1 /*- 2 * BSD LICENSE 3 * 4 * Copyright(c) 2010-2014 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_tailq.h> 46 #include <rte_eal.h> 47 #include <rte_byteorder.h> 48 #include <rte_atomic.h> 49 #include <rte_launch.h> 50 #include <rte_per_lcore.h> 51 #include <rte_lcore.h> 52 #include <rte_branch_prediction.h> 53 #include <rte_debug.h> 54 #include <rte_ring.h> 55 #include <rte_log.h> 56 #include <rte_mempool.h> 57 #include <rte_memcpy.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_malloc.h> 64 #include <rte_fbk_hash.h> 65 #include <rte_string_fns.h> 66 #include <rte_cycles.h> 67 68 #include "common.h" 69 #include "init_drivers.h" 70 #include "args.h" 71 #include "init.h" 72 #include "main.h" 73 74 #define MBUFS_PER_CLIENT 1536 75 #define MBUFS_PER_PORT 1536 76 #define MBUF_CACHE_SIZE 512 77 #define MBUF_OVERHEAD (sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM) 78 #define RX_MBUF_DATA_SIZE 2048 79 #define MBUF_SIZE (RX_MBUF_DATA_SIZE + MBUF_OVERHEAD) 80 81 #define RTE_MP_RX_DESC_DEFAULT 512 82 #define RTE_MP_TX_DESC_DEFAULT 512 83 #define CLIENT_QUEUE_RINGSIZE 128 84 85 #define NO_FLAGS 0 86 87 /* 88 * RX and TX Prefetch, Host, and Write-back threshold values should be 89 * carefully set for optimal performance. Consult the network 90 * controller's datasheet and supporting DPDK documentation for guidance 91 * on how these parameters should be set. 92 */ 93 /* Default configuration for rx and tx thresholds etc. */ 94 /* 95 * These default values are optimized for use with the Intel(R) 82599 10 GbE 96 * Controller and the DPDK ixgbe PMD. Consider using other values for other 97 * network controllers and/or network drivers. 98 */ 99 #define MP_DEFAULT_PTHRESH 36 100 #define MP_DEFAULT_RX_HTHRESH 8 101 #define MP_DEFAULT_TX_HTHRESH 0 102 #define MP_DEFAULT_WTHRESH 0 103 104 static const struct rte_eth_rxconf rx_conf_default = { 105 .rx_thresh = { 106 .pthresh = MP_DEFAULT_PTHRESH, 107 .hthresh = MP_DEFAULT_RX_HTHRESH, 108 .wthresh = MP_DEFAULT_WTHRESH, 109 }, 110 }; 111 112 static const struct rte_eth_txconf tx_conf_default = { 113 .tx_thresh = { 114 .pthresh = MP_DEFAULT_PTHRESH, 115 .hthresh = MP_DEFAULT_TX_HTHRESH, 116 .wthresh = MP_DEFAULT_WTHRESH, 117 }, 118 .tx_free_thresh = 0, /* Use PMD default values */ 119 .tx_rs_thresh = 0, /* Use PMD default values */ 120 }; 121 122 /* The mbuf pool for packet rx */ 123 struct rte_mempool *pktmbuf_pool; 124 125 /* array of info/queues for clients */ 126 struct client *clients = NULL; 127 128 /* the port details */ 129 struct port_info *ports; 130 131 /** 132 * Initialise the mbuf pool for packet reception for the NIC, and any other 133 * buffer pools needed by the app - currently none. 134 */ 135 static int 136 init_mbuf_pools(void) 137 { 138 const unsigned num_mbufs = (num_clients * MBUFS_PER_CLIENT) \ 139 + (ports->num_ports * MBUFS_PER_PORT); 140 141 /* don't pass single-producer/single-consumer flags to mbuf create as it 142 * seems faster to use a cache instead */ 143 printf("Creating mbuf pool '%s' [%u mbufs] ...\n", 144 PKTMBUF_POOL_NAME, num_mbufs); 145 pktmbuf_pool = rte_mempool_create(PKTMBUF_POOL_NAME, num_mbufs, 146 MBUF_SIZE, MBUF_CACHE_SIZE, 147 sizeof(struct rte_pktmbuf_pool_private), rte_pktmbuf_pool_init, 148 NULL, rte_pktmbuf_init, NULL, rte_socket_id(), NO_FLAGS ); 149 150 return (pktmbuf_pool == NULL); /* 0 on success */ 151 } 152 153 /** 154 * Initialise an individual port: 155 * - configure number of rx and tx rings 156 * - set up each rx ring, to pull from the main mbuf pool 157 * - set up each tx ring 158 * - start the port and report its status to stdout 159 */ 160 static int 161 init_port(uint8_t port_num) 162 { 163 /* for port configuration all features are off by default */ 164 const struct rte_eth_conf port_conf = { 165 .rxmode = { 166 .mq_mode = ETH_MQ_RX_RSS 167 } 168 }; 169 const uint16_t rx_rings = 1, tx_rings = num_clients; 170 const uint16_t rx_ring_size = RTE_MP_RX_DESC_DEFAULT; 171 const uint16_t tx_ring_size = RTE_MP_TX_DESC_DEFAULT; 172 173 uint16_t q; 174 int retval; 175 176 printf("Port %u init ... ", (unsigned)port_num); 177 fflush(stdout); 178 179 /* Standard DPDK port initialisation - config port, then set up 180 * rx and tx rings */ 181 if ((retval = rte_eth_dev_configure(port_num, rx_rings, tx_rings, 182 &port_conf)) != 0) 183 return retval; 184 185 for (q = 0; q < rx_rings; q++) { 186 retval = rte_eth_rx_queue_setup(port_num, q, rx_ring_size, 187 rte_eth_dev_socket_id(port_num), &rx_conf_default, pktmbuf_pool); 188 if (retval < 0) return retval; 189 } 190 191 for ( q = 0; q < tx_rings; q ++ ) { 192 retval = rte_eth_tx_queue_setup(port_num, q, tx_ring_size, 193 rte_eth_dev_socket_id(port_num), &tx_conf_default); 194 if (retval < 0) return retval; 195 } 196 197 rte_eth_promiscuous_enable(port_num); 198 199 retval = rte_eth_dev_start(port_num); 200 if (retval < 0) return retval; 201 202 printf( "done: \n"); 203 204 return 0; 205 } 206 207 /** 208 * Set up the DPDK rings which will be used to pass packets, via 209 * pointers, between the multi-process server and client processes. 