xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision bb9cd910)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2014-2021 Broadcom
3  * All rights reserved.
4  */
5 
6 #include <inttypes.h>
7 #include <stdbool.h>
8 
9 #include <rte_dev.h>
10 #include <ethdev_driver.h>
11 #include <ethdev_pci.h>
12 #include <rte_malloc.h>
13 #include <rte_cycles.h>
14 #include <rte_alarm.h>
15 #include <rte_kvargs.h>
16 #include <rte_vect.h>
17 
18 #include "bnxt.h"
19 #include "bnxt_filter.h"
20 #include "bnxt_hwrm.h"
21 #include "bnxt_irq.h"
22 #include "bnxt_reps.h"
23 #include "bnxt_ring.h"
24 #include "bnxt_rxq.h"
25 #include "bnxt_rxr.h"
26 #include "bnxt_stats.h"
27 #include "bnxt_txq.h"
28 #include "bnxt_txr.h"
29 #include "bnxt_vnic.h"
30 #include "hsi_struct_def_dpdk.h"
31 #include "bnxt_nvm_defs.h"
32 #include "bnxt_tf_common.h"
33 #include "ulp_flow_db.h"
34 #include "rte_pmd_bnxt.h"
35 
36 #define DRV_MODULE_NAME		"bnxt"
37 static const char bnxt_version[] =
38 	"Broadcom NetXtreme driver " DRV_MODULE_NAME;
39 
40 /*
41  * The set of PCI devices this driver supports
42  */
43 static const struct rte_pci_id bnxt_pci_id_map[] = {
44 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
45 			 BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
46 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
47 			 BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
48 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
49 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
50 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
51 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
52 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
53 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
54 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
55 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
56 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
57 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
58 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
59 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
60 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
61 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
62 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
63 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
64 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
65 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
66 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
67 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
68 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
69 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
70 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
71 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
72 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
73 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
74 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
75 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
76 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
77 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
78 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
79 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
80 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
81 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
82 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
83 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58812) },
84 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58814) },
85 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818) },
86 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818_VF) },
87 	{ .vendor_id = 0, /* sentinel */ },
88 };
89 
90 #define BNXT_DEVARG_FLOW_XSTAT	"flow-xstat"
91 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
92 #define BNXT_DEVARG_REPRESENTOR	"representor"
93 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
94 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
95 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
96 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
97 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
98 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
99 #define BNXT_DEVARG_APP_ID	"app-id"
100 
101 static const char *const bnxt_dev_args[] = {
102 	BNXT_DEVARG_REPRESENTOR,
103 	BNXT_DEVARG_FLOW_XSTAT,
104 	BNXT_DEVARG_MAX_NUM_KFLOWS,
105 	BNXT_DEVARG_REP_BASED_PF,
106 	BNXT_DEVARG_REP_IS_PF,
107 	BNXT_DEVARG_REP_Q_R2F,
108 	BNXT_DEVARG_REP_Q_F2R,
109 	BNXT_DEVARG_REP_FC_R2F,
110 	BNXT_DEVARG_REP_FC_F2R,
111 	BNXT_DEVARG_APP_ID,
112 	NULL
113 };
114 
115 /*
116  * app-id = an non-negative 8-bit number
117  */
118 #define BNXT_DEVARG_APP_ID_INVALID(val)			((val) > 255)
119 
120 /*
121  * flow_xstat == false to disable the feature
122  * flow_xstat == true to enable the feature
123  */
124 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
125 
126 /*
127  * rep_is_pf == false to indicate VF representor
128  * rep_is_pf == true to indicate PF representor
129  */
130 #define	BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)	((rep_is_pf) > 1)
131 
132 /*
133  * rep_based_pf == Physical index of the PF
134  */
135 #define	BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)	((rep_based_pf) > 15)
136 /*
137  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
138  */
139 #define	BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)	((rep_q_r2f) > 3)
140 
141 /*
142  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
143  */
144 #define	BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)	((rep_q_f2r) > 3)
145 
146 /*
147  * rep_fc_r2f == Flow control for the representor to endpoint direction
148  */
149 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)	((rep_fc_r2f) > 1)
150 
151 /*
152  * rep_fc_f2r == Flow control for the endpoint to representor direction
153  */
154 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)	((rep_fc_f2r) > 1)
155 
156 int bnxt_cfa_code_dynfield_offset = -1;
157 
158 /*
159  * max_num_kflows must be >= 32
160  * and must be a power-of-2 supported value
161  * return: 1 -> invalid
162  *         0 -> valid
163  */
164 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
165 {
166 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
167 		return 1;
168 	return 0;
169 }
170 
171 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
172 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
173 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
174 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
175 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
176 static int bnxt_restore_vlan_filters(struct bnxt *bp);
177 static void bnxt_dev_recover(void *arg);
178 static void bnxt_free_error_recovery_info(struct bnxt *bp);
179 static void bnxt_free_rep_info(struct bnxt *bp);
180 
181 int is_bnxt_in_error(struct bnxt *bp)
182 {
183 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
184 		return -EIO;
185 	if (bp->flags & BNXT_FLAG_FW_RESET)
186 		return -EBUSY;
187 
188 	return 0;
189 }
190 
191 /***********************/
192 
193 /*
194  * High level utility functions
195  */
196 
197 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
198 {
199 	unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
200 					     BNXT_RSS_TBL_SIZE_P5);
201 
202 	if (!BNXT_CHIP_P5(bp))
203 		return 1;
204 
205 	return RTE_ALIGN_MUL_CEIL(num_rss_rings,
206 				  BNXT_RSS_ENTRIES_PER_CTX_P5) /
207 				  BNXT_RSS_ENTRIES_PER_CTX_P5;
208 }
209 
210 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
211 {
212 	if (!BNXT_CHIP_P5(bp))
213 		return HW_HASH_INDEX_SIZE;
214 
215 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
216 }
217 
218 static void bnxt_free_parent_info(struct bnxt *bp)
219 {
220 	rte_free(bp->parent);
221 	bp->parent = NULL;
222 }
223 
224 static void bnxt_free_pf_info(struct bnxt *bp)
225 {
226 	rte_free(bp->pf);
227 	bp->pf = NULL;
228 }
229 
230 static void bnxt_free_link_info(struct bnxt *bp)
231 {
232 	rte_free(bp->link_info);
233 	bp->link_info = NULL;
234 }
235 
236 static void bnxt_free_leds_info(struct bnxt *bp)
237 {
238 	if (BNXT_VF(bp))
239 		return;
240 
241 	rte_free(bp->leds);
242 	bp->leds = NULL;
243 }
244 
245 static void bnxt_free_flow_stats_info(struct bnxt *bp)
246 {
247 	rte_free(bp->flow_stat);
248 	bp->flow_stat = NULL;
249 }
250 
251 static void bnxt_free_cos_queues(struct bnxt *bp)
252 {
253 	rte_free(bp->rx_cos_queue);
254 	bp->rx_cos_queue = NULL;
255 	rte_free(bp->tx_cos_queue);
256 	bp->tx_cos_queue = NULL;
257 }
258 
259 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
260 {
261 	bnxt_free_filter_mem(bp);
262 	bnxt_free_vnic_attributes(bp);
263 	bnxt_free_vnic_mem(bp);
264 
265 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
266 	 * If the number of rings change across fw update,
267 	 * we don't have much choice except to warn the user.
268 	 */
269 	if (!reconfig) {
270 		bnxt_free_stats(bp);
271 		bnxt_free_tx_rings(bp);
272 		bnxt_free_rx_rings(bp);
273 	}
274 	bnxt_free_async_cp_ring(bp);
275 	bnxt_free_rxtx_nq_ring(bp);
276 
277 	rte_free(bp->grp_info);
278 	bp->grp_info = NULL;
279 }
280 
281 static int bnxt_alloc_parent_info(struct bnxt *bp)
282 {
283 	bp->parent = rte_zmalloc("bnxt_parent_info",
284 				 sizeof(struct bnxt_parent_info), 0);
285 	if (bp->parent == NULL)
286 		return -ENOMEM;
287 
288 	return 0;
289 }
290 
291 static int bnxt_alloc_pf_info(struct bnxt *bp)
292 {
293 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
294 	if (bp->pf == NULL)
295 		return -ENOMEM;
296 
297 	return 0;
298 }
299 
300 static int bnxt_alloc_link_info(struct bnxt *bp)
301 {
302 	bp->link_info =
303 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
304 	if (bp->link_info == NULL)
305 		return -ENOMEM;
306 
307 	return 0;
308 }
309 
310 static int bnxt_alloc_leds_info(struct bnxt *bp)
311 {
312 	if (BNXT_VF(bp))
313 		return 0;
314 
315 	bp->leds = rte_zmalloc("bnxt_leds",
316 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
317 			       0);
318 	if (bp->leds == NULL)
319 		return -ENOMEM;
320 
321 	return 0;
322 }
323 
324 static int bnxt_alloc_cos_queues(struct bnxt *bp)
325 {
326 	bp->rx_cos_queue =
327 		rte_zmalloc("bnxt_rx_cosq",
328 			    BNXT_COS_QUEUE_COUNT *
329 			    sizeof(struct bnxt_cos_queue_info),
330 			    0);
331 	if (bp->rx_cos_queue == NULL)
332 		return -ENOMEM;
333 
334 	bp->tx_cos_queue =
335 		rte_zmalloc("bnxt_tx_cosq",
336 			    BNXT_COS_QUEUE_COUNT *
337 			    sizeof(struct bnxt_cos_queue_info),
338 			    0);
339 	if (bp->tx_cos_queue == NULL)
340 		return -ENOMEM;
341 
342 	return 0;
343 }
344 
345 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
346 {
347 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
348 				    sizeof(struct bnxt_flow_stat_info), 0);
349 	if (bp->flow_stat == NULL)
350 		return -ENOMEM;
351 
352 	return 0;
353 }
354 
355 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
356 {
357 	int rc;
358 
359 	rc = bnxt_alloc_ring_grps(bp);
360 	if (rc)
361 		goto alloc_mem_err;
362 
363 	rc = bnxt_alloc_async_ring_struct(bp);
364 	if (rc)
365 		goto alloc_mem_err;
366 
367 	rc = bnxt_alloc_vnic_mem(bp);
368 	if (rc)
369 		goto alloc_mem_err;
370 
371 	rc = bnxt_alloc_vnic_attributes(bp);
372 	if (rc)
373 		goto alloc_mem_err;
374 
375 	rc = bnxt_alloc_filter_mem(bp);
376 	if (rc)
377 		goto alloc_mem_err;
378 
379 	rc = bnxt_alloc_async_cp_ring(bp);
380 	if (rc)
381 		goto alloc_mem_err;
382 
383 	rc = bnxt_alloc_rxtx_nq_ring(bp);
384 	if (rc)
385 		goto alloc_mem_err;
386 
387 	if (BNXT_FLOW_XSTATS_EN(bp)) {
388 		rc = bnxt_alloc_flow_stats_info(bp);
389 		if (rc)
390 			goto alloc_mem_err;
391 	}
392 
393 	return 0;
394 
395 alloc_mem_err:
396 	bnxt_free_mem(bp, reconfig);
397 	return rc;
398 }
399 
400 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
401 {
402 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
403 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
404 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
405 	struct bnxt_rx_queue *rxq;
406 	unsigned int j;
407 	int rc;
408 
409 	rc = bnxt_vnic_grp_alloc(bp, vnic);
410 	if (rc)
411 		goto err_out;
412 
413 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
414 		    vnic_id, vnic, vnic->fw_grp_ids);
415 
416 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
417 	if (rc)
418 		goto err_out;
419 
420 	/* Alloc RSS context only if RSS mode is enabled */
421 	if (dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS) {
422 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
423 
424 		/* RSS table size in Thor is 512.
425 		 * Cap max Rx rings to same value
426 		 */
427 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
428 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
429 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
430 			goto err_out;
431 		}
432 
433 		rc = 0;
434 		for (j = 0; j < nr_ctxs; j++) {
435 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
436 			if (rc)
437 				break;
438 		}
439 		if (rc) {
440 			PMD_DRV_LOG(ERR,
441 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
442 				    vnic_id, j, rc);
443 			goto err_out;
444 		}
445 		vnic->num_lb_ctxts = nr_ctxs;
446 	}
447 
448 	/*
449 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
450 	 * setting is not available at this time, it will not be
451 	 * configured correctly in the CFA.
452 	 */
453 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
454 		vnic->vlan_strip = true;
455 	else
456 		vnic->vlan_strip = false;
457 
458 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
459 	if (rc)
460 		goto err_out;
461 
462 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
463 	if (rc)
464 		goto err_out;
465 
466 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
467 		rxq = bp->eth_dev->data->rx_queues[j];
468 
469 		PMD_DRV_LOG(DEBUG,
470 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
471 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
472 
473 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
474 			rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
475 		else
476 			vnic->rx_queue_cnt++;
477 	}
478 
479 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
480 
481 	rc = bnxt_vnic_rss_configure(bp, vnic);
482 	if (rc)
483 		goto err_out;
484 
485 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
486 
487 	rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
488 				    (rx_offloads & RTE_ETH_RX_OFFLOAD_TCP_LRO) ?
489 				    true : false);
490 	if (rc)
491 		goto err_out;
492 
493 	return 0;
494 err_out:
495 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
496 		    vnic_id, rc);
497 	return rc;
498 }
499 
500 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
501 {
502 	int rc = 0;
503 
504 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
505 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
506 	if (rc)
507 		return rc;
508 
509 	PMD_DRV_LOG(DEBUG,
510 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
511 		    " rx_fc_in_tbl.ctx_id = %d\n",
512 		    bp->flow_stat->rx_fc_in_tbl.va,
513 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
514 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
515 
516 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
517 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
518 	if (rc)
519 		return rc;
520 
521 	PMD_DRV_LOG(DEBUG,
522 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
523 		    " rx_fc_out_tbl.ctx_id = %d\n",
524 		    bp->flow_stat->rx_fc_out_tbl.va,
525 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
526 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
527 
528 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
529 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
530 	if (rc)
531 		return rc;
532 
533 	PMD_DRV_LOG(DEBUG,
534 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
535 		    " tx_fc_in_tbl.ctx_id = %d\n",
536 		    bp->flow_stat->tx_fc_in_tbl.va,
537 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
538 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
539 
540 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
541 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
542 	if (rc)
543 		return rc;
544 
545 	PMD_DRV_LOG(DEBUG,
546 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
547 		    " tx_fc_out_tbl.ctx_id = %d\n",
548 		    bp->flow_stat->tx_fc_out_tbl.va,
549 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
550 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
551 
552 	memset(bp->flow_stat->rx_fc_out_tbl.va,
553 	       0,
554 	       bp->flow_stat->rx_fc_out_tbl.size);
555 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
556 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
557 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
558 				       bp->flow_stat->max_fc,
559 				       true);
560 	if (rc)
561 		return rc;
562 
563 	memset(bp->flow_stat->tx_fc_out_tbl.va,
564 	       0,
565 	       bp->flow_stat->tx_fc_out_tbl.size);
566 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
567 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
568 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
569 				       bp->flow_stat->max_fc,
570 				       true);
571 
572 	return rc;
573 }
574 
575 static int bnxt_alloc_ctx_mem_buf(struct bnxt *bp, char *type, size_t size,
576 				  struct bnxt_ctx_mem_buf_info *ctx)
577 {
578 	if (!ctx)
579 		return -EINVAL;
580 
581 	ctx->va = rte_zmalloc_socket(type, size, 0,
582 				     bp->eth_dev->device->numa_node);
583 	if (ctx->va == NULL)
584 		return -ENOMEM;
585 	rte_mem_lock_page(ctx->va);
586 	ctx->size = size;
587 	ctx->dma = rte_mem_virt2iova(ctx->va);
588 	if (ctx->dma == RTE_BAD_IOVA)
589 		return -ENOMEM;
590 
591 	return 0;
592 }
593 
594 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
595 {
596 	struct rte_pci_device *pdev = bp->pdev;
597 	char type[RTE_MEMZONE_NAMESIZE];
598 	uint16_t max_fc;
599 	int rc = 0;
600 
601 	max_fc = bp->flow_stat->max_fc;
602 
603 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
604 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
605 	/* 4 bytes for each counter-id */
606 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
607 				    max_fc * 4,
608 				    &bp->flow_stat->rx_fc_in_tbl);
609 	if (rc)
610 		return rc;
611 
612 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
613 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
614 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
615 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
616 				    max_fc * 16,
617 				    &bp->flow_stat->rx_fc_out_tbl);
618 	if (rc)
619 		return rc;
620 
621 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
622 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
623 	/* 4 bytes for each counter-id */
624 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
625 				    max_fc * 4,
626 				    &bp->flow_stat->tx_fc_in_tbl);
627 	if (rc)
628 		return rc;
629 
630 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
631 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
632 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
633 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
634 				    max_fc * 16,
635 				    &bp->flow_stat->tx_fc_out_tbl);
636 	if (rc)
637 		return rc;
638 
639 	rc = bnxt_register_fc_ctx_mem(bp);
640 
641 	return rc;
642 }
643 
644 static int bnxt_init_ctx_mem(struct bnxt *bp)
645 {
646 	int rc = 0;
647 
648 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
649 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
650 	    !BNXT_FLOW_XSTATS_EN(bp))
651 		return 0;
652 
653 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
654 	if (rc)
655 		return rc;
656 
657 	rc = bnxt_init_fc_ctx_mem(bp);
658 
659 	return rc;
660 }
661 
662 static int bnxt_update_phy_setting(struct bnxt *bp)
663 {
664 	struct rte_eth_link new;
665 	int rc;
666 
667 	rc = bnxt_get_hwrm_link_config(bp, &new);
668 	if (rc) {
669 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
670 		return rc;
671 	}
672 
673 	/*
674 	 * On BCM957508-N2100 adapters, FW will not allow any user other
675 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
676 	 * always returns link up. Force phy update always in that case.
677 	 */
678 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
679 		rc = bnxt_set_hwrm_link_config(bp, true);
680 		if (rc) {
681 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
682 			return rc;
683 		}
684 	}
685 
686 	return rc;
687 }
688 
689 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
690 {
691 	rte_free(bp->prev_rx_ring_stats);
692 	rte_free(bp->prev_tx_ring_stats);
693 
694 	bp->prev_rx_ring_stats = NULL;
695 	bp->prev_tx_ring_stats = NULL;
696 }
697 
698 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
699 {
700 	bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
701 					      sizeof(struct bnxt_ring_stats) *
702 					      bp->rx_cp_nr_rings,
703 					      0);
704 	if (bp->prev_rx_ring_stats == NULL)
705 		return -ENOMEM;
706 
707 	bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
708 					     sizeof(struct bnxt_ring_stats) *
709 					     bp->tx_cp_nr_rings,
710 					     0);
711 	if (bp->prev_tx_ring_stats == NULL)
712 		goto error;
713 
714 	return 0;
715 
716 error:
717 	bnxt_free_prev_ring_stats(bp);
718 	return -ENOMEM;
719 }
720 
721 static int bnxt_start_nic(struct bnxt *bp)
722 {
723 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
724 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
725 	uint32_t intr_vector = 0;
726 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
727 	uint32_t vec = BNXT_MISC_VEC_ID;
728 	unsigned int i, j;
729 	int rc;
730 
731 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU)
732 		bp->flags |= BNXT_FLAG_JUMBO;
733 	else
734 		bp->flags &= ~BNXT_FLAG_JUMBO;
735 
736 	/* THOR does not support ring groups.
737 	 * But we will use the array to save RSS context IDs.
