xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision 081e42dab11d1add2d038fdf2bd4c86b20043d08)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2014-2021 Broadcom
3  * All rights reserved.
4  */
5 
6 #include <inttypes.h>
7 #include <stdbool.h>
8 
9 #include <rte_dev.h>
10 #include <ethdev_driver.h>
11 #include <ethdev_pci.h>
12 #include <rte_malloc.h>
13 #include <rte_cycles.h>
14 #include <rte_alarm.h>
15 #include <rte_kvargs.h>
16 #include <rte_vect.h>
17 
18 #include "bnxt.h"
19 #include "bnxt_filter.h"
20 #include "bnxt_hwrm.h"
21 #include "bnxt_irq.h"
22 #include "bnxt_reps.h"
23 #include "bnxt_ring.h"
24 #include "bnxt_rxq.h"
25 #include "bnxt_rxr.h"
26 #include "bnxt_stats.h"
27 #include "bnxt_txq.h"
28 #include "bnxt_txr.h"
29 #include "bnxt_vnic.h"
30 #include "hsi_struct_def_dpdk.h"
31 #include "bnxt_nvm_defs.h"
32 #include "bnxt_tf_common.h"
33 #include "ulp_flow_db.h"
34 #include "rte_pmd_bnxt.h"
35 
36 #define DRV_MODULE_NAME		"bnxt"
37 static const char bnxt_version[] =
38 	"Broadcom NetXtreme driver " DRV_MODULE_NAME;
39 
40 /*
41  * The set of PCI devices this driver supports
42  */
43 static const struct rte_pci_id bnxt_pci_id_map[] = {
44 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
45 			 BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
46 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
47 			 BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
48 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
49 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
50 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
51 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
52 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
53 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
54 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
55 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
56 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
57 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
58 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
59 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
60 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
61 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
62 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
63 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
64 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
65 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
66 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
67 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
68 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
69 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
70 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
71 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
72 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
73 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
74 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
75 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
76 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
77 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
78 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
79 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
80 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
81 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
82 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
83 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58812) },
84 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58814) },
85 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818) },
86 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818_VF) },
87 	{ .vendor_id = 0, /* sentinel */ },
88 };
89 
90 #define	BNXT_DEVARG_ACCUM_STATS	"accum-stats"
91 #define BNXT_DEVARG_FLOW_XSTAT	"flow-xstat"
92 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
93 #define BNXT_DEVARG_REPRESENTOR	"representor"
94 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
95 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
96 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
97 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
98 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
99 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
100 #define BNXT_DEVARG_APP_ID	"app-id"
101 
102 static const char *const bnxt_dev_args[] = {
103 	BNXT_DEVARG_REPRESENTOR,
104 	BNXT_DEVARG_ACCUM_STATS,
105 	BNXT_DEVARG_FLOW_XSTAT,
106 	BNXT_DEVARG_MAX_NUM_KFLOWS,
107 	BNXT_DEVARG_REP_BASED_PF,
108 	BNXT_DEVARG_REP_IS_PF,
109 	BNXT_DEVARG_REP_Q_R2F,
110 	BNXT_DEVARG_REP_Q_F2R,
111 	BNXT_DEVARG_REP_FC_R2F,
112 	BNXT_DEVARG_REP_FC_F2R,
113 	BNXT_DEVARG_APP_ID,
114 	NULL
115 };
116 
117 /*
118  * accum-stats == false to disable flow counter accumulation
119  * accum-stats == true to enable flow counter accumulation
120  */
121 #define	BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)	((accum_stats) > 1)
122 
123 /*
124  * app-id = an non-negative 8-bit number
125  */
126 #define BNXT_DEVARG_APP_ID_INVALID(val)			((val) > 255)
127 
128 /*
129  * flow_xstat == false to disable the feature
130  * flow_xstat == true to enable the feature
131  */
132 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
133 
134 /*
135  * rep_is_pf == false to indicate VF representor
136  * rep_is_pf == true to indicate PF representor
137  */
138 #define	BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)	((rep_is_pf) > 1)
139 
140 /*
141  * rep_based_pf == Physical index of the PF
142  */
143 #define	BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)	((rep_based_pf) > 15)
144 /*
145  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
146  */
147 #define	BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)	((rep_q_r2f) > 3)
148 
149 /*
150  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
151  */
152 #define	BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)	((rep_q_f2r) > 3)
153 
154 /*
155  * rep_fc_r2f == Flow control for the representor to endpoint direction
156  */
157 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)	((rep_fc_r2f) > 1)
158 
159 /*
160  * rep_fc_f2r == Flow control for the endpoint to representor direction
161  */
162 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)	((rep_fc_f2r) > 1)
163 
164 int bnxt_cfa_code_dynfield_offset = -1;
165 
166 /*
167  * max_num_kflows must be >= 32
168  * and must be a power-of-2 supported value
169  * return: 1 -> invalid
170  *         0 -> valid
171  */
172 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
173 {
174 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
175 		return 1;
176 	return 0;
177 }
178 
179 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
180 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
181 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
182 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
183 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
184 static int bnxt_restore_vlan_filters(struct bnxt *bp);
185 static void bnxt_dev_recover(void *arg);
186 static void bnxt_free_error_recovery_info(struct bnxt *bp);
187 static void bnxt_free_rep_info(struct bnxt *bp);
188 
189 int is_bnxt_in_error(struct bnxt *bp)
190 {
191 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
192 		return -EIO;
193 	if (bp->flags & BNXT_FLAG_FW_RESET)
194 		return -EBUSY;
195 
196 	return 0;
197 }
198 
199 /***********************/
200 
201 /*
202  * High level utility functions
203  */
204 
205 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
206 {
207 	unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
208 					     BNXT_RSS_TBL_SIZE_P5);
209 
210 	if (!BNXT_CHIP_P5(bp))
211 		return 1;
212 
213 	return RTE_ALIGN_MUL_CEIL(num_rss_rings,
214 				  BNXT_RSS_ENTRIES_PER_CTX_P5) /
215 				  BNXT_RSS_ENTRIES_PER_CTX_P5;
216 }
217 
218 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
219 {
220 	if (!BNXT_CHIP_P5(bp))
221 		return HW_HASH_INDEX_SIZE;
222 
223 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
224 }
225 
226 static void bnxt_free_parent_info(struct bnxt *bp)
227 {
228 	rte_free(bp->parent);
229 	bp->parent = NULL;
230 }
231 
232 static void bnxt_free_pf_info(struct bnxt *bp)
233 {
234 	rte_free(bp->pf);
235 	bp->pf = NULL;
236 }
237 
238 static void bnxt_free_link_info(struct bnxt *bp)
239 {
240 	rte_free(bp->link_info);
241 	bp->link_info = NULL;
242 }
243 
244 static void bnxt_free_leds_info(struct bnxt *bp)
245 {
246 	if (BNXT_VF(bp))
247 		return;
248 
249 	rte_free(bp->leds);
250 	bp->leds = NULL;
251 }
252 
253 static void bnxt_free_flow_stats_info(struct bnxt *bp)
254 {
255 	rte_free(bp->flow_stat);
256 	bp->flow_stat = NULL;
257 }
258 
259 static void bnxt_free_cos_queues(struct bnxt *bp)
260 {
261 	rte_free(bp->rx_cos_queue);
262 	bp->rx_cos_queue = NULL;
263 	rte_free(bp->tx_cos_queue);
264 	bp->tx_cos_queue = NULL;
265 }
266 
267 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
268 {
269 	bnxt_free_filter_mem(bp);
270 	bnxt_free_vnic_attributes(bp);
271 	bnxt_free_vnic_mem(bp);
272 
273 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
274 	 * If the number of rings change across fw update,
275 	 * we don't have much choice except to warn the user.
276 	 */
277 	if (!reconfig) {
278 		bnxt_free_stats(bp);
279 		bnxt_free_tx_rings(bp);
280 		bnxt_free_rx_rings(bp);
281 	}
282 	bnxt_free_async_cp_ring(bp);
283 	bnxt_free_rxtx_nq_ring(bp);
284 
285 	rte_free(bp->grp_info);
286 	bp->grp_info = NULL;
287 }
288 
289 static int bnxt_alloc_parent_info(struct bnxt *bp)
290 {
291 	bp->parent = rte_zmalloc("bnxt_parent_info",
292 				 sizeof(struct bnxt_parent_info), 0);
293 	if (bp->parent == NULL)
294 		return -ENOMEM;
295 
296 	return 0;
297 }
298 
299 static int bnxt_alloc_pf_info(struct bnxt *bp)
300 {
301 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
302 	if (bp->pf == NULL)
303 		return -ENOMEM;
304 
305 	return 0;
306 }
307 
308 static int bnxt_alloc_link_info(struct bnxt *bp)
309 {
310 	bp->link_info =
311 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
312 	if (bp->link_info == NULL)
313 		return -ENOMEM;
314 
315 	return 0;
316 }
317 
318 static int bnxt_alloc_leds_info(struct bnxt *bp)
319 {
320 	if (BNXT_VF(bp))
321 		return 0;
322 
323 	bp->leds = rte_zmalloc("bnxt_leds",
324 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
325 			       0);
326 	if (bp->leds == NULL)
327 		return -ENOMEM;
328 
329 	return 0;
330 }
331 
332 static int bnxt_alloc_cos_queues(struct bnxt *bp)
333 {
334 	bp->rx_cos_queue =
335 		rte_zmalloc("bnxt_rx_cosq",
336 			    BNXT_COS_QUEUE_COUNT *
337 			    sizeof(struct bnxt_cos_queue_info),
338 			    0);
339 	if (bp->rx_cos_queue == NULL)
340 		return -ENOMEM;
341 
342 	bp->tx_cos_queue =
343 		rte_zmalloc("bnxt_tx_cosq",
344 			    BNXT_COS_QUEUE_COUNT *
345 			    sizeof(struct bnxt_cos_queue_info),
346 			    0);
347 	if (bp->tx_cos_queue == NULL)
348 		return -ENOMEM;
349 
350 	return 0;
351 }
352 
353 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
354 {
355 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
356 				    sizeof(struct bnxt_flow_stat_info), 0);
357 	if (bp->flow_stat == NULL)
358 		return -ENOMEM;
359 
360 	return 0;
361 }
362 
363 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
364 {
365 	int rc;
366 
367 	rc = bnxt_alloc_ring_grps(bp);
368 	if (rc)
369 		goto alloc_mem_err;
370 
371 	rc = bnxt_alloc_async_ring_struct(bp);
372 	if (rc)
373 		goto alloc_mem_err;
374 
375 	rc = bnxt_alloc_vnic_mem(bp);
376 	if (rc)
377 		goto alloc_mem_err;
378 
379 	rc = bnxt_alloc_vnic_attributes(bp);
380 	if (rc)
381 		goto alloc_mem_err;
382 
383 	rc = bnxt_alloc_filter_mem(bp);
384 	if (rc)
385 		goto alloc_mem_err;
386 
387 	rc = bnxt_alloc_async_cp_ring(bp);
388 	if (rc)
389 		goto alloc_mem_err;
390 
391 	rc = bnxt_alloc_rxtx_nq_ring(bp);
392 	if (rc)
393 		goto alloc_mem_err;
394 
395 	if (BNXT_FLOW_XSTATS_EN(bp)) {
396 		rc = bnxt_alloc_flow_stats_info(bp);
397 		if (rc)
398 			goto alloc_mem_err;
399 	}
400 
401 	return 0;
402 
403 alloc_mem_err:
404 	bnxt_free_mem(bp, reconfig);
405 	return rc;
406 }
407 
408 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
409 {
410 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
411 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
412 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
413 	struct bnxt_rx_queue *rxq;
414 	unsigned int j;
415 	int rc;
416 
417 	rc = bnxt_vnic_grp_alloc(bp, vnic);
418 	if (rc)
419 		goto err_out;
420 
421 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
422 		    vnic_id, vnic, vnic->fw_grp_ids);
423 
424 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
425 	if (rc)
426 		goto err_out;
427 
428 	/* Alloc RSS context only if RSS mode is enabled */
429 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
430 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
431 
432 		/* RSS table size in Thor is 512.
433 		 * Cap max Rx rings to same value
434 		 */
435 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
436 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
437 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
438 			goto err_out;
439 		}
440 
441 		rc = 0;
442 		for (j = 0; j < nr_ctxs; j++) {
443 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
444 			if (rc)
445 				break;
446 		}
447 		if (rc) {
448 			PMD_DRV_LOG(ERR,
449 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
450 				    vnic_id, j, rc);
451 			goto err_out;
452 		}
453 		vnic->num_lb_ctxts = nr_ctxs;
454 	}
455 
456 	/*
457 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
458 	 * setting is not available at this time, it will not be
459 	 * configured correctly in the CFA.
460 	 */
461 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
462 		vnic->vlan_strip = true;
463 	else
464 		vnic->vlan_strip = false;
465 
466 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
467 	if (rc)
468 		goto err_out;
469 
470 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
471 	if (rc)
472 		goto err_out;
473 
474 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
475 		rxq = bp->eth_dev->data->rx_queues[j];
476 
477 		PMD_DRV_LOG(DEBUG,
478 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
479 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
480 
481 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
482 			rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
483 		else
484 			vnic->rx_queue_cnt++;
485 	}
486 
487 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
488 
489 	rc = bnxt_vnic_rss_configure(bp, vnic);
490 	if (rc)
491 		goto err_out;
492 
493 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
494 
495 	rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
496 				    (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO) ?
497 				    true : false);
498 	if (rc)
499 		goto err_out;
500 
501 	return 0;
502 err_out:
503 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
504 		    vnic_id, rc);
505 	return rc;
506 }
507 
508 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
509 {
510 	int rc = 0;
511 
512 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
513 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
514 	if (rc)
515 		return rc;
516 
517 	PMD_DRV_LOG(DEBUG,
518 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
519 		    " rx_fc_in_tbl.ctx_id = %d\n",
520 		    bp->flow_stat->rx_fc_in_tbl.va,
521 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
522 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
523 
524 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
525 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
526 	if (rc)
527 		return rc;
528 
529 	PMD_DRV_LOG(DEBUG,
530 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
531 		    " rx_fc_out_tbl.ctx_id = %d\n",
532 		    bp->flow_stat->rx_fc_out_tbl.va,
533 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
534 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
535 
536 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
537 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
538 	if (rc)
539 		return rc;
540 
541 	PMD_DRV_LOG(DEBUG,
542 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
543 		    " tx_fc_in_tbl.ctx_id = %d\n",
544 		    bp->flow_stat->tx_fc_in_tbl.va,
545 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
546 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
547 
548 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
549 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
550 	if (rc)
551 		return rc;
552 
553 	PMD_DRV_LOG(DEBUG,
554 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
555 		    " tx_fc_out_tbl.ctx_id = %d\n",
556 		    bp->flow_stat->tx_fc_out_tbl.va,
557 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
558 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
559 
560 	memset(bp->flow_stat->rx_fc_out_tbl.va,
561 	       0,
562 	       bp->flow_stat->rx_fc_out_tbl.size);
563 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
564 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
565 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
566 				       bp->flow_stat->max_fc,
567 				       true);
568 	if (rc)
569 		return rc;
570 
571 	memset(bp->flow_stat->tx_fc_out_tbl.va,
572 	       0,
573 	       bp->flow_stat->tx_fc_out_tbl.size);
574 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
575 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
576 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
577 				       bp->flow_stat->max_fc,
578 				       true);
579 
580 	return rc;
581 }
582 
583 static int bnxt_alloc_ctx_mem_buf(struct bnxt *bp, char *type, size_t size,
584 				  struct bnxt_ctx_mem_buf_info *ctx)
585 {
586 	if (!ctx)
587 		return -EINVAL;
588 
589 	ctx->va = rte_zmalloc_socket(type, size, 0,
590 				     bp->eth_dev->device->numa_node);
591 	if (ctx->va == NULL)
592 		return -ENOMEM;
593 	rte_mem_lock_page(ctx->va);
594 	ctx->size = size;
595 	ctx->dma = rte_mem_virt2iova(ctx->va);
596 	if (ctx->dma == RTE_BAD_IOVA)
597 		return -ENOMEM;
598 
599 	return 0;
600 }
601 
602 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
603 {
604 	struct rte_pci_device *pdev = bp->pdev;
605 	char type[RTE_MEMZONE_NAMESIZE];
606 	uint16_t max_fc;
607 	int rc = 0;
608 
609 	max_fc = bp->flow_stat->max_fc;
610 
611 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
612 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
613 	/* 4 bytes for each counter-id */
614 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
615 				    max_fc * 4,
616 				    &bp->flow_stat->rx_fc_in_tbl);
617 	if (rc)
618 		return rc;
619 
620 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
621 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
622 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
623 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
624 				    max_fc * 16,
625 				    &bp->flow_stat->rx_fc_out_tbl);
626 	if (rc)
627 		return rc;
628 
629 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
630 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
631 	/* 4 bytes for each counter-id */
632 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
633 				    max_fc * 4,
634 				    &bp->flow_stat->tx_fc_in_tbl);
635 	if (rc)
636 		return rc;
637 
638 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
639 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
640 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
641 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
642 				    max_fc * 16,
643 				    &bp->flow_stat->tx_fc_out_tbl);
644 	if (rc)
645 		return rc;
646 
647 	rc = bnxt_register_fc_ctx_mem(bp);
648 
649 	return rc;
650 }
651 
652 static int bnxt_init_ctx_mem(struct bnxt *bp)
653 {
654 	int rc = 0;
655 
656 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
657 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
658 	    !BNXT_FLOW_XSTATS_EN(bp))
659 		return 0;
660 
661 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
662 	if (rc)
663 		return rc;
664 
665 	rc = bnxt_init_fc_ctx_mem(bp);
666 
667 	return rc;
668 }
669 
670 static int bnxt_update_phy_setting(struct bnxt *bp)
671 {
672 	struct rte_eth_link new;
673 	int rc;
674 
675 	rc = bnxt_get_hwrm_link_config(bp, &new);
676 	if (rc) {
677 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
678 		return rc;
679 	}
680 
681 	/*
682 	 * On BCM957508-N2100 adapters, FW will not allow any user other
683 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
684 	 * always returns link up. Force phy update always in that case.
685 	 */
686 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
687 		rc = bnxt_set_hwrm_link_config(bp, true);
688 		if (rc) {
689 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
690 			return rc;
691 		}
692 	}
693 
694 	return rc;
695 }
696 
697 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
698 {
699 	rte_free(bp->prev_rx_ring_stats);
700 	rte_free(bp->prev_tx_ring_stats);
701 
702 	bp->prev_rx_ring_stats = NULL;
703 	bp->prev_tx_ring_stats = NULL;
704 }
705 
706 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
707 {
708 	bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
709 					      sizeof(struct bnxt_ring_stats) *
710 					      bp->rx_cp_nr_rings,
711 					      0);
712 	if (bp->prev_rx_ring_stats == NULL)
713 		return -ENOMEM;
714 
715 	bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
716 					     sizeof(struct bnxt_ring_stats) *
717 					     bp->tx_cp_nr_rings,
718 					     0);
719 	if (bp->prev_tx_ring_stats == NULL)
720 		goto error;
721 
722 	return 0;
723 
724 error:
725 	bnxt_free_prev_ring_stats(bp);
726 	return -ENOMEM;
727 }
728 
729 static int bnxt_start_nic(struct bnxt *bp)
730 {
731 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
732 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
733 	uint32_t intr_vector = 0;
734 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
735 	uint32_t vec = BNXT_MISC_VEC_ID;
736 	unsigned int i, j;
737 	int rc;
738 
739 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
740 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
741 			DEV_RX_OFFLOAD_JUMBO_FRAME;
742 		bp->flags |= BNXT_FLAG_JUMBO;
743 	} else {
744 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
745 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
746 		bp->flags &= ~BNXT_FLAG_JUMBO;
747 	}
748 
749 	/* THOR does not support ring groups.
750 	 * But we will use the array to save RSS context IDs.
