suricata
runmode-dpdk.c
Go to the documentation of this file.
1 /* Copyright (C) 2021-2025 Open Information Security Foundation
2  *
3  * You can copy, redistribute or modify this Program under the terms of
4  * the GNU General Public License version 2 as published by the Free
5  * Software Foundation.
6  *
7  * This program is distributed in the hope that it will be useful,
8  * but WITHOUT ANY WARRANTY; without even the implied warranty of
9  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10  * GNU General Public License for more details.
11  *
12  * You should have received a copy of the GNU General Public License
13  * version 2 along with this program; if not, write to the Free Software
14  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
15  * 02110-1301, USA.
16  */
17 
18 /**
19  * \ingroup dpdk
20  *
21  * @{
22  */
23 
24 /**
25  * \file
26  *
27  * \author Lukas Sismis <lukas.sismis@gmail.com>
28  *
29  * DPDK runmode
30  *
31  */
32 
33 #include "suricata-common.h"
34 #include "runmodes.h"
35 #include "runmode-dpdk.h"
36 #include "decode.h"
37 #include "source-dpdk.h"
38 #include "util-runmodes.h"
39 #include "util-byte.h"
40 #include "util-cpu.h"
41 #include "util-debug.h"
42 #include "util-device-private.h"
43 #include "util-dpdk.h"
44 #include "util-dpdk-bonding.h"
45 #include "util-dpdk-common.h"
46 #include "util-dpdk-i40e.h"
47 #include "util-dpdk-ice.h"
48 #include "util-dpdk-ixgbe.h"
49 #include "util-dpdk-rss.h"
50 #include "util-time.h"
51 #include "util-conf.h"
52 #include "suricata.h"
53 #include "util-affinity.h"
54 
55 #ifdef HAVE_DPDK
56 
57 // Calculates the closest multiple of y from x
58 #define ROUNDUP(x, y) ((((x) + ((y)-1)) / (y)) * (y))
59 
60 /* Maximum DPDK EAL parameters count. */
61 #define EAL_ARGS 48
62 
63 struct Arguments {
64  uint16_t capacity;
65  char **argv;
66  uint16_t argc;
67 };
68 
69 static char *AllocArgument(size_t arg_len);
70 static char *AllocAndSetArgument(const char *arg);
71 static char *AllocAndSetOption(const char *arg);
72 
73 static void ArgumentsInit(struct Arguments *args, uint16_t capacity);
74 static void ArgumentsCleanup(struct Arguments *args);
75 static void ArgumentsAdd(struct Arguments *args, char *value);
76 static void ArgumentsAddOptionAndArgument(struct Arguments *args, const char *opt, const char *arg);
77 static void InitEal(void);
78 
79 static void ConfigSetIface(DPDKIfaceConfig *iconf, const char *entry_str);
80 static int ConfigSetThreads(DPDKIfaceConfig *iconf, const char *entry_str);
81 static int ConfigSetRxQueues(DPDKIfaceConfig *iconf, uint16_t nb_queues, uint16_t max_queues);
82 static int ConfigSetTxQueues(
83  DPDKIfaceConfig *iconf, uint16_t nb_queues, uint16_t max_queues, bool iface_sends_pkts);
84 static int ConfigSetMempoolSize(DPDKIfaceConfig *iconf, const char *entry_str);
85 static int ConfigSetMempoolCacheSize(DPDKIfaceConfig *iconf, const char *entry_str);
86 static int ConfigSetRxDescriptors(DPDKIfaceConfig *iconf, const char *entry_str, uint16_t max_desc);
87 static int ConfigSetTxDescriptors(
88  DPDKIfaceConfig *iconf, const char *entry_str, uint16_t max_desc, bool iface_sends_pkts);
89 static int ConfigSetMtu(DPDKIfaceConfig *iconf, intmax_t entry_int);
90 static bool ConfigSetPromiscuousMode(DPDKIfaceConfig *iconf, int entry_bool);
91 static bool ConfigSetMulticast(DPDKIfaceConfig *iconf, int entry_bool);
92 static int ConfigSetChecksumChecks(DPDKIfaceConfig *iconf, int entry_bool);
93 static int ConfigSetChecksumOffload(DPDKIfaceConfig *iconf, int entry_bool);
94 static int ConfigSetCopyIface(DPDKIfaceConfig *iconf, const char *entry_str);
95 static int ConfigSetCopyMode(DPDKIfaceConfig *iconf, const char *entry_str);
96 static int ConfigSetCopyIfaceSettings(DPDKIfaceConfig *iconf, const char *iface, const char *mode);
97 static void ConfigInit(DPDKIfaceConfig **iconf);
98 static int ConfigLoad(DPDKIfaceConfig *iconf, const char *iface);
99 static DPDKIfaceConfig *ConfigParse(const char *iface);
100 
101 static void DeviceInitPortConf(const DPDKIfaceConfig *iconf,
102  const struct rte_eth_dev_info *dev_info, struct rte_eth_conf *port_conf);
103 static int DeviceConfigureQueues(DPDKIfaceConfig *iconf, const struct rte_eth_dev_info *dev_info,
104  const struct rte_eth_conf *port_conf);
105 static int DeviceValidateOutIfaceConfig(DPDKIfaceConfig *iconf);
106 static int DeviceConfigureIPS(DPDKIfaceConfig *iconf);
107 static int DeviceConfigure(DPDKIfaceConfig *iconf);
108 static void *ParseDpdkConfigAndConfigureDevice(const char *iface);
109 static void DPDKDerefConfig(void *conf);
110 
111 #define DPDK_CONFIG_DEFAULT_THREADS "auto"
112 #define DPDK_CONFIG_DEFAULT_INTERRUPT_MODE false
113 #define DPDK_CONFIG_DEFAULT_MEMPOOL_SIZE "auto"
114 #define DPDK_CONFIG_DEFAULT_MEMPOOL_CACHE_SIZE "auto"
115 #define DPDK_CONFIG_DEFAULT_RX_DESCRIPTORS "auto"
116 #define DPDK_CONFIG_DEFAULT_TX_DESCRIPTORS "auto"
117 #define DPDK_CONFIG_DEFAULT_RSS_HASH_FUNCTIONS RTE_ETH_RSS_IP
118 #define DPDK_CONFIG_DEFAULT_MTU 1500
119 #define DPDK_CONFIG_DEFAULT_PROMISCUOUS_MODE 1
120 #define DPDK_CONFIG_DEFAULT_MULTICAST_MODE 1
121 #define DPDK_CONFIG_DEFAULT_CHECKSUM_VALIDATION 1
122 #define DPDK_CONFIG_DEFAULT_CHECKSUM_VALIDATION_OFFLOAD 1
123 #define DPDK_CONFIG_DEFAULT_VLAN_STRIP 0
124 #define DPDK_CONFIG_DEFAULT_LINKUP_TIMEOUT 0
125 #define DPDK_CONFIG_DEFAULT_COPY_MODE "none"
126 #define DPDK_CONFIG_DEFAULT_COPY_INTERFACE "none"
127 
128 DPDKIfaceConfigAttributes dpdk_yaml = {
129  .threads = "threads",
130  .irq_mode = "interrupt-mode",
131  .promisc = "promisc",
132  .multicast = "multicast",
133  .checksum_checks = "checksum-checks",
134  .checksum_checks_offload = "checksum-checks-offload",
135  .mtu = "mtu",
136  .vlan_strip_offload = "vlan-strip-offload",
137  .rss_hf = "rss-hash-functions",
138  .linkup_timeout = "linkup-timeout",
139  .mempool_size = "mempool-size",
140  .mempool_cache_size = "mempool-cache-size",
141  .rx_descriptors = "rx-descriptors",
142  .tx_descriptors = "tx-descriptors",
143  .copy_mode = "copy-mode",
144  .copy_iface = "copy-iface",
145 };
146 
147 /**
148  * \brief Input is a number of which we want to find the greatest divisor up to max_num (inclusive).
149  * The divisor is returned.
150  */
151 static uint32_t GreatestDivisorUpTo(uint32_t num, uint32_t max_num)
152 {
153  for (uint32_t i = max_num; i >= 2; i--) {
154  if (num % i == 0) {
155  return i;
156  }
157  }
158  return 1;
159 }
160 
161 /**
162  * \brief Input is a number of which we want to find the greatest power of 2 up to num. The power of
163  * 2 is returned or 0 if no valid power of 2 is found.
164  */
165 static uint64_t GreatestPowOf2UpTo(uint64_t num)
166 {
167  if (num == 0) {
168  return 0; // No power of 2 exists for 0
169  }
170 
171  // Bit manipulation to isolate the highest set bit
172  num |= (num >> 1);
173  num |= (num >> 2);
174  num |= (num >> 4);
175  num |= (num >> 8);
176  num |= (num >> 16);
177  num |= (num >> 32);
178  num = num - (num >> 1);
179 
180  return num;
181 }
182 
183 static char *AllocArgument(size_t arg_len)
184 {
185  SCEnter();
186  char *ptr;
187 
188  arg_len += 1; // null character
189  ptr = (char *)SCCalloc(arg_len, sizeof(char));
190  if (ptr == NULL)
191  FatalError("Could not allocate memory for an argument");
192 
193  SCReturnPtr(ptr, "char *");
194 }
195 
196 /**
197  * Allocates space for length of the given string and then copies contents
198  * @param arg String to set to the newly allocated space
199  * @return memory address if no error otherwise NULL (with errno set)
200  */
201 static char *AllocAndSetArgument(const char *arg)
202 {
203  SCEnter();
204  if (arg == NULL)
205  FatalError("Passed argument is NULL in DPDK config initialization");
206 
207  char *ptr;
208  size_t arg_len = strlen(arg);
209 
210  ptr = AllocArgument(arg_len);
211  strlcpy(ptr, arg, arg_len + 1);
212  SCReturnPtr(ptr, "char *");
213 }
214 
215 static char *AllocAndSetOption(const char *arg)
216 {
217  SCEnter();
218  if (arg == NULL)
219  FatalError("Passed option is NULL in DPDK config initialization");
220 
221  char *ptr = NULL;
222  size_t arg_len = strlen(arg);
223  uint8_t is_long_arg = arg_len > 1;
224  const char *dash_prefix = is_long_arg ? "--" : "-";
225  size_t full_len = arg_len + strlen(dash_prefix);
226 
227  ptr = AllocArgument(full_len);
228  strlcpy(ptr, dash_prefix, full_len);
229  strlcat(ptr, arg, full_len);
230  SCReturnPtr(ptr, "char *");
231 }
232 
233 static void ArgumentsInit(struct Arguments *args, uint16_t capacity)
234 {
235  SCEnter();
236  args->argv = SCCalloc(capacity, sizeof(*args->argv)); // alloc array of pointers
237  if (args->argv == NULL)
238  FatalError("Could not allocate memory for Arguments structure");
239 
240  args->capacity = capacity;
241  args->argc = 0;
242  SCReturn;
243 }
244 
245 static void ArgumentsCleanup(struct Arguments *args)
246 {
247  SCEnter();
248  for (int i = 0; i < args->argc; i++) {
249  if (args->argv[i] != NULL) {
250  SCFree(args->argv[i]);
251  args->argv[i] = NULL;
252  }
253  }
254 
255  SCFree(args->argv);
256  args->argv = NULL;
257  args->argc = 0;
258  args->capacity = 0;
259 }
260 
261 static void ArgumentsAdd(struct Arguments *args, char *value)
262 {
263  SCEnter();
264  if (args->argc + 1 > args->capacity)
265  FatalError("No capacity for more arguments (Max: %" PRIu32 ")", EAL_ARGS);
266 
267  args->argv[args->argc++] = value;
268  SCReturn;
269 }
270 
271 static void ArgumentsAddOptionAndArgument(struct Arguments *args, const char *opt, const char *arg)
272 {
273  SCEnter();
274  char *option;
275  char *argument;
276 
277  option = AllocAndSetOption(opt);
278  ArgumentsAdd(args, option);
279 
280  // Empty argument could mean option only (e.g. --no-huge)
281  if (arg == NULL || arg[0] == '\0')
282  SCReturn;
283 
284  argument = AllocAndSetArgument(arg);
285  ArgumentsAdd(args, argument);
286  SCReturn;
287 }
288 
289 static void InitEal(void)
290 {
291  SCEnter();
292  int retval;
293  SCConfNode *param;