738 	 */
739 	if (BNXT_CHIP_P5(bp))
740 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
741 
742 	rc = bnxt_alloc_hwrm_rings(bp);
743 	if (rc) {
744 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
745 		goto err_out;
746 	}
747 
748 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
749 	if (rc) {
750 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
751 		goto err_out;
752 	}
753 
754 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
755 		goto skip_cosq_cfg;
756 
757 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
758 		if (bp->rx_cos_queue[i].id != 0xff) {
759 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
760 
761 			if (!vnic) {
762 				PMD_DRV_LOG(ERR,
763 					    "Num pools more than FW profile\n");
764 				rc = -EINVAL;
765 				goto err_out;
766 			}
767 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
768 			bp->rx_cosq_cnt++;
769 		}
770 	}
771 
772 skip_cosq_cfg:
773 	rc = bnxt_mq_rx_configure(bp);
774 	if (rc) {
775 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
776 		goto err_out;
777 	}
778 
779 	for (j = 0; j < bp->rx_nr_rings; j++) {
780 		struct bnxt_rx_queue *rxq = bp->rx_queues[j];
781 
782 		if (!rxq->rx_deferred_start) {
783 			bp->eth_dev->data->rx_queue_state[j] =
784 				RTE_ETH_QUEUE_STATE_STARTED;
785 			rxq->rx_started = true;
786 		}
787 	}
788 
789 	/* default vnic 0 */
790 	rc = bnxt_setup_one_vnic(bp, 0);
791 	if (rc)
792 		goto err_out;
793 	/* VNIC configuration */
794 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
795 		for (i = 1; i < bp->nr_vnics; i++) {
796 			rc = bnxt_setup_one_vnic(bp, i);
797 			if (rc)
798 				goto err_out;
799 		}
800 	}
801 
802 	for (j = 0; j < bp->tx_nr_rings; j++) {
803 		struct bnxt_tx_queue *txq = bp->tx_queues[j];
804 
805 		if (!txq->tx_deferred_start) {
806 			bp->eth_dev->data->tx_queue_state[j] =
807 				RTE_ETH_QUEUE_STATE_STARTED;
808 			txq->tx_started = true;
809 		}
810 	}
811 
812 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
813 	if (rc) {
814 		PMD_DRV_LOG(ERR,
815 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
816 		goto err_out;
817 	}
818 
819 	/* check and configure queue intr-vector mapping */
820 	if ((rte_intr_cap_multiple(intr_handle) ||
821 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
822 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
823 		intr_vector = bp->eth_dev->data->nb_rx_queues;
824 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
825 		if (intr_vector > bp->rx_cp_nr_rings) {
826 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
827 					bp->rx_cp_nr_rings);
828 			return -ENOTSUP;
829 		}
830 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
831 		if (rc)
832 			return rc;
833 	}
834 
835 	if (rte_intr_dp_is_en(intr_handle)) {
836 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
837 					bp->eth_dev->data->nb_rx_queues)) {
838 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
839 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
840 			rc = -ENOMEM;
841 			goto err_out;
842 		}
843 		PMD_DRV_LOG(DEBUG, "intr_handle->nb_efd = %d "
844 			    "intr_handle->max_intr = %d\n",
845 			    rte_intr_nb_efd_get(intr_handle),
846 			    rte_intr_max_intr_get(intr_handle));
847 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
848 		     queue_id++) {
849 			rte_intr_vec_list_index_set(intr_handle,
850 					queue_id, vec + BNXT_RX_VEC_START);
851 			if (vec < base + rte_intr_nb_efd_get(intr_handle)
852 			    - 1)
853 				vec++;
854 		}
855 	}
856 
857 	/* enable uio/vfio intr/eventfd mapping */
858 	rc = rte_intr_enable(intr_handle);
859 #ifndef RTE_EXEC_ENV_FREEBSD
860 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
861 	if (rc)
862 		goto err_out;
863 #endif
864 
865 	rc = bnxt_update_phy_setting(bp);
866 	if (rc)
867 		goto err_out;
868 
869 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
870 	if (!bp->mark_table)
871 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
872 
873 	return 0;
874 
875 err_out:
876 	/* Some of the error status returned by FW may not be from errno.h */
877 	if (rc > 0)
878 		rc = -EIO;
879 
880 	return rc;
881 }
882 
883 static int bnxt_shutdown_nic(struct bnxt *bp)
884 {
885 	bnxt_free_all_hwrm_resources(bp);
886 	bnxt_free_all_filters(bp);
887 	bnxt_free_all_vnics(bp);
888 	return 0;
889 }
890 
891 /*
892  * Device configuration and status function
893  */
894 
895 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
896 {
897 	uint32_t link_speed = 0;
898 	uint32_t speed_capa = 0;
899 
900 	if (bp->link_info == NULL)
901 		return 0;
902 
903 	link_speed = bp->link_info->support_speeds;
904 
905 	/* If PAM4 is configured, use PAM4 supported speed */
906 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
907 		link_speed = bp->link_info->support_pam4_speeds;
908 
909 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
910 		speed_capa |= RTE_ETH_LINK_SPEED_100M;
911 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
912 		speed_capa |= RTE_ETH_LINK_SPEED_100M_HD;
913 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
914 		speed_capa |= RTE_ETH_LINK_SPEED_1G;
915 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
916 		speed_capa |= RTE_ETH_LINK_SPEED_2_5G;
917 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
918 		speed_capa |= RTE_ETH_LINK_SPEED_10G;
919 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
920 		speed_capa |= RTE_ETH_LINK_SPEED_20G;
921 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
922 		speed_capa |= RTE_ETH_LINK_SPEED_25G;
923 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
924 		speed_capa |= RTE_ETH_LINK_SPEED_40G;
925 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
926 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
927 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
928 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
929 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
930 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
931 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
932 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
933 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
934 		speed_capa |= RTE_ETH_LINK_SPEED_200G;
935 
936 	if (bp->link_info->auto_mode ==
937 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
938 		speed_capa |= RTE_ETH_LINK_SPEED_FIXED;
939 
940 	return speed_capa;
941 }
942 
943 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
944 				struct rte_eth_dev_info *dev_info)
945 {
946 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
947 	struct bnxt *bp = eth_dev->data->dev_private;
948 	uint16_t max_vnics, i, j, vpool, vrxq;
949 	unsigned int max_rx_rings;
950 	int rc;
951 
952 	rc = is_bnxt_in_error(bp);
953 	if (rc)
954 		return rc;
955 
956 	/* MAC Specifics */
957 	dev_info->max_mac_addrs = bp->max_l2_ctx;
958 	dev_info->max_hash_mac_addrs = 0;
959 
960 	/* PF/VF specifics */
961 	if (BNXT_PF(bp))
962 		dev_info->max_vfs = pdev->max_vfs;
963 
964 	max_rx_rings = bnxt_max_rings(bp);
965 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
966 	dev_info->max_rx_queues = max_rx_rings;
967 	dev_info->max_tx_queues = max_rx_rings;
968 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
969 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
970 	max_vnics = bp->max_vnics;
971 
972 	/* MTU specifics */
973 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
974 	dev_info->max_mtu = BNXT_MAX_MTU;
975 
976 	/* Fast path specifics */
977 	dev_info->min_rx_bufsize = 1;
978 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
979 
980 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
981 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
982 		dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_TIMESTAMP;
983 	if (bp->vnic_cap_flags & BNXT_VNIC_CAP_VLAN_RX_STRIP)
984 		dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
985 	dev_info->tx_queue_offload_capa = RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
986 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
987 				    dev_info->tx_queue_offload_capa;
988 	if (bp->fw_cap & BNXT_FW_CAP_VLAN_TX_INSERT)
989 		dev_info->tx_offload_capa |= RTE_ETH_TX_OFFLOAD_VLAN_INSERT;
990 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
991 
992 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
993 	dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
994 			     RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
995 	dev_info->dev_capa &= ~RTE_ETH_DEV_CAPA_FLOW_RULE_KEEP;
996 
997 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
998 		.rx_thresh = {
999 			.pthresh = 8,
1000 			.hthresh = 8,
1001 			.wthresh = 0,
1002 		},
1003 		.rx_free_thresh = 32,
1004 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
1005 	};
1006 
1007 	dev_info->default_txconf = (struct rte_eth_txconf) {
1008 		.tx_thresh = {
1009 			.pthresh = 32,
1010 			.hthresh = 0,
1011 			.wthresh = 0,
1012 		},
1013 		.tx_free_thresh = 32,
1014 		.tx_rs_thresh = 32,
1015 	};
1016 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
1017 
1018 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1019 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1020 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1021 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1022 
1023 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1024 		dev_info->switch_info.name = eth_dev->device->name;
1025 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1026 		dev_info->switch_info.port_id =
1027 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1028 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1029 	}
1030 
1031 	/*
1032 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1033 	 *       need further investigation.
1034 	 */
1035 
1036 	/* VMDq resources */
1037 	vpool = 64; /* RTE_ETH_64_POOLS */
1038 	vrxq = 128; /* RTE_ETH_VMDQ_DCB_NUM_QUEUES */
1039 	for (i = 0; i < 4; vpool >>= 1, i++) {
1040 		if (max_vnics > vpool) {
1041 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1042 				if (dev_info->max_rx_queues > vrxq) {
1043 					if (vpool > vrxq)
1044 						vpool = vrxq;
1045 					goto found;
1046 				}
1047 			}
1048 			/* Not enough resources to support VMDq */
1049 			break;
1050 		}
1051 	}
1052 	/* Not enough resources to support VMDq */
1053 	vpool = 0;
1054 	vrxq = 0;
1055 found:
1056 	dev_info->max_vmdq_pools = vpool;
1057 	dev_info->vmdq_queue_num = vrxq;
1058 
1059 	dev_info->vmdq_pool_base = 0;
1060 	dev_info->vmdq_queue_base = 0;
1061 
1062 	return 0;
1063 }
1064 
1065 /* Configure the device based on the configuration provided */
1066 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1067 {
1068 	struct bnxt *bp = eth_dev->data->dev_private;
1069 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1070 	int rc;
1071 
1072 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1073 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1074 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1075 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1076 
1077 	rc = is_bnxt_in_error(bp);
1078 	if (rc)
1079 		return rc;
1080 
1081 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1082 		rc = bnxt_hwrm_check_vf_rings(bp);
1083 		if (rc) {
1084 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1085 			return -ENOSPC;
1086 		}
1087 
1088 		/* If a resource has already been allocated - in this case
1089 		 * it is the async completion ring, free it. Reallocate it after
1090 		 * resource reservation. This will ensure the resource counts
1091 		 * are calculated correctly.
1092 		 */
1093 
1094 		pthread_mutex_lock(&bp->def_cp_lock);
1095 
1096 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1097 			bnxt_disable_int(bp);
1098 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1099 		}
1100 
1101 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1102 		if (rc) {
1103 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1104 			pthread_mutex_unlock(&bp->def_cp_lock);
1105 			return -ENOSPC;
1106 		}
1107 
1108 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1109 			rc = bnxt_alloc_async_cp_ring(bp);
1110 			if (rc) {
1111 				pthread_mutex_unlock(&bp->def_cp_lock);
1112 				return rc;
1113 			}
1114 			bnxt_enable_int(bp);
1115 		}
1116 
1117 		pthread_mutex_unlock(&bp->def_cp_lock);
1118 	}
1119 
1120 	/* Inherit new configurations */
1121 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1122 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1123 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1124 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1125 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1126 	    bp->max_stat_ctx)
1127 		goto resource_error;
1128 
1129 	if (BNXT_HAS_RING_GRPS(bp) &&
1130 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1131 		goto resource_error;
1132 
1133 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS) &&
1134 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1135 		goto resource_error;
1136 
1137 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1138 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1139 
1140 	if (eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
1141 		rx_offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
1142 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1143 
1144 	bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1145 
1146 	return 0;
1147 
1148 resource_error:
1149 	PMD_DRV_LOG(ERR,
1150 		    "Insufficient resources to support requested config\n");
1151 	PMD_DRV_LOG(ERR,
1152 		    "Num Queues Requested: Tx %d, Rx %d\n",
1153 		    eth_dev->data->nb_tx_queues,
1154 		    eth_dev->data->nb_rx_queues);
1155 	PMD_DRV_LOG(ERR,
1156 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1157 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1158 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1159 	return -ENOSPC;
1160 }
1161 
1162 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1163 {
1164 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1165 
1166 	if (link->link_status)
1167 		PMD_DRV_LOG(DEBUG, "Port %d Link Up - speed %u Mbps - %s\n",
1168 			eth_dev->data->port_id,
1169 			(uint32_t)link->link_speed,
1170 			(link->link_duplex == RTE_ETH_LINK_FULL_DUPLEX) ?
1171 			("full-duplex") : ("half-duplex\n"));
1172 	else
1173 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1174 			eth_dev->data->port_id);
1175 }
1176 
1177 /*
1178  * Determine whether the current configuration requires support for scattered
1179  * receive; return 1 if scattered receive is required and 0 if not.
1180  */
1181 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1182 {
1183 	uint32_t overhead = BNXT_MAX_PKT_LEN - BNXT_MAX_MTU;
1184 	uint16_t buf_size;
1185 	int i;
1186 
1187 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_SCATTER)
1188 		return 1;
1189 
1190 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_TCP_LRO)
1191 		return 1;
1192 
1193 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1194 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1195 
1196 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1197 				      RTE_PKTMBUF_HEADROOM);
1198 		if (eth_dev->data->mtu + overhead > buf_size)
1199 			return 1;
1200 	}
1201 	return 0;
1202 }
1203 
1204 static eth_rx_burst_t
1205 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1206 {
1207 	struct bnxt *bp = eth_dev->data->dev_private;
1208 
1209 	/* Disable vector mode RX for Stingray2 for now */
1210 	if (BNXT_CHIP_SR2(bp)) {
1211 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1212 		return bnxt_recv_pkts;
1213 	}
1214 
1215 #if (defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)) && \
1216 	!defined(RTE_LIBRTE_IEEE1588)
1217 
1218 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1219 	if (eth_dev->data->scattered_rx)
1220 		goto use_scalar_rx;
1221 
1222 	/*
1223 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1224 	 * asynchronous completions and receive completions can be placed in
1225 	 * the same completion ring.
1226 	 */
1227 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1228 		goto use_scalar_rx;
1229 
1230 	/*
1231 	 * Vector mode receive cannot be enabled if any receive offloads outside
1232 	 * a limited subset have been enabled.
1233 	 */
1234 	if (eth_dev->data->dev_conf.rxmode.offloads &
1235 		~(RTE_ETH_RX_OFFLOAD_VLAN_STRIP |
1236 		  RTE_ETH_RX_OFFLOAD_KEEP_CRC |
1237 		  RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
1238 		  RTE_ETH_RX_OFFLOAD_UDP_CKSUM |
1239 		  RTE_ETH_RX_OFFLOAD_TCP_CKSUM |
1240 		  RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1241 		  RTE_ETH_RX_OFFLOAD_OUTER_UDP_CKSUM |
1242 		  RTE_ETH_RX_OFFLOAD_RSS_HASH |
1243 		  RTE_ETH_RX_OFFLOAD_VLAN_FILTER))
1244 		goto use_scalar_rx;
1245 
1246 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1247 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1248 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1249 		PMD_DRV_LOG(INFO,
1250 			    "Using AVX2 vector mode receive for port %d\n",
1251 			    eth_dev->data->port_id);
1252 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1253 		return bnxt_recv_pkts_vec_avx2;
1254 	}
1255  #endif
1256 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1257 		PMD_DRV_LOG(INFO,
1258 			    "Using SSE vector mode receive for port %d\n",
1259 			    eth_dev->data->port_id);
1260 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1261 		return bnxt_recv_pkts_vec;
1262 	}
1263 
1264 use_scalar_rx:
1265 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1266 		    eth_dev->data->port_id);
1267 	PMD_DRV_LOG(INFO,
1268 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1269 		    eth_dev->data->port_id,
1270 		    eth_dev->data->scattered_rx,
1271 		    eth_dev->data->dev_conf.rxmode.offloads);
1272 #endif
1273 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1274 	return bnxt_recv_pkts;
1275 }
1276 
1277 static eth_tx_burst_t
1278 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1279 {
1280 	struct bnxt *bp = eth_dev->data->dev_private;
1281 
1282 	/* Disable vector mode TX for Stingray2 for now */
1283 	if (BNXT_CHIP_SR2(bp))
1284 		return bnxt_xmit_pkts;
1285 
1286 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64) && \
1287 	!defined(RTE_LIBRTE_IEEE1588)
1288 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1289 
1290 	/*
1291 	 * Vector mode transmit can be enabled only if not using scatter rx
1292 	 * or tx offloads.
1293 	 */
1294 	if (eth_dev->data->scattered_rx ||
1295 	    (offloads & ~RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE) ||
1296 	    BNXT_TRUFLOW_EN(bp))
1297 		goto use_scalar_tx;
1298 
1299 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1300 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1301 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1302 		PMD_DRV_LOG(INFO,
1303 			    "Using AVX2 vector mode transmit for port %d\n",
1304 			    eth_dev->data->port_id);
1305 		return bnxt_xmit_pkts_vec_avx2;
1306 	}
1307 #endif
1308 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1309 		PMD_DRV_LOG(INFO,
1310 			    "Using SSE vector mode transmit for port %d\n",
1311 			    eth_dev->data->port_id);
1312 		return bnxt_xmit_pkts_vec;
1313 	}
1314 
1315 use_scalar_tx:
1316 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1317 		    eth_dev->data->port_id);
1318 	PMD_DRV_LOG(INFO,
1319 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1320 		    eth_dev->data->port_id,
1321 		    eth_dev->data->scattered_rx,
1322 		    offloads);
1323 #endif
1324 	return bnxt_xmit_pkts;
1325 }
1326 
1327 static int bnxt_handle_if_change_status(struct bnxt *bp)
1328 {
1329 	int rc;
1330 
1331 	/* Since fw has undergone a reset and lost all contexts,
1332 	 * set fatal flag to not issue hwrm during cleanup
1333 	 */
1334 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1335 	bnxt_uninit_resources(bp, true);
1336 
1337 	/* clear fatal flag so that re-init happens */
1338 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1339 	rc = bnxt_init_resources(bp, true);
1340 
1341 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1342 
1343 	return rc;
1344 }
1345 
1346 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1347 {
1348 	struct bnxt *bp = eth_dev->data->dev_private;
1349 	int rc = 0;
1350 
1351 	if (!BNXT_SINGLE_PF(bp))
1352 		return -ENOTSUP;
1353 
1354 	if (!bp->link_info->link_up)
1355 		rc = bnxt_set_hwrm_link_config(bp, true);
1356 	if (!rc)
1357 		eth_dev->data->dev_link.link_status = 1;
1358 
1359 	bnxt_print_link_info(eth_dev);
1360 	return rc;
1361 }
1362 
1363 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1364 {
1365 	struct bnxt *bp = eth_dev->data->dev_private;
1366 
1367 	if (!BNXT_SINGLE_PF(bp))
1368 		return -ENOTSUP;
1369 
1370 	eth_dev->data->dev_link.link_status = 0;
1371 	bnxt_set_hwrm_link_config(bp, false);
1372 	bp->link_info->link_up = 0;
1373 
1374 	return 0;
1375 }
1376 
1377 static void bnxt_free_switch_domain(struct bnxt *bp)
1378 {
1379 	int rc = 0;
1380 
1381 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1382 		return;
1383 
1384 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1385 	if (rc)
1386 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1387 			    bp->switch_domain_id, rc);
1388 }
1389 
1390 static void bnxt_ptp_get_current_time(void *arg)
1391 {
1392 	struct bnxt *bp = arg;
1393 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1394 	int rc;
1395 
1396 	rc = is_bnxt_in_error(bp);
1397 	if (rc)
1398 		return;
1399 
1400 	if (!ptp)
1401 		return;
1402 
1403 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1404 				&ptp->current_time);
1405 
1406 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1407 	if (rc != 0) {
1408 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1409 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1410 	}
1411 }
1412 
1413 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1414 {
1415 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1416 	int rc;
1417 
1418 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1419 		return 0;
1420 
1421 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1422 				&ptp->current_time);
1423 
1424 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1425 	return rc;
1426 }
1427 
1428 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1429 {
1430 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1431 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1432 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1433 	}
1434 }
1435 
1436 static void bnxt_ptp_stop(struct bnxt *bp)
1437 {
1438 	bnxt_cancel_ptp_alarm(bp);
1439 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1440 }
1441 
1442 static int bnxt_ptp_start(struct bnxt *bp)
1443 {
1444 	int rc;
1445 
1446 	rc = bnxt_schedule_ptp_alarm(bp);
1447 	if (rc != 0) {
1448 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1449 	} else {
1450 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1451 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1452 	}
1453 
1454 	return rc;
1455 }
1456 
1457 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1458 {
1459 	struct bnxt *bp = eth_dev->data->dev_private;
1460 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1461 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1462 	struct rte_eth_link link;
1463 	int ret;
1464 
1465 	eth_dev->data->dev_started = 0;
1466 
1467 	/* Prevent crashes when queues are still in use */
1468 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1469 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1470 
1471 	bnxt_disable_int(bp);
1472 
1473 	/* disable uio/vfio intr/eventfd mapping */
1474 	rte_intr_disable(intr_handle);
1475 
1476 	/* Stop the child representors for this device */
1477 	ret = bnxt_rep_stop_all(bp);
1478 	if (ret != 0)
1479 		return ret;
1480 
1481 	/* delete the bnxt ULP port details */
1482 	bnxt_ulp_port_deinit(bp);
1483 
1484 	bnxt_cancel_fw_health_check(bp);
1485 
1486 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1487 		bnxt_cancel_ptp_alarm(bp);
1488 
1489 	/* Do not bring link down during reset recovery */
1490 	if (!is_bnxt_in_error(bp)) {
1491 		bnxt_dev_set_link_down_op(eth_dev);
1492 		/* Wait for link to be reset */
1493 		if (BNXT_SINGLE_PF(bp))
1494 			rte_delay_ms(500);
1495 		/* clear the recorded link status */
1496 		memset(&link, 0, sizeof(link));
1497 		rte_eth_linkstatus_set(eth_dev, &link);
1498 	}
1499 
1500 	/* Clean queue intr-vector mapping */
1501 	rte_intr_efd_disable(intr_handle);
1502 	rte_intr_vec_list_free(intr_handle);
1503 
1504 	bnxt_hwrm_port_clr_stats(bp);
1505 	bnxt_free_tx_mbufs(bp);
1506 	bnxt_free_rx_mbufs(bp);
1507 	/* Process any remaining notifications in default completion queue */
1508 	bnxt_int_handler(eth_dev);
1509 	bnxt_shutdown_nic(bp);
1510 	bnxt_hwrm_if_change(bp, false);
1511 
1512 	bnxt_free_prev_ring_stats(bp);
1513 	rte_free(bp->mark_table);
1514 	bp->mark_table = NULL;
1515 
1516 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1517 	bp->rx_cosq_cnt = 0;
1518 	/* All filters are deleted on a port stop. */
1519 	if (BNXT_FLOW_XSTATS_EN(bp))
1520 		bp->flow_stat->flow_count = 0;
1521 
1522 	eth_dev->data->scattered_rx = 0;
1523 
1524 	return 0;
1525 }
1526 
1527 /* Unload the driver, release resources */
1528 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1529 {
1530 	struct bnxt *bp = eth_dev->data->dev_private;
1531 
1532 	pthread_mutex_lock(&bp->err_recovery_lock);
1533 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1534 		PMD_DRV_LOG(ERR,
1535 			    "Adapter recovering from error..Please retry\n");
1536 		pthread_mutex_unlock(&bp->err_recovery_lock);
1537 		return -EAGAIN;
1538 	}
1539 	pthread_mutex_unlock(&bp->err_recovery_lock);
1540 
1541 	return bnxt_dev_stop(eth_dev);
1542 }
1543 
1544 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1545 {
1546 	struct bnxt *bp = eth_dev->data->dev_private;
1547 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1548 	int vlan_mask = 0;
1549 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1550 
1551 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1552 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1553 		return -EINVAL;
1554 	}
1555 
1556 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1557 		PMD_DRV_LOG(ERR,
1558 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1559 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1560 
1561 	do {
1562 		rc = bnxt_hwrm_if_change(bp, true);
1563 		if (rc == 0 || rc != -EAGAIN)
1564 			break;
1565 
1566 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1567 	} while (retry_cnt--);
1568 
1569 	if (rc)
1570 		return rc;
1571 
1572 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1573 		rc = bnxt_handle_if_change_status(bp);
