xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision c9902a15bd005b6d4fe072cf7b60fe4ee679155f)
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 		if (!rxq->rx_deferred_start) {
487 			bp->eth_dev->data->rx_queue_state[j] =
488 				RTE_ETH_QUEUE_STATE_STARTED;
489 			rxq->rx_started = true;
490 		}
491 	}
492 
493 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
494 
495 	rc = bnxt_vnic_rss_configure(bp, vnic);
496 	if (rc)
497 		goto err_out;
498 
499 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
500 
501 	rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
502 				    (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO) ?
503 				    true : false);
504 	if (rc)
505 		goto err_out;
506 
507 	return 0;
508 err_out:
509 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
510 		    vnic_id, rc);
511 	return rc;
512 }
513 
514 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
515 {
516 	int rc = 0;
517 
518 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
519 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
520 	if (rc)
521 		return rc;
522 
523 	PMD_DRV_LOG(DEBUG,
524 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
525 		    " rx_fc_in_tbl.ctx_id = %d\n",
526 		    bp->flow_stat->rx_fc_in_tbl.va,
527 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
528 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
529 
530 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
531 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
532 	if (rc)
533 		return rc;
534 
535 	PMD_DRV_LOG(DEBUG,
536 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
537 		    " rx_fc_out_tbl.ctx_id = %d\n",
538 		    bp->flow_stat->rx_fc_out_tbl.va,
539 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
540 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
541 
542 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
543 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
544 	if (rc)
545 		return rc;
546 
547 	PMD_DRV_LOG(DEBUG,
548 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
549 		    " tx_fc_in_tbl.ctx_id = %d\n",
550 		    bp->flow_stat->tx_fc_in_tbl.va,
551 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
552 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
553 
554 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
555 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
556 	if (rc)
557 		return rc;
558 
559 	PMD_DRV_LOG(DEBUG,
560 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
561 		    " tx_fc_out_tbl.ctx_id = %d\n",
562 		    bp->flow_stat->tx_fc_out_tbl.va,
563 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
564 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
565 
566 	memset(bp->flow_stat->rx_fc_out_tbl.va,
567 	       0,
568 	       bp->flow_stat->rx_fc_out_tbl.size);
569 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
570 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
571 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
572 				       bp->flow_stat->max_fc,
573 				       true);
574 	if (rc)
575 		return rc;
576 
577 	memset(bp->flow_stat->tx_fc_out_tbl.va,
578 	       0,
579 	       bp->flow_stat->tx_fc_out_tbl.size);
580 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
581 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
582 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
583 				       bp->flow_stat->max_fc,
584 				       true);
585 
586 	return rc;
587 }
588 
589 static int bnxt_alloc_ctx_mem_buf(struct bnxt *bp, char *type, size_t size,
590 				  struct bnxt_ctx_mem_buf_info *ctx)
591 {
592 	if (!ctx)
593 		return -EINVAL;
594 
595 	ctx->va = rte_zmalloc_socket(type, size, 0,
596 				     bp->eth_dev->device->numa_node);
597 	if (ctx->va == NULL)
598 		return -ENOMEM;
599 	rte_mem_lock_page(ctx->va);
600 	ctx->size = size;
601 	ctx->dma = rte_mem_virt2iova(ctx->va);
602 	if (ctx->dma == RTE_BAD_IOVA)
603 		return -ENOMEM;
604 
605 	return 0;
606 }
607 
608 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
609 {
610 	struct rte_pci_device *pdev = bp->pdev;
611 	char type[RTE_MEMZONE_NAMESIZE];
612 	uint16_t max_fc;
613 	int rc = 0;
614 
615 	max_fc = bp->flow_stat->max_fc;
616 
617 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
618 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
619 	/* 4 bytes for each counter-id */
620 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
621 				    max_fc * 4,
622 				    &bp->flow_stat->rx_fc_in_tbl);
623 	if (rc)
624 		return rc;
625 
626 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
627 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
628 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
629 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
630 				    max_fc * 16,
631 				    &bp->flow_stat->rx_fc_out_tbl);
632 	if (rc)
633 		return rc;
634 
635 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
636 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
637 	/* 4 bytes for each counter-id */
638 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
639 				    max_fc * 4,
640 				    &bp->flow_stat->tx_fc_in_tbl);
641 	if (rc)
642 		return rc;
643 
644 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
645 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
646 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
647 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
648 				    max_fc * 16,
649 				    &bp->flow_stat->tx_fc_out_tbl);
650 	if (rc)
651 		return rc;
652 
653 	rc = bnxt_register_fc_ctx_mem(bp);
654 
655 	return rc;
656 }
657 
658 static int bnxt_init_ctx_mem(struct bnxt *bp)
659 {
660 	int rc = 0;
661 
662 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
663 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
664 	    !BNXT_FLOW_XSTATS_EN(bp))
665 		return 0;
666 
667 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
668 	if (rc)
669 		return rc;
670 
671 	rc = bnxt_init_fc_ctx_mem(bp);
672 
673 	return rc;
674 }
675 
676 static int bnxt_update_phy_setting(struct bnxt *bp)
677 {
678 	struct rte_eth_link new;
679 	int rc;
680 
681 	rc = bnxt_get_hwrm_link_config(bp, &new);
682 	if (rc) {
683 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
684 		return rc;
685 	}
686 
687 	/*
688 	 * On BCM957508-N2100 adapters, FW will not allow any user other
689 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
690 	 * always returns link up. Force phy update always in that case.
691 	 */
692 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
693 		rc = bnxt_set_hwrm_link_config(bp, true);
694 		if (rc) {
695 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
696 			return rc;
697 		}
698 	}
699 
700 	return rc;
701 }
702 
703 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
704 {
705 	rte_free(bp->prev_rx_ring_stats);
706 	rte_free(bp->prev_tx_ring_stats);
707 
708 	bp->prev_rx_ring_stats = NULL;
709 	bp->prev_tx_ring_stats = NULL;
710 }
711 
712 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
713 {
714 	bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
715 					      sizeof(struct bnxt_ring_stats) *
716 					      bp->rx_cp_nr_rings,
717 					      0);
718 	if (bp->prev_rx_ring_stats == NULL)
719 		return -ENOMEM;
720 
721 	bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
722 					     sizeof(struct bnxt_ring_stats) *
723 					     bp->tx_cp_nr_rings,
724 					     0);
725 	if (bp->prev_tx_ring_stats == NULL)
726 		goto error;
727 
728 	return 0;
729 
730 error:
731 	bnxt_free_prev_ring_stats(bp);
732 	return -ENOMEM;
733 }
734 
735 static int bnxt_start_nic(struct bnxt *bp)
736 {
737 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
738 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
739 	uint32_t intr_vector = 0;
740 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
741 	uint32_t vec = BNXT_MISC_VEC_ID;
742 	unsigned int i, j;
743 	int rc;
744 
745 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
746 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
747 			DEV_RX_OFFLOAD_JUMBO_FRAME;
748 		bp->flags |= BNXT_FLAG_JUMBO;
749 	} else {
750 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
751 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
752 		bp->flags &= ~BNXT_FLAG_JUMBO;
753 	}
754 
755 	/* THOR does not support ring groups.
756 	 * But we will use the array to save RSS context IDs.
757 	 */
758 	if (BNXT_CHIP_P5(bp))
759 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
760 
761 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
762 	if (rc) {
763 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
764 		goto err_out;
765 	}
766 
767 	rc = bnxt_alloc_hwrm_rings(bp);
768 	if (rc) {
769 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
770 		goto err_out;
771 	}
772 
773 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
774 	if (rc) {
775 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
776 		goto err_out;
777 	}
778 
779 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
780 		goto skip_cosq_cfg;
781 
782 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
783 		if (bp->rx_cos_queue[i].id != 0xff) {
784 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
785 
786 			if (!vnic) {
787 				PMD_DRV_LOG(ERR,
788 					    "Num pools more than FW profile\n");
789 				rc = -EINVAL;
790 				goto err_out;
791 			}
792 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
793 			bp->rx_cosq_cnt++;
794 		}
795 	}
796 
797 skip_cosq_cfg:
798 	rc = bnxt_mq_rx_configure(bp);
799 	if (rc) {
800 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
801 		goto err_out;
802 	}
803 
804 	/* default vnic 0 */
805 	rc = bnxt_setup_one_vnic(bp, 0);
806 	if (rc)
807 		goto err_out;
808 	/* VNIC configuration */
809 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
810 		for (i = 1; i < bp->nr_vnics; i++) {
811 			rc = bnxt_setup_one_vnic(bp, i);
812 			if (rc)
813 				goto err_out;
814 		}
815 	}
816 
817 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
818 	if (rc) {
819 		PMD_DRV_LOG(ERR,
820 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
821 		goto err_out;
822 	}
823 
824 	/* check and configure queue intr-vector mapping */
825 	if ((rte_intr_cap_multiple(intr_handle) ||
826 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
827 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
828 		intr_vector = bp->eth_dev->data->nb_rx_queues;
829 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
830 		if (intr_vector > bp->rx_cp_nr_rings) {
831 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
832 					bp->rx_cp_nr_rings);
833 			return -ENOTSUP;
834 		}
835 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
836 		if (rc)
837 			return rc;
838 	}
839 
840 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
841 		intr_handle->intr_vec =
842 			rte_zmalloc("intr_vec",
843 				    bp->eth_dev->data->nb_rx_queues *
844 				    sizeof(int), 0);
845 		if (intr_handle->intr_vec == NULL) {
846 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
847 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
848 			rc = -ENOMEM;
849 			goto err_out;
850 		}
851 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
852 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
853 			 intr_handle->intr_vec, intr_handle->nb_efd,
854 			intr_handle->max_intr);
855 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
856 		     queue_id++) {
857 			intr_handle->intr_vec[queue_id] =
858 							vec + BNXT_RX_VEC_START;
859 			if (vec < base + intr_handle->nb_efd - 1)
860 				vec++;
861 		}
862 	}
863 
864 	/* enable uio/vfio intr/eventfd mapping */
865 	rc = rte_intr_enable(intr_handle);
866 #ifndef RTE_EXEC_ENV_FREEBSD
867 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
868 	if (rc)
869 		goto err_out;
870 #endif
871 
872 	rc = bnxt_update_phy_setting(bp);
873 	if (rc)
874 		goto err_out;
875 
876 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
877 	if (!bp->mark_table)
878 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
879 
880 	return 0;
881 
882 err_out:
883 	/* Some of the error status returned by FW may not be from errno.h */
884 	if (rc > 0)
885 		rc = -EIO;
886 
887 	return rc;
888 }
889 
890 static int bnxt_shutdown_nic(struct bnxt *bp)
891 {
892 	bnxt_free_all_hwrm_resources(bp);
893 	bnxt_free_all_filters(bp);
894 	bnxt_free_all_vnics(bp);
895 	return 0;
896 }
897 
898 /*
899  * Device configuration and status function
900  */
901 
902 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
903 {
904 	uint32_t link_speed = 0;
905 	uint32_t speed_capa = 0;
906 
907 	if (bp->link_info == NULL)
908 		return 0;
909 
910 	link_speed = bp->link_info->support_speeds;
911 
912 	/* If PAM4 is configured, use PAM4 supported speed */
913 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
914 		link_speed = bp->link_info->support_pam4_speeds;
915 
916 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
917 		speed_capa |= ETH_LINK_SPEED_100M;
918 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
919 		speed_capa |= ETH_LINK_SPEED_100M_HD;
920 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
921 		speed_capa |= ETH_LINK_SPEED_1G;
922 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
923 		speed_capa |= ETH_LINK_SPEED_2_5G;
924 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
925 		speed_capa |= ETH_LINK_SPEED_10G;
926 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
927 		speed_capa |= ETH_LINK_SPEED_20G;
928 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
929 		speed_capa |= ETH_LINK_SPEED_25G;
930 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
931 		speed_capa |= ETH_LINK_SPEED_40G;
932 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
933 		speed_capa |= ETH_LINK_SPEED_50G;
934 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
935 		speed_capa |= ETH_LINK_SPEED_100G;
936 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
937 		speed_capa |= ETH_LINK_SPEED_50G;
938 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
939 		speed_capa |= ETH_LINK_SPEED_100G;
940 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
941 		speed_capa |= ETH_LINK_SPEED_200G;
942 
943 	if (bp->link_info->auto_mode ==
944 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
945 		speed_capa |= ETH_LINK_SPEED_FIXED;
946 
947 	return speed_capa;
948 }
949 
950 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
951 				struct rte_eth_dev_info *dev_info)
952 {
953 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
954 	struct bnxt *bp = eth_dev->data->dev_private;
955 	uint16_t max_vnics, i, j, vpool, vrxq;
956 	unsigned int max_rx_rings;
957 	int rc;
958 
959 	rc = is_bnxt_in_error(bp);
960 	if (rc)
961 		return rc;
962 
963 	/* MAC Specifics */
964 	dev_info->max_mac_addrs = bp->max_l2_ctx;
965 	dev_info->max_hash_mac_addrs = 0;
966 
967 	/* PF/VF specifics */
968 	if (BNXT_PF(bp))
969 		dev_info->max_vfs = pdev->max_vfs;
970 
971 	max_rx_rings = bnxt_max_rings(bp);
972 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
973 	dev_info->max_rx_queues = max_rx_rings;
974 	dev_info->max_tx_queues = max_rx_rings;
975 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
976 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
977 	max_vnics = bp->max_vnics;
978 
979 	/* MTU specifics */
980 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
981 	dev_info->max_mtu = BNXT_MAX_MTU;
982 
983 	/* Fast path specifics */
984 	dev_info->min_rx_bufsize = 1;
985 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
986 
987 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
988 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
989 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
990 	if (bp->vnic_cap_flags & BNXT_VNIC_CAP_VLAN_RX_STRIP)
991 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_VLAN_STRIP;
992 	dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
993 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
994 				    dev_info->tx_queue_offload_capa;
995 	if (bp->fw_cap & BNXT_FW_CAP_VLAN_TX_INSERT)
996 		dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_VLAN_INSERT;
997 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
998 
999 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
1000 	dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
1001 			     RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
1002 
1003 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
1004 		.rx_thresh = {
1005 			.pthresh = 8,
1006 			.hthresh = 8,
1007 			.wthresh = 0,
1008 		},
1009 		.rx_free_thresh = 32,
1010 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
1011 	};
1012 
1013 	dev_info->default_txconf = (struct rte_eth_txconf) {
1014 		.tx_thresh = {
1015 			.pthresh = 32,
1016 			.hthresh = 0,
1017 			.wthresh = 0,
1018 		},
1019 		.tx_free_thresh = 32,
1020 		.tx_rs_thresh = 32,
1021 	};
1022 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
1023 
1024 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1025 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1026 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1027 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1028 
1029 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1030 		dev_info->switch_info.name = eth_dev->device->name;
1031 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1032 		dev_info->switch_info.port_id =
1033 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1034 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1035 	}
1036 
1037 	/*
1038 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1039 	 *       need further investigation.
1040 	 */
1041 
1042 	/* VMDq resources */
1043 	vpool = 64; /* ETH_64_POOLS */
1044 	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
1045 	for (i = 0; i < 4; vpool >>= 1, i++) {
1046 		if (max_vnics > vpool) {
1047 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1048 				if (dev_info->max_rx_queues > vrxq) {
1049 					if (vpool > vrxq)
1050 						vpool = vrxq;
1051 					goto found;
1052 				}
1053 			}
1054 			/* Not enough resources to support VMDq */
1055 			break;
1056 		}
1057 	}
1058 	/* Not enough resources to support VMDq */
1059 	vpool = 0;
1060 	vrxq = 0;
1061 found:
1062 	dev_info->max_vmdq_pools = vpool;
1063 	dev_info->vmdq_queue_num = vrxq;
1064 
1065 	dev_info->vmdq_pool_base = 0;
1066 	dev_info->vmdq_queue_base = 0;
1067 
1068 	return 0;
1069 }
1070 
1071 /* Configure the device based on the configuration provided */
1072 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1073 {
1074 	struct bnxt *bp = eth_dev->data->dev_private;
1075 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1076 	int rc;
1077 
1078 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1079 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1080 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1081 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1082 
1083 	rc = is_bnxt_in_error(bp);
1084 	if (rc)
1085 		return rc;
1086 
1087 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1088 		rc = bnxt_hwrm_check_vf_rings(bp);
1089 		if (rc) {
1090 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1091 			return -ENOSPC;
1092 		}
1093 
1094 		/* If a resource has already been allocated - in this case
1095 		 * it is the async completion ring, free it. Reallocate it after
1096 		 * resource reservation. This will ensure the resource counts
1097 		 * are calculated correctly.
1098 		 */
1099 
1100 		pthread_mutex_lock(&bp->def_cp_lock);
1101 
1102 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1103 			bnxt_disable_int(bp);
1104 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1105 		}
1106 
1107 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1108 		if (rc) {
1109 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1110 			pthread_mutex_unlock(&bp->def_cp_lock);
1111 			return -ENOSPC;
1112 		}
1113 
1114 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1115 			rc = bnxt_alloc_async_cp_ring(bp);
1116 			if (rc) {
1117 				pthread_mutex_unlock(&bp->def_cp_lock);
1118 				return rc;
1119 			}
1120 			bnxt_enable_int(bp);
1121 		}
1122 
1123 		pthread_mutex_unlock(&bp->def_cp_lock);
1124 	}
1125 
1126 	/* Inherit new configurations */
1127 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1128 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1129 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1130 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1131 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1132 	    bp->max_stat_ctx)
1133 		goto resource_error;
1134 
1135 	if (BNXT_HAS_RING_GRPS(bp) &&
1136 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1137 		goto resource_error;
1138 
1139 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1140 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1141 		goto resource_error;
1142 
1143 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1144 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1145 
1146 	if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1147 		rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1148 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1149 
1150 	if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1151 		eth_dev->data->mtu =
1152 			eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1153 			RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1154 			BNXT_NUM_VLANS;
1155 		bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1156 	}
1157 	return 0;
1158 
1159 resource_error:
1160 	PMD_DRV_LOG(ERR,
1161 		    "Insufficient resources to support requested config\n");
1162 	PMD_DRV_LOG(ERR,
1163 		    "Num Queues Requested: Tx %d, Rx %d\n",
1164 		    eth_dev->data->nb_tx_queues,
1165 		    eth_dev->data->nb_rx_queues);
1166 	PMD_DRV_LOG(ERR,
1167 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1168 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1169 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1170 	return -ENOSPC;
1171 }
1172 
1173 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1174 {
1175 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1176 
1177 	if (link->link_status)
1178 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1179 			eth_dev->data->port_id,
1180 			(uint32_t)link->link_speed,
1181 			(link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1182 			("full-duplex") : ("half-duplex\n"));
1183 	else
1184 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1185 			eth_dev->data->port_id);
1186 }
1187 
1188 /*
1189  * Determine whether the current configuration requires support for scattered
1190  * receive; return 1 if scattered receive is required and 0 if not.
1191  */
1192 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1193 {
1194 	uint16_t buf_size;
1195 	int i;
1196 
1197 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1198 		return 1;
1199 
1200 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1201 		return 1;
1202 
1203 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1204 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1205 
1206 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1207 				      RTE_PKTMBUF_HEADROOM);
1208 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1209 			return 1;
1210 	}
1211 	return 0;
1212 }
1213 
1214 static eth_rx_burst_t
1215 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1216 {
1217 	struct bnxt *bp = eth_dev->data->dev_private;
1218 
1219 	/* Disable vector mode RX for Stingray2 for now */
1220 	if (BNXT_CHIP_SR2(bp)) {
1221 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1222 		return bnxt_recv_pkts;
1223 	}
1224 
1225 #if (defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)) && \
1226 	!defined(RTE_LIBRTE_IEEE1588)
1227 
1228 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1229 	if (eth_dev->data->scattered_rx)
1230 		goto use_scalar_rx;
1231 
1232 	/*
1233 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1234 	 * asynchronous completions and receive completions can be placed in
1235 	 * the same completion ring.
1236 	 */
1237 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1238 		goto use_scalar_rx;
1239 
1240 	/*
1241 	 * Vector mode receive cannot be enabled if any receive offloads outside
1242 	 * a limited subset have been enabled.