294  const SCConfNode *eal_params = SCConfGetNode("dpdk.eal-params");
295  struct Arguments args;
296  char **eal_argv;
297 
298  if (eal_params == NULL) {
299  FatalError("DPDK EAL parameters not found in the config");
300  }
301 
302  ArgumentsInit(&args, EAL_ARGS);
303  ArgumentsAdd(&args, AllocAndSetArgument("suricata"));
304 
305  TAILQ_FOREACH (param, &eal_params->head, next) {
306  if (SCConfNodeIsSequence(param)) {
307  const char *key = param->name;
308  SCConfNode *val;
309  TAILQ_FOREACH (val, &param->head, next) {
310  ArgumentsAddOptionAndArgument(&args, key, (const char *)val->val);
311  }
312  continue;
313  }
314  ArgumentsAddOptionAndArgument(&args, param->name, param->val);
315  }
316 
317  // creating a shallow copy for cleanup because rte_eal_init changes array contents
318  eal_argv = SCCalloc(args.argc, sizeof(*args.argv));
319  if (eal_argv == NULL) {
320  FatalError("Failed to allocate memory for the array of DPDK EAL arguments");
321  }
322  memcpy(eal_argv, args.argv, args.argc * sizeof(*args.argv));
323 
324  rte_log_set_global_level(RTE_LOG_WARNING);
325  retval = rte_eal_init(args.argc, eal_argv);
326 
327  ArgumentsCleanup(&args);
328  SCFree(eal_argv);
329 
330  if (retval < 0) { // retval bound to the result of rte_eal_init
331  FatalError("DPDK EAL initialization error: %s", rte_strerror(-retval));
332  }
333 }
334 
335 static void DPDKDerefConfig(void *conf)
336 {
337  SCEnter();
338  DPDKIfaceConfig *iconf = (DPDKIfaceConfig *)conf;
339 
340  if (SC_ATOMIC_SUB(iconf->ref, 1) == 1) {
341  DPDKDeviceResourcesDeinit(&iconf->pkt_mempools);
342  SCFree(iconf);
343  }
344  SCReturn;
345 }
346 
347 static void ConfigInit(DPDKIfaceConfig **iconf)
348 {
349  SCEnter();
350  DPDKIfaceConfig *ptr = NULL;
351  ptr = SCCalloc(1, sizeof(DPDKIfaceConfig));
352  if (ptr == NULL)
353  FatalError("Could not allocate memory for DPDKIfaceConfig");
354 
355  ptr->out_port_id = UINT16_MAX; // make sure no port is set
356  SC_ATOMIC_INIT(ptr->ref);
357  (void)SC_ATOMIC_ADD(ptr->ref, 1);
358  ptr->DerefFunc = DPDKDerefConfig;
359  ptr->flags = 0;
360 
361  *iconf = ptr;
362  SCReturn;
363 }
364 
365 static void ConfigSetIface(DPDKIfaceConfig *iconf, const char *entry_str)
366 {
367  SCEnter();
368  int retval;
369 
370  if (entry_str == NULL || entry_str[0] == '\0')
371  FatalError("Interface name in DPDK config is NULL or empty");
372 
373  retval = rte_eth_dev_get_port_by_name(entry_str, &iconf->port_id);
374  if (retval < 0)
375  FatalError("%s: interface not found: %s", entry_str, rte_strerror(-retval));
376 
377  strlcpy(iconf->iface, entry_str, sizeof(iconf->iface));
378  SCReturn;
379 }
380 
381 static int ConfigSetThreads(DPDKIfaceConfig *iconf, const char *entry_str)
382 {
383  SCEnter();
384  static uint16_t remaining_auto_cpus = UINT16_MAX; // uninitialized
386  SCLogError("DPDK runmode requires configured thread affinity");
387  SCReturnInt(-EINVAL);
388  }
389 
390  bool wtaf_periface = true;
391  ThreadsAffinityType *wtaf = GetAffinityTypeForNameAndIface("worker-cpu-set", iconf->iface);
392  if (wtaf == NULL) {
393  wtaf_periface = false;
394  wtaf = GetAffinityTypeForNameAndIface("worker-cpu-set", NULL); // mandatory
395  if (wtaf == NULL) {
396  SCLogError("Specify worker-cpu-set list in the threading section");
397  SCReturnInt(-EINVAL);
398  }
399  }
400  ThreadsAffinityType *mtaf = GetAffinityTypeForNameAndIface("management-cpu-set", NULL);
401  if (mtaf == NULL) {
402  SCLogError("Specify management-cpu-set list in the threading section");
403  SCReturnInt(-EINVAL);
404  }
405  uint16_t sched_cpus = UtilAffinityGetAffinedCPUNum(wtaf);
406  if (sched_cpus == UtilCpuGetNumProcessorsOnline()) {
407  SCLogWarning(
408  "\"all\" specified in worker CPU cores affinity, excluding management threads");
409  UtilAffinityCpusExclude(wtaf, mtaf);
410  sched_cpus = UtilAffinityGetAffinedCPUNum(wtaf);
411  }
412 
413  if (sched_cpus == 0) {
414  SCLogError("No worker CPU cores with configured affinity were configured");
415  SCReturnInt(-EINVAL);
416  } else if (UtilAffinityCpusOverlap(wtaf, mtaf) != 0) {
417  SCLogWarning("Worker threads should not overlap with management threads in the CPU core "
418  "affinity configuration");
419  }
420 
421  const char *active_runmode = RunmodeGetActive();
422  if (active_runmode && !strcmp("single", active_runmode)) {
423  iconf->threads = 1;
424  SCReturnInt(0);
425  }
426 
427  if (entry_str == NULL) {
428  SCLogError("Number of threads for interface \"%s\" not specified", iconf->iface);
429  SCReturnInt(-EINVAL);
430  }
431 
432  if (strcmp(entry_str, "auto") == 0) {
433  if (wtaf_periface) {
434  iconf->threads = (uint16_t)sched_cpus;
435  SCLogConfig("%s: auto-assigned %u threads", iconf->iface, iconf->threads);
436  SCReturnInt(0);
437  }
438 
439  uint16_t live_dev_count = (uint16_t)LiveGetDeviceCountWithoutAssignedThreading();
440  if (live_dev_count == 0) {
441  SCLogError("No live devices found, cannot auto-assign threads");
442  SCReturnInt(-EINVAL);
443  }
444  iconf->threads = sched_cpus / live_dev_count;
445  if (iconf->threads == 0) {
446  SCLogError("Not enough worker CPU cores with affinity were configured");
447  SCReturnInt(-ERANGE);
448  }
449 
450  if (remaining_auto_cpus > 0) {
451  iconf->threads++;
452  remaining_auto_cpus--;
453  } else if (remaining_auto_cpus == UINT16_MAX) {
454  // first time auto-assignment
455  remaining_auto_cpus = sched_cpus % live_dev_count;
456  if (remaining_auto_cpus > 0) {
457  iconf->threads++;
458  remaining_auto_cpus--;
459  }
460  }
461  SCLogConfig("%s: auto-assigned %u threads", iconf->iface, iconf->threads);
462  SCReturnInt(0);
463  }
464 
465  if (StringParseUint16(&iconf->threads, 10, 0, entry_str) < 0) {
466  SCLogError("Threads entry for interface %s contain non-numerical characters - \"%s\"",
467  iconf->iface, entry_str);
468  SCReturnInt(-EINVAL);
469  }
470 
471  if (iconf->threads <= 0) {
472  SCLogError("%s: positive number of threads required", iconf->iface);
473  SCReturnInt(-ERANGE);
474  }
475 
476  SCReturnInt(0);
477 }
478 
479 static bool ConfigSetInterruptMode(DPDKIfaceConfig *iconf, bool enable)
480 {
481  SCEnter();
482  if (enable)
483  iconf->flags |= DPDK_IRQ_MODE;
484 
485  SCReturnBool(true);
486 }
487 
488 static int ConfigSetRxQueues(DPDKIfaceConfig *iconf, uint16_t nb_queues, uint16_t max_queues)
489 {
490  SCEnter();
491  if (nb_queues == 0) {
492  SCLogInfo("%s: positive number of RX queues is required", iconf->iface);
493  SCReturnInt(-ERANGE);
494  }
495 
496  if (nb_queues > max_queues) {
497  SCLogInfo("%s: number of RX queues cannot exceed %" PRIu16, iconf->iface, max_queues);
498  SCReturnInt(-ERANGE);
499  }
500 
501  iconf->nb_rx_queues = nb_queues;
502  SCReturnInt(0);
503 }
504 
505 static bool ConfigIfaceSendsPkts(const char *mode)
506 {
507  SCEnter();
508  if (strcmp(mode, "ips") == 0 || strcmp(mode, "tap") == 0) {
509  SCReturnBool(true);
510  }
511  SCReturnBool(false);
512 }
513 
514 static int ConfigSetTxQueues(
515  DPDKIfaceConfig *iconf, uint16_t nb_queues, uint16_t max_queues, bool iface_sends_pkts)
516 {
517  SCEnter();
518  if (nb_queues == 0 && iface_sends_pkts) {
519  SCLogInfo("%s: positive number of TX queues is required", iconf->iface);
520  SCReturnInt(-ERANGE);
521  }
522 
523  if (nb_queues > max_queues) {
524  SCLogInfo("%s: number of TX queues cannot exceed %" PRIu16, iconf->iface, max_queues);
525  SCReturnInt(-ERANGE);
526  }
527 
528  iconf->nb_tx_queues = nb_queues;
529  SCReturnInt(0);
530 }
531 
532 static uint32_t MempoolSizeCalculate(
533  uint32_t rx_queues, uint32_t rx_desc, uint32_t tx_queues, uint32_t tx_desc)
534 {
535  uint32_t sz = rx_queues * rx_desc + tx_queues * tx_desc;
536  if (!tx_queues || !tx_desc)
537  sz *= 2; // double to have enough space for RX descriptors
538 
539  return sz;
540 }
541 
542 static int ConfigSetMempoolSize(DPDKIfaceConfig *iconf, const char *entry_str)
543 {
544  SCEnter();
545  if (entry_str == NULL || entry_str[0] == '\0' || strcmp(entry_str, "auto") == 0) {
546  // calculate the mempool size based on the number of:
547  // - RX / TX queues
548  // - RX / TX descriptors
549  bool err = false;
550  if (iconf->nb_rx_queues == 0) {
551  // in IDS mode, we don't need TX queues
552  SCLogError("%s: cannot autocalculate mempool size without RX queues", iconf->iface);
553  err = true;
554  }
555 
556  if (iconf->nb_rx_desc == 0) {
557  SCLogError(
558  "%s: cannot autocalculate mempool size without RX descriptors", iconf->iface);
559  err = true;
560  }
561 
562  if (err) {
563  SCReturnInt(-EINVAL);
564  }
565 
566  iconf->mempool_size = MempoolSizeCalculate(
567  iconf->nb_rx_queues, iconf->nb_rx_desc, iconf->nb_tx_queues, iconf->nb_tx_desc);
568  SCReturnInt(0);
569  }
570 
571  if (StringParseUint32(&iconf->mempool_size, 10, 0, entry_str) < 0) {
572  SCLogError("%s: mempool size entry contain non-numerical characters - \"%s\"", iconf->iface,
573  entry_str);
574  SCReturnInt(-EINVAL);
575  }
576 
577  if (MempoolSizeCalculate(
578  iconf->nb_rx_queues, iconf->nb_rx_desc, iconf->nb_tx_queues, iconf->nb_tx_desc) >
579  iconf->mempool_size + 1) { // +1 to mask mempool size advice given in Suricata 7.0.x -
580  // mp_size should be n = (2^q - 1)
581  SCLogError("%s: mempool size is likely too small for the number of descriptors and queues, "
582  "set to \"auto\" or adjust to the value of \"%" PRIu32 "\"",
583  iconf->iface,
584  MempoolSizeCalculate(iconf->nb_rx_queues, iconf->nb_rx_desc, iconf->nb_tx_queues,
585  iconf->nb_tx_desc));
586  SCReturnInt(-ERANGE);
587  }
588 
589  if (iconf->mempool_size == 0) {
590  SCLogError("%s: mempool size requires a positive integer", iconf->iface);
591  SCReturnInt(-ERANGE);
592  }
593 
594  SCReturnInt(0);
595 }
596 
597 static uint32_t MempoolCacheSizeCalculate(uint32_t mp_sz)
598 {
599  // It is advised to have mempool cache size lower or equal to:
600  // RTE_MEMPOOL_CACHE_MAX_SIZE (by default 512) and "mempool-size / 1.5"
601  // and at the same time "mempool-size modulo cache_size == 0".
