VNFs/DPPD-PROX/main.c

   1 /*
   2 // Copyright (c) 2010-2017 Intel Corporation
   3 //
   4 // Licensed under the Apache License, Version 2.0 (the "License");
   5 // you may not use this file except in compliance with the License.
   6 // You may obtain a copy of the License at
   7 //
   8 //     http://www.apache.org/licenses/LICENSE-2.0
   9 //
  10 // Unless required by applicable law or agreed to in writing, software
  11 // distributed under the License is distributed on an "AS IS" BASIS,
  12 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13 // See the License for the specific language governing permissions and
  14 // limitations under the License.
  15 */
  16
  17 #include <string.h>
  18 #include <locale.h>
  19 #include <unistd.h>
  20 #include <signal.h>
  21
  22 #include <rte_cycles.h>
  23 #include <rte_atomic.h>
  24 #include <rte_table_hash.h>
  25 #include <rte_memzone.h>
  26 #include <rte_errno.h>
  27
  28 #include "prox_malloc.h"
  29 #include "run.h"
  30 #include "main.h"
  31 #include "log.h"
  32 #include "quit.h"
  33 #include "clock.h"
  34 #include "defines.h"
  35 #include "version.h"
  36 #include "prox_args.h"
  37 #include "prox_assert.h"
  38 #include "prox_cfg.h"
  39 #include "prox_shared.h"
  40 #include "prox_port_cfg.h"
  41 #include "toeplitz.h"
  42 #include "hash_utils.h"
  43 #include "handle_lb_net.h"
  44 #include "prox_cksum.h"
  45 #include "thread_nop.h"
  46 #include "thread_generic.h"
  47 #include "thread_pipeline.h"
  48 #include "cqm.h"
  49 #include "handle_master.h"
  50
  51 #if RTE_VERSION < RTE_VERSION_NUM(1,8,0,0)
  52 #define RTE_CACHE_LINE_SIZE CACHE_LINE_SIZE
  53 #endif
  54
  55 uint8_t lb_nb_txrings = 0xff;
  56 struct rte_ring *ctrl_rings[RTE_MAX_LCORE*MAX_TASKS_PER_CORE];
  57
  58 static void __attribute__((noreturn)) prox_usage(const char *prgname)
  59 {
  60         plog_info("\nUsage: %s [-f CONFIG_FILE] [-a|-e] [-m|-s|-i] [-w DEF] [-u] [-t]\n"
  61                   "\t-f CONFIG_FILE : configuration file to load, ./prox.cfg by default\n"
  62                   "\t-l LOG_FILE : log file name, ./prox.log by default\n"
  63                   "\t-p : include PID in log file name if default log file is used\n"
  64                   "\t-o DISPLAY: Set display to use, can be 'curses' (default), 'cli' or 'none'\n"
  65                   "\t-v verbosity : initial logging verbosity\n"
  66                   "\t-a : autostart all cores (by default)\n"
  67                   "\t-e : don't autostart\n"
  68                   "\t-n : Create NULL devices instead of using PCI devices, useful together with -i\n"
  69                   "\t-m : list supported task modes and exit\n"
  70                   "\t-s : check configuration file syntax and exit\n"
  71                   "\t-i : check initialization sequence and exit\n"
  72                   "\t-u : Listen on UDS /tmp/prox.sock\n"
  73                   "\t-t : Listen on TCP port 8474\n"
  74                   "\t-q : Pass argument to Lua interpreter, useful to define variables\n"
  75                   "\t-w : define variable using syntax varname=value\n"
  76                   "\t     takes precedence over variables defined in CONFIG_FILE\n"
  77                   "\t-k : Log statistics to file \"stats_dump\" in current directory\n"
  78                   "\t-d : Run as daemon, the parent process will block until PROX is not initialized\n"
  79                   "\t-z : Ignore CPU topology, implies -i\n"
  80                   "\t-r : Change initial screen refresh rate. If set to a lower than 0.001 seconds,\n"
  81                   "\t     screen refreshing will be disabled\n"
  82                   , prgname);
  83         exit(EXIT_FAILURE);
  84 }
  85
  86 static void check_mixed_normal_pipeline(void)
  87 {
  88         struct lcore_cfg *lconf = NULL;
  89         uint32_t lcore_id = -1;
  90
  91         while (prox_core_next(&lcore_id, 0) == 0) {
  92                 lconf = &lcore_cfg[lcore_id];
  93
  94                 int all_thread_nop = 1;
  95                 int generic = 0;
  96                 int pipeline = 0;
  97                 for (uint8_t task_id = 0; task_id < lconf->n_tasks_all; ++task_id) {
  98                         struct task_args *targ = &lconf->targs[task_id];
  99                         all_thread_nop = all_thread_nop &&
 100                                 targ->task_init->thread_x == thread_nop;
 101
 102                         pipeline = pipeline || targ->task_init->thread_x == thread_pipeline;
 103                         generic = generic || targ->task_init->thread_x == thread_generic;
 104                 }
 105                 PROX_PANIC(generic && pipeline, "Can't run both pipeline and normal thread on same core\n");
 106
 107                 if (all_thread_nop)
 108                         lconf->thread_x = thread_nop;
 109                 else {
 110                         lconf->thread_x = thread_generic;
 111                 }
 112         }
 113 }
 114
 115 static void check_zero_rx(void)
 116 {
 117         struct lcore_cfg *lconf = NULL;
 118         struct task_args *targ;
 119
 120         while (core_targ_next(&lconf, &targ, 0) == 0) {
 121                 if (targ->nb_rxports != 0) {
 122                         PROX_PANIC(task_init_flag_set(targ->task_init, TASK_FEATURE_NO_RX),
 123                            "\tCore %u task %u: rx_ports configured while mode %s does not use it\n", lconf->id, targ->id, targ->task_init->mode_str);
 124                 }
 125         }
 126 }
 127
 128 static void check_missing_rx(void)
 129 {
 130         struct lcore_cfg *lconf = NULL, *rx_lconf = NULL;
 131         struct task_args *targ, *rx_targ = NULL;
 132         struct prox_port_cfg *port;
 133         uint8_t port_id, rx_port_id, ok;
 134
 135         while (core_targ_next(&lconf, &targ, 0) == 0) {
 136                 PROX_PANIC((targ->flags & TASK_ARG_RX_RING) && targ->rx_rings[0] == 0 && !targ->tx_opt_ring_task,
