xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision b7b78a089c454d42eb654360eeecb1e2f15e6cd8)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2014-2021 Broadcom
3  * All rights reserved.
4  */
5 
6 #include <inttypes.h>
7 #include <stdbool.h>
8 
9 #include <rte_dev.h>
10 #include <ethdev_driver.h>
11 #include <ethdev_pci.h>
12 #include <rte_malloc.h>
13 #include <rte_cycles.h>
14 #include <rte_alarm.h>
15 #include <rte_kvargs.h>
16 #include <rte_vect.h>
17 
18 #include "bnxt.h"
19 #include "bnxt_filter.h"
20 #include "bnxt_hwrm.h"
21 #include "bnxt_irq.h"
22 #include "bnxt_reps.h"
23 #include "bnxt_ring.h"
24 #include "bnxt_rxq.h"
25 #include "bnxt_rxr.h"
26 #include "bnxt_stats.h"
27 #include "bnxt_txq.h"
28 #include "bnxt_txr.h"
29 #include "bnxt_vnic.h"
30 #include "hsi_struct_def_dpdk.h"
31 #include "bnxt_nvm_defs.h"
32 #include "bnxt_tf_common.h"
33 #include "ulp_flow_db.h"
34 #include "rte_pmd_bnxt.h"
35 
36 #define DRV_MODULE_NAME		"bnxt"
37 static const char bnxt_version[] =
38 	"Broadcom NetXtreme driver " DRV_MODULE_NAME;
39 
40 /*
41  * The set of PCI devices this driver supports
42  */
43 static const struct rte_pci_id bnxt_pci_id_map[] = {
44 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
45 			 BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
46 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
47 			 BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
48 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
49 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
50 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
51 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
52 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
53 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
54 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
55 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
56 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
57 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
58 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
59 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
60 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
61 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
62 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
63 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
64 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
65 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
66 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
67 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
68 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
69 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
70 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
71 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
72 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
73 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
74 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
75 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
76 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
77 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
78 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
79 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
80 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
81 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
82 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
83 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58812) },
84 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58814) },
85 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818) },
86 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818_VF) },
87 	{ .vendor_id = 0, /* sentinel */ },
88 };
89 
90 #define BNXT_DEVARG_TRUFLOW	"host-based-truflow"
91 #define BNXT_DEVARG_FLOW_XSTAT	"flow-xstat"
92 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
93 #define BNXT_DEVARG_REPRESENTOR	"representor"
94 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
95 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
96 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
97 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
98 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
99 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
100 
101 static const char *const bnxt_dev_args[] = {
102 	BNXT_DEVARG_REPRESENTOR,
103 	BNXT_DEVARG_TRUFLOW,
104 	BNXT_DEVARG_FLOW_XSTAT,
105 	BNXT_DEVARG_MAX_NUM_KFLOWS,
106 	BNXT_DEVARG_REP_BASED_PF,
107 	BNXT_DEVARG_REP_IS_PF,
108 	BNXT_DEVARG_REP_Q_R2F,
109 	BNXT_DEVARG_REP_Q_F2R,
110 	BNXT_DEVARG_REP_FC_R2F,
111 	BNXT_DEVARG_REP_FC_F2R,
112 	NULL
113 };
114 
115 /*
116  * truflow == false to disable the feature
117  * truflow == true to enable the feature
118  */
119 #define	BNXT_DEVARG_TRUFLOW_INVALID(truflow)	((truflow) > 1)
120 
121 /*
122  * flow_xstat == false to disable the feature
123  * flow_xstat == true to enable the feature
124  */
125 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
126 
127 /*
128  * rep_is_pf == false to indicate VF representor
129  * rep_is_pf == true to indicate PF representor
130  */
131 #define	BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)	((rep_is_pf) > 1)
132 
133 /*
134  * rep_based_pf == Physical index of the PF
135  */
136 #define	BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)	((rep_based_pf) > 15)
137 /*
138  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
139  */
140 #define	BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)	((rep_q_r2f) > 3)
141 
142 /*
143  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
144  */
145 #define	BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)	((rep_q_f2r) > 3)
146 
147 /*
148  * rep_fc_r2f == Flow control for the representor to endpoint direction
149  */
150 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)	((rep_fc_r2f) > 1)
151 
152 /*
153  * rep_fc_f2r == Flow control for the endpoint to representor direction
154  */
155 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)	((rep_fc_f2r) > 1)
156 
157 int bnxt_cfa_code_dynfield_offset = -1;
158 
159 /*
160  * max_num_kflows must be >= 32
161  * and must be a power-of-2 supported value
162  * return: 1 -> invalid
163  *         0 -> valid
164  */
165 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
166 {
167 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
168 		return 1;
169 	return 0;
170 }
171 
172 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
173 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
174 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
175 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
176 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
177 static int bnxt_restore_vlan_filters(struct bnxt *bp);
178 static void bnxt_dev_recover(void *arg);
179 static void bnxt_free_error_recovery_info(struct bnxt *bp);
180 static void bnxt_free_rep_info(struct bnxt *bp);
181 
182 int is_bnxt_in_error(struct bnxt *bp)
183 {
184 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
185 		return -EIO;
186 	if (bp->flags & BNXT_FLAG_FW_RESET)
187 		return -EBUSY;
188 
189 	return 0;
190 }
191 
192 /***********************/
193 
194 /*
195  * High level utility functions
196  */
197 
198 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
199 {
200 	unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
201 					     BNXT_RSS_TBL_SIZE_P5);
202 
203 	if (!BNXT_CHIP_P5(bp))
204 		return 1;
205 
206 	return RTE_ALIGN_MUL_CEIL(num_rss_rings,
207 				  BNXT_RSS_ENTRIES_PER_CTX_P5) /
208 				  BNXT_RSS_ENTRIES_PER_CTX_P5;
209 }
210 
211 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
212 {
213 	if (!BNXT_CHIP_P5(bp))
214 		return HW_HASH_INDEX_SIZE;
215 
216 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
217 }
218 
219 static void bnxt_free_parent_info(struct bnxt *bp)
220 {
221 	rte_free(bp->parent);
222 }
223 
224 static void bnxt_free_pf_info(struct bnxt *bp)
225 {
226 	rte_free(bp->pf);
227 }
228 
229 static void bnxt_free_link_info(struct bnxt *bp)
230 {
231 	rte_free(bp->link_info);
232 }
233 
234 static void bnxt_free_leds_info(struct bnxt *bp)
235 {
236 	if (BNXT_VF(bp))
237 		return;
238 
239 	rte_free(bp->leds);
240 	bp->leds = NULL;
241 }
242 
243 static void bnxt_free_flow_stats_info(struct bnxt *bp)
244 {
245 	rte_free(bp->flow_stat);
246 	bp->flow_stat = NULL;
247 }
248 
249 static void bnxt_free_cos_queues(struct bnxt *bp)
250 {
251 	rte_free(bp->rx_cos_queue);
252 	rte_free(bp->tx_cos_queue);
253 }
254 
255 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
256 {
257 	bnxt_free_filter_mem(bp);
258 	bnxt_free_vnic_attributes(bp);
259 	bnxt_free_vnic_mem(bp);
260 
261 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
262 	 * If the number of rings change across fw update,
263 	 * we don't have much choice except to warn the user.
264 	 */
265 	if (!reconfig) {
266 		bnxt_free_stats(bp);
267 		bnxt_free_tx_rings(bp);
268 		bnxt_free_rx_rings(bp);
269 	}
270 	bnxt_free_async_cp_ring(bp);
271 	bnxt_free_rxtx_nq_ring(bp);
272 
273 	rte_free(bp->grp_info);
274 	bp->grp_info = NULL;
275 }
276 
277 static int bnxt_alloc_parent_info(struct bnxt *bp)
278 {
279 	bp->parent = rte_zmalloc("bnxt_parent_info",
280 				 sizeof(struct bnxt_parent_info), 0);
281 	if (bp->parent == NULL)
282 		return -ENOMEM;
283 
284 	return 0;
285 }
286 
287 static int bnxt_alloc_pf_info(struct bnxt *bp)
288 {
289 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
290 	if (bp->pf == NULL)
291 		return -ENOMEM;
292 
293 	return 0;
294 }
295 
296 static int bnxt_alloc_link_info(struct bnxt *bp)
297 {
298 	bp->link_info =
299 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
300 	if (bp->link_info == NULL)
301 		return -ENOMEM;
302 
303 	return 0;
304 }
305 
306 static int bnxt_alloc_leds_info(struct bnxt *bp)
307 {
308 	if (BNXT_VF(bp))
309 		return 0;
310 
311 	bp->leds = rte_zmalloc("bnxt_leds",
312 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
313 			       0);
314 	if (bp->leds == NULL)
315 		return -ENOMEM;
316 
317 	return 0;
318 }
319 
320 static int bnxt_alloc_cos_queues(struct bnxt *bp)
321 {
322 	bp->rx_cos_queue =
323 		rte_zmalloc("bnxt_rx_cosq",
324 			    BNXT_COS_QUEUE_COUNT *
325 			    sizeof(struct bnxt_cos_queue_info),
326 			    0);
327 	if (bp->rx_cos_queue == NULL)
328 		return -ENOMEM;
329 
330 	bp->tx_cos_queue =
331 		rte_zmalloc("bnxt_tx_cosq",
332 			    BNXT_COS_QUEUE_COUNT *
333 			    sizeof(struct bnxt_cos_queue_info),
334 			    0);
335 	if (bp->tx_cos_queue == NULL)
336 		return -ENOMEM;
337 
338 	return 0;
339 }
340 
341 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
342 {
343 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
344 				    sizeof(struct bnxt_flow_stat_info), 0);
345 	if (bp->flow_stat == NULL)
346 		return -ENOMEM;
347 
348 	return 0;
349 }
350 
351 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
352 {
353 	int rc;
354 
355 	rc = bnxt_alloc_ring_grps(bp);
356 	if (rc)
357 		goto alloc_mem_err;
358 
359 	rc = bnxt_alloc_async_ring_struct(bp);
360 	if (rc)
361 		goto alloc_mem_err;
362 
363 	rc = bnxt_alloc_vnic_mem(bp);
364 	if (rc)
365 		goto alloc_mem_err;
366 
367 	rc = bnxt_alloc_vnic_attributes(bp);
368 	if (rc)
369 		goto alloc_mem_err;
370 
371 	rc = bnxt_alloc_filter_mem(bp);
372 	if (rc)
373 		goto alloc_mem_err;
374 
375 	rc = bnxt_alloc_async_cp_ring(bp);
376 	if (rc)
377 		goto alloc_mem_err;
378 
379 	rc = bnxt_alloc_rxtx_nq_ring(bp);
380 	if (rc)
381 		goto alloc_mem_err;
382 
383 	if (BNXT_FLOW_XSTATS_EN(bp)) {
384 		rc = bnxt_alloc_flow_stats_info(bp);
385 		if (rc)
386 			goto alloc_mem_err;
387 	}
388 
389 	return 0;
390 
391 alloc_mem_err:
392 	bnxt_free_mem(bp, reconfig);
393 	return rc;
394 }
395 
396 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
397 {
398 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
399 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
400 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
401 	struct bnxt_rx_queue *rxq;
402 	unsigned int j;
403 	int rc;
404 
405 	rc = bnxt_vnic_grp_alloc(bp, vnic);
406 	if (rc)
407 		goto err_out;
408 
409 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
410 		    vnic_id, vnic, vnic->fw_grp_ids);
411 
412 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
413 	if (rc)
414 		goto err_out;
415 
416 	/* Alloc RSS context only if RSS mode is enabled */
417 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
418 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
419 
420 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
421 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
422 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
423 			PMD_DRV_LOG(ERR,
424 				    "Only queues 0-%d will be in RSS table\n",
425 				    BNXT_RSS_TBL_SIZE_P5 - 1);
426 		}
427 
428 		rc = 0;
429 		for (j = 0; j < nr_ctxs; j++) {
430 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
431 			if (rc)
432 				break;
433 		}
434 		if (rc) {
435 			PMD_DRV_LOG(ERR,
436 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
437 				    vnic_id, j, rc);
438 			goto err_out;
439 		}
440 		vnic->num_lb_ctxts = nr_ctxs;
441 	}
442 
443 	/*
444 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
445 	 * setting is not available at this time, it will not be
446 	 * configured correctly in the CFA.
447 	 */
448 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
449 		vnic->vlan_strip = true;
450 	else
451 		vnic->vlan_strip = false;
452 
453 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
454 	if (rc)
455 		goto err_out;
456 
457 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
458 	if (rc)
459 		goto err_out;
460 
461 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
462 		rxq = bp->eth_dev->data->rx_queues[j];
463 
464 		PMD_DRV_LOG(DEBUG,
465 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
466 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
467 
468 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
469 			rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
470 		else
471 			vnic->rx_queue_cnt++;
472 	}
473 
474 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
475 
476 	rc = bnxt_vnic_rss_configure(bp, vnic);
477 	if (rc)
478 		goto err_out;
479 
480 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
481 
482 	if (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO)
483 		bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 1);
484 	else
485 		bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 0);
486 
487 	return 0;
488 err_out:
489 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
490 		    vnic_id, rc);
491 	return rc;
492 }
493 
494 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
495 {
496 	int rc = 0;
497 
498 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
499 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
500 	if (rc)
501 		return rc;
502 
503 	PMD_DRV_LOG(DEBUG,
504 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
505 		    " rx_fc_in_tbl.ctx_id = %d\n",
506 		    bp->flow_stat->rx_fc_in_tbl.va,
507 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
508 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
509 
510 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
511 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
512 	if (rc)
513 		return rc;
514 
515 	PMD_DRV_LOG(DEBUG,
516 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
517 		    " rx_fc_out_tbl.ctx_id = %d\n",
518 		    bp->flow_stat->rx_fc_out_tbl.va,
519 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
520 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
521 
522 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
523 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
524 	if (rc)
525 		return rc;
526 
527 	PMD_DRV_LOG(DEBUG,
528 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
529 		    " tx_fc_in_tbl.ctx_id = %d\n",
530 		    bp->flow_stat->tx_fc_in_tbl.va,
531 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
532 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
533 
534 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
535 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
536 	if (rc)
537 		return rc;
538 
539 	PMD_DRV_LOG(DEBUG,
540 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
541 		    " tx_fc_out_tbl.ctx_id = %d\n",
542 		    bp->flow_stat->tx_fc_out_tbl.va,
543 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
544 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
545 
546 	memset(bp->flow_stat->rx_fc_out_tbl.va,
547 	       0,
548 	       bp->flow_stat->rx_fc_out_tbl.size);
549 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
550 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
551 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
552 				       bp->flow_stat->max_fc,
553 				       true);
554 	if (rc)
555 		return rc;
556 
557 	memset(bp->flow_stat->tx_fc_out_tbl.va,
558 	       0,
559 	       bp->flow_stat->tx_fc_out_tbl.size);
560 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
561 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
562 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
563 				       bp->flow_stat->max_fc,
564 				       true);
565 
566 	return rc;
567 }
568 
569 static int bnxt_alloc_ctx_mem_buf(char *type, size_t size,
570 				  struct bnxt_ctx_mem_buf_info *ctx)
571 {
572 	if (!ctx)
573 		return -EINVAL;
574 
575 	ctx->va = rte_zmalloc(type, size, 0);
576 	if (ctx->va == NULL)
577 		return -ENOMEM;
578 	rte_mem_lock_page(ctx->va);
579 	ctx->size = size;
580 	ctx->dma = rte_mem_virt2iova(ctx->va);
581 	if (ctx->dma == RTE_BAD_IOVA)
582 		return -ENOMEM;
583 
584 	return 0;
585 }
586 
587 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
588 {
589 	struct rte_pci_device *pdev = bp->pdev;
590 	char type[RTE_MEMZONE_NAMESIZE];
591 	uint16_t max_fc;
592 	int rc = 0;
593 
594 	max_fc = bp->flow_stat->max_fc;
595 
596 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
597 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
598 	/* 4 bytes for each counter-id */
599 	rc = bnxt_alloc_ctx_mem_buf(type,
600 				    max_fc * 4,
601 				    &bp->flow_stat->rx_fc_in_tbl);
602 	if (rc)
603 		return rc;
604 
605 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
606 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
607 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
608 	rc = bnxt_alloc_ctx_mem_buf(type,
609 				    max_fc * 16,
610 				    &bp->flow_stat->rx_fc_out_tbl);
611 	if (rc)
612 		return rc;
613 
614 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
615 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
616 	/* 4 bytes for each counter-id */
617 	rc = bnxt_alloc_ctx_mem_buf(type,
618 				    max_fc * 4,
619 				    &bp->flow_stat->tx_fc_in_tbl);
620 	if (rc)
621 		return rc;
622 
623 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
624 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
625 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
626 	rc = bnxt_alloc_ctx_mem_buf(type,
627 				    max_fc * 16,
628 				    &bp->flow_stat->tx_fc_out_tbl);
629 	if (rc)
630 		return rc;
631 
632 	rc = bnxt_register_fc_ctx_mem(bp);
633 
634 	return rc;
635 }
636 
637 static int bnxt_init_ctx_mem(struct bnxt *bp)
638 {
639 	int rc = 0;
640 
641 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
642 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
643 	    !BNXT_FLOW_XSTATS_EN(bp))
644 		return 0;
645 
646 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
647 	if (rc)
648 		return rc;
649 
650 	rc = bnxt_init_fc_ctx_mem(bp);
651 
652 	return rc;
653 }
654 
655 static int bnxt_update_phy_setting(struct bnxt *bp)
656 {
657 	struct rte_eth_link new;
658 	int rc;
659 
660 	rc = bnxt_get_hwrm_link_config(bp, &new);
661 	if (rc) {
662 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
663 		return rc;
664 	}
665 
666 	/*
667 	 * On BCM957508-N2100 adapters, FW will not allow any user other
668 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
669 	 * always returns link up. Force phy update always in that case.
670 	 */
671 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
672 		rc = bnxt_set_hwrm_link_config(bp, true);
673 		if (rc) {
674 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
675 			return rc;
676 		}
677 	}
678 
679 	return rc;
680 }
681 
682 static int bnxt_start_nic(struct bnxt *bp)
683 {
684 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
685 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
686 	uint32_t intr_vector = 0;
687 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
688 	uint32_t vec = BNXT_MISC_VEC_ID;
689 	unsigned int i, j;
690 	int rc;
691 
692 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
693 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
694 			DEV_RX_OFFLOAD_JUMBO_FRAME;
695 		bp->flags |= BNXT_FLAG_JUMBO;
696 	} else {
697 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
698 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
699 		bp->flags &= ~BNXT_FLAG_JUMBO;
700 	}
701 
702 	/* THOR does not support ring groups.
703 	 * But we will use the array to save RSS context IDs.