602  uint32_t max_cache_size = MIN(RTE_MEMPOOL_CACHE_MAX_SIZE, (uint32_t)(mp_sz / 1.5));
603  return GreatestDivisorUpTo(mp_sz, max_cache_size);
604 }
605 
606 static int ConfigSetMempoolCacheSize(DPDKIfaceConfig *iconf, const char *entry_str)
607 {
608  SCEnter();
609  if (entry_str == NULL || entry_str[0] == '\0' || strcmp(entry_str, "auto") == 0) {
610  // calculate the mempool size based on the mempool size (it needs to be already filled in)
611  if (iconf->mempool_size == 0) {
612  SCLogError("%s: cannot calculate mempool cache size of a mempool with size %d",
613  iconf->iface, iconf->mempool_size);
614  SCReturnInt(-EINVAL);
615  }
616 
617  iconf->mempool_cache_size = MempoolCacheSizeCalculate(iconf->mempool_size);
618  SCReturnInt(0);
619  }
620 
621  if (StringParseUint32(&iconf->mempool_cache_size, 10, 0, entry_str) < 0) {
622  SCLogError("%s: mempool cache size entry contain non-numerical characters - \"%s\"",
623  iconf->iface, entry_str);
624  SCReturnInt(-EINVAL);
625  }
626 
627  if (iconf->mempool_cache_size <= 0 || iconf->mempool_cache_size > RTE_MEMPOOL_CACHE_MAX_SIZE) {
628  SCLogError("%s: mempool cache size requires a positive number smaller than %" PRIu32,
629  iconf->iface, RTE_MEMPOOL_CACHE_MAX_SIZE);
630  SCReturnInt(-ERANGE);
631  }
632 
633  SCReturnInt(0);
634 }
635 
636 static int ConfigSetRxDescriptors(DPDKIfaceConfig *iconf, const char *entry_str, uint16_t max_desc)
637 {
638  SCEnter();
639  if (entry_str == NULL || entry_str[0] == '\0') {
640  SCLogInfo("%s: number of RX descriptors not found, going with: %s", iconf->iface,
641  DPDK_CONFIG_DEFAULT_RX_DESCRIPTORS);
642  entry_str = DPDK_CONFIG_DEFAULT_RX_DESCRIPTORS;
643  }
644 
645  if (strcmp(entry_str, "auto") == 0) {
646  iconf->nb_rx_desc = (uint16_t)GreatestPowOf2UpTo(max_desc);
647  SCReturnInt(0);
648  }
649 
650  if (StringParseUint16(&iconf->nb_rx_desc, 10, 0, entry_str) < 0) {
651  SCLogError("%s: RX descriptors entry contains non-numerical characters - \"%s\"",
652  iconf->iface, entry_str);
653  SCReturnInt(-EINVAL);
654  }
655 
656  if (iconf->nb_rx_desc == 0) {
657  SCLogError("%s: positive number of RX descriptors is required", iconf->iface);
658  SCReturnInt(-ERANGE);
659  } else if (iconf->nb_rx_desc > max_desc) {
660  SCLogError("%s: number of RX descriptors cannot exceed %" PRIu16, iconf->iface, max_desc);
661  SCReturnInt(-ERANGE);
662  }
663 
664  SCReturnInt(0);
665 }
666 
667 static int ConfigSetTxDescriptors(
668  DPDKIfaceConfig *iconf, const char *entry_str, uint16_t max_desc, bool iface_sends_pkts)
669 {
670  SCEnter();
671  if (entry_str == NULL || entry_str[0] == '\0') {
672  SCLogInfo("%s: number of TX descriptors not found, going with: %s", iconf->iface,
673  DPDK_CONFIG_DEFAULT_TX_DESCRIPTORS);
674  entry_str = DPDK_CONFIG_DEFAULT_TX_DESCRIPTORS;
675  }
676 
677  if (strcmp(entry_str, "auto") == 0) {
678  if (iface_sends_pkts) {
679  iconf->nb_tx_desc = (uint16_t)GreatestPowOf2UpTo(max_desc);
680  } else {
681  iconf->nb_tx_desc = 0;
682  }
683  SCReturnInt(0);
684  }
685 
686  if (StringParseUint16(&iconf->nb_tx_desc, 10, 0, entry_str) < 0) {
687  SCLogError("%s: TX descriptors entry contains non-numerical characters - \"%s\"",
688  iconf->iface, entry_str);
689  SCReturnInt(-EINVAL);
690  }
691 
692  if (iconf->nb_tx_desc == 0 && iface_sends_pkts) {
693  SCLogError("%s: positive number of TX descriptors is required", iconf->iface);
694  SCReturnInt(-ERANGE);
695  } else if (iconf->nb_tx_desc > max_desc) {
696  SCLogError("%s: number of TX descriptors cannot exceed %" PRIu16, iconf->iface, max_desc);
697  SCReturnInt(-ERANGE);
698  }
699 
700  SCReturnInt(0);
701 }
702 
703 static int ConfigSetRSSHashFunctions(DPDKIfaceConfig *iconf, const char *entry_str)
704 {
705  SCEnter();
706  if (entry_str == NULL || entry_str[0] == '\0' || strcmp(entry_str, "auto") == 0) {
707  iconf->rss_hf = DPDK_CONFIG_DEFAULT_RSS_HASH_FUNCTIONS;
708  SCReturnInt(0);
709  }
710 
711  if (StringParseUint64(&iconf->rss_hf, 0, 0, entry_str) < 0) {
712  SCLogError("%s: RSS hash functions entry contain non-numerical characters - \"%s\"",
713  iconf->iface, entry_str);
714  SCReturnInt(-EINVAL);
715  }
716 
717  SCReturnInt(0);
718 }
719 
720 static int ConfigSetMtu(DPDKIfaceConfig *iconf, intmax_t entry_int)
721 {
722  SCEnter();
723  if (entry_int < RTE_ETHER_MIN_MTU || entry_int > RTE_ETHER_MAX_JUMBO_FRAME_LEN) {
724  SCLogError("%s: MTU size can only be between %" PRIu32 " and %" PRIu32, iconf->iface,
725  RTE_ETHER_MIN_MTU, RTE_ETHER_MAX_JUMBO_FRAME_LEN);
726  SCReturnInt(-ERANGE);
727  }
728 
729  iconf->mtu = (uint16_t)entry_int;
730  SCReturnInt(0);
731 }
732 
733 static int ConfigSetLinkupTimeout(DPDKIfaceConfig *iconf, intmax_t entry_int)
734 {
735  SCEnter();
736  if (entry_int < 0 || entry_int > UINT16_MAX) {
737  SCLogError("%s: Link-up waiting timeout needs to be a positive number (up to %u) or 0 to "
738  "disable",
739  iconf->iface, UINT16_MAX);
740  SCReturnInt(-ERANGE);
741  }
742 
743  iconf->linkup_timeout = (uint16_t)entry_int;
744  SCReturnInt(0);
745 }
746 
747 static bool ConfigSetPromiscuousMode(DPDKIfaceConfig *iconf, int entry_bool)
748 {
749  SCEnter();
750  if (entry_bool)
751  iconf->flags |= DPDK_PROMISC;
752 
753  SCReturnBool(true);
754 }
755 
756 static bool ConfigSetMulticast(DPDKIfaceConfig *iconf, int entry_bool)
757 {
758  SCEnter();
759  if (entry_bool)
760  iconf->flags |= DPDK_MULTICAST; // enable
761 
762  SCReturnBool(true);
763 }
764 
765 static int ConfigSetChecksumChecks(DPDKIfaceConfig *iconf, int entry_bool)
766 {
767  SCEnter();
768  if (entry_bool)
769  iconf->checksum_mode = CHECKSUM_VALIDATION_ENABLE;
770 
771  SCReturnInt(0);
772 }
773 
774 static int ConfigSetChecksumOffload(DPDKIfaceConfig *iconf, int entry_bool)
775 {
776  SCEnter();
777  if (entry_bool)
778  iconf->flags |= DPDK_RX_CHECKSUM_OFFLOAD;
779 
780  SCReturnInt(0);
781 }
782 
783 static void ConfigSetVlanStrip(DPDKIfaceConfig *iconf, int entry_bool)
784 {
785  SCEnter();
786  iconf->vlan_strip_enabled = entry_bool;
787  SCReturn;
788 }
789 
790 static int ConfigSetCopyIface(DPDKIfaceConfig *iconf, const char *entry_str)
791 {
792  SCEnter();
793  int retval;
794 
795  if (entry_str == NULL || entry_str[0] == '\0' || strcmp(entry_str, "none") == 0) {
796  iconf->out_iface = NULL;
797  SCReturnInt(0);
798  }
799 
800  retval = rte_eth_dev_get_port_by_name(entry_str, &iconf->out_port_id);
801  if (retval < 0) {
802  SCLogError("%s: copy interface (%s) not found: %s", iconf->iface, entry_str,
803  rte_strerror(-retval));
804  SCReturnInt(retval);
805  }
806 
807  iconf->out_iface = entry_str;
808  SCReturnInt(0);
809 }
810 
811 static int ConfigSetCopyMode(DPDKIfaceConfig *iconf, const char *entry_str)
812 {
813  SCEnter();
814  if (entry_str == NULL) {
815  SCLogWarning("%s: no copy mode specified, changing to %s ", iconf->iface,
816  DPDK_CONFIG_DEFAULT_COPY_MODE);
817  entry_str = DPDK_CONFIG_DEFAULT_COPY_MODE;
818  }
819 
820  if (strcmp(entry_str, "none") != 0 && strcmp(entry_str, "tap") != 0 &&
821  strcmp(entry_str, "ips") != 0) {
822  SCLogWarning("%s: copy mode \"%s\" is not one of the possible values (none|tap|ips). "
823  "Changing to %s",
824  entry_str, iconf->iface, DPDK_CONFIG_DEFAULT_COPY_MODE);
825  entry_str = DPDK_CONFIG_DEFAULT_COPY_MODE;
826  }
827 
828  if (strcmp(entry_str, "none") == 0) {
829  iconf->copy_mode = DPDK_COPY_MODE_NONE;
830  } else if (strcmp(entry_str, "tap") == 0) {
831  iconf->copy_mode = DPDK_COPY_MODE_TAP;
832  } else if (strcmp(entry_str, "ips") == 0) {
833  iconf->copy_mode = DPDK_COPY_MODE_IPS;
834  }
835 
836  SCReturnInt(0);
837 }
838 
839 static int ConfigSetCopyIfaceSettings(DPDKIfaceConfig *iconf, const char *iface, const char *mode)
840 {
841  SCEnter();
842  int retval;
843 
844  retval = ConfigSetCopyIface(iconf, iface);
845  if (retval < 0)
846  SCReturnInt(retval);
847 
848  retval = ConfigSetCopyMode(iconf, mode);
849  if (retval < 0)
850  SCReturnInt(retval);
851 
852  if (iconf->copy_mode == DPDK_COPY_MODE_NONE) {
853  if (iconf->out_iface != NULL)
854  iconf->out_iface = NULL;
855  SCReturnInt(0);
856  }
857 
858  if (iconf->out_iface == NULL || strlen(iconf->out_iface) <= 0) {
859  SCLogError("%s: copy mode enabled but interface not set", iconf->iface);
860  SCReturnInt(-EINVAL);
861  }
862 
863  SCReturnInt(0);
864 }
865 
866 static int ConfigLoad(DPDKIfaceConfig *iconf, const char *iface)
867 {
868  SCEnter();
869  int retval;
870  SCConfNode *if_root;
871  SCConfNode *if_default;
872  const char *entry_str = NULL;
873  intmax_t entry_int = 0;
874  int entry_bool = 0;
875  const char *copy_iface_str = NULL;
876  const char *copy_mode_str = NULL;
877 
878  ConfigSetIface(iconf, iface);
879  struct rte_eth_dev_info dev_info = { 0 };
880  retval = rte_eth_dev_info_get(iconf->port_id, &dev_info);
881  if (retval < 0) {
882  SCLogError("%s: getting device info failed: %s", iconf->iface, rte_strerror(-retval));
883  SCReturnInt(retval);
884  }
885 
886  retval = SCConfSetRootAndDefaultNodes("dpdk.interfaces", iconf->iface, &if_root, &if_default);
887  if (retval < 0) {
888  FatalError("failed to find DPDK configuration for the interface %s", iconf->iface);
889  }
890 
891  retval = SCConfGetChildValueWithDefault(if_root, if_default, dpdk_yaml.threads, &entry_str) != 1
892  ? ConfigSetThreads(iconf, DPDK_CONFIG_DEFAULT_THREADS)
893  : ConfigSetThreads(iconf, entry_str);
894  if (retval < 0)
895  SCReturnInt(retval);
896 
897  bool irq_enable;
899  if_root, if_default, dpdk_yaml.irq_mode, &entry_bool);
900  if (retval != 1) {
901  irq_enable = DPDK_CONFIG_DEFAULT_INTERRUPT_MODE;
902  } else {
903  irq_enable = entry_bool ? true : false;
904  }
905  retval = ConfigSetInterruptMode(iconf, irq_enable);
906  if (retval != true)
907  SCReturnInt(-EINVAL);
908 
910  if_root, if_default, dpdk_yaml.copy_mode, &copy_mode_str);
911  if (retval != 1) {
912  copy_mode_str = DPDK_CONFIG_DEFAULT_COPY_MODE;
913  }
914 
916  if_root, if_default, dpdk_yaml.rx_descriptors, &entry_str) != 1
917  ? ConfigSetRxDescriptors(iconf, DPDK_CONFIG_DEFAULT_RX_DESCRIPTORS,
918  dev_info.rx_desc_lim.nb_max)
919  : ConfigSetRxDescriptors(iconf, entry_str, dev_info.rx_desc_lim.nb_max);
920  if (retval < 0)
921  SCReturnInt(retval);
922 
923  bool iface_sends_pkts = ConfigIfaceSendsPkts(copy_mode_str);
925  if_root, if_default, dpdk_yaml.tx_descriptors, &entry_str) != 1
926  ? ConfigSetTxDescriptors(iconf, DPDK_CONFIG_DEFAULT_TX_DESCRIPTORS,
