xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision edc13af9a8bbdca92bd8974165df2ff7049f45b7)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2014-2023 Broadcom
3  * All rights reserved.
4  */
5 
6 #include <inttypes.h>
7 #include <stdbool.h>
8 
9 #include <dev_driver.h>
10 #include <ethdev_driver.h>
11 #include <ethdev_pci.h>
12 #include <rte_malloc.h>
13 #include <rte_cycles.h>
14 #include <rte_alarm.h>
15 #include <rte_kvargs.h>
16 #include <rte_vect.h>
17 
18 #include "bnxt.h"
19 #include "bnxt_filter.h"
20 #include "bnxt_hwrm.h"
21 #include "bnxt_irq.h"
22 #include "bnxt_reps.h"
23 #include "bnxt_ring.h"
24 #include "bnxt_rxq.h"
25 #include "bnxt_rxr.h"
26 #include "bnxt_stats.h"
27 #include "bnxt_txq.h"
28 #include "bnxt_txr.h"
29 #include "bnxt_vnic.h"
30 #include "hsi_struct_def_dpdk.h"
31 #include "bnxt_nvm_defs.h"
32 #include "bnxt_tf_common.h"
33 #include "ulp_flow_db.h"
34 #include "rte_pmd_bnxt.h"
35 
36 #define DRV_MODULE_NAME		"bnxt"
37 static const char bnxt_version[] =
38 	"Broadcom NetXtreme driver " DRV_MODULE_NAME;
39 
40 /*
41  * The set of PCI devices this driver supports
42  */
43 static const struct rte_pci_id bnxt_pci_id_map[] = {
44 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
45 			 BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
46 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
47 			 BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
48 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
49 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
50 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
51 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
52 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
53 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
54 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
55 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
56 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
57 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
58 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
59 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
60 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
61 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
62 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
63 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
64 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
65 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
66 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
67 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
68 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
69 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
70 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
71 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
72 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
73 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
74 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
75 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
76 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
77 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
78 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
79 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
80 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
81 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
82 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
83 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58812) },
84 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58814) },
85 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818) },
86 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818_VF) },
87 	{ .vendor_id = 0, /* sentinel */ },
88 };
89 
90 #define BNXT_DEVARG_FLOW_XSTAT	"flow-xstat"
91 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
92 #define BNXT_DEVARG_REPRESENTOR	"representor"
93 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
94 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
95 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
96 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
97 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
98 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
99 #define BNXT_DEVARG_APP_ID	"app-id"
100 #define BNXT_DEVARG_IEEE_1588	"ieee-1588"
101 
102 static const char *const bnxt_dev_args[] = {
103 	BNXT_DEVARG_REPRESENTOR,
104 	BNXT_DEVARG_FLOW_XSTAT,
105 	BNXT_DEVARG_MAX_NUM_KFLOWS,
106 	BNXT_DEVARG_REP_BASED_PF,
107 	BNXT_DEVARG_REP_IS_PF,
108 	BNXT_DEVARG_REP_Q_R2F,
109 	BNXT_DEVARG_REP_Q_F2R,
110 	BNXT_DEVARG_REP_FC_R2F,
111 	BNXT_DEVARG_REP_FC_F2R,
112 	BNXT_DEVARG_APP_ID,
113 	BNXT_DEVARG_IEEE_1588,
114 	NULL
115 };
116 
117 /*
118  * app-id = an non-negative 8-bit number
119  */
120 #define BNXT_DEVARG_APP_ID_INVALID(val)			((val) > 255)
121 
122 /*
123  * ieee-1588 = an non-negative 8-bit number
124  */
125 #define BNXT_DEVARG_IEEE_1588_INVALID(val)			((val) > 255)
126 
127 /*
128  * flow_xstat == false to disable the feature
129  * flow_xstat == true to enable the feature
130  */
131 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
132 
133 /*
134  * rep_is_pf == false to indicate VF representor
135  * rep_is_pf == true to indicate PF representor
136  */
137 #define	BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)	((rep_is_pf) > 1)
138 
139 /*
140  * rep_based_pf == Physical index of the PF
141  */
142 #define	BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)	((rep_based_pf) > 15)
143 /*
144  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
145  */
146 #define	BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)	((rep_q_r2f) > 3)
147 
148 /*
149  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
150  */
151 #define	BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)	((rep_q_f2r) > 3)
152 
153 /*
154  * rep_fc_r2f == Flow control for the representor to endpoint direction
155  */
156 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)	((rep_fc_r2f) > 1)
157 
158 /*
159  * rep_fc_f2r == Flow control for the endpoint to representor direction
160  */
161 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)	((rep_fc_f2r) > 1)
162 
163 int bnxt_cfa_code_dynfield_offset = -1;
164 
165 /*
166  * max_num_kflows must be >= 32
167  * and must be a power-of-2 supported value
168  * return: 1 -> invalid
169  *         0 -> valid
170  */
171 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
172 {
173 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
174 		return 1;
175 	return 0;
176 }
177 
178 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
179 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
180 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
181 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
182 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
183 static int bnxt_restore_vlan_filters(struct bnxt *bp);
184 static void bnxt_dev_recover(void *arg);
185 static void bnxt_free_error_recovery_info(struct bnxt *bp);
186 static void bnxt_free_rep_info(struct bnxt *bp);
187 static int bnxt_check_fw_ready(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 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, reconfig);
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 	/* populate the fw group table */
425 	bnxt_vnic_ring_grp_populate(bp, vnic);
426 	bnxt_vnic_rules_init(vnic);
427 
428 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
429 	if (rc)
430 		goto err_out;
431 
432 	/* Alloc RSS context only if RSS mode is enabled */
433 	if (dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS) {
434 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
435 
436 		/* RSS table size in P5 is 512.
437 		 * Cap max Rx rings to same value
438 		 */
439 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
440 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
441 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
442 			goto err_out;
443 		}
444 
445 		rc = 0;
446 		for (j = 0; j < nr_ctxs; j++) {
447 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
448 			if (rc)
449 				break;
450 		}
451 		if (rc) {
452 			PMD_DRV_LOG(ERR,
453 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
454 				    vnic_id, j, rc);
455 			goto err_out;
456 		}
457 		vnic->num_lb_ctxts = nr_ctxs;
458 	}
459 
460 	/*
461 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
462 	 * setting is not available at this time, it will not be
463 	 * configured correctly in the CFA.
464 	 */
465 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
466 		vnic->vlan_strip = true;
467 	else
468 		vnic->vlan_strip = false;
469 
470 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
471 	if (rc)
472 		goto err_out;
473 
474 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
475 	if (rc)
476 		goto err_out;
477 
478 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
479 		rxq = bp->eth_dev->data->rx_queues[j];
480 
481 		PMD_DRV_LOG(DEBUG,
482 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
483 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
484 
485 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
486 			vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
487 	}
488 
489 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
490 
491 	rc = bnxt_vnic_rss_configure(bp, vnic);
492 	if (rc)
493 		goto err_out;
494 
495 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
496 
497 	rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
498 				    (rx_offloads & RTE_ETH_RX_OFFLOAD_TCP_LRO) ?
499 				    true : false);
500 	if (rc)
501 		goto err_out;
502 
503 	return 0;
504 err_out:
505 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
506 		    vnic_id, rc);
507 	return rc;
508 }
509 
510 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
511 {
512 	int rc = 0;
513 
514 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
515 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
516 	if (rc)
517 		return rc;
518 
519 	PMD_DRV_LOG(DEBUG,
520 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
521 		    " rx_fc_in_tbl.ctx_id = %d\n",
522 		    bp->flow_stat->rx_fc_in_tbl.va,
523 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
524 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
525 
526 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
527 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
528 	if (rc)
529 		return rc;
530 
531 	PMD_DRV_LOG(DEBUG,
532 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
533 		    " rx_fc_out_tbl.ctx_id = %d\n",
534 		    bp->flow_stat->rx_fc_out_tbl.va,
535 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
536 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
537 
538 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
539 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
540 	if (rc)
541 		return rc;
542 
543 	PMD_DRV_LOG(DEBUG,
544 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
545 		    " tx_fc_in_tbl.ctx_id = %d\n",
546 		    bp->flow_stat->tx_fc_in_tbl.va,
547 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
548 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
549 
550 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
551 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
552 	if (rc)
553 		return rc;
554 
555 	PMD_DRV_LOG(DEBUG,
556 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
557 		    " tx_fc_out_tbl.ctx_id = %d\n",
558 		    bp->flow_stat->tx_fc_out_tbl.va,
559 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
560 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
561 
562 	memset(bp->flow_stat->rx_fc_out_tbl.va,
563 	       0,
564 	       bp->flow_stat->rx_fc_out_tbl.size);
565 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
566 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
567 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
568 				       bp->flow_stat->max_fc,
569 				       true);
570 	if (rc)
571 		return rc;
572 
573 	memset(bp->flow_stat->tx_fc_out_tbl.va,
574 	       0,
575 	       bp->flow_stat->tx_fc_out_tbl.size);
576 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
577 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
578 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
579 				       bp->flow_stat->max_fc,
580 				       true);
581 
582 	return rc;
583 }
584 
585 static int bnxt_alloc_ctx_mem_buf(struct bnxt *bp, char *type, size_t size,
586 				  struct bnxt_ctx_mem_buf_info *ctx)
587 {
588 	if (!ctx)
589 		return -EINVAL;
590 
591 	ctx->va = rte_zmalloc_socket(type, size, 0,
592 				     bp->eth_dev->device->numa_node);
593 	if (ctx->va == NULL)
594 		return -ENOMEM;
595 	rte_mem_lock_page(ctx->va);
596 	ctx->size = size;
597 	ctx->dma = rte_mem_virt2iova(ctx->va);
598 	if (ctx->dma == RTE_BAD_IOVA)
599 		return -ENOMEM;
600 
601 	return 0;
602 }
603 
604 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
605 {
606 	struct rte_pci_device *pdev = bp->pdev;
607 	char type[RTE_MEMZONE_NAMESIZE];
608 	uint16_t max_fc;
609 	int rc = 0;
610 
611 	max_fc = bp->flow_stat->max_fc;
612 
613 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
614 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
615 	/* 4 bytes for each counter-id */
616 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
617 				    max_fc * 4,
618 				    &bp->flow_stat->rx_fc_in_tbl);
619 	if (rc)
620 		return rc;
621 
622 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
623 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
624 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
625 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
626 				    max_fc * 16,
627 				    &bp->flow_stat->rx_fc_out_tbl);
628 	if (rc)
629 		return rc;
630 
631 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
632 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
633 	/* 4 bytes for each counter-id */
634 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
635 				    max_fc * 4,
636 				    &bp->flow_stat->tx_fc_in_tbl);
637 	if (rc)
638 		return rc;
639 
640 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
641 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
642 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
643 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
644 				    max_fc * 16,
645 				    &bp->flow_stat->tx_fc_out_tbl);
646 	if (rc)
647 		return rc;
648 
649 	rc = bnxt_register_fc_ctx_mem(bp);
650 
651 	return rc;
652 }
653 
654 static int bnxt_init_ctx_mem(struct bnxt *bp)
655 {
656 	int rc = 0;
657 
658 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
659 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
660 	    !BNXT_FLOW_XSTATS_EN(bp))
661 		return 0;
662 
663 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
664 	if (rc)
665 		return rc;
666 
667 	rc = bnxt_init_fc_ctx_mem(bp);
668 
669 	return rc;
670 }
671 
672 static inline bool bnxt_force_link_config(struct bnxt *bp)
673 {
674 	uint16_t subsystem_device_id = bp->pdev->id.subsystem_device_id;
675 
676 	switch (subsystem_device_id) {
677 	case BROADCOM_DEV_957508_N2100:
678 	case BROADCOM_DEV_957414_N225:
679 		return true;
680 	default:
681 		return false;
682 	}
683 }
684 
685 static int bnxt_update_phy_setting(struct bnxt *bp)
686 {
687 	struct rte_eth_link new;
688 	int rc;
689 
690 	rc = bnxt_get_hwrm_link_config(bp, &new);
691 	if (rc) {
692 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
693 		return rc;
694 	}
695 
696 	/*
697 	 * Device is not obliged link down in certain scenarios, even
698 	 * when forced. When FW does not allow any user other than BMC
699 	 * to shutdown the port, bnxt_get_hwrm_link_config() call always
700 	 * returns link up. Force phy update always in that case.
701 	 */
702 	if (!new.link_status || bnxt_force_link_config(bp)) {
703 		rc = bnxt_set_hwrm_link_config(bp, true);
704 		if (rc) {
705 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
706 			return rc;
707 		}
708 	}
709 
710 	return rc;
711 }
712 
713 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
714 {
715 	rte_free(bp->prev_rx_ring_stats);
716 	rte_free(bp->prev_tx_ring_stats);
717 
718 	bp->prev_rx_ring_stats = NULL;
719 	bp->prev_tx_ring_stats = NULL;
720 }
721 
722 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
723 {
724 	bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
725 					      sizeof(struct bnxt_ring_stats) *
726 					      bp->rx_cp_nr_rings,
727 					      0);
728 	if (bp->prev_rx_ring_stats == NULL)
729 		return -ENOMEM;
730 
731 	bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
732 					     sizeof(struct bnxt_ring_stats) *
733 					     bp->tx_cp_nr_rings,
734 					     0);
735 	if (bp->tx_cp_nr_rings > 0 && bp->prev_tx_ring_stats == NULL)
736 		goto error;
737 
738 	return 0;
739 
740 error:
741 	bnxt_free_prev_ring_stats(bp);
742 	return -ENOMEM;
743 }
744 
745 static int bnxt_start_nic(struct bnxt *bp)
746 {
747 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
748 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
749 	uint32_t intr_vector = 0;
750 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
751 	uint32_t vec = BNXT_MISC_VEC_ID;
752 	unsigned int i, j;
753 	int rc;
754 
755 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU)
756 		bp->flags |= BNXT_FLAG_JUMBO;
757 	else
758 		bp->flags &= ~BNXT_FLAG_JUMBO;
759 
760 	/* P5 does not support ring groups.
761 	 * But we will use the array to save RSS context IDs.
762 	 */
763 	if (BNXT_CHIP_P5(bp))
764 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
765 
766 	rc = bnxt_vnic_queue_db_init(bp);
767 	if (rc) {
768 		PMD_DRV_LOG(ERR, "could not allocate vnic db\n");
769 		goto err_out;
770 	}
771 
772 	rc = bnxt_alloc_hwrm_rings(bp);
773 	if (rc) {
774 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
775 		goto err_out;
776 	}
777 
778 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
779 	if (rc) {
780 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
781 		goto err_out;
782 	}
783 
784 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
785 		goto skip_cosq_cfg;
786 
787 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
788 		if (bp->rx_cos_queue[i].id != 0xff) {
789 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
790 
791 			if (!vnic) {
792 				PMD_DRV_LOG(ERR,
793 					    "Num pools more than FW profile\n");
794 				rc = -EINVAL;
795 				goto err_out;
796 			}
797 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
798 			bp->rx_cosq_cnt++;
799 		}
800 	}
801 
802 skip_cosq_cfg:
803 	rc = bnxt_mq_rx_configure(bp);
804 	if (rc) {
805 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
806 		goto err_out;
807 	}
808 
809 	for (j = 0; j < bp->rx_nr_rings; j++) {
810 		struct bnxt_rx_queue *rxq = bp->rx_queues[j];
811 
812 		if (!rxq->rx_deferred_start) {
813 			bp->eth_dev->data->rx_queue_state[j] =
814 				RTE_ETH_QUEUE_STATE_STARTED;
815 			rxq->rx_started = true;
816 		}
817 	}
818 
819 	/* setup the default vnic details*/
820 	bnxt_vnic_queue_db_update_dlft_vnic(bp);
821 
822 	/* VNIC configuration */
823 	for (i = 0; i < bp->nr_vnics; i++) {
824 		rc = bnxt_setup_one_vnic(bp, i);
825 		if (rc)
826 			goto err_out;
827 	}
828 
829 	for (j = 0; j < bp->tx_nr_rings; j++) {
830 		struct bnxt_tx_queue *txq = bp->tx_queues[j];
831 
832 		if (!txq->tx_deferred_start) {
833 			bp->eth_dev->data->tx_queue_state[j] =
834 				RTE_ETH_QUEUE_STATE_STARTED;
835 			txq->tx_started = true;
836 		}
837 	}
838 
839 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
840 	if (rc) {
841 		PMD_DRV_LOG(ERR,
842 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
843 		goto err_out;
844 	}
845 
846 	/* check and configure queue intr-vector mapping */
847 	if ((rte_intr_cap_multiple(intr_handle) ||
848 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
849 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
850 		intr_vector = bp->eth_dev->data->nb_rx_queues;
851 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
852 		if (intr_vector > bp->rx_cp_nr_rings) {
853 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
854 					bp->rx_cp_nr_rings);
855 			return -ENOTSUP;
856 		}
857 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
858 		if (rc)
859 			return rc;
860 	}
861 
862 	if (rte_intr_dp_is_en(intr_handle)) {
863 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
864 					bp->eth_dev->data->nb_rx_queues)) {
865 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
866 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
867 			rc = -ENOMEM;
868 			goto err_out;
869 		}
870 		PMD_DRV_LOG(DEBUG, "intr_handle->nb_efd = %d "
871 			    "intr_handle->max_intr = %d\n",
872 			    rte_intr_nb_efd_get(intr_handle),
873 			    rte_intr_max_intr_get(intr_handle));
874 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
875 		     queue_id++) {
876 			rte_intr_vec_list_index_set(intr_handle,
877 					queue_id, vec + BNXT_RX_VEC_START);
878 			if (vec < base + rte_intr_nb_efd_get(intr_handle)
879 			    - 1)
880 				vec++;
881 		}
882 	}
883 
884 	/* enable uio/vfio intr/eventfd mapping */
885 	rc = rte_intr_enable(intr_handle);
886 #ifndef RTE_EXEC_ENV_FREEBSD
887 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
888 	if (rc)
889 		goto err_out;
890 #endif
891 
892 	rc = bnxt_update_phy_setting(bp);
893 	if (rc)
894 		goto err_out;
895 
896 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
897 	if (!bp->mark_table)
898 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
899 
900 	return 0;
901 
902 err_out:
903 	/* Some of the error status returned by FW may not be from errno.h */
904 	if (rc > 0)
905 		rc = -EIO;
906 
907 	return rc;
908 }
909 
910 static int bnxt_shutdown_nic(struct bnxt *bp)
911 {
912 	bnxt_free_all_hwrm_resources(bp);
913 	bnxt_free_all_filters(bp);
914 	bnxt_free_all_vnics(bp);
915 	bnxt_vnic_queue_db_deinit(bp);
916 	return 0;
917 }
918 
919 /*
920  * Device configuration and status function
921  */
922 
923 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
924 {
925 	uint32_t pam4_link_speed = 0;
926 	uint32_t link_speed = 0;
927 	uint32_t speed_capa = 0;
928 
929 	if (bp->link_info == NULL)
930 		return 0;
931 
932 	link_speed = bp->link_info->support_speeds;
933 
934 	/* If PAM4 is configured, use PAM4 supported speed */
935 	if (bp->link_info->support_pam4_speeds > 0)
936 		pam4_link_speed = bp->link_info->support_pam4_speeds;
937 
938 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
939 		speed_capa |= RTE_ETH_LINK_SPEED_100M;
940 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
941 		speed_capa |= RTE_ETH_LINK_SPEED_100M_HD;
942 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
943 		speed_capa |= RTE_ETH_LINK_SPEED_1G;
944 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
945 		speed_capa |= RTE_ETH_LINK_SPEED_2_5G;
946 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
947 		speed_capa |= RTE_ETH_LINK_SPEED_10G;
948 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
949 		speed_capa |= RTE_ETH_LINK_SPEED_20G;
950 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
951 		speed_capa |= RTE_ETH_LINK_SPEED_25G;
952 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
953 		speed_capa |= RTE_ETH_LINK_SPEED_40G;
954 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
955 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
956 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
957 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
958 	if (pam4_link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
959 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
960 	if (pam4_link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
961 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
962 	if (pam4_link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
963 		speed_capa |= RTE_ETH_LINK_SPEED_200G;
964 
965 	if (bp->link_info->auto_mode ==
966 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
967 		speed_capa |= RTE_ETH_LINK_SPEED_FIXED;
968 
969 	return speed_capa;
970 }
971 
972 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
973 				struct rte_eth_dev_info *dev_info)
974 {
975 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
976 	struct bnxt *bp = eth_dev->data->dev_private;
977 	uint16_t max_vnics, i, j, vpool, vrxq;
978 	unsigned int max_rx_rings;
979 	int rc;
980 
981 	rc = is_bnxt_in_error(bp);
982 	if (rc)
983 		return rc;
984 
985 	/* MAC Specifics */
986 	dev_info->max_mac_addrs = RTE_MIN(bp->max_l2_ctx, RTE_ETH_NUM_RECEIVE_MAC_ADDR);
987 	dev_info->max_hash_mac_addrs = 0;
988 
989 	/* PF/VF specifics */
990 	if (BNXT_PF(bp))
991 		dev_info->max_vfs = pdev->max_vfs;
992 
993 	max_rx_rings = bnxt_max_rings(bp);
994 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
995 	dev_info->max_rx_queues = max_rx_rings;
996 	dev_info->max_tx_queues = max_rx_rings;
997 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
998 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
999 	max_vnics = bp->max_vnics;
1000 
1001 	/* MTU specifics */
1002 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
1003 	dev_info->max_mtu = BNXT_MAX_MTU;
1004 
1005 	/* Fast path specifics */
1006 	dev_info->min_rx_bufsize = 1;
1007 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
1008 
1009 	dev_info->rx_offload_capa = bnxt_get_rx_port_offloads(bp);
1010 	dev_info->tx_queue_offload_capa = RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
1011 	dev_info->tx_offload_capa = bnxt_get_tx_port_offloads(bp) |
1012 				    dev_info->tx_queue_offload_capa;
1013 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
1014 
1015 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
1016 	dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
1017 			     RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
1018 	dev_info->dev_capa &= ~RTE_ETH_DEV_CAPA_FLOW_RULE_KEEP;
1019 
1020 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
1021 		.rx_thresh = {
1022 			.pthresh = 8,
1023 			.hthresh = 8,
1024 			.wthresh = 0,
1025 		},
1026 		.rx_free_thresh = 32,
1027 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
1028 	};
1029 
1030 	dev_info->default_txconf = (struct rte_eth_txconf) {
1031 		.tx_thresh = {
1032 			.pthresh = 32,
1033 			.hthresh = 0,
1034 			.wthresh = 0,
1035 		},
1036 		.tx_free_thresh = 32,
1037 		.tx_rs_thresh = 32,
1038 	};
1039 
1040 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1041 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1042 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1043 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1044 
1045 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1046 		dev_info->switch_info.name = eth_dev->device->name;
1047 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1048 		dev_info->switch_info.port_id =
1049 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1050 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1051 	}
1052 
1053 	/*
1054 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1055 	 *       need further investigation.
