xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision 68a03efeed657e6e05f281479b33b51102797e15)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2014-2021 Broadcom
3  * All rights reserved.
4  */
5 
6 #include <inttypes.h>
7 #include <stdbool.h>
8 
9 #include <rte_dev.h>
10 #include <ethdev_driver.h>
11 #include <ethdev_pci.h>
12 #include <rte_malloc.h>
13 #include <rte_cycles.h>
14 #include <rte_alarm.h>
15 #include <rte_kvargs.h>
16 #include <rte_vect.h>
17 
18 #include "bnxt.h"
19 #include "bnxt_filter.h"
20 #include "bnxt_hwrm.h"
21 #include "bnxt_irq.h"
22 #include "bnxt_reps.h"
23 #include "bnxt_ring.h"
24 #include "bnxt_rxq.h"
25 #include "bnxt_rxr.h"
26 #include "bnxt_stats.h"
27 #include "bnxt_txq.h"
28 #include "bnxt_txr.h"
29 #include "bnxt_vnic.h"
30 #include "hsi_struct_def_dpdk.h"
31 #include "bnxt_nvm_defs.h"
32 #include "bnxt_tf_common.h"
33 #include "ulp_flow_db.h"
34 #include "rte_pmd_bnxt.h"
35 
36 #define DRV_MODULE_NAME		"bnxt"
37 static const char bnxt_version[] =
38 	"Broadcom NetXtreme driver " DRV_MODULE_NAME;
39 
40 /*
41  * The set of PCI devices this driver supports
42  */
43 static const struct rte_pci_id bnxt_pci_id_map[] = {
44 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
45 			 BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
46 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
47 			 BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
48 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
49 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
50 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
51 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
52 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
53 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
54 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
55 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
56 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
57 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
58 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
59 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
60 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
61 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
62 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
63 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
64 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
65 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
66 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
67 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
68 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
69 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
70 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
71 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
72 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
73 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
74 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
75 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
76 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
77 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
78 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
79 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
80 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
81 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
82 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
83 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58812) },
84 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58814) },
85 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818) },
86 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818_VF) },
87 	{ .vendor_id = 0, /* sentinel */ },
88 };
89 
90 #define BNXT_DEVARG_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_start_nic(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 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1147 		return 1;
1148 
1149 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1150 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1151 
1152 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1153 				      RTE_PKTMBUF_HEADROOM);
1154 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1155 			return 1;
1156 	}
1157 	return 0;
1158 }
1159 
1160 static eth_rx_burst_t
1161 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1162 {
1163 	struct bnxt *bp = eth_dev->data->dev_private;
1164 
1165 	/* Disable vector mode RX for Stingray2 for now */
1166 	if (BNXT_CHIP_SR2(bp)) {
1167 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1168 		return bnxt_recv_pkts;
1169 	}
1170 
1171 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1172 #ifndef RTE_LIBRTE_IEEE1588
1173 	/*
1174 	 * Vector mode receive can be enabled only if scatter rx is not
1175 	 * in use and rx offloads are limited to VLAN stripping and
1176 	 * CRC stripping.
1177 	 */
1178 	if (!eth_dev->data->scattered_rx &&
1179 	    !(eth_dev->data->dev_conf.rxmode.offloads &
1180 	      ~(DEV_RX_OFFLOAD_VLAN_STRIP |
1181 		DEV_RX_OFFLOAD_KEEP_CRC |
1182 		DEV_RX_OFFLOAD_JUMBO_FRAME |
1183 		DEV_RX_OFFLOAD_IPV4_CKSUM |
1184 		DEV_RX_OFFLOAD_UDP_CKSUM |
1185 		DEV_RX_OFFLOAD_TCP_CKSUM |
1186 		DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1187 		DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1188 		DEV_RX_OFFLOAD_RSS_HASH |
1189 		DEV_RX_OFFLOAD_VLAN_FILTER)) &&
1190 	    !BNXT_TRUFLOW_EN(bp) && BNXT_NUM_ASYNC_CPR(bp) &&
1191 	    rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1192 		PMD_DRV_LOG(INFO, "Using vector mode receive for port %d\n",
1193 			    eth_dev->data->port_id);
1194 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1195 		return bnxt_recv_pkts_vec;
1196 	}
1197 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1198 		    eth_dev->data->port_id);
1199 	PMD_DRV_LOG(INFO,
1200 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1201 		    eth_dev->data->port_id,
1202 		    eth_dev->data->scattered_rx,
1203 		    eth_dev->data->dev_conf.rxmode.offloads);
1204 #endif
1205 #endif
1206 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1207 	return bnxt_recv_pkts;
1208 }
1209 
1210 static eth_tx_burst_t
1211 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1212 {
1213 	struct bnxt *bp = eth_dev->data->dev_private;
1214 
1215 	/* Disable vector mode TX for Stingray2 for now */
1216 	if (BNXT_CHIP_SR2(bp))
1217 		return bnxt_xmit_pkts;
1218 
1219 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1220 #ifndef RTE_LIBRTE_IEEE1588
1221 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1222 
1223 	/*
1224 	 * Vector mode transmit can be enabled only if not using scatter rx
1225 	 * or tx offloads.
1226 	 */
1227 	if (!eth_dev->data->scattered_rx &&
1228 	    !(offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) &&
1229 	    !BNXT_TRUFLOW_EN(bp) &&
1230 	    rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1231 		PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
1232 			    eth_dev->data->port_id);
1233 		return bnxt_xmit_pkts_vec;
1234 	}
1235 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1236 		    eth_dev->data->port_id);
1237 	PMD_DRV_LOG(INFO,
1238 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1239 		    eth_dev->data->port_id,
1240 		    eth_dev->data->scattered_rx,
1241 		    offloads);
1242 #endif
1243 #endif
1244 	return bnxt_xmit_pkts;
1245 }
1246 
1247 static int bnxt_handle_if_change_status(struct bnxt *bp)
1248 {
1249 	int rc;
1250 
1251 	/* Since fw has undergone a reset and lost all contexts,
1252 	 * set fatal flag to not issue hwrm during cleanup
1253 	 */
1254 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1255 	bnxt_uninit_resources(bp, true);
1256 
1257 	/* clear fatal flag so that re-init happens */
1258 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1259 	rc = bnxt_init_resources(bp, true);
1260 
1261 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1262 
1263 	return rc;
1264 }
1265 
1266 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1267 {
1268 	struct bnxt *bp = eth_dev->data->dev_private;
1269 	int rc = 0;
1270 
1271 	if (!BNXT_SINGLE_PF(bp))
1272 		return -ENOTSUP;
1273 
1274 	if (!bp->link_info->link_up)
1275 		rc = bnxt_set_hwrm_link_config(bp, true);
1276 	if (!rc)
1277 		eth_dev->data->dev_link.link_status = 1;
1278 
1279 	bnxt_print_link_info(eth_dev);
1280 	return rc;
1281 }
1282 
1283 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1284 {
1285 	struct bnxt *bp = eth_dev->data->dev_private;
1286 
1287 	if (!BNXT_SINGLE_PF(bp))
1288 		return -ENOTSUP;
1289 
1290 	eth_dev->data->dev_link.link_status = 0;
1291 	bnxt_set_hwrm_link_config(bp, false);
1292 	bp->link_info->link_up = 0;
1293 
1294 	return 0;
1295 }
1296 
1297 static void bnxt_free_switch_domain(struct bnxt *bp)
1298 {
1299 	int rc = 0;
1300 
1301 	if (bp->switch_domain_id) {
1302 		rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1303 		if (rc)
1304 			PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1305 				    bp->switch_domain_id, rc);
1306 	}
1307 }
1308 
1309 static void bnxt_ptp_get_current_time(void *arg)
1310 {
1311 	struct bnxt *bp = arg;
1312 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1313 	int rc;
1314 
1315 	rc = is_bnxt_in_error(bp);
1316 	if (rc)
1317 		return;
1318 
1319 	if (!ptp)
1320 		return;
1321 
1322 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1323 				&ptp->current_time);
1324 
1325 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1326 	if (rc != 0) {
1327 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1328 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1329 	}
1330 }
1331 
1332 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1333 {
1334 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1335 	int rc;
1336 
1337 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1338 		return 0;
1339 
1340 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1341 				&ptp->current_time);
1342 
1343 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1344 	return rc;
1345 }
1346 
1347 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1348 {
1349 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1350 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1351 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1352 	}
1353 }
1354 
1355 static void bnxt_ptp_stop(struct bnxt *bp)
1356 {
1357 	bnxt_cancel_ptp_alarm(bp);
1358 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1359 }
1360 
1361 static int bnxt_ptp_start(struct bnxt *bp)
1362 {
1363 	int rc;
1364 
1365 	rc = bnxt_schedule_ptp_alarm(bp);
1366 	if (rc != 0) {
1367 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1368 	} else {
1369 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1370 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1371 	}
1372 
1373 	return rc;
1374 }
1375 
1376 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1377 {
1378 	struct bnxt *bp = eth_dev->data->dev_private;
1379 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1380 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1381 	struct rte_eth_link link;
1382 	int ret;
1383 
1384 	eth_dev->data->dev_started = 0;
1385 	eth_dev->data->scattered_rx = 0;
1386 
1387 	/* Prevent crashes when queues are still in use */
1388 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1389 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1390 
1391 	bnxt_disable_int(bp);
1392 
1393 	/* disable uio/vfio intr/eventfd mapping */
1394 	rte_intr_disable(intr_handle);
1395 
1396 	/* Stop the child representors for this device */
1397 	ret = bnxt_rep_stop_all(bp);
1398 	if (ret != 0)
1399 		return ret;
1400 
1401 	/* delete the bnxt ULP port details */
1402 	bnxt_ulp_port_deinit(bp);
1403 
1404 	bnxt_cancel_fw_health_check(bp);
1405 
1406 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1407 		bnxt_cancel_ptp_alarm(bp);
1408 
1409 	/* Do not bring link down during reset recovery */
1410 	if (!is_bnxt_in_error(bp)) {
1411 		bnxt_dev_set_link_down_op(eth_dev);
1412 		/* Wait for link to be reset */
1413 		if (BNXT_SINGLE_PF(bp))
1414 			rte_delay_ms(500);
1415 		/* clear the recorded link status */
1416 		memset(&link, 0, sizeof(link));
1417 		rte_eth_linkstatus_set(eth_dev, &link);
1418 	}
1419 
1420 	/* Clean queue intr-vector mapping */
1421 	rte_intr_efd_disable(intr_handle);
1422 	if (intr_handle->intr_vec != NULL) {
1423 		rte_free(intr_handle->intr_vec);
1424 		intr_handle->intr_vec = NULL;
1425 	}
1426 
1427 	bnxt_hwrm_port_clr_stats(bp);
1428 	bnxt_free_tx_mbufs(bp);
1429 	bnxt_free_rx_mbufs(bp);
1430 	/* Process any remaining notifications in default completion queue */
1431 	bnxt_int_handler(eth_dev);
1432 	bnxt_shutdown_nic(bp);
1433 	bnxt_hwrm_if_change(bp, false);
1434 
1435 	rte_free(bp->mark_table);
1436 	bp->mark_table = NULL;
1437 
1438 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1439 	bp->rx_cosq_cnt = 0;
1440 	/* All filters are deleted on a port stop. */
1441 	if (BNXT_FLOW_XSTATS_EN(bp))
1442 		bp->flow_stat->flow_count = 0;
1443 
1444 	return 0;
1445 }
1446 
1447 /* Unload the driver, release resources */
1448 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1449 {
1450 	struct bnxt *bp = eth_dev->data->dev_private;
1451 
1452 	pthread_mutex_lock(&bp->err_recovery_lock);
1453 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1454 		PMD_DRV_LOG(ERR,
1455 			    "Adapter recovering from error..Please retry\n");
1456 		pthread_mutex_unlock(&bp->err_recovery_lock);
1457 		return -EAGAIN;
1458 	}
1459 	pthread_mutex_unlock(&bp->err_recovery_lock);
1460 
1461 	return bnxt_dev_stop(eth_dev);
1462 }
1463 
1464 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1465 {
1466 	struct bnxt *bp = eth_dev->data->dev_private;
1467 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1468 	int vlan_mask = 0;
1469 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1470 
1471 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1472 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1473 		return -EINVAL;
1474 	}
1475 
1476 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1477 		PMD_DRV_LOG(ERR,
1478 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1479 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1480 
1481 	do {
1482 		rc = bnxt_hwrm_if_change(bp, true);
1483 		if (rc == 0 || rc != -EAGAIN)
1484 			break;
1485 
1486 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1487 	} while (retry_cnt--);
1488 
1489 	if (rc)
1490 		return rc;
1491 
1492 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1493 		rc = bnxt_handle_if_change_status(bp);
1494 		if (rc)
1495 			return rc;
1496 	}
1497 
1498 	bnxt_enable_int(bp);
1499 
1500 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1501 
1502 	rc = bnxt_start_nic(bp);
1503 	if (rc)
1504 		goto error;
1505 
1506 	eth_dev->data->dev_started = 1;
1507 
1508 	bnxt_link_update_op(eth_dev, 1);
1509 
1510 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1511 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1512 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1513 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1514 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1515 	if (rc)
1516 		goto error;
1517 
1518 	/* Initialize bnxt ULP port details */
1519 	rc = bnxt_ulp_port_init(bp);
1520 	if (rc)
1521 		goto error;
1522 
1523 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1524 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1525 
1526 	bnxt_schedule_fw_health_check(bp);
1527 
1528 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1529 		bnxt_schedule_ptp_alarm(bp);
1530 
1531 	return 0;
1532 
1533 error:
1534 	bnxt_dev_stop(eth_dev);
1535 	return rc;
1536 }
1537 
1538 static void
1539 bnxt_uninit_locks(struct bnxt *bp)
1540 {
1541 	pthread_mutex_destroy(&bp->flow_lock);
1542 	pthread_mutex_destroy(&bp->def_cp_lock);
1543 	pthread_mutex_destroy(&bp->health_check_lock);
1544 	pthread_mutex_destroy(&bp->err_recovery_lock);
1545 	if (bp->rep_info) {
1546 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1547 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1548 	}
1549 }
1550 
1551 static void bnxt_drv_uninit(struct bnxt *bp)
1552 {
1553 	bnxt_free_switch_domain(bp);
1554 	bnxt_free_leds_info(bp);
1555 	bnxt_free_cos_queues(bp);
1556 	bnxt_free_link_info(bp);
1557 	bnxt_free_pf_info(bp);
1558 	bnxt_free_parent_info(bp);
1559 	bnxt_uninit_locks(bp);
1560 
1561 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1562 	bp->tx_mem_zone = NULL;
1563 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1564 	bp->rx_mem_zone = NULL;
1565 
1566 	bnxt_free_vf_info(bp);
1567 
1568 	rte_free(bp->grp_info);
1569 	bp->grp_info = NULL;
1570 }
1571 
1572 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1573 {
1574 	struct bnxt *bp = eth_dev->data->dev_private;
1575 	int ret = 0;
1576 
1577 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1578 		return 0;
1579 
1580 	pthread_mutex_lock(&bp->err_recovery_lock);
1581 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1582 		PMD_DRV_LOG(ERR,
1583 			    "Adapter recovering from error...Please retry\n");
1584 		pthread_mutex_unlock(&bp->err_recovery_lock);
1585 		return -EAGAIN;
1586 	}
1587 	pthread_mutex_unlock(&bp->err_recovery_lock);
1588 
1589 	/* cancel the recovery handler before remove dev */
1590 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1591 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1592 	bnxt_cancel_fc_thread(bp);
1593 
1594 	if (eth_dev->data->dev_started)
1595 		ret = bnxt_dev_stop(eth_dev);
1596 
1597 	bnxt_uninit_resources(bp, false);
1598 
1599 	bnxt_drv_uninit(bp);
1600 
1601 	return ret;
1602 }
1603 
1604 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1605 				    uint32_t index)
1606 {
1607 	struct bnxt *bp = eth_dev->data->dev_private;
1608 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1609 	struct bnxt_vnic_info *vnic;
1610 	struct bnxt_filter_info *filter, *temp_filter;
1611 	uint32_t i;
1612 
1613 	if (is_bnxt_in_error(bp))
1614 		return;
1615 
1616 	/*
1617 	 * Loop through all VNICs from the specified filter flow pools to
1618 	 * remove the corresponding MAC addr filter
1619 	 */
1620 	for (i = 0; i < bp->nr_vnics; i++) {
1621 		if (!(pool_mask & (1ULL << i)))
1622 			continue;
1623 
1624 		vnic = &bp->vnic_info[i];
1625 		filter = STAILQ_FIRST(&vnic->filter);
1626 		while (filter) {
1627 			temp_filter = STAILQ_NEXT(filter, next);
1628 			if (filter->mac_index == index) {
1629 				STAILQ_REMOVE(&vnic->filter, filter,
1630 						bnxt_filter_info, next);
1631 				bnxt_hwrm_clear_l2_filter(bp, filter);
1632 				bnxt_free_filter(bp, filter);
1633 			}
1634 			filter = temp_filter;
1635 		}
1636 	}
1637 }
1638 
1639 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1640 			       struct rte_ether_addr *mac_addr, uint32_t index,
1641 			       uint32_t pool)
1642 {
1643 	struct bnxt_filter_info *filter;
1644 	int rc = 0;
1645 
1646 	/* Attach requested MAC address to the new l2_filter */
1647 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1648 		if (filter->mac_index == index) {
1649 			PMD_DRV_LOG(DEBUG,
1650 				    "MAC addr already existed for pool %d\n",
1651 				    pool);
1652 			return 0;
1653 		}
1654 	}
1655 
1656 	filter = bnxt_alloc_filter(bp);
1657 	if (!filter) {
1658 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1659 		return -ENODEV;
1660 	}
1661 
1662 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1663 	 * if the MAC that's been programmed now is a different one, then,
1664 	 * copy that addr to filter->l2_addr
1665 	 */
1666 	if (mac_addr)
1667 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1668 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1669 
1670 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1671 	if (!rc) {
1672 		filter->mac_index = index;
1673 		if (filter->mac_index == 0)
1674 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1675 		else
1676 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1677 	} else {
1678 		bnxt_free_filter(bp, filter);
1679 	}
1680 
1681 	return rc;
1682 }
1683 
1684 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1685 				struct rte_ether_addr *mac_addr,
1686 				uint32_t index, uint32_t pool)
1687 {
1688 	struct bnxt *bp = eth_dev->data->dev_private;
1689 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1690 	int rc = 0;
1691 
1692 	rc = is_bnxt_in_error(bp);
1693 	if (rc)
1694 		return rc;
1695 
1696 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1697 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1698 		return -ENOTSUP;
1699 	}
1700 
1701 	if (!vnic) {
1702 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1703 		return -EINVAL;
1704 	}
1705 
1706 	/* Filter settings will get applied when port is started */
1707 	if (!eth_dev->data->dev_started)
1708 		return 0;
1709 
1710 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1711 
1712 	return rc;
1713 }
1714 
1715 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1716 {
1717 	int rc = 0;
1718 	struct bnxt *bp = eth_dev->data->dev_private;
1719 	struct rte_eth_link new;
1720 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1721 			BNXT_MIN_LINK_WAIT_CNT;
1722 
1723 	rc = is_bnxt_in_error(bp);
1724 	if (rc)
1725 		return rc;
1726 
1727 	memset(&new, 0, sizeof(new));
1728 	do {
1729 		/* Retrieve link info from hardware */
1730 		rc = bnxt_get_hwrm_link_config(bp, &new);
1731 		if (rc) {
1732 			new.link_speed = ETH_LINK_SPEED_100M;
1733 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1734 			PMD_DRV_LOG(ERR,
1735 				"Failed to retrieve link rc = 0x%x!\n", rc);
1736 			goto out;
1737 		}
1738 
1739 		if (!wait_to_complete || new.link_status)
1740 			break;
1741 
1742 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1743 	} while (cnt--);
1744 
1745 	/* Only single function PF can bring phy down.