210 * Each client needs one RX queue. 211 */ 212 static int 213 init_shm_rings(void) 214 { 215 unsigned i; 216 unsigned socket_id; 217 const char * q_name; 218 const unsigned ringsize = CLIENT_QUEUE_RINGSIZE; 219 220 clients = rte_malloc("client details", 221 sizeof(*clients) * num_clients, 0); 222 if (clients == NULL) 223 rte_exit(EXIT_FAILURE, "Cannot allocate memory for client program details\n"); 224 225 for (i = 0; i < num_clients; i++) { 226 /* Create an RX queue for each client */ 227 socket_id = rte_socket_id(); 228 q_name = get_rx_queue_name(i); 229 clients[i].rx_q = rte_ring_create(q_name, 230 ringsize, socket_id, 231 RING_F_SP_ENQ | RING_F_SC_DEQ ); /* single prod, single cons */ 232 if (clients[i].rx_q == NULL) 233 rte_exit(EXIT_FAILURE, "Cannot create rx ring queue for client %u\n", i); 234 } 235 return 0; 236 } 237 238 /* Check the link status of all ports in up to 9s, and print them finally */ 239 static void 240 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask) 241 { 242 #define CHECK_INTERVAL 100 /* 100ms */ 243 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */ 244 uint8_t portid, count, all_ports_up, print_flag = 0; 245 struct rte_eth_link link; 246 247 printf("\nChecking link status"); 248 fflush(stdout); 249 for (count = 0; count <= MAX_CHECK_TIME; count++) { 250 all_ports_up = 1; 251 for (portid = 0; portid < port_num; portid++) { 252 if ((port_mask & (1 << ports->id[portid])) == 0) 253 continue; 254 memset(&link, 0, sizeof(link)); 255 rte_eth_link_get_nowait(ports->id[portid], &link); 256 /* print link status if flag set */ 257 if (print_flag == 1) { 258 if (link.link_status) 259 printf("Port %d Link Up - speed %u " 260 "Mbps - %s\n", ports->id[portid], 261 (unsigned)link.link_speed, 262 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ? 263 ("full-duplex") : ("half-duplex\n")); 264 else 265 printf("Port %d Link Down\n", 266 (uint8_t)ports->id[portid]); 267 continue; 268 } 269 /* clear all_ports_up flag if any link down */ 270 if (link.link_status == 0) { 271 all_ports_up = 0; 272 break; 273 } 274 } 275 /* after finally printing all link status, get out */ 276 if (print_flag == 1) 277 break; 278 279 if (all_ports_up == 0) { 280 printf("."); 281 fflush(stdout); 282 rte_delay_ms(CHECK_INTERVAL); 283 } 284 285 /* set the print_flag if all ports up or timeout */ 286 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) { 287 print_flag = 1; 288 printf("done\n"); 289 } 290 } 291 } 292 293 /** 294 * Main init function for the multi-process server app, 295 * calls subfunctions to do each stage of the initialisation. 296 */ 297 int 298 init(int argc, char *argv[]) 299 { 300 int retval; 301 const struct rte_memzone *mz; 302 uint8_t i, total_ports; 303 304 /* init EAL, parsing EAL args */ 305 retval = rte_eal_init(argc, argv); 306 if (retval < 0) 307 return -1; 308 argc -= retval; 309 argv += retval; 310 311 /* initialise the nic drivers */ 312 retval = init_drivers(); 313 if (retval != 0) 314 rte_exit(EXIT_FAILURE, "Cannot initialise drivers\n"); 315 316 /* get total number of ports */ 317 total_ports = rte_eth_dev_count(); 318 319 /* set up array for port data */ 320 mz = rte_memzone_reserve(MZ_PORT_INFO, sizeof(*ports), 321 rte_socket_id(), NO_FLAGS); 322 if (mz == NULL) 323 rte_exit(EXIT_FAILURE, "Cannot reserve memory zone for port information\n"); 324 memset(mz->addr, 0, sizeof(*ports)); 325 ports = mz->addr; 326 327 /* parse additional, application arguments */ 328 retval = parse_app_args(total_ports, argc, argv); 329 if (retval != 0) 330 return -1; 331 332 /* initialise mbuf pools */ 333 retval = init_mbuf_pools(); 334 if (retval != 0) 335 rte_exit(EXIT_FAILURE, "Cannot create needed mbuf pools\n"); 336 337 /* now initialise the ports we will use */ 338 for (i = 0; i < ports->num_ports; i++) { 339 retval = init_port(ports->id[i]); 340 if (retval != 0) 341 rte_exit(EXIT_FAILURE, "Cannot initialise port %u\n", 342 (unsigned)i); 343 } 344 345 check_all_ports_link_status(ports->num_ports, (~0x0)); 346 347 /* initialise the client queues/rings for inter-eu comms */ 348 init_shm_rings(); 349 350 return 0; 351 } 352