1243 	 */
1244 	if (eth_dev->data->dev_conf.rxmode.offloads &
1245 		~(DEV_RX_OFFLOAD_VLAN_STRIP |
1246 		  DEV_RX_OFFLOAD_KEEP_CRC |
1247 		  DEV_RX_OFFLOAD_JUMBO_FRAME |
1248 		  DEV_RX_OFFLOAD_IPV4_CKSUM |
1249 		  DEV_RX_OFFLOAD_UDP_CKSUM |
1250 		  DEV_RX_OFFLOAD_TCP_CKSUM |
1251 		  DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1252 		  DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1253 		  DEV_RX_OFFLOAD_RSS_HASH |
1254 		  DEV_RX_OFFLOAD_VLAN_FILTER))
1255 		goto use_scalar_rx;
1256 
1257 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1258 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1259 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1260 		PMD_DRV_LOG(INFO,
1261 			    "Using AVX2 vector mode receive for port %d\n",
1262 			    eth_dev->data->port_id);
1263 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1264 		return bnxt_recv_pkts_vec_avx2;
1265 	}
1266  #endif
1267 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1268 		PMD_DRV_LOG(INFO,
1269 			    "Using SSE vector mode receive for port %d\n",
1270 			    eth_dev->data->port_id);
1271 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1272 		return bnxt_recv_pkts_vec;
1273 	}
1274 
1275 use_scalar_rx:
1276 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1277 		    eth_dev->data->port_id);
1278 	PMD_DRV_LOG(INFO,
1279 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1280 		    eth_dev->data->port_id,
1281 		    eth_dev->data->scattered_rx,
1282 		    eth_dev->data->dev_conf.rxmode.offloads);
1283 #endif
1284 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1285 	return bnxt_recv_pkts;
1286 }
1287 
1288 static eth_tx_burst_t
1289 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1290 {
1291 	struct bnxt *bp = eth_dev->data->dev_private;
1292 
1293 	/* Disable vector mode TX for Stingray2 for now */
1294 	if (BNXT_CHIP_SR2(bp))
1295 		return bnxt_xmit_pkts;
1296 
1297 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64) && \
1298 	!defined(RTE_LIBRTE_IEEE1588)
1299 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1300 
1301 	/*
1302 	 * Vector mode transmit can be enabled only if not using scatter rx
1303 	 * or tx offloads.
1304 	 */
1305 	if (eth_dev->data->scattered_rx ||
1306 	    (offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) ||
1307 	    BNXT_TRUFLOW_EN(bp))
1308 		goto use_scalar_tx;
1309 
1310 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1311 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1312 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1313 		PMD_DRV_LOG(INFO,
1314 			    "Using AVX2 vector mode transmit for port %d\n",
1315 			    eth_dev->data->port_id);
1316 		return bnxt_xmit_pkts_vec_avx2;
1317 	}
1318 #endif
1319 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1320 		PMD_DRV_LOG(INFO,
1321 			    "Using SSE vector mode transmit for port %d\n",
1322 			    eth_dev->data->port_id);
1323 		return bnxt_xmit_pkts_vec;
1324 	}
1325 
1326 use_scalar_tx:
1327 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1328 		    eth_dev->data->port_id);
1329 	PMD_DRV_LOG(INFO,
1330 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1331 		    eth_dev->data->port_id,
1332 		    eth_dev->data->scattered_rx,
1333 		    offloads);
1334 #endif
1335 	return bnxt_xmit_pkts;
1336 }
1337 
1338 static int bnxt_handle_if_change_status(struct bnxt *bp)
1339 {
1340 	int rc;
1341 
1342 	/* Since fw has undergone a reset and lost all contexts,
1343 	 * set fatal flag to not issue hwrm during cleanup
1344 	 */
1345 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1346 	bnxt_uninit_resources(bp, true);
1347 
1348 	/* clear fatal flag so that re-init happens */
1349 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1350 	rc = bnxt_init_resources(bp, true);
1351 
1352 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1353 
1354 	return rc;
1355 }
1356 
1357 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1358 {
1359 	struct bnxt *bp = eth_dev->data->dev_private;
1360 	int rc = 0;
1361 
1362 	if (!BNXT_SINGLE_PF(bp))
1363 		return -ENOTSUP;
1364 
1365 	if (!bp->link_info->link_up)
1366 		rc = bnxt_set_hwrm_link_config(bp, true);
1367 	if (!rc)
1368 		eth_dev->data->dev_link.link_status = 1;
1369 
1370 	bnxt_print_link_info(eth_dev);
1371 	return rc;
1372 }
1373 
1374 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1375 {
1376 	struct bnxt *bp = eth_dev->data->dev_private;
1377 
1378 	if (!BNXT_SINGLE_PF(bp))
1379 		return -ENOTSUP;
1380 
1381 	eth_dev->data->dev_link.link_status = 0;
1382 	bnxt_set_hwrm_link_config(bp, false);
1383 	bp->link_info->link_up = 0;
1384 
1385 	return 0;
1386 }
1387 
1388 static void bnxt_free_switch_domain(struct bnxt *bp)
1389 {
1390 	int rc = 0;
1391 
1392 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1393 		return;
1394 
1395 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1396 	if (rc)
1397 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1398 			    bp->switch_domain_id, rc);
1399 }
1400 
1401 static void bnxt_ptp_get_current_time(void *arg)
1402 {
1403 	struct bnxt *bp = arg;
1404 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1405 	int rc;
1406 
1407 	rc = is_bnxt_in_error(bp);
1408 	if (rc)
1409 		return;
1410 
1411 	if (!ptp)
1412 		return;
1413 
1414 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1415 				&ptp->current_time);
1416 
1417 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1418 	if (rc != 0) {
1419 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1420 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1421 	}
1422 }
1423 
1424 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1425 {
1426 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1427 	int rc;
1428 
1429 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1430 		return 0;
1431 
1432 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1433 				&ptp->current_time);
1434 
1435 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1436 	return rc;
1437 }
1438 
1439 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1440 {
1441 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1442 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1443 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1444 	}
1445 }
1446 
1447 static void bnxt_ptp_stop(struct bnxt *bp)
1448 {
1449 	bnxt_cancel_ptp_alarm(bp);
1450 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1451 }
1452 
1453 static int bnxt_ptp_start(struct bnxt *bp)
1454 {
1455 	int rc;
1456 
1457 	rc = bnxt_schedule_ptp_alarm(bp);
1458 	if (rc != 0) {
1459 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1460 	} else {
1461 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1462 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1463 	}
1464 
1465 	return rc;
1466 }
1467 
1468 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1469 {
1470 	struct bnxt *bp = eth_dev->data->dev_private;
1471 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1472 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1473 	struct rte_eth_link link;
1474 	int ret;
1475 
1476 	eth_dev->data->dev_started = 0;
1477 	eth_dev->data->scattered_rx = 0;
1478 
1479 	/* Prevent crashes when queues are still in use */
1480 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1481 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1482 
1483 	bnxt_disable_int(bp);
1484 
1485 	/* disable uio/vfio intr/eventfd mapping */
1486 	rte_intr_disable(intr_handle);
1487 
1488 	/* Stop the child representors for this device */
1489 	ret = bnxt_rep_stop_all(bp);
1490 	if (ret != 0)
1491 		return ret;
1492 
1493 	/* delete the bnxt ULP port details */
1494 	bnxt_ulp_port_deinit(bp);
1495 
1496 	bnxt_cancel_fw_health_check(bp);
1497 
1498 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1499 		bnxt_cancel_ptp_alarm(bp);
1500 
1501 	/* Do not bring link down during reset recovery */
1502 	if (!is_bnxt_in_error(bp)) {
1503 		bnxt_dev_set_link_down_op(eth_dev);
1504 		/* Wait for link to be reset */
1505 		if (BNXT_SINGLE_PF(bp))
1506 			rte_delay_ms(500);
1507 		/* clear the recorded link status */
1508 		memset(&link, 0, sizeof(link));
1509 		rte_eth_linkstatus_set(eth_dev, &link);
1510 	}
1511 
1512 	/* Clean queue intr-vector mapping */
1513 	rte_intr_efd_disable(intr_handle);
1514 	if (intr_handle->intr_vec != NULL) {
1515 		rte_free(intr_handle->intr_vec);
1516 		intr_handle->intr_vec = NULL;
1517 	}
1518 
1519 	bnxt_hwrm_port_clr_stats(bp);
1520 	bnxt_free_tx_mbufs(bp);
1521 	bnxt_free_rx_mbufs(bp);
1522 	/* Process any remaining notifications in default completion queue */
1523 	bnxt_int_handler(eth_dev);
1524 	bnxt_shutdown_nic(bp);
1525 	bnxt_hwrm_if_change(bp, false);
1526 
1527 	bnxt_free_prev_ring_stats(bp);
1528 	rte_free(bp->mark_table);
1529 	bp->mark_table = NULL;
1530 
1531 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1532 	bp->rx_cosq_cnt = 0;
1533 	/* All filters are deleted on a port stop. */
1534 	if (BNXT_FLOW_XSTATS_EN(bp))
1535 		bp->flow_stat->flow_count = 0;
1536 
1537 	return 0;
1538 }
1539 
1540 /* Unload the driver, release resources */
1541 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1542 {
1543 	struct bnxt *bp = eth_dev->data->dev_private;
1544 
1545 	pthread_mutex_lock(&bp->err_recovery_lock);
1546 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1547 		PMD_DRV_LOG(ERR,
1548 			    "Adapter recovering from error..Please retry\n");
1549 		pthread_mutex_unlock(&bp->err_recovery_lock);
1550 		return -EAGAIN;
1551 	}
1552 	pthread_mutex_unlock(&bp->err_recovery_lock);
1553 
1554 	return bnxt_dev_stop(eth_dev);
1555 }
1556 
1557 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1558 {
1559 	struct bnxt *bp = eth_dev->data->dev_private;
1560 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1561 	int vlan_mask = 0;
1562 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1563 
1564 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1565 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1566 		return -EINVAL;
1567 	}
1568 
1569 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1570 		PMD_DRV_LOG(ERR,
1571 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1572 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1573 
1574 	do {
1575 		rc = bnxt_hwrm_if_change(bp, true);
1576 		if (rc == 0 || rc != -EAGAIN)
1577 			break;
1578 
1579 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1580 	} while (retry_cnt--);
1581 
1582 	if (rc)
1583 		return rc;
1584 
1585 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1586 		rc = bnxt_handle_if_change_status(bp);
1587 		if (rc)
1588 			return rc;
1589 	}
1590 
1591 	bnxt_enable_int(bp);
1592 
1593 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1594 
1595 	rc = bnxt_start_nic(bp);
1596 	if (rc)
1597 		goto error;
1598 
1599 	rc = bnxt_alloc_prev_ring_stats(bp);
1600 	if (rc)
1601 		goto error;
1602 
1603 	eth_dev->data->dev_started = 1;
1604 
1605 	bnxt_link_update_op(eth_dev, 1);
1606 
1607 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1608 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1609 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1610 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1611 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1612 	if (rc)
1613 		goto error;
1614 
1615 	/* Initialize bnxt ULP port details */
1616 	rc = bnxt_ulp_port_init(bp);
1617 	if (rc)
1618 		goto error;
1619 
1620 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1621 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1622 
1623 	bnxt_schedule_fw_health_check(bp);
1624 
1625 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1626 		bnxt_schedule_ptp_alarm(bp);
1627 
1628 	return 0;
1629 
1630 error:
1631 	bnxt_dev_stop(eth_dev);
1632 	return rc;
1633 }
1634 
1635 static void
1636 bnxt_uninit_locks(struct bnxt *bp)
1637 {
1638 	pthread_mutex_destroy(&bp->flow_lock);
1639 	pthread_mutex_destroy(&bp->def_cp_lock);
1640 	pthread_mutex_destroy(&bp->health_check_lock);
1641 	pthread_mutex_destroy(&bp->err_recovery_lock);
1642 	if (bp->rep_info) {
1643 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1644 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1645 	}
1646 }
1647 
1648 static void bnxt_drv_uninit(struct bnxt *bp)
1649 {
1650 	bnxt_free_leds_info(bp);
1651 	bnxt_free_cos_queues(bp);
1652 	bnxt_free_link_info(bp);
1653 	bnxt_free_parent_info(bp);
1654 	bnxt_uninit_locks(bp);
1655 
1656 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1657 	bp->tx_mem_zone = NULL;
1658 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1659 	bp->rx_mem_zone = NULL;
1660 
1661 	bnxt_free_vf_info(bp);
1662 	bnxt_free_pf_info(bp);
1663 
1664 	rte_free(bp->grp_info);
1665 	bp->grp_info = NULL;
1666 }
1667 
1668 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1669 {
1670 	struct bnxt *bp = eth_dev->data->dev_private;
1671 	int ret = 0;
1672 
1673 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1674 		return 0;
1675 
1676 	pthread_mutex_lock(&bp->err_recovery_lock);
1677 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1678 		PMD_DRV_LOG(ERR,
1679 			    "Adapter recovering from error...Please retry\n");
1680 		pthread_mutex_unlock(&bp->err_recovery_lock);
1681 		return -EAGAIN;
1682 	}
1683 	pthread_mutex_unlock(&bp->err_recovery_lock);
1684 
1685 	/* cancel the recovery handler before remove dev */
1686 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1687 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1688 	bnxt_cancel_fc_thread(bp);
1689 
1690 	if (eth_dev->data->dev_started)
1691 		ret = bnxt_dev_stop(eth_dev);
1692 
1693 	bnxt_uninit_resources(bp, false);
1694 
1695 	bnxt_drv_uninit(bp);
1696 
1697 	return ret;
1698 }
1699 
1700 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1701 				    uint32_t index)
1702 {
1703 	struct bnxt *bp = eth_dev->data->dev_private;
1704 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1705 	struct bnxt_vnic_info *vnic;
1706 	struct bnxt_filter_info *filter, *temp_filter;
1707 	uint32_t i;
1708 
1709 	if (is_bnxt_in_error(bp))
1710 		return;
1711 
1712 	/*
1713 	 * Loop through all VNICs from the specified filter flow pools to
1714 	 * remove the corresponding MAC addr filter
1715 	 */
1716 	for (i = 0; i < bp->nr_vnics; i++) {
1717 		if (!(pool_mask & (1ULL << i)))
1718 			continue;
1719 
1720 		vnic = &bp->vnic_info[i];
1721 		filter = STAILQ_FIRST(&vnic->filter);
1722 		while (filter) {
1723 			temp_filter = STAILQ_NEXT(filter, next);
1724 			if (filter->mac_index == index) {
1725 				STAILQ_REMOVE(&vnic->filter, filter,
1726 						bnxt_filter_info, next);
1727 				bnxt_hwrm_clear_l2_filter(bp, filter);
1728 				bnxt_free_filter(bp, filter);
1729 			}
1730 			filter = temp_filter;
1731 		}
1732 	}
1733 }
1734 
1735 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1736 			       struct rte_ether_addr *mac_addr, uint32_t index,
1737 			       uint32_t pool)
1738 {
1739 	struct bnxt_filter_info *filter;
1740 	int rc = 0;
1741 
1742 	/* Attach requested MAC address to the new l2_filter */
1743 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1744 		if (filter->mac_index == index) {
1745 			PMD_DRV_LOG(DEBUG,
1746 				    "MAC addr already existed for pool %d\n",
1747 				    pool);
1748 			return 0;
1749 		}
1750 	}
1751 
1752 	filter = bnxt_alloc_filter(bp);
1753 	if (!filter) {
1754 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1755 		return -ENODEV;
1756 	}
1757 
1758 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1759 	 * if the MAC that's been programmed now is a different one, then,
1760 	 * copy that addr to filter->l2_addr
1761 	 */
1762 	if (mac_addr)
1763 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1764 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1765 
1766 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1767 	if (!rc) {
1768 		filter->mac_index = index;
1769 		if (filter->mac_index == 0)
1770 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1771 		else
1772 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1773 	} else {
1774 		bnxt_free_filter(bp, filter);
1775 	}
1776 
1777 	return rc;
1778 }
1779 
1780 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1781 				struct rte_ether_addr *mac_addr,
1782 				uint32_t index, uint32_t pool)
1783 {
1784 	struct bnxt *bp = eth_dev->data->dev_private;
1785 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1786 	int rc = 0;
1787 
1788 	rc = is_bnxt_in_error(bp);
1789 	if (rc)
1790 		return rc;
1791 
1792 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1793 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1794 		return -ENOTSUP;
1795 	}
1796 
1797 	if (!vnic) {
1798 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1799 		return -EINVAL;
1800 	}
1801 
1802 	/* Filter settings will get applied when port is started */
1803 	if (!eth_dev->data->dev_started)
1804 		return 0;
1805 
1806 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1807 
1808 	return rc;
1809 }
1810 
1811 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1812 {
1813 	int rc = 0;
1814 	struct bnxt *bp = eth_dev->data->dev_private;
1815 	struct rte_eth_link new;
1816 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1817 			BNXT_MIN_LINK_WAIT_CNT;
1818 
1819 	rc = is_bnxt_in_error(bp);
1820 	if (rc)
1821 		return rc;
1822 
1823 	memset(&new, 0, sizeof(new));
1824 
1825 	if (bp->link_info == NULL)
1826 		goto out;
1827 
1828 	do {
1829 		/* Retrieve link info from hardware */
1830 		rc = bnxt_get_hwrm_link_config(bp, &new);
1831 		if (rc) {
1832 			new.link_speed = ETH_LINK_SPEED_100M;
1833 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1834 			PMD_DRV_LOG(ERR,
1835 				"Failed to retrieve link rc = 0x%x!\n", rc);
1836 			goto out;
1837 		}
1838 
1839 		if (!wait_to_complete || new.link_status)
1840 			break;
1841 
1842 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1843 	} while (cnt--);
1844 
1845 	/* Only single function PF can bring phy down.