927  dev_info.tx_desc_lim.nb_max, iface_sends_pkts)
928  : ConfigSetTxDescriptors(
929  iconf, entry_str, dev_info.tx_desc_lim.nb_max, iface_sends_pkts);
930  if (retval < 0)
931  SCReturnInt(retval);
932 
933  // currently only mapping "1 thread == 1 RX (and 1 TX queue in IPS mode)" is supported
934  retval = ConfigSetRxQueues(iconf, iconf->threads, dev_info.max_rx_queues);
935  if (retval < 0) {
936  SCLogError("%s: too many threads configured - reduce thread count to: %" PRIu16,
937  iconf->iface, dev_info.max_rx_queues);
938  SCReturnInt(retval);
939  }
940 
941  // currently only mapping "1 thread == 1 RX (and 1 TX queue in IPS mode)" is supported
942  uint16_t tx_queues = iconf->nb_tx_desc > 0 ? iconf->threads : 0;
943  retval = ConfigSetTxQueues(iconf, tx_queues, dev_info.max_tx_queues, iface_sends_pkts);
944  if (retval < 0) {
945  SCLogError("%s: too many threads configured - reduce thread count to: %" PRIu16,
946  iconf->iface, dev_info.max_tx_queues);
947  SCReturnInt(retval);
948  }
949 
951  if_root, if_default, dpdk_yaml.mempool_size, &entry_str) != 1
952  ? ConfigSetMempoolSize(iconf, DPDK_CONFIG_DEFAULT_MEMPOOL_SIZE)
953  : ConfigSetMempoolSize(iconf, entry_str);
954  if (retval < 0)
955  SCReturnInt(retval);
956 
958  if_root, if_default, dpdk_yaml.mempool_cache_size, &entry_str) != 1
959  ? ConfigSetMempoolCacheSize(iconf, DPDK_CONFIG_DEFAULT_MEMPOOL_CACHE_SIZE)
960  : ConfigSetMempoolCacheSize(iconf, entry_str);
961  if (retval < 0)
962  SCReturnInt(retval);
963 
964  retval = SCConfGetChildValueIntWithDefault(if_root, if_default, dpdk_yaml.mtu, &entry_int) != 1
965  ? ConfigSetMtu(iconf, DPDK_CONFIG_DEFAULT_MTU)
966  : ConfigSetMtu(iconf, entry_int);
967  if (retval < 0)
968  SCReturnInt(retval);
969 
970  retval = SCConfGetChildValueWithDefault(if_root, if_default, dpdk_yaml.rss_hf, &entry_str) != 1
971  ? ConfigSetRSSHashFunctions(iconf, NULL)
972  : ConfigSetRSSHashFunctions(iconf, entry_str);
973  if (retval < 0)
974  SCReturnInt(retval);
975 
977  if_root, if_default, dpdk_yaml.promisc, &entry_bool) != 1
978  ? ConfigSetPromiscuousMode(iconf, DPDK_CONFIG_DEFAULT_PROMISCUOUS_MODE)
979  : ConfigSetPromiscuousMode(iconf, entry_bool);
980  if (retval != true)
981  SCReturnInt(-EINVAL);
982 
984  if_root, if_default, dpdk_yaml.multicast, &entry_bool) != 1
985  ? ConfigSetMulticast(iconf, DPDK_CONFIG_DEFAULT_MULTICAST_MODE)
986  : ConfigSetMulticast(iconf, entry_bool);
987  if (retval != true)
988  SCReturnInt(-EINVAL);
989 
991  if_root, if_default, dpdk_yaml.checksum_checks, &entry_bool) != 1
992  ? ConfigSetChecksumChecks(iconf, DPDK_CONFIG_DEFAULT_CHECKSUM_VALIDATION)
993  : ConfigSetChecksumChecks(iconf, entry_bool);
994  if (retval < 0)
995  SCReturnInt(retval);
996 
998  if_root, if_default, dpdk_yaml.checksum_checks_offload, &entry_bool) != 1
999  ? ConfigSetChecksumOffload(
1000  iconf, DPDK_CONFIG_DEFAULT_CHECKSUM_VALIDATION_OFFLOAD)
1001  : ConfigSetChecksumOffload(iconf, entry_bool);
1002  if (retval < 0)
1003  SCReturnInt(retval);
1004 
1006  if_root, if_default, dpdk_yaml.vlan_strip_offload, &entry_bool);
1007  if (retval != 1) {
1008  ConfigSetVlanStrip(iconf, DPDK_CONFIG_DEFAULT_VLAN_STRIP);
1009  } else {
1010  ConfigSetVlanStrip(iconf, entry_bool);
1011  }
1012 
1014  if_root, if_default, dpdk_yaml.linkup_timeout, &entry_int) != 1
1015  ? ConfigSetLinkupTimeout(iconf, DPDK_CONFIG_DEFAULT_LINKUP_TIMEOUT)
1016  : ConfigSetLinkupTimeout(iconf, entry_int);
1017  if (retval < 0)
1018  SCReturnInt(retval);
1019 
1021  if_root, if_default, dpdk_yaml.copy_iface, &copy_iface_str);
1022  if (retval != 1) {
1023  copy_iface_str = DPDK_CONFIG_DEFAULT_COPY_INTERFACE;
1024  }
1025 
1026  retval = ConfigSetCopyIfaceSettings(iconf, copy_iface_str, copy_mode_str);
1027  if (retval < 0)
1028  SCReturnInt(retval);
1029 
1030  SCReturnInt(0);
1031 }
1032 
1033 static bool ConfigThreadsGenericIsValid(uint16_t iface_threads, ThreadsAffinityType *wtaf)
1034 {
1035  static uint32_t total_cpus = 0;
1036  total_cpus += iface_threads;
1037  if (wtaf == NULL) {
1038  SCLogError("Specify worker-cpu-set list in the threading section");
1039  return false;
1040  }
1041  if (total_cpus > UtilAffinityGetAffinedCPUNum(wtaf)) {
1042  SCLogError("Interfaces requested more cores than configured in the worker-cpu-set "
1043  "threading section (requested %d configured %d",
1044  total_cpus, UtilAffinityGetAffinedCPUNum(wtaf));
1045  return false;
1046  }
1047 
1048  return true;
1049 }
1050 
1051 static bool ConfigThreadsInterfaceIsValid(uint16_t iface_threads, ThreadsAffinityType *itaf)
1052 {
1053  if (iface_threads > UtilAffinityGetAffinedCPUNum(itaf)) {
1054  SCLogError("Interface requested more cores than configured in the interface-specific "
1055  "threading section (requested %d configured %d",
1056  iface_threads, UtilAffinityGetAffinedCPUNum(itaf));
1057  return false;
1058  }
1059 
1060  return true;
1061 }
1062 
1063 static bool ConfigIsThreadingValid(uint16_t iface_threads, const char *iface)
1064 {
1065  ThreadsAffinityType *itaf = GetAffinityTypeForNameAndIface("worker-cpu-set", iface);
1066  ThreadsAffinityType *wtaf = GetAffinityTypeForNameAndIface("worker-cpu-set", NULL);
1067  if (itaf && !ConfigThreadsInterfaceIsValid(iface_threads, itaf)) {
1068  return false;
1069  } else if (itaf == NULL && !ConfigThreadsGenericIsValid(iface_threads, wtaf)) {
1070  return false;
1071  }
1072  return true;
1073 }
1074 
1075 static DPDKIfaceConfig *ConfigParse(const char *iface)
1076 {
1077  SCEnter();
1078  int retval;
1079  DPDKIfaceConfig *iconf = NULL;
1080  if (iface == NULL)
1081  FatalError("DPDK interface is NULL");
1082 
1083  ConfigInit(&iconf);
1084  retval = ConfigLoad(iconf, iface);
1085  if (retval < 0 || !ConfigIsThreadingValid(iconf->threads, iface)) {
1086  iconf->DerefFunc(iconf);
1087  SCReturnPtr(NULL, "void *");
1088  }
1089 
1090  SCReturnPtr(iconf, "DPDKIfaceConfig *");
1091 }
1092 
1093 static void DeviceSetPMDSpecificRSS(struct rte_eth_rss_conf *rss_conf, const char *driver_name)
1094 {
1095  if (strcmp(driver_name, "net_i40e") == 0)
1096  i40eDeviceSetRSSConf(rss_conf);
1097  if (strcmp(driver_name, "net_ice") == 0)
1098  iceDeviceSetRSSConf(rss_conf);
1099  if (strcmp(driver_name, "net_ixgbe") == 0)
1100  ixgbeDeviceSetRSSHashFunction(&rss_conf->rss_hf);
1101  if (strcmp(driver_name, "net_e1000_igb") == 0)
1102  rss_conf->rss_hf = (RTE_ETH_RSS_IPV4 | RTE_ETH_RSS_IPV6 | RTE_ETH_RSS_IPV6_EX);
1103 }
1104 
1105 // Returns -1 if no bit is set
1106 static int32_t GetFirstSetBitPosition(uint64_t bits)
1107 {
1108  for (int32_t i = 0; i < 64; i++) {
1109  if (bits & BIT_U64(i))
1110  return i;
1111  }
1112  return -1;
1113 }
1114 
1115 static void DumpRSSFlags(const uint64_t requested, const uint64_t actual)
1116 {
1117  SCLogConfig("REQUESTED (groups):");
1118 
1119  SCLogConfig(
1120  "RTE_ETH_RSS_IP %sset", ((requested & RTE_ETH_RSS_IP) == RTE_ETH_RSS_IP) ? "" : "NOT ");
1121  SCLogConfig("RTE_ETH_RSS_TCP %sset",
1122  ((requested & RTE_ETH_RSS_TCP) == RTE_ETH_RSS_TCP) ? "" : "NOT ");
1123  SCLogConfig("RTE_ETH_RSS_UDP %sset",
1124  ((requested & RTE_ETH_RSS_UDP) == RTE_ETH_RSS_UDP) ? "" : "NOT ");
1125  SCLogConfig("RTE_ETH_RSS_SCTP %sset",
1126  ((requested & RTE_ETH_RSS_SCTP) == RTE_ETH_RSS_SCTP) ? "" : "NOT ");
1127  SCLogConfig("RTE_ETH_RSS_TUNNEL %sset",
1128  ((requested & RTE_ETH_RSS_TUNNEL) == RTE_ETH_RSS_TUNNEL) ? "" : "NOT ");
1129 
1130  SCLogConfig("REQUESTED (individual):");
1131  SCLogConfig("RTE_ETH_RSS_IPV4 (Bit position: %d) %sset",
1132  GetFirstSetBitPosition(RTE_ETH_RSS_IPV4), (requested & RTE_ETH_RSS_IPV4) ? "" : "NOT ");
1133  SCLogConfig("RTE_ETH_RSS_FRAG_IPV4 (Bit position: %d) %sset",
1134  GetFirstSetBitPosition(RTE_ETH_RSS_FRAG_IPV4),
1135  (requested & RTE_ETH_RSS_FRAG_IPV4) ? "" : "NOT ");
1136  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV4_TCP (Bit position: %d) %sset",
1137  GetFirstSetBitPosition(RTE_ETH_RSS_NONFRAG_IPV4_TCP),
1138  (requested & RTE_ETH_RSS_NONFRAG_IPV4_TCP) ? "" : "NOT ");
1139  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV4_UDP (Bit position: %d) %sset",
1140  GetFirstSetBitPosition(RTE_ETH_RSS_NONFRAG_IPV4_UDP),
1141  (requested & RTE_ETH_RSS_NONFRAG_IPV4_UDP) ? "" : "NOT ");
1142  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV4_SCTP (Bit position: %d) %sset",
1143  GetFirstSetBitPosition(RTE_ETH_RSS_NONFRAG_IPV4_SCTP),
1144  (requested & RTE_ETH_RSS_NONFRAG_IPV4_SCTP) ? "" : "NOT ");
1145  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV4_OTHER (Bit position: %d) %sset",
1146  GetFirstSetBitPosition(RTE_ETH_RSS_NONFRAG_IPV4_OTHER),
1147  (requested & RTE_ETH_RSS_NONFRAG_IPV4_OTHER) ? "" : "NOT ");
1148  SCLogConfig("RTE_ETH_RSS_IPV6 (Bit position: %d) %sset",
1149  GetFirstSetBitPosition(RTE_ETH_RSS_IPV6), (requested & RTE_ETH_RSS_IPV6) ? "" : "NOT ");
1150  SCLogConfig("RTE_ETH_RSS_FRAG_IPV6 (Bit position: %d) %sset",
1151  GetFirstSetBitPosition(RTE_ETH_RSS_FRAG_IPV6),
1152  (requested & RTE_ETH_RSS_FRAG_IPV6) ? "" : "NOT ");
1153  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV6_TCP (Bit position: %d) %sset",
1154  GetFirstSetBitPosition(RTE_ETH_RSS_NONFRAG_IPV6_TCP),
1155  (requested & RTE_ETH_RSS_NONFRAG_IPV6_TCP) ? "" : "NOT ");
1156  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV6_UDP (Bit position: %d) %sset",
1157  GetFirstSetBitPosition(RTE_ETH_RSS_NONFRAG_IPV6_UDP),
1158  (requested & RTE_ETH_RSS_NONFRAG_IPV6_UDP) ? "" : "NOT ");
1159  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV6_SCTP (Bit position: %d) %sset",
1160  GetFirstSetBitPosition(RTE_ETH_RSS_NONFRAG_IPV6_SCTP),
1161  (requested & RTE_ETH_RSS_NONFRAG_IPV6_SCTP) ? "" : "NOT ");
1162  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV6_OTHER (Bit position: %d) %sset",
1163  GetFirstSetBitPosition(RTE_ETH_RSS_NONFRAG_IPV6_OTHER),
1164  (requested & RTE_ETH_RSS_NONFRAG_IPV6_OTHER) ? "" : "NOT ");
1165 
1166  SCLogConfig("RTE_ETH_RSS_L2_PAYLOAD (Bit position: %d) %sset",
1167  GetFirstSetBitPosition(RTE_ETH_RSS_L2_PAYLOAD),
1168  (requested & RTE_ETH_RSS_L2_PAYLOAD) ? "" : "NOT ");
1169  SCLogConfig("RTE_ETH_RSS_IPV6_EX (Bit position: %d) %sset",
1170  GetFirstSetBitPosition(RTE_ETH_RSS_IPV6_EX),
1171  (requested & RTE_ETH_RSS_IPV6_EX) ? "" : "NOT ");
1172  SCLogConfig("RTE_ETH_RSS_IPV6_TCP_EX (Bit position: %d) %sset",
1173  GetFirstSetBitPosition(RTE_ETH_RSS_IPV6_TCP_EX),
1174  (requested & RTE_ETH_RSS_IPV6_TCP_EX) ? "" : "NOT ");
1175  SCLogConfig("RTE_ETH_RSS_IPV6_UDP_EX (Bit position: %d) %sset",
