xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision 49c19c94)
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 		if (!rxq->rx_deferred_start) {
487 			bp->eth_dev->data->rx_queue_state[j] =
488 				RTE_ETH_QUEUE_STATE_STARTED;
489 			rxq->rx_started = true;
490 		}
491 	}
492 
493 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
494 
495 	rc = bnxt_vnic_rss_configure(bp, vnic);
496 	if (rc)
497 		goto err_out;
498 
499 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
500 
501 	rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
502 				    (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO) ?
503 				    true : false);
504 	if (rc)
505 		goto err_out;
506 
507 	return 0;
508 err_out:
509 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
510 		    vnic_id, rc);
511 	return rc;
512 }
513 
514 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
515 {
516 	int rc = 0;
517 
518 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
519 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
520 	if (rc)
521 		return rc;
522 
523 	PMD_DRV_LOG(DEBUG,
524 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
525 		    " rx_fc_in_tbl.ctx_id = %d\n",
526 		    bp->flow_stat->rx_fc_in_tbl.va,
527 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
528 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
529 
530 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
531 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
532 	if (rc)
533 		return rc;
534 
535 	PMD_DRV_LOG(DEBUG,
536 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
537 		    " rx_fc_out_tbl.ctx_id = %d\n",
538 		    bp->flow_stat->rx_fc_out_tbl.va,
539 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
540 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
541 
542 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
543 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
544 	if (rc)
545 		return rc;
546 
547 	PMD_DRV_LOG(DEBUG,
548 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
549 		    " tx_fc_in_tbl.ctx_id = %d\n",
550 		    bp->flow_stat->tx_fc_in_tbl.va,
551 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
552 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
553 
554 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
555 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
556 	if (rc)
557 		return rc;
558 
559 	PMD_DRV_LOG(DEBUG,
560 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
561 		    " tx_fc_out_tbl.ctx_id = %d\n",
562 		    bp->flow_stat->tx_fc_out_tbl.va,
563 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
564 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
565 
566 	memset(bp->flow_stat->rx_fc_out_tbl.va,
567 	       0,
568 	       bp->flow_stat->rx_fc_out_tbl.size);
569 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
570 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
571 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
572 				       bp->flow_stat->max_fc,
573 				       true);
574 	if (rc)
575 		return rc;
576 
577 	memset(bp->flow_stat->tx_fc_out_tbl.va,
578 	       0,
579 	       bp->flow_stat->tx_fc_out_tbl.size);
580 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
581 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
582 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
583 				       bp->flow_stat->max_fc,
584 				       true);
585 
586 	return rc;
587 }
588 
589 static int bnxt_alloc_ctx_mem_buf(struct bnxt *bp, char *type, size_t size,
590 				  struct bnxt_ctx_mem_buf_info *ctx)
591 {
592 	if (!ctx)
593 		return -EINVAL;
594 
595 	ctx->va = rte_zmalloc_socket(type, size, 0,
596 				     bp->eth_dev->device->numa_node);
597 	if (ctx->va == NULL)
598 		return -ENOMEM;
599 	rte_mem_lock_page(ctx->va);
600 	ctx->size = size;
601 	ctx->dma = rte_mem_virt2iova(ctx->va);
602 	if (ctx->dma == RTE_BAD_IOVA)
603 		return -ENOMEM;
604 
605 	return 0;
606 }
607 
608 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
609 {
610 	struct rte_pci_device *pdev = bp->pdev;
611 	char type[RTE_MEMZONE_NAMESIZE];
612 	uint16_t max_fc;
613 	int rc = 0;
614 
615 	max_fc = bp->flow_stat->max_fc;
616 
617 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
618 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
619 	/* 4 bytes for each counter-id */
620 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
621 				    max_fc * 4,
622 				    &bp->flow_stat->rx_fc_in_tbl);
623 	if (rc)
624 		return rc;
625 
626 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
627 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
628 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
629 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
630 				    max_fc * 16,
631 				    &bp->flow_stat->rx_fc_out_tbl);
632 	if (rc)
633 		return rc;
634 
635 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
636 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
637 	/* 4 bytes for each counter-id */
638 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
639 				    max_fc * 4,
640 				    &bp->flow_stat->tx_fc_in_tbl);
641 	if (rc)
642 		return rc;
643 
644 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
645 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
646 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
647 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
648 				    max_fc * 16,
649 				    &bp->flow_stat->tx_fc_out_tbl);
650 	if (rc)
651 		return rc;
652 
653 	rc = bnxt_register_fc_ctx_mem(bp);
654 
655 	return rc;
656 }
657 
658 static int bnxt_init_ctx_mem(struct bnxt *bp)
659 {
660 	int rc = 0;
661 
662 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
663 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
664 	    !BNXT_FLOW_XSTATS_EN(bp))
665 		return 0;
666 
667 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
668 	if (rc)
669 		return rc;
670 
671 	rc = bnxt_init_fc_ctx_mem(bp);
672 
673 	return rc;
674 }
675 
676 static int bnxt_update_phy_setting(struct bnxt *bp)
677 {
678 	struct rte_eth_link new;
679 	int rc;
680 
681 	rc = bnxt_get_hwrm_link_config(bp, &new);
682 	if (rc) {
683 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
684 		return rc;
685 	}
686 
687 	/*
688 	 * On BCM957508-N2100 adapters, FW will not allow any user other
689 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
690 	 * always returns link up. Force phy update always in that case.
691 	 */
692 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
693 		rc = bnxt_set_hwrm_link_config(bp, true);
694 		if (rc) {
695 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
696 			return rc;
697 		}
698 	}
699 
700 	return rc;
701 }
702 
703 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
704 {
705 	rte_free(bp->prev_rx_ring_stats);
706 	rte_free(bp->prev_tx_ring_stats);
707 
708 	bp->prev_rx_ring_stats = NULL;
709 	bp->prev_tx_ring_stats = NULL;
710 }
711 
712 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
713 {
714 	bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
715 					      sizeof(struct bnxt_ring_stats) *
716 					      bp->rx_cp_nr_rings,
717 					      0);
718 	if (bp->prev_rx_ring_stats == NULL)
719 		return -ENOMEM;
720 
721 	bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
722 					     sizeof(struct bnxt_ring_stats) *
723 					     bp->tx_cp_nr_rings,
724 					     0);
725 	if (bp->prev_tx_ring_stats == NULL)
726 		goto error;
727 
728 	return 0;
729 
730 error:
731 	bnxt_free_prev_ring_stats(bp);
732 	return -ENOMEM;
733 }
734 
735 static int bnxt_start_nic(struct bnxt *bp)
736 {
737 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
738 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
739 	uint32_t intr_vector = 0;
740 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
741 	uint32_t vec = BNXT_MISC_VEC_ID;
742 	unsigned int i, j;
743 	int rc;
744 
745 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
746 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
747 			DEV_RX_OFFLOAD_JUMBO_FRAME;
748 		bp->flags |= BNXT_FLAG_JUMBO;
749 	} else {
750 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
751 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
752 		bp->flags &= ~BNXT_FLAG_JUMBO;
753 	}
754 
755 	/* THOR does not support ring groups.
756 	 * But we will use the array to save RSS context IDs.
757 	 */
758 	if (BNXT_CHIP_P5(bp))
759 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
760 
761 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
762 	if (rc) {
763 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
764 		goto err_out;
765 	}
766 
767 	rc = bnxt_alloc_hwrm_rings(bp);
768 	if (rc) {
769 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
770 		goto err_out;
771 	}
772 
773 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
774 	if (rc) {
775 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
776 		goto err_out;
777 	}
778 
779 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
780 		goto skip_cosq_cfg;
781 
782 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
783 		if (bp->rx_cos_queue[i].id != 0xff) {
784 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
785 
786 			if (!vnic) {
787 				PMD_DRV_LOG(ERR,
788 					    "Num pools more than FW profile\n");
789 				rc = -EINVAL;
790 				goto err_out;
791 			}
792 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
793 			bp->rx_cosq_cnt++;
794 		}
795 	}
796 
797 skip_cosq_cfg:
798 	rc = bnxt_mq_rx_configure(bp);
799 	if (rc) {
800 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
801 		goto err_out;
802 	}
803 
804 	/* default vnic 0 */
805 	rc = bnxt_setup_one_vnic(bp, 0);
806 	if (rc)
807 		goto err_out;
808 	/* VNIC configuration */
809 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
810 		for (i = 1; i < bp->nr_vnics; i++) {
811 			rc = bnxt_setup_one_vnic(bp, i);
812 			if (rc)
813 				goto err_out;
814 		}
815 	}
816 
817 	for (j = 0; j < bp->tx_nr_rings; j++) {
818 		struct bnxt_tx_queue *txq = bp->tx_queues[j];
819 
820 		if (!txq->tx_deferred_start) {
821 			bp->eth_dev->data->tx_queue_state[j] =
822 				RTE_ETH_QUEUE_STATE_STARTED;
823 			txq->tx_started = true;
824 		}
825 	}
826 
827 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
828 	if (rc) {
829 		PMD_DRV_LOG(ERR,
830 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
831 		goto err_out;
832 	}
833 
834 	/* check and configure queue intr-vector mapping */
835 	if ((rte_intr_cap_multiple(intr_handle) ||
836 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
837 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
838 		intr_vector = bp->eth_dev->data->nb_rx_queues;
839 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
840 		if (intr_vector > bp->rx_cp_nr_rings) {
841 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
842 					bp->rx_cp_nr_rings);
843 			return -ENOTSUP;
844 		}
845 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
846 		if (rc)
847 			return rc;
848 	}
849 
850 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
851 		intr_handle->intr_vec =
852 			rte_zmalloc("intr_vec",
853 				    bp->eth_dev->data->nb_rx_queues *
854 				    sizeof(int), 0);
855 		if (intr_handle->intr_vec == NULL) {
856 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
857 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
858 			rc = -ENOMEM;
859 			goto err_out;
860 		}
861 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
862 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
863 			 intr_handle->intr_vec, intr_handle->nb_efd,
864 			intr_handle->max_intr);
865 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
866 		     queue_id++) {
867 			intr_handle->intr_vec[queue_id] =
868 							vec + BNXT_RX_VEC_START;
869 			if (vec < base + intr_handle->nb_efd - 1)
870 				vec++;
871 		}
872 	}
873 
874 	/* enable uio/vfio intr/eventfd mapping */
875 	rc = rte_intr_enable(intr_handle);
876 #ifndef RTE_EXEC_ENV_FREEBSD
877 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
878 	if (rc)
879 		goto err_out;
880 #endif
881 
882 	rc = bnxt_update_phy_setting(bp);
883 	if (rc)
884 		goto err_out;
885 
886 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
887 	if (!bp->mark_table)
888 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
889 
890 	return 0;
891 
892 err_out:
893 	/* Some of the error status returned by FW may not be from errno.h */
894 	if (rc > 0)
895 		rc = -EIO;
896 
897 	return rc;
898 }
899 
900 static int bnxt_shutdown_nic(struct bnxt *bp)
901 {
902 	bnxt_free_all_hwrm_resources(bp);
903 	bnxt_free_all_filters(bp);
904 	bnxt_free_all_vnics(bp);
905 	return 0;
906 }
907 
908 /*
909  * Device configuration and status function
910  */
911 
912 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
913 {
914 	uint32_t link_speed = 0;
915 	uint32_t speed_capa = 0;
916 
917 	if (bp->link_info == NULL)
918 		return 0;
919 
920 	link_speed = bp->link_info->support_speeds;
921 
922 	/* If PAM4 is configured, use PAM4 supported speed */
923 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
924 		link_speed = bp->link_info->support_pam4_speeds;
925 
926 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
927 		speed_capa |= ETH_LINK_SPEED_100M;
928 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
929 		speed_capa |= ETH_LINK_SPEED_100M_HD;
930 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
931 		speed_capa |= ETH_LINK_SPEED_1G;
932 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
933 		speed_capa |= ETH_LINK_SPEED_2_5G;
934 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
935 		speed_capa |= ETH_LINK_SPEED_10G;
936 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
937 		speed_capa |= ETH_LINK_SPEED_20G;
938 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
939 		speed_capa |= ETH_LINK_SPEED_25G;
940 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
941 		speed_capa |= ETH_LINK_SPEED_40G;
942 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
943 		speed_capa |= ETH_LINK_SPEED_50G;
944 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
945 		speed_capa |= ETH_LINK_SPEED_100G;
946 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
947 		speed_capa |= ETH_LINK_SPEED_50G;
948 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
949 		speed_capa |= ETH_LINK_SPEED_100G;
950 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
951 		speed_capa |= ETH_LINK_SPEED_200G;
952 
953 	if (bp->link_info->auto_mode ==
954 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
955 		speed_capa |= ETH_LINK_SPEED_FIXED;
956 
957 	return speed_capa;
958 }
959 
960 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
961 				struct rte_eth_dev_info *dev_info)
962 {
963 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
964 	struct bnxt *bp = eth_dev->data->dev_private;
965 	uint16_t max_vnics, i, j, vpool, vrxq;
966 	unsigned int max_rx_rings;
967 	int rc;
968 
969 	rc = is_bnxt_in_error(bp);
970 	if (rc)
971 		return rc;
972 
973 	/* MAC Specifics */
974 	dev_info->max_mac_addrs = bp->max_l2_ctx;
975 	dev_info->max_hash_mac_addrs = 0;
976 
977 	/* PF/VF specifics */
978 	if (BNXT_PF(bp))
979 		dev_info->max_vfs = pdev->max_vfs;
980 
981 	max_rx_rings = bnxt_max_rings(bp);
982 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
983 	dev_info->max_rx_queues = max_rx_rings;
984 	dev_info->max_tx_queues = max_rx_rings;
985 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
986 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
987 	max_vnics = bp->max_vnics;
988 
989 	/* MTU specifics */
990 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
991 	dev_info->max_mtu = BNXT_MAX_MTU;
992 
993 	/* Fast path specifics */
994 	dev_info->min_rx_bufsize = 1;
995 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
996 
997 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
998 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
999 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
1000 	if (bp->vnic_cap_flags & BNXT_VNIC_CAP_VLAN_RX_STRIP)
1001 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_VLAN_STRIP;
1002 	dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
1003 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
1004 				    dev_info->tx_queue_offload_capa;
1005 	if (bp->fw_cap & BNXT_FW_CAP_VLAN_TX_INSERT)
1006 		dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_VLAN_INSERT;
1007 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
1008 
1009 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
1010 	dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
1011 			     RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
1012 
1013 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
1014 		.rx_thresh = {
1015 			.pthresh = 8,
1016 			.hthresh = 8,
1017 			.wthresh = 0,
1018 		},
1019 		.rx_free_thresh = 32,
1020 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
1021 	};
1022 
1023 	dev_info->default_txconf = (struct rte_eth_txconf) {
1024 		.tx_thresh = {
1025 			.pthresh = 32,
1026 			.hthresh = 0,
1027 			.wthresh = 0,
1028 		},
1029 		.tx_free_thresh = 32,
1030 		.tx_rs_thresh = 32,
1031 	};
1032 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
1033 
1034 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1035 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1036 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1037 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1038 
1039 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1040 		dev_info->switch_info.name = eth_dev->device->name;
1041 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1042 		dev_info->switch_info.port_id =
1043 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1044 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1045 	}
1046 
1047 	/*
1048 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1049 	 *       need further investigation.
1050 	 */
1051 
1052 	/* VMDq resources */
1053 	vpool = 64; /* ETH_64_POOLS */
1054 	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
1055 	for (i = 0; i < 4; vpool >>= 1, i++) {
1056 		if (max_vnics > vpool) {
1057 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1058 				if (dev_info->max_rx_queues > vrxq) {
1059 					if (vpool > vrxq)
1060 						vpool = vrxq;
1061 					goto found;
1062 				}
1063 			}
1064 			/* Not enough resources to support VMDq */
1065 			break;
1066 		}
1067 	}
1068 	/* Not enough resources to support VMDq */
1069 	vpool = 0;
1070 	vrxq = 0;
1071 found:
1072 	dev_info->max_vmdq_pools = vpool;
1073 	dev_info->vmdq_queue_num = vrxq;
1074 
1075 	dev_info->vmdq_pool_base = 0;
1076 	dev_info->vmdq_queue_base = 0;
1077 
1078 	return 0;
1079 }
1080 
1081 /* Configure the device based on the configuration provided */
1082 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1083 {
1084 	struct bnxt *bp = eth_dev->data->dev_private;
1085 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1086 	int rc;
1087 
1088 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1089 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1090 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1091 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1092 
1093 	rc = is_bnxt_in_error(bp);
1094 	if (rc)
1095 		return rc;
1096 
1097 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1098 		rc = bnxt_hwrm_check_vf_rings(bp);
1099 		if (rc) {
1100 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1101 			return -ENOSPC;
1102 		}
1103 
1104 		/* If a resource has already been allocated - in this case
1105 		 * it is the async completion ring, free it. Reallocate it after
1106 		 * resource reservation. This will ensure the resource counts
1107 		 * are calculated correctly.
1108 		 */
1109 
1110 		pthread_mutex_lock(&bp->def_cp_lock);
1111 
1112 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1113 			bnxt_disable_int(bp);
1114 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1115 		}
1116 
1117 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1118 		if (rc) {
1119 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1120 			pthread_mutex_unlock(&bp->def_cp_lock);
1121 			return -ENOSPC;
1122 		}
1123 
1124 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1125 			rc = bnxt_alloc_async_cp_ring(bp);
1126 			if (rc) {
1127 				pthread_mutex_unlock(&bp->def_cp_lock);
1128 				return rc;
1129 			}
1130 			bnxt_enable_int(bp);
1131 		}
1132 
1133 		pthread_mutex_unlock(&bp->def_cp_lock);
1134 	}
1135 
1136 	/* Inherit new configurations */
1137 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1138 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1139 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1140 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1141 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1142 	    bp->max_stat_ctx)
1143 		goto resource_error;
1144 
1145 	if (BNXT_HAS_RING_GRPS(bp) &&
1146 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1147 		goto resource_error;
1148 
1149 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1150 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1151 		goto resource_error;
1152 
1153 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1154 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1155 
1156 	if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1157 		rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1158 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1159 
1160 	if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1161 		eth_dev->data->mtu =
1162 			eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1163 			RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1164 			BNXT_NUM_VLANS;
1165 		bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1166 	}
1167 	return 0;
1168 
1169 resource_error:
1170 	PMD_DRV_LOG(ERR,
1171 		    "Insufficient resources to support requested config\n");
1172 	PMD_DRV_LOG(ERR,
1173 		    "Num Queues Requested: Tx %d, Rx %d\n",
1174 		    eth_dev->data->nb_tx_queues,
1175 		    eth_dev->data->nb_rx_queues);
1176 	PMD_DRV_LOG(ERR,
1177 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1178 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1179 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1180 	return -ENOSPC;
1181 }
1182 
1183 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1184 {
1185 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1186 
1187 	if (link->link_status)
1188 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1189 			eth_dev->data->port_id,
1190 			(uint32_t)link->link_speed,
1191 			(link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1192 			("full-duplex") : ("half-duplex\n"));
1193 	else
1194 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1195 			eth_dev->data->port_id);
1196 }
1197 
1198 /*
1199  * Determine whether the current configuration requires support for scattered
1200  * receive; return 1 if scattered receive is required and 0 if not.
1201  */
1202 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1203 {
1204 	uint16_t buf_size;
1205 	int i;
1206 
1207 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1208 		return 1;
1209 
1210 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1211 		return 1;
1212 
1213 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1214 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1215 
1216 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1217 				      RTE_PKTMBUF_HEADROOM);
1218 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1219 			return 1;
1220 	}
1221 	return 0;
1222 }
1223 
1224 static eth_rx_burst_t
1225 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1226 {
1227 	struct bnxt *bp = eth_dev->data->dev_private;
1228 
1229 	/* Disable vector mode RX for Stingray2 for now */
1230 	if (BNXT_CHIP_SR2(bp)) {
1231 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1232 		return bnxt_recv_pkts;
1233 	}
1234 
1235 #if (defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)) && \
1236 	!defined(RTE_LIBRTE_IEEE1588)
1237 
1238 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1239 	if (eth_dev->data->scattered_rx)
1240 		goto use_scalar_rx;
1241 
1242 	/*
1243 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1244 	 * asynchronous completions and receive completions can be placed in
1245 	 * the same completion ring.
1246 	 */
1247 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1248 		goto use_scalar_rx;
1249 
1250 	/*
1251 	 * Vector mode receive cannot be enabled if any receive offloads outside
1252 	 * a limited subset have been enabled.