1746 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1747 	 */
1748 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1749 		memset(&new, 0, sizeof(new));
1750 
1751 out:
1752 	/* Timed out or success */
1753 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1754 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1755 		rte_eth_linkstatus_set(eth_dev, &new);
1756 
1757 		rte_eth_dev_callback_process(eth_dev,
1758 					     RTE_ETH_EVENT_INTR_LSC,
1759 					     NULL);
1760 
1761 		bnxt_print_link_info(eth_dev);
1762 	}
1763 
1764 	return rc;
1765 }
1766 
1767 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1768 {
1769 	struct bnxt *bp = eth_dev->data->dev_private;
1770 	struct bnxt_vnic_info *vnic;
1771 	uint32_t old_flags;
1772 	int rc;
1773 
1774 	rc = is_bnxt_in_error(bp);
1775 	if (rc)
1776 		return rc;
1777 
1778 	/* Filter settings will get applied when port is started */
1779 	if (!eth_dev->data->dev_started)
1780 		return 0;
1781 
1782 	if (bp->vnic_info == NULL)
1783 		return 0;
1784 
1785 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1786 
1787 	old_flags = vnic->flags;
1788 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1789 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1790 	if (rc != 0)
1791 		vnic->flags = old_flags;
1792 
1793 	return rc;
1794 }
1795 
1796 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1797 {
1798 	struct bnxt *bp = eth_dev->data->dev_private;
1799 	struct bnxt_vnic_info *vnic;
1800 	uint32_t old_flags;
1801 	int rc;
1802 
1803 	rc = is_bnxt_in_error(bp);
1804 	if (rc)
1805 		return rc;
1806 
1807 	/* Filter settings will get applied when port is started */
1808 	if (!eth_dev->data->dev_started)
1809 		return 0;
1810 
1811 	if (bp->vnic_info == NULL)
1812 		return 0;
1813 
1814 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1815 
1816 	old_flags = vnic->flags;
1817 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1818 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1819 	if (rc != 0)
1820 		vnic->flags = old_flags;
1821 
1822 	return rc;
1823 }
1824 
1825 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1826 {
1827 	struct bnxt *bp = eth_dev->data->dev_private;
1828 	struct bnxt_vnic_info *vnic;
1829 	uint32_t old_flags;
1830 	int rc;
1831 
1832 	rc = is_bnxt_in_error(bp);
1833 	if (rc)
1834 		return rc;
1835 
1836 	/* Filter settings will get applied when port is started */
1837 	if (!eth_dev->data->dev_started)
1838 		return 0;
1839 
1840 	if (bp->vnic_info == NULL)
1841 		return 0;
1842 
1843 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1844 
1845 	old_flags = vnic->flags;
1846 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1847 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1848 	if (rc != 0)
1849 		vnic->flags = old_flags;
1850 
1851 	return rc;
1852 }
1853 
1854 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1855 {
1856 	struct bnxt *bp = eth_dev->data->dev_private;
1857 	struct bnxt_vnic_info *vnic;
1858 	uint32_t old_flags;
1859 	int rc;
1860 
1861 	rc = is_bnxt_in_error(bp);
1862 	if (rc)
1863 		return rc;
1864 
1865 	/* Filter settings will get applied when port is started */
1866 	if (!eth_dev->data->dev_started)
1867 		return 0;
1868 
1869 	if (bp->vnic_info == NULL)
1870 		return 0;
1871 
1872 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1873 
1874 	old_flags = vnic->flags;
1875 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1876 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1877 	if (rc != 0)
1878 		vnic->flags = old_flags;
1879 
1880 	return rc;
1881 }
1882 
1883 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1884 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1885 {
1886 	if (qid >= bp->rx_nr_rings)
1887 		return NULL;
1888 
1889 	return bp->eth_dev->data->rx_queues[qid];
1890 }
1891 
1892 /* Return rxq corresponding to a given rss table ring/group ID. */
1893 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1894 {
1895 	struct bnxt_rx_queue *rxq;
1896 	unsigned int i;
1897 
1898 	if (!BNXT_HAS_RING_GRPS(bp)) {
1899 		for (i = 0; i < bp->rx_nr_rings; i++) {
1900 			rxq = bp->eth_dev->data->rx_queues[i];
1901 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1902 				return rxq->index;
1903 		}
1904 	} else {
1905 		for (i = 0; i < bp->rx_nr_rings; i++) {
1906 			if (bp->grp_info[i].fw_grp_id == fwr)
1907 				return i;
1908 		}
1909 	}
1910 
1911 	return INVALID_HW_RING_ID;
1912 }
1913 
1914 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1915 			    struct rte_eth_rss_reta_entry64 *reta_conf,
1916 			    uint16_t reta_size)
1917 {
1918 	struct bnxt *bp = eth_dev->data->dev_private;
1919 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1920 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1921 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1922 	uint16_t idx, sft;
1923 	int i, rc;
1924 
1925 	rc = is_bnxt_in_error(bp);
1926 	if (rc)
1927 		return rc;
1928 
1929 	if (!vnic->rss_table)
1930 		return -EINVAL;
1931 
1932 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
1933 		return -EINVAL;
1934 
1935 	if (reta_size != tbl_size) {
1936 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1937 			"(%d) must equal the size supported by the hardware "
1938 			"(%d)\n", reta_size, tbl_size);
1939 		return -EINVAL;
1940 	}
1941 
1942 	for (i = 0; i < reta_size; i++) {
1943 		struct bnxt_rx_queue *rxq;
1944 
1945 		idx = i / RTE_RETA_GROUP_SIZE;
1946 		sft = i % RTE_RETA_GROUP_SIZE;
1947 
1948 		if (!(reta_conf[idx].mask & (1ULL << sft)))
1949 			continue;
1950 
1951 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
1952 		if (!rxq) {
1953 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
1954 			return -EINVAL;
1955 		}
1956 
1957 		if (BNXT_CHIP_P5(bp)) {
1958 			vnic->rss_table[i * 2] =
1959 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
1960 			vnic->rss_table[i * 2 + 1] =
1961 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
1962 		} else {
1963 			vnic->rss_table[i] =
1964 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
1965 		}
1966 	}
1967 
1968 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1969 	return rc;
1970 }
1971 
1972 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
1973 			      struct rte_eth_rss_reta_entry64 *reta_conf,
1974 			      uint16_t reta_size)
1975 {
1976 	struct bnxt *bp = eth_dev->data->dev_private;
1977 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1978 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1979 	uint16_t idx, sft, i;
1980 	int rc;
1981 
1982 	rc = is_bnxt_in_error(bp);
1983 	if (rc)
1984 		return rc;
1985 
1986 	/* Retrieve from the default VNIC */
1987 	if (!vnic)
1988 		return -EINVAL;
1989 	if (!vnic->rss_table)
1990 		return -EINVAL;
1991 
1992 	if (reta_size != tbl_size) {
1993 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1994 			"(%d) must equal the size supported by the hardware "
1995 			"(%d)\n", reta_size, tbl_size);
1996 		return -EINVAL;
1997 	}
1998 
1999 	for (idx = 0, i = 0; i < reta_size; i++) {
2000 		idx = i / RTE_RETA_GROUP_SIZE;
2001 		sft = i % RTE_RETA_GROUP_SIZE;
2002 
2003 		if (reta_conf[idx].mask & (1ULL << sft)) {
2004 			uint16_t qid;
2005 
2006 			if (BNXT_CHIP_P5(bp))
2007 				qid = bnxt_rss_to_qid(bp,
2008 						      vnic->rss_table[i * 2]);
2009 			else
2010 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2011 
2012 			if (qid == INVALID_HW_RING_ID) {
2013 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2014 				return -EINVAL;
2015 			}
2016 			reta_conf[idx].reta[sft] = qid;
2017 		}
2018 	}
2019 
2020 	return 0;
2021 }
2022 
2023 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2024 				   struct rte_eth_rss_conf *rss_conf)
2025 {
2026 	struct bnxt *bp = eth_dev->data->dev_private;
2027 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2028 	struct bnxt_vnic_info *vnic;
2029 	int rc;
2030 
2031 	rc = is_bnxt_in_error(bp);
2032 	if (rc)
2033 		return rc;
2034 
2035 	/*
2036 	 * If RSS enablement were different than dev_configure,
2037 	 * then return -EINVAL
2038 	 */
2039 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
2040 		if (!rss_conf->rss_hf)
2041 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2042 	} else {
2043 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2044 			return -EINVAL;
2045 	}
2046 
2047 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
2048 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2049 	       rss_conf,
2050 	       sizeof(*rss_conf));
2051 
2052 	/* Update the default RSS VNIC(s) */
2053 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2054 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2055 	vnic->hash_mode =
2056 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2057 					    ETH_RSS_LEVEL(rss_conf->rss_hf));
2058 
2059 	/*
2060 	 * If hashkey is not specified, use the previously configured
2061 	 * hashkey
2062 	 */
2063 	if (!rss_conf->rss_key)
2064 		goto rss_config;
2065 
2066 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2067 		PMD_DRV_LOG(ERR,
2068 			    "Invalid hashkey length, should be 16 bytes\n");
2069 		return -EINVAL;
2070 	}
2071 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2072 
2073 rss_config:
2074 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2075 	return rc;
2076 }
2077 
2078 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2079 				     struct rte_eth_rss_conf *rss_conf)
2080 {
2081 	struct bnxt *bp = eth_dev->data->dev_private;
2082 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2083 	int len, rc;
2084 	uint32_t hash_types;
2085 
2086 	rc = is_bnxt_in_error(bp);
2087 	if (rc)
2088 		return rc;
2089 
2090 	/* RSS configuration is the same for all VNICs */
2091 	if (vnic && vnic->rss_hash_key) {
2092 		if (rss_conf->rss_key) {
2093 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2094 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2095 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2096 		}
2097 
2098 		hash_types = vnic->hash_type;
2099 		rss_conf->rss_hf = 0;
2100 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2101 			rss_conf->rss_hf |= ETH_RSS_IPV4;
2102 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2103 		}
2104 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2105 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2106 			hash_types &=
2107 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2108 		}
2109 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2110 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2111 			hash_types &=
2112 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2113 		}
2114 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2115 			rss_conf->rss_hf |= ETH_RSS_IPV6;
2116 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2117 		}
2118 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2119 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2120 			hash_types &=
2121 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2122 		}
2123 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2124 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2125 			hash_types &=
2126 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2127 		}
2128 
2129 		rss_conf->rss_hf |=
2130 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2131 
2132 		if (hash_types) {
2133 			PMD_DRV_LOG(ERR,
2134 				"Unknown RSS config from firmware (%08x), RSS disabled",
2135 				vnic->hash_type);
2136 			return -ENOTSUP;
2137 		}
2138 	} else {
2139 		rss_conf->rss_hf = 0;
2140 	}
2141 	return 0;
2142 }
2143 
2144 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2145 			       struct rte_eth_fc_conf *fc_conf)
2146 {
2147 	struct bnxt *bp = dev->data->dev_private;
2148 	struct rte_eth_link link_info;
2149 	int rc;
2150 
2151 	rc = is_bnxt_in_error(bp);
2152 	if (rc)
2153 		return rc;
2154 
2155 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2156 	if (rc)
2157 		return rc;
2158 
2159 	memset(fc_conf, 0, sizeof(*fc_conf));
2160 	if (bp->link_info->auto_pause)
2161 		fc_conf->autoneg = 1;
2162 	switch (bp->link_info->pause) {
2163 	case 0:
2164 		fc_conf->mode = RTE_FC_NONE;
2165 		break;
2166 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2167 		fc_conf->mode = RTE_FC_TX_PAUSE;
2168 		break;
2169 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2170 		fc_conf->mode = RTE_FC_RX_PAUSE;
2171 		break;
2172 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2173 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2174 		fc_conf->mode = RTE_FC_FULL;
2175 		break;
2176 	}
2177 	return 0;
2178 }
2179 
2180 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2181 			       struct rte_eth_fc_conf *fc_conf)
2182 {
2183 	struct bnxt *bp = dev->data->dev_private;
2184 	int rc;
2185 
2186 	rc = is_bnxt_in_error(bp);
2187 	if (rc)
2188 		return rc;
2189 
2190 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2191 		PMD_DRV_LOG(ERR, "Flow Control Settings cannot be modified\n");
2192 		return -ENOTSUP;
2193 	}
2194 
2195 	switch (fc_conf->mode) {
2196 	case RTE_FC_NONE:
2197 		bp->link_info->auto_pause = 0;
2198 		bp->link_info->force_pause = 0;
2199 		break;
2200 	case RTE_FC_RX_PAUSE:
2201 		if (fc_conf->autoneg) {
2202 			bp->link_info->auto_pause =
2203 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2204 			bp->link_info->force_pause = 0;
2205 		} else {
2206 			bp->link_info->auto_pause = 0;
2207 			bp->link_info->force_pause =
2208 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2209 		}
2210 		break;
2211 	case RTE_FC_TX_PAUSE:
2212 		if (fc_conf->autoneg) {
2213 			bp->link_info->auto_pause =
2214 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2215 			bp->link_info->force_pause = 0;
2216 		} else {
2217 			bp->link_info->auto_pause = 0;
2218 			bp->link_info->force_pause =
2219 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2220 		}
2221 		break;
2222 	case RTE_FC_FULL:
2223 		if (fc_conf->autoneg) {
2224 			bp->link_info->auto_pause =
2225 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2226 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2227 			bp->link_info->force_pause = 0;
2228 		} else {
2229 			bp->link_info->auto_pause = 0;
2230 			bp->link_info->force_pause =
2231 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2232 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2233 		}
2234 		break;
2235 	}
2236 	return bnxt_set_hwrm_link_config(bp, true);
2237 }
2238 
2239 /* Add UDP tunneling port */
2240 static int
2241 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2242 			 struct rte_eth_udp_tunnel *udp_tunnel)
2243 {
2244 	struct bnxt *bp = eth_dev->data->dev_private;
2245 	uint16_t tunnel_type = 0;
2246 	int rc = 0;
2247 
2248 	rc = is_bnxt_in_error(bp);
2249 	if (rc)
2250 		return rc;
2251 
2252 	switch (udp_tunnel->prot_type) {
2253 	case RTE_TUNNEL_TYPE_VXLAN:
2254 		if (bp->vxlan_port_cnt) {
2255 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2256 				udp_tunnel->udp_port);
2257 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2258 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2259 				return -ENOSPC;
2260 			}
2261 			bp->vxlan_port_cnt++;
2262 			return 0;
2263 		}
2264 		tunnel_type =
2265 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2266 		bp->vxlan_port_cnt++;
2267 		break;
2268 	case RTE_TUNNEL_TYPE_GENEVE:
2269 		if (bp->geneve_port_cnt) {
2270 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2271 				udp_tunnel->udp_port);
2272 			if (bp->geneve_port != udp_tunnel->udp_port) {
2273 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2274 				return -ENOSPC;
2275 			}
2276 			bp->geneve_port_cnt++;
2277 			return 0;
2278 		}
2279 		tunnel_type =
2280 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2281 		bp->geneve_port_cnt++;
2282 		break;
2283 	default:
2284 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2285 		return -ENOTSUP;
2286 	}
2287 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2288 					     tunnel_type);
2289 	return rc;
2290 }
2291 
2292 static int
2293 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2294 			 struct rte_eth_udp_tunnel *udp_tunnel)
2295 {
2296 	struct bnxt *bp = eth_dev->data->dev_private;
2297 	uint16_t tunnel_type = 0;
2298 	uint16_t port = 0;
2299 	int rc = 0;
2300 
2301 	rc = is_bnxt_in_error(bp);
2302 	if (rc)
2303 		return rc;
2304 
2305 	switch (udp_tunnel->prot_type) {
2306 	case RTE_TUNNEL_TYPE_VXLAN:
2307 		if (!bp->vxlan_port_cnt) {
2308 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2309 			return -EINVAL;
2310 		}
2311 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2312 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2313 				udp_tunnel->udp_port, bp->vxlan_port);
2314 			return -EINVAL;
2315 		}
2316 		if (--bp->vxlan_port_cnt)
2317 			return 0;
2318 
2319 		tunnel_type =
2320 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2321 		port = bp->vxlan_fw_dst_port_id;
2322 		break;
2323 	case RTE_TUNNEL_TYPE_GENEVE:
2324 		if (!bp->geneve_port_cnt) {
2325 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2326 			return -EINVAL;
2327 		}
2328 		if (bp->geneve_port != udp_tunnel->udp_port) {
2329 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2330 				udp_tunnel->udp_port, bp->geneve_port);
2331 			return -EINVAL;
2332 		}
2333 		if (--bp->geneve_port_cnt)
2334 			return 0;
2335 
2336 		tunnel_type =
2337 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2338 		port = bp->geneve_fw_dst_port_id;
2339 		break;
2340 	default:
2341 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2342 		return -ENOTSUP;
2343 	}
2344 
2345 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2346 	return rc;
2347 }
2348 
2349 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2350 {
2351 	struct bnxt_filter_info *filter;
2352 	struct bnxt_vnic_info *vnic;
2353 	int rc = 0;
2354 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2355 
2356 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2357 	filter = STAILQ_FIRST(&vnic->filter);
2358 	while (filter) {
2359 		/* Search for this matching MAC+VLAN filter */
2360 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2361 			/* Delete the filter */
2362 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2363 			if (rc)
2364 				return rc;
2365 			STAILQ_REMOVE(&vnic->filter, filter,
2366 				      bnxt_filter_info, next);
2367 			bnxt_free_filter(bp, filter);
2368 			PMD_DRV_LOG(INFO,
2369 				    "Deleted vlan filter for %d\n",
2370 				    vlan_id);
2371 			return 0;
2372 		}
2373 		filter = STAILQ_NEXT(filter, next);
2374 	}
2375 	return -ENOENT;
2376 }
2377 
2378 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2379 {
2380 	struct bnxt_filter_info *filter;
2381 	struct bnxt_vnic_info *vnic;
2382 	int rc = 0;
2383 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2384 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2385 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2386 
2387 	/* Implementation notes on the use of VNIC in this command:
2388 	 *
2389 	 * By default, these filters belong to default vnic for the function.
