = 0;
int_fld[i] = strtoul(str_fld[i], &end, 0);
if (errno != 0 || end == str_fld[i] || int_fld[i] > 255)
return -1;
}
if (nb_rx_thread_params >= MAX_LCORE_PARAMS) {
printf("exceeded max number of rx params: %hu\n",
nb_rx_thread_params);
return -1;
}
rx_thread_params_array[nb_rx_thread_params].port_id =
(uint8_t)int_fld[FLD_PORT];
rx_thread_params_array[nb_rx_thread_params].queue_id =
(uint8_t)int_fld[FLD_QUEUE];
rx_thread_params_array[nb_rx_thread_params].lcore_id =
(uint8_t)int_fld[FLD_LCORE];
rx_thread_params_array[nb_rx_thread_params].thread_id =
(uint8_t)int_fld[FLD_THREAD];
++nb_rx_thread_params;
}
rx_thread_params = rx_thread_params_array;
return 0;
}
static int
parse_tx_config(const char *q_arg)
{
char s[256];
const char *p, *p0 = q_arg;
char *end;
enum fieldnames {
FLD_LCORE = 0,
FLD_THREAD,
_NUM_FLD
};
unsigned long int_fld[_NUM_FLD];
char *str_fld[_NUM_FLD];
int i;
unsigned size;
nb_tx_thread_params = 0;
while ((p = strchr(p0, '(')) != NULL) {
++p;
p0 = strchr(p, ')');
if (p0 == NULL)
return -1;
size = p0 - p;
if (size >= sizeof(s))
return -1;
snprintf(s, sizeof(s), "%.*s", size, p);
if (
rte_strsplit(s,
sizeof(s), str_fld, _NUM_FLD,
',') != _NUM_FLD)
return -1;
for (i = 0; i < _NUM_FLD; i++) {
errno = 0;
int_fld[i] = strtoul(str_fld[i], &end, 0);
if (errno != 0 || end == str_fld[i] || int_fld[i] > 255)
return -1;
}
if (nb_tx_thread_params >= MAX_LCORE_PARAMS) {
printf("exceeded max number of tx params: %hu\n",
nb_tx_thread_params);
return -1;
}
tx_thread_params_array[nb_tx_thread_params].lcore_id =
(uint8_t)int_fld[FLD_LCORE];
tx_thread_params_array[nb_tx_thread_params].thread_id =
(uint8_t)int_fld[FLD_THREAD];
++nb_tx_thread_params;
}
tx_thread_params = tx_thread_params_array;
return 0;
}
#if (APP_CPU_LOAD > 0)
static int
parse_stat_lcore(const char *stat_lcore)
{
char *end = NULL;
unsigned long lcore_id;
lcore_id = strtoul(stat_lcore, &end, 10);
if ((stat_lcore[0] == '\0') || (end == NULL) || (*end != '\0'))
return -1;
return lcore_id;
}
#endif
static void
parse_eth_dest(const char *optarg)
{
uint8_t portid;
char *port_end;
uint8_t c, *dest, peer_addr[6];
errno = 0;
portid = strtoul(optarg, &port_end, 10);
if (errno != 0 || port_end == optarg || *port_end++ != ',')
"Invalid eth-dest: %s", optarg);
if (portid >= RTE_MAX_ETHPORTS)
"eth-dest: port %d >= RTE_MAX_ETHPORTS(%d)\n",
portid, RTE_MAX_ETHPORTS);
if (cmdline_parse_etheraddr(NULL, port_end,
&peer_addr, sizeof(peer_addr)) < 0)
"Invalid ethernet address: %s\n",
port_end);
dest = (uint8_t *)&dest_eth_addr[portid];
for (c = 0; c < 6; c++)
dest[c] = peer_addr[c];
*(uint64_t *)(val_eth + portid) = dest_eth_addr[portid];
}
#define CMD_LINE_OPT_RX_CONFIG "rx"
#define CMD_LINE_OPT_TX_CONFIG "tx"
#define CMD_LINE_OPT_STAT_LCORE "stat-lcore"
#define CMD_LINE_OPT_ETH_DEST "eth-dest"
#define CMD_LINE_OPT_NO_NUMA "no-numa"
#define CMD_LINE_OPT_IPV6 "ipv6"
#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"
static int
parse_args(int argc, char **argv)
{
int opt, ret;
char **argvopt;
int option_index;
char *prgname = argv[0];
static struct option lgopts[] = {
