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 <inttypes.h> 42 #include <sys/queue.h> 43 #include <errno.h> 44 #include <netinet/ip.h> 45 46 #include <rte_common.h> 47 #include <rte_memory.h> 48 #include <rte_memzone.h> 49 #include <rte_tailq.h> 50 #include <rte_eal.h> 51 #include <rte_byteorder.h> 52 #include <rte_launch.h> 53 #include <rte_per_lcore.h> 54 #include <rte_lcore.h> 55 #include <rte_branch_prediction.h> 56 #include <rte_atomic.h> 57 #include <rte_ring.h> 58 #include <rte_log.h> 59 #include <rte_debug.h> 60 #include <rte_mempool.h> 61 #include <rte_memcpy.h> 62 #include <rte_mbuf.h> 63 #include <rte_ether.h> 64 #include <rte_interrupts.h> 65 #include <rte_pci.h> 66 #include <rte_ethdev.h> 67 #include <rte_byteorder.h> 68 #include <rte_malloc.h> 69 #include <rte_fbk_hash.h> 70 #include <rte_string_fns.h> 71 72 #include "common.h" 73 #include "args.h" 74 #include "init.h" 75 76 /* 77 * When doing reads from the NIC or the client queues, 78 * use this batch size 79 */ 80 #define PACKET_READ_SIZE 32 81 82 /* 83 * Local buffers to put packets in, used to send packets in bursts to the 84 * clients 85 */ 86 struct client_rx_buf { 87 struct rte_mbuf *buffer[PACKET_READ_SIZE]; 88 uint16_t count; 89 }; 90 91 /* One buffer per client rx queue - dynamically allocate array */ 92 static struct client_rx_buf *cl_rx_buf; 93 94 static const char * 95 get_printable_mac_addr(uint8_t port) 96 { 97 static const char err_address[] = "00:00:00:00:00:00"; 98 static char addresses[RTE_MAX_ETHPORTS][sizeof(err_address)]; 99 100 if (unlikely(port >= RTE_MAX_ETHPORTS)) 101 return err_address; 102 if (unlikely(addresses[port][0]=='\0')){ 103 struct ether_addr mac; 104 rte_eth_macaddr_get(port, &mac); 105 snprintf(addresses[port], sizeof(addresses[port]), 106 "%02x:%02x:%02x:%02x:%02x:%02x\n", 107 mac.addr_bytes[0], mac.addr_bytes[1], mac.addr_bytes[2], 108 mac.addr_bytes[3], mac.addr_bytes[4], mac.addr_bytes[5]); 109 } 110 return addresses[port]; 111 } 112 113 /* 114 * This function displays the recorded statistics for each port 115 * and for each client. It uses ANSI terminal codes to clear 116 * screen when called. It is called from a single non-master 117 * thread in the server process, when the process is run with more 118 * than one lcore enabled. 119 */ 120 static void 121 do_stats_display(void) 122 { 123 unsigned i, j; 124 const char clr[] = { 27, '[', '2', 'J', '\0' }; 125 const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' }; 126 uint64_t port_tx[RTE_MAX_ETHPORTS], port_tx_drop[RTE_MAX_ETHPORTS]; 127 uint64_t client_tx[MAX_CLIENTS], client_tx_drop[MAX_CLIENTS]; 128 129 /* to get TX stats, we need to do some summing calculations */ 130 memset(port_tx, 0, sizeof(port_tx)); 131 memset(port_tx_drop, 0, sizeof(port_tx_drop)); 132 memset(client_tx, 0, sizeof(client_tx)); 133 memset(client_tx_drop, 0, sizeof(client_tx_drop)); 134 135 for (i = 0; i < num_clients; i++){ 136 const volatile struct tx_stats *tx = &ports->tx_stats[i]; 137 for (j = 0; j < ports->num_ports; j++){ 138 /* assign to local variables here, save re-reading volatile vars */ 139 const uint64_t tx_val = tx->tx[ports->id[j]]; 140 const uint64_t drop_val = tx->tx_drop[ports->id[j]]; 141 port_tx[j] += tx_val; 142 port_tx_drop[j] += drop_val; 143 client_tx[i] += tx_val; 144 client_tx_drop[i] += drop_val; 145 } 146 } 147 148 /* Clear screen and move to top left */ 149 printf("%s%s", clr, topLeft); 150 151 printf("PORTS\n"); 152 printf("-----\n"); 153 for (i = 0; i < ports->num_ports; i++) 154 printf("Port %u: '%s'\t", (unsigned)ports->id[i], 155 get_printable_mac_addr(ports->id[i])); 156 printf("\n\n"); 157 for (i = 0; i < ports->num_ports; i++){ 158 printf("Port %u - rx: %9"PRIu64"\t" 159 "tx: %9"PRIu64"\n", 160 (unsigned)ports->id[i], ports->rx_stats.rx[i], 161 port_tx[i]); 162 } 163 164 printf("\nCLIENTS\n"); 165 printf("-------\n"); 166 for (i = 0; i < num_clients; i++){ 167 const unsigned long long rx = clients[i].stats.rx; 168 const unsigned long long rx_drop = clients[i].stats.rx_drop; 169 printf("Client %2u - rx: %9llu, rx_drop: %9llu\n" 170 " tx: %9"PRIu64", tx_drop: %9"PRIu64"\n", 171 i, rx, rx_drop, client_tx[i], client_tx_drop[i]); 172 } 173 174 printf("\n"); 175 } 176 177 /* 178 * The function called from each non-master lcore used by the process. 