xref: /dpdk/drivers/net/octeontx/octeontx_ethdev.c (revision 1bb4a528c41f4af4847bd3d58cc2b2b9f1ec9a27)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2017 Cavium, Inc
3  */
4 
5 #include <stdio.h>
6 #include <stdarg.h>
7 #include <stdbool.h>
8 #include <stdint.h>
9 #include <string.h>
10 #include <unistd.h>
11 
12 #include <rte_alarm.h>
13 #include <rte_branch_prediction.h>
14 #include <rte_bus_vdev.h>
15 #include <rte_cycles.h>
16 #include <rte_debug.h>
17 #include <rte_devargs.h>
18 #include <rte_dev.h>
19 #include <rte_kvargs.h>
20 #include <rte_malloc.h>
21 #include <rte_mbuf_pool_ops.h>
22 #include <rte_prefetch.h>
23 
24 #include "octeontx_ethdev.h"
25 #include "octeontx_rxtx.h"
26 #include "octeontx_logs.h"
27 
28 struct evdev_priv_data {
29 	OFFLOAD_FLAGS; /*Sequence should not be changed */
30 } __rte_cache_aligned;
31 
32 struct octeontx_vdev_init_params {
33 	uint8_t	nr_port;
34 };
35 
36 uint16_t
37 rte_octeontx_pchan_map[OCTEONTX_MAX_BGX_PORTS][OCTEONTX_MAX_LMAC_PER_BGX];
38 
39 enum octeontx_link_speed {
40 	OCTEONTX_LINK_SPEED_SGMII,
41 	OCTEONTX_LINK_SPEED_XAUI,
42 	OCTEONTX_LINK_SPEED_RXAUI,
43 	OCTEONTX_LINK_SPEED_10G_R,
44 	OCTEONTX_LINK_SPEED_40G_R,
45 	OCTEONTX_LINK_SPEED_RESERVE1,
46 	OCTEONTX_LINK_SPEED_QSGMII,
47 	OCTEONTX_LINK_SPEED_RESERVE2
48 };
49 
50 RTE_LOG_REGISTER_SUFFIX(otx_net_logtype_mbox, mbox, NOTICE);
51 RTE_LOG_REGISTER_SUFFIX(otx_net_logtype_init, init, NOTICE);
52 RTE_LOG_REGISTER_SUFFIX(otx_net_logtype_driver, driver, NOTICE);
53 
54 /* Parse integer from integer argument */
55 static int
56 parse_integer_arg(const char *key __rte_unused,
57 		const char *value, void *extra_args)
58 {
59 	int *i = (int *)extra_args;
60 
61 	*i = atoi(value);
62 	if (*i < 0) {
63 		octeontx_log_err("argument has to be positive.");
64 		return -1;
65 	}
66 
67 	return 0;
68 }
69 
70 static int
71 octeontx_parse_vdev_init_params(struct octeontx_vdev_init_params *params,
72 				struct rte_vdev_device *dev)
73 {
74 	struct rte_kvargs *kvlist = NULL;
75 	int ret = 0;
76 
77 	static const char * const octeontx_vdev_valid_params[] = {
78 		OCTEONTX_VDEV_NR_PORT_ARG,
79 		NULL
80 	};
81 
82 	const char *input_args = rte_vdev_device_args(dev);
83 	if (params == NULL)
84 		return -EINVAL;
85 
86 
87 	if (input_args) {
88 		kvlist = rte_kvargs_parse(input_args,
89 				octeontx_vdev_valid_params);
90 		if (kvlist == NULL)
91 			return -1;
92 
93 		ret = rte_kvargs_process(kvlist,
94 					OCTEONTX_VDEV_NR_PORT_ARG,
95 					&parse_integer_arg,
96 					&params->nr_port);
97 		if (ret < 0)
98 			goto free_kvlist;
99 	}
100 
101 free_kvlist:
102 	rte_kvargs_free(kvlist);
103 	return ret;
104 }
105 
106 static int
107 octeontx_port_open(struct octeontx_nic *nic)
108 {
109 	octeontx_mbox_bgx_port_conf_t bgx_port_conf;
110 	octeontx_mbox_bgx_port_fifo_cfg_t fifo_cfg;
111 	int res;
112 
113 	res = 0;
114 	memset(&bgx_port_conf, 0x0, sizeof(bgx_port_conf));
115 	PMD_INIT_FUNC_TRACE();
116 
117 	res = octeontx_bgx_port_open(nic->port_id, &bgx_port_conf);
118 	if (res < 0) {
119 		octeontx_log_err("failed to open port %d", res);
120 		return res;
121 	}
122 
123 	nic->node = bgx_port_conf.node;
124 	nic->port_ena = bgx_port_conf.enable;
125 	nic->base_ichan = bgx_port_conf.base_chan;
126 	nic->base_ochan = bgx_port_conf.base_chan;
127 	nic->num_ichans = bgx_port_conf.num_chans;
128 	nic->num_ochans = bgx_port_conf.num_chans;
129 	nic->bgx_mtu = bgx_port_conf.mtu;
130 	nic->bpen = bgx_port_conf.bpen;
131 	nic->fcs_strip = bgx_port_conf.fcs_strip;
132 	nic->bcast_mode = bgx_port_conf.bcast_mode;
133 	nic->mcast_mode = bgx_port_conf.mcast_mode;
134 	nic->speed	= bgx_port_conf.mode;
135 
136 	memset(&fifo_cfg, 0x0, sizeof(fifo_cfg));
137 
138 	res = octeontx_bgx_port_get_fifo_cfg(nic->port_id, &fifo_cfg);
139 	if (res < 0) {
140 		octeontx_log_err("failed to get port %d fifo cfg", res);
141 		return res;
142 	}
143 
144 	nic->fc.rx_fifosz = fifo_cfg.rx_fifosz;
145 
146 	memcpy(&nic->mac_addr[0], &bgx_port_conf.macaddr[0],
147 		RTE_ETHER_ADDR_LEN);
148 
149 	octeontx_log_dbg("port opened %d", nic->port_id);
150 	return res;
151 }
152 
153 static void
154 octeontx_link_status_print(struct rte_eth_dev *eth_dev,
155 			   struct rte_eth_link *link)
156 {
157 	if (link && link->link_status)
158 		octeontx_log_info("Port %u: Link Up - speed %u Mbps - %s",
159 			  (eth_dev->data->port_id),
160 			  link->link_speed,
161 			  link->link_duplex == ETH_LINK_FULL_DUPLEX ?
