xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision 04ed01f1)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2014-2018 Broadcom
3  * All rights reserved.
4  */
5 
6 #include <inttypes.h>
7 #include <stdbool.h>
8 
9 #include <rte_dev.h>
10 #include <rte_ethdev_driver.h>
11 #include <rte_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_TRUFLOW	"host-based-truflow"
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 
101 static const char *const bnxt_dev_args[] = {
102 	BNXT_DEVARG_REPRESENTOR,
103 	BNXT_DEVARG_TRUFLOW,
104 	BNXT_DEVARG_FLOW_XSTAT,
105 	BNXT_DEVARG_MAX_NUM_KFLOWS,
106 	BNXT_DEVARG_REP_BASED_PF,
107 	BNXT_DEVARG_REP_IS_PF,
108 	BNXT_DEVARG_REP_Q_R2F,
109 	BNXT_DEVARG_REP_Q_F2R,
110 	BNXT_DEVARG_REP_FC_R2F,
111 	BNXT_DEVARG_REP_FC_F2R,
112 	NULL
113 };
114 
115 /*
116  * truflow == false to disable the feature
117  * truflow == true to enable the feature
118  */
119 #define	BNXT_DEVARG_TRUFLOW_INVALID(truflow)	((truflow) > 1)
120 
121 /*
122  * flow_xstat == false to disable the feature
123  * flow_xstat == true to enable the feature
124  */
125 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
126 
127 /*
128  * rep_is_pf == false to indicate VF representor
129  * rep_is_pf == true to indicate PF representor
130  */
131 #define	BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)	((rep_is_pf) > 1)
132 
133 /*
134  * rep_based_pf == Physical index of the PF
135  */
136 #define	BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)	((rep_based_pf) > 15)
137 /*
138  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
139  */
140 #define	BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)	((rep_q_r2f) > 3)
141 
142 /*
143  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
144  */
145 #define	BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)	((rep_q_f2r) > 3)
146 
147 /*
148  * rep_fc_r2f == Flow control for the representor to endpoint direction
149  */
150 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)	((rep_fc_r2f) > 1)
151 
152 /*
153  * rep_fc_f2r == Flow control for the endpoint to representor direction
154  */
155 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)	((rep_fc_f2r) > 1)
156 
157 int bnxt_cfa_code_dynfield_offset = -1;
158 
159 /*
160  * max_num_kflows must be >= 32
161  * and must be a power-of-2 supported value
162  * return: 1 -> invalid
163  *         0 -> valid
164  */
165 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
166 {
167 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
168 		return 1;
169 	return 0;
170 }
171 
172 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
173 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
174 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
175 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
176 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
177 static int bnxt_restore_vlan_filters(struct bnxt *bp);
178 static void bnxt_dev_recover(void *arg);
179 static void bnxt_free_error_recovery_info(struct bnxt *bp);
180 static void bnxt_free_rep_info(struct bnxt *bp);
181 
182 int is_bnxt_in_error(struct bnxt *bp)
183 {
184 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
185 		return -EIO;
186 	if (bp->flags & BNXT_FLAG_FW_RESET)
187 		return -EBUSY;
188 
189 	return 0;
190 }
191 
192 /***********************/
193 
194 /*
195  * High level utility functions
196  */
197 
198 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
199 {
200 	unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
201 					     BNXT_RSS_TBL_SIZE_P5);
202 
203 	if (!BNXT_CHIP_P5(bp))
204 		return 1;
205 
206 	return RTE_ALIGN_MUL_CEIL(num_rss_rings,
207 				  BNXT_RSS_ENTRIES_PER_CTX_P5) /
208 				  BNXT_RSS_ENTRIES_PER_CTX_P5;
209 }
210 
211 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
212 {
213 	if (!BNXT_CHIP_P5(bp))
214 		return HW_HASH_INDEX_SIZE;
215 
216 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
217 }
218 
219 static void bnxt_free_parent_info(struct bnxt *bp)
220 {
221 	rte_free(bp->parent);
222 }
223 
224 static void bnxt_free_pf_info(struct bnxt *bp)
225 {
226 	rte_free(bp->pf);
227 }
228 
229 static void bnxt_free_link_info(struct bnxt *bp)
230 {
231 	rte_free(bp->link_info);
232 }
233 
234 static void bnxt_free_leds_info(struct bnxt *bp)
235 {
236 	if (BNXT_VF(bp))
237 		return;
238 
239 	rte_free(bp->leds);
240 	bp->leds = NULL;
241 }
242 
243 static void bnxt_free_flow_stats_info(struct bnxt *bp)
244 {
245 	rte_free(bp->flow_stat);
246 	bp->flow_stat = NULL;
247 }
248 
249 static void bnxt_free_cos_queues(struct bnxt *bp)
250 {
251 	rte_free(bp->rx_cos_queue);
252 	rte_free(bp->tx_cos_queue);
253 }
254 
255 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
256 {
257 	bnxt_free_filter_mem(bp);
258 	bnxt_free_vnic_attributes(bp);
259 	bnxt_free_vnic_mem(bp);
260 
261 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
262 	 * If the number of rings change across fw update,
263 	 * we don't have much choice except to warn the user.
264 	 */
265 	if (!reconfig) {
266 		bnxt_free_stats(bp);
267 		bnxt_free_tx_rings(bp);
268 		bnxt_free_rx_rings(bp);
269 	}
270 	bnxt_free_async_cp_ring(bp);
271 	bnxt_free_rxtx_nq_ring(bp);
272 
273 	rte_free(bp->grp_info);
274 	bp->grp_info = NULL;
275 }
276 
277 static int bnxt_alloc_parent_info(struct bnxt *bp)
278 {
279 	bp->parent = rte_zmalloc("bnxt_parent_info",
280 				 sizeof(struct bnxt_parent_info), 0);
281 	if (bp->parent == NULL)
282 		return -ENOMEM;
283 
284 	return 0;
285 }
286 
287 static int bnxt_alloc_pf_info(struct bnxt *bp)
288 {
289 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
290 	if (bp->pf == NULL)
291 		return -ENOMEM;
292 
293 	return 0;
294 }
295 
296 static int bnxt_alloc_link_info(struct bnxt *bp)
297 {
298 	bp->link_info =
299 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
300 	if (bp->link_info == NULL)
301 		return -ENOMEM;
302 
303 	return 0;
304 }
305 
306 static int bnxt_alloc_leds_info(struct bnxt *bp)
307 {
308 	if (BNXT_VF(bp))
309 		return 0;
310 
311 	bp->leds = rte_zmalloc("bnxt_leds",
312 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
313 			       0);
314 	if (bp->leds == NULL)
315 		return -ENOMEM;
316 
317 	return 0;
318 }
319 
320 static int bnxt_alloc_cos_queues(struct bnxt *bp)
321 {
322 	bp->rx_cos_queue =
323 		rte_zmalloc("bnxt_rx_cosq",
324 			    BNXT_COS_QUEUE_COUNT *
325 			    sizeof(struct bnxt_cos_queue_info),
326 			    0);
327 	if (bp->rx_cos_queue == NULL)
328 		return -ENOMEM;
329 
330 	bp->tx_cos_queue =
331 		rte_zmalloc("bnxt_tx_cosq",
332 			    BNXT_COS_QUEUE_COUNT *
333 			    sizeof(struct bnxt_cos_queue_info),
334 			    0);
335 	if (bp->tx_cos_queue == NULL)
336 		return -ENOMEM;
337 
338 	return 0;
339 }
340 
341 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
342 {
343 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
344 				    sizeof(struct bnxt_flow_stat_info), 0);
345 	if (bp->flow_stat == NULL)
346 		return -ENOMEM;
347 
348 	return 0;
349 }
350 
351 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
352 {
353 	int rc;
354 
355 	rc = bnxt_alloc_ring_grps(bp);
356 	if (rc)
357 		goto alloc_mem_err;
358 
359 	rc = bnxt_alloc_async_ring_struct(bp);
360 	if (rc)
361 		goto alloc_mem_err;
362 
363 	rc = bnxt_alloc_vnic_mem(bp);
364 	if (rc)
365 		goto alloc_mem_err;
366 
367 	rc = bnxt_alloc_vnic_attributes(bp);
368 	if (rc)
369 		goto alloc_mem_err;
370 
371 	rc = bnxt_alloc_filter_mem(bp);
372 	if (rc)
373 		goto alloc_mem_err;
374 
375 	rc = bnxt_alloc_async_cp_ring(bp);
376 	if (rc)
377 		goto alloc_mem_err;
378 
379 	rc = bnxt_alloc_rxtx_nq_ring(bp);
380 	if (rc)
381 		goto alloc_mem_err;
382 
383 	if (BNXT_FLOW_XSTATS_EN(bp)) {
384 		rc = bnxt_alloc_flow_stats_info(bp);
385 		if (rc)
386 			goto alloc_mem_err;
387 	}
388 
389 	return 0;
390 
391 alloc_mem_err:
392 	bnxt_free_mem(bp, reconfig);
393 	return rc;
394 }
395 
396 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
397 {
398 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
399 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
400 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
401 	struct bnxt_rx_queue *rxq;
402 	unsigned int j;
403 	int rc;
404 
405 	rc = bnxt_vnic_grp_alloc(bp, vnic);
406 	if (rc)
407 		goto err_out;
408 
409 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
410 		    vnic_id, vnic, vnic->fw_grp_ids);
411 
412 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
413 	if (rc)
414 		goto err_out;
415 
416 	/* Alloc RSS context only if RSS mode is enabled */
417 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
418 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
419 
420 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
421 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
422 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
423 			PMD_DRV_LOG(ERR,
424 				    "Only queues 0-%d will be in RSS table\n",
425 				    BNXT_RSS_TBL_SIZE_P5 - 1);
426 		}
427 
428 		rc = 0;
429 		for (j = 0; j < nr_ctxs; j++) {
430 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
431 			if (rc)
432 				break;
433 		}
434 		if (rc) {
435 			PMD_DRV_LOG(ERR,
436 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
437 				    vnic_id, j, rc);
438 			goto err_out;
439 		}
440 		vnic->num_lb_ctxts = nr_ctxs;
441 	}
442 
443 	/*
444 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
445 	 * setting is not available at this time, it will not be
446 	 * configured correctly in the CFA.
447 	 */
448 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
449 		vnic->vlan_strip = true;
450 	else
451 		vnic->vlan_strip = false;
452 
453 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
454 	if (rc)
455 		goto err_out;
456 
457 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
458 	if (rc)
459 		goto err_out;
460 
461 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
462 		rxq = bp->eth_dev->data->rx_queues[j];
463 
464 		PMD_DRV_LOG(DEBUG,
465 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
466 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
467 
468 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
469 			rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
470 		else
471 			vnic->rx_queue_cnt++;
472 	}
473 
474 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
475 
476 	rc = bnxt_vnic_rss_configure(bp, vnic);
477 	if (rc)
478 		goto err_out;
479 
480 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
481 
482 	if (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO)
483 		bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 1);
484 	else
485 		bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 0);
486 
487 	return 0;
488 err_out:
489 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
490 		    vnic_id, rc);
491 	return rc;
492 }
493 
494 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
495 {
496 	int rc = 0;
497 
498 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
499 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
500 	if (rc)
501 		return rc;
502 
503 	PMD_DRV_LOG(DEBUG,
504 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
505 		    " rx_fc_in_tbl.ctx_id = %d\n",
506 		    bp->flow_stat->rx_fc_in_tbl.va,
507 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
508 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
509 
510 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
511 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
512 	if (rc)
513 		return rc;
514 
515 	PMD_DRV_LOG(DEBUG,
516 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
517 		    " rx_fc_out_tbl.ctx_id = %d\n",
518 		    bp->flow_stat->rx_fc_out_tbl.va,
519 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
520 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
521 
522 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
523 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
524 	if (rc)
525 		return rc;
526 
527 	PMD_DRV_LOG(DEBUG,
528 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
529 		    " tx_fc_in_tbl.ctx_id = %d\n",
530 		    bp->flow_stat->tx_fc_in_tbl.va,
531 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
532 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
533 
534 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
535 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
536 	if (rc)
537 		return rc;
538 
539 	PMD_DRV_LOG(DEBUG,
540 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
541 		    " tx_fc_out_tbl.ctx_id = %d\n",
542 		    bp->flow_stat->tx_fc_out_tbl.va,
543 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
544 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
545 
546 	memset(bp->flow_stat->rx_fc_out_tbl.va,
547 	       0,
548 	       bp->flow_stat->rx_fc_out_tbl.size);
549 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
550 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
551 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
552 				       bp->flow_stat->max_fc,
553 				       true);
554 	if (rc)
555 		return rc;
556 
557 	memset(bp->flow_stat->tx_fc_out_tbl.va,
558 	       0,
559 	       bp->flow_stat->tx_fc_out_tbl.size);
560 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
561 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
562 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
563 				       bp->flow_stat->max_fc,
564 				       true);
565 
566 	return rc;
567 }
568 
569 static int bnxt_alloc_ctx_mem_buf(char *type, size_t size,
570 				  struct bnxt_ctx_mem_buf_info *ctx)
571 {
572 	if (!ctx)
573 		return -EINVAL;
574 
575 	ctx->va = rte_zmalloc(type, size, 0);
576 	if (ctx->va == NULL)
577 		return -ENOMEM;
578 	rte_mem_lock_page(ctx->va);
579 	ctx->size = size;
580 	ctx->dma = rte_mem_virt2iova(ctx->va);
581 	if (ctx->dma == RTE_BAD_IOVA)
582 		return -ENOMEM;
583 
584 	return 0;
585 }
586 
587 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
588 {
589 	struct rte_pci_device *pdev = bp->pdev;
590 	char type[RTE_MEMZONE_NAMESIZE];
591 	uint16_t max_fc;
592 	int rc = 0;
593 
594 	max_fc = bp->flow_stat->max_fc;
595 
596 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
597 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
598 	/* 4 bytes for each counter-id */
599 	rc = bnxt_alloc_ctx_mem_buf(type,
600 				    max_fc * 4,
601 				    &bp->flow_stat->rx_fc_in_tbl);
602 	if (rc)
603 		return rc;
604 
605 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
606 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
607 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
608 	rc = bnxt_alloc_ctx_mem_buf(type,
609 				    max_fc * 16,
610 				    &bp->flow_stat->rx_fc_out_tbl);
611 	if (rc)
612 		return rc;
613 
614 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
615 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
616 	/* 4 bytes for each counter-id */
617 	rc = bnxt_alloc_ctx_mem_buf(type,
618 				    max_fc * 4,
619 				    &bp->flow_stat->tx_fc_in_tbl);
620 	if (rc)
621 		return rc;
622 
623 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
624 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
625 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
626 	rc = bnxt_alloc_ctx_mem_buf(type,
627 				    max_fc * 16,
628 				    &bp->flow_stat->tx_fc_out_tbl);
629 	if (rc)
630 		return rc;
631 
632 	rc = bnxt_register_fc_ctx_mem(bp);
633 
634 	return rc;
635 }
636 
637 static int bnxt_init_ctx_mem(struct bnxt *bp)
638 {
639 	int rc = 0;
640 
641 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
642 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
643 	    !BNXT_FLOW_XSTATS_EN(bp))
644 		return 0;
645 
646 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
647 	if (rc)
648 		return rc;
649 
650 	rc = bnxt_init_fc_ctx_mem(bp);
651 
652 	return rc;
653 }
654 
655 static int bnxt_update_phy_setting(struct bnxt *bp)
656 {
657 	struct rte_eth_link new;
658 	int rc;
659 
660 	rc = bnxt_get_hwrm_link_config(bp, &new);
661 	if (rc) {
662 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
663 		return rc;
664 	}
665 
666 	/*
667 	 * On BCM957508-N2100 adapters, FW will not allow any user other
668 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
669 	 * always returns link up. Force phy update always in that case.
670 	 */
671 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
672 		rc = bnxt_set_hwrm_link_config(bp, true);
673 		if (rc) {
674 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
675 			return rc;
676 		}
677 	}
678 
679 	return rc;
680 }
681 
682 static int bnxt_init_chip(struct bnxt *bp)
683 {
684 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
685 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
686 	uint32_t intr_vector = 0;
687 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
688 	uint32_t vec = BNXT_MISC_VEC_ID;
689 	unsigned int i, j;
690 	int rc;
691 
692 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
693 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
694 			DEV_RX_OFFLOAD_JUMBO_FRAME;
695 		bp->flags |= BNXT_FLAG_JUMBO;
696 	} else {
697 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
698 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
699 		bp->flags &= ~BNXT_FLAG_JUMBO;
700 	}
701 
702 	/* THOR does not support ring groups.
703 	 * But we will use the array to save RSS context IDs.
704 	 */
705 	if (BNXT_CHIP_P5(bp))
706 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
707 
708 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
709 	if (rc) {
710 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
711 		goto err_out;
712 	}
713 
714 	rc = bnxt_alloc_hwrm_rings(bp);
715 	if (rc) {
716 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
717 		goto err_out;
718 	}
719 
720 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
721 	if (rc) {
722 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
723 		goto err_out;
724 	}
725 
726 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
727 		goto skip_cosq_cfg;
728 
729 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
730 		if (bp->rx_cos_queue[i].id != 0xff) {
731 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
732 
733 			if (!vnic) {
734 				PMD_DRV_LOG(ERR,
735 					    "Num pools more than FW profile\n");
736 				rc = -EINVAL;
737 				goto err_out;
738 			}
739 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
740 			bp->rx_cosq_cnt++;
741 		}
742 	}
743 
744 skip_cosq_cfg:
745 	rc = bnxt_mq_rx_configure(bp);
746 	if (rc) {
747 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
748 		goto err_out;
749 	}
750 
751 	/* default vnic 0 */
752 	rc = bnxt_setup_one_vnic(bp, 0);
753 	if (rc)
754 		goto err_out;
755 	/* VNIC configuration */
756 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
757 		for (i = 1; i < bp->nr_vnics; i++) {
758 			rc = bnxt_setup_one_vnic(bp, i);
759 			if (rc)
760 				goto err_out;
761 		}
762 	}
763 
764 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
765 	if (rc) {
766 		PMD_DRV_LOG(ERR,
767 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
768 		goto err_out;
769 	}
770 
771 	/* check and configure queue intr-vector mapping */
772 	if ((rte_intr_cap_multiple(intr_handle) ||
773 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
774 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
775 		intr_vector = bp->eth_dev->data->nb_rx_queues;
776 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
777 		if (intr_vector > bp->rx_cp_nr_rings) {
778 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
779 					bp->rx_cp_nr_rings);
780 			return -ENOTSUP;
781 		}
782 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
783 		if (rc)
784 			return rc;
785 	}
786 
787 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
788 		intr_handle->intr_vec =
789 			rte_zmalloc("intr_vec",
790 				    bp->eth_dev->data->nb_rx_queues *
791 				    sizeof(int), 0);
792 		if (intr_handle->intr_vec == NULL) {
793 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
794 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
795 			rc = -ENOMEM;
796 			goto err_disable;
797 		}
798 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
799 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
800 			 intr_handle->intr_vec, intr_handle->nb_efd,
801 			intr_handle->max_intr);
802 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
803 		     queue_id++) {
804 			intr_handle->intr_vec[queue_id] =
805 							vec + BNXT_RX_VEC_START;
806 			if (vec < base + intr_handle->nb_efd - 1)
807 				vec++;
808 		}
809 	}
810 
811 	/* enable uio/vfio intr/eventfd mapping */
812 	rc = rte_intr_enable(intr_handle);
813 #ifndef RTE_EXEC_ENV_FREEBSD
814 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
815 	if (rc)
816 		goto err_free;
817 #endif
818 
819 	rc = bnxt_update_phy_setting(bp);
820 	if (rc)
821 		goto err_free;
822 
823 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
824 	if (!bp->mark_table)
825 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
826 
827 	return 0;
828 
829 err_free:
830 	rte_free(intr_handle->intr_vec);
831 err_disable:
832 	rte_intr_efd_disable(intr_handle);
833 err_out:
834 	/* Some of the error status returned by FW may not be from errno.h */
835 	if (rc > 0)
836 		rc = -EIO;
837 
838 	return rc;
839 }
840 
841 static int bnxt_shutdown_nic(struct bnxt *bp)
842 {
843 	bnxt_free_all_hwrm_resources(bp);
844 	bnxt_free_all_filters(bp);
845 	bnxt_free_all_vnics(bp);
846 	return 0;
847 }
848 
849 /*
850  * Device configuration and status function
851  */
852 
853 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
854 {
855 	uint32_t link_speed = bp->link_info->support_speeds;
856 	uint32_t speed_capa = 0;
857 
858 	/* If PAM4 is configured, use PAM4 supported speed */
859 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
860 		link_speed = bp->link_info->support_pam4_speeds;
861 
862 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
863 		speed_capa |= ETH_LINK_SPEED_100M;
864 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
865 		speed_capa |= ETH_LINK_SPEED_100M_HD;
866 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
867 		speed_capa |= ETH_LINK_SPEED_1G;
868 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
869 		speed_capa |= ETH_LINK_SPEED_2_5G;
870 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
871 		speed_capa |= ETH_LINK_SPEED_10G;
872 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
873 		speed_capa |= ETH_LINK_SPEED_20G;
874 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
875 		speed_capa |= ETH_LINK_SPEED_25G;
876 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
877 		speed_capa |= ETH_LINK_SPEED_40G;
878 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
879 		speed_capa |= ETH_LINK_SPEED_50G;
880 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
881 		speed_capa |= ETH_LINK_SPEED_100G;
882 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
883 		speed_capa |= ETH_LINK_SPEED_50G;
884 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
885 		speed_capa |= ETH_LINK_SPEED_100G;
886 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
887 		speed_capa |= ETH_LINK_SPEED_200G;
888 
889 	if (bp->link_info->auto_mode ==
890 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
891 		speed_capa |= ETH_LINK_SPEED_FIXED;
892 	else
893 		speed_capa |= ETH_LINK_SPEED_AUTONEG;
894 
895 	return speed_capa;
896 }
897 
898 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
899 				struct rte_eth_dev_info *dev_info)
900 {
901 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
902 	struct bnxt *bp = eth_dev->data->dev_private;
903 	uint16_t max_vnics, i, j, vpool, vrxq;
904 	unsigned int max_rx_rings;
905 	int rc;
906 
907 	rc = is_bnxt_in_error(bp);
908 	if (rc)
909 		return rc;
910 
911 	/* MAC Specifics */
912 	dev_info->max_mac_addrs = bp->max_l2_ctx;
913 	dev_info->max_hash_mac_addrs = 0;
914 
915 	/* PF/VF specifics */
916 	if (BNXT_PF(bp))
917 		dev_info->max_vfs = pdev->max_vfs;
918 
919 	max_rx_rings = bnxt_max_rings(bp);
920 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
921 	dev_info->max_rx_queues = max_rx_rings;
922 	dev_info->max_tx_queues = max_rx_rings;
923 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
924 	dev_info->hash_key_size = 40;
925 	max_vnics = bp->max_vnics;
926 
927 	/* MTU specifics */
928 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
929 	dev_info->max_mtu = BNXT_MAX_MTU;
930 
931 	/* Fast path specifics */
932 	dev_info->min_rx_bufsize = 1;
933 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
934 
935 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
936 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
937 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
938 	dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
939 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
940 				    dev_info->tx_queue_offload_capa;
941 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
942 
943 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
944 
945 	/* *INDENT-OFF* */
946 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
947 		.rx_thresh = {
948 			.pthresh = 8,
949 			.hthresh = 8,
950 			.wthresh = 0,
951 		},
952 		.rx_free_thresh = 32,
953 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
954 	};
955 
956 	dev_info->default_txconf = (struct rte_eth_txconf) {
957 		.tx_thresh = {
958 			.pthresh = 32,
959 			.hthresh = 0,
960 			.wthresh = 0,
961 		},
962 		.tx_free_thresh = 32,
963 		.tx_rs_thresh = 32,
964 	};
965 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
966 
967 	eth_dev->data->dev_conf.intr_conf.rxq = 1;
968 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
969 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
970 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
971 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
972 
973 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
974 		dev_info->switch_info.name = eth_dev->device->name;
975 		dev_info->switch_info.domain_id = bp->switch_domain_id;
976 		dev_info->switch_info.port_id =
977 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
978 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
979 	}
980 
981 	/* *INDENT-ON* */
982 
983 	/*
984 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
985 	 *       need further investigation.