704 	 */
705 	if (BNXT_CHIP_P5(bp))
706 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
707 
708 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
709 	if (rc) {
710 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
711 		goto err_out;
712 	}
713 
714 	rc = bnxt_alloc_hwrm_rings(bp);
715 	if (rc) {
716 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
717 		goto err_out;
718 	}
719 
720 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
721 	if (rc) {
722 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
723 		goto err_out;
724 	}
725 
726 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
727 		goto skip_cosq_cfg;
728 
729 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
730 		if (bp->rx_cos_queue[i].id != 0xff) {
731 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
732 
733 			if (!vnic) {
734 				PMD_DRV_LOG(ERR,
735 					    "Num pools more than FW profile\n");
736 				rc = -EINVAL;
737 				goto err_out;
738 			}
739 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
740 			bp->rx_cosq_cnt++;
741 		}
742 	}
743 
744 skip_cosq_cfg:
745 	rc = bnxt_mq_rx_configure(bp);
746 	if (rc) {
747 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
748 		goto err_out;
749 	}
750 
751 	/* default vnic 0 */
752 	rc = bnxt_setup_one_vnic(bp, 0);
753 	if (rc)
754 		goto err_out;
755 	/* VNIC configuration */
756 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
757 		for (i = 1; i < bp->nr_vnics; i++) {
758 			rc = bnxt_setup_one_vnic(bp, i);
759 			if (rc)
760 				goto err_out;
761 		}
762 	}
763 
764 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
765 	if (rc) {
766 		PMD_DRV_LOG(ERR,
767 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
768 		goto err_out;
769 	}
770 
771 	/* check and configure queue intr-vector mapping */
772 	if ((rte_intr_cap_multiple(intr_handle) ||
773 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
774 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
775 		intr_vector = bp->eth_dev->data->nb_rx_queues;
776 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
777 		if (intr_vector > bp->rx_cp_nr_rings) {
778 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
779 					bp->rx_cp_nr_rings);
780 			return -ENOTSUP;
781 		}
782 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
783 		if (rc)
784 			return rc;
785 	}
786 
787 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
788 		intr_handle->intr_vec =
789 			rte_zmalloc("intr_vec",
790 				    bp->eth_dev->data->nb_rx_queues *
791 				    sizeof(int), 0);
792 		if (intr_handle->intr_vec == NULL) {
793 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
794 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
795 			rc = -ENOMEM;
796 			goto err_disable;
797 		}
798 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
799 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
800 			 intr_handle->intr_vec, intr_handle->nb_efd,
801 			intr_handle->max_intr);
802 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
803 		     queue_id++) {
804 			intr_handle->intr_vec[queue_id] =
805 							vec + BNXT_RX_VEC_START;
806 			if (vec < base + intr_handle->nb_efd - 1)
807 				vec++;
808 		}
809 	}
810 
811 	/* enable uio/vfio intr/eventfd mapping */
812 	rc = rte_intr_enable(intr_handle);
813 #ifndef RTE_EXEC_ENV_FREEBSD
814 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
815 	if (rc)
816 		goto err_free;
817 #endif
818 
819 	rc = bnxt_update_phy_setting(bp);
820 	if (rc)
821 		goto err_free;
822 
823 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
824 	if (!bp->mark_table)
825 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
826 
827 	return 0;
828 
829 err_free:
830 	rte_free(intr_handle->intr_vec);
831 err_disable:
832 	rte_intr_efd_disable(intr_handle);
833 err_out:
834 	/* Some of the error status returned by FW may not be from errno.h */
835 	if (rc > 0)
836 		rc = -EIO;
837 
838 	return rc;
839 }
840 
841 static int bnxt_shutdown_nic(struct bnxt *bp)
842 {
843 	bnxt_free_all_hwrm_resources(bp);
844 	bnxt_free_all_filters(bp);
845 	bnxt_free_all_vnics(bp);
846 	return 0;
847 }
848 
849 /*
850  * Device configuration and status function
851  */
852 
853 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
854 {
855 	uint32_t link_speed = bp->link_info->support_speeds;
856 	uint32_t speed_capa = 0;
857 
858 	/* If PAM4 is configured, use PAM4 supported speed */
859 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
860 		link_speed = bp->link_info->support_pam4_speeds;
861 
862 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
863 		speed_capa |= ETH_LINK_SPEED_100M;
864 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
865 		speed_capa |= ETH_LINK_SPEED_100M_HD;
866 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
867 		speed_capa |= ETH_LINK_SPEED_1G;
868 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
869 		speed_capa |= ETH_LINK_SPEED_2_5G;
870 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
871 		speed_capa |= ETH_LINK_SPEED_10G;
872 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
873 		speed_capa |= ETH_LINK_SPEED_20G;
874 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
875 		speed_capa |= ETH_LINK_SPEED_25G;
876 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
877 		speed_capa |= ETH_LINK_SPEED_40G;
878 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
879 		speed_capa |= ETH_LINK_SPEED_50G;
880 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
881 		speed_capa |= ETH_LINK_SPEED_100G;
882 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
883 		speed_capa |= ETH_LINK_SPEED_50G;
884 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
885 		speed_capa |= ETH_LINK_SPEED_100G;
886 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
887 		speed_capa |= ETH_LINK_SPEED_200G;
888 
889 	if (bp->link_info->auto_mode ==
890 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
891 		speed_capa |= ETH_LINK_SPEED_FIXED;
892 	else
893 		speed_capa |= ETH_LINK_SPEED_AUTONEG;
894 
895 	return speed_capa;
896 }
897 
898 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
899 				struct rte_eth_dev_info *dev_info)
900 {
901 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
902 	struct bnxt *bp = eth_dev->data->dev_private;
903 	uint16_t max_vnics, i, j, vpool, vrxq;
904 	unsigned int max_rx_rings;
905 	int rc;
906 
907 	rc = is_bnxt_in_error(bp);
908 	if (rc)
909 		return rc;
910 
911 	/* MAC Specifics */
912 	dev_info->max_mac_addrs = bp->max_l2_ctx;
913 	dev_info->max_hash_mac_addrs = 0;
914 
915 	/* PF/VF specifics */
916 	if (BNXT_PF(bp))
917 		dev_info->max_vfs = pdev->max_vfs;
918 
919 	max_rx_rings = bnxt_max_rings(bp);
920 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
921 	dev_info->max_rx_queues = max_rx_rings;
922 	dev_info->max_tx_queues = max_rx_rings;
923 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
924 	dev_info->hash_key_size = 40;
925 	max_vnics = bp->max_vnics;
926 
927 	/* MTU specifics */
928 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
929 	dev_info->max_mtu = BNXT_MAX_MTU;
930 
931 	/* Fast path specifics */
932 	dev_info->min_rx_bufsize = 1;
933 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
934 
935 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
936 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
937 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
938 	dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
939 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
940 				    dev_info->tx_queue_offload_capa;
941 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
942 
943 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
944 
945 	/* *INDENT-OFF* */
946 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
947 		.rx_thresh = {
948 			.pthresh = 8,
949 			.hthresh = 8,
950 			.wthresh = 0,
951 		},
952 		.rx_free_thresh = 32,
953 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
954 	};
955 
956 	dev_info->default_txconf = (struct rte_eth_txconf) {
957 		.tx_thresh = {
958 			.pthresh = 32,
959 			.hthresh = 0,
960 			.wthresh = 0,
961 		},
962 		.tx_free_thresh = 32,
963 		.tx_rs_thresh = 32,
964 	};
965 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
966 
967 	eth_dev->data->dev_conf.intr_conf.rxq = 1;
968 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
969 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
970 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
971 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
972 
973 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
974 		dev_info->switch_info.name = eth_dev->device->name;
975 		dev_info->switch_info.domain_id = bp->switch_domain_id;
976 		dev_info->switch_info.port_id =
977 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
978 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
979 	}
980 
981 	/* *INDENT-ON* */
982 
983 	/*
984 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
985 	 *       need further investigation.
986 	 */
987 
988 	/* VMDq resources */
989 	vpool = 64; /* ETH_64_POOLS */
990 	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
991 	for (i = 0; i < 4; vpool >>= 1, i++) {
992 		if (max_vnics > vpool) {
993 			for (j = 0; j < 5; vrxq >>= 1, j++) {
994 				if (dev_info->max_rx_queues > vrxq) {
995 					if (vpool > vrxq)
996 						vpool = vrxq;
997 					goto found;
998 				}
999 			}
1000 			/* Not enough resources to support VMDq */
1001 			break;
1002 		}
1003 	}
1004 	/* Not enough resources to support VMDq */
1005 	vpool = 0;
1006 	vrxq = 0;
1007 found:
1008 	dev_info->max_vmdq_pools = vpool;
1009 	dev_info->vmdq_queue_num = vrxq;
1010 
1011 	dev_info->vmdq_pool_base = 0;
1012 	dev_info->vmdq_queue_base = 0;
1013 
1014 	return 0;
1015 }
1016 
1017 /* Configure the device based on the configuration provided */
1018 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1019 {
1020 	struct bnxt *bp = eth_dev->data->dev_private;
1021 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1022 	int rc;
1023 
1024 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1025 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1026 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1027 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1028 
1029 	rc = is_bnxt_in_error(bp);
1030 	if (rc)
1031 		return rc;
1032 
1033 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1034 		rc = bnxt_hwrm_check_vf_rings(bp);
1035 		if (rc) {
1036 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1037 			return -ENOSPC;
1038 		}
1039 
1040 		/* If a resource has already been allocated - in this case
1041 		 * it is the async completion ring, free it. Reallocate it after
1042 		 * resource reservation. This will ensure the resource counts
1043 		 * are calculated correctly.
1044 		 */
1045 
1046 		pthread_mutex_lock(&bp->def_cp_lock);
1047 
1048 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1049 			bnxt_disable_int(bp);
1050 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1051 		}
1052 
1053 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1054 		if (rc) {
1055 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1056 			pthread_mutex_unlock(&bp->def_cp_lock);
1057 			return -ENOSPC;
1058 		}
1059 
1060 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1061 			rc = bnxt_alloc_async_cp_ring(bp);
1062 			if (rc) {
1063 				pthread_mutex_unlock(&bp->def_cp_lock);
1064 				return rc;
1065 			}
1066 			bnxt_enable_int(bp);
1067 		}
1068 
1069 		pthread_mutex_unlock(&bp->def_cp_lock);
1070 	}
1071 
1072 	/* Inherit new configurations */
1073 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1074 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1075 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1076 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1077 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1078 	    bp->max_stat_ctx)
1079 		goto resource_error;
1080 
1081 	if (BNXT_HAS_RING_GRPS(bp) &&
1082 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1083 		goto resource_error;
1084 
1085 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1086 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1087 		goto resource_error;
1088 
1089 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1090 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1091 
1092 	if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1093 		rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1094 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1095 
1096 	if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1097 		eth_dev->data->mtu =
1098 			eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1099 			RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1100 			BNXT_NUM_VLANS;
1101 		bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1102 	}
1103 	return 0;
1104 
1105 resource_error:
1106 	PMD_DRV_LOG(ERR,
1107 		    "Insufficient resources to support requested config\n");
1108 	PMD_DRV_LOG(ERR,
1109 		    "Num Queues Requested: Tx %d, Rx %d\n",
1110 		    eth_dev->data->nb_tx_queues,
1111 		    eth_dev->data->nb_rx_queues);
1112 	PMD_DRV_LOG(ERR,
1113 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1114 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1115 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1116 	return -ENOSPC;
1117 }
1118 
1119 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1120 {
1121 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1122 
1123 	if (link->link_status)
1124 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1125 			eth_dev->data->port_id,
1126 			(uint32_t)link->link_speed,
1127 			(link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1128 			("full-duplex") : ("half-duplex\n"));
1129 	else
1130 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1131 			eth_dev->data->port_id);
1132 }
1133 
1134 /*
1135  * Determine whether the current configuration requires support for scattered
1136  * receive; return 1 if scattered receive is required and 0 if not.
1137  */
1138 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1139 {
1140 	uint16_t buf_size;
1141 	int i;
1142 
1143 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1144 		return 1;
1145 
1146 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1147 		return 1;
1148 
1149 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1150 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1151 
1152 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1153 				      RTE_PKTMBUF_HEADROOM);
1154 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1155 			return 1;
1156 	}
1157 	return 0;
1158 }
1159 
1160 static eth_rx_burst_t
1161 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1162 {
1163 	struct bnxt *bp = eth_dev->data->dev_private;
1164 
1165 	/* Disable vector mode RX for Stingray2 for now */
1166 	if (BNXT_CHIP_SR2(bp)) {
1167 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1168 		return bnxt_recv_pkts;
1169 	}
1170 
1171 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1172 #ifndef RTE_LIBRTE_IEEE1588
1173 	/*
1174 	 * Vector mode receive can be enabled only if scatter rx is not
1175 	 * in use and rx offloads are limited to VLAN stripping and
1176 	 * CRC stripping.
1177 	 */
1178 	if (!eth_dev->data->scattered_rx &&
1179 	    !(eth_dev->data->dev_conf.rxmode.offloads &
1180 	      ~(DEV_RX_OFFLOAD_VLAN_STRIP |
1181 		DEV_RX_OFFLOAD_KEEP_CRC |
1182 		DEV_RX_OFFLOAD_JUMBO_FRAME |
1183 		DEV_RX_OFFLOAD_IPV4_CKSUM |
1184 		DEV_RX_OFFLOAD_UDP_CKSUM |
1185 		DEV_RX_OFFLOAD_TCP_CKSUM |
1186 		DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1187 		DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1188 		DEV_RX_OFFLOAD_RSS_HASH |
1189 		DEV_RX_OFFLOAD_VLAN_FILTER)) &&
1190 	    !BNXT_TRUFLOW_EN(bp) && BNXT_NUM_ASYNC_CPR(bp) &&
1191 	    rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1192 		PMD_DRV_LOG(INFO, "Using vector mode receive for port %d\n",
1193 			    eth_dev->data->port_id);
1194 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1195 		return bnxt_recv_pkts_vec;
1196 	}
1197 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1198 		    eth_dev->data->port_id);
1199 	PMD_DRV_LOG(INFO,
1200 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1201 		    eth_dev->data->port_id,
1202 		    eth_dev->data->scattered_rx,
1203 		    eth_dev->data->dev_conf.rxmode.offloads);
1204 #endif
1205 #endif
1206 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1207 	return bnxt_recv_pkts;
1208 }
1209 
1210 static eth_tx_burst_t
1211 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1212 {
1213 	struct bnxt *bp = eth_dev->data->dev_private;
1214 
1215 	/* Disable vector mode TX for Stingray2 for now */
1216 	if (BNXT_CHIP_SR2(bp))
1217 		return bnxt_xmit_pkts;
1218 
1219 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1220 #ifndef RTE_LIBRTE_IEEE1588
1221 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1222 
1223 	/*
1224 	 * Vector mode transmit can be enabled only if not using scatter rx
1225 	 * or tx offloads.
1226 	 */
1227 	if (!eth_dev->data->scattered_rx &&
1228 	    !(offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) &&
1229 	    !BNXT_TRUFLOW_EN(bp) &&
1230 	    rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1231 		PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
1232 			    eth_dev->data->port_id);
1233 		return bnxt_xmit_pkts_vec;
1234 	}
1235 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1236 		    eth_dev->data->port_id);
1237 	PMD_DRV_LOG(INFO,
1238 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1239 		    eth_dev->data->port_id,
1240 		    eth_dev->data->scattered_rx,
1241 		    offloads);
1242 #endif
1243 #endif
1244 	return bnxt_xmit_pkts;
1245 }
1246 
1247 static int bnxt_handle_if_change_status(struct bnxt *bp)
1248 {
1249 	int rc;
1250 
1251 	/* Since fw has undergone a reset and lost all contexts,
1252 	 * set fatal flag to not issue hwrm during cleanup
1253 	 */
1254 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1255 	bnxt_uninit_resources(bp, true);
1256 
1257 	/* clear fatal flag so that re-init happens */
1258 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1259 	rc = bnxt_init_resources(bp, true);
1260 
1261 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1262 
1263 	return rc;
1264 }
1265 
1266 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1267 {
1268 	struct bnxt *bp = eth_dev->data->dev_private;
1269 	int rc = 0;
1270 
1271 	if (!bp->link_info->link_up)
1272 		rc = bnxt_set_hwrm_link_config(bp, true);
1273 	if (!rc)
1274 		eth_dev->data->dev_link.link_status = 1;
1275 
1276 	bnxt_print_link_info(eth_dev);
1277 	return rc;
1278 }
1279 
1280 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1281 {
1282 	struct bnxt *bp = eth_dev->data->dev_private;
1283 
1284 	eth_dev->data->dev_link.link_status = 0;
1285 	bnxt_set_hwrm_link_config(bp, false);
1286 	bp->link_info->link_up = 0;
1287 
1288 	return 0;
1289 }
1290 
1291 static void bnxt_free_switch_domain(struct bnxt *bp)
1292 {
1293 	int rc = 0;
1294 
1295 	if (bp->switch_domain_id) {
1296 		rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1297 		if (rc)
1298 			PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1299 				    bp->switch_domain_id, rc);
1300 	}
1301 }
1302 
1303 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1304 {
1305 	struct bnxt *bp = eth_dev->data->dev_private;
1306 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1307 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1308 	struct rte_eth_link link;
1309 	int ret;
1310 
1311 	eth_dev->data->dev_started = 0;
1312 	eth_dev->data->scattered_rx = 0;
1313 
1314 	/* Prevent crashes when queues are still in use */
1315 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1316 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1317 
1318 	bnxt_disable_int(bp);
1319 
1320 	/* disable uio/vfio intr/eventfd mapping */
1321 	rte_intr_disable(intr_handle);
1322 
1323 	/* Stop the child representors for this device */
1324 	ret = bnxt_rep_stop_all(bp);
1325 	if (ret != 0)
1326 		return ret;
1327 
1328 	/* delete the bnxt ULP port details */
1329 	bnxt_ulp_port_deinit(bp);
1330 
1331 	bnxt_cancel_fw_health_check(bp);
1332 
1333 	/* Do not bring link down during reset recovery */
1334 	if (!is_bnxt_in_error(bp)) {
1335 		bnxt_dev_set_link_down_op(eth_dev);
1336 		/* Wait for link to be reset */
1337 		if (BNXT_SINGLE_PF(bp))
1338 			rte_delay_ms(500);
1339 		/* clear the recorded link status */
1340 		memset(&link, 0, sizeof(link));
1341 		rte_eth_linkstatus_set(eth_dev, &link);
1342 	}
1343 
1344 	/* Clean queue intr-vector mapping */
1345 	rte_intr_efd_disable(intr_handle);
1346 	if (intr_handle->intr_vec != NULL) {
1347 		rte_free(intr_handle->intr_vec);
1348 		intr_handle->intr_vec = NULL;
1349 	}
1350 
1351 	bnxt_hwrm_port_clr_stats(bp);
1352 	bnxt_free_tx_mbufs(bp);
1353 	bnxt_free_rx_mbufs(bp);
1354 	/* Process any remaining notifications in default completion queue */
1355 	bnxt_int_handler(eth_dev);
1356 	bnxt_shutdown_nic(bp);
1357 	bnxt_hwrm_if_change(bp, false);
1358 
1359 	rte_free(bp->mark_table);
1360 	bp->mark_table = NULL;
1361 
1362 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1363 	bp->rx_cosq_cnt = 0;
1364 	/* All filters are deleted on a port stop. */
1365 	if (BNXT_FLOW_XSTATS_EN(bp))
1366 		bp->flow_stat->flow_count = 0;
1367 
1368 	return 0;
1369 }
1370 
1371 /* Unload the driver, release resources */
1372 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1373 {
1374 	struct bnxt *bp = eth_dev->data->dev_private;
1375 
1376 	pthread_mutex_lock(&bp->err_recovery_lock);
1377 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1378 		PMD_DRV_LOG(ERR,
1379 			    "Adapter recovering from error..Please retry\n");
1380 		pthread_mutex_unlock(&bp->err_recovery_lock);
1381 		return -EAGAIN;
1382 	}
1383 	pthread_mutex_unlock(&bp->err_recovery_lock);
1384 
1385 	return bnxt_dev_stop(eth_dev);
1386 }
1387 
1388 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1389 {
1390 	struct bnxt *bp = eth_dev->data->dev_private;
1391 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1392 	int vlan_mask = 0;
1393 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1394 
1395 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1396 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1397 		return -EINVAL;
1398 	}
1399 
1400 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1401 		PMD_DRV_LOG(ERR,
1402 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1403 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1404 
1405 	do {
1406 		rc = bnxt_hwrm_if_change(bp, true);
1407 		if (rc == 0 || rc != -EAGAIN)
1408 			break;
1409 
1410 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1411 	} while (retry_cnt--);
1412 
1413 	if (rc)
1414 		return rc;
1415 
1416 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1417 		rc = bnxt_handle_if_change_status(bp);
1418 		if (rc)
1419 			return rc;
1420 	}
1421 
1422 	bnxt_enable_int(bp);
1423 
1424 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1425 
1426 	rc = bnxt_start_nic(bp);
1427 	if (rc)
1428 		goto error;
1429 
1430 	eth_dev->data->dev_started = 1;
1431 
1432 	bnxt_link_update_op(eth_dev, 1);
1433 
1434 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1435 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1436 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1437 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1438 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1439 	if (rc)
1440 		goto error;
1441 
1442 	/* Initialize bnxt ULP port details */
1443 	rc = bnxt_ulp_port_init(bp);
1444 	if (rc)
1445 		goto error;
1446 
1447 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1448 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1449 
1450 	bnxt_schedule_fw_health_check(bp);
1451 
1452 	return 0;
1453 
1454 error:
1455 	bnxt_dev_stop(eth_dev);
1456 	return rc;
1457 }
1458 
1459 static void
1460 bnxt_uninit_locks(struct bnxt *bp)
1461 {
1462 	pthread_mutex_destroy(&bp->flow_lock);
1463 	pthread_mutex_destroy(&bp->def_cp_lock);
1464 	pthread_mutex_destroy(&bp->health_check_lock);
1465 	pthread_mutex_destroy(&bp->err_recovery_lock);
1466 	if (bp->rep_info) {
1467 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1468 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1469 	}
1470 }
1471 
1472 static void bnxt_drv_uninit(struct bnxt *bp)
1473 {
1474 	bnxt_free_switch_domain(bp);
1475 	bnxt_free_leds_info(bp);
1476 	bnxt_free_cos_queues(bp);
1477 	bnxt_free_link_info(bp);
1478 	bnxt_free_pf_info(bp);
1479 	bnxt_free_parent_info(bp);
1480 	bnxt_uninit_locks(bp);
1481 
1482 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1483 	bp->tx_mem_zone = NULL;
1484 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1485 	bp->rx_mem_zone = NULL;
1486 
1487 	bnxt_hwrm_free_vf_info(bp);
1488 
1489 	rte_free(bp->grp_info);
1490 	bp->grp_info = NULL;
1491 }
1492 
1493 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1494 {
1495 	struct bnxt *bp = eth_dev->data->dev_private;
1496 	int ret = 0;
1497 
1498 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1499 		return 0;
1500 
1501 	pthread_mutex_lock(&bp->err_recovery_lock);
1502 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1503 		PMD_DRV_LOG(ERR,
1504 			    "Adapter recovering from error...Please retry\n");
1505 		pthread_mutex_unlock(&bp->err_recovery_lock);
1506 		return -EAGAIN;
1507 	}
1508 	pthread_mutex_unlock(&bp->err_recovery_lock);
1509 
1510 	/* cancel the recovery handler before remove dev */
1511 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1512 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1513 	bnxt_cancel_fc_thread(bp);
1514 
1515 	if (eth_dev->data->dev_started)
1516 		ret = bnxt_dev_stop(eth_dev);
1517 
1518 	bnxt_uninit_resources(bp, false);
1519 
1520 	bnxt_drv_uninit(bp);
1521 
1522 	return ret;
1523 }
1524 
1525 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1526 				    uint32_t index)
1527 {
1528 	struct bnxt *bp = eth_dev->data->dev_private;
1529 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1530 	struct bnxt_vnic_info *vnic;
1531 	struct bnxt_filter_info *filter, *temp_filter;
1532 	uint32_t i;
1533 
1534 	if (is_bnxt_in_error(bp))
1535 		return;
1536 
1537 	/*
1538 	 * Loop through all VNICs from the specified filter flow pools to
1539 	 * remove the corresponding MAC addr filter
1540 	 */
1541 	for (i = 0; i < bp->nr_vnics; i++) {
1542 		if (!(pool_mask & (1ULL << i)))
1543 			continue;
1544 
1545 		vnic = &bp->vnic_info[i];
1546 		filter = STAILQ_FIRST(&vnic->filter);
1547 		while (filter) {
1548 			temp_filter = STAILQ_NEXT(filter, next);
1549 			if (filter->mac_index == index) {
1550 				STAILQ_REMOVE(&vnic->filter, filter,
1551 						bnxt_filter_info, next);
1552 				bnxt_hwrm_clear_l2_filter(bp, filter);
1553 				bnxt_free_filter(bp, filter);
1554 			}
1555 			filter = temp_filter;
1556 		}
1557 	}
1558 }
1559 
1560 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1561 			       struct rte_ether_addr *mac_addr, uint32_t index,
1562 			       uint32_t pool)
1563 {
1564 	struct bnxt_filter_info *filter;
1565 	int rc = 0;
1566 
1567 	/* Attach requested MAC address to the new l2_filter */
1568 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1569 		if (filter->mac_index == index) {
1570 			PMD_DRV_LOG(DEBUG,
1571 				    "MAC addr already existed for pool %d\n",
1572 				    pool);
1573 			return 0;
1574 		}
1575 	}
1576 
1577 	filter = bnxt_alloc_filter(bp);
1578 	if (!filter) {
1579 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1580 		return -ENODEV;
1581 	}
1582 
1583 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1584 	 * if the MAC that's been programmed now is a different one, then,
1585 	 * copy that addr to filter->l2_addr
1586 	 */
1587 	if (mac_addr)
1588 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1589 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1590 
1591 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1592 	if (!rc) {
1593 		filter->mac_index = index;
1594 		if (filter->mac_index == 0)
1595 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1596 		else
1597 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1598 	} else {
1599 		bnxt_free_filter(bp, filter);
1600 	}
1601 
1602 	return rc;
1603 }
1604 
1605 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1606 				struct rte_ether_addr *mac_addr,
1607 				uint32_t index, uint32_t pool)
1608 {
1609 	struct bnxt *bp = eth_dev->data->dev_private;
1610 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1611 	int rc = 0;
1612 
1613 	rc = is_bnxt_in_error(bp);
1614 	if (rc)
1615 		return rc;
1616 
1617 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1618 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1619 		return -ENOTSUP;
1620 	}
1621 
1622 	if (!vnic) {
1623 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1624 		return -EINVAL;
1625 	}
1626 
1627 	/* Filter settings will get applied when port is started */
1628 	if (!eth_dev->data->dev_started)
1629 		return 0;
1630 
1631 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1632 
1633 	return rc;
1634 }
1635 
1636 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1637 {
1638 	int rc = 0;
1639 	struct bnxt *bp = eth_dev->data->dev_private;
1640 	struct rte_eth_link new;
1641 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1642 			BNXT_MIN_LINK_WAIT_CNT;
1643 
1644 	rc = is_bnxt_in_error(bp);
1645 	if (rc)
1646 		return rc;
1647 
1648 	memset(&new, 0, sizeof(new));
1649 	do {
1650 		/* Retrieve link info from hardware */
1651 		rc = bnxt_get_hwrm_link_config(bp, &new);
1652 		if (rc) {
1653 			new.link_speed = ETH_LINK_SPEED_100M;
1654 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1655 			PMD_DRV_LOG(ERR,
1656 				"Failed to retrieve link rc = 0x%x!\n", rc);
1657 			goto out;
1658 		}
1659 
1660 		if (!wait_to_complete || new.link_status)
1661 			break;
1662 
1663 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1664 	} while (cnt--);
1665 
1666 	/* Only single function PF can bring phy down.
