xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision b6f30094e511c22f038907a507d16f7cc022c682)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2014-2023 Broadcom
3  * All rights reserved.
4  */
5 
6 #include <inttypes.h>
7 #include <stdalign.h>
8 #include <stdbool.h>
9 
10 #include <dev_driver.h>
11 #include <ethdev_driver.h>
12 #include <ethdev_pci.h>
13 #include <rte_malloc.h>
14 #include <rte_cycles.h>
15 #include <rte_alarm.h>
16 #include <rte_kvargs.h>
17 #include <rte_vect.h>
18 
19 #include "bnxt.h"
20 #include "bnxt_filter.h"
21 #include "bnxt_hwrm.h"
22 #include "bnxt_irq.h"
23 #include "bnxt_reps.h"
24 #include "bnxt_ring.h"
25 #include "bnxt_rxq.h"
26 #include "bnxt_rxr.h"
27 #include "bnxt_stats.h"
28 #include "bnxt_txq.h"
29 #include "bnxt_txr.h"
30 #include "bnxt_vnic.h"
31 #include "hsi_struct_def_dpdk.h"
32 #include "bnxt_nvm_defs.h"
33 #include "bnxt_tf_common.h"
34 #include "ulp_flow_db.h"
35 #include "rte_pmd_bnxt.h"
36 
37 #define DRV_MODULE_NAME		"bnxt"
38 static const char bnxt_version[] =
39 	"Broadcom NetXtreme driver " DRV_MODULE_NAME;
40 
41 /*
42  * The set of PCI devices this driver supports
43  */
44 static const struct rte_pci_id bnxt_pci_id_map[] = {
45 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
46 			 BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
47 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
48 			 BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
49 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
50 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
51 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
52 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
53 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
54 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
55 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
56 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
57 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
58 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
59 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
60 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
61 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
62 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
63 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
64 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
65 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
66 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
67 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
68 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
69 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
70 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
71 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
72 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
73 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
74 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
75 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
76 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
77 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
78 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
79 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
80 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
81 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
82 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
83 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
84 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58812) },
85 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58814) },
86 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818) },
87 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818_VF) },
88 	{ .vendor_id = 0, /* sentinel */ },
89 };
90 
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 #define BNXT_DEVARG_IEEE_1588	"ieee-1588"
102 
103 static const char *const bnxt_dev_args[] = {
104 	BNXT_DEVARG_REPRESENTOR,
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 	BNXT_DEVARG_IEEE_1588,
115 	NULL
116 };
117 
118 /*
119  * app-id = an non-negative 8-bit number
120  */
121 #define BNXT_DEVARG_APP_ID_INVALID(val)			((val) > 255)
122 
123 /*
124  * ieee-1588 = an non-negative 8-bit number
125  */
126 #define BNXT_DEVARG_IEEE_1588_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 static int bnxt_check_fw_ready(struct bnxt *bp);
189 
190 int is_bnxt_in_error(struct bnxt *bp)
191 {
192 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
193 		return -EIO;
194 	if (bp->flags & BNXT_FLAG_FW_RESET)
195 		return -EBUSY;
196 
197 	return 0;
198 }
199 
200 /***********************/
201 
202 /*
203  * High level utility functions
204  */
205 
206 uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
207 {
208 	unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
209 					     BNXT_RSS_TBL_SIZE_P5);
210 
211 	if (!BNXT_CHIP_P5(bp))
212 		return 1;
213 
214 	return RTE_ALIGN_MUL_CEIL(num_rss_rings,
215 				  BNXT_RSS_ENTRIES_PER_CTX_P5) /
216 				  BNXT_RSS_ENTRIES_PER_CTX_P5;
217 }
218 
219 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
220 {
221 	if (!BNXT_CHIP_P5(bp))
222 		return HW_HASH_INDEX_SIZE;
223 
224 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
225 }
226 
227 static void bnxt_free_parent_info(struct bnxt *bp)
228 {
229 	rte_free(bp->parent);
230 	bp->parent = NULL;
231 }
232 
233 static void bnxt_free_pf_info(struct bnxt *bp)
234 {
235 	rte_free(bp->pf);
236 	bp->pf = NULL;
237 }
238 
239 static void bnxt_free_link_info(struct bnxt *bp)
240 {
241 	rte_free(bp->link_info);
242 	bp->link_info = NULL;
243 }
244 
245 static void bnxt_free_leds_info(struct bnxt *bp)
246 {
247 	if (BNXT_VF(bp))
248 		return;
249 
250 	rte_free(bp->leds);
251 	bp->leds = NULL;
252 }
253 
254 static void bnxt_free_flow_stats_info(struct bnxt *bp)
255 {
256 	rte_free(bp->flow_stat);
257 	bp->flow_stat = NULL;
258 }
259 
260 static void bnxt_free_cos_queues(struct bnxt *bp)
261 {
262 	rte_free(bp->rx_cos_queue);
263 	bp->rx_cos_queue = NULL;
264 	rte_free(bp->tx_cos_queue);
265 	bp->tx_cos_queue = NULL;
266 }
267 
268 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
269 {
270 	bnxt_free_filter_mem(bp);
271 	bnxt_free_vnic_attributes(bp);
272 	bnxt_free_vnic_mem(bp);
273 
274 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
275 	 * If the number of rings change across fw update,
276 	 * we don't have much choice except to warn the user.
277 	 */
278 	if (!reconfig) {
279 		bnxt_free_stats(bp);
280 		bnxt_free_tx_rings(bp);
281 		bnxt_free_rx_rings(bp);
282 	}
283 	bnxt_free_async_cp_ring(bp);
284 	bnxt_free_rxtx_nq_ring(bp);
285 
286 	rte_free(bp->grp_info);
287 	bp->grp_info = NULL;
288 }
289 
290 static int bnxt_alloc_parent_info(struct bnxt *bp)
291 {
292 	bp->parent = rte_zmalloc("bnxt_parent_info",
293 				 sizeof(struct bnxt_parent_info), 0);
294 	if (bp->parent == NULL)
295 		return -ENOMEM;
296 
297 	return 0;
298 }
299 
300 static int bnxt_alloc_pf_info(struct bnxt *bp)
301 {
302 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
303 	if (bp->pf == NULL)
304 		return -ENOMEM;
305 
306 	return 0;
307 }
308 
309 static int bnxt_alloc_link_info(struct bnxt *bp)
310 {
311 	bp->link_info =
312 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
313 	if (bp->link_info == NULL)
314 		return -ENOMEM;
315 
316 	return 0;
317 }
318 
319 static int bnxt_alloc_leds_info(struct bnxt *bp)
320 {
321 	if (BNXT_VF(bp))
322 		return 0;
323 
324 	bp->leds = rte_zmalloc("bnxt_leds",
325 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
326 			       0);
327 	if (bp->leds == NULL)
328 		return -ENOMEM;
329 
330 	return 0;
331 }
332 
333 static int bnxt_alloc_cos_queues(struct bnxt *bp)
334 {
335 	bp->rx_cos_queue =
336 		rte_zmalloc("bnxt_rx_cosq",
337 			    BNXT_COS_QUEUE_COUNT *
338 			    sizeof(struct bnxt_cos_queue_info),
339 			    0);
340 	if (bp->rx_cos_queue == NULL)
341 		return -ENOMEM;
342 
343 	bp->tx_cos_queue =
344 		rte_zmalloc("bnxt_tx_cosq",
345 			    BNXT_COS_QUEUE_COUNT *
346 			    sizeof(struct bnxt_cos_queue_info),
347 			    0);
348 	if (bp->tx_cos_queue == NULL)
349 		return -ENOMEM;
350 
351 	return 0;
352 }
353 
354 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
355 {
356 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
357 				    sizeof(struct bnxt_flow_stat_info), 0);
358 	if (bp->flow_stat == NULL)
359 		return -ENOMEM;
360 
361 	return 0;
362 }
363 
364 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
365 {
366 	int rc;
367 
368 	rc = bnxt_alloc_ring_grps(bp);
369 	if (rc)
370 		goto alloc_mem_err;
371 
372 	rc = bnxt_alloc_async_ring_struct(bp);
373 	if (rc)
374 		goto alloc_mem_err;
375 
376 	rc = bnxt_alloc_vnic_mem(bp);
377 	if (rc)
378 		goto alloc_mem_err;
379 
380 	rc = bnxt_alloc_vnic_attributes(bp, reconfig);
381 	if (rc)
382 		goto alloc_mem_err;
383 
384 	rc = bnxt_alloc_filter_mem(bp);
385 	if (rc)
386 		goto alloc_mem_err;
387 
388 	rc = bnxt_alloc_async_cp_ring(bp);
389 	if (rc)
390 		goto alloc_mem_err;
391 
392 	rc = bnxt_alloc_rxtx_nq_ring(bp);
393 	if (rc)
394 		goto alloc_mem_err;
395 
396 	if (BNXT_FLOW_XSTATS_EN(bp)) {
397 		rc = bnxt_alloc_flow_stats_info(bp);
398 		if (rc)
399 			goto alloc_mem_err;
400 	}
401 
402 	return 0;
403 
404 alloc_mem_err:
405 	bnxt_free_mem(bp, reconfig);
406 	return rc;
407 }
408 
409 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
410 {
411 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
412 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
413 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
414 	struct bnxt_rx_queue *rxq;
415 	unsigned int j;
416 	int rc;
417 
418 	rc = bnxt_vnic_grp_alloc(bp, vnic);
419 	if (rc)
420 		goto err_out;
421 
422 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
423 		    vnic_id, vnic, vnic->fw_grp_ids);
424 
425 	/* populate the fw group table */
426 	bnxt_vnic_ring_grp_populate(bp, vnic);
427 	bnxt_vnic_rules_init(vnic);
428 
429 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
430 	if (rc)
431 		goto err_out;
432 
433 	/* Alloc RSS context only if RSS mode is enabled */
434 	if (dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS) {
435 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
436 
437 		/* RSS table size in P5 is 512.
438 		 * Cap max Rx rings to same value
439 		 */
440 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
441 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
442 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
443 			goto err_out;
444 		}
445 
446 		rc = 0;
447 		for (j = 0; j < nr_ctxs; j++) {
448 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
449 			if (rc)
450 				break;
451 		}
452 		if (rc) {
453 			PMD_DRV_LOG(ERR,
454 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
455 				    vnic_id, j, rc);
456 			goto err_out;
457 		}
458 		vnic->num_lb_ctxts = nr_ctxs;
459 	}
460 
461 	/*
462 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
463 	 * setting is not available at this time, it will not be
464 	 * configured correctly in the CFA.
465 	 */
466 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
467 		vnic->vlan_strip = true;
468 	else
469 		vnic->vlan_strip = false;
470 
471 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
472 	if (rc)
473 		goto err_out;
474 
475 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
476 	if (rc)
477 		goto err_out;
478 
479 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
480 		rxq = bp->eth_dev->data->rx_queues[j];
481 
482 		PMD_DRV_LOG(DEBUG,
483 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
484 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
485 
486 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
487 			vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
488 	}
489 
490 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
491 
492 	rc = bnxt_vnic_rss_configure(bp, vnic);
493 	if (rc)
494 		goto err_out;
495 
496 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
497 
498 	rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
499 				    (rx_offloads & RTE_ETH_RX_OFFLOAD_TCP_LRO) ?
500 				    true : false);
501 	if (rc)
502 		goto err_out;
503 
504 	return 0;
505 err_out:
506 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
507 		    vnic_id, rc);
508 	return rc;
509 }
510 
511 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
512 {
513 	int rc = 0;
514 
515 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
516 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
517 	if (rc)
518 		return rc;
519 
520 	PMD_DRV_LOG(DEBUG,
521 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
522 		    " rx_fc_in_tbl.ctx_id = %d\n",
523 		    bp->flow_stat->rx_fc_in_tbl.va,
524 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
525 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
526 
527 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
528 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
529 	if (rc)
530 		return rc;
531 
532 	PMD_DRV_LOG(DEBUG,
533 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
534 		    " rx_fc_out_tbl.ctx_id = %d\n",
535 		    bp->flow_stat->rx_fc_out_tbl.va,
536 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
537 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
538 
539 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
540 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
541 	if (rc)
542 		return rc;
543 
544 	PMD_DRV_LOG(DEBUG,
545 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
546 		    " tx_fc_in_tbl.ctx_id = %d\n",
547 		    bp->flow_stat->tx_fc_in_tbl.va,
548 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
549 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
550 
551 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
552 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
553 	if (rc)
554 		return rc;
555 
556 	PMD_DRV_LOG(DEBUG,
557 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
558 		    " tx_fc_out_tbl.ctx_id = %d\n",
559 		    bp->flow_stat->tx_fc_out_tbl.va,
560 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
561 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
562 
563 	memset(bp->flow_stat->rx_fc_out_tbl.va,
564 	       0,
565 	       bp->flow_stat->rx_fc_out_tbl.size);
566 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
567 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
568 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
569 				       bp->flow_stat->max_fc,
570 				       true);
571 	if (rc)
572 		return rc;
573 
574 	memset(bp->flow_stat->tx_fc_out_tbl.va,
575 	       0,
576 	       bp->flow_stat->tx_fc_out_tbl.size);
577 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
578 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
579 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
580 				       bp->flow_stat->max_fc,
581 				       true);
582 
583 	return rc;
584 }
585 
586 static int bnxt_alloc_ctx_mem_buf(struct bnxt *bp, char *type, size_t size,
587 				  struct bnxt_ctx_mem_buf_info *ctx)
588 {
589 	if (!ctx)
590 		return -EINVAL;
591 
592 	ctx->va = rte_zmalloc_socket(type, size, 0,
593 				     bp->eth_dev->device->numa_node);
594 	if (ctx->va == NULL)
595 		return -ENOMEM;
596 	rte_mem_lock_page(ctx->va);
597 	ctx->size = size;
598 	ctx->dma = rte_mem_virt2iova(ctx->va);
599 	if (ctx->dma == RTE_BAD_IOVA)
600 		return -ENOMEM;
601 
602 	return 0;
603 }
604 
605 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
606 {
607 	struct rte_pci_device *pdev = bp->pdev;
608 	char type[RTE_MEMZONE_NAMESIZE];
609 	uint16_t max_fc;
610 	int rc = 0;
611 
612 	max_fc = bp->flow_stat->max_fc;
613 
614 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
615 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
616 	/* 4 bytes for each counter-id */
617 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
618 				    max_fc * 4,
619 				    &bp->flow_stat->rx_fc_in_tbl);
620 	if (rc)
621 		return rc;
622 
623 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
624 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
625 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
626 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
627 				    max_fc * 16,
628 				    &bp->flow_stat->rx_fc_out_tbl);
629 	if (rc)
630 		return rc;
631 
632 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
633 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
634 	/* 4 bytes for each counter-id */
635 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
636 				    max_fc * 4,
637 				    &bp->flow_stat->tx_fc_in_tbl);
638 	if (rc)
639 		return rc;
640 
641 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
642 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
643 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
644 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
645 				    max_fc * 16,
646 				    &bp->flow_stat->tx_fc_out_tbl);
647 	if (rc)
648 		return rc;
649 
650 	rc = bnxt_register_fc_ctx_mem(bp);
651 
652 	return rc;
653 }
654 
655 static int bnxt_init_ctx_mem(struct bnxt *bp)
656 {
657 	int rc = 0;
658 
659 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
660 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
661 	    !BNXT_FLOW_XSTATS_EN(bp))
662 		return 0;
663 
664 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
665 	if (rc)
666 		return rc;
667 
668 	rc = bnxt_init_fc_ctx_mem(bp);
669 
670 	return rc;
671 }
672 
673 static inline bool bnxt_force_link_config(struct bnxt *bp)
674 {
675 	uint16_t subsystem_device_id = bp->pdev->id.subsystem_device_id;
676 
677 	switch (subsystem_device_id) {
678 	case BROADCOM_DEV_957508_N2100:
679 	case BROADCOM_DEV_957414_N225:
680 		return true;
681 	default:
682 		return false;
683 	}
684 }
685 
686 static int bnxt_update_phy_setting(struct bnxt *bp)
687 {
688 	struct rte_eth_link new;
689 	int rc;
690 
691 	rc = bnxt_get_hwrm_link_config(bp, &new);
692 	if (rc) {
693 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
694 		return rc;
695 	}
696 
697 	/*
698 	 * Device is not obliged link down in certain scenarios, even
699 	 * when forced. When FW does not allow any user other than BMC
700 	 * to shutdown the port, bnxt_get_hwrm_link_config() call always
701 	 * returns link up. Force phy update always in that case.
