xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision 2addc463)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2014-2021 Broadcom
3  * All rights reserved.
4  */
5 
6 #include <inttypes.h>
7 #include <stdbool.h>
8 
9 #include <rte_dev.h>
10 #include <ethdev_driver.h>
11 #include <ethdev_pci.h>
12 #include <rte_malloc.h>
13 #include <rte_cycles.h>
14 #include <rte_alarm.h>
15 #include <rte_kvargs.h>
16 #include <rte_vect.h>
17 
18 #include "bnxt.h"
19 #include "bnxt_filter.h"
20 #include "bnxt_hwrm.h"
21 #include "bnxt_irq.h"
22 #include "bnxt_reps.h"
23 #include "bnxt_ring.h"
24 #include "bnxt_rxq.h"
25 #include "bnxt_rxr.h"
26 #include "bnxt_stats.h"
27 #include "bnxt_txq.h"
28 #include "bnxt_txr.h"
29 #include "bnxt_vnic.h"
30 #include "hsi_struct_def_dpdk.h"
31 #include "bnxt_nvm_defs.h"
32 #include "bnxt_tf_common.h"
33 #include "ulp_flow_db.h"
34 #include "rte_pmd_bnxt.h"
35 
36 #define DRV_MODULE_NAME		"bnxt"
37 static const char bnxt_version[] =
38 	"Broadcom NetXtreme driver " DRV_MODULE_NAME;
39 
40 /*
41  * The set of PCI devices this driver supports
42  */
43 static const struct rte_pci_id bnxt_pci_id_map[] = {
44 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
45 			 BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
46 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
47 			 BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
48 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
49 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
50 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
51 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
52 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
53 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
54 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
55 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
56 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
57 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
58 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
59 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
60 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
61 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
62 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
63 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
64 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
65 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
66 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
67 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
68 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
69 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
70 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
71 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
72 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
73 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
74 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
75 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
76 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
77 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
78 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
79 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
80 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
81 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
82 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
83 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58812) },
84 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58814) },
85 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818) },
86 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818_VF) },
87 	{ .vendor_id = 0, /* sentinel */ },
88 };
89 
90 #define BNXT_DEVARG_TRUFLOW	"host-based-truflow"
91 #define BNXT_DEVARG_FLOW_XSTAT	"flow-xstat"
92 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
93 #define BNXT_DEVARG_REPRESENTOR	"representor"
94 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
95 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
96 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
97 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
98 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
99 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
100 
101 static const char *const bnxt_dev_args[] = {
102 	BNXT_DEVARG_REPRESENTOR,
103 	BNXT_DEVARG_TRUFLOW,
104 	BNXT_DEVARG_FLOW_XSTAT,
105 	BNXT_DEVARG_MAX_NUM_KFLOWS,
106 	BNXT_DEVARG_REP_BASED_PF,
107 	BNXT_DEVARG_REP_IS_PF,
108 	BNXT_DEVARG_REP_Q_R2F,
109 	BNXT_DEVARG_REP_Q_F2R,
110 	BNXT_DEVARG_REP_FC_R2F,
111 	BNXT_DEVARG_REP_FC_F2R,
112 	NULL
113 };
114 
115 /*
116  * truflow == false to disable the feature
117  * truflow == true to enable the feature
118  */
119 #define	BNXT_DEVARG_TRUFLOW_INVALID(truflow)	((truflow) > 1)
120 
121 /*
122  * flow_xstat == false to disable the feature
123  * flow_xstat == true to enable the feature
124  */
125 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
126 
127 /*
128  * rep_is_pf == false to indicate VF representor
129  * rep_is_pf == true to indicate PF representor
130  */
131 #define	BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)	((rep_is_pf) > 1)
132 
133 /*
134  * rep_based_pf == Physical index of the PF
135  */
136 #define	BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)	((rep_based_pf) > 15)
137 /*
138  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
139  */
140 #define	BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)	((rep_q_r2f) > 3)
141 
142 /*
143  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
144  */
145 #define	BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)	((rep_q_f2r) > 3)
146 
147 /*
148  * rep_fc_r2f == Flow control for the representor to endpoint direction
149  */
150 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)	((rep_fc_r2f) > 1)
151 
152 /*
153  * rep_fc_f2r == Flow control for the endpoint to representor direction
154  */
155 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)	((rep_fc_f2r) > 1)
156 
157 int bnxt_cfa_code_dynfield_offset = -1;
158 
159 /*
160  * max_num_kflows must be >= 32
161  * and must be a power-of-2 supported value
162  * return: 1 -> invalid
163  *         0 -> valid
164  */
165 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
166 {
167 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
168 		return 1;
169 	return 0;
170 }
171 
172 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
173 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
174 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
175 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
176 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
177 static int bnxt_restore_vlan_filters(struct bnxt *bp);
178 static void bnxt_dev_recover(void *arg);
179 static void bnxt_free_error_recovery_info(struct bnxt *bp);
180 static void bnxt_free_rep_info(struct bnxt *bp);
181 
182 int is_bnxt_in_error(struct bnxt *bp)
183 {
184 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
185 		return -EIO;
186 	if (bp->flags & BNXT_FLAG_FW_RESET)
187 		return -EBUSY;
188 
189 	return 0;
190 }
191 
192 /***********************/
193 
194 /*
195  * High level utility functions
196  */
197 
198 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
199 {
200 	unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
201 					     BNXT_RSS_TBL_SIZE_P5);
202 
203 	if (!BNXT_CHIP_P5(bp))
204 		return 1;
205 
206 	return RTE_ALIGN_MUL_CEIL(num_rss_rings,
207 				  BNXT_RSS_ENTRIES_PER_CTX_P5) /
208 				  BNXT_RSS_ENTRIES_PER_CTX_P5;
209 }
210 
211 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
212 {
213 	if (!BNXT_CHIP_P5(bp))
214 		return HW_HASH_INDEX_SIZE;
215 
216 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
217 }
218 
219 static void bnxt_free_parent_info(struct bnxt *bp)
220 {
221 	rte_free(bp->parent);
222 }
223 
224 static void bnxt_free_pf_info(struct bnxt *bp)
225 {
226 	rte_free(bp->pf);
227 }
228 
229 static void bnxt_free_link_info(struct bnxt *bp)
230 {
231 	rte_free(bp->link_info);
232 }
233 
234 static void bnxt_free_leds_info(struct bnxt *bp)
235 {
236 	if (BNXT_VF(bp))
237 		return;
238 
239 	rte_free(bp->leds);
240 	bp->leds = NULL;
241 }
242 
243 static void bnxt_free_flow_stats_info(struct bnxt *bp)
244 {
245 	rte_free(bp->flow_stat);
246 	bp->flow_stat = NULL;
247 }
248 
249 static void bnxt_free_cos_queues(struct bnxt *bp)
250 {
251 	rte_free(bp->rx_cos_queue);
252 	rte_free(bp->tx_cos_queue);
253 }
254 
255 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
256 {
257 	bnxt_free_filter_mem(bp);
258 	bnxt_free_vnic_attributes(bp);
259 	bnxt_free_vnic_mem(bp);
260 
261 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
262 	 * If the number of rings change across fw update,
263 	 * we don't have much choice except to warn the user.
264 	 */
265 	if (!reconfig) {
266 		bnxt_free_stats(bp);
267 		bnxt_free_tx_rings(bp);
268 		bnxt_free_rx_rings(bp);
269 	}
270 	bnxt_free_async_cp_ring(bp);
271 	bnxt_free_rxtx_nq_ring(bp);
272 
273 	rte_free(bp->grp_info);
274 	bp->grp_info = NULL;
275 }
276 
277 static int bnxt_alloc_parent_info(struct bnxt *bp)
278 {
279 	bp->parent = rte_zmalloc("bnxt_parent_info",
280 				 sizeof(struct bnxt_parent_info), 0);
281 	if (bp->parent == NULL)
282 		return -ENOMEM;
283 
284 	return 0;
285 }
286 
287 static int bnxt_alloc_pf_info(struct bnxt *bp)
288 {
289 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
290 	if (bp->pf == NULL)
291 		return -ENOMEM;
292 
293 	return 0;
294 }
295 
296 static int bnxt_alloc_link_info(struct bnxt *bp)
297 {
298 	bp->link_info =
299 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
300 	if (bp->link_info == NULL)
301 		return -ENOMEM;
302 
303 	return 0;
304 }
305 
306 static int bnxt_alloc_leds_info(struct bnxt *bp)
307 {
308 	if (BNXT_VF(bp))
309 		return 0;
310 
311 	bp->leds = rte_zmalloc("bnxt_leds",
312 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
313 			       0);
314 	if (bp->leds == NULL)
315 		return -ENOMEM;
316 
317 	return 0;
318 }
319 
320 static int bnxt_alloc_cos_queues(struct bnxt *bp)
321 {
322 	bp->rx_cos_queue =
323 		rte_zmalloc("bnxt_rx_cosq",
324 			    BNXT_COS_QUEUE_COUNT *
325 			    sizeof(struct bnxt_cos_queue_info),
326 			    0);
327 	if (bp->rx_cos_queue == NULL)
328 		return -ENOMEM;
329 
330 	bp->tx_cos_queue =
331 		rte_zmalloc("bnxt_tx_cosq",
332 			    BNXT_COS_QUEUE_COUNT *
333 			    sizeof(struct bnxt_cos_queue_info),
334 			    0);
335 	if (bp->tx_cos_queue == NULL)
336 		return -ENOMEM;
337 
338 	return 0;
339 }
340 
341 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
342 {
343 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
344 				    sizeof(struct bnxt_flow_stat_info), 0);
345 	if (bp->flow_stat == NULL)
346 		return -ENOMEM;
347 
348 	return 0;
349 }
350 
351 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
352 {
353 	int rc;
354 
355 	rc = bnxt_alloc_ring_grps(bp);
356 	if (rc)
357 		goto alloc_mem_err;
358 
359 	rc = bnxt_alloc_async_ring_struct(bp);
360 	if (rc)
361 		goto alloc_mem_err;
362 
363 	rc = bnxt_alloc_vnic_mem(bp);
364 	if (rc)
365 		goto alloc_mem_err;
366 
367 	rc = bnxt_alloc_vnic_attributes(bp);
368 	if (rc)
369 		goto alloc_mem_err;
370 
371 	rc = bnxt_alloc_filter_mem(bp);
372 	if (rc)
373 		goto alloc_mem_err;
374 
375 	rc = bnxt_alloc_async_cp_ring(bp);
376 	if (rc)
377 		goto alloc_mem_err;
378 
379 	rc = bnxt_alloc_rxtx_nq_ring(bp);
380 	if (rc)
381 		goto alloc_mem_err;
382 
383 	if (BNXT_FLOW_XSTATS_EN(bp)) {
384 		rc = bnxt_alloc_flow_stats_info(bp);
385 		if (rc)
386 			goto alloc_mem_err;
387 	}
388 
389 	return 0;
390 
391 alloc_mem_err:
392 	bnxt_free_mem(bp, reconfig);
393 	return rc;
394 }
395 
396 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
397 {
398 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
399 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
400 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
401 	struct bnxt_rx_queue *rxq;
402 	unsigned int j;
403 	int rc;
404 
405 	rc = bnxt_vnic_grp_alloc(bp, vnic);
406 	if (rc)
407 		goto err_out;
408 
409 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
410 		    vnic_id, vnic, vnic->fw_grp_ids);
411 
412 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
413 	if (rc)
414 		goto err_out;
415 
416 	/* Alloc RSS context only if RSS mode is enabled */
417 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
418 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
419 
420 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
421 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
422 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
423 			PMD_DRV_LOG(ERR,
424 				    "Only queues 0-%d will be in RSS table\n",
425 				    BNXT_RSS_TBL_SIZE_P5 - 1);
426 		}
427 
428 		rc = 0;
429 		for (j = 0; j < nr_ctxs; j++) {
430 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
431 			if (rc)
432 				break;
433 		}
434 		if (rc) {
435 			PMD_DRV_LOG(ERR,
436 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
437 				    vnic_id, j, rc);
438 			goto err_out;
439 		}
440 		vnic->num_lb_ctxts = nr_ctxs;
441 	}
442 
443 	/*
444 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
445 	 * setting is not available at this time, it will not be
446 	 * configured correctly in the CFA.
447 	 */
448 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
449 		vnic->vlan_strip = true;
450 	else
451 		vnic->vlan_strip = false;
452 
453 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
454 	if (rc)
455 		goto err_out;
456 
457 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
458 	if (rc)
459 		goto err_out;
460 
461 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
462 		rxq = bp->eth_dev->data->rx_queues[j];
463 
464 		PMD_DRV_LOG(DEBUG,
465 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
466 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
467 
468 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
469 			rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
470 		else
471 			vnic->rx_queue_cnt++;
472 	}
473 
474 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
475 
476 	rc = bnxt_vnic_rss_configure(bp, vnic);
477 	if (rc)
478 		goto err_out;
479 
480 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
481 
482 	if (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO)
483 		bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 1);
484 	else
485 		bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 0);
486 
487 	return 0;
488 err_out:
489 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
490 		    vnic_id, rc);
491 	return rc;
492 }
493 
494 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
495 {
496 	int rc = 0;
497 
498 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
499 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
500 	if (rc)
501 		return rc;
502 
503 	PMD_DRV_LOG(DEBUG,
504 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
505 		    " rx_fc_in_tbl.ctx_id = %d\n",
506 		    bp->flow_stat->rx_fc_in_tbl.va,
507 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
508 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
509 
510 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
511 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
512 	if (rc)
513 		return rc;
514 
515 	PMD_DRV_LOG(DEBUG,
516 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
517 		    " rx_fc_out_tbl.ctx_id = %d\n",
518 		    bp->flow_stat->rx_fc_out_tbl.va,
519 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
520 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
521 
522 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
523 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
524 	if (rc)
525 		return rc;
526 
527 	PMD_DRV_LOG(DEBUG,
528 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
529 		    " tx_fc_in_tbl.ctx_id = %d\n",
530 		    bp->flow_stat->tx_fc_in_tbl.va,
531 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
532 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
533 
534 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
535 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
536 	if (rc)
537 		return rc;
538 
539 	PMD_DRV_LOG(DEBUG,
540 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
541 		    " tx_fc_out_tbl.ctx_id = %d\n",
542 		    bp->flow_stat->tx_fc_out_tbl.va,
543 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
544 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
545 
546 	memset(bp->flow_stat->rx_fc_out_tbl.va,
547 	       0,
548 	       bp->flow_stat->rx_fc_out_tbl.size);
549 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
550 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
551 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
552 				       bp->flow_stat->max_fc,
553 				       true);
554 	if (rc)
555 		return rc;
556 
557 	memset(bp->flow_stat->tx_fc_out_tbl.va,
558 	       0,
559 	       bp->flow_stat->tx_fc_out_tbl.size);
560 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
561 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
562 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
563 				       bp->flow_stat->max_fc,
564 				       true);
565 
566 	return rc;
567 }
568 
569 static int bnxt_alloc_ctx_mem_buf(char *type, size_t size,
570 				  struct bnxt_ctx_mem_buf_info *ctx)
571 {
572 	if (!ctx)
573 		return -EINVAL;
574 
575 	ctx->va = rte_zmalloc(type, size, 0);
576 	if (ctx->va == NULL)
577 		return -ENOMEM;
578 	rte_mem_lock_page(ctx->va);
579 	ctx->size = size;
580 	ctx->dma = rte_mem_virt2iova(ctx->va);
581 	if (ctx->dma == RTE_BAD_IOVA)
582 		return -ENOMEM;
583 
584 	return 0;
585 }
586 
587 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
588 {
589 	struct rte_pci_device *pdev = bp->pdev;
590 	char type[RTE_MEMZONE_NAMESIZE];
591 	uint16_t max_fc;
592 	int rc = 0;
593 
594 	max_fc = bp->flow_stat->max_fc;
595 
596 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
597 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
598 	/* 4 bytes for each counter-id */
599 	rc = bnxt_alloc_ctx_mem_buf(type,
600 				    max_fc * 4,
601 				    &bp->flow_stat->rx_fc_in_tbl);
602 	if (rc)
603 		return rc;
604 
605 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
606 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
607 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
608 	rc = bnxt_alloc_ctx_mem_buf(type,
609 				    max_fc * 16,
610 				    &bp->flow_stat->rx_fc_out_tbl);
611 	if (rc)
612 		return rc;
613 
614 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
615 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
616 	/* 4 bytes for each counter-id */
617 	rc = bnxt_alloc_ctx_mem_buf(type,
618 				    max_fc * 4,
619 				    &bp->flow_stat->tx_fc_in_tbl);
620 	if (rc)
621 		return rc;
622 
623 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
624 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
625 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
626 	rc = bnxt_alloc_ctx_mem_buf(type,
627 				    max_fc * 16,
628 				    &bp->flow_stat->tx_fc_out_tbl);
629 	if (rc)
630 		return rc;
631 
632 	rc = bnxt_register_fc_ctx_mem(bp);
633 
634 	return rc;
635 }
636 
637 static int bnxt_init_ctx_mem(struct bnxt *bp)
638 {
639 	int rc = 0;
640 
641 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
642 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
643 	    !BNXT_FLOW_XSTATS_EN(bp))
644 		return 0;
645 
646 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
647 	if (rc)
648 		return rc;
649 
650 	rc = bnxt_init_fc_ctx_mem(bp);
651 
652 	return rc;
653 }
654 
655 static int bnxt_update_phy_setting(struct bnxt *bp)
656 {
657 	struct rte_eth_link new;
658 	int rc;
659 
660 	rc = bnxt_get_hwrm_link_config(bp, &new);
661 	if (rc) {
662 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
663 		return rc;
664 	}
665 
666 	/*
667 	 * On BCM957508-N2100 adapters, FW will not allow any user other
668 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
669 	 * always returns link up. Force phy update always in that case.
670 	 */
671 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
672 		rc = bnxt_set_hwrm_link_config(bp, true);
673 		if (rc) {
674 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
675 			return rc;
676 		}
677 	}
678 
679 	return rc;
680 }
681 
682 static int bnxt_start_nic(struct bnxt *bp)
683 {
684 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
685 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
686 	uint32_t intr_vector = 0;
687 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
688 	uint32_t vec = BNXT_MISC_VEC_ID;
689 	unsigned int i, j;
690 	int rc;
691 
692 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
693 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
694 			DEV_RX_OFFLOAD_JUMBO_FRAME;
695 		bp->flags |= BNXT_FLAG_JUMBO;
696 	} else {
697 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
698 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
699 		bp->flags &= ~BNXT_FLAG_JUMBO;
700 	}
701 
702 	/* THOR does not support ring groups.
703 	 * But we will use the array to save RSS context IDs.