1667 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1668 	 */
1669 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1670 		memset(&new, 0, sizeof(new));
1671 
1672 out:
1673 	/* Timed out or success */
1674 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1675 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1676 		rte_eth_linkstatus_set(eth_dev, &new);
1677 
1678 		rte_eth_dev_callback_process(eth_dev,
1679 					     RTE_ETH_EVENT_INTR_LSC,
1680 					     NULL);
1681 
1682 		bnxt_print_link_info(eth_dev);
1683 	}
1684 
1685 	return rc;
1686 }
1687 
1688 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1689 {
1690 	struct bnxt *bp = eth_dev->data->dev_private;
1691 	struct bnxt_vnic_info *vnic;
1692 	uint32_t old_flags;
1693 	int rc;
1694 
1695 	rc = is_bnxt_in_error(bp);
1696 	if (rc)
1697 		return rc;
1698 
1699 	/* Filter settings will get applied when port is started */
1700 	if (!eth_dev->data->dev_started)
1701 		return 0;
1702 
1703 	if (bp->vnic_info == NULL)
1704 		return 0;
1705 
1706 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1707 
1708 	old_flags = vnic->flags;
1709 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1710 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1711 	if (rc != 0)
1712 		vnic->flags = old_flags;
1713 
1714 	return rc;
1715 }
1716 
1717 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1718 {
1719 	struct bnxt *bp = eth_dev->data->dev_private;
1720 	struct bnxt_vnic_info *vnic;
1721 	uint32_t old_flags;
1722 	int rc;
1723 
1724 	rc = is_bnxt_in_error(bp);
1725 	if (rc)
1726 		return rc;
1727 
1728 	/* Filter settings will get applied when port is started */
1729 	if (!eth_dev->data->dev_started)
1730 		return 0;
1731 
1732 	if (bp->vnic_info == NULL)
1733 		return 0;
1734 
1735 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1736 
1737 	old_flags = vnic->flags;
1738 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1739 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1740 	if (rc != 0)
1741 		vnic->flags = old_flags;
1742 
1743 	return rc;
1744 }
1745 
1746 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1747 {
1748 	struct bnxt *bp = eth_dev->data->dev_private;
1749 	struct bnxt_vnic_info *vnic;
1750 	uint32_t old_flags;
1751 	int rc;
1752 
1753 	rc = is_bnxt_in_error(bp);
1754 	if (rc)
1755 		return rc;
1756 
1757 	/* Filter settings will get applied when port is started */
1758 	if (!eth_dev->data->dev_started)
1759 		return 0;
1760 
1761 	if (bp->vnic_info == NULL)
1762 		return 0;
1763 
1764 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1765 
1766 	old_flags = vnic->flags;
1767 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1768 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1769 	if (rc != 0)
1770 		vnic->flags = old_flags;
1771 
1772 	return rc;
1773 }
1774 
1775 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1776 {
1777 	struct bnxt *bp = eth_dev->data->dev_private;
1778 	struct bnxt_vnic_info *vnic;
1779 	uint32_t old_flags;
1780 	int rc;
1781 
1782 	rc = is_bnxt_in_error(bp);
1783 	if (rc)
1784 		return rc;
1785 
1786 	/* Filter settings will get applied when port is started */
1787 	if (!eth_dev->data->dev_started)
1788 		return 0;
1789 
1790 	if (bp->vnic_info == NULL)
1791 		return 0;
1792 
1793 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1794 
1795 	old_flags = vnic->flags;
1796 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1797 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1798 	if (rc != 0)
1799 		vnic->flags = old_flags;
1800 
1801 	return rc;
1802 }
1803 
1804 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1805 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1806 {
1807 	if (qid >= bp->rx_nr_rings)
1808 		return NULL;
1809 
1810 	return bp->eth_dev->data->rx_queues[qid];
1811 }
1812 
1813 /* Return rxq corresponding to a given rss table ring/group ID. */
1814 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1815 {
1816 	struct bnxt_rx_queue *rxq;
1817 	unsigned int i;
1818 
1819 	if (!BNXT_HAS_RING_GRPS(bp)) {
1820 		for (i = 0; i < bp->rx_nr_rings; i++) {
1821 			rxq = bp->eth_dev->data->rx_queues[i];
1822 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1823 				return rxq->index;
1824 		}
1825 	} else {
1826 		for (i = 0; i < bp->rx_nr_rings; i++) {
1827 			if (bp->grp_info[i].fw_grp_id == fwr)
1828 				return i;
1829 		}
1830 	}
1831 
1832 	return INVALID_HW_RING_ID;
1833 }
1834 
1835 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1836 			    struct rte_eth_rss_reta_entry64 *reta_conf,
1837 			    uint16_t reta_size)
1838 {
1839 	struct bnxt *bp = eth_dev->data->dev_private;
1840 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1841 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1842 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1843 	uint16_t idx, sft;
1844 	int i, rc;
1845 
1846 	rc = is_bnxt_in_error(bp);
1847 	if (rc)
1848 		return rc;
1849 
1850 	if (!vnic->rss_table)
1851 		return -EINVAL;
1852 
1853 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
1854 		return -EINVAL;
1855 
1856 	if (reta_size != tbl_size) {
1857 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1858 			"(%d) must equal the size supported by the hardware "
1859 			"(%d)\n", reta_size, tbl_size);
1860 		return -EINVAL;
1861 	}
1862 
1863 	for (i = 0; i < reta_size; i++) {
1864 		struct bnxt_rx_queue *rxq;
1865 
1866 		idx = i / RTE_RETA_GROUP_SIZE;
1867 		sft = i % RTE_RETA_GROUP_SIZE;
1868 
1869 		if (!(reta_conf[idx].mask & (1ULL << sft)))
1870 			continue;
1871 
1872 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
1873 		if (!rxq) {
1874 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
1875 			return -EINVAL;
1876 		}
1877 
1878 		if (BNXT_CHIP_P5(bp)) {
1879 			vnic->rss_table[i * 2] =
1880 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
1881 			vnic->rss_table[i * 2 + 1] =
1882 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
1883 		} else {
1884 			vnic->rss_table[i] =
1885 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
1886 		}
1887 	}
1888 
1889 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1890 	return rc;
1891 }
1892 
1893 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
1894 			      struct rte_eth_rss_reta_entry64 *reta_conf,
1895 			      uint16_t reta_size)
1896 {
1897 	struct bnxt *bp = eth_dev->data->dev_private;
1898 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1899 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1900 	uint16_t idx, sft, i;
1901 	int rc;
1902 
1903 	rc = is_bnxt_in_error(bp);
1904 	if (rc)
1905 		return rc;
1906 
1907 	/* Retrieve from the default VNIC */
1908 	if (!vnic)
1909 		return -EINVAL;
1910 	if (!vnic->rss_table)
1911 		return -EINVAL;
1912 
1913 	if (reta_size != tbl_size) {
1914 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1915 			"(%d) must equal the size supported by the hardware "
1916 			"(%d)\n", reta_size, tbl_size);
1917 		return -EINVAL;
1918 	}
1919 
1920 	for (idx = 0, i = 0; i < reta_size; i++) {
1921 		idx = i / RTE_RETA_GROUP_SIZE;
1922 		sft = i % RTE_RETA_GROUP_SIZE;
1923 
1924 		if (reta_conf[idx].mask & (1ULL << sft)) {
1925 			uint16_t qid;
1926 
1927 			if (BNXT_CHIP_P5(bp))
1928 				qid = bnxt_rss_to_qid(bp,
1929 						      vnic->rss_table[i * 2]);
1930 			else
1931 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
1932 
1933 			if (qid == INVALID_HW_RING_ID) {
1934 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
1935 				return -EINVAL;
1936 			}
1937 			reta_conf[idx].reta[sft] = qid;
1938 		}
1939 	}
1940 
1941 	return 0;
1942 }
1943 
1944 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
1945 				   struct rte_eth_rss_conf *rss_conf)
1946 {
1947 	struct bnxt *bp = eth_dev->data->dev_private;
1948 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1949 	struct bnxt_vnic_info *vnic;
1950 	int rc;
1951 
1952 	rc = is_bnxt_in_error(bp);
1953 	if (rc)
1954 		return rc;
1955 
1956 	/*
1957 	 * If RSS enablement were different than dev_configure,
1958 	 * then return -EINVAL
1959 	 */
1960 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
1961 		if (!rss_conf->rss_hf)
1962 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
1963 	} else {
1964 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
1965 			return -EINVAL;
1966 	}
1967 
1968 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
1969 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
1970 	       rss_conf,
1971 	       sizeof(*rss_conf));
1972 
1973 	/* Update the default RSS VNIC(s) */
1974 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1975 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
1976 	vnic->hash_mode =
1977 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
1978 					    ETH_RSS_LEVEL(rss_conf->rss_hf));
1979 
1980 	/*
1981 	 * If hashkey is not specified, use the previously configured
1982 	 * hashkey
1983 	 */
1984 	if (!rss_conf->rss_key)
1985 		goto rss_config;
1986 
1987 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
1988 		PMD_DRV_LOG(ERR,
1989 			    "Invalid hashkey length, should be 16 bytes\n");
1990 		return -EINVAL;
1991 	}
1992 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
1993 
1994 rss_config:
1995 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1996 	return rc;
1997 }
1998 
1999 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2000 				     struct rte_eth_rss_conf *rss_conf)
2001 {
2002 	struct bnxt *bp = eth_dev->data->dev_private;
2003 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2004 	int len, rc;
2005 	uint32_t hash_types;
2006 
2007 	rc = is_bnxt_in_error(bp);
2008 	if (rc)
2009 		return rc;
2010 
2011 	/* RSS configuration is the same for all VNICs */
2012 	if (vnic && vnic->rss_hash_key) {
2013 		if (rss_conf->rss_key) {
2014 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2015 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2016 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2017 		}
2018 
2019 		hash_types = vnic->hash_type;
2020 		rss_conf->rss_hf = 0;
2021 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2022 			rss_conf->rss_hf |= ETH_RSS_IPV4;
2023 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2024 		}
2025 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2026 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2027 			hash_types &=
2028 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2029 		}
2030 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2031 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2032 			hash_types &=
2033 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2034 		}
2035 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2036 			rss_conf->rss_hf |= ETH_RSS_IPV6;
2037 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2038 		}
2039 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2040 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2041 			hash_types &=
2042 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2043 		}
2044 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2045 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2046 			hash_types &=
2047 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2048 		}
2049 
2050 		rss_conf->rss_hf |=
2051 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2052 
2053 		if (hash_types) {
2054 			PMD_DRV_LOG(ERR,
2055 				"Unknown RSS config from firmware (%08x), RSS disabled",
2056 				vnic->hash_type);
2057 			return -ENOTSUP;
2058 		}
2059 	} else {
2060 		rss_conf->rss_hf = 0;
2061 	}
2062 	return 0;
2063 }
2064 
2065 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2066 			       struct rte_eth_fc_conf *fc_conf)
2067 {
2068 	struct bnxt *bp = dev->data->dev_private;
2069 	struct rte_eth_link link_info;
2070 	int rc;
2071 
2072 	rc = is_bnxt_in_error(bp);
2073 	if (rc)
2074 		return rc;
2075 
2076 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2077 	if (rc)
2078 		return rc;
2079 
2080 	memset(fc_conf, 0, sizeof(*fc_conf));
2081 	if (bp->link_info->auto_pause)
2082 		fc_conf->autoneg = 1;
2083 	switch (bp->link_info->pause) {
2084 	case 0:
2085 		fc_conf->mode = RTE_FC_NONE;
2086 		break;
2087 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2088 		fc_conf->mode = RTE_FC_TX_PAUSE;
2089 		break;
2090 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2091 		fc_conf->mode = RTE_FC_RX_PAUSE;
2092 		break;
2093 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2094 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2095 		fc_conf->mode = RTE_FC_FULL;
2096 		break;
2097 	}
2098 	return 0;
2099 }
2100 
2101 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2102 			       struct rte_eth_fc_conf *fc_conf)
2103 {
2104 	struct bnxt *bp = dev->data->dev_private;
2105 	int rc;
2106 
2107 	rc = is_bnxt_in_error(bp);
2108 	if (rc)
2109 		return rc;
2110 
2111 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2112 		PMD_DRV_LOG(ERR, "Flow Control Settings cannot be modified\n");
2113 		return -ENOTSUP;
2114 	}
2115 
2116 	switch (fc_conf->mode) {
2117 	case RTE_FC_NONE:
2118 		bp->link_info->auto_pause = 0;
2119 		bp->link_info->force_pause = 0;
2120 		break;
2121 	case RTE_FC_RX_PAUSE:
2122 		if (fc_conf->autoneg) {
2123 			bp->link_info->auto_pause =
2124 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2125 			bp->link_info->force_pause = 0;
2126 		} else {
2127 			bp->link_info->auto_pause = 0;
2128 			bp->link_info->force_pause =
2129 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2130 		}
2131 		break;
2132 	case RTE_FC_TX_PAUSE:
2133 		if (fc_conf->autoneg) {
2134 			bp->link_info->auto_pause =
2135 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2136 			bp->link_info->force_pause = 0;
2137 		} else {
2138 			bp->link_info->auto_pause = 0;
2139 			bp->link_info->force_pause =
2140 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2141 		}
2142 		break;
2143 	case RTE_FC_FULL:
2144 		if (fc_conf->autoneg) {
2145 			bp->link_info->auto_pause =
2146 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2147 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2148 			bp->link_info->force_pause = 0;
2149 		} else {
2150 			bp->link_info->auto_pause = 0;
2151 			bp->link_info->force_pause =
2152 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2153 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2154 		}
2155 		break;
2156 	}
2157 	return bnxt_set_hwrm_link_config(bp, true);
2158 }
2159 
2160 /* Add UDP tunneling port */
2161 static int
2162 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2163 			 struct rte_eth_udp_tunnel *udp_tunnel)
2164 {
2165 	struct bnxt *bp = eth_dev->data->dev_private;
2166 	uint16_t tunnel_type = 0;
2167 	int rc = 0;
2168 
2169 	rc = is_bnxt_in_error(bp);
2170 	if (rc)
2171 		return rc;
2172 
2173 	switch (udp_tunnel->prot_type) {
2174 	case RTE_TUNNEL_TYPE_VXLAN:
2175 		if (bp->vxlan_port_cnt) {
2176 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2177 				udp_tunnel->udp_port);
2178 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2179 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2180 				return -ENOSPC;
2181 			}
2182 			bp->vxlan_port_cnt++;
2183 			return 0;
2184 		}
2185 		tunnel_type =
2186 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2187 		bp->vxlan_port_cnt++;
2188 		break;
2189 	case RTE_TUNNEL_TYPE_GENEVE:
2190 		if (bp->geneve_port_cnt) {
2191 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2192 				udp_tunnel->udp_port);
2193 			if (bp->geneve_port != udp_tunnel->udp_port) {
2194 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2195 				return -ENOSPC;
2196 			}
2197 			bp->geneve_port_cnt++;
2198 			return 0;
2199 		}
2200 		tunnel_type =
2201 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2202 		bp->geneve_port_cnt++;
2203 		break;
2204 	default:
2205 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2206 		return -ENOTSUP;
2207 	}
2208 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2209 					     tunnel_type);
2210 	return rc;
2211 }
2212 
2213 static int
2214 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2215 			 struct rte_eth_udp_tunnel *udp_tunnel)
2216 {
2217 	struct bnxt *bp = eth_dev->data->dev_private;
2218 	uint16_t tunnel_type = 0;
2219 	uint16_t port = 0;
2220 	int rc = 0;
2221 
2222 	rc = is_bnxt_in_error(bp);
2223 	if (rc)
2224 		return rc;
2225 
2226 	switch (udp_tunnel->prot_type) {
2227 	case RTE_TUNNEL_TYPE_VXLAN:
2228 		if (!bp->vxlan_port_cnt) {
2229 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2230 			return -EINVAL;
2231 		}
2232 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2233 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2234 				udp_tunnel->udp_port, bp->vxlan_port);
2235 			return -EINVAL;
2236 		}
2237 		if (--bp->vxlan_port_cnt)
2238 			return 0;
2239 
2240 		tunnel_type =
2241 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2242 		port = bp->vxlan_fw_dst_port_id;
2243 		break;
2244 	case RTE_TUNNEL_TYPE_GENEVE:
2245 		if (!bp->geneve_port_cnt) {
2246 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2247 			return -EINVAL;
2248 		}
2249 		if (bp->geneve_port != udp_tunnel->udp_port) {
2250 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2251 				udp_tunnel->udp_port, bp->geneve_port);
2252 			return -EINVAL;
2253 		}
2254 		if (--bp->geneve_port_cnt)
2255 			return 0;
2256 
2257 		tunnel_type =
2258 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2259 		port = bp->geneve_fw_dst_port_id;
2260 		break;
2261 	default:
2262 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2263 		return -ENOTSUP;
2264 	}
2265 
2266 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2267 	return rc;
2268 }
2269 
2270 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2271 {
2272 	struct bnxt_filter_info *filter;
2273 	struct bnxt_vnic_info *vnic;
2274 	int rc = 0;
2275 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2276 
2277 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2278 	filter = STAILQ_FIRST(&vnic->filter);
2279 	while (filter) {
2280 		/* Search for this matching MAC+VLAN filter */
2281 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2282 			/* Delete the filter */
2283 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2284 			if (rc)
2285 				return rc;
2286 			STAILQ_REMOVE(&vnic->filter, filter,
2287 				      bnxt_filter_info, next);
2288 			bnxt_free_filter(bp, filter);
2289 			PMD_DRV_LOG(INFO,
2290 				    "Deleted vlan filter for %d\n",
2291 				    vlan_id);
2292 			return 0;
2293 		}
2294 		filter = STAILQ_NEXT(filter, next);
2295 	}
2296 	return -ENOENT;
2297 }
2298 
2299 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2300 {
2301 	struct bnxt_filter_info *filter;
2302 	struct bnxt_vnic_info *vnic;
2303 	int rc = 0;
2304 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2305 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2306 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2307 
2308 	/* Implementation notes on the use of VNIC in this command:
2309 	 *
2310 	 * By default, these filters belong to default vnic for the function.