 137                            "Configuration Error - Core %u task %u Receiving from ring, but nobody xmitting to this ring\n", lconf->id, targ->id);
 138                 if (targ->nb_rxports == 0 && targ->nb_rxrings == 0) {
 139                         PROX_PANIC(!task_init_flag_set(targ->task_init, TASK_FEATURE_NO_RX),
 140                                    "\tCore %u task %u: no rx_ports and no rx_rings configured while required by mode %s\n", lconf->id, targ->id, targ->task_init->mode_str);
 141                 }
 142         }
 143
 144         lconf = NULL;
 145         while (core_targ_next(&lconf, &targ, 0) == 0) {
 146                 if (strcmp(targ->task_init->sub_mode_str, "l3") != 0)
 147                         continue;
 148                 port = find_reachable_port(targ);
 149                 if (port == NULL)
 150                         continue;
 151                 port_id = port - prox_port_cfg;
 152                 rx_lconf = NULL;
 153                 ok = 0;
 154                 plog_info("\tCore %d task %d transmitting to port %d in L3 mode\n", lconf->id, targ->id, port_id);
 155                 while (core_targ_next(&rx_lconf, &rx_targ, 0) == 0) {
 156                         for (uint8_t i = 0; i < rx_targ->nb_rxports; ++i) {
 157                                 rx_port_id = rx_targ->rx_port_queue[i].port;
 158                                 if ((rx_port_id == port_id) && (rx_targ->task_init->flag_features & TASK_FEATURE_L3)){
 159                                         ok = 1;
 160                                         break;
 161                                 }
 162                         }
 163                         if (ok == 1) {
 164                                 plog_info("\tCore %d task %d has found core %d task %d receiving from port %d\n", lconf->id, targ->id, rx_lconf->id, rx_targ->id, port_id);
 165                                 break;
 166                         }
 167                 }
 168                 PROX_PANIC(ok == 0, "L3 sub mode for port %d on core %d task %d, but no core/task receiving on that port\n", port_id, lconf->id, targ->id);
 169         }
 170 }
 171
 172 static void check_cfg_consistent(void)
 173 {
 174         check_missing_rx();
 175         check_zero_rx();
 176         check_mixed_normal_pipeline();
 177 }
 178
 179 static void plog_all_rings(void)
 180 {
 181         struct lcore_cfg *lconf = NULL;
 182         struct task_args *targ;
 183
 184         while (core_targ_next(&lconf, &targ, 0) == 0) {
 185                 for (uint8_t ring_idx = 0; ring_idx < targ->nb_rxrings; ++ring_idx) {
 186                         plog_info("\tCore %u, task %u, rx_ring[%u] %p\n", lconf->id, targ->id, ring_idx, targ->rx_rings[ring_idx]);
 187                 }
 188         }
 189 }
 190
 191 static int chain_flag_state(struct task_args *targ, uint64_t flag, int is_set)
 192 {
 193         if (task_init_flag_set(targ->task_init, flag) == is_set)
 194                 return 1;
 195
 196         int ret = 0;
 197
 198         for (uint32_t i = 0; i < targ->n_prev_tasks; ++i) {
 199                 ret = chain_flag_state(targ->prev_tasks[i], flag, is_set);
 200                 if (ret)
 201                         return 1;
 202         }
 203         return 0;
 204 }
 205
 206 static void configure_if_tx_queues(struct task_args *targ, uint8_t socket)
 207 {
 208         uint8_t if_port;
 209
 210         for (uint8_t i = 0; i < targ->nb_txports; ++i) {
 211                 if_port = targ->tx_port_queue[i].port;
 212
 213                 PROX_PANIC(if_port == OUT_DISCARD, "port misconfigured, exiting\n");
 214
 215                 PROX_PANIC(!prox_port_cfg[if_port].active, "\tPort %u not used, skipping...\n", if_port);
 216
 217                 int dsocket = prox_port_cfg[if_port].socket;
 218                 if (dsocket != -1 && dsocket != socket) {
 219                         plog_warn("TX core on socket %d while device on socket %d\n", socket, dsocket);
 220                 }
 221
 222                 if (prox_port_cfg[if_port].tx_ring[0] == '\0') {  // Rings-backed port can use single queue
 223                         targ->tx_port_queue[i].queue = prox_port_cfg[if_port].n_txq;
 224                         prox_port_cfg[if_port].n_txq++;
 225                 } else {
 226                         prox_port_cfg[if_port].n_txq = 1;
 227                         targ->tx_port_queue[i].queue = 0;
 228                 }
 229                 /* Set the ETH_TXQ_FLAGS_NOREFCOUNT flag if none of
 230                    the tasks up to the task transmitting to the port
 231                    does not use refcnt. */
 232                 if (!chain_flag_state(targ, TASK_FEATURE_TXQ_FLAGS_REFCOUNT, 1)) {
 233                         prox_port_cfg[if_port].tx_conf.txq_flags |= ETH_TXQ_FLAGS_NOREFCOUNT;
 234                         plog_info("\t\tEnabling No refcnt on port %d\n", if_port);
 235                 }
 236                 else {
 237                         plog_info("\t\tRefcnt used on port %d\n", if_port);
 238                 }
 239
 240                 /* By default OFFLOAD is enabled, but if the whole
 241                    chain has NOOFFLOADS set all the way until the
 242                    first task that receives from a port, it will be
 243                    disabled for the destination port. */
 244                 if (chain_flag_state(targ, TASK_FEATURE_TXQ_FLAGS_NOOFFLOADS, 1)) {
 245                         prox_port_cfg[if_port].tx_conf.txq_flags |= ETH_TXQ_FLAGS_NOOFFLOADS;
 246                         plog_info("\t\tDisabling TX offloads on port %d\n", if_port);
 247                 } else {
 248                         plog_info("\t\tEnabling TX offloads on port %d\n", if_port);
 249                 }
 250
 251                 /* By default NOMULTSEGS is disabled, as drivers/NIC might split packets on RX
