xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision 29fd052d)
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_FLOW_XSTAT	"flow-xstat"
91 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
92 #define BNXT_DEVARG_REPRESENTOR	"representor"
93 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
94 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
95 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
96 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
97 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
98 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
99 #define BNXT_DEVARG_APP_ID	"app-id"
100 
101 static const char *const bnxt_dev_args[] = {
102 	BNXT_DEVARG_REPRESENTOR,
103 	BNXT_DEVARG_FLOW_XSTAT,
104 	BNXT_DEVARG_MAX_NUM_KFLOWS,
105 	BNXT_DEVARG_REP_BASED_PF,
106 	BNXT_DEVARG_REP_IS_PF,
107 	BNXT_DEVARG_REP_Q_R2F,
108 	BNXT_DEVARG_REP_Q_F2R,
109 	BNXT_DEVARG_REP_FC_R2F,
110 	BNXT_DEVARG_REP_FC_F2R,
111 	BNXT_DEVARG_APP_ID,
112 	NULL
113 };
114 
115 /*
116  * app-id = an non-negative 8-bit number
117  */
118 #define BNXT_DEVARG_APP_ID_INVALID(val)			((val) > 255)
119 
120 /*
121  * flow_xstat == false to disable the feature
122  * flow_xstat == true to enable the feature
123  */
124 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
125 
126 /*
127  * rep_is_pf == false to indicate VF representor
128  * rep_is_pf == true to indicate PF representor
129  */
130 #define	BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)	((rep_is_pf) > 1)
131 
132 /*
133  * rep_based_pf == Physical index of the PF
134  */
135 #define	BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)	((rep_based_pf) > 15)
136 /*
137  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
138  */
139 #define	BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)	((rep_q_r2f) > 3)
140 
141 /*
142  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
143  */
144 #define	BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)	((rep_q_f2r) > 3)
145 
146 /*
147  * rep_fc_r2f == Flow control for the representor to endpoint direction
148  */
149 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)	((rep_fc_r2f) > 1)
150 
151 /*
152  * rep_fc_f2r == Flow control for the endpoint to representor direction
153  */
154 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)	((rep_fc_f2r) > 1)
155 
156 int bnxt_cfa_code_dynfield_offset = -1;
157 
158 /*
159  * max_num_kflows must be >= 32
160  * and must be a power-of-2 supported value
161  * return: 1 -> invalid
162  *         0 -> valid
163  */
164 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
165 {
166 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
167 		return 1;
168 	return 0;
169 }
170 
171 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
172 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
173 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
174 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
175 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
176 static int bnxt_restore_vlan_filters(struct bnxt *bp);
177 static void bnxt_dev_recover(void *arg);
178 static void bnxt_free_error_recovery_info(struct bnxt *bp);
179 static void bnxt_free_rep_info(struct bnxt *bp);
180 
181 int is_bnxt_in_error(struct bnxt *bp)
182 {
183 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
184 		return -EIO;
185 	if (bp->flags & BNXT_FLAG_FW_RESET)
186 		return -EBUSY;
187 
188 	return 0;
189 }
190 
191 /***********************/
192 
193 /*
194  * High level utility functions
195  */
196 
197 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
198 {
199 	unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
200 					     BNXT_RSS_TBL_SIZE_P5);
201 
202 	if (!BNXT_CHIP_P5(bp))
203 		return 1;
204 
205 	return RTE_ALIGN_MUL_CEIL(num_rss_rings,
206 				  BNXT_RSS_ENTRIES_PER_CTX_P5) /
207 				  BNXT_RSS_ENTRIES_PER_CTX_P5;
208 }
209 
210 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
211 {
212 	if (!BNXT_CHIP_P5(bp))
213 		return HW_HASH_INDEX_SIZE;
214 
215 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
216 }
217 
218 static void bnxt_free_parent_info(struct bnxt *bp)
219 {
220 	rte_free(bp->parent);
221 	bp->parent = NULL;
222 }
223 
224 static void bnxt_free_pf_info(struct bnxt *bp)
225 {
226 	rte_free(bp->pf);
227 	bp->pf = NULL;
228 }
229 
230 static void bnxt_free_link_info(struct bnxt *bp)
231 {
232 	rte_free(bp->link_info);
233 	bp->link_info = NULL;
234 }
235 
236 static void bnxt_free_leds_info(struct bnxt *bp)
237 {
238 	if (BNXT_VF(bp))
239 		return;
240 
241 	rte_free(bp->leds);
242 	bp->leds = NULL;
243 }
244 
245 static void bnxt_free_flow_stats_info(struct bnxt *bp)
246 {
247 	rte_free(bp->flow_stat);
248 	bp->flow_stat = NULL;
249 }
250 
251 static void bnxt_free_cos_queues(struct bnxt *bp)
252 {
253 	rte_free(bp->rx_cos_queue);
254 	bp->rx_cos_queue = NULL;
255 	rte_free(bp->tx_cos_queue);
256 	bp->tx_cos_queue = NULL;
257 }
258 
259 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
260 {
261 	bnxt_free_filter_mem(bp);
262 	bnxt_free_vnic_attributes(bp);
263 	bnxt_free_vnic_mem(bp);
264 
265 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
266 	 * If the number of rings change across fw update,
267 	 * we don't have much choice except to warn the user.
268 	 */
269 	if (!reconfig) {
270 		bnxt_free_stats(bp);
271 		bnxt_free_tx_rings(bp);
272 		bnxt_free_rx_rings(bp);
273 	}
274 	bnxt_free_async_cp_ring(bp);
275 	bnxt_free_rxtx_nq_ring(bp);
276 
277 	rte_free(bp->grp_info);
278 	bp->grp_info = NULL;
279 }
280 
281 static int bnxt_alloc_parent_info(struct bnxt *bp)
282 {
283 	bp->parent = rte_zmalloc("bnxt_parent_info",
284 				 sizeof(struct bnxt_parent_info), 0);
285 	if (bp->parent == NULL)
286 		return -ENOMEM;
287 
288 	return 0;
289 }
290 
291 static int bnxt_alloc_pf_info(struct bnxt *bp)
292 {
293 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
294 	if (bp->pf == NULL)
295 		return -ENOMEM;
296 
297 	return 0;
298 }
299 
300 static int bnxt_alloc_link_info(struct bnxt *bp)
301 {
302 	bp->link_info =
303 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
304 	if (bp->link_info == NULL)
305 		return -ENOMEM;
306 
307 	return 0;
308 }
309 
310 static int bnxt_alloc_leds_info(struct bnxt *bp)
311 {
312 	if (BNXT_VF(bp))
313 		return 0;
314 
315 	bp->leds = rte_zmalloc("bnxt_leds",
316 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
317 			       0);
318 	if (bp->leds == NULL)
319 		return -ENOMEM;
320 
321 	return 0;
322 }
323 
324 static int bnxt_alloc_cos_queues(struct bnxt *bp)
325 {
326 	bp->rx_cos_queue =
327 		rte_zmalloc("bnxt_rx_cosq",
328 			    BNXT_COS_QUEUE_COUNT *
329 			    sizeof(struct bnxt_cos_queue_info),
330 			    0);
331 	if (bp->rx_cos_queue == NULL)
332 		return -ENOMEM;
333 
334 	bp->tx_cos_queue =
335 		rte_zmalloc("bnxt_tx_cosq",
336 			    BNXT_COS_QUEUE_COUNT *
337 			    sizeof(struct bnxt_cos_queue_info),
338 			    0);
339 	if (bp->tx_cos_queue == NULL)
340 		return -ENOMEM;
341 
342 	return 0;
343 }
344 
345 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
346 {
347 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
348 				    sizeof(struct bnxt_flow_stat_info), 0);
349 	if (bp->flow_stat == NULL)
350 		return -ENOMEM;
351 
352 	return 0;
353 }
354 
355 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
356 {
357 	int rc;
358 
359 	rc = bnxt_alloc_ring_grps(bp);
360 	if (rc)
361 		goto alloc_mem_err;
362 
363 	rc = bnxt_alloc_async_ring_struct(bp);
364 	if (rc)
365 		goto alloc_mem_err;
366 
367 	rc = bnxt_alloc_vnic_mem(bp);
368 	if (rc)
369 		goto alloc_mem_err;
370 
371 	rc = bnxt_alloc_vnic_attributes(bp, reconfig);
372 	if (rc)
373 		goto alloc_mem_err;
374 
375 	rc = bnxt_alloc_filter_mem(bp);
376 	if (rc)
377 		goto alloc_mem_err;
378 
379 	rc = bnxt_alloc_async_cp_ring(bp);
380 	if (rc)
381 		goto alloc_mem_err;
382 
383 	rc = bnxt_alloc_rxtx_nq_ring(bp);
384 	if (rc)
385 		goto alloc_mem_err;
386 
387 	if (BNXT_FLOW_XSTATS_EN(bp)) {
388 		rc = bnxt_alloc_flow_stats_info(bp);
389 		if (rc)
390 			goto alloc_mem_err;
391 	}
392 
393 	return 0;
394 
395 alloc_mem_err:
396 	bnxt_free_mem(bp, reconfig);
397 	return rc;
398 }
399 
400 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
401 {
402 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
403 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
404 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
405 	struct bnxt_rx_queue *rxq;
406 	unsigned int j;
407 	int rc;
408 
409 	rc = bnxt_vnic_grp_alloc(bp, vnic);
410 	if (rc)
411 		goto err_out;
412 
413 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
414 		    vnic_id, vnic, vnic->fw_grp_ids);
415 
416 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
417 	if (rc)
418 		goto err_out;
419 
420 	/* Alloc RSS context only if RSS mode is enabled */
421 	if (dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS) {
422 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
423 
424 		/* RSS table size in Thor is 512.
425 		 * Cap max Rx rings to same value
426 		 */
427 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
428 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
429 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
430 			goto err_out;
431 		}
432 
433 		rc = 0;
434 		for (j = 0; j < nr_ctxs; j++) {
435 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
436 			if (rc)
437 				break;
438 		}
439 		if (rc) {
440 			PMD_DRV_LOG(ERR,
441 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
442 				    vnic_id, j, rc);
443 			goto err_out;
444 		}
445 		vnic->num_lb_ctxts = nr_ctxs;
446 	}
447 
448 	/*
449 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
450 	 * setting is not available at this time, it will not be
451 	 * configured correctly in the CFA.
452 	 */
453 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
454 		vnic->vlan_strip = true;
455 	else
456 		vnic->vlan_strip = false;
457 
458 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
459 	if (rc)
460 		goto err_out;
461 
462 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
463 	if (rc)
464 		goto err_out;
465 
466 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
467 		rxq = bp->eth_dev->data->rx_queues[j];
468 
469 		PMD_DRV_LOG(DEBUG,
470 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
471 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
472 
473 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
474 			rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
475 		else
476 			vnic->rx_queue_cnt++;
477 	}
478 
479 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
480 
481 	rc = bnxt_vnic_rss_configure(bp, vnic);
482 	if (rc)
483 		goto err_out;
484 
485 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
486 
487 	rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
488 				    (rx_offloads & RTE_ETH_RX_OFFLOAD_TCP_LRO) ?
489 				    true : false);
490 	if (rc)
491 		goto err_out;
492 
493 	return 0;
494 err_out:
495 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
496 		    vnic_id, rc);
497 	return rc;
498 }
499 
500 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
501 {
502 	int rc = 0;
503 
504 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
505 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
506 	if (rc)
507 		return rc;
508 
509 	PMD_DRV_LOG(DEBUG,
510 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
511 		    " rx_fc_in_tbl.ctx_id = %d\n",
512 		    bp->flow_stat->rx_fc_in_tbl.va,
513 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
514 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
515 
516 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
517 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
518 	if (rc)
519 		return rc;
520 
521 	PMD_DRV_LOG(DEBUG,
522 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
523 		    " rx_fc_out_tbl.ctx_id = %d\n",
524 		    bp->flow_stat->rx_fc_out_tbl.va,
525 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
526 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
527 
528 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
529 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
530 	if (rc)
531 		return rc;
532 
533 	PMD_DRV_LOG(DEBUG,
534 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
535 		    " tx_fc_in_tbl.ctx_id = %d\n",
536 		    bp->flow_stat->tx_fc_in_tbl.va,
537 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
538 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
539 
540 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
541 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
542 	if (rc)
543 		return rc;
544 
545 	PMD_DRV_LOG(DEBUG,
546 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
547 		    " tx_fc_out_tbl.ctx_id = %d\n",
548 		    bp->flow_stat->tx_fc_out_tbl.va,
549 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
550 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
551 
552 	memset(bp->flow_stat->rx_fc_out_tbl.va,
553 	       0,
554 	       bp->flow_stat->rx_fc_out_tbl.size);
555 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
556 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
557 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
558 				       bp->flow_stat->max_fc,
559 				       true);
560 	if (rc)
561 		return rc;
562 
563 	memset(bp->flow_stat->tx_fc_out_tbl.va,
564 	       0,
565 	       bp->flow_stat->tx_fc_out_tbl.size);
566 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
567 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
568 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
569 				       bp->flow_stat->max_fc,
570 				       true);
571 
572 	return rc;
573 }
574 
575 static int bnxt_alloc_ctx_mem_buf(struct bnxt *bp, char *type, size_t size,
576 				  struct bnxt_ctx_mem_buf_info *ctx)
577 {
578 	if (!ctx)
579 		return -EINVAL;
580 
581 	ctx->va = rte_zmalloc_socket(type, size, 0,
582 				     bp->eth_dev->device->numa_node);
583 	if (ctx->va == NULL)
584 		return -ENOMEM;
585 	rte_mem_lock_page(ctx->va);
586 	ctx->size = size;
587 	ctx->dma = rte_mem_virt2iova(ctx->va);
588 	if (ctx->dma == RTE_BAD_IOVA)
589 		return -ENOMEM;
590 
591 	return 0;
592 }
593 
594 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
595 {
596 	struct rte_pci_device *pdev = bp->pdev;
597 	char type[RTE_MEMZONE_NAMESIZE];
598 	uint16_t max_fc;
599 	int rc = 0;
600 
601 	max_fc = bp->flow_stat->max_fc;
602 
603 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
604 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
605 	/* 4 bytes for each counter-id */
606 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
607 				    max_fc * 4,
608 				    &bp->flow_stat->rx_fc_in_tbl);
609 	if (rc)
610 		return rc;
611 
612 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
613 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
614 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
615 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
616 				    max_fc * 16,
617 				    &bp->flow_stat->rx_fc_out_tbl);
618 	if (rc)
619 		return rc;
620 
621 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
622 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
623 	/* 4 bytes for each counter-id */
624 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
625 				    max_fc * 4,
626 				    &bp->flow_stat->tx_fc_in_tbl);
627 	if (rc)
628 		return rc;
629 
630 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
631 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
632 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
633 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
634 				    max_fc * 16,
635 				    &bp->flow_stat->tx_fc_out_tbl);
636 	if (rc)
637 		return rc;
638 
639 	rc = bnxt_register_fc_ctx_mem(bp);
640 
641 	return rc;
642 }
643 
644 static int bnxt_init_ctx_mem(struct bnxt *bp)
645 {
646 	int rc = 0;
647 
648 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
649 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
650 	    !BNXT_FLOW_XSTATS_EN(bp))
651 		return 0;
652 
653 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
654 	if (rc)
655 		return rc;
656 
657 	rc = bnxt_init_fc_ctx_mem(bp);
658 
659 	return rc;
660 }
661 
662 static int bnxt_update_phy_setting(struct bnxt *bp)
663 {
664 	struct rte_eth_link new;
665 	int rc;
666 
667 	rc = bnxt_get_hwrm_link_config(bp, &new);
668 	if (rc) {
669 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
670 		return rc;
671 	}
672 
673 	/*
674 	 * On BCM957508-N2100 adapters, FW will not allow any user other
675 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
676 	 * always returns link up. Force phy update always in that case.
677 	 */
678 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
679 		rc = bnxt_set_hwrm_link_config(bp, true);
680 		if (rc) {
681 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
682 			return rc;
683 		}
684 	}
685 
686 	return rc;
687 }
688 
689 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
690 {
691 	rte_free(bp->prev_rx_ring_stats);
692 	rte_free(bp->prev_tx_ring_stats);
693 
694 	bp->prev_rx_ring_stats = NULL;
695 	bp->prev_tx_ring_stats = NULL;
696 }
697 
698 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
699 {
700 	bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
701 					      sizeof(struct bnxt_ring_stats) *
702 					      bp->rx_cp_nr_rings,
703 					      0);
704 	if (bp->prev_rx_ring_stats == NULL)
705 		return -ENOMEM;
706 
707 	bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
708 					     sizeof(struct bnxt_ring_stats) *
709 					     bp->tx_cp_nr_rings,
710 					     0);
711 	if (bp->prev_tx_ring_stats == NULL)
712 		goto error;
713 
714 	return 0;
715 
716 error:
717 	bnxt_free_prev_ring_stats(bp);
718 	return -ENOMEM;
719 }
720 
721 static int bnxt_start_nic(struct bnxt *bp)
722 {
723 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
724 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
725 	uint32_t intr_vector = 0;
726 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
727 	uint32_t vec = BNXT_MISC_VEC_ID;
728 	unsigned int i, j;
729 	int rc;
730 
731 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU)
732 		bp->flags |= BNXT_FLAG_JUMBO;
733 	else
734 		bp->flags &= ~BNXT_FLAG_JUMBO;
735 
736 	/* THOR does not support ring groups.
737 	 * But we will use the array to save RSS context IDs.
738 	 */
739 	if (BNXT_CHIP_P5(bp))
740 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
741 
742 	rc = bnxt_alloc_hwrm_rings(bp);
743 	if (rc) {
744 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
745 		goto err_out;
746 	}
747 
748 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
749 	if (rc) {
750 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
751 		goto err_out;
752 	}
753 
754 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
755 		goto skip_cosq_cfg;
756 
757 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
758 		if (bp->rx_cos_queue[i].id != 0xff) {
759 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
760 
761 			if (!vnic) {
762 				PMD_DRV_LOG(ERR,
763 					    "Num pools more than FW profile\n");
764 				rc = -EINVAL;
765 				goto err_out;
766 			}
767 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
768 			bp->rx_cosq_cnt++;
769 		}
770 	}
771 
772 skip_cosq_cfg:
773 	rc = bnxt_mq_rx_configure(bp);
774 	if (rc) {
775 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
776 		goto err_out;
777 	}
778 
779 	for (j = 0; j < bp->rx_nr_rings; j++) {
780 		struct bnxt_rx_queue *rxq = bp->rx_queues[j];
781 
782 		if (!rxq->rx_deferred_start) {
783 			bp->eth_dev->data->rx_queue_state[j] =
784 				RTE_ETH_QUEUE_STATE_STARTED;
785 			rxq->rx_started = true;
786 		}
787 	}
788 
789 	/* VNIC configuration */
790 	for (i = 0; i < bp->nr_vnics; i++) {
791 		rc = bnxt_setup_one_vnic(bp, i);
792 		if (rc)
793 			goto err_out;
794 	}
795 
796 	for (j = 0; j < bp->tx_nr_rings; j++) {
797 		struct bnxt_tx_queue *txq = bp->tx_queues[j];
798 
799 		if (!txq->tx_deferred_start) {
800 			bp->eth_dev->data->tx_queue_state[j] =
801 				RTE_ETH_QUEUE_STATE_STARTED;
802 			txq->tx_started = true;
803 		}
804 	}
805 
806 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
807 	if (rc) {
808 		PMD_DRV_LOG(ERR,
809 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
810 		goto err_out;
811 	}
812 
813 	/* check and configure queue intr-vector mapping */
814 	if ((rte_intr_cap_multiple(intr_handle) ||
815 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
816 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
817 		intr_vector = bp->eth_dev->data->nb_rx_queues;
818 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
819 		if (intr_vector > bp->rx_cp_nr_rings) {
820 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
821 					bp->rx_cp_nr_rings);
822 			return -ENOTSUP;
823 		}
824 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
825 		if (rc)
826 			return rc;
827 	}
828 
829 	if (rte_intr_dp_is_en(intr_handle)) {
830 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
831 					bp->eth_dev->data->nb_rx_queues)) {
832 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
833 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
834 			rc = -ENOMEM;
835 			goto err_out;
836 		}
837 		PMD_DRV_LOG(DEBUG, "intr_handle->nb_efd = %d "
838 			    "intr_handle->max_intr = %d\n",
839 			    rte_intr_nb_efd_get(intr_handle),
840 			    rte_intr_max_intr_get(intr_handle));
841 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
842 		     queue_id++) {
843 			rte_intr_vec_list_index_set(intr_handle,
844 					queue_id, vec + BNXT_RX_VEC_START);
845 			if (vec < base + rte_intr_nb_efd_get(intr_handle)
846 			    - 1)
847 				vec++;
848 		}
849 	}
850 
851 	/* enable uio/vfio intr/eventfd mapping */
852 	rc = rte_intr_enable(intr_handle);
853 #ifndef RTE_EXEC_ENV_FREEBSD
854 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
855 	if (rc)
856 		goto err_out;
857 #endif
858 
859 	rc = bnxt_update_phy_setting(bp);
860 	if (rc)
861 		goto err_out;
862 
863 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
864 	if (!bp->mark_table)
865 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
866 
867 	return 0;
868 
869 err_out:
870 	/* Some of the error status returned by FW may not be from errno.h */
871 	if (rc > 0)
872 		rc = -EIO;
873 
874 	return rc;
875 }
876 
877 static int bnxt_shutdown_nic(struct bnxt *bp)
878 {
879 	bnxt_free_all_hwrm_resources(bp);
880 	bnxt_free_all_filters(bp);
881 	bnxt_free_all_vnics(bp);
882 	return 0;
883 }
884 
885 /*
886  * Device configuration and status function
887  */
888 
889 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
890 {
891 	uint32_t link_speed = 0;
892 	uint32_t speed_capa = 0;
893 
894 	if (bp->link_info == NULL)
895 		return 0;
896 
897 	link_speed = bp->link_info->support_speeds;
898 
899 	/* If PAM4 is configured, use PAM4 supported speed */
900 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
901 		link_speed = bp->link_info->support_pam4_speeds;
902 
903 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
904 		speed_capa |= RTE_ETH_LINK_SPEED_100M;
905 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
906 		speed_capa |= RTE_ETH_LINK_SPEED_100M_HD;
907 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
908 		speed_capa |= RTE_ETH_LINK_SPEED_1G;
909 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
910 		speed_capa |= RTE_ETH_LINK_SPEED_2_5G;
911 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
912 		speed_capa |= RTE_ETH_LINK_SPEED_10G;
913 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
914 		speed_capa |= RTE_ETH_LINK_SPEED_20G;
915 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
916 		speed_capa |= RTE_ETH_LINK_SPEED_25G;
917 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
918 		speed_capa |= RTE_ETH_LINK_SPEED_40G;
919 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
920 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
921 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
922 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
923 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
924 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
925 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
926 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
927 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
928 		speed_capa |= RTE_ETH_LINK_SPEED_200G;
929 
930 	if (bp->link_info->auto_mode ==
931 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
932 		speed_capa |= RTE_ETH_LINK_SPEED_FIXED;
933 
934 	return speed_capa;
935 }
936 
937 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
938 				struct rte_eth_dev_info *dev_info)
939 {
940 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
941 	struct bnxt *bp = eth_dev->data->dev_private;
942 	uint16_t max_vnics, i, j, vpool, vrxq;
943 	unsigned int max_rx_rings;
944 	int rc;
945 
946 	rc = is_bnxt_in_error(bp);
947 	if (rc)
948 		return rc;
949 
950 	/* MAC Specifics */
951 	dev_info->max_mac_addrs = RTE_MIN(bp->max_l2_ctx, RTE_ETH_NUM_RECEIVE_MAC_ADDR);
952 	dev_info->max_hash_mac_addrs = 0;
953 
954 	/* PF/VF specifics */
955 	if (BNXT_PF(bp))
956 		dev_info->max_vfs = pdev->max_vfs;
957 
958 	max_rx_rings = bnxt_max_rings(bp);
959 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
960 	dev_info->max_rx_queues = max_rx_rings;
961 	dev_info->max_tx_queues = max_rx_rings;
962 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
963 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
964 	max_vnics = bp->max_vnics;
965 
966 	/* MTU specifics */
967 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
968 	dev_info->max_mtu = BNXT_MAX_MTU;
969 
970 	/* Fast path specifics */
971 	dev_info->min_rx_bufsize = 1;
972 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
973 
974 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
975 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
976 		dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_TIMESTAMP;
977 	if (bp->vnic_cap_flags & BNXT_VNIC_CAP_VLAN_RX_STRIP)
978 		dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
979 	dev_info->tx_queue_offload_capa = RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
980 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
981 				    dev_info->tx_queue_offload_capa;
982 	if (bp->fw_cap & BNXT_FW_CAP_VLAN_TX_INSERT)
983 		dev_info->tx_offload_capa |= RTE_ETH_TX_OFFLOAD_VLAN_INSERT;
984 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
985 
986 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
987 	dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
988 			     RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
989 	dev_info->dev_capa &= ~RTE_ETH_DEV_CAPA_FLOW_RULE_KEEP;
990 
991 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
992 		.rx_thresh = {
993 			.pthresh = 8,
994 			.hthresh = 8,
995 			.wthresh = 0,
996 		},
997 		.rx_free_thresh = 32,
998 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
999 	};
1000 
1001 	dev_info->default_txconf = (struct rte_eth_txconf) {
1002 		.tx_thresh = {
1003 			.pthresh = 32,
1004 			.hthresh = 0,
1005 			.wthresh = 0,
1006 		},
1007 		.tx_free_thresh = 32,
1008 		.tx_rs_thresh = 32,
1009 	};
1010 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
1011 
1012 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1013 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1014 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1015 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1016 
1017 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1018 		dev_info->switch_info.name = eth_dev->device->name;
1019 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1020 		dev_info->switch_info.port_id =
1021 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1022 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1023 	}
1024 
1025 	/*
1026 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1027 	 *       need further investigation.