702 	 */
703 	if (!new.link_status || bnxt_force_link_config(bp)) {
704 		rc = bnxt_set_hwrm_link_config(bp, true);
705 		if (rc) {
706 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
707 			return rc;
708 		}
709 	}
710 
711 	return rc;
712 }
713 
714 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
715 {
716 	rte_free(bp->prev_rx_ring_stats);
717 	rte_free(bp->prev_tx_ring_stats);
718 
719 	bp->prev_rx_ring_stats = NULL;
720 	bp->prev_tx_ring_stats = NULL;
721 }
722 
723 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
724 {
725 	bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
726 					      sizeof(struct bnxt_ring_stats) *
727 					      bp->rx_cp_nr_rings,
728 					      0);
729 	if (bp->prev_rx_ring_stats == NULL)
730 		return -ENOMEM;
731 
732 	bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
733 					     sizeof(struct bnxt_ring_stats) *
734 					     bp->tx_cp_nr_rings,
735 					     0);
736 	if (bp->tx_cp_nr_rings > 0 && bp->prev_tx_ring_stats == NULL)
737 		goto error;
738 
739 	return 0;
740 
741 error:
742 	bnxt_free_prev_ring_stats(bp);
743 	return -ENOMEM;
744 }
745 
746 static int bnxt_start_nic(struct bnxt *bp)
747 {
748 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
749 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
750 	uint32_t intr_vector = 0;
751 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
752 	uint32_t vec = BNXT_MISC_VEC_ID;
753 	unsigned int i, j;
754 	int rc;
755 
756 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU)
757 		bp->flags |= BNXT_FLAG_JUMBO;
758 	else
759 		bp->flags &= ~BNXT_FLAG_JUMBO;
760 
761 	/* P5 does not support ring groups.
762 	 * But we will use the array to save RSS context IDs.
763 	 */
764 	if (BNXT_CHIP_P5(bp))
765 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
766 
767 	rc = bnxt_vnic_queue_db_init(bp);
768 	if (rc) {
769 		PMD_DRV_LOG(ERR, "could not allocate vnic db\n");
770 		goto err_out;
771 	}
772 
773 	rc = bnxt_alloc_hwrm_rings(bp);
774 	if (rc) {
775 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
776 		goto err_out;
777 	}
778 
779 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
780 	if (rc) {
781 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
782 		goto err_out;
783 	}
784 
785 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
786 		goto skip_cosq_cfg;
787 
788 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
789 		if (bp->rx_cos_queue[i].id != 0xff) {
790 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
791 
792 			if (!vnic) {
793 				PMD_DRV_LOG(ERR,
794 					    "Num pools more than FW profile\n");
795 				rc = -EINVAL;
796 				goto err_out;
797 			}
798 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
799 			bp->rx_cosq_cnt++;
800 		}
801 	}
802 
803 skip_cosq_cfg:
804 	rc = bnxt_mq_rx_configure(bp);
805 	if (rc) {
806 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
807 		goto err_out;
808 	}
809 
810 	for (j = 0; j < bp->rx_nr_rings; j++) {
811 		struct bnxt_rx_queue *rxq = bp->rx_queues[j];
812 
813 		if (!rxq->rx_deferred_start) {
814 			bp->eth_dev->data->rx_queue_state[j] =
815 				RTE_ETH_QUEUE_STATE_STARTED;
816 			rxq->rx_started = true;
817 		}
818 	}
819 
820 	/* setup the default vnic details*/
821 	bnxt_vnic_queue_db_update_dlft_vnic(bp);
822 
823 	/* VNIC configuration */
824 	for (i = 0; i < bp->nr_vnics; i++) {
825 		rc = bnxt_setup_one_vnic(bp, i);
826 		if (rc)
827 			goto err_out;
828 	}
829 
830 	for (j = 0; j < bp->tx_nr_rings; j++) {
831 		struct bnxt_tx_queue *txq = bp->tx_queues[j];
832 
833 		if (!txq->tx_deferred_start) {
834 			bp->eth_dev->data->tx_queue_state[j] =
835 				RTE_ETH_QUEUE_STATE_STARTED;
836 			txq->tx_started = true;
837 		}
838 	}
839 
840 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
841 	if (rc) {
842 		PMD_DRV_LOG(ERR,
843 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
844 		goto err_out;
845 	}
846 
847 	/* check and configure queue intr-vector mapping */
848 	if ((rte_intr_cap_multiple(intr_handle) ||
849 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
850 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
851 		intr_vector = bp->eth_dev->data->nb_rx_queues;
852 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
853 		if (intr_vector > bp->rx_cp_nr_rings) {
854 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
855 					bp->rx_cp_nr_rings);
856 			return -ENOTSUP;
857 		}
858 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
859 		if (rc)
860 			return rc;
861 	}
862 
863 	if (rte_intr_dp_is_en(intr_handle)) {
864 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
865 					bp->eth_dev->data->nb_rx_queues)) {
866 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
867 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
868 			rc = -ENOMEM;
869 			goto err_out;
870 		}
871 		PMD_DRV_LOG(DEBUG, "intr_handle->nb_efd = %d "
872 			    "intr_handle->max_intr = %d\n",
873 			    rte_intr_nb_efd_get(intr_handle),
874 			    rte_intr_max_intr_get(intr_handle));
875 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
876 		     queue_id++) {
877 			rte_intr_vec_list_index_set(intr_handle,
878 					queue_id, vec + BNXT_RX_VEC_START);
879 			if (vec < base + rte_intr_nb_efd_get(intr_handle)
880 			    - 1)
881 				vec++;
882 		}
883 	}
884 
885 	/* enable uio/vfio intr/eventfd mapping */
886 	rc = rte_intr_enable(intr_handle);
887 #ifndef RTE_EXEC_ENV_FREEBSD
888 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
889 	if (rc)
890 		goto err_out;
891 #endif
892 
893 	rc = bnxt_update_phy_setting(bp);
894 	if (rc)
895 		goto err_out;
896 
897 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
898 	if (!bp->mark_table)
899 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
900 
901 	return 0;
902 
903 err_out:
904 	/* Some of the error status returned by FW may not be from errno.h */
905 	if (rc > 0)
906 		rc = -EIO;
907 
908 	return rc;
909 }
910 
911 static int bnxt_shutdown_nic(struct bnxt *bp)
912 {
913 	bnxt_free_all_hwrm_resources(bp);
914 	bnxt_free_all_filters(bp);
915 	bnxt_free_all_vnics(bp);
916 	bnxt_vnic_queue_db_deinit(bp);
917 	return 0;
918 }
919 
920 /*
921  * Device configuration and status function
922  */
923 
924 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
925 {
926 	uint32_t pam4_link_speed = 0;
927 	uint32_t link_speed = 0;
928 	uint32_t speed_capa = 0;
929 
930 	if (bp->link_info == NULL)
931 		return 0;
932 
933 	link_speed = bp->link_info->support_speeds;
934 
935 	/* If PAM4 is configured, use PAM4 supported speed */
936 	if (bp->link_info->support_pam4_speeds > 0)
937 		pam4_link_speed = bp->link_info->support_pam4_speeds;
938 
939 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
940 		speed_capa |= RTE_ETH_LINK_SPEED_100M;
941 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
942 		speed_capa |= RTE_ETH_LINK_SPEED_100M_HD;
943 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
944 		speed_capa |= RTE_ETH_LINK_SPEED_1G;
945 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
946 		speed_capa |= RTE_ETH_LINK_SPEED_2_5G;
947 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
948 		speed_capa |= RTE_ETH_LINK_SPEED_10G;
949 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
950 		speed_capa |= RTE_ETH_LINK_SPEED_20G;
951 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
952 		speed_capa |= RTE_ETH_LINK_SPEED_25G;
953 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
954 		speed_capa |= RTE_ETH_LINK_SPEED_40G;
955 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
956 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
957 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
958 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
959 	if (pam4_link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
960 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
961 	if (pam4_link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
962 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
963 	if (pam4_link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
964 		speed_capa |= RTE_ETH_LINK_SPEED_200G;
965 
966 	if (bp->link_info->auto_mode ==
967 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
968 		speed_capa |= RTE_ETH_LINK_SPEED_FIXED;
969 
970 	return speed_capa;
971 }
972 
973 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
974 				struct rte_eth_dev_info *dev_info)
975 {
976 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
977 	struct bnxt *bp = eth_dev->data->dev_private;
978 	uint16_t max_vnics, i, j, vpool, vrxq;
979 	unsigned int max_rx_rings;
980 	int rc;
981 
982 	rc = is_bnxt_in_error(bp);
983 	if (rc)
984 		return rc;
985 
986 	/* MAC Specifics */
987 	dev_info->max_mac_addrs = RTE_MIN(bp->max_l2_ctx, RTE_ETH_NUM_RECEIVE_MAC_ADDR);
988 	dev_info->max_hash_mac_addrs = 0;
989 
990 	/* PF/VF specifics */
991 	if (BNXT_PF(bp))
992 		dev_info->max_vfs = pdev->max_vfs;
993 
994 	max_rx_rings = bnxt_max_rings(bp);
995 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
996 	dev_info->max_rx_queues = max_rx_rings;
997 	dev_info->max_tx_queues = max_rx_rings;
998 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
999 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
1000 	max_vnics = bp->max_vnics;
1001 
1002 	/* MTU specifics */
1003 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
1004 	dev_info->max_mtu = BNXT_MAX_MTU;
1005 
1006 	/* Fast path specifics */
1007 	dev_info->min_rx_bufsize = 1;
1008 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
1009 
1010 	dev_info->rx_offload_capa = bnxt_get_rx_port_offloads(bp);
1011 	dev_info->tx_queue_offload_capa = RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
1012 	dev_info->tx_offload_capa = bnxt_get_tx_port_offloads(bp) |
1013 				    dev_info->tx_queue_offload_capa;
1014 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
1015 
1016 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
1017 	dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
1018 			     RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
1019 	dev_info->dev_capa &= ~RTE_ETH_DEV_CAPA_FLOW_RULE_KEEP;
1020 
1021 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
1022 		.rx_thresh = {
1023 			.pthresh = 8,
1024 			.hthresh = 8,
1025 			.wthresh = 0,
1026 		},
1027 		.rx_free_thresh = 32,
1028 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
1029 	};
1030 
1031 	dev_info->default_txconf = (struct rte_eth_txconf) {
1032 		.tx_thresh = {
1033 			.pthresh = 32,
1034 			.hthresh = 0,
1035 			.wthresh = 0,
1036 		},
1037 		.tx_free_thresh = 32,
1038 		.tx_rs_thresh = 32,
1039 	};
1040 
1041 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1042 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1043 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1044 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1045 
1046 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1047 		dev_info->switch_info.name = eth_dev->device->name;
1048 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1049 		dev_info->switch_info.port_id =
1050 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1051 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1052 	}
1053 
1054 	/*
1055 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1056 	 *       need further investigation.
1057 	 */
1058 
1059 	/* VMDq resources */
1060 	vpool = 64; /* RTE_ETH_64_POOLS */
1061 	vrxq = 128; /* RTE_ETH_VMDQ_DCB_NUM_QUEUES */
1062 	for (i = 0; i < 4; vpool >>= 1, i++) {
1063 		if (max_vnics > vpool) {
1064 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1065 				if (dev_info->max_rx_queues > vrxq) {
1066 					if (vpool > vrxq)
1067 						vpool = vrxq;
1068 					goto found;
1069 				}
1070 			}
1071 			/* Not enough resources to support VMDq */
1072 			break;
1073 		}
1074 	}
1075 	/* Not enough resources to support VMDq */
1076 	vpool = 0;
1077 	vrxq = 0;
1078 found:
1079 	dev_info->max_vmdq_pools = vpool;
1080 	dev_info->vmdq_queue_num = vrxq;
1081 
1082 	dev_info->vmdq_pool_base = 0;
1083 	dev_info->vmdq_queue_base = 0;
1084 
1085 	dev_info->err_handle_mode = RTE_ETH_ERROR_HANDLE_MODE_PROACTIVE;
1086 
1087 	return 0;
1088 }
1089 
1090 /* Configure the device based on the configuration provided */
1091 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1092 {
1093 	struct bnxt *bp = eth_dev->data->dev_private;
1094 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1095 	struct rte_eth_rss_conf *rss_conf = &eth_dev->data->dev_conf.rx_adv_conf.rss_conf;
1096 	int rc;
1097 
1098 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1099 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1100 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1101 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1102 
1103 	rc = is_bnxt_in_error(bp);
1104 	if (rc)
1105 		return rc;
1106 
1107 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1108 		rc = bnxt_hwrm_check_vf_rings(bp);
1109 		if (rc) {
1110 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1111 			return -ENOSPC;
1112 		}
1113 
1114 		/* If a resource has already been allocated - in this case
1115 		 * it is the async completion ring, free it. Reallocate it after
1116 		 * resource reservation. This will ensure the resource counts
1117 		 * are calculated correctly.
1118 		 */
1119 
1120 		pthread_mutex_lock(&bp->def_cp_lock);
1121 
1122 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1123 			bnxt_disable_int(bp);
1124 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1125 		}
1126 
1127 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1128 		if (rc) {
1129 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1130 			pthread_mutex_unlock(&bp->def_cp_lock);
1131 			return -ENOSPC;
1132 		}
1133 
1134 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1135 			rc = bnxt_alloc_async_cp_ring(bp);
1136 			if (rc) {
1137 				pthread_mutex_unlock(&bp->def_cp_lock);
1138 				return rc;
1139 			}
1140 			bnxt_enable_int(bp);
1141 		}
1142 
1143 		pthread_mutex_unlock(&bp->def_cp_lock);
1144 	}
1145 
1146 	/* Inherit new configurations */
1147 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1148 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1149 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1150 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1151 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1152 	    bp->max_stat_ctx)
1153 		goto resource_error;
1154 
1155 	if (BNXT_HAS_RING_GRPS(bp) &&
1156 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1157 		goto resource_error;
1158 
1159 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS) &&
1160 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1161 		goto resource_error;
1162 
1163 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1164 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1165 
1166 	if (eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
1167 		rx_offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
1168 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1169 
1170 	/* application provides the hash key to program */
1171 	if (rss_conf->rss_key != NULL) {
1172 		if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE)
1173 			PMD_DRV_LOG(WARNING, "port %u RSS key len must be %d bytes long",
1174 				    eth_dev->data->port_id, HW_HASH_KEY_SIZE);
1175 		else
1176 			memcpy(bp->rss_conf.rss_key, rss_conf->rss_key, HW_HASH_KEY_SIZE);
1177 	}
1178 	bp->rss_conf.rss_key_len = HW_HASH_KEY_SIZE;
1179 	bp->rss_conf.rss_hf = rss_conf->rss_hf;
1180 
1181 	bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1182 
1183 	return 0;
1184 
1185 resource_error:
1186 	PMD_DRV_LOG(ERR,
1187 		    "Insufficient resources to support requested config\n");
1188 	PMD_DRV_LOG(ERR,
1189 		    "Num Queues Requested: Tx %d, Rx %d\n",
1190 		    eth_dev->data->nb_tx_queues,
1191 		    eth_dev->data->nb_rx_queues);
1192 	PMD_DRV_LOG(ERR,
1193 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1194 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1195 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1196 	return -ENOSPC;
1197 }
1198 
1199 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1200 {
1201 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1202 
1203 	if (link->link_status)
1204 		PMD_DRV_LOG(DEBUG, "Port %d Link Up - speed %u Mbps - %s\n",
1205 			eth_dev->data->port_id,
1206 			(uint32_t)link->link_speed,
1207 			(link->link_duplex == RTE_ETH_LINK_FULL_DUPLEX) ?
1208 			("full-duplex") : ("half-duplex\n"));
1209 	else
1210 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1211 			eth_dev->data->port_id);
1212 }
1213 
1214 /*
1215  * Determine whether the current configuration requires support for scattered
1216  * receive; return 1 if scattered receive is required and 0 if not.
