xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision ba57777d)
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_ACCUM_STATS	"accum-stats"
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 #define BNXT_DEVARG_APP_ID	"app-id"
101 
102 static const char *const bnxt_dev_args[] = {
103 	BNXT_DEVARG_REPRESENTOR,
104 	BNXT_DEVARG_ACCUM_STATS,
105 	BNXT_DEVARG_FLOW_XSTAT,
106 	BNXT_DEVARG_MAX_NUM_KFLOWS,
107 	BNXT_DEVARG_REP_BASED_PF,
108 	BNXT_DEVARG_REP_IS_PF,
109 	BNXT_DEVARG_REP_Q_R2F,
110 	BNXT_DEVARG_REP_Q_F2R,
111 	BNXT_DEVARG_REP_FC_R2F,
112 	BNXT_DEVARG_REP_FC_F2R,
113 	BNXT_DEVARG_APP_ID,
114 	NULL
115 };
116 
117 /*
118  * accum-stats == false to disable flow counter accumulation
119  * accum-stats == true to enable flow counter accumulation
120  */
121 #define	BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)	((accum_stats) > 1)
122 
123 /*
124  * app-id = an non-negative 8-bit number
125  */
126 #define BNXT_DEVARG_APP_ID_INVALID(val)			((val) > 255)
127 
128 /*
129  * flow_xstat == false to disable the feature
130  * flow_xstat == true to enable the feature
131  */
132 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
133 
134 /*
135  * rep_is_pf == false to indicate VF representor
136  * rep_is_pf == true to indicate PF representor
137  */
138 #define	BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)	((rep_is_pf) > 1)
139 
140 /*
141  * rep_based_pf == Physical index of the PF
142  */
143 #define	BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)	((rep_based_pf) > 15)
144 /*
145  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
146  */
147 #define	BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)	((rep_q_r2f) > 3)
148 
149 /*
150  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
151  */
152 #define	BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)	((rep_q_f2r) > 3)
153 
154 /*
155  * rep_fc_r2f == Flow control for the representor to endpoint direction
156  */
157 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)	((rep_fc_r2f) > 1)
158 
159 /*
160  * rep_fc_f2r == Flow control for the endpoint to representor direction
161  */
162 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)	((rep_fc_f2r) > 1)
163 
164 int bnxt_cfa_code_dynfield_offset = -1;
165 
166 /*
167  * max_num_kflows must be >= 32
168  * and must be a power-of-2 supported value
169  * return: 1 -> invalid
170  *         0 -> valid
171  */
172 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
173 {
174 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
175 		return 1;
176 	return 0;
177 }
178 
179 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
180 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
181 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
182 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
183 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
184 static int bnxt_restore_vlan_filters(struct bnxt *bp);
185 static void bnxt_dev_recover(void *arg);
186 static void bnxt_free_error_recovery_info(struct bnxt *bp);
187 static void bnxt_free_rep_info(struct bnxt *bp);
188 
189 int is_bnxt_in_error(struct bnxt *bp)
190 {
191 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
192 		return -EIO;
193 	if (bp->flags & BNXT_FLAG_FW_RESET)
194 		return -EBUSY;
195 
196 	return 0;
197 }
198 
199 /***********************/
200 
201 /*
202  * High level utility functions
203  */
204 
205 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
206 {
207 	unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
208 					     BNXT_RSS_TBL_SIZE_P5);
209 
210 	if (!BNXT_CHIP_P5(bp))
211 		return 1;
212 
213 	return RTE_ALIGN_MUL_CEIL(num_rss_rings,
214 				  BNXT_RSS_ENTRIES_PER_CTX_P5) /
215 				  BNXT_RSS_ENTRIES_PER_CTX_P5;
216 }
217 
218 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
219 {
220 	if (!BNXT_CHIP_P5(bp))
221 		return HW_HASH_INDEX_SIZE;
222 
223 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
224 }
225 
226 static void bnxt_free_parent_info(struct bnxt *bp)
227 {
228 	rte_free(bp->parent);
229 	bp->parent = NULL;
230 }
231 
232 static void bnxt_free_pf_info(struct bnxt *bp)
233 {
234 	rte_free(bp->pf);
235 	bp->pf = NULL;
236 }
237 
238 static void bnxt_free_link_info(struct bnxt *bp)
239 {
240 	rte_free(bp->link_info);
241 	bp->link_info = NULL;
242 }
243 
244 static void bnxt_free_leds_info(struct bnxt *bp)
245 {
246 	if (BNXT_VF(bp))
247 		return;
248 
249 	rte_free(bp->leds);
250 	bp->leds = NULL;
251 }
252 
253 static void bnxt_free_flow_stats_info(struct bnxt *bp)
254 {
255 	rte_free(bp->flow_stat);
256 	bp->flow_stat = NULL;
257 }
258 
259 static void bnxt_free_cos_queues(struct bnxt *bp)
260 {
261 	rte_free(bp->rx_cos_queue);
262 	bp->rx_cos_queue = NULL;
263 	rte_free(bp->tx_cos_queue);
264 	bp->tx_cos_queue = NULL;
265 }
266 
267 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
268 {
269 	bnxt_free_filter_mem(bp);
270 	bnxt_free_vnic_attributes(bp);
271 	bnxt_free_vnic_mem(bp);
272 
273 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
274 	 * If the number of rings change across fw update,
275 	 * we don't have much choice except to warn the user.
276 	 */
277 	if (!reconfig) {
278 		bnxt_free_stats(bp);
279 		bnxt_free_tx_rings(bp);
280 		bnxt_free_rx_rings(bp);
281 	}
282 	bnxt_free_async_cp_ring(bp);
283 	bnxt_free_rxtx_nq_ring(bp);
284 
285 	rte_free(bp->grp_info);
286 	bp->grp_info = NULL;
287 }
288 
289 static int bnxt_alloc_parent_info(struct bnxt *bp)
290 {
291 	bp->parent = rte_zmalloc("bnxt_parent_info",
292 				 sizeof(struct bnxt_parent_info), 0);
293 	if (bp->parent == NULL)
294 		return -ENOMEM;
295 
296 	return 0;
297 }
298 
299 static int bnxt_alloc_pf_info(struct bnxt *bp)
300 {
301 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
302 	if (bp->pf == NULL)
303 		return -ENOMEM;
304 
305 	return 0;
306 }
307 
308 static int bnxt_alloc_link_info(struct bnxt *bp)
309 {
310 	bp->link_info =
311 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
312 	if (bp->link_info == NULL)
313 		return -ENOMEM;
314 
315 	return 0;
316 }
317 
318 static int bnxt_alloc_leds_info(struct bnxt *bp)
319 {
320 	if (BNXT_VF(bp))
321 		return 0;
322 
323 	bp->leds = rte_zmalloc("bnxt_leds",
324 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
325 			       0);
326 	if (bp->leds == NULL)
327 		return -ENOMEM;
328 
329 	return 0;
330 }
331 
332 static int bnxt_alloc_cos_queues(struct bnxt *bp)
333 {
334 	bp->rx_cos_queue =
335 		rte_zmalloc("bnxt_rx_cosq",
336 			    BNXT_COS_QUEUE_COUNT *
337 			    sizeof(struct bnxt_cos_queue_info),
338 			    0);
339 	if (bp->rx_cos_queue == NULL)
340 		return -ENOMEM;
341 
342 	bp->tx_cos_queue =
343 		rte_zmalloc("bnxt_tx_cosq",
344 			    BNXT_COS_QUEUE_COUNT *
345 			    sizeof(struct bnxt_cos_queue_info),
346 			    0);
347 	if (bp->tx_cos_queue == NULL)
348 		return -ENOMEM;
349 
350 	return 0;
351 }
352 
353 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
354 {
355 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
356 				    sizeof(struct bnxt_flow_stat_info), 0);
357 	if (bp->flow_stat == NULL)
358 		return -ENOMEM;
359 
360 	return 0;
361 }
362 
363 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
364 {
365 	int rc;
366 
367 	rc = bnxt_alloc_ring_grps(bp);
368 	if (rc)
369 		goto alloc_mem_err;
370 
371 	rc = bnxt_alloc_async_ring_struct(bp);
372 	if (rc)
373 		goto alloc_mem_err;
374 
375 	rc = bnxt_alloc_vnic_mem(bp);
376 	if (rc)
377 		goto alloc_mem_err;
378 
379 	rc = bnxt_alloc_vnic_attributes(bp);
380 	if (rc)
381 		goto alloc_mem_err;
382 
383 	rc = bnxt_alloc_filter_mem(bp);
384 	if (rc)
385 		goto alloc_mem_err;
386 
387 	rc = bnxt_alloc_async_cp_ring(bp);
388 	if (rc)
389 		goto alloc_mem_err;
390 
391 	rc = bnxt_alloc_rxtx_nq_ring(bp);
392 	if (rc)
393 		goto alloc_mem_err;
394 
395 	if (BNXT_FLOW_XSTATS_EN(bp)) {
396 		rc = bnxt_alloc_flow_stats_info(bp);
397 		if (rc)
398 			goto alloc_mem_err;
399 	}
400 
401 	return 0;
402 
403 alloc_mem_err:
404 	bnxt_free_mem(bp, reconfig);
405 	return rc;
406 }
407 
408 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
409 {
410 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
411 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
412 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
413 	struct bnxt_rx_queue *rxq;
414 	unsigned int j;
415 	int rc;
416 
417 	rc = bnxt_vnic_grp_alloc(bp, vnic);
418 	if (rc)
419 		goto err_out;
420 
421 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
422 		    vnic_id, vnic, vnic->fw_grp_ids);
423 
424 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
425 	if (rc)
426 		goto err_out;
427 
428 	/* Alloc RSS context only if RSS mode is enabled */
429 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
430 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
431 
432 		/* RSS table size in Thor is 512.
433 		 * Cap max Rx rings to same value
434 		 */
435 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
436 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
437 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
438 			goto err_out;
439 		}
440 
441 		rc = 0;
442 		for (j = 0; j < nr_ctxs; j++) {
443 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
444 			if (rc)
445 				break;
446 		}
447 		if (rc) {
448 			PMD_DRV_LOG(ERR,
449 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
450 				    vnic_id, j, rc);
451 			goto err_out;
452 		}
453 		vnic->num_lb_ctxts = nr_ctxs;
454 	}
455 
456 	/*
457 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
458 	 * setting is not available at this time, it will not be
459 	 * configured correctly in the CFA.
460 	 */
461 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
462 		vnic->vlan_strip = true;
463 	else
464 		vnic->vlan_strip = false;
465 
466 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
467 	if (rc)
468 		goto err_out;
469 
470 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
471 	if (rc)
472 		goto err_out;
473 
474 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
475 		rxq = bp->eth_dev->data->rx_queues[j];
476 
477 		PMD_DRV_LOG(DEBUG,
478 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
479 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
480 
481 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
482 			rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
483 		else
484 			vnic->rx_queue_cnt++;
485 	}
486 
487 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
488 
489 	rc = bnxt_vnic_rss_configure(bp, vnic);
490 	if (rc)
491 		goto err_out;
492 
493 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
494 
495 	rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
496 				    (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO) ?
497 				    true : false);
498 	if (rc)
499 		goto err_out;
500 
501 	return 0;
502 err_out:
503 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
504 		    vnic_id, rc);
505 	return rc;
506 }
507 
508 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
509 {
510 	int rc = 0;
511 
512 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
513 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
514 	if (rc)
515 		return rc;
516 
517 	PMD_DRV_LOG(DEBUG,
518 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
519 		    " rx_fc_in_tbl.ctx_id = %d\n",
520 		    bp->flow_stat->rx_fc_in_tbl.va,
521 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
522 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
523 
524 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
525 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
526 	if (rc)
527 		return rc;
528 
529 	PMD_DRV_LOG(DEBUG,
530 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
531 		    " rx_fc_out_tbl.ctx_id = %d\n",
532 		    bp->flow_stat->rx_fc_out_tbl.va,
533 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
534 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
535 
536 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
537 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
538 	if (rc)
539 		return rc;
540 
541 	PMD_DRV_LOG(DEBUG,
542 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
543 		    " tx_fc_in_tbl.ctx_id = %d\n",
544 		    bp->flow_stat->tx_fc_in_tbl.va,
545 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
546 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
547 
548 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
549 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
550 	if (rc)
551 		return rc;
552 
553 	PMD_DRV_LOG(DEBUG,
554 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
555 		    " tx_fc_out_tbl.ctx_id = %d\n",
556 		    bp->flow_stat->tx_fc_out_tbl.va,
557 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
558 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
559 
560 	memset(bp->flow_stat->rx_fc_out_tbl.va,
561 	       0,
562 	       bp->flow_stat->rx_fc_out_tbl.size);
563 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
564 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
565 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
566 				       bp->flow_stat->max_fc,
567 				       true);
568 	if (rc)
569 		return rc;
570 
571 	memset(bp->flow_stat->tx_fc_out_tbl.va,
572 	       0,
573 	       bp->flow_stat->tx_fc_out_tbl.size);
574 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
575 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
576 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
577 				       bp->flow_stat->max_fc,
578 				       true);
579 
580 	return rc;
581 }
582 
583 static int bnxt_alloc_ctx_mem_buf(struct bnxt *bp, char *type, size_t size,
584 				  struct bnxt_ctx_mem_buf_info *ctx)
585 {
586 	if (!ctx)
587 		return -EINVAL;
588 
589 	ctx->va = rte_zmalloc_socket(type, size, 0,
590 				     bp->eth_dev->device->numa_node);
591 	if (ctx->va == NULL)
592 		return -ENOMEM;
593 	rte_mem_lock_page(ctx->va);
594 	ctx->size = size;
595 	ctx->dma = rte_mem_virt2iova(ctx->va);
596 	if (ctx->dma == RTE_BAD_IOVA)
597 		return -ENOMEM;
598 
599 	return 0;
600 }
601 
602 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
603 {
604 	struct rte_pci_device *pdev = bp->pdev;
605 	char type[RTE_MEMZONE_NAMESIZE];
606 	uint16_t max_fc;
607 	int rc = 0;
608 
609 	max_fc = bp->flow_stat->max_fc;
610 
611 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
612 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
613 	/* 4 bytes for each counter-id */
614 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
615 				    max_fc * 4,
616 				    &bp->flow_stat->rx_fc_in_tbl);
617 	if (rc)
618 		return rc;
619 
620 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
621 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
622 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
623 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
624 				    max_fc * 16,
625 				    &bp->flow_stat->rx_fc_out_tbl);
626 	if (rc)
627 		return rc;
628 
629 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
630 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
631 	/* 4 bytes for each counter-id */
632 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
633 				    max_fc * 4,
634 				    &bp->flow_stat->tx_fc_in_tbl);
635 	if (rc)
636 		return rc;
637 
638 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
639 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
640 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
641 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
642 				    max_fc * 16,
643 				    &bp->flow_stat->tx_fc_out_tbl);
644 	if (rc)
645 		return rc;
646 
647 	rc = bnxt_register_fc_ctx_mem(bp);
648 
649 	return rc;
650 }
651 
652 static int bnxt_init_ctx_mem(struct bnxt *bp)
653 {
654 	int rc = 0;
655 
656 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
657 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
658 	    !BNXT_FLOW_XSTATS_EN(bp))
659 		return 0;
660 
661 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
662 	if (rc)
663 		return rc;
664 
665 	rc = bnxt_init_fc_ctx_mem(bp);
666 
667 	return rc;
668 }
669 
670 static int bnxt_update_phy_setting(struct bnxt *bp)
671 {
672 	struct rte_eth_link new;
673 	int rc;
674 
675 	rc = bnxt_get_hwrm_link_config(bp, &new);
676 	if (rc) {
677 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
678 		return rc;
679 	}
680 
681 	/*
682 	 * On BCM957508-N2100 adapters, FW will not allow any user other
683 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
684 	 * always returns link up. Force phy update always in that case.
685 	 */
686 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
687 		rc = bnxt_set_hwrm_link_config(bp, true);
688 		if (rc) {
689 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
690 			return rc;
691 		}
692 	}
693 
694 	return rc;
695 }
696 
697 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
698 {
699 	rte_free(bp->prev_rx_ring_stats);
700 	rte_free(bp->prev_tx_ring_stats);
701 
702 	bp->prev_rx_ring_stats = NULL;
703 	bp->prev_tx_ring_stats = NULL;
704 }
705 
706 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
707 {
708 	bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
709 					      sizeof(struct bnxt_ring_stats) *
710 					      bp->rx_cp_nr_rings,
711 					      0);
712 	if (bp->prev_rx_ring_stats == NULL)
713 		return -ENOMEM;
714 
715 	bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
716 					     sizeof(struct bnxt_ring_stats) *
717 					     bp->tx_cp_nr_rings,
718 					     0);
719 	if (bp->prev_tx_ring_stats == NULL)
720 		goto error;
721 
722 	return 0;
723 
724 error:
725 	bnxt_free_prev_ring_stats(bp);
726 	return -ENOMEM;
727 }
728 
729 static int bnxt_start_nic(struct bnxt *bp)
730 {
731 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
732 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
733 	uint32_t intr_vector = 0;
734 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
735 	uint32_t vec = BNXT_MISC_VEC_ID;
736 	unsigned int i, j;
737 	int rc;
738 
739 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
740 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
741 			DEV_RX_OFFLOAD_JUMBO_FRAME;
742 		bp->flags |= BNXT_FLAG_JUMBO;
743 	} else {
744 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
745 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
746 		bp->flags &= ~BNXT_FLAG_JUMBO;
747 	}
748 
749 	/* THOR does not support ring groups.
750 	 * But we will use the array to save RSS context IDs.
751 	 */
752 	if (BNXT_CHIP_P5(bp))
753 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
754 
755 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
756 	if (rc) {
757 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
758 		goto err_out;
759 	}
760 
761 	rc = bnxt_alloc_hwrm_rings(bp);
762 	if (rc) {
763 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
764 		goto err_out;
765 	}
766 
767 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
768 	if (rc) {
769 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
770 		goto err_out;
771 	}
772 
773 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
774 		goto skip_cosq_cfg;
775 
776 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
777 		if (bp->rx_cos_queue[i].id != 0xff) {
778 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
779 
780 			if (!vnic) {
781 				PMD_DRV_LOG(ERR,
782 					    "Num pools more than FW profile\n");
783 				rc = -EINVAL;
784 				goto err_out;
785 			}
786 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
787 			bp->rx_cosq_cnt++;
788 		}
789 	}
790 
791 skip_cosq_cfg:
792 	rc = bnxt_mq_rx_configure(bp);
793 	if (rc) {
794 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
795 		goto err_out;
796 	}
797 
798 	/* default vnic 0 */
799 	rc = bnxt_setup_one_vnic(bp, 0);
800 	if (rc)
801 		goto err_out;
802 	/* VNIC configuration */
803 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
804 		for (i = 1; i < bp->nr_vnics; i++) {
805 			rc = bnxt_setup_one_vnic(bp, i);
806 			if (rc)
807 				goto err_out;
808 		}
809 	}
810 
811 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
812 	if (rc) {
813 		PMD_DRV_LOG(ERR,
814 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
815 		goto err_out;
816 	}
817 
818 	/* check and configure queue intr-vector mapping */
819 	if ((rte_intr_cap_multiple(intr_handle) ||
820 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
821 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
822 		intr_vector = bp->eth_dev->data->nb_rx_queues;
823 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
824 		if (intr_vector > bp->rx_cp_nr_rings) {
825 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
826 					bp->rx_cp_nr_rings);
827 			return -ENOTSUP;
828 		}
829 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
830 		if (rc)
831 			return rc;
832 	}
833 
834 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
835 		intr_handle->intr_vec =
836 			rte_zmalloc("intr_vec",
837 				    bp->eth_dev->data->nb_rx_queues *
838 				    sizeof(int), 0);
839 		if (intr_handle->intr_vec == NULL) {
840 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
841 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
842 			rc = -ENOMEM;
843 			goto err_out;
844 		}
845 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
846 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
847 			 intr_handle->intr_vec, intr_handle->nb_efd,
848 			intr_handle->max_intr);
849 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
850 		     queue_id++) {
851 			intr_handle->intr_vec[queue_id] =
852 							vec + BNXT_RX_VEC_START;
853 			if (vec < base + intr_handle->nb_efd - 1)
854 				vec++;
855 		}
856 	}
857 
858 	/* enable uio/vfio intr/eventfd mapping */
859 	rc = rte_intr_enable(intr_handle);
860 #ifndef RTE_EXEC_ENV_FREEBSD
861 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
862 	if (rc)
863 		goto err_out;
864 #endif
865 
866 	rc = bnxt_update_phy_setting(bp);
867 	if (rc)
868 		goto err_out;
869 
870 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
871 	if (!bp->mark_table)
872 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
873 
874 	return 0;
875 
876 err_out:
877 	/* Some of the error status returned by FW may not be from errno.h */
878 	if (rc > 0)
879 		rc = -EIO;
880 
881 	return rc;
882 }
883 
884 static int bnxt_shutdown_nic(struct bnxt *bp)
885 {
886 	bnxt_free_all_hwrm_resources(bp);
887 	bnxt_free_all_filters(bp);
888 	bnxt_free_all_vnics(bp);
889 	return 0;
890 }
891 
892 /*
893  * Device configuration and status function
894  */
895 
896 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
897 {
898 	uint32_t link_speed = 0;
899 	uint32_t speed_capa = 0;
900 
901 	if (bp->link_info == NULL)
902 		return 0;
903 
904 	link_speed = bp->link_info->support_speeds;
905 
906 	/* If PAM4 is configured, use PAM4 supported speed */
907 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
908 		link_speed = bp->link_info->support_pam4_speeds;
909 
910 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
911 		speed_capa |= ETH_LINK_SPEED_100M;
912 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
913 		speed_capa |= ETH_LINK_SPEED_100M_HD;
914 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
915 		speed_capa |= ETH_LINK_SPEED_1G;
916 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
917 		speed_capa |= ETH_LINK_SPEED_2_5G;
918 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
919 		speed_capa |= ETH_LINK_SPEED_10G;
920 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
921 		speed_capa |= ETH_LINK_SPEED_20G;
922 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
923 		speed_capa |= ETH_LINK_SPEED_25G;
924 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
925 		speed_capa |= ETH_LINK_SPEED_40G;
926 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
927 		speed_capa |= ETH_LINK_SPEED_50G;
928 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
929 		speed_capa |= ETH_LINK_SPEED_100G;
930 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
931 		speed_capa |= ETH_LINK_SPEED_50G;
932 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
933 		speed_capa |= ETH_LINK_SPEED_100G;
934 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
935 		speed_capa |= ETH_LINK_SPEED_200G;
936 
937 	if (bp->link_info->auto_mode ==
938 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
939 		speed_capa |= ETH_LINK_SPEED_FIXED;
940 
941 	return speed_capa;
942 }
943 
944 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
945 				struct rte_eth_dev_info *dev_info)
946 {
947 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
948 	struct bnxt *bp = eth_dev->data->dev_private;
949 	uint16_t max_vnics, i, j, vpool, vrxq;
950 	unsigned int max_rx_rings;
951 	int rc;
952 
953 	rc = is_bnxt_in_error(bp);
954 	if (rc)
955 		return rc;
956 
957 	/* MAC Specifics */
958 	dev_info->max_mac_addrs = bp->max_l2_ctx;
959 	dev_info->max_hash_mac_addrs = 0;
960 
961 	/* PF/VF specifics */
962 	if (BNXT_PF(bp))
963 		dev_info->max_vfs = pdev->max_vfs;
964 
965 	max_rx_rings = bnxt_max_rings(bp);
966 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
967 	dev_info->max_rx_queues = max_rx_rings;
968 	dev_info->max_tx_queues = max_rx_rings;
969 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
970 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
971 	max_vnics = bp->max_vnics;
972 
973 	/* MTU specifics */
974 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
975 	dev_info->max_mtu = BNXT_MAX_MTU;
976 
977 	/* Fast path specifics */
978 	dev_info->min_rx_bufsize = 1;
979 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
980 
981 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
982 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
983 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
984 	dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
985 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
986 				    dev_info->tx_queue_offload_capa;
987 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
988 
989 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
990 	dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
991 			     RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
992 
993 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
994 		.rx_thresh = {
995 			.pthresh = 8,
996 			.hthresh = 8,
997 			.wthresh = 0,
998 		},
999 		.rx_free_thresh = 32,
1000 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
1001 	};
1002 
1003 	dev_info->default_txconf = (struct rte_eth_txconf) {
1004 		.tx_thresh = {
1005 			.pthresh = 32,
1006 			.hthresh = 0,
1007 			.wthresh = 0,
1008 		},
1009 		.tx_free_thresh = 32,
1010 		.tx_rs_thresh = 32,
1011 	};
1012 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
1013 
1014 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1015 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1016 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1017 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1018 
1019 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1020 		dev_info->switch_info.name = eth_dev->device->name;
1021 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1022 		dev_info->switch_info.port_id =
1023 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1024 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1025 	}
1026 
1027 	/*
1028 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1029 	 *       need further investigation.