{CMD_LINE_OPT_RX_CONFIG, 1, 0, 0},
{CMD_LINE_OPT_TX_CONFIG, 1, 0, 0},
{CMD_LINE_OPT_STAT_LCORE, 1, 0, 0},
{CMD_LINE_OPT_ETH_DEST, 1, 0, 0},
{CMD_LINE_OPT_NO_NUMA, 0, 0, 0},
{CMD_LINE_OPT_IPV6, 0, 0, 0},
{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},
{NULL, 0, 0, 0}
};
argvopt = argv;
while ((opt = getopt_long(argc, argvopt, "p:P",
lgopts, &option_index)) != EOF) {
switch (opt) {
case 'p':
enabled_port_mask = parse_portmask(optarg);
if (enabled_port_mask == 0) {
printf("invalid portmask\n");
print_usage(prgname);
return -1;
}
break;
case 'P':
printf("Promiscuous mode selected\n");
promiscuous_on = 1;
break;
case 0:
if (!strncmp(lgopts[option_index].name, CMD_LINE_OPT_RX_CONFIG,
sizeof(CMD_LINE_OPT_RX_CONFIG))) {
ret = parse_rx_config(optarg);
if (ret) {
printf("invalid rx-config\n");
print_usage(prgname);
return -1;
}
}
if (!strncmp(lgopts[option_index].name, CMD_LINE_OPT_TX_CONFIG,
sizeof(CMD_LINE_OPT_TX_CONFIG))) {
ret = parse_tx_config(optarg);
if (ret) {
printf("invalid tx-config\n");
print_usage(prgname);
return -1;
}
}
#if (APP_CPU_LOAD > 0)
if (!strncmp(lgopts[option_index].name, CMD_LINE_OPT_STAT_LCORE,
sizeof(CMD_LINE_OPT_STAT_LCORE))) {
cpu_load_lcore_id = parse_stat_lcore(optarg);
}
#endif
if (!strncmp(lgopts[option_index].name, CMD_LINE_OPT_ETH_DEST,
sizeof(CMD_LINE_OPT_ETH_DEST)))
parse_eth_dest(optarg);
if (!strncmp(lgopts[option_index].name, CMD_LINE_OPT_NO_NUMA,
sizeof(CMD_LINE_OPT_NO_NUMA))) {
printf("numa is disabled\n");
numa_on = 0;
}
#if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
if (!strncmp(lgopts[option_index].name, CMD_LINE_OPT_IPV6,
sizeof(CMD_LINE_OPT_IPV6))) {
printf("ipv6 is specified\n");
ipv6 = 1;
}
#endif
if (!strncmp(lgopts[option_index].name, CMD_LINE_OPT_NO_LTHREADS,
sizeof(CMD_LINE_OPT_NO_LTHREADS))) {
printf("l-threads model is disabled\n");
lthreads_on = 0;
}
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,
0};
printf("jumbo frame is enabled - disabling simple TX path\n");
if (0 == getopt_long(argc, argvopt, "", &lenopts,
&option_index)) {
ret = parse_max_pkt_len(optarg);
if ((ret < 64) || (ret > MAX_JUMBO_PKT_LEN)) {
printf("invalid packet length\n");
print_usage(prgname);
return -1;
}
}
printf("set jumbo frame max packet length to %u\n",
}
#if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
if (!strncmp(lgopts[option_index].name, CMD_LINE_OPT_HASH_ENTRY_NUM,
sizeof(CMD_LINE_OPT_HASH_ENTRY_NUM))) {
ret = parse_hash_entry_number(optarg);
if ((ret > 0) && (ret <= L3FWD_HASH_ENTRIES)) {
hash_entry_number = ret;
} else {
printf("invalid hash entry number\n");
print_usage(prgname);
return -1;
}
}
#endif
break;
default:
print_usage(prgname);
return -1;
}
}
if (optind >= 0)
argv[optind-1] = prgname;
ret = optind-1;
optind = 0;
return ret;
}
static void
print_ethaddr(
const char *name,
const struct ether_addr *eth_addr)
{
char buf[ETHER_ADDR_FMT_SIZE];
printf("%s%s", name, buf);
}
#if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
static void convert_ipv4_5tuple(struct ipv4_5tuple *key1,
union ipv4_5tuple_host *key2)
{
key2->proto = key1->proto;