1028 	 */
1029 
1030 	/* VMDq resources */
1031 	vpool = 64; /* RTE_ETH_64_POOLS */
1032 	vrxq = 128; /* RTE_ETH_VMDQ_DCB_NUM_QUEUES */
1033 	for (i = 0; i < 4; vpool >>= 1, i++) {
1034 		if (max_vnics > vpool) {
1035 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1036 				if (dev_info->max_rx_queues > vrxq) {
1037 					if (vpool > vrxq)
1038 						vpool = vrxq;
1039 					goto found;
1040 				}
1041 			}
1042 			/* Not enough resources to support VMDq */
1043 			break;
1044 		}
1045 	}
1046 	/* Not enough resources to support VMDq */
1047 	vpool = 0;
1048 	vrxq = 0;
1049 found:
1050 	dev_info->max_vmdq_pools = vpool;
1051 	dev_info->vmdq_queue_num = vrxq;
1052 
1053 	dev_info->vmdq_pool_base = 0;
1054 	dev_info->vmdq_queue_base = 0;
1055 
1056 	return 0;
1057 }
1058 
1059 /* Configure the device based on the configuration provided */
1060 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1061 {
1062 	struct bnxt *bp = eth_dev->data->dev_private;
1063 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1064 	struct rte_eth_rss_conf *rss_conf = &eth_dev->data->dev_conf.rx_adv_conf.rss_conf;
1065 	int rc;
1066 
1067 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1068 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1069 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1070 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1071 
1072 	rc = is_bnxt_in_error(bp);
1073 	if (rc)
1074 		return rc;
1075 
1076 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1077 		rc = bnxt_hwrm_check_vf_rings(bp);
1078 		if (rc) {
1079 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1080 			return -ENOSPC;
1081 		}
1082 
1083 		/* If a resource has already been allocated - in this case
1084 		 * it is the async completion ring, free it. Reallocate it after
1085 		 * resource reservation. This will ensure the resource counts
1086 		 * are calculated correctly.
1087 		 */
1088 
1089 		pthread_mutex_lock(&bp->def_cp_lock);
1090 
1091 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1092 			bnxt_disable_int(bp);
1093 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1094 		}
1095 
1096 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1097 		if (rc) {
1098 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1099 			pthread_mutex_unlock(&bp->def_cp_lock);
1100 			return -ENOSPC;
1101 		}
1102 
1103 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1104 			rc = bnxt_alloc_async_cp_ring(bp);
1105 			if (rc) {
1106 				pthread_mutex_unlock(&bp->def_cp_lock);
1107 				return rc;
1108 			}
1109 			bnxt_enable_int(bp);
1110 		}
1111 
1112 		pthread_mutex_unlock(&bp->def_cp_lock);
1113 	}
1114 
1115 	/* Inherit new configurations */
1116 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1117 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1118 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1119 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1120 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1121 	    bp->max_stat_ctx)
1122 		goto resource_error;
1123 
1124 	if (BNXT_HAS_RING_GRPS(bp) &&
1125 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1126 		goto resource_error;
1127 
1128 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS) &&
1129 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1130 		goto resource_error;
1131 
1132 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1133 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1134 
1135 	if (eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
1136 		rx_offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
1137 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1138 
1139 	/* application provides the hash key to program */
1140 	if (rss_conf->rss_key != NULL) {
1141 		if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE)
1142 			PMD_DRV_LOG(WARNING, "port %u RSS key len must be %d bytes long",
1143 				    eth_dev->data->port_id, HW_HASH_KEY_SIZE);
1144 		else
1145 			memcpy(bp->rss_conf.rss_key, rss_conf->rss_key, HW_HASH_KEY_SIZE);
1146 	}
1147 	bp->rss_conf.rss_key_len = HW_HASH_KEY_SIZE;
1148 	bp->rss_conf.rss_hf = rss_conf->rss_hf;
1149 
1150 	bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1151 
1152 	return 0;
1153 
1154 resource_error:
1155 	PMD_DRV_LOG(ERR,
1156 		    "Insufficient resources to support requested config\n");
1157 	PMD_DRV_LOG(ERR,
1158 		    "Num Queues Requested: Tx %d, Rx %d\n",
1159 		    eth_dev->data->nb_tx_queues,
1160 		    eth_dev->data->nb_rx_queues);
1161 	PMD_DRV_LOG(ERR,
1162 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1163 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1164 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1165 	return -ENOSPC;
1166 }
1167 
1168 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1169 {
1170 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1171 
1172 	if (link->link_status)
1173 		PMD_DRV_LOG(DEBUG, "Port %d Link Up - speed %u Mbps - %s\n",
1174 			eth_dev->data->port_id,
1175 			(uint32_t)link->link_speed,
1176 			(link->link_duplex == RTE_ETH_LINK_FULL_DUPLEX) ?
1177 			("full-duplex") : ("half-duplex\n"));
1178 	else
1179 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1180 			eth_dev->data->port_id);
1181 }
1182 
1183 /*
1184  * Determine whether the current configuration requires support for scattered
1185  * receive; return 1 if scattered receive is required and 0 if not.
1186  */
1187 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1188 {
1189 	uint32_t overhead = BNXT_MAX_PKT_LEN - BNXT_MAX_MTU;
1190 	uint16_t buf_size;
1191 	int i;
1192 
1193 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_SCATTER)
1194 		return 1;
1195 
1196 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_TCP_LRO)
1197 		return 1;
1198 
1199 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1200 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1201 
1202 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1203 				      RTE_PKTMBUF_HEADROOM);
1204 		if (eth_dev->data->mtu + overhead > buf_size)
1205 			return 1;
1206 	}
1207 	return 0;
1208 }
1209 
1210 static eth_rx_burst_t
1211 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1212 {
1213 	struct bnxt *bp = eth_dev->data->dev_private;
1214 
1215 	/* Disable vector mode RX for Stingray2 for now */
1216 	if (BNXT_CHIP_SR2(bp)) {
1217 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1218 		return bnxt_recv_pkts;
1219 	}
1220 
1221 #if (defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)) && \
1222 	!defined(RTE_LIBRTE_IEEE1588)
1223 
1224 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1225 	if (eth_dev->data->scattered_rx)
1226 		goto use_scalar_rx;
1227 
1228 	/*
1229 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1230 	 * asynchronous completions and receive completions can be placed in
1231 	 * the same completion ring.
1232 	 */
1233 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1234 		goto use_scalar_rx;
1235 
1236 	/*
1237 	 * Vector mode receive cannot be enabled if any receive offloads outside
1238 	 * a limited subset have been enabled.
1239 	 */
1240 	if (eth_dev->data->dev_conf.rxmode.offloads &
1241 		~(RTE_ETH_RX_OFFLOAD_VLAN_STRIP |
1242 		  RTE_ETH_RX_OFFLOAD_KEEP_CRC |
1243 		  RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
1244 		  RTE_ETH_RX_OFFLOAD_UDP_CKSUM |
1245 		  RTE_ETH_RX_OFFLOAD_TCP_CKSUM |
1246 		  RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1247 		  RTE_ETH_RX_OFFLOAD_OUTER_UDP_CKSUM |
1248 		  RTE_ETH_RX_OFFLOAD_RSS_HASH |
1249 		  RTE_ETH_RX_OFFLOAD_VLAN_FILTER))
1250 		goto use_scalar_rx;
1251 
1252 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1253 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1254 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1255 		PMD_DRV_LOG(INFO,
1256 			    "Using AVX2 vector mode receive for port %d\n",
1257 			    eth_dev->data->port_id);
1258 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1259 		return bnxt_recv_pkts_vec_avx2;
1260 	}
1261  #endif
1262 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1263 		PMD_DRV_LOG(INFO,
1264 			    "Using SSE vector mode receive for port %d\n",
1265 			    eth_dev->data->port_id);
1266 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1267 		return bnxt_recv_pkts_vec;
1268 	}
1269 
1270 use_scalar_rx:
1271 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1272 		    eth_dev->data->port_id);
1273 	PMD_DRV_LOG(INFO,
1274 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1275 		    eth_dev->data->port_id,
1276 		    eth_dev->data->scattered_rx,
1277 		    eth_dev->data->dev_conf.rxmode.offloads);
1278 #endif
1279 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1280 	return bnxt_recv_pkts;
1281 }
1282 
1283 static eth_tx_burst_t
1284 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1285 {
1286 	struct bnxt *bp = eth_dev->data->dev_private;
1287 
1288 	/* Disable vector mode TX for Stingray2 for now */
1289 	if (BNXT_CHIP_SR2(bp))
1290 		return bnxt_xmit_pkts;
1291 
1292 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64) && \
1293 	!defined(RTE_LIBRTE_IEEE1588)
1294 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1295 
1296 	/*
1297 	 * Vector mode transmit can be enabled only if not using scatter rx
1298 	 * or tx offloads.
1299 	 */
1300 	if (eth_dev->data->scattered_rx ||
1301 	    (offloads & ~RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE) ||
1302 	    BNXT_TRUFLOW_EN(bp))
1303 		goto use_scalar_tx;
1304 
1305 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1306 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1307 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1308 		PMD_DRV_LOG(INFO,
1309 			    "Using AVX2 vector mode transmit for port %d\n",
1310 			    eth_dev->data->port_id);
1311 		return bnxt_xmit_pkts_vec_avx2;
1312 	}
1313 #endif
1314 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1315 		PMD_DRV_LOG(INFO,
1316 			    "Using SSE vector mode transmit for port %d\n",
1317 			    eth_dev->data->port_id);
1318 		return bnxt_xmit_pkts_vec;
1319 	}
1320 
1321 use_scalar_tx:
1322 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1323 		    eth_dev->data->port_id);
1324 	PMD_DRV_LOG(INFO,
1325 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1326 		    eth_dev->data->port_id,
1327 		    eth_dev->data->scattered_rx,
1328 		    offloads);
1329 #endif
1330 	return bnxt_xmit_pkts;
1331 }
1332 
1333 static int bnxt_handle_if_change_status(struct bnxt *bp)
1334 {
1335 	int rc;
1336 
1337 	/* Since fw has undergone a reset and lost all contexts,
1338 	 * set fatal flag to not issue hwrm during cleanup
1339 	 */
1340 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1341 	bnxt_uninit_resources(bp, true);
1342 
1343 	/* clear fatal flag so that re-init happens */
1344 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1345 	rc = bnxt_init_resources(bp, true);
1346 
1347 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1348 
1349 	return rc;
1350 }
1351 
1352 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1353 {
1354 	struct bnxt *bp = eth_dev->data->dev_private;
1355 	int rc = 0;
1356 
1357 	if (!BNXT_SINGLE_PF(bp))
1358 		return -ENOTSUP;
1359 
1360 	if (!bp->link_info->link_up)
1361 		rc = bnxt_set_hwrm_link_config(bp, true);
1362 	if (!rc)
1363 		eth_dev->data->dev_link.link_status = 1;
1364 
1365 	bnxt_print_link_info(eth_dev);
1366 	return rc;
1367 }
1368 
1369 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1370 {
1371 	struct bnxt *bp = eth_dev->data->dev_private;
1372 
1373 	if (!BNXT_SINGLE_PF(bp))
1374 		return -ENOTSUP;
1375 
1376 	eth_dev->data->dev_link.link_status = 0;
1377 	bnxt_set_hwrm_link_config(bp, false);
1378 	bp->link_info->link_up = 0;
1379 
1380 	return 0;
1381 }
1382 
1383 static void bnxt_free_switch_domain(struct bnxt *bp)
1384 {
1385 	int rc = 0;
1386 
1387 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1388 		return;
1389 
1390 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1391 	if (rc)
1392 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1393 			    bp->switch_domain_id, rc);
1394 }
1395 
1396 static void bnxt_ptp_get_current_time(void *arg)
1397 {
1398 	struct bnxt *bp = arg;
1399 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1400 	int rc;
1401 
1402 	rc = is_bnxt_in_error(bp);
1403 	if (rc)
1404 		return;
1405 
1406 	if (!ptp)
1407 		return;
1408 
1409 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1410 				&ptp->current_time);
1411 
1412 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1413 	if (rc != 0) {
1414 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1415 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1416 	}
1417 }
1418 
1419 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1420 {
1421 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1422 	int rc;
1423 
1424 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1425 		return 0;
1426 
1427 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1428 				&ptp->current_time);
1429 
1430 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1431 	return rc;
1432 }
1433 
1434 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1435 {
1436 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1437 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1438 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1439 	}
1440 }
1441 
1442 static void bnxt_ptp_stop(struct bnxt *bp)
1443 {
1444 	bnxt_cancel_ptp_alarm(bp);
1445 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1446 }
1447 
1448 static int bnxt_ptp_start(struct bnxt *bp)
1449 {
1450 	int rc;
1451 
1452 	rc = bnxt_schedule_ptp_alarm(bp);
1453 	if (rc != 0) {
1454 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1455 	} else {
1456 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1457 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1458 	}
1459 
1460 	return rc;
1461 }
1462 
1463 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1464 {
1465 	struct bnxt *bp = eth_dev->data->dev_private;
1466 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1467 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1468 	struct rte_eth_link link;
1469 	int ret;
1470 
1471 	eth_dev->data->dev_started = 0;
1472 
1473 	/* Prevent crashes when queues are still in use */
1474 	bnxt_stop_rxtx(eth_dev);
1475 
1476 	bnxt_disable_int(bp);
1477 
1478 	/* disable uio/vfio intr/eventfd mapping */
1479 	rte_intr_disable(intr_handle);
1480 
1481 	/* Stop the child representors for this device */
1482 	ret = bnxt_rep_stop_all(bp);
1483 	if (ret != 0)
1484 		return ret;
1485 
1486 	/* delete the bnxt ULP port details */
1487 	bnxt_ulp_port_deinit(bp);
1488 
1489 	bnxt_cancel_fw_health_check(bp);
1490 
1491 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1492 		bnxt_cancel_ptp_alarm(bp);
1493 
1494 	/* Do not bring link down during reset recovery */
1495 	if (!is_bnxt_in_error(bp)) {
1496 		bnxt_dev_set_link_down_op(eth_dev);
1497 		/* Wait for link to be reset */
1498 		if (BNXT_SINGLE_PF(bp))
1499 			rte_delay_ms(500);
1500 		/* clear the recorded link status */
1501 		memset(&link, 0, sizeof(link));
1502 		rte_eth_linkstatus_set(eth_dev, &link);
1503 	}
1504 
1505 	/* Clean queue intr-vector mapping */
1506 	rte_intr_efd_disable(intr_handle);
1507 	rte_intr_vec_list_free(intr_handle);
1508 
1509 	bnxt_hwrm_port_clr_stats(bp);
1510 	bnxt_free_tx_mbufs(bp);
1511 	bnxt_free_rx_mbufs(bp);
1512 	/* Process any remaining notifications in default completion queue */
1513 	bnxt_int_handler(eth_dev);
1514 	bnxt_shutdown_nic(bp);
1515 	bnxt_hwrm_if_change(bp, false);
1516 
1517 	bnxt_free_prev_ring_stats(bp);
1518 	rte_free(bp->mark_table);
1519 	bp->mark_table = NULL;
1520 
1521 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1522 	bp->rx_cosq_cnt = 0;
1523 	/* All filters are deleted on a port stop. */
1524 	if (BNXT_FLOW_XSTATS_EN(bp))
1525 		bp->flow_stat->flow_count = 0;
1526 
1527 	eth_dev->data->scattered_rx = 0;
1528 
1529 	return 0;
1530 }
1531 
1532 /* Unload the driver, release resources */
1533 int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1534 {
1535 	struct bnxt *bp = eth_dev->data->dev_private;
1536 
1537 	pthread_mutex_lock(&bp->err_recovery_lock);
1538 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1539 		PMD_DRV_LOG(ERR,
1540 			    "Adapter recovering from error..Please retry\n");
1541 		pthread_mutex_unlock(&bp->err_recovery_lock);
1542 		return -EAGAIN;
1543 	}
1544 	pthread_mutex_unlock(&bp->err_recovery_lock);
1545 
1546 	return bnxt_dev_stop(eth_dev);
1547 }
1548 
1549 int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1550 {
1551 	struct bnxt *bp = eth_dev->data->dev_private;
1552 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1553 	int vlan_mask = 0;
1554 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1555 
1556 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1557 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1558 		return -EINVAL;
1559 	}
1560 
1561 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1562 		PMD_DRV_LOG(ERR,
1563 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1564 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1565 
1566 	do {
1567 		rc = bnxt_hwrm_if_change(bp, true);
1568 		if (rc == 0 || rc != -EAGAIN)
1569 			break;
1570 
1571 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1572 	} while (retry_cnt--);
1573 
1574 	if (rc)
1575 		return rc;
1576 
1577 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1578 		rc = bnxt_handle_if_change_status(bp);
1579 		if (rc)
1580 			return rc;
1581 	}
1582 
1583 	bnxt_enable_int(bp);
1584 
1585 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1586 
1587 	rc = bnxt_start_nic(bp);
1588 	if (rc)
1589 		goto error;
1590 
1591 	rc = bnxt_alloc_prev_ring_stats(bp);
1592 	if (rc)
1593 		goto error;
1594 
1595 	eth_dev->data->dev_started = 1;
1596 
1597 	bnxt_link_update_op(eth_dev, 1);
1598 
1599 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
1600 		vlan_mask |= RTE_ETH_VLAN_FILTER_MASK;
1601 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
1602 		vlan_mask |= RTE_ETH_VLAN_STRIP_MASK;
1603 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1604 	if (rc)
1605 		goto error;
1606 
1607 	/* Initialize bnxt ULP port details */
1608 	rc = bnxt_ulp_port_init(bp);
1609 	if (rc)
1610 		goto error;
1611 
1612 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1613 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1614 
1615 	bnxt_schedule_fw_health_check(bp);
1616 
1617 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1618 		bnxt_schedule_ptp_alarm(bp);
1619 
1620 	return 0;
1621 
1622 error:
1623 	bnxt_dev_stop(eth_dev);
1624 	return rc;
1625 }
1626 
1627 static void
1628 bnxt_uninit_locks(struct bnxt *bp)
1629 {
1630 	pthread_mutex_destroy(&bp->flow_lock);
1631 	pthread_mutex_destroy(&bp->def_cp_lock);
1632 	pthread_mutex_destroy(&bp->health_check_lock);
1633 	pthread_mutex_destroy(&bp->err_recovery_lock);
1634 	if (bp->rep_info) {
1635 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1636 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1637 	}
1638 }
1639 
1640 static void bnxt_drv_uninit(struct bnxt *bp)
1641 {
1642 	bnxt_free_leds_info(bp);
1643 	bnxt_free_cos_queues(bp);
1644 	bnxt_free_link_info(bp);
1645 	bnxt_free_parent_info(bp);
1646 	bnxt_uninit_locks(bp);
1647 
1648 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1649 	bp->tx_mem_zone = NULL;
1650 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1651 	bp->rx_mem_zone = NULL;
1652 
1653 	bnxt_free_vf_info(bp);
1654 	bnxt_free_pf_info(bp);
1655 
1656 	rte_free(bp->grp_info);
1657 	bp->grp_info = NULL;
1658 }
1659 
1660 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1661 {
1662 	struct bnxt *bp = eth_dev->data->dev_private;
1663 	int ret = 0;
1664 
1665 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1666 		return 0;
1667 
1668 	pthread_mutex_lock(&bp->err_recovery_lock);
1669 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1670 		PMD_DRV_LOG(ERR,
1671 			    "Adapter recovering from error...Please retry\n");
1672 		pthread_mutex_unlock(&bp->err_recovery_lock);
1673 		return -EAGAIN;
1674 	}
1675 	pthread_mutex_unlock(&bp->err_recovery_lock);
1676 
1677 	/* cancel the recovery handler before remove dev */
1678 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1679 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1680 	bnxt_cancel_fc_thread(bp);
1681 	rte_eal_alarm_cancel(bnxt_handle_vf_cfg_change, (void *)bp);
1682 
1683 	if (eth_dev->data->dev_started)
1684 		ret = bnxt_dev_stop(eth_dev);
1685 
1686 	bnxt_uninit_resources(bp, false);
1687 
1688 	bnxt_drv_uninit(bp);
1689 
1690 	return ret;
1691 }
1692 
1693 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1694 				    uint32_t index)
1695 {
1696 	struct bnxt *bp = eth_dev->data->dev_private;
1697 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1698 	struct bnxt_vnic_info *vnic;
1699 	struct bnxt_filter_info *filter, *temp_filter;
1700 	uint32_t i;
1701 
1702 	if (is_bnxt_in_error(bp))
1703 		return;
1704 
1705 	/*
1706 	 * Loop through all VNICs from the specified filter flow pools to
1707 	 * remove the corresponding MAC addr filter
1708 	 */
1709 	for (i = 0; i < bp->nr_vnics; i++) {
1710 		if (!(pool_mask & (1ULL << i)))
1711 			continue;
1712 
1713 		vnic = &bp->vnic_info[i];
1714 		filter = STAILQ_FIRST(&vnic->filter);
1715 		while (filter) {
1716 			temp_filter = STAILQ_NEXT(filter, next);
1717 			if (filter->mac_index == index) {
1718 				STAILQ_REMOVE(&vnic->filter, filter,
1719 						bnxt_filter_info, next);
1720 				bnxt_hwrm_clear_l2_filter(bp, filter);
1721 				bnxt_free_filter(bp, filter);
1722 			}
1723 			filter = temp_filter;
1724 		}
1725 	}
1726 }
1727 
1728 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1729 			       struct rte_ether_addr *mac_addr, uint32_t index,
1730 			       uint32_t pool)
1731 {
1732 	struct bnxt_filter_info *filter;
1733 	int rc = 0;
1734 
1735 	/* Attach requested MAC address to the new l2_filter */
1736 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1737 		if (filter->mac_index == index) {
1738 			PMD_DRV_LOG(DEBUG,
1739 				    "MAC addr already existed for pool %d\n",
1740 				    pool);
1741 			return 0;
1742 		}
1743 	}
1744 
1745 	filter = bnxt_alloc_filter(bp);
1746 	if (!filter) {
1747 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1748 		return -ENODEV;
1749 	}
1750 
1751 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1752 	 * if the MAC that's been programmed now is a different one, then,
1753 	 * copy that addr to filter->l2_addr
1754 	 */
1755 	if (mac_addr)
1756 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1757 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1758 
1759 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1760 	if (!rc) {
1761 		filter->mac_index = index;
1762 		if (filter->mac_index == 0)
1763 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1764 		else
1765 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1766 	} else {
1767 		bnxt_free_filter(bp, filter);
1768 	}
1769 
1770 	return rc;
1771 }
1772 
1773 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1774 				struct rte_ether_addr *mac_addr,
1775 				uint32_t index, uint32_t pool)
1776 {
1777 	struct bnxt *bp = eth_dev->data->dev_private;
1778 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1779 	int rc = 0;
1780 
1781 	rc = is_bnxt_in_error(bp);
1782 	if (rc)
1783 		return rc;
1784 
1785 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1786 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1787 		return -ENOTSUP;
1788 	}
1789 
1790 	if (!vnic) {
1791 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1792 		return -EINVAL;
1793 	}
1794 
1795 	/* Filter settings will get applied when port is started */
1796 	if (!eth_dev->data->dev_started)
1797 		return 0;
1798 
1799 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1800 
1801 	return rc;
1802 }
1803 
1804 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1805 {
1806 	int rc = 0;
1807 	struct bnxt *bp = eth_dev->data->dev_private;
1808 	struct rte_eth_link new;
1809 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1810 			BNXT_MIN_LINK_WAIT_CNT;
1811 
1812 	rc = is_bnxt_in_error(bp);
1813 	if (rc)
1814 		return rc;
1815 
1816 	memset(&new, 0, sizeof(new));
1817 
1818 	if (bp->link_info == NULL)
1819 		goto out;
1820 
1821 	do {
1822 		/* Retrieve link info from hardware */
1823 		rc = bnxt_get_hwrm_link_config(bp, &new);
1824 		if (rc) {
1825 			new.link_speed = RTE_ETH_LINK_SPEED_100M;
1826 			new.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
1827 			PMD_DRV_LOG(ERR,
1828 				"Failed to retrieve link rc = 0x%x!\n", rc);
1829 			goto out;
1830 		}
1831 
1832 		if (!wait_to_complete || new.link_status)
1833 			break;
1834 
1835 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1836 	} while (cnt--);
1837 
1838 	/* Only single function PF can bring phy down.