 252                    It should only be enabled when we know for sure that the RX does not split packets.
 253                    Set the ETH_TXQ_FLAGS_NOMULTSEGS flag if none of the tasks up to the task
 254                    transmitting to the port does not use multsegs. */
 255                 if (!chain_flag_state(targ, TASK_FEATURE_TXQ_FLAGS_NOMULTSEGS, 0)) {
 256                         prox_port_cfg[if_port].tx_conf.txq_flags |= ETH_TXQ_FLAGS_NOMULTSEGS;
 257                         plog_info("\t\tEnabling No MultiSegs on port %d\n", if_port);
 258                 }
 259                 else {
 260                         plog_info("\t\tMultiSegs used on port %d\n", if_port);
 261                 }
 262         }
 263 }
 264
 265 static void configure_if_rx_queues(struct task_args *targ, uint8_t socket)
 266 {
 267         for (int i = 0; i < targ->nb_rxports; i++) {
 268                 uint8_t if_port = targ->rx_port_queue[i].port;
 269
 270                 if (if_port == OUT_DISCARD) {
 271                         return;
 272                 }
 273
 274                 PROX_PANIC(!prox_port_cfg[if_port].active, "Port %u not used, aborting...\n", if_port);
 275
 276                 if(prox_port_cfg[if_port].rx_ring[0] != '\0') {
 277                         prox_port_cfg[if_port].n_rxq = 0;
 278                 }
 279
 280                 targ->rx_port_queue[i].queue = prox_port_cfg[if_port].n_rxq;
 281                 prox_port_cfg[if_port].pool[targ->rx_port_queue[i].queue] = targ->pool;
 282                 prox_port_cfg[if_port].pool_size[targ->rx_port_queue[i].queue] = targ->nb_mbuf - 1;
 283                 prox_port_cfg[if_port].n_rxq++;
 284
 285                 int dsocket = prox_port_cfg[if_port].socket;
 286                 if (dsocket != -1 && dsocket != socket) {
 287                         plog_warn("RX core on socket %d while device on socket %d\n", socket, dsocket);
 288                 }
 289         }
 290 }
 291
 292 static void configure_if_queues(void)
 293 {
 294         struct lcore_cfg *lconf = NULL;
 295         struct task_args *targ;
 296         uint8_t socket;
 297
 298         while (core_targ_next(&lconf, &targ, 0) == 0) {
 299                 socket = rte_lcore_to_socket_id(lconf->id);
 300
 301                 configure_if_tx_queues(targ, socket);
 302                 configure_if_rx_queues(targ, socket);
 303         }
 304 }
 305
 306 static const char *gen_ring_name(void)
 307 {
 308         static char retval[] = "XX";
 309         static const char* ring_names =
 310                 "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
 311                 "abcdefghijklmnopqrstuvwxyz"
 312                 "[\\]^_`!\"#$%&'()*+,-./:;<="
 313                 ">?@{|}0123456789";
 314         static int idx2 = 0;
 315
 316         int idx = idx2;
 317
 318         retval[0] = ring_names[idx % strlen(ring_names)];
 319         idx /= strlen(ring_names);
 320         retval[1] = idx ? ring_names[(idx - 1) % strlen(ring_names)] : 0;
 321
 322         idx2++;
 323
 324         return retval;
 325 }
 326
 327 struct ring_init_stats {
 328         uint32_t n_pkt_rings;
 329         uint32_t n_ctrl_rings;
 330         uint32_t n_opt_rings;
 331 };
 332
 333 static uint32_t ring_init_stats_total(const struct ring_init_stats *ris)
 334 {
 335         return ris->n_pkt_rings + ris->n_ctrl_rings + ris->n_opt_rings;
 336 }
 337
 338 static uint32_t count_incoming_tasks(uint32_t lcore_worker, uint32_t dest_task)
 339 {
 340         struct lcore_cfg *lconf = NULL;
 341         struct task_args *targ;
 342         uint32_t ret = 0;
 343         struct core_task ct;
 344
 345         while (core_targ_next(&lconf, &targ, 0) == 0) {
 346                 for (uint8_t idxx = 0; idxx < MAX_PROTOCOLS; ++idxx) {
 347                         for (uint8_t ridx = 0; ridx < targ->core_task_set[idxx].n_elems; ++ridx) {
 348                                 ct = targ->core_task_set[idxx].core_task[ridx];
 349
 350                                 if (dest_task == ct.task && lcore_worker == ct.core)
 351                                         ret++;
 352                         }
 353                 }
 354         }
 355         return ret;
 356 }
 357
 358 static struct rte_ring *get_existing_ring(uint32_t lcore_id, uint32_t task_id)
 359 {
 360         if (!prox_core_active(lcore_id, 0))
 361                 return NULL;
 362
 363         struct lcore_cfg *lconf = &lcore_cfg[lcore_id];
 364
 365         if (task_id >= lconf->n_tasks_all)
 366                 return NULL;
 367
 368         if (lconf->targs[task_id].nb_rxrings == 0)
 369                 return NULL;
 370
 371         return lconf->targs[task_id].rx_rings[0];
 372 }
 373
 374 static struct rte_ring *init_ring_between_tasks(struct lcore_cfg *lconf, struct task_args *starg,
 375                                     const struct core_task ct, uint8_t ring_idx, int idx,
 376                                     struct ring_init_stats *ris)
 377 {
 378         uint8_t socket;
 379         struct rte_ring *ring = NULL;
 380         struct lcore_cfg *lworker;
 381         struct task_args *dtarg;
 382
 383         PROX_ASSERT(prox_core_active(ct.core, 0));
 384         lworker = &lcore_cfg[ct.core];
 385
 386         /* socket used is the one that the sending core resides on */
 387         socket = rte_lcore_to_socket_id(lconf->id);
 388
 389         plog_info("\t\tCreating ring on socket %u with size %u\n"
 390                   "\t\t\tsource core, task and socket = %u, %u, %u\n"
 391                   "\t\t\tdestination core, task and socket = %u, %u, %u\n"
 392                   "\t\t\tdestination worker id = %u\n",
 393                   socket, starg->ring_size,
 394                   lconf->id, starg->id, socket,
 395                   ct.core, ct.task, rte_lcore_to_socket_id(ct.core),
 396                   ring_idx);
 397
 398         if (ct.type) {
 399                 struct rte_ring **dring = NULL;
 400
 401                 if (ct.type == CTRL_TYPE_MSG)
 402                         dring = &lworker->ctrl_rings_m[ct.task];
 403                 else if (ct.type == CTRL_TYPE_PKT) {
 404                         dring = &lworker->ctrl_rings_p[ct.task];
 405                         starg->flags |= TASK_ARG_CTRL_RINGS_P;
 406                 }
 407
 408                 if (*dring == NULL)
 409                         ring = rte_ring_create(gen_ring_name(), starg->ring_size, socket, RING_F_SC_DEQ);
 410                 else
 411                         ring = *dring;
 412                 PROX_PANIC(ring == NULL, "Cannot create ring to connect I/O core %u with worker core %u\n", lconf->id, ct.core);
 413
 414                 starg->tx_rings[starg->tot_n_txrings_inited] = ring;
 415                 starg->tot_n_txrings_inited++;
 416                 *dring = ring;
 417                 if (lconf->id == prox_cfg.master) {
 418                         ctrl_rings[ct.core*MAX_TASKS_PER_CORE + ct.task] = ring;
 419                 } else if (ct.core == prox_cfg.master) {
 420                         starg->ctrl_plane_ring = ring;
 421                 }
 422
 423                 plog_info("\t\tCore %u task %u to -> core %u task %u ctrl_ring %s %p %s\n",
 424                           lconf->id, starg->id, ct.core, ct.task, ct.type == CTRL_TYPE_PKT?