162 			  "full-duplex" : "half-duplex");
163 	else
164 		octeontx_log_info("Port %d: Link Down",
165 				  (int)(eth_dev->data->port_id));
166 }
167 
168 static void
169 octeontx_link_status_update(struct octeontx_nic *nic,
170 			 struct rte_eth_link *link)
171 {
172 	memset(link, 0, sizeof(*link));
173 
174 	link->link_status = nic->link_up ? ETH_LINK_UP : ETH_LINK_DOWN;
175 
176 	switch (nic->speed) {
177 	case OCTEONTX_LINK_SPEED_SGMII:
178 		link->link_speed = ETH_SPEED_NUM_1G;
179 		break;
180 
181 	case OCTEONTX_LINK_SPEED_XAUI:
182 		link->link_speed = ETH_SPEED_NUM_10G;
183 		break;
184 
185 	case OCTEONTX_LINK_SPEED_RXAUI:
186 	case OCTEONTX_LINK_SPEED_10G_R:
187 		link->link_speed = ETH_SPEED_NUM_10G;
188 		break;
189 	case OCTEONTX_LINK_SPEED_QSGMII:
190 		link->link_speed = ETH_SPEED_NUM_5G;
191 		break;
192 	case OCTEONTX_LINK_SPEED_40G_R:
193 		link->link_speed = ETH_SPEED_NUM_40G;
194 		break;
195 
196 	case OCTEONTX_LINK_SPEED_RESERVE1:
197 	case OCTEONTX_LINK_SPEED_RESERVE2:
198 	default:
199 		link->link_speed = ETH_SPEED_NUM_NONE;
200 		octeontx_log_err("incorrect link speed %d", nic->speed);
201 		break;
202 	}
203 
204 	link->link_duplex = ETH_LINK_FULL_DUPLEX;
205 	link->link_autoneg = ETH_LINK_AUTONEG;
206 }
207 
208 static void
209 octeontx_link_status_poll(void *arg)
210 {
211 	struct octeontx_nic *nic = arg;
212 	struct rte_eth_link link;
213 	struct rte_eth_dev *dev;
214 	int res;
215 
216 	PMD_INIT_FUNC_TRACE();
217 
218 	dev = nic->dev;
219 
220 	res = octeontx_bgx_port_link_status(nic->port_id);
221 	if (res < 0) {
222 		octeontx_log_err("Failed to get port %d link status",
223 				nic->port_id);
224 	} else {
225 		if (nic->link_up != (uint8_t)res) {
226 			nic->link_up = (uint8_t)res;
227 			octeontx_link_status_update(nic, &link);
228 			octeontx_link_status_print(dev, &link);
229 			rte_eth_linkstatus_set(dev, &link);
230 			rte_eth_dev_callback_process(dev,
231 						     RTE_ETH_EVENT_INTR_LSC,
232 						     NULL);
233 		}
234 	}
235 
236 	res = rte_eal_alarm_set(OCCTX_INTR_POLL_INTERVAL_MS * 1000,
237 				octeontx_link_status_poll, nic);
238 	if (res < 0)
239 		octeontx_log_err("Failed to restart alarm for port %d, err: %d",
240 				nic->port_id, res);
241 }
242 
243 static void
244 octeontx_port_close(struct octeontx_nic *nic)
245 {
246 	PMD_INIT_FUNC_TRACE();
247 
248 	rte_eal_alarm_cancel(octeontx_link_status_poll, nic);
249 	octeontx_bgx_port_close(nic->port_id);
250 	octeontx_log_dbg("port closed %d", nic->port_id);
251 }
252 
253 static int
254 octeontx_port_start(struct octeontx_nic *nic)
255 {
256 	PMD_INIT_FUNC_TRACE();
257 
258 	return octeontx_bgx_port_start(nic->port_id);
259 }
260 
261 static int
262 octeontx_port_stop(struct octeontx_nic *nic)
263 {
264 	PMD_INIT_FUNC_TRACE();
265 
266 	return octeontx_bgx_port_stop(nic->port_id);
267 }
268 
269 static int
270 octeontx_port_promisc_set(struct octeontx_nic *nic, int en)
271 {
272 	struct rte_eth_dev *dev;
273 	int res;
274 
275 	res = 0;
276 	PMD_INIT_FUNC_TRACE();
277 	dev = nic->dev;
278 
279 	res = octeontx_bgx_port_promisc_set(nic->port_id, en);
280 	if (res < 0) {
281 		octeontx_log_err("failed to set promiscuous mode %d",
282 				nic->port_id);
283 		return res;
284 	}
285 
286 	/* Set proper flag for the mode */
287 	dev->data->promiscuous = (en != 0) ? 1 : 0;
288 
289 	octeontx_log_dbg("port %d : promiscuous mode %s",
290 			nic->port_id, en ? "set" : "unset");
291 
292 	return 0;
293 }
294 
295 static int
296 octeontx_port_stats(struct octeontx_nic *nic, struct rte_eth_stats *stats)
297 {
298 	octeontx_mbox_bgx_port_stats_t bgx_stats;
299 	int res;
300 
301 	PMD_INIT_FUNC_TRACE();
302 
303 	res = octeontx_bgx_port_stats(nic->port_id, &bgx_stats);
304 	if (res < 0) {
305 		octeontx_log_err("failed to get port stats %d", nic->port_id);
306 		return res;
307 	}
308 
309 	stats->ipackets = bgx_stats.rx_packets;
310 	stats->ibytes = bgx_stats.rx_bytes;
311 	stats->imissed = bgx_stats.rx_dropped;
312 	stats->ierrors = bgx_stats.rx_errors;
313 	stats->opackets = bgx_stats.tx_packets;
314 	stats->obytes = bgx_stats.tx_bytes;
315 	stats->oerrors = bgx_stats.tx_errors;
316 
317 	octeontx_log_dbg("port%d stats inpkts=%" PRIx64 " outpkts=%" PRIx64 "",
318 			nic->port_id, stats->ipackets, stats->opackets);
319 
320 	return 0;
321 }
322 
323 static int
324 octeontx_port_stats_clr(struct octeontx_nic *nic)
325 {
326 	PMD_INIT_FUNC_TRACE();
327 
328 	return octeontx_bgx_port_stats_clr(nic->port_id);
329 }
330 
331 static inline void
332 devconf_set_default_sane_values(struct rte_event_dev_config *dev_conf,
333 				struct rte_event_dev_info *info)
334 {
335 	memset(dev_conf, 0, sizeof(struct rte_event_dev_config));
336 	dev_conf->dequeue_timeout_ns = info->min_dequeue_timeout_ns;
337 
338 	dev_conf->nb_event_ports = info->max_event_ports;
339 	dev_conf->nb_event_queues = info->max_event_queues;
340 
341 	dev_conf->nb_event_queue_flows = info->max_event_queue_flows;
342 	dev_conf->nb_event_port_dequeue_depth =
343 			info->max_event_port_dequeue_depth;
344 	dev_conf->nb_event_port_enqueue_depth =
345 			info->max_event_port_enqueue_depth;
346 	dev_conf->nb_event_port_enqueue_depth =
347 			info->max_event_port_enqueue_depth;
348 	dev_conf->nb_events_limit =
349 			info->max_num_events;
350 }
351 
352 static uint16_t
353 octeontx_tx_offload_flags(struct rte_eth_dev *eth_dev)
354 {
355 	struct octeontx_nic *nic = octeontx_pmd_priv(eth_dev);
356 	uint16_t flags = 0;
357 
358 	if (nic->tx_offloads & DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM ||
359 	    nic->tx_offloads & DEV_TX_OFFLOAD_OUTER_UDP_CKSUM)
360 		flags |= OCCTX_TX_OFFLOAD_OL3_OL4_CSUM_F;
361 
362 	if (nic->tx_offloads & DEV_TX_OFFLOAD_IPV4_CKSUM ||
363 	    nic->tx_offloads & DEV_TX_OFFLOAD_TCP_CKSUM ||
364 	    nic->tx_offloads & DEV_TX_OFFLOAD_UDP_CKSUM ||
365 	    nic->tx_offloads & DEV_TX_OFFLOAD_SCTP_CKSUM)
366 		flags |= OCCTX_TX_OFFLOAD_L3_L4_CSUM_F;
367 
368 	if (!(nic->tx_offloads & DEV_TX_OFFLOAD_MBUF_FAST_FREE))
369 		flags |= OCCTX_TX_OFFLOAD_MBUF_NOFF_F;
370 
371 	if (nic->tx_offloads & DEV_TX_OFFLOAD_MULTI_SEGS)
372 		flags |= OCCTX_TX_MULTI_SEG_F;
373 
374 	return flags;
375 }
376 
377 static uint16_t
378 octeontx_rx_offload_flags(struct rte_eth_dev *eth_dev)
379 {
380 	struct octeontx_nic *nic = octeontx_pmd_priv(eth_dev);
381 	uint16_t flags = 0;
382 
383 	if (nic->rx_offloads & (DEV_RX_OFFLOAD_TCP_CKSUM |
384 			 DEV_RX_OFFLOAD_UDP_CKSUM))
385 		flags |= OCCTX_RX_OFFLOAD_CSUM_F;
386 
387 	if (nic->rx_offloads & (DEV_RX_OFFLOAD_IPV4_CKSUM |
388 				DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM))