1217  */
1218 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1219 {
1220 	uint32_t overhead = BNXT_MAX_PKT_LEN - BNXT_MAX_MTU;
1221 	uint16_t buf_size;
1222 	int i;
1223 
1224 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_SCATTER)
1225 		return 1;
1226 
1227 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_TCP_LRO)
1228 		return 1;
1229 
1230 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1231 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1232 
1233 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1234 				      RTE_PKTMBUF_HEADROOM);
1235 		if (eth_dev->data->mtu + overhead > buf_size)
1236 			return 1;
1237 	}
1238 	return 0;
1239 }
1240 
1241 static eth_rx_burst_t
1242 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1243 {
1244 	struct bnxt *bp = eth_dev->data->dev_private;
1245 
1246 	/* Disable vector mode RX for Stingray2 for now */
1247 	if (BNXT_CHIP_SR2(bp)) {
1248 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1249 		return bnxt_recv_pkts;
1250 	}
1251 
1252 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1253 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1254 	if (eth_dev->data->scattered_rx)
1255 		goto use_scalar_rx;
1256 
1257 	/*
1258 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1259 	 * asynchronous completions and receive completions can be placed in
1260 	 * the same completion ring.
1261 	 */
1262 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1263 		goto use_scalar_rx;
1264 
1265 	/*
1266 	 * Vector mode receive cannot be enabled if any receive offloads outside
1267 	 * a limited subset have been enabled.
1268 	 */
1269 	if (eth_dev->data->dev_conf.rxmode.offloads &
1270 		~(RTE_ETH_RX_OFFLOAD_VLAN_STRIP |
1271 		  RTE_ETH_RX_OFFLOAD_KEEP_CRC |
1272 		  RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
1273 		  RTE_ETH_RX_OFFLOAD_UDP_CKSUM |
1274 		  RTE_ETH_RX_OFFLOAD_TCP_CKSUM |
1275 		  RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1276 		  RTE_ETH_RX_OFFLOAD_OUTER_UDP_CKSUM |
1277 		  RTE_ETH_RX_OFFLOAD_RSS_HASH |
1278 		  RTE_ETH_RX_OFFLOAD_VLAN_FILTER))
1279 		goto use_scalar_rx;
1280 
1281 	if (bp->ieee_1588)
1282 		goto use_scalar_rx;
1283 
1284 #if defined(RTE_ARCH_X86)
1285 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1286 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1287 		PMD_DRV_LOG(INFO,
1288 			    "Using AVX2 vector mode receive for port %d\n",
1289 			    eth_dev->data->port_id);
1290 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1291 		return bnxt_recv_pkts_vec_avx2;
1292 	}
1293  #endif
1294 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1295 		PMD_DRV_LOG(INFO,
1296 			    "Using SSE vector mode receive for port %d\n",
1297 			    eth_dev->data->port_id);
1298 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1299 		return bnxt_recv_pkts_vec;
1300 	}
1301 
1302 use_scalar_rx:
1303 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1304 		    eth_dev->data->port_id);
1305 	PMD_DRV_LOG(INFO,
1306 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1307 		    eth_dev->data->port_id,
1308 		    eth_dev->data->scattered_rx,
1309 		    eth_dev->data->dev_conf.rxmode.offloads);
1310 #endif
1311 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1312 	return bnxt_recv_pkts;
1313 }
1314 
1315 static eth_tx_burst_t
1316 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1317 {
1318 	struct bnxt *bp = eth_dev->data->dev_private;
1319 
1320 	/* Disable vector mode TX for Stingray2 for now */
1321 	if (BNXT_CHIP_SR2(bp))
1322 		return bnxt_xmit_pkts;
1323 
1324 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1325 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1326 
1327 	/*
1328 	 * Vector mode transmit can be enabled only if not using scatter rx
1329 	 * or tx offloads.
1330 	 */
1331 	if (eth_dev->data->scattered_rx ||
1332 	    (offloads & ~RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE) ||
1333 	    BNXT_TRUFLOW_EN(bp) || bp->ieee_1588)
1334 		goto use_scalar_tx;
1335 
1336 #if defined(RTE_ARCH_X86)
1337 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1338 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1339 		PMD_DRV_LOG(INFO,
1340 			    "Using AVX2 vector mode transmit for port %d\n",
1341 			    eth_dev->data->port_id);
1342 		return bnxt_xmit_pkts_vec_avx2;
1343 	}
1344 #endif
1345 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1346 		PMD_DRV_LOG(INFO,
1347 			    "Using SSE vector mode transmit for port %d\n",
1348 			    eth_dev->data->port_id);
1349 		return bnxt_xmit_pkts_vec;
1350 	}
1351 
1352 use_scalar_tx:
1353 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1354 		    eth_dev->data->port_id);
1355 	PMD_DRV_LOG(INFO,
1356 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1357 		    eth_dev->data->port_id,
1358 		    eth_dev->data->scattered_rx,
1359 		    offloads);
1360 #endif
1361 	return bnxt_xmit_pkts;
1362 }
1363 
1364 static int bnxt_handle_if_change_status(struct bnxt *bp)
1365 {
1366 	int rc;
1367 
1368 	/* Since fw has undergone a reset and lost all contexts,
1369 	 * set fatal flag to not issue hwrm during cleanup
1370 	 */
1371 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1372 	bnxt_uninit_resources(bp, true);
1373 
1374 	/* clear fatal flag so that re-init happens */
1375 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1376 
1377 	rc = bnxt_check_fw_ready(bp);
1378 	if (rc)
1379 		return rc;
1380 
1381 	rc = bnxt_init_resources(bp, true);
1382 
1383 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1384 
1385 	return rc;
1386 }
1387 
1388 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1389 {
1390 	struct bnxt *bp = eth_dev->data->dev_private;
1391 	int rc = 0;
1392 
1393 	if (!BNXT_SINGLE_PF(bp))
1394 		return -ENOTSUP;
1395 
1396 	if (!bp->link_info->link_up)
1397 		rc = bnxt_set_hwrm_link_config(bp, true);
1398 	if (!rc)
1399 		eth_dev->data->dev_link.link_status = 1;
1400 
1401 	bnxt_print_link_info(eth_dev);
1402 	return rc;
1403 }
1404 
1405 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1406 {
1407 	struct bnxt *bp = eth_dev->data->dev_private;
1408 
1409 	if (!BNXT_SINGLE_PF(bp))
1410 		return -ENOTSUP;
1411 
1412 	eth_dev->data->dev_link.link_status = 0;
1413 	bnxt_set_hwrm_link_config(bp, false);
1414 	bp->link_info->link_up = 0;
1415 
1416 	return 0;
1417 }
1418 
1419 static void bnxt_free_switch_domain(struct bnxt *bp)
1420 {
1421 	int rc = 0;
1422 
1423 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1424 		return;
1425 
1426 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1427 	if (rc)
1428 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1429 			    bp->switch_domain_id, rc);
1430 }
1431 
1432 static void bnxt_ptp_get_current_time(void *arg)
1433 {
1434 	struct bnxt *bp = arg;
1435 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1436 	int rc;
1437 
1438 	rc = is_bnxt_in_error(bp);
1439 	if (rc)
1440 		return;
1441 
1442 	if (!ptp)
1443 		return;
1444 
1445 	rte_spinlock_lock(&ptp->ptp_lock);
1446 	ptp->old_time = ptp->current_time;
1447 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1448 				&ptp->current_time);
1449 	rte_spinlock_unlock(&ptp->ptp_lock);
1450 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1451 	if (rc != 0) {
1452 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1453 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1454 	}
1455 }
1456 
1457 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1458 {
1459 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1460 	int rc;
1461 
1462 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1463 		return 0;
1464 
1465 	rte_spinlock_lock(&ptp->ptp_lock);
1466 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1467 				&ptp->current_time);
1468 	ptp->old_time = ptp->current_time;
1469 	rte_spinlock_unlock(&ptp->ptp_lock);
1470 
1471 
1472 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1473 	return rc;
1474 }
1475 
1476 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1477 {
1478 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1479 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1480 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1481 	}
1482 }
1483 
1484 static void bnxt_ptp_stop(struct bnxt *bp)
1485 {
1486 	bnxt_cancel_ptp_alarm(bp);
1487 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1488 }
1489 
1490 static int bnxt_ptp_start(struct bnxt *bp)
1491 {
1492 	int rc;
1493 
1494 	rc = bnxt_schedule_ptp_alarm(bp);
1495 	if (rc != 0) {
1496 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1497 	} else {
1498 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1499 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1500 	}
1501 
1502 	return rc;
1503 }
1504 
1505 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1506 {
1507 	struct bnxt *bp = eth_dev->data->dev_private;
1508 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1509 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1510 	struct rte_eth_link link;
1511 	uint16_t i;
1512 	int ret;
1513 
1514 	eth_dev->data->dev_started = 0;
1515 
1516 	/* Prevent crashes when queues are still in use */
1517 	bnxt_stop_rxtx(eth_dev);
1518 
1519 	bnxt_disable_int(bp);
1520 
1521 	/* disable uio/vfio intr/eventfd mapping */
1522 	rte_intr_disable(intr_handle);
1523 
1524 	/* Stop the child representors for this device */
1525 	ret = bnxt_rep_stop_all(bp);
1526 	if (ret != 0)
1527 		return ret;
1528 
1529 	/* delete the bnxt ULP port details */
1530 	bnxt_ulp_port_deinit(bp);
1531 
1532 	bnxt_cancel_fw_health_check(bp);
1533 
1534 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1535 		bnxt_cancel_ptp_alarm(bp);
1536 
1537 	/* Do not bring link down during reset recovery */
1538 	if (!is_bnxt_in_error(bp)) {
1539 		bnxt_dev_set_link_down_op(eth_dev);
1540 		/* Wait for link to be reset */
1541 		if (BNXT_SINGLE_PF(bp))
1542 			rte_delay_ms(500);
1543 		/* clear the recorded link status */
1544 		memset(&link, 0, sizeof(link));
1545 		rte_eth_linkstatus_set(eth_dev, &link);
1546 	}
1547 
1548 	/* Clean queue intr-vector mapping */
1549 	rte_intr_efd_disable(intr_handle);
1550 	rte_intr_vec_list_free(intr_handle);
1551 
1552 	bnxt_hwrm_port_clr_stats(bp);
1553 	bnxt_free_tx_mbufs(bp);
1554 	bnxt_free_rx_mbufs(bp);
1555 	/* Process any remaining notifications in default completion queue */
1556 	bnxt_int_handler(eth_dev);
1557 	bnxt_shutdown_nic(bp);
1558 	bnxt_hwrm_if_change(bp, false);
1559 
1560 	bnxt_free_prev_ring_stats(bp);
1561 	rte_free(bp->mark_table);
1562 	bp->mark_table = NULL;
1563 
1564 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1565 	bp->rx_cosq_cnt = 0;
1566 	/* All filters are deleted on a port stop. */
1567 	if (BNXT_FLOW_XSTATS_EN(bp))
1568 		bp->flow_stat->flow_count = 0;
1569 
1570 	eth_dev->data->scattered_rx = 0;
1571 
1572 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++)
1573 		eth_dev->data->rx_queue_state[i] = RTE_ETH_QUEUE_STATE_STOPPED;
1574 	for (i = 0; i < eth_dev->data->nb_tx_queues; i++)
1575 		eth_dev->data->tx_queue_state[i] = RTE_ETH_QUEUE_STATE_STOPPED;
1576 
1577 	return 0;
1578 }
1579 
1580 /* Unload the driver, release resources */
1581 int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1582 {
1583 	struct bnxt *bp = eth_dev->data->dev_private;
1584 
1585 	pthread_mutex_lock(&bp->err_recovery_lock);
1586 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1587 		PMD_DRV_LOG(ERR,
1588 			    "Adapter recovering from error..Please retry\n");
1589 		pthread_mutex_unlock(&bp->err_recovery_lock);
1590 		return -EAGAIN;
1591 	}
1592 	pthread_mutex_unlock(&bp->err_recovery_lock);
1593 
1594 	return bnxt_dev_stop(eth_dev);
1595 }
1596 
1597 int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1598 {
1599 	struct bnxt *bp = eth_dev->data->dev_private;
1600 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1601 	int vlan_mask = 0;
1602 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1603 
1604 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1605 		PMD_DRV_LOG(ERR,
1606 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1607 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1608 
1609 	do {
1610 		rc = bnxt_hwrm_if_change(bp, true);
1611 		if (rc == 0 || rc != -EAGAIN)
1612 			break;
1613 
1614 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1615 	} while (retry_cnt--);
1616 
1617 	if (rc)
1618 		return rc;
1619 
1620 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1621 		rc = bnxt_handle_if_change_status(bp);
1622 		if (rc)
1623 			return rc;
1624 	}
1625 
1626 	bnxt_enable_int(bp);
1627 
1628 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1629 
1630 	rc = bnxt_start_nic(bp);
1631 	if (rc)
1632 		goto error;
1633 
1634 	rc = bnxt_alloc_prev_ring_stats(bp);
1635 	if (rc)
1636 		goto error;
1637 
1638 	eth_dev->data->dev_started = 1;
1639 
1640 	bnxt_link_update_op(eth_dev, 0);
1641 
1642 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
1643 		vlan_mask |= RTE_ETH_VLAN_FILTER_MASK;
1644 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
1645 		vlan_mask |= RTE_ETH_VLAN_STRIP_MASK;
1646 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1647 	if (rc)
1648 		goto error;
1649 
1650 	/* Initialize bnxt ULP port details */
1651 	rc = bnxt_ulp_port_init(bp);
1652 	if (rc)
1653 		goto error;
1654 
1655 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1656 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1657 
1658 	bnxt_schedule_fw_health_check(bp);
1659 
1660 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1661 		bnxt_schedule_ptp_alarm(bp);
1662 
1663 	return 0;
1664 
1665 error:
1666 	bnxt_dev_stop(eth_dev);
1667 	return rc;
1668 }
1669 
1670 static void
1671 bnxt_uninit_locks(struct bnxt *bp)
1672 {
1673 	pthread_mutex_destroy(&bp->flow_lock);
1674 	pthread_mutex_destroy(&bp->def_cp_lock);
1675 	pthread_mutex_destroy(&bp->health_check_lock);
1676 	pthread_mutex_destroy(&bp->err_recovery_lock);
1677 	if (bp->rep_info) {
1678 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1679 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1680 	}
1681 }
1682 
1683 static void bnxt_drv_uninit(struct bnxt *bp)
1684 {
1685 	bnxt_free_leds_info(bp);
1686 	bnxt_free_cos_queues(bp);
1687 	bnxt_free_link_info(bp);
1688 	bnxt_free_parent_info(bp);
1689 	bnxt_uninit_locks(bp);
1690 	bnxt_free_rep_info(bp);
1691 
1692 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1693 	bp->tx_mem_zone = NULL;
1694 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1695 	bp->rx_mem_zone = NULL;
1696 
1697 	bnxt_free_vf_info(bp);
1698 	bnxt_free_pf_info(bp);
1699 
1700 	rte_free(bp->grp_info);
1701 	bp->grp_info = NULL;
1702 }
1703 
1704 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1705 {
1706 	struct bnxt *bp = eth_dev->data->dev_private;
1707 	int ret = 0;
1708 
1709 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1710 		return 0;
1711 
1712 	pthread_mutex_lock(&bp->err_recovery_lock);
1713 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1714 		PMD_DRV_LOG(ERR,
1715 			    "Adapter recovering from error...Please retry\n");
1716 		pthread_mutex_unlock(&bp->err_recovery_lock);
1717 		return -EAGAIN;
1718 	}
1719 	pthread_mutex_unlock(&bp->err_recovery_lock);
1720 
1721 	/* cancel the recovery handler before remove dev */
1722 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1723 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1724 	bnxt_cancel_fc_thread(bp);
1725 	rte_eal_alarm_cancel(bnxt_handle_vf_cfg_change, (void *)bp);
1726 
1727 	if (eth_dev->data->dev_started)
1728 		ret = bnxt_dev_stop(eth_dev);
1729 
1730 	bnxt_uninit_resources(bp, false);
1731 
1732 	bnxt_drv_uninit(bp);
1733 
1734 	return ret;
1735 }
1736 
1737 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1738 				    uint32_t index)
1739 {
1740 	struct bnxt *bp = eth_dev->data->dev_private;
1741 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1742 	struct bnxt_vnic_info *vnic;
1743 	struct bnxt_filter_info *filter, *temp_filter;
1744 	uint32_t i;
1745 
1746 	if (is_bnxt_in_error(bp))
1747 		return;
1748 
1749 	/*
1750 	 * Loop through all VNICs from the specified filter flow pools to
1751 	 * remove the corresponding MAC addr filter
1752 	 */
1753 	for (i = 0; i < bp->nr_vnics; i++) {
1754 		if (!(pool_mask & (1ULL << i)))
1755 			continue;
1756 
1757 		vnic = &bp->vnic_info[i];
1758 		filter = STAILQ_FIRST(&vnic->filter);
1759 		while (filter) {
1760 			temp_filter = STAILQ_NEXT(filter, next);
1761 			if (filter->mac_index == index) {
1762 				STAILQ_REMOVE(&vnic->filter, filter,
1763 						bnxt_filter_info, next);
1764 				bnxt_hwrm_clear_l2_filter(bp, filter);
1765 				bnxt_free_filter(bp, filter);
1766 			}
1767 			filter = temp_filter;
1768 		}
1769 	}
1770 }
1771 
1772 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1773 			       struct rte_ether_addr *mac_addr, uint32_t index,
1774 			       uint32_t pool)
1775 {
1776 	struct bnxt_filter_info *filter;
1777 	int rc = 0;
1778 
1779 	/* Attach requested MAC address to the new l2_filter */
1780 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1781 		if (filter->mac_index == index) {
1782 			PMD_DRV_LOG(DEBUG,
1783 				    "MAC addr already existed for pool %d\n",
1784 				    pool);
1785 			return 0;
1786 		}
1787 	}
1788 
1789 	filter = bnxt_alloc_filter(bp);
1790 	if (!filter) {
1791 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1792 		return -ENODEV;
1793 	}
1794 
1795 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1796 	 * if the MAC that's been programmed now is a different one, then,
1797 	 * copy that addr to filter->l2_addr
1798 	 */
1799 	if (mac_addr)
1800 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1801 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1802 
1803 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1804 	if (!rc) {
1805 		filter->mac_index = index;
1806 		if (filter->mac_index == 0)
1807 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1808 		else
1809 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1810 	} else {
1811 		bnxt_free_filter(bp, filter);
1812 	}
1813 
1814 	return rc;
1815 }
1816 
1817 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1818 				struct rte_ether_addr *mac_addr,
1819 				uint32_t index, uint32_t pool)
1820 {
1821 	struct bnxt *bp = eth_dev->data->dev_private;
1822 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1823 	int rc = 0;
1824 
1825 	rc = is_bnxt_in_error(bp);
1826 	if (rc)
1827 		return rc;
1828 
1829 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1830 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1831 		return -ENOTSUP;
1832 	}
1833 
1834 	if (!vnic) {
1835 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1836 		return -EINVAL;
1837 	}
1838 
1839 	/* Filter settings will get applied when port is started */
1840 	if (!eth_dev->data->dev_started)
1841 		return 0;
1842 
1843 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1844 
1845 	return rc;
1846 }
1847 
1848 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1849 {
1850 	int rc = 0;
1851 	struct bnxt *bp = eth_dev->data->dev_private;
1852 	struct rte_eth_link new;
1853 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1854 			BNXT_MIN_LINK_WAIT_CNT;
1855 
1856 	rc = is_bnxt_in_error(bp);
1857 	if (rc)
1858 		return rc;
1859 
1860 	memset(&new, 0, sizeof(new));
1861 
1862 	if (bp->link_info == NULL)
1863 		goto out;
1864 
1865 	/* Only single function PF can bring the phy down.