1030 	 */
1031 
1032 	/* VMDq resources */
1033 	vpool = 64; /* ETH_64_POOLS */
1034 	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
1035 	for (i = 0; i < 4; vpool >>= 1, i++) {
1036 		if (max_vnics > vpool) {
1037 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1038 				if (dev_info->max_rx_queues > vrxq) {
1039 					if (vpool > vrxq)
1040 						vpool = vrxq;
1041 					goto found;
1042 				}
1043 			}
1044 			/* Not enough resources to support VMDq */
1045 			break;
1046 		}
1047 	}
1048 	/* Not enough resources to support VMDq */
1049 	vpool = 0;
1050 	vrxq = 0;
1051 found:
1052 	dev_info->max_vmdq_pools = vpool;
1053 	dev_info->vmdq_queue_num = vrxq;
1054 
1055 	dev_info->vmdq_pool_base = 0;
1056 	dev_info->vmdq_queue_base = 0;
1057 
1058 	return 0;
1059 }
1060 
1061 /* Configure the device based on the configuration provided */
1062 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1063 {
1064 	struct bnxt *bp = eth_dev->data->dev_private;
1065 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1066 	int rc;
1067 
1068 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1069 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1070 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1071 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1072 
1073 	rc = is_bnxt_in_error(bp);
1074 	if (rc)
1075 		return rc;
1076 
1077 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1078 		rc = bnxt_hwrm_check_vf_rings(bp);
1079 		if (rc) {
1080 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1081 			return -ENOSPC;
1082 		}
1083 
1084 		/* If a resource has already been allocated - in this case
1085 		 * it is the async completion ring, free it. Reallocate it after
1086 		 * resource reservation. This will ensure the resource counts
1087 		 * are calculated correctly.
1088 		 */
1089 
1090 		pthread_mutex_lock(&bp->def_cp_lock);
1091 
1092 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1093 			bnxt_disable_int(bp);
1094 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1095 		}
1096 
1097 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1098 		if (rc) {
1099 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1100 			pthread_mutex_unlock(&bp->def_cp_lock);
1101 			return -ENOSPC;
1102 		}
1103 
1104 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1105 			rc = bnxt_alloc_async_cp_ring(bp);
1106 			if (rc) {
1107 				pthread_mutex_unlock(&bp->def_cp_lock);
1108 				return rc;
1109 			}
1110 			bnxt_enable_int(bp);
1111 		}
1112 
1113 		pthread_mutex_unlock(&bp->def_cp_lock);
1114 	}
1115 
1116 	/* Inherit new configurations */
1117 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1118 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1119 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1120 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1121 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1122 	    bp->max_stat_ctx)
1123 		goto resource_error;
1124 
1125 	if (BNXT_HAS_RING_GRPS(bp) &&
1126 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1127 		goto resource_error;
1128 
1129 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1130 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1131 		goto resource_error;
1132 
1133 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1134 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1135 
1136 	if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1137 		rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1138 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1139 
1140 	if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1141 		eth_dev->data->mtu =
1142 			eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1143 			RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1144 			BNXT_NUM_VLANS;
1145 		bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1146 	}
1147 	return 0;
1148 
1149 resource_error:
1150 	PMD_DRV_LOG(ERR,
1151 		    "Insufficient resources to support requested config\n");
1152 	PMD_DRV_LOG(ERR,
1153 		    "Num Queues Requested: Tx %d, Rx %d\n",
1154 		    eth_dev->data->nb_tx_queues,
1155 		    eth_dev->data->nb_rx_queues);
1156 	PMD_DRV_LOG(ERR,
1157 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1158 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1159 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1160 	return -ENOSPC;
1161 }
1162 
1163 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1164 {
1165 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1166 
1167 	if (link->link_status)
1168 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1169 			eth_dev->data->port_id,
1170 			(uint32_t)link->link_speed,
1171 			(link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1172 			("full-duplex") : ("half-duplex\n"));
1173 	else
1174 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1175 			eth_dev->data->port_id);
1176 }
1177 
1178 /*
1179  * Determine whether the current configuration requires support for scattered
1180  * receive; return 1 if scattered receive is required and 0 if not.
1181  */
1182 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1183 {
1184 	uint16_t buf_size;
1185 	int i;
1186 
1187 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1188 		return 1;
1189 
1190 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1191 		return 1;
1192 
1193 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1194 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1195 
1196 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1197 				      RTE_PKTMBUF_HEADROOM);
1198 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1199 			return 1;
1200 	}
1201 	return 0;
1202 }
1203 
1204 static eth_rx_burst_t
1205 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1206 {
1207 	struct bnxt *bp = eth_dev->data->dev_private;
1208 
1209 	/* Disable vector mode RX for Stingray2 for now */
1210 	if (BNXT_CHIP_SR2(bp)) {
1211 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1212 		return bnxt_recv_pkts;
1213 	}
1214 
1215 #if (defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)) && \
1216 	!defined(RTE_LIBRTE_IEEE1588)
1217 
1218 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1219 	if (eth_dev->data->scattered_rx)
1220 		goto use_scalar_rx;
1221 
1222 	/*
1223 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1224 	 * asynchronous completions and receive completions can be placed in
1225 	 * the same completion ring.
1226 	 */
1227 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1228 		goto use_scalar_rx;
1229 
1230 	/*
1231 	 * Vector mode receive cannot be enabled if any receive offloads outside
1232 	 * a limited subset have been enabled.
1233 	 */
1234 	if (eth_dev->data->dev_conf.rxmode.offloads &
1235 		~(DEV_RX_OFFLOAD_VLAN_STRIP |
1236 		  DEV_RX_OFFLOAD_KEEP_CRC |
1237 		  DEV_RX_OFFLOAD_JUMBO_FRAME |
1238 		  DEV_RX_OFFLOAD_IPV4_CKSUM |
1239 		  DEV_RX_OFFLOAD_UDP_CKSUM |
1240 		  DEV_RX_OFFLOAD_TCP_CKSUM |
1241 		  DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1242 		  DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1243 		  DEV_RX_OFFLOAD_RSS_HASH |
1244 		  DEV_RX_OFFLOAD_VLAN_FILTER))
1245 		goto use_scalar_rx;
1246 
1247 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1248 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1249 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1250 		PMD_DRV_LOG(INFO,
1251 			    "Using AVX2 vector mode receive for port %d\n",
1252 			    eth_dev->data->port_id);
1253 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1254 		return bnxt_recv_pkts_vec_avx2;
1255 	}
1256  #endif
1257 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1258 		PMD_DRV_LOG(INFO,
1259 			    "Using SSE vector mode receive for port %d\n",
1260 			    eth_dev->data->port_id);
1261 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1262 		return bnxt_recv_pkts_vec;
1263 	}
1264 
1265 use_scalar_rx:
1266 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1267 		    eth_dev->data->port_id);
1268 	PMD_DRV_LOG(INFO,
1269 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1270 		    eth_dev->data->port_id,
1271 		    eth_dev->data->scattered_rx,
1272 		    eth_dev->data->dev_conf.rxmode.offloads);
1273 #endif
1274 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1275 	return bnxt_recv_pkts;
1276 }
1277 
1278 static eth_tx_burst_t
1279 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1280 {
1281 	struct bnxt *bp = eth_dev->data->dev_private;
1282 
1283 	/* Disable vector mode TX for Stingray2 for now */
1284 	if (BNXT_CHIP_SR2(bp))
1285 		return bnxt_xmit_pkts;
1286 
1287 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64) && \
1288 	!defined(RTE_LIBRTE_IEEE1588)
1289 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1290 
1291 	/*
1292 	 * Vector mode transmit can be enabled only if not using scatter rx
1293 	 * or tx offloads.
1294 	 */
1295 	if (eth_dev->data->scattered_rx ||
1296 	    (offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) ||
1297 	    BNXT_TRUFLOW_EN(bp))
1298 		goto use_scalar_tx;
1299 
1300 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1301 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1302 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1303 		PMD_DRV_LOG(INFO,
1304 			    "Using AVX2 vector mode transmit for port %d\n",
1305 			    eth_dev->data->port_id);
1306 		return bnxt_xmit_pkts_vec_avx2;
1307 	}
1308 #endif
1309 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1310 		PMD_DRV_LOG(INFO,
1311 			    "Using SSE vector mode transmit for port %d\n",
1312 			    eth_dev->data->port_id);
1313 		return bnxt_xmit_pkts_vec;
1314 	}
1315 
1316 use_scalar_tx:
1317 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1318 		    eth_dev->data->port_id);
1319 	PMD_DRV_LOG(INFO,
1320 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1321 		    eth_dev->data->port_id,
1322 		    eth_dev->data->scattered_rx,
1323 		    offloads);
1324 #endif
1325 	return bnxt_xmit_pkts;
1326 }
1327 
1328 static int bnxt_handle_if_change_status(struct bnxt *bp)
1329 {
1330 	int rc;
1331 
1332 	/* Since fw has undergone a reset and lost all contexts,
1333 	 * set fatal flag to not issue hwrm during cleanup
1334 	 */
1335 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1336 	bnxt_uninit_resources(bp, true);
1337 
1338 	/* clear fatal flag so that re-init happens */
1339 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1340 	rc = bnxt_init_resources(bp, true);
1341 
1342 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1343 
1344 	return rc;
1345 }
1346 
1347 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1348 {
1349 	struct bnxt *bp = eth_dev->data->dev_private;
1350 	int rc = 0;
1351 
1352 	if (!BNXT_SINGLE_PF(bp))
1353 		return -ENOTSUP;
1354 
1355 	if (!bp->link_info->link_up)
1356 		rc = bnxt_set_hwrm_link_config(bp, true);
1357 	if (!rc)
1358 		eth_dev->data->dev_link.link_status = 1;
1359 
1360 	bnxt_print_link_info(eth_dev);
1361 	return rc;
1362 }
1363 
1364 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1365 {
1366 	struct bnxt *bp = eth_dev->data->dev_private;
1367 
1368 	if (!BNXT_SINGLE_PF(bp))
1369 		return -ENOTSUP;
1370 
1371 	eth_dev->data->dev_link.link_status = 0;
1372 	bnxt_set_hwrm_link_config(bp, false);
1373 	bp->link_info->link_up = 0;
1374 
1375 	return 0;
1376 }
1377 
1378 static void bnxt_free_switch_domain(struct bnxt *bp)
1379 {
1380 	int rc = 0;
1381 
1382 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1383 		return;
1384 
1385 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1386 	if (rc)
1387 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1388 			    bp->switch_domain_id, rc);
1389 }
1390 
1391 static void bnxt_ptp_get_current_time(void *arg)
1392 {
1393 	struct bnxt *bp = arg;
1394 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1395 	int rc;
1396 
1397 	rc = is_bnxt_in_error(bp);
1398 	if (rc)
1399 		return;
1400 
1401 	if (!ptp)
1402 		return;
1403 
1404 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1405 				&ptp->current_time);
1406 
1407 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1408 	if (rc != 0) {
1409 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1410 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1411 	}
1412 }
1413 
1414 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1415 {
1416 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1417 	int rc;
1418 
1419 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1420 		return 0;
1421 
1422 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1423 				&ptp->current_time);
1424 
1425 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1426 	return rc;
1427 }
1428 
1429 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1430 {
1431 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1432 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1433 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1434 	}
1435 }
1436 
1437 static void bnxt_ptp_stop(struct bnxt *bp)
1438 {
1439 	bnxt_cancel_ptp_alarm(bp);
1440 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1441 }
1442 
1443 static int bnxt_ptp_start(struct bnxt *bp)
1444 {
1445 	int rc;
1446 
1447 	rc = bnxt_schedule_ptp_alarm(bp);
1448 	if (rc != 0) {
1449 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1450 	} else {
1451 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1452 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1453 	}
1454 
1455 	return rc;
1456 }
1457 
1458 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1459 {
1460 	struct bnxt *bp = eth_dev->data->dev_private;
1461 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1462 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1463 	struct rte_eth_link link;
1464 	int ret;
1465 
1466 	eth_dev->data->dev_started = 0;
1467 	eth_dev->data->scattered_rx = 0;
1468 
1469 	/* Prevent crashes when queues are still in use */
1470 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1471 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1472 
1473 	bnxt_disable_int(bp);
1474 
1475 	/* disable uio/vfio intr/eventfd mapping */
1476 	rte_intr_disable(intr_handle);
1477 
1478 	/* Stop the child representors for this device */
1479 	ret = bnxt_rep_stop_all(bp);
1480 	if (ret != 0)
1481 		return ret;
1482 
1483 	/* delete the bnxt ULP port details */
1484 	bnxt_ulp_port_deinit(bp);
1485 
1486 	bnxt_cancel_fw_health_check(bp);
1487 
1488 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1489 		bnxt_cancel_ptp_alarm(bp);
1490 
1491 	/* Do not bring link down during reset recovery */
1492 	if (!is_bnxt_in_error(bp)) {
1493 		bnxt_dev_set_link_down_op(eth_dev);
1494 		/* Wait for link to be reset */
1495 		if (BNXT_SINGLE_PF(bp))
1496 			rte_delay_ms(500);
1497 		/* clear the recorded link status */
1498 		memset(&link, 0, sizeof(link));
1499 		rte_eth_linkstatus_set(eth_dev, &link);
1500 	}
1501 
1502 	/* Clean queue intr-vector mapping */
1503 	rte_intr_efd_disable(intr_handle);
1504 	if (intr_handle->intr_vec != NULL) {
1505 		rte_free(intr_handle->intr_vec);
1506 		intr_handle->intr_vec = NULL;
1507 	}
1508 
1509 	bnxt_hwrm_port_clr_stats(bp);
1510 	bnxt_free_tx_mbufs(bp);
1511 	bnxt_free_rx_mbufs(bp);
1512 	/* Process any remaining notifications in default completion queue */
1513 	bnxt_int_handler(eth_dev);
1514 	bnxt_shutdown_nic(bp);
1515 	bnxt_hwrm_if_change(bp, false);
1516 
1517 	bnxt_free_prev_ring_stats(bp);
1518 	rte_free(bp->mark_table);
1519 	bp->mark_table = NULL;
1520 
1521 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1522 	bp->rx_cosq_cnt = 0;
1523 	/* All filters are deleted on a port stop. */
1524 	if (BNXT_FLOW_XSTATS_EN(bp))
1525 		bp->flow_stat->flow_count = 0;
1526 
1527 	return 0;
1528 }
1529 
1530 /* Unload the driver, release resources */
1531 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1532 {
1533 	struct bnxt *bp = eth_dev->data->dev_private;
1534 
1535 	pthread_mutex_lock(&bp->err_recovery_lock);
1536 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1537 		PMD_DRV_LOG(ERR,
1538 			    "Adapter recovering from error..Please retry\n");
1539 		pthread_mutex_unlock(&bp->err_recovery_lock);
1540 		return -EAGAIN;
1541 	}
1542 	pthread_mutex_unlock(&bp->err_recovery_lock);
1543 
1544 	return bnxt_dev_stop(eth_dev);
1545 }
1546 
1547 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1548 {
1549 	struct bnxt *bp = eth_dev->data->dev_private;
1550 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1551 	int vlan_mask = 0;
1552 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1553 
1554 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1555 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1556 		return -EINVAL;
1557 	}
1558 
1559 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1560 		PMD_DRV_LOG(ERR,
1561 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1562 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1563 
1564 	do {
1565 		rc = bnxt_hwrm_if_change(bp, true);
1566 		if (rc == 0 || rc != -EAGAIN)
1567 			break;
1568 
1569 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1570 	} while (retry_cnt--);
1571 
1572 	if (rc)
1573 		return rc;
1574 
1575 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1576 		rc = bnxt_handle_if_change_status(bp);
1577 		if (rc)
1578 			return rc;
1579 	}
1580 
1581 	bnxt_enable_int(bp);
1582 
1583 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1584 
1585 	rc = bnxt_start_nic(bp);
1586 	if (rc)
1587 		goto error;
1588 
1589 	rc = bnxt_alloc_prev_ring_stats(bp);
1590 	if (rc)
1591 		goto error;
1592 
1593 	eth_dev->data->dev_started = 1;
1594 
1595 	bnxt_link_update_op(eth_dev, 1);
1596 
1597 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1598 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1599 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1600 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1601 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1602 	if (rc)
1603 		goto error;
1604 
1605 	/* Initialize bnxt ULP port details */
1606 	rc = bnxt_ulp_port_init(bp);
1607 	if (rc)
1608 		goto error;
1609 
1610 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1611 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1612 
1613 	bnxt_schedule_fw_health_check(bp);
1614 
1615 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1616 		bnxt_schedule_ptp_alarm(bp);
1617 
1618 	return 0;
1619 
1620 error:
1621 	bnxt_dev_stop(eth_dev);
1622 	return rc;
1623 }
1624 
1625 static void
1626 bnxt_uninit_locks(struct bnxt *bp)
1627 {
1628 	pthread_mutex_destroy(&bp->flow_lock);
1629 	pthread_mutex_destroy(&bp->def_cp_lock);
1630 	pthread_mutex_destroy(&bp->health_check_lock);
1631 	pthread_mutex_destroy(&bp->err_recovery_lock);
1632 	if (bp->rep_info) {
1633 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1634 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1635 	}
1636 }
1637 
1638 static void bnxt_drv_uninit(struct bnxt *bp)
1639 {
1640 	bnxt_free_leds_info(bp);
1641 	bnxt_free_cos_queues(bp);
1642 	bnxt_free_link_info(bp);
1643 	bnxt_free_parent_info(bp);
1644 	bnxt_uninit_locks(bp);
1645 
1646 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1647 	bp->tx_mem_zone = NULL;
1648 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1649 	bp->rx_mem_zone = NULL;
1650 
1651 	bnxt_free_vf_info(bp);
1652 	bnxt_free_pf_info(bp);
1653 
1654 	rte_free(bp->grp_info);
1655 	bp->grp_info = NULL;
1656 }
1657 
1658 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1659 {
1660 	struct bnxt *bp = eth_dev->data->dev_private;
1661 	int ret = 0;
1662 
1663 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1664 		return 0;
1665 
1666 	pthread_mutex_lock(&bp->err_recovery_lock);
1667 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1668 		PMD_DRV_LOG(ERR,
1669 			    "Adapter recovering from error...Please retry\n");
1670 		pthread_mutex_unlock(&bp->err_recovery_lock);
1671 		return -EAGAIN;
1672 	}
1673 	pthread_mutex_unlock(&bp->err_recovery_lock);
1674 
1675 	/* cancel the recovery handler before remove dev */
1676 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1677 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1678 	bnxt_cancel_fc_thread(bp);
1679 
1680 	if (eth_dev->data->dev_started)
1681 		ret = bnxt_dev_stop(eth_dev);
1682 
1683 	bnxt_uninit_resources(bp, false);
1684 
1685 	bnxt_drv_uninit(bp);
1686 
1687 	return ret;
1688 }
1689 
1690 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1691 				    uint32_t index)
1692 {
1693 	struct bnxt *bp = eth_dev->data->dev_private;
1694 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1695 	struct bnxt_vnic_info *vnic;
1696 	struct bnxt_filter_info *filter, *temp_filter;
1697 	uint32_t i;
1698 
1699 	if (is_bnxt_in_error(bp))
1700 		return;
1701 
1702 	/*
1703 	 * Loop through all VNICs from the specified filter flow pools to
1704 	 * remove the corresponding MAC addr filter
1705 	 */
1706 	for (i = 0; i < bp->nr_vnics; i++) {
1707 		if (!(pool_mask & (1ULL << i)))
1708 			continue;
1709 
1710 		vnic = &bp->vnic_info[i];
1711 		filter = STAILQ_FIRST(&vnic->filter);
1712 		while (filter) {
1713 			temp_filter = STAILQ_NEXT(filter, next);
1714 			if (filter->mac_index == index) {
1715 				STAILQ_REMOVE(&vnic->filter, filter,
1716 						bnxt_filter_info, next);
1717 				bnxt_hwrm_clear_l2_filter(bp, filter);
1718 				bnxt_free_filter(bp, filter);
1719 			}
1720 			filter = temp_filter;
1721 		}
1722 	}
1723 }
1724 
1725 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1726 			       struct rte_ether_addr *mac_addr, uint32_t index,
1727 			       uint32_t pool)
1728 {
1729 	struct bnxt_filter_info *filter;
1730 	int rc = 0;
1731 
1732 	/* Attach requested MAC address to the new l2_filter */
1733 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1734 		if (filter->mac_index == index) {
1735 			PMD_DRV_LOG(DEBUG,
1736 				    "MAC addr already existed for pool %d\n",
1737 				    pool);
1738 			return 0;
1739 		}
1740 	}
1741 
1742 	filter = bnxt_alloc_filter(bp);
1743 	if (!filter) {
1744 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1745 		return -ENODEV;
1746 	}
1747 
1748 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1749 	 * if the MAC that's been programmed now is a different one, then,
1750 	 * copy that addr to filter->l2_addr
1751 	 */
1752 	if (mac_addr)
1753 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1754 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1755 
1756 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1757 	if (!rc) {
1758 		filter->mac_index = index;
1759 		if (filter->mac_index == 0)
1760 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1761 		else
1762 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1763 	} else {
1764 		bnxt_free_filter(bp, filter);
1765 	}
1766 
1767 	return rc;
1768 }
1769 
1770 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1771 				struct rte_ether_addr *mac_addr,
1772 				uint32_t index, uint32_t pool)
1773 {
1774 	struct bnxt *bp = eth_dev->data->dev_private;
1775 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1776 	int rc = 0;
1777 
1778 	rc = is_bnxt_in_error(bp);
1779 	if (rc)
1780 		return rc;
1781 
1782 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1783 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1784 		return -ENOTSUP;
1785 	}
1786 
1787 	if (!vnic) {
1788 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1789 		return -EINVAL;
1790 	}
1791 
1792 	/* Filter settings will get applied when port is started */
1793 	if (!eth_dev->data->dev_started)
1794 		return 0;
1795 
1796 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1797 
1798 	return rc;
1799 }
1800 
1801 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1802 {
1803 	int rc = 0;
1804 	struct bnxt *bp = eth_dev->data->dev_private;
1805 	struct rte_eth_link new;
1806 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1807 			BNXT_MIN_LINK_WAIT_CNT;
1808 
1809 	rc = is_bnxt_in_error(bp);
1810 	if (rc)
1811 		return rc;
1812 
1813 	memset(&new, 0, sizeof(new));
1814 
1815 	if (bp->link_info == NULL)
1816 		goto out;
1817 
1818 	do {
1819 		/* Retrieve link info from hardware */
1820 		rc = bnxt_get_hwrm_link_config(bp, &new);
1821 		if (rc) {
1822 			new.link_speed = ETH_LINK_SPEED_100M;
1823 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1824 			PMD_DRV_LOG(ERR,
1825 				"Failed to retrieve link rc = 0x%x!\n", rc);
1826 			goto out;
1827 		}
1828 
1829 		if (!wait_to_complete || new.link_status)
1830 			break;
1831 
1832 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1833 	} while (cnt--);
1834 
1835 	/* Only single function PF can bring phy down.