986 	 */
987 
988 	/* VMDq resources */
989 	vpool = 64; /* ETH_64_POOLS */
990 	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
991 	for (i = 0; i < 4; vpool >>= 1, i++) {
992 		if (max_vnics > vpool) {
993 			for (j = 0; j < 5; vrxq >>= 1, j++) {
994 				if (dev_info->max_rx_queues > vrxq) {
995 					if (vpool > vrxq)
996 						vpool = vrxq;
997 					goto found;
998 				}
999 			}
1000 			/* Not enough resources to support VMDq */
1001 			break;
1002 		}
1003 	}
1004 	/* Not enough resources to support VMDq */
1005 	vpool = 0;
1006 	vrxq = 0;
1007 found:
1008 	dev_info->max_vmdq_pools = vpool;
1009 	dev_info->vmdq_queue_num = vrxq;
1010 
1011 	dev_info->vmdq_pool_base = 0;
1012 	dev_info->vmdq_queue_base = 0;
1013 
1014 	return 0;
1015 }
1016 
1017 /* Configure the device based on the configuration provided */
1018 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1019 {
1020 	struct bnxt *bp = eth_dev->data->dev_private;
1021 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1022 	int rc;
1023 
1024 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1025 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1026 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1027 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1028 
1029 	rc = is_bnxt_in_error(bp);
1030 	if (rc)
1031 		return rc;
1032 
1033 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1034 		rc = bnxt_hwrm_check_vf_rings(bp);
1035 		if (rc) {
1036 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1037 			return -ENOSPC;
1038 		}
1039 
1040 		/* If a resource has already been allocated - in this case
1041 		 * it is the async completion ring, free it. Reallocate it after
1042 		 * resource reservation. This will ensure the resource counts
1043 		 * are calculated correctly.
1044 		 */
1045 
1046 		pthread_mutex_lock(&bp->def_cp_lock);
1047 
1048 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1049 			bnxt_disable_int(bp);
1050 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1051 		}
1052 
1053 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1054 		if (rc) {
1055 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1056 			pthread_mutex_unlock(&bp->def_cp_lock);
1057 			return -ENOSPC;
1058 		}
1059 
1060 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1061 			rc = bnxt_alloc_async_cp_ring(bp);
1062 			if (rc) {
1063 				pthread_mutex_unlock(&bp->def_cp_lock);
1064 				return rc;
1065 			}
1066 			bnxt_enable_int(bp);
1067 		}
1068 
1069 		pthread_mutex_unlock(&bp->def_cp_lock);
1070 	}
1071 
1072 	/* Inherit new configurations */
1073 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1074 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1075 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1076 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1077 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1078 	    bp->max_stat_ctx)
1079 		goto resource_error;
1080 
1081 	if (BNXT_HAS_RING_GRPS(bp) &&
1082 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1083 		goto resource_error;
1084 
1085 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1086 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1087 		goto resource_error;
1088 
1089 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1090 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1091 
1092 	if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1093 		rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1094 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1095 
1096 	if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1097 		eth_dev->data->mtu =
1098 			eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1099 			RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1100 			BNXT_NUM_VLANS;
1101 		bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1102 	}
1103 	return 0;
1104 
1105 resource_error:
1106 	PMD_DRV_LOG(ERR,
1107 		    "Insufficient resources to support requested config\n");
1108 	PMD_DRV_LOG(ERR,
1109 		    "Num Queues Requested: Tx %d, Rx %d\n",
1110 		    eth_dev->data->nb_tx_queues,
1111 		    eth_dev->data->nb_rx_queues);
1112 	PMD_DRV_LOG(ERR,
1113 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1114 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1115 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1116 	return -ENOSPC;
1117 }
1118 
1119 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1120 {
1121 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1122 
1123 	if (link->link_status)
1124 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1125 			eth_dev->data->port_id,
1126 			(uint32_t)link->link_speed,
1127 			(link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1128 			("full-duplex") : ("half-duplex\n"));
1129 	else
1130 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1131 			eth_dev->data->port_id);
1132 }
1133 
1134 /*
1135  * Determine whether the current configuration requires support for scattered
1136  * receive; return 1 if scattered receive is required and 0 if not.
1137  */
1138 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1139 {
1140 	uint16_t buf_size;
1141 	int i;
1142 
1143 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1144 		return 1;
1145 
1146 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1147 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1148 
1149 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1150 				      RTE_PKTMBUF_HEADROOM);
1151 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1152 			return 1;
1153 	}
1154 	return 0;
1155 }
1156 
1157 static eth_rx_burst_t
1158 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1159 {
1160 	struct bnxt *bp = eth_dev->data->dev_private;
1161 
1162 	/* Disable vector mode RX for Stingray2 for now */
1163 	if (BNXT_CHIP_SR2(bp)) {
1164 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1165 		return bnxt_recv_pkts;
1166 	}
1167 
1168 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1169 #ifndef RTE_LIBRTE_IEEE1588
1170 	/*
1171 	 * Vector mode receive can be enabled only if scatter rx is not
1172 	 * in use and rx offloads are limited to VLAN stripping and
1173 	 * CRC stripping.
1174 	 */
1175 	if (!eth_dev->data->scattered_rx &&
1176 	    !(eth_dev->data->dev_conf.rxmode.offloads &
1177 	      ~(DEV_RX_OFFLOAD_VLAN_STRIP |
1178 		DEV_RX_OFFLOAD_KEEP_CRC |
1179 		DEV_RX_OFFLOAD_JUMBO_FRAME |
1180 		DEV_RX_OFFLOAD_IPV4_CKSUM |
1181 		DEV_RX_OFFLOAD_UDP_CKSUM |
1182 		DEV_RX_OFFLOAD_TCP_CKSUM |
1183 		DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1184 		DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1185 		DEV_RX_OFFLOAD_RSS_HASH |
1186 		DEV_RX_OFFLOAD_VLAN_FILTER)) &&
1187 	    !BNXT_TRUFLOW_EN(bp) && BNXT_NUM_ASYNC_CPR(bp) &&
1188 	    rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1189 		PMD_DRV_LOG(INFO, "Using vector mode receive for port %d\n",
1190 			    eth_dev->data->port_id);
1191 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1192 		return bnxt_recv_pkts_vec;
1193 	}
1194 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1195 		    eth_dev->data->port_id);
1196 	PMD_DRV_LOG(INFO,
1197 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1198 		    eth_dev->data->port_id,
1199 		    eth_dev->data->scattered_rx,
1200 		    eth_dev->data->dev_conf.rxmode.offloads);
1201 #endif
1202 #endif
1203 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1204 	return bnxt_recv_pkts;
1205 }
1206 
1207 static eth_tx_burst_t
1208 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1209 {
1210 	struct bnxt *bp = eth_dev->data->dev_private;
1211 
1212 	/* Disable vector mode TX for Stingray2 for now */
1213 	if (BNXT_CHIP_SR2(bp))
1214 		return bnxt_xmit_pkts;
1215 
1216 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1217 #ifndef RTE_LIBRTE_IEEE1588
1218 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1219 
1220 	/*
1221 	 * Vector mode transmit can be enabled only if not using scatter rx
1222 	 * or tx offloads.
1223 	 */
1224 	if (!eth_dev->data->scattered_rx &&
1225 	    !(offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) &&
1226 	    !BNXT_TRUFLOW_EN(bp) &&
1227 	    rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1228 		PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
1229 			    eth_dev->data->port_id);
1230 		return bnxt_xmit_pkts_vec;
1231 	}
1232 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1233 		    eth_dev->data->port_id);
1234 	PMD_DRV_LOG(INFO,
1235 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1236 		    eth_dev->data->port_id,
1237 		    eth_dev->data->scattered_rx,
1238 		    offloads);
1239 #endif
1240 #endif
1241 	return bnxt_xmit_pkts;
1242 }
1243 
1244 static int bnxt_handle_if_change_status(struct bnxt *bp)
1245 {
1246 	int rc;
1247 
1248 	/* Since fw has undergone a reset and lost all contexts,
1249 	 * set fatal flag to not issue hwrm during cleanup
1250 	 */
1251 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1252 	bnxt_uninit_resources(bp, true);
1253 
1254 	/* clear fatal flag so that re-init happens */
1255 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1256 	rc = bnxt_init_resources(bp, true);
1257 
1258 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1259 
1260 	return rc;
1261 }
1262 
1263 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1264 {
1265 	struct bnxt *bp = eth_dev->data->dev_private;
1266 	int rc = 0;
1267 
1268 	if (!bp->link_info->link_up)
1269 		rc = bnxt_set_hwrm_link_config(bp, true);
1270 	if (!rc)
1271 		eth_dev->data->dev_link.link_status = 1;
1272 
1273 	bnxt_print_link_info(eth_dev);
1274 	return rc;
1275 }
1276 
1277 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1278 {
1279 	struct bnxt *bp = eth_dev->data->dev_private;
1280 
1281 	eth_dev->data->dev_link.link_status = 0;
1282 	bnxt_set_hwrm_link_config(bp, false);
1283 	bp->link_info->link_up = 0;
1284 
1285 	return 0;
1286 }
1287 
1288 static void bnxt_free_switch_domain(struct bnxt *bp)
1289 {
1290 	int rc = 0;
1291 
1292 	if (bp->switch_domain_id) {
1293 		rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1294 		if (rc)
1295 			PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1296 				    bp->switch_domain_id, rc);
1297 	}
1298 }
1299 
1300 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1301 {
1302 	struct bnxt *bp = eth_dev->data->dev_private;
1303 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1304 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1305 	struct rte_eth_link link;
1306 	int ret;
1307 
1308 	eth_dev->data->dev_started = 0;
1309 	eth_dev->data->scattered_rx = 0;
1310 
1311 	/* Prevent crashes when queues are still in use */
1312 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1313 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1314 
1315 	bnxt_disable_int(bp);
1316 
1317 	/* disable uio/vfio intr/eventfd mapping */
1318 	rte_intr_disable(intr_handle);
1319 
1320 	/* Stop the child representors for this device */
1321 	ret = bnxt_rep_stop_all(bp);
1322 	if (ret != 0)
1323 		return ret;
1324 
1325 	/* delete the bnxt ULP port details */
1326 	bnxt_ulp_port_deinit(bp);
1327 
1328 	bnxt_cancel_fw_health_check(bp);
1329 
1330 	/* Do not bring link down during reset recovery */
1331 	if (!is_bnxt_in_error(bp)) {
1332 		bnxt_dev_set_link_down_op(eth_dev);
1333 		/* Wait for link to be reset */
1334 		if (BNXT_SINGLE_PF(bp))
1335 			rte_delay_ms(500);
1336 		/* clear the recorded link status */
1337 		memset(&link, 0, sizeof(link));
1338 		rte_eth_linkstatus_set(eth_dev, &link);
1339 	}
1340 
1341 	/* Clean queue intr-vector mapping */
1342 	rte_intr_efd_disable(intr_handle);
1343 	if (intr_handle->intr_vec != NULL) {
1344 		rte_free(intr_handle->intr_vec);
1345 		intr_handle->intr_vec = NULL;
1346 	}
1347 
1348 	bnxt_hwrm_port_clr_stats(bp);
1349 	bnxt_free_tx_mbufs(bp);
1350 	bnxt_free_rx_mbufs(bp);
1351 	/* Process any remaining notifications in default completion queue */
1352 	bnxt_int_handler(eth_dev);
1353 	bnxt_shutdown_nic(bp);
1354 	bnxt_hwrm_if_change(bp, false);
1355 
1356 	rte_free(bp->mark_table);
1357 	bp->mark_table = NULL;
1358 
1359 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1360 	bp->rx_cosq_cnt = 0;
1361 	/* All filters are deleted on a port stop. */
1362 	if (BNXT_FLOW_XSTATS_EN(bp))
1363 		bp->flow_stat->flow_count = 0;
1364 
1365 	return 0;
1366 }
1367 
1368 /* Unload the driver, release resources */
1369 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1370 {
1371 	struct bnxt *bp = eth_dev->data->dev_private;
1372 
1373 	pthread_mutex_lock(&bp->err_recovery_lock);
1374 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1375 		PMD_DRV_LOG(ERR,
1376 			    "Adapter recovering from error..Please retry\n");
1377 		return -EAGAIN;
1378 	}
1379 	pthread_mutex_unlock(&bp->err_recovery_lock);
1380 
1381 	return bnxt_dev_stop(eth_dev);
1382 }
1383 
1384 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1385 {
1386 	struct bnxt *bp = eth_dev->data->dev_private;
1387 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1388 	int vlan_mask = 0;
1389 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1390 
1391 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1392 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1393 		return -EINVAL;
1394 	}
1395 
1396 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1397 		PMD_DRV_LOG(ERR,
1398 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1399 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1400 
1401 	do {
1402 		rc = bnxt_hwrm_if_change(bp, true);
1403 		if (rc == 0 || rc != -EAGAIN)
1404 			break;
1405 
1406 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1407 	} while (retry_cnt--);
1408 
1409 	if (rc)
1410 		return rc;
1411 
1412 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1413 		rc = bnxt_handle_if_change_status(bp);
1414 		if (rc)
1415 			return rc;
1416 	}
1417 
1418 	bnxt_enable_int(bp);
1419 
1420 	rc = bnxt_init_chip(bp);
1421 	if (rc)
1422 		goto error;
1423 
1424 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1425 	eth_dev->data->dev_started = 1;
1426 
1427 	bnxt_link_update_op(eth_dev, 1);
1428 
1429 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1430 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1431 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1432 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1433 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1434 	if (rc)
1435 		goto error;
1436 
1437 	/* Initialize bnxt ULP port details */
1438 	rc = bnxt_ulp_port_init(bp);
1439 	if (rc)
1440 		goto error;
1441 
1442 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1443 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1444 
1445 	bnxt_schedule_fw_health_check(bp);
1446 
1447 	return 0;
1448 
1449 error:
1450 	bnxt_dev_stop(eth_dev);
1451 	return rc;
1452 }
1453 
1454 static void
1455 bnxt_uninit_locks(struct bnxt *bp)
1456 {
1457 	pthread_mutex_destroy(&bp->flow_lock);
1458 	pthread_mutex_destroy(&bp->def_cp_lock);
1459 	pthread_mutex_destroy(&bp->health_check_lock);
1460 	pthread_mutex_destroy(&bp->err_recovery_lock);
1461 	if (bp->rep_info) {
1462 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1463 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1464 	}
1465 }
1466 
1467 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1468 {
1469 	struct bnxt *bp = eth_dev->data->dev_private;
1470 	int ret = 0;
1471 
1472 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1473 		return 0;
1474 
1475 	pthread_mutex_lock(&bp->err_recovery_lock);
1476 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1477 		PMD_DRV_LOG(ERR,
1478 			    "Adapter recovering from error...Please retry\n");
1479 		return -EAGAIN;
1480 	}
1481 	pthread_mutex_unlock(&bp->err_recovery_lock);
1482 
1483 	/* cancel the recovery handler before remove dev */
1484 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1485 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1486 	bnxt_cancel_fc_thread(bp);
1487 
1488 	if (eth_dev->data->dev_started)
1489 		ret = bnxt_dev_stop(eth_dev);
1490 
1491 	bnxt_free_switch_domain(bp);
1492 
1493 	bnxt_uninit_resources(bp, false);
1494 
1495 	bnxt_free_leds_info(bp);
1496 	bnxt_free_cos_queues(bp);
1497 	bnxt_free_link_info(bp);
1498 	bnxt_free_pf_info(bp);
1499 	bnxt_free_parent_info(bp);
1500 	bnxt_uninit_locks(bp);
1501 
1502 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1503 	bp->tx_mem_zone = NULL;
1504 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1505 	bp->rx_mem_zone = NULL;
1506 
1507 	bnxt_hwrm_free_vf_info(bp);
1508 
1509 	rte_free(bp->grp_info);
1510 	bp->grp_info = NULL;
1511 
1512 	return ret;
1513 }
1514 
1515 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1516 				    uint32_t index)
1517 {
1518 	struct bnxt *bp = eth_dev->data->dev_private;
1519 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1520 	struct bnxt_vnic_info *vnic;
1521 	struct bnxt_filter_info *filter, *temp_filter;
1522 	uint32_t i;
1523 
1524 	if (is_bnxt_in_error(bp))
1525 		return;
1526 
1527 	/*
1528 	 * Loop through all VNICs from the specified filter flow pools to
1529 	 * remove the corresponding MAC addr filter
1530 	 */
1531 	for (i = 0; i < bp->nr_vnics; i++) {
1532 		if (!(pool_mask & (1ULL << i)))
1533 			continue;
1534 
1535 		vnic = &bp->vnic_info[i];
1536 		filter = STAILQ_FIRST(&vnic->filter);
1537 		while (filter) {
1538 			temp_filter = STAILQ_NEXT(filter, next);
1539 			if (filter->mac_index == index) {
1540 				STAILQ_REMOVE(&vnic->filter, filter,
1541 						bnxt_filter_info, next);
1542 				bnxt_hwrm_clear_l2_filter(bp, filter);
1543 				bnxt_free_filter(bp, filter);
1544 			}
1545 			filter = temp_filter;
1546 		}
1547 	}
1548 }
1549 
1550 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1551 			       struct rte_ether_addr *mac_addr, uint32_t index,
1552 			       uint32_t pool)
1553 {
1554 	struct bnxt_filter_info *filter;
1555 	int rc = 0;
1556 
1557 	/* Attach requested MAC address to the new l2_filter */
1558 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1559 		if (filter->mac_index == index) {
1560 			PMD_DRV_LOG(DEBUG,
1561 				    "MAC addr already existed for pool %d\n",
1562 				    pool);
1563 			return 0;
1564 		}
1565 	}
1566 
1567 	filter = bnxt_alloc_filter(bp);
1568 	if (!filter) {
1569 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1570 		return -ENODEV;
1571 	}
1572 
1573 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1574 	 * if the MAC that's been programmed now is a different one, then,
1575 	 * copy that addr to filter->l2_addr
1576 	 */
1577 	if (mac_addr)
1578 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1579 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1580 
1581 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1582 	if (!rc) {
1583 		filter->mac_index = index;
1584 		if (filter->mac_index == 0)
1585 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1586 		else
1587 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1588 	} else {
1589 		bnxt_free_filter(bp, filter);
1590 	}
1591 
1592 	return rc;
1593 }
1594 
1595 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1596 				struct rte_ether_addr *mac_addr,
1597 				uint32_t index, uint32_t pool)
1598 {
1599 	struct bnxt *bp = eth_dev->data->dev_private;
1600 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1601 	int rc = 0;
1602 
1603 	rc = is_bnxt_in_error(bp);
1604 	if (rc)
1605 		return rc;
1606 
1607 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1608 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1609 		return -ENOTSUP;
1610 	}
1611 
1612 	if (!vnic) {
1613 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1614 		return -EINVAL;
1615 	}
1616 
1617 	/* Filter settings will get applied when port is started */
1618 	if (!eth_dev->data->dev_started)
1619 		return 0;
1620 
1621 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1622 
1623 	return rc;
1624 }
1625 
1626 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1627 {
1628 	int rc = 0;
1629 	struct bnxt *bp = eth_dev->data->dev_private;
1630 	struct rte_eth_link new;
1631 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1632 			BNXT_MIN_LINK_WAIT_CNT;
1633 
1634 	rc = is_bnxt_in_error(bp);
1635 	if (rc)
1636 		return rc;
1637 
1638 	memset(&new, 0, sizeof(new));
1639 	do {
1640 		/* Retrieve link info from hardware */
1641 		rc = bnxt_get_hwrm_link_config(bp, &new);
1642 		if (rc) {
1643 			new.link_speed = ETH_LINK_SPEED_100M;
1644 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1645 			PMD_DRV_LOG(ERR,
1646 				"Failed to retrieve link rc = 0x%x!\n", rc);
1647 			goto out;
1648 		}
1649 
1650 		if (!wait_to_complete || new.link_status)
1651 			break;
1652 
1653 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1654 	} while (cnt--);
1655 
1656 	/* Only single function PF can bring phy down.