1839 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1840 	 */
1841 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1842 		memset(&new, 0, sizeof(new));
1843 
1844 out:
1845 	/* Timed out or success */
1846 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1847 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1848 		rte_eth_linkstatus_set(eth_dev, &new);
1849 		bnxt_print_link_info(eth_dev);
1850 	}
1851 
1852 	return rc;
1853 }
1854 
1855 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1856 {
1857 	struct bnxt *bp = eth_dev->data->dev_private;
1858 	struct bnxt_vnic_info *vnic;
1859 	uint32_t old_flags;
1860 	int rc;
1861 
1862 	rc = is_bnxt_in_error(bp);
1863 	if (rc)
1864 		return rc;
1865 
1866 	/* Filter settings will get applied when port is started */
1867 	if (!eth_dev->data->dev_started)
1868 		return 0;
1869 
1870 	if (bp->vnic_info == NULL)
1871 		return 0;
1872 
1873 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1874 
1875 	old_flags = vnic->flags;
1876 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1877 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1878 	if (rc != 0)
1879 		vnic->flags = old_flags;
1880 
1881 	return rc;
1882 }
1883 
1884 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1885 {
1886 	struct bnxt *bp = eth_dev->data->dev_private;
1887 	struct bnxt_vnic_info *vnic;
1888 	uint32_t old_flags;
1889 	int rc;
1890 
1891 	rc = is_bnxt_in_error(bp);
1892 	if (rc)
1893 		return rc;
1894 
1895 	/* Filter settings will get applied when port is started */
1896 	if (!eth_dev->data->dev_started)
1897 		return 0;
1898 
1899 	if (bp->vnic_info == NULL)
1900 		return 0;
1901 
1902 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1903 
1904 	old_flags = vnic->flags;
1905 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1906 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1907 	if (rc != 0)
1908 		vnic->flags = old_flags;
1909 
1910 	return rc;
1911 }
1912 
1913 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1914 {
1915 	struct bnxt *bp = eth_dev->data->dev_private;
1916 	struct bnxt_vnic_info *vnic;
1917 	uint32_t old_flags;
1918 	int rc;
1919 
1920 	rc = is_bnxt_in_error(bp);
1921 	if (rc)
1922 		return rc;
1923 
1924 	/* Filter settings will get applied when port is started */
1925 	if (!eth_dev->data->dev_started)
1926 		return 0;
1927 
1928 	if (bp->vnic_info == NULL)
1929 		return 0;
1930 
1931 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1932 
1933 	old_flags = vnic->flags;
1934 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1935 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1936 	if (rc != 0)
1937 		vnic->flags = old_flags;
1938 
1939 	return rc;
1940 }
1941 
1942 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1943 {
1944 	struct bnxt *bp = eth_dev->data->dev_private;
1945 	struct bnxt_vnic_info *vnic;
1946 	uint32_t old_flags;
1947 	int rc;
1948 
1949 	rc = is_bnxt_in_error(bp);
1950 	if (rc)
1951 		return rc;
1952 
1953 	/* Filter settings will get applied when port is started */
1954 	if (!eth_dev->data->dev_started)
1955 		return 0;
1956 
1957 	if (bp->vnic_info == NULL)
1958 		return 0;
1959 
1960 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1961 
1962 	old_flags = vnic->flags;
1963 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1964 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1965 	if (rc != 0)
1966 		vnic->flags = old_flags;
1967 
1968 	return rc;
1969 }
1970 
1971 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1972 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1973 {
1974 	if (qid >= bp->rx_nr_rings)
1975 		return NULL;
1976 
1977 	return bp->eth_dev->data->rx_queues[qid];
1978 }
1979 
1980 /* Return rxq corresponding to a given rss table ring/group ID. */
1981 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1982 {
1983 	struct bnxt_rx_queue *rxq;
1984 	unsigned int i;
1985 
1986 	if (!BNXT_HAS_RING_GRPS(bp)) {
1987 		for (i = 0; i < bp->rx_nr_rings; i++) {
1988 			rxq = bp->eth_dev->data->rx_queues[i];
1989 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1990 				return rxq->index;
1991 		}
1992 	} else {
1993 		for (i = 0; i < bp->rx_nr_rings; i++) {
1994 			if (bp->grp_info[i].fw_grp_id == fwr)
1995 				return i;
1996 		}
1997 	}
1998 
1999 	return INVALID_HW_RING_ID;
2000 }
2001 
2002 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
2003 			    struct rte_eth_rss_reta_entry64 *reta_conf,
2004 			    uint16_t reta_size)
2005 {
2006 	struct bnxt *bp = eth_dev->data->dev_private;
2007 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2008 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2009 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2010 	uint16_t idx, sft;
2011 	int i, rc;
2012 
2013 	rc = is_bnxt_in_error(bp);
2014 	if (rc)
2015 		return rc;
2016 
2017 	if (!vnic->rss_table)
2018 		return -EINVAL;
2019 
2020 	if (!(dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG))
2021 		return -EINVAL;
2022 
2023 	if (reta_size != tbl_size) {
2024 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2025 			"(%d) must equal the size supported by the hardware "
2026 			"(%d)\n", reta_size, tbl_size);
2027 		return -EINVAL;
2028 	}
2029 
2030 	for (i = 0; i < reta_size; i++) {
2031 		struct bnxt_rx_queue *rxq;
2032 
2033 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2034 		sft = i % RTE_ETH_RETA_GROUP_SIZE;
2035 
2036 		if (!(reta_conf[idx].mask & (1ULL << sft)))
2037 			continue;
2038 
2039 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2040 		if (!rxq) {
2041 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
2042 			return -EINVAL;
2043 		}
2044 
2045 		if (BNXT_CHIP_P5(bp)) {
2046 			vnic->rss_table[i * 2] =
2047 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
2048 			vnic->rss_table[i * 2 + 1] =
2049 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
2050 		} else {
2051 			vnic->rss_table[i] =
2052 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2053 		}
2054 	}
2055 
2056 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2057 	return rc;
2058 }
2059 
2060 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2061 			      struct rte_eth_rss_reta_entry64 *reta_conf,
2062 			      uint16_t reta_size)
2063 {
2064 	struct bnxt *bp = eth_dev->data->dev_private;
2065 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2066 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2067 	uint16_t idx, sft, i;
2068 	int rc;
2069 
2070 	rc = is_bnxt_in_error(bp);
2071 	if (rc)
2072 		return rc;
2073 
2074 	if (!vnic)
2075 		return -EINVAL;
2076 	if (!vnic->rss_table)
2077 		return -EINVAL;
2078 
2079 	if (reta_size != tbl_size) {
2080 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2081 			"(%d) must equal the size supported by the hardware "
2082 			"(%d)\n", reta_size, tbl_size);
2083 		return -EINVAL;
2084 	}
2085 
2086 	for (idx = 0, i = 0; i < reta_size; i++) {
2087 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2088 		sft = i % RTE_ETH_RETA_GROUP_SIZE;
2089 
2090 		if (reta_conf[idx].mask & (1ULL << sft)) {
2091 			uint16_t qid;
2092 
2093 			if (BNXT_CHIP_P5(bp))
2094 				qid = bnxt_rss_to_qid(bp,
2095 						      vnic->rss_table[i * 2]);
2096 			else
2097 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2098 
2099 			if (qid == INVALID_HW_RING_ID) {
2100 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2101 				return -EINVAL;
2102 			}
2103 			reta_conf[idx].reta[sft] = qid;
2104 		}
2105 	}
2106 
2107 	return 0;
2108 }
2109 
2110 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2111 				   struct rte_eth_rss_conf *rss_conf)
2112 {
2113 	struct bnxt *bp = eth_dev->data->dev_private;
2114 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2115 	struct bnxt_vnic_info *vnic;
2116 	int rc;
2117 
2118 	rc = is_bnxt_in_error(bp);
2119 	if (rc)
2120 		return rc;
2121 
2122 	/*
2123 	 * If RSS enablement were different than dev_configure,
2124 	 * then return -EINVAL
2125 	 */
2126 	if (dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG) {
2127 		if (!rss_conf->rss_hf)
2128 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2129 	} else {
2130 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2131 			return -EINVAL;
2132 	}
2133 
2134 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
2135 
2136 	/* Update the default RSS VNIC(s) */
2137 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2138 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2139 	vnic->hash_mode =
2140 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2141 					    RTE_ETH_RSS_LEVEL(rss_conf->rss_hf));
2142 
2143 	/* Cache the hash function */
2144 	bp->rss_conf.rss_hf = rss_conf->rss_hf;
2145 
2146 	/*
2147 	 * If hashkey is not specified, use the previously configured
2148 	 * hashkey
2149 	 */
2150 	if (!rss_conf->rss_key)
2151 		goto rss_config;
2152 
2153 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2154 		PMD_DRV_LOG(ERR,
2155 			    "Invalid hashkey length, should be %d bytes\n",
2156 			    HW_HASH_KEY_SIZE);
2157 		return -EINVAL;
2158 	}
2159 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2160 
2161 	/* Cache the hash key */
2162 	memcpy(bp->rss_conf.rss_key, rss_conf->rss_key, HW_HASH_KEY_SIZE);
2163 
2164 rss_config:
2165 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2166 	return rc;
2167 }
2168 
2169 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2170 				     struct rte_eth_rss_conf *rss_conf)
2171 {
2172 	struct bnxt *bp = eth_dev->data->dev_private;
2173 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2174 	int len, rc;
2175 	uint32_t hash_types;
2176 
2177 	rc = is_bnxt_in_error(bp);
2178 	if (rc)
2179 		return rc;
2180 
2181 	/* RSS configuration is the same for all VNICs */
2182 	if (vnic && vnic->rss_hash_key) {
2183 		if (rss_conf->rss_key) {
2184 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2185 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2186 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2187 		}
2188 
2189 		hash_types = vnic->hash_type;
2190 		rss_conf->rss_hf = 0;
2191 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2192 			rss_conf->rss_hf |= RTE_ETH_RSS_IPV4;
2193 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2194 		}
2195 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2196 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_TCP;
2197 			hash_types &=
2198 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2199 		}
2200 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2201 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_UDP;
2202 			hash_types &=
2203 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2204 		}
2205 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2206 			rss_conf->rss_hf |= RTE_ETH_RSS_IPV6;
2207 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2208 		}
2209 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2210 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_TCP;
2211 			hash_types &=
2212 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2213 		}
2214 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2215 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_UDP;
2216 			hash_types &=
2217 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2218 		}
2219 
2220 		rss_conf->rss_hf |=
2221 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2222 
2223 		if (hash_types) {
2224 			PMD_DRV_LOG(ERR,
2225 				"Unknown RSS config from firmware (%08x), RSS disabled",
2226 				vnic->hash_type);
2227 			return -ENOTSUP;
2228 		}
2229 	} else {
2230 		rss_conf->rss_hf = 0;
2231 	}
2232 	return 0;
2233 }
2234 
2235 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2236 			       struct rte_eth_fc_conf *fc_conf)
2237 {
2238 	struct bnxt *bp = dev->data->dev_private;
2239 	struct rte_eth_link link_info;
2240 	int rc;
2241 
2242 	rc = is_bnxt_in_error(bp);
2243 	if (rc)
2244 		return rc;
2245 
2246 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2247 	if (rc)
2248 		return rc;
2249 
2250 	memset(fc_conf, 0, sizeof(*fc_conf));
2251 	if (bp->link_info->auto_pause)
2252 		fc_conf->autoneg = 1;
2253 	switch (bp->link_info->pause) {
2254 	case 0:
2255 		fc_conf->mode = RTE_ETH_FC_NONE;
2256 		break;
2257 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2258 		fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
2259 		break;
2260 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2261 		fc_conf->mode = RTE_ETH_FC_RX_PAUSE;
2262 		break;
2263 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2264 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2265 		fc_conf->mode = RTE_ETH_FC_FULL;
2266 		break;
2267 	}
2268 	return 0;
2269 }
2270 
2271 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2272 			       struct rte_eth_fc_conf *fc_conf)
2273 {
2274 	struct bnxt *bp = dev->data->dev_private;
2275 	int rc;
2276 
2277 	rc = is_bnxt_in_error(bp);
2278 	if (rc)
2279 		return rc;
2280 
2281 	if (!BNXT_SINGLE_PF(bp)) {
2282 		PMD_DRV_LOG(ERR,
2283 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2284 		return -ENOTSUP;
2285 	}
2286 
2287 	switch (fc_conf->mode) {
2288 	case RTE_ETH_FC_NONE:
2289 		bp->link_info->auto_pause = 0;
2290 		bp->link_info->force_pause = 0;
2291 		break;
2292 	case RTE_ETH_FC_RX_PAUSE:
2293 		if (fc_conf->autoneg) {
2294 			bp->link_info->auto_pause =
2295 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2296 			bp->link_info->force_pause = 0;
2297 		} else {
2298 			bp->link_info->auto_pause = 0;
2299 			bp->link_info->force_pause =
2300 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2301 		}
2302 		break;
2303 	case RTE_ETH_FC_TX_PAUSE:
2304 		if (fc_conf->autoneg) {
2305 			bp->link_info->auto_pause =
2306 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2307 			bp->link_info->force_pause = 0;
2308 		} else {
2309 			bp->link_info->auto_pause = 0;
2310 			bp->link_info->force_pause =
2311 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2312 		}
2313 		break;
2314 	case RTE_ETH_FC_FULL:
2315 		if (fc_conf->autoneg) {
2316 			bp->link_info->auto_pause =
2317 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2318 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2319 			bp->link_info->force_pause = 0;
2320 		} else {
2321 			bp->link_info->auto_pause = 0;
2322 			bp->link_info->force_pause =
2323 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2324 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2325 		}
2326 		break;
2327 	}
2328 	return bnxt_set_hwrm_link_config(bp, true);
2329 }
2330 
2331 /* Add UDP tunneling port */
2332 static int
2333 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2334 			 struct rte_eth_udp_tunnel *udp_tunnel)
2335 {
2336 	struct bnxt *bp = eth_dev->data->dev_private;
2337 	uint16_t tunnel_type = 0;
2338 	int rc = 0;
2339 
2340 	rc = is_bnxt_in_error(bp);
2341 	if (rc)
2342 		return rc;
2343 
2344 	switch (udp_tunnel->prot_type) {
2345 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
2346 		if (bp->vxlan_port_cnt) {
2347 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2348 				udp_tunnel->udp_port);
2349 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2350 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2351 				return -ENOSPC;
2352 			}
2353 			bp->vxlan_port_cnt++;
2354 			return 0;
2355 		}
2356 		tunnel_type =
2357 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2358 		break;
2359 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2360 		if (bp->geneve_port_cnt) {
2361 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2362 				udp_tunnel->udp_port);
2363 			if (bp->geneve_port != udp_tunnel->udp_port) {
2364 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2365 				return -ENOSPC;
2366 			}
2367 			bp->geneve_port_cnt++;
2368 			return 0;
2369 		}
2370 		tunnel_type =
2371 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2372 		break;
2373 	default:
2374 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2375 		return -ENOTSUP;
2376 	}
2377 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2378 					     tunnel_type);
2379 
2380 	if (rc != 0)
2381 		return rc;
2382 
2383 	if (tunnel_type ==
2384 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN)
2385 		bp->vxlan_port_cnt++;
2386 
2387 	if (tunnel_type ==
2388 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE)
2389 		bp->geneve_port_cnt++;
2390 
2391 	return rc;
2392 }
2393 
2394 static int
2395 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2396 			 struct rte_eth_udp_tunnel *udp_tunnel)
2397 {
2398 	struct bnxt *bp = eth_dev->data->dev_private;
2399 	uint16_t tunnel_type = 0;
2400 	uint16_t port = 0;
2401 	int rc = 0;
2402 
2403 	rc = is_bnxt_in_error(bp);
2404 	if (rc)
2405 		return rc;
2406 
2407 	switch (udp_tunnel->prot_type) {
2408 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
2409 		if (!bp->vxlan_port_cnt) {
2410 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2411 			return -EINVAL;
2412 		}
2413 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2414 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2415 				udp_tunnel->udp_port, bp->vxlan_port);
2416 			return -EINVAL;
2417 		}
2418 		if (--bp->vxlan_port_cnt)
2419 			return 0;
2420 
2421 		tunnel_type =
2422 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2423 		port = bp->vxlan_fw_dst_port_id;
2424 		break;
2425 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2426 		if (!bp->geneve_port_cnt) {
2427 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2428 			return -EINVAL;
2429 		}
2430 		if (bp->geneve_port != udp_tunnel->udp_port) {
2431 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2432 				udp_tunnel->udp_port, bp->geneve_port);
2433 			return -EINVAL;
2434 		}
2435 		if (--bp->geneve_port_cnt)
2436 			return 0;
2437 
2438 		tunnel_type =
2439 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2440 		port = bp->geneve_fw_dst_port_id;
2441 		break;
2442 	default:
2443 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2444 		return -ENOTSUP;
2445 	}
2446 
2447 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2448 	return rc;
2449 }
2450 
2451 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2452 {
2453 	struct bnxt_filter_info *filter;
2454 	struct bnxt_vnic_info *vnic;
2455 	int rc = 0;
2456 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2457 
2458 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2459 	filter = STAILQ_FIRST(&vnic->filter);
2460 	while (filter) {
2461 		/* Search for this matching MAC+VLAN filter */
2462 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2463 			/* Delete the filter */
2464 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2465 			if (rc)
2466 				return rc;
2467 			STAILQ_REMOVE(&vnic->filter, filter,
2468 				      bnxt_filter_info, next);
2469 			bnxt_free_filter(bp, filter);
2470 			PMD_DRV_LOG(INFO,
2471 				    "Deleted vlan filter for %d\n",
2472 				    vlan_id);
2473 			return 0;
2474 		}
2475 		filter = STAILQ_NEXT(filter, next);
2476 	}
2477 	return -ENOENT;
2478 }
2479 
2480 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2481 {
2482 	struct bnxt_filter_info *filter;
2483 	struct bnxt_vnic_info *vnic;
2484 	int rc = 0;
2485 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2486 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2487 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2488 
2489 	/* Implementation notes on the use of VNIC in this command:
2490 	 *
2491 	 * By default, these filters belong to default vnic for the function.