1836 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1837 	 */
1838 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1839 		memset(&new, 0, sizeof(new));
1840 
1841 out:
1842 	/* Timed out or success */
1843 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1844 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1845 		rte_eth_linkstatus_set(eth_dev, &new);
1846 		bnxt_print_link_info(eth_dev);
1847 	}
1848 
1849 	return rc;
1850 }
1851 
1852 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1853 {
1854 	struct bnxt *bp = eth_dev->data->dev_private;
1855 	struct bnxt_vnic_info *vnic;
1856 	uint32_t old_flags;
1857 	int rc;
1858 
1859 	rc = is_bnxt_in_error(bp);
1860 	if (rc)
1861 		return rc;
1862 
1863 	/* Filter settings will get applied when port is started */
1864 	if (!eth_dev->data->dev_started)
1865 		return 0;
1866 
1867 	if (bp->vnic_info == NULL)
1868 		return 0;
1869 
1870 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1871 
1872 	old_flags = vnic->flags;
1873 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1874 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1875 	if (rc != 0)
1876 		vnic->flags = old_flags;
1877 
1878 	return rc;
1879 }
1880 
1881 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1882 {
1883 	struct bnxt *bp = eth_dev->data->dev_private;
1884 	struct bnxt_vnic_info *vnic;
1885 	uint32_t old_flags;
1886 	int rc;
1887 
1888 	rc = is_bnxt_in_error(bp);
1889 	if (rc)
1890 		return rc;
1891 
1892 	/* Filter settings will get applied when port is started */
1893 	if (!eth_dev->data->dev_started)
1894 		return 0;
1895 
1896 	if (bp->vnic_info == NULL)
1897 		return 0;
1898 
1899 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1900 
1901 	old_flags = vnic->flags;
1902 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1903 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1904 	if (rc != 0)
1905 		vnic->flags = old_flags;
1906 
1907 	return rc;
1908 }
1909 
1910 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1911 {
1912 	struct bnxt *bp = eth_dev->data->dev_private;
1913 	struct bnxt_vnic_info *vnic;
1914 	uint32_t old_flags;
1915 	int rc;
1916 
1917 	rc = is_bnxt_in_error(bp);
1918 	if (rc)
1919 		return rc;
1920 
1921 	/* Filter settings will get applied when port is started */
1922 	if (!eth_dev->data->dev_started)
1923 		return 0;
1924 
1925 	if (bp->vnic_info == NULL)
1926 		return 0;
1927 
1928 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1929 
1930 	old_flags = vnic->flags;
1931 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1932 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1933 	if (rc != 0)
1934 		vnic->flags = old_flags;
1935 
1936 	return rc;
1937 }
1938 
1939 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1940 {
1941 	struct bnxt *bp = eth_dev->data->dev_private;
1942 	struct bnxt_vnic_info *vnic;
1943 	uint32_t old_flags;
1944 	int rc;
1945 
1946 	rc = is_bnxt_in_error(bp);
1947 	if (rc)
1948 		return rc;
1949 
1950 	/* Filter settings will get applied when port is started */
1951 	if (!eth_dev->data->dev_started)
1952 		return 0;
1953 
1954 	if (bp->vnic_info == NULL)
1955 		return 0;
1956 
1957 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1958 
1959 	old_flags = vnic->flags;
1960 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1961 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1962 	if (rc != 0)
1963 		vnic->flags = old_flags;
1964 
1965 	return rc;
1966 }
1967 
1968 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1969 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1970 {
1971 	if (qid >= bp->rx_nr_rings)
1972 		return NULL;
1973 
1974 	return bp->eth_dev->data->rx_queues[qid];
1975 }
1976 
1977 /* Return rxq corresponding to a given rss table ring/group ID. */
1978 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1979 {
1980 	struct bnxt_rx_queue *rxq;
1981 	unsigned int i;
1982 
1983 	if (!BNXT_HAS_RING_GRPS(bp)) {
1984 		for (i = 0; i < bp->rx_nr_rings; i++) {
1985 			rxq = bp->eth_dev->data->rx_queues[i];
1986 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1987 				return rxq->index;
1988 		}
1989 	} else {
1990 		for (i = 0; i < bp->rx_nr_rings; i++) {
1991 			if (bp->grp_info[i].fw_grp_id == fwr)
1992 				return i;
1993 		}
1994 	}
1995 
1996 	return INVALID_HW_RING_ID;
1997 }
1998 
1999 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
2000 			    struct rte_eth_rss_reta_entry64 *reta_conf,
2001 			    uint16_t reta_size)
2002 {
2003 	struct bnxt *bp = eth_dev->data->dev_private;
2004 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2005 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2006 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2007 	uint16_t idx, sft;
2008 	int i, rc;
2009 
2010 	rc = is_bnxt_in_error(bp);
2011 	if (rc)
2012 		return rc;
2013 
2014 	if (!vnic->rss_table)
2015 		return -EINVAL;
2016 
2017 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
2018 		return -EINVAL;
2019 
2020 	if (reta_size != tbl_size) {
2021 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2022 			"(%d) must equal the size supported by the hardware "
2023 			"(%d)\n", reta_size, tbl_size);
2024 		return -EINVAL;
2025 	}
2026 
2027 	for (i = 0; i < reta_size; i++) {
2028 		struct bnxt_rx_queue *rxq;
2029 
2030 		idx = i / RTE_RETA_GROUP_SIZE;
2031 		sft = i % RTE_RETA_GROUP_SIZE;
2032 
2033 		if (!(reta_conf[idx].mask & (1ULL << sft)))
2034 			continue;
2035 
2036 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2037 		if (!rxq) {
2038 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
2039 			return -EINVAL;
2040 		}
2041 
2042 		if (BNXT_CHIP_P5(bp)) {
2043 			vnic->rss_table[i * 2] =
2044 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
2045 			vnic->rss_table[i * 2 + 1] =
2046 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
2047 		} else {
2048 			vnic->rss_table[i] =
2049 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2050 		}
2051 	}
2052 
2053 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2054 	return rc;
2055 }
2056 
2057 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2058 			      struct rte_eth_rss_reta_entry64 *reta_conf,
2059 			      uint16_t reta_size)
2060 {
2061 	struct bnxt *bp = eth_dev->data->dev_private;
2062 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2063 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2064 	uint16_t idx, sft, i;
2065 	int rc;
2066 
2067 	rc = is_bnxt_in_error(bp);
2068 	if (rc)
2069 		return rc;
2070 
2071 	if (!vnic)
2072 		return -EINVAL;
2073 	if (!vnic->rss_table)
2074 		return -EINVAL;
2075 
2076 	if (reta_size != tbl_size) {
2077 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2078 			"(%d) must equal the size supported by the hardware "
2079 			"(%d)\n", reta_size, tbl_size);
2080 		return -EINVAL;
2081 	}
2082 
2083 	for (idx = 0, i = 0; i < reta_size; i++) {
2084 		idx = i / RTE_RETA_GROUP_SIZE;
2085 		sft = i % RTE_RETA_GROUP_SIZE;
2086 
2087 		if (reta_conf[idx].mask & (1ULL << sft)) {
2088 			uint16_t qid;
2089 
2090 			if (BNXT_CHIP_P5(bp))
2091 				qid = bnxt_rss_to_qid(bp,
2092 						      vnic->rss_table[i * 2]);
2093 			else
2094 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2095 
2096 			if (qid == INVALID_HW_RING_ID) {
2097 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2098 				return -EINVAL;
2099 			}
2100 			reta_conf[idx].reta[sft] = qid;
2101 		}
2102 	}
2103 
2104 	return 0;
2105 }
2106 
2107 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2108 				   struct rte_eth_rss_conf *rss_conf)
2109 {
2110 	struct bnxt *bp = eth_dev->data->dev_private;
2111 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2112 	struct bnxt_vnic_info *vnic;
2113 	int rc;
2114 
2115 	rc = is_bnxt_in_error(bp);
2116 	if (rc)
2117 		return rc;
2118 
2119 	/*
2120 	 * If RSS enablement were different than dev_configure,
2121 	 * then return -EINVAL
2122 	 */
2123 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
2124 		if (!rss_conf->rss_hf)
2125 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2126 	} else {
2127 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2128 			return -EINVAL;
2129 	}
2130 
2131 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
2132 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2133 	       rss_conf,
2134 	       sizeof(*rss_conf));
2135 
2136 	/* Update the default RSS VNIC(s) */
2137 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2138 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2139 	vnic->hash_mode =
2140 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2141 					    ETH_RSS_LEVEL(rss_conf->rss_hf));
2142 
2143 	/*
2144 	 * If hashkey is not specified, use the previously configured
2145 	 * hashkey
2146 	 */
2147 	if (!rss_conf->rss_key)
2148 		goto rss_config;
2149 
2150 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2151 		PMD_DRV_LOG(ERR,
2152 			    "Invalid hashkey length, should be %d bytes\n",
2153 			    HW_HASH_KEY_SIZE);
2154 		return -EINVAL;
2155 	}
2156 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2157 
2158 rss_config:
2159 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2160 	return rc;
2161 }
2162 
2163 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2164 				     struct rte_eth_rss_conf *rss_conf)
2165 {
2166 	struct bnxt *bp = eth_dev->data->dev_private;
2167 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2168 	int len, rc;
2169 	uint32_t hash_types;
2170 
2171 	rc = is_bnxt_in_error(bp);
2172 	if (rc)
2173 		return rc;
2174 
2175 	/* RSS configuration is the same for all VNICs */
2176 	if (vnic && vnic->rss_hash_key) {
2177 		if (rss_conf->rss_key) {
2178 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2179 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2180 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2181 		}
2182 
2183 		hash_types = vnic->hash_type;
2184 		rss_conf->rss_hf = 0;
2185 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2186 			rss_conf->rss_hf |= ETH_RSS_IPV4;
2187 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2188 		}
2189 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2190 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2191 			hash_types &=
2192 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2193 		}
2194 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2195 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2196 			hash_types &=
2197 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2198 		}
2199 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2200 			rss_conf->rss_hf |= ETH_RSS_IPV6;
2201 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2202 		}
2203 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2204 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2205 			hash_types &=
2206 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2207 		}
2208 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2209 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2210 			hash_types &=
2211 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2212 		}
2213 
2214 		rss_conf->rss_hf |=
2215 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2216 
2217 		if (hash_types) {
2218 			PMD_DRV_LOG(ERR,
2219 				"Unknown RSS config from firmware (%08x), RSS disabled",
2220 				vnic->hash_type);
2221 			return -ENOTSUP;
2222 		}
2223 	} else {
2224 		rss_conf->rss_hf = 0;
2225 	}
2226 	return 0;
2227 }
2228 
2229 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2230 			       struct rte_eth_fc_conf *fc_conf)
2231 {
2232 	struct bnxt *bp = dev->data->dev_private;
2233 	struct rte_eth_link link_info;
2234 	int rc;
2235 
2236 	rc = is_bnxt_in_error(bp);
2237 	if (rc)
2238 		return rc;
2239 
2240 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2241 	if (rc)
2242 		return rc;
2243 
2244 	memset(fc_conf, 0, sizeof(*fc_conf));
2245 	if (bp->link_info->auto_pause)
2246 		fc_conf->autoneg = 1;
2247 	switch (bp->link_info->pause) {
2248 	case 0:
2249 		fc_conf->mode = RTE_FC_NONE;
2250 		break;
2251 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2252 		fc_conf->mode = RTE_FC_TX_PAUSE;
2253 		break;
2254 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2255 		fc_conf->mode = RTE_FC_RX_PAUSE;
2256 		break;
2257 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2258 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2259 		fc_conf->mode = RTE_FC_FULL;
2260 		break;
2261 	}
2262 	return 0;
2263 }
2264 
2265 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2266 			       struct rte_eth_fc_conf *fc_conf)
2267 {
2268 	struct bnxt *bp = dev->data->dev_private;
2269 	int rc;
2270 
2271 	rc = is_bnxt_in_error(bp);
2272 	if (rc)
2273 		return rc;
2274 
2275 	if (!BNXT_SINGLE_PF(bp)) {
2276 		PMD_DRV_LOG(ERR,
2277 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2278 		return -ENOTSUP;
2279 	}
2280 
2281 	switch (fc_conf->mode) {
2282 	case RTE_FC_NONE:
2283 		bp->link_info->auto_pause = 0;
2284 		bp->link_info->force_pause = 0;
2285 		break;
2286 	case RTE_FC_RX_PAUSE:
2287 		if (fc_conf->autoneg) {
2288 			bp->link_info->auto_pause =
2289 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2290 			bp->link_info->force_pause = 0;
2291 		} else {
2292 			bp->link_info->auto_pause = 0;
2293 			bp->link_info->force_pause =
2294 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2295 		}
2296 		break;
2297 	case RTE_FC_TX_PAUSE:
2298 		if (fc_conf->autoneg) {
2299 			bp->link_info->auto_pause =
2300 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2301 			bp->link_info->force_pause = 0;
2302 		} else {
2303 			bp->link_info->auto_pause = 0;
2304 			bp->link_info->force_pause =
2305 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2306 		}
2307 		break;
2308 	case RTE_FC_FULL:
2309 		if (fc_conf->autoneg) {
2310 			bp->link_info->auto_pause =
2311 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2312 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2313 			bp->link_info->force_pause = 0;
2314 		} else {
2315 			bp->link_info->auto_pause = 0;
2316 			bp->link_info->force_pause =
2317 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2318 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2319 		}
2320 		break;
2321 	}
2322 	return bnxt_set_hwrm_link_config(bp, true);
2323 }
2324 
2325 /* Add UDP tunneling port */
2326 static int
2327 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2328 			 struct rte_eth_udp_tunnel *udp_tunnel)
2329 {
2330 	struct bnxt *bp = eth_dev->data->dev_private;
2331 	uint16_t tunnel_type = 0;
2332 	int rc = 0;
2333 
2334 	rc = is_bnxt_in_error(bp);
2335 	if (rc)
2336 		return rc;
2337 
2338 	switch (udp_tunnel->prot_type) {
2339 	case RTE_TUNNEL_TYPE_VXLAN:
2340 		if (bp->vxlan_port_cnt) {
2341 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2342 				udp_tunnel->udp_port);
2343 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2344 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2345 				return -ENOSPC;
2346 			}
2347 			bp->vxlan_port_cnt++;
2348 			return 0;
2349 		}
2350 		tunnel_type =
2351 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2352 		bp->vxlan_port_cnt++;
2353 		break;
2354 	case RTE_TUNNEL_TYPE_GENEVE:
2355 		if (bp->geneve_port_cnt) {
2356 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2357 				udp_tunnel->udp_port);
2358 			if (bp->geneve_port != udp_tunnel->udp_port) {
2359 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2360 				return -ENOSPC;
2361 			}
2362 			bp->geneve_port_cnt++;
2363 			return 0;
2364 		}
2365 		tunnel_type =
2366 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2367 		bp->geneve_port_cnt++;
2368 		break;
2369 	default:
2370 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2371 		return -ENOTSUP;
2372 	}
2373 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2374 					     tunnel_type);
2375 	return rc;
2376 }
2377 
2378 static int
2379 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2380 			 struct rte_eth_udp_tunnel *udp_tunnel)
2381 {
2382 	struct bnxt *bp = eth_dev->data->dev_private;
2383 	uint16_t tunnel_type = 0;
2384 	uint16_t port = 0;
2385 	int rc = 0;
2386 
2387 	rc = is_bnxt_in_error(bp);
2388 	if (rc)
2389 		return rc;
2390 
2391 	switch (udp_tunnel->prot_type) {
2392 	case RTE_TUNNEL_TYPE_VXLAN:
2393 		if (!bp->vxlan_port_cnt) {
2394 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2395 			return -EINVAL;
2396 		}
2397 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2398 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2399 				udp_tunnel->udp_port, bp->vxlan_port);
2400 			return -EINVAL;
2401 		}
2402 		if (--bp->vxlan_port_cnt)
2403 			return 0;
2404 
2405 		tunnel_type =
2406 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2407 		port = bp->vxlan_fw_dst_port_id;
2408 		break;
2409 	case RTE_TUNNEL_TYPE_GENEVE:
2410 		if (!bp->geneve_port_cnt) {
2411 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2412 			return -EINVAL;
2413 		}
2414 		if (bp->geneve_port != udp_tunnel->udp_port) {
2415 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2416 				udp_tunnel->udp_port, bp->geneve_port);
2417 			return -EINVAL;
2418 		}
2419 		if (--bp->geneve_port_cnt)
2420 			return 0;
2421 
2422 		tunnel_type =
2423 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2424 		port = bp->geneve_fw_dst_port_id;
2425 		break;
2426 	default:
2427 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2428 		return -ENOTSUP;
2429 	}
2430 
2431 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2432 	return rc;
2433 }
2434 
2435 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2436 {
2437 	struct bnxt_filter_info *filter;
2438 	struct bnxt_vnic_info *vnic;
2439 	int rc = 0;
2440 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2441 
2442 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2443 	filter = STAILQ_FIRST(&vnic->filter);
2444 	while (filter) {
2445 		/* Search for this matching MAC+VLAN filter */
2446 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2447 			/* Delete the filter */
2448 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2449 			if (rc)
2450 				return rc;
2451 			STAILQ_REMOVE(&vnic->filter, filter,
2452 				      bnxt_filter_info, next);
2453 			bnxt_free_filter(bp, filter);
2454 			PMD_DRV_LOG(INFO,
2455 				    "Deleted vlan filter for %d\n",
2456 				    vlan_id);
2457 			return 0;
2458 		}
2459 		filter = STAILQ_NEXT(filter, next);
2460 	}
2461 	return -ENOENT;
2462 }
2463 
2464 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2465 {
2466 	struct bnxt_filter_info *filter;
2467 	struct bnxt_vnic_info *vnic;
2468 	int rc = 0;
2469 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2470 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2471 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2472 
2473 	/* Implementation notes on the use of VNIC in this command:
2474 	 *
2475 	 * By default, these filters belong to default vnic for the function.
2476 	 * Once these filters are set up, only destination VNIC can be modified.
2477 	 * If the destination VNIC is not specified in this command,
2478 	 * then the HWRM shall only create an l2 context id.
2479 	 */
2480 
2481 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2482 	filter = STAILQ_FIRST(&vnic->filter);
2483 	/* Check if the VLAN has already been added */
2484 	while (filter) {
2485 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2486 			return -EEXIST;
2487 
2488 		filter = STAILQ_NEXT(filter, next);
2489 	}
2490 
2491 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2492 	 * command to create MAC+VLAN filter with the right flags, enables set.
2493 	 */
2494 	filter = bnxt_alloc_filter(bp);
2495 	if (!filter) {
2496 		PMD_DRV_LOG(ERR,
2497 			    "MAC/VLAN filter alloc failed\n");
2498 		return -ENOMEM;
2499 	}
2500 	/* MAC + VLAN ID filter */
2501 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2502 	 * untagged packets are received
2503 	 *
2504 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2505 	 * packets and only the programmed vlan's packets are received
2506 	 */
2507 	filter->l2_ivlan = vlan_id;
2508 	filter->l2_ivlan_mask = 0x0FFF;
2509 	filter->enables |= en;
2510 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2511 
2512 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2513 	if (rc) {
2514 		/* Free the newly allocated filter as we were
2515 		 * not able to create the filter in hardware.