1657 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1658 	 */
1659 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1660 		memset(&new, 0, sizeof(new));
1661 
1662 out:
1663 	/* Timed out or success */
1664 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1665 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1666 		rte_eth_linkstatus_set(eth_dev, &new);
1667 
1668 		rte_eth_dev_callback_process(eth_dev,
1669 					     RTE_ETH_EVENT_INTR_LSC,
1670 					     NULL);
1671 
1672 		bnxt_print_link_info(eth_dev);
1673 	}
1674 
1675 	return rc;
1676 }
1677 
1678 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1679 {
1680 	struct bnxt *bp = eth_dev->data->dev_private;
1681 	struct bnxt_vnic_info *vnic;
1682 	uint32_t old_flags;
1683 	int rc;
1684 
1685 	rc = is_bnxt_in_error(bp);
1686 	if (rc)
1687 		return rc;
1688 
1689 	/* Filter settings will get applied when port is started */
1690 	if (!eth_dev->data->dev_started)
1691 		return 0;
1692 
1693 	if (bp->vnic_info == NULL)
1694 		return 0;
1695 
1696 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1697 
1698 	old_flags = vnic->flags;
1699 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1700 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1701 	if (rc != 0)
1702 		vnic->flags = old_flags;
1703 
1704 	return rc;
1705 }
1706 
1707 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1708 {
1709 	struct bnxt *bp = eth_dev->data->dev_private;
1710 	struct bnxt_vnic_info *vnic;
1711 	uint32_t old_flags;
1712 	int rc;
1713 
1714 	rc = is_bnxt_in_error(bp);
1715 	if (rc)
1716 		return rc;
1717 
1718 	/* Filter settings will get applied when port is started */
1719 	if (!eth_dev->data->dev_started)
1720 		return 0;
1721 
1722 	if (bp->vnic_info == NULL)
1723 		return 0;
1724 
1725 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1726 
1727 	old_flags = vnic->flags;
1728 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1729 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1730 	if (rc != 0)
1731 		vnic->flags = old_flags;
1732 
1733 	return rc;
1734 }
1735 
1736 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1737 {
1738 	struct bnxt *bp = eth_dev->data->dev_private;
1739 	struct bnxt_vnic_info *vnic;
1740 	uint32_t old_flags;
1741 	int rc;
1742 
1743 	rc = is_bnxt_in_error(bp);
1744 	if (rc)
1745 		return rc;
1746 
1747 	/* Filter settings will get applied when port is started */
1748 	if (!eth_dev->data->dev_started)
1749 		return 0;
1750 
1751 	if (bp->vnic_info == NULL)
1752 		return 0;
1753 
1754 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1755 
1756 	old_flags = vnic->flags;
1757 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1758 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1759 	if (rc != 0)
1760 		vnic->flags = old_flags;
1761 
1762 	return rc;
1763 }
1764 
1765 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1766 {
1767 	struct bnxt *bp = eth_dev->data->dev_private;
1768 	struct bnxt_vnic_info *vnic;
1769 	uint32_t old_flags;
1770 	int rc;
1771 
1772 	rc = is_bnxt_in_error(bp);
1773 	if (rc)
1774 		return rc;
1775 
1776 	/* Filter settings will get applied when port is started */
1777 	if (!eth_dev->data->dev_started)
1778 		return 0;
1779 
1780 	if (bp->vnic_info == NULL)
1781 		return 0;
1782 
1783 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1784 
1785 	old_flags = vnic->flags;
1786 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1787 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1788 	if (rc != 0)
1789 		vnic->flags = old_flags;
1790 
1791 	return rc;
1792 }
1793 
1794 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1795 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1796 {
1797 	if (qid >= bp->rx_nr_rings)
1798 		return NULL;
1799 
1800 	return bp->eth_dev->data->rx_queues[qid];
1801 }
1802 
1803 /* Return rxq corresponding to a given rss table ring/group ID. */
1804 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1805 {
1806 	struct bnxt_rx_queue *rxq;
1807 	unsigned int i;
1808 
1809 	if (!BNXT_HAS_RING_GRPS(bp)) {
1810 		for (i = 0; i < bp->rx_nr_rings; i++) {
1811 			rxq = bp->eth_dev->data->rx_queues[i];
1812 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1813 				return rxq->index;
1814 		}
1815 	} else {
1816 		for (i = 0; i < bp->rx_nr_rings; i++) {
1817 			if (bp->grp_info[i].fw_grp_id == fwr)
1818 				return i;
1819 		}
1820 	}
1821 
1822 	return INVALID_HW_RING_ID;
1823 }
1824 
1825 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1826 			    struct rte_eth_rss_reta_entry64 *reta_conf,
1827 			    uint16_t reta_size)
1828 {
1829 	struct bnxt *bp = eth_dev->data->dev_private;
1830 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1831 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1832 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1833 	uint16_t idx, sft;
1834 	int i, rc;
1835 
1836 	rc = is_bnxt_in_error(bp);
1837 	if (rc)
1838 		return rc;
1839 
1840 	if (!vnic->rss_table)
1841 		return -EINVAL;
1842 
1843 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
1844 		return -EINVAL;
1845 
1846 	if (reta_size != tbl_size) {
1847 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1848 			"(%d) must equal the size supported by the hardware "
1849 			"(%d)\n", reta_size, tbl_size);
1850 		return -EINVAL;
1851 	}
1852 
1853 	for (i = 0; i < reta_size; i++) {
1854 		struct bnxt_rx_queue *rxq;
1855 
1856 		idx = i / RTE_RETA_GROUP_SIZE;
1857 		sft = i % RTE_RETA_GROUP_SIZE;
1858 
1859 		if (!(reta_conf[idx].mask & (1ULL << sft)))
1860 			continue;
1861 
1862 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
1863 		if (!rxq) {
1864 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
1865 			return -EINVAL;
1866 		}
1867 
1868 		if (BNXT_CHIP_P5(bp)) {
1869 			vnic->rss_table[i * 2] =
1870 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
1871 			vnic->rss_table[i * 2 + 1] =
1872 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
1873 		} else {
1874 			vnic->rss_table[i] =
1875 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
1876 		}
1877 	}
1878 
1879 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1880 	return rc;
1881 }
1882 
1883 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
1884 			      struct rte_eth_rss_reta_entry64 *reta_conf,
1885 			      uint16_t reta_size)
1886 {
1887 	struct bnxt *bp = eth_dev->data->dev_private;
1888 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1889 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1890 	uint16_t idx, sft, i;
1891 	int rc;
1892 
1893 	rc = is_bnxt_in_error(bp);
1894 	if (rc)
1895 		return rc;
1896 
1897 	/* Retrieve from the default VNIC */
1898 	if (!vnic)
1899 		return -EINVAL;
1900 	if (!vnic->rss_table)
1901 		return -EINVAL;
1902 
1903 	if (reta_size != tbl_size) {
1904 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1905 			"(%d) must equal the size supported by the hardware "
1906 			"(%d)\n", reta_size, tbl_size);
1907 		return -EINVAL;
1908 	}
1909 
1910 	for (idx = 0, i = 0; i < reta_size; i++) {
1911 		idx = i / RTE_RETA_GROUP_SIZE;
1912 		sft = i % RTE_RETA_GROUP_SIZE;
1913 
1914 		if (reta_conf[idx].mask & (1ULL << sft)) {
1915 			uint16_t qid;
1916 
1917 			if (BNXT_CHIP_P5(bp))
1918 				qid = bnxt_rss_to_qid(bp,
1919 						      vnic->rss_table[i * 2]);
1920 			else
1921 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
1922 
1923 			if (qid == INVALID_HW_RING_ID) {
1924 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
1925 				return -EINVAL;
1926 			}
1927 			reta_conf[idx].reta[sft] = qid;
1928 		}
1929 	}
1930 
1931 	return 0;
1932 }
1933 
1934 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
1935 				   struct rte_eth_rss_conf *rss_conf)
1936 {
1937 	struct bnxt *bp = eth_dev->data->dev_private;
1938 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1939 	struct bnxt_vnic_info *vnic;
1940 	int rc;
1941 
1942 	rc = is_bnxt_in_error(bp);
1943 	if (rc)
1944 		return rc;
1945 
1946 	/*
1947 	 * If RSS enablement were different than dev_configure,
1948 	 * then return -EINVAL
1949 	 */
1950 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
1951 		if (!rss_conf->rss_hf)
1952 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
1953 	} else {
1954 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
1955 			return -EINVAL;
1956 	}
1957 
1958 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
1959 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
1960 	       rss_conf,
1961 	       sizeof(*rss_conf));
1962 
1963 	/* Update the default RSS VNIC(s) */
1964 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1965 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
1966 	vnic->hash_mode =
1967 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
1968 					    ETH_RSS_LEVEL(rss_conf->rss_hf));
1969 
1970 	/*
1971 	 * If hashkey is not specified, use the previously configured
1972 	 * hashkey
1973 	 */
1974 	if (!rss_conf->rss_key)
1975 		goto rss_config;
1976 
1977 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
1978 		PMD_DRV_LOG(ERR,
1979 			    "Invalid hashkey length, should be 16 bytes\n");
1980 		return -EINVAL;
1981 	}
1982 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
1983 
1984 rss_config:
1985 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1986 	return rc;
1987 }
1988 
1989 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
1990 				     struct rte_eth_rss_conf *rss_conf)
1991 {
1992 	struct bnxt *bp = eth_dev->data->dev_private;
1993 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1994 	int len, rc;
1995 	uint32_t hash_types;
1996 
1997 	rc = is_bnxt_in_error(bp);
1998 	if (rc)
1999 		return rc;
2000 
2001 	/* RSS configuration is the same for all VNICs */
2002 	if (vnic && vnic->rss_hash_key) {
2003 		if (rss_conf->rss_key) {
2004 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2005 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2006 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2007 		}
2008 
2009 		hash_types = vnic->hash_type;
2010 		rss_conf->rss_hf = 0;
2011 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2012 			rss_conf->rss_hf |= ETH_RSS_IPV4;
2013 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2014 		}
2015 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2016 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2017 			hash_types &=
2018 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2019 		}
2020 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2021 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2022 			hash_types &=
2023 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2024 		}
2025 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2026 			rss_conf->rss_hf |= ETH_RSS_IPV6;
2027 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2028 		}
2029 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2030 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2031 			hash_types &=
2032 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2033 		}
2034 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2035 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2036 			hash_types &=
2037 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2038 		}
2039 
2040 		rss_conf->rss_hf |=
2041 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2042 
2043 		if (hash_types) {
2044 			PMD_DRV_LOG(ERR,
2045 				"Unknown RSS config from firmware (%08x), RSS disabled",
2046 				vnic->hash_type);
2047 			return -ENOTSUP;
2048 		}
2049 	} else {
2050 		rss_conf->rss_hf = 0;
2051 	}
2052 	return 0;
2053 }
2054 
2055 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2056 			       struct rte_eth_fc_conf *fc_conf)
2057 {
2058 	struct bnxt *bp = dev->data->dev_private;
2059 	struct rte_eth_link link_info;
2060 	int rc;
2061 
2062 	rc = is_bnxt_in_error(bp);
2063 	if (rc)
2064 		return rc;
2065 
2066 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2067 	if (rc)
2068 		return rc;
2069 
2070 	memset(fc_conf, 0, sizeof(*fc_conf));
2071 	if (bp->link_info->auto_pause)
2072 		fc_conf->autoneg = 1;
2073 	switch (bp->link_info->pause) {
2074 	case 0:
2075 		fc_conf->mode = RTE_FC_NONE;
2076 		break;
2077 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2078 		fc_conf->mode = RTE_FC_TX_PAUSE;
2079 		break;
2080 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2081 		fc_conf->mode = RTE_FC_RX_PAUSE;
2082 		break;
2083 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2084 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2085 		fc_conf->mode = RTE_FC_FULL;
2086 		break;
2087 	}
2088 	return 0;
2089 }
2090 
2091 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2092 			       struct rte_eth_fc_conf *fc_conf)
2093 {
2094 	struct bnxt *bp = dev->data->dev_private;
2095 	int rc;
2096 
2097 	rc = is_bnxt_in_error(bp);
2098 	if (rc)
2099 		return rc;
2100 
2101 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2102 		PMD_DRV_LOG(ERR, "Flow Control Settings cannot be modified\n");
2103 		return -ENOTSUP;
2104 	}
2105 
2106 	switch (fc_conf->mode) {
2107 	case RTE_FC_NONE:
2108 		bp->link_info->auto_pause = 0;
2109 		bp->link_info->force_pause = 0;
2110 		break;
2111 	case RTE_FC_RX_PAUSE:
2112 		if (fc_conf->autoneg) {
2113 			bp->link_info->auto_pause =
2114 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2115 			bp->link_info->force_pause = 0;
2116 		} else {
2117 			bp->link_info->auto_pause = 0;
2118 			bp->link_info->force_pause =
2119 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2120 		}
2121 		break;
2122 	case RTE_FC_TX_PAUSE:
2123 		if (fc_conf->autoneg) {
2124 			bp->link_info->auto_pause =
2125 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2126 			bp->link_info->force_pause = 0;
2127 		} else {
2128 			bp->link_info->auto_pause = 0;
2129 			bp->link_info->force_pause =
2130 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2131 		}
2132 		break;
2133 	case RTE_FC_FULL:
2134 		if (fc_conf->autoneg) {
2135 			bp->link_info->auto_pause =
2136 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2137 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2138 			bp->link_info->force_pause = 0;
2139 		} else {
2140 			bp->link_info->auto_pause = 0;
2141 			bp->link_info->force_pause =
2142 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2143 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2144 		}
2145 		break;
2146 	}
2147 	return bnxt_set_hwrm_link_config(bp, true);
2148 }
2149 
2150 /* Add UDP tunneling port */
2151 static int
2152 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2153 			 struct rte_eth_udp_tunnel *udp_tunnel)
2154 {
2155 	struct bnxt *bp = eth_dev->data->dev_private;
2156 	uint16_t tunnel_type = 0;
2157 	int rc = 0;
2158 
2159 	rc = is_bnxt_in_error(bp);
2160 	if (rc)
2161 		return rc;
2162 
2163 	switch (udp_tunnel->prot_type) {
2164 	case RTE_TUNNEL_TYPE_VXLAN:
2165 		if (bp->vxlan_port_cnt) {
2166 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2167 				udp_tunnel->udp_port);
2168 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2169 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2170 				return -ENOSPC;
2171 			}
2172 			bp->vxlan_port_cnt++;
2173 			return 0;
2174 		}
2175 		tunnel_type =
2176 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2177 		bp->vxlan_port_cnt++;
2178 		break;
2179 	case RTE_TUNNEL_TYPE_GENEVE:
2180 		if (bp->geneve_port_cnt) {
2181 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2182 				udp_tunnel->udp_port);
2183 			if (bp->geneve_port != udp_tunnel->udp_port) {
2184 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2185 				return -ENOSPC;
2186 			}
2187 			bp->geneve_port_cnt++;
2188 			return 0;
2189 		}
2190 		tunnel_type =
2191 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2192 		bp->geneve_port_cnt++;
2193 		break;
2194 	default:
2195 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2196 		return -ENOTSUP;
2197 	}
2198 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2199 					     tunnel_type);
2200 	return rc;
2201 }
2202 
2203 static int
2204 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2205 			 struct rte_eth_udp_tunnel *udp_tunnel)
2206 {
2207 	struct bnxt *bp = eth_dev->data->dev_private;
2208 	uint16_t tunnel_type = 0;
2209 	uint16_t port = 0;
2210 	int rc = 0;
2211 
2212 	rc = is_bnxt_in_error(bp);
2213 	if (rc)
2214 		return rc;
2215 
2216 	switch (udp_tunnel->prot_type) {
2217 	case RTE_TUNNEL_TYPE_VXLAN:
2218 		if (!bp->vxlan_port_cnt) {
2219 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2220 			return -EINVAL;
2221 		}
2222 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2223 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2224 				udp_tunnel->udp_port, bp->vxlan_port);
2225 			return -EINVAL;
2226 		}
2227 		if (--bp->vxlan_port_cnt)
2228 			return 0;
2229 
2230 		tunnel_type =
2231 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2232 		port = bp->vxlan_fw_dst_port_id;
2233 		break;
2234 	case RTE_TUNNEL_TYPE_GENEVE:
2235 		if (!bp->geneve_port_cnt) {
2236 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2237 			return -EINVAL;
2238 		}
2239 		if (bp->geneve_port != udp_tunnel->udp_port) {
2240 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2241 				udp_tunnel->udp_port, bp->geneve_port);
2242 			return -EINVAL;
2243 		}
2244 		if (--bp->geneve_port_cnt)
2245 			return 0;
2246 
2247 		tunnel_type =
2248 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2249 		port = bp->geneve_fw_dst_port_id;
2250 		break;
2251 	default:
2252 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2253 		return -ENOTSUP;
2254 	}
2255 
2256 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2257 	return rc;
2258 }
2259 
2260 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2261 {
2262 	struct bnxt_filter_info *filter;
2263 	struct bnxt_vnic_info *vnic;
2264 	int rc = 0;
2265 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2266 
2267 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2268 	filter = STAILQ_FIRST(&vnic->filter);
2269 	while (filter) {
2270 		/* Search for this matching MAC+VLAN filter */
2271 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2272 			/* Delete the filter */
2273 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2274 			if (rc)
2275 				return rc;
2276 			STAILQ_REMOVE(&vnic->filter, filter,
2277 				      bnxt_filter_info, next);
2278 			bnxt_free_filter(bp, filter);
2279 			PMD_DRV_LOG(INFO,
2280 				    "Deleted vlan filter for %d\n",
2281 				    vlan_id);
2282 			return 0;
2283 		}
2284 		filter = STAILQ_NEXT(filter, next);
2285 	}
2286 	return -ENOENT;
2287 }
2288 
2289 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2290 {
2291 	struct bnxt_filter_info *filter;
2292 	struct bnxt_vnic_info *vnic;
2293 	int rc = 0;
2294 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2295 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2296 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2297 
2298 	/* Implementation notes on the use of VNIC in this command:
2299 	 *
2300 	 * By default, these filters belong to default vnic for the function.