2492 	 * Once these filters are set up, only destination VNIC can be modified.
2493 	 * If the destination VNIC is not specified in this command,
2494 	 * then the HWRM shall only create an l2 context id.
2495 	 */
2496 
2497 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2498 	filter = STAILQ_FIRST(&vnic->filter);
2499 	/* Check if the VLAN has already been added */
2500 	while (filter) {
2501 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2502 			return -EEXIST;
2503 
2504 		filter = STAILQ_NEXT(filter, next);
2505 	}
2506 
2507 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2508 	 * command to create MAC+VLAN filter with the right flags, enables set.
2509 	 */
2510 	filter = bnxt_alloc_filter(bp);
2511 	if (!filter) {
2512 		PMD_DRV_LOG(ERR,
2513 			    "MAC/VLAN filter alloc failed\n");
2514 		return -ENOMEM;
2515 	}
2516 	/* MAC + VLAN ID filter */
2517 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2518 	 * untagged packets are received
2519 	 *
2520 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2521 	 * packets and only the programmed vlan's packets are received
2522 	 */
2523 	filter->l2_ivlan = vlan_id;
2524 	filter->l2_ivlan_mask = 0x0FFF;
2525 	filter->enables |= en;
2526 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2527 
2528 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2529 	if (rc) {
2530 		/* Free the newly allocated filter as we were
2531 		 * not able to create the filter in hardware.
2532 		 */
2533 		bnxt_free_filter(bp, filter);
2534 		return rc;
2535 	}
2536 
2537 	filter->mac_index = 0;
2538 	/* Add this new filter to the list */
2539 	if (vlan_id == 0)
2540 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2541 	else
2542 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2543 
2544 	PMD_DRV_LOG(INFO,
2545 		    "Added Vlan filter for %d\n", vlan_id);
2546 	return rc;
2547 }
2548 
2549 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2550 		uint16_t vlan_id, int on)
2551 {
2552 	struct bnxt *bp = eth_dev->data->dev_private;
2553 	int rc;
2554 
2555 	rc = is_bnxt_in_error(bp);
2556 	if (rc)
2557 		return rc;
2558 
2559 	if (!eth_dev->data->dev_started) {
2560 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2561 		return -EINVAL;
2562 	}
2563 
2564 	/* These operations apply to ALL existing MAC/VLAN filters */
2565 	if (on)
2566 		return bnxt_add_vlan_filter(bp, vlan_id);
2567 	else
2568 		return bnxt_del_vlan_filter(bp, vlan_id);
2569 }
2570 
2571 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2572 				    struct bnxt_vnic_info *vnic)
2573 {
2574 	struct bnxt_filter_info *filter;
2575 	int rc;
2576 
2577 	filter = STAILQ_FIRST(&vnic->filter);
2578 	while (filter) {
2579 		if (filter->mac_index == 0 &&
2580 		    !memcmp(filter->l2_addr, bp->mac_addr,
2581 			    RTE_ETHER_ADDR_LEN)) {
2582 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2583 			if (!rc) {
2584 				STAILQ_REMOVE(&vnic->filter, filter,
2585 					      bnxt_filter_info, next);
2586 				bnxt_free_filter(bp, filter);
2587 			}
2588 			return rc;
2589 		}
2590 		filter = STAILQ_NEXT(filter, next);
2591 	}
2592 	return 0;
2593 }
2594 
2595 static int
2596 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2597 {
2598 	struct bnxt_vnic_info *vnic;
2599 	unsigned int i;
2600 	int rc;
2601 
2602 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2603 	if (!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)) {
2604 		/* Remove any VLAN filters programmed */
2605 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2606 			bnxt_del_vlan_filter(bp, i);
2607 
2608 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2609 		if (rc)
2610 			return rc;
2611 	} else {
2612 		/* Default filter will allow packets that match the
2613 		 * dest mac. So, it has to be deleted, otherwise, we
2614 		 * will endup receiving vlan packets for which the
2615 		 * filter is not programmed, when hw-vlan-filter
2616 		 * configuration is ON
2617 		 */
2618 		bnxt_del_dflt_mac_filter(bp, vnic);
2619 		/* This filter will allow only untagged packets */
2620 		bnxt_add_vlan_filter(bp, 0);
2621 	}
2622 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2623 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER));
2624 
2625 	return 0;
2626 }
2627 
2628 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2629 {
2630 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2631 	unsigned int i;
2632 	int rc;
2633 
2634 	/* Destroy vnic filters and vnic */
2635 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2636 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2637 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2638 			bnxt_del_vlan_filter(bp, i);
2639 	}
2640 	bnxt_del_dflt_mac_filter(bp, vnic);
2641 
2642 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2643 	if (rc)
2644 		return rc;
2645 
2646 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2647 	if (rc)
2648 		return rc;
2649 
2650 	rte_free(vnic->fw_grp_ids);
2651 	vnic->fw_grp_ids = NULL;
2652 
2653 	vnic->rx_queue_cnt = 0;
2654 
2655 	return 0;
2656 }
2657 
2658 static int
2659 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2660 {
2661 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2662 	int rc;
2663 
2664 	/* Destroy, recreate and reconfigure the default vnic */
2665 	rc = bnxt_free_one_vnic(bp, 0);
2666 	if (rc)
2667 		return rc;
2668 
2669 	/* default vnic 0 */
2670 	rc = bnxt_setup_one_vnic(bp, 0);
2671 	if (rc)
2672 		return rc;
2673 
2674 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2675 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2676 		rc = bnxt_add_vlan_filter(bp, 0);
2677 		if (rc)
2678 			return rc;
2679 		rc = bnxt_restore_vlan_filters(bp);
2680 		if (rc)
2681 			return rc;
2682 	} else {
2683 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2684 		if (rc)
2685 			return rc;
2686 	}
2687 
2688 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2689 	if (rc)
2690 		return rc;
2691 
2692 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2693 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP));
2694 
2695 	return rc;
2696 }
2697 
2698 static int
2699 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2700 {
2701 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2702 	struct bnxt *bp = dev->data->dev_private;
2703 	int rc;
2704 
2705 	rc = is_bnxt_in_error(bp);
2706 	if (rc)
2707 		return rc;
2708 
2709 	/* Filter settings will get applied when port is started */
2710 	if (!dev->data->dev_started)
2711 		return 0;
2712 
2713 	if (mask & RTE_ETH_VLAN_FILTER_MASK) {
2714 		/* Enable or disable VLAN filtering */
2715 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2716 		if (rc)
2717 			return rc;
2718 	}
2719 
2720 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
2721 		/* Enable or disable VLAN stripping */
2722 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2723 		if (rc)
2724 			return rc;
2725 	}
2726 
2727 	if (mask & RTE_ETH_VLAN_EXTEND_MASK) {
2728 		if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND)
2729 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2730 		else
2731 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2732 	}
2733 
2734 	return 0;
2735 }
2736 
2737 static int
2738 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2739 		      uint16_t tpid)
2740 {
2741 	struct bnxt *bp = dev->data->dev_private;
2742 	int qinq = dev->data->dev_conf.rxmode.offloads &
2743 		   RTE_ETH_RX_OFFLOAD_VLAN_EXTEND;
2744 
2745 	if (vlan_type != RTE_ETH_VLAN_TYPE_INNER &&
2746 	    vlan_type != RTE_ETH_VLAN_TYPE_OUTER) {
2747 		PMD_DRV_LOG(ERR,
2748 			    "Unsupported vlan type.");
2749 		return -EINVAL;
2750 	}
2751 	if (!qinq) {
2752 		PMD_DRV_LOG(ERR,
2753 			    "QinQ not enabled. Needs to be ON as we can "
2754 			    "accelerate only outer vlan\n");
2755 		return -EINVAL;
2756 	}
2757 
2758 	if (vlan_type == RTE_ETH_VLAN_TYPE_OUTER) {
2759 		switch (tpid) {
2760 		case RTE_ETHER_TYPE_QINQ:
2761 			bp->outer_tpid_bd =
2762 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2763 				break;
2764 		case RTE_ETHER_TYPE_VLAN:
2765 			bp->outer_tpid_bd =
2766 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2767 				break;
2768 		case RTE_ETHER_TYPE_QINQ1:
2769 			bp->outer_tpid_bd =
2770 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2771 				break;
2772 		case RTE_ETHER_TYPE_QINQ2:
2773 			bp->outer_tpid_bd =
2774 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2775 				break;
2776 		case RTE_ETHER_TYPE_QINQ3:
2777 			bp->outer_tpid_bd =
2778 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2779 				break;
2780 		default:
2781 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2782 			return -EINVAL;
2783 		}
2784 		bp->outer_tpid_bd |= tpid;
2785 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2786 	} else if (vlan_type == RTE_ETH_VLAN_TYPE_INNER) {
2787 		PMD_DRV_LOG(ERR,
2788 			    "Can accelerate only outer vlan in QinQ\n");
2789 		return -EINVAL;
2790 	}
2791 
2792 	return 0;
2793 }
2794 
2795 static int
2796 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2797 			     struct rte_ether_addr *addr)
2798 {
2799 	struct bnxt *bp = dev->data->dev_private;
2800 	/* Default Filter is tied to VNIC 0 */
2801 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2802 	int rc;
2803 
2804 	rc = is_bnxt_in_error(bp);
2805 	if (rc)
2806 		return rc;
2807 
2808 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2809 		return -EPERM;
2810 
2811 	if (rte_is_zero_ether_addr(addr))
2812 		return -EINVAL;
2813 
2814 	/* Filter settings will get applied when port is started */
2815 	if (!dev->data->dev_started)
2816 		return 0;
2817 
2818 	/* Check if the requested MAC is already added */
2819 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2820 		return 0;
2821 
2822 	/* Destroy filter and re-create it */
2823 	bnxt_del_dflt_mac_filter(bp, vnic);
2824 
2825 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2826 	if (dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2827 		/* This filter will allow only untagged packets */
2828 		rc = bnxt_add_vlan_filter(bp, 0);
2829 	} else {
2830 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2831 	}
2832 
2833 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2834 	return rc;
2835 }
2836 
2837 static int
2838 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2839 			  struct rte_ether_addr *mc_addr_set,
2840 			  uint32_t nb_mc_addr)
2841 {
2842 	struct bnxt *bp = eth_dev->data->dev_private;
2843 	struct bnxt_vnic_info *vnic;
2844 	uint32_t i = 0;
2845 	int rc;
2846 
2847 	rc = is_bnxt_in_error(bp);
2848 	if (rc)
2849 		return rc;
2850 
2851 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2852 
2853 	bp->nb_mc_addr = nb_mc_addr;
2854 
2855 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2856 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2857 		goto allmulti;
2858 	}
2859 
2860 	/* TODO Check for Duplicate mcast addresses */
2861 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2862 	for (i = 0; i < nb_mc_addr; i++)
2863 		rte_ether_addr_copy(&mc_addr_set[i], &bp->mcast_addr_list[i]);
2864 
2865 	if (bp->nb_mc_addr)
2866 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2867 	else
2868 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2869 
2870 allmulti:
2871 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2872 }
2873 
2874 static int
2875 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2876 {
2877 	struct bnxt *bp = dev->data->dev_private;
2878 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2879 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2880 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2881 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2882 	int ret;
2883 
2884 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2885 			fw_major, fw_minor, fw_updt, fw_rsvd);
2886 	if (ret < 0)
2887 		return -EINVAL;
2888 
2889 	ret += 1; /* add the size of '\0' */
2890 	if (fw_size < (size_t)ret)
2891 		return ret;
2892 	else
2893 		return 0;
2894 }
2895 
2896 static void
2897 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2898 	struct rte_eth_rxq_info *qinfo)
2899 {
2900 	struct bnxt *bp = dev->data->dev_private;
2901 	struct bnxt_rx_queue *rxq;
2902 
2903 	if (is_bnxt_in_error(bp))
2904 		return;
2905 
2906 	rxq = dev->data->rx_queues[queue_id];
2907 
2908 	qinfo->mp = rxq->mb_pool;
2909 	qinfo->scattered_rx = dev->data->scattered_rx;
2910 	qinfo->nb_desc = rxq->nb_rx_desc;
2911 
2912 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2913 	qinfo->conf.rx_drop_en = rxq->drop_en;
2914 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2915 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2916 }
2917 
2918 static void
2919 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2920 	struct rte_eth_txq_info *qinfo)
2921 {
2922 	struct bnxt *bp = dev->data->dev_private;
2923 	struct bnxt_tx_queue *txq;
2924 
2925 	if (is_bnxt_in_error(bp))
2926 		return;
2927 
2928 	txq = dev->data->tx_queues[queue_id];
2929 
2930 	qinfo->nb_desc = txq->nb_tx_desc;
2931 
2932 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2933 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2934 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2935 
2936 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2937 	qinfo->conf.tx_rs_thresh = 0;
2938 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2939 	qinfo->conf.offloads = txq->offloads;
2940 }
2941 
2942 static const struct {
2943 	eth_rx_burst_t pkt_burst;
2944 	const char *info;
2945 } bnxt_rx_burst_info[] = {
2946 	{bnxt_recv_pkts,		"Scalar"},
2947 #if defined(RTE_ARCH_X86)
2948 	{bnxt_recv_pkts_vec,		"Vector SSE"},
2949 #endif
2950 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2951 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
2952 #endif
2953 #if defined(RTE_ARCH_ARM64)
2954 	{bnxt_recv_pkts_vec,		"Vector Neon"},
2955 #endif
2956 };
2957 
2958 static int
2959 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2960 		       struct rte_eth_burst_mode *mode)
2961 {
2962 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2963 	size_t i;
2964 
2965 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2966 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2967 			snprintf(mode->info, sizeof(mode->info), "%s",
2968 				 bnxt_rx_burst_info[i].info);
2969 			return 0;
2970 		}
2971 	}
2972 
2973 	return -EINVAL;
2974 }
2975 
2976 static const struct {
2977 	eth_tx_burst_t pkt_burst;
2978 	const char *info;
2979 } bnxt_tx_burst_info[] = {
2980 	{bnxt_xmit_pkts,		"Scalar"},
2981 #if defined(RTE_ARCH_X86)
2982 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
2983 #endif
2984 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2985 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
2986 #endif
2987 #if defined(RTE_ARCH_ARM64)
2988 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
2989 #endif
2990 };
2991 
2992 static int
2993 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2994 		       struct rte_eth_burst_mode *mode)
2995 {
2996 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2997 	size_t i;
2998 
2999 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
3000 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
3001 			snprintf(mode->info, sizeof(mode->info), "%s",
3002 				 bnxt_tx_burst_info[i].info);
3003 			return 0;
3004 		}
3005 	}
3006 
3007 	return -EINVAL;
3008 }
3009 
3010 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3011 {
3012 	uint32_t overhead = BNXT_MAX_PKT_LEN - BNXT_MAX_MTU;
3013 	struct bnxt *bp = eth_dev->data->dev_private;
3014 	uint32_t new_pkt_size;
3015 	uint32_t rc;
3016 	uint32_t i;
3017 
3018 	rc = is_bnxt_in_error(bp);
3019 	if (rc)
3020 		return rc;
3021 
3022 	/* Exit if receive queues are not configured yet */
3023 	if (!eth_dev->data->nb_rx_queues)
3024 		return rc;
3025 
3026 	new_pkt_size = new_mtu + overhead;
3027 
3028 	/*
3029 	 * Disallow any MTU change that would require scattered receive support
3030 	 * if it is not already enabled.
3031 	 */
3032 	if (eth_dev->data->dev_started &&
3033 	    !eth_dev->data->scattered_rx &&
3034 	    (new_pkt_size >
3035 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
3036 		PMD_DRV_LOG(ERR,
3037 			    "MTU change would require scattered rx support. ");
3038 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
3039 		return -EINVAL;
3040 	}
3041 
3042 	if (new_mtu > RTE_ETHER_MTU)
3043 		bp->flags |= BNXT_FLAG_JUMBO;
3044 	else
3045 		bp->flags &= ~BNXT_FLAG_JUMBO;
3046 
3047 	/* Is there a change in mtu setting? */
3048 	if (eth_dev->data->mtu == new_mtu)
3049 		return rc;
3050 
3051 	for (i = 0; i < bp->nr_vnics; i++) {
3052 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3053 		uint16_t size = 0;
3054 
3055 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
3056 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
3057 		if (rc)
3058 			break;
3059 
3060 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
3061 		size -= RTE_PKTMBUF_HEADROOM;
3062 
3063 		if (size < new_mtu) {
3064 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
3065 			if (rc)
3066 				return rc;
3067 		}
3068 	}
3069 
3070 	if (bnxt_hwrm_config_host_mtu(bp))
3071 		PMD_DRV_LOG(WARNING, "Failed to configure host MTU\n");
3072 
3073 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3074 
3075 	return rc;
3076 }
3077 
3078 static int
3079 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3080 {
3081 	struct bnxt *bp = dev->data->dev_private;
3082 	uint16_t vlan = bp->vlan;
3083 	int rc;
3084 
3085 	rc = is_bnxt_in_error(bp);
3086 	if (rc)
3087 		return rc;
3088 
3089 	if (!BNXT_SINGLE_PF(bp)) {
3090 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3091 		return -ENOTSUP;
3092 	}
3093 	bp->vlan = on ? pvid : 0;
3094 
3095 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3096 	if (rc)
3097 		bp->vlan = vlan;
3098 	return rc;
3099 }
3100 
3101 static int
3102 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3103 {
3104 	struct bnxt *bp = dev->data->dev_private;
3105 	int rc;
3106 
3107 	rc = is_bnxt_in_error(bp);
3108 	if (rc)
3109 		return rc;
3110 
3111 	return bnxt_hwrm_port_led_cfg(bp, true);
3112 }
3113 
3114 static int
3115 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3116 {
3117 	struct bnxt *bp = dev->data->dev_private;
3118 	int rc;
3119 
3120 	rc = is_bnxt_in_error(bp);
3121 	if (rc)
3122 		return rc;
3123 
3124 	return bnxt_hwrm_port_led_cfg(bp, false);
3125 }
3126 
3127 static uint32_t
3128 bnxt_rx_queue_count_op(void *rx_queue)
3129 {
3130 	struct bnxt *bp;
3131 	struct bnxt_cp_ring_info *cpr;
3132 	uint32_t desc = 0, raw_cons, cp_ring_size;
3133 	struct bnxt_rx_queue *rxq;
3134 	struct rx_pkt_cmpl *rxcmp;
3135 	int rc;
3136 
3137 	rxq = rx_queue;
3138 	bp = rxq->bp;
3139 
3140 	rc = is_bnxt_in_error(bp);
3141 	if (rc)
3142 		return rc;
3143 
3144 	cpr = rxq->cp_ring;
3145 	raw_cons = cpr->cp_raw_cons;
3146 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3147 
3148 	while (1) {
3149 		uint32_t agg_cnt, cons, cmpl_type;
3150 
3151 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3152 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3153 
3154 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3155 			break;
3156 
3157 		cmpl_type = CMP_TYPE(rxcmp);
3158 
3159 		switch (cmpl_type) {
3160 		case CMPL_BASE_TYPE_RX_L2:
3161 		case CMPL_BASE_TYPE_RX_L2_V2:
3162 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3163 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3164 			desc++;
3165 			break;
3166 
3167 		case CMPL_BASE_TYPE_RX_TPA_END:
3168 			if (BNXT_CHIP_P5(rxq->bp)) {
3169 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3170 
3171 				p5_tpa_end = (void *)rxcmp;
3172 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3173 			} else {
3174 				struct rx_tpa_end_cmpl *tpa_end;
3175 
3176 				tpa_end = (void *)rxcmp;
3177 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3178 			}
3179 
3180 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3181 			desc++;
3182 			break;
3183 
3184 		default:
3185 			raw_cons += CMP_LEN(cmpl_type);
3186 		}
3187 	}
3188 
3189 	return desc;
3190 }
3191 
3192 static int
3193 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3194 {
3195 	struct bnxt_rx_queue *rxq = rx_queue;
3196 	struct bnxt_cp_ring_info *cpr;
3197 	struct bnxt_rx_ring_info *rxr;
3198 	uint32_t desc, raw_cons, cp_ring_size;
3199 	struct bnxt *bp = rxq->bp;
3200 	struct rx_pkt_cmpl *rxcmp;
3201 	int rc;
3202 
3203 	rc = is_bnxt_in_error(bp);
3204 	if (rc)
3205 		return rc;
3206 
3207 	if (offset >= rxq->nb_rx_desc)
3208 		return -EINVAL;
3209 
3210 	rxr = rxq->rx_ring;
3211 	cpr = rxq->cp_ring;
3212 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3213 
3214 	/*
3215 	 * For the vector receive case, the completion at the requested
3216 	 * offset can be indexed directly.
