1
2
3
4
5
6#include "test_pipeline_common.h"
7
8
9
10static __rte_always_inline int
11pipeline_atq_nb_event_queues(struct evt_options *opt)
12{
13 RTE_SET_USED(opt);
14
15 return rte_eth_dev_count_avail();
16}
17
18typedef int (*pipeline_atq_worker_t)(void *arg);
19
20static __rte_noinline int
21pipeline_atq_worker_single_stage_tx(void *arg)
22{
23 PIPELINE_WORKER_SINGLE_STAGE_INIT;
24 uint8_t enq = 0, deq = 0;
25
26 while (t->done == false) {
27 deq = rte_event_dequeue_burst(dev, port, &ev, 1, 0);
28
29 if (!deq) {
30 rte_pause();
31 continue;
32 }
33
34 deq = pipeline_event_tx(dev, port, &ev, t);
35 w->processed_pkts++;
36 }
37 pipeline_worker_cleanup(dev, port, &ev, enq, deq);
38
39 return 0;
40}
41
42static __rte_noinline int
43pipeline_atq_worker_single_stage_fwd(void *arg)
44{
45 PIPELINE_WORKER_SINGLE_STAGE_INIT;
46 const uint8_t *tx_queue = t->tx_evqueue_id;
47 uint8_t enq = 0, deq = 0;
48
49 while (t->done == false) {
50 deq = rte_event_dequeue_burst(dev, port, &ev, 1, 0);
51
52 if (!deq) {
53 rte_pause();
54 continue;
55 }
56
57 ev.queue_id = tx_queue[ev.mbuf->port];
58 pipeline_fwd_event(&ev, RTE_SCHED_TYPE_ATOMIC);
59 enq = pipeline_event_enqueue(dev, port, &ev, t);
60 w->processed_pkts++;
61 }
62 pipeline_worker_cleanup(dev, port, &ev, enq, deq);
63
64 return 0;
65}
66
67static __rte_noinline int
68pipeline_atq_worker_single_stage_burst_tx(void *arg)
69{
70 PIPELINE_WORKER_SINGLE_STAGE_BURST_INIT;
71 uint16_t nb_rx = 0, nb_tx = 0;
72
73 while (t->done == false) {
74 nb_rx = rte_event_dequeue_burst(dev, port, ev, BURST_SIZE, 0);
75
76 if (!nb_rx) {
77 rte_pause();
78 continue;
79 }
80
81 for (i = 0; i < nb_rx; i++) {
82 rte_prefetch0(ev[i + 1].mbuf);
83 rte_event_eth_tx_adapter_txq_set(ev[i].mbuf, 0);
84 }
85
86 nb_tx = pipeline_event_tx_burst(dev, port, ev, nb_rx, t);
87 w->processed_pkts += nb_tx;
88 }
89 pipeline_worker_cleanup(dev, port, ev, nb_tx, nb_rx);
90
91 return 0;
92}
93
94static __rte_noinline int
95pipeline_atq_worker_single_stage_burst_fwd(void *arg)
96{
97 PIPELINE_WORKER_SINGLE_STAGE_BURST_INIT;
98 const uint8_t *tx_queue = t->tx_evqueue_id;
99 uint16_t nb_rx = 0, nb_tx = 0;
100
101 while (t->done == false) {
102 nb_rx = rte_event_dequeue_burst(dev, port, ev, BURST_SIZE, 0);
103
104 if (!nb_rx) {
105 rte_pause();
106 continue;
107 }
108
109 for (i = 0; i < nb_rx; i++) {
110 rte_prefetch0(ev[i + 1].mbuf);
111 rte_event_eth_tx_adapter_txq_set(ev[i].mbuf, 0);
112 ev[i].queue_id = tx_queue[ev[i].mbuf->port];
113 pipeline_fwd_event(&ev[i], RTE_SCHED_TYPE_ATOMIC);
114 }
115
116 nb_tx = pipeline_event_enqueue_burst(dev, port, ev, nb_rx, t);
117 w->processed_pkts += nb_tx;
118 }
119 pipeline_worker_cleanup(dev, port, ev, nb_tx, nb_rx);
120
121 return 0;
122}
123
124static __rte_noinline int
125pipeline_atq_worker_single_stage_tx_vector(void *arg)
126{
127 PIPELINE_WORKER_SINGLE_STAGE_INIT;