2301 	 * Once these filters are set up, only destination VNIC can be modified.
2302 	 * If the destination VNIC is not specified in this command,
2303 	 * then the HWRM shall only create an l2 context id.
2304 	 */
2305 
2306 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2307 	filter = STAILQ_FIRST(&vnic->filter);
2308 	/* Check if the VLAN has already been added */
2309 	while (filter) {
2310 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2311 			return -EEXIST;
2312 
2313 		filter = STAILQ_NEXT(filter, next);
2314 	}
2315 
2316 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2317 	 * command to create MAC+VLAN filter with the right flags, enables set.
2318 	 */
2319 	filter = bnxt_alloc_filter(bp);
2320 	if (!filter) {
2321 		PMD_DRV_LOG(ERR,
2322 			    "MAC/VLAN filter alloc failed\n");
2323 		return -ENOMEM;
2324 	}
2325 	/* MAC + VLAN ID filter */
2326 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2327 	 * untagged packets are received
2328 	 *
2329 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2330 	 * packets and only the programmed vlan's packets are received
2331 	 */
2332 	filter->l2_ivlan = vlan_id;
2333 	filter->l2_ivlan_mask = 0x0FFF;
2334 	filter->enables |= en;
2335 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2336 
2337 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2338 	if (rc) {
2339 		/* Free the newly allocated filter as we were
2340 		 * not able to create the filter in hardware.
2341 		 */
2342 		bnxt_free_filter(bp, filter);
2343 		return rc;
2344 	}
2345 
2346 	filter->mac_index = 0;
2347 	/* Add this new filter to the list */
2348 	if (vlan_id == 0)
2349 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2350 	else
2351 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2352 
2353 	PMD_DRV_LOG(INFO,
2354 		    "Added Vlan filter for %d\n", vlan_id);
2355 	return rc;
2356 }
2357 
2358 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2359 		uint16_t vlan_id, int on)
2360 {
2361 	struct bnxt *bp = eth_dev->data->dev_private;
2362 	int rc;
2363 
2364 	rc = is_bnxt_in_error(bp);
2365 	if (rc)
2366 		return rc;
2367 
2368 	if (!eth_dev->data->dev_started) {
2369 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2370 		return -EINVAL;
2371 	}
2372 
2373 	/* These operations apply to ALL existing MAC/VLAN filters */
2374 	if (on)
2375 		return bnxt_add_vlan_filter(bp, vlan_id);
2376 	else
2377 		return bnxt_del_vlan_filter(bp, vlan_id);
2378 }
2379 
2380 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2381 				    struct bnxt_vnic_info *vnic)
2382 {
2383 	struct bnxt_filter_info *filter;
2384 	int rc;
2385 
2386 	filter = STAILQ_FIRST(&vnic->filter);
2387 	while (filter) {
2388 		if (filter->mac_index == 0 &&
2389 		    !memcmp(filter->l2_addr, bp->mac_addr,
2390 			    RTE_ETHER_ADDR_LEN)) {
2391 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2392 			if (!rc) {
2393 				STAILQ_REMOVE(&vnic->filter, filter,
2394 					      bnxt_filter_info, next);
2395 				bnxt_free_filter(bp, filter);
2396 			}
2397 			return rc;
2398 		}
2399 		filter = STAILQ_NEXT(filter, next);
2400 	}
2401 	return 0;
2402 }
2403 
2404 static int
2405 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2406 {
2407 	struct bnxt_vnic_info *vnic;
2408 	unsigned int i;
2409 	int rc;
2410 
2411 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2412 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2413 		/* Remove any VLAN filters programmed */
2414 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2415 			bnxt_del_vlan_filter(bp, i);
2416 
2417 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2418 		if (rc)
2419 			return rc;
2420 	} else {
2421 		/* Default filter will allow packets that match the
2422 		 * dest mac. So, it has to be deleted, otherwise, we
2423 		 * will endup receiving vlan packets for which the
2424 		 * filter is not programmed, when hw-vlan-filter
2425 		 * configuration is ON
2426 		 */
2427 		bnxt_del_dflt_mac_filter(bp, vnic);
2428 		/* This filter will allow only untagged packets */
2429 		bnxt_add_vlan_filter(bp, 0);
2430 	}
2431 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2432 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2433 
2434 	return 0;
2435 }
2436 
2437 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2438 {
2439 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2440 	unsigned int i;
2441 	int rc;
2442 
2443 	/* Destroy vnic filters and vnic */
2444 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2445 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2446 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2447 			bnxt_del_vlan_filter(bp, i);
2448 	}
2449 	bnxt_del_dflt_mac_filter(bp, vnic);
2450 
2451 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2452 	if (rc)
2453 		return rc;
2454 
2455 	rte_free(vnic->fw_grp_ids);
2456 	vnic->fw_grp_ids = NULL;
2457 
2458 	vnic->rx_queue_cnt = 0;
2459 
2460 	return 0;
2461 }
2462 
2463 static int
2464 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2465 {
2466 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2467 	int rc;
2468 
2469 	/* Destroy, recreate and reconfigure the default vnic */
2470 	rc = bnxt_free_one_vnic(bp, 0);
2471 	if (rc)
2472 		return rc;
2473 
2474 	/* default vnic 0 */
2475 	rc = bnxt_setup_one_vnic(bp, 0);
2476 	if (rc)
2477 		return rc;
2478 
2479 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2480 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2481 		rc = bnxt_add_vlan_filter(bp, 0);
2482 		if (rc)
2483 			return rc;
2484 		rc = bnxt_restore_vlan_filters(bp);
2485 		if (rc)
2486 			return rc;
2487 	} else {
2488 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2489 		if (rc)
2490 			return rc;
2491 	}
2492 
2493 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2494 	if (rc)
2495 		return rc;
2496 
2497 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2498 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2499 
2500 	return rc;
2501 }
2502 
2503 static int
2504 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2505 {
2506 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2507 	struct bnxt *bp = dev->data->dev_private;
2508 	int rc;
2509 
2510 	rc = is_bnxt_in_error(bp);
2511 	if (rc)
2512 		return rc;
2513 
2514 	/* Filter settings will get applied when port is started */
2515 	if (!dev->data->dev_started)
2516 		return 0;
2517 
2518 	if (mask & ETH_VLAN_FILTER_MASK) {
2519 		/* Enable or disable VLAN filtering */
2520 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2521 		if (rc)
2522 			return rc;
2523 	}
2524 
2525 	if (mask & ETH_VLAN_STRIP_MASK) {
2526 		/* Enable or disable VLAN stripping */
2527 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2528 		if (rc)
2529 			return rc;
2530 	}
2531 
2532 	if (mask & ETH_VLAN_EXTEND_MASK) {
2533 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2534 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2535 		else
2536 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2537 	}
2538 
2539 	return 0;
2540 }
2541 
2542 static int
2543 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2544 		      uint16_t tpid)
2545 {
2546 	struct bnxt *bp = dev->data->dev_private;
2547 	int qinq = dev->data->dev_conf.rxmode.offloads &
2548 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2549 
2550 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2551 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2552 		PMD_DRV_LOG(ERR,
2553 			    "Unsupported vlan type.");
2554 		return -EINVAL;
2555 	}
2556 	if (!qinq) {
2557 		PMD_DRV_LOG(ERR,
2558 			    "QinQ not enabled. Needs to be ON as we can "
2559 			    "accelerate only outer vlan\n");
2560 		return -EINVAL;
2561 	}
2562 
2563 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2564 		switch (tpid) {
2565 		case RTE_ETHER_TYPE_QINQ:
2566 			bp->outer_tpid_bd =
2567 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2568 				break;
2569 		case RTE_ETHER_TYPE_VLAN:
2570 			bp->outer_tpid_bd =
2571 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2572 				break;
2573 		case RTE_ETHER_TYPE_QINQ1:
2574 			bp->outer_tpid_bd =
2575 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2576 				break;
2577 		case RTE_ETHER_TYPE_QINQ2:
2578 			bp->outer_tpid_bd =
2579 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2580 				break;
2581 		case RTE_ETHER_TYPE_QINQ3:
2582 			bp->outer_tpid_bd =
2583 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2584 				break;
2585 		default:
2586 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2587 			return -EINVAL;
2588 		}
2589 		bp->outer_tpid_bd |= tpid;
2590 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2591 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2592 		PMD_DRV_LOG(ERR,
2593 			    "Can accelerate only outer vlan in QinQ\n");
2594 		return -EINVAL;
2595 	}
2596 
2597 	return 0;
2598 }
2599 
2600 static int
2601 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2602 			     struct rte_ether_addr *addr)
2603 {
2604 	struct bnxt *bp = dev->data->dev_private;
2605 	/* Default Filter is tied to VNIC 0 */
2606 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2607 	int rc;
2608 
2609 	rc = is_bnxt_in_error(bp);
2610 	if (rc)
2611 		return rc;
2612 
2613 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2614 		return -EPERM;
2615 
2616 	if (rte_is_zero_ether_addr(addr))
2617 		return -EINVAL;
2618 
2619 	/* Filter settings will get applied when port is started */
2620 	if (!dev->data->dev_started)
2621 		return 0;
2622 
2623 	/* Check if the requested MAC is already added */
2624 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2625 		return 0;
2626 
2627 	/* Destroy filter and re-create it */
2628 	bnxt_del_dflt_mac_filter(bp, vnic);
2629 
2630 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2631 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2632 		/* This filter will allow only untagged packets */
2633 		rc = bnxt_add_vlan_filter(bp, 0);
2634 	} else {
2635 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2636 	}
2637 
2638 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2639 	return rc;
2640 }
2641 
2642 static int
2643 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2644 			  struct rte_ether_addr *mc_addr_set,
2645 			  uint32_t nb_mc_addr)
2646 {
2647 	struct bnxt *bp = eth_dev->data->dev_private;
2648 	char *mc_addr_list = (char *)mc_addr_set;
2649 	struct bnxt_vnic_info *vnic;
2650 	uint32_t off = 0, i = 0;
2651 	int rc;
2652 
2653 	rc = is_bnxt_in_error(bp);
2654 	if (rc)
2655 		return rc;
2656 
2657 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2658 
2659 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2660 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2661 		goto allmulti;
2662 	}
2663 
2664 	/* TODO Check for Duplicate mcast addresses */
2665 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2666 	for (i = 0; i < nb_mc_addr; i++) {
2667 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2668 			RTE_ETHER_ADDR_LEN);
2669 		off += RTE_ETHER_ADDR_LEN;
2670 	}
2671 
2672 	vnic->mc_addr_cnt = i;
2673 	if (vnic->mc_addr_cnt)
2674 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2675 	else
2676 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2677 
2678 allmulti:
2679 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2680 }
2681 
2682 static int
2683 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2684 {
2685 	struct bnxt *bp = dev->data->dev_private;
2686 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2687 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2688 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2689 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2690 	int ret;
2691 
2692 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2693 			fw_major, fw_minor, fw_updt, fw_rsvd);
2694 
2695 	ret += 1; /* add the size of '\0' */
2696 	if (fw_size < (uint32_t)ret)
2697 		return ret;
2698 	else
2699 		return 0;
2700 }
2701 
2702 static void
2703 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2704 	struct rte_eth_rxq_info *qinfo)
2705 {
2706 	struct bnxt *bp = dev->data->dev_private;
2707 	struct bnxt_rx_queue *rxq;
2708 
2709 	if (is_bnxt_in_error(bp))
2710 		return;
2711 
2712 	rxq = dev->data->rx_queues[queue_id];
2713 
2714 	qinfo->mp = rxq->mb_pool;
2715 	qinfo->scattered_rx = dev->data->scattered_rx;
2716 	qinfo->nb_desc = rxq->nb_rx_desc;
2717 
2718 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2719 	qinfo->conf.rx_drop_en = rxq->drop_en;
2720 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2721 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2722 }
2723 
2724 static void
2725 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2726 	struct rte_eth_txq_info *qinfo)
2727 {
2728 	struct bnxt *bp = dev->data->dev_private;
2729 	struct bnxt_tx_queue *txq;
2730 
2731 	if (is_bnxt_in_error(bp))
2732 		return;
2733 
2734 	txq = dev->data->tx_queues[queue_id];
2735 
2736 	qinfo->nb_desc = txq->nb_tx_desc;
2737 
2738 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2739 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2740 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2741 
2742 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2743 	qinfo->conf.tx_rs_thresh = 0;
2744 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2745 	qinfo->conf.offloads = txq->offloads;
2746 }
2747 
2748 static const struct {
2749 	eth_rx_burst_t pkt_burst;
2750 	const char *info;
2751 } bnxt_rx_burst_info[] = {
2752 	{bnxt_recv_pkts,	"Scalar"},
2753 #if defined(RTE_ARCH_X86)
2754 	{bnxt_recv_pkts_vec,	"Vector SSE"},
2755 #elif defined(RTE_ARCH_ARM64)
2756 	{bnxt_recv_pkts_vec,	"Vector Neon"},
2757 #endif
2758 };
2759 
2760 static int
2761 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2762 		       struct rte_eth_burst_mode *mode)
2763 {
2764 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2765 	size_t i;
2766 
2767 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2768 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2769 			snprintf(mode->info, sizeof(mode->info), "%s",
2770 				 bnxt_rx_burst_info[i].info);
2771 			return 0;
2772 		}
2773 	}
2774 
2775 	return -EINVAL;
2776 }
2777 
2778 static const struct {
2779 	eth_tx_burst_t pkt_burst;
2780 	const char *info;
2781 } bnxt_tx_burst_info[] = {
2782 	{bnxt_xmit_pkts,	"Scalar"},
2783 #if defined(RTE_ARCH_X86)
2784 	{bnxt_xmit_pkts_vec,	"Vector SSE"},
2785 #elif defined(RTE_ARCH_ARM64)
2786 	{bnxt_xmit_pkts_vec,	"Vector Neon"},
2787 #endif
2788 };
2789 
2790 static int
2791 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2792 		       struct rte_eth_burst_mode *mode)
2793 {
2794 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2795 	size_t i;
2796 
2797 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2798 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2799 			snprintf(mode->info, sizeof(mode->info), "%s",
2800 				 bnxt_tx_burst_info[i].info);
2801 			return 0;
2802 		}
2803 	}
2804 
2805 	return -EINVAL;
2806 }
2807 
2808 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2809 {
2810 	struct bnxt *bp = eth_dev->data->dev_private;
2811 	uint32_t new_pkt_size;
2812 	uint32_t rc = 0;
2813 	uint32_t i;
2814 
2815 	rc = is_bnxt_in_error(bp);
2816 	if (rc)
2817 		return rc;
2818 
2819 	/* Exit if receive queues are not configured yet */
2820 	if (!eth_dev->data->nb_rx_queues)
2821 		return rc;
2822 
2823 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2824 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2825 
2826 	/*
2827 	 * Disallow any MTU change that would require scattered receive support
2828 	 * if it is not already enabled.
2829 	 */
2830 	if (eth_dev->data->dev_started &&
2831 	    !eth_dev->data->scattered_rx &&
2832 	    (new_pkt_size >
2833 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2834 		PMD_DRV_LOG(ERR,
2835 			    "MTU change would require scattered rx support. ");
2836 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2837 		return -EINVAL;
2838 	}
2839 
2840 	if (new_mtu > RTE_ETHER_MTU) {
2841 		bp->flags |= BNXT_FLAG_JUMBO;
2842 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
2843 			DEV_RX_OFFLOAD_JUMBO_FRAME;
2844 	} else {
2845 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
2846 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
2847 		bp->flags &= ~BNXT_FLAG_JUMBO;
2848 	}
2849 
2850 	/* Is there a change in mtu setting? */
2851 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2852 		return rc;
2853 
2854 	for (i = 0; i < bp->nr_vnics; i++) {
2855 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2856 		uint16_t size = 0;
2857 
2858 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
2859 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2860 		if (rc)
2861 			break;
2862 
2863 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2864 		size -= RTE_PKTMBUF_HEADROOM;
2865 
2866 		if (size < new_mtu) {
2867 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2868 			if (rc)
2869 				return rc;
2870 		}
2871 	}
2872 
2873 	if (!rc)
2874 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2875 
2876 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2877 
2878 	return rc;
2879 }
2880 
2881 static int
2882 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2883 {
2884 	struct bnxt *bp = dev->data->dev_private;
2885 	uint16_t vlan = bp->vlan;
2886 	int rc;
2887 
2888 	rc = is_bnxt_in_error(bp);
2889 	if (rc)
2890 		return rc;
2891 
2892 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2893 		PMD_DRV_LOG(ERR,
2894 			"PVID cannot be modified for this function\n");
2895 		return -ENOTSUP;
2896 	}
2897 	bp->vlan = on ? pvid : 0;
2898 
2899 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
2900 	if (rc)
2901 		bp->vlan = vlan;
2902 	return rc;
2903 }
2904 
2905 static int
2906 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
2907 {
2908 	struct bnxt *bp = dev->data->dev_private;
2909 	int rc;
2910 
2911 	rc = is_bnxt_in_error(bp);
2912 	if (rc)
2913 		return rc;
2914 
2915 	return bnxt_hwrm_port_led_cfg(bp, true);
2916 }
2917 
2918 static int
2919 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
2920 {
2921 	struct bnxt *bp = dev->data->dev_private;
2922 	int rc;
2923 
2924 	rc = is_bnxt_in_error(bp);
2925 	if (rc)
2926 		return rc;
2927 
2928 	return bnxt_hwrm_port_led_cfg(bp, false);
2929 }
2930 
2931 static uint32_t
2932 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
2933 {
2934 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
2935 	uint32_t desc = 0, raw_cons = 0, cons;
2936 	struct bnxt_cp_ring_info *cpr;
2937 	struct bnxt_rx_queue *rxq;
2938 	struct rx_pkt_cmpl *rxcmp;
2939 	int rc;
2940 
2941 	rc = is_bnxt_in_error(bp);
2942 	if (rc)
2943 		return rc;
2944 
2945 	rxq = dev->data->rx_queues[rx_queue_id];
2946 	cpr = rxq->cp_ring;
2947 	raw_cons = cpr->cp_raw_cons;
2948 
2949 	while (1) {
2950 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
2951 		rte_prefetch0(&cpr->cp_desc_ring[cons]);
2952 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2953 
2954 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct)) {
2955 			break;
2956 		} else {
2957 			raw_cons++;
2958 			desc++;
2959 		}
2960 	}
2961 
2962 	return desc;
2963 }
2964 
2965 static int
2966 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
2967 {
2968 	struct bnxt_rx_queue *rxq = (struct bnxt_rx_queue *)rx_queue;
2969 	struct bnxt_rx_ring_info *rxr;
2970 	struct bnxt_cp_ring_info *cpr;
2971 	struct rte_mbuf *rx_buf;
2972 	struct rx_pkt_cmpl *rxcmp;
2973 	uint32_t cons, cp_cons;
2974 	int rc;
2975 
2976 	if (!rxq)
2977 		return -EINVAL;
2978 
2979 	rc = is_bnxt_in_error(rxq->bp);
2980 	if (rc)
2981 		return rc;
2982 
2983 	cpr = rxq->cp_ring;
2984 	rxr = rxq->rx_ring;
2985 
2986 	if (offset >= rxq->nb_rx_desc)
2987 		return -EINVAL;
2988 
2989 	cons = RING_CMP(cpr->cp_ring_struct, offset);
2990 	cp_cons = cpr->cp_raw_cons;
2991 	rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2992 
2993 	if (cons > cp_cons) {
2994 		if (CMPL_VALID(rxcmp, cpr->valid))
2995 			return RTE_ETH_RX_DESC_DONE;
2996 	} else {
2997 		if (CMPL_VALID(rxcmp, !cpr->valid))
2998 			return RTE_ETH_RX_DESC_DONE;
2999 	}
3000 	rx_buf = rxr->rx_buf_ring[cons];
3001 	if (rx_buf == NULL || rx_buf == &rxq->fake_mbuf)
3002 		return RTE_ETH_RX_DESC_UNAVAIL;
3003 
3004 
3005 	return RTE_ETH_RX_DESC_AVAIL;
3006 }
3007 
3008 static int
3009 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3010 {
3011 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3012 	struct bnxt_tx_ring_info *txr;
3013 	struct bnxt_cp_ring_info *cpr;
3014 	struct bnxt_sw_tx_bd *tx_buf;
3015 	struct tx_pkt_cmpl *txcmp;
3016 	uint32_t cons, cp_cons;
3017 	int rc;
3018 
3019 	if (!txq)
3020 		return -EINVAL;
3021 
3022 	rc = is_bnxt_in_error(txq->bp);
3023 	if (rc)
3024 		return rc;
3025 
3026 	cpr = txq->cp_ring;
3027 	txr = txq->tx_ring;
3028 
3029 	if (offset >= txq->nb_tx_desc)
3030 		return -EINVAL;
3031 
3032 	cons = RING_CMP(cpr->cp_ring_struct, offset);
3033 	txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3034 	cp_cons = cpr->cp_raw_cons;
3035 
3036 	if (cons > cp_cons) {
3037 		if (CMPL_VALID(txcmp, cpr->valid))
3038 			return RTE_ETH_TX_DESC_UNAVAIL;
3039 	} else {
3040 		if (CMPL_VALID(txcmp, !cpr->valid))
3041 			return RTE_ETH_TX_DESC_UNAVAIL;
3042 	}
3043 	tx_buf = &txr->tx_buf_ring[cons];
3044 	if (tx_buf->mbuf == NULL)
3045 		return RTE_ETH_TX_DESC_DONE;
3046 
3047 	return RTE_ETH_TX_DESC_FULL;
3048 }
3049 
3050 int
3051 bnxt_filter_ctrl_op(struct rte_eth_dev *dev,
3052 		    enum rte_filter_type filter_type,
3053 		    enum rte_filter_op filter_op, void *arg)
3054 {
3055 	struct bnxt *bp = dev->data->dev_private;
3056 	int ret = 0;
3057 
3058 	if (!bp)
3059 		return -EIO;
3060 
3061 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3062 		struct bnxt_representor *vfr = dev->data->dev_private;
3063 		bp = vfr->parent_dev->data->dev_private;
3064 		/* parent is deleted while children are still valid */
3065 		if (!bp) {
3066 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error %d:%d\n",
3067 				    dev->data->port_id,
3068 				    filter_type,
3069 				    filter_op);
3070 			return -EIO;
3071 		}
3072 	}
3073 
3074 	ret = is_bnxt_in_error(bp);
3075 	if (ret)
3076 		return ret;
3077 
3078 	switch (filter_type) {
3079 	case RTE_ETH_FILTER_GENERIC:
3080 		if (filter_op != RTE_ETH_FILTER_GET)
3081 			return -EINVAL;
3082 
3083 		/* PMD supports thread-safe flow operations.  rte_flow API
3084 		 * functions can avoid mutex for multi-thread safety.
