xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision 452c1916)
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_ACCUM_STATS	"accum-stats"
91 #define BNXT_DEVARG_FLOW_XSTAT	"flow-xstat"
92 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
93 #define BNXT_DEVARG_REPRESENTOR	"representor"
94 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
95 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
96 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
97 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
98 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
99 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
100 #define BNXT_DEVARG_APP_ID	"app-id"
101 
102 static const char *const bnxt_dev_args[] = {
103 	BNXT_DEVARG_REPRESENTOR,
104 	BNXT_DEVARG_ACCUM_STATS,
105 	BNXT_DEVARG_FLOW_XSTAT,
106 	BNXT_DEVARG_MAX_NUM_KFLOWS,
107 	BNXT_DEVARG_REP_BASED_PF,
108 	BNXT_DEVARG_REP_IS_PF,
109 	BNXT_DEVARG_REP_Q_R2F,
110 	BNXT_DEVARG_REP_Q_F2R,
111 	BNXT_DEVARG_REP_FC_R2F,
112 	BNXT_DEVARG_REP_FC_F2R,
113 	BNXT_DEVARG_APP_ID,
114 	NULL
115 };
116 
117 /*
118  * accum-stats == false to disable flow counter accumulation
119  * accum-stats == true to enable flow counter accumulation
120  */
121 #define	BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)	((accum_stats) > 1)
122 
123 /*
124  * app-id = an non-negative 8-bit number
125  */
126 #define BNXT_DEVARG_APP_ID_INVALID(val)			((val) > 255)
127 
128 /*
129  * flow_xstat == false to disable the feature
130  * flow_xstat == true to enable the feature
131  */
132 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
133 
134 /*
135  * rep_is_pf == false to indicate VF representor
136  * rep_is_pf == true to indicate PF representor
137  */
138 #define	BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)	((rep_is_pf) > 1)
139 
140 /*
141  * rep_based_pf == Physical index of the PF
142  */
143 #define	BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)	((rep_based_pf) > 15)
144 /*
145  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
146  */
147 #define	BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)	((rep_q_r2f) > 3)
148 
149 /*
150  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
151  */
152 #define	BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)	((rep_q_f2r) > 3)
153 
154 /*
155  * rep_fc_r2f == Flow control for the representor to endpoint direction
156  */
157 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)	((rep_fc_r2f) > 1)
158 
159 /*
160  * rep_fc_f2r == Flow control for the endpoint to representor direction
161  */
162 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)	((rep_fc_f2r) > 1)
163 
164 int bnxt_cfa_code_dynfield_offset = -1;
165 
166 /*
167  * max_num_kflows must be >= 32
168  * and must be a power-of-2 supported value
169  * return: 1 -> invalid
170  *         0 -> valid
171  */
172 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
173 {
174 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
175 		return 1;
176 	return 0;
177 }
178 
179 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
180 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
181 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
182 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
183 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
184 static int bnxt_restore_vlan_filters(struct bnxt *bp);
185 static void bnxt_dev_recover(void *arg);
186 static void bnxt_free_error_recovery_info(struct bnxt *bp);
187 static void bnxt_free_rep_info(struct bnxt *bp);
188 
189 int is_bnxt_in_error(struct bnxt *bp)
190 {
191 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
192 		return -EIO;
193 	if (bp->flags & BNXT_FLAG_FW_RESET)
194 		return -EBUSY;
195 
196 	return 0;
197 }
198 
199 /***********************/
200 
201 /*
202  * High level utility functions
203  */
204 
205 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
206 {
207 	unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
208 					     BNXT_RSS_TBL_SIZE_P5);
209 
210 	if (!BNXT_CHIP_P5(bp))
211 		return 1;
212 
213 	return RTE_ALIGN_MUL_CEIL(num_rss_rings,
214 				  BNXT_RSS_ENTRIES_PER_CTX_P5) /
215 				  BNXT_RSS_ENTRIES_PER_CTX_P5;
216 }
217 
218 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
219 {
220 	if (!BNXT_CHIP_P5(bp))
221 		return HW_HASH_INDEX_SIZE;
222 
223 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
224 }
225 
226 static void bnxt_free_parent_info(struct bnxt *bp)
227 {
228 	rte_free(bp->parent);
229 	bp->parent = NULL;
230 }
231 
232 static void bnxt_free_pf_info(struct bnxt *bp)
233 {
234 	rte_free(bp->pf);
235 	bp->pf = NULL;
236 }
237 
238 static void bnxt_free_link_info(struct bnxt *bp)
239 {
240 	rte_free(bp->link_info);
241 	bp->link_info = NULL;
242 }
243 
244 static void bnxt_free_leds_info(struct bnxt *bp)
245 {
246 	if (BNXT_VF(bp))
247 		return;
248 
249 	rte_free(bp->leds);
250 	bp->leds = NULL;
251 }
252 
253 static void bnxt_free_flow_stats_info(struct bnxt *bp)
254 {
255 	rte_free(bp->flow_stat);
256 	bp->flow_stat = NULL;
257 }
258 
259 static void bnxt_free_cos_queues(struct bnxt *bp)
260 {
261 	rte_free(bp->rx_cos_queue);
262 	bp->rx_cos_queue = NULL;
263 	rte_free(bp->tx_cos_queue);
264 	bp->tx_cos_queue = NULL;
265 }
266 
267 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
268 {
269 	bnxt_free_filter_mem(bp);
270 	bnxt_free_vnic_attributes(bp);
271 	bnxt_free_vnic_mem(bp);
272 
273 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
274 	 * If the number of rings change across fw update,
275 	 * we don't have much choice except to warn the user.
276 	 */
277 	if (!reconfig) {
278 		bnxt_free_stats(bp);
279 		bnxt_free_tx_rings(bp);
280 		bnxt_free_rx_rings(bp);
281 	}
282 	bnxt_free_async_cp_ring(bp);
283 	bnxt_free_rxtx_nq_ring(bp);
284 
285 	rte_free(bp->grp_info);
286 	bp->grp_info = NULL;
287 }
288 
289 static int bnxt_alloc_parent_info(struct bnxt *bp)
290 {
291 	bp->parent = rte_zmalloc("bnxt_parent_info",
292 				 sizeof(struct bnxt_parent_info), 0);
293 	if (bp->parent == NULL)
294 		return -ENOMEM;
295 
296 	return 0;
297 }
298 
299 static int bnxt_alloc_pf_info(struct bnxt *bp)
300 {
301 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
302 	if (bp->pf == NULL)
303 		return -ENOMEM;
304 
305 	return 0;
306 }
307 
308 static int bnxt_alloc_link_info(struct bnxt *bp)
309 {
310 	bp->link_info =
311 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
312 	if (bp->link_info == NULL)
313 		return -ENOMEM;
314 
315 	return 0;
316 }
317 
318 static int bnxt_alloc_leds_info(struct bnxt *bp)
319 {
320 	if (BNXT_VF(bp))
321 		return 0;
322 
323 	bp->leds = rte_zmalloc("bnxt_leds",
324 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
325 			       0);
326 	if (bp->leds == NULL)
327 		return -ENOMEM;
328 
329 	return 0;
330 }
331 
332 static int bnxt_alloc_cos_queues(struct bnxt *bp)
333 {
334 	bp->rx_cos_queue =
335 		rte_zmalloc("bnxt_rx_cosq",
336 			    BNXT_COS_QUEUE_COUNT *
337 			    sizeof(struct bnxt_cos_queue_info),
338 			    0);
339 	if (bp->rx_cos_queue == NULL)
340 		return -ENOMEM;
341 
342 	bp->tx_cos_queue =
343 		rte_zmalloc("bnxt_tx_cosq",
344 			    BNXT_COS_QUEUE_COUNT *
345 			    sizeof(struct bnxt_cos_queue_info),
346 			    0);
347 	if (bp->tx_cos_queue == NULL)
348 		return -ENOMEM;
349 
350 	return 0;
351 }
352 
353 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
354 {
355 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
356 				    sizeof(struct bnxt_flow_stat_info), 0);
357 	if (bp->flow_stat == NULL)
358 		return -ENOMEM;
359 
360 	return 0;
361 }
362 
363 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
364 {
365 	int rc;
366 
367 	rc = bnxt_alloc_ring_grps(bp);
368 	if (rc)
369 		goto alloc_mem_err;
370 
371 	rc = bnxt_alloc_async_ring_struct(bp);
372 	if (rc)
373 		goto alloc_mem_err;
374 
375 	rc = bnxt_alloc_vnic_mem(bp);
376 	if (rc)
377 		goto alloc_mem_err;
378 
379 	rc = bnxt_alloc_vnic_attributes(bp);
380 	if (rc)
381 		goto alloc_mem_err;
382 
383 	rc = bnxt_alloc_filter_mem(bp);
384 	if (rc)
385 		goto alloc_mem_err;
386 
387 	rc = bnxt_alloc_async_cp_ring(bp);
388 	if (rc)
389 		goto alloc_mem_err;
390 
391 	rc = bnxt_alloc_rxtx_nq_ring(bp);
392 	if (rc)
393 		goto alloc_mem_err;
394 
395 	if (BNXT_FLOW_XSTATS_EN(bp)) {
396 		rc = bnxt_alloc_flow_stats_info(bp);
397 		if (rc)
398 			goto alloc_mem_err;
399 	}
400 
401 	return 0;
402 
403 alloc_mem_err:
404 	bnxt_free_mem(bp, reconfig);
405 	return rc;
406 }
407 
408 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
409 {
410 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
411 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
412 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
413 	struct bnxt_rx_queue *rxq;
414 	unsigned int j;
415 	int rc;
416 
417 	rc = bnxt_vnic_grp_alloc(bp, vnic);
418 	if (rc)
419 		goto err_out;
420 
421 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
422 		    vnic_id, vnic, vnic->fw_grp_ids);
423 
424 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
425 	if (rc)
426 		goto err_out;
427 
428 	/* Alloc RSS context only if RSS mode is enabled */
429 	if (dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS) {
430 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
431 
432 		/* RSS table size in Thor is 512.
433 		 * Cap max Rx rings to same value
434 		 */
435 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
436 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
437 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
438 			goto err_out;
439 		}
440 
441 		rc = 0;
442 		for (j = 0; j < nr_ctxs; j++) {
443 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
444 			if (rc)
445 				break;
446 		}
447 		if (rc) {
448 			PMD_DRV_LOG(ERR,
449 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
450 				    vnic_id, j, rc);
451 			goto err_out;
452 		}
453 		vnic->num_lb_ctxts = nr_ctxs;
454 	}
455 
456 	/*
457 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
458 	 * setting is not available at this time, it will not be
459 	 * configured correctly in the CFA.
460 	 */
461 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
462 		vnic->vlan_strip = true;
463 	else
464 		vnic->vlan_strip = false;
465 
466 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
467 	if (rc)
468 		goto err_out;
469 
470 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
471 	if (rc)
472 		goto err_out;
473 
474 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
475 		rxq = bp->eth_dev->data->rx_queues[j];
476 
477 		PMD_DRV_LOG(DEBUG,
478 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
479 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
480 
481 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
482 			rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
483 		else
484 			vnic->rx_queue_cnt++;
485 	}
486 
487 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
488 
489 	rc = bnxt_vnic_rss_configure(bp, vnic);
490 	if (rc)
491 		goto err_out;
492 
493 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
494 
495 	rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
496 				    (rx_offloads & RTE_ETH_RX_OFFLOAD_TCP_LRO) ?
497 				    true : false);
498 	if (rc)
499 		goto err_out;
500 
501 	return 0;
502 err_out:
503 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
504 		    vnic_id, rc);
505 	return rc;
506 }
507 
508 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
509 {
510 	int rc = 0;
511 
512 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
513 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
514 	if (rc)
515 		return rc;
516 
517 	PMD_DRV_LOG(DEBUG,
518 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
519 		    " rx_fc_in_tbl.ctx_id = %d\n",
520 		    bp->flow_stat->rx_fc_in_tbl.va,
521 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
522 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
523 
524 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
525 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
526 	if (rc)
527 		return rc;
528 
529 	PMD_DRV_LOG(DEBUG,
530 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
531 		    " rx_fc_out_tbl.ctx_id = %d\n",
532 		    bp->flow_stat->rx_fc_out_tbl.va,
533 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
534 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
535 
536 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
537 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
538 	if (rc)
539 		return rc;
540 
541 	PMD_DRV_LOG(DEBUG,
542 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
543 		    " tx_fc_in_tbl.ctx_id = %d\n",
544 		    bp->flow_stat->tx_fc_in_tbl.va,
545 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
546 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
547 
548 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
549 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
550 	if (rc)
551 		return rc;
552 
553 	PMD_DRV_LOG(DEBUG,
554 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
555 		    " tx_fc_out_tbl.ctx_id = %d\n",
556 		    bp->flow_stat->tx_fc_out_tbl.va,
557 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
558 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
559 
560 	memset(bp->flow_stat->rx_fc_out_tbl.va,
561 	       0,
562 	       bp->flow_stat->rx_fc_out_tbl.size);
563 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
564 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
565 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
566 				       bp->flow_stat->max_fc,
567 				       true);
568 	if (rc)
569 		return rc;
570 
571 	memset(bp->flow_stat->tx_fc_out_tbl.va,
572 	       0,
573 	       bp->flow_stat->tx_fc_out_tbl.size);
574 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
575 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
576 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
577 				       bp->flow_stat->max_fc,
578 				       true);
579 
580 	return rc;
581 }
582 
583 static int bnxt_alloc_ctx_mem_buf(struct bnxt *bp, char *type, size_t size,
584 				  struct bnxt_ctx_mem_buf_info *ctx)
585 {
586 	if (!ctx)
587 		return -EINVAL;
588 
589 	ctx->va = rte_zmalloc_socket(type, size, 0,
590 				     bp->eth_dev->device->numa_node);
591 	if (ctx->va == NULL)
592 		return -ENOMEM;
593 	rte_mem_lock_page(ctx->va);
594 	ctx->size = size;
595 	ctx->dma = rte_mem_virt2iova(ctx->va);
596 	if (ctx->dma == RTE_BAD_IOVA)
597 		return -ENOMEM;
598 
599 	return 0;
600 }
601 
602 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
603 {
604 	struct rte_pci_device *pdev = bp->pdev;
605 	char type[RTE_MEMZONE_NAMESIZE];
606 	uint16_t max_fc;
607 	int rc = 0;
608 
609 	max_fc = bp->flow_stat->max_fc;
610 
611 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
612 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
613 	/* 4 bytes for each counter-id */
614 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
615 				    max_fc * 4,
616 				    &bp->flow_stat->rx_fc_in_tbl);
617 	if (rc)
618 		return rc;
619 
620 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
621 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
622 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
623 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
624 				    max_fc * 16,
625 				    &bp->flow_stat->rx_fc_out_tbl);
626 	if (rc)
627 		return rc;
628 
629 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
630 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
631 	/* 4 bytes for each counter-id */
632 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
633 				    max_fc * 4,
634 				    &bp->flow_stat->tx_fc_in_tbl);
635 	if (rc)
636 		return rc;
637 
638 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
639 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
640 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
641 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
642 				    max_fc * 16,
643 				    &bp->flow_stat->tx_fc_out_tbl);
644 	if (rc)
645 		return rc;
646 
647 	rc = bnxt_register_fc_ctx_mem(bp);
648 
649 	return rc;
650 }
651 
652 static int bnxt_init_ctx_mem(struct bnxt *bp)
653 {
654 	int rc = 0;
655 
656 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
657 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
658 	    !BNXT_FLOW_XSTATS_EN(bp))
659 		return 0;
660 
661 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
662 	if (rc)
663 		return rc;
664 
665 	rc = bnxt_init_fc_ctx_mem(bp);
666 
667 	return rc;
668 }
669 
670 static int bnxt_update_phy_setting(struct bnxt *bp)
671 {
672 	struct rte_eth_link new;
673 	int rc;
674 
675 	rc = bnxt_get_hwrm_link_config(bp, &new);
676 	if (rc) {
677 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
678 		return rc;
679 	}
680 
681 	/*
682 	 * On BCM957508-N2100 adapters, FW will not allow any user other
683 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
684 	 * always returns link up. Force phy update always in that case.
685 	 */
686 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
687 		rc = bnxt_set_hwrm_link_config(bp, true);
688 		if (rc) {
689 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
690 			return rc;
691 		}
692 	}
693 
694 	return rc;
695 }
696 
697 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
698 {
699 	rte_free(bp->prev_rx_ring_stats);
700 	rte_free(bp->prev_tx_ring_stats);
701 
702 	bp->prev_rx_ring_stats = NULL;
703 	bp->prev_tx_ring_stats = NULL;
704 }
705 
706 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
707 {
708 	bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
709 					      sizeof(struct bnxt_ring_stats) *
710 					      bp->rx_cp_nr_rings,
711 					      0);
712 	if (bp->prev_rx_ring_stats == NULL)
713 		return -ENOMEM;
714 
715 	bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
716 					     sizeof(struct bnxt_ring_stats) *
717 					     bp->tx_cp_nr_rings,
718 					     0);
719 	if (bp->prev_tx_ring_stats == NULL)
720 		goto error;
721 
722 	return 0;
723 
724 error:
725 	bnxt_free_prev_ring_stats(bp);
726 	return -ENOMEM;
727 }
728 
729 static int bnxt_start_nic(struct bnxt *bp)
730 {
731 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
732 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
733 	uint32_t intr_vector = 0;
734 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
735 	uint32_t vec = BNXT_MISC_VEC_ID;
736 	unsigned int i, j;
737 	int rc;
738 
739 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU)
740 		bp->flags |= BNXT_FLAG_JUMBO;
741 	else
742 		bp->flags &= ~BNXT_FLAG_JUMBO;
743 
744 	/* THOR does not support ring groups.
745 	 * But we will use the array to save RSS context IDs.
746 	 */
747 	if (BNXT_CHIP_P5(bp))
748 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
749 
750 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
751 	if (rc) {
752 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
753 		goto err_out;
754 	}
755 
756 	rc = bnxt_alloc_hwrm_rings(bp);
757 	if (rc) {
758 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
759 		goto err_out;
760 	}
761 
762 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
763 	if (rc) {
764 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
765 		goto err_out;
766 	}
767 
768 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
769 		goto skip_cosq_cfg;
770 
771 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
772 		if (bp->rx_cos_queue[i].id != 0xff) {
773 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
774 
775 			if (!vnic) {
776 				PMD_DRV_LOG(ERR,
777 					    "Num pools more than FW profile\n");
778 				rc = -EINVAL;
779 				goto err_out;
780 			}
781 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
782 			bp->rx_cosq_cnt++;
783 		}
784 	}
785 
786 skip_cosq_cfg:
787 	rc = bnxt_mq_rx_configure(bp);
788 	if (rc) {
789 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
790 		goto err_out;
791 	}
792 
793 	/* default vnic 0 */
794 	rc = bnxt_setup_one_vnic(bp, 0);
795 	if (rc)
796 		goto err_out;
797 	/* VNIC configuration */
798 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
799 		for (i = 1; i < bp->nr_vnics; i++) {
800 			rc = bnxt_setup_one_vnic(bp, i);
801 			if (rc)
802 				goto err_out;
803 		}
804 	}
805 
806 	for (j = 0; j < bp->tx_nr_rings; j++) {
807 		struct bnxt_tx_queue *txq = bp->tx_queues[j];
808 
809 		if (!txq->tx_deferred_start) {
810 			bp->eth_dev->data->tx_queue_state[j] =
811 				RTE_ETH_QUEUE_STATE_STARTED;
812 			txq->tx_started = true;
813 		}
814 	}
815 
816 	for (j = 0; j < bp->rx_nr_rings; j++) {
817 		struct bnxt_rx_queue *rxq = bp->rx_queues[j];
818 
819 		if (!rxq->rx_deferred_start) {
820 			bp->eth_dev->data->rx_queue_state[j] =
821 				RTE_ETH_QUEUE_STATE_STARTED;
822 			rxq->rx_started = true;
823 		}
824 	}
825 
826 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
827 	if (rc) {
828 		PMD_DRV_LOG(ERR,
829 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
830 		goto err_out;
831 	}
832 
833 	/* check and configure queue intr-vector mapping */
834 	if ((rte_intr_cap_multiple(intr_handle) ||
835 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
836 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
837 		intr_vector = bp->eth_dev->data->nb_rx_queues;
838 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
839 		if (intr_vector > bp->rx_cp_nr_rings) {
840 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
841 					bp->rx_cp_nr_rings);
842 			return -ENOTSUP;
843 		}
844 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
845 		if (rc)
846 			return rc;
847 	}
848 
849 	if (rte_intr_dp_is_en(intr_handle)) {
850 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
851 					bp->eth_dev->data->nb_rx_queues)) {
852 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
853 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
854 			rc = -ENOMEM;
855 			goto err_out;
856 		}
857 		PMD_DRV_LOG(DEBUG, "intr_handle->nb_efd = %d "
858 			    "intr_handle->max_intr = %d\n",
859 			    rte_intr_nb_efd_get(intr_handle),
860 			    rte_intr_max_intr_get(intr_handle));
861 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
862 		     queue_id++) {
863 			rte_intr_vec_list_index_set(intr_handle,
864 					queue_id, vec + BNXT_RX_VEC_START);
865 			if (vec < base + rte_intr_nb_efd_get(intr_handle)
866 			    - 1)
867 				vec++;
868 		}
869 	}
870 
871 	/* enable uio/vfio intr/eventfd mapping */
872 	rc = rte_intr_enable(intr_handle);
873 #ifndef RTE_EXEC_ENV_FREEBSD
874 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
875 	if (rc)
876 		goto err_out;
877 #endif
878 
879 	rc = bnxt_update_phy_setting(bp);
880 	if (rc)
881 		goto err_out;
882 
883 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
884 	if (!bp->mark_table)
885 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
886 
887 	return 0;
888 
889 err_out:
890 	/* Some of the error status returned by FW may not be from errno.h */
891 	if (rc > 0)
892 		rc = -EIO;
893 
894 	return rc;
895 }
896 
897 static int bnxt_shutdown_nic(struct bnxt *bp)
898 {
899 	bnxt_free_all_hwrm_resources(bp);
900 	bnxt_free_all_filters(bp);
901 	bnxt_free_all_vnics(bp);
902 	return 0;
903 }
904 
905 /*
906  * Device configuration and status function
907  */
908 
909 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
910 {
911 	uint32_t link_speed = 0;
912 	uint32_t speed_capa = 0;
913 
914 	if (bp->link_info == NULL)
915 		return 0;
916 
917 	link_speed = bp->link_info->support_speeds;
918 
919 	/* If PAM4 is configured, use PAM4 supported speed */
920 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
921 		link_speed = bp->link_info->support_pam4_speeds;
922 
923 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
924 		speed_capa |= RTE_ETH_LINK_SPEED_100M;
925 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
926 		speed_capa |= RTE_ETH_LINK_SPEED_100M_HD;
927 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
928 		speed_capa |= RTE_ETH_LINK_SPEED_1G;
929 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
930 		speed_capa |= RTE_ETH_LINK_SPEED_2_5G;
931 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
932 		speed_capa |= RTE_ETH_LINK_SPEED_10G;
933 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
934 		speed_capa |= RTE_ETH_LINK_SPEED_20G;
935 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
936 		speed_capa |= RTE_ETH_LINK_SPEED_25G;
937 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
938 		speed_capa |= RTE_ETH_LINK_SPEED_40G;
939 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
940 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
941 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
942 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
943 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
944 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
945 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
946 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
947 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
948 		speed_capa |= RTE_ETH_LINK_SPEED_200G;
949 
950 	if (bp->link_info->auto_mode ==
951 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
952 		speed_capa |= RTE_ETH_LINK_SPEED_FIXED;
953 
954 	return speed_capa;
955 }
956 
957 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
958 				struct rte_eth_dev_info *dev_info)
959 {
960 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
961 	struct bnxt *bp = eth_dev->data->dev_private;
962 	uint16_t max_vnics, i, j, vpool, vrxq;
963 	unsigned int max_rx_rings;
964 	int rc;
965 
966 	rc = is_bnxt_in_error(bp);
967 	if (rc)
968 		return rc;
969 
970 	/* MAC Specifics */
971 	dev_info->max_mac_addrs = bp->max_l2_ctx;
972 	dev_info->max_hash_mac_addrs = 0;
973 
974 	/* PF/VF specifics */
975 	if (BNXT_PF(bp))
976 		dev_info->max_vfs = pdev->max_vfs;
977 
978 	max_rx_rings = bnxt_max_rings(bp);
979 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
980 	dev_info->max_rx_queues = max_rx_rings;
981 	dev_info->max_tx_queues = max_rx_rings;
982 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
983 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
984 	max_vnics = bp->max_vnics;
985 
986 	/* MTU specifics */
987 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
988 	dev_info->max_mtu = BNXT_MAX_MTU;
989 
990 	/* Fast path specifics */
991 	dev_info->min_rx_bufsize = 1;
992 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
993 
994 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
995 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
996 		dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_TIMESTAMP;
997 	if (bp->vnic_cap_flags & BNXT_VNIC_CAP_VLAN_RX_STRIP)
998 		dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
999 	dev_info->tx_queue_offload_capa = RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
1000 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
1001 				    dev_info->tx_queue_offload_capa;
1002 	if (bp->fw_cap & BNXT_FW_CAP_VLAN_TX_INSERT)
1003 		dev_info->tx_offload_capa |= RTE_ETH_TX_OFFLOAD_VLAN_INSERT;
1004 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
1005 
1006 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
1007 	dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
1008 			     RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
1009 
1010 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
1011 		.rx_thresh = {
1012 			.pthresh = 8,
1013 			.hthresh = 8,
1014 			.wthresh = 0,
1015 		},
1016 		.rx_free_thresh = 32,
1017 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
1018 	};
1019 
1020 	dev_info->default_txconf = (struct rte_eth_txconf) {
1021 		.tx_thresh = {
1022 			.pthresh = 32,
1023 			.hthresh = 0,
1024 			.wthresh = 0,
1025 		},
1026 		.tx_free_thresh = 32,
1027 		.tx_rs_thresh = 32,
1028 	};
1029 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
1030 
1031 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1032 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1033 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1034 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1035 
1036 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1037 		dev_info->switch_info.name = eth_dev->device->name;
1038 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1039 		dev_info->switch_info.port_id =
1040 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1041 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1042 	}
1043 
1044 	/*
1045 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1046 	 *       need further investigation.