389 		flags |= OCCTX_RX_OFFLOAD_CSUM_F;
390 
391 	if (nic->rx_offloads & DEV_RX_OFFLOAD_SCATTER) {
392 		flags |= OCCTX_RX_MULTI_SEG_F;
393 		eth_dev->data->scattered_rx = 1;
394 		/* If scatter mode is enabled, TX should also be in multi
395 		 * seg mode, else memory leak will occur
396 		 */
397 		nic->tx_offloads |= DEV_TX_OFFLOAD_MULTI_SEGS;
398 	}
399 
400 	return flags;
401 }
402 
403 static int
404 octeontx_dev_configure(struct rte_eth_dev *dev)
405 {
406 	struct rte_eth_dev_data *data = dev->data;
407 	struct rte_eth_conf *conf = &data->dev_conf;
408 	struct rte_eth_rxmode *rxmode = &conf->rxmode;
409 	struct rte_eth_txmode *txmode = &conf->txmode;
410 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
411 	int ret;
412 
413 	PMD_INIT_FUNC_TRACE();
414 	RTE_SET_USED(conf);
415 
416 	if (!rte_eal_has_hugepages()) {
417 		octeontx_log_err("huge page is not configured");
418 		return -EINVAL;
419 	}
420 
421 	if (txmode->mq_mode) {
422 		octeontx_log_err("tx mq_mode DCB or VMDq not supported");
423 		return -EINVAL;
424 	}
425 
426 	if (rxmode->mq_mode != ETH_MQ_RX_NONE &&
427 		rxmode->mq_mode != ETH_MQ_RX_RSS) {
428 		octeontx_log_err("unsupported rx qmode %d", rxmode->mq_mode);
429 		return -EINVAL;
430 	}
431 
432 	if (!(txmode->offloads & DEV_TX_OFFLOAD_MT_LOCKFREE)) {
433 		PMD_INIT_LOG(NOTICE, "cant disable lockfree tx");
434 		txmode->offloads |= DEV_TX_OFFLOAD_MT_LOCKFREE;
435 	}
436 
437 	if (conf->link_speeds & ETH_LINK_SPEED_FIXED) {
438 		octeontx_log_err("setting link speed/duplex not supported");
439 		return -EINVAL;
440 	}
441 
442 	if (conf->dcb_capability_en) {
443 		octeontx_log_err("DCB enable not supported");
444 		return -EINVAL;
445 	}
446 
447 	if (conf->fdir_conf.mode != RTE_FDIR_MODE_NONE) {
448 		octeontx_log_err("flow director not supported");
449 		return -EINVAL;
450 	}
451 
452 	nic->num_tx_queues = dev->data->nb_tx_queues;
453 
454 	ret = octeontx_pko_channel_open(nic->pko_vfid * PKO_VF_NUM_DQ,
455 					nic->num_tx_queues,
456 					nic->base_ochan);
457 	if (ret) {
458 		octeontx_log_err("failed to open channel %d no-of-txq %d",
459 			   nic->base_ochan, nic->num_tx_queues);
460 		return -EFAULT;
461 	}
462 
463 	ret = octeontx_dev_vlan_offload_init(dev);
464 	if (ret) {
465 		octeontx_log_err("failed to initialize vlan offload");
466 		return -EFAULT;
467 	}
468 
469 	nic->pki.classifier_enable = false;
470 	nic->pki.hash_enable = true;
471 	nic->pki.initialized = false;
472 
473 	nic->rx_offloads |= rxmode->offloads;
474 	nic->tx_offloads |= txmode->offloads;
475 	nic->rx_offload_flags |= octeontx_rx_offload_flags(dev);
476 	nic->tx_offload_flags |= octeontx_tx_offload_flags(dev);
477 
478 	return 0;
479 }
480 
481 static int
482 octeontx_dev_close(struct rte_eth_dev *dev)
483 {
484 	struct octeontx_txq *txq = NULL;
485 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
486 	unsigned int i;
487 	int ret;
488 
489 	PMD_INIT_FUNC_TRACE();
490 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
491 		return 0;
492 
493 	rte_event_dev_close(nic->evdev);
494 
495 	octeontx_dev_flow_ctrl_fini(dev);
496 
497 	octeontx_dev_vlan_offload_fini(dev);
498 
499 	ret = octeontx_pko_channel_close(nic->base_ochan);
500 	if (ret < 0) {
501 		octeontx_log_err("failed to close channel %d VF%d %d %d",
502 			     nic->base_ochan, nic->port_id, nic->num_tx_queues,
503 			     ret);
504 	}
505 	/* Free txq resources for this port */
506 	for (i = 0; i < nic->num_tx_queues; i++) {
507 		txq = dev->data->tx_queues[i];
508 		if (!txq)
509 			continue;
510 
511 		rte_free(txq);
512 	}
513 
514 	octeontx_port_close(nic);
515 
516 	return 0;
517 }
518 
519 static int
520 octeontx_dev_mtu_set(struct rte_eth_dev *eth_dev, uint16_t mtu)
521 {
522 	uint32_t buffsz, frame_size = mtu + OCCTX_L2_OVERHEAD;
523 	struct octeontx_nic *nic = octeontx_pmd_priv(eth_dev);
524 	struct rte_eth_dev_data *data = eth_dev->data;
525 	int rc = 0;
526 
527 	/* Check if MTU is within the allowed range */
528 	if (frame_size < OCCTX_MIN_FRS || frame_size > OCCTX_MAX_FRS)
529 		return -EINVAL;
530 
531 	buffsz = data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM;
532 
533 	/* Refuse MTU that requires the support of scattered packets
534 	 * when this feature has not been enabled before.
535 	 */
536 	if (data->dev_started && frame_size > buffsz &&
537 	    !(nic->rx_offloads & DEV_RX_OFFLOAD_SCATTER)) {
538 		octeontx_log_err("Scatter mode is disabled");
539 		return -EINVAL;
540 	}
541 
542 	/* Check <seg size> * <max_seg>  >= max_frame */
543 	if ((nic->rx_offloads & DEV_RX_OFFLOAD_SCATTER)	&&
544 	    (frame_size > buffsz * OCCTX_RX_NB_SEG_MAX))
545 		return -EINVAL;
546 
547 	rc = octeontx_pko_send_mtu(nic->port_id, frame_size);
548 	if (rc)
549 		return rc;
550 
551 	rc = octeontx_bgx_port_mtu_set(nic->port_id, frame_size);
552 	if (rc)
553 		return rc;
554 
555 	if (mtu > RTE_ETHER_MTU)
556 		nic->rx_offloads |= DEV_RX_OFFLOAD_JUMBO_FRAME;
557 	else
558 		nic->rx_offloads &= ~DEV_RX_OFFLOAD_JUMBO_FRAME;
559 
560 	octeontx_log_info("Received pkt beyond  maxlen %d will be dropped",
561 			  frame_size);
562 
563 	return rc;
564 }
565 
566 static int
567 octeontx_recheck_rx_offloads(struct octeontx_rxq *rxq)
568 {
569 	struct rte_eth_dev *eth_dev = rxq->eth_dev;
570 	struct octeontx_nic *nic = octeontx_pmd_priv(eth_dev);
571 	struct rte_eth_dev_data *data = eth_dev->data;
572 	struct rte_pktmbuf_pool_private *mbp_priv;
573 	struct evdev_priv_data *evdev_priv;
574 	struct rte_eventdev *dev;
575 	uint32_t buffsz;
576 
577 	/* Get rx buffer size */
578 	mbp_priv = rte_mempool_get_priv(rxq->pool);
579 	buffsz = mbp_priv->mbuf_data_room_size - RTE_PKTMBUF_HEADROOM;
580 
581 	/* Setup scatter mode if needed by jumbo */
582 	if (data->mtu > buffsz) {
583 		nic->rx_offloads |= DEV_RX_OFFLOAD_SCATTER;
584 		nic->rx_offload_flags |= octeontx_rx_offload_flags(eth_dev);
585 		nic->tx_offload_flags |= octeontx_tx_offload_flags(eth_dev);
586 	}
587 
588 	/* Sharing offload flags via eventdev priv region */
589 	dev = &rte_eventdevs[rxq->evdev];
590 	evdev_priv = dev->data->dev_private;
591 	evdev_priv->rx_offload_flags = nic->rx_offload_flags;
592 	evdev_priv->tx_offload_flags = nic->tx_offload_flags;
593 
594 	/* Setup MTU */
595 	nic->mtu = data->mtu;
596 
597 	return 0;
598 }
599 
600 static int
601 octeontx_dev_start(struct rte_eth_dev *dev)
602 {
603 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
604 	struct octeontx_rxq *rxq;
605 	int ret, i;
606 
607 	PMD_INIT_FUNC_TRACE();
608 	/* Rechecking if any new offload set to update
609 	 * rx/tx burst function pointer accordingly.