3217 	 */
3218 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3219 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3220 		struct rx_pkt_cmpl *rxcmp;
3221 		uint32_t cons;
3222 
3223 		/* Check status of completion descriptor. */
3224 		raw_cons = cpr->cp_raw_cons +
3225 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3226 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3227 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3228 
3229 		if (bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3230 			return RTE_ETH_RX_DESC_DONE;
3231 
3232 		/* Check whether rx desc has an mbuf attached. */
3233 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3234 		if (cons >= rxq->rxrearm_start &&
3235 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3236 			return RTE_ETH_RX_DESC_UNAVAIL;
3237 		}
3238 
3239 		return RTE_ETH_RX_DESC_AVAIL;
3240 	}
3241 #endif
3242 
3243 	/*
3244 	 * For the non-vector receive case, scan the completion ring to
3245 	 * locate the completion descriptor for the requested offset.
3246 	 */
3247 	raw_cons = cpr->cp_raw_cons;
3248 	desc = 0;
3249 	while (1) {
3250 		uint32_t agg_cnt, cons, cmpl_type;
3251 
3252 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3253 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3254 
3255 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3256 			break;
3257 
3258 		cmpl_type = CMP_TYPE(rxcmp);
3259 
3260 		switch (cmpl_type) {
3261 		case CMPL_BASE_TYPE_RX_L2:
3262 		case CMPL_BASE_TYPE_RX_L2_V2:
3263 			if (desc == offset) {
3264 				cons = rxcmp->opaque;
3265 				if (rxr->rx_buf_ring[cons])
3266 					return RTE_ETH_RX_DESC_DONE;
3267 				else
3268 					return RTE_ETH_RX_DESC_UNAVAIL;
3269 			}
3270 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3271 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3272 			desc++;
3273 			break;
3274 
3275 		case CMPL_BASE_TYPE_RX_TPA_END:
3276 			if (desc == offset)
3277 				return RTE_ETH_RX_DESC_DONE;
3278 
3279 			if (BNXT_CHIP_P5(rxq->bp)) {
3280 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3281 
3282 				p5_tpa_end = (void *)rxcmp;
3283 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3284 			} else {
3285 				struct rx_tpa_end_cmpl *tpa_end;
3286 
3287 				tpa_end = (void *)rxcmp;
3288 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3289 			}
3290 
3291 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3292 			desc++;
3293 			break;
3294 
3295 		default:
3296 			raw_cons += CMP_LEN(cmpl_type);
3297 		}
3298 	}
3299 
3300 	return RTE_ETH_RX_DESC_AVAIL;
3301 }
3302 
3303 static int
3304 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3305 {
3306 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3307 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3308 	uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3309 	struct cmpl_base *cp_desc_ring;
3310 	int rc;
3311 
3312 	rc = is_bnxt_in_error(txq->bp);
3313 	if (rc)
3314 		return rc;
3315 
3316 	if (offset >= txq->nb_tx_desc)
3317 		return -EINVAL;
3318 
3319 	/* Return "desc done" if descriptor is available for use. */
3320 	if (bnxt_tx_bds_in_hw(txq) <= offset)
3321 		return RTE_ETH_TX_DESC_DONE;
3322 
3323 	raw_cons = cpr->cp_raw_cons;
3324 	cp_desc_ring = cpr->cp_desc_ring;
3325 	ring_mask = cpr->cp_ring_struct->ring_mask;
3326 
3327 	/* Check to see if hw has posted a completion for the descriptor. */
3328 	while (1) {
3329 		struct tx_cmpl *txcmp;
3330 		uint32_t cons;
3331 
3332 		cons = RING_CMPL(ring_mask, raw_cons);
3333 		txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3334 
3335 		if (!bnxt_cpr_cmp_valid(txcmp, raw_cons, ring_mask + 1))
3336 			break;
3337 
3338 		if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3339 			nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3340 
3341 		if (nb_tx_pkts > offset)
3342 			return RTE_ETH_TX_DESC_DONE;
3343 
3344 		raw_cons = NEXT_RAW_CMP(raw_cons);
3345 	}
3346 
3347 	/* Descriptor is pending transmit, not yet completed by hardware. */
3348 	return RTE_ETH_TX_DESC_FULL;
3349 }
3350 
3351 int
3352 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3353 		     const struct rte_flow_ops **ops)
3354 {
3355 	struct bnxt *bp = dev->data->dev_private;
3356 	int ret = 0;
3357 
3358 	if (!bp)
3359 		return -EIO;
3360 
3361 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3362 		struct bnxt_representor *vfr = dev->data->dev_private;
3363 		bp = vfr->parent_dev->data->dev_private;
3364 		/* parent is deleted while children are still valid */
3365 		if (!bp) {
3366 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3367 				    dev->data->port_id);
3368 			return -EIO;
3369 		}
3370 	}
3371 
3372 	ret = is_bnxt_in_error(bp);
3373 	if (ret)
3374 		return ret;
3375 
3376 	/* PMD supports thread-safe flow operations.  rte_flow API
3377 	 * functions can avoid mutex for multi-thread safety.
3378 	 */
3379 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3380 
3381 	if (BNXT_TRUFLOW_EN(bp))
3382 		*ops = &bnxt_ulp_rte_flow_ops;
3383 	else
3384 		*ops = &bnxt_flow_ops;
3385 
3386 	return ret;
3387 }
3388 
3389 static const uint32_t *
3390 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3391 {
3392 	static const uint32_t ptypes[] = {
3393 		RTE_PTYPE_L2_ETHER_VLAN,
3394 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3395 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3396 		RTE_PTYPE_L4_ICMP,
3397 		RTE_PTYPE_L4_TCP,
3398 		RTE_PTYPE_L4_UDP,
3399 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3400 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3401 		RTE_PTYPE_INNER_L4_ICMP,
3402 		RTE_PTYPE_INNER_L4_TCP,
3403 		RTE_PTYPE_INNER_L4_UDP,
3404 		RTE_PTYPE_UNKNOWN
3405 	};
3406 
3407 	if (!dev->rx_pkt_burst)
3408 		return NULL;
3409 
3410 	return ptypes;
3411 }
3412 
3413 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3414 			 int reg_win)
3415 {
3416 	uint32_t reg_base = *reg_arr & 0xfffff000;
3417 	uint32_t win_off;
3418 	int i;
3419 
3420 	for (i = 0; i < count; i++) {
3421 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3422 			return -ERANGE;
3423 	}
3424 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3425 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3426 	return 0;
3427 }
3428 
3429 static int bnxt_map_ptp_regs(struct bnxt *bp)
3430 {
3431 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3432 	uint32_t *reg_arr;
3433 	int rc, i;
3434 
3435 	reg_arr = ptp->rx_regs;
3436 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3437 	if (rc)
3438 		return rc;
3439 
3440 	reg_arr = ptp->tx_regs;
3441 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3442 	if (rc)
3443 		return rc;
3444 
3445 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3446 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3447 
3448 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3449 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3450 
3451 	return 0;
3452 }
3453 
3454 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3455 {
3456 	rte_write32(0, (uint8_t *)bp->bar0 +
3457 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3458 	rte_write32(0, (uint8_t *)bp->bar0 +
3459 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3460 }
3461 
3462 static uint64_t bnxt_cc_read(struct bnxt *bp)
3463 {
3464 	uint64_t ns;
3465 
3466 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3467 			      BNXT_GRCPF_REG_SYNC_TIME));
3468 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3469 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3470 	return ns;
3471 }
3472 
3473 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3474 {
3475 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3476 	uint32_t fifo;
3477 
3478 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3479 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3480 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3481 		return -EAGAIN;
3482 
3483 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3484 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3485 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3486 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3487 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3488 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3489 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3490 
3491 	return 0;
3492 }
3493 
3494 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3495 {
3496 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3497 	struct bnxt_pf_info *pf = bp->pf;
3498 	uint16_t port_id;
3499 	int i = 0;
3500 	uint32_t fifo;
3501 
3502 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3503 		return -EINVAL;
3504 
3505 	port_id = pf->port_id;
3506 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3507 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3508 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3509 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3510 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3511 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3512 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3513 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3514 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3515 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3516 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3517 		i++;
3518 	}
3519 
3520 	if (i >= BNXT_PTP_RX_PND_CNT)
3521 		return -EBUSY;
3522 
3523 	return 0;
3524 }
3525 
3526 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3527 {
3528 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3529 	struct bnxt_pf_info *pf = bp->pf;
3530 	uint16_t port_id;
3531 	uint32_t fifo;
3532 
3533 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3534 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3535 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3536 		return -EAGAIN;
3537 
3538 	port_id = pf->port_id;
3539 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3540 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3541 
3542 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3543 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3544 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3545 		return bnxt_clr_rx_ts(bp, ts);
3546 
3547 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3548 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3549 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3550 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3551 
3552 	return 0;
3553 }
3554 
3555 static int
3556 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3557 {
3558 	uint64_t ns;
3559 	struct bnxt *bp = dev->data->dev_private;
3560 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3561 
3562 	if (!ptp)
3563 		return -ENOTSUP;
3564 
3565 	ns = rte_timespec_to_ns(ts);
3566 	/* Set the timecounters to a new value. */
3567 	ptp->tc.nsec = ns;
3568 	ptp->tx_tstamp_tc.nsec = ns;
3569 	ptp->rx_tstamp_tc.nsec = ns;
3570 
3571 	return 0;
3572 }
3573 
3574 static int
3575 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3576 {
3577 	struct bnxt *bp = dev->data->dev_private;
3578 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3579 	uint64_t ns, systime_cycles = 0;
3580 	int rc = 0;
3581 
3582 	if (!ptp)
3583 		return -ENOTSUP;
3584 
3585 	if (BNXT_CHIP_P5(bp))
3586 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3587 					     &systime_cycles);
3588 	else
3589 		systime_cycles = bnxt_cc_read(bp);
3590 
3591 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3592 	*ts = rte_ns_to_timespec(ns);
3593 
3594 	return rc;
3595 }
3596 static int
3597 bnxt_timesync_enable(struct rte_eth_dev *dev)
3598 {
3599 	struct bnxt *bp = dev->data->dev_private;
3600 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3601 	uint32_t shift = 0;
3602 	int rc;
3603 
3604 	if (!ptp)
3605 		return -ENOTSUP;
3606 
3607 	ptp->rx_filter = 1;
3608 	ptp->tx_tstamp_en = 1;
3609 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3610 
3611 	rc = bnxt_hwrm_ptp_cfg(bp);
3612 	if (rc)
3613 		return rc;
3614 
3615 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3616 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3617 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3618 
3619 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3620 	ptp->tc.cc_shift = shift;
3621 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3622 
3623 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3624 	ptp->rx_tstamp_tc.cc_shift = shift;
3625 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3626 
3627 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3628 	ptp->tx_tstamp_tc.cc_shift = shift;
3629 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3630 
3631 	if (!BNXT_CHIP_P5(bp))
3632 		bnxt_map_ptp_regs(bp);
3633 	else
3634 		rc = bnxt_ptp_start(bp);
3635 
3636 	return rc;
3637 }
3638 
3639 static int
3640 bnxt_timesync_disable(struct rte_eth_dev *dev)
3641 {
3642 	struct bnxt *bp = dev->data->dev_private;
3643 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3644 
3645 	if (!ptp)
3646 		return -ENOTSUP;
3647 
3648 	ptp->rx_filter = 0;
3649 	ptp->tx_tstamp_en = 0;
3650 	ptp->rxctl = 0;
3651 
3652 	bnxt_hwrm_ptp_cfg(bp);
3653 
3654 	if (!BNXT_CHIP_P5(bp))
3655 		bnxt_unmap_ptp_regs(bp);
3656 	else
3657 		bnxt_ptp_stop(bp);
3658 
3659 	return 0;
3660 }
3661 
3662 static int
3663 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3664 				 struct timespec *timestamp,
3665 				 uint32_t flags __rte_unused)
3666 {
3667 	struct bnxt *bp = dev->data->dev_private;
3668 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3669 	uint64_t rx_tstamp_cycles = 0;
3670 	uint64_t ns;
3671 
3672 	if (!ptp)
3673 		return -ENOTSUP;
3674 
3675 	if (BNXT_CHIP_P5(bp))
3676 		rx_tstamp_cycles = ptp->rx_timestamp;
3677 	else
3678 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3679 
3680 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3681 	*timestamp = rte_ns_to_timespec(ns);
3682 	return  0;
3683 }
3684 
3685 static int
3686 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3687 				 struct timespec *timestamp)
3688 {
3689 	struct bnxt *bp = dev->data->dev_private;
3690 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3691 	uint64_t tx_tstamp_cycles = 0;
3692 	uint64_t ns;
3693 	int rc = 0;
3694 
3695 	if (!ptp)
3696 		return -ENOTSUP;
3697 
3698 	if (BNXT_CHIP_P5(bp))
3699 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3700 					     &tx_tstamp_cycles);
3701 	else
3702 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3703 
3704 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3705 	*timestamp = rte_ns_to_timespec(ns);
3706 
3707 	return rc;
3708 }
3709 
3710 static int
3711 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3712 {
3713 	struct bnxt *bp = dev->data->dev_private;
3714 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3715 
3716 	if (!ptp)
3717 		return -ENOTSUP;
3718 
3719 	ptp->tc.nsec += delta;
3720 	ptp->tx_tstamp_tc.nsec += delta;
3721 	ptp->rx_tstamp_tc.nsec += delta;
3722 
3723 	return 0;
3724 }
3725 
3726 static int
3727 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3728 {
3729 	struct bnxt *bp = dev->data->dev_private;
3730 	int rc;
3731 	uint32_t dir_entries;
3732 	uint32_t entry_length;
3733 
3734 	rc = is_bnxt_in_error(bp);
3735 	if (rc)
3736 		return rc;
3737 
3738 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3739 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3740 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3741 
3742 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3743 	if (rc != 0)
3744 		return rc;
3745 
3746 	return dir_entries * entry_length;
3747 }
3748 
3749 static int
3750 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3751 		struct rte_dev_eeprom_info *in_eeprom)
3752 {
3753 	struct bnxt *bp = dev->data->dev_private;
3754 	uint32_t index;
3755 	uint32_t offset;
3756 	int rc;
3757 
3758 	rc = is_bnxt_in_error(bp);
3759 	if (rc)
3760 		return rc;
3761 
3762 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3763 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3764 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3765 		    in_eeprom->offset, in_eeprom->length);
3766 
3767 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3768 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3769 						in_eeprom->data);
3770 
3771 	index = in_eeprom->offset >> 24;
3772 	offset = in_eeprom->offset & 0xffffff;
3773 
3774 	if (index != 0)
3775 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3776 					   in_eeprom->length, in_eeprom->data);
3777 
3778 	return 0;
3779 }
3780 
3781 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3782 {
3783 	switch (dir_type) {
3784 	case BNX_DIR_TYPE_CHIMP_PATCH:
3785 	case BNX_DIR_TYPE_BOOTCODE:
3786 	case BNX_DIR_TYPE_BOOTCODE_2:
3787 	case BNX_DIR_TYPE_APE_FW:
3788 	case BNX_DIR_TYPE_APE_PATCH:
3789 	case BNX_DIR_TYPE_KONG_FW:
3790 	case BNX_DIR_TYPE_KONG_PATCH:
3791 	case BNX_DIR_TYPE_BONO_FW:
3792 	case BNX_DIR_TYPE_BONO_PATCH:
3793 		/* FALLTHROUGH */
3794 		return true;
3795 	}
3796 
3797 	return false;
3798 }
3799 
3800 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3801 {
3802 	switch (dir_type) {
3803 	case BNX_DIR_TYPE_AVS:
3804 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3805 	case BNX_DIR_TYPE_PCIE:
3806 	case BNX_DIR_TYPE_TSCF_UCODE:
3807 	case BNX_DIR_TYPE_EXT_PHY:
3808 	case BNX_DIR_TYPE_CCM:
3809 	case BNX_DIR_TYPE_ISCSI_BOOT:
3810 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3811 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3812 		/* FALLTHROUGH */
3813 		return true;
3814 	}
3815 
3816 	return false;
3817 }
3818 
3819 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3820 {
3821 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3822 		bnxt_dir_type_is_other_exec_format(dir_type);
3823 }
3824 
3825 static int
3826 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3827 		struct rte_dev_eeprom_info *in_eeprom)
3828 {
3829 	struct bnxt *bp = dev->data->dev_private;
3830 	uint8_t index, dir_op;
3831 	uint16_t type, ext, ordinal, attr;
3832 	int rc;
3833 
3834 	rc = is_bnxt_in_error(bp);
3835 	if (rc)
3836 		return rc;
3837 
3838 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3839 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3840 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3841 		    in_eeprom->offset, in_eeprom->length);
3842 
3843 	if (!BNXT_PF(bp)) {
3844 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3845 		return -EINVAL;
3846 	}
3847 
3848 	type = in_eeprom->magic >> 16;
3849 
3850 	if (type == 0xffff) { /* special value for directory operations */
3851 		index = in_eeprom->magic & 0xff;
3852 		dir_op = in_eeprom->magic >> 8;
3853 		if (index == 0)
3854 			return -EINVAL;
3855 		switch (dir_op) {
3856 		case 0x0e: /* erase */
3857 			if (in_eeprom->offset != ~in_eeprom->magic)
3858 				return -EINVAL;
3859 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3860 		default:
3861 			return -EINVAL;
3862 		}
3863 	}
3864 
3865 	/* Create or re-write an NVM item: */
3866 	if (bnxt_dir_type_is_executable(type) == true)
3867 		return -EOPNOTSUPP;
3868 	ext = in_eeprom->magic & 0xffff;
3869 	ordinal = in_eeprom->offset >> 16;
3870 	attr = in_eeprom->offset & 0xffff;
3871 
3872 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3873 				     in_eeprom->data, in_eeprom->length);
3874 }
3875 
3876 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3877 				struct rte_eth_dev_module_info *modinfo)
3878 {
3879 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3880 	struct bnxt *bp = dev->data->dev_private;
3881 	int rc;
3882 
3883 	/* No point in going further if phy status indicates
3884 	 * module is not inserted or if it is powered down or
3885 	 * if it is of type 10GBase-T
3886 	 */
3887 	if (bp->link_info->module_status >
3888 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3889 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3890 			    dev->data->port_id);
3891 		return -ENOTSUP;
3892 	}
3893 
3894 	/* This feature is not supported in older firmware versions */
3895 	if (bp->hwrm_spec_code < 0x10202) {
3896 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3897 			    dev->data->port_id);
3898 		return -ENOTSUP;
3899 	}
3900 
3901 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3902 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3903 						   module_info);
3904 
3905 	if (rc)
3906 		return rc;
3907 
3908 	switch (module_info[0]) {
3909 	case SFF_MODULE_ID_SFP:
3910 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3911 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3912 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3913 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3914 		break;
3915 	case SFF_MODULE_ID_QSFP:
3916 	case SFF_MODULE_ID_QSFP_PLUS:
3917 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3918 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3919 		break;
3920 	case SFF_MODULE_ID_QSFP28:
3921 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3922 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3923 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3924 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3925 		break;
3926 	default:
3927 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3928 		return -ENOTSUP;
3929 	}
3930 
3931 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3932 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3933 
3934 	return 0;
3935 }
3936 
3937 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3938 				  struct rte_dev_eeprom_info *info)
3939 {
3940 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3941 	uint32_t offset = info->offset, length = info->length;
3942 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3943 	struct bnxt *bp = dev->data->dev_private;
3944 	uint8_t *data = info->data;
3945 	uint8_t page = offset >> 7;
3946 	uint8_t max_pages = 2;
3947 	uint8_t opt_pages;
3948 	int rc;
3949 
3950 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3951 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3952 						   module_info);
3953 	if (rc)
3954 		return rc;
3955 
3956 	switch (module_info[0]) {
3957 	case SFF_MODULE_ID_SFP:
3958 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
3959 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
3960 			pg_addr[2] = I2C_DEV_ADDR_A2;
3961 			pg_addr[3] = I2C_DEV_ADDR_A2;
3962 			max_pages = 4;
3963 		}
3964 		break;
3965 	case SFF_MODULE_ID_QSFP28:
3966 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
3967 							   SFF8636_OPT_PAGES_OFFSET,
3968 							   1, &opt_pages);
3969 		if (rc)
3970 			return rc;
3971 
3972 		if (opt_pages & SFF8636_PAGE1_MASK) {
3973 			pg_addr[2] = I2C_DEV_ADDR_A0;
3974 			max_pages = 3;
3975 		}
3976 		if (opt_pages & SFF8636_PAGE2_MASK) {
3977 			pg_addr[3] = I2C_DEV_ADDR_A0;
3978 			max_pages = 4;
3979 		}
3980 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
3981 			pg_addr[4] = I2C_DEV_ADDR_A0;
3982 			max_pages = 5;
3983 		}
3984 		break;
3985 	default:
3986 		break;
3987 	}
3988 
3989 	memset(data, 0, length);
3990 
3991 	offset &= 0xff;
3992 	while (length && page < max_pages) {
3993 		uint8_t raw_page = page ? page - 1 : 0;
3994 		uint16_t chunk;
3995 
3996 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
3997 			raw_page = 0;
3998 		else if (page)
3999 			offset |= 0x80;
4000 		chunk = RTE_MIN(length, 256 - offset);
4001 
4002 		if (pg_addr[page]) {
4003 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
4004 								   raw_page, offset,
4005 								   chunk, data);
4006 			if (rc)
4007 				return rc;
4008 		}
4009 
4010 		data += chunk;
4011 		length -= chunk;
4012 		offset = 0;
4013 		page += 1 + (chunk > 128);
4014 	}
4015 
4016 	return length ? -EINVAL : 0;
4017 }
4018 
4019 /*
4020  * Initialization
4021  */
4022 
4023 static const struct eth_dev_ops bnxt_dev_ops = {
4024 	.dev_infos_get = bnxt_dev_info_get_op,
4025 	.dev_close = bnxt_dev_close_op,
4026 	.dev_configure = bnxt_dev_configure_op,
4027 	.dev_start = bnxt_dev_start_op,
4028 	.dev_stop = bnxt_dev_stop_op,
4029 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4030 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4031 	.stats_get = bnxt_stats_get_op,
4032 	.stats_reset = bnxt_stats_reset_op,
4033 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4034 	.rx_queue_release = bnxt_rx_queue_release_op,
4035 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4036 	.tx_queue_release = bnxt_tx_queue_release_op,
4037 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4038 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4039 	.reta_update = bnxt_reta_update_op,
4040 	.reta_query = bnxt_reta_query_op,
4041 	.rss_hash_update = bnxt_rss_hash_update_op,
4042 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4043 	.link_update = bnxt_link_update_op,
4044 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4045 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4046 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4047 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4048 	.mac_addr_add = bnxt_mac_addr_add_op,
4049 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4050 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4051 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4052 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4053 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4054 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4055 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4056 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4057 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4058 	.mtu_set = bnxt_mtu_set_op,
4059 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4060 	.xstats_get = bnxt_dev_xstats_get_op,
4061 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4062 	.xstats_reset = bnxt_dev_xstats_reset_op,
4063 	.fw_version_get = bnxt_fw_version_get,