 425                           "pkt" : "msg", ring, ring->name);
 426                 ris->n_ctrl_rings++;
 427                 return ring;
 428         }
 429
 430         dtarg = &lworker->targs[ct.task];
 431         lworker->targs[ct.task].worker_thread_id = ring_idx;
 432         PROX_ASSERT(dtarg->flags & TASK_ARG_RX_RING);
 433         PROX_ASSERT(ct.task < lworker->n_tasks_all);
 434
 435         /* If all the following conditions are met, the ring can be
 436            optimized away. */
 437         if (!task_is_master(starg) && !task_is_master(dtarg) && starg->lconf->id == dtarg->lconf->id &&
 438             starg->nb_txrings == 1 && idx == 0 && dtarg->task &&
 439             dtarg->tot_rxrings == 1 && starg->task == dtarg->task - 1) {
 440                 plog_info("\t\tOptimizing away ring on core %u from task %u to task %u\n",
 441                           dtarg->lconf->id, starg->task, dtarg->task);
 442                 /* No need to set up ws_mbuf. */
 443                 starg->tx_opt_ring = 1;
 444                 /* During init of destination task, the buffer in the
 445                    source task will be initialized. */
 446                 dtarg->tx_opt_ring_task = starg;
 447                 ris->n_opt_rings++;
 448                 ++dtarg->nb_rxrings;
 449                 return NULL;
 450         }
 451
 452         int ring_created = 1;
 453         /* Only create multi-producer rings if configured to do so AND
 454            there is only one task sending to the task */
 455         if ((prox_cfg.flags & DSF_MP_RINGS && count_incoming_tasks(ct.core, ct.task) > 1)
 456                 || (prox_cfg.flags & DSF_ENABLE_BYPASS)) {
 457                 ring = get_existing_ring(ct.core, ct.task);
 458
 459                 if (ring) {
 460                         plog_info("\t\tCore %u task %u creatign MP ring %p to core %u task %u\n",
 461                                   lconf->id, starg->id, ring, ct.core, ct.task);
 462                         ring_created = 0;
 463                 }
 464                 else {
 465                         ring = rte_ring_create(gen_ring_name(), starg->ring_size, socket, RING_F_SC_DEQ);
 466                         plog_info("\t\tCore %u task %u using MP ring %p from core %u task %u\n",
 467                                   lconf->id, starg->id, ring, ct.core, ct.task);
 468                 }
 469         }
 470         else
 471                 ring = rte_ring_create(gen_ring_name(), starg->ring_size, socket, RING_F_SP_ENQ | RING_F_SC_DEQ);
 472
 473         PROX_PANIC(ring == NULL, "Cannot create ring to connect I/O core %u with worker core %u\n", lconf->id, ct.core);
 474
 475         starg->tx_rings[starg->tot_n_txrings_inited] = ring;
 476         starg->tot_n_txrings_inited++;
 477
 478         if (ring_created) {
 479                 PROX_ASSERT(dtarg->nb_rxrings < MAX_RINGS_PER_TASK);
 480                 dtarg->rx_rings[dtarg->nb_rxrings] = ring;
 481                 ++dtarg->nb_rxrings;
 482         }
 483         dtarg->nb_slave_threads = starg->core_task_set[idx].n_elems;
 484         dtarg->lb_friend_core = lconf->id;
 485         dtarg->lb_friend_task = starg->id;
 486         plog_info("\t\tWorker thread %d has core %d, task %d as a lb friend\n", ct.core, lconf->id, starg->id);
 487         plog_info("\t\tCore %u task %u tx_ring[%u] -> core %u task %u rx_ring[%u] %p %s %u WT\n",
 488                   lconf->id, starg->id, ring_idx, ct.core, ct.task, dtarg->nb_rxrings, ring, ring->name,
 489                   dtarg->nb_slave_threads);
 490         ++ris->n_pkt_rings;
 491         return ring;
 492 }
 493
 494 static void init_rings(void)
 495 {
 496         struct lcore_cfg *lconf = NULL;
 497         struct task_args *starg;
 498         struct ring_init_stats ris = {0};
 499
 500         while (core_targ_next(&lconf, &starg, 1) == 0) {
 501                 plog_info("\t*** Initializing rings on core %u, task %u ***\n", lconf->id, starg->id);
 502                 for (uint8_t idx = 0; idx < MAX_PROTOCOLS; ++idx) {
 503                         for (uint8_t ring_idx = 0; ring_idx < starg->core_task_set[idx].n_elems; ++ring_idx) {
 504                                 PROX_ASSERT(ring_idx < MAX_WT_PER_LB);
 505                                 PROX_ASSERT(starg->tot_n_txrings_inited < MAX_RINGS_PER_TASK);
 506
 507                                 struct core_task ct = starg->core_task_set[idx].core_task[ring_idx];
 508                                 init_ring_between_tasks(lconf, starg, ct, ring_idx, idx, &ris);
 509                         }
 510                 }
 511         }
 512
 513         plog_info("\tInitialized %d rings:\n"
 514                   "\t\tNumber of packet rings: %u\n"
 515                   "\t\tNumber of control rings: %u\n"
 516                   "\t\tNumber of optimized rings: %u\n",
 517                   ring_init_stats_total(&ris),
 518                   ris.n_pkt_rings,
 519                   ris.n_ctrl_rings,
 520                   ris.n_opt_rings);
 521
 522         lconf = NULL;
 523         struct prox_port_cfg *port;
 524         while (core_targ_next(&lconf, &starg, 1) == 0) {
 525                 if ((starg->task_init) && (starg->task_init->flag_features & TASK_FEATURE_L3)) {
 526                         struct core_task ct;
 527                         ct.core = prox_cfg.master;
 528                         ct.task = 0;
 529                         ct.type = CTRL_TYPE_PKT;
 530                         struct rte_ring *rx_ring = init_ring_between_tasks(lconf, starg, ct, 0, 0, &ris);
 531
 532                         ct.core = lconf->id;
 533                         ct.task = starg->id;;
 534                         struct rte_ring *tx_ring = init_ring_between_tasks(lcore_cfg, lcore_cfg[prox_cfg.master].targs, ct, 0, 0, &ris);
 535                 }
 536         }
 537 }
 538
