1 /*- 2 * BSD LICENSE 3 * 4 * Copyright(c) 2010-2013 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 35 #include <stdio.h> 36 #include <stdlib.h> 37 #include <string.h> 38 #include <unistd.h> 39 #include <stdint.h> 40 #include <stdarg.h> 41 #include <inttypes.h> 42 #include <inttypes.h> 43 #include <sys/queue.h> 44 #include <errno.h> 45 #include <netinet/ip.h> 46 47 #include <rte_common.h> 48 #include <rte_memory.h> 49 #include <rte_memzone.h> 50 #include <rte_tailq.h> 51 #include <rte_eal.h> 52 #include <rte_byteorder.h> 53 #include <rte_launch.h> 54 #include <rte_per_lcore.h> 55 #include <rte_lcore.h> 56 #include <rte_branch_prediction.h> 57 #include <rte_atomic.h> 58 #include <rte_ring.h> 59 #include <rte_log.h> 60 #include <rte_debug.h> 61 #include <rte_mempool.h> 62 #include <rte_memcpy.h> 63 #include <rte_mbuf.h> 64 #include <rte_ether.h> 65 #include <rte_interrupts.h> 66 #include <rte_pci.h> 67 #include <rte_ethdev.h> 68 #include <rte_byteorder.h> 69 #include <rte_malloc.h> 70 #include <rte_fbk_hash.h> 71 #include <rte_string_fns.h> 72 73 #include "common.h" 74 #include "args.h" 75 #include "init.h" 76 #include "main.h" 77 78 /* 79 * When doing reads from the NIC or the client queues, 80 * use this batch size 81 */ 82 #define PACKET_READ_SIZE 32 83 84 /* 85 * Local buffers to put packets in, used to send packets in bursts to the 86 * clients 87 */ 88 struct client_rx_buf { 89 struct rte_mbuf *buffer[PACKET_READ_SIZE]; 90 uint16_t count; 91 }; 92 93 /* One buffer per client rx queue - dynamically allocate array */ 94 static struct client_rx_buf *cl_rx_buf; 95 96 static const char * 97 get_printable_mac_addr(uint8_t port) 98 { 99 static const char err_address[] = "00:00:00:00:00:00"; 100 static char addresses[RTE_MAX_ETHPORTS][sizeof(err_address)]; 101 102 if (unlikely(port >= RTE_MAX_ETHPORTS)) 103 return err_address; 104 if (unlikely(addresses[port][0]=='\0')){ 105 struct ether_addr mac; 106 rte_eth_macaddr_get(port, &mac); 107 rte_snprintf(addresses[port], sizeof(addresses[port]), 108 "%02x:%02x:%02x:%02x:%02x:%02x\n", 109 mac.addr_bytes[0], mac.addr_bytes[1], mac.addr_bytes[2], 110 mac.addr_bytes[3], mac.addr_bytes[4], mac.addr_bytes[5]); 111 } 112 return addresses[port]; 113 } 114 115 /* 116 * This function displays the recorded statistics for each port 117 * and for each client. It uses ANSI terminal codes to clear 118 * screen when called. It is called from a single non-master 119 * thread in the server process, when the process is run with more 120 * than one lcore enabled. 121 */ 122 static void 123 do_stats_display(void) 124 { 125 unsigned i, j; 126 const char clr[] = { 27, '[', '2', 'J', '\0' }; 127 const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' }; 128 uint64_t port_tx[RTE_MAX_ETHPORTS], port_tx_drop[RTE_MAX_ETHPORTS]; 129 uint64_t client_tx[MAX_CLIENTS], client_tx_drop[MAX_CLIENTS]; 130 131 /* to get TX stats, we need to do some summing calculations */ 132 memset(port_tx, 0, sizeof(port_tx)); 133 memset(port_tx_drop, 0, sizeof(port_tx_drop)); 134 memset(client_tx, 0, sizeof(client_tx)); 135 memset(client_tx_drop, 0, sizeof(client_tx_drop)); 136 137 for (i = 0; i < num_clients; i++){ 138 const volatile struct tx_stats *tx = &ports->tx_stats[i]; 139 for (j = 0; j < ports->num_ports; j++){ 140 /* assign to local variables here, save re-reading volatile vars */ 141 const uint64_t tx_val = tx->tx[ports->id[j]]; 142 const uint64_t drop_val = tx->tx_drop[ports->id[j]]; 143 port_tx[j] += tx_val; 144 port_tx_drop[j] += drop_val; 145 client_tx[i] += tx_val; 146 client_tx_drop[i] += drop_val; 147 } 148 } 149 150 /* Clear screen and move to top left */ 151 printf("%s%s", clr, topLeft); 152 153 printf("PORTS\n"); 154 printf("-----\n"); 155 for (i = 0; i < ports->num_ports; i++) 156 printf("Port %u: '%s'\t", (unsigned)ports->id[i], 157 get_printable_mac_addr(ports->id[i])); 158 printf("\n\n"); 159 for (i = 0; i < ports->num_ports; i++){ 160 printf("Port %u - rx: %9"PRIu64"\t" 161 "tx: %9"PRIu64"\n", 162 (unsigned)ports->id[i], ports->rx_stats.rx[i], 163 port_tx[i]); 164 } 165 166 printf("\nCLIENTS\n"); 167 printf("-------\n"); 168 for (i = 0; i < num_clients; i++){ 169 const unsigned long long rx = clients[i].stats.rx; 170 const unsigned long long rx_drop = clients[i].stats.rx_drop; 171 printf("Client %2u - rx: %9llu, rx_drop: %9llu\n" 172 " tx: %9"PRIu64", tx_drop: %9"PRIu64"\n", 173 i, rx, rx_drop, client_tx[i], client_tx_drop[i]); 174 } 175 176 printf("\n"); 177 } 178 179 /* 180 * The function called from each non-master lcore used by the process. 