704 	 */
705 	if (BNXT_CHIP_P5(bp))
706 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
707 
708 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
709 	if (rc) {
710 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
711 		goto err_out;
712 	}
713 
714 	rc = bnxt_alloc_hwrm_rings(bp);
715 	if (rc) {
716 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
717 		goto err_out;
718 	}
719 
720 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
721 	if (rc) {
722 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
723 		goto err_out;
724 	}
725 
726 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
727 		goto skip_cosq_cfg;
728 
729 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
730 		if (bp->rx_cos_queue[i].id != 0xff) {
731 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
732 
733 			if (!vnic) {
734 				PMD_DRV_LOG(ERR,
735 					    "Num pools more than FW profile\n");
736 				rc = -EINVAL;
737 				goto err_out;
738 			}
739 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
740 			bp->rx_cosq_cnt++;
741 		}
742 	}
743 
744 skip_cosq_cfg:
745 	rc = bnxt_mq_rx_configure(bp);
746 	if (rc) {
747 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
748 		goto err_out;
749 	}
750 
751 	/* default vnic 0 */
752 	rc = bnxt_setup_one_vnic(bp, 0);
753 	if (rc)
754 		goto err_out;
755 	/* VNIC configuration */
756 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
757 		for (i = 1; i < bp->nr_vnics; i++) {
758 			rc = bnxt_setup_one_vnic(bp, i);
759 			if (rc)
760 				goto err_out;
761 		}
762 	}
763 
764 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
765 	if (rc) {
766 		PMD_DRV_LOG(ERR,
767 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
768 		goto err_out;
769 	}
770 
771 	/* check and configure queue intr-vector mapping */
772 	if ((rte_intr_cap_multiple(intr_handle) ||
773 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
774 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
775 		intr_vector = bp->eth_dev->data->nb_rx_queues;
776 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
777 		if (intr_vector > bp->rx_cp_nr_rings) {
778 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
779 					bp->rx_cp_nr_rings);
780 			return -ENOTSUP;
781 		}
782 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
783 		if (rc)
784 			return rc;
785 	}
786 
787 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
788 		intr_handle->intr_vec =
789 			rte_zmalloc("intr_vec",
790 				    bp->eth_dev->data->nb_rx_queues *
791 				    sizeof(int), 0);
792 		if (intr_handle->intr_vec == NULL) {
793 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
794 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
795 			rc = -ENOMEM;
796 			goto err_disable;
797 		}
798 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
799 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
800 			 intr_handle->intr_vec, intr_handle->nb_efd,
801 			intr_handle->max_intr);
802 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
803 		     queue_id++) {
804 			intr_handle->intr_vec[queue_id] =
805 							vec + BNXT_RX_VEC_START;
806 			if (vec < base + intr_handle->nb_efd - 1)
807 				vec++;
808 		}
809 	}
810 
811 	/* enable uio/vfio intr/eventfd mapping */
812 	rc = rte_intr_enable(intr_handle);
813 #ifndef RTE_EXEC_ENV_FREEBSD
814 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
815 	if (rc)
816 		goto err_free;
817 #endif
818 
819 	rc = bnxt_update_phy_setting(bp);
820 	if (rc)
821 		goto err_free;
822 
823 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
824 	if (!bp->mark_table)
825 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
826 
827 	return 0;
828 
829 err_free:
830 	rte_free(intr_handle->intr_vec);
831 err_disable:
832 	rte_intr_efd_disable(intr_handle);
833 err_out:
834 	/* Some of the error status returned by FW may not be from errno.h */
835 	if (rc > 0)
836 		rc = -EIO;
837 
838 	return rc;
839 }
840 
841 static int bnxt_shutdown_nic(struct bnxt *bp)
842 {
843 	bnxt_free_all_hwrm_resources(bp);
844 	bnxt_free_all_filters(bp);
845 	bnxt_free_all_vnics(bp);
846 	return 0;
847 }
848 
849 /*
850  * Device configuration and status function
851  */
852 
853 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
854 {
855 	uint32_t link_speed = bp->link_info->support_speeds;
856 	uint32_t speed_capa = 0;
857 
858 	/* If PAM4 is configured, use PAM4 supported speed */
859 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
860 		link_speed = bp->link_info->support_pam4_speeds;
861 
862 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
863 		speed_capa |= ETH_LINK_SPEED_100M;
864 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
865 		speed_capa |= ETH_LINK_SPEED_100M_HD;
866 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
867 		speed_capa |= ETH_LINK_SPEED_1G;
868 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
869 		speed_capa |= ETH_LINK_SPEED_2_5G;
870 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
871 		speed_capa |= ETH_LINK_SPEED_10G;
872 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
873 		speed_capa |= ETH_LINK_SPEED_20G;
874 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
875 		speed_capa |= ETH_LINK_SPEED_25G;
876 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
877 		speed_capa |= ETH_LINK_SPEED_40G;
878 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
879 		speed_capa |= ETH_LINK_SPEED_50G;
880 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
881 		speed_capa |= ETH_LINK_SPEED_100G;
882 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
883 		speed_capa |= ETH_LINK_SPEED_50G;
884 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
885 		speed_capa |= ETH_LINK_SPEED_100G;
886 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
887 		speed_capa |= ETH_LINK_SPEED_200G;
888 
889 	if (bp->link_info->auto_mode ==
890 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
891 		speed_capa |= ETH_LINK_SPEED_FIXED;
892 	else
893 		speed_capa |= ETH_LINK_SPEED_AUTONEG;
894 
895 	return speed_capa;
896 }
897 
898 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
899 				struct rte_eth_dev_info *dev_info)
900 {
901 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
902 	struct bnxt *bp = eth_dev->data->dev_private;
903 	uint16_t max_vnics, i, j, vpool, vrxq;
904 	unsigned int max_rx_rings;
905 	int rc;
906 
907 	rc = is_bnxt_in_error(bp);
908 	if (rc)
909 		return rc;
910 
911 	/* MAC Specifics */
912 	dev_info->max_mac_addrs = bp->max_l2_ctx;
913 	dev_info->max_hash_mac_addrs = 0;
914 
915 	/* PF/VF specifics */
916 	if (BNXT_PF(bp))
917 		dev_info->max_vfs = pdev->max_vfs;
918 
919 	max_rx_rings = bnxt_max_rings(bp);
920 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
921 	dev_info->max_rx_queues = max_rx_rings;
922 	dev_info->max_tx_queues = max_rx_rings;
923 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
924 	dev_info->hash_key_size = 40;
925 	max_vnics = bp->max_vnics;
926 
927 	/* MTU specifics */
928 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
929 	dev_info->max_mtu = BNXT_MAX_MTU;
930 
931 	/* Fast path specifics */
932 	dev_info->min_rx_bufsize = 1;
933 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
934 
935 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
936 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
937 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
938 	dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
939 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
940 				    dev_info->tx_queue_offload_capa;
941 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
942 
943 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
944 
945 	/* *INDENT-OFF* */
946 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
947 		.rx_thresh = {
948 			.pthresh = 8,
949 			.hthresh = 8,
950 			.wthresh = 0,
951 		},
952 		.rx_free_thresh = 32,
953 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
954 	};
955 
956 	dev_info->default_txconf = (struct rte_eth_txconf) {
957 		.tx_thresh = {
958 			.pthresh = 32,
959 			.hthresh = 0,
960 			.wthresh = 0,
961 		},
962 		.tx_free_thresh = 32,
963 		.tx_rs_thresh = 32,
964 	};
965 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
966 
967 	eth_dev->data->dev_conf.intr_conf.rxq = 1;
968 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
969 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
970 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
971 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
972 
973 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
974 		dev_info->switch_info.name = eth_dev->device->name;
975 		dev_info->switch_info.domain_id = bp->switch_domain_id;
976 		dev_info->switch_info.port_id =
977 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
978 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
979 	}
980 
981 	/* *INDENT-ON* */
982 
983 	/*
984 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
985 	 *       need further investigation.
986 	 */
987 
988 	/* VMDq resources */
989 	vpool = 64; /* ETH_64_POOLS */
990 	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
991 	for (i = 0; i < 4; vpool >>= 1, i++) {
992 		if (max_vnics > vpool) {
993 			for (j = 0; j < 5; vrxq >>= 1, j++) {
994 				if (dev_info->max_rx_queues > vrxq) {
995 					if (vpool > vrxq)
996 						vpool = vrxq;
997 					goto found;
998 				}
999 			}
1000 			/* Not enough resources to support VMDq */
1001 			break;
1002 		}
1003 	}
1004 	/* Not enough resources to support VMDq */
1005 	vpool = 0;
1006 	vrxq = 0;
1007 found:
1008 	dev_info->max_vmdq_pools = vpool;
1009 	dev_info->vmdq_queue_num = vrxq;
1010 
1011 	dev_info->vmdq_pool_base = 0;
1012 	dev_info->vmdq_queue_base = 0;
1013 
1014 	return 0;
1015 }
1016 
1017 /* Configure the device based on the configuration provided */
1018 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1019 {
1020 	struct bnxt *bp = eth_dev->data->dev_private;
1021 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1022 	int rc;
1023 
1024 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1025 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1026 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1027 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1028 
1029 	rc = is_bnxt_in_error(bp);
1030 	if (rc)
1031 		return rc;
1032 
1033 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1034 		rc = bnxt_hwrm_check_vf_rings(bp);
1035 		if (rc) {
1036 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1037 			return -ENOSPC;
1038 		}
1039 
1040 		/* If a resource has already been allocated - in this case
1041 		 * it is the async completion ring, free it. Reallocate it after
1042 		 * resource reservation. This will ensure the resource counts
1043 		 * are calculated correctly.
1044 		 */
1045 
1046 		pthread_mutex_lock(&bp->def_cp_lock);
1047 
1048 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1049 			bnxt_disable_int(bp);
1050 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1051 		}
1052 
1053 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1054 		if (rc) {
1055 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1056 			pthread_mutex_unlock(&bp->def_cp_lock);
1057 			return -ENOSPC;
1058 		}
1059 
1060 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1061 			rc = bnxt_alloc_async_cp_ring(bp);
1062 			if (rc) {
1063 				pthread_mutex_unlock(&bp->def_cp_lock);
1064 				return rc;
1065 			}
1066 			bnxt_enable_int(bp);
1067 		}
1068 
1069 		pthread_mutex_unlock(&bp->def_cp_lock);
1070 	}
1071 
1072 	/* Inherit new configurations */
1073 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1074 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1075 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1076 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1077 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1078 	    bp->max_stat_ctx)
1079 		goto resource_error;
1080 
1081 	if (BNXT_HAS_RING_GRPS(bp) &&
1082 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1083 		goto resource_error;
1084 
1085 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1086 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1087 		goto resource_error;
1088 
1089 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1090 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1091 
1092 	if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1093 		rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1094 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1095 
1096 	if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1097 		eth_dev->data->mtu =
1098 			eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1099 			RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1100 			BNXT_NUM_VLANS;
1101 		bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1102 	}
1103 	return 0;
1104 
1105 resource_error:
1106 	PMD_DRV_LOG(ERR,
1107 		    "Insufficient resources to support requested config\n");
1108 	PMD_DRV_LOG(ERR,
1109 		    "Num Queues Requested: Tx %d, Rx %d\n",
1110 		    eth_dev->data->nb_tx_queues,
1111 		    eth_dev->data->nb_rx_queues);
1112 	PMD_DRV_LOG(ERR,
1113 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1114 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1115 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1116 	return -ENOSPC;
1117 }
1118 
1119 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1120 {
1121 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1122 
1123 	if (link->link_status)
1124 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1125 			eth_dev->data->port_id,
1126 			(uint32_t)link->link_speed,
1127 			(link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1128 			("full-duplex") : ("half-duplex\n"));
1129 	else
1130 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1131 			eth_dev->data->port_id);
1132 }
1133 
1134 /*
1135  * Determine whether the current configuration requires support for scattered
1136  * receive; return 1 if scattered receive is required and 0 if not.
1137  */
1138 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1139 {
1140 	uint16_t buf_size;
1141 	int i;
1142 
1143 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1144 		return 1;
1145 
1146 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1147 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1148 
1149 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1150 				      RTE_PKTMBUF_HEADROOM);
1151 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1152 			return 1;
1153 	}
1154 	return 0;
1155 }
1156 
1157 static eth_rx_burst_t
1158 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1159 {
1160 	struct bnxt *bp = eth_dev->data->dev_private;
1161 
1162 	/* Disable vector mode RX for Stingray2 for now */
1163 	if (BNXT_CHIP_SR2(bp)) {
1164 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1165 		return bnxt_recv_pkts;
1166 	}
1167 
1168 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1169 #ifndef RTE_LIBRTE_IEEE1588
1170 	/*
1171 	 * Vector mode receive can be enabled only if scatter rx is not
1172 	 * in use and rx offloads are limited to VLAN stripping and
1173 	 * CRC stripping.
1174 	 */
1175 	if (!eth_dev->data->scattered_rx &&
1176 	    !(eth_dev->data->dev_conf.rxmode.offloads &
1177 	      ~(DEV_RX_OFFLOAD_VLAN_STRIP |
1178 		DEV_RX_OFFLOAD_KEEP_CRC |
1179 		DEV_RX_OFFLOAD_JUMBO_FRAME |
1180 		DEV_RX_OFFLOAD_IPV4_CKSUM |
1181 		DEV_RX_OFFLOAD_UDP_CKSUM |
1182 		DEV_RX_OFFLOAD_TCP_CKSUM |
1183 		DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1184 		DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1185 		DEV_RX_OFFLOAD_RSS_HASH |
1186 		DEV_RX_OFFLOAD_VLAN_FILTER)) &&
1187 	    !BNXT_TRUFLOW_EN(bp) && BNXT_NUM_ASYNC_CPR(bp) &&
1188 	    rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1189 		PMD_DRV_LOG(INFO, "Using vector mode receive for port %d\n",
1190 			    eth_dev->data->port_id);
1191 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1192 		return bnxt_recv_pkts_vec;
1193 	}
1194 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1195 		    eth_dev->data->port_id);
1196 	PMD_DRV_LOG(INFO,
1197 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1198 		    eth_dev->data->port_id,
1199 		    eth_dev->data->scattered_rx,
1200 		    eth_dev->data->dev_conf.rxmode.offloads);
1201 #endif
1202 #endif
1203 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1204 	return bnxt_recv_pkts;
1205 }
1206 
1207 static eth_tx_burst_t
1208 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1209 {
1210 	struct bnxt *bp = eth_dev->data->dev_private;
1211 
1212 	/* Disable vector mode TX for Stingray2 for now */
1213 	if (BNXT_CHIP_SR2(bp))
1214 		return bnxt_xmit_pkts;
1215 
1216 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1217 #ifndef RTE_LIBRTE_IEEE1588
1218 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1219 
1220 	/*
1221 	 * Vector mode transmit can be enabled only if not using scatter rx
1222 	 * or tx offloads.
1223 	 */
1224 	if (!eth_dev->data->scattered_rx &&
1225 	    !(offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) &&
1226 	    !BNXT_TRUFLOW_EN(bp) &&
1227 	    rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1228 		PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
1229 			    eth_dev->data->port_id);
1230 		return bnxt_xmit_pkts_vec;
1231 	}
1232 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1233 		    eth_dev->data->port_id);
1234 	PMD_DRV_LOG(INFO,
1235 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1236 		    eth_dev->data->port_id,
1237 		    eth_dev->data->scattered_rx,
1238 		    offloads);
1239 #endif
1240 #endif
1241 	return bnxt_xmit_pkts;
1242 }
1243 
1244 static int bnxt_handle_if_change_status(struct bnxt *bp)
1245 {
1246 	int rc;
1247 
1248 	/* Since fw has undergone a reset and lost all contexts,
1249 	 * set fatal flag to not issue hwrm during cleanup
1250 	 */
1251 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1252 	bnxt_uninit_resources(bp, true);
1253 
1254 	/* clear fatal flag so that re-init happens */
1255 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1256 	rc = bnxt_init_resources(bp, true);
1257 
1258 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1259 
1260 	return rc;
1261 }
1262 
1263 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1264 {
1265 	struct bnxt *bp = eth_dev->data->dev_private;
1266 	int rc = 0;
1267 
1268 	if (!bp->link_info->link_up)
1269 		rc = bnxt_set_hwrm_link_config(bp, true);
1270 	if (!rc)
1271 		eth_dev->data->dev_link.link_status = 1;
1272 
1273 	bnxt_print_link_info(eth_dev);
1274 	return rc;
1275 }
1276 
1277 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1278 {
1279 	struct bnxt *bp = eth_dev->data->dev_private;
1280 
1281 	eth_dev->data->dev_link.link_status = 0;
1282 	bnxt_set_hwrm_link_config(bp, false);
1283 	bp->link_info->link_up = 0;
1284 
1285 	return 0;
1286 }
1287 
1288 static void bnxt_free_switch_domain(struct bnxt *bp)
1289 {
1290 	int rc = 0;
1291 
1292 	if (bp->switch_domain_id) {
1293 		rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1294 		if (rc)
1295 			PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1296 				    bp->switch_domain_id, rc);
1297 	}
1298 }
1299 
1300 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1301 {
1302 	struct bnxt *bp = eth_dev->data->dev_private;
1303 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1304 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1305 	struct rte_eth_link link;
1306 	int ret;
1307 
1308 	eth_dev->data->dev_started = 0;
1309 	eth_dev->data->scattered_rx = 0;
1310 
1311 	/* Prevent crashes when queues are still in use */
1312 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1313 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1314 
1315 	bnxt_disable_int(bp);
1316 
1317 	/* disable uio/vfio intr/eventfd mapping */
1318 	rte_intr_disable(intr_handle);
1319 
1320 	/* Stop the child representors for this device */
1321 	ret = bnxt_rep_stop_all(bp);
1322 	if (ret != 0)
1323 		return ret;
1324 
1325 	/* delete the bnxt ULP port details */
1326 	bnxt_ulp_port_deinit(bp);
1327 
1328 	bnxt_cancel_fw_health_check(bp);
1329 
1330 	/* Do not bring link down during reset recovery */
1331 	if (!is_bnxt_in_error(bp)) {
1332 		bnxt_dev_set_link_down_op(eth_dev);
1333 		/* Wait for link to be reset */
1334 		if (BNXT_SINGLE_PF(bp))
1335 			rte_delay_ms(500);
1336 		/* clear the recorded link status */
1337 		memset(&link, 0, sizeof(link));
1338 		rte_eth_linkstatus_set(eth_dev, &link);
1339 	}
1340 
1341 	/* Clean queue intr-vector mapping */
1342 	rte_intr_efd_disable(intr_handle);
1343 	if (intr_handle->intr_vec != NULL) {
1344 		rte_free(intr_handle->intr_vec);
1345 		intr_handle->intr_vec = NULL;
1346 	}
1347 
1348 	bnxt_hwrm_port_clr_stats(bp);
1349 	bnxt_free_tx_mbufs(bp);
1350 	bnxt_free_rx_mbufs(bp);
1351 	/* Process any remaining notifications in default completion queue */
1352 	bnxt_int_handler(eth_dev);
1353 	bnxt_shutdown_nic(bp);
1354 	bnxt_hwrm_if_change(bp, false);
1355 
1356 	rte_free(bp->mark_table);
1357 	bp->mark_table = NULL;
1358 
1359 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1360 	bp->rx_cosq_cnt = 0;
1361 	/* All filters are deleted on a port stop. */
1362 	if (BNXT_FLOW_XSTATS_EN(bp))
1363 		bp->flow_stat->flow_count = 0;
1364 
1365 	return 0;
1366 }
1367 
1368 /* Unload the driver, release resources */
1369 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1370 {
1371 	struct bnxt *bp = eth_dev->data->dev_private;
1372 
1373 	pthread_mutex_lock(&bp->err_recovery_lock);
1374 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1375 		PMD_DRV_LOG(ERR,
1376 			    "Adapter recovering from error..Please retry\n");
1377 		pthread_mutex_unlock(&bp->err_recovery_lock);
1378 		return -EAGAIN;
1379 	}
1380 	pthread_mutex_unlock(&bp->err_recovery_lock);
1381 
1382 	return bnxt_dev_stop(eth_dev);
1383 }
1384 
1385 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1386 {
1387 	struct bnxt *bp = eth_dev->data->dev_private;
1388 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1389 	int vlan_mask = 0;
1390 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1391 
1392 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1393 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1394 		return -EINVAL;
1395 	}
1396 
1397 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1398 		PMD_DRV_LOG(ERR,
1399 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1400 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1401 
1402 	do {
1403 		rc = bnxt_hwrm_if_change(bp, true);
1404 		if (rc == 0 || rc != -EAGAIN)
1405 			break;
1406 
1407 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1408 	} while (retry_cnt--);
1409 
1410 	if (rc)
1411 		return rc;
1412 
1413 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1414 		rc = bnxt_handle_if_change_status(bp);
1415 		if (rc)
1416 			return rc;
1417 	}
1418 
1419 	bnxt_enable_int(bp);
1420 
1421 	rc = bnxt_start_nic(bp);
1422 	if (rc)
1423 		goto error;
1424 
1425 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1426 	eth_dev->data->dev_started = 1;
1427 
1428 	bnxt_link_update_op(eth_dev, 1);
1429 
1430 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1431 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1432 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1433 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1434 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1435 	if (rc)
1436 		goto error;
1437 
1438 	/* Initialize bnxt ULP port details */
1439 	rc = bnxt_ulp_port_init(bp);
1440 	if (rc)
1441 		goto error;
1442 
1443 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1444 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1445 
1446 	bnxt_schedule_fw_health_check(bp);
1447 
1448 	return 0;
1449 
1450 error:
1451 	bnxt_dev_stop(eth_dev);
1452 	return rc;
1453 }
1454 
1455 static void
1456 bnxt_uninit_locks(struct bnxt *bp)
1457 {
1458 	pthread_mutex_destroy(&bp->flow_lock);
1459 	pthread_mutex_destroy(&bp->def_cp_lock);
1460 	pthread_mutex_destroy(&bp->health_check_lock);
1461 	pthread_mutex_destroy(&bp->err_recovery_lock);
1462 	if (bp->rep_info) {
1463 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1464 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1465 	}
1466 }
1467 
1468 static void bnxt_drv_uninit(struct bnxt *bp)
1469 {
1470 	bnxt_free_switch_domain(bp);
1471 	bnxt_free_leds_info(bp);
1472 	bnxt_free_cos_queues(bp);
1473 	bnxt_free_link_info(bp);
1474 	bnxt_free_pf_info(bp);
1475 	bnxt_free_parent_info(bp);
1476 	bnxt_uninit_locks(bp);
1477 
1478 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1479 	bp->tx_mem_zone = NULL;
1480 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1481 	bp->rx_mem_zone = NULL;
1482 
1483 	bnxt_hwrm_free_vf_info(bp);
1484 
1485 	rte_free(bp->grp_info);
1486 	bp->grp_info = NULL;
1487 }
1488 
1489 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1490 {
1491 	struct bnxt *bp = eth_dev->data->dev_private;
1492 	int ret = 0;
1493 
1494 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1495 		return 0;
1496 
1497 	pthread_mutex_lock(&bp->err_recovery_lock);
1498 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1499 		PMD_DRV_LOG(ERR,
1500 			    "Adapter recovering from error...Please retry\n");
1501 		pthread_mutex_unlock(&bp->err_recovery_lock);
1502 		return -EAGAIN;
1503 	}
1504 	pthread_mutex_unlock(&bp->err_recovery_lock);
1505 
1506 	/* cancel the recovery handler before remove dev */
1507 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1508 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1509 	bnxt_cancel_fc_thread(bp);
1510 
1511 	if (eth_dev->data->dev_started)
1512 		ret = bnxt_dev_stop(eth_dev);
1513 
1514 	bnxt_uninit_resources(bp, false);
1515 
1516 	bnxt_drv_uninit(bp);
1517 
1518 	return ret;
1519 }
1520 
1521 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1522 				    uint32_t index)
1523 {
1524 	struct bnxt *bp = eth_dev->data->dev_private;
1525 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1526 	struct bnxt_vnic_info *vnic;
1527 	struct bnxt_filter_info *filter, *temp_filter;
1528 	uint32_t i;
1529 
1530 	if (is_bnxt_in_error(bp))
1531 		return;
1532 
1533 	/*
1534 	 * Loop through all VNICs from the specified filter flow pools to
1535 	 * remove the corresponding MAC addr filter
1536 	 */
1537 	for (i = 0; i < bp->nr_vnics; i++) {
1538 		if (!(pool_mask & (1ULL << i)))
1539 			continue;
1540 
1541 		vnic = &bp->vnic_info[i];
1542 		filter = STAILQ_FIRST(&vnic->filter);
1543 		while (filter) {
1544 			temp_filter = STAILQ_NEXT(filter, next);
1545 			if (filter->mac_index == index) {
1546 				STAILQ_REMOVE(&vnic->filter, filter,
1547 						bnxt_filter_info, next);
1548 				bnxt_hwrm_clear_l2_filter(bp, filter);
1549 				bnxt_free_filter(bp, filter);
1550 			}
1551 			filter = temp_filter;
1552 		}
1553 	}
1554 }
1555 
1556 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1557 			       struct rte_ether_addr *mac_addr, uint32_t index,
1558 			       uint32_t pool)
1559 {
1560 	struct bnxt_filter_info *filter;
1561 	int rc = 0;
1562 
1563 	/* Attach requested MAC address to the new l2_filter */
1564 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1565 		if (filter->mac_index == index) {
1566 			PMD_DRV_LOG(DEBUG,
1567 				    "MAC addr already existed for pool %d\n",
1568 				    pool);
1569 			return 0;
1570 		}
1571 	}
1572 
1573 	filter = bnxt_alloc_filter(bp);
1574 	if (!filter) {
1575 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1576 		return -ENODEV;
1577 	}
1578 
1579 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1580 	 * if the MAC that's been programmed now is a different one, then,
1581 	 * copy that addr to filter->l2_addr
1582 	 */
1583 	if (mac_addr)
1584 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1585 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1586 
1587 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1588 	if (!rc) {
1589 		filter->mac_index = index;
1590 		if (filter->mac_index == 0)
1591 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1592 		else
1593 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1594 	} else {
1595 		bnxt_free_filter(bp, filter);
1596 	}
1597 
1598 	return rc;
1599 }
1600 
1601 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1602 				struct rte_ether_addr *mac_addr,
1603 				uint32_t index, uint32_t pool)
1604 {
1605 	struct bnxt *bp = eth_dev->data->dev_private;
1606 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1607 	int rc = 0;
1608 
1609 	rc = is_bnxt_in_error(bp);
1610 	if (rc)
1611 		return rc;
1612 
1613 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1614 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1615 		return -ENOTSUP;
1616 	}
1617 
1618 	if (!vnic) {
1619 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1620 		return -EINVAL;
1621 	}
1622 
1623 	/* Filter settings will get applied when port is started */
1624 	if (!eth_dev->data->dev_started)
1625 		return 0;
1626 
1627 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1628 
1629 	return rc;
1630 }
1631 
1632 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1633 {
1634 	int rc = 0;
1635 	struct bnxt *bp = eth_dev->data->dev_private;
1636 	struct rte_eth_link new;
1637 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1638 			BNXT_MIN_LINK_WAIT_CNT;
1639 
1640 	rc = is_bnxt_in_error(bp);
1641 	if (rc)
1642 		return rc;
1643 
1644 	memset(&new, 0, sizeof(new));
1645 	do {
1646 		/* Retrieve link info from hardware */
1647 		rc = bnxt_get_hwrm_link_config(bp, &new);
1648 		if (rc) {
1649 			new.link_speed = ETH_LINK_SPEED_100M;
1650 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1651 			PMD_DRV_LOG(ERR,
1652 				"Failed to retrieve link rc = 0x%x!\n", rc);
1653 			goto out;
1654 		}
1655 
1656 		if (!wait_to_complete || new.link_status)
1657 			break;
1658 
1659 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1660 	} while (cnt--);
1661 
1662 	/* Only single function PF can bring phy down.