1253 	 */
1254 	if (eth_dev->data->dev_conf.rxmode.offloads &
1255 		~(DEV_RX_OFFLOAD_VLAN_STRIP |
1256 		  DEV_RX_OFFLOAD_KEEP_CRC |
1257 		  DEV_RX_OFFLOAD_JUMBO_FRAME |
1258 		  DEV_RX_OFFLOAD_IPV4_CKSUM |
1259 		  DEV_RX_OFFLOAD_UDP_CKSUM |
1260 		  DEV_RX_OFFLOAD_TCP_CKSUM |
1261 		  DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1262 		  DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1263 		  DEV_RX_OFFLOAD_RSS_HASH |
1264 		  DEV_RX_OFFLOAD_VLAN_FILTER))
1265 		goto use_scalar_rx;
1266 
1267 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1268 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1269 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1270 		PMD_DRV_LOG(INFO,
1271 			    "Using AVX2 vector mode receive for port %d\n",
1272 			    eth_dev->data->port_id);
1273 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1274 		return bnxt_recv_pkts_vec_avx2;
1275 	}
1276  #endif
1277 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1278 		PMD_DRV_LOG(INFO,
1279 			    "Using SSE vector mode receive for port %d\n",
1280 			    eth_dev->data->port_id);
1281 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1282 		return bnxt_recv_pkts_vec;
1283 	}
1284 
1285 use_scalar_rx:
1286 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1287 		    eth_dev->data->port_id);
1288 	PMD_DRV_LOG(INFO,
1289 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1290 		    eth_dev->data->port_id,
1291 		    eth_dev->data->scattered_rx,
1292 		    eth_dev->data->dev_conf.rxmode.offloads);
1293 #endif
1294 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1295 	return bnxt_recv_pkts;
1296 }
1297 
1298 static eth_tx_burst_t
1299 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1300 {
1301 	struct bnxt *bp = eth_dev->data->dev_private;
1302 
1303 	/* Disable vector mode TX for Stingray2 for now */
1304 	if (BNXT_CHIP_SR2(bp))
1305 		return bnxt_xmit_pkts;
1306 
1307 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64) && \
1308 	!defined(RTE_LIBRTE_IEEE1588)
1309 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1310 
1311 	/*
1312 	 * Vector mode transmit can be enabled only if not using scatter rx
1313 	 * or tx offloads.
1314 	 */
1315 	if (eth_dev->data->scattered_rx ||
1316 	    (offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) ||
1317 	    BNXT_TRUFLOW_EN(bp))
1318 		goto use_scalar_tx;
1319 
1320 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1321 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1322 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1323 		PMD_DRV_LOG(INFO,
1324 			    "Using AVX2 vector mode transmit for port %d\n",
1325 			    eth_dev->data->port_id);
1326 		return bnxt_xmit_pkts_vec_avx2;
1327 	}
1328 #endif
1329 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1330 		PMD_DRV_LOG(INFO,
1331 			    "Using SSE vector mode transmit for port %d\n",
1332 			    eth_dev->data->port_id);
1333 		return bnxt_xmit_pkts_vec;
1334 	}
1335 
1336 use_scalar_tx:
1337 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1338 		    eth_dev->data->port_id);
1339 	PMD_DRV_LOG(INFO,
1340 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1341 		    eth_dev->data->port_id,
1342 		    eth_dev->data->scattered_rx,
1343 		    offloads);
1344 #endif
1345 	return bnxt_xmit_pkts;
1346 }
1347 
1348 static int bnxt_handle_if_change_status(struct bnxt *bp)
1349 {
1350 	int rc;
1351 
1352 	/* Since fw has undergone a reset and lost all contexts,
1353 	 * set fatal flag to not issue hwrm during cleanup
1354 	 */
1355 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1356 	bnxt_uninit_resources(bp, true);
1357 
1358 	/* clear fatal flag so that re-init happens */
1359 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1360 	rc = bnxt_init_resources(bp, true);
1361 
1362 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1363 
1364 	return rc;
1365 }
1366 
1367 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1368 {
1369 	struct bnxt *bp = eth_dev->data->dev_private;
1370 	int rc = 0;
1371 
1372 	if (!BNXT_SINGLE_PF(bp))
1373 		return -ENOTSUP;
1374 
1375 	if (!bp->link_info->link_up)
1376 		rc = bnxt_set_hwrm_link_config(bp, true);
1377 	if (!rc)
1378 		eth_dev->data->dev_link.link_status = 1;
1379 
1380 	bnxt_print_link_info(eth_dev);
1381 	return rc;
1382 }
1383 
1384 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1385 {
1386 	struct bnxt *bp = eth_dev->data->dev_private;
1387 
1388 	if (!BNXT_SINGLE_PF(bp))
1389 		return -ENOTSUP;
1390 
1391 	eth_dev->data->dev_link.link_status = 0;
1392 	bnxt_set_hwrm_link_config(bp, false);
1393 	bp->link_info->link_up = 0;
1394 
1395 	return 0;
1396 }
1397 
1398 static void bnxt_free_switch_domain(struct bnxt *bp)
1399 {
1400 	int rc = 0;
1401 
1402 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1403 		return;
1404 
1405 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1406 	if (rc)
1407 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1408 			    bp->switch_domain_id, rc);
1409 }
1410 
1411 static void bnxt_ptp_get_current_time(void *arg)
1412 {
1413 	struct bnxt *bp = arg;
1414 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1415 	int rc;
1416 
1417 	rc = is_bnxt_in_error(bp);
1418 	if (rc)
1419 		return;
1420 
1421 	if (!ptp)
1422 		return;
1423 
1424 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1425 				&ptp->current_time);
1426 
1427 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1428 	if (rc != 0) {
1429 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1430 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1431 	}
1432 }
1433 
1434 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1435 {
1436 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1437 	int rc;
1438 
1439 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1440 		return 0;
1441 
1442 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1443 				&ptp->current_time);
1444 
1445 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1446 	return rc;
1447 }
1448 
1449 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1450 {
1451 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1452 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1453 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1454 	}
1455 }
1456 
1457 static void bnxt_ptp_stop(struct bnxt *bp)
1458 {
1459 	bnxt_cancel_ptp_alarm(bp);
1460 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1461 }
1462 
1463 static int bnxt_ptp_start(struct bnxt *bp)
1464 {
1465 	int rc;
1466 
1467 	rc = bnxt_schedule_ptp_alarm(bp);
1468 	if (rc != 0) {
1469 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1470 	} else {
1471 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1472 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1473 	}
1474 
1475 	return rc;
1476 }
1477 
1478 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1479 {
1480 	struct bnxt *bp = eth_dev->data->dev_private;
1481 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1482 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1483 	struct rte_eth_link link;
1484 	int ret;
1485 
1486 	eth_dev->data->dev_started = 0;
1487 	eth_dev->data->scattered_rx = 0;
1488 
1489 	/* Prevent crashes when queues are still in use */
1490 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1491 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1492 
1493 	bnxt_disable_int(bp);
1494 
1495 	/* disable uio/vfio intr/eventfd mapping */
1496 	rte_intr_disable(intr_handle);
1497 
1498 	/* Stop the child representors for this device */
1499 	ret = bnxt_rep_stop_all(bp);
1500 	if (ret != 0)
1501 		return ret;
1502 
1503 	/* delete the bnxt ULP port details */
1504 	bnxt_ulp_port_deinit(bp);
1505 
1506 	bnxt_cancel_fw_health_check(bp);
1507 
1508 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1509 		bnxt_cancel_ptp_alarm(bp);
1510 
1511 	/* Do not bring link down during reset recovery */
1512 	if (!is_bnxt_in_error(bp)) {
1513 		bnxt_dev_set_link_down_op(eth_dev);
1514 		/* Wait for link to be reset */
1515 		if (BNXT_SINGLE_PF(bp))
1516 			rte_delay_ms(500);
1517 		/* clear the recorded link status */
1518 		memset(&link, 0, sizeof(link));
1519 		rte_eth_linkstatus_set(eth_dev, &link);
1520 	}
1521 
1522 	/* Clean queue intr-vector mapping */
1523 	rte_intr_efd_disable(intr_handle);
1524 	if (intr_handle->intr_vec != NULL) {
1525 		rte_free(intr_handle->intr_vec);
1526 		intr_handle->intr_vec = NULL;
1527 	}
1528 
1529 	bnxt_hwrm_port_clr_stats(bp);
1530 	bnxt_free_tx_mbufs(bp);
1531 	bnxt_free_rx_mbufs(bp);
1532 	/* Process any remaining notifications in default completion queue */
1533 	bnxt_int_handler(eth_dev);
1534 	bnxt_shutdown_nic(bp);
1535 	bnxt_hwrm_if_change(bp, false);
1536 
1537 	bnxt_free_prev_ring_stats(bp);
1538 	rte_free(bp->mark_table);
1539 	bp->mark_table = NULL;
1540 
1541 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1542 	bp->rx_cosq_cnt = 0;
1543 	/* All filters are deleted on a port stop. */
1544 	if (BNXT_FLOW_XSTATS_EN(bp))
1545 		bp->flow_stat->flow_count = 0;
1546 
1547 	return 0;
1548 }
1549 
1550 /* Unload the driver, release resources */
1551 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1552 {
1553 	struct bnxt *bp = eth_dev->data->dev_private;
1554 
1555 	pthread_mutex_lock(&bp->err_recovery_lock);
1556 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1557 		PMD_DRV_LOG(ERR,
1558 			    "Adapter recovering from error..Please retry\n");
1559 		pthread_mutex_unlock(&bp->err_recovery_lock);
1560 		return -EAGAIN;
1561 	}
1562 	pthread_mutex_unlock(&bp->err_recovery_lock);
1563 
1564 	return bnxt_dev_stop(eth_dev);
1565 }
1566 
1567 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1568 {
1569 	struct bnxt *bp = eth_dev->data->dev_private;
1570 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1571 	int vlan_mask = 0;
1572 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1573 
1574 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1575 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1576 		return -EINVAL;
1577 	}
1578 
1579 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1580 		PMD_DRV_LOG(ERR,
1581 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1582 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1583 
1584 	do {
1585 		rc = bnxt_hwrm_if_change(bp, true);
1586 		if (rc == 0 || rc != -EAGAIN)
1587 			break;
1588 
1589 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1590 	} while (retry_cnt--);
1591 
1592 	if (rc)
1593 		return rc;
1594 
1595 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1596 		rc = bnxt_handle_if_change_status(bp);
1597 		if (rc)
1598 			return rc;
1599 	}
1600 
1601 	bnxt_enable_int(bp);
1602 
1603 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1604 
1605 	rc = bnxt_start_nic(bp);
1606 	if (rc)
1607 		goto error;
1608 
1609 	rc = bnxt_alloc_prev_ring_stats(bp);
1610 	if (rc)
1611 		goto error;
1612 
1613 	eth_dev->data->dev_started = 1;
1614 
1615 	bnxt_link_update_op(eth_dev, 1);
1616 
1617 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1618 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1619 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1620 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1621 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1622 	if (rc)
1623 		goto error;
1624 
1625 	/* Initialize bnxt ULP port details */
1626 	rc = bnxt_ulp_port_init(bp);
1627 	if (rc)
1628 		goto error;
1629 
1630 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1631 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1632 
1633 	bnxt_schedule_fw_health_check(bp);
1634 
1635 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1636 		bnxt_schedule_ptp_alarm(bp);
1637 
1638 	return 0;
1639 
1640 error:
1641 	bnxt_dev_stop(eth_dev);
1642 	return rc;
1643 }
1644 
1645 static void
1646 bnxt_uninit_locks(struct bnxt *bp)
1647 {
1648 	pthread_mutex_destroy(&bp->flow_lock);
1649 	pthread_mutex_destroy(&bp->def_cp_lock);
1650 	pthread_mutex_destroy(&bp->health_check_lock);
1651 	pthread_mutex_destroy(&bp->err_recovery_lock);
1652 	if (bp->rep_info) {
1653 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1654 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1655 	}
1656 }
1657 
1658 static void bnxt_drv_uninit(struct bnxt *bp)
1659 {
1660 	bnxt_free_leds_info(bp);
1661 	bnxt_free_cos_queues(bp);
1662 	bnxt_free_link_info(bp);
1663 	bnxt_free_parent_info(bp);
1664 	bnxt_uninit_locks(bp);
1665 
1666 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1667 	bp->tx_mem_zone = NULL;
1668 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1669 	bp->rx_mem_zone = NULL;
1670 
1671 	bnxt_free_vf_info(bp);
1672 	bnxt_free_pf_info(bp);
1673 
1674 	rte_free(bp->grp_info);
1675 	bp->grp_info = NULL;
1676 }
1677 
1678 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1679 {
1680 	struct bnxt *bp = eth_dev->data->dev_private;
1681 	int ret = 0;
1682 
1683 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1684 		return 0;
1685 
1686 	pthread_mutex_lock(&bp->err_recovery_lock);
1687 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1688 		PMD_DRV_LOG(ERR,
1689 			    "Adapter recovering from error...Please retry\n");
1690 		pthread_mutex_unlock(&bp->err_recovery_lock);
1691 		return -EAGAIN;
1692 	}
1693 	pthread_mutex_unlock(&bp->err_recovery_lock);
1694 
1695 	/* cancel the recovery handler before remove dev */
1696 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1697 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1698 	bnxt_cancel_fc_thread(bp);
1699 
1700 	if (eth_dev->data->dev_started)
1701 		ret = bnxt_dev_stop(eth_dev);
1702 
1703 	bnxt_uninit_resources(bp, false);
1704 
1705 	bnxt_drv_uninit(bp);
1706 
1707 	return ret;
1708 }
1709 
1710 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1711 				    uint32_t index)
1712 {
1713 	struct bnxt *bp = eth_dev->data->dev_private;
1714 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1715 	struct bnxt_vnic_info *vnic;
1716 	struct bnxt_filter_info *filter, *temp_filter;
1717 	uint32_t i;
1718 
1719 	if (is_bnxt_in_error(bp))
1720 		return;
1721 
1722 	/*
1723 	 * Loop through all VNICs from the specified filter flow pools to
1724 	 * remove the corresponding MAC addr filter
1725 	 */
1726 	for (i = 0; i < bp->nr_vnics; i++) {
1727 		if (!(pool_mask & (1ULL << i)))
1728 			continue;
1729 
1730 		vnic = &bp->vnic_info[i];
1731 		filter = STAILQ_FIRST(&vnic->filter);
1732 		while (filter) {
1733 			temp_filter = STAILQ_NEXT(filter, next);
1734 			if (filter->mac_index == index) {
1735 				STAILQ_REMOVE(&vnic->filter, filter,
1736 						bnxt_filter_info, next);
1737 				bnxt_hwrm_clear_l2_filter(bp, filter);
1738 				bnxt_free_filter(bp, filter);
1739 			}
1740 			filter = temp_filter;
1741 		}
1742 	}
1743 }
1744 
1745 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1746 			       struct rte_ether_addr *mac_addr, uint32_t index,
1747 			       uint32_t pool)
1748 {
1749 	struct bnxt_filter_info *filter;
1750 	int rc = 0;
1751 
1752 	/* Attach requested MAC address to the new l2_filter */
1753 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1754 		if (filter->mac_index == index) {
1755 			PMD_DRV_LOG(DEBUG,
1756 				    "MAC addr already existed for pool %d\n",
1757 				    pool);
1758 			return 0;
1759 		}
1760 	}
1761 
1762 	filter = bnxt_alloc_filter(bp);
1763 	if (!filter) {
1764 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1765 		return -ENODEV;
1766 	}
1767 
1768 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1769 	 * if the MAC that's been programmed now is a different one, then,
1770 	 * copy that addr to filter->l2_addr
1771 	 */
1772 	if (mac_addr)
1773 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1774 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1775 
1776 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1777 	if (!rc) {
1778 		filter->mac_index = index;
1779 		if (filter->mac_index == 0)
1780 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1781 		else
1782 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1783 	} else {
1784 		bnxt_free_filter(bp, filter);
1785 	}
1786 
1787 	return rc;
1788 }
1789 
1790 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1791 				struct rte_ether_addr *mac_addr,
1792 				uint32_t index, uint32_t pool)
1793 {
1794 	struct bnxt *bp = eth_dev->data->dev_private;
1795 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1796 	int rc = 0;
1797 
1798 	rc = is_bnxt_in_error(bp);
1799 	if (rc)
1800 		return rc;
1801 
1802 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1803 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1804 		return -ENOTSUP;
1805 	}
1806 
1807 	if (!vnic) {
1808 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1809 		return -EINVAL;
1810 	}
1811 
1812 	/* Filter settings will get applied when port is started */
1813 	if (!eth_dev->data->dev_started)
1814 		return 0;
1815 
1816 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1817 
1818 	return rc;
1819 }
1820 
1821 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1822 {
1823 	int rc = 0;
1824 	struct bnxt *bp = eth_dev->data->dev_private;
1825 	struct rte_eth_link new;
1826 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1827 			BNXT_MIN_LINK_WAIT_CNT;
1828 
1829 	rc = is_bnxt_in_error(bp);
1830 	if (rc)
1831 		return rc;
1832 
1833 	memset(&new, 0, sizeof(new));
1834 
1835 	if (bp->link_info == NULL)
1836 		goto out;
1837 
1838 	do {
1839 		/* Retrieve link info from hardware */
1840 		rc = bnxt_get_hwrm_link_config(bp, &new);
1841 		if (rc) {
1842 			new.link_speed = ETH_LINK_SPEED_100M;
1843 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1844 			PMD_DRV_LOG(ERR,
1845 				"Failed to retrieve link rc = 0x%x!\n", rc);
1846 			goto out;
1847 		}
1848 
1849 		if (!wait_to_complete || new.link_status)
1850 			break;
1851 
1852 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1853 	} while (cnt--);
1854 
1855 	/* Only single function PF can bring phy down.