1176  GetFirstSetBitPosition(RTE_ETH_RSS_IPV6_UDP_EX),
1177  (requested & RTE_ETH_RSS_IPV6_UDP_EX) ? "" : "NOT ");
1178 
1179  SCLogConfig("RTE_ETH_RSS_PORT (Bit position: %d) %sset",
1180  GetFirstSetBitPosition(RTE_ETH_RSS_PORT), (requested & RTE_ETH_RSS_PORT) ? "" : "NOT ");
1181  SCLogConfig("RTE_ETH_RSS_VXLAN (Bit position: %d) %sset",
1182  GetFirstSetBitPosition(RTE_ETH_RSS_VXLAN),
1183  (requested & RTE_ETH_RSS_VXLAN) ? "" : "NOT ");
1184  SCLogConfig("RTE_ETH_RSS_NVGRE (Bit position: %d) %sset",
1185  GetFirstSetBitPosition(RTE_ETH_RSS_NVGRE),
1186  (requested & RTE_ETH_RSS_NVGRE) ? "" : "NOT ");
1187  SCLogConfig("RTE_ETH_RSS_GTPU (Bit position: %d) %sset",
1188  GetFirstSetBitPosition(RTE_ETH_RSS_GTPU), (requested & RTE_ETH_RSS_GTPU) ? "" : "NOT ");
1189 
1190  SCLogConfig("RTE_ETH_RSS_L3_SRC_ONLY (Bit position: %d) %sset",
1191  GetFirstSetBitPosition(RTE_ETH_RSS_L3_SRC_ONLY),
1192  (requested & RTE_ETH_RSS_L3_SRC_ONLY) ? "" : "NOT ");
1193  SCLogConfig("RTE_ETH_RSS_L3_DST_ONLY (Bit position: %d) %sset",
1194  GetFirstSetBitPosition(RTE_ETH_RSS_L3_DST_ONLY),
1195  (requested & RTE_ETH_RSS_L3_DST_ONLY) ? "" : "NOT ");
1196  SCLogConfig("RTE_ETH_RSS_L4_SRC_ONLY (Bit position: %d) %sset",
1197  GetFirstSetBitPosition(RTE_ETH_RSS_L4_SRC_ONLY),
1198  (requested & RTE_ETH_RSS_L4_SRC_ONLY) ? "" : "NOT ");
1199  SCLogConfig("RTE_ETH_RSS_L4_DST_ONLY (Bit position: %d) %sset",
1200  GetFirstSetBitPosition(RTE_ETH_RSS_L4_DST_ONLY),
1201  (requested & RTE_ETH_RSS_L4_DST_ONLY) ? "" : "NOT ");
1202  SCLogConfig("ACTUAL (group):");
1203  SCLogConfig(
1204  "RTE_ETH_RSS_IP %sset", ((actual & RTE_ETH_RSS_IP) == RTE_ETH_RSS_IP) ? "" : "NOT ");
1205  SCLogConfig(
1206  "RTE_ETH_RSS_TCP %sset", ((actual & RTE_ETH_RSS_TCP) == RTE_ETH_RSS_TCP) ? "" : "NOT ");
1207  SCLogConfig(
1208  "RTE_ETH_RSS_UDP %sset", ((actual & RTE_ETH_RSS_UDP) == RTE_ETH_RSS_UDP) ? "" : "NOT ");
1209  SCLogConfig("RTE_ETH_RSS_SCTP %sset",
1210  ((actual & RTE_ETH_RSS_SCTP) == RTE_ETH_RSS_SCTP) ? "" : "NOT ");
1211  SCLogConfig("RTE_ETH_RSS_TUNNEL %sset",
1212  ((actual & RTE_ETH_RSS_TUNNEL) == RTE_ETH_RSS_TUNNEL) ? "" : "NOT ");
1213 
1214  SCLogConfig("ACTUAL (individual flags):");
1215  SCLogConfig("RTE_ETH_RSS_IPV4 %sset", (actual & RTE_ETH_RSS_IPV4) ? "" : "NOT ");
1216  SCLogConfig("RTE_ETH_RSS_FRAG_IPV4 %sset", (actual & RTE_ETH_RSS_FRAG_IPV4) ? "" : "NOT ");
1217  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV4_TCP %sset",
1218  (actual & RTE_ETH_RSS_NONFRAG_IPV4_TCP) ? "" : "NOT ");
1219  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV4_UDP %sset",
1220  (actual & RTE_ETH_RSS_NONFRAG_IPV4_UDP) ? "" : "NOT ");
1221  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV4_SCTP %sset",
1222  (actual & RTE_ETH_RSS_NONFRAG_IPV4_SCTP) ? "" : "NOT ");
1223  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV4_OTHER %sset",
1224  (actual & RTE_ETH_RSS_NONFRAG_IPV4_OTHER) ? "" : "NOT ");
1225  SCLogConfig("RTE_ETH_RSS_IPV6 %sset", (actual & RTE_ETH_RSS_IPV6) ? "" : "NOT ");
1226  SCLogConfig("RTE_ETH_RSS_FRAG_IPV6 %sset", (actual & RTE_ETH_RSS_FRAG_IPV6) ? "" : "NOT ");
1227  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV6_TCP %sset",
1228  (actual & RTE_ETH_RSS_NONFRAG_IPV6_TCP) ? "" : "NOT ");
1229  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV6_UDP %sset",
1230  (actual & RTE_ETH_RSS_NONFRAG_IPV6_UDP) ? "" : "NOT ");
1231  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV6_SCTP %sset",
1232  (actual & RTE_ETH_RSS_NONFRAG_IPV6_SCTP) ? "" : "NOT ");
1233  SCLogConfig("RTE_ETH_RSS_NONFRAG_IPV6_OTHER %sset",
1234  (actual & RTE_ETH_RSS_NONFRAG_IPV6_OTHER) ? "" : "NOT ");
1235 
1236  SCLogConfig("RTE_ETH_RSS_L2_PAYLOAD %sset", (actual & RTE_ETH_RSS_L2_PAYLOAD) ? "" : "NOT ");
1237  SCLogConfig("RTE_ETH_RSS_IPV6_EX %sset", (actual & RTE_ETH_RSS_IPV6_EX) ? "" : "NOT ");
1238  SCLogConfig("RTE_ETH_RSS_IPV6_TCP_EX %sset", (actual & RTE_ETH_RSS_IPV6_TCP_EX) ? "" : "NOT ");
1239  SCLogConfig("RTE_ETH_RSS_IPV6_UDP_EX %sset", (actual & RTE_ETH_RSS_IPV6_UDP_EX) ? "" : "NOT ");
1240 
1241  SCLogConfig("RTE_ETH_RSS_PORT %sset", (actual & RTE_ETH_RSS_PORT) ? "" : "NOT ");
1242  SCLogConfig("RTE_ETH_RSS_VXLAN %sset", (actual & RTE_ETH_RSS_VXLAN) ? "" : "NOT ");
1243  SCLogConfig("RTE_ETH_RSS_NVGRE %sset", (actual & RTE_ETH_RSS_NVGRE) ? "" : "NOT ");
1244  SCLogConfig("RTE_ETH_RSS_GTPU %sset", (actual & RTE_ETH_RSS_GTPU) ? "" : "NOT ");
1245 
1246  SCLogConfig("RTE_ETH_RSS_L3_SRC_ONLY %sset", (actual & RTE_ETH_RSS_L3_SRC_ONLY) ? "" : "NOT ");
1247  SCLogConfig("RTE_ETH_RSS_L3_DST_ONLY %sset", (actual & RTE_ETH_RSS_L3_DST_ONLY) ? "" : "NOT ");
1248  SCLogConfig("RTE_ETH_RSS_L4_SRC_ONLY %sset", (actual & RTE_ETH_RSS_L4_SRC_ONLY) ? "" : "NOT ");
1249  SCLogConfig("RTE_ETH_RSS_L4_DST_ONLY %sset", (actual & RTE_ETH_RSS_L4_DST_ONLY) ? "" : "NOT ");
1250 }
1251 
1252 static void DumpRXOffloadCapabilities(const uint64_t rx_offld_capa)
1253 {
1254  SCLogConfig("RTE_ETH_RX_OFFLOAD_VLAN_STRIP - %savailable",
1255  rx_offld_capa & RTE_ETH_RX_OFFLOAD_VLAN_STRIP ? "" : "NOT ");
1256  SCLogConfig("RTE_ETH_RX_OFFLOAD_IPV4_CKSUM - %savailable",
1257  rx_offld_capa & RTE_ETH_RX_OFFLOAD_IPV4_CKSUM ? "" : "NOT ");
1258  SCLogConfig("RTE_ETH_RX_OFFLOAD_UDP_CKSUM - %savailable",
1259  rx_offld_capa & RTE_ETH_RX_OFFLOAD_UDP_CKSUM ? "" : "NOT ");
1260  SCLogConfig("RTE_ETH_RX_OFFLOAD_TCP_CKSUM - %savailable",
1261  rx_offld_capa & RTE_ETH_RX_OFFLOAD_TCP_CKSUM ? "" : "NOT ");
1262  SCLogConfig("RTE_ETH_RX_OFFLOAD_TCP_LRO - %savailable",
1263  rx_offld_capa & RTE_ETH_RX_OFFLOAD_TCP_LRO ? "" : "NOT ");
1264  SCLogConfig("RTE_ETH_RX_OFFLOAD_QINQ_STRIP - %savailable",
1265  rx_offld_capa & RTE_ETH_RX_OFFLOAD_QINQ_STRIP ? "" : "NOT ");
1266  SCLogConfig("RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM - %savailable",
1267  rx_offld_capa & RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM ? "" : "NOT ");
1268  SCLogConfig("RTE_ETH_RX_OFFLOAD_MACSEC_STRIP - %savailable",
1269  rx_offld_capa & RTE_ETH_RX_OFFLOAD_MACSEC_STRIP ? "" : "NOT ");
1270 #if RTE_VERSION < RTE_VERSION_NUM(22, 11, 0, 0)
1271  SCLogConfig("RTE_ETH_RX_OFFLOAD_HEADER_SPLIT - %savailable",
1272  rx_offld_capa & RTE_ETH_RX_OFFLOAD_HEADER_SPLIT ? "" : "NOT ");
1273 #endif
1274  SCLogConfig("RTE_ETH_RX_OFFLOAD_VLAN_FILTER - %savailable",
1275  rx_offld_capa & RTE_ETH_RX_OFFLOAD_VLAN_FILTER ? "" : "NOT ");
1276  SCLogConfig("RTE_ETH_RX_OFFLOAD_VLAN_EXTEND - %savailable",
1277  rx_offld_capa & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND ? "" : "NOT ");
1278  SCLogConfig("RTE_ETH_RX_OFFLOAD_SCATTER - %savailable",
1279  rx_offld_capa & RTE_ETH_RX_OFFLOAD_SCATTER ? "" : "NOT ");
1280  SCLogConfig("RTE_ETH_RX_OFFLOAD_TIMESTAMP - %savailable",
1281  rx_offld_capa & RTE_ETH_RX_OFFLOAD_TIMESTAMP ? "" : "NOT ");
1282  SCLogConfig("RTE_ETH_RX_OFFLOAD_SECURITY - %savailable",
1283  rx_offld_capa & RTE_ETH_RX_OFFLOAD_SECURITY ? "" : "NOT ");
1284  SCLogConfig("RTE_ETH_RX_OFFLOAD_KEEP_CRC - %savailable",
1285  rx_offld_capa & RTE_ETH_RX_OFFLOAD_KEEP_CRC ? "" : "NOT ");
1286  SCLogConfig("RTE_ETH_RX_OFFLOAD_SCTP_CKSUM - %savailable",
1287  rx_offld_capa & RTE_ETH_RX_OFFLOAD_SCTP_CKSUM ? "" : "NOT ");
1288  SCLogConfig("RTE_ETH_RX_OFFLOAD_OUTER_UDP_CKSUM - %savailable",
1289  rx_offld_capa & RTE_ETH_RX_OFFLOAD_OUTER_UDP_CKSUM ? "" : "NOT ");
1290  SCLogConfig("RTE_ETH_RX_OFFLOAD_RSS_HASH - %savailable",
1291  rx_offld_capa & RTE_ETH_RX_OFFLOAD_RSS_HASH ? "" : "NOT ");
1292 #if RTE_VERSION >= RTE_VERSION_NUM(20, 11, 0, 0)
1293  SCLogConfig("RTE_ETH_RX_OFFLOAD_BUFFER_SPLIT - %savailable",
1294  rx_offld_capa & RTE_ETH_RX_OFFLOAD_BUFFER_SPLIT ? "" : "NOT ");
1295 #endif
1296 }
1297 
1298 static int DeviceValidateMTU(const DPDKIfaceConfig *iconf, const struct rte_eth_dev_info *dev_info)
1299 {
1300  SCEnter();
1301  if (iconf->mtu > dev_info->max_mtu || iconf->mtu < dev_info->min_mtu) {
1302  SCLogError("%s: MTU out of bounds. "
1303  "Min MTU: %" PRIu16 " Max MTU: %" PRIu16,
1304  iconf->iface, dev_info->min_mtu, dev_info->max_mtu);
1305  SCReturnInt(-ERANGE);
1306  }
1307 
1308 #if RTE_VERSION < RTE_VERSION_NUM(21, 11, 0, 0)
1309  // check if jumbo frames are set and are available
1310  if (iconf->mtu > RTE_ETHER_MAX_LEN &&
1311  !(dev_info->rx_offload_capa & DEV_RX_OFFLOAD_JUMBO_FRAME)) {
1312  SCLogError("%s: jumbo frames not supported, set MTU to 1500", iconf->iface);
1313  SCReturnInt(-EINVAL);
1314  }
1315 #endif
1316 
1317  SCReturnInt(0);
1318 }
1319 
1320 static void DeviceSetMTU(struct rte_eth_conf *port_conf, uint16_t mtu)
1321 {
1322 #if RTE_VERSION >= RTE_VERSION_NUM(21, 11, 0, 0)
1323  port_conf->rxmode.mtu = mtu;
1324 #else
1325  port_conf->rxmode.max_rx_pkt_len = mtu;
1326  if (mtu > RTE_ETHER_MAX_LEN) {
1327  port_conf->rxmode.offloads |= DEV_RX_OFFLOAD_JUMBO_FRAME;
1328  }
1329 #endif
1330 }
1331 
1332 static void PortConfSetInterruptMode(const DPDKIfaceConfig *iconf, struct rte_eth_conf *port_conf)
1333 {
1334  SCLogConfig("%s: interrupt mode is %s", iconf->iface,
1335  iconf->flags & DPDK_IRQ_MODE ? "enabled" : "disabled");
1336  if (iconf->flags & DPDK_IRQ_MODE)
1337  port_conf->intr_conf.rxq = 1;
1338 }
1339 
1340 static void PortConfSetRSSConf(const DPDKIfaceConfig *iconf,
1341  const struct rte_eth_dev_info *dev_info, struct rte_eth_conf *port_conf)
1342 {
1343  if (dev_info->rx_offload_capa & RTE_ETH_RX_OFFLOAD_RSS_HASH) {
1344  if (iconf->nb_rx_queues > 1) {
1345  SCLogConfig("%s: RSS enabled for %d queues", iconf->iface, iconf->nb_rx_queues);
1346  port_conf->rx_adv_conf.rss_conf = (struct rte_eth_rss_conf){
1347  .rss_key = RSS_HKEY,
1348  .rss_key_len = RSS_HKEY_LEN,
1349  .rss_hf = iconf->rss_hf,
1350  };
1351 
1352  const char *dev_driver = dev_info->driver_name;
1353  if (strcmp(dev_info->driver_name, "net_bonding") == 0) {
1354  dev_driver = BondingDeviceDriverGet(iconf->port_id);
1355  }
1356 
1357  DeviceSetPMDSpecificRSS(&port_conf->rx_adv_conf.rss_conf, dev_driver);
1358 
1359  uint64_t rss_hf_tmp =
1360  port_conf->rx_adv_conf.rss_conf.rss_hf & dev_info->flow_type_rss_offloads;
1361  if (port_conf->rx_adv_conf.rss_conf.rss_hf != rss_hf_tmp) {
1362  DumpRSSFlags(port_conf->rx_adv_conf.rss_conf.rss_hf, rss_hf_tmp);
1363 
1364  SCLogWarning("%s: modified RSS hash function based on hardware support: "
1365  "requested:%#" PRIx64 ", configured:%#" PRIx64,
1366  iconf->iface, port_conf->rx_adv_conf.rss_conf.rss_hf, rss_hf_tmp);
1367  port_conf->rx_adv_conf.rss_conf.rss_hf = rss_hf_tmp;