2390 	 * Once these filters are set up, only destination VNIC can be modified.
2391 	 * If the destination VNIC is not specified in this command,
2392 	 * then the HWRM shall only create an l2 context id.
2393 	 */
2394 
2395 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2396 	filter = STAILQ_FIRST(&vnic->filter);
2397 	/* Check if the VLAN has already been added */
2398 	while (filter) {
2399 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2400 			return -EEXIST;
2401 
2402 		filter = STAILQ_NEXT(filter, next);
2403 	}
2404 
2405 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2406 	 * command to create MAC+VLAN filter with the right flags, enables set.
2407 	 */
2408 	filter = bnxt_alloc_filter(bp);
2409 	if (!filter) {
2410 		PMD_DRV_LOG(ERR,
2411 			    "MAC/VLAN filter alloc failed\n");
2412 		return -ENOMEM;
2413 	}
2414 	/* MAC + VLAN ID filter */
2415 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2416 	 * untagged packets are received
2417 	 *
2418 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2419 	 * packets and only the programmed vlan's packets are received
2420 	 */
2421 	filter->l2_ivlan = vlan_id;
2422 	filter->l2_ivlan_mask = 0x0FFF;
2423 	filter->enables |= en;
2424 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2425 
2426 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2427 	if (rc) {
2428 		/* Free the newly allocated filter as we were
2429 		 * not able to create the filter in hardware.
2430 		 */
2431 		bnxt_free_filter(bp, filter);
2432 		return rc;
2433 	}
2434 
2435 	filter->mac_index = 0;
2436 	/* Add this new filter to the list */
2437 	if (vlan_id == 0)
2438 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2439 	else
2440 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2441 
2442 	PMD_DRV_LOG(INFO,
2443 		    "Added Vlan filter for %d\n", vlan_id);
2444 	return rc;
2445 }
2446 
2447 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2448 		uint16_t vlan_id, int on)
2449 {
2450 	struct bnxt *bp = eth_dev->data->dev_private;
2451 	int rc;
2452 
2453 	rc = is_bnxt_in_error(bp);
2454 	if (rc)
2455 		return rc;
2456 
2457 	if (!eth_dev->data->dev_started) {
2458 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2459 		return -EINVAL;
2460 	}
2461 
2462 	/* These operations apply to ALL existing MAC/VLAN filters */
2463 	if (on)
2464 		return bnxt_add_vlan_filter(bp, vlan_id);
2465 	else
2466 		return bnxt_del_vlan_filter(bp, vlan_id);
2467 }
2468 
2469 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2470 				    struct bnxt_vnic_info *vnic)
2471 {
2472 	struct bnxt_filter_info *filter;
2473 	int rc;
2474 
2475 	filter = STAILQ_FIRST(&vnic->filter);
2476 	while (filter) {
2477 		if (filter->mac_index == 0 &&
2478 		    !memcmp(filter->l2_addr, bp->mac_addr,
2479 			    RTE_ETHER_ADDR_LEN)) {
2480 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2481 			if (!rc) {
2482 				STAILQ_REMOVE(&vnic->filter, filter,
2483 					      bnxt_filter_info, next);
2484 				bnxt_free_filter(bp, filter);
2485 			}
2486 			return rc;
2487 		}
2488 		filter = STAILQ_NEXT(filter, next);
2489 	}
2490 	return 0;
2491 }
2492 
2493 static int
2494 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2495 {
2496 	struct bnxt_vnic_info *vnic;
2497 	unsigned int i;
2498 	int rc;
2499 
2500 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2501 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2502 		/* Remove any VLAN filters programmed */
2503 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2504 			bnxt_del_vlan_filter(bp, i);
2505 
2506 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2507 		if (rc)
2508 			return rc;
2509 	} else {
2510 		/* Default filter will allow packets that match the
2511 		 * dest mac. So, it has to be deleted, otherwise, we
2512 		 * will endup receiving vlan packets for which the
2513 		 * filter is not programmed, when hw-vlan-filter
2514 		 * configuration is ON
2515 		 */
2516 		bnxt_del_dflt_mac_filter(bp, vnic);
2517 		/* This filter will allow only untagged packets */
2518 		bnxt_add_vlan_filter(bp, 0);
2519 	}
2520 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2521 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2522 
2523 	return 0;
2524 }
2525 
2526 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2527 {
2528 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2529 	unsigned int i;
2530 	int rc;
2531 
2532 	/* Destroy vnic filters and vnic */
2533 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2534 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2535 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2536 			bnxt_del_vlan_filter(bp, i);
2537 	}
2538 	bnxt_del_dflt_mac_filter(bp, vnic);
2539 
2540 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2541 	if (rc)
2542 		return rc;
2543 
2544 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2545 	if (rc)
2546 		return rc;
2547 
2548 	rte_free(vnic->fw_grp_ids);
2549 	vnic->fw_grp_ids = NULL;
2550 
2551 	vnic->rx_queue_cnt = 0;
2552 
2553 	return 0;
2554 }
2555 
2556 static int
2557 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2558 {
2559 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2560 	int rc;
2561 
2562 	/* Destroy, recreate and reconfigure the default vnic */
2563 	rc = bnxt_free_one_vnic(bp, 0);
2564 	if (rc)
2565 		return rc;
2566 
2567 	/* default vnic 0 */
2568 	rc = bnxt_setup_one_vnic(bp, 0);
2569 	if (rc)
2570 		return rc;
2571 
2572 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2573 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2574 		rc = bnxt_add_vlan_filter(bp, 0);
2575 		if (rc)
2576 			return rc;
2577 		rc = bnxt_restore_vlan_filters(bp);
2578 		if (rc)
2579 			return rc;
2580 	} else {
2581 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2582 		if (rc)
2583 			return rc;
2584 	}
2585 
2586 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2587 	if (rc)
2588 		return rc;
2589 
2590 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2591 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2592 
2593 	return rc;
2594 }
2595 
2596 static int
2597 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2598 {
2599 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2600 	struct bnxt *bp = dev->data->dev_private;
2601 	int rc;
2602 
2603 	rc = is_bnxt_in_error(bp);
2604 	if (rc)
2605 		return rc;
2606 
2607 	/* Filter settings will get applied when port is started */
2608 	if (!dev->data->dev_started)
2609 		return 0;
2610 
2611 	if (mask & ETH_VLAN_FILTER_MASK) {
2612 		/* Enable or disable VLAN filtering */
2613 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2614 		if (rc)
2615 			return rc;
2616 	}
2617 
2618 	if (mask & ETH_VLAN_STRIP_MASK) {
2619 		/* Enable or disable VLAN stripping */
2620 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2621 		if (rc)
2622 			return rc;
2623 	}
2624 
2625 	if (mask & ETH_VLAN_EXTEND_MASK) {
2626 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2627 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2628 		else
2629 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2630 	}
2631 
2632 	return 0;
2633 }
2634 
2635 static int
2636 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2637 		      uint16_t tpid)
2638 {
2639 	struct bnxt *bp = dev->data->dev_private;
2640 	int qinq = dev->data->dev_conf.rxmode.offloads &
2641 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2642 
2643 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2644 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2645 		PMD_DRV_LOG(ERR,
2646 			    "Unsupported vlan type.");
2647 		return -EINVAL;
2648 	}
2649 	if (!qinq) {
2650 		PMD_DRV_LOG(ERR,
2651 			    "QinQ not enabled. Needs to be ON as we can "
2652 			    "accelerate only outer vlan\n");
2653 		return -EINVAL;
2654 	}
2655 
2656 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2657 		switch (tpid) {
2658 		case RTE_ETHER_TYPE_QINQ:
2659 			bp->outer_tpid_bd =
2660 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2661 				break;
2662 		case RTE_ETHER_TYPE_VLAN:
2663 			bp->outer_tpid_bd =
2664 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2665 				break;
2666 		case RTE_ETHER_TYPE_QINQ1:
2667 			bp->outer_tpid_bd =
2668 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2669 				break;
2670 		case RTE_ETHER_TYPE_QINQ2:
2671 			bp->outer_tpid_bd =
2672 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2673 				break;
2674 		case RTE_ETHER_TYPE_QINQ3:
2675 			bp->outer_tpid_bd =
2676 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2677 				break;
2678 		default:
2679 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2680 			return -EINVAL;
2681 		}
2682 		bp->outer_tpid_bd |= tpid;
2683 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2684 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2685 		PMD_DRV_LOG(ERR,
2686 			    "Can accelerate only outer vlan in QinQ\n");
2687 		return -EINVAL;
2688 	}
2689 
2690 	return 0;
2691 }
2692 
2693 static int
2694 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2695 			     struct rte_ether_addr *addr)
2696 {
2697 	struct bnxt *bp = dev->data->dev_private;
2698 	/* Default Filter is tied to VNIC 0 */
2699 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2700 	int rc;
2701 
2702 	rc = is_bnxt_in_error(bp);
2703 	if (rc)
2704 		return rc;
2705 
2706 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2707 		return -EPERM;
2708 
2709 	if (rte_is_zero_ether_addr(addr))
2710 		return -EINVAL;
2711 
2712 	/* Filter settings will get applied when port is started */
2713 	if (!dev->data->dev_started)
2714 		return 0;
2715 
2716 	/* Check if the requested MAC is already added */
2717 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2718 		return 0;
2719 
2720 	/* Destroy filter and re-create it */
2721 	bnxt_del_dflt_mac_filter(bp, vnic);
2722 
2723 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2724 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2725 		/* This filter will allow only untagged packets */
2726 		rc = bnxt_add_vlan_filter(bp, 0);
2727 	} else {
2728 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2729 	}
2730 
2731 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2732 	return rc;
2733 }
2734 
2735 static int
2736 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2737 			  struct rte_ether_addr *mc_addr_set,
2738 			  uint32_t nb_mc_addr)
2739 {
2740 	struct bnxt *bp = eth_dev->data->dev_private;
2741 	char *mc_addr_list = (char *)mc_addr_set;
2742 	struct bnxt_vnic_info *vnic;
2743 	uint32_t off = 0, i = 0;
2744 	int rc;
2745 
2746 	rc = is_bnxt_in_error(bp);
2747 	if (rc)
2748 		return rc;
2749 
2750 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2751 
2752 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2753 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2754 		goto allmulti;
2755 	}
2756 
2757 	/* TODO Check for Duplicate mcast addresses */
2758 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2759 	for (i = 0; i < nb_mc_addr; i++) {
2760 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2761 			RTE_ETHER_ADDR_LEN);
2762 		off += RTE_ETHER_ADDR_LEN;
2763 	}
2764 
2765 	vnic->mc_addr_cnt = i;
2766 	if (vnic->mc_addr_cnt)
2767 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2768 	else
2769 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2770 
2771 allmulti:
2772 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2773 }
2774 
2775 static int
2776 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2777 {
2778 	struct bnxt *bp = dev->data->dev_private;
2779 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2780 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2781 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2782 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2783 	int ret;
2784 
2785 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2786 			fw_major, fw_minor, fw_updt, fw_rsvd);
2787 
2788 	ret += 1; /* add the size of '\0' */
2789 	if (fw_size < (uint32_t)ret)
2790 		return ret;
2791 	else
2792 		return 0;
2793 }
2794 
2795 static void
2796 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2797 	struct rte_eth_rxq_info *qinfo)
2798 {
2799 	struct bnxt *bp = dev->data->dev_private;
2800 	struct bnxt_rx_queue *rxq;
2801 
2802 	if (is_bnxt_in_error(bp))
2803 		return;
2804 
2805 	rxq = dev->data->rx_queues[queue_id];
2806 
2807 	qinfo->mp = rxq->mb_pool;
2808 	qinfo->scattered_rx = dev->data->scattered_rx;
2809 	qinfo->nb_desc = rxq->nb_rx_desc;
2810 
2811 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2812 	qinfo->conf.rx_drop_en = rxq->drop_en;
2813 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2814 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2815 }
2816 
2817 static void
2818 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2819 	struct rte_eth_txq_info *qinfo)
2820 {
2821 	struct bnxt *bp = dev->data->dev_private;
2822 	struct bnxt_tx_queue *txq;
2823 
2824 	if (is_bnxt_in_error(bp))
2825 		return;
2826 
2827 	txq = dev->data->tx_queues[queue_id];
2828 
2829 	qinfo->nb_desc = txq->nb_tx_desc;
2830 
2831 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2832 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2833 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2834 
2835 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2836 	qinfo->conf.tx_rs_thresh = 0;
2837 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2838 	qinfo->conf.offloads = txq->offloads;
2839 }
2840 
2841 static const struct {
2842 	eth_rx_burst_t pkt_burst;
2843 	const char *info;
2844 } bnxt_rx_burst_info[] = {
2845 	{bnxt_recv_pkts,	"Scalar"},
2846 #if defined(RTE_ARCH_X86)
2847 	{bnxt_recv_pkts_vec,	"Vector SSE"},
2848 #elif defined(RTE_ARCH_ARM64)
2849 	{bnxt_recv_pkts_vec,	"Vector Neon"},
2850 #endif
2851 };
2852 
2853 static int
2854 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2855 		       struct rte_eth_burst_mode *mode)
2856 {
2857 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2858 	size_t i;
2859 
2860 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2861 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2862 			snprintf(mode->info, sizeof(mode->info), "%s",
2863 				 bnxt_rx_burst_info[i].info);
2864 			return 0;
2865 		}
2866 	}
2867 
2868 	return -EINVAL;
2869 }
2870 
2871 static const struct {
2872 	eth_tx_burst_t pkt_burst;
2873 	const char *info;
2874 } bnxt_tx_burst_info[] = {
2875 	{bnxt_xmit_pkts,	"Scalar"},
2876 #if defined(RTE_ARCH_X86)
2877 	{bnxt_xmit_pkts_vec,	"Vector SSE"},
2878 #elif defined(RTE_ARCH_ARM64)
2879 	{bnxt_xmit_pkts_vec,	"Vector Neon"},
2880 #endif
2881 };
2882 
2883 static int
2884 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2885 		       struct rte_eth_burst_mode *mode)
2886 {
2887 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2888 	size_t i;
2889 
2890 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2891 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2892 			snprintf(mode->info, sizeof(mode->info), "%s",
2893 				 bnxt_tx_burst_info[i].info);
2894 			return 0;
2895 		}
2896 	}
2897 
2898 	return -EINVAL;
2899 }
2900 
2901 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2902 {
2903 	struct bnxt *bp = eth_dev->data->dev_private;
2904 	uint32_t new_pkt_size;
2905 	uint32_t rc = 0;
2906 	uint32_t i;
2907 
2908 	rc = is_bnxt_in_error(bp);
2909 	if (rc)
2910 		return rc;
2911 
2912 	/* Exit if receive queues are not configured yet */
2913 	if (!eth_dev->data->nb_rx_queues)
2914 		return rc;
2915 
2916 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2917 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2918 
2919 	/*
2920 	 * Disallow any MTU change that would require scattered receive support
2921 	 * if it is not already enabled.