1056 	 */
1057 
1058 	/* VMDq resources */
1059 	vpool = 64; /* RTE_ETH_64_POOLS */
1060 	vrxq = 128; /* RTE_ETH_VMDQ_DCB_NUM_QUEUES */
1061 	for (i = 0; i < 4; vpool >>= 1, i++) {
1062 		if (max_vnics > vpool) {
1063 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1064 				if (dev_info->max_rx_queues > vrxq) {
1065 					if (vpool > vrxq)
1066 						vpool = vrxq;
1067 					goto found;
1068 				}
1069 			}
1070 			/* Not enough resources to support VMDq */
1071 			break;
1072 		}
1073 	}
1074 	/* Not enough resources to support VMDq */
1075 	vpool = 0;
1076 	vrxq = 0;
1077 found:
1078 	dev_info->max_vmdq_pools = vpool;
1079 	dev_info->vmdq_queue_num = vrxq;
1080 
1081 	dev_info->vmdq_pool_base = 0;
1082 	dev_info->vmdq_queue_base = 0;
1083 
1084 	dev_info->err_handle_mode = RTE_ETH_ERROR_HANDLE_MODE_PROACTIVE;
1085 
1086 	return 0;
1087 }
1088 
1089 /* Configure the device based on the configuration provided */
1090 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1091 {
1092 	struct bnxt *bp = eth_dev->data->dev_private;
1093 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1094 	struct rte_eth_rss_conf *rss_conf = &eth_dev->data->dev_conf.rx_adv_conf.rss_conf;
1095 	int rc;
1096 
1097 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1098 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1099 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1100 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1101 
1102 	rc = is_bnxt_in_error(bp);
1103 	if (rc)
1104 		return rc;
1105 
1106 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1107 		rc = bnxt_hwrm_check_vf_rings(bp);
1108 		if (rc) {
1109 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1110 			return -ENOSPC;
1111 		}
1112 
1113 		/* If a resource has already been allocated - in this case
1114 		 * it is the async completion ring, free it. Reallocate it after
1115 		 * resource reservation. This will ensure the resource counts
1116 		 * are calculated correctly.
1117 		 */
1118 
1119 		pthread_mutex_lock(&bp->def_cp_lock);
1120 
1121 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1122 			bnxt_disable_int(bp);
1123 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1124 		}
1125 
1126 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1127 		if (rc) {
1128 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1129 			pthread_mutex_unlock(&bp->def_cp_lock);
1130 			return -ENOSPC;
1131 		}
1132 
1133 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1134 			rc = bnxt_alloc_async_cp_ring(bp);
1135 			if (rc) {
1136 				pthread_mutex_unlock(&bp->def_cp_lock);
1137 				return rc;
1138 			}
1139 			bnxt_enable_int(bp);
1140 		}
1141 
1142 		pthread_mutex_unlock(&bp->def_cp_lock);
1143 	}
1144 
1145 	/* Inherit new configurations */
1146 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1147 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1148 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1149 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1150 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1151 	    bp->max_stat_ctx)
1152 		goto resource_error;
1153 
1154 	if (BNXT_HAS_RING_GRPS(bp) &&
1155 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1156 		goto resource_error;
1157 
1158 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS) &&
1159 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1160 		goto resource_error;
1161 
1162 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1163 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1164 
1165 	if (eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
1166 		rx_offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
1167 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1168 
1169 	/* application provides the hash key to program */
1170 	if (rss_conf->rss_key != NULL) {
1171 		if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE)
1172 			PMD_DRV_LOG(WARNING, "port %u RSS key len must be %d bytes long",
1173 				    eth_dev->data->port_id, HW_HASH_KEY_SIZE);
1174 		else
1175 			memcpy(bp->rss_conf.rss_key, rss_conf->rss_key, HW_HASH_KEY_SIZE);
1176 	}
1177 	bp->rss_conf.rss_key_len = HW_HASH_KEY_SIZE;
1178 	bp->rss_conf.rss_hf = rss_conf->rss_hf;
1179 
1180 	bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1181 
1182 	return 0;
1183 
1184 resource_error:
1185 	PMD_DRV_LOG(ERR,
1186 		    "Insufficient resources to support requested config\n");
1187 	PMD_DRV_LOG(ERR,
1188 		    "Num Queues Requested: Tx %d, Rx %d\n",
1189 		    eth_dev->data->nb_tx_queues,
1190 		    eth_dev->data->nb_rx_queues);
1191 	PMD_DRV_LOG(ERR,
1192 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1193 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1194 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1195 	return -ENOSPC;
1196 }
1197 
1198 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1199 {
1200 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1201 
1202 	if (link->link_status)
1203 		PMD_DRV_LOG(DEBUG, "Port %d Link Up - speed %u Mbps - %s\n",
1204 			eth_dev->data->port_id,
1205 			(uint32_t)link->link_speed,
1206 			(link->link_duplex == RTE_ETH_LINK_FULL_DUPLEX) ?
1207 			("full-duplex") : ("half-duplex\n"));
1208 	else
1209 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1210 			eth_dev->data->port_id);
1211 }
1212 
1213 /*
1214  * Determine whether the current configuration requires support for scattered
1215  * receive; return 1 if scattered receive is required and 0 if not.
1216  */
1217 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1218 {
1219 	uint32_t overhead = BNXT_MAX_PKT_LEN - BNXT_MAX_MTU;
1220 	uint16_t buf_size;
1221 	int i;
1222 
1223 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_SCATTER)
1224 		return 1;
1225 
1226 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_TCP_LRO)
1227 		return 1;
1228 
1229 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1230 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1231 
1232 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1233 				      RTE_PKTMBUF_HEADROOM);
1234 		if (eth_dev->data->mtu + overhead > buf_size)
1235 			return 1;
1236 	}
1237 	return 0;
1238 }
1239 
1240 static eth_rx_burst_t
1241 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1242 {
1243 	struct bnxt *bp = eth_dev->data->dev_private;
1244 
1245 	/* Disable vector mode RX for Stingray2 for now */
1246 	if (BNXT_CHIP_SR2(bp)) {
1247 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1248 		return bnxt_recv_pkts;
1249 	}
1250 
1251 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1252 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1253 	if (eth_dev->data->scattered_rx)
1254 		goto use_scalar_rx;
1255 
1256 	/*
1257 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1258 	 * asynchronous completions and receive completions can be placed in
1259 	 * the same completion ring.
1260 	 */
1261 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1262 		goto use_scalar_rx;
1263 
1264 	/*
1265 	 * Vector mode receive cannot be enabled if any receive offloads outside
1266 	 * a limited subset have been enabled.
1267 	 */
1268 	if (eth_dev->data->dev_conf.rxmode.offloads &
1269 		~(RTE_ETH_RX_OFFLOAD_VLAN_STRIP |
1270 		  RTE_ETH_RX_OFFLOAD_KEEP_CRC |
1271 		  RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
1272 		  RTE_ETH_RX_OFFLOAD_UDP_CKSUM |
1273 		  RTE_ETH_RX_OFFLOAD_TCP_CKSUM |
1274 		  RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1275 		  RTE_ETH_RX_OFFLOAD_OUTER_UDP_CKSUM |
1276 		  RTE_ETH_RX_OFFLOAD_RSS_HASH |
1277 		  RTE_ETH_RX_OFFLOAD_VLAN_FILTER))
1278 		goto use_scalar_rx;
1279 
1280 	if (bp->ieee_1588)
1281 		goto use_scalar_rx;
1282 
1283 #if defined(RTE_ARCH_X86)
1284 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1285 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1286 		PMD_DRV_LOG(INFO,
1287 			    "Using AVX2 vector mode receive for port %d\n",
1288 			    eth_dev->data->port_id);
1289 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1290 		return bnxt_recv_pkts_vec_avx2;
1291 	}
1292  #endif
1293 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1294 		PMD_DRV_LOG(INFO,
1295 			    "Using SSE vector mode receive for port %d\n",
1296 			    eth_dev->data->port_id);
1297 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1298 		return bnxt_recv_pkts_vec;
1299 	}
1300 
1301 use_scalar_rx:
1302 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1303 		    eth_dev->data->port_id);
1304 	PMD_DRV_LOG(INFO,
1305 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1306 		    eth_dev->data->port_id,
1307 		    eth_dev->data->scattered_rx,
1308 		    eth_dev->data->dev_conf.rxmode.offloads);
1309 #endif
1310 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1311 	return bnxt_recv_pkts;
1312 }
1313 
1314 static eth_tx_burst_t
1315 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1316 {
1317 	struct bnxt *bp = eth_dev->data->dev_private;
1318 
1319 	/* Disable vector mode TX for Stingray2 for now */
1320 	if (BNXT_CHIP_SR2(bp))
1321 		return bnxt_xmit_pkts;
1322 
1323 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1324 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1325 
1326 	/*
1327 	 * Vector mode transmit can be enabled only if not using scatter rx
1328 	 * or tx offloads.
1329 	 */
1330 	if (eth_dev->data->scattered_rx ||
1331 	    (offloads & ~RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE) ||
1332 	    BNXT_TRUFLOW_EN(bp) || bp->ieee_1588)
1333 		goto use_scalar_tx;
1334 
1335 #if defined(RTE_ARCH_X86)
1336 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1337 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1338 		PMD_DRV_LOG(INFO,
1339 			    "Using AVX2 vector mode transmit for port %d\n",
1340 			    eth_dev->data->port_id);
1341 		return bnxt_xmit_pkts_vec_avx2;
1342 	}
1343 #endif
1344 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1345 		PMD_DRV_LOG(INFO,
1346 			    "Using SSE vector mode transmit for port %d\n",
1347 			    eth_dev->data->port_id);
1348 		return bnxt_xmit_pkts_vec;
1349 	}
1350 
1351 use_scalar_tx:
1352 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1353 		    eth_dev->data->port_id);
1354 	PMD_DRV_LOG(INFO,
1355 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1356 		    eth_dev->data->port_id,
1357 		    eth_dev->data->scattered_rx,
1358 		    offloads);
1359 #endif
1360 	return bnxt_xmit_pkts;
1361 }
1362 
1363 static int bnxt_handle_if_change_status(struct bnxt *bp)
1364 {
1365 	int rc;
1366 
1367 	/* Since fw has undergone a reset and lost all contexts,
1368 	 * set fatal flag to not issue hwrm during cleanup
1369 	 */
1370 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1371 	bnxt_uninit_resources(bp, true);
1372 
1373 	/* clear fatal flag so that re-init happens */
1374 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1375 
1376 	rc = bnxt_check_fw_ready(bp);
1377 	if (rc)
1378 		return rc;
1379 
1380 	rc = bnxt_init_resources(bp, true);
1381 
1382 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1383 
1384 	return rc;
1385 }
1386 
1387 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1388 {
1389 	struct bnxt *bp = eth_dev->data->dev_private;
1390 	int rc = 0;
1391 
1392 	if (!BNXT_SINGLE_PF(bp))
1393 		return -ENOTSUP;
1394 
1395 	if (!bp->link_info->link_up)
1396 		rc = bnxt_set_hwrm_link_config(bp, true);
1397 	if (!rc)
1398 		eth_dev->data->dev_link.link_status = 1;
1399 
1400 	bnxt_print_link_info(eth_dev);
1401 	return rc;
1402 }
1403 
1404 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1405 {
1406 	struct bnxt *bp = eth_dev->data->dev_private;
1407 
1408 	if (!BNXT_SINGLE_PF(bp))
1409 		return -ENOTSUP;
1410 
1411 	eth_dev->data->dev_link.link_status = 0;
1412 	bnxt_set_hwrm_link_config(bp, false);
1413 	bp->link_info->link_up = 0;
1414 
1415 	return 0;
1416 }
1417 
1418 static void bnxt_free_switch_domain(struct bnxt *bp)
1419 {
1420 	int rc = 0;
1421 
1422 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1423 		return;
1424 
1425 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1426 	if (rc)
1427 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1428 			    bp->switch_domain_id, rc);
1429 }
1430 
1431 static void bnxt_ptp_get_current_time(void *arg)
1432 {
1433 	struct bnxt *bp = arg;
1434 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1435 	int rc;
1436 
1437 	rc = is_bnxt_in_error(bp);
1438 	if (rc)
1439 		return;
1440 
1441 	if (!ptp)
1442 		return;
1443 
1444 	rte_spinlock_lock(&ptp->ptp_lock);
1445 	ptp->old_time = ptp->current_time;
1446 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1447 				&ptp->current_time);
1448 	rte_spinlock_unlock(&ptp->ptp_lock);
1449 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1450 	if (rc != 0) {
1451 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1452 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1453 	}
1454 }
1455 
1456 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1457 {
1458 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1459 	int rc;
1460 
1461 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1462 		return 0;
1463 
1464 	rte_spinlock_lock(&ptp->ptp_lock);
1465 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1466 				&ptp->current_time);
1467 	ptp->old_time = ptp->current_time;
1468 	rte_spinlock_unlock(&ptp->ptp_lock);
1469 
1470 
1471 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1472 	return rc;
1473 }
1474 
1475 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1476 {
1477 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1478 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1479 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1480 	}
1481 }
1482 
1483 static void bnxt_ptp_stop(struct bnxt *bp)
1484 {
1485 	bnxt_cancel_ptp_alarm(bp);
1486 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1487 }
1488 
1489 static int bnxt_ptp_start(struct bnxt *bp)
1490 {
1491 	int rc;
1492 
1493 	rc = bnxt_schedule_ptp_alarm(bp);
1494 	if (rc != 0) {
1495 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1496 	} else {
1497 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1498 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1499 	}
1500 
1501 	return rc;
1502 }
1503 
1504 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1505 {
1506 	struct bnxt *bp = eth_dev->data->dev_private;
1507 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1508 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1509 	struct rte_eth_link link;
1510 	int ret;
1511 
1512 	eth_dev->data->dev_started = 0;
1513 
1514 	/* Prevent crashes when queues are still in use */
1515 	bnxt_stop_rxtx(eth_dev);
1516 
1517 	bnxt_disable_int(bp);
1518 
1519 	/* disable uio/vfio intr/eventfd mapping */
1520 	rte_intr_disable(intr_handle);
1521 
1522 	/* Stop the child representors for this device */
1523 	ret = bnxt_rep_stop_all(bp);
1524 	if (ret != 0)
1525 		return ret;
1526 
1527 	/* delete the bnxt ULP port details */
1528 	bnxt_ulp_port_deinit(bp);
1529 
1530 	bnxt_cancel_fw_health_check(bp);
1531 
1532 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1533 		bnxt_cancel_ptp_alarm(bp);
1534 
1535 	/* Do not bring link down during reset recovery */
1536 	if (!is_bnxt_in_error(bp)) {
1537 		bnxt_dev_set_link_down_op(eth_dev);
1538 		/* Wait for link to be reset */
1539 		if (BNXT_SINGLE_PF(bp))
1540 			rte_delay_ms(500);
1541 		/* clear the recorded link status */
1542 		memset(&link, 0, sizeof(link));
1543 		rte_eth_linkstatus_set(eth_dev, &link);
1544 	}
1545 
1546 	/* Clean queue intr-vector mapping */
1547 	rte_intr_efd_disable(intr_handle);
1548 	rte_intr_vec_list_free(intr_handle);
1549 
1550 	bnxt_hwrm_port_clr_stats(bp);
1551 	bnxt_free_tx_mbufs(bp);
1552 	bnxt_free_rx_mbufs(bp);
1553 	/* Process any remaining notifications in default completion queue */
1554 	bnxt_int_handler(eth_dev);
1555 	bnxt_shutdown_nic(bp);
1556 	bnxt_hwrm_if_change(bp, false);
1557 
1558 	bnxt_free_prev_ring_stats(bp);
1559 	rte_free(bp->mark_table);
1560 	bp->mark_table = NULL;
1561 
1562 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1563 	bp->rx_cosq_cnt = 0;
1564 	/* All filters are deleted on a port stop. */
1565 	if (BNXT_FLOW_XSTATS_EN(bp))
1566 		bp->flow_stat->flow_count = 0;
1567 
1568 	eth_dev->data->scattered_rx = 0;
1569 
1570 	return 0;
1571 }
1572 
1573 /* Unload the driver, release resources */
1574 int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1575 {
1576 	struct bnxt *bp = eth_dev->data->dev_private;
1577 
1578 	pthread_mutex_lock(&bp->err_recovery_lock);
1579 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1580 		PMD_DRV_LOG(ERR,
1581 			    "Adapter recovering from error..Please retry\n");
1582 		pthread_mutex_unlock(&bp->err_recovery_lock);
1583 		return -EAGAIN;
1584 	}
1585 	pthread_mutex_unlock(&bp->err_recovery_lock);
1586 
1587 	return bnxt_dev_stop(eth_dev);
1588 }
1589 
1590 int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1591 {
1592 	struct bnxt *bp = eth_dev->data->dev_private;
1593 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1594 	int vlan_mask = 0;
1595 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1596 
1597 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1598 		PMD_DRV_LOG(ERR,
1599 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1600 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1601 
1602 	do {
1603 		rc = bnxt_hwrm_if_change(bp, true);
1604 		if (rc == 0 || rc != -EAGAIN)
1605 			break;
1606 
1607 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1608 	} while (retry_cnt--);
1609 
1610 	if (rc)
1611 		return rc;
1612 
1613 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1614 		rc = bnxt_handle_if_change_status(bp);
1615 		if (rc)
1616 			return rc;
1617 	}
1618 
1619 	bnxt_enable_int(bp);
1620 
1621 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1622 
1623 	rc = bnxt_start_nic(bp);
1624 	if (rc)
1625 		goto error;
1626 
1627 	rc = bnxt_alloc_prev_ring_stats(bp);
1628 	if (rc)
1629 		goto error;
1630 
1631 	eth_dev->data->dev_started = 1;
1632 
1633 	bnxt_link_update_op(eth_dev, 0);
1634 
1635 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
1636 		vlan_mask |= RTE_ETH_VLAN_FILTER_MASK;
1637 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
1638 		vlan_mask |= RTE_ETH_VLAN_STRIP_MASK;
1639 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1640 	if (rc)
1641 		goto error;
1642 
1643 	/* Initialize bnxt ULP port details */
1644 	rc = bnxt_ulp_port_init(bp);
1645 	if (rc)
1646 		goto error;
1647 
1648 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1649 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1650 
1651 	bnxt_schedule_fw_health_check(bp);
1652 
1653 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1654 		bnxt_schedule_ptp_alarm(bp);
1655 
1656 	return 0;
1657 
1658 error:
1659 	bnxt_dev_stop(eth_dev);
1660 	return rc;
1661 }
1662 
1663 static void
1664 bnxt_uninit_locks(struct bnxt *bp)
1665 {
1666 	pthread_mutex_destroy(&bp->flow_lock);
1667 	pthread_mutex_destroy(&bp->def_cp_lock);
1668 	pthread_mutex_destroy(&bp->health_check_lock);
1669 	pthread_mutex_destroy(&bp->err_recovery_lock);
1670 	if (bp->rep_info) {
1671 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1672 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1673 	}
1674 }
1675 
1676 static void bnxt_drv_uninit(struct bnxt *bp)
1677 {
1678 	bnxt_free_leds_info(bp);
1679 	bnxt_free_cos_queues(bp);
1680 	bnxt_free_link_info(bp);
1681 	bnxt_free_parent_info(bp);
1682 	bnxt_uninit_locks(bp);
1683 	bnxt_free_rep_info(bp);
1684 
1685 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1686 	bp->tx_mem_zone = NULL;
1687 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1688 	bp->rx_mem_zone = NULL;
1689 
1690 	bnxt_free_vf_info(bp);
1691 	bnxt_free_pf_info(bp);
1692 
1693 	rte_free(bp->grp_info);
1694 	bp->grp_info = NULL;
1695 }
1696 
1697 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1698 {
1699 	struct bnxt *bp = eth_dev->data->dev_private;
1700 	int ret = 0;
1701 
1702 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1703 		return 0;
1704 
1705 	pthread_mutex_lock(&bp->err_recovery_lock);
1706 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1707 		PMD_DRV_LOG(ERR,
1708 			    "Adapter recovering from error...Please retry\n");
1709 		pthread_mutex_unlock(&bp->err_recovery_lock);
1710 		return -EAGAIN;
1711 	}
1712 	pthread_mutex_unlock(&bp->err_recovery_lock);
1713 
1714 	/* cancel the recovery handler before remove dev */
1715 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1716 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1717 	bnxt_cancel_fc_thread(bp);
1718 	rte_eal_alarm_cancel(bnxt_handle_vf_cfg_change, (void *)bp);
1719 
1720 	if (eth_dev->data->dev_started)
1721 		ret = bnxt_dev_stop(eth_dev);
1722 
1723 	bnxt_uninit_resources(bp, false);
1724 
1725 	bnxt_drv_uninit(bp);
1726 
1727 	return ret;
1728 }
1729 
1730 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1731 				    uint32_t index)
1732 {
1733 	struct bnxt *bp = eth_dev->data->dev_private;
1734 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1735 	struct bnxt_vnic_info *vnic;
1736 	struct bnxt_filter_info *filter, *temp_filter;
1737 	uint32_t i;
1738 
1739 	if (is_bnxt_in_error(bp))
1740 		return;
1741 
1742 	/*
1743 	 * Loop through all VNICs from the specified filter flow pools to
1744 	 * remove the corresponding MAC addr filter
1745 	 */
1746 	for (i = 0; i < bp->nr_vnics; i++) {
1747 		if (!(pool_mask & (1ULL << i)))
1748 			continue;
1749 
1750 		vnic = &bp->vnic_info[i];
1751 		filter = STAILQ_FIRST(&vnic->filter);
1752 		while (filter) {
1753 			temp_filter = STAILQ_NEXT(filter, next);
1754 			if (filter->mac_index == index) {
1755 				STAILQ_REMOVE(&vnic->filter, filter,
1756 						bnxt_filter_info, next);
1757 				bnxt_hwrm_clear_l2_filter(bp, filter);
1758 				bnxt_free_filter(bp, filter);
1759 			}
1760 			filter = temp_filter;
1761 		}
1762 	}
1763 }
1764 
1765 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1766 			       struct rte_ether_addr *mac_addr, uint32_t index,
1767 			       uint32_t pool)
1768 {
1769 	struct bnxt_filter_info *filter;
1770 	int rc = 0;
1771 
1772 	/* Attach requested MAC address to the new l2_filter */
1773 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1774 		if (filter->mac_index == index) {
1775 			PMD_DRV_LOG(DEBUG,
1776 				    "MAC addr already existed for pool %d\n",
1777 				    pool);
1778 			return 0;
1779 		}
1780 	}
1781 
1782 	filter = bnxt_alloc_filter(bp);
1783 	if (!filter) {
1784 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1785 		return -ENODEV;
1786 	}
1787 
1788 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1789 	 * if the MAC that's been programmed now is a different one, then,
1790 	 * copy that addr to filter->l2_addr
1791 	 */
1792 	if (mac_addr)
1793 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1794 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1795 
1796 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1797 	if (!rc) {
1798 		filter->mac_index = index;
1799 		if (filter->mac_index == 0)
1800 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1801 		else
1802 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1803 	} else {
1804 		bnxt_free_filter(bp, filter);
1805 	}
1806 
1807 	return rc;
1808 }
1809 
1810 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1811 				struct rte_ether_addr *mac_addr,
1812 				uint32_t index, uint32_t pool)
1813 {
1814 	struct bnxt *bp = eth_dev->data->dev_private;
1815 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1816 	int rc = 0;
1817 
1818 	rc = is_bnxt_in_error(bp);
1819 	if (rc)
1820 		return rc;
1821 
1822 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1823 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1824 		return -ENOTSUP;
1825 	}
1826 
1827 	if (!vnic) {
1828 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1829 		return -EINVAL;
1830 	}
1831 
1832 	/* Filter settings will get applied when port is started */
1833 	if (!eth_dev->data->dev_started)
1834 		return 0;
1835 
1836 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1837 
1838 	return rc;
1839 }
1840 
1841 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1842 {
1843 	int rc = 0;
1844 	struct bnxt *bp = eth_dev->data->dev_private;
1845 	struct rte_eth_link new;
1846 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1847 			BNXT_MIN_LINK_WAIT_CNT;
1848 
1849 	rc = is_bnxt_in_error(bp);
1850 	if (rc)
1851 		return rc;
1852 
1853 	memset(&new, 0, sizeof(new));
1854 
1855 	if (bp->link_info == NULL)
1856 		goto out;
1857 
1858 	/* Only single function PF can bring the phy down.