2516 		 */
2517 		bnxt_free_filter(bp, filter);
2518 		return rc;
2519 	}
2520 
2521 	filter->mac_index = 0;
2522 	/* Add this new filter to the list */
2523 	if (vlan_id == 0)
2524 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2525 	else
2526 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2527 
2528 	PMD_DRV_LOG(INFO,
2529 		    "Added Vlan filter for %d\n", vlan_id);
2530 	return rc;
2531 }
2532 
2533 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2534 		uint16_t vlan_id, int on)
2535 {
2536 	struct bnxt *bp = eth_dev->data->dev_private;
2537 	int rc;
2538 
2539 	rc = is_bnxt_in_error(bp);
2540 	if (rc)
2541 		return rc;
2542 
2543 	if (!eth_dev->data->dev_started) {
2544 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2545 		return -EINVAL;
2546 	}
2547 
2548 	/* These operations apply to ALL existing MAC/VLAN filters */
2549 	if (on)
2550 		return bnxt_add_vlan_filter(bp, vlan_id);
2551 	else
2552 		return bnxt_del_vlan_filter(bp, vlan_id);
2553 }
2554 
2555 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2556 				    struct bnxt_vnic_info *vnic)
2557 {
2558 	struct bnxt_filter_info *filter;
2559 	int rc;
2560 
2561 	filter = STAILQ_FIRST(&vnic->filter);
2562 	while (filter) {
2563 		if (filter->mac_index == 0 &&
2564 		    !memcmp(filter->l2_addr, bp->mac_addr,
2565 			    RTE_ETHER_ADDR_LEN)) {
2566 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2567 			if (!rc) {
2568 				STAILQ_REMOVE(&vnic->filter, filter,
2569 					      bnxt_filter_info, next);
2570 				bnxt_free_filter(bp, filter);
2571 			}
2572 			return rc;
2573 		}
2574 		filter = STAILQ_NEXT(filter, next);
2575 	}
2576 	return 0;
2577 }
2578 
2579 static int
2580 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2581 {
2582 	struct bnxt_vnic_info *vnic;
2583 	unsigned int i;
2584 	int rc;
2585 
2586 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2587 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2588 		/* Remove any VLAN filters programmed */
2589 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2590 			bnxt_del_vlan_filter(bp, i);
2591 
2592 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2593 		if (rc)
2594 			return rc;
2595 	} else {
2596 		/* Default filter will allow packets that match the
2597 		 * dest mac. So, it has to be deleted, otherwise, we
2598 		 * will endup receiving vlan packets for which the
2599 		 * filter is not programmed, when hw-vlan-filter
2600 		 * configuration is ON
2601 		 */
2602 		bnxt_del_dflt_mac_filter(bp, vnic);
2603 		/* This filter will allow only untagged packets */
2604 		bnxt_add_vlan_filter(bp, 0);
2605 	}
2606 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2607 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2608 
2609 	return 0;
2610 }
2611 
2612 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2613 {
2614 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2615 	unsigned int i;
2616 	int rc;
2617 
2618 	/* Destroy vnic filters and vnic */
2619 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2620 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2621 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2622 			bnxt_del_vlan_filter(bp, i);
2623 	}
2624 	bnxt_del_dflt_mac_filter(bp, vnic);
2625 
2626 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2627 	if (rc)
2628 		return rc;
2629 
2630 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2631 	if (rc)
2632 		return rc;
2633 
2634 	rte_free(vnic->fw_grp_ids);
2635 	vnic->fw_grp_ids = NULL;
2636 
2637 	vnic->rx_queue_cnt = 0;
2638 
2639 	return 0;
2640 }
2641 
2642 static int
2643 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2644 {
2645 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2646 	int rc;
2647 
2648 	/* Destroy, recreate and reconfigure the default vnic */
2649 	rc = bnxt_free_one_vnic(bp, 0);
2650 	if (rc)
2651 		return rc;
2652 
2653 	/* default vnic 0 */
2654 	rc = bnxt_setup_one_vnic(bp, 0);
2655 	if (rc)
2656 		return rc;
2657 
2658 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2659 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2660 		rc = bnxt_add_vlan_filter(bp, 0);
2661 		if (rc)
2662 			return rc;
2663 		rc = bnxt_restore_vlan_filters(bp);
2664 		if (rc)
2665 			return rc;
2666 	} else {
2667 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2668 		if (rc)
2669 			return rc;
2670 	}
2671 
2672 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2673 	if (rc)
2674 		return rc;
2675 
2676 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2677 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2678 
2679 	return rc;
2680 }
2681 
2682 static int
2683 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2684 {
2685 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2686 	struct bnxt *bp = dev->data->dev_private;
2687 	int rc;
2688 
2689 	rc = is_bnxt_in_error(bp);
2690 	if (rc)
2691 		return rc;
2692 
2693 	/* Filter settings will get applied when port is started */
2694 	if (!dev->data->dev_started)
2695 		return 0;
2696 
2697 	if (mask & ETH_VLAN_FILTER_MASK) {
2698 		/* Enable or disable VLAN filtering */
2699 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2700 		if (rc)
2701 			return rc;
2702 	}
2703 
2704 	if (mask & ETH_VLAN_STRIP_MASK) {
2705 		/* Enable or disable VLAN stripping */
2706 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2707 		if (rc)
2708 			return rc;
2709 	}
2710 
2711 	if (mask & ETH_VLAN_EXTEND_MASK) {
2712 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2713 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2714 		else
2715 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2716 	}
2717 
2718 	return 0;
2719 }
2720 
2721 static int
2722 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2723 		      uint16_t tpid)
2724 {
2725 	struct bnxt *bp = dev->data->dev_private;
2726 	int qinq = dev->data->dev_conf.rxmode.offloads &
2727 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2728 
2729 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2730 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2731 		PMD_DRV_LOG(ERR,
2732 			    "Unsupported vlan type.");
2733 		return -EINVAL;
2734 	}
2735 	if (!qinq) {
2736 		PMD_DRV_LOG(ERR,
2737 			    "QinQ not enabled. Needs to be ON as we can "
2738 			    "accelerate only outer vlan\n");
2739 		return -EINVAL;
2740 	}
2741 
2742 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2743 		switch (tpid) {
2744 		case RTE_ETHER_TYPE_QINQ:
2745 			bp->outer_tpid_bd =
2746 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2747 				break;
2748 		case RTE_ETHER_TYPE_VLAN:
2749 			bp->outer_tpid_bd =
2750 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2751 				break;
2752 		case RTE_ETHER_TYPE_QINQ1:
2753 			bp->outer_tpid_bd =
2754 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2755 				break;
2756 		case RTE_ETHER_TYPE_QINQ2:
2757 			bp->outer_tpid_bd =
2758 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2759 				break;
2760 		case RTE_ETHER_TYPE_QINQ3:
2761 			bp->outer_tpid_bd =
2762 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2763 				break;
2764 		default:
2765 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2766 			return -EINVAL;
2767 		}
2768 		bp->outer_tpid_bd |= tpid;
2769 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2770 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2771 		PMD_DRV_LOG(ERR,
2772 			    "Can accelerate only outer vlan in QinQ\n");
2773 		return -EINVAL;
2774 	}
2775 
2776 	return 0;
2777 }
2778 
2779 static int
2780 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2781 			     struct rte_ether_addr *addr)
2782 {
2783 	struct bnxt *bp = dev->data->dev_private;
2784 	/* Default Filter is tied to VNIC 0 */
2785 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2786 	int rc;
2787 
2788 	rc = is_bnxt_in_error(bp);
2789 	if (rc)
2790 		return rc;
2791 
2792 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2793 		return -EPERM;
2794 
2795 	if (rte_is_zero_ether_addr(addr))
2796 		return -EINVAL;
2797 
2798 	/* Filter settings will get applied when port is started */
2799 	if (!dev->data->dev_started)
2800 		return 0;
2801 
2802 	/* Check if the requested MAC is already added */
2803 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2804 		return 0;
2805 
2806 	/* Destroy filter and re-create it */
2807 	bnxt_del_dflt_mac_filter(bp, vnic);
2808 
2809 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2810 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2811 		/* This filter will allow only untagged packets */
2812 		rc = bnxt_add_vlan_filter(bp, 0);
2813 	} else {
2814 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2815 	}
2816 
2817 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2818 	return rc;
2819 }
2820 
2821 static int
2822 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2823 			  struct rte_ether_addr *mc_addr_set,
2824 			  uint32_t nb_mc_addr)
2825 {
2826 	struct bnxt *bp = eth_dev->data->dev_private;
2827 	char *mc_addr_list = (char *)mc_addr_set;
2828 	struct bnxt_vnic_info *vnic;
2829 	uint32_t off = 0, i = 0;
2830 	int rc;
2831 
2832 	rc = is_bnxt_in_error(bp);
2833 	if (rc)
2834 		return rc;
2835 
2836 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2837 
2838 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2839 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2840 		goto allmulti;
2841 	}
2842 
2843 	/* TODO Check for Duplicate mcast addresses */
2844 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2845 	for (i = 0; i < nb_mc_addr; i++) {
2846 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2847 			RTE_ETHER_ADDR_LEN);
2848 		off += RTE_ETHER_ADDR_LEN;
2849 	}
2850 
2851 	vnic->mc_addr_cnt = i;
2852 	if (vnic->mc_addr_cnt)
2853 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2854 	else
2855 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2856 
2857 allmulti:
2858 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2859 }
2860 
2861 static int
2862 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2863 {
2864 	struct bnxt *bp = dev->data->dev_private;
2865 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2866 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2867 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2868 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2869 	int ret;
2870 
2871 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2872 			fw_major, fw_minor, fw_updt, fw_rsvd);
2873 	if (ret < 0)
2874 		return -EINVAL;
2875 
2876 	ret += 1; /* add the size of '\0' */
2877 	if (fw_size < (size_t)ret)
2878 		return ret;
2879 	else
2880 		return 0;
2881 }
2882 
2883 static void
2884 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2885 	struct rte_eth_rxq_info *qinfo)
2886 {
2887 	struct bnxt *bp = dev->data->dev_private;
2888 	struct bnxt_rx_queue *rxq;
2889 
2890 	if (is_bnxt_in_error(bp))
2891 		return;
2892 
2893 	rxq = dev->data->rx_queues[queue_id];
2894 
2895 	qinfo->mp = rxq->mb_pool;
2896 	qinfo->scattered_rx = dev->data->scattered_rx;
2897 	qinfo->nb_desc = rxq->nb_rx_desc;
2898 
2899 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2900 	qinfo->conf.rx_drop_en = rxq->drop_en;
2901 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2902 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2903 }
2904 
2905 static void
2906 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2907 	struct rte_eth_txq_info *qinfo)
2908 {
2909 	struct bnxt *bp = dev->data->dev_private;
2910 	struct bnxt_tx_queue *txq;
2911 
2912 	if (is_bnxt_in_error(bp))
2913 		return;
2914 
2915 	txq = dev->data->tx_queues[queue_id];
2916 
2917 	qinfo->nb_desc = txq->nb_tx_desc;
2918 
2919 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2920 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2921 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2922 
2923 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2924 	qinfo->conf.tx_rs_thresh = 0;
2925 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2926 	qinfo->conf.offloads = txq->offloads;
2927 }
2928 
2929 static const struct {
2930 	eth_rx_burst_t pkt_burst;
2931 	const char *info;
2932 } bnxt_rx_burst_info[] = {
2933 	{bnxt_recv_pkts,		"Scalar"},
2934 #if defined(RTE_ARCH_X86)
2935 	{bnxt_recv_pkts_vec,		"Vector SSE"},
2936 #endif
2937 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2938 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
2939 #endif
2940 #if defined(RTE_ARCH_ARM64)
2941 	{bnxt_recv_pkts_vec,		"Vector Neon"},
2942 #endif
2943 };
2944 
2945 static int
2946 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2947 		       struct rte_eth_burst_mode *mode)
2948 {
2949 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2950 	size_t i;
2951 
2952 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2953 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2954 			snprintf(mode->info, sizeof(mode->info), "%s",
2955 				 bnxt_rx_burst_info[i].info);
2956 			return 0;
2957 		}
2958 	}
2959 
2960 	return -EINVAL;
2961 }
2962 
2963 static const struct {
2964 	eth_tx_burst_t pkt_burst;
2965 	const char *info;
2966 } bnxt_tx_burst_info[] = {
2967 	{bnxt_xmit_pkts,		"Scalar"},
2968 #if defined(RTE_ARCH_X86)
2969 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
2970 #endif
2971 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2972 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
2973 #endif
2974 #if defined(RTE_ARCH_ARM64)
2975 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
2976 #endif
2977 };
2978 
2979 static int
2980 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2981 		       struct rte_eth_burst_mode *mode)
2982 {
2983 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2984 	size_t i;
2985 
2986 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2987 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2988 			snprintf(mode->info, sizeof(mode->info), "%s",
2989 				 bnxt_tx_burst_info[i].info);
2990 			return 0;
2991 		}
2992 	}
2993 
2994 	return -EINVAL;
2995 }
2996 
2997 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2998 {
2999 	struct bnxt *bp = eth_dev->data->dev_private;
3000 	uint32_t new_pkt_size;
3001 	uint32_t rc = 0;
3002 	uint32_t i;
3003 
3004 	rc = is_bnxt_in_error(bp);
3005 	if (rc)
3006 		return rc;
3007 
3008 	/* Exit if receive queues are not configured yet */
3009 	if (!eth_dev->data->nb_rx_queues)
3010 		return rc;
3011 
3012 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
3013 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
3014 
3015 	/*
3016 	 * Disallow any MTU change that would require scattered receive support
3017 	 * if it is not already enabled.
3018 	 */
3019 	if (eth_dev->data->dev_started &&
3020 	    !eth_dev->data->scattered_rx &&
3021 	    (new_pkt_size >
3022 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
3023 		PMD_DRV_LOG(ERR,
3024 			    "MTU change would require scattered rx support. ");
3025 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
3026 		return -EINVAL;
3027 	}
3028 
3029 	if (new_mtu > RTE_ETHER_MTU) {
3030 		bp->flags |= BNXT_FLAG_JUMBO;
3031 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
3032 			DEV_RX_OFFLOAD_JUMBO_FRAME;
3033 	} else {
3034 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
3035 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
3036 		bp->flags &= ~BNXT_FLAG_JUMBO;
3037 	}
3038 
3039 	/* Is there a change in mtu setting? */
3040 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
3041 		return rc;
3042 
3043 	for (i = 0; i < bp->nr_vnics; i++) {
3044 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3045 		uint16_t size = 0;
3046 
3047 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
3048 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
3049 		if (rc)
3050 			break;
3051 
3052 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
3053 		size -= RTE_PKTMBUF_HEADROOM;
3054 
3055 		if (size < new_mtu) {
3056 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
3057 			if (rc)
3058 				return rc;
3059 		}
3060 	}
3061 
3062 	if (!rc)
3063 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
3064 
3065 	if (bnxt_hwrm_config_host_mtu(bp))
3066 		PMD_DRV_LOG(WARNING, "Failed to configure host MTU\n");
3067 
3068 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3069 
3070 	return rc;
3071 }
3072 
3073 static int
3074 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3075 {
3076 	struct bnxt *bp = dev->data->dev_private;
3077 	uint16_t vlan = bp->vlan;
3078 	int rc;
3079 
3080 	rc = is_bnxt_in_error(bp);
3081 	if (rc)
3082 		return rc;
3083 
3084 	if (!BNXT_SINGLE_PF(bp)) {
3085 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3086 		return -ENOTSUP;
3087 	}
3088 	bp->vlan = on ? pvid : 0;
3089 
3090 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3091 	if (rc)
3092 		bp->vlan = vlan;
3093 	return rc;
3094 }
3095 
3096 static int
3097 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3098 {
3099 	struct bnxt *bp = dev->data->dev_private;
3100 	int rc;
3101 
3102 	rc = is_bnxt_in_error(bp);
3103 	if (rc)
3104 		return rc;
3105 
3106 	return bnxt_hwrm_port_led_cfg(bp, true);
3107 }
3108 
3109 static int
3110 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3111 {
3112 	struct bnxt *bp = dev->data->dev_private;
3113 	int rc;
3114 
3115 	rc = is_bnxt_in_error(bp);
3116 	if (rc)
3117 		return rc;
3118 
3119 	return bnxt_hwrm_port_led_cfg(bp, false);
3120 }
3121 
3122 static uint32_t
3123 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
3124 {
3125 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
3126 	struct bnxt_cp_ring_info *cpr;
3127 	uint32_t desc = 0, raw_cons, cp_ring_size;
3128 	struct bnxt_rx_queue *rxq;
3129 	struct rx_pkt_cmpl *rxcmp;
3130 	int rc;
3131 
3132 	rc = is_bnxt_in_error(bp);
3133 	if (rc)
3134 		return rc;
3135 
3136 	rxq = dev->data->rx_queues[rx_queue_id];
3137 	cpr = rxq->cp_ring;
3138 	raw_cons = cpr->cp_raw_cons;
3139 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3140 
3141 	while (1) {
3142 		uint32_t agg_cnt, cons, cmpl_type;
3143 
3144 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3145 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3146 
3147 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3148 			break;
3149 
3150 		cmpl_type = CMP_TYPE(rxcmp);
3151 
3152 		switch (cmpl_type) {
3153 		case CMPL_BASE_TYPE_RX_L2:
3154 		case CMPL_BASE_TYPE_RX_L2_V2:
3155 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3156 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3157 			desc++;
3158 			break;
3159 
3160 		case CMPL_BASE_TYPE_RX_TPA_END:
3161 			if (BNXT_CHIP_P5(rxq->bp)) {
3162 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3163 
3164 				p5_tpa_end = (void *)rxcmp;
3165 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3166 			} else {
3167 				struct rx_tpa_end_cmpl *tpa_end;
3168 
3169 				tpa_end = (void *)rxcmp;
3170 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3171 			}
3172 
3173 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3174 			desc++;
3175 			break;
3176 
3177 		default:
3178 			raw_cons += CMP_LEN(cmpl_type);
3179 		}
3180 	}
3181 
3182 	return desc;
3183 }
3184 
3185 static int
3186 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3187 {
3188 	struct bnxt_rx_queue *rxq = rx_queue;
3189 	struct bnxt_cp_ring_info *cpr;
3190 	struct bnxt_rx_ring_info *rxr;
3191 	uint32_t desc, raw_cons, cp_ring_size;
3192 	struct bnxt *bp = rxq->bp;
3193 	struct rx_pkt_cmpl *rxcmp;
3194 	int rc;
3195 
3196 	rc = is_bnxt_in_error(bp);
3197 	if (rc)
3198 		return rc;
3199 
3200 	if (offset >= rxq->nb_rx_desc)
3201 		return -EINVAL;
3202 
3203 	rxr = rxq->rx_ring;
3204 	cpr = rxq->cp_ring;
3205 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3206 
3207 	/*
3208 	 * For the vector receive case, the completion at the requested
3209 	 * offset can be indexed directly.
3210 	 */
3211 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3212 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3213 		struct rx_pkt_cmpl *rxcmp;
3214 		uint32_t cons;
3215 
3216 		/* Check status of completion descriptor. */
3217 		raw_cons = cpr->cp_raw_cons +
3218 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3219 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3220 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3221 
3222 		if (bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3223 			return RTE_ETH_RX_DESC_DONE;
3224 
3225 		/* Check whether rx desc has an mbuf attached. */
3226 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3227 		if (cons >= rxq->rxrearm_start &&
3228 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3229 			return RTE_ETH_RX_DESC_UNAVAIL;
3230 		}
3231 
3232 		return RTE_ETH_RX_DESC_AVAIL;
3233 	}
3234 #endif
3235 
3236 	/*
3237 	 * For the non-vector receive case, scan the completion ring to
3238 	 * locate the completion descriptor for the requested offset.
3239 	 */
3240 	raw_cons = cpr->cp_raw_cons;
3241 	desc = 0;
3242 	while (1) {
3243 		uint32_t agg_cnt, cons, cmpl_type;
3244 
3245 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3246 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3247 
3248 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3249 			break;
3250 
3251 		cmpl_type = CMP_TYPE(rxcmp);
3252 
3253 		switch (cmpl_type) {
3254 		case CMPL_BASE_TYPE_RX_L2:
3255 		case CMPL_BASE_TYPE_RX_L2_V2:
3256 			if (desc == offset) {
3257 				cons = rxcmp->opaque;
3258 				if (rxr->rx_buf_ring[cons])
3259 					return RTE_ETH_RX_DESC_DONE;
3260 				else
3261 					return RTE_ETH_RX_DESC_UNAVAIL;
3262 			}
3263 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3264 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3265 			desc++;
3266 			break;
3267 
3268 		case CMPL_BASE_TYPE_RX_TPA_END:
3269 			if (desc == offset)
3270 				return RTE_ETH_RX_DESC_DONE;
3271 
3272 			if (BNXT_CHIP_P5(rxq->bp)) {
3273 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3274 
3275 				p5_tpa_end = (void *)rxcmp;
3276 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3277 			} else {
3278 				struct rx_tpa_end_cmpl *tpa_end;
3279 
3280 				tpa_end = (void *)rxcmp;
3281 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3282 			}
3283 
3284 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3285 			desc++;
3286 			break;
3287 
3288 		default:
3289 			raw_cons += CMP_LEN(cmpl_type);
3290 		}
3291 	}
3292 
3293 	return RTE_ETH_RX_DESC_AVAIL;
3294 }
3295 
3296 static int
3297 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3298 {
3299 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3300 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3301 	uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3302 	struct cmpl_base *cp_desc_ring;
3303 	int rc;
3304 
3305 	rc = is_bnxt_in_error(txq->bp);
3306 	if (rc)
3307 		return rc;
3308 
3309 	if (offset >= txq->nb_tx_desc)
3310 		return -EINVAL;
3311 
3312 	/* Return "desc done" if descriptor is available for use. */
3313 	if (bnxt_tx_bds_in_hw(txq) <= offset)
3314 		return RTE_ETH_TX_DESC_DONE;
3315 
3316 	raw_cons = cpr->cp_raw_cons;
3317 	cp_desc_ring = cpr->cp_desc_ring;
3318 	ring_mask = cpr->cp_ring_struct->ring_mask;
3319 
3320 	/* Check to see if hw has posted a completion for the descriptor. */
3321 	while (1) {
3322 		struct tx_cmpl *txcmp;
3323 		uint32_t cons;
3324 
3325 		cons = RING_CMPL(ring_mask, raw_cons);
3326 		txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3327 
3328 		if (!bnxt_cpr_cmp_valid(txcmp, raw_cons, ring_mask + 1))
3329 			break;
3330 
3331 		if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3332 			nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3333 
3334 		if (nb_tx_pkts > offset)
3335 			return RTE_ETH_TX_DESC_DONE;
3336 
3337 		raw_cons = NEXT_RAW_CMP(raw_cons);
3338 	}
3339 
3340 	/* Descriptor is pending transmit, not yet completed by hardware. */
3341 	return RTE_ETH_TX_DESC_FULL;
3342 }
3343 
3344 int
3345 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3346 		     const struct rte_flow_ops **ops)
3347 {
3348 	struct bnxt *bp = dev->data->dev_private;
3349 	int ret = 0;
3350 
3351 	if (!bp)
3352 		return -EIO;
3353 
3354 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3355 		struct bnxt_representor *vfr = dev->data->dev_private;
3356 		bp = vfr->parent_dev->data->dev_private;
3357 		/* parent is deleted while children are still valid */
3358 		if (!bp) {
3359 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3360 				    dev->data->port_id);
3361 			return -EIO;
3362 		}
3363 	}
3364 
3365 	ret = is_bnxt_in_error(bp);
3366 	if (ret)
3367 		return ret;
3368 
3369 	/* PMD supports thread-safe flow operations.  rte_flow API
3370 	 * functions can avoid mutex for multi-thread safety.