751 	 */
752 	if (BNXT_CHIP_P5(bp))
753 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
754 
755 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
756 	if (rc) {
757 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
758 		goto err_out;
759 	}
760 
761 	rc = bnxt_alloc_hwrm_rings(bp);
762 	if (rc) {
763 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
764 		goto err_out;
765 	}
766 
767 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
768 	if (rc) {
769 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
770 		goto err_out;
771 	}
772 
773 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
774 		goto skip_cosq_cfg;
775 
776 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
777 		if (bp->rx_cos_queue[i].id != 0xff) {
778 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
779 
780 			if (!vnic) {
781 				PMD_DRV_LOG(ERR,
782 					    "Num pools more than FW profile\n");
783 				rc = -EINVAL;
784 				goto err_out;
785 			}
786 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
787 			bp->rx_cosq_cnt++;
788 		}
789 	}
790 
791 skip_cosq_cfg:
792 	rc = bnxt_mq_rx_configure(bp);
793 	if (rc) {
794 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
795 		goto err_out;
796 	}
797 
798 	/* default vnic 0 */
799 	rc = bnxt_setup_one_vnic(bp, 0);
800 	if (rc)
801 		goto err_out;
802 	/* VNIC configuration */
803 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
804 		for (i = 1; i < bp->nr_vnics; i++) {
805 			rc = bnxt_setup_one_vnic(bp, i);
806 			if (rc)
807 				goto err_out;
808 		}
809 	}
810 
811 	for (j = 0; j < bp->tx_nr_rings; j++) {
812 		struct bnxt_tx_queue *txq = bp->tx_queues[j];
813 
814 		if (!txq->tx_deferred_start) {
815 			bp->eth_dev->data->tx_queue_state[j] =
816 				RTE_ETH_QUEUE_STATE_STARTED;
817 			txq->tx_started = true;
818 		}
819 	}
820 
821 	for (j = 0; j < bp->rx_nr_rings; j++) {
822 		struct bnxt_rx_queue *rxq = bp->rx_queues[j];
823 
824 		if (!rxq->rx_deferred_start) {
825 			bp->eth_dev->data->rx_queue_state[j] =
826 				RTE_ETH_QUEUE_STATE_STARTED;
827 			rxq->rx_started = true;
828 		}
829 	}
830 
831 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
832 	if (rc) {
833 		PMD_DRV_LOG(ERR,
834 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
835 		goto err_out;
836 	}
837 
838 	/* check and configure queue intr-vector mapping */
839 	if ((rte_intr_cap_multiple(intr_handle) ||
840 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
841 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
842 		intr_vector = bp->eth_dev->data->nb_rx_queues;
843 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
844 		if (intr_vector > bp->rx_cp_nr_rings) {
845 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
846 					bp->rx_cp_nr_rings);
847 			return -ENOTSUP;
848 		}
849 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
850 		if (rc)
851 			return rc;
852 	}
853 
854 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
855 		intr_handle->intr_vec =
856 			rte_zmalloc("intr_vec",
857 				    bp->eth_dev->data->nb_rx_queues *
858 				    sizeof(int), 0);
859 		if (intr_handle->intr_vec == NULL) {
860 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
861 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
862 			rc = -ENOMEM;
863 			goto err_out;
864 		}
865 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
866 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
867 			 intr_handle->intr_vec, intr_handle->nb_efd,
868 			intr_handle->max_intr);
869 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
870 		     queue_id++) {
871 			intr_handle->intr_vec[queue_id] =
872 							vec + BNXT_RX_VEC_START;
873 			if (vec < base + intr_handle->nb_efd - 1)
874 				vec++;
875 		}
876 	}
877 
878 	/* enable uio/vfio intr/eventfd mapping */
879 	rc = rte_intr_enable(intr_handle);
880 #ifndef RTE_EXEC_ENV_FREEBSD
881 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
882 	if (rc)
883 		goto err_out;
884 #endif
885 
886 	rc = bnxt_update_phy_setting(bp);
887 	if (rc)
888 		goto err_out;
889 
890 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
891 	if (!bp->mark_table)
892 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
893 
894 	return 0;
895 
896 err_out:
897 	/* Some of the error status returned by FW may not be from errno.h */
898 	if (rc > 0)
899 		rc = -EIO;
900 
901 	return rc;
902 }
903 
904 static int bnxt_shutdown_nic(struct bnxt *bp)
905 {
906 	bnxt_free_all_hwrm_resources(bp);
907 	bnxt_free_all_filters(bp);
908 	bnxt_free_all_vnics(bp);
909 	return 0;
910 }
911 
912 /*
913  * Device configuration and status function
914  */
915 
916 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
917 {
918 	uint32_t link_speed = 0;
919 	uint32_t speed_capa = 0;
920 
921 	if (bp->link_info == NULL)
922 		return 0;
923 
924 	link_speed = bp->link_info->support_speeds;
925 
926 	/* If PAM4 is configured, use PAM4 supported speed */
927 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
928 		link_speed = bp->link_info->support_pam4_speeds;
929 
930 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
931 		speed_capa |= ETH_LINK_SPEED_100M;
932 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
933 		speed_capa |= ETH_LINK_SPEED_100M_HD;
934 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
935 		speed_capa |= ETH_LINK_SPEED_1G;
936 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
937 		speed_capa |= ETH_LINK_SPEED_2_5G;
938 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
939 		speed_capa |= ETH_LINK_SPEED_10G;
940 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
941 		speed_capa |= ETH_LINK_SPEED_20G;
942 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
943 		speed_capa |= ETH_LINK_SPEED_25G;
944 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
945 		speed_capa |= ETH_LINK_SPEED_40G;
946 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
947 		speed_capa |= ETH_LINK_SPEED_50G;
948 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
949 		speed_capa |= ETH_LINK_SPEED_100G;
950 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
951 		speed_capa |= ETH_LINK_SPEED_50G;
952 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
953 		speed_capa |= ETH_LINK_SPEED_100G;
954 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
955 		speed_capa |= ETH_LINK_SPEED_200G;
956 
957 	if (bp->link_info->auto_mode ==
958 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
959 		speed_capa |= ETH_LINK_SPEED_FIXED;
960 
961 	return speed_capa;
962 }
963 
964 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
965 				struct rte_eth_dev_info *dev_info)
966 {
967 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
968 	struct bnxt *bp = eth_dev->data->dev_private;
969 	uint16_t max_vnics, i, j, vpool, vrxq;
970 	unsigned int max_rx_rings;
971 	int rc;
972 
973 	rc = is_bnxt_in_error(bp);
974 	if (rc)
975 		return rc;
976 
977 	/* MAC Specifics */
978 	dev_info->max_mac_addrs = bp->max_l2_ctx;
979 	dev_info->max_hash_mac_addrs = 0;
980 
981 	/* PF/VF specifics */
982 	if (BNXT_PF(bp))
983 		dev_info->max_vfs = pdev->max_vfs;
984 
985 	max_rx_rings = bnxt_max_rings(bp);
986 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
987 	dev_info->max_rx_queues = max_rx_rings;
988 	dev_info->max_tx_queues = max_rx_rings;
989 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
990 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
991 	max_vnics = bp->max_vnics;
992 
993 	/* MTU specifics */
994 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
995 	dev_info->max_mtu = BNXT_MAX_MTU;
996 
997 	/* Fast path specifics */
998 	dev_info->min_rx_bufsize = 1;
999 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
1000 
1001 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
1002 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
1003 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
1004 	if (bp->vnic_cap_flags & BNXT_VNIC_CAP_VLAN_RX_STRIP)
1005 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_VLAN_STRIP;
1006 	dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
1007 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
1008 				    dev_info->tx_queue_offload_capa;
1009 	if (bp->fw_cap & BNXT_FW_CAP_VLAN_TX_INSERT)
1010 		dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_VLAN_INSERT;
1011 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
1012 
1013 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
1014 	dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
1015 			     RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
1016 
1017 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
1018 		.rx_thresh = {
1019 			.pthresh = 8,
1020 			.hthresh = 8,
1021 			.wthresh = 0,
1022 		},
1023 		.rx_free_thresh = 32,
1024 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
1025 	};
1026 
1027 	dev_info->default_txconf = (struct rte_eth_txconf) {
1028 		.tx_thresh = {
1029 			.pthresh = 32,
1030 			.hthresh = 0,
1031 			.wthresh = 0,
1032 		},
1033 		.tx_free_thresh = 32,
1034 		.tx_rs_thresh = 32,
1035 	};
1036 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
1037 
1038 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1039 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1040 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1041 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1042 
1043 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1044 		dev_info->switch_info.name = eth_dev->device->name;
1045 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1046 		dev_info->switch_info.port_id =
1047 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1048 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1049 	}
1050 
1051 	/*
1052 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1053 	 *       need further investigation.
1054 	 */
1055 
1056 	/* VMDq resources */
1057 	vpool = 64; /* ETH_64_POOLS */
1058 	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
1059 	for (i = 0; i < 4; vpool >>= 1, i++) {
1060 		if (max_vnics > vpool) {
1061 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1062 				if (dev_info->max_rx_queues > vrxq) {
1063 					if (vpool > vrxq)
1064 						vpool = vrxq;
1065 					goto found;
1066 				}
1067 			}
1068 			/* Not enough resources to support VMDq */
1069 			break;
1070 		}
1071 	}
1072 	/* Not enough resources to support VMDq */
1073 	vpool = 0;
1074 	vrxq = 0;
1075 found:
1076 	dev_info->max_vmdq_pools = vpool;
1077 	dev_info->vmdq_queue_num = vrxq;
1078 
1079 	dev_info->vmdq_pool_base = 0;
1080 	dev_info->vmdq_queue_base = 0;
1081 
1082 	return 0;
1083 }
1084 
1085 /* Configure the device based on the configuration provided */
1086 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1087 {
1088 	struct bnxt *bp = eth_dev->data->dev_private;
1089 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1090 	int rc;
1091 
1092 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1093 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1094 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1095 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1096 
1097 	rc = is_bnxt_in_error(bp);
1098 	if (rc)
1099 		return rc;
1100 
1101 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1102 		rc = bnxt_hwrm_check_vf_rings(bp);
1103 		if (rc) {
1104 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1105 			return -ENOSPC;
1106 		}
1107 
1108 		/* If a resource has already been allocated - in this case
1109 		 * it is the async completion ring, free it. Reallocate it after
1110 		 * resource reservation. This will ensure the resource counts
1111 		 * are calculated correctly.
1112 		 */
1113 
1114 		pthread_mutex_lock(&bp->def_cp_lock);
1115 
1116 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1117 			bnxt_disable_int(bp);
1118 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1119 		}
1120 
1121 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1122 		if (rc) {
1123 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1124 			pthread_mutex_unlock(&bp->def_cp_lock);
1125 			return -ENOSPC;
1126 		}
1127 
1128 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1129 			rc = bnxt_alloc_async_cp_ring(bp);
1130 			if (rc) {
1131 				pthread_mutex_unlock(&bp->def_cp_lock);
1132 				return rc;
1133 			}
1134 			bnxt_enable_int(bp);
1135 		}
1136 
1137 		pthread_mutex_unlock(&bp->def_cp_lock);
1138 	}
1139 
1140 	/* Inherit new configurations */
1141 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1142 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1143 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1144 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1145 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1146 	    bp->max_stat_ctx)
1147 		goto resource_error;
1148 
1149 	if (BNXT_HAS_RING_GRPS(bp) &&
1150 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1151 		goto resource_error;
1152 
1153 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1154 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1155 		goto resource_error;
1156 
1157 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1158 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1159 
1160 	if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1161 		rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1162 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1163 
1164 	if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1165 		eth_dev->data->mtu =
1166 			eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1167 			RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1168 			BNXT_NUM_VLANS;
1169 		bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1170 	}
1171 	return 0;
1172 
1173 resource_error:
1174 	PMD_DRV_LOG(ERR,
1175 		    "Insufficient resources to support requested config\n");
1176 	PMD_DRV_LOG(ERR,
1177 		    "Num Queues Requested: Tx %d, Rx %d\n",
1178 		    eth_dev->data->nb_tx_queues,
1179 		    eth_dev->data->nb_rx_queues);
1180 	PMD_DRV_LOG(ERR,
1181 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1182 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1183 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1184 	return -ENOSPC;
1185 }
1186 
1187 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1188 {
1189 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1190 
1191 	if (link->link_status)
1192 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1193 			eth_dev->data->port_id,
1194 			(uint32_t)link->link_speed,
1195 			(link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1196 			("full-duplex") : ("half-duplex\n"));
1197 	else
1198 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1199 			eth_dev->data->port_id);
1200 }
1201 
1202 /*
1203  * Determine whether the current configuration requires support for scattered
1204  * receive; return 1 if scattered receive is required and 0 if not.
1205  */
1206 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1207 {
1208 	uint16_t buf_size;
1209 	int i;
1210 
1211 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1212 		return 1;
1213 
1214 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1215 		return 1;
1216 
1217 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1218 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1219 
1220 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1221 				      RTE_PKTMBUF_HEADROOM);
1222 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1223 			return 1;
1224 	}
1225 	return 0;
1226 }
1227 
1228 static eth_rx_burst_t
1229 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1230 {
1231 	struct bnxt *bp = eth_dev->data->dev_private;
1232 
1233 	/* Disable vector mode RX for Stingray2 for now */
1234 	if (BNXT_CHIP_SR2(bp)) {
1235 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1236 		return bnxt_recv_pkts;
1237 	}
1238 
1239 #if (defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)) && \
1240 	!defined(RTE_LIBRTE_IEEE1588)
1241 
1242 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1243 	if (eth_dev->data->scattered_rx)
1244 		goto use_scalar_rx;
1245 
1246 	/*
1247 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1248 	 * asynchronous completions and receive completions can be placed in
1249 	 * the same completion ring.
1250 	 */
1251 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1252 		goto use_scalar_rx;
1253 
1254 	/*
1255 	 * Vector mode receive cannot be enabled if any receive offloads outside
1256 	 * a limited subset have been enabled.
1257 	 */
1258 	if (eth_dev->data->dev_conf.rxmode.offloads &
1259 		~(DEV_RX_OFFLOAD_VLAN_STRIP |
1260 		  DEV_RX_OFFLOAD_KEEP_CRC |
1261 		  DEV_RX_OFFLOAD_JUMBO_FRAME |
1262 		  DEV_RX_OFFLOAD_IPV4_CKSUM |
1263 		  DEV_RX_OFFLOAD_UDP_CKSUM |
1264 		  DEV_RX_OFFLOAD_TCP_CKSUM |
1265 		  DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1266 		  DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1267 		  DEV_RX_OFFLOAD_RSS_HASH |
1268 		  DEV_RX_OFFLOAD_VLAN_FILTER))
1269 		goto use_scalar_rx;
1270 
1271 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1272 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1273 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1274 		PMD_DRV_LOG(INFO,
1275 			    "Using AVX2 vector mode receive for port %d\n",
1276 			    eth_dev->data->port_id);
1277 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1278 		return bnxt_recv_pkts_vec_avx2;
1279 	}
1280  #endif
1281 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1282 		PMD_DRV_LOG(INFO,
1283 			    "Using SSE vector mode receive for port %d\n",
1284 			    eth_dev->data->port_id);
1285 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1286 		return bnxt_recv_pkts_vec;
1287 	}
1288 
1289 use_scalar_rx:
1290 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1291 		    eth_dev->data->port_id);
1292 	PMD_DRV_LOG(INFO,
1293 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1294 		    eth_dev->data->port_id,
1295 		    eth_dev->data->scattered_rx,
1296 		    eth_dev->data->dev_conf.rxmode.offloads);
1297 #endif
1298 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1299 	return bnxt_recv_pkts;
1300 }
1301 
1302 static eth_tx_burst_t
1303 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1304 {
1305 	struct bnxt *bp = eth_dev->data->dev_private;
1306 
1307 	/* Disable vector mode TX for Stingray2 for now */
1308 	if (BNXT_CHIP_SR2(bp))
1309 		return bnxt_xmit_pkts;
1310 
1311 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64) && \
1312 	!defined(RTE_LIBRTE_IEEE1588)
1313 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1314 
1315 	/*
1316 	 * Vector mode transmit can be enabled only if not using scatter rx
1317 	 * or tx offloads.
1318 	 */
1319 	if (eth_dev->data->scattered_rx ||
1320 	    (offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) ||
1321 	    BNXT_TRUFLOW_EN(bp))
1322 		goto use_scalar_tx;
1323 
1324 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1325 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1326 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1327 		PMD_DRV_LOG(INFO,
1328 			    "Using AVX2 vector mode transmit for port %d\n",
1329 			    eth_dev->data->port_id);
1330 		return bnxt_xmit_pkts_vec_avx2;
1331 	}
1332 #endif
1333 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1334 		PMD_DRV_LOG(INFO,
1335 			    "Using SSE vector mode transmit for port %d\n",
1336 			    eth_dev->data->port_id);
1337 		return bnxt_xmit_pkts_vec;
1338 	}
1339 
1340 use_scalar_tx:
1341 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1342 		    eth_dev->data->port_id);
1343 	PMD_DRV_LOG(INFO,
1344 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1345 		    eth_dev->data->port_id,
1346 		    eth_dev->data->scattered_rx,
1347 		    offloads);
1348 #endif
1349 	return bnxt_xmit_pkts;
1350 }
1351 
1352 static int bnxt_handle_if_change_status(struct bnxt *bp)
1353 {
1354 	int rc;
1355 
1356 	/* Since fw has undergone a reset and lost all contexts,
1357 	 * set fatal flag to not issue hwrm during cleanup
1358 	 */
1359 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1360 	bnxt_uninit_resources(bp, true);
1361 
1362 	/* clear fatal flag so that re-init happens */
1363 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1364 	rc = bnxt_init_resources(bp, true);
1365 
1366 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1367 
1368 	return rc;
1369 }
1370 
1371 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1372 {
1373 	struct bnxt *bp = eth_dev->data->dev_private;
1374 	int rc = 0;
1375 
1376 	if (!BNXT_SINGLE_PF(bp))
1377 		return -ENOTSUP;
1378 
1379 	if (!bp->link_info->link_up)
1380 		rc = bnxt_set_hwrm_link_config(bp, true);
1381 	if (!rc)
1382 		eth_dev->data->dev_link.link_status = 1;
1383 
1384 	bnxt_print_link_info(eth_dev);
1385 	return rc;
1386 }
1387 
1388 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1389 {
1390 	struct bnxt *bp = eth_dev->data->dev_private;
1391 
1392 	if (!BNXT_SINGLE_PF(bp))
1393 		return -ENOTSUP;
1394 
1395 	eth_dev->data->dev_link.link_status = 0;
1396 	bnxt_set_hwrm_link_config(bp, false);
1397 	bp->link_info->link_up = 0;
1398 
1399 	return 0;
1400 }
1401 
1402 static void bnxt_free_switch_domain(struct bnxt *bp)
1403 {
1404 	int rc = 0;
1405 
1406 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1407 		return;
1408 
1409 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1410 	if (rc)
1411 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1412 			    bp->switch_domain_id, rc);
1413 }
1414 
1415 static void bnxt_ptp_get_current_time(void *arg)
1416 {
1417 	struct bnxt *bp = arg;
1418 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1419 	int rc;
1420 
1421 	rc = is_bnxt_in_error(bp);
1422 	if (rc)
1423 		return;
1424 
1425 	if (!ptp)
1426 		return;
1427 
1428 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1429 				&ptp->current_time);
1430 
1431 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1432 	if (rc != 0) {
1433 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1434 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1435 	}
1436 }
1437 
1438 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1439 {
1440 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1441 	int rc;
1442 
1443 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1444 		return 0;
1445 
1446 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1447 				&ptp->current_time);
1448 
1449 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1450 	return rc;
1451 }
1452 
1453 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1454 {
1455 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1456 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1457 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1458 	}
1459 }
1460 
1461 static void bnxt_ptp_stop(struct bnxt *bp)
1462 {
1463 	bnxt_cancel_ptp_alarm(bp);
1464 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1465 }
1466 
1467 static int bnxt_ptp_start(struct bnxt *bp)
1468 {
1469 	int rc;
1470 
1471 	rc = bnxt_schedule_ptp_alarm(bp);
1472 	if (rc != 0) {
1473 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1474 	} else {
1475 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1476 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1477 	}
1478 
1479 	return rc;
1480 }
1481 
1482 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1483 {
1484 	struct bnxt *bp = eth_dev->data->dev_private;
1485 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1486 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1487 	struct rte_eth_link link;
1488 	int ret;
1489 
1490 	eth_dev->data->dev_started = 0;
1491 	eth_dev->data->scattered_rx = 0;
1492 
1493 	/* Prevent crashes when queues are still in use */
1494 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1495 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1496 
1497 	bnxt_disable_int(bp);
1498 
1499 	/* disable uio/vfio intr/eventfd mapping */
1500 	rte_intr_disable(intr_handle);
1501 
1502 	/* Stop the child representors for this device */
1503 	ret = bnxt_rep_stop_all(bp);
1504 	if (ret != 0)
1505 		return ret;
1506 
1507 	/* delete the bnxt ULP port details */
1508 	bnxt_ulp_port_deinit(bp);
1509 
1510 	bnxt_cancel_fw_health_check(bp);
1511 
1512 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1513 		bnxt_cancel_ptp_alarm(bp);
1514 
1515 	/* Do not bring link down during reset recovery */
1516 	if (!is_bnxt_in_error(bp)) {
1517 		bnxt_dev_set_link_down_op(eth_dev);
1518 		/* Wait for link to be reset */
1519 		if (BNXT_SINGLE_PF(bp))
1520 			rte_delay_ms(500);
1521 		/* clear the recorded link status */
1522 		memset(&link, 0, sizeof(link));
1523 		rte_eth_linkstatus_set(eth_dev, &link);
1524 	}
1525 
1526 	/* Clean queue intr-vector mapping */
1527 	rte_intr_efd_disable(intr_handle);
1528 	if (intr_handle->intr_vec != NULL) {
1529 		rte_free(intr_handle->intr_vec);
1530 		intr_handle->intr_vec = NULL;
1531 	}
1532 
1533 	bnxt_hwrm_port_clr_stats(bp);
1534 	bnxt_free_tx_mbufs(bp);
1535 	bnxt_free_rx_mbufs(bp);
1536 	/* Process any remaining notifications in default completion queue */
1537 	bnxt_int_handler(eth_dev);
1538 	bnxt_shutdown_nic(bp);
1539 	bnxt_hwrm_if_change(bp, false);
1540 
1541 	bnxt_free_prev_ring_stats(bp);
1542 	rte_free(bp->mark_table);
1543 	bp->mark_table = NULL;
1544 
1545 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1546 	bp->rx_cosq_cnt = 0;
1547 	/* All filters are deleted on a port stop. */
1548 	if (BNXT_FLOW_XSTATS_EN(bp))
1549 		bp->flow_stat->flow_count = 0;
1550 
1551 	return 0;
1552 }
1553 
1554 /* Unload the driver, release resources */
1555 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1556 {
1557 	struct bnxt *bp = eth_dev->data->dev_private;
1558 
1559 	pthread_mutex_lock(&bp->err_recovery_lock);
1560 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1561 		PMD_DRV_LOG(ERR,
1562 			    "Adapter recovering from error..Please retry\n");
1563 		pthread_mutex_unlock(&bp->err_recovery_lock);
1564 		return -EAGAIN;
1565 	}
1566 	pthread_mutex_unlock(&bp->err_recovery_lock);
1567 
1568 	return bnxt_dev_stop(eth_dev);
1569 }
1570 
1571 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1572 {
1573 	struct bnxt *bp = eth_dev->data->dev_private;
1574 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1575 	int vlan_mask = 0;
1576 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1577 
1578 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1579 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1580 		return -EINVAL;
1581 	}
1582 
1583 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1584 		PMD_DRV_LOG(ERR,
1585 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1586 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1587 
1588 	do {
1589 		rc = bnxt_hwrm_if_change(bp, true);
1590 		if (rc == 0 || rc != -EAGAIN)
1591 			break;
1592 
1593 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1594 	} while (retry_cnt--);
1595 
1596 	if (rc)
1597 		return rc;
1598 
1599 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1600 		rc = bnxt_handle_if_change_status(bp);
1601 		if (rc)
1602 			return rc;
1603 	}
1604 
1605 	bnxt_enable_int(bp);
1606 
1607 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1608 
1609 	rc = bnxt_start_nic(bp);
1610 	if (rc)
1611 		goto error;
1612 
1613 	rc = bnxt_alloc_prev_ring_stats(bp);
1614 	if (rc)
1615 		goto error;
1616 
1617 	eth_dev->data->dev_started = 1;
1618 
1619 	bnxt_link_update_op(eth_dev, 1);
1620 
1621 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1622 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1623 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1624 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1625 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1626 	if (rc)
1627 		goto error;
1628 
1629 	/* Initialize bnxt ULP port details */
1630 	rc = bnxt_ulp_port_init(bp);
1631 	if (rc)
1632 		goto error;
1633 
1634 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1635 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1636 
1637 	bnxt_schedule_fw_health_check(bp);
1638 
1639 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1640 		bnxt_schedule_ptp_alarm(bp);
1641 
1642 	return 0;
1643 
1644 error:
1645 	bnxt_dev_stop(eth_dev);
1646 	return rc;
1647 }
1648 
1649 static void
1650 bnxt_uninit_locks(struct bnxt *bp)
1651 {
1652 	pthread_mutex_destroy(&bp->flow_lock);
1653 	pthread_mutex_destroy(&bp->def_cp_lock);
1654 	pthread_mutex_destroy(&bp->health_check_lock);
1655 	pthread_mutex_destroy(&bp->err_recovery_lock);
1656 	if (bp->rep_info) {
1657 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1658 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1659 	}
1660 }
1661 
1662 static void bnxt_drv_uninit(struct bnxt *bp)
1663 {
1664 	bnxt_free_leds_info(bp);
1665 	bnxt_free_cos_queues(bp);
1666 	bnxt_free_link_info(bp);
1667 	bnxt_free_parent_info(bp);
1668 	bnxt_uninit_locks(bp);
1669 
1670 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1671 	bp->tx_mem_zone = NULL;
1672 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1673 	bp->rx_mem_zone = NULL;
1674 
1675 	bnxt_free_vf_info(bp);
1676 	bnxt_free_pf_info(bp);
1677 
1678 	rte_free(bp->grp_info);
1679 	bp->grp_info = NULL;
1680 }
1681 
1682 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1683 {
1684 	struct bnxt *bp = eth_dev->data->dev_private;
1685 	int ret = 0;
1686 
1687 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1688 		return 0;
1689 
1690 	pthread_mutex_lock(&bp->err_recovery_lock);
1691 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1692 		PMD_DRV_LOG(ERR,
1693 			    "Adapter recovering from error...Please retry\n");
1694 		pthread_mutex_unlock(&bp->err_recovery_lock);
1695 		return -EAGAIN;
1696 	}
1697 	pthread_mutex_unlock(&bp->err_recovery_lock);
1698 
1699 	/* cancel the recovery handler before remove dev */
1700 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1701 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1702 	bnxt_cancel_fc_thread(bp);
1703 
1704 	if (eth_dev->data->dev_started)
1705 		ret = bnxt_dev_stop(eth_dev);
1706 
1707 	bnxt_uninit_resources(bp, false);
1708 
1709 	bnxt_drv_uninit(bp);
1710 
1711 	return ret;
1712 }
1713 
1714 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1715 				    uint32_t index)
1716 {
1717 	struct bnxt *bp = eth_dev->data->dev_private;
1718 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1719 	struct bnxt_vnic_info *vnic;
1720 	struct bnxt_filter_info *filter, *temp_filter;
1721 	uint32_t i;
1722 
1723 	if (is_bnxt_in_error(bp))
1724 		return;
1725 
1726 	/*
1727 	 * Loop through all VNICs from the specified filter flow pools to
1728 	 * remove the corresponding MAC addr filter
1729 	 */
1730 	for (i = 0; i < bp->nr_vnics; i++) {
1731 		if (!(pool_mask & (1ULL << i)))
1732 			continue;
1733 
1734 		vnic = &bp->vnic_info[i];
1735 		filter = STAILQ_FIRST(&vnic->filter);
1736 		while (filter) {
1737 			temp_filter = STAILQ_NEXT(filter, next);
1738 			if (filter->mac_index == index) {
1739 				STAILQ_REMOVE(&vnic->filter, filter,
1740 						bnxt_filter_info, next);
1741 				bnxt_hwrm_clear_l2_filter(bp, filter);
1742 				bnxt_free_filter(bp, filter);
1743 			}
1744 			filter = temp_filter;
1745 		}
1746 	}
1747 }
1748 
1749 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1750 			       struct rte_ether_addr *mac_addr, uint32_t index,
1751 			       uint32_t pool)
1752 {
1753 	struct bnxt_filter_info *filter;
1754 	int rc = 0;
1755 
1756 	/* Attach requested MAC address to the new l2_filter */
1757 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1758 		if (filter->mac_index == index) {
1759 			PMD_DRV_LOG(DEBUG,
1760 				    "MAC addr already existed for pool %d\n",
1761 				    pool);
1762 			return 0;
1763 		}
1764 	}
1765 
1766 	filter = bnxt_alloc_filter(bp);
1767 	if (!filter) {
1768 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1769 		return -ENODEV;
1770 	}
1771 
1772 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1773 	 * if the MAC that's been programmed now is a different one, then,
1774 	 * copy that addr to filter->l2_addr
1775 	 */
1776 	if (mac_addr)
1777 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1778 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1779 
1780 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1781 	if (!rc) {
1782 		filter->mac_index = index;
1783 		if (filter->mac_index == 0)
1784 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1785 		else
1786 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1787 	} else {
1788 		bnxt_free_filter(bp, filter);
1789 	}
1790 
1791 	return rc;
1792 }
1793 
1794 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1795 				struct rte_ether_addr *mac_addr,
1796 				uint32_t index, uint32_t pool)
1797 {
1798 	struct bnxt *bp = eth_dev->data->dev_private;
1799 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1800 	int rc = 0;
1801 
1802 	rc = is_bnxt_in_error(bp);
1803 	if (rc)
1804 		return rc;
1805 
1806 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1807 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1808 		return -ENOTSUP;
1809 	}
1810 
1811 	if (!vnic) {
1812 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1813 		return -EINVAL;
1814 	}
1815 
1816 	/* Filter settings will get applied when port is started */
1817 	if (!eth_dev->data->dev_started)
1818 		return 0;
1819 
1820 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1821 
1822 	return rc;
1823 }
1824 
1825 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1826 {
1827 	int rc = 0;
1828 	struct bnxt *bp = eth_dev->data->dev_private;
1829 	struct rte_eth_link new;
1830 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1831 			BNXT_MIN_LINK_WAIT_CNT;
1832 
1833 	rc = is_bnxt_in_error(bp);
1834 	if (rc)
1835 		return rc;
1836 
1837 	memset(&new, 0, sizeof(new));
1838 
1839 	if (bp->link_info == NULL)
1840 		goto out;
1841 
1842 	do {
1843 		/* Retrieve link info from hardware */
1844 		rc = bnxt_get_hwrm_link_config(bp, &new);
1845 		if (rc) {
1846 			new.link_speed = ETH_LINK_SPEED_100M;
1847 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1848 			PMD_DRV_LOG(ERR,
1849 				"Failed to retrieve link rc = 0x%x!\n", rc);
1850 			goto out;
1851 		}
1852 
1853 		if (!wait_to_complete || new.link_status)
1854 			break;
1855 
1856 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1857 	} while (cnt--);
1858 
1859 	/* Only single function PF can bring phy down.