1846 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1847 	 */
1848 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1849 		memset(&new, 0, sizeof(new));
1850 
1851 out:
1852 	/* Timed out or success */
1853 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1854 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1855 		rte_eth_linkstatus_set(eth_dev, &new);
1856 		bnxt_print_link_info(eth_dev);
1857 	}
1858 
1859 	return rc;
1860 }
1861 
1862 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1863 {
1864 	struct bnxt *bp = eth_dev->data->dev_private;
1865 	struct bnxt_vnic_info *vnic;
1866 	uint32_t old_flags;
1867 	int rc;
1868 
1869 	rc = is_bnxt_in_error(bp);
1870 	if (rc)
1871 		return rc;
1872 
1873 	/* Filter settings will get applied when port is started */
1874 	if (!eth_dev->data->dev_started)
1875 		return 0;
1876 
1877 	if (bp->vnic_info == NULL)
1878 		return 0;
1879 
1880 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1881 
1882 	old_flags = vnic->flags;
1883 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1884 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1885 	if (rc != 0)
1886 		vnic->flags = old_flags;
1887 
1888 	return rc;
1889 }
1890 
1891 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1892 {
1893 	struct bnxt *bp = eth_dev->data->dev_private;
1894 	struct bnxt_vnic_info *vnic;
1895 	uint32_t old_flags;
1896 	int rc;
1897 
1898 	rc = is_bnxt_in_error(bp);
1899 	if (rc)
1900 		return rc;
1901 
1902 	/* Filter settings will get applied when port is started */
1903 	if (!eth_dev->data->dev_started)
1904 		return 0;
1905 
1906 	if (bp->vnic_info == NULL)
1907 		return 0;
1908 
1909 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1910 
1911 	old_flags = vnic->flags;
1912 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1913 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1914 	if (rc != 0)
1915 		vnic->flags = old_flags;
1916 
1917 	return rc;
1918 }
1919 
1920 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1921 {
1922 	struct bnxt *bp = eth_dev->data->dev_private;
1923 	struct bnxt_vnic_info *vnic;
1924 	uint32_t old_flags;
1925 	int rc;
1926 
1927 	rc = is_bnxt_in_error(bp);
1928 	if (rc)
1929 		return rc;
1930 
1931 	/* Filter settings will get applied when port is started */
1932 	if (!eth_dev->data->dev_started)
1933 		return 0;
1934 
1935 	if (bp->vnic_info == NULL)
1936 		return 0;
1937 
1938 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1939 
1940 	old_flags = vnic->flags;
1941 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1942 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1943 	if (rc != 0)
1944 		vnic->flags = old_flags;
1945 
1946 	return rc;
1947 }
1948 
1949 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1950 {
1951 	struct bnxt *bp = eth_dev->data->dev_private;
1952 	struct bnxt_vnic_info *vnic;
1953 	uint32_t old_flags;
1954 	int rc;
1955 
1956 	rc = is_bnxt_in_error(bp);
1957 	if (rc)
1958 		return rc;
1959 
1960 	/* Filter settings will get applied when port is started */
1961 	if (!eth_dev->data->dev_started)
1962 		return 0;
1963 
1964 	if (bp->vnic_info == NULL)
1965 		return 0;
1966 
1967 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1968 
1969 	old_flags = vnic->flags;
1970 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1971 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1972 	if (rc != 0)
1973 		vnic->flags = old_flags;
1974 
1975 	return rc;
1976 }
1977 
1978 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1979 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1980 {
1981 	if (qid >= bp->rx_nr_rings)
1982 		return NULL;
1983 
1984 	return bp->eth_dev->data->rx_queues[qid];
1985 }
1986 
1987 /* Return rxq corresponding to a given rss table ring/group ID. */
1988 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1989 {
1990 	struct bnxt_rx_queue *rxq;
1991 	unsigned int i;
1992 
1993 	if (!BNXT_HAS_RING_GRPS(bp)) {
1994 		for (i = 0; i < bp->rx_nr_rings; i++) {
1995 			rxq = bp->eth_dev->data->rx_queues[i];
1996 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1997 				return rxq->index;
1998 		}
1999 	} else {
2000 		for (i = 0; i < bp->rx_nr_rings; i++) {
2001 			if (bp->grp_info[i].fw_grp_id == fwr)
2002 				return i;
2003 		}
2004 	}
2005 
2006 	return INVALID_HW_RING_ID;
2007 }
2008 
2009 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
2010 			    struct rte_eth_rss_reta_entry64 *reta_conf,
2011 			    uint16_t reta_size)
2012 {
2013 	struct bnxt *bp = eth_dev->data->dev_private;
2014 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2015 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2016 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2017 	uint16_t idx, sft;
2018 	int i, rc;
2019 
2020 	rc = is_bnxt_in_error(bp);
2021 	if (rc)
2022 		return rc;
2023 
2024 	if (!vnic->rss_table)
2025 		return -EINVAL;
2026 
2027 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
2028 		return -EINVAL;
2029 
2030 	if (reta_size != tbl_size) {
2031 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2032 			"(%d) must equal the size supported by the hardware "
2033 			"(%d)\n", reta_size, tbl_size);
2034 		return -EINVAL;
2035 	}
2036 
2037 	for (i = 0; i < reta_size; i++) {
2038 		struct bnxt_rx_queue *rxq;
2039 
2040 		idx = i / RTE_RETA_GROUP_SIZE;
2041 		sft = i % RTE_RETA_GROUP_SIZE;
2042 
2043 		if (!(reta_conf[idx].mask & (1ULL << sft)))
2044 			continue;
2045 
2046 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2047 		if (!rxq) {
2048 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
2049 			return -EINVAL;
2050 		}
2051 
2052 		if (BNXT_CHIP_P5(bp)) {
2053 			vnic->rss_table[i * 2] =
2054 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
2055 			vnic->rss_table[i * 2 + 1] =
2056 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
2057 		} else {
2058 			vnic->rss_table[i] =
2059 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2060 		}
2061 	}
2062 
2063 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2064 	return rc;
2065 }
2066 
2067 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2068 			      struct rte_eth_rss_reta_entry64 *reta_conf,
2069 			      uint16_t reta_size)
2070 {
2071 	struct bnxt *bp = eth_dev->data->dev_private;
2072 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2073 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2074 	uint16_t idx, sft, i;
2075 	int rc;
2076 
2077 	rc = is_bnxt_in_error(bp);
2078 	if (rc)
2079 		return rc;
2080 
2081 	if (!vnic)
2082 		return -EINVAL;
2083 	if (!vnic->rss_table)
2084 		return -EINVAL;
2085 
2086 	if (reta_size != tbl_size) {
2087 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2088 			"(%d) must equal the size supported by the hardware "
2089 			"(%d)\n", reta_size, tbl_size);
2090 		return -EINVAL;
2091 	}
2092 
2093 	for (idx = 0, i = 0; i < reta_size; i++) {
2094 		idx = i / RTE_RETA_GROUP_SIZE;
2095 		sft = i % RTE_RETA_GROUP_SIZE;
2096 
2097 		if (reta_conf[idx].mask & (1ULL << sft)) {
2098 			uint16_t qid;
2099 
2100 			if (BNXT_CHIP_P5(bp))
2101 				qid = bnxt_rss_to_qid(bp,
2102 						      vnic->rss_table[i * 2]);
2103 			else
2104 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2105 
2106 			if (qid == INVALID_HW_RING_ID) {
2107 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2108 				return -EINVAL;
2109 			}
2110 			reta_conf[idx].reta[sft] = qid;
2111 		}
2112 	}
2113 
2114 	return 0;
2115 }
2116 
2117 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2118 				   struct rte_eth_rss_conf *rss_conf)
2119 {
2120 	struct bnxt *bp = eth_dev->data->dev_private;
2121 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2122 	struct bnxt_vnic_info *vnic;
2123 	int rc;
2124 
2125 	rc = is_bnxt_in_error(bp);
2126 	if (rc)
2127 		return rc;
2128 
2129 	/*
2130 	 * If RSS enablement were different than dev_configure,
2131 	 * then return -EINVAL
2132 	 */
2133 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
2134 		if (!rss_conf->rss_hf)
2135 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2136 	} else {
2137 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2138 			return -EINVAL;
2139 	}
2140 
2141 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
2142 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2143 	       rss_conf,
2144 	       sizeof(*rss_conf));
2145 
2146 	/* Update the default RSS VNIC(s) */
2147 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2148 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2149 	vnic->hash_mode =
2150 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2151 					    ETH_RSS_LEVEL(rss_conf->rss_hf));
2152 
2153 	/*
2154 	 * If hashkey is not specified, use the previously configured
2155 	 * hashkey
2156 	 */
2157 	if (!rss_conf->rss_key)
2158 		goto rss_config;
2159 
2160 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2161 		PMD_DRV_LOG(ERR,
2162 			    "Invalid hashkey length, should be %d bytes\n",
2163 			    HW_HASH_KEY_SIZE);
2164 		return -EINVAL;
2165 	}
2166 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2167 
2168 rss_config:
2169 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2170 	return rc;
2171 }
2172 
2173 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2174 				     struct rte_eth_rss_conf *rss_conf)
2175 {
2176 	struct bnxt *bp = eth_dev->data->dev_private;
2177 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2178 	int len, rc;
2179 	uint32_t hash_types;
2180 
2181 	rc = is_bnxt_in_error(bp);
2182 	if (rc)
2183 		return rc;
2184 
2185 	/* RSS configuration is the same for all VNICs */
2186 	if (vnic && vnic->rss_hash_key) {
2187 		if (rss_conf->rss_key) {
2188 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2189 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2190 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2191 		}
2192 
2193 		hash_types = vnic->hash_type;
2194 		rss_conf->rss_hf = 0;
2195 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2196 			rss_conf->rss_hf |= ETH_RSS_IPV4;
2197 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2198 		}
2199 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2200 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2201 			hash_types &=
2202 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2203 		}
2204 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2205 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2206 			hash_types &=
2207 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2208 		}
2209 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2210 			rss_conf->rss_hf |= ETH_RSS_IPV6;
2211 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2212 		}
2213 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2214 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2215 			hash_types &=
2216 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2217 		}
2218 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2219 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2220 			hash_types &=
2221 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2222 		}
2223 
2224 		rss_conf->rss_hf |=
2225 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2226 
2227 		if (hash_types) {
2228 			PMD_DRV_LOG(ERR,
2229 				"Unknown RSS config from firmware (%08x), RSS disabled",
2230 				vnic->hash_type);
2231 			return -ENOTSUP;
2232 		}
2233 	} else {
2234 		rss_conf->rss_hf = 0;
2235 	}
2236 	return 0;
2237 }
2238 
2239 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2240 			       struct rte_eth_fc_conf *fc_conf)
2241 {
2242 	struct bnxt *bp = dev->data->dev_private;
2243 	struct rte_eth_link link_info;
2244 	int rc;
2245 
2246 	rc = is_bnxt_in_error(bp);
2247 	if (rc)
2248 		return rc;
2249 
2250 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2251 	if (rc)
2252 		return rc;
2253 
2254 	memset(fc_conf, 0, sizeof(*fc_conf));
2255 	if (bp->link_info->auto_pause)
2256 		fc_conf->autoneg = 1;
2257 	switch (bp->link_info->pause) {
2258 	case 0:
2259 		fc_conf->mode = RTE_FC_NONE;
2260 		break;
2261 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2262 		fc_conf->mode = RTE_FC_TX_PAUSE;
2263 		break;
2264 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2265 		fc_conf->mode = RTE_FC_RX_PAUSE;
2266 		break;
2267 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2268 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2269 		fc_conf->mode = RTE_FC_FULL;
2270 		break;
2271 	}
2272 	return 0;
2273 }
2274 
2275 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2276 			       struct rte_eth_fc_conf *fc_conf)
2277 {
2278 	struct bnxt *bp = dev->data->dev_private;
2279 	int rc;
2280 
2281 	rc = is_bnxt_in_error(bp);
2282 	if (rc)
2283 		return rc;
2284 
2285 	if (!BNXT_SINGLE_PF(bp)) {
2286 		PMD_DRV_LOG(ERR,
2287 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2288 		return -ENOTSUP;
2289 	}
2290 
2291 	switch (fc_conf->mode) {
2292 	case RTE_FC_NONE:
2293 		bp->link_info->auto_pause = 0;
2294 		bp->link_info->force_pause = 0;
2295 		break;
2296 	case RTE_FC_RX_PAUSE:
2297 		if (fc_conf->autoneg) {
2298 			bp->link_info->auto_pause =
2299 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2300 			bp->link_info->force_pause = 0;
2301 		} else {
2302 			bp->link_info->auto_pause = 0;
2303 			bp->link_info->force_pause =
2304 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2305 		}
2306 		break;
2307 	case RTE_FC_TX_PAUSE:
2308 		if (fc_conf->autoneg) {
2309 			bp->link_info->auto_pause =
2310 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2311 			bp->link_info->force_pause = 0;
2312 		} else {
2313 			bp->link_info->auto_pause = 0;
2314 			bp->link_info->force_pause =
2315 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2316 		}
2317 		break;
2318 	case RTE_FC_FULL:
2319 		if (fc_conf->autoneg) {
2320 			bp->link_info->auto_pause =
2321 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2322 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2323 			bp->link_info->force_pause = 0;
2324 		} else {
2325 			bp->link_info->auto_pause = 0;
2326 			bp->link_info->force_pause =
2327 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2328 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2329 		}
2330 		break;
2331 	}
2332 	return bnxt_set_hwrm_link_config(bp, true);
2333 }
2334 
2335 /* Add UDP tunneling port */
2336 static int
2337 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2338 			 struct rte_eth_udp_tunnel *udp_tunnel)
2339 {
2340 	struct bnxt *bp = eth_dev->data->dev_private;
2341 	uint16_t tunnel_type = 0;
2342 	int rc = 0;
2343 
2344 	rc = is_bnxt_in_error(bp);
2345 	if (rc)
2346 		return rc;
2347 
2348 	switch (udp_tunnel->prot_type) {
2349 	case RTE_TUNNEL_TYPE_VXLAN:
2350 		if (bp->vxlan_port_cnt) {
2351 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2352 				udp_tunnel->udp_port);
2353 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2354 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2355 				return -ENOSPC;
2356 			}
2357 			bp->vxlan_port_cnt++;
2358 			return 0;
2359 		}
2360 		tunnel_type =
2361 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2362 		bp->vxlan_port_cnt++;
2363 		break;
2364 	case RTE_TUNNEL_TYPE_GENEVE:
2365 		if (bp->geneve_port_cnt) {
2366 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2367 				udp_tunnel->udp_port);
2368 			if (bp->geneve_port != udp_tunnel->udp_port) {
2369 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2370 				return -ENOSPC;
2371 			}
2372 			bp->geneve_port_cnt++;
2373 			return 0;
2374 		}
2375 		tunnel_type =
2376 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2377 		bp->geneve_port_cnt++;
2378 		break;
2379 	default:
2380 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2381 		return -ENOTSUP;
2382 	}
2383 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2384 					     tunnel_type);
2385 	return rc;
2386 }
2387 
2388 static int
2389 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2390 			 struct rte_eth_udp_tunnel *udp_tunnel)
2391 {
2392 	struct bnxt *bp = eth_dev->data->dev_private;
2393 	uint16_t tunnel_type = 0;
2394 	uint16_t port = 0;
2395 	int rc = 0;
2396 
2397 	rc = is_bnxt_in_error(bp);
2398 	if (rc)
2399 		return rc;
2400 
2401 	switch (udp_tunnel->prot_type) {
2402 	case RTE_TUNNEL_TYPE_VXLAN:
2403 		if (!bp->vxlan_port_cnt) {
2404 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2405 			return -EINVAL;
2406 		}
2407 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2408 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2409 				udp_tunnel->udp_port, bp->vxlan_port);
2410 			return -EINVAL;
2411 		}
2412 		if (--bp->vxlan_port_cnt)
2413 			return 0;
2414 
2415 		tunnel_type =
2416 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2417 		port = bp->vxlan_fw_dst_port_id;
2418 		break;
2419 	case RTE_TUNNEL_TYPE_GENEVE:
2420 		if (!bp->geneve_port_cnt) {
2421 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2422 			return -EINVAL;
2423 		}
2424 		if (bp->geneve_port != udp_tunnel->udp_port) {
2425 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2426 				udp_tunnel->udp_port, bp->geneve_port);
2427 			return -EINVAL;
2428 		}
2429 		if (--bp->geneve_port_cnt)
2430 			return 0;
2431 
2432 		tunnel_type =
2433 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2434 		port = bp->geneve_fw_dst_port_id;
2435 		break;
2436 	default:
2437 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2438 		return -ENOTSUP;
2439 	}
2440 
2441 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2442 	return rc;
2443 }
2444 
2445 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2446 {
2447 	struct bnxt_filter_info *filter;
2448 	struct bnxt_vnic_info *vnic;
2449 	int rc = 0;
2450 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2451 
2452 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2453 	filter = STAILQ_FIRST(&vnic->filter);
2454 	while (filter) {
2455 		/* Search for this matching MAC+VLAN filter */
2456 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2457 			/* Delete the filter */
2458 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2459 			if (rc)
2460 				return rc;
2461 			STAILQ_REMOVE(&vnic->filter, filter,
2462 				      bnxt_filter_info, next);
2463 			bnxt_free_filter(bp, filter);
2464 			PMD_DRV_LOG(INFO,
2465 				    "Deleted vlan filter for %d\n",
2466 				    vlan_id);
2467 			return 0;
2468 		}
2469 		filter = STAILQ_NEXT(filter, next);
2470 	}
2471 	return -ENOENT;
2472 }
2473 
2474 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2475 {
2476 	struct bnxt_filter_info *filter;
2477 	struct bnxt_vnic_info *vnic;
2478 	int rc = 0;
2479 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2480 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2481 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2482 
2483 	/* Implementation notes on the use of VNIC in this command:
2484 	 *
2485 	 * By default, these filters belong to default vnic for the function.
2486 	 * Once these filters are set up, only destination VNIC can be modified.
2487 	 * If the destination VNIC is not specified in this command,
2488 	 * then the HWRM shall only create an l2 context id.
2489 	 */
2490 
2491 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2492 	filter = STAILQ_FIRST(&vnic->filter);
2493 	/* Check if the VLAN has already been added */
2494 	while (filter) {
2495 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2496 			return -EEXIST;
2497 
2498 		filter = STAILQ_NEXT(filter, next);
2499 	}
2500 
2501 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2502 	 * command to create MAC+VLAN filter with the right flags, enables set.
2503 	 */
2504 	filter = bnxt_alloc_filter(bp);
2505 	if (!filter) {
2506 		PMD_DRV_LOG(ERR,
2507 			    "MAC/VLAN filter alloc failed\n");
2508 		return -ENOMEM;
2509 	}
2510 	/* MAC + VLAN ID filter */
2511 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2512 	 * untagged packets are received
2513 	 *
2514 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2515 	 * packets and only the programmed vlan's packets are received
2516 	 */
2517 	filter->l2_ivlan = vlan_id;
2518 	filter->l2_ivlan_mask = 0x0FFF;
2519 	filter->enables |= en;
2520 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2521 
2522 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2523 	if (rc) {
2524 		/* Free the newly allocated filter as we were
2525 		 * not able to create the filter in hardware.
2526 		 */
2527 		bnxt_free_filter(bp, filter);
2528 		return rc;
2529 	}
2530 
2531 	filter->mac_index = 0;
2532 	/* Add this new filter to the list */
2533 	if (vlan_id == 0)
2534 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2535 	else
2536 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2537 
2538 	PMD_DRV_LOG(INFO,
2539 		    "Added Vlan filter for %d\n", vlan_id);
2540 	return rc;
2541 }
2542 
2543 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2544 		uint16_t vlan_id, int on)
2545 {
2546 	struct bnxt *bp = eth_dev->data->dev_private;
2547 	int rc;
2548 
2549 	rc = is_bnxt_in_error(bp);
2550 	if (rc)
2551 		return rc;
2552 
2553 	if (!eth_dev->data->dev_started) {
2554 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2555 		return -EINVAL;
2556 	}
2557 
2558 	/* These operations apply to ALL existing MAC/VLAN filters */
2559 	if (on)
2560 		return bnxt_add_vlan_filter(bp, vlan_id);
2561 	else
2562 		return bnxt_del_vlan_filter(bp, vlan_id);
2563 }
2564 
2565 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2566 				    struct bnxt_vnic_info *vnic)
2567 {
2568 	struct bnxt_filter_info *filter;
2569 	int rc;
2570 
2571 	filter = STAILQ_FIRST(&vnic->filter);
2572 	while (filter) {
2573 		if (filter->mac_index == 0 &&
2574 		    !memcmp(filter->l2_addr, bp->mac_addr,
2575 			    RTE_ETHER_ADDR_LEN)) {
2576 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2577 			if (!rc) {
2578 				STAILQ_REMOVE(&vnic->filter, filter,
2579 					      bnxt_filter_info, next);
2580 				bnxt_free_filter(bp, filter);
2581 			}
2582 			return rc;
2583 		}
2584 		filter = STAILQ_NEXT(filter, next);
2585 	}
2586 	return 0;
2587 }
2588 
2589 static int
2590 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2591 {
2592 	struct bnxt_vnic_info *vnic;
2593 	unsigned int i;
2594 	int rc;
2595 
2596 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2597 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2598 		/* Remove any VLAN filters programmed */
2599 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2600 			bnxt_del_vlan_filter(bp, i);
2601 
2602 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2603 		if (rc)