2311 	 * Once these filters are set up, only destination VNIC can be modified.
2312 	 * If the destination VNIC is not specified in this command,
2313 	 * then the HWRM shall only create an l2 context id.
2314 	 */
2315 
2316 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2317 	filter = STAILQ_FIRST(&vnic->filter);
2318 	/* Check if the VLAN has already been added */
2319 	while (filter) {
2320 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2321 			return -EEXIST;
2322 
2323 		filter = STAILQ_NEXT(filter, next);
2324 	}
2325 
2326 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2327 	 * command to create MAC+VLAN filter with the right flags, enables set.
2328 	 */
2329 	filter = bnxt_alloc_filter(bp);
2330 	if (!filter) {
2331 		PMD_DRV_LOG(ERR,
2332 			    "MAC/VLAN filter alloc failed\n");
2333 		return -ENOMEM;
2334 	}
2335 	/* MAC + VLAN ID filter */
2336 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2337 	 * untagged packets are received
2338 	 *
2339 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2340 	 * packets and only the programmed vlan's packets are received
2341 	 */
2342 	filter->l2_ivlan = vlan_id;
2343 	filter->l2_ivlan_mask = 0x0FFF;
2344 	filter->enables |= en;
2345 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2346 
2347 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2348 	if (rc) {
2349 		/* Free the newly allocated filter as we were
2350 		 * not able to create the filter in hardware.
2351 		 */
2352 		bnxt_free_filter(bp, filter);
2353 		return rc;
2354 	}
2355 
2356 	filter->mac_index = 0;
2357 	/* Add this new filter to the list */
2358 	if (vlan_id == 0)
2359 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2360 	else
2361 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2362 
2363 	PMD_DRV_LOG(INFO,
2364 		    "Added Vlan filter for %d\n", vlan_id);
2365 	return rc;
2366 }
2367 
2368 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2369 		uint16_t vlan_id, int on)
2370 {
2371 	struct bnxt *bp = eth_dev->data->dev_private;
2372 	int rc;
2373 
2374 	rc = is_bnxt_in_error(bp);
2375 	if (rc)
2376 		return rc;
2377 
2378 	if (!eth_dev->data->dev_started) {
2379 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2380 		return -EINVAL;
2381 	}
2382 
2383 	/* These operations apply to ALL existing MAC/VLAN filters */
2384 	if (on)
2385 		return bnxt_add_vlan_filter(bp, vlan_id);
2386 	else
2387 		return bnxt_del_vlan_filter(bp, vlan_id);
2388 }
2389 
2390 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2391 				    struct bnxt_vnic_info *vnic)
2392 {
2393 	struct bnxt_filter_info *filter;
2394 	int rc;
2395 
2396 	filter = STAILQ_FIRST(&vnic->filter);
2397 	while (filter) {
2398 		if (filter->mac_index == 0 &&
2399 		    !memcmp(filter->l2_addr, bp->mac_addr,
2400 			    RTE_ETHER_ADDR_LEN)) {
2401 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2402 			if (!rc) {
2403 				STAILQ_REMOVE(&vnic->filter, filter,
2404 					      bnxt_filter_info, next);
2405 				bnxt_free_filter(bp, filter);
2406 			}
2407 			return rc;
2408 		}
2409 		filter = STAILQ_NEXT(filter, next);
2410 	}
2411 	return 0;
2412 }
2413 
2414 static int
2415 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2416 {
2417 	struct bnxt_vnic_info *vnic;
2418 	unsigned int i;
2419 	int rc;
2420 
2421 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2422 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2423 		/* Remove any VLAN filters programmed */
2424 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2425 			bnxt_del_vlan_filter(bp, i);
2426 
2427 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2428 		if (rc)
2429 			return rc;
2430 	} else {
2431 		/* Default filter will allow packets that match the
2432 		 * dest mac. So, it has to be deleted, otherwise, we
2433 		 * will endup receiving vlan packets for which the
2434 		 * filter is not programmed, when hw-vlan-filter
2435 		 * configuration is ON
2436 		 */
2437 		bnxt_del_dflt_mac_filter(bp, vnic);
2438 		/* This filter will allow only untagged packets */
2439 		bnxt_add_vlan_filter(bp, 0);
2440 	}
2441 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2442 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2443 
2444 	return 0;
2445 }
2446 
2447 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2448 {
2449 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2450 	unsigned int i;
2451 	int rc;
2452 
2453 	/* Destroy vnic filters and vnic */
2454 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2455 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2456 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2457 			bnxt_del_vlan_filter(bp, i);
2458 	}
2459 	bnxt_del_dflt_mac_filter(bp, vnic);
2460 
2461 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2462 	if (rc)
2463 		return rc;
2464 
2465 	rte_free(vnic->fw_grp_ids);
2466 	vnic->fw_grp_ids = NULL;
2467 
2468 	vnic->rx_queue_cnt = 0;
2469 
2470 	return 0;
2471 }
2472 
2473 static int
2474 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2475 {
2476 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2477 	int rc;
2478 
2479 	/* Destroy, recreate and reconfigure the default vnic */
2480 	rc = bnxt_free_one_vnic(bp, 0);
2481 	if (rc)
2482 		return rc;
2483 
2484 	/* default vnic 0 */
2485 	rc = bnxt_setup_one_vnic(bp, 0);
2486 	if (rc)
2487 		return rc;
2488 
2489 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2490 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2491 		rc = bnxt_add_vlan_filter(bp, 0);
2492 		if (rc)
2493 			return rc;
2494 		rc = bnxt_restore_vlan_filters(bp);
2495 		if (rc)
2496 			return rc;
2497 	} else {
2498 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2499 		if (rc)
2500 			return rc;
2501 	}
2502 
2503 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2504 	if (rc)
2505 		return rc;
2506 
2507 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2508 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2509 
2510 	return rc;
2511 }
2512 
2513 static int
2514 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2515 {
2516 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2517 	struct bnxt *bp = dev->data->dev_private;
2518 	int rc;
2519 
2520 	rc = is_bnxt_in_error(bp);
2521 	if (rc)
2522 		return rc;
2523 
2524 	/* Filter settings will get applied when port is started */
2525 	if (!dev->data->dev_started)
2526 		return 0;
2527 
2528 	if (mask & ETH_VLAN_FILTER_MASK) {
2529 		/* Enable or disable VLAN filtering */
2530 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2531 		if (rc)
2532 			return rc;
2533 	}
2534 
2535 	if (mask & ETH_VLAN_STRIP_MASK) {
2536 		/* Enable or disable VLAN stripping */
2537 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2538 		if (rc)
2539 			return rc;
2540 	}
2541 
2542 	if (mask & ETH_VLAN_EXTEND_MASK) {
2543 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2544 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2545 		else
2546 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2547 	}
2548 
2549 	return 0;
2550 }
2551 
2552 static int
2553 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2554 		      uint16_t tpid)
2555 {
2556 	struct bnxt *bp = dev->data->dev_private;
2557 	int qinq = dev->data->dev_conf.rxmode.offloads &
2558 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2559 
2560 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2561 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2562 		PMD_DRV_LOG(ERR,
2563 			    "Unsupported vlan type.");
2564 		return -EINVAL;
2565 	}
2566 	if (!qinq) {
2567 		PMD_DRV_LOG(ERR,
2568 			    "QinQ not enabled. Needs to be ON as we can "
2569 			    "accelerate only outer vlan\n");
2570 		return -EINVAL;
2571 	}
2572 
2573 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2574 		switch (tpid) {
2575 		case RTE_ETHER_TYPE_QINQ:
2576 			bp->outer_tpid_bd =
2577 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2578 				break;
2579 		case RTE_ETHER_TYPE_VLAN:
2580 			bp->outer_tpid_bd =
2581 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2582 				break;
2583 		case RTE_ETHER_TYPE_QINQ1:
2584 			bp->outer_tpid_bd =
2585 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2586 				break;
2587 		case RTE_ETHER_TYPE_QINQ2:
2588 			bp->outer_tpid_bd =
2589 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2590 				break;
2591 		case RTE_ETHER_TYPE_QINQ3:
2592 			bp->outer_tpid_bd =
2593 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2594 				break;
2595 		default:
2596 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2597 			return -EINVAL;
2598 		}
2599 		bp->outer_tpid_bd |= tpid;
2600 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2601 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2602 		PMD_DRV_LOG(ERR,
2603 			    "Can accelerate only outer vlan in QinQ\n");
2604 		return -EINVAL;
2605 	}
2606 
2607 	return 0;
2608 }
2609 
2610 static int
2611 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2612 			     struct rte_ether_addr *addr)
2613 {
2614 	struct bnxt *bp = dev->data->dev_private;
2615 	/* Default Filter is tied to VNIC 0 */
2616 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2617 	int rc;
2618 
2619 	rc = is_bnxt_in_error(bp);
2620 	if (rc)
2621 		return rc;
2622 
2623 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2624 		return -EPERM;
2625 
2626 	if (rte_is_zero_ether_addr(addr))
2627 		return -EINVAL;
2628 
2629 	/* Filter settings will get applied when port is started */
2630 	if (!dev->data->dev_started)
2631 		return 0;
2632 
2633 	/* Check if the requested MAC is already added */
2634 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2635 		return 0;
2636 
2637 	/* Destroy filter and re-create it */
2638 	bnxt_del_dflt_mac_filter(bp, vnic);
2639 
2640 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2641 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2642 		/* This filter will allow only untagged packets */
2643 		rc = bnxt_add_vlan_filter(bp, 0);
2644 	} else {
2645 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2646 	}
2647 
2648 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2649 	return rc;
2650 }
2651 
2652 static int
2653 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2654 			  struct rte_ether_addr *mc_addr_set,
2655 			  uint32_t nb_mc_addr)
2656 {
2657 	struct bnxt *bp = eth_dev->data->dev_private;
2658 	char *mc_addr_list = (char *)mc_addr_set;
2659 	struct bnxt_vnic_info *vnic;
2660 	uint32_t off = 0, i = 0;
2661 	int rc;
2662 
2663 	rc = is_bnxt_in_error(bp);
2664 	if (rc)
2665 		return rc;
2666 
2667 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2668 
2669 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2670 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2671 		goto allmulti;
2672 	}
2673 
2674 	/* TODO Check for Duplicate mcast addresses */
2675 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2676 	for (i = 0; i < nb_mc_addr; i++) {
2677 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2678 			RTE_ETHER_ADDR_LEN);
2679 		off += RTE_ETHER_ADDR_LEN;
2680 	}
2681 
2682 	vnic->mc_addr_cnt = i;
2683 	if (vnic->mc_addr_cnt)
2684 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2685 	else
2686 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2687 
2688 allmulti:
2689 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2690 }
2691 
2692 static int
2693 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2694 {
2695 	struct bnxt *bp = dev->data->dev_private;
2696 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2697 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2698 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2699 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2700 	int ret;
2701 
2702 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2703 			fw_major, fw_minor, fw_updt, fw_rsvd);
2704 
2705 	ret += 1; /* add the size of '\0' */
2706 	if (fw_size < (uint32_t)ret)
2707 		return ret;
2708 	else
2709 		return 0;
2710 }
2711 
2712 static void
2713 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2714 	struct rte_eth_rxq_info *qinfo)
2715 {
2716 	struct bnxt *bp = dev->data->dev_private;
2717 	struct bnxt_rx_queue *rxq;
2718 
2719 	if (is_bnxt_in_error(bp))
2720 		return;
2721 
2722 	rxq = dev->data->rx_queues[queue_id];
2723 
2724 	qinfo->mp = rxq->mb_pool;
2725 	qinfo->scattered_rx = dev->data->scattered_rx;
2726 	qinfo->nb_desc = rxq->nb_rx_desc;
2727 
2728 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2729 	qinfo->conf.rx_drop_en = rxq->drop_en;
2730 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2731 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2732 }
2733 
2734 static void
2735 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2736 	struct rte_eth_txq_info *qinfo)
2737 {
2738 	struct bnxt *bp = dev->data->dev_private;
2739 	struct bnxt_tx_queue *txq;
2740 
2741 	if (is_bnxt_in_error(bp))
2742 		return;
2743 
2744 	txq = dev->data->tx_queues[queue_id];
2745 
2746 	qinfo->nb_desc = txq->nb_tx_desc;
2747 
2748 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2749 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2750 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2751 
2752 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2753 	qinfo->conf.tx_rs_thresh = 0;
2754 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2755 	qinfo->conf.offloads = txq->offloads;
2756 }
2757 
2758 static const struct {
2759 	eth_rx_burst_t pkt_burst;
2760 	const char *info;
2761 } bnxt_rx_burst_info[] = {
2762 	{bnxt_recv_pkts,	"Scalar"},
2763 #if defined(RTE_ARCH_X86)
2764 	{bnxt_recv_pkts_vec,	"Vector SSE"},
2765 #elif defined(RTE_ARCH_ARM64)
2766 	{bnxt_recv_pkts_vec,	"Vector Neon"},
2767 #endif
2768 };
2769 
2770 static int
2771 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2772 		       struct rte_eth_burst_mode *mode)
2773 {
2774 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2775 	size_t i;
2776 
2777 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2778 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2779 			snprintf(mode->info, sizeof(mode->info), "%s",
2780 				 bnxt_rx_burst_info[i].info);
2781 			return 0;
2782 		}
2783 	}
2784 
2785 	return -EINVAL;
2786 }
2787 
2788 static const struct {
2789 	eth_tx_burst_t pkt_burst;
2790 	const char *info;
2791 } bnxt_tx_burst_info[] = {
2792 	{bnxt_xmit_pkts,	"Scalar"},
2793 #if defined(RTE_ARCH_X86)
2794 	{bnxt_xmit_pkts_vec,	"Vector SSE"},
2795 #elif defined(RTE_ARCH_ARM64)
2796 	{bnxt_xmit_pkts_vec,	"Vector Neon"},
2797 #endif
2798 };
2799 
2800 static int
2801 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2802 		       struct rte_eth_burst_mode *mode)
2803 {
2804 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2805 	size_t i;
2806 
2807 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2808 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2809 			snprintf(mode->info, sizeof(mode->info), "%s",
2810 				 bnxt_tx_burst_info[i].info);
2811 			return 0;
2812 		}
2813 	}
2814 
2815 	return -EINVAL;
2816 }
2817 
2818 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2819 {
2820 	struct bnxt *bp = eth_dev->data->dev_private;
2821 	uint32_t new_pkt_size;
2822 	uint32_t rc = 0;
2823 	uint32_t i;
2824 
2825 	rc = is_bnxt_in_error(bp);
2826 	if (rc)
2827 		return rc;
2828 
2829 	/* Exit if receive queues are not configured yet */
2830 	if (!eth_dev->data->nb_rx_queues)
2831 		return rc;
2832 
2833 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2834 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2835 
2836 	/*
2837 	 * Disallow any MTU change that would require scattered receive support
2838 	 * if it is not already enabled.
2839 	 */
2840 	if (eth_dev->data->dev_started &&
2841 	    !eth_dev->data->scattered_rx &&
2842 	    (new_pkt_size >
2843 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2844 		PMD_DRV_LOG(ERR,
2845 			    "MTU change would require scattered rx support. ");
2846 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2847 		return -EINVAL;
2848 	}
2849 
2850 	if (new_mtu > RTE_ETHER_MTU) {
2851 		bp->flags |= BNXT_FLAG_JUMBO;
2852 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
2853 			DEV_RX_OFFLOAD_JUMBO_FRAME;
2854 	} else {
2855 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
2856 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
2857 		bp->flags &= ~BNXT_FLAG_JUMBO;
2858 	}
2859 
2860 	/* Is there a change in mtu setting? */
2861 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2862 		return rc;
2863 
2864 	for (i = 0; i < bp->nr_vnics; i++) {
2865 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2866 		uint16_t size = 0;
2867 
2868 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
2869 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2870 		if (rc)
2871 			break;
2872 
2873 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2874 		size -= RTE_PKTMBUF_HEADROOM;
2875 
2876 		if (size < new_mtu) {
2877 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2878 			if (rc)
2879 				return rc;
2880 		}
2881 	}
2882 
2883 	if (!rc)
2884 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2885 
2886 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2887 
2888 	return rc;
2889 }
2890 
2891 static int
2892 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2893 {
2894 	struct bnxt *bp = dev->data->dev_private;
2895 	uint16_t vlan = bp->vlan;
2896 	int rc;
2897 
2898 	rc = is_bnxt_in_error(bp);
2899 	if (rc)
2900 		return rc;
2901 
2902 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2903 		PMD_DRV_LOG(ERR,
2904 			"PVID cannot be modified for this function\n");
2905 		return -ENOTSUP;
2906 	}
2907 	bp->vlan = on ? pvid : 0;
2908 
2909 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
2910 	if (rc)
2911 		bp->vlan = vlan;
2912 	return rc;
2913 }
2914 
2915 static int
2916 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
2917 {
2918 	struct bnxt *bp = dev->data->dev_private;
2919 	int rc;
2920 
2921 	rc = is_bnxt_in_error(bp);
2922 	if (rc)
2923 		return rc;
2924 
2925 	return bnxt_hwrm_port_led_cfg(bp, true);
2926 }
2927 
2928 static int
2929 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
2930 {
2931 	struct bnxt *bp = dev->data->dev_private;
2932 	int rc;
2933 
2934 	rc = is_bnxt_in_error(bp);
2935 	if (rc)
2936 		return rc;
2937 
2938 	return bnxt_hwrm_port_led_cfg(bp, false);
2939 }
2940 
2941 static uint32_t
2942 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
2943 {
2944 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
2945 	struct bnxt_cp_ring_info *cpr;
2946 	uint32_t desc = 0, raw_cons;
2947 	struct bnxt_rx_queue *rxq;
2948 	struct rx_pkt_cmpl *rxcmp;
2949 	int rc;
2950 
2951 	rc = is_bnxt_in_error(bp);
2952 	if (rc)
2953 		return rc;
2954 
2955 	rxq = dev->data->rx_queues[rx_queue_id];
2956 	cpr = rxq->cp_ring;
2957 	raw_cons = cpr->cp_raw_cons;
2958 
2959 	while (1) {
2960 		uint32_t agg_cnt, cons, cmpl_type;
2961 
2962 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
2963 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2964 
2965 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
2966 			break;
2967 
2968 		cmpl_type = CMP_TYPE(rxcmp);
2969 
2970 		switch (cmpl_type) {
2971 		case CMPL_BASE_TYPE_RX_L2:
2972 		case CMPL_BASE_TYPE_RX_L2_V2:
2973 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
2974 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
2975 			desc++;
2976 			break;
2977 
2978 		case CMPL_BASE_TYPE_RX_TPA_END:
2979 			if (BNXT_CHIP_P5(rxq->bp)) {
2980 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
2981 
2982 				p5_tpa_end = (void *)rxcmp;
2983 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
2984 			} else {
2985 				struct rx_tpa_end_cmpl *tpa_end;
2986 
2987 				tpa_end = (void *)rxcmp;
2988 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
2989 			}
2990 
2991 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
2992 			desc++;
2993 			break;
2994 
2995 		default:
2996 			raw_cons += CMP_LEN(cmpl_type);
2997 		}
2998 	}
2999 
3000 	return desc;
3001 }
3002 
3003 static int
3004 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3005 {
3006 	struct bnxt_rx_queue *rxq = (struct bnxt_rx_queue *)rx_queue;
3007 	struct bnxt_rx_ring_info *rxr;
3008 	struct bnxt_cp_ring_info *cpr;
3009 	struct rte_mbuf *rx_buf;
3010 	struct rx_pkt_cmpl *rxcmp;
3011 	uint32_t cons, cp_cons;
3012 	int rc;
3013 
3014 	if (!rxq)
3015 		return -EINVAL;
3016 
3017 	rc = is_bnxt_in_error(rxq->bp);
3018 	if (rc)
3019 		return rc;
3020 
3021 	cpr = rxq->cp_ring;
3022 	rxr = rxq->rx_ring;
3023 
3024 	if (offset >= rxq->nb_rx_desc)
3025 		return -EINVAL;
3026 
3027 	cons = RING_CMP(cpr->cp_ring_struct, offset);
3028 	cp_cons = cpr->cp_raw_cons;
3029 	rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3030 
3031 	if (cons > cp_cons) {
3032 		if (CMPL_VALID(rxcmp, cpr->valid))
3033 			return RTE_ETH_RX_DESC_DONE;
3034 	} else {
3035 		if (CMPL_VALID(rxcmp, !cpr->valid))
3036 			return RTE_ETH_RX_DESC_DONE;
3037 	}
3038 	rx_buf = rxr->rx_buf_ring[cons];
3039 	if (rx_buf == NULL || rx_buf == &rxq->fake_mbuf)
3040 		return RTE_ETH_RX_DESC_UNAVAIL;
3041 
3042 
3043 	return RTE_ETH_RX_DESC_AVAIL;
3044 }
3045 
3046 static int
3047 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3048 {
3049 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3050 	struct bnxt_tx_ring_info *txr;
3051 	struct bnxt_cp_ring_info *cpr;
3052 	struct bnxt_sw_tx_bd *tx_buf;
3053 	struct tx_pkt_cmpl *txcmp;
3054 	uint32_t cons, cp_cons;
3055 	int rc;
3056 
3057 	if (!txq)
3058 		return -EINVAL;
3059 
3060 	rc = is_bnxt_in_error(txq->bp);
3061 	if (rc)
3062 		return rc;
3063 
3064 	cpr = txq->cp_ring;
3065 	txr = txq->tx_ring;
3066 
3067 	if (offset >= txq->nb_tx_desc)
3068 		return -EINVAL;
3069 
3070 	cons = RING_CMP(cpr->cp_ring_struct, offset);
3071 	txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3072 	cp_cons = cpr->cp_raw_cons;
3073 
3074 	if (cons > cp_cons) {
3075 		if (CMPL_VALID(txcmp, cpr->valid))
3076 			return RTE_ETH_TX_DESC_UNAVAIL;
3077 	} else {
3078 		if (CMPL_VALID(txcmp, !cpr->valid))
3079 			return RTE_ETH_TX_DESC_UNAVAIL;
3080 	}
3081 	tx_buf = &txr->tx_buf_ring[cons];
3082 	if (tx_buf->mbuf == NULL)
3083 		return RTE_ETH_TX_DESC_DONE;
3084 
3085 	return RTE_ETH_TX_DESC_FULL;
3086 }
3087 
3088 int
3089 bnxt_filter_ctrl_op(struct rte_eth_dev *dev,
3090 		    enum rte_filter_type filter_type,
3091 		    enum rte_filter_op filter_op, void *arg)
3092 {
3093 	struct bnxt *bp = dev->data->dev_private;
3094 	int ret = 0;
3095 
3096 	if (!bp)
3097 		return -EIO;
3098 
3099 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3100 		struct bnxt_representor *vfr = dev->data->dev_private;
3101 		bp = vfr->parent_dev->data->dev_private;
3102 		/* parent is deleted while children are still valid */
3103 		if (!bp) {
3104 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error %d:%d\n",
3105 				    dev->data->port_id,
3106 				    filter_type,
3107 				    filter_op);
3108 			return -EIO;
3109 		}
3110 	}
3111 
3112 	ret = is_bnxt_in_error(bp);
3113 	if (ret)
3114 		return ret;
3115 
3116 	switch (filter_type) {
3117 	case RTE_ETH_FILTER_GENERIC:
3118 		if (filter_op != RTE_ETH_FILTER_GET)
3119 			return -EINVAL;
3120 
3121 		/* PMD supports thread-safe flow operations.  rte_flow API
3122 		 * functions can avoid mutex for multi-thread safety.