1860 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1861 	 */
1862 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1863 		memset(&new, 0, sizeof(new));
1864 
1865 out:
1866 	/* Timed out or success */
1867 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1868 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1869 		rte_eth_linkstatus_set(eth_dev, &new);
1870 		bnxt_print_link_info(eth_dev);
1871 	}
1872 
1873 	return rc;
1874 }
1875 
1876 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1877 {
1878 	struct bnxt *bp = eth_dev->data->dev_private;
1879 	struct bnxt_vnic_info *vnic;
1880 	uint32_t old_flags;
1881 	int rc;
1882 
1883 	rc = is_bnxt_in_error(bp);
1884 	if (rc)
1885 		return rc;
1886 
1887 	/* Filter settings will get applied when port is started */
1888 	if (!eth_dev->data->dev_started)
1889 		return 0;
1890 
1891 	if (bp->vnic_info == NULL)
1892 		return 0;
1893 
1894 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1895 
1896 	old_flags = vnic->flags;
1897 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1898 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1899 	if (rc != 0)
1900 		vnic->flags = old_flags;
1901 
1902 	return rc;
1903 }
1904 
1905 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1906 {
1907 	struct bnxt *bp = eth_dev->data->dev_private;
1908 	struct bnxt_vnic_info *vnic;
1909 	uint32_t old_flags;
1910 	int rc;
1911 
1912 	rc = is_bnxt_in_error(bp);
1913 	if (rc)
1914 		return rc;
1915 
1916 	/* Filter settings will get applied when port is started */
1917 	if (!eth_dev->data->dev_started)
1918 		return 0;
1919 
1920 	if (bp->vnic_info == NULL)
1921 		return 0;
1922 
1923 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1924 
1925 	old_flags = vnic->flags;
1926 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1927 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1928 	if (rc != 0)
1929 		vnic->flags = old_flags;
1930 
1931 	return rc;
1932 }
1933 
1934 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1935 {
1936 	struct bnxt *bp = eth_dev->data->dev_private;
1937 	struct bnxt_vnic_info *vnic;
1938 	uint32_t old_flags;
1939 	int rc;
1940 
1941 	rc = is_bnxt_in_error(bp);
1942 	if (rc)
1943 		return rc;
1944 
1945 	/* Filter settings will get applied when port is started */
1946 	if (!eth_dev->data->dev_started)
1947 		return 0;
1948 
1949 	if (bp->vnic_info == NULL)
1950 		return 0;
1951 
1952 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1953 
1954 	old_flags = vnic->flags;
1955 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1956 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1957 	if (rc != 0)
1958 		vnic->flags = old_flags;
1959 
1960 	return rc;
1961 }
1962 
1963 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1964 {
1965 	struct bnxt *bp = eth_dev->data->dev_private;
1966 	struct bnxt_vnic_info *vnic;
1967 	uint32_t old_flags;
1968 	int rc;
1969 
1970 	rc = is_bnxt_in_error(bp);
1971 	if (rc)
1972 		return rc;
1973 
1974 	/* Filter settings will get applied when port is started */
1975 	if (!eth_dev->data->dev_started)
1976 		return 0;
1977 
1978 	if (bp->vnic_info == NULL)
1979 		return 0;
1980 
1981 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1982 
1983 	old_flags = vnic->flags;
1984 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1985 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1986 	if (rc != 0)
1987 		vnic->flags = old_flags;
1988 
1989 	return rc;
1990 }
1991 
1992 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1993 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1994 {
1995 	if (qid >= bp->rx_nr_rings)
1996 		return NULL;
1997 
1998 	return bp->eth_dev->data->rx_queues[qid];
1999 }
2000 
2001 /* Return rxq corresponding to a given rss table ring/group ID. */
2002 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
2003 {
2004 	struct bnxt_rx_queue *rxq;
2005 	unsigned int i;
2006 
2007 	if (!BNXT_HAS_RING_GRPS(bp)) {
2008 		for (i = 0; i < bp->rx_nr_rings; i++) {
2009 			rxq = bp->eth_dev->data->rx_queues[i];
2010 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
2011 				return rxq->index;
2012 		}
2013 	} else {
2014 		for (i = 0; i < bp->rx_nr_rings; i++) {
2015 			if (bp->grp_info[i].fw_grp_id == fwr)
2016 				return i;
2017 		}
2018 	}
2019 
2020 	return INVALID_HW_RING_ID;
2021 }
2022 
2023 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
2024 			    struct rte_eth_rss_reta_entry64 *reta_conf,
2025 			    uint16_t reta_size)
2026 {
2027 	struct bnxt *bp = eth_dev->data->dev_private;
2028 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2029 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2030 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2031 	uint16_t idx, sft;
2032 	int i, rc;
2033 
2034 	rc = is_bnxt_in_error(bp);
2035 	if (rc)
2036 		return rc;
2037 
2038 	if (!vnic->rss_table)
2039 		return -EINVAL;
2040 
2041 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
2042 		return -EINVAL;
2043 
2044 	if (reta_size != tbl_size) {
2045 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2046 			"(%d) must equal the size supported by the hardware "
2047 			"(%d)\n", reta_size, tbl_size);
2048 		return -EINVAL;
2049 	}
2050 
2051 	for (i = 0; i < reta_size; i++) {
2052 		struct bnxt_rx_queue *rxq;
2053 
2054 		idx = i / RTE_RETA_GROUP_SIZE;
2055 		sft = i % RTE_RETA_GROUP_SIZE;
2056 
2057 		if (!(reta_conf[idx].mask & (1ULL << sft)))
2058 			continue;
2059 
2060 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2061 		if (!rxq) {
2062 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
2063 			return -EINVAL;
2064 		}
2065 
2066 		if (BNXT_CHIP_P5(bp)) {
2067 			vnic->rss_table[i * 2] =
2068 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
2069 			vnic->rss_table[i * 2 + 1] =
2070 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
2071 		} else {
2072 			vnic->rss_table[i] =
2073 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2074 		}
2075 	}
2076 
2077 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2078 	return rc;
2079 }
2080 
2081 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2082 			      struct rte_eth_rss_reta_entry64 *reta_conf,
2083 			      uint16_t reta_size)
2084 {
2085 	struct bnxt *bp = eth_dev->data->dev_private;
2086 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2087 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2088 	uint16_t idx, sft, i;
2089 	int rc;
2090 
2091 	rc = is_bnxt_in_error(bp);
2092 	if (rc)
2093 		return rc;
2094 
2095 	if (!vnic)
2096 		return -EINVAL;
2097 	if (!vnic->rss_table)
2098 		return -EINVAL;
2099 
2100 	if (reta_size != tbl_size) {
2101 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2102 			"(%d) must equal the size supported by the hardware "
2103 			"(%d)\n", reta_size, tbl_size);
2104 		return -EINVAL;
2105 	}
2106 
2107 	for (idx = 0, i = 0; i < reta_size; i++) {
2108 		idx = i / RTE_RETA_GROUP_SIZE;
2109 		sft = i % RTE_RETA_GROUP_SIZE;
2110 
2111 		if (reta_conf[idx].mask & (1ULL << sft)) {
2112 			uint16_t qid;
2113 
2114 			if (BNXT_CHIP_P5(bp))
2115 				qid = bnxt_rss_to_qid(bp,
2116 						      vnic->rss_table[i * 2]);
2117 			else
2118 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2119 
2120 			if (qid == INVALID_HW_RING_ID) {
2121 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2122 				return -EINVAL;
2123 			}
2124 			reta_conf[idx].reta[sft] = qid;
2125 		}
2126 	}
2127 
2128 	return 0;
2129 }
2130 
2131 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2132 				   struct rte_eth_rss_conf *rss_conf)
2133 {
2134 	struct bnxt *bp = eth_dev->data->dev_private;
2135 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2136 	struct bnxt_vnic_info *vnic;
2137 	int rc;
2138 
2139 	rc = is_bnxt_in_error(bp);
2140 	if (rc)
2141 		return rc;
2142 
2143 	/*
2144 	 * If RSS enablement were different than dev_configure,
2145 	 * then return -EINVAL
2146 	 */
2147 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
2148 		if (!rss_conf->rss_hf)
2149 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2150 	} else {
2151 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2152 			return -EINVAL;
2153 	}
2154 
2155 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
2156 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2157 	       rss_conf,
2158 	       sizeof(*rss_conf));
2159 
2160 	/* Update the default RSS VNIC(s) */
2161 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2162 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2163 	vnic->hash_mode =
2164 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2165 					    ETH_RSS_LEVEL(rss_conf->rss_hf));
2166 
2167 	/*
2168 	 * If hashkey is not specified, use the previously configured
2169 	 * hashkey
2170 	 */
2171 	if (!rss_conf->rss_key)
2172 		goto rss_config;
2173 
2174 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2175 		PMD_DRV_LOG(ERR,
2176 			    "Invalid hashkey length, should be %d bytes\n",
2177 			    HW_HASH_KEY_SIZE);
2178 		return -EINVAL;
2179 	}
2180 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2181 
2182 rss_config:
2183 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2184 	return rc;
2185 }
2186 
2187 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2188 				     struct rte_eth_rss_conf *rss_conf)
2189 {
2190 	struct bnxt *bp = eth_dev->data->dev_private;
2191 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2192 	int len, rc;
2193 	uint32_t hash_types;
2194 
2195 	rc = is_bnxt_in_error(bp);
2196 	if (rc)
2197 		return rc;
2198 
2199 	/* RSS configuration is the same for all VNICs */
2200 	if (vnic && vnic->rss_hash_key) {
2201 		if (rss_conf->rss_key) {
2202 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2203 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2204 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2205 		}
2206 
2207 		hash_types = vnic->hash_type;
2208 		rss_conf->rss_hf = 0;
2209 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2210 			rss_conf->rss_hf |= ETH_RSS_IPV4;
2211 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2212 		}
2213 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2214 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2215 			hash_types &=
2216 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2217 		}
2218 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2219 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2220 			hash_types &=
2221 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2222 		}
2223 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2224 			rss_conf->rss_hf |= ETH_RSS_IPV6;
2225 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2226 		}
2227 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2228 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2229 			hash_types &=
2230 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2231 		}
2232 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2233 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2234 			hash_types &=
2235 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2236 		}
2237 
2238 		rss_conf->rss_hf |=
2239 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2240 
2241 		if (hash_types) {
2242 			PMD_DRV_LOG(ERR,
2243 				"Unknown RSS config from firmware (%08x), RSS disabled",
2244 				vnic->hash_type);
2245 			return -ENOTSUP;
2246 		}
2247 	} else {
2248 		rss_conf->rss_hf = 0;
2249 	}
2250 	return 0;
2251 }
2252 
2253 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2254 			       struct rte_eth_fc_conf *fc_conf)
2255 {
2256 	struct bnxt *bp = dev->data->dev_private;
2257 	struct rte_eth_link link_info;
2258 	int rc;
2259 
2260 	rc = is_bnxt_in_error(bp);
2261 	if (rc)
2262 		return rc;
2263 
2264 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2265 	if (rc)
2266 		return rc;
2267 
2268 	memset(fc_conf, 0, sizeof(*fc_conf));
2269 	if (bp->link_info->auto_pause)
2270 		fc_conf->autoneg = 1;
2271 	switch (bp->link_info->pause) {
2272 	case 0:
2273 		fc_conf->mode = RTE_FC_NONE;
2274 		break;
2275 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2276 		fc_conf->mode = RTE_FC_TX_PAUSE;
2277 		break;
2278 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2279 		fc_conf->mode = RTE_FC_RX_PAUSE;
2280 		break;
2281 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2282 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2283 		fc_conf->mode = RTE_FC_FULL;
2284 		break;
2285 	}
2286 	return 0;
2287 }
2288 
2289 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2290 			       struct rte_eth_fc_conf *fc_conf)
2291 {
2292 	struct bnxt *bp = dev->data->dev_private;
2293 	int rc;
2294 
2295 	rc = is_bnxt_in_error(bp);
2296 	if (rc)
2297 		return rc;
2298 
2299 	if (!BNXT_SINGLE_PF(bp)) {
2300 		PMD_DRV_LOG(ERR,
2301 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2302 		return -ENOTSUP;
2303 	}
2304 
2305 	switch (fc_conf->mode) {
2306 	case RTE_FC_NONE:
2307 		bp->link_info->auto_pause = 0;
2308 		bp->link_info->force_pause = 0;
2309 		break;
2310 	case RTE_FC_RX_PAUSE:
2311 		if (fc_conf->autoneg) {
2312 			bp->link_info->auto_pause =
2313 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2314 			bp->link_info->force_pause = 0;
2315 		} else {
2316 			bp->link_info->auto_pause = 0;
2317 			bp->link_info->force_pause =
2318 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2319 		}
2320 		break;
2321 	case RTE_FC_TX_PAUSE:
2322 		if (fc_conf->autoneg) {
2323 			bp->link_info->auto_pause =
2324 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2325 			bp->link_info->force_pause = 0;
2326 		} else {
2327 			bp->link_info->auto_pause = 0;
2328 			bp->link_info->force_pause =
2329 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2330 		}
2331 		break;
2332 	case RTE_FC_FULL:
2333 		if (fc_conf->autoneg) {
2334 			bp->link_info->auto_pause =
2335 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2336 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2337 			bp->link_info->force_pause = 0;
2338 		} else {
2339 			bp->link_info->auto_pause = 0;
2340 			bp->link_info->force_pause =
2341 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2342 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2343 		}
2344 		break;
2345 	}
2346 	return bnxt_set_hwrm_link_config(bp, true);
2347 }
2348 
2349 /* Add UDP tunneling port */
2350 static int
2351 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2352 			 struct rte_eth_udp_tunnel *udp_tunnel)
2353 {
2354 	struct bnxt *bp = eth_dev->data->dev_private;
2355 	uint16_t tunnel_type = 0;
2356 	int rc = 0;
2357 
2358 	rc = is_bnxt_in_error(bp);
2359 	if (rc)
2360 		return rc;
2361 
2362 	switch (udp_tunnel->prot_type) {
2363 	case RTE_TUNNEL_TYPE_VXLAN:
2364 		if (bp->vxlan_port_cnt) {
2365 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2366 				udp_tunnel->udp_port);
2367 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2368 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2369 				return -ENOSPC;
2370 			}
2371 			bp->vxlan_port_cnt++;
2372 			return 0;
2373 		}
2374 		tunnel_type =
2375 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2376 		break;
2377 	case RTE_TUNNEL_TYPE_GENEVE:
2378 		if (bp->geneve_port_cnt) {
2379 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2380 				udp_tunnel->udp_port);
2381 			if (bp->geneve_port != udp_tunnel->udp_port) {
2382 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2383 				return -ENOSPC;
2384 			}
2385 			bp->geneve_port_cnt++;
2386 			return 0;
2387 		}
2388 		tunnel_type =
2389 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2390 		break;
2391 	default:
2392 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2393 		return -ENOTSUP;
2394 	}
2395 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2396 					     tunnel_type);
2397 
2398 	if (rc != 0)
2399 		return rc;
2400 
2401 	if (tunnel_type ==
2402 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN)
2403 		bp->vxlan_port_cnt++;
2404 
2405 	if (tunnel_type ==
2406 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE)
2407 		bp->geneve_port_cnt++;
2408 
2409 	return rc;
2410 }
2411 
2412 static int
2413 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2414 			 struct rte_eth_udp_tunnel *udp_tunnel)
2415 {
2416 	struct bnxt *bp = eth_dev->data->dev_private;
2417 	uint16_t tunnel_type = 0;
2418 	uint16_t port = 0;
2419 	int rc = 0;
2420 
2421 	rc = is_bnxt_in_error(bp);
2422 	if (rc)
2423 		return rc;
2424 
2425 	switch (udp_tunnel->prot_type) {
2426 	case RTE_TUNNEL_TYPE_VXLAN:
2427 		if (!bp->vxlan_port_cnt) {
2428 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2429 			return -EINVAL;
2430 		}
2431 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2432 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2433 				udp_tunnel->udp_port, bp->vxlan_port);
2434 			return -EINVAL;
2435 		}
2436 		if (--bp->vxlan_port_cnt)
2437 			return 0;
2438 
2439 		tunnel_type =
2440 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2441 		port = bp->vxlan_fw_dst_port_id;
2442 		break;
2443 	case RTE_TUNNEL_TYPE_GENEVE:
2444 		if (!bp->geneve_port_cnt) {
2445 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2446 			return -EINVAL;
2447 		}
2448 		if (bp->geneve_port != udp_tunnel->udp_port) {
2449 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2450 				udp_tunnel->udp_port, bp->geneve_port);
2451 			return -EINVAL;
2452 		}
2453 		if (--bp->geneve_port_cnt)
2454 			return 0;
2455 
2456 		tunnel_type =
2457 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2458 		port = bp->geneve_fw_dst_port_id;
2459 		break;
2460 	default:
2461 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2462 		return -ENOTSUP;
2463 	}
2464 
2465 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2466 	return rc;
2467 }
2468 
2469 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2470 {
2471 	struct bnxt_filter_info *filter;
2472 	struct bnxt_vnic_info *vnic;
2473 	int rc = 0;
2474 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2475 
2476 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2477 	filter = STAILQ_FIRST(&vnic->filter);
2478 	while (filter) {
2479 		/* Search for this matching MAC+VLAN filter */
2480 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2481 			/* Delete the filter */
2482 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2483 			if (rc)
2484 				return rc;
2485 			STAILQ_REMOVE(&vnic->filter, filter,
2486 				      bnxt_filter_info, next);
2487 			bnxt_free_filter(bp, filter);
2488 			PMD_DRV_LOG(INFO,
2489 				    "Deleted vlan filter for %d\n",
2490 				    vlan_id);
2491 			return 0;
2492 		}
2493 		filter = STAILQ_NEXT(filter, next);
2494 	}
2495 	return -ENOENT;
2496 }
2497 
2498 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2499 {
2500 	struct bnxt_filter_info *filter;
2501 	struct bnxt_vnic_info *vnic;
2502 	int rc = 0;
2503 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2504 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2505 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2506 
2507 	/* Implementation notes on the use of VNIC in this command:
2508 	 *
2509 	 * By default, these filters belong to default vnic for the function.