2604 			return rc;
2605 	} else {
2606 		/* Default filter will allow packets that match the
2607 		 * dest mac. So, it has to be deleted, otherwise, we
2608 		 * will endup receiving vlan packets for which the
2609 		 * filter is not programmed, when hw-vlan-filter
2610 		 * configuration is ON
2611 		 */
2612 		bnxt_del_dflt_mac_filter(bp, vnic);
2613 		/* This filter will allow only untagged packets */
2614 		bnxt_add_vlan_filter(bp, 0);
2615 	}
2616 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2617 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2618 
2619 	return 0;
2620 }
2621 
2622 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2623 {
2624 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2625 	unsigned int i;
2626 	int rc;
2627 
2628 	/* Destroy vnic filters and vnic */
2629 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2630 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2631 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2632 			bnxt_del_vlan_filter(bp, i);
2633 	}
2634 	bnxt_del_dflt_mac_filter(bp, vnic);
2635 
2636 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2637 	if (rc)
2638 		return rc;
2639 
2640 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2641 	if (rc)
2642 		return rc;
2643 
2644 	rte_free(vnic->fw_grp_ids);
2645 	vnic->fw_grp_ids = NULL;
2646 
2647 	vnic->rx_queue_cnt = 0;
2648 
2649 	return 0;
2650 }
2651 
2652 static int
2653 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2654 {
2655 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2656 	int rc;
2657 
2658 	/* Destroy, recreate and reconfigure the default vnic */
2659 	rc = bnxt_free_one_vnic(bp, 0);
2660 	if (rc)
2661 		return rc;
2662 
2663 	/* default vnic 0 */
2664 	rc = bnxt_setup_one_vnic(bp, 0);
2665 	if (rc)
2666 		return rc;
2667 
2668 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2669 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2670 		rc = bnxt_add_vlan_filter(bp, 0);
2671 		if (rc)
2672 			return rc;
2673 		rc = bnxt_restore_vlan_filters(bp);
2674 		if (rc)
2675 			return rc;
2676 	} else {
2677 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2678 		if (rc)
2679 			return rc;
2680 	}
2681 
2682 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2683 	if (rc)
2684 		return rc;
2685 
2686 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2687 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2688 
2689 	return rc;
2690 }
2691 
2692 static int
2693 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2694 {
2695 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2696 	struct bnxt *bp = dev->data->dev_private;
2697 	int rc;
2698 
2699 	rc = is_bnxt_in_error(bp);
2700 	if (rc)
2701 		return rc;
2702 
2703 	/* Filter settings will get applied when port is started */
2704 	if (!dev->data->dev_started)
2705 		return 0;
2706 
2707 	if (mask & ETH_VLAN_FILTER_MASK) {
2708 		/* Enable or disable VLAN filtering */
2709 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2710 		if (rc)
2711 			return rc;
2712 	}
2713 
2714 	if (mask & ETH_VLAN_STRIP_MASK) {
2715 		/* Enable or disable VLAN stripping */
2716 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2717 		if (rc)
2718 			return rc;
2719 	}
2720 
2721 	if (mask & ETH_VLAN_EXTEND_MASK) {
2722 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2723 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2724 		else
2725 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2726 	}
2727 
2728 	return 0;
2729 }
2730 
2731 static int
2732 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2733 		      uint16_t tpid)
2734 {
2735 	struct bnxt *bp = dev->data->dev_private;
2736 	int qinq = dev->data->dev_conf.rxmode.offloads &
2737 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2738 
2739 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2740 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2741 		PMD_DRV_LOG(ERR,
2742 			    "Unsupported vlan type.");
2743 		return -EINVAL;
2744 	}
2745 	if (!qinq) {
2746 		PMD_DRV_LOG(ERR,
2747 			    "QinQ not enabled. Needs to be ON as we can "
2748 			    "accelerate only outer vlan\n");
2749 		return -EINVAL;
2750 	}
2751 
2752 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2753 		switch (tpid) {
2754 		case RTE_ETHER_TYPE_QINQ:
2755 			bp->outer_tpid_bd =
2756 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2757 				break;
2758 		case RTE_ETHER_TYPE_VLAN:
2759 			bp->outer_tpid_bd =
2760 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2761 				break;
2762 		case RTE_ETHER_TYPE_QINQ1:
2763 			bp->outer_tpid_bd =
2764 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2765 				break;
2766 		case RTE_ETHER_TYPE_QINQ2:
2767 			bp->outer_tpid_bd =
2768 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2769 				break;
2770 		case RTE_ETHER_TYPE_QINQ3:
2771 			bp->outer_tpid_bd =
2772 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2773 				break;
2774 		default:
2775 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2776 			return -EINVAL;
2777 		}
2778 		bp->outer_tpid_bd |= tpid;
2779 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2780 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2781 		PMD_DRV_LOG(ERR,
2782 			    "Can accelerate only outer vlan in QinQ\n");
2783 		return -EINVAL;
2784 	}
2785 
2786 	return 0;
2787 }
2788 
2789 static int
2790 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2791 			     struct rte_ether_addr *addr)
2792 {
2793 	struct bnxt *bp = dev->data->dev_private;
2794 	/* Default Filter is tied to VNIC 0 */
2795 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2796 	int rc;
2797 
2798 	rc = is_bnxt_in_error(bp);
2799 	if (rc)
2800 		return rc;
2801 
2802 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2803 		return -EPERM;
2804 
2805 	if (rte_is_zero_ether_addr(addr))
2806 		return -EINVAL;
2807 
2808 	/* Filter settings will get applied when port is started */
2809 	if (!dev->data->dev_started)
2810 		return 0;
2811 
2812 	/* Check if the requested MAC is already added */
2813 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2814 		return 0;
2815 
2816 	/* Destroy filter and re-create it */
2817 	bnxt_del_dflt_mac_filter(bp, vnic);
2818 
2819 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2820 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2821 		/* This filter will allow only untagged packets */
2822 		rc = bnxt_add_vlan_filter(bp, 0);
2823 	} else {
2824 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2825 	}
2826 
2827 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2828 	return rc;
2829 }
2830 
2831 static int
2832 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2833 			  struct rte_ether_addr *mc_addr_set,
2834 			  uint32_t nb_mc_addr)
2835 {
2836 	struct bnxt *bp = eth_dev->data->dev_private;
2837 	char *mc_addr_list = (char *)mc_addr_set;
2838 	struct bnxt_vnic_info *vnic;
2839 	uint32_t off = 0, i = 0;
2840 	int rc;
2841 
2842 	rc = is_bnxt_in_error(bp);
2843 	if (rc)
2844 		return rc;
2845 
2846 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2847 
2848 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2849 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2850 		goto allmulti;
2851 	}
2852 
2853 	/* TODO Check for Duplicate mcast addresses */
2854 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2855 	for (i = 0; i < nb_mc_addr; i++) {
2856 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2857 			RTE_ETHER_ADDR_LEN);
2858 		off += RTE_ETHER_ADDR_LEN;
2859 	}
2860 
2861 	vnic->mc_addr_cnt = i;
2862 	if (vnic->mc_addr_cnt)
2863 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2864 	else
2865 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2866 
2867 allmulti:
2868 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2869 }
2870 
2871 static int
2872 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2873 {
2874 	struct bnxt *bp = dev->data->dev_private;
2875 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2876 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2877 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2878 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2879 	int ret;
2880 
2881 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2882 			fw_major, fw_minor, fw_updt, fw_rsvd);
2883 	if (ret < 0)
2884 		return -EINVAL;
2885 
2886 	ret += 1; /* add the size of '\0' */
2887 	if (fw_size < (size_t)ret)
2888 		return ret;
2889 	else
2890 		return 0;
2891 }
2892 
2893 static void
2894 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2895 	struct rte_eth_rxq_info *qinfo)
2896 {
2897 	struct bnxt *bp = dev->data->dev_private;
2898 	struct bnxt_rx_queue *rxq;
2899 
2900 	if (is_bnxt_in_error(bp))
2901 		return;
2902 
2903 	rxq = dev->data->rx_queues[queue_id];
2904 
2905 	qinfo->mp = rxq->mb_pool;
2906 	qinfo->scattered_rx = dev->data->scattered_rx;
2907 	qinfo->nb_desc = rxq->nb_rx_desc;
2908 
2909 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2910 	qinfo->conf.rx_drop_en = rxq->drop_en;
2911 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2912 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2913 }
2914 
2915 static void
2916 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2917 	struct rte_eth_txq_info *qinfo)
2918 {
2919 	struct bnxt *bp = dev->data->dev_private;
2920 	struct bnxt_tx_queue *txq;
2921 
2922 	if (is_bnxt_in_error(bp))
2923 		return;
2924 
2925 	txq = dev->data->tx_queues[queue_id];
2926 
2927 	qinfo->nb_desc = txq->nb_tx_desc;
2928 
2929 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2930 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2931 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2932 
2933 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2934 	qinfo->conf.tx_rs_thresh = 0;
2935 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2936 	qinfo->conf.offloads = txq->offloads;
2937 }
2938 
2939 static const struct {
2940 	eth_rx_burst_t pkt_burst;
2941 	const char *info;
2942 } bnxt_rx_burst_info[] = {
2943 	{bnxt_recv_pkts,		"Scalar"},
2944 #if defined(RTE_ARCH_X86)
2945 	{bnxt_recv_pkts_vec,		"Vector SSE"},
2946 #endif
2947 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2948 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
2949 #endif
2950 #if defined(RTE_ARCH_ARM64)
2951 	{bnxt_recv_pkts_vec,		"Vector Neon"},
2952 #endif
2953 };
2954 
2955 static int
2956 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2957 		       struct rte_eth_burst_mode *mode)
2958 {
2959 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2960 	size_t i;
2961 
2962 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2963 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2964 			snprintf(mode->info, sizeof(mode->info), "%s",
2965 				 bnxt_rx_burst_info[i].info);
2966 			return 0;
2967 		}
2968 	}
2969 
2970 	return -EINVAL;
2971 }
2972 
2973 static const struct {
2974 	eth_tx_burst_t pkt_burst;
2975 	const char *info;
2976 } bnxt_tx_burst_info[] = {
2977 	{bnxt_xmit_pkts,		"Scalar"},
2978 #if defined(RTE_ARCH_X86)
2979 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
2980 #endif
2981 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2982 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
2983 #endif
2984 #if defined(RTE_ARCH_ARM64)
2985 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
2986 #endif
2987 };
2988 
2989 static int
2990 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2991 		       struct rte_eth_burst_mode *mode)
2992 {
2993 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2994 	size_t i;
2995 
2996 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2997 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2998 			snprintf(mode->info, sizeof(mode->info), "%s",
2999 				 bnxt_tx_burst_info[i].info);
3000 			return 0;
3001 		}
3002 	}
3003 
3004 	return -EINVAL;
3005 }
3006 
3007 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3008 {
3009 	struct bnxt *bp = eth_dev->data->dev_private;
3010 	uint32_t new_pkt_size;
3011 	uint32_t rc = 0;
3012 	uint32_t i;
3013 
3014 	rc = is_bnxt_in_error(bp);
3015 	if (rc)
3016 		return rc;
3017 
3018 	/* Exit if receive queues are not configured yet */
3019 	if (!eth_dev->data->nb_rx_queues)
3020 		return rc;
3021 
3022 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
3023 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
3024 
3025 	/*
3026 	 * Disallow any MTU change that would require scattered receive support
3027 	 * if it is not already enabled.
3028 	 */
3029 	if (eth_dev->data->dev_started &&
3030 	    !eth_dev->data->scattered_rx &&
3031 	    (new_pkt_size >
3032 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
3033 		PMD_DRV_LOG(ERR,
3034 			    "MTU change would require scattered rx support. ");
3035 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
3036 		return -EINVAL;
3037 	}
3038 
3039 	if (new_mtu > RTE_ETHER_MTU) {
3040 		bp->flags |= BNXT_FLAG_JUMBO;
3041 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
3042 			DEV_RX_OFFLOAD_JUMBO_FRAME;
3043 	} else {
3044 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
3045 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
3046 		bp->flags &= ~BNXT_FLAG_JUMBO;
3047 	}
3048 
3049 	/* Is there a change in mtu setting? */
3050 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
3051 		return rc;
3052 
3053 	for (i = 0; i < bp->nr_vnics; i++) {
3054 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3055 		uint16_t size = 0;
3056 
3057 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
3058 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
3059 		if (rc)
3060 			break;
3061 
3062 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
3063 		size -= RTE_PKTMBUF_HEADROOM;
3064 
3065 		if (size < new_mtu) {
3066 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
3067 			if (rc)
3068 				return rc;
3069 		}
3070 	}
3071 
3072 	if (!rc)
3073 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
3074 
3075 	if (bnxt_hwrm_config_host_mtu(bp))
3076 		PMD_DRV_LOG(WARNING, "Failed to configure host MTU\n");
3077 
3078 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3079 
3080 	return rc;
3081 }
3082 
3083 static int
3084 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3085 {
3086 	struct bnxt *bp = dev->data->dev_private;
3087 	uint16_t vlan = bp->vlan;
3088 	int rc;
3089 
3090 	rc = is_bnxt_in_error(bp);
3091 	if (rc)
3092 		return rc;
3093 
3094 	if (!BNXT_SINGLE_PF(bp)) {
3095 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3096 		return -ENOTSUP;
3097 	}
3098 	bp->vlan = on ? pvid : 0;
3099 
3100 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3101 	if (rc)
3102 		bp->vlan = vlan;
3103 	return rc;
3104 }
3105 
3106 static int
3107 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3108 {
3109 	struct bnxt *bp = dev->data->dev_private;
3110 	int rc;
3111 
3112 	rc = is_bnxt_in_error(bp);
3113 	if (rc)
3114 		return rc;
3115 
3116 	return bnxt_hwrm_port_led_cfg(bp, true);
3117 }
3118 
3119 static int
3120 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3121 {
3122 	struct bnxt *bp = dev->data->dev_private;
3123 	int rc;
3124 
3125 	rc = is_bnxt_in_error(bp);
3126 	if (rc)
3127 		return rc;
3128 
3129 	return bnxt_hwrm_port_led_cfg(bp, false);
3130 }
3131 
3132 static uint32_t
3133 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
3134 {
3135 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
3136 	struct bnxt_cp_ring_info *cpr;
3137 	uint32_t desc = 0, raw_cons, cp_ring_size;
3138 	struct bnxt_rx_queue *rxq;
3139 	struct rx_pkt_cmpl *rxcmp;
3140 	int rc;
3141 
3142 	rc = is_bnxt_in_error(bp);
3143 	if (rc)
3144 		return rc;
3145 
3146 	rxq = dev->data->rx_queues[rx_queue_id];
3147 	cpr = rxq->cp_ring;
3148 	raw_cons = cpr->cp_raw_cons;
3149 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3150 
3151 	while (1) {
3152 		uint32_t agg_cnt, cons, cmpl_type;
3153 
3154 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3155 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3156 
3157 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3158 			break;
3159 
3160 		cmpl_type = CMP_TYPE(rxcmp);
3161 
3162 		switch (cmpl_type) {
3163 		case CMPL_BASE_TYPE_RX_L2:
3164 		case CMPL_BASE_TYPE_RX_L2_V2:
3165 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3166 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3167 			desc++;
3168 			break;
3169 
3170 		case CMPL_BASE_TYPE_RX_TPA_END:
3171 			if (BNXT_CHIP_P5(rxq->bp)) {
3172 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3173 
3174 				p5_tpa_end = (void *)rxcmp;
3175 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3176 			} else {
3177 				struct rx_tpa_end_cmpl *tpa_end;
3178 
3179 				tpa_end = (void *)rxcmp;
3180 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3181 			}
3182 
3183 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3184 			desc++;
3185 			break;
3186 
3187 		default:
3188 			raw_cons += CMP_LEN(cmpl_type);
3189 		}
3190 	}
3191 
3192 	return desc;
3193 }
3194 
3195 static int
3196 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3197 {
3198 	struct bnxt_rx_queue *rxq = rx_queue;
3199 	struct bnxt_cp_ring_info *cpr;
3200 	struct bnxt_rx_ring_info *rxr;
3201 	uint32_t desc, raw_cons, cp_ring_size;
3202 	struct bnxt *bp = rxq->bp;
3203 	struct rx_pkt_cmpl *rxcmp;
3204 	int rc;
3205 
3206 	rc = is_bnxt_in_error(bp);
3207 	if (rc)
3208 		return rc;
3209 
3210 	if (offset >= rxq->nb_rx_desc)
3211 		return -EINVAL;
3212 
3213 	rxr = rxq->rx_ring;
3214 	cpr = rxq->cp_ring;
3215 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3216 
3217 	/*
3218 	 * For the vector receive case, the completion at the requested
3219 	 * offset can be indexed directly.
3220 	 */
3221 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3222 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3223 		struct rx_pkt_cmpl *rxcmp;
3224 		uint32_t cons;
3225 
3226 		/* Check status of completion descriptor. */
3227 		raw_cons = cpr->cp_raw_cons +
3228 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3229 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3230 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3231 
3232 		if (bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3233 			return RTE_ETH_RX_DESC_DONE;
3234 
3235 		/* Check whether rx desc has an mbuf attached. */
3236 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3237 		if (cons >= rxq->rxrearm_start &&
3238 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3239 			return RTE_ETH_RX_DESC_UNAVAIL;
3240 		}
3241 
3242 		return RTE_ETH_RX_DESC_AVAIL;
3243 	}
3244 #endif
3245 
3246 	/*
3247 	 * For the non-vector receive case, scan the completion ring to
3248 	 * locate the completion descriptor for the requested offset.
3249 	 */
3250 	raw_cons = cpr->cp_raw_cons;
3251 	desc = 0;
3252 	while (1) {
3253 		uint32_t agg_cnt, cons, cmpl_type;
3254 
3255 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3256 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3257 
3258 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3259 			break;
3260 
3261 		cmpl_type = CMP_TYPE(rxcmp);
3262 
3263 		switch (cmpl_type) {
3264 		case CMPL_BASE_TYPE_RX_L2:
3265 		case CMPL_BASE_TYPE_RX_L2_V2:
3266 			if (desc == offset) {
3267 				cons = rxcmp->opaque;
3268 				if (rxr->rx_buf_ring[cons])
3269 					return RTE_ETH_RX_DESC_DONE;
3270 				else
3271 					return RTE_ETH_RX_DESC_UNAVAIL;
3272 			}
3273 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3274 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3275 			desc++;
3276 			break;
3277 
3278 		case CMPL_BASE_TYPE_RX_TPA_END:
3279 			if (desc == offset)
3280 				return RTE_ETH_RX_DESC_DONE;
3281 
3282 			if (BNXT_CHIP_P5(rxq->bp)) {
3283 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3284 
3285 				p5_tpa_end = (void *)rxcmp;
3286 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3287 			} else {
3288 				struct rx_tpa_end_cmpl *tpa_end;
3289 
3290 				tpa_end = (void *)rxcmp;
3291 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3292 			}
3293 
3294 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3295 			desc++;
3296 			break;
3297 
3298 		default:
3299 			raw_cons += CMP_LEN(cmpl_type);
3300 		}
3301 	}
3302 
3303 	return RTE_ETH_RX_DESC_AVAIL;
3304 }
3305 
3306 static int
3307 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3308 {
3309 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3310 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3311 	uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3312 	struct cmpl_base *cp_desc_ring;
3313 	int rc;
3314 
3315 	rc = is_bnxt_in_error(txq->bp);
3316 	if (rc)
3317 		return rc;
3318 
3319 	if (offset >= txq->nb_tx_desc)
3320 		return -EINVAL;
3321 
3322 	/* Return "desc done" if descriptor is available for use. */
3323 	if (bnxt_tx_bds_in_hw(txq) <= offset)
3324 		return RTE_ETH_TX_DESC_DONE;
3325 
3326 	raw_cons = cpr->cp_raw_cons;
3327 	cp_desc_ring = cpr->cp_desc_ring;
3328 	ring_mask = cpr->cp_ring_struct->ring_mask;
3329 
3330 	/* Check to see if hw has posted a completion for the descriptor. */
3331 	while (1) {
3332 		struct tx_cmpl *txcmp;
3333 		uint32_t cons;
3334 
3335 		cons = RING_CMPL(ring_mask, raw_cons);
3336 		txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3337 
3338 		if (!bnxt_cpr_cmp_valid(txcmp, raw_cons, ring_mask + 1))
3339 			break;
3340 
3341 		if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3342 			nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3343 
3344 		if (nb_tx_pkts > offset)
3345 			return RTE_ETH_TX_DESC_DONE;
3346 
3347 		raw_cons = NEXT_RAW_CMP(raw_cons);
3348 	}
3349 
3350 	/* Descriptor is pending transmit, not yet completed by hardware. */
3351 	return RTE_ETH_TX_DESC_FULL;
3352 }
3353 
3354 int
3355 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3356 		     const struct rte_flow_ops **ops)
3357 {
3358 	struct bnxt *bp = dev->data->dev_private;
3359 	int ret = 0;
3360 
3361 	if (!bp)
3362 		return -EIO;
3363 
3364 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3365 		struct bnxt_representor *vfr = dev->data->dev_private;
3366 		bp = vfr->parent_dev->data->dev_private;
3367 		/* parent is deleted while children are still valid */
3368 		if (!bp) {
3369 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3370 				    dev->data->port_id);
3371 			return -EIO;
3372 		}
3373 	}
3374 
3375 	ret = is_bnxt_in_error(bp);
3376 	if (ret)
3377 		return ret;
3378 
3379 	/* PMD supports thread-safe flow operations.  rte_flow API
3380 	 * functions can avoid mutex for multi-thread safety.