2922 	 */
2923 	if (eth_dev->data->dev_started &&
2924 	    !eth_dev->data->scattered_rx &&
2925 	    (new_pkt_size >
2926 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2927 		PMD_DRV_LOG(ERR,
2928 			    "MTU change would require scattered rx support. ");
2929 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2930 		return -EINVAL;
2931 	}
2932 
2933 	if (new_mtu > RTE_ETHER_MTU) {
2934 		bp->flags |= BNXT_FLAG_JUMBO;
2935 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
2936 			DEV_RX_OFFLOAD_JUMBO_FRAME;
2937 	} else {
2938 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
2939 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
2940 		bp->flags &= ~BNXT_FLAG_JUMBO;
2941 	}
2942 
2943 	/* Is there a change in mtu setting? */
2944 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2945 		return rc;
2946 
2947 	for (i = 0; i < bp->nr_vnics; i++) {
2948 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2949 		uint16_t size = 0;
2950 
2951 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
2952 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2953 		if (rc)
2954 			break;
2955 
2956 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2957 		size -= RTE_PKTMBUF_HEADROOM;
2958 
2959 		if (size < new_mtu) {
2960 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2961 			if (rc)
2962 				return rc;
2963 		}
2964 	}
2965 
2966 	if (!rc)
2967 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2968 
2969 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2970 
2971 	return rc;
2972 }
2973 
2974 static int
2975 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2976 {
2977 	struct bnxt *bp = dev->data->dev_private;
2978 	uint16_t vlan = bp->vlan;
2979 	int rc;
2980 
2981 	rc = is_bnxt_in_error(bp);
2982 	if (rc)
2983 		return rc;
2984 
2985 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2986 		PMD_DRV_LOG(ERR,
2987 			"PVID cannot be modified for this function\n");
2988 		return -ENOTSUP;
2989 	}
2990 	bp->vlan = on ? pvid : 0;
2991 
2992 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
2993 	if (rc)
2994 		bp->vlan = vlan;
2995 	return rc;
2996 }
2997 
2998 static int
2999 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3000 {
3001 	struct bnxt *bp = dev->data->dev_private;
3002 	int rc;
3003 
3004 	rc = is_bnxt_in_error(bp);
3005 	if (rc)
3006 		return rc;
3007 
3008 	return bnxt_hwrm_port_led_cfg(bp, true);
3009 }
3010 
3011 static int
3012 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3013 {
3014 	struct bnxt *bp = dev->data->dev_private;
3015 	int rc;
3016 
3017 	rc = is_bnxt_in_error(bp);
3018 	if (rc)
3019 		return rc;
3020 
3021 	return bnxt_hwrm_port_led_cfg(bp, false);
3022 }
3023 
3024 static uint32_t
3025 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
3026 {
3027 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
3028 	struct bnxt_cp_ring_info *cpr;
3029 	uint32_t desc = 0, raw_cons;
3030 	struct bnxt_rx_queue *rxq;
3031 	struct rx_pkt_cmpl *rxcmp;
3032 	int rc;
3033 
3034 	rc = is_bnxt_in_error(bp);
3035 	if (rc)
3036 		return rc;
3037 
3038 	rxq = dev->data->rx_queues[rx_queue_id];
3039 	cpr = rxq->cp_ring;
3040 	raw_cons = cpr->cp_raw_cons;
3041 
3042 	while (1) {
3043 		uint32_t agg_cnt, cons, cmpl_type;
3044 
3045 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3046 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3047 
3048 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3049 			break;
3050 
3051 		cmpl_type = CMP_TYPE(rxcmp);
3052 
3053 		switch (cmpl_type) {
3054 		case CMPL_BASE_TYPE_RX_L2:
3055 		case CMPL_BASE_TYPE_RX_L2_V2:
3056 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3057 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3058 			desc++;
3059 			break;
3060 
3061 		case CMPL_BASE_TYPE_RX_TPA_END:
3062 			if (BNXT_CHIP_P5(rxq->bp)) {
3063 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3064 
3065 				p5_tpa_end = (void *)rxcmp;
3066 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3067 			} else {
3068 				struct rx_tpa_end_cmpl *tpa_end;
3069 
3070 				tpa_end = (void *)rxcmp;
3071 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3072 			}
3073 
3074 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3075 			desc++;
3076 			break;
3077 
3078 		default:
3079 			raw_cons += CMP_LEN(cmpl_type);
3080 		}
3081 	}
3082 
3083 	return desc;
3084 }
3085 
3086 static int
3087 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3088 {
3089 	struct bnxt_rx_queue *rxq = rx_queue;
3090 	struct bnxt_cp_ring_info *cpr;
3091 	struct bnxt_rx_ring_info *rxr;
3092 	uint32_t desc, raw_cons;
3093 	struct bnxt *bp = rxq->bp;
3094 	struct rx_pkt_cmpl *rxcmp;
3095 	int rc;
3096 
3097 	rc = is_bnxt_in_error(bp);
3098 	if (rc)
3099 		return rc;
3100 
3101 	if (offset >= rxq->nb_rx_desc)
3102 		return -EINVAL;
3103 
3104 	rxr = rxq->rx_ring;
3105 	cpr = rxq->cp_ring;
3106 
3107 	/*
3108 	 * For the vector receive case, the completion at the requested
3109 	 * offset can be indexed directly.
3110 	 */
3111 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3112 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3113 		struct rx_pkt_cmpl *rxcmp;
3114 		uint32_t cons;
3115 
3116 		/* Check status of completion descriptor. */
3117 		raw_cons = cpr->cp_raw_cons +
3118 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3119 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3120 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3121 
3122 		if (CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3123 			return RTE_ETH_RX_DESC_DONE;
3124 
3125 		/* Check whether rx desc has an mbuf attached. */
3126 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3127 		if (cons >= rxq->rxrearm_start &&
3128 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3129 			return RTE_ETH_RX_DESC_UNAVAIL;
3130 		}
3131 
3132 		return RTE_ETH_RX_DESC_AVAIL;
3133 	}
3134 #endif
3135 
3136 	/*
3137 	 * For the non-vector receive case, scan the completion ring to
3138 	 * locate the completion descriptor for the requested offset.
3139 	 */
3140 	raw_cons = cpr->cp_raw_cons;
3141 	desc = 0;
3142 	while (1) {
3143 		uint32_t agg_cnt, cons, cmpl_type;
3144 
3145 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3146 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3147 
3148 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3149 			break;
3150 
3151 		cmpl_type = CMP_TYPE(rxcmp);
3152 
3153 		switch (cmpl_type) {
3154 		case CMPL_BASE_TYPE_RX_L2:
3155 		case CMPL_BASE_TYPE_RX_L2_V2:
3156 			if (desc == offset) {
3157 				cons = rxcmp->opaque;
3158 				if (rxr->rx_buf_ring[cons])
3159 					return RTE_ETH_RX_DESC_DONE;
3160 				else
3161 					return RTE_ETH_RX_DESC_UNAVAIL;
3162 			}
3163 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3164 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3165 			desc++;
3166 			break;
3167 
3168 		case CMPL_BASE_TYPE_RX_TPA_END:
3169 			if (desc == offset)
3170 				return RTE_ETH_RX_DESC_DONE;
3171 
3172 			if (BNXT_CHIP_P5(rxq->bp)) {
3173 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3174 
3175 				p5_tpa_end = (void *)rxcmp;
3176 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3177 			} else {
3178 				struct rx_tpa_end_cmpl *tpa_end;
3179 
3180 				tpa_end = (void *)rxcmp;
3181 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3182 			}
3183 
3184 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3185 			desc++;
3186 			break;
3187 
3188 		default:
3189 			raw_cons += CMP_LEN(cmpl_type);
3190 		}
3191 	}
3192 
3193 	return RTE_ETH_RX_DESC_AVAIL;
3194 }
3195 
3196 static int
3197 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3198 {
3199 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3200 	struct bnxt_tx_ring_info *txr;
3201 	struct bnxt_cp_ring_info *cpr;
3202 	struct rte_mbuf **tx_buf;
3203 	struct tx_pkt_cmpl *txcmp;
3204 	uint32_t cons, cp_cons;
3205 	int rc;
3206 
3207 	if (!txq)
3208 		return -EINVAL;
3209 
3210 	rc = is_bnxt_in_error(txq->bp);
3211 	if (rc)
3212 		return rc;
3213 
3214 	cpr = txq->cp_ring;
3215 	txr = txq->tx_ring;
3216 
3217 	if (offset >= txq->nb_tx_desc)
3218 		return -EINVAL;
3219 
3220 	cons = RING_CMP(cpr->cp_ring_struct, offset);
3221 	txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3222 	cp_cons = cpr->cp_raw_cons;
3223 
3224 	if (cons > cp_cons) {
3225 		if (CMPL_VALID(txcmp, cpr->valid))
3226 			return RTE_ETH_TX_DESC_UNAVAIL;
3227 	} else {
3228 		if (CMPL_VALID(txcmp, !cpr->valid))
3229 			return RTE_ETH_TX_DESC_UNAVAIL;
3230 	}
3231 	tx_buf = &txr->tx_buf_ring[cons];
3232 	if (*tx_buf == NULL)
3233 		return RTE_ETH_TX_DESC_DONE;
3234 
3235 	return RTE_ETH_TX_DESC_FULL;
3236 }
3237 
3238 int
3239 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3240 		     const struct rte_flow_ops **ops)
3241 {
3242 	struct bnxt *bp = dev->data->dev_private;
3243 	int ret = 0;
3244 
3245 	if (!bp)
3246 		return -EIO;
3247 
3248 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3249 		struct bnxt_representor *vfr = dev->data->dev_private;
3250 		bp = vfr->parent_dev->data->dev_private;
3251 		/* parent is deleted while children are still valid */
3252 		if (!bp) {
3253 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3254 				    dev->data->port_id);
3255 			return -EIO;
3256 		}
3257 	}
3258 
3259 	ret = is_bnxt_in_error(bp);
3260 	if (ret)
3261 		return ret;
3262 
3263 	/* PMD supports thread-safe flow operations.  rte_flow API
3264 	 * functions can avoid mutex for multi-thread safety.
3265 	 */
3266 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3267 
3268 	if (BNXT_TRUFLOW_EN(bp))
3269 		*ops = &bnxt_ulp_rte_flow_ops;
3270 	else
3271 		*ops = &bnxt_flow_ops;
3272 
3273 	return ret;
3274 }
3275 
3276 static const uint32_t *
3277 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3278 {
3279 	static const uint32_t ptypes[] = {
3280 		RTE_PTYPE_L2_ETHER_VLAN,
3281 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3282 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3283 		RTE_PTYPE_L4_ICMP,
3284 		RTE_PTYPE_L4_TCP,
3285 		RTE_PTYPE_L4_UDP,
3286 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3287 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3288 		RTE_PTYPE_INNER_L4_ICMP,
3289 		RTE_PTYPE_INNER_L4_TCP,
3290 		RTE_PTYPE_INNER_L4_UDP,
3291 		RTE_PTYPE_UNKNOWN
3292 	};
3293 
3294 	if (!dev->rx_pkt_burst)
3295 		return NULL;
3296 
3297 	return ptypes;
3298 }
3299 
3300 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3301 			 int reg_win)
3302 {
3303 	uint32_t reg_base = *reg_arr & 0xfffff000;
3304 	uint32_t win_off;
3305 	int i;
3306 
3307 	for (i = 0; i < count; i++) {
3308 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3309 			return -ERANGE;
3310 	}
3311 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3312 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3313 	return 0;
3314 }
3315 
3316 static int bnxt_map_ptp_regs(struct bnxt *bp)
3317 {
3318 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3319 	uint32_t *reg_arr;
3320 	int rc, i;
3321 
3322 	reg_arr = ptp->rx_regs;
3323 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3324 	if (rc)
3325 		return rc;
3326 
3327 	reg_arr = ptp->tx_regs;
3328 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3329 	if (rc)
3330 		return rc;
3331 
3332 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3333 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3334 
3335 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3336 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3337 
3338 	return 0;
3339 }
3340 
3341 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3342 {
3343 	rte_write32(0, (uint8_t *)bp->bar0 +
3344 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3345 	rte_write32(0, (uint8_t *)bp->bar0 +
3346 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3347 }
3348 
3349 static uint64_t bnxt_cc_read(struct bnxt *bp)
3350 {
3351 	uint64_t ns;
3352 
3353 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3354 			      BNXT_GRCPF_REG_SYNC_TIME));
3355 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3356 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3357 	return ns;
3358 }
3359 
3360 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3361 {
3362 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3363 	uint32_t fifo;
3364 
3365 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3366 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3367 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3368 		return -EAGAIN;
3369 
3370 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3371 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3372 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3373 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3374 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3375 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3376 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3377 
3378 	return 0;
3379 }
3380 
3381 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3382 {
3383 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3384 	struct bnxt_pf_info *pf = bp->pf;
3385 	uint16_t port_id;
3386 	uint32_t fifo;
3387 
3388 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3389 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3390 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3391 		return -EAGAIN;
3392 
3393 	port_id = pf->port_id;
3394 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3395 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3396 
3397 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3398 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3399 	if (fifo & BNXT_PTP_RX_FIFO_PENDING) {
3400 /*		bnxt_clr_rx_ts(bp);	  TBD  */
3401 		return -EBUSY;
3402 	}
3403 
3404 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3405 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3406 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3407 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3408 
3409 	return 0;
3410 }
3411 
3412 static int
3413 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3414 {
3415 	uint64_t ns;
3416 	struct bnxt *bp = dev->data->dev_private;
3417 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3418 
3419 	if (!ptp)
3420 		return -ENOTSUP;
3421 
3422 	ns = rte_timespec_to_ns(ts);
3423 	/* Set the timecounters to a new value. */
3424 	ptp->tc.nsec = ns;
3425 	ptp->tx_tstamp_tc.nsec = ns;
3426 	ptp->rx_tstamp_tc.nsec = ns;
3427 
3428 	return 0;
3429 }
3430 
3431 static int
3432 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3433 {
3434 	struct bnxt *bp = dev->data->dev_private;
3435 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3436 	uint64_t ns, systime_cycles = 0;
3437 	int rc = 0;
3438 
3439 	if (!ptp)
3440 		return -ENOTSUP;
3441 
3442 	if (BNXT_CHIP_P5(bp))
3443 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3444 					     &systime_cycles);
3445 	else
3446 		systime_cycles = bnxt_cc_read(bp);
3447 
3448 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3449 	*ts = rte_ns_to_timespec(ns);
3450 
3451 	return rc;
3452 }
3453 static int
3454 bnxt_timesync_enable(struct rte_eth_dev *dev)
3455 {
3456 	struct bnxt *bp = dev->data->dev_private;
3457 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3458 	uint32_t shift = 0;
3459 	int rc;
3460 
3461 	if (!ptp)
3462 		return -ENOTSUP;
3463 
3464 	ptp->rx_filter = 1;
3465 	ptp->tx_tstamp_en = 1;
3466 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3467 
3468 	rc = bnxt_hwrm_ptp_cfg(bp);
3469 	if (rc)
3470 		return rc;
3471 
3472 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3473 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3474 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3475 
3476 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3477 	ptp->tc.cc_shift = shift;
3478 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3479 
3480 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3481 	ptp->rx_tstamp_tc.cc_shift = shift;
3482 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3483 
3484 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3485 	ptp->tx_tstamp_tc.cc_shift = shift;
3486 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3487 
3488 	if (!BNXT_CHIP_P5(bp))
3489 		bnxt_map_ptp_regs(bp);
3490 	else
3491 		rc = bnxt_ptp_start(bp);
3492 
3493 	return rc;
3494 }
3495 
3496 static int
3497 bnxt_timesync_disable(struct rte_eth_dev *dev)
3498 {
3499 	struct bnxt *bp = dev->data->dev_private;
3500 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3501 
3502 	if (!ptp)
3503 		return -ENOTSUP;
3504 
3505 	ptp->rx_filter = 0;
3506 	ptp->tx_tstamp_en = 0;
3507 	ptp->rxctl = 0;
3508 
3509 	bnxt_hwrm_ptp_cfg(bp);
3510 
3511 	if (!BNXT_CHIP_P5(bp))
3512 		bnxt_unmap_ptp_regs(bp);
3513 	else
3514 		bnxt_ptp_stop(bp);
3515 
3516 	return 0;
3517 }
3518 
3519 static int
3520 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3521 				 struct timespec *timestamp,
3522 				 uint32_t flags __rte_unused)
3523 {
3524 	struct bnxt *bp = dev->data->dev_private;
3525 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3526 	uint64_t rx_tstamp_cycles = 0;
3527 	uint64_t ns;
3528 
3529 	if (!ptp)
3530 		return -ENOTSUP;
3531 
3532 	if (BNXT_CHIP_P5(bp))
3533 		rx_tstamp_cycles = ptp->rx_timestamp;
3534 	else
3535 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3536 
3537 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3538 	*timestamp = rte_ns_to_timespec(ns);
3539 	return  0;
3540 }
3541 
3542 static int
3543 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3544 				 struct timespec *timestamp)
3545 {
3546 	struct bnxt *bp = dev->data->dev_private;
3547 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3548 	uint64_t tx_tstamp_cycles = 0;
3549 	uint64_t ns;
3550 	int rc = 0;
3551 
3552 	if (!ptp)
3553 		return -ENOTSUP;
3554 
3555 	if (BNXT_CHIP_P5(bp))
3556 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3557 					     &tx_tstamp_cycles);
3558 	else
3559 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3560 
3561 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3562 	*timestamp = rte_ns_to_timespec(ns);
3563 
3564 	return rc;
3565 }
3566 
3567 static int
3568 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3569 {
3570 	struct bnxt *bp = dev->data->dev_private;
3571 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3572 
3573 	if (!ptp)
3574 		return -ENOTSUP;
3575 
3576 	ptp->tc.nsec += delta;
3577 	ptp->tx_tstamp_tc.nsec += delta;
3578 	ptp->rx_tstamp_tc.nsec += delta;
3579 
3580 	return 0;
3581 }
3582 
3583 static int
3584 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3585 {
3586 	struct bnxt *bp = dev->data->dev_private;
3587 	int rc;
3588 	uint32_t dir_entries;
3589 	uint32_t entry_length;
3590 
3591 	rc = is_bnxt_in_error(bp);
3592 	if (rc)
3593 		return rc;
3594 
3595 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3596 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3597 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3598 
