1 /*- 2 * BSD LICENSE 3 * 4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 11 * * Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * * Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in 15 * the documentation and/or other materials provided with the 16 * distribution. 17 * * Neither the name of Intel Corporation nor the names of its 18 * contributors may be used to endorse or promote products derived 19 * from this software without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 32 */ 33 34 #include <stdio.h> 35 #include <stdlib.h> 36 #include <string.h> 37 #include <unistd.h> 38 #include <stdint.h> 39 #include <stdarg.h> 40 #include <inttypes.h> 41 #include <sys/queue.h> 42 #include <errno.h> 43 #include <netinet/ip.h> 44 45 #include <rte_common.h> 46 #include <rte_memory.h> 47 #include <rte_memzone.h> 48 #include <rte_eal.h> 49 #include <rte_launch.h> 50 #include <rte_per_lcore.h> 51 #include <rte_lcore.h> 52 #include <rte_branch_prediction.h> 53 #include <rte_atomic.h> 54 #include <rte_ring.h> 55 #include <rte_log.h> 56 #include <rte_debug.h> 57 #include <rte_mempool.h> 58 #include <rte_memcpy.h> 59 #include <rte_mbuf.h> 60 #include <rte_ether.h> 61 #include <rte_interrupts.h> 62 #include <rte_pci.h> 63 #include <rte_ethdev.h> 64 #include <rte_byteorder.h> 65 #include <rte_malloc.h> 66 #include <rte_string_fns.h> 67 68 #include "common.h" 69 #include "args.h" 70 #include "init.h" 71 72 /* 73 * When doing reads from the NIC or the client queues, 74 * use this batch size 75 */ 76 #define PACKET_READ_SIZE 32 77 78 /* 79 * Local buffers to put packets in, used to send packets in bursts to the 80 * clients 81 */ 82 struct client_rx_buf { 83 struct rte_mbuf *buffer[PACKET_READ_SIZE]; 84 uint16_t count; 85 }; 86 87 /* One buffer per client rx queue - dynamically allocate array */ 88 static struct client_rx_buf *cl_rx_buf; 89 90 static const char * 91 get_printable_mac_addr(uint8_t port) 92 { 93 static const char err_address[] = "00:00:00:00:00:00"; 94 static char addresses[RTE_MAX_ETHPORTS][sizeof(err_address)]; 95 96 if (unlikely(port >= RTE_MAX_ETHPORTS)) 97 return err_address; 98 if (unlikely(addresses[port][0]=='\0')){ 99 struct ether_addr mac; 100 rte_eth_macaddr_get(port, &mac); 101 snprintf(addresses[port], sizeof(addresses[port]), 102 "%02x:%02x:%02x:%02x:%02x:%02x\n", 103 mac.addr_bytes[0], mac.addr_bytes[1], mac.addr_bytes[2], 104 mac.addr_bytes[3], mac.addr_bytes[4], mac.addr_bytes[5]); 105 } 106 return addresses[port]; 107 } 108 109 /* 110 * This function displays the recorded statistics for each port 111 * and for each client. It uses ANSI terminal codes to clear 112 * screen when called. It is called from a single non-master 113 * thread in the server process, when the process is run with more 114 * than one lcore enabled. 115 */ 116 static void 117 do_stats_display(void) 118 { 119 unsigned i, j; 120 const char clr[] = { 27, '[', '2', 'J', '\0' }; 121 const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' }; 122 uint64_t port_tx[RTE_MAX_ETHPORTS], port_tx_drop[RTE_MAX_ETHPORTS]; 123 uint64_t client_tx[MAX_CLIENTS], client_tx_drop[MAX_CLIENTS]; 124 125 /* to get TX stats, we need to do some summing calculations */ 126 memset(port_tx, 0, sizeof(port_tx)); 127 memset(port_tx_drop, 0, sizeof(port_tx_drop)); 128 memset(client_tx, 0, sizeof(client_tx)); 129 memset(client_tx_drop, 0, sizeof(client_tx_drop)); 130 131 for (i = 0; i < num_clients; i++){ 132 const volatile struct tx_stats *tx = &ports->tx_stats[i]; 133 for (j = 0; j < ports->num_ports; j++){ 134 /* assign to local variables here, save re-reading volatile vars */ 135 const uint64_t tx_val = tx->tx[ports->id[j]]; 136 const uint64_t drop_val = tx->tx_drop[ports->id[j]]; 137 port_tx[j] += tx_val; 138 port_tx_drop[j] += drop_val; 139 client_tx[i] += tx_val; 140 client_tx_drop[i] += drop_val; 141 } 142 } 143 144 /* Clear screen and move to top left */ 145 printf("%s%s", clr, topLeft); 146 147 printf("PORTS\n"); 148 printf("-----\n"); 149 for (i = 0; i < ports->num_ports; i++) 150 printf("Port %u: '%s'\t", (unsigned)ports->id[i], 151 get_printable_mac_addr(ports->id[i])); 152 printf("\n\n"); 153 for (i = 0; i < ports->num_ports; i++){ 154 printf("Port %u - rx: %9"PRIu64"\t" 155 "tx: %9"PRIu64"\n", 156 (unsigned)ports->id[i], ports->rx_stats.rx[i], 157 port_tx[i]); 158 } 159 160 printf("\nCLIENTS\n"); 161 printf("-------\n"); 162 for (i = 0; i < num_clients; i++){ 163 const unsigned long long rx = clients[i].stats.rx; 164 const unsigned long long rx_drop = clients[i].stats.rx_drop; 165 printf("Client %2u - rx: %9llu, rx_drop: %9llu\n" 166 " tx: %9"PRIu64", tx_drop: %9"PRIu64"\n", 167 i, rx, rx_drop, client_tx[i], client_tx_drop[i]); 168 } 169 170 printf("\n"); 171 } 172 173 /* 174 * The function called from each non-master lcore used by the process. 