128 uint8_t enq = 0, deq = 0;
129 uint16_t vector_sz;
130
131 while (!t->done) {
132 deq = rte_event_dequeue_burst(dev, port, &ev, 1, 0);
133
134 if (!deq) {
135 rte_pause();
136 continue;
137 }
138 vector_sz = ev.vec->nb_elem;
139 enq = pipeline_event_tx_vector(dev, port, &ev, t);
140 w->processed_pkts += vector_sz;
141 }
142 pipeline_worker_cleanup(dev, port, &ev, enq, deq);
143
144 return 0;
145}
146
147static __rte_noinline int
148pipeline_atq_worker_single_stage_fwd_vector(void *arg)
149{
150 PIPELINE_WORKER_SINGLE_STAGE_INIT;
151 const uint8_t *tx_queue = t->tx_evqueue_id;
152 uint8_t enq = 0, deq = 0;
153 uint16_t vector_sz;
154
155 while (!t->done) {
156 deq = rte_event_dequeue_burst(dev, port, &ev, 1, 0);
157
158 if (!deq) {
159 rte_pause();
160 continue;
161 }
162
163 vector_sz = ev.vec->nb_elem;
164 ev.queue_id = tx_queue[ev.vec->port];
165 ev.vec->queue = 0;
166 pipeline_fwd_event_vector(&ev, RTE_SCHED_TYPE_ATOMIC);
167 enq = pipeline_event_enqueue(dev, port, &ev, t);
168 w->processed_pkts += vector_sz;
169 }
170 pipeline_worker_cleanup(dev, port, &ev, enq, deq);
171
172 return 0;
173}
174
175static __rte_noinline int
176pipeline_atq_worker_single_stage_burst_tx_vector(void *arg)
177{
178 PIPELINE_WORKER_SINGLE_STAGE_BURST_INIT;
179 uint16_t nb_rx = 0, nb_tx = 0;
180 uint16_t vector_sz;
181
182 while (!t->done) {
183 nb_rx = rte_event_dequeue_burst(dev, port, ev, BURST_SIZE, 0);
184
185 if (!nb_rx) {
186 rte_pause();
187 continue;
188 }
189 vector_sz = 0;
190 for (i = 0; i < nb_rx; i++) {
191 vector_sz += ev[i].vec->nb_elem;
192 ev[i].vec->queue = 0;
193 }
194
195 nb_tx = pipeline_event_tx_burst(dev, port, ev, nb_rx, t);
196 w->processed_pkts += vector_sz;
197 }
198 pipeline_worker_cleanup(dev, port, ev, nb_tx, nb_rx);
199
200 return 0;
201}
202
203static __rte_noinline int
204pipeline_atq_worker_single_stage_burst_fwd_vector(void *arg)
205{
206 PIPELINE_WORKER_SINGLE_STAGE_BURST_INIT;
207 const uint8_t *tx_queue = t->tx_evqueue_id;
208 uint16_t nb_rx = 0, nb_tx = 0;
209 uint16_t vector_sz;
210
211 while (!t->done) {
212 nb_rx = rte_event_dequeue_burst(dev, port, ev, BURST_SIZE, 0);
213
214 if (!nb_rx) {
215 rte_pause();
216 continue;
217 }
218
219 vector_sz = 0;
220 for (i = 0; i < nb_rx; i++) {
221 ev[i].queue_id = tx_queue[ev[i].vec->port];
222 ev[i].vec->queue = 0;
223 vector_sz += ev[i].vec->nb_elem;
224 pipeline_fwd_event_vector(&ev[i],
225 RTE_SCHED_TYPE_ATOMIC);
226 }
227
228 nb_tx = pipeline_event_enqueue_burst(dev, port, ev, nb_rx, t);
229 w->processed_pkts += vector_sz;
230 }
231 pipeline_worker_cleanup(dev, port, ev, nb_tx, nb_rx);
232
233 return 0;
234}
235
236static __rte_noinline int
237pipeline_atq_worker_multi_stage_tx(void *arg)
238{
239 PIPELINE_WORKER_MULTI_STAGE_INIT;
240 uint8_t enq = 0, deq = 0;
241
242 while (t->done == false) {
243 deq = rte_event_dequeue_burst(dev, port, &ev, 1, 0);
244