1856 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1857 	 */
1858 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1859 		memset(&new, 0, sizeof(new));
1860 
1861 out:
1862 	/* Timed out or success */
1863 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1864 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1865 		rte_eth_linkstatus_set(eth_dev, &new);
1866 		bnxt_print_link_info(eth_dev);
1867 	}
1868 
1869 	return rc;
1870 }
1871 
1872 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1873 {
1874 	struct bnxt *bp = eth_dev->data->dev_private;
1875 	struct bnxt_vnic_info *vnic;
1876 	uint32_t old_flags;
1877 	int rc;
1878 
1879 	rc = is_bnxt_in_error(bp);
1880 	if (rc)
1881 		return rc;
1882 
1883 	/* Filter settings will get applied when port is started */
1884 	if (!eth_dev->data->dev_started)
1885 		return 0;
1886 
1887 	if (bp->vnic_info == NULL)
1888 		return 0;
1889 
1890 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1891 
1892 	old_flags = vnic->flags;
1893 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1894 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1895 	if (rc != 0)
1896 		vnic->flags = old_flags;
1897 
1898 	return rc;
1899 }
1900 
1901 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1902 {
1903 	struct bnxt *bp = eth_dev->data->dev_private;
1904 	struct bnxt_vnic_info *vnic;
1905 	uint32_t old_flags;
1906 	int rc;
1907 
1908 	rc = is_bnxt_in_error(bp);
1909 	if (rc)
1910 		return rc;
1911 
1912 	/* Filter settings will get applied when port is started */
1913 	if (!eth_dev->data->dev_started)
1914 		return 0;
1915 
1916 	if (bp->vnic_info == NULL)
1917 		return 0;
1918 
1919 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1920 
1921 	old_flags = vnic->flags;
1922 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1923 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1924 	if (rc != 0)
1925 		vnic->flags = old_flags;
1926 
1927 	return rc;
1928 }
1929 
1930 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1931 {
1932 	struct bnxt *bp = eth_dev->data->dev_private;
1933 	struct bnxt_vnic_info *vnic;
1934 	uint32_t old_flags;
1935 	int rc;
1936 
1937 	rc = is_bnxt_in_error(bp);
1938 	if (rc)
1939 		return rc;
1940 
1941 	/* Filter settings will get applied when port is started */
1942 	if (!eth_dev->data->dev_started)
1943 		return 0;
1944 
1945 	if (bp->vnic_info == NULL)
1946 		return 0;
1947 
1948 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1949 
1950 	old_flags = vnic->flags;
1951 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1952 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1953 	if (rc != 0)
1954 		vnic->flags = old_flags;
1955 
1956 	return rc;
1957 }
1958 
1959 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1960 {
1961 	struct bnxt *bp = eth_dev->data->dev_private;
1962 	struct bnxt_vnic_info *vnic;
1963 	uint32_t old_flags;
1964 	int rc;
1965 
1966 	rc = is_bnxt_in_error(bp);
1967 	if (rc)
1968 		return rc;
1969 
1970 	/* Filter settings will get applied when port is started */
1971 	if (!eth_dev->data->dev_started)
1972 		return 0;
1973 
1974 	if (bp->vnic_info == NULL)
1975 		return 0;
1976 
1977 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1978 
1979 	old_flags = vnic->flags;
1980 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1981 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1982 	if (rc != 0)
1983 		vnic->flags = old_flags;
1984 
1985 	return rc;
1986 }
1987 
1988 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1989 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1990 {
1991 	if (qid >= bp->rx_nr_rings)
1992 		return NULL;
1993 
1994 	return bp->eth_dev->data->rx_queues[qid];
1995 }
1996 
1997 /* Return rxq corresponding to a given rss table ring/group ID. */
1998 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1999 {
2000 	struct bnxt_rx_queue *rxq;
2001 	unsigned int i;
2002 
2003 	if (!BNXT_HAS_RING_GRPS(bp)) {
2004 		for (i = 0; i < bp->rx_nr_rings; i++) {
2005 			rxq = bp->eth_dev->data->rx_queues[i];
2006 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
2007 				return rxq->index;
2008 		}
2009 	} else {
2010 		for (i = 0; i < bp->rx_nr_rings; i++) {
2011 			if (bp->grp_info[i].fw_grp_id == fwr)
2012 				return i;
2013 		}
2014 	}
2015 
2016 	return INVALID_HW_RING_ID;
2017 }
2018 
2019 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
2020 			    struct rte_eth_rss_reta_entry64 *reta_conf,
2021 			    uint16_t reta_size)
2022 {
2023 	struct bnxt *bp = eth_dev->data->dev_private;
2024 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2025 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2026 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2027 	uint16_t idx, sft;
2028 	int i, rc;
2029 
2030 	rc = is_bnxt_in_error(bp);
2031 	if (rc)
2032 		return rc;
2033 
2034 	if (!vnic->rss_table)
2035 		return -EINVAL;
2036 
2037 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
2038 		return -EINVAL;
2039 
2040 	if (reta_size != tbl_size) {
2041 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2042 			"(%d) must equal the size supported by the hardware "
2043 			"(%d)\n", reta_size, tbl_size);
2044 		return -EINVAL;
2045 	}
2046 
2047 	for (i = 0; i < reta_size; i++) {
2048 		struct bnxt_rx_queue *rxq;
2049 
2050 		idx = i / RTE_RETA_GROUP_SIZE;
2051 		sft = i % RTE_RETA_GROUP_SIZE;
2052 
2053 		if (!(reta_conf[idx].mask & (1ULL << sft)))
2054 			continue;
2055 
2056 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2057 		if (!rxq) {
2058 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
2059 			return -EINVAL;
2060 		}
2061 
2062 		if (BNXT_CHIP_P5(bp)) {
2063 			vnic->rss_table[i * 2] =
2064 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
2065 			vnic->rss_table[i * 2 + 1] =
2066 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
2067 		} else {
2068 			vnic->rss_table[i] =
2069 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2070 		}
2071 	}
2072 
2073 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2074 	return rc;
2075 }
2076 
2077 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2078 			      struct rte_eth_rss_reta_entry64 *reta_conf,
2079 			      uint16_t reta_size)
2080 {
2081 	struct bnxt *bp = eth_dev->data->dev_private;
2082 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2083 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2084 	uint16_t idx, sft, i;
2085 	int rc;
2086 
2087 	rc = is_bnxt_in_error(bp);
2088 	if (rc)
2089 		return rc;
2090 
2091 	if (!vnic)
2092 		return -EINVAL;
2093 	if (!vnic->rss_table)
2094 		return -EINVAL;
2095 
2096 	if (reta_size != tbl_size) {
2097 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2098 			"(%d) must equal the size supported by the hardware "
2099 			"(%d)\n", reta_size, tbl_size);
2100 		return -EINVAL;
2101 	}
2102 
2103 	for (idx = 0, i = 0; i < reta_size; i++) {
2104 		idx = i / RTE_RETA_GROUP_SIZE;
2105 		sft = i % RTE_RETA_GROUP_SIZE;
2106 
2107 		if (reta_conf[idx].mask & (1ULL << sft)) {
2108 			uint16_t qid;
2109 
2110 			if (BNXT_CHIP_P5(bp))
2111 				qid = bnxt_rss_to_qid(bp,
2112 						      vnic->rss_table[i * 2]);
2113 			else
2114 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2115 
2116 			if (qid == INVALID_HW_RING_ID) {
2117 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2118 				return -EINVAL;
2119 			}
2120 			reta_conf[idx].reta[sft] = qid;
2121 		}
2122 	}
2123 
2124 	return 0;
2125 }
2126 
2127 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2128 				   struct rte_eth_rss_conf *rss_conf)
2129 {
2130 	struct bnxt *bp = eth_dev->data->dev_private;
2131 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2132 	struct bnxt_vnic_info *vnic;
2133 	int rc;
2134 
2135 	rc = is_bnxt_in_error(bp);
2136 	if (rc)
2137 		return rc;
2138 
2139 	/*
2140 	 * If RSS enablement were different than dev_configure,
2141 	 * then return -EINVAL
2142 	 */
2143 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
2144 		if (!rss_conf->rss_hf)
2145 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2146 	} else {
2147 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2148 			return -EINVAL;
2149 	}
2150 
2151 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
2152 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2153 	       rss_conf,
2154 	       sizeof(*rss_conf));
2155 
2156 	/* Update the default RSS VNIC(s) */
2157 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2158 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2159 	vnic->hash_mode =
2160 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2161 					    ETH_RSS_LEVEL(rss_conf->rss_hf));
2162 
2163 	/*
2164 	 * If hashkey is not specified, use the previously configured
2165 	 * hashkey
2166 	 */
2167 	if (!rss_conf->rss_key)
2168 		goto rss_config;
2169 
2170 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2171 		PMD_DRV_LOG(ERR,
2172 			    "Invalid hashkey length, should be %d bytes\n",
2173 			    HW_HASH_KEY_SIZE);
2174 		return -EINVAL;
2175 	}
2176 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2177 
2178 rss_config:
2179 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2180 	return rc;
2181 }
2182 
2183 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2184 				     struct rte_eth_rss_conf *rss_conf)
2185 {
2186 	struct bnxt *bp = eth_dev->data->dev_private;
2187 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2188 	int len, rc;
2189 	uint32_t hash_types;
2190 
2191 	rc = is_bnxt_in_error(bp);
2192 	if (rc)
2193 		return rc;
2194 
2195 	/* RSS configuration is the same for all VNICs */
2196 	if (vnic && vnic->rss_hash_key) {
2197 		if (rss_conf->rss_key) {
2198 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2199 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2200 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2201 		}
2202 
2203 		hash_types = vnic->hash_type;
2204 		rss_conf->rss_hf = 0;
2205 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2206 			rss_conf->rss_hf |= ETH_RSS_IPV4;
2207 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2208 		}
2209 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2210 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2211 			hash_types &=
2212 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2213 		}
2214 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2215 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2216 			hash_types &=
2217 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2218 		}
2219 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2220 			rss_conf->rss_hf |= ETH_RSS_IPV6;
2221 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2222 		}
2223 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2224 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2225 			hash_types &=
2226 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2227 		}
2228 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2229 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2230 			hash_types &=
2231 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2232 		}
2233 
2234 		rss_conf->rss_hf |=
2235 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2236 
2237 		if (hash_types) {
2238 			PMD_DRV_LOG(ERR,
2239 				"Unknown RSS config from firmware (%08x), RSS disabled",
2240 				vnic->hash_type);
2241 			return -ENOTSUP;
2242 		}
2243 	} else {
2244 		rss_conf->rss_hf = 0;
2245 	}
2246 	return 0;
2247 }
2248 
2249 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2250 			       struct rte_eth_fc_conf *fc_conf)
2251 {
2252 	struct bnxt *bp = dev->data->dev_private;
2253 	struct rte_eth_link link_info;
2254 	int rc;
2255 
2256 	rc = is_bnxt_in_error(bp);
2257 	if (rc)
2258 		return rc;
2259 
2260 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2261 	if (rc)
2262 		return rc;
2263 
2264 	memset(fc_conf, 0, sizeof(*fc_conf));
2265 	if (bp->link_info->auto_pause)
2266 		fc_conf->autoneg = 1;
2267 	switch (bp->link_info->pause) {
2268 	case 0:
2269 		fc_conf->mode = RTE_FC_NONE;
2270 		break;
2271 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2272 		fc_conf->mode = RTE_FC_TX_PAUSE;
2273 		break;
2274 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2275 		fc_conf->mode = RTE_FC_RX_PAUSE;
2276 		break;
2277 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2278 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2279 		fc_conf->mode = RTE_FC_FULL;
2280 		break;
2281 	}
2282 	return 0;
2283 }
2284 
2285 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2286 			       struct rte_eth_fc_conf *fc_conf)
2287 {
2288 	struct bnxt *bp = dev->data->dev_private;
2289 	int rc;
2290 
2291 	rc = is_bnxt_in_error(bp);
2292 	if (rc)
2293 		return rc;
2294 
2295 	if (!BNXT_SINGLE_PF(bp)) {
2296 		PMD_DRV_LOG(ERR,
2297 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2298 		return -ENOTSUP;
2299 	}
2300 
2301 	switch (fc_conf->mode) {
2302 	case RTE_FC_NONE:
2303 		bp->link_info->auto_pause = 0;
2304 		bp->link_info->force_pause = 0;
2305 		break;
2306 	case RTE_FC_RX_PAUSE:
2307 		if (fc_conf->autoneg) {
2308 			bp->link_info->auto_pause =
2309 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2310 			bp->link_info->force_pause = 0;
2311 		} else {
2312 			bp->link_info->auto_pause = 0;
2313 			bp->link_info->force_pause =
2314 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2315 		}
2316 		break;
2317 	case RTE_FC_TX_PAUSE:
2318 		if (fc_conf->autoneg) {
2319 			bp->link_info->auto_pause =
2320 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2321 			bp->link_info->force_pause = 0;
2322 		} else {
2323 			bp->link_info->auto_pause = 0;
2324 			bp->link_info->force_pause =
2325 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2326 		}
2327 		break;
2328 	case RTE_FC_FULL:
2329 		if (fc_conf->autoneg) {
2330 			bp->link_info->auto_pause =
2331 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2332 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2333 			bp->link_info->force_pause = 0;
2334 		} else {
2335 			bp->link_info->auto_pause = 0;
2336 			bp->link_info->force_pause =
2337 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2338 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2339 		}
2340 		break;
2341 	}
2342 	return bnxt_set_hwrm_link_config(bp, true);
2343 }
2344 
2345 /* Add UDP tunneling port */
2346 static int
2347 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2348 			 struct rte_eth_udp_tunnel *udp_tunnel)
2349 {
2350 	struct bnxt *bp = eth_dev->data->dev_private;
2351 	uint16_t tunnel_type = 0;
2352 	int rc = 0;
2353 
2354 	rc = is_bnxt_in_error(bp);
2355 	if (rc)
2356 		return rc;
2357 
2358 	switch (udp_tunnel->prot_type) {
2359 	case RTE_TUNNEL_TYPE_VXLAN:
2360 		if (bp->vxlan_port_cnt) {
2361 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2362 				udp_tunnel->udp_port);
2363 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2364 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2365 				return -ENOSPC;
2366 			}
2367 			bp->vxlan_port_cnt++;
2368 			return 0;
2369 		}
2370 		tunnel_type =
2371 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2372 		break;
2373 	case RTE_TUNNEL_TYPE_GENEVE:
2374 		if (bp->geneve_port_cnt) {
2375 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2376 				udp_tunnel->udp_port);
2377 			if (bp->geneve_port != udp_tunnel->udp_port) {
2378 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2379 				return -ENOSPC;
2380 			}
2381 			bp->geneve_port_cnt++;
2382 			return 0;
2383 		}
2384 		tunnel_type =
2385 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2386 		break;
2387 	default:
2388 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2389 		return -ENOTSUP;
2390 	}
2391 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2392 					     tunnel_type);
2393 
2394 	if (rc != 0)
2395 		return rc;
2396 
2397 	if (tunnel_type ==
2398 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN)
2399 		bp->vxlan_port_cnt++;
2400 
2401 	if (tunnel_type ==
2402 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE)
2403 		bp->geneve_port_cnt++;
2404 
2405 	return rc;
2406 }
2407 
2408 static int
2409 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2410 			 struct rte_eth_udp_tunnel *udp_tunnel)
2411 {
2412 	struct bnxt *bp = eth_dev->data->dev_private;
2413 	uint16_t tunnel_type = 0;
2414 	uint16_t port = 0;
2415 	int rc = 0;
2416 
2417 	rc = is_bnxt_in_error(bp);
2418 	if (rc)
2419 		return rc;
2420 
2421 	switch (udp_tunnel->prot_type) {
2422 	case RTE_TUNNEL_TYPE_VXLAN:
2423 		if (!bp->vxlan_port_cnt) {
2424 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2425 			return -EINVAL;
2426 		}
2427 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2428 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2429 				udp_tunnel->udp_port, bp->vxlan_port);
2430 			return -EINVAL;
2431 		}
2432 		if (--bp->vxlan_port_cnt)
2433 			return 0;
2434 
2435 		tunnel_type =
2436 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2437 		port = bp->vxlan_fw_dst_port_id;
2438 		break;
2439 	case RTE_TUNNEL_TYPE_GENEVE:
2440 		if (!bp->geneve_port_cnt) {
2441 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2442 			return -EINVAL;
2443 		}
2444 		if (bp->geneve_port != udp_tunnel->udp_port) {
2445 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2446 				udp_tunnel->udp_port, bp->geneve_port);
2447 			return -EINVAL;
2448 		}
2449 		if (--bp->geneve_port_cnt)
2450 			return 0;
2451 
2452 		tunnel_type =
2453 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2454 		port = bp->geneve_fw_dst_port_id;
2455 		break;
2456 	default:
2457 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2458 		return -ENOTSUP;
2459 	}
2460 
2461 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2462 	return rc;
2463 }
2464 
2465 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2466 {
2467 	struct bnxt_filter_info *filter;
2468 	struct bnxt_vnic_info *vnic;
2469 	int rc = 0;
2470 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2471 
2472 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2473 	filter = STAILQ_FIRST(&vnic->filter);
2474 	while (filter) {
2475 		/* Search for this matching MAC+VLAN filter */
2476 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2477 			/* Delete the filter */
2478 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2479 			if (rc)
2480 				return rc;
2481 			STAILQ_REMOVE(&vnic->filter, filter,
2482 				      bnxt_filter_info, next);
2483 			bnxt_free_filter(bp, filter);
2484 			PMD_DRV_LOG(INFO,
2485 				    "Deleted vlan filter for %d\n",
2486 				    vlan_id);
2487 			return 0;
2488 		}
2489 		filter = STAILQ_NEXT(filter, next);
2490 	}
2491 	return -ENOENT;
2492 }
2493 
2494 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2495 {
2496 	struct bnxt_filter_info *filter;
2497 	struct bnxt_vnic_info *vnic;
2498 	int rc = 0;
2499 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2500 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2501 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2502 
2503 	/* Implementation notes on the use of VNIC in this command:
2504 	 *
2505 	 * By default, these filters belong to default vnic for the function.
2506 	 * Once these filters are set up, only destination VNIC can be modified.
2507 	 * If the destination VNIC is not specified in this command,
2508 	 * then the HWRM shall only create an l2 context id.
2509 	 */
2510 
2511 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2512 	filter = STAILQ_FIRST(&vnic->filter);
2513 	/* Check if the VLAN has already been added */
2514 	while (filter) {
2515 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2516 			return -EEXIST;
2517 
2518 		filter = STAILQ_NEXT(filter, next);
2519 	}
2520 
2521 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2522 	 * command to create MAC+VLAN filter with the right flags, enables set.
2523 	 */
2524 	filter = bnxt_alloc_filter(bp);
2525 	if (!filter) {
2526 		PMD_DRV_LOG(ERR,
2527 			    "MAC/VLAN filter alloc failed\n");
2528 		return -ENOMEM;
2529 	}
2530 	/* MAC + VLAN ID filter */
2531 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2532 	 * untagged packets are received
2533 	 *
2534 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2535 	 * packets and only the programmed vlan's packets are received
2536 	 */
2537 	filter->l2_ivlan = vlan_id;
2538 	filter->l2_ivlan_mask = 0x0FFF;
2539 	filter->enables |= en;
2540 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2541 
2542 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2543 	if (rc) {
2544 		/* Free the newly allocated filter as we were
2545 		 * not able to create the filter in hardware.