1663 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1664 	 */
1665 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1666 		memset(&new, 0, sizeof(new));
1667 
1668 out:
1669 	/* Timed out or success */
1670 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1671 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1672 		rte_eth_linkstatus_set(eth_dev, &new);
1673 
1674 		rte_eth_dev_callback_process(eth_dev,
1675 					     RTE_ETH_EVENT_INTR_LSC,
1676 					     NULL);
1677 
1678 		bnxt_print_link_info(eth_dev);
1679 	}
1680 
1681 	return rc;
1682 }
1683 
1684 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1685 {
1686 	struct bnxt *bp = eth_dev->data->dev_private;
1687 	struct bnxt_vnic_info *vnic;
1688 	uint32_t old_flags;
1689 	int rc;
1690 
1691 	rc = is_bnxt_in_error(bp);
1692 	if (rc)
1693 		return rc;
1694 
1695 	/* Filter settings will get applied when port is started */
1696 	if (!eth_dev->data->dev_started)
1697 		return 0;
1698 
1699 	if (bp->vnic_info == NULL)
1700 		return 0;
1701 
1702 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1703 
1704 	old_flags = vnic->flags;
1705 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1706 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1707 	if (rc != 0)
1708 		vnic->flags = old_flags;
1709 
1710 	return rc;
1711 }
1712 
1713 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1714 {
1715 	struct bnxt *bp = eth_dev->data->dev_private;
1716 	struct bnxt_vnic_info *vnic;
1717 	uint32_t old_flags;
1718 	int rc;
1719 
1720 	rc = is_bnxt_in_error(bp);
1721 	if (rc)
1722 		return rc;
1723 
1724 	/* Filter settings will get applied when port is started */
1725 	if (!eth_dev->data->dev_started)
1726 		return 0;
1727 
1728 	if (bp->vnic_info == NULL)
1729 		return 0;
1730 
1731 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1732 
1733 	old_flags = vnic->flags;
1734 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1735 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1736 	if (rc != 0)
1737 		vnic->flags = old_flags;
1738 
1739 	return rc;
1740 }
1741 
1742 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1743 {
1744 	struct bnxt *bp = eth_dev->data->dev_private;
1745 	struct bnxt_vnic_info *vnic;
1746 	uint32_t old_flags;
1747 	int rc;
1748 
1749 	rc = is_bnxt_in_error(bp);
1750 	if (rc)
1751 		return rc;
1752 
1753 	/* Filter settings will get applied when port is started */
1754 	if (!eth_dev->data->dev_started)
1755 		return 0;
1756 
1757 	if (bp->vnic_info == NULL)
1758 		return 0;
1759 
1760 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1761 
1762 	old_flags = vnic->flags;
1763 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1764 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1765 	if (rc != 0)
1766 		vnic->flags = old_flags;
1767 
1768 	return rc;
1769 }
1770 
1771 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1772 {
1773 	struct bnxt *bp = eth_dev->data->dev_private;
1774 	struct bnxt_vnic_info *vnic;
1775 	uint32_t old_flags;
1776 	int rc;
1777 
1778 	rc = is_bnxt_in_error(bp);
1779 	if (rc)
1780 		return rc;
1781 
1782 	/* Filter settings will get applied when port is started */
1783 	if (!eth_dev->data->dev_started)
1784 		return 0;
1785 
1786 	if (bp->vnic_info == NULL)
1787 		return 0;
1788 
1789 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1790 
1791 	old_flags = vnic->flags;
1792 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1793 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1794 	if (rc != 0)
1795 		vnic->flags = old_flags;
1796 
1797 	return rc;
1798 }
1799 
1800 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1801 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1802 {
1803 	if (qid >= bp->rx_nr_rings)
1804 		return NULL;
1805 
1806 	return bp->eth_dev->data->rx_queues[qid];
1807 }
1808 
1809 /* Return rxq corresponding to a given rss table ring/group ID. */
1810 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1811 {
1812 	struct bnxt_rx_queue *rxq;
1813 	unsigned int i;
1814 
1815 	if (!BNXT_HAS_RING_GRPS(bp)) {
1816 		for (i = 0; i < bp->rx_nr_rings; i++) {
1817 			rxq = bp->eth_dev->data->rx_queues[i];
1818 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1819 				return rxq->index;
1820 		}
1821 	} else {
1822 		for (i = 0; i < bp->rx_nr_rings; i++) {
1823 			if (bp->grp_info[i].fw_grp_id == fwr)
1824 				return i;
1825 		}
1826 	}
1827 
1828 	return INVALID_HW_RING_ID;
1829 }
1830 
1831 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1832 			    struct rte_eth_rss_reta_entry64 *reta_conf,
1833 			    uint16_t reta_size)
1834 {
1835 	struct bnxt *bp = eth_dev->data->dev_private;
1836 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1837 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1838 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1839 	uint16_t idx, sft;
1840 	int i, rc;
1841 
1842 	rc = is_bnxt_in_error(bp);
1843 	if (rc)
1844 		return rc;
1845 
1846 	if (!vnic->rss_table)
1847 		return -EINVAL;
1848 
1849 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
1850 		return -EINVAL;
1851 
1852 	if (reta_size != tbl_size) {
1853 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1854 			"(%d) must equal the size supported by the hardware "
1855 			"(%d)\n", reta_size, tbl_size);
1856 		return -EINVAL;
1857 	}
1858 
1859 	for (i = 0; i < reta_size; i++) {
1860 		struct bnxt_rx_queue *rxq;
1861 
1862 		idx = i / RTE_RETA_GROUP_SIZE;
1863 		sft = i % RTE_RETA_GROUP_SIZE;
1864 
1865 		if (!(reta_conf[idx].mask & (1ULL << sft)))
1866 			continue;
1867 
1868 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
1869 		if (!rxq) {
1870 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
1871 			return -EINVAL;
1872 		}
1873 
1874 		if (BNXT_CHIP_P5(bp)) {
1875 			vnic->rss_table[i * 2] =
1876 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
1877 			vnic->rss_table[i * 2 + 1] =
1878 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
1879 		} else {
1880 			vnic->rss_table[i] =
1881 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
1882 		}
1883 	}
1884 
1885 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1886 	return rc;
1887 }
1888 
1889 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
1890 			      struct rte_eth_rss_reta_entry64 *reta_conf,
1891 			      uint16_t reta_size)
1892 {
1893 	struct bnxt *bp = eth_dev->data->dev_private;
1894 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1895 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1896 	uint16_t idx, sft, i;
1897 	int rc;
1898 
1899 	rc = is_bnxt_in_error(bp);
1900 	if (rc)
1901 		return rc;
1902 
1903 	/* Retrieve from the default VNIC */
1904 	if (!vnic)
1905 		return -EINVAL;
1906 	if (!vnic->rss_table)
1907 		return -EINVAL;
1908 
1909 	if (reta_size != tbl_size) {
1910 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1911 			"(%d) must equal the size supported by the hardware "
1912 			"(%d)\n", reta_size, tbl_size);
1913 		return -EINVAL;
1914 	}
1915 
1916 	for (idx = 0, i = 0; i < reta_size; i++) {
1917 		idx = i / RTE_RETA_GROUP_SIZE;
1918 		sft = i % RTE_RETA_GROUP_SIZE;
1919 
1920 		if (reta_conf[idx].mask & (1ULL << sft)) {
1921 			uint16_t qid;
1922 
1923 			if (BNXT_CHIP_P5(bp))
1924 				qid = bnxt_rss_to_qid(bp,
1925 						      vnic->rss_table[i * 2]);
1926 			else
1927 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
1928 
1929 			if (qid == INVALID_HW_RING_ID) {
1930 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
1931 				return -EINVAL;
1932 			}
1933 			reta_conf[idx].reta[sft] = qid;
1934 		}
1935 	}
1936 
1937 	return 0;
1938 }
1939 
1940 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
1941 				   struct rte_eth_rss_conf *rss_conf)
1942 {
1943 	struct bnxt *bp = eth_dev->data->dev_private;
1944 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1945 	struct bnxt_vnic_info *vnic;
1946 	int rc;
1947 
1948 	rc = is_bnxt_in_error(bp);
1949 	if (rc)
1950 		return rc;
1951 
1952 	/*
1953 	 * If RSS enablement were different than dev_configure,
1954 	 * then return -EINVAL
1955 	 */
1956 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
1957 		if (!rss_conf->rss_hf)
1958 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
1959 	} else {
1960 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
1961 			return -EINVAL;
1962 	}
1963 
1964 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
1965 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
1966 	       rss_conf,
1967 	       sizeof(*rss_conf));
1968 
1969 	/* Update the default RSS VNIC(s) */
1970 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1971 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
1972 	vnic->hash_mode =
1973 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
1974 					    ETH_RSS_LEVEL(rss_conf->rss_hf));
1975 
1976 	/*
1977 	 * If hashkey is not specified, use the previously configured
1978 	 * hashkey
1979 	 */
1980 	if (!rss_conf->rss_key)
1981 		goto rss_config;
1982 
1983 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
1984 		PMD_DRV_LOG(ERR,
1985 			    "Invalid hashkey length, should be 16 bytes\n");
1986 		return -EINVAL;
1987 	}
1988 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
1989 
1990 rss_config:
1991 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1992 	return rc;
1993 }
1994 
1995 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
1996 				     struct rte_eth_rss_conf *rss_conf)
1997 {
1998 	struct bnxt *bp = eth_dev->data->dev_private;
1999 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2000 	int len, rc;
2001 	uint32_t hash_types;
2002 
2003 	rc = is_bnxt_in_error(bp);
2004 	if (rc)
2005 		return rc;
2006 
2007 	/* RSS configuration is the same for all VNICs */
2008 	if (vnic && vnic->rss_hash_key) {
2009 		if (rss_conf->rss_key) {
2010 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2011 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2012 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2013 		}
2014 
2015 		hash_types = vnic->hash_type;
2016 		rss_conf->rss_hf = 0;
2017 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2018 			rss_conf->rss_hf |= ETH_RSS_IPV4;
2019 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2020 		}
2021 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2022 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2023 			hash_types &=
2024 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2025 		}
2026 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2027 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2028 			hash_types &=
2029 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2030 		}
2031 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2032 			rss_conf->rss_hf |= ETH_RSS_IPV6;
2033 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2034 		}
2035 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2036 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2037 			hash_types &=
2038 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2039 		}
2040 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2041 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2042 			hash_types &=
2043 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2044 		}
2045 
2046 		rss_conf->rss_hf |=
2047 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2048 
2049 		if (hash_types) {
2050 			PMD_DRV_LOG(ERR,
2051 				"Unknown RSS config from firmware (%08x), RSS disabled",
2052 				vnic->hash_type);
2053 			return -ENOTSUP;
2054 		}
2055 	} else {
2056 		rss_conf->rss_hf = 0;
2057 	}
2058 	return 0;
2059 }
2060 
2061 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2062 			       struct rte_eth_fc_conf *fc_conf)
2063 {
2064 	struct bnxt *bp = dev->data->dev_private;
2065 	struct rte_eth_link link_info;
2066 	int rc;
2067 
2068 	rc = is_bnxt_in_error(bp);
2069 	if (rc)
2070 		return rc;
2071 
2072 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2073 	if (rc)
2074 		return rc;
2075 
2076 	memset(fc_conf, 0, sizeof(*fc_conf));
2077 	if (bp->link_info->auto_pause)
2078 		fc_conf->autoneg = 1;
2079 	switch (bp->link_info->pause) {
2080 	case 0:
2081 		fc_conf->mode = RTE_FC_NONE;
2082 		break;
2083 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2084 		fc_conf->mode = RTE_FC_TX_PAUSE;
2085 		break;
2086 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2087 		fc_conf->mode = RTE_FC_RX_PAUSE;
2088 		break;
2089 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2090 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2091 		fc_conf->mode = RTE_FC_FULL;
2092 		break;
2093 	}
2094 	return 0;
2095 }
2096 
2097 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2098 			       struct rte_eth_fc_conf *fc_conf)
2099 {
2100 	struct bnxt *bp = dev->data->dev_private;
2101 	int rc;
2102 
2103 	rc = is_bnxt_in_error(bp);
2104 	if (rc)
2105 		return rc;
2106 
2107 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2108 		PMD_DRV_LOG(ERR, "Flow Control Settings cannot be modified\n");
2109 		return -ENOTSUP;
2110 	}
2111 
2112 	switch (fc_conf->mode) {
2113 	case RTE_FC_NONE:
2114 		bp->link_info->auto_pause = 0;
2115 		bp->link_info->force_pause = 0;
2116 		break;
2117 	case RTE_FC_RX_PAUSE:
2118 		if (fc_conf->autoneg) {
2119 			bp->link_info->auto_pause =
2120 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2121 			bp->link_info->force_pause = 0;
2122 		} else {
2123 			bp->link_info->auto_pause = 0;
2124 			bp->link_info->force_pause =
2125 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2126 		}
2127 		break;
2128 	case RTE_FC_TX_PAUSE:
2129 		if (fc_conf->autoneg) {
2130 			bp->link_info->auto_pause =
2131 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2132 			bp->link_info->force_pause = 0;
2133 		} else {
2134 			bp->link_info->auto_pause = 0;
2135 			bp->link_info->force_pause =
2136 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2137 		}
2138 		break;
2139 	case RTE_FC_FULL:
2140 		if (fc_conf->autoneg) {
2141 			bp->link_info->auto_pause =
2142 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2143 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2144 			bp->link_info->force_pause = 0;
2145 		} else {
2146 			bp->link_info->auto_pause = 0;
2147 			bp->link_info->force_pause =
2148 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2149 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2150 		}
2151 		break;
2152 	}
2153 	return bnxt_set_hwrm_link_config(bp, true);
2154 }
2155 
2156 /* Add UDP tunneling port */
2157 static int
2158 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2159 			 struct rte_eth_udp_tunnel *udp_tunnel)
2160 {
2161 	struct bnxt *bp = eth_dev->data->dev_private;
2162 	uint16_t tunnel_type = 0;
2163 	int rc = 0;
2164 
2165 	rc = is_bnxt_in_error(bp);
2166 	if (rc)
2167 		return rc;
2168 
2169 	switch (udp_tunnel->prot_type) {
2170 	case RTE_TUNNEL_TYPE_VXLAN:
2171 		if (bp->vxlan_port_cnt) {
2172 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2173 				udp_tunnel->udp_port);
2174 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2175 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2176 				return -ENOSPC;
2177 			}
2178 			bp->vxlan_port_cnt++;
2179 			return 0;
2180 		}
2181 		tunnel_type =
2182 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2183 		bp->vxlan_port_cnt++;
2184 		break;
2185 	case RTE_TUNNEL_TYPE_GENEVE:
2186 		if (bp->geneve_port_cnt) {
2187 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2188 				udp_tunnel->udp_port);
2189 			if (bp->geneve_port != udp_tunnel->udp_port) {
2190 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2191 				return -ENOSPC;
2192 			}
2193 			bp->geneve_port_cnt++;
2194 			return 0;
2195 		}
2196 		tunnel_type =
2197 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2198 		bp->geneve_port_cnt++;
2199 		break;
2200 	default:
2201 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2202 		return -ENOTSUP;
2203 	}
2204 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2205 					     tunnel_type);
2206 	return rc;
2207 }
2208 
2209 static int
2210 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2211 			 struct rte_eth_udp_tunnel *udp_tunnel)
2212 {
2213 	struct bnxt *bp = eth_dev->data->dev_private;
2214 	uint16_t tunnel_type = 0;
2215 	uint16_t port = 0;
2216 	int rc = 0;
2217 
2218 	rc = is_bnxt_in_error(bp);
2219 	if (rc)
2220 		return rc;
2221 
2222 	switch (udp_tunnel->prot_type) {
2223 	case RTE_TUNNEL_TYPE_VXLAN:
2224 		if (!bp->vxlan_port_cnt) {
2225 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2226 			return -EINVAL;
2227 		}
2228 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2229 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2230 				udp_tunnel->udp_port, bp->vxlan_port);
2231 			return -EINVAL;
2232 		}
2233 		if (--bp->vxlan_port_cnt)
2234 			return 0;
2235 
2236 		tunnel_type =
2237 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2238 		port = bp->vxlan_fw_dst_port_id;
2239 		break;
2240 	case RTE_TUNNEL_TYPE_GENEVE:
2241 		if (!bp->geneve_port_cnt) {
2242 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2243 			return -EINVAL;
2244 		}
2245 		if (bp->geneve_port != udp_tunnel->udp_port) {
2246 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2247 				udp_tunnel->udp_port, bp->geneve_port);
2248 			return -EINVAL;
2249 		}
2250 		if (--bp->geneve_port_cnt)
2251 			return 0;
2252 
2253 		tunnel_type =
2254 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2255 		port = bp->geneve_fw_dst_port_id;
2256 		break;
2257 	default:
2258 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2259 		return -ENOTSUP;
2260 	}
2261 
2262 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2263 	return rc;
2264 }
2265 
2266 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2267 {
2268 	struct bnxt_filter_info *filter;
2269 	struct bnxt_vnic_info *vnic;
2270 	int rc = 0;
2271 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2272 
2273 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2274 	filter = STAILQ_FIRST(&vnic->filter);
2275 	while (filter) {
2276 		/* Search for this matching MAC+VLAN filter */
2277 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2278 			/* Delete the filter */
2279 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2280 			if (rc)
2281 				return rc;
2282 			STAILQ_REMOVE(&vnic->filter, filter,
2283 				      bnxt_filter_info, next);
2284 			bnxt_free_filter(bp, filter);
2285 			PMD_DRV_LOG(INFO,
2286 				    "Deleted vlan filter for %d\n",
2287 				    vlan_id);
2288 			return 0;
2289 		}
2290 		filter = STAILQ_NEXT(filter, next);
2291 	}
2292 	return -ENOENT;
2293 }
2294 
2295 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2296 {
2297 	struct bnxt_filter_info *filter;
2298 	struct bnxt_vnic_info *vnic;
2299 	int rc = 0;
2300 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2301 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2302 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2303 
2304 	/* Implementation notes on the use of VNIC in this command:
2305 	 *
2306 	 * By default, these filters belong to default vnic for the function.