3381 	 */
3382 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3383 
3384 	if (BNXT_TRUFLOW_EN(bp))
3385 		*ops = &bnxt_ulp_rte_flow_ops;
3386 	else
3387 		*ops = &bnxt_flow_ops;
3388 
3389 	return ret;
3390 }
3391 
3392 static const uint32_t *
3393 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3394 {
3395 	static const uint32_t ptypes[] = {
3396 		RTE_PTYPE_L2_ETHER_VLAN,
3397 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3398 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3399 		RTE_PTYPE_L4_ICMP,
3400 		RTE_PTYPE_L4_TCP,
3401 		RTE_PTYPE_L4_UDP,
3402 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3403 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3404 		RTE_PTYPE_INNER_L4_ICMP,
3405 		RTE_PTYPE_INNER_L4_TCP,
3406 		RTE_PTYPE_INNER_L4_UDP,
3407 		RTE_PTYPE_UNKNOWN
3408 	};
3409 
3410 	if (!dev->rx_pkt_burst)
3411 		return NULL;
3412 
3413 	return ptypes;
3414 }
3415 
3416 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3417 			 int reg_win)
3418 {
3419 	uint32_t reg_base = *reg_arr & 0xfffff000;
3420 	uint32_t win_off;
3421 	int i;
3422 
3423 	for (i = 0; i < count; i++) {
3424 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3425 			return -ERANGE;
3426 	}
3427 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3428 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3429 	return 0;
3430 }
3431 
3432 static int bnxt_map_ptp_regs(struct bnxt *bp)
3433 {
3434 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3435 	uint32_t *reg_arr;
3436 	int rc, i;
3437 
3438 	reg_arr = ptp->rx_regs;
3439 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3440 	if (rc)
3441 		return rc;
3442 
3443 	reg_arr = ptp->tx_regs;
3444 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3445 	if (rc)
3446 		return rc;
3447 
3448 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3449 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3450 
3451 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3452 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3453 
3454 	return 0;
3455 }
3456 
3457 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3458 {
3459 	rte_write32(0, (uint8_t *)bp->bar0 +
3460 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3461 	rte_write32(0, (uint8_t *)bp->bar0 +
3462 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3463 }
3464 
3465 static uint64_t bnxt_cc_read(struct bnxt *bp)
3466 {
3467 	uint64_t ns;
3468 
3469 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3470 			      BNXT_GRCPF_REG_SYNC_TIME));
3471 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3472 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3473 	return ns;
3474 }
3475 
3476 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3477 {
3478 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3479 	uint32_t fifo;
3480 
3481 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3482 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3483 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3484 		return -EAGAIN;
3485 
3486 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3487 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3488 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3489 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3490 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3491 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3492 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3493 
3494 	return 0;
3495 }
3496 
3497 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3498 {
3499 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3500 	struct bnxt_pf_info *pf = bp->pf;
3501 	uint16_t port_id;
3502 	int i = 0;
3503 	uint32_t fifo;
3504 
3505 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3506 		return -EINVAL;
3507 
3508 	port_id = pf->port_id;
3509 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3510 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3511 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3512 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3513 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3514 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3515 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3516 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3517 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3518 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3519 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3520 		i++;
3521 	}
3522 
3523 	if (i >= BNXT_PTP_RX_PND_CNT)
3524 		return -EBUSY;
3525 
3526 	return 0;
3527 }
3528 
3529 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3530 {
3531 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3532 	struct bnxt_pf_info *pf = bp->pf;
3533 	uint16_t port_id;
3534 	uint32_t fifo;
3535 
3536 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3537 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3538 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3539 		return -EAGAIN;
3540 
3541 	port_id = pf->port_id;
3542 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3543 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3544 
3545 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3546 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3547 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3548 		return bnxt_clr_rx_ts(bp, ts);
3549 
3550 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3551 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3552 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3553 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3554 
3555 	return 0;
3556 }
3557 
3558 static int
3559 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3560 {
3561 	uint64_t ns;
3562 	struct bnxt *bp = dev->data->dev_private;
3563 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3564 
3565 	if (!ptp)
3566 		return -ENOTSUP;
3567 
3568 	ns = rte_timespec_to_ns(ts);
3569 	/* Set the timecounters to a new value. */
3570 	ptp->tc.nsec = ns;
3571 	ptp->tx_tstamp_tc.nsec = ns;
3572 	ptp->rx_tstamp_tc.nsec = ns;
3573 
3574 	return 0;
3575 }
3576 
3577 static int
3578 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3579 {
3580 	struct bnxt *bp = dev->data->dev_private;
3581 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3582 	uint64_t ns, systime_cycles = 0;
3583 	int rc = 0;
3584 
3585 	if (!ptp)
3586 		return -ENOTSUP;
3587 
3588 	if (BNXT_CHIP_P5(bp))
3589 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3590 					     &systime_cycles);
3591 	else
3592 		systime_cycles = bnxt_cc_read(bp);
3593 
3594 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3595 	*ts = rte_ns_to_timespec(ns);
3596 
3597 	return rc;
3598 }
3599 static int
3600 bnxt_timesync_enable(struct rte_eth_dev *dev)
3601 {
3602 	struct bnxt *bp = dev->data->dev_private;
3603 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3604 	uint32_t shift = 0;
3605 	int rc;
3606 
3607 	if (!ptp)
3608 		return -ENOTSUP;
3609 
3610 	ptp->rx_filter = 1;
3611 	ptp->tx_tstamp_en = 1;
3612 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3613 
3614 	rc = bnxt_hwrm_ptp_cfg(bp);
3615 	if (rc)
3616 		return rc;
3617 
3618 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3619 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3620 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3621 
3622 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3623 	ptp->tc.cc_shift = shift;
3624 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3625 
3626 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3627 	ptp->rx_tstamp_tc.cc_shift = shift;
3628 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3629 
3630 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3631 	ptp->tx_tstamp_tc.cc_shift = shift;
3632 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3633 
3634 	if (!BNXT_CHIP_P5(bp))
3635 		bnxt_map_ptp_regs(bp);
3636 	else
3637 		rc = bnxt_ptp_start(bp);
3638 
3639 	return rc;
3640 }
3641 
3642 static int
3643 bnxt_timesync_disable(struct rte_eth_dev *dev)
3644 {
3645 	struct bnxt *bp = dev->data->dev_private;
3646 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3647 
3648 	if (!ptp)
3649 		return -ENOTSUP;
3650 
3651 	ptp->rx_filter = 0;
3652 	ptp->tx_tstamp_en = 0;
3653 	ptp->rxctl = 0;
3654 
3655 	bnxt_hwrm_ptp_cfg(bp);
3656 
3657 	if (!BNXT_CHIP_P5(bp))
3658 		bnxt_unmap_ptp_regs(bp);
3659 	else
3660 		bnxt_ptp_stop(bp);
3661 
3662 	return 0;
3663 }
3664 
3665 static int
3666 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3667 				 struct timespec *timestamp,
3668 				 uint32_t flags __rte_unused)
3669 {
3670 	struct bnxt *bp = dev->data->dev_private;
3671 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3672 	uint64_t rx_tstamp_cycles = 0;
3673 	uint64_t ns;
3674 
3675 	if (!ptp)
3676 		return -ENOTSUP;
3677 
3678 	if (BNXT_CHIP_P5(bp))
3679 		rx_tstamp_cycles = ptp->rx_timestamp;
3680 	else
3681 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3682 
3683 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3684 	*timestamp = rte_ns_to_timespec(ns);
3685 	return  0;
3686 }
3687 
3688 static int
3689 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3690 				 struct timespec *timestamp)
3691 {
3692 	struct bnxt *bp = dev->data->dev_private;
3693 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3694 	uint64_t tx_tstamp_cycles = 0;
3695 	uint64_t ns;
3696 	int rc = 0;
3697 
3698 	if (!ptp)
3699 		return -ENOTSUP;
3700 
3701 	if (BNXT_CHIP_P5(bp))
3702 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3703 					     &tx_tstamp_cycles);
3704 	else
3705 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3706 
3707 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3708 	*timestamp = rte_ns_to_timespec(ns);
3709 
3710 	return rc;
3711 }
3712 
3713 static int
3714 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3715 {
3716 	struct bnxt *bp = dev->data->dev_private;
3717 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3718 
3719 	if (!ptp)
3720 		return -ENOTSUP;
3721 
3722 	ptp->tc.nsec += delta;
3723 	ptp->tx_tstamp_tc.nsec += delta;
3724 	ptp->rx_tstamp_tc.nsec += delta;
3725 
3726 	return 0;
3727 }
3728 
3729 static int
3730 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3731 {
3732 	struct bnxt *bp = dev->data->dev_private;
3733 	int rc;
3734 	uint32_t dir_entries;
3735 	uint32_t entry_length;
3736 
3737 	rc = is_bnxt_in_error(bp);
3738 	if (rc)
3739 		return rc;
3740 
3741 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3742 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3743 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3744 
3745 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3746 	if (rc != 0)
3747 		return rc;
3748 
3749 	return dir_entries * entry_length;
3750 }
3751 
3752 static int
3753 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3754 		struct rte_dev_eeprom_info *in_eeprom)
3755 {
3756 	struct bnxt *bp = dev->data->dev_private;
3757 	uint32_t index;
3758 	uint32_t offset;
3759 	int rc;
3760 
3761 	rc = is_bnxt_in_error(bp);
3762 	if (rc)
3763 		return rc;
3764 
3765 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3766 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3767 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3768 		    in_eeprom->offset, in_eeprom->length);
3769 
3770 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3771 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3772 						in_eeprom->data);
3773 
3774 	index = in_eeprom->offset >> 24;
3775 	offset = in_eeprom->offset & 0xffffff;
3776 
3777 	if (index != 0)
3778 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3779 					   in_eeprom->length, in_eeprom->data);
3780 
3781 	return 0;
3782 }
3783 
3784 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3785 {
3786 	switch (dir_type) {
3787 	case BNX_DIR_TYPE_CHIMP_PATCH:
3788 	case BNX_DIR_TYPE_BOOTCODE:
3789 	case BNX_DIR_TYPE_BOOTCODE_2:
3790 	case BNX_DIR_TYPE_APE_FW:
3791 	case BNX_DIR_TYPE_APE_PATCH:
3792 	case BNX_DIR_TYPE_KONG_FW:
3793 	case BNX_DIR_TYPE_KONG_PATCH:
3794 	case BNX_DIR_TYPE_BONO_FW:
3795 	case BNX_DIR_TYPE_BONO_PATCH:
3796 		/* FALLTHROUGH */
3797 		return true;
3798 	}
3799 
3800 	return false;
3801 }
3802 
3803 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3804 {
3805 	switch (dir_type) {
3806 	case BNX_DIR_TYPE_AVS:
3807 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3808 	case BNX_DIR_TYPE_PCIE:
3809 	case BNX_DIR_TYPE_TSCF_UCODE:
3810 	case BNX_DIR_TYPE_EXT_PHY:
3811 	case BNX_DIR_TYPE_CCM:
3812 	case BNX_DIR_TYPE_ISCSI_BOOT:
3813 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3814 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3815 		/* FALLTHROUGH */
3816 		return true;
3817 	}
3818 
3819 	return false;
3820 }
3821 
3822 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3823 {
3824 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3825 		bnxt_dir_type_is_other_exec_format(dir_type);
3826 }
3827 
3828 static int
3829 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3830 		struct rte_dev_eeprom_info *in_eeprom)
3831 {
3832 	struct bnxt *bp = dev->data->dev_private;
3833 	uint8_t index, dir_op;
3834 	uint16_t type, ext, ordinal, attr;
3835 	int rc;
3836 
3837 	rc = is_bnxt_in_error(bp);
3838 	if (rc)
3839 		return rc;
3840 
3841 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3842 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3843 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3844 		    in_eeprom->offset, in_eeprom->length);
3845 
3846 	if (!BNXT_PF(bp)) {
3847 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3848 		return -EINVAL;
3849 	}
3850 
3851 	type = in_eeprom->magic >> 16;
3852 
3853 	if (type == 0xffff) { /* special value for directory operations */
3854 		index = in_eeprom->magic & 0xff;
3855 		dir_op = in_eeprom->magic >> 8;
3856 		if (index == 0)
3857 			return -EINVAL;
3858 		switch (dir_op) {
3859 		case 0x0e: /* erase */
3860 			if (in_eeprom->offset != ~in_eeprom->magic)
3861 				return -EINVAL;
3862 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3863 		default:
3864 			return -EINVAL;
3865 		}
3866 	}
3867 
3868 	/* Create or re-write an NVM item: */
3869 	if (bnxt_dir_type_is_executable(type) == true)
3870 		return -EOPNOTSUPP;
3871 	ext = in_eeprom->magic & 0xffff;
3872 	ordinal = in_eeprom->offset >> 16;
3873 	attr = in_eeprom->offset & 0xffff;
3874 
3875 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3876 				     in_eeprom->data, in_eeprom->length);
3877 }
3878 
3879 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3880 				struct rte_eth_dev_module_info *modinfo)
3881 {
3882 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3883 	struct bnxt *bp = dev->data->dev_private;
3884 	int rc;
3885 
3886 	/* No point in going further if phy status indicates
3887 	 * module is not inserted or if it is powered down or
3888 	 * if it is of type 10GBase-T
3889 	 */
3890 	if (bp->link_info->module_status >
3891 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3892 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3893 			    dev->data->port_id);
3894 		return -ENOTSUP;
3895 	}
3896 
3897 	/* This feature is not supported in older firmware versions */
3898 	if (bp->hwrm_spec_code < 0x10202) {
3899 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3900 			    dev->data->port_id);
3901 		return -ENOTSUP;
3902 	}
3903 
3904 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3905 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3906 						   module_info);
3907 
3908 	if (rc)
3909 		return rc;
3910 
3911 	switch (module_info[0]) {
3912 	case SFF_MODULE_ID_SFP:
3913 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3914 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3915 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3916 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3917 		break;
3918 	case SFF_MODULE_ID_QSFP:
3919 	case SFF_MODULE_ID_QSFP_PLUS:
3920 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3921 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3922 		break;
3923 	case SFF_MODULE_ID_QSFP28:
3924 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3925 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3926 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3927 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3928 		break;
3929 	default:
3930 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3931 		return -ENOTSUP;
3932 	}
3933 
3934 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3935 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3936 
3937 	return 0;
3938 }
3939 
3940 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3941 				  struct rte_dev_eeprom_info *info)
3942 {
3943 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3944 	uint32_t offset = info->offset, length = info->length;
3945 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3946 	struct bnxt *bp = dev->data->dev_private;
3947 	uint8_t *data = info->data;
3948 	uint8_t page = offset >> 7;
3949 	uint8_t max_pages = 2;
3950 	uint8_t opt_pages;
3951 	int rc;
3952 
3953 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3954 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3955 						   module_info);
3956 	if (rc)
3957 		return rc;
3958 
3959 	switch (module_info[0]) {
3960 	case SFF_MODULE_ID_SFP:
3961 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
3962 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
3963 			pg_addr[2] = I2C_DEV_ADDR_A2;
3964 			pg_addr[3] = I2C_DEV_ADDR_A2;
3965 			max_pages = 4;
3966 		}
3967 		break;
3968 	case SFF_MODULE_ID_QSFP28:
3969 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
3970 							   SFF8636_OPT_PAGES_OFFSET,
3971 							   1, &opt_pages);
3972 		if (rc)
3973 			return rc;
3974 
3975 		if (opt_pages & SFF8636_PAGE1_MASK) {
3976 			pg_addr[2] = I2C_DEV_ADDR_A0;
3977 			max_pages = 3;
3978 		}
3979 		if (opt_pages & SFF8636_PAGE2_MASK) {
3980 			pg_addr[3] = I2C_DEV_ADDR_A0;
3981 			max_pages = 4;
3982 		}
3983 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
3984 			pg_addr[4] = I2C_DEV_ADDR_A0;
3985 			max_pages = 5;
3986 		}
3987 		break;
3988 	default:
3989 		break;
3990 	}
3991 
3992 	memset(data, 0, length);
3993 
3994 	offset &= 0xff;
3995 	while (length && page < max_pages) {
3996 		uint8_t raw_page = page ? page - 1 : 0;
3997 		uint16_t chunk;
3998 
3999 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
4000 			raw_page = 0;
4001 		else if (page)
4002 			offset |= 0x80;
4003 		chunk = RTE_MIN(length, 256 - offset);
4004 
4005 		if (pg_addr[page]) {
4006 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
4007 								   raw_page, offset,
4008 								   chunk, data);
4009 			if (rc)
4010 				return rc;
4011 		}
4012 
4013 		data += chunk;
4014 		length -= chunk;
4015 		offset = 0;
4016 		page += 1 + (chunk > 128);
4017 	}
4018 
4019 	return length ? -EINVAL : 0;
4020 }
4021 
4022 /*
4023  * Initialization
4024  */
4025 
4026 static const struct eth_dev_ops bnxt_dev_ops = {
4027 	.dev_infos_get = bnxt_dev_info_get_op,
4028 	.dev_close = bnxt_dev_close_op,
4029 	.dev_configure = bnxt_dev_configure_op,
4030 	.dev_start = bnxt_dev_start_op,
4031 	.dev_stop = bnxt_dev_stop_op,
4032 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4033 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4034 	.stats_get = bnxt_stats_get_op,
4035 	.stats_reset = bnxt_stats_reset_op,
4036 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4037 	.rx_queue_release = bnxt_rx_queue_release_op,
4038 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4039 	.tx_queue_release = bnxt_tx_queue_release_op,
4040 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4041 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4042 	.reta_update = bnxt_reta_update_op,
4043 	.reta_query = bnxt_reta_query_op,
4044 	.rss_hash_update = bnxt_rss_hash_update_op,
4045 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4046 	.link_update = bnxt_link_update_op,
4047 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4048 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4049 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4050 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4051 	.mac_addr_add = bnxt_mac_addr_add_op,
4052 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4053 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4054 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4055 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4056 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4057 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4058 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4059 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4060 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4061 	.mtu_set = bnxt_mtu_set_op,
4062 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4063 	.xstats_get = bnxt_dev_xstats_get_op,
4064 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4065 	.xstats_reset = bnxt_dev_xstats_reset_op,
4066 	.fw_version_get = bnxt_fw_version_get,
4067 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4068 	.rxq_info_get = bnxt_rxq_info_get_op,
4069 	.txq_info_get = bnxt_txq_info_get_op,
4070 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4071 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4072 	.dev_led_on = bnxt_dev_led_on_op,
4073 	.dev_led_off = bnxt_dev_led_off_op,
4074 	.rx_queue_start = bnxt_rx_queue_start,
4075 	.rx_queue_stop = bnxt_rx_queue_stop,
4076 	.tx_queue_start = bnxt_tx_queue_start,
4077 	.tx_queue_stop = bnxt_tx_queue_stop,
4078 	.flow_ops_get = bnxt_flow_ops_get_op,
4079 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4080 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4081 	.get_eeprom           = bnxt_get_eeprom_op,
4082 	.set_eeprom           = bnxt_set_eeprom_op,
4083 	.get_module_info = bnxt_get_module_info,
4084 	.get_module_eeprom = bnxt_get_module_eeprom,
4085 	.timesync_enable      = bnxt_timesync_enable,
4086 	.timesync_disable     = bnxt_timesync_disable,
4087 	.timesync_read_time   = bnxt_timesync_read_time,
4088 	.timesync_write_time   = bnxt_timesync_write_time,
4089 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4090 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4091 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4092 };
4093 
4094 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4095 {
4096 	uint32_t offset;
4097 
4098 	/* Only pre-map the reset GRC registers using window 3 */
4099 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4100 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4101 
4102 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4103 
4104 	return offset;
4105 }
4106 
4107 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4108 {
4109 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4110 	uint32_t reg_base = 0xffffffff;
4111 	int i;
4112 
4113 	/* Only pre-map the monitoring GRC registers using window 2 */
4114 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4115 		uint32_t reg = info->status_regs[i];
4116 
4117 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4118 			continue;
4119 
4120 		if (reg_base == 0xffffffff)
4121 			reg_base = reg & 0xfffff000;
4122 		if ((reg & 0xfffff000) != reg_base)
4123 			return -ERANGE;
4124 
4125 		/* Use mask 0xffc as the Lower 2 bits indicates
4126 		 * address space location
4127 		 */
4128 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4129 						(reg & 0xffc);
4130 	}
4131 
4132 	if (reg_base == 0xffffffff)
4133 		return 0;
4134 
4135 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4136 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4137 
4138 	return 0;
4139 }
4140 
4141 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4142 {
4143 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4144 	uint32_t delay = info->delay_after_reset[index];
4145 	uint32_t val = info->reset_reg_val[index];
4146 	uint32_t reg = info->reset_reg[index];
4147 	uint32_t type, offset;
4148 	int ret;
4149 
4150 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4151 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4152 
4153 	switch (type) {
4154 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4155 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4156 		if (ret < 0) {
4157 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4158 				    val, offset);
4159 			return;
4160 		}
4161 		break;
4162 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4163 		offset = bnxt_map_reset_regs(bp, offset);
4164 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4165 		break;
4166 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4167 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4168 		break;
4169 	}
4170 	/* wait on a specific interval of time until core reset is complete */
4171 	if (delay)
4172 		rte_delay_ms(delay);
4173 }
4174 
4175 static void bnxt_dev_cleanup(struct bnxt *bp)
4176 {
4177 	bp->eth_dev->data->dev_link.link_status = 0;
4178 	bp->link_info->link_up = 0;
4179 	if (bp->eth_dev->data->dev_started)
4180 		bnxt_dev_stop(bp->eth_dev);
4181 
4182 	bnxt_uninit_resources(bp, true);
4183 }
4184 
4185 static int
4186 bnxt_check_fw_reset_done(struct bnxt *bp)
4187 {
4188 	int timeout = bp->fw_reset_max_msecs;
4189 	uint16_t val = 0;
4190 	int rc;
4191 
4192 	do {
4193 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4194 		if (rc < 0) {
4195 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4196 			return rc;
4197 		}
4198 		if (val != 0xffff)
4199 			break;
4200 		rte_delay_ms(1);
4201 	} while (timeout--);
4202 
4203 	if (val == 0xffff) {
4204 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4205 		return -1;
4206 	}
4207 
4208 	return 0;
4209 }
4210 
4211 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4212 {
4213 	struct rte_eth_dev *dev = bp->eth_dev;
4214 	struct rte_vlan_filter_conf *vfc;
4215 	int vidx, vbit, rc;
4216 	uint16_t vlan_id;
4217 
4218 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4219 		vfc = &dev->data->vlan_filter_conf;
4220 		vidx = vlan_id / 64;
4221 		vbit = vlan_id % 64;
4222 
4223 		/* Each bit corresponds to a VLAN id */
4224 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4225 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4226 			if (rc)
4227 				return rc;
4228 		}
4229 	}
4230 
4231 	return 0;
4232 }
4233 
4234 static int bnxt_restore_mac_filters(struct bnxt *bp)
4235 {
4236 	struct rte_eth_dev *dev = bp->eth_dev;
4237 	struct rte_eth_dev_info dev_info;
4238 	struct rte_ether_addr *addr;
4239 	uint64_t pool_mask;
4240 	uint32_t pool = 0;
4241 	uint32_t i;
4242 	int rc;
4243 
4244 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4245 		return 0;
4246 
4247 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4248 	if (rc)
4249 		return rc;
4250 
4251 	/* replay MAC address configuration */
4252 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4253 		addr = &dev->data->mac_addrs[i];
4254 
4255 		/* skip zero address */
4256 		if (rte_is_zero_ether_addr(addr))
4257 			continue;
4258 
4259 		pool = 0;
4260 		pool_mask = dev->data->mac_pool_sel[i];
4261 
4262 		do {
4263 			if (pool_mask & 1ULL) {
4264 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4265 				if (rc)
4266 					return rc;
4267 			}
4268 			pool_mask >>= 1;
4269 			pool++;
4270 		} while (pool_mask);
4271 	}
4272 
4273 	return 0;
4274 }
4275 
4276 static int bnxt_restore_filters(struct bnxt *bp)
4277 {
4278 	struct rte_eth_dev *dev = bp->eth_dev;
4279 	int ret = 0;
4280 
4281 	if (dev->data->all_multicast) {
4282 		ret = bnxt_allmulticast_enable_op(dev);
4283 		if (ret)
4284 			return ret;
4285 	}
4286 	if (dev->data->promiscuous) {
4287 		ret = bnxt_promiscuous_enable_op(dev);
4288 		if (ret)
4289 			return ret;
4290 	}
4291 
4292 	ret = bnxt_restore_mac_filters(bp);
4293 	if (ret)
4294 		return ret;
4295 
4296 	ret = bnxt_restore_vlan_filters(bp);
4297 	/* TODO restore other filters as well */
4298 	return ret;
4299 }
4300 
4301 static int bnxt_check_fw_ready(struct bnxt *bp)
4302 {
4303 	int timeout = bp->fw_reset_max_msecs;
4304 	int rc = 0;
4305 
4306 	do {
4307 		rc = bnxt_hwrm_poll_ver_get(bp);
4308 		if (rc == 0)
4309 			break;
4310 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4311 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4312 	} while (rc && timeout > 0);
4313 
4314 	if (rc)
4315 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4316 
4317 	return rc;
4318 }
4319 
4320 static void bnxt_dev_recover(void *arg)
4321 {
4322 	struct bnxt *bp = arg;
4323 	int rc = 0;
4324 
4325 	pthread_mutex_lock(&bp->err_recovery_lock);
4326 
4327 	if (!bp->fw_reset_min_msecs) {
4328 		rc = bnxt_check_fw_reset_done(bp);
4329 		if (rc)
4330 			goto err;
4331 	}
4332 
4333 	/* Clear Error flag so that device re-init should happen */
4334 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4335 
4336 	rc = bnxt_check_fw_ready(bp);
4337 	if (rc)
4338 		goto err;
4339 
4340 	rc = bnxt_init_resources(bp, true);
4341 	if (rc) {
4342 		PMD_DRV_LOG(ERR,
4343 			    "Failed to initialize resources after reset\n");
4344 		goto err;
4345 	}
4346 	/* clear reset flag as the device is initialized now */
4347 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4348 
4349 	rc = bnxt_dev_start_op(bp->eth_dev);
4350 	if (rc) {
4351 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4352 		goto err_start;
4353 	}
4354 
4355 	rc = bnxt_restore_filters(bp);
4356 	if (rc)
4357 		goto err_start;
4358 
4359 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4360 	pthread_mutex_unlock(&bp->err_recovery_lock);
4361 
4362 	return;
4363 err_start:
4364 	bnxt_dev_stop(bp->eth_dev);
4365 err:
4366 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4367 	bnxt_uninit_resources(bp, false);
4368 	if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4369 		rte_eth_dev_callback_process(bp->eth_dev,
4370 					     RTE_ETH_EVENT_INTR_RMV,
4371 					     NULL);
4372 	pthread_mutex_unlock(&bp->err_recovery_lock);
4373 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4374 }
4375 
4376 void bnxt_dev_reset_and_resume(void *arg)
4377 {
4378 	struct bnxt *bp = arg;
4379 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4380 	uint16_t val = 0;
4381 	int rc;
4382 
4383 	bnxt_dev_cleanup(bp);
4384 
4385 	bnxt_wait_for_device_shutdown(bp);
4386 
4387 	/* During some fatal firmware error conditions, the PCI config space
4388 	 * register 0x2e which normally contains the subsystem ID will become
4389 	 * 0xffff. This register will revert back to the normal value after
4390 	 * the chip has completed core reset. If we detect this condition,
4391 	 * we can poll this config register immediately for the value to revert.