2510 	 * Once these filters are set up, only destination VNIC can be modified.
2511 	 * If the destination VNIC is not specified in this command,
2512 	 * then the HWRM shall only create an l2 context id.
2513 	 */
2514 
2515 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2516 	filter = STAILQ_FIRST(&vnic->filter);
2517 	/* Check if the VLAN has already been added */
2518 	while (filter) {
2519 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2520 			return -EEXIST;
2521 
2522 		filter = STAILQ_NEXT(filter, next);
2523 	}
2524 
2525 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2526 	 * command to create MAC+VLAN filter with the right flags, enables set.
2527 	 */
2528 	filter = bnxt_alloc_filter(bp);
2529 	if (!filter) {
2530 		PMD_DRV_LOG(ERR,
2531 			    "MAC/VLAN filter alloc failed\n");
2532 		return -ENOMEM;
2533 	}
2534 	/* MAC + VLAN ID filter */
2535 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2536 	 * untagged packets are received
2537 	 *
2538 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2539 	 * packets and only the programmed vlan's packets are received
2540 	 */
2541 	filter->l2_ivlan = vlan_id;
2542 	filter->l2_ivlan_mask = 0x0FFF;
2543 	filter->enables |= en;
2544 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2545 
2546 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2547 	if (rc) {
2548 		/* Free the newly allocated filter as we were
2549 		 * not able to create the filter in hardware.
2550 		 */
2551 		bnxt_free_filter(bp, filter);
2552 		return rc;
2553 	}
2554 
2555 	filter->mac_index = 0;
2556 	/* Add this new filter to the list */
2557 	if (vlan_id == 0)
2558 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2559 	else
2560 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2561 
2562 	PMD_DRV_LOG(INFO,
2563 		    "Added Vlan filter for %d\n", vlan_id);
2564 	return rc;
2565 }
2566 
2567 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2568 		uint16_t vlan_id, int on)
2569 {
2570 	struct bnxt *bp = eth_dev->data->dev_private;
2571 	int rc;
2572 
2573 	rc = is_bnxt_in_error(bp);
2574 	if (rc)
2575 		return rc;
2576 
2577 	if (!eth_dev->data->dev_started) {
2578 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2579 		return -EINVAL;
2580 	}
2581 
2582 	/* These operations apply to ALL existing MAC/VLAN filters */
2583 	if (on)
2584 		return bnxt_add_vlan_filter(bp, vlan_id);
2585 	else
2586 		return bnxt_del_vlan_filter(bp, vlan_id);
2587 }
2588 
2589 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2590 				    struct bnxt_vnic_info *vnic)
2591 {
2592 	struct bnxt_filter_info *filter;
2593 	int rc;
2594 
2595 	filter = STAILQ_FIRST(&vnic->filter);
2596 	while (filter) {
2597 		if (filter->mac_index == 0 &&
2598 		    !memcmp(filter->l2_addr, bp->mac_addr,
2599 			    RTE_ETHER_ADDR_LEN)) {
2600 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2601 			if (!rc) {
2602 				STAILQ_REMOVE(&vnic->filter, filter,
2603 					      bnxt_filter_info, next);
2604 				bnxt_free_filter(bp, filter);
2605 			}
2606 			return rc;
2607 		}
2608 		filter = STAILQ_NEXT(filter, next);
2609 	}
2610 	return 0;
2611 }
2612 
2613 static int
2614 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2615 {
2616 	struct bnxt_vnic_info *vnic;
2617 	unsigned int i;
2618 	int rc;
2619 
2620 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2621 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2622 		/* Remove any VLAN filters programmed */
2623 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2624 			bnxt_del_vlan_filter(bp, i);
2625 
2626 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2627 		if (rc)
2628 			return rc;
2629 	} else {
2630 		/* Default filter will allow packets that match the
2631 		 * dest mac. So, it has to be deleted, otherwise, we
2632 		 * will endup receiving vlan packets for which the
2633 		 * filter is not programmed, when hw-vlan-filter
2634 		 * configuration is ON
2635 		 */
2636 		bnxt_del_dflt_mac_filter(bp, vnic);
2637 		/* This filter will allow only untagged packets */
2638 		bnxt_add_vlan_filter(bp, 0);
2639 	}
2640 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2641 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2642 
2643 	return 0;
2644 }
2645 
2646 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2647 {
2648 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2649 	unsigned int i;
2650 	int rc;
2651 
2652 	/* Destroy vnic filters and vnic */
2653 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2654 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2655 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2656 			bnxt_del_vlan_filter(bp, i);
2657 	}
2658 	bnxt_del_dflt_mac_filter(bp, vnic);
2659 
2660 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2661 	if (rc)
2662 		return rc;
2663 
2664 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2665 	if (rc)
2666 		return rc;
2667 
2668 	rte_free(vnic->fw_grp_ids);
2669 	vnic->fw_grp_ids = NULL;
2670 
2671 	vnic->rx_queue_cnt = 0;
2672 
2673 	return 0;
2674 }
2675 
2676 static int
2677 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2678 {
2679 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2680 	int rc;
2681 
2682 	/* Destroy, recreate and reconfigure the default vnic */
2683 	rc = bnxt_free_one_vnic(bp, 0);
2684 	if (rc)
2685 		return rc;
2686 
2687 	/* default vnic 0 */
2688 	rc = bnxt_setup_one_vnic(bp, 0);
2689 	if (rc)
2690 		return rc;
2691 
2692 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2693 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2694 		rc = bnxt_add_vlan_filter(bp, 0);
2695 		if (rc)
2696 			return rc;
2697 		rc = bnxt_restore_vlan_filters(bp);
2698 		if (rc)
2699 			return rc;
2700 	} else {
2701 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2702 		if (rc)
2703 			return rc;
2704 	}
2705 
2706 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2707 	if (rc)
2708 		return rc;
2709 
2710 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2711 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2712 
2713 	return rc;
2714 }
2715 
2716 static int
2717 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2718 {
2719 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2720 	struct bnxt *bp = dev->data->dev_private;
2721 	int rc;
2722 
2723 	rc = is_bnxt_in_error(bp);
2724 	if (rc)
2725 		return rc;
2726 
2727 	/* Filter settings will get applied when port is started */
2728 	if (!dev->data->dev_started)
2729 		return 0;
2730 
2731 	if (mask & ETH_VLAN_FILTER_MASK) {
2732 		/* Enable or disable VLAN filtering */
2733 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2734 		if (rc)
2735 			return rc;
2736 	}
2737 
2738 	if (mask & ETH_VLAN_STRIP_MASK) {
2739 		/* Enable or disable VLAN stripping */
2740 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2741 		if (rc)
2742 			return rc;
2743 	}
2744 
2745 	if (mask & ETH_VLAN_EXTEND_MASK) {
2746 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2747 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2748 		else
2749 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2750 	}
2751 
2752 	return 0;
2753 }
2754 
2755 static int
2756 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2757 		      uint16_t tpid)
2758 {
2759 	struct bnxt *bp = dev->data->dev_private;
2760 	int qinq = dev->data->dev_conf.rxmode.offloads &
2761 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2762 
2763 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2764 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2765 		PMD_DRV_LOG(ERR,
2766 			    "Unsupported vlan type.");
2767 		return -EINVAL;
2768 	}
2769 	if (!qinq) {
2770 		PMD_DRV_LOG(ERR,
2771 			    "QinQ not enabled. Needs to be ON as we can "
2772 			    "accelerate only outer vlan\n");
2773 		return -EINVAL;
2774 	}
2775 
2776 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2777 		switch (tpid) {
2778 		case RTE_ETHER_TYPE_QINQ:
2779 			bp->outer_tpid_bd =
2780 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2781 				break;
2782 		case RTE_ETHER_TYPE_VLAN:
2783 			bp->outer_tpid_bd =
2784 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2785 				break;
2786 		case RTE_ETHER_TYPE_QINQ1:
2787 			bp->outer_tpid_bd =
2788 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2789 				break;
2790 		case RTE_ETHER_TYPE_QINQ2:
2791 			bp->outer_tpid_bd =
2792 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2793 				break;
2794 		case RTE_ETHER_TYPE_QINQ3:
2795 			bp->outer_tpid_bd =
2796 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2797 				break;
2798 		default:
2799 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2800 			return -EINVAL;
2801 		}
2802 		bp->outer_tpid_bd |= tpid;
2803 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2804 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2805 		PMD_DRV_LOG(ERR,
2806 			    "Can accelerate only outer vlan in QinQ\n");
2807 		return -EINVAL;
2808 	}
2809 
2810 	return 0;
2811 }
2812 
2813 static int
2814 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2815 			     struct rte_ether_addr *addr)
2816 {
2817 	struct bnxt *bp = dev->data->dev_private;
2818 	/* Default Filter is tied to VNIC 0 */
2819 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2820 	int rc;
2821 
2822 	rc = is_bnxt_in_error(bp);
2823 	if (rc)
2824 		return rc;
2825 
2826 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2827 		return -EPERM;
2828 
2829 	if (rte_is_zero_ether_addr(addr))
2830 		return -EINVAL;
2831 
2832 	/* Filter settings will get applied when port is started */
2833 	if (!dev->data->dev_started)
2834 		return 0;
2835 
2836 	/* Check if the requested MAC is already added */
2837 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2838 		return 0;
2839 
2840 	/* Destroy filter and re-create it */
2841 	bnxt_del_dflt_mac_filter(bp, vnic);
2842 
2843 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2844 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2845 		/* This filter will allow only untagged packets */
2846 		rc = bnxt_add_vlan_filter(bp, 0);
2847 	} else {
2848 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2849 	}
2850 
2851 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2852 	return rc;
2853 }
2854 
2855 static int
2856 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2857 			  struct rte_ether_addr *mc_addr_set,
2858 			  uint32_t nb_mc_addr)
2859 {
2860 	struct bnxt *bp = eth_dev->data->dev_private;
2861 	char *mc_addr_list = (char *)mc_addr_set;
2862 	struct bnxt_vnic_info *vnic;
2863 	uint32_t off = 0, i = 0;
2864 	int rc;
2865 
2866 	rc = is_bnxt_in_error(bp);
2867 	if (rc)
2868 		return rc;
2869 
2870 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2871 
2872 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2873 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2874 		goto allmulti;
2875 	}
2876 
2877 	/* TODO Check for Duplicate mcast addresses */
2878 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2879 	for (i = 0; i < nb_mc_addr; i++) {
2880 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2881 			RTE_ETHER_ADDR_LEN);
2882 		off += RTE_ETHER_ADDR_LEN;
2883 	}
2884 
2885 	vnic->mc_addr_cnt = i;
2886 	if (vnic->mc_addr_cnt)
2887 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2888 	else
2889 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2890 
2891 allmulti:
2892 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2893 }
2894 
2895 static int
2896 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2897 {
2898 	struct bnxt *bp = dev->data->dev_private;
2899 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2900 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2901 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2902 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2903 	int ret;
2904 
2905 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2906 			fw_major, fw_minor, fw_updt, fw_rsvd);
2907 	if (ret < 0)
2908 		return -EINVAL;
2909 
2910 	ret += 1; /* add the size of '\0' */
2911 	if (fw_size < (size_t)ret)
2912 		return ret;
2913 	else
2914 		return 0;
2915 }
2916 
2917 static void
2918 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2919 	struct rte_eth_rxq_info *qinfo)
2920 {
2921 	struct bnxt *bp = dev->data->dev_private;
2922 	struct bnxt_rx_queue *rxq;
2923 
2924 	if (is_bnxt_in_error(bp))
2925 		return;
2926 
2927 	rxq = dev->data->rx_queues[queue_id];
2928 
2929 	qinfo->mp = rxq->mb_pool;
2930 	qinfo->scattered_rx = dev->data->scattered_rx;
2931 	qinfo->nb_desc = rxq->nb_rx_desc;
2932 
2933 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2934 	qinfo->conf.rx_drop_en = rxq->drop_en;
2935 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2936 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2937 }
2938 
2939 static void
2940 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2941 	struct rte_eth_txq_info *qinfo)
2942 {
2943 	struct bnxt *bp = dev->data->dev_private;
2944 	struct bnxt_tx_queue *txq;
2945 
2946 	if (is_bnxt_in_error(bp))
2947 		return;
2948 
2949 	txq = dev->data->tx_queues[queue_id];
2950 
2951 	qinfo->nb_desc = txq->nb_tx_desc;
2952 
2953 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2954 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2955 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2956 
2957 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2958 	qinfo->conf.tx_rs_thresh = 0;
2959 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2960 	qinfo->conf.offloads = txq->offloads;
2961 }
2962 
2963 static const struct {
2964 	eth_rx_burst_t pkt_burst;
2965 	const char *info;
2966 } bnxt_rx_burst_info[] = {
2967 	{bnxt_recv_pkts,		"Scalar"},
2968 #if defined(RTE_ARCH_X86)
2969 	{bnxt_recv_pkts_vec,		"Vector SSE"},
2970 #endif
2971 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2972 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
2973 #endif
2974 #if defined(RTE_ARCH_ARM64)
2975 	{bnxt_recv_pkts_vec,		"Vector Neon"},
2976 #endif
2977 };
2978 
2979 static int
2980 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2981 		       struct rte_eth_burst_mode *mode)
2982 {
2983 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2984 	size_t i;
2985 
2986 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2987 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2988 			snprintf(mode->info, sizeof(mode->info), "%s",
2989 				 bnxt_rx_burst_info[i].info);
2990 			return 0;
2991 		}
2992 	}
2993 
2994 	return -EINVAL;
2995 }
2996 
2997 static const struct {
2998 	eth_tx_burst_t pkt_burst;
2999 	const char *info;
3000 } bnxt_tx_burst_info[] = {
3001 	{bnxt_xmit_pkts,		"Scalar"},
3002 #if defined(RTE_ARCH_X86)
3003 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
3004 #endif
3005 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
3006 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
3007 #endif
3008 #if defined(RTE_ARCH_ARM64)
3009 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
3010 #endif
3011 };
3012 
3013 static int
3014 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
3015 		       struct rte_eth_burst_mode *mode)
3016 {
3017 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
3018 	size_t i;
3019 
3020 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
3021 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
3022 			snprintf(mode->info, sizeof(mode->info), "%s",
3023 				 bnxt_tx_burst_info[i].info);
3024 			return 0;
3025 		}
3026 	}
3027 
3028 	return -EINVAL;
3029 }
3030 
3031 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3032 {
3033 	struct bnxt *bp = eth_dev->data->dev_private;
3034 	uint32_t new_pkt_size;
3035 	uint32_t rc = 0;
3036 	uint32_t i;
3037 
3038 	rc = is_bnxt_in_error(bp);
3039 	if (rc)
3040 		return rc;
3041 
3042 	/* Exit if receive queues are not configured yet */
3043 	if (!eth_dev->data->nb_rx_queues)
3044 		return rc;
3045 
3046 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
3047 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
3048 
3049 	/*
3050 	 * Disallow any MTU change that would require scattered receive support
3051 	 * if it is not already enabled.
3052 	 */
3053 	if (eth_dev->data->dev_started &&
3054 	    !eth_dev->data->scattered_rx &&
3055 	    (new_pkt_size >
3056 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
3057 		PMD_DRV_LOG(ERR,
3058 			    "MTU change would require scattered rx support. ");
3059 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
3060 		return -EINVAL;
3061 	}
3062 
3063 	if (new_mtu > RTE_ETHER_MTU) {
3064 		bp->flags |= BNXT_FLAG_JUMBO;
3065 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
3066 			DEV_RX_OFFLOAD_JUMBO_FRAME;
3067 	} else {
3068 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
3069 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
3070 		bp->flags &= ~BNXT_FLAG_JUMBO;
3071 	}
3072 
3073 	/* Is there a change in mtu setting? */
3074 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
3075 		return rc;
3076 
3077 	for (i = 0; i < bp->nr_vnics; i++) {
3078 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3079 		uint16_t size = 0;
3080 
3081 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
3082 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
3083 		if (rc)
3084 			break;
3085 
3086 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
3087 		size -= RTE_PKTMBUF_HEADROOM;
3088 
3089 		if (size < new_mtu) {
3090 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
3091 			if (rc)
3092 				return rc;
3093 		}
3094 	}
3095 
3096 	if (!rc)
3097 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
3098 
3099 	if (bnxt_hwrm_config_host_mtu(bp))
3100 		PMD_DRV_LOG(WARNING, "Failed to configure host MTU\n");
3101 
3102 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3103 
3104 	return rc;
3105 }
3106 
3107 static int
3108 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3109 {
3110 	struct bnxt *bp = dev->data->dev_private;
3111 	uint16_t vlan = bp->vlan;
3112 	int rc;
3113 
3114 	rc = is_bnxt_in_error(bp);
3115 	if (rc)
3116 		return rc;
3117 
3118 	if (!BNXT_SINGLE_PF(bp)) {
3119 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3120 		return -ENOTSUP;
3121 	}
3122 	bp->vlan = on ? pvid : 0;
3123 
3124 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3125 	if (rc)
3126 		bp->vlan = vlan;
3127 	return rc;
3128 }
3129 
3130 static int
3131 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3132 {
3133 	struct bnxt *bp = dev->data->dev_private;
3134 	int rc;
3135 
3136 	rc = is_bnxt_in_error(bp);
3137 	if (rc)
3138 		return rc;
3139 
3140 	return bnxt_hwrm_port_led_cfg(bp, true);
3141 }
3142 
3143 static int
3144 bnxt_dev_led_off_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, false);
3154 }
3155 
3156 static uint32_t
3157 bnxt_rx_queue_count_op(void *rx_queue)
3158 {
3159 	struct bnxt *bp;
3160 	struct bnxt_cp_ring_info *cpr;
3161 	uint32_t desc = 0, raw_cons, cp_ring_size;
3162 	struct bnxt_rx_queue *rxq;
3163 	struct rx_pkt_cmpl *rxcmp;
3164 	int rc;
3165 
3166 	rxq = rx_queue;
3167 	bp = rxq->bp;
3168 
3169 	rc = is_bnxt_in_error(bp);
3170 	if (rc)
3171 		return rc;
3172 
3173 	cpr = rxq->cp_ring;
3174 	raw_cons = cpr->cp_raw_cons;
3175 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3176 
3177 	while (1) {
3178 		uint32_t agg_cnt, cons, cmpl_type;
3179 
3180 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3181 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3182 
3183 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3184 			break;
3185 
3186 		cmpl_type = CMP_TYPE(rxcmp);
3187 
3188 		switch (cmpl_type) {
3189 		case CMPL_BASE_TYPE_RX_L2:
3190 		case CMPL_BASE_TYPE_RX_L2_V2:
3191 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3192 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3193 			desc++;
3194 			break;
3195 
3196 		case CMPL_BASE_TYPE_RX_TPA_END:
3197 			if (BNXT_CHIP_P5(rxq->bp)) {
3198 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3199 
3200 				p5_tpa_end = (void *)rxcmp;
3201 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3202 			} else {
3203 				struct rx_tpa_end_cmpl *tpa_end;
3204 
3205 				tpa_end = (void *)rxcmp;
3206 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3207 			}
3208 
3209 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3210 			desc++;
3211 			break;
3212 
3213 		default:
3214 			raw_cons += CMP_LEN(cmpl_type);
3215 		}
3216 	}
3217 
3218 	return desc;
3219 }
3220 
3221 static int
3222 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3223 {
3224 	struct bnxt_rx_queue *rxq = rx_queue;
3225 	struct bnxt_cp_ring_info *cpr;
3226 	struct bnxt_rx_ring_info *rxr;
3227 	uint32_t desc, raw_cons, cp_ring_size;
3228 	struct bnxt *bp = rxq->bp;
3229 	struct rx_pkt_cmpl *rxcmp;
3230 	int rc;
3231 
3232 	rc = is_bnxt_in_error(bp);
3233 	if (rc)
3234 		return rc;
3235 
3236 	if (offset >= rxq->nb_rx_desc)
3237 		return -EINVAL;
3238 
3239 	rxr = rxq->rx_ring;
3240 	cpr = rxq->cp_ring;
3241 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3242 
3243 	/*
3244 	 * For the vector receive case, the completion at the requested
3245 	 * offset can be indexed directly.
3246 	 */
3247 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3248 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3249 		struct rx_pkt_cmpl *rxcmp;
3250 		uint32_t cons;
3251 
3252 		/* Check status of completion descriptor. */
3253 		raw_cons = cpr->cp_raw_cons +
3254 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3255 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3256 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3257 
3258 		if (bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3259 			return RTE_ETH_RX_DESC_DONE;
3260 
3261 		/* Check whether rx desc has an mbuf attached. */
3262 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3263 		if (cons >= rxq->rxrearm_start &&
3264 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3265 			return RTE_ETH_RX_DESC_UNAVAIL;
3266 		}
3267 
3268 		return RTE_ETH_RX_DESC_AVAIL;
3269 	}
3270 #endif
3271 
3272 	/*
3273 	 * For the non-vector receive case, scan the completion ring to
3274 	 * locate the completion descriptor for the requested offset.