2307 	 * Once these filters are set up, only destination VNIC can be modified.
2308 	 * If the destination VNIC is not specified in this command,
2309 	 * then the HWRM shall only create an l2 context id.
2310 	 */
2311 
2312 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2313 	filter = STAILQ_FIRST(&vnic->filter);
2314 	/* Check if the VLAN has already been added */
2315 	while (filter) {
2316 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2317 			return -EEXIST;
2318 
2319 		filter = STAILQ_NEXT(filter, next);
2320 	}
2321 
2322 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2323 	 * command to create MAC+VLAN filter with the right flags, enables set.
2324 	 */
2325 	filter = bnxt_alloc_filter(bp);
2326 	if (!filter) {
2327 		PMD_DRV_LOG(ERR,
2328 			    "MAC/VLAN filter alloc failed\n");
2329 		return -ENOMEM;
2330 	}
2331 	/* MAC + VLAN ID filter */
2332 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2333 	 * untagged packets are received
2334 	 *
2335 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2336 	 * packets and only the programmed vlan's packets are received
2337 	 */
2338 	filter->l2_ivlan = vlan_id;
2339 	filter->l2_ivlan_mask = 0x0FFF;
2340 	filter->enables |= en;
2341 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2342 
2343 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2344 	if (rc) {
2345 		/* Free the newly allocated filter as we were
2346 		 * not able to create the filter in hardware.
2347 		 */
2348 		bnxt_free_filter(bp, filter);
2349 		return rc;
2350 	}
2351 
2352 	filter->mac_index = 0;
2353 	/* Add this new filter to the list */
2354 	if (vlan_id == 0)
2355 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2356 	else
2357 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2358 
2359 	PMD_DRV_LOG(INFO,
2360 		    "Added Vlan filter for %d\n", vlan_id);
2361 	return rc;
2362 }
2363 
2364 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2365 		uint16_t vlan_id, int on)
2366 {
2367 	struct bnxt *bp = eth_dev->data->dev_private;
2368 	int rc;
2369 
2370 	rc = is_bnxt_in_error(bp);
2371 	if (rc)
2372 		return rc;
2373 
2374 	if (!eth_dev->data->dev_started) {
2375 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2376 		return -EINVAL;
2377 	}
2378 
2379 	/* These operations apply to ALL existing MAC/VLAN filters */
2380 	if (on)
2381 		return bnxt_add_vlan_filter(bp, vlan_id);
2382 	else
2383 		return bnxt_del_vlan_filter(bp, vlan_id);
2384 }
2385 
2386 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2387 				    struct bnxt_vnic_info *vnic)
2388 {
2389 	struct bnxt_filter_info *filter;
2390 	int rc;
2391 
2392 	filter = STAILQ_FIRST(&vnic->filter);
2393 	while (filter) {
2394 		if (filter->mac_index == 0 &&
2395 		    !memcmp(filter->l2_addr, bp->mac_addr,
2396 			    RTE_ETHER_ADDR_LEN)) {
2397 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2398 			if (!rc) {
2399 				STAILQ_REMOVE(&vnic->filter, filter,
2400 					      bnxt_filter_info, next);
2401 				bnxt_free_filter(bp, filter);
2402 			}
2403 			return rc;
2404 		}
2405 		filter = STAILQ_NEXT(filter, next);
2406 	}
2407 	return 0;
2408 }
2409 
2410 static int
2411 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2412 {
2413 	struct bnxt_vnic_info *vnic;
2414 	unsigned int i;
2415 	int rc;
2416 
2417 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2418 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2419 		/* Remove any VLAN filters programmed */
2420 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2421 			bnxt_del_vlan_filter(bp, i);
2422 
2423 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2424 		if (rc)
2425 			return rc;
2426 	} else {
2427 		/* Default filter will allow packets that match the
2428 		 * dest mac. So, it has to be deleted, otherwise, we
2429 		 * will endup receiving vlan packets for which the
2430 		 * filter is not programmed, when hw-vlan-filter
2431 		 * configuration is ON
2432 		 */
2433 		bnxt_del_dflt_mac_filter(bp, vnic);
2434 		/* This filter will allow only untagged packets */
2435 		bnxt_add_vlan_filter(bp, 0);
2436 	}
2437 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2438 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2439 
2440 	return 0;
2441 }
2442 
2443 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2444 {
2445 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2446 	unsigned int i;
2447 	int rc;
2448 
2449 	/* Destroy vnic filters and vnic */
2450 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2451 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2452 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2453 			bnxt_del_vlan_filter(bp, i);
2454 	}
2455 	bnxt_del_dflt_mac_filter(bp, vnic);
2456 
2457 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2458 	if (rc)
2459 		return rc;
2460 
2461 	rte_free(vnic->fw_grp_ids);
2462 	vnic->fw_grp_ids = NULL;
2463 
2464 	vnic->rx_queue_cnt = 0;
2465 
2466 	return 0;
2467 }
2468 
2469 static int
2470 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2471 {
2472 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2473 	int rc;
2474 
2475 	/* Destroy, recreate and reconfigure the default vnic */
2476 	rc = bnxt_free_one_vnic(bp, 0);
2477 	if (rc)
2478 		return rc;
2479 
2480 	/* default vnic 0 */
2481 	rc = bnxt_setup_one_vnic(bp, 0);
2482 	if (rc)
2483 		return rc;
2484 
2485 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2486 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2487 		rc = bnxt_add_vlan_filter(bp, 0);
2488 		if (rc)
2489 			return rc;
2490 		rc = bnxt_restore_vlan_filters(bp);
2491 		if (rc)
2492 			return rc;
2493 	} else {
2494 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2495 		if (rc)
2496 			return rc;
2497 	}
2498 
2499 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2500 	if (rc)
2501 		return rc;
2502 
2503 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2504 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2505 
2506 	return rc;
2507 }
2508 
2509 static int
2510 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2511 {
2512 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2513 	struct bnxt *bp = dev->data->dev_private;
2514 	int rc;
2515 
2516 	rc = is_bnxt_in_error(bp);
2517 	if (rc)
2518 		return rc;
2519 
2520 	/* Filter settings will get applied when port is started */
2521 	if (!dev->data->dev_started)
2522 		return 0;
2523 
2524 	if (mask & ETH_VLAN_FILTER_MASK) {
2525 		/* Enable or disable VLAN filtering */
2526 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2527 		if (rc)
2528 			return rc;
2529 	}
2530 
2531 	if (mask & ETH_VLAN_STRIP_MASK) {
2532 		/* Enable or disable VLAN stripping */
2533 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2534 		if (rc)
2535 			return rc;
2536 	}
2537 
2538 	if (mask & ETH_VLAN_EXTEND_MASK) {
2539 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2540 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2541 		else
2542 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2543 	}
2544 
2545 	return 0;
2546 }
2547 
2548 static int
2549 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2550 		      uint16_t tpid)
2551 {
2552 	struct bnxt *bp = dev->data->dev_private;
2553 	int qinq = dev->data->dev_conf.rxmode.offloads &
2554 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2555 
2556 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2557 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2558 		PMD_DRV_LOG(ERR,
2559 			    "Unsupported vlan type.");
2560 		return -EINVAL;
2561 	}
2562 	if (!qinq) {
2563 		PMD_DRV_LOG(ERR,
2564 			    "QinQ not enabled. Needs to be ON as we can "
2565 			    "accelerate only outer vlan\n");
2566 		return -EINVAL;
2567 	}
2568 
2569 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2570 		switch (tpid) {
2571 		case RTE_ETHER_TYPE_QINQ:
2572 			bp->outer_tpid_bd =
2573 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2574 				break;
2575 		case RTE_ETHER_TYPE_VLAN:
2576 			bp->outer_tpid_bd =
2577 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2578 				break;
2579 		case RTE_ETHER_TYPE_QINQ1:
2580 			bp->outer_tpid_bd =
2581 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2582 				break;
2583 		case RTE_ETHER_TYPE_QINQ2:
2584 			bp->outer_tpid_bd =
2585 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2586 				break;
2587 		case RTE_ETHER_TYPE_QINQ3:
2588 			bp->outer_tpid_bd =
2589 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2590 				break;
2591 		default:
2592 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2593 			return -EINVAL;
2594 		}
2595 		bp->outer_tpid_bd |= tpid;
2596 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2597 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2598 		PMD_DRV_LOG(ERR,
2599 			    "Can accelerate only outer vlan in QinQ\n");
2600 		return -EINVAL;
2601 	}
2602 
2603 	return 0;
2604 }
2605 
2606 static int
2607 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2608 			     struct rte_ether_addr *addr)
2609 {
2610 	struct bnxt *bp = dev->data->dev_private;
2611 	/* Default Filter is tied to VNIC 0 */
2612 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2613 	int rc;
2614 
2615 	rc = is_bnxt_in_error(bp);
2616 	if (rc)
2617 		return rc;
2618 
2619 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2620 		return -EPERM;
2621 
2622 	if (rte_is_zero_ether_addr(addr))
2623 		return -EINVAL;
2624 
2625 	/* Filter settings will get applied when port is started */
2626 	if (!dev->data->dev_started)
2627 		return 0;
2628 
2629 	/* Check if the requested MAC is already added */
2630 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2631 		return 0;
2632 
2633 	/* Destroy filter and re-create it */
2634 	bnxt_del_dflt_mac_filter(bp, vnic);
2635 
2636 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2637 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2638 		/* This filter will allow only untagged packets */
2639 		rc = bnxt_add_vlan_filter(bp, 0);
2640 	} else {
2641 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2642 	}
2643 
2644 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2645 	return rc;
2646 }
2647 
2648 static int
2649 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2650 			  struct rte_ether_addr *mc_addr_set,
2651 			  uint32_t nb_mc_addr)
2652 {
2653 	struct bnxt *bp = eth_dev->data->dev_private;
2654 	char *mc_addr_list = (char *)mc_addr_set;
2655 	struct bnxt_vnic_info *vnic;
2656 	uint32_t off = 0, i = 0;
2657 	int rc;
2658 
2659 	rc = is_bnxt_in_error(bp);
2660 	if (rc)
2661 		return rc;
2662 
2663 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2664 
2665 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2666 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2667 		goto allmulti;
2668 	}
2669 
2670 	/* TODO Check for Duplicate mcast addresses */
2671 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2672 	for (i = 0; i < nb_mc_addr; i++) {
2673 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2674 			RTE_ETHER_ADDR_LEN);
2675 		off += RTE_ETHER_ADDR_LEN;
2676 	}
2677 
2678 	vnic->mc_addr_cnt = i;
2679 	if (vnic->mc_addr_cnt)
2680 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2681 	else
2682 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2683 
2684 allmulti:
2685 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2686 }
2687 
2688 static int
2689 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2690 {
2691 	struct bnxt *bp = dev->data->dev_private;
2692 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2693 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2694 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2695 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2696 	int ret;
2697 
2698 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2699 			fw_major, fw_minor, fw_updt, fw_rsvd);
2700 
2701 	ret += 1; /* add the size of '\0' */
2702 	if (fw_size < (uint32_t)ret)
2703 		return ret;
2704 	else
2705 		return 0;
2706 }
2707 
2708 static void
2709 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2710 	struct rte_eth_rxq_info *qinfo)
2711 {
2712 	struct bnxt *bp = dev->data->dev_private;
2713 	struct bnxt_rx_queue *rxq;
2714 
2715 	if (is_bnxt_in_error(bp))
2716 		return;
2717 
2718 	rxq = dev->data->rx_queues[queue_id];
2719 
2720 	qinfo->mp = rxq->mb_pool;
2721 	qinfo->scattered_rx = dev->data->scattered_rx;
2722 	qinfo->nb_desc = rxq->nb_rx_desc;
2723 
2724 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2725 	qinfo->conf.rx_drop_en = rxq->drop_en;
2726 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2727 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2728 }
2729 
2730 static void
2731 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2732 	struct rte_eth_txq_info *qinfo)
2733 {
2734 	struct bnxt *bp = dev->data->dev_private;
2735 	struct bnxt_tx_queue *txq;
2736 
2737 	if (is_bnxt_in_error(bp))
2738 		return;
2739 
2740 	txq = dev->data->tx_queues[queue_id];
2741 
2742 	qinfo->nb_desc = txq->nb_tx_desc;
2743 
2744 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2745 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2746 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2747 
2748 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2749 	qinfo->conf.tx_rs_thresh = 0;
2750 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2751 	qinfo->conf.offloads = txq->offloads;
2752 }
2753 
2754 static const struct {
2755 	eth_rx_burst_t pkt_burst;
2756 	const char *info;
2757 } bnxt_rx_burst_info[] = {
2758 	{bnxt_recv_pkts,	"Scalar"},
2759 #if defined(RTE_ARCH_X86)
2760 	{bnxt_recv_pkts_vec,	"Vector SSE"},
2761 #elif defined(RTE_ARCH_ARM64)
2762 	{bnxt_recv_pkts_vec,	"Vector Neon"},
2763 #endif
2764 };
2765 
2766 static int
2767 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2768 		       struct rte_eth_burst_mode *mode)
2769 {
2770 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2771 	size_t i;
2772 
2773 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2774 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2775 			snprintf(mode->info, sizeof(mode->info), "%s",
2776 				 bnxt_rx_burst_info[i].info);
2777 			return 0;
2778 		}
2779 	}
2780 
2781 	return -EINVAL;
2782 }
2783 
2784 static const struct {
2785 	eth_tx_burst_t pkt_burst;
2786 	const char *info;
2787 } bnxt_tx_burst_info[] = {
2788 	{bnxt_xmit_pkts,	"Scalar"},
2789 #if defined(RTE_ARCH_X86)
2790 	{bnxt_xmit_pkts_vec,	"Vector SSE"},
2791 #elif defined(RTE_ARCH_ARM64)
2792 	{bnxt_xmit_pkts_vec,	"Vector Neon"},
2793 #endif
2794 };
2795 
2796 static int
2797 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2798 		       struct rte_eth_burst_mode *mode)
2799 {
2800 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2801 	size_t i;
2802 
2803 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2804 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2805 			snprintf(mode->info, sizeof(mode->info), "%s",
2806 				 bnxt_tx_burst_info[i].info);
2807 			return 0;
2808 		}
2809 	}
2810 
2811 	return -EINVAL;
2812 }
2813 
2814 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2815 {
2816 	struct bnxt *bp = eth_dev->data->dev_private;
2817 	uint32_t new_pkt_size;
2818 	uint32_t rc = 0;
2819 	uint32_t i;
2820 
2821 	rc = is_bnxt_in_error(bp);
2822 	if (rc)
2823 		return rc;
2824 
2825 	/* Exit if receive queues are not configured yet */
2826 	if (!eth_dev->data->nb_rx_queues)
2827 		return rc;
2828 
2829 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2830 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2831 
2832 	/*
2833 	 * Disallow any MTU change that would require scattered receive support
2834 	 * if it is not already enabled.
2835 	 */
2836 	if (eth_dev->data->dev_started &&
2837 	    !eth_dev->data->scattered_rx &&
2838 	    (new_pkt_size >
2839 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2840 		PMD_DRV_LOG(ERR,
2841 			    "MTU change would require scattered rx support. ");
2842 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2843 		return -EINVAL;
2844 	}
2845 
2846 	if (new_mtu > RTE_ETHER_MTU) {
2847 		bp->flags |= BNXT_FLAG_JUMBO;
2848 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
2849 			DEV_RX_OFFLOAD_JUMBO_FRAME;
2850 	} else {
2851 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
2852 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
2853 		bp->flags &= ~BNXT_FLAG_JUMBO;
2854 	}
2855 
2856 	/* Is there a change in mtu setting? */
2857 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2858 		return rc;
2859 
2860 	for (i = 0; i < bp->nr_vnics; i++) {
2861 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2862 		uint16_t size = 0;
2863 
2864 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
2865 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2866 		if (rc)
2867 			break;
2868 
2869 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2870 		size -= RTE_PKTMBUF_HEADROOM;
2871 
2872 		if (size < new_mtu) {
2873 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2874 			if (rc)
2875 				return rc;
2876 		}
2877 	}
2878 
2879 	if (!rc)
2880 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2881 
2882 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2883 
2884 	return rc;
2885 }
2886 
2887 static int
2888 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2889 {
2890 	struct bnxt *bp = dev->data->dev_private;
2891 	uint16_t vlan = bp->vlan;
2892 	int rc;
2893 
2894 	rc = is_bnxt_in_error(bp);
2895 	if (rc)
2896 		return rc;
2897 
2898 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2899 		PMD_DRV_LOG(ERR,
2900 			"PVID cannot be modified for this function\n");
2901 		return -ENOTSUP;
2902 	}
2903 	bp->vlan = on ? pvid : 0;
2904 
2905 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
2906 	if (rc)
2907 		bp->vlan = vlan;
2908 	return rc;
2909 }
2910 
2911 static int
2912 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
2913 {
2914 	struct bnxt *bp = dev->data->dev_private;
2915 	int rc;
2916 
2917 	rc = is_bnxt_in_error(bp);
2918 	if (rc)
2919 		return rc;
2920 
2921 	return bnxt_hwrm_port_led_cfg(bp, true);
2922 }
2923 
2924 static int
2925 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
2926 {
2927 	struct bnxt *bp = dev->data->dev_private;
2928 	int rc;
2929 
2930 	rc = is_bnxt_in_error(bp);
2931 	if (rc)
2932 		return rc;
2933 
2934 	return bnxt_hwrm_port_led_cfg(bp, false);
2935 }
2936 
2937 static uint32_t
2938 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
2939 {
2940 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
2941 	uint32_t desc = 0, raw_cons = 0, cons;
2942 	struct bnxt_cp_ring_info *cpr;
2943 	struct bnxt_rx_queue *rxq;
2944 	struct rx_pkt_cmpl *rxcmp;
2945 	int rc;
2946 
2947 	rc = is_bnxt_in_error(bp);
2948 	if (rc)
2949 		return rc;
2950 
2951 	rxq = dev->data->rx_queues[rx_queue_id];
2952 	cpr = rxq->cp_ring;
2953 	raw_cons = cpr->cp_raw_cons;
2954 
2955 	while (1) {
2956 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
2957 		rte_prefetch0(&cpr->cp_desc_ring[cons]);
2958 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2959 
2960 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct)) {
2961 			break;
2962 		} else {
2963 			raw_cons++;
2964 			desc++;
2965 		}
2966 	}
2967 
2968 	return desc;
2969 }
2970 
2971 static int
2972 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
2973 {
2974 	struct bnxt_rx_queue *rxq = (struct bnxt_rx_queue *)rx_queue;
2975 	struct bnxt_rx_ring_info *rxr;
2976 	struct bnxt_cp_ring_info *cpr;
2977 	struct rte_mbuf *rx_buf;
2978 	struct rx_pkt_cmpl *rxcmp;
2979 	uint32_t cons, cp_cons;
2980 	int rc;
2981 
2982 	if (!rxq)
2983 		return -EINVAL;
2984 
2985 	rc = is_bnxt_in_error(rxq->bp);
2986 	if (rc)
2987 		return rc;
2988 
2989 	cpr = rxq->cp_ring;
2990 	rxr = rxq->rx_ring;
2991 
2992 	if (offset >= rxq->nb_rx_desc)
2993 		return -EINVAL;
2994 
2995 	cons = RING_CMP(cpr->cp_ring_struct, offset);
2996 	cp_cons = cpr->cp_raw_cons;
2997 	rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2998 
2999 	if (cons > cp_cons) {
3000 		if (CMPL_VALID(rxcmp, cpr->valid))
3001 			return RTE_ETH_RX_DESC_DONE;
3002 	} else {
3003 		if (CMPL_VALID(rxcmp, !cpr->valid))
3004 			return RTE_ETH_RX_DESC_DONE;
3005 	}
3006 	rx_buf = rxr->rx_buf_ring[cons];
3007 	if (rx_buf == NULL || rx_buf == &rxq->fake_mbuf)
3008 		return RTE_ETH_RX_DESC_UNAVAIL;
3009 
3010 
3011 	return RTE_ETH_RX_DESC_AVAIL;
3012 }
3013 
3014 static int
3015 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3016 {
3017 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3018 	struct bnxt_tx_ring_info *txr;
3019 	struct bnxt_cp_ring_info *cpr;
3020 	struct bnxt_sw_tx_bd *tx_buf;
3021 	struct tx_pkt_cmpl *txcmp;
3022 	uint32_t cons, cp_cons;
3023 	int rc;
3024 
3025 	if (!txq)
3026 		return -EINVAL;
3027 
3028 	rc = is_bnxt_in_error(txq->bp);
3029 	if (rc)
3030 		return rc;
3031 
3032 	cpr = txq->cp_ring;
3033 	txr = txq->tx_ring;
3034 
3035 	if (offset >= txq->nb_tx_desc)
3036 		return -EINVAL;
3037 
3038 	cons = RING_CMP(cpr->cp_ring_struct, offset);
3039 	txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3040 	cp_cons = cpr->cp_raw_cons;
3041 
3042 	if (cons > cp_cons) {
3043 		if (CMPL_VALID(txcmp, cpr->valid))
3044 			return RTE_ETH_TX_DESC_UNAVAIL;
3045 	} else {
3046 		if (CMPL_VALID(txcmp, !cpr->valid))
3047 			return RTE_ETH_TX_DESC_UNAVAIL;
3048 	}
3049 	tx_buf = &txr->tx_buf_ring[cons];
3050 	if (tx_buf->mbuf == NULL)
3051 		return RTE_ETH_TX_DESC_DONE;
3052 
3053 	return RTE_ETH_TX_DESC_FULL;
3054 }
3055 
3056 int
3057 bnxt_filter_ctrl_op(struct rte_eth_dev *dev,
3058 		    enum rte_filter_type filter_type,
3059 		    enum rte_filter_op filter_op, void *arg)
3060 {
3061 	struct bnxt *bp = dev->data->dev_private;
3062 	int ret = 0;
3063 
3064 	if (!bp)
3065 		return -EIO;
3066 
3067 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3068 		struct bnxt_representor *vfr = dev->data->dev_private;
3069 		bp = vfr->parent_dev->data->dev_private;
3070 		/* parent is deleted while children are still valid */
3071 		if (!bp) {
3072 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error %d:%d\n",
3073 				    dev->data->port_id,
3074 				    filter_type,
3075 				    filter_op);
3076 			return -EIO;
3077 		}
3078 	}
3079 
3080 	ret = is_bnxt_in_error(bp);
3081 	if (ret)
3082 		return ret;
3083 
3084 	switch (filter_type) {
3085 	case RTE_ETH_FILTER_GENERIC:
3086 		if (filter_op != RTE_ETH_FILTER_GET)
3087 			return -EINVAL;
3088 
3089 		/* PMD supports thread-safe flow operations.  rte_flow API
3090 		 * functions can avoid mutex for multi-thread safety.