2546 		 */
2547 		bnxt_free_filter(bp, filter);
2548 		return rc;
2549 	}
2550 
2551 	filter->mac_index = 0;
2552 	/* Add this new filter to the list */
2553 	if (vlan_id == 0)
2554 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2555 	else
2556 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2557 
2558 	PMD_DRV_LOG(INFO,
2559 		    "Added Vlan filter for %d\n", vlan_id);
2560 	return rc;
2561 }
2562 
2563 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2564 		uint16_t vlan_id, int on)
2565 {
2566 	struct bnxt *bp = eth_dev->data->dev_private;
2567 	int rc;
2568 
2569 	rc = is_bnxt_in_error(bp);
2570 	if (rc)
2571 		return rc;
2572 
2573 	if (!eth_dev->data->dev_started) {
2574 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2575 		return -EINVAL;
2576 	}
2577 
2578 	/* These operations apply to ALL existing MAC/VLAN filters */
2579 	if (on)
2580 		return bnxt_add_vlan_filter(bp, vlan_id);
2581 	else
2582 		return bnxt_del_vlan_filter(bp, vlan_id);
2583 }
2584 
2585 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2586 				    struct bnxt_vnic_info *vnic)
2587 {
2588 	struct bnxt_filter_info *filter;
2589 	int rc;
2590 
2591 	filter = STAILQ_FIRST(&vnic->filter);
2592 	while (filter) {
2593 		if (filter->mac_index == 0 &&
2594 		    !memcmp(filter->l2_addr, bp->mac_addr,
2595 			    RTE_ETHER_ADDR_LEN)) {
2596 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2597 			if (!rc) {
2598 				STAILQ_REMOVE(&vnic->filter, filter,
2599 					      bnxt_filter_info, next);
2600 				bnxt_free_filter(bp, filter);
2601 			}
2602 			return rc;
2603 		}
2604 		filter = STAILQ_NEXT(filter, next);
2605 	}
2606 	return 0;
2607 }
2608 
2609 static int
2610 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2611 {
2612 	struct bnxt_vnic_info *vnic;
2613 	unsigned int i;
2614 	int rc;
2615 
2616 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2617 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2618 		/* Remove any VLAN filters programmed */
2619 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2620 			bnxt_del_vlan_filter(bp, i);
2621 
2622 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2623 		if (rc)
2624 			return rc;
2625 	} else {
2626 		/* Default filter will allow packets that match the
2627 		 * dest mac. So, it has to be deleted, otherwise, we
2628 		 * will endup receiving vlan packets for which the
2629 		 * filter is not programmed, when hw-vlan-filter
2630 		 * configuration is ON
2631 		 */
2632 		bnxt_del_dflt_mac_filter(bp, vnic);
2633 		/* This filter will allow only untagged packets */
2634 		bnxt_add_vlan_filter(bp, 0);
2635 	}
2636 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2637 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2638 
2639 	return 0;
2640 }
2641 
2642 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2643 {
2644 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2645 	unsigned int i;
2646 	int rc;
2647 
2648 	/* Destroy vnic filters and vnic */
2649 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2650 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2651 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2652 			bnxt_del_vlan_filter(bp, i);
2653 	}
2654 	bnxt_del_dflt_mac_filter(bp, vnic);
2655 
2656 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2657 	if (rc)
2658 		return rc;
2659 
2660 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2661 	if (rc)
2662 		return rc;
2663 
2664 	rte_free(vnic->fw_grp_ids);
2665 	vnic->fw_grp_ids = NULL;
2666 
2667 	vnic->rx_queue_cnt = 0;
2668 
2669 	return 0;
2670 }
2671 
2672 static int
2673 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2674 {
2675 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2676 	int rc;
2677 
2678 	/* Destroy, recreate and reconfigure the default vnic */
2679 	rc = bnxt_free_one_vnic(bp, 0);
2680 	if (rc)
2681 		return rc;
2682 
2683 	/* default vnic 0 */
2684 	rc = bnxt_setup_one_vnic(bp, 0);
2685 	if (rc)
2686 		return rc;
2687 
2688 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2689 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2690 		rc = bnxt_add_vlan_filter(bp, 0);
2691 		if (rc)
2692 			return rc;
2693 		rc = bnxt_restore_vlan_filters(bp);
2694 		if (rc)
2695 			return rc;
2696 	} else {
2697 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2698 		if (rc)
2699 			return rc;
2700 	}
2701 
2702 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2703 	if (rc)
2704 		return rc;
2705 
2706 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2707 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2708 
2709 	return rc;
2710 }
2711 
2712 static int
2713 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2714 {
2715 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2716 	struct bnxt *bp = dev->data->dev_private;
2717 	int rc;
2718 
2719 	rc = is_bnxt_in_error(bp);
2720 	if (rc)
2721 		return rc;
2722 
2723 	/* Filter settings will get applied when port is started */
2724 	if (!dev->data->dev_started)
2725 		return 0;
2726 
2727 	if (mask & ETH_VLAN_FILTER_MASK) {
2728 		/* Enable or disable VLAN filtering */
2729 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2730 		if (rc)
2731 			return rc;
2732 	}
2733 
2734 	if (mask & ETH_VLAN_STRIP_MASK) {
2735 		/* Enable or disable VLAN stripping */
2736 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2737 		if (rc)
2738 			return rc;
2739 	}
2740 
2741 	if (mask & ETH_VLAN_EXTEND_MASK) {
2742 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2743 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2744 		else
2745 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2746 	}
2747 
2748 	return 0;
2749 }
2750 
2751 static int
2752 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2753 		      uint16_t tpid)
2754 {
2755 	struct bnxt *bp = dev->data->dev_private;
2756 	int qinq = dev->data->dev_conf.rxmode.offloads &
2757 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2758 
2759 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2760 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2761 		PMD_DRV_LOG(ERR,
2762 			    "Unsupported vlan type.");
2763 		return -EINVAL;
2764 	}
2765 	if (!qinq) {
2766 		PMD_DRV_LOG(ERR,
2767 			    "QinQ not enabled. Needs to be ON as we can "
2768 			    "accelerate only outer vlan\n");
2769 		return -EINVAL;
2770 	}
2771 
2772 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2773 		switch (tpid) {
2774 		case RTE_ETHER_TYPE_QINQ:
2775 			bp->outer_tpid_bd =
2776 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2777 				break;
2778 		case RTE_ETHER_TYPE_VLAN:
2779 			bp->outer_tpid_bd =
2780 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2781 				break;
2782 		case RTE_ETHER_TYPE_QINQ1:
2783 			bp->outer_tpid_bd =
2784 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2785 				break;
2786 		case RTE_ETHER_TYPE_QINQ2:
2787 			bp->outer_tpid_bd =
2788 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2789 				break;
2790 		case RTE_ETHER_TYPE_QINQ3:
2791 			bp->outer_tpid_bd =
2792 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2793 				break;
2794 		default:
2795 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2796 			return -EINVAL;
2797 		}
2798 		bp->outer_tpid_bd |= tpid;
2799 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2800 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2801 		PMD_DRV_LOG(ERR,
2802 			    "Can accelerate only outer vlan in QinQ\n");
2803 		return -EINVAL;
2804 	}
2805 
2806 	return 0;
2807 }
2808 
2809 static int
2810 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2811 			     struct rte_ether_addr *addr)
2812 {
2813 	struct bnxt *bp = dev->data->dev_private;
2814 	/* Default Filter is tied to VNIC 0 */
2815 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2816 	int rc;
2817 
2818 	rc = is_bnxt_in_error(bp);
2819 	if (rc)
2820 		return rc;
2821 
2822 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2823 		return -EPERM;
2824 
2825 	if (rte_is_zero_ether_addr(addr))
2826 		return -EINVAL;
2827 
2828 	/* Filter settings will get applied when port is started */
2829 	if (!dev->data->dev_started)
2830 		return 0;
2831 
2832 	/* Check if the requested MAC is already added */
2833 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2834 		return 0;
2835 
2836 	/* Destroy filter and re-create it */
2837 	bnxt_del_dflt_mac_filter(bp, vnic);
2838 
2839 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2840 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2841 		/* This filter will allow only untagged packets */
2842 		rc = bnxt_add_vlan_filter(bp, 0);
2843 	} else {
2844 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2845 	}
2846 
2847 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2848 	return rc;
2849 }
2850 
2851 static int
2852 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2853 			  struct rte_ether_addr *mc_addr_set,
2854 			  uint32_t nb_mc_addr)
2855 {
2856 	struct bnxt *bp = eth_dev->data->dev_private;
2857 	char *mc_addr_list = (char *)mc_addr_set;
2858 	struct bnxt_vnic_info *vnic;
2859 	uint32_t off = 0, i = 0;
2860 	int rc;
2861 
2862 	rc = is_bnxt_in_error(bp);
2863 	if (rc)
2864 		return rc;
2865 
2866 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2867 
2868 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2869 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2870 		goto allmulti;
2871 	}
2872 
2873 	/* TODO Check for Duplicate mcast addresses */
2874 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2875 	for (i = 0; i < nb_mc_addr; i++) {
2876 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2877 			RTE_ETHER_ADDR_LEN);
2878 		off += RTE_ETHER_ADDR_LEN;
2879 	}
2880 
2881 	vnic->mc_addr_cnt = i;
2882 	if (vnic->mc_addr_cnt)
2883 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2884 	else
2885 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2886 
2887 allmulti:
2888 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2889 }
2890 
2891 static int
2892 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2893 {
2894 	struct bnxt *bp = dev->data->dev_private;
2895 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2896 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2897 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2898 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2899 	int ret;
2900 
2901 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2902 			fw_major, fw_minor, fw_updt, fw_rsvd);
2903 	if (ret < 0)
2904 		return -EINVAL;
2905 
2906 	ret += 1; /* add the size of '\0' */
2907 	if (fw_size < (size_t)ret)
2908 		return ret;
2909 	else
2910 		return 0;
2911 }
2912 
2913 static void
2914 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2915 	struct rte_eth_rxq_info *qinfo)
2916 {
2917 	struct bnxt *bp = dev->data->dev_private;
2918 	struct bnxt_rx_queue *rxq;
2919 
2920 	if (is_bnxt_in_error(bp))
2921 		return;
2922 
2923 	rxq = dev->data->rx_queues[queue_id];
2924 
2925 	qinfo->mp = rxq->mb_pool;
2926 	qinfo->scattered_rx = dev->data->scattered_rx;
2927 	qinfo->nb_desc = rxq->nb_rx_desc;
2928 
2929 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2930 	qinfo->conf.rx_drop_en = rxq->drop_en;
2931 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2932 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2933 }
2934 
2935 static void
2936 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2937 	struct rte_eth_txq_info *qinfo)
2938 {
2939 	struct bnxt *bp = dev->data->dev_private;
2940 	struct bnxt_tx_queue *txq;
2941 
2942 	if (is_bnxt_in_error(bp))
2943 		return;
2944 
2945 	txq = dev->data->tx_queues[queue_id];
2946 
2947 	qinfo->nb_desc = txq->nb_tx_desc;
2948 
2949 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2950 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2951 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2952 
2953 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2954 	qinfo->conf.tx_rs_thresh = 0;
2955 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2956 	qinfo->conf.offloads = txq->offloads;
2957 }
2958 
2959 static const struct {
2960 	eth_rx_burst_t pkt_burst;
2961 	const char *info;
2962 } bnxt_rx_burst_info[] = {
2963 	{bnxt_recv_pkts,		"Scalar"},
2964 #if defined(RTE_ARCH_X86)
2965 	{bnxt_recv_pkts_vec,		"Vector SSE"},
2966 #endif
2967 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2968 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
2969 #endif
2970 #if defined(RTE_ARCH_ARM64)
2971 	{bnxt_recv_pkts_vec,		"Vector Neon"},
2972 #endif
2973 };
2974 
2975 static int
2976 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2977 		       struct rte_eth_burst_mode *mode)
2978 {
2979 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2980 	size_t i;
2981 
2982 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2983 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2984 			snprintf(mode->info, sizeof(mode->info), "%s",
2985 				 bnxt_rx_burst_info[i].info);
2986 			return 0;
2987 		}
2988 	}
2989 
2990 	return -EINVAL;
2991 }
2992 
2993 static const struct {
2994 	eth_tx_burst_t pkt_burst;
2995 	const char *info;
2996 } bnxt_tx_burst_info[] = {
2997 	{bnxt_xmit_pkts,		"Scalar"},
2998 #if defined(RTE_ARCH_X86)
2999 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
3000 #endif
3001 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
3002 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
3003 #endif
3004 #if defined(RTE_ARCH_ARM64)
3005 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
3006 #endif
3007 };
3008 
3009 static int
3010 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
3011 		       struct rte_eth_burst_mode *mode)
3012 {
3013 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
3014 	size_t i;
3015 
3016 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
3017 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
3018 			snprintf(mode->info, sizeof(mode->info), "%s",
3019 				 bnxt_tx_burst_info[i].info);
3020 			return 0;
3021 		}
3022 	}
3023 
3024 	return -EINVAL;
3025 }
3026 
3027 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3028 {
3029 	struct bnxt *bp = eth_dev->data->dev_private;
3030 	uint32_t new_pkt_size;
3031 	uint32_t rc = 0;
3032 	uint32_t i;
3033 
3034 	rc = is_bnxt_in_error(bp);
3035 	if (rc)
3036 		return rc;
3037 
3038 	/* Exit if receive queues are not configured yet */
3039 	if (!eth_dev->data->nb_rx_queues)
3040 		return rc;
3041 
3042 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
3043 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
3044 
3045 	/*
3046 	 * Disallow any MTU change that would require scattered receive support
3047 	 * if it is not already enabled.
3048 	 */
3049 	if (eth_dev->data->dev_started &&
3050 	    !eth_dev->data->scattered_rx &&
3051 	    (new_pkt_size >
3052 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
3053 		PMD_DRV_LOG(ERR,
3054 			    "MTU change would require scattered rx support. ");
3055 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
3056 		return -EINVAL;
3057 	}
3058 
3059 	if (new_mtu > RTE_ETHER_MTU) {
3060 		bp->flags |= BNXT_FLAG_JUMBO;
3061 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
3062 			DEV_RX_OFFLOAD_JUMBO_FRAME;
3063 	} else {
3064 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
3065 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
3066 		bp->flags &= ~BNXT_FLAG_JUMBO;
3067 	}
3068 
3069 	/* Is there a change in mtu setting? */
3070 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
3071 		return rc;
3072 
3073 	for (i = 0; i < bp->nr_vnics; i++) {
3074 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3075 		uint16_t size = 0;
3076 
3077 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
3078 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
3079 		if (rc)
3080 			break;
3081 
3082 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
3083 		size -= RTE_PKTMBUF_HEADROOM;
3084 
3085 		if (size < new_mtu) {
3086 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
3087 			if (rc)
3088 				return rc;
3089 		}
3090 	}
3091 
3092 	if (!rc)
3093 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
3094 
3095 	if (bnxt_hwrm_config_host_mtu(bp))
3096 		PMD_DRV_LOG(WARNING, "Failed to configure host MTU\n");
3097 
3098 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3099 
3100 	return rc;
3101 }
3102 
3103 static int
3104 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3105 {
3106 	struct bnxt *bp = dev->data->dev_private;
3107 	uint16_t vlan = bp->vlan;
3108 	int rc;
3109 
3110 	rc = is_bnxt_in_error(bp);
3111 	if (rc)
3112 		return rc;
3113 
3114 	if (!BNXT_SINGLE_PF(bp)) {
3115 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3116 		return -ENOTSUP;
3117 	}
3118 	bp->vlan = on ? pvid : 0;
3119 
3120 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3121 	if (rc)
3122 		bp->vlan = vlan;
3123 	return rc;
3124 }
3125 
3126 static int
3127 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3128 {
3129 	struct bnxt *bp = dev->data->dev_private;
3130 	int rc;
3131 
3132 	rc = is_bnxt_in_error(bp);
3133 	if (rc)
3134 		return rc;
3135 
3136 	return bnxt_hwrm_port_led_cfg(bp, true);
3137 }
3138 
3139 static int
3140 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3141 {
3142 	struct bnxt *bp = dev->data->dev_private;
3143 	int rc;
3144 
3145 	rc = is_bnxt_in_error(bp);
3146 	if (rc)
3147 		return rc;
3148 
3149 	return bnxt_hwrm_port_led_cfg(bp, false);
3150 }
3151 
3152 static uint32_t
3153 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
3154 {
3155 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
3156 	struct bnxt_cp_ring_info *cpr;
3157 	uint32_t desc = 0, raw_cons, cp_ring_size;
3158 	struct bnxt_rx_queue *rxq;
3159 	struct rx_pkt_cmpl *rxcmp;
3160 	int rc;
3161 
3162 	rc = is_bnxt_in_error(bp);
3163 	if (rc)
3164 		return rc;
3165 
3166 	rxq = dev->data->rx_queues[rx_queue_id];
3167 	cpr = rxq->cp_ring;
3168 	raw_cons = cpr->cp_raw_cons;
3169 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3170 
3171 	while (1) {
3172 		uint32_t agg_cnt, cons, cmpl_type;
3173 
3174 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3175 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3176 
3177 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3178 			break;
3179 
3180 		cmpl_type = CMP_TYPE(rxcmp);
3181 
3182 		switch (cmpl_type) {
3183 		case CMPL_BASE_TYPE_RX_L2:
3184 		case CMPL_BASE_TYPE_RX_L2_V2:
3185 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3186 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3187 			desc++;
3188 			break;
3189 
3190 		case CMPL_BASE_TYPE_RX_TPA_END:
3191 			if (BNXT_CHIP_P5(rxq->bp)) {
3192 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3193 
3194 				p5_tpa_end = (void *)rxcmp;
3195 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3196 			} else {
3197 				struct rx_tpa_end_cmpl *tpa_end;
3198 
3199 				tpa_end = (void *)rxcmp;
3200 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3201 			}
3202 
3203 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3204 			desc++;
3205 			break;
3206 
3207 		default:
3208 			raw_cons += CMP_LEN(cmpl_type);
3209 		}
3210 	}
3211 
3212 	return desc;
3213 }
3214 
3215 static int
3216 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3217 {
3218 	struct bnxt_rx_queue *rxq = rx_queue;
3219 	struct bnxt_cp_ring_info *cpr;
3220 	struct bnxt_rx_ring_info *rxr;
3221 	uint32_t desc, raw_cons, cp_ring_size;
3222 	struct bnxt *bp = rxq->bp;
3223 	struct rx_pkt_cmpl *rxcmp;
3224 	int rc;
3225 
3226 	rc = is_bnxt_in_error(bp);
3227 	if (rc)
3228 		return rc;
3229 
3230 	if (offset >= rxq->nb_rx_desc)
3231 		return -EINVAL;
3232 
3233 	rxr = rxq->rx_ring;
3234 	cpr = rxq->cp_ring;
3235 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3236 
3237 	/*
3238 	 * For the vector receive case, the completion at the requested
3239 	 * offset can be indexed directly.
3240 	 */
3241 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3242 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3243 		struct rx_pkt_cmpl *rxcmp;
3244 		uint32_t cons;
3245 
3246 		/* Check status of completion descriptor. */
3247 		raw_cons = cpr->cp_raw_cons +
3248 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3249 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3250 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3251 
3252 		if (bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3253 			return RTE_ETH_RX_DESC_DONE;
3254 
3255 		/* Check whether rx desc has an mbuf attached. */
3256 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3257 		if (cons >= rxq->rxrearm_start &&
3258 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3259 			return RTE_ETH_RX_DESC_UNAVAIL;
3260 		}
3261 
3262 		return RTE_ETH_RX_DESC_AVAIL;
3263 	}
3264 #endif
3265 
3266 	/*
3267 	 * For the non-vector receive case, scan the completion ring to
3268 	 * locate the completion descriptor for the requested offset.
3269 	 */
3270 	raw_cons = cpr->cp_raw_cons;
3271 	desc = 0;
3272 	while (1) {
3273 		uint32_t agg_cnt, cons, cmpl_type;
3274 
3275 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3276 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3277 
3278 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3279 			break;
3280 
3281 		cmpl_type = CMP_TYPE(rxcmp);
3282 
3283 		switch (cmpl_type) {
3284 		case CMPL_BASE_TYPE_RX_L2:
3285 		case CMPL_BASE_TYPE_RX_L2_V2:
3286 			if (desc == offset) {
3287 				cons = rxcmp->opaque;
3288 				if (rxr->rx_buf_ring[cons])
3289 					return RTE_ETH_RX_DESC_DONE;
3290 				else
3291 					return RTE_ETH_RX_DESC_UNAVAIL;
3292 			}
3293 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3294 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3295 			desc++;
3296 			break;
3297 
3298 		case CMPL_BASE_TYPE_RX_TPA_END:
3299 			if (desc == offset)
3300 				return RTE_ETH_RX_DESC_DONE;
3301 
3302 			if (BNXT_CHIP_P5(rxq->bp)) {
3303 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3304 
3305 				p5_tpa_end = (void *)rxcmp;
3306 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3307 			} else {
3308 				struct rx_tpa_end_cmpl *tpa_end;
3309 
3310 				tpa_end = (void *)rxcmp;
3311 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3312 			}
3313 
3314 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3315 			desc++;
3316 			break;
3317 
3318 		default:
3319 			raw_cons += CMP_LEN(cmpl_type);
3320 		}
3321 	}
3322 
3323 	return RTE_ETH_RX_DESC_AVAIL;
3324 }
3325 
3326 static int
3327 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3328 {
3329 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3330 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3331 	uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3332 	struct cmpl_base *cp_desc_ring;
3333 	int rc;
3334 
3335 	rc = is_bnxt_in_error(txq->bp);
3336 	if (rc)
3337 		return rc;
3338 
3339 	if (offset >= txq->nb_tx_desc)
3340 		return -EINVAL;
3341 
3342 	/* Return "desc done" if descriptor is available for use. */
3343 	if (bnxt_tx_bds_in_hw(txq) <= offset)
3344 		return RTE_ETH_TX_DESC_DONE;
3345 
3346 	raw_cons = cpr->cp_raw_cons;
3347 	cp_desc_ring = cpr->cp_desc_ring;
3348 	ring_mask = cpr->cp_ring_struct->ring_mask;
3349 
3350 	/* Check to see if hw has posted a completion for the descriptor. */
3351 	while (1) {
3352 		struct tx_cmpl *txcmp;
3353 		uint32_t cons;
3354 
3355 		cons = RING_CMPL(ring_mask, raw_cons);
3356 		txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3357 
3358 		if (!bnxt_cpr_cmp_valid(txcmp, raw_cons, ring_mask + 1))
3359 			break;
3360 
3361 		if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3362 			nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3363 
3364 		if (nb_tx_pkts > offset)
3365 			return RTE_ETH_TX_DESC_DONE;
3366 
3367 		raw_cons = NEXT_RAW_CMP(raw_cons);
3368 	}
3369 
3370 	/* Descriptor is pending transmit, not yet completed by hardware. */
3371 	return RTE_ETH_TX_DESC_FULL;
3372 }
3373 
3374 int
3375 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3376 		     const struct rte_flow_ops **ops)
3377 {
3378 	struct bnxt *bp = dev->data->dev_private;
3379 	int ret = 0;
3380 
3381 	if (!bp)
3382 		return -EIO;
3383 
3384 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3385 		struct bnxt_representor *vfr = dev->data->dev_private;
3386 		bp = vfr->parent_dev->data->dev_private;
3387 		/* parent is deleted while children are still valid */
3388 		if (!bp) {
3389 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3390 				    dev->data->port_id);
3391 			return -EIO;
3392 		}
3393 	}
3394 
3395 	ret = is_bnxt_in_error(bp);
3396 	if (ret)
3397 		return ret;
3398 
3399 	/* PMD supports thread-safe flow operations.  rte_flow API
3400 	 * functions can avoid mutex for multi-thread safety.