4064 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4065 	.rxq_info_get = bnxt_rxq_info_get_op,
4066 	.txq_info_get = bnxt_txq_info_get_op,
4067 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4068 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4069 	.dev_led_on = bnxt_dev_led_on_op,
4070 	.dev_led_off = bnxt_dev_led_off_op,
4071 	.rx_queue_start = bnxt_rx_queue_start,
4072 	.rx_queue_stop = bnxt_rx_queue_stop,
4073 	.tx_queue_start = bnxt_tx_queue_start,
4074 	.tx_queue_stop = bnxt_tx_queue_stop,
4075 	.flow_ops_get = bnxt_flow_ops_get_op,
4076 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4077 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4078 	.get_eeprom           = bnxt_get_eeprom_op,
4079 	.set_eeprom           = bnxt_set_eeprom_op,
4080 	.get_module_info = bnxt_get_module_info,
4081 	.get_module_eeprom = bnxt_get_module_eeprom,
4082 	.timesync_enable      = bnxt_timesync_enable,
4083 	.timesync_disable     = bnxt_timesync_disable,
4084 	.timesync_read_time   = bnxt_timesync_read_time,
4085 	.timesync_write_time   = bnxt_timesync_write_time,
4086 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4087 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4088 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4089 };
4090 
4091 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4092 {
4093 	uint32_t offset;
4094 
4095 	/* Only pre-map the reset GRC registers using window 3 */
4096 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4097 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4098 
4099 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4100 
4101 	return offset;
4102 }
4103 
4104 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4105 {
4106 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4107 	uint32_t reg_base = 0xffffffff;
4108 	int i;
4109 
4110 	/* Only pre-map the monitoring GRC registers using window 2 */
4111 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4112 		uint32_t reg = info->status_regs[i];
4113 
4114 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4115 			continue;
4116 
4117 		if (reg_base == 0xffffffff)
4118 			reg_base = reg & 0xfffff000;
4119 		if ((reg & 0xfffff000) != reg_base)
4120 			return -ERANGE;
4121 
4122 		/* Use mask 0xffc as the Lower 2 bits indicates
4123 		 * address space location
4124 		 */
4125 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4126 						(reg & 0xffc);
4127 	}
4128 
4129 	if (reg_base == 0xffffffff)
4130 		return 0;
4131 
4132 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4133 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4134 
4135 	return 0;
4136 }
4137 
4138 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4139 {
4140 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4141 	uint32_t delay = info->delay_after_reset[index];
4142 	uint32_t val = info->reset_reg_val[index];
4143 	uint32_t reg = info->reset_reg[index];
4144 	uint32_t type, offset;
4145 	int ret;
4146 
4147 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4148 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4149 
4150 	switch (type) {
4151 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4152 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4153 		if (ret < 0) {
4154 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4155 				    val, offset);
4156 			return;
4157 		}
4158 		break;
4159 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4160 		offset = bnxt_map_reset_regs(bp, offset);
4161 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4162 		break;
4163 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4164 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4165 		break;
4166 	}
4167 	/* wait on a specific interval of time until core reset is complete */
4168 	if (delay)
4169 		rte_delay_ms(delay);
4170 }
4171 
4172 static void bnxt_dev_cleanup(struct bnxt *bp)
4173 {
4174 	bp->eth_dev->data->dev_link.link_status = 0;
4175 	bp->link_info->link_up = 0;
4176 	if (bp->eth_dev->data->dev_started)
4177 		bnxt_dev_stop(bp->eth_dev);
4178 
4179 	bnxt_uninit_resources(bp, true);
4180 }
4181 
4182 static int
4183 bnxt_check_fw_reset_done(struct bnxt *bp)
4184 {
4185 	int timeout = bp->fw_reset_max_msecs;
4186 	uint16_t val = 0;
4187 	int rc;
4188 
4189 	do {
4190 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4191 		if (rc < 0) {
4192 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4193 			return rc;
4194 		}
4195 		if (val != 0xffff)
4196 			break;
4197 		rte_delay_ms(1);
4198 	} while (timeout--);
4199 
4200 	if (val == 0xffff) {
4201 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4202 		return -1;
4203 	}
4204 
4205 	return 0;
4206 }
4207 
4208 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4209 {
4210 	struct rte_eth_dev *dev = bp->eth_dev;
4211 	struct rte_vlan_filter_conf *vfc;
4212 	int vidx, vbit, rc;
4213 	uint16_t vlan_id;
4214 
4215 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4216 		vfc = &dev->data->vlan_filter_conf;
4217 		vidx = vlan_id / 64;
4218 		vbit = vlan_id % 64;
4219 
4220 		/* Each bit corresponds to a VLAN id */
4221 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4222 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4223 			if (rc)
4224 				return rc;
4225 		}
4226 	}
4227 
4228 	return 0;
4229 }
4230 
4231 static int bnxt_restore_mac_filters(struct bnxt *bp)
4232 {
4233 	struct rte_eth_dev *dev = bp->eth_dev;
4234 	struct rte_eth_dev_info dev_info;
4235 	struct rte_ether_addr *addr;
4236 	uint64_t pool_mask;
4237 	uint32_t pool = 0;
4238 	uint32_t i;
4239 	int rc;
4240 
4241 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4242 		return 0;
4243 
4244 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4245 	if (rc)
4246 		return rc;
4247 
4248 	/* replay MAC address configuration */
4249 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4250 		addr = &dev->data->mac_addrs[i];
4251 
4252 		/* skip zero address */
4253 		if (rte_is_zero_ether_addr(addr))
4254 			continue;
4255 
4256 		pool = 0;
4257 		pool_mask = dev->data->mac_pool_sel[i];
4258 
4259 		do {
4260 			if (pool_mask & 1ULL) {
4261 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4262 				if (rc)
4263 					return rc;
4264 			}
4265 			pool_mask >>= 1;
4266 			pool++;
4267 		} while (pool_mask);
4268 	}
4269 
4270 	return 0;
4271 }
4272 
4273 static int bnxt_restore_mcast_mac_filters(struct bnxt *bp)
4274 {
4275 	int ret = 0;
4276 
4277 	ret = bnxt_dev_set_mc_addr_list_op(bp->eth_dev, bp->mcast_addr_list,
4278 					   bp->nb_mc_addr);
4279 	if (ret)
4280 		PMD_DRV_LOG(ERR, "Failed to restore multicast MAC addreeses\n");
4281 
4282 	return ret;
4283 }
4284 
4285 static int bnxt_restore_filters(struct bnxt *bp)
4286 {
4287 	struct rte_eth_dev *dev = bp->eth_dev;
4288 	int ret = 0;
4289 
4290 	if (dev->data->all_multicast) {
4291 		ret = bnxt_allmulticast_enable_op(dev);
4292 		if (ret)
4293 			return ret;
4294 	}
4295 	if (dev->data->promiscuous) {
4296 		ret = bnxt_promiscuous_enable_op(dev);
4297 		if (ret)
4298 			return ret;
4299 	}
4300 
4301 	ret = bnxt_restore_mac_filters(bp);
4302 	if (ret)
4303 		return ret;
4304 
4305 	/* if vlans are already programmed, this can fail with -EEXIST */
4306 	ret = bnxt_restore_vlan_filters(bp);
4307 	if (ret && ret != -EEXIST)
4308 		return ret;
4309 
4310 	ret = bnxt_restore_mcast_mac_filters(bp);
4311 	if (ret)
4312 		return ret;
4313 
4314 	return ret;
4315 }
4316 
4317 static int bnxt_check_fw_ready(struct bnxt *bp)
4318 {
4319 	int timeout = bp->fw_reset_max_msecs;
4320 	int rc = 0;
4321 
4322 	do {
4323 		rc = bnxt_hwrm_poll_ver_get(bp);
4324 		if (rc == 0)
4325 			break;
4326 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4327 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4328 	} while (rc && timeout > 0);
4329 
4330 	if (rc)
4331 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4332 
4333 	return rc;
4334 }
4335 
4336 static void bnxt_dev_recover(void *arg)
4337 {
4338 	struct bnxt *bp = arg;
4339 	int rc = 0;
4340 
4341 	pthread_mutex_lock(&bp->err_recovery_lock);
4342 
4343 	if (!bp->fw_reset_min_msecs) {
4344 		rc = bnxt_check_fw_reset_done(bp);
4345 		if (rc)
4346 			goto err;
4347 	}
4348 
4349 	/* Clear Error flag so that device re-init should happen */
4350 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4351 	PMD_DRV_LOG(INFO, "Port: %u Starting recovery...\n",
4352 		    bp->eth_dev->data->port_id);
4353 
4354 	rc = bnxt_check_fw_ready(bp);
4355 	if (rc)
4356 		goto err;
4357 
4358 	rc = bnxt_init_resources(bp, true);
4359 	if (rc) {
4360 		PMD_DRV_LOG(ERR,
4361 			    "Failed to initialize resources after reset\n");
4362 		goto err;
4363 	}
4364 	/* clear reset flag as the device is initialized now */
4365 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4366 
4367 	rc = bnxt_dev_start_op(bp->eth_dev);
4368 	if (rc) {
4369 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4370 		goto err_start;
4371 	}
4372 
4373 	rc = bnxt_restore_filters(bp);
4374 	if (rc)
4375 		goto err_start;
4376 
4377 	rte_eth_fp_ops[bp->eth_dev->data->port_id].rx_pkt_burst =
4378 		bp->eth_dev->rx_pkt_burst;
4379 	rte_eth_fp_ops[bp->eth_dev->data->port_id].tx_pkt_burst =
4380 		bp->eth_dev->tx_pkt_burst;
4381 	rte_mb();
4382 
4383 	PMD_DRV_LOG(INFO, "Port: %u Recovered from FW reset\n",
4384 		    bp->eth_dev->data->port_id);
4385 	pthread_mutex_unlock(&bp->err_recovery_lock);
4386 
4387 	return;
4388 err_start:
4389 	bnxt_dev_stop(bp->eth_dev);
4390 err:
4391 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4392 	bnxt_uninit_resources(bp, false);
4393 	if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4394 		rte_eth_dev_callback_process(bp->eth_dev,
4395 					     RTE_ETH_EVENT_INTR_RMV,
4396 					     NULL);
4397 	pthread_mutex_unlock(&bp->err_recovery_lock);
4398 	PMD_DRV_LOG(ERR, "Port %u: Failed to recover from FW reset\n",
4399 		    bp->eth_dev->data->port_id);
4400 }
4401 
4402 void bnxt_dev_reset_and_resume(void *arg)
4403 {
4404 	struct bnxt *bp = arg;
4405 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4406 	uint16_t val = 0;
4407 	int rc;
4408 
4409 	bnxt_dev_cleanup(bp);
4410 	PMD_DRV_LOG(INFO, "Port: %u Finished bnxt_dev_cleanup\n",
4411 		    bp->eth_dev->data->port_id);
4412 
4413 	bnxt_wait_for_device_shutdown(bp);
4414 
4415 	/* During some fatal firmware error conditions, the PCI config space
4416 	 * register 0x2e which normally contains the subsystem ID will become
4417 	 * 0xffff. This register will revert back to the normal value after
4418 	 * the chip has completed core reset. If we detect this condition,
4419 	 * we can poll this config register immediately for the value to revert.
4420 	 */
4421 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4422 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4423 		if (rc < 0) {
4424 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4425 			return;
4426 		}
4427 		if (val == 0xffff) {
4428 			bp->fw_reset_min_msecs = 0;
4429 			us = 1;
4430 		}
4431 	}
4432 
4433 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4434 	if (rc)
4435 		PMD_DRV_LOG(ERR, "Port %u: Error setting recovery alarm",
4436 			    bp->eth_dev->data->port_id);
4437 }
4438 
4439 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4440 {
4441 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4442 	uint32_t reg = info->status_regs[index];
4443 	uint32_t type, offset, val = 0;
4444 	int ret = 0;
4445 
4446 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4447 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4448 
4449 	switch (type) {
4450 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4451 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4452 		if (ret < 0)
4453 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4454 				    offset);
4455 		break;
4456 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4457 		offset = info->mapped_status_regs[index];
4458 		/* FALLTHROUGH */
4459 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4460 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4461 				       offset));
4462 		break;
4463 	}
4464 
4465 	return val;
4466 }
4467 
4468 static int bnxt_fw_reset_all(struct bnxt *bp)
4469 {
4470 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4471 	uint32_t i;
4472 	int rc = 0;
4473 
4474 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4475 		/* Reset through primary function driver */
4476 		for (i = 0; i < info->reg_array_cnt; i++)
4477 			bnxt_write_fw_reset_reg(bp, i);
4478 		/* Wait for time specified by FW after triggering reset */
4479 		rte_delay_ms(info->primary_func_wait_period_after_reset);
4480 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4481 		/* Reset with the help of Kong processor */
4482 		rc = bnxt_hwrm_fw_reset(bp);
4483 		if (rc)
4484 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4485 	}
4486 
4487 	return rc;
4488 }
4489 
4490 static void bnxt_fw_reset_cb(void *arg)
4491 {
4492 	struct bnxt *bp = arg;
4493 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4494 	int rc = 0;
4495 
4496 	/* Only Primary function can do FW reset */
4497 	if (bnxt_is_primary_func(bp) &&
4498 	    bnxt_is_recovery_enabled(bp)) {
4499 		rc = bnxt_fw_reset_all(bp);
4500 		if (rc) {
4501 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4502 			return;
4503 		}
4504 	}
4505 
4506 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4507 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4508 	 * (including MASTER FUNC). After receiving this Async, all the active
4509 	 * drivers should treat this case as FW initiated recovery
4510 	 */
4511 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4512 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4513 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4514 
4515 		/* To recover from error */
4516 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4517 				  (void *)bp);
4518 	}
4519 }
4520 
4521 /* Driver should poll FW heartbeat, reset_counter with the frequency
4522  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4523  * When the driver detects heartbeat stop or change in reset_counter,
4524  * it has to trigger a reset to recover from the error condition.
4525  * A “primary function” is the function who will have the privilege to
4526  * initiate the chimp reset. The primary function will be elected by the
4527  * firmware and will be notified through async message.
4528  */
4529 static void bnxt_check_fw_health(void *arg)
4530 {
4531 	struct bnxt *bp = arg;
4532 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4533 	uint32_t val = 0, wait_msec;
4534 
4535 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4536 	    is_bnxt_in_error(bp))
4537 		return;
4538 
4539 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4540 	if (val == info->last_heart_beat)
4541 		goto reset;
4542 
4543 	info->last_heart_beat = val;
4544 
4545 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4546 	if (val != info->last_reset_counter)
4547 		goto reset;
4548 
4549 	info->last_reset_counter = val;
4550 
4551 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4552 			  bnxt_check_fw_health, (void *)bp);
4553 
4554 	return;
4555 reset:
4556 	/* Stop DMA to/from device */
4557 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4558 	bp->flags |= BNXT_FLAG_FW_RESET;
4559 
4560 	bnxt_stop_rxtx(bp->eth_dev);
4561 
4562 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4563 
4564 	if (bnxt_is_primary_func(bp))
4565 		wait_msec = info->primary_func_wait_period;
4566 	else
4567 		wait_msec = info->normal_func_wait_period;
4568 
4569 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4570 			  bnxt_fw_reset_cb, (void *)bp);
4571 }
4572 
4573 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4574 {
4575 	uint32_t polling_freq;
4576 
4577 	pthread_mutex_lock(&bp->health_check_lock);
4578 
4579 	if (!bnxt_is_recovery_enabled(bp))
4580 		goto done;
4581 
4582 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4583 		goto done;
4584 
4585 	polling_freq = bp->recovery_info->driver_polling_freq;
4586 
4587 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4588 			  bnxt_check_fw_health, (void *)bp);
4589 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4590 
4591 done:
4592 	pthread_mutex_unlock(&bp->health_check_lock);
4593 }
4594 
4595 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4596 {
4597 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4598 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4599 }
4600 
4601 static bool bnxt_vf_pciid(uint16_t device_id)
4602 {
4603 	switch (device_id) {
4604 	case BROADCOM_DEV_ID_57304_VF:
4605 	case BROADCOM_DEV_ID_57406_VF:
4606 	case BROADCOM_DEV_ID_5731X_VF:
4607 	case BROADCOM_DEV_ID_5741X_VF:
4608 	case BROADCOM_DEV_ID_57414_VF:
4609 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4610 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4611 	case BROADCOM_DEV_ID_58802_VF:
4612 	case BROADCOM_DEV_ID_57500_VF1:
4613 	case BROADCOM_DEV_ID_57500_VF2:
4614 	case BROADCOM_DEV_ID_58818_VF:
4615 		/* FALLTHROUGH */
4616 		return true;
4617 	default:
4618 		return false;
4619 	}
4620 }
4621 
4622 /* Phase 5 device */
4623 static bool bnxt_p5_device(uint16_t device_id)
4624 {
4625 	switch (device_id) {
4626 	case BROADCOM_DEV_ID_57508:
4627 	case BROADCOM_DEV_ID_57504:
4628 	case BROADCOM_DEV_ID_57502:
4629 	case BROADCOM_DEV_ID_57508_MF1:
4630 	case BROADCOM_DEV_ID_57504_MF1:
4631 	case BROADCOM_DEV_ID_57502_MF1:
4632 	case BROADCOM_DEV_ID_57508_MF2:
4633 	case BROADCOM_DEV_ID_57504_MF2:
4634 	case BROADCOM_DEV_ID_57502_MF2:
4635 	case BROADCOM_DEV_ID_57500_VF1:
4636 	case BROADCOM_DEV_ID_57500_VF2:
4637 	case BROADCOM_DEV_ID_58812:
4638 	case BROADCOM_DEV_ID_58814:
4639 	case BROADCOM_DEV_ID_58818:
4640 	case BROADCOM_DEV_ID_58818_VF:
4641 		/* FALLTHROUGH */
4642 		return true;
4643 	default:
4644 		return false;
4645 	}
4646 }
4647 
4648 bool bnxt_stratus_device(struct bnxt *bp)
4649 {
4650 	uint16_t device_id = bp->pdev->id.device_id;
4651 
4652 	switch (device_id) {
4653 	case BROADCOM_DEV_ID_STRATUS_NIC:
4654 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4655 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4656 		/* FALLTHROUGH */
4657 		return true;
4658 	default:
4659 		return false;
4660 	}
4661 }
4662 
4663 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4664 {
4665 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4666 	struct bnxt *bp = eth_dev->data->dev_private;
4667 
4668 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4669 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4670 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4671 	if (!bp->bar0 || !bp->doorbell_base) {
4672 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4673 		return -ENODEV;
4674 	}
4675 
4676 	bp->eth_dev = eth_dev;
4677 	bp->pdev = pci_dev;
4678 
4679 	return 0;
4680 }
4681 
4682 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4683 				  struct bnxt_ctx_pg_info *ctx_pg,
4684 				  uint32_t mem_size,
4685 				  const char *suffix,
4686 				  uint16_t idx)
4687 {
4688 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4689 	const struct rte_memzone *mz = NULL;
4690 	char mz_name[RTE_MEMZONE_NAMESIZE];
4691 	rte_iova_t mz_phys_addr;
4692 	uint64_t valid_bits = 0;
4693 	uint32_t sz;
4694 	int i;
4695 
4696 	if (!mem_size)
4697 		return 0;
4698 
4699 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4700 			 BNXT_PAGE_SIZE;
4701 	rmem->page_size = BNXT_PAGE_SIZE;
4702 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4703 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4704 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4705 
4706 	valid_bits = PTU_PTE_VALID;
4707 
4708 	if (rmem->nr_pages > 1) {
4709 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4710 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4711 			 suffix, idx, bp->eth_dev->data->port_id);
4712 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4713 		mz = rte_memzone_lookup(mz_name);
4714 		if (!mz) {
4715 			mz = rte_memzone_reserve_aligned(mz_name,
4716 						rmem->nr_pages * 8,
4717 						bp->eth_dev->device->numa_node,
4718 						RTE_MEMZONE_2MB |
4719 						RTE_MEMZONE_SIZE_HINT_ONLY |
4720 						RTE_MEMZONE_IOVA_CONTIG,
4721 						BNXT_PAGE_SIZE);
4722 			if (mz == NULL)
4723 				return -ENOMEM;
4724 		}
4725 
4726 		memset(mz->addr, 0, mz->len);
4727 		mz_phys_addr = mz->iova;
4728 
4729 		rmem->pg_tbl = mz->addr;
4730 		rmem->pg_tbl_map = mz_phys_addr;
4731 		rmem->pg_tbl_mz = mz;
4732 	}
4733 
4734 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4735 		 suffix, idx, bp->eth_dev->data->port_id);
4736 	mz = rte_memzone_lookup(mz_name);
4737 	if (!mz) {
4738 		mz = rte_memzone_reserve_aligned(mz_name,
4739 						 mem_size,
4740 						 bp->eth_dev->device->numa_node,
4741 						 RTE_MEMZONE_1GB |
4742 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4743 						 RTE_MEMZONE_IOVA_CONTIG,
4744 						 BNXT_PAGE_SIZE);
4745 		if (mz == NULL)
4746 			return -ENOMEM;
4747 	}
4748 
4749 	memset(mz->addr, 0, mz->len);
4750 	mz_phys_addr = mz->iova;
4751 
4752 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4753 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4754 		rmem->dma_arr[i] = mz_phys_addr + sz;
4755 
4756 		if (rmem->nr_pages > 1) {
4757 			if (i == rmem->nr_pages - 2 &&
4758 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4759 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4760 			else if (i == rmem->nr_pages - 1 &&
4761 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4762 				valid_bits |= PTU_PTE_LAST;
4763 
4764 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4765 							   valid_bits);
4766 		}
4767 	}
4768 
4769 	rmem->mz = mz;
4770 	if (rmem->vmem_size)
4771 		rmem->vmem = (void **)mz->addr;
4772 	rmem->dma_arr[0] = mz_phys_addr;
4773 	return 0;
4774 }
4775 
4776 static void bnxt_free_ctx_mem(struct bnxt *bp)
4777 {
4778 	int i;
4779 
4780 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4781 		return;
4782 
4783 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4784 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4785 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4786 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4787 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4788 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4789 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4790 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4791 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4792 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4793 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4794 
4795 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4796 		if (bp->ctx->tqm_mem[i])
4797 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4798 	}
4799 
4800 	rte_free(bp->ctx);
4801 	bp->ctx = NULL;
4802 }
4803 
4804 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4805 
4806 #define min_t(type, x, y) ({                    \
4807 	type __min1 = (x);                      \
4808 	type __min2 = (y);                      \
4809 	__min1 < __min2 ? __min1 : __min2; })
4810 
4811 #define max_t(type, x, y) ({                    \
4812 	type __max1 = (x);                      \
4813 	type __max2 = (y);                      \
4814 	__max1 > __max2 ? __max1 : __max2; })
4815 
4816 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4817 
4818 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4819 {
4820 	struct bnxt_ctx_pg_info *ctx_pg;
4821 	struct bnxt_ctx_mem_info *ctx;
4822 	uint32_t mem_size, ena, entries;
4823 	uint32_t entries_sp, min;
4824 	int i, rc;
4825 
4826 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4827 	if (rc) {
4828 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4829 		return rc;
4830 	}
4831 	ctx = bp->ctx;
4832 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4833 		return 0;
4834 
4835 	ctx_pg = &ctx->qp_mem;
4836 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4837 	if (ctx->qp_entry_size) {
4838 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4839 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4840 		if (rc)
4841 			return rc;
4842 	}
4843 
4844 	ctx_pg = &ctx->srq_mem;
4845 	ctx_pg->entries = ctx->srq_max_l2_entries;
4846 	if (ctx->srq_entry_size) {
4847 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4848 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4849 		if (rc)
4850 			return rc;
4851 	}
4852 
4853 	ctx_pg = &ctx->cq_mem;
4854 	ctx_pg->entries = ctx->cq_max_l2_entries;
4855 	if (ctx->cq_entry_size) {
4856 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4857 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4858 		if (rc)
4859 			return rc;
4860 	}
4861 
4862 	ctx_pg = &ctx->vnic_mem;
4863 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4864 		ctx->vnic_max_ring_table_entries;
4865 	if (ctx->vnic_entry_size) {
4866 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4867 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4868 		if (rc)
4869 			return rc;
4870 	}
4871 
4872 	ctx_pg = &ctx->stat_mem;
4873 	ctx_pg->entries = ctx->stat_max_entries;
4874 	if (ctx->stat_entry_size) {
4875 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4876 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4877 		if (rc)
4878 			return rc;
4879 	}
4880 
4881 	min = ctx->tqm_min_entries_per_ring;
4882 
4883 	entries_sp = ctx->qp_max_l2_entries +
4884 		     ctx->vnic_max_vnic_entries +
4885 		     2 * ctx->qp_min_qp1_entries + min;
4886 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4887 
4888 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4889 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4890 	entries = clamp_t(uint32_t, entries, min,
4891 			  ctx->tqm_max_entries_per_ring);
4892 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4893 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4894 		 * i > 8 is other ext rings.