1047 	 */
1048 
1049 	/* VMDq resources */
1050 	vpool = 64; /* RTE_ETH_64_POOLS */
1051 	vrxq = 128; /* RTE_ETH_VMDQ_DCB_NUM_QUEUES */
1052 	for (i = 0; i < 4; vpool >>= 1, i++) {
1053 		if (max_vnics > vpool) {
1054 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1055 				if (dev_info->max_rx_queues > vrxq) {
1056 					if (vpool > vrxq)
1057 						vpool = vrxq;
1058 					goto found;
1059 				}
1060 			}
1061 			/* Not enough resources to support VMDq */
1062 			break;
1063 		}
1064 	}
1065 	/* Not enough resources to support VMDq */
1066 	vpool = 0;
1067 	vrxq = 0;
1068 found:
1069 	dev_info->max_vmdq_pools = vpool;
1070 	dev_info->vmdq_queue_num = vrxq;
1071 
1072 	dev_info->vmdq_pool_base = 0;
1073 	dev_info->vmdq_queue_base = 0;
1074 
1075 	return 0;
1076 }
1077 
1078 /* Configure the device based on the configuration provided */
1079 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1080 {
1081 	struct bnxt *bp = eth_dev->data->dev_private;
1082 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1083 	int rc;
1084 
1085 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1086 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1087 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1088 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1089 
1090 	rc = is_bnxt_in_error(bp);
1091 	if (rc)
1092 		return rc;
1093 
1094 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1095 		rc = bnxt_hwrm_check_vf_rings(bp);
1096 		if (rc) {
1097 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1098 			return -ENOSPC;
1099 		}
1100 
1101 		/* If a resource has already been allocated - in this case
1102 		 * it is the async completion ring, free it. Reallocate it after
1103 		 * resource reservation. This will ensure the resource counts
1104 		 * are calculated correctly.
1105 		 */
1106 
1107 		pthread_mutex_lock(&bp->def_cp_lock);
1108 
1109 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1110 			bnxt_disable_int(bp);
1111 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1112 		}
1113 
1114 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1115 		if (rc) {
1116 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1117 			pthread_mutex_unlock(&bp->def_cp_lock);
1118 			return -ENOSPC;
1119 		}
1120 
1121 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1122 			rc = bnxt_alloc_async_cp_ring(bp);
1123 			if (rc) {
1124 				pthread_mutex_unlock(&bp->def_cp_lock);
1125 				return rc;
1126 			}
1127 			bnxt_enable_int(bp);
1128 		}
1129 
1130 		pthread_mutex_unlock(&bp->def_cp_lock);
1131 	}
1132 
1133 	/* Inherit new configurations */
1134 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1135 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1136 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1137 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1138 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1139 	    bp->max_stat_ctx)
1140 		goto resource_error;
1141 
1142 	if (BNXT_HAS_RING_GRPS(bp) &&
1143 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1144 		goto resource_error;
1145 
1146 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS) &&
1147 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1148 		goto resource_error;
1149 
1150 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1151 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1152 
1153 	if (eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
1154 		rx_offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
1155 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1156 
1157 	bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1158 
1159 	return 0;
1160 
1161 resource_error:
1162 	PMD_DRV_LOG(ERR,
1163 		    "Insufficient resources to support requested config\n");
1164 	PMD_DRV_LOG(ERR,
1165 		    "Num Queues Requested: Tx %d, Rx %d\n",
1166 		    eth_dev->data->nb_tx_queues,
1167 		    eth_dev->data->nb_rx_queues);
1168 	PMD_DRV_LOG(ERR,
1169 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1170 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1171 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1172 	return -ENOSPC;
1173 }
1174 
1175 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1176 {
1177 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1178 
1179 	if (link->link_status)
1180 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1181 			eth_dev->data->port_id,
1182 			(uint32_t)link->link_speed,
1183 			(link->link_duplex == RTE_ETH_LINK_FULL_DUPLEX) ?
1184 			("full-duplex") : ("half-duplex\n"));
1185 	else
1186 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1187 			eth_dev->data->port_id);
1188 }
1189 
1190 /*
1191  * Determine whether the current configuration requires support for scattered
1192  * receive; return 1 if scattered receive is required and 0 if not.
1193  */
1194 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1195 {
1196 	uint32_t overhead = BNXT_MAX_PKT_LEN - BNXT_MAX_MTU;
1197 	uint16_t buf_size;
1198 	int i;
1199 
1200 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_SCATTER)
1201 		return 1;
1202 
1203 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_TCP_LRO)
1204 		return 1;
1205 
1206 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1207 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1208 
1209 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1210 				      RTE_PKTMBUF_HEADROOM);
1211 		if (eth_dev->data->mtu + overhead > buf_size)
1212 			return 1;
1213 	}
1214 	return 0;
1215 }
1216 
1217 static eth_rx_burst_t
1218 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1219 {
1220 	struct bnxt *bp = eth_dev->data->dev_private;
1221 
1222 	/* Disable vector mode RX for Stingray2 for now */
1223 	if (BNXT_CHIP_SR2(bp)) {
1224 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1225 		return bnxt_recv_pkts;
1226 	}
1227 
1228 #if (defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)) && \
1229 	!defined(RTE_LIBRTE_IEEE1588)
1230 
1231 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1232 	if (eth_dev->data->scattered_rx)
1233 		goto use_scalar_rx;
1234 
1235 	/*
1236 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1237 	 * asynchronous completions and receive completions can be placed in
1238 	 * the same completion ring.
1239 	 */
1240 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1241 		goto use_scalar_rx;
1242 
1243 	/*
1244 	 * Vector mode receive cannot be enabled if any receive offloads outside
1245 	 * a limited subset have been enabled.
1246 	 */
1247 	if (eth_dev->data->dev_conf.rxmode.offloads &
1248 		~(RTE_ETH_RX_OFFLOAD_VLAN_STRIP |
1249 		  RTE_ETH_RX_OFFLOAD_KEEP_CRC |
1250 		  RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
1251 		  RTE_ETH_RX_OFFLOAD_UDP_CKSUM |
1252 		  RTE_ETH_RX_OFFLOAD_TCP_CKSUM |
1253 		  RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1254 		  RTE_ETH_RX_OFFLOAD_OUTER_UDP_CKSUM |
1255 		  RTE_ETH_RX_OFFLOAD_RSS_HASH |
1256 		  RTE_ETH_RX_OFFLOAD_VLAN_FILTER))
1257 		goto use_scalar_rx;
1258 
1259 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1260 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1261 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1262 		PMD_DRV_LOG(INFO,
1263 			    "Using AVX2 vector mode receive for port %d\n",
1264 			    eth_dev->data->port_id);
1265 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1266 		return bnxt_recv_pkts_vec_avx2;
1267 	}
1268  #endif
1269 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1270 		PMD_DRV_LOG(INFO,
1271 			    "Using SSE vector mode receive for port %d\n",
1272 			    eth_dev->data->port_id);
1273 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1274 		return bnxt_recv_pkts_vec;
1275 	}
1276 
1277 use_scalar_rx:
1278 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1279 		    eth_dev->data->port_id);
1280 	PMD_DRV_LOG(INFO,
1281 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1282 		    eth_dev->data->port_id,
1283 		    eth_dev->data->scattered_rx,
1284 		    eth_dev->data->dev_conf.rxmode.offloads);
1285 #endif
1286 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1287 	return bnxt_recv_pkts;
1288 }
1289 
1290 static eth_tx_burst_t
1291 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1292 {
1293 	struct bnxt *bp = eth_dev->data->dev_private;
1294 
1295 	/* Disable vector mode TX for Stingray2 for now */
1296 	if (BNXT_CHIP_SR2(bp))
1297 		return bnxt_xmit_pkts;
1298 
1299 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64) && \
1300 	!defined(RTE_LIBRTE_IEEE1588)
1301 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1302 
1303 	/*
1304 	 * Vector mode transmit can be enabled only if not using scatter rx
1305 	 * or tx offloads.
1306 	 */
1307 	if (eth_dev->data->scattered_rx ||
1308 	    (offloads & ~RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE) ||
1309 	    BNXT_TRUFLOW_EN(bp))
1310 		goto use_scalar_tx;
1311 
1312 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1313 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1314 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1315 		PMD_DRV_LOG(INFO,
1316 			    "Using AVX2 vector mode transmit for port %d\n",
1317 			    eth_dev->data->port_id);
1318 		return bnxt_xmit_pkts_vec_avx2;
1319 	}
1320 #endif
1321 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1322 		PMD_DRV_LOG(INFO,
1323 			    "Using SSE vector mode transmit for port %d\n",
1324 			    eth_dev->data->port_id);
1325 		return bnxt_xmit_pkts_vec;
1326 	}
1327 
1328 use_scalar_tx:
1329 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1330 		    eth_dev->data->port_id);
1331 	PMD_DRV_LOG(INFO,
1332 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1333 		    eth_dev->data->port_id,
1334 		    eth_dev->data->scattered_rx,
1335 		    offloads);
1336 #endif
1337 	return bnxt_xmit_pkts;
1338 }
1339 
1340 static int bnxt_handle_if_change_status(struct bnxt *bp)
1341 {
1342 	int rc;
1343 
1344 	/* Since fw has undergone a reset and lost all contexts,
1345 	 * set fatal flag to not issue hwrm during cleanup
1346 	 */
1347 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1348 	bnxt_uninit_resources(bp, true);
1349 
1350 	/* clear fatal flag so that re-init happens */
1351 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1352 	rc = bnxt_init_resources(bp, true);
1353 
1354 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1355 
1356 	return rc;
1357 }
1358 
1359 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1360 {
1361 	struct bnxt *bp = eth_dev->data->dev_private;
1362 	int rc = 0;
1363 
1364 	if (!BNXT_SINGLE_PF(bp))
1365 		return -ENOTSUP;
1366 
1367 	if (!bp->link_info->link_up)
1368 		rc = bnxt_set_hwrm_link_config(bp, true);
1369 	if (!rc)
1370 		eth_dev->data->dev_link.link_status = 1;
1371 
1372 	bnxt_print_link_info(eth_dev);
1373 	return rc;
1374 }
1375 
1376 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1377 {
1378 	struct bnxt *bp = eth_dev->data->dev_private;
1379 
1380 	if (!BNXT_SINGLE_PF(bp))
1381 		return -ENOTSUP;
1382 
1383 	eth_dev->data->dev_link.link_status = 0;
1384 	bnxt_set_hwrm_link_config(bp, false);
1385 	bp->link_info->link_up = 0;
1386 
1387 	return 0;
1388 }
1389 
1390 static void bnxt_free_switch_domain(struct bnxt *bp)
1391 {
1392 	int rc = 0;
1393 
1394 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1395 		return;
1396 
1397 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1398 	if (rc)
1399 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1400 			    bp->switch_domain_id, rc);
1401 }
1402 
1403 static void bnxt_ptp_get_current_time(void *arg)
1404 {
1405 	struct bnxt *bp = arg;
1406 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1407 	int rc;
1408 
1409 	rc = is_bnxt_in_error(bp);
1410 	if (rc)
1411 		return;
1412 
1413 	if (!ptp)
1414 		return;
1415 
1416 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1417 				&ptp->current_time);
1418 
1419 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1420 	if (rc != 0) {
1421 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1422 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1423 	}
1424 }
1425 
1426 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1427 {
1428 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1429 	int rc;
1430 
1431 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1432 		return 0;
1433 
1434 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1435 				&ptp->current_time);
1436 
1437 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1438 	return rc;
1439 }
1440 
1441 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1442 {
1443 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1444 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1445 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1446 	}
1447 }
1448 
1449 static void bnxt_ptp_stop(struct bnxt *bp)
1450 {
1451 	bnxt_cancel_ptp_alarm(bp);
1452 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1453 }
1454 
1455 static int bnxt_ptp_start(struct bnxt *bp)
1456 {
1457 	int rc;
1458 
1459 	rc = bnxt_schedule_ptp_alarm(bp);
1460 	if (rc != 0) {
1461 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1462 	} else {
1463 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1464 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1465 	}
1466 
1467 	return rc;
1468 }
1469 
1470 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1471 {
1472 	struct bnxt *bp = eth_dev->data->dev_private;
1473 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1474 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1475 	struct rte_eth_link link;
1476 	int ret;
1477 
1478 	eth_dev->data->dev_started = 0;
1479 	eth_dev->data->scattered_rx = 0;
1480 
1481 	/* Prevent crashes when queues are still in use */
1482 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1483 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1484 
1485 	bnxt_disable_int(bp);
1486 
1487 	/* disable uio/vfio intr/eventfd mapping */
1488 	rte_intr_disable(intr_handle);
1489 
1490 	/* Stop the child representors for this device */
1491 	ret = bnxt_rep_stop_all(bp);
1492 	if (ret != 0)
1493 		return ret;
1494 
1495 	/* delete the bnxt ULP port details */
1496 	bnxt_ulp_port_deinit(bp);
1497 
1498 	bnxt_cancel_fw_health_check(bp);
1499 
1500 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1501 		bnxt_cancel_ptp_alarm(bp);
1502 
1503 	/* Do not bring link down during reset recovery */
1504 	if (!is_bnxt_in_error(bp)) {
1505 		bnxt_dev_set_link_down_op(eth_dev);
1506 		/* Wait for link to be reset */
1507 		if (BNXT_SINGLE_PF(bp))
1508 			rte_delay_ms(500);
1509 		/* clear the recorded link status */
1510 		memset(&link, 0, sizeof(link));
1511 		rte_eth_linkstatus_set(eth_dev, &link);
1512 	}
1513 
1514 	/* Clean queue intr-vector mapping */
1515 	rte_intr_efd_disable(intr_handle);
1516 	rte_intr_vec_list_free(intr_handle);
1517 
1518 	bnxt_hwrm_port_clr_stats(bp);
1519 	bnxt_free_tx_mbufs(bp);
1520 	bnxt_free_rx_mbufs(bp);
1521 	/* Process any remaining notifications in default completion queue */
1522 	bnxt_int_handler(eth_dev);
1523 	bnxt_shutdown_nic(bp);
1524 	bnxt_hwrm_if_change(bp, false);
1525 
1526 	bnxt_free_prev_ring_stats(bp);
1527 	rte_free(bp->mark_table);
1528 	bp->mark_table = NULL;
1529 
1530 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1531 	bp->rx_cosq_cnt = 0;
1532 	/* All filters are deleted on a port stop. */
1533 	if (BNXT_FLOW_XSTATS_EN(bp))
1534 		bp->flow_stat->flow_count = 0;
1535 
1536 	return 0;
1537 }
1538 
1539 /* Unload the driver, release resources */
1540 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1541 {
1542 	struct bnxt *bp = eth_dev->data->dev_private;
1543 
1544 	pthread_mutex_lock(&bp->err_recovery_lock);
1545 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1546 		PMD_DRV_LOG(ERR,
1547 			    "Adapter recovering from error..Please retry\n");
1548 		pthread_mutex_unlock(&bp->err_recovery_lock);
1549 		return -EAGAIN;
1550 	}
1551 	pthread_mutex_unlock(&bp->err_recovery_lock);
1552 
1553 	return bnxt_dev_stop(eth_dev);
1554 }
1555 
1556 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1557 {
1558 	struct bnxt *bp = eth_dev->data->dev_private;
1559 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1560 	int vlan_mask = 0;
1561 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1562 
1563 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1564 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1565 		return -EINVAL;
1566 	}
1567 
1568 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1569 		PMD_DRV_LOG(ERR,
1570 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1571 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1572 
1573 	do {
1574 		rc = bnxt_hwrm_if_change(bp, true);
1575 		if (rc == 0 || rc != -EAGAIN)
1576 			break;
1577 
1578 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1579 	} while (retry_cnt--);
1580 
1581 	if (rc)
1582 		return rc;
1583 
1584 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1585 		rc = bnxt_handle_if_change_status(bp);
1586 		if (rc)
1587 			return rc;
1588 	}
1589 
1590 	bnxt_enable_int(bp);
1591 
1592 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1593 
1594 	rc = bnxt_start_nic(bp);
1595 	if (rc)
1596 		goto error;
1597 
1598 	rc = bnxt_alloc_prev_ring_stats(bp);
1599 	if (rc)
1600 		goto error;
1601 
1602 	eth_dev->data->dev_started = 1;
1603 
1604 	bnxt_link_update_op(eth_dev, 1);
1605 
1606 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
1607 		vlan_mask |= RTE_ETH_VLAN_FILTER_MASK;
1608 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
1609 		vlan_mask |= RTE_ETH_VLAN_STRIP_MASK;
1610 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1611 	if (rc)
1612 		goto error;
1613 
1614 	/* Initialize bnxt ULP port details */
1615 	rc = bnxt_ulp_port_init(bp);
1616 	if (rc)
1617 		goto error;
1618 
1619 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1620 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1621 
1622 	bnxt_schedule_fw_health_check(bp);
1623 
1624 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1625 		bnxt_schedule_ptp_alarm(bp);
1626 
1627 	return 0;
1628 
1629 error:
1630 	bnxt_dev_stop(eth_dev);
1631 	return rc;
1632 }
1633 
1634 static void
1635 bnxt_uninit_locks(struct bnxt *bp)
1636 {
1637 	pthread_mutex_destroy(&bp->flow_lock);
1638 	pthread_mutex_destroy(&bp->def_cp_lock);
1639 	pthread_mutex_destroy(&bp->health_check_lock);
1640 	pthread_mutex_destroy(&bp->err_recovery_lock);
1641 	if (bp->rep_info) {
1642 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1643 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1644 	}
1645 }
1646 
1647 static void bnxt_drv_uninit(struct bnxt *bp)
1648 {
1649 	bnxt_free_leds_info(bp);
1650 	bnxt_free_cos_queues(bp);
1651 	bnxt_free_link_info(bp);
1652 	bnxt_free_parent_info(bp);
1653 	bnxt_uninit_locks(bp);
1654 
1655 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1656 	bp->tx_mem_zone = NULL;
1657 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1658 	bp->rx_mem_zone = NULL;
1659 
1660 	bnxt_free_vf_info(bp);
1661 	bnxt_free_pf_info(bp);
1662 
1663 	rte_free(bp->grp_info);
1664 	bp->grp_info = NULL;
1665 }
1666 
1667 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1668 {
1669 	struct bnxt *bp = eth_dev->data->dev_private;
1670 	int ret = 0;
1671 
1672 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1673 		return 0;
1674 
1675 	pthread_mutex_lock(&bp->err_recovery_lock);
1676 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1677 		PMD_DRV_LOG(ERR,
1678 			    "Adapter recovering from error...Please retry\n");
1679 		pthread_mutex_unlock(&bp->err_recovery_lock);
1680 		return -EAGAIN;
1681 	}
1682 	pthread_mutex_unlock(&bp->err_recovery_lock);
1683 
1684 	/* cancel the recovery handler before remove dev */
1685 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1686 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1687 	bnxt_cancel_fc_thread(bp);
1688 
1689 	if (eth_dev->data->dev_started)
1690 		ret = bnxt_dev_stop(eth_dev);
1691 
1692 	bnxt_uninit_resources(bp, false);
1693 
1694 	bnxt_drv_uninit(bp);
1695 
1696 	return ret;
1697 }
1698 
1699 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1700 				    uint32_t index)
1701 {
1702 	struct bnxt *bp = eth_dev->data->dev_private;
1703 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1704 	struct bnxt_vnic_info *vnic;
1705 	struct bnxt_filter_info *filter, *temp_filter;
1706 	uint32_t i;
1707 
1708 	if (is_bnxt_in_error(bp))
1709 		return;
1710 
1711 	/*
1712 	 * Loop through all VNICs from the specified filter flow pools to
1713 	 * remove the corresponding MAC addr filter
1714 	 */
1715 	for (i = 0; i < bp->nr_vnics; i++) {
1716 		if (!(pool_mask & (1ULL << i)))
1717 			continue;
1718 
1719 		vnic = &bp->vnic_info[i];
1720 		filter = STAILQ_FIRST(&vnic->filter);
1721 		while (filter) {
1722 			temp_filter = STAILQ_NEXT(filter, next);
1723 			if (filter->mac_index == index) {
1724 				STAILQ_REMOVE(&vnic->filter, filter,
1725 						bnxt_filter_info, next);
1726 				bnxt_hwrm_clear_l2_filter(bp, filter);
1727 				bnxt_free_filter(bp, filter);
1728 			}
1729 			filter = temp_filter;
1730 		}
1731 	}
1732 }
1733 
1734 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1735 			       struct rte_ether_addr *mac_addr, uint32_t index,
1736 			       uint32_t pool)
1737 {
1738 	struct bnxt_filter_info *filter;
1739 	int rc = 0;
1740 
1741 	/* Attach requested MAC address to the new l2_filter */
1742 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1743 		if (filter->mac_index == index) {
1744 			PMD_DRV_LOG(DEBUG,
1745 				    "MAC addr already existed for pool %d\n",
1746 				    pool);
1747 			return 0;
1748 		}
1749 	}
1750 
1751 	filter = bnxt_alloc_filter(bp);
1752 	if (!filter) {
1753 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1754 		return -ENODEV;
1755 	}
1756 
1757 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1758 	 * if the MAC that's been programmed now is a different one, then,
1759 	 * copy that addr to filter->l2_addr
1760 	 */
1761 	if (mac_addr)
1762 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1763 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1764 
1765 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1766 	if (!rc) {
1767 		filter->mac_index = index;
1768 		if (filter->mac_index == 0)
1769 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1770 		else
1771 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1772 	} else {
1773 		bnxt_free_filter(bp, filter);
1774 	}
1775 
1776 	return rc;
1777 }
1778 
1779 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1780 				struct rte_ether_addr *mac_addr,
1781 				uint32_t index, uint32_t pool)
1782 {
1783 	struct bnxt *bp = eth_dev->data->dev_private;
1784 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1785 	int rc = 0;
1786 
1787 	rc = is_bnxt_in_error(bp);
1788 	if (rc)
1789 		return rc;
1790 
1791 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1792 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1793 		return -ENOTSUP;
1794 	}
1795 
1796 	if (!vnic) {
1797 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1798 		return -EINVAL;
1799 	}
1800 
1801 	/* Filter settings will get applied when port is started */
1802 	if (!eth_dev->data->dev_started)
1803 		return 0;
1804 
1805 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1806 
1807 	return rc;
1808 }
1809 
1810 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1811 {
1812 	int rc = 0;
1813 	struct bnxt *bp = eth_dev->data->dev_private;
1814 	struct rte_eth_link new;
1815 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1816 			BNXT_MIN_LINK_WAIT_CNT;
1817 
1818 	rc = is_bnxt_in_error(bp);
1819 	if (rc)
1820 		return rc;
1821 
1822 	memset(&new, 0, sizeof(new));
1823 
1824 	if (bp->link_info == NULL)
1825 		goto out;
1826 
1827 	do {
1828 		/* Retrieve link info from hardware */
1829 		rc = bnxt_get_hwrm_link_config(bp, &new);
1830 		if (rc) {
1831 			new.link_speed = RTE_ETH_LINK_SPEED_100M;
1832 			new.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
1833 			PMD_DRV_LOG(ERR,
1834 				"Failed to retrieve link rc = 0x%x!\n", rc);
1835 			goto out;
1836 		}
1837 
1838 		if (!wait_to_complete || new.link_status)
1839 			break;
1840 
1841 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1842 	} while (cnt--);
1843 
1844 	/* Only single function PF can bring phy down.