3401 	 */
3402 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3403 
3404 	if (BNXT_TRUFLOW_EN(bp))
3405 		*ops = &bnxt_ulp_rte_flow_ops;
3406 	else
3407 		*ops = &bnxt_flow_ops;
3408 
3409 	return ret;
3410 }
3411 
3412 static const uint32_t *
3413 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3414 {
3415 	static const uint32_t ptypes[] = {
3416 		RTE_PTYPE_L2_ETHER_VLAN,
3417 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3418 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3419 		RTE_PTYPE_L4_ICMP,
3420 		RTE_PTYPE_L4_TCP,
3421 		RTE_PTYPE_L4_UDP,
3422 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3423 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3424 		RTE_PTYPE_INNER_L4_ICMP,
3425 		RTE_PTYPE_INNER_L4_TCP,
3426 		RTE_PTYPE_INNER_L4_UDP,
3427 		RTE_PTYPE_UNKNOWN
3428 	};
3429 
3430 	if (!dev->rx_pkt_burst)
3431 		return NULL;
3432 
3433 	return ptypes;
3434 }
3435 
3436 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3437 			 int reg_win)
3438 {
3439 	uint32_t reg_base = *reg_arr & 0xfffff000;
3440 	uint32_t win_off;
3441 	int i;
3442 
3443 	for (i = 0; i < count; i++) {
3444 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3445 			return -ERANGE;
3446 	}
3447 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3448 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3449 	return 0;
3450 }
3451 
3452 static int bnxt_map_ptp_regs(struct bnxt *bp)
3453 {
3454 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3455 	uint32_t *reg_arr;
3456 	int rc, i;
3457 
3458 	reg_arr = ptp->rx_regs;
3459 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3460 	if (rc)
3461 		return rc;
3462 
3463 	reg_arr = ptp->tx_regs;
3464 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3465 	if (rc)
3466 		return rc;
3467 
3468 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3469 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3470 
3471 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3472 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3473 
3474 	return 0;
3475 }
3476 
3477 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3478 {
3479 	rte_write32(0, (uint8_t *)bp->bar0 +
3480 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3481 	rte_write32(0, (uint8_t *)bp->bar0 +
3482 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3483 }
3484 
3485 static uint64_t bnxt_cc_read(struct bnxt *bp)
3486 {
3487 	uint64_t ns;
3488 
3489 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3490 			      BNXT_GRCPF_REG_SYNC_TIME));
3491 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3492 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3493 	return ns;
3494 }
3495 
3496 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3497 {
3498 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3499 	uint32_t fifo;
3500 
3501 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3502 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3503 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3504 		return -EAGAIN;
3505 
3506 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3507 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3508 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3509 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3510 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3511 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3512 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3513 
3514 	return 0;
3515 }
3516 
3517 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3518 {
3519 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3520 	struct bnxt_pf_info *pf = bp->pf;
3521 	uint16_t port_id;
3522 	int i = 0;
3523 	uint32_t fifo;
3524 
3525 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3526 		return -EINVAL;
3527 
3528 	port_id = pf->port_id;
3529 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3530 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3531 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3532 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3533 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3534 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3535 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3536 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3537 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3538 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3539 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3540 		i++;
3541 	}
3542 
3543 	if (i >= BNXT_PTP_RX_PND_CNT)
3544 		return -EBUSY;
3545 
3546 	return 0;
3547 }
3548 
3549 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3550 {
3551 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3552 	struct bnxt_pf_info *pf = bp->pf;
3553 	uint16_t port_id;
3554 	uint32_t fifo;
3555 
3556 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3557 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3558 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3559 		return -EAGAIN;
3560 
3561 	port_id = pf->port_id;
3562 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3563 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3564 
3565 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3566 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3567 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3568 		return bnxt_clr_rx_ts(bp, ts);
3569 
3570 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3571 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3572 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3573 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3574 
3575 	return 0;
3576 }
3577 
3578 static int
3579 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3580 {
3581 	uint64_t ns;
3582 	struct bnxt *bp = dev->data->dev_private;
3583 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3584 
3585 	if (!ptp)
3586 		return -ENOTSUP;
3587 
3588 	ns = rte_timespec_to_ns(ts);
3589 	/* Set the timecounters to a new value. */
3590 	ptp->tc.nsec = ns;
3591 	ptp->tx_tstamp_tc.nsec = ns;
3592 	ptp->rx_tstamp_tc.nsec = ns;
3593 
3594 	return 0;
3595 }
3596 
3597 static int
3598 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3599 {
3600 	struct bnxt *bp = dev->data->dev_private;
3601 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3602 	uint64_t ns, systime_cycles = 0;
3603 	int rc = 0;
3604 
3605 	if (!ptp)
3606 		return -ENOTSUP;
3607 
3608 	if (BNXT_CHIP_P5(bp))
3609 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3610 					     &systime_cycles);
3611 	else
3612 		systime_cycles = bnxt_cc_read(bp);
3613 
3614 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3615 	*ts = rte_ns_to_timespec(ns);
3616 
3617 	return rc;
3618 }
3619 static int
3620 bnxt_timesync_enable(struct rte_eth_dev *dev)
3621 {
3622 	struct bnxt *bp = dev->data->dev_private;
3623 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3624 	uint32_t shift = 0;
3625 	int rc;
3626 
3627 	if (!ptp)
3628 		return -ENOTSUP;
3629 
3630 	ptp->rx_filter = 1;
3631 	ptp->tx_tstamp_en = 1;
3632 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3633 
3634 	rc = bnxt_hwrm_ptp_cfg(bp);
3635 	if (rc)
3636 		return rc;
3637 
3638 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3639 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3640 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3641 
3642 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3643 	ptp->tc.cc_shift = shift;
3644 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3645 
3646 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3647 	ptp->rx_tstamp_tc.cc_shift = shift;
3648 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3649 
3650 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3651 	ptp->tx_tstamp_tc.cc_shift = shift;
3652 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3653 
3654 	if (!BNXT_CHIP_P5(bp))
3655 		bnxt_map_ptp_regs(bp);
3656 	else
3657 		rc = bnxt_ptp_start(bp);
3658 
3659 	return rc;
3660 }
3661 
3662 static int
3663 bnxt_timesync_disable(struct rte_eth_dev *dev)
3664 {
3665 	struct bnxt *bp = dev->data->dev_private;
3666 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3667 
3668 	if (!ptp)
3669 		return -ENOTSUP;
3670 
3671 	ptp->rx_filter = 0;
3672 	ptp->tx_tstamp_en = 0;
3673 	ptp->rxctl = 0;
3674 
3675 	bnxt_hwrm_ptp_cfg(bp);
3676 
3677 	if (!BNXT_CHIP_P5(bp))
3678 		bnxt_unmap_ptp_regs(bp);
3679 	else
3680 		bnxt_ptp_stop(bp);
3681 
3682 	return 0;
3683 }
3684 
3685 static int
3686 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3687 				 struct timespec *timestamp,
3688 				 uint32_t flags __rte_unused)
3689 {
3690 	struct bnxt *bp = dev->data->dev_private;
3691 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3692 	uint64_t rx_tstamp_cycles = 0;
3693 	uint64_t ns;
3694 
3695 	if (!ptp)
3696 		return -ENOTSUP;
3697 
3698 	if (BNXT_CHIP_P5(bp))
3699 		rx_tstamp_cycles = ptp->rx_timestamp;
3700 	else
3701 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3702 
3703 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3704 	*timestamp = rte_ns_to_timespec(ns);
3705 	return  0;
3706 }
3707 
3708 static int
3709 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3710 				 struct timespec *timestamp)
3711 {
3712 	struct bnxt *bp = dev->data->dev_private;
3713 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3714 	uint64_t tx_tstamp_cycles = 0;
3715 	uint64_t ns;
3716 	int rc = 0;
3717 
3718 	if (!ptp)
3719 		return -ENOTSUP;
3720 
3721 	if (BNXT_CHIP_P5(bp))
3722 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3723 					     &tx_tstamp_cycles);
3724 	else
3725 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3726 
3727 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3728 	*timestamp = rte_ns_to_timespec(ns);
3729 
3730 	return rc;
3731 }
3732 
3733 static int
3734 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3735 {
3736 	struct bnxt *bp = dev->data->dev_private;
3737 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3738 
3739 	if (!ptp)
3740 		return -ENOTSUP;
3741 
3742 	ptp->tc.nsec += delta;
3743 	ptp->tx_tstamp_tc.nsec += delta;
3744 	ptp->rx_tstamp_tc.nsec += delta;
3745 
3746 	return 0;
3747 }
3748 
3749 static int
3750 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3751 {
3752 	struct bnxt *bp = dev->data->dev_private;
3753 	int rc;
3754 	uint32_t dir_entries;
3755 	uint32_t entry_length;
3756 
3757 	rc = is_bnxt_in_error(bp);
3758 	if (rc)
3759 		return rc;
3760 
3761 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3762 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3763 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3764 
3765 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3766 	if (rc != 0)
3767 		return rc;
3768 
3769 	return dir_entries * entry_length;
3770 }
3771 
3772 static int
3773 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3774 		struct rte_dev_eeprom_info *in_eeprom)
3775 {
3776 	struct bnxt *bp = dev->data->dev_private;
3777 	uint32_t index;
3778 	uint32_t offset;
3779 	int rc;
3780 
3781 	rc = is_bnxt_in_error(bp);
3782 	if (rc)
3783 		return rc;
3784 
3785 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3786 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3787 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3788 		    in_eeprom->offset, in_eeprom->length);
3789 
3790 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3791 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3792 						in_eeprom->data);
3793 
3794 	index = in_eeprom->offset >> 24;
3795 	offset = in_eeprom->offset & 0xffffff;
3796 
3797 	if (index != 0)
3798 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3799 					   in_eeprom->length, in_eeprom->data);
3800 
3801 	return 0;
3802 }
3803 
3804 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3805 {
3806 	switch (dir_type) {
3807 	case BNX_DIR_TYPE_CHIMP_PATCH:
3808 	case BNX_DIR_TYPE_BOOTCODE:
3809 	case BNX_DIR_TYPE_BOOTCODE_2:
3810 	case BNX_DIR_TYPE_APE_FW:
3811 	case BNX_DIR_TYPE_APE_PATCH:
3812 	case BNX_DIR_TYPE_KONG_FW:
3813 	case BNX_DIR_TYPE_KONG_PATCH:
3814 	case BNX_DIR_TYPE_BONO_FW:
3815 	case BNX_DIR_TYPE_BONO_PATCH:
3816 		/* FALLTHROUGH */
3817 		return true;
3818 	}
3819 
3820 	return false;
3821 }
3822 
3823 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3824 {
3825 	switch (dir_type) {
3826 	case BNX_DIR_TYPE_AVS:
3827 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3828 	case BNX_DIR_TYPE_PCIE:
3829 	case BNX_DIR_TYPE_TSCF_UCODE:
3830 	case BNX_DIR_TYPE_EXT_PHY:
3831 	case BNX_DIR_TYPE_CCM:
3832 	case BNX_DIR_TYPE_ISCSI_BOOT:
3833 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3834 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3835 		/* FALLTHROUGH */
3836 		return true;
3837 	}
3838 
3839 	return false;
3840 }
3841 
3842 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3843 {
3844 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3845 		bnxt_dir_type_is_other_exec_format(dir_type);
3846 }
3847 
3848 static int
3849 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3850 		struct rte_dev_eeprom_info *in_eeprom)
3851 {
3852 	struct bnxt *bp = dev->data->dev_private;
3853 	uint8_t index, dir_op;
3854 	uint16_t type, ext, ordinal, attr;
3855 	int rc;
3856 
3857 	rc = is_bnxt_in_error(bp);
3858 	if (rc)
3859 		return rc;
3860 
3861 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3862 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3863 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3864 		    in_eeprom->offset, in_eeprom->length);
3865 
3866 	if (!BNXT_PF(bp)) {
3867 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3868 		return -EINVAL;
3869 	}
3870 
3871 	type = in_eeprom->magic >> 16;
3872 
3873 	if (type == 0xffff) { /* special value for directory operations */
3874 		index = in_eeprom->magic & 0xff;
3875 		dir_op = in_eeprom->magic >> 8;
3876 		if (index == 0)
3877 			return -EINVAL;
3878 		switch (dir_op) {
3879 		case 0x0e: /* erase */
3880 			if (in_eeprom->offset != ~in_eeprom->magic)
3881 				return -EINVAL;
3882 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3883 		default:
3884 			return -EINVAL;
3885 		}
3886 	}
3887 
3888 	/* Create or re-write an NVM item: */
3889 	if (bnxt_dir_type_is_executable(type) == true)
3890 		return -EOPNOTSUPP;
3891 	ext = in_eeprom->magic & 0xffff;
3892 	ordinal = in_eeprom->offset >> 16;
3893 	attr = in_eeprom->offset & 0xffff;
3894 
3895 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3896 				     in_eeprom->data, in_eeprom->length);
3897 }
3898 
3899 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3900 				struct rte_eth_dev_module_info *modinfo)
3901 {
3902 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3903 	struct bnxt *bp = dev->data->dev_private;
3904 	int rc;
3905 
3906 	/* No point in going further if phy status indicates
3907 	 * module is not inserted or if it is powered down or
3908 	 * if it is of type 10GBase-T
3909 	 */
3910 	if (bp->link_info->module_status >
3911 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3912 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3913 			    dev->data->port_id);
3914 		return -ENOTSUP;
3915 	}
3916 
3917 	/* This feature is not supported in older firmware versions */
3918 	if (bp->hwrm_spec_code < 0x10202) {
3919 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3920 			    dev->data->port_id);
3921 		return -ENOTSUP;
3922 	}
3923 
3924 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3925 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3926 						   module_info);
3927 
3928 	if (rc)
3929 		return rc;
3930 
3931 	switch (module_info[0]) {
3932 	case SFF_MODULE_ID_SFP:
3933 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3934 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3935 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3936 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3937 		break;
3938 	case SFF_MODULE_ID_QSFP:
3939 	case SFF_MODULE_ID_QSFP_PLUS:
3940 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3941 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3942 		break;
3943 	case SFF_MODULE_ID_QSFP28:
3944 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3945 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3946 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3947 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3948 		break;
3949 	default:
3950 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3951 		return -ENOTSUP;
3952 	}
3953 
3954 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3955 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3956 
3957 	return 0;
3958 }
3959 
3960 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3961 				  struct rte_dev_eeprom_info *info)
3962 {
3963 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3964 	uint32_t offset = info->offset, length = info->length;
3965 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3966 	struct bnxt *bp = dev->data->dev_private;
3967 	uint8_t *data = info->data;
3968 	uint8_t page = offset >> 7;
3969 	uint8_t max_pages = 2;
3970 	uint8_t opt_pages;
3971 	int rc;
3972 
3973 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3974 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3975 						   module_info);
3976 	if (rc)
3977 		return rc;
3978 
3979 	switch (module_info[0]) {
3980 	case SFF_MODULE_ID_SFP:
3981 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
3982 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
3983 			pg_addr[2] = I2C_DEV_ADDR_A2;
3984 			pg_addr[3] = I2C_DEV_ADDR_A2;
3985 			max_pages = 4;
3986 		}
3987 		break;
3988 	case SFF_MODULE_ID_QSFP28:
3989 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
3990 							   SFF8636_OPT_PAGES_OFFSET,
3991 							   1, &opt_pages);
3992 		if (rc)
3993 			return rc;
3994 
3995 		if (opt_pages & SFF8636_PAGE1_MASK) {
3996 			pg_addr[2] = I2C_DEV_ADDR_A0;
3997 			max_pages = 3;
3998 		}
3999 		if (opt_pages & SFF8636_PAGE2_MASK) {
4000 			pg_addr[3] = I2C_DEV_ADDR_A0;
4001 			max_pages = 4;
4002 		}
4003 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
4004 			pg_addr[4] = I2C_DEV_ADDR_A0;
4005 			max_pages = 5;
4006 		}
4007 		break;
4008 	default:
4009 		break;
4010 	}
4011 
4012 	memset(data, 0, length);
4013 
4014 	offset &= 0xff;
4015 	while (length && page < max_pages) {
4016 		uint8_t raw_page = page ? page - 1 : 0;
4017 		uint16_t chunk;
4018 
4019 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
4020 			raw_page = 0;
4021 		else if (page)
4022 			offset |= 0x80;
4023 		chunk = RTE_MIN(length, 256 - offset);
4024 
4025 		if (pg_addr[page]) {
4026 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
4027 								   raw_page, offset,
4028 								   chunk, data);
4029 			if (rc)
4030 				return rc;
4031 		}
4032 
4033 		data += chunk;
4034 		length -= chunk;
4035 		offset = 0;
4036 		page += 1 + (chunk > 128);
4037 	}
4038 
4039 	return length ? -EINVAL : 0;
4040 }
4041 
4042 /*
4043  * Initialization
4044  */
4045 
4046 static const struct eth_dev_ops bnxt_dev_ops = {
4047 	.dev_infos_get = bnxt_dev_info_get_op,
4048 	.dev_close = bnxt_dev_close_op,
4049 	.dev_configure = bnxt_dev_configure_op,
4050 	.dev_start = bnxt_dev_start_op,
4051 	.dev_stop = bnxt_dev_stop_op,
4052 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4053 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4054 	.stats_get = bnxt_stats_get_op,
4055 	.stats_reset = bnxt_stats_reset_op,
4056 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4057 	.rx_queue_release = bnxt_rx_queue_release_op,
4058 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4059 	.tx_queue_release = bnxt_tx_queue_release_op,
4060 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4061 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4062 	.reta_update = bnxt_reta_update_op,
4063 	.reta_query = bnxt_reta_query_op,
4064 	.rss_hash_update = bnxt_rss_hash_update_op,
4065 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4066 	.link_update = bnxt_link_update_op,
4067 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4068 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4069 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4070 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4071 	.mac_addr_add = bnxt_mac_addr_add_op,
4072 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4073 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4074 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4075 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4076 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4077 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4078 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4079 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4080 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4081 	.mtu_set = bnxt_mtu_set_op,
4082 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4083 	.xstats_get = bnxt_dev_xstats_get_op,
4084 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4085 	.xstats_reset = bnxt_dev_xstats_reset_op,
4086 	.fw_version_get = bnxt_fw_version_get,
4087 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4088 	.rxq_info_get = bnxt_rxq_info_get_op,
4089 	.txq_info_get = bnxt_txq_info_get_op,
4090 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4091 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4092 	.dev_led_on = bnxt_dev_led_on_op,
4093 	.dev_led_off = bnxt_dev_led_off_op,
4094 	.rx_queue_start = bnxt_rx_queue_start,
4095 	.rx_queue_stop = bnxt_rx_queue_stop,
4096 	.tx_queue_start = bnxt_tx_queue_start,
4097 	.tx_queue_stop = bnxt_tx_queue_stop,
4098 	.flow_ops_get = bnxt_flow_ops_get_op,
4099 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4100 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4101 	.get_eeprom           = bnxt_get_eeprom_op,
4102 	.set_eeprom           = bnxt_set_eeprom_op,
4103 	.get_module_info = bnxt_get_module_info,
4104 	.get_module_eeprom = bnxt_get_module_eeprom,
4105 	.timesync_enable      = bnxt_timesync_enable,
4106 	.timesync_disable     = bnxt_timesync_disable,
4107 	.timesync_read_time   = bnxt_timesync_read_time,
4108 	.timesync_write_time   = bnxt_timesync_write_time,
4109 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4110 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4111 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4112 };
4113 
4114 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4115 {
4116 	uint32_t offset;
4117 
4118 	/* Only pre-map the reset GRC registers using window 3 */
4119 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4120 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4121 
4122 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4123 
4124 	return offset;
4125 }
4126 
4127 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4128 {
4129 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4130 	uint32_t reg_base = 0xffffffff;
4131 	int i;
4132 
4133 	/* Only pre-map the monitoring GRC registers using window 2 */
4134 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4135 		uint32_t reg = info->status_regs[i];
4136 
4137 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4138 			continue;
4139 
4140 		if (reg_base == 0xffffffff)
4141 			reg_base = reg & 0xfffff000;
4142 		if ((reg & 0xfffff000) != reg_base)
4143 			return -ERANGE;
4144 
4145 		/* Use mask 0xffc as the Lower 2 bits indicates
4146 		 * address space location
4147 		 */
4148 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4149 						(reg & 0xffc);
4150 	}
4151 
4152 	if (reg_base == 0xffffffff)
4153 		return 0;
4154 
4155 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4156 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4157 
4158 	return 0;
4159 }
4160 
4161 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4162 {
4163 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4164 	uint32_t delay = info->delay_after_reset[index];
4165 	uint32_t val = info->reset_reg_val[index];
4166 	uint32_t reg = info->reset_reg[index];
4167 	uint32_t type, offset;
4168 	int ret;
4169 
4170 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4171 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4172 
4173 	switch (type) {
4174 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4175 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4176 		if (ret < 0) {
4177 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4178 				    val, offset);
4179 			return;
4180 		}
4181 		break;
4182 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4183 		offset = bnxt_map_reset_regs(bp, offset);
4184 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4185 		break;
4186 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4187 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4188 		break;
4189 	}
4190 	/* wait on a specific interval of time until core reset is complete */
4191 	if (delay)
4192 		rte_delay_ms(delay);
4193 }
4194 
4195 static void bnxt_dev_cleanup(struct bnxt *bp)
4196 {
4197 	bp->eth_dev->data->dev_link.link_status = 0;
4198 	bp->link_info->link_up = 0;
4199 	if (bp->eth_dev->data->dev_started)
4200 		bnxt_dev_stop(bp->eth_dev);
4201 
4202 	bnxt_uninit_resources(bp, true);
4203 }
4204 
4205 static int
4206 bnxt_check_fw_reset_done(struct bnxt *bp)
4207 {
4208 	int timeout = bp->fw_reset_max_msecs;
4209 	uint16_t val = 0;
4210 	int rc;
4211 
4212 	do {
4213 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4214 		if (rc < 0) {
4215 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4216 			return rc;
4217 		}
4218 		if (val != 0xffff)
4219 			break;
4220 		rte_delay_ms(1);
4221 	} while (timeout--);
4222 
4223 	if (val == 0xffff) {
4224 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4225 		return -1;
4226 	}
4227 
4228 	return 0;
4229 }
4230 
4231 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4232 {
4233 	struct rte_eth_dev *dev = bp->eth_dev;
4234 	struct rte_vlan_filter_conf *vfc;
4235 	int vidx, vbit, rc;
4236 	uint16_t vlan_id;
4237 
4238 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4239 		vfc = &dev->data->vlan_filter_conf;
4240 		vidx = vlan_id / 64;
4241 		vbit = vlan_id % 64;
4242 
4243 		/* Each bit corresponds to a VLAN id */
4244 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4245 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4246 			if (rc)
4247 				return rc;
4248 		}
4249 	}
4250 
4251 	return 0;
4252 }
4253 
4254 static int bnxt_restore_mac_filters(struct bnxt *bp)
4255 {
4256 	struct rte_eth_dev *dev = bp->eth_dev;
4257 	struct rte_eth_dev_info dev_info;
4258 	struct rte_ether_addr *addr;
4259 	uint64_t pool_mask;
4260 	uint32_t pool = 0;
4261 	uint32_t i;
4262 	int rc;
4263 
4264 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4265 		return 0;
4266 
4267 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4268 	if (rc)
4269 		return rc;
4270 
4271 	/* replay MAC address configuration */
4272 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4273 		addr = &dev->data->mac_addrs[i];
4274 
4275 		/* skip zero address */
4276 		if (rte_is_zero_ether_addr(addr))
4277 			continue;
4278 
4279 		pool = 0;
4280 		pool_mask = dev->data->mac_pool_sel[i];
4281 
4282 		do {
4283 			if (pool_mask & 1ULL) {
4284 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4285 				if (rc)
4286 					return rc;
4287 			}
4288 			pool_mask >>= 1;
4289 			pool++;
4290 		} while (pool_mask);
4291 	}
4292 
4293 	return 0;
4294 }
4295 
4296 static int bnxt_restore_filters(struct bnxt *bp)
4297 {
4298 	struct rte_eth_dev *dev = bp->eth_dev;
4299 	int ret = 0;
4300 
4301 	if (dev->data->all_multicast) {
4302 		ret = bnxt_allmulticast_enable_op(dev);
4303 		if (ret)
4304 			return ret;
4305 	}
4306 	if (dev->data->promiscuous) {
4307 		ret = bnxt_promiscuous_enable_op(dev);
4308 		if (ret)
4309 			return ret;
4310 	}
4311 
4312 	ret = bnxt_restore_mac_filters(bp);
4313 	if (ret)
4314 		return ret;
4315 
4316 	ret = bnxt_restore_vlan_filters(bp);
4317 	/* TODO restore other filters as well */
4318 	return ret;
4319 }
4320 
4321 static int bnxt_check_fw_ready(struct bnxt *bp)
4322 {
4323 	int timeout = bp->fw_reset_max_msecs;
4324 	int rc = 0;
4325 
4326 	do {
4327 		rc = bnxt_hwrm_poll_ver_get(bp);
4328 		if (rc == 0)
4329 			break;
4330 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4331 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4332 	} while (rc && timeout > 0);
4333 
4334 	if (rc)
4335 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4336 
4337 	return rc;
4338 }
4339 
4340 static void bnxt_dev_recover(void *arg)
4341 {
4342 	struct bnxt *bp = arg;
4343 	int rc = 0;
4344 
4345 	pthread_mutex_lock(&bp->err_recovery_lock);
4346 
4347 	if (!bp->fw_reset_min_msecs) {
4348 		rc = bnxt_check_fw_reset_done(bp);
4349 		if (rc)
4350 			goto err;
4351 	}
4352 
4353 	/* Clear Error flag so that device re-init should happen */
4354 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4355 
4356 	rc = bnxt_check_fw_ready(bp);
4357 	if (rc)
4358 		goto err;
4359 
4360 	rc = bnxt_init_resources(bp, true);
4361 	if (rc) {
4362 		PMD_DRV_LOG(ERR,
4363 			    "Failed to initialize resources after reset\n");
4364 		goto err;
4365 	}
4366 	/* clear reset flag as the device is initialized now */
4367 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4368 
4369 	rc = bnxt_dev_start_op(bp->eth_dev);
4370 	if (rc) {
4371 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4372 		goto err_start;
4373 	}
4374 
4375 	rc = bnxt_restore_filters(bp);
4376 	if (rc)
4377 		goto err_start;
4378 
4379 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4380 	pthread_mutex_unlock(&bp->err_recovery_lock);
4381 
4382 	return;
4383 err_start:
4384 	bnxt_dev_stop(bp->eth_dev);
4385 err:
4386 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4387 	bnxt_uninit_resources(bp, false);
4388 	if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4389 		rte_eth_dev_callback_process(bp->eth_dev,
4390 					     RTE_ETH_EVENT_INTR_RMV,
4391 					     NULL);
4392 	pthread_mutex_unlock(&bp->err_recovery_lock);
4393 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4394 }
4395 
4396 void bnxt_dev_reset_and_resume(void *arg)
4397 {
4398 	struct bnxt *bp = arg;
4399 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4400 	uint16_t val = 0;
4401 	int rc;
4402 
4403 	bnxt_dev_cleanup(bp);
4404 
4405 	bnxt_wait_for_device_shutdown(bp);
4406 
4407 	/* During some fatal firmware error conditions, the PCI config space
4408 	 * register 0x2e which normally contains the subsystem ID will become
4409 	 * 0xffff. This register will revert back to the normal value after
4410 	 * the chip has completed core reset. If we detect this condition,
4411 	 * we can poll this config register immediately for the value to revert.