key2->pad0 = 0;
key2->pad1 = 0;
}
static void convert_ipv6_5tuple(struct ipv6_5tuple *key1,
union ipv6_5tuple_host *key2)
{
uint32_t i;
for (i = 0; i < 16; i++) {
key2->ip_dst[i] = key1->ip_dst[i];
key2->ip_src[i] = key1->ip_src[i];
}
key2->proto = key1->proto;
key2->pad0 = 0;
key2->pad1 = 0;
key2->reserve = 0;
}
#define BYTE_VALUE_MAX 256
#define ALL_32_BITS 0xffffffff
#define BIT_8_TO_15 0x0000ff00
static inline void
populate_ipv4_few_flow_into_table(const struct rte_hash *h)
{
uint32_t i;
int32_t ret;
uint32_t array_len =
RTE_DIM(ipv4_l3fwd_route_array);
mask0 = _mm_set_epi32(ALL_32_BITS, ALL_32_BITS, ALL_32_BITS, BIT_8_TO_15);
for (i = 0; i < array_len; i++) {
struct ipv4_l3fwd_route entry;
union ipv4_5tuple_host newkey;
entry = ipv4_l3fwd_route_array[i];
convert_ipv4_5tuple(&entry.key, &newkey);
if (ret < 0) {
rte_exit(EXIT_FAILURE,
"Unable to add entry %" PRIu32
" to the l3fwd hash.\n", i);
}
ipv4_l3fwd_out_if[ret] = entry.if_out;
}
printf("Hash: Adding 0x%" PRIx32 " keys\n", array_len);
}
#define BIT_16_TO_23 0x00ff0000
static inline void
populate_ipv6_few_flow_into_table(const struct rte_hash *h)
{
uint32_t i;
int32_t ret;
uint32_t array_len =
RTE_DIM(ipv6_l3fwd_route_array);
mask1 = _mm_set_epi32(ALL_32_BITS, ALL_32_BITS, ALL_32_BITS, BIT_16_TO_23);
mask2 = _mm_set_epi32(0, 0, ALL_32_BITS, ALL_32_BITS);
for (i = 0; i < array_len; i++) {
struct ipv6_l3fwd_route entry;
union ipv6_5tuple_host newkey;
entry = ipv6_l3fwd_route_array[i];
convert_ipv6_5tuple(&entry.key, &newkey);
if (ret < 0) {
rte_exit(EXIT_FAILURE,
"Unable to add entry %" PRIu32
" to the l3fwd hash.\n", i);
}
ipv6_l3fwd_out_if[ret] = entry.if_out;
}
printf("Hash: Adding 0x%" PRIx32 "keys\n", array_len);
}
#define NUMBER_PORT_USED 4
static inline void
populate_ipv4_many_flow_into_table(const struct rte_hash *h,
unsigned int nr_flow)
{
unsigned i;
mask0 = _mm_set_epi32(ALL_32_BITS, ALL_32_BITS, ALL_32_BITS, BIT_8_TO_15);
for (i = 0; i < nr_flow; i++) {
struct ipv4_l3fwd_route entry;
union ipv4_5tuple_host newkey;
uint8_t a = (uint8_t)((i / NUMBER_PORT_USED) % BYTE_VALUE_MAX);
uint8_t b = (uint8_t)(((i / NUMBER_PORT_USED) / BYTE_VALUE_MAX) %
BYTE_VALUE_MAX);
uint8_t c = (uint8_t)((i / NUMBER_PORT_USED) / (BYTE_VALUE_MAX *
BYTE_VALUE_MAX));
memset(&entry, 0, sizeof(entry));
switch (i & (NUMBER_PORT_USED - 1)) {
case 0:
entry = ipv4_l3fwd_route_array[0];
entry.key.ip_dst =
IPv4(101, c, b, a);
break;
case 1:
entry = ipv4_l3fwd_route_array[1];
entry.key.ip_dst =
IPv4(201, c, b, a);
break;
case 2:
entry = ipv4_l3fwd_route_array[2];
entry.key.ip_dst =
IPv4(111, c, b, a);
break;
case 3:
entry = ipv4_l3fwd_route_array[3];
entry.key.ip_dst =
IPv4(211, c, b, a);
break;
};
convert_ipv4_5tuple(&entry.key, &newkey);
if (ret < 0)
rte_exit(EXIT_FAILURE,
"Unable to add entry %u\n", i);
ipv4_l3fwd_out_if[ret] = (uint8_t)entry.if_out;
}
printf("Hash: Adding 0x%x keys\n", nr_flow);
}
static inline void
populate_ipv6_many_flow_into_table(const struct rte_hash *h,
unsigned int nr_flow)
{
unsigned i;
mask1 = _mm_set_epi32(ALL_32_BITS, ALL_32_BITS, ALL_32_BITS, BIT_16_TO_23);
