1 /*- 2 * BSD LICENSE 3 * 4 * Copyright(c) 2016-2017 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 <stdint.h> 35 #include <stdio.h> 36 #include <inttypes.h> 37 #include <stdarg.h> 38 #include <errno.h> 39 #include <sys/queue.h> 40 #include <stdlib.h> 41 #include <getopt.h> 42 #include <string.h> 43 44 #include <rte_common.h> 45 #include <rte_malloc.h> 46 #include <rte_memory.h> 47 #include <rte_memzone.h> 48 #include <rte_eal.h> 49 #include <rte_atomic.h> 50 #include <rte_branch_prediction.h> 51 #include <rte_log.h> 52 #include <rte_per_lcore.h> 53 #include <rte_launch.h> 54 #include <rte_lcore.h> 55 #include <rte_ring.h> 56 #include <rte_launch.h> 57 #include <rte_lcore.h> 58 #include <rte_debug.h> 59 #include <rte_mempool.h> 60 #include <rte_mbuf.h> 61 #include <rte_interrupts.h> 62 #include <rte_pci.h> 63 #include <rte_ether.h> 64 #include <rte_ethdev.h> 65 #include <rte_string_fns.h> 66 #include <rte_ip.h> 67 68 #include "common.h" 69 70 /* Number of packets to attempt to read from queue */ 71 #define PKT_READ_SIZE ((uint16_t)32) 72 73 /* 74 * Our node id number - tells us which rx queue to read, and NIC TX 75 * queue to write to. 76 */ 77 static uint8_t node_id; 78 79 #define MBQ_CAPACITY 32 80 81 /* maps input ports to output ports for packets */ 82 static uint8_t output_ports[RTE_MAX_ETHPORTS]; 83 84 /* buffers up a set of packet that are ready to send */ 85 struct rte_eth_dev_tx_buffer *tx_buffer[RTE_MAX_ETHPORTS]; 86 87 /* shared data from server. We update statistics here */ 88 static struct tx_stats *tx_stats; 89 90 static struct filter_stats *filter_stats; 91 92 /* 93 * print a usage message 94 */ 95 static void 96 usage(const char *progname) 97 { 98 printf("Usage: %s [EAL args] -- -n <node_id>\n\n", progname); 99 } 100 101 /* 102 * Convert the node id number from a string to an int. 103 */ 104 static int 105 parse_node_num(const char *node) 106 { 107 char *end = NULL; 108 unsigned long temp; 109 110 if (node == NULL || *node == '\0') 111 return -1; 112 113 temp = strtoul(node, &end, 10); 114 if (end == NULL || *end != '\0') 115 return -1; 116 117 node_id = (uint8_t)temp; 118 return 0; 119 } 120 121 /* 122 * Parse the application arguments to the node app. 123 */ 124 static int 125 parse_app_args(int argc, char *argv[]) 126 { 127 int option_index, opt; 128 char **argvopt = argv; 129 const char *progname = NULL; 130 static struct option lgopts[] = { /* no long options */ 131 {NULL, 0, 0, 0 } 132 }; 133 progname = argv[0]; 134 135 while ((opt = getopt_long(argc, argvopt, "n:", lgopts, 136 &option_index)) != EOF) { 137 switch (opt) { 138 case 'n': 139 if (parse_node_num(optarg) != 0) { 140 usage(progname); 141 return -1; 142 } 143 break; 144 default: 145 usage(progname); 146 return -1; 147 } 148 } 149 return 0; 150 } 151 152 /* 153 * Tx buffer error callback 154 */ 155 static void 156 flush_tx_error_callback(struct rte_mbuf **unsent, uint16_t count, 157 void *userdata) { 158 int i; 159 uint8_t port_id = (uintptr_t)userdata; 160 161 tx_stats->tx_drop[port_id] += count; 162 163 /* free the mbufs which failed from transmit */ 164 for (i = 0; i < count; i++) 165 rte_pktmbuf_free(unsent[i]); 166 167 } 168 169 static void 170 configure_tx_buffer(uint8_t port_id, uint16_t size) 171 { 172 int ret; 173 174 /* Initialize TX buffers */ 175 tx_buffer[port_id] = rte_zmalloc_socket("tx_buffer", 176 RTE_ETH_TX_BUFFER_SIZE(size), 