3371 	 */
3372 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3373 
3374 	if (BNXT_TRUFLOW_EN(bp))
3375 		*ops = &bnxt_ulp_rte_flow_ops;
3376 	else
3377 		*ops = &bnxt_flow_ops;
3378 
3379 	return ret;
3380 }
3381 
3382 static const uint32_t *
3383 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3384 {
3385 	static const uint32_t ptypes[] = {
3386 		RTE_PTYPE_L2_ETHER_VLAN,
3387 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3388 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3389 		RTE_PTYPE_L4_ICMP,
3390 		RTE_PTYPE_L4_TCP,
3391 		RTE_PTYPE_L4_UDP,
3392 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3393 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3394 		RTE_PTYPE_INNER_L4_ICMP,
3395 		RTE_PTYPE_INNER_L4_TCP,
3396 		RTE_PTYPE_INNER_L4_UDP,
3397 		RTE_PTYPE_UNKNOWN
3398 	};
3399 
3400 	if (!dev->rx_pkt_burst)
3401 		return NULL;
3402 
3403 	return ptypes;
3404 }
3405 
3406 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3407 			 int reg_win)
3408 {
3409 	uint32_t reg_base = *reg_arr & 0xfffff000;
3410 	uint32_t win_off;
3411 	int i;
3412 
3413 	for (i = 0; i < count; i++) {
3414 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3415 			return -ERANGE;
3416 	}
3417 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3418 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3419 	return 0;
3420 }
3421 
3422 static int bnxt_map_ptp_regs(struct bnxt *bp)
3423 {
3424 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3425 	uint32_t *reg_arr;
3426 	int rc, i;
3427 
3428 	reg_arr = ptp->rx_regs;
3429 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3430 	if (rc)
3431 		return rc;
3432 
3433 	reg_arr = ptp->tx_regs;
3434 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3435 	if (rc)
3436 		return rc;
3437 
3438 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3439 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3440 
3441 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3442 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3443 
3444 	return 0;
3445 }
3446 
3447 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3448 {
3449 	rte_write32(0, (uint8_t *)bp->bar0 +
3450 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3451 	rte_write32(0, (uint8_t *)bp->bar0 +
3452 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3453 }
3454 
3455 static uint64_t bnxt_cc_read(struct bnxt *bp)
3456 {
3457 	uint64_t ns;
3458 
3459 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3460 			      BNXT_GRCPF_REG_SYNC_TIME));
3461 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3462 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3463 	return ns;
3464 }
3465 
3466 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3467 {
3468 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3469 	uint32_t fifo;
3470 
3471 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3472 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3473 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3474 		return -EAGAIN;
3475 
3476 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3477 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3478 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3479 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3480 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3481 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3482 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3483 
3484 	return 0;
3485 }
3486 
3487 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3488 {
3489 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3490 	struct bnxt_pf_info *pf = bp->pf;
3491 	uint16_t port_id;
3492 	int i = 0;
3493 	uint32_t fifo;
3494 
3495 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3496 		return -EINVAL;
3497 
3498 	port_id = pf->port_id;
3499 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3500 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3501 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3502 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3503 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3504 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3505 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3506 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3507 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3508 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3509 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3510 		i++;
3511 	}
3512 
3513 	if (i >= BNXT_PTP_RX_PND_CNT)
3514 		return -EBUSY;
3515 
3516 	return 0;
3517 }
3518 
3519 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3520 {
3521 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3522 	struct bnxt_pf_info *pf = bp->pf;
3523 	uint16_t port_id;
3524 	uint32_t fifo;
3525 
3526 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3527 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3528 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3529 		return -EAGAIN;
3530 
3531 	port_id = pf->port_id;
3532 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3533 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3534 
3535 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3536 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3537 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3538 		return bnxt_clr_rx_ts(bp, ts);
3539 
3540 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3541 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3542 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3543 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3544 
3545 	return 0;
3546 }
3547 
3548 static int
3549 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3550 {
3551 	uint64_t ns;
3552 	struct bnxt *bp = dev->data->dev_private;
3553 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3554 
3555 	if (!ptp)
3556 		return -ENOTSUP;
3557 
3558 	ns = rte_timespec_to_ns(ts);
3559 	/* Set the timecounters to a new value. */
3560 	ptp->tc.nsec = ns;
3561 	ptp->tx_tstamp_tc.nsec = ns;
3562 	ptp->rx_tstamp_tc.nsec = ns;
3563 
3564 	return 0;
3565 }
3566 
3567 static int
3568 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3569 {
3570 	struct bnxt *bp = dev->data->dev_private;
3571 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3572 	uint64_t ns, systime_cycles = 0;
3573 	int rc = 0;
3574 
3575 	if (!ptp)
3576 		return -ENOTSUP;
3577 
3578 	if (BNXT_CHIP_P5(bp))
3579 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3580 					     &systime_cycles);
3581 	else
3582 		systime_cycles = bnxt_cc_read(bp);
3583 
3584 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3585 	*ts = rte_ns_to_timespec(ns);
3586 
3587 	return rc;
3588 }
3589 static int
3590 bnxt_timesync_enable(struct rte_eth_dev *dev)
3591 {
3592 	struct bnxt *bp = dev->data->dev_private;
3593 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3594 	uint32_t shift = 0;
3595 	int rc;
3596 
3597 	if (!ptp)
3598 		return -ENOTSUP;
3599 
3600 	ptp->rx_filter = 1;
3601 	ptp->tx_tstamp_en = 1;
3602 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3603 
3604 	rc = bnxt_hwrm_ptp_cfg(bp);
3605 	if (rc)
3606 		return rc;
3607 
3608 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3609 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3610 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3611 
3612 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3613 	ptp->tc.cc_shift = shift;
3614 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3615 
3616 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3617 	ptp->rx_tstamp_tc.cc_shift = shift;
3618 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3619 
3620 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3621 	ptp->tx_tstamp_tc.cc_shift = shift;
3622 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3623 
3624 	if (!BNXT_CHIP_P5(bp))
3625 		bnxt_map_ptp_regs(bp);
3626 	else
3627 		rc = bnxt_ptp_start(bp);
3628 
3629 	return rc;
3630 }
3631 
3632 static int
3633 bnxt_timesync_disable(struct rte_eth_dev *dev)
3634 {
3635 	struct bnxt *bp = dev->data->dev_private;
3636 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3637 
3638 	if (!ptp)
3639 		return -ENOTSUP;
3640 
3641 	ptp->rx_filter = 0;
3642 	ptp->tx_tstamp_en = 0;
3643 	ptp->rxctl = 0;
3644 
3645 	bnxt_hwrm_ptp_cfg(bp);
3646 
3647 	if (!BNXT_CHIP_P5(bp))
3648 		bnxt_unmap_ptp_regs(bp);
3649 	else
3650 		bnxt_ptp_stop(bp);
3651 
3652 	return 0;
3653 }
3654 
3655 static int
3656 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3657 				 struct timespec *timestamp,
3658 				 uint32_t flags __rte_unused)
3659 {
3660 	struct bnxt *bp = dev->data->dev_private;
3661 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3662 	uint64_t rx_tstamp_cycles = 0;
3663 	uint64_t ns;
3664 
3665 	if (!ptp)
3666 		return -ENOTSUP;
3667 
3668 	if (BNXT_CHIP_P5(bp))
3669 		rx_tstamp_cycles = ptp->rx_timestamp;
3670 	else
3671 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3672 
3673 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3674 	*timestamp = rte_ns_to_timespec(ns);
3675 	return  0;
3676 }
3677 
3678 static int
3679 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3680 				 struct timespec *timestamp)
3681 {
3682 	struct bnxt *bp = dev->data->dev_private;
3683 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3684 	uint64_t tx_tstamp_cycles = 0;
3685 	uint64_t ns;
3686 	int rc = 0;
3687 
3688 	if (!ptp)
3689 		return -ENOTSUP;
3690 
3691 	if (BNXT_CHIP_P5(bp))
3692 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3693 					     &tx_tstamp_cycles);
3694 	else
3695 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3696 
3697 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3698 	*timestamp = rte_ns_to_timespec(ns);
3699 
3700 	return rc;
3701 }
3702 
3703 static int
3704 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3705 {
3706 	struct bnxt *bp = dev->data->dev_private;
3707 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3708 
3709 	if (!ptp)
3710 		return -ENOTSUP;
3711 
3712 	ptp->tc.nsec += delta;
3713 	ptp->tx_tstamp_tc.nsec += delta;
3714 	ptp->rx_tstamp_tc.nsec += delta;
3715 
3716 	return 0;
3717 }
3718 
3719 static int
3720 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3721 {
3722 	struct bnxt *bp = dev->data->dev_private;
3723 	int rc;
3724 	uint32_t dir_entries;
3725 	uint32_t entry_length;
3726 
3727 	rc = is_bnxt_in_error(bp);
3728 	if (rc)
3729 		return rc;
3730 
3731 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3732 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3733 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3734 
3735 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3736 	if (rc != 0)
3737 		return rc;
3738 
3739 	return dir_entries * entry_length;
3740 }
3741 
3742 static int
3743 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3744 		struct rte_dev_eeprom_info *in_eeprom)
3745 {
3746 	struct bnxt *bp = dev->data->dev_private;
3747 	uint32_t index;
3748 	uint32_t offset;
3749 	int rc;
3750 
3751 	rc = is_bnxt_in_error(bp);
3752 	if (rc)
3753 		return rc;
3754 
3755 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3756 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3757 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3758 		    in_eeprom->offset, in_eeprom->length);
3759 
3760 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3761 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3762 						in_eeprom->data);
3763 
3764 	index = in_eeprom->offset >> 24;
3765 	offset = in_eeprom->offset & 0xffffff;
3766 
3767 	if (index != 0)
3768 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3769 					   in_eeprom->length, in_eeprom->data);
3770 
3771 	return 0;
3772 }
3773 
3774 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3775 {
3776 	switch (dir_type) {
3777 	case BNX_DIR_TYPE_CHIMP_PATCH:
3778 	case BNX_DIR_TYPE_BOOTCODE:
3779 	case BNX_DIR_TYPE_BOOTCODE_2:
3780 	case BNX_DIR_TYPE_APE_FW:
3781 	case BNX_DIR_TYPE_APE_PATCH:
3782 	case BNX_DIR_TYPE_KONG_FW:
3783 	case BNX_DIR_TYPE_KONG_PATCH:
3784 	case BNX_DIR_TYPE_BONO_FW:
3785 	case BNX_DIR_TYPE_BONO_PATCH:
3786 		/* FALLTHROUGH */
3787 		return true;
3788 	}
3789 
3790 	return false;
3791 }
3792 
3793 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3794 {
3795 	switch (dir_type) {
3796 	case BNX_DIR_TYPE_AVS:
3797 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3798 	case BNX_DIR_TYPE_PCIE:
3799 	case BNX_DIR_TYPE_TSCF_UCODE:
3800 	case BNX_DIR_TYPE_EXT_PHY:
3801 	case BNX_DIR_TYPE_CCM:
3802 	case BNX_DIR_TYPE_ISCSI_BOOT:
3803 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3804 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3805 		/* FALLTHROUGH */
3806 		return true;
3807 	}
3808 
3809 	return false;
3810 }
3811 
3812 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3813 {
3814 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3815 		bnxt_dir_type_is_other_exec_format(dir_type);
3816 }
3817 
3818 static int
3819 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3820 		struct rte_dev_eeprom_info *in_eeprom)
3821 {
3822 	struct bnxt *bp = dev->data->dev_private;
3823 	uint8_t index, dir_op;
3824 	uint16_t type, ext, ordinal, attr;
3825 	int rc;
3826 
3827 	rc = is_bnxt_in_error(bp);
3828 	if (rc)
3829 		return rc;
3830 
3831 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3832 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3833 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3834 		    in_eeprom->offset, in_eeprom->length);
3835 
3836 	if (!BNXT_PF(bp)) {
3837 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3838 		return -EINVAL;
3839 	}
3840 
3841 	type = in_eeprom->magic >> 16;
3842 
3843 	if (type == 0xffff) { /* special value for directory operations */
3844 		index = in_eeprom->magic & 0xff;
3845 		dir_op = in_eeprom->magic >> 8;
3846 		if (index == 0)
3847 			return -EINVAL;
3848 		switch (dir_op) {
3849 		case 0x0e: /* erase */
3850 			if (in_eeprom->offset != ~in_eeprom->magic)
3851 				return -EINVAL;
3852 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3853 		default:
3854 			return -EINVAL;
3855 		}
3856 	}
3857 
3858 	/* Create or re-write an NVM item: */
3859 	if (bnxt_dir_type_is_executable(type) == true)
3860 		return -EOPNOTSUPP;
3861 	ext = in_eeprom->magic & 0xffff;
3862 	ordinal = in_eeprom->offset >> 16;
3863 	attr = in_eeprom->offset & 0xffff;
3864 
3865 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3866 				     in_eeprom->data, in_eeprom->length);
3867 }
3868 
3869 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3870 				struct rte_eth_dev_module_info *modinfo)
3871 {
3872 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3873 	struct bnxt *bp = dev->data->dev_private;
3874 	int rc;
3875 
3876 	/* No point in going further if phy status indicates
3877 	 * module is not inserted or if it is powered down or
3878 	 * if it is of type 10GBase-T
3879 	 */
3880 	if (bp->link_info->module_status >
3881 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3882 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3883 			    dev->data->port_id);
3884 		return -ENOTSUP;
3885 	}
3886 
3887 	/* This feature is not supported in older firmware versions */
3888 	if (bp->hwrm_spec_code < 0x10202) {
3889 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3890 			    dev->data->port_id);
3891 		return -ENOTSUP;
3892 	}
3893 
3894 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3895 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3896 						   module_info);
3897 
3898 	if (rc)
3899 		return rc;
3900 
3901 	switch (module_info[0]) {
3902 	case SFF_MODULE_ID_SFP:
3903 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3904 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3905 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3906 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3907 		break;
3908 	case SFF_MODULE_ID_QSFP:
3909 	case SFF_MODULE_ID_QSFP_PLUS:
3910 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3911 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3912 		break;
3913 	case SFF_MODULE_ID_QSFP28:
3914 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3915 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3916 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3917 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3918 		break;
3919 	default:
3920 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3921 		return -ENOTSUP;
3922 	}
3923 
3924 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3925 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3926 
3927 	return 0;
3928 }
3929 
3930 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3931 				  struct rte_dev_eeprom_info *info)
3932 {
3933 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3934 	uint32_t offset = info->offset, length = info->length;
3935 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3936 	struct bnxt *bp = dev->data->dev_private;
3937 	uint8_t *data = info->data;
3938 	uint8_t page = offset >> 7;
3939 	uint8_t max_pages = 2;
3940 	uint8_t opt_pages;
3941 	int rc;
3942 
3943 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3944 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3945 						   module_info);
3946 	if (rc)
3947 		return rc;
3948 
3949 	switch (module_info[0]) {
3950 	case SFF_MODULE_ID_SFP:
3951 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
3952 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
3953 			pg_addr[2] = I2C_DEV_ADDR_A2;
3954 			pg_addr[3] = I2C_DEV_ADDR_A2;
3955 			max_pages = 4;
3956 		}
3957 		break;
3958 	case SFF_MODULE_ID_QSFP28:
3959 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
3960 							   SFF8636_OPT_PAGES_OFFSET,
3961 							   1, &opt_pages);
3962 		if (rc)
3963 			return rc;
3964 
3965 		if (opt_pages & SFF8636_PAGE1_MASK) {
3966 			pg_addr[2] = I2C_DEV_ADDR_A0;
3967 			max_pages = 3;
3968 		}
3969 		if (opt_pages & SFF8636_PAGE2_MASK) {
3970 			pg_addr[3] = I2C_DEV_ADDR_A0;
3971 			max_pages = 4;
3972 		}
3973 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
3974 			pg_addr[4] = I2C_DEV_ADDR_A0;
3975 			max_pages = 5;
3976 		}
3977 		break;
3978 	default:
3979 		break;
3980 	}
3981 
3982 	memset(data, 0, length);
3983 
3984 	offset &= 0xff;
3985 	while (length && page < max_pages) {
3986 		uint8_t raw_page = page ? page - 1 : 0;
3987 		uint16_t chunk;
3988 
3989 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
3990 			raw_page = 0;
3991 		else if (page)
3992 			offset |= 0x80;
3993 		chunk = RTE_MIN(length, 256 - offset);
3994 
3995 		if (pg_addr[page]) {
3996 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
3997 								   raw_page, offset,
3998 								   chunk, data);
3999 			if (rc)
4000 				return rc;
4001 		}
4002 
4003 		data += chunk;
4004 		length -= chunk;
4005 		offset = 0;
4006 		page += 1 + (chunk > 128);
4007 	}
4008 
4009 	return length ? -EINVAL : 0;
4010 }
4011 
4012 /*
4013  * Initialization
4014  */
4015 
4016 static const struct eth_dev_ops bnxt_dev_ops = {
4017 	.dev_infos_get = bnxt_dev_info_get_op,
4018 	.dev_close = bnxt_dev_close_op,
4019 	.dev_configure = bnxt_dev_configure_op,
4020 	.dev_start = bnxt_dev_start_op,
4021 	.dev_stop = bnxt_dev_stop_op,
4022 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4023 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4024 	.stats_get = bnxt_stats_get_op,
4025 	.stats_reset = bnxt_stats_reset_op,
4026 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4027 	.rx_queue_release = bnxt_rx_queue_release_op,
4028 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4029 	.tx_queue_release = bnxt_tx_queue_release_op,
4030 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4031 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4032 	.reta_update = bnxt_reta_update_op,
4033 	.reta_query = bnxt_reta_query_op,
4034 	.rss_hash_update = bnxt_rss_hash_update_op,
4035 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4036 	.link_update = bnxt_link_update_op,
4037 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4038 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4039 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4040 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4041 	.mac_addr_add = bnxt_mac_addr_add_op,
4042 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4043 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4044 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4045 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4046 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4047 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4048 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4049 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4050 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4051 	.mtu_set = bnxt_mtu_set_op,
4052 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4053 	.xstats_get = bnxt_dev_xstats_get_op,
4054 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4055 	.xstats_reset = bnxt_dev_xstats_reset_op,
4056 	.fw_version_get = bnxt_fw_version_get,
4057 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4058 	.rxq_info_get = bnxt_rxq_info_get_op,
4059 	.txq_info_get = bnxt_txq_info_get_op,
4060 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4061 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4062 	.dev_led_on = bnxt_dev_led_on_op,
4063 	.dev_led_off = bnxt_dev_led_off_op,
4064 	.rx_queue_start = bnxt_rx_queue_start,
4065 	.rx_queue_stop = bnxt_rx_queue_stop,
4066 	.tx_queue_start = bnxt_tx_queue_start,
4067 	.tx_queue_stop = bnxt_tx_queue_stop,
4068 	.flow_ops_get = bnxt_flow_ops_get_op,
4069 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4070 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4071 	.get_eeprom           = bnxt_get_eeprom_op,
4072 	.set_eeprom           = bnxt_set_eeprom_op,
4073 	.get_module_info = bnxt_get_module_info,
4074 	.get_module_eeprom = bnxt_get_module_eeprom,
4075 	.timesync_enable      = bnxt_timesync_enable,
4076 	.timesync_disable     = bnxt_timesync_disable,
4077 	.timesync_read_time   = bnxt_timesync_read_time,
4078 	.timesync_write_time   = bnxt_timesync_write_time,
4079 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4080 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4081 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4082 };
4083 
4084 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4085 {
4086 	uint32_t offset;
4087 
4088 	/* Only pre-map the reset GRC registers using window 3 */
4089 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4090 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4091 
4092 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4093 
4094 	return offset;
4095 }
4096 
4097 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4098 {
4099 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4100 	uint32_t reg_base = 0xffffffff;
4101 	int i;
4102 
4103 	/* Only pre-map the monitoring GRC registers using window 2 */
4104 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4105 		uint32_t reg = info->status_regs[i];
4106 
4107 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4108 			continue;
4109 
4110 		if (reg_base == 0xffffffff)
4111 			reg_base = reg & 0xfffff000;
4112 		if ((reg & 0xfffff000) != reg_base)
4113 			return -ERANGE;
4114 
4115 		/* Use mask 0xffc as the Lower 2 bits indicates
4116 		 * address space location
4117 		 */
4118 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4119 						(reg & 0xffc);
4120 	}
4121 
4122 	if (reg_base == 0xffffffff)
4123 		return 0;
4124 
4125 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4126 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4127 
4128 	return 0;
4129 }
4130 
4131 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4132 {
4133 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4134 	uint32_t delay = info->delay_after_reset[index];
4135 	uint32_t val = info->reset_reg_val[index];
4136 	uint32_t reg = info->reset_reg[index];
4137 	uint32_t type, offset;
4138 	int ret;
4139 
4140 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4141 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4142 
4143 	switch (type) {
4144 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4145 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4146 		if (ret < 0) {
4147 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4148 				    val, offset);
4149 			return;
4150 		}
4151 		break;
4152 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4153 		offset = bnxt_map_reset_regs(bp, offset);
4154 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4155 		break;
4156 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4157 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4158 		break;
4159 	}
4160 	/* wait on a specific interval of time until core reset is complete */
4161 	if (delay)
4162 		rte_delay_ms(delay);
4163 }
4164 
4165 static void bnxt_dev_cleanup(struct bnxt *bp)
4166 {
4167 	bp->eth_dev->data->dev_link.link_status = 0;
4168 	bp->link_info->link_up = 0;
4169 	if (bp->eth_dev->data->dev_started)
4170 		bnxt_dev_stop(bp->eth_dev);
4171 
4172 	bnxt_uninit_resources(bp, true);
4173 }
4174 
4175 static int
4176 bnxt_check_fw_reset_done(struct bnxt *bp)
4177 {
4178 	int timeout = bp->fw_reset_max_msecs;
4179 	uint16_t val = 0;
4180 	int rc;
4181 
4182 	do {
4183 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4184 		if (rc < 0) {
4185 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4186 			return rc;
4187 		}
4188 		if (val != 0xffff)
4189 			break;
4190 		rte_delay_ms(1);
4191 	} while (timeout--);
4192 
4193 	if (val == 0xffff) {
4194 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4195 		return -1;
4196 	}
4197 
4198 	return 0;
4199 }
4200 
4201 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4202 {
4203 	struct rte_eth_dev *dev = bp->eth_dev;
4204 	struct rte_vlan_filter_conf *vfc;
4205 	int vidx, vbit, rc;
4206 	uint16_t vlan_id;
4207 
4208 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4209 		vfc = &dev->data->vlan_filter_conf;
4210 		vidx = vlan_id / 64;
4211 		vbit = vlan_id % 64;
4212 
4213 		/* Each bit corresponds to a VLAN id */
4214 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4215 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4216 			if (rc)
4217 				return rc;
4218 		}
4219 	}
4220 
4221 	return 0;
4222 }
4223 
4224 static int bnxt_restore_mac_filters(struct bnxt *bp)
4225 {
4226 	struct rte_eth_dev *dev = bp->eth_dev;
4227 	struct rte_eth_dev_info dev_info;
4228 	struct rte_ether_addr *addr;
4229 	uint64_t pool_mask;
4230 	uint32_t pool = 0;
4231 	uint32_t i;
4232 	int rc;
4233 
4234 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4235 		return 0;
4236 
4237 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4238 	if (rc)
4239 		return rc;
4240 
4241 	/* replay MAC address configuration */
4242 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4243 		addr = &dev->data->mac_addrs[i];
4244 
4245 		/* skip zero address */
4246 		if (rte_is_zero_ether_addr(addr))
4247 			continue;
4248 
4249 		pool = 0;
4250 		pool_mask = dev->data->mac_pool_sel[i];
4251 
4252 		do {
4253 			if (pool_mask & 1ULL) {
4254 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4255 				if (rc)
4256 					return rc;
4257 			}
4258 			pool_mask >>= 1;
4259 			pool++;
4260 		} while (pool_mask);
4261 	}
4262 
4263 	return 0;
4264 }
4265 
4266 static int bnxt_restore_filters(struct bnxt *bp)
4267 {
4268 	struct rte_eth_dev *dev = bp->eth_dev;
4269 	int ret = 0;
4270 
4271 	if (dev->data->all_multicast) {
4272 		ret = bnxt_allmulticast_enable_op(dev);
4273 		if (ret)
4274 			return ret;
4275 	}
4276 	if (dev->data->promiscuous) {
4277 		ret = bnxt_promiscuous_enable_op(dev);
4278 		if (ret)
4279 			return ret;
4280 	}
4281 
4282 	ret = bnxt_restore_mac_filters(bp);
4283 	if (ret)
4284 		return ret;
4285 
4286 	ret = bnxt_restore_vlan_filters(bp);
4287 	/* TODO restore other filters as well */
4288 	return ret;
4289 }
4290 
4291 static int bnxt_check_fw_ready(struct bnxt *bp)
4292 {
4293 	int timeout = bp->fw_reset_max_msecs;
4294 	int rc = 0;
4295 
4296 	do {
4297 		rc = bnxt_hwrm_poll_ver_get(bp);
4298 		if (rc == 0)
4299 			break;
4300 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4301 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4302 	} while (rc && timeout > 0);
4303 
4304 	if (rc)
4305 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4306 
4307 	return rc;
4308 }
4309 
4310 static void bnxt_dev_recover(void *arg)
4311 {
4312 	struct bnxt *bp = arg;
4313 	int rc = 0;
4314 
4315 	pthread_mutex_lock(&bp->err_recovery_lock);
4316 
4317 	if (!bp->fw_reset_min_msecs) {
4318 		rc = bnxt_check_fw_reset_done(bp);
4319 		if (rc)
4320 			goto err;
4321 	}
4322 
4323 	/* Clear Error flag so that device re-init should happen */
4324 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4325 
4326 	rc = bnxt_check_fw_ready(bp);
4327 	if (rc)
4328 		goto err;
4329 
4330 	rc = bnxt_init_resources(bp, true);
4331 	if (rc) {
4332 		PMD_DRV_LOG(ERR,
4333 			    "Failed to initialize resources after reset\n");
4334 		goto err;
4335 	}
4336 	/* clear reset flag as the device is initialized now */
4337 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4338 
4339 	rc = bnxt_dev_start_op(bp->eth_dev);
4340 	if (rc) {
4341 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4342 		goto err_start;
4343 	}
4344 
4345 	rc = bnxt_restore_filters(bp);
4346 	if (rc)
4347 		goto err_start;
4348 
4349 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4350 	pthread_mutex_unlock(&bp->err_recovery_lock);
4351 
4352 	return;
4353 err_start:
4354 	bnxt_dev_stop(bp->eth_dev);
4355 err:
4356 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4357 	bnxt_uninit_resources(bp, false);
4358 	if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4359 		rte_eth_dev_callback_process(bp->eth_dev,
4360 					     RTE_ETH_EVENT_INTR_RMV,
4361 					     NULL);
4362 	pthread_mutex_unlock(&bp->err_recovery_lock);
4363 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4364 }
4365 
4366 void bnxt_dev_reset_and_resume(void *arg)
4367 {
4368 	struct bnxt *bp = arg;
4369 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4370 	uint16_t val = 0;
4371 	int rc;
4372 
4373 	bnxt_dev_cleanup(bp);
4374 
4375 	bnxt_wait_for_device_shutdown(bp);
4376 
4377 	/* During some fatal firmware error conditions, the PCI config space
4378 	 * register 0x2e which normally contains the subsystem ID will become
4379 	 * 0xffff. This register will revert back to the normal value after
4380 	 * the chip has completed core reset. If we detect this condition,
4381 	 * we can poll this config register immediately for the value to revert.