4412 	 */
4413 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4414 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4415 		if (rc < 0) {
4416 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4417 			return;
4418 		}
4419 		if (val == 0xffff) {
4420 			bp->fw_reset_min_msecs = 0;
4421 			us = 1;
4422 		}
4423 	}
4424 
4425 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4426 	if (rc)
4427 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4428 }
4429 
4430 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4431 {
4432 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4433 	uint32_t reg = info->status_regs[index];
4434 	uint32_t type, offset, val = 0;
4435 	int ret = 0;
4436 
4437 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4438 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4439 
4440 	switch (type) {
4441 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4442 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4443 		if (ret < 0)
4444 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4445 				    offset);
4446 		break;
4447 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4448 		offset = info->mapped_status_regs[index];
4449 		/* FALLTHROUGH */
4450 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4451 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4452 				       offset));
4453 		break;
4454 	}
4455 
4456 	return val;
4457 }
4458 
4459 static int bnxt_fw_reset_all(struct bnxt *bp)
4460 {
4461 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4462 	uint32_t i;
4463 	int rc = 0;
4464 
4465 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4466 		/* Reset through primary function driver */
4467 		for (i = 0; i < info->reg_array_cnt; i++)
4468 			bnxt_write_fw_reset_reg(bp, i);
4469 		/* Wait for time specified by FW after triggering reset */
4470 		rte_delay_ms(info->primary_func_wait_period_after_reset);
4471 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4472 		/* Reset with the help of Kong processor */
4473 		rc = bnxt_hwrm_fw_reset(bp);
4474 		if (rc)
4475 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4476 	}
4477 
4478 	return rc;
4479 }
4480 
4481 static void bnxt_fw_reset_cb(void *arg)
4482 {
4483 	struct bnxt *bp = arg;
4484 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4485 	int rc = 0;
4486 
4487 	/* Only Primary function can do FW reset */
4488 	if (bnxt_is_primary_func(bp) &&
4489 	    bnxt_is_recovery_enabled(bp)) {
4490 		rc = bnxt_fw_reset_all(bp);
4491 		if (rc) {
4492 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4493 			return;
4494 		}
4495 	}
4496 
4497 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4498 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4499 	 * (including MASTER FUNC). After receiving this Async, all the active
4500 	 * drivers should treat this case as FW initiated recovery
4501 	 */
4502 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4503 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4504 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4505 
4506 		/* To recover from error */
4507 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4508 				  (void *)bp);
4509 	}
4510 }
4511 
4512 /* Driver should poll FW heartbeat, reset_counter with the frequency
4513  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4514  * When the driver detects heartbeat stop or change in reset_counter,
4515  * it has to trigger a reset to recover from the error condition.
4516  * A “primary function” is the function who will have the privilege to
4517  * initiate the chimp reset. The primary function will be elected by the
4518  * firmware and will be notified through async message.
4519  */
4520 static void bnxt_check_fw_health(void *arg)
4521 {
4522 	struct bnxt *bp = arg;
4523 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4524 	uint32_t val = 0, wait_msec;
4525 
4526 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4527 	    is_bnxt_in_error(bp))
4528 		return;
4529 
4530 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4531 	if (val == info->last_heart_beat)
4532 		goto reset;
4533 
4534 	info->last_heart_beat = val;
4535 
4536 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4537 	if (val != info->last_reset_counter)
4538 		goto reset;
4539 
4540 	info->last_reset_counter = val;
4541 
4542 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4543 			  bnxt_check_fw_health, (void *)bp);
4544 
4545 	return;
4546 reset:
4547 	/* Stop DMA to/from device */
4548 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4549 	bp->flags |= BNXT_FLAG_FW_RESET;
4550 
4551 	bnxt_stop_rxtx(bp);
4552 
4553 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4554 
4555 	if (bnxt_is_primary_func(bp))
4556 		wait_msec = info->primary_func_wait_period;
4557 	else
4558 		wait_msec = info->normal_func_wait_period;
4559 
4560 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4561 			  bnxt_fw_reset_cb, (void *)bp);
4562 }
4563 
4564 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4565 {
4566 	uint32_t polling_freq;
4567 
4568 	pthread_mutex_lock(&bp->health_check_lock);
4569 
4570 	if (!bnxt_is_recovery_enabled(bp))
4571 		goto done;
4572 
4573 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4574 		goto done;
4575 
4576 	polling_freq = bp->recovery_info->driver_polling_freq;
4577 
4578 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4579 			  bnxt_check_fw_health, (void *)bp);
4580 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4581 
4582 done:
4583 	pthread_mutex_unlock(&bp->health_check_lock);
4584 }
4585 
4586 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4587 {
4588 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4589 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4590 }
4591 
4592 static bool bnxt_vf_pciid(uint16_t device_id)
4593 {
4594 	switch (device_id) {
4595 	case BROADCOM_DEV_ID_57304_VF:
4596 	case BROADCOM_DEV_ID_57406_VF:
4597 	case BROADCOM_DEV_ID_5731X_VF:
4598 	case BROADCOM_DEV_ID_5741X_VF:
4599 	case BROADCOM_DEV_ID_57414_VF:
4600 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4601 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4602 	case BROADCOM_DEV_ID_58802_VF:
4603 	case BROADCOM_DEV_ID_57500_VF1:
4604 	case BROADCOM_DEV_ID_57500_VF2:
4605 	case BROADCOM_DEV_ID_58818_VF:
4606 		/* FALLTHROUGH */
4607 		return true;
4608 	default:
4609 		return false;
4610 	}
4611 }
4612 
4613 /* Phase 5 device */
4614 static bool bnxt_p5_device(uint16_t device_id)
4615 {
4616 	switch (device_id) {
4617 	case BROADCOM_DEV_ID_57508:
4618 	case BROADCOM_DEV_ID_57504:
4619 	case BROADCOM_DEV_ID_57502:
4620 	case BROADCOM_DEV_ID_57508_MF1:
4621 	case BROADCOM_DEV_ID_57504_MF1:
4622 	case BROADCOM_DEV_ID_57502_MF1:
4623 	case BROADCOM_DEV_ID_57508_MF2:
4624 	case BROADCOM_DEV_ID_57504_MF2:
4625 	case BROADCOM_DEV_ID_57502_MF2:
4626 	case BROADCOM_DEV_ID_57500_VF1:
4627 	case BROADCOM_DEV_ID_57500_VF2:
4628 	case BROADCOM_DEV_ID_58812:
4629 	case BROADCOM_DEV_ID_58814:
4630 	case BROADCOM_DEV_ID_58818:
4631 	case BROADCOM_DEV_ID_58818_VF:
4632 		/* FALLTHROUGH */
4633 		return true;
4634 	default:
4635 		return false;
4636 	}
4637 }
4638 
4639 bool bnxt_stratus_device(struct bnxt *bp)
4640 {
4641 	uint16_t device_id = bp->pdev->id.device_id;
4642 
4643 	switch (device_id) {
4644 	case BROADCOM_DEV_ID_STRATUS_NIC:
4645 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4646 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4647 		/* FALLTHROUGH */
4648 		return true;
4649 	default:
4650 		return false;
4651 	}
4652 }
4653 
4654 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4655 {
4656 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4657 	struct bnxt *bp = eth_dev->data->dev_private;
4658 
4659 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4660 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4661 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4662 	if (!bp->bar0 || !bp->doorbell_base) {
4663 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4664 		return -ENODEV;
4665 	}
4666 
4667 	bp->eth_dev = eth_dev;
4668 	bp->pdev = pci_dev;
4669 
4670 	return 0;
4671 }
4672 
4673 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4674 				  struct bnxt_ctx_pg_info *ctx_pg,
4675 				  uint32_t mem_size,
4676 				  const char *suffix,
4677 				  uint16_t idx)
4678 {
4679 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4680 	const struct rte_memzone *mz = NULL;
4681 	char mz_name[RTE_MEMZONE_NAMESIZE];
4682 	rte_iova_t mz_phys_addr;
4683 	uint64_t valid_bits = 0;
4684 	uint32_t sz;
4685 	int i;
4686 
4687 	if (!mem_size)
4688 		return 0;
4689 
4690 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4691 			 BNXT_PAGE_SIZE;
4692 	rmem->page_size = BNXT_PAGE_SIZE;
4693 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4694 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4695 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4696 
4697 	valid_bits = PTU_PTE_VALID;
4698 
4699 	if (rmem->nr_pages > 1) {
4700 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4701 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4702 			 suffix, idx, bp->eth_dev->data->port_id);
4703 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4704 		mz = rte_memzone_lookup(mz_name);
4705 		if (!mz) {
4706 			mz = rte_memzone_reserve_aligned(mz_name,
4707 						rmem->nr_pages * 8,
4708 						bp->eth_dev->device->numa_node,
4709 						RTE_MEMZONE_2MB |
4710 						RTE_MEMZONE_SIZE_HINT_ONLY |
4711 						RTE_MEMZONE_IOVA_CONTIG,
4712 						BNXT_PAGE_SIZE);
4713 			if (mz == NULL)
4714 				return -ENOMEM;
4715 		}
4716 
4717 		memset(mz->addr, 0, mz->len);
4718 		mz_phys_addr = mz->iova;
4719 
4720 		rmem->pg_tbl = mz->addr;
4721 		rmem->pg_tbl_map = mz_phys_addr;
4722 		rmem->pg_tbl_mz = mz;
4723 	}
4724 
4725 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4726 		 suffix, idx, bp->eth_dev->data->port_id);
4727 	mz = rte_memzone_lookup(mz_name);
4728 	if (!mz) {
4729 		mz = rte_memzone_reserve_aligned(mz_name,
4730 						 mem_size,
4731 						 bp->eth_dev->device->numa_node,
4732 						 RTE_MEMZONE_1GB |
4733 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4734 						 RTE_MEMZONE_IOVA_CONTIG,
4735 						 BNXT_PAGE_SIZE);
4736 		if (mz == NULL)
4737 			return -ENOMEM;
4738 	}
4739 
4740 	memset(mz->addr, 0, mz->len);
4741 	mz_phys_addr = mz->iova;
4742 
4743 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4744 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4745 		rmem->dma_arr[i] = mz_phys_addr + sz;
4746 
4747 		if (rmem->nr_pages > 1) {
4748 			if (i == rmem->nr_pages - 2 &&
4749 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4750 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4751 			else if (i == rmem->nr_pages - 1 &&
4752 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4753 				valid_bits |= PTU_PTE_LAST;
4754 
4755 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4756 							   valid_bits);
4757 		}
4758 	}
4759 
4760 	rmem->mz = mz;
4761 	if (rmem->vmem_size)
4762 		rmem->vmem = (void **)mz->addr;
4763 	rmem->dma_arr[0] = mz_phys_addr;
4764 	return 0;
4765 }
4766 
4767 static void bnxt_free_ctx_mem(struct bnxt *bp)
4768 {
4769 	int i;
4770 
4771 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4772 		return;
4773 
4774 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4775 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4776 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4777 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4778 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4779 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4780 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4781 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4782 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4783 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4784 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4785 
4786 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4787 		if (bp->ctx->tqm_mem[i])
4788 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4789 	}
4790 
4791 	rte_free(bp->ctx);
4792 	bp->ctx = NULL;
4793 }
4794 
4795 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4796 
4797 #define min_t(type, x, y) ({                    \
4798 	type __min1 = (x);                      \
4799 	type __min2 = (y);                      \
4800 	__min1 < __min2 ? __min1 : __min2; })
4801 
4802 #define max_t(type, x, y) ({                    \
4803 	type __max1 = (x);                      \
4804 	type __max2 = (y);                      \
4805 	__max1 > __max2 ? __max1 : __max2; })
4806 
4807 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4808 
4809 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4810 {
4811 	struct bnxt_ctx_pg_info *ctx_pg;
4812 	struct bnxt_ctx_mem_info *ctx;
4813 	uint32_t mem_size, ena, entries;
4814 	uint32_t entries_sp, min;
4815 	int i, rc;
4816 
4817 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4818 	if (rc) {
4819 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4820 		return rc;
4821 	}
4822 	ctx = bp->ctx;
4823 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4824 		return 0;
4825 
4826 	ctx_pg = &ctx->qp_mem;
4827 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4828 	if (ctx->qp_entry_size) {
4829 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4830 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4831 		if (rc)
4832 			return rc;
4833 	}
4834 
4835 	ctx_pg = &ctx->srq_mem;
4836 	ctx_pg->entries = ctx->srq_max_l2_entries;
4837 	if (ctx->srq_entry_size) {
4838 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4839 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4840 		if (rc)
4841 			return rc;
4842 	}
4843 
4844 	ctx_pg = &ctx->cq_mem;
4845 	ctx_pg->entries = ctx->cq_max_l2_entries;
4846 	if (ctx->cq_entry_size) {
4847 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4848 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4849 		if (rc)
4850 			return rc;
4851 	}
4852 
4853 	ctx_pg = &ctx->vnic_mem;
4854 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4855 		ctx->vnic_max_ring_table_entries;
4856 	if (ctx->vnic_entry_size) {
4857 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4858 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4859 		if (rc)
4860 			return rc;
4861 	}
4862 
4863 	ctx_pg = &ctx->stat_mem;
4864 	ctx_pg->entries = ctx->stat_max_entries;
4865 	if (ctx->stat_entry_size) {
4866 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4867 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4868 		if (rc)
4869 			return rc;
4870 	}
4871 
4872 	min = ctx->tqm_min_entries_per_ring;
4873 
4874 	entries_sp = ctx->qp_max_l2_entries +
4875 		     ctx->vnic_max_vnic_entries +
4876 		     2 * ctx->qp_min_qp1_entries + min;
4877 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4878 
4879 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4880 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4881 	entries = clamp_t(uint32_t, entries, min,
4882 			  ctx->tqm_max_entries_per_ring);
4883 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4884 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4885 		 * i > 8 is other ext rings.