3091 		 */
3092 		dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3093 
3094 		if (BNXT_TRUFLOW_EN(bp))
3095 			*(const void **)arg = &bnxt_ulp_rte_flow_ops;
3096 		else
3097 			*(const void **)arg = &bnxt_flow_ops;
3098 		break;
3099 	default:
3100 		PMD_DRV_LOG(ERR,
3101 			"Filter type (%d) not supported", filter_type);
3102 		ret = -EINVAL;
3103 		break;
3104 	}
3105 	return ret;
3106 }
3107 
3108 static const uint32_t *
3109 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3110 {
3111 	static const uint32_t ptypes[] = {
3112 		RTE_PTYPE_L2_ETHER_VLAN,
3113 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3114 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3115 		RTE_PTYPE_L4_ICMP,
3116 		RTE_PTYPE_L4_TCP,
3117 		RTE_PTYPE_L4_UDP,
3118 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3119 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3120 		RTE_PTYPE_INNER_L4_ICMP,
3121 		RTE_PTYPE_INNER_L4_TCP,
3122 		RTE_PTYPE_INNER_L4_UDP,
3123 		RTE_PTYPE_UNKNOWN
3124 	};
3125 
3126 	if (!dev->rx_pkt_burst)
3127 		return NULL;
3128 
3129 	return ptypes;
3130 }
3131 
3132 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3133 			 int reg_win)
3134 {
3135 	uint32_t reg_base = *reg_arr & 0xfffff000;
3136 	uint32_t win_off;
3137 	int i;
3138 
3139 	for (i = 0; i < count; i++) {
3140 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3141 			return -ERANGE;
3142 	}
3143 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3144 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3145 	return 0;
3146 }
3147 
3148 static int bnxt_map_ptp_regs(struct bnxt *bp)
3149 {
3150 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3151 	uint32_t *reg_arr;
3152 	int rc, i;
3153 
3154 	reg_arr = ptp->rx_regs;
3155 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3156 	if (rc)
3157 		return rc;
3158 
3159 	reg_arr = ptp->tx_regs;
3160 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3161 	if (rc)
3162 		return rc;
3163 
3164 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3165 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3166 
3167 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3168 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3169 
3170 	return 0;
3171 }
3172 
3173 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3174 {
3175 	rte_write32(0, (uint8_t *)bp->bar0 +
3176 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3177 	rte_write32(0, (uint8_t *)bp->bar0 +
3178 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3179 }
3180 
3181 static uint64_t bnxt_cc_read(struct bnxt *bp)
3182 {
3183 	uint64_t ns;
3184 
3185 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3186 			      BNXT_GRCPF_REG_SYNC_TIME));
3187 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3188 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3189 	return ns;
3190 }
3191 
3192 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3193 {
3194 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3195 	uint32_t fifo;
3196 
3197 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3198 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3199 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3200 		return -EAGAIN;
3201 
3202 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3203 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3204 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3205 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3206 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3207 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3208 
3209 	return 0;
3210 }
3211 
3212 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3213 {
3214 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3215 	struct bnxt_pf_info *pf = bp->pf;
3216 	uint16_t port_id;
3217 	uint32_t fifo;
3218 
3219 	if (!ptp)
3220 		return -ENODEV;
3221 
3222 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3223 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3224 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3225 		return -EAGAIN;
3226 
3227 	port_id = pf->port_id;
3228 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3229 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3230 
3231 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3232 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3233 	if (fifo & BNXT_PTP_RX_FIFO_PENDING) {
3234 /*		bnxt_clr_rx_ts(bp);	  TBD  */
3235 		return -EBUSY;
3236 	}
3237 
3238 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3239 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3240 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3241 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3242 
3243 	return 0;
3244 }
3245 
3246 static int
3247 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3248 {
3249 	uint64_t ns;
3250 	struct bnxt *bp = dev->data->dev_private;
3251 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3252 
3253 	if (!ptp)
3254 		return 0;
3255 
3256 	ns = rte_timespec_to_ns(ts);
3257 	/* Set the timecounters to a new value. */
3258 	ptp->tc.nsec = ns;
3259 
3260 	return 0;
3261 }
3262 
3263 static int
3264 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3265 {
3266 	struct bnxt *bp = dev->data->dev_private;
3267 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3268 	uint64_t ns, systime_cycles = 0;
3269 	int rc = 0;
3270 
3271 	if (!ptp)
3272 		return 0;
3273 
3274 	if (BNXT_CHIP_P5(bp))
3275 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3276 					     &systime_cycles);
3277 	else
3278 		systime_cycles = bnxt_cc_read(bp);
3279 
3280 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3281 	*ts = rte_ns_to_timespec(ns);
3282 
3283 	return rc;
3284 }
3285 static int
3286 bnxt_timesync_enable(struct rte_eth_dev *dev)
3287 {
3288 	struct bnxt *bp = dev->data->dev_private;
3289 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3290 	uint32_t shift = 0;
3291 	int rc;
3292 
3293 	if (!ptp)
3294 		return 0;
3295 
3296 	ptp->rx_filter = 1;
3297 	ptp->tx_tstamp_en = 1;
3298 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3299 
3300 	rc = bnxt_hwrm_ptp_cfg(bp);
3301 	if (rc)
3302 		return rc;
3303 
3304 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3305 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3306 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3307 
3308 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3309 	ptp->tc.cc_shift = shift;
3310 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3311 
3312 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3313 	ptp->rx_tstamp_tc.cc_shift = shift;
3314 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3315 
3316 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3317 	ptp->tx_tstamp_tc.cc_shift = shift;
3318 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3319 
3320 	if (!BNXT_CHIP_P5(bp))
3321 		bnxt_map_ptp_regs(bp);
3322 
3323 	return 0;
3324 }
3325 
3326 static int
3327 bnxt_timesync_disable(struct rte_eth_dev *dev)
3328 {
3329 	struct bnxt *bp = dev->data->dev_private;
3330 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3331 
3332 	if (!ptp)
3333 		return 0;
3334 
3335 	ptp->rx_filter = 0;
3336 	ptp->tx_tstamp_en = 0;
3337 	ptp->rxctl = 0;
3338 
3339 	bnxt_hwrm_ptp_cfg(bp);
3340 
3341 	if (!BNXT_CHIP_P5(bp))
3342 		bnxt_unmap_ptp_regs(bp);
3343 
3344 	return 0;
3345 }
3346 
3347 static int
3348 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3349 				 struct timespec *timestamp,
3350 				 uint32_t flags __rte_unused)
3351 {
3352 	struct bnxt *bp = dev->data->dev_private;
3353 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3354 	uint64_t rx_tstamp_cycles = 0;
3355 	uint64_t ns;
3356 
3357 	if (!ptp)
3358 		return 0;
3359 
3360 	if (BNXT_CHIP_P5(bp))
3361 		rx_tstamp_cycles = ptp->rx_timestamp;
3362 	else
3363 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3364 
3365 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3366 	*timestamp = rte_ns_to_timespec(ns);
3367 	return  0;
3368 }
3369 
3370 static int
3371 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3372 				 struct timespec *timestamp)
3373 {
3374 	struct bnxt *bp = dev->data->dev_private;
3375 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3376 	uint64_t tx_tstamp_cycles = 0;
3377 	uint64_t ns;
3378 	int rc = 0;
3379 
3380 	if (!ptp)
3381 		return 0;
3382 
3383 	if (BNXT_CHIP_P5(bp))
3384 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3385 					     &tx_tstamp_cycles);
3386 	else
3387 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3388 
3389 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3390 	*timestamp = rte_ns_to_timespec(ns);
3391 
3392 	return rc;
3393 }
3394 
3395 static int
3396 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3397 {
3398 	struct bnxt *bp = dev->data->dev_private;
3399 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3400 
3401 	if (!ptp)
3402 		return 0;
3403 
3404 	ptp->tc.nsec += delta;
3405 
3406 	return 0;
3407 }
3408 
3409 static int
3410 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3411 {
3412 	struct bnxt *bp = dev->data->dev_private;
3413 	int rc;
3414 	uint32_t dir_entries;
3415 	uint32_t entry_length;
3416 
3417 	rc = is_bnxt_in_error(bp);
3418 	if (rc)
3419 		return rc;
3420 
3421 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3422 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3423 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3424 
3425 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3426 	if (rc != 0)
3427 		return rc;
3428 
3429 	return dir_entries * entry_length;
3430 }
3431 
3432 static int
3433 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3434 		struct rte_dev_eeprom_info *in_eeprom)
3435 {
3436 	struct bnxt *bp = dev->data->dev_private;
3437 	uint32_t index;
3438 	uint32_t offset;
3439 	int rc;
3440 
3441 	rc = is_bnxt_in_error(bp);
3442 	if (rc)
3443 		return rc;
3444 
3445 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3446 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3447 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3448 		    in_eeprom->offset, in_eeprom->length);
3449 
3450 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3451 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3452 						in_eeprom->data);
3453 
3454 	index = in_eeprom->offset >> 24;
3455 	offset = in_eeprom->offset & 0xffffff;
3456 
3457 	if (index != 0)
3458 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3459 					   in_eeprom->length, in_eeprom->data);
3460 
3461 	return 0;
3462 }
3463 
3464 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3465 {
3466 	switch (dir_type) {
3467 	case BNX_DIR_TYPE_CHIMP_PATCH:
3468 	case BNX_DIR_TYPE_BOOTCODE:
3469 	case BNX_DIR_TYPE_BOOTCODE_2:
3470 	case BNX_DIR_TYPE_APE_FW:
3471 	case BNX_DIR_TYPE_APE_PATCH:
3472 	case BNX_DIR_TYPE_KONG_FW:
3473 	case BNX_DIR_TYPE_KONG_PATCH:
3474 	case BNX_DIR_TYPE_BONO_FW:
3475 	case BNX_DIR_TYPE_BONO_PATCH:
3476 		/* FALLTHROUGH */
3477 		return true;
3478 	}
3479 
3480 	return false;
3481 }
3482 
3483 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3484 {
3485 	switch (dir_type) {
3486 	case BNX_DIR_TYPE_AVS:
3487 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3488 	case BNX_DIR_TYPE_PCIE:
3489 	case BNX_DIR_TYPE_TSCF_UCODE:
3490 	case BNX_DIR_TYPE_EXT_PHY:
3491 	case BNX_DIR_TYPE_CCM:
3492 	case BNX_DIR_TYPE_ISCSI_BOOT:
3493 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3494 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3495 		/* FALLTHROUGH */
3496 		return true;
3497 	}
3498 
3499 	return false;
3500 }
3501 
3502 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3503 {
3504 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3505 		bnxt_dir_type_is_other_exec_format(dir_type);
3506 }
3507 
3508 static int
3509 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3510 		struct rte_dev_eeprom_info *in_eeprom)
3511 {
3512 	struct bnxt *bp = dev->data->dev_private;
3513 	uint8_t index, dir_op;
3514 	uint16_t type, ext, ordinal, attr;
3515 	int rc;
3516 
3517 	rc = is_bnxt_in_error(bp);
3518 	if (rc)
3519 		return rc;
3520 
3521 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3522 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3523 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3524 		    in_eeprom->offset, in_eeprom->length);
3525 
3526 	if (!BNXT_PF(bp)) {
3527 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3528 		return -EINVAL;
3529 	}
3530 
3531 	type = in_eeprom->magic >> 16;
3532 
3533 	if (type == 0xffff) { /* special value for directory operations */
3534 		index = in_eeprom->magic & 0xff;
3535 		dir_op = in_eeprom->magic >> 8;
3536 		if (index == 0)
3537 			return -EINVAL;
3538 		switch (dir_op) {
3539 		case 0x0e: /* erase */
3540 			if (in_eeprom->offset != ~in_eeprom->magic)
3541 				return -EINVAL;
3542 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3543 		default:
3544 			return -EINVAL;
3545 		}
3546 	}
3547 
3548 	/* Create or re-write an NVM item: */
3549 	if (bnxt_dir_type_is_executable(type) == true)
3550 		return -EOPNOTSUPP;
3551 	ext = in_eeprom->magic & 0xffff;
3552 	ordinal = in_eeprom->offset >> 16;
3553 	attr = in_eeprom->offset & 0xffff;
3554 
3555 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3556 				     in_eeprom->data, in_eeprom->length);
3557 }
3558 
3559 /*
3560  * Initialization
3561  */
3562 
3563 static const struct eth_dev_ops bnxt_dev_ops = {
3564 	.dev_infos_get = bnxt_dev_info_get_op,
3565 	.dev_close = bnxt_dev_close_op,
3566 	.dev_configure = bnxt_dev_configure_op,
3567 	.dev_start = bnxt_dev_start_op,
3568 	.dev_stop = bnxt_dev_stop_op,
3569 	.dev_set_link_up = bnxt_dev_set_link_up_op,
3570 	.dev_set_link_down = bnxt_dev_set_link_down_op,
3571 	.stats_get = bnxt_stats_get_op,
3572 	.stats_reset = bnxt_stats_reset_op,
3573 	.rx_queue_setup = bnxt_rx_queue_setup_op,
3574 	.rx_queue_release = bnxt_rx_queue_release_op,
3575 	.tx_queue_setup = bnxt_tx_queue_setup_op,
3576 	.tx_queue_release = bnxt_tx_queue_release_op,
3577 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
3578 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
3579 	.reta_update = bnxt_reta_update_op,
3580 	.reta_query = bnxt_reta_query_op,
3581 	.rss_hash_update = bnxt_rss_hash_update_op,
3582 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
3583 	.link_update = bnxt_link_update_op,
3584 	.promiscuous_enable = bnxt_promiscuous_enable_op,
3585 	.promiscuous_disable = bnxt_promiscuous_disable_op,
3586 	.allmulticast_enable = bnxt_allmulticast_enable_op,
3587 	.allmulticast_disable = bnxt_allmulticast_disable_op,
3588 	.mac_addr_add = bnxt_mac_addr_add_op,
3589 	.mac_addr_remove = bnxt_mac_addr_remove_op,
3590 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
3591 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
3592 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
3593 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
3594 	.vlan_filter_set = bnxt_vlan_filter_set_op,
3595 	.vlan_offload_set = bnxt_vlan_offload_set_op,
3596 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
3597 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
3598 	.mtu_set = bnxt_mtu_set_op,
3599 	.mac_addr_set = bnxt_set_default_mac_addr_op,
3600 	.xstats_get = bnxt_dev_xstats_get_op,
3601 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
3602 	.xstats_reset = bnxt_dev_xstats_reset_op,
3603 	.fw_version_get = bnxt_fw_version_get,
3604 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
3605 	.rxq_info_get = bnxt_rxq_info_get_op,
3606 	.txq_info_get = bnxt_txq_info_get_op,
3607 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
3608 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
3609 	.dev_led_on = bnxt_dev_led_on_op,
3610 	.dev_led_off = bnxt_dev_led_off_op,
3611 	.rx_queue_start = bnxt_rx_queue_start,
3612 	.rx_queue_stop = bnxt_rx_queue_stop,
3613 	.tx_queue_start = bnxt_tx_queue_start,
3614 	.tx_queue_stop = bnxt_tx_queue_stop,
3615 	.filter_ctrl = bnxt_filter_ctrl_op,
3616 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
3617 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
3618 	.get_eeprom           = bnxt_get_eeprom_op,
3619 	.set_eeprom           = bnxt_set_eeprom_op,
3620 	.timesync_enable      = bnxt_timesync_enable,
3621 	.timesync_disable     = bnxt_timesync_disable,
3622 	.timesync_read_time   = bnxt_timesync_read_time,
3623 	.timesync_write_time   = bnxt_timesync_write_time,
3624 	.timesync_adjust_time = bnxt_timesync_adjust_time,
3625 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
3626 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
3627 };
3628 
3629 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
3630 {
3631 	uint32_t offset;
3632 
3633 	/* Only pre-map the reset GRC registers using window 3 */
3634 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
3635 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
3636 
3637 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
3638 
3639 	return offset;
3640 }
3641 
3642 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
3643 {
3644 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3645 	uint32_t reg_base = 0xffffffff;
3646 	int i;
3647 
3648 	/* Only pre-map the monitoring GRC registers using window 2 */
3649 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
3650 		uint32_t reg = info->status_regs[i];
3651 
3652 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
3653 			continue;
3654 
3655 		if (reg_base == 0xffffffff)
3656 			reg_base = reg & 0xfffff000;
3657 		if ((reg & 0xfffff000) != reg_base)
3658 			return -ERANGE;
3659 
3660 		/* Use mask 0xffc as the Lower 2 bits indicates
3661 		 * address space location
3662 		 */
3663 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
3664 						(reg & 0xffc);
3665 	}
3666 
3667 	if (reg_base == 0xffffffff)
3668 		return 0;
3669 
3670 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
3671 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
3672 
3673 	return 0;
3674 }
3675 
3676 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
3677 {
3678 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3679 	uint32_t delay = info->delay_after_reset[index];
3680 	uint32_t val = info->reset_reg_val[index];
3681 	uint32_t reg = info->reset_reg[index];
3682 	uint32_t type, offset;
3683 
3684 	type = BNXT_FW_STATUS_REG_TYPE(reg);
3685 	offset = BNXT_FW_STATUS_REG_OFF(reg);
3686 
3687 	switch (type) {
3688 	case BNXT_FW_STATUS_REG_TYPE_CFG:
3689 		rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
3690 		break;
3691 	case BNXT_FW_STATUS_REG_TYPE_GRC:
3692 		offset = bnxt_map_reset_regs(bp, offset);
3693 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3694 		break;
3695 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
3696 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3697 		break;
3698 	}
3699 	/* wait on a specific interval of time until core reset is complete */
3700 	if (delay)
3701 		rte_delay_ms(delay);
3702 }
3703 
3704 static void bnxt_dev_cleanup(struct bnxt *bp)
3705 {
3706 	bp->eth_dev->data->dev_link.link_status = 0;
3707 	bp->link_info->link_up = 0;
3708 	if (bp->eth_dev->data->dev_started)
3709 		bnxt_dev_stop(bp->eth_dev);
3710 
3711 	bnxt_uninit_resources(bp, true);
3712 }
3713 
3714 static int bnxt_restore_vlan_filters(struct bnxt *bp)
3715 {
3716 	struct rte_eth_dev *dev = bp->eth_dev;
3717 	struct rte_vlan_filter_conf *vfc;
3718 	int vidx, vbit, rc;
3719 	uint16_t vlan_id;
3720 
3721 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
3722 		vfc = &dev->data->vlan_filter_conf;
3723 		vidx = vlan_id / 64;
3724 		vbit = vlan_id % 64;
3725 
3726 		/* Each bit corresponds to a VLAN id */
3727 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
3728 			rc = bnxt_add_vlan_filter(bp, vlan_id);
3729 			if (rc)
3730 				return rc;
3731 		}
3732 	}
3733 
3734 	return 0;
3735 }
3736 
3737 static int bnxt_restore_mac_filters(struct bnxt *bp)
3738 {
3739 	struct rte_eth_dev *dev = bp->eth_dev;
3740 	struct rte_eth_dev_info dev_info;
3741 	struct rte_ether_addr *addr;
3742 	uint64_t pool_mask;
3743 	uint32_t pool = 0;
3744 	uint16_t i;
3745 	int rc;
3746 
3747 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
3748 		return 0;
3749 
3750 	rc = bnxt_dev_info_get_op(dev, &dev_info);
3751 	if (rc)
3752 		return rc;
3753 
3754 	/* replay MAC address configuration */
3755 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
3756 		addr = &dev->data->mac_addrs[i];
3757 
3758 		/* skip zero address */
3759 		if (rte_is_zero_ether_addr(addr))
3760 			continue;
3761 
3762 		pool = 0;
3763 		pool_mask = dev->data->mac_pool_sel[i];
3764 
3765 		do {
3766 			if (pool_mask & 1ULL) {
3767 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
3768 				if (rc)
3769 					return rc;
3770 			}
3771 			pool_mask >>= 1;
3772 			pool++;
3773 		} while (pool_mask);
3774 	}
3775 
3776 	return 0;
3777 }
3778 
3779 static int bnxt_restore_filters(struct bnxt *bp)
3780 {
3781 	struct rte_eth_dev *dev = bp->eth_dev;
3782 	int ret = 0;
3783 
3784 	if (dev->data->all_multicast) {
3785 		ret = bnxt_allmulticast_enable_op(dev);
3786 		if (ret)
3787 			return ret;
3788 	}
3789 	if (dev->data->promiscuous) {
3790 		ret = bnxt_promiscuous_enable_op(dev);
3791 		if (ret)
3792 			return ret;
3793 	}
3794 
3795 	ret = bnxt_restore_mac_filters(bp);
3796 	if (ret)
3797 		return ret;
3798 
3799 	ret = bnxt_restore_vlan_filters(bp);