0, 177 rte_eth_dev_socket_id(port_id)); 178 if (tx_buffer[port_id] == NULL) 179 rte_exit(EXIT_FAILURE, "Cannot allocate buffer for tx " 180 "on port %u\n", (unsigned int) port_id); 181 182 rte_eth_tx_buffer_init(tx_buffer[port_id], size); 183 184 ret = rte_eth_tx_buffer_set_err_callback(tx_buffer[port_id], 185 flush_tx_error_callback, (void *)(intptr_t)port_id); 186 if (ret < 0) 187 rte_exit(EXIT_FAILURE, "Cannot set error callback for " 188 "tx buffer on port %u\n", (unsigned int) port_id); 189 } 190 191 /* 192 * set up output ports so that all traffic on port gets sent out 193 * its paired port. Index using actual port numbers since that is 194 * what comes in the mbuf structure. 195 */ 196 static void 197 configure_output_ports(const struct shared_info *info) 198 { 199 int i; 200 201 if (info->num_ports > RTE_MAX_ETHPORTS) 202 rte_exit(EXIT_FAILURE, "Too many ethernet ports. " 203 "RTE_MAX_ETHPORTS = %u\n", 204 (unsigned int)RTE_MAX_ETHPORTS); 205 for (i = 0; i < info->num_ports - 1; i += 2) { 206 uint8_t p1 = info->id[i]; 207 uint8_t p2 = info->id[i+1]; 208 209 output_ports[p1] = p2; 210 output_ports[p2] = p1; 211 212 configure_tx_buffer(p1, MBQ_CAPACITY); 213 configure_tx_buffer(p2, MBQ_CAPACITY); 214 215 } 216 } 217 218 /* 219 * Create the hash table that will contain the flows that 220 * the node will handle, which will be used to decide if packet 221 * is transmitted or dropped. 222 */ 223 static struct rte_hash * 224 create_hash_table(const struct shared_info *info) 225 { 226 uint32_t num_flows_node = info->num_flows / info->num_nodes; 227 char name[RTE_HASH_NAMESIZE]; 228 struct rte_hash *h; 229 230 /* create table */ 231 struct rte_hash_parameters hash_params = { 232 .entries = num_flows_node * 2, /* table load = 50% */ 233 .key_len = sizeof(uint32_t), /* Store IPv4 dest IP address */ 234 .socket_id = rte_socket_id(), 235 .hash_func_init_val = 0, 236 }; 237 238 snprintf(name, sizeof(name), "hash_table_%d", node_id); 239 hash_params.name = name; 240 h = rte_hash_create(&hash_params); 241 242 if (h == NULL) 243 rte_exit(EXIT_FAILURE, 244 "Problem creating the hash table for node %d\n", 245 node_id); 246 return h; 247 } 248 249 static void 250 populate_hash_table(const struct rte_hash *h, const struct shared_info *info) 251 { 252 unsigned int i; 253 int32_t ret; 254 uint32_t ip_dst; 255 uint32_t num_flows_node = 0; 256 uint64_t target_node; 257 258 /* Add flows in table */ 259 for (i = 0; i < info->num_flows; i++) { 260 target_node = i % info->num_nodes; 261 if (target_node != node_id) 262 continue; 263 264 ip_dst = rte_cpu_to_be_32(i); 265 266 ret = rte_hash_add_key(h, (void *) &ip_dst); 267 if (ret < 0) 268 rte_exit(EXIT_FAILURE, "Unable to add entry %u " 269 "in hash table\n", i); 270 else 271 num_flows_node++; 272 273 } 274 275 printf("Hash table: Adding 0x%x keys\n", num_flows_node); 276 } 277 278 /* 279 * This function performs routing of packets 280 * Just sends each input packet out an output port based solely on the input 281 * port it arrived on. 282 */ 283 static inline void 284 transmit_packet(struct rte_mbuf *buf) 285 { 286 int sent; 287 const uint8_t in_port = buf->port; 288 const uint8_t out_port = output_ports[in_port]; 289 struct rte_eth_dev_tx_buffer *buffer = tx_buffer[out_port]; 290 291 sent = rte_eth_tx_buffer(out_port, node_id, buffer, buf); 292 if (sent) 293 tx_stats->tx[out_port] += sent; 294 295 } 296 297 static inline void 298 handle_packets(struct rte_hash *h, struct rte_mbuf **bufs, uint16_t num_packets) 299 { 300 struct ipv4_hdr *ipv4_hdr; 301 uint32_t ipv4_dst_ip[PKT_READ_SIZE]; 302 const void *key_ptrs[PKT_READ_SIZE]; 303 unsigned int i; 304 int32_t positions[PKT_READ_SIZE] = {0}; 305 306 for (i = 0; i < num_packets; i++) { 307 /* Handle IPv4 header.