3123 		 */
3124 		dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3125 
3126 		if (BNXT_TRUFLOW_EN(bp))
3127 			*(const void **)arg = &bnxt_ulp_rte_flow_ops;
3128 		else
3129 			*(const void **)arg = &bnxt_flow_ops;
3130 		break;
3131 	default:
3132 		PMD_DRV_LOG(ERR,
3133 			"Filter type (%d) not supported", filter_type);
3134 		ret = -EINVAL;
3135 		break;
3136 	}
3137 	return ret;
3138 }
3139 
3140 static const uint32_t *
3141 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3142 {
3143 	static const uint32_t ptypes[] = {
3144 		RTE_PTYPE_L2_ETHER_VLAN,
3145 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3146 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3147 		RTE_PTYPE_L4_ICMP,
3148 		RTE_PTYPE_L4_TCP,
3149 		RTE_PTYPE_L4_UDP,
3150 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3151 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3152 		RTE_PTYPE_INNER_L4_ICMP,
3153 		RTE_PTYPE_INNER_L4_TCP,
3154 		RTE_PTYPE_INNER_L4_UDP,
3155 		RTE_PTYPE_UNKNOWN
3156 	};
3157 
3158 	if (!dev->rx_pkt_burst)
3159 		return NULL;
3160 
3161 	return ptypes;
3162 }
3163 
3164 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3165 			 int reg_win)
3166 {
3167 	uint32_t reg_base = *reg_arr & 0xfffff000;
3168 	uint32_t win_off;
3169 	int i;
3170 
3171 	for (i = 0; i < count; i++) {
3172 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3173 			return -ERANGE;
3174 	}
3175 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3176 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3177 	return 0;
3178 }
3179 
3180 static int bnxt_map_ptp_regs(struct bnxt *bp)
3181 {
3182 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3183 	uint32_t *reg_arr;
3184 	int rc, i;
3185 
3186 	reg_arr = ptp->rx_regs;
3187 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3188 	if (rc)
3189 		return rc;
3190 
3191 	reg_arr = ptp->tx_regs;
3192 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3193 	if (rc)
3194 		return rc;
3195 
3196 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3197 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3198 
3199 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3200 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3201 
3202 	return 0;
3203 }
3204 
3205 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3206 {
3207 	rte_write32(0, (uint8_t *)bp->bar0 +
3208 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3209 	rte_write32(0, (uint8_t *)bp->bar0 +
3210 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3211 }
3212 
3213 static uint64_t bnxt_cc_read(struct bnxt *bp)
3214 {
3215 	uint64_t ns;
3216 
3217 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3218 			      BNXT_GRCPF_REG_SYNC_TIME));
3219 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3220 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3221 	return ns;
3222 }
3223 
3224 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3225 {
3226 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3227 	uint32_t fifo;
3228 
3229 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3230 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3231 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3232 		return -EAGAIN;
3233 
3234 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3235 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3236 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3237 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3238 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3239 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3240 
3241 	return 0;
3242 }
3243 
3244 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3245 {
3246 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3247 	struct bnxt_pf_info *pf = bp->pf;
3248 	uint16_t port_id;
3249 	uint32_t fifo;
3250 
3251 	if (!ptp)
3252 		return -ENODEV;
3253 
3254 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3255 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3256 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3257 		return -EAGAIN;
3258 
3259 	port_id = pf->port_id;
3260 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3261 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3262 
3263 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3264 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3265 	if (fifo & BNXT_PTP_RX_FIFO_PENDING) {
3266 /*		bnxt_clr_rx_ts(bp);	  TBD  */
3267 		return -EBUSY;
3268 	}
3269 
3270 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3271 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3272 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3273 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3274 
3275 	return 0;
3276 }
3277 
3278 static int
3279 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3280 {
3281 	uint64_t ns;
3282 	struct bnxt *bp = dev->data->dev_private;
3283 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3284 
3285 	if (!ptp)
3286 		return 0;
3287 
3288 	ns = rte_timespec_to_ns(ts);
3289 	/* Set the timecounters to a new value. */
3290 	ptp->tc.nsec = ns;
3291 
3292 	return 0;
3293 }
3294 
3295 static int
3296 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3297 {
3298 	struct bnxt *bp = dev->data->dev_private;
3299 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3300 	uint64_t ns, systime_cycles = 0;
3301 	int rc = 0;
3302 
3303 	if (!ptp)
3304 		return 0;
3305 
3306 	if (BNXT_CHIP_P5(bp))
3307 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3308 					     &systime_cycles);
3309 	else
3310 		systime_cycles = bnxt_cc_read(bp);
3311 
3312 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3313 	*ts = rte_ns_to_timespec(ns);
3314 
3315 	return rc;
3316 }
3317 static int
3318 bnxt_timesync_enable(struct rte_eth_dev *dev)
3319 {
3320 	struct bnxt *bp = dev->data->dev_private;
3321 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3322 	uint32_t shift = 0;
3323 	int rc;
3324 
3325 	if (!ptp)
3326 		return 0;
3327 
3328 	ptp->rx_filter = 1;
3329 	ptp->tx_tstamp_en = 1;
3330 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3331 
3332 	rc = bnxt_hwrm_ptp_cfg(bp);
3333 	if (rc)
3334 		return rc;
3335 
3336 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3337 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3338 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3339 
3340 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3341 	ptp->tc.cc_shift = shift;
3342 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3343 
3344 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3345 	ptp->rx_tstamp_tc.cc_shift = shift;
3346 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3347 
3348 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3349 	ptp->tx_tstamp_tc.cc_shift = shift;
3350 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3351 
3352 	if (!BNXT_CHIP_P5(bp))
3353 		bnxt_map_ptp_regs(bp);
3354 
3355 	return 0;
3356 }
3357 
3358 static int
3359 bnxt_timesync_disable(struct rte_eth_dev *dev)
3360 {
3361 	struct bnxt *bp = dev->data->dev_private;
3362 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3363 
3364 	if (!ptp)
3365 		return 0;
3366 
3367 	ptp->rx_filter = 0;
3368 	ptp->tx_tstamp_en = 0;
3369 	ptp->rxctl = 0;
3370 
3371 	bnxt_hwrm_ptp_cfg(bp);
3372 
3373 	if (!BNXT_CHIP_P5(bp))
3374 		bnxt_unmap_ptp_regs(bp);
3375 
3376 	return 0;
3377 }
3378 
3379 static int
3380 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3381 				 struct timespec *timestamp,
3382 				 uint32_t flags __rte_unused)
3383 {
3384 	struct bnxt *bp = dev->data->dev_private;
3385 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3386 	uint64_t rx_tstamp_cycles = 0;
3387 	uint64_t ns;
3388 
3389 	if (!ptp)
3390 		return 0;
3391 
3392 	if (BNXT_CHIP_P5(bp))
3393 		rx_tstamp_cycles = ptp->rx_timestamp;
3394 	else
3395 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3396 
3397 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3398 	*timestamp = rte_ns_to_timespec(ns);
3399 	return  0;
3400 }
3401 
3402 static int
3403 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3404 				 struct timespec *timestamp)
3405 {
3406 	struct bnxt *bp = dev->data->dev_private;
3407 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3408 	uint64_t tx_tstamp_cycles = 0;
3409 	uint64_t ns;
3410 	int rc = 0;
3411 
3412 	if (!ptp)
3413 		return 0;
3414 
3415 	if (BNXT_CHIP_P5(bp))
3416 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3417 					     &tx_tstamp_cycles);
3418 	else
3419 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3420 
3421 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3422 	*timestamp = rte_ns_to_timespec(ns);
3423 
3424 	return rc;
3425 }
3426 
3427 static int
3428 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3429 {
3430 	struct bnxt *bp = dev->data->dev_private;
3431 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3432 
3433 	if (!ptp)
3434 		return 0;
3435 
3436 	ptp->tc.nsec += delta;
3437 
3438 	return 0;
3439 }
3440 
3441 static int
3442 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3443 {
3444 	struct bnxt *bp = dev->data->dev_private;
3445 	int rc;
3446 	uint32_t dir_entries;
3447 	uint32_t entry_length;
3448 
3449 	rc = is_bnxt_in_error(bp);
3450 	if (rc)
3451 		return rc;
3452 
3453 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3454 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3455 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3456 
3457 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3458 	if (rc != 0)
3459 		return rc;
3460 
3461 	return dir_entries * entry_length;
3462 }
3463 
3464 static int
3465 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3466 		struct rte_dev_eeprom_info *in_eeprom)
3467 {
3468 	struct bnxt *bp = dev->data->dev_private;
3469 	uint32_t index;
3470 	uint32_t offset;
3471 	int rc;
3472 
3473 	rc = is_bnxt_in_error(bp);
3474 	if (rc)
3475 		return rc;
3476 
3477 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3478 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3479 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3480 		    in_eeprom->offset, in_eeprom->length);
3481 
3482 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3483 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3484 						in_eeprom->data);
3485 
3486 	index = in_eeprom->offset >> 24;
3487 	offset = in_eeprom->offset & 0xffffff;
3488 
3489 	if (index != 0)
3490 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3491 					   in_eeprom->length, in_eeprom->data);
3492 
3493 	return 0;
3494 }
3495 
3496 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3497 {
3498 	switch (dir_type) {
3499 	case BNX_DIR_TYPE_CHIMP_PATCH:
3500 	case BNX_DIR_TYPE_BOOTCODE:
3501 	case BNX_DIR_TYPE_BOOTCODE_2:
3502 	case BNX_DIR_TYPE_APE_FW:
3503 	case BNX_DIR_TYPE_APE_PATCH:
3504 	case BNX_DIR_TYPE_KONG_FW:
3505 	case BNX_DIR_TYPE_KONG_PATCH:
3506 	case BNX_DIR_TYPE_BONO_FW:
3507 	case BNX_DIR_TYPE_BONO_PATCH:
3508 		/* FALLTHROUGH */
3509 		return true;
3510 	}
3511 
3512 	return false;
3513 }
3514 
3515 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3516 {
3517 	switch (dir_type) {
3518 	case BNX_DIR_TYPE_AVS:
3519 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3520 	case BNX_DIR_TYPE_PCIE:
3521 	case BNX_DIR_TYPE_TSCF_UCODE:
3522 	case BNX_DIR_TYPE_EXT_PHY:
3523 	case BNX_DIR_TYPE_CCM:
3524 	case BNX_DIR_TYPE_ISCSI_BOOT:
3525 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3526 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3527 		/* FALLTHROUGH */
3528 		return true;
3529 	}
3530 
3531 	return false;
3532 }
3533 
3534 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3535 {
3536 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3537 		bnxt_dir_type_is_other_exec_format(dir_type);
3538 }
3539 
3540 static int
3541 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3542 		struct rte_dev_eeprom_info *in_eeprom)
3543 {
3544 	struct bnxt *bp = dev->data->dev_private;
3545 	uint8_t index, dir_op;
3546 	uint16_t type, ext, ordinal, attr;
3547 	int rc;
3548 
3549 	rc = is_bnxt_in_error(bp);
3550 	if (rc)
3551 		return rc;
3552 
3553 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3554 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3555 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3556 		    in_eeprom->offset, in_eeprom->length);
3557 
3558 	if (!BNXT_PF(bp)) {
3559 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3560 		return -EINVAL;
3561 	}
3562 
3563 	type = in_eeprom->magic >> 16;
3564 
3565 	if (type == 0xffff) { /* special value for directory operations */
3566 		index = in_eeprom->magic & 0xff;
3567 		dir_op = in_eeprom->magic >> 8;
3568 		if (index == 0)
3569 			return -EINVAL;
3570 		switch (dir_op) {
3571 		case 0x0e: /* erase */
3572 			if (in_eeprom->offset != ~in_eeprom->magic)
3573 				return -EINVAL;
3574 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3575 		default:
3576 			return -EINVAL;
3577 		}
3578 	}
3579 
3580 	/* Create or re-write an NVM item: */
3581 	if (bnxt_dir_type_is_executable(type) == true)
3582 		return -EOPNOTSUPP;
3583 	ext = in_eeprom->magic & 0xffff;
3584 	ordinal = in_eeprom->offset >> 16;
3585 	attr = in_eeprom->offset & 0xffff;
3586 
3587 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3588 				     in_eeprom->data, in_eeprom->length);
3589 }
3590 
3591 /*
3592  * Initialization
3593  */
3594 
3595 static const struct eth_dev_ops bnxt_dev_ops = {
3596 	.dev_infos_get = bnxt_dev_info_get_op,
3597 	.dev_close = bnxt_dev_close_op,
3598 	.dev_configure = bnxt_dev_configure_op,
3599 	.dev_start = bnxt_dev_start_op,
3600 	.dev_stop = bnxt_dev_stop_op,
3601 	.dev_set_link_up = bnxt_dev_set_link_up_op,
3602 	.dev_set_link_down = bnxt_dev_set_link_down_op,
3603 	.stats_get = bnxt_stats_get_op,
3604 	.stats_reset = bnxt_stats_reset_op,
3605 	.rx_queue_setup = bnxt_rx_queue_setup_op,
3606 	.rx_queue_release = bnxt_rx_queue_release_op,
3607 	.tx_queue_setup = bnxt_tx_queue_setup_op,
3608 	.tx_queue_release = bnxt_tx_queue_release_op,
3609 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
3610 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
3611 	.reta_update = bnxt_reta_update_op,
3612 	.reta_query = bnxt_reta_query_op,
3613 	.rss_hash_update = bnxt_rss_hash_update_op,
3614 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
3615 	.link_update = bnxt_link_update_op,
3616 	.promiscuous_enable = bnxt_promiscuous_enable_op,
3617 	.promiscuous_disable = bnxt_promiscuous_disable_op,
3618 	.allmulticast_enable = bnxt_allmulticast_enable_op,
3619 	.allmulticast_disable = bnxt_allmulticast_disable_op,
3620 	.mac_addr_add = bnxt_mac_addr_add_op,
3621 	.mac_addr_remove = bnxt_mac_addr_remove_op,
3622 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
3623 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
3624 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
3625 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
3626 	.vlan_filter_set = bnxt_vlan_filter_set_op,
3627 	.vlan_offload_set = bnxt_vlan_offload_set_op,
3628 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
3629 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
3630 	.mtu_set = bnxt_mtu_set_op,
3631 	.mac_addr_set = bnxt_set_default_mac_addr_op,
3632 	.xstats_get = bnxt_dev_xstats_get_op,
3633 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
3634 	.xstats_reset = bnxt_dev_xstats_reset_op,
3635 	.fw_version_get = bnxt_fw_version_get,
3636 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
3637 	.rxq_info_get = bnxt_rxq_info_get_op,
3638 	.txq_info_get = bnxt_txq_info_get_op,
3639 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
3640 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
3641 	.dev_led_on = bnxt_dev_led_on_op,
3642 	.dev_led_off = bnxt_dev_led_off_op,
3643 	.rx_queue_start = bnxt_rx_queue_start,
3644 	.rx_queue_stop = bnxt_rx_queue_stop,
3645 	.tx_queue_start = bnxt_tx_queue_start,
3646 	.tx_queue_stop = bnxt_tx_queue_stop,
3647 	.filter_ctrl = bnxt_filter_ctrl_op,
3648 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
3649 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
3650 	.get_eeprom           = bnxt_get_eeprom_op,
3651 	.set_eeprom           = bnxt_set_eeprom_op,
3652 	.timesync_enable      = bnxt_timesync_enable,
3653 	.timesync_disable     = bnxt_timesync_disable,
3654 	.timesync_read_time   = bnxt_timesync_read_time,
3655 	.timesync_write_time   = bnxt_timesync_write_time,
3656 	.timesync_adjust_time = bnxt_timesync_adjust_time,
3657 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
3658 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
3659 };
3660 
3661 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
3662 {
3663 	uint32_t offset;
3664 
3665 	/* Only pre-map the reset GRC registers using window 3 */
3666 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
3667 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
3668 
3669 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
3670 
3671 	return offset;
3672 }
3673 
3674 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
3675 {
3676 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3677 	uint32_t reg_base = 0xffffffff;
3678 	int i;
3679 
3680 	/* Only pre-map the monitoring GRC registers using window 2 */
3681 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
3682 		uint32_t reg = info->status_regs[i];
3683 
3684 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
3685 			continue;
3686 
3687 		if (reg_base == 0xffffffff)
3688 			reg_base = reg & 0xfffff000;
3689 		if ((reg & 0xfffff000) != reg_base)
3690 			return -ERANGE;
3691 
3692 		/* Use mask 0xffc as the Lower 2 bits indicates
3693 		 * address space location
3694 		 */
3695 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
3696 						(reg & 0xffc);
3697 	}
3698 
3699 	if (reg_base == 0xffffffff)
3700 		return 0;
3701 
3702 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
3703 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
3704 
3705 	return 0;
3706 }
3707 
3708 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
3709 {
3710 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3711 	uint32_t delay = info->delay_after_reset[index];
3712 	uint32_t val = info->reset_reg_val[index];
3713 	uint32_t reg = info->reset_reg[index];
3714 	uint32_t type, offset;
3715 
3716 	type = BNXT_FW_STATUS_REG_TYPE(reg);
3717 	offset = BNXT_FW_STATUS_REG_OFF(reg);
3718 
3719 	switch (type) {
3720 	case BNXT_FW_STATUS_REG_TYPE_CFG:
3721 		rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
3722 		break;
3723 	case BNXT_FW_STATUS_REG_TYPE_GRC:
3724 		offset = bnxt_map_reset_regs(bp, offset);
3725 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3726 		break;
3727 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
3728 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3729 		break;
3730 	}
3731 	/* wait on a specific interval of time until core reset is complete */
3732 	if (delay)
3733 		rte_delay_ms(delay);
3734 }
3735 
3736 static void bnxt_dev_cleanup(struct bnxt *bp)
3737 {
3738 	bp->eth_dev->data->dev_link.link_status = 0;
3739 	bp->link_info->link_up = 0;
3740 	if (bp->eth_dev->data->dev_started)
3741 		bnxt_dev_stop(bp->eth_dev);
3742 
3743 	bnxt_uninit_resources(bp, true);
3744 }
3745 
3746 static int bnxt_restore_vlan_filters(struct bnxt *bp)
3747 {
3748 	struct rte_eth_dev *dev = bp->eth_dev;
3749 	struct rte_vlan_filter_conf *vfc;
3750 	int vidx, vbit, rc;
3751 	uint16_t vlan_id;
3752 
3753 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
3754 		vfc = &dev->data->vlan_filter_conf;
3755 		vidx = vlan_id / 64;
3756 		vbit = vlan_id % 64;
3757 
3758 		/* Each bit corresponds to a VLAN id */
3759 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
3760 			rc = bnxt_add_vlan_filter(bp, vlan_id);
3761 			if (rc)
3762 				return rc;
3763 		}
3764 	}
3765 
3766 	return 0;
3767 }
3768 
3769 static int bnxt_restore_mac_filters(struct bnxt *bp)
3770 {
3771 	struct rte_eth_dev *dev = bp->eth_dev;
3772 	struct rte_eth_dev_info dev_info;
3773 	struct rte_ether_addr *addr;
3774 	uint64_t pool_mask;
3775 	uint32_t pool = 0;
3776 	uint16_t i;
3777 	int rc;
3778 
3779 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
3780 		return 0;
3781 
3782 	rc = bnxt_dev_info_get_op(dev, &dev_info);
3783 	if (rc)
3784 		return rc;
3785 
3786 	/* replay MAC address configuration */
3787 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
3788 		addr = &dev->data->mac_addrs[i];
3789 
3790 		/* skip zero address */
3791 		if (rte_is_zero_ether_addr(addr))
3792 			continue;
3793 
3794 		pool = 0;
3795 		pool_mask = dev->data->mac_pool_sel[i];
3796 
3797 		do {
3798 			if (pool_mask & 1ULL) {
3799 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
3800 				if (rc)
3801 					return rc;
3802 			}
3803 			pool_mask >>= 1;
3804 			pool++;
3805 		} while (pool_mask);
3806 	}
3807 
3808 	return 0;
3809 }
3810 
3811 static int bnxt_restore_filters(struct bnxt *bp)
3812 {
3813 	struct rte_eth_dev *dev = bp->eth_dev;
3814 	int ret = 0;
3815 
3816 	if (dev->data->all_multicast) {
3817 		ret = bnxt_allmulticast_enable_op(dev);
3818 		if (ret)
3819 			return ret;
3820 	}
3821 	if (dev->data->promiscuous) {
3822 		ret = bnxt_promiscuous_enable_op(dev);
3823 		if (ret)
3824 			return ret;
3825 	}
3826 
3827 	ret = bnxt_restore_mac_filters(bp);
3828 	if (ret)
3829 		return ret;
3830 
3831 	ret = bnxt_restore_vlan_filters(bp);