3800 	/* TODO restore other filters as well */
3801 	return ret;
3802 }
3803 
3804 static void bnxt_dev_recover(void *arg)
3805 {
3806 	struct bnxt *bp = arg;
3807 	int timeout = bp->fw_reset_max_msecs;
3808 	int rc = 0;
3809 
3810 	pthread_mutex_lock(&bp->err_recovery_lock);
3811 	/* Clear Error flag so that device re-init should happen */
3812 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
3813 
3814 	do {
3815 		rc = bnxt_hwrm_ver_get(bp, SHORT_HWRM_CMD_TIMEOUT);
3816 		if (rc == 0)
3817 			break;
3818 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
3819 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
3820 	} while (rc && timeout);
3821 
3822 	if (rc) {
3823 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
3824 		goto err;
3825 	}
3826 
3827 	rc = bnxt_init_resources(bp, true);
3828 	if (rc) {
3829 		PMD_DRV_LOG(ERR,
3830 			    "Failed to initialize resources after reset\n");
3831 		goto err;
3832 	}
3833 	/* clear reset flag as the device is initialized now */
3834 	bp->flags &= ~BNXT_FLAG_FW_RESET;
3835 
3836 	rc = bnxt_dev_start_op(bp->eth_dev);
3837 	if (rc) {
3838 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
3839 		goto err_start;
3840 	}
3841 
3842 	rc = bnxt_restore_filters(bp);
3843 	if (rc)
3844 		goto err_start;
3845 
3846 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
3847 	pthread_mutex_unlock(&bp->err_recovery_lock);
3848 
3849 	return;
3850 err_start:
3851 	bnxt_dev_stop(bp->eth_dev);
3852 err:
3853 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
3854 	bnxt_uninit_resources(bp, false);
3855 	pthread_mutex_unlock(&bp->err_recovery_lock);
3856 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
3857 }
3858 
3859 void bnxt_dev_reset_and_resume(void *arg)
3860 {
3861 	struct bnxt *bp = arg;
3862 	int rc;
3863 
3864 	bnxt_dev_cleanup(bp);
3865 
3866 	bnxt_wait_for_device_shutdown(bp);
3867 
3868 	rc = rte_eal_alarm_set(US_PER_MS * bp->fw_reset_min_msecs,
3869 			       bnxt_dev_recover, (void *)bp);
3870 	if (rc)
3871 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
3872 }
3873 
3874 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
3875 {
3876 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3877 	uint32_t reg = info->status_regs[index];
3878 	uint32_t type, offset, val = 0;
3879 
3880 	type = BNXT_FW_STATUS_REG_TYPE(reg);
3881 	offset = BNXT_FW_STATUS_REG_OFF(reg);
3882 
3883 	switch (type) {
3884 	case BNXT_FW_STATUS_REG_TYPE_CFG:
3885 		rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
3886 		break;
3887 	case BNXT_FW_STATUS_REG_TYPE_GRC:
3888 		offset = info->mapped_status_regs[index];
3889 		/* FALLTHROUGH */
3890 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
3891 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3892 				       offset));
3893 		break;
3894 	}
3895 
3896 	return val;
3897 }
3898 
3899 static int bnxt_fw_reset_all(struct bnxt *bp)
3900 {
3901 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3902 	uint32_t i;
3903 	int rc = 0;
3904 
3905 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
3906 		/* Reset through master function driver */
3907 		for (i = 0; i < info->reg_array_cnt; i++)
3908 			bnxt_write_fw_reset_reg(bp, i);
3909 		/* Wait for time specified by FW after triggering reset */
3910 		rte_delay_ms(info->master_func_wait_period_after_reset);
3911 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
3912 		/* Reset with the help of Kong processor */
3913 		rc = bnxt_hwrm_fw_reset(bp);
3914 		if (rc)
3915 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
3916 	}
3917 
3918 	return rc;
3919 }
3920 
3921 static void bnxt_fw_reset_cb(void *arg)
3922 {
3923 	struct bnxt *bp = arg;
3924 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3925 	int rc = 0;
3926 
3927 	/* Only Master function can do FW reset */
3928 	if (bnxt_is_master_func(bp) &&
3929 	    bnxt_is_recovery_enabled(bp)) {
3930 		rc = bnxt_fw_reset_all(bp);
3931 		if (rc) {
3932 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
3933 			return;
3934 		}
3935 	}
3936 
3937 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
3938 	 * EXCEPTION_FATAL_ASYNC event to all the functions
3939 	 * (including MASTER FUNC). After receiving this Async, all the active
3940 	 * drivers should treat this case as FW initiated recovery
3941 	 */
3942 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
3943 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
3944 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
3945 
3946 		/* To recover from error */
3947 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
3948 				  (void *)bp);
3949 	}
3950 }
3951 
3952 /* Driver should poll FW heartbeat, reset_counter with the frequency
3953  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
3954  * When the driver detects heartbeat stop or change in reset_counter,
3955  * it has to trigger a reset to recover from the error condition.
3956  * A “master PF” is the function who will have the privilege to
3957  * initiate the chimp reset. The master PF will be elected by the
3958  * firmware and will be notified through async message.
3959  */
3960 static void bnxt_check_fw_health(void *arg)
3961 {
3962 	struct bnxt *bp = arg;
3963 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3964 	uint32_t val = 0, wait_msec;
3965 
3966 	if (!info || !bnxt_is_recovery_enabled(bp) ||
3967 	    is_bnxt_in_error(bp))
3968 		return;
3969 
3970 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
3971 	if (val == info->last_heart_beat)
3972 		goto reset;
3973 
3974 	info->last_heart_beat = val;
3975 
3976 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
3977 	if (val != info->last_reset_counter)
3978 		goto reset;
3979 
3980 	info->last_reset_counter = val;
3981 
3982 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
3983 			  bnxt_check_fw_health, (void *)bp);
3984 
3985 	return;
3986 reset:
3987 	/* Stop DMA to/from device */
3988 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
3989 	bp->flags |= BNXT_FLAG_FW_RESET;
3990 
3991 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
3992 
3993 	if (bnxt_is_master_func(bp))
3994 		wait_msec = info->master_func_wait_period;
3995 	else
3996 		wait_msec = info->normal_func_wait_period;
3997 
3998 	rte_eal_alarm_set(US_PER_MS * wait_msec,
3999 			  bnxt_fw_reset_cb, (void *)bp);
4000 }
4001 
4002 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4003 {
4004 	uint32_t polling_freq;
4005 
4006 	pthread_mutex_lock(&bp->health_check_lock);
4007 
4008 	if (!bnxt_is_recovery_enabled(bp))
4009 		goto done;
4010 
4011 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4012 		goto done;
4013 
4014 	polling_freq = bp->recovery_info->driver_polling_freq;
4015 
4016 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4017 			  bnxt_check_fw_health, (void *)bp);
4018 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4019 
4020 done:
4021 	pthread_mutex_unlock(&bp->health_check_lock);
4022 }
4023 
4024 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4025 {
4026 	if (!bnxt_is_recovery_enabled(bp))
4027 		return;
4028 
4029 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4030 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4031 }
4032 
4033 static bool bnxt_vf_pciid(uint16_t device_id)
4034 {
4035 	switch (device_id) {
4036 	case BROADCOM_DEV_ID_57304_VF:
4037 	case BROADCOM_DEV_ID_57406_VF:
4038 	case BROADCOM_DEV_ID_5731X_VF:
4039 	case BROADCOM_DEV_ID_5741X_VF:
4040 	case BROADCOM_DEV_ID_57414_VF:
4041 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4042 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4043 	case BROADCOM_DEV_ID_58802_VF:
4044 	case BROADCOM_DEV_ID_57500_VF1:
4045 	case BROADCOM_DEV_ID_57500_VF2:
4046 	case BROADCOM_DEV_ID_58818_VF:
4047 		/* FALLTHROUGH */
4048 		return true;
4049 	default:
4050 		return false;
4051 	}
4052 }
4053 
4054 /* Phase 5 device */
4055 static bool bnxt_p5_device(uint16_t device_id)
4056 {
4057 	switch (device_id) {
4058 	case BROADCOM_DEV_ID_57508:
4059 	case BROADCOM_DEV_ID_57504:
4060 	case BROADCOM_DEV_ID_57502:
4061 	case BROADCOM_DEV_ID_57508_MF1:
4062 	case BROADCOM_DEV_ID_57504_MF1:
4063 	case BROADCOM_DEV_ID_57502_MF1:
4064 	case BROADCOM_DEV_ID_57508_MF2:
4065 	case BROADCOM_DEV_ID_57504_MF2:
4066 	case BROADCOM_DEV_ID_57502_MF2:
4067 	case BROADCOM_DEV_ID_57500_VF1:
4068 	case BROADCOM_DEV_ID_57500_VF2:
4069 	case BROADCOM_DEV_ID_58812:
4070 	case BROADCOM_DEV_ID_58814:
4071 	case BROADCOM_DEV_ID_58818:
4072 	case BROADCOM_DEV_ID_58818_VF:
4073 		/* FALLTHROUGH */
4074 		return true;
4075 	default:
4076 		return false;
4077 	}
4078 }
4079 
4080 bool bnxt_stratus_device(struct bnxt *bp)
4081 {
4082 	uint16_t device_id = bp->pdev->id.device_id;
4083 
4084 	switch (device_id) {
4085 	case BROADCOM_DEV_ID_STRATUS_NIC:
4086 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4087 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4088 		/* FALLTHROUGH */
4089 		return true;
4090 	default:
4091 		return false;
4092 	}
4093 }
4094 
4095 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4096 {
4097 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4098 	struct bnxt *bp = eth_dev->data->dev_private;
4099 
4100 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4101 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4102 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4103 	if (!bp->bar0 || !bp->doorbell_base) {
4104 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4105 		return -ENODEV;
4106 	}
4107 
4108 	bp->eth_dev = eth_dev;
4109 	bp->pdev = pci_dev;
4110 
4111 	return 0;
4112 }
4113 
4114 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4115 				  struct bnxt_ctx_pg_info *ctx_pg,
4116 				  uint32_t mem_size,
4117 				  const char *suffix,
4118 				  uint16_t idx)
4119 {
4120 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4121 	const struct rte_memzone *mz = NULL;
4122 	char mz_name[RTE_MEMZONE_NAMESIZE];
4123 	rte_iova_t mz_phys_addr;
4124 	uint64_t valid_bits = 0;
4125 	uint32_t sz;
4126 	int i;
4127 
4128 	if (!mem_size)
4129 		return 0;
4130 
4131 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4132 			 BNXT_PAGE_SIZE;
4133 	rmem->page_size = BNXT_PAGE_SIZE;
4134 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4135 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4136 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4137 
4138 	valid_bits = PTU_PTE_VALID;
4139 
4140 	if (rmem->nr_pages > 1) {
4141 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4142 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4143 			 suffix, idx, bp->eth_dev->data->port_id);
4144 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4145 		mz = rte_memzone_lookup(mz_name);
4146 		if (!mz) {
4147 			mz = rte_memzone_reserve_aligned(mz_name,
4148 						rmem->nr_pages * 8,
4149 						SOCKET_ID_ANY,
4150 						RTE_MEMZONE_2MB |
4151 						RTE_MEMZONE_SIZE_HINT_ONLY |
4152 						RTE_MEMZONE_IOVA_CONTIG,
4153 						BNXT_PAGE_SIZE);
4154 			if (mz == NULL)
4155 				return -ENOMEM;
4156 		}
4157 
4158 		memset(mz->addr, 0, mz->len);
4159 		mz_phys_addr = mz->iova;
4160 
4161 		rmem->pg_tbl = mz->addr;
4162 		rmem->pg_tbl_map = mz_phys_addr;
4163 		rmem->pg_tbl_mz = mz;
4164 	}
4165 
4166 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4167 		 suffix, idx, bp->eth_dev->data->port_id);
4168 	mz = rte_memzone_lookup(mz_name);
4169 	if (!mz) {
4170 		mz = rte_memzone_reserve_aligned(mz_name,
4171 						 mem_size,
4172 						 SOCKET_ID_ANY,
4173 						 RTE_MEMZONE_1GB |
4174 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4175 						 RTE_MEMZONE_IOVA_CONTIG,
4176 						 BNXT_PAGE_SIZE);
4177 		if (mz == NULL)
4178 			return -ENOMEM;
4179 	}
4180 
4181 	memset(mz->addr, 0, mz->len);
4182 	mz_phys_addr = mz->iova;
4183 
4184 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4185 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4186 		rmem->dma_arr[i] = mz_phys_addr + sz;
4187 
4188 		if (rmem->nr_pages > 1) {
4189 			if (i == rmem->nr_pages - 2 &&
4190 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4191 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4192 			else if (i == rmem->nr_pages - 1 &&
4193 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4194 				valid_bits |= PTU_PTE_LAST;
4195 
4196 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4197 							   valid_bits);
4198 		}
4199 	}
4200 
4201 	rmem->mz = mz;
4202 	if (rmem->vmem_size)
4203 		rmem->vmem = (void **)mz->addr;
4204 	rmem->dma_arr[0] = mz_phys_addr;
4205 	return 0;
4206 }
4207 
4208 static void bnxt_free_ctx_mem(struct bnxt *bp)
4209 {
4210 	int i;
4211 
4212 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4213 		return;
4214 
4215 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4216 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4217 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4218 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4219 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4220 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4221 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4222 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4223 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4224 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4225 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4226 
4227 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4228 		if (bp->ctx->tqm_mem[i])
4229 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4230 	}
4231 
4232 	rte_free(bp->ctx);
4233 	bp->ctx = NULL;
4234 }
4235 
4236 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4237 
4238 #define min_t(type, x, y) ({                    \
4239 	type __min1 = (x);                      \
4240 	type __min2 = (y);                      \
4241 	__min1 < __min2 ? __min1 : __min2; })
4242 
4243 #define max_t(type, x, y) ({                    \
4244 	type __max1 = (x);                      \
4245 	type __max2 = (y);                      \
4246 	__max1 > __max2 ? __max1 : __max2; })
4247 
4248 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4249 
4250 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4251 {
4252 	struct bnxt_ctx_pg_info *ctx_pg;
4253 	struct bnxt_ctx_mem_info *ctx;
4254 	uint32_t mem_size, ena, entries;
4255 	uint32_t entries_sp, min;
4256 	int i, rc;
4257 
4258 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4259 	if (rc) {
4260 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4261 		return rc;
4262 	}
4263 	ctx = bp->ctx;
4264 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4265 		return 0;
4266 
4267 	ctx_pg = &ctx->qp_mem;
4268 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4269 	if (ctx->qp_entry_size) {
4270 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4271 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4272 		if (rc)
4273 			return rc;
4274 	}
4275 
4276 	ctx_pg = &ctx->srq_mem;
4277 	ctx_pg->entries = ctx->srq_max_l2_entries;
4278 	if (ctx->srq_entry_size) {
4279 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4280 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4281 		if (rc)
4282 			return rc;
4283 	}
4284 
4285 	ctx_pg = &ctx->cq_mem;
4286 	ctx_pg->entries = ctx->cq_max_l2_entries;
4287 	if (ctx->cq_entry_size) {
4288 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4289 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4290 		if (rc)
4291 			return rc;
4292 	}
4293 
4294 	ctx_pg = &ctx->vnic_mem;
4295 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4296 		ctx->vnic_max_ring_table_entries;
4297 	if (ctx->vnic_entry_size) {
4298 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4299 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4300 		if (rc)
4301 			return rc;
4302 	}
4303 
4304 	ctx_pg = &ctx->stat_mem;
4305 	ctx_pg->entries = ctx->stat_max_entries;
4306 	if (ctx->stat_entry_size) {
4307 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4308 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4309 		if (rc)
4310 			return rc;
4311 	}
4312 
4313 	min = ctx->tqm_min_entries_per_ring;
4314 
4315 	entries_sp = ctx->qp_max_l2_entries +
4316 		     ctx->vnic_max_vnic_entries +
4317 		     2 * ctx->qp_min_qp1_entries + min;
4318 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4319 
4320 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4321 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4322 	entries = clamp_t(uint32_t, entries, min,
4323 			  ctx->tqm_max_entries_per_ring);
4324 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4325 		ctx_pg = ctx->tqm_mem[i];
4326 		ctx_pg->entries = i ? entries : entries_sp;
4327 		if (ctx->tqm_entry_size) {
4328 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4329 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "tqm_mem", i);
4330 			if (rc)
4331 				return rc;
4332 		}
4333 		ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4334 	}
4335 
4336 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4337 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4338 	if (rc)
4339 		PMD_DRV_LOG(ERR,
4340 			    "Failed to configure context mem: rc = %d\n", rc);
4341 	else
4342 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4343 
4344 	return rc;
4345 }
4346 
4347 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4348 {
4349 	struct rte_pci_device *pci_dev = bp->pdev;
4350 	char mz_name[RTE_MEMZONE_NAMESIZE];
4351 	const struct rte_memzone *mz = NULL;
4352 	uint32_t total_alloc_len;
4353 	rte_iova_t mz_phys_addr;
4354 
4355 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4356 		return 0;
4357 
4358 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4359 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4360 		 pci_dev->addr.bus, pci_dev->addr.devid,
4361 		 pci_dev->addr.function, "rx_port_stats");
4362 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4363 	mz = rte_memzone_lookup(mz_name);
4364 	total_alloc_len =
4365 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4366 				       sizeof(struct rx_port_stats_ext) + 512);
4367 	if (!mz) {
4368 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4369 					 SOCKET_ID_ANY,
4370 					 RTE_MEMZONE_2MB |
4371 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4372 					 RTE_MEMZONE_IOVA_CONTIG);
4373 		if (mz == NULL)
4374 			return -ENOMEM;
4375 	}
4376 	memset(mz->addr, 0, mz->len);
4377 	mz_phys_addr = mz->iova;
4378 
4379 	bp->rx_mem_zone = (const void *)mz;
4380 	bp->hw_rx_port_stats = mz->addr;
4381 	bp->hw_rx_port_stats_map = mz_phys_addr;
4382 
4383 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4384 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4385 		 pci_dev->addr.bus, pci_dev->addr.devid,
4386 		 pci_dev->addr.function, "tx_port_stats");
4387 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4388 	mz = rte_memzone_lookup(mz_name);
4389 	total_alloc_len =
4390 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4391 				       sizeof(struct tx_port_stats_ext) + 512);
4392 	if (!mz) {
4393 		mz = rte_memzone_reserve(mz_name,
4394 					 total_alloc_len,
4395 					 SOCKET_ID_ANY,
4396 					 RTE_MEMZONE_2MB |
4397 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4398 					 RTE_MEMZONE_IOVA_CONTIG);
4399 		if (mz == NULL)
4400 			return -ENOMEM;
4401 	}
4402 	memset(mz->addr, 0, mz->len);
4403 	mz_phys_addr = mz->iova;
4404 
4405 	bp->tx_mem_zone = (const void *)mz;
4406 	bp->hw_tx_port_stats = mz->addr;
4407 	bp->hw_tx_port_stats_map = mz_phys_addr;
4408 	bp->flags |= BNXT_FLAG_PORT_STATS;
4409 
4410 	/* Display extended statistics if FW supports it */
4411 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4412 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4413 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4414 		return 0;
4415 
4416 	bp->hw_rx_port_stats_ext = (void *)
4417 		((uint8_t *)bp->hw_rx_port_stats +
4418 		 sizeof(struct rx_port_stats));
4419 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4420 		sizeof(struct rx_port_stats);
4421 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4422 
4423 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4424 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4425 		bp->hw_tx_port_stats_ext = (void *)
4426 			((uint8_t *)bp->hw_tx_port_stats +
4427 			 sizeof(struct tx_port_stats));
4428 		bp->hw_tx_port_stats_ext_map =
4429 			bp->hw_tx_port_stats_map +
4430 			sizeof(struct tx_port_stats);
4431 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4432 	}
4433 
4434 	return 0;
4435 }
4436 
4437 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4438 {
4439 	struct bnxt *bp = eth_dev->data->dev_private;
4440 	int rc = 0;
4441 
4442 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4443 					       RTE_ETHER_ADDR_LEN *
4444 					       bp->max_l2_ctx,
4445 					       0);
4446 	if (eth_dev->data->mac_addrs == NULL) {
4447 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4448 		return -ENOMEM;
4449 	}
4450 
4451 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4452 		if (BNXT_PF(bp))
4453 			return -EINVAL;
4454 
4455 		/* Generate a random MAC address, if none was assigned by PF */
4456 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4457 		bnxt_eth_hw_addr_random(bp->mac_addr);
4458 		PMD_DRV_LOG(INFO,
4459 			    "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4460 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4461 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4462 