1866 	 * In certain scenarios, device is not obliged link down even when forced.
1867 	 * When port is stopped, report link down in those cases.
1868 	 */
1869 	if (!eth_dev->data->dev_started &&
1870 	    (!BNXT_SINGLE_PF(bp) || bnxt_force_link_config(bp)))
1871 		goto out;
1872 
1873 	do {
1874 		/* Retrieve link info from hardware */
1875 		rc = bnxt_get_hwrm_link_config(bp, &new);
1876 		if (rc) {
1877 			new.link_speed = RTE_ETH_LINK_SPEED_100M;
1878 			new.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
1879 			PMD_DRV_LOG(ERR,
1880 				"Failed to retrieve link rc = 0x%x!\n", rc);
1881 			goto out;
1882 		}
1883 
1884 		if (!wait_to_complete || new.link_status)
1885 			break;
1886 
1887 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1888 	} while (cnt--);
1889 
1890 out:
1891 	/* Timed out or success */
1892 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1893 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1894 		rte_eth_linkstatus_set(eth_dev, &new);
1895 		bnxt_print_link_info(eth_dev);
1896 	}
1897 
1898 	return rc;
1899 }
1900 
1901 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1902 {
1903 	struct bnxt *bp = eth_dev->data->dev_private;
1904 	struct bnxt_vnic_info *vnic;
1905 	uint32_t old_flags;
1906 	int rc;
1907 
1908 	rc = is_bnxt_in_error(bp);
1909 	if (rc)
1910 		return rc;
1911 
1912 	/* Filter settings will get applied when port is started */
1913 	if (!eth_dev->data->dev_started)
1914 		return 0;
1915 
1916 	if (bp->vnic_info == NULL)
1917 		return 0;
1918 
1919 	vnic = bnxt_get_default_vnic(bp);
1920 
1921 	old_flags = vnic->flags;
1922 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1923 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1924 	if (rc != 0)
1925 		vnic->flags = old_flags;
1926 
1927 	return rc;
1928 }
1929 
1930 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1931 {
1932 	struct bnxt *bp = eth_dev->data->dev_private;
1933 	struct bnxt_vnic_info *vnic;
1934 	uint32_t old_flags;
1935 	int rc;
1936 
1937 	rc = is_bnxt_in_error(bp);
1938 	if (rc)
1939 		return rc;
1940 
1941 	/* Filter settings will get applied when port is started */
1942 	if (!eth_dev->data->dev_started)
1943 		return 0;
1944 
1945 	if (bp->vnic_info == NULL)
1946 		return 0;
1947 
1948 	vnic = bnxt_get_default_vnic(bp);
1949 
1950 	old_flags = vnic->flags;
1951 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1952 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1953 	if (rc != 0)
1954 		vnic->flags = old_flags;
1955 
1956 	return rc;
1957 }
1958 
1959 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1960 {
1961 	struct bnxt *bp = eth_dev->data->dev_private;
1962 	struct bnxt_vnic_info *vnic;
1963 	uint32_t old_flags;
1964 	int rc;
1965 
1966 	rc = is_bnxt_in_error(bp);
1967 	if (rc)
1968 		return rc;
1969 
1970 	/* Filter settings will get applied when port is started */
1971 	if (!eth_dev->data->dev_started)
1972 		return 0;
1973 
1974 	if (bp->vnic_info == NULL)
1975 		return 0;
1976 
1977 	vnic = bnxt_get_default_vnic(bp);
1978 
1979 	old_flags = vnic->flags;
1980 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1981 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1982 	if (rc != 0)
1983 		vnic->flags = old_flags;
1984 
1985 	return rc;
1986 }
1987 
1988 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1989 {
1990 	struct bnxt *bp = eth_dev->data->dev_private;
1991 	struct bnxt_vnic_info *vnic;
1992 	uint32_t old_flags;
1993 	int rc;
1994 
1995 	rc = is_bnxt_in_error(bp);
1996 	if (rc)
1997 		return rc;
1998 
1999 	/* Filter settings will get applied when port is started */
2000 	if (!eth_dev->data->dev_started)
2001 		return 0;
2002 
2003 	if (bp->vnic_info == NULL)
2004 		return 0;
2005 
2006 	vnic = bnxt_get_default_vnic(bp);
2007 
2008 	old_flags = vnic->flags;
2009 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2010 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2011 	if (rc != 0)
2012 		vnic->flags = old_flags;
2013 
2014 	return rc;
2015 }
2016 
2017 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
2018 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
2019 {
2020 	if (qid >= bp->rx_nr_rings)
2021 		return NULL;
2022 
2023 	return bp->eth_dev->data->rx_queues[qid];
2024 }
2025 
2026 /* Return rxq corresponding to a given rss table ring/group ID. */
2027 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
2028 {
2029 	struct bnxt_rx_queue *rxq;
2030 	unsigned int i;
2031 
2032 	if (!BNXT_HAS_RING_GRPS(bp)) {
2033 		for (i = 0; i < bp->rx_nr_rings; i++) {
2034 			rxq = bp->eth_dev->data->rx_queues[i];
2035 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
2036 				return rxq->index;
2037 		}
2038 	} else {
2039 		for (i = 0; i < bp->rx_nr_rings; i++) {
2040 			if (bp->grp_info[i].fw_grp_id == fwr)
2041 				return i;
2042 		}
2043 	}
2044 
2045 	return INVALID_HW_RING_ID;
2046 }
2047 
2048 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
2049 			    struct rte_eth_rss_reta_entry64 *reta_conf,
2050 			    uint16_t reta_size)
2051 {
2052 	struct bnxt *bp = eth_dev->data->dev_private;
2053 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2054 	struct bnxt_vnic_info *vnic = bnxt_get_default_vnic(bp);
2055 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2056 	uint16_t idx, sft;
2057 	int i, rc;
2058 
2059 	rc = is_bnxt_in_error(bp);
2060 	if (rc)
2061 		return rc;
2062 
2063 	if (!vnic->rss_table)
2064 		return -EINVAL;
2065 
2066 	if (!(dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG))
2067 		return -EINVAL;
2068 
2069 	if (reta_size != tbl_size) {
2070 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2071 			"(%d) must equal the size supported by the hardware "
2072 			"(%d)\n", reta_size, tbl_size);
2073 		return -EINVAL;
2074 	}
2075 
2076 	if (bnxt_vnic_reta_config_update(bp, vnic, reta_conf, reta_size)) {
2077 		PMD_DRV_LOG(ERR, "Error in setting the reta config\n");
2078 		return -EINVAL;
2079 	}
2080 	for (i = 0; i < reta_size; i++) {
2081 		struct bnxt_rx_queue *rxq;
2082 
2083 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2084 		sft = i % RTE_ETH_RETA_GROUP_SIZE;
2085 
2086 		if (!(reta_conf[idx].mask & (1ULL << sft)))
2087 			continue;
2088 
2089 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2090 		if (BNXT_CHIP_P5(bp)) {
2091 			vnic->rss_table[i * 2] =
2092 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
2093 			vnic->rss_table[i * 2 + 1] =
2094 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
2095 		} else {
2096 			vnic->rss_table[i] =
2097 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2098 		}
2099 	}
2100 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2101 	return rc;
2102 }
2103 
2104 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2105 			      struct rte_eth_rss_reta_entry64 *reta_conf,
2106 			      uint16_t reta_size)
2107 {
2108 	struct bnxt *bp = eth_dev->data->dev_private;
2109 	struct bnxt_vnic_info *vnic = bnxt_get_default_vnic(bp);
2110 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2111 	uint16_t idx, sft, i;
2112 	int rc;
2113 
2114 	rc = is_bnxt_in_error(bp);
2115 	if (rc)
2116 		return rc;
2117 
2118 	if (!vnic)
2119 		return -EINVAL;
2120 	if (!vnic->rss_table)
2121 		return -EINVAL;
2122 
2123 	if (reta_size != tbl_size) {
2124 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2125 			"(%d) must equal the size supported by the hardware "
2126 			"(%d)\n", reta_size, tbl_size);
2127 		return -EINVAL;
2128 	}
2129 
2130 	for (idx = 0, i = 0; i < reta_size; i++) {
2131 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2132 		sft = i % RTE_ETH_RETA_GROUP_SIZE;
2133 
2134 		if (reta_conf[idx].mask & (1ULL << sft)) {
2135 			uint16_t qid;
2136 
2137 			if (BNXT_CHIP_P5(bp))
2138 				qid = bnxt_rss_to_qid(bp,
2139 						      vnic->rss_table[i * 2]);
2140 			else
2141 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2142 
2143 			if (qid == INVALID_HW_RING_ID) {
2144 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2145 				return -EINVAL;
2146 			}
2147 			reta_conf[idx].reta[sft] = qid;
2148 		}
2149 	}
2150 
2151 	return 0;
2152 }
2153 
2154 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2155 				   struct rte_eth_rss_conf *rss_conf)
2156 {
2157 	struct bnxt *bp = eth_dev->data->dev_private;
2158 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2159 	struct bnxt_vnic_info *vnic;
2160 	int rc;
2161 
2162 	rc = is_bnxt_in_error(bp);
2163 	if (rc)
2164 		return rc;
2165 
2166 	/*
2167 	 * If RSS enablement were different than dev_configure,
2168 	 * then return -EINVAL
2169 	 */
2170 	if (dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG) {
2171 		if (!rss_conf->rss_hf)
2172 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2173 	} else {
2174 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2175 			return -EINVAL;
2176 	}
2177 
2178 	/* Update the default RSS VNIC(s) */
2179 	vnic = bnxt_get_default_vnic(bp);
2180 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2181 	vnic->hash_mode =
2182 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2183 					    RTE_ETH_RSS_LEVEL(rss_conf->rss_hf));
2184 
2185 	/* Cache the hash function */
2186 	bp->rss_conf.rss_hf = rss_conf->rss_hf;
2187 
2188 	/*
2189 	 * If hashkey is not specified, use the previously configured
2190 	 * hashkey
2191 	 */
2192 	if (!rss_conf->rss_key)
2193 		goto rss_config;
2194 
2195 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2196 		PMD_DRV_LOG(ERR,
2197 			    "Invalid hashkey length, should be %d bytes\n",
2198 			    HW_HASH_KEY_SIZE);
2199 		return -EINVAL;
2200 	}
2201 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2202 
2203 	/* Cache the hash key */
2204 	memcpy(bp->rss_conf.rss_key, rss_conf->rss_key, HW_HASH_KEY_SIZE);
2205 
2206 rss_config:
2207 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2208 	return rc;
2209 }
2210 
2211 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2212 				     struct rte_eth_rss_conf *rss_conf)
2213 {
2214 	struct bnxt *bp = eth_dev->data->dev_private;
2215 	struct bnxt_vnic_info *vnic = bnxt_get_default_vnic(bp);
2216 	int len, rc;
2217 	uint32_t hash_types;
2218 
2219 	rc = is_bnxt_in_error(bp);
2220 	if (rc)
2221 		return rc;
2222 
2223 	/* RSS configuration is the same for all VNICs */
2224 	if (vnic && vnic->rss_hash_key) {
2225 		if (rss_conf->rss_key) {
2226 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2227 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2228 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2229 		}
2230 
2231 		hash_types = vnic->hash_type;
2232 		rss_conf->rss_hf = 0;
2233 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2234 			rss_conf->rss_hf |= RTE_ETH_RSS_IPV4;
2235 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2236 		}
2237 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2238 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_TCP;
2239 			hash_types &=
2240 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2241 		}
2242 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2243 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_UDP;
2244 			hash_types &=
2245 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2246 		}
2247 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2248 			rss_conf->rss_hf |= RTE_ETH_RSS_IPV6;
2249 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2250 		}
2251 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2252 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_TCP;
2253 			hash_types &=
2254 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2255 		}
2256 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2257 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_UDP;
2258 			hash_types &=
2259 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2260 		}
2261 
2262 		rss_conf->rss_hf |=
2263 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2264 
2265 		if (hash_types) {
2266 			PMD_DRV_LOG(ERR,
2267 				"Unknown RSS config from firmware (%08x), RSS disabled",
2268 				vnic->hash_type);
2269 			return -ENOTSUP;
2270 		}
2271 	} else {
2272 		rss_conf->rss_hf = 0;
2273 	}
2274 	return 0;
2275 }
2276 
2277 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2278 			       struct rte_eth_fc_conf *fc_conf)
2279 {
2280 	struct bnxt *bp = dev->data->dev_private;
2281 	struct rte_eth_link link_info;
2282 	int rc;
2283 
2284 	rc = is_bnxt_in_error(bp);
2285 	if (rc)
2286 		return rc;
2287 
2288 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2289 	if (rc)
2290 		return rc;
2291 
2292 	memset(fc_conf, 0, sizeof(*fc_conf));
2293 	if (bp->link_info->auto_pause)
2294 		fc_conf->autoneg = 1;
2295 	switch (bp->link_info->pause) {
2296 	case 0:
2297 		fc_conf->mode = RTE_ETH_FC_NONE;
2298 		break;
2299 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2300 		fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
2301 		break;
2302 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2303 		fc_conf->mode = RTE_ETH_FC_RX_PAUSE;
2304 		break;
2305 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2306 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2307 		fc_conf->mode = RTE_ETH_FC_FULL;
2308 		break;
2309 	}
2310 	return 0;
2311 }
2312 
2313 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2314 			       struct rte_eth_fc_conf *fc_conf)
2315 {
2316 	struct bnxt *bp = dev->data->dev_private;
2317 	int rc;
2318 
2319 	rc = is_bnxt_in_error(bp);
2320 	if (rc)
2321 		return rc;
2322 
2323 	if (!BNXT_SINGLE_PF(bp)) {
2324 		PMD_DRV_LOG(ERR,
2325 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2326 		return -ENOTSUP;
2327 	}
2328 
2329 	switch (fc_conf->mode) {
2330 	case RTE_ETH_FC_NONE:
2331 		bp->link_info->auto_pause = 0;
2332 		bp->link_info->force_pause = 0;
2333 		break;
2334 	case RTE_ETH_FC_RX_PAUSE:
2335 		if (fc_conf->autoneg) {
2336 			bp->link_info->auto_pause =
2337 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2338 			bp->link_info->force_pause = 0;
2339 		} else {
2340 			bp->link_info->auto_pause = 0;
2341 			bp->link_info->force_pause =
2342 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2343 		}
2344 		break;
2345 	case RTE_ETH_FC_TX_PAUSE:
2346 		if (fc_conf->autoneg) {
2347 			bp->link_info->auto_pause =
2348 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2349 			bp->link_info->force_pause = 0;
2350 		} else {
2351 			bp->link_info->auto_pause = 0;
2352 			bp->link_info->force_pause =
2353 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2354 		}
2355 		break;
2356 	case RTE_ETH_FC_FULL:
2357 		if (fc_conf->autoneg) {
2358 			bp->link_info->auto_pause =
2359 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2360 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2361 			bp->link_info->force_pause = 0;
2362 		} else {
2363 			bp->link_info->auto_pause = 0;
2364 			bp->link_info->force_pause =
2365 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2366 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2367 		}
2368 		break;
2369 	}
2370 	return bnxt_set_hwrm_link_config(bp, true);
2371 }
2372 
2373 /* Add UDP tunneling port */
2374 int
2375 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2376 			 struct rte_eth_udp_tunnel *udp_tunnel)
2377 {
2378 	struct bnxt *bp = eth_dev->data->dev_private;
2379 	uint16_t tunnel_type = 0;
2380 	int rc = 0;
2381 
2382 	rc = is_bnxt_in_error(bp);