610 	 */
611 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
612 		rxq = dev->data->rx_queues[i];
613 		octeontx_recheck_rx_offloads(rxq);
614 	}
615 
616 	/* Setting up the mtu */
617 	ret = octeontx_dev_mtu_set(dev, nic->mtu);
618 	if (ret) {
619 		octeontx_log_err("Failed to set default MTU size %d", ret);
620 		goto error;
621 	}
622 
623 	/*
624 	 * Tx start
625 	 */
626 	octeontx_set_tx_function(dev);
627 	ret = octeontx_pko_channel_start(nic->base_ochan);
628 	if (ret < 0) {
629 		octeontx_log_err("fail to conf VF%d no. txq %d chan %d ret %d",
630 			   nic->port_id, nic->num_tx_queues, nic->base_ochan,
631 			   ret);
632 		goto error;
633 	}
634 
635 	/*
636 	 * Rx start
637 	 */
638 	dev->rx_pkt_burst = octeontx_recv_pkts;
639 	ret = octeontx_pki_port_start(nic->port_id);
640 	if (ret < 0) {
641 		octeontx_log_err("fail to start Rx on port %d", nic->port_id);
642 		goto channel_stop_error;
643 	}
644 
645 	/*
646 	 * Start port
647 	 */
648 	ret = octeontx_port_start(nic);
649 	if (ret < 0) {
650 		octeontx_log_err("failed start port %d", ret);
651 		goto pki_port_stop_error;
652 	}
653 
654 	PMD_TX_LOG(DEBUG, "pko: start channel %d no.of txq %d port %d",
655 			nic->base_ochan, nic->num_tx_queues, nic->port_id);
656 
657 	ret = rte_event_dev_start(nic->evdev);
658 	if (ret < 0) {
659 		octeontx_log_err("failed to start evdev: ret (%d)", ret);
660 		goto pki_port_stop_error;
661 	}
662 
663 	/* Success */
664 	return ret;
665 
666 pki_port_stop_error:
667 	octeontx_pki_port_stop(nic->port_id);
668 channel_stop_error:
669 	octeontx_pko_channel_stop(nic->base_ochan);
670 error:
671 	return ret;
672 }
673 
674 static int
675 octeontx_dev_stop(struct rte_eth_dev *dev)
676 {
677 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
678 	int ret;
679 
680 	PMD_INIT_FUNC_TRACE();
681 
682 	rte_event_dev_stop(nic->evdev);
683 
684 	ret = octeontx_port_stop(nic);
685 	if (ret < 0) {
686 		octeontx_log_err("failed to req stop port %d res=%d",
687 					nic->port_id, ret);
688 		return ret;
689 	}
690 
691 	ret = octeontx_pki_port_stop(nic->port_id);
692 	if (ret < 0) {
693 		octeontx_log_err("failed to stop pki port %d res=%d",
694 					nic->port_id, ret);
695 		return ret;
696 	}
697 
698 	ret = octeontx_pko_channel_stop(nic->base_ochan);
699 	if (ret < 0) {
700 		octeontx_log_err("failed to stop channel %d VF%d %d %d",
701 			     nic->base_ochan, nic->port_id, nic->num_tx_queues,
702 			     ret);
703 		return ret;
704 	}
705 
706 	return 0;
707 }
708 
709 static int
710 octeontx_dev_promisc_enable(struct rte_eth_dev *dev)
711 {
712 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
713 
714 	PMD_INIT_FUNC_TRACE();
715 	return octeontx_port_promisc_set(nic, 1);
716 }
717 
718 static int
719 octeontx_dev_promisc_disable(struct rte_eth_dev *dev)
720 {
721 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
722 
723 	PMD_INIT_FUNC_TRACE();
724 	return octeontx_port_promisc_set(nic, 0);
725 }
726 
727 static int
728 octeontx_port_link_status(struct octeontx_nic *nic)
729 {
730 	int res;
731 
732 	PMD_INIT_FUNC_TRACE();
733 	res = octeontx_bgx_port_link_status(nic->port_id);
734 	if (res < 0) {
735 		octeontx_log_err("failed to get port %d link status",
736 				nic->port_id);
737 		return res;
738 	}
739 
740 	if (nic->link_up != (uint8_t)res || nic->print_flag == -1) {
741 		nic->link_up = (uint8_t)res;
742 		nic->print_flag = 1;
743 	}
744 	octeontx_log_dbg("port %d link status %d", nic->port_id, nic->link_up);
745 
746 	return res;
747 }
748 
749 /*
750  * Return 0 means link status changed, -1 means not changed
751  */
752 static int
753 octeontx_dev_link_update(struct rte_eth_dev *dev,
754 			 int wait_to_complete __rte_unused)
755 {
756 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
757 	struct rte_eth_link link;
758 	int res;
759 
760 	PMD_INIT_FUNC_TRACE();
761 
762 	res = octeontx_port_link_status(nic);
763 	if (res < 0) {
764 		octeontx_log_err("failed to request link status %d", res);
765 		return res;
766 	}
767 
768 	octeontx_link_status_update(nic, &link);
769 	if (nic->print_flag) {
770 		octeontx_link_status_print(nic->dev, &link);
771 		nic->print_flag = 0;
772 	}
773 
774 	return rte_eth_linkstatus_set(dev, &link);
775 }
776 
777 static int
778 octeontx_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
779 {
780 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
781 
782 	PMD_INIT_FUNC_TRACE();
783 	return octeontx_port_stats(nic, stats);
784 }
785 
786 static int
787 octeontx_dev_stats_reset(struct rte_eth_dev *dev)
788 {
789 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
790 
791 	PMD_INIT_FUNC_TRACE();
792 	return octeontx_port_stats_clr(nic);
793 }
794 
795 static void
796 octeontx_dev_mac_addr_del(struct rte_eth_dev *dev, uint32_t index)
797 {
798 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
799 	int ret;
800 
801 	ret = octeontx_bgx_port_mac_del(nic->port_id, index);
802 	if (ret != 0)
803 		octeontx_log_err("failed to del MAC address filter on port %d",
804 				 nic->port_id);
805 }
806 
807 static int
808 octeontx_dev_mac_addr_add(struct rte_eth_dev *dev,
809 			  struct rte_ether_addr *mac_addr,
810 			  uint32_t index,
811 			  __rte_unused uint32_t vmdq)
812 {
813 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
814 	int ret;
815 
816 	ret = octeontx_bgx_port_mac_add(nic->port_id, mac_addr->addr_bytes,
817 					index);
818 	if (ret < 0) {
819 		octeontx_log_err("failed to add MAC address filter on port %d",
820 				 nic->port_id);
821 		return ret;
822 	}
823 
824 	return 0;
825 }
826 
827 static int
828 octeontx_dev_default_mac_addr_set(struct rte_eth_dev *dev,
829 					struct rte_ether_addr *addr)
830 {
831 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
832 	int ret;
833 
834 	ret = octeontx_bgx_port_mac_set(nic->port_id, addr->addr_bytes);
835 	if (ret == 0) {
836 		/* Update same mac address to BGX CAM table */
837 		ret = octeontx_bgx_port_mac_add(nic->port_id, addr->addr_bytes,
838 						0);
839 	}
840 	if (ret < 0) {
841 		octeontx_log_err("failed to set MAC address on port %d",
842 				 nic->port_id);
843 	}
844 
845 	return ret;
846 }
847 
848 static int
849 octeontx_dev_info(struct rte_eth_dev *dev,