1859 	 * In certain scenarios, device is not obliged link down even when forced.
1860 	 * When port is stopped, report link down in those cases.
1861 	 */
1862 	if (!eth_dev->data->dev_started &&
1863 	    (!BNXT_SINGLE_PF(bp) || bnxt_force_link_config(bp)))
1864 		goto out;
1865 
1866 	do {
1867 		/* Retrieve link info from hardware */
1868 		rc = bnxt_get_hwrm_link_config(bp, &new);
1869 		if (rc) {
1870 			new.link_speed = RTE_ETH_LINK_SPEED_100M;
1871 			new.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
1872 			PMD_DRV_LOG(ERR,
1873 				"Failed to retrieve link rc = 0x%x!\n", rc);
1874 			goto out;
1875 		}
1876 
1877 		if (!wait_to_complete || new.link_status)
1878 			break;
1879 
1880 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1881 	} while (cnt--);
1882 
1883 out:
1884 	/* Timed out or success */
1885 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1886 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1887 		rte_eth_linkstatus_set(eth_dev, &new);
1888 		bnxt_print_link_info(eth_dev);
1889 	}
1890 
1891 	return rc;
1892 }
1893 
1894 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1895 {
1896 	struct bnxt *bp = eth_dev->data->dev_private;
1897 	struct bnxt_vnic_info *vnic;
1898 	uint32_t old_flags;
1899 	int rc;
1900 
1901 	rc = is_bnxt_in_error(bp);
1902 	if (rc)
1903 		return rc;
1904 
1905 	/* Filter settings will get applied when port is started */
1906 	if (!eth_dev->data->dev_started)
1907 		return 0;
1908 
1909 	if (bp->vnic_info == NULL)
1910 		return 0;
1911 
1912 	vnic = bnxt_get_default_vnic(bp);
1913 
1914 	old_flags = vnic->flags;
1915 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1916 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1917 	if (rc != 0)
1918 		vnic->flags = old_flags;
1919 
1920 	return rc;
1921 }
1922 
1923 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1924 {
1925 	struct bnxt *bp = eth_dev->data->dev_private;
1926 	struct bnxt_vnic_info *vnic;
1927 	uint32_t old_flags;
1928 	int rc;
1929 
1930 	rc = is_bnxt_in_error(bp);
1931 	if (rc)
1932 		return rc;
1933 
1934 	/* Filter settings will get applied when port is started */
1935 	if (!eth_dev->data->dev_started)
1936 		return 0;
1937 
1938 	if (bp->vnic_info == NULL)
1939 		return 0;
1940 
1941 	vnic = bnxt_get_default_vnic(bp);
1942 
1943 	old_flags = vnic->flags;
1944 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1945 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1946 	if (rc != 0)
1947 		vnic->flags = old_flags;
1948 
1949 	return rc;
1950 }
1951 
1952 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1953 {
1954 	struct bnxt *bp = eth_dev->data->dev_private;
1955 	struct bnxt_vnic_info *vnic;
1956 	uint32_t old_flags;
1957 	int rc;
1958 
1959 	rc = is_bnxt_in_error(bp);
1960 	if (rc)
1961 		return rc;
1962 
1963 	/* Filter settings will get applied when port is started */
1964 	if (!eth_dev->data->dev_started)
1965 		return 0;
1966 
1967 	if (bp->vnic_info == NULL)
1968 		return 0;
1969 
1970 	vnic = bnxt_get_default_vnic(bp);
1971 
1972 	old_flags = vnic->flags;
1973 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1974 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1975 	if (rc != 0)
1976 		vnic->flags = old_flags;
1977 
1978 	return rc;
1979 }
1980 
1981 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1982 {
1983 	struct bnxt *bp = eth_dev->data->dev_private;
1984 	struct bnxt_vnic_info *vnic;
1985 	uint32_t old_flags;
1986 	int rc;
1987 
1988 	rc = is_bnxt_in_error(bp);
1989 	if (rc)
1990 		return rc;
1991 
1992 	/* Filter settings will get applied when port is started */
1993 	if (!eth_dev->data->dev_started)
1994 		return 0;
1995 
1996 	if (bp->vnic_info == NULL)
1997 		return 0;
1998 
1999 	vnic = bnxt_get_default_vnic(bp);
2000 
2001 	old_flags = vnic->flags;
2002 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2003 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2004 	if (rc != 0)
2005 		vnic->flags = old_flags;
2006 
2007 	return rc;
2008 }
2009 
2010 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
2011 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
2012 {
2013 	if (qid >= bp->rx_nr_rings)
2014 		return NULL;
2015 
2016 	return bp->eth_dev->data->rx_queues[qid];
2017 }
2018 
2019 /* Return rxq corresponding to a given rss table ring/group ID. */
2020 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
2021 {
2022 	struct bnxt_rx_queue *rxq;
2023 	unsigned int i;
2024 
2025 	if (!BNXT_HAS_RING_GRPS(bp)) {
2026 		for (i = 0; i < bp->rx_nr_rings; i++) {
2027 			rxq = bp->eth_dev->data->rx_queues[i];
2028 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
2029 				return rxq->index;
2030 		}
2031 	} else {
2032 		for (i = 0; i < bp->rx_nr_rings; i++) {
2033 			if (bp->grp_info[i].fw_grp_id == fwr)
2034 				return i;
2035 		}
2036 	}
2037 
2038 	return INVALID_HW_RING_ID;
2039 }
2040 
2041 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
2042 			    struct rte_eth_rss_reta_entry64 *reta_conf,
2043 			    uint16_t reta_size)
2044 {
2045 	struct bnxt *bp = eth_dev->data->dev_private;
2046 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2047 	struct bnxt_vnic_info *vnic = bnxt_get_default_vnic(bp);
2048 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2049 	uint16_t idx, sft;
2050 	int i, rc;
2051 
2052 	rc = is_bnxt_in_error(bp);
2053 	if (rc)
2054 		return rc;
2055 
2056 	if (!vnic->rss_table)
2057 		return -EINVAL;
2058 
2059 	if (!(dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG))
2060 		return -EINVAL;
2061 
2062 	if (reta_size != tbl_size) {
2063 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2064 			"(%d) must equal the size supported by the hardware "
2065 			"(%d)\n", reta_size, tbl_size);
2066 		return -EINVAL;
2067 	}
2068 
2069 	if (bnxt_vnic_reta_config_update(bp, vnic, reta_conf, reta_size)) {
2070 		PMD_DRV_LOG(ERR, "Error in setting the reta config\n");
2071 		return -EINVAL;
2072 	}
2073 	for (i = 0; i < reta_size; i++) {
2074 		struct bnxt_rx_queue *rxq;
2075 
2076 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2077 		sft = i % RTE_ETH_RETA_GROUP_SIZE;
2078 
2079 		if (!(reta_conf[idx].mask & (1ULL << sft)))
2080 			continue;
2081 
2082 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2083 		if (BNXT_CHIP_P5(bp)) {
2084 			vnic->rss_table[i * 2] =
2085 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
2086 			vnic->rss_table[i * 2 + 1] =
2087 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
2088 		} else {
2089 			vnic->rss_table[i] =
2090 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2091 		}
2092 	}
2093 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2094 	return rc;
2095 }
2096 
2097 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2098 			      struct rte_eth_rss_reta_entry64 *reta_conf,
2099 			      uint16_t reta_size)
2100 {
2101 	struct bnxt *bp = eth_dev->data->dev_private;
2102 	struct bnxt_vnic_info *vnic = bnxt_get_default_vnic(bp);
2103 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2104 	uint16_t idx, sft, i;
2105 	int rc;
2106 
2107 	rc = is_bnxt_in_error(bp);
2108 	if (rc)
2109 		return rc;
2110 
2111 	if (!vnic)
2112 		return -EINVAL;
2113 	if (!vnic->rss_table)
2114 		return -EINVAL;
2115 
2116 	if (reta_size != tbl_size) {
2117 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2118 			"(%d) must equal the size supported by the hardware "
2119 			"(%d)\n", reta_size, tbl_size);
2120 		return -EINVAL;
2121 	}
2122 
2123 	for (idx = 0, i = 0; i < reta_size; i++) {
2124 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2125 		sft = i % RTE_ETH_RETA_GROUP_SIZE;
2126 
2127 		if (reta_conf[idx].mask & (1ULL << sft)) {
2128 			uint16_t qid;
2129 
2130 			if (BNXT_CHIP_P5(bp))
2131 				qid = bnxt_rss_to_qid(bp,
2132 						      vnic->rss_table[i * 2]);
2133 			else
2134 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2135 
2136 			if (qid == INVALID_HW_RING_ID) {
2137 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2138 				return -EINVAL;
2139 			}
2140 			reta_conf[idx].reta[sft] = qid;
2141 		}
2142 	}
2143 
2144 	return 0;
2145 }
2146 
2147 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2148 				   struct rte_eth_rss_conf *rss_conf)
2149 {
2150 	struct bnxt *bp = eth_dev->data->dev_private;
2151 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2152 	struct bnxt_vnic_info *vnic;
2153 	int rc;
2154 
2155 	rc = is_bnxt_in_error(bp);
2156 	if (rc)
2157 		return rc;
2158 
2159 	/*
2160 	 * If RSS enablement were different than dev_configure,
2161 	 * then return -EINVAL
2162 	 */
2163 	if (dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG) {
2164 		if (!rss_conf->rss_hf)
2165 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2166 	} else {
2167 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2168 			return -EINVAL;
2169 	}
2170 
2171 	/* Update the default RSS VNIC(s) */
2172 	vnic = bnxt_get_default_vnic(bp);
2173 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2174 	vnic->hash_mode =
2175 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2176 					    RTE_ETH_RSS_LEVEL(rss_conf->rss_hf));
2177 
2178 	/* Cache the hash function */
2179 	bp->rss_conf.rss_hf = rss_conf->rss_hf;
2180 
2181 	/*
2182 	 * If hashkey is not specified, use the previously configured
2183 	 * hashkey
2184 	 */
2185 	if (!rss_conf->rss_key)
2186 		goto rss_config;
2187 
2188 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2189 		PMD_DRV_LOG(ERR,
2190 			    "Invalid hashkey length, should be %d bytes\n",
2191 			    HW_HASH_KEY_SIZE);
2192 		return -EINVAL;
2193 	}
2194 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2195 
2196 	/* Cache the hash key */
2197 	memcpy(bp->rss_conf.rss_key, rss_conf->rss_key, HW_HASH_KEY_SIZE);
2198 
2199 rss_config:
2200 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2201 	return rc;
2202 }
2203 
2204 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2205 				     struct rte_eth_rss_conf *rss_conf)
2206 {
2207 	struct bnxt *bp = eth_dev->data->dev_private;
2208 	struct bnxt_vnic_info *vnic = bnxt_get_default_vnic(bp);
2209 	int len, rc;
2210 	uint32_t hash_types;
2211 
2212 	rc = is_bnxt_in_error(bp);
2213 	if (rc)
2214 		return rc;
2215 
2216 	/* RSS configuration is the same for all VNICs */
2217 	if (vnic && vnic->rss_hash_key) {
2218 		if (rss_conf->rss_key) {
2219 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2220 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2221 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2222 		}
2223 
2224 		hash_types = vnic->hash_type;
2225 		rss_conf->rss_hf = 0;
2226 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2227 			rss_conf->rss_hf |= RTE_ETH_RSS_IPV4;
2228 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2229 		}
2230 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2231 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_TCP;
2232 			hash_types &=
2233 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2234 		}
2235 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2236 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_UDP;
2237 			hash_types &=
2238 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2239 		}
2240 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2241 			rss_conf->rss_hf |= RTE_ETH_RSS_IPV6;
2242 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2243 		}
2244 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2245 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_TCP;
2246 			hash_types &=
2247 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2248 		}
2249 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2250 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_UDP;
2251 			hash_types &=
2252 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2253 		}
2254 
2255 		rss_conf->rss_hf |=
2256 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2257 
2258 		if (hash_types) {
2259 			PMD_DRV_LOG(ERR,
2260 				"Unknown RSS config from firmware (%08x), RSS disabled",
2261 				vnic->hash_type);
2262 			return -ENOTSUP;
2263 		}
2264 	} else {
2265 		rss_conf->rss_hf = 0;
2266 	}
2267 	return 0;
2268 }
2269 
2270 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2271 			       struct rte_eth_fc_conf *fc_conf)
2272 {
2273 	struct bnxt *bp = dev->data->dev_private;
2274 	struct rte_eth_link link_info;
2275 	int rc;
2276 
2277 	rc = is_bnxt_in_error(bp);
2278 	if (rc)
2279 		return rc;
2280 
2281 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2282 	if (rc)
2283 		return rc;
2284 
2285 	memset(fc_conf, 0, sizeof(*fc_conf));
2286 	if (bp->link_info->auto_pause)
2287 		fc_conf->autoneg = 1;
2288 	switch (bp->link_info->pause) {
2289 	case 0:
2290 		fc_conf->mode = RTE_ETH_FC_NONE;
2291 		break;
2292 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2293 		fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
2294 		break;
2295 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2296 		fc_conf->mode = RTE_ETH_FC_RX_PAUSE;
2297 		break;
2298 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2299 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2300 		fc_conf->mode = RTE_ETH_FC_FULL;
2301 		break;
2302 	}
2303 	return 0;
2304 }
2305 
2306 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2307 			       struct rte_eth_fc_conf *fc_conf)
2308 {
2309 	struct bnxt *bp = dev->data->dev_private;
2310 	int rc;
2311 
2312 	rc = is_bnxt_in_error(bp);
2313 	if (rc)
2314 		return rc;
2315 
2316 	if (!BNXT_SINGLE_PF(bp)) {
2317 		PMD_DRV_LOG(ERR,
2318 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2319 		return -ENOTSUP;
2320 	}
2321 
2322 	switch (fc_conf->mode) {
2323 	case RTE_ETH_FC_NONE:
2324 		bp->link_info->auto_pause = 0;
2325 		bp->link_info->force_pause = 0;
2326 		break;
2327 	case RTE_ETH_FC_RX_PAUSE:
2328 		if (fc_conf->autoneg) {
2329 			bp->link_info->auto_pause =
2330 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2331 			bp->link_info->force_pause = 0;
2332 		} else {
2333 			bp->link_info->auto_pause = 0;
2334 			bp->link_info->force_pause =
2335 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2336 		}
2337 		break;
2338 	case RTE_ETH_FC_TX_PAUSE:
2339 		if (fc_conf->autoneg) {
2340 			bp->link_info->auto_pause =
2341 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2342 			bp->link_info->force_pause = 0;
2343 		} else {
2344 			bp->link_info->auto_pause = 0;
2345 			bp->link_info->force_pause =
2346 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2347 		}
2348 		break;
2349 	case RTE_ETH_FC_FULL:
2350 		if (fc_conf->autoneg) {
2351 			bp->link_info->auto_pause =
2352 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2353 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2354 			bp->link_info->force_pause = 0;
2355 		} else {
2356 			bp->link_info->auto_pause = 0;
2357 			bp->link_info->force_pause =
2358 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2359 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2360 		}
2361 		break;
2362 	}
2363 	return bnxt_set_hwrm_link_config(bp, true);
2364 }
2365 
2366 /* Add UDP tunneling port */
2367 int
2368 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2369 			 struct rte_eth_udp_tunnel *udp_tunnel)
2370 {
2371 	struct bnxt *bp = eth_dev->data->dev_private;
2372 	uint16_t tunnel_type = 0;
2373 	int rc = 0;
2374 
2375 	rc = is_bnxt_in_error(bp);
2376 	if (rc)
2377 		return rc;
2378 
2379 	switch (udp_tunnel->prot_type) {
2380 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
2381 		if (bp->vxlan_port_cnt) {
2382 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2383 				udp_tunnel->udp_port);
2384 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2385 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2386 				return -ENOSPC;
2387 			}
2388 			bp->vxlan_port_cnt++;
2389 			return 0;
2390 		}
2391 		tunnel_type =
2392 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2393 		break;
2394 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2395 		if (bp->geneve_port_cnt) {
2396 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2397 				udp_tunnel->udp_port);
2398 			if (bp->geneve_port != udp_tunnel->udp_port) {
2399 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2400 				return -ENOSPC;
2401 			}
2402 			bp->geneve_port_cnt++;
2403 			return 0;
2404 		}
2405 		tunnel_type =
2406 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2407 		break;
2408 	case RTE_ETH_TUNNEL_TYPE_ECPRI:
2409 		if (bp->ecpri_port_cnt) {
2410 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2411 				udp_tunnel->udp_port);
2412 			if (bp->ecpri_port != udp_tunnel->udp_port) {
2413 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2414 				return -ENOSPC;
2415 			}
2416 			bp->ecpri_port_cnt++;
2417 			return 0;
2418 		}
2419 		tunnel_type =
2420 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_ECPRI;
2421 		break;
2422 	default:
2423 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2424 		return -ENOTSUP;
2425 	}
2426 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2427 					     tunnel_type);
2428 
2429 	if (rc != 0)
2430 		return rc;
2431 
2432 	if (tunnel_type ==
2433 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN)
2434 		bp->vxlan_port_cnt++;
2435 
2436 	if (tunnel_type ==
2437 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE)
2438 		bp->geneve_port_cnt++;
2439 
2440 	if (tunnel_type ==
2441 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_ECPRI)
2442 		bp->ecpri_port_cnt++;
2443 
2444 	return rc;
2445 }
2446 
2447 int
2448 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2449 			 struct rte_eth_udp_tunnel *udp_tunnel)
2450 {
2451 	struct bnxt *bp = eth_dev->data->dev_private;
2452 	uint16_t tunnel_type = 0;
2453 	uint16_t port = 0;
2454 	int rc = 0;
2455 
2456 	rc = is_bnxt_in_error(bp);
2457 	if (rc)
2458 		return rc;
2459 
2460 	switch (udp_tunnel->prot_type) {
2461 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
2462 		if (!bp->vxlan_port_cnt) {
2463 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2464 			return -EINVAL;
2465 		}
2466 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2467 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2468 				udp_tunnel->udp_port, bp->vxlan_port);
2469 			return -EINVAL;
2470 		}
2471 		if (--bp->vxlan_port_cnt)
2472 			return 0;
2473 
2474 		tunnel_type =
2475 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2476 		port = bp->vxlan_fw_dst_port_id;
2477 		break;
2478 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2479 		if (!bp->geneve_port_cnt) {
2480 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2481 			return -EINVAL;
2482 		}
2483 		if (bp->geneve_port != udp_tunnel->udp_port) {
2484 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2485 				udp_tunnel->udp_port, bp->geneve_port);
2486 			return -EINVAL;
2487 		}
2488 		if (--bp->geneve_port_cnt)
2489 			return 0;
2490 
2491 		tunnel_type =
2492 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2493 		port = bp->geneve_fw_dst_port_id;
2494 		break;
2495 	case RTE_ETH_TUNNEL_TYPE_ECPRI:
2496 		if (!bp->ecpri_port_cnt) {
2497 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2498 			return -EINVAL;
2499 		}
2500 		if (bp->ecpri_port != udp_tunnel->udp_port) {
2501 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2502 				udp_tunnel->udp_port, bp->ecpri_port);
2503 			return -EINVAL;
2504 		}
2505 		if (--bp->ecpri_port_cnt)
2506 			return 0;
2507 
2508 		tunnel_type =
2509 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_ECPRI;
2510 		port = bp->ecpri_fw_dst_port_id;
2511 		break;
2512 	default:
2513 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2514 		return -ENOTSUP;
2515 	}
2516 
2517 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2518 	return rc;
2519 }
2520 
2521 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2522 {
2523 	struct bnxt_filter_info *filter;
2524 	struct bnxt_vnic_info *vnic;
2525 	int rc = 0;
2526 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2527 
2528 	vnic = bnxt_get_default_vnic(bp);
2529 	filter = STAILQ_FIRST(&vnic->filter);
2530 	while (filter) {
2531 		/* Search for this matching MAC+VLAN filter */
2532 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2533 			/* Delete the filter */
2534 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2535 			if (rc)
2536 				return rc;
2537 			STAILQ_REMOVE(&vnic->filter, filter,
2538 				      bnxt_filter_info, next);
2539 			bnxt_free_filter(bp, filter);
2540 			PMD_DRV_LOG(INFO,
2541 				    "Deleted vlan filter for %d\n",
2542 				    vlan_id);
2543 			return 0;
2544 		}
2545 		filter = STAILQ_NEXT(filter, next);
2546 	}
2547 	return -ENOENT;
2548 }
2549 
2550 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2551 {
2552 	struct bnxt_filter_info *filter;
2553 	struct bnxt_vnic_info *vnic;
2554 	int rc = 0;
2555 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2556 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2557 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2558 
2559 	/* Implementation notes on the use of VNIC in this command:
2560 	 *
2561 	 * By default, these filters belong to default vnic for the function.
2562 	 * Once these filters are set up, only destination VNIC can be modified.
