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