3085 		 */
3086 		dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3087 
3088 		if (BNXT_TRUFLOW_EN(bp))
3089 			*(const void **)arg = &bnxt_ulp_rte_flow_ops;
3090 		else
3091 			*(const void **)arg = &bnxt_flow_ops;
3092 		break;
3093 	default:
3094 		PMD_DRV_LOG(ERR,
3095 			"Filter type (%d) not supported", filter_type);
3096 		ret = -EINVAL;
3097 		break;
3098 	}
3099 	return ret;
3100 }
3101 
3102 static const uint32_t *
3103 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3104 {
3105 	static const uint32_t ptypes[] = {
3106 		RTE_PTYPE_L2_ETHER_VLAN,
3107 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3108 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3109 		RTE_PTYPE_L4_ICMP,
3110 		RTE_PTYPE_L4_TCP,
3111 		RTE_PTYPE_L4_UDP,
3112 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3113 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3114 		RTE_PTYPE_INNER_L4_ICMP,
3115 		RTE_PTYPE_INNER_L4_TCP,
3116 		RTE_PTYPE_INNER_L4_UDP,
3117 		RTE_PTYPE_UNKNOWN
3118 	};
3119 
3120 	if (!dev->rx_pkt_burst)
3121 		return NULL;
3122 
3123 	return ptypes;
3124 }
3125 
3126 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3127 			 int reg_win)
3128 {
3129 	uint32_t reg_base = *reg_arr & 0xfffff000;
3130 	uint32_t win_off;
3131 	int i;
3132 
3133 	for (i = 0; i < count; i++) {
3134 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3135 			return -ERANGE;
3136 	}
3137 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3138 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3139 	return 0;
3140 }
3141 
3142 static int bnxt_map_ptp_regs(struct bnxt *bp)
3143 {
3144 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3145 	uint32_t *reg_arr;
3146 	int rc, i;
3147 
3148 	reg_arr = ptp->rx_regs;
3149 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3150 	if (rc)
3151 		return rc;
3152 
3153 	reg_arr = ptp->tx_regs;
3154 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3155 	if (rc)
3156 		return rc;
3157 
3158 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3159 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3160 
3161 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3162 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3163 
3164 	return 0;
3165 }
3166 
3167 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3168 {
3169 	rte_write32(0, (uint8_t *)bp->bar0 +
3170 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3171 	rte_write32(0, (uint8_t *)bp->bar0 +
3172 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3173 }
3174 
3175 static uint64_t bnxt_cc_read(struct bnxt *bp)
3176 {
3177 	uint64_t ns;
3178 
3179 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3180 			      BNXT_GRCPF_REG_SYNC_TIME));
3181 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3182 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3183 	return ns;
3184 }
3185 
3186 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3187 {
3188 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3189 	uint32_t fifo;
3190 
3191 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3192 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3193 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3194 		return -EAGAIN;
3195 
3196 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3197 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3198 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3199 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3200 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3201 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3202 
3203 	return 0;
3204 }
3205 
3206 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3207 {
3208 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3209 	struct bnxt_pf_info *pf = bp->pf;
3210 	uint16_t port_id;
3211 	uint32_t fifo;
3212 
3213 	if (!ptp)
3214 		return -ENODEV;
3215 
3216 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3217 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3218 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3219 		return -EAGAIN;
3220 
3221 	port_id = pf->port_id;
3222 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3223 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3224 
3225 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3226 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3227 	if (fifo & BNXT_PTP_RX_FIFO_PENDING) {
3228 /*		bnxt_clr_rx_ts(bp);	  TBD  */
3229 		return -EBUSY;
3230 	}
3231 
3232 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3233 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3234 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3235 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3236 
3237 	return 0;
3238 }
3239 
3240 static int
3241 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3242 {
3243 	uint64_t ns;
3244 	struct bnxt *bp = dev->data->dev_private;
3245 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3246 
3247 	if (!ptp)
3248 		return 0;
3249 
3250 	ns = rte_timespec_to_ns(ts);
3251 	/* Set the timecounters to a new value. */
3252 	ptp->tc.nsec = ns;
3253 
3254 	return 0;
3255 }
3256 
3257 static int
3258 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3259 {
3260 	struct bnxt *bp = dev->data->dev_private;
3261 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3262 	uint64_t ns, systime_cycles = 0;
3263 	int rc = 0;
3264 
3265 	if (!ptp)
3266 		return 0;
3267 
3268 	if (BNXT_CHIP_P5(bp))
3269 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3270 					     &systime_cycles);
3271 	else
3272 		systime_cycles = bnxt_cc_read(bp);
3273 
3274 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3275 	*ts = rte_ns_to_timespec(ns);
3276 
3277 	return rc;
3278 }
3279 static int
3280 bnxt_timesync_enable(struct rte_eth_dev *dev)
3281 {
3282 	struct bnxt *bp = dev->data->dev_private;
3283 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3284 	uint32_t shift = 0;
3285 	int rc;
3286 
3287 	if (!ptp)
3288 		return 0;
3289 
3290 	ptp->rx_filter = 1;
3291 	ptp->tx_tstamp_en = 1;
3292 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3293 
3294 	rc = bnxt_hwrm_ptp_cfg(bp);
3295 	if (rc)
3296 		return rc;
3297 
3298 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3299 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3300 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3301 
3302 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3303 	ptp->tc.cc_shift = shift;
3304 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3305 
3306 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3307 	ptp->rx_tstamp_tc.cc_shift = shift;
3308 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3309 
3310 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3311 	ptp->tx_tstamp_tc.cc_shift = shift;
3312 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3313 
3314 	if (!BNXT_CHIP_P5(bp))
3315 		bnxt_map_ptp_regs(bp);
3316 
3317 	return 0;
3318 }
3319 
3320 static int
3321 bnxt_timesync_disable(struct rte_eth_dev *dev)
3322 {
3323 	struct bnxt *bp = dev->data->dev_private;
3324 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3325 
3326 	if (!ptp)
3327 		return 0;
3328 
3329 	ptp->rx_filter = 0;
3330 	ptp->tx_tstamp_en = 0;
3331 	ptp->rxctl = 0;
3332 
3333 	bnxt_hwrm_ptp_cfg(bp);
3334 
3335 	if (!BNXT_CHIP_P5(bp))
3336 		bnxt_unmap_ptp_regs(bp);
3337 
3338 	return 0;
3339 }
3340 
3341 static int
3342 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3343 				 struct timespec *timestamp,
3344 				 uint32_t flags __rte_unused)
3345 {
3346 	struct bnxt *bp = dev->data->dev_private;
3347 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3348 	uint64_t rx_tstamp_cycles = 0;
3349 	uint64_t ns;
3350 
3351 	if (!ptp)
3352 		return 0;
3353 
3354 	if (BNXT_CHIP_P5(bp))
3355 		rx_tstamp_cycles = ptp->rx_timestamp;
3356 	else
3357 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3358 
3359 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3360 	*timestamp = rte_ns_to_timespec(ns);
3361 	return  0;
3362 }
3363 
3364 static int
3365 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3366 				 struct timespec *timestamp)
3367 {
3368 	struct bnxt *bp = dev->data->dev_private;
3369 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3370 	uint64_t tx_tstamp_cycles = 0;
3371 	uint64_t ns;
3372 	int rc = 0;
3373 
3374 	if (!ptp)
3375 		return 0;
3376 
3377 	if (BNXT_CHIP_P5(bp))
3378 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3379 					     &tx_tstamp_cycles);
3380 	else
3381 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3382 
3383 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3384 	*timestamp = rte_ns_to_timespec(ns);
3385 
3386 	return rc;
3387 }
3388 
3389 static int
3390 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3391 {
3392 	struct bnxt *bp = dev->data->dev_private;
3393 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3394 
3395 	if (!ptp)
3396 		return 0;
3397 
3398 	ptp->tc.nsec += delta;
3399 
3400 	return 0;
3401 }
3402 
3403 static int
3404 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3405 {
3406 	struct bnxt *bp = dev->data->dev_private;
3407 	int rc;
3408 	uint32_t dir_entries;
3409 	uint32_t entry_length;
3410 
3411 	rc = is_bnxt_in_error(bp);
3412 	if (rc)
3413 		return rc;
3414 
3415 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3416 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3417 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3418 
3419 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3420 	if (rc != 0)
3421 		return rc;
3422 
3423 	return dir_entries * entry_length;
3424 }
3425 
3426 static int
3427 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3428 		struct rte_dev_eeprom_info *in_eeprom)
3429 {
3430 	struct bnxt *bp = dev->data->dev_private;
3431 	uint32_t index;
3432 	uint32_t offset;
3433 	int rc;
3434 
3435 	rc = is_bnxt_in_error(bp);
3436 	if (rc)
3437 		return rc;
3438 
3439 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3440 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3441 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3442 		    in_eeprom->offset, in_eeprom->length);
3443 
3444 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3445 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3446 						in_eeprom->data);
3447 
3448 	index = in_eeprom->offset >> 24;
3449 	offset = in_eeprom->offset & 0xffffff;
3450 
3451 	if (index != 0)
3452 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3453 					   in_eeprom->length, in_eeprom->data);
3454 
3455 	return 0;
3456 }
3457 
3458 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3459 {
3460 	switch (dir_type) {
3461 	case BNX_DIR_TYPE_CHIMP_PATCH:
3462 	case BNX_DIR_TYPE_BOOTCODE:
3463 	case BNX_DIR_TYPE_BOOTCODE_2:
3464 	case BNX_DIR_TYPE_APE_FW:
3465 	case BNX_DIR_TYPE_APE_PATCH:
3466 	case BNX_DIR_TYPE_KONG_FW:
3467 	case BNX_DIR_TYPE_KONG_PATCH:
3468 	case BNX_DIR_TYPE_BONO_FW:
3469 	case BNX_DIR_TYPE_BONO_PATCH:
3470 		/* FALLTHROUGH */
3471 		return true;
3472 	}
3473 
3474 	return false;
3475 }
3476 
3477 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3478 {
3479 	switch (dir_type) {
3480 	case BNX_DIR_TYPE_AVS:
3481 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3482 	case BNX_DIR_TYPE_PCIE:
3483 	case BNX_DIR_TYPE_TSCF_UCODE:
3484 	case BNX_DIR_TYPE_EXT_PHY:
3485 	case BNX_DIR_TYPE_CCM:
3486 	case BNX_DIR_TYPE_ISCSI_BOOT:
3487 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3488 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3489 		/* FALLTHROUGH */
3490 		return true;
3491 	}
3492 
3493 	return false;
3494 }
3495 
3496 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3497 {
3498 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3499 		bnxt_dir_type_is_other_exec_format(dir_type);
3500 }
3501 
3502 static int
3503 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3504 		struct rte_dev_eeprom_info *in_eeprom)
3505 {
3506 	struct bnxt *bp = dev->data->dev_private;
3507 	uint8_t index, dir_op;
3508 	uint16_t type, ext, ordinal, attr;
3509 	int rc;
3510 
3511 	rc = is_bnxt_in_error(bp);
3512 	if (rc)
3513 		return rc;
3514 
3515 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3516 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3517 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3518 		    in_eeprom->offset, in_eeprom->length);
3519 
3520 	if (!BNXT_PF(bp)) {
3521 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3522 		return -EINVAL;
3523 	}
3524 
3525 	type = in_eeprom->magic >> 16;
3526 
3527 	if (type == 0xffff) { /* special value for directory operations */
3528 		index = in_eeprom->magic & 0xff;
3529 		dir_op = in_eeprom->magic >> 8;
3530 		if (index == 0)
3531 			return -EINVAL;
3532 		switch (dir_op) {
3533 		case 0x0e: /* erase */
3534 			if (in_eeprom->offset != ~in_eeprom->magic)
3535 				return -EINVAL;
3536 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3537 		default:
3538 			return -EINVAL;
3539 		}
3540 	}
3541 
3542 	/* Create or re-write an NVM item: */
3543 	if (bnxt_dir_type_is_executable(type) == true)
3544 		return -EOPNOTSUPP;
3545 	ext = in_eeprom->magic & 0xffff;
3546 	ordinal = in_eeprom->offset >> 16;
3547 	attr = in_eeprom->offset & 0xffff;
3548 
3549 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3550 				     in_eeprom->data, in_eeprom->length);
3551 }
3552 
3553 /*
3554  * Initialization
3555  */
3556 
3557 static const struct eth_dev_ops bnxt_dev_ops = {
3558 	.dev_infos_get = bnxt_dev_info_get_op,
3559 	.dev_close = bnxt_dev_close_op,
3560 	.dev_configure = bnxt_dev_configure_op,
3561 	.dev_start = bnxt_dev_start_op,
3562 	.dev_stop = bnxt_dev_stop_op,
3563 	.dev_set_link_up = bnxt_dev_set_link_up_op,
3564 	.dev_set_link_down = bnxt_dev_set_link_down_op,
3565 	.stats_get = bnxt_stats_get_op,
3566 	.stats_reset = bnxt_stats_reset_op,
3567 	.rx_queue_setup = bnxt_rx_queue_setup_op,
3568 	.rx_queue_release = bnxt_rx_queue_release_op,
3569 	.tx_queue_setup = bnxt_tx_queue_setup_op,
3570 	.tx_queue_release = bnxt_tx_queue_release_op,
3571 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
3572 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
3573 	.reta_update = bnxt_reta_update_op,
3574 	.reta_query = bnxt_reta_query_op,
3575 	.rss_hash_update = bnxt_rss_hash_update_op,
3576 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
3577 	.link_update = bnxt_link_update_op,
3578 	.promiscuous_enable = bnxt_promiscuous_enable_op,
3579 	.promiscuous_disable = bnxt_promiscuous_disable_op,
3580 	.allmulticast_enable = bnxt_allmulticast_enable_op,
3581 	.allmulticast_disable = bnxt_allmulticast_disable_op,
3582 	.mac_addr_add = bnxt_mac_addr_add_op,
3583 	.mac_addr_remove = bnxt_mac_addr_remove_op,
3584 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
3585 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
3586 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
3587 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
3588 	.vlan_filter_set = bnxt_vlan_filter_set_op,
3589 	.vlan_offload_set = bnxt_vlan_offload_set_op,
3590 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
3591 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
3592 	.mtu_set = bnxt_mtu_set_op,
3593 	.mac_addr_set = bnxt_set_default_mac_addr_op,
3594 	.xstats_get = bnxt_dev_xstats_get_op,
3595 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
3596 	.xstats_reset = bnxt_dev_xstats_reset_op,
3597 	.fw_version_get = bnxt_fw_version_get,
3598 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
3599 	.rxq_info_get = bnxt_rxq_info_get_op,
3600 	.txq_info_get = bnxt_txq_info_get_op,
3601 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
3602 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
3603 	.dev_led_on = bnxt_dev_led_on_op,
3604 	.dev_led_off = bnxt_dev_led_off_op,
3605 	.rx_queue_start = bnxt_rx_queue_start,
3606 	.rx_queue_stop = bnxt_rx_queue_stop,
3607 	.tx_queue_start = bnxt_tx_queue_start,
3608 	.tx_queue_stop = bnxt_tx_queue_stop,
3609 	.filter_ctrl = bnxt_filter_ctrl_op,
3610 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
3611 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
3612 	.get_eeprom           = bnxt_get_eeprom_op,
3613 	.set_eeprom           = bnxt_set_eeprom_op,
3614 	.timesync_enable      = bnxt_timesync_enable,
3615 	.timesync_disable     = bnxt_timesync_disable,
3616 	.timesync_read_time   = bnxt_timesync_read_time,
3617 	.timesync_write_time   = bnxt_timesync_write_time,
3618 	.timesync_adjust_time = bnxt_timesync_adjust_time,
3619 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
3620 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
3621 };
3622 
3623 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
3624 {
3625 	uint32_t offset;
3626 
3627 	/* Only pre-map the reset GRC registers using window 3 */
3628 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
3629 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
3630 
3631 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
3632 
3633 	return offset;
3634 }
3635 
3636 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
3637 {
3638 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3639 	uint32_t reg_base = 0xffffffff;
3640 	int i;
3641 
3642 	/* Only pre-map the monitoring GRC registers using window 2 */
3643 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
3644 		uint32_t reg = info->status_regs[i];
3645 
3646 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
3647 			continue;
3648 
3649 		if (reg_base == 0xffffffff)
3650 			reg_base = reg & 0xfffff000;
3651 		if ((reg & 0xfffff000) != reg_base)
3652 			return -ERANGE;
3653 
3654 		/* Use mask 0xffc as the Lower 2 bits indicates
3655 		 * address space location
3656 		 */
3657 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
3658 						(reg & 0xffc);
3659 	}
3660 
3661 	if (reg_base == 0xffffffff)
3662 		return 0;
3663 
3664 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
3665 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
3666 
3667 	return 0;
3668 }
3669 
3670 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
3671 {
3672 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3673 	uint32_t delay = info->delay_after_reset[index];
3674 	uint32_t val = info->reset_reg_val[index];
3675 	uint32_t reg = info->reset_reg[index];
3676 	uint32_t type, offset;
3677 
3678 	type = BNXT_FW_STATUS_REG_TYPE(reg);
3679 	offset = BNXT_FW_STATUS_REG_OFF(reg);
3680 
3681 	switch (type) {
3682 	case BNXT_FW_STATUS_REG_TYPE_CFG:
3683 		rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
3684 		break;
3685 	case BNXT_FW_STATUS_REG_TYPE_GRC:
3686 		offset = bnxt_map_reset_regs(bp, offset);
3687 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3688 		break;
3689 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
3690 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3691 		break;
3692 	}
3693 	/* wait on a specific interval of time until core reset is complete */
3694 	if (delay)
3695 		rte_delay_ms(delay);
3696 }
3697 
3698 static void bnxt_dev_cleanup(struct bnxt *bp)
3699 {
3700 	bp->eth_dev->data->dev_link.link_status = 0;
3701 	bp->link_info->link_up = 0;
3702 	if (bp->eth_dev->data->dev_started)
3703 		bnxt_dev_stop(bp->eth_dev);
3704 
3705 	bnxt_uninit_resources(bp, true);
3706 }
3707 
3708 static int bnxt_restore_vlan_filters(struct bnxt *bp)
3709 {
3710 	struct rte_eth_dev *dev = bp->eth_dev;
3711 	struct rte_vlan_filter_conf *vfc;
3712 	int vidx, vbit, rc;
3713 	uint16_t vlan_id;
3714 
3715 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
3716 		vfc = &dev->data->vlan_filter_conf;
3717 		vidx = vlan_id / 64;
3718 		vbit = vlan_id % 64;
3719 
3720 		/* Each bit corresponds to a VLAN id */
3721 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
3722 			rc = bnxt_add_vlan_filter(bp, vlan_id);
3723 			if (rc)
3724 				return rc;
3725 		}
3726 	}
3727 
3728 	return 0;
3729 }
3730 
3731 static int bnxt_restore_mac_filters(struct bnxt *bp)
3732 {
3733 	struct rte_eth_dev *dev = bp->eth_dev;
3734 	struct rte_eth_dev_info dev_info;
3735 	struct rte_ether_addr *addr;
3736 	uint64_t pool_mask;
3737 	uint32_t pool = 0;
3738 	uint16_t i;
3739 	int rc;
3740 
3741 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
3742 		return 0;
3743 
3744 	rc = bnxt_dev_info_get_op(dev, &dev_info);
3745 	if (rc)
3746 		return rc;
3747 
3748 	/* replay MAC address configuration */
3749 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
3750 		addr = &dev->data->mac_addrs[i];
3751 
3752 		/* skip zero address */
3753 		if (rte_is_zero_ether_addr(addr))
3754 			continue;
3755 
3756 		pool = 0;
3757 		pool_mask = dev->data->mac_pool_sel[i];
3758 
3759 		do {
3760 			if (pool_mask & 1ULL) {
3761 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
3762 				if (rc)
3763 					return rc;
3764 			}
3765 			pool_mask >>= 1;
3766 			pool++;
3767 		} while (pool_mask);
3768 	}
3769 
3770 	return 0;
3771 }
3772 
3773 static int bnxt_restore_filters(struct bnxt *bp)
3774 {
3775 	struct rte_eth_dev *dev = bp->eth_dev;
3776 	int ret = 0;
3777 
3778 	if (dev->data->all_multicast) {
3779 		ret = bnxt_allmulticast_enable_op(dev);
3780 		if (ret)
3781 			return ret;
3782 	}
3783 	if (dev->data->promiscuous) {
3784 		ret = bnxt_promiscuous_enable_op(dev);
3785 		if (ret)
3786 			return ret;
3787 	}
3788 
3789 	ret = bnxt_restore_mac_filters(bp);