850 		struct rte_eth_dev_info *dev_info)
851 {
852 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
853 
854 	/* Autonegotiation may be disabled */
855 	dev_info->speed_capa = ETH_LINK_SPEED_FIXED;
856 	dev_info->speed_capa |= ETH_LINK_SPEED_10M | ETH_LINK_SPEED_100M |
857 			ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G |
858 			ETH_LINK_SPEED_40G;
859 
860 	/* Min/Max MTU supported */
861 	dev_info->min_rx_bufsize = OCCTX_MIN_FRS;
862 	dev_info->max_rx_pktlen = OCCTX_MAX_FRS;
863 	dev_info->max_mtu = dev_info->max_rx_pktlen - OCCTX_L2_OVERHEAD;
864 	dev_info->min_mtu = dev_info->min_rx_bufsize - OCCTX_L2_OVERHEAD;
865 
866 	dev_info->max_mac_addrs =
867 				octeontx_bgx_port_mac_entries_get(nic->port_id);
868 	dev_info->max_rx_queues = 1;
869 	dev_info->max_tx_queues = PKO_MAX_NUM_DQ;
870 	dev_info->min_rx_bufsize = 0;
871 
872 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
873 		.rx_free_thresh = 0,
874 		.rx_drop_en = 0,
875 		.offloads = OCTEONTX_RX_OFFLOADS,
876 	};
877 
878 	dev_info->default_txconf = (struct rte_eth_txconf) {
879 		.tx_free_thresh = 0,
880 		.offloads = OCTEONTX_TX_OFFLOADS,
881 	};
882 
883 	dev_info->rx_offload_capa = OCTEONTX_RX_OFFLOADS;
884 	dev_info->tx_offload_capa = OCTEONTX_TX_OFFLOADS;
885 	dev_info->rx_queue_offload_capa = OCTEONTX_RX_OFFLOADS;
886 	dev_info->tx_queue_offload_capa = OCTEONTX_TX_OFFLOADS;
887 
888 	return 0;
889 }
890 
891 static void
892 octeontx_dq_info_getter(octeontx_dq_t *dq, void *out)
893 {
894 	((octeontx_dq_t *)out)->lmtline_va = dq->lmtline_va;
895 	((octeontx_dq_t *)out)->ioreg_va = dq->ioreg_va;
896 	((octeontx_dq_t *)out)->fc_status_va = dq->fc_status_va;
897 }
898 
899 static int
900 octeontx_vf_start_tx_queue(struct rte_eth_dev *dev, struct octeontx_nic *nic,
901 				uint16_t qidx)
902 {
903 	struct octeontx_txq *txq;
904 	int res;
905 
906 	PMD_INIT_FUNC_TRACE();
907 
908 	if (dev->data->tx_queue_state[qidx] == RTE_ETH_QUEUE_STATE_STARTED)
909 		return 0;
910 
911 	txq = dev->data->tx_queues[qidx];
912 
913 	res = octeontx_pko_channel_query_dqs(nic->base_ochan,
914 						&txq->dq,
915 						sizeof(octeontx_dq_t),
916 						txq->queue_id,
917 						octeontx_dq_info_getter);
918 	if (res < 0) {
919 		res = -EFAULT;
920 		goto close_port;
921 	}
922 
923 	dev->data->tx_queue_state[qidx] = RTE_ETH_QUEUE_STATE_STARTED;
924 	return res;
925 
926 close_port:
927 	(void)octeontx_port_stop(nic);
928 	octeontx_pko_channel_stop(nic->base_ochan);
929 	octeontx_pko_channel_close(nic->base_ochan);
930 	dev->data->tx_queue_state[qidx] = RTE_ETH_QUEUE_STATE_STOPPED;
931 	return res;
932 }
933 
934 int
935 octeontx_dev_tx_queue_start(struct rte_eth_dev *dev, uint16_t qidx)
936 {
937 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
938 
939 	PMD_INIT_FUNC_TRACE();
940 	qidx = qidx % PKO_VF_NUM_DQ;
941 	return octeontx_vf_start_tx_queue(dev, nic, qidx);
942 }
943 
944 static inline int
945 octeontx_vf_stop_tx_queue(struct rte_eth_dev *dev, struct octeontx_nic *nic,
946 			  uint16_t qidx)
947 {
948 	int ret = 0;
949 
950 	RTE_SET_USED(nic);
951 	PMD_INIT_FUNC_TRACE();
952 
953 	if (dev->data->tx_queue_state[qidx] == RTE_ETH_QUEUE_STATE_STOPPED)
954 		return 0;
955 
956 	dev->data->tx_queue_state[qidx] = RTE_ETH_QUEUE_STATE_STOPPED;
957 	return ret;
958 }
959 
960 int
961 octeontx_dev_tx_queue_stop(struct rte_eth_dev *dev, uint16_t qidx)
962 {
963 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
964 
965 	PMD_INIT_FUNC_TRACE();
966 	qidx = qidx % PKO_VF_NUM_DQ;
967 
968 	return octeontx_vf_stop_tx_queue(dev, nic, qidx);
969 }
970 
971 static void
972 octeontx_dev_tx_queue_release(struct rte_eth_dev *dev, uint16_t qid)
973 {
974 	int res;
975 
976 	PMD_INIT_FUNC_TRACE();
977 
978 	if (dev->data->tx_queues[qid]) {
979 		res = octeontx_dev_tx_queue_stop(dev, qid);
980 		if (res < 0)
981 			octeontx_log_err("failed stop tx_queue(%d)\n", qid);
982 
983 		rte_free(dev->data->tx_queues[qid]);
984 	}
985 }
986 
987 static int
988 octeontx_dev_tx_queue_setup(struct rte_eth_dev *dev, uint16_t qidx,
989 			    uint16_t nb_desc, unsigned int socket_id,
990 			    const struct rte_eth_txconf *tx_conf __rte_unused)
991 {
992 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
993 	struct octeontx_txq *txq = NULL;
994 	uint16_t dq_num;
995 	int res = 0;
996 
997 	RTE_SET_USED(nb_desc);
998 	RTE_SET_USED(socket_id);
999 
1000 	dq_num = (nic->pko_vfid * PKO_VF_NUM_DQ) + qidx;
1001 
1002 	/* Socket id check */
1003 	if (socket_id != (unsigned int)SOCKET_ID_ANY &&
1004 			socket_id != (unsigned int)nic->node)
1005 		PMD_TX_LOG(INFO, "socket_id expected %d, configured %d",
1006 						socket_id, nic->node);
1007 
1008 	/* Free memory prior to re-allocation if needed. */
1009 	if (dev->data->tx_queues[qidx] != NULL) {
1010 		PMD_TX_LOG(DEBUG, "freeing memory prior to re-allocation %d",
1011 				qidx);
1012 		octeontx_dev_tx_queue_release(dev, qidx);
1013 		dev->data->tx_queues[qidx] = NULL;
1014 	}
1015 
1016 	/* Allocating tx queue data structure */
1017 	txq = rte_zmalloc_socket("ethdev TX queue", sizeof(struct octeontx_txq),
1018 				 RTE_CACHE_LINE_SIZE, nic->node);
1019 	if (txq == NULL) {
1020 		octeontx_log_err("failed to allocate txq=%d", qidx);
1021 		res = -ENOMEM;
1022 		goto err;
1023 	}
1024 
1025 	txq->eth_dev = dev;
1026 	txq->queue_id = dq_num;
1027 	dev->data->tx_queues[qidx] = txq;
1028 	dev->data->tx_queue_state[qidx] = RTE_ETH_QUEUE_STATE_STOPPED;
1029 
1030 	res = octeontx_pko_channel_query_dqs(nic->base_ochan,
1031 						&txq->dq,
1032 						sizeof(octeontx_dq_t),
1033 						txq->queue_id,
1034 						octeontx_dq_info_getter);
1035 	if (res < 0) {
1036 		res = -EFAULT;
1037 		goto err;
1038 	}
1039 
1040 	PMD_TX_LOG(DEBUG, "[%d]:[%d] txq=%p nb_desc=%d lmtline=%p ioreg_va=%p fc_status_va=%p",
1041 			qidx, txq->queue_id, txq, nb_desc, txq->dq.lmtline_va,
1042 			txq->dq.ioreg_va,
1043 			txq->dq.fc_status_va);
1044 
1045 	return res;
1046 
1047 err:
1048 	if (txq)
1049 		rte_free(txq);
1050 
1051 	return res;
1052 }
1053 
1054 static int
1055 octeontx_dev_rx_queue_setup(struct rte_eth_dev *dev, uint16_t qidx,
1056 				uint16_t nb_desc, unsigned int socket_id,
1057 				const struct rte_eth_rxconf *rx_conf,
