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 <string.h> 37 #include <sys/queue.h> 38 #include <errno.h> 39 #include <stdarg.h> 40 #include <inttypes.h> 41 42 #include <rte_common.h> 43 #include <rte_memory.h> 44 #include <rte_memzone.h> 45 #include <rte_eal.h> 46 #include <rte_byteorder.h> 47 #include <rte_atomic.h> 48 #include <rte_launch.h> 49 #include <rte_per_lcore.h> 50 #include <rte_lcore.h> 51 #include <rte_branch_prediction.h> 52 #include <rte_debug.h> 53 #include <rte_ring.h> 54 #include <rte_log.h> 55 #include <rte_mempool.h> 56 #include <rte_memcpy.h> 57 #include <rte_mbuf.h> 58 #include <rte_interrupts.h> 59 #include <rte_ether.h> 60 #include <rte_ethdev.h> 61 #include <rte_malloc.h> 62 #include <rte_string_fns.h> 63 #include <rte_cycles.h> 64 #include <rte_efd.h> 65 #include <rte_hash.h> 66 67 #include "common.h" 68 #include "args.h" 69 #include "init.h" 70 71 #define MBUFS_PER_NODE 1536 72 #define MBUFS_PER_PORT 1536 73 #define MBUF_CACHE_SIZE 512 74 75 #define RTE_MP_RX_DESC_DEFAULT 512 76 #define RTE_MP_TX_DESC_DEFAULT 512 77 #define NODE_QUEUE_RINGSIZE 128 78 79 #define NO_FLAGS 0 80 81 /* The mbuf pool for packet rx */ 82 struct rte_mempool *pktmbuf_pool; 83 84 /* array of info/queues for nodes */ 85 struct node *nodes; 86 87 /* EFD table */ 88 struct rte_efd_table *efd_table; 89 90 /* Shared info between server and nodes */ 91 struct shared_info *info; 92 93 /** 94 * Initialise the mbuf pool for packet reception for the NIC, and any other 95 * buffer pools needed by the app - currently none. 96 */ 97 static int 98 init_mbuf_pools(void) 99 { 100 const unsigned int num_mbufs = (num_nodes * MBUFS_PER_NODE) + 101 (info->num_ports * MBUFS_PER_PORT); 102 103 /* 104 * Don't pass single-producer/single-consumer flags to mbuf create as it 105 * seems faster to use a cache instead 106 */ 107 printf("Creating mbuf pool '%s' [%u mbufs] ...\n", 108 PKTMBUF_POOL_NAME, num_mbufs); 109 pktmbuf_pool = rte_pktmbuf_pool_create(PKTMBUF_POOL_NAME, num_mbufs, 110 MBUF_CACHE_SIZE, 0, RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id()); 111 112 return pktmbuf_pool == NULL; /* 0 on success */ 113 } 114 115 /** 116 * Initialise an individual port: 117 * - configure number of rx and tx rings 118 * - set up each rx ring, to pull from the main mbuf pool 119 * - set up each tx ring 120 * - start the port and report its status to stdout 121 */ 122 static int 123 init_port(uint16_t port_num) 124 { 125 /* for port configuration all features are off by default */ 126 const struct rte_eth_conf port_conf = { 127 .rxmode = { 128 .mq_mode = ETH_MQ_RX_RSS 129 } 130 }; 131 const uint16_t rx_rings = 1, tx_rings = num_nodes; 132 uint16_t rx_ring_size = RTE_MP_RX_DESC_DEFAULT; 133 uint16_t tx_ring_size = RTE_MP_TX_DESC_DEFAULT; 134 135 uint16_t q; 136 int retval; 137 138 printf("Port %u init ... ", port_num); 139 fflush(stdout); 140 141 /* 142 * Standard DPDK port initialisation - config port, then set up 143 * rx and tx rings. 144 */ 145 retval = rte_eth_dev_configure(port_num, rx_rings, tx_rings, &port_conf); 146 if (retval != 0) 147 return retval; 148 149 retval = rte_eth_dev_adjust_nb_rx_tx_desc(port_num, &rx_ring_size, 150 &tx_ring_size); 151 if (retval != 0) 152 return retval; 153 154 for (q = 0; q < rx_rings; q++) { 155 retval = rte_eth_rx_queue_setup(port_num, q, rx_ring_size, 156 rte_eth_dev_socket_id(port_num), 157 NULL, pktmbuf_pool); 158 if (retval < 0) 159 return retval; 160 } 161 162 for (q = 0; q < tx_rings; q++) { 163 retval = rte_eth_tx_queue_setup(port_num, q, tx_ring_size, 164 rte_eth_dev_socket_id(port_num), 165 NULL); 166 if (retval < 0) 167 return retval; 168 } 169 170 rte_eth_promiscuous_enable(port_num); 171 172 retval = rte_eth_dev_start(port_num); 173 if (retval < 0) 174 return retval; 175 176 printf("done:\n"); 177 178 return 0; 179 } 180 181 /** 182 * Set up the DPDK rings which will be used to pass packets, via 183 * pointers, between the multi-process server and node processes. 184 * Each node needs one RX queue. 