3275 	 */
3276 	raw_cons = cpr->cp_raw_cons;
3277 	desc = 0;
3278 	while (1) {
3279 		uint32_t agg_cnt, cons, cmpl_type;
3280 
3281 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3282 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3283 
3284 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3285 			break;
3286 
3287 		cmpl_type = CMP_TYPE(rxcmp);
3288 
3289 		switch (cmpl_type) {
3290 		case CMPL_BASE_TYPE_RX_L2:
3291 		case CMPL_BASE_TYPE_RX_L2_V2:
3292 			if (desc == offset) {
3293 				cons = rxcmp->opaque;
3294 				if (rxr->rx_buf_ring[cons])
3295 					return RTE_ETH_RX_DESC_DONE;
3296 				else
3297 					return RTE_ETH_RX_DESC_UNAVAIL;
3298 			}
3299 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3300 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3301 			desc++;
3302 			break;
3303 
3304 		case CMPL_BASE_TYPE_RX_TPA_END:
3305 			if (desc == offset)
3306 				return RTE_ETH_RX_DESC_DONE;
3307 
3308 			if (BNXT_CHIP_P5(rxq->bp)) {
3309 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3310 
3311 				p5_tpa_end = (void *)rxcmp;
3312 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3313 			} else {
3314 				struct rx_tpa_end_cmpl *tpa_end;
3315 
3316 				tpa_end = (void *)rxcmp;
3317 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3318 			}
3319 
3320 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3321 			desc++;
3322 			break;
3323 
3324 		default:
3325 			raw_cons += CMP_LEN(cmpl_type);
3326 		}
3327 	}
3328 
3329 	return RTE_ETH_RX_DESC_AVAIL;
3330 }
3331 
3332 static int
3333 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3334 {
3335 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3336 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3337 	uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3338 	struct cmpl_base *cp_desc_ring;
3339 	int rc;
3340 
3341 	rc = is_bnxt_in_error(txq->bp);
3342 	if (rc)
3343 		return rc;
3344 
3345 	if (offset >= txq->nb_tx_desc)
3346 		return -EINVAL;
3347 
3348 	/* Return "desc done" if descriptor is available for use. */
3349 	if (bnxt_tx_bds_in_hw(txq) <= offset)
3350 		return RTE_ETH_TX_DESC_DONE;
3351 
3352 	raw_cons = cpr->cp_raw_cons;
3353 	cp_desc_ring = cpr->cp_desc_ring;
3354 	ring_mask = cpr->cp_ring_struct->ring_mask;
3355 
3356 	/* Check to see if hw has posted a completion for the descriptor. */
3357 	while (1) {
3358 		struct tx_cmpl *txcmp;
3359 		uint32_t cons;
3360 
3361 		cons = RING_CMPL(ring_mask, raw_cons);
3362 		txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3363 
3364 		if (!bnxt_cpr_cmp_valid(txcmp, raw_cons, ring_mask + 1))
3365 			break;
3366 
3367 		if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3368 			nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3369 
3370 		if (nb_tx_pkts > offset)
3371 			return RTE_ETH_TX_DESC_DONE;
3372 
3373 		raw_cons = NEXT_RAW_CMP(raw_cons);
3374 	}
3375 
3376 	/* Descriptor is pending transmit, not yet completed by hardware. */
3377 	return RTE_ETH_TX_DESC_FULL;
3378 }
3379 
3380 int
3381 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3382 		     const struct rte_flow_ops **ops)
3383 {
3384 	struct bnxt *bp = dev->data->dev_private;
3385 	int ret = 0;
3386 
3387 	if (!bp)
3388 		return -EIO;
3389 
3390 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3391 		struct bnxt_representor *vfr = dev->data->dev_private;
3392 		bp = vfr->parent_dev->data->dev_private;
3393 		/* parent is deleted while children are still valid */
3394 		if (!bp) {
3395 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3396 				    dev->data->port_id);
3397 			return -EIO;
3398 		}
3399 	}
3400 
3401 	ret = is_bnxt_in_error(bp);
3402 	if (ret)
3403 		return ret;
3404 
3405 	/* PMD supports thread-safe flow operations.  rte_flow API
3406 	 * functions can avoid mutex for multi-thread safety.
3407 	 */
3408 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3409 
3410 	if (BNXT_TRUFLOW_EN(bp))
3411 		*ops = &bnxt_ulp_rte_flow_ops;
3412 	else
3413 		*ops = &bnxt_flow_ops;
3414 
3415 	return ret;
3416 }
3417 
3418 static const uint32_t *
3419 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3420 {
3421 	static const uint32_t ptypes[] = {
3422 		RTE_PTYPE_L2_ETHER_VLAN,
3423 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3424 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3425 		RTE_PTYPE_L4_ICMP,
3426 		RTE_PTYPE_L4_TCP,
3427 		RTE_PTYPE_L4_UDP,
3428 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3429 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3430 		RTE_PTYPE_INNER_L4_ICMP,
3431 		RTE_PTYPE_INNER_L4_TCP,
3432 		RTE_PTYPE_INNER_L4_UDP,
3433 		RTE_PTYPE_UNKNOWN
3434 	};
3435 
3436 	if (!dev->rx_pkt_burst)
3437 		return NULL;
3438 
3439 	return ptypes;
3440 }
3441 
3442 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3443 			 int reg_win)
3444 {
3445 	uint32_t reg_base = *reg_arr & 0xfffff000;
3446 	uint32_t win_off;
3447 	int i;
3448 
3449 	for (i = 0; i < count; i++) {
3450 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3451 			return -ERANGE;
3452 	}
3453 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3454 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3455 	return 0;
3456 }
3457 
3458 static int bnxt_map_ptp_regs(struct bnxt *bp)
3459 {
3460 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3461 	uint32_t *reg_arr;
3462 	int rc, i;
3463 
3464 	reg_arr = ptp->rx_regs;
3465 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3466 	if (rc)
3467 		return rc;
3468 
3469 	reg_arr = ptp->tx_regs;
3470 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3471 	if (rc)
3472 		return rc;
3473 
3474 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3475 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3476 
3477 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3478 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3479 
3480 	return 0;
3481 }
3482 
3483 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3484 {
3485 	rte_write32(0, (uint8_t *)bp->bar0 +
3486 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3487 	rte_write32(0, (uint8_t *)bp->bar0 +
3488 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3489 }
3490 
3491 static uint64_t bnxt_cc_read(struct bnxt *bp)
3492 {
3493 	uint64_t ns;
3494 
3495 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3496 			      BNXT_GRCPF_REG_SYNC_TIME));
3497 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3498 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3499 	return ns;
3500 }
3501 
3502 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3503 {
3504 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3505 	uint32_t fifo;
3506 
3507 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3508 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3509 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3510 		return -EAGAIN;
3511 
3512 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3513 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3514 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3515 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3516 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3517 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3518 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3519 
3520 	return 0;
3521 }
3522 
3523 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3524 {
3525 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3526 	struct bnxt_pf_info *pf = bp->pf;
3527 	uint16_t port_id;
3528 	int i = 0;
3529 	uint32_t fifo;
3530 
3531 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3532 		return -EINVAL;
3533 
3534 	port_id = pf->port_id;
3535 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3536 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3537 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3538 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3539 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3540 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3541 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3542 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3543 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3544 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3545 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3546 		i++;
3547 	}
3548 
3549 	if (i >= BNXT_PTP_RX_PND_CNT)
3550 		return -EBUSY;
3551 
3552 	return 0;
3553 }
3554 
3555 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3556 {
3557 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3558 	struct bnxt_pf_info *pf = bp->pf;
3559 	uint16_t port_id;
3560 	uint32_t fifo;
3561 
3562 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3563 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3564 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3565 		return -EAGAIN;
3566 
3567 	port_id = pf->port_id;
3568 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3569 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3570 
3571 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3572 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3573 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3574 		return bnxt_clr_rx_ts(bp, ts);
3575 
3576 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3577 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3578 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3579 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3580 
3581 	return 0;
3582 }
3583 
3584 static int
3585 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3586 {
3587 	uint64_t ns;
3588 	struct bnxt *bp = dev->data->dev_private;
3589 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3590 
3591 	if (!ptp)
3592 		return -ENOTSUP;
3593 
3594 	ns = rte_timespec_to_ns(ts);
3595 	/* Set the timecounters to a new value. */
3596 	ptp->tc.nsec = ns;
3597 	ptp->tx_tstamp_tc.nsec = ns;
3598 	ptp->rx_tstamp_tc.nsec = ns;
3599 
3600 	return 0;
3601 }
3602 
3603 static int
3604 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3605 {
3606 	struct bnxt *bp = dev->data->dev_private;
3607 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3608 	uint64_t ns, systime_cycles = 0;
3609 	int rc = 0;
3610 
3611 	if (!ptp)
3612 		return -ENOTSUP;
3613 
3614 	if (BNXT_CHIP_P5(bp))
3615 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3616 					     &systime_cycles);
3617 	else
3618 		systime_cycles = bnxt_cc_read(bp);
3619 
3620 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3621 	*ts = rte_ns_to_timespec(ns);
3622 
3623 	return rc;
3624 }
3625 static int
3626 bnxt_timesync_enable(struct rte_eth_dev *dev)
3627 {
3628 	struct bnxt *bp = dev->data->dev_private;
3629 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3630 	uint32_t shift = 0;
3631 	int rc;
3632 
3633 	if (!ptp)
3634 		return -ENOTSUP;
3635 
3636 	ptp->rx_filter = 1;
3637 	ptp->tx_tstamp_en = 1;
3638 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3639 
3640 	rc = bnxt_hwrm_ptp_cfg(bp);
3641 	if (rc)
3642 		return rc;
3643 
3644 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3645 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3646 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3647 
3648 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3649 	ptp->tc.cc_shift = shift;
3650 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3651 
3652 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3653 	ptp->rx_tstamp_tc.cc_shift = shift;
3654 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3655 
3656 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3657 	ptp->tx_tstamp_tc.cc_shift = shift;
3658 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3659 
3660 	if (!BNXT_CHIP_P5(bp))
3661 		bnxt_map_ptp_regs(bp);
3662 	else
3663 		rc = bnxt_ptp_start(bp);
3664 
3665 	return rc;
3666 }
3667 
3668 static int
3669 bnxt_timesync_disable(struct rte_eth_dev *dev)
3670 {
3671 	struct bnxt *bp = dev->data->dev_private;
3672 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3673 
3674 	if (!ptp)
3675 		return -ENOTSUP;
3676 
3677 	ptp->rx_filter = 0;
3678 	ptp->tx_tstamp_en = 0;
3679 	ptp->rxctl = 0;
3680 
3681 	bnxt_hwrm_ptp_cfg(bp);
3682 
3683 	if (!BNXT_CHIP_P5(bp))
3684 		bnxt_unmap_ptp_regs(bp);
3685 	else
3686 		bnxt_ptp_stop(bp);
3687 
3688 	return 0;
3689 }
3690 
3691 static int
3692 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3693 				 struct timespec *timestamp,
3694 				 uint32_t flags __rte_unused)
3695 {
3696 	struct bnxt *bp = dev->data->dev_private;
3697 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3698 	uint64_t rx_tstamp_cycles = 0;
3699 	uint64_t ns;
3700 
3701 	if (!ptp)
3702 		return -ENOTSUP;
3703 
3704 	if (BNXT_CHIP_P5(bp))
3705 		rx_tstamp_cycles = ptp->rx_timestamp;
3706 	else
3707 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3708 
3709 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3710 	*timestamp = rte_ns_to_timespec(ns);
3711 	return  0;
3712 }
3713 
3714 static int
3715 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3716 				 struct timespec *timestamp)
3717 {
3718 	struct bnxt *bp = dev->data->dev_private;
3719 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3720 	uint64_t tx_tstamp_cycles = 0;
3721 	uint64_t ns;
3722 	int rc = 0;
3723 
3724 	if (!ptp)
3725 		return -ENOTSUP;
3726 
3727 	if (BNXT_CHIP_P5(bp))
3728 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3729 					     &tx_tstamp_cycles);
3730 	else
3731 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3732 
3733 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3734 	*timestamp = rte_ns_to_timespec(ns);
3735 
3736 	return rc;
3737 }
3738 
3739 static int
3740 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3741 {
3742 	struct bnxt *bp = dev->data->dev_private;
3743 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3744 
3745 	if (!ptp)
3746 		return -ENOTSUP;
3747 
3748 	ptp->tc.nsec += delta;
3749 	ptp->tx_tstamp_tc.nsec += delta;
3750 	ptp->rx_tstamp_tc.nsec += delta;
3751 
3752 	return 0;
3753 }
3754 
3755 static int
3756 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3757 {
3758 	struct bnxt *bp = dev->data->dev_private;
3759 	int rc;
3760 	uint32_t dir_entries;
3761 	uint32_t entry_length;
3762 
3763 	rc = is_bnxt_in_error(bp);
3764 	if (rc)
3765 		return rc;
3766 
3767 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3768 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3769 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3770 
3771 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3772 	if (rc != 0)
3773 		return rc;
3774 
3775 	return dir_entries * entry_length;
3776 }
3777 
3778 static int
3779 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3780 		struct rte_dev_eeprom_info *in_eeprom)
3781 {
3782 	struct bnxt *bp = dev->data->dev_private;
3783 	uint32_t index;
3784 	uint32_t offset;
3785 	int rc;
3786 
3787 	rc = is_bnxt_in_error(bp);
3788 	if (rc)
3789 		return rc;
3790 
3791 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3792 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3793 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3794 		    in_eeprom->offset, in_eeprom->length);
3795 
3796 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3797 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3798 						in_eeprom->data);
3799 
3800 	index = in_eeprom->offset >> 24;
3801 	offset = in_eeprom->offset & 0xffffff;
3802 
3803 	if (index != 0)
3804 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3805 					   in_eeprom->length, in_eeprom->data);
3806 
3807 	return 0;
3808 }
3809 
3810 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3811 {
3812 	switch (dir_type) {
3813 	case BNX_DIR_TYPE_CHIMP_PATCH:
3814 	case BNX_DIR_TYPE_BOOTCODE:
3815 	case BNX_DIR_TYPE_BOOTCODE_2:
3816 	case BNX_DIR_TYPE_APE_FW:
3817 	case BNX_DIR_TYPE_APE_PATCH:
3818 	case BNX_DIR_TYPE_KONG_FW:
3819 	case BNX_DIR_TYPE_KONG_PATCH:
3820 	case BNX_DIR_TYPE_BONO_FW:
3821 	case BNX_DIR_TYPE_BONO_PATCH:
3822 		/* FALLTHROUGH */
3823 		return true;
3824 	}
3825 
3826 	return false;
3827 }
3828 
3829 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3830 {
3831 	switch (dir_type) {
3832 	case BNX_DIR_TYPE_AVS:
3833 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3834 	case BNX_DIR_TYPE_PCIE:
3835 	case BNX_DIR_TYPE_TSCF_UCODE:
3836 	case BNX_DIR_TYPE_EXT_PHY:
3837 	case BNX_DIR_TYPE_CCM:
3838 	case BNX_DIR_TYPE_ISCSI_BOOT:
3839 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3840 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3841 		/* FALLTHROUGH */
3842 		return true;
3843 	}
3844 
3845 	return false;
3846 }
3847 
3848 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3849 {
3850 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3851 		bnxt_dir_type_is_other_exec_format(dir_type);
3852 }
3853 
3854 static int
3855 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3856 		struct rte_dev_eeprom_info *in_eeprom)
3857 {
3858 	struct bnxt *bp = dev->data->dev_private;
3859 	uint8_t index, dir_op;
3860 	uint16_t type, ext, ordinal, attr;
3861 	int rc;
3862 
3863 	rc = is_bnxt_in_error(bp);
3864 	if (rc)
3865 		return rc;
3866 
3867 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3868 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3869 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3870 		    in_eeprom->offset, in_eeprom->length);
3871 
3872 	if (!BNXT_PF(bp)) {
3873 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3874 		return -EINVAL;
3875 	}
3876 
3877 	type = in_eeprom->magic >> 16;
3878 
3879 	if (type == 0xffff) { /* special value for directory operations */
3880 		index = in_eeprom->magic & 0xff;
3881 		dir_op = in_eeprom->magic >> 8;
3882 		if (index == 0)
3883 			return -EINVAL;
3884 		switch (dir_op) {
3885 		case 0x0e: /* erase */
3886 			if (in_eeprom->offset != ~in_eeprom->magic)
3887 				return -EINVAL;
3888 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3889 		default:
3890 			return -EINVAL;
3891 		}
3892 	}
3893 
3894 	/* Create or re-write an NVM item: */
3895 	if (bnxt_dir_type_is_executable(type) == true)
3896 		return -EOPNOTSUPP;
3897 	ext = in_eeprom->magic & 0xffff;
3898 	ordinal = in_eeprom->offset >> 16;
3899 	attr = in_eeprom->offset & 0xffff;
3900 
3901 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3902 				     in_eeprom->data, in_eeprom->length);
3903 }
3904 
3905 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3906 				struct rte_eth_dev_module_info *modinfo)
3907 {
3908 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3909 	struct bnxt *bp = dev->data->dev_private;
3910 	int rc;
3911 
3912 	/* No point in going further if phy status indicates
3913 	 * module is not inserted or if it is powered down or
3914 	 * if it is of type 10GBase-T
3915 	 */
3916 	if (bp->link_info->module_status >
3917 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3918 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3919 			    dev->data->port_id);
3920 		return -ENOTSUP;
3921 	}
3922 
3923 	/* This feature is not supported in older firmware versions */
3924 	if (bp->hwrm_spec_code < 0x10202) {
3925 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3926 			    dev->data->port_id);
3927 		return -ENOTSUP;
3928 	}
3929 
3930 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3931 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3932 						   module_info);
3933 
3934 	if (rc)
3935 		return rc;
3936 
3937 	switch (module_info[0]) {
3938 	case SFF_MODULE_ID_SFP:
3939 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3940 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3941 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3942 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3943 		break;
3944 	case SFF_MODULE_ID_QSFP:
3945 	case SFF_MODULE_ID_QSFP_PLUS:
3946 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3947 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3948 		break;
3949 	case SFF_MODULE_ID_QSFP28:
3950 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3951 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3952 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3953 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3954 		break;
3955 	default:
3956 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3957 		return -ENOTSUP;
3958 	}
3959 
3960 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3961 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3962 
3963 	return 0;
3964 }
3965 
3966 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3967 				  struct rte_dev_eeprom_info *info)
3968 {
3969 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3970 	uint32_t offset = info->offset, length = info->length;
3971 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3972 	struct bnxt *bp = dev->data->dev_private;
3973 	uint8_t *data = info->data;
3974 	uint8_t page = offset >> 7;
3975 	uint8_t max_pages = 2;
3976 	uint8_t opt_pages;
3977 	int rc;
3978 
3979 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3980 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3981 						   module_info);
3982 	if (rc)
3983 		return rc;
3984 
3985 	switch (module_info[0]) {
3986 	case SFF_MODULE_ID_SFP:
3987 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
3988 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
3989 			pg_addr[2] = I2C_DEV_ADDR_A2;
3990 			pg_addr[3] = I2C_DEV_ADDR_A2;
3991 			max_pages = 4;
3992 		}
3993 		break;
3994 	case SFF_MODULE_ID_QSFP28:
3995 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
3996 							   SFF8636_OPT_PAGES_OFFSET,
3997 							   1, &opt_pages);
3998 		if (rc)
3999 			return rc;
4000 
4001 		if (opt_pages & SFF8636_PAGE1_MASK) {
4002 			pg_addr[2] = I2C_DEV_ADDR_A0;
4003 			max_pages = 3;
4004 		}
4005 		if (opt_pages & SFF8636_PAGE2_MASK) {
4006 			pg_addr[3] = I2C_DEV_ADDR_A0;
4007 			max_pages = 4;
4008 		}
4009 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
4010 			pg_addr[4] = I2C_DEV_ADDR_A0;
4011 			max_pages = 5;
4012 		}
4013 		break;
4014 	default:
4015 		break;
4016 	}
4017 
4018 	memset(data, 0, length);
4019 
4020 	offset &= 0xff;
4021 	while (length && page < max_pages) {
4022 		uint8_t raw_page = page ? page - 1 : 0;
4023 		uint16_t chunk;
4024 
4025 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
4026 			raw_page = 0;
4027 		else if (page)
4028 			offset |= 0x80;
4029 		chunk = RTE_MIN(length, 256 - offset);
4030 
4031 		if (pg_addr[page]) {
4032 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
4033 								   raw_page, offset,
4034 								   chunk, data);
4035 			if (rc)
4036 				return rc;
4037 		}
4038 
4039 		data += chunk;
4040 		length -= chunk;
4041 		offset = 0;
4042 		page += 1 + (chunk > 128);
4043 	}
4044 
4045 	return length ? -EINVAL : 0;
4046 }
4047 
4048 /*
4049  * Initialization
4050  */
4051 
4052 static const struct eth_dev_ops bnxt_dev_ops = {
4053 	.dev_infos_get = bnxt_dev_info_get_op,
4054 	.dev_close = bnxt_dev_close_op,
4055 	.dev_configure = bnxt_dev_configure_op,
4056 	.dev_start = bnxt_dev_start_op,
4057 	.dev_stop = bnxt_dev_stop_op,
4058 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4059 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4060 	.stats_get = bnxt_stats_get_op,
4061 	.stats_reset = bnxt_stats_reset_op,
4062 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4063 	.rx_queue_release = bnxt_rx_queue_release_op,
4064 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4065 	.tx_queue_release = bnxt_tx_queue_release_op,
4066 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4067 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4068 	.reta_update = bnxt_reta_update_op,
4069 	.reta_query = bnxt_reta_query_op,
4070 	.rss_hash_update = bnxt_rss_hash_update_op,
4071 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4072 	.link_update = bnxt_link_update_op,
4073 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4074 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4075 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4076 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4077 	.mac_addr_add = bnxt_mac_addr_add_op,
4078 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4079 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4080 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4081 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4082 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4083 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4084 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4085 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4086 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4087 	.mtu_set = bnxt_mtu_set_op,
4088 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4089 	.xstats_get = bnxt_dev_xstats_get_op,
4090 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4091 	.xstats_reset = bnxt_dev_xstats_reset_op,
4092 	.fw_version_get = bnxt_fw_version_get,