3790 	if (ret)
3791 		return ret;
3792 
3793 	ret = bnxt_restore_vlan_filters(bp);
3794 	/* TODO restore other filters as well */
3795 	return ret;
3796 }
3797 
3798 static void bnxt_dev_recover(void *arg)
3799 {
3800 	struct bnxt *bp = arg;
3801 	int timeout = bp->fw_reset_max_msecs;
3802 	int rc = 0;
3803 
3804 	pthread_mutex_lock(&bp->err_recovery_lock);
3805 	/* Clear Error flag so that device re-init should happen */
3806 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
3807 
3808 	do {
3809 		rc = bnxt_hwrm_ver_get(bp, SHORT_HWRM_CMD_TIMEOUT);
3810 		if (rc == 0)
3811 			break;
3812 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
3813 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
3814 	} while (rc && timeout);
3815 
3816 	if (rc) {
3817 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
3818 		goto err;
3819 	}
3820 
3821 	rc = bnxt_init_resources(bp, true);
3822 	if (rc) {
3823 		PMD_DRV_LOG(ERR,
3824 			    "Failed to initialize resources after reset\n");
3825 		goto err;
3826 	}
3827 	/* clear reset flag as the device is initialized now */
3828 	bp->flags &= ~BNXT_FLAG_FW_RESET;
3829 
3830 	rc = bnxt_dev_start_op(bp->eth_dev);
3831 	if (rc) {
3832 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
3833 		goto err_start;
3834 	}
3835 
3836 	rc = bnxt_restore_filters(bp);
3837 	if (rc)
3838 		goto err_start;
3839 
3840 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
3841 	pthread_mutex_unlock(&bp->err_recovery_lock);
3842 
3843 	return;
3844 err_start:
3845 	bnxt_dev_stop(bp->eth_dev);
3846 err:
3847 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
3848 	bnxt_uninit_resources(bp, false);
3849 	pthread_mutex_unlock(&bp->err_recovery_lock);
3850 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
3851 }
3852 
3853 void bnxt_dev_reset_and_resume(void *arg)
3854 {
3855 	struct bnxt *bp = arg;
3856 	int rc;
3857 
3858 	bnxt_dev_cleanup(bp);
3859 
3860 	bnxt_wait_for_device_shutdown(bp);
3861 
3862 	rc = rte_eal_alarm_set(US_PER_MS * bp->fw_reset_min_msecs,
3863 			       bnxt_dev_recover, (void *)bp);
3864 	if (rc)
3865 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
3866 }
3867 
3868 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
3869 {
3870 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3871 	uint32_t reg = info->status_regs[index];
3872 	uint32_t type, offset, val = 0;
3873 
3874 	type = BNXT_FW_STATUS_REG_TYPE(reg);
3875 	offset = BNXT_FW_STATUS_REG_OFF(reg);
3876 
3877 	switch (type) {
3878 	case BNXT_FW_STATUS_REG_TYPE_CFG:
3879 		rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
3880 		break;
3881 	case BNXT_FW_STATUS_REG_TYPE_GRC:
3882 		offset = info->mapped_status_regs[index];
3883 		/* FALLTHROUGH */
3884 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
3885 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3886 				       offset));
3887 		break;
3888 	}
3889 
3890 	return val;
3891 }
3892 
3893 static int bnxt_fw_reset_all(struct bnxt *bp)
3894 {
3895 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3896 	uint32_t i;
3897 	int rc = 0;
3898 
3899 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
3900 		/* Reset through master function driver */
3901 		for (i = 0; i < info->reg_array_cnt; i++)
3902 			bnxt_write_fw_reset_reg(bp, i);
3903 		/* Wait for time specified by FW after triggering reset */
3904 		rte_delay_ms(info->master_func_wait_period_after_reset);
3905 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
3906 		/* Reset with the help of Kong processor */
3907 		rc = bnxt_hwrm_fw_reset(bp);
3908 		if (rc)
3909 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
3910 	}
3911 
3912 	return rc;
3913 }
3914 
3915 static void bnxt_fw_reset_cb(void *arg)
3916 {
3917 	struct bnxt *bp = arg;
3918 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3919 	int rc = 0;
3920 
3921 	/* Only Master function can do FW reset */
3922 	if (bnxt_is_master_func(bp) &&
3923 	    bnxt_is_recovery_enabled(bp)) {
3924 		rc = bnxt_fw_reset_all(bp);
3925 		if (rc) {
3926 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
3927 			return;
3928 		}
3929 	}
3930 
3931 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
3932 	 * EXCEPTION_FATAL_ASYNC event to all the functions
3933 	 * (including MASTER FUNC). After receiving this Async, all the active
3934 	 * drivers should treat this case as FW initiated recovery
3935 	 */
3936 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
3937 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
3938 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
3939 
3940 		/* To recover from error */
3941 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
3942 				  (void *)bp);
3943 	}
3944 }
3945 
3946 /* Driver should poll FW heartbeat, reset_counter with the frequency
3947  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
3948  * When the driver detects heartbeat stop or change in reset_counter,
3949  * it has to trigger a reset to recover from the error condition.
3950  * A “master PF” is the function who will have the privilege to
3951  * initiate the chimp reset. The master PF will be elected by the
3952  * firmware and will be notified through async message.
3953  */
3954 static void bnxt_check_fw_health(void *arg)
3955 {
3956 	struct bnxt *bp = arg;
3957 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3958 	uint32_t val = 0, wait_msec;
3959 
3960 	if (!info || !bnxt_is_recovery_enabled(bp) ||
3961 	    is_bnxt_in_error(bp))
3962 		return;
3963 
3964 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
3965 	if (val == info->last_heart_beat)
3966 		goto reset;
3967 
3968 	info->last_heart_beat = val;
3969 
3970 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
3971 	if (val != info->last_reset_counter)
3972 		goto reset;
3973 
3974 	info->last_reset_counter = val;
3975 
3976 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
3977 			  bnxt_check_fw_health, (void *)bp);
3978 
3979 	return;
3980 reset:
3981 	/* Stop DMA to/from device */
3982 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
3983 	bp->flags |= BNXT_FLAG_FW_RESET;
3984 
3985 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
3986 
3987 	if (bnxt_is_master_func(bp))
3988 		wait_msec = info->master_func_wait_period;
3989 	else
3990 		wait_msec = info->normal_func_wait_period;
3991 
3992 	rte_eal_alarm_set(US_PER_MS * wait_msec,
3993 			  bnxt_fw_reset_cb, (void *)bp);
3994 }
3995 
3996 void bnxt_schedule_fw_health_check(struct bnxt *bp)
3997 {
3998 	uint32_t polling_freq;
3999 
4000 	pthread_mutex_lock(&bp->health_check_lock);
4001 
4002 	if (!bnxt_is_recovery_enabled(bp))
4003 		goto done;
4004 
4005 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4006 		goto done;
4007 
4008 	polling_freq = bp->recovery_info->driver_polling_freq;
4009 
4010 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4011 			  bnxt_check_fw_health, (void *)bp);
4012 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4013 
4014 done:
4015 	pthread_mutex_unlock(&bp->health_check_lock);
4016 }
4017 
4018 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4019 {
4020 	if (!bnxt_is_recovery_enabled(bp))
4021 		return;
4022 
4023 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4024 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4025 }
4026 
4027 static bool bnxt_vf_pciid(uint16_t device_id)
4028 {
4029 	switch (device_id) {
4030 	case BROADCOM_DEV_ID_57304_VF:
4031 	case BROADCOM_DEV_ID_57406_VF:
4032 	case BROADCOM_DEV_ID_5731X_VF:
4033 	case BROADCOM_DEV_ID_5741X_VF:
4034 	case BROADCOM_DEV_ID_57414_VF:
4035 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4036 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4037 	case BROADCOM_DEV_ID_58802_VF:
4038 	case BROADCOM_DEV_ID_57500_VF1:
4039 	case BROADCOM_DEV_ID_57500_VF2:
4040 	case BROADCOM_DEV_ID_58818_VF:
4041 		/* FALLTHROUGH */
4042 		return true;
4043 	default:
4044 		return false;
4045 	}
4046 }
4047 
4048 /* Phase 5 device */
4049 static bool bnxt_p5_device(uint16_t device_id)
4050 {
4051 	switch (device_id) {
4052 	case BROADCOM_DEV_ID_57508:
4053 	case BROADCOM_DEV_ID_57504:
4054 	case BROADCOM_DEV_ID_57502:
4055 	case BROADCOM_DEV_ID_57508_MF1:
4056 	case BROADCOM_DEV_ID_57504_MF1:
4057 	case BROADCOM_DEV_ID_57502_MF1:
4058 	case BROADCOM_DEV_ID_57508_MF2:
4059 	case BROADCOM_DEV_ID_57504_MF2:
4060 	case BROADCOM_DEV_ID_57502_MF2:
4061 	case BROADCOM_DEV_ID_57500_VF1:
4062 	case BROADCOM_DEV_ID_57500_VF2:
4063 	case BROADCOM_DEV_ID_58812:
4064 	case BROADCOM_DEV_ID_58814:
4065 	case BROADCOM_DEV_ID_58818:
4066 	case BROADCOM_DEV_ID_58818_VF:
4067 		/* FALLTHROUGH */
4068 		return true;
4069 	default:
4070 		return false;
4071 	}
4072 }
4073 
4074 bool bnxt_stratus_device(struct bnxt *bp)
4075 {
4076 	uint16_t device_id = bp->pdev->id.device_id;
4077 
4078 	switch (device_id) {
4079 	case BROADCOM_DEV_ID_STRATUS_NIC:
4080 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4081 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4082 		/* FALLTHROUGH */
4083 		return true;
4084 	default:
4085 		return false;
4086 	}
4087 }
4088 
4089 static int bnxt_init_board(struct rte_eth_dev *eth_dev)
4090 {
4091 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4092 	struct bnxt *bp = eth_dev->data->dev_private;
4093 
4094 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4095 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4096 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4097 	if (!bp->bar0 || !bp->doorbell_base) {
4098 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4099 		return -ENODEV;
4100 	}
4101 
4102 	bp->eth_dev = eth_dev;
4103 	bp->pdev = pci_dev;
4104 
4105 	return 0;
4106 }
4107 
4108 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4109 				  struct bnxt_ctx_pg_info *ctx_pg,
4110 				  uint32_t mem_size,
4111 				  const char *suffix,
4112 				  uint16_t idx)
4113 {
4114 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4115 	const struct rte_memzone *mz = NULL;
4116 	char mz_name[RTE_MEMZONE_NAMESIZE];
4117 	rte_iova_t mz_phys_addr;
4118 	uint64_t valid_bits = 0;
4119 	uint32_t sz;
4120 	int i;
4121 
4122 	if (!mem_size)
4123 		return 0;
4124 
4125 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4126 			 BNXT_PAGE_SIZE;
4127 	rmem->page_size = BNXT_PAGE_SIZE;
4128 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4129 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4130 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4131 
4132 	valid_bits = PTU_PTE_VALID;
4133 
4134 	if (rmem->nr_pages > 1) {
4135 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4136 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4137 			 suffix, idx, bp->eth_dev->data->port_id);
4138 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4139 		mz = rte_memzone_lookup(mz_name);
4140 		if (!mz) {
4141 			mz = rte_memzone_reserve_aligned(mz_name,
4142 						rmem->nr_pages * 8,
4143 						SOCKET_ID_ANY,
4144 						RTE_MEMZONE_2MB |
4145 						RTE_MEMZONE_SIZE_HINT_ONLY |
4146 						RTE_MEMZONE_IOVA_CONTIG,
4147 						BNXT_PAGE_SIZE);
4148 			if (mz == NULL)
4149 				return -ENOMEM;
4150 		}
4151 
4152 		memset(mz->addr, 0, mz->len);
4153 		mz_phys_addr = mz->iova;
4154 
4155 		rmem->pg_tbl = mz->addr;
4156 		rmem->pg_tbl_map = mz_phys_addr;
4157 		rmem->pg_tbl_mz = mz;
4158 	}
4159 
4160 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4161 		 suffix, idx, bp->eth_dev->data->port_id);
4162 	mz = rte_memzone_lookup(mz_name);
4163 	if (!mz) {
4164 		mz = rte_memzone_reserve_aligned(mz_name,
4165 						 mem_size,
4166 						 SOCKET_ID_ANY,
4167 						 RTE_MEMZONE_1GB |
4168 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4169 						 RTE_MEMZONE_IOVA_CONTIG,
4170 						 BNXT_PAGE_SIZE);
4171 		if (mz == NULL)
4172 			return -ENOMEM;
4173 	}
4174 
4175 	memset(mz->addr, 0, mz->len);
4176 	mz_phys_addr = mz->iova;
4177 
4178 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4179 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4180 		rmem->dma_arr[i] = mz_phys_addr + sz;
4181 
4182 		if (rmem->nr_pages > 1) {
4183 			if (i == rmem->nr_pages - 2 &&
4184 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4185 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4186 			else if (i == rmem->nr_pages - 1 &&
4187 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4188 				valid_bits |= PTU_PTE_LAST;
4189 
4190 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4191 							   valid_bits);
4192 		}
4193 	}
4194 
4195 	rmem->mz = mz;
4196 	if (rmem->vmem_size)
4197 		rmem->vmem = (void **)mz->addr;
4198 	rmem->dma_arr[0] = mz_phys_addr;
4199 	return 0;
4200 }
4201 
4202 static void bnxt_free_ctx_mem(struct bnxt *bp)
4203 {
4204 	int i;
4205 
4206 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4207 		return;
4208 
4209 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4210 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4211 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4212 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4213 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4214 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4215 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4216 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4217 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4218 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4219 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4220 
4221 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4222 		if (bp->ctx->tqm_mem[i])
4223 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4224 	}
4225 
4226 	rte_free(bp->ctx);
4227 	bp->ctx = NULL;
4228 }
4229 
4230 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4231 
4232 #define min_t(type, x, y) ({                    \
4233 	type __min1 = (x);                      \
4234 	type __min2 = (y);                      \
4235 	__min1 < __min2 ? __min1 : __min2; })
4236 
4237 #define max_t(type, x, y) ({                    \
4238 	type __max1 = (x);                      \
4239 	type __max2 = (y);                      \
4240 	__max1 > __max2 ? __max1 : __max2; })
4241 
4242 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4243 
4244 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4245 {
4246 	struct bnxt_ctx_pg_info *ctx_pg;
4247 	struct bnxt_ctx_mem_info *ctx;
4248 	uint32_t mem_size, ena, entries;
4249 	uint32_t entries_sp, min;
4250 	int i, rc;
4251 
4252 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4253 	if (rc) {
4254 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4255 		return rc;
4256 	}
4257 	ctx = bp->ctx;
4258 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4259 		return 0;
4260 
4261 	ctx_pg = &ctx->qp_mem;
4262 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4263 	if (ctx->qp_entry_size) {
4264 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4265 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4266 		if (rc)
4267 			return rc;
4268 	}
4269 
4270 	ctx_pg = &ctx->srq_mem;
4271 	ctx_pg->entries = ctx->srq_max_l2_entries;
4272 	if (ctx->srq_entry_size) {
4273 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4274 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4275 		if (rc)
4276 			return rc;
4277 	}
4278 
4279 	ctx_pg = &ctx->cq_mem;
4280 	ctx_pg->entries = ctx->cq_max_l2_entries;
4281 	if (ctx->cq_entry_size) {
4282 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4283 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4284 		if (rc)
4285 			return rc;
4286 	}
4287 
4288 	ctx_pg = &ctx->vnic_mem;
4289 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4290 		ctx->vnic_max_ring_table_entries;
4291 	if (ctx->vnic_entry_size) {
4292 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4293 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4294 		if (rc)
4295 			return rc;
4296 	}
4297 
4298 	ctx_pg = &ctx->stat_mem;
4299 	ctx_pg->entries = ctx->stat_max_entries;
4300 	if (ctx->stat_entry_size) {
4301 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4302 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4303 		if (rc)
4304 			return rc;
4305 	}
4306 
4307 	min = ctx->tqm_min_entries_per_ring;
4308 
4309 	entries_sp = ctx->qp_max_l2_entries +
4310 		     ctx->vnic_max_vnic_entries +
4311 		     2 * ctx->qp_min_qp1_entries + min;
4312 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4313 
4314 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4315 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4316 	entries = clamp_t(uint32_t, entries, min,
4317 			  ctx->tqm_max_entries_per_ring);
4318 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4319 		ctx_pg = ctx->tqm_mem[i];
4320 		ctx_pg->entries = i ? entries : entries_sp;
4321 		if (ctx->tqm_entry_size) {
4322 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4323 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "tqm_mem", i);
4324 			if (rc)
4325 				return rc;
4326 		}
4327 		ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4328 	}
4329 
4330 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4331 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4332 	if (rc)
4333 		PMD_DRV_LOG(ERR,
4334 			    "Failed to configure context mem: rc = %d\n", rc);
4335 	else
4336 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4337 
4338 	return rc;
4339 }
4340 
4341 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4342 {
4343 	struct rte_pci_device *pci_dev = bp->pdev;
4344 	char mz_name[RTE_MEMZONE_NAMESIZE];
4345 	const struct rte_memzone *mz = NULL;
4346 	uint32_t total_alloc_len;
4347 	rte_iova_t mz_phys_addr;
4348 
4349 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4350 		return 0;
4351 
4352 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4353 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4354 		 pci_dev->addr.bus, pci_dev->addr.devid,
4355 		 pci_dev->addr.function, "rx_port_stats");
4356 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4357 	mz = rte_memzone_lookup(mz_name);
4358 	total_alloc_len =
4359 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4360 				       sizeof(struct rx_port_stats_ext) + 512);
4361 	if (!mz) {
4362 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4363 					 SOCKET_ID_ANY,
4364 					 RTE_MEMZONE_2MB |
4365 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4366 					 RTE_MEMZONE_IOVA_CONTIG);
4367 		if (mz == NULL)
4368 			return -ENOMEM;
4369 	}
4370 	memset(mz->addr, 0, mz->len);
4371 	mz_phys_addr = mz->iova;
4372 
4373 	bp->rx_mem_zone = (const void *)mz;
4374 	bp->hw_rx_port_stats = mz->addr;
4375 	bp->hw_rx_port_stats_map = mz_phys_addr;
4376 
4377 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4378 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4379 		 pci_dev->addr.bus, pci_dev->addr.devid,
4380 		 pci_dev->addr.function, "tx_port_stats");
4381 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4382 	mz = rte_memzone_lookup(mz_name);
4383 	total_alloc_len =
4384 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4385 				       sizeof(struct tx_port_stats_ext) + 512);
4386 	if (!mz) {
4387 		mz = rte_memzone_reserve(mz_name,
4388 					 total_alloc_len,
4389 					 SOCKET_ID_ANY,
4390 					 RTE_MEMZONE_2MB |
4391 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4392 					 RTE_MEMZONE_IOVA_CONTIG);
4393 		if (mz == NULL)
4394 			return -ENOMEM;
4395 	}
4396 	memset(mz->addr, 0, mz->len);
4397 	mz_phys_addr = mz->iova;
4398 
4399 	bp->tx_mem_zone = (const void *)mz;
4400 	bp->hw_tx_port_stats = mz->addr;
4401 	bp->hw_tx_port_stats_map = mz_phys_addr;
4402 	bp->flags |= BNXT_FLAG_PORT_STATS;
4403 
4404 	/* Display extended statistics if FW supports it */
4405 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4406 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4407 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4408 		return 0;
4409 
4410 	bp->hw_rx_port_stats_ext = (void *)
4411 		((uint8_t *)bp->hw_rx_port_stats +
4412 		 sizeof(struct rx_port_stats));
4413 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4414 		sizeof(struct rx_port_stats);
4415 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4416 
4417 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4418 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4419 		bp->hw_tx_port_stats_ext = (void *)
4420 			((uint8_t *)bp->hw_tx_port_stats +
4421 			 sizeof(struct tx_port_stats));
4422 		bp->hw_tx_port_stats_ext_map =
4423 			bp->hw_tx_port_stats_map +
4424 			sizeof(struct tx_port_stats);
4425 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4426 	}
4427 
4428 	return 0;
4429 }
4430 
4431 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4432 {
4433 	struct bnxt *bp = eth_dev->data->dev_private;
4434 	int rc = 0;
4435 
4436 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4437 					       RTE_ETHER_ADDR_LEN *
4438 					       bp->max_l2_ctx,
4439 					       0);
4440 	if (eth_dev->data->mac_addrs == NULL) {
4441 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4442 		return -ENOMEM;
4443 	}
4444 
4445 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4446 		if (BNXT_PF(bp))
4447 			return -EINVAL;
4448 
4449 		/* Generate a random MAC address, if none was assigned by PF */
4450 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4451 		bnxt_eth_hw_addr_random(bp->mac_addr);
4452 		PMD_DRV_LOG(INFO,