1368  }
1369  port_conf->rxmode.mq_mode = RTE_ETH_MQ_RX_RSS;
1370  } else {
1371  SCLogConfig("%s: RSS not enabled", iconf->iface);
1372  port_conf->rx_adv_conf.rss_conf.rss_key = NULL;
1373  port_conf->rx_adv_conf.rss_conf.rss_hf = 0;
1374  }
1375  } else {
1376  SCLogConfig("%s: RSS not supported", iconf->iface);
1377  }
1378 }
1379 
1380 static void PortConfSetChsumOffload(const DPDKIfaceConfig *iconf,
1381  const struct rte_eth_dev_info *dev_info, struct rte_eth_conf *port_conf)
1382 {
1383  if (iconf->checksum_mode == CHECKSUM_VALIDATION_DISABLE) {
1384  SCLogConfig("%s: checksum validation disabled", iconf->iface);
1385  } else if ((dev_info->rx_offload_capa & RTE_ETH_RX_OFFLOAD_CHECKSUM) ==
1386  RTE_ETH_RX_OFFLOAD_CHECKSUM) { // multibit comparison to make sure all bits are set
1387  if (iconf->checksum_mode == CHECKSUM_VALIDATION_ENABLE &&
1388  iconf->flags & DPDK_RX_CHECKSUM_OFFLOAD) {
1389  SCLogConfig("%s: IP, TCP and UDP checksum validation offloaded", iconf->iface);
1390  port_conf->rxmode.offloads |= RTE_ETH_RX_OFFLOAD_CHECKSUM;
1391  } else if (iconf->checksum_mode == CHECKSUM_VALIDATION_ENABLE &&
1392  !(iconf->flags & DPDK_RX_CHECKSUM_OFFLOAD)) {
1393  SCLogConfig("%s: checksum validation enabled (but can be offloaded)", iconf->iface);
1394  }
1395  }
1396 }
1397 
1398 static void PortConfSetVlanOffload(const DPDKIfaceConfig *iconf,
1399  const struct rte_eth_dev_info *dev_info, struct rte_eth_conf *port_conf)
1400 {
1401  if (iconf->vlan_strip_enabled) {
1402  if (dev_info->rx_offload_capa & RTE_ETH_RX_OFFLOAD_VLAN_STRIP) {
1403  port_conf->rxmode.offloads |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
1404  SCLogConfig("%s: hardware VLAN stripping enabled", iconf->iface);
1405  } else {
1406  SCLogWarning("%s: hardware VLAN stripping enabled but not supported, disabling",
1407  iconf->iface);
1408  }
1409  }
1410 }
1411 
1412 static void DeviceInitPortConf(const DPDKIfaceConfig *iconf,
1413  const struct rte_eth_dev_info *dev_info, struct rte_eth_conf *port_conf)
1414 {
1415  DumpRXOffloadCapabilities(dev_info->rx_offload_capa);
1416  *port_conf = (struct rte_eth_conf){
1417  .rxmode = {
1418  .mq_mode = RTE_ETH_MQ_RX_NONE,
1419  .offloads = 0, // turn every offload off to prevent any packet modification
1420  },
1421  .txmode = {
1422  .mq_mode = RTE_ETH_MQ_TX_NONE,
1423  .offloads = 0,
1424  },
1425  };
1426 
1427  PortConfSetInterruptMode(iconf, port_conf);
1428 
1429  // configure RX offloads
1430  PortConfSetRSSConf(iconf, dev_info, port_conf);
1431  PortConfSetChsumOffload(iconf, dev_info, port_conf);
1432  DeviceSetMTU(port_conf, iconf->mtu);
1433  PortConfSetVlanOffload(iconf, dev_info, port_conf);
1434 
1435  if (dev_info->tx_offload_capa & RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE) {
1436  port_conf->txmode.offloads |= RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
1437  }
1438 }
1439 
1440 static int DeviceConfigureQueues(DPDKIfaceConfig *iconf, const struct rte_eth_dev_info *dev_info,
1441  const struct rte_eth_conf *port_conf)
1442 {
1443  SCEnter();
1444  int retval;
1445  struct rte_eth_rxconf rxq_conf;
1446  struct rte_eth_txconf txq_conf;
1447 
1448  retval = DPDKDeviceResourcesInit(&(iconf->pkt_mempools), iconf->nb_rx_queues);
1449  if (retval < 0) {
1450  goto cleanup;
1451  }
1452 
1453  // +4 for VLAN header
1454  uint16_t mtu_size = iconf->mtu + RTE_ETHER_CRC_LEN + RTE_ETHER_HDR_LEN + 4;
1455  uint16_t mbuf_size = ROUNDUP(mtu_size, 1024) + RTE_PKTMBUF_HEADROOM;
1456  // ceil the div so e.g. mp_size of 262144 and 262143 both lead to 65535 on 4 rx queues
1457  uint32_t q_mp_sz = (iconf->mempool_size + iconf->nb_rx_queues - 1) / iconf->nb_rx_queues - 1;
1458  uint32_t q_mp_cache_sz = MempoolCacheSizeCalculate(q_mp_sz);
1459  SCLogInfo("%s: creating %u packet mempools of size %u, cache size %u, mbuf size %u",
1460  iconf->iface, iconf->nb_rx_queues, q_mp_sz, q_mp_cache_sz, mbuf_size);
1461  for (int i = 0; i < iconf->nb_rx_queues; i++) {
1462  char mempool_name[64];
1463  snprintf(mempool_name, sizeof(mempool_name), "mp_%d_%.20s", i, iconf->iface);
1464  iconf->pkt_mempools->pkt_mp[i] = rte_pktmbuf_pool_create(
1465  mempool_name, q_mp_sz, q_mp_cache_sz, 0, mbuf_size, (int)iconf->socket_id);
1466  if (iconf->pkt_mempools->pkt_mp[i] == NULL) {
1467  retval = -rte_errno;
1468  SCLogError("%s: rte_pktmbuf_pool_create failed with code %d (mempool: %s) - %s",
1469  iconf->iface, rte_errno, mempool_name, rte_strerror(rte_errno));
1470  goto cleanup;
1471  }
1472  }
1473 
1474  for (uint16_t queue_id = 0; queue_id < iconf->nb_rx_queues; queue_id++) {
1475  rxq_conf = dev_info->default_rxconf;
1476  rxq_conf.offloads = port_conf->rxmode.offloads;
1477  rxq_conf.rx_thresh.hthresh = 0;
1478  rxq_conf.rx_thresh.pthresh = 0;
1479  rxq_conf.rx_thresh.wthresh = 0;
1480  rxq_conf.rx_free_thresh = 0;
1481  rxq_conf.rx_drop_en = 0;
1482  SCLogConfig("%s: setting up RX queue %d: rx_desc: %u offloads: 0x%" PRIx64
1483  " hthresh: %" PRIu8 " pthresh: %" PRIu8 " wthresh: %" PRIu8
1484  " free_thresh: %" PRIu16 " drop_en: %" PRIu8,
1485  iconf->iface, queue_id, iconf->nb_rx_desc, rxq_conf.offloads,
1486  rxq_conf.rx_thresh.hthresh, rxq_conf.rx_thresh.pthresh, rxq_conf.rx_thresh.wthresh,
1487  rxq_conf.rx_free_thresh, rxq_conf.rx_drop_en);
1488 
1489  retval = rte_eth_rx_queue_setup(iconf->port_id, queue_id, iconf->nb_rx_desc,
1490  (unsigned int)iconf->socket_id, &rxq_conf, iconf->pkt_mempools->pkt_mp[queue_id]);
1491  if (retval < 0) {
1492  SCLogError("%s: failed to setup RX queue %u: %s", iconf->iface, queue_id,
1493  rte_strerror(-retval));
1494  goto cleanup;
1495  }
1496  }
1497 
1498  for (uint16_t queue_id = 0; queue_id < iconf->nb_tx_queues; queue_id++) {
1499  txq_conf = dev_info->default_txconf;
1500  txq_conf.offloads = port_conf->txmode.offloads;
1501  SCLogConfig("%s: setting up TX queue %d: tx_desc: %" PRIu16 " tx: offloads: 0x%" PRIx64
1502  " hthresh: %" PRIu8 " pthresh: %" PRIu8 " wthresh: %" PRIu8
1503  " tx_free_thresh: %" PRIu16 " tx_rs_thresh: %" PRIu16
1504  " txq_deferred_start: %" PRIu8,
1505  iconf->iface, queue_id, iconf->nb_tx_desc, txq_conf.offloads,
1506  txq_conf.tx_thresh.hthresh, txq_conf.tx_thresh.pthresh, txq_conf.tx_thresh.wthresh,
1507  txq_conf.tx_free_thresh, txq_conf.tx_rs_thresh, txq_conf.tx_deferred_start);
1508  retval = rte_eth_tx_queue_setup(iconf->port_id, queue_id, iconf->nb_tx_desc,
1509  (unsigned int)iconf->socket_id, &txq_conf);
1510  if (retval < 0) {
1511  SCLogError("%s: failed to setup TX queue %u: %s", iconf->iface, queue_id,
1512  rte_strerror(-retval));
1513  retval = -1; // the error code explained, informing about failure
1514  goto cleanup;
1515  }
1516  }
1517 
1518  SCReturnInt(0);
1519 
1520 cleanup:
1521  DPDKDeviceResourcesDeinit(&iconf->pkt_mempools);
1522  SCReturnInt(retval);
1523 }
1524 
1525 static int DeviceValidateOutIfaceConfig(DPDKIfaceConfig *iconf)
1526 {
1527  SCEnter();
1528  int retval;
1529  DPDKIfaceConfig *out_iconf = NULL;
1530  ConfigInit(&out_iconf);
1531  if (out_iconf == NULL) {
1532  FatalError("Copy interface of the interface \"%s\" is NULL", iconf->iface);
1533  }
1534 
1535  retval = ConfigLoad(out_iconf, iconf->out_iface);
1536  if (retval < 0) {
1537  SCLogError("%s: fail to load config of interface", iconf->out_iface);
1538  out_iconf->DerefFunc(out_iconf);
1539  SCReturnInt(-EINVAL);
1540  }
1541 
1542  if (iconf->nb_rx_queues != out_iconf->nb_tx_queues) {
1543  // the other direction is validated when the copy interface is configured
1544  SCLogError("%s: configured %d RX queues but copy interface %s has %d TX queues"
1545  " - number of queues must be equal",
1546  iconf->iface, iconf->nb_rx_queues, out_iconf->iface, out_iconf->nb_tx_queues);
1547  out_iconf->DerefFunc(out_iconf);
1548  SCReturnInt(-EINVAL);
1549  } else if (iconf->mtu != out_iconf->mtu) {
1550  SCLogError("%s: configured MTU of %d but copy interface %s has MTU set to %d"
1551  " - MTU must be equal",
1552  iconf->iface, iconf->mtu, out_iconf->iface, out_iconf->mtu);
1553  out_iconf->DerefFunc(out_iconf);
1554  SCReturnInt(-EINVAL);
1555  } else if (iconf->copy_mode != out_iconf->copy_mode) {
1556  SCLogError("%s: copy modes of interfaces %s and %s are not equal", iconf->iface,
1557  iconf->iface, out_iconf->iface);
1558  out_iconf->DerefFunc(out_iconf);
1559  SCReturnInt(-EINVAL);
1560  } else if (strcmp(iconf->iface, out_iconf->out_iface) != 0) {
1561  // check if the other iface has the current iface set as a copy iface
1562  SCLogError("%s: copy interface of %s is not set to %s", iconf->iface, out_iconf->iface,
1563  iconf->iface);
1564  out_iconf->DerefFunc(out_iconf);
1565  SCReturnInt(-EINVAL);
1566  }
1567 
1568  out_iconf->DerefFunc(out_iconf);
1569  SCReturnInt(0);
1570 }
1571 
1572 static int DeviceConfigureIPS(DPDKIfaceConfig *iconf)
1573 {
1574  SCEnter();
1575  if (iconf->out_iface != NULL) {
1576  if (!rte_eth_dev_is_valid_port(iconf->out_port_id)) {
1577  SCLogError("%s: retrieved copy interface port ID \"%d\" is invalid or the device is "
1578  "not attached ",
1579  iconf->iface, iconf->out_port_id);
1580  SCReturnInt(-ENODEV);
1581  }
1582  int32_t out_port_socket_id;
1583  int retval = DPDKDeviceSetSocketID(iconf->out_port_id, &out_port_socket_id);
1584  if (retval < 0) {
1585  SCLogError("%s: invalid socket id: %s", iconf->out_iface, rte_strerror(-retval));
1586  SCReturnInt(retval);
1587  }
1588 
1589  if (iconf->socket_id != out_port_socket_id) {
1590  SCLogWarning(
1591  "%s: out iface %s is not on the same NUMA node (%s - NUMA %d, %s - NUMA %d)",
1592  iconf->iface, iconf->out_iface, iconf->iface, iconf->socket_id,
1593  iconf->out_iface, out_port_socket_id);
1594  }
1595 
1596  retval = DeviceValidateOutIfaceConfig(iconf);
1597  if (retval != 0) {
1598  // Error will be written out by the validation function
1599  SCReturnInt(retval);
1600  }
1601 
1602  if (iconf->copy_mode == DPDK_COPY_MODE_IPS)
1603  SCLogInfo("%s: DPDK IPS mode activated: %s->%s", iconf->iface, iconf->iface,
1604  iconf->out_iface);
1605  else if (iconf->copy_mode == DPDK_COPY_MODE_TAP)
1606  SCLogInfo("%s: DPDK TAP mode activated: %s->%s", iconf->iface, iconf->iface,
1607  iconf->out_iface);
1608  }
1609  SCReturnInt(0);
1610 }
1611 
1612 /**
1613  * Function verifies changes in e.g. device info after configuration has
1614  * happened. Sometimes (e.g. DPDK Bond PMD with Intel NICs i40e/ixgbe) change
1615  * device info only after the device configuration.