245 if (!deq) {
246 rte_pause();
247 continue;
248 }
249
250 cq_id = ev.sub_event_type % nb_stages;
251
252 if (cq_id == last_queue) {
253 enq = pipeline_event_tx(dev, port, &ev, t);
254 w->processed_pkts++;
255 continue;
256 }
257
258 ev.sub_event_type++;
259 pipeline_fwd_event(&ev, sched_type_list[cq_id]);
260 enq = pipeline_event_enqueue(dev, port, &ev, t);
261 }
262 pipeline_worker_cleanup(dev, port, &ev, enq, deq);
263
264 return 0;
265}
266
267static __rte_noinline int
268pipeline_atq_worker_multi_stage_fwd(void *arg)
269{
270 PIPELINE_WORKER_MULTI_STAGE_INIT;
271 const uint8_t *tx_queue = t->tx_evqueue_id;
272 uint8_t enq = 0, deq = 0;
273
274 while (t->done == false) {
275 deq = rte_event_dequeue_burst(dev, port, &ev, 1, 0);
276
277 if (!deq) {
278 rte_pause();
279 continue;
280 }
281
282 cq_id = ev.sub_event_type % nb_stages;
283
284 if (cq_id == last_queue) {
285 ev.queue_id = tx_queue[ev.mbuf->port];
286 pipeline_fwd_event(&ev, RTE_SCHED_TYPE_ATOMIC);
287 w->processed_pkts++;
288 } else {
289 ev.sub_event_type++;
290 pipeline_fwd_event(&ev, sched_type_list[cq_id]);
291 }
292
293 enq = pipeline_event_enqueue(dev, port, &ev, t);
294 }
295 pipeline_worker_cleanup(dev, port, &ev, enq, deq);
296
297 return 0;
298}
299
300static __rte_noinline int
301pipeline_atq_worker_multi_stage_burst_tx(void *arg)
302{
303 PIPELINE_WORKER_MULTI_STAGE_BURST_INIT;
304 uint16_t nb_rx = 0, nb_tx = 0;
305
306 while (t->done == false) {
307 nb_rx = rte_event_dequeue_burst(dev, port, ev, BURST_SIZE, 0);
308
309 if (!nb_rx) {
310 rte_pause();
311 continue;
312 }
313
314 for (i = 0; i < nb_rx; i++) {
315 rte_prefetch0(ev[i + 1].mbuf);
316 cq_id = ev[i].sub_event_type % nb_stages;
317
318 if (cq_id == last_queue) {
319 pipeline_event_tx(dev, port, &ev[i], t);
320 ev[i].op = RTE_EVENT_OP_RELEASE;
321 w->processed_pkts++;
322 continue;
323 }
324
325 ev[i].sub_event_type++;
326 pipeline_fwd_event(&ev[i], sched_type_list[cq_id]);
327 }
328
329 nb_tx = pipeline_event_enqueue_burst(dev, port, ev, nb_rx, t);
330 }
331 pipeline_worker_cleanup(dev, port, ev, nb_tx, nb_rx);
332
333 return 0;
334}
335
336static __rte_noinline int
337pipeline_atq_worker_multi_stage_burst_fwd(void *arg)
338{
339 PIPELINE_WORKER_MULTI_STAGE_BURST_INIT;
340 const uint8_t *tx_queue = t->tx_evqueue_id;
341 uint16_t nb_rx = 0, nb_tx = 0;
342
343 while (t->done == false) {
344 nb_rx = rte_event_dequeue_burst(dev, port, ev, BURST_SIZE, 0);
345
346 if (!nb_rx) {
347 rte_pause();
348 continue;
349 }
350
351 for (i = 0; i < nb_rx; i++) {
352 rte_prefetch0(ev[i + 1].mbuf);
353 cq_id = ev[i].sub_event_type % nb_stages;
354
355 if (cq_id == last_queue) {
356 w->processed_pkts++;
357 ev[i].queue_id = tx_queue[ev[i].mbuf->port];
358 pipeline_fwd_event(&ev[i],
359 RTE_SCHED_TYPE_ATOMIC);
360 } else {
361 ev[i].sub_event_type++;
362 pipeline_fwd_event(&ev[i],
363 sched_type_list[cq_id]);