1574 		if (rc)
1575 			return rc;
1576 	}
1577 
1578 	bnxt_enable_int(bp);
1579 
1580 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1581 
1582 	rc = bnxt_start_nic(bp);
1583 	if (rc)
1584 		goto error;
1585 
1586 	rc = bnxt_alloc_prev_ring_stats(bp);
1587 	if (rc)
1588 		goto error;
1589 
1590 	eth_dev->data->dev_started = 1;
1591 
1592 	bnxt_link_update_op(eth_dev, 1);
1593 
1594 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
1595 		vlan_mask |= RTE_ETH_VLAN_FILTER_MASK;
1596 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
1597 		vlan_mask |= RTE_ETH_VLAN_STRIP_MASK;
1598 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1599 	if (rc)
1600 		goto error;
1601 
1602 	/* Initialize bnxt ULP port details */
1603 	rc = bnxt_ulp_port_init(bp);
1604 	if (rc)
1605 		goto error;
1606 
1607 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1608 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1609 
1610 	bnxt_schedule_fw_health_check(bp);
1611 
1612 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1613 		bnxt_schedule_ptp_alarm(bp);
1614 
1615 	return 0;
1616 
1617 error:
1618 	bnxt_dev_stop(eth_dev);
1619 	return rc;
1620 }
1621 
1622 static void
1623 bnxt_uninit_locks(struct bnxt *bp)
1624 {
1625 	pthread_mutex_destroy(&bp->flow_lock);
1626 	pthread_mutex_destroy(&bp->def_cp_lock);
1627 	pthread_mutex_destroy(&bp->health_check_lock);
1628 	pthread_mutex_destroy(&bp->err_recovery_lock);
1629 	if (bp->rep_info) {
1630 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1631 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1632 	}
1633 }
1634 
1635 static void bnxt_drv_uninit(struct bnxt *bp)
1636 {
1637 	bnxt_free_leds_info(bp);
1638 	bnxt_free_cos_queues(bp);
1639 	bnxt_free_link_info(bp);
1640 	bnxt_free_parent_info(bp);
1641 	bnxt_uninit_locks(bp);
1642 
1643 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1644 	bp->tx_mem_zone = NULL;
1645 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1646 	bp->rx_mem_zone = NULL;
1647 
1648 	bnxt_free_vf_info(bp);
1649 	bnxt_free_pf_info(bp);
1650 
1651 	rte_free(bp->grp_info);
1652 	bp->grp_info = NULL;
1653 }
1654 
1655 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1656 {
1657 	struct bnxt *bp = eth_dev->data->dev_private;
1658 	int ret = 0;
1659 
1660 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1661 		return 0;
1662 
1663 	pthread_mutex_lock(&bp->err_recovery_lock);
1664 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1665 		PMD_DRV_LOG(ERR,
1666 			    "Adapter recovering from error...Please retry\n");
1667 		pthread_mutex_unlock(&bp->err_recovery_lock);
1668 		return -EAGAIN;
1669 	}
1670 	pthread_mutex_unlock(&bp->err_recovery_lock);
1671 
1672 	/* cancel the recovery handler before remove dev */
1673 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1674 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1675 	bnxt_cancel_fc_thread(bp);
1676 
1677 	if (eth_dev->data->dev_started)
1678 		ret = bnxt_dev_stop(eth_dev);
1679 
1680 	bnxt_uninit_resources(bp, false);
1681 
1682 	bnxt_drv_uninit(bp);
1683 
1684 	return ret;
1685 }
1686 
1687 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1688 				    uint32_t index)
1689 {
1690 	struct bnxt *bp = eth_dev->data->dev_private;
1691 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1692 	struct bnxt_vnic_info *vnic;
1693 	struct bnxt_filter_info *filter, *temp_filter;
1694 	uint32_t i;
1695 
1696 	if (is_bnxt_in_error(bp))
1697 		return;
1698 
1699 	/*
1700 	 * Loop through all VNICs from the specified filter flow pools to
1701 	 * remove the corresponding MAC addr filter
1702 	 */
1703 	for (i = 0; i < bp->nr_vnics; i++) {
1704 		if (!(pool_mask & (1ULL << i)))
1705 			continue;
1706 
1707 		vnic = &bp->vnic_info[i];
1708 		filter = STAILQ_FIRST(&vnic->filter);
1709 		while (filter) {
1710 			temp_filter = STAILQ_NEXT(filter, next);
1711 			if (filter->mac_index == index) {
1712 				STAILQ_REMOVE(&vnic->filter, filter,
1713 						bnxt_filter_info, next);
1714 				bnxt_hwrm_clear_l2_filter(bp, filter);
1715 				bnxt_free_filter(bp, filter);
1716 			}
1717 			filter = temp_filter;
1718 		}
1719 	}
1720 }
1721 
1722 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1723 			       struct rte_ether_addr *mac_addr, uint32_t index,
1724 			       uint32_t pool)
1725 {
1726 	struct bnxt_filter_info *filter;
1727 	int rc = 0;
1728 
1729 	/* Attach requested MAC address to the new l2_filter */
1730 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1731 		if (filter->mac_index == index) {
1732 			PMD_DRV_LOG(DEBUG,
1733 				    "MAC addr already existed for pool %d\n",
1734 				    pool);
1735 			return 0;
1736 		}
1737 	}
1738 
1739 	filter = bnxt_alloc_filter(bp);
1740 	if (!filter) {
1741 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1742 		return -ENODEV;
1743 	}
1744 
1745 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1746 	 * if the MAC that's been programmed now is a different one, then,
1747 	 * copy that addr to filter->l2_addr
1748 	 */
1749 	if (mac_addr)
1750 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1751 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1752 
1753 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1754 	if (!rc) {
1755 		filter->mac_index = index;
1756 		if (filter->mac_index == 0)
1757 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1758 		else
1759 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1760 	} else {
1761 		bnxt_free_filter(bp, filter);
1762 	}
1763 
1764 	return rc;
1765 }
1766 
1767 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1768 				struct rte_ether_addr *mac_addr,
1769 				uint32_t index, uint32_t pool)
1770 {
1771 	struct bnxt *bp = eth_dev->data->dev_private;
1772 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1773 	int rc = 0;
1774 
1775 	rc = is_bnxt_in_error(bp);
1776 	if (rc)
1777 		return rc;
1778 
1779 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1780 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1781 		return -ENOTSUP;
1782 	}
1783 
1784 	if (!vnic) {
1785 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1786 		return -EINVAL;
1787 	}
1788 
1789 	/* Filter settings will get applied when port is started */
1790 	if (!eth_dev->data->dev_started)
1791 		return 0;
1792 
1793 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1794 
1795 	return rc;
1796 }
1797 
1798 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1799 {
1800 	int rc = 0;
1801 	struct bnxt *bp = eth_dev->data->dev_private;
1802 	struct rte_eth_link new;
1803 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1804 			BNXT_MIN_LINK_WAIT_CNT;
1805 
1806 	rc = is_bnxt_in_error(bp);
1807 	if (rc)
1808 		return rc;
1809 
1810 	memset(&new, 0, sizeof(new));
1811 
1812 	if (bp->link_info == NULL)
1813 		goto out;
1814 
1815 	do {
1816 		/* Retrieve link info from hardware */
1817 		rc = bnxt_get_hwrm_link_config(bp, &new);
1818 		if (rc) {
1819 			new.link_speed = RTE_ETH_LINK_SPEED_100M;
1820 			new.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
1821 			PMD_DRV_LOG(ERR,
1822 				"Failed to retrieve link rc = 0x%x!\n", rc);
1823 			goto out;
1824 		}
1825 
1826 		if (!wait_to_complete || new.link_status)
1827 			break;
1828 
1829 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1830 	} while (cnt--);
1831 
1832 	/* Only single function PF can bring phy down.
1833 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1834 	 */
1835 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1836 		memset(&new, 0, sizeof(new));
1837 
1838 out:
1839 	/* Timed out or success */
1840 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1841 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1842 		rte_eth_linkstatus_set(eth_dev, &new);
1843 		bnxt_print_link_info(eth_dev);
1844 	}
1845 
1846 	return rc;
1847 }
1848 
1849 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1850 {
1851 	struct bnxt *bp = eth_dev->data->dev_private;
1852 	struct bnxt_vnic_info *vnic;
1853 	uint32_t old_flags;
1854 	int rc;
1855 
1856 	rc = is_bnxt_in_error(bp);
1857 	if (rc)
1858 		return rc;
1859 
1860 	/* Filter settings will get applied when port is started */
1861 	if (!eth_dev->data->dev_started)
1862 		return 0;
1863 
1864 	if (bp->vnic_info == NULL)
1865 		return 0;
1866 
1867 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1868 
1869 	old_flags = vnic->flags;
1870 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1871 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1872 	if (rc != 0)
1873 		vnic->flags = old_flags;
1874 
1875 	return rc;
1876 }
1877 
1878 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1879 {
1880 	struct bnxt *bp = eth_dev->data->dev_private;
1881 	struct bnxt_vnic_info *vnic;
1882 	uint32_t old_flags;
1883 	int rc;
1884 
1885 	rc = is_bnxt_in_error(bp);
1886 	if (rc)
1887 		return rc;
1888 
1889 	/* Filter settings will get applied when port is started */
1890 	if (!eth_dev->data->dev_started)
1891 		return 0;
1892 
1893 	if (bp->vnic_info == NULL)
1894 		return 0;
1895 
1896 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1897 
1898 	old_flags = vnic->flags;
1899 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1900 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1901 	if (rc != 0)
1902 		vnic->flags = old_flags;
1903 
1904 	return rc;
1905 }
1906 
1907 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1908 {
1909 	struct bnxt *bp = eth_dev->data->dev_private;
1910 	struct bnxt_vnic_info *vnic;
1911 	uint32_t old_flags;
1912 	int rc;
1913 
1914 	rc = is_bnxt_in_error(bp);
1915 	if (rc)
1916 		return rc;
1917 
1918 	/* Filter settings will get applied when port is started */
1919 	if (!eth_dev->data->dev_started)
1920 		return 0;
1921 
1922 	if (bp->vnic_info == NULL)
1923 		return 0;
1924 
1925 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1926 
1927 	old_flags = vnic->flags;
1928 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1929 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1930 	if (rc != 0)
1931 		vnic->flags = old_flags;
1932 
1933 	return rc;
1934 }
1935 
1936 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1937 {
1938 	struct bnxt *bp = eth_dev->data->dev_private;
1939 	struct bnxt_vnic_info *vnic;
1940 	uint32_t old_flags;
1941 	int rc;
1942 
1943 	rc = is_bnxt_in_error(bp);
1944 	if (rc)
1945 		return rc;
1946 
1947 	/* Filter settings will get applied when port is started */
1948 	if (!eth_dev->data->dev_started)
1949 		return 0;
1950 
1951 	if (bp->vnic_info == NULL)
1952 		return 0;
1953 
1954 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1955 
1956 	old_flags = vnic->flags;
1957 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1958 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1959 	if (rc != 0)
1960 		vnic->flags = old_flags;
1961 
1962 	return rc;
1963 }
1964 
1965 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1966 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1967 {
1968 	if (qid >= bp->rx_nr_rings)
1969 		return NULL;
1970 
1971 	return bp->eth_dev->data->rx_queues[qid];
1972 }
1973 
1974 /* Return rxq corresponding to a given rss table ring/group ID. */
1975 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1976 {
1977 	struct bnxt_rx_queue *rxq;
1978 	unsigned int i;
1979 
1980 	if (!BNXT_HAS_RING_GRPS(bp)) {
1981 		for (i = 0; i < bp->rx_nr_rings; i++) {
1982 			rxq = bp->eth_dev->data->rx_queues[i];
1983 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1984 				return rxq->index;
1985 		}
1986 	} else {
1987 		for (i = 0; i < bp->rx_nr_rings; i++) {
1988 			if (bp->grp_info[i].fw_grp_id == fwr)
1989 				return i;
1990 		}
1991 	}
1992 
1993 	return INVALID_HW_RING_ID;
1994 }
1995 
1996 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1997 			    struct rte_eth_rss_reta_entry64 *reta_conf,
1998 			    uint16_t reta_size)
1999 {
2000 	struct bnxt *bp = eth_dev->data->dev_private;
2001 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2002 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2003 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2004 	uint16_t idx, sft;
2005 	int i, rc;
2006 
2007 	rc = is_bnxt_in_error(bp);
2008 	if (rc)
2009 		return rc;
2010 
2011 	if (!vnic->rss_table)
2012 		return -EINVAL;
2013 
2014 	if (!(dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG))
2015 		return -EINVAL;
2016 
2017 	if (reta_size != tbl_size) {
2018 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2019 			"(%d) must equal the size supported by the hardware "
2020 			"(%d)\n", reta_size, tbl_size);
2021 		return -EINVAL;
2022 	}
2023 
2024 	for (i = 0; i < reta_size; i++) {
2025 		struct bnxt_rx_queue *rxq;
2026 
2027 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2028 		sft = i % RTE_ETH_RETA_GROUP_SIZE;
2029 
2030 		if (!(reta_conf[idx].mask & (1ULL << sft)))
2031 			continue;
2032 
2033 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2034 		if (!rxq) {
2035 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
2036 			return -EINVAL;
2037 		}
2038 
2039 		if (BNXT_CHIP_P5(bp)) {
2040 			vnic->rss_table[i * 2] =
2041 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
2042 			vnic->rss_table[i * 2 + 1] =
2043 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
2044 		} else {
2045 			vnic->rss_table[i] =
2046 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2047 		}
2048 	}
2049 
2050 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2051 	return rc;
2052 }
2053 
2054 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2055 			      struct rte_eth_rss_reta_entry64 *reta_conf,
2056 			      uint16_t reta_size)
2057 {
2058 	struct bnxt *bp = eth_dev->data->dev_private;
2059 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2060 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2061 	uint16_t idx, sft, i;
2062 	int rc;
2063 
2064 	rc = is_bnxt_in_error(bp);
2065 	if (rc)
2066 		return rc;
2067 
2068 	if (!vnic)
2069 		return -EINVAL;
2070 	if (!vnic->rss_table)
2071 		return -EINVAL;
2072 
2073 	if (reta_size != tbl_size) {
2074 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2075 			"(%d) must equal the size supported by the hardware "
2076 			"(%d)\n", reta_size, tbl_size);
2077 		return -EINVAL;
2078 	}
2079 
2080 	for (idx = 0, i = 0; i < reta_size; i++) {
2081 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2082 		sft = i % RTE_ETH_RETA_GROUP_SIZE;
2083 
2084 		if (reta_conf[idx].mask & (1ULL << sft)) {
2085 			uint16_t qid;
2086 
2087 			if (BNXT_CHIP_P5(bp))
2088 				qid = bnxt_rss_to_qid(bp,
2089 						      vnic->rss_table[i * 2]);
2090 			else
2091 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2092 
2093 			if (qid == INVALID_HW_RING_ID) {
2094 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2095 				return -EINVAL;
2096 			}
2097 			reta_conf[idx].reta[sft] = qid;
2098 		}
2099 	}
2100 
2101 	return 0;
2102 }
2103 
2104 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2105 				   struct rte_eth_rss_conf *rss_conf)
2106 {
2107 	struct bnxt *bp = eth_dev->data->dev_private;
2108 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2109 	struct bnxt_vnic_info *vnic;
2110 	int rc;
2111 
2112 	rc = is_bnxt_in_error(bp);
2113 	if (rc)
2114 		return rc;
2115 
2116 	/*
2117 	 * If RSS enablement were different than dev_configure,
2118 	 * then return -EINVAL
2119 	 */
2120 	if (dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG) {
2121 		if (!rss_conf->rss_hf)
2122 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2123 	} else {
2124 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2125 			return -EINVAL;
2126 	}
2127 
2128 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
2129 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2130 	       rss_conf,
2131 	       sizeof(*rss_conf));
2132 
2133 	/* Update the default RSS VNIC(s) */
2134 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2135 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2136 	vnic->hash_mode =
2137 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2138 					    RTE_ETH_RSS_LEVEL(rss_conf->rss_hf));
2139 
2140 	/*
2141 	 * If hashkey is not specified, use the previously configured
2142 	 * hashkey
2143 	 */
2144 	if (!rss_conf->rss_key)
2145 		goto rss_config;
2146 
2147 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2148 		PMD_DRV_LOG(ERR,
2149 			    "Invalid hashkey length, should be %d bytes\n",
2150 			    HW_HASH_KEY_SIZE);
2151 		return -EINVAL;
2152 	}
2153 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2154 
2155 rss_config:
2156 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2157 	return rc;
2158 }
2159 
2160 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2161 				     struct rte_eth_rss_conf *rss_conf)
2162 {
2163 	struct bnxt *bp = eth_dev->data->dev_private;
2164 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2165 	int len, rc;
2166 	uint32_t hash_types;
2167 
2168 	rc = is_bnxt_in_error(bp);
2169 	if (rc)
2170 		return rc;
2171 
2172 	/* RSS configuration is the same for all VNICs */
2173 	if (vnic && vnic->rss_hash_key) {
2174 		if (rss_conf->rss_key) {
2175 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2176 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2177 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2178 		}
2179 
2180 		hash_types = vnic->hash_type;
2181 		rss_conf->rss_hf = 0;
2182 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2183 			rss_conf->rss_hf |= RTE_ETH_RSS_IPV4;
2184 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2185 		}
2186 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2187 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_TCP;
2188 			hash_types &=
2189 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2190 		}
2191 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2192 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_UDP;
2193 			hash_types &=
2194 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2195 		}
2196 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2197 			rss_conf->rss_hf |= RTE_ETH_RSS_IPV6;
2198 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2199 		}
2200 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2201 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_TCP;
2202 			hash_types &=
2203 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2204 		}
2205 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2206 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_UDP;
2207 			hash_types &=
2208 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2209 		}
2210 
2211 		rss_conf->rss_hf |=
2212 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2213 
2214 		if (hash_types) {
2215 			PMD_DRV_LOG(ERR,
2216 				"Unknown RSS config from firmware (%08x), RSS disabled",
2217 				vnic->hash_type);
2218 			return -ENOTSUP;
2219 		}
2220 	} else {
2221 		rss_conf->rss_hf = 0;
2222 	}
2223 	return 0;
2224 }
2225 
2226 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2227 			       struct rte_eth_fc_conf *fc_conf)
2228 {
2229 	struct bnxt *bp = dev->data->dev_private;
2230 	struct rte_eth_link link_info;
2231 	int rc;
2232 
2233 	rc = is_bnxt_in_error(bp);
2234 	if (rc)
2235 		return rc;
2236 
2237 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2238 	if (rc)
2239 		return rc;
2240 
2241 	memset(fc_conf, 0, sizeof(*fc_conf));
2242 	if (bp->link_info->auto_pause)
2243 		fc_conf->autoneg = 1;
2244 	switch (bp->link_info->pause) {
2245 	case 0:
2246 		fc_conf->mode = RTE_ETH_FC_NONE;
2247 		break;
2248 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2249 		fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
2250 		break;
2251 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2252 		fc_conf->mode = RTE_ETH_FC_RX_PAUSE;
2253 		break;
2254 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2255 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2256 		fc_conf->mode = RTE_ETH_FC_FULL;
2257 		break;
2258 	}
2259 	return 0;
2260 }
2261 
2262 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2263 			       struct rte_eth_fc_conf *fc_conf)
2264 {
2265 	struct bnxt *bp = dev->data->dev_private;
2266 	int rc;
2267 
2268 	rc = is_bnxt_in_error(bp);
2269 	if (rc)
2270 		return rc;
2271 
2272 	if (!BNXT_SINGLE_PF(bp)) {
2273 		PMD_DRV_LOG(ERR,
2274 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2275 		return -ENOTSUP;
2276 	}
2277 
2278 	switch (fc_conf->mode) {
2279 	case RTE_ETH_FC_NONE:
2280 		bp->link_info->auto_pause = 0;
2281 		bp->link_info->force_pause = 0;
2282 		break;
2283 	case RTE_ETH_FC_RX_PAUSE:
2284 		if (fc_conf->autoneg) {
2285 			bp->link_info->auto_pause =
2286 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2287 			bp->link_info->force_pause = 0;
2288 		} else {
2289 			bp->link_info->auto_pause = 0;
2290 			bp->link_info->force_pause =
2291 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2292 		}
2293 		break;
2294 	case RTE_ETH_FC_TX_PAUSE:
2295 		if (fc_conf->autoneg) {
2296 			bp->link_info->auto_pause =
2297 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2298 			bp->link_info->force_pause = 0;
2299 		} else {
2300 			bp->link_info->auto_pause = 0;
2301 			bp->link_info->force_pause =
2302 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2303 		}
2304 		break;
2305 	case RTE_ETH_FC_FULL:
2306 		if (fc_conf->autoneg) {
2307 			bp->link_info->auto_pause =
2308 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2309 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2310 			bp->link_info->force_pause = 0;
2311 		} else {
2312 			bp->link_info->auto_pause = 0;
2313 			bp->link_info->force_pause =
2314 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2315 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2316 		}
2317 		break;
2318 	}
2319 	return bnxt_set_hwrm_link_config(bp, true);
2320 }
2321 
2322 /* Add UDP tunneling port */
2323 static int
2324 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2325 			 struct rte_eth_udp_tunnel *udp_tunnel)
2326 {
2327 	struct bnxt *bp = eth_dev->data->dev_private;
2328 	uint16_t tunnel_type = 0;
2329 	int rc = 0;
2330 
2331 	rc = is_bnxt_in_error(bp);
2332 	if (rc)
2333 		return rc;
2334 
2335 	switch (udp_tunnel->prot_type) {
2336 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
2337 		if (bp->vxlan_port_cnt) {
2338 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2339 				udp_tunnel->udp_port);
2340 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2341 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2342 				return -ENOSPC;
2343 			}
2344 			bp->vxlan_port_cnt++;
2345 			return 0;
2346 		}
2347 		tunnel_type =
2348 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2349 		break;
2350 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2351 		if (bp->geneve_port_cnt) {
2352 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2353 				udp_tunnel->udp_port);
2354 			if (bp->geneve_port != udp_tunnel->udp_port) {
2355 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2356 				return -ENOSPC;
2357 			}
2358 			bp->geneve_port_cnt++;
2359 			return 0;
2360 		}
2361 		tunnel_type =
2362 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2363 		break;
2364 	default:
2365 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2366 		return -ENOTSUP;
2367 	}
2368 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2369 					     tunnel_type);
2370 
2371 	if (rc != 0)
2372 		return rc;
2373 
2374 	if (tunnel_type ==
2375 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN)
2376 		bp->vxlan_port_cnt++;
2377 
2378 	if (tunnel_type ==
2379 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE)
2380 		bp->geneve_port_cnt++;
2381 
2382 	return rc;
2383 }
2384 
2385 static int
2386 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2387 			 struct rte_eth_udp_tunnel *udp_tunnel)
2388 {
2389 	struct bnxt *bp = eth_dev->data->dev_private;
2390 	uint16_t tunnel_type = 0;
2391 	uint16_t port = 0;
2392 	int rc = 0;
2393 
2394 	rc = is_bnxt_in_error(bp);
2395 	if (rc)
2396 		return rc;
2397 
2398 	switch (udp_tunnel->prot_type) {
2399 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
2400 		if (!bp->vxlan_port_cnt) {
2401 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2402 			return -EINVAL;
2403 		}
2404 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2405 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2406 				udp_tunnel->udp_port, bp->vxlan_port);
2407 			return -EINVAL;
2408 		}
2409 		if (--bp->vxlan_port_cnt)
2410 			return 0;
2411 
2412 		tunnel_type =
2413 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2414 		port = bp->vxlan_fw_dst_port_id;
2415 		break;
2416 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2417 		if (!bp->geneve_port_cnt) {
2418 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2419 			return -EINVAL;
2420 		}
2421 		if (bp->geneve_port != udp_tunnel->udp_port) {
2422 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2423 				udp_tunnel->udp_port, bp->geneve_port);
2424 			return -EINVAL;
2425 		}
2426 		if (--bp->geneve_port_cnt)
2427 			return 0;
2428 
2429 		tunnel_type =
2430 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2431 		port = bp->geneve_fw_dst_port_id;
2432 		break;
2433 	default:
2434 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2435 		return -ENOTSUP;
2436 	}
2437 
2438 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2439 	return rc;
2440 }
2441 
2442 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2443 {
2444 	struct bnxt_filter_info *filter;
2445 	struct bnxt_vnic_info *vnic;
2446 	int rc = 0;
2447 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2448 
2449 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2450 	filter = STAILQ_FIRST(&vnic->filter);
2451 	while (filter) {
2452 		/* Search for this matching MAC+VLAN filter */
2453 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2454 			/* Delete the filter */
2455 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2456 			if (rc)
2457 				return rc;
2458 			STAILQ_REMOVE(&vnic->filter, filter,
2459 				      bnxt_filter_info, next);
2460 			bnxt_free_filter(bp, filter);
2461 			PMD_DRV_LOG(INFO,
2462 				    "Deleted vlan filter for %d\n",
2463 				    vlan_id);
2464 			return 0;
2465 		}
2466 		filter = STAILQ_NEXT(filter, next);
2467 	}
2468 	return -ENOENT;
2469 }
2470 
2471 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2472 {
2473 	struct bnxt_filter_info *filter;
2474 	struct bnxt_vnic_info *vnic;
2475 	int rc = 0;
2476 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2477 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2478 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2479 
2480 	/* Implementation notes on the use of VNIC in this command:
2481 	 *
2482 	 * By default, these filters belong to default vnic for the function.