4392 	 */
4393 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4394 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4395 		if (rc < 0) {
4396 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4397 			return;
4398 		}
4399 		if (val == 0xffff) {
4400 			bp->fw_reset_min_msecs = 0;
4401 			us = 1;
4402 		}
4403 	}
4404 
4405 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4406 	if (rc)
4407 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4408 }
4409 
4410 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4411 {
4412 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4413 	uint32_t reg = info->status_regs[index];
4414 	uint32_t type, offset, val = 0;
4415 	int ret = 0;
4416 
4417 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4418 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4419 
4420 	switch (type) {
4421 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4422 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4423 		if (ret < 0)
4424 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4425 				    offset);
4426 		break;
4427 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4428 		offset = info->mapped_status_regs[index];
4429 		/* FALLTHROUGH */
4430 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4431 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4432 				       offset));
4433 		break;
4434 	}
4435 
4436 	return val;
4437 }
4438 
4439 static int bnxt_fw_reset_all(struct bnxt *bp)
4440 {
4441 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4442 	uint32_t i;
4443 	int rc = 0;
4444 
4445 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4446 		/* Reset through master function driver */
4447 		for (i = 0; i < info->reg_array_cnt; i++)
4448 			bnxt_write_fw_reset_reg(bp, i);
4449 		/* Wait for time specified by FW after triggering reset */
4450 		rte_delay_ms(info->master_func_wait_period_after_reset);
4451 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4452 		/* Reset with the help of Kong processor */
4453 		rc = bnxt_hwrm_fw_reset(bp);
4454 		if (rc)
4455 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4456 	}
4457 
4458 	return rc;
4459 }
4460 
4461 static void bnxt_fw_reset_cb(void *arg)
4462 {
4463 	struct bnxt *bp = arg;
4464 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4465 	int rc = 0;
4466 
4467 	/* Only Master function can do FW reset */
4468 	if (bnxt_is_master_func(bp) &&
4469 	    bnxt_is_recovery_enabled(bp)) {
4470 		rc = bnxt_fw_reset_all(bp);
4471 		if (rc) {
4472 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4473 			return;
4474 		}
4475 	}
4476 
4477 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4478 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4479 	 * (including MASTER FUNC). After receiving this Async, all the active
4480 	 * drivers should treat this case as FW initiated recovery
4481 	 */
4482 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4483 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4484 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4485 
4486 		/* To recover from error */
4487 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4488 				  (void *)bp);
4489 	}
4490 }
4491 
4492 /* Driver should poll FW heartbeat, reset_counter with the frequency
4493  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4494  * When the driver detects heartbeat stop or change in reset_counter,
4495  * it has to trigger a reset to recover from the error condition.
4496  * A “master PF” is the function who will have the privilege to
4497  * initiate the chimp reset. The master PF will be elected by the
4498  * firmware and will be notified through async message.
4499  */
4500 static void bnxt_check_fw_health(void *arg)
4501 {
4502 	struct bnxt *bp = arg;
4503 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4504 	uint32_t val = 0, wait_msec;
4505 
4506 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4507 	    is_bnxt_in_error(bp))
4508 		return;
4509 
4510 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4511 	if (val == info->last_heart_beat)
4512 		goto reset;
4513 
4514 	info->last_heart_beat = val;
4515 
4516 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4517 	if (val != info->last_reset_counter)
4518 		goto reset;
4519 
4520 	info->last_reset_counter = val;
4521 
4522 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4523 			  bnxt_check_fw_health, (void *)bp);
4524 
4525 	return;
4526 reset:
4527 	/* Stop DMA to/from device */
4528 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4529 	bp->flags |= BNXT_FLAG_FW_RESET;
4530 
4531 	bnxt_stop_rxtx(bp);
4532 
4533 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4534 
4535 	if (bnxt_is_master_func(bp))
4536 		wait_msec = info->master_func_wait_period;
4537 	else
4538 		wait_msec = info->normal_func_wait_period;
4539 
4540 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4541 			  bnxt_fw_reset_cb, (void *)bp);
4542 }
4543 
4544 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4545 {
4546 	uint32_t polling_freq;
4547 
4548 	pthread_mutex_lock(&bp->health_check_lock);
4549 
4550 	if (!bnxt_is_recovery_enabled(bp))
4551 		goto done;
4552 
4553 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4554 		goto done;
4555 
4556 	polling_freq = bp->recovery_info->driver_polling_freq;
4557 
4558 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4559 			  bnxt_check_fw_health, (void *)bp);
4560 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4561 
4562 done:
4563 	pthread_mutex_unlock(&bp->health_check_lock);
4564 }
4565 
4566 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4567 {
4568 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4569 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4570 }
4571 
4572 static bool bnxt_vf_pciid(uint16_t device_id)
4573 {
4574 	switch (device_id) {
4575 	case BROADCOM_DEV_ID_57304_VF:
4576 	case BROADCOM_DEV_ID_57406_VF:
4577 	case BROADCOM_DEV_ID_5731X_VF:
4578 	case BROADCOM_DEV_ID_5741X_VF:
4579 	case BROADCOM_DEV_ID_57414_VF:
4580 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4581 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4582 	case BROADCOM_DEV_ID_58802_VF:
4583 	case BROADCOM_DEV_ID_57500_VF1:
4584 	case BROADCOM_DEV_ID_57500_VF2:
4585 	case BROADCOM_DEV_ID_58818_VF:
4586 		/* FALLTHROUGH */
4587 		return true;
4588 	default:
4589 		return false;
4590 	}
4591 }
4592 
4593 /* Phase 5 device */
4594 static bool bnxt_p5_device(uint16_t device_id)
4595 {
4596 	switch (device_id) {
4597 	case BROADCOM_DEV_ID_57508:
4598 	case BROADCOM_DEV_ID_57504:
4599 	case BROADCOM_DEV_ID_57502:
4600 	case BROADCOM_DEV_ID_57508_MF1:
4601 	case BROADCOM_DEV_ID_57504_MF1:
4602 	case BROADCOM_DEV_ID_57502_MF1:
4603 	case BROADCOM_DEV_ID_57508_MF2:
4604 	case BROADCOM_DEV_ID_57504_MF2:
4605 	case BROADCOM_DEV_ID_57502_MF2:
4606 	case BROADCOM_DEV_ID_57500_VF1:
4607 	case BROADCOM_DEV_ID_57500_VF2:
4608 	case BROADCOM_DEV_ID_58812:
4609 	case BROADCOM_DEV_ID_58814:
4610 	case BROADCOM_DEV_ID_58818:
4611 	case BROADCOM_DEV_ID_58818_VF:
4612 		/* FALLTHROUGH */
4613 		return true;
4614 	default:
4615 		return false;
4616 	}
4617 }
4618 
4619 bool bnxt_stratus_device(struct bnxt *bp)
4620 {
4621 	uint16_t device_id = bp->pdev->id.device_id;
4622 
4623 	switch (device_id) {
4624 	case BROADCOM_DEV_ID_STRATUS_NIC:
4625 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4626 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4627 		/* FALLTHROUGH */
4628 		return true;
4629 	default:
4630 		return false;
4631 	}
4632 }
4633 
4634 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4635 {
4636 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4637 	struct bnxt *bp = eth_dev->data->dev_private;
4638 
4639 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4640 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4641 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4642 	if (!bp->bar0 || !bp->doorbell_base) {
4643 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4644 		return -ENODEV;
4645 	}
4646 
4647 	bp->eth_dev = eth_dev;
4648 	bp->pdev = pci_dev;
4649 
4650 	return 0;
4651 }
4652 
4653 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4654 				  struct bnxt_ctx_pg_info *ctx_pg,
4655 				  uint32_t mem_size,
4656 				  const char *suffix,
4657 				  uint16_t idx)
4658 {
4659 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4660 	const struct rte_memzone *mz = NULL;
4661 	char mz_name[RTE_MEMZONE_NAMESIZE];
4662 	rte_iova_t mz_phys_addr;
4663 	uint64_t valid_bits = 0;
4664 	uint32_t sz;
4665 	int i;
4666 
4667 	if (!mem_size)
4668 		return 0;
4669 
4670 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4671 			 BNXT_PAGE_SIZE;
4672 	rmem->page_size = BNXT_PAGE_SIZE;
4673 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4674 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4675 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4676 
4677 	valid_bits = PTU_PTE_VALID;
4678 
4679 	if (rmem->nr_pages > 1) {
4680 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4681 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4682 			 suffix, idx, bp->eth_dev->data->port_id);
4683 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4684 		mz = rte_memzone_lookup(mz_name);
4685 		if (!mz) {
4686 			mz = rte_memzone_reserve_aligned(mz_name,
4687 						rmem->nr_pages * 8,
4688 						bp->eth_dev->device->numa_node,
4689 						RTE_MEMZONE_2MB |
4690 						RTE_MEMZONE_SIZE_HINT_ONLY |
4691 						RTE_MEMZONE_IOVA_CONTIG,
4692 						BNXT_PAGE_SIZE);
4693 			if (mz == NULL)
4694 				return -ENOMEM;
4695 		}
4696 
4697 		memset(mz->addr, 0, mz->len);
4698 		mz_phys_addr = mz->iova;
4699 
4700 		rmem->pg_tbl = mz->addr;
4701 		rmem->pg_tbl_map = mz_phys_addr;
4702 		rmem->pg_tbl_mz = mz;
4703 	}
4704 
4705 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4706 		 suffix, idx, bp->eth_dev->data->port_id);
4707 	mz = rte_memzone_lookup(mz_name);
4708 	if (!mz) {
4709 		mz = rte_memzone_reserve_aligned(mz_name,
4710 						 mem_size,
4711 						 bp->eth_dev->device->numa_node,
4712 						 RTE_MEMZONE_1GB |
4713 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4714 						 RTE_MEMZONE_IOVA_CONTIG,
4715 						 BNXT_PAGE_SIZE);
4716 		if (mz == NULL)
4717 			return -ENOMEM;
4718 	}
4719 
4720 	memset(mz->addr, 0, mz->len);
4721 	mz_phys_addr = mz->iova;
4722 
4723 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4724 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4725 		rmem->dma_arr[i] = mz_phys_addr + sz;
4726 
4727 		if (rmem->nr_pages > 1) {
4728 			if (i == rmem->nr_pages - 2 &&
4729 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4730 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4731 			else if (i == rmem->nr_pages - 1 &&
4732 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4733 				valid_bits |= PTU_PTE_LAST;
4734 
4735 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4736 							   valid_bits);
4737 		}
4738 	}
4739 
4740 	rmem->mz = mz;
4741 	if (rmem->vmem_size)
4742 		rmem->vmem = (void **)mz->addr;
4743 	rmem->dma_arr[0] = mz_phys_addr;
4744 	return 0;
4745 }
4746 
4747 static void bnxt_free_ctx_mem(struct bnxt *bp)
4748 {
4749 	int i;
4750 
4751 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4752 		return;
4753 
4754 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4755 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4756 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4757 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4758 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4759 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4760 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4761 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4762 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4763 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4764 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4765 
4766 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4767 		if (bp->ctx->tqm_mem[i])
4768 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4769 	}
4770 
4771 	rte_free(bp->ctx);
4772 	bp->ctx = NULL;
4773 }
4774 
4775 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4776 
4777 #define min_t(type, x, y) ({                    \
4778 	type __min1 = (x);                      \
4779 	type __min2 = (y);                      \
4780 	__min1 < __min2 ? __min1 : __min2; })
4781 
4782 #define max_t(type, x, y) ({                    \
4783 	type __max1 = (x);                      \
4784 	type __max2 = (y);                      \
4785 	__max1 > __max2 ? __max1 : __max2; })
4786 
4787 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4788 
4789 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4790 {
4791 	struct bnxt_ctx_pg_info *ctx_pg;
4792 	struct bnxt_ctx_mem_info *ctx;
4793 	uint32_t mem_size, ena, entries;
4794 	uint32_t entries_sp, min;
4795 	int i, rc;
4796 
4797 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4798 	if (rc) {
4799 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4800 		return rc;
4801 	}
4802 	ctx = bp->ctx;
4803 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4804 		return 0;
4805 
4806 	ctx_pg = &ctx->qp_mem;
4807 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4808 	if (ctx->qp_entry_size) {
4809 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4810 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4811 		if (rc)
4812 			return rc;
4813 	}
4814 
4815 	ctx_pg = &ctx->srq_mem;
4816 	ctx_pg->entries = ctx->srq_max_l2_entries;
4817 	if (ctx->srq_entry_size) {
4818 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4819 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4820 		if (rc)
4821 			return rc;
4822 	}
4823 
4824 	ctx_pg = &ctx->cq_mem;
4825 	ctx_pg->entries = ctx->cq_max_l2_entries;
4826 	if (ctx->cq_entry_size) {
4827 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4828 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4829 		if (rc)
4830 			return rc;
4831 	}
4832 
4833 	ctx_pg = &ctx->vnic_mem;
4834 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4835 		ctx->vnic_max_ring_table_entries;
4836 	if (ctx->vnic_entry_size) {
4837 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4838 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4839 		if (rc)
4840 			return rc;
4841 	}
4842 
4843 	ctx_pg = &ctx->stat_mem;
4844 	ctx_pg->entries = ctx->stat_max_entries;
4845 	if (ctx->stat_entry_size) {
4846 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4847 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4848 		if (rc)
4849 			return rc;
4850 	}
4851 
4852 	min = ctx->tqm_min_entries_per_ring;
4853 
4854 	entries_sp = ctx->qp_max_l2_entries +
4855 		     ctx->vnic_max_vnic_entries +
4856 		     2 * ctx->qp_min_qp1_entries + min;
4857 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4858 
4859 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4860 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4861 	entries = clamp_t(uint32_t, entries, min,
4862 			  ctx->tqm_max_entries_per_ring);
4863 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4864 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4865 		 * i > 8 is other ext rings.