4895 		 */
4896 		ctx_pg = ctx->tqm_mem[i];
4897 		ctx_pg->entries = i ? entries : entries_sp;
4898 		if (ctx->tqm_entry_size) {
4899 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4900 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4901 						    "tqm_mem", i);
4902 			if (rc)
4903 				return rc;
4904 		}
4905 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4906 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4907 		else
4908 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4909 	}
4910 
4911 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4912 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4913 	if (rc)
4914 		PMD_DRV_LOG(ERR,
4915 			    "Failed to configure context mem: rc = %d\n", rc);
4916 	else
4917 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4918 
4919 	return rc;
4920 }
4921 
4922 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4923 {
4924 	struct rte_pci_device *pci_dev = bp->pdev;
4925 	char mz_name[RTE_MEMZONE_NAMESIZE];
4926 	const struct rte_memzone *mz = NULL;
4927 	uint32_t total_alloc_len;
4928 	rte_iova_t mz_phys_addr;
4929 
4930 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4931 		return 0;
4932 
4933 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4934 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4935 		 pci_dev->addr.bus, pci_dev->addr.devid,
4936 		 pci_dev->addr.function, "rx_port_stats");
4937 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4938 	mz = rte_memzone_lookup(mz_name);
4939 	total_alloc_len =
4940 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4941 				       sizeof(struct rx_port_stats_ext) + 512);
4942 	if (!mz) {
4943 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4944 					 SOCKET_ID_ANY,
4945 					 RTE_MEMZONE_2MB |
4946 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4947 					 RTE_MEMZONE_IOVA_CONTIG);
4948 		if (mz == NULL)
4949 			return -ENOMEM;
4950 	}
4951 	memset(mz->addr, 0, mz->len);
4952 	mz_phys_addr = mz->iova;
4953 
4954 	bp->rx_mem_zone = (const void *)mz;
4955 	bp->hw_rx_port_stats = mz->addr;
4956 	bp->hw_rx_port_stats_map = mz_phys_addr;
4957 
4958 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4959 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4960 		 pci_dev->addr.bus, pci_dev->addr.devid,
4961 		 pci_dev->addr.function, "tx_port_stats");
4962 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4963 	mz = rte_memzone_lookup(mz_name);
4964 	total_alloc_len =
4965 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4966 				       sizeof(struct tx_port_stats_ext) + 512);
4967 	if (!mz) {
4968 		mz = rte_memzone_reserve(mz_name,
4969 					 total_alloc_len,
4970 					 SOCKET_ID_ANY,
4971 					 RTE_MEMZONE_2MB |
4972 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4973 					 RTE_MEMZONE_IOVA_CONTIG);
4974 		if (mz == NULL)
4975 			return -ENOMEM;
4976 	}
4977 	memset(mz->addr, 0, mz->len);
4978 	mz_phys_addr = mz->iova;
4979 
4980 	bp->tx_mem_zone = (const void *)mz;
4981 	bp->hw_tx_port_stats = mz->addr;
4982 	bp->hw_tx_port_stats_map = mz_phys_addr;
4983 	bp->flags |= BNXT_FLAG_PORT_STATS;
4984 
4985 	/* Display extended statistics if FW supports it */
4986 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4987 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4988 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4989 		return 0;
4990 
4991 	bp->hw_rx_port_stats_ext = (void *)
4992 		((uint8_t *)bp->hw_rx_port_stats +
4993 		 sizeof(struct rx_port_stats));
4994 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4995 		sizeof(struct rx_port_stats);
4996 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4997 
4998 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4999 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
5000 		bp->hw_tx_port_stats_ext = (void *)
5001 			((uint8_t *)bp->hw_tx_port_stats +
5002 			 sizeof(struct tx_port_stats));
5003 		bp->hw_tx_port_stats_ext_map =
5004 			bp->hw_tx_port_stats_map +
5005 			sizeof(struct tx_port_stats);
5006 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
5007 	}
5008 
5009 	return 0;
5010 }
5011 
5012 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
5013 {
5014 	struct bnxt *bp = eth_dev->data->dev_private;
5015 	size_t max_mac_addr = RTE_MIN(bp->max_l2_ctx, RTE_ETH_NUM_RECEIVE_MAC_ADDR);
5016 	int rc = 0;
5017 
5018 	if (bp->max_l2_ctx > RTE_ETH_NUM_RECEIVE_MAC_ADDR)
5019 		PMD_DRV_LOG(INFO, "Max number of MAC addrs supported is %d, but will be limited to %d\n",
5020 			    bp->max_l2_ctx, RTE_ETH_NUM_RECEIVE_MAC_ADDR);
5021 
5022 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
5023 					       RTE_ETHER_ADDR_LEN * max_mac_addr,
5024 					       0);
5025 	if (eth_dev->data->mac_addrs == NULL) {
5026 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
5027 		return -ENOMEM;
5028 	}
5029 
5030 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
5031 		if (BNXT_PF(bp))
5032 			return -EINVAL;
5033 
5034 		/* Generate a random MAC address, if none was assigned by PF */
5035 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
5036 		bnxt_eth_hw_addr_random(bp->mac_addr);
5037 		PMD_DRV_LOG(INFO,
5038 			    "Assign random MAC:" RTE_ETHER_ADDR_PRT_FMT "\n",
5039 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
5040 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
5041 
5042 		rc = bnxt_hwrm_set_mac(bp);
5043 		if (rc)
5044 			return rc;
5045 	}
5046 
5047 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5048 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5049 
5050 	/*
5051 	 *  Allocate memory to hold multicast mac addresses added.
5052 	 *  Used to restore them during reset recovery
5053 	 */
5054 	bp->mcast_addr_list = rte_zmalloc("bnxt_mcast_addr_tbl",
5055 					  sizeof(struct rte_ether_addr) *
5056 					  BNXT_MAX_MC_ADDRS, 0);
5057 	if (bp->mcast_addr_list == NULL) {
5058 		PMD_DRV_LOG(ERR, "Failed to allocate multicast addr table\n");
5059 		return -ENOMEM;
5060 	}
5061 	bp->mc_list_dma_addr = rte_malloc_virt2iova(bp->mcast_addr_list);
5062 	if (bp->mc_list_dma_addr == RTE_BAD_IOVA) {
5063 		PMD_DRV_LOG(ERR, "Fail to map mcast_addr_list to physical memory\n");
5064 		return -ENOMEM;
5065 	}
5066 
5067 	return rc;
5068 }
5069 
5070 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5071 {
5072 	int rc = 0;
5073 
5074 	/* MAC is already configured in FW */
5075 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5076 		return 0;
5077 
5078 	/* Restore the old MAC configured */
5079 	rc = bnxt_hwrm_set_mac(bp);
5080 	if (rc)
5081 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5082 
5083 	return rc;
5084 }
5085 
5086 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5087 {
5088 	if (!BNXT_PF(bp))
5089 		return;
5090 
5091 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5092 
5093 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5094 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5095 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5096 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5097 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5098 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5099 }
5100 
5101 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5102 {
5103 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5104 
5105 	if (info) {
5106 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5107 			memset(info, 0, sizeof(*info));
5108 		return;
5109 	}
5110 
5111 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5112 		return;
5113 
5114 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5115 			   sizeof(*info), 0);
5116 	if (!info)
5117 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5118 
5119 	bp->recovery_info = info;
5120 }
5121 
5122 static void bnxt_check_fw_status(struct bnxt *bp)
5123 {
5124 	uint32_t fw_status;
5125 
5126 	if (!(bp->recovery_info &&
5127 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5128 		return;
5129 
5130 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5131 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5132 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5133 			    fw_status);
5134 }
5135 
5136 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5137 {
5138 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5139 	uint32_t status_loc;
5140 	uint32_t sig_ver;
5141 
5142 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5143 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5144 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5145 				   BNXT_GRCP_WINDOW_2_BASE +
5146 				   offsetof(struct hcomm_status,
5147 					    sig_ver)));
5148 	/* If the signature is absent, then FW does not support this feature */
5149 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5150 	    HCOMM_STATUS_SIGNATURE_VAL)
5151 		return 0;
5152 
5153 	if (!info) {
5154 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5155 				   sizeof(*info), 0);
5156 		if (!info)
5157 			return -ENOMEM;
5158 		bp->recovery_info = info;
5159 	} else {
5160 		memset(info, 0, sizeof(*info));
5161 	}
5162 
5163 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5164 				      BNXT_GRCP_WINDOW_2_BASE +
5165 				      offsetof(struct hcomm_status,
5166 					       fw_status_loc)));
5167 
5168 	/* Only pre-map the FW health status GRC register */
5169 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5170 		return 0;
5171 
5172 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5173 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5174 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5175 
5176 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5177 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5178 
5179 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5180 
5181 	return 0;
5182 }
5183 
5184 /* This function gets the FW version along with the
5185  * capabilities(MAX and current) of the function, vnic,
5186  * error recovery, phy and other chip related info
5187  */
5188 static int bnxt_get_config(struct bnxt *bp)
5189 {
5190 	uint16_t mtu;
5191 	int rc = 0;
5192 
5193 	bp->fw_cap = 0;
5194 
5195 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5196 	if (rc)
5197 		return rc;
5198 
5199 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5200 	if (rc) {
5201 		bnxt_check_fw_status(bp);
5202 		return rc;
5203 	}
5204 
5205 	rc = bnxt_hwrm_func_reset(bp);
5206 	if (rc)
5207 		return -EIO;
5208 
5209 	rc = bnxt_hwrm_vnic_qcaps(bp);
5210 	if (rc)
5211 		return rc;
5212 
5213 	rc = bnxt_hwrm_queue_qportcfg(bp);
5214 	if (rc)
5215 		return rc;
5216 
5217 	/* Get the MAX capabilities for this function.
5218 	 * This function also allocates context memory for TQM rings and
5219 	 * informs the firmware about this allocated backing store memory.
5220 	 */
5221 	rc = bnxt_hwrm_func_qcaps(bp);
5222 	if (rc)
5223 		return rc;
5224 
5225 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5226 	if (rc)
5227 		return rc;
5228 
5229 	bnxt_hwrm_port_mac_qcfg(bp);
5230 
5231 	bnxt_hwrm_parent_pf_qcfg(bp);
5232 
5233 	bnxt_hwrm_port_phy_qcaps(bp);
5234 
5235 	bnxt_alloc_error_recovery_info(bp);
5236 	/* Get the adapter error recovery support info */
5237 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5238 	if (rc)
5239 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5240 
5241 	bnxt_hwrm_port_led_qcaps(bp);
5242 
5243 	return 0;
5244 }
5245 
5246 static int
5247 bnxt_init_locks(struct bnxt *bp)
5248 {
5249 	int err;
5250 
5251 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5252 	if (err) {
5253 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5254 		return err;
5255 	}
5256 
5257 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5258 	if (err) {
5259 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5260 		return err;
5261 	}
5262 
5263 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5264 	if (err) {
5265 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5266 		return err;
5267 	}
5268 
5269 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5270 	if (err)
5271 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5272 
5273 	return err;
5274 }
5275 
5276 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5277 {
5278 	int rc = 0;
5279 
5280 	rc = bnxt_get_config(bp);
5281 	if (rc)
5282 		return rc;
5283 
5284 	if (!reconfig_dev) {
5285 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5286 		if (rc)
5287 			return rc;
5288 	} else {
5289 		rc = bnxt_restore_dflt_mac(bp);
5290 		if (rc)
5291 			return rc;
5292 	}
5293 
5294 	bnxt_config_vf_req_fwd(bp);
5295 
5296 	rc = bnxt_hwrm_func_driver_register(bp);
5297 	if (rc) {
5298 		PMD_DRV_LOG(ERR, "Failed to register driver");
5299 		return -EBUSY;
5300 	}
5301 
5302 	if (BNXT_PF(bp)) {
5303 		if (bp->pdev->max_vfs) {
5304 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5305 			if (rc) {
5306 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5307 				return rc;
5308 			}
5309 		} else {
5310 			rc = bnxt_hwrm_allocate_pf_only(bp);
5311 			if (rc) {
5312 				PMD_DRV_LOG(ERR,
5313 					    "Failed to allocate PF resources");
5314 				return rc;
5315 			}
5316 		}
5317 	}
5318 
5319 	if (!reconfig_dev) {
5320 		bp->rss_conf.rss_key = rte_zmalloc("bnxt_rss_key",
5321 						   HW_HASH_KEY_SIZE, 0);
5322 		if (bp->rss_conf.rss_key == NULL) {
5323 			PMD_DRV_LOG(ERR, "port %u cannot allocate RSS hash key memory",
5324 				    bp->eth_dev->data->port_id);
5325 			return -ENOMEM;
5326 		}
5327 	}
5328 
5329 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5330 	if (rc)
5331 		return rc;
5332 
5333 	rc = bnxt_setup_int(bp);
5334 	if (rc)
5335 		return rc;
5336 
5337 	rc = bnxt_request_int(bp);
5338 	if (rc)
5339 		return rc;
5340 
5341 	rc = bnxt_init_ctx_mem(bp);
5342 	if (rc) {
5343 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5344 		return rc;
5345 	}
5346 
5347 	return 0;
5348 }
5349 
5350 static int
5351 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5352 			     const char *value, void *opaque_arg)
5353 {
5354 	struct bnxt *bp = opaque_arg;
5355 	unsigned long flow_xstat;
5356 	char *end = NULL;
5357 
5358 	if (!value || !opaque_arg) {
5359 		PMD_DRV_LOG(ERR,
5360 			    "Invalid parameter passed to flow_xstat devarg.\n");
5361 		return -EINVAL;
5362 	}
5363 
5364 	flow_xstat = strtoul(value, &end, 10);
5365 	if (end == NULL || *end != '\0' ||
5366 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5367 		PMD_DRV_LOG(ERR,
5368 			    "Invalid parameter passed to flow_xstat devarg.\n");
5369 		return -EINVAL;
5370 	}
5371 
5372 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5373 		PMD_DRV_LOG(ERR,
5374 			    "Invalid value passed to flow_xstat devarg.\n");
5375 		return -EINVAL;
5376 	}
5377 
5378 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5379 	if (BNXT_FLOW_XSTATS_EN(bp))
5380 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5381 
5382 	return 0;
5383 }
5384 
5385 static int
5386 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5387 					const char *value, void *opaque_arg)
5388 {
5389 	struct bnxt *bp = opaque_arg;
5390 	unsigned long max_num_kflows;
5391 	char *end = NULL;
5392 
5393 	if (!value || !opaque_arg) {
5394 		PMD_DRV_LOG(ERR,
5395 			"Invalid parameter passed to max_num_kflows devarg.\n");
5396 		return -EINVAL;
5397 	}
5398 
5399 	max_num_kflows = strtoul(value, &end, 10);
5400 	if (end == NULL || *end != '\0' ||
5401 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5402 		PMD_DRV_LOG(ERR,
5403 			"Invalid parameter passed to max_num_kflows devarg.\n");
5404 		return -EINVAL;
5405 	}
5406 
5407 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5408 		PMD_DRV_LOG(ERR,
5409 			"Invalid value passed to max_num_kflows devarg.\n");
5410 		return -EINVAL;
5411 	}
5412 
5413 	bp->max_num_kflows = max_num_kflows;
5414 	if (bp->max_num_kflows)
5415 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5416 				max_num_kflows);
5417 
5418 	return 0;
5419 }
5420 
5421 static int
5422 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5423 				 const char *value, void *opaque_arg)
5424 {
5425 	struct bnxt *bp = opaque_arg;
5426 	unsigned long app_id;
5427 	char *end = NULL;
5428 
5429 	if (!value || !opaque_arg) {
5430 		PMD_DRV_LOG(ERR,
5431 			    "Invalid parameter passed to app-id "
5432 			    "devargs.\n");
5433 		return -EINVAL;
5434 	}
5435 
5436 	app_id = strtoul(value, &end, 10);
5437 	if (end == NULL || *end != '\0' ||
5438 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5439 		PMD_DRV_LOG(ERR,
5440 			    "Invalid parameter passed to app_id "
5441 			    "devargs.\n");
5442 		return -EINVAL;
5443 	}
5444 
5445 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5446 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5447 			    (uint16_t)app_id);
5448 		return -EINVAL;
5449 	}
5450 
5451 	bp->app_id = app_id;
5452 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5453 
5454 	return 0;
5455 }
5456 
5457 static int
5458 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5459 			    const char *value, void *opaque_arg)
5460 {
5461 	struct bnxt_representor *vfr_bp = opaque_arg;
5462 	unsigned long rep_is_pf;
5463 	char *end = NULL;
5464 
5465 	if (!value || !opaque_arg) {
5466 		PMD_DRV_LOG(ERR,
5467 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5468 		return -EINVAL;
5469 	}
5470 
5471 	rep_is_pf = strtoul(value, &end, 10);
5472 	if (end == NULL || *end != '\0' ||
5473 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5474 		PMD_DRV_LOG(ERR,
5475 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5476 		return -EINVAL;
5477 	}
5478 
5479 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5480 		PMD_DRV_LOG(ERR,
5481 			    "Invalid value passed to rep_is_pf devargs.\n");
5482 		return -EINVAL;
5483 	}
5484 
5485 	vfr_bp->flags |= rep_is_pf;
5486 	if (BNXT_REP_PF(vfr_bp))
5487 		PMD_DRV_LOG(INFO, "PF representor\n");
5488 	else
5489 		PMD_DRV_LOG(INFO, "VF representor\n");
5490 
5491 	return 0;
5492 }
5493 
5494 static int
5495 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5496 			       const char *value, void *opaque_arg)
5497 {
5498 	struct bnxt_representor *vfr_bp = opaque_arg;
5499 	unsigned long rep_based_pf;
5500 	char *end = NULL;
5501 
5502 	if (!value || !opaque_arg) {
5503 		PMD_DRV_LOG(ERR,
5504 			    "Invalid parameter passed to rep_based_pf "
5505 			    "devargs.\n");
5506 		return -EINVAL;
5507 	}
5508 
5509 	rep_based_pf = strtoul(value, &end, 10);
5510 	if (end == NULL || *end != '\0' ||
5511 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5512 		PMD_DRV_LOG(ERR,
5513 			    "Invalid parameter passed to rep_based_pf "
5514 			    "devargs.\n");
5515 		return -EINVAL;
5516 	}
5517 
5518 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5519 		PMD_DRV_LOG(ERR,
5520 			    "Invalid value passed to rep_based_pf devargs.\n");
5521 		return -EINVAL;
5522 	}
5523 
5524 	vfr_bp->rep_based_pf = rep_based_pf;
5525 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5526 
5527 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5528 
5529 	return 0;
5530 }
5531 
5532 static int
5533 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5534 			    const char *value, void *opaque_arg)
5535 {
5536 	struct bnxt_representor *vfr_bp = opaque_arg;
5537 	unsigned long rep_q_r2f;
5538 	char *end = NULL;
5539 
5540 	if (!value || !opaque_arg) {
5541 		PMD_DRV_LOG(ERR,
5542 			    "Invalid parameter passed to rep_q_r2f "
5543 			    "devargs.\n");
5544 		return -EINVAL;
5545 	}
5546 
5547 	rep_q_r2f = strtoul(value, &end, 10);
5548 	if (end == NULL || *end != '\0' ||
5549 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5550 		PMD_DRV_LOG(ERR,
5551 			    "Invalid parameter passed to rep_q_r2f "
5552 			    "devargs.\n");
5553 		return -EINVAL;
5554 	}
5555 
5556 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5557 		PMD_DRV_LOG(ERR,
5558 			    "Invalid value passed to rep_q_r2f devargs.\n");
5559 		return -EINVAL;
5560 	}
5561 
5562 	vfr_bp->rep_q_r2f = rep_q_r2f;
5563 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5564 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5565 
5566 	return 0;
5567 }
5568 
5569 static int
5570 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5571 			    const char *value, void *opaque_arg)
5572 {
5573 	struct bnxt_representor *vfr_bp = opaque_arg;
5574 	unsigned long rep_q_f2r;
5575 	char *end = NULL;
5576 
5577 	if (!value || !opaque_arg) {
5578 		PMD_DRV_LOG(ERR,
5579 			    "Invalid parameter passed to rep_q_f2r "
5580 			    "devargs.\n");
5581 		return -EINVAL;
5582 	}
5583 
5584 	rep_q_f2r = strtoul(value, &end, 10);
5585 	if (end == NULL || *end != '\0' ||
5586 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5587 		PMD_DRV_LOG(ERR,
5588 			    "Invalid parameter passed to rep_q_f2r "
5589 			    "devargs.\n");
5590 		return -EINVAL;
5591 	}
5592 
5593 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5594 		PMD_DRV_LOG(ERR,
5595 			    "Invalid value passed to rep_q_f2r devargs.\n");
5596 		return -EINVAL;
5597 	}
5598 
5599 	vfr_bp->rep_q_f2r = rep_q_f2r;
5600 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5601 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5602 
5603 	return 0;
5604 }
5605 
5606 static int
5607 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5608 			     const char *value, void *opaque_arg)
5609 {
5610 	struct bnxt_representor *vfr_bp = opaque_arg;
5611 	unsigned long rep_fc_r2f;
5612 	char *end = NULL;
5613 
5614 	if (!value || !opaque_arg) {
5615 		PMD_DRV_LOG(ERR,
5616 			    "Invalid parameter passed to rep_fc_r2f "
5617 			    "devargs.\n");
5618 		return -EINVAL;
5619 	}
5620 
5621 	rep_fc_r2f = strtoul(value, &end, 10);
5622 	if (end == NULL || *end != '\0' ||
5623 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5624 		PMD_DRV_LOG(ERR,
5625 			    "Invalid parameter passed to rep_fc_r2f "
5626 			    "devargs.\n");
5627 		return -EINVAL;
5628 	}
5629 
5630 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5631 		PMD_DRV_LOG(ERR,
5632 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5633 		return -EINVAL;
5634 	}
5635 
5636 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5637 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5638 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5639 
5640 	return 0;
5641 }
5642 
5643 static int
5644 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5645 			     const char *value, void *opaque_arg)
5646 {
5647 	struct bnxt_representor *vfr_bp = opaque_arg;
5648 	unsigned long rep_fc_f2r;
5649 	char *end = NULL;
5650 
5651 	if (!value || !opaque_arg) {
5652 		PMD_DRV_LOG(ERR,
5653 			    "Invalid parameter passed to rep_fc_f2r "
5654 			    "devargs.\n");
5655 		return -EINVAL;
5656 	}
5657 
5658 	rep_fc_f2r = strtoul(value, &end, 10);
5659 	if (end == NULL || *end != '\0' ||
5660 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5661 		PMD_DRV_LOG(ERR,
5662 			    "Invalid parameter passed to rep_fc_f2r "
5663 			    "devargs.\n");
5664 		return -EINVAL;
5665 	}
5666 
5667 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5668 		PMD_DRV_LOG(ERR,
5669 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5670 		return -EINVAL;
5671 	}
5672 
5673 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5674 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5675 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5676 
5677 	return 0;
5678 }
5679 
5680 static int
5681 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5682 {
5683 	struct rte_kvargs *kvlist;
5684 	int ret;
5685 
5686 	if (devargs == NULL)
5687 		return 0;
5688 
5689 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5690 	if (kvlist == NULL)
5691 		return -EINVAL;
5692 
5693 	/*
5694 	 * Handler for "flow_xstat" devarg.