2483 	 * Once these filters are set up, only destination VNIC can be modified.
2484 	 * If the destination VNIC is not specified in this command,
2485 	 * then the HWRM shall only create an l2 context id.
2486 	 */
2487 
2488 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2489 	filter = STAILQ_FIRST(&vnic->filter);
2490 	/* Check if the VLAN has already been added */
2491 	while (filter) {
2492 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2493 			return -EEXIST;
2494 
2495 		filter = STAILQ_NEXT(filter, next);
2496 	}
2497 
2498 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2499 	 * command to create MAC+VLAN filter with the right flags, enables set.
2500 	 */
2501 	filter = bnxt_alloc_filter(bp);
2502 	if (!filter) {
2503 		PMD_DRV_LOG(ERR,
2504 			    "MAC/VLAN filter alloc failed\n");
2505 		return -ENOMEM;
2506 	}
2507 	/* MAC + VLAN ID filter */
2508 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2509 	 * untagged packets are received
2510 	 *
2511 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2512 	 * packets and only the programmed vlan's packets are received
2513 	 */
2514 	filter->l2_ivlan = vlan_id;
2515 	filter->l2_ivlan_mask = 0x0FFF;
2516 	filter->enables |= en;
2517 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2518 
2519 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2520 	if (rc) {
2521 		/* Free the newly allocated filter as we were
2522 		 * not able to create the filter in hardware.
2523 		 */
2524 		bnxt_free_filter(bp, filter);
2525 		return rc;
2526 	}
2527 
2528 	filter->mac_index = 0;
2529 	/* Add this new filter to the list */
2530 	if (vlan_id == 0)
2531 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2532 	else
2533 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2534 
2535 	PMD_DRV_LOG(INFO,
2536 		    "Added Vlan filter for %d\n", vlan_id);
2537 	return rc;
2538 }
2539 
2540 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2541 		uint16_t vlan_id, int on)
2542 {
2543 	struct bnxt *bp = eth_dev->data->dev_private;
2544 	int rc;
2545 
2546 	rc = is_bnxt_in_error(bp);
2547 	if (rc)
2548 		return rc;
2549 
2550 	if (!eth_dev->data->dev_started) {
2551 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2552 		return -EINVAL;
2553 	}
2554 
2555 	/* These operations apply to ALL existing MAC/VLAN filters */
2556 	if (on)
2557 		return bnxt_add_vlan_filter(bp, vlan_id);
2558 	else
2559 		return bnxt_del_vlan_filter(bp, vlan_id);
2560 }
2561 
2562 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2563 				    struct bnxt_vnic_info *vnic)
2564 {
2565 	struct bnxt_filter_info *filter;
2566 	int rc;
2567 
2568 	filter = STAILQ_FIRST(&vnic->filter);
2569 	while (filter) {
2570 		if (filter->mac_index == 0 &&
2571 		    !memcmp(filter->l2_addr, bp->mac_addr,
2572 			    RTE_ETHER_ADDR_LEN)) {
2573 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2574 			if (!rc) {
2575 				STAILQ_REMOVE(&vnic->filter, filter,
2576 					      bnxt_filter_info, next);
2577 				bnxt_free_filter(bp, filter);
2578 			}
2579 			return rc;
2580 		}
2581 		filter = STAILQ_NEXT(filter, next);
2582 	}
2583 	return 0;
2584 }
2585 
2586 static int
2587 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2588 {
2589 	struct bnxt_vnic_info *vnic;
2590 	unsigned int i;
2591 	int rc;
2592 
2593 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2594 	if (!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)) {
2595 		/* Remove any VLAN filters programmed */
2596 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2597 			bnxt_del_vlan_filter(bp, i);
2598 
2599 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2600 		if (rc)
2601 			return rc;
2602 	} else {
2603 		/* Default filter will allow packets that match the
2604 		 * dest mac. So, it has to be deleted, otherwise, we
2605 		 * will endup receiving vlan packets for which the
2606 		 * filter is not programmed, when hw-vlan-filter
2607 		 * configuration is ON
2608 		 */
2609 		bnxt_del_dflt_mac_filter(bp, vnic);
2610 		/* This filter will allow only untagged packets */
2611 		bnxt_add_vlan_filter(bp, 0);
2612 	}
2613 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2614 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER));
2615 
2616 	return 0;
2617 }
2618 
2619 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2620 {
2621 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2622 	unsigned int i;
2623 	int rc;
2624 
2625 	/* Destroy vnic filters and vnic */
2626 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2627 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2628 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2629 			bnxt_del_vlan_filter(bp, i);
2630 	}
2631 	bnxt_del_dflt_mac_filter(bp, vnic);
2632 
2633 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2634 	if (rc)
2635 		return rc;
2636 
2637 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2638 	if (rc)
2639 		return rc;
2640 
2641 	rte_free(vnic->fw_grp_ids);
2642 	vnic->fw_grp_ids = NULL;
2643 
2644 	vnic->rx_queue_cnt = 0;
2645 
2646 	return 0;
2647 }
2648 
2649 static int
2650 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2651 {
2652 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2653 	int rc;
2654 
2655 	/* Destroy, recreate and reconfigure the default vnic */
2656 	rc = bnxt_free_one_vnic(bp, 0);
2657 	if (rc)
2658 		return rc;
2659 
2660 	/* default vnic 0 */
2661 	rc = bnxt_setup_one_vnic(bp, 0);
2662 	if (rc)
2663 		return rc;
2664 
2665 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2666 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2667 		rc = bnxt_add_vlan_filter(bp, 0);
2668 		if (rc)
2669 			return rc;
2670 		rc = bnxt_restore_vlan_filters(bp);
2671 		if (rc)
2672 			return rc;
2673 	} else {
2674 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2675 		if (rc)
2676 			return rc;
2677 	}
2678 
2679 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2680 	if (rc)
2681 		return rc;
2682 
2683 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2684 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP));
2685 
2686 	return rc;
2687 }
2688 
2689 static int
2690 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2691 {
2692 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2693 	struct bnxt *bp = dev->data->dev_private;
2694 	int rc;
2695 
2696 	rc = is_bnxt_in_error(bp);
2697 	if (rc)
2698 		return rc;
2699 
2700 	/* Filter settings will get applied when port is started */
2701 	if (!dev->data->dev_started)
2702 		return 0;
2703 
2704 	if (mask & RTE_ETH_VLAN_FILTER_MASK) {
2705 		/* Enable or disable VLAN filtering */
2706 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2707 		if (rc)
2708 			return rc;
2709 	}
2710 
2711 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
2712 		/* Enable or disable VLAN stripping */
2713 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2714 		if (rc)
2715 			return rc;
2716 	}
2717 
2718 	if (mask & RTE_ETH_VLAN_EXTEND_MASK) {
2719 		if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND)
2720 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2721 		else
2722 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2723 	}
2724 
2725 	return 0;
2726 }
2727 
2728 static int
2729 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2730 		      uint16_t tpid)
2731 {
2732 	struct bnxt *bp = dev->data->dev_private;
2733 	int qinq = dev->data->dev_conf.rxmode.offloads &
2734 		   RTE_ETH_RX_OFFLOAD_VLAN_EXTEND;
2735 
2736 	if (vlan_type != RTE_ETH_VLAN_TYPE_INNER &&
2737 	    vlan_type != RTE_ETH_VLAN_TYPE_OUTER) {
2738 		PMD_DRV_LOG(ERR,
2739 			    "Unsupported vlan type.");
2740 		return -EINVAL;
2741 	}
2742 	if (!qinq) {
2743 		PMD_DRV_LOG(ERR,
2744 			    "QinQ not enabled. Needs to be ON as we can "
2745 			    "accelerate only outer vlan\n");
2746 		return -EINVAL;
2747 	}
2748 
2749 	if (vlan_type == RTE_ETH_VLAN_TYPE_OUTER) {
2750 		switch (tpid) {
2751 		case RTE_ETHER_TYPE_QINQ:
2752 			bp->outer_tpid_bd =
2753 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2754 				break;
2755 		case RTE_ETHER_TYPE_VLAN:
2756 			bp->outer_tpid_bd =
2757 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2758 				break;
2759 		case RTE_ETHER_TYPE_QINQ1:
2760 			bp->outer_tpid_bd =
2761 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2762 				break;
2763 		case RTE_ETHER_TYPE_QINQ2:
2764 			bp->outer_tpid_bd =
2765 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2766 				break;
2767 		case RTE_ETHER_TYPE_QINQ3:
2768 			bp->outer_tpid_bd =
2769 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2770 				break;
2771 		default:
2772 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2773 			return -EINVAL;
2774 		}
2775 		bp->outer_tpid_bd |= tpid;
2776 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2777 	} else if (vlan_type == RTE_ETH_VLAN_TYPE_INNER) {
2778 		PMD_DRV_LOG(ERR,
2779 			    "Can accelerate only outer vlan in QinQ\n");
2780 		return -EINVAL;
2781 	}
2782 
2783 	return 0;
2784 }
2785 
2786 static int
2787 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2788 			     struct rte_ether_addr *addr)
2789 {
2790 	struct bnxt *bp = dev->data->dev_private;
2791 	/* Default Filter is tied to VNIC 0 */
2792 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2793 	int rc;
2794 
2795 	rc = is_bnxt_in_error(bp);
2796 	if (rc)
2797 		return rc;
2798 
2799 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2800 		return -EPERM;
2801 
2802 	if (rte_is_zero_ether_addr(addr))
2803 		return -EINVAL;
2804 
2805 	/* Filter settings will get applied when port is started */
2806 	if (!dev->data->dev_started)
2807 		return 0;
2808 
2809 	/* Check if the requested MAC is already added */
2810 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2811 		return 0;
2812 
2813 	/* Destroy filter and re-create it */
2814 	bnxt_del_dflt_mac_filter(bp, vnic);
2815 
2816 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2817 	if (dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2818 		/* This filter will allow only untagged packets */
2819 		rc = bnxt_add_vlan_filter(bp, 0);
2820 	} else {
2821 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2822 	}
2823 
2824 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2825 	return rc;
2826 }
2827 
2828 static int
2829 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2830 			  struct rte_ether_addr *mc_addr_set,
2831 			  uint32_t nb_mc_addr)
2832 {
2833 	struct bnxt *bp = eth_dev->data->dev_private;
2834 	char *mc_addr_list = (char *)mc_addr_set;
2835 	struct bnxt_vnic_info *vnic;
2836 	uint32_t off = 0, i = 0;
2837 	int rc;
2838 
2839 	rc = is_bnxt_in_error(bp);
2840 	if (rc)
2841 		return rc;
2842 
2843 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2844 
2845 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2846 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2847 		goto allmulti;
2848 	}
2849 
2850 	/* TODO Check for Duplicate mcast addresses */
2851 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2852 	for (i = 0; i < nb_mc_addr; i++) {
2853 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2854 			RTE_ETHER_ADDR_LEN);
2855 		off += RTE_ETHER_ADDR_LEN;
2856 	}
2857 
2858 	vnic->mc_addr_cnt = i;
2859 	if (vnic->mc_addr_cnt)
2860 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2861 	else
2862 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2863 
2864 allmulti:
2865 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2866 }
2867 
2868 static int
2869 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2870 {
2871 	struct bnxt *bp = dev->data->dev_private;
2872 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2873 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2874 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2875 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2876 	int ret;
2877 
2878 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2879 			fw_major, fw_minor, fw_updt, fw_rsvd);
2880 	if (ret < 0)
2881 		return -EINVAL;
2882 
2883 	ret += 1; /* add the size of '\0' */
2884 	if (fw_size < (size_t)ret)
2885 		return ret;
2886 	else
2887 		return 0;
2888 }
2889 
2890 static void
2891 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2892 	struct rte_eth_rxq_info *qinfo)
2893 {
2894 	struct bnxt *bp = dev->data->dev_private;
2895 	struct bnxt_rx_queue *rxq;
2896 
2897 	if (is_bnxt_in_error(bp))
2898 		return;
2899 
2900 	rxq = dev->data->rx_queues[queue_id];
2901 
2902 	qinfo->mp = rxq->mb_pool;
2903 	qinfo->scattered_rx = dev->data->scattered_rx;
2904 	qinfo->nb_desc = rxq->nb_rx_desc;
2905 
2906 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2907 	qinfo->conf.rx_drop_en = rxq->drop_en;
2908 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2909 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2910 }
2911 
2912 static void
2913 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2914 	struct rte_eth_txq_info *qinfo)
2915 {
2916 	struct bnxt *bp = dev->data->dev_private;
2917 	struct bnxt_tx_queue *txq;
2918 
2919 	if (is_bnxt_in_error(bp))
2920 		return;
2921 
2922 	txq = dev->data->tx_queues[queue_id];
2923 
2924 	qinfo->nb_desc = txq->nb_tx_desc;
2925 
2926 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2927 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2928 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2929 
2930 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2931 	qinfo->conf.tx_rs_thresh = 0;
2932 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2933 	qinfo->conf.offloads = txq->offloads;
2934 }
2935 
2936 static const struct {
2937 	eth_rx_burst_t pkt_burst;
2938 	const char *info;
2939 } bnxt_rx_burst_info[] = {
2940 	{bnxt_recv_pkts,		"Scalar"},
2941 #if defined(RTE_ARCH_X86)
2942 	{bnxt_recv_pkts_vec,		"Vector SSE"},
2943 #endif
2944 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2945 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
2946 #endif
2947 #if defined(RTE_ARCH_ARM64)
2948 	{bnxt_recv_pkts_vec,		"Vector Neon"},
2949 #endif
2950 };
2951 
2952 static int
2953 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2954 		       struct rte_eth_burst_mode *mode)
2955 {
2956 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2957 	size_t i;
2958 
2959 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2960 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2961 			snprintf(mode->info, sizeof(mode->info), "%s",
2962 				 bnxt_rx_burst_info[i].info);
2963 			return 0;
2964 		}
2965 	}
2966 
2967 	return -EINVAL;
2968 }
2969 
2970 static const struct {
2971 	eth_tx_burst_t pkt_burst;
2972 	const char *info;
2973 } bnxt_tx_burst_info[] = {
2974 	{bnxt_xmit_pkts,		"Scalar"},
2975 #if defined(RTE_ARCH_X86)
2976 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
2977 #endif
2978 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2979 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
2980 #endif
2981 #if defined(RTE_ARCH_ARM64)
2982 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
2983 #endif
2984 };
2985 
2986 static int
2987 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2988 		       struct rte_eth_burst_mode *mode)
2989 {
2990 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2991 	size_t i;
2992 
2993 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2994 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2995 			snprintf(mode->info, sizeof(mode->info), "%s",
2996 				 bnxt_tx_burst_info[i].info);
2997 			return 0;
2998 		}
2999 	}
3000 
3001 	return -EINVAL;
3002 }
3003 
3004 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3005 {
3006 	uint32_t overhead = BNXT_MAX_PKT_LEN - BNXT_MAX_MTU;
3007 	struct bnxt *bp = eth_dev->data->dev_private;
3008 	uint32_t new_pkt_size;
3009 	uint32_t rc;
3010 	uint32_t i;
3011 
3012 	rc = is_bnxt_in_error(bp);
3013 	if (rc)
3014 		return rc;
3015 
3016 	/* Exit if receive queues are not configured yet */
3017 	if (!eth_dev->data->nb_rx_queues)
3018 		return rc;
3019 
3020 	new_pkt_size = new_mtu + overhead;
3021 
3022 	/*
3023 	 * Disallow any MTU change that would require scattered receive support
3024 	 * if it is not already enabled.
3025 	 */
3026 	if (eth_dev->data->dev_started &&
3027 	    !eth_dev->data->scattered_rx &&
3028 	    (new_pkt_size >
3029 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
3030 		PMD_DRV_LOG(ERR,
3031 			    "MTU change would require scattered rx support. ");
3032 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
3033 		return -EINVAL;
3034 	}
3035 
3036 	if (new_mtu > RTE_ETHER_MTU)
3037 		bp->flags |= BNXT_FLAG_JUMBO;
3038 	else
3039 		bp->flags &= ~BNXT_FLAG_JUMBO;
3040 
3041 	/* Is there a change in mtu setting? */
3042 	if (eth_dev->data->mtu == new_mtu)
3043 		return rc;
3044 
3045 	for (i = 0; i < bp->nr_vnics; i++) {
3046 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3047 		uint16_t size = 0;
3048 
3049 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
3050 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
3051 		if (rc)
3052 			break;
3053 
3054 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
3055 		size -= RTE_PKTMBUF_HEADROOM;
3056 
3057 		if (size < new_mtu) {
3058 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
3059 			if (rc)
3060 				return rc;
3061 		}
3062 	}
3063 
3064 	if (bnxt_hwrm_config_host_mtu(bp))
3065 		PMD_DRV_LOG(WARNING, "Failed to configure host MTU\n");
3066 
3067 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3068 
3069 	return rc;
3070 }
3071 
3072 static int
3073 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3074 {
3075 	struct bnxt *bp = dev->data->dev_private;
3076 	uint16_t vlan = bp->vlan;
3077 	int rc;
3078 
3079 	rc = is_bnxt_in_error(bp);
3080 	if (rc)
3081 		return rc;
3082 
3083 	if (!BNXT_SINGLE_PF(bp)) {
3084 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3085 		return -ENOTSUP;
3086 	}
3087 	bp->vlan = on ? pvid : 0;
3088 
3089 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3090 	if (rc)
3091 		bp->vlan = vlan;
3092 	return rc;
3093 }
3094 
3095 static int
3096 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3097 {
3098 	struct bnxt *bp = dev->data->dev_private;
3099 	int rc;
3100 
3101 	rc = is_bnxt_in_error(bp);
3102 	if (rc)
3103 		return rc;
3104 
3105 	return bnxt_hwrm_port_led_cfg(bp, true);
3106 }
3107 
3108 static int
3109 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3110 {
3111 	struct bnxt *bp = dev->data->dev_private;
3112 	int rc;
3113 
3114 	rc = is_bnxt_in_error(bp);
3115 	if (rc)
3116 		return rc;
3117 
3118 	return bnxt_hwrm_port_led_cfg(bp, false);
3119 }
3120 
3121 static uint32_t
3122 bnxt_rx_queue_count_op(void *rx_queue)
3123 {
3124 	struct bnxt *bp;
3125 	struct bnxt_cp_ring_info *cpr;
3126 	uint32_t desc = 0, raw_cons, cp_ring_size;
3127 	struct bnxt_rx_queue *rxq;
3128 	struct rx_pkt_cmpl *rxcmp;
3129 	int rc;
3130 
3131 	rxq = rx_queue;
3132 	bp = rxq->bp;
3133 
3134 	rc = is_bnxt_in_error(bp);
3135 	if (rc)
3136 		return rc;
3137 
3138 	cpr = rxq->cp_ring;
3139 	raw_cons = cpr->cp_raw_cons;
3140 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3141 
3142 	while (1) {
3143 		uint32_t agg_cnt, cons, cmpl_type;
3144 
3145 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3146 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3147 
3148 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3149 			break;
3150 
3151 		cmpl_type = CMP_TYPE(rxcmp);
3152 
3153 		switch (cmpl_type) {
3154 		case CMPL_BASE_TYPE_RX_L2:
3155 		case CMPL_BASE_TYPE_RX_L2_V2:
3156 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3157 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3158 			desc++;
3159 			break;
3160 
3161 		case CMPL_BASE_TYPE_RX_TPA_END:
3162 			if (BNXT_CHIP_P5(rxq->bp)) {
3163 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3164 
3165 				p5_tpa_end = (void *)rxcmp;
3166 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3167 			} else {
3168 				struct rx_tpa_end_cmpl *tpa_end;
3169 
3170 				tpa_end = (void *)rxcmp;
3171 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3172 			}
3173 
3174 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3175 			desc++;
3176 			break;
3177 
3178 		default:
3179 			raw_cons += CMP_LEN(cmpl_type);
3180 		}
3181 	}
3182 
3183 	return desc;
3184 }
3185 
3186 static int
3187 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3188 {
3189 	struct bnxt_rx_queue *rxq = rx_queue;
3190 	struct bnxt_cp_ring_info *cpr;
3191 	struct bnxt_rx_ring_info *rxr;
3192 	uint32_t desc, raw_cons, cp_ring_size;
3193 	struct bnxt *bp = rxq->bp;
3194 	struct rx_pkt_cmpl *rxcmp;
3195 	int rc;
3196 
3197 	rc = is_bnxt_in_error(bp);
3198 	if (rc)
3199 		return rc;
3200 
3201 	if (offset >= rxq->nb_rx_desc)
3202 		return -EINVAL;
3203 
3204 	rxr = rxq->rx_ring;
3205 	cpr = rxq->cp_ring;
3206 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3207 
3208 	/*
3209 	 * For the vector receive case, the completion at the requested
3210 	 * offset can be indexed directly.