179 * The test_and_set function is used to randomly pick a single lcore on which 180 * the code to display the statistics will run. Otherwise, the code just 181 * repeatedly sleeps. 182 */ 183 static int 184 sleep_lcore(__attribute__((unused)) void *dummy) 185 { 186 /* Used to pick a display thread - static, so zero-initialised */ 187 static rte_atomic32_t display_stats; 188 189 /* Only one core should display stats */ 190 if (rte_atomic32_test_and_set(&display_stats)) { 191 const unsigned sleeptime = 1; 192 printf("Core %u displaying statistics\n", rte_lcore_id()); 193 194 /* Longer initial pause so above printf is seen */ 195 sleep(sleeptime * 3); 196 197 /* Loop forever: sleep always returns 0 or <= param */ 198 while (sleep(sleeptime) <= sleeptime) 199 do_stats_display(); 200 } 201 return 0; 202 } 203 204 /* 205 * Function to set all the client statistic values to zero. 206 * Called at program startup. 207 */ 208 static void 209 clear_stats(void) 210 { 211 unsigned i; 212 213 for (i = 0; i < num_clients; i++) 214 clients[i].stats.rx = clients[i].stats.rx_drop = 0; 215 } 216 217 /* 218 * send a burst of traffic to a client, assuming there are packets 219 * available to be sent to this client 220 */ 221 static void 222 flush_rx_queue(uint16_t client) 223 { 224 uint16_t j; 225 struct client *cl; 226 227 if (cl_rx_buf[client].count == 0) 228 return; 229 230 cl = &clients[client]; 231 if (rte_ring_enqueue_bulk(cl->rx_q, (void **)cl_rx_buf[client].buffer, 232 cl_rx_buf[client].count) != 0){ 233 for (j = 0; j < cl_rx_buf[client].count; j++) 234 rte_pktmbuf_free(cl_rx_buf[client].buffer[j]); 235 cl->stats.rx_drop += cl_rx_buf[client].count; 236 } 237 else 238 cl->stats.rx += cl_rx_buf[client].count; 239 240 cl_rx_buf[client].count = 0; 241 } 242 243 /* 244 * marks a packet down to be sent to a particular client process 245 */ 246 static inline void 247 enqueue_rx_packet(uint8_t client, struct rte_mbuf *buf) 248 { 249 cl_rx_buf[client].buffer[cl_rx_buf[client].count++] = buf; 250 } 251 252 /* 253 * This function takes a group of packets and routes them 254 * individually to the client process. Very simply round-robins the packets 255 * without checking any of the packet contents. 256 */ 257 static void 258 process_packets(uint32_t port_num __rte_unused, 259 struct rte_mbuf *pkts[], uint16_t rx_count) 260 { 261 uint16_t i; 262 uint8_t client = 0; 263 264 for (i = 0; i < rx_count; i++) { 265 enqueue_rx_packet(client, pkts[i]); 266 267 if (++client == num_clients) 268 client = 0; 269 } 270 271 for (i = 0; i < num_clients; i++) 272 flush_rx_queue(i); 273 } 274 275 /* 276 * Function called by the master lcore of the DPDK process. 277 */ 278 static void 279 do_packet_forwarding(void) 280 { 281 unsigned port_num = 0; /* indexes the port[] array */ 282 283 for (;;) { 284 struct rte_mbuf *buf[PACKET_READ_SIZE]; 285 uint16_t rx_count; 286 287 /* read a port */ 288 rx_count = rte_eth_rx_burst(ports->id[port_num], 0, \ 289 buf, PACKET_READ_SIZE); 290 ports->rx_stats.rx[port_num] += rx_count; 291 292 /* Now process the NIC packets read */ 293 if (likely(rx_count > 0)) 294 process_packets(port_num, buf, rx_count); 295 296 /* move to next port */ 297 if (++port_num == ports->num_ports) 298 port_num = 0; 299 } 300 } 301 302 int 303 main(int argc, char *argv[]) 304 { 305 /* initialise the system */ 306 if (init(argc, argv) < 0 ) 307 return -1; 308 RTE_LOG(INFO, APP, "Finished Process Init.\n"); 309 310 cl_rx_buf = calloc(num_clients, sizeof(cl_rx_buf[0])); 311 312 /* clear statistics */ 313 clear_stats(); 314 315 /* put all other cores to sleep bar master */ 316 rte_eal_mp_remote_launch(sleep_lcore, NULL, SKIP_MASTER); 317 318 do_packet_forwarding(); 319 return 0; 320 } 321