3832 	/* TODO restore other filters as well */
3833 	return ret;
3834 }
3835 
3836 static void bnxt_dev_recover(void *arg)
3837 {
3838 	struct bnxt *bp = arg;
3839 	int timeout = bp->fw_reset_max_msecs;
3840 	int rc = 0;
3841 
3842 	pthread_mutex_lock(&bp->err_recovery_lock);
3843 	/* Clear Error flag so that device re-init should happen */
3844 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
3845 
3846 	do {
3847 		rc = bnxt_hwrm_ver_get(bp, SHORT_HWRM_CMD_TIMEOUT);
3848 		if (rc == 0)
3849 			break;
3850 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
3851 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
3852 	} while (rc && timeout);
3853 
3854 	if (rc) {
3855 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
3856 		goto err;
3857 	}
3858 
3859 	rc = bnxt_init_resources(bp, true);
3860 	if (rc) {
3861 		PMD_DRV_LOG(ERR,
3862 			    "Failed to initialize resources after reset\n");
3863 		goto err;
3864 	}
3865 	/* clear reset flag as the device is initialized now */
3866 	bp->flags &= ~BNXT_FLAG_FW_RESET;
3867 
3868 	rc = bnxt_dev_start_op(bp->eth_dev);
3869 	if (rc) {
3870 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
3871 		goto err_start;
3872 	}
3873 
3874 	rc = bnxt_restore_filters(bp);
3875 	if (rc)
3876 		goto err_start;
3877 
3878 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
3879 	pthread_mutex_unlock(&bp->err_recovery_lock);
3880 
3881 	return;
3882 err_start:
3883 	bnxt_dev_stop(bp->eth_dev);
3884 err:
3885 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
3886 	bnxt_uninit_resources(bp, false);
3887 	pthread_mutex_unlock(&bp->err_recovery_lock);
3888 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
3889 }
3890 
3891 void bnxt_dev_reset_and_resume(void *arg)
3892 {
3893 	struct bnxt *bp = arg;
3894 	int rc;
3895 
3896 	bnxt_dev_cleanup(bp);
3897 
3898 	bnxt_wait_for_device_shutdown(bp);
3899 
3900 	rc = rte_eal_alarm_set(US_PER_MS * bp->fw_reset_min_msecs,
3901 			       bnxt_dev_recover, (void *)bp);
3902 	if (rc)
3903 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
3904 }
3905 
3906 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
3907 {
3908 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3909 	uint32_t reg = info->status_regs[index];
3910 	uint32_t type, offset, val = 0;
3911 
3912 	type = BNXT_FW_STATUS_REG_TYPE(reg);
3913 	offset = BNXT_FW_STATUS_REG_OFF(reg);
3914 
3915 	switch (type) {
3916 	case BNXT_FW_STATUS_REG_TYPE_CFG:
3917 		rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
3918 		break;
3919 	case BNXT_FW_STATUS_REG_TYPE_GRC:
3920 		offset = info->mapped_status_regs[index];
3921 		/* FALLTHROUGH */
3922 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
3923 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3924 				       offset));
3925 		break;
3926 	}
3927 
3928 	return val;
3929 }
3930 
3931 static int bnxt_fw_reset_all(struct bnxt *bp)
3932 {
3933 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3934 	uint32_t i;
3935 	int rc = 0;
3936 
3937 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
3938 		/* Reset through master function driver */
3939 		for (i = 0; i < info->reg_array_cnt; i++)
3940 			bnxt_write_fw_reset_reg(bp, i);
3941 		/* Wait for time specified by FW after triggering reset */
3942 		rte_delay_ms(info->master_func_wait_period_after_reset);
3943 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
3944 		/* Reset with the help of Kong processor */
3945 		rc = bnxt_hwrm_fw_reset(bp);
3946 		if (rc)
3947 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
3948 	}
3949 
3950 	return rc;
3951 }
3952 
3953 static void bnxt_fw_reset_cb(void *arg)
3954 {
3955 	struct bnxt *bp = arg;
3956 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3957 	int rc = 0;
3958 
3959 	/* Only Master function can do FW reset */
3960 	if (bnxt_is_master_func(bp) &&
3961 	    bnxt_is_recovery_enabled(bp)) {
3962 		rc = bnxt_fw_reset_all(bp);
3963 		if (rc) {
3964 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
3965 			return;
3966 		}
3967 	}
3968 
3969 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
3970 	 * EXCEPTION_FATAL_ASYNC event to all the functions
3971 	 * (including MASTER FUNC). After receiving this Async, all the active
3972 	 * drivers should treat this case as FW initiated recovery
3973 	 */
3974 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
3975 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
3976 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
3977 
3978 		/* To recover from error */
3979 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
3980 				  (void *)bp);
3981 	}
3982 }
3983 
3984 /* Driver should poll FW heartbeat, reset_counter with the frequency
3985  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
3986  * When the driver detects heartbeat stop or change in reset_counter,
3987  * it has to trigger a reset to recover from the error condition.
3988  * A “master PF” is the function who will have the privilege to
3989  * initiate the chimp reset. The master PF will be elected by the
3990  * firmware and will be notified through async message.
3991  */
3992 static void bnxt_check_fw_health(void *arg)
3993 {
3994 	struct bnxt *bp = arg;
3995 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3996 	uint32_t val = 0, wait_msec;
3997 
3998 	if (!info || !bnxt_is_recovery_enabled(bp) ||
3999 	    is_bnxt_in_error(bp))
4000 		return;
4001 
4002 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4003 	if (val == info->last_heart_beat)
4004 		goto reset;
4005 
4006 	info->last_heart_beat = val;
4007 
4008 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4009 	if (val != info->last_reset_counter)
4010 		goto reset;
4011 
4012 	info->last_reset_counter = val;
4013 
4014 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4015 			  bnxt_check_fw_health, (void *)bp);
4016 
4017 	return;
4018 reset:
4019 	/* Stop DMA to/from device */
4020 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4021 	bp->flags |= BNXT_FLAG_FW_RESET;
4022 
4023 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4024 
4025 	if (bnxt_is_master_func(bp))
4026 		wait_msec = info->master_func_wait_period;
4027 	else
4028 		wait_msec = info->normal_func_wait_period;
4029 
4030 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4031 			  bnxt_fw_reset_cb, (void *)bp);
4032 }
4033 
4034 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4035 {
4036 	uint32_t polling_freq;
4037 
4038 	pthread_mutex_lock(&bp->health_check_lock);
4039 
4040 	if (!bnxt_is_recovery_enabled(bp))
4041 		goto done;
4042 
4043 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4044 		goto done;
4045 
4046 	polling_freq = bp->recovery_info->driver_polling_freq;
4047 
4048 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4049 			  bnxt_check_fw_health, (void *)bp);
4050 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4051 
4052 done:
4053 	pthread_mutex_unlock(&bp->health_check_lock);
4054 }
4055 
4056 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4057 {
4058 	if (!bnxt_is_recovery_enabled(bp))
4059 		return;
4060 
4061 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4062 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4063 }
4064 
4065 static bool bnxt_vf_pciid(uint16_t device_id)
4066 {
4067 	switch (device_id) {
4068 	case BROADCOM_DEV_ID_57304_VF:
4069 	case BROADCOM_DEV_ID_57406_VF:
4070 	case BROADCOM_DEV_ID_5731X_VF:
4071 	case BROADCOM_DEV_ID_5741X_VF:
4072 	case BROADCOM_DEV_ID_57414_VF:
4073 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4074 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4075 	case BROADCOM_DEV_ID_58802_VF:
4076 	case BROADCOM_DEV_ID_57500_VF1:
4077 	case BROADCOM_DEV_ID_57500_VF2:
4078 	case BROADCOM_DEV_ID_58818_VF:
4079 		/* FALLTHROUGH */
4080 		return true;
4081 	default:
4082 		return false;
4083 	}
4084 }
4085 
4086 /* Phase 5 device */
4087 static bool bnxt_p5_device(uint16_t device_id)
4088 {
4089 	switch (device_id) {
4090 	case BROADCOM_DEV_ID_57508:
4091 	case BROADCOM_DEV_ID_57504:
4092 	case BROADCOM_DEV_ID_57502:
4093 	case BROADCOM_DEV_ID_57508_MF1:
4094 	case BROADCOM_DEV_ID_57504_MF1:
4095 	case BROADCOM_DEV_ID_57502_MF1:
4096 	case BROADCOM_DEV_ID_57508_MF2:
4097 	case BROADCOM_DEV_ID_57504_MF2:
4098 	case BROADCOM_DEV_ID_57502_MF2:
4099 	case BROADCOM_DEV_ID_57500_VF1:
4100 	case BROADCOM_DEV_ID_57500_VF2:
4101 	case BROADCOM_DEV_ID_58812:
4102 	case BROADCOM_DEV_ID_58814:
4103 	case BROADCOM_DEV_ID_58818:
4104 	case BROADCOM_DEV_ID_58818_VF:
4105 		/* FALLTHROUGH */
4106 		return true;
4107 	default:
4108 		return false;
4109 	}
4110 }
4111 
4112 bool bnxt_stratus_device(struct bnxt *bp)
4113 {
4114 	uint16_t device_id = bp->pdev->id.device_id;
4115 
4116 	switch (device_id) {
4117 	case BROADCOM_DEV_ID_STRATUS_NIC:
4118 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4119 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4120 		/* FALLTHROUGH */
4121 		return true;
4122 	default:
4123 		return false;
4124 	}
4125 }
4126 
4127 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4128 {
4129 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4130 	struct bnxt *bp = eth_dev->data->dev_private;
4131 
4132 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4133 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4134 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4135 	if (!bp->bar0 || !bp->doorbell_base) {
4136 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4137 		return -ENODEV;
4138 	}
4139 
4140 	bp->eth_dev = eth_dev;
4141 	bp->pdev = pci_dev;
4142 
4143 	return 0;
4144 }
4145 
4146 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4147 				  struct bnxt_ctx_pg_info *ctx_pg,
4148 				  uint32_t mem_size,
4149 				  const char *suffix,
4150 				  uint16_t idx)
4151 {
4152 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4153 	const struct rte_memzone *mz = NULL;
4154 	char mz_name[RTE_MEMZONE_NAMESIZE];
4155 	rte_iova_t mz_phys_addr;
4156 	uint64_t valid_bits = 0;
4157 	uint32_t sz;
4158 	int i;
4159 
4160 	if (!mem_size)
4161 		return 0;
4162 
4163 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4164 			 BNXT_PAGE_SIZE;
4165 	rmem->page_size = BNXT_PAGE_SIZE;
4166 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4167 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4168 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4169 
4170 	valid_bits = PTU_PTE_VALID;
4171 
4172 	if (rmem->nr_pages > 1) {
4173 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4174 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4175 			 suffix, idx, bp->eth_dev->data->port_id);
4176 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4177 		mz = rte_memzone_lookup(mz_name);
4178 		if (!mz) {
4179 			mz = rte_memzone_reserve_aligned(mz_name,
4180 						rmem->nr_pages * 8,
4181 						SOCKET_ID_ANY,
4182 						RTE_MEMZONE_2MB |
4183 						RTE_MEMZONE_SIZE_HINT_ONLY |
4184 						RTE_MEMZONE_IOVA_CONTIG,
4185 						BNXT_PAGE_SIZE);
4186 			if (mz == NULL)
4187 				return -ENOMEM;
4188 		}
4189 
4190 		memset(mz->addr, 0, mz->len);
4191 		mz_phys_addr = mz->iova;
4192 
4193 		rmem->pg_tbl = mz->addr;
4194 		rmem->pg_tbl_map = mz_phys_addr;
4195 		rmem->pg_tbl_mz = mz;
4196 	}
4197 
4198 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4199 		 suffix, idx, bp->eth_dev->data->port_id);
4200 	mz = rte_memzone_lookup(mz_name);
4201 	if (!mz) {
4202 		mz = rte_memzone_reserve_aligned(mz_name,
4203 						 mem_size,
4204 						 SOCKET_ID_ANY,
4205 						 RTE_MEMZONE_1GB |
4206 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4207 						 RTE_MEMZONE_IOVA_CONTIG,
4208 						 BNXT_PAGE_SIZE);
4209 		if (mz == NULL)
4210 			return -ENOMEM;
4211 	}
4212 
4213 	memset(mz->addr, 0, mz->len);
4214 	mz_phys_addr = mz->iova;
4215 
4216 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4217 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4218 		rmem->dma_arr[i] = mz_phys_addr + sz;
4219 
4220 		if (rmem->nr_pages > 1) {
4221 			if (i == rmem->nr_pages - 2 &&
4222 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4223 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4224 			else if (i == rmem->nr_pages - 1 &&
4225 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4226 				valid_bits |= PTU_PTE_LAST;
4227 
4228 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4229 							   valid_bits);
4230 		}
4231 	}
4232 
4233 	rmem->mz = mz;
4234 	if (rmem->vmem_size)
4235 		rmem->vmem = (void **)mz->addr;
4236 	rmem->dma_arr[0] = mz_phys_addr;
4237 	return 0;
4238 }
4239 
4240 static void bnxt_free_ctx_mem(struct bnxt *bp)
4241 {
4242 	int i;
4243 
4244 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4245 		return;
4246 
4247 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4248 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4249 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4250 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4251 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4252 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4253 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4254 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4255 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4256 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4257 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4258 
4259 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4260 		if (bp->ctx->tqm_mem[i])
4261 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4262 	}
4263 
4264 	rte_free(bp->ctx);
4265 	bp->ctx = NULL;
4266 }
4267 
4268 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4269 
4270 #define min_t(type, x, y) ({                    \
4271 	type __min1 = (x);                      \
4272 	type __min2 = (y);                      \
4273 	__min1 < __min2 ? __min1 : __min2; })
4274 
4275 #define max_t(type, x, y) ({                    \
4276 	type __max1 = (x);                      \
4277 	type __max2 = (y);                      \
4278 	__max1 > __max2 ? __max1 : __max2; })
4279 
4280 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4281 
4282 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4283 {
4284 	struct bnxt_ctx_pg_info *ctx_pg;
4285 	struct bnxt_ctx_mem_info *ctx;
4286 	uint32_t mem_size, ena, entries;
4287 	uint32_t entries_sp, min;
4288 	int i, rc;
4289 
4290 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4291 	if (rc) {
4292 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4293 		return rc;
4294 	}
4295 	ctx = bp->ctx;
4296 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4297 		return 0;
4298 
4299 	ctx_pg = &ctx->qp_mem;
4300 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4301 	if (ctx->qp_entry_size) {
4302 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4303 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4304 		if (rc)
4305 			return rc;
4306 	}
4307 
4308 	ctx_pg = &ctx->srq_mem;
4309 	ctx_pg->entries = ctx->srq_max_l2_entries;
4310 	if (ctx->srq_entry_size) {
4311 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4312 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4313 		if (rc)
4314 			return rc;
4315 	}
4316 
4317 	ctx_pg = &ctx->cq_mem;
4318 	ctx_pg->entries = ctx->cq_max_l2_entries;
4319 	if (ctx->cq_entry_size) {
4320 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4321 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4322 		if (rc)
4323 			return rc;
4324 	}
4325 
4326 	ctx_pg = &ctx->vnic_mem;
4327 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4328 		ctx->vnic_max_ring_table_entries;
4329 	if (ctx->vnic_entry_size) {
4330 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4331 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4332 		if (rc)
4333 			return rc;
4334 	}
4335 
4336 	ctx_pg = &ctx->stat_mem;
4337 	ctx_pg->entries = ctx->stat_max_entries;
4338 	if (ctx->stat_entry_size) {
4339 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4340 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4341 		if (rc)
4342 			return rc;
4343 	}
4344 
4345 	min = ctx->tqm_min_entries_per_ring;
4346 
4347 	entries_sp = ctx->qp_max_l2_entries +
4348 		     ctx->vnic_max_vnic_entries +
4349 		     2 * ctx->qp_min_qp1_entries + min;
4350 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4351 
4352 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4353 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4354 	entries = clamp_t(uint32_t, entries, min,
4355 			  ctx->tqm_max_entries_per_ring);
4356 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4357 		ctx_pg = ctx->tqm_mem[i];
4358 		ctx_pg->entries = i ? entries : entries_sp;
4359 		if (ctx->tqm_entry_size) {
4360 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4361 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "tqm_mem", i);
4362 			if (rc)
4363 				return rc;
4364 		}
4365 		ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4366 	}
4367 
4368 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4369 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4370 	if (rc)
4371 		PMD_DRV_LOG(ERR,
4372 			    "Failed to configure context mem: rc = %d\n", rc);
4373 	else
4374 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4375 
4376 	return rc;
4377 }
4378 
4379 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4380 {
4381 	struct rte_pci_device *pci_dev = bp->pdev;
4382 	char mz_name[RTE_MEMZONE_NAMESIZE];
4383 	const struct rte_memzone *mz = NULL;
4384 	uint32_t total_alloc_len;
4385 	rte_iova_t mz_phys_addr;
4386 
4387 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4388 		return 0;
4389 
4390 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4391 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4392 		 pci_dev->addr.bus, pci_dev->addr.devid,
4393 		 pci_dev->addr.function, "rx_port_stats");
4394 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4395 	mz = rte_memzone_lookup(mz_name);
4396 	total_alloc_len =
4397 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4398 				       sizeof(struct rx_port_stats_ext) + 512);
4399 	if (!mz) {
4400 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4401 					 SOCKET_ID_ANY,
4402 					 RTE_MEMZONE_2MB |
4403 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4404 					 RTE_MEMZONE_IOVA_CONTIG);
4405 		if (mz == NULL)
4406 			return -ENOMEM;
4407 	}
4408 	memset(mz->addr, 0, mz->len);
4409 	mz_phys_addr = mz->iova;
4410 
4411 	bp->rx_mem_zone = (const void *)mz;
4412 	bp->hw_rx_port_stats = mz->addr;
4413 	bp->hw_rx_port_stats_map = mz_phys_addr;
4414 
4415 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4416 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4417 		 pci_dev->addr.bus, pci_dev->addr.devid,
4418 		 pci_dev->addr.function, "tx_port_stats");
4419 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4420 	mz = rte_memzone_lookup(mz_name);
4421 	total_alloc_len =
4422 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4423 				       sizeof(struct tx_port_stats_ext) + 512);
4424 	if (!mz) {
4425 		mz = rte_memzone_reserve(mz_name,
4426 					 total_alloc_len,
4427 					 SOCKET_ID_ANY,
4428 					 RTE_MEMZONE_2MB |
4429 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4430 					 RTE_MEMZONE_IOVA_CONTIG);
4431 		if (mz == NULL)
4432 			return -ENOMEM;
4433 	}
4434 	memset(mz->addr, 0, mz->len);
4435 	mz_phys_addr = mz->iova;
4436 
4437 	bp->tx_mem_zone = (const void *)mz;
4438 	bp->hw_tx_port_stats = mz->addr;
4439 	bp->hw_tx_port_stats_map = mz_phys_addr;
4440 	bp->flags |= BNXT_FLAG_PORT_STATS;
4441 
4442 	/* Display extended statistics if FW supports it */
4443 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4444 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4445 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4446 		return 0;
4447 
4448 	bp->hw_rx_port_stats_ext = (void *)
4449 		((uint8_t *)bp->hw_rx_port_stats +
4450 		 sizeof(struct rx_port_stats));
4451 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4452 		sizeof(struct rx_port_stats);
4453 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4454 
4455 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4456 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4457 		bp->hw_tx_port_stats_ext = (void *)
4458 			((uint8_t *)bp->hw_tx_port_stats +
4459 			 sizeof(struct tx_port_stats));
4460 		bp->hw_tx_port_stats_ext_map =
4461 			bp->hw_tx_port_stats_map +
4462 			sizeof(struct tx_port_stats);
4463 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4464 	}
4465 
4466 	return 0;
4467 }
4468 
4469 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4470 {
4471 	struct bnxt *bp = eth_dev->data->dev_private;
4472 	int rc = 0;
4473 
4474 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4475 					       RTE_ETHER_ADDR_LEN *
4476 					       bp->max_l2_ctx,
4477 					       0);
4478 	if (eth_dev->data->mac_addrs == NULL) {
4479 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4480 		return -ENOMEM;
4481 	}
4482 
4483 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4484 		if (BNXT_PF(bp))
4485 			return -EINVAL;
4486 
4487 		/* Generate a random MAC address, if none was assigned by PF */
4488 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4489 		bnxt_eth_hw_addr_random(bp->mac_addr);
4490 		PMD_DRV_LOG(INFO,
4491 			    "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4492 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4493 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4494 