3211 	 */
3212 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3213 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3214 		struct rx_pkt_cmpl *rxcmp;
3215 		uint32_t cons;
3216 
3217 		/* Check status of completion descriptor. */
3218 		raw_cons = cpr->cp_raw_cons +
3219 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3220 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3221 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3222 
3223 		if (bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3224 			return RTE_ETH_RX_DESC_DONE;
3225 
3226 		/* Check whether rx desc has an mbuf attached. */
3227 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3228 		if (cons >= rxq->rxrearm_start &&
3229 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3230 			return RTE_ETH_RX_DESC_UNAVAIL;
3231 		}
3232 
3233 		return RTE_ETH_RX_DESC_AVAIL;
3234 	}
3235 #endif
3236 
3237 	/*
3238 	 * For the non-vector receive case, scan the completion ring to
3239 	 * locate the completion descriptor for the requested offset.
3240 	 */
3241 	raw_cons = cpr->cp_raw_cons;
3242 	desc = 0;
3243 	while (1) {
3244 		uint32_t agg_cnt, cons, cmpl_type;
3245 
3246 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3247 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3248 
3249 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3250 			break;
3251 
3252 		cmpl_type = CMP_TYPE(rxcmp);
3253 
3254 		switch (cmpl_type) {
3255 		case CMPL_BASE_TYPE_RX_L2:
3256 		case CMPL_BASE_TYPE_RX_L2_V2:
3257 			if (desc == offset) {
3258 				cons = rxcmp->opaque;
3259 				if (rxr->rx_buf_ring[cons])
3260 					return RTE_ETH_RX_DESC_DONE;
3261 				else
3262 					return RTE_ETH_RX_DESC_UNAVAIL;
3263 			}
3264 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3265 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3266 			desc++;
3267 			break;
3268 
3269 		case CMPL_BASE_TYPE_RX_TPA_END:
3270 			if (desc == offset)
3271 				return RTE_ETH_RX_DESC_DONE;
3272 
3273 			if (BNXT_CHIP_P5(rxq->bp)) {
3274 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3275 
3276 				p5_tpa_end = (void *)rxcmp;
3277 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3278 			} else {
3279 				struct rx_tpa_end_cmpl *tpa_end;
3280 
3281 				tpa_end = (void *)rxcmp;
3282 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3283 			}
3284 
3285 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3286 			desc++;
3287 			break;
3288 
3289 		default:
3290 			raw_cons += CMP_LEN(cmpl_type);
3291 		}
3292 	}
3293 
3294 	return RTE_ETH_RX_DESC_AVAIL;
3295 }
3296 
3297 static int
3298 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3299 {
3300 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3301 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3302 	uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3303 	struct cmpl_base *cp_desc_ring;
3304 	int rc;
3305 
3306 	rc = is_bnxt_in_error(txq->bp);
3307 	if (rc)
3308 		return rc;
3309 
3310 	if (offset >= txq->nb_tx_desc)
3311 		return -EINVAL;
3312 
3313 	/* Return "desc done" if descriptor is available for use. */
3314 	if (bnxt_tx_bds_in_hw(txq) <= offset)
3315 		return RTE_ETH_TX_DESC_DONE;
3316 
3317 	raw_cons = cpr->cp_raw_cons;
3318 	cp_desc_ring = cpr->cp_desc_ring;
3319 	ring_mask = cpr->cp_ring_struct->ring_mask;
3320 
3321 	/* Check to see if hw has posted a completion for the descriptor. */
3322 	while (1) {
3323 		struct tx_cmpl *txcmp;
3324 		uint32_t cons;
3325 
3326 		cons = RING_CMPL(ring_mask, raw_cons);
3327 		txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3328 
3329 		if (!bnxt_cpr_cmp_valid(txcmp, raw_cons, ring_mask + 1))
3330 			break;
3331 
3332 		if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3333 			nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3334 
3335 		if (nb_tx_pkts > offset)
3336 			return RTE_ETH_TX_DESC_DONE;
3337 
3338 		raw_cons = NEXT_RAW_CMP(raw_cons);
3339 	}
3340 
3341 	/* Descriptor is pending transmit, not yet completed by hardware. */
3342 	return RTE_ETH_TX_DESC_FULL;
3343 }
3344 
3345 int
3346 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3347 		     const struct rte_flow_ops **ops)
3348 {
3349 	struct bnxt *bp = dev->data->dev_private;
3350 	int ret = 0;
3351 
3352 	if (!bp)
3353 		return -EIO;
3354 
3355 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3356 		struct bnxt_representor *vfr = dev->data->dev_private;
3357 		bp = vfr->parent_dev->data->dev_private;
3358 		/* parent is deleted while children are still valid */
3359 		if (!bp) {
3360 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3361 				    dev->data->port_id);
3362 			return -EIO;
3363 		}
3364 	}
3365 
3366 	ret = is_bnxt_in_error(bp);
3367 	if (ret)
3368 		return ret;
3369 
3370 	/* PMD supports thread-safe flow operations.  rte_flow API
3371 	 * functions can avoid mutex for multi-thread safety.
3372 	 */
3373 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3374 
3375 	if (BNXT_TRUFLOW_EN(bp))
3376 		*ops = &bnxt_ulp_rte_flow_ops;
3377 	else
3378 		*ops = &bnxt_flow_ops;
3379 
3380 	return ret;
3381 }
3382 
3383 static const uint32_t *
3384 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3385 {
3386 	static const uint32_t ptypes[] = {
3387 		RTE_PTYPE_L2_ETHER_VLAN,
3388 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3389 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3390 		RTE_PTYPE_L4_ICMP,
3391 		RTE_PTYPE_L4_TCP,
3392 		RTE_PTYPE_L4_UDP,
3393 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3394 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3395 		RTE_PTYPE_INNER_L4_ICMP,
3396 		RTE_PTYPE_INNER_L4_TCP,
3397 		RTE_PTYPE_INNER_L4_UDP,
3398 		RTE_PTYPE_UNKNOWN
3399 	};
3400 
3401 	if (!dev->rx_pkt_burst)
3402 		return NULL;
3403 
3404 	return ptypes;
3405 }
3406 
3407 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3408 			 int reg_win)
3409 {
3410 	uint32_t reg_base = *reg_arr & 0xfffff000;
3411 	uint32_t win_off;
3412 	int i;
3413 
3414 	for (i = 0; i < count; i++) {
3415 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3416 			return -ERANGE;
3417 	}
3418 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3419 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3420 	return 0;
3421 }
3422 
3423 static int bnxt_map_ptp_regs(struct bnxt *bp)
3424 {
3425 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3426 	uint32_t *reg_arr;
3427 	int rc, i;
3428 
3429 	reg_arr = ptp->rx_regs;
3430 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3431 	if (rc)
3432 		return rc;
3433 
3434 	reg_arr = ptp->tx_regs;
3435 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3436 	if (rc)
3437 		return rc;
3438 
3439 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3440 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3441 
3442 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3443 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3444 
3445 	return 0;
3446 }
3447 
3448 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3449 {
3450 	rte_write32(0, (uint8_t *)bp->bar0 +
3451 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3452 	rte_write32(0, (uint8_t *)bp->bar0 +
3453 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3454 }
3455 
3456 static uint64_t bnxt_cc_read(struct bnxt *bp)
3457 {
3458 	uint64_t ns;
3459 
3460 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3461 			      BNXT_GRCPF_REG_SYNC_TIME));
3462 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3463 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3464 	return ns;
3465 }
3466 
3467 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3468 {
3469 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3470 	uint32_t fifo;
3471 
3472 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3473 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3474 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3475 		return -EAGAIN;
3476 
3477 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3478 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3479 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3480 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3481 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3482 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3483 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3484 
3485 	return 0;
3486 }
3487 
3488 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3489 {
3490 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3491 	struct bnxt_pf_info *pf = bp->pf;
3492 	uint16_t port_id;
3493 	int i = 0;
3494 	uint32_t fifo;
3495 
3496 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3497 		return -EINVAL;
3498 
3499 	port_id = pf->port_id;
3500 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3501 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3502 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3503 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3504 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3505 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3506 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3507 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3508 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3509 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3510 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3511 		i++;
3512 	}
3513 
3514 	if (i >= BNXT_PTP_RX_PND_CNT)
3515 		return -EBUSY;
3516 
3517 	return 0;
3518 }
3519 
3520 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3521 {
3522 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3523 	struct bnxt_pf_info *pf = bp->pf;
3524 	uint16_t port_id;
3525 	uint32_t fifo;
3526 
3527 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3528 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3529 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3530 		return -EAGAIN;
3531 
3532 	port_id = pf->port_id;
3533 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3534 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3535 
3536 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3537 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3538 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3539 		return bnxt_clr_rx_ts(bp, ts);
3540 
3541 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3542 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3543 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3544 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3545 
3546 	return 0;
3547 }
3548 
3549 static int
3550 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3551 {
3552 	uint64_t ns;
3553 	struct bnxt *bp = dev->data->dev_private;
3554 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3555 
3556 	if (!ptp)
3557 		return -ENOTSUP;
3558 
3559 	ns = rte_timespec_to_ns(ts);
3560 	/* Set the timecounters to a new value. */
3561 	ptp->tc.nsec = ns;
3562 	ptp->tx_tstamp_tc.nsec = ns;
3563 	ptp->rx_tstamp_tc.nsec = ns;
3564 
3565 	return 0;
3566 }
3567 
3568 static int
3569 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3570 {
3571 	struct bnxt *bp = dev->data->dev_private;
3572 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3573 	uint64_t ns, systime_cycles = 0;
3574 	int rc = 0;
3575 
3576 	if (!ptp)
3577 		return -ENOTSUP;
3578 
3579 	if (BNXT_CHIP_P5(bp))
3580 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3581 					     &systime_cycles);
3582 	else
3583 		systime_cycles = bnxt_cc_read(bp);
3584 
3585 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3586 	*ts = rte_ns_to_timespec(ns);
3587 
3588 	return rc;
3589 }
3590 static int
3591 bnxt_timesync_enable(struct rte_eth_dev *dev)
3592 {
3593 	struct bnxt *bp = dev->data->dev_private;
3594 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3595 	uint32_t shift = 0;
3596 	int rc;
3597 
3598 	if (!ptp)
3599 		return -ENOTSUP;
3600 
3601 	ptp->rx_filter = 1;
3602 	ptp->tx_tstamp_en = 1;
3603 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3604 
3605 	rc = bnxt_hwrm_ptp_cfg(bp);
3606 	if (rc)
3607 		return rc;
3608 
3609 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3610 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3611 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3612 
3613 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3614 	ptp->tc.cc_shift = shift;
3615 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3616 
3617 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3618 	ptp->rx_tstamp_tc.cc_shift = shift;
3619 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3620 
3621 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3622 	ptp->tx_tstamp_tc.cc_shift = shift;
3623 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3624 
3625 	if (!BNXT_CHIP_P5(bp))
3626 		bnxt_map_ptp_regs(bp);
3627 	else
3628 		rc = bnxt_ptp_start(bp);
3629 
3630 	return rc;
3631 }
3632 
3633 static int
3634 bnxt_timesync_disable(struct rte_eth_dev *dev)
3635 {
3636 	struct bnxt *bp = dev->data->dev_private;
3637 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3638 
3639 	if (!ptp)
3640 		return -ENOTSUP;
3641 
3642 	ptp->rx_filter = 0;
3643 	ptp->tx_tstamp_en = 0;
3644 	ptp->rxctl = 0;
3645 
3646 	bnxt_hwrm_ptp_cfg(bp);
3647 
3648 	if (!BNXT_CHIP_P5(bp))
3649 		bnxt_unmap_ptp_regs(bp);
3650 	else
3651 		bnxt_ptp_stop(bp);
3652 
3653 	return 0;
3654 }
3655 
3656 static int
3657 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3658 				 struct timespec *timestamp,
3659 				 uint32_t flags __rte_unused)
3660 {
3661 	struct bnxt *bp = dev->data->dev_private;
3662 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3663 	uint64_t rx_tstamp_cycles = 0;
3664 	uint64_t ns;
3665 
3666 	if (!ptp)
3667 		return -ENOTSUP;
3668 
3669 	if (BNXT_CHIP_P5(bp))
3670 		rx_tstamp_cycles = ptp->rx_timestamp;
3671 	else
3672 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3673 
3674 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3675 	*timestamp = rte_ns_to_timespec(ns);
3676 	return  0;
3677 }
3678 
3679 static int
3680 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3681 				 struct timespec *timestamp)
3682 {
3683 	struct bnxt *bp = dev->data->dev_private;
3684 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3685 	uint64_t tx_tstamp_cycles = 0;
3686 	uint64_t ns;
3687 	int rc = 0;
3688 
3689 	if (!ptp)
3690 		return -ENOTSUP;
3691 
3692 	if (BNXT_CHIP_P5(bp))
3693 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3694 					     &tx_tstamp_cycles);
3695 	else
3696 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3697 
3698 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3699 	*timestamp = rte_ns_to_timespec(ns);
3700 
3701 	return rc;
3702 }
3703 
3704 static int
3705 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3706 {
3707 	struct bnxt *bp = dev->data->dev_private;
3708 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3709 
3710 	if (!ptp)
3711 		return -ENOTSUP;
3712 
3713 	ptp->tc.nsec += delta;
3714 	ptp->tx_tstamp_tc.nsec += delta;
3715 	ptp->rx_tstamp_tc.nsec += delta;
3716 
3717 	return 0;
3718 }
3719 
3720 static int
3721 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3722 {
3723 	struct bnxt *bp = dev->data->dev_private;
3724 	int rc;
3725 	uint32_t dir_entries;
3726 	uint32_t entry_length;
3727 
3728 	rc = is_bnxt_in_error(bp);
3729 	if (rc)
3730 		return rc;
3731 
3732 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3733 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3734 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3735 
3736 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3737 	if (rc != 0)
3738 		return rc;
3739 
3740 	return dir_entries * entry_length;
3741 }
3742 
3743 static int
3744 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3745 		struct rte_dev_eeprom_info *in_eeprom)
3746 {
3747 	struct bnxt *bp = dev->data->dev_private;
3748 	uint32_t index;
3749 	uint32_t offset;
3750 	int rc;
3751 
3752 	rc = is_bnxt_in_error(bp);
3753 	if (rc)
3754 		return rc;
3755 
3756 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3757 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3758 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3759 		    in_eeprom->offset, in_eeprom->length);
3760 
3761 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3762 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3763 						in_eeprom->data);
3764 
3765 	index = in_eeprom->offset >> 24;
3766 	offset = in_eeprom->offset & 0xffffff;
3767 
3768 	if (index != 0)
3769 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3770 					   in_eeprom->length, in_eeprom->data);
3771 
3772 	return 0;
3773 }
3774 
3775 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3776 {
3777 	switch (dir_type) {
3778 	case BNX_DIR_TYPE_CHIMP_PATCH:
3779 	case BNX_DIR_TYPE_BOOTCODE:
3780 	case BNX_DIR_TYPE_BOOTCODE_2:
3781 	case BNX_DIR_TYPE_APE_FW:
3782 	case BNX_DIR_TYPE_APE_PATCH:
3783 	case BNX_DIR_TYPE_KONG_FW:
3784 	case BNX_DIR_TYPE_KONG_PATCH:
3785 	case BNX_DIR_TYPE_BONO_FW:
3786 	case BNX_DIR_TYPE_BONO_PATCH:
3787 		/* FALLTHROUGH */
3788 		return true;
3789 	}
3790 
3791 	return false;
3792 }
3793 
3794 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3795 {
3796 	switch (dir_type) {
3797 	case BNX_DIR_TYPE_AVS:
3798 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3799 	case BNX_DIR_TYPE_PCIE:
3800 	case BNX_DIR_TYPE_TSCF_UCODE:
3801 	case BNX_DIR_TYPE_EXT_PHY:
3802 	case BNX_DIR_TYPE_CCM:
3803 	case BNX_DIR_TYPE_ISCSI_BOOT:
3804 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3805 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3806 		/* FALLTHROUGH */
3807 		return true;
3808 	}
3809 
3810 	return false;
3811 }
3812 
3813 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3814 {
3815 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3816 		bnxt_dir_type_is_other_exec_format(dir_type);
3817 }
3818 
3819 static int
3820 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3821 		struct rte_dev_eeprom_info *in_eeprom)
3822 {
3823 	struct bnxt *bp = dev->data->dev_private;
3824 	uint8_t index, dir_op;
3825 	uint16_t type, ext, ordinal, attr;
3826 	int rc;
3827 
3828 	rc = is_bnxt_in_error(bp);
3829 	if (rc)
3830 		return rc;
3831 
3832 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3833 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3834 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3835 		    in_eeprom->offset, in_eeprom->length);
3836 
3837 	if (!BNXT_PF(bp)) {
3838 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3839 		return -EINVAL;
3840 	}
3841 
3842 	type = in_eeprom->magic >> 16;
3843 
3844 	if (type == 0xffff) { /* special value for directory operations */
3845 		index = in_eeprom->magic & 0xff;
3846 		dir_op = in_eeprom->magic >> 8;
3847 		if (index == 0)
3848 			return -EINVAL;
3849 		switch (dir_op) {
3850 		case 0x0e: /* erase */
3851 			if (in_eeprom->offset != ~in_eeprom->magic)
3852 				return -EINVAL;
3853 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3854 		default:
3855 			return -EINVAL;
3856 		}
3857 	}
3858 
3859 	/* Create or re-write an NVM item: */
3860 	if (bnxt_dir_type_is_executable(type) == true)
3861 		return -EOPNOTSUPP;
3862 	ext = in_eeprom->magic & 0xffff;
3863 	ordinal = in_eeprom->offset >> 16;
3864 	attr = in_eeprom->offset & 0xffff;
3865 
3866 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3867 				     in_eeprom->data, in_eeprom->length);
3868 }
3869 
3870 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3871 				struct rte_eth_dev_module_info *modinfo)
3872 {
3873 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3874 	struct bnxt *bp = dev->data->dev_private;
3875 	int rc;
3876 
3877 	/* No point in going further if phy status indicates
3878 	 * module is not inserted or if it is powered down or
3879 	 * if it is of type 10GBase-T
3880 	 */
3881 	if (bp->link_info->module_status >
3882 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3883 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3884 			    dev->data->port_id);
3885 		return -ENOTSUP;
3886 	}
3887 
3888 	/* This feature is not supported in older firmware versions */
3889 	if (bp->hwrm_spec_code < 0x10202) {
3890 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3891 			    dev->data->port_id);
3892 		return -ENOTSUP;
3893 	}
3894 
3895 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3896 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3897 						   module_info);
3898 
3899 	if (rc)
3900 		return rc;
3901 
3902 	switch (module_info[0]) {
3903 	case SFF_MODULE_ID_SFP:
3904 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3905 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3906 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3907 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3908 		break;
3909 	case SFF_MODULE_ID_QSFP:
3910 	case SFF_MODULE_ID_QSFP_PLUS:
3911 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3912 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3913 		break;
3914 	case SFF_MODULE_ID_QSFP28:
3915 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3916 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3917 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3918 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3919 		break;
3920 	default:
3921 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3922 		return -ENOTSUP;
3923 	}
3924 
3925 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3926 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3927 
3928 	return 0;
3929 }
3930 
3931 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3932 				  struct rte_dev_eeprom_info *info)
3933 {
3934 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3935 	uint32_t offset = info->offset, length = info->length;
3936 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3937 	struct bnxt *bp = dev->data->dev_private;
3938 	uint8_t *data = info->data;
3939 	uint8_t page = offset >> 7;
3940 	uint8_t max_pages = 2;
3941 	uint8_t opt_pages;
3942 	int rc;
3943 
3944 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3945 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3946 						   module_info);
3947 	if (rc)
3948 		return rc;
3949 
3950 	switch (module_info[0]) {
3951 	case SFF_MODULE_ID_SFP:
3952 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
3953 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
3954 			pg_addr[2] = I2C_DEV_ADDR_A2;
3955 			pg_addr[3] = I2C_DEV_ADDR_A2;
3956 			max_pages = 4;
3957 		}
3958 		break;
3959 	case SFF_MODULE_ID_QSFP28:
3960 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
3961 							   SFF8636_OPT_PAGES_OFFSET,
3962 							   1, &opt_pages);
3963 		if (rc)
3964 			return rc;
3965 
3966 		if (opt_pages & SFF8636_PAGE1_MASK) {
3967 			pg_addr[2] = I2C_DEV_ADDR_A0;
3968 			max_pages = 3;
3969 		}
3970 		if (opt_pages & SFF8636_PAGE2_MASK) {
3971 			pg_addr[3] = I2C_DEV_ADDR_A0;
3972 			max_pages = 4;
3973 		}
3974 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
3975 			pg_addr[4] = I2C_DEV_ADDR_A0;
3976 			max_pages = 5;
3977 		}
3978 		break;
3979 	default:
3980 		break;
3981 	}
3982 
3983 	memset(data, 0, length);
3984 
3985 	offset &= 0xff;
3986 	while (length && page < max_pages) {
3987 		uint8_t raw_page = page ? page - 1 : 0;
3988 		uint16_t chunk;
3989 
3990 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
3991 			raw_page = 0;
3992 		else if (page)
3993 			offset |= 0x80;
3994 		chunk = RTE_MIN(length, 256 - offset);
3995 
3996 		if (pg_addr[page]) {
3997 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
3998 								   raw_page, offset,
3999 								   chunk, data);
4000 			if (rc)
4001 				return rc;
4002 		}
4003 
4004 		data += chunk;
4005 		length -= chunk;
4006 		offset = 0;
4007 		page += 1 + (chunk > 128);
4008 	}
4009 
4010 	return length ? -EINVAL : 0;
4011 }
4012 
4013 /*
4014  * Initialization
4015  */
4016 
4017 static const struct eth_dev_ops bnxt_dev_ops = {
4018 	.dev_infos_get = bnxt_dev_info_get_op,
4019 	.dev_close = bnxt_dev_close_op,
4020 	.dev_configure = bnxt_dev_configure_op,
4021 	.dev_start = bnxt_dev_start_op,
4022 	.dev_stop = bnxt_dev_stop_op,
4023 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4024 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4025 	.stats_get = bnxt_stats_get_op,
4026 	.stats_reset = bnxt_stats_reset_op,
4027 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4028 	.rx_queue_release = bnxt_rx_queue_release_op,
4029 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4030 	.tx_queue_release = bnxt_tx_queue_release_op,
4031 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4032 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4033 	.reta_update = bnxt_reta_update_op,
4034 	.reta_query = bnxt_reta_query_op,
4035 	.rss_hash_update = bnxt_rss_hash_update_op,
4036 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4037 	.link_update = bnxt_link_update_op,
4038 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4039 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4040 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4041 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4042 	.mac_addr_add = bnxt_mac_addr_add_op,
4043 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4044 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4045 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4046 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4047 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4048 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4049 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4050 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4051 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4052 	.mtu_set = bnxt_mtu_set_op,