1845 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1846 	 */
1847 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1848 		memset(&new, 0, sizeof(new));
1849 
1850 out:
1851 	/* Timed out or success */
1852 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1853 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1854 		rte_eth_linkstatus_set(eth_dev, &new);
1855 		bnxt_print_link_info(eth_dev);
1856 	}
1857 
1858 	return rc;
1859 }
1860 
1861 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1862 {
1863 	struct bnxt *bp = eth_dev->data->dev_private;
1864 	struct bnxt_vnic_info *vnic;
1865 	uint32_t old_flags;
1866 	int rc;
1867 
1868 	rc = is_bnxt_in_error(bp);
1869 	if (rc)
1870 		return rc;
1871 
1872 	/* Filter settings will get applied when port is started */
1873 	if (!eth_dev->data->dev_started)
1874 		return 0;
1875 
1876 	if (bp->vnic_info == NULL)
1877 		return 0;
1878 
1879 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1880 
1881 	old_flags = vnic->flags;
1882 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1883 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1884 	if (rc != 0)
1885 		vnic->flags = old_flags;
1886 
1887 	return rc;
1888 }
1889 
1890 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1891 {
1892 	struct bnxt *bp = eth_dev->data->dev_private;
1893 	struct bnxt_vnic_info *vnic;
1894 	uint32_t old_flags;
1895 	int rc;
1896 
1897 	rc = is_bnxt_in_error(bp);
1898 	if (rc)
1899 		return rc;
1900 
1901 	/* Filter settings will get applied when port is started */
1902 	if (!eth_dev->data->dev_started)
1903 		return 0;
1904 
1905 	if (bp->vnic_info == NULL)
1906 		return 0;
1907 
1908 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1909 
1910 	old_flags = vnic->flags;
1911 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1912 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1913 	if (rc != 0)
1914 		vnic->flags = old_flags;
1915 
1916 	return rc;
1917 }
1918 
1919 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1920 {
1921 	struct bnxt *bp = eth_dev->data->dev_private;
1922 	struct bnxt_vnic_info *vnic;
1923 	uint32_t old_flags;
1924 	int rc;
1925 
1926 	rc = is_bnxt_in_error(bp);
1927 	if (rc)
1928 		return rc;
1929 
1930 	/* Filter settings will get applied when port is started */
1931 	if (!eth_dev->data->dev_started)
1932 		return 0;
1933 
1934 	if (bp->vnic_info == NULL)
1935 		return 0;
1936 
1937 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1938 
1939 	old_flags = vnic->flags;
1940 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1941 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1942 	if (rc != 0)
1943 		vnic->flags = old_flags;
1944 
1945 	return rc;
1946 }
1947 
1948 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1949 {
1950 	struct bnxt *bp = eth_dev->data->dev_private;
1951 	struct bnxt_vnic_info *vnic;
1952 	uint32_t old_flags;
1953 	int rc;
1954 
1955 	rc = is_bnxt_in_error(bp);
1956 	if (rc)
1957 		return rc;
1958 
1959 	/* Filter settings will get applied when port is started */
1960 	if (!eth_dev->data->dev_started)
1961 		return 0;
1962 
1963 	if (bp->vnic_info == NULL)
1964 		return 0;
1965 
1966 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1967 
1968 	old_flags = vnic->flags;
1969 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1970 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1971 	if (rc != 0)
1972 		vnic->flags = old_flags;
1973 
1974 	return rc;
1975 }
1976 
1977 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1978 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1979 {
1980 	if (qid >= bp->rx_nr_rings)
1981 		return NULL;
1982 
1983 	return bp->eth_dev->data->rx_queues[qid];
1984 }
1985 
1986 /* Return rxq corresponding to a given rss table ring/group ID. */
1987 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1988 {
1989 	struct bnxt_rx_queue *rxq;
1990 	unsigned int i;
1991 
1992 	if (!BNXT_HAS_RING_GRPS(bp)) {
1993 		for (i = 0; i < bp->rx_nr_rings; i++) {
1994 			rxq = bp->eth_dev->data->rx_queues[i];
1995 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1996 				return rxq->index;
1997 		}
1998 	} else {
1999 		for (i = 0; i < bp->rx_nr_rings; i++) {
2000 			if (bp->grp_info[i].fw_grp_id == fwr)
2001 				return i;
2002 		}
2003 	}
2004 
2005 	return INVALID_HW_RING_ID;
2006 }
2007 
2008 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
2009 			    struct rte_eth_rss_reta_entry64 *reta_conf,
2010 			    uint16_t reta_size)
2011 {
2012 	struct bnxt *bp = eth_dev->data->dev_private;
2013 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2014 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2015 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2016 	uint16_t idx, sft;
2017 	int i, rc;
2018 
2019 	rc = is_bnxt_in_error(bp);
2020 	if (rc)
2021 		return rc;
2022 
2023 	if (!vnic->rss_table)
2024 		return -EINVAL;
2025 
2026 	if (!(dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG))
2027 		return -EINVAL;
2028 
2029 	if (reta_size != tbl_size) {
2030 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2031 			"(%d) must equal the size supported by the hardware "
2032 			"(%d)\n", reta_size, tbl_size);
2033 		return -EINVAL;
2034 	}
2035 
2036 	for (i = 0; i < reta_size; i++) {
2037 		struct bnxt_rx_queue *rxq;
2038 
2039 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2040 		sft = i % RTE_ETH_RETA_GROUP_SIZE;
2041 
2042 		if (!(reta_conf[idx].mask & (1ULL << sft)))
2043 			continue;
2044 
2045 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2046 		if (!rxq) {
2047 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
2048 			return -EINVAL;
2049 		}
2050 
2051 		if (BNXT_CHIP_P5(bp)) {
2052 			vnic->rss_table[i * 2] =
2053 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
2054 			vnic->rss_table[i * 2 + 1] =
2055 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
2056 		} else {
2057 			vnic->rss_table[i] =
2058 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2059 		}
2060 	}
2061 
2062 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2063 	return rc;
2064 }
2065 
2066 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2067 			      struct rte_eth_rss_reta_entry64 *reta_conf,
2068 			      uint16_t reta_size)
2069 {
2070 	struct bnxt *bp = eth_dev->data->dev_private;
2071 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2072 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2073 	uint16_t idx, sft, i;
2074 	int rc;
2075 
2076 	rc = is_bnxt_in_error(bp);
2077 	if (rc)
2078 		return rc;
2079 
2080 	if (!vnic)
2081 		return -EINVAL;
2082 	if (!vnic->rss_table)
2083 		return -EINVAL;
2084 
2085 	if (reta_size != tbl_size) {
2086 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2087 			"(%d) must equal the size supported by the hardware "
2088 			"(%d)\n", reta_size, tbl_size);
2089 		return -EINVAL;
2090 	}
2091 
2092 	for (idx = 0, i = 0; i < reta_size; i++) {
2093 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2094 		sft = i % RTE_ETH_RETA_GROUP_SIZE;
2095 
2096 		if (reta_conf[idx].mask & (1ULL << sft)) {
2097 			uint16_t qid;
2098 
2099 			if (BNXT_CHIP_P5(bp))
2100 				qid = bnxt_rss_to_qid(bp,
2101 						      vnic->rss_table[i * 2]);
2102 			else
2103 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2104 
2105 			if (qid == INVALID_HW_RING_ID) {
2106 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2107 				return -EINVAL;
2108 			}
2109 			reta_conf[idx].reta[sft] = qid;
2110 		}
2111 	}
2112 
2113 	return 0;
2114 }
2115 
2116 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2117 				   struct rte_eth_rss_conf *rss_conf)
2118 {
2119 	struct bnxt *bp = eth_dev->data->dev_private;
2120 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2121 	struct bnxt_vnic_info *vnic;
2122 	int rc;
2123 
2124 	rc = is_bnxt_in_error(bp);
2125 	if (rc)
2126 		return rc;
2127 
2128 	/*
2129 	 * If RSS enablement were different than dev_configure,
2130 	 * then return -EINVAL
2131 	 */
2132 	if (dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG) {
2133 		if (!rss_conf->rss_hf)
2134 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2135 	} else {
2136 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2137 			return -EINVAL;
2138 	}
2139 
2140 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
2141 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2142 	       rss_conf,
2143 	       sizeof(*rss_conf));
2144 
2145 	/* Update the default RSS VNIC(s) */
2146 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2147 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2148 	vnic->hash_mode =
2149 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2150 					    RTE_ETH_RSS_LEVEL(rss_conf->rss_hf));
2151 
2152 	/*
2153 	 * If hashkey is not specified, use the previously configured
2154 	 * hashkey
2155 	 */
2156 	if (!rss_conf->rss_key)
2157 		goto rss_config;
2158 
2159 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2160 		PMD_DRV_LOG(ERR,
2161 			    "Invalid hashkey length, should be %d bytes\n",
2162 			    HW_HASH_KEY_SIZE);
2163 		return -EINVAL;
2164 	}
2165 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2166 
2167 rss_config:
2168 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2169 	return rc;
2170 }
2171 
2172 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2173 				     struct rte_eth_rss_conf *rss_conf)
2174 {
2175 	struct bnxt *bp = eth_dev->data->dev_private;
2176 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2177 	int len, rc;
2178 	uint32_t hash_types;
2179 
2180 	rc = is_bnxt_in_error(bp);
2181 	if (rc)
2182 		return rc;
2183 
2184 	/* RSS configuration is the same for all VNICs */
2185 	if (vnic && vnic->rss_hash_key) {
2186 		if (rss_conf->rss_key) {
2187 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2188 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2189 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2190 		}
2191 
2192 		hash_types = vnic->hash_type;
2193 		rss_conf->rss_hf = 0;
2194 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2195 			rss_conf->rss_hf |= RTE_ETH_RSS_IPV4;
2196 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2197 		}
2198 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2199 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_TCP;
2200 			hash_types &=
2201 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2202 		}
2203 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2204 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_UDP;
2205 			hash_types &=
2206 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2207 		}
2208 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2209 			rss_conf->rss_hf |= RTE_ETH_RSS_IPV6;
2210 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2211 		}
2212 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2213 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_TCP;
2214 			hash_types &=
2215 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2216 		}
2217 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2218 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_UDP;
2219 			hash_types &=
2220 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2221 		}
2222 
2223 		rss_conf->rss_hf |=
2224 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2225 
2226 		if (hash_types) {
2227 			PMD_DRV_LOG(ERR,
2228 				"Unknown RSS config from firmware (%08x), RSS disabled",
2229 				vnic->hash_type);
2230 			return -ENOTSUP;
2231 		}
2232 	} else {
2233 		rss_conf->rss_hf = 0;
2234 	}
2235 	return 0;
2236 }
2237 
2238 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2239 			       struct rte_eth_fc_conf *fc_conf)
2240 {
2241 	struct bnxt *bp = dev->data->dev_private;
2242 	struct rte_eth_link link_info;
2243 	int rc;
2244 
2245 	rc = is_bnxt_in_error(bp);
2246 	if (rc)
2247 		return rc;
2248 
2249 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2250 	if (rc)
2251 		return rc;
2252 
2253 	memset(fc_conf, 0, sizeof(*fc_conf));
2254 	if (bp->link_info->auto_pause)
2255 		fc_conf->autoneg = 1;
2256 	switch (bp->link_info->pause) {
2257 	case 0:
2258 		fc_conf->mode = RTE_ETH_FC_NONE;
2259 		break;
2260 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2261 		fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
2262 		break;
2263 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2264 		fc_conf->mode = RTE_ETH_FC_RX_PAUSE;
2265 		break;
2266 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2267 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2268 		fc_conf->mode = RTE_ETH_FC_FULL;
2269 		break;
2270 	}
2271 	return 0;
2272 }
2273 
2274 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2275 			       struct rte_eth_fc_conf *fc_conf)
2276 {
2277 	struct bnxt *bp = dev->data->dev_private;
2278 	int rc;
2279 
2280 	rc = is_bnxt_in_error(bp);
2281 	if (rc)
2282 		return rc;
2283 
2284 	if (!BNXT_SINGLE_PF(bp)) {
2285 		PMD_DRV_LOG(ERR,
2286 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2287 		return -ENOTSUP;
2288 	}
2289 
2290 	switch (fc_conf->mode) {
2291 	case RTE_ETH_FC_NONE:
2292 		bp->link_info->auto_pause = 0;
2293 		bp->link_info->force_pause = 0;
2294 		break;
2295 	case RTE_ETH_FC_RX_PAUSE:
2296 		if (fc_conf->autoneg) {
2297 			bp->link_info->auto_pause =
2298 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2299 			bp->link_info->force_pause = 0;
2300 		} else {
2301 			bp->link_info->auto_pause = 0;
2302 			bp->link_info->force_pause =
2303 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2304 		}
2305 		break;
2306 	case RTE_ETH_FC_TX_PAUSE:
2307 		if (fc_conf->autoneg) {
2308 			bp->link_info->auto_pause =
2309 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
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 		}
2316 		break;
2317 	case RTE_ETH_FC_FULL:
2318 		if (fc_conf->autoneg) {
2319 			bp->link_info->auto_pause =
2320 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2321 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2322 			bp->link_info->force_pause = 0;
2323 		} else {
2324 			bp->link_info->auto_pause = 0;
2325 			bp->link_info->force_pause =
2326 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2327 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2328 		}
2329 		break;
2330 	}
2331 	return bnxt_set_hwrm_link_config(bp, true);
2332 }
2333 
2334 /* Add UDP tunneling port */
2335 static int
2336 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2337 			 struct rte_eth_udp_tunnel *udp_tunnel)
2338 {
2339 	struct bnxt *bp = eth_dev->data->dev_private;
2340 	uint16_t tunnel_type = 0;
2341 	int rc = 0;
2342 
2343 	rc = is_bnxt_in_error(bp);
2344 	if (rc)
2345 		return rc;
2346 
2347 	switch (udp_tunnel->prot_type) {
2348 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
2349 		if (bp->vxlan_port_cnt) {
2350 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2351 				udp_tunnel->udp_port);
2352 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2353 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2354 				return -ENOSPC;
2355 			}
2356 			bp->vxlan_port_cnt++;
2357 			return 0;
2358 		}
2359 		tunnel_type =
2360 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2361 		break;
2362 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2363 		if (bp->geneve_port_cnt) {
2364 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2365 				udp_tunnel->udp_port);
2366 			if (bp->geneve_port != udp_tunnel->udp_port) {
2367 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2368 				return -ENOSPC;
2369 			}
2370 			bp->geneve_port_cnt++;
2371 			return 0;
2372 		}
2373 		tunnel_type =
2374 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2375 		break;
2376 	default:
2377 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2378 		return -ENOTSUP;
2379 	}
2380 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2381 					     tunnel_type);
2382 
2383 	if (rc != 0)
2384 		return rc;
2385 
2386 	if (tunnel_type ==
2387 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN)
2388 		bp->vxlan_port_cnt++;
2389 
2390 	if (tunnel_type ==
2391 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE)
2392 		bp->geneve_port_cnt++;
2393 
2394 	return rc;
2395 }
2396 
2397 static int
2398 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2399 			 struct rte_eth_udp_tunnel *udp_tunnel)
2400 {
2401 	struct bnxt *bp = eth_dev->data->dev_private;
2402 	uint16_t tunnel_type = 0;
2403 	uint16_t port = 0;
2404 	int rc = 0;
2405 
2406 	rc = is_bnxt_in_error(bp);
2407 	if (rc)
2408 		return rc;
2409 
2410 	switch (udp_tunnel->prot_type) {
2411 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
2412 		if (!bp->vxlan_port_cnt) {
2413 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2414 			return -EINVAL;
2415 		}
2416 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2417 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2418 				udp_tunnel->udp_port, bp->vxlan_port);
2419 			return -EINVAL;
2420 		}
2421 		if (--bp->vxlan_port_cnt)
2422 			return 0;
2423 
2424 		tunnel_type =
2425 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2426 		port = bp->vxlan_fw_dst_port_id;
2427 		break;
2428 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2429 		if (!bp->geneve_port_cnt) {
2430 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2431 			return -EINVAL;
2432 		}
2433 		if (bp->geneve_port != udp_tunnel->udp_port) {
2434 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2435 				udp_tunnel->udp_port, bp->geneve_port);
2436 			return -EINVAL;
2437 		}
2438 		if (--bp->geneve_port_cnt)
2439 			return 0;
2440 
2441 		tunnel_type =
2442 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2443 		port = bp->geneve_fw_dst_port_id;
2444 		break;
2445 	default:
2446 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2447 		return -ENOTSUP;
2448 	}
2449 
2450 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2451 	return rc;
2452 }
2453 
2454 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2455 {
2456 	struct bnxt_filter_info *filter;
2457 	struct bnxt_vnic_info *vnic;
2458 	int rc = 0;
2459 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2460 
2461 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2462 	filter = STAILQ_FIRST(&vnic->filter);
2463 	while (filter) {
2464 		/* Search for this matching MAC+VLAN filter */
2465 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2466 			/* Delete the filter */
2467 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2468 			if (rc)
2469 				return rc;
2470 			STAILQ_REMOVE(&vnic->filter, filter,
2471 				      bnxt_filter_info, next);
2472 			bnxt_free_filter(bp, filter);
2473 			PMD_DRV_LOG(INFO,
2474 				    "Deleted vlan filter for %d\n",
2475 				    vlan_id);
2476 			return 0;
2477 		}
2478 		filter = STAILQ_NEXT(filter, next);
2479 	}
2480 	return -ENOENT;
2481 }
2482 
2483 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2484 {
2485 	struct bnxt_filter_info *filter;
2486 	struct bnxt_vnic_info *vnic;
2487 	int rc = 0;
2488 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2489 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2490 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2491 
2492 	/* Implementation notes on the use of VNIC in this command:
2493 	 *
2494 	 * By default, these filters belong to default vnic for the function.
2495 	 * Once these filters are set up, only destination VNIC can be modified.
2496 	 * If the destination VNIC is not specified in this command,
2497 	 * then the HWRM shall only create an l2 context id.
2498 	 */
2499 
2500 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2501 	filter = STAILQ_FIRST(&vnic->filter);
2502 	/* Check if the VLAN has already been added */
2503 	while (filter) {
2504 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2505 			return -EEXIST;
2506 
2507 		filter = STAILQ_NEXT(filter, next);
2508 	}
2509 
2510 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2511 	 * command to create MAC+VLAN filter with the right flags, enables set.