2563 	 * If the destination VNIC is not specified in this command,
2564 	 * then the HWRM shall only create an l2 context id.
2565 	 */
2566 
2567 	vnic = bnxt_get_default_vnic(bp);
2568 	filter = STAILQ_FIRST(&vnic->filter);
2569 	/* Check if the VLAN has already been added */
2570 	while (filter) {
2571 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2572 			return -EEXIST;
2573 
2574 		filter = STAILQ_NEXT(filter, next);
2575 	}
2576 
2577 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2578 	 * command to create MAC+VLAN filter with the right flags, enables set.
2579 	 */
2580 	filter = bnxt_alloc_filter(bp);
2581 	if (!filter) {
2582 		PMD_DRV_LOG(ERR,
2583 			    "MAC/VLAN filter alloc failed\n");
2584 		return -ENOMEM;
2585 	}
2586 	/* MAC + VLAN ID filter */
2587 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2588 	 * untagged packets are received
2589 	 *
2590 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2591 	 * packets and only the programmed vlan's packets are received
2592 	 */
2593 	filter->l2_ivlan = vlan_id;
2594 	filter->l2_ivlan_mask = 0x0FFF;
2595 	filter->enables |= en;
2596 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2597 
2598 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2599 	if (rc) {
2600 		/* Free the newly allocated filter as we were
2601 		 * not able to create the filter in hardware.
2602 		 */
2603 		bnxt_free_filter(bp, filter);
2604 		return rc;
2605 	}
2606 
2607 	filter->mac_index = 0;
2608 	/* Add this new filter to the list */
2609 	if (vlan_id == 0)
2610 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2611 	else
2612 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2613 
2614 	PMD_DRV_LOG(INFO,
2615 		    "Added Vlan filter for %d\n", vlan_id);
2616 	return rc;
2617 }
2618 
2619 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2620 		uint16_t vlan_id, int on)
2621 {
2622 	struct bnxt *bp = eth_dev->data->dev_private;
2623 	int rc;
2624 
2625 	rc = is_bnxt_in_error(bp);
2626 	if (rc)
2627 		return rc;
2628 
2629 	if (!eth_dev->data->dev_started) {
2630 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2631 		return -EINVAL;
2632 	}
2633 
2634 	/* These operations apply to ALL existing MAC/VLAN filters */
2635 	if (on)
2636 		return bnxt_add_vlan_filter(bp, vlan_id);
2637 	else
2638 		return bnxt_del_vlan_filter(bp, vlan_id);
2639 }
2640 
2641 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2642 				    struct bnxt_vnic_info *vnic)
2643 {
2644 	struct bnxt_filter_info *filter;
2645 	int rc;
2646 
2647 	filter = STAILQ_FIRST(&vnic->filter);
2648 	while (filter) {
2649 		if (filter->mac_index == 0 &&
2650 		    !memcmp(filter->l2_addr, bp->mac_addr,
2651 			    RTE_ETHER_ADDR_LEN)) {
2652 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2653 			if (!rc) {
2654 				STAILQ_REMOVE(&vnic->filter, filter,
2655 					      bnxt_filter_info, next);
2656 				bnxt_free_filter(bp, filter);
2657 			}
2658 			return rc;
2659 		}
2660 		filter = STAILQ_NEXT(filter, next);
2661 	}
2662 	return 0;
2663 }
2664 
2665 static int
2666 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2667 {
2668 	struct bnxt_vnic_info *vnic;
2669 	unsigned int i;
2670 	int rc;
2671 
2672 	vnic = bnxt_get_default_vnic(bp);
2673 	if (!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)) {
2674 		/* Remove any VLAN filters programmed */
2675 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2676 			bnxt_del_vlan_filter(bp, i);
2677 
2678 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2679 		if (rc)
2680 			return rc;
2681 	} else {
2682 		/* Default filter will allow packets that match the
2683 		 * dest mac. So, it has to be deleted, otherwise, we
2684 		 * will endup receiving vlan packets for which the
2685 		 * filter is not programmed, when hw-vlan-filter
2686 		 * configuration is ON
2687 		 */
2688 		bnxt_del_dflt_mac_filter(bp, vnic);
2689 		/* This filter will allow only untagged packets */
2690 		bnxt_add_vlan_filter(bp, 0);
2691 	}
2692 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2693 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER));
2694 
2695 	return 0;
2696 }
2697 
2698 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2699 {
2700 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2701 	unsigned int i;
2702 	int rc;
2703 
2704 	/* Destroy vnic filters and vnic */
2705 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2706 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2707 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2708 			bnxt_del_vlan_filter(bp, i);
2709 	}
2710 	bnxt_del_dflt_mac_filter(bp, vnic);
2711 
2712 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2713 	if (rc)
2714 		return rc;
2715 
2716 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2717 	if (rc)
2718 		return rc;
2719 
2720 	rte_free(vnic->fw_grp_ids);
2721 	vnic->fw_grp_ids = NULL;
2722 
2723 	vnic->rx_queue_cnt = 0;
2724 
2725 	return 0;
2726 }
2727 
2728 static int
2729 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2730 {
2731 	struct bnxt_vnic_info *vnic = bnxt_get_default_vnic(bp);
2732 	int rc;
2733 
2734 	/* Destroy, recreate and reconfigure the default vnic */
2735 	rc = bnxt_free_one_vnic(bp, bp->vnic_queue_db.dflt_vnic_id);
2736 	if (rc)
2737 		return rc;
2738 
2739 	/* setup the default vnic details*/
2740 	bnxt_vnic_queue_db_update_dlft_vnic(bp);
2741 
2742 	rc = bnxt_setup_one_vnic(bp, bp->vnic_queue_db.dflt_vnic_id);
2743 	if (rc)
2744 		return rc;
2745 
2746 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2747 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2748 		rc = bnxt_add_vlan_filter(bp, 0);
2749 		if (rc)
2750 			return rc;
2751 		rc = bnxt_restore_vlan_filters(bp);
2752 		if (rc)
2753 			return rc;
2754 	} else {
2755 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2756 		if (rc)
2757 			return rc;
2758 	}
2759 
2760 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2761 	if (rc)
2762 		return rc;
2763 
2764 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2765 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP));
2766 
2767 	return rc;
2768 }
2769 
2770 static int
2771 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2772 {
2773 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2774 	struct bnxt *bp = dev->data->dev_private;
2775 	int rc;
2776 
2777 	rc = is_bnxt_in_error(bp);
2778 	if (rc)
2779 		return rc;
2780 
2781 	/* Filter settings will get applied when port is started */
2782 	if (!dev->data->dev_started)
2783 		return 0;
2784 
2785 	if (mask & RTE_ETH_VLAN_FILTER_MASK) {
2786 		/* Enable or disable VLAN filtering */
2787 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2788 		if (rc)
2789 			return rc;
2790 	}
2791 
2792 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
2793 		/* Enable or disable VLAN stripping */
2794 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2795 		if (rc)
2796 			return rc;
2797 	}
2798 
2799 	if (mask & RTE_ETH_VLAN_EXTEND_MASK) {
2800 		if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND)
2801 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2802 		else
2803 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2804 	}
2805 
2806 	return 0;
2807 }
2808 
2809 static int
2810 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2811 		      uint16_t tpid)
2812 {
2813 	struct bnxt *bp = dev->data->dev_private;
2814 	int qinq = dev->data->dev_conf.rxmode.offloads &
2815 		   RTE_ETH_RX_OFFLOAD_VLAN_EXTEND;
2816 
2817 	if (vlan_type != RTE_ETH_VLAN_TYPE_INNER &&
2818 	    vlan_type != RTE_ETH_VLAN_TYPE_OUTER) {
2819 		PMD_DRV_LOG(ERR,
2820 			    "Unsupported vlan type.");
2821 		return -EINVAL;
2822 	}
2823 	if (!qinq) {
2824 		PMD_DRV_LOG(ERR,
2825 			    "QinQ not enabled. Needs to be ON as we can "
2826 			    "accelerate only outer vlan\n");
2827 		return -EINVAL;
2828 	}
2829 
2830 	if (vlan_type == RTE_ETH_VLAN_TYPE_OUTER) {
2831 		switch (tpid) {
2832 		case RTE_ETHER_TYPE_QINQ:
2833 			bp->outer_tpid_bd =
2834 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2835 				break;
2836 		case RTE_ETHER_TYPE_VLAN:
2837 			bp->outer_tpid_bd =
2838 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2839 				break;
2840 		case RTE_ETHER_TYPE_QINQ1:
2841 			bp->outer_tpid_bd =
2842 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2843 				break;
2844 		case RTE_ETHER_TYPE_QINQ2:
2845 			bp->outer_tpid_bd =
2846 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2847 				break;
2848 		case RTE_ETHER_TYPE_QINQ3:
2849 			bp->outer_tpid_bd =
2850 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2851 				break;
2852 		default:
2853 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2854 			return -EINVAL;
2855 		}
2856 		bp->outer_tpid_bd |= tpid;
2857 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2858 	} else if (vlan_type == RTE_ETH_VLAN_TYPE_INNER) {
2859 		PMD_DRV_LOG(ERR,
2860 			    "Can accelerate only outer vlan in QinQ\n");
2861 		return -EINVAL;
2862 	}
2863 
2864 	return 0;
2865 }
2866 
2867 static int
2868 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2869 			     struct rte_ether_addr *addr)
2870 {
2871 	struct bnxt *bp = dev->data->dev_private;
2872 	/* Default Filter is tied to VNIC 0 */
2873 	struct bnxt_vnic_info *vnic = bnxt_get_default_vnic(bp);
2874 	int rc;
2875 
2876 	rc = is_bnxt_in_error(bp);
2877 	if (rc)
2878 		return rc;
2879 
2880 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2881 		return -EPERM;
2882 
2883 	if (rte_is_zero_ether_addr(addr))
2884 		return -EINVAL;
2885 
2886 	/* Filter settings will get applied when port is started */
2887 	if (!dev->data->dev_started)
2888 		return 0;
2889 
2890 	/* Check if the requested MAC is already added */
2891 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2892 		return 0;
2893 
2894 	/* Destroy filter and re-create it */
2895 	bnxt_del_dflt_mac_filter(bp, vnic);
2896 
2897 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2898 	if (dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2899 		/* This filter will allow only untagged packets */
2900 		rc = bnxt_add_vlan_filter(bp, 0);
2901 	} else {
2902 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2903 	}
2904 
2905 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2906 	return rc;
2907 }
2908 
2909 static int
2910 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2911 			  struct rte_ether_addr *mc_addr_set,
2912 			  uint32_t nb_mc_addr)
2913 {
2914 	struct bnxt *bp = eth_dev->data->dev_private;
2915 	struct bnxt_vnic_info *vnic;
2916 	uint32_t i = 0;
2917 	int rc;
2918 
2919 	rc = is_bnxt_in_error(bp);
2920 	if (rc)
2921 		return rc;
2922 
2923 	vnic = bnxt_get_default_vnic(bp);
2924 
2925 	bp->nb_mc_addr = nb_mc_addr;
2926 
2927 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2928 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2929 		goto allmulti;
2930 	}
2931 
2932 	/* TODO Check for Duplicate mcast addresses */
2933 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2934 	for (i = 0; i < nb_mc_addr; i++)
2935 		rte_ether_addr_copy(&mc_addr_set[i], &bp->mcast_addr_list[i]);
2936 
2937 	if (bp->nb_mc_addr)
2938 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2939 	else
2940 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2941 
2942 allmulti:
2943 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2944 }
2945 
2946 static int
2947 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2948 {
2949 	struct bnxt *bp = dev->data->dev_private;
2950 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2951 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2952 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2953 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2954 	int ret;
2955 
2956 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2957 			fw_major, fw_minor, fw_updt, fw_rsvd);
2958 	if (ret < 0)
2959 		return -EINVAL;
2960 
2961 	ret += 1; /* add the size of '\0' */
2962 	if (fw_size < (size_t)ret)
2963 		return ret;
2964 	else
2965 		return 0;
2966 }
2967 
2968 static void
2969 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2970 	struct rte_eth_rxq_info *qinfo)
2971 {
2972 	struct bnxt *bp = dev->data->dev_private;
2973 	struct bnxt_rx_queue *rxq;
2974 
2975 	if (is_bnxt_in_error(bp))
2976 		return;
2977 
2978 	rxq = dev->data->rx_queues[queue_id];
2979 
2980 	qinfo->mp = rxq->mb_pool;
2981 	qinfo->scattered_rx = dev->data->scattered_rx;
2982 	qinfo->nb_desc = rxq->nb_rx_desc;
2983 
2984 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2985 	qinfo->conf.rx_drop_en = rxq->drop_en;
2986 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2987 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2988 }
2989 
2990 static void
2991 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2992 	struct rte_eth_txq_info *qinfo)
2993 {
2994 	struct bnxt *bp = dev->data->dev_private;
2995 	struct bnxt_tx_queue *txq;
2996 
2997 	if (is_bnxt_in_error(bp))
2998 		return;
2999 
3000 	txq = dev->data->tx_queues[queue_id];
3001 
3002 	qinfo->nb_desc = txq->nb_tx_desc;
3003 
3004 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
3005 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
3006 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
3007 
3008 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
3009 	qinfo->conf.tx_rs_thresh = 0;
3010 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
3011 	qinfo->conf.offloads = txq->offloads;
3012 }
3013 
3014 static const struct {
3015 	eth_rx_burst_t pkt_burst;
3016 	const char *info;
3017 } bnxt_rx_burst_info[] = {
3018 	{bnxt_recv_pkts,		"Scalar"},
3019 #if defined(RTE_ARCH_X86)
3020 	{bnxt_recv_pkts_vec,		"Vector SSE"},
3021 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
3022 #endif
3023 #if defined(RTE_ARCH_ARM64)
3024 	{bnxt_recv_pkts_vec,		"Vector Neon"},
3025 #endif
3026 };
3027 
3028 static int
3029 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
3030 		       struct rte_eth_burst_mode *mode)
3031 {
3032 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
3033 	size_t i;
3034 
3035 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
3036 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
3037 			snprintf(mode->info, sizeof(mode->info), "%s",
3038 				 bnxt_rx_burst_info[i].info);
3039 			return 0;
3040 		}
3041 	}
3042 
3043 	return -EINVAL;
3044 }
3045 
3046 static const struct {
3047 	eth_tx_burst_t pkt_burst;
3048 	const char *info;
3049 } bnxt_tx_burst_info[] = {
3050 	{bnxt_xmit_pkts,		"Scalar"},
3051 #if defined(RTE_ARCH_X86)
3052 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
3053 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
3054 #endif
3055 #if defined(RTE_ARCH_ARM64)
3056 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
3057 #endif
3058 };
3059 
3060 static int
3061 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
3062 		       struct rte_eth_burst_mode *mode)
3063 {
3064 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
3065 	size_t i;
3066 
3067 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
3068 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
3069 			snprintf(mode->info, sizeof(mode->info), "%s",
3070 				 bnxt_tx_burst_info[i].info);
3071 			return 0;
3072 		}
3073 	}
3074 
3075 	return -EINVAL;
3076 }
3077 
3078 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3079 {
3080 	struct bnxt *bp = eth_dev->data->dev_private;
3081 	uint32_t rc = 0;
3082 
3083 	rc = is_bnxt_in_error(bp);
3084 	if (rc)
3085 		return rc;
3086 
3087 	/* Return if port is active */
3088 	if (eth_dev->data->dev_started) {
3089 		PMD_DRV_LOG(ERR, "Stop port before changing MTU\n");
3090 		return -EBUSY;
3091 	}
3092 
3093 	/* Exit if receive queues are not configured yet */
3094 	if (!eth_dev->data->nb_rx_queues)
3095 		return -ENOTSUP;
3096 
3097 	/* Is there a change in mtu setting? */
3098 	if (eth_dev->data->mtu == new_mtu)
3099 		return rc;
3100 
3101 	if (new_mtu > RTE_ETHER_MTU)
3102 		bp->flags |= BNXT_FLAG_JUMBO;
3103 	else
3104 		bp->flags &= ~BNXT_FLAG_JUMBO;
3105 
3106 	rc = bnxt_vnic_mru_config(bp, new_mtu);
3107 	if (rc) {
3108 		PMD_DRV_LOG(ERR, "failed to update mtu in vnic context\n");
3109 		return rc;
3110 	}
3111 
3112 	if (bnxt_hwrm_config_host_mtu(bp))
3113 		PMD_DRV_LOG(WARNING, "Failed to configure host MTU\n");
3114 
3115 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3116 
3117 	return rc;
3118 }
3119 
3120 static int
3121 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3122 {
3123 	struct bnxt *bp = dev->data->dev_private;
3124 	uint16_t vlan = bp->vlan;
3125 	int rc;
3126 
3127 	rc = is_bnxt_in_error(bp);
3128 	if (rc)
3129 		return rc;
3130 
3131 	if (!BNXT_SINGLE_PF(bp)) {
3132 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3133 		return -ENOTSUP;
3134 	}
3135 	bp->vlan = on ? pvid : 0;
3136 
3137 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3138 	if (rc)
3139 		bp->vlan = vlan;
3140 	return rc;
3141 }
3142 
3143 static int
3144 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3145 {
3146 	struct bnxt *bp = dev->data->dev_private;
3147 	int rc;
3148 
3149 	rc = is_bnxt_in_error(bp);
3150 	if (rc)
3151 		return rc;
3152 
3153 	return bnxt_hwrm_port_led_cfg(bp, true);
3154 }
3155 
3156 static int
3157 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3158 {
3159 	struct bnxt *bp = dev->data->dev_private;
3160 	int rc;
3161 
3162 	rc = is_bnxt_in_error(bp);
3163 	if (rc)
3164 		return rc;
3165 
3166 	return bnxt_hwrm_port_led_cfg(bp, false);
3167 }
3168 
3169 static uint32_t
3170 bnxt_rx_queue_count_op(void *rx_queue)
3171 {
3172 	struct bnxt *bp;
3173 	struct bnxt_cp_ring_info *cpr;
3174 	uint32_t desc = 0, raw_cons, cp_ring_size;
3175 	struct bnxt_rx_queue *rxq;
3176 	struct rx_pkt_cmpl *rxcmp;
3177 	int rc;
3178 
3179 	rxq = rx_queue;
3180 	bp = rxq->bp;
3181 
3182 	rc = is_bnxt_in_error(bp);
3183 	if (rc)
3184 		return rc;
3185 
3186 	cpr = rxq->cp_ring;
3187 	raw_cons = cpr->cp_raw_cons;
3188 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3189 
3190 	while (1) {
3191 		uint32_t agg_cnt, cons, cmpl_type;
3192 
3193 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3194 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3195 
3196 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3197 			break;
3198 
3199 		cmpl_type = CMP_TYPE(rxcmp);
3200 
3201 		switch (cmpl_type) {
3202 		case CMPL_BASE_TYPE_RX_L2:
3203 		case CMPL_BASE_TYPE_RX_L2_V2:
3204 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3205 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3206 			desc++;
3207 			break;
3208 
3209 		case CMPL_BASE_TYPE_RX_TPA_END:
3210 			if (BNXT_CHIP_P5(rxq->bp)) {
3211 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3212 
3213 				p5_tpa_end = (void *)rxcmp;
3214 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3215 			} else {
3216 				struct rx_tpa_end_cmpl *tpa_end;
3217 
3218 				tpa_end = (void *)rxcmp;
3219 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3220 			}
3221 
3222 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3223 			desc++;
3224 			break;
3225 
3226 		default:
3227 			raw_cons += CMP_LEN(cmpl_type);
3228 		}
3229 	}
3230 
3231 	return desc;
3232 }
3233 
3234 static int
3235 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3236 {
3237 	struct bnxt_rx_queue *rxq = rx_queue;
3238 	struct bnxt_cp_ring_info *cpr;
3239 	struct bnxt_rx_ring_info *rxr;
3240 	uint32_t desc, raw_cons, cp_ring_size;
3241 	struct bnxt *bp = rxq->bp;
3242 	struct rx_pkt_cmpl *rxcmp;
3243 	int rc;
3244 
3245 	rc = is_bnxt_in_error(bp);
3246 	if (rc)
3247 		return rc;
3248 
3249 	if (offset >= rxq->nb_rx_desc)
3250 		return -EINVAL;
3251 
3252 	rxr = rxq->rx_ring;
3253 	cpr = rxq->cp_ring;
3254 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3255 
3256 	/*
3257 	 * For the vector receive case, the completion at the requested
3258 	 * offset can be indexed directly.
3259 	 */
3260 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3261 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3262 		struct rx_pkt_cmpl *rxcmp;
3263 		uint32_t cons;
3264 
3265 		/* Check status of completion descriptor. */
3266 		raw_cons = cpr->cp_raw_cons +
3267 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3268 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3269 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3270 
3271 		if (bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3272 			return RTE_ETH_RX_DESC_DONE;
3273 
3274 		/* Check whether rx desc has an mbuf attached. */
3275 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3276 		if (cons >= rxq->rxrearm_start &&
3277 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3278 			return RTE_ETH_RX_DESC_UNAVAIL;
3279 		}
3280 
3281 		return RTE_ETH_RX_DESC_AVAIL;
3282 	}
3283 #endif
3284 
3285 	/*
3286 	 * For the non-vector receive case, scan the completion ring to
3287 	 * locate the completion descriptor for the requested offset.