364 }
365 }
366
367 nb_tx = pipeline_event_enqueue_burst(dev, port, ev, nb_rx, t);
368 }
369 pipeline_worker_cleanup(dev, port, ev, nb_tx, nb_rx);
370
371 return 0;
372}
373
374static __rte_noinline int
375pipeline_atq_worker_multi_stage_tx_vector(void *arg)
376{
377 PIPELINE_WORKER_MULTI_STAGE_INIT;
378 uint8_t enq = 0, deq = 0;
379 uint16_t vector_sz;
380
381 while (!t->done) {
382 deq = rte_event_dequeue_burst(dev, port, &ev, 1, 0);
383
384 if (!deq) {
385 rte_pause();
386 continue;
387 }
388
389 cq_id = ev.sub_event_type % nb_stages;
390
391 if (cq_id == last_queue) {
392 vector_sz = ev.vec->nb_elem;
393 enq = pipeline_event_tx_vector(dev, port, &ev, t);
394 w->processed_pkts += vector_sz;
395 continue;
396 }
397
398 ev.sub_event_type++;
399 pipeline_fwd_event_vector(&ev, sched_type_list[cq_id]);
400 enq = pipeline_event_enqueue(dev, port, &ev, t);
401 }
402 pipeline_worker_cleanup(dev, port, &ev, enq, deq);
403
404 return 0;
405}
406
407static __rte_noinline int
408pipeline_atq_worker_multi_stage_fwd_vector(void *arg)
409{
410 PIPELINE_WORKER_MULTI_STAGE_INIT;
411 const uint8_t *tx_queue = t->tx_evqueue_id;
412 uint8_t enq = 0, deq = 0;
413 uint16_t vector_sz;
414
415 while (!t->done) {
416 deq = rte_event_dequeue_burst(dev, port, &ev, 1, 0);
417
418 if (!deq) {
419 rte_pause();
420 continue;
421 }
422
423 cq_id = ev.sub_event_type % nb_stages;
424
425 if (cq_id == last_queue) {
426 ev.queue_id = tx_queue[ev.vec->port];
427 ev.vec->queue = 0;
428 vector_sz = ev.vec->nb_elem;
429 pipeline_fwd_event_vector(&ev, RTE_SCHED_TYPE_ATOMIC);
430 enq = pipeline_event_enqueue(dev, port, &ev, t);
431 w->processed_pkts += vector_sz;
432 } else {
433 ev.sub_event_type++;
434 pipeline_fwd_event_vector(&ev, sched_type_list[cq_id]);
435 enq = pipeline_event_enqueue(dev, port, &ev, t);
436 }
437 }
438 pipeline_worker_cleanup(dev, port, &ev, enq, deq);
439
440 return 0;
441}
442
443static __rte_noinline int
444pipeline_atq_worker_multi_stage_burst_tx_vector(void *arg)
445{
446 PIPELINE_WORKER_MULTI_STAGE_BURST_INIT;
447 uint16_t nb_rx = 0, nb_tx = 0;
448 uint16_t vector_sz;
449
450 while (!t->done) {
451 nb_rx = rte_event_dequeue_burst(dev, port, ev, BURST_SIZE, 0);
452
453 if (!nb_rx) {
454 rte_pause();
455 continue;
456 }
457
458 for (i = 0; i < nb_rx; i++) {
459 cq_id = ev[i].sub_event_type % nb_stages;
460
461 if (cq_id == last_queue) {
462 vector_sz = ev[i].vec->nb_elem;
463 pipeline_event_tx_vector(dev, port, &ev[i], t);
464 ev[i].op = RTE_EVENT_OP_RELEASE;
465 w->processed_pkts += vector_sz;
466 continue;
467 }
468
469 ev[i].sub_event_type++;
470 pipeline_fwd_event_vector(&ev[i],
471 sched_type_list[cq_id]);
472 }
473
474 nb_tx = pipeline_event_enqueue_burst(dev, port, ev, nb_rx, t);
475 }
476 pipeline_worker_cleanup(dev, port, ev, nb_tx, nb_rx);
477
478 return 0;
479}
480
481static __rte_noinline int