4886 		 */
4887 		ctx_pg = ctx->tqm_mem[i];
4888 		ctx_pg->entries = i ? entries : entries_sp;
4889 		if (ctx->tqm_entry_size) {
4890 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4891 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4892 						    "tqm_mem", i);
4893 			if (rc)
4894 				return rc;
4895 		}
4896 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4897 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4898 		else
4899 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4900 	}
4901 
4902 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4903 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4904 	if (rc)
4905 		PMD_DRV_LOG(ERR,
4906 			    "Failed to configure context mem: rc = %d\n", rc);
4907 	else
4908 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4909 
4910 	return rc;
4911 }
4912 
4913 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4914 {
4915 	struct rte_pci_device *pci_dev = bp->pdev;
4916 	char mz_name[RTE_MEMZONE_NAMESIZE];
4917 	const struct rte_memzone *mz = NULL;
4918 	uint32_t total_alloc_len;
4919 	rte_iova_t mz_phys_addr;
4920 
4921 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4922 		return 0;
4923 
4924 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4925 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4926 		 pci_dev->addr.bus, pci_dev->addr.devid,
4927 		 pci_dev->addr.function, "rx_port_stats");
4928 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4929 	mz = rte_memzone_lookup(mz_name);
4930 	total_alloc_len =
4931 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4932 				       sizeof(struct rx_port_stats_ext) + 512);
4933 	if (!mz) {
4934 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4935 					 SOCKET_ID_ANY,
4936 					 RTE_MEMZONE_2MB |
4937 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4938 					 RTE_MEMZONE_IOVA_CONTIG);
4939 		if (mz == NULL)
4940 			return -ENOMEM;
4941 	}
4942 	memset(mz->addr, 0, mz->len);
4943 	mz_phys_addr = mz->iova;
4944 
4945 	bp->rx_mem_zone = (const void *)mz;
4946 	bp->hw_rx_port_stats = mz->addr;
4947 	bp->hw_rx_port_stats_map = mz_phys_addr;
4948 
4949 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4950 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4951 		 pci_dev->addr.bus, pci_dev->addr.devid,
4952 		 pci_dev->addr.function, "tx_port_stats");
4953 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4954 	mz = rte_memzone_lookup(mz_name);
4955 	total_alloc_len =
4956 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4957 				       sizeof(struct tx_port_stats_ext) + 512);
4958 	if (!mz) {
4959 		mz = rte_memzone_reserve(mz_name,
4960 					 total_alloc_len,
4961 					 SOCKET_ID_ANY,
4962 					 RTE_MEMZONE_2MB |
4963 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4964 					 RTE_MEMZONE_IOVA_CONTIG);
4965 		if (mz == NULL)
4966 			return -ENOMEM;
4967 	}
4968 	memset(mz->addr, 0, mz->len);
4969 	mz_phys_addr = mz->iova;
4970 
4971 	bp->tx_mem_zone = (const void *)mz;
4972 	bp->hw_tx_port_stats = mz->addr;
4973 	bp->hw_tx_port_stats_map = mz_phys_addr;
4974 	bp->flags |= BNXT_FLAG_PORT_STATS;
4975 
4976 	/* Display extended statistics if FW supports it */
4977 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4978 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4979 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4980 		return 0;
4981 
4982 	bp->hw_rx_port_stats_ext = (void *)
4983 		((uint8_t *)bp->hw_rx_port_stats +
4984 		 sizeof(struct rx_port_stats));
4985 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4986 		sizeof(struct rx_port_stats);
4987 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4988 
4989 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4990 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4991 		bp->hw_tx_port_stats_ext = (void *)
4992 			((uint8_t *)bp->hw_tx_port_stats +
4993 			 sizeof(struct tx_port_stats));
4994 		bp->hw_tx_port_stats_ext_map =
4995 			bp->hw_tx_port_stats_map +
4996 			sizeof(struct tx_port_stats);
4997 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4998 	}
4999 
5000 	return 0;
5001 }
5002 
5003 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
5004 {
5005 	struct bnxt *bp = eth_dev->data->dev_private;
5006 	int rc = 0;
5007 
5008 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
5009 					       RTE_ETHER_ADDR_LEN *
5010 					       bp->max_l2_ctx,
5011 					       0);
5012 	if (eth_dev->data->mac_addrs == NULL) {
5013 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
5014 		return -ENOMEM;
5015 	}
5016 
5017 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
5018 		if (BNXT_PF(bp))
5019 			return -EINVAL;
5020 
5021 		/* Generate a random MAC address, if none was assigned by PF */
5022 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
5023 		bnxt_eth_hw_addr_random(bp->mac_addr);
5024 		PMD_DRV_LOG(INFO,
5025 			    "Assign random MAC:" RTE_ETHER_ADDR_PRT_FMT "\n",
5026 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
5027 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
5028 
5029 		rc = bnxt_hwrm_set_mac(bp);
5030 		if (rc)
5031 			return rc;
5032 	}
5033 
5034 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5035 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5036 
5037 	return rc;
5038 }
5039 
5040 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5041 {
5042 	int rc = 0;
5043 
5044 	/* MAC is already configured in FW */
5045 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5046 		return 0;
5047 
5048 	/* Restore the old MAC configured */
5049 	rc = bnxt_hwrm_set_mac(bp);
5050 	if (rc)
5051 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5052 
5053 	return rc;
5054 }
5055 
5056 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5057 {
5058 	if (!BNXT_PF(bp))
5059 		return;
5060 
5061 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5062 
5063 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5064 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5065 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5066 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5067 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5068 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5069 }
5070 
5071 struct bnxt *
5072 bnxt_get_bp(uint16_t port)
5073 {
5074 	struct bnxt *bp;
5075 	struct rte_eth_dev *dev;
5076 
5077 	if (!rte_eth_dev_is_valid_port(port)) {
5078 		PMD_DRV_LOG(ERR, "Invalid port %d\n", port);
5079 		return NULL;
5080 	}
5081 
5082 	dev = &rte_eth_devices[port];
5083 	if (!is_bnxt_supported(dev)) {
5084 		PMD_DRV_LOG(ERR, "Device %d not supported\n", port);
5085 		return NULL;
5086 	}
5087 
5088 	bp = (struct bnxt *)dev->data->dev_private;
5089 	if (!BNXT_TRUFLOW_EN(bp)) {
5090 		PMD_DRV_LOG(ERR, "TRUFLOW not enabled\n");
5091 		return NULL;
5092 	}
5093 
5094 	return bp;
5095 }
5096 
5097 uint16_t
5098 bnxt_get_svif(uint16_t port_id, bool func_svif,
5099 	      enum bnxt_ulp_intf_type type)
5100 {
5101 	struct rte_eth_dev *eth_dev;
5102 	struct bnxt *bp;
5103 
5104 	eth_dev = &rte_eth_devices[port_id];
5105 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5106 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5107 		if (!vfr)
5108 			return 0;
5109 
5110 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5111 			return vfr->svif;
5112 
5113 		eth_dev = vfr->parent_dev;
5114 	}
5115 
5116 	bp = eth_dev->data->dev_private;
5117 
5118 	return func_svif ? bp->func_svif : bp->port_svif;
5119 }
5120 
5121 void
5122 bnxt_get_iface_mac(uint16_t port, enum bnxt_ulp_intf_type type,
5123 		   uint8_t *mac, uint8_t *parent_mac)
5124 {
5125 	struct rte_eth_dev *eth_dev;
5126 	struct bnxt *bp;
5127 
5128 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF &&
5129 	    type != BNXT_ULP_INTF_TYPE_PF)
5130 		return;
5131 
5132 	eth_dev = &rte_eth_devices[port];
5133 	bp = eth_dev->data->dev_private;
5134 	memcpy(mac, bp->mac_addr, RTE_ETHER_ADDR_LEN);
5135 
5136 	if (type == BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5137 		memcpy(parent_mac, bp->parent->mac_addr, RTE_ETHER_ADDR_LEN);
5138 }
5139 
5140 uint16_t
5141 bnxt_get_parent_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5142 {
5143 	struct rte_eth_dev *eth_dev;
5144 	struct bnxt *bp;
5145 
5146 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5147 		return 0;
5148 
5149 	eth_dev = &rte_eth_devices[port];
5150 	bp = eth_dev->data->dev_private;
5151 
5152 	return bp->parent->vnic;
5153 }
5154 uint16_t
5155 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5156 {
5157 	struct rte_eth_dev *eth_dev;
5158 	struct bnxt_vnic_info *vnic;
5159 	struct bnxt *bp;
5160 
5161 	eth_dev = &rte_eth_devices[port];
5162 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5163 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5164 		if (!vfr)
5165 			return 0;
5166 
5167 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5168 			return vfr->dflt_vnic_id;
5169 
5170 		eth_dev = vfr->parent_dev;
5171 	}
5172 
5173 	bp = eth_dev->data->dev_private;
5174 
5175 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
5176 
5177 	return vnic->fw_vnic_id;
5178 }
5179 
5180 uint16_t
5181 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
5182 {
5183 	struct rte_eth_dev *eth_dev;
5184 	struct bnxt *bp;
5185 
5186 	eth_dev = &rte_eth_devices[port];
5187 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5188 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5189 		if (!vfr)
5190 			return 0;
5191 
5192 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5193 			return vfr->fw_fid;
5194 
5195 		eth_dev = vfr->parent_dev;
5196 	}
5197 
5198 	bp = eth_dev->data->dev_private;
5199 
5200 	return bp->fw_fid;
5201 }
5202 
5203 enum bnxt_ulp_intf_type
5204 bnxt_get_interface_type(uint16_t port)
5205 {
5206 	struct rte_eth_dev *eth_dev;
5207 	struct bnxt *bp;
5208 
5209 	eth_dev = &rte_eth_devices[port];
5210 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
5211 		return BNXT_ULP_INTF_TYPE_VF_REP;
5212 
5213 	bp = eth_dev->data->dev_private;
5214 	if (BNXT_PF(bp))
5215 		return BNXT_ULP_INTF_TYPE_PF;
5216 	else if (BNXT_VF_IS_TRUSTED(bp))
5217 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
5218 	else if (BNXT_VF(bp))
5219 		return BNXT_ULP_INTF_TYPE_VF;
5220 
5221 	return BNXT_ULP_INTF_TYPE_INVALID;
5222 }
5223 
5224 uint16_t
5225 bnxt_get_phy_port_id(uint16_t port_id)
5226 {
5227 	struct bnxt_representor *vfr;
5228 	struct rte_eth_dev *eth_dev;
5229 	struct bnxt *bp;
5230 
5231 	eth_dev = &rte_eth_devices[port_id];
5232 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5233 		vfr = eth_dev->data->dev_private;
5234 		if (!vfr)
5235 			return 0;
5236 
5237 		eth_dev = vfr->parent_dev;
5238 	}
5239 
5240 	bp = eth_dev->data->dev_private;
5241 
5242 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
5243 }
5244 
5245 uint16_t
5246 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
5247 {
5248 	struct rte_eth_dev *eth_dev;
5249 	struct bnxt *bp;
5250 
5251 	eth_dev = &rte_eth_devices[port_id];
5252 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5253 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5254 		if (!vfr)
5255 			return 0;
5256 
5257 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5258 			return vfr->fw_fid - 1;
5259 
5260 		eth_dev = vfr->parent_dev;
5261 	}
5262 
5263 	bp = eth_dev->data->dev_private;
5264 
5265 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
5266 }
5267 
5268 uint16_t
5269 bnxt_get_vport(uint16_t port_id)
5270 {
5271 	return (1 << bnxt_get_phy_port_id(port_id));
5272 }
5273 
5274 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5275 {
5276 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5277 
5278 	if (info) {
5279 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5280 			memset(info, 0, sizeof(*info));
5281 		return;
5282 	}
5283 
5284 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5285 		return;
5286 
5287 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5288 			   sizeof(*info), 0);
5289 	if (!info)
5290 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5291 
5292 	bp->recovery_info = info;
5293 }
5294 
5295 static void bnxt_check_fw_status(struct bnxt *bp)
5296 {
5297 	uint32_t fw_status;
5298 
5299 	if (!(bp->recovery_info &&
5300 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5301 		return;
5302 
5303 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5304 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5305 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5306 			    fw_status);
5307 }
5308 
5309 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5310 {
5311 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5312 	uint32_t status_loc;
5313 	uint32_t sig_ver;
5314 
5315 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5316 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5317 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5318 				   BNXT_GRCP_WINDOW_2_BASE +
5319 				   offsetof(struct hcomm_status,
5320 					    sig_ver)));
5321 	/* If the signature is absent, then FW does not support this feature */
5322 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5323 	    HCOMM_STATUS_SIGNATURE_VAL)
5324 		return 0;
5325 
5326 	if (!info) {
5327 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5328 				   sizeof(*info), 0);
5329 		if (!info)
5330 			return -ENOMEM;
5331 		bp->recovery_info = info;
5332 	} else {
5333 		memset(info, 0, sizeof(*info));
5334 	}
5335 
5336 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5337 				      BNXT_GRCP_WINDOW_2_BASE +
5338 				      offsetof(struct hcomm_status,
5339 					       fw_status_loc)));
5340 
5341 	/* Only pre-map the FW health status GRC register */
5342 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5343 		return 0;
5344 
5345 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5346 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5347 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5348 
5349 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5350 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5351 
5352 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5353 
5354 	return 0;
5355 }
5356 
5357 /* This function gets the FW version along with the
5358  * capabilities(MAX and current) of the function, vnic,
5359  * error recovery, phy and other chip related info
5360  */
5361 static int bnxt_get_config(struct bnxt *bp)
5362 {
5363 	uint16_t mtu;
5364 	int rc = 0;
5365 
5366 	bp->fw_cap = 0;
5367 
5368 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5369 	if (rc)
5370 		return rc;
5371 
5372 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5373 	if (rc) {
5374 		bnxt_check_fw_status(bp);
5375 		return rc;
5376 	}
5377 
5378 	rc = bnxt_hwrm_func_reset(bp);
5379 	if (rc)
5380 		return -EIO;
5381 
5382 	rc = bnxt_hwrm_vnic_qcaps(bp);
5383 	if (rc)
5384 		return rc;
5385 
5386 	rc = bnxt_hwrm_queue_qportcfg(bp);
5387 	if (rc)
5388 		return rc;
5389 
5390 	/* Get the MAX capabilities for this function.
5391 	 * This function also allocates context memory for TQM rings and
5392 	 * informs the firmware about this allocated backing store memory.
5393 	 */
5394 	rc = bnxt_hwrm_func_qcaps(bp);
5395 	if (rc)
5396 		return rc;
5397 
5398 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5399 	if (rc)
5400 		return rc;
5401 
5402 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5403 	if (rc)
5404 		return rc;
5405 
5406 	bnxt_hwrm_port_mac_qcfg(bp);
5407 
5408 	bnxt_hwrm_parent_pf_qcfg(bp);
5409 
5410 	bnxt_hwrm_port_phy_qcaps(bp);
5411 
5412 	bnxt_alloc_error_recovery_info(bp);
5413 	/* Get the adapter error recovery support info */
5414 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5415 	if (rc)
5416 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5417 
5418 	bnxt_hwrm_port_led_qcaps(bp);
5419 
5420 	return 0;
5421 }
5422 
5423 static int
5424 bnxt_init_locks(struct bnxt *bp)
5425 {
5426 	int err;
5427 
5428 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5429 	if (err) {
5430 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5431 		return err;
5432 	}
5433 
5434 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5435 	if (err) {
5436 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5437 		return err;
5438 	}
5439 
5440 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5441 	if (err) {
5442 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5443 		return err;
5444 	}
5445 
5446 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5447 	if (err)
5448 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5449 
5450 	return err;
5451 }
5452 
5453 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5454 {
5455 	int rc = 0;
5456 
5457 	rc = bnxt_get_config(bp);
5458 	if (rc)
5459 		return rc;
5460 
5461 	if (!reconfig_dev) {
5462 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5463 		if (rc)
5464 			return rc;
5465 	} else {
5466 		rc = bnxt_restore_dflt_mac(bp);
5467 		if (rc)
5468 			return rc;
5469 	}
5470 
5471 	bnxt_config_vf_req_fwd(bp);
5472 
5473 	rc = bnxt_hwrm_func_driver_register(bp);
5474 	if (rc) {
5475 		PMD_DRV_LOG(ERR, "Failed to register driver");
5476 		return -EBUSY;
5477 	}
5478 
5479 	if (BNXT_PF(bp)) {
5480 		if (bp->pdev->max_vfs) {
5481 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5482 			if (rc) {
5483 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5484 				return rc;
5485 			}
5486 		} else {
5487 			rc = bnxt_hwrm_allocate_pf_only(bp);
5488 			if (rc) {
5489 				PMD_DRV_LOG(ERR,
5490 					    "Failed to allocate PF resources");
5491 				return rc;
5492 			}
5493 		}
5494 	}
5495 
5496 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5497 	if (rc)
5498 		return rc;
5499 
5500 	rc = bnxt_setup_int(bp);
5501 	if (rc)
5502 		return rc;
5503 
5504 	rc = bnxt_request_int(bp);
5505 	if (rc)
5506 		return rc;
5507 
5508 	rc = bnxt_init_ctx_mem(bp);
5509 	if (rc) {
5510 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5511 		return rc;
5512 	}
5513 
5514 	return 0;
5515 }
5516 
5517 static int
5518 bnxt_parse_devarg_accum_stats(__rte_unused const char *key,
5519 			      const char *value, void *opaque_arg)
5520 {
5521 	struct bnxt *bp = opaque_arg;
5522 	unsigned long accum_stats;
5523 	char *end = NULL;
5524 
5525 	if (!value || !opaque_arg) {
5526 		PMD_DRV_LOG(ERR,
5527 			    "Invalid parameter passed to accum-stats devargs.\n");
5528 		return -EINVAL;
5529 	}
5530 
5531 	accum_stats = strtoul(value, &end, 10);
5532 	if (end == NULL || *end != '\0' ||
5533 	    (accum_stats == ULONG_MAX && errno == ERANGE)) {
5534 		PMD_DRV_LOG(ERR,
5535 			    "Invalid parameter passed to accum-stats devargs.\n");
5536 		return -EINVAL;
5537 	}
5538 
5539 	if (BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)) {
5540 		PMD_DRV_LOG(ERR,
5541 			    "Invalid value passed to accum-stats devargs.\n");
5542 		return -EINVAL;
5543 	}
5544 
5545 	if (accum_stats) {
5546 		bp->flags2 |= BNXT_FLAGS2_ACCUM_STATS_EN;
5547 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature enabled.\n");
5548 	} else {
5549 		bp->flags2 &= ~BNXT_FLAGS2_ACCUM_STATS_EN;
5550 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature disabled.\n");
5551 	}
5552 
5553 	return 0;
5554 }
5555 
5556 static int
5557 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5558 			     const char *value, void *opaque_arg)
5559 {
5560 	struct bnxt *bp = opaque_arg;
5561 	unsigned long flow_xstat;
5562 	char *end = NULL;
5563 
5564 	if (!value || !opaque_arg) {
5565 		PMD_DRV_LOG(ERR,
5566 			    "Invalid parameter passed to flow_xstat devarg.\n");
5567 		return -EINVAL;
5568 	}
5569 
5570 	flow_xstat = strtoul(value, &end, 10);
5571 	if (end == NULL || *end != '\0' ||
5572 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5573 		PMD_DRV_LOG(ERR,
5574 			    "Invalid parameter passed to flow_xstat devarg.\n");
5575 		return -EINVAL;
5576 	}
5577 
5578 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5579 		PMD_DRV_LOG(ERR,
5580 			    "Invalid value passed to flow_xstat devarg.\n");
5581 		return -EINVAL;
5582 	}
5583 
5584 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5585 	if (BNXT_FLOW_XSTATS_EN(bp))
5586 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5587 
5588 	return 0;
5589 }
5590 
5591 static int
5592 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5593 					const char *value, void *opaque_arg)
5594 {
5595 	struct bnxt *bp = opaque_arg;
5596 	unsigned long max_num_kflows;
5597 	char *end = NULL;
5598 
5599 	if (!value || !opaque_arg) {
5600 		PMD_DRV_LOG(ERR,
5601 			"Invalid parameter passed to max_num_kflows devarg.\n");
5602 		return -EINVAL;
5603 	}
5604 
5605 	max_num_kflows = strtoul(value, &end, 10);
5606 	if (end == NULL || *end != '\0' ||
5607 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5608 		PMD_DRV_LOG(ERR,
5609 			"Invalid parameter passed to max_num_kflows devarg.\n");
5610 		return -EINVAL;
5611 	}
5612 
5613 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5614 		PMD_DRV_LOG(ERR,
5615 			"Invalid value passed to max_num_kflows devarg.\n");
5616 		return -EINVAL;
5617 	}
5618 
5619 	bp->max_num_kflows = max_num_kflows;
5620 	if (bp->max_num_kflows)
5621 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5622 				max_num_kflows);
5623 
5624 	return 0;
5625 }
5626 
5627 static int
5628 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5629 				 const char *value, void *opaque_arg)
5630 {
5631 	struct bnxt *bp = opaque_arg;
5632 	unsigned long app_id;
5633 	char *end = NULL;
5634 
5635 	if (!value || !opaque_arg) {
5636 		PMD_DRV_LOG(ERR,
5637 			    "Invalid parameter passed to app-id "
5638 			    "devargs.\n");
5639 		return -EINVAL;
5640 	}
5641 
5642 	app_id = strtoul(value, &end, 10);
5643 	if (end == NULL || *end != '\0' ||
5644 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5645 		PMD_DRV_LOG(ERR,
5646 			    "Invalid parameter passed to app_id "
5647 			    "devargs.\n");
5648 		return -EINVAL;
5649 	}
5650 
5651 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5652 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5653 			    (uint16_t)app_id);
5654 		return -EINVAL;
5655 	}
5656 
5657 	bp->app_id = app_id;
5658 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5659 
5660 	return 0;
5661 }
5662 
5663 static int
5664 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5665 			    const char *value, void *opaque_arg)
5666 {
5667 	struct bnxt_representor *vfr_bp = opaque_arg;
5668 	unsigned long rep_is_pf;
5669 	char *end = NULL;
5670 
5671 	if (!value || !opaque_arg) {
5672 		PMD_DRV_LOG(ERR,
5673 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5674 		return -EINVAL;
5675 	}
5676 
5677 	rep_is_pf = strtoul(value, &end, 10);
5678 	if (end == NULL || *end != '\0' ||
5679 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5680 		PMD_DRV_LOG(ERR,
5681 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5682 		return -EINVAL;
5683 	}
5684 
5685 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5686 		PMD_DRV_LOG(ERR,
5687 			    "Invalid value passed to rep_is_pf devargs.\n");
5688 		return -EINVAL;
5689 	}
5690 
5691 	vfr_bp->flags |= rep_is_pf;
5692 	if (BNXT_REP_PF(vfr_bp))
5693 		PMD_DRV_LOG(INFO, "PF representor\n");
5694 	else
5695 		PMD_DRV_LOG(INFO, "VF representor\n");
5696 
5697 	return 0;
5698 }
5699 
5700 static int
5701 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5702 			       const char *value, void *opaque_arg)
5703 {
5704 	struct bnxt_representor *vfr_bp = opaque_arg;
5705 	unsigned long rep_based_pf;
5706 	char *end = NULL;
5707 
5708 	if (!value || !opaque_arg) {
5709 		PMD_DRV_LOG(ERR,
5710 			    "Invalid parameter passed to rep_based_pf "
5711 			    "devargs.\n");
5712 		return -EINVAL;
5713 	}
5714 
5715 	rep_based_pf = strtoul(value, &end, 10);
5716 	if (end == NULL || *end != '\0' ||
5717 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5718 		PMD_DRV_LOG(ERR,
5719 			    "Invalid parameter passed to rep_based_pf "
5720 			    "devargs.\n");
5721 		return -EINVAL;
5722 	}
5723 
5724 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5725 		PMD_DRV_LOG(ERR,
5726 			    "Invalid value passed to rep_based_pf devargs.\n");
5727 		return -EINVAL;
5728 	}
5729 
5730 	vfr_bp->rep_based_pf = rep_based_pf;
5731 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5732 
5733 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5734 
5735 	return 0;
5736 }
5737 
5738 static int
5739 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5740 			    const char *value, void *opaque_arg)
5741 {
5742 	struct bnxt_representor *vfr_bp = opaque_arg;
5743 	unsigned long rep_q_r2f;
5744 	char *end = NULL;
5745 
5746 	if (!value || !opaque_arg) {
5747 		PMD_DRV_LOG(ERR,
5748 			    "Invalid parameter passed to rep_q_r2f "
5749 			    "devargs.\n");
5750 		return -EINVAL;
5751 	}
5752 
5753 	rep_q_r2f = strtoul(value, &end, 10);
5754 	if (end == NULL || *end != '\0' ||
5755 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5756 		PMD_DRV_LOG(ERR,
5757 			    "Invalid parameter passed to rep_q_r2f "
5758 			    "devargs.\n");
5759 		return -EINVAL;
5760 	}
5761 
5762 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5763 		PMD_DRV_LOG(ERR,
5764 			    "Invalid value passed to rep_q_r2f devargs.\n");
5765 		return -EINVAL;
5766 	}
5767 
5768 	vfr_bp->rep_q_r2f = rep_q_r2f;
5769 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5770 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5771 
5772 	return 0;
5773 }
5774 
5775 static int
5776 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5777 			    const char *value, void *opaque_arg)
5778 {
5779 	struct bnxt_representor *vfr_bp = opaque_arg;
5780 	unsigned long rep_q_f2r;
5781 	char *end = NULL;
5782 
5783 	if (!value || !opaque_arg) {
5784 		PMD_DRV_LOG(ERR,
5785 			    "Invalid parameter passed to rep_q_f2r "
5786 			    "devargs.\n");
5787 		return -EINVAL;
5788 	}
5789 
5790 	rep_q_f2r = strtoul(value, &end, 10);
5791 	if (end == NULL || *end != '\0' ||
5792 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5793 		PMD_DRV_LOG(ERR,
5794 			    "Invalid parameter passed to rep_q_f2r "
5795 			    "devargs.\n");
5796 		return -EINVAL;
5797 	}
5798 
5799 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5800 		PMD_DRV_LOG(ERR,
5801 			    "Invalid value passed to rep_q_f2r devargs.\n");
5802 		return -EINVAL;
5803 	}
5804 
5805 	vfr_bp->rep_q_f2r = rep_q_f2r;
5806 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5807 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5808 
5809 	return 0;
5810 }
5811 
5812 static int
5813 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5814 			     const char *value, void *opaque_arg)
5815 {
5816 	struct bnxt_representor *vfr_bp = opaque_arg;
5817 	unsigned long rep_fc_r2f;
5818 	char *end = NULL;
5819 
5820 	if (!value || !opaque_arg) {
5821 		PMD_DRV_LOG(ERR,
5822 			    "Invalid parameter passed to rep_fc_r2f "
5823 			    "devargs.\n");
5824 		return -EINVAL;
5825 	}
5826 
5827 	rep_fc_r2f = strtoul(value, &end, 10);
5828 	if (end == NULL || *end != '\0' ||
5829 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5830 		PMD_DRV_LOG(ERR,
5831 			    "Invalid parameter passed to rep_fc_r2f "
5832 			    "devargs.\n");
5833 		return -EINVAL;
5834 	}
5835 
5836 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5837 		PMD_DRV_LOG(ERR,
5838 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5839 		return -EINVAL;
5840 	}
5841 
5842 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5843 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5844 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5845 
5846 	return 0;
5847 }
5848 
5849 static int
5850 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5851 			     const char *value, void *opaque_arg)
5852 {
5853 	struct bnxt_representor *vfr_bp = opaque_arg;
5854 	unsigned long rep_fc_f2r;
5855 	char *end = NULL;
5856 
5857 	if (!value || !opaque_arg) {
5858 		PMD_DRV_LOG(ERR,
5859 			    "Invalid parameter passed to rep_fc_f2r "
5860 			    "devargs.\n");
5861 		return -EINVAL;
5862 	}
5863 
5864 	rep_fc_f2r = strtoul(value, &end, 10);
5865 	if (end == NULL || *end != '\0' ||
5866 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5867 		PMD_DRV_LOG(ERR,
5868 			    "Invalid parameter passed to rep_fc_f2r "
5869 			    "devargs.\n");
5870 		return -EINVAL;
5871 	}
5872 
5873 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5874 		PMD_DRV_LOG(ERR,
5875 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5876 		return -EINVAL;
5877 	}
5878 
5879 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5880 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5881 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5882 
5883 	return 0;
5884 }
5885 
5886 static int
5887 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5888 {
5889 	struct rte_kvargs *kvlist;
5890 	int ret;
5891 
5892 	if (devargs == NULL)
5893 		return 0;
5894 
5895 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5896 	if (kvlist == NULL)
5897 		return -EINVAL;
5898 
5899 	/*
5900 	 * Handler for "flow_xstat" devarg.