4053 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4054 	.xstats_get = bnxt_dev_xstats_get_op,
4055 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4056 	.xstats_reset = bnxt_dev_xstats_reset_op,
4057 	.fw_version_get = bnxt_fw_version_get,
4058 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4059 	.rxq_info_get = bnxt_rxq_info_get_op,
4060 	.txq_info_get = bnxt_txq_info_get_op,
4061 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4062 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4063 	.dev_led_on = bnxt_dev_led_on_op,
4064 	.dev_led_off = bnxt_dev_led_off_op,
4065 	.rx_queue_start = bnxt_rx_queue_start,
4066 	.rx_queue_stop = bnxt_rx_queue_stop,
4067 	.tx_queue_start = bnxt_tx_queue_start,
4068 	.tx_queue_stop = bnxt_tx_queue_stop,
4069 	.flow_ops_get = bnxt_flow_ops_get_op,
4070 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4071 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4072 	.get_eeprom           = bnxt_get_eeprom_op,
4073 	.set_eeprom           = bnxt_set_eeprom_op,
4074 	.get_module_info = bnxt_get_module_info,
4075 	.get_module_eeprom = bnxt_get_module_eeprom,
4076 	.timesync_enable      = bnxt_timesync_enable,
4077 	.timesync_disable     = bnxt_timesync_disable,
4078 	.timesync_read_time   = bnxt_timesync_read_time,
4079 	.timesync_write_time   = bnxt_timesync_write_time,
4080 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4081 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4082 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4083 };
4084 
4085 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4086 {
4087 	uint32_t offset;
4088 
4089 	/* Only pre-map the reset GRC registers using window 3 */
4090 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4091 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4092 
4093 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4094 
4095 	return offset;
4096 }
4097 
4098 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4099 {
4100 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4101 	uint32_t reg_base = 0xffffffff;
4102 	int i;
4103 
4104 	/* Only pre-map the monitoring GRC registers using window 2 */
4105 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4106 		uint32_t reg = info->status_regs[i];
4107 
4108 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4109 			continue;
4110 
4111 		if (reg_base == 0xffffffff)
4112 			reg_base = reg & 0xfffff000;
4113 		if ((reg & 0xfffff000) != reg_base)
4114 			return -ERANGE;
4115 
4116 		/* Use mask 0xffc as the Lower 2 bits indicates
4117 		 * address space location
4118 		 */
4119 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4120 						(reg & 0xffc);
4121 	}
4122 
4123 	if (reg_base == 0xffffffff)
4124 		return 0;
4125 
4126 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4127 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4128 
4129 	return 0;
4130 }
4131 
4132 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4133 {
4134 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4135 	uint32_t delay = info->delay_after_reset[index];
4136 	uint32_t val = info->reset_reg_val[index];
4137 	uint32_t reg = info->reset_reg[index];
4138 	uint32_t type, offset;
4139 	int ret;
4140 
4141 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4142 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4143 
4144 	switch (type) {
4145 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4146 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4147 		if (ret < 0) {
4148 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4149 				    val, offset);
4150 			return;
4151 		}
4152 		break;
4153 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4154 		offset = bnxt_map_reset_regs(bp, offset);
4155 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4156 		break;
4157 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4158 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4159 		break;
4160 	}
4161 	/* wait on a specific interval of time until core reset is complete */
4162 	if (delay)
4163 		rte_delay_ms(delay);
4164 }
4165 
4166 static void bnxt_dev_cleanup(struct bnxt *bp)
4167 {
4168 	bp->eth_dev->data->dev_link.link_status = 0;
4169 	bp->link_info->link_up = 0;
4170 	if (bp->eth_dev->data->dev_started)
4171 		bnxt_dev_stop(bp->eth_dev);
4172 
4173 	bnxt_uninit_resources(bp, true);
4174 }
4175 
4176 static int
4177 bnxt_check_fw_reset_done(struct bnxt *bp)
4178 {
4179 	int timeout = bp->fw_reset_max_msecs;
4180 	uint16_t val = 0;
4181 	int rc;
4182 
4183 	do {
4184 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4185 		if (rc < 0) {
4186 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4187 			return rc;
4188 		}
4189 		if (val != 0xffff)
4190 			break;
4191 		rte_delay_ms(1);
4192 	} while (timeout--);
4193 
4194 	if (val == 0xffff) {
4195 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4196 		return -1;
4197 	}
4198 
4199 	return 0;
4200 }
4201 
4202 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4203 {
4204 	struct rte_eth_dev *dev = bp->eth_dev;
4205 	struct rte_vlan_filter_conf *vfc;
4206 	int vidx, vbit, rc;
4207 	uint16_t vlan_id;
4208 
4209 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4210 		vfc = &dev->data->vlan_filter_conf;
4211 		vidx = vlan_id / 64;
4212 		vbit = vlan_id % 64;
4213 
4214 		/* Each bit corresponds to a VLAN id */
4215 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4216 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4217 			if (rc)
4218 				return rc;
4219 		}
4220 	}
4221 
4222 	return 0;
4223 }
4224 
4225 static int bnxt_restore_mac_filters(struct bnxt *bp)
4226 {
4227 	struct rte_eth_dev *dev = bp->eth_dev;
4228 	struct rte_eth_dev_info dev_info;
4229 	struct rte_ether_addr *addr;
4230 	uint64_t pool_mask;
4231 	uint32_t pool = 0;
4232 	uint32_t i;
4233 	int rc;
4234 
4235 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4236 		return 0;
4237 
4238 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4239 	if (rc)
4240 		return rc;
4241 
4242 	/* replay MAC address configuration */
4243 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4244 		addr = &dev->data->mac_addrs[i];
4245 
4246 		/* skip zero address */
4247 		if (rte_is_zero_ether_addr(addr))
4248 			continue;
4249 
4250 		pool = 0;
4251 		pool_mask = dev->data->mac_pool_sel[i];
4252 
4253 		do {
4254 			if (pool_mask & 1ULL) {
4255 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4256 				if (rc)
4257 					return rc;
4258 			}
4259 			pool_mask >>= 1;
4260 			pool++;
4261 		} while (pool_mask);
4262 	}
4263 
4264 	return 0;
4265 }
4266 
4267 static int bnxt_restore_filters(struct bnxt *bp)
4268 {
4269 	struct rte_eth_dev *dev = bp->eth_dev;
4270 	int ret = 0;
4271 
4272 	if (dev->data->all_multicast) {
4273 		ret = bnxt_allmulticast_enable_op(dev);
4274 		if (ret)
4275 			return ret;
4276 	}
4277 	if (dev->data->promiscuous) {
4278 		ret = bnxt_promiscuous_enable_op(dev);
4279 		if (ret)
4280 			return ret;
4281 	}
4282 
4283 	ret = bnxt_restore_mac_filters(bp);
4284 	if (ret)
4285 		return ret;
4286 
4287 	ret = bnxt_restore_vlan_filters(bp);
4288 	/* TODO restore other filters as well */
4289 	return ret;
4290 }
4291 
4292 static int bnxt_check_fw_ready(struct bnxt *bp)
4293 {
4294 	int timeout = bp->fw_reset_max_msecs;
4295 	int rc = 0;
4296 
4297 	do {
4298 		rc = bnxt_hwrm_poll_ver_get(bp);
4299 		if (rc == 0)
4300 			break;
4301 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4302 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4303 	} while (rc && timeout > 0);
4304 
4305 	if (rc)
4306 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4307 
4308 	return rc;
4309 }
4310 
4311 static void bnxt_dev_recover(void *arg)
4312 {
4313 	struct bnxt *bp = arg;
4314 	int rc = 0;
4315 
4316 	pthread_mutex_lock(&bp->err_recovery_lock);
4317 
4318 	if (!bp->fw_reset_min_msecs) {
4319 		rc = bnxt_check_fw_reset_done(bp);
4320 		if (rc)
4321 			goto err;
4322 	}
4323 
4324 	/* Clear Error flag so that device re-init should happen */
4325 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4326 
4327 	rc = bnxt_check_fw_ready(bp);
4328 	if (rc)
4329 		goto err;
4330 
4331 	rc = bnxt_init_resources(bp, true);
4332 	if (rc) {
4333 		PMD_DRV_LOG(ERR,
4334 			    "Failed to initialize resources after reset\n");
4335 		goto err;
4336 	}
4337 	/* clear reset flag as the device is initialized now */
4338 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4339 
4340 	rc = bnxt_dev_start_op(bp->eth_dev);
4341 	if (rc) {
4342 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4343 		goto err_start;
4344 	}
4345 
4346 	rc = bnxt_restore_filters(bp);
4347 	if (rc)
4348 		goto err_start;
4349 
4350 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4351 	pthread_mutex_unlock(&bp->err_recovery_lock);
4352 
4353 	return;
4354 err_start:
4355 	bnxt_dev_stop(bp->eth_dev);
4356 err:
4357 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4358 	bnxt_uninit_resources(bp, false);
4359 	if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4360 		rte_eth_dev_callback_process(bp->eth_dev,
4361 					     RTE_ETH_EVENT_INTR_RMV,
4362 					     NULL);
4363 	pthread_mutex_unlock(&bp->err_recovery_lock);
4364 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4365 }
4366 
4367 void bnxt_dev_reset_and_resume(void *arg)
4368 {
4369 	struct bnxt *bp = arg;
4370 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4371 	uint16_t val = 0;
4372 	int rc;
4373 
4374 	bnxt_dev_cleanup(bp);
4375 
4376 	bnxt_wait_for_device_shutdown(bp);
4377 
4378 	/* During some fatal firmware error conditions, the PCI config space
4379 	 * register 0x2e which normally contains the subsystem ID will become
4380 	 * 0xffff. This register will revert back to the normal value after
4381 	 * the chip has completed core reset. If we detect this condition,
4382 	 * we can poll this config register immediately for the value to revert.
4383 	 */
4384 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4385 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4386 		if (rc < 0) {
4387 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4388 			return;
4389 		}
4390 		if (val == 0xffff) {
4391 			bp->fw_reset_min_msecs = 0;
4392 			us = 1;
4393 		}
4394 	}
4395 
4396 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4397 	if (rc)
4398 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4399 }
4400 
4401 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4402 {
4403 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4404 	uint32_t reg = info->status_regs[index];
4405 	uint32_t type, offset, val = 0;
4406 	int ret = 0;
4407 
4408 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4409 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4410 
4411 	switch (type) {
4412 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4413 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4414 		if (ret < 0)
4415 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4416 				    offset);
4417 		break;
4418 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4419 		offset = info->mapped_status_regs[index];
4420 		/* FALLTHROUGH */
4421 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4422 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4423 				       offset));
4424 		break;
4425 	}
4426 
4427 	return val;
4428 }
4429 
4430 static int bnxt_fw_reset_all(struct bnxt *bp)
4431 {
4432 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4433 	uint32_t i;
4434 	int rc = 0;
4435 
4436 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4437 		/* Reset through primary function driver */
4438 		for (i = 0; i < info->reg_array_cnt; i++)
4439 			bnxt_write_fw_reset_reg(bp, i);
4440 		/* Wait for time specified by FW after triggering reset */
4441 		rte_delay_ms(info->primary_func_wait_period_after_reset);
4442 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4443 		/* Reset with the help of Kong processor */
4444 		rc = bnxt_hwrm_fw_reset(bp);
4445 		if (rc)
4446 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4447 	}
4448 
4449 	return rc;
4450 }
4451 
4452 static void bnxt_fw_reset_cb(void *arg)
4453 {
4454 	struct bnxt *bp = arg;
4455 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4456 	int rc = 0;
4457 
4458 	/* Only Primary function can do FW reset */
4459 	if (bnxt_is_primary_func(bp) &&
4460 	    bnxt_is_recovery_enabled(bp)) {
4461 		rc = bnxt_fw_reset_all(bp);
4462 		if (rc) {
4463 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4464 			return;
4465 		}
4466 	}
4467 
4468 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4469 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4470 	 * (including MASTER FUNC). After receiving this Async, all the active
4471 	 * drivers should treat this case as FW initiated recovery
4472 	 */
4473 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4474 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4475 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4476 
4477 		/* To recover from error */
4478 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4479 				  (void *)bp);
4480 	}
4481 }
4482 
4483 /* Driver should poll FW heartbeat, reset_counter with the frequency
4484  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4485  * When the driver detects heartbeat stop or change in reset_counter,
4486  * it has to trigger a reset to recover from the error condition.
4487  * A “primary function” is the function who will have the privilege to
4488  * initiate the chimp reset. The primary function will be elected by the
4489  * firmware and will be notified through async message.
4490  */
4491 static void bnxt_check_fw_health(void *arg)
4492 {
4493 	struct bnxt *bp = arg;
4494 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4495 	uint32_t val = 0, wait_msec;
4496 
4497 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4498 	    is_bnxt_in_error(bp))
4499 		return;
4500 
4501 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4502 	if (val == info->last_heart_beat)
4503 		goto reset;
4504 
4505 	info->last_heart_beat = val;
4506 
4507 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4508 	if (val != info->last_reset_counter)
4509 		goto reset;
4510 
4511 	info->last_reset_counter = val;
4512 
4513 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4514 			  bnxt_check_fw_health, (void *)bp);
4515 
4516 	return;
4517 reset:
4518 	/* Stop DMA to/from device */
4519 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4520 	bp->flags |= BNXT_FLAG_FW_RESET;
4521 
4522 	bnxt_stop_rxtx(bp);
4523 
4524 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4525 
4526 	if (bnxt_is_primary_func(bp))
4527 		wait_msec = info->primary_func_wait_period;
4528 	else
4529 		wait_msec = info->normal_func_wait_period;
4530 
4531 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4532 			  bnxt_fw_reset_cb, (void *)bp);
4533 }
4534 
4535 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4536 {
4537 	uint32_t polling_freq;
4538 
4539 	pthread_mutex_lock(&bp->health_check_lock);
4540 
4541 	if (!bnxt_is_recovery_enabled(bp))
4542 		goto done;
4543 
4544 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4545 		goto done;
4546 
4547 	polling_freq = bp->recovery_info->driver_polling_freq;
4548 
4549 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4550 			  bnxt_check_fw_health, (void *)bp);
4551 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4552 
4553 done:
4554 	pthread_mutex_unlock(&bp->health_check_lock);
4555 }
4556 
4557 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4558 {
4559 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4560 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4561 }
4562 
4563 static bool bnxt_vf_pciid(uint16_t device_id)
4564 {
4565 	switch (device_id) {
4566 	case BROADCOM_DEV_ID_57304_VF:
4567 	case BROADCOM_DEV_ID_57406_VF:
4568 	case BROADCOM_DEV_ID_5731X_VF:
4569 	case BROADCOM_DEV_ID_5741X_VF:
4570 	case BROADCOM_DEV_ID_57414_VF:
4571 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4572 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4573 	case BROADCOM_DEV_ID_58802_VF:
4574 	case BROADCOM_DEV_ID_57500_VF1:
4575 	case BROADCOM_DEV_ID_57500_VF2:
4576 	case BROADCOM_DEV_ID_58818_VF:
4577 		/* FALLTHROUGH */
4578 		return true;
4579 	default:
4580 		return false;
4581 	}
4582 }
4583 
4584 /* Phase 5 device */
4585 static bool bnxt_p5_device(uint16_t device_id)
4586 {
4587 	switch (device_id) {
4588 	case BROADCOM_DEV_ID_57508:
4589 	case BROADCOM_DEV_ID_57504:
4590 	case BROADCOM_DEV_ID_57502:
4591 	case BROADCOM_DEV_ID_57508_MF1:
4592 	case BROADCOM_DEV_ID_57504_MF1:
4593 	case BROADCOM_DEV_ID_57502_MF1:
4594 	case BROADCOM_DEV_ID_57508_MF2:
4595 	case BROADCOM_DEV_ID_57504_MF2:
4596 	case BROADCOM_DEV_ID_57502_MF2:
4597 	case BROADCOM_DEV_ID_57500_VF1:
4598 	case BROADCOM_DEV_ID_57500_VF2:
4599 	case BROADCOM_DEV_ID_58812:
4600 	case BROADCOM_DEV_ID_58814:
4601 	case BROADCOM_DEV_ID_58818:
4602 	case BROADCOM_DEV_ID_58818_VF:
4603 		/* FALLTHROUGH */
4604 		return true;
4605 	default:
4606 		return false;
4607 	}
4608 }
4609 
4610 bool bnxt_stratus_device(struct bnxt *bp)
4611 {
4612 	uint16_t device_id = bp->pdev->id.device_id;
4613 
4614 	switch (device_id) {
4615 	case BROADCOM_DEV_ID_STRATUS_NIC:
4616 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4617 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4618 		/* FALLTHROUGH */
4619 		return true;
4620 	default:
4621 		return false;
4622 	}
4623 }
4624 
4625 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4626 {
4627 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4628 	struct bnxt *bp = eth_dev->data->dev_private;
4629 
4630 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4631 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4632 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4633 	if (!bp->bar0 || !bp->doorbell_base) {
4634 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4635 		return -ENODEV;
4636 	}
4637 
4638 	bp->eth_dev = eth_dev;
4639 	bp->pdev = pci_dev;
4640 
4641 	return 0;
4642 }
4643 
4644 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4645 				  struct bnxt_ctx_pg_info *ctx_pg,
4646 				  uint32_t mem_size,
4647 				  const char *suffix,
4648 				  uint16_t idx)
4649 {
4650 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4651 	const struct rte_memzone *mz = NULL;
4652 	char mz_name[RTE_MEMZONE_NAMESIZE];
4653 	rte_iova_t mz_phys_addr;
4654 	uint64_t valid_bits = 0;
4655 	uint32_t sz;
4656 	int i;
4657 
4658 	if (!mem_size)
4659 		return 0;
4660 
4661 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4662 			 BNXT_PAGE_SIZE;
4663 	rmem->page_size = BNXT_PAGE_SIZE;
4664 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4665 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4666 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4667 
4668 	valid_bits = PTU_PTE_VALID;
4669 
4670 	if (rmem->nr_pages > 1) {
4671 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4672 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4673 			 suffix, idx, bp->eth_dev->data->port_id);
4674 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4675 		mz = rte_memzone_lookup(mz_name);
4676 		if (!mz) {
4677 			mz = rte_memzone_reserve_aligned(mz_name,
4678 						rmem->nr_pages * 8,
4679 						bp->eth_dev->device->numa_node,
4680 						RTE_MEMZONE_2MB |
4681 						RTE_MEMZONE_SIZE_HINT_ONLY |
4682 						RTE_MEMZONE_IOVA_CONTIG,
4683 						BNXT_PAGE_SIZE);
4684 			if (mz == NULL)
4685 				return -ENOMEM;
4686 		}
4687 
4688 		memset(mz->addr, 0, mz->len);
4689 		mz_phys_addr = mz->iova;
4690 
4691 		rmem->pg_tbl = mz->addr;
4692 		rmem->pg_tbl_map = mz_phys_addr;
4693 		rmem->pg_tbl_mz = mz;
4694 	}
4695 
4696 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4697 		 suffix, idx, bp->eth_dev->data->port_id);
4698 	mz = rte_memzone_lookup(mz_name);
4699 	if (!mz) {
4700 		mz = rte_memzone_reserve_aligned(mz_name,
4701 						 mem_size,
4702 						 bp->eth_dev->device->numa_node,
4703 						 RTE_MEMZONE_1GB |
4704 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4705 						 RTE_MEMZONE_IOVA_CONTIG,
4706 						 BNXT_PAGE_SIZE);
4707 		if (mz == NULL)
4708 			return -ENOMEM;
4709 	}
4710 
4711 	memset(mz->addr, 0, mz->len);
4712 	mz_phys_addr = mz->iova;
4713 
4714 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4715 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4716 		rmem->dma_arr[i] = mz_phys_addr + sz;
4717 
4718 		if (rmem->nr_pages > 1) {
4719 			if (i == rmem->nr_pages - 2 &&
4720 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4721 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4722 			else if (i == rmem->nr_pages - 1 &&
4723 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4724 				valid_bits |= PTU_PTE_LAST;
4725 
4726 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4727 							   valid_bits);
4728 		}
4729 	}
4730 
4731 	rmem->mz = mz;
4732 	if (rmem->vmem_size)
4733 		rmem->vmem = (void **)mz->addr;
4734 	rmem->dma_arr[0] = mz_phys_addr;
4735 	return 0;
4736 }
4737 
4738 static void bnxt_free_ctx_mem(struct bnxt *bp)
4739 {
4740 	int i;
4741 
4742 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4743 		return;
4744 
4745 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4746 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4747 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4748 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4749 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4750 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4751 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4752 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4753 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4754 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4755 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4756 
4757 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4758 		if (bp->ctx->tqm_mem[i])
4759 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4760 	}
4761 
4762 	rte_free(bp->ctx);
4763 	bp->ctx = NULL;
4764 }
4765 
4766 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4767 
4768 #define min_t(type, x, y) ({                    \
4769 	type __min1 = (x);                      \
4770 	type __min2 = (y);                      \
4771 	__min1 < __min2 ? __min1 : __min2; })
4772 
4773 #define max_t(type, x, y) ({                    \
4774 	type __max1 = (x);                      \
4775 	type __max2 = (y);                      \
4776 	__max1 > __max2 ? __max1 : __max2; })
4777 
4778 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4779 
4780 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4781 {
4782 	struct bnxt_ctx_pg_info *ctx_pg;
4783 	struct bnxt_ctx_mem_info *ctx;
4784 	uint32_t mem_size, ena, entries;
4785 	uint32_t entries_sp, min;
4786 	int i, rc;
4787 
4788 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4789 	if (rc) {
4790 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4791 		return rc;
4792 	}
4793 	ctx = bp->ctx;
4794 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4795 		return 0;
4796 
4797 	ctx_pg = &ctx->qp_mem;
4798 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4799 	if (ctx->qp_entry_size) {
4800 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4801 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4802 		if (rc)
4803 			return rc;
4804 	}
4805 
4806 	ctx_pg = &ctx->srq_mem;
4807 	ctx_pg->entries = ctx->srq_max_l2_entries;
4808 	if (ctx->srq_entry_size) {
4809 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4810 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4811 		if (rc)
4812 			return rc;
4813 	}
4814 
4815 	ctx_pg = &ctx->cq_mem;
4816 	ctx_pg->entries = ctx->cq_max_l2_entries;
4817 	if (ctx->cq_entry_size) {
4818 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4819 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4820 		if (rc)
4821 			return rc;
4822 	}
4823 
4824 	ctx_pg = &ctx->vnic_mem;
4825 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4826 		ctx->vnic_max_ring_table_entries;
4827 	if (ctx->vnic_entry_size) {
4828 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4829 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4830 		if (rc)
4831 			return rc;
4832 	}
4833 
4834 	ctx_pg = &ctx->stat_mem;
4835 	ctx_pg->entries = ctx->stat_max_entries;
4836 	if (ctx->stat_entry_size) {
4837 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4838 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4839 		if (rc)
4840 			return rc;
4841 	}
4842 
4843 	min = ctx->tqm_min_entries_per_ring;
4844 
4845 	entries_sp = ctx->qp_max_l2_entries +
4846 		     ctx->vnic_max_vnic_entries +
4847 		     2 * ctx->qp_min_qp1_entries + min;
4848 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4849 
4850 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4851 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4852 	entries = clamp_t(uint32_t, entries, min,
4853 			  ctx->tqm_max_entries_per_ring);
4854 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4855 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4856 		 * i > 8 is other ext rings.