3599 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3600 	if (rc != 0)
3601 		return rc;
3602 
3603 	return dir_entries * entry_length;
3604 }
3605 
3606 static int
3607 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3608 		struct rte_dev_eeprom_info *in_eeprom)
3609 {
3610 	struct bnxt *bp = dev->data->dev_private;
3611 	uint32_t index;
3612 	uint32_t offset;
3613 	int rc;
3614 
3615 	rc = is_bnxt_in_error(bp);
3616 	if (rc)
3617 		return rc;
3618 
3619 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3620 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3621 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3622 		    in_eeprom->offset, in_eeprom->length);
3623 
3624 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3625 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3626 						in_eeprom->data);
3627 
3628 	index = in_eeprom->offset >> 24;
3629 	offset = in_eeprom->offset & 0xffffff;
3630 
3631 	if (index != 0)
3632 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3633 					   in_eeprom->length, in_eeprom->data);
3634 
3635 	return 0;
3636 }
3637 
3638 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3639 {
3640 	switch (dir_type) {
3641 	case BNX_DIR_TYPE_CHIMP_PATCH:
3642 	case BNX_DIR_TYPE_BOOTCODE:
3643 	case BNX_DIR_TYPE_BOOTCODE_2:
3644 	case BNX_DIR_TYPE_APE_FW:
3645 	case BNX_DIR_TYPE_APE_PATCH:
3646 	case BNX_DIR_TYPE_KONG_FW:
3647 	case BNX_DIR_TYPE_KONG_PATCH:
3648 	case BNX_DIR_TYPE_BONO_FW:
3649 	case BNX_DIR_TYPE_BONO_PATCH:
3650 		/* FALLTHROUGH */
3651 		return true;
3652 	}
3653 
3654 	return false;
3655 }
3656 
3657 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3658 {
3659 	switch (dir_type) {
3660 	case BNX_DIR_TYPE_AVS:
3661 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3662 	case BNX_DIR_TYPE_PCIE:
3663 	case BNX_DIR_TYPE_TSCF_UCODE:
3664 	case BNX_DIR_TYPE_EXT_PHY:
3665 	case BNX_DIR_TYPE_CCM:
3666 	case BNX_DIR_TYPE_ISCSI_BOOT:
3667 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3668 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3669 		/* FALLTHROUGH */
3670 		return true;
3671 	}
3672 
3673 	return false;
3674 }
3675 
3676 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3677 {
3678 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3679 		bnxt_dir_type_is_other_exec_format(dir_type);
3680 }
3681 
3682 static int
3683 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3684 		struct rte_dev_eeprom_info *in_eeprom)
3685 {
3686 	struct bnxt *bp = dev->data->dev_private;
3687 	uint8_t index, dir_op;
3688 	uint16_t type, ext, ordinal, attr;
3689 	int rc;
3690 
3691 	rc = is_bnxt_in_error(bp);
3692 	if (rc)
3693 		return rc;
3694 
3695 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3696 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3697 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3698 		    in_eeprom->offset, in_eeprom->length);
3699 
3700 	if (!BNXT_PF(bp)) {
3701 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3702 		return -EINVAL;
3703 	}
3704 
3705 	type = in_eeprom->magic >> 16;
3706 
3707 	if (type == 0xffff) { /* special value for directory operations */
3708 		index = in_eeprom->magic & 0xff;
3709 		dir_op = in_eeprom->magic >> 8;
3710 		if (index == 0)
3711 			return -EINVAL;
3712 		switch (dir_op) {
3713 		case 0x0e: /* erase */
3714 			if (in_eeprom->offset != ~in_eeprom->magic)
3715 				return -EINVAL;
3716 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3717 		default:
3718 			return -EINVAL;
3719 		}
3720 	}
3721 
3722 	/* Create or re-write an NVM item: */
3723 	if (bnxt_dir_type_is_executable(type) == true)
3724 		return -EOPNOTSUPP;
3725 	ext = in_eeprom->magic & 0xffff;
3726 	ordinal = in_eeprom->offset >> 16;
3727 	attr = in_eeprom->offset & 0xffff;
3728 
3729 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3730 				     in_eeprom->data, in_eeprom->length);
3731 }
3732 
3733 /*
3734  * Initialization
3735  */
3736 
3737 static const struct eth_dev_ops bnxt_dev_ops = {
3738 	.dev_infos_get = bnxt_dev_info_get_op,
3739 	.dev_close = bnxt_dev_close_op,
3740 	.dev_configure = bnxt_dev_configure_op,
3741 	.dev_start = bnxt_dev_start_op,
3742 	.dev_stop = bnxt_dev_stop_op,
3743 	.dev_set_link_up = bnxt_dev_set_link_up_op,
3744 	.dev_set_link_down = bnxt_dev_set_link_down_op,
3745 	.stats_get = bnxt_stats_get_op,
3746 	.stats_reset = bnxt_stats_reset_op,
3747 	.rx_queue_setup = bnxt_rx_queue_setup_op,
3748 	.rx_queue_release = bnxt_rx_queue_release_op,
3749 	.tx_queue_setup = bnxt_tx_queue_setup_op,
3750 	.tx_queue_release = bnxt_tx_queue_release_op,
3751 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
3752 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
3753 	.reta_update = bnxt_reta_update_op,
3754 	.reta_query = bnxt_reta_query_op,
3755 	.rss_hash_update = bnxt_rss_hash_update_op,
3756 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
3757 	.link_update = bnxt_link_update_op,
3758 	.promiscuous_enable = bnxt_promiscuous_enable_op,
3759 	.promiscuous_disable = bnxt_promiscuous_disable_op,
3760 	.allmulticast_enable = bnxt_allmulticast_enable_op,
3761 	.allmulticast_disable = bnxt_allmulticast_disable_op,
3762 	.mac_addr_add = bnxt_mac_addr_add_op,
3763 	.mac_addr_remove = bnxt_mac_addr_remove_op,
3764 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
3765 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
3766 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
3767 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
3768 	.vlan_filter_set = bnxt_vlan_filter_set_op,
3769 	.vlan_offload_set = bnxt_vlan_offload_set_op,
3770 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
3771 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
3772 	.mtu_set = bnxt_mtu_set_op,
3773 	.mac_addr_set = bnxt_set_default_mac_addr_op,
3774 	.xstats_get = bnxt_dev_xstats_get_op,
3775 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
3776 	.xstats_reset = bnxt_dev_xstats_reset_op,
3777 	.fw_version_get = bnxt_fw_version_get,
3778 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
3779 	.rxq_info_get = bnxt_rxq_info_get_op,
3780 	.txq_info_get = bnxt_txq_info_get_op,
3781 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
3782 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
3783 	.dev_led_on = bnxt_dev_led_on_op,
3784 	.dev_led_off = bnxt_dev_led_off_op,
3785 	.rx_queue_start = bnxt_rx_queue_start,
3786 	.rx_queue_stop = bnxt_rx_queue_stop,
3787 	.tx_queue_start = bnxt_tx_queue_start,
3788 	.tx_queue_stop = bnxt_tx_queue_stop,
3789 	.flow_ops_get = bnxt_flow_ops_get_op,
3790 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
3791 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
3792 	.get_eeprom           = bnxt_get_eeprom_op,
3793 	.set_eeprom           = bnxt_set_eeprom_op,
3794 	.timesync_enable      = bnxt_timesync_enable,
3795 	.timesync_disable     = bnxt_timesync_disable,
3796 	.timesync_read_time   = bnxt_timesync_read_time,
3797 	.timesync_write_time   = bnxt_timesync_write_time,
3798 	.timesync_adjust_time = bnxt_timesync_adjust_time,
3799 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
3800 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
3801 };
3802 
3803 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
3804 {
3805 	uint32_t offset;
3806 
3807 	/* Only pre-map the reset GRC registers using window 3 */
3808 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
3809 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
3810 
3811 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
3812 
3813 	return offset;
3814 }
3815 
3816 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
3817 {
3818 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3819 	uint32_t reg_base = 0xffffffff;
3820 	int i;
3821 
3822 	/* Only pre-map the monitoring GRC registers using window 2 */
3823 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
3824 		uint32_t reg = info->status_regs[i];
3825 
3826 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
3827 			continue;
3828 
3829 		if (reg_base == 0xffffffff)
3830 			reg_base = reg & 0xfffff000;
3831 		if ((reg & 0xfffff000) != reg_base)
3832 			return -ERANGE;
3833 
3834 		/* Use mask 0xffc as the Lower 2 bits indicates
3835 		 * address space location
3836 		 */
3837 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
3838 						(reg & 0xffc);
3839 	}
3840 
3841 	if (reg_base == 0xffffffff)
3842 		return 0;
3843 
3844 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
3845 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
3846 
3847 	return 0;
3848 }
3849 
3850 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
3851 {
3852 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3853 	uint32_t delay = info->delay_after_reset[index];
3854 	uint32_t val = info->reset_reg_val[index];
3855 	uint32_t reg = info->reset_reg[index];
3856 	uint32_t type, offset;
3857 	int ret;
3858 
3859 	type = BNXT_FW_STATUS_REG_TYPE(reg);
3860 	offset = BNXT_FW_STATUS_REG_OFF(reg);
3861 
3862 	switch (type) {
3863 	case BNXT_FW_STATUS_REG_TYPE_CFG:
3864 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
3865 		if (ret < 0) {
3866 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
3867 				    val, offset);
3868 			return;
3869 		}
3870 		break;
3871 	case BNXT_FW_STATUS_REG_TYPE_GRC:
3872 		offset = bnxt_map_reset_regs(bp, offset);
3873 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3874 		break;
3875 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
3876 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3877 		break;
3878 	}
3879 	/* wait on a specific interval of time until core reset is complete */
3880 	if (delay)
3881 		rte_delay_ms(delay);
3882 }
3883 
3884 static void bnxt_dev_cleanup(struct bnxt *bp)
3885 {
3886 	bp->eth_dev->data->dev_link.link_status = 0;
3887 	bp->link_info->link_up = 0;
3888 	if (bp->eth_dev->data->dev_started)
3889 		bnxt_dev_stop(bp->eth_dev);
3890 
3891 	bnxt_uninit_resources(bp, true);
3892 }
3893 
3894 static int
3895 bnxt_check_fw_reset_done(struct bnxt *bp)
3896 {
3897 	int timeout = bp->fw_reset_max_msecs;
3898 	uint16_t val = 0;
3899 	int rc;
3900 
3901 	do {
3902 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
3903 		if (rc < 0) {
3904 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
3905 			return rc;
3906 		}
3907 		if (val != 0xffff)
3908 			break;
3909 		rte_delay_ms(1);
3910 	} while (timeout--);
3911 
3912 	if (val == 0xffff) {
3913 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
3914 		return -1;
3915 	}
3916 
3917 	return 0;
3918 }
3919 
3920 static int bnxt_restore_vlan_filters(struct bnxt *bp)
3921 {
3922 	struct rte_eth_dev *dev = bp->eth_dev;
3923 	struct rte_vlan_filter_conf *vfc;
3924 	int vidx, vbit, rc;
3925 	uint16_t vlan_id;
3926 
3927 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
3928 		vfc = &dev->data->vlan_filter_conf;
3929 		vidx = vlan_id / 64;
3930 		vbit = vlan_id % 64;
3931 
3932 		/* Each bit corresponds to a VLAN id */
3933 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
3934 			rc = bnxt_add_vlan_filter(bp, vlan_id);
3935 			if (rc)
3936 				return rc;
3937 		}
3938 	}
3939 
3940 	return 0;
3941 }
3942 
3943 static int bnxt_restore_mac_filters(struct bnxt *bp)
3944 {
3945 	struct rte_eth_dev *dev = bp->eth_dev;
3946 	struct rte_eth_dev_info dev_info;
3947 	struct rte_ether_addr *addr;
3948 	uint64_t pool_mask;
3949 	uint32_t pool = 0;
3950 	uint16_t i;
3951 	int rc;
3952 
3953 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
3954 		return 0;
3955 
3956 	rc = bnxt_dev_info_get_op(dev, &dev_info);
3957 	if (rc)
3958 		return rc;
3959 
3960 	/* replay MAC address configuration */
3961 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
3962 		addr = &dev->data->mac_addrs[i];
3963 
3964 		/* skip zero address */
3965 		if (rte_is_zero_ether_addr(addr))
3966 			continue;
3967 
3968 		pool = 0;
3969 		pool_mask = dev->data->mac_pool_sel[i];
3970 
3971 		do {
3972 			if (pool_mask & 1ULL) {
3973 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
3974 				if (rc)
3975 					return rc;
3976 			}
3977 			pool_mask >>= 1;
3978 			pool++;
3979 		} while (pool_mask);
3980 	}
3981 
3982 	return 0;
3983 }
3984 
3985 static int bnxt_restore_filters(struct bnxt *bp)
3986 {
3987 	struct rte_eth_dev *dev = bp->eth_dev;
3988 	int ret = 0;
3989 
3990 	if (dev->data->all_multicast) {
3991 		ret = bnxt_allmulticast_enable_op(dev);
3992 		if (ret)
3993 			return ret;
3994 	}
3995 	if (dev->data->promiscuous) {
3996 		ret = bnxt_promiscuous_enable_op(dev);
3997 		if (ret)
3998 			return ret;
3999 	}
4000 
4001 	ret = bnxt_restore_mac_filters(bp);
4002 	if (ret)
4003 		return ret;
4004 
4005 	ret = bnxt_restore_vlan_filters(bp);
4006 	/* TODO restore other filters as well */
4007 	return ret;
4008 }
4009 
4010 static int bnxt_check_fw_ready(struct bnxt *bp)
4011 {
4012 	int timeout = bp->fw_reset_max_msecs;
4013 	int rc = 0;
4014 
4015 	do {
4016 		rc = bnxt_hwrm_poll_ver_get(bp);
4017 		if (rc == 0)
4018 			break;
4019 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4020 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4021 	} while (rc && timeout > 0);
4022 
4023 	if (rc)
4024 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4025 
4026 	return rc;
4027 }
4028 
4029 static void bnxt_dev_recover(void *arg)
4030 {
4031 	struct bnxt *bp = arg;
4032 	int rc = 0;
4033 
4034 	pthread_mutex_lock(&bp->err_recovery_lock);
4035 
4036 	if (!bp->fw_reset_min_msecs) {
4037 		rc = bnxt_check_fw_reset_done(bp);
4038 		if (rc)
4039 			goto err;
4040 	}
4041 
4042 	/* Clear Error flag so that device re-init should happen */
4043 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4044 
4045 	rc = bnxt_check_fw_ready(bp);
4046 	if (rc)
4047 		goto err;
4048 
4049 	rc = bnxt_init_resources(bp, true);
4050 	if (rc) {
4051 		PMD_DRV_LOG(ERR,
4052 			    "Failed to initialize resources after reset\n");
4053 		goto err;
4054 	}
4055 	/* clear reset flag as the device is initialized now */
4056 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4057 
4058 	rc = bnxt_dev_start_op(bp->eth_dev);
4059 	if (rc) {
4060 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4061 		goto err_start;
4062 	}
4063 
4064 	rc = bnxt_restore_filters(bp);
4065 	if (rc)
4066 		goto err_start;
4067 
4068 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4069 	pthread_mutex_unlock(&bp->err_recovery_lock);
4070 
4071 	return;
4072 err_start:
4073 	bnxt_dev_stop(bp->eth_dev);
4074 err:
4075 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4076 	bnxt_uninit_resources(bp, false);
4077 	pthread_mutex_unlock(&bp->err_recovery_lock);
4078 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4079 }
4080 
4081 void bnxt_dev_reset_and_resume(void *arg)
4082 {
4083 	struct bnxt *bp = arg;
4084 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4085 	uint16_t val = 0;
4086 	int rc;
4087 
4088 	bnxt_dev_cleanup(bp);
4089 
4090 	bnxt_wait_for_device_shutdown(bp);
4091 
4092 	/* During some fatal firmware error conditions, the PCI config space
4093 	 * register 0x2e which normally contains the subsystem ID will become
4094 	 * 0xffff. This register will revert back to the normal value after
4095 	 * the chip has completed core reset. If we detect this condition,
4096 	 * we can poll this config register immediately for the value to revert.
4097 	 */
4098 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4099 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4100 		if (rc < 0) {
4101 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4102 			return;
4103 		}
4104 		if (val == 0xffff) {
4105 			bp->fw_reset_min_msecs = 0;
4106 			us = 1;
4107 		}
4108 	}
4109 
4110 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4111 	if (rc)
4112 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4113 }
4114 
4115 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4116 {
4117 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4118 	uint32_t reg = info->status_regs[index];
4119 	uint32_t type, offset, val = 0;
4120 
4121 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4122 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4123 
4124 	switch (type) {
4125 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4126 		rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4127 		break;
4128 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4129 		offset = info->mapped_status_regs[index];
4130 		/* FALLTHROUGH */
4131 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4132 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4133 				       offset));
4134 		break;
4135 	}
4136 
4137 	return val;
4138 }
4139 
4140 static int bnxt_fw_reset_all(struct bnxt *bp)
4141 {
4142 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4143 	uint32_t i;
4144 	int rc = 0;
4145 
4146 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4147 		/* Reset through master function driver */
4148 		for (i = 0; i < info->reg_array_cnt; i++)
4149 			bnxt_write_fw_reset_reg(bp, i);
4150 		/* Wait for time specified by FW after triggering reset */
4151 		rte_delay_ms(info->master_func_wait_period_after_reset);
4152 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4153 		/* Reset with the help of Kong processor */
4154 		rc = bnxt_hwrm_fw_reset(bp);
4155 		if (rc)
4156 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4157 	}
4158 
4159 	return rc;
4160 }
4161 
4162 static void bnxt_fw_reset_cb(void *arg)
4163 {
4164 	struct bnxt *bp = arg;
4165 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4166 	int rc = 0;
4167 
4168 	/* Only Master function can do FW reset */
4169 	if (bnxt_is_master_func(bp) &&
4170 	    bnxt_is_recovery_enabled(bp)) {
4171 		rc = bnxt_fw_reset_all(bp);
4172 		if (rc) {
4173 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4174 			return;
4175 		}
4176 	}
4177 
4178 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4179 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4180 	 * (including MASTER FUNC). After receiving this Async, all the active
4181 	 * drivers should treat this case as FW initiated recovery
4182 	 */
4183 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4184 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4185 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4186 
4187 		/* To recover from error */
4188 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4189 				  (void *)bp);
4190 	}
4191 }
4192 
4193 /* Driver should poll FW heartbeat, reset_counter with the frequency
4194  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4195  * When the driver detects heartbeat stop or change in reset_counter,
4196  * it has to trigger a reset to recover from the error condition.
4197  * A “master PF” is the function who will have the privilege to
4198  * initiate the chimp reset. The master PF will be elected by the
4199  * firmware and will be notified through async message.