4453 			    "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4454 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4455 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4456 
4457 		rc = bnxt_hwrm_set_mac(bp);
4458 		if (rc)
4459 			return rc;
4460 	}
4461 
4462 	/* Copy the permanent MAC from the FUNC_QCAPS response */
4463 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
4464 
4465 	return rc;
4466 }
4467 
4468 static int bnxt_restore_dflt_mac(struct bnxt *bp)
4469 {
4470 	int rc = 0;
4471 
4472 	/* MAC is already configured in FW */
4473 	if (BNXT_HAS_DFLT_MAC_SET(bp))
4474 		return 0;
4475 
4476 	/* Restore the old MAC configured */
4477 	rc = bnxt_hwrm_set_mac(bp);
4478 	if (rc)
4479 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
4480 
4481 	return rc;
4482 }
4483 
4484 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
4485 {
4486 	if (!BNXT_PF(bp))
4487 		return;
4488 
4489 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
4490 
4491 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
4492 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
4493 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
4494 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
4495 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
4496 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
4497 }
4498 
4499 uint16_t
4500 bnxt_get_svif(uint16_t port_id, bool func_svif,
4501 	      enum bnxt_ulp_intf_type type)
4502 {
4503 	struct rte_eth_dev *eth_dev;
4504 	struct bnxt *bp;
4505 
4506 	eth_dev = &rte_eth_devices[port_id];
4507 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4508 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4509 		if (!vfr)
4510 			return 0;
4511 
4512 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4513 			return vfr->svif;
4514 
4515 		eth_dev = vfr->parent_dev;
4516 	}
4517 
4518 	bp = eth_dev->data->dev_private;
4519 
4520 	return func_svif ? bp->func_svif : bp->port_svif;
4521 }
4522 
4523 uint16_t
4524 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
4525 {
4526 	struct rte_eth_dev *eth_dev;
4527 	struct bnxt_vnic_info *vnic;
4528 	struct bnxt *bp;
4529 
4530 	eth_dev = &rte_eth_devices[port];
4531 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4532 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4533 		if (!vfr)
4534 			return 0;
4535 
4536 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4537 			return vfr->dflt_vnic_id;
4538 
4539 		eth_dev = vfr->parent_dev;
4540 	}
4541 
4542 	bp = eth_dev->data->dev_private;
4543 
4544 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
4545 
4546 	return vnic->fw_vnic_id;
4547 }
4548 
4549 uint16_t
4550 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
4551 {
4552 	struct rte_eth_dev *eth_dev;
4553 	struct bnxt *bp;
4554 
4555 	eth_dev = &rte_eth_devices[port];
4556 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4557 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4558 		if (!vfr)
4559 			return 0;
4560 
4561 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4562 			return vfr->fw_fid;
4563 
4564 		eth_dev = vfr->parent_dev;
4565 	}
4566 
4567 	bp = eth_dev->data->dev_private;
4568 
4569 	return bp->fw_fid;
4570 }
4571 
4572 enum bnxt_ulp_intf_type
4573 bnxt_get_interface_type(uint16_t port)
4574 {
4575 	struct rte_eth_dev *eth_dev;
4576 	struct bnxt *bp;
4577 
4578 	eth_dev = &rte_eth_devices[port];
4579 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
4580 		return BNXT_ULP_INTF_TYPE_VF_REP;
4581 
4582 	bp = eth_dev->data->dev_private;
4583 	if (BNXT_PF(bp))
4584 		return BNXT_ULP_INTF_TYPE_PF;
4585 	else if (BNXT_VF_IS_TRUSTED(bp))
4586 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
4587 	else if (BNXT_VF(bp))
4588 		return BNXT_ULP_INTF_TYPE_VF;
4589 
4590 	return BNXT_ULP_INTF_TYPE_INVALID;
4591 }
4592 
4593 uint16_t
4594 bnxt_get_phy_port_id(uint16_t port_id)
4595 {
4596 	struct bnxt_representor *vfr;
4597 	struct rte_eth_dev *eth_dev;
4598 	struct bnxt *bp;
4599 
4600 	eth_dev = &rte_eth_devices[port_id];
4601 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4602 		vfr = eth_dev->data->dev_private;
4603 		if (!vfr)
4604 			return 0;
4605 
4606 		eth_dev = vfr->parent_dev;
4607 	}
4608 
4609 	bp = eth_dev->data->dev_private;
4610 
4611 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
4612 }
4613 
4614 uint16_t
4615 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
4616 {
4617 	struct rte_eth_dev *eth_dev;
4618 	struct bnxt *bp;
4619 
4620 	eth_dev = &rte_eth_devices[port_id];
4621 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4622 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4623 		if (!vfr)
4624 			return 0;
4625 
4626 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4627 			return vfr->fw_fid - 1;
4628 
4629 		eth_dev = vfr->parent_dev;
4630 	}
4631 
4632 	bp = eth_dev->data->dev_private;
4633 
4634 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
4635 }
4636 
4637 uint16_t
4638 bnxt_get_vport(uint16_t port_id)
4639 {
4640 	return (1 << bnxt_get_phy_port_id(port_id));
4641 }
4642 
4643 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
4644 {
4645 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4646 
4647 	if (info) {
4648 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
4649 			memset(info, 0, sizeof(*info));
4650 		return;
4651 	}
4652 
4653 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
4654 		return;
4655 
4656 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4657 			   sizeof(*info), 0);
4658 	if (!info)
4659 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4660 
4661 	bp->recovery_info = info;
4662 }
4663 
4664 static void bnxt_check_fw_status(struct bnxt *bp)
4665 {
4666 	uint32_t fw_status;
4667 
4668 	if (!(bp->recovery_info &&
4669 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
4670 		return;
4671 
4672 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
4673 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
4674 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
4675 			    fw_status);
4676 }
4677 
4678 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
4679 {
4680 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4681 	uint32_t status_loc;
4682 	uint32_t sig_ver;
4683 
4684 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
4685 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4686 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4687 				   BNXT_GRCP_WINDOW_2_BASE +
4688 				   offsetof(struct hcomm_status,
4689 					    sig_ver)));
4690 	/* If the signature is absent, then FW does not support this feature */
4691 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
4692 	    HCOMM_STATUS_SIGNATURE_VAL)
4693 		return 0;
4694 
4695 	if (!info) {
4696 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4697 				   sizeof(*info), 0);
4698 		if (!info)
4699 			return -ENOMEM;
4700 		bp->recovery_info = info;
4701 	} else {
4702 		memset(info, 0, sizeof(*info));
4703 	}
4704 
4705 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4706 				      BNXT_GRCP_WINDOW_2_BASE +
4707 				      offsetof(struct hcomm_status,
4708 					       fw_status_loc)));
4709 
4710 	/* Only pre-map the FW health status GRC register */
4711 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
4712 		return 0;
4713 
4714 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
4715 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
4716 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
4717 
4718 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
4719 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4720 
4721 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
4722 
4723 	return 0;
4724 }
4725 
4726 static int bnxt_init_fw(struct bnxt *bp)
4727 {
4728 	uint16_t mtu;
4729 	int rc = 0;
4730 
4731 	bp->fw_cap = 0;
4732 
4733 	rc = bnxt_map_hcomm_fw_status_reg(bp);
4734 	if (rc)
4735 		return rc;
4736 
4737 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
4738 	if (rc) {
4739 		bnxt_check_fw_status(bp);
4740 		return rc;
4741 	}
4742 
4743 	rc = bnxt_hwrm_func_reset(bp);
4744 	if (rc)
4745 		return -EIO;
4746 
4747 	rc = bnxt_hwrm_vnic_qcaps(bp);
4748 	if (rc)
4749 		return rc;
4750 
4751 	rc = bnxt_hwrm_queue_qportcfg(bp);
4752 	if (rc)
4753 		return rc;
4754 
4755 	/* Get the MAX capabilities for this function.
4756 	 * This function also allocates context memory for TQM rings and
4757 	 * informs the firmware about this allocated backing store memory.
4758 	 */
4759 	rc = bnxt_hwrm_func_qcaps(bp);
4760 	if (rc)
4761 		return rc;
4762 
4763 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
4764 	if (rc)
4765 		return rc;
4766 
4767 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
4768 	if (rc)
4769 		return rc;
4770 
4771 	bnxt_hwrm_port_mac_qcfg(bp);
4772 
4773 	bnxt_hwrm_parent_pf_qcfg(bp);
4774 
4775 	bnxt_hwrm_port_phy_qcaps(bp);
4776 
4777 	bnxt_alloc_error_recovery_info(bp);
4778 	/* Get the adapter error recovery support info */
4779 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
4780 	if (rc)
4781 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4782 
4783 	bnxt_hwrm_port_led_qcaps(bp);
4784 
4785 	return 0;
4786 }
4787 
4788 static int
4789 bnxt_init_locks(struct bnxt *bp)
4790 {
4791 	int err;
4792 
4793 	err = pthread_mutex_init(&bp->flow_lock, NULL);
4794 	if (err) {
4795 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
4796 		return err;
4797 	}
4798 
4799 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
4800 	if (err) {
4801 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
4802 		return err;
4803 	}
4804 
4805 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
4806 	if (err) {
4807 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
4808 		return err;
4809 	}
4810 
4811 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
4812 	if (err)
4813 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
4814 
4815 	return err;
4816 }
4817 
4818 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
4819 {
4820 	int rc = 0;
4821 
4822 	rc = bnxt_init_fw(bp);
4823 	if (rc)
4824 		return rc;
4825 
4826 	if (!reconfig_dev) {
4827 		rc = bnxt_setup_mac_addr(bp->eth_dev);
4828 		if (rc)
4829 			return rc;
4830 	} else {
4831 		rc = bnxt_restore_dflt_mac(bp);
4832 		if (rc)
4833 			return rc;
4834 	}
4835 
4836 	bnxt_config_vf_req_fwd(bp);
4837 
4838 	rc = bnxt_hwrm_func_driver_register(bp);
4839 	if (rc) {
4840 		PMD_DRV_LOG(ERR, "Failed to register driver");
4841 		return -EBUSY;
4842 	}
4843 
4844 	if (BNXT_PF(bp)) {
4845 		if (bp->pdev->max_vfs) {
4846 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
4847 			if (rc) {
4848 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
4849 				return rc;
4850 			}
4851 		} else {
4852 			rc = bnxt_hwrm_allocate_pf_only(bp);
4853 			if (rc) {
4854 				PMD_DRV_LOG(ERR,
4855 					    "Failed to allocate PF resources");
4856 				return rc;
4857 			}
4858 		}
4859 	}
4860 
4861 	rc = bnxt_alloc_mem(bp, reconfig_dev);
4862 	if (rc)
4863 		return rc;
4864 
4865 	rc = bnxt_setup_int(bp);
4866 	if (rc)
4867 		return rc;
4868 
4869 	rc = bnxt_request_int(bp);
4870 	if (rc)
4871 		return rc;
4872 
4873 	rc = bnxt_init_ctx_mem(bp);
4874 	if (rc) {
4875 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
4876 		return rc;
4877 	}
4878 
4879 	return 0;
4880 }
4881 
4882 static int
4883 bnxt_parse_devarg_truflow(__rte_unused const char *key,
4884 			  const char *value, void *opaque_arg)
4885 {
4886 	struct bnxt *bp = opaque_arg;
4887 	unsigned long truflow;
4888 	char *end = NULL;
4889 
4890 	if (!value || !opaque_arg) {
4891 		PMD_DRV_LOG(ERR,
4892 			    "Invalid parameter passed to truflow devargs.\n");
4893 		return -EINVAL;
4894 	}
4895 
4896 	truflow = strtoul(value, &end, 10);
4897 	if (end == NULL || *end != '\0' ||
4898 	    (truflow == ULONG_MAX && errno == ERANGE)) {
4899 		PMD_DRV_LOG(ERR,
4900 			    "Invalid parameter passed to truflow devargs.\n");
4901 		return -EINVAL;
4902 	}
4903 
4904 	if (BNXT_DEVARG_TRUFLOW_INVALID(truflow)) {
4905 		PMD_DRV_LOG(ERR,
4906 			    "Invalid value passed to truflow devargs.\n");
4907 		return -EINVAL;
4908 	}
4909 
4910 	if (truflow) {
4911 		bp->flags |= BNXT_FLAG_TRUFLOW_EN;
4912 		PMD_DRV_LOG(INFO, "Host-based truflow feature enabled.\n");
4913 	} else {
4914 		bp->flags &= ~BNXT_FLAG_TRUFLOW_EN;
4915 		PMD_DRV_LOG(INFO, "Host-based truflow feature disabled.\n");
4916 	}
4917 
4918 	return 0;
4919 }
4920 
4921 static int
4922 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
4923 			     const char *value, void *opaque_arg)
4924 {
4925 	struct bnxt *bp = opaque_arg;
4926 	unsigned long flow_xstat;
4927 	char *end = NULL;
4928 
4929 	if (!value || !opaque_arg) {
4930 		PMD_DRV_LOG(ERR,
4931 			    "Invalid parameter passed to flow_xstat devarg.\n");
4932 		return -EINVAL;
4933 	}
4934 
4935 	flow_xstat = strtoul(value, &end, 10);
4936 	if (end == NULL || *end != '\0' ||
4937 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
4938 		PMD_DRV_LOG(ERR,
4939 			    "Invalid parameter passed to flow_xstat devarg.\n");
4940 		return -EINVAL;
4941 	}
4942 
4943 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
4944 		PMD_DRV_LOG(ERR,
4945 			    "Invalid value passed to flow_xstat devarg.\n");
4946 		return -EINVAL;
4947 	}
4948 
4949 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
4950 	if (BNXT_FLOW_XSTATS_EN(bp))
4951 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
4952 
4953 	return 0;
4954 }
4955 
4956 static int
4957 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
4958 					const char *value, void *opaque_arg)
4959 {
4960 	struct bnxt *bp = opaque_arg;
4961 	unsigned long max_num_kflows;
4962 	char *end = NULL;
4963 
4964 	if (!value || !opaque_arg) {
4965 		PMD_DRV_LOG(ERR,
4966 			"Invalid parameter passed to max_num_kflows devarg.\n");
4967 		return -EINVAL;
4968 	}
4969 
4970 	max_num_kflows = strtoul(value, &end, 10);
4971 	if (end == NULL || *end != '\0' ||
4972 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
4973 		PMD_DRV_LOG(ERR,
4974 			"Invalid parameter passed to max_num_kflows devarg.\n");
4975 		return -EINVAL;
4976 	}
4977 
4978 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
4979 		PMD_DRV_LOG(ERR,
4980 			"Invalid value passed to max_num_kflows devarg.\n");
4981 		return -EINVAL;
4982 	}
4983 
4984 	bp->max_num_kflows = max_num_kflows;
4985 	if (bp->max_num_kflows)
4986 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
4987 				max_num_kflows);
4988 
4989 	return 0;
4990 }
4991 
4992 static int
4993 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
4994 			    const char *value, void *opaque_arg)
4995 {
4996 	struct bnxt_representor *vfr_bp = opaque_arg;
4997 	unsigned long rep_is_pf;
4998 	char *end = NULL;
4999 
5000 	if (!value || !opaque_arg) {
5001 		PMD_DRV_LOG(ERR,
5002 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5003 		return -EINVAL;
5004 	}
5005 
5006 	rep_is_pf = strtoul(value, &end, 10);
5007 	if (end == NULL || *end != '\0' ||
5008 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5009 		PMD_DRV_LOG(ERR,
5010 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5011 		return -EINVAL;
5012 	}
5013 
5014 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5015 		PMD_DRV_LOG(ERR,
5016 			    "Invalid value passed to rep_is_pf devargs.\n");
5017 		return -EINVAL;
5018 	}
5019 
5020 	vfr_bp->flags |= rep_is_pf;
5021 	if (BNXT_REP_PF(vfr_bp))
5022 		PMD_DRV_LOG(INFO, "PF representor\n");
5023 	else
5024 		PMD_DRV_LOG(INFO, "VF representor\n");
5025 
5026 	return 0;
5027 }
5028 
5029 static int
5030 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5031 			       const char *value, void *opaque_arg)
5032 {
5033 	struct bnxt_representor *vfr_bp = opaque_arg;
5034 	unsigned long rep_based_pf;
5035 	char *end = NULL;
5036 
5037 	if (!value || !opaque_arg) {
5038 		PMD_DRV_LOG(ERR,
5039 			    "Invalid parameter passed to rep_based_pf "
5040 			    "devargs.\n");
5041 		return -EINVAL;
5042 	}
5043 
5044 	rep_based_pf = strtoul(value, &end, 10);
5045 	if (end == NULL || *end != '\0' ||
5046 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5047 		PMD_DRV_LOG(ERR,
5048 			    "Invalid parameter passed to rep_based_pf "
5049 			    "devargs.\n");
5050 		return -EINVAL;
5051 	}
5052 
5053 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5054 		PMD_DRV_LOG(ERR,
5055 			    "Invalid value passed to rep_based_pf devargs.\n");
5056 		return -EINVAL;
5057 	}
5058 
5059 	vfr_bp->rep_based_pf = rep_based_pf;
5060 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5061 
5062 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5063 
5064 	return 0;
5065 }
5066 
5067 static int
5068 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5069 			    const char *value, void *opaque_arg)
5070 {
5071 	struct bnxt_representor *vfr_bp = opaque_arg;
5072 	unsigned long rep_q_r2f;
5073 	char *end = NULL;
5074 
5075 	if (!value || !opaque_arg) {
5076 		PMD_DRV_LOG(ERR,
5077 			    "Invalid parameter passed to rep_q_r2f "
5078 			    "devargs.\n");
5079 		return -EINVAL;
5080 	}
5081 
5082 	rep_q_r2f = strtoul(value, &end, 10);
5083 	if (end == NULL || *end != '\0' ||
5084 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5085 		PMD_DRV_LOG(ERR,
5086 			    "Invalid parameter passed to rep_q_r2f "
5087 			    "devargs.\n");
5088 		return -EINVAL;
5089 	}
5090 
5091 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5092 		PMD_DRV_LOG(ERR,
5093 			    "Invalid value passed to rep_q_r2f devargs.\n");
5094 		return -EINVAL;
5095 	}
5096 
5097 	vfr_bp->rep_q_r2f = rep_q_r2f;
5098 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5099 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5100 
5101 	return 0;
5102 }
5103 
5104 static int
5105 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5106 			    const char *value, void *opaque_arg)
5107 {
5108 	struct bnxt_representor *vfr_bp = opaque_arg;
5109 	unsigned long rep_q_f2r;
5110 	char *end = NULL;
5111 
5112 	if (!value || !opaque_arg) {
5113 		PMD_DRV_LOG(ERR,
5114 			    "Invalid parameter passed to rep_q_f2r "
5115 			    "devargs.\n");
5116 		return -EINVAL;
5117 	}
5118 
5119 	rep_q_f2r = strtoul(value, &end, 10);
5120 	if (end == NULL || *end != '\0' ||
5121 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5122 		PMD_DRV_LOG(ERR,
5123 			    "Invalid parameter passed to rep_q_f2r "
5124 			    "devargs.\n");
5125 		return -EINVAL;
5126 	}
5127 
5128 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5129 		PMD_DRV_LOG(ERR,
5130 			    "Invalid value passed to rep_q_f2r devargs.\n");
5131 		return -EINVAL;
5132 	}
5133 
5134 	vfr_bp->rep_q_f2r = rep_q_f2r;
5135 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5136 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5137 
5138 	return 0;
5139 }
5140 
5141 static int
5142 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5143 			     const char *value, void *opaque_arg)
5144 {
5145 	struct bnxt_representor *vfr_bp = opaque_arg;
5146 	unsigned long rep_fc_r2f;
5147 	char *end = NULL;
5148 
5149 	if (!value || !opaque_arg) {
5150 		PMD_DRV_LOG(ERR,
5151 			    "Invalid parameter passed to rep_fc_r2f "
5152 			    "devargs.\n");
5153 		return -EINVAL;
5154 	}
5155 
5156 	rep_fc_r2f = strtoul(value, &end, 10);
5157 	if (end == NULL || *end != '\0' ||
5158 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5159 		PMD_DRV_LOG(ERR,
5160 			    "Invalid parameter passed to rep_fc_r2f "
5161 			    "devargs.\n");
5162 		return -EINVAL;
5163 	}
5164 
5165 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5166 		PMD_DRV_LOG(ERR,
5167 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5168 		return -EINVAL;
5169 	}
5170 
5171 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5172 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5173 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5174 
5175 	return 0;
5176 }
5177 
5178 static int
5179 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5180 			     const char *value, void *opaque_arg)
5181 {
5182 	struct bnxt_representor *vfr_bp = opaque_arg;
5183 	unsigned long rep_fc_f2r;
5184 	char *end = NULL;
5185 
5186 	if (!value || !opaque_arg) {
5187 		PMD_DRV_LOG(ERR,
5188 			    "Invalid parameter passed to rep_fc_f2r "
5189 			    "devargs.\n");
5190 		return -EINVAL;
5191 	}
5192 
5193 	rep_fc_f2r = strtoul(value, &end, 10);
5194 	if (end == NULL || *end != '\0' ||
5195 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5196 		PMD_DRV_LOG(ERR,
5197 			    "Invalid parameter passed to rep_fc_f2r "
5198 			    "devargs.\n");
5199 		return -EINVAL;
5200 	}
5201 
5202 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5203 		PMD_DRV_LOG(ERR,
5204 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5205 		return -EINVAL;
5206 	}
5207 
5208 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5209 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5210 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5211 
5212 	return 0;
5213 }
5214 
5215 static void
5216 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5217 {
5218 	struct rte_kvargs *kvlist;
5219 
5220 	if (devargs == NULL)
5221 		return;
5222 
5223 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5224 	if (kvlist == NULL)
5225 		return;
5226 
5227 	/*
5228 	 * Handler for "truflow" devarg.