3288 	 */
3289 	raw_cons = cpr->cp_raw_cons;
3290 	desc = 0;
3291 	while (1) {
3292 		uint32_t agg_cnt, cons, cmpl_type;
3293 
3294 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3295 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3296 
3297 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3298 			break;
3299 
3300 		cmpl_type = CMP_TYPE(rxcmp);
3301 
3302 		switch (cmpl_type) {
3303 		case CMPL_BASE_TYPE_RX_L2:
3304 		case CMPL_BASE_TYPE_RX_L2_V2:
3305 			if (desc == offset) {
3306 				cons = rxcmp->opaque;
3307 				if (rxr->rx_buf_ring[cons])
3308 					return RTE_ETH_RX_DESC_DONE;
3309 				else
3310 					return RTE_ETH_RX_DESC_UNAVAIL;
3311 			}
3312 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3313 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3314 			desc++;
3315 			break;
3316 
3317 		case CMPL_BASE_TYPE_RX_TPA_END:
3318 			if (desc == offset)
3319 				return RTE_ETH_RX_DESC_DONE;
3320 
3321 			if (BNXT_CHIP_P5(rxq->bp)) {
3322 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3323 
3324 				p5_tpa_end = (void *)rxcmp;
3325 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3326 			} else {
3327 				struct rx_tpa_end_cmpl *tpa_end;
3328 
3329 				tpa_end = (void *)rxcmp;
3330 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3331 			}
3332 
3333 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3334 			desc++;
3335 			break;
3336 
3337 		default:
3338 			raw_cons += CMP_LEN(cmpl_type);
3339 		}
3340 	}
3341 
3342 	return RTE_ETH_RX_DESC_AVAIL;
3343 }
3344 
3345 static int
3346 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3347 {
3348 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3349 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3350 	uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3351 	struct cmpl_base *cp_desc_ring;
3352 	int rc;
3353 
3354 	rc = is_bnxt_in_error(txq->bp);
3355 	if (rc)
3356 		return rc;
3357 
3358 	if (offset >= txq->nb_tx_desc)
3359 		return -EINVAL;
3360 
3361 	/* Return "desc done" if descriptor is available for use. */
3362 	if (bnxt_tx_bds_in_hw(txq) <= offset)
3363 		return RTE_ETH_TX_DESC_DONE;
3364 
3365 	raw_cons = cpr->cp_raw_cons;
3366 	cp_desc_ring = cpr->cp_desc_ring;
3367 	ring_mask = cpr->cp_ring_struct->ring_mask;
3368 
3369 	/* Check to see if hw has posted a completion for the descriptor. */
3370 	while (1) {
3371 		struct tx_cmpl *txcmp;
3372 		uint32_t cons;
3373 
3374 		cons = RING_CMPL(ring_mask, raw_cons);
3375 		txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3376 
3377 		if (!bnxt_cpr_cmp_valid(txcmp, raw_cons, ring_mask + 1))
3378 			break;
3379 
3380 		if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3381 			nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3382 
3383 		if (nb_tx_pkts > offset)
3384 			return RTE_ETH_TX_DESC_DONE;
3385 
3386 		raw_cons = NEXT_RAW_CMP(raw_cons);
3387 	}
3388 
3389 	/* Descriptor is pending transmit, not yet completed by hardware. */
3390 	return RTE_ETH_TX_DESC_FULL;
3391 }
3392 
3393 int
3394 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3395 		     const struct rte_flow_ops **ops)
3396 {
3397 	struct bnxt *bp = dev->data->dev_private;
3398 	int ret = 0;
3399 
3400 	if (!bp)
3401 		return -EIO;
3402 
3403 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3404 		struct bnxt_representor *vfr = dev->data->dev_private;
3405 		bp = vfr->parent_dev->data->dev_private;
3406 		/* parent is deleted while children are still valid */
3407 		if (!bp) {
3408 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3409 				    dev->data->port_id);
3410 			return -EIO;
3411 		}
3412 	}
3413 
3414 	ret = is_bnxt_in_error(bp);
3415 	if (ret)
3416 		return ret;
3417 
3418 	/* PMD supports thread-safe flow operations.  rte_flow API
3419 	 * functions can avoid mutex for multi-thread safety.
3420 	 */
3421 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3422 
3423 	if (BNXT_TRUFLOW_EN(bp))
3424 		*ops = &bnxt_ulp_rte_flow_ops;
3425 	else
3426 		*ops = &bnxt_flow_ops;
3427 
3428 	return ret;
3429 }
3430 
3431 static const uint32_t *
3432 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3433 {
3434 	static const uint32_t ptypes[] = {
3435 		RTE_PTYPE_L2_ETHER_VLAN,
3436 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3437 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3438 		RTE_PTYPE_L4_ICMP,
3439 		RTE_PTYPE_L4_TCP,
3440 		RTE_PTYPE_L4_UDP,
3441 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3442 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3443 		RTE_PTYPE_INNER_L4_ICMP,
3444 		RTE_PTYPE_INNER_L4_TCP,
3445 		RTE_PTYPE_INNER_L4_UDP,
3446 		RTE_PTYPE_UNKNOWN
3447 	};
3448 
3449 	if (!dev->rx_pkt_burst)
3450 		return NULL;
3451 
3452 	return ptypes;
3453 }
3454 
3455 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3456 			 int reg_win)
3457 {
3458 	uint32_t reg_base = *reg_arr & 0xfffff000;
3459 	uint32_t win_off;
3460 	int i;
3461 
3462 	for (i = 0; i < count; i++) {
3463 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3464 			return -ERANGE;
3465 	}
3466 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3467 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3468 	return 0;
3469 }
3470 
3471 static int bnxt_map_ptp_regs(struct bnxt *bp)
3472 {
3473 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3474 	uint32_t *reg_arr;
3475 	int rc, i;
3476 
3477 	reg_arr = ptp->rx_regs;
3478 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3479 	if (rc)
3480 		return rc;
3481 
3482 	reg_arr = ptp->tx_regs;
3483 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3484 	if (rc)
3485 		return rc;
3486 
3487 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3488 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3489 
3490 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3491 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3492 
3493 	return 0;
3494 }
3495 
3496 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3497 {
3498 	rte_write32(0, (uint8_t *)bp->bar0 +
3499 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3500 	rte_write32(0, (uint8_t *)bp->bar0 +
3501 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3502 }
3503 
3504 static uint64_t bnxt_cc_read(struct bnxt *bp)
3505 {
3506 	uint64_t ns;
3507 
3508 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3509 			      BNXT_GRCPF_REG_SYNC_TIME));
3510 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3511 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3512 	return ns;
3513 }
3514 
3515 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3516 {
3517 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3518 	uint32_t fifo;
3519 
3520 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3521 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3522 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3523 		return -EAGAIN;
3524 
3525 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3526 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3527 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3528 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3529 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3530 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3531 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3532 
3533 	return 0;
3534 }
3535 
3536 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3537 {
3538 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3539 	struct bnxt_pf_info *pf = bp->pf;
3540 	uint16_t port_id;
3541 	int i = 0;
3542 	uint32_t fifo;
3543 
3544 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3545 		return -EINVAL;
3546 
3547 	port_id = pf->port_id;
3548 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3549 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3550 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3551 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3552 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3553 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3554 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3555 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3556 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3557 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3558 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3559 		i++;
3560 	}
3561 
3562 	if (i >= BNXT_PTP_RX_PND_CNT)
3563 		return -EBUSY;
3564 
3565 	return 0;
3566 }
3567 
3568 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3569 {
3570 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3571 	struct bnxt_pf_info *pf = bp->pf;
3572 	uint16_t port_id;
3573 	uint32_t fifo;
3574 
3575 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3576 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3577 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3578 		return -EAGAIN;
3579 
3580 	port_id = pf->port_id;
3581 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3582 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3583 
3584 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3585 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3586 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3587 		return bnxt_clr_rx_ts(bp, ts);
3588 
3589 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3590 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3591 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3592 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3593 
3594 	return 0;
3595 }
3596 
3597 static int
3598 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3599 {
3600 	uint64_t ns;
3601 	struct bnxt *bp = dev->data->dev_private;
3602 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3603 
3604 	if (!ptp)
3605 		return -ENOTSUP;
3606 
3607 	ns = rte_timespec_to_ns(ts);
3608 	/* Set the timecounters to a new value. */
3609 	ptp->tc.nsec = ns;
3610 	ptp->tx_tstamp_tc.nsec = ns;
3611 	ptp->rx_tstamp_tc.nsec = ns;
3612 
3613 	return 0;
3614 }
3615 
3616 static int
3617 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3618 {
3619 	struct bnxt *bp = dev->data->dev_private;
3620 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3621 	uint64_t ns, systime_cycles = 0;
3622 	int rc = 0;
3623 
3624 	if (!ptp)
3625 		return -ENOTSUP;
3626 
3627 	if (BNXT_CHIP_P5(bp))
3628 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3629 					     &systime_cycles);
3630 	else
3631 		systime_cycles = bnxt_cc_read(bp);
3632 
3633 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3634 	*ts = rte_ns_to_timespec(ns);
3635 
3636 	return rc;
3637 }
3638 static int
3639 bnxt_timesync_enable(struct rte_eth_dev *dev)
3640 {
3641 	struct bnxt *bp = dev->data->dev_private;
3642 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3643 	uint32_t shift = 0;
3644 	int rc;
3645 
3646 	if (!ptp)
3647 		return -ENOTSUP;
3648 
3649 	ptp->rx_filter = 1;
3650 	ptp->tx_tstamp_en = 1;
3651 	ptp->filter_all = 1;
3652 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3653 
3654 	rc = bnxt_hwrm_ptp_cfg(bp);
3655 	if (rc)
3656 		return rc;
3657 
3658 	rte_spinlock_init(&ptp->ptp_lock);
3659 	bp->ptp_all_rx_tstamp = 1;
3660 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3661 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3662 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3663 
3664 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3665 	ptp->tc.cc_shift = shift;
3666 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3667 
3668 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3669 	ptp->rx_tstamp_tc.cc_shift = shift;
3670 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3671 
3672 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3673 	ptp->tx_tstamp_tc.cc_shift = shift;
3674 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3675 
3676 	if (!BNXT_CHIP_P5(bp))
3677 		bnxt_map_ptp_regs(bp);
3678 	else
3679 		rc = bnxt_ptp_start(bp);
3680 
3681 	return rc;
3682 }
3683 
3684 static int
3685 bnxt_timesync_disable(struct rte_eth_dev *dev)
3686 {
3687 	struct bnxt *bp = dev->data->dev_private;
3688 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3689 
3690 	if (!ptp)
3691 		return -ENOTSUP;
3692 
3693 	ptp->rx_filter = 0;
3694 	ptp->tx_tstamp_en = 0;
3695 	ptp->rxctl = 0;
3696 	ptp->filter_all = 0;
3697 
3698 	bnxt_hwrm_ptp_cfg(bp);
3699 
3700 	bp->ptp_all_rx_tstamp = 0;
3701 	if (!BNXT_CHIP_P5(bp))
3702 		bnxt_unmap_ptp_regs(bp);
3703 	else
3704 		bnxt_ptp_stop(bp);
3705 
3706 	return 0;
3707 }
3708 
3709 static int
3710 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3711 				 struct timespec *timestamp,
3712 				 uint32_t flags __rte_unused)
3713 {
3714 	struct bnxt *bp = dev->data->dev_private;
3715 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3716 	uint64_t rx_tstamp_cycles = 0;
3717 	uint64_t ns;
3718 
3719 	if (!ptp)
3720 		return -ENOTSUP;
3721 
3722 	if (BNXT_CHIP_P5(bp))
3723 		rx_tstamp_cycles = ptp->rx_timestamp;
3724 	else
3725 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3726 
3727 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3728 	*timestamp = rte_ns_to_timespec(ns);
3729 	return  0;
3730 }
3731 
3732 static int
3733 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3734 				 struct timespec *timestamp)
3735 {
3736 	struct bnxt *bp = dev->data->dev_private;
3737 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3738 	uint64_t tx_tstamp_cycles = 0;
3739 	uint64_t ns;
3740 	int rc = 0;
3741 
3742 	if (!ptp)
3743 		return -ENOTSUP;
3744 
3745 	if (BNXT_CHIP_P5(bp))
3746 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3747 					     &tx_tstamp_cycles);
3748 	else
3749 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3750 
3751 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3752 	*timestamp = rte_ns_to_timespec(ns);
3753 
3754 	return rc;
3755 }
3756 
3757 static int
3758 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3759 {
3760 	struct bnxt *bp = dev->data->dev_private;
3761 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3762 
3763 	if (!ptp)
3764 		return -ENOTSUP;
3765 
3766 	ptp->tc.nsec += delta;
3767 	ptp->tx_tstamp_tc.nsec += delta;
3768 	ptp->rx_tstamp_tc.nsec += delta;
3769 
3770 	return 0;
3771 }
3772 
3773 static int
3774 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3775 {
3776 	struct bnxt *bp = dev->data->dev_private;
3777 	int rc;
3778 	uint32_t dir_entries;
3779 	uint32_t entry_length;
3780 
3781 	rc = is_bnxt_in_error(bp);
3782 	if (rc)
3783 		return rc;
3784 
3785 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3786 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3787 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3788 
3789 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3790 	if (rc != 0)
3791 		return rc;
3792 
3793 	return dir_entries * entry_length;
3794 }
3795 
3796 static int
3797 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3798 		struct rte_dev_eeprom_info *in_eeprom)
3799 {
3800 	struct bnxt *bp = dev->data->dev_private;
3801 	uint32_t index;
3802 	uint32_t offset;
3803 	int rc;
3804 
3805 	rc = is_bnxt_in_error(bp);
3806 	if (rc)
3807 		return rc;
3808 
3809 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3810 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3811 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3812 		    in_eeprom->offset, in_eeprom->length);
3813 
3814 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3815 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3816 						in_eeprom->data);
3817 
3818 	index = in_eeprom->offset >> 24;
3819 	offset = in_eeprom->offset & 0xffffff;
3820 
3821 	if (index != 0)
3822 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3823 					   in_eeprom->length, in_eeprom->data);
3824 
3825 	return 0;
3826 }
3827 
3828 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3829 {
3830 	switch (dir_type) {
3831 	case BNX_DIR_TYPE_CHIMP_PATCH:
3832 	case BNX_DIR_TYPE_BOOTCODE:
3833 	case BNX_DIR_TYPE_BOOTCODE_2:
3834 	case BNX_DIR_TYPE_APE_FW:
3835 	case BNX_DIR_TYPE_APE_PATCH:
3836 	case BNX_DIR_TYPE_KONG_FW:
3837 	case BNX_DIR_TYPE_KONG_PATCH:
3838 	case BNX_DIR_TYPE_BONO_FW:
3839 	case BNX_DIR_TYPE_BONO_PATCH:
3840 		/* FALLTHROUGH */
3841 		return true;
3842 	}
3843 
3844 	return false;
3845 }
3846 
3847 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3848 {
3849 	switch (dir_type) {
3850 	case BNX_DIR_TYPE_AVS:
3851 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3852 	case BNX_DIR_TYPE_PCIE:
3853 	case BNX_DIR_TYPE_TSCF_UCODE:
3854 	case BNX_DIR_TYPE_EXT_PHY:
3855 	case BNX_DIR_TYPE_CCM:
3856 	case BNX_DIR_TYPE_ISCSI_BOOT:
3857 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3858 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3859 		/* FALLTHROUGH */
3860 		return true;
3861 	}
3862 
3863 	return false;
3864 }
3865 
3866 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3867 {
3868 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3869 		bnxt_dir_type_is_other_exec_format(dir_type);
3870 }
3871 
3872 static int
3873 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3874 		struct rte_dev_eeprom_info *in_eeprom)
3875 {
3876 	struct bnxt *bp = dev->data->dev_private;
3877 	uint8_t index, dir_op;
3878 	uint16_t type, ext, ordinal, attr;
3879 	int rc;
3880 
3881 	rc = is_bnxt_in_error(bp);
3882 	if (rc)
3883 		return rc;
3884 
3885 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3886 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3887 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3888 		    in_eeprom->offset, in_eeprom->length);
3889 
3890 	if (!BNXT_PF(bp)) {
3891 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3892 		return -EINVAL;
3893 	}
3894 
3895 	type = in_eeprom->magic >> 16;
3896 
3897 	if (type == 0xffff) { /* special value for directory operations */
3898 		index = in_eeprom->magic & 0xff;
3899 		dir_op = in_eeprom->magic >> 8;
3900 		if (index == 0)
3901 			return -EINVAL;
3902 		switch (dir_op) {
3903 		case 0x0e: /* erase */
3904 			if (in_eeprom->offset != ~in_eeprom->magic)
3905 				return -EINVAL;
3906 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3907 		default:
3908 			return -EINVAL;
3909 		}
3910 	}
3911 
3912 	/* Create or re-write an NVM item: */
3913 	if (bnxt_dir_type_is_executable(type) == true)
3914 		return -EOPNOTSUPP;
3915 	ext = in_eeprom->magic & 0xffff;
3916 	ordinal = in_eeprom->offset >> 16;
3917 	attr = in_eeprom->offset & 0xffff;
3918 
3919 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3920 				     in_eeprom->data, in_eeprom->length);
3921 }
3922 
3923 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3924 				struct rte_eth_dev_module_info *modinfo)
3925 {
3926 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3927 	struct bnxt *bp = dev->data->dev_private;
3928 	int rc;
3929 
3930 	/* No point in going further if phy status indicates
3931 	 * module is not inserted or if it is powered down or
3932 	 * if it is of type 10GBase-T
3933 	 */
3934 	if (bp->link_info->module_status >
3935 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3936 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3937 			    dev->data->port_id);
3938 		return -ENOTSUP;
3939 	}
3940 
3941 	/* This feature is not supported in older firmware versions */
3942 	if (bp->hwrm_spec_code < 0x10202) {
3943 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3944 			    dev->data->port_id);
3945 		return -ENOTSUP;
3946 	}
3947 
3948 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3949 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3950 						   module_info);
3951 
3952 	if (rc)
3953 		return rc;
3954 
3955 	switch (module_info[0]) {
3956 	case SFF_MODULE_ID_SFP:
3957 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3958 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3959 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3960 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3961 		break;
3962 	case SFF_MODULE_ID_QSFP:
3963 	case SFF_MODULE_ID_QSFP_PLUS:
3964 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3965 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3966 		break;
3967 	case SFF_MODULE_ID_QSFP28:
3968 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3969 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3970 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3971 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3972 		break;
3973 	default:
3974 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3975 		return -ENOTSUP;
3976 	}
3977 
3978 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3979 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3980 
3981 	return 0;
3982 }
3983 
3984 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3985 				  struct rte_dev_eeprom_info *info)
3986 {
3987 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3988 	uint32_t offset = info->offset, length = info->length;
3989 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3990 	struct bnxt *bp = dev->data->dev_private;
3991 	uint8_t *data = info->data;
3992 	uint8_t page = offset >> 7;
3993 	uint8_t max_pages = 2;
3994 	uint8_t opt_pages;
3995 	int rc;
3996 
3997 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3998 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3999 						   module_info);
4000 	if (rc)
4001 		return rc;
4002 
4003 	switch (module_info[0]) {
4004 	case SFF_MODULE_ID_SFP:
4005 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
4006 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
4007 			pg_addr[2] = I2C_DEV_ADDR_A2;
4008 			pg_addr[3] = I2C_DEV_ADDR_A2;
4009 			max_pages = 4;
4010 		}
4011 		break;
4012 	case SFF_MODULE_ID_QSFP28:
4013 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
4014 							   SFF8636_OPT_PAGES_OFFSET,
4015 							   1, &opt_pages);
4016 		if (rc)
4017 			return rc;
4018 
4019 		if (opt_pages & SFF8636_PAGE1_MASK) {
4020 			pg_addr[2] = I2C_DEV_ADDR_A0;
4021 			max_pages = 3;
4022 		}
4023 		if (opt_pages & SFF8636_PAGE2_MASK) {
4024 			pg_addr[3] = I2C_DEV_ADDR_A0;
4025 			max_pages = 4;
4026 		}
4027 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
4028 			pg_addr[4] = I2C_DEV_ADDR_A0;
4029 			max_pages = 5;
4030 		}
4031 		break;
4032 	default:
4033 		break;
4034 	}
4035 
4036 	memset(data, 0, length);
4037 
4038 	offset &= 0xff;
4039 	while (length && page < max_pages) {
4040 		uint8_t raw_page = page ? page - 1 : 0;
4041 		uint16_t chunk;
4042 
4043 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
4044 			raw_page = 0;
4045 		else if (page)
4046 			offset |= 0x80;
4047 		chunk = RTE_MIN(length, 256 - offset);
4048 
4049 		if (pg_addr[page]) {
4050 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
4051 								   raw_page, offset,
4052 								   chunk, data);
4053 			if (rc)
4054 				return rc;
4055 		}
4056 
4057 		data += chunk;
4058 		length -= chunk;
4059 		offset = 0;
4060 		page += 1 + (chunk > 128);
4061 	}
4062 
4063 	return length ? -EINVAL : 0;
4064 }
4065 
4066 /*
4067  * Initialization
4068  */
4069 
4070 static const struct eth_dev_ops bnxt_dev_ops = {
4071 	.dev_infos_get = bnxt_dev_info_get_op,
4072 	.dev_close = bnxt_dev_close_op,
4073 	.dev_configure = bnxt_dev_configure_op,
4074 	.dev_start = bnxt_dev_start_op,
4075 	.dev_stop = bnxt_dev_stop_op,
4076 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4077 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4078 	.stats_get = bnxt_stats_get_op,
4079 	.stats_reset = bnxt_stats_reset_op,
4080 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4081 	.rx_queue_release = bnxt_rx_queue_release_op,
4082 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4083 	.tx_queue_release = bnxt_tx_queue_release_op,
4084 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4085 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4086 	.reta_update = bnxt_reta_update_op,
4087 	.reta_query = bnxt_reta_query_op,
4088 	.rss_hash_update = bnxt_rss_hash_update_op,
4089 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4090 	.link_update = bnxt_link_update_op,
4091 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4092 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4093 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4094 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4095 	.mac_addr_add = bnxt_mac_addr_add_op,
4096 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4097 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4098 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4099 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4100 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4101 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4102 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4103 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4104 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4105 	.mtu_set = bnxt_mtu_set_op,
4106 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4107 	.xstats_get = bnxt_dev_xstats_get_op,
4108 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4109 	.xstats_reset = bnxt_dev_xstats_reset_op,
4110 	.fw_version_get = bnxt_fw_version_get,
4111 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4112 	.rxq_info_get = bnxt_rxq_info_get_op,
4113 	.txq_info_get = bnxt_txq_info_get_op,
4114 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4115 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4116 	.dev_led_on = bnxt_dev_led_on_op,
4117 	.dev_led_off = bnxt_dev_led_off_op,
4118 	.rx_queue_start = bnxt_rx_queue_start,
4119 	.rx_queue_stop = bnxt_rx_queue_stop,
4120 	.tx_queue_start = bnxt_tx_queue_start,
4121 	.tx_queue_stop = bnxt_tx_queue_stop,
4122 	.flow_ops_get = bnxt_flow_ops_get_op,
4123 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4124 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4125 	.get_eeprom           = bnxt_get_eeprom_op,