2383 	if (rc)
2384 		return rc;
2385 
2386 	switch (udp_tunnel->prot_type) {
2387 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
2388 		if (bp->vxlan_port_cnt) {
2389 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2390 				udp_tunnel->udp_port);
2391 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2392 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2393 				return -ENOSPC;
2394 			}
2395 			bp->vxlan_port_cnt++;
2396 			return 0;
2397 		}
2398 		tunnel_type =
2399 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2400 		break;
2401 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2402 		if (bp->geneve_port_cnt) {
2403 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2404 				udp_tunnel->udp_port);
2405 			if (bp->geneve_port != udp_tunnel->udp_port) {
2406 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2407 				return -ENOSPC;
2408 			}
2409 			bp->geneve_port_cnt++;
2410 			return 0;
2411 		}
2412 		tunnel_type =
2413 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2414 		break;
2415 	case RTE_ETH_TUNNEL_TYPE_ECPRI:
2416 		if (bp->ecpri_port_cnt) {
2417 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2418 				udp_tunnel->udp_port);
2419 			if (bp->ecpri_port != udp_tunnel->udp_port) {
2420 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2421 				return -ENOSPC;
2422 			}
2423 			bp->ecpri_port_cnt++;
2424 			return 0;
2425 		}
2426 		tunnel_type =
2427 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_ECPRI;
2428 		break;
2429 	default:
2430 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2431 		return -ENOTSUP;
2432 	}
2433 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2434 					     tunnel_type);
2435 
2436 	if (rc != 0)
2437 		return rc;
2438 
2439 	if (tunnel_type ==
2440 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN)
2441 		bp->vxlan_port_cnt++;
2442 
2443 	if (tunnel_type ==
2444 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE)
2445 		bp->geneve_port_cnt++;
2446 
2447 	if (tunnel_type ==
2448 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_ECPRI)
2449 		bp->ecpri_port_cnt++;
2450 
2451 	return rc;
2452 }
2453 
2454 int
2455 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2456 			 struct rte_eth_udp_tunnel *udp_tunnel)
2457 {
2458 	struct bnxt *bp = eth_dev->data->dev_private;
2459 	uint16_t tunnel_type = 0;
2460 	uint16_t port = 0;
2461 	int rc = 0;
2462 
2463 	rc = is_bnxt_in_error(bp);
2464 	if (rc)
2465 		return rc;
2466 
2467 	switch (udp_tunnel->prot_type) {
2468 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
2469 		if (!bp->vxlan_port_cnt) {
2470 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2471 			return -EINVAL;
2472 		}
2473 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2474 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2475 				udp_tunnel->udp_port, bp->vxlan_port);
2476 			return -EINVAL;
2477 		}
2478 		if (--bp->vxlan_port_cnt)
2479 			return 0;
2480 
2481 		tunnel_type =
2482 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2483 		port = bp->vxlan_fw_dst_port_id;
2484 		break;
2485 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2486 		if (!bp->geneve_port_cnt) {
2487 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2488 			return -EINVAL;
2489 		}
2490 		if (bp->geneve_port != udp_tunnel->udp_port) {
2491 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2492 				udp_tunnel->udp_port, bp->geneve_port);
2493 			return -EINVAL;
2494 		}
2495 		if (--bp->geneve_port_cnt)
2496 			return 0;
2497 
2498 		tunnel_type =
2499 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2500 		port = bp->geneve_fw_dst_port_id;
2501 		break;
2502 	case RTE_ETH_TUNNEL_TYPE_ECPRI:
2503 		if (!bp->ecpri_port_cnt) {
2504 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2505 			return -EINVAL;
2506 		}
2507 		if (bp->ecpri_port != udp_tunnel->udp_port) {
2508 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2509 				udp_tunnel->udp_port, bp->ecpri_port);
2510 			return -EINVAL;
2511 		}
2512 		if (--bp->ecpri_port_cnt)
2513 			return 0;
2514 
2515 		tunnel_type =
2516 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_ECPRI;
2517 		port = bp->ecpri_fw_dst_port_id;
2518 		break;
2519 	default:
2520 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2521 		return -ENOTSUP;
2522 	}
2523 
2524 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2525 	return rc;
2526 }
2527 
2528 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2529 {
2530 	struct bnxt_filter_info *filter;
2531 	struct bnxt_vnic_info *vnic;
2532 	int rc = 0;
2533 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2534 
2535 	vnic = bnxt_get_default_vnic(bp);
2536 	filter = STAILQ_FIRST(&vnic->filter);
2537 	while (filter) {
2538 		/* Search for this matching MAC+VLAN filter */
2539 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2540 			/* Delete the filter */
2541 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2542 			if (rc)
2543 				return rc;
2544 			STAILQ_REMOVE(&vnic->filter, filter,
2545 				      bnxt_filter_info, next);
2546 			bnxt_free_filter(bp, filter);
2547 			PMD_DRV_LOG(INFO,
2548 				    "Deleted vlan filter for %d\n",
2549 				    vlan_id);
2550 			return 0;
2551 		}
2552 		filter = STAILQ_NEXT(filter, next);
2553 	}
2554 	return -ENOENT;
2555 }
2556 
2557 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2558 {
2559 	struct bnxt_filter_info *filter;
2560 	struct bnxt_vnic_info *vnic;
2561 	int rc = 0;
2562 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2563 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2564 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2565 
2566 	/* Implementation notes on the use of VNIC in this command:
2567 	 *
2568 	 * By default, these filters belong to default vnic for the function.
2569 	 * Once these filters are set up, only destination VNIC can be modified.
2570 	 * If the destination VNIC is not specified in this command,
2571 	 * then the HWRM shall only create an l2 context id.
2572 	 */
2573 
2574 	vnic = bnxt_get_default_vnic(bp);
2575 	filter = STAILQ_FIRST(&vnic->filter);
2576 	/* Check if the VLAN has already been added */
2577 	while (filter) {
2578 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2579 			return -EEXIST;
2580 
2581 		filter = STAILQ_NEXT(filter, next);
2582 	}
2583 
2584 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2585 	 * command to create MAC+VLAN filter with the right flags, enables set.
2586 	 */
2587 	filter = bnxt_alloc_filter(bp);
2588 	if (!filter) {
2589 		PMD_DRV_LOG(ERR,
2590 			    "MAC/VLAN filter alloc failed\n");
2591 		return -ENOMEM;
2592 	}
2593 	/* MAC + VLAN ID filter */
2594 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2595 	 * untagged packets are received
2596 	 *
2597 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2598 	 * packets and only the programmed vlan's packets are received
2599 	 */
2600 	filter->l2_ivlan = vlan_id;
2601 	filter->l2_ivlan_mask = 0x0FFF;
2602 	filter->enables |= en;
2603 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2604 
2605 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2606 	if (rc) {
2607 		/* Free the newly allocated filter as we were
2608 		 * not able to create the filter in hardware.
2609 		 */
2610 		bnxt_free_filter(bp, filter);
2611 		return rc;
2612 	}
2613 
2614 	filter->mac_index = 0;
2615 	/* Add this new filter to the list */
2616 	if (vlan_id == 0)
2617 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2618 	else
2619 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2620 
2621 	PMD_DRV_LOG(INFO,
2622 		    "Added Vlan filter for %d\n", vlan_id);
2623 	return rc;
2624 }
2625 
2626 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2627 		uint16_t vlan_id, int on)
2628 {
2629 	struct bnxt *bp = eth_dev->data->dev_private;
2630 	int rc;
2631 
2632 	rc = is_bnxt_in_error(bp);
2633 	if (rc)
2634 		return rc;
2635 
2636 	if (!eth_dev->data->dev_started) {
2637 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2638 		return -EINVAL;
2639 	}
2640 
2641 	/* These operations apply to ALL existing MAC/VLAN filters */
2642 	if (on)
2643 		return bnxt_add_vlan_filter(bp, vlan_id);
2644 	else
2645 		return bnxt_del_vlan_filter(bp, vlan_id);
2646 }
2647 
2648 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2649 				    struct bnxt_vnic_info *vnic)
2650 {
2651 	struct bnxt_filter_info *filter;
2652 	int rc;
2653 
2654 	filter = STAILQ_FIRST(&vnic->filter);
2655 	while (filter) {
2656 		if (filter->mac_index == 0 &&
2657 		    !memcmp(filter->l2_addr, bp->mac_addr,
2658 			    RTE_ETHER_ADDR_LEN)) {
2659 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2660 			if (!rc) {
2661 				STAILQ_REMOVE(&vnic->filter, filter,
2662 					      bnxt_filter_info, next);
2663 				bnxt_free_filter(bp, filter);
2664 			}
2665 			return rc;
2666 		}
2667 		filter = STAILQ_NEXT(filter, next);
2668 	}
2669 	return 0;
2670 }
2671 
2672 static int
2673 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2674 {
2675 	struct bnxt_vnic_info *vnic;
2676 	unsigned int i;
2677 	int rc;
2678 
2679 	vnic = bnxt_get_default_vnic(bp);
2680 	if (!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)) {
2681 		/* Remove any VLAN filters programmed */
2682 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2683 			bnxt_del_vlan_filter(bp, i);
2684 
2685 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2686 		if (rc)
2687 			return rc;
2688 	} else {
2689 		/* Default filter will allow packets that match the
2690 		 * dest mac. So, it has to be deleted, otherwise, we
2691 		 * will endup receiving vlan packets for which the
2692 		 * filter is not programmed, when hw-vlan-filter
2693 		 * configuration is ON
2694 		 */
2695 		bnxt_del_dflt_mac_filter(bp, vnic);
2696 		/* This filter will allow only untagged packets */
2697 		bnxt_add_vlan_filter(bp, 0);
2698 	}
2699 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2700 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER));
2701 
2702 	return 0;
2703 }
2704 
2705 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2706 {
2707 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2708 	unsigned int i;
2709 	int rc;
2710 
2711 	/* Destroy vnic filters and vnic */
2712 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2713 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2714 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2715 			bnxt_del_vlan_filter(bp, i);
2716 	}
2717 	bnxt_del_dflt_mac_filter(bp, vnic);
2718 
2719 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2720 	if (rc)
2721 		return rc;
2722 
2723 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2724 	if (rc)
2725 		return rc;
2726 
2727 	rte_free(vnic->fw_grp_ids);
2728 	vnic->fw_grp_ids = NULL;
2729 
2730 	vnic->rx_queue_cnt = 0;
2731 
2732 	return 0;
2733 }
2734 
2735 static int
2736 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2737 {
2738 	struct bnxt_vnic_info *vnic = bnxt_get_default_vnic(bp);
2739 	int rc;
2740 
2741 	/* Destroy, recreate and reconfigure the default vnic */
2742 	rc = bnxt_free_one_vnic(bp, bp->vnic_queue_db.dflt_vnic_id);
2743 	if (rc)
2744 		return rc;
2745 
2746 	/* setup the default vnic details*/
2747 	bnxt_vnic_queue_db_update_dlft_vnic(bp);
2748 
2749 	rc = bnxt_setup_one_vnic(bp, bp->vnic_queue_db.dflt_vnic_id);
2750 	if (rc)
2751 		return rc;
2752 
2753 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2754 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2755 		rc = bnxt_add_vlan_filter(bp, 0);
2756 		if (rc)
2757 			return rc;
2758 		rc = bnxt_restore_vlan_filters(bp);
2759 		if (rc)
2760 			return rc;
2761 	} else {
2762 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2763 		if (rc)
2764 			return rc;
2765 	}
2766 
2767 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2768 	if (rc)
2769 		return rc;
2770 
2771 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2772 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP));
2773 
2774 	return rc;
2775 }
2776 
2777 static int
2778 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2779 {
2780 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2781 	struct bnxt *bp = dev->data->dev_private;
2782 	int rc;
2783 
2784 	rc = is_bnxt_in_error(bp);
2785 	if (rc)
2786 		return rc;
2787 
2788 	/* Filter settings will get applied when port is started */
2789 	if (!dev->data->dev_started)
2790 		return 0;
2791 
2792 	if (mask & RTE_ETH_VLAN_FILTER_MASK) {
2793 		/* Enable or disable VLAN filtering */
2794 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2795 		if (rc)
2796 			return rc;
2797 	}
2798 
2799 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
2800 		/* Enable or disable VLAN stripping */
2801 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2802 		if (rc)
2803 			return rc;
2804 	}
2805 
2806 	if (mask & RTE_ETH_VLAN_EXTEND_MASK) {
2807 		if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND)
2808 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2809 		else
2810 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2811 	}
2812 
2813 	return 0;
2814 }
2815 
2816 static int
2817 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2818 		      uint16_t tpid)
2819 {
2820 	struct bnxt *bp = dev->data->dev_private;
2821 	int qinq = dev->data->dev_conf.rxmode.offloads &
2822 		   RTE_ETH_RX_OFFLOAD_VLAN_EXTEND;
2823 
2824 	if (vlan_type != RTE_ETH_VLAN_TYPE_INNER &&
2825 	    vlan_type != RTE_ETH_VLAN_TYPE_OUTER) {
2826 		PMD_DRV_LOG(ERR,
2827 			    "Unsupported vlan type.");
2828 		return -EINVAL;
2829 	}
2830 	if (!qinq) {
2831 		PMD_DRV_LOG(ERR,
2832 			    "QinQ not enabled. Needs to be ON as we can "
2833 			    "accelerate only outer vlan\n");
2834 		return -EINVAL;
2835 	}
2836 
2837 	if (vlan_type == RTE_ETH_VLAN_TYPE_OUTER) {
2838 		switch (tpid) {
2839 		case RTE_ETHER_TYPE_QINQ:
2840 			bp->outer_tpid_bd =
2841 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2842 				break;
2843 		case RTE_ETHER_TYPE_VLAN:
2844 			bp->outer_tpid_bd =
2845 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2846 				break;
2847 		case RTE_ETHER_TYPE_QINQ1:
2848 			bp->outer_tpid_bd =
2849 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2850 				break;
2851 		case RTE_ETHER_TYPE_QINQ2:
2852 			bp->outer_tpid_bd =
2853 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2854 				break;
2855 		case RTE_ETHER_TYPE_QINQ3:
2856 			bp->outer_tpid_bd =
2857 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2858 				break;
2859 		default:
2860 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2861 			return -EINVAL;
2862 		}
2863 		bp->outer_tpid_bd |= tpid;
2864 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2865 	} else if (vlan_type == RTE_ETH_VLAN_TYPE_INNER) {
2866 		PMD_DRV_LOG(ERR,
2867 			    "Can accelerate only outer vlan in QinQ\n");
2868 		return -EINVAL;
2869 	}
2870 
2871 	return 0;
2872 }
2873 
2874 static int
2875 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2876 			     struct rte_ether_addr *addr)
2877 {
2878 	struct bnxt *bp = dev->data->dev_private;
2879 	/* Default Filter is tied to VNIC 0 */
2880 	struct bnxt_vnic_info *vnic = bnxt_get_default_vnic(bp);
2881 	int rc;
2882 
2883 	rc = is_bnxt_in_error(bp);
2884 	if (rc)
2885 		return rc;
2886 
2887 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2888 		return -EPERM;