5901 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5902 	 */
5903 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5904 				 bnxt_parse_devarg_flow_xstat, bp);
5905 	if (ret)
5906 		goto err;
5907 
5908 	/*
5909 	 * Handler for "accum-stats" devarg.
5910 	 * Invoked as for ex: "-a 0000:00:0d.0,accum-stats=1"
5911 	 */
5912 	rte_kvargs_process(kvlist, BNXT_DEVARG_ACCUM_STATS,
5913 			   bnxt_parse_devarg_accum_stats, bp);
5914 	/*
5915 	 * Handler for "max_num_kflows" devarg.
5916 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5917 	 */
5918 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5919 				 bnxt_parse_devarg_max_num_kflows, bp);
5920 	if (ret)
5921 		goto err;
5922 
5923 err:
5924 	/*
5925 	 * Handler for "app-id" devarg.
5926 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5927 	 */
5928 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5929 			   bnxt_parse_devarg_app_id, bp);
5930 
5931 	rte_kvargs_free(kvlist);
5932 	return ret;
5933 }
5934 
5935 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5936 {
5937 	int rc = 0;
5938 
5939 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5940 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5941 		if (rc)
5942 			PMD_DRV_LOG(ERR,
5943 				    "Failed to alloc switch domain: %d\n", rc);
5944 		else
5945 			PMD_DRV_LOG(INFO,
5946 				    "Switch domain allocated %d\n",
5947 				    bp->switch_domain_id);
5948 	}
5949 
5950 	return rc;
5951 }
5952 
5953 /* Allocate and initialize various fields in bnxt struct that
5954  * need to be allocated/destroyed only once in the lifetime of the driver
5955  */
5956 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5957 {
5958 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5959 	struct bnxt *bp = eth_dev->data->dev_private;
5960 	int rc = 0;
5961 
5962 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5963 
5964 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5965 		bp->flags |= BNXT_FLAG_VF;
5966 
5967 	if (bnxt_p5_device(pci_dev->id.device_id))
5968 		bp->flags |= BNXT_FLAG_CHIP_P5;
5969 
5970 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5971 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5972 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5973 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5974 		bp->flags |= BNXT_FLAG_STINGRAY;
5975 
5976 	if (BNXT_TRUFLOW_EN(bp)) {
5977 		/* extra mbuf field is required to store CFA code from mark */
5978 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5979 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5980 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5981 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5982 		};
5983 		bnxt_cfa_code_dynfield_offset =
5984 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5985 		if (bnxt_cfa_code_dynfield_offset < 0) {
5986 			PMD_DRV_LOG(ERR,
5987 			    "Failed to register mbuf field for TruFlow mark\n");
5988 			return -rte_errno;
5989 		}
5990 	}
5991 
5992 	rc = bnxt_map_pci_bars(eth_dev);
5993 	if (rc) {
5994 		PMD_DRV_LOG(ERR,
5995 			    "Failed to initialize board rc: %x\n", rc);
5996 		return rc;
5997 	}
5998 
5999 	rc = bnxt_alloc_pf_info(bp);
6000 	if (rc)
6001 		return rc;
6002 
6003 	rc = bnxt_alloc_link_info(bp);
6004 	if (rc)
6005 		return rc;
6006 
6007 	rc = bnxt_alloc_parent_info(bp);
6008 	if (rc)
6009 		return rc;
6010 
6011 	rc = bnxt_alloc_hwrm_resources(bp);
6012 	if (rc) {
6013 		PMD_DRV_LOG(ERR,
6014 			    "Failed to allocate response buffer rc: %x\n", rc);
6015 		return rc;
6016 	}
6017 	rc = bnxt_alloc_leds_info(bp);
6018 	if (rc)
6019 		return rc;
6020 
6021 	rc = bnxt_alloc_cos_queues(bp);
6022 	if (rc)
6023 		return rc;
6024 
6025 	rc = bnxt_init_locks(bp);
6026 	if (rc)
6027 		return rc;
6028 
6029 	rc = bnxt_alloc_switch_domain(bp);
6030 	if (rc)
6031 		return rc;
6032 
6033 	return rc;
6034 }
6035 
6036 static int
6037 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
6038 {
6039 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
6040 	static int version_printed;
6041 	struct bnxt *bp;
6042 	int rc;
6043 
6044 	if (version_printed++ == 0)
6045 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
6046 
6047 	eth_dev->dev_ops = &bnxt_dev_ops;
6048 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
6049 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
6050 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
6051 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
6052 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
6053 
6054 	/*
6055 	 * For secondary processes, we don't initialise any further
6056 	 * as primary has already done this work.
6057 	 */
6058 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6059 		return 0;
6060 
6061 	rte_eth_copy_pci_info(eth_dev, pci_dev);
6062 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
6063 
6064 	bp = eth_dev->data->dev_private;
6065 
6066 	/* Parse dev arguments passed on when starting the DPDK application. */
6067 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
6068 	if (rc)
6069 		goto error_free;
6070 
6071 	rc = bnxt_drv_init(eth_dev);
6072 	if (rc)
6073 		goto error_free;
6074 
6075 	rc = bnxt_init_resources(bp, false);
6076 	if (rc)
6077 		goto error_free;
6078 
6079 	rc = bnxt_alloc_stats_mem(bp);
6080 	if (rc)
6081 		goto error_free;
6082 
6083 	PMD_DRV_LOG(INFO,
6084 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
6085 		    DRV_MODULE_NAME,
6086 		    pci_dev->mem_resource[0].phys_addr,
6087 		    pci_dev->mem_resource[0].addr);
6088 
6089 	return 0;
6090 
6091 error_free:
6092 	bnxt_dev_uninit(eth_dev);
6093 	return rc;
6094 }
6095 
6096 
6097 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
6098 {
6099 	if (!ctx)
6100 		return;
6101 
6102 	if (ctx->va)
6103 		rte_free(ctx->va);
6104 
6105 	ctx->va = NULL;
6106 	ctx->dma = RTE_BAD_IOVA;
6107 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
6108 }
6109 
6110 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
6111 {
6112 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
6113 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6114 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
6115 				  bp->flow_stat->max_fc,
6116 				  false);
6117 
6118 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
6119 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6120 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
6121 				  bp->flow_stat->max_fc,
6122 				  false);
6123 
6124 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6125 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
6126 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6127 
6128 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6129 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
6130 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6131 
6132 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6133 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
6134 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6135 
6136 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6137 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
6138 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6139 }
6140 
6141 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
6142 {
6143 	bnxt_unregister_fc_ctx_mem(bp);
6144 
6145 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
6146 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
6147 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
6148 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
6149 }
6150 
6151 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
6152 {
6153 	if (BNXT_FLOW_XSTATS_EN(bp))
6154 		bnxt_uninit_fc_ctx_mem(bp);
6155 }
6156 
6157 static void
6158 bnxt_free_error_recovery_info(struct bnxt *bp)
6159 {
6160 	rte_free(bp->recovery_info);
6161 	bp->recovery_info = NULL;
6162 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
6163 }
6164 
6165 static int
6166 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
6167 {
6168 	int rc;
6169 
6170 	bnxt_free_int(bp);
6171 	bnxt_free_mem(bp, reconfig_dev);
6172 
6173 	bnxt_hwrm_func_buf_unrgtr(bp);
6174 	if (bp->pf != NULL) {
6175 		rte_free(bp->pf->vf_req_buf);
6176 		bp->pf->vf_req_buf = NULL;
6177 	}
6178 
6179 	rc = bnxt_hwrm_func_driver_unregister(bp);
6180 	bp->flags &= ~BNXT_FLAG_REGISTERED;
6181 	bnxt_free_ctx_mem(bp);
6182 	if (!reconfig_dev) {
6183 		bnxt_free_hwrm_resources(bp);
6184 		bnxt_free_error_recovery_info(bp);
6185 	}
6186 
6187 	bnxt_uninit_ctx_mem(bp);
6188 
6189 	bnxt_free_flow_stats_info(bp);
6190 	if (bp->rep_info != NULL)
6191 		bnxt_free_switch_domain(bp);
6192 	bnxt_free_rep_info(bp);
6193 	rte_free(bp->ptp_cfg);
6194 	bp->ptp_cfg = NULL;
6195 	return rc;
6196 }
6197 
6198 static int
6199 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
6200 {
6201 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6202 		return -EPERM;
6203 
6204 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
6205 
6206 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
6207 		bnxt_dev_close_op(eth_dev);
6208 
6209 	return 0;
6210 }
6211 
6212 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6213 {
6214 	struct bnxt *bp = eth_dev->data->dev_private;
6215 	struct rte_eth_dev *vf_rep_eth_dev;
6216 	int ret = 0, i;
6217 
6218 	if (!bp)
6219 		return -EINVAL;
6220 
6221 	for (i = 0; i < bp->num_reps; i++) {
6222 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6223 		if (!vf_rep_eth_dev)
6224 			continue;
6225 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6226 			    vf_rep_eth_dev->data->port_id);
6227 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6228 	}
6229 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6230 		    eth_dev->data->port_id);
6231 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6232 
6233 	return ret;
6234 }
6235 
6236 static void bnxt_free_rep_info(struct bnxt *bp)
6237 {
6238 	rte_free(bp->rep_info);
6239 	bp->rep_info = NULL;
6240 	rte_free(bp->cfa_code_map);
6241 	bp->cfa_code_map = NULL;
6242 }
6243 
6244 static int bnxt_init_rep_info(struct bnxt *bp)
6245 {
6246 	int i = 0, rc;
6247 
6248 	if (bp->rep_info)
6249 		return 0;
6250 
6251 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
6252 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
6253 				   0);
6254 	if (!bp->rep_info) {
6255 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6256 		return -ENOMEM;
6257 	}
6258 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6259 				       sizeof(*bp->cfa_code_map) *
6260 				       BNXT_MAX_CFA_CODE, 0);
6261 	if (!bp->cfa_code_map) {
6262 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6263 		bnxt_free_rep_info(bp);
6264 		return -ENOMEM;
6265 	}
6266 
6267 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6268 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6269 
6270 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6271 	if (rc) {
6272 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6273 		bnxt_free_rep_info(bp);
6274 		return rc;
6275 	}
6276 
6277 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6278 	if (rc) {
6279 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6280 		bnxt_free_rep_info(bp);
6281 		return rc;
6282 	}
6283 
6284 	return rc;
6285 }
6286 
6287 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6288 			       struct rte_eth_devargs *eth_da,
6289 			       struct rte_eth_dev *backing_eth_dev,
6290 			       const char *dev_args)
6291 {
6292 	struct rte_eth_dev *vf_rep_eth_dev;
6293 	char name[RTE_ETH_NAME_MAX_LEN];
6294 	struct bnxt *backing_bp;
6295 	uint16_t num_rep;
6296 	int i, ret = 0;
6297 	struct rte_kvargs *kvlist = NULL;
6298 
6299 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6300 		return 0;
6301 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6302 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6303 			    eth_da->type);
6304 		return -ENOTSUP;
6305 	}
6306 	num_rep = eth_da->nb_representor_ports;
6307 	if (num_rep > BNXT_MAX_VF_REPS) {
6308 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6309 			    num_rep, BNXT_MAX_VF_REPS);
6310 		return -EINVAL;
6311 	}
6312 
6313 	if (num_rep >= RTE_MAX_ETHPORTS) {
6314 		PMD_DRV_LOG(ERR,
6315 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6316 			    num_rep, RTE_MAX_ETHPORTS);
6317 		return -EINVAL;
6318 	}
6319 
6320 	backing_bp = backing_eth_dev->data->dev_private;
6321 
6322 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6323 		PMD_DRV_LOG(ERR,
6324 			    "Not a PF or trusted VF. No Representor support\n");
6325 		/* Returning an error is not an option.
6326 		 * Applications are not handling this correctly
6327 		 */
6328 		return 0;
6329 	}
6330 
6331 	if (bnxt_init_rep_info(backing_bp))
6332 		return 0;
6333 
6334 	for (i = 0; i < num_rep; i++) {
6335 		struct bnxt_representor representor = {
6336 			.vf_id = eth_da->representor_ports[i],
6337 			.switch_domain_id = backing_bp->switch_domain_id,
6338 			.parent_dev = backing_eth_dev
6339 		};
6340 
6341 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
6342 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6343 				    representor.vf_id, BNXT_MAX_VF_REPS);
6344 			continue;
6345 		}
6346 
6347 		/* representor port net_bdf_port */
6348 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6349 			 pci_dev->device.name, eth_da->representor_ports[i]);
6350 
6351 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6352 		if (kvlist) {
6353 			/*
6354 			 * Handler for "rep_is_pf" devarg.
6355 			 * Invoked as for ex: "-a 000:00:0d.0,
6356 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6357 			 */
6358 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6359 						 bnxt_parse_devarg_rep_is_pf,
6360 						 (void *)&representor);
6361 			if (ret) {
6362 				ret = -EINVAL;
6363 				goto err;
6364 			}
6365 			/*
6366 			 * Handler for "rep_based_pf" devarg.
6367 			 * Invoked as for ex: "-a 000:00:0d.0,
6368 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6369 			 */
6370 			ret = rte_kvargs_process(kvlist,
6371 						 BNXT_DEVARG_REP_BASED_PF,
6372 						 bnxt_parse_devarg_rep_based_pf,
6373 						 (void *)&representor);
6374 			if (ret) {
6375 				ret = -EINVAL;
6376 				goto err;
6377 			}
6378 			/*
6379 			 * Handler for "rep_based_pf" devarg.
6380 			 * Invoked as for ex: "-a 000:00:0d.0,
6381 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6382 			 */
6383 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6384 						 bnxt_parse_devarg_rep_q_r2f,
6385 						 (void *)&representor);
6386 			if (ret) {
6387 				ret = -EINVAL;
6388 				goto err;
6389 			}
6390 			/*
6391 			 * Handler for "rep_based_pf" devarg.
6392 			 * Invoked as for ex: "-a 000:00:0d.0,
6393 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6394 			 */
6395 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6396 						 bnxt_parse_devarg_rep_q_f2r,
6397 						 (void *)&representor);
6398 			if (ret) {
6399 				ret = -EINVAL;
6400 				goto err;
6401 			}
6402 			/*
6403 			 * Handler for "rep_based_pf" devarg.
6404 			 * Invoked as for ex: "-a 000:00:0d.0,
6405 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6406 			 */
6407 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6408 						 bnxt_parse_devarg_rep_fc_r2f,
6409 						 (void *)&representor);
6410 			if (ret) {
6411 				ret = -EINVAL;
6412 				goto err;
6413 			}
6414 			/*
6415 			 * Handler for "rep_based_pf" devarg.
6416 			 * Invoked as for ex: "-a 000:00:0d.0,
6417 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6418 			 */
6419 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6420 						 bnxt_parse_devarg_rep_fc_f2r,
6421 						 (void *)&representor);
6422 			if (ret) {
6423 				ret = -EINVAL;
6424 				goto err;
6425 			}
6426 		}
6427 
6428 		ret = rte_eth_dev_create(&pci_dev->device, name,
6429 					 sizeof(struct bnxt_representor),
6430 					 NULL, NULL,
6431 					 bnxt_representor_init,
6432 					 &representor);
6433 		if (ret) {
6434 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6435 				    "representor %s.", name);
6436 			goto err;
6437 		}
6438 
6439 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6440 		if (!vf_rep_eth_dev) {
6441 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6442 				    " for VF-Rep: %s.", name);
6443 			ret = -ENODEV;
6444 			goto err;
6445 		}
6446 
6447 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6448 			    backing_eth_dev->data->port_id);
6449 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6450 							 vf_rep_eth_dev;
6451 		backing_bp->num_reps++;
6452 
6453 	}
6454 
6455 	rte_kvargs_free(kvlist);
6456 	return 0;
6457 
6458 err:
6459 	/* If num_rep > 1, then rollback already created
6460 	 * ports, since we'll be failing the probe anyway
6461 	 */
6462 	if (num_rep > 1)
6463 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6464 	rte_errno = -ret;
6465 	rte_kvargs_free(kvlist);
6466 
6467 	return ret;
6468 }
6469 
6470 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6471 			  struct rte_pci_device *pci_dev)
6472 {
6473 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6474 	struct rte_eth_dev *backing_eth_dev;
6475 	uint16_t num_rep;
6476 	int ret = 0;
6477 
6478 	if (pci_dev->device.devargs) {
6479 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6480 					    &eth_da);
6481 		if (ret)
6482 			return ret;
6483 	}
6484 
6485 	num_rep = eth_da.nb_representor_ports;
6486 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6487 		    num_rep);
6488 
6489 	/* We could come here after first level of probe is already invoked
6490 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6491 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6492 	 */
6493 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6494 	if (backing_eth_dev == NULL) {
6495 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6496 					 sizeof(struct bnxt),
6497 					 eth_dev_pci_specific_init, pci_dev,
6498 					 bnxt_dev_init, NULL);
6499 
6500 		if (ret || !num_rep)
6501 			return ret;
6502 
6503 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6504 	}
6505 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6506 		    backing_eth_dev->data->port_id);
6507 
6508 	if (!num_rep)
6509 		return ret;
6510 
6511 	/* probe representor ports now */
6512 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6513 				  pci_dev->device.devargs->args);
6514 
6515 	return ret;
6516 }
6517 
6518 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6519 {
6520 	struct rte_eth_dev *eth_dev;
6521 
6522 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6523 	if (!eth_dev)
6524 		return 0; /* Invoked typically only by OVS-DPDK, by the
6525 			   * time it comes here the eth_dev is already
6526 			   * deleted by rte_eth_dev_close(), so returning
6527 			   * +ve value will at least help in proper cleanup
6528 			   */
6529 
6530 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6531 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6532 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6533 			return rte_eth_dev_destroy(eth_dev,
6534 						   bnxt_representor_uninit);
6535 		else
6536 			return rte_eth_dev_destroy(eth_dev,
6537 						   bnxt_dev_uninit);
6538 	} else {
6539 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6540 	}
6541 }
6542 
6543 static struct rte_pci_driver bnxt_rte_pmd = {
6544 	.id_table = bnxt_pci_id_map,
6545 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6546 			RTE_PCI_DRV_INTR_RMV |
6547 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6548 						  * and OVS-DPDK
6549 						  */
6550 	.probe = bnxt_pci_probe,
6551 	.remove = bnxt_pci_remove,
6552 };
6553 
6554 static bool
6555 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6556 {
6557 	if (strcmp(dev->device->driver->name, drv->driver.name))
6558 		return false;
6559 
6560 	return true;
6561 }
6562 
6563 bool is_bnxt_supported(struct rte_eth_dev *dev)
6564 {
6565 	return is_device_supported(dev, &bnxt_rte_pmd);
6566 }
6567 
6568 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6569 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6570 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6571 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6572