4126 	.set_eeprom           = bnxt_set_eeprom_op,
4127 	.get_module_info = bnxt_get_module_info,
4128 	.get_module_eeprom = bnxt_get_module_eeprom,
4129 	.timesync_enable      = bnxt_timesync_enable,
4130 	.timesync_disable     = bnxt_timesync_disable,
4131 	.timesync_read_time   = bnxt_timesync_read_time,
4132 	.timesync_write_time   = bnxt_timesync_write_time,
4133 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4134 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4135 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4136 	.mtr_ops_get = bnxt_flow_meter_ops_get,
4137 };
4138 
4139 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4140 {
4141 	uint32_t offset;
4142 
4143 	/* Only pre-map the reset GRC registers using window 3 */
4144 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4145 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4146 
4147 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4148 
4149 	return offset;
4150 }
4151 
4152 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4153 {
4154 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4155 	uint32_t reg_base = 0xffffffff;
4156 	int i;
4157 
4158 	/* Only pre-map the monitoring GRC registers using window 2 */
4159 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4160 		uint32_t reg = info->status_regs[i];
4161 
4162 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4163 			continue;
4164 
4165 		if (reg_base == 0xffffffff)
4166 			reg_base = reg & 0xfffff000;
4167 		if ((reg & 0xfffff000) != reg_base)
4168 			return -ERANGE;
4169 
4170 		/* Use mask 0xffc as the Lower 2 bits indicates
4171 		 * address space location
4172 		 */
4173 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4174 						(reg & 0xffc);
4175 	}
4176 
4177 	if (reg_base == 0xffffffff)
4178 		return 0;
4179 
4180 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4181 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4182 
4183 	return 0;
4184 }
4185 
4186 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4187 {
4188 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4189 	uint32_t delay = info->delay_after_reset[index];
4190 	uint32_t val = info->reset_reg_val[index];
4191 	uint32_t reg = info->reset_reg[index];
4192 	uint32_t type, offset;
4193 	int ret;
4194 
4195 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4196 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4197 
4198 	switch (type) {
4199 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4200 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4201 		if (ret < 0) {
4202 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4203 				    val, offset);
4204 			return;
4205 		}
4206 		break;
4207 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4208 		offset = bnxt_map_reset_regs(bp, offset);
4209 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4210 		break;
4211 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4212 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4213 		break;
4214 	}
4215 	/* wait on a specific interval of time until core reset is complete */
4216 	if (delay)
4217 		rte_delay_ms(delay);
4218 }
4219 
4220 static void bnxt_dev_cleanup(struct bnxt *bp)
4221 {
4222 	bp->eth_dev->data->dev_link.link_status = 0;
4223 	bp->link_info->link_up = 0;
4224 	if (bp->eth_dev->data->dev_started)
4225 		bnxt_dev_stop(bp->eth_dev);
4226 
4227 	bnxt_uninit_resources(bp, true);
4228 }
4229 
4230 static int
4231 bnxt_check_fw_reset_done(struct bnxt *bp)
4232 {
4233 	int timeout = bp->fw_reset_max_msecs;
4234 	uint16_t val = 0;
4235 	int rc;
4236 
4237 	do {
4238 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4239 		if (rc < 0) {
4240 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4241 			return rc;
4242 		}
4243 		if (val != 0xffff)
4244 			break;
4245 		rte_delay_ms(1);
4246 	} while (timeout--);
4247 
4248 	if (val == 0xffff) {
4249 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4250 		return -1;
4251 	}
4252 
4253 	return 0;
4254 }
4255 
4256 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4257 {
4258 	struct rte_eth_dev *dev = bp->eth_dev;
4259 	struct rte_vlan_filter_conf *vfc;
4260 	int vidx, vbit, rc;
4261 	uint16_t vlan_id;
4262 
4263 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4264 		vfc = &dev->data->vlan_filter_conf;
4265 		vidx = vlan_id / 64;
4266 		vbit = vlan_id % 64;
4267 
4268 		/* Each bit corresponds to a VLAN id */
4269 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4270 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4271 			if (rc)
4272 				return rc;
4273 		}
4274 	}
4275 
4276 	return 0;
4277 }
4278 
4279 static int bnxt_restore_mac_filters(struct bnxt *bp)
4280 {
4281 	struct rte_eth_dev *dev = bp->eth_dev;
4282 	struct rte_eth_dev_info dev_info;
4283 	struct rte_ether_addr *addr;
4284 	uint64_t pool_mask;
4285 	uint32_t pool = 0;
4286 	uint32_t i;
4287 	int rc;
4288 
4289 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4290 		return 0;
4291 
4292 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4293 	if (rc)
4294 		return rc;
4295 
4296 	/* replay MAC address configuration */
4297 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4298 		addr = &dev->data->mac_addrs[i];
4299 
4300 		/* skip zero address */
4301 		if (rte_is_zero_ether_addr(addr))
4302 			continue;
4303 
4304 		pool = 0;
4305 		pool_mask = dev->data->mac_pool_sel[i];
4306 
4307 		do {
4308 			if (pool_mask & 1ULL) {
4309 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4310 				if (rc)
4311 					return rc;
4312 			}
4313 			pool_mask >>= 1;
4314 			pool++;
4315 		} while (pool_mask);
4316 	}
4317 
4318 	return 0;
4319 }
4320 
4321 static int bnxt_restore_mcast_mac_filters(struct bnxt *bp)
4322 {
4323 	int ret = 0;
4324 
4325 	ret = bnxt_dev_set_mc_addr_list_op(bp->eth_dev, bp->mcast_addr_list,
4326 					   bp->nb_mc_addr);
4327 	if (ret)
4328 		PMD_DRV_LOG(ERR, "Failed to restore multicast MAC addreeses\n");
4329 
4330 	return ret;
4331 }
4332 
4333 static int bnxt_restore_filters(struct bnxt *bp)
4334 {
4335 	struct rte_eth_dev *dev = bp->eth_dev;
4336 	int ret = 0;
4337 
4338 	if (dev->data->all_multicast) {
4339 		ret = bnxt_allmulticast_enable_op(dev);
4340 		if (ret)
4341 			return ret;
4342 	}
4343 	if (dev->data->promiscuous) {
4344 		ret = bnxt_promiscuous_enable_op(dev);
4345 		if (ret)
4346 			return ret;
4347 	}
4348 
4349 	ret = bnxt_restore_mac_filters(bp);
4350 	if (ret)
4351 		return ret;
4352 
4353 	/* if vlans are already programmed, this can fail with -EEXIST */
4354 	ret = bnxt_restore_vlan_filters(bp);
4355 	if (ret && ret != -EEXIST)
4356 		return ret;
4357 
4358 	ret = bnxt_restore_mcast_mac_filters(bp);
4359 	if (ret)
4360 		return ret;
4361 
4362 	return ret;
4363 }
4364 
4365 static int bnxt_check_fw_ready(struct bnxt *bp)
4366 {
4367 	int timeout = bp->fw_reset_max_msecs ? : BNXT_MAX_FW_RESET_TIMEOUT;
4368 	int rc = 0;
4369 
4370 	do {
4371 		rc = bnxt_hwrm_poll_ver_get(bp);
4372 		if (rc == 0)
4373 			break;
4374 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4375 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4376 	} while (rc && timeout > 0);
4377 
4378 	if (rc)
4379 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4380 
4381 	return rc;
4382 }
4383 
4384 static void bnxt_dev_recover(void *arg)
4385 {
4386 	struct bnxt *bp = arg;
4387 	int rc = 0;
4388 
4389 	pthread_mutex_lock(&bp->err_recovery_lock);
4390 
4391 	if (!bp->fw_reset_min_msecs) {
4392 		rc = bnxt_check_fw_reset_done(bp);
4393 		if (rc)
4394 			goto err;
4395 	}
4396 
4397 	/* Clear Error flag so that device re-init should happen */
4398 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4399 	PMD_DRV_LOG(INFO, "Port: %u Starting recovery...\n",
4400 		    bp->eth_dev->data->port_id);
4401 
4402 	rc = bnxt_check_fw_ready(bp);
4403 	if (rc)
4404 		goto err;
4405 
4406 	rc = bnxt_init_resources(bp, true);
4407 	if (rc) {
4408 		PMD_DRV_LOG(ERR,
4409 			    "Failed to initialize resources after reset\n");
4410 		goto err;
4411 	}
4412 	/* clear reset flag as the device is initialized now */
4413 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4414 
4415 	rc = bnxt_dev_start_op(bp->eth_dev);
4416 	if (rc) {
4417 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4418 		goto err_start;
4419 	}
4420 
4421 	rc = bnxt_restore_filters(bp);
4422 	if (rc)
4423 		goto err_start;
4424 
4425 	rte_eth_fp_ops[bp->eth_dev->data->port_id].rx_pkt_burst =
4426 		bp->eth_dev->rx_pkt_burst;
4427 	rte_eth_fp_ops[bp->eth_dev->data->port_id].tx_pkt_burst =
4428 		bp->eth_dev->tx_pkt_burst;
4429 	rte_mb();
4430 
4431 	PMD_DRV_LOG(INFO, "Port: %u Recovered from FW reset\n",
4432 		    bp->eth_dev->data->port_id);
4433 	pthread_mutex_unlock(&bp->err_recovery_lock);
4434 	rte_eth_dev_callback_process(bp->eth_dev,
4435 				     RTE_ETH_EVENT_RECOVERY_SUCCESS,
4436 				     NULL);
4437 	return;
4438 err_start:
4439 	bnxt_dev_stop(bp->eth_dev);
4440 err:
4441 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4442 	bnxt_uninit_resources(bp, false);
4443 	rte_eth_dev_callback_process(bp->eth_dev,
4444 				     RTE_ETH_EVENT_RECOVERY_FAILED,
4445 				     NULL);
4446 	if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4447 		rte_eth_dev_callback_process(bp->eth_dev,
4448 					     RTE_ETH_EVENT_INTR_RMV,
4449 					     NULL);
4450 	pthread_mutex_unlock(&bp->err_recovery_lock);
4451 	PMD_DRV_LOG(ERR, "Port %u: Failed to recover from FW reset\n",
4452 		    bp->eth_dev->data->port_id);
4453 }
4454 
4455 void bnxt_dev_reset_and_resume(void *arg)
4456 {
4457 	struct bnxt *bp = arg;
4458 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4459 	uint16_t val = 0;
4460 	int rc;
4461 
4462 	bnxt_dev_cleanup(bp);
4463 	PMD_DRV_LOG(INFO, "Port: %u Finished bnxt_dev_cleanup\n",
4464 		    bp->eth_dev->data->port_id);
4465 
4466 	bnxt_wait_for_device_shutdown(bp);
4467 
4468 	/* During some fatal firmware error conditions, the PCI config space
4469 	 * register 0x2e which normally contains the subsystem ID will become
4470 	 * 0xffff. This register will revert back to the normal value after
4471 	 * the chip has completed core reset. If we detect this condition,
4472 	 * we can poll this config register immediately for the value to revert.
4473 	 */
4474 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4475 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4476 		if (rc < 0) {
4477 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4478 			return;
4479 		}
4480 		if (val == 0xffff) {
4481 			bp->fw_reset_min_msecs = 0;
4482 			us = 1;
4483 		}
4484 	}
4485 
4486 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4487 	if (rc)
4488 		PMD_DRV_LOG(ERR, "Port %u: Error setting recovery alarm",
4489 			    bp->eth_dev->data->port_id);
4490 }
4491 
4492 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4493 {
4494 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4495 	uint32_t reg = info->status_regs[index];
4496 	uint32_t type, offset, val = 0;
4497 	int ret = 0;
4498 
4499 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4500 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4501 
4502 	switch (type) {
4503 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4504 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4505 		if (ret < 0)
4506 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4507 				    offset);
4508 		break;
4509 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4510 		offset = info->mapped_status_regs[index];
4511 		/* FALLTHROUGH */
4512 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4513 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4514 				       offset));
4515 		break;
4516 	}
4517 
4518 	return val;
4519 }
4520 
4521 static int bnxt_fw_reset_all(struct bnxt *bp)
4522 {
4523 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4524 	uint32_t i;
4525 	int rc = 0;
4526 
4527 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4528 		/* Reset through primary function driver */
4529 		for (i = 0; i < info->reg_array_cnt; i++)
4530 			bnxt_write_fw_reset_reg(bp, i);
4531 		/* Wait for time specified by FW after triggering reset */
4532 		rte_delay_ms(info->primary_func_wait_period_after_reset);
4533 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4534 		/* Reset with the help of Kong processor */
4535 		rc = bnxt_hwrm_fw_reset(bp);
4536 		if (rc)
4537 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4538 	}
4539 
4540 	return rc;
4541 }
4542 
4543 static void bnxt_fw_reset_cb(void *arg)
4544 {
4545 	struct bnxt *bp = arg;
4546 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4547 	int rc = 0;
4548 
4549 	/* Only Primary function can do FW reset */
4550 	if (bnxt_is_primary_func(bp) &&
4551 	    bnxt_is_recovery_enabled(bp)) {
4552 		rc = bnxt_fw_reset_all(bp);
4553 		if (rc) {
4554 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4555 			return;
4556 		}
4557 	}
4558 
4559 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4560 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4561 	 * (including MASTER FUNC). After receiving this Async, all the active
4562 	 * drivers should treat this case as FW initiated recovery
4563 	 */
4564 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4565 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4566 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4567 
4568 		/* To recover from error */
4569 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4570 				  (void *)bp);
4571 	}
4572 }
4573 
4574 /* Driver should poll FW heartbeat, reset_counter with the frequency
4575  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4576  * When the driver detects heartbeat stop or change in reset_counter,
4577  * it has to trigger a reset to recover from the error condition.
4578  * A “primary function” is the function who will have the privilege to
4579  * initiate the chimp reset. The primary function will be elected by the
4580  * firmware and will be notified through async message.
4581  */
4582 static void bnxt_check_fw_health(void *arg)
4583 {
4584 	struct bnxt *bp = arg;
4585 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4586 	uint32_t val = 0, wait_msec;
4587 
4588 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4589 	    is_bnxt_in_error(bp))
4590 		return;
4591 
4592 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4593 	if (val == info->last_heart_beat)
4594 		goto reset;
4595 
4596 	info->last_heart_beat = val;
4597 
4598 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4599 	if (val != info->last_reset_counter)
4600 		goto reset;
4601 
4602 	info->last_reset_counter = val;
4603 
4604 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4605 			  bnxt_check_fw_health, (void *)bp);
4606 
4607 	return;
4608 reset:
4609 	/* Stop DMA to/from device */
4610 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4611 	bp->flags |= BNXT_FLAG_FW_RESET;
4612 
4613 	bnxt_stop_rxtx(bp->eth_dev);
4614 
4615 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4616 
4617 	rte_eth_dev_callback_process(bp->eth_dev,
4618 				     RTE_ETH_EVENT_ERR_RECOVERING,
4619 				     NULL);
4620 
4621 	if (bnxt_is_primary_func(bp))
4622 		wait_msec = info->primary_func_wait_period;
4623 	else
4624 		wait_msec = info->normal_func_wait_period;
4625 
4626 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4627 			  bnxt_fw_reset_cb, (void *)bp);
4628 }
4629 
4630 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4631 {
4632 	uint32_t polling_freq;
4633 
4634 	pthread_mutex_lock(&bp->health_check_lock);
4635 
4636 	if (!bnxt_is_recovery_enabled(bp))
4637 		goto done;
4638 
4639 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4640 		goto done;
4641 
4642 	polling_freq = bp->recovery_info->driver_polling_freq;
4643 
4644 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4645 			  bnxt_check_fw_health, (void *)bp);
4646 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4647 
4648 done:
4649 	pthread_mutex_unlock(&bp->health_check_lock);
4650 }
4651 
4652 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4653 {
4654 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4655 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4656 }
4657 
4658 static bool bnxt_vf_pciid(uint16_t device_id)
4659 {
4660 	switch (device_id) {
4661 	case BROADCOM_DEV_ID_57304_VF:
4662 	case BROADCOM_DEV_ID_57406_VF:
4663 	case BROADCOM_DEV_ID_5731X_VF:
4664 	case BROADCOM_DEV_ID_5741X_VF:
4665 	case BROADCOM_DEV_ID_57414_VF:
4666 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4667 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4668 	case BROADCOM_DEV_ID_58802_VF:
4669 	case BROADCOM_DEV_ID_57500_VF1:
4670 	case BROADCOM_DEV_ID_57500_VF2:
4671 	case BROADCOM_DEV_ID_58818_VF:
4672 		/* FALLTHROUGH */
4673 		return true;
4674 	default:
4675 		return false;
4676 	}
4677 }
4678 
4679 /* Phase 5 device */
4680 static bool bnxt_p5_device(uint16_t device_id)
4681 {
4682 	switch (device_id) {
4683 	case BROADCOM_DEV_ID_57508:
4684 	case BROADCOM_DEV_ID_57504:
4685 	case BROADCOM_DEV_ID_57502:
4686 	case BROADCOM_DEV_ID_57508_MF1:
4687 	case BROADCOM_DEV_ID_57504_MF1:
4688 	case BROADCOM_DEV_ID_57502_MF1:
4689 	case BROADCOM_DEV_ID_57508_MF2:
4690 	case BROADCOM_DEV_ID_57504_MF2:
4691 	case BROADCOM_DEV_ID_57502_MF2:
4692 	case BROADCOM_DEV_ID_57500_VF1:
4693 	case BROADCOM_DEV_ID_57500_VF2:
4694 	case BROADCOM_DEV_ID_58812:
4695 	case BROADCOM_DEV_ID_58814:
4696 	case BROADCOM_DEV_ID_58818:
4697 	case BROADCOM_DEV_ID_58818_VF:
4698 		/* FALLTHROUGH */
4699 		return true;
4700 	default:
4701 		return false;
4702 	}
4703 }
4704 
4705 bool bnxt_stratus_device(struct bnxt *bp)
4706 {
4707 	uint16_t device_id = bp->pdev->id.device_id;
4708 
4709 	switch (device_id) {
4710 	case BROADCOM_DEV_ID_STRATUS_NIC:
4711 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4712 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4713 		/* FALLTHROUGH */
4714 		return true;
4715 	default:
4716 		return false;
4717 	}
4718 }
4719 
4720 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4721 {
4722 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4723 	struct bnxt *bp = eth_dev->data->dev_private;
4724 
4725 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4726 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4727 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4728 	if (!bp->bar0 || !bp->doorbell_base) {
4729 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4730 		return -ENODEV;
4731 	}
4732 
4733 	bp->eth_dev = eth_dev;
4734 	bp->pdev = pci_dev;
4735 
4736 	return 0;
4737 }
4738 
4739 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4740 				  struct bnxt_ctx_pg_info *ctx_pg,
4741 				  uint32_t mem_size,
4742 				  const char *suffix,
4743 				  uint16_t idx)
4744 {
4745 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4746 	const struct rte_memzone *mz = NULL;
4747 	char mz_name[RTE_MEMZONE_NAMESIZE];
4748 	rte_iova_t mz_phys_addr;
4749 	uint64_t valid_bits = 0;
4750 	uint32_t sz;
4751 	int i;
4752 
4753 	if (!mem_size)
4754 		return 0;
4755 
4756 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4757 			 BNXT_PAGE_SIZE;
4758 	rmem->page_size = BNXT_PAGE_SIZE;
4759 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4760 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4761 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4762 
4763 	valid_bits = PTU_PTE_VALID;
4764 
4765 	if (rmem->nr_pages > 1) {
4766 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4767 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4768 			 suffix, idx, bp->eth_dev->data->port_id);
4769 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4770 		mz = rte_memzone_lookup(mz_name);
4771 		if (!mz) {
4772 			mz = rte_memzone_reserve_aligned(mz_name,
4773 						rmem->nr_pages * 8,
4774 						bp->eth_dev->device->numa_node,
4775 						RTE_MEMZONE_2MB |
4776 						RTE_MEMZONE_SIZE_HINT_ONLY |
4777 						RTE_MEMZONE_IOVA_CONTIG,
4778 						BNXT_PAGE_SIZE);
4779 			if (mz == NULL)
4780 				return -ENOMEM;
4781 		}
4782 
4783 		memset(mz->addr, 0, mz->len);
4784 		mz_phys_addr = mz->iova;
4785 
4786 		rmem->pg_tbl = mz->addr;
4787 		rmem->pg_tbl_map = mz_phys_addr;
4788 		rmem->pg_tbl_mz = mz;
4789 	}
4790 
4791 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4792 		 suffix, idx, bp->eth_dev->data->port_id);
4793 	mz = rte_memzone_lookup(mz_name);
4794 	if (!mz) {
4795 		mz = rte_memzone_reserve_aligned(mz_name,
4796 						 mem_size,
4797 						 bp->eth_dev->device->numa_node,
4798 						 RTE_MEMZONE_1GB |
4799 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4800 						 RTE_MEMZONE_IOVA_CONTIG,
4801 						 BNXT_PAGE_SIZE);
4802 		if (mz == NULL)
4803 			return -ENOMEM;
4804 	}
4805 
4806 	memset(mz->addr, 0, mz->len);
4807 	mz_phys_addr = mz->iova;
4808 
4809 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4810 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4811 		rmem->dma_arr[i] = mz_phys_addr + sz;
4812 
4813 		if (rmem->nr_pages > 1) {
4814 			if (i == rmem->nr_pages - 2 &&
4815 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4816 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4817 			else if (i == rmem->nr_pages - 1 &&
4818 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4819 				valid_bits |= PTU_PTE_LAST;
4820 
4821 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4822 							   valid_bits);
4823 		}
4824 	}
4825 
4826 	rmem->mz = mz;
4827 	if (rmem->vmem_size)
4828 		rmem->vmem = (void **)mz->addr;
4829 	rmem->dma_arr[0] = mz_phys_addr;
4830 	return 0;
4831 }
4832 
4833 static void bnxt_free_ctx_mem(struct bnxt *bp)
4834 {
4835 	int i;
4836 
4837 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4838 		return;
4839 
4840 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4841 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4842 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4843 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4844 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4845 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4846 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4847 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4848 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4849 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4850 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4851 
4852 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4853 		if (bp->ctx->tqm_mem[i])
4854 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4855 	}
4856 
4857 	rte_free(bp->ctx);
4858 	bp->ctx = NULL;
4859 }
4860 
4861 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4862 
4863 #define min_t(type, x, y) ({                    \
4864 	type __min1 = (x);                      \
4865 	type __min2 = (y);                      \
4866 	__min1 < __min2 ? __min1 : __min2; })
4867 
4868 #define max_t(type, x, y) ({                    \
4869 	type __max1 = (x);                      \
4870 	type __max2 = (y);                      \
4871 	__max1 > __max2 ? __max1 : __max2; })
4872 
4873 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4874 
4875 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4876 {
4877 	struct bnxt_ctx_pg_info *ctx_pg;
4878 	struct bnxt_ctx_mem_info *ctx;
4879 	uint32_t mem_size, ena, entries;
4880 	uint32_t entries_sp, min;
4881 	int i, rc;
4882 
4883 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4884 	if (rc) {
4885 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4886 		return rc;
4887 	}
4888 	ctx = bp->ctx;
4889 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4890 		return 0;
4891 
4892 	ctx_pg = &ctx->qp_mem;
4893 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4894 	if (ctx->qp_entry_size) {
4895 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4896 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4897 		if (rc)
4898 			return rc;
4899 	}
4900 
4901 	ctx_pg = &ctx->srq_mem;
4902 	ctx_pg->entries = ctx->srq_max_l2_entries;
4903 	if (ctx->srq_entry_size) {
4904 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4905 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4906 		if (rc)
4907 			return rc;
4908 	}
4909 
4910 	ctx_pg = &ctx->cq_mem;
4911 	ctx_pg->entries = ctx->cq_max_l2_entries;
4912 	if (ctx->cq_entry_size) {
4913 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4914 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4915 		if (rc)
4916 			return rc;
4917 	}
4918 
4919 	ctx_pg = &ctx->vnic_mem;
4920 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4921 		ctx->vnic_max_ring_table_entries;
4922 	if (ctx->vnic_entry_size) {
4923 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4924 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4925 		if (rc)
4926 			return rc;
4927 	}
4928 
4929 	ctx_pg = &ctx->stat_mem;
4930 	ctx_pg->entries = ctx->stat_max_entries;
4931 	if (ctx->stat_entry_size) {
4932 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4933 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4934 		if (rc)
4935 			return rc;
4936 	}
4937 
4938 	min = ctx->tqm_min_entries_per_ring;
4939 
4940 	entries_sp = ctx->qp_max_l2_entries +
4941 		     ctx->vnic_max_vnic_entries +
4942 		     2 * ctx->qp_min_qp1_entries + min;
4943 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4944 
4945 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4946 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4947 	entries = clamp_t(uint32_t, entries, min,
4948 			  ctx->tqm_max_entries_per_ring);
4949 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4950 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4951 		 * i > 8 is other ext rings.