1616  * @param iconf
1617  * @param dev_info
1618  * @return 0 on success, -EAGAIN when reconfiguration is needed, <0 on failure
1619  */
1620 static int32_t DeviceVerifyPostConfigure(
1621  const DPDKIfaceConfig *iconf, const struct rte_eth_dev_info *dev_info)
1622 {
1623  SCEnter();
1624  struct rte_eth_dev_info post_conf_dev_info = { 0 };
1625  int32_t ret = rte_eth_dev_info_get(iconf->port_id, &post_conf_dev_info);
1626  if (ret < 0) {
1627  SCLogError("%s: getting device info failed: %s", iconf->iface, rte_strerror(-ret));
1628  SCReturnInt(ret);
1629  }
1630 
1631  if (dev_info->flow_type_rss_offloads != post_conf_dev_info.flow_type_rss_offloads ||
1632  dev_info->rx_offload_capa != post_conf_dev_info.rx_offload_capa ||
1633  dev_info->tx_offload_capa != post_conf_dev_info.tx_offload_capa ||
1634  dev_info->max_rx_queues != post_conf_dev_info.max_rx_queues ||
1635  dev_info->max_tx_queues != post_conf_dev_info.max_tx_queues ||
1636  dev_info->max_mtu != post_conf_dev_info.max_mtu) {
1637  SCLogWarning("%s: device information severely changed after configuration, reconfiguring",
1638  iconf->iface);
1639  return -EAGAIN;
1640  }
1641 
1642  if (strcmp(dev_info->driver_name, "net_bonding") == 0) {
1643  ret = BondingAllDevicesSameDriver(iconf->port_id);
1644  if (ret < 0) {
1645  SCLogError("%s: bond port uses port with different DPDK drivers", iconf->iface);
1646  SCReturnInt(ret);
1647  }
1648  }
1649 
1650  return 0;
1651 }
1652 
1653 static int DeviceConfigure(DPDKIfaceConfig *iconf)
1654 {
1655  SCEnter();
1656  if (!rte_eth_dev_is_valid_port(iconf->port_id)) {
1657  SCLogError("%s: retrieved port ID \"%d\" is invalid or the device is not attached ",
1658  iconf->iface, iconf->port_id);
1659  SCReturnInt(-ENODEV);
1660  }
1661 
1662  int32_t retval = DPDKDeviceSetSocketID(iconf->port_id, &iconf->socket_id);
1663  if (retval < 0) {
1664  SCLogError("%s: invalid socket id: %s", iconf->iface, rte_strerror(-retval));
1665  SCReturnInt(retval);
1666  }
1667 
1668  struct rte_eth_dev_info dev_info = { 0 };
1669  retval = rte_eth_dev_info_get(iconf->port_id, &dev_info);
1670  if (retval < 0) {
1671  SCLogError("%s: getting device info failed: %s", iconf->iface, rte_strerror(-retval));
1672  SCReturnInt(retval);
1673  }
1674 
1675  if (iconf->nb_rx_queues > dev_info.max_rx_queues) {
1676  SCLogError("%s: configured RX queues %u is higher than device maximum (%" PRIu16 ")",
1677  iconf->iface, iconf->nb_rx_queues, dev_info.max_rx_queues);
1678  SCReturnInt(-ERANGE);
1679  }
1680 
1681  if (iconf->nb_tx_queues > dev_info.max_tx_queues) {
1682  SCLogError("%s: configured TX queues %u is higher than device maximum (%" PRIu16 ")",
1683  iconf->iface, iconf->nb_tx_queues, dev_info.max_tx_queues);
1684  SCReturnInt(-ERANGE);
1685  }
1686 
1687  retval = DeviceValidateMTU(iconf, &dev_info);
1688  if (retval < 0)
1689  return retval;
1690 
1691  struct rte_eth_conf port_conf = { 0 };
1692  DeviceInitPortConf(iconf, &dev_info, &port_conf);
1693  if (port_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_CHECKSUM) {
1694  // Suricata does not need recalc checksums now
1695  iconf->checksum_mode = CHECKSUM_VALIDATION_OFFLOAD;
1696  }
1697 
1698  retval = rte_eth_dev_configure(
1699  iconf->port_id, iconf->nb_rx_queues, iconf->nb_tx_queues, &port_conf);
1700  if (retval < 0) {
1701  SCLogError("%s: failed to configure the device: %s", iconf->iface, rte_strerror(-retval));
1702  SCReturnInt(retval);
1703  }
1704 
1705  retval = DeviceVerifyPostConfigure(iconf, &dev_info);
1706  if (retval < 0)
1707  return retval;
1708 
1709  uint16_t tmp_nb_rx_desc = iconf->nb_rx_desc;
1710  uint16_t tmp_nb_tx_desc = iconf->nb_tx_desc;
1711  retval = rte_eth_dev_adjust_nb_rx_tx_desc(
1712  iconf->port_id, &iconf->nb_rx_desc, &iconf->nb_tx_desc);
1713  if (retval != 0) {
1714  SCLogError("%s: failed to adjust device queue descriptors: %s", iconf->iface,
1715  rte_strerror(-retval));
1716  SCReturnInt(retval);
1717  } else if (tmp_nb_rx_desc != iconf->nb_rx_desc || tmp_nb_tx_desc != iconf->nb_tx_desc) {
1718  SCLogWarning("%s: device queue descriptors adjusted (RX: from %u to %u, TX: from %u to %u)",
1719  iconf->iface, tmp_nb_rx_desc, iconf->nb_rx_desc, tmp_nb_tx_desc, iconf->nb_tx_desc);
1720  }
1721 
1722  retval = iconf->flags & DPDK_MULTICAST ? rte_eth_allmulticast_enable(iconf->port_id)
1723  : rte_eth_allmulticast_disable(iconf->port_id);
1724  if (retval == -ENOTSUP) {
1725  retval = rte_eth_allmulticast_get(iconf->port_id);
1726  // when multicast is enabled but set to disable or vice versa
1727  if ((retval == 1 && !(iconf->flags & DPDK_MULTICAST)) ||
1728  (retval == 0 && (iconf->flags & DPDK_MULTICAST))) {
1729  SCLogWarning("%s: cannot configure allmulticast, the port is %sin allmulticast mode",
1730  iconf->iface, retval == 1 ? "" : "not ");
1731  } else if (retval < 0) {
1732  SCLogError("%s: failed to get multicast mode: %s", iconf->iface, rte_strerror(-retval));
1733  SCReturnInt(retval);
1734  }
1735  } else if (retval < 0) {
1736  SCLogError("%s: error when changing multicast setting: %s", iconf->iface,
1737  rte_strerror(-retval));
1738  SCReturnInt(retval);
1739  }
1740 
1741  retval = iconf->flags & DPDK_PROMISC ? rte_eth_promiscuous_enable(iconf->port_id)
1742  : rte_eth_promiscuous_disable(iconf->port_id);
1743  if (retval == -ENOTSUP) {
1744  retval = rte_eth_promiscuous_get(iconf->port_id);
1745  if ((retval == 1 && !(iconf->flags & DPDK_PROMISC)) ||
1746  (retval == 0 && (iconf->flags & DPDK_PROMISC))) {
1747  SCLogError("%s: cannot configure promiscuous mode, the port is in %spromiscuous mode",
1748  iconf->iface, retval == 1 ? "" : "non-");
1750  } else if (retval < 0) {
1751  SCLogError(
1752  "%s: failed to get promiscuous mode: %s", iconf->iface, rte_strerror(-retval));
1753  SCReturnInt(retval);
1754  }
1755  } else if (retval < 0) {
1756  SCLogError("%s: error when changing promiscuous setting: %s", iconf->iface,
1757  rte_strerror(-retval));
1759  }
1760 
1761  // set maximum transmission unit
1762  SCLogConfig("%s: setting MTU to %d", iconf->iface, iconf->mtu);
1763  retval = rte_eth_dev_set_mtu(iconf->port_id, iconf->mtu);
1764  if (retval == -ENOTSUP) {
1765  // if it is not possible to set the MTU, retrieve it
1766  retval = rte_eth_dev_get_mtu(iconf->port_id, &iconf->mtu);
1767  if (retval < 0) {
1768  SCLogError("%s: failed to retrieve MTU: %s", iconf->iface, rte_strerror(-retval));
1769  SCReturnInt(retval);
1770  }
1771  SCLogWarning(
1772  "%s: changing MTU is not supported, current MTU: %u", iconf->iface, iconf->mtu);
1773  } else if (retval < 0) {
1774  SCLogError(
1775  "%s: failed to set MTU to %u: %s", iconf->iface, iconf->mtu, rte_strerror(-retval));
1776  SCReturnInt(retval);
1777  }
1778 
1779  retval = DeviceConfigureQueues(iconf, &dev_info, &port_conf);
1780  if (retval < 0) {
1781  SCReturnInt(retval);
1782  }
1783 
1784  retval = DeviceConfigureIPS(iconf);
1785  if (retval < 0) {
1786  SCReturnInt(retval);
1787  }
1788 
1789  SCReturnInt(0);
1790 }
1791 
1792 static void *ParseDpdkConfigAndConfigureDevice(const char *iface)
1793 {
1794  int retval;
1795  DPDKIfaceConfig *iconf = ConfigParse(iface);
1796  if (iconf == NULL) {
1797  FatalError("DPDK configuration could not be parsed");
1798  }
1799 
1800  retval = DeviceConfigure(iconf);
1801  if (retval == -EAGAIN) {
1802  // for e.g. bonding PMD it needs to be reconfigured
1803  retval = DeviceConfigure(iconf);
1804  }
1805 
1806  if (retval < 0) { // handles both configure attempts
1807  iconf->DerefFunc(iconf);
1808  if (rte_eal_cleanup() != 0)
1809  FatalError("EAL cleanup failed: %s", rte_strerror(-retval));
1810 
1811  if (retval == -ENOMEM) {
1812  FatalError("%s: memory allocation failed - consider"
1813  "%s freeing up some memory.",
1814  iface,
1815  rte_eal_has_hugepages() != 0 ? " increasing the number of hugepages or" : "");
1816  } else {
1817  FatalError("%s: failed to configure", iface);
1818  }
1819  }
1820 
1821  SC_ATOMIC_RESET(iconf->ref);
1822  (void)SC_ATOMIC_ADD(iconf->ref, iconf->threads);
1823  // This counter is increased by worker threads that individually pick queue IDs.
1824  SC_ATOMIC_RESET(iconf->queue_id);
1825  SC_ATOMIC_RESET(iconf->inconsistent_numa_cnt);
1826  iconf->workers_sync = SCCalloc(1, sizeof(*iconf->workers_sync));
1827  if (iconf->workers_sync == NULL) {
1828  FatalError("Failed to allocate memory for workers_sync");
1829  }
1830  SC_ATOMIC_RESET(iconf->workers_sync->worker_checked_in);
1831  iconf->workers_sync->worker_cnt = iconf->threads;
1832 
1833  // initialize LiveDev DPDK values
1834  LiveDevice *ldev_instance = LiveGetDevice(iface);
1835  if (ldev_instance == NULL) {
1836  FatalError("Device %s is not registered as a live device", iface);
1837  }
1838  for (uint16_t i = 0; i < iconf->threads; i++) {
1839  ldev_instance->dpdk_vars = iconf->pkt_mempools;
1840  iconf->pkt_mempools = NULL;
1841  }
1842  return iconf;
1843 }
1844 
1845 /**
1846  * \brief extract information from config file
1847  *
1848  * The returned structure will be freed by the thread init function.
1849  * This is thus necessary to or copy the structure before giving it
1850  * to thread or to reparse the file for each thread (and thus have
1851  * new structure.
1852  *
1853  * After configuration is loaded, DPDK also configures the device according to the settings.
1854  *
1855  * \return a DPDKIfaceConfig corresponding to the interface name
1856  */
1857 
1858 static int DPDKConfigGetThreadsCount(void *conf)
1859 {
1860  if (conf == NULL)
1861  FatalError("Configuration file is NULL");
1862 
1863  DPDKIfaceConfig *dpdk_conf = (DPDKIfaceConfig *)conf;
1864  return dpdk_conf->threads;
1865 }
1866 
1867 #endif /* HAVE_DPDK */
1868 
1869 static int DPDKRunModeIsIPS(void)
1870 {
1871  /* Find initial node */
1872  const char dpdk_node_query[] = "dpdk.interfaces";
1873  SCConfNode *dpdk_node = SCConfGetNode(dpdk_node_query);
1874  if (dpdk_node == NULL) {
1875  FatalError("Unable to get %s configuration node", dpdk_node_query);
1876  }
1877 
1878  const char default_iface[] = "default";
1879  SCConfNode *if_default = SCConfNodeLookupKeyValue(dpdk_node, "interface", default_iface);
1880  int nlive = LiveGetDeviceCount();
1881  bool has_ips = false;
1882  bool has_ids = false;
1883  for (int ldev = 0; ldev < nlive; ldev++) {
1884  const char *live_dev = LiveGetDeviceName(ldev);
1885  if (live_dev == NULL)
1886  FatalError("Unable to get device id %d from LiveDevice list", ldev);
1887 
1888  SCConfNode *if_root = ConfFindDeviceConfig(dpdk_node, live_dev);
1889  if (if_root == NULL) {
1890  if (if_default == NULL)
1891  FatalError("Unable to get %s or %s interface", live_dev, default_iface);
1892 
1893  if_root = if_default;
1894  }
1895 
1896  const char *copymodestr = NULL;
1897  const char *copyifacestr = NULL;
1898  if (SCConfGetChildValueWithDefault(if_root, if_default, "copy-mode", &copymodestr) == 1 &&
1899  SCConfGetChildValue(if_root, "copy-iface", &copyifacestr) == 1) {
1900  if (strcmp(copymodestr, "ips") == 0) {
1901  has_ips = true;
1902  } else {
1903  has_ids = true;
1904  }
1905  } else {
1906  has_ids = true;
1907  }
1908 
1909  if (has_ids && has_ips) {
1910  FatalError("Copy-mode of interface %s mixes with the previously set copy-modes "
1911  "(only IDS/TAP and IPS copy-mode combinations are allowed in DPDK",
1912  live_dev);
1913  }
1914  }
1915 
1916  return has_ips;
1917 }
1918 
1919 static int DPDKRunModeEnableIPS(void)
1920 {
1921  int r = DPDKRunModeIsIPS();
1922  if (r == 1) {
1923  SCLogInfo("Setting IPS mode");
1924  EngineModeSetIPS();
1925  }
1926  return r;
1927 }
1928 
1929 const char *RunModeDpdkGetDefaultMode(void)
1930 {
1931  return "workers";
1932 }
1933 
1935 {
1937  "Workers DPDK mode, each thread does all"
1938  " tasks from acquisition to logging",
1939  RunModeIdsDpdkWorkers, DPDKRunModeEnableIPS);
1940 }
1941 
1942 /**
1943  * \brief Workers version of the DPDK processing.