 539 static void shuffle_mempool(struct rte_mempool* mempool, uint32_t nb_mbuf)
 540 {
 541         struct rte_mbuf** pkts = prox_zmalloc(nb_mbuf * sizeof(*pkts), rte_socket_id());
 542         uint64_t got = 0;
 543
 544         while (rte_mempool_get_bulk(mempool, (void**)(pkts + got), 1) == 0)
 545                 ++got;
 546
 547         while (got) {
 548                 int idx;
 549                 do {
 550                         idx = rand() % nb_mbuf - 1;
 551                 } while (pkts[idx] == 0);
 552
 553                 rte_mempool_put_bulk(mempool, (void**)&pkts[idx], 1);
 554                 pkts[idx] = 0;
 555                 --got;
 556         };
 557         prox_free(pkts);
 558 }
 559
 560 static void setup_mempools_unique_per_socket(void)
 561 {
 562         uint32_t flags = 0;
 563         char name[64];
 564         struct lcore_cfg *lconf = NULL;
 565         struct task_args *targ;
 566
 567         struct rte_mempool     *pool[MAX_SOCKETS];
 568         uint32_t mbuf_count[MAX_SOCKETS] = {0};
 569         uint32_t nb_cache_mbuf[MAX_SOCKETS] = {0};
 570         uint32_t mbuf_size[MAX_SOCKETS] = {0};
 571
 572         while (core_targ_next_early(&lconf, &targ, 0) == 0) {
 573                 PROX_PANIC(targ->task_init == NULL, "task_init = NULL, is mode specified for core %d, task %d ?\n", lconf->id, targ->id);
 574                 uint8_t socket = rte_lcore_to_socket_id(lconf->id);
 575                 PROX_ASSERT(socket < MAX_SOCKETS);
 576
 577                 if (targ->mbuf_size_set_explicitely)
 578                         flags = MEMPOOL_F_NO_SPREAD;
 579                 if ((!targ->mbuf_size_set_explicitely) && (targ->task_init->mbuf_size != 0)) {
 580                         targ->mbuf_size = targ->task_init->mbuf_size;
 581                 }
 582                 if (targ->rx_port_queue[0].port != OUT_DISCARD) {
 583                         struct prox_port_cfg* port_cfg = &prox_port_cfg[targ->rx_port_queue[0].port];
 584                         PROX_ASSERT(targ->nb_mbuf != 0);
 585                         mbuf_count[socket] += targ->nb_mbuf;
 586                         if (nb_cache_mbuf[socket] == 0)
 587                                 nb_cache_mbuf[socket] = targ->nb_cache_mbuf;
 588                         else {
 589                                 PROX_PANIC(nb_cache_mbuf[socket] != targ->nb_cache_mbuf,
 590                                            "all mbuf_cache must have the same size if using a unique mempool per socket\n");
 591                         }
 592                         if (mbuf_size[socket] == 0)
 593                                 mbuf_size[socket] = targ->mbuf_size;
 594                         else {
 595                                 PROX_PANIC(mbuf_size[socket] != targ->mbuf_size,
 596                                            "all mbuf_size must have the same size if using a unique mempool per socket\n");
 597                         }
 598                         if ((!targ->mbuf_size_set_explicitely) && (strcmp(port_cfg->short_name, "vmxnet3") == 0)) {
 599                                 if (mbuf_size[socket] < MBUF_SIZE + RTE_PKTMBUF_HEADROOM)
 600                                         mbuf_size[socket] = MBUF_SIZE + RTE_PKTMBUF_HEADROOM;
 601                         }
 602                 }
 603         }
 604         for (int i = 0 ; i < MAX_SOCKETS; i++) {
 605                 if (mbuf_count[i] != 0) {
 606                         sprintf(name, "socket_%u_pool", i);
 607                         pool[i] = rte_mempool_create(name,
 608                                                      mbuf_count[i] - 1, mbuf_size[i],
 609                                                      nb_cache_mbuf[i],
 610                                                      sizeof(struct rte_pktmbuf_pool_private),
 611                                                      rte_pktmbuf_pool_init, NULL,
 612                                                      prox_pktmbuf_init, NULL,
 613                                                      i, flags);
 614                         PROX_PANIC(pool[i] == NULL, "\t\tError: cannot create mempool for socket %u\n", i);
 615                         plog_info("\t\tMempool %p size = %u * %u cache %u, socket %d\n", pool[i],
 616                                   mbuf_count[i], mbuf_size[i], nb_cache_mbuf[i], i);
 617
 618                         if (prox_cfg.flags & DSF_SHUFFLE) {
 619                                 shuffle_mempool(pool[i], mbuf_count[i]);
 620                         }
 621                 }
 622         }
 623
 624         lconf = NULL;
 625         while (core_targ_next_early(&lconf, &targ, 0) == 0) {
 626                 uint8_t socket = rte_lcore_to_socket_id(lconf->id);
 627
 628                 if (targ->rx_port_queue[0].port != OUT_DISCARD) {
 629                         /* use this pool for the interface that the core is receiving from */
 630                         /* If one core receives from multiple ports, all the ports use the same mempool */
 631                         targ->pool = pool[socket];
 632                         /* Set the number of mbuf to the number of the unique mempool, so that the used and free work */
 633                         targ->nb_mbuf = mbuf_count[socket];
 634                         plog_info("\t\tMempool %p size = %u * %u cache %u, socket %d\n", targ->pool,
 635                                   targ->nb_mbuf, mbuf_size[socket], targ->nb_cache_mbuf, socket);
 636                 }
 637         }
 638 }
 639
 640 static void setup_mempool_for_rx_task(struct lcore_cfg *lconf, struct task_args *targ)
 641 {
 642         const uint8_t socket = rte_lcore_to_socket_id(lconf->id);
 643         struct prox_port_cfg *port_cfg = &prox_port_cfg[targ->rx_port_queue[0].port];
 644         const struct rte_memzone *mz;
 645         struct rte_mempool *mp = NULL;
 646         uint32_t flags = 0;
 647         char memzone_name[64];