4463 		rc = bnxt_hwrm_set_mac(bp);
4464 		if (rc)
4465 			return rc;
4466 	}
4467 
4468 	/* Copy the permanent MAC from the FUNC_QCAPS response */
4469 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
4470 
4471 	return rc;
4472 }
4473 
4474 static int bnxt_restore_dflt_mac(struct bnxt *bp)
4475 {
4476 	int rc = 0;
4477 
4478 	/* MAC is already configured in FW */
4479 	if (BNXT_HAS_DFLT_MAC_SET(bp))
4480 		return 0;
4481 
4482 	/* Restore the old MAC configured */
4483 	rc = bnxt_hwrm_set_mac(bp);
4484 	if (rc)
4485 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
4486 
4487 	return rc;
4488 }
4489 
4490 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
4491 {
4492 	if (!BNXT_PF(bp))
4493 		return;
4494 
4495 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
4496 
4497 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
4498 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
4499 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
4500 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
4501 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
4502 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
4503 }
4504 
4505 uint16_t
4506 bnxt_get_svif(uint16_t port_id, bool func_svif,
4507 	      enum bnxt_ulp_intf_type type)
4508 {
4509 	struct rte_eth_dev *eth_dev;
4510 	struct bnxt *bp;
4511 
4512 	eth_dev = &rte_eth_devices[port_id];
4513 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4514 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4515 		if (!vfr)
4516 			return 0;
4517 
4518 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4519 			return vfr->svif;
4520 
4521 		eth_dev = vfr->parent_dev;
4522 	}
4523 
4524 	bp = eth_dev->data->dev_private;
4525 
4526 	return func_svif ? bp->func_svif : bp->port_svif;
4527 }
4528 
4529 uint16_t
4530 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
4531 {
4532 	struct rte_eth_dev *eth_dev;
4533 	struct bnxt_vnic_info *vnic;
4534 	struct bnxt *bp;
4535 
4536 	eth_dev = &rte_eth_devices[port];
4537 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4538 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4539 		if (!vfr)
4540 			return 0;
4541 
4542 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4543 			return vfr->dflt_vnic_id;
4544 
4545 		eth_dev = vfr->parent_dev;
4546 	}
4547 
4548 	bp = eth_dev->data->dev_private;
4549 
4550 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
4551 
4552 	return vnic->fw_vnic_id;
4553 }
4554 
4555 uint16_t
4556 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
4557 {
4558 	struct rte_eth_dev *eth_dev;
4559 	struct bnxt *bp;
4560 
4561 	eth_dev = &rte_eth_devices[port];
4562 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4563 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4564 		if (!vfr)
4565 			return 0;
4566 
4567 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4568 			return vfr->fw_fid;
4569 
4570 		eth_dev = vfr->parent_dev;
4571 	}
4572 
4573 	bp = eth_dev->data->dev_private;
4574 
4575 	return bp->fw_fid;
4576 }
4577 
4578 enum bnxt_ulp_intf_type
4579 bnxt_get_interface_type(uint16_t port)
4580 {
4581 	struct rte_eth_dev *eth_dev;
4582 	struct bnxt *bp;
4583 
4584 	eth_dev = &rte_eth_devices[port];
4585 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
4586 		return BNXT_ULP_INTF_TYPE_VF_REP;
4587 
4588 	bp = eth_dev->data->dev_private;
4589 	if (BNXT_PF(bp))
4590 		return BNXT_ULP_INTF_TYPE_PF;
4591 	else if (BNXT_VF_IS_TRUSTED(bp))
4592 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
4593 	else if (BNXT_VF(bp))
4594 		return BNXT_ULP_INTF_TYPE_VF;
4595 
4596 	return BNXT_ULP_INTF_TYPE_INVALID;
4597 }
4598 
4599 uint16_t
4600 bnxt_get_phy_port_id(uint16_t port_id)
4601 {
4602 	struct bnxt_representor *vfr;
4603 	struct rte_eth_dev *eth_dev;
4604 	struct bnxt *bp;
4605 
4606 	eth_dev = &rte_eth_devices[port_id];
4607 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4608 		vfr = eth_dev->data->dev_private;
4609 		if (!vfr)
4610 			return 0;
4611 
4612 		eth_dev = vfr->parent_dev;
4613 	}
4614 
4615 	bp = eth_dev->data->dev_private;
4616 
4617 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
4618 }
4619 
4620 uint16_t
4621 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
4622 {
4623 	struct rte_eth_dev *eth_dev;
4624 	struct bnxt *bp;
4625 
4626 	eth_dev = &rte_eth_devices[port_id];
4627 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4628 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4629 		if (!vfr)
4630 			return 0;
4631 
4632 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4633 			return vfr->fw_fid - 1;
4634 
4635 		eth_dev = vfr->parent_dev;
4636 	}
4637 
4638 	bp = eth_dev->data->dev_private;
4639 
4640 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
4641 }
4642 
4643 uint16_t
4644 bnxt_get_vport(uint16_t port_id)
4645 {
4646 	return (1 << bnxt_get_phy_port_id(port_id));
4647 }
4648 
4649 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
4650 {
4651 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4652 
4653 	if (info) {
4654 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
4655 			memset(info, 0, sizeof(*info));
4656 		return;
4657 	}
4658 
4659 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
4660 		return;
4661 
4662 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4663 			   sizeof(*info), 0);
4664 	if (!info)
4665 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4666 
4667 	bp->recovery_info = info;
4668 }
4669 
4670 static void bnxt_check_fw_status(struct bnxt *bp)
4671 {
4672 	uint32_t fw_status;
4673 
4674 	if (!(bp->recovery_info &&
4675 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
4676 		return;
4677 
4678 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
4679 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
4680 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
4681 			    fw_status);
4682 }
4683 
4684 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
4685 {
4686 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4687 	uint32_t status_loc;
4688 	uint32_t sig_ver;
4689 
4690 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
4691 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4692 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4693 				   BNXT_GRCP_WINDOW_2_BASE +
4694 				   offsetof(struct hcomm_status,
4695 					    sig_ver)));
4696 	/* If the signature is absent, then FW does not support this feature */
4697 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
4698 	    HCOMM_STATUS_SIGNATURE_VAL)
4699 		return 0;
4700 
4701 	if (!info) {
4702 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4703 				   sizeof(*info), 0);
4704 		if (!info)
4705 			return -ENOMEM;
4706 		bp->recovery_info = info;
4707 	} else {
4708 		memset(info, 0, sizeof(*info));
4709 	}
4710 
4711 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4712 				      BNXT_GRCP_WINDOW_2_BASE +
4713 				      offsetof(struct hcomm_status,
4714 					       fw_status_loc)));
4715 
4716 	/* Only pre-map the FW health status GRC register */
4717 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
4718 		return 0;
4719 
4720 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
4721 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
4722 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
4723 
4724 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
4725 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4726 
4727 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
4728 
4729 	return 0;
4730 }
4731 
4732 /* This function gets the FW version along with the
4733  * capabilities(MAX and current) of the function, vnic,
4734  * error recovery, phy and other chip related info
4735  */
4736 static int bnxt_get_config(struct bnxt *bp)
4737 {
4738 	uint16_t mtu;
4739 	int rc = 0;
4740 
4741 	bp->fw_cap = 0;
4742 
4743 	rc = bnxt_map_hcomm_fw_status_reg(bp);
4744 	if (rc)
4745 		return rc;
4746 
4747 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
4748 	if (rc) {
4749 		bnxt_check_fw_status(bp);
4750 		return rc;
4751 	}
4752 
4753 	rc = bnxt_hwrm_func_reset(bp);
4754 	if (rc)
4755 		return -EIO;
4756 
4757 	rc = bnxt_hwrm_vnic_qcaps(bp);
4758 	if (rc)
4759 		return rc;
4760 
4761 	rc = bnxt_hwrm_queue_qportcfg(bp);
4762 	if (rc)
4763 		return rc;
4764 
4765 	/* Get the MAX capabilities for this function.
4766 	 * This function also allocates context memory for TQM rings and
4767 	 * informs the firmware about this allocated backing store memory.
4768 	 */
4769 	rc = bnxt_hwrm_func_qcaps(bp);
4770 	if (rc)
4771 		return rc;
4772 
4773 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
4774 	if (rc)
4775 		return rc;
4776 
4777 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
4778 	if (rc)
4779 		return rc;
4780 
4781 	bnxt_hwrm_port_mac_qcfg(bp);
4782 
4783 	bnxt_hwrm_parent_pf_qcfg(bp);
4784 
4785 	bnxt_hwrm_port_phy_qcaps(bp);
4786 
4787 	bnxt_alloc_error_recovery_info(bp);
4788 	/* Get the adapter error recovery support info */
4789 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
4790 	if (rc)
4791 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4792 
4793 	bnxt_hwrm_port_led_qcaps(bp);
4794 
4795 	return 0;
4796 }
4797 
4798 static int
4799 bnxt_init_locks(struct bnxt *bp)
4800 {
4801 	int err;
4802 
4803 	err = pthread_mutex_init(&bp->flow_lock, NULL);
4804 	if (err) {
4805 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
4806 		return err;
4807 	}
4808 
4809 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
4810 	if (err) {
4811 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
4812 		return err;
4813 	}
4814 
4815 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
4816 	if (err) {
4817 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
4818 		return err;
4819 	}
4820 
4821 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
4822 	if (err)
4823 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
4824 
4825 	return err;
4826 }
4827 
4828 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
4829 {
4830 	int rc = 0;
4831 
4832 	rc = bnxt_get_config(bp);
4833 	if (rc)
4834 		return rc;
4835 
4836 	if (!reconfig_dev) {
4837 		rc = bnxt_setup_mac_addr(bp->eth_dev);
4838 		if (rc)
4839 			return rc;
4840 	} else {
4841 		rc = bnxt_restore_dflt_mac(bp);
4842 		if (rc)
4843 			return rc;
4844 	}
4845 
4846 	bnxt_config_vf_req_fwd(bp);
4847 
4848 	rc = bnxt_hwrm_func_driver_register(bp);
4849 	if (rc) {
4850 		PMD_DRV_LOG(ERR, "Failed to register driver");
4851 		return -EBUSY;
4852 	}
4853 
4854 	if (BNXT_PF(bp)) {
4855 		if (bp->pdev->max_vfs) {
4856 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
4857 			if (rc) {
4858 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
4859 				return rc;
4860 			}
4861 		} else {
4862 			rc = bnxt_hwrm_allocate_pf_only(bp);
4863 			if (rc) {
4864 				PMD_DRV_LOG(ERR,
4865 					    "Failed to allocate PF resources");
4866 				return rc;
4867 			}
4868 		}
4869 	}
4870 
4871 	rc = bnxt_alloc_mem(bp, reconfig_dev);
4872 	if (rc)
4873 		return rc;
4874 
4875 	rc = bnxt_setup_int(bp);
4876 	if (rc)
4877 		return rc;
4878 
4879 	rc = bnxt_request_int(bp);
4880 	if (rc)
4881 		return rc;
4882 
4883 	rc = bnxt_init_ctx_mem(bp);
4884 	if (rc) {
4885 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
4886 		return rc;
4887 	}
4888 
4889 	return 0;
4890 }
4891 
4892 static int
4893 bnxt_parse_devarg_truflow(__rte_unused const char *key,
4894 			  const char *value, void *opaque_arg)
4895 {
4896 	struct bnxt *bp = opaque_arg;
4897 	unsigned long truflow;
4898 	char *end = NULL;
4899 
4900 	if (!value || !opaque_arg) {
4901 		PMD_DRV_LOG(ERR,
4902 			    "Invalid parameter passed to truflow devargs.\n");
4903 		return -EINVAL;
4904 	}
4905 
4906 	truflow = strtoul(value, &end, 10);
4907 	if (end == NULL || *end != '\0' ||
4908 	    (truflow == ULONG_MAX && errno == ERANGE)) {
4909 		PMD_DRV_LOG(ERR,
4910 			    "Invalid parameter passed to truflow devargs.\n");
4911 		return -EINVAL;
4912 	}
4913 
4914 	if (BNXT_DEVARG_TRUFLOW_INVALID(truflow)) {
4915 		PMD_DRV_LOG(ERR,
4916 			    "Invalid value passed to truflow devargs.\n");
4917 		return -EINVAL;
4918 	}
4919 
4920 	if (truflow) {
4921 		bp->flags |= BNXT_FLAG_TRUFLOW_EN;
4922 		PMD_DRV_LOG(INFO, "Host-based truflow feature enabled.\n");
4923 	} else {
4924 		bp->flags &= ~BNXT_FLAG_TRUFLOW_EN;
4925 		PMD_DRV_LOG(INFO, "Host-based truflow feature disabled.\n");
4926 	}
4927 
4928 	return 0;
4929 }
4930 
4931 static int
4932 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
4933 			     const char *value, void *opaque_arg)
4934 {
4935 	struct bnxt *bp = opaque_arg;
4936 	unsigned long flow_xstat;
4937 	char *end = NULL;
4938 
4939 	if (!value || !opaque_arg) {
4940 		PMD_DRV_LOG(ERR,
4941 			    "Invalid parameter passed to flow_xstat devarg.\n");
4942 		return -EINVAL;
4943 	}
4944 
4945 	flow_xstat = strtoul(value, &end, 10);
4946 	if (end == NULL || *end != '\0' ||
4947 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
4948 		PMD_DRV_LOG(ERR,
4949 			    "Invalid parameter passed to flow_xstat devarg.\n");
4950 		return -EINVAL;
4951 	}
4952 
4953 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
4954 		PMD_DRV_LOG(ERR,
4955 			    "Invalid value passed to flow_xstat devarg.\n");
4956 		return -EINVAL;
4957 	}
4958 
4959 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
4960 	if (BNXT_FLOW_XSTATS_EN(bp))
4961 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
4962 
4963 	return 0;
4964 }
4965 
4966 static int
4967 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
4968 					const char *value, void *opaque_arg)
4969 {
4970 	struct bnxt *bp = opaque_arg;
4971 	unsigned long max_num_kflows;
4972 	char *end = NULL;
4973 
4974 	if (!value || !opaque_arg) {
4975 		PMD_DRV_LOG(ERR,
4976 			"Invalid parameter passed to max_num_kflows devarg.\n");
4977 		return -EINVAL;
4978 	}
4979 
4980 	max_num_kflows = strtoul(value, &end, 10);
4981 	if (end == NULL || *end != '\0' ||
4982 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
4983 		PMD_DRV_LOG(ERR,
4984 			"Invalid parameter passed to max_num_kflows devarg.\n");
4985 		return -EINVAL;
4986 	}
4987 
4988 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
4989 		PMD_DRV_LOG(ERR,
4990 			"Invalid value passed to max_num_kflows devarg.\n");
4991 		return -EINVAL;
4992 	}
4993 
4994 	bp->max_num_kflows = max_num_kflows;
4995 	if (bp->max_num_kflows)
4996 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
4997 				max_num_kflows);
4998 
4999 	return 0;
5000 }
5001 
5002 static int
5003 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5004 			    const char *value, void *opaque_arg)
5005 {
5006 	struct bnxt_representor *vfr_bp = opaque_arg;
5007 	unsigned long rep_is_pf;
5008 	char *end = NULL;
5009 
5010 	if (!value || !opaque_arg) {
5011 		PMD_DRV_LOG(ERR,
5012 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5013 		return -EINVAL;
5014 	}
5015 
5016 	rep_is_pf = strtoul(value, &end, 10);
5017 	if (end == NULL || *end != '\0' ||
5018 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5019 		PMD_DRV_LOG(ERR,
5020 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5021 		return -EINVAL;
5022 	}
5023 
5024 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5025 		PMD_DRV_LOG(ERR,
5026 			    "Invalid value passed to rep_is_pf devargs.\n");
5027 		return -EINVAL;
5028 	}
5029 
5030 	vfr_bp->flags |= rep_is_pf;
5031 	if (BNXT_REP_PF(vfr_bp))
5032 		PMD_DRV_LOG(INFO, "PF representor\n");
5033 	else
5034 		PMD_DRV_LOG(INFO, "VF representor\n");
5035 
5036 	return 0;
5037 }
5038 
5039 static int
5040 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5041 			       const char *value, void *opaque_arg)
5042 {
5043 	struct bnxt_representor *vfr_bp = opaque_arg;
5044 	unsigned long rep_based_pf;
5045 	char *end = NULL;
5046 
5047 	if (!value || !opaque_arg) {
5048 		PMD_DRV_LOG(ERR,
5049 			    "Invalid parameter passed to rep_based_pf "
5050 			    "devargs.\n");
5051 		return -EINVAL;
5052 	}
5053 
5054 	rep_based_pf = strtoul(value, &end, 10);
5055 	if (end == NULL || *end != '\0' ||
5056 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5057 		PMD_DRV_LOG(ERR,
5058 			    "Invalid parameter passed to rep_based_pf "
5059 			    "devargs.\n");
5060 		return -EINVAL;
5061 	}
5062 
5063 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5064 		PMD_DRV_LOG(ERR,
5065 			    "Invalid value passed to rep_based_pf devargs.\n");
5066 		return -EINVAL;
5067 	}
5068 
5069 	vfr_bp->rep_based_pf = rep_based_pf;
5070 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5071 
5072 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5073 
5074 	return 0;
5075 }
5076 
5077 static int
5078 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5079 			    const char *value, void *opaque_arg)
5080 {
5081 	struct bnxt_representor *vfr_bp = opaque_arg;
5082 	unsigned long rep_q_r2f;
5083 	char *end = NULL;
5084 
5085 	if (!value || !opaque_arg) {
5086 		PMD_DRV_LOG(ERR,
5087 			    "Invalid parameter passed to rep_q_r2f "
5088 			    "devargs.\n");
5089 		return -EINVAL;
5090 	}
5091 
5092 	rep_q_r2f = strtoul(value, &end, 10);
5093 	if (end == NULL || *end != '\0' ||
5094 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5095 		PMD_DRV_LOG(ERR,
5096 			    "Invalid parameter passed to rep_q_r2f "
5097 			    "devargs.\n");
5098 		return -EINVAL;
5099 	}
5100 
5101 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5102 		PMD_DRV_LOG(ERR,
5103 			    "Invalid value passed to rep_q_r2f devargs.\n");
5104 		return -EINVAL;
5105 	}
5106 
5107 	vfr_bp->rep_q_r2f = rep_q_r2f;
5108 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5109 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5110 
5111 	return 0;
5112 }
5113 
5114 static int
5115 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5116 			    const char *value, void *opaque_arg)
5117 {
5118 	struct bnxt_representor *vfr_bp = opaque_arg;
5119 	unsigned long rep_q_f2r;
5120 	char *end = NULL;
5121 
5122 	if (!value || !opaque_arg) {
5123 		PMD_DRV_LOG(ERR,
5124 			    "Invalid parameter passed to rep_q_f2r "
5125 			    "devargs.\n");
5126 		return -EINVAL;
5127 	}
5128 
5129 	rep_q_f2r = strtoul(value, &end, 10);
5130 	if (end == NULL || *end != '\0' ||
5131 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5132 		PMD_DRV_LOG(ERR,
5133 			    "Invalid parameter passed to rep_q_f2r "
5134 			    "devargs.\n");
5135 		return -EINVAL;
5136 	}
5137 
5138 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5139 		PMD_DRV_LOG(ERR,
5140 			    "Invalid value passed to rep_q_f2r devargs.\n");
5141 		return -EINVAL;
5142 	}
5143 
5144 	vfr_bp->rep_q_f2r = rep_q_f2r;
5145 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5146 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5147 
5148 	return 0;
5149 }
5150 
5151 static int
5152 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5153 			     const char *value, void *opaque_arg)
5154 {
5155 	struct bnxt_representor *vfr_bp = opaque_arg;
5156 	unsigned long rep_fc_r2f;
5157 	char *end = NULL;
5158 
5159 	if (!value || !opaque_arg) {
5160 		PMD_DRV_LOG(ERR,
5161 			    "Invalid parameter passed to rep_fc_r2f "
5162 			    "devargs.\n");
5163 		return -EINVAL;
5164 	}
5165 
5166 	rep_fc_r2f = strtoul(value, &end, 10);
5167 	if (end == NULL || *end != '\0' ||
5168 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5169 		PMD_DRV_LOG(ERR,
5170 			    "Invalid parameter passed to rep_fc_r2f "
5171 			    "devargs.\n");
5172 		return -EINVAL;
5173 	}
5174 
5175 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5176 		PMD_DRV_LOG(ERR,
5177 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5178 		return -EINVAL;
5179 	}
5180 
5181 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5182 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5183 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5184 
5185 	return 0;
5186 }
5187 
5188 static int
5189 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5190 			     const char *value, void *opaque_arg)
5191 {
5192 	struct bnxt_representor *vfr_bp = opaque_arg;
5193 	unsigned long rep_fc_f2r;
5194 	char *end = NULL;
5195 
5196 	if (!value || !opaque_arg) {
5197 		PMD_DRV_LOG(ERR,
5198 			    "Invalid parameter passed to rep_fc_f2r "
5199 			    "devargs.\n");
5200 		return -EINVAL;
5201 	}
5202 
5203 	rep_fc_f2r = strtoul(value, &end, 10);
5204 	if (end == NULL || *end != '\0' ||
5205 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5206 		PMD_DRV_LOG(ERR,
5207 			    "Invalid parameter passed to rep_fc_f2r "
5208 			    "devargs.\n");
5209 		return -EINVAL;
5210 	}
5211 
5212 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5213 		PMD_DRV_LOG(ERR,
5214 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5215 		return -EINVAL;
5216 	}
5217 
5218 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5219 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5220 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5221 
5222 	return 0;
5223 }
5224 
5225 static void
5226 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5227 {
5228 	struct rte_kvargs *kvlist;
5229 
5230 	if (devargs == NULL)
5231 		return;
5232 
5233 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5234 	if (kvlist == NULL)
5235 		return;
5236 
5237 	/*
5238 	 * Handler for "truflow" devarg.