185 */ 186 static int 187 init_shm_rings(void) 188 { 189 unsigned int i; 190 unsigned int socket_id; 191 const char *q_name; 192 const unsigned int ringsize = NODE_QUEUE_RINGSIZE; 193 194 nodes = rte_malloc("node details", 195 sizeof(*nodes) * num_nodes, 0); 196 if (nodes == NULL) 197 rte_exit(EXIT_FAILURE, "Cannot allocate memory for " 198 "node program details\n"); 199 200 for (i = 0; i < num_nodes; i++) { 201 /* Create an RX queue for each node */ 202 socket_id = rte_socket_id(); 203 q_name = get_rx_queue_name(i); 204 nodes[i].rx_q = rte_ring_create(q_name, 205 ringsize, socket_id, 206 RING_F_SP_ENQ | RING_F_SC_DEQ); 207 if (nodes[i].rx_q == NULL) 208 rte_exit(EXIT_FAILURE, "Cannot create rx ring queue " 209 "for node %u\n", i); 210 } 211 return 0; 212 } 213 214 /* 215 * Create EFD table which will contain all the flows 216 * that will be distributed among the nodes 217 */ 218 static void 219 create_efd_table(void) 220 { 221 uint8_t socket_id = rte_socket_id(); 222 223 /* create table */ 224 efd_table = rte_efd_create("flow table", num_flows * 2, sizeof(uint32_t), 225 1 << socket_id, socket_id); 226 227 if (efd_table == NULL) 228 rte_exit(EXIT_FAILURE, "Problem creating the flow table\n"); 229 } 230 231 static void 232 populate_efd_table(void) 233 { 234 unsigned int i; 235 int32_t ret; 236 uint32_t ip_dst; 237 uint8_t socket_id = rte_socket_id(); 238 uint64_t node_id; 239 240 /* Add flows in table */ 241 for (i = 0; i < num_flows; i++) { 242 node_id = i % num_nodes; 243 244 ip_dst = rte_cpu_to_be_32(i); 245 ret = rte_efd_update(efd_table, socket_id, 246 (void *)&ip_dst, (efd_value_t)node_id); 247 if (ret < 0) 248 rte_exit(EXIT_FAILURE, "Unable to add entry %u in " 249 "EFD table\n", i); 250 } 251 252 printf("EFD table: Adding 0x%x keys\n", num_flows); 253 } 254 255 /* Check the link status of all ports in up to 9s, and print them finally */ 256 static void 257 check_all_ports_link_status(uint16_t port_num, uint32_t port_mask) 258 { 259 #define CHECK_INTERVAL 100 /* 100ms */ 260 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */ 261 uint8_t count, all_ports_up, print_flag = 0; 262 uint16_t portid; 263 struct rte_eth_link link; 264 265 printf("\nChecking link status"); 266 fflush(stdout); 267 for (count = 0; count <= MAX_CHECK_TIME; count++) { 268 all_ports_up = 1; 269 for (portid = 0; portid < port_num; portid++) { 270 if ((port_mask & (1 << info->id[portid])) == 0) 271 continue; 272 memset(&link, 0, sizeof(link)); 273 rte_eth_link_get_nowait(info->id[portid], &link); 274 /* print link status if flag set */ 275 if (print_flag == 1) { 276 if (link.link_status) 277 printf( 278 "Port%d Link Up. Speed %u Mbps - %s\n", 279 info->id[portid], 280 link.link_speed, 281 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ? 282 ("full-duplex") : ("half-duplex\n")); 283 else 284 printf("Port %d Link Down\n", 285 info->id[portid]); 286 continue; 287 } 288 /* clear all_ports_up flag if any link down */ 289 if (link.link_status == ETH_LINK_DOWN) { 290 all_ports_up = 0; 291 break; 292 } 293 } 294 /* after finally printing all link status, get out */ 295 if (print_flag == 1) 296 break; 297 298 if (all_ports_up == 0) { 299 printf("."); 300 fflush(stdout); 301 rte_delay_ms(CHECK_INTERVAL); 302 } 303 304 /* set the print_flag if all ports up or timeout */ 305 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) { 306 print_flag = 1; 307 printf("done\n"); 308 } 309 } 310 } 311 312 /** 313 * Main init function for the multi-process server app, 314 * calls subfunctions to do each stage of the initialisation. 315 */ 316 int 317 init(int argc, char *argv[]) 318 { 319 int retval; 320 const struct rte_memzone *mz; 321 uint8_t i, total_ports; 322 323 /* init EAL, parsing EAL args */ 324 retval = rte_eal_init(argc, argv); 325 if (retval < 0) 326 return -1; 327 argc -= retval; 328 argv += retval; 329 330 /* get total number of ports */ 331 total_ports = rte_eth_dev_count(); 332 333 /* set up array for port data */ 334 mz = rte_memzone_reserve(MZ_SHARED_INFO, sizeof(*info), 335 rte_socket_id(), NO_FLAGS); 336 if (mz == NULL) 337 rte_exit(EXIT_FAILURE, "Cannot reserve memory zone " 338 "for port information\n"); 339 memset(mz->addr, 0, sizeof(*info)); 340 info = mz->addr; 341 342 /* parse additional, application arguments */ 343 retval = parse_app_args(total_ports, argc, argv); 344 if (retval != 0) 345 return -1; 346 347 /* initialise mbuf pools */ 348 retval = init_mbuf_pools(); 349 if (retval != 0) 350 rte_exit(EXIT_FAILURE, "Cannot create needed mbuf pools\n"); 351 352 /* now initialise the ports we will use */ 353 for (i = 0; i < info->num_ports; i++) { 354 retval = init_port(info->id[i]); 355 if (retval != 0) 356 rte_exit(EXIT_FAILURE, "Cannot initialise port %u\n", 357 (unsigned int) i); 358 } 359 360 check_all_ports_link_status(info->num_ports, (~0x0)); 361 362 /* initialise the node queues/rings for inter-eu comms */ 363 init_shm_rings(); 364 365 /* Create the EFD table */ 366 create_efd_table(); 367 368 /* Populate the EFD table */ 369 populate_efd_table(); 370 371 /* Share the total number of nodes */ 372 info->num_nodes = num_nodes; 373 374 /* Share the total number of flows */ 375 info->num_flows = num_flows; 376 return 0; 377 } 378