1058 				struct rte_mempool *mb_pool)
1059 {
1060 	struct octeontx_nic *nic = octeontx_pmd_priv(dev);
1061 	struct rte_mempool_ops *mp_ops = NULL;
1062 	struct octeontx_rxq *rxq = NULL;
1063 	pki_pktbuf_cfg_t pktbuf_conf;
1064 	pki_hash_cfg_t pki_hash;
1065 	pki_qos_cfg_t pki_qos;
1066 	uintptr_t pool;
1067 	int ret, port;
1068 	uint16_t gaura;
1069 	unsigned int ev_queues = (nic->ev_queues * nic->port_id) + qidx;
1070 	unsigned int ev_ports = (nic->ev_ports * nic->port_id) + qidx;
1071 
1072 	RTE_SET_USED(nb_desc);
1073 
1074 	memset(&pktbuf_conf, 0, sizeof(pktbuf_conf));
1075 	memset(&pki_hash, 0, sizeof(pki_hash));
1076 	memset(&pki_qos, 0, sizeof(pki_qos));
1077 
1078 	mp_ops = rte_mempool_get_ops(mb_pool->ops_index);
1079 	if (strcmp(mp_ops->name, "octeontx_fpavf")) {
1080 		octeontx_log_err("failed to find octeontx_fpavf mempool");
1081 		return -ENOTSUP;
1082 	}
1083 
1084 	/* Handle forbidden configurations */
1085 	if (nic->pki.classifier_enable) {
1086 		octeontx_log_err("cannot setup queue %d. "
1087 					"Classifier option unsupported", qidx);
1088 		return -EINVAL;
1089 	}
1090 
1091 	port = nic->port_id;
1092 
1093 	/* Rx deferred start is not supported */
1094 	if (rx_conf->rx_deferred_start) {
1095 		octeontx_log_err("rx deferred start not supported");
1096 		return -EINVAL;
1097 	}
1098 
1099 	/* Verify queue index */
1100 	if (qidx >= dev->data->nb_rx_queues) {
1101 		octeontx_log_err("QID %d not supporteded (0 - %d available)\n",
1102 				qidx, (dev->data->nb_rx_queues - 1));
1103 		return -ENOTSUP;
1104 	}
1105 
1106 	/* Socket id check */
1107 	if (socket_id != (unsigned int)SOCKET_ID_ANY &&
1108 			socket_id != (unsigned int)nic->node)
1109 		PMD_RX_LOG(INFO, "socket_id expected %d, configured %d",
1110 						socket_id, nic->node);
1111 
1112 	/* Allocating rx queue data structure */
1113 	rxq = rte_zmalloc_socket("ethdev RX queue", sizeof(struct octeontx_rxq),
1114 				 RTE_CACHE_LINE_SIZE, nic->node);
1115 	if (rxq == NULL) {
1116 		octeontx_log_err("failed to allocate rxq=%d", qidx);
1117 		return -ENOMEM;
1118 	}
1119 
1120 	if (!nic->pki.initialized) {
1121 		pktbuf_conf.port_type = 0;
1122 		pki_hash.port_type = 0;
1123 		pki_qos.port_type = 0;
1124 
1125 		pktbuf_conf.mmask.f_wqe_skip = 1;
1126 		pktbuf_conf.mmask.f_first_skip = 1;
1127 		pktbuf_conf.mmask.f_later_skip = 1;
1128 		pktbuf_conf.mmask.f_mbuff_size = 1;
1129 		pktbuf_conf.mmask.f_cache_mode = 1;
1130 
1131 		pktbuf_conf.wqe_skip = OCTTX_PACKET_WQE_SKIP;
1132 		pktbuf_conf.first_skip = OCTTX_PACKET_FIRST_SKIP(mb_pool);
1133 		pktbuf_conf.later_skip = OCTTX_PACKET_LATER_SKIP;
1134 		pktbuf_conf.mbuff_size = (mb_pool->elt_size -
1135 					RTE_PKTMBUF_HEADROOM -
1136 					rte_pktmbuf_priv_size(mb_pool) -
1137 					sizeof(struct rte_mbuf));
1138 
1139 		pktbuf_conf.cache_mode = PKI_OPC_MODE_STF2_STT;
1140 
1141 		ret = octeontx_pki_port_pktbuf_config(port, &pktbuf_conf);
1142 		if (ret != 0) {
1143 			octeontx_log_err("fail to configure pktbuf for port %d",
1144 					port);
1145 			rte_free(rxq);
1146 			return ret;
1147 		}
1148 		PMD_RX_LOG(DEBUG, "Port %d Rx pktbuf configured:\n"
1149 				"\tmbuf_size:\t0x%0x\n"
1150 				"\twqe_skip:\t0x%0x\n"
1151 				"\tfirst_skip:\t0x%0x\n"
1152 				"\tlater_skip:\t0x%0x\n"
1153 				"\tcache_mode:\t%s\n",
1154 				port,
1155 				pktbuf_conf.mbuff_size,
1156 				pktbuf_conf.wqe_skip,
1157 				pktbuf_conf.first_skip,
1158 				pktbuf_conf.later_skip,
1159 				(pktbuf_conf.cache_mode ==
1160 						PKI_OPC_MODE_STT) ?
1161 				"STT" :
1162 				(pktbuf_conf.cache_mode ==
1163 						PKI_OPC_MODE_STF) ?
1164 				"STF" :
1165 				(pktbuf_conf.cache_mode ==
1166 						PKI_OPC_MODE_STF1_STT) ?
1167 				"STF1_STT" : "STF2_STT");
1168 
1169 		if (nic->pki.hash_enable) {
1170 			pki_hash.tag_dlc = 1;
1171 			pki_hash.tag_slc = 1;
1172 			pki_hash.tag_dlf = 1;
1173 			pki_hash.tag_slf = 1;
1174 			pki_hash.tag_prt = 1;
1175 			octeontx_pki_port_hash_config(port, &pki_hash);
1176 		}
1177 
1178 		pool = (uintptr_t)mb_pool->pool_id;
1179 
1180 		/* Get the gaura Id */
1181 		gaura = octeontx_fpa_bufpool_gaura(pool);
1182 
1183 		pki_qos.qpg_qos = PKI_QPG_QOS_NONE;
1184 		pki_qos.num_entry = 1;
1185 		pki_qos.drop_policy = 0;
1186 		pki_qos.tag_type = 0L;
1187 		pki_qos.qos_entry[0].port_add = 0;
1188 		pki_qos.qos_entry[0].gaura = gaura;
1189 		pki_qos.qos_entry[0].ggrp_ok = ev_queues;
1190 		pki_qos.qos_entry[0].ggrp_bad = ev_queues;
1191 		pki_qos.qos_entry[0].grptag_bad = 0;
1192 		pki_qos.qos_entry[0].grptag_ok = 0;
1193 
1194 		ret = octeontx_pki_port_create_qos(port, &pki_qos);
1195 		if (ret < 0) {
1196 			octeontx_log_err("failed to create QOS port=%d, q=%d",
1197 					port, qidx);
1198 			rte_free(rxq);
1199 			return ret;
1200 		}
1201 		nic->pki.initialized = true;
1202 	}
1203 
1204 	rxq->port_id = nic->port_id;
1205 	rxq->eth_dev = dev;
1206 	rxq->queue_id = qidx;
1207 	rxq->evdev = nic->evdev;
1208 	rxq->ev_queues = ev_queues;
1209 	rxq->ev_ports = ev_ports;
1210 	rxq->pool = mb_pool;
1211 
1212 	octeontx_recheck_rx_offloads(rxq);
1213 	dev->data->rx_queues[qidx] = rxq;
1214 	dev->data->rx_queue_state[qidx] = RTE_ETH_QUEUE_STATE_STOPPED;
1215 
1216 	return 0;
1217 }
1218 
1219 static void
1220 octeontx_dev_rx_queue_release(struct rte_eth_dev *dev, uint16_t qid)
1221 {
1222 	rte_free(dev->data->rx_queues[qid]);
1223 }
1224 
1225 static const uint32_t *
1226 octeontx_dev_supported_ptypes_get(struct rte_eth_dev *dev)
1227 {
1228 	static const uint32_t ptypes[] = {
1229 		RTE_PTYPE_L3_IPV4,
1230 		RTE_PTYPE_L3_IPV4_EXT,
1231 		RTE_PTYPE_L3_IPV6,
1232 		RTE_PTYPE_L3_IPV6_EXT,
1233 		RTE_PTYPE_L4_TCP,
1234 		RTE_PTYPE_L4_UDP,
1235 		RTE_PTYPE_L4_FRAG,
1236 		RTE_PTYPE_UNKNOWN
1237 	};
1238 
1239 	if (dev->rx_pkt_burst == octeontx_recv_pkts)
1240 		return ptypes;
1241 
1242 	return NULL;
1243 }
1244 
1245 static int
1246 octeontx_pool_ops(struct rte_eth_dev *dev, const char *pool)
1247 {
1248 	RTE_SET_USED(dev);
1249 
1250 	if (!strcmp(pool, "octeontx_fpavf"))
1251 		return 0;
1252 
1253 	return -ENOTSUP;
1254 }
1255 
1256 /* Initialize and register driver with DPDK Application */
1257 static const struct eth_dev_ops octeontx_dev_ops = {