4495 		rc = bnxt_hwrm_set_mac(bp);
4496 		if (rc)
4497 			return rc;
4498 	}
4499 
4500 	/* Copy the permanent MAC from the FUNC_QCAPS response */
4501 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
4502 
4503 	return rc;
4504 }
4505 
4506 static int bnxt_restore_dflt_mac(struct bnxt *bp)
4507 {
4508 	int rc = 0;
4509 
4510 	/* MAC is already configured in FW */
4511 	if (BNXT_HAS_DFLT_MAC_SET(bp))
4512 		return 0;
4513 
4514 	/* Restore the old MAC configured */
4515 	rc = bnxt_hwrm_set_mac(bp);
4516 	if (rc)
4517 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
4518 
4519 	return rc;
4520 }
4521 
4522 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
4523 {
4524 	if (!BNXT_PF(bp))
4525 		return;
4526 
4527 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
4528 
4529 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
4530 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
4531 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
4532 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
4533 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
4534 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
4535 }
4536 
4537 uint16_t
4538 bnxt_get_svif(uint16_t port_id, bool func_svif,
4539 	      enum bnxt_ulp_intf_type type)
4540 {
4541 	struct rte_eth_dev *eth_dev;
4542 	struct bnxt *bp;
4543 
4544 	eth_dev = &rte_eth_devices[port_id];
4545 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4546 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4547 		if (!vfr)
4548 			return 0;
4549 
4550 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4551 			return vfr->svif;
4552 
4553 		eth_dev = vfr->parent_dev;
4554 	}
4555 
4556 	bp = eth_dev->data->dev_private;
4557 
4558 	return func_svif ? bp->func_svif : bp->port_svif;
4559 }
4560 
4561 uint16_t
4562 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
4563 {
4564 	struct rte_eth_dev *eth_dev;
4565 	struct bnxt_vnic_info *vnic;
4566 	struct bnxt *bp;
4567 
4568 	eth_dev = &rte_eth_devices[port];
4569 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4570 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4571 		if (!vfr)
4572 			return 0;
4573 
4574 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4575 			return vfr->dflt_vnic_id;
4576 
4577 		eth_dev = vfr->parent_dev;
4578 	}
4579 
4580 	bp = eth_dev->data->dev_private;
4581 
4582 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
4583 
4584 	return vnic->fw_vnic_id;
4585 }
4586 
4587 uint16_t
4588 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
4589 {
4590 	struct rte_eth_dev *eth_dev;
4591 	struct bnxt *bp;
4592 
4593 	eth_dev = &rte_eth_devices[port];
4594 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4595 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4596 		if (!vfr)
4597 			return 0;
4598 
4599 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4600 			return vfr->fw_fid;
4601 
4602 		eth_dev = vfr->parent_dev;
4603 	}
4604 
4605 	bp = eth_dev->data->dev_private;
4606 
4607 	return bp->fw_fid;
4608 }
4609 
4610 enum bnxt_ulp_intf_type
4611 bnxt_get_interface_type(uint16_t port)
4612 {
4613 	struct rte_eth_dev *eth_dev;
4614 	struct bnxt *bp;
4615 
4616 	eth_dev = &rte_eth_devices[port];
4617 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
4618 		return BNXT_ULP_INTF_TYPE_VF_REP;
4619 
4620 	bp = eth_dev->data->dev_private;
4621 	if (BNXT_PF(bp))
4622 		return BNXT_ULP_INTF_TYPE_PF;
4623 	else if (BNXT_VF_IS_TRUSTED(bp))
4624 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
4625 	else if (BNXT_VF(bp))
4626 		return BNXT_ULP_INTF_TYPE_VF;
4627 
4628 	return BNXT_ULP_INTF_TYPE_INVALID;
4629 }
4630 
4631 uint16_t
4632 bnxt_get_phy_port_id(uint16_t port_id)
4633 {
4634 	struct bnxt_representor *vfr;
4635 	struct rte_eth_dev *eth_dev;
4636 	struct bnxt *bp;
4637 
4638 	eth_dev = &rte_eth_devices[port_id];
4639 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4640 		vfr = eth_dev->data->dev_private;
4641 		if (!vfr)
4642 			return 0;
4643 
4644 		eth_dev = vfr->parent_dev;
4645 	}
4646 
4647 	bp = eth_dev->data->dev_private;
4648 
4649 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
4650 }
4651 
4652 uint16_t
4653 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
4654 {
4655 	struct rte_eth_dev *eth_dev;
4656 	struct bnxt *bp;
4657 
4658 	eth_dev = &rte_eth_devices[port_id];
4659 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4660 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4661 		if (!vfr)
4662 			return 0;
4663 
4664 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4665 			return vfr->fw_fid - 1;
4666 
4667 		eth_dev = vfr->parent_dev;
4668 	}
4669 
4670 	bp = eth_dev->data->dev_private;
4671 
4672 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
4673 }
4674 
4675 uint16_t
4676 bnxt_get_vport(uint16_t port_id)
4677 {
4678 	return (1 << bnxt_get_phy_port_id(port_id));
4679 }
4680 
4681 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
4682 {
4683 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4684 
4685 	if (info) {
4686 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
4687 			memset(info, 0, sizeof(*info));
4688 		return;
4689 	}
4690 
4691 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
4692 		return;
4693 
4694 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4695 			   sizeof(*info), 0);
4696 	if (!info)
4697 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4698 
4699 	bp->recovery_info = info;
4700 }
4701 
4702 static void bnxt_check_fw_status(struct bnxt *bp)
4703 {
4704 	uint32_t fw_status;
4705 
4706 	if (!(bp->recovery_info &&
4707 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
4708 		return;
4709 
4710 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
4711 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
4712 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
4713 			    fw_status);
4714 }
4715 
4716 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
4717 {
4718 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4719 	uint32_t status_loc;
4720 	uint32_t sig_ver;
4721 
4722 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
4723 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4724 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4725 				   BNXT_GRCP_WINDOW_2_BASE +
4726 				   offsetof(struct hcomm_status,
4727 					    sig_ver)));
4728 	/* If the signature is absent, then FW does not support this feature */
4729 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
4730 	    HCOMM_STATUS_SIGNATURE_VAL)
4731 		return 0;
4732 
4733 	if (!info) {
4734 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4735 				   sizeof(*info), 0);
4736 		if (!info)
4737 			return -ENOMEM;
4738 		bp->recovery_info = info;
4739 	} else {
4740 		memset(info, 0, sizeof(*info));
4741 	}
4742 
4743 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4744 				      BNXT_GRCP_WINDOW_2_BASE +
4745 				      offsetof(struct hcomm_status,
4746 					       fw_status_loc)));
4747 
4748 	/* Only pre-map the FW health status GRC register */
4749 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
4750 		return 0;
4751 
4752 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
4753 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
4754 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
4755 
4756 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
4757 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4758 
4759 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
4760 
4761 	return 0;
4762 }
4763 
4764 /* This function gets the FW version along with the
4765  * capabilities(MAX and current) of the function, vnic,
4766  * error recovery, phy and other chip related info
4767  */
4768 static int bnxt_get_config(struct bnxt *bp)
4769 {
4770 	uint16_t mtu;
4771 	int rc = 0;
4772 
4773 	bp->fw_cap = 0;
4774 
4775 	rc = bnxt_map_hcomm_fw_status_reg(bp);
4776 	if (rc)
4777 		return rc;
4778 
4779 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
4780 	if (rc) {
4781 		bnxt_check_fw_status(bp);
4782 		return rc;
4783 	}
4784 
4785 	rc = bnxt_hwrm_func_reset(bp);
4786 	if (rc)
4787 		return -EIO;
4788 
4789 	rc = bnxt_hwrm_vnic_qcaps(bp);
4790 	if (rc)
4791 		return rc;
4792 
4793 	rc = bnxt_hwrm_queue_qportcfg(bp);
4794 	if (rc)
4795 		return rc;
4796 
4797 	/* Get the MAX capabilities for this function.
4798 	 * This function also allocates context memory for TQM rings and
4799 	 * informs the firmware about this allocated backing store memory.
4800 	 */
4801 	rc = bnxt_hwrm_func_qcaps(bp);
4802 	if (rc)
4803 		return rc;
4804 
4805 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
4806 	if (rc)
4807 		return rc;
4808 
4809 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
4810 	if (rc)
4811 		return rc;
4812 
4813 	bnxt_hwrm_port_mac_qcfg(bp);
4814 
4815 	bnxt_hwrm_parent_pf_qcfg(bp);
4816 
4817 	bnxt_hwrm_port_phy_qcaps(bp);
4818 
4819 	bnxt_alloc_error_recovery_info(bp);
4820 	/* Get the adapter error recovery support info */
4821 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
4822 	if (rc)
4823 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4824 
4825 	bnxt_hwrm_port_led_qcaps(bp);
4826 
4827 	return 0;
4828 }
4829 
4830 static int
4831 bnxt_init_locks(struct bnxt *bp)
4832 {
4833 	int err;
4834 
4835 	err = pthread_mutex_init(&bp->flow_lock, NULL);
4836 	if (err) {
4837 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
4838 		return err;
4839 	}
4840 
4841 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
4842 	if (err) {
4843 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
4844 		return err;
4845 	}
4846 
4847 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
4848 	if (err) {
4849 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
4850 		return err;
4851 	}
4852 
4853 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
4854 	if (err)
4855 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
4856 
4857 	return err;
4858 }
4859 
4860 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
4861 {
4862 	int rc = 0;
4863 
4864 	rc = bnxt_get_config(bp);
4865 	if (rc)
4866 		return rc;
4867 
4868 	if (!reconfig_dev) {
4869 		rc = bnxt_setup_mac_addr(bp->eth_dev);
4870 		if (rc)
4871 			return rc;
4872 	} else {
4873 		rc = bnxt_restore_dflt_mac(bp);
4874 		if (rc)
4875 			return rc;
4876 	}
4877 
4878 	bnxt_config_vf_req_fwd(bp);
4879 
4880 	rc = bnxt_hwrm_func_driver_register(bp);
4881 	if (rc) {
4882 		PMD_DRV_LOG(ERR, "Failed to register driver");
4883 		return -EBUSY;
4884 	}
4885 
4886 	if (BNXT_PF(bp)) {
4887 		if (bp->pdev->max_vfs) {
4888 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
4889 			if (rc) {
4890 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
4891 				return rc;
4892 			}
4893 		} else {
4894 			rc = bnxt_hwrm_allocate_pf_only(bp);
4895 			if (rc) {
4896 				PMD_DRV_LOG(ERR,
4897 					    "Failed to allocate PF resources");
4898 				return rc;
4899 			}
4900 		}
4901 	}
4902 
4903 	rc = bnxt_alloc_mem(bp, reconfig_dev);
4904 	if (rc)
4905 		return rc;
4906 
4907 	rc = bnxt_setup_int(bp);
4908 	if (rc)
4909 		return rc;
4910 
4911 	rc = bnxt_request_int(bp);
4912 	if (rc)
4913 		return rc;
4914 
4915 	rc = bnxt_init_ctx_mem(bp);
4916 	if (rc) {
4917 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
4918 		return rc;
4919 	}
4920 
4921 	return 0;
4922 }
4923 
4924 static int
4925 bnxt_parse_devarg_truflow(__rte_unused const char *key,
4926 			  const char *value, void *opaque_arg)
4927 {
4928 	struct bnxt *bp = opaque_arg;
4929 	unsigned long truflow;
4930 	char *end = NULL;
4931 
4932 	if (!value || !opaque_arg) {
4933 		PMD_DRV_LOG(ERR,
4934 			    "Invalid parameter passed to truflow devargs.\n");
4935 		return -EINVAL;
4936 	}
4937 
4938 	truflow = strtoul(value, &end, 10);
4939 	if (end == NULL || *end != '\0' ||
4940 	    (truflow == ULONG_MAX && errno == ERANGE)) {
4941 		PMD_DRV_LOG(ERR,
4942 			    "Invalid parameter passed to truflow devargs.\n");
4943 		return -EINVAL;
4944 	}
4945 
4946 	if (BNXT_DEVARG_TRUFLOW_INVALID(truflow)) {
4947 		PMD_DRV_LOG(ERR,
4948 			    "Invalid value passed to truflow devargs.\n");
4949 		return -EINVAL;
4950 	}
4951 
4952 	if (truflow) {
4953 		bp->flags |= BNXT_FLAG_TRUFLOW_EN;
4954 		PMD_DRV_LOG(INFO, "Host-based truflow feature enabled.\n");
4955 	} else {
4956 		bp->flags &= ~BNXT_FLAG_TRUFLOW_EN;
4957 		PMD_DRV_LOG(INFO, "Host-based truflow feature disabled.\n");
4958 	}
4959 
4960 	return 0;
4961 }
4962 
4963 static int
4964 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
4965 			     const char *value, void *opaque_arg)
4966 {
4967 	struct bnxt *bp = opaque_arg;
4968 	unsigned long flow_xstat;
4969 	char *end = NULL;
4970 
4971 	if (!value || !opaque_arg) {
4972 		PMD_DRV_LOG(ERR,
4973 			    "Invalid parameter passed to flow_xstat devarg.\n");
4974 		return -EINVAL;
4975 	}
4976 
4977 	flow_xstat = strtoul(value, &end, 10);
4978 	if (end == NULL || *end != '\0' ||
4979 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
4980 		PMD_DRV_LOG(ERR,
4981 			    "Invalid parameter passed to flow_xstat devarg.\n");
4982 		return -EINVAL;
4983 	}
4984 
4985 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
4986 		PMD_DRV_LOG(ERR,
4987 			    "Invalid value passed to flow_xstat devarg.\n");
4988 		return -EINVAL;
4989 	}
4990 
4991 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
4992 	if (BNXT_FLOW_XSTATS_EN(bp))
4993 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
4994 
4995 	return 0;
4996 }
4997 
4998 static int
4999 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5000 					const char *value, void *opaque_arg)
5001 {
5002 	struct bnxt *bp = opaque_arg;
5003 	unsigned long max_num_kflows;
5004 	char *end = NULL;
5005 
5006 	if (!value || !opaque_arg) {
5007 		PMD_DRV_LOG(ERR,
5008 			"Invalid parameter passed to max_num_kflows devarg.\n");
5009 		return -EINVAL;
5010 	}
5011 
5012 	max_num_kflows = strtoul(value, &end, 10);
5013 	if (end == NULL || *end != '\0' ||
5014 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5015 		PMD_DRV_LOG(ERR,
5016 			"Invalid parameter passed to max_num_kflows devarg.\n");
5017 		return -EINVAL;
5018 	}
5019 
5020 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5021 		PMD_DRV_LOG(ERR,
5022 			"Invalid value passed to max_num_kflows devarg.\n");
5023 		return -EINVAL;
5024 	}
5025 
5026 	bp->max_num_kflows = max_num_kflows;
5027 	if (bp->max_num_kflows)
5028 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5029 				max_num_kflows);
5030 
5031 	return 0;
5032 }
5033 
5034 static int
5035 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5036 			    const char *value, void *opaque_arg)
5037 {
5038 	struct bnxt_representor *vfr_bp = opaque_arg;
5039 	unsigned long rep_is_pf;
5040 	char *end = NULL;
5041 
5042 	if (!value || !opaque_arg) {
5043 		PMD_DRV_LOG(ERR,
5044 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5045 		return -EINVAL;
5046 	}
5047 
5048 	rep_is_pf = strtoul(value, &end, 10);
5049 	if (end == NULL || *end != '\0' ||
5050 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5051 		PMD_DRV_LOG(ERR,
5052 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5053 		return -EINVAL;
5054 	}
5055 
5056 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5057 		PMD_DRV_LOG(ERR,
5058 			    "Invalid value passed to rep_is_pf devargs.\n");
5059 		return -EINVAL;
5060 	}
5061 
5062 	vfr_bp->flags |= rep_is_pf;
5063 	if (BNXT_REP_PF(vfr_bp))
5064 		PMD_DRV_LOG(INFO, "PF representor\n");
5065 	else
5066 		PMD_DRV_LOG(INFO, "VF representor\n");
5067 
5068 	return 0;
5069 }
5070 
5071 static int
5072 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5073 			       const char *value, void *opaque_arg)
5074 {
5075 	struct bnxt_representor *vfr_bp = opaque_arg;
5076 	unsigned long rep_based_pf;
5077 	char *end = NULL;
5078 
5079 	if (!value || !opaque_arg) {
5080 		PMD_DRV_LOG(ERR,
5081 			    "Invalid parameter passed to rep_based_pf "
5082 			    "devargs.\n");
5083 		return -EINVAL;
5084 	}
5085 
5086 	rep_based_pf = strtoul(value, &end, 10);
5087 	if (end == NULL || *end != '\0' ||
5088 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5089 		PMD_DRV_LOG(ERR,
5090 			    "Invalid parameter passed to rep_based_pf "
5091 			    "devargs.\n");
5092 		return -EINVAL;
5093 	}
5094 
5095 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5096 		PMD_DRV_LOG(ERR,
5097 			    "Invalid value passed to rep_based_pf devargs.\n");
5098 		return -EINVAL;
5099 	}
5100 
5101 	vfr_bp->rep_based_pf = rep_based_pf;
5102 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5103 
5104 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5105 
5106 	return 0;
5107 }
5108 
5109 static int
5110 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5111 			    const char *value, void *opaque_arg)
5112 {
5113 	struct bnxt_representor *vfr_bp = opaque_arg;
5114 	unsigned long rep_q_r2f;
5115 	char *end = NULL;
5116 
5117 	if (!value || !opaque_arg) {
5118 		PMD_DRV_LOG(ERR,
5119 			    "Invalid parameter passed to rep_q_r2f "
5120 			    "devargs.\n");
5121 		return -EINVAL;
5122 	}
5123 
5124 	rep_q_r2f = strtoul(value, &end, 10);
5125 	if (end == NULL || *end != '\0' ||
5126 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5127 		PMD_DRV_LOG(ERR,
5128 			    "Invalid parameter passed to rep_q_r2f "
5129 			    "devargs.\n");
5130 		return -EINVAL;
5131 	}
5132 
5133 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5134 		PMD_DRV_LOG(ERR,
5135 			    "Invalid value passed to rep_q_r2f devargs.\n");
5136 		return -EINVAL;
5137 	}
5138 
5139 	vfr_bp->rep_q_r2f = rep_q_r2f;
5140 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5141 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5142 
5143 	return 0;
5144 }
5145 
5146 static int
5147 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5148 			    const char *value, void *opaque_arg)
5149 {
5150 	struct bnxt_representor *vfr_bp = opaque_arg;
5151 	unsigned long rep_q_f2r;
5152 	char *end = NULL;
5153 
5154 	if (!value || !opaque_arg) {
5155 		PMD_DRV_LOG(ERR,
5156 			    "Invalid parameter passed to rep_q_f2r "
5157 			    "devargs.\n");
5158 		return -EINVAL;
5159 	}
5160 
5161 	rep_q_f2r = strtoul(value, &end, 10);
5162 	if (end == NULL || *end != '\0' ||
5163 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5164 		PMD_DRV_LOG(ERR,
5165 			    "Invalid parameter passed to rep_q_f2r "
5166 			    "devargs.\n");
5167 		return -EINVAL;
5168 	}
5169 
5170 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5171 		PMD_DRV_LOG(ERR,
5172 			    "Invalid value passed to rep_q_f2r devargs.\n");
5173 		return -EINVAL;
5174 	}
5175 
5176 	vfr_bp->rep_q_f2r = rep_q_f2r;
5177 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5178 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5179 
5180 	return 0;
5181 }
5182 
5183 static int
5184 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5185 			     const char *value, void *opaque_arg)
5186 {
5187 	struct bnxt_representor *vfr_bp = opaque_arg;
5188 	unsigned long rep_fc_r2f;
5189 	char *end = NULL;
5190 
5191 	if (!value || !opaque_arg) {
5192 		PMD_DRV_LOG(ERR,
5193 			    "Invalid parameter passed to rep_fc_r2f "
5194 			    "devargs.\n");
5195 		return -EINVAL;
5196 	}
5197 
5198 	rep_fc_r2f = strtoul(value, &end, 10);
5199 	if (end == NULL || *end != '\0' ||
5200 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5201 		PMD_DRV_LOG(ERR,
5202 			    "Invalid parameter passed to rep_fc_r2f "
5203 			    "devargs.\n");
5204 		return -EINVAL;
5205 	}
5206 
5207 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5208 		PMD_DRV_LOG(ERR,
5209 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5210 		return -EINVAL;
5211 	}
5212 
5213 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5214 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5215 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5216 
5217 	return 0;
5218 }
5219 
5220 static int
5221 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5222 			     const char *value, void *opaque_arg)
5223 {
5224 	struct bnxt_representor *vfr_bp = opaque_arg;
5225 	unsigned long rep_fc_f2r;
5226 	char *end = NULL;
5227 
5228 	if (!value || !opaque_arg) {
5229 		PMD_DRV_LOG(ERR,
5230 			    "Invalid parameter passed to rep_fc_f2r "
5231 			    "devargs.\n");
5232 		return -EINVAL;
5233 	}
5234 
5235 	rep_fc_f2r = strtoul(value, &end, 10);
5236 	if (end == NULL || *end != '\0' ||
5237 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5238 		PMD_DRV_LOG(ERR,
5239 			    "Invalid parameter passed to rep_fc_f2r "
5240 			    "devargs.\n");
5241 		return -EINVAL;
5242 	}
5243 
5244 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5245 		PMD_DRV_LOG(ERR,
5246 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5247 		return -EINVAL;
5248 	}
5249 
5250 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5251 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5252 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5253 
5254 	return 0;
5255 }
5256 
5257 static void
5258 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5259 {
5260 	struct rte_kvargs *kvlist;
5261 
5262 	if (devargs == NULL)
5263 		return;
5264 
5265 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5266 	if (kvlist == NULL)
5267 		return;
5268 
5269 	/*
5270 	 * Handler for "truflow" devarg.