mask2 = _mm_set_epi32(0, 0, ALL_32_BITS, ALL_32_BITS);
for (i = 0; i < nr_flow; i++) {
struct ipv6_l3fwd_route entry;
union ipv6_5tuple_host newkey;
uint8_t a = (uint8_t) ((i / NUMBER_PORT_USED) % BYTE_VALUE_MAX);
uint8_t b = (uint8_t) (((i / NUMBER_PORT_USED) / BYTE_VALUE_MAX) %
BYTE_VALUE_MAX);
uint8_t c = (uint8_t) ((i / NUMBER_PORT_USED) / (BYTE_VALUE_MAX *
BYTE_VALUE_MAX));
memset(&entry, 0, sizeof(entry));
switch (i & (NUMBER_PORT_USED - 1)) {
case 0:
entry = ipv6_l3fwd_route_array[0];
break;
case 1:
entry = ipv6_l3fwd_route_array[1];
break;
case 2:
entry = ipv6_l3fwd_route_array[2];
break;
case 3:
entry = ipv6_l3fwd_route_array[3];
break;
};
entry.key.ip_dst[13] = c;
entry.key.ip_dst[14] = b;
entry.key.ip_dst[15] = a;
convert_ipv6_5tuple(&entry.key, &newkey);
if (ret < 0)
rte_exit(EXIT_FAILURE,
"Unable to add entry %u\n", i);
ipv6_l3fwd_out_if[ret] = (uint8_t) entry.if_out;
}
printf("Hash: Adding 0x%x keys\n", nr_flow);
}
static void
setup_hash(int socketid)
{
.entries = L3FWD_HASH_ENTRIES,
.key_len = sizeof(union ipv4_5tuple_host),
.hash_func = ipv4_hash_crc,
.hash_func_init_val = 0,
};
.entries = L3FWD_HASH_ENTRIES,
.key_len = sizeof(union ipv6_5tuple_host),
.hash_func = ipv6_hash_crc,
.hash_func_init_val = 0,
};
char s[64];
snprintf(s, sizeof(s), "ipv4_l3fwd_hash_%d", socketid);
ipv4_l3fwd_hash_params.
name = s;
ipv4_l3fwd_lookup_struct[socketid] =
if (ipv4_l3fwd_lookup_struct[socketid] == NULL)
rte_exit(EXIT_FAILURE,
"Unable to create the l3fwd hash on "
"socket %d\n", socketid);
snprintf(s, sizeof(s), "ipv6_l3fwd_hash_%d", socketid);
ipv6_l3fwd_hash_params.
name = s;
ipv6_l3fwd_lookup_struct[socketid] =
if (ipv6_l3fwd_lookup_struct[socketid] == NULL)
rte_exit(EXIT_FAILURE,
"Unable to create the l3fwd hash on "
"socket %d\n", socketid);
if (hash_entry_number != HASH_ENTRY_NUMBER_DEFAULT) {
if (ipv6 == 0) {
populate_ipv4_many_flow_into_table(
ipv4_l3fwd_lookup_struct[socketid], hash_entry_number);
} else {
populate_ipv6_many_flow_into_table(
ipv6_l3fwd_lookup_struct[socketid], hash_entry_number);
}
} else {
if (ipv6 == 0) {
populate_ipv4_few_flow_into_table(
ipv4_l3fwd_lookup_struct[socketid]);
} else {
populate_ipv6_few_flow_into_table(
ipv6_l3fwd_lookup_struct[socketid]);
}
}
}
#endif
#if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
static void
setup_lpm(int socketid)
{
unsigned i;
int ret;
char s[64];
snprintf(s, sizeof(s), "IPV4_L3FWD_LPM_%d", socketid);
IPV4_L3FWD_LPM_MAX_RULES, 0);
if (ipv4_l3fwd_lookup_struct[socketid] == NULL)
rte_exit(EXIT_FAILURE,
"Unable to create the l3fwd LPM table"
" on socket %d\n", socketid);
for (i = 0; i < IPV4_L3FWD_NUM_ROUTES; i++) {
if ((1 << ipv4_l3fwd_route_array[i].if_out &
enabled_port_mask) == 0)
continue;
ipv4_l3fwd_route_array[i].ip,
ipv4_l3fwd_route_array[i].depth,
ipv4_l3fwd_route_array[i].if_out);
if (ret < 0) {
rte_exit(EXIT_FAILURE,
"Unable to add entry %u to the "
"l3fwd LPM table on socket %d\n",
i, socketid);
}
printf("LPM: Adding route 0x%08x / %d (%d)\n",
(unsigned)ipv4_l3fwd_route_array[i].ip,
ipv4_l3fwd_route_array[i].depth,