2889 
2890 	if (rte_is_zero_ether_addr(addr))
2891 		return -EINVAL;
2892 
2893 	/* Filter settings will get applied when port is started */
2894 	if (!dev->data->dev_started)
2895 		return 0;
2896 
2897 	/* Check if the requested MAC is already added */
2898 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2899 		return 0;
2900 
2901 	/* Destroy filter and re-create it */
2902 	bnxt_del_dflt_mac_filter(bp, vnic);
2903 
2904 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2905 	if (dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2906 		/* This filter will allow only untagged packets */
2907 		rc = bnxt_add_vlan_filter(bp, 0);
2908 	} else {
2909 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2910 	}
2911 
2912 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2913 	return rc;
2914 }
2915 
2916 static int
2917 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2918 			  struct rte_ether_addr *mc_addr_set,
2919 			  uint32_t nb_mc_addr)
2920 {
2921 	struct bnxt *bp = eth_dev->data->dev_private;
2922 	struct bnxt_vnic_info *vnic;
2923 	uint32_t i = 0;
2924 	int rc;
2925 
2926 	rc = is_bnxt_in_error(bp);
2927 	if (rc)
2928 		return rc;
2929 
2930 	vnic = bnxt_get_default_vnic(bp);
2931 
2932 	bp->nb_mc_addr = nb_mc_addr;
2933 
2934 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2935 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2936 		goto allmulti;
2937 	}
2938 
2939 	/* TODO Check for Duplicate mcast addresses */
2940 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2941 	for (i = 0; i < nb_mc_addr; i++)
2942 		rte_ether_addr_copy(&mc_addr_set[i], &bp->mcast_addr_list[i]);
2943 
2944 	if (bp->nb_mc_addr)
2945 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2946 	else
2947 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2948 
2949 allmulti:
2950 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2951 }
2952 
2953 static int
2954 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2955 {
2956 	struct bnxt *bp = dev->data->dev_private;
2957 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2958 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2959 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2960 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2961 	int ret;
2962 
2963 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2964 			fw_major, fw_minor, fw_updt, fw_rsvd);
2965 	if (ret < 0)
2966 		return -EINVAL;
2967 
2968 	ret += 1; /* add the size of '\0' */
2969 	if (fw_size < (size_t)ret)
2970 		return ret;
2971 	else
2972 		return 0;
2973 }
2974 
2975 static void
2976 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2977 	struct rte_eth_rxq_info *qinfo)
2978 {
2979 	struct bnxt *bp = dev->data->dev_private;
2980 	struct bnxt_rx_queue *rxq;
2981 
2982 	if (is_bnxt_in_error(bp))
2983 		return;
2984 
2985 	rxq = dev->data->rx_queues[queue_id];
2986 
2987 	qinfo->mp = rxq->mb_pool;
2988 	qinfo->scattered_rx = dev->data->scattered_rx;
2989 	qinfo->nb_desc = rxq->nb_rx_desc;
2990 
2991 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2992 	qinfo->conf.rx_drop_en = rxq->drop_en;
2993 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2994 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2995 }
2996 
2997 static void
2998 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2999 	struct rte_eth_txq_info *qinfo)
3000 {
3001 	struct bnxt *bp = dev->data->dev_private;
3002 	struct bnxt_tx_queue *txq;
3003 
3004 	if (is_bnxt_in_error(bp))
3005 		return;
3006 
3007 	txq = dev->data->tx_queues[queue_id];
3008 
3009 	qinfo->nb_desc = txq->nb_tx_desc;
3010 
3011 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
3012 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
3013 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
3014 
3015 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
3016 	qinfo->conf.tx_rs_thresh = 0;
3017 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
3018 	qinfo->conf.offloads = txq->offloads;
3019 }
3020 
3021 static const struct {
3022 	eth_rx_burst_t pkt_burst;
3023 	const char *info;
3024 } bnxt_rx_burst_info[] = {
3025 	{bnxt_recv_pkts,		"Scalar"},
3026 #if defined(RTE_ARCH_X86)
3027 	{bnxt_recv_pkts_vec,		"Vector SSE"},
3028 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
3029 #endif
3030 #if defined(RTE_ARCH_ARM64)
3031 	{bnxt_recv_pkts_vec,		"Vector Neon"},
3032 #endif
3033 };
3034 
3035 static int
3036 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
3037 		       struct rte_eth_burst_mode *mode)
3038 {
3039 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
3040 	size_t i;
3041 
3042 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
3043 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
3044 			snprintf(mode->info, sizeof(mode->info), "%s",
3045 				 bnxt_rx_burst_info[i].info);
3046 			return 0;
3047 		}
3048 	}
3049 
3050 	return -EINVAL;
3051 }
3052 
3053 static const struct {
3054 	eth_tx_burst_t pkt_burst;
3055 	const char *info;
3056 } bnxt_tx_burst_info[] = {
3057 	{bnxt_xmit_pkts,		"Scalar"},
3058 #if defined(RTE_ARCH_X86)
3059 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
3060 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
3061 #endif
3062 #if defined(RTE_ARCH_ARM64)
3063 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
3064 #endif
3065 };
3066 
3067 static int
3068 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
3069 		       struct rte_eth_burst_mode *mode)
3070 {
3071 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
3072 	size_t i;
3073 
3074 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
3075 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
3076 			snprintf(mode->info, sizeof(mode->info), "%s",
3077 				 bnxt_tx_burst_info[i].info);
3078 			return 0;
3079 		}
3080 	}
3081 
3082 	return -EINVAL;
3083 }
3084 
3085 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3086 {
3087 	struct bnxt *bp = eth_dev->data->dev_private;
3088 	uint32_t rc = 0;
3089 
3090 	rc = is_bnxt_in_error(bp);
3091 	if (rc)
3092 		return rc;
3093 
3094 	/* Return if port is active */
3095 	if (eth_dev->data->dev_started) {
3096 		PMD_DRV_LOG(ERR, "Stop port before changing MTU\n");
3097 		return -EBUSY;
3098 	}
3099 
3100 	/* Exit if receive queues are not configured yet */
3101 	if (!eth_dev->data->nb_rx_queues)
3102 		return -ENOTSUP;
3103 
3104 	/* Is there a change in mtu setting? */
3105 	if (eth_dev->data->mtu == new_mtu)
3106 		return rc;
3107 
3108 	if (new_mtu > RTE_ETHER_MTU)
3109 		bp->flags |= BNXT_FLAG_JUMBO;
3110 	else
3111 		bp->flags &= ~BNXT_FLAG_JUMBO;
3112 
3113 	rc = bnxt_vnic_mru_config(bp, new_mtu);
3114 	if (rc) {
3115 		PMD_DRV_LOG(ERR, "failed to update mtu in vnic context\n");
3116 		return rc;
3117 	}
3118 
3119 	if (bnxt_hwrm_config_host_mtu(bp))
3120 		PMD_DRV_LOG(WARNING, "Failed to configure host MTU\n");
3121 
3122 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3123 
3124 	return rc;
3125 }
3126 
3127 static int
3128 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3129 {
3130 	struct bnxt *bp = dev->data->dev_private;
3131 	uint16_t vlan = bp->vlan;
3132 	int rc;
3133 
3134 	rc = is_bnxt_in_error(bp);
3135 	if (rc)
3136 		return rc;
3137 
3138 	if (!BNXT_SINGLE_PF(bp)) {
3139 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3140 		return -ENOTSUP;
3141 	}
3142 	bp->vlan = on ? pvid : 0;
3143 
3144 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3145 	if (rc)
3146 		bp->vlan = vlan;
3147 	return rc;
3148 }
3149 
3150 static int
3151 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3152 {
3153 	struct bnxt *bp = dev->data->dev_private;
3154 	int rc;
3155 
3156 	rc = is_bnxt_in_error(bp);
3157 	if (rc)
3158 		return rc;
3159 
3160 	return bnxt_hwrm_port_led_cfg(bp, true);
3161 }
3162 
3163 static int
3164 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3165 {
3166 	struct bnxt *bp = dev->data->dev_private;
3167 	int rc;
3168 
3169 	rc = is_bnxt_in_error(bp);
3170 	if (rc)
3171 		return rc;
3172 
3173 	return bnxt_hwrm_port_led_cfg(bp, false);
3174 }
3175 
3176 static uint32_t
3177 bnxt_rx_queue_count_op(void *rx_queue)
3178 {
3179 	struct bnxt *bp;
3180 	struct bnxt_cp_ring_info *cpr;
3181 	uint32_t desc = 0, raw_cons, cp_ring_size;
3182 	struct bnxt_rx_queue *rxq;
3183 	struct rx_pkt_cmpl *rxcmp;
3184 	int rc;
3185 
3186 	rxq = rx_queue;
3187 	bp = rxq->bp;
3188 
3189 	rc = is_bnxt_in_error(bp);
3190 	if (rc)
3191 		return rc;
3192 
3193 	cpr = rxq->cp_ring;
3194 	raw_cons = cpr->cp_raw_cons;
3195 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3196 
3197 	while (1) {
3198 		uint32_t agg_cnt, cons, cmpl_type;
3199 
3200 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3201 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3202 
3203 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3204 			break;
3205 
3206 		cmpl_type = CMP_TYPE(rxcmp);
3207 
3208 		switch (cmpl_type) {
3209 		case CMPL_BASE_TYPE_RX_L2:
3210 		case CMPL_BASE_TYPE_RX_L2_V2:
3211 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3212 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3213 			desc++;
3214 			break;
3215 
3216 		case CMPL_BASE_TYPE_RX_TPA_END:
3217 			if (BNXT_CHIP_P5(rxq->bp)) {
3218 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3219 
3220 				p5_tpa_end = (void *)rxcmp;
3221 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3222 			} else {
3223 				struct rx_tpa_end_cmpl *tpa_end;
3224 
3225 				tpa_end = (void *)rxcmp;
3226 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3227 			}
3228 
3229 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3230 			desc++;
3231 			break;
3232 
3233 		default:
3234 			raw_cons += CMP_LEN(cmpl_type);
3235 		}
3236 	}
3237 
3238 	return desc;
3239 }
3240 
3241 static int
3242 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3243 {
3244 	struct bnxt_rx_queue *rxq = rx_queue;
3245 	struct bnxt_cp_ring_info *cpr;
3246 	struct bnxt_rx_ring_info *rxr;
3247 	uint32_t desc, raw_cons, cp_ring_size;
3248 	struct bnxt *bp = rxq->bp;
3249 	struct rx_pkt_cmpl *rxcmp;
3250 	int rc;
3251 
3252 	rc = is_bnxt_in_error(bp);
3253 	if (rc)
3254 		return rc;
3255 
3256 	if (offset >= rxq->nb_rx_desc)
3257 		return -EINVAL;
3258 
3259 	rxr = rxq->rx_ring;
3260 	cpr = rxq->cp_ring;
3261 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3262 
3263 	/*
3264 	 * For the vector receive case, the completion at the requested
3265 	 * offset can be indexed directly.
3266 	 */
3267 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3268 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3269 		struct rx_pkt_cmpl *rxcmp;
3270 		uint32_t cons;
3271 
3272 		/* Check status of completion descriptor. */
3273 		raw_cons = cpr->cp_raw_cons +
3274 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3275 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3276 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3277 
3278 		if (bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3279 			return RTE_ETH_RX_DESC_DONE;
3280 
3281 		/* Check whether rx desc has an mbuf attached. */
3282 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3283 		if (cons >= rxq->rxrearm_start &&
3284 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3285 			return RTE_ETH_RX_DESC_UNAVAIL;
3286 		}
3287 
3288 		return RTE_ETH_RX_DESC_AVAIL;
3289 	}
3290 #endif
3291 
3292 	/*
3293 	 * For the non-vector receive case, scan the completion ring to
3294 	 * locate the completion descriptor for the requested offset.
3295 	 */
3296 	raw_cons = cpr->cp_raw_cons;
3297 	desc = 0;
3298 	while (1) {
3299 		uint32_t agg_cnt, cons, cmpl_type;
3300 
3301 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3302 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3303 
3304 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3305 			break;
3306 
3307 		cmpl_type = CMP_TYPE(rxcmp);
3308 
3309 		switch (cmpl_type) {
3310 		case CMPL_BASE_TYPE_RX_L2:
3311 		case CMPL_BASE_TYPE_RX_L2_V2:
3312 			if (desc == offset) {
3313 				cons = rxcmp->opaque;
3314 				if (rxr->rx_buf_ring[cons])
3315 					return RTE_ETH_RX_DESC_DONE;
3316 				else
3317 					return RTE_ETH_RX_DESC_UNAVAIL;
3318 			}
3319 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3320 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3321 			desc++;
3322 			break;
3323 
3324 		case CMPL_BASE_TYPE_RX_TPA_END:
3325 			if (desc == offset)
3326 				return RTE_ETH_RX_DESC_DONE;
3327 
3328 			if (BNXT_CHIP_P5(rxq->bp)) {
3329 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3330 
3331 				p5_tpa_end = (void *)rxcmp;
3332 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3333 			} else {
3334 				struct rx_tpa_end_cmpl *tpa_end;
3335 
3336 				tpa_end = (void *)rxcmp;
3337 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3338 			}
3339 
3340 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3341 			desc++;
3342 			break;
3343 
3344 		default:
3345 			raw_cons += CMP_LEN(cmpl_type);
3346 		}
3347 	}
3348 
3349 	return RTE_ETH_RX_DESC_AVAIL;
3350 }
3351 
3352 static int
3353 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3354 {
3355 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3356 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3357 	uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3358 	struct cmpl_base *cp_desc_ring;
3359 	int rc;
3360 
3361 	rc = is_bnxt_in_error(txq->bp);
3362 	if (rc)
3363 		return rc;
3364 
3365 	if (offset >= txq->nb_tx_desc)
3366 		return -EINVAL;
3367 
3368 	/* Return "desc done" if descriptor is available for use. */
3369 	if (bnxt_tx_bds_in_hw(txq) <= offset)
3370 		return RTE_ETH_TX_DESC_DONE;
3371 
3372 	raw_cons = cpr->cp_raw_cons;
3373 	cp_desc_ring = cpr->cp_desc_ring;
3374 	ring_mask = cpr->cp_ring_struct->ring_mask;
3375 
3376 	/* Check to see if hw has posted a completion for the descriptor. */
3377 	while (1) {
3378 		struct tx_cmpl *txcmp;
3379 		uint32_t cons;
3380 
3381 		cons = RING_CMPL(ring_mask, raw_cons);
3382 		txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3383 
3384 		if (!bnxt_cpr_cmp_valid(txcmp, raw_cons, ring_mask + 1))
3385 			break;
3386 
3387 		if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3388 			nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3389 
3390 		if (nb_tx_pkts > offset)
3391 			return RTE_ETH_TX_DESC_DONE;
3392 
3393 		raw_cons = NEXT_RAW_CMP(raw_cons);
3394 	}
3395 
3396 	/* Descriptor is pending transmit, not yet completed by hardware. */
3397 	return RTE_ETH_TX_DESC_FULL;
3398 }
3399 
3400 int
3401 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3402 		     const struct rte_flow_ops **ops)
3403 {
3404 	struct bnxt *bp = dev->data->dev_private;
3405 	int ret = 0;
3406 
3407 	if (!bp)
3408 		return -EIO;
3409 
3410 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3411 		struct bnxt_representor *vfr = dev->data->dev_private;
3412 		bp = vfr->parent_dev->data->dev_private;
3413 		/* parent is deleted while children are still valid */
3414 		if (!bp) {
3415 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3416 				    dev->data->port_id);
3417 			return -EIO;
3418 		}
3419 	}
3420 
3421 	ret = is_bnxt_in_error(bp);
3422 	if (ret)
3423 		return ret;
3424 
3425 	/* PMD supports thread-safe flow operations.  rte_flow API
3426 	 * functions can avoid mutex for multi-thread safety.