 648         char name[64];
 649
 650         /* mbuf size can be set
 651          *  - from config file (highest priority, overwriting any other config) - should only be used as workaround
 652          *  - through each 'mode', overwriting the default mbuf_size
 653          *  - defaulted to MBUF_SIZE i.e. 1518 Bytes
 654          * Except is set expliciteky, ensure that size is big enough for vmxnet3 driver
 655          */
 656         if (targ->mbuf_size_set_explicitely) {
 657                 flags = MEMPOOL_F_NO_SPREAD;
 658                 /* targ->mbuf_size already set */
 659         }
 660         else if (targ->task_init->mbuf_size != 0) {
 661                 /* mbuf_size not set through config file but set through mode */
 662                 targ->mbuf_size = targ->task_init->mbuf_size;
 663         }
 664         else if (strcmp(port_cfg->short_name, "vmxnet3") == 0) {
 665                 if (targ->mbuf_size < MBUF_SIZE + RTE_PKTMBUF_HEADROOM)
 666                         targ->mbuf_size = MBUF_SIZE + RTE_PKTMBUF_HEADROOM;
 667         }
 668
 669         /* allocate memory pool for packets */
 670         PROX_ASSERT(targ->nb_mbuf != 0);
 671
 672         if (targ->pool_name[0] == '\0') {
 673                 sprintf(name, "core_%u_port_%u_pool", lconf->id, targ->id);
 674         }
 675
 676         snprintf(memzone_name, sizeof(memzone_name)-1, "MP_%s", targ->pool_name);
 677         mz = rte_memzone_lookup(memzone_name);
 678
 679         if (mz != NULL) {
 680                 mp = (struct rte_mempool*)mz->addr;
 681
 682                 targ->nb_mbuf = mp->size;
 683                 targ->pool = mp;
 684         }
 685
 686 #ifdef RTE_LIBRTE_IVSHMEM_FALSE
 687         if (mz != NULL && mp != NULL && mp->phys_addr != mz->ioremap_addr) {
 688                 /* Init mbufs with ioremap_addr for dma */
 689                 mp->phys_addr = mz->ioremap_addr;
 690                 mp->elt_pa[0] = mp->phys_addr + (mp->elt_va_start - (uintptr_t)mp);
 691
 692                 struct prox_pktmbuf_reinit_args init_args;
 693                 init_args.mp = mp;
 694                 init_args.lconf = lconf;
 695
 696                 uint32_t elt_sz = mp->elt_size + mp->header_size + mp->trailer_size;
 697                 rte_mempool_obj_iter((void*)mp->elt_va_start, mp->size, elt_sz, 1,
 698                                      mp->elt_pa, mp->pg_num, mp->pg_shift, prox_pktmbuf_reinit, &init_args);
 699         }
 700 #endif
 701
 702         /* Use this pool for the interface that the core is
 703            receiving from if one core receives from multiple
 704            ports, all the ports use the same mempool */
 705         if (targ->pool == NULL) {
 706                 plog_info("\t\tCreating mempool with name '%s'\n", name);
 707                 targ->pool = rte_mempool_create(name,
 708                                                 targ->nb_mbuf - 1, targ->mbuf_size,
 709                                                 targ->nb_cache_mbuf,
 710                                                 sizeof(struct rte_pktmbuf_pool_private),
 711                                                 rte_pktmbuf_pool_init, NULL,
 712                                                 prox_pktmbuf_init, lconf,
 713                                                 socket, flags);
 714         }
 715
 716         PROX_PANIC(targ->pool == NULL,
 717                    "\t\tError: cannot create mempool for core %u port %u: %s\n", lconf->id, targ->id, rte_strerror(rte_errno));
 718
 719         plog_info("\t\tMempool %p size = %u * %u cache %u, socket %d\n", targ->pool,
 720                   targ->nb_mbuf, targ->mbuf_size, targ->nb_cache_mbuf, socket);
 721         if (prox_cfg.flags & DSF_SHUFFLE) {
 722                 shuffle_mempool(targ->pool, targ->nb_mbuf);
 723         }
 724 }
 725
 726 static void setup_mempools_multiple_per_socket(void)
 727 {
 728         struct lcore_cfg *lconf = NULL;
 729         struct task_args *targ;
 730
 731         while (core_targ_next_early(&lconf, &targ, 0) == 0) {
 732                 PROX_PANIC(targ->task_init == NULL, "task_init = NULL, is mode specified for core %d, task %d ?\n", lconf->id, targ->id);
 733                 if (targ->rx_port_queue[0].port == OUT_DISCARD)
 734                         continue;
 735                 setup_mempool_for_rx_task(lconf, targ);
 736         }
 737 }
 738
 739 static void setup_mempools(void)
 740 {
 741         if (prox_cfg.flags & UNIQUE_MEMPOOL_PER_SOCKET)
 742                 setup_mempools_unique_per_socket();
 743         else
 744                 setup_mempools_multiple_per_socket();
 745 }
 746
 747 static void set_task_lconf(void)
 748 {
 749         struct lcore_cfg *lconf;
 750         uint32_t lcore_id = -1;
 751
 752         while(prox_core_next(&lcore_id, 1) == 0) {
 753                 lconf = &lcore_cfg[lcore_id];
 754                 for (uint8_t task_id = 0; task_id < lconf->n_tasks_all; ++task_id) {
 755                         lconf->targs[task_id].lconf = lconf;
 756                 }
 757         }
 758 }
 759
 760 static void set_dest_threads(void)
 761 {
 762         struct lcore_cfg *lconf = NULL;
 763         struct task_args *targ;
 764
 765         while (core_targ_next(&lconf, &targ, 0) == 0) {
 766                 for (uint8_t idx = 0; idx < MAX_PROTOCOLS; ++idx) {
 767                         for (uint8_t ring_idx = 0; ring_idx < targ->core_task_set[idx].n_elems; ++ring_idx) {
 768                                 struct core_task ct = targ->core_task_set[idx].core_task[ring_idx];
 769
 770                                 struct task_args *dest_task = core_targ_get(ct.core, ct.task);
 771                                 dest_task->prev_tasks[dest_task->n_prev_tasks++] = targ;
 772                         }
 773                 }
 774         }
 775 }
 776
 777 static void setup_all_task_structs_early_init(void)