482pipeline_atq_worker_multi_stage_burst_fwd_vector(void *arg)
483{
484 PIPELINE_WORKER_MULTI_STAGE_BURST_INIT;
485 const uint8_t *tx_queue = t->tx_evqueue_id;
486 uint16_t nb_rx = 0, nb_tx = 0;
487 uint16_t vector_sz;
488
489 while (!t->done) {
490 nb_rx = rte_event_dequeue_burst(dev, port, ev, BURST_SIZE, 0);
491
492 if (!nb_rx) {
493 rte_pause();
494 continue;
495 }
496
497 for (i = 0; i < nb_rx; i++) {
498 cq_id = ev[i].sub_event_type % nb_stages;
499
500 if (cq_id == last_queue) {
501 vector_sz = ev[i].vec->nb_elem;
502 ev[i].queue_id = tx_queue[ev[i].vec->port];
503 ev[i].vec->queue = 0;
504 pipeline_fwd_event_vector(
505 &ev[i], RTE_SCHED_TYPE_ATOMIC);
506 w->processed_pkts += vector_sz;
507 } else {
508 ev[i].sub_event_type++;
509 pipeline_fwd_event_vector(
510 &ev[i], sched_type_list[cq_id]);
511 }
512 }
513
514 nb_tx = pipeline_event_enqueue_burst(dev, port, ev, nb_rx, t);
515 }
516 pipeline_worker_cleanup(dev, port, ev, nb_tx, nb_rx);
517
518 return 0;
519}
520
521static int
522worker_wrapper(void *arg)
523{
524 struct worker_data *w = arg;
525 struct evt_options *opt = w->t->opt;
526 const bool burst = evt_has_burst_mode(w->dev_id);
527 const bool internal_port = w->t->internal_port;
528 const uint8_t nb_stages = opt->nb_stages;
529
530 const pipeline_atq_worker_t
531 pipeline_atq_worker_single_stage[2][2][2] = {
532 [0][0][0] = pipeline_atq_worker_single_stage_fwd,
533 [0][0][1] = pipeline_atq_worker_single_stage_tx,
534 [0][1][0] = pipeline_atq_worker_single_stage_burst_fwd,
535 [0][1][1] = pipeline_atq_worker_single_stage_burst_tx,
536 [1][0][0] = pipeline_atq_worker_single_stage_fwd_vector,
537 [1][0][1] = pipeline_atq_worker_single_stage_tx_vector,
538 [1][1][0] = pipeline_atq_worker_single_stage_burst_fwd_vector,
539 [1][1][1] = pipeline_atq_worker_single_stage_burst_tx_vector,
540 };
541 const pipeline_atq_worker_t
542 pipeline_atq_worker_multi_stage[2][2][2] = {
543 [0][0][0] = pipeline_atq_worker_multi_stage_fwd,
544 [0][0][1] = pipeline_atq_worker_multi_stage_tx,
545 [0][1][0] = pipeline_atq_worker_multi_stage_burst_fwd,
546 [0][1][1] = pipeline_atq_worker_multi_stage_burst_tx,
547 [1][0][0] = pipeline_atq_worker_multi_stage_fwd_vector,
548 [1][0][1] = pipeline_atq_worker_multi_stage_tx_vector,
549 [1][1][0] = pipeline_atq_worker_multi_stage_burst_fwd_vector,
550 [1][1][1] = pipeline_atq_worker_multi_stage_burst_tx_vector,
551 };
552
553 if (nb_stages == 1)
554 return (pipeline_atq_worker_single_stage[opt->ena_vector][burst]
555 [internal_port])(arg);
556 else
557 return (pipeline_atq_worker_multi_stage[opt->ena_vector][burst]
558 [internal_port])(arg);
559
560 rte_panic("invalid worker\n");
561}
562
563static int
564pipeline_atq_launch_lcores(struct evt_test *test, struct evt_options *opt)
565{
566 return pipeline_launch_lcores(test, opt, worker_wrapper);
567}
568
569static int