5239 	 * Invoked as for ex: "-a 0000:00:0d.0,host-based-truflow=1"
5240 	 */
5241 	rte_kvargs_process(kvlist, BNXT_DEVARG_TRUFLOW,
5242 			   bnxt_parse_devarg_truflow, bp);
5243 
5244 	/*
5245 	 * Handler for "flow_xstat" devarg.
5246 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5247 	 */
5248 	rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5249 			   bnxt_parse_devarg_flow_xstat, bp);
5250 
5251 	/*
5252 	 * Handler for "max_num_kflows" devarg.
5253 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5254 	 */
5255 	rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5256 			   bnxt_parse_devarg_max_num_kflows, bp);
5257 
5258 	rte_kvargs_free(kvlist);
5259 }
5260 
5261 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5262 {
5263 	int rc = 0;
5264 
5265 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5266 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5267 		if (rc)
5268 			PMD_DRV_LOG(ERR,
5269 				    "Failed to alloc switch domain: %d\n", rc);
5270 		else
5271 			PMD_DRV_LOG(INFO,
5272 				    "Switch domain allocated %d\n",
5273 				    bp->switch_domain_id);
5274 	}
5275 
5276 	return rc;
5277 }
5278 
5279 /* Allocate and initialize various fields in bnxt struct that
5280  * need to be allocated/destroyed only once in the lifetime of the driver
5281  */
5282 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5283 {
5284 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5285 	struct bnxt *bp = eth_dev->data->dev_private;
5286 	int rc = 0;
5287 
5288 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5289 
5290 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5291 		bp->flags |= BNXT_FLAG_VF;
5292 
5293 	if (bnxt_p5_device(pci_dev->id.device_id))
5294 		bp->flags |= BNXT_FLAG_CHIP_P5;
5295 
5296 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5297 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5298 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5299 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5300 		bp->flags |= BNXT_FLAG_STINGRAY;
5301 
5302 	if (BNXT_TRUFLOW_EN(bp)) {
5303 		/* extra mbuf field is required to store CFA code from mark */
5304 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5305 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5306 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5307 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5308 		};
5309 		bnxt_cfa_code_dynfield_offset =
5310 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5311 		if (bnxt_cfa_code_dynfield_offset < 0) {
5312 			PMD_DRV_LOG(ERR,
5313 			    "Failed to register mbuf field for TruFlow mark\n");
5314 			return -rte_errno;
5315 		}
5316 	}
5317 
5318 	rc = bnxt_map_pci_bars(eth_dev);
5319 	if (rc) {
5320 		PMD_DRV_LOG(ERR,
5321 			    "Failed to initialize board rc: %x\n", rc);
5322 		return rc;
5323 	}
5324 
5325 	rc = bnxt_alloc_pf_info(bp);
5326 	if (rc)
5327 		return rc;
5328 
5329 	rc = bnxt_alloc_link_info(bp);
5330 	if (rc)
5331 		return rc;
5332 
5333 	rc = bnxt_alloc_parent_info(bp);
5334 	if (rc)
5335 		return rc;
5336 
5337 	rc = bnxt_alloc_hwrm_resources(bp);
5338 	if (rc) {
5339 		PMD_DRV_LOG(ERR,
5340 			    "Failed to allocate hwrm resource rc: %x\n", rc);
5341 		return rc;
5342 	}
5343 	rc = bnxt_alloc_leds_info(bp);
5344 	if (rc)
5345 		return rc;
5346 
5347 	rc = bnxt_alloc_cos_queues(bp);
5348 	if (rc)
5349 		return rc;
5350 
5351 	rc = bnxt_init_locks(bp);
5352 	if (rc)
5353 		return rc;
5354 
5355 	rc = bnxt_alloc_switch_domain(bp);
5356 	if (rc)
5357 		return rc;
5358 
5359 	return rc;
5360 }
5361 
5362 static int
5363 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5364 {
5365 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5366 	static int version_printed;
5367 	struct bnxt *bp;
5368 	int rc;
5369 
5370 	if (version_printed++ == 0)
5371 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5372 
5373 	eth_dev->dev_ops = &bnxt_dev_ops;
5374 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5375 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5376 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5377 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5378 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5379 
5380 	/*
5381 	 * For secondary processes, we don't initialise any further
5382 	 * as primary has already done this work.
5383 	 */
5384 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5385 		return 0;
5386 
5387 	rte_eth_copy_pci_info(eth_dev, pci_dev);
5388 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5389 
5390 	bp = eth_dev->data->dev_private;
5391 
5392 	/* Parse dev arguments passed on when starting the DPDK application. */
5393 	bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5394 
5395 	rc = bnxt_drv_init(eth_dev);
5396 	if (rc)
5397 		goto error_free;
5398 
5399 	rc = bnxt_init_resources(bp, false);
5400 	if (rc)
5401 		goto error_free;
5402 
5403 	rc = bnxt_alloc_stats_mem(bp);
5404 	if (rc)
5405 		goto error_free;
5406 
5407 	PMD_DRV_LOG(INFO,
5408 		    DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
5409 		    pci_dev->mem_resource[0].phys_addr,
5410 		    pci_dev->mem_resource[0].addr);
5411 
5412 	return 0;
5413 
5414 error_free:
5415 	bnxt_dev_uninit(eth_dev);
5416 	return rc;
5417 }
5418 
5419 
5420 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5421 {
5422 	if (!ctx)
5423 		return;
5424 
5425 	if (ctx->va)
5426 		rte_free(ctx->va);
5427 
5428 	ctx->va = NULL;
5429 	ctx->dma = RTE_BAD_IOVA;
5430 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5431 }
5432 
5433 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5434 {
5435 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5436 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5437 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
5438 				  bp->flow_stat->max_fc,
5439 				  false);
5440 
5441 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5442 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5443 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
5444 				  bp->flow_stat->max_fc,
5445 				  false);
5446 
5447 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5448 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5449 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5450 
5451 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5452 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5453 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5454 
5455 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5456 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5457 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5458 
5459 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5460 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5461 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5462 }
5463 
5464 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5465 {
5466 	bnxt_unregister_fc_ctx_mem(bp);
5467 
5468 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5469 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5470 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5471 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5472 }
5473 
5474 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5475 {
5476 	if (BNXT_FLOW_XSTATS_EN(bp))
5477 		bnxt_uninit_fc_ctx_mem(bp);
5478 }
5479 
5480 static void
5481 bnxt_free_error_recovery_info(struct bnxt *bp)
5482 {
5483 	rte_free(bp->recovery_info);
5484 	bp->recovery_info = NULL;
5485 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5486 }
5487 
5488 static int
5489 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5490 {
5491 	int rc;
5492 
5493 	bnxt_free_int(bp);
5494 	bnxt_free_mem(bp, reconfig_dev);
5495 
5496 	bnxt_hwrm_func_buf_unrgtr(bp);
5497 	rte_free(bp->pf->vf_req_buf);
5498 
5499 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
5500 	bp->flags &= ~BNXT_FLAG_REGISTERED;
5501 	bnxt_free_ctx_mem(bp);
5502 	if (!reconfig_dev) {
5503 		bnxt_free_hwrm_resources(bp);
5504 		bnxt_free_error_recovery_info(bp);
5505 	}
5506 
5507 	bnxt_uninit_ctx_mem(bp);
5508 
5509 	bnxt_free_flow_stats_info(bp);
5510 	bnxt_free_rep_info(bp);
5511 	rte_free(bp->ptp_cfg);
5512 	bp->ptp_cfg = NULL;
5513 	return rc;
5514 }
5515 
5516 static int
5517 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5518 {
5519 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5520 		return -EPERM;
5521 
5522 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5523 
5524 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5525 		bnxt_dev_close_op(eth_dev);
5526 
5527 	return 0;
5528 }
5529 
5530 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
5531 {
5532 	struct bnxt *bp = eth_dev->data->dev_private;
5533 	struct rte_eth_dev *vf_rep_eth_dev;
5534 	int ret = 0, i;
5535 
5536 	if (!bp)
5537 		return -EINVAL;
5538 
5539 	for (i = 0; i < bp->num_reps; i++) {
5540 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
5541 		if (!vf_rep_eth_dev)
5542 			continue;
5543 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
5544 			    vf_rep_eth_dev->data->port_id);
5545 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
5546 	}
5547 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
5548 		    eth_dev->data->port_id);
5549 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
5550 
5551 	return ret;
5552 }
5553 
5554 static void bnxt_free_rep_info(struct bnxt *bp)
5555 {
5556 	rte_free(bp->rep_info);
5557 	bp->rep_info = NULL;
5558 	rte_free(bp->cfa_code_map);
5559 	bp->cfa_code_map = NULL;
5560 }
5561 
5562 static int bnxt_init_rep_info(struct bnxt *bp)
5563 {
5564 	int i = 0, rc;
5565 
5566 	if (bp->rep_info)
5567 		return 0;
5568 
5569 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
5570 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
5571 				   0);
5572 	if (!bp->rep_info) {
5573 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
5574 		return -ENOMEM;
5575 	}
5576 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
5577 				       sizeof(*bp->cfa_code_map) *
5578 				       BNXT_MAX_CFA_CODE, 0);
5579 	if (!bp->cfa_code_map) {
5580 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
5581 		bnxt_free_rep_info(bp);
5582 		return -ENOMEM;
5583 	}
5584 
5585 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
5586 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
5587 
5588 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
5589 	if (rc) {
5590 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
5591 		bnxt_free_rep_info(bp);
5592 		return rc;
5593 	}
5594 
5595 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
5596 	if (rc) {
5597 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
5598 		bnxt_free_rep_info(bp);
5599 		return rc;
5600 	}
5601 
5602 	return rc;
5603 }
5604 
5605 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
5606 			       struct rte_eth_devargs *eth_da,
5607 			       struct rte_eth_dev *backing_eth_dev,
5608 			       const char *dev_args)
5609 {
5610 	struct rte_eth_dev *vf_rep_eth_dev;
5611 	char name[RTE_ETH_NAME_MAX_LEN];
5612 	struct bnxt *backing_bp;
5613 	uint16_t num_rep;
5614 	int i, ret = 0;
5615 	struct rte_kvargs *kvlist = NULL;
5616 
5617 	num_rep = eth_da->nb_representor_ports;
5618 	if (num_rep > BNXT_MAX_VF_REPS) {
5619 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
5620 			    num_rep, BNXT_MAX_VF_REPS);
5621 		return -EINVAL;
5622 	}
5623 
5624 	if (num_rep >= RTE_MAX_ETHPORTS) {
5625 		PMD_DRV_LOG(ERR,
5626 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
5627 			    num_rep, RTE_MAX_ETHPORTS);
5628 		return -EINVAL;
5629 	}
5630 
5631 	backing_bp = backing_eth_dev->data->dev_private;
5632 
5633 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
5634 		PMD_DRV_LOG(ERR,
5635 			    "Not a PF or trusted VF. No Representor support\n");
5636 		/* Returning an error is not an option.
5637 		 * Applications are not handling this correctly
5638 		 */
5639 		return 0;
5640 	}
5641 
5642 	if (bnxt_init_rep_info(backing_bp))
5643 		return 0;
5644 
5645 	for (i = 0; i < num_rep; i++) {
5646 		struct bnxt_representor representor = {
5647 			.vf_id = eth_da->representor_ports[i],
5648 			.switch_domain_id = backing_bp->switch_domain_id,
5649 			.parent_dev = backing_eth_dev
5650 		};
5651 
5652 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
5653 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
5654 				    representor.vf_id, BNXT_MAX_VF_REPS);
5655 			continue;
5656 		}
5657 
5658 		/* representor port net_bdf_port */
5659 		snprintf(name, sizeof(name), "net_%s_representor_%d",
5660 			 pci_dev->device.name, eth_da->representor_ports[i]);
5661 
5662 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
5663 		if (kvlist) {
5664 			/*
5665 			 * Handler for "rep_is_pf" devarg.
5666 			 * Invoked as for ex: "-a 000:00:0d.0,
5667 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5668 			 */
5669 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
5670 						 bnxt_parse_devarg_rep_is_pf,
5671 						 (void *)&representor);
5672 			if (ret) {
5673 				ret = -EINVAL;
5674 				goto err;
5675 			}
5676 			/*
5677 			 * Handler for "rep_based_pf" devarg.
5678 			 * Invoked as for ex: "-a 000:00:0d.0,
5679 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5680 			 */
5681 			ret = rte_kvargs_process(kvlist,
5682 						 BNXT_DEVARG_REP_BASED_PF,
5683 						 bnxt_parse_devarg_rep_based_pf,
5684 						 (void *)&representor);
5685 			if (ret) {
5686 				ret = -EINVAL;
5687 				goto err;
5688 			}
5689 			/*
5690 			 * Handler for "rep_based_pf" devarg.
5691 			 * Invoked as for ex: "-a 000:00:0d.0,
5692 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5693 			 */
5694 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
5695 						 bnxt_parse_devarg_rep_q_r2f,
5696 						 (void *)&representor);
5697 			if (ret) {
5698 				ret = -EINVAL;
5699 				goto err;
5700 			}
5701 			/*
5702 			 * Handler for "rep_based_pf" devarg.
5703 			 * Invoked as for ex: "-a 000:00:0d.0,
5704 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5705 			 */
5706 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
5707 						 bnxt_parse_devarg_rep_q_f2r,
5708 						 (void *)&representor);
5709 			if (ret) {
5710 				ret = -EINVAL;
5711 				goto err;
5712 			}
5713 			/*
5714 			 * Handler for "rep_based_pf" devarg.
5715 			 * Invoked as for ex: "-a 000:00:0d.0,
5716 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5717 			 */
5718 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
5719 						 bnxt_parse_devarg_rep_fc_r2f,
5720 						 (void *)&representor);
5721 			if (ret) {
5722 				ret = -EINVAL;
5723 				goto err;
5724 			}
5725 			/*
5726 			 * Handler for "rep_based_pf" devarg.
5727 			 * Invoked as for ex: "-a 000:00:0d.0,
5728 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5729 			 */
5730 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
5731 						 bnxt_parse_devarg_rep_fc_f2r,
5732 						 (void *)&representor);
5733 			if (ret) {
5734 				ret = -EINVAL;
5735 				goto err;
5736 			}
5737 		}
5738 
5739 		ret = rte_eth_dev_create(&pci_dev->device, name,
5740 					 sizeof(struct bnxt_representor),
5741 					 NULL, NULL,
5742 					 bnxt_representor_init,
5743 					 &representor);
5744 		if (ret) {
5745 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
5746 				    "representor %s.", name);
5747 			goto err;
5748 		}
5749 
5750 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
5751 		if (!vf_rep_eth_dev) {
5752 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
5753 				    " for VF-Rep: %s.", name);
5754 			ret = -ENODEV;
5755 			goto err;
5756 		}
5757 
5758 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
5759 			    backing_eth_dev->data->port_id);
5760 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
5761 							 vf_rep_eth_dev;
5762 		backing_bp->num_reps++;
5763 
5764 	}
5765 
5766 	rte_kvargs_free(kvlist);
5767 	return 0;
5768 
5769 err:
5770 	/* If num_rep > 1, then rollback already created
5771 	 * ports, since we'll be failing the probe anyway
5772 	 */
5773 	if (num_rep > 1)
5774 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
5775 	rte_errno = -ret;
5776 	rte_kvargs_free(kvlist);
5777 
5778 	return ret;
5779 }
5780 
5781 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
5782 			  struct rte_pci_device *pci_dev)
5783 {
5784 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
5785 	struct rte_eth_dev *backing_eth_dev;
5786 	uint16_t num_rep;
5787 	int ret = 0;
5788 
5789 	if (pci_dev->device.devargs) {
5790 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
5791 					    &eth_da);
5792 		if (ret)
5793 			return ret;
5794 	}
5795 
5796 	num_rep = eth_da.nb_representor_ports;
5797 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
5798 		    num_rep);
5799 
5800 	/* We could come here after first level of probe is already invoked
5801 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
5802 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
5803 	 */
5804 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
5805 	if (backing_eth_dev == NULL) {
5806 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
5807 					 sizeof(struct bnxt),
5808 					 eth_dev_pci_specific_init, pci_dev,
5809 					 bnxt_dev_init, NULL);
5810 
5811 		if (ret || !num_rep)
5812 			return ret;
5813 
5814 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
5815 	}
5816 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
5817 		    backing_eth_dev->data->port_id);
5818 
5819 	if (!num_rep)
5820 		return ret;
5821 
5822 	/* probe representor ports now */
5823 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
5824 				  pci_dev->device.devargs->args);
5825 
5826 	return ret;
5827 }
5828 
5829 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
5830 {
5831 	struct rte_eth_dev *eth_dev;
5832 
5833 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
5834 	if (!eth_dev)
5835 		return 0; /* Invoked typically only by OVS-DPDK, by the
5836 			   * time it comes here the eth_dev is already
5837 			   * deleted by rte_eth_dev_close(), so returning
5838 			   * +ve value will at least help in proper cleanup
5839 			   */
5840 
5841 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
5842 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
5843 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
5844 			return rte_eth_dev_destroy(eth_dev,
5845 						   bnxt_representor_uninit);
5846 		else
5847 			return rte_eth_dev_destroy(eth_dev,
5848 						   bnxt_dev_uninit);
5849 	} else {
5850 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
5851 	}
5852 }
5853 
5854 static struct rte_pci_driver bnxt_rte_pmd = {
5855 	.id_table = bnxt_pci_id_map,
5856 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
5857 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
5858 						  * and OVS-DPDK
5859 						  */
5860 	.probe = bnxt_pci_probe,
5861 	.remove = bnxt_pci_remove,
5862 };
5863 
5864 static bool
5865 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
5866 {
5867 	if (strcmp(dev->device->driver->name, drv->driver.name))
5868 		return false;
5869 
5870 	return true;
5871 }
5872 
5873 bool is_bnxt_supported(struct rte_eth_dev *dev)
5874 {
5875 	return is_device_supported(dev, &bnxt_rte_pmd);
5876 }
5877 
5878 RTE_LOG_REGISTER(bnxt_logtype_driver, pmd.net.bnxt.driver, NOTICE);
5879 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
5880 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
5881 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
5882