4952 		 */
4953 		ctx_pg = ctx->tqm_mem[i];
4954 		ctx_pg->entries = i ? entries : entries_sp;
4955 		if (ctx->tqm_entry_size) {
4956 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4957 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4958 						    "tqm_mem", i);
4959 			if (rc)
4960 				return rc;
4961 		}
4962 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4963 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4964 		else
4965 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4966 	}
4967 
4968 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4969 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4970 	if (rc)
4971 		PMD_DRV_LOG(ERR,
4972 			    "Failed to configure context mem: rc = %d\n", rc);
4973 	else
4974 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4975 
4976 	return rc;
4977 }
4978 
4979 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4980 {
4981 	struct rte_pci_device *pci_dev = bp->pdev;
4982 	char mz_name[RTE_MEMZONE_NAMESIZE];
4983 	const struct rte_memzone *mz = NULL;
4984 	uint32_t total_alloc_len;
4985 	rte_iova_t mz_phys_addr;
4986 
4987 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4988 		return 0;
4989 
4990 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4991 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4992 		 pci_dev->addr.bus, pci_dev->addr.devid,
4993 		 pci_dev->addr.function, "rx_port_stats");
4994 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4995 	mz = rte_memzone_lookup(mz_name);
4996 	total_alloc_len =
4997 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4998 				       sizeof(struct rx_port_stats_ext) + 512);
4999 	if (!mz) {
5000 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
5001 					 SOCKET_ID_ANY,
5002 					 RTE_MEMZONE_2MB |
5003 					 RTE_MEMZONE_SIZE_HINT_ONLY |
5004 					 RTE_MEMZONE_IOVA_CONTIG);
5005 		if (mz == NULL)
5006 			return -ENOMEM;
5007 	}
5008 	memset(mz->addr, 0, mz->len);
5009 	mz_phys_addr = mz->iova;
5010 
5011 	bp->rx_mem_zone = (const void *)mz;
5012 	bp->hw_rx_port_stats = mz->addr;
5013 	bp->hw_rx_port_stats_map = mz_phys_addr;
5014 
5015 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
5016 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
5017 		 pci_dev->addr.bus, pci_dev->addr.devid,
5018 		 pci_dev->addr.function, "tx_port_stats");
5019 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
5020 	mz = rte_memzone_lookup(mz_name);
5021 	total_alloc_len =
5022 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
5023 				       sizeof(struct tx_port_stats_ext) + 512);
5024 	if (!mz) {
5025 		mz = rte_memzone_reserve(mz_name,
5026 					 total_alloc_len,
5027 					 SOCKET_ID_ANY,
5028 					 RTE_MEMZONE_2MB |
5029 					 RTE_MEMZONE_SIZE_HINT_ONLY |
5030 					 RTE_MEMZONE_IOVA_CONTIG);
5031 		if (mz == NULL)
5032 			return -ENOMEM;
5033 	}
5034 	memset(mz->addr, 0, mz->len);
5035 	mz_phys_addr = mz->iova;
5036 
5037 	bp->tx_mem_zone = (const void *)mz;
5038 	bp->hw_tx_port_stats = mz->addr;
5039 	bp->hw_tx_port_stats_map = mz_phys_addr;
5040 	bp->flags |= BNXT_FLAG_PORT_STATS;
5041 
5042 	/* Display extended statistics if FW supports it */
5043 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
5044 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
5045 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
5046 		return 0;
5047 
5048 	bp->hw_rx_port_stats_ext = (void *)
5049 		((uint8_t *)bp->hw_rx_port_stats +
5050 		 sizeof(struct rx_port_stats));
5051 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
5052 		sizeof(struct rx_port_stats);
5053 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
5054 
5055 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
5056 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
5057 		bp->hw_tx_port_stats_ext = (void *)
5058 			((uint8_t *)bp->hw_tx_port_stats +
5059 			 sizeof(struct tx_port_stats));
5060 		bp->hw_tx_port_stats_ext_map =
5061 			bp->hw_tx_port_stats_map +
5062 			sizeof(struct tx_port_stats);
5063 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
5064 	}
5065 
5066 	return 0;
5067 }
5068 
5069 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
5070 {
5071 	struct bnxt *bp = eth_dev->data->dev_private;
5072 	size_t max_mac_addr = RTE_MIN(bp->max_l2_ctx, RTE_ETH_NUM_RECEIVE_MAC_ADDR);
5073 	int rc = 0;
5074 
5075 	if (bp->max_l2_ctx > RTE_ETH_NUM_RECEIVE_MAC_ADDR)
5076 		PMD_DRV_LOG(INFO, "Max number of MAC addrs supported is %d, but will be limited to %d\n",
5077 			    bp->max_l2_ctx, RTE_ETH_NUM_RECEIVE_MAC_ADDR);
5078 
5079 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
5080 					       RTE_ETHER_ADDR_LEN * max_mac_addr,
5081 					       0);
5082 	if (eth_dev->data->mac_addrs == NULL) {
5083 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
5084 		return -ENOMEM;
5085 	}
5086 
5087 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
5088 		if (BNXT_PF(bp))
5089 			return -EINVAL;
5090 
5091 		/* Generate a random MAC address, if none was assigned by PF */
5092 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
5093 		bnxt_eth_hw_addr_random(bp->mac_addr);
5094 		PMD_DRV_LOG(INFO,
5095 			    "Assign random MAC:" RTE_ETHER_ADDR_PRT_FMT "\n",
5096 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
5097 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
5098 
5099 		rc = bnxt_hwrm_set_mac(bp);
5100 		if (rc)
5101 			return rc;
5102 	}
5103 
5104 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5105 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5106 
5107 	/*
5108 	 *  Allocate memory to hold multicast mac addresses added.
5109 	 *  Used to restore them during reset recovery
5110 	 */
5111 	bp->mcast_addr_list = rte_zmalloc("bnxt_mcast_addr_tbl",
5112 					  sizeof(struct rte_ether_addr) *
5113 					  BNXT_MAX_MC_ADDRS, 0);
5114 	if (bp->mcast_addr_list == NULL) {
5115 		PMD_DRV_LOG(ERR, "Failed to allocate multicast addr table\n");
5116 		return -ENOMEM;
5117 	}
5118 	bp->mc_list_dma_addr = rte_malloc_virt2iova(bp->mcast_addr_list);
5119 	if (bp->mc_list_dma_addr == RTE_BAD_IOVA) {
5120 		PMD_DRV_LOG(ERR, "Fail to map mcast_addr_list to physical memory\n");
5121 		return -ENOMEM;
5122 	}
5123 
5124 	return rc;
5125 }
5126 
5127 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5128 {
5129 	int rc = 0;
5130 
5131 	/* MAC is already configured in FW */
5132 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5133 		return 0;
5134 
5135 	/* Restore the old MAC configured */
5136 	rc = bnxt_hwrm_set_mac(bp);
5137 	if (rc)
5138 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5139 
5140 	return rc;
5141 }
5142 
5143 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5144 {
5145 	if (!BNXT_PF(bp))
5146 		return;
5147 
5148 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5149 
5150 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5151 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5152 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5153 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5154 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5155 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5156 }
5157 
5158 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5159 {
5160 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5161 
5162 	if (info) {
5163 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5164 			memset(info, 0, sizeof(*info));
5165 		return;
5166 	}
5167 
5168 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5169 		return;
5170 
5171 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5172 			   sizeof(*info), 0);
5173 	if (!info)
5174 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5175 
5176 	bp->recovery_info = info;
5177 }
5178 
5179 static void bnxt_check_fw_status(struct bnxt *bp)
5180 {
5181 	uint32_t fw_status;
5182 
5183 	if (!(bp->recovery_info &&
5184 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5185 		return;
5186 
5187 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5188 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5189 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5190 			    fw_status);
5191 }
5192 
5193 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5194 {
5195 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5196 	uint32_t status_loc;
5197 	uint32_t sig_ver;
5198 
5199 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5200 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5201 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5202 				   BNXT_GRCP_WINDOW_2_BASE +
5203 				   offsetof(struct hcomm_status,
5204 					    sig_ver)));
5205 	/* If the signature is absent, then FW does not support this feature */
5206 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5207 	    HCOMM_STATUS_SIGNATURE_VAL)
5208 		return 0;
5209 
5210 	if (!info) {
5211 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5212 				   sizeof(*info), 0);
5213 		if (!info)
5214 			return -ENOMEM;
5215 		bp->recovery_info = info;
5216 	} else {
5217 		memset(info, 0, sizeof(*info));
5218 	}
5219 
5220 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5221 				      BNXT_GRCP_WINDOW_2_BASE +
5222 				      offsetof(struct hcomm_status,
5223 					       fw_status_loc)));
5224 
5225 	/* Only pre-map the FW health status GRC register */
5226 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5227 		return 0;
5228 
5229 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5230 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5231 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5232 
5233 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5234 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5235 
5236 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5237 
5238 	return 0;
5239 }
5240 
5241 /* This function gets the FW version along with the
5242  * capabilities(MAX and current) of the function, vnic,
5243  * error recovery, phy and other chip related info
5244  */
5245 static int bnxt_get_config(struct bnxt *bp)
5246 {
5247 	uint16_t mtu;
5248 	int rc = 0;
5249 
5250 	bp->fw_cap = 0;
5251 
5252 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5253 	if (rc)
5254 		return rc;
5255 
5256 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5257 	if (rc) {
5258 		bnxt_check_fw_status(bp);
5259 		return rc;
5260 	}
5261 
5262 	rc = bnxt_hwrm_func_reset(bp);
5263 	if (rc)
5264 		return -EIO;
5265 
5266 	rc = bnxt_hwrm_vnic_qcaps(bp);
5267 	if (rc)
5268 		return rc;
5269 
5270 	rc = bnxt_hwrm_queue_qportcfg(bp);
5271 	if (rc)
5272 		return rc;
5273 
5274 	/* Get the MAX capabilities for this function.
5275 	 * This function also allocates context memory for TQM rings and
5276 	 * informs the firmware about this allocated backing store memory.
5277 	 */
5278 	rc = bnxt_hwrm_func_qcaps(bp);
5279 	if (rc)
5280 		return rc;
5281 
5282 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5283 	if (rc)
5284 		return rc;
5285 
5286 	bnxt_hwrm_port_mac_qcfg(bp);
5287 
5288 	bnxt_hwrm_parent_pf_qcfg(bp);
5289 
5290 	bnxt_hwrm_port_phy_qcaps(bp);
5291 
5292 	bnxt_alloc_error_recovery_info(bp);
5293 	/* Get the adapter error recovery support info */
5294 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5295 	if (rc)
5296 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5297 
5298 	bnxt_hwrm_port_led_qcaps(bp);
5299 
5300 	return 0;
5301 }
5302 
5303 static int
5304 bnxt_init_locks(struct bnxt *bp)
5305 {
5306 	int err;
5307 
5308 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5309 	if (err) {
5310 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5311 		return err;
5312 	}
5313 
5314 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5315 	if (err) {
5316 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5317 		return err;
5318 	}
5319 
5320 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5321 	if (err) {
5322 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5323 		return err;
5324 	}
5325 
5326 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5327 	if (err)
5328 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5329 
5330 	return err;
5331 }
5332 
5333 /* This should be called after we have queried trusted VF cap */
5334 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5335 {
5336 	int rc = 0;
5337 
5338 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5339 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5340 		if (rc)
5341 			PMD_DRV_LOG(ERR,
5342 				    "Failed to alloc switch domain: %d\n", rc);
5343 		else
5344 			PMD_DRV_LOG(INFO,
5345 				    "Switch domain allocated %d\n",
5346 				    bp->switch_domain_id);
5347 	}
5348 
5349 	return rc;
5350 }
5351 
5352 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5353 {
5354 	int rc = 0;
5355 
5356 	if (reconfig_dev) {
5357 		rc = bnxt_get_config(bp);
5358 		if (rc)
5359 			return rc;
5360 	}
5361 
5362 	rc = bnxt_alloc_switch_domain(bp);
5363 	if (rc)
5364 		return rc;
5365 
5366 	if (!reconfig_dev) {
5367 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5368 		if (rc)
5369 			return rc;
5370 	} else {
5371 		rc = bnxt_restore_dflt_mac(bp);
5372 		if (rc)
5373 			return rc;
5374 	}
5375 
5376 	bnxt_config_vf_req_fwd(bp);
5377 
5378 	rc = bnxt_hwrm_func_driver_register(bp);
5379 	if (rc) {
5380 		PMD_DRV_LOG(ERR, "Failed to register driver");
5381 		return -EBUSY;
5382 	}
5383 
5384 	if (BNXT_PF(bp)) {
5385 		if (bp->pdev->max_vfs) {
5386 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5387 			if (rc) {
5388 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5389 				return rc;
5390 			}
5391 		} else {
5392 			rc = bnxt_hwrm_allocate_pf_only(bp);
5393 			if (rc) {
5394 				PMD_DRV_LOG(ERR,
5395 					    "Failed to allocate PF resources");
5396 				return rc;
5397 			}
5398 		}
5399 	}
5400 
5401 	if (!reconfig_dev) {
5402 		bp->rss_conf.rss_key = rte_zmalloc("bnxt_rss_key",
5403 						   HW_HASH_KEY_SIZE, 0);
5404 		if (bp->rss_conf.rss_key == NULL) {
5405 			PMD_DRV_LOG(ERR, "port %u cannot allocate RSS hash key memory",
5406 				    bp->eth_dev->data->port_id);
5407 			return -ENOMEM;
5408 		}
5409 	}
5410 
5411 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5412 	if (rc)
5413 		return rc;
5414 
5415 	rc = bnxt_setup_int(bp);
5416 	if (rc)
5417 		return rc;
5418 
5419 	rc = bnxt_request_int(bp);
5420 	if (rc)
5421 		return rc;
5422 
5423 	rc = bnxt_init_ctx_mem(bp);
5424 	if (rc) {
5425 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5426 		return rc;
5427 	}
5428 
5429 	return 0;
5430 }
5431 
5432 static int
5433 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5434 			     const char *value, void *opaque_arg)
5435 {
5436 	struct bnxt *bp = opaque_arg;
5437 	unsigned long flow_xstat;
5438 	char *end = NULL;
5439 
5440 	if (!value || !opaque_arg) {
5441 		PMD_DRV_LOG(ERR,
5442 			    "Invalid parameter passed to flow_xstat devarg.\n");
5443 		return -EINVAL;
5444 	}
5445 
5446 	flow_xstat = strtoul(value, &end, 10);
5447 	if (end == NULL || *end != '\0' ||
5448 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5449 		PMD_DRV_LOG(ERR,
5450 			    "Invalid parameter passed to flow_xstat devarg.\n");
5451 		return -EINVAL;
5452 	}
5453 
5454 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5455 		PMD_DRV_LOG(ERR,
5456 			    "Invalid value passed to flow_xstat devarg.\n");
5457 		return -EINVAL;
5458 	}
5459 
5460 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5461 	if (BNXT_FLOW_XSTATS_EN(bp))
5462 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5463 
5464 	return 0;
5465 }
5466 
5467 static int
5468 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5469 					const char *value, void *opaque_arg)
5470 {
5471 	struct bnxt *bp = opaque_arg;
5472 	unsigned long max_num_kflows;
5473 	char *end = NULL;
5474 
5475 	if (!value || !opaque_arg) {
5476 		PMD_DRV_LOG(ERR,
5477 			"Invalid parameter passed to max_num_kflows devarg.\n");
5478 		return -EINVAL;
5479 	}
5480 
5481 	max_num_kflows = strtoul(value, &end, 10);
5482 	if (end == NULL || *end != '\0' ||
5483 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5484 		PMD_DRV_LOG(ERR,
5485 			"Invalid parameter passed to max_num_kflows devarg.\n");
5486 		return -EINVAL;
5487 	}
5488 
5489 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5490 		PMD_DRV_LOG(ERR,
5491 			"Invalid value passed to max_num_kflows devarg.\n");
5492 		return -EINVAL;
5493 	}
5494 
5495 	bp->max_num_kflows = max_num_kflows;
5496 	if (bp->max_num_kflows)
5497 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5498 				max_num_kflows);
5499 
5500 	return 0;
5501 }
5502 
5503 static int
5504 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5505 				 const char *value, void *opaque_arg)
5506 {
5507 	struct bnxt *bp = opaque_arg;
5508 	unsigned long app_id;
5509 	char *end = NULL;
5510 
5511 	if (!value || !opaque_arg) {
5512 		PMD_DRV_LOG(ERR,
5513 			    "Invalid parameter passed to app-id "
5514 			    "devargs.\n");
5515 		return -EINVAL;
5516 	}
5517 
5518 	app_id = strtoul(value, &end, 10);
5519 	if (end == NULL || *end != '\0' ||
5520 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5521 		PMD_DRV_LOG(ERR,
5522 			    "Invalid parameter passed to app_id "
5523 			    "devargs.\n");
5524 		return -EINVAL;
5525 	}
5526 
5527 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5528 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5529 			    (uint16_t)app_id);
5530 		return -EINVAL;
5531 	}
5532 
5533 	bp->app_id = app_id;
5534 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5535 
5536 	return 0;
5537 }
5538 
5539 static int
5540 bnxt_parse_devarg_ieee_1588(__rte_unused const char *key,
5541 			    const char *value, void *opaque_arg)
5542 {
5543 	struct bnxt *bp = opaque_arg;
5544 	unsigned long ieee_1588;
5545 	char *end = NULL;
5546 
5547 	if (!value || !opaque_arg) {
5548 		PMD_DRV_LOG(ERR,
5549 			    "Invalid parameter passed to ieee-1588 "
5550 			    "devargs.\n");
5551 		return -EINVAL;
5552 	}
5553 
5554 	ieee_1588 = strtoul(value, &end, 10);
5555 	if (end == NULL || *end != '\0' ||
5556 	    (ieee_1588 == ULONG_MAX && errno == ERANGE)) {
5557 		PMD_DRV_LOG(ERR,
5558 			    "Invalid parameter passed to ieee_1588 "
5559 			    "devargs.\n");
5560 		return -EINVAL;
5561 	}
5562 
5563 	if (BNXT_DEVARG_IEEE_1588_INVALID(ieee_1588)) {
5564 		PMD_DRV_LOG(ERR, "Invalid ieee-1588(%d) devargs.\n",
5565 			    (uint16_t)ieee_1588);
5566 		return -EINVAL;
5567 	}
5568 
5569 	bp->ieee_1588 = ieee_1588;
5570 	PMD_DRV_LOG(INFO, "ieee-1588=%d feature enabled.\n", (uint16_t)ieee_1588);
5571 
5572 	return 0;
5573 }
5574 
5575 static int
5576 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5577 			    const char *value, void *opaque_arg)
5578 {
5579 	struct bnxt_representor *vfr_bp = opaque_arg;
5580 	unsigned long rep_is_pf;
5581 	char *end = NULL;
5582 
5583 	if (!value || !opaque_arg) {
5584 		PMD_DRV_LOG(ERR,
5585 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5586 		return -EINVAL;
5587 	}
5588 
5589 	rep_is_pf = strtoul(value, &end, 10);
5590 	if (end == NULL || *end != '\0' ||
5591 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5592 		PMD_DRV_LOG(ERR,
5593 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5594 		return -EINVAL;
5595 	}
5596 
5597 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5598 		PMD_DRV_LOG(ERR,
5599 			    "Invalid value passed to rep_is_pf devargs.\n");
5600 		return -EINVAL;
5601 	}
5602 
5603 	vfr_bp->flags |= rep_is_pf;
5604 	if (BNXT_REP_PF(vfr_bp))
5605 		PMD_DRV_LOG(INFO, "PF representor\n");
5606 	else
5607 		PMD_DRV_LOG(INFO, "VF representor\n");
5608 
5609 	return 0;
5610 }
5611 
5612 static int
5613 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5614 			       const char *value, void *opaque_arg)
5615 {
5616 	struct bnxt_representor *vfr_bp = opaque_arg;
5617 	unsigned long rep_based_pf;
5618 	char *end = NULL;
5619 
5620 	if (!value || !opaque_arg) {
5621 		PMD_DRV_LOG(ERR,
5622 			    "Invalid parameter passed to rep_based_pf "
5623 			    "devargs.\n");
5624 		return -EINVAL;
5625 	}
5626 
5627 	rep_based_pf = strtoul(value, &end, 10);
5628 	if (end == NULL || *end != '\0' ||
5629 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5630 		PMD_DRV_LOG(ERR,
5631 			    "Invalid parameter passed to rep_based_pf "
5632 			    "devargs.\n");
5633 		return -EINVAL;
5634 	}
5635 
5636 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5637 		PMD_DRV_LOG(ERR,
5638 			    "Invalid value passed to rep_based_pf devargs.\n");
5639 		return -EINVAL;
5640 	}
5641 
5642 	vfr_bp->rep_based_pf = rep_based_pf;
5643 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5644 
5645 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5646 
5647 	return 0;
5648 }
5649 
5650 static int
5651 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5652 			    const char *value, void *opaque_arg)
5653 {
5654 	struct bnxt_representor *vfr_bp = opaque_arg;
5655 	unsigned long rep_q_r2f;
5656 	char *end = NULL;
5657 
5658 	if (!value || !opaque_arg) {
5659 		PMD_DRV_LOG(ERR,
5660 			    "Invalid parameter passed to rep_q_r2f "
5661 			    "devargs.\n");
5662 		return -EINVAL;
5663 	}
5664 
5665 	rep_q_r2f = strtoul(value, &end, 10);
5666 	if (end == NULL || *end != '\0' ||
5667 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5668 		PMD_DRV_LOG(ERR,
5669 			    "Invalid parameter passed to rep_q_r2f "
5670 			    "devargs.\n");
5671 		return -EINVAL;
5672 	}
5673 
5674 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5675 		PMD_DRV_LOG(ERR,
5676 			    "Invalid value passed to rep_q_r2f devargs.\n");
5677 		return -EINVAL;
5678 	}
5679 
5680 	vfr_bp->rep_q_r2f = rep_q_r2f;
5681 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5682 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5683 
5684 	return 0;
5685 }
5686 
5687 static int
5688 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5689 			    const char *value, void *opaque_arg)
5690 {
5691 	struct bnxt_representor *vfr_bp = opaque_arg;
5692 	unsigned long rep_q_f2r;
5693 	char *end = NULL;
5694 
5695 	if (!value || !opaque_arg) {
5696 		PMD_DRV_LOG(ERR,
5697 			    "Invalid parameter passed to rep_q_f2r "
5698 			    "devargs.\n");
5699 		return -EINVAL;
5700 	}
5701 
5702 	rep_q_f2r = strtoul(value, &end, 10);
5703 	if (end == NULL || *end != '\0' ||
5704 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5705 		PMD_DRV_LOG(ERR,
5706 			    "Invalid parameter passed to rep_q_f2r "
5707 			    "devargs.\n");
5708 		return -EINVAL;
5709 	}
5710 
5711 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5712 		PMD_DRV_LOG(ERR,
5713 			    "Invalid value passed to rep_q_f2r devargs.\n");
5714 		return -EINVAL;
5715 	}
5716 
5717 	vfr_bp->rep_q_f2r = rep_q_f2r;
5718 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5719 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5720 
5721 	return 0;
5722 }
5723 
5724 static int
5725 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5726 			     const char *value, void *opaque_arg)
5727 {
5728 	struct bnxt_representor *vfr_bp = opaque_arg;
5729 	unsigned long rep_fc_r2f;
5730 	char *end = NULL;
5731 
5732 	if (!value || !opaque_arg) {
5733 		PMD_DRV_LOG(ERR,
5734 			    "Invalid parameter passed to rep_fc_r2f "
5735 			    "devargs.\n");
5736 		return -EINVAL;
5737 	}
5738 
5739 	rep_fc_r2f = strtoul(value, &end, 10);
5740 	if (end == NULL || *end != '\0' ||
5741 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5742 		PMD_DRV_LOG(ERR,
5743 			    "Invalid parameter passed to rep_fc_r2f "
5744 			    "devargs.\n");
5745 		return -EINVAL;
5746 	}
5747 
5748 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5749 		PMD_DRV_LOG(ERR,
5750 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5751 		return -EINVAL;
5752 	}
5753 
5754 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5755 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5756 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5757 
5758 	return 0;
5759 }
5760 
5761 static int
5762 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5763 			     const char *value, void *opaque_arg)
5764 {
5765 	struct bnxt_representor *vfr_bp = opaque_arg;
5766 	unsigned long rep_fc_f2r;
5767 	char *end = NULL;
5768 
5769 	if (!value || !opaque_arg) {
5770 		PMD_DRV_LOG(ERR,
5771 			    "Invalid parameter passed to rep_fc_f2r "
5772 			    "devargs.\n");
5773 		return -EINVAL;
5774 	}
5775 
5776 	rep_fc_f2r = strtoul(value, &end, 10);
5777 	if (end == NULL || *end != '\0' ||
5778 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5779 		PMD_DRV_LOG(ERR,
5780 			    "Invalid parameter passed to rep_fc_f2r "
5781 			    "devargs.\n");
5782 		return -EINVAL;
5783 	}
5784 
5785 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5786 		PMD_DRV_LOG(ERR,
5787 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5788 		return -EINVAL;
5789 	}
5790 
5791 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5792 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5793 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5794 
5795 	return 0;
5796 }
5797 
5798 static int
5799 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5800 {
5801 	struct rte_kvargs *kvlist;
5802 	int ret = 0;
5803 
5804 	if (devargs == NULL)
5805 		return 0;
5806 
5807 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5808 	if (kvlist == NULL)
5809 		return -EINVAL;
5810 
5811 	/*
5812 	 * Handler for "flow_xstat" devarg.