5271 	 * Invoked as for ex: "-a 0000:00:0d.0,host-based-truflow=1"
5272 	 */
5273 	rte_kvargs_process(kvlist, BNXT_DEVARG_TRUFLOW,
5274 			   bnxt_parse_devarg_truflow, bp);
5275 
5276 	/*
5277 	 * Handler for "flow_xstat" devarg.
5278 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5279 	 */
5280 	rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5281 			   bnxt_parse_devarg_flow_xstat, bp);
5282 
5283 	/*
5284 	 * Handler for "max_num_kflows" devarg.
5285 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5286 	 */
5287 	rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5288 			   bnxt_parse_devarg_max_num_kflows, bp);
5289 
5290 	rte_kvargs_free(kvlist);
5291 }
5292 
5293 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5294 {
5295 	int rc = 0;
5296 
5297 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5298 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5299 		if (rc)
5300 			PMD_DRV_LOG(ERR,
5301 				    "Failed to alloc switch domain: %d\n", rc);
5302 		else
5303 			PMD_DRV_LOG(INFO,
5304 				    "Switch domain allocated %d\n",
5305 				    bp->switch_domain_id);
5306 	}
5307 
5308 	return rc;
5309 }
5310 
5311 /* Allocate and initialize various fields in bnxt struct that
5312  * need to be allocated/destroyed only once in the lifetime of the driver
5313  */
5314 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5315 {
5316 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5317 	struct bnxt *bp = eth_dev->data->dev_private;
5318 	int rc = 0;
5319 
5320 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5321 
5322 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5323 		bp->flags |= BNXT_FLAG_VF;
5324 
5325 	if (bnxt_p5_device(pci_dev->id.device_id))
5326 		bp->flags |= BNXT_FLAG_CHIP_P5;
5327 
5328 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5329 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5330 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5331 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5332 		bp->flags |= BNXT_FLAG_STINGRAY;
5333 
5334 	if (BNXT_TRUFLOW_EN(bp)) {
5335 		/* extra mbuf field is required to store CFA code from mark */
5336 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5337 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5338 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5339 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5340 		};
5341 		bnxt_cfa_code_dynfield_offset =
5342 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5343 		if (bnxt_cfa_code_dynfield_offset < 0) {
5344 			PMD_DRV_LOG(ERR,
5345 			    "Failed to register mbuf field for TruFlow mark\n");
5346 			return -rte_errno;
5347 		}
5348 	}
5349 
5350 	rc = bnxt_map_pci_bars(eth_dev);
5351 	if (rc) {
5352 		PMD_DRV_LOG(ERR,
5353 			    "Failed to initialize board rc: %x\n", rc);
5354 		return rc;
5355 	}
5356 
5357 	rc = bnxt_alloc_pf_info(bp);
5358 	if (rc)
5359 		return rc;
5360 
5361 	rc = bnxt_alloc_link_info(bp);
5362 	if (rc)
5363 		return rc;
5364 
5365 	rc = bnxt_alloc_parent_info(bp);
5366 	if (rc)
5367 		return rc;
5368 
5369 	rc = bnxt_alloc_hwrm_resources(bp);
5370 	if (rc) {
5371 		PMD_DRV_LOG(ERR,
5372 			    "Failed to allocate hwrm resource rc: %x\n", rc);
5373 		return rc;
5374 	}
5375 	rc = bnxt_alloc_leds_info(bp);
5376 	if (rc)
5377 		return rc;
5378 
5379 	rc = bnxt_alloc_cos_queues(bp);
5380 	if (rc)
5381 		return rc;
5382 
5383 	rc = bnxt_init_locks(bp);
5384 	if (rc)
5385 		return rc;
5386 
5387 	rc = bnxt_alloc_switch_domain(bp);
5388 	if (rc)
5389 		return rc;
5390 
5391 	return rc;
5392 }
5393 
5394 static int
5395 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5396 {
5397 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5398 	static int version_printed;
5399 	struct bnxt *bp;
5400 	int rc;
5401 
5402 	if (version_printed++ == 0)
5403 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5404 
5405 	eth_dev->dev_ops = &bnxt_dev_ops;
5406 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5407 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5408 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5409 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5410 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5411 
5412 	/*
5413 	 * For secondary processes, we don't initialise any further
5414 	 * as primary has already done this work.
5415 	 */
5416 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5417 		return 0;
5418 
5419 	rte_eth_copy_pci_info(eth_dev, pci_dev);
5420 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5421 
5422 	bp = eth_dev->data->dev_private;
5423 
5424 	/* Parse dev arguments passed on when starting the DPDK application. */
5425 	bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5426 
5427 	rc = bnxt_drv_init(eth_dev);
5428 	if (rc)
5429 		goto error_free;
5430 
5431 	rc = bnxt_init_resources(bp, false);
5432 	if (rc)
5433 		goto error_free;
5434 
5435 	rc = bnxt_alloc_stats_mem(bp);
5436 	if (rc)
5437 		goto error_free;
5438 
5439 	PMD_DRV_LOG(INFO,
5440 		    DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
5441 		    pci_dev->mem_resource[0].phys_addr,
5442 		    pci_dev->mem_resource[0].addr);
5443 
5444 	return 0;
5445 
5446 error_free:
5447 	bnxt_dev_uninit(eth_dev);
5448 	return rc;
5449 }
5450 
5451 
5452 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5453 {
5454 	if (!ctx)
5455 		return;
5456 
5457 	if (ctx->va)
5458 		rte_free(ctx->va);
5459 
5460 	ctx->va = NULL;
5461 	ctx->dma = RTE_BAD_IOVA;
5462 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5463 }
5464 
5465 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5466 {
5467 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5468 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5469 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
5470 				  bp->flow_stat->max_fc,
5471 				  false);
5472 
5473 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5474 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5475 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
5476 				  bp->flow_stat->max_fc,
5477 				  false);
5478 
5479 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5480 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5481 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5482 
5483 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5484 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5485 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5486 
5487 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5488 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5489 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5490 
5491 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5492 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5493 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5494 }
5495 
5496 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5497 {
5498 	bnxt_unregister_fc_ctx_mem(bp);
5499 
5500 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5501 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5502 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5503 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5504 }
5505 
5506 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5507 {
5508 	if (BNXT_FLOW_XSTATS_EN(bp))
5509 		bnxt_uninit_fc_ctx_mem(bp);
5510 }
5511 
5512 static void
5513 bnxt_free_error_recovery_info(struct bnxt *bp)
5514 {
5515 	rte_free(bp->recovery_info);
5516 	bp->recovery_info = NULL;
5517 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5518 }
5519 
5520 static int
5521 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5522 {
5523 	int rc;
5524 
5525 	bnxt_free_int(bp);
5526 	bnxt_free_mem(bp, reconfig_dev);
5527 
5528 	bnxt_hwrm_func_buf_unrgtr(bp);
5529 	rte_free(bp->pf->vf_req_buf);
5530 
5531 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
5532 	bp->flags &= ~BNXT_FLAG_REGISTERED;
5533 	bnxt_free_ctx_mem(bp);
5534 	if (!reconfig_dev) {
5535 		bnxt_free_hwrm_resources(bp);
5536 		bnxt_free_error_recovery_info(bp);
5537 	}
5538 
5539 	bnxt_uninit_ctx_mem(bp);
5540 
5541 	bnxt_free_flow_stats_info(bp);
5542 	bnxt_free_rep_info(bp);
5543 	rte_free(bp->ptp_cfg);
5544 	bp->ptp_cfg = NULL;
5545 	return rc;
5546 }
5547 
5548 static int
5549 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5550 {
5551 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5552 		return -EPERM;
5553 
5554 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5555 
5556 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5557 		bnxt_dev_close_op(eth_dev);
5558 
5559 	return 0;
5560 }
5561 
5562 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
5563 {
5564 	struct bnxt *bp = eth_dev->data->dev_private;
5565 	struct rte_eth_dev *vf_rep_eth_dev;
5566 	int ret = 0, i;
5567 
5568 	if (!bp)
5569 		return -EINVAL;
5570 
5571 	for (i = 0; i < bp->num_reps; i++) {
5572 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
5573 		if (!vf_rep_eth_dev)
5574 			continue;
5575 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
5576 			    vf_rep_eth_dev->data->port_id);
5577 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
5578 	}
5579 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
5580 		    eth_dev->data->port_id);
5581 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
5582 
5583 	return ret;
5584 }
5585 
5586 static void bnxt_free_rep_info(struct bnxt *bp)
5587 {
5588 	rte_free(bp->rep_info);
5589 	bp->rep_info = NULL;
5590 	rte_free(bp->cfa_code_map);
5591 	bp->cfa_code_map = NULL;
5592 }
5593 
5594 static int bnxt_init_rep_info(struct bnxt *bp)
5595 {
5596 	int i = 0, rc;
5597 
5598 	if (bp->rep_info)
5599 		return 0;
5600 
5601 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
5602 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
5603 				   0);
5604 	if (!bp->rep_info) {
5605 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
5606 		return -ENOMEM;
5607 	}
5608 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
5609 				       sizeof(*bp->cfa_code_map) *
5610 				       BNXT_MAX_CFA_CODE, 0);
5611 	if (!bp->cfa_code_map) {
5612 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
5613 		bnxt_free_rep_info(bp);
5614 		return -ENOMEM;
5615 	}
5616 
5617 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
5618 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
5619 
5620 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
5621 	if (rc) {
5622 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
5623 		bnxt_free_rep_info(bp);
5624 		return rc;
5625 	}
5626 
5627 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
5628 	if (rc) {
5629 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
5630 		bnxt_free_rep_info(bp);
5631 		return rc;
5632 	}
5633 
5634 	return rc;
5635 }
5636 
5637 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
5638 			       struct rte_eth_devargs *eth_da,
5639 			       struct rte_eth_dev *backing_eth_dev,
5640 			       const char *dev_args)
5641 {
5642 	struct rte_eth_dev *vf_rep_eth_dev;
5643 	char name[RTE_ETH_NAME_MAX_LEN];
5644 	struct bnxt *backing_bp;
5645 	uint16_t num_rep;
5646 	int i, ret = 0;
5647 	struct rte_kvargs *kvlist = NULL;
5648 
5649 	num_rep = eth_da->nb_representor_ports;
5650 	if (num_rep > BNXT_MAX_VF_REPS) {
5651 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
5652 			    num_rep, BNXT_MAX_VF_REPS);
5653 		return -EINVAL;
5654 	}
5655 
5656 	if (num_rep >= RTE_MAX_ETHPORTS) {
5657 		PMD_DRV_LOG(ERR,
5658 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
5659 			    num_rep, RTE_MAX_ETHPORTS);
5660 		return -EINVAL;
5661 	}
5662 
5663 	backing_bp = backing_eth_dev->data->dev_private;
5664 
5665 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
5666 		PMD_DRV_LOG(ERR,
5667 			    "Not a PF or trusted VF. No Representor support\n");
5668 		/* Returning an error is not an option.
5669 		 * Applications are not handling this correctly
5670 		 */
5671 		return 0;
5672 	}
5673 
5674 	if (bnxt_init_rep_info(backing_bp))
5675 		return 0;
5676 
5677 	for (i = 0; i < num_rep; i++) {
5678 		struct bnxt_representor representor = {
5679 			.vf_id = eth_da->representor_ports[i],
5680 			.switch_domain_id = backing_bp->switch_domain_id,
5681 			.parent_dev = backing_eth_dev
5682 		};
5683 
5684 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
5685 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
5686 				    representor.vf_id, BNXT_MAX_VF_REPS);
5687 			continue;
5688 		}
5689 
5690 		/* representor port net_bdf_port */
5691 		snprintf(name, sizeof(name), "net_%s_representor_%d",
5692 			 pci_dev->device.name, eth_da->representor_ports[i]);
5693 
5694 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
5695 		if (kvlist) {
5696 			/*
5697 			 * Handler for "rep_is_pf" devarg.
5698 			 * Invoked as for ex: "-a 000:00:0d.0,
5699 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5700 			 */
5701 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
5702 						 bnxt_parse_devarg_rep_is_pf,
5703 						 (void *)&representor);
5704 			if (ret) {
5705 				ret = -EINVAL;
5706 				goto err;
5707 			}
5708 			/*
5709 			 * Handler for "rep_based_pf" devarg.
5710 			 * Invoked as for ex: "-a 000:00:0d.0,
5711 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5712 			 */
5713 			ret = rte_kvargs_process(kvlist,
5714 						 BNXT_DEVARG_REP_BASED_PF,
5715 						 bnxt_parse_devarg_rep_based_pf,
5716 						 (void *)&representor);
5717 			if (ret) {
5718 				ret = -EINVAL;
5719 				goto err;
5720 			}
5721 			/*
5722 			 * Handler for "rep_based_pf" devarg.
5723 			 * Invoked as for ex: "-a 000:00:0d.0,
5724 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5725 			 */
5726 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
5727 						 bnxt_parse_devarg_rep_q_r2f,
5728 						 (void *)&representor);
5729 			if (ret) {
5730 				ret = -EINVAL;
5731 				goto err;
5732 			}
5733 			/*
5734 			 * Handler for "rep_based_pf" devarg.
5735 			 * Invoked as for ex: "-a 000:00:0d.0,
5736 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5737 			 */
5738 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
5739 						 bnxt_parse_devarg_rep_q_f2r,
5740 						 (void *)&representor);
5741 			if (ret) {
5742 				ret = -EINVAL;
5743 				goto err;
5744 			}
5745 			/*
5746 			 * Handler for "rep_based_pf" devarg.
5747 			 * Invoked as for ex: "-a 000:00:0d.0,
5748 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5749 			 */
5750 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
5751 						 bnxt_parse_devarg_rep_fc_r2f,
5752 						 (void *)&representor);
5753 			if (ret) {
5754 				ret = -EINVAL;
5755 				goto err;
5756 			}
5757 			/*
5758 			 * Handler for "rep_based_pf" devarg.
5759 			 * Invoked as for ex: "-a 000:00:0d.0,
5760 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5761 			 */
5762 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
5763 						 bnxt_parse_devarg_rep_fc_f2r,
5764 						 (void *)&representor);
5765 			if (ret) {
5766 				ret = -EINVAL;
5767 				goto err;
5768 			}
5769 		}
5770 
5771 		ret = rte_eth_dev_create(&pci_dev->device, name,
5772 					 sizeof(struct bnxt_representor),
5773 					 NULL, NULL,
5774 					 bnxt_representor_init,
5775 					 &representor);
5776 		if (ret) {
5777 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
5778 				    "representor %s.", name);
5779 			goto err;
5780 		}
5781 
5782 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
5783 		if (!vf_rep_eth_dev) {
5784 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
5785 				    " for VF-Rep: %s.", name);
5786 			ret = -ENODEV;
5787 			goto err;
5788 		}
5789 
5790 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
5791 			    backing_eth_dev->data->port_id);
5792 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
5793 							 vf_rep_eth_dev;
5794 		backing_bp->num_reps++;
5795 
5796 	}
5797 
5798 	rte_kvargs_free(kvlist);
5799 	return 0;
5800 
5801 err:
5802 	/* If num_rep > 1, then rollback already created
5803 	 * ports, since we'll be failing the probe anyway
5804 	 */
5805 	if (num_rep > 1)
5806 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
5807 	rte_errno = -ret;
5808 	rte_kvargs_free(kvlist);
5809 
5810 	return ret;
5811 }
5812 
5813 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
5814 			  struct rte_pci_device *pci_dev)
5815 {
5816 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
5817 	struct rte_eth_dev *backing_eth_dev;
5818 	uint16_t num_rep;
5819 	int ret = 0;
5820 
5821 	if (pci_dev->device.devargs) {
5822 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
5823 					    &eth_da);
5824 		if (ret)
5825 			return ret;
5826 	}
5827 
5828 	num_rep = eth_da.nb_representor_ports;
5829 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
5830 		    num_rep);
5831 
5832 	/* We could come here after first level of probe is already invoked
5833 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
5834 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
5835 	 */
5836 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
5837 	if (backing_eth_dev == NULL) {
5838 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
5839 					 sizeof(struct bnxt),
5840 					 eth_dev_pci_specific_init, pci_dev,
5841 					 bnxt_dev_init, NULL);
5842 
5843 		if (ret || !num_rep)
5844 			return ret;
5845 
5846 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
5847 	}
5848 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
5849 		    backing_eth_dev->data->port_id);
5850 
5851 	if (!num_rep)
5852 		return ret;
5853 
5854 	/* probe representor ports now */
5855 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
5856 				  pci_dev->device.devargs->args);
5857 
5858 	return ret;
5859 }
5860 
5861 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
5862 {
5863 	struct rte_eth_dev *eth_dev;
5864 
5865 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
5866 	if (!eth_dev)
5867 		return 0; /* Invoked typically only by OVS-DPDK, by the
5868 			   * time it comes here the eth_dev is already
5869 			   * deleted by rte_eth_dev_close(), so returning
5870 			   * +ve value will at least help in proper cleanup
5871 			   */
5872 
5873 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
5874 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
5875 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
5876 			return rte_eth_dev_destroy(eth_dev,
5877 						   bnxt_representor_uninit);
5878 		else
5879 			return rte_eth_dev_destroy(eth_dev,
5880 						   bnxt_dev_uninit);
5881 	} else {
5882 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
5883 	}
5884 }
5885 
5886 static struct rte_pci_driver bnxt_rte_pmd = {
5887 	.id_table = bnxt_pci_id_map,
5888 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
5889 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
5890 						  * and OVS-DPDK
5891 						  */
5892 	.probe = bnxt_pci_probe,
5893 	.remove = bnxt_pci_remove,
5894 };
5895 
5896 static bool
5897 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
5898 {
5899 	if (strcmp(dev->device->driver->name, drv->driver.name))
5900 		return false;
5901 
5902 	return true;
5903 }
5904 
5905 bool is_bnxt_supported(struct rte_eth_dev *dev)
5906 {
5907 	return is_device_supported(dev, &bnxt_rte_pmd);
5908 }
5909 
5910 RTE_LOG_REGISTER(bnxt_logtype_driver, pmd.net.bnxt.driver, NOTICE);
5911 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
5912 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
5913 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
5914