4200  */
4201 static void bnxt_check_fw_health(void *arg)
4202 {
4203 	struct bnxt *bp = arg;
4204 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4205 	uint32_t val = 0, wait_msec;
4206 
4207 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4208 	    is_bnxt_in_error(bp))
4209 		return;
4210 
4211 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4212 	if (val == info->last_heart_beat)
4213 		goto reset;
4214 
4215 	info->last_heart_beat = val;
4216 
4217 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4218 	if (val != info->last_reset_counter)
4219 		goto reset;
4220 
4221 	info->last_reset_counter = val;
4222 
4223 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4224 			  bnxt_check_fw_health, (void *)bp);
4225 
4226 	return;
4227 reset:
4228 	/* Stop DMA to/from device */
4229 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4230 	bp->flags |= BNXT_FLAG_FW_RESET;
4231 
4232 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4233 
4234 	if (bnxt_is_master_func(bp))
4235 		wait_msec = info->master_func_wait_period;
4236 	else
4237 		wait_msec = info->normal_func_wait_period;
4238 
4239 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4240 			  bnxt_fw_reset_cb, (void *)bp);
4241 }
4242 
4243 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4244 {
4245 	uint32_t polling_freq;
4246 
4247 	pthread_mutex_lock(&bp->health_check_lock);
4248 
4249 	if (!bnxt_is_recovery_enabled(bp))
4250 		goto done;
4251 
4252 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4253 		goto done;
4254 
4255 	polling_freq = bp->recovery_info->driver_polling_freq;
4256 
4257 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4258 			  bnxt_check_fw_health, (void *)bp);
4259 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4260 
4261 done:
4262 	pthread_mutex_unlock(&bp->health_check_lock);
4263 }
4264 
4265 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4266 {
4267 	if (!bnxt_is_recovery_enabled(bp))
4268 		return;
4269 
4270 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4271 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4272 }
4273 
4274 static bool bnxt_vf_pciid(uint16_t device_id)
4275 {
4276 	switch (device_id) {
4277 	case BROADCOM_DEV_ID_57304_VF:
4278 	case BROADCOM_DEV_ID_57406_VF:
4279 	case BROADCOM_DEV_ID_5731X_VF:
4280 	case BROADCOM_DEV_ID_5741X_VF:
4281 	case BROADCOM_DEV_ID_57414_VF:
4282 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4283 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4284 	case BROADCOM_DEV_ID_58802_VF:
4285 	case BROADCOM_DEV_ID_57500_VF1:
4286 	case BROADCOM_DEV_ID_57500_VF2:
4287 	case BROADCOM_DEV_ID_58818_VF:
4288 		/* FALLTHROUGH */
4289 		return true;
4290 	default:
4291 		return false;
4292 	}
4293 }
4294 
4295 /* Phase 5 device */
4296 static bool bnxt_p5_device(uint16_t device_id)
4297 {
4298 	switch (device_id) {
4299 	case BROADCOM_DEV_ID_57508:
4300 	case BROADCOM_DEV_ID_57504:
4301 	case BROADCOM_DEV_ID_57502:
4302 	case BROADCOM_DEV_ID_57508_MF1:
4303 	case BROADCOM_DEV_ID_57504_MF1:
4304 	case BROADCOM_DEV_ID_57502_MF1:
4305 	case BROADCOM_DEV_ID_57508_MF2:
4306 	case BROADCOM_DEV_ID_57504_MF2:
4307 	case BROADCOM_DEV_ID_57502_MF2:
4308 	case BROADCOM_DEV_ID_57500_VF1:
4309 	case BROADCOM_DEV_ID_57500_VF2:
4310 	case BROADCOM_DEV_ID_58812:
4311 	case BROADCOM_DEV_ID_58814:
4312 	case BROADCOM_DEV_ID_58818:
4313 	case BROADCOM_DEV_ID_58818_VF:
4314 		/* FALLTHROUGH */
4315 		return true;
4316 	default:
4317 		return false;
4318 	}
4319 }
4320 
4321 bool bnxt_stratus_device(struct bnxt *bp)
4322 {
4323 	uint16_t device_id = bp->pdev->id.device_id;
4324 
4325 	switch (device_id) {
4326 	case BROADCOM_DEV_ID_STRATUS_NIC:
4327 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4328 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4329 		/* FALLTHROUGH */
4330 		return true;
4331 	default:
4332 		return false;
4333 	}
4334 }
4335 
4336 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4337 {
4338 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4339 	struct bnxt *bp = eth_dev->data->dev_private;
4340 
4341 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4342 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4343 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4344 	if (!bp->bar0 || !bp->doorbell_base) {
4345 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4346 		return -ENODEV;
4347 	}
4348 
4349 	bp->eth_dev = eth_dev;
4350 	bp->pdev = pci_dev;
4351 
4352 	return 0;
4353 }
4354 
4355 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4356 				  struct bnxt_ctx_pg_info *ctx_pg,
4357 				  uint32_t mem_size,
4358 				  const char *suffix,
4359 				  uint16_t idx)
4360 {
4361 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4362 	const struct rte_memzone *mz = NULL;
4363 	char mz_name[RTE_MEMZONE_NAMESIZE];
4364 	rte_iova_t mz_phys_addr;
4365 	uint64_t valid_bits = 0;
4366 	uint32_t sz;
4367 	int i;
4368 
4369 	if (!mem_size)
4370 		return 0;
4371 
4372 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4373 			 BNXT_PAGE_SIZE;
4374 	rmem->page_size = BNXT_PAGE_SIZE;
4375 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4376 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4377 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4378 
4379 	valid_bits = PTU_PTE_VALID;
4380 
4381 	if (rmem->nr_pages > 1) {
4382 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4383 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4384 			 suffix, idx, bp->eth_dev->data->port_id);
4385 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4386 		mz = rte_memzone_lookup(mz_name);
4387 		if (!mz) {
4388 			mz = rte_memzone_reserve_aligned(mz_name,
4389 						rmem->nr_pages * 8,
4390 						SOCKET_ID_ANY,
4391 						RTE_MEMZONE_2MB |
4392 						RTE_MEMZONE_SIZE_HINT_ONLY |
4393 						RTE_MEMZONE_IOVA_CONTIG,
4394 						BNXT_PAGE_SIZE);
4395 			if (mz == NULL)
4396 				return -ENOMEM;
4397 		}
4398 
4399 		memset(mz->addr, 0, mz->len);
4400 		mz_phys_addr = mz->iova;
4401 
4402 		rmem->pg_tbl = mz->addr;
4403 		rmem->pg_tbl_map = mz_phys_addr;
4404 		rmem->pg_tbl_mz = mz;
4405 	}
4406 
4407 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4408 		 suffix, idx, bp->eth_dev->data->port_id);
4409 	mz = rte_memzone_lookup(mz_name);
4410 	if (!mz) {
4411 		mz = rte_memzone_reserve_aligned(mz_name,
4412 						 mem_size,
4413 						 SOCKET_ID_ANY,
4414 						 RTE_MEMZONE_1GB |
4415 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4416 						 RTE_MEMZONE_IOVA_CONTIG,
4417 						 BNXT_PAGE_SIZE);
4418 		if (mz == NULL)
4419 			return -ENOMEM;
4420 	}
4421 
4422 	memset(mz->addr, 0, mz->len);
4423 	mz_phys_addr = mz->iova;
4424 
4425 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4426 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4427 		rmem->dma_arr[i] = mz_phys_addr + sz;
4428 
4429 		if (rmem->nr_pages > 1) {
4430 			if (i == rmem->nr_pages - 2 &&
4431 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4432 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4433 			else if (i == rmem->nr_pages - 1 &&
4434 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4435 				valid_bits |= PTU_PTE_LAST;
4436 
4437 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4438 							   valid_bits);
4439 		}
4440 	}
4441 
4442 	rmem->mz = mz;
4443 	if (rmem->vmem_size)
4444 		rmem->vmem = (void **)mz->addr;
4445 	rmem->dma_arr[0] = mz_phys_addr;
4446 	return 0;
4447 }
4448 
4449 static void bnxt_free_ctx_mem(struct bnxt *bp)
4450 {
4451 	int i;
4452 
4453 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4454 		return;
4455 
4456 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4457 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4458 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4459 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4460 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4461 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4462 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4463 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4464 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4465 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4466 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4467 
4468 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4469 		if (bp->ctx->tqm_mem[i])
4470 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4471 	}
4472 
4473 	rte_free(bp->ctx);
4474 	bp->ctx = NULL;
4475 }
4476 
4477 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4478 
4479 #define min_t(type, x, y) ({                    \
4480 	type __min1 = (x);                      \
4481 	type __min2 = (y);                      \
4482 	__min1 < __min2 ? __min1 : __min2; })
4483 
4484 #define max_t(type, x, y) ({                    \
4485 	type __max1 = (x);                      \
4486 	type __max2 = (y);                      \
4487 	__max1 > __max2 ? __max1 : __max2; })
4488 
4489 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4490 
4491 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4492 {
4493 	struct bnxt_ctx_pg_info *ctx_pg;
4494 	struct bnxt_ctx_mem_info *ctx;
4495 	uint32_t mem_size, ena, entries;
4496 	uint32_t entries_sp, min;
4497 	int i, rc;
4498 
4499 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4500 	if (rc) {
4501 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4502 		return rc;
4503 	}
4504 	ctx = bp->ctx;
4505 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4506 		return 0;
4507 
4508 	ctx_pg = &ctx->qp_mem;
4509 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4510 	if (ctx->qp_entry_size) {
4511 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4512 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4513 		if (rc)
4514 			return rc;
4515 	}
4516 
4517 	ctx_pg = &ctx->srq_mem;
4518 	ctx_pg->entries = ctx->srq_max_l2_entries;
4519 	if (ctx->srq_entry_size) {
4520 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4521 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4522 		if (rc)
4523 			return rc;
4524 	}
4525 
4526 	ctx_pg = &ctx->cq_mem;
4527 	ctx_pg->entries = ctx->cq_max_l2_entries;
4528 	if (ctx->cq_entry_size) {
4529 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4530 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4531 		if (rc)
4532 			return rc;
4533 	}
4534 
4535 	ctx_pg = &ctx->vnic_mem;
4536 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4537 		ctx->vnic_max_ring_table_entries;
4538 	if (ctx->vnic_entry_size) {
4539 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4540 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4541 		if (rc)
4542 			return rc;
4543 	}
4544 
4545 	ctx_pg = &ctx->stat_mem;
4546 	ctx_pg->entries = ctx->stat_max_entries;
4547 	if (ctx->stat_entry_size) {
4548 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4549 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4550 		if (rc)
4551 			return rc;
4552 	}
4553 
4554 	min = ctx->tqm_min_entries_per_ring;
4555 
4556 	entries_sp = ctx->qp_max_l2_entries +
4557 		     ctx->vnic_max_vnic_entries +
4558 		     2 * ctx->qp_min_qp1_entries + min;
4559 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4560 
4561 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4562 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4563 	entries = clamp_t(uint32_t, entries, min,
4564 			  ctx->tqm_max_entries_per_ring);
4565 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4566 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4567 		 * i > 8 is other ext rings.
4568 		 */
4569 		ctx_pg = ctx->tqm_mem[i];
4570 		ctx_pg->entries = i ? entries : entries_sp;
4571 		if (ctx->tqm_entry_size) {
4572 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4573 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4574 						    "tqm_mem", i);
4575 			if (rc)
4576 				return rc;
4577 		}
4578 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4579 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4580 		else
4581 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4582 	}
4583 
4584 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4585 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4586 	if (rc)
4587 		PMD_DRV_LOG(ERR,
4588 			    "Failed to configure context mem: rc = %d\n", rc);
4589 	else
4590 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4591 
4592 	return rc;
4593 }
4594 
4595 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4596 {
4597 	struct rte_pci_device *pci_dev = bp->pdev;
4598 	char mz_name[RTE_MEMZONE_NAMESIZE];
4599 	const struct rte_memzone *mz = NULL;
4600 	uint32_t total_alloc_len;
4601 	rte_iova_t mz_phys_addr;
4602 
4603 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4604 		return 0;
4605 
4606 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4607 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4608 		 pci_dev->addr.bus, pci_dev->addr.devid,
4609 		 pci_dev->addr.function, "rx_port_stats");
4610 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4611 	mz = rte_memzone_lookup(mz_name);
4612 	total_alloc_len =
4613 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4614 				       sizeof(struct rx_port_stats_ext) + 512);
4615 	if (!mz) {
4616 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4617 					 SOCKET_ID_ANY,
4618 					 RTE_MEMZONE_2MB |
4619 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4620 					 RTE_MEMZONE_IOVA_CONTIG);
4621 		if (mz == NULL)
4622 			return -ENOMEM;
4623 	}
4624 	memset(mz->addr, 0, mz->len);
4625 	mz_phys_addr = mz->iova;
4626 
4627 	bp->rx_mem_zone = (const void *)mz;
4628 	bp->hw_rx_port_stats = mz->addr;
4629 	bp->hw_rx_port_stats_map = mz_phys_addr;
4630 
4631 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4632 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4633 		 pci_dev->addr.bus, pci_dev->addr.devid,
4634 		 pci_dev->addr.function, "tx_port_stats");
4635 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4636 	mz = rte_memzone_lookup(mz_name);
4637 	total_alloc_len =
4638 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4639 				       sizeof(struct tx_port_stats_ext) + 512);
4640 	if (!mz) {
4641 		mz = rte_memzone_reserve(mz_name,
4642 					 total_alloc_len,
4643 					 SOCKET_ID_ANY,
4644 					 RTE_MEMZONE_2MB |
4645 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4646 					 RTE_MEMZONE_IOVA_CONTIG);
4647 		if (mz == NULL)
4648 			return -ENOMEM;
4649 	}
4650 	memset(mz->addr, 0, mz->len);
4651 	mz_phys_addr = mz->iova;
4652 
4653 	bp->tx_mem_zone = (const void *)mz;
4654 	bp->hw_tx_port_stats = mz->addr;
4655 	bp->hw_tx_port_stats_map = mz_phys_addr;
4656 	bp->flags |= BNXT_FLAG_PORT_STATS;
4657 
4658 	/* Display extended statistics if FW supports it */
4659 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4660 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4661 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4662 		return 0;
4663 
4664 	bp->hw_rx_port_stats_ext = (void *)
4665 		((uint8_t *)bp->hw_rx_port_stats +
4666 		 sizeof(struct rx_port_stats));
4667 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4668 		sizeof(struct rx_port_stats);
4669 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4670 
4671 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4672 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4673 		bp->hw_tx_port_stats_ext = (void *)
4674 			((uint8_t *)bp->hw_tx_port_stats +
4675 			 sizeof(struct tx_port_stats));
4676 		bp->hw_tx_port_stats_ext_map =
4677 			bp->hw_tx_port_stats_map +
4678 			sizeof(struct tx_port_stats);
4679 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4680 	}
4681 
4682 	return 0;
4683 }
4684 
4685 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4686 {
4687 	struct bnxt *bp = eth_dev->data->dev_private;
4688 	int rc = 0;
4689 
4690 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4691 					       RTE_ETHER_ADDR_LEN *
4692 					       bp->max_l2_ctx,
4693 					       0);
4694 	if (eth_dev->data->mac_addrs == NULL) {
4695 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4696 		return -ENOMEM;
4697 	}
4698 
4699 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4700 		if (BNXT_PF(bp))
4701 			return -EINVAL;
4702 
4703 		/* Generate a random MAC address, if none was assigned by PF */
4704 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4705 		bnxt_eth_hw_addr_random(bp->mac_addr);
4706 		PMD_DRV_LOG(INFO,
4707 			    "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4708 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4709 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4710 
4711 		rc = bnxt_hwrm_set_mac(bp);
4712 		if (rc)
4713 			return rc;
4714 	}
4715 
4716 	/* Copy the permanent MAC from the FUNC_QCAPS response */
4717 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
4718 
4719 	return rc;
4720 }
4721 
4722 static int bnxt_restore_dflt_mac(struct bnxt *bp)
4723 {
4724 	int rc = 0;
4725 
4726 	/* MAC is already configured in FW */
4727 	if (BNXT_HAS_DFLT_MAC_SET(bp))
4728 		return 0;
4729 
4730 	/* Restore the old MAC configured */
4731 	rc = bnxt_hwrm_set_mac(bp);
4732 	if (rc)
4733 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
4734 
4735 	return rc;
4736 }
4737 
4738 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
4739 {
4740 	if (!BNXT_PF(bp))
4741 		return;
4742 
4743 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
4744 
4745 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
4746 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
4747 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
4748 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
4749 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
4750 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
4751 }
4752 
4753 uint16_t
4754 bnxt_get_svif(uint16_t port_id, bool func_svif,
4755 	      enum bnxt_ulp_intf_type type)
4756 {
4757 	struct rte_eth_dev *eth_dev;
4758 	struct bnxt *bp;
4759 
4760 	eth_dev = &rte_eth_devices[port_id];
4761 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4762 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4763 		if (!vfr)
4764 			return 0;
4765 
4766 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4767 			return vfr->svif;
4768 
4769 		eth_dev = vfr->parent_dev;
4770 	}
4771 
4772 	bp = eth_dev->data->dev_private;
4773 
4774 	return func_svif ? bp->func_svif : bp->port_svif;
4775 }
4776 
4777 uint16_t
4778 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
4779 {
4780 	struct rte_eth_dev *eth_dev;
4781 	struct bnxt_vnic_info *vnic;
4782 	struct bnxt *bp;
4783 
4784 	eth_dev = &rte_eth_devices[port];
4785 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4786 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4787 		if (!vfr)
4788 			return 0;
4789 
4790 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4791 			return vfr->dflt_vnic_id;
4792 
4793 		eth_dev = vfr->parent_dev;
4794 	}
4795 
4796 	bp = eth_dev->data->dev_private;
4797 
4798 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
4799 
4800 	return vnic->fw_vnic_id;
4801 }
4802 
4803 uint16_t
4804 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
4805 {
4806 	struct rte_eth_dev *eth_dev;
4807 	struct bnxt *bp;
4808 
4809 	eth_dev = &rte_eth_devices[port];
4810 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4811 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4812 		if (!vfr)
4813 			return 0;
4814 
4815 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4816 			return vfr->fw_fid;
4817 
4818 		eth_dev = vfr->parent_dev;
4819 	}
4820 
4821 	bp = eth_dev->data->dev_private;
4822 
4823 	return bp->fw_fid;
4824 }
4825 
4826 enum bnxt_ulp_intf_type
4827 bnxt_get_interface_type(uint16_t port)
4828 {
4829 	struct rte_eth_dev *eth_dev;
4830 	struct bnxt *bp;
4831 
4832 	eth_dev = &rte_eth_devices[port];
4833 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
4834 		return BNXT_ULP_INTF_TYPE_VF_REP;
4835 
4836 	bp = eth_dev->data->dev_private;
4837 	if (BNXT_PF(bp))
4838 		return BNXT_ULP_INTF_TYPE_PF;
4839 	else if (BNXT_VF_IS_TRUSTED(bp))
4840 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
4841 	else if (BNXT_VF(bp))
4842 		return BNXT_ULP_INTF_TYPE_VF;
4843 
4844 	return BNXT_ULP_INTF_TYPE_INVALID;
4845 }
4846 
4847 uint16_t
4848 bnxt_get_phy_port_id(uint16_t port_id)
4849 {
4850 	struct bnxt_representor *vfr;
4851 	struct rte_eth_dev *eth_dev;
4852 	struct bnxt *bp;
4853 
4854 	eth_dev = &rte_eth_devices[port_id];
4855 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4856 		vfr = eth_dev->data->dev_private;
4857 		if (!vfr)
4858 			return 0;
4859 
4860 		eth_dev = vfr->parent_dev;
4861 	}
4862 
4863 	bp = eth_dev->data->dev_private;
4864 
4865 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
4866 }
4867 
4868 uint16_t
4869 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
4870 {
4871 	struct rte_eth_dev *eth_dev;
4872 	struct bnxt *bp;
4873 
4874 	eth_dev = &rte_eth_devices[port_id];
4875 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4876 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4877 		if (!vfr)
4878 			return 0;
4879 
4880 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4881 			return vfr->fw_fid - 1;
4882 
4883 		eth_dev = vfr->parent_dev;
4884 	}
4885 
4886 	bp = eth_dev->data->dev_private;
4887 
4888 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
4889 }
4890 
4891 uint16_t
4892 bnxt_get_vport(uint16_t port_id)
4893 {
4894 	return (1 << bnxt_get_phy_port_id(port_id));
4895 }
4896 
4897 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
4898 {
4899 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4900 
4901 	if (info) {
4902 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
4903 			memset(info, 0, sizeof(*info));
4904 		return;
4905 	}
4906 
4907 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
4908 		return;
4909 
4910 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4911 			   sizeof(*info), 0);
4912 	if (!info)
4913 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4914 
4915 	bp->recovery_info = info;
4916 }
4917 
4918 static void bnxt_check_fw_status(struct bnxt *bp)
4919 {
4920 	uint32_t fw_status;
4921 
4922 	if (!(bp->recovery_info &&
4923 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
4924 		return;
4925 
4926 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
4927 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
4928 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
4929 			    fw_status);
4930 }
4931 
4932 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
4933 {
4934 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4935 	uint32_t status_loc;
4936 	uint32_t sig_ver;
4937 
4938 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
4939 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4940 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4941 				   BNXT_GRCP_WINDOW_2_BASE +
4942 				   offsetof(struct hcomm_status,
4943 					    sig_ver)));
4944 	/* If the signature is absent, then FW does not support this feature */
4945 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
4946 	    HCOMM_STATUS_SIGNATURE_VAL)
4947 		return 0;
4948 
4949 	if (!info) {
4950 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4951 				   sizeof(*info), 0);
4952 		if (!info)
4953 			return -ENOMEM;
4954 		bp->recovery_info = info;
4955 	} else {
4956 		memset(info, 0, sizeof(*info));
4957 	}
4958 
4959 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4960 				      BNXT_GRCP_WINDOW_2_BASE +
4961 				      offsetof(struct hcomm_status,
4962 					       fw_status_loc)));
4963 
4964 	/* Only pre-map the FW health status GRC register */
4965 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
4966 		return 0;
4967 
4968 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
4969 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
4970 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
4971 
4972 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
4973 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4974 
4975 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
4976 
4977 	return 0;
4978 }
4979 
4980 /* This function gets the FW version along with the
4981  * capabilities(MAX and current) of the function, vnic,
4982  * error recovery, phy and other chip related info
4983  */
4984 static int bnxt_get_config(struct bnxt *bp)
4985 {
4986 	uint16_t mtu;
4987 	int rc = 0;
4988 
4989 	bp->fw_cap = 0;
4990 
4991 	rc = bnxt_map_hcomm_fw_status_reg(bp);
4992 	if (rc)
4993 		return rc;
4994 
4995 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
4996 	if (rc) {
4997 		bnxt_check_fw_status(bp);
4998 		return rc;
4999 	}
5000 
5001 	rc = bnxt_hwrm_func_reset(bp);
5002 	if (rc)
5003 		return -EIO;
5004 
5005 	rc = bnxt_hwrm_vnic_qcaps(bp);
5006 	if (rc)
5007 		return rc;
5008 
5009 	rc = bnxt_hwrm_queue_qportcfg(bp);
5010 	if (rc)
5011 		return rc;
5012 
5013 	/* Get the MAX capabilities for this function.