1258 	.dev_configure		 = octeontx_dev_configure,
1259 	.dev_infos_get		 = octeontx_dev_info,
1260 	.dev_close		 = octeontx_dev_close,
1261 	.dev_start		 = octeontx_dev_start,
1262 	.dev_stop		 = octeontx_dev_stop,
1263 	.promiscuous_enable	 = octeontx_dev_promisc_enable,
1264 	.promiscuous_disable	 = octeontx_dev_promisc_disable,
1265 	.link_update		 = octeontx_dev_link_update,
1266 	.stats_get		 = octeontx_dev_stats_get,
1267 	.stats_reset		 = octeontx_dev_stats_reset,
1268 	.mac_addr_remove	 = octeontx_dev_mac_addr_del,
1269 	.mac_addr_add		 = octeontx_dev_mac_addr_add,
1270 	.mac_addr_set		 = octeontx_dev_default_mac_addr_set,
1271 	.vlan_offload_set	 = octeontx_dev_vlan_offload_set,
1272 	.vlan_filter_set	 = octeontx_dev_vlan_filter_set,
1273 	.tx_queue_start		 = octeontx_dev_tx_queue_start,
1274 	.tx_queue_stop		 = octeontx_dev_tx_queue_stop,
1275 	.tx_queue_setup		 = octeontx_dev_tx_queue_setup,
1276 	.tx_queue_release	 = octeontx_dev_tx_queue_release,
1277 	.rx_queue_setup		 = octeontx_dev_rx_queue_setup,
1278 	.rx_queue_release	 = octeontx_dev_rx_queue_release,
1279 	.dev_set_link_up          = octeontx_dev_set_link_up,
1280 	.dev_set_link_down        = octeontx_dev_set_link_down,
1281 	.dev_supported_ptypes_get = octeontx_dev_supported_ptypes_get,
1282 	.mtu_set                 = octeontx_dev_mtu_set,
1283 	.pool_ops_supported      = octeontx_pool_ops,
1284 	.flow_ctrl_get           = octeontx_dev_flow_ctrl_get,
1285 	.flow_ctrl_set           = octeontx_dev_flow_ctrl_set,
1286 };
1287 
1288 /* Create Ethdev interface per BGX LMAC ports */
1289 static int
1290 octeontx_create(struct rte_vdev_device *dev, int port, uint8_t evdev,
1291 			int socket_id)
1292 {
1293 	int res;
1294 	size_t pko_vfid;
1295 	char octtx_name[OCTEONTX_MAX_NAME_LEN];
1296 	struct octeontx_nic *nic = NULL;
1297 	struct rte_eth_dev *eth_dev = NULL;
1298 	struct rte_eth_dev_data *data;
1299 	const char *name = rte_vdev_device_name(dev);
1300 	int max_entries;
1301 
1302 	PMD_INIT_FUNC_TRACE();
1303 
1304 	sprintf(octtx_name, "%s_%d", name, port);
1305 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1306 		eth_dev = rte_eth_dev_attach_secondary(octtx_name);
1307 		if (eth_dev == NULL)
1308 			return -ENODEV;
1309 
1310 		eth_dev->dev_ops = &octeontx_dev_ops;
1311 		eth_dev->device = &dev->device;
1312 		octeontx_set_tx_function(eth_dev);
1313 		eth_dev->rx_pkt_burst = octeontx_recv_pkts;
1314 		rte_eth_dev_probing_finish(eth_dev);
1315 		return 0;
1316 	}
1317 
1318 	/* Reserve an ethdev entry */
1319 	eth_dev = rte_eth_dev_allocate(octtx_name);
1320 	if (eth_dev == NULL) {
1321 		octeontx_log_err("failed to allocate rte_eth_dev");
1322 		res = -ENOMEM;
1323 		goto err;
1324 	}
1325 	data = eth_dev->data;
1326 
1327 	nic = rte_zmalloc_socket(octtx_name, sizeof(*nic), 0, socket_id);
1328 	if (nic == NULL) {
1329 		octeontx_log_err("failed to allocate nic structure");
1330 		res = -ENOMEM;
1331 		goto err;
1332 	}
1333 	data->dev_private = nic;
1334 	pko_vfid = octeontx_pko_get_vfid();
1335 
1336 	if (pko_vfid == SIZE_MAX) {
1337 		octeontx_log_err("failed to get pko vfid");
1338 		res = -ENODEV;
1339 		goto err;
1340 	}
1341 
1342 	nic->pko_vfid = pko_vfid;
1343 	nic->port_id = port;
1344 	nic->evdev = evdev;
1345 
1346 	res = octeontx_port_open(nic);
1347 	if (res < 0)
1348 		goto err;
1349 
1350 	/* Rx side port configuration */
1351 	res = octeontx_pki_port_open(port);
1352 	if (res != 0) {
1353 		octeontx_log_err("failed to open PKI port %d", port);
1354 		res = -ENODEV;
1355 		goto err;
1356 	}
1357 
1358 	eth_dev->device = &dev->device;
1359 	eth_dev->intr_handle = NULL;
1360 	eth_dev->data->numa_node = dev->device.numa_node;
1361 
1362 	data->port_id = eth_dev->data->port_id;
1363 
1364 	nic->ev_queues = 1;
1365 	nic->ev_ports = 1;
1366 	nic->print_flag = -1;
1367 
1368 	data->dev_link.link_status = ETH_LINK_DOWN;
1369 	data->dev_started = 0;
1370 	data->promiscuous = 0;
1371 	data->all_multicast = 0;
1372 	data->scattered_rx = 0;
1373 
1374 	/* Get maximum number of supported MAC entries */
1375 	max_entries = octeontx_bgx_port_mac_entries_get(nic->port_id);
1376 	if (max_entries < 0) {
1377 		octeontx_log_err("Failed to get max entries for mac addr");
1378 		res = -ENOTSUP;
1379 		goto err;
1380 	}
1381 
1382 	data->mac_addrs = rte_zmalloc_socket(octtx_name, max_entries *
1383 					     RTE_ETHER_ADDR_LEN, 0,
1384 							socket_id);
1385 	if (data->mac_addrs == NULL) {
1386 		octeontx_log_err("failed to allocate memory for mac_addrs");
1387 		res = -ENOMEM;
1388 		goto err;
1389 	}
1390 
1391 	eth_dev->dev_ops = &octeontx_dev_ops;
1392 
1393 	/* Finally save ethdev pointer to the NIC structure */
1394 	nic->dev = eth_dev;
1395 
1396 	if (nic->port_id != data->port_id) {
1397 		octeontx_log_err("eth_dev->port_id (%d) is diff to orig (%d)",
1398 				data->port_id, nic->port_id);
1399 		res = -EINVAL;
1400 		goto free_mac_addrs;
1401 	}
1402 
1403 	res = rte_eal_alarm_set(OCCTX_INTR_POLL_INTERVAL_MS * 1000,
1404 				octeontx_link_status_poll, nic);
1405 	if (res) {
1406 		octeontx_log_err("Failed to start link polling alarm");
1407 		goto err;
1408 	}
1409 
1410 	/* Update port_id mac to eth_dev */
1411 	memcpy(data->mac_addrs, nic->mac_addr, RTE_ETHER_ADDR_LEN);
1412 
1413 	/* Update same mac address to BGX CAM table at index 0 */
1414 	octeontx_bgx_port_mac_add(nic->port_id, nic->mac_addr, 0);
1415 
1416 	res = octeontx_dev_flow_ctrl_init(eth_dev);
1417 	if (res < 0)
1418 		goto err;
1419 
1420 	PMD_INIT_LOG(DEBUG, "ethdev info: ");
1421 	PMD_INIT_LOG(DEBUG, "port %d, port_ena %d ochan %d num_ochan %d tx_q %d",
1422 				nic->port_id, nic->port_ena,
1423 				nic->base_ochan, nic->num_ochans,
1424 				nic->num_tx_queues);
1425 	PMD_INIT_LOG(DEBUG, "speed %d mtu %d", nic->speed, nic->bgx_mtu);
1426 
1427 	rte_octeontx_pchan_map[(nic->base_ochan >> 8) & 0x7]
1428 		[(nic->base_ochan >> 4) & 0xF] = data->port_id;
1429 
1430 	rte_eth_dev_probing_finish(eth_dev);
1431 	return data->port_id;
1432 
1433 free_mac_addrs:
1434 	rte_free(data->mac_addrs);
1435 	data->mac_addrs = NULL;
1436 err:
1437 	if (nic)
1438 		octeontx_port_close(nic);
1439 
1440 	rte_eth_dev_release_port(eth_dev);
1441 