4382 	 */
4383 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4384 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4385 		if (rc < 0) {
4386 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4387 			return;
4388 		}
4389 		if (val == 0xffff) {
4390 			bp->fw_reset_min_msecs = 0;
4391 			us = 1;
4392 		}
4393 	}
4394 
4395 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4396 	if (rc)
4397 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4398 }
4399 
4400 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4401 {
4402 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4403 	uint32_t reg = info->status_regs[index];
4404 	uint32_t type, offset, val = 0;
4405 	int ret = 0;
4406 
4407 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4408 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4409 
4410 	switch (type) {
4411 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4412 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4413 		if (ret < 0)
4414 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4415 				    offset);
4416 		break;
4417 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4418 		offset = info->mapped_status_regs[index];
4419 		/* FALLTHROUGH */
4420 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4421 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4422 				       offset));
4423 		break;
4424 	}
4425 
4426 	return val;
4427 }
4428 
4429 static int bnxt_fw_reset_all(struct bnxt *bp)
4430 {
4431 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4432 	uint32_t i;
4433 	int rc = 0;
4434 
4435 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4436 		/* Reset through master function driver */
4437 		for (i = 0; i < info->reg_array_cnt; i++)
4438 			bnxt_write_fw_reset_reg(bp, i);
4439 		/* Wait for time specified by FW after triggering reset */
4440 		rte_delay_ms(info->master_func_wait_period_after_reset);
4441 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4442 		/* Reset with the help of Kong processor */
4443 		rc = bnxt_hwrm_fw_reset(bp);
4444 		if (rc)
4445 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4446 	}
4447 
4448 	return rc;
4449 }
4450 
4451 static void bnxt_fw_reset_cb(void *arg)
4452 {
4453 	struct bnxt *bp = arg;
4454 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4455 	int rc = 0;
4456 
4457 	/* Only Master function can do FW reset */
4458 	if (bnxt_is_master_func(bp) &&
4459 	    bnxt_is_recovery_enabled(bp)) {
4460 		rc = bnxt_fw_reset_all(bp);
4461 		if (rc) {
4462 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4463 			return;
4464 		}
4465 	}
4466 
4467 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4468 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4469 	 * (including MASTER FUNC). After receiving this Async, all the active
4470 	 * drivers should treat this case as FW initiated recovery
4471 	 */
4472 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4473 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4474 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4475 
4476 		/* To recover from error */
4477 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4478 				  (void *)bp);
4479 	}
4480 }
4481 
4482 /* Driver should poll FW heartbeat, reset_counter with the frequency
4483  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4484  * When the driver detects heartbeat stop or change in reset_counter,
4485  * it has to trigger a reset to recover from the error condition.
4486  * A “master PF” is the function who will have the privilege to
4487  * initiate the chimp reset. The master PF will be elected by the
4488  * firmware and will be notified through async message.
4489  */
4490 static void bnxt_check_fw_health(void *arg)
4491 {
4492 	struct bnxt *bp = arg;
4493 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4494 	uint32_t val = 0, wait_msec;
4495 
4496 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4497 	    is_bnxt_in_error(bp))
4498 		return;
4499 
4500 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4501 	if (val == info->last_heart_beat)
4502 		goto reset;
4503 
4504 	info->last_heart_beat = val;
4505 
4506 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4507 	if (val != info->last_reset_counter)
4508 		goto reset;
4509 
4510 	info->last_reset_counter = val;
4511 
4512 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4513 			  bnxt_check_fw_health, (void *)bp);
4514 
4515 	return;
4516 reset:
4517 	/* Stop DMA to/from device */
4518 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4519 	bp->flags |= BNXT_FLAG_FW_RESET;
4520 
4521 	bnxt_stop_rxtx(bp);
4522 
4523 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4524 
4525 	if (bnxt_is_master_func(bp))
4526 		wait_msec = info->master_func_wait_period;
4527 	else
4528 		wait_msec = info->normal_func_wait_period;
4529 
4530 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4531 			  bnxt_fw_reset_cb, (void *)bp);
4532 }
4533 
4534 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4535 {
4536 	uint32_t polling_freq;
4537 
4538 	pthread_mutex_lock(&bp->health_check_lock);
4539 
4540 	if (!bnxt_is_recovery_enabled(bp))
4541 		goto done;
4542 
4543 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4544 		goto done;
4545 
4546 	polling_freq = bp->recovery_info->driver_polling_freq;
4547 
4548 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4549 			  bnxt_check_fw_health, (void *)bp);
4550 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4551 
4552 done:
4553 	pthread_mutex_unlock(&bp->health_check_lock);
4554 }
4555 
4556 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4557 {
4558 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4559 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4560 }
4561 
4562 static bool bnxt_vf_pciid(uint16_t device_id)
4563 {
4564 	switch (device_id) {
4565 	case BROADCOM_DEV_ID_57304_VF:
4566 	case BROADCOM_DEV_ID_57406_VF:
4567 	case BROADCOM_DEV_ID_5731X_VF:
4568 	case BROADCOM_DEV_ID_5741X_VF:
4569 	case BROADCOM_DEV_ID_57414_VF:
4570 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4571 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4572 	case BROADCOM_DEV_ID_58802_VF:
4573 	case BROADCOM_DEV_ID_57500_VF1:
4574 	case BROADCOM_DEV_ID_57500_VF2:
4575 	case BROADCOM_DEV_ID_58818_VF:
4576 		/* FALLTHROUGH */
4577 		return true;
4578 	default:
4579 		return false;
4580 	}
4581 }
4582 
4583 /* Phase 5 device */
4584 static bool bnxt_p5_device(uint16_t device_id)
4585 {
4586 	switch (device_id) {
4587 	case BROADCOM_DEV_ID_57508:
4588 	case BROADCOM_DEV_ID_57504:
4589 	case BROADCOM_DEV_ID_57502:
4590 	case BROADCOM_DEV_ID_57508_MF1:
4591 	case BROADCOM_DEV_ID_57504_MF1:
4592 	case BROADCOM_DEV_ID_57502_MF1:
4593 	case BROADCOM_DEV_ID_57508_MF2:
4594 	case BROADCOM_DEV_ID_57504_MF2:
4595 	case BROADCOM_DEV_ID_57502_MF2:
4596 	case BROADCOM_DEV_ID_57500_VF1:
4597 	case BROADCOM_DEV_ID_57500_VF2:
4598 	case BROADCOM_DEV_ID_58812:
4599 	case BROADCOM_DEV_ID_58814:
4600 	case BROADCOM_DEV_ID_58818:
4601 	case BROADCOM_DEV_ID_58818_VF:
4602 		/* FALLTHROUGH */
4603 		return true;
4604 	default:
4605 		return false;
4606 	}
4607 }
4608 
4609 bool bnxt_stratus_device(struct bnxt *bp)
4610 {
4611 	uint16_t device_id = bp->pdev->id.device_id;
4612 
4613 	switch (device_id) {
4614 	case BROADCOM_DEV_ID_STRATUS_NIC:
4615 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4616 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4617 		/* FALLTHROUGH */
4618 		return true;
4619 	default:
4620 		return false;
4621 	}
4622 }
4623 
4624 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4625 {
4626 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4627 	struct bnxt *bp = eth_dev->data->dev_private;
4628 
4629 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4630 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4631 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4632 	if (!bp->bar0 || !bp->doorbell_base) {
4633 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4634 		return -ENODEV;
4635 	}
4636 
4637 	bp->eth_dev = eth_dev;
4638 	bp->pdev = pci_dev;
4639 
4640 	return 0;
4641 }
4642 
4643 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4644 				  struct bnxt_ctx_pg_info *ctx_pg,
4645 				  uint32_t mem_size,
4646 				  const char *suffix,
4647 				  uint16_t idx)
4648 {
4649 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4650 	const struct rte_memzone *mz = NULL;
4651 	char mz_name[RTE_MEMZONE_NAMESIZE];
4652 	rte_iova_t mz_phys_addr;
4653 	uint64_t valid_bits = 0;
4654 	uint32_t sz;
4655 	int i;
4656 
4657 	if (!mem_size)
4658 		return 0;
4659 
4660 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4661 			 BNXT_PAGE_SIZE;
4662 	rmem->page_size = BNXT_PAGE_SIZE;
4663 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4664 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4665 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4666 
4667 	valid_bits = PTU_PTE_VALID;
4668 
4669 	if (rmem->nr_pages > 1) {
4670 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4671 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4672 			 suffix, idx, bp->eth_dev->data->port_id);
4673 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4674 		mz = rte_memzone_lookup(mz_name);
4675 		if (!mz) {
4676 			mz = rte_memzone_reserve_aligned(mz_name,
4677 						rmem->nr_pages * 8,
4678 						bp->eth_dev->device->numa_node,
4679 						RTE_MEMZONE_2MB |
4680 						RTE_MEMZONE_SIZE_HINT_ONLY |
4681 						RTE_MEMZONE_IOVA_CONTIG,
4682 						BNXT_PAGE_SIZE);
4683 			if (mz == NULL)
4684 				return -ENOMEM;
4685 		}
4686 
4687 		memset(mz->addr, 0, mz->len);
4688 		mz_phys_addr = mz->iova;
4689 
4690 		rmem->pg_tbl = mz->addr;
4691 		rmem->pg_tbl_map = mz_phys_addr;
4692 		rmem->pg_tbl_mz = mz;
4693 	}
4694 
4695 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4696 		 suffix, idx, bp->eth_dev->data->port_id);
4697 	mz = rte_memzone_lookup(mz_name);
4698 	if (!mz) {
4699 		mz = rte_memzone_reserve_aligned(mz_name,
4700 						 mem_size,
4701 						 bp->eth_dev->device->numa_node,
4702 						 RTE_MEMZONE_1GB |
4703 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4704 						 RTE_MEMZONE_IOVA_CONTIG,
4705 						 BNXT_PAGE_SIZE);
4706 		if (mz == NULL)
4707 			return -ENOMEM;
4708 	}
4709 
4710 	memset(mz->addr, 0, mz->len);
4711 	mz_phys_addr = mz->iova;
4712 
4713 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4714 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4715 		rmem->dma_arr[i] = mz_phys_addr + sz;
4716 
4717 		if (rmem->nr_pages > 1) {
4718 			if (i == rmem->nr_pages - 2 &&
4719 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4720 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4721 			else if (i == rmem->nr_pages - 1 &&
4722 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4723 				valid_bits |= PTU_PTE_LAST;
4724 
4725 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4726 							   valid_bits);
4727 		}
4728 	}
4729 
4730 	rmem->mz = mz;
4731 	if (rmem->vmem_size)
4732 		rmem->vmem = (void **)mz->addr;
4733 	rmem->dma_arr[0] = mz_phys_addr;
4734 	return 0;
4735 }
4736 
4737 static void bnxt_free_ctx_mem(struct bnxt *bp)
4738 {
4739 	int i;
4740 
4741 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4742 		return;
4743 
4744 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4745 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4746 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4747 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4748 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4749 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4750 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4751 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4752 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4753 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4754 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4755 
4756 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4757 		if (bp->ctx->tqm_mem[i])
4758 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4759 	}
4760 
4761 	rte_free(bp->ctx);
4762 	bp->ctx = NULL;
4763 }
4764 
4765 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4766 
4767 #define min_t(type, x, y) ({                    \
4768 	type __min1 = (x);                      \
4769 	type __min2 = (y);                      \
4770 	__min1 < __min2 ? __min1 : __min2; })
4771 
4772 #define max_t(type, x, y) ({                    \
4773 	type __max1 = (x);                      \
4774 	type __max2 = (y);                      \
4775 	__max1 > __max2 ? __max1 : __max2; })
4776 
4777 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4778 
4779 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4780 {
4781 	struct bnxt_ctx_pg_info *ctx_pg;
4782 	struct bnxt_ctx_mem_info *ctx;
4783 	uint32_t mem_size, ena, entries;
4784 	uint32_t entries_sp, min;
4785 	int i, rc;
4786 
4787 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4788 	if (rc) {
4789 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4790 		return rc;
4791 	}
4792 	ctx = bp->ctx;
4793 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4794 		return 0;
4795 
4796 	ctx_pg = &ctx->qp_mem;
4797 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4798 	if (ctx->qp_entry_size) {
4799 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4800 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4801 		if (rc)
4802 			return rc;
4803 	}
4804 
4805 	ctx_pg = &ctx->srq_mem;
4806 	ctx_pg->entries = ctx->srq_max_l2_entries;
4807 	if (ctx->srq_entry_size) {
4808 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4809 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4810 		if (rc)
4811 			return rc;
4812 	}
4813 
4814 	ctx_pg = &ctx->cq_mem;
4815 	ctx_pg->entries = ctx->cq_max_l2_entries;
4816 	if (ctx->cq_entry_size) {
4817 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4818 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4819 		if (rc)
4820 			return rc;
4821 	}
4822 
4823 	ctx_pg = &ctx->vnic_mem;
4824 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4825 		ctx->vnic_max_ring_table_entries;
4826 	if (ctx->vnic_entry_size) {
4827 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4828 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4829 		if (rc)
4830 			return rc;
4831 	}
4832 
4833 	ctx_pg = &ctx->stat_mem;
4834 	ctx_pg->entries = ctx->stat_max_entries;
4835 	if (ctx->stat_entry_size) {
4836 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4837 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4838 		if (rc)
4839 			return rc;
4840 	}
4841 
4842 	min = ctx->tqm_min_entries_per_ring;
4843 
4844 	entries_sp = ctx->qp_max_l2_entries +
4845 		     ctx->vnic_max_vnic_entries +
4846 		     2 * ctx->qp_min_qp1_entries + min;
4847 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4848 
4849 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4850 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4851 	entries = clamp_t(uint32_t, entries, min,
4852 			  ctx->tqm_max_entries_per_ring);
4853 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4854 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4855 		 * i > 8 is other ext rings.