4866 		 */
4867 		ctx_pg = ctx->tqm_mem[i];
4868 		ctx_pg->entries = i ? entries : entries_sp;
4869 		if (ctx->tqm_entry_size) {
4870 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4871 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4872 						    "tqm_mem", i);
4873 			if (rc)
4874 				return rc;
4875 		}
4876 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4877 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4878 		else
4879 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4880 	}
4881 
4882 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4883 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4884 	if (rc)
4885 		PMD_DRV_LOG(ERR,
4886 			    "Failed to configure context mem: rc = %d\n", rc);
4887 	else
4888 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4889 
4890 	return rc;
4891 }
4892 
4893 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4894 {
4895 	struct rte_pci_device *pci_dev = bp->pdev;
4896 	char mz_name[RTE_MEMZONE_NAMESIZE];
4897 	const struct rte_memzone *mz = NULL;
4898 	uint32_t total_alloc_len;
4899 	rte_iova_t mz_phys_addr;
4900 
4901 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4902 		return 0;
4903 
4904 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4905 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4906 		 pci_dev->addr.bus, pci_dev->addr.devid,
4907 		 pci_dev->addr.function, "rx_port_stats");
4908 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4909 	mz = rte_memzone_lookup(mz_name);
4910 	total_alloc_len =
4911 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4912 				       sizeof(struct rx_port_stats_ext) + 512);
4913 	if (!mz) {
4914 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4915 					 SOCKET_ID_ANY,
4916 					 RTE_MEMZONE_2MB |
4917 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4918 					 RTE_MEMZONE_IOVA_CONTIG);
4919 		if (mz == NULL)
4920 			return -ENOMEM;
4921 	}
4922 	memset(mz->addr, 0, mz->len);
4923 	mz_phys_addr = mz->iova;
4924 
4925 	bp->rx_mem_zone = (const void *)mz;
4926 	bp->hw_rx_port_stats = mz->addr;
4927 	bp->hw_rx_port_stats_map = mz_phys_addr;
4928 
4929 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4930 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4931 		 pci_dev->addr.bus, pci_dev->addr.devid,
4932 		 pci_dev->addr.function, "tx_port_stats");
4933 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4934 	mz = rte_memzone_lookup(mz_name);
4935 	total_alloc_len =
4936 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4937 				       sizeof(struct tx_port_stats_ext) + 512);
4938 	if (!mz) {
4939 		mz = rte_memzone_reserve(mz_name,
4940 					 total_alloc_len,
4941 					 SOCKET_ID_ANY,
4942 					 RTE_MEMZONE_2MB |
4943 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4944 					 RTE_MEMZONE_IOVA_CONTIG);
4945 		if (mz == NULL)
4946 			return -ENOMEM;
4947 	}
4948 	memset(mz->addr, 0, mz->len);
4949 	mz_phys_addr = mz->iova;
4950 
4951 	bp->tx_mem_zone = (const void *)mz;
4952 	bp->hw_tx_port_stats = mz->addr;
4953 	bp->hw_tx_port_stats_map = mz_phys_addr;
4954 	bp->flags |= BNXT_FLAG_PORT_STATS;
4955 
4956 	/* Display extended statistics if FW supports it */
4957 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4958 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4959 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4960 		return 0;
4961 
4962 	bp->hw_rx_port_stats_ext = (void *)
4963 		((uint8_t *)bp->hw_rx_port_stats +
4964 		 sizeof(struct rx_port_stats));
4965 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4966 		sizeof(struct rx_port_stats);
4967 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4968 
4969 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4970 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4971 		bp->hw_tx_port_stats_ext = (void *)
4972 			((uint8_t *)bp->hw_tx_port_stats +
4973 			 sizeof(struct tx_port_stats));
4974 		bp->hw_tx_port_stats_ext_map =
4975 			bp->hw_tx_port_stats_map +
4976 			sizeof(struct tx_port_stats);
4977 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4978 	}
4979 
4980 	return 0;
4981 }
4982 
4983 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4984 {
4985 	struct bnxt *bp = eth_dev->data->dev_private;
4986 	int rc = 0;
4987 
4988 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4989 					       RTE_ETHER_ADDR_LEN *
4990 					       bp->max_l2_ctx,
4991 					       0);
4992 	if (eth_dev->data->mac_addrs == NULL) {
4993 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4994 		return -ENOMEM;
4995 	}
4996 
4997 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4998 		if (BNXT_PF(bp))
4999 			return -EINVAL;
5000 
5001 		/* Generate a random MAC address, if none was assigned by PF */
5002 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
5003 		bnxt_eth_hw_addr_random(bp->mac_addr);
5004 		PMD_DRV_LOG(INFO,
5005 			    "Assign random MAC:" RTE_ETHER_ADDR_PRT_FMT "\n",
5006 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
5007 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
5008 
5009 		rc = bnxt_hwrm_set_mac(bp);
5010 		if (rc)
5011 			return rc;
5012 	}
5013 
5014 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5015 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5016 
5017 	return rc;
5018 }
5019 
5020 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5021 {
5022 	int rc = 0;
5023 
5024 	/* MAC is already configured in FW */
5025 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5026 		return 0;
5027 
5028 	/* Restore the old MAC configured */
5029 	rc = bnxt_hwrm_set_mac(bp);
5030 	if (rc)
5031 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5032 
5033 	return rc;
5034 }
5035 
5036 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5037 {
5038 	if (!BNXT_PF(bp))
5039 		return;
5040 
5041 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5042 
5043 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5044 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5045 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5046 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5047 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5048 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5049 }
5050 
5051 struct bnxt *
5052 bnxt_get_bp(uint16_t port)
5053 {
5054 	struct bnxt *bp;
5055 	struct rte_eth_dev *dev;
5056 
5057 	if (!rte_eth_dev_is_valid_port(port)) {
5058 		PMD_DRV_LOG(ERR, "Invalid port %d\n", port);
5059 		return NULL;
5060 	}
5061 
5062 	dev = &rte_eth_devices[port];
5063 	if (!is_bnxt_supported(dev)) {
5064 		PMD_DRV_LOG(ERR, "Device %d not supported\n", port);
5065 		return NULL;
5066 	}
5067 
5068 	bp = (struct bnxt *)dev->data->dev_private;
5069 	if (!BNXT_TRUFLOW_EN(bp)) {
5070 		PMD_DRV_LOG(ERR, "TRUFLOW not enabled\n");
5071 		return NULL;
5072 	}
5073 
5074 	return bp;
5075 }
5076 
5077 uint16_t
5078 bnxt_get_svif(uint16_t port_id, bool func_svif,
5079 	      enum bnxt_ulp_intf_type type)
5080 {
5081 	struct rte_eth_dev *eth_dev;
5082 	struct bnxt *bp;
5083 
5084 	eth_dev = &rte_eth_devices[port_id];
5085 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5086 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5087 		if (!vfr)
5088 			return 0;
5089 
5090 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5091 			return vfr->svif;
5092 
5093 		eth_dev = vfr->parent_dev;
5094 	}
5095 
5096 	bp = eth_dev->data->dev_private;
5097 
5098 	return func_svif ? bp->func_svif : bp->port_svif;
5099 }
5100 
5101 void
5102 bnxt_get_iface_mac(uint16_t port, enum bnxt_ulp_intf_type type,
5103 		   uint8_t *mac, uint8_t *parent_mac)
5104 {
5105 	struct rte_eth_dev *eth_dev;
5106 	struct bnxt *bp;
5107 
5108 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF &&
5109 	    type != BNXT_ULP_INTF_TYPE_PF)
5110 		return;
5111 
5112 	eth_dev = &rte_eth_devices[port];
5113 	bp = eth_dev->data->dev_private;
5114 	memcpy(mac, bp->mac_addr, RTE_ETHER_ADDR_LEN);
5115 
5116 	if (type == BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5117 		memcpy(parent_mac, bp->parent->mac_addr, RTE_ETHER_ADDR_LEN);
5118 }
5119 
5120 uint16_t
5121 bnxt_get_parent_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5122 {
5123 	struct rte_eth_dev *eth_dev;
5124 	struct bnxt *bp;
5125 
5126 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5127 		return 0;
5128 
5129 	eth_dev = &rte_eth_devices[port];
5130 	bp = eth_dev->data->dev_private;
5131 
5132 	return bp->parent->vnic;
5133 }
5134 uint16_t
5135 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5136 {
5137 	struct rte_eth_dev *eth_dev;
5138 	struct bnxt_vnic_info *vnic;
5139 	struct bnxt *bp;
5140 
5141 	eth_dev = &rte_eth_devices[port];
5142 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5143 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5144 		if (!vfr)
5145 			return 0;
5146 
5147 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5148 			return vfr->dflt_vnic_id;
5149 
5150 		eth_dev = vfr->parent_dev;
5151 	}
5152 
5153 	bp = eth_dev->data->dev_private;
5154 
5155 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
5156 
5157 	return vnic->fw_vnic_id;
5158 }
5159 
5160 uint16_t
5161 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
5162 {
5163 	struct rte_eth_dev *eth_dev;
5164 	struct bnxt *bp;
5165 
5166 	eth_dev = &rte_eth_devices[port];
5167 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5168 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5169 		if (!vfr)
5170 			return 0;
5171 
5172 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5173 			return vfr->fw_fid;
5174 
5175 		eth_dev = vfr->parent_dev;
5176 	}
5177 
5178 	bp = eth_dev->data->dev_private;
5179 
5180 	return bp->fw_fid;
5181 }
5182 
5183 enum bnxt_ulp_intf_type
5184 bnxt_get_interface_type(uint16_t port)
5185 {
5186 	struct rte_eth_dev *eth_dev;
5187 	struct bnxt *bp;
5188 
5189 	eth_dev = &rte_eth_devices[port];
5190 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
5191 		return BNXT_ULP_INTF_TYPE_VF_REP;
5192 
5193 	bp = eth_dev->data->dev_private;
5194 	if (BNXT_PF(bp))
5195 		return BNXT_ULP_INTF_TYPE_PF;
5196 	else if (BNXT_VF_IS_TRUSTED(bp))
5197 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
5198 	else if (BNXT_VF(bp))
5199 		return BNXT_ULP_INTF_TYPE_VF;
5200 
5201 	return BNXT_ULP_INTF_TYPE_INVALID;
5202 }
5203 
5204 uint16_t
5205 bnxt_get_phy_port_id(uint16_t port_id)
5206 {
5207 	struct bnxt_representor *vfr;
5208 	struct rte_eth_dev *eth_dev;
5209 	struct bnxt *bp;
5210 
5211 	eth_dev = &rte_eth_devices[port_id];
5212 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5213 		vfr = eth_dev->data->dev_private;
5214 		if (!vfr)
5215 			return 0;
5216 
5217 		eth_dev = vfr->parent_dev;
5218 	}
5219 
5220 	bp = eth_dev->data->dev_private;
5221 
5222 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
5223 }
5224 
5225 uint16_t
5226 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
5227 {
5228 	struct rte_eth_dev *eth_dev;
5229 	struct bnxt *bp;
5230 
5231 	eth_dev = &rte_eth_devices[port_id];
5232 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5233 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5234 		if (!vfr)
5235 			return 0;
5236 
5237 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5238 			return vfr->fw_fid - 1;
5239 
5240 		eth_dev = vfr->parent_dev;
5241 	}
5242 
5243 	bp = eth_dev->data->dev_private;
5244 
5245 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
5246 }
5247 
5248 uint16_t
5249 bnxt_get_vport(uint16_t port_id)
5250 {
5251 	return (1 << bnxt_get_phy_port_id(port_id));
5252 }
5253 
5254 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5255 {
5256 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5257 
5258 	if (info) {
5259 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5260 			memset(info, 0, sizeof(*info));
5261 		return;
5262 	}
5263 
5264 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5265 		return;
5266 
5267 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5268 			   sizeof(*info), 0);
5269 	if (!info)
5270 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5271 
5272 	bp->recovery_info = info;
5273 }
5274 
5275 static void bnxt_check_fw_status(struct bnxt *bp)
5276 {
5277 	uint32_t fw_status;
5278 
5279 	if (!(bp->recovery_info &&
5280 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5281 		return;
5282 
5283 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5284 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5285 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5286 			    fw_status);
5287 }
5288 
5289 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5290 {
5291 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5292 	uint32_t status_loc;
5293 	uint32_t sig_ver;
5294 
5295 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5296 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5297 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5298 				   BNXT_GRCP_WINDOW_2_BASE +
5299 				   offsetof(struct hcomm_status,
5300 					    sig_ver)));
5301 	/* If the signature is absent, then FW does not support this feature */
5302 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5303 	    HCOMM_STATUS_SIGNATURE_VAL)
5304 		return 0;
5305 
5306 	if (!info) {
5307 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5308 				   sizeof(*info), 0);
5309 		if (!info)
5310 			return -ENOMEM;
5311 		bp->recovery_info = info;
5312 	} else {
5313 		memset(info, 0, sizeof(*info));
5314 	}
5315 
5316 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5317 				      BNXT_GRCP_WINDOW_2_BASE +
5318 				      offsetof(struct hcomm_status,
5319 					       fw_status_loc)));
5320 
5321 	/* Only pre-map the FW health status GRC register */
5322 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5323 		return 0;
5324 
5325 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5326 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5327 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5328 
5329 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5330 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5331 
5332 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5333 
5334 	return 0;
5335 }
5336 
5337 /* This function gets the FW version along with the
5338  * capabilities(MAX and current) of the function, vnic,
5339  * error recovery, phy and other chip related info
5340  */
5341 static int bnxt_get_config(struct bnxt *bp)
5342 {
5343 	uint16_t mtu;
5344 	int rc = 0;
5345 
5346 	bp->fw_cap = 0;
5347 
5348 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5349 	if (rc)
5350 		return rc;
5351 
5352 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5353 	if (rc) {
5354 		bnxt_check_fw_status(bp);
5355 		return rc;
5356 	}
5357 
5358 	rc = bnxt_hwrm_func_reset(bp);
5359 	if (rc)
5360 		return -EIO;
5361 
5362 	rc = bnxt_hwrm_vnic_qcaps(bp);
5363 	if (rc)
5364 		return rc;
5365 
5366 	rc = bnxt_hwrm_queue_qportcfg(bp);
5367 	if (rc)
5368 		return rc;
5369 
5370 	/* Get the MAX capabilities for this function.
5371 	 * This function also allocates context memory for TQM rings and
5372 	 * informs the firmware about this allocated backing store memory.
5373 	 */
5374 	rc = bnxt_hwrm_func_qcaps(bp);
5375 	if (rc)
5376 		return rc;
5377 
5378 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5379 	if (rc)
5380 		return rc;
5381 
5382 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5383 	if (rc)
5384 		return rc;
5385 
5386 	bnxt_hwrm_port_mac_qcfg(bp);
5387 
5388 	bnxt_hwrm_parent_pf_qcfg(bp);
5389 
5390 	bnxt_hwrm_port_phy_qcaps(bp);
5391 
5392 	bnxt_alloc_error_recovery_info(bp);
5393 	/* Get the adapter error recovery support info */
5394 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5395 	if (rc)
5396 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5397 
5398 	bnxt_hwrm_port_led_qcaps(bp);
5399 
5400 	return 0;
5401 }
5402 
5403 static int
5404 bnxt_init_locks(struct bnxt *bp)
5405 {
5406 	int err;
5407 
5408 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5409 	if (err) {
5410 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5411 		return err;
5412 	}
5413 
5414 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5415 	if (err) {
5416 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5417 		return err;
5418 	}
5419 
5420 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5421 	if (err) {
5422 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5423 		return err;
5424 	}
5425 
5426 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5427 	if (err)
5428 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5429 
5430 	return err;
5431 }
5432 
5433 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5434 {
5435 	int rc = 0;
5436 
5437 	rc = bnxt_get_config(bp);
5438 	if (rc)
5439 		return rc;
5440 
5441 	if (!reconfig_dev) {
5442 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5443 		if (rc)
5444 			return rc;
5445 	} else {
5446 		rc = bnxt_restore_dflt_mac(bp);
5447 		if (rc)
5448 			return rc;
5449 	}
5450 
5451 	bnxt_config_vf_req_fwd(bp);
5452 
5453 	rc = bnxt_hwrm_func_driver_register(bp);
5454 	if (rc) {
5455 		PMD_DRV_LOG(ERR, "Failed to register driver");
5456 		return -EBUSY;
5457 	}
5458 
5459 	if (BNXT_PF(bp)) {
5460 		if (bp->pdev->max_vfs) {
5461 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5462 			if (rc) {
5463 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5464 				return rc;
5465 			}
5466 		} else {
5467 			rc = bnxt_hwrm_allocate_pf_only(bp);
5468 			if (rc) {
5469 				PMD_DRV_LOG(ERR,
5470 					    "Failed to allocate PF resources");
5471 				return rc;
5472 			}
5473 		}
5474 	}
5475 
5476 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5477 	if (rc)
5478 		return rc;
5479 
5480 	rc = bnxt_setup_int(bp);
5481 	if (rc)
5482 		return rc;
5483 
5484 	rc = bnxt_request_int(bp);
5485 	if (rc)
5486 		return rc;
5487 
5488 	rc = bnxt_init_ctx_mem(bp);
5489 	if (rc) {
5490 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5491 		return rc;
5492 	}
5493 
5494 	return 0;
5495 }
5496 
5497 static int
5498 bnxt_parse_devarg_accum_stats(__rte_unused const char *key,
5499 			      const char *value, void *opaque_arg)
5500 {
5501 	struct bnxt *bp = opaque_arg;
5502 	unsigned long accum_stats;
5503 	char *end = NULL;
5504 
5505 	if (!value || !opaque_arg) {
5506 		PMD_DRV_LOG(ERR,
5507 			    "Invalid parameter passed to accum-stats devargs.\n");
5508 		return -EINVAL;
5509 	}
5510 
5511 	accum_stats = strtoul(value, &end, 10);
5512 	if (end == NULL || *end != '\0' ||
5513 	    (accum_stats == ULONG_MAX && errno == ERANGE)) {
5514 		PMD_DRV_LOG(ERR,
5515 			    "Invalid parameter passed to accum-stats devargs.\n");
5516 		return -EINVAL;
5517 	}
5518 
5519 	if (BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)) {
5520 		PMD_DRV_LOG(ERR,
5521 			    "Invalid value passed to accum-stats devargs.\n");
5522 		return -EINVAL;
5523 	}
5524 
5525 	if (accum_stats) {
5526 		bp->flags2 |= BNXT_FLAGS2_ACCUM_STATS_EN;
5527 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature enabled.\n");
5528 	} else {
5529 		bp->flags2 &= ~BNXT_FLAGS2_ACCUM_STATS_EN;
5530 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature disabled.\n");
5531 	}
5532 
5533 	return 0;
5534 }
5535 
5536 static int
5537 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5538 			     const char *value, void *opaque_arg)
5539 {
5540 	struct bnxt *bp = opaque_arg;
5541 	unsigned long flow_xstat;
5542 	char *end = NULL;
5543 
5544 	if (!value || !opaque_arg) {
5545 		PMD_DRV_LOG(ERR,
5546 			    "Invalid parameter passed to flow_xstat devarg.\n");
5547 		return -EINVAL;
5548 	}
5549 
5550 	flow_xstat = strtoul(value, &end, 10);
5551 	if (end == NULL || *end != '\0' ||
5552 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5553 		PMD_DRV_LOG(ERR,
5554 			    "Invalid parameter passed to flow_xstat devarg.\n");
5555 		return -EINVAL;
5556 	}
5557 
5558 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5559 		PMD_DRV_LOG(ERR,
5560 			    "Invalid value passed to flow_xstat devarg.\n");
5561 		return -EINVAL;
5562 	}
5563 
5564 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5565 	if (BNXT_FLOW_XSTATS_EN(bp))
5566 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5567 
5568 	return 0;
5569 }
5570 
5571 static int
5572 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5573 					const char *value, void *opaque_arg)
5574 {
5575 	struct bnxt *bp = opaque_arg;
5576 	unsigned long max_num_kflows;
5577 	char *end = NULL;
5578 
5579 	if (!value || !opaque_arg) {
5580 		PMD_DRV_LOG(ERR,
5581 			"Invalid parameter passed to max_num_kflows devarg.\n");
5582 		return -EINVAL;
5583 	}
5584 
5585 	max_num_kflows = strtoul(value, &end, 10);
5586 	if (end == NULL || *end != '\0' ||
5587 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5588 		PMD_DRV_LOG(ERR,
5589 			"Invalid parameter passed to max_num_kflows devarg.\n");
5590 		return -EINVAL;
5591 	}
5592 
5593 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5594 		PMD_DRV_LOG(ERR,
5595 			"Invalid value passed to max_num_kflows devarg.\n");
5596 		return -EINVAL;
5597 	}
5598 
5599 	bp->max_num_kflows = max_num_kflows;
5600 	if (bp->max_num_kflows)
5601 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5602 				max_num_kflows);
5603 
5604 	return 0;
5605 }
5606 
5607 static int
5608 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5609 				 const char *value, void *opaque_arg)
5610 {
5611 	struct bnxt *bp = opaque_arg;
5612 	unsigned long app_id;
5613 	char *end = NULL;
5614 
5615 	if (!value || !opaque_arg) {
5616 		PMD_DRV_LOG(ERR,
5617 			    "Invalid parameter passed to app-id "
5618 			    "devargs.\n");
5619 		return -EINVAL;
5620 	}
5621 
5622 	app_id = strtoul(value, &end, 10);
5623 	if (end == NULL || *end != '\0' ||
5624 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5625 		PMD_DRV_LOG(ERR,
5626 			    "Invalid parameter passed to app_id "
5627 			    "devargs.\n");
5628 		return -EINVAL;
5629 	}
5630 
5631 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5632 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5633 			    (uint16_t)app_id);
5634 		return -EINVAL;
5635 	}
5636 
5637 	bp->app_id = app_id;
5638 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5639 
5640 	return 0;
5641 }
5642 
5643 static int
5644 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5645 			    const char *value, void *opaque_arg)
5646 {
5647 	struct bnxt_representor *vfr_bp = opaque_arg;
5648 	unsigned long rep_is_pf;
5649 	char *end = NULL;
5650 
5651 	if (!value || !opaque_arg) {
5652 		PMD_DRV_LOG(ERR,
5653 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5654 		return -EINVAL;
5655 	}
5656 
5657 	rep_is_pf = strtoul(value, &end, 10);
5658 	if (end == NULL || *end != '\0' ||
5659 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5660 		PMD_DRV_LOG(ERR,
5661 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5662 		return -EINVAL;
5663 	}
5664 
5665 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5666 		PMD_DRV_LOG(ERR,
5667 			    "Invalid value passed to rep_is_pf devargs.\n");
5668 		return -EINVAL;
5669 	}
5670 
5671 	vfr_bp->flags |= rep_is_pf;
5672 	if (BNXT_REP_PF(vfr_bp))
5673 		PMD_DRV_LOG(INFO, "PF representor\n");
5674 	else
5675 		PMD_DRV_LOG(INFO, "VF representor\n");
5676 
5677 	return 0;
5678 }
5679 
5680 static int
5681 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5682 			       const char *value, void *opaque_arg)
5683 {
5684 	struct bnxt_representor *vfr_bp = opaque_arg;
5685 	unsigned long rep_based_pf;
5686 	char *end = NULL;
5687 
5688 	if (!value || !opaque_arg) {
5689 		PMD_DRV_LOG(ERR,
5690 			    "Invalid parameter passed to rep_based_pf "
5691 			    "devargs.\n");
5692 		return -EINVAL;
5693 	}
5694 
5695 	rep_based_pf = strtoul(value, &end, 10);
5696 	if (end == NULL || *end != '\0' ||
5697 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5698 		PMD_DRV_LOG(ERR,
5699 			    "Invalid parameter passed to rep_based_pf "
5700 			    "devargs.\n");
5701 		return -EINVAL;
5702 	}
5703 
5704 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5705 		PMD_DRV_LOG(ERR,
5706 			    "Invalid value passed to rep_based_pf devargs.\n");
5707 		return -EINVAL;
5708 	}
5709 
5710 	vfr_bp->rep_based_pf = rep_based_pf;
5711 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5712 
5713 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5714 
5715 	return 0;
5716 }
5717 
5718 static int
5719 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5720 			    const char *value, void *opaque_arg)
5721 {
5722 	struct bnxt_representor *vfr_bp = opaque_arg;
5723 	unsigned long rep_q_r2f;
5724 	char *end = NULL;
5725 
5726 	if (!value || !opaque_arg) {
5727 		PMD_DRV_LOG(ERR,
5728 			    "Invalid parameter passed to rep_q_r2f "
5729 			    "devargs.\n");
5730 		return -EINVAL;
5731 	}
5732 
5733 	rep_q_r2f = strtoul(value, &end, 10);
5734 	if (end == NULL || *end != '\0' ||
5735 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5736 		PMD_DRV_LOG(ERR,
5737 			    "Invalid parameter passed to rep_q_r2f "
5738 			    "devargs.\n");
5739 		return -EINVAL;
5740 	}
5741 
5742 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5743 		PMD_DRV_LOG(ERR,
5744 			    "Invalid value passed to rep_q_r2f devargs.\n");
5745 		return -EINVAL;
5746 	}
5747 
5748 	vfr_bp->rep_q_r2f = rep_q_r2f;
5749 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5750 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5751 
5752 	return 0;
5753 }
5754 
5755 static int
5756 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5757 			    const char *value, void *opaque_arg)
5758 {
5759 	struct bnxt_representor *vfr_bp = opaque_arg;
5760 	unsigned long rep_q_f2r;
5761 	char *end = NULL;
5762 
5763 	if (!value || !opaque_arg) {
5764 		PMD_DRV_LOG(ERR,
5765 			    "Invalid parameter passed to rep_q_f2r "
5766 			    "devargs.\n");
5767 		return -EINVAL;
5768 	}
5769 
5770 	rep_q_f2r = strtoul(value, &end, 10);
5771 	if (end == NULL || *end != '\0' ||
5772 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5773 		PMD_DRV_LOG(ERR,
5774 			    "Invalid parameter passed to rep_q_f2r "
5775 			    "devargs.\n");
5776 		return -EINVAL;
5777 	}
5778 
5779 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5780 		PMD_DRV_LOG(ERR,
5781 			    "Invalid value passed to rep_q_f2r devargs.\n");
5782 		return -EINVAL;
5783 	}
5784 
5785 	vfr_bp->rep_q_f2r = rep_q_f2r;
5786 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5787 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5788 
5789 	return 0;
5790 }
5791 
5792 static int
5793 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5794 			     const char *value, void *opaque_arg)
5795 {
5796 	struct bnxt_representor *vfr_bp = opaque_arg;
5797 	unsigned long rep_fc_r2f;
5798 	char *end = NULL;
5799 
5800 	if (!value || !opaque_arg) {
5801 		PMD_DRV_LOG(ERR,
5802 			    "Invalid parameter passed to rep_fc_r2f "
5803 			    "devargs.\n");
5804 		return -EINVAL;
5805 	}
5806 
5807 	rep_fc_r2f = strtoul(value, &end, 10);
5808 	if (end == NULL || *end != '\0' ||
5809 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5810 		PMD_DRV_LOG(ERR,
5811 			    "Invalid parameter passed to rep_fc_r2f "
5812 			    "devargs.\n");
5813 		return -EINVAL;
5814 	}
5815 
5816 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5817 		PMD_DRV_LOG(ERR,
5818 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5819 		return -EINVAL;
5820 	}
5821 
5822 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5823 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5824 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5825 
5826 	return 0;
5827 }
5828 
5829 static int
5830 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5831 			     const char *value, void *opaque_arg)
5832 {
5833 	struct bnxt_representor *vfr_bp = opaque_arg;
5834 	unsigned long rep_fc_f2r;
5835 	char *end = NULL;
5836 
5837 	if (!value || !opaque_arg) {
5838 		PMD_DRV_LOG(ERR,
5839 			    "Invalid parameter passed to rep_fc_f2r "
5840 			    "devargs.\n");
5841 		return -EINVAL;
5842 	}
5843 
5844 	rep_fc_f2r = strtoul(value, &end, 10);
5845 	if (end == NULL || *end != '\0' ||
5846 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5847 		PMD_DRV_LOG(ERR,
5848 			    "Invalid parameter passed to rep_fc_f2r "
5849 			    "devargs.\n");
5850 		return -EINVAL;
5851 	}
5852 
5853 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5854 		PMD_DRV_LOG(ERR,
5855 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5856 		return -EINVAL;
5857 	}
5858 
5859 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5860 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5861 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5862 
5863 	return 0;
5864 }
5865 
5866 static int
5867 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5868 {
5869 	struct rte_kvargs *kvlist;
5870 	int ret;
5871 
5872 	if (devargs == NULL)
5873 		return 0;
5874 
5875 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5876 	if (kvlist == NULL)
5877 		return -EINVAL;
5878 
5879 	/*
5880 	 * Handler for "flow_xstat" devarg.