5229 	 * Invoked as for ex: "-a 0000:00:0d.0,host-based-truflow=1"
5230 	 */
5231 	rte_kvargs_process(kvlist, BNXT_DEVARG_TRUFLOW,
5232 			   bnxt_parse_devarg_truflow, bp);
5233 
5234 	/*
5235 	 * Handler for "flow_xstat" devarg.
5236 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5237 	 */
5238 	rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5239 			   bnxt_parse_devarg_flow_xstat, bp);
5240 
5241 	/*
5242 	 * Handler for "max_num_kflows" devarg.
5243 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5244 	 */
5245 	rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5246 			   bnxt_parse_devarg_max_num_kflows, bp);
5247 
5248 	rte_kvargs_free(kvlist);
5249 }
5250 
5251 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5252 {
5253 	int rc = 0;
5254 
5255 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5256 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5257 		if (rc)
5258 			PMD_DRV_LOG(ERR,
5259 				    "Failed to alloc switch domain: %d\n", rc);
5260 		else
5261 			PMD_DRV_LOG(INFO,
5262 				    "Switch domain allocated %d\n",
5263 				    bp->switch_domain_id);
5264 	}
5265 
5266 	return rc;
5267 }
5268 
5269 static int
5270 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5271 {
5272 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5273 	static int version_printed;
5274 	struct bnxt *bp;
5275 	int rc;
5276 
5277 	if (version_printed++ == 0)
5278 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5279 
5280 	eth_dev->dev_ops = &bnxt_dev_ops;
5281 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5282 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5283 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5284 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5285 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5286 
5287 	/*
5288 	 * For secondary processes, we don't initialise any further
5289 	 * as primary has already done this work.
5290 	 */
5291 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5292 		return 0;
5293 
5294 	rte_eth_copy_pci_info(eth_dev, pci_dev);
5295 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5296 
5297 	bp = eth_dev->data->dev_private;
5298 
5299 	/* Parse dev arguments passed on when starting the DPDK application. */
5300 	bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5301 
5302 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5303 
5304 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5305 		bp->flags |= BNXT_FLAG_VF;
5306 
5307 	if (bnxt_p5_device(pci_dev->id.device_id))
5308 		bp->flags |= BNXT_FLAG_CHIP_P5;
5309 
5310 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5311 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5312 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5313 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5314 		bp->flags |= BNXT_FLAG_STINGRAY;
5315 
5316 	if (BNXT_TRUFLOW_EN(bp)) {
5317 		/* extra mbuf field is required to store CFA code from mark */
5318 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5319 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5320 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5321 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5322 		};
5323 		bnxt_cfa_code_dynfield_offset =
5324 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5325 		if (bnxt_cfa_code_dynfield_offset < 0) {
5326 			PMD_DRV_LOG(ERR,
5327 			    "Failed to register mbuf field for TruFlow mark\n");
5328 			return -rte_errno;
5329 		}
5330 	}
5331 
5332 	rc = bnxt_init_board(eth_dev);
5333 	if (rc) {
5334 		PMD_DRV_LOG(ERR,
5335 			    "Failed to initialize board rc: %x\n", rc);
5336 		return rc;
5337 	}
5338 
5339 	rc = bnxt_alloc_pf_info(bp);
5340 	if (rc)
5341 		goto error_free;
5342 
5343 	rc = bnxt_alloc_link_info(bp);
5344 	if (rc)
5345 		goto error_free;
5346 
5347 	rc = bnxt_alloc_parent_info(bp);
5348 	if (rc)
5349 		goto error_free;
5350 
5351 	rc = bnxt_alloc_hwrm_resources(bp);
5352 	if (rc) {
5353 		PMD_DRV_LOG(ERR,
5354 			    "Failed to allocate hwrm resource rc: %x\n", rc);
5355 		goto error_free;
5356 	}
5357 	rc = bnxt_alloc_leds_info(bp);
5358 	if (rc)
5359 		goto error_free;
5360 
5361 	rc = bnxt_alloc_cos_queues(bp);
5362 	if (rc)
5363 		goto error_free;
5364 
5365 	rc = bnxt_init_locks(bp);
5366 	if (rc)
5367 		goto error_free;
5368 
5369 	rc = bnxt_init_resources(bp, false);
5370 	if (rc)
5371 		goto error_free;
5372 
5373 	rc = bnxt_alloc_stats_mem(bp);
5374 	if (rc)
5375 		goto error_free;
5376 
5377 	bnxt_alloc_switch_domain(bp);
5378 
5379 	PMD_DRV_LOG(INFO,
5380 		    DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
5381 		    pci_dev->mem_resource[0].phys_addr,
5382 		    pci_dev->mem_resource[0].addr);
5383 
5384 	return 0;
5385 
5386 error_free:
5387 	bnxt_dev_uninit(eth_dev);
5388 	return rc;
5389 }
5390 
5391 
5392 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5393 {
5394 	if (!ctx)
5395 		return;
5396 
5397 	if (ctx->va)
5398 		rte_free(ctx->va);
5399 
5400 	ctx->va = NULL;
5401 	ctx->dma = RTE_BAD_IOVA;
5402 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5403 }
5404 
5405 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5406 {
5407 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5408 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5409 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
5410 				  bp->flow_stat->max_fc,
5411 				  false);
5412 
5413 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5414 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5415 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
5416 				  bp->flow_stat->max_fc,
5417 				  false);
5418 
5419 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5420 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5421 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5422 
5423 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5424 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5425 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5426 
5427 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5428 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5429 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5430 
5431 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5432 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5433 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5434 }
5435 
5436 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5437 {
5438 	bnxt_unregister_fc_ctx_mem(bp);
5439 
5440 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5441 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5442 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5443 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5444 }
5445 
5446 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5447 {
5448 	if (BNXT_FLOW_XSTATS_EN(bp))
5449 		bnxt_uninit_fc_ctx_mem(bp);
5450 }
5451 
5452 static void
5453 bnxt_free_error_recovery_info(struct bnxt *bp)
5454 {
5455 	rte_free(bp->recovery_info);
5456 	bp->recovery_info = NULL;
5457 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5458 }
5459 
5460 static int
5461 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5462 {
5463 	int rc;
5464 
5465 	bnxt_free_int(bp);
5466 	bnxt_free_mem(bp, reconfig_dev);
5467 
5468 	bnxt_hwrm_func_buf_unrgtr(bp);
5469 	rte_free(bp->pf->vf_req_buf);
5470 
5471 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
5472 	bp->flags &= ~BNXT_FLAG_REGISTERED;
5473 	bnxt_free_ctx_mem(bp);
5474 	if (!reconfig_dev) {
5475 		bnxt_free_hwrm_resources(bp);
5476 		bnxt_free_error_recovery_info(bp);
5477 	}
5478 
5479 	bnxt_uninit_ctx_mem(bp);
5480 
5481 	bnxt_free_flow_stats_info(bp);
5482 	bnxt_free_rep_info(bp);
5483 	rte_free(bp->ptp_cfg);
5484 	bp->ptp_cfg = NULL;
5485 	return rc;
5486 }
5487 
5488 static int
5489 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5490 {
5491 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5492 		return -EPERM;
5493 
5494 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5495 
5496 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5497 		bnxt_dev_close_op(eth_dev);
5498 
5499 	return 0;
5500 }
5501 
5502 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
5503 {
5504 	struct bnxt *bp = eth_dev->data->dev_private;
5505 	struct rte_eth_dev *vf_rep_eth_dev;
5506 	int ret = 0, i;
5507 
5508 	if (!bp)
5509 		return -EINVAL;
5510 
5511 	for (i = 0; i < bp->num_reps; i++) {
5512 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
5513 		if (!vf_rep_eth_dev)
5514 			continue;
5515 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
5516 			    vf_rep_eth_dev->data->port_id);
5517 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
5518 	}
5519 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
5520 		    eth_dev->data->port_id);
5521 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
5522 
5523 	return ret;
5524 }
5525 
5526 static void bnxt_free_rep_info(struct bnxt *bp)
5527 {
5528 	rte_free(bp->rep_info);
5529 	bp->rep_info = NULL;
5530 	rte_free(bp->cfa_code_map);
5531 	bp->cfa_code_map = NULL;
5532 }
5533 
5534 static int bnxt_init_rep_info(struct bnxt *bp)
5535 {
5536 	int i = 0, rc;
5537 
5538 	if (bp->rep_info)
5539 		return 0;
5540 
5541 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
5542 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
5543 				   0);
5544 	if (!bp->rep_info) {
5545 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
5546 		return -ENOMEM;
5547 	}
5548 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
5549 				       sizeof(*bp->cfa_code_map) *
5550 				       BNXT_MAX_CFA_CODE, 0);
5551 	if (!bp->cfa_code_map) {
5552 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
5553 		bnxt_free_rep_info(bp);
5554 		return -ENOMEM;
5555 	}
5556 
5557 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
5558 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
5559 
5560 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
5561 	if (rc) {
5562 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
5563 		bnxt_free_rep_info(bp);
5564 		return rc;
5565 	}
5566 
5567 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
5568 	if (rc) {
5569 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
5570 		bnxt_free_rep_info(bp);
5571 		return rc;
5572 	}
5573 
5574 	return rc;
5575 }
5576 
5577 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
5578 			       struct rte_eth_devargs *eth_da,
5579 			       struct rte_eth_dev *backing_eth_dev,
5580 			       const char *dev_args)
5581 {
5582 	struct rte_eth_dev *vf_rep_eth_dev;
5583 	char name[RTE_ETH_NAME_MAX_LEN];
5584 	struct bnxt *backing_bp;
5585 	uint16_t num_rep;
5586 	int i, ret = 0;
5587 	struct rte_kvargs *kvlist = NULL;
5588 
5589 	num_rep = eth_da->nb_representor_ports;
5590 	if (num_rep > BNXT_MAX_VF_REPS) {
5591 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
5592 			    num_rep, BNXT_MAX_VF_REPS);
5593 		return -EINVAL;
5594 	}
5595 
5596 	if (num_rep >= RTE_MAX_ETHPORTS) {
5597 		PMD_DRV_LOG(ERR,
5598 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
5599 			    num_rep, RTE_MAX_ETHPORTS);
5600 		return -EINVAL;
5601 	}
5602 
5603 	backing_bp = backing_eth_dev->data->dev_private;
5604 
5605 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
5606 		PMD_DRV_LOG(ERR,
5607 			    "Not a PF or trusted VF. No Representor support\n");
5608 		/* Returning an error is not an option.
5609 		 * Applications are not handling this correctly
5610 		 */
5611 		return 0;
5612 	}
5613 
5614 	if (bnxt_init_rep_info(backing_bp))
5615 		return 0;
5616 
5617 	for (i = 0; i < num_rep; i++) {
5618 		struct bnxt_representor representor = {
5619 			.vf_id = eth_da->representor_ports[i],
5620 			.switch_domain_id = backing_bp->switch_domain_id,
5621 			.parent_dev = backing_eth_dev
5622 		};
5623 
5624 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
5625 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
5626 				    representor.vf_id, BNXT_MAX_VF_REPS);
5627 			continue;
5628 		}
5629 
5630 		/* representor port net_bdf_port */
5631 		snprintf(name, sizeof(name), "net_%s_representor_%d",
5632 			 pci_dev->device.name, eth_da->representor_ports[i]);
5633 
5634 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
5635 		if (kvlist) {
5636 			/*
5637 			 * Handler for "rep_is_pf" devarg.
5638 			 * Invoked as for ex: "-a 000:00:0d.0,
5639 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5640 			 */
5641 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
5642 						 bnxt_parse_devarg_rep_is_pf,
5643 						 (void *)&representor);
5644 			if (ret) {
5645 				ret = -EINVAL;
5646 				goto err;
5647 			}
5648 			/*
5649 			 * Handler for "rep_based_pf" devarg.
5650 			 * Invoked as for ex: "-a 000:00:0d.0,
5651 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5652 			 */
5653 			ret = rte_kvargs_process(kvlist,
5654 						 BNXT_DEVARG_REP_BASED_PF,
5655 						 bnxt_parse_devarg_rep_based_pf,
5656 						 (void *)&representor);
5657 			if (ret) {
5658 				ret = -EINVAL;
5659 				goto err;
5660 			}
5661 			/*
5662 			 * Handler for "rep_based_pf" devarg.
5663 			 * Invoked as for ex: "-a 000:00:0d.0,
5664 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5665 			 */
5666 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
5667 						 bnxt_parse_devarg_rep_q_r2f,
5668 						 (void *)&representor);
5669 			if (ret) {
5670 				ret = -EINVAL;
5671 				goto err;
5672 			}
5673 			/*
5674 			 * Handler for "rep_based_pf" devarg.
5675 			 * Invoked as for ex: "-a 000:00:0d.0,
5676 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5677 			 */
5678 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
5679 						 bnxt_parse_devarg_rep_q_f2r,
5680 						 (void *)&representor);
5681 			if (ret) {
5682 				ret = -EINVAL;
5683 				goto err;
5684 			}
5685 			/*
5686 			 * Handler for "rep_based_pf" devarg.
5687 			 * Invoked as for ex: "-a 000:00:0d.0,
5688 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5689 			 */
5690 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
5691 						 bnxt_parse_devarg_rep_fc_r2f,
5692 						 (void *)&representor);
5693 			if (ret) {
5694 				ret = -EINVAL;
5695 				goto err;
5696 			}
5697 			/*
5698 			 * Handler for "rep_based_pf" devarg.
5699 			 * Invoked as for ex: "-a 000:00:0d.0,
5700 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5701 			 */
5702 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
5703 						 bnxt_parse_devarg_rep_fc_f2r,
5704 						 (void *)&representor);
5705 			if (ret) {
5706 				ret = -EINVAL;
5707 				goto err;
5708 			}
5709 		}
5710 
5711 		ret = rte_eth_dev_create(&pci_dev->device, name,
5712 					 sizeof(struct bnxt_representor),
5713 					 NULL, NULL,
5714 					 bnxt_representor_init,
5715 					 &representor);
5716 		if (ret) {
5717 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
5718 				    "representor %s.", name);
5719 			goto err;
5720 		}
5721 
5722 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
5723 		if (!vf_rep_eth_dev) {
5724 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
5725 				    " for VF-Rep: %s.", name);
5726 			ret = -ENODEV;
5727 			goto err;
5728 		}
5729 
5730 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
5731 			    backing_eth_dev->data->port_id);
5732 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
5733 							 vf_rep_eth_dev;
5734 		backing_bp->num_reps++;
5735 
5736 	}
5737 
5738 	rte_kvargs_free(kvlist);
5739 	return 0;
5740 
5741 err:
5742 	/* If num_rep > 1, then rollback already created
5743 	 * ports, since we'll be failing the probe anyway
5744 	 */
5745 	if (num_rep > 1)
5746 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
5747 	rte_errno = -ret;
5748 	rte_kvargs_free(kvlist);
5749 
5750 	return ret;
5751 }
5752 
5753 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
5754 			  struct rte_pci_device *pci_dev)
5755 {
5756 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
5757 	struct rte_eth_dev *backing_eth_dev;
5758 	uint16_t num_rep;
5759 	int ret = 0;
5760 
5761 	if (pci_dev->device.devargs) {
5762 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
5763 					    &eth_da);
5764 		if (ret)
5765 			return ret;
5766 	}
5767 
5768 	num_rep = eth_da.nb_representor_ports;
5769 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
5770 		    num_rep);
5771 
5772 	/* We could come here after first level of probe is already invoked
5773 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
5774 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
5775 	 */
5776 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
5777 	if (backing_eth_dev == NULL) {
5778 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
5779 					 sizeof(struct bnxt),
5780 					 eth_dev_pci_specific_init, pci_dev,
5781 					 bnxt_dev_init, NULL);
5782 
5783 		if (ret || !num_rep)
5784 			return ret;
5785 
5786 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
5787 	}
5788 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
5789 		    backing_eth_dev->data->port_id);
5790 
5791 	if (!num_rep)
5792 		return ret;
5793 
5794 	/* probe representor ports now */
5795 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
5796 				  pci_dev->device.devargs->args);
5797 
5798 	return ret;
5799 }
5800 
5801 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
5802 {
5803 	struct rte_eth_dev *eth_dev;
5804 
5805 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
5806 	if (!eth_dev)
5807 		return 0; /* Invoked typically only by OVS-DPDK, by the
5808 			   * time it comes here the eth_dev is already
5809 			   * deleted by rte_eth_dev_close(), so returning
5810 			   * +ve value will at least help in proper cleanup
5811 			   */
5812 
5813 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
5814 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
5815 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
5816 			return rte_eth_dev_destroy(eth_dev,
5817 						   bnxt_representor_uninit);
5818 		else
5819 			return rte_eth_dev_destroy(eth_dev,
5820 						   bnxt_dev_uninit);
5821 	} else {
5822 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
5823 	}
5824 }
5825 
5826 static struct rte_pci_driver bnxt_rte_pmd = {
5827 	.id_table = bnxt_pci_id_map,
5828 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
5829 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
5830 						  * and OVS-DPDK
5831 						  */
5832 	.probe = bnxt_pci_probe,
5833 	.remove = bnxt_pci_remove,
5834 };
5835 
5836 static bool
5837 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
5838 {
5839 	if (strcmp(dev->device->driver->name, drv->driver.name))
5840 		return false;
5841 
5842 	return true;
5843 }
5844 
5845 bool is_bnxt_supported(struct rte_eth_dev *dev)
5846 {
5847 	return is_device_supported(dev, &bnxt_rte_pmd);
5848 }
5849 
5850 RTE_LOG_REGISTER(bnxt_logtype_driver, pmd.net.bnxt.driver, NOTICE);
5851 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
5852 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
5853 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
5854