4093 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4094 	.rxq_info_get = bnxt_rxq_info_get_op,
4095 	.txq_info_get = bnxt_txq_info_get_op,
4096 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4097 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4098 	.dev_led_on = bnxt_dev_led_on_op,
4099 	.dev_led_off = bnxt_dev_led_off_op,
4100 	.rx_queue_start = bnxt_rx_queue_start,
4101 	.rx_queue_stop = bnxt_rx_queue_stop,
4102 	.tx_queue_start = bnxt_tx_queue_start,
4103 	.tx_queue_stop = bnxt_tx_queue_stop,
4104 	.flow_ops_get = bnxt_flow_ops_get_op,
4105 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4106 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4107 	.get_eeprom           = bnxt_get_eeprom_op,
4108 	.set_eeprom           = bnxt_set_eeprom_op,
4109 	.get_module_info = bnxt_get_module_info,
4110 	.get_module_eeprom = bnxt_get_module_eeprom,
4111 	.timesync_enable      = bnxt_timesync_enable,
4112 	.timesync_disable     = bnxt_timesync_disable,
4113 	.timesync_read_time   = bnxt_timesync_read_time,
4114 	.timesync_write_time   = bnxt_timesync_write_time,
4115 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4116 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4117 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4118 };
4119 
4120 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4121 {
4122 	uint32_t offset;
4123 
4124 	/* Only pre-map the reset GRC registers using window 3 */
4125 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4126 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4127 
4128 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4129 
4130 	return offset;
4131 }
4132 
4133 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4134 {
4135 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4136 	uint32_t reg_base = 0xffffffff;
4137 	int i;
4138 
4139 	/* Only pre-map the monitoring GRC registers using window 2 */
4140 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4141 		uint32_t reg = info->status_regs[i];
4142 
4143 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4144 			continue;
4145 
4146 		if (reg_base == 0xffffffff)
4147 			reg_base = reg & 0xfffff000;
4148 		if ((reg & 0xfffff000) != reg_base)
4149 			return -ERANGE;
4150 
4151 		/* Use mask 0xffc as the Lower 2 bits indicates
4152 		 * address space location
4153 		 */
4154 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4155 						(reg & 0xffc);
4156 	}
4157 
4158 	if (reg_base == 0xffffffff)
4159 		return 0;
4160 
4161 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4162 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4163 
4164 	return 0;
4165 }
4166 
4167 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4168 {
4169 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4170 	uint32_t delay = info->delay_after_reset[index];
4171 	uint32_t val = info->reset_reg_val[index];
4172 	uint32_t reg = info->reset_reg[index];
4173 	uint32_t type, offset;
4174 	int ret;
4175 
4176 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4177 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4178 
4179 	switch (type) {
4180 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4181 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4182 		if (ret < 0) {
4183 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4184 				    val, offset);
4185 			return;
4186 		}
4187 		break;
4188 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4189 		offset = bnxt_map_reset_regs(bp, offset);
4190 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4191 		break;
4192 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4193 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4194 		break;
4195 	}
4196 	/* wait on a specific interval of time until core reset is complete */
4197 	if (delay)
4198 		rte_delay_ms(delay);
4199 }
4200 
4201 static void bnxt_dev_cleanup(struct bnxt *bp)
4202 {
4203 	bp->eth_dev->data->dev_link.link_status = 0;
4204 	bp->link_info->link_up = 0;
4205 	if (bp->eth_dev->data->dev_started)
4206 		bnxt_dev_stop(bp->eth_dev);
4207 
4208 	bnxt_uninit_resources(bp, true);
4209 }
4210 
4211 static int
4212 bnxt_check_fw_reset_done(struct bnxt *bp)
4213 {
4214 	int timeout = bp->fw_reset_max_msecs;
4215 	uint16_t val = 0;
4216 	int rc;
4217 
4218 	do {
4219 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4220 		if (rc < 0) {
4221 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4222 			return rc;
4223 		}
4224 		if (val != 0xffff)
4225 			break;
4226 		rte_delay_ms(1);
4227 	} while (timeout--);
4228 
4229 	if (val == 0xffff) {
4230 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4231 		return -1;
4232 	}
4233 
4234 	return 0;
4235 }
4236 
4237 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4238 {
4239 	struct rte_eth_dev *dev = bp->eth_dev;
4240 	struct rte_vlan_filter_conf *vfc;
4241 	int vidx, vbit, rc;
4242 	uint16_t vlan_id;
4243 
4244 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4245 		vfc = &dev->data->vlan_filter_conf;
4246 		vidx = vlan_id / 64;
4247 		vbit = vlan_id % 64;
4248 
4249 		/* Each bit corresponds to a VLAN id */
4250 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4251 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4252 			if (rc)
4253 				return rc;
4254 		}
4255 	}
4256 
4257 	return 0;
4258 }
4259 
4260 static int bnxt_restore_mac_filters(struct bnxt *bp)
4261 {
4262 	struct rte_eth_dev *dev = bp->eth_dev;
4263 	struct rte_eth_dev_info dev_info;
4264 	struct rte_ether_addr *addr;
4265 	uint64_t pool_mask;
4266 	uint32_t pool = 0;
4267 	uint32_t i;
4268 	int rc;
4269 
4270 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4271 		return 0;
4272 
4273 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4274 	if (rc)
4275 		return rc;
4276 
4277 	/* replay MAC address configuration */
4278 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4279 		addr = &dev->data->mac_addrs[i];
4280 
4281 		/* skip zero address */
4282 		if (rte_is_zero_ether_addr(addr))
4283 			continue;
4284 
4285 		pool = 0;
4286 		pool_mask = dev->data->mac_pool_sel[i];
4287 
4288 		do {
4289 			if (pool_mask & 1ULL) {
4290 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4291 				if (rc)
4292 					return rc;
4293 			}
4294 			pool_mask >>= 1;
4295 			pool++;
4296 		} while (pool_mask);
4297 	}
4298 
4299 	return 0;
4300 }
4301 
4302 static int bnxt_restore_filters(struct bnxt *bp)
4303 {
4304 	struct rte_eth_dev *dev = bp->eth_dev;
4305 	int ret = 0;
4306 
4307 	if (dev->data->all_multicast) {
4308 		ret = bnxt_allmulticast_enable_op(dev);
4309 		if (ret)
4310 			return ret;
4311 	}
4312 	if (dev->data->promiscuous) {
4313 		ret = bnxt_promiscuous_enable_op(dev);
4314 		if (ret)
4315 			return ret;
4316 	}
4317 
4318 	ret = bnxt_restore_mac_filters(bp);
4319 	if (ret)
4320 		return ret;
4321 
4322 	ret = bnxt_restore_vlan_filters(bp);
4323 	/* TODO restore other filters as well */
4324 	return ret;
4325 }
4326 
4327 static int bnxt_check_fw_ready(struct bnxt *bp)
4328 {
4329 	int timeout = bp->fw_reset_max_msecs;
4330 	int rc = 0;
4331 
4332 	do {
4333 		rc = bnxt_hwrm_poll_ver_get(bp);
4334 		if (rc == 0)
4335 			break;
4336 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4337 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4338 	} while (rc && timeout > 0);
4339 
4340 	if (rc)
4341 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4342 
4343 	return rc;
4344 }
4345 
4346 static void bnxt_dev_recover(void *arg)
4347 {
4348 	struct bnxt *bp = arg;
4349 	int rc = 0;
4350 
4351 	pthread_mutex_lock(&bp->err_recovery_lock);
4352 
4353 	if (!bp->fw_reset_min_msecs) {
4354 		rc = bnxt_check_fw_reset_done(bp);
4355 		if (rc)
4356 			goto err;
4357 	}
4358 
4359 	/* Clear Error flag so that device re-init should happen */
4360 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4361 
4362 	rc = bnxt_check_fw_ready(bp);
4363 	if (rc)
4364 		goto err;
4365 
4366 	rc = bnxt_init_resources(bp, true);
4367 	if (rc) {
4368 		PMD_DRV_LOG(ERR,
4369 			    "Failed to initialize resources after reset\n");
4370 		goto err;
4371 	}
4372 	/* clear reset flag as the device is initialized now */
4373 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4374 
4375 	rc = bnxt_dev_start_op(bp->eth_dev);
4376 	if (rc) {
4377 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4378 		goto err_start;
4379 	}
4380 
4381 	rc = bnxt_restore_filters(bp);
4382 	if (rc)
4383 		goto err_start;
4384 
4385 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4386 	pthread_mutex_unlock(&bp->err_recovery_lock);
4387 
4388 	return;
4389 err_start:
4390 	bnxt_dev_stop(bp->eth_dev);
4391 err:
4392 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4393 	bnxt_uninit_resources(bp, false);
4394 	if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4395 		rte_eth_dev_callback_process(bp->eth_dev,
4396 					     RTE_ETH_EVENT_INTR_RMV,
4397 					     NULL);
4398 	pthread_mutex_unlock(&bp->err_recovery_lock);
4399 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4400 }
4401 
4402 void bnxt_dev_reset_and_resume(void *arg)
4403 {
4404 	struct bnxt *bp = arg;
4405 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4406 	uint16_t val = 0;
4407 	int rc;
4408 
4409 	bnxt_dev_cleanup(bp);
4410 
4411 	bnxt_wait_for_device_shutdown(bp);
4412 
4413 	/* During some fatal firmware error conditions, the PCI config space
4414 	 * register 0x2e which normally contains the subsystem ID will become
4415 	 * 0xffff. This register will revert back to the normal value after
4416 	 * the chip has completed core reset. If we detect this condition,
4417 	 * we can poll this config register immediately for the value to revert.
4418 	 */
4419 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4420 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4421 		if (rc < 0) {
4422 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4423 			return;
4424 		}
4425 		if (val == 0xffff) {
4426 			bp->fw_reset_min_msecs = 0;
4427 			us = 1;
4428 		}
4429 	}
4430 
4431 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4432 	if (rc)
4433 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4434 }
4435 
4436 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4437 {
4438 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4439 	uint32_t reg = info->status_regs[index];
4440 	uint32_t type, offset, val = 0;
4441 	int ret = 0;
4442 
4443 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4444 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4445 
4446 	switch (type) {
4447 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4448 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4449 		if (ret < 0)
4450 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4451 				    offset);
4452 		break;
4453 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4454 		offset = info->mapped_status_regs[index];
4455 		/* FALLTHROUGH */
4456 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4457 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4458 				       offset));
4459 		break;
4460 	}
4461 
4462 	return val;
4463 }
4464 
4465 static int bnxt_fw_reset_all(struct bnxt *bp)
4466 {
4467 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4468 	uint32_t i;
4469 	int rc = 0;
4470 
4471 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4472 		/* Reset through primary function driver */
4473 		for (i = 0; i < info->reg_array_cnt; i++)
4474 			bnxt_write_fw_reset_reg(bp, i);
4475 		/* Wait for time specified by FW after triggering reset */
4476 		rte_delay_ms(info->primary_func_wait_period_after_reset);
4477 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4478 		/* Reset with the help of Kong processor */
4479 		rc = bnxt_hwrm_fw_reset(bp);
4480 		if (rc)
4481 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4482 	}
4483 
4484 	return rc;
4485 }
4486 
4487 static void bnxt_fw_reset_cb(void *arg)
4488 {
4489 	struct bnxt *bp = arg;
4490 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4491 	int rc = 0;
4492 
4493 	/* Only Primary function can do FW reset */
4494 	if (bnxt_is_primary_func(bp) &&
4495 	    bnxt_is_recovery_enabled(bp)) {
4496 		rc = bnxt_fw_reset_all(bp);
4497 		if (rc) {
4498 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4499 			return;
4500 		}
4501 	}
4502 
4503 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4504 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4505 	 * (including MASTER FUNC). After receiving this Async, all the active
4506 	 * drivers should treat this case as FW initiated recovery
4507 	 */
4508 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4509 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4510 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4511 
4512 		/* To recover from error */
4513 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4514 				  (void *)bp);
4515 	}
4516 }
4517 
4518 /* Driver should poll FW heartbeat, reset_counter with the frequency
4519  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4520  * When the driver detects heartbeat stop or change in reset_counter,
4521  * it has to trigger a reset to recover from the error condition.
4522  * A “primary function” is the function who will have the privilege to
4523  * initiate the chimp reset. The primary function will be elected by the
4524  * firmware and will be notified through async message.
4525  */
4526 static void bnxt_check_fw_health(void *arg)
4527 {
4528 	struct bnxt *bp = arg;
4529 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4530 	uint32_t val = 0, wait_msec;
4531 
4532 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4533 	    is_bnxt_in_error(bp))
4534 		return;
4535 
4536 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4537 	if (val == info->last_heart_beat)
4538 		goto reset;
4539 
4540 	info->last_heart_beat = val;
4541 
4542 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4543 	if (val != info->last_reset_counter)
4544 		goto reset;
4545 
4546 	info->last_reset_counter = val;
4547 
4548 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4549 			  bnxt_check_fw_health, (void *)bp);
4550 
4551 	return;
4552 reset:
4553 	/* Stop DMA to/from device */
4554 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4555 	bp->flags |= BNXT_FLAG_FW_RESET;
4556 
4557 	bnxt_stop_rxtx(bp);
4558 
4559 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4560 
4561 	if (bnxt_is_primary_func(bp))
4562 		wait_msec = info->primary_func_wait_period;
4563 	else
4564 		wait_msec = info->normal_func_wait_period;
4565 
4566 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4567 			  bnxt_fw_reset_cb, (void *)bp);
4568 }
4569 
4570 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4571 {
4572 	uint32_t polling_freq;
4573 
4574 	pthread_mutex_lock(&bp->health_check_lock);
4575 
4576 	if (!bnxt_is_recovery_enabled(bp))
4577 		goto done;
4578 
4579 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4580 		goto done;
4581 
4582 	polling_freq = bp->recovery_info->driver_polling_freq;
4583 
4584 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4585 			  bnxt_check_fw_health, (void *)bp);
4586 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4587 
4588 done:
4589 	pthread_mutex_unlock(&bp->health_check_lock);
4590 }
4591 
4592 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4593 {
4594 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4595 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4596 }
4597 
4598 static bool bnxt_vf_pciid(uint16_t device_id)
4599 {
4600 	switch (device_id) {
4601 	case BROADCOM_DEV_ID_57304_VF:
4602 	case BROADCOM_DEV_ID_57406_VF:
4603 	case BROADCOM_DEV_ID_5731X_VF:
4604 	case BROADCOM_DEV_ID_5741X_VF:
4605 	case BROADCOM_DEV_ID_57414_VF:
4606 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4607 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4608 	case BROADCOM_DEV_ID_58802_VF:
4609 	case BROADCOM_DEV_ID_57500_VF1:
4610 	case BROADCOM_DEV_ID_57500_VF2:
4611 	case BROADCOM_DEV_ID_58818_VF:
4612 		/* FALLTHROUGH */
4613 		return true;
4614 	default:
4615 		return false;
4616 	}
4617 }
4618 
4619 /* Phase 5 device */
4620 static bool bnxt_p5_device(uint16_t device_id)
4621 {
4622 	switch (device_id) {
4623 	case BROADCOM_DEV_ID_57508:
4624 	case BROADCOM_DEV_ID_57504:
4625 	case BROADCOM_DEV_ID_57502:
4626 	case BROADCOM_DEV_ID_57508_MF1:
4627 	case BROADCOM_DEV_ID_57504_MF1:
4628 	case BROADCOM_DEV_ID_57502_MF1:
4629 	case BROADCOM_DEV_ID_57508_MF2:
4630 	case BROADCOM_DEV_ID_57504_MF2:
4631 	case BROADCOM_DEV_ID_57502_MF2:
4632 	case BROADCOM_DEV_ID_57500_VF1:
4633 	case BROADCOM_DEV_ID_57500_VF2:
4634 	case BROADCOM_DEV_ID_58812:
4635 	case BROADCOM_DEV_ID_58814:
4636 	case BROADCOM_DEV_ID_58818:
4637 	case BROADCOM_DEV_ID_58818_VF:
4638 		/* FALLTHROUGH */
4639 		return true;
4640 	default:
4641 		return false;
4642 	}
4643 }
4644 
4645 bool bnxt_stratus_device(struct bnxt *bp)
4646 {
4647 	uint16_t device_id = bp->pdev->id.device_id;
4648 
4649 	switch (device_id) {
4650 	case BROADCOM_DEV_ID_STRATUS_NIC:
4651 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4652 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4653 		/* FALLTHROUGH */
4654 		return true;
4655 	default:
4656 		return false;
4657 	}
4658 }
4659 
4660 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4661 {
4662 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4663 	struct bnxt *bp = eth_dev->data->dev_private;
4664 
4665 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4666 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4667 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4668 	if (!bp->bar0 || !bp->doorbell_base) {
4669 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4670 		return -ENODEV;
4671 	}
4672 
4673 	bp->eth_dev = eth_dev;
4674 	bp->pdev = pci_dev;
4675 
4676 	return 0;
4677 }
4678 
4679 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4680 				  struct bnxt_ctx_pg_info *ctx_pg,
4681 				  uint32_t mem_size,
4682 				  const char *suffix,
4683 				  uint16_t idx)
4684 {
4685 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4686 	const struct rte_memzone *mz = NULL;
4687 	char mz_name[RTE_MEMZONE_NAMESIZE];
4688 	rte_iova_t mz_phys_addr;
4689 	uint64_t valid_bits = 0;
4690 	uint32_t sz;
4691 	int i;
4692 
4693 	if (!mem_size)
4694 		return 0;
4695 
4696 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4697 			 BNXT_PAGE_SIZE;
4698 	rmem->page_size = BNXT_PAGE_SIZE;
4699 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4700 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4701 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4702 
4703 	valid_bits = PTU_PTE_VALID;
4704 
4705 	if (rmem->nr_pages > 1) {
4706 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4707 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4708 			 suffix, idx, bp->eth_dev->data->port_id);
4709 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4710 		mz = rte_memzone_lookup(mz_name);
4711 		if (!mz) {
4712 			mz = rte_memzone_reserve_aligned(mz_name,
4713 						rmem->nr_pages * 8,
4714 						bp->eth_dev->device->numa_node,
4715 						RTE_MEMZONE_2MB |
4716 						RTE_MEMZONE_SIZE_HINT_ONLY |
4717 						RTE_MEMZONE_IOVA_CONTIG,
4718 						BNXT_PAGE_SIZE);
4719 			if (mz == NULL)
4720 				return -ENOMEM;
4721 		}
4722 
4723 		memset(mz->addr, 0, mz->len);
4724 		mz_phys_addr = mz->iova;
4725 
4726 		rmem->pg_tbl = mz->addr;
4727 		rmem->pg_tbl_map = mz_phys_addr;
4728 		rmem->pg_tbl_mz = mz;
4729 	}
4730 
4731 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4732 		 suffix, idx, bp->eth_dev->data->port_id);
4733 	mz = rte_memzone_lookup(mz_name);
4734 	if (!mz) {
4735 		mz = rte_memzone_reserve_aligned(mz_name,
4736 						 mem_size,
4737 						 bp->eth_dev->device->numa_node,
4738 						 RTE_MEMZONE_1GB |
4739 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4740 						 RTE_MEMZONE_IOVA_CONTIG,
4741 						 BNXT_PAGE_SIZE);
4742 		if (mz == NULL)
4743 			return -ENOMEM;
4744 	}
4745 
4746 	memset(mz->addr, 0, mz->len);
4747 	mz_phys_addr = mz->iova;
4748 
4749 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4750 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4751 		rmem->dma_arr[i] = mz_phys_addr + sz;
4752 
4753 		if (rmem->nr_pages > 1) {
4754 			if (i == rmem->nr_pages - 2 &&
4755 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4756 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4757 			else if (i == rmem->nr_pages - 1 &&
4758 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4759 				valid_bits |= PTU_PTE_LAST;
4760 
4761 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4762 							   valid_bits);
4763 		}
4764 	}
4765 
4766 	rmem->mz = mz;
4767 	if (rmem->vmem_size)
4768 		rmem->vmem = (void **)mz->addr;
4769 	rmem->dma_arr[0] = mz_phys_addr;
4770 	return 0;
4771 }
4772 
4773 static void bnxt_free_ctx_mem(struct bnxt *bp)
4774 {
4775 	int i;
4776 
4777 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4778 		return;
4779 
4780 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4781 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4782 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4783 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4784 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4785 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4786 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4787 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4788 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4789 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4790 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4791 
4792 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4793 		if (bp->ctx->tqm_mem[i])
4794 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4795 	}
4796 
4797 	rte_free(bp->ctx);
4798 	bp->ctx = NULL;
4799 }
4800 
4801 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4802 
4803 #define min_t(type, x, y) ({                    \
4804 	type __min1 = (x);                      \
4805 	type __min2 = (y);                      \
4806 	__min1 < __min2 ? __min1 : __min2; })
4807 
4808 #define max_t(type, x, y) ({                    \
4809 	type __max1 = (x);                      \
4810 	type __max2 = (y);                      \
4811 	__max1 > __max2 ? __max1 : __max2; })
4812 
4813 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4814 
4815 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4816 {
4817 	struct bnxt_ctx_pg_info *ctx_pg;
4818 	struct bnxt_ctx_mem_info *ctx;
4819 	uint32_t mem_size, ena, entries;
4820 	uint32_t entries_sp, min;
4821 	int i, rc;
4822 
4823 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4824 	if (rc) {
4825 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4826 		return rc;
4827 	}
4828 	ctx = bp->ctx;
4829 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4830 		return 0;
4831 
4832 	ctx_pg = &ctx->qp_mem;
4833 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4834 	if (ctx->qp_entry_size) {
4835 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4836 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4837 		if (rc)
4838 			return rc;
4839 	}
4840 
4841 	ctx_pg = &ctx->srq_mem;
4842 	ctx_pg->entries = ctx->srq_max_l2_entries;
4843 	if (ctx->srq_entry_size) {
4844 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4845 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4846 		if (rc)
4847 			return rc;
4848 	}
4849 
4850 	ctx_pg = &ctx->cq_mem;
4851 	ctx_pg->entries = ctx->cq_max_l2_entries;
4852 	if (ctx->cq_entry_size) {
4853 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4854 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4855 		if (rc)
4856 			return rc;
4857 	}
4858 
4859 	ctx_pg = &ctx->vnic_mem;
4860 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4861 		ctx->vnic_max_ring_table_entries;
4862 	if (ctx->vnic_entry_size) {
4863 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4864 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4865 		if (rc)
4866 			return rc;
4867 	}
4868 
4869 	ctx_pg = &ctx->stat_mem;
4870 	ctx_pg->entries = ctx->stat_max_entries;
4871 	if (ctx->stat_entry_size) {
4872 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4873 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4874 		if (rc)
4875 			return rc;
4876 	}
4877 
4878 	min = ctx->tqm_min_entries_per_ring;
4879 
4880 	entries_sp = ctx->qp_max_l2_entries +
4881 		     ctx->vnic_max_vnic_entries +
4882 		     2 * ctx->qp_min_qp1_entries + min;
4883 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4884 
4885 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4886 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4887 	entries = clamp_t(uint32_t, entries, min,
4888 			  ctx->tqm_max_entries_per_ring);
4889 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4890 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4891 		 * i > 8 is other ext rings.