5695 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5696 	 */
5697 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5698 				 bnxt_parse_devarg_flow_xstat, bp);
5699 	if (ret)
5700 		goto err;
5701 
5702 	/*
5703 	 * Handler for "max_num_kflows" devarg.
5704 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5705 	 */
5706 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5707 				 bnxt_parse_devarg_max_num_kflows, bp);
5708 	if (ret)
5709 		goto err;
5710 
5711 err:
5712 	/*
5713 	 * Handler for "app-id" devarg.
5714 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5715 	 */
5716 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5717 			   bnxt_parse_devarg_app_id, bp);
5718 
5719 	rte_kvargs_free(kvlist);
5720 	return ret;
5721 }
5722 
5723 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5724 {
5725 	int rc = 0;
5726 
5727 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5728 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5729 		if (rc)
5730 			PMD_DRV_LOG(ERR,
5731 				    "Failed to alloc switch domain: %d\n", rc);
5732 		else
5733 			PMD_DRV_LOG(INFO,
5734 				    "Switch domain allocated %d\n",
5735 				    bp->switch_domain_id);
5736 	}
5737 
5738 	return rc;
5739 }
5740 
5741 /* Allocate and initialize various fields in bnxt struct that
5742  * need to be allocated/destroyed only once in the lifetime of the driver
5743  */
5744 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5745 {
5746 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5747 	struct bnxt *bp = eth_dev->data->dev_private;
5748 	int rc = 0;
5749 
5750 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5751 
5752 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5753 		bp->flags |= BNXT_FLAG_VF;
5754 
5755 	if (bnxt_p5_device(pci_dev->id.device_id))
5756 		bp->flags |= BNXT_FLAG_CHIP_P5;
5757 
5758 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5759 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5760 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5761 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5762 		bp->flags |= BNXT_FLAG_STINGRAY;
5763 
5764 	if (BNXT_TRUFLOW_EN(bp)) {
5765 		/* extra mbuf field is required to store CFA code from mark */
5766 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5767 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5768 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5769 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5770 		};
5771 		bnxt_cfa_code_dynfield_offset =
5772 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5773 		if (bnxt_cfa_code_dynfield_offset < 0) {
5774 			PMD_DRV_LOG(ERR,
5775 			    "Failed to register mbuf field for TruFlow mark\n");
5776 			return -rte_errno;
5777 		}
5778 	}
5779 
5780 	rc = bnxt_map_pci_bars(eth_dev);
5781 	if (rc) {
5782 		PMD_DRV_LOG(ERR,
5783 			    "Failed to initialize board rc: %x\n", rc);
5784 		return rc;
5785 	}
5786 
5787 	rc = bnxt_alloc_pf_info(bp);
5788 	if (rc)
5789 		return rc;
5790 
5791 	rc = bnxt_alloc_link_info(bp);
5792 	if (rc)
5793 		return rc;
5794 
5795 	rc = bnxt_alloc_parent_info(bp);
5796 	if (rc)
5797 		return rc;
5798 
5799 	rc = bnxt_alloc_hwrm_resources(bp);
5800 	if (rc) {
5801 		PMD_DRV_LOG(ERR,
5802 			    "Failed to allocate response buffer rc: %x\n", rc);
5803 		return rc;
5804 	}
5805 	rc = bnxt_alloc_leds_info(bp);
5806 	if (rc)
5807 		return rc;
5808 
5809 	rc = bnxt_alloc_cos_queues(bp);
5810 	if (rc)
5811 		return rc;
5812 
5813 	rc = bnxt_init_locks(bp);
5814 	if (rc)
5815 		return rc;
5816 
5817 	rc = bnxt_alloc_switch_domain(bp);
5818 	if (rc)
5819 		return rc;
5820 
5821 	return rc;
5822 }
5823 
5824 static int
5825 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5826 {
5827 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5828 	static int version_printed;
5829 	struct bnxt *bp;
5830 	int rc;
5831 
5832 	if (version_printed++ == 0)
5833 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5834 
5835 	eth_dev->dev_ops = &bnxt_dev_ops;
5836 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5837 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5838 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5839 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5840 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5841 
5842 	/*
5843 	 * For secondary processes, we don't initialise any further
5844 	 * as primary has already done this work.
5845 	 */
5846 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5847 		return 0;
5848 
5849 	rte_eth_copy_pci_info(eth_dev, pci_dev);
5850 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5851 
5852 	bp = eth_dev->data->dev_private;
5853 
5854 	/* Parse dev arguments passed on when starting the DPDK application. */
5855 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5856 	if (rc)
5857 		goto error_free;
5858 
5859 	rc = bnxt_drv_init(eth_dev);
5860 	if (rc)
5861 		goto error_free;
5862 
5863 	rc = bnxt_init_resources(bp, false);
5864 	if (rc)
5865 		goto error_free;
5866 
5867 	rc = bnxt_alloc_stats_mem(bp);
5868 	if (rc)
5869 		goto error_free;
5870 
5871 	PMD_DRV_LOG(INFO,
5872 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
5873 		    DRV_MODULE_NAME,
5874 		    pci_dev->mem_resource[0].phys_addr,
5875 		    pci_dev->mem_resource[0].addr);
5876 
5877 	return 0;
5878 
5879 error_free:
5880 	bnxt_dev_uninit(eth_dev);
5881 	return rc;
5882 }
5883 
5884 
5885 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5886 {
5887 	if (!ctx)
5888 		return;
5889 
5890 	rte_free(ctx->va);
5891 
5892 	ctx->va = NULL;
5893 	ctx->dma = RTE_BAD_IOVA;
5894 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5895 }
5896 
5897 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5898 {
5899 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5900 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5901 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
5902 				  bp->flow_stat->max_fc,
5903 				  false);
5904 
5905 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5906 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5907 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
5908 				  bp->flow_stat->max_fc,
5909 				  false);
5910 
5911 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5912 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5913 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5914 
5915 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5916 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5917 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5918 
5919 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5920 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5921 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5922 
5923 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5924 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5925 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5926 }
5927 
5928 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5929 {
5930 	bnxt_unregister_fc_ctx_mem(bp);
5931 
5932 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5933 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5934 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5935 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5936 }
5937 
5938 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5939 {
5940 	if (BNXT_FLOW_XSTATS_EN(bp))
5941 		bnxt_uninit_fc_ctx_mem(bp);
5942 }
5943 
5944 static void
5945 bnxt_free_error_recovery_info(struct bnxt *bp)
5946 {
5947 	rte_free(bp->recovery_info);
5948 	bp->recovery_info = NULL;
5949 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5950 }
5951 
5952 static int
5953 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5954 {
5955 	int rc;
5956 
5957 	bnxt_free_int(bp);
5958 	bnxt_free_mem(bp, reconfig_dev);
5959 
5960 	bnxt_hwrm_func_buf_unrgtr(bp);
5961 	if (bp->pf != NULL) {
5962 		rte_free(bp->pf->vf_req_buf);
5963 		bp->pf->vf_req_buf = NULL;
5964 	}
5965 
5966 	rc = bnxt_hwrm_func_driver_unregister(bp);
5967 	bp->flags &= ~BNXT_FLAG_REGISTERED;
5968 	bnxt_free_ctx_mem(bp);
5969 	if (!reconfig_dev) {
5970 		bnxt_free_hwrm_resources(bp);
5971 		bnxt_free_error_recovery_info(bp);
5972 		rte_free(bp->mcast_addr_list);
5973 		bp->mcast_addr_list = NULL;
5974 		rte_free(bp->rss_conf.rss_key);
5975 		bp->rss_conf.rss_key = NULL;
5976 	}
5977 
5978 	bnxt_uninit_ctx_mem(bp);
5979 
5980 	bnxt_free_flow_stats_info(bp);
5981 	if (bp->rep_info != NULL)
5982 		bnxt_free_switch_domain(bp);
5983 	bnxt_free_rep_info(bp);
5984 	rte_free(bp->ptp_cfg);
5985 	bp->ptp_cfg = NULL;
5986 	return rc;
5987 }
5988 
5989 static int
5990 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5991 {
5992 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5993 		return -EPERM;
5994 
5995 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5996 
5997 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5998 		bnxt_dev_close_op(eth_dev);
5999 
6000 	return 0;
6001 }
6002 
6003 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6004 {
6005 	struct bnxt *bp = eth_dev->data->dev_private;
6006 	struct rte_eth_dev *vf_rep_eth_dev;
6007 	int ret = 0, i;
6008 
6009 	if (!bp)
6010 		return -EINVAL;
6011 
6012 	for (i = 0; i < bp->num_reps; i++) {
6013 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6014 		if (!vf_rep_eth_dev)
6015 			continue;
6016 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6017 			    vf_rep_eth_dev->data->port_id);
6018 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6019 	}
6020 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6021 		    eth_dev->data->port_id);
6022 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6023 
6024 	return ret;
6025 }
6026 
6027 static void bnxt_free_rep_info(struct bnxt *bp)
6028 {
6029 	rte_free(bp->rep_info);
6030 	bp->rep_info = NULL;
6031 	rte_free(bp->cfa_code_map);
6032 	bp->cfa_code_map = NULL;
6033 }
6034 
6035 static int bnxt_init_rep_info(struct bnxt *bp)
6036 {
6037 	int i = 0, rc;
6038 
6039 	if (bp->rep_info)
6040 		return 0;
6041 
6042 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
6043 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS(bp),
6044 				   0);
6045 	if (!bp->rep_info) {
6046 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6047 		return -ENOMEM;
6048 	}
6049 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6050 				       sizeof(*bp->cfa_code_map) *
6051 				       BNXT_MAX_CFA_CODE, 0);
6052 	if (!bp->cfa_code_map) {
6053 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6054 		bnxt_free_rep_info(bp);
6055 		return -ENOMEM;
6056 	}
6057 
6058 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6059 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6060 
6061 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6062 	if (rc) {
6063 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6064 		bnxt_free_rep_info(bp);
6065 		return rc;
6066 	}
6067 
6068 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6069 	if (rc) {
6070 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6071 		bnxt_free_rep_info(bp);
6072 		return rc;
6073 	}
6074 
6075 	return rc;
6076 }
6077 
6078 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6079 			       struct rte_eth_devargs *eth_da,
6080 			       struct rte_eth_dev *backing_eth_dev,
6081 			       const char *dev_args)
6082 {
6083 	struct rte_eth_dev *vf_rep_eth_dev;
6084 	char name[RTE_ETH_NAME_MAX_LEN];
6085 	struct bnxt *backing_bp = backing_eth_dev->data->dev_private;
6086 	uint16_t max_vf_reps = BNXT_MAX_VF_REPS(backing_bp);
6087 
6088 	uint16_t num_rep;
6089 	int i, ret = 0;
6090 	struct rte_kvargs *kvlist = NULL;
6091 
6092 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6093 		return 0;
6094 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6095 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6096 			    eth_da->type);
6097 		return -ENOTSUP;
6098 	}
6099 	num_rep = eth_da->nb_representor_ports;
6100 	if (num_rep > max_vf_reps) {
6101 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6102 			    num_rep, max_vf_reps);
6103 		return -EINVAL;
6104 	}
6105 
6106 	if (num_rep >= RTE_MAX_ETHPORTS) {
6107 		PMD_DRV_LOG(ERR,
6108 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6109 			    num_rep, RTE_MAX_ETHPORTS);
6110 		return -EINVAL;
6111 	}
6112 
6113 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6114 		PMD_DRV_LOG(ERR,
6115 			    "Not a PF or trusted VF. No Representor support\n");
6116 		/* Returning an error is not an option.
6117 		 * Applications are not handling this correctly
6118 		 */
6119 		return 0;
6120 	}
6121 
6122 	if (bnxt_init_rep_info(backing_bp))
6123 		return 0;
6124 
6125 	for (i = 0; i < num_rep; i++) {
6126 		struct bnxt_representor representor = {
6127 			.vf_id = eth_da->representor_ports[i],
6128 			.switch_domain_id = backing_bp->switch_domain_id,
6129 			.parent_dev = backing_eth_dev
6130 		};
6131 
6132 		if (representor.vf_id >= max_vf_reps) {
6133 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6134 				    representor.vf_id, max_vf_reps);
6135 			continue;
6136 		}
6137 
6138 		/* representor port net_bdf_port */
6139 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6140 			 pci_dev->device.name, eth_da->representor_ports[i]);
6141 
6142 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6143 		if (kvlist) {
6144 			/*
6145 			 * Handler for "rep_is_pf" devarg.
6146 			 * Invoked as for ex: "-a 000:00:0d.0,
6147 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6148 			 */
6149 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6150 						 bnxt_parse_devarg_rep_is_pf,
6151 						 (void *)&representor);
6152 			if (ret) {
6153 				ret = -EINVAL;
6154 				goto err;
6155 			}
6156 			/*
6157 			 * Handler for "rep_based_pf" devarg.
6158 			 * Invoked as for ex: "-a 000:00:0d.0,
6159 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6160 			 */
6161 			ret = rte_kvargs_process(kvlist,
6162 						 BNXT_DEVARG_REP_BASED_PF,
6163 						 bnxt_parse_devarg_rep_based_pf,
6164 						 (void *)&representor);
6165 			if (ret) {
6166 				ret = -EINVAL;
6167 				goto err;
6168 			}
6169 			/*
6170 			 * Handler for "rep_based_pf" devarg.
6171 			 * Invoked as for ex: "-a 000:00:0d.0,
6172 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6173 			 */
6174 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6175 						 bnxt_parse_devarg_rep_q_r2f,
6176 						 (void *)&representor);
6177 			if (ret) {
6178 				ret = -EINVAL;
6179 				goto err;
6180 			}
6181 			/*
6182 			 * Handler for "rep_based_pf" devarg.
6183 			 * Invoked as for ex: "-a 000:00:0d.0,
6184 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6185 			 */
6186 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6187 						 bnxt_parse_devarg_rep_q_f2r,
6188 						 (void *)&representor);
6189 			if (ret) {
6190 				ret = -EINVAL;
6191 				goto err;
6192 			}
6193 			/*
6194 			 * Handler for "rep_based_pf" devarg.
6195 			 * Invoked as for ex: "-a 000:00:0d.0,
6196 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6197 			 */
6198 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6199 						 bnxt_parse_devarg_rep_fc_r2f,
6200 						 (void *)&representor);
6201 			if (ret) {
6202 				ret = -EINVAL;
6203 				goto err;
6204 			}
6205 			/*
6206 			 * Handler for "rep_based_pf" devarg.
6207 			 * Invoked as for ex: "-a 000:00:0d.0,
6208 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6209 			 */
6210 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6211 						 bnxt_parse_devarg_rep_fc_f2r,
6212 						 (void *)&representor);
6213 			if (ret) {
6214 				ret = -EINVAL;
6215 				goto err;
6216 			}
6217 		}
6218 
6219 		ret = rte_eth_dev_create(&pci_dev->device, name,
6220 					 sizeof(struct bnxt_representor),
6221 					 NULL, NULL,
6222 					 bnxt_representor_init,
6223 					 &representor);
6224 		if (ret) {
6225 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6226 				    "representor %s.", name);
6227 			goto err;
6228 		}
6229 
6230 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6231 		if (!vf_rep_eth_dev) {
6232 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6233 				    " for VF-Rep: %s.", name);
6234 			ret = -ENODEV;
6235 			goto err;
6236 		}
6237 
6238 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6239 			    backing_eth_dev->data->port_id);
6240 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6241 							 vf_rep_eth_dev;
6242 		backing_bp->num_reps++;
6243 
6244 	}
6245 
6246 	rte_kvargs_free(kvlist);
6247 	return 0;
6248 
6249 err:
6250 	/* If num_rep > 1, then rollback already created
6251 	 * ports, since we'll be failing the probe anyway
6252 	 */
6253 	if (num_rep > 1)
6254 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6255 	rte_errno = -ret;
6256 	rte_kvargs_free(kvlist);
6257 
6258 	return ret;
6259 }
6260 
6261 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6262 			  struct rte_pci_device *pci_dev)
6263 {
6264 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6265 	struct rte_eth_dev *backing_eth_dev;
6266 	uint16_t num_rep;
6267 	int ret = 0;
6268 
6269 	if (pci_dev->device.devargs) {
6270 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6271 					    &eth_da);
6272 		if (ret)
6273 			return ret;
6274 	}
6275 
6276 	num_rep = eth_da.nb_representor_ports;
6277 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6278 		    num_rep);
6279 
6280 	/* We could come here after first level of probe is already invoked
6281 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6282 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6283 	 */
6284 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6285 	if (backing_eth_dev == NULL) {
6286 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6287 					 sizeof(struct bnxt),
6288 					 eth_dev_pci_specific_init, pci_dev,
6289 					 bnxt_dev_init, NULL);
6290 
6291 		if (ret || !num_rep)
6292 			return ret;
6293 
6294 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6295 	}
6296 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6297 		    backing_eth_dev->data->port_id);
6298 
6299 	if (!num_rep)
6300 		return ret;
6301 
6302 	/* probe representor ports now */
6303 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6304 				  pci_dev->device.devargs->args);
6305 
6306 	return ret;
6307 }
6308 
6309 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6310 {
6311 	struct rte_eth_dev *eth_dev;
6312 
6313 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6314 	if (!eth_dev)
6315 		return 0; /* Invoked typically only by OVS-DPDK, by the
6316 			   * time it comes here the eth_dev is already
6317 			   * deleted by rte_eth_dev_close(), so returning
6318 			   * +ve value will at least help in proper cleanup
6319 			   */
6320 
6321 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6322 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6323 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6324 			return rte_eth_dev_destroy(eth_dev,
6325 						   bnxt_representor_uninit);
6326 		else
6327 			return rte_eth_dev_destroy(eth_dev,
6328 						   bnxt_dev_uninit);
6329 	} else {
6330 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6331 	}
6332 }
6333 
6334 static struct rte_pci_driver bnxt_rte_pmd = {
6335 	.id_table = bnxt_pci_id_map,
6336 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6337 			RTE_PCI_DRV_INTR_RMV |
6338 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6339 						  * and OVS-DPDK
6340 						  */
6341 	.probe = bnxt_pci_probe,
6342 	.remove = bnxt_pci_remove,
6343 };
6344 
6345 static bool
6346 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6347 {
6348 	if (strcmp(dev->device->driver->name, drv->driver.name))
6349 		return false;
6350 
6351 	return true;
6352 }
6353 
6354 bool is_bnxt_supported(struct rte_eth_dev *dev)
6355 {
6356 	return is_device_supported(dev, &bnxt_rte_pmd);
6357 }
6358 
6359 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6360 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6361 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6362 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6363