2512 	 */
2513 	filter = bnxt_alloc_filter(bp);
2514 	if (!filter) {
2515 		PMD_DRV_LOG(ERR,
2516 			    "MAC/VLAN filter alloc failed\n");
2517 		return -ENOMEM;
2518 	}
2519 	/* MAC + VLAN ID filter */
2520 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2521 	 * untagged packets are received
2522 	 *
2523 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2524 	 * packets and only the programmed vlan's packets are received
2525 	 */
2526 	filter->l2_ivlan = vlan_id;
2527 	filter->l2_ivlan_mask = 0x0FFF;
2528 	filter->enables |= en;
2529 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2530 
2531 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2532 	if (rc) {
2533 		/* Free the newly allocated filter as we were
2534 		 * not able to create the filter in hardware.
2535 		 */
2536 		bnxt_free_filter(bp, filter);
2537 		return rc;
2538 	}
2539 
2540 	filter->mac_index = 0;
2541 	/* Add this new filter to the list */
2542 	if (vlan_id == 0)
2543 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2544 	else
2545 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2546 
2547 	PMD_DRV_LOG(INFO,
2548 		    "Added Vlan filter for %d\n", vlan_id);
2549 	return rc;
2550 }
2551 
2552 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2553 		uint16_t vlan_id, int on)
2554 {
2555 	struct bnxt *bp = eth_dev->data->dev_private;
2556 	int rc;
2557 
2558 	rc = is_bnxt_in_error(bp);
2559 	if (rc)
2560 		return rc;
2561 
2562 	if (!eth_dev->data->dev_started) {
2563 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2564 		return -EINVAL;
2565 	}
2566 
2567 	/* These operations apply to ALL existing MAC/VLAN filters */
2568 	if (on)
2569 		return bnxt_add_vlan_filter(bp, vlan_id);
2570 	else
2571 		return bnxt_del_vlan_filter(bp, vlan_id);
2572 }
2573 
2574 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2575 				    struct bnxt_vnic_info *vnic)
2576 {
2577 	struct bnxt_filter_info *filter;
2578 	int rc;
2579 
2580 	filter = STAILQ_FIRST(&vnic->filter);
2581 	while (filter) {
2582 		if (filter->mac_index == 0 &&
2583 		    !memcmp(filter->l2_addr, bp->mac_addr,
2584 			    RTE_ETHER_ADDR_LEN)) {
2585 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2586 			if (!rc) {
2587 				STAILQ_REMOVE(&vnic->filter, filter,
2588 					      bnxt_filter_info, next);
2589 				bnxt_free_filter(bp, filter);
2590 			}
2591 			return rc;
2592 		}
2593 		filter = STAILQ_NEXT(filter, next);
2594 	}
2595 	return 0;
2596 }
2597 
2598 static int
2599 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2600 {
2601 	struct bnxt_vnic_info *vnic;
2602 	unsigned int i;
2603 	int rc;
2604 
2605 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2606 	if (!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)) {
2607 		/* Remove any VLAN filters programmed */
2608 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2609 			bnxt_del_vlan_filter(bp, i);
2610 
2611 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2612 		if (rc)
2613 			return rc;
2614 	} else {
2615 		/* Default filter will allow packets that match the
2616 		 * dest mac. So, it has to be deleted, otherwise, we
2617 		 * will endup receiving vlan packets for which the
2618 		 * filter is not programmed, when hw-vlan-filter
2619 		 * configuration is ON
2620 		 */
2621 		bnxt_del_dflt_mac_filter(bp, vnic);
2622 		/* This filter will allow only untagged packets */
2623 		bnxt_add_vlan_filter(bp, 0);
2624 	}
2625 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2626 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER));
2627 
2628 	return 0;
2629 }
2630 
2631 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2632 {
2633 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2634 	unsigned int i;
2635 	int rc;
2636 
2637 	/* Destroy vnic filters and vnic */
2638 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2639 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2640 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2641 			bnxt_del_vlan_filter(bp, i);
2642 	}
2643 	bnxt_del_dflt_mac_filter(bp, vnic);
2644 
2645 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2646 	if (rc)
2647 		return rc;
2648 
2649 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2650 	if (rc)
2651 		return rc;
2652 
2653 	rte_free(vnic->fw_grp_ids);
2654 	vnic->fw_grp_ids = NULL;
2655 
2656 	vnic->rx_queue_cnt = 0;
2657 
2658 	return 0;
2659 }
2660 
2661 static int
2662 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2663 {
2664 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2665 	int rc;
2666 
2667 	/* Destroy, recreate and reconfigure the default vnic */
2668 	rc = bnxt_free_one_vnic(bp, 0);
2669 	if (rc)
2670 		return rc;
2671 
2672 	/* default vnic 0 */
2673 	rc = bnxt_setup_one_vnic(bp, 0);
2674 	if (rc)
2675 		return rc;
2676 
2677 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2678 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2679 		rc = bnxt_add_vlan_filter(bp, 0);
2680 		if (rc)
2681 			return rc;
2682 		rc = bnxt_restore_vlan_filters(bp);
2683 		if (rc)
2684 			return rc;
2685 	} else {
2686 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2687 		if (rc)
2688 			return rc;
2689 	}
2690 
2691 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2692 	if (rc)
2693 		return rc;
2694 
2695 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2696 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP));
2697 
2698 	return rc;
2699 }
2700 
2701 static int
2702 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2703 {
2704 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2705 	struct bnxt *bp = dev->data->dev_private;
2706 	int rc;
2707 
2708 	rc = is_bnxt_in_error(bp);
2709 	if (rc)
2710 		return rc;
2711 
2712 	/* Filter settings will get applied when port is started */
2713 	if (!dev->data->dev_started)
2714 		return 0;
2715 
2716 	if (mask & RTE_ETH_VLAN_FILTER_MASK) {
2717 		/* Enable or disable VLAN filtering */
2718 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2719 		if (rc)
2720 			return rc;
2721 	}
2722 
2723 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
2724 		/* Enable or disable VLAN stripping */
2725 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2726 		if (rc)
2727 			return rc;
2728 	}
2729 
2730 	if (mask & RTE_ETH_VLAN_EXTEND_MASK) {
2731 		if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND)
2732 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2733 		else
2734 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2735 	}
2736 
2737 	return 0;
2738 }
2739 
2740 static int
2741 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2742 		      uint16_t tpid)
2743 {
2744 	struct bnxt *bp = dev->data->dev_private;
2745 	int qinq = dev->data->dev_conf.rxmode.offloads &
2746 		   RTE_ETH_RX_OFFLOAD_VLAN_EXTEND;
2747 
2748 	if (vlan_type != RTE_ETH_VLAN_TYPE_INNER &&
2749 	    vlan_type != RTE_ETH_VLAN_TYPE_OUTER) {
2750 		PMD_DRV_LOG(ERR,
2751 			    "Unsupported vlan type.");
2752 		return -EINVAL;
2753 	}
2754 	if (!qinq) {
2755 		PMD_DRV_LOG(ERR,
2756 			    "QinQ not enabled. Needs to be ON as we can "
2757 			    "accelerate only outer vlan\n");
2758 		return -EINVAL;
2759 	}
2760 
2761 	if (vlan_type == RTE_ETH_VLAN_TYPE_OUTER) {
2762 		switch (tpid) {
2763 		case RTE_ETHER_TYPE_QINQ:
2764 			bp->outer_tpid_bd =
2765 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2766 				break;
2767 		case RTE_ETHER_TYPE_VLAN:
2768 			bp->outer_tpid_bd =
2769 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2770 				break;
2771 		case RTE_ETHER_TYPE_QINQ1:
2772 			bp->outer_tpid_bd =
2773 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2774 				break;
2775 		case RTE_ETHER_TYPE_QINQ2:
2776 			bp->outer_tpid_bd =
2777 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2778 				break;
2779 		case RTE_ETHER_TYPE_QINQ3:
2780 			bp->outer_tpid_bd =
2781 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2782 				break;
2783 		default:
2784 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2785 			return -EINVAL;
2786 		}
2787 		bp->outer_tpid_bd |= tpid;
2788 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2789 	} else if (vlan_type == RTE_ETH_VLAN_TYPE_INNER) {
2790 		PMD_DRV_LOG(ERR,
2791 			    "Can accelerate only outer vlan in QinQ\n");
2792 		return -EINVAL;
2793 	}
2794 
2795 	return 0;
2796 }
2797 
2798 static int
2799 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2800 			     struct rte_ether_addr *addr)
2801 {
2802 	struct bnxt *bp = dev->data->dev_private;
2803 	/* Default Filter is tied to VNIC 0 */
2804 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2805 	int rc;
2806 
2807 	rc = is_bnxt_in_error(bp);
2808 	if (rc)
2809 		return rc;
2810 
2811 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2812 		return -EPERM;
2813 
2814 	if (rte_is_zero_ether_addr(addr))
2815 		return -EINVAL;
2816 
2817 	/* Filter settings will get applied when port is started */
2818 	if (!dev->data->dev_started)
2819 		return 0;
2820 
2821 	/* Check if the requested MAC is already added */
2822 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2823 		return 0;
2824 
2825 	/* Destroy filter and re-create it */
2826 	bnxt_del_dflt_mac_filter(bp, vnic);
2827 
2828 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2829 	if (dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2830 		/* This filter will allow only untagged packets */
2831 		rc = bnxt_add_vlan_filter(bp, 0);
2832 	} else {
2833 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2834 	}
2835 
2836 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2837 	return rc;
2838 }
2839 
2840 static int
2841 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2842 			  struct rte_ether_addr *mc_addr_set,
2843 			  uint32_t nb_mc_addr)
2844 {
2845 	struct bnxt *bp = eth_dev->data->dev_private;
2846 	char *mc_addr_list = (char *)mc_addr_set;
2847 	struct bnxt_vnic_info *vnic;
2848 	uint32_t off = 0, i = 0;
2849 	int rc;
2850 
2851 	rc = is_bnxt_in_error(bp);
2852 	if (rc)
2853 		return rc;
2854 
2855 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2856 
2857 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2858 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2859 		goto allmulti;
2860 	}
2861 
2862 	/* TODO Check for Duplicate mcast addresses */
2863 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2864 	for (i = 0; i < nb_mc_addr; i++) {
2865 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2866 			RTE_ETHER_ADDR_LEN);
2867 		off += RTE_ETHER_ADDR_LEN;
2868 	}
2869 
2870 	vnic->mc_addr_cnt = i;
2871 	if (vnic->mc_addr_cnt)
2872 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2873 	else
2874 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2875 
2876 allmulti:
2877 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2878 }
2879 
2880 static int
2881 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2882 {
2883 	struct bnxt *bp = dev->data->dev_private;
2884 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2885 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2886 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2887 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2888 	int ret;
2889 
2890 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2891 			fw_major, fw_minor, fw_updt, fw_rsvd);
2892 	if (ret < 0)
2893 		return -EINVAL;
2894 
2895 	ret += 1; /* add the size of '\0' */
2896 	if (fw_size < (size_t)ret)
2897 		return ret;
2898 	else
2899 		return 0;
2900 }
2901 
2902 static void
2903 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2904 	struct rte_eth_rxq_info *qinfo)
2905 {
2906 	struct bnxt *bp = dev->data->dev_private;
2907 	struct bnxt_rx_queue *rxq;
2908 
2909 	if (is_bnxt_in_error(bp))
2910 		return;
2911 
2912 	rxq = dev->data->rx_queues[queue_id];
2913 
2914 	qinfo->mp = rxq->mb_pool;
2915 	qinfo->scattered_rx = dev->data->scattered_rx;
2916 	qinfo->nb_desc = rxq->nb_rx_desc;
2917 
2918 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2919 	qinfo->conf.rx_drop_en = rxq->drop_en;
2920 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2921 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2922 }
2923 
2924 static void
2925 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2926 	struct rte_eth_txq_info *qinfo)
2927 {
2928 	struct bnxt *bp = dev->data->dev_private;
2929 	struct bnxt_tx_queue *txq;
2930 
2931 	if (is_bnxt_in_error(bp))
2932 		return;
2933 
2934 	txq = dev->data->tx_queues[queue_id];
2935 
2936 	qinfo->nb_desc = txq->nb_tx_desc;
2937 
2938 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2939 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2940 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2941 
2942 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2943 	qinfo->conf.tx_rs_thresh = 0;
2944 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2945 	qinfo->conf.offloads = txq->offloads;
2946 }
2947 
2948 static const struct {
2949 	eth_rx_burst_t pkt_burst;
2950 	const char *info;
2951 } bnxt_rx_burst_info[] = {
2952 	{bnxt_recv_pkts,		"Scalar"},
2953 #if defined(RTE_ARCH_X86)
2954 	{bnxt_recv_pkts_vec,		"Vector SSE"},
2955 #endif
2956 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2957 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
2958 #endif
2959 #if defined(RTE_ARCH_ARM64)
2960 	{bnxt_recv_pkts_vec,		"Vector Neon"},
2961 #endif
2962 };
2963 
2964 static int
2965 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2966 		       struct rte_eth_burst_mode *mode)
2967 {
2968 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2969 	size_t i;
2970 
2971 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2972 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2973 			snprintf(mode->info, sizeof(mode->info), "%s",
2974 				 bnxt_rx_burst_info[i].info);
2975 			return 0;
2976 		}
2977 	}
2978 
2979 	return -EINVAL;
2980 }
2981 
2982 static const struct {
2983 	eth_tx_burst_t pkt_burst;
2984 	const char *info;
2985 } bnxt_tx_burst_info[] = {
2986 	{bnxt_xmit_pkts,		"Scalar"},
2987 #if defined(RTE_ARCH_X86)
2988 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
2989 #endif
2990 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2991 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
2992 #endif
2993 #if defined(RTE_ARCH_ARM64)
2994 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
2995 #endif
2996 };
2997 
2998 static int
2999 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
3000 		       struct rte_eth_burst_mode *mode)
3001 {
3002 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
3003 	size_t i;
3004 
3005 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
3006 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
3007 			snprintf(mode->info, sizeof(mode->info), "%s",
3008 				 bnxt_tx_burst_info[i].info);
3009 			return 0;
3010 		}
3011 	}
3012 
3013 	return -EINVAL;
3014 }
3015 
3016 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3017 {
3018 	uint32_t overhead = BNXT_MAX_PKT_LEN - BNXT_MAX_MTU;
3019 	struct bnxt *bp = eth_dev->data->dev_private;
3020 	uint32_t new_pkt_size;
3021 	uint32_t rc;
3022 	uint32_t i;
3023 
3024 	rc = is_bnxt_in_error(bp);
3025 	if (rc)
3026 		return rc;
3027 
3028 	/* Exit if receive queues are not configured yet */
3029 	if (!eth_dev->data->nb_rx_queues)
3030 		return rc;
3031 
3032 	new_pkt_size = new_mtu + overhead;
3033 
3034 	/*
3035 	 * Disallow any MTU change that would require scattered receive support
3036 	 * if it is not already enabled.
3037 	 */
3038 	if (eth_dev->data->dev_started &&
3039 	    !eth_dev->data->scattered_rx &&
3040 	    (new_pkt_size >
3041 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
3042 		PMD_DRV_LOG(ERR,
3043 			    "MTU change would require scattered rx support. ");
3044 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
3045 		return -EINVAL;
3046 	}
3047 
3048 	if (new_mtu > RTE_ETHER_MTU)
3049 		bp->flags |= BNXT_FLAG_JUMBO;
3050 	else
3051 		bp->flags &= ~BNXT_FLAG_JUMBO;
3052 
3053 	/* Is there a change in mtu setting? */
3054 	if (eth_dev->data->mtu == new_mtu)
3055 		return rc;
3056 
3057 	for (i = 0; i < bp->nr_vnics; i++) {
3058 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3059 		uint16_t size = 0;
3060 
3061 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
3062 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
3063 		if (rc)
3064 			break;
3065 
3066 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
3067 		size -= RTE_PKTMBUF_HEADROOM;
3068 
3069 		if (size < new_mtu) {
3070 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
3071 			if (rc)
3072 				return rc;
3073 		}
3074 	}
3075 
3076 	if (bnxt_hwrm_config_host_mtu(bp))
3077 		PMD_DRV_LOG(WARNING, "Failed to configure host MTU\n");
3078 
3079 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3080 
3081 	return rc;
3082 }
3083 
3084 static int
3085 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3086 {
3087 	struct bnxt *bp = dev->data->dev_private;
3088 	uint16_t vlan = bp->vlan;
3089 	int rc;
3090 
3091 	rc = is_bnxt_in_error(bp);
3092 	if (rc)
3093 		return rc;
3094 
3095 	if (!BNXT_SINGLE_PF(bp)) {
3096 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3097 		return -ENOTSUP;
3098 	}
3099 	bp->vlan = on ? pvid : 0;
3100 
3101 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3102 	if (rc)
3103 		bp->vlan = vlan;
3104 	return rc;
3105 }
3106 
3107 static int
3108 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3109 {
3110 	struct bnxt *bp = dev->data->dev_private;
3111 	int rc;
3112 
3113 	rc = is_bnxt_in_error(bp);
3114 	if (rc)
3115 		return rc;
3116 
3117 	return bnxt_hwrm_port_led_cfg(bp, true);
3118 }
3119 
3120 static int
3121 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3122 {
3123 	struct bnxt *bp = dev->data->dev_private;
3124 	int rc;
3125 
3126 	rc = is_bnxt_in_error(bp);
3127 	if (rc)
3128 		return rc;
3129 
3130 	return bnxt_hwrm_port_led_cfg(bp, false);
3131 }
3132 
3133 static uint32_t
3134 bnxt_rx_queue_count_op(void *rx_queue)
3135 {
3136 	struct bnxt *bp;
3137 	struct bnxt_cp_ring_info *cpr;
3138 	uint32_t desc = 0, raw_cons, cp_ring_size;
3139 	struct bnxt_rx_queue *rxq;
3140 	struct rx_pkt_cmpl *rxcmp;
3141 	int rc;
3142 
3143 	rxq = rx_queue;
3144 	bp = rxq->bp;
3145 
3146 	rc = is_bnxt_in_error(bp);
3147 	if (rc)
3148 		return rc;
3149 
3150 	cpr = rxq->cp_ring;
3151 	raw_cons = cpr->cp_raw_cons;
3152 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3153 
3154 	while (1) {
3155 		uint32_t agg_cnt, cons, cmpl_type;
3156 
3157 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3158 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3159 
3160 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3161 			break;
3162 
3163 		cmpl_type = CMP_TYPE(rxcmp);
3164 
3165 		switch (cmpl_type) {
3166 		case CMPL_BASE_TYPE_RX_L2:
3167 		case CMPL_BASE_TYPE_RX_L2_V2:
3168 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3169 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3170 			desc++;
3171 			break;
3172 
3173 		case CMPL_BASE_TYPE_RX_TPA_END:
3174 			if (BNXT_CHIP_P5(rxq->bp)) {
3175 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3176 
3177 				p5_tpa_end = (void *)rxcmp;
3178 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3179 			} else {
3180 				struct rx_tpa_end_cmpl *tpa_end;
3181 
3182 				tpa_end = (void *)rxcmp;
3183 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3184 			}
3185 
3186 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3187 			desc++;
3188 			break;
3189 
3190 		default:
3191 			raw_cons += CMP_LEN(cmpl_type);
3192 		}
3193 	}
3194 
3195 	return desc;
3196 }
3197 
3198 static int
3199 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3200 {
3201 	struct bnxt_rx_queue *rxq = rx_queue;
3202 	struct bnxt_cp_ring_info *cpr;
3203 	struct bnxt_rx_ring_info *rxr;
3204 	uint32_t desc, raw_cons, cp_ring_size;
3205 	struct bnxt *bp = rxq->bp;
3206 	struct rx_pkt_cmpl *rxcmp;
3207 	int rc;
3208 
3209 	rc = is_bnxt_in_error(bp);
3210 	if (rc)
3211 		return rc;
3212 
3213 	if (offset >= rxq->nb_rx_desc)
3214 		return -EINVAL;
3215 
3216 	rxr = rxq->rx_ring;
3217 	cpr = rxq->cp_ring;
3218 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3219 
3220 	/*
3221 	 * For the vector receive case, the completion at the requested
3222 	 * offset can be indexed directly.
3223 	 */
3224 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3225 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3226 		struct rx_pkt_cmpl *rxcmp;
3227 		uint32_t cons;
3228 
3229 		/* Check status of completion descriptor. */
3230 		raw_cons = cpr->cp_raw_cons +
3231 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3232 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3233 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3234 
3235 		if (bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3236 			return RTE_ETH_RX_DESC_DONE;
3237 
3238 		/* Check whether rx desc has an mbuf attached. */
3239 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3240 		if (cons >= rxq->rxrearm_start &&
3241 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3242 			return RTE_ETH_RX_DESC_UNAVAIL;
3243 		}
3244 
3245 		return RTE_ETH_RX_DESC_AVAIL;
3246 	}
3247 #endif
3248 
3249 	/*
3250 	 * For the non-vector receive case, scan the completion ring to
3251 	 * locate the completion descriptor for the requested offset.
