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