 778 {
 779         struct lcore_cfg *lconf = NULL;
 780         struct task_args *targ;
 781
 782         plog_info("\t*** Calling early init on all tasks ***\n");
 783         while (core_targ_next(&lconf, &targ, 0) == 0) {
 784                 if (targ->task_init->early_init) {
 785                         targ->task_init->early_init(targ);
 786                 }
 787         }
 788 }
 789
 790 static void setup_all_task_structs(void)
 791 {
 792         struct lcore_cfg *lconf;
 793         uint32_t lcore_id = -1;
 794         struct task_base *tmaster = NULL;
 795
 796         while(prox_core_next(&lcore_id, 1) == 0) {
 797                 lconf = &lcore_cfg[lcore_id];
 798                 for (uint8_t task_id = 0; task_id < lconf->n_tasks_all; ++task_id) {
 799                         if (task_is_master(&lconf->targs[task_id])) {
 800                                 plog_info("\tInitializing MASTER struct for core %d task %d\n", lcore_id, task_id);
 801                                 lconf->tasks_all[task_id] = init_task_struct(&lconf->targs[task_id]);
 802                                 tmaster = lconf->tasks_all[task_id];
 803                         }
 804                 }
 805         }
 806         PROX_PANIC(tmaster == NULL, "Can't initialize master task\n");
 807         lcore_id = -1;
 808
 809         while(prox_core_next(&lcore_id, 1) == 0) {
 810                 lconf = &lcore_cfg[lcore_id];
 811                 plog_info("\tInitializing struct for core %d with %d task\n", lcore_id, lconf->n_tasks_all);
 812                 for (uint8_t task_id = 0; task_id < lconf->n_tasks_all; ++task_id) {
 813                         if (!task_is_master(&lconf->targs[task_id])) {
 814                                 plog_info("\tInitializing struct for core %d task %d\n", lcore_id, task_id);
 815                                 lconf->targs[task_id].tmaster = tmaster;
 816                                 lconf->tasks_all[task_id] = init_task_struct(&lconf->targs[task_id]);
 817                         }
 818                 }
 819         }
 820 }
 821
 822 static void init_port_activate(void)
 823 {
 824         struct lcore_cfg *lconf = NULL;
 825         struct task_args *targ;
 826         uint8_t port_id = 0;
 827
 828         while (core_targ_next_early(&lconf, &targ, 0) == 0) {
 829                 for (int i = 0; i < targ->nb_rxports; i++) {
 830                         port_id = targ->rx_port_queue[i].port;
 831                         prox_port_cfg[port_id].active = 1;
 832                 }
 833
 834                 for (int i = 0; i < targ->nb_txports; i++) {
 835                         port_id = targ->tx_port_queue[i].port;
 836                         prox_port_cfg[port_id].active = 1;
 837                 }
 838         }
 839 }
 840
 841 /* Initialize cores and allocate mempools */
 842 static void init_lcores(void)
 843 {
 844         struct lcore_cfg *lconf = 0;
 845         uint32_t lcore_id = -1;
 846
 847         while(prox_core_next(&lcore_id, 0) == 0) {
 848                 uint8_t socket = rte_lcore_to_socket_id(lcore_id);
 849                 PROX_PANIC(socket + 1 > MAX_SOCKETS, "Can't configure core %u (on socket %u). MAX_SOCKET is set to %d\n", lcore_id, socket, MAX_SOCKETS);
 850         }
 851
 852         /* need to allocate mempools as the first thing to use the lowest possible address range */
 853         plog_info("=== Initializing mempools ===\n");
 854         setup_mempools();
 855
 856         lcore_cfg_alloc_hp();
 857
 858         set_dest_threads();
 859         set_task_lconf();
 860
 861         plog_info("=== Initializing port addresses ===\n");
 862         init_port_addr();
 863
 864         plog_info("=== Initializing queue numbers on cores ===\n");
 865         configure_if_queues();
 866
 867         plog_info("=== Initializing rings on cores ===\n");
 868         init_rings();
 869
 870         plog_info("=== Checking configuration consistency ===\n");
 871         check_cfg_consistent();
 872
 873         plog_all_rings();
 874
 875         setup_all_task_structs_early_init();
 876         plog_info("=== Initializing tasks ===\n");
 877         setup_all_task_structs();
 878 }
 879
 880 static int setup_prox(int argc, char **argv)
 881 {
 882         if (prox_read_config_file() != 0 ||
 883             prox_setup_rte(argv[0]) != 0) {
 884                 return -1;
 885         }
 886
 887         if (prox_cfg.flags & DSF_CHECK_SYNTAX) {
 888                 plog_info("=== Configuration file syntax has been checked ===\n\n");
 889                 exit(EXIT_SUCCESS);
 890         }
 891
 892         init_port_activate();
 893         plog_info("=== Initializing rte devices ===\n");
 894         if (!(prox_cfg.flags & DSF_USE_DUMMY_DEVICES))
 895                 init_rte_ring_dev();
 896         init_rte_dev(prox_cfg.flags & DSF_USE_DUMMY_DEVICES);
 897         plog_info("=== Calibrating TSC overhead ===\n");
 898         clock_init();
 899         plog_info("\tTSC running at %"PRIu64" Hz\n", rte_get_tsc_hz());
 900
 901         init_lcores();
 902         plog_info("=== Initializing ports ===\n");
 903         init_port_all();
 904
 905         if (prox_cfg.logbuf_size) {
 906                 prox_cfg.logbuf = prox_zmalloc(prox_cfg.logbuf_size, rte_socket_id());
 907                 PROX_PANIC(prox_cfg.logbuf == NULL, "Failed to allocate memory for logbuf with size = %d\n", prox_cfg.logbuf_size);
 908         }
 909
 910         if (prox_cfg.flags & DSF_CHECK_INIT) {
 911                 plog_info("=== Initialization sequence completed ===\n\n");
 912                 exit(EXIT_SUCCESS);
 913         }
 914
 915         /* Current way that works to disable DPDK logging */
 916         FILE *f = fopen("/dev/null", "r");
 917         rte_openlog_stream(f);
 918         plog_info("=== PROX started ===\n");
 919         return 0;
 920 }
 921
 922 static int success = 0;