3252 	 */
3253 	raw_cons = cpr->cp_raw_cons;
3254 	desc = 0;
3255 	while (1) {
3256 		uint32_t agg_cnt, cons, cmpl_type;
3257 
3258 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3259 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3260 
3261 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3262 			break;
3263 
3264 		cmpl_type = CMP_TYPE(rxcmp);
3265 
3266 		switch (cmpl_type) {
3267 		case CMPL_BASE_TYPE_RX_L2:
3268 		case CMPL_BASE_TYPE_RX_L2_V2:
3269 			if (desc == offset) {
3270 				cons = rxcmp->opaque;
3271 				if (rxr->rx_buf_ring[cons])
3272 					return RTE_ETH_RX_DESC_DONE;
3273 				else
3274 					return RTE_ETH_RX_DESC_UNAVAIL;
3275 			}
3276 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3277 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3278 			desc++;
3279 			break;
3280 
3281 		case CMPL_BASE_TYPE_RX_TPA_END:
3282 			if (desc == offset)
3283 				return RTE_ETH_RX_DESC_DONE;
3284 
3285 			if (BNXT_CHIP_P5(rxq->bp)) {
3286 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3287 
3288 				p5_tpa_end = (void *)rxcmp;
3289 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3290 			} else {
3291 				struct rx_tpa_end_cmpl *tpa_end;
3292 
3293 				tpa_end = (void *)rxcmp;
3294 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3295 			}
3296 
3297 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3298 			desc++;
3299 			break;
3300 
3301 		default:
3302 			raw_cons += CMP_LEN(cmpl_type);
3303 		}
3304 	}
3305 
3306 	return RTE_ETH_RX_DESC_AVAIL;
3307 }
3308 
3309 static int
3310 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3311 {
3312 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3313 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3314 	uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3315 	struct cmpl_base *cp_desc_ring;
3316 	int rc;
3317 
3318 	rc = is_bnxt_in_error(txq->bp);
3319 	if (rc)
3320 		return rc;
3321 
3322 	if (offset >= txq->nb_tx_desc)
3323 		return -EINVAL;
3324 
3325 	/* Return "desc done" if descriptor is available for use. */
3326 	if (bnxt_tx_bds_in_hw(txq) <= offset)
3327 		return RTE_ETH_TX_DESC_DONE;
3328 
3329 	raw_cons = cpr->cp_raw_cons;
3330 	cp_desc_ring = cpr->cp_desc_ring;
3331 	ring_mask = cpr->cp_ring_struct->ring_mask;
3332 
3333 	/* Check to see if hw has posted a completion for the descriptor. */
3334 	while (1) {
3335 		struct tx_cmpl *txcmp;
3336 		uint32_t cons;
3337 
3338 		cons = RING_CMPL(ring_mask, raw_cons);
3339 		txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3340 
3341 		if (!bnxt_cpr_cmp_valid(txcmp, raw_cons, ring_mask + 1))
3342 			break;
3343 
3344 		if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3345 			nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3346 
3347 		if (nb_tx_pkts > offset)
3348 			return RTE_ETH_TX_DESC_DONE;
3349 
3350 		raw_cons = NEXT_RAW_CMP(raw_cons);
3351 	}
3352 
3353 	/* Descriptor is pending transmit, not yet completed by hardware. */
3354 	return RTE_ETH_TX_DESC_FULL;
3355 }
3356 
3357 int
3358 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3359 		     const struct rte_flow_ops **ops)
3360 {
3361 	struct bnxt *bp = dev->data->dev_private;
3362 	int ret = 0;
3363 
3364 	if (!bp)
3365 		return -EIO;
3366 
3367 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3368 		struct bnxt_representor *vfr = dev->data->dev_private;
3369 		bp = vfr->parent_dev->data->dev_private;
3370 		/* parent is deleted while children are still valid */
3371 		if (!bp) {
3372 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3373 				    dev->data->port_id);
3374 			return -EIO;
3375 		}
3376 	}
3377 
3378 	ret = is_bnxt_in_error(bp);
3379 	if (ret)
3380 		return ret;
3381 
3382 	/* PMD supports thread-safe flow operations.  rte_flow API
3383 	 * functions can avoid mutex for multi-thread safety.
3384 	 */
3385 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3386 
3387 	if (BNXT_TRUFLOW_EN(bp))
3388 		*ops = &bnxt_ulp_rte_flow_ops;
3389 	else
3390 		*ops = &bnxt_flow_ops;
3391 
3392 	return ret;
3393 }
3394 
3395 static const uint32_t *
3396 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3397 {
3398 	static const uint32_t ptypes[] = {
3399 		RTE_PTYPE_L2_ETHER_VLAN,
3400 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3401 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3402 		RTE_PTYPE_L4_ICMP,
3403 		RTE_PTYPE_L4_TCP,
3404 		RTE_PTYPE_L4_UDP,
3405 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3406 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3407 		RTE_PTYPE_INNER_L4_ICMP,
3408 		RTE_PTYPE_INNER_L4_TCP,
3409 		RTE_PTYPE_INNER_L4_UDP,
3410 		RTE_PTYPE_UNKNOWN
3411 	};
3412 
3413 	if (!dev->rx_pkt_burst)
3414 		return NULL;
3415 
3416 	return ptypes;
3417 }
3418 
3419 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3420 			 int reg_win)
3421 {
3422 	uint32_t reg_base = *reg_arr & 0xfffff000;
3423 	uint32_t win_off;
3424 	int i;
3425 
3426 	for (i = 0; i < count; i++) {
3427 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3428 			return -ERANGE;
3429 	}
3430 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3431 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3432 	return 0;
3433 }
3434 
3435 static int bnxt_map_ptp_regs(struct bnxt *bp)
3436 {
3437 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3438 	uint32_t *reg_arr;
3439 	int rc, i;
3440 
3441 	reg_arr = ptp->rx_regs;
3442 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3443 	if (rc)
3444 		return rc;
3445 
3446 	reg_arr = ptp->tx_regs;
3447 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3448 	if (rc)
3449 		return rc;
3450 
3451 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3452 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3453 
3454 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3455 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3456 
3457 	return 0;
3458 }
3459 
3460 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3461 {
3462 	rte_write32(0, (uint8_t *)bp->bar0 +
3463 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3464 	rte_write32(0, (uint8_t *)bp->bar0 +
3465 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3466 }
3467 
3468 static uint64_t bnxt_cc_read(struct bnxt *bp)
3469 {
3470 	uint64_t ns;
3471 
3472 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3473 			      BNXT_GRCPF_REG_SYNC_TIME));
3474 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3475 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3476 	return ns;
3477 }
3478 
3479 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3480 {
3481 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3482 	uint32_t fifo;
3483 
3484 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3485 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3486 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3487 		return -EAGAIN;
3488 
3489 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3490 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3491 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3492 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3493 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3494 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3495 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3496 
3497 	return 0;
3498 }
3499 
3500 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3501 {
3502 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3503 	struct bnxt_pf_info *pf = bp->pf;
3504 	uint16_t port_id;
3505 	int i = 0;
3506 	uint32_t fifo;
3507 
3508 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3509 		return -EINVAL;
3510 
3511 	port_id = pf->port_id;
3512 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3513 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3514 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3515 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3516 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3517 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3518 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3519 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3520 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3521 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3522 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3523 		i++;
3524 	}
3525 
3526 	if (i >= BNXT_PTP_RX_PND_CNT)
3527 		return -EBUSY;
3528 
3529 	return 0;
3530 }
3531 
3532 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3533 {
3534 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3535 	struct bnxt_pf_info *pf = bp->pf;
3536 	uint16_t port_id;
3537 	uint32_t fifo;
3538 
3539 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3540 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3541 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3542 		return -EAGAIN;
3543 
3544 	port_id = pf->port_id;
3545 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3546 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3547 
3548 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3549 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3550 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3551 		return bnxt_clr_rx_ts(bp, ts);
3552 
3553 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3554 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3555 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3556 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3557 
3558 	return 0;
3559 }
3560 
3561 static int
3562 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3563 {
3564 	uint64_t ns;
3565 	struct bnxt *bp = dev->data->dev_private;
3566 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3567 
3568 	if (!ptp)
3569 		return -ENOTSUP;
3570 
3571 	ns = rte_timespec_to_ns(ts);
3572 	/* Set the timecounters to a new value. */
3573 	ptp->tc.nsec = ns;
3574 	ptp->tx_tstamp_tc.nsec = ns;
3575 	ptp->rx_tstamp_tc.nsec = ns;
3576 
3577 	return 0;
3578 }
3579 
3580 static int
3581 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3582 {
3583 	struct bnxt *bp = dev->data->dev_private;
3584 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3585 	uint64_t ns, systime_cycles = 0;
3586 	int rc = 0;
3587 
3588 	if (!ptp)
3589 		return -ENOTSUP;
3590 
3591 	if (BNXT_CHIP_P5(bp))
3592 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3593 					     &systime_cycles);
3594 	else
3595 		systime_cycles = bnxt_cc_read(bp);
3596 
3597 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3598 	*ts = rte_ns_to_timespec(ns);
3599 
3600 	return rc;
3601 }
3602 static int
3603 bnxt_timesync_enable(struct rte_eth_dev *dev)
3604 {
3605 	struct bnxt *bp = dev->data->dev_private;
3606 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3607 	uint32_t shift = 0;
3608 	int rc;
3609 
3610 	if (!ptp)
3611 		return -ENOTSUP;
3612 
3613 	ptp->rx_filter = 1;
3614 	ptp->tx_tstamp_en = 1;
3615 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3616 
3617 	rc = bnxt_hwrm_ptp_cfg(bp);
3618 	if (rc)
3619 		return rc;
3620 
3621 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3622 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3623 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3624 
3625 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3626 	ptp->tc.cc_shift = shift;
3627 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3628 
3629 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3630 	ptp->rx_tstamp_tc.cc_shift = shift;
3631 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3632 
3633 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3634 	ptp->tx_tstamp_tc.cc_shift = shift;
3635 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3636 
3637 	if (!BNXT_CHIP_P5(bp))
3638 		bnxt_map_ptp_regs(bp);
3639 	else
3640 		rc = bnxt_ptp_start(bp);
3641 
3642 	return rc;
3643 }
3644 
3645 static int
3646 bnxt_timesync_disable(struct rte_eth_dev *dev)
3647 {
3648 	struct bnxt *bp = dev->data->dev_private;
3649 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3650 
3651 	if (!ptp)
3652 		return -ENOTSUP;
3653 
3654 	ptp->rx_filter = 0;
3655 	ptp->tx_tstamp_en = 0;
3656 	ptp->rxctl = 0;
3657 
3658 	bnxt_hwrm_ptp_cfg(bp);
3659 
3660 	if (!BNXT_CHIP_P5(bp))
3661 		bnxt_unmap_ptp_regs(bp);
3662 	else
3663 		bnxt_ptp_stop(bp);
3664 
3665 	return 0;
3666 }
3667 
3668 static int
3669 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3670 				 struct timespec *timestamp,
3671 				 uint32_t flags __rte_unused)
3672 {
3673 	struct bnxt *bp = dev->data->dev_private;
3674 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3675 	uint64_t rx_tstamp_cycles = 0;
3676 	uint64_t ns;
3677 
3678 	if (!ptp)
3679 		return -ENOTSUP;
3680 
3681 	if (BNXT_CHIP_P5(bp))
3682 		rx_tstamp_cycles = ptp->rx_timestamp;
3683 	else
3684 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3685 
3686 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3687 	*timestamp = rte_ns_to_timespec(ns);
3688 	return  0;
3689 }
3690 
3691 static int
3692 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3693 				 struct timespec *timestamp)
3694 {
3695 	struct bnxt *bp = dev->data->dev_private;
3696 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3697 	uint64_t tx_tstamp_cycles = 0;
3698 	uint64_t ns;
3699 	int rc = 0;
3700 
3701 	if (!ptp)
3702 		return -ENOTSUP;
3703 
3704 	if (BNXT_CHIP_P5(bp))
3705 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3706 					     &tx_tstamp_cycles);
3707 	else
3708 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3709 
3710 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3711 	*timestamp = rte_ns_to_timespec(ns);
3712 
3713 	return rc;
3714 }
3715 
3716 static int
3717 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3718 {
3719 	struct bnxt *bp = dev->data->dev_private;
3720 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3721 
3722 	if (!ptp)
3723 		return -ENOTSUP;
3724 
3725 	ptp->tc.nsec += delta;
3726 	ptp->tx_tstamp_tc.nsec += delta;
3727 	ptp->rx_tstamp_tc.nsec += delta;
3728 
3729 	return 0;
3730 }
3731 
3732 static int
3733 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3734 {
3735 	struct bnxt *bp = dev->data->dev_private;
3736 	int rc;
3737 	uint32_t dir_entries;
3738 	uint32_t entry_length;
3739 
3740 	rc = is_bnxt_in_error(bp);
3741 	if (rc)
3742 		return rc;
3743 
3744 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3745 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3746 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3747 
3748 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3749 	if (rc != 0)
3750 		return rc;
3751 
3752 	return dir_entries * entry_length;
3753 }
3754 
3755 static int
3756 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3757 		struct rte_dev_eeprom_info *in_eeprom)
3758 {
3759 	struct bnxt *bp = dev->data->dev_private;
3760 	uint32_t index;
3761 	uint32_t offset;
3762 	int rc;
3763 
3764 	rc = is_bnxt_in_error(bp);
3765 	if (rc)
3766 		return rc;
3767 
3768 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3769 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3770 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3771 		    in_eeprom->offset, in_eeprom->length);
3772 
3773 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3774 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3775 						in_eeprom->data);
3776 
3777 	index = in_eeprom->offset >> 24;
3778 	offset = in_eeprom->offset & 0xffffff;
3779 
3780 	if (index != 0)
3781 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3782 					   in_eeprom->length, in_eeprom->data);
3783 
3784 	return 0;
3785 }
3786 
3787 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3788 {
3789 	switch (dir_type) {
3790 	case BNX_DIR_TYPE_CHIMP_PATCH:
3791 	case BNX_DIR_TYPE_BOOTCODE:
3792 	case BNX_DIR_TYPE_BOOTCODE_2:
3793 	case BNX_DIR_TYPE_APE_FW:
3794 	case BNX_DIR_TYPE_APE_PATCH:
3795 	case BNX_DIR_TYPE_KONG_FW:
3796 	case BNX_DIR_TYPE_KONG_PATCH:
3797 	case BNX_DIR_TYPE_BONO_FW:
3798 	case BNX_DIR_TYPE_BONO_PATCH:
3799 		/* FALLTHROUGH */
3800 		return true;
3801 	}
3802 
3803 	return false;
3804 }
3805 
3806 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3807 {
3808 	switch (dir_type) {
3809 	case BNX_DIR_TYPE_AVS:
3810 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3811 	case BNX_DIR_TYPE_PCIE:
3812 	case BNX_DIR_TYPE_TSCF_UCODE:
3813 	case BNX_DIR_TYPE_EXT_PHY:
3814 	case BNX_DIR_TYPE_CCM:
3815 	case BNX_DIR_TYPE_ISCSI_BOOT:
3816 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3817 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3818 		/* FALLTHROUGH */
3819 		return true;
3820 	}
3821 
3822 	return false;
3823 }
3824 
3825 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3826 {
3827 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3828 		bnxt_dir_type_is_other_exec_format(dir_type);
3829 }
3830 
3831 static int
3832 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3833 		struct rte_dev_eeprom_info *in_eeprom)
3834 {
3835 	struct bnxt *bp = dev->data->dev_private;
3836 	uint8_t index, dir_op;
3837 	uint16_t type, ext, ordinal, attr;
3838 	int rc;
3839 
3840 	rc = is_bnxt_in_error(bp);
3841 	if (rc)
3842 		return rc;
3843 
3844 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3845 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3846 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3847 		    in_eeprom->offset, in_eeprom->length);
3848 
3849 	if (!BNXT_PF(bp)) {
3850 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3851 		return -EINVAL;
3852 	}
3853 
3854 	type = in_eeprom->magic >> 16;
3855 
3856 	if (type == 0xffff) { /* special value for directory operations */
3857 		index = in_eeprom->magic & 0xff;
3858 		dir_op = in_eeprom->magic >> 8;
3859 		if (index == 0)
3860 			return -EINVAL;
3861 		switch (dir_op) {
3862 		case 0x0e: /* erase */
3863 			if (in_eeprom->offset != ~in_eeprom->magic)
3864 				return -EINVAL;
3865 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3866 		default:
3867 			return -EINVAL;
3868 		}
3869 	}
3870 
3871 	/* Create or re-write an NVM item: */
3872 	if (bnxt_dir_type_is_executable(type) == true)
3873 		return -EOPNOTSUPP;
3874 	ext = in_eeprom->magic & 0xffff;
3875 	ordinal = in_eeprom->offset >> 16;
3876 	attr = in_eeprom->offset & 0xffff;
3877 
3878 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3879 				     in_eeprom->data, in_eeprom->length);
3880 }
3881 
3882 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3883 				struct rte_eth_dev_module_info *modinfo)
3884 {
3885 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3886 	struct bnxt *bp = dev->data->dev_private;
3887 	int rc;
3888 
3889 	/* No point in going further if phy status indicates
3890 	 * module is not inserted or if it is powered down or
3891 	 * if it is of type 10GBase-T
3892 	 */
3893 	if (bp->link_info->module_status >
3894 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3895 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3896 			    dev->data->port_id);
3897 		return -ENOTSUP;
3898 	}
3899 
3900 	/* This feature is not supported in older firmware versions */
3901 	if (bp->hwrm_spec_code < 0x10202) {
3902 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3903 			    dev->data->port_id);
3904 		return -ENOTSUP;
3905 	}
3906 
3907 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3908 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3909 						   module_info);
3910 
3911 	if (rc)
3912 		return rc;
3913 
3914 	switch (module_info[0]) {
3915 	case SFF_MODULE_ID_SFP:
3916 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3917 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3918 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3919 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3920 		break;
3921 	case SFF_MODULE_ID_QSFP:
3922 	case SFF_MODULE_ID_QSFP_PLUS:
3923 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3924 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3925 		break;
3926 	case SFF_MODULE_ID_QSFP28:
3927 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3928 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3929 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3930 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3931 		break;
3932 	default:
3933 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3934 		return -ENOTSUP;
3935 	}
3936 
3937 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3938 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3939 
3940 	return 0;
3941 }
3942 
3943 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3944 				  struct rte_dev_eeprom_info *info)
3945 {
3946 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3947 	uint32_t offset = info->offset, length = info->length;
3948 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3949 	struct bnxt *bp = dev->data->dev_private;
3950 	uint8_t *data = info->data;
3951 	uint8_t page = offset >> 7;
3952 	uint8_t max_pages = 2;
3953 	uint8_t opt_pages;
3954 	int rc;
3955 
3956 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3957 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3958 						   module_info);
3959 	if (rc)
3960 		return rc;
3961 
3962 	switch (module_info[0]) {
3963 	case SFF_MODULE_ID_SFP:
3964 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
3965 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
3966 			pg_addr[2] = I2C_DEV_ADDR_A2;
3967 			pg_addr[3] = I2C_DEV_ADDR_A2;
3968 			max_pages = 4;
3969 		}
3970 		break;
3971 	case SFF_MODULE_ID_QSFP28:
3972 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
3973 							   SFF8636_OPT_PAGES_OFFSET,
3974 							   1, &opt_pages);
3975 		if (rc)
3976 			return rc;
3977 
3978 		if (opt_pages & SFF8636_PAGE1_MASK) {
3979 			pg_addr[2] = I2C_DEV_ADDR_A0;
3980 			max_pages = 3;
3981 		}
3982 		if (opt_pages & SFF8636_PAGE2_MASK) {
3983 			pg_addr[3] = I2C_DEV_ADDR_A0;
3984 			max_pages = 4;
3985 		}
3986 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
3987 			pg_addr[4] = I2C_DEV_ADDR_A0;
3988 			max_pages = 5;
3989 		}
3990 		break;
3991 	default:
3992 		break;
3993 	}
3994 
3995 	memset(data, 0, length);
3996 
3997 	offset &= 0xff;
3998 	while (length && page < max_pages) {
3999 		uint8_t raw_page = page ? page - 1 : 0;
4000 		uint16_t chunk;
4001 
4002 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
4003 			raw_page = 0;
4004 		else if (page)
4005 			offset |= 0x80;
4006 		chunk = RTE_MIN(length, 256 - offset);
4007 
4008 		if (pg_addr[page]) {
4009 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
4010 								   raw_page, offset,
4011 								   chunk, data);
4012 			if (rc)
4013 				return rc;
4014 		}
4015 
4016 		data += chunk;
4017 		length -= chunk;
4018 		offset = 0;
4019 		page += 1 + (chunk > 128);
4020 	}
4021 
4022 	return length ? -EINVAL : 0;
4023 }
4024 
4025 /*
4026  * Initialization
4027  */
4028 
4029 static const struct eth_dev_ops bnxt_dev_ops = {
4030 	.dev_infos_get = bnxt_dev_info_get_op,
4031 	.dev_close = bnxt_dev_close_op,
4032 	.dev_configure = bnxt_dev_configure_op,
4033 	.dev_start = bnxt_dev_start_op,
4034 	.dev_stop = bnxt_dev_stop_op,
4035 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4036 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4037 	.stats_get = bnxt_stats_get_op,
4038 	.stats_reset = bnxt_stats_reset_op,
4039 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4040 	.rx_queue_release = bnxt_rx_queue_release_op,
4041 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4042 	.tx_queue_release = bnxt_tx_queue_release_op,
4043 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4044 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4045 	.reta_update = bnxt_reta_update_op,
4046 	.reta_query = bnxt_reta_query_op,
4047 	.rss_hash_update = bnxt_rss_hash_update_op,
4048 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4049 	.link_update = bnxt_link_update_op,
4050 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4051 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4052 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4053 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4054 	.mac_addr_add = bnxt_mac_addr_add_op,
4055 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4056 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4057 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4058 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4059 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4060 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4061 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4062 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4063 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4064 	.mtu_set = bnxt_mtu_set_op,
4065 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4066 	.xstats_get = bnxt_dev_xstats_get_op,
4067 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4068 	.xstats_reset = bnxt_dev_xstats_reset_op,
4069 	.fw_version_get = bnxt_fw_version_get,