3427 	 */
3428 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3429 
3430 	if (BNXT_TRUFLOW_EN(bp))
3431 		*ops = &bnxt_ulp_rte_flow_ops;
3432 	else
3433 		*ops = &bnxt_flow_ops;
3434 
3435 	return ret;
3436 }
3437 
3438 static const uint32_t *
3439 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3440 {
3441 	static const uint32_t ptypes[] = {
3442 		RTE_PTYPE_L2_ETHER_VLAN,
3443 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3444 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3445 		RTE_PTYPE_L4_ICMP,
3446 		RTE_PTYPE_L4_TCP,
3447 		RTE_PTYPE_L4_UDP,
3448 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3449 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3450 		RTE_PTYPE_INNER_L4_ICMP,
3451 		RTE_PTYPE_INNER_L4_TCP,
3452 		RTE_PTYPE_INNER_L4_UDP,
3453 		RTE_PTYPE_UNKNOWN
3454 	};
3455 
3456 	if (!dev->rx_pkt_burst)
3457 		return NULL;
3458 
3459 	return ptypes;
3460 }
3461 
3462 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3463 			 int reg_win)
3464 {
3465 	uint32_t reg_base = *reg_arr & 0xfffff000;
3466 	uint32_t win_off;
3467 	int i;
3468 
3469 	for (i = 0; i < count; i++) {
3470 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3471 			return -ERANGE;
3472 	}
3473 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3474 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3475 	return 0;
3476 }
3477 
3478 static int bnxt_map_ptp_regs(struct bnxt *bp)
3479 {
3480 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3481 	uint32_t *reg_arr;
3482 	int rc, i;
3483 
3484 	reg_arr = ptp->rx_regs;
3485 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3486 	if (rc)
3487 		return rc;
3488 
3489 	reg_arr = ptp->tx_regs;
3490 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3491 	if (rc)
3492 		return rc;
3493 
3494 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3495 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3496 
3497 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3498 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3499 
3500 	return 0;
3501 }
3502 
3503 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3504 {
3505 	rte_write32(0, (uint8_t *)bp->bar0 +
3506 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3507 	rte_write32(0, (uint8_t *)bp->bar0 +
3508 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3509 }
3510 
3511 static uint64_t bnxt_cc_read(struct bnxt *bp)
3512 {
3513 	uint64_t ns;
3514 
3515 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3516 			      BNXT_GRCPF_REG_SYNC_TIME));
3517 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3518 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3519 	return ns;
3520 }
3521 
3522 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3523 {
3524 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3525 	uint32_t fifo;
3526 
3527 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3528 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3529 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3530 		return -EAGAIN;
3531 
3532 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3533 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3534 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3535 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3536 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3537 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3538 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3539 
3540 	return 0;
3541 }
3542 
3543 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3544 {
3545 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3546 	struct bnxt_pf_info *pf = bp->pf;
3547 	uint16_t port_id;
3548 	int i = 0;
3549 	uint32_t fifo;
3550 
3551 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3552 		return -EINVAL;
3553 
3554 	port_id = pf->port_id;
3555 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3556 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3557 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3558 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3559 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3560 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3561 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3562 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3563 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3564 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3565 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3566 		i++;
3567 	}
3568 
3569 	if (i >= BNXT_PTP_RX_PND_CNT)
3570 		return -EBUSY;
3571 
3572 	return 0;
3573 }
3574 
3575 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3576 {
3577 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3578 	struct bnxt_pf_info *pf = bp->pf;
3579 	uint16_t port_id;
3580 	uint32_t fifo;
3581 
3582 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3583 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3584 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3585 		return -EAGAIN;
3586 
3587 	port_id = pf->port_id;
3588 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3589 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3590 
3591 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3592 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3593 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3594 		return bnxt_clr_rx_ts(bp, ts);
3595 
3596 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3597 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3598 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3599 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3600 
3601 	return 0;
3602 }
3603 
3604 static int
3605 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3606 {
3607 	uint64_t ns;
3608 	struct bnxt *bp = dev->data->dev_private;
3609 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3610 
3611 	if (!ptp)
3612 		return -ENOTSUP;
3613 
3614 	ns = rte_timespec_to_ns(ts);
3615 	/* Set the timecounters to a new value. */
3616 	ptp->tc.nsec = ns;
3617 	ptp->tx_tstamp_tc.nsec = ns;
3618 	ptp->rx_tstamp_tc.nsec = ns;
3619 
3620 	return 0;
3621 }
3622 
3623 static int
3624 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3625 {
3626 	struct bnxt *bp = dev->data->dev_private;
3627 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3628 	uint64_t ns, systime_cycles = 0;
3629 	int rc = 0;
3630 
3631 	if (!ptp)
3632 		return -ENOTSUP;
3633 
3634 	if (BNXT_CHIP_P5(bp))
3635 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3636 					     &systime_cycles);
3637 	else
3638 		systime_cycles = bnxt_cc_read(bp);
3639 
3640 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3641 	*ts = rte_ns_to_timespec(ns);
3642 
3643 	return rc;
3644 }
3645 static int
3646 bnxt_timesync_enable(struct rte_eth_dev *dev)
3647 {
3648 	struct bnxt *bp = dev->data->dev_private;
3649 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3650 	uint32_t shift = 0;
3651 	int rc;
3652 
3653 	if (!ptp)
3654 		return -ENOTSUP;
3655 
3656 	ptp->rx_filter = 1;
3657 	ptp->tx_tstamp_en = 1;
3658 	ptp->filter_all = 1;
3659 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3660 
3661 	rc = bnxt_hwrm_ptp_cfg(bp);
3662 	if (rc)
3663 		return rc;
3664 
3665 	rte_spinlock_init(&ptp->ptp_lock);
3666 	bp->ptp_all_rx_tstamp = 1;
3667 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3668 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3669 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3670 
3671 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3672 	ptp->tc.cc_shift = shift;
3673 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3674 
3675 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3676 	ptp->rx_tstamp_tc.cc_shift = shift;
3677 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3678 
3679 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3680 	ptp->tx_tstamp_tc.cc_shift = shift;
3681 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3682 
3683 	if (!BNXT_CHIP_P5(bp))
3684 		bnxt_map_ptp_regs(bp);
3685 	else
3686 		rc = bnxt_ptp_start(bp);
3687 
3688 	return rc;
3689 }
3690 
3691 static int
3692 bnxt_timesync_disable(struct rte_eth_dev *dev)
3693 {
3694 	struct bnxt *bp = dev->data->dev_private;
3695 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3696 
3697 	if (!ptp)
3698 		return -ENOTSUP;
3699 
3700 	ptp->rx_filter = 0;
3701 	ptp->tx_tstamp_en = 0;
3702 	ptp->rxctl = 0;
3703 	ptp->filter_all = 0;
3704 
3705 	bnxt_hwrm_ptp_cfg(bp);
3706 
3707 	bp->ptp_all_rx_tstamp = 0;
3708 	if (!BNXT_CHIP_P5(bp))
3709 		bnxt_unmap_ptp_regs(bp);
3710 	else
3711 		bnxt_ptp_stop(bp);
3712 
3713 	return 0;
3714 }
3715 
3716 static int
3717 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3718 				 struct timespec *timestamp,
3719 				 uint32_t flags __rte_unused)
3720 {
3721 	struct bnxt *bp = dev->data->dev_private;
3722 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3723 	uint64_t rx_tstamp_cycles = 0;
3724 	uint64_t ns;
3725 
3726 	if (!ptp)
3727 		return -ENOTSUP;
3728 
3729 	if (BNXT_CHIP_P5(bp))
3730 		rx_tstamp_cycles = ptp->rx_timestamp;
3731 	else
3732 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3733 
3734 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3735 	*timestamp = rte_ns_to_timespec(ns);
3736 	return  0;
3737 }
3738 
3739 static int
3740 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3741 				 struct timespec *timestamp)
3742 {
3743 	struct bnxt *bp = dev->data->dev_private;
3744 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3745 	uint64_t tx_tstamp_cycles = 0;
3746 	uint64_t ns;
3747 	int rc = 0;
3748 
3749 	if (!ptp)
3750 		return -ENOTSUP;
3751 
3752 	if (BNXT_CHIP_P5(bp))
3753 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3754 					     &tx_tstamp_cycles);
3755 	else
3756 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3757 
3758 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3759 	*timestamp = rte_ns_to_timespec(ns);
3760 
3761 	return rc;
3762 }
3763 
3764 static int
3765 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3766 {
3767 	struct bnxt *bp = dev->data->dev_private;
3768 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3769 
3770 	if (!ptp)
3771 		return -ENOTSUP;
3772 
3773 	ptp->tc.nsec += delta;
3774 	ptp->tx_tstamp_tc.nsec += delta;
3775 	ptp->rx_tstamp_tc.nsec += delta;
3776 
3777 	return 0;
3778 }
3779 
3780 static int
3781 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3782 {
3783 	struct bnxt *bp = dev->data->dev_private;
3784 	int rc;
3785 	uint32_t dir_entries;
3786 	uint32_t entry_length;
3787 
3788 	rc = is_bnxt_in_error(bp);
3789 	if (rc)
3790 		return rc;
3791 
3792 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3793 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3794 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3795 
3796 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3797 	if (rc != 0)
3798 		return rc;
3799 
3800 	return dir_entries * entry_length;
3801 }
3802 
3803 static int
3804 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3805 		struct rte_dev_eeprom_info *in_eeprom)
3806 {
3807 	struct bnxt *bp = dev->data->dev_private;
3808 	uint32_t index;
3809 	uint32_t offset;
3810 	int rc;
3811 
3812 	rc = is_bnxt_in_error(bp);
3813 	if (rc)
3814 		return rc;
3815 
3816 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3817 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3818 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3819 		    in_eeprom->offset, in_eeprom->length);
3820 
3821 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3822 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3823 						in_eeprom->data);
3824 
3825 	index = in_eeprom->offset >> 24;
3826 	offset = in_eeprom->offset & 0xffffff;
3827 
3828 	if (index != 0)
3829 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3830 					   in_eeprom->length, in_eeprom->data);
3831 
3832 	return 0;
3833 }
3834 
3835 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3836 {
3837 	switch (dir_type) {
3838 	case BNX_DIR_TYPE_CHIMP_PATCH:
3839 	case BNX_DIR_TYPE_BOOTCODE:
3840 	case BNX_DIR_TYPE_BOOTCODE_2:
3841 	case BNX_DIR_TYPE_APE_FW:
3842 	case BNX_DIR_TYPE_APE_PATCH:
3843 	case BNX_DIR_TYPE_KONG_FW:
3844 	case BNX_DIR_TYPE_KONG_PATCH:
3845 	case BNX_DIR_TYPE_BONO_FW:
3846 	case BNX_DIR_TYPE_BONO_PATCH:
3847 		/* FALLTHROUGH */
3848 		return true;
3849 	}
3850 
3851 	return false;
3852 }
3853 
3854 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3855 {
3856 	switch (dir_type) {
3857 	case BNX_DIR_TYPE_AVS:
3858 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3859 	case BNX_DIR_TYPE_PCIE:
3860 	case BNX_DIR_TYPE_TSCF_UCODE:
3861 	case BNX_DIR_TYPE_EXT_PHY:
3862 	case BNX_DIR_TYPE_CCM:
3863 	case BNX_DIR_TYPE_ISCSI_BOOT:
3864 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3865 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3866 		/* FALLTHROUGH */
3867 		return true;
3868 	}
3869 
3870 	return false;
3871 }
3872 
3873 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3874 {
3875 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3876 		bnxt_dir_type_is_other_exec_format(dir_type);
3877 }
3878 
3879 static int
3880 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3881 		struct rte_dev_eeprom_info *in_eeprom)
3882 {
3883 	struct bnxt *bp = dev->data->dev_private;
3884 	uint8_t index, dir_op;
3885 	uint16_t type, ext, ordinal, attr;
3886 	int rc;
3887 
3888 	rc = is_bnxt_in_error(bp);
3889 	if (rc)
3890 		return rc;
3891 
3892 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3893 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3894 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3895 		    in_eeprom->offset, in_eeprom->length);
3896 
3897 	if (!BNXT_PF(bp)) {
3898 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3899 		return -EINVAL;
3900 	}
3901 
3902 	type = in_eeprom->magic >> 16;
3903 
3904 	if (type == 0xffff) { /* special value for directory operations */
3905 		index = in_eeprom->magic & 0xff;
3906 		dir_op = in_eeprom->magic >> 8;
3907 		if (index == 0)
3908 			return -EINVAL;
3909 		switch (dir_op) {
3910 		case 0x0e: /* erase */
3911 			if (in_eeprom->offset != ~in_eeprom->magic)
3912 				return -EINVAL;
3913 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3914 		default:
3915 			return -EINVAL;
3916 		}
3917 	}
3918 
3919 	/* Create or re-write an NVM item: */
3920 	if (bnxt_dir_type_is_executable(type) == true)
3921 		return -EOPNOTSUPP;
3922 	ext = in_eeprom->magic & 0xffff;
3923 	ordinal = in_eeprom->offset >> 16;
3924 	attr = in_eeprom->offset & 0xffff;
3925 
3926 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3927 				     in_eeprom->data, in_eeprom->length);
3928 }
3929 
3930 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3931 				struct rte_eth_dev_module_info *modinfo)
3932 {
3933 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3934 	struct bnxt *bp = dev->data->dev_private;
3935 	int rc;
3936 
3937 	/* No point in going further if phy status indicates
3938 	 * module is not inserted or if it is powered down or
3939 	 * if it is of type 10GBase-T
3940 	 */
3941 	if (bp->link_info->module_status >
3942 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3943 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3944 			    dev->data->port_id);
3945 		return -ENOTSUP;
3946 	}
3947 
3948 	/* This feature is not supported in older firmware versions */
3949 	if (bp->hwrm_spec_code < 0x10202) {
3950 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3951 			    dev->data->port_id);
3952 		return -ENOTSUP;
3953 	}
3954 
3955 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3956 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3957 						   module_info);
3958 
3959 	if (rc)
3960 		return rc;
3961 
3962 	switch (module_info[0]) {
3963 	case SFF_MODULE_ID_SFP:
3964 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3965 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3966 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3967 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3968 		break;
3969 	case SFF_MODULE_ID_QSFP:
3970 	case SFF_MODULE_ID_QSFP_PLUS:
3971 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3972 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3973 		break;
3974 	case SFF_MODULE_ID_QSFP28:
3975 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3976 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3977 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3978 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3979 		break;
3980 	default:
3981 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3982 		return -ENOTSUP;
3983 	}
3984 
3985 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3986 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3987 
3988 	return 0;
3989 }
3990 
3991 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3992 				  struct rte_dev_eeprom_info *info)
3993 {
3994 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3995 	uint32_t offset = info->offset, length = info->length;
3996 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3997 	struct bnxt *bp = dev->data->dev_private;
3998 	uint8_t *data = info->data;
3999 	uint8_t page = offset >> 7;
4000 	uint8_t max_pages = 2;
4001 	uint8_t opt_pages;
4002 	int rc;
4003 
4004 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
4005 						   SFF_DIAG_SUPPORT_OFFSET + 1,
4006 						   module_info);
4007 	if (rc)
4008 		return rc;
4009 
4010 	switch (module_info[0]) {
4011 	case SFF_MODULE_ID_SFP:
4012 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
4013 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
4014 			pg_addr[2] = I2C_DEV_ADDR_A2;
4015 			pg_addr[3] = I2C_DEV_ADDR_A2;
4016 			max_pages = 4;
4017 		}
4018 		break;
4019 	case SFF_MODULE_ID_QSFP28:
4020 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