1442 	return res;
1443 }
1444 
1445 /* Un initialize octeontx device */
1446 static int
1447 octeontx_remove(struct rte_vdev_device *dev)
1448 {
1449 	char octtx_name[OCTEONTX_MAX_NAME_LEN];
1450 	struct rte_eth_dev *eth_dev = NULL;
1451 	struct octeontx_nic *nic = NULL;
1452 	int i;
1453 
1454 	if (dev == NULL)
1455 		return -EINVAL;
1456 
1457 	for (i = 0; i < OCTEONTX_VDEV_DEFAULT_MAX_NR_PORT; i++) {
1458 		sprintf(octtx_name, "eth_octeontx_%d", i);
1459 
1460 		eth_dev = rte_eth_dev_allocated(octtx_name);
1461 		if (eth_dev == NULL)
1462 			continue; /* port already released */
1463 
1464 		if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1465 			rte_eth_dev_release_port(eth_dev);
1466 			continue;
1467 		}
1468 
1469 		nic = octeontx_pmd_priv(eth_dev);
1470 		rte_event_dev_stop(nic->evdev);
1471 		PMD_INIT_LOG(INFO, "Closing octeontx device %s", octtx_name);
1472 		octeontx_dev_close(eth_dev);
1473 		rte_eth_dev_release_port(eth_dev);
1474 	}
1475 
1476 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1477 		return 0;
1478 
1479 	/* Free FC resource */
1480 	octeontx_pko_fc_free();
1481 
1482 	return 0;
1483 }
1484 
1485 /* Initialize octeontx device */
1486 static int
1487 octeontx_probe(struct rte_vdev_device *dev)
1488 {
1489 	const char *dev_name;
1490 	static int probe_once;
1491 	uint8_t socket_id, qlist;
1492 	int tx_vfcnt, port_id, evdev, qnum, pnum, res, i;
1493 	struct rte_event_dev_config dev_conf;
1494 	const char *eventdev_name = "event_octeontx";
1495 	struct rte_event_dev_info info;
1496 	struct rte_eth_dev *eth_dev;
1497 
1498 	struct octeontx_vdev_init_params init_params = {
1499 		OCTEONTX_VDEV_DEFAULT_MAX_NR_PORT
1500 	};
1501 
1502 	dev_name = rte_vdev_device_name(dev);
1503 
1504 	if (rte_eal_process_type() == RTE_PROC_SECONDARY &&
1505 	    strlen(rte_vdev_device_args(dev)) == 0) {
1506 		eth_dev = rte_eth_dev_attach_secondary(dev_name);
1507 		if (!eth_dev) {
1508 			PMD_INIT_LOG(ERR, "Failed to probe %s", dev_name);
1509 			return -1;
1510 		}
1511 		/* TODO: request info from primary to set up Rx and Tx */
1512 		eth_dev->dev_ops = &octeontx_dev_ops;
1513 		eth_dev->device = &dev->device;
1514 		rte_eth_dev_probing_finish(eth_dev);
1515 		return 0;
1516 	}
1517 
1518 	res = octeontx_parse_vdev_init_params(&init_params, dev);
1519 	if (res < 0)
1520 		return -EINVAL;
1521 
1522 	if (init_params.nr_port > OCTEONTX_VDEV_DEFAULT_MAX_NR_PORT) {
1523 		octeontx_log_err("nr_port (%d) > max (%d)", init_params.nr_port,
1524 				OCTEONTX_VDEV_DEFAULT_MAX_NR_PORT);
1525 		return -ENOTSUP;
1526 	}
1527 
1528 	PMD_INIT_LOG(DEBUG, "initializing %s pmd", dev_name);
1529 
1530 	socket_id = rte_socket_id();
1531 
1532 	tx_vfcnt = octeontx_pko_vf_count();
1533 
1534 	if (tx_vfcnt < init_params.nr_port) {
1535 		octeontx_log_err("not enough PKO (%d) for port number (%d)",
1536 				tx_vfcnt, init_params.nr_port);
1537 		return -EINVAL;
1538 	}
1539 	evdev = rte_event_dev_get_dev_id(eventdev_name);
1540 	if (evdev < 0) {
1541 		octeontx_log_err("eventdev %s not found", eventdev_name);
1542 		return -ENODEV;
1543 	}
1544 
1545 	res = rte_event_dev_info_get(evdev, &info);
1546 	if (res < 0) {
1547 		octeontx_log_err("failed to eventdev info %d", res);
1548 		return -EINVAL;
1549 	}
1550 
1551 	PMD_INIT_LOG(DEBUG, "max_queue %d max_port %d",
1552 			info.max_event_queues, info.max_event_ports);
1553 
1554 	if (octeontx_pko_init_fc(tx_vfcnt))
1555 		return -ENOMEM;
1556 
1557 	devconf_set_default_sane_values(&dev_conf, &info);
1558 	res = rte_event_dev_configure(evdev, &dev_conf);
1559 	if (res < 0)
1560 		goto parse_error;
1561 
1562 	rte_event_dev_attr_get(evdev, RTE_EVENT_DEV_ATTR_PORT_COUNT,
1563 			(uint32_t *)&pnum);
1564 	rte_event_dev_attr_get(evdev, RTE_EVENT_DEV_ATTR_QUEUE_COUNT,
1565 			(uint32_t *)&qnum);
1566 	if (pnum < qnum) {
1567 		octeontx_log_err("too few event ports (%d) for event_q(%d)",
1568 				pnum, qnum);
1569 		res = -EINVAL;
1570 		goto parse_error;
1571 	}
1572 
1573 	/* Enable all queues available */
1574 	for (i = 0; i < qnum; i++) {
1575 		res = rte_event_queue_setup(evdev, i, NULL);
1576 		if (res < 0) {
1577 			octeontx_log_err("failed to setup event_q(%d): res %d",
1578 					i, res);
1579 			goto parse_error;
1580 		}
1581 	}
1582 
1583 	/* Enable all ports available */
1584 	for (i = 0; i < pnum; i++) {
1585 		res = rte_event_port_setup(evdev, i, NULL);
1586 		if (res < 0) {
1587 			res = -ENODEV;
1588 			octeontx_log_err("failed to setup ev port(%d) res=%d",
1589 						i, res);
1590 			goto parse_error;
1591 		}
1592 	}
1593 
1594 	/*
1595 	 * Do 1:1 links for ports & queues. All queues would be mapped to
1596 	 * one port. If there are more ports than queues, then some ports
1597 	 * won't be linked to any queue.
1598 	 */
1599 	for (i = 0; i < qnum; i++) {
1600 		/* Link one queue to one event port */
1601 		qlist = i;
1602 		res = rte_event_port_link(evdev, i, &qlist, NULL, 1);
1603 		if (res < 0) {
1604 			res = -ENODEV;
1605 			octeontx_log_err("failed to link port (%d): res=%d",
1606 					i, res);
1607 			goto parse_error;
1608 		}
1609 	}
1610 
1611 	/* Create ethdev interface */
1612 	for (i = 0; i < init_params.nr_port; i++) {
1613 		port_id = octeontx_create(dev, i, evdev, socket_id);
1614 		if (port_id < 0) {
1615 			octeontx_log_err("failed to create device %s",
1616 					dev_name);
1617 			res = -ENODEV;
1618 			goto parse_error;
1619 		}
1620 
1621 		PMD_INIT_LOG(INFO, "created ethdev %s for port %d", dev_name,
1622 					port_id);
1623 	}
1624 
1625 	if (probe_once) {
1626 		octeontx_log_err("interface %s not supported", dev_name);
1627 		octeontx_remove(dev);
1628 		res = -ENOTSUP;
1629 		goto parse_error;
1630 	}
1631 	rte_mbuf_set_platform_mempool_ops("octeontx_fpavf");
1632 	probe_once = 1;
1633 
1634 	return 0;
1635 
1636 parse_error:
1637 	octeontx_pko_fc_free();
1638 	return res;
1639 }
1640 
1641 static struct rte_vdev_driver octeontx_pmd_drv = {
1642 	.probe = octeontx_probe,
1643 	.remove = octeontx_remove,
1644 };
1645 
1646 RTE_PMD_REGISTER_VDEV(OCTEONTX_PMD, octeontx_pmd_drv);
1647 RTE_PMD_REGISTER_ALIAS(OCTEONTX_PMD, eth_octeontx);
1648 RTE_PMD_REGISTER_PARAM_STRING(OCTEONTX_PMD, "nr_port=<int> ");
1649