 923 static void siguser_handler(int signal)
 924 {
 925         if (signal == SIGUSR1)
 926                 success = 1;
 927         else
 928                 success = 0;
 929 }
 930
 931 static void sigabrt_handler(__attribute__((unused)) int signum)
 932 {
 933         /* restore default disposition for SIGABRT and SIGPIPE */
 934         signal(SIGABRT, SIG_DFL);
 935         signal(SIGPIPE, SIG_DFL);
 936
 937         /* ignore further Ctrl-C */
 938         signal(SIGINT, SIG_IGN);
 939
 940         /* more drastic exit on tedious termination signal */
 941         plog_info("Aborting...\n");
 942         if (lcore_cfg != NULL) {
 943                 uint32_t lcore_id;
 944                 pthread_t thread_id, tid0, tid = pthread_self();
 945                 memset(&tid0, 0, sizeof(tid0));
 946
 947                 /* cancel all threads except current one */
 948                 lcore_id = -1;
 949                 while (prox_core_next(&lcore_id, 1) == 0) {
 950                         thread_id = lcore_cfg[lcore_id].thread_id;
 951                         if (pthread_equal(thread_id, tid0))
 952                                 continue;
 953                         if (pthread_equal(thread_id, tid))
 954                                 continue;
 955                         pthread_cancel(thread_id);
 956                 }
 957
 958                 /* wait for cancelled threads to terminate */
 959                 lcore_id = -1;
 960                 while (prox_core_next(&lcore_id, 1) == 0) {
 961                         thread_id = lcore_cfg[lcore_id].thread_id;
 962                         if (pthread_equal(thread_id, tid0))
 963                                 continue;
 964                         if (pthread_equal(thread_id, tid))
 965                                 continue;
 966                         pthread_join(thread_id, NULL);
 967                 }
 968         }
 969
 970         /* close ncurses */
 971         display_end();
 972
 973         /* close ports on termination signal */
 974         close_ports_atexit();
 975
 976         /* terminate now */
 977         abort();
 978 }
 979
 980 static void sigterm_handler(int signum)
 981 {
 982         /* abort on second Ctrl-C */
 983         if (signum == SIGINT)
 984                 signal(SIGINT, sigabrt_handler);
 985
 986         /* gracefully quit on harmless termination signal */
 987         /* ports will subsequently get closed at resulting exit */
 988         quit();
 989 }
 990
 991 int main(int argc, char **argv)
 992 {
 993         /* set en_US locale to print big numbers with ',' */
 994         setlocale(LC_NUMERIC, "en_US.utf-8");
 995
 996         if (prox_parse_args(argc, argv) != 0){
 997                 prox_usage(argv[0]);
 998         }
 999
1000         plog_init(prox_cfg.log_name, prox_cfg.log_name_pid);
1001         plog_info("=== " PROGRAM_NAME " " VERSION_STR " ===\n");
1002         plog_info("\tUsing DPDK %s\n", rte_version() + sizeof(RTE_VER_PREFIX));
1003         read_rdt_info();
1004
1005         if (prox_cfg.flags & DSF_LIST_TASK_MODES) {
1006                 /* list supported task modes and exit */
1007                 tasks_list();
1008                 return EXIT_SUCCESS;
1009         }
1010
1011         /* close ports at normal exit */
1012         atexit(close_ports_atexit);
1013         /* gracefully quit on harmless termination signals */
1014         signal(SIGHUP, sigterm_handler);
1015         signal(SIGINT, sigterm_handler);
1016         signal(SIGQUIT, sigterm_handler);
1017         signal(SIGTERM, sigterm_handler);
1018         signal(SIGUSR1, sigterm_handler);
1019         signal(SIGUSR2, sigterm_handler);
1020         /* more drastic exit on tedious termination signals */
1021         signal(SIGABRT, sigabrt_handler);
1022         signal(SIGPIPE, sigabrt_handler);
1023
1024         if (prox_cfg.flags & DSF_DAEMON) {
1025                 signal(SIGUSR1, siguser_handler);
1026                 signal(SIGUSR2, siguser_handler);
1027                 plog_info("=== Running in Daemon mode ===\n");
1028                 plog_info("\tForking child and waiting for setup completion\n");
1029
1030                 pid_t ppid = getpid();
1031                 pid_t pid = fork();
1032                 if (pid < 0) {
1033                         plog_err("Failed to fork process to run in daemon mode\n");
1034                         return EXIT_FAILURE;
1035                 }
1036
1037                 if (pid == 0) {
1038                         fclose(stdin);
1039                         fclose(stdout);
1040                         fclose(stderr);
1041                         if (setsid() < 0) {
1042                                 kill(ppid, SIGUSR2);
1043                                 return EXIT_FAILURE;
1044                         }
1045                         if (setup_prox(argc, argv) != 0) {
1046                                 kill(ppid, SIGUSR2);
1047                                 return EXIT_FAILURE;
1048                         }
1049                         else {
1050                                 kill(ppid, SIGUSR1);
1051                                 run(prox_cfg.flags);
1052                                 return EXIT_SUCCESS;
1053                         }
1054                 }
1055                 else {
1056                         /* Before exiting the parent, wait until the
1057                            child process has finished setting up */
1058                         pause();
1059                         if (prox_cfg.logbuf) {
1060                                 file_print(prox_cfg.logbuf);
1061                         }
1062                         return success? EXIT_SUCCESS : EXIT_FAILURE;
1063                 }
1064         }
1065
1066         if (setup_prox(argc, argv) != 0)
1067                 return EXIT_FAILURE;
1068         run(prox_cfg.flags);
1069         return EXIT_SUCCESS;
1070 }