1944  *
1945  * Start N threads with each thread doing all the work.
1946  *
1947  */
1949 {
1950  SCEnter();
1951 #ifdef HAVE_DPDK
1952  int ret;
1953 
1954  TimeModeSetLive();
1955 
1956  InitEal();
1957  ret = RunModeSetLiveCaptureWorkers(ParseDpdkConfigAndConfigureDevice, DPDKConfigGetThreadsCount,
1958  "ReceiveDPDK", "DecodeDPDK", thread_name_workers, NULL);
1959  if (ret != 0) {
1960  FatalError("Unable to start runmode");
1961  }
1962 
1963  SCLogDebug("RunModeIdsDpdkWorkers initialised");
1964 
1965 #endif /* HAVE_DPDK */
1966  SCReturnInt(0);
1967 }
1968 
1969 /**
1970  * @}
1971  */
thread_name_workers
const char * thread_name_workers
Definition: runmodes.c:68
DPDKIfaceConfigAttributes_::checksum_checks_offload
const char * checksum_checks_offload
Definition: runmode-dpdk.h:32
LiveGetDeviceCountWithoutAssignedThreading
int LiveGetDeviceCountWithoutAssignedThreading(void)
Definition: util-device.c:182
util-device-private.h
util-byte.h
DPDKIfaceConfigAttributes_::mempool_size
const char * mempool_size
Definition: runmode-dpdk.h:37
threading_set_cpu_affinity
bool threading_set_cpu_affinity
Definition: runmodes.c:62
CHECKSUM_VALIDATION_OFFLOAD
@ CHECKSUM_VALIDATION_OFFLOAD
Definition: decode.h:48
GetAffinityTypeForNameAndIface
ThreadsAffinityType * GetAffinityTypeForNameAndIface(const char *name, const char *interface_name)
Find affinity by name (*-cpu-set name) and an interface name.
Definition: util-affinity.c:138
RunModeSetLiveCaptureWorkers
int RunModeSetLiveCaptureWorkers(ConfigIfaceParserFunc ConfigParser, ConfigIfaceThreadsCountFunc ModThreadsCount, const char *recv_mod_name, const char *decode_mod_name, const char *thread_name, const char *live_dev)
Definition: util-runmodes.c:330
DPDKIfaceConfigAttributes_::promisc
const char * promisc
Definition: runmode-dpdk.h:29
SC_ATOMIC_INIT
#define SC_ATOMIC_INIT(name)
wrapper for initializing an atomic variable.
Definition: util-atomic.h:314
DPDKIfaceConfigAttributes_::mempool_cache_size
const char * mempool_cache_size
Definition: runmode-dpdk.h:38
DPDK_COPY_MODE_IPS
@ DPDK_COPY_MODE_IPS
Definition: source-dpdk.h:34
SCLogDebug
#define SCLogDebug(...)
Definition: util-debug.h:275
next
struct HtpBodyChunk_ * next
Definition: app-layer-htp.h:0
UtilAffinityGetAffinedCPUNum
uint16_t UtilAffinityGetAffinedCPUNum(ThreadsAffinityType *taf)
Return the total number of CPUs in a given affinity.
Definition: util-affinity.c:1043
LiveDevice_
Definition: util-device-private.h:32
SC_ATOMIC_ADD
#define SC_ATOMIC_ADD(name, val)
add a value to our atomic variable
Definition: util-atomic.h:332
DPDKIfaceConfigAttributes_::copy_iface
const char * copy_iface
Definition: runmode-dpdk.h:42
StringParseUint16
int StringParseUint16(uint16_t *res, int base, size_t len, const char *str)
Definition: util-byte.c:337
RunModeIdsDpdkWorkers
int RunModeIdsDpdkWorkers(void)
Workers version of the DPDK processing.
Definition: runmode-dpdk.c:1948
util-runmodes.h
TAILQ_FOREACH
#define TAILQ_FOREACH(var, head, field)
Definition: queue.h:252
MIN
#define MIN(x, y)
Definition: suricata-common.h:408
CHECKSUM_VALIDATION_DISABLE
@ CHECKSUM_VALIDATION_DISABLE
Definition: decode.h:43
RunmodeGetActive
char * RunmodeGetActive(void)
Definition: runmodes.c:199
util-dpdk-rss.h
DPDKIfaceConfig_
Definition: source-dpdk.h:53
util-dpdk-ice.h
TM_ECODE_FAILED
@ TM_ECODE_FAILED
Definition: tm-threads-common.h:82
SCConfNodeIsSequence
int SCConfNodeIsSequence(const SCConfNode *node)
Check if a node is a sequence or node.
Definition: conf.c:925
strlcpy
size_t strlcpy(char *dst, const char *src, size_t siz)
Definition: util-strlcpyu.c:43
SCReturnBool
#define SCReturnBool(x)
Definition: util-debug.h:295
CHECKSUM_VALIDATION_ENABLE
@ CHECKSUM_VALIDATION_ENABLE
Definition: decode.h:44
DPDKIfaceConfigAttributes_::linkup_timeout
const char * linkup_timeout
Definition: runmode-dpdk.h:36
RunModeDpdkGetDefaultMode
const char * RunModeDpdkGetDefaultMode(void)
Definition: runmode-dpdk.c:1929
decode.h
util-debug.h
SCConfGetChildValueIntWithDefault
int SCConfGetChildValueIntWithDefault(const SCConfNode *base, const SCConfNode *dflt, const char *name, intmax_t *val)
Definition: conf.c:476
DPDKIfaceConfigAttributes_::rx_descriptors
const char * rx_descriptors
Definition: runmode-dpdk.h:39
strlcat
size_t strlcat(char *, const char *src, size_t siz)
Definition: util-strlcatu.c:45
util-cpu.h
DPDKIfaceConfigAttributes_::checksum_checks
const char * checksum_checks
Definition: runmode-dpdk.h:31
DPDKDeviceSetSocketID
int32_t DPDKDeviceSetSocketID(uint16_t port_id, int32_t *socket_id)
Definition: util-dpdk.c:75
LiveGetDevice
LiveDevice * LiveGetDevice(const char *name)
Get a pointer to the device at idx.
Definition: util-device.c:268
DPDK_RX_CHECKSUM_OFFLOAD
#define DPDK_RX_CHECKSUM_OFFLOAD
Definition: source-dpdk.h:44
SCEnter
#define SCEnter(...)
Definition: util-debug.h:277
util-affinity.h
EngineModeSetIPS
void EngineModeSetIPS(void)
Definition: suricata.c:259
StringParseUint32
int StringParseUint32(uint32_t *res, int base, size_t len, const char *str)
Definition: util-byte.c:313
util-time.h
SCLogWarning
#define SCLogWarning(...)
Macro used to log WARNING messages.
Definition: util-debug.h:255
DPDKIfaceConfigAttributes_
Definition: runmode-dpdk.h:26
SC_ATOMIC_SUB
#define SC_ATOMIC_SUB(name, val)
sub a value from our atomic variable
Definition: util-atomic.h:341
SCReturn
#define SCReturn
Definition: util-debug.h:279
DPDKIfaceConfigAttributes_::irq_mode
const char * irq_mode
Definition: runmode-dpdk.h:28
RunModeRegisterNewRunMode
void RunModeRegisterNewRunMode(enum SCRunModes runmode, const char *name, const char *description, int(*RunModeFunc)(void), int(*RunModeIsIPSEnabled)(void))
Registers a new runmode.
Definition: runmodes.c:486
SCReturnPtr
#define SCReturnPtr(x, type)
Definition: util-debug.h:293
DPDKIfaceConfigAttributes_::vlan_strip_offload
const char * vlan_strip_offload
Definition: runmode-dpdk.h:34
BIT_U64
#define BIT_U64(n)
Definition: suricata-common.h:418
runmodes.h
SCLogInfo
#define SCLogInfo(...)
Macro used to log INFORMATIONAL messages.
Definition: util-debug.h:225
DPDKIfaceConfigAttributes_::threads
const char * threads
Definition: runmode-dpdk.h:27
TimeModeSetLive
void TimeModeSetLive(void)
Definition: util-time.c:99
DPDKIfaceConfigAttributes_::tx_descriptors
const char * tx_descriptors
Definition: runmode-dpdk.h:40
DPDK_IRQ_MODE
#define DPDK_IRQ_MODE
Definition: source-dpdk.h:42
util-dpdk.h
util-conf.h
SCConfNodeLookupKeyValue
SCConfNode * SCConfNodeLookupKeyValue(const SCConfNode *base, const char *key, const char *value)
Lookup for a key value under a specific node.
Definition: conf.c:841
SCConfGetChildValue
int SCConfGetChildValue(const SCConfNode *base, const char *name, const char **vptr)
Definition: conf.c:363
StringParseUint64
int StringParseUint64(uint64_t *res, int base, size_t len, const char *str)
Definition: util-byte.c:308
source-dpdk.h
suricata-common.h
LiveGetDeviceName
const char * LiveGetDeviceName(int number)
Get a pointer to the device name at idx.
Definition: util-device.c:204
DPDK_PROMISC
#define DPDK_PROMISC
Definition: source-dpdk.h:40
FatalError
#define FatalError(...)
Definition: util-debug.h:510
SC_ATOMIC_RESET
#define SC_ATOMIC_RESET(name)
wrapper for reinitializing an atomic variable.
Definition: util-atomic.h:323
ThreadsAffinityType_
Definition: util-affinity.h:72
SCLogConfig
struct SCLogConfig_ SCLogConfig
Holds the config state used by the logging api.
SCConfGetNode
SCConfNode * SCConfGetNode(const char *name)
Get a SCConfNode by name.
Definition: conf.c:181
SCLogError
#define SCLogError(...)
Macro used to log ERROR messages.
Definition: util-debug.h:267
SCFree
#define SCFree(p)
Definition: util-mem.h:61
RUNMODE_DPDK
@ RUNMODE_DPDK
Definition: runmodes.h:39
util-dpdk-bonding.h
DPDKIfaceConfigAttributes_::multicast
const char * multicast
Definition: runmode-dpdk.h:30
RunModeDpdkRegister
void RunModeDpdkRegister(void)
Definition: runmode-dpdk.c:1934
util-dpdk-ixgbe.h
SCConfSetRootAndDefaultNodes
int SCConfSetRootAndDefaultNodes(const char *ifaces_node_name, const char *iface, SCConfNode **if_root, SCConfNode **if_default)
Finds and sets root and default node of the interface.
Definition: conf.c:962
util-dpdk-i40e.h
suricata.h
runmode-dpdk.h
SCConfNode_::name
char * name
Definition: conf.h:38
DPDKIfaceConfigAttributes_::rss_hf
const char * rss_hf
Definition: runmode-dpdk.h:35
ConfFindDeviceConfig
SCConfNode * ConfFindDeviceConfig(SCConfNode *node, const char *iface)
Find the configuration node for a specific device.
Definition: util-conf.c:121
LiveGetDeviceCount
int LiveGetDeviceCount(void)
Get the number of registered devices.
Definition: util-device.c:170
SCConfGetChildValueWithDefault
int SCConfGetChildValueWithDefault(const SCConfNode *base, const SCConfNode *dflt, const char *name, const char **vptr)
Definition: conf.c:393
DPDKIfaceConfigAttributes_::mtu
const char * mtu
Definition: runmode-dpdk.h:33
DPDKIfaceConfigAttributes_::copy_mode
const char * copy_mode
Definition: runmode-dpdk.h:41
UtilCpuGetNumProcessorsOnline
uint16_t UtilCpuGetNumProcessorsOnline(void)
Get the number of cpus online in the system.
Definition: util-cpu.c:108
SCConfGetChildValueBoolWithDefault
int SCConfGetChildValueBoolWithDefault(const SCConfNode *base, const SCConfNode *dflt, const char *name, int *val)
Definition: conf.c:528
util-dpdk-common.h
SCCalloc
#define SCCalloc(nm, sz)
Definition: util-mem.h:53
SCReturnInt
#define SCReturnInt(x)
Definition: util-debug.h:281
SCConfNode_
Definition: conf.h:37
SCConfNode_::val
char * val
Definition: conf.h:39
DPDK_COPY_MODE_TAP
@ DPDK_COPY_MODE_TAP
Definition: source-dpdk.h:34
DPDK_COPY_MODE_NONE
@ DPDK_COPY_MODE_NONE
Definition: source-dpdk.h:34
DPDK_MULTICAST
#define DPDK_MULTICAST
Definition: source-dpdk.h:41