4857 		 */
4858 		ctx_pg = ctx->tqm_mem[i];
4859 		ctx_pg->entries = i ? entries : entries_sp;
4860 		if (ctx->tqm_entry_size) {
4861 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4862 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4863 						    "tqm_mem", i);
4864 			if (rc)
4865 				return rc;
4866 		}
4867 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4868 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4869 		else
4870 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4871 	}
4872 
4873 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4874 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4875 	if (rc)
4876 		PMD_DRV_LOG(ERR,
4877 			    "Failed to configure context mem: rc = %d\n", rc);
4878 	else
4879 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4880 
4881 	return rc;
4882 }
4883 
4884 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4885 {
4886 	struct rte_pci_device *pci_dev = bp->pdev;
4887 	char mz_name[RTE_MEMZONE_NAMESIZE];
4888 	const struct rte_memzone *mz = NULL;
4889 	uint32_t total_alloc_len;
4890 	rte_iova_t mz_phys_addr;
4891 
4892 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4893 		return 0;
4894 
4895 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4896 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4897 		 pci_dev->addr.bus, pci_dev->addr.devid,
4898 		 pci_dev->addr.function, "rx_port_stats");
4899 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4900 	mz = rte_memzone_lookup(mz_name);
4901 	total_alloc_len =
4902 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4903 				       sizeof(struct rx_port_stats_ext) + 512);
4904 	if (!mz) {
4905 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4906 					 SOCKET_ID_ANY,
4907 					 RTE_MEMZONE_2MB |
4908 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4909 					 RTE_MEMZONE_IOVA_CONTIG);
4910 		if (mz == NULL)
4911 			return -ENOMEM;
4912 	}
4913 	memset(mz->addr, 0, mz->len);
4914 	mz_phys_addr = mz->iova;
4915 
4916 	bp->rx_mem_zone = (const void *)mz;
4917 	bp->hw_rx_port_stats = mz->addr;
4918 	bp->hw_rx_port_stats_map = mz_phys_addr;
4919 
4920 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4921 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4922 		 pci_dev->addr.bus, pci_dev->addr.devid,
4923 		 pci_dev->addr.function, "tx_port_stats");
4924 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4925 	mz = rte_memzone_lookup(mz_name);
4926 	total_alloc_len =
4927 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4928 				       sizeof(struct tx_port_stats_ext) + 512);
4929 	if (!mz) {
4930 		mz = rte_memzone_reserve(mz_name,
4931 					 total_alloc_len,
4932 					 SOCKET_ID_ANY,
4933 					 RTE_MEMZONE_2MB |
4934 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4935 					 RTE_MEMZONE_IOVA_CONTIG);
4936 		if (mz == NULL)
4937 			return -ENOMEM;
4938 	}
4939 	memset(mz->addr, 0, mz->len);
4940 	mz_phys_addr = mz->iova;
4941 
4942 	bp->tx_mem_zone = (const void *)mz;
4943 	bp->hw_tx_port_stats = mz->addr;
4944 	bp->hw_tx_port_stats_map = mz_phys_addr;
4945 	bp->flags |= BNXT_FLAG_PORT_STATS;
4946 
4947 	/* Display extended statistics if FW supports it */
4948 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4949 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4950 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4951 		return 0;
4952 
4953 	bp->hw_rx_port_stats_ext = (void *)
4954 		((uint8_t *)bp->hw_rx_port_stats +
4955 		 sizeof(struct rx_port_stats));
4956 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4957 		sizeof(struct rx_port_stats);
4958 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4959 
4960 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4961 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4962 		bp->hw_tx_port_stats_ext = (void *)
4963 			((uint8_t *)bp->hw_tx_port_stats +
4964 			 sizeof(struct tx_port_stats));
4965 		bp->hw_tx_port_stats_ext_map =
4966 			bp->hw_tx_port_stats_map +
4967 			sizeof(struct tx_port_stats);
4968 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4969 	}
4970 
4971 	return 0;
4972 }
4973 
4974 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4975 {
4976 	struct bnxt *bp = eth_dev->data->dev_private;
4977 	int rc = 0;
4978 
4979 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4980 					       RTE_ETHER_ADDR_LEN *
4981 					       bp->max_l2_ctx,
4982 					       0);
4983 	if (eth_dev->data->mac_addrs == NULL) {
4984 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4985 		return -ENOMEM;
4986 	}
4987 
4988 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4989 		if (BNXT_PF(bp))
4990 			return -EINVAL;
4991 
4992 		/* Generate a random MAC address, if none was assigned by PF */
4993 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4994 		bnxt_eth_hw_addr_random(bp->mac_addr);
4995 		PMD_DRV_LOG(INFO,
4996 			    "Assign random MAC:" RTE_ETHER_ADDR_PRT_FMT "\n",
4997 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4998 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4999 
5000 		rc = bnxt_hwrm_set_mac(bp);
5001 		if (rc)
5002 			return rc;
5003 	}
5004 
5005 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5006 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5007 
5008 	return rc;
5009 }
5010 
5011 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5012 {
5013 	int rc = 0;
5014 
5015 	/* MAC is already configured in FW */
5016 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5017 		return 0;
5018 
5019 	/* Restore the old MAC configured */
5020 	rc = bnxt_hwrm_set_mac(bp);
5021 	if (rc)
5022 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5023 
5024 	return rc;
5025 }
5026 
5027 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5028 {
5029 	if (!BNXT_PF(bp))
5030 		return;
5031 
5032 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5033 
5034 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5035 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5036 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5037 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5038 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5039 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5040 }
5041 
5042 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5043 {
5044 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5045 
5046 	if (info) {
5047 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5048 			memset(info, 0, sizeof(*info));
5049 		return;
5050 	}
5051 
5052 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5053 		return;
5054 
5055 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5056 			   sizeof(*info), 0);
5057 	if (!info)
5058 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5059 
5060 	bp->recovery_info = info;
5061 }
5062 
5063 static void bnxt_check_fw_status(struct bnxt *bp)
5064 {
5065 	uint32_t fw_status;
5066 
5067 	if (!(bp->recovery_info &&
5068 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5069 		return;
5070 
5071 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5072 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5073 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5074 			    fw_status);
5075 }
5076 
5077 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5078 {
5079 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5080 	uint32_t status_loc;
5081 	uint32_t sig_ver;
5082 
5083 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5084 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5085 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5086 				   BNXT_GRCP_WINDOW_2_BASE +
5087 				   offsetof(struct hcomm_status,
5088 					    sig_ver)));
5089 	/* If the signature is absent, then FW does not support this feature */
5090 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5091 	    HCOMM_STATUS_SIGNATURE_VAL)
5092 		return 0;
5093 
5094 	if (!info) {
5095 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5096 				   sizeof(*info), 0);
5097 		if (!info)
5098 			return -ENOMEM;
5099 		bp->recovery_info = info;
5100 	} else {
5101 		memset(info, 0, sizeof(*info));
5102 	}
5103 
5104 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5105 				      BNXT_GRCP_WINDOW_2_BASE +
5106 				      offsetof(struct hcomm_status,
5107 					       fw_status_loc)));
5108 
5109 	/* Only pre-map the FW health status GRC register */
5110 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5111 		return 0;
5112 
5113 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5114 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5115 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5116 
5117 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5118 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5119 
5120 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5121 
5122 	return 0;
5123 }
5124 
5125 /* This function gets the FW version along with the
5126  * capabilities(MAX and current) of the function, vnic,
5127  * error recovery, phy and other chip related info
5128  */
5129 static int bnxt_get_config(struct bnxt *bp)
5130 {
5131 	uint16_t mtu;
5132 	int rc = 0;
5133 
5134 	bp->fw_cap = 0;
5135 
5136 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5137 	if (rc)
5138 		return rc;
5139 
5140 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5141 	if (rc) {
5142 		bnxt_check_fw_status(bp);
5143 		return rc;
5144 	}
5145 
5146 	rc = bnxt_hwrm_func_reset(bp);
5147 	if (rc)
5148 		return -EIO;
5149 
5150 	rc = bnxt_hwrm_vnic_qcaps(bp);
5151 	if (rc)
5152 		return rc;
5153 
5154 	rc = bnxt_hwrm_queue_qportcfg(bp);
5155 	if (rc)
5156 		return rc;
5157 
5158 	/* Get the MAX capabilities for this function.
5159 	 * This function also allocates context memory for TQM rings and
5160 	 * informs the firmware about this allocated backing store memory.
5161 	 */
5162 	rc = bnxt_hwrm_func_qcaps(bp);
5163 	if (rc)
5164 		return rc;
5165 
5166 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5167 	if (rc)
5168 		return rc;
5169 
5170 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5171 	if (rc)
5172 		return rc;
5173 
5174 	bnxt_hwrm_port_mac_qcfg(bp);
5175 
5176 	bnxt_hwrm_parent_pf_qcfg(bp);
5177 
5178 	bnxt_hwrm_port_phy_qcaps(bp);
5179 
5180 	bnxt_alloc_error_recovery_info(bp);
5181 	/* Get the adapter error recovery support info */
5182 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5183 	if (rc)
5184 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5185 
5186 	bnxt_hwrm_port_led_qcaps(bp);
5187 
5188 	return 0;
5189 }
5190 
5191 static int
5192 bnxt_init_locks(struct bnxt *bp)
5193 {
5194 	int err;
5195 
5196 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5197 	if (err) {
5198 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5199 		return err;
5200 	}
5201 
5202 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5203 	if (err) {
5204 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5205 		return err;
5206 	}
5207 
5208 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5209 	if (err) {
5210 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5211 		return err;
5212 	}
5213 
5214 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5215 	if (err)
5216 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5217 
5218 	return err;
5219 }
5220 
5221 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5222 {
5223 	int rc = 0;
5224 
5225 	rc = bnxt_get_config(bp);
5226 	if (rc)
5227 		return rc;
5228 
5229 	if (!reconfig_dev) {
5230 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5231 		if (rc)
5232 			return rc;
5233 	} else {
5234 		rc = bnxt_restore_dflt_mac(bp);
5235 		if (rc)
5236 			return rc;
5237 	}
5238 
5239 	bnxt_config_vf_req_fwd(bp);
5240 
5241 	rc = bnxt_hwrm_func_driver_register(bp);
5242 	if (rc) {
5243 		PMD_DRV_LOG(ERR, "Failed to register driver");
5244 		return -EBUSY;
5245 	}
5246 
5247 	if (BNXT_PF(bp)) {
5248 		if (bp->pdev->max_vfs) {
5249 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5250 			if (rc) {
5251 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5252 				return rc;
5253 			}
5254 		} else {
5255 			rc = bnxt_hwrm_allocate_pf_only(bp);
5256 			if (rc) {
5257 				PMD_DRV_LOG(ERR,
5258 					    "Failed to allocate PF resources");
5259 				return rc;
5260 			}
5261 		}
5262 	}
5263 
5264 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5265 	if (rc)
5266 		return rc;
5267 
5268 	rc = bnxt_setup_int(bp);
5269 	if (rc)
5270 		return rc;
5271 
5272 	rc = bnxt_request_int(bp);
5273 	if (rc)
5274 		return rc;
5275 
5276 	rc = bnxt_init_ctx_mem(bp);
5277 	if (rc) {
5278 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5279 		return rc;
5280 	}
5281 
5282 	return 0;
5283 }
5284 
5285 static int
5286 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5287 			     const char *value, void *opaque_arg)
5288 {
5289 	struct bnxt *bp = opaque_arg;
5290 	unsigned long flow_xstat;
5291 	char *end = NULL;
5292 
5293 	if (!value || !opaque_arg) {
5294 		PMD_DRV_LOG(ERR,
5295 			    "Invalid parameter passed to flow_xstat devarg.\n");
5296 		return -EINVAL;
5297 	}
5298 
5299 	flow_xstat = strtoul(value, &end, 10);
5300 	if (end == NULL || *end != '\0' ||
5301 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5302 		PMD_DRV_LOG(ERR,
5303 			    "Invalid parameter passed to flow_xstat devarg.\n");
5304 		return -EINVAL;
5305 	}
5306 
5307 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5308 		PMD_DRV_LOG(ERR,
5309 			    "Invalid value passed to flow_xstat devarg.\n");
5310 		return -EINVAL;
5311 	}
5312 
5313 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5314 	if (BNXT_FLOW_XSTATS_EN(bp))
5315 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5316 
5317 	return 0;
5318 }
5319 
5320 static int
5321 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5322 					const char *value, void *opaque_arg)
5323 {
5324 	struct bnxt *bp = opaque_arg;
5325 	unsigned long max_num_kflows;
5326 	char *end = NULL;
5327 
5328 	if (!value || !opaque_arg) {
5329 		PMD_DRV_LOG(ERR,
5330 			"Invalid parameter passed to max_num_kflows devarg.\n");
5331 		return -EINVAL;
5332 	}
5333 
5334 	max_num_kflows = strtoul(value, &end, 10);
5335 	if (end == NULL || *end != '\0' ||
5336 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5337 		PMD_DRV_LOG(ERR,
5338 			"Invalid parameter passed to max_num_kflows devarg.\n");
5339 		return -EINVAL;
5340 	}
5341 
5342 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5343 		PMD_DRV_LOG(ERR,
5344 			"Invalid value passed to max_num_kflows devarg.\n");
5345 		return -EINVAL;
5346 	}
5347 
5348 	bp->max_num_kflows = max_num_kflows;
5349 	if (bp->max_num_kflows)
5350 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5351 				max_num_kflows);
5352 
5353 	return 0;
5354 }
5355 
5356 static int
5357 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5358 				 const char *value, void *opaque_arg)
5359 {
5360 	struct bnxt *bp = opaque_arg;
5361 	unsigned long app_id;
5362 	char *end = NULL;
5363 
5364 	if (!value || !opaque_arg) {
5365 		PMD_DRV_LOG(ERR,
5366 			    "Invalid parameter passed to app-id "
5367 			    "devargs.\n");
5368 		return -EINVAL;
5369 	}
5370 
5371 	app_id = strtoul(value, &end, 10);
5372 	if (end == NULL || *end != '\0' ||
5373 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5374 		PMD_DRV_LOG(ERR,
5375 			    "Invalid parameter passed to app_id "
5376 			    "devargs.\n");
5377 		return -EINVAL;
5378 	}
5379 
5380 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5381 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5382 			    (uint16_t)app_id);
5383 		return -EINVAL;
5384 	}
5385 
5386 	bp->app_id = app_id;
5387 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5388 
5389 	return 0;
5390 }
5391 
5392 static int
5393 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5394 			    const char *value, void *opaque_arg)
5395 {
5396 	struct bnxt_representor *vfr_bp = opaque_arg;
5397 	unsigned long rep_is_pf;
5398 	char *end = NULL;
5399 
5400 	if (!value || !opaque_arg) {
5401 		PMD_DRV_LOG(ERR,
5402 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5403 		return -EINVAL;
5404 	}
5405 
5406 	rep_is_pf = strtoul(value, &end, 10);
5407 	if (end == NULL || *end != '\0' ||
5408 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5409 		PMD_DRV_LOG(ERR,
5410 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5411 		return -EINVAL;
5412 	}
5413 
5414 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5415 		PMD_DRV_LOG(ERR,
5416 			    "Invalid value passed to rep_is_pf devargs.\n");
5417 		return -EINVAL;
5418 	}
5419 
5420 	vfr_bp->flags |= rep_is_pf;
5421 	if (BNXT_REP_PF(vfr_bp))
5422 		PMD_DRV_LOG(INFO, "PF representor\n");
5423 	else
5424 		PMD_DRV_LOG(INFO, "VF representor\n");
5425 
5426 	return 0;
5427 }
5428 
5429 static int
5430 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5431 			       const char *value, void *opaque_arg)
5432 {
5433 	struct bnxt_representor *vfr_bp = opaque_arg;
5434 	unsigned long rep_based_pf;
5435 	char *end = NULL;
5436 
5437 	if (!value || !opaque_arg) {
5438 		PMD_DRV_LOG(ERR,
5439 			    "Invalid parameter passed to rep_based_pf "
5440 			    "devargs.\n");
5441 		return -EINVAL;
5442 	}
5443 
5444 	rep_based_pf = strtoul(value, &end, 10);
5445 	if (end == NULL || *end != '\0' ||
5446 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5447 		PMD_DRV_LOG(ERR,
5448 			    "Invalid parameter passed to rep_based_pf "
5449 			    "devargs.\n");
5450 		return -EINVAL;
5451 	}
5452 
5453 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5454 		PMD_DRV_LOG(ERR,
5455 			    "Invalid value passed to rep_based_pf devargs.\n");
5456 		return -EINVAL;
5457 	}
5458 
5459 	vfr_bp->rep_based_pf = rep_based_pf;
5460 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5461 
5462 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5463 
5464 	return 0;
5465 }
5466 
5467 static int
5468 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5469 			    const char *value, void *opaque_arg)
5470 {
5471 	struct bnxt_representor *vfr_bp = opaque_arg;
5472 	unsigned long rep_q_r2f;
5473 	char *end = NULL;
5474 
5475 	if (!value || !opaque_arg) {
5476 		PMD_DRV_LOG(ERR,
5477 			    "Invalid parameter passed to rep_q_r2f "
5478 			    "devargs.\n");
5479 		return -EINVAL;
5480 	}
5481 
5482 	rep_q_r2f = strtoul(value, &end, 10);
5483 	if (end == NULL || *end != '\0' ||
5484 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5485 		PMD_DRV_LOG(ERR,
5486 			    "Invalid parameter passed to rep_q_r2f "
5487 			    "devargs.\n");
5488 		return -EINVAL;
5489 	}
5490 
5491 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5492 		PMD_DRV_LOG(ERR,
5493 			    "Invalid value passed to rep_q_r2f devargs.\n");
5494 		return -EINVAL;
5495 	}
5496 
5497 	vfr_bp->rep_q_r2f = rep_q_r2f;
5498 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5499 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5500 
5501 	return 0;
5502 }
5503 
5504 static int
5505 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5506 			    const char *value, void *opaque_arg)
5507 {
5508 	struct bnxt_representor *vfr_bp = opaque_arg;
5509 	unsigned long rep_q_f2r;
5510 	char *end = NULL;
5511 
5512 	if (!value || !opaque_arg) {
5513 		PMD_DRV_LOG(ERR,
5514 			    "Invalid parameter passed to rep_q_f2r "
5515 			    "devargs.\n");
5516 		return -EINVAL;
5517 	}
5518 
5519 	rep_q_f2r = strtoul(value, &end, 10);
5520 	if (end == NULL || *end != '\0' ||
5521 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5522 		PMD_DRV_LOG(ERR,
5523 			    "Invalid parameter passed to rep_q_f2r "
5524 			    "devargs.\n");
5525 		return -EINVAL;
5526 	}
5527 
5528 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5529 		PMD_DRV_LOG(ERR,
5530 			    "Invalid value passed to rep_q_f2r devargs.\n");
5531 		return -EINVAL;
5532 	}
5533 
5534 	vfr_bp->rep_q_f2r = rep_q_f2r;
5535 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5536 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5537 
5538 	return 0;
5539 }
5540 
5541 static int
5542 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5543 			     const char *value, void *opaque_arg)
5544 {
5545 	struct bnxt_representor *vfr_bp = opaque_arg;
5546 	unsigned long rep_fc_r2f;
5547 	char *end = NULL;
5548 
5549 	if (!value || !opaque_arg) {
5550 		PMD_DRV_LOG(ERR,
5551 			    "Invalid parameter passed to rep_fc_r2f "
5552 			    "devargs.\n");
5553 		return -EINVAL;
5554 	}
5555 
5556 	rep_fc_r2f = strtoul(value, &end, 10);
5557 	if (end == NULL || *end != '\0' ||
5558 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5559 		PMD_DRV_LOG(ERR,
5560 			    "Invalid parameter passed to rep_fc_r2f "
5561 			    "devargs.\n");
5562 		return -EINVAL;
5563 	}
5564 
5565 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5566 		PMD_DRV_LOG(ERR,
5567 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5568 		return -EINVAL;
5569 	}
5570 
5571 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5572 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5573 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5574 
5575 	return 0;
5576 }
5577 
5578 static int
5579 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5580 			     const char *value, void *opaque_arg)
5581 {
5582 	struct bnxt_representor *vfr_bp = opaque_arg;
5583 	unsigned long rep_fc_f2r;
5584 	char *end = NULL;
5585 
5586 	if (!value || !opaque_arg) {
5587 		PMD_DRV_LOG(ERR,
5588 			    "Invalid parameter passed to rep_fc_f2r "
5589 			    "devargs.\n");
5590 		return -EINVAL;
5591 	}
5592 
5593 	rep_fc_f2r = strtoul(value, &end, 10);
5594 	if (end == NULL || *end != '\0' ||
5595 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5596 		PMD_DRV_LOG(ERR,
5597 			    "Invalid parameter passed to rep_fc_f2r "
5598 			    "devargs.\n");
5599 		return -EINVAL;
5600 	}
5601 
5602 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5603 		PMD_DRV_LOG(ERR,
5604 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5605 		return -EINVAL;
5606 	}
5607 
5608 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5609 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5610 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5611 
5612 	return 0;
5613 }
5614 
5615 static int
5616 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5617 {
5618 	struct rte_kvargs *kvlist;
5619 	int ret;
5620 
5621 	if (devargs == NULL)
5622 		return 0;
5623 
5624 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5625 	if (kvlist == NULL)
5626 		return -EINVAL;
5627 
5628 	/*
5629 	 * Handler for "flow_xstat" devarg.