4856 		 */
4857 		ctx_pg = ctx->tqm_mem[i];
4858 		ctx_pg->entries = i ? entries : entries_sp;
4859 		if (ctx->tqm_entry_size) {
4860 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4861 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4862 						    "tqm_mem", i);
4863 			if (rc)
4864 				return rc;
4865 		}
4866 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4867 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4868 		else
4869 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4870 	}
4871 
4872 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4873 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4874 	if (rc)
4875 		PMD_DRV_LOG(ERR,
4876 			    "Failed to configure context mem: rc = %d\n", rc);
4877 	else
4878 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4879 
4880 	return rc;
4881 }
4882 
4883 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4884 {
4885 	struct rte_pci_device *pci_dev = bp->pdev;
4886 	char mz_name[RTE_MEMZONE_NAMESIZE];
4887 	const struct rte_memzone *mz = NULL;
4888 	uint32_t total_alloc_len;
4889 	rte_iova_t mz_phys_addr;
4890 
4891 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4892 		return 0;
4893 
4894 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4895 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4896 		 pci_dev->addr.bus, pci_dev->addr.devid,
4897 		 pci_dev->addr.function, "rx_port_stats");
4898 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4899 	mz = rte_memzone_lookup(mz_name);
4900 	total_alloc_len =
4901 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4902 				       sizeof(struct rx_port_stats_ext) + 512);
4903 	if (!mz) {
4904 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4905 					 SOCKET_ID_ANY,
4906 					 RTE_MEMZONE_2MB |
4907 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4908 					 RTE_MEMZONE_IOVA_CONTIG);
4909 		if (mz == NULL)
4910 			return -ENOMEM;
4911 	}
4912 	memset(mz->addr, 0, mz->len);
4913 	mz_phys_addr = mz->iova;
4914 
4915 	bp->rx_mem_zone = (const void *)mz;
4916 	bp->hw_rx_port_stats = mz->addr;
4917 	bp->hw_rx_port_stats_map = mz_phys_addr;
4918 
4919 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4920 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4921 		 pci_dev->addr.bus, pci_dev->addr.devid,
4922 		 pci_dev->addr.function, "tx_port_stats");
4923 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4924 	mz = rte_memzone_lookup(mz_name);
4925 	total_alloc_len =
4926 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4927 				       sizeof(struct tx_port_stats_ext) + 512);
4928 	if (!mz) {
4929 		mz = rte_memzone_reserve(mz_name,
4930 					 total_alloc_len,
4931 					 SOCKET_ID_ANY,
4932 					 RTE_MEMZONE_2MB |
4933 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4934 					 RTE_MEMZONE_IOVA_CONTIG);
4935 		if (mz == NULL)
4936 			return -ENOMEM;
4937 	}
4938 	memset(mz->addr, 0, mz->len);
4939 	mz_phys_addr = mz->iova;
4940 
4941 	bp->tx_mem_zone = (const void *)mz;
4942 	bp->hw_tx_port_stats = mz->addr;
4943 	bp->hw_tx_port_stats_map = mz_phys_addr;
4944 	bp->flags |= BNXT_FLAG_PORT_STATS;
4945 
4946 	/* Display extended statistics if FW supports it */
4947 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4948 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4949 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4950 		return 0;
4951 
4952 	bp->hw_rx_port_stats_ext = (void *)
4953 		((uint8_t *)bp->hw_rx_port_stats +
4954 		 sizeof(struct rx_port_stats));
4955 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4956 		sizeof(struct rx_port_stats);
4957 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4958 
4959 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4960 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4961 		bp->hw_tx_port_stats_ext = (void *)
4962 			((uint8_t *)bp->hw_tx_port_stats +
4963 			 sizeof(struct tx_port_stats));
4964 		bp->hw_tx_port_stats_ext_map =
4965 			bp->hw_tx_port_stats_map +
4966 			sizeof(struct tx_port_stats);
4967 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4968 	}
4969 
4970 	return 0;
4971 }
4972 
4973 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4974 {
4975 	struct bnxt *bp = eth_dev->data->dev_private;
4976 	int rc = 0;
4977 
4978 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4979 					       RTE_ETHER_ADDR_LEN *
4980 					       bp->max_l2_ctx,
4981 					       0);
4982 	if (eth_dev->data->mac_addrs == NULL) {
4983 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4984 		return -ENOMEM;
4985 	}
4986 
4987 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4988 		if (BNXT_PF(bp))
4989 			return -EINVAL;
4990 
4991 		/* Generate a random MAC address, if none was assigned by PF */
4992 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4993 		bnxt_eth_hw_addr_random(bp->mac_addr);
4994 		PMD_DRV_LOG(INFO,
4995 			    "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4996 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4997 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4998 
4999 		rc = bnxt_hwrm_set_mac(bp);
5000 		if (rc)
5001 			return rc;
5002 	}
5003 
5004 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5005 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5006 
5007 	return rc;
5008 }
5009 
5010 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5011 {
5012 	int rc = 0;
5013 
5014 	/* MAC is already configured in FW */
5015 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5016 		return 0;
5017 
5018 	/* Restore the old MAC configured */
5019 	rc = bnxt_hwrm_set_mac(bp);
5020 	if (rc)
5021 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5022 
5023 	return rc;
5024 }
5025 
5026 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5027 {
5028 	if (!BNXT_PF(bp))
5029 		return;
5030 
5031 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5032 
5033 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5034 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5035 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5036 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5037 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5038 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5039 }
5040 
5041 struct bnxt *
5042 bnxt_get_bp(uint16_t port)
5043 {
5044 	struct bnxt *bp;
5045 	struct rte_eth_dev *dev;
5046 
5047 	if (!rte_eth_dev_is_valid_port(port)) {
5048 		PMD_DRV_LOG(ERR, "Invalid port %d\n", port);
5049 		return NULL;
5050 	}
5051 
5052 	dev = &rte_eth_devices[port];
5053 	if (!is_bnxt_supported(dev)) {
5054 		PMD_DRV_LOG(ERR, "Device %d not supported\n", port);
5055 		return NULL;
5056 	}
5057 
5058 	bp = (struct bnxt *)dev->data->dev_private;
5059 	if (!BNXT_TRUFLOW_EN(bp)) {
5060 		PMD_DRV_LOG(ERR, "TRUFLOW not enabled\n");
5061 		return NULL;
5062 	}
5063 
5064 	return bp;
5065 }
5066 
5067 uint16_t
5068 bnxt_get_svif(uint16_t port_id, bool func_svif,
5069 	      enum bnxt_ulp_intf_type type)
5070 {
5071 	struct rte_eth_dev *eth_dev;
5072 	struct bnxt *bp;
5073 
5074 	eth_dev = &rte_eth_devices[port_id];
5075 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5076 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5077 		if (!vfr)
5078 			return 0;
5079 
5080 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5081 			return vfr->svif;
5082 
5083 		eth_dev = vfr->parent_dev;
5084 	}
5085 
5086 	bp = eth_dev->data->dev_private;
5087 
5088 	return func_svif ? bp->func_svif : bp->port_svif;
5089 }
5090 
5091 void
5092 bnxt_get_iface_mac(uint16_t port, enum bnxt_ulp_intf_type type,
5093 		   uint8_t *mac, uint8_t *parent_mac)
5094 {
5095 	struct rte_eth_dev *eth_dev;
5096 	struct bnxt *bp;
5097 
5098 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF &&
5099 	    type != BNXT_ULP_INTF_TYPE_PF)
5100 		return;
5101 
5102 	eth_dev = &rte_eth_devices[port];
5103 	bp = eth_dev->data->dev_private;
5104 	memcpy(mac, bp->mac_addr, RTE_ETHER_ADDR_LEN);
5105 
5106 	if (type == BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5107 		memcpy(parent_mac, bp->parent->mac_addr, RTE_ETHER_ADDR_LEN);
5108 }
5109 
5110 uint16_t
5111 bnxt_get_parent_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5112 {
5113 	struct rte_eth_dev *eth_dev;
5114 	struct bnxt *bp;
5115 
5116 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5117 		return 0;
5118 
5119 	eth_dev = &rte_eth_devices[port];
5120 	bp = eth_dev->data->dev_private;
5121 
5122 	return bp->parent->vnic;
5123 }
5124 uint16_t
5125 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5126 {
5127 	struct rte_eth_dev *eth_dev;
5128 	struct bnxt_vnic_info *vnic;
5129 	struct bnxt *bp;
5130 
5131 	eth_dev = &rte_eth_devices[port];
5132 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5133 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5134 		if (!vfr)
5135 			return 0;
5136 
5137 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5138 			return vfr->dflt_vnic_id;
5139 
5140 		eth_dev = vfr->parent_dev;
5141 	}
5142 
5143 	bp = eth_dev->data->dev_private;
5144 
5145 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
5146 
5147 	return vnic->fw_vnic_id;
5148 }
5149 
5150 uint16_t
5151 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
5152 {
5153 	struct rte_eth_dev *eth_dev;
5154 	struct bnxt *bp;
5155 
5156 	eth_dev = &rte_eth_devices[port];
5157 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5158 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5159 		if (!vfr)
5160 			return 0;
5161 
5162 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5163 			return vfr->fw_fid;
5164 
5165 		eth_dev = vfr->parent_dev;
5166 	}
5167 
5168 	bp = eth_dev->data->dev_private;
5169 
5170 	return bp->fw_fid;
5171 }
5172 
5173 enum bnxt_ulp_intf_type
5174 bnxt_get_interface_type(uint16_t port)
5175 {
5176 	struct rte_eth_dev *eth_dev;
5177 	struct bnxt *bp;
5178 
5179 	eth_dev = &rte_eth_devices[port];
5180 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
5181 		return BNXT_ULP_INTF_TYPE_VF_REP;
5182 
5183 	bp = eth_dev->data->dev_private;
5184 	if (BNXT_PF(bp))
5185 		return BNXT_ULP_INTF_TYPE_PF;
5186 	else if (BNXT_VF_IS_TRUSTED(bp))
5187 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
5188 	else if (BNXT_VF(bp))
5189 		return BNXT_ULP_INTF_TYPE_VF;
5190 
5191 	return BNXT_ULP_INTF_TYPE_INVALID;
5192 }
5193 
5194 uint16_t
5195 bnxt_get_phy_port_id(uint16_t port_id)
5196 {
5197 	struct bnxt_representor *vfr;
5198 	struct rte_eth_dev *eth_dev;
5199 	struct bnxt *bp;
5200 
5201 	eth_dev = &rte_eth_devices[port_id];
5202 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5203 		vfr = eth_dev->data->dev_private;
5204 		if (!vfr)
5205 			return 0;
5206 
5207 		eth_dev = vfr->parent_dev;
5208 	}
5209 
5210 	bp = eth_dev->data->dev_private;
5211 
5212 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
5213 }
5214 
5215 uint16_t
5216 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
5217 {
5218 	struct rte_eth_dev *eth_dev;
5219 	struct bnxt *bp;
5220 
5221 	eth_dev = &rte_eth_devices[port_id];
5222 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5223 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5224 		if (!vfr)
5225 			return 0;
5226 
5227 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5228 			return vfr->fw_fid - 1;
5229 
5230 		eth_dev = vfr->parent_dev;
5231 	}
5232 
5233 	bp = eth_dev->data->dev_private;
5234 
5235 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
5236 }
5237 
5238 uint16_t
5239 bnxt_get_vport(uint16_t port_id)
5240 {
5241 	return (1 << bnxt_get_phy_port_id(port_id));
5242 }
5243 
5244 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5245 {
5246 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5247 
5248 	if (info) {
5249 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5250 			memset(info, 0, sizeof(*info));
5251 		return;
5252 	}
5253 
5254 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5255 		return;
5256 
5257 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5258 			   sizeof(*info), 0);
5259 	if (!info)
5260 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5261 
5262 	bp->recovery_info = info;
5263 }
5264 
5265 static void bnxt_check_fw_status(struct bnxt *bp)
5266 {
5267 	uint32_t fw_status;
5268 
5269 	if (!(bp->recovery_info &&
5270 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5271 		return;
5272 
5273 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5274 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5275 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5276 			    fw_status);
5277 }
5278 
5279 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5280 {
5281 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5282 	uint32_t status_loc;
5283 	uint32_t sig_ver;
5284 
5285 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5286 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5287 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5288 				   BNXT_GRCP_WINDOW_2_BASE +
5289 				   offsetof(struct hcomm_status,
5290 					    sig_ver)));
5291 	/* If the signature is absent, then FW does not support this feature */
5292 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5293 	    HCOMM_STATUS_SIGNATURE_VAL)
5294 		return 0;
5295 
5296 	if (!info) {
5297 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5298 				   sizeof(*info), 0);
5299 		if (!info)
5300 			return -ENOMEM;
5301 		bp->recovery_info = info;
5302 	} else {
5303 		memset(info, 0, sizeof(*info));
5304 	}
5305 
5306 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5307 				      BNXT_GRCP_WINDOW_2_BASE +
5308 				      offsetof(struct hcomm_status,
5309 					       fw_status_loc)));
5310 
5311 	/* Only pre-map the FW health status GRC register */
5312 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5313 		return 0;
5314 
5315 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5316 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5317 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5318 
5319 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5320 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5321 
5322 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5323 
5324 	return 0;
5325 }
5326 
5327 /* This function gets the FW version along with the
5328  * capabilities(MAX and current) of the function, vnic,
5329  * error recovery, phy and other chip related info
5330  */
5331 static int bnxt_get_config(struct bnxt *bp)
5332 {
5333 	uint16_t mtu;
5334 	int rc = 0;
5335 
5336 	bp->fw_cap = 0;
5337 
5338 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5339 	if (rc)
5340 		return rc;
5341 
5342 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5343 	if (rc) {
5344 		bnxt_check_fw_status(bp);
5345 		return rc;
5346 	}
5347 
5348 	rc = bnxt_hwrm_func_reset(bp);
5349 	if (rc)
5350 		return -EIO;
5351 
5352 	rc = bnxt_hwrm_vnic_qcaps(bp);
5353 	if (rc)
5354 		return rc;
5355 
5356 	rc = bnxt_hwrm_queue_qportcfg(bp);
5357 	if (rc)
5358 		return rc;
5359 
5360 	/* Get the MAX capabilities for this function.
5361 	 * This function also allocates context memory for TQM rings and
5362 	 * informs the firmware about this allocated backing store memory.
5363 	 */
5364 	rc = bnxt_hwrm_func_qcaps(bp);
5365 	if (rc)
5366 		return rc;
5367 
5368 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5369 	if (rc)
5370 		return rc;
5371 
5372 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5373 	if (rc)
5374 		return rc;
5375 
5376 	bnxt_hwrm_port_mac_qcfg(bp);
5377 
5378 	bnxt_hwrm_parent_pf_qcfg(bp);
5379 
5380 	bnxt_hwrm_port_phy_qcaps(bp);
5381 
5382 	bnxt_alloc_error_recovery_info(bp);
5383 	/* Get the adapter error recovery support info */
5384 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5385 	if (rc)
5386 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5387 
5388 	bnxt_hwrm_port_led_qcaps(bp);
5389 
5390 	return 0;
5391 }
5392 
5393 static int
5394 bnxt_init_locks(struct bnxt *bp)
5395 {
5396 	int err;
5397 
5398 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5399 	if (err) {
5400 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5401 		return err;
5402 	}
5403 
5404 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5405 	if (err) {
5406 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5407 		return err;
5408 	}
5409 
5410 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5411 	if (err) {
5412 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5413 		return err;
5414 	}
5415 
5416 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5417 	if (err)
5418 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5419 
5420 	return err;
5421 }
5422 
5423 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5424 {
5425 	int rc = 0;
5426 
5427 	rc = bnxt_get_config(bp);
5428 	if (rc)
5429 		return rc;
5430 
5431 	if (!reconfig_dev) {
5432 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5433 		if (rc)
5434 			return rc;
5435 	} else {
5436 		rc = bnxt_restore_dflt_mac(bp);
5437 		if (rc)
5438 			return rc;
5439 	}
5440 
5441 	bnxt_config_vf_req_fwd(bp);
5442 
5443 	rc = bnxt_hwrm_func_driver_register(bp);
5444 	if (rc) {
5445 		PMD_DRV_LOG(ERR, "Failed to register driver");
5446 		return -EBUSY;
5447 	}
5448 
5449 	if (BNXT_PF(bp)) {
5450 		if (bp->pdev->max_vfs) {
5451 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5452 			if (rc) {
5453 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5454 				return rc;
5455 			}
5456 		} else {
5457 			rc = bnxt_hwrm_allocate_pf_only(bp);
5458 			if (rc) {
5459 				PMD_DRV_LOG(ERR,
5460 					    "Failed to allocate PF resources");
5461 				return rc;
5462 			}
5463 		}
5464 	}
5465 
5466 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5467 	if (rc)
5468 		return rc;
5469 
5470 	rc = bnxt_setup_int(bp);
5471 	if (rc)
5472 		return rc;
5473 
5474 	rc = bnxt_request_int(bp);
5475 	if (rc)
5476 		return rc;
5477 
5478 	rc = bnxt_init_ctx_mem(bp);
5479 	if (rc) {
5480 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5481 		return rc;
5482 	}
5483 
5484 	return 0;
5485 }
5486 
5487 static int
5488 bnxt_parse_devarg_accum_stats(__rte_unused const char *key,
5489 			      const char *value, void *opaque_arg)
5490 {
5491 	struct bnxt *bp = opaque_arg;
5492 	unsigned long accum_stats;
5493 	char *end = NULL;
5494 
5495 	if (!value || !opaque_arg) {
5496 		PMD_DRV_LOG(ERR,
5497 			    "Invalid parameter passed to accum-stats devargs.\n");
5498 		return -EINVAL;
5499 	}
5500 
5501 	accum_stats = strtoul(value, &end, 10);
5502 	if (end == NULL || *end != '\0' ||
5503 	    (accum_stats == ULONG_MAX && errno == ERANGE)) {
5504 		PMD_DRV_LOG(ERR,
5505 			    "Invalid parameter passed to accum-stats devargs.\n");
5506 		return -EINVAL;
5507 	}
5508 
5509 	if (BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)) {
5510 		PMD_DRV_LOG(ERR,
5511 			    "Invalid value passed to accum-stats devargs.\n");
5512 		return -EINVAL;
5513 	}
5514 
5515 	if (accum_stats) {
5516 		bp->flags2 |= BNXT_FLAGS2_ACCUM_STATS_EN;
5517 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature enabled.\n");
5518 	} else {
5519 		bp->flags2 &= ~BNXT_FLAGS2_ACCUM_STATS_EN;
5520 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature disabled.\n");
5521 	}
5522 
5523 	return 0;
5524 }
5525 
5526 static int
5527 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5528 			     const char *value, void *opaque_arg)
5529 {
5530 	struct bnxt *bp = opaque_arg;
5531 	unsigned long flow_xstat;
5532 	char *end = NULL;
5533 
5534 	if (!value || !opaque_arg) {
5535 		PMD_DRV_LOG(ERR,
5536 			    "Invalid parameter passed to flow_xstat devarg.\n");
5537 		return -EINVAL;
5538 	}
5539 
5540 	flow_xstat = strtoul(value, &end, 10);
5541 	if (end == NULL || *end != '\0' ||
5542 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5543 		PMD_DRV_LOG(ERR,
5544 			    "Invalid parameter passed to flow_xstat devarg.\n");
5545 		return -EINVAL;
5546 	}
5547 
5548 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5549 		PMD_DRV_LOG(ERR,
5550 			    "Invalid value passed to flow_xstat devarg.\n");
5551 		return -EINVAL;
5552 	}
5553 
5554 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5555 	if (BNXT_FLOW_XSTATS_EN(bp))
5556 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5557 
5558 	return 0;
5559 }
5560 
5561 static int
5562 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5563 					const char *value, void *opaque_arg)
5564 {
5565 	struct bnxt *bp = opaque_arg;
5566 	unsigned long max_num_kflows;
5567 	char *end = NULL;
5568 
5569 	if (!value || !opaque_arg) {
5570 		PMD_DRV_LOG(ERR,
5571 			"Invalid parameter passed to max_num_kflows devarg.\n");
5572 		return -EINVAL;
5573 	}
5574 
5575 	max_num_kflows = strtoul(value, &end, 10);
5576 	if (end == NULL || *end != '\0' ||
5577 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5578 		PMD_DRV_LOG(ERR,
5579 			"Invalid parameter passed to max_num_kflows devarg.\n");
5580 		return -EINVAL;
5581 	}
5582 
5583 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5584 		PMD_DRV_LOG(ERR,
5585 			"Invalid value passed to max_num_kflows devarg.\n");
5586 		return -EINVAL;
5587 	}
5588 
5589 	bp->max_num_kflows = max_num_kflows;
5590 	if (bp->max_num_kflows)
5591 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5592 				max_num_kflows);
5593 
5594 	return 0;
5595 }
5596 
5597 static int
5598 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5599 				 const char *value, void *opaque_arg)
5600 {
5601 	struct bnxt *bp = opaque_arg;
5602 	unsigned long app_id;
5603 	char *end = NULL;
5604 
5605 	if (!value || !opaque_arg) {
5606 		PMD_DRV_LOG(ERR,
5607 			    "Invalid parameter passed to app-id "
5608 			    "devargs.\n");
5609 		return -EINVAL;
5610 	}
5611 
5612 	app_id = strtoul(value, &end, 10);
5613 	if (end == NULL || *end != '\0' ||
5614 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5615 		PMD_DRV_LOG(ERR,
5616 			    "Invalid parameter passed to app_id "
5617 			    "devargs.\n");
5618 		return -EINVAL;
5619 	}
5620 
5621 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5622 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5623 			    (uint16_t)app_id);
5624 		return -EINVAL;
5625 	}
5626 
5627 	bp->app_id = app_id;
5628 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5629 
5630 	return 0;
5631 }
5632 
5633 static int
5634 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5635 			    const char *value, void *opaque_arg)
5636 {
5637 	struct bnxt_representor *vfr_bp = opaque_arg;
5638 	unsigned long rep_is_pf;
5639 	char *end = NULL;
5640 
5641 	if (!value || !opaque_arg) {
5642 		PMD_DRV_LOG(ERR,
5643 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5644 		return -EINVAL;
5645 	}
5646 
5647 	rep_is_pf = strtoul(value, &end, 10);
5648 	if (end == NULL || *end != '\0' ||
5649 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5650 		PMD_DRV_LOG(ERR,
5651 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5652 		return -EINVAL;
5653 	}
5654 
5655 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5656 		PMD_DRV_LOG(ERR,
5657 			    "Invalid value passed to rep_is_pf devargs.\n");
5658 		return -EINVAL;
5659 	}
5660 
5661 	vfr_bp->flags |= rep_is_pf;
5662 	if (BNXT_REP_PF(vfr_bp))
5663 		PMD_DRV_LOG(INFO, "PF representor\n");
5664 	else
5665 		PMD_DRV_LOG(INFO, "VF representor\n");
5666 
5667 	return 0;
5668 }
5669 
5670 static int
5671 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5672 			       const char *value, void *opaque_arg)
5673 {
5674 	struct bnxt_representor *vfr_bp = opaque_arg;
5675 	unsigned long rep_based_pf;
5676 	char *end = NULL;
5677 
5678 	if (!value || !opaque_arg) {
5679 		PMD_DRV_LOG(ERR,
5680 			    "Invalid parameter passed to rep_based_pf "
5681 			    "devargs.\n");
5682 		return -EINVAL;
5683 	}
5684 
5685 	rep_based_pf = strtoul(value, &end, 10);
5686 	if (end == NULL || *end != '\0' ||
5687 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5688 		PMD_DRV_LOG(ERR,
5689 			    "Invalid parameter passed to rep_based_pf "
5690 			    "devargs.\n");
5691 		return -EINVAL;
5692 	}
5693 
5694 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5695 		PMD_DRV_LOG(ERR,
5696 			    "Invalid value passed to rep_based_pf devargs.\n");
5697 		return -EINVAL;
5698 	}
5699 
5700 	vfr_bp->rep_based_pf = rep_based_pf;
5701 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5702 
5703 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5704 
5705 	return 0;
5706 }
5707 
5708 static int
5709 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5710 			    const char *value, void *opaque_arg)
5711 {
5712 	struct bnxt_representor *vfr_bp = opaque_arg;
5713 	unsigned long rep_q_r2f;
5714 	char *end = NULL;
5715 
5716 	if (!value || !opaque_arg) {
5717 		PMD_DRV_LOG(ERR,
5718 			    "Invalid parameter passed to rep_q_r2f "
5719 			    "devargs.\n");
5720 		return -EINVAL;
5721 	}
5722 
5723 	rep_q_r2f = strtoul(value, &end, 10);
5724 	if (end == NULL || *end != '\0' ||
5725 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5726 		PMD_DRV_LOG(ERR,
5727 			    "Invalid parameter passed to rep_q_r2f "
5728 			    "devargs.\n");
5729 		return -EINVAL;
5730 	}
5731 
5732 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5733 		PMD_DRV_LOG(ERR,
5734 			    "Invalid value passed to rep_q_r2f devargs.\n");
5735 		return -EINVAL;
5736 	}
5737 
5738 	vfr_bp->rep_q_r2f = rep_q_r2f;
5739 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5740 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5741 
5742 	return 0;
5743 }
5744 
5745 static int
5746 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5747 			    const char *value, void *opaque_arg)
5748 {
5749 	struct bnxt_representor *vfr_bp = opaque_arg;
5750 	unsigned long rep_q_f2r;
5751 	char *end = NULL;
5752 
5753 	if (!value || !opaque_arg) {
5754 		PMD_DRV_LOG(ERR,
5755 			    "Invalid parameter passed to rep_q_f2r "
5756 			    "devargs.\n");
5757 		return -EINVAL;
5758 	}
5759 
5760 	rep_q_f2r = strtoul(value, &end, 10);
5761 	if (end == NULL || *end != '\0' ||
5762 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5763 		PMD_DRV_LOG(ERR,
5764 			    "Invalid parameter passed to rep_q_f2r "
5765 			    "devargs.\n");
5766 		return -EINVAL;
5767 	}
5768 
5769 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5770 		PMD_DRV_LOG(ERR,
5771 			    "Invalid value passed to rep_q_f2r devargs.\n");
5772 		return -EINVAL;
5773 	}
5774 
5775 	vfr_bp->rep_q_f2r = rep_q_f2r;
5776 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5777 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5778 
5779 	return 0;
5780 }
5781 
5782 static int
5783 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5784 			     const char *value, void *opaque_arg)
5785 {
5786 	struct bnxt_representor *vfr_bp = opaque_arg;
5787 	unsigned long rep_fc_r2f;
5788 	char *end = NULL;
5789 
5790 	if (!value || !opaque_arg) {
5791 		PMD_DRV_LOG(ERR,
5792 			    "Invalid parameter passed to rep_fc_r2f "
5793 			    "devargs.\n");
5794 		return -EINVAL;
5795 	}
5796 
5797 	rep_fc_r2f = strtoul(value, &end, 10);
5798 	if (end == NULL || *end != '\0' ||
5799 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5800 		PMD_DRV_LOG(ERR,
5801 			    "Invalid parameter passed to rep_fc_r2f "
5802 			    "devargs.\n");
5803 		return -EINVAL;
5804 	}
5805 
5806 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5807 		PMD_DRV_LOG(ERR,
5808 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5809 		return -EINVAL;
5810 	}
5811 
5812 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5813 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5814 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5815 
5816 	return 0;
5817 }
5818 
5819 static int
5820 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5821 			     const char *value, void *opaque_arg)
5822 {
5823 	struct bnxt_representor *vfr_bp = opaque_arg;
5824 	unsigned long rep_fc_f2r;
5825 	char *end = NULL;
5826 
5827 	if (!value || !opaque_arg) {
5828 		PMD_DRV_LOG(ERR,
5829 			    "Invalid parameter passed to rep_fc_f2r "
5830 			    "devargs.\n");
5831 		return -EINVAL;
5832 	}
5833 
5834 	rep_fc_f2r = strtoul(value, &end, 10);
5835 	if (end == NULL || *end != '\0' ||
5836 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5837 		PMD_DRV_LOG(ERR,
5838 			    "Invalid parameter passed to rep_fc_f2r "
5839 			    "devargs.\n");
5840 		return -EINVAL;
5841 	}
5842 
5843 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5844 		PMD_DRV_LOG(ERR,
5845 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5846 		return -EINVAL;
5847 	}
5848 
5849 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5850 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5851 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5852 
5853 	return 0;
5854 }
5855 
5856 static int
5857 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5858 {
5859 	struct rte_kvargs *kvlist;
5860 	int ret;
5861 
5862 	if (devargs == NULL)
5863 		return 0;
5864 
5865 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5866 	if (kvlist == NULL)
5867 		return -EINVAL;
5868 
5869 	/*
5870 	 * Handler for "flow_xstat" devarg.