4070 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4071 	.rxq_info_get = bnxt_rxq_info_get_op,
4072 	.txq_info_get = bnxt_txq_info_get_op,
4073 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4074 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4075 	.dev_led_on = bnxt_dev_led_on_op,
4076 	.dev_led_off = bnxt_dev_led_off_op,
4077 	.rx_queue_start = bnxt_rx_queue_start,
4078 	.rx_queue_stop = bnxt_rx_queue_stop,
4079 	.tx_queue_start = bnxt_tx_queue_start,
4080 	.tx_queue_stop = bnxt_tx_queue_stop,
4081 	.flow_ops_get = bnxt_flow_ops_get_op,
4082 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4083 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4084 	.get_eeprom           = bnxt_get_eeprom_op,
4085 	.set_eeprom           = bnxt_set_eeprom_op,
4086 	.get_module_info = bnxt_get_module_info,
4087 	.get_module_eeprom = bnxt_get_module_eeprom,
4088 	.timesync_enable      = bnxt_timesync_enable,
4089 	.timesync_disable     = bnxt_timesync_disable,
4090 	.timesync_read_time   = bnxt_timesync_read_time,
4091 	.timesync_write_time   = bnxt_timesync_write_time,
4092 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4093 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4094 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4095 };
4096 
4097 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4098 {
4099 	uint32_t offset;
4100 
4101 	/* Only pre-map the reset GRC registers using window 3 */
4102 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4103 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4104 
4105 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4106 
4107 	return offset;
4108 }
4109 
4110 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4111 {
4112 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4113 	uint32_t reg_base = 0xffffffff;
4114 	int i;
4115 
4116 	/* Only pre-map the monitoring GRC registers using window 2 */
4117 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4118 		uint32_t reg = info->status_regs[i];
4119 
4120 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4121 			continue;
4122 
4123 		if (reg_base == 0xffffffff)
4124 			reg_base = reg & 0xfffff000;
4125 		if ((reg & 0xfffff000) != reg_base)
4126 			return -ERANGE;
4127 
4128 		/* Use mask 0xffc as the Lower 2 bits indicates
4129 		 * address space location
4130 		 */
4131 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4132 						(reg & 0xffc);
4133 	}
4134 
4135 	if (reg_base == 0xffffffff)
4136 		return 0;
4137 
4138 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4139 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4140 
4141 	return 0;
4142 }
4143 
4144 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4145 {
4146 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4147 	uint32_t delay = info->delay_after_reset[index];
4148 	uint32_t val = info->reset_reg_val[index];
4149 	uint32_t reg = info->reset_reg[index];
4150 	uint32_t type, offset;
4151 	int ret;
4152 
4153 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4154 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4155 
4156 	switch (type) {
4157 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4158 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4159 		if (ret < 0) {
4160 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4161 				    val, offset);
4162 			return;
4163 		}
4164 		break;
4165 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4166 		offset = bnxt_map_reset_regs(bp, offset);
4167 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4168 		break;
4169 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4170 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4171 		break;
4172 	}
4173 	/* wait on a specific interval of time until core reset is complete */
4174 	if (delay)
4175 		rte_delay_ms(delay);
4176 }
4177 
4178 static void bnxt_dev_cleanup(struct bnxt *bp)
4179 {
4180 	bp->eth_dev->data->dev_link.link_status = 0;
4181 	bp->link_info->link_up = 0;
4182 	if (bp->eth_dev->data->dev_started)
4183 		bnxt_dev_stop(bp->eth_dev);
4184 
4185 	bnxt_uninit_resources(bp, true);
4186 }
4187 
4188 static int
4189 bnxt_check_fw_reset_done(struct bnxt *bp)
4190 {
4191 	int timeout = bp->fw_reset_max_msecs;
4192 	uint16_t val = 0;
4193 	int rc;
4194 
4195 	do {
4196 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4197 		if (rc < 0) {
4198 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4199 			return rc;
4200 		}
4201 		if (val != 0xffff)
4202 			break;
4203 		rte_delay_ms(1);
4204 	} while (timeout--);
4205 
4206 	if (val == 0xffff) {
4207 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4208 		return -1;
4209 	}
4210 
4211 	return 0;
4212 }
4213 
4214 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4215 {
4216 	struct rte_eth_dev *dev = bp->eth_dev;
4217 	struct rte_vlan_filter_conf *vfc;
4218 	int vidx, vbit, rc;
4219 	uint16_t vlan_id;
4220 
4221 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4222 		vfc = &dev->data->vlan_filter_conf;
4223 		vidx = vlan_id / 64;
4224 		vbit = vlan_id % 64;
4225 
4226 		/* Each bit corresponds to a VLAN id */
4227 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4228 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4229 			if (rc)
4230 				return rc;
4231 		}
4232 	}
4233 
4234 	return 0;
4235 }
4236 
4237 static int bnxt_restore_mac_filters(struct bnxt *bp)
4238 {
4239 	struct rte_eth_dev *dev = bp->eth_dev;
4240 	struct rte_eth_dev_info dev_info;
4241 	struct rte_ether_addr *addr;
4242 	uint64_t pool_mask;
4243 	uint32_t pool = 0;
4244 	uint32_t i;
4245 	int rc;
4246 
4247 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4248 		return 0;
4249 
4250 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4251 	if (rc)
4252 		return rc;
4253 
4254 	/* replay MAC address configuration */
4255 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4256 		addr = &dev->data->mac_addrs[i];
4257 
4258 		/* skip zero address */
4259 		if (rte_is_zero_ether_addr(addr))
4260 			continue;
4261 
4262 		pool = 0;
4263 		pool_mask = dev->data->mac_pool_sel[i];
4264 
4265 		do {
4266 			if (pool_mask & 1ULL) {
4267 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4268 				if (rc)
4269 					return rc;
4270 			}
4271 			pool_mask >>= 1;
4272 			pool++;
4273 		} while (pool_mask);
4274 	}
4275 
4276 	return 0;
4277 }
4278 
4279 static int bnxt_restore_filters(struct bnxt *bp)
4280 {
4281 	struct rte_eth_dev *dev = bp->eth_dev;
4282 	int ret = 0;
4283 
4284 	if (dev->data->all_multicast) {
4285 		ret = bnxt_allmulticast_enable_op(dev);
4286 		if (ret)
4287 			return ret;
4288 	}
4289 	if (dev->data->promiscuous) {
4290 		ret = bnxt_promiscuous_enable_op(dev);
4291 		if (ret)
4292 			return ret;
4293 	}
4294 
4295 	ret = bnxt_restore_mac_filters(bp);
4296 	if (ret)
4297 		return ret;
4298 
4299 	ret = bnxt_restore_vlan_filters(bp);
4300 	/* TODO restore other filters as well */
4301 	return ret;
4302 }
4303 
4304 static int bnxt_check_fw_ready(struct bnxt *bp)
4305 {
4306 	int timeout = bp->fw_reset_max_msecs;
4307 	int rc = 0;
4308 
4309 	do {
4310 		rc = bnxt_hwrm_poll_ver_get(bp);
4311 		if (rc == 0)
4312 			break;
4313 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4314 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4315 	} while (rc && timeout > 0);
4316 
4317 	if (rc)
4318 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4319 
4320 	return rc;
4321 }
4322 
4323 static void bnxt_dev_recover(void *arg)
4324 {
4325 	struct bnxt *bp = arg;
4326 	int rc = 0;
4327 
4328 	pthread_mutex_lock(&bp->err_recovery_lock);
4329 
4330 	if (!bp->fw_reset_min_msecs) {
4331 		rc = bnxt_check_fw_reset_done(bp);
4332 		if (rc)
4333 			goto err;
4334 	}
4335 
4336 	/* Clear Error flag so that device re-init should happen */
4337 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4338 
4339 	rc = bnxt_check_fw_ready(bp);
4340 	if (rc)
4341 		goto err;
4342 
4343 	rc = bnxt_init_resources(bp, true);
4344 	if (rc) {
4345 		PMD_DRV_LOG(ERR,
4346 			    "Failed to initialize resources after reset\n");
4347 		goto err;
4348 	}
4349 	/* clear reset flag as the device is initialized now */
4350 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4351 
4352 	rc = bnxt_dev_start_op(bp->eth_dev);
4353 	if (rc) {
4354 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4355 		goto err_start;
4356 	}
4357 
4358 	rc = bnxt_restore_filters(bp);
4359 	if (rc)
4360 		goto err_start;
4361 
4362 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4363 	pthread_mutex_unlock(&bp->err_recovery_lock);
4364 
4365 	return;
4366 err_start:
4367 	bnxt_dev_stop(bp->eth_dev);
4368 err:
4369 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4370 	bnxt_uninit_resources(bp, false);
4371 	if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4372 		rte_eth_dev_callback_process(bp->eth_dev,
4373 					     RTE_ETH_EVENT_INTR_RMV,
4374 					     NULL);
4375 	pthread_mutex_unlock(&bp->err_recovery_lock);
4376 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4377 }
4378 
4379 void bnxt_dev_reset_and_resume(void *arg)
4380 {
4381 	struct bnxt *bp = arg;
4382 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4383 	uint16_t val = 0;
4384 	int rc;
4385 
4386 	bnxt_dev_cleanup(bp);
4387 
4388 	bnxt_wait_for_device_shutdown(bp);
4389 
4390 	/* During some fatal firmware error conditions, the PCI config space
4391 	 * register 0x2e which normally contains the subsystem ID will become
4392 	 * 0xffff. This register will revert back to the normal value after
4393 	 * the chip has completed core reset. If we detect this condition,
4394 	 * we can poll this config register immediately for the value to revert.
4395 	 */
4396 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4397 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4398 		if (rc < 0) {
4399 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4400 			return;
4401 		}
4402 		if (val == 0xffff) {
4403 			bp->fw_reset_min_msecs = 0;
4404 			us = 1;
4405 		}
4406 	}
4407 
4408 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4409 	if (rc)
4410 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4411 }
4412 
4413 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4414 {
4415 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4416 	uint32_t reg = info->status_regs[index];
4417 	uint32_t type, offset, val = 0;
4418 	int ret = 0;
4419 
4420 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4421 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4422 
4423 	switch (type) {
4424 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4425 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4426 		if (ret < 0)
4427 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4428 				    offset);
4429 		break;
4430 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4431 		offset = info->mapped_status_regs[index];
4432 		/* FALLTHROUGH */
4433 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4434 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4435 				       offset));
4436 		break;
4437 	}
4438 
4439 	return val;
4440 }
4441 
4442 static int bnxt_fw_reset_all(struct bnxt *bp)
4443 {
4444 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4445 	uint32_t i;
4446 	int rc = 0;
4447 
4448 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4449 		/* Reset through primary function driver */
4450 		for (i = 0; i < info->reg_array_cnt; i++)
4451 			bnxt_write_fw_reset_reg(bp, i);
4452 		/* Wait for time specified by FW after triggering reset */
4453 		rte_delay_ms(info->primary_func_wait_period_after_reset);
4454 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4455 		/* Reset with the help of Kong processor */
4456 		rc = bnxt_hwrm_fw_reset(bp);
4457 		if (rc)
4458 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4459 	}
4460 
4461 	return rc;
4462 }
4463 
4464 static void bnxt_fw_reset_cb(void *arg)
4465 {
4466 	struct bnxt *bp = arg;
4467 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4468 	int rc = 0;
4469 
4470 	/* Only Primary function can do FW reset */
4471 	if (bnxt_is_primary_func(bp) &&
4472 	    bnxt_is_recovery_enabled(bp)) {
4473 		rc = bnxt_fw_reset_all(bp);
4474 		if (rc) {
4475 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4476 			return;
4477 		}
4478 	}
4479 
4480 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4481 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4482 	 * (including MASTER FUNC). After receiving this Async, all the active
4483 	 * drivers should treat this case as FW initiated recovery
4484 	 */
4485 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4486 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4487 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4488 
4489 		/* To recover from error */
4490 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4491 				  (void *)bp);
4492 	}
4493 }
4494 
4495 /* Driver should poll FW heartbeat, reset_counter with the frequency
4496  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4497  * When the driver detects heartbeat stop or change in reset_counter,
4498  * it has to trigger a reset to recover from the error condition.
4499  * A “primary function” is the function who will have the privilege to
4500  * initiate the chimp reset. The primary function will be elected by the
4501  * firmware and will be notified through async message.
4502  */
4503 static void bnxt_check_fw_health(void *arg)
4504 {
4505 	struct bnxt *bp = arg;
4506 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4507 	uint32_t val = 0, wait_msec;
4508 
4509 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4510 	    is_bnxt_in_error(bp))
4511 		return;
4512 
4513 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4514 	if (val == info->last_heart_beat)
4515 		goto reset;
4516 
4517 	info->last_heart_beat = val;
4518 
4519 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4520 	if (val != info->last_reset_counter)
4521 		goto reset;
4522 
4523 	info->last_reset_counter = val;
4524 
4525 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4526 			  bnxt_check_fw_health, (void *)bp);
4527 
4528 	return;
4529 reset:
4530 	/* Stop DMA to/from device */
4531 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4532 	bp->flags |= BNXT_FLAG_FW_RESET;
4533 
4534 	bnxt_stop_rxtx(bp);
4535 
4536 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4537 
4538 	if (bnxt_is_primary_func(bp))
4539 		wait_msec = info->primary_func_wait_period;
4540 	else
4541 		wait_msec = info->normal_func_wait_period;
4542 
4543 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4544 			  bnxt_fw_reset_cb, (void *)bp);
4545 }
4546 
4547 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4548 {
4549 	uint32_t polling_freq;
4550 
4551 	pthread_mutex_lock(&bp->health_check_lock);
4552 
4553 	if (!bnxt_is_recovery_enabled(bp))
4554 		goto done;
4555 
4556 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4557 		goto done;
4558 
4559 	polling_freq = bp->recovery_info->driver_polling_freq;
4560 
4561 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4562 			  bnxt_check_fw_health, (void *)bp);
4563 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4564 
4565 done:
4566 	pthread_mutex_unlock(&bp->health_check_lock);
4567 }
4568 
4569 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4570 {
4571 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4572 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4573 }
4574 
4575 static bool bnxt_vf_pciid(uint16_t device_id)
4576 {
4577 	switch (device_id) {
4578 	case BROADCOM_DEV_ID_57304_VF:
4579 	case BROADCOM_DEV_ID_57406_VF:
4580 	case BROADCOM_DEV_ID_5731X_VF:
4581 	case BROADCOM_DEV_ID_5741X_VF:
4582 	case BROADCOM_DEV_ID_57414_VF:
4583 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4584 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4585 	case BROADCOM_DEV_ID_58802_VF:
4586 	case BROADCOM_DEV_ID_57500_VF1:
4587 	case BROADCOM_DEV_ID_57500_VF2:
4588 	case BROADCOM_DEV_ID_58818_VF:
4589 		/* FALLTHROUGH */
4590 		return true;
4591 	default:
4592 		return false;
4593 	}
4594 }
4595 
4596 /* Phase 5 device */
4597 static bool bnxt_p5_device(uint16_t device_id)
4598 {
4599 	switch (device_id) {
4600 	case BROADCOM_DEV_ID_57508:
4601 	case BROADCOM_DEV_ID_57504:
4602 	case BROADCOM_DEV_ID_57502:
4603 	case BROADCOM_DEV_ID_57508_MF1:
4604 	case BROADCOM_DEV_ID_57504_MF1:
4605 	case BROADCOM_DEV_ID_57502_MF1:
4606 	case BROADCOM_DEV_ID_57508_MF2:
4607 	case BROADCOM_DEV_ID_57504_MF2:
4608 	case BROADCOM_DEV_ID_57502_MF2:
4609 	case BROADCOM_DEV_ID_57500_VF1:
4610 	case BROADCOM_DEV_ID_57500_VF2:
4611 	case BROADCOM_DEV_ID_58812:
4612 	case BROADCOM_DEV_ID_58814:
4613 	case BROADCOM_DEV_ID_58818:
4614 	case BROADCOM_DEV_ID_58818_VF:
4615 		/* FALLTHROUGH */
4616 		return true;
4617 	default:
4618 		return false;
4619 	}
4620 }
4621 
4622 bool bnxt_stratus_device(struct bnxt *bp)
4623 {
4624 	uint16_t device_id = bp->pdev->id.device_id;
4625 
4626 	switch (device_id) {
4627 	case BROADCOM_DEV_ID_STRATUS_NIC:
4628 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4629 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4630 		/* FALLTHROUGH */
4631 		return true;
4632 	default:
4633 		return false;
4634 	}
4635 }
4636 
4637 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4638 {
4639 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4640 	struct bnxt *bp = eth_dev->data->dev_private;
4641 
4642 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4643 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4644 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4645 	if (!bp->bar0 || !bp->doorbell_base) {
4646 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4647 		return -ENODEV;
4648 	}
4649 
4650 	bp->eth_dev = eth_dev;
4651 	bp->pdev = pci_dev;
4652 
4653 	return 0;
4654 }
4655 
4656 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4657 				  struct bnxt_ctx_pg_info *ctx_pg,
4658 				  uint32_t mem_size,
4659 				  const char *suffix,
4660 				  uint16_t idx)
4661 {
4662 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4663 	const struct rte_memzone *mz = NULL;
4664 	char mz_name[RTE_MEMZONE_NAMESIZE];
4665 	rte_iova_t mz_phys_addr;
4666 	uint64_t valid_bits = 0;
4667 	uint32_t sz;
4668 	int i;
4669 
4670 	if (!mem_size)
4671 		return 0;
4672 
4673 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4674 			 BNXT_PAGE_SIZE;
4675 	rmem->page_size = BNXT_PAGE_SIZE;
4676 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4677 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4678 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4679 
4680 	valid_bits = PTU_PTE_VALID;
4681 
4682 	if (rmem->nr_pages > 1) {
4683 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4684 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4685 			 suffix, idx, bp->eth_dev->data->port_id);
4686 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4687 		mz = rte_memzone_lookup(mz_name);
4688 		if (!mz) {
4689 			mz = rte_memzone_reserve_aligned(mz_name,
4690 						rmem->nr_pages * 8,
4691 						bp->eth_dev->device->numa_node,
4692 						RTE_MEMZONE_2MB |
4693 						RTE_MEMZONE_SIZE_HINT_ONLY |
4694 						RTE_MEMZONE_IOVA_CONTIG,
4695 						BNXT_PAGE_SIZE);
4696 			if (mz == NULL)
4697 				return -ENOMEM;
4698 		}
4699 
4700 		memset(mz->addr, 0, mz->len);
4701 		mz_phys_addr = mz->iova;
4702 
4703 		rmem->pg_tbl = mz->addr;
4704 		rmem->pg_tbl_map = mz_phys_addr;
4705 		rmem->pg_tbl_mz = mz;
4706 	}
4707 
4708 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4709 		 suffix, idx, bp->eth_dev->data->port_id);
4710 	mz = rte_memzone_lookup(mz_name);
4711 	if (!mz) {
4712 		mz = rte_memzone_reserve_aligned(mz_name,
4713 						 mem_size,
4714 						 bp->eth_dev->device->numa_node,
4715 						 RTE_MEMZONE_1GB |
4716 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4717 						 RTE_MEMZONE_IOVA_CONTIG,
4718 						 BNXT_PAGE_SIZE);
4719 		if (mz == NULL)
4720 			return -ENOMEM;
4721 	}
4722 
4723 	memset(mz->addr, 0, mz->len);
4724 	mz_phys_addr = mz->iova;
4725 
4726 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4727 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4728 		rmem->dma_arr[i] = mz_phys_addr + sz;
4729 
4730 		if (rmem->nr_pages > 1) {
4731 			if (i == rmem->nr_pages - 2 &&
4732 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4733 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4734 			else if (i == rmem->nr_pages - 1 &&
4735 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4736 				valid_bits |= PTU_PTE_LAST;
4737 
4738 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4739 							   valid_bits);
4740 		}
4741 	}
4742 
4743 	rmem->mz = mz;
4744 	if (rmem->vmem_size)
4745 		rmem->vmem = (void **)mz->addr;
4746 	rmem->dma_arr[0] = mz_phys_addr;
4747 	return 0;
4748 }
4749 
4750 static void bnxt_free_ctx_mem(struct bnxt *bp)
4751 {
4752 	int i;
4753 
4754 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4755 		return;
4756 
4757 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4758 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4759 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4760 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4761 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4762 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4763 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4764 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4765 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4766 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4767 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4768 
4769 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4770 		if (bp->ctx->tqm_mem[i])
4771 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4772 	}
4773 
4774 	rte_free(bp->ctx);
4775 	bp->ctx = NULL;
4776 }
4777 
4778 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4779 
4780 #define min_t(type, x, y) ({                    \
4781 	type __min1 = (x);                      \
4782 	type __min2 = (y);                      \
4783 	__min1 < __min2 ? __min1 : __min2; })
4784 
4785 #define max_t(type, x, y) ({                    \
4786 	type __max1 = (x);                      \
4787 	type __max2 = (y);                      \
4788 	__max1 > __max2 ? __max1 : __max2; })
4789 
4790 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4791 
4792 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4793 {
4794 	struct bnxt_ctx_pg_info *ctx_pg;
4795 	struct bnxt_ctx_mem_info *ctx;
4796 	uint32_t mem_size, ena, entries;
4797 	uint32_t entries_sp, min;
4798 	int i, rc;
4799 
4800 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4801 	if (rc) {
4802 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4803 		return rc;
4804 	}
4805 	ctx = bp->ctx;
4806 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4807 		return 0;
4808 
4809 	ctx_pg = &ctx->qp_mem;
4810 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4811 	if (ctx->qp_entry_size) {
4812 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4813 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4814 		if (rc)
4815 			return rc;
4816 	}
4817 
4818 	ctx_pg = &ctx->srq_mem;
4819 	ctx_pg->entries = ctx->srq_max_l2_entries;
4820 	if (ctx->srq_entry_size) {
4821 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4822 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4823 		if (rc)
4824 			return rc;
4825 	}
4826 
4827 	ctx_pg = &ctx->cq_mem;
4828 	ctx_pg->entries = ctx->cq_max_l2_entries;
4829 	if (ctx->cq_entry_size) {
4830 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4831 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4832 		if (rc)
4833 			return rc;
4834 	}
4835 
4836 	ctx_pg = &ctx->vnic_mem;
4837 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4838 		ctx->vnic_max_ring_table_entries;
4839 	if (ctx->vnic_entry_size) {
4840 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4841 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4842 		if (rc)
4843 			return rc;
4844 	}
4845 
4846 	ctx_pg = &ctx->stat_mem;
4847 	ctx_pg->entries = ctx->stat_max_entries;
4848 	if (ctx->stat_entry_size) {
4849 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4850 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4851 		if (rc)
4852 			return rc;
4853 	}
4854 
4855 	min = ctx->tqm_min_entries_per_ring;
4856 
4857 	entries_sp = ctx->qp_max_l2_entries +
4858 		     ctx->vnic_max_vnic_entries +
4859 		     2 * ctx->qp_min_qp1_entries + min;
4860 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4861 
4862 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4863 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4864 	entries = clamp_t(uint32_t, entries, min,
4865 			  ctx->tqm_max_entries_per_ring);
4866 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4867 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4868 		 * i > 8 is other ext rings.