5630 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5631 	 */
5632 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5633 				 bnxt_parse_devarg_flow_xstat, bp);
5634 	if (ret)
5635 		goto err;
5636 
5637 	/*
5638 	 * Handler for "max_num_kflows" devarg.
5639 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5640 	 */
5641 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5642 				 bnxt_parse_devarg_max_num_kflows, bp);
5643 	if (ret)
5644 		goto err;
5645 
5646 err:
5647 	/*
5648 	 * Handler for "app-id" devarg.
5649 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5650 	 */
5651 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5652 			   bnxt_parse_devarg_app_id, bp);
5653 
5654 	rte_kvargs_free(kvlist);
5655 	return ret;
5656 }
5657 
5658 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5659 {
5660 	int rc = 0;
5661 
5662 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5663 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5664 		if (rc)
5665 			PMD_DRV_LOG(ERR,
5666 				    "Failed to alloc switch domain: %d\n", rc);
5667 		else
5668 			PMD_DRV_LOG(INFO,
5669 				    "Switch domain allocated %d\n",
5670 				    bp->switch_domain_id);
5671 	}
5672 
5673 	return rc;
5674 }
5675 
5676 /* Allocate and initialize various fields in bnxt struct that
5677  * need to be allocated/destroyed only once in the lifetime of the driver
5678  */
5679 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5680 {
5681 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5682 	struct bnxt *bp = eth_dev->data->dev_private;
5683 	int rc = 0;
5684 
5685 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5686 
5687 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5688 		bp->flags |= BNXT_FLAG_VF;
5689 
5690 	if (bnxt_p5_device(pci_dev->id.device_id))
5691 		bp->flags |= BNXT_FLAG_CHIP_P5;
5692 
5693 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5694 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5695 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5696 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5697 		bp->flags |= BNXT_FLAG_STINGRAY;
5698 
5699 	if (BNXT_TRUFLOW_EN(bp)) {
5700 		/* extra mbuf field is required to store CFA code from mark */
5701 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5702 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5703 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5704 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5705 		};
5706 		bnxt_cfa_code_dynfield_offset =
5707 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5708 		if (bnxt_cfa_code_dynfield_offset < 0) {
5709 			PMD_DRV_LOG(ERR,
5710 			    "Failed to register mbuf field for TruFlow mark\n");
5711 			return -rte_errno;
5712 		}
5713 	}
5714 
5715 	rc = bnxt_map_pci_bars(eth_dev);
5716 	if (rc) {
5717 		PMD_DRV_LOG(ERR,
5718 			    "Failed to initialize board rc: %x\n", rc);
5719 		return rc;
5720 	}
5721 
5722 	rc = bnxt_alloc_pf_info(bp);
5723 	if (rc)
5724 		return rc;
5725 
5726 	rc = bnxt_alloc_link_info(bp);
5727 	if (rc)
5728 		return rc;
5729 
5730 	rc = bnxt_alloc_parent_info(bp);
5731 	if (rc)
5732 		return rc;
5733 
5734 	rc = bnxt_alloc_hwrm_resources(bp);
5735 	if (rc) {
5736 		PMD_DRV_LOG(ERR,
5737 			    "Failed to allocate response buffer rc: %x\n", rc);
5738 		return rc;
5739 	}
5740 	rc = bnxt_alloc_leds_info(bp);
5741 	if (rc)
5742 		return rc;
5743 
5744 	rc = bnxt_alloc_cos_queues(bp);
5745 	if (rc)
5746 		return rc;
5747 
5748 	rc = bnxt_init_locks(bp);
5749 	if (rc)
5750 		return rc;
5751 
5752 	rc = bnxt_alloc_switch_domain(bp);
5753 	if (rc)
5754 		return rc;
5755 
5756 	return rc;
5757 }
5758 
5759 static int
5760 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5761 {
5762 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5763 	static int version_printed;
5764 	struct bnxt *bp;
5765 	int rc;
5766 
5767 	if (version_printed++ == 0)
5768 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5769 
5770 	eth_dev->dev_ops = &bnxt_dev_ops;
5771 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5772 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5773 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5774 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5775 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5776 
5777 	/*
5778 	 * For secondary processes, we don't initialise any further
5779 	 * as primary has already done this work.
5780 	 */
5781 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5782 		return 0;
5783 
5784 	rte_eth_copy_pci_info(eth_dev, pci_dev);
5785 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5786 
5787 	bp = eth_dev->data->dev_private;
5788 
5789 	/* Parse dev arguments passed on when starting the DPDK application. */
5790 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5791 	if (rc)
5792 		goto error_free;
5793 
5794 	rc = bnxt_drv_init(eth_dev);
5795 	if (rc)
5796 		goto error_free;
5797 
5798 	rc = bnxt_init_resources(bp, false);
5799 	if (rc)
5800 		goto error_free;
5801 
5802 	rc = bnxt_alloc_stats_mem(bp);
5803 	if (rc)
5804 		goto error_free;
5805 
5806 	PMD_DRV_LOG(INFO,
5807 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
5808 		    DRV_MODULE_NAME,
5809 		    pci_dev->mem_resource[0].phys_addr,
5810 		    pci_dev->mem_resource[0].addr);
5811 
5812 	return 0;
5813 
5814 error_free:
5815 	bnxt_dev_uninit(eth_dev);
5816 	return rc;
5817 }
5818 
5819 
5820 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5821 {
5822 	if (!ctx)
5823 		return;
5824 
5825 	if (ctx->va)
5826 		rte_free(ctx->va);
5827 
5828 	ctx->va = NULL;
5829 	ctx->dma = RTE_BAD_IOVA;
5830 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5831 }
5832 
5833 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5834 {
5835 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5836 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5837 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
5838 				  bp->flow_stat->max_fc,
5839 				  false);
5840 
5841 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5842 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5843 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
5844 				  bp->flow_stat->max_fc,
5845 				  false);
5846 
5847 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5848 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5849 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5850 
5851 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5852 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5853 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5854 
5855 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5856 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5857 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5858 
5859 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5860 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5861 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5862 }
5863 
5864 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5865 {
5866 	bnxt_unregister_fc_ctx_mem(bp);
5867 
5868 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5869 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5870 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5871 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5872 }
5873 
5874 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5875 {
5876 	if (BNXT_FLOW_XSTATS_EN(bp))
5877 		bnxt_uninit_fc_ctx_mem(bp);
5878 }
5879 
5880 static void
5881 bnxt_free_error_recovery_info(struct bnxt *bp)
5882 {
5883 	rte_free(bp->recovery_info);
5884 	bp->recovery_info = NULL;
5885 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5886 }
5887 
5888 static int
5889 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5890 {
5891 	int rc;
5892 
5893 	bnxt_free_int(bp);
5894 	bnxt_free_mem(bp, reconfig_dev);
5895 
5896 	bnxt_hwrm_func_buf_unrgtr(bp);
5897 	if (bp->pf != NULL) {
5898 		rte_free(bp->pf->vf_req_buf);
5899 		bp->pf->vf_req_buf = NULL;
5900 	}
5901 
5902 	rc = bnxt_hwrm_func_driver_unregister(bp);
5903 	bp->flags &= ~BNXT_FLAG_REGISTERED;
5904 	bnxt_free_ctx_mem(bp);
5905 	if (!reconfig_dev) {
5906 		bnxt_free_hwrm_resources(bp);
5907 		bnxt_free_error_recovery_info(bp);
5908 	}
5909 
5910 	bnxt_uninit_ctx_mem(bp);
5911 
5912 	bnxt_free_flow_stats_info(bp);
5913 	if (bp->rep_info != NULL)
5914 		bnxt_free_switch_domain(bp);
5915 	bnxt_free_rep_info(bp);
5916 	rte_free(bp->ptp_cfg);
5917 	bp->ptp_cfg = NULL;
5918 	return rc;
5919 }
5920 
5921 static int
5922 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5923 {
5924 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5925 		return -EPERM;
5926 
5927 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5928 
5929 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5930 		bnxt_dev_close_op(eth_dev);
5931 
5932 	return 0;
5933 }
5934 
5935 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
5936 {
5937 	struct bnxt *bp = eth_dev->data->dev_private;
5938 	struct rte_eth_dev *vf_rep_eth_dev;
5939 	int ret = 0, i;
5940 
5941 	if (!bp)
5942 		return -EINVAL;
5943 
5944 	for (i = 0; i < bp->num_reps; i++) {
5945 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
5946 		if (!vf_rep_eth_dev)
5947 			continue;
5948 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
5949 			    vf_rep_eth_dev->data->port_id);
5950 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
5951 	}
5952 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
5953 		    eth_dev->data->port_id);
5954 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
5955 
5956 	return ret;
5957 }
5958 
5959 static void bnxt_free_rep_info(struct bnxt *bp)
5960 {
5961 	rte_free(bp->rep_info);
5962 	bp->rep_info = NULL;
5963 	rte_free(bp->cfa_code_map);
5964 	bp->cfa_code_map = NULL;
5965 }
5966 
5967 static int bnxt_init_rep_info(struct bnxt *bp)
5968 {
5969 	int i = 0, rc;
5970 
5971 	if (bp->rep_info)
5972 		return 0;
5973 
5974 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
5975 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS(bp),
5976 				   0);
5977 	if (!bp->rep_info) {
5978 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
5979 		return -ENOMEM;
5980 	}
5981 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
5982 				       sizeof(*bp->cfa_code_map) *
5983 				       BNXT_MAX_CFA_CODE, 0);
5984 	if (!bp->cfa_code_map) {
5985 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
5986 		bnxt_free_rep_info(bp);
5987 		return -ENOMEM;
5988 	}
5989 
5990 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
5991 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
5992 
5993 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
5994 	if (rc) {
5995 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
5996 		bnxt_free_rep_info(bp);
5997 		return rc;
5998 	}
5999 
6000 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6001 	if (rc) {
6002 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6003 		bnxt_free_rep_info(bp);
6004 		return rc;
6005 	}
6006 
6007 	return rc;
6008 }
6009 
6010 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6011 			       struct rte_eth_devargs *eth_da,
6012 			       struct rte_eth_dev *backing_eth_dev,
6013 			       const char *dev_args)
6014 {
6015 	struct rte_eth_dev *vf_rep_eth_dev;
6016 	char name[RTE_ETH_NAME_MAX_LEN];
6017 	struct bnxt *backing_bp = backing_eth_dev->data->dev_private;
6018 	uint16_t max_vf_reps = BNXT_MAX_VF_REPS(backing_bp);
6019 
6020 	uint16_t num_rep;
6021 	int i, ret = 0;
6022 	struct rte_kvargs *kvlist = NULL;
6023 
6024 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6025 		return 0;
6026 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6027 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6028 			    eth_da->type);
6029 		return -ENOTSUP;
6030 	}
6031 	num_rep = eth_da->nb_representor_ports;
6032 	if (num_rep > max_vf_reps) {
6033 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6034 			    num_rep, max_vf_reps);
6035 		return -EINVAL;
6036 	}
6037 
6038 	if (num_rep >= RTE_MAX_ETHPORTS) {
6039 		PMD_DRV_LOG(ERR,
6040 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6041 			    num_rep, RTE_MAX_ETHPORTS);
6042 		return -EINVAL;
6043 	}
6044 
6045 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6046 		PMD_DRV_LOG(ERR,
6047 			    "Not a PF or trusted VF. No Representor support\n");
6048 		/* Returning an error is not an option.
6049 		 * Applications are not handling this correctly
6050 		 */
6051 		return 0;
6052 	}
6053 
6054 	if (bnxt_init_rep_info(backing_bp))
6055 		return 0;
6056 
6057 	for (i = 0; i < num_rep; i++) {
6058 		struct bnxt_representor representor = {
6059 			.vf_id = eth_da->representor_ports[i],
6060 			.switch_domain_id = backing_bp->switch_domain_id,
6061 			.parent_dev = backing_eth_dev
6062 		};
6063 
6064 		if (representor.vf_id >= max_vf_reps) {
6065 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6066 				    representor.vf_id, max_vf_reps);
6067 			continue;
6068 		}
6069 
6070 		/* representor port net_bdf_port */
6071 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6072 			 pci_dev->device.name, eth_da->representor_ports[i]);
6073 
6074 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6075 		if (kvlist) {
6076 			/*
6077 			 * Handler for "rep_is_pf" devarg.
6078 			 * Invoked as for ex: "-a 000:00:0d.0,
6079 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6080 			 */
6081 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6082 						 bnxt_parse_devarg_rep_is_pf,
6083 						 (void *)&representor);
6084 			if (ret) {
6085 				ret = -EINVAL;
6086 				goto err;
6087 			}
6088 			/*
6089 			 * Handler for "rep_based_pf" devarg.
6090 			 * Invoked as for ex: "-a 000:00:0d.0,
6091 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6092 			 */
6093 			ret = rte_kvargs_process(kvlist,
6094 						 BNXT_DEVARG_REP_BASED_PF,
6095 						 bnxt_parse_devarg_rep_based_pf,
6096 						 (void *)&representor);
6097 			if (ret) {
6098 				ret = -EINVAL;
6099 				goto err;
6100 			}
6101 			/*
6102 			 * Handler for "rep_based_pf" devarg.
6103 			 * Invoked as for ex: "-a 000:00:0d.0,
6104 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6105 			 */
6106 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6107 						 bnxt_parse_devarg_rep_q_r2f,
6108 						 (void *)&representor);
6109 			if (ret) {
6110 				ret = -EINVAL;
6111 				goto err;
6112 			}
6113 			/*
6114 			 * Handler for "rep_based_pf" devarg.
6115 			 * Invoked as for ex: "-a 000:00:0d.0,
6116 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6117 			 */
6118 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6119 						 bnxt_parse_devarg_rep_q_f2r,
6120 						 (void *)&representor);
6121 			if (ret) {
6122 				ret = -EINVAL;
6123 				goto err;
6124 			}
6125 			/*
6126 			 * Handler for "rep_based_pf" devarg.
6127 			 * Invoked as for ex: "-a 000:00:0d.0,
6128 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6129 			 */
6130 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6131 						 bnxt_parse_devarg_rep_fc_r2f,
6132 						 (void *)&representor);
6133 			if (ret) {
6134 				ret = -EINVAL;
6135 				goto err;
6136 			}
6137 			/*
6138 			 * Handler for "rep_based_pf" devarg.
6139 			 * Invoked as for ex: "-a 000:00:0d.0,
6140 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6141 			 */
6142 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6143 						 bnxt_parse_devarg_rep_fc_f2r,
6144 						 (void *)&representor);
6145 			if (ret) {
6146 				ret = -EINVAL;
6147 				goto err;
6148 			}
6149 		}
6150 
6151 		ret = rte_eth_dev_create(&pci_dev->device, name,
6152 					 sizeof(struct bnxt_representor),
6153 					 NULL, NULL,
6154 					 bnxt_representor_init,
6155 					 &representor);
6156 		if (ret) {
6157 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6158 				    "representor %s.", name);
6159 			goto err;
6160 		}
6161 
6162 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6163 		if (!vf_rep_eth_dev) {
6164 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6165 				    " for VF-Rep: %s.", name);
6166 			ret = -ENODEV;
6167 			goto err;
6168 		}
6169 
6170 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6171 			    backing_eth_dev->data->port_id);
6172 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6173 							 vf_rep_eth_dev;
6174 		backing_bp->num_reps++;
6175 
6176 	}
6177 
6178 	rte_kvargs_free(kvlist);
6179 	return 0;
6180 
6181 err:
6182 	/* If num_rep > 1, then rollback already created
6183 	 * ports, since we'll be failing the probe anyway
6184 	 */
6185 	if (num_rep > 1)
6186 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6187 	rte_errno = -ret;
6188 	rte_kvargs_free(kvlist);
6189 
6190 	return ret;
6191 }
6192 
6193 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6194 			  struct rte_pci_device *pci_dev)
6195 {
6196 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6197 	struct rte_eth_dev *backing_eth_dev;
6198 	uint16_t num_rep;
6199 	int ret = 0;
6200 
6201 	if (pci_dev->device.devargs) {
6202 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6203 					    &eth_da);
6204 		if (ret)
6205 			return ret;
6206 	}
6207 
6208 	num_rep = eth_da.nb_representor_ports;
6209 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6210 		    num_rep);
6211 
6212 	/* We could come here after first level of probe is already invoked
6213 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6214 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6215 	 */
6216 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6217 	if (backing_eth_dev == NULL) {
6218 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6219 					 sizeof(struct bnxt),
6220 					 eth_dev_pci_specific_init, pci_dev,
6221 					 bnxt_dev_init, NULL);
6222 
6223 		if (ret || !num_rep)
6224 			return ret;
6225 
6226 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6227 	}
6228 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6229 		    backing_eth_dev->data->port_id);
6230 
6231 	if (!num_rep)
6232 		return ret;
6233 
6234 	/* probe representor ports now */
6235 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6236 				  pci_dev->device.devargs->args);
6237 
6238 	return ret;
6239 }
6240 
6241 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6242 {
6243 	struct rte_eth_dev *eth_dev;
6244 
6245 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6246 	if (!eth_dev)
6247 		return 0; /* Invoked typically only by OVS-DPDK, by the
6248 			   * time it comes here the eth_dev is already
6249 			   * deleted by rte_eth_dev_close(), so returning
6250 			   * +ve value will at least help in proper cleanup
6251 			   */
6252 
6253 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6254 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6255 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6256 			return rte_eth_dev_destroy(eth_dev,
6257 						   bnxt_representor_uninit);
6258 		else
6259 			return rte_eth_dev_destroy(eth_dev,
6260 						   bnxt_dev_uninit);
6261 	} else {
6262 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6263 	}
6264 }
6265 
6266 static struct rte_pci_driver bnxt_rte_pmd = {
6267 	.id_table = bnxt_pci_id_map,
6268 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6269 			RTE_PCI_DRV_INTR_RMV |
6270 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6271 						  * and OVS-DPDK
6272 						  */
6273 	.probe = bnxt_pci_probe,
6274 	.remove = bnxt_pci_remove,
6275 };
6276 
6277 static bool
6278 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6279 {
6280 	if (strcmp(dev->device->driver->name, drv->driver.name))
6281 		return false;
6282 
6283 	return true;
6284 }
6285 
6286 bool is_bnxt_supported(struct rte_eth_dev *dev)
6287 {
6288 	return is_device_supported(dev, &bnxt_rte_pmd);
6289 }
6290 
6291 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6292 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6293 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6294 
6295