4892 		 */
4893 		ctx_pg = ctx->tqm_mem[i];
4894 		ctx_pg->entries = i ? entries : entries_sp;
4895 		if (ctx->tqm_entry_size) {
4896 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4897 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4898 						    "tqm_mem", i);
4899 			if (rc)
4900 				return rc;
4901 		}
4902 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4903 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4904 		else
4905 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4906 	}
4907 
4908 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4909 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4910 	if (rc)
4911 		PMD_DRV_LOG(ERR,
4912 			    "Failed to configure context mem: rc = %d\n", rc);
4913 	else
4914 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4915 
4916 	return rc;
4917 }
4918 
4919 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4920 {
4921 	struct rte_pci_device *pci_dev = bp->pdev;
4922 	char mz_name[RTE_MEMZONE_NAMESIZE];
4923 	const struct rte_memzone *mz = NULL;
4924 	uint32_t total_alloc_len;
4925 	rte_iova_t mz_phys_addr;
4926 
4927 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4928 		return 0;
4929 
4930 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4931 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4932 		 pci_dev->addr.bus, pci_dev->addr.devid,
4933 		 pci_dev->addr.function, "rx_port_stats");
4934 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4935 	mz = rte_memzone_lookup(mz_name);
4936 	total_alloc_len =
4937 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4938 				       sizeof(struct rx_port_stats_ext) + 512);
4939 	if (!mz) {
4940 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4941 					 SOCKET_ID_ANY,
4942 					 RTE_MEMZONE_2MB |
4943 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4944 					 RTE_MEMZONE_IOVA_CONTIG);
4945 		if (mz == NULL)
4946 			return -ENOMEM;
4947 	}
4948 	memset(mz->addr, 0, mz->len);
4949 	mz_phys_addr = mz->iova;
4950 
4951 	bp->rx_mem_zone = (const void *)mz;
4952 	bp->hw_rx_port_stats = mz->addr;
4953 	bp->hw_rx_port_stats_map = mz_phys_addr;
4954 
4955 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4956 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4957 		 pci_dev->addr.bus, pci_dev->addr.devid,
4958 		 pci_dev->addr.function, "tx_port_stats");
4959 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4960 	mz = rte_memzone_lookup(mz_name);
4961 	total_alloc_len =
4962 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4963 				       sizeof(struct tx_port_stats_ext) + 512);
4964 	if (!mz) {
4965 		mz = rte_memzone_reserve(mz_name,
4966 					 total_alloc_len,
4967 					 SOCKET_ID_ANY,
4968 					 RTE_MEMZONE_2MB |
4969 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4970 					 RTE_MEMZONE_IOVA_CONTIG);
4971 		if (mz == NULL)
4972 			return -ENOMEM;
4973 	}
4974 	memset(mz->addr, 0, mz->len);
4975 	mz_phys_addr = mz->iova;
4976 
4977 	bp->tx_mem_zone = (const void *)mz;
4978 	bp->hw_tx_port_stats = mz->addr;
4979 	bp->hw_tx_port_stats_map = mz_phys_addr;
4980 	bp->flags |= BNXT_FLAG_PORT_STATS;
4981 
4982 	/* Display extended statistics if FW supports it */
4983 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4984 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4985 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4986 		return 0;
4987 
4988 	bp->hw_rx_port_stats_ext = (void *)
4989 		((uint8_t *)bp->hw_rx_port_stats +
4990 		 sizeof(struct rx_port_stats));
4991 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4992 		sizeof(struct rx_port_stats);
4993 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4994 
4995 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4996 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4997 		bp->hw_tx_port_stats_ext = (void *)
4998 			((uint8_t *)bp->hw_tx_port_stats +
4999 			 sizeof(struct tx_port_stats));
5000 		bp->hw_tx_port_stats_ext_map =
5001 			bp->hw_tx_port_stats_map +
5002 			sizeof(struct tx_port_stats);
5003 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
5004 	}
5005 
5006 	return 0;
5007 }
5008 
5009 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
5010 {
5011 	struct bnxt *bp = eth_dev->data->dev_private;
5012 	int rc = 0;
5013 
5014 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
5015 					       RTE_ETHER_ADDR_LEN *
5016 					       bp->max_l2_ctx,
5017 					       0);
5018 	if (eth_dev->data->mac_addrs == NULL) {
5019 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
5020 		return -ENOMEM;
5021 	}
5022 
5023 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
5024 		if (BNXT_PF(bp))
5025 			return -EINVAL;
5026 
5027 		/* Generate a random MAC address, if none was assigned by PF */
5028 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
5029 		bnxt_eth_hw_addr_random(bp->mac_addr);
5030 		PMD_DRV_LOG(INFO,
5031 			    "Assign random MAC:" RTE_ETHER_ADDR_PRT_FMT "\n",
5032 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
5033 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
5034 
5035 		rc = bnxt_hwrm_set_mac(bp);
5036 		if (rc)
5037 			return rc;
5038 	}
5039 
5040 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5041 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5042 
5043 	return rc;
5044 }
5045 
5046 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5047 {
5048 	int rc = 0;
5049 
5050 	/* MAC is already configured in FW */
5051 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5052 		return 0;
5053 
5054 	/* Restore the old MAC configured */
5055 	rc = bnxt_hwrm_set_mac(bp);
5056 	if (rc)
5057 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5058 
5059 	return rc;
5060 }
5061 
5062 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5063 {
5064 	if (!BNXT_PF(bp))
5065 		return;
5066 
5067 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5068 
5069 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5070 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5071 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5072 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5073 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5074 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5075 }
5076 
5077 struct bnxt *
5078 bnxt_get_bp(uint16_t port)
5079 {
5080 	struct bnxt *bp;
5081 	struct rte_eth_dev *dev;
5082 
5083 	if (!rte_eth_dev_is_valid_port(port)) {
5084 		PMD_DRV_LOG(ERR, "Invalid port %d\n", port);
5085 		return NULL;
5086 	}
5087 
5088 	dev = &rte_eth_devices[port];
5089 	if (!is_bnxt_supported(dev)) {
5090 		PMD_DRV_LOG(ERR, "Device %d not supported\n", port);
5091 		return NULL;
5092 	}
5093 
5094 	bp = (struct bnxt *)dev->data->dev_private;
5095 	if (!BNXT_TRUFLOW_EN(bp)) {
5096 		PMD_DRV_LOG(ERR, "TRUFLOW not enabled\n");
5097 		return NULL;
5098 	}
5099 
5100 	return bp;
5101 }
5102 
5103 uint16_t
5104 bnxt_get_svif(uint16_t port_id, bool func_svif,
5105 	      enum bnxt_ulp_intf_type type)
5106 {
5107 	struct rte_eth_dev *eth_dev;
5108 	struct bnxt *bp;
5109 
5110 	eth_dev = &rte_eth_devices[port_id];
5111 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5112 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5113 		if (!vfr)
5114 			return 0;
5115 
5116 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5117 			return vfr->svif;
5118 
5119 		eth_dev = vfr->parent_dev;
5120 	}
5121 
5122 	bp = eth_dev->data->dev_private;
5123 
5124 	return func_svif ? bp->func_svif : bp->port_svif;
5125 }
5126 
5127 void
5128 bnxt_get_iface_mac(uint16_t port, enum bnxt_ulp_intf_type type,
5129 		   uint8_t *mac, uint8_t *parent_mac)
5130 {
5131 	struct rte_eth_dev *eth_dev;
5132 	struct bnxt *bp;
5133 
5134 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF &&
5135 	    type != BNXT_ULP_INTF_TYPE_PF)
5136 		return;
5137 
5138 	eth_dev = &rte_eth_devices[port];
5139 	bp = eth_dev->data->dev_private;
5140 	memcpy(mac, bp->mac_addr, RTE_ETHER_ADDR_LEN);
5141 
5142 	if (type == BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5143 		memcpy(parent_mac, bp->parent->mac_addr, RTE_ETHER_ADDR_LEN);
5144 }
5145 
5146 uint16_t
5147 bnxt_get_parent_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5148 {
5149 	struct rte_eth_dev *eth_dev;
5150 	struct bnxt *bp;
5151 
5152 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5153 		return 0;
5154 
5155 	eth_dev = &rte_eth_devices[port];
5156 	bp = eth_dev->data->dev_private;
5157 
5158 	return bp->parent->vnic;
5159 }
5160 uint16_t
5161 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5162 {
5163 	struct rte_eth_dev *eth_dev;
5164 	struct bnxt_vnic_info *vnic;
5165 	struct bnxt *bp;
5166 
5167 	eth_dev = &rte_eth_devices[port];
5168 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5169 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5170 		if (!vfr)
5171 			return 0;
5172 
5173 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5174 			return vfr->dflt_vnic_id;
5175 
5176 		eth_dev = vfr->parent_dev;
5177 	}
5178 
5179 	bp = eth_dev->data->dev_private;
5180 
5181 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
5182 
5183 	return vnic->fw_vnic_id;
5184 }
5185 
5186 uint16_t
5187 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
5188 {
5189 	struct rte_eth_dev *eth_dev;
5190 	struct bnxt *bp;
5191 
5192 	eth_dev = &rte_eth_devices[port];
5193 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5194 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5195 		if (!vfr)
5196 			return 0;
5197 
5198 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5199 			return vfr->fw_fid;
5200 
5201 		eth_dev = vfr->parent_dev;
5202 	}
5203 
5204 	bp = eth_dev->data->dev_private;
5205 
5206 	return bp->fw_fid;
5207 }
5208 
5209 enum bnxt_ulp_intf_type
5210 bnxt_get_interface_type(uint16_t port)
5211 {
5212 	struct rte_eth_dev *eth_dev;
5213 	struct bnxt *bp;
5214 
5215 	eth_dev = &rte_eth_devices[port];
5216 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
5217 		return BNXT_ULP_INTF_TYPE_VF_REP;
5218 
5219 	bp = eth_dev->data->dev_private;
5220 	if (BNXT_PF(bp))
5221 		return BNXT_ULP_INTF_TYPE_PF;
5222 	else if (BNXT_VF_IS_TRUSTED(bp))
5223 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
5224 	else if (BNXT_VF(bp))
5225 		return BNXT_ULP_INTF_TYPE_VF;
5226 
5227 	return BNXT_ULP_INTF_TYPE_INVALID;
5228 }
5229 
5230 uint16_t
5231 bnxt_get_phy_port_id(uint16_t port_id)
5232 {
5233 	struct bnxt_representor *vfr;
5234 	struct rte_eth_dev *eth_dev;
5235 	struct bnxt *bp;
5236 
5237 	eth_dev = &rte_eth_devices[port_id];
5238 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5239 		vfr = eth_dev->data->dev_private;
5240 		if (!vfr)
5241 			return 0;
5242 
5243 		eth_dev = vfr->parent_dev;
5244 	}
5245 
5246 	bp = eth_dev->data->dev_private;
5247 
5248 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
5249 }
5250 
5251 uint16_t
5252 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
5253 {
5254 	struct rte_eth_dev *eth_dev;
5255 	struct bnxt *bp;
5256 
5257 	eth_dev = &rte_eth_devices[port_id];
5258 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5259 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5260 		if (!vfr)
5261 			return 0;
5262 
5263 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5264 			return vfr->fw_fid - 1;
5265 
5266 		eth_dev = vfr->parent_dev;
5267 	}
5268 
5269 	bp = eth_dev->data->dev_private;
5270 
5271 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
5272 }
5273 
5274 uint16_t
5275 bnxt_get_vport(uint16_t port_id)
5276 {
5277 	return (1 << bnxt_get_phy_port_id(port_id));
5278 }
5279 
5280 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5281 {
5282 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5283 
5284 	if (info) {
5285 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5286 			memset(info, 0, sizeof(*info));
5287 		return;
5288 	}
5289 
5290 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5291 		return;
5292 
5293 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5294 			   sizeof(*info), 0);
5295 	if (!info)
5296 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5297 
5298 	bp->recovery_info = info;
5299 }
5300 
5301 static void bnxt_check_fw_status(struct bnxt *bp)
5302 {
5303 	uint32_t fw_status;
5304 
5305 	if (!(bp->recovery_info &&
5306 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5307 		return;
5308 
5309 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5310 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5311 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5312 			    fw_status);
5313 }
5314 
5315 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5316 {
5317 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5318 	uint32_t status_loc;
5319 	uint32_t sig_ver;
5320 
5321 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5322 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5323 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5324 				   BNXT_GRCP_WINDOW_2_BASE +
5325 				   offsetof(struct hcomm_status,
5326 					    sig_ver)));
5327 	/* If the signature is absent, then FW does not support this feature */
5328 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5329 	    HCOMM_STATUS_SIGNATURE_VAL)
5330 		return 0;
5331 
5332 	if (!info) {
5333 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5334 				   sizeof(*info), 0);
5335 		if (!info)
5336 			return -ENOMEM;
5337 		bp->recovery_info = info;
5338 	} else {
5339 		memset(info, 0, sizeof(*info));
5340 	}
5341 
5342 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5343 				      BNXT_GRCP_WINDOW_2_BASE +
5344 				      offsetof(struct hcomm_status,
5345 					       fw_status_loc)));
5346 
5347 	/* Only pre-map the FW health status GRC register */
5348 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5349 		return 0;
5350 
5351 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5352 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5353 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5354 
5355 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5356 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5357 
5358 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5359 
5360 	return 0;
5361 }
5362 
5363 /* This function gets the FW version along with the
5364  * capabilities(MAX and current) of the function, vnic,
5365  * error recovery, phy and other chip related info
5366  */
5367 static int bnxt_get_config(struct bnxt *bp)
5368 {
5369 	uint16_t mtu;
5370 	int rc = 0;
5371 
5372 	bp->fw_cap = 0;
5373 
5374 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5375 	if (rc)
5376 		return rc;
5377 
5378 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5379 	if (rc) {
5380 		bnxt_check_fw_status(bp);
5381 		return rc;
5382 	}
5383 
5384 	rc = bnxt_hwrm_func_reset(bp);
5385 	if (rc)
5386 		return -EIO;
5387 
5388 	rc = bnxt_hwrm_vnic_qcaps(bp);
5389 	if (rc)
5390 		return rc;
5391 
5392 	rc = bnxt_hwrm_queue_qportcfg(bp);
5393 	if (rc)
5394 		return rc;
5395 
5396 	/* Get the MAX capabilities for this function.
5397 	 * This function also allocates context memory for TQM rings and
5398 	 * informs the firmware about this allocated backing store memory.
5399 	 */
5400 	rc = bnxt_hwrm_func_qcaps(bp);
5401 	if (rc)
5402 		return rc;
5403 
5404 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5405 	if (rc)
5406 		return rc;
5407 
5408 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5409 	if (rc)
5410 		return rc;
5411 
5412 	bnxt_hwrm_port_mac_qcfg(bp);
5413 
5414 	bnxt_hwrm_parent_pf_qcfg(bp);
5415 
5416 	bnxt_hwrm_port_phy_qcaps(bp);
5417 
5418 	bnxt_alloc_error_recovery_info(bp);
5419 	/* Get the adapter error recovery support info */
5420 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5421 	if (rc)
5422 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5423 
5424 	bnxt_hwrm_port_led_qcaps(bp);
5425 
5426 	return 0;
5427 }
5428 
5429 static int
5430 bnxt_init_locks(struct bnxt *bp)
5431 {
5432 	int err;
5433 
5434 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5435 	if (err) {
5436 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5437 		return err;
5438 	}
5439 
5440 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5441 	if (err) {
5442 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5443 		return err;
5444 	}
5445 
5446 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5447 	if (err) {
5448 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5449 		return err;
5450 	}
5451 
5452 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5453 	if (err)
5454 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5455 
5456 	return err;
5457 }
5458 
5459 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5460 {
5461 	int rc = 0;
5462 
5463 	rc = bnxt_get_config(bp);
5464 	if (rc)
5465 		return rc;
5466 
5467 	if (!reconfig_dev) {
5468 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5469 		if (rc)
5470 			return rc;
5471 	} else {
5472 		rc = bnxt_restore_dflt_mac(bp);
5473 		if (rc)
5474 			return rc;
5475 	}
5476 
5477 	bnxt_config_vf_req_fwd(bp);
5478 
5479 	rc = bnxt_hwrm_func_driver_register(bp);
5480 	if (rc) {
5481 		PMD_DRV_LOG(ERR, "Failed to register driver");
5482 		return -EBUSY;
5483 	}
5484 
5485 	if (BNXT_PF(bp)) {
5486 		if (bp->pdev->max_vfs) {
5487 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5488 			if (rc) {
5489 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5490 				return rc;
5491 			}
5492 		} else {
5493 			rc = bnxt_hwrm_allocate_pf_only(bp);
5494 			if (rc) {
5495 				PMD_DRV_LOG(ERR,
5496 					    "Failed to allocate PF resources");
5497 				return rc;
5498 			}
5499 		}
5500 	}
5501 
5502 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5503 	if (rc)
5504 		return rc;
5505 
5506 	rc = bnxt_setup_int(bp);
5507 	if (rc)
5508 		return rc;
5509 
5510 	rc = bnxt_request_int(bp);
5511 	if (rc)
5512 		return rc;
5513 
5514 	rc = bnxt_init_ctx_mem(bp);
5515 	if (rc) {
5516 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5517 		return rc;
5518 	}
5519 
5520 	return 0;
5521 }
5522 
5523 static int
5524 bnxt_parse_devarg_accum_stats(__rte_unused const char *key,
5525 			      const char *value, void *opaque_arg)
5526 {
5527 	struct bnxt *bp = opaque_arg;
5528 	unsigned long accum_stats;
5529 	char *end = NULL;
5530 
5531 	if (!value || !opaque_arg) {
5532 		PMD_DRV_LOG(ERR,
5533 			    "Invalid parameter passed to accum-stats devargs.\n");
5534 		return -EINVAL;
5535 	}
5536 
5537 	accum_stats = strtoul(value, &end, 10);
5538 	if (end == NULL || *end != '\0' ||
5539 	    (accum_stats == ULONG_MAX && errno == ERANGE)) {
5540 		PMD_DRV_LOG(ERR,
5541 			    "Invalid parameter passed to accum-stats devargs.\n");
5542 		return -EINVAL;
5543 	}
5544 
5545 	if (BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)) {
5546 		PMD_DRV_LOG(ERR,
5547 			    "Invalid value passed to accum-stats devargs.\n");
5548 		return -EINVAL;
5549 	}
5550 
5551 	if (accum_stats) {
5552 		bp->flags2 |= BNXT_FLAGS2_ACCUM_STATS_EN;
5553 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature enabled.\n");
5554 	} else {
5555 		bp->flags2 &= ~BNXT_FLAGS2_ACCUM_STATS_EN;
5556 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature disabled.\n");
5557 	}
5558 
5559 	return 0;
5560 }
5561 
5562 static int
5563 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5564 			     const char *value, void *opaque_arg)
5565 {
5566 	struct bnxt *bp = opaque_arg;
5567 	unsigned long flow_xstat;
5568 	char *end = NULL;
5569 
5570 	if (!value || !opaque_arg) {
5571 		PMD_DRV_LOG(ERR,
5572 			    "Invalid parameter passed to flow_xstat devarg.\n");
5573 		return -EINVAL;
5574 	}
5575 
5576 	flow_xstat = strtoul(value, &end, 10);
5577 	if (end == NULL || *end != '\0' ||
5578 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5579 		PMD_DRV_LOG(ERR,
5580 			    "Invalid parameter passed to flow_xstat devarg.\n");
5581 		return -EINVAL;
5582 	}
5583 
5584 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5585 		PMD_DRV_LOG(ERR,
5586 			    "Invalid value passed to flow_xstat devarg.\n");
5587 		return -EINVAL;
5588 	}
5589 
5590 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5591 	if (BNXT_FLOW_XSTATS_EN(bp))
5592 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5593 
5594 	return 0;
5595 }
5596 
5597 static int
5598 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5599 					const char *value, void *opaque_arg)
5600 {
5601 	struct bnxt *bp = opaque_arg;
5602 	unsigned long max_num_kflows;
5603 	char *end = NULL;
5604 
5605 	if (!value || !opaque_arg) {
5606 		PMD_DRV_LOG(ERR,
5607 			"Invalid parameter passed to max_num_kflows devarg.\n");
5608 		return -EINVAL;
5609 	}
5610 
5611 	max_num_kflows = strtoul(value, &end, 10);
5612 	if (end == NULL || *end != '\0' ||
5613 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5614 		PMD_DRV_LOG(ERR,
5615 			"Invalid parameter passed to max_num_kflows devarg.\n");
5616 		return -EINVAL;
5617 	}
5618 
5619 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5620 		PMD_DRV_LOG(ERR,
5621 			"Invalid value passed to max_num_kflows devarg.\n");
5622 		return -EINVAL;
5623 	}
5624 
5625 	bp->max_num_kflows = max_num_kflows;
5626 	if (bp->max_num_kflows)
5627 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5628 				max_num_kflows);
5629 
5630 	return 0;
5631 }
5632 
5633 static int
5634 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5635 				 const char *value, void *opaque_arg)
5636 {
5637 	struct bnxt *bp = opaque_arg;
5638 	unsigned long app_id;
5639 	char *end = NULL;
5640 
5641 	if (!value || !opaque_arg) {
5642 		PMD_DRV_LOG(ERR,
5643 			    "Invalid parameter passed to app-id "
5644 			    "devargs.\n");
5645 		return -EINVAL;
5646 	}
5647 
5648 	app_id = strtoul(value, &end, 10);
5649 	if (end == NULL || *end != '\0' ||
5650 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5651 		PMD_DRV_LOG(ERR,
5652 			    "Invalid parameter passed to app_id "
5653 			    "devargs.\n");
5654 		return -EINVAL;
5655 	}
5656 
5657 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5658 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5659 			    (uint16_t)app_id);
5660 		return -EINVAL;
5661 	}
5662 
5663 	bp->app_id = app_id;
5664 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5665 
5666 	return 0;
5667 }
5668 
5669 static int
5670 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5671 			    const char *value, void *opaque_arg)
5672 {
5673 	struct bnxt_representor *vfr_bp = opaque_arg;
5674 	unsigned long rep_is_pf;
5675 	char *end = NULL;
5676 
5677 	if (!value || !opaque_arg) {
5678 		PMD_DRV_LOG(ERR,
5679 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5680 		return -EINVAL;
5681 	}
5682 
5683 	rep_is_pf = strtoul(value, &end, 10);
5684 	if (end == NULL || *end != '\0' ||
5685 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5686 		PMD_DRV_LOG(ERR,
5687 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5688 		return -EINVAL;
5689 	}
5690 
5691 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5692 		PMD_DRV_LOG(ERR,
5693 			    "Invalid value passed to rep_is_pf devargs.\n");
5694 		return -EINVAL;
5695 	}
5696 
5697 	vfr_bp->flags |= rep_is_pf;
5698 	if (BNXT_REP_PF(vfr_bp))
5699 		PMD_DRV_LOG(INFO, "PF representor\n");
5700 	else
5701 		PMD_DRV_LOG(INFO, "VF representor\n");
5702 
5703 	return 0;
5704 }
5705 
5706 static int
5707 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5708 			       const char *value, void *opaque_arg)
5709 {
5710 	struct bnxt_representor *vfr_bp = opaque_arg;
5711 	unsigned long rep_based_pf;
5712 	char *end = NULL;
5713 
5714 	if (!value || !opaque_arg) {
5715 		PMD_DRV_LOG(ERR,
5716 			    "Invalid parameter passed to rep_based_pf "
5717 			    "devargs.\n");
5718 		return -EINVAL;
5719 	}
5720 
5721 	rep_based_pf = strtoul(value, &end, 10);
5722 	if (end == NULL || *end != '\0' ||
5723 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5724 		PMD_DRV_LOG(ERR,
5725 			    "Invalid parameter passed to rep_based_pf "
5726 			    "devargs.\n");
5727 		return -EINVAL;
5728 	}
5729 
5730 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5731 		PMD_DRV_LOG(ERR,
5732 			    "Invalid value passed to rep_based_pf devargs.\n");
5733 		return -EINVAL;
5734 	}
5735 
5736 	vfr_bp->rep_based_pf = rep_based_pf;
5737 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5738 
5739 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5740 
5741 	return 0;
5742 }
5743 
5744 static int
5745 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5746 			    const char *value, void *opaque_arg)
5747 {
5748 	struct bnxt_representor *vfr_bp = opaque_arg;
5749 	unsigned long rep_q_r2f;
5750 	char *end = NULL;
5751 
5752 	if (!value || !opaque_arg) {
5753 		PMD_DRV_LOG(ERR,
5754 			    "Invalid parameter passed to rep_q_r2f "
5755 			    "devargs.\n");
5756 		return -EINVAL;
5757 	}
5758 
5759 	rep_q_r2f = strtoul(value, &end, 10);
5760 	if (end == NULL || *end != '\0' ||
5761 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5762 		PMD_DRV_LOG(ERR,
5763 			    "Invalid parameter passed to rep_q_r2f "
5764 			    "devargs.\n");
5765 		return -EINVAL;
5766 	}
5767 
5768 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5769 		PMD_DRV_LOG(ERR,
5770 			    "Invalid value passed to rep_q_r2f devargs.\n");
5771 		return -EINVAL;
5772 	}
5773 
5774 	vfr_bp->rep_q_r2f = rep_q_r2f;
5775 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5776 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5777 
5778 	return 0;
5779 }
5780 
5781 static int
5782 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5783 			    const char *value, void *opaque_arg)
5784 {
5785 	struct bnxt_representor *vfr_bp = opaque_arg;
5786 	unsigned long rep_q_f2r;
5787 	char *end = NULL;
5788 
5789 	if (!value || !opaque_arg) {
5790 		PMD_DRV_LOG(ERR,
5791 			    "Invalid parameter passed to rep_q_f2r "
5792 			    "devargs.\n");
5793 		return -EINVAL;
5794 	}
5795 
5796 	rep_q_f2r = strtoul(value, &end, 10);
5797 	if (end == NULL || *end != '\0' ||
5798 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5799 		PMD_DRV_LOG(ERR,
5800 			    "Invalid parameter passed to rep_q_f2r "
5801 			    "devargs.\n");
5802 		return -EINVAL;
5803 	}
5804 
5805 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5806 		PMD_DRV_LOG(ERR,
5807 			    "Invalid value passed to rep_q_f2r devargs.\n");
5808 		return -EINVAL;
5809 	}
5810 
5811 	vfr_bp->rep_q_f2r = rep_q_f2r;
5812 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5813 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5814 
5815 	return 0;
5816 }
5817 
5818 static int
5819 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5820 			     const char *value, void *opaque_arg)
5821 {
5822 	struct bnxt_representor *vfr_bp = opaque_arg;
5823 	unsigned long rep_fc_r2f;
5824 	char *end = NULL;
5825 
5826 	if (!value || !opaque_arg) {
5827 		PMD_DRV_LOG(ERR,
5828 			    "Invalid parameter passed to rep_fc_r2f "
5829 			    "devargs.\n");
5830 		return -EINVAL;
5831 	}
5832 
5833 	rep_fc_r2f = strtoul(value, &end, 10);
5834 	if (end == NULL || *end != '\0' ||
5835 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5836 		PMD_DRV_LOG(ERR,
5837 			    "Invalid parameter passed to rep_fc_r2f "
5838 			    "devargs.\n");
5839 		return -EINVAL;
5840 	}
5841 
5842 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5843 		PMD_DRV_LOG(ERR,
5844 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5845 		return -EINVAL;
5846 	}
5847 
5848 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5849 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5850 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5851 
5852 	return 0;
5853 }
5854 
5855 static int
5856 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5857 			     const char *value, void *opaque_arg)
5858 {
5859 	struct bnxt_representor *vfr_bp = opaque_arg;
5860 	unsigned long rep_fc_f2r;
5861 	char *end = NULL;
5862 
5863 	if (!value || !opaque_arg) {
5864 		PMD_DRV_LOG(ERR,
5865 			    "Invalid parameter passed to rep_fc_f2r "
5866 			    "devargs.\n");
5867 		return -EINVAL;
5868 	}
5869 
5870 	rep_fc_f2r = strtoul(value, &end, 10);
5871 	if (end == NULL || *end != '\0' ||
5872 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5873 		PMD_DRV_LOG(ERR,
5874 			    "Invalid parameter passed to rep_fc_f2r "
5875 			    "devargs.\n");
5876 		return -EINVAL;
5877 	}
5878 
5879 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5880 		PMD_DRV_LOG(ERR,
5881 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5882 		return -EINVAL;
5883 	}
5884 
5885 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5886 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5887 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5888 
5889 	return 0;
5890 }
5891 
5892 static int
5893 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5894 {
5895 	struct rte_kvargs *kvlist;
5896 	int ret;
5897 
5898 	if (devargs == NULL)
5899 		return 0;
5900 
5901 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5902 	if (kvlist == NULL)
5903 		return -EINVAL;
5904 
5905 	/*
5906 	 * Handler for "flow_xstat" devarg.