5813 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5814 	 */
5815 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5816 				 bnxt_parse_devarg_flow_xstat, bp);
5817 	if (ret)
5818 		goto err;
5819 
5820 	/*
5821 	 * Handler for "max_num_kflows" devarg.
5822 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5823 	 */
5824 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5825 				 bnxt_parse_devarg_max_num_kflows, bp);
5826 	if (ret)
5827 		goto err;
5828 
5829 err:
5830 	/*
5831 	 * Handler for "app-id" devarg.
5832 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5833 	 */
5834 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5835 			   bnxt_parse_devarg_app_id, bp);
5836 
5837 	/*
5838 	 * Handler for "ieee-1588" devarg.
5839 	 * Invoked as for ex: "-a 000:00:0d.0,ieee-1588=1"
5840 	 */
5841 	rte_kvargs_process(kvlist, BNXT_DEVARG_IEEE_1588,
5842 			   bnxt_parse_devarg_ieee_1588, bp);
5843 
5844 	rte_kvargs_free(kvlist);
5845 	return ret;
5846 }
5847 
5848 /* Allocate and initialize various fields in bnxt struct that
5849  * need to be allocated/destroyed only once in the lifetime of the driver
5850  */
5851 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5852 {
5853 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5854 	struct bnxt *bp = eth_dev->data->dev_private;
5855 	int rc = 0;
5856 
5857 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5858 
5859 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5860 		bp->flags |= BNXT_FLAG_VF;
5861 
5862 	if (bnxt_p5_device(pci_dev->id.device_id))
5863 		bp->flags |= BNXT_FLAG_CHIP_P5;
5864 
5865 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5866 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5867 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5868 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5869 		bp->flags |= BNXT_FLAG_STINGRAY;
5870 
5871 	rc = bnxt_map_pci_bars(eth_dev);
5872 	if (rc) {
5873 		PMD_DRV_LOG(ERR,
5874 			    "Failed to initialize board rc: %x\n", rc);
5875 		return rc;
5876 	}
5877 
5878 	rc = bnxt_alloc_pf_info(bp);
5879 	if (rc)
5880 		return rc;
5881 
5882 	rc = bnxt_alloc_link_info(bp);
5883 	if (rc)
5884 		return rc;
5885 
5886 	rc = bnxt_alloc_parent_info(bp);
5887 	if (rc)
5888 		return rc;
5889 
5890 	rc = bnxt_alloc_hwrm_resources(bp);
5891 	if (rc) {
5892 		PMD_DRV_LOG(ERR,
5893 			    "Failed to allocate response buffer rc: %x\n", rc);
5894 		return rc;
5895 	}
5896 	rc = bnxt_alloc_leds_info(bp);
5897 	if (rc)
5898 		return rc;
5899 
5900 	rc = bnxt_alloc_cos_queues(bp);
5901 	if (rc)
5902 		return rc;
5903 
5904 	rc = bnxt_init_locks(bp);
5905 	if (rc)
5906 		return rc;
5907 
5908 	rc = bnxt_get_config(bp);
5909 	if (rc)
5910 		return rc;
5911 
5912 	if (BNXT_TRUFLOW_EN(bp)) {
5913 		/* extra mbuf field is required to store CFA code from mark */
5914 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5915 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5916 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5917 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5918 		};
5919 		bnxt_cfa_code_dynfield_offset =
5920 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5921 		if (bnxt_cfa_code_dynfield_offset < 0) {
5922 			PMD_DRV_LOG(ERR,
5923 			    "Failed to register mbuf field for TruFlow mark\n");
5924 			return -rte_errno;
5925 		}
5926 	}
5927 
5928 	return rc;
5929 }
5930 
5931 static int
5932 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5933 {
5934 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5935 	static int version_printed;
5936 	struct bnxt *bp;
5937 	int rc;
5938 
5939 	if (version_printed++ == 0)
5940 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5941 
5942 	eth_dev->dev_ops = &bnxt_dev_ops;
5943 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5944 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5945 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5946 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5947 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5948 
5949 	/*
5950 	 * For secondary processes, we don't initialise any further
5951 	 * as primary has already done this work.
5952 	 */
5953 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5954 		return 0;
5955 
5956 	rte_eth_copy_pci_info(eth_dev, pci_dev);
5957 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5958 	eth_dev->data->dev_flags |= RTE_ETH_DEV_INTR_LSC;
5959 
5960 	bp = eth_dev->data->dev_private;
5961 
5962 	/* set the default app id */
5963 	bp->app_id = bnxt_ulp_default_app_id_get();
5964 
5965 	/* Parse dev arguments passed on when starting the DPDK application. */
5966 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5967 	if (rc)
5968 		goto error_free;
5969 
5970 	rc = bnxt_drv_init(eth_dev);
5971 	if (rc)
5972 		goto error_free;
5973 
5974 	rc = bnxt_init_resources(bp, false);
5975 	if (rc)
5976 		goto error_free;
5977 
5978 	rc = bnxt_alloc_stats_mem(bp);
5979 	if (rc)
5980 		goto error_free;
5981 
5982 	PMD_DRV_LOG(INFO,
5983 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
5984 		    DRV_MODULE_NAME,
5985 		    pci_dev->mem_resource[0].phys_addr,
5986 		    pci_dev->mem_resource[0].addr);
5987 
5988 	return 0;
5989 
5990 error_free:
5991 	bnxt_dev_uninit(eth_dev);
5992 	return rc;
5993 }
5994 
5995 
5996 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5997 {
5998 	if (!ctx)
5999 		return;
6000 
6001 	if (ctx->va)
6002 		rte_free(ctx->va);
6003 
6004 	ctx->va = NULL;
6005 	ctx->dma = RTE_BAD_IOVA;
6006 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
6007 }
6008 
6009 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
6010 {
6011 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
6012 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6013 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
6014 				  bp->flow_stat->max_fc,
6015 				  false);
6016 
6017 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
6018 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6019 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
6020 				  bp->flow_stat->max_fc,
6021 				  false);
6022 
6023 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6024 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
6025 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6026 
6027 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6028 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
6029 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6030 
6031 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6032 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
6033 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6034 
6035 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6036 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
6037 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6038 }
6039 
6040 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
6041 {
6042 	bnxt_unregister_fc_ctx_mem(bp);
6043 
6044 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
6045 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
6046 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
6047 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
6048 }
6049 
6050 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
6051 {
6052 	if (BNXT_FLOW_XSTATS_EN(bp))
6053 		bnxt_uninit_fc_ctx_mem(bp);
6054 }
6055 
6056 static void
6057 bnxt_free_error_recovery_info(struct bnxt *bp)
6058 {
6059 	rte_free(bp->recovery_info);
6060 	bp->recovery_info = NULL;
6061 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
6062 }
6063 
6064 static int
6065 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
6066 {
6067 	int rc;
6068 
6069 	bnxt_free_int(bp);
6070 	bnxt_free_mem(bp, reconfig_dev);
6071 
6072 	bnxt_hwrm_func_buf_unrgtr(bp);
6073 	if (bp->pf != NULL) {
6074 		rte_free(bp->pf->vf_req_buf);
6075 		bp->pf->vf_req_buf = NULL;
6076 	}
6077 
6078 	rc = bnxt_hwrm_func_driver_unregister(bp);
6079 	bp->flags &= ~BNXT_FLAG_REGISTERED;
6080 	bnxt_free_ctx_mem(bp);
6081 	if (!reconfig_dev) {
6082 		bnxt_free_hwrm_resources(bp);
6083 		bnxt_free_error_recovery_info(bp);
6084 		rte_free(bp->mcast_addr_list);
6085 		bp->mcast_addr_list = NULL;
6086 		rte_free(bp->rss_conf.rss_key);
6087 		bp->rss_conf.rss_key = NULL;
6088 	}
6089 
6090 	bnxt_uninit_ctx_mem(bp);
6091 
6092 	bnxt_free_flow_stats_info(bp);
6093 	bnxt_free_switch_domain(bp);
6094 	rte_free(bp->ptp_cfg);
6095 	bp->ptp_cfg = NULL;
6096 	return rc;
6097 }
6098 
6099 static int
6100 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
6101 {
6102 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6103 		return -EPERM;
6104 
6105 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
6106 
6107 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
6108 		bnxt_dev_close_op(eth_dev);
6109 
6110 	return 0;
6111 }
6112 
6113 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6114 {
6115 	struct bnxt *bp = eth_dev->data->dev_private;
6116 	struct rte_eth_dev *vf_rep_eth_dev;
6117 	int ret = 0, i;
6118 
6119 	if (!bp)
6120 		return -EINVAL;
6121 
6122 	for (i = 0; i < bp->num_reps; i++) {
6123 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6124 		if (!vf_rep_eth_dev)
6125 			continue;
6126 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6127 			    vf_rep_eth_dev->data->port_id);
6128 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6129 	}
6130 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6131 		    eth_dev->data->port_id);
6132 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6133 
6134 	return ret;
6135 }
6136 
6137 static void bnxt_free_rep_info(struct bnxt *bp)
6138 {
6139 	rte_free(bp->rep_info);
6140 	bp->rep_info = NULL;
6141 	rte_free(bp->cfa_code_map);
6142 	bp->cfa_code_map = NULL;
6143 }
6144 
6145 static int bnxt_init_rep_info(struct bnxt *bp)
6146 {
6147 	int i = 0, rc;
6148 
6149 	if (bp->rep_info)
6150 		return 0;
6151 
6152 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
6153 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS(bp),
6154 				   0);
6155 	if (!bp->rep_info) {
6156 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6157 		return -ENOMEM;
6158 	}
6159 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6160 				       sizeof(*bp->cfa_code_map) *
6161 				       BNXT_MAX_CFA_CODE, 0);
6162 	if (!bp->cfa_code_map) {
6163 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6164 		bnxt_free_rep_info(bp);
6165 		return -ENOMEM;
6166 	}
6167 
6168 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6169 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6170 
6171 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6172 	if (rc) {
6173 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6174 		bnxt_free_rep_info(bp);
6175 		return rc;
6176 	}
6177 
6178 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6179 	if (rc) {
6180 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6181 		bnxt_free_rep_info(bp);
6182 		return rc;
6183 	}
6184 
6185 	return rc;
6186 }
6187 
6188 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6189 			       struct rte_eth_devargs *eth_da,
6190 			       struct rte_eth_dev *backing_eth_dev,
6191 			       const char *dev_args)
6192 {
6193 	struct rte_eth_dev *vf_rep_eth_dev;
6194 	char name[RTE_ETH_NAME_MAX_LEN];
6195 	struct bnxt *backing_bp = backing_eth_dev->data->dev_private;
6196 	uint16_t max_vf_reps = BNXT_MAX_VF_REPS(backing_bp);
6197 
6198 	uint16_t num_rep;
6199 	int i, ret = 0;
6200 	struct rte_kvargs *kvlist = NULL;
6201 
6202 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6203 		return 0;
6204 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6205 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6206 			    eth_da->type);
6207 		return -ENOTSUP;
6208 	}
6209 	num_rep = eth_da->nb_representor_ports;
6210 	if (num_rep > max_vf_reps) {
6211 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6212 			    num_rep, max_vf_reps);
6213 		return -EINVAL;
6214 	}
6215 
6216 	if (num_rep >= RTE_MAX_ETHPORTS) {
6217 		PMD_DRV_LOG(ERR,
6218 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6219 			    num_rep, RTE_MAX_ETHPORTS);
6220 		return -EINVAL;
6221 	}
6222 
6223 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6224 		PMD_DRV_LOG(ERR,
6225 			    "Not a PF or trusted VF. No Representor support\n");
6226 		/* Returning an error is not an option.
6227 		 * Applications are not handling this correctly
6228 		 */
6229 		return 0;
6230 	}
6231 
6232 	if (bnxt_init_rep_info(backing_bp))
6233 		return 0;
6234 
6235 	for (i = 0; i < num_rep; i++) {
6236 		struct bnxt_representor representor = {
6237 			.vf_id = eth_da->representor_ports[i],
6238 			.switch_domain_id = backing_bp->switch_domain_id,
6239 			.parent_dev = backing_eth_dev
6240 		};
6241 
6242 		if (representor.vf_id >= max_vf_reps) {
6243 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6244 				    representor.vf_id, max_vf_reps);
6245 			continue;
6246 		}
6247 
6248 		/* representor port net_bdf_port */
6249 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6250 			 pci_dev->device.name, eth_da->representor_ports[i]);
6251 
6252 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6253 		if (kvlist) {
6254 			/*
6255 			 * Handler for "rep_is_pf" devarg.
6256 			 * Invoked as for ex: "-a 000:00:0d.0,
6257 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6258 			 */
6259 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6260 						 bnxt_parse_devarg_rep_is_pf,
6261 						 (void *)&representor);
6262 			if (ret) {
6263 				ret = -EINVAL;
6264 				goto err;
6265 			}
6266 			/*
6267 			 * Handler for "rep_based_pf" devarg.
6268 			 * Invoked as for ex: "-a 000:00:0d.0,
6269 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6270 			 */
6271 			ret = rte_kvargs_process(kvlist,
6272 						 BNXT_DEVARG_REP_BASED_PF,
6273 						 bnxt_parse_devarg_rep_based_pf,
6274 						 (void *)&representor);
6275 			if (ret) {
6276 				ret = -EINVAL;
6277 				goto err;
6278 			}
6279 			/*
6280 			 * Handler for "rep_based_pf" devarg.
6281 			 * Invoked as for ex: "-a 000:00:0d.0,
6282 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6283 			 */
6284 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6285 						 bnxt_parse_devarg_rep_q_r2f,
6286 						 (void *)&representor);
6287 			if (ret) {
6288 				ret = -EINVAL;
6289 				goto err;
6290 			}
6291 			/*
6292 			 * Handler for "rep_based_pf" devarg.
6293 			 * Invoked as for ex: "-a 000:00:0d.0,
6294 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6295 			 */
6296 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6297 						 bnxt_parse_devarg_rep_q_f2r,
6298 						 (void *)&representor);
6299 			if (ret) {
6300 				ret = -EINVAL;
6301 				goto err;
6302 			}
6303 			/*
6304 			 * Handler for "rep_based_pf" devarg.
6305 			 * Invoked as for ex: "-a 000:00:0d.0,
6306 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6307 			 */
6308 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6309 						 bnxt_parse_devarg_rep_fc_r2f,
6310 						 (void *)&representor);
6311 			if (ret) {
6312 				ret = -EINVAL;
6313 				goto err;
6314 			}
6315 			/*
6316 			 * Handler for "rep_based_pf" devarg.
6317 			 * Invoked as for ex: "-a 000:00:0d.0,
6318 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6319 			 */
6320 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6321 						 bnxt_parse_devarg_rep_fc_f2r,
6322 						 (void *)&representor);
6323 			if (ret) {
6324 				ret = -EINVAL;
6325 				goto err;
6326 			}
6327 		}
6328 
6329 		ret = rte_eth_dev_create(&pci_dev->device, name,
6330 					 sizeof(struct bnxt_representor),
6331 					 NULL, NULL,
6332 					 bnxt_representor_init,
6333 					 &representor);
6334 		if (ret) {
6335 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6336 				    "representor %s.", name);
6337 			goto err;
6338 		}
6339 
6340 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6341 		if (!vf_rep_eth_dev) {
6342 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6343 				    " for VF-Rep: %s.", name);
6344 			ret = -ENODEV;
6345 			goto err;
6346 		}
6347 
6348 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6349 			    backing_eth_dev->data->port_id);
6350 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6351 							 vf_rep_eth_dev;
6352 		backing_bp->num_reps++;
6353 
6354 	}
6355 
6356 	rte_kvargs_free(kvlist);
6357 	return 0;
6358 
6359 err:
6360 	/* If num_rep > 1, then rollback already created
6361 	 * ports, since we'll be failing the probe anyway
6362 	 */
6363 	if (num_rep > 1)
6364 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6365 	rte_errno = -ret;
6366 	rte_kvargs_free(kvlist);
6367 
6368 	return ret;
6369 }
6370 
6371 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6372 			  struct rte_pci_device *pci_dev)
6373 {
6374 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6375 	struct rte_eth_dev *backing_eth_dev;
6376 	uint16_t num_rep;
6377 	int ret = 0;
6378 
6379 	if (pci_dev->device.devargs) {
6380 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6381 					    &eth_da);
6382 		if (ret)
6383 			return ret;
6384 	}
6385 
6386 	num_rep = eth_da.nb_representor_ports;
6387 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6388 		    num_rep);
6389 
6390 	/* We could come here after first level of probe is already invoked
6391 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6392 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6393 	 */
6394 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6395 	if (backing_eth_dev == NULL) {
6396 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6397 					 sizeof(struct bnxt),
6398 					 eth_dev_pci_specific_init, pci_dev,
6399 					 bnxt_dev_init, NULL);
6400 
6401 		if (ret || !num_rep)
6402 			return ret;
6403 
6404 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6405 	}
6406 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6407 		    backing_eth_dev->data->port_id);
6408 
6409 	if (!num_rep)
6410 		return ret;
6411 
6412 	/* probe representor ports now */
6413 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6414 				  pci_dev->device.devargs->args);
6415 
6416 	return ret;
6417 }
6418 
6419 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6420 {
6421 	struct rte_eth_dev *eth_dev;
6422 
6423 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6424 	if (!eth_dev)
6425 		return 0; /* Invoked typically only by OVS-DPDK, by the
6426 			   * time it comes here the eth_dev is already
6427 			   * deleted by rte_eth_dev_close(), so returning
6428 			   * +ve value will at least help in proper cleanup
6429 			   */
6430 
6431 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6432 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6433 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6434 			return rte_eth_dev_destroy(eth_dev,
6435 						   bnxt_representor_uninit);
6436 		else
6437 			return rte_eth_dev_destroy(eth_dev,
6438 						   bnxt_dev_uninit);
6439 	} else {
6440 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6441 	}
6442 }
6443 
6444 static struct rte_pci_driver bnxt_rte_pmd = {
6445 	.id_table = bnxt_pci_id_map,
6446 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6447 			RTE_PCI_DRV_INTR_RMV |
6448 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6449 						  * and OVS-DPDK
6450 						  */
6451 	.probe = bnxt_pci_probe,
6452 	.remove = bnxt_pci_remove,
6453 };
6454 
6455 static bool
6456 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6457 {
6458 	if (strcmp(dev->device->driver->name, drv->driver.name))
6459 		return false;
6460 
6461 	return true;
6462 }
6463 
6464 bool is_bnxt_supported(struct rte_eth_dev *dev)
6465 {
6466 	return is_device_supported(dev, &bnxt_rte_pmd);
6467 }
6468 
6469 struct tf *bnxt_get_tfp_session(struct bnxt *bp, enum bnxt_session_type type)
6470 {
6471 	return (type >= BNXT_SESSION_TYPE_LAST) ?
6472 		&bp->tfp[BNXT_SESSION_TYPE_REGULAR] : &bp->tfp[type];
6473 }
6474 
6475 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6476 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6477 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6478 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6479