570pipeline_atq_eventdev_setup(struct evt_test *test, struct evt_options *opt)
571{
572 int ret;
573 int nb_ports;
574 int nb_queues;
575 uint8_t queue, is_prod;
576 uint8_t tx_evqueue_id[RTE_MAX_ETHPORTS];
577 uint8_t queue_arr[RTE_EVENT_MAX_QUEUES_PER_DEV];
578 uint8_t nb_worker_queues = 0;
579 uint8_t tx_evport_id = 0;
580 uint16_t prod = 0;
581 struct rte_event_dev_info info;
582 struct test_pipeline *t = evt_test_priv(test);
583
584 nb_ports = evt_nr_active_lcores(opt->wlcores);
585 nb_queues = rte_eth_dev_count_avail();
586
587 memset(tx_evqueue_id, 0, sizeof(uint8_t) * RTE_MAX_ETHPORTS);
588 memset(queue_arr, 0, sizeof(uint8_t) * RTE_EVENT_MAX_QUEUES_PER_DEV);
589
590 if (!t->internal_port) {
591 RTE_ETH_FOREACH_DEV(prod) {
592 tx_evqueue_id[prod] = nb_queues;
593 nb_queues++;
594 }
595 }
596
597 rte_event_dev_info_get(opt->dev_id, &info);
598
599 ret = evt_configure_eventdev(opt, nb_queues, nb_ports);
600 if (ret) {
601 evt_err("failed to configure eventdev %d", opt->dev_id);
602 return ret;
603 }
604
605 struct rte_event_queue_conf q_conf = {
606 .priority = RTE_EVENT_DEV_PRIORITY_NORMAL,
607 .nb_atomic_flows = opt->nb_flows,
608 .nb_atomic_order_sequences = opt->nb_flows,
609 };
610
611 for (queue = 0; queue < nb_queues; queue++) {
612 q_conf.event_queue_cfg = RTE_EVENT_QUEUE_CFG_ALL_TYPES;
613
614 if (!t->internal_port) {
615 is_prod = false;
616 RTE_ETH_FOREACH_DEV(prod) {
617 if (queue == tx_evqueue_id[prod]) {
618 q_conf.event_queue_cfg =
619 RTE_EVENT_QUEUE_CFG_SINGLE_LINK;
620 is_prod = true;
621 break;
622 }
623 }
624 if (!is_prod) {
625 queue_arr[nb_worker_queues] = queue;
626 nb_worker_queues++;
627 }
628 }
629
630 ret = rte_event_queue_setup(opt->dev_id, queue, &q_conf);
631 if (ret) {
632 evt_err("failed to setup queue=%d", queue);
633 return ret;
634 }
635 }
636
637 if (opt->wkr_deq_dep > info.max_event_port_dequeue_depth)
638 opt->wkr_deq_dep = info.max_event_port_dequeue_depth;
639
640
641 const struct rte_event_port_conf p_conf = {
642 .dequeue_depth = opt->wkr_deq_dep,
643 .enqueue_depth = info.max_event_port_dequeue_depth,
644 .new_event_threshold = info.max_num_events,
645 };
646
647 if (!t->internal_port)
648 ret = pipeline_event_port_setup(test, opt, queue_arr,
649 nb_worker_queues, p_conf);
650 else
651 ret = pipeline_event_port_setup(test, opt, NULL, nb_queues,
652 p_conf);
653
654 if (ret)
655 return ret;
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684 ret = pipeline_event_rx_adapter_setup(opt, 1, p_conf);
685 if (ret)
686 return ret;
687 ret = pipeline_event_tx_adapter_setup(opt, p_conf);
688 if (ret)
689 return ret;
690
691 if (!evt_has_distributed_sched(opt->dev_id)) {
692 uint32_t service_id;
693 rte_event_dev_service_id_get(opt->dev_id, &service_id);
694 ret = evt_service_setup(service_id);
695 if (ret) {
696 evt_err("No service lcore found to run event dev.");
697 return ret;
698 }
699 }
700
701