175 * The test_and_set function is used to randomly pick a single lcore on which 176 * the code to display the statistics will run. Otherwise, the code just 177 * repeatedly sleeps. 178 */ 179 static int 180 sleep_lcore(__attribute__((unused)) void *dummy) 181 { 182 /* Used to pick a display thread - static, so zero-initialised */ 183 static rte_atomic32_t display_stats; 184 185 /* Only one core should display stats */ 186 if (rte_atomic32_test_and_set(&display_stats)) { 187 const unsigned sleeptime = 1; 188 printf("Core %u displaying statistics\n", rte_lcore_id()); 189 190 /* Longer initial pause so above printf is seen */ 191 sleep(sleeptime * 3); 192 193 /* Loop forever: sleep always returns 0 or <= param */ 194 while (sleep(sleeptime) <= sleeptime) 195 do_stats_display(); 196 } 197 return 0; 198 } 199 200 /* 201 * Function to set all the client statistic values to zero. 202 * Called at program startup. 203 */ 204 static void 205 clear_stats(void) 206 { 207 unsigned i; 208 209 for (i = 0; i < num_clients; i++) 210 clients[i].stats.rx = clients[i].stats.rx_drop = 0; 211 } 212 213 /* 214 * send a burst of traffic to a client, assuming there are packets 215 * available to be sent to this client 216 */ 217 static void 218 flush_rx_queue(uint16_t client) 219 { 220 uint16_t j; 221 struct client *cl; 222 223 if (cl_rx_buf[client].count == 0) 224 return; 225 226 cl = &clients[client]; 227 if (rte_ring_enqueue_bulk(cl->rx_q, (void **)cl_rx_buf[client].buffer, 228 cl_rx_buf[client].count, NULL) == 0){ 229 for (j = 0; j < cl_rx_buf[client].count; j++) 230 rte_pktmbuf_free(cl_rx_buf[client].buffer[j]); 231 cl->stats.rx_drop += cl_rx_buf[client].count; 232 } 233 else 234 cl->stats.rx += cl_rx_buf[client].count; 235 236 cl_rx_buf[client].count = 0; 237 } 238 239 /* 240 * marks a packet down to be sent to a particular client process 241 */ 242 static inline void 243 enqueue_rx_packet(uint8_t client, struct rte_mbuf *buf) 244 { 245 cl_rx_buf[client].buffer[cl_rx_buf[client].count++] = buf; 246 } 247 248 /* 249 * This function takes a group of packets and routes them 250 * individually to the client process. Very simply round-robins the packets 251 * without checking any of the packet contents. 252 */ 253 static void 254 process_packets(uint32_t port_num __rte_unused, 255 struct rte_mbuf *pkts[], uint16_t rx_count) 256 { 257 uint16_t i; 258 uint8_t client = 0; 259 260 for (i = 0; i < rx_count; i++) { 261 enqueue_rx_packet(client, pkts[i]); 262 263 if (++client == num_clients) 264 client = 0; 265 } 266 267 for (i = 0; i < num_clients; i++) 268 flush_rx_queue(i); 269 } 270 271 /* 272 * Function called by the master lcore of the DPDK process. 273 */ 274 static void 275 do_packet_forwarding(void) 276 { 277 unsigned port_num = 0; /* indexes the port[] array */ 278 279 for (;;) { 280 struct rte_mbuf *buf[PACKET_READ_SIZE]; 281 uint16_t rx_count; 282 283 /* read a port */ 284 rx_count = rte_eth_rx_burst(ports->id[port_num], 0, \ 285 buf, PACKET_READ_SIZE); 286 ports->rx_stats.rx[port_num] += rx_count; 287 288 /* Now process the NIC packets read */ 289 if (likely(rx_count > 0)) 290 process_packets(port_num, buf, rx_count); 291 292 /* move to next port */ 293 if (++port_num == ports->num_ports) 294 port_num = 0; 295 } 296 } 297 298 int 299 main(int argc, char *argv[]) 300 { 301 /* initialise the system */ 302 if (init(argc, argv) < 0 ) 303 return -1; 304 RTE_LOG(INFO, APP, "Finished Process Init.\n"); 305 306 cl_rx_buf = calloc(num_clients, sizeof(cl_rx_buf[0])); 307 308 /* clear statistics */ 309 clear_stats(); 310 311 /* put all other cores to sleep bar master */ 312 rte_eal_mp_remote_launch(sleep_lcore, NULL, SKIP_MASTER); 313 314 do_packet_forwarding(); 315 return 0; 316 } 317