5871 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5872 	 */
5873 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5874 				 bnxt_parse_devarg_flow_xstat, bp);
5875 	if (ret)
5876 		goto err;
5877 
5878 	/*
5879 	 * Handler for "accum-stats" devarg.
5880 	 * Invoked as for ex: "-a 0000:00:0d.0,accum-stats=1"
5881 	 */
5882 	rte_kvargs_process(kvlist, BNXT_DEVARG_ACCUM_STATS,
5883 			   bnxt_parse_devarg_accum_stats, bp);
5884 	/*
5885 	 * Handler for "max_num_kflows" devarg.
5886 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5887 	 */
5888 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5889 				 bnxt_parse_devarg_max_num_kflows, bp);
5890 	if (ret)
5891 		goto err;
5892 
5893 err:
5894 	/*
5895 	 * Handler for "app-id" devarg.
5896 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5897 	 */
5898 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5899 			   bnxt_parse_devarg_app_id, bp);
5900 
5901 	rte_kvargs_free(kvlist);
5902 	return ret;
5903 }
5904 
5905 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5906 {
5907 	int rc = 0;
5908 
5909 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5910 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5911 		if (rc)
5912 			PMD_DRV_LOG(ERR,
5913 				    "Failed to alloc switch domain: %d\n", rc);
5914 		else
5915 			PMD_DRV_LOG(INFO,
5916 				    "Switch domain allocated %d\n",
5917 				    bp->switch_domain_id);
5918 	}
5919 
5920 	return rc;
5921 }
5922 
5923 /* Allocate and initialize various fields in bnxt struct that
5924  * need to be allocated/destroyed only once in the lifetime of the driver
5925  */
5926 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5927 {
5928 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5929 	struct bnxt *bp = eth_dev->data->dev_private;
5930 	int rc = 0;
5931 
5932 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5933 
5934 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5935 		bp->flags |= BNXT_FLAG_VF;
5936 
5937 	if (bnxt_p5_device(pci_dev->id.device_id))
5938 		bp->flags |= BNXT_FLAG_CHIP_P5;
5939 
5940 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5941 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5942 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5943 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5944 		bp->flags |= BNXT_FLAG_STINGRAY;
5945 
5946 	if (BNXT_TRUFLOW_EN(bp)) {
5947 		/* extra mbuf field is required to store CFA code from mark */
5948 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5949 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5950 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5951 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5952 		};
5953 		bnxt_cfa_code_dynfield_offset =
5954 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5955 		if (bnxt_cfa_code_dynfield_offset < 0) {
5956 			PMD_DRV_LOG(ERR,
5957 			    "Failed to register mbuf field for TruFlow mark\n");
5958 			return -rte_errno;
5959 		}
5960 	}
5961 
5962 	rc = bnxt_map_pci_bars(eth_dev);
5963 	if (rc) {
5964 		PMD_DRV_LOG(ERR,
5965 			    "Failed to initialize board rc: %x\n", rc);
5966 		return rc;
5967 	}
5968 
5969 	rc = bnxt_alloc_pf_info(bp);
5970 	if (rc)
5971 		return rc;
5972 
5973 	rc = bnxt_alloc_link_info(bp);
5974 	if (rc)
5975 		return rc;
5976 
5977 	rc = bnxt_alloc_parent_info(bp);
5978 	if (rc)
5979 		return rc;
5980 
5981 	rc = bnxt_alloc_hwrm_resources(bp);
5982 	if (rc) {
5983 		PMD_DRV_LOG(ERR,
5984 			    "Failed to allocate response buffer rc: %x\n", rc);
5985 		return rc;
5986 	}
5987 	rc = bnxt_alloc_leds_info(bp);
5988 	if (rc)
5989 		return rc;
5990 
5991 	rc = bnxt_alloc_cos_queues(bp);
5992 	if (rc)
5993 		return rc;
5994 
5995 	rc = bnxt_init_locks(bp);
5996 	if (rc)
5997 		return rc;
5998 
5999 	rc = bnxt_alloc_switch_domain(bp);
6000 	if (rc)
6001 		return rc;
6002 
6003 	return rc;
6004 }
6005 
6006 static int
6007 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
6008 {
6009 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
6010 	static int version_printed;
6011 	struct bnxt *bp;
6012 	int rc;
6013 
6014 	if (version_printed++ == 0)
6015 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
6016 
6017 	eth_dev->dev_ops = &bnxt_dev_ops;
6018 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
6019 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
6020 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
6021 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
6022 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
6023 
6024 	/*
6025 	 * For secondary processes, we don't initialise any further
6026 	 * as primary has already done this work.
6027 	 */
6028 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6029 		return 0;
6030 
6031 	rte_eth_copy_pci_info(eth_dev, pci_dev);
6032 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
6033 
6034 	bp = eth_dev->data->dev_private;
6035 
6036 	/* Parse dev arguments passed on when starting the DPDK application. */
6037 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
6038 	if (rc)
6039 		goto error_free;
6040 
6041 	rc = bnxt_drv_init(eth_dev);
6042 	if (rc)
6043 		goto error_free;
6044 
6045 	rc = bnxt_init_resources(bp, false);
6046 	if (rc)
6047 		goto error_free;
6048 
6049 	rc = bnxt_alloc_stats_mem(bp);
6050 	if (rc)
6051 		goto error_free;
6052 
6053 	PMD_DRV_LOG(INFO,
6054 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
6055 		    DRV_MODULE_NAME,
6056 		    pci_dev->mem_resource[0].phys_addr,
6057 		    pci_dev->mem_resource[0].addr);
6058 
6059 	return 0;
6060 
6061 error_free:
6062 	bnxt_dev_uninit(eth_dev);
6063 	return rc;
6064 }
6065 
6066 
6067 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
6068 {
6069 	if (!ctx)
6070 		return;
6071 
6072 	if (ctx->va)
6073 		rte_free(ctx->va);
6074 
6075 	ctx->va = NULL;
6076 	ctx->dma = RTE_BAD_IOVA;
6077 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
6078 }
6079 
6080 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
6081 {
6082 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
6083 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6084 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
6085 				  bp->flow_stat->max_fc,
6086 				  false);
6087 
6088 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
6089 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6090 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
6091 				  bp->flow_stat->max_fc,
6092 				  false);
6093 
6094 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6095 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
6096 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6097 
6098 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6099 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
6100 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6101 
6102 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6103 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
6104 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6105 
6106 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6107 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
6108 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6109 }
6110 
6111 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
6112 {
6113 	bnxt_unregister_fc_ctx_mem(bp);
6114 
6115 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
6116 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
6117 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
6118 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
6119 }
6120 
6121 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
6122 {
6123 	if (BNXT_FLOW_XSTATS_EN(bp))
6124 		bnxt_uninit_fc_ctx_mem(bp);
6125 }
6126 
6127 static void
6128 bnxt_free_error_recovery_info(struct bnxt *bp)
6129 {
6130 	rte_free(bp->recovery_info);
6131 	bp->recovery_info = NULL;
6132 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
6133 }
6134 
6135 static int
6136 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
6137 {
6138 	int rc;
6139 
6140 	bnxt_free_int(bp);
6141 	bnxt_free_mem(bp, reconfig_dev);
6142 
6143 	bnxt_hwrm_func_buf_unrgtr(bp);
6144 	if (bp->pf != NULL) {
6145 		rte_free(bp->pf->vf_req_buf);
6146 		bp->pf->vf_req_buf = NULL;
6147 	}
6148 
6149 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
6150 	bp->flags &= ~BNXT_FLAG_REGISTERED;
6151 	bnxt_free_ctx_mem(bp);
6152 	if (!reconfig_dev) {
6153 		bnxt_free_hwrm_resources(bp);
6154 		bnxt_free_error_recovery_info(bp);
6155 	}
6156 
6157 	bnxt_uninit_ctx_mem(bp);
6158 
6159 	bnxt_free_flow_stats_info(bp);
6160 	if (bp->rep_info != NULL)
6161 		bnxt_free_switch_domain(bp);
6162 	bnxt_free_rep_info(bp);
6163 	rte_free(bp->ptp_cfg);
6164 	bp->ptp_cfg = NULL;
6165 	return rc;
6166 }
6167 
6168 static int
6169 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
6170 {
6171 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6172 		return -EPERM;
6173 
6174 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
6175 
6176 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
6177 		bnxt_dev_close_op(eth_dev);
6178 
6179 	return 0;
6180 }
6181 
6182 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6183 {
6184 	struct bnxt *bp = eth_dev->data->dev_private;
6185 	struct rte_eth_dev *vf_rep_eth_dev;
6186 	int ret = 0, i;
6187 
6188 	if (!bp)
6189 		return -EINVAL;
6190 
6191 	for (i = 0; i < bp->num_reps; i++) {
6192 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6193 		if (!vf_rep_eth_dev)
6194 			continue;
6195 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6196 			    vf_rep_eth_dev->data->port_id);
6197 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6198 	}
6199 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6200 		    eth_dev->data->port_id);
6201 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6202 
6203 	return ret;
6204 }
6205 
6206 static void bnxt_free_rep_info(struct bnxt *bp)
6207 {
6208 	rte_free(bp->rep_info);
6209 	bp->rep_info = NULL;
6210 	rte_free(bp->cfa_code_map);
6211 	bp->cfa_code_map = NULL;
6212 }
6213 
6214 static int bnxt_init_rep_info(struct bnxt *bp)
6215 {
6216 	int i = 0, rc;
6217 
6218 	if (bp->rep_info)
6219 		return 0;
6220 
6221 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
6222 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
6223 				   0);
6224 	if (!bp->rep_info) {
6225 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6226 		return -ENOMEM;
6227 	}
6228 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6229 				       sizeof(*bp->cfa_code_map) *
6230 				       BNXT_MAX_CFA_CODE, 0);
6231 	if (!bp->cfa_code_map) {
6232 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6233 		bnxt_free_rep_info(bp);
6234 		return -ENOMEM;
6235 	}
6236 
6237 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6238 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6239 
6240 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6241 	if (rc) {
6242 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6243 		bnxt_free_rep_info(bp);
6244 		return rc;
6245 	}
6246 
6247 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6248 	if (rc) {
6249 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6250 		bnxt_free_rep_info(bp);
6251 		return rc;
6252 	}
6253 
6254 	return rc;
6255 }
6256 
6257 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6258 			       struct rte_eth_devargs *eth_da,
6259 			       struct rte_eth_dev *backing_eth_dev,
6260 			       const char *dev_args)
6261 {
6262 	struct rte_eth_dev *vf_rep_eth_dev;
6263 	char name[RTE_ETH_NAME_MAX_LEN];
6264 	struct bnxt *backing_bp;
6265 	uint16_t num_rep;
6266 	int i, ret = 0;
6267 	struct rte_kvargs *kvlist = NULL;
6268 
6269 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6270 		return 0;
6271 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6272 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6273 			    eth_da->type);
6274 		return -ENOTSUP;
6275 	}
6276 	num_rep = eth_da->nb_representor_ports;
6277 	if (num_rep > BNXT_MAX_VF_REPS) {
6278 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6279 			    num_rep, BNXT_MAX_VF_REPS);
6280 		return -EINVAL;
6281 	}
6282 
6283 	if (num_rep >= RTE_MAX_ETHPORTS) {
6284 		PMD_DRV_LOG(ERR,
6285 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6286 			    num_rep, RTE_MAX_ETHPORTS);
6287 		return -EINVAL;
6288 	}
6289 
6290 	backing_bp = backing_eth_dev->data->dev_private;
6291 
6292 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6293 		PMD_DRV_LOG(ERR,
6294 			    "Not a PF or trusted VF. No Representor support\n");
6295 		/* Returning an error is not an option.
6296 		 * Applications are not handling this correctly
6297 		 */
6298 		return 0;
6299 	}
6300 
6301 	if (bnxt_init_rep_info(backing_bp))
6302 		return 0;
6303 
6304 	for (i = 0; i < num_rep; i++) {
6305 		struct bnxt_representor representor = {
6306 			.vf_id = eth_da->representor_ports[i],
6307 			.switch_domain_id = backing_bp->switch_domain_id,
6308 			.parent_dev = backing_eth_dev
6309 		};
6310 
6311 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
6312 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6313 				    representor.vf_id, BNXT_MAX_VF_REPS);
6314 			continue;
6315 		}
6316 
6317 		/* representor port net_bdf_port */
6318 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6319 			 pci_dev->device.name, eth_da->representor_ports[i]);
6320 
6321 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6322 		if (kvlist) {
6323 			/*
6324 			 * Handler for "rep_is_pf" devarg.
6325 			 * Invoked as for ex: "-a 000:00:0d.0,
6326 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6327 			 */
6328 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6329 						 bnxt_parse_devarg_rep_is_pf,
6330 						 (void *)&representor);
6331 			if (ret) {
6332 				ret = -EINVAL;
6333 				goto err;
6334 			}
6335 			/*
6336 			 * Handler for "rep_based_pf" devarg.
6337 			 * Invoked as for ex: "-a 000:00:0d.0,
6338 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6339 			 */
6340 			ret = rte_kvargs_process(kvlist,
6341 						 BNXT_DEVARG_REP_BASED_PF,
6342 						 bnxt_parse_devarg_rep_based_pf,
6343 						 (void *)&representor);
6344 			if (ret) {
6345 				ret = -EINVAL;
6346 				goto err;
6347 			}
6348 			/*
6349 			 * Handler for "rep_based_pf" devarg.
6350 			 * Invoked as for ex: "-a 000:00:0d.0,
6351 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6352 			 */
6353 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6354 						 bnxt_parse_devarg_rep_q_r2f,
6355 						 (void *)&representor);
6356 			if (ret) {
6357 				ret = -EINVAL;
6358 				goto err;
6359 			}
6360 			/*
6361 			 * Handler for "rep_based_pf" devarg.
6362 			 * Invoked as for ex: "-a 000:00:0d.0,
6363 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6364 			 */
6365 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6366 						 bnxt_parse_devarg_rep_q_f2r,
6367 						 (void *)&representor);
6368 			if (ret) {
6369 				ret = -EINVAL;
6370 				goto err;
6371 			}
6372 			/*
6373 			 * Handler for "rep_based_pf" devarg.
6374 			 * Invoked as for ex: "-a 000:00:0d.0,
6375 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6376 			 */
6377 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6378 						 bnxt_parse_devarg_rep_fc_r2f,
6379 						 (void *)&representor);
6380 			if (ret) {
6381 				ret = -EINVAL;
6382 				goto err;
6383 			}
6384 			/*
6385 			 * Handler for "rep_based_pf" devarg.
6386 			 * Invoked as for ex: "-a 000:00:0d.0,
6387 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6388 			 */
6389 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6390 						 bnxt_parse_devarg_rep_fc_f2r,
6391 						 (void *)&representor);
6392 			if (ret) {
6393 				ret = -EINVAL;
6394 				goto err;
6395 			}
6396 		}
6397 
6398 		ret = rte_eth_dev_create(&pci_dev->device, name,
6399 					 sizeof(struct bnxt_representor),
6400 					 NULL, NULL,
6401 					 bnxt_representor_init,
6402 					 &representor);
6403 		if (ret) {
6404 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6405 				    "representor %s.", name);
6406 			goto err;
6407 		}
6408 
6409 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6410 		if (!vf_rep_eth_dev) {
6411 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6412 				    " for VF-Rep: %s.", name);
6413 			ret = -ENODEV;
6414 			goto err;
6415 		}
6416 
6417 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6418 			    backing_eth_dev->data->port_id);
6419 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6420 							 vf_rep_eth_dev;
6421 		backing_bp->num_reps++;
6422 
6423 	}
6424 
6425 	rte_kvargs_free(kvlist);
6426 	return 0;
6427 
6428 err:
6429 	/* If num_rep > 1, then rollback already created
6430 	 * ports, since we'll be failing the probe anyway
6431 	 */
6432 	if (num_rep > 1)
6433 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6434 	rte_errno = -ret;
6435 	rte_kvargs_free(kvlist);
6436 
6437 	return ret;
6438 }
6439 
6440 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6441 			  struct rte_pci_device *pci_dev)
6442 {
6443 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6444 	struct rte_eth_dev *backing_eth_dev;
6445 	uint16_t num_rep;
6446 	int ret = 0;
6447 
6448 	if (pci_dev->device.devargs) {
6449 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6450 					    &eth_da);
6451 		if (ret)
6452 			return ret;
6453 	}
6454 
6455 	num_rep = eth_da.nb_representor_ports;
6456 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6457 		    num_rep);
6458 
6459 	/* We could come here after first level of probe is already invoked
6460 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6461 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6462 	 */
6463 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6464 	if (backing_eth_dev == NULL) {
6465 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6466 					 sizeof(struct bnxt),
6467 					 eth_dev_pci_specific_init, pci_dev,
6468 					 bnxt_dev_init, NULL);
6469 
6470 		if (ret || !num_rep)
6471 			return ret;
6472 
6473 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6474 	}
6475 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6476 		    backing_eth_dev->data->port_id);
6477 
6478 	if (!num_rep)
6479 		return ret;
6480 
6481 	/* probe representor ports now */
6482 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6483 				  pci_dev->device.devargs->args);
6484 
6485 	return ret;
6486 }
6487 
6488 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6489 {
6490 	struct rte_eth_dev *eth_dev;
6491 
6492 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6493 	if (!eth_dev)
6494 		return 0; /* Invoked typically only by OVS-DPDK, by the
6495 			   * time it comes here the eth_dev is already
6496 			   * deleted by rte_eth_dev_close(), so returning
6497 			   * +ve value will at least help in proper cleanup
6498 			   */
6499 
6500 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6501 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6502 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6503 			return rte_eth_dev_destroy(eth_dev,
6504 						   bnxt_representor_uninit);
6505 		else
6506 			return rte_eth_dev_destroy(eth_dev,
6507 						   bnxt_dev_uninit);
6508 	} else {
6509 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6510 	}
6511 }
6512 
6513 static struct rte_pci_driver bnxt_rte_pmd = {
6514 	.id_table = bnxt_pci_id_map,
6515 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6516 			RTE_PCI_DRV_INTR_RMV |
6517 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6518 						  * and OVS-DPDK
6519 						  */
6520 	.probe = bnxt_pci_probe,
6521 	.remove = bnxt_pci_remove,
6522 };
6523 
6524 static bool
6525 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6526 {
6527 	if (strcmp(dev->device->driver->name, drv->driver.name))
6528 		return false;
6529 
6530 	return true;
6531 }
6532 
6533 bool is_bnxt_supported(struct rte_eth_dev *dev)
6534 {
6535 	return is_device_supported(dev, &bnxt_rte_pmd);
6536 }
6537 
6538 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6539 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6540 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6541 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6542