702 if (!t->internal_port) {
703 RTE_ETH_FOREACH_DEV(prod) {
704 ret = rte_event_eth_tx_adapter_event_port_get(prod,
705 &tx_evport_id);
706 if (ret) {
707 evt_err("Unable to get Tx adapter[%d]", prod);
708 return ret;
709 }
710
711 if (rte_event_port_link(opt->dev_id, tx_evport_id,
712 &tx_evqueue_id[prod],
713 NULL, 1) != 1) {
714 evt_err("Unable to link Tx adptr[%d] evprt[%d]",
715 prod, tx_evport_id);
716 return ret;
717 }
718 }
719 }
720
721 ret = rte_event_dev_start(opt->dev_id);
722 if (ret) {
723 evt_err("failed to start eventdev %d", opt->dev_id);
724 return ret;
725 }
726
727
728 RTE_ETH_FOREACH_DEV(prod) {
729 ret = rte_eth_dev_start(prod);
730 if (ret) {
731 evt_err("Ethernet dev [%d] failed to start."
732 " Using synthetic producer", prod);
733 return ret;
734 }
735 }
736
737 RTE_ETH_FOREACH_DEV(prod) {
738 ret = rte_event_eth_rx_adapter_start(prod);
739 if (ret) {
740 evt_err("Rx adapter[%d] start failed", prod);
741 return ret;
742 }
743
744 ret = rte_event_eth_tx_adapter_start(prod);
745 if (ret) {
746 evt_err("Tx adapter[%d] start failed", prod);
747 return ret;
748 }
749 }
750
751 memcpy(t->tx_evqueue_id, tx_evqueue_id, sizeof(uint8_t) *
752 RTE_MAX_ETHPORTS);
753
754 return 0;
755}
756
757static void
758pipeline_atq_opt_dump(struct evt_options *opt)
759{
760 pipeline_opt_dump(opt, pipeline_atq_nb_event_queues(opt));
761}
762
763static int
764pipeline_atq_opt_check(struct evt_options *opt)
765{
766 return pipeline_opt_check(opt, pipeline_atq_nb_event_queues(opt));
767}
768
769static bool
770pipeline_atq_capability_check(struct evt_options *opt)
771{
772 struct rte_event_dev_info dev_info;
773
774 rte_event_dev_info_get(opt->dev_id, &dev_info);
775 if (dev_info.max_event_queues < pipeline_atq_nb_event_queues(opt) ||
776 dev_info.max_event_ports <
777 evt_nr_active_lcores(opt->wlcores)) {
778 evt_err("not enough eventdev queues=%d/%d or ports=%d/%d",
779 pipeline_atq_nb_event_queues(opt),
780 dev_info.max_event_queues,
781 evt_nr_active_lcores(opt->wlcores),
782 dev_info.max_event_ports);
783 }
784 if (!evt_has_all_types_queue(opt->dev_id))
785 return false;
786
787 return true;
788}
789
790static const struct evt_test_ops pipeline_atq = {
791 .cap_check = pipeline_atq_capability_check,
792 .opt_check = pipeline_atq_opt_check,
793 .opt_dump = pipeline_atq_opt_dump,
794 .test_setup = pipeline_test_setup,
795 .mempool_setup = pipeline_mempool_setup,
796 .ethdev_setup = pipeline_ethdev_setup,
797 .eventdev_setup = pipeline_atq_eventdev_setup,
798 .launch_lcores = pipeline_atq_launch_lcores,
799 .ethdev_rx_stop = pipeline_ethdev_rx_stop,
800 .eventdev_destroy = pipeline_eventdev_destroy,
801 .mempool_destroy = pipeline_mempool_destroy,
802 .ethdev_destroy = pipeline_ethdev_destroy,
803 .test_result = pipeline_test_result,
804 .test_destroy = pipeline_test_destroy,
805};
806
807EVT_TEST_REGISTER(pipeline_atq);
808