4869 		 */
4870 		ctx_pg = ctx->tqm_mem[i];
4871 		ctx_pg->entries = i ? entries : entries_sp;
4872 		if (ctx->tqm_entry_size) {
4873 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4874 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4875 						    "tqm_mem", i);
4876 			if (rc)
4877 				return rc;
4878 		}
4879 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4880 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4881 		else
4882 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4883 	}
4884 
4885 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4886 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4887 	if (rc)
4888 		PMD_DRV_LOG(ERR,
4889 			    "Failed to configure context mem: rc = %d\n", rc);
4890 	else
4891 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4892 
4893 	return rc;
4894 }
4895 
4896 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4897 {
4898 	struct rte_pci_device *pci_dev = bp->pdev;
4899 	char mz_name[RTE_MEMZONE_NAMESIZE];
4900 	const struct rte_memzone *mz = NULL;
4901 	uint32_t total_alloc_len;
4902 	rte_iova_t mz_phys_addr;
4903 
4904 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4905 		return 0;
4906 
4907 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4908 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4909 		 pci_dev->addr.bus, pci_dev->addr.devid,
4910 		 pci_dev->addr.function, "rx_port_stats");
4911 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4912 	mz = rte_memzone_lookup(mz_name);
4913 	total_alloc_len =
4914 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4915 				       sizeof(struct rx_port_stats_ext) + 512);
4916 	if (!mz) {
4917 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4918 					 SOCKET_ID_ANY,
4919 					 RTE_MEMZONE_2MB |
4920 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4921 					 RTE_MEMZONE_IOVA_CONTIG);
4922 		if (mz == NULL)
4923 			return -ENOMEM;
4924 	}
4925 	memset(mz->addr, 0, mz->len);
4926 	mz_phys_addr = mz->iova;
4927 
4928 	bp->rx_mem_zone = (const void *)mz;
4929 	bp->hw_rx_port_stats = mz->addr;
4930 	bp->hw_rx_port_stats_map = mz_phys_addr;
4931 
4932 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4933 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4934 		 pci_dev->addr.bus, pci_dev->addr.devid,
4935 		 pci_dev->addr.function, "tx_port_stats");
4936 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4937 	mz = rte_memzone_lookup(mz_name);
4938 	total_alloc_len =
4939 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4940 				       sizeof(struct tx_port_stats_ext) + 512);
4941 	if (!mz) {
4942 		mz = rte_memzone_reserve(mz_name,
4943 					 total_alloc_len,
4944 					 SOCKET_ID_ANY,
4945 					 RTE_MEMZONE_2MB |
4946 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4947 					 RTE_MEMZONE_IOVA_CONTIG);
4948 		if (mz == NULL)
4949 			return -ENOMEM;
4950 	}
4951 	memset(mz->addr, 0, mz->len);
4952 	mz_phys_addr = mz->iova;
4953 
4954 	bp->tx_mem_zone = (const void *)mz;
4955 	bp->hw_tx_port_stats = mz->addr;
4956 	bp->hw_tx_port_stats_map = mz_phys_addr;
4957 	bp->flags |= BNXT_FLAG_PORT_STATS;
4958 
4959 	/* Display extended statistics if FW supports it */
4960 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4961 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4962 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4963 		return 0;
4964 
4965 	bp->hw_rx_port_stats_ext = (void *)
4966 		((uint8_t *)bp->hw_rx_port_stats +
4967 		 sizeof(struct rx_port_stats));
4968 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4969 		sizeof(struct rx_port_stats);
4970 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4971 
4972 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4973 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4974 		bp->hw_tx_port_stats_ext = (void *)
4975 			((uint8_t *)bp->hw_tx_port_stats +
4976 			 sizeof(struct tx_port_stats));
4977 		bp->hw_tx_port_stats_ext_map =
4978 			bp->hw_tx_port_stats_map +
4979 			sizeof(struct tx_port_stats);
4980 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4981 	}
4982 
4983 	return 0;
4984 }
4985 
4986 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4987 {
4988 	struct bnxt *bp = eth_dev->data->dev_private;
4989 	int rc = 0;
4990 
4991 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4992 					       RTE_ETHER_ADDR_LEN *
4993 					       bp->max_l2_ctx,
4994 					       0);
4995 	if (eth_dev->data->mac_addrs == NULL) {
4996 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4997 		return -ENOMEM;
4998 	}
4999 
5000 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
5001 		if (BNXT_PF(bp))
5002 			return -EINVAL;
5003 
5004 		/* Generate a random MAC address, if none was assigned by PF */
5005 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
5006 		bnxt_eth_hw_addr_random(bp->mac_addr);
5007 		PMD_DRV_LOG(INFO,
5008 			    "Assign random MAC:" RTE_ETHER_ADDR_PRT_FMT "\n",
5009 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
5010 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
5011 
5012 		rc = bnxt_hwrm_set_mac(bp);
5013 		if (rc)
5014 			return rc;
5015 	}
5016 
5017 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5018 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5019 
5020 	return rc;
5021 }
5022 
5023 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5024 {
5025 	int rc = 0;
5026 
5027 	/* MAC is already configured in FW */
5028 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5029 		return 0;
5030 
5031 	/* Restore the old MAC configured */
5032 	rc = bnxt_hwrm_set_mac(bp);
5033 	if (rc)
5034 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5035 
5036 	return rc;
5037 }
5038 
5039 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5040 {
5041 	if (!BNXT_PF(bp))
5042 		return;
5043 
5044 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5045 
5046 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5047 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5048 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5049 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5050 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5051 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5052 }
5053 
5054 struct bnxt *
5055 bnxt_get_bp(uint16_t port)
5056 {
5057 	struct bnxt *bp;
5058 	struct rte_eth_dev *dev;
5059 
5060 	if (!rte_eth_dev_is_valid_port(port)) {
5061 		PMD_DRV_LOG(ERR, "Invalid port %d\n", port);
5062 		return NULL;
5063 	}
5064 
5065 	dev = &rte_eth_devices[port];
5066 	if (!is_bnxt_supported(dev)) {
5067 		PMD_DRV_LOG(ERR, "Device %d not supported\n", port);
5068 		return NULL;
5069 	}
5070 
5071 	bp = (struct bnxt *)dev->data->dev_private;
5072 	if (!BNXT_TRUFLOW_EN(bp)) {
5073 		PMD_DRV_LOG(ERR, "TRUFLOW not enabled\n");
5074 		return NULL;
5075 	}
5076 
5077 	return bp;
5078 }
5079 
5080 uint16_t
5081 bnxt_get_svif(uint16_t port_id, bool func_svif,
5082 	      enum bnxt_ulp_intf_type type)
5083 {
5084 	struct rte_eth_dev *eth_dev;
5085 	struct bnxt *bp;
5086 
5087 	eth_dev = &rte_eth_devices[port_id];
5088 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5089 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5090 		if (!vfr)
5091 			return 0;
5092 
5093 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5094 			return vfr->svif;
5095 
5096 		eth_dev = vfr->parent_dev;
5097 	}
5098 
5099 	bp = eth_dev->data->dev_private;
5100 
5101 	return func_svif ? bp->func_svif : bp->port_svif;
5102 }
5103 
5104 void
5105 bnxt_get_iface_mac(uint16_t port, enum bnxt_ulp_intf_type type,
5106 		   uint8_t *mac, uint8_t *parent_mac)
5107 {
5108 	struct rte_eth_dev *eth_dev;
5109 	struct bnxt *bp;
5110 
5111 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF &&
5112 	    type != BNXT_ULP_INTF_TYPE_PF)
5113 		return;
5114 
5115 	eth_dev = &rte_eth_devices[port];
5116 	bp = eth_dev->data->dev_private;
5117 	memcpy(mac, bp->mac_addr, RTE_ETHER_ADDR_LEN);
5118 
5119 	if (type == BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5120 		memcpy(parent_mac, bp->parent->mac_addr, RTE_ETHER_ADDR_LEN);
5121 }
5122 
5123 uint16_t
5124 bnxt_get_parent_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5125 {
5126 	struct rte_eth_dev *eth_dev;
5127 	struct bnxt *bp;
5128 
5129 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5130 		return 0;
5131 
5132 	eth_dev = &rte_eth_devices[port];
5133 	bp = eth_dev->data->dev_private;
5134 
5135 	return bp->parent->vnic;
5136 }
5137 uint16_t
5138 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5139 {
5140 	struct rte_eth_dev *eth_dev;
5141 	struct bnxt_vnic_info *vnic;
5142 	struct bnxt *bp;
5143 
5144 	eth_dev = &rte_eth_devices[port];
5145 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5146 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5147 		if (!vfr)
5148 			return 0;
5149 
5150 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5151 			return vfr->dflt_vnic_id;
5152 
5153 		eth_dev = vfr->parent_dev;
5154 	}
5155 
5156 	bp = eth_dev->data->dev_private;
5157 
5158 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
5159 
5160 	return vnic->fw_vnic_id;
5161 }
5162 
5163 uint16_t
5164 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
5165 {
5166 	struct rte_eth_dev *eth_dev;
5167 	struct bnxt *bp;
5168 
5169 	eth_dev = &rte_eth_devices[port];
5170 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5171 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5172 		if (!vfr)
5173 			return 0;
5174 
5175 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5176 			return vfr->fw_fid;
5177 
5178 		eth_dev = vfr->parent_dev;
5179 	}
5180 
5181 	bp = eth_dev->data->dev_private;
5182 
5183 	return bp->fw_fid;
5184 }
5185 
5186 enum bnxt_ulp_intf_type
5187 bnxt_get_interface_type(uint16_t port)
5188 {
5189 	struct rte_eth_dev *eth_dev;
5190 	struct bnxt *bp;
5191 
5192 	eth_dev = &rte_eth_devices[port];
5193 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
5194 		return BNXT_ULP_INTF_TYPE_VF_REP;
5195 
5196 	bp = eth_dev->data->dev_private;
5197 	if (BNXT_PF(bp))
5198 		return BNXT_ULP_INTF_TYPE_PF;
5199 	else if (BNXT_VF_IS_TRUSTED(bp))
5200 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
5201 	else if (BNXT_VF(bp))
5202 		return BNXT_ULP_INTF_TYPE_VF;
5203 
5204 	return BNXT_ULP_INTF_TYPE_INVALID;
5205 }
5206 
5207 uint16_t
5208 bnxt_get_phy_port_id(uint16_t port_id)
5209 {
5210 	struct bnxt_representor *vfr;
5211 	struct rte_eth_dev *eth_dev;
5212 	struct bnxt *bp;
5213 
5214 	eth_dev = &rte_eth_devices[port_id];
5215 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5216 		vfr = eth_dev->data->dev_private;
5217 		if (!vfr)
5218 			return 0;
5219 
5220 		eth_dev = vfr->parent_dev;
5221 	}
5222 
5223 	bp = eth_dev->data->dev_private;
5224 
5225 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
5226 }
5227 
5228 uint16_t
5229 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
5230 {
5231 	struct rte_eth_dev *eth_dev;
5232 	struct bnxt *bp;
5233 
5234 	eth_dev = &rte_eth_devices[port_id];
5235 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5236 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5237 		if (!vfr)
5238 			return 0;
5239 
5240 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5241 			return vfr->fw_fid - 1;
5242 
5243 		eth_dev = vfr->parent_dev;
5244 	}
5245 
5246 	bp = eth_dev->data->dev_private;
5247 
5248 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
5249 }
5250 
5251 uint16_t
5252 bnxt_get_vport(uint16_t port_id)
5253 {
5254 	return (1 << bnxt_get_phy_port_id(port_id));
5255 }
5256 
5257 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5258 {
5259 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5260 
5261 	if (info) {
5262 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5263 			memset(info, 0, sizeof(*info));
5264 		return;
5265 	}
5266 
5267 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5268 		return;
5269 
5270 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5271 			   sizeof(*info), 0);
5272 	if (!info)
5273 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5274 
5275 	bp->recovery_info = info;
5276 }
5277 
5278 static void bnxt_check_fw_status(struct bnxt *bp)
5279 {
5280 	uint32_t fw_status;
5281 
5282 	if (!(bp->recovery_info &&
5283 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5284 		return;
5285 
5286 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5287 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5288 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5289 			    fw_status);
5290 }
5291 
5292 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5293 {
5294 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5295 	uint32_t status_loc;
5296 	uint32_t sig_ver;
5297 
5298 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5299 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5300 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5301 				   BNXT_GRCP_WINDOW_2_BASE +
5302 				   offsetof(struct hcomm_status,
5303 					    sig_ver)));
5304 	/* If the signature is absent, then FW does not support this feature */
5305 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5306 	    HCOMM_STATUS_SIGNATURE_VAL)
5307 		return 0;
5308 
5309 	if (!info) {
5310 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5311 				   sizeof(*info), 0);
5312 		if (!info)
5313 			return -ENOMEM;
5314 		bp->recovery_info = info;
5315 	} else {
5316 		memset(info, 0, sizeof(*info));
5317 	}
5318 
5319 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5320 				      BNXT_GRCP_WINDOW_2_BASE +
5321 				      offsetof(struct hcomm_status,
5322 					       fw_status_loc)));
5323 
5324 	/* Only pre-map the FW health status GRC register */
5325 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5326 		return 0;
5327 
5328 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5329 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5330 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5331 
5332 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5333 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5334 
5335 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5336 
5337 	return 0;
5338 }
5339 
5340 /* This function gets the FW version along with the
5341  * capabilities(MAX and current) of the function, vnic,
5342  * error recovery, phy and other chip related info
5343  */
5344 static int bnxt_get_config(struct bnxt *bp)
5345 {
5346 	uint16_t mtu;
5347 	int rc = 0;
5348 
5349 	bp->fw_cap = 0;
5350 
5351 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5352 	if (rc)
5353 		return rc;
5354 
5355 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5356 	if (rc) {
5357 		bnxt_check_fw_status(bp);
5358 		return rc;
5359 	}
5360 
5361 	rc = bnxt_hwrm_func_reset(bp);
5362 	if (rc)
5363 		return -EIO;
5364 
5365 	rc = bnxt_hwrm_vnic_qcaps(bp);
5366 	if (rc)
5367 		return rc;
5368 
5369 	rc = bnxt_hwrm_queue_qportcfg(bp);
5370 	if (rc)
5371 		return rc;
5372 
5373 	/* Get the MAX capabilities for this function.
5374 	 * This function also allocates context memory for TQM rings and
5375 	 * informs the firmware about this allocated backing store memory.
5376 	 */
5377 	rc = bnxt_hwrm_func_qcaps(bp);
5378 	if (rc)
5379 		return rc;
5380 
5381 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5382 	if (rc)
5383 		return rc;
5384 
5385 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5386 	if (rc)
5387 		return rc;
5388 
5389 	bnxt_hwrm_port_mac_qcfg(bp);
5390 
5391 	bnxt_hwrm_parent_pf_qcfg(bp);
5392 
5393 	bnxt_hwrm_port_phy_qcaps(bp);
5394 
5395 	bnxt_alloc_error_recovery_info(bp);
5396 	/* Get the adapter error recovery support info */
5397 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5398 	if (rc)
5399 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5400 
5401 	bnxt_hwrm_port_led_qcaps(bp);
5402 
5403 	return 0;
5404 }
5405 
5406 static int
5407 bnxt_init_locks(struct bnxt *bp)
5408 {
5409 	int err;
5410 
5411 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5412 	if (err) {
5413 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5414 		return err;
5415 	}
5416 
5417 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5418 	if (err) {
5419 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5420 		return err;
5421 	}
5422 
5423 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5424 	if (err) {
5425 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5426 		return err;
5427 	}
5428 
5429 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5430 	if (err)
5431 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5432 
5433 	return err;
5434 }
5435 
5436 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5437 {
5438 	int rc = 0;
5439 
5440 	rc = bnxt_get_config(bp);
5441 	if (rc)
5442 		return rc;
5443 
5444 	if (!reconfig_dev) {
5445 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5446 		if (rc)
5447 			return rc;
5448 	} else {
5449 		rc = bnxt_restore_dflt_mac(bp);
5450 		if (rc)
5451 			return rc;
5452 	}
5453 
5454 	bnxt_config_vf_req_fwd(bp);
5455 
5456 	rc = bnxt_hwrm_func_driver_register(bp);
5457 	if (rc) {
5458 		PMD_DRV_LOG(ERR, "Failed to register driver");
5459 		return -EBUSY;
5460 	}
5461 
5462 	if (BNXT_PF(bp)) {
5463 		if (bp->pdev->max_vfs) {
5464 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5465 			if (rc) {
5466 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5467 				return rc;
5468 			}
5469 		} else {
5470 			rc = bnxt_hwrm_allocate_pf_only(bp);
5471 			if (rc) {
5472 				PMD_DRV_LOG(ERR,
5473 					    "Failed to allocate PF resources");
5474 				return rc;
5475 			}
5476 		}
5477 	}
5478 
5479 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5480 	if (rc)
5481 		return rc;
5482 
5483 	rc = bnxt_setup_int(bp);
5484 	if (rc)
5485 		return rc;
5486 
5487 	rc = bnxt_request_int(bp);
5488 	if (rc)
5489 		return rc;
5490 
5491 	rc = bnxt_init_ctx_mem(bp);
5492 	if (rc) {
5493 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5494 		return rc;
5495 	}
5496 
5497 	return 0;
5498 }
5499 
5500 static int
5501 bnxt_parse_devarg_accum_stats(__rte_unused const char *key,
5502 			      const char *value, void *opaque_arg)
5503 {
5504 	struct bnxt *bp = opaque_arg;
5505 	unsigned long accum_stats;
5506 	char *end = NULL;
5507 
5508 	if (!value || !opaque_arg) {
5509 		PMD_DRV_LOG(ERR,
5510 			    "Invalid parameter passed to accum-stats devargs.\n");
5511 		return -EINVAL;
5512 	}
5513 
5514 	accum_stats = strtoul(value, &end, 10);
5515 	if (end == NULL || *end != '\0' ||
5516 	    (accum_stats == ULONG_MAX && errno == ERANGE)) {
5517 		PMD_DRV_LOG(ERR,
5518 			    "Invalid parameter passed to accum-stats devargs.\n");
5519 		return -EINVAL;
5520 	}
5521 
5522 	if (BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)) {
5523 		PMD_DRV_LOG(ERR,
5524 			    "Invalid value passed to accum-stats devargs.\n");
5525 		return -EINVAL;
5526 	}
5527 
5528 	if (accum_stats) {
5529 		bp->flags2 |= BNXT_FLAGS2_ACCUM_STATS_EN;
5530 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature enabled.\n");
5531 	} else {
5532 		bp->flags2 &= ~BNXT_FLAGS2_ACCUM_STATS_EN;
5533 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature disabled.\n");
5534 	}
5535 
5536 	return 0;
5537 }
5538 
5539 static int
5540 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5541 			     const char *value, void *opaque_arg)
5542 {
5543 	struct bnxt *bp = opaque_arg;
5544 	unsigned long flow_xstat;
5545 	char *end = NULL;
5546 
5547 	if (!value || !opaque_arg) {
5548 		PMD_DRV_LOG(ERR,
5549 			    "Invalid parameter passed to flow_xstat devarg.\n");
5550 		return -EINVAL;
5551 	}
5552 
5553 	flow_xstat = strtoul(value, &end, 10);
5554 	if (end == NULL || *end != '\0' ||
5555 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5556 		PMD_DRV_LOG(ERR,
5557 			    "Invalid parameter passed to flow_xstat devarg.\n");
5558 		return -EINVAL;
5559 	}
5560 
5561 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5562 		PMD_DRV_LOG(ERR,
5563 			    "Invalid value passed to flow_xstat devarg.\n");
5564 		return -EINVAL;
5565 	}
5566 
5567 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5568 	if (BNXT_FLOW_XSTATS_EN(bp))
5569 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5570 
5571 	return 0;
5572 }
5573 
5574 static int
5575 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5576 					const char *value, void *opaque_arg)
5577 {
5578 	struct bnxt *bp = opaque_arg;
5579 	unsigned long max_num_kflows;
5580 	char *end = NULL;
5581 
5582 	if (!value || !opaque_arg) {
5583 		PMD_DRV_LOG(ERR,
5584 			"Invalid parameter passed to max_num_kflows devarg.\n");
5585 		return -EINVAL;
5586 	}
5587 
5588 	max_num_kflows = strtoul(value, &end, 10);
5589 	if (end == NULL || *end != '\0' ||
5590 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5591 		PMD_DRV_LOG(ERR,
5592 			"Invalid parameter passed to max_num_kflows devarg.\n");
5593 		return -EINVAL;
5594 	}
5595 
5596 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5597 		PMD_DRV_LOG(ERR,
5598 			"Invalid value passed to max_num_kflows devarg.\n");
5599 		return -EINVAL;
5600 	}
5601 
5602 	bp->max_num_kflows = max_num_kflows;
5603 	if (bp->max_num_kflows)
5604 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5605 				max_num_kflows);
5606 
5607 	return 0;
5608 }
5609 
5610 static int
5611 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5612 				 const char *value, void *opaque_arg)
5613 {
5614 	struct bnxt *bp = opaque_arg;
5615 	unsigned long app_id;
5616 	char *end = NULL;
5617 
5618 	if (!value || !opaque_arg) {
5619 		PMD_DRV_LOG(ERR,
5620 			    "Invalid parameter passed to app-id "
5621 			    "devargs.\n");
5622 		return -EINVAL;
5623 	}
5624 
5625 	app_id = strtoul(value, &end, 10);
5626 	if (end == NULL || *end != '\0' ||
5627 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5628 		PMD_DRV_LOG(ERR,
5629 			    "Invalid parameter passed to app_id "
5630 			    "devargs.\n");
5631 		return -EINVAL;
5632 	}
5633 
5634 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5635 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5636 			    (uint16_t)app_id);
5637 		return -EINVAL;
5638 	}
5639 
5640 	bp->app_id = app_id;
5641 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5642 
5643 	return 0;
5644 }
5645 
5646 static int
5647 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5648 			    const char *value, void *opaque_arg)
5649 {
5650 	struct bnxt_representor *vfr_bp = opaque_arg;
5651 	unsigned long rep_is_pf;
5652 	char *end = NULL;
5653 
5654 	if (!value || !opaque_arg) {
5655 		PMD_DRV_LOG(ERR,
5656 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5657 		return -EINVAL;
5658 	}
5659 
5660 	rep_is_pf = strtoul(value, &end, 10);
5661 	if (end == NULL || *end != '\0' ||
5662 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5663 		PMD_DRV_LOG(ERR,
5664 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5665 		return -EINVAL;
5666 	}
5667 
5668 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5669 		PMD_DRV_LOG(ERR,
5670 			    "Invalid value passed to rep_is_pf devargs.\n");
5671 		return -EINVAL;
5672 	}
5673 
5674 	vfr_bp->flags |= rep_is_pf;
5675 	if (BNXT_REP_PF(vfr_bp))
5676 		PMD_DRV_LOG(INFO, "PF representor\n");
5677 	else
5678 		PMD_DRV_LOG(INFO, "VF representor\n");
5679 
5680 	return 0;
5681 }
5682 
5683 static int
5684 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5685 			       const char *value, void *opaque_arg)
5686 {
5687 	struct bnxt_representor *vfr_bp = opaque_arg;
5688 	unsigned long rep_based_pf;
5689 	char *end = NULL;
5690 
5691 	if (!value || !opaque_arg) {
5692 		PMD_DRV_LOG(ERR,
5693 			    "Invalid parameter passed to rep_based_pf "
5694 			    "devargs.\n");
5695 		return -EINVAL;
5696 	}
5697 
5698 	rep_based_pf = strtoul(value, &end, 10);
5699 	if (end == NULL || *end != '\0' ||
5700 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5701 		PMD_DRV_LOG(ERR,
5702 			    "Invalid parameter passed to rep_based_pf "
5703 			    "devargs.\n");
5704 		return -EINVAL;
5705 	}
5706 
5707 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5708 		PMD_DRV_LOG(ERR,
5709 			    "Invalid value passed to rep_based_pf devargs.\n");
5710 		return -EINVAL;
5711 	}
5712 
5713 	vfr_bp->rep_based_pf = rep_based_pf;
5714 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5715 
5716 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5717 
5718 	return 0;
5719 }
5720 
5721 static int
5722 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5723 			    const char *value, void *opaque_arg)
5724 {
5725 	struct bnxt_representor *vfr_bp = opaque_arg;
5726 	unsigned long rep_q_r2f;
5727 	char *end = NULL;
5728 
5729 	if (!value || !opaque_arg) {
5730 		PMD_DRV_LOG(ERR,
5731 			    "Invalid parameter passed to rep_q_r2f "
5732 			    "devargs.\n");
5733 		return -EINVAL;
5734 	}
5735 
5736 	rep_q_r2f = strtoul(value, &end, 10);
5737 	if (end == NULL || *end != '\0' ||
5738 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5739 		PMD_DRV_LOG(ERR,
5740 			    "Invalid parameter passed to rep_q_r2f "
5741 			    "devargs.\n");
5742 		return -EINVAL;
5743 	}
5744 
5745 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5746 		PMD_DRV_LOG(ERR,
5747 			    "Invalid value passed to rep_q_r2f devargs.\n");
5748 		return -EINVAL;
5749 	}
5750 
5751 	vfr_bp->rep_q_r2f = rep_q_r2f;
5752 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5753 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5754 
5755 	return 0;
5756 }
5757 
5758 static int
5759 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5760 			    const char *value, void *opaque_arg)
5761 {
5762 	struct bnxt_representor *vfr_bp = opaque_arg;
5763 	unsigned long rep_q_f2r;
5764 	char *end = NULL;
5765 
5766 	if (!value || !opaque_arg) {
5767 		PMD_DRV_LOG(ERR,
5768 			    "Invalid parameter passed to rep_q_f2r "
5769 			    "devargs.\n");
5770 		return -EINVAL;
5771 	}
5772 
5773 	rep_q_f2r = strtoul(value, &end, 10);
5774 	if (end == NULL || *end != '\0' ||
5775 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5776 		PMD_DRV_LOG(ERR,
5777 			    "Invalid parameter passed to rep_q_f2r "
5778 			    "devargs.\n");
5779 		return -EINVAL;
5780 	}
5781 
5782 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5783 		PMD_DRV_LOG(ERR,
5784 			    "Invalid value passed to rep_q_f2r devargs.\n");
5785 		return -EINVAL;
5786 	}
5787 
5788 	vfr_bp->rep_q_f2r = rep_q_f2r;
5789 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5790 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5791 
5792 	return 0;
5793 }
5794 
5795 static int
5796 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5797 			     const char *value, void *opaque_arg)
5798 {
5799 	struct bnxt_representor *vfr_bp = opaque_arg;
5800 	unsigned long rep_fc_r2f;
5801 	char *end = NULL;
5802 
5803 	if (!value || !opaque_arg) {
5804 		PMD_DRV_LOG(ERR,
5805 			    "Invalid parameter passed to rep_fc_r2f "
5806 			    "devargs.\n");
5807 		return -EINVAL;
5808 	}
5809 
5810 	rep_fc_r2f = strtoul(value, &end, 10);
5811 	if (end == NULL || *end != '\0' ||
5812 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5813 		PMD_DRV_LOG(ERR,
5814 			    "Invalid parameter passed to rep_fc_r2f "
5815 			    "devargs.\n");
5816 		return -EINVAL;
5817 	}
5818 
5819 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5820 		PMD_DRV_LOG(ERR,
5821 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5822 		return -EINVAL;
5823 	}
5824 
5825 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5826 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5827 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5828 
5829 	return 0;
5830 }
5831 
5832 static int
5833 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5834 			     const char *value, void *opaque_arg)
5835 {
5836 	struct bnxt_representor *vfr_bp = opaque_arg;
5837 	unsigned long rep_fc_f2r;
5838 	char *end = NULL;
5839 
5840 	if (!value || !opaque_arg) {
5841 		PMD_DRV_LOG(ERR,
5842 			    "Invalid parameter passed to rep_fc_f2r "
5843 			    "devargs.\n");
5844 		return -EINVAL;
5845 	}
5846 
5847 	rep_fc_f2r = strtoul(value, &end, 10);
5848 	if (end == NULL || *end != '\0' ||
5849 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5850 		PMD_DRV_LOG(ERR,
5851 			    "Invalid parameter passed to rep_fc_f2r "
5852 			    "devargs.\n");
5853 		return -EINVAL;
5854 	}
5855 
5856 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5857 		PMD_DRV_LOG(ERR,
5858 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5859 		return -EINVAL;
5860 	}
5861 
5862 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5863 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5864 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5865 
5866 	return 0;
5867 }
5868 
5869 static int
5870 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5871 {
5872 	struct rte_kvargs *kvlist;
5873 	int ret;
5874 
5875 	if (devargs == NULL)
5876 		return 0;
5877 
5878 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5879 	if (kvlist == NULL)
5880 		return -EINVAL;
5881 
5882 	/*
5883 	 * Handler for "flow_xstat" devarg.