5014 	 * This function also allocates context memory for TQM rings and
5015 	 * informs the firmware about this allocated backing store memory.
5016 	 */
5017 	rc = bnxt_hwrm_func_qcaps(bp);
5018 	if (rc)
5019 		return rc;
5020 
5021 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5022 	if (rc)
5023 		return rc;
5024 
5025 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5026 	if (rc)
5027 		return rc;
5028 
5029 	bnxt_hwrm_port_mac_qcfg(bp);
5030 
5031 	bnxt_hwrm_parent_pf_qcfg(bp);
5032 
5033 	bnxt_hwrm_port_phy_qcaps(bp);
5034 
5035 	bnxt_alloc_error_recovery_info(bp);
5036 	/* Get the adapter error recovery support info */
5037 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5038 	if (rc)
5039 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5040 
5041 	bnxt_hwrm_port_led_qcaps(bp);
5042 
5043 	return 0;
5044 }
5045 
5046 static int
5047 bnxt_init_locks(struct bnxt *bp)
5048 {
5049 	int err;
5050 
5051 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5052 	if (err) {
5053 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5054 		return err;
5055 	}
5056 
5057 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5058 	if (err) {
5059 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5060 		return err;
5061 	}
5062 
5063 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5064 	if (err) {
5065 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5066 		return err;
5067 	}
5068 
5069 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5070 	if (err)
5071 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5072 
5073 	return err;
5074 }
5075 
5076 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5077 {
5078 	int rc = 0;
5079 
5080 	rc = bnxt_get_config(bp);
5081 	if (rc)
5082 		return rc;
5083 
5084 	if (!reconfig_dev) {
5085 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5086 		if (rc)
5087 			return rc;
5088 	} else {
5089 		rc = bnxt_restore_dflt_mac(bp);
5090 		if (rc)
5091 			return rc;
5092 	}
5093 
5094 	bnxt_config_vf_req_fwd(bp);
5095 
5096 	rc = bnxt_hwrm_func_driver_register(bp);
5097 	if (rc) {
5098 		PMD_DRV_LOG(ERR, "Failed to register driver");
5099 		return -EBUSY;
5100 	}
5101 
5102 	if (BNXT_PF(bp)) {
5103 		if (bp->pdev->max_vfs) {
5104 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5105 			if (rc) {
5106 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5107 				return rc;
5108 			}
5109 		} else {
5110 			rc = bnxt_hwrm_allocate_pf_only(bp);
5111 			if (rc) {
5112 				PMD_DRV_LOG(ERR,
5113 					    "Failed to allocate PF resources");
5114 				return rc;
5115 			}
5116 		}
5117 	}
5118 
5119 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5120 	if (rc)
5121 		return rc;
5122 
5123 	rc = bnxt_setup_int(bp);
5124 	if (rc)
5125 		return rc;
5126 
5127 	rc = bnxt_request_int(bp);
5128 	if (rc)
5129 		return rc;
5130 
5131 	rc = bnxt_init_ctx_mem(bp);
5132 	if (rc) {
5133 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5134 		return rc;
5135 	}
5136 
5137 	return 0;
5138 }
5139 
5140 static int
5141 bnxt_parse_devarg_truflow(__rte_unused const char *key,
5142 			  const char *value, void *opaque_arg)
5143 {
5144 	struct bnxt *bp = opaque_arg;
5145 	unsigned long truflow;
5146 	char *end = NULL;
5147 
5148 	if (!value || !opaque_arg) {
5149 		PMD_DRV_LOG(ERR,
5150 			    "Invalid parameter passed to truflow devargs.\n");
5151 		return -EINVAL;
5152 	}
5153 
5154 	truflow = strtoul(value, &end, 10);
5155 	if (end == NULL || *end != '\0' ||
5156 	    (truflow == ULONG_MAX && errno == ERANGE)) {
5157 		PMD_DRV_LOG(ERR,
5158 			    "Invalid parameter passed to truflow devargs.\n");
5159 		return -EINVAL;
5160 	}
5161 
5162 	if (BNXT_DEVARG_TRUFLOW_INVALID(truflow)) {
5163 		PMD_DRV_LOG(ERR,
5164 			    "Invalid value passed to truflow devargs.\n");
5165 		return -EINVAL;
5166 	}
5167 
5168 	if (truflow) {
5169 		bp->flags |= BNXT_FLAG_TRUFLOW_EN;
5170 		PMD_DRV_LOG(INFO, "Host-based truflow feature enabled.\n");
5171 	} else {
5172 		bp->flags &= ~BNXT_FLAG_TRUFLOW_EN;
5173 		PMD_DRV_LOG(INFO, "Host-based truflow feature disabled.\n");
5174 	}
5175 
5176 	return 0;
5177 }
5178 
5179 static int
5180 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5181 			     const char *value, void *opaque_arg)
5182 {
5183 	struct bnxt *bp = opaque_arg;
5184 	unsigned long flow_xstat;
5185 	char *end = NULL;
5186 
5187 	if (!value || !opaque_arg) {
5188 		PMD_DRV_LOG(ERR,
5189 			    "Invalid parameter passed to flow_xstat devarg.\n");
5190 		return -EINVAL;
5191 	}
5192 
5193 	flow_xstat = strtoul(value, &end, 10);
5194 	if (end == NULL || *end != '\0' ||
5195 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5196 		PMD_DRV_LOG(ERR,
5197 			    "Invalid parameter passed to flow_xstat devarg.\n");
5198 		return -EINVAL;
5199 	}
5200 
5201 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5202 		PMD_DRV_LOG(ERR,
5203 			    "Invalid value passed to flow_xstat devarg.\n");
5204 		return -EINVAL;
5205 	}
5206 
5207 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5208 	if (BNXT_FLOW_XSTATS_EN(bp))
5209 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5210 
5211 	return 0;
5212 }
5213 
5214 static int
5215 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5216 					const char *value, void *opaque_arg)
5217 {
5218 	struct bnxt *bp = opaque_arg;
5219 	unsigned long max_num_kflows;
5220 	char *end = NULL;
5221 
5222 	if (!value || !opaque_arg) {
5223 		PMD_DRV_LOG(ERR,
5224 			"Invalid parameter passed to max_num_kflows devarg.\n");
5225 		return -EINVAL;
5226 	}
5227 
5228 	max_num_kflows = strtoul(value, &end, 10);
5229 	if (end == NULL || *end != '\0' ||
5230 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5231 		PMD_DRV_LOG(ERR,
5232 			"Invalid parameter passed to max_num_kflows devarg.\n");
5233 		return -EINVAL;
5234 	}
5235 
5236 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5237 		PMD_DRV_LOG(ERR,
5238 			"Invalid value passed to max_num_kflows devarg.\n");
5239 		return -EINVAL;
5240 	}
5241 
5242 	bp->max_num_kflows = max_num_kflows;
5243 	if (bp->max_num_kflows)
5244 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5245 				max_num_kflows);
5246 
5247 	return 0;
5248 }
5249 
5250 static int
5251 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5252 			    const char *value, void *opaque_arg)
5253 {
5254 	struct bnxt_representor *vfr_bp = opaque_arg;
5255 	unsigned long rep_is_pf;
5256 	char *end = NULL;
5257 
5258 	if (!value || !opaque_arg) {
5259 		PMD_DRV_LOG(ERR,
5260 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5261 		return -EINVAL;
5262 	}
5263 
5264 	rep_is_pf = strtoul(value, &end, 10);
5265 	if (end == NULL || *end != '\0' ||
5266 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5267 		PMD_DRV_LOG(ERR,
5268 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5269 		return -EINVAL;
5270 	}
5271 
5272 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5273 		PMD_DRV_LOG(ERR,
5274 			    "Invalid value passed to rep_is_pf devargs.\n");
5275 		return -EINVAL;
5276 	}
5277 
5278 	vfr_bp->flags |= rep_is_pf;
5279 	if (BNXT_REP_PF(vfr_bp))
5280 		PMD_DRV_LOG(INFO, "PF representor\n");
5281 	else
5282 		PMD_DRV_LOG(INFO, "VF representor\n");
5283 
5284 	return 0;
5285 }
5286 
5287 static int
5288 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5289 			       const char *value, void *opaque_arg)
5290 {
5291 	struct bnxt_representor *vfr_bp = opaque_arg;
5292 	unsigned long rep_based_pf;
5293 	char *end = NULL;
5294 
5295 	if (!value || !opaque_arg) {
5296 		PMD_DRV_LOG(ERR,
5297 			    "Invalid parameter passed to rep_based_pf "
5298 			    "devargs.\n");
5299 		return -EINVAL;
5300 	}
5301 
5302 	rep_based_pf = strtoul(value, &end, 10);
5303 	if (end == NULL || *end != '\0' ||
5304 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5305 		PMD_DRV_LOG(ERR,
5306 			    "Invalid parameter passed to rep_based_pf "
5307 			    "devargs.\n");
5308 		return -EINVAL;
5309 	}
5310 
5311 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5312 		PMD_DRV_LOG(ERR,
5313 			    "Invalid value passed to rep_based_pf devargs.\n");
5314 		return -EINVAL;
5315 	}
5316 
5317 	vfr_bp->rep_based_pf = rep_based_pf;
5318 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5319 
5320 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5321 
5322 	return 0;
5323 }
5324 
5325 static int
5326 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5327 			    const char *value, void *opaque_arg)
5328 {
5329 	struct bnxt_representor *vfr_bp = opaque_arg;
5330 	unsigned long rep_q_r2f;
5331 	char *end = NULL;
5332 
5333 	if (!value || !opaque_arg) {
5334 		PMD_DRV_LOG(ERR,
5335 			    "Invalid parameter passed to rep_q_r2f "
5336 			    "devargs.\n");
5337 		return -EINVAL;
5338 	}
5339 
5340 	rep_q_r2f = strtoul(value, &end, 10);
5341 	if (end == NULL || *end != '\0' ||
5342 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5343 		PMD_DRV_LOG(ERR,
5344 			    "Invalid parameter passed to rep_q_r2f "
5345 			    "devargs.\n");
5346 		return -EINVAL;
5347 	}
5348 
5349 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5350 		PMD_DRV_LOG(ERR,
5351 			    "Invalid value passed to rep_q_r2f devargs.\n");
5352 		return -EINVAL;
5353 	}
5354 
5355 	vfr_bp->rep_q_r2f = rep_q_r2f;
5356 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5357 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5358 
5359 	return 0;
5360 }
5361 
5362 static int
5363 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5364 			    const char *value, void *opaque_arg)
5365 {
5366 	struct bnxt_representor *vfr_bp = opaque_arg;
5367 	unsigned long rep_q_f2r;
5368 	char *end = NULL;
5369 
5370 	if (!value || !opaque_arg) {
5371 		PMD_DRV_LOG(ERR,
5372 			    "Invalid parameter passed to rep_q_f2r "
5373 			    "devargs.\n");
5374 		return -EINVAL;
5375 	}
5376 
5377 	rep_q_f2r = strtoul(value, &end, 10);
5378 	if (end == NULL || *end != '\0' ||
5379 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5380 		PMD_DRV_LOG(ERR,
5381 			    "Invalid parameter passed to rep_q_f2r "
5382 			    "devargs.\n");
5383 		return -EINVAL;
5384 	}
5385 
5386 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5387 		PMD_DRV_LOG(ERR,
5388 			    "Invalid value passed to rep_q_f2r devargs.\n");
5389 		return -EINVAL;
5390 	}
5391 
5392 	vfr_bp->rep_q_f2r = rep_q_f2r;
5393 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5394 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5395 
5396 	return 0;
5397 }
5398 
5399 static int
5400 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5401 			     const char *value, void *opaque_arg)
5402 {
5403 	struct bnxt_representor *vfr_bp = opaque_arg;
5404 	unsigned long rep_fc_r2f;
5405 	char *end = NULL;
5406 
5407 	if (!value || !opaque_arg) {
5408 		PMD_DRV_LOG(ERR,
5409 			    "Invalid parameter passed to rep_fc_r2f "
5410 			    "devargs.\n");
5411 		return -EINVAL;
5412 	}
5413 
5414 	rep_fc_r2f = strtoul(value, &end, 10);
5415 	if (end == NULL || *end != '\0' ||
5416 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5417 		PMD_DRV_LOG(ERR,
5418 			    "Invalid parameter passed to rep_fc_r2f "
5419 			    "devargs.\n");
5420 		return -EINVAL;
5421 	}
5422 
5423 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5424 		PMD_DRV_LOG(ERR,
5425 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5426 		return -EINVAL;
5427 	}
5428 
5429 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5430 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5431 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5432 
5433 	return 0;
5434 }
5435 
5436 static int
5437 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5438 			     const char *value, void *opaque_arg)
5439 {
5440 	struct bnxt_representor *vfr_bp = opaque_arg;
5441 	unsigned long rep_fc_f2r;
5442 	char *end = NULL;
5443 
5444 	if (!value || !opaque_arg) {
5445 		PMD_DRV_LOG(ERR,
5446 			    "Invalid parameter passed to rep_fc_f2r "
5447 			    "devargs.\n");
5448 		return -EINVAL;
5449 	}
5450 
5451 	rep_fc_f2r = strtoul(value, &end, 10);
5452 	if (end == NULL || *end != '\0' ||
5453 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5454 		PMD_DRV_LOG(ERR,
5455 			    "Invalid parameter passed to rep_fc_f2r "
5456 			    "devargs.\n");
5457 		return -EINVAL;
5458 	}
5459 
5460 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5461 		PMD_DRV_LOG(ERR,
5462 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5463 		return -EINVAL;
5464 	}
5465 
5466 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5467 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5468 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5469 
5470 	return 0;
5471 }
5472 
5473 static int
5474 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5475 {
5476 	struct rte_kvargs *kvlist;
5477 	int ret;
5478 
5479 	if (devargs == NULL)
5480 		return 0;
5481 
5482 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5483 	if (kvlist == NULL)
5484 		return -EINVAL;
5485 
5486 	/*
5487 	 * Handler for "truflow" devarg.