ipv4_l3fwd_route_array[i].if_out);
}
snprintf(s, sizeof(s), "IPV6_L3FWD_LPM_%d", socketid);
config.max_rules = IPV6_L3FWD_LPM_MAX_RULES;
config.number_tbl8s = IPV6_L3FWD_LPM_NUMBER_TBL8S;
config.flags = 0;
&config);
if (ipv6_l3fwd_lookup_struct[socketid] == NULL)
rte_exit(EXIT_FAILURE,
"Unable to create the l3fwd LPM table"
" on socket %d\n", socketid);
for (i = 0; i < IPV6_L3FWD_NUM_ROUTES; i++) {
if ((1 << ipv6_l3fwd_route_array[i].if_out &
enabled_port_mask) == 0)
continue;
ipv6_l3fwd_route_array[i].ip,
ipv6_l3fwd_route_array[i].depth,
ipv6_l3fwd_route_array[i].if_out);
if (ret < 0) {
rte_exit(EXIT_FAILURE,
"Unable to add entry %u to the "
"l3fwd LPM table on socket %d\n",
i, socketid);
}
printf("LPM: Adding route %s / %d (%d)\n",
"IPV6",
ipv6_l3fwd_route_array[i].depth,
ipv6_l3fwd_route_array[i].if_out);
}
}
#endif
static int
init_mem(unsigned nb_mbuf)
{
struct lcore_conf *qconf;
int socketid;
unsigned lcore_id;
char s[64];
for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
continue;
if (numa_on)
else
socketid = 0;
if (socketid >= NB_SOCKETS) {
rte_exit(EXIT_FAILURE,
"Socket %d of lcore %u is out of range %d\n",
socketid, lcore_id, NB_SOCKETS);
}
if (pktmbuf_pool[socketid] == NULL) {
snprintf(s, sizeof(s), "mbuf_pool_%d", socketid);
pktmbuf_pool[socketid] =
MEMPOOL_CACHE_SIZE, 0,
RTE_MBUF_DEFAULT_BUF_SIZE, socketid);
if (pktmbuf_pool[socketid] == NULL)
"Cannot init mbuf pool on socket %d\n", socketid);
else
printf("Allocated mbuf pool on socket %d\n", socketid);
#if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
setup_lpm(socketid);
#else
setup_hash(socketid);
#endif
}
qconf = &lcore_conf[lcore_id];
qconf->ipv4_lookup_struct = ipv4_l3fwd_lookup_struct[socketid];
qconf->ipv6_lookup_struct = ipv6_l3fwd_lookup_struct[socketid];
}
return 0;
}
static void
check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
{
#define CHECK_INTERVAL 100
#define MAX_CHECK_TIME 90
uint8_t portid, count, all_ports_up, print_flag = 0;
printf("\nChecking link status");
fflush(stdout);
for (count = 0; count <= MAX_CHECK_TIME; count++) {
all_ports_up = 1;
for (portid = 0; portid < port_num; portid++) {
if ((port_mask & (1 << portid)) == 0)
continue;
memset(&link, 0, sizeof(link));
if (print_flag == 1) {
if (link.link_status)
printf("Port %d Link Up - speed %u "
"Mbps - %s\n", (uint8_t)portid,
(unsigned)link.link_speed,
("full-duplex") : ("half-duplex\n"));
else
printf("Port %d Link Down\n",
(uint8_t)portid);
continue;
}
if (link.link_status == 0) {
all_ports_up = 0;
break;
}
}
if (print_flag == 1)
break;
if (all_ports_up == 0) {
printf(".");
fflush(stdout);
}
if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
print_flag = 1;
printf("done\n");
}
}
}
int
main(int argc, char **argv)
{
struct rte_eth_dev_info dev_info;
int ret;
int i;
unsigned nb_ports;
uint16_t queueid;
unsigned lcore_id;
uint32_t n_tx_queue, nb_lcores;
uint8_t portid, nb_rx_queue, queue, socketid;
if (ret < 0)
rte_exit(EXIT_FAILURE,
"Invalid EAL parameters\n");
argc -= ret;
argv += ret;
for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