5884 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5885 	 */
5886 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5887 				 bnxt_parse_devarg_flow_xstat, bp);
5888 	if (ret)
5889 		goto err;
5890 
5891 	/*
5892 	 * Handler for "accum-stats" devarg.
5893 	 * Invoked as for ex: "-a 0000:00:0d.0,accum-stats=1"
5894 	 */
5895 	rte_kvargs_process(kvlist, BNXT_DEVARG_ACCUM_STATS,
5896 			   bnxt_parse_devarg_accum_stats, bp);
5897 	/*
5898 	 * Handler for "max_num_kflows" devarg.
5899 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5900 	 */
5901 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5902 				 bnxt_parse_devarg_max_num_kflows, bp);
5903 	if (ret)
5904 		goto err;
5905 
5906 err:
5907 	/*
5908 	 * Handler for "app-id" devarg.
5909 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5910 	 */
5911 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5912 			   bnxt_parse_devarg_app_id, bp);
5913 
5914 	rte_kvargs_free(kvlist);
5915 	return ret;
5916 }
5917 
5918 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5919 {
5920 	int rc = 0;
5921 
5922 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5923 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5924 		if (rc)
5925 			PMD_DRV_LOG(ERR,
5926 				    "Failed to alloc switch domain: %d\n", rc);
5927 		else
5928 			PMD_DRV_LOG(INFO,
5929 				    "Switch domain allocated %d\n",
5930 				    bp->switch_domain_id);
5931 	}
5932 
5933 	return rc;
5934 }
5935 
5936 /* Allocate and initialize various fields in bnxt struct that
5937  * need to be allocated/destroyed only once in the lifetime of the driver
5938  */
5939 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5940 {
5941 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5942 	struct bnxt *bp = eth_dev->data->dev_private;
5943 	int rc = 0;
5944 
5945 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5946 
5947 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5948 		bp->flags |= BNXT_FLAG_VF;
5949 
5950 	if (bnxt_p5_device(pci_dev->id.device_id))
5951 		bp->flags |= BNXT_FLAG_CHIP_P5;
5952 
5953 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5954 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5955 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5956 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5957 		bp->flags |= BNXT_FLAG_STINGRAY;
5958 
5959 	if (BNXT_TRUFLOW_EN(bp)) {
5960 		/* extra mbuf field is required to store CFA code from mark */
5961 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5962 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5963 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5964 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5965 		};
5966 		bnxt_cfa_code_dynfield_offset =
5967 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5968 		if (bnxt_cfa_code_dynfield_offset < 0) {
5969 			PMD_DRV_LOG(ERR,
5970 			    "Failed to register mbuf field for TruFlow mark\n");
5971 			return -rte_errno;
5972 		}
5973 	}
5974 
5975 	rc = bnxt_map_pci_bars(eth_dev);
5976 	if (rc) {
5977 		PMD_DRV_LOG(ERR,
5978 			    "Failed to initialize board rc: %x\n", rc);
5979 		return rc;
5980 	}
5981 
5982 	rc = bnxt_alloc_pf_info(bp);
5983 	if (rc)
5984 		return rc;
5985 
5986 	rc = bnxt_alloc_link_info(bp);
5987 	if (rc)
5988 		return rc;
5989 
5990 	rc = bnxt_alloc_parent_info(bp);
5991 	if (rc)
5992 		return rc;
5993 
5994 	rc = bnxt_alloc_hwrm_resources(bp);
5995 	if (rc) {
5996 		PMD_DRV_LOG(ERR,
5997 			    "Failed to allocate response buffer rc: %x\n", rc);
5998 		return rc;
5999 	}
6000 	rc = bnxt_alloc_leds_info(bp);
6001 	if (rc)
6002 		return rc;
6003 
6004 	rc = bnxt_alloc_cos_queues(bp);
6005 	if (rc)
6006 		return rc;
6007 
6008 	rc = bnxt_init_locks(bp);
6009 	if (rc)
6010 		return rc;
6011 
6012 	rc = bnxt_alloc_switch_domain(bp);
6013 	if (rc)
6014 		return rc;
6015 
6016 	return rc;
6017 }
6018 
6019 static int
6020 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
6021 {
6022 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
6023 	static int version_printed;
6024 	struct bnxt *bp;
6025 	int rc;
6026 
6027 	if (version_printed++ == 0)
6028 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
6029 
6030 	eth_dev->dev_ops = &bnxt_dev_ops;
6031 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
6032 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
6033 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
6034 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
6035 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
6036 
6037 	/*
6038 	 * For secondary processes, we don't initialise any further
6039 	 * as primary has already done this work.
6040 	 */
6041 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6042 		return 0;
6043 
6044 	rte_eth_copy_pci_info(eth_dev, pci_dev);
6045 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
6046 
6047 	bp = eth_dev->data->dev_private;
6048 
6049 	/* Parse dev arguments passed on when starting the DPDK application. */
6050 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
6051 	if (rc)
6052 		goto error_free;
6053 
6054 	rc = bnxt_drv_init(eth_dev);
6055 	if (rc)
6056 		goto error_free;
6057 
6058 	rc = bnxt_init_resources(bp, false);
6059 	if (rc)
6060 		goto error_free;
6061 
6062 	rc = bnxt_alloc_stats_mem(bp);
6063 	if (rc)
6064 		goto error_free;
6065 
6066 	PMD_DRV_LOG(INFO,
6067 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
6068 		    DRV_MODULE_NAME,
6069 		    pci_dev->mem_resource[0].phys_addr,
6070 		    pci_dev->mem_resource[0].addr);
6071 
6072 	return 0;
6073 
6074 error_free:
6075 	bnxt_dev_uninit(eth_dev);
6076 	return rc;
6077 }
6078 
6079 
6080 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
6081 {
6082 	if (!ctx)
6083 		return;
6084 
6085 	if (ctx->va)
6086 		rte_free(ctx->va);
6087 
6088 	ctx->va = NULL;
6089 	ctx->dma = RTE_BAD_IOVA;
6090 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
6091 }
6092 
6093 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
6094 {
6095 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
6096 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6097 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
6098 				  bp->flow_stat->max_fc,
6099 				  false);
6100 
6101 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
6102 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6103 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
6104 				  bp->flow_stat->max_fc,
6105 				  false);
6106 
6107 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6108 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
6109 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6110 
6111 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6112 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
6113 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6114 
6115 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6116 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
6117 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6118 
6119 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6120 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
6121 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6122 }
6123 
6124 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
6125 {
6126 	bnxt_unregister_fc_ctx_mem(bp);
6127 
6128 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
6129 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
6130 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
6131 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
6132 }
6133 
6134 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
6135 {
6136 	if (BNXT_FLOW_XSTATS_EN(bp))
6137 		bnxt_uninit_fc_ctx_mem(bp);
6138 }
6139 
6140 static void
6141 bnxt_free_error_recovery_info(struct bnxt *bp)
6142 {
6143 	rte_free(bp->recovery_info);
6144 	bp->recovery_info = NULL;
6145 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
6146 }
6147 
6148 static int
6149 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
6150 {
6151 	int rc;
6152 
6153 	bnxt_free_int(bp);
6154 	bnxt_free_mem(bp, reconfig_dev);
6155 
6156 	bnxt_hwrm_func_buf_unrgtr(bp);
6157 	if (bp->pf != NULL) {
6158 		rte_free(bp->pf->vf_req_buf);
6159 		bp->pf->vf_req_buf = NULL;
6160 	}
6161 
6162 	rc = bnxt_hwrm_func_driver_unregister(bp);
6163 	bp->flags &= ~BNXT_FLAG_REGISTERED;
6164 	bnxt_free_ctx_mem(bp);
6165 	if (!reconfig_dev) {
6166 		bnxt_free_hwrm_resources(bp);
6167 		bnxt_free_error_recovery_info(bp);
6168 	}
6169 
6170 	bnxt_uninit_ctx_mem(bp);
6171 
6172 	bnxt_free_flow_stats_info(bp);
6173 	if (bp->rep_info != NULL)
6174 		bnxt_free_switch_domain(bp);
6175 	bnxt_free_rep_info(bp);
6176 	rte_free(bp->ptp_cfg);
6177 	bp->ptp_cfg = NULL;
6178 	return rc;
6179 }
6180 
6181 static int
6182 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
6183 {
6184 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6185 		return -EPERM;
6186 
6187 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
6188 
6189 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
6190 		bnxt_dev_close_op(eth_dev);
6191 
6192 	return 0;
6193 }
6194 
6195 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6196 {
6197 	struct bnxt *bp = eth_dev->data->dev_private;
6198 	struct rte_eth_dev *vf_rep_eth_dev;
6199 	int ret = 0, i;
6200 
6201 	if (!bp)
6202 		return -EINVAL;
6203 
6204 	for (i = 0; i < bp->num_reps; i++) {
6205 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6206 		if (!vf_rep_eth_dev)
6207 			continue;
6208 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6209 			    vf_rep_eth_dev->data->port_id);
6210 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6211 	}
6212 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6213 		    eth_dev->data->port_id);
6214 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6215 
6216 	return ret;
6217 }
6218 
6219 static void bnxt_free_rep_info(struct bnxt *bp)
6220 {
6221 	rte_free(bp->rep_info);
6222 	bp->rep_info = NULL;
6223 	rte_free(bp->cfa_code_map);
6224 	bp->cfa_code_map = NULL;
6225 }
6226 
6227 static int bnxt_init_rep_info(struct bnxt *bp)
6228 {
6229 	int i = 0, rc;
6230 
6231 	if (bp->rep_info)
6232 		return 0;
6233 
6234 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
6235 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
6236 				   0);
6237 	if (!bp->rep_info) {
6238 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6239 		return -ENOMEM;
6240 	}
6241 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6242 				       sizeof(*bp->cfa_code_map) *
6243 				       BNXT_MAX_CFA_CODE, 0);
6244 	if (!bp->cfa_code_map) {
6245 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6246 		bnxt_free_rep_info(bp);
6247 		return -ENOMEM;
6248 	}
6249 
6250 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6251 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6252 
6253 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6254 	if (rc) {
6255 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6256 		bnxt_free_rep_info(bp);
6257 		return rc;
6258 	}
6259 
6260 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6261 	if (rc) {
6262 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6263 		bnxt_free_rep_info(bp);
6264 		return rc;
6265 	}
6266 
6267 	return rc;
6268 }
6269 
6270 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6271 			       struct rte_eth_devargs *eth_da,
6272 			       struct rte_eth_dev *backing_eth_dev,
6273 			       const char *dev_args)
6274 {
6275 	struct rte_eth_dev *vf_rep_eth_dev;
6276 	char name[RTE_ETH_NAME_MAX_LEN];
6277 	struct bnxt *backing_bp;
6278 	uint16_t num_rep;
6279 	int i, ret = 0;
6280 	struct rte_kvargs *kvlist = NULL;
6281 
6282 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6283 		return 0;
6284 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6285 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6286 			    eth_da->type);
6287 		return -ENOTSUP;
6288 	}
6289 	num_rep = eth_da->nb_representor_ports;
6290 	if (num_rep > BNXT_MAX_VF_REPS) {
6291 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6292 			    num_rep, BNXT_MAX_VF_REPS);
6293 		return -EINVAL;
6294 	}
6295 
6296 	if (num_rep >= RTE_MAX_ETHPORTS) {
6297 		PMD_DRV_LOG(ERR,
6298 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6299 			    num_rep, RTE_MAX_ETHPORTS);
6300 		return -EINVAL;
6301 	}
6302 
6303 	backing_bp = backing_eth_dev->data->dev_private;
6304 
6305 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6306 		PMD_DRV_LOG(ERR,
6307 			    "Not a PF or trusted VF. No Representor support\n");
6308 		/* Returning an error is not an option.
6309 		 * Applications are not handling this correctly
6310 		 */
6311 		return 0;
6312 	}
6313 
6314 	if (bnxt_init_rep_info(backing_bp))
6315 		return 0;
6316 
6317 	for (i = 0; i < num_rep; i++) {
6318 		struct bnxt_representor representor = {
6319 			.vf_id = eth_da->representor_ports[i],
6320 			.switch_domain_id = backing_bp->switch_domain_id,
6321 			.parent_dev = backing_eth_dev
6322 		};
6323 
6324 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
6325 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6326 				    representor.vf_id, BNXT_MAX_VF_REPS);
6327 			continue;
6328 		}
6329 
6330 		/* representor port net_bdf_port */
6331 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6332 			 pci_dev->device.name, eth_da->representor_ports[i]);
6333 
6334 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6335 		if (kvlist) {
6336 			/*
6337 			 * Handler for "rep_is_pf" devarg.
6338 			 * Invoked as for ex: "-a 000:00:0d.0,
6339 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6340 			 */
6341 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6342 						 bnxt_parse_devarg_rep_is_pf,
6343 						 (void *)&representor);
6344 			if (ret) {
6345 				ret = -EINVAL;
6346 				goto err;
6347 			}
6348 			/*
6349 			 * Handler for "rep_based_pf" devarg.
6350 			 * Invoked as for ex: "-a 000:00:0d.0,
6351 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6352 			 */
6353 			ret = rte_kvargs_process(kvlist,
6354 						 BNXT_DEVARG_REP_BASED_PF,
6355 						 bnxt_parse_devarg_rep_based_pf,
6356 						 (void *)&representor);
6357 			if (ret) {
6358 				ret = -EINVAL;
6359 				goto err;
6360 			}
6361 			/*
6362 			 * Handler for "rep_based_pf" devarg.
6363 			 * Invoked as for ex: "-a 000:00:0d.0,
6364 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6365 			 */
6366 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6367 						 bnxt_parse_devarg_rep_q_r2f,
6368 						 (void *)&representor);
6369 			if (ret) {
6370 				ret = -EINVAL;
6371 				goto err;
6372 			}
6373 			/*
6374 			 * Handler for "rep_based_pf" devarg.
6375 			 * Invoked as for ex: "-a 000:00:0d.0,
6376 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6377 			 */
6378 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6379 						 bnxt_parse_devarg_rep_q_f2r,
6380 						 (void *)&representor);
6381 			if (ret) {
6382 				ret = -EINVAL;
6383 				goto err;
6384 			}
6385 			/*
6386 			 * Handler for "rep_based_pf" devarg.
6387 			 * Invoked as for ex: "-a 000:00:0d.0,
6388 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6389 			 */
6390 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6391 						 bnxt_parse_devarg_rep_fc_r2f,
6392 						 (void *)&representor);
6393 			if (ret) {
6394 				ret = -EINVAL;
6395 				goto err;
6396 			}
6397 			/*
6398 			 * Handler for "rep_based_pf" devarg.
6399 			 * Invoked as for ex: "-a 000:00:0d.0,
6400 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6401 			 */
6402 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6403 						 bnxt_parse_devarg_rep_fc_f2r,
6404 						 (void *)&representor);
6405 			if (ret) {
6406 				ret = -EINVAL;
6407 				goto err;
6408 			}
6409 		}
6410 
6411 		ret = rte_eth_dev_create(&pci_dev->device, name,
6412 					 sizeof(struct bnxt_representor),
6413 					 NULL, NULL,
6414 					 bnxt_representor_init,
6415 					 &representor);
6416 		if (ret) {
6417 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6418 				    "representor %s.", name);
6419 			goto err;
6420 		}
6421 
6422 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6423 		if (!vf_rep_eth_dev) {
6424 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6425 				    " for VF-Rep: %s.", name);
6426 			ret = -ENODEV;
6427 			goto err;
6428 		}
6429 
6430 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6431 			    backing_eth_dev->data->port_id);
6432 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6433 							 vf_rep_eth_dev;
6434 		backing_bp->num_reps++;
6435 
6436 	}
6437 
6438 	rte_kvargs_free(kvlist);
6439 	return 0;
6440 
6441 err:
6442 	/* If num_rep > 1, then rollback already created
6443 	 * ports, since we'll be failing the probe anyway
6444 	 */
6445 	if (num_rep > 1)
6446 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6447 	rte_errno = -ret;
6448 	rte_kvargs_free(kvlist);
6449 
6450 	return ret;
6451 }
6452 
6453 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6454 			  struct rte_pci_device *pci_dev)
6455 {
6456 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6457 	struct rte_eth_dev *backing_eth_dev;
6458 	uint16_t num_rep;
6459 	int ret = 0;
6460 
6461 	if (pci_dev->device.devargs) {
6462 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6463 					    &eth_da);
6464 		if (ret)
6465 			return ret;
6466 	}
6467 
6468 	num_rep = eth_da.nb_representor_ports;
6469 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6470 		    num_rep);
6471 
6472 	/* We could come here after first level of probe is already invoked
6473 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6474 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6475 	 */
6476 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6477 	if (backing_eth_dev == NULL) {
6478 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6479 					 sizeof(struct bnxt),
6480 					 eth_dev_pci_specific_init, pci_dev,
6481 					 bnxt_dev_init, NULL);
6482 
6483 		if (ret || !num_rep)
6484 			return ret;
6485 
6486 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6487 	}
6488 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6489 		    backing_eth_dev->data->port_id);
6490 
6491 	if (!num_rep)
6492 		return ret;
6493 
6494 	/* probe representor ports now */
6495 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6496 				  pci_dev->device.devargs->args);
6497 
6498 	return ret;
6499 }
6500 
6501 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6502 {
6503 	struct rte_eth_dev *eth_dev;
6504 
6505 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6506 	if (!eth_dev)
6507 		return 0; /* Invoked typically only by OVS-DPDK, by the
6508 			   * time it comes here the eth_dev is already
6509 			   * deleted by rte_eth_dev_close(), so returning
6510 			   * +ve value will at least help in proper cleanup
6511 			   */
6512 
6513 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6514 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6515 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6516 			return rte_eth_dev_destroy(eth_dev,
6517 						   bnxt_representor_uninit);
6518 		else
6519 			return rte_eth_dev_destroy(eth_dev,
6520 						   bnxt_dev_uninit);
6521 	} else {
6522 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6523 	}
6524 }
6525 
6526 static struct rte_pci_driver bnxt_rte_pmd = {
6527 	.id_table = bnxt_pci_id_map,
6528 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6529 			RTE_PCI_DRV_INTR_RMV |
6530 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6531 						  * and OVS-DPDK
6532 						  */
6533 	.probe = bnxt_pci_probe,
6534 	.remove = bnxt_pci_remove,
6535 };
6536 
6537 static bool
6538 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6539 {
6540 	if (strcmp(dev->device->driver->name, drv->driver.name))
6541 		return false;
6542 
6543 	return true;
6544 }
6545 
6546 bool is_bnxt_supported(struct rte_eth_dev *dev)
6547 {
6548 	return is_device_supported(dev, &bnxt_rte_pmd);
6549 }
6550 
6551 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6552 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6553 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6554 
6555