5488 	 * Invoked as for ex: "-a 0000:00:0d.0,host-based-truflow=1"
5489 	 */
5490 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_TRUFLOW,
5491 				 bnxt_parse_devarg_truflow, bp);
5492 	if (ret)
5493 		goto err;
5494 
5495 	/*
5496 	 * Handler for "flow_xstat" devarg.
5497 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5498 	 */
5499 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5500 				 bnxt_parse_devarg_flow_xstat, bp);
5501 	if (ret)
5502 		goto err;
5503 
5504 	/*
5505 	 * Handler for "max_num_kflows" devarg.
5506 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5507 	 */
5508 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5509 				 bnxt_parse_devarg_max_num_kflows, bp);
5510 	if (ret)
5511 		goto err;
5512 
5513 err:
5514 	rte_kvargs_free(kvlist);
5515 	return ret;
5516 }
5517 
5518 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5519 {
5520 	int rc = 0;
5521 
5522 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5523 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5524 		if (rc)
5525 			PMD_DRV_LOG(ERR,
5526 				    "Failed to alloc switch domain: %d\n", rc);
5527 		else
5528 			PMD_DRV_LOG(INFO,
5529 				    "Switch domain allocated %d\n",
5530 				    bp->switch_domain_id);
5531 	}
5532 
5533 	return rc;
5534 }
5535 
5536 /* Allocate and initialize various fields in bnxt struct that
5537  * need to be allocated/destroyed only once in the lifetime of the driver
5538  */
5539 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5540 {
5541 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5542 	struct bnxt *bp = eth_dev->data->dev_private;
5543 	int rc = 0;
5544 
5545 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5546 
5547 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5548 		bp->flags |= BNXT_FLAG_VF;
5549 
5550 	if (bnxt_p5_device(pci_dev->id.device_id))
5551 		bp->flags |= BNXT_FLAG_CHIP_P5;
5552 
5553 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5554 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5555 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5556 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5557 		bp->flags |= BNXT_FLAG_STINGRAY;
5558 
5559 	if (BNXT_TRUFLOW_EN(bp)) {
5560 		/* extra mbuf field is required to store CFA code from mark */
5561 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5562 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5563 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5564 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5565 		};
5566 		bnxt_cfa_code_dynfield_offset =
5567 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5568 		if (bnxt_cfa_code_dynfield_offset < 0) {
5569 			PMD_DRV_LOG(ERR,
5570 			    "Failed to register mbuf field for TruFlow mark\n");
5571 			return -rte_errno;
5572 		}
5573 	}
5574 
5575 	rc = bnxt_map_pci_bars(eth_dev);
5576 	if (rc) {
5577 		PMD_DRV_LOG(ERR,
5578 			    "Failed to initialize board rc: %x\n", rc);
5579 		return rc;
5580 	}
5581 
5582 	rc = bnxt_alloc_pf_info(bp);
5583 	if (rc)
5584 		return rc;
5585 
5586 	rc = bnxt_alloc_link_info(bp);
5587 	if (rc)
5588 		return rc;
5589 
5590 	rc = bnxt_alloc_parent_info(bp);
5591 	if (rc)
5592 		return rc;
5593 
5594 	rc = bnxt_alloc_hwrm_resources(bp);
5595 	if (rc) {
5596 		PMD_DRV_LOG(ERR,
5597 			    "Failed to allocate response buffer rc: %x\n", rc);
5598 		return rc;
5599 	}
5600 	rc = bnxt_alloc_leds_info(bp);
5601 	if (rc)
5602 		return rc;
5603 
5604 	rc = bnxt_alloc_cos_queues(bp);
5605 	if (rc)
5606 		return rc;
5607 
5608 	rc = bnxt_init_locks(bp);
5609 	if (rc)
5610 		return rc;
5611 
5612 	rc = bnxt_alloc_switch_domain(bp);
5613 	if (rc)
5614 		return rc;
5615 
5616 	return rc;
5617 }
5618 
5619 static int
5620 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5621 {
5622 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5623 	static int version_printed;
5624 	struct bnxt *bp;
5625 	int rc;
5626 
5627 	if (version_printed++ == 0)
5628 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5629 
5630 	eth_dev->dev_ops = &bnxt_dev_ops;
5631 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5632 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5633 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5634 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5635 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5636 
5637 	/*
5638 	 * For secondary processes, we don't initialise any further
5639 	 * as primary has already done this work.
5640 	 */
5641 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5642 		return 0;
5643 
5644 	rte_eth_copy_pci_info(eth_dev, pci_dev);
5645 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5646 
5647 	bp = eth_dev->data->dev_private;
5648 
5649 	/* Parse dev arguments passed on when starting the DPDK application. */
5650 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5651 	if (rc)
5652 		goto error_free;
5653 
5654 	rc = bnxt_drv_init(eth_dev);
5655 	if (rc)
5656 		goto error_free;
5657 
5658 	rc = bnxt_init_resources(bp, false);
5659 	if (rc)
5660 		goto error_free;
5661 
5662 	rc = bnxt_alloc_stats_mem(bp);
5663 	if (rc)
5664 		goto error_free;
5665 
5666 	PMD_DRV_LOG(INFO,
5667 		    DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
5668 		    pci_dev->mem_resource[0].phys_addr,
5669 		    pci_dev->mem_resource[0].addr);
5670 
5671 	return 0;
5672 
5673 error_free:
5674 	bnxt_dev_uninit(eth_dev);
5675 	return rc;
5676 }
5677 
5678 
5679 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5680 {
5681 	if (!ctx)
5682 		return;
5683 
5684 	if (ctx->va)
5685 		rte_free(ctx->va);
5686 
5687 	ctx->va = NULL;
5688 	ctx->dma = RTE_BAD_IOVA;
5689 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5690 }
5691 
5692 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5693 {
5694 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5695 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5696 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
5697 				  bp->flow_stat->max_fc,
5698 				  false);
5699 
5700 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5701 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5702 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
5703 				  bp->flow_stat->max_fc,
5704 				  false);
5705 
5706 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5707 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5708 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5709 
5710 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5711 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5712 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5713 
5714 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5715 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5716 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5717 
5718 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5719 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5720 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5721 }
5722 
5723 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5724 {
5725 	bnxt_unregister_fc_ctx_mem(bp);
5726 
5727 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5728 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5729 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5730 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5731 }
5732 
5733 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5734 {
5735 	if (BNXT_FLOW_XSTATS_EN(bp))
5736 		bnxt_uninit_fc_ctx_mem(bp);
5737 }
5738 
5739 static void
5740 bnxt_free_error_recovery_info(struct bnxt *bp)
5741 {
5742 	rte_free(bp->recovery_info);
5743 	bp->recovery_info = NULL;
5744 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5745 }
5746 
5747 static int
5748 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5749 {
5750 	int rc;
5751 
5752 	bnxt_free_int(bp);
5753 	bnxt_free_mem(bp, reconfig_dev);
5754 
5755 	bnxt_hwrm_func_buf_unrgtr(bp);
5756 	rte_free(bp->pf->vf_req_buf);
5757 
5758 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
5759 	bp->flags &= ~BNXT_FLAG_REGISTERED;
5760 	bnxt_free_ctx_mem(bp);
5761 	if (!reconfig_dev) {
5762 		bnxt_free_hwrm_resources(bp);
5763 		bnxt_free_error_recovery_info(bp);
5764 	}
5765 
5766 	bnxt_uninit_ctx_mem(bp);
5767 
5768 	bnxt_free_flow_stats_info(bp);
5769 	bnxt_free_rep_info(bp);
5770 	rte_free(bp->ptp_cfg);
5771 	bp->ptp_cfg = NULL;
5772 	return rc;
5773 }
5774 
5775 static int
5776 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5777 {
5778 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5779 		return -EPERM;
5780 
5781 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5782 
5783 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5784 		bnxt_dev_close_op(eth_dev);
5785 
5786 	return 0;
5787 }
5788 
5789 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
5790 {
5791 	struct bnxt *bp = eth_dev->data->dev_private;
5792 	struct rte_eth_dev *vf_rep_eth_dev;
5793 	int ret = 0, i;
5794 
5795 	if (!bp)
5796 		return -EINVAL;
5797 
5798 	for (i = 0; i < bp->num_reps; i++) {
5799 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
5800 		if (!vf_rep_eth_dev)
5801 			continue;
5802 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
5803 			    vf_rep_eth_dev->data->port_id);
5804 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
5805 	}
5806 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
5807 		    eth_dev->data->port_id);
5808 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
5809 
5810 	return ret;
5811 }
5812 
5813 static void bnxt_free_rep_info(struct bnxt *bp)
5814 {
5815 	rte_free(bp->rep_info);
5816 	bp->rep_info = NULL;
5817 	rte_free(bp->cfa_code_map);
5818 	bp->cfa_code_map = NULL;
5819 }
5820 
5821 static int bnxt_init_rep_info(struct bnxt *bp)
5822 {
5823 	int i = 0, rc;
5824 
5825 	if (bp->rep_info)
5826 		return 0;
5827 
5828 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
5829 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
5830 				   0);
5831 	if (!bp->rep_info) {
5832 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
5833 		return -ENOMEM;
5834 	}
5835 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
5836 				       sizeof(*bp->cfa_code_map) *
5837 				       BNXT_MAX_CFA_CODE, 0);
5838 	if (!bp->cfa_code_map) {
5839 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
5840 		bnxt_free_rep_info(bp);
5841 		return -ENOMEM;
5842 	}
5843 
5844 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
5845 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
5846 
5847 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
5848 	if (rc) {
5849 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
5850 		bnxt_free_rep_info(bp);
5851 		return rc;
5852 	}
5853 
5854 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
5855 	if (rc) {
5856 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
5857 		bnxt_free_rep_info(bp);
5858 		return rc;
5859 	}
5860 
5861 	return rc;
5862 }
5863 
5864 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
5865 			       struct rte_eth_devargs *eth_da,
5866 			       struct rte_eth_dev *backing_eth_dev,
5867 			       const char *dev_args)
5868 {
5869 	struct rte_eth_dev *vf_rep_eth_dev;
5870 	char name[RTE_ETH_NAME_MAX_LEN];
5871 	struct bnxt *backing_bp;
5872 	uint16_t num_rep;
5873 	int i, ret = 0;
5874 	struct rte_kvargs *kvlist = NULL;
5875 
5876 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
5877 		return 0;
5878 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
5879 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
5880 			    eth_da->type);
5881 		return -ENOTSUP;
5882 	}
5883 	num_rep = eth_da->nb_representor_ports;
5884 	if (num_rep > BNXT_MAX_VF_REPS) {
5885 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
5886 			    num_rep, BNXT_MAX_VF_REPS);
5887 		return -EINVAL;
5888 	}
5889 
5890 	if (num_rep >= RTE_MAX_ETHPORTS) {
5891 		PMD_DRV_LOG(ERR,
5892 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
5893 			    num_rep, RTE_MAX_ETHPORTS);
5894 		return -EINVAL;
5895 	}
5896 
5897 	backing_bp = backing_eth_dev->data->dev_private;
5898 
5899 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
5900 		PMD_DRV_LOG(ERR,
5901 			    "Not a PF or trusted VF. No Representor support\n");
5902 		/* Returning an error is not an option.
5903 		 * Applications are not handling this correctly
5904 		 */
5905 		return 0;
5906 	}
5907 
5908 	if (bnxt_init_rep_info(backing_bp))
5909 		return 0;
5910 
5911 	for (i = 0; i < num_rep; i++) {
5912 		struct bnxt_representor representor = {
5913 			.vf_id = eth_da->representor_ports[i],
5914 			.switch_domain_id = backing_bp->switch_domain_id,
5915 			.parent_dev = backing_eth_dev
5916 		};
5917 
5918 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
5919 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
5920 				    representor.vf_id, BNXT_MAX_VF_REPS);
5921 			continue;
5922 		}
5923 
5924 		/* representor port net_bdf_port */
5925 		snprintf(name, sizeof(name), "net_%s_representor_%d",
5926 			 pci_dev->device.name, eth_da->representor_ports[i]);
5927 
5928 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
5929 		if (kvlist) {
5930 			/*
5931 			 * Handler for "rep_is_pf" devarg.
5932 			 * Invoked as for ex: "-a 000:00:0d.0,
5933 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5934 			 */
5935 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
5936 						 bnxt_parse_devarg_rep_is_pf,
5937 						 (void *)&representor);
5938 			if (ret) {
5939 				ret = -EINVAL;
5940 				goto err;
5941 			}
5942 			/*
5943 			 * Handler for "rep_based_pf" devarg.
5944 			 * Invoked as for ex: "-a 000:00:0d.0,
5945 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5946 			 */
5947 			ret = rte_kvargs_process(kvlist,
5948 						 BNXT_DEVARG_REP_BASED_PF,
5949 						 bnxt_parse_devarg_rep_based_pf,
5950 						 (void *)&representor);
5951 			if (ret) {
5952 				ret = -EINVAL;
5953 				goto err;
5954 			}
5955 			/*
5956 			 * Handler for "rep_based_pf" devarg.
5957 			 * Invoked as for ex: "-a 000:00:0d.0,
5958 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5959 			 */
5960 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
5961 						 bnxt_parse_devarg_rep_q_r2f,
5962 						 (void *)&representor);
5963 			if (ret) {
5964 				ret = -EINVAL;
5965 				goto err;
5966 			}
5967 			/*
5968 			 * Handler for "rep_based_pf" devarg.
5969 			 * Invoked as for ex: "-a 000:00:0d.0,
5970 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5971 			 */
5972 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
5973 						 bnxt_parse_devarg_rep_q_f2r,
5974 						 (void *)&representor);
5975 			if (ret) {
5976 				ret = -EINVAL;
5977 				goto err;
5978 			}
5979 			/*
5980 			 * Handler for "rep_based_pf" devarg.
5981 			 * Invoked as for ex: "-a 000:00:0d.0,
5982 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5983 			 */
5984 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
5985 						 bnxt_parse_devarg_rep_fc_r2f,
5986 						 (void *)&representor);
5987 			if (ret) {
5988 				ret = -EINVAL;
5989 				goto err;
5990 			}
5991 			/*
5992 			 * Handler for "rep_based_pf" devarg.
5993 			 * Invoked as for ex: "-a 000:00:0d.0,
5994 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5995 			 */
5996 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
5997 						 bnxt_parse_devarg_rep_fc_f2r,
5998 						 (void *)&representor);
5999 			if (ret) {
6000 				ret = -EINVAL;
6001 				goto err;
6002 			}
6003 		}
6004 
6005 		ret = rte_eth_dev_create(&pci_dev->device, name,
6006 					 sizeof(struct bnxt_representor),
6007 					 NULL, NULL,
6008 					 bnxt_representor_init,
6009 					 &representor);
6010 		if (ret) {
6011 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6012 				    "representor %s.", name);
6013 			goto err;
6014 		}
6015 
6016 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6017 		if (!vf_rep_eth_dev) {
6018 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6019 				    " for VF-Rep: %s.", name);
6020 			ret = -ENODEV;
6021 			goto err;
6022 		}
6023 
6024 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6025 			    backing_eth_dev->data->port_id);
6026 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6027 							 vf_rep_eth_dev;
6028 		backing_bp->num_reps++;
6029 
6030 	}
6031 
6032 	rte_kvargs_free(kvlist);
6033 	return 0;
6034 
6035 err:
6036 	/* If num_rep > 1, then rollback already created
6037 	 * ports, since we'll be failing the probe anyway
6038 	 */
6039 	if (num_rep > 1)
6040 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6041 	rte_errno = -ret;
6042 	rte_kvargs_free(kvlist);
6043 
6044 	return ret;
6045 }
6046 
6047 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6048 			  struct rte_pci_device *pci_dev)
6049 {
6050 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6051 	struct rte_eth_dev *backing_eth_dev;
6052 	uint16_t num_rep;
6053 	int ret = 0;
6054 
6055 	if (pci_dev->device.devargs) {
6056 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6057 					    &eth_da);
6058 		if (ret)
6059 			return ret;
6060 	}
6061 
6062 	num_rep = eth_da.nb_representor_ports;
6063 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6064 		    num_rep);
6065 
6066 	/* We could come here after first level of probe is already invoked
6067 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6068 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6069 	 */
6070 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6071 	if (backing_eth_dev == NULL) {
6072 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6073 					 sizeof(struct bnxt),
6074 					 eth_dev_pci_specific_init, pci_dev,
6075 					 bnxt_dev_init, NULL);
6076 
6077 		if (ret || !num_rep)
6078 			return ret;
6079 
6080 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6081 	}
6082 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6083 		    backing_eth_dev->data->port_id);
6084 
6085 	if (!num_rep)
6086 		return ret;
6087 
6088 	/* probe representor ports now */
6089 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6090 				  pci_dev->device.devargs->args);
6091 
6092 	return ret;
6093 }
6094 
6095 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6096 {
6097 	struct rte_eth_dev *eth_dev;
6098 
6099 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6100 	if (!eth_dev)
6101 		return 0; /* Invoked typically only by OVS-DPDK, by the
6102 			   * time it comes here the eth_dev is already
6103 			   * deleted by rte_eth_dev_close(), so returning
6104 			   * +ve value will at least help in proper cleanup
6105 			   */
6106 
6107 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6108 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6109 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6110 			return rte_eth_dev_destroy(eth_dev,
6111 						   bnxt_representor_uninit);
6112 		else
6113 			return rte_eth_dev_destroy(eth_dev,
6114 						   bnxt_dev_uninit);
6115 	} else {
6116 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6117 	}
6118 }
6119 
6120 static struct rte_pci_driver bnxt_rte_pmd = {
6121 	.id_table = bnxt_pci_id_map,
6122 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6123 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6124 						  * and OVS-DPDK
6125 						  */
6126 	.probe = bnxt_pci_probe,
6127 	.remove = bnxt_pci_remove,
6128 };
6129 
6130 static bool
6131 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6132 {
6133 	if (strcmp(dev->device->driver->name, drv->driver.name))
6134 		return false;
6135 
6136 	return true;
6137 }
6138 
6139 bool is_bnxt_supported(struct rte_eth_dev *dev)
6140 {
6141 	return is_device_supported(dev, &bnxt_rte_pmd);
6142 }
6143 
6144 RTE_LOG_REGISTER(bnxt_logtype_driver, pmd.net.bnxt.driver, NOTICE);
6145 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6146 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6147 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6148