((uint64_t)portid << 40);
*(uint64_t *)(val_eth + portid) = dest_eth_addr[portid];
}
ret = parse_args(argc, argv);
if (ret < 0)
rte_exit(EXIT_FAILURE,
"Invalid L3FWD parameters\n");
if (check_lcore_params() < 0)
rte_exit(EXIT_FAILURE,
"check_lcore_params failed\n");
printf("Initializing rx-queues...\n");
ret = init_rx_queues();
if (ret < 0)
rte_exit(EXIT_FAILURE,
"init_rx_queues failed\n");
printf("Initializing tx-threads...\n");
ret = init_tx_threads();
if (ret < 0)
rte_exit(EXIT_FAILURE,
"init_tx_threads failed\n");
printf("Initializing rings...\n");
ret = init_rx_rings();
if (ret < 0)
rte_exit(EXIT_FAILURE,
"init_rx_rings failed\n");
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");
for (portid = 0; portid < nb_ports; portid++) {
if ((enabled_port_mask & (1 << portid)) == 0) {
printf("\nSkipping disabled port %d\n", portid);
continue;
}
printf("Initializing port %d ... ", portid);
fflush(stdout);
nb_rx_queue = get_port_n_rx_queues(portid);
n_tx_queue = nb_lcores;
if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
n_tx_queue = MAX_TX_QUEUE_PER_PORT;
printf("Creating queues: nb_rxq=%d nb_txq=%u... ",
nb_rx_queue, (unsigned)n_tx_queue);
(uint16_t)n_tx_queue, &port_conf);
if (ret < 0)
rte_exit(EXIT_FAILURE,
"Cannot configure device: err=%d, port=%d\n",
ret, portid);
print_ethaddr(" Address:", &ports_eth_addr[portid]);
printf(", ");
print_ethaddr("Destination:",
(
const struct ether_addr *)&dest_eth_addr[portid]);
printf(", ");
ret = init_mem(NB_MBUF);
if (ret < 0)
rte_exit(EXIT_FAILURE,
"init_mem failed\n");
queueid = 0;
for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
continue;
if (numa_on)
else
socketid = 0;
printf("txq=%u,%d,%d ", lcore_id, queueid, socketid);
fflush(stdout);
txconf = &dev_info.default_txconf;
socketid, txconf);
if (ret < 0)
rte_exit(EXIT_FAILURE,
"rte_eth_tx_queue_setup: err=%d, "
"port=%d\n", ret, portid);
tx_thread[lcore_id].tx_queue_id[portid] = queueid;
queueid++;
}
printf("\n");
}
for (i = 0; i < n_rx_thread; i++) {
lcore_id = rx_thread[i].conf.lcore_id;
"Cannot start Rx thread on lcore %u: lcore disabled\n",
lcore_id
);
}
printf("\nInitializing rx queues for Rx thread %d on lcore %u ... ",
i, lcore_id);
fflush(stdout);
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 (numa_on)
else
socketid = 0;
printf("rxq=%d,%d,%d ", portid, queueid, socketid);
fflush(stdout);
socketid,
NULL,
pktmbuf_pool[socketid]);
if (ret < 0)
rte_exit(EXIT_FAILURE,
"rte_eth_rx_queue_setup: err=%d, "
"port=%d\n", ret, portid);
}
}
printf("\n");
for (portid = 0; portid < nb_ports; portid++) {
if ((enabled_port_mask & (1 << portid)) == 0)
continue;
if (ret < 0)
rte_exit(EXIT_FAILURE,
"rte_eth_dev_start: err=%d, port=%d\n",
ret, portid);
if (promiscuous_on)
}
check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask);
if (lthreads_on) {
printf("Starting L-Threading Model\n");
#if (APP_CPU_LOAD > 0)
if (cpu_load_lcore_id > 0)
nb_lcores--;
#endif
lthread_num_schedulers_set(nb_lcores);
lthread_master_spawner(NULL);
} else {
printf("Starting P-Threading Model\n");
return -1;
}
}
return 0;
}