181 * The test_and_set function is used to randomly pick a single lcore on which 182 * the code to display the statistics will run. Otherwise, the code just 183 * repeatedly sleeps. 184 */ 185 static int 186 sleep_lcore(__attribute__((unused)) void *dummy) 187 { 188 /* Used to pick a display thread - static, so zero-initialised */ 189 static rte_atomic32_t display_stats; 190 191 /* Only one core should display stats */ 192 if (rte_atomic32_test_and_set(&display_stats)) { 193 const unsigned sleeptime = 1; 194 printf("Core %u displaying statistics\n", rte_lcore_id()); 195 196 /* Longer initial pause so above printf is seen */ 197 sleep(sleeptime * 3); 198 199 /* Loop forever: sleep always returns 0 or <= param */ 200 while (sleep(sleeptime) <= sleeptime) 201 do_stats_display(); 202 } 203 return 0; 204 } 205 206 /* 207 * Function to set all the client statistic values to zero. 208 * Called at program startup. 209 */ 210 static void 211 clear_stats(void) 212 { 213 unsigned i; 214 215 for (i = 0; i < num_clients; i++) 216 clients[i].stats.rx = clients[i].stats.rx_drop = 0; 217 } 218 219 /* 220 * send a burst of traffic to a client, assuming there are packets 221 * available to be sent to this client 222 */ 223 static void 224 flush_rx_queue(uint16_t client) 225 { 226 uint16_t j; 227 struct client *cl; 228 229 if (cl_rx_buf[client].count == 0) 230 return; 231 232 cl = &clients[client]; 233 if (rte_ring_enqueue_bulk(cl->rx_q, (void **)cl_rx_buf[client].buffer, 234 cl_rx_buf[client].count) != 0){ 235 for (j = 0; j < cl_rx_buf[client].count; j++) 236 rte_pktmbuf_free(cl_rx_buf[client].buffer[j]); 237 cl->stats.rx_drop += cl_rx_buf[client].count; 238 } 239 else 240 cl->stats.rx += cl_rx_buf[client].count; 241 242 cl_rx_buf[client].count = 0; 243 } 244 245 /* 246 * marks a packet down to be sent to a particular client process 247 */ 248 static inline void 249 enqueue_rx_packet(uint8_t client, struct rte_mbuf *buf) 250 { 251 cl_rx_buf[client].buffer[cl_rx_buf[client].count++] = buf; 252 } 253 254 /* 255 * This function takes a group of packets and routes them 256 * individually to the client process. Very simply round-robins the packets 257 * without checking any of the packet contents. 258 */ 259 static void 260 process_packets(uint32_t port_num __rte_unused, 261 struct rte_mbuf *pkts[], uint16_t rx_count) 262 { 263 uint16_t i; 264 uint8_t client = 0; 265 266 for (i = 0; i < rx_count; i++) { 267 enqueue_rx_packet(client, pkts[i]); 268 269 if (++client == num_clients) 270 client = 0; 271 } 272 273 for (i = 0; i < num_clients; i++) 274 flush_rx_queue(i); 275 } 276 277 /* 278 * Function called by the master lcore of the DPDK process. 279 */ 280 static void 281 do_packet_forwarding(void) 282 { 283 unsigned port_num = 0; /* indexes the port[] array */ 284 285 for (;;) { 286 struct rte_mbuf *buf[PACKET_READ_SIZE]; 287 uint16_t rx_count; 288 289 /* read a port */ 290 rx_count = rte_eth_rx_burst(ports->id[port_num], 0, \ 291 buf, PACKET_READ_SIZE); 292 ports->rx_stats.rx[port_num] += rx_count; 293 294 /* Now process the NIC packets read */ 295 if (likely(rx_count > 0)) 296 process_packets(port_num, buf, rx_count); 297 298 /* move to next port */ 299 if (++port_num == ports->num_ports) 300 port_num = 0; 301 } 302 } 303 304 int 305 MAIN(int argc, char *argv[]) 306 { 307 /* initialise the system */ 308 if (init(argc, argv) < 0 ) 309 return -1; 310 RTE_LOG(INFO, APP, "Finished Process Init.\n"); 311 312 cl_rx_buf = calloc(num_clients, sizeof(cl_rx_buf[0])); 313 314 /* clear statistics */ 315 clear_stats(); 316 317 /* put all other cores to sleep bar master */ 318 rte_eal_mp_remote_launch(sleep_lcore, NULL, SKIP_MASTER); 319 320 do_packet_forwarding(); 321 return 0; 322 } 323