5881 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5882 	 */
5883 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5884 				 bnxt_parse_devarg_flow_xstat, bp);
5885 	if (ret)
5886 		goto err;
5887 
5888 	/*
5889 	 * Handler for "accum-stats" devarg.
5890 	 * Invoked as for ex: "-a 0000:00:0d.0,accum-stats=1"
5891 	 */
5892 	rte_kvargs_process(kvlist, BNXT_DEVARG_ACCUM_STATS,
5893 			   bnxt_parse_devarg_accum_stats, bp);
5894 	/*
5895 	 * Handler for "max_num_kflows" devarg.
5896 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5897 	 */
5898 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5899 				 bnxt_parse_devarg_max_num_kflows, bp);
5900 	if (ret)
5901 		goto err;
5902 
5903 err:
5904 	/*
5905 	 * Handler for "app-id" devarg.
5906 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5907 	 */
5908 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5909 			   bnxt_parse_devarg_app_id, bp);
5910 
5911 	rte_kvargs_free(kvlist);
5912 	return ret;
5913 }
5914 
5915 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5916 {
5917 	int rc = 0;
5918 
5919 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5920 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5921 		if (rc)
5922 			PMD_DRV_LOG(ERR,
5923 				    "Failed to alloc switch domain: %d\n", rc);
5924 		else
5925 			PMD_DRV_LOG(INFO,
5926 				    "Switch domain allocated %d\n",
5927 				    bp->switch_domain_id);
5928 	}
5929 
5930 	return rc;
5931 }
5932 
5933 /* Allocate and initialize various fields in bnxt struct that
5934  * need to be allocated/destroyed only once in the lifetime of the driver
5935  */
5936 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5937 {
5938 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5939 	struct bnxt *bp = eth_dev->data->dev_private;
5940 	int rc = 0;
5941 
5942 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5943 
5944 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5945 		bp->flags |= BNXT_FLAG_VF;
5946 
5947 	if (bnxt_p5_device(pci_dev->id.device_id))
5948 		bp->flags |= BNXT_FLAG_CHIP_P5;
5949 
5950 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5951 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5952 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5953 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5954 		bp->flags |= BNXT_FLAG_STINGRAY;
5955 
5956 	if (BNXT_TRUFLOW_EN(bp)) {
5957 		/* extra mbuf field is required to store CFA code from mark */
5958 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5959 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5960 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5961 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5962 		};
5963 		bnxt_cfa_code_dynfield_offset =
5964 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5965 		if (bnxt_cfa_code_dynfield_offset < 0) {
5966 			PMD_DRV_LOG(ERR,
5967 			    "Failed to register mbuf field for TruFlow mark\n");
5968 			return -rte_errno;
5969 		}
5970 	}
5971 
5972 	rc = bnxt_map_pci_bars(eth_dev);
5973 	if (rc) {
5974 		PMD_DRV_LOG(ERR,
5975 			    "Failed to initialize board rc: %x\n", rc);
5976 		return rc;
5977 	}
5978 
5979 	rc = bnxt_alloc_pf_info(bp);
5980 	if (rc)
5981 		return rc;
5982 
5983 	rc = bnxt_alloc_link_info(bp);
5984 	if (rc)
5985 		return rc;
5986 
5987 	rc = bnxt_alloc_parent_info(bp);
5988 	if (rc)
5989 		return rc;
5990 
5991 	rc = bnxt_alloc_hwrm_resources(bp);
5992 	if (rc) {
5993 		PMD_DRV_LOG(ERR,
5994 			    "Failed to allocate response buffer rc: %x\n", rc);
5995 		return rc;
5996 	}
5997 	rc = bnxt_alloc_leds_info(bp);
5998 	if (rc)
5999 		return rc;
6000 
6001 	rc = bnxt_alloc_cos_queues(bp);
6002 	if (rc)
6003 		return rc;
6004 
6005 	rc = bnxt_init_locks(bp);
6006 	if (rc)
6007 		return rc;
6008 
6009 	rc = bnxt_alloc_switch_domain(bp);
6010 	if (rc)
6011 		return rc;
6012 
6013 	return rc;
6014 }
6015 
6016 static int
6017 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
6018 {
6019 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
6020 	static int version_printed;
6021 	struct bnxt *bp;
6022 	int rc;
6023 
6024 	if (version_printed++ == 0)
6025 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
6026 
6027 	eth_dev->dev_ops = &bnxt_dev_ops;
6028 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
6029 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
6030 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
6031 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
6032 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
6033 
6034 	/*
6035 	 * For secondary processes, we don't initialise any further
6036 	 * as primary has already done this work.
6037 	 */
6038 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6039 		return 0;
6040 
6041 	rte_eth_copy_pci_info(eth_dev, pci_dev);
6042 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
6043 
6044 	bp = eth_dev->data->dev_private;
6045 
6046 	/* Parse dev arguments passed on when starting the DPDK application. */
6047 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
6048 	if (rc)
6049 		goto error_free;
6050 
6051 	rc = bnxt_drv_init(eth_dev);
6052 	if (rc)
6053 		goto error_free;
6054 
6055 	rc = bnxt_init_resources(bp, false);
6056 	if (rc)
6057 		goto error_free;
6058 
6059 	rc = bnxt_alloc_stats_mem(bp);
6060 	if (rc)
6061 		goto error_free;
6062 
6063 	PMD_DRV_LOG(INFO,
6064 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
6065 		    DRV_MODULE_NAME,
6066 		    pci_dev->mem_resource[0].phys_addr,
6067 		    pci_dev->mem_resource[0].addr);
6068 
6069 	return 0;
6070 
6071 error_free:
6072 	bnxt_dev_uninit(eth_dev);
6073 	return rc;
6074 }
6075 
6076 
6077 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
6078 {
6079 	if (!ctx)
6080 		return;
6081 
6082 	if (ctx->va)
6083 		rte_free(ctx->va);
6084 
6085 	ctx->va = NULL;
6086 	ctx->dma = RTE_BAD_IOVA;
6087 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
6088 }
6089 
6090 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
6091 {
6092 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
6093 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6094 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
6095 				  bp->flow_stat->max_fc,
6096 				  false);
6097 
6098 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
6099 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6100 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
6101 				  bp->flow_stat->max_fc,
6102 				  false);
6103 
6104 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6105 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
6106 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6107 
6108 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6109 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
6110 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6111 
6112 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6113 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
6114 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6115 
6116 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6117 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
6118 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6119 }
6120 
6121 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
6122 {
6123 	bnxt_unregister_fc_ctx_mem(bp);
6124 
6125 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
6126 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
6127 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
6128 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
6129 }
6130 
6131 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
6132 {
6133 	if (BNXT_FLOW_XSTATS_EN(bp))
6134 		bnxt_uninit_fc_ctx_mem(bp);
6135 }
6136 
6137 static void
6138 bnxt_free_error_recovery_info(struct bnxt *bp)
6139 {
6140 	rte_free(bp->recovery_info);
6141 	bp->recovery_info = NULL;
6142 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
6143 }
6144 
6145 static int
6146 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
6147 {
6148 	int rc;
6149 
6150 	bnxt_free_int(bp);
6151 	bnxt_free_mem(bp, reconfig_dev);
6152 
6153 	bnxt_hwrm_func_buf_unrgtr(bp);
6154 	if (bp->pf != NULL) {
6155 		rte_free(bp->pf->vf_req_buf);
6156 		bp->pf->vf_req_buf = NULL;
6157 	}
6158 
6159 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
6160 	bp->flags &= ~BNXT_FLAG_REGISTERED;
6161 	bnxt_free_ctx_mem(bp);
6162 	if (!reconfig_dev) {
6163 		bnxt_free_hwrm_resources(bp);
6164 		bnxt_free_error_recovery_info(bp);
6165 	}
6166 
6167 	bnxt_uninit_ctx_mem(bp);
6168 
6169 	bnxt_free_flow_stats_info(bp);
6170 	if (bp->rep_info != NULL)
6171 		bnxt_free_switch_domain(bp);
6172 	bnxt_free_rep_info(bp);
6173 	rte_free(bp->ptp_cfg);
6174 	bp->ptp_cfg = NULL;
6175 	return rc;
6176 }
6177 
6178 static int
6179 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
6180 {
6181 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6182 		return -EPERM;
6183 
6184 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
6185 
6186 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
6187 		bnxt_dev_close_op(eth_dev);
6188 
6189 	return 0;
6190 }
6191 
6192 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6193 {
6194 	struct bnxt *bp = eth_dev->data->dev_private;
6195 	struct rte_eth_dev *vf_rep_eth_dev;
6196 	int ret = 0, i;
6197 
6198 	if (!bp)
6199 		return -EINVAL;
6200 
6201 	for (i = 0; i < bp->num_reps; i++) {
6202 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6203 		if (!vf_rep_eth_dev)
6204 			continue;
6205 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6206 			    vf_rep_eth_dev->data->port_id);
6207 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6208 	}
6209 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6210 		    eth_dev->data->port_id);
6211 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6212 
6213 	return ret;
6214 }
6215 
6216 static void bnxt_free_rep_info(struct bnxt *bp)
6217 {
6218 	rte_free(bp->rep_info);
6219 	bp->rep_info = NULL;
6220 	rte_free(bp->cfa_code_map);
6221 	bp->cfa_code_map = NULL;
6222 }
6223 
6224 static int bnxt_init_rep_info(struct bnxt *bp)
6225 {
6226 	int i = 0, rc;
6227 
6228 	if (bp->rep_info)
6229 		return 0;
6230 
6231 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
6232 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
6233 				   0);
6234 	if (!bp->rep_info) {
6235 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6236 		return -ENOMEM;
6237 	}
6238 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6239 				       sizeof(*bp->cfa_code_map) *
6240 				       BNXT_MAX_CFA_CODE, 0);
6241 	if (!bp->cfa_code_map) {
6242 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6243 		bnxt_free_rep_info(bp);
6244 		return -ENOMEM;
6245 	}
6246 
6247 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6248 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6249 
6250 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6251 	if (rc) {
6252 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6253 		bnxt_free_rep_info(bp);
6254 		return rc;
6255 	}
6256 
6257 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6258 	if (rc) {
6259 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6260 		bnxt_free_rep_info(bp);
6261 		return rc;
6262 	}
6263 
6264 	return rc;
6265 }
6266 
6267 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6268 			       struct rte_eth_devargs *eth_da,
6269 			       struct rte_eth_dev *backing_eth_dev,
6270 			       const char *dev_args)
6271 {
6272 	struct rte_eth_dev *vf_rep_eth_dev;
6273 	char name[RTE_ETH_NAME_MAX_LEN];
6274 	struct bnxt *backing_bp;
6275 	uint16_t num_rep;
6276 	int i, ret = 0;
6277 	struct rte_kvargs *kvlist = NULL;
6278 
6279 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6280 		return 0;
6281 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6282 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6283 			    eth_da->type);
6284 		return -ENOTSUP;
6285 	}
6286 	num_rep = eth_da->nb_representor_ports;
6287 	if (num_rep > BNXT_MAX_VF_REPS) {
6288 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6289 			    num_rep, BNXT_MAX_VF_REPS);
6290 		return -EINVAL;
6291 	}
6292 
6293 	if (num_rep >= RTE_MAX_ETHPORTS) {
6294 		PMD_DRV_LOG(ERR,
6295 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6296 			    num_rep, RTE_MAX_ETHPORTS);
6297 		return -EINVAL;
6298 	}
6299 
6300 	backing_bp = backing_eth_dev->data->dev_private;
6301 
6302 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6303 		PMD_DRV_LOG(ERR,
6304 			    "Not a PF or trusted VF. No Representor support\n");
6305 		/* Returning an error is not an option.
6306 		 * Applications are not handling this correctly
6307 		 */
6308 		return 0;
6309 	}
6310 
6311 	if (bnxt_init_rep_info(backing_bp))
6312 		return 0;
6313 
6314 	for (i = 0; i < num_rep; i++) {
6315 		struct bnxt_representor representor = {
6316 			.vf_id = eth_da->representor_ports[i],
6317 			.switch_domain_id = backing_bp->switch_domain_id,
6318 			.parent_dev = backing_eth_dev
6319 		};
6320 
6321 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
6322 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6323 				    representor.vf_id, BNXT_MAX_VF_REPS);
6324 			continue;
6325 		}
6326 
6327 		/* representor port net_bdf_port */
6328 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6329 			 pci_dev->device.name, eth_da->representor_ports[i]);
6330 
6331 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6332 		if (kvlist) {
6333 			/*
6334 			 * Handler for "rep_is_pf" devarg.
6335 			 * Invoked as for ex: "-a 000:00:0d.0,
6336 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6337 			 */
6338 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6339 						 bnxt_parse_devarg_rep_is_pf,
6340 						 (void *)&representor);
6341 			if (ret) {
6342 				ret = -EINVAL;
6343 				goto err;
6344 			}
6345 			/*
6346 			 * Handler for "rep_based_pf" devarg.
6347 			 * Invoked as for ex: "-a 000:00:0d.0,
6348 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6349 			 */
6350 			ret = rte_kvargs_process(kvlist,
6351 						 BNXT_DEVARG_REP_BASED_PF,
6352 						 bnxt_parse_devarg_rep_based_pf,
6353 						 (void *)&representor);
6354 			if (ret) {
6355 				ret = -EINVAL;
6356 				goto err;
6357 			}
6358 			/*
6359 			 * Handler for "rep_based_pf" devarg.
6360 			 * Invoked as for ex: "-a 000:00:0d.0,
6361 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6362 			 */
6363 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6364 						 bnxt_parse_devarg_rep_q_r2f,
6365 						 (void *)&representor);
6366 			if (ret) {
6367 				ret = -EINVAL;
6368 				goto err;
6369 			}
6370 			/*
6371 			 * Handler for "rep_based_pf" devarg.
6372 			 * Invoked as for ex: "-a 000:00:0d.0,
6373 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6374 			 */
6375 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6376 						 bnxt_parse_devarg_rep_q_f2r,
6377 						 (void *)&representor);
6378 			if (ret) {
6379 				ret = -EINVAL;
6380 				goto err;
6381 			}
6382 			/*
6383 			 * Handler for "rep_based_pf" devarg.
6384 			 * Invoked as for ex: "-a 000:00:0d.0,
6385 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6386 			 */
6387 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6388 						 bnxt_parse_devarg_rep_fc_r2f,
6389 						 (void *)&representor);
6390 			if (ret) {
6391 				ret = -EINVAL;
6392 				goto err;
6393 			}
6394 			/*
6395 			 * Handler for "rep_based_pf" devarg.
6396 			 * Invoked as for ex: "-a 000:00:0d.0,
6397 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6398 			 */
6399 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6400 						 bnxt_parse_devarg_rep_fc_f2r,
6401 						 (void *)&representor);
6402 			if (ret) {
6403 				ret = -EINVAL;
6404 				goto err;
6405 			}
6406 		}
6407 
6408 		ret = rte_eth_dev_create(&pci_dev->device, name,
6409 					 sizeof(struct bnxt_representor),
6410 					 NULL, NULL,
6411 					 bnxt_representor_init,
6412 					 &representor);
6413 		if (ret) {
6414 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6415 				    "representor %s.", name);
6416 			goto err;
6417 		}
6418 
6419 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6420 		if (!vf_rep_eth_dev) {
6421 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6422 				    " for VF-Rep: %s.", name);
6423 			ret = -ENODEV;
6424 			goto err;
6425 		}
6426 
6427 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6428 			    backing_eth_dev->data->port_id);
6429 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6430 							 vf_rep_eth_dev;
6431 		backing_bp->num_reps++;
6432 
6433 	}
6434 
6435 	rte_kvargs_free(kvlist);
6436 	return 0;
6437 
6438 err:
6439 	/* If num_rep > 1, then rollback already created
6440 	 * ports, since we'll be failing the probe anyway
6441 	 */
6442 	if (num_rep > 1)
6443 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6444 	rte_errno = -ret;
6445 	rte_kvargs_free(kvlist);
6446 
6447 	return ret;
6448 }
6449 
6450 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6451 			  struct rte_pci_device *pci_dev)
6452 {
6453 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6454 	struct rte_eth_dev *backing_eth_dev;
6455 	uint16_t num_rep;
6456 	int ret = 0;
6457 
6458 	if (pci_dev->device.devargs) {
6459 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6460 					    &eth_da);
6461 		if (ret)
6462 			return ret;
6463 	}
6464 
6465 	num_rep = eth_da.nb_representor_ports;
6466 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6467 		    num_rep);
6468 
6469 	/* We could come here after first level of probe is already invoked
6470 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6471 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6472 	 */
6473 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6474 	if (backing_eth_dev == NULL) {
6475 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6476 					 sizeof(struct bnxt),
6477 					 eth_dev_pci_specific_init, pci_dev,
6478 					 bnxt_dev_init, NULL);
6479 
6480 		if (ret || !num_rep)
6481 			return ret;
6482 
6483 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6484 	}
6485 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6486 		    backing_eth_dev->data->port_id);
6487 
6488 	if (!num_rep)
6489 		return ret;
6490 
6491 	/* probe representor ports now */
6492 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6493 				  pci_dev->device.devargs->args);
6494 
6495 	return ret;
6496 }
6497 
6498 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6499 {
6500 	struct rte_eth_dev *eth_dev;
6501 
6502 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6503 	if (!eth_dev)
6504 		return 0; /* Invoked typically only by OVS-DPDK, by the
6505 			   * time it comes here the eth_dev is already
6506 			   * deleted by rte_eth_dev_close(), so returning
6507 			   * +ve value will at least help in proper cleanup
6508 			   */
6509 
6510 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6511 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6512 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6513 			return rte_eth_dev_destroy(eth_dev,
6514 						   bnxt_representor_uninit);
6515 		else
6516 			return rte_eth_dev_destroy(eth_dev,
6517 						   bnxt_dev_uninit);
6518 	} else {
6519 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6520 	}
6521 }
6522 
6523 static struct rte_pci_driver bnxt_rte_pmd = {
6524 	.id_table = bnxt_pci_id_map,
6525 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6526 			RTE_PCI_DRV_INTR_RMV |
6527 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6528 						  * and OVS-DPDK
6529 						  */
6530 	.probe = bnxt_pci_probe,
6531 	.remove = bnxt_pci_remove,
6532 };
6533 
6534 static bool
6535 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6536 {
6537 	if (strcmp(dev->device->driver->name, drv->driver.name))
6538 		return false;
6539 
6540 	return true;
6541 }
6542 
6543 bool is_bnxt_supported(struct rte_eth_dev *dev)
6544 {
6545 	return is_device_supported(dev, &bnxt_rte_pmd);
6546 }
6547 
6548 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6549 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6550 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6551 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6552