*/ 308 ipv4_hdr = rte_pktmbuf_mtod_offset(bufs[i], struct ipv4_hdr *, 309 sizeof(struct ether_hdr)); 310 ipv4_dst_ip[i] = ipv4_hdr->dst_addr; 311 key_ptrs[i] = &ipv4_dst_ip[i]; 312 } 313 /* Check if packets belongs to any flows handled by this node */ 314 rte_hash_lookup_bulk(h, key_ptrs, num_packets, positions); 315 316 for (i = 0; i < num_packets; i++) { 317 if (likely(positions[i] >= 0)) { 318 filter_stats->passed++; 319 transmit_packet(bufs[i]); 320 } else { 321 filter_stats->drop++; 322 /* Drop packet, as flow is not handled by this node */ 323 rte_pktmbuf_free(bufs[i]); 324 } 325 } 326 } 327 328 /* 329 * Application main function - loops through 330 * receiving and processing packets. Never returns 331 */ 332 int 333 main(int argc, char *argv[]) 334 { 335 const struct rte_memzone *mz; 336 struct rte_ring *rx_ring; 337 struct rte_hash *h; 338 struct rte_mempool *mp; 339 struct shared_info *info; 340 int need_flush = 0; /* indicates whether we have unsent packets */ 341 int retval; 342 void *pkts[PKT_READ_SIZE]; 343 uint16_t sent; 344 345 retval = rte_eal_init(argc, argv); 346 if (retval < 0) 347 return -1; 348 argc -= retval; 349 argv += retval; 350 351 if (parse_app_args(argc, argv) < 0) 352 rte_exit(EXIT_FAILURE, "Invalid command-line arguments\n"); 353 354 if (rte_eth_dev_count() == 0) 355 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n"); 356 357 rx_ring = rte_ring_lookup(get_rx_queue_name(node_id)); 358 if (rx_ring == NULL) 359 rte_exit(EXIT_FAILURE, "Cannot get RX ring - " 360 "is server process running?\n"); 361 362 mp = rte_mempool_lookup(PKTMBUF_POOL_NAME); 363 if (mp == NULL) 364 rte_exit(EXIT_FAILURE, "Cannot get mempool for mbufs\n"); 365 366 mz = rte_memzone_lookup(MZ_SHARED_INFO); 367 if (mz == NULL) 368 rte_exit(EXIT_FAILURE, "Cannot get port info structure\n"); 369 info = mz->addr; 370 tx_stats = &(info->tx_stats[node_id]); 371 filter_stats = &(info->filter_stats[node_id]); 372 373 configure_output_ports(info); 374 375 h = create_hash_table(info); 376 377 populate_hash_table(h, info); 378 379 RTE_LOG(INFO, APP, "Finished Process Init.\n"); 380 381 printf("\nNode process %d handling packets\n", node_id); 382 printf("[Press Ctrl-C to quit ...]\n"); 383 384 for (;;) { 385 uint16_t rx_pkts = PKT_READ_SIZE; 386 uint8_t port; 387 388 /* 389 * Try dequeuing max possible packets first, if that fails, 390 * get the most we can. Loop body should only execute once, 391 * maximum 392 */ 393 while (rx_pkts > 0 && 394 unlikely(rte_ring_dequeue_bulk(rx_ring, pkts, 395 rx_pkts) != 0)) 396 rx_pkts = (uint16_t)RTE_MIN(rte_ring_count(rx_ring), 397 PKT_READ_SIZE); 398 399 if (unlikely(rx_pkts == 0)) { 400 if (need_flush) 401 for (port = 0; port < info->num_ports; port++) { 402 sent = rte_eth_tx_buffer_flush( 403 info->id[port], 404 node_id, 405 tx_buffer[port]); 406 if (unlikely(sent)) 407 tx_stats->tx[port] += sent; 408 } 409 need_flush = 0; 410 continue; 411 } 412 413 handle_packets(h, (struct rte_mbuf **)pkts, rx_pkts); 414 415 need_flush = 1; 416 } 417 } 418