5907 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5908 	 */
5909 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5910 				 bnxt_parse_devarg_flow_xstat, bp);
5911 	if (ret)
5912 		goto err;
5913 
5914 	/*
5915 	 * Handler for "accum-stats" devarg.
5916 	 * Invoked as for ex: "-a 0000:00:0d.0,accum-stats=1"
5917 	 */
5918 	rte_kvargs_process(kvlist, BNXT_DEVARG_ACCUM_STATS,
5919 			   bnxt_parse_devarg_accum_stats, bp);
5920 	/*
5921 	 * Handler for "max_num_kflows" devarg.
5922 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5923 	 */
5924 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5925 				 bnxt_parse_devarg_max_num_kflows, bp);
5926 	if (ret)
5927 		goto err;
5928 
5929 err:
5930 	/*
5931 	 * Handler for "app-id" devarg.
5932 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5933 	 */
5934 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5935 			   bnxt_parse_devarg_app_id, bp);
5936 
5937 	rte_kvargs_free(kvlist);
5938 	return ret;
5939 }
5940 
5941 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5942 {
5943 	int rc = 0;
5944 
5945 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5946 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5947 		if (rc)
5948 			PMD_DRV_LOG(ERR,
5949 				    "Failed to alloc switch domain: %d\n", rc);
5950 		else
5951 			PMD_DRV_LOG(INFO,
5952 				    "Switch domain allocated %d\n",
5953 				    bp->switch_domain_id);
5954 	}
5955 
5956 	return rc;
5957 }
5958 
5959 /* Allocate and initialize various fields in bnxt struct that
5960  * need to be allocated/destroyed only once in the lifetime of the driver
5961  */
5962 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5963 {
5964 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5965 	struct bnxt *bp = eth_dev->data->dev_private;
5966 	int rc = 0;
5967 
5968 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5969 
5970 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5971 		bp->flags |= BNXT_FLAG_VF;
5972 
5973 	if (bnxt_p5_device(pci_dev->id.device_id))
5974 		bp->flags |= BNXT_FLAG_CHIP_P5;
5975 
5976 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5977 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5978 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5979 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5980 		bp->flags |= BNXT_FLAG_STINGRAY;
5981 
5982 	if (BNXT_TRUFLOW_EN(bp)) {
5983 		/* extra mbuf field is required to store CFA code from mark */
5984 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5985 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5986 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5987 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5988 		};
5989 		bnxt_cfa_code_dynfield_offset =
5990 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5991 		if (bnxt_cfa_code_dynfield_offset < 0) {
5992 			PMD_DRV_LOG(ERR,
5993 			    "Failed to register mbuf field for TruFlow mark\n");
5994 			return -rte_errno;
5995 		}
5996 	}
5997 
5998 	rc = bnxt_map_pci_bars(eth_dev);
5999 	if (rc) {
6000 		PMD_DRV_LOG(ERR,
6001 			    "Failed to initialize board rc: %x\n", rc);
6002 		return rc;
6003 	}
6004 
6005 	rc = bnxt_alloc_pf_info(bp);
6006 	if (rc)
6007 		return rc;
6008 
6009 	rc = bnxt_alloc_link_info(bp);
6010 	if (rc)
6011 		return rc;
6012 
6013 	rc = bnxt_alloc_parent_info(bp);
6014 	if (rc)
6015 		return rc;
6016 
6017 	rc = bnxt_alloc_hwrm_resources(bp);
6018 	if (rc) {
6019 		PMD_DRV_LOG(ERR,
6020 			    "Failed to allocate response buffer rc: %x\n", rc);
6021 		return rc;
6022 	}
6023 	rc = bnxt_alloc_leds_info(bp);
6024 	if (rc)
6025 		return rc;
6026 
6027 	rc = bnxt_alloc_cos_queues(bp);
6028 	if (rc)
6029 		return rc;
6030 
6031 	rc = bnxt_init_locks(bp);
6032 	if (rc)
6033 		return rc;
6034 
6035 	rc = bnxt_alloc_switch_domain(bp);
6036 	if (rc)
6037 		return rc;
6038 
6039 	return rc;
6040 }
6041 
6042 static int
6043 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
6044 {
6045 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
6046 	static int version_printed;
6047 	struct bnxt *bp;
6048 	int rc;
6049 
6050 	if (version_printed++ == 0)
6051 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
6052 
6053 	eth_dev->dev_ops = &bnxt_dev_ops;
6054 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
6055 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
6056 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
6057 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
6058 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
6059 
6060 	/*
6061 	 * For secondary processes, we don't initialise any further
6062 	 * as primary has already done this work.
6063 	 */
6064 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6065 		return 0;
6066 
6067 	rte_eth_copy_pci_info(eth_dev, pci_dev);
6068 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
6069 
6070 	bp = eth_dev->data->dev_private;
6071 
6072 	/* Parse dev arguments passed on when starting the DPDK application. */
6073 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
6074 	if (rc)
6075 		goto error_free;
6076 
6077 	rc = bnxt_drv_init(eth_dev);
6078 	if (rc)
6079 		goto error_free;
6080 
6081 	rc = bnxt_init_resources(bp, false);
6082 	if (rc)
6083 		goto error_free;
6084 
6085 	rc = bnxt_alloc_stats_mem(bp);
6086 	if (rc)
6087 		goto error_free;
6088 
6089 	PMD_DRV_LOG(INFO,
6090 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
6091 		    DRV_MODULE_NAME,
6092 		    pci_dev->mem_resource[0].phys_addr,
6093 		    pci_dev->mem_resource[0].addr);
6094 
6095 	return 0;
6096 
6097 error_free:
6098 	bnxt_dev_uninit(eth_dev);
6099 	return rc;
6100 }
6101 
6102 
6103 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
6104 {
6105 	if (!ctx)
6106 		return;
6107 
6108 	if (ctx->va)
6109 		rte_free(ctx->va);
6110 
6111 	ctx->va = NULL;
6112 	ctx->dma = RTE_BAD_IOVA;
6113 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
6114 }
6115 
6116 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
6117 {
6118 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
6119 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6120 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
6121 				  bp->flow_stat->max_fc,
6122 				  false);
6123 
6124 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
6125 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6126 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
6127 				  bp->flow_stat->max_fc,
6128 				  false);
6129 
6130 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6131 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
6132 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6133 
6134 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6135 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
6136 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6137 
6138 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6139 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
6140 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6141 
6142 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6143 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
6144 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6145 }
6146 
6147 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
6148 {
6149 	bnxt_unregister_fc_ctx_mem(bp);
6150 
6151 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
6152 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
6153 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
6154 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
6155 }
6156 
6157 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
6158 {
6159 	if (BNXT_FLOW_XSTATS_EN(bp))
6160 		bnxt_uninit_fc_ctx_mem(bp);
6161 }
6162 
6163 static void
6164 bnxt_free_error_recovery_info(struct bnxt *bp)
6165 {
6166 	rte_free(bp->recovery_info);
6167 	bp->recovery_info = NULL;
6168 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
6169 }
6170 
6171 static int
6172 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
6173 {
6174 	int rc;
6175 
6176 	bnxt_free_int(bp);
6177 	bnxt_free_mem(bp, reconfig_dev);
6178 
6179 	bnxt_hwrm_func_buf_unrgtr(bp);
6180 	if (bp->pf != NULL) {
6181 		rte_free(bp->pf->vf_req_buf);
6182 		bp->pf->vf_req_buf = NULL;
6183 	}
6184 
6185 	rc = bnxt_hwrm_func_driver_unregister(bp);
6186 	bp->flags &= ~BNXT_FLAG_REGISTERED;
6187 	bnxt_free_ctx_mem(bp);
6188 	if (!reconfig_dev) {
6189 		bnxt_free_hwrm_resources(bp);
6190 		bnxt_free_error_recovery_info(bp);
6191 	}
6192 
6193 	bnxt_uninit_ctx_mem(bp);
6194 
6195 	bnxt_free_flow_stats_info(bp);
6196 	if (bp->rep_info != NULL)
6197 		bnxt_free_switch_domain(bp);
6198 	bnxt_free_rep_info(bp);
6199 	rte_free(bp->ptp_cfg);
6200 	bp->ptp_cfg = NULL;
6201 	return rc;
6202 }
6203 
6204 static int
6205 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
6206 {
6207 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6208 		return -EPERM;
6209 
6210 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
6211 
6212 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
6213 		bnxt_dev_close_op(eth_dev);
6214 
6215 	return 0;
6216 }
6217 
6218 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6219 {
6220 	struct bnxt *bp = eth_dev->data->dev_private;
6221 	struct rte_eth_dev *vf_rep_eth_dev;
6222 	int ret = 0, i;
6223 
6224 	if (!bp)
6225 		return -EINVAL;
6226 
6227 	for (i = 0; i < bp->num_reps; i++) {
6228 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6229 		if (!vf_rep_eth_dev)
6230 			continue;
6231 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6232 			    vf_rep_eth_dev->data->port_id);
6233 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6234 	}
6235 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6236 		    eth_dev->data->port_id);
6237 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6238 
6239 	return ret;
6240 }
6241 
6242 static void bnxt_free_rep_info(struct bnxt *bp)
6243 {
6244 	rte_free(bp->rep_info);
6245 	bp->rep_info = NULL;
6246 	rte_free(bp->cfa_code_map);
6247 	bp->cfa_code_map = NULL;
6248 }
6249 
6250 static int bnxt_init_rep_info(struct bnxt *bp)
6251 {
6252 	int i = 0, rc;
6253 
6254 	if (bp->rep_info)
6255 		return 0;
6256 
6257 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
6258 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
6259 				   0);
6260 	if (!bp->rep_info) {
6261 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6262 		return -ENOMEM;
6263 	}
6264 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6265 				       sizeof(*bp->cfa_code_map) *
6266 				       BNXT_MAX_CFA_CODE, 0);
6267 	if (!bp->cfa_code_map) {
6268 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6269 		bnxt_free_rep_info(bp);
6270 		return -ENOMEM;
6271 	}
6272 
6273 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6274 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6275 
6276 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6277 	if (rc) {
6278 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6279 		bnxt_free_rep_info(bp);
6280 		return rc;
6281 	}
6282 
6283 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6284 	if (rc) {
6285 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6286 		bnxt_free_rep_info(bp);
6287 		return rc;
6288 	}
6289 
6290 	return rc;
6291 }
6292 
6293 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6294 			       struct rte_eth_devargs *eth_da,
6295 			       struct rte_eth_dev *backing_eth_dev,
6296 			       const char *dev_args)
6297 {
6298 	struct rte_eth_dev *vf_rep_eth_dev;
6299 	char name[RTE_ETH_NAME_MAX_LEN];
6300 	struct bnxt *backing_bp;
6301 	uint16_t num_rep;
6302 	int i, ret = 0;
6303 	struct rte_kvargs *kvlist = NULL;
6304 
6305 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6306 		return 0;
6307 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6308 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6309 			    eth_da->type);
6310 		return -ENOTSUP;
6311 	}
6312 	num_rep = eth_da->nb_representor_ports;
6313 	if (num_rep > BNXT_MAX_VF_REPS) {
6314 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6315 			    num_rep, BNXT_MAX_VF_REPS);
6316 		return -EINVAL;
6317 	}
6318 
6319 	if (num_rep >= RTE_MAX_ETHPORTS) {
6320 		PMD_DRV_LOG(ERR,
6321 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6322 			    num_rep, RTE_MAX_ETHPORTS);
6323 		return -EINVAL;
6324 	}
6325 
6326 	backing_bp = backing_eth_dev->data->dev_private;
6327 
6328 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6329 		PMD_DRV_LOG(ERR,
6330 			    "Not a PF or trusted VF. No Representor support\n");
6331 		/* Returning an error is not an option.
6332 		 * Applications are not handling this correctly
6333 		 */
6334 		return 0;
6335 	}
6336 
6337 	if (bnxt_init_rep_info(backing_bp))
6338 		return 0;
6339 
6340 	for (i = 0; i < num_rep; i++) {
6341 		struct bnxt_representor representor = {
6342 			.vf_id = eth_da->representor_ports[i],
6343 			.switch_domain_id = backing_bp->switch_domain_id,
6344 			.parent_dev = backing_eth_dev
6345 		};
6346 
6347 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
6348 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6349 				    representor.vf_id, BNXT_MAX_VF_REPS);
6350 			continue;
6351 		}
6352 
6353 		/* representor port net_bdf_port */
6354 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6355 			 pci_dev->device.name, eth_da->representor_ports[i]);
6356 
6357 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6358 		if (kvlist) {
6359 			/*
6360 			 * Handler for "rep_is_pf" devarg.
6361 			 * Invoked as for ex: "-a 000:00:0d.0,
6362 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6363 			 */
6364 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6365 						 bnxt_parse_devarg_rep_is_pf,
6366 						 (void *)&representor);
6367 			if (ret) {
6368 				ret = -EINVAL;
6369 				goto err;
6370 			}
6371 			/*
6372 			 * Handler for "rep_based_pf" devarg.
6373 			 * Invoked as for ex: "-a 000:00:0d.0,
6374 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6375 			 */
6376 			ret = rte_kvargs_process(kvlist,
6377 						 BNXT_DEVARG_REP_BASED_PF,
6378 						 bnxt_parse_devarg_rep_based_pf,
6379 						 (void *)&representor);
6380 			if (ret) {
6381 				ret = -EINVAL;
6382 				goto err;
6383 			}
6384 			/*
6385 			 * Handler for "rep_based_pf" devarg.
6386 			 * Invoked as for ex: "-a 000:00:0d.0,
6387 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6388 			 */
6389 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6390 						 bnxt_parse_devarg_rep_q_r2f,
6391 						 (void *)&representor);
6392 			if (ret) {
6393 				ret = -EINVAL;
6394 				goto err;
6395 			}
6396 			/*
6397 			 * Handler for "rep_based_pf" devarg.
6398 			 * Invoked as for ex: "-a 000:00:0d.0,
6399 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6400 			 */
6401 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6402 						 bnxt_parse_devarg_rep_q_f2r,
6403 						 (void *)&representor);
6404 			if (ret) {
6405 				ret = -EINVAL;
6406 				goto err;
6407 			}
6408 			/*
6409 			 * Handler for "rep_based_pf" devarg.
6410 			 * Invoked as for ex: "-a 000:00:0d.0,
6411 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6412 			 */
6413 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6414 						 bnxt_parse_devarg_rep_fc_r2f,
6415 						 (void *)&representor);
6416 			if (ret) {
6417 				ret = -EINVAL;
6418 				goto err;
6419 			}
6420 			/*
6421 			 * Handler for "rep_based_pf" devarg.
6422 			 * Invoked as for ex: "-a 000:00:0d.0,
6423 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6424 			 */
6425 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6426 						 bnxt_parse_devarg_rep_fc_f2r,
6427 						 (void *)&representor);
6428 			if (ret) {
6429 				ret = -EINVAL;
6430 				goto err;
6431 			}
6432 		}
6433 
6434 		ret = rte_eth_dev_create(&pci_dev->device, name,
6435 					 sizeof(struct bnxt_representor),
6436 					 NULL, NULL,
6437 					 bnxt_representor_init,
6438 					 &representor);
6439 		if (ret) {
6440 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6441 				    "representor %s.", name);
6442 			goto err;
6443 		}
6444 
6445 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6446 		if (!vf_rep_eth_dev) {
6447 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6448 				    " for VF-Rep: %s.", name);
6449 			ret = -ENODEV;
6450 			goto err;
6451 		}
6452 
6453 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6454 			    backing_eth_dev->data->port_id);
6455 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6456 							 vf_rep_eth_dev;
6457 		backing_bp->num_reps++;
6458 
6459 	}
6460 
6461 	rte_kvargs_free(kvlist);
6462 	return 0;
6463 
6464 err:
6465 	/* If num_rep > 1, then rollback already created
6466 	 * ports, since we'll be failing the probe anyway
6467 	 */
6468 	if (num_rep > 1)
6469 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6470 	rte_errno = -ret;
6471 	rte_kvargs_free(kvlist);
6472 
6473 	return ret;
6474 }
6475 
6476 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6477 			  struct rte_pci_device *pci_dev)
6478 {
6479 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6480 	struct rte_eth_dev *backing_eth_dev;
6481 	uint16_t num_rep;
6482 	int ret = 0;
6483 
6484 	if (pci_dev->device.devargs) {
6485 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6486 					    &eth_da);
6487 		if (ret)
6488 			return ret;
6489 	}
6490 
6491 	num_rep = eth_da.nb_representor_ports;
6492 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6493 		    num_rep);
6494 
6495 	/* We could come here after first level of probe is already invoked
6496 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6497 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6498 	 */
6499 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6500 	if (backing_eth_dev == NULL) {
6501 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6502 					 sizeof(struct bnxt),
6503 					 eth_dev_pci_specific_init, pci_dev,
6504 					 bnxt_dev_init, NULL);
6505 
6506 		if (ret || !num_rep)
6507 			return ret;
6508 
6509 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6510 	}
6511 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6512 		    backing_eth_dev->data->port_id);
6513 
6514 	if (!num_rep)
6515 		return ret;
6516 
6517 	/* probe representor ports now */
6518 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6519 				  pci_dev->device.devargs->args);
6520 
6521 	return ret;
6522 }
6523 
6524 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6525 {
6526 	struct rte_eth_dev *eth_dev;
6527 
6528 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6529 	if (!eth_dev)
6530 		return 0; /* Invoked typically only by OVS-DPDK, by the
6531 			   * time it comes here the eth_dev is already
6532 			   * deleted by rte_eth_dev_close(), so returning
6533 			   * +ve value will at least help in proper cleanup
6534 			   */
6535 
6536 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6537 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6538 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6539 			return rte_eth_dev_destroy(eth_dev,
6540 						   bnxt_representor_uninit);
6541 		else
6542 			return rte_eth_dev_destroy(eth_dev,
6543 						   bnxt_dev_uninit);
6544 	} else {
6545 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6546 	}
6547 }
6548 
6549 static struct rte_pci_driver bnxt_rte_pmd = {
6550 	.id_table = bnxt_pci_id_map,
6551 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6552 			RTE_PCI_DRV_INTR_RMV |
6553 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6554 						  * and OVS-DPDK
6555 						  */
6556 	.probe = bnxt_pci_probe,
6557 	.remove = bnxt_pci_remove,
6558 };
6559 
6560 static bool
6561 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6562 {
6563 	if (strcmp(dev->device->driver->name, drv->driver.name))
6564 		return false;
6565 
6566 	return true;
6567 }
6568 
6569 bool is_bnxt_supported(struct rte_eth_dev *dev)
6570 {
6571 	return is_device_supported(dev, &bnxt_rte_pmd);
6572 }
6573 
6574 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6575 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6576 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6577 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6578