4021 							   SFF8636_OPT_PAGES_OFFSET,
4022 							   1, &opt_pages);
4023 		if (rc)
4024 			return rc;
4025 
4026 		if (opt_pages & SFF8636_PAGE1_MASK) {
4027 			pg_addr[2] = I2C_DEV_ADDR_A0;
4028 			max_pages = 3;
4029 		}
4030 		if (opt_pages & SFF8636_PAGE2_MASK) {
4031 			pg_addr[3] = I2C_DEV_ADDR_A0;
4032 			max_pages = 4;
4033 		}
4034 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
4035 			pg_addr[4] = I2C_DEV_ADDR_A0;
4036 			max_pages = 5;
4037 		}
4038 		break;
4039 	default:
4040 		break;
4041 	}
4042 
4043 	memset(data, 0, length);
4044 
4045 	offset &= 0xff;
4046 	while (length && page < max_pages) {
4047 		uint8_t raw_page = page ? page - 1 : 0;
4048 		uint16_t chunk;
4049 
4050 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
4051 			raw_page = 0;
4052 		else if (page)
4053 			offset |= 0x80;
4054 		chunk = RTE_MIN(length, 256 - offset);
4055 
4056 		if (pg_addr[page]) {
4057 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
4058 								   raw_page, offset,
4059 								   chunk, data);
4060 			if (rc)
4061 				return rc;
4062 		}
4063 
4064 		data += chunk;
4065 		length -= chunk;
4066 		offset = 0;
4067 		page += 1 + (chunk > 128);
4068 	}
4069 
4070 	return length ? -EINVAL : 0;
4071 }
4072 
4073 /*
4074  * Initialization
4075  */
4076 
4077 static const struct eth_dev_ops bnxt_dev_ops = {
4078 	.dev_infos_get = bnxt_dev_info_get_op,
4079 	.dev_close = bnxt_dev_close_op,
4080 	.dev_configure = bnxt_dev_configure_op,
4081 	.dev_start = bnxt_dev_start_op,
4082 	.dev_stop = bnxt_dev_stop_op,
4083 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4084 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4085 	.stats_get = bnxt_stats_get_op,
4086 	.stats_reset = bnxt_stats_reset_op,
4087 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4088 	.rx_queue_release = bnxt_rx_queue_release_op,
4089 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4090 	.tx_queue_release = bnxt_tx_queue_release_op,
4091 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4092 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4093 	.reta_update = bnxt_reta_update_op,
4094 	.reta_query = bnxt_reta_query_op,
4095 	.rss_hash_update = bnxt_rss_hash_update_op,
4096 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4097 	.link_update = bnxt_link_update_op,
4098 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4099 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4100 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4101 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4102 	.mac_addr_add = bnxt_mac_addr_add_op,
4103 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4104 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4105 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4106 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4107 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4108 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4109 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4110 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4111 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4112 	.mtu_set = bnxt_mtu_set_op,
4113 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4114 	.xstats_get = bnxt_dev_xstats_get_op,
4115 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4116 	.xstats_reset = bnxt_dev_xstats_reset_op,
4117 	.fw_version_get = bnxt_fw_version_get,
4118 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4119 	.rxq_info_get = bnxt_rxq_info_get_op,
4120 	.txq_info_get = bnxt_txq_info_get_op,
4121 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4122 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4123 	.dev_led_on = bnxt_dev_led_on_op,
4124 	.dev_led_off = bnxt_dev_led_off_op,
4125 	.rx_queue_start = bnxt_rx_queue_start,
4126 	.rx_queue_stop = bnxt_rx_queue_stop,
4127 	.tx_queue_start = bnxt_tx_queue_start,
4128 	.tx_queue_stop = bnxt_tx_queue_stop,
4129 	.flow_ops_get = bnxt_flow_ops_get_op,
4130 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4131 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4132 	.get_eeprom           = bnxt_get_eeprom_op,
4133 	.set_eeprom           = bnxt_set_eeprom_op,
4134 	.get_module_info = bnxt_get_module_info,
4135 	.get_module_eeprom = bnxt_get_module_eeprom,
4136 	.timesync_enable      = bnxt_timesync_enable,
4137 	.timesync_disable     = bnxt_timesync_disable,
4138 	.timesync_read_time   = bnxt_timesync_read_time,
4139 	.timesync_write_time   = bnxt_timesync_write_time,
4140 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4141 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4142 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4143 	.mtr_ops_get = bnxt_flow_meter_ops_get,
4144 };
4145 
4146 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4147 {
4148 	uint32_t offset;
4149 
4150 	/* Only pre-map the reset GRC registers using window 3 */
4151 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4152 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4153 
4154 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4155 
4156 	return offset;
4157 }
4158 
4159 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4160 {
4161 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4162 	uint32_t reg_base = 0xffffffff;
4163 	int i;
4164 
4165 	/* Only pre-map the monitoring GRC registers using window 2 */
4166 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4167 		uint32_t reg = info->status_regs[i];
4168 
4169 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4170 			continue;
4171 
4172 		if (reg_base == 0xffffffff)
4173 			reg_base = reg & 0xfffff000;
4174 		if ((reg & 0xfffff000) != reg_base)
4175 			return -ERANGE;
4176 
4177 		/* Use mask 0xffc as the Lower 2 bits indicates
4178 		 * address space location
4179 		 */
4180 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4181 						(reg & 0xffc);
4182 	}
4183 
4184 	if (reg_base == 0xffffffff)
4185 		return 0;
4186 
4187 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4188 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4189 
4190 	return 0;
4191 }
4192 
4193 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4194 {
4195 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4196 	uint32_t delay = info->delay_after_reset[index];
4197 	uint32_t val = info->reset_reg_val[index];
4198 	uint32_t reg = info->reset_reg[index];
4199 	uint32_t type, offset;
4200 	int ret;
4201 
4202 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4203 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4204 
4205 	switch (type) {
4206 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4207 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4208 		if (ret < 0) {
4209 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4210 				    val, offset);
4211 			return;
4212 		}
4213 		break;
4214 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4215 		offset = bnxt_map_reset_regs(bp, offset);
4216 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4217 		break;
4218 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4219 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4220 		break;
4221 	}
4222 	/* wait on a specific interval of time until core reset is complete */
4223 	if (delay)
4224 		rte_delay_ms(delay);
4225 }
4226 
4227 static void bnxt_dev_cleanup(struct bnxt *bp)
4228 {
4229 	bp->eth_dev->data->dev_link.link_status = 0;
4230 	bp->link_info->link_up = 0;
4231 	if (bp->eth_dev->data->dev_started)
4232 		bnxt_dev_stop(bp->eth_dev);
4233 
4234 	bnxt_uninit_resources(bp, true);
4235 }
4236 
4237 static int
4238 bnxt_check_fw_reset_done(struct bnxt *bp)
4239 {
4240 	int timeout = bp->fw_reset_max_msecs;
4241 	uint16_t val = 0;
4242 	int rc;
4243 
4244 	do {
4245 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4246 		if (rc < 0) {
4247 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4248 			return rc;
4249 		}
4250 		if (val != 0xffff)
4251 			break;
4252 		rte_delay_ms(1);
4253 	} while (timeout--);
4254 
4255 	if (val == 0xffff) {
4256 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4257 		return -1;
4258 	}
4259 
4260 	return 0;
4261 }
4262 
4263 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4264 {
4265 	struct rte_eth_dev *dev = bp->eth_dev;
4266 	struct rte_vlan_filter_conf *vfc;
4267 	int vidx, vbit, rc;
4268 	uint16_t vlan_id;
4269 
4270 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4271 		vfc = &dev->data->vlan_filter_conf;
4272 		vidx = vlan_id / 64;
4273 		vbit = vlan_id % 64;
4274 
4275 		/* Each bit corresponds to a VLAN id */
4276 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4277 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4278 			if (rc)
4279 				return rc;
4280 		}
4281 	}
4282 
4283 	return 0;
4284 }
4285 
4286 static int bnxt_restore_mac_filters(struct bnxt *bp)
4287 {
4288 	struct rte_eth_dev *dev = bp->eth_dev;
4289 	struct rte_eth_dev_info dev_info;
4290 	struct rte_ether_addr *addr;
4291 	uint64_t pool_mask;
4292 	uint32_t pool = 0;
4293 	uint32_t i;
4294 	int rc;
4295 
4296 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4297 		return 0;
4298 
4299 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4300 	if (rc)
4301 		return rc;
4302 
4303 	/* replay MAC address configuration */
4304 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4305 		addr = &dev->data->mac_addrs[i];
4306 
4307 		/* skip zero address */
4308 		if (rte_is_zero_ether_addr(addr))
4309 			continue;
4310 
4311 		pool = 0;
4312 		pool_mask = dev->data->mac_pool_sel[i];
4313 
4314 		do {
4315 			if (pool_mask & 1ULL) {
4316 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4317 				if (rc)
4318 					return rc;
4319 			}
4320 			pool_mask >>= 1;
4321 			pool++;
4322 		} while (pool_mask);
4323 	}
4324 
4325 	return 0;
4326 }
4327 
4328 static int bnxt_restore_mcast_mac_filters(struct bnxt *bp)
4329 {
4330 	int ret = 0;
4331 
4332 	ret = bnxt_dev_set_mc_addr_list_op(bp->eth_dev, bp->mcast_addr_list,
4333 					   bp->nb_mc_addr);
4334 	if (ret)
4335 		PMD_DRV_LOG(ERR, "Failed to restore multicast MAC addreeses\n");
4336 
4337 	return ret;
4338 }
4339 
4340 static int bnxt_restore_filters(struct bnxt *bp)
4341 {
4342 	struct rte_eth_dev *dev = bp->eth_dev;
4343 	int ret = 0;
4344 
4345 	if (dev->data->all_multicast) {
4346 		ret = bnxt_allmulticast_enable_op(dev);
4347 		if (ret)
4348 			return ret;
4349 	}
4350 	if (dev->data->promiscuous) {
4351 		ret = bnxt_promiscuous_enable_op(dev);
4352 		if (ret)
4353 			return ret;
4354 	}
4355 
4356 	ret = bnxt_restore_mac_filters(bp);
4357 	if (ret)
4358 		return ret;
4359 
4360 	/* if vlans are already programmed, this can fail with -EEXIST */
4361 	ret = bnxt_restore_vlan_filters(bp);
4362 	if (ret && ret != -EEXIST)
4363 		return ret;
4364 
4365 	ret = bnxt_restore_mcast_mac_filters(bp);
4366 	if (ret)
4367 		return ret;
4368 
4369 	return ret;
4370 }
4371 
4372 static int bnxt_check_fw_ready(struct bnxt *bp)
4373 {
4374 	int timeout = bp->fw_reset_max_msecs ? : BNXT_MAX_FW_RESET_TIMEOUT;
4375 	int rc = 0;
4376 
4377 	do {
4378 		rc = bnxt_hwrm_poll_ver_get(bp);
4379 		if (rc == 0)
4380 			break;
4381 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4382 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4383 	} while (rc && timeout > 0);
4384 
4385 	if (rc)
4386 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4387 
4388 	return rc;
4389 }
4390 
4391 static void bnxt_dev_recover(void *arg)
4392 {
4393 	struct bnxt *bp = arg;
4394 	int rc = 0;
4395 
4396 	pthread_mutex_lock(&bp->err_recovery_lock);
4397 
4398 	if (!bp->fw_reset_min_msecs) {
4399 		rc = bnxt_check_fw_reset_done(bp);
4400 		if (rc)
4401 			goto err;
4402 	}
4403 
4404 	/* Clear Error flag so that device re-init should happen */
4405 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4406 	PMD_DRV_LOG(INFO, "Port: %u Starting recovery...\n",
4407 		    bp->eth_dev->data->port_id);
4408 
4409 	rc = bnxt_check_fw_ready(bp);
4410 	if (rc)
4411 		goto err;
4412 
4413 	rc = bnxt_init_resources(bp, true);
4414 	if (rc) {
4415 		PMD_DRV_LOG(ERR,
4416 			    "Failed to initialize resources after reset\n");
4417 		goto err;
4418 	}
4419 	/* clear reset flag as the device is initialized now */
4420 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4421 
4422 	rc = bnxt_dev_start_op(bp->eth_dev);
4423 	if (rc) {
4424 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4425 		goto err_start;
4426 	}
4427 
4428 	rc = bnxt_restore_filters(bp);
4429 	if (rc)
4430 		goto err_start;
4431 
4432 	rte_eth_fp_ops[bp->eth_dev->data->port_id].rx_pkt_burst =
4433 		bp->eth_dev->rx_pkt_burst;
4434 	rte_eth_fp_ops[bp->eth_dev->data->port_id].tx_pkt_burst =
4435 		bp->eth_dev->tx_pkt_burst;
4436 	rte_mb();
4437 
4438 	PMD_DRV_LOG(INFO, "Port: %u Recovered from FW reset\n",
4439 		    bp->eth_dev->data->port_id);
4440 	pthread_mutex_unlock(&bp->err_recovery_lock);
4441 	rte_eth_dev_callback_process(bp->eth_dev,
4442 				     RTE_ETH_EVENT_RECOVERY_SUCCESS,
4443 				     NULL);
4444 	return;
4445 err_start:
4446 	bnxt_dev_stop(bp->eth_dev);
4447 err:
4448 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4449 	bnxt_uninit_resources(bp, false);
4450 	rte_eth_dev_callback_process(bp->eth_dev,
4451 				     RTE_ETH_EVENT_RECOVERY_FAILED,
4452 				     NULL);
4453 	if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4454 		rte_eth_dev_callback_process(bp->eth_dev,
4455 					     RTE_ETH_EVENT_INTR_RMV,
4456 					     NULL);
4457 	pthread_mutex_unlock(&bp->err_recovery_lock);
4458 	PMD_DRV_LOG(ERR, "Port %u: Failed to recover from FW reset\n",
4459 		    bp->eth_dev->data->port_id);
4460 }
4461 
4462 void bnxt_dev_reset_and_resume(void *arg)
4463 {
4464 	struct bnxt *bp = arg;
4465 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4466 	uint16_t val = 0;
4467 	int rc;
4468 
4469 	bnxt_dev_cleanup(bp);
4470 	PMD_DRV_LOG(INFO, "Port: %u Finished bnxt_dev_cleanup\n",
4471 		    bp->eth_dev->data->port_id);
4472 
4473 	bnxt_wait_for_device_shutdown(bp);
4474 
4475 	/* During some fatal firmware error conditions, the PCI config space
4476 	 * register 0x2e which normally contains the subsystem ID will become
4477 	 * 0xffff. This register will revert back to the normal value after
4478 	 * the chip has completed core reset. If we detect this condition,
4479 	 * we can poll this config register immediately for the value to revert.
4480 	 */
4481 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4482 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4483 		if (rc < 0) {
4484 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4485 			return;
4486 		}
4487 		if (val == 0xffff) {
4488 			bp->fw_reset_min_msecs = 0;
4489 			us = 1;
4490 		}
4491 	}
4492 
4493 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4494 	if (rc)
4495 		PMD_DRV_LOG(ERR, "Port %u: Error setting recovery alarm",
4496 			    bp->eth_dev->data->port_id);
4497 }
4498 
4499 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4500 {
4501 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4502 	uint32_t reg = info->status_regs[index];
4503 	uint32_t type, offset, val = 0;
4504 	int ret = 0;
4505 
4506 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4507 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4508 
4509 	switch (type) {
4510 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4511 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4512 		if (ret < 0)
4513 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4514 				    offset);
4515 		break;
4516 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4517 		offset = info->mapped_status_regs[index];
4518 		/* FALLTHROUGH */
4519 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4520 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4521 				       offset));
4522 		break;
4523 	}
4524 
4525 	return val;
4526 }
4527 
4528 static int bnxt_fw_reset_all(struct bnxt *bp)
4529 {
4530 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4531 	uint32_t i;
4532 	int rc = 0;
4533 
4534 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4535 		/* Reset through primary function driver */
4536 		for (i = 0; i < info->reg_array_cnt; i++)
4537 			bnxt_write_fw_reset_reg(bp, i);
4538 		/* Wait for time specified by FW after triggering reset */
4539 		rte_delay_ms(info->primary_func_wait_period_after_reset);
4540 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4541 		/* Reset with the help of Kong processor */
4542 		rc = bnxt_hwrm_fw_reset(bp);
4543 		if (rc)
4544 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4545 	}
4546 
4547 	return rc;
4548 }
4549 
4550 static void bnxt_fw_reset_cb(void *arg)
4551 {
4552 	struct bnxt *bp = arg;
4553 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4554 	int rc = 0;
4555 
4556 	/* Only Primary function can do FW reset */
4557 	if (bnxt_is_primary_func(bp) &&
4558 	    bnxt_is_recovery_enabled(bp)) {
4559 		rc = bnxt_fw_reset_all(bp);
4560 		if (rc) {
4561 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4562 			return;
4563 		}
4564 	}
4565 
4566 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4567 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4568 	 * (including MASTER FUNC). After receiving this Async, all the active
4569 	 * drivers should treat this case as FW initiated recovery
4570 	 */
4571 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4572 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4573 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4574 
4575 		/* To recover from error */
4576 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4577 				  (void *)bp);
4578 	}
4579 }
4580 
4581 /* Driver should poll FW heartbeat, reset_counter with the frequency
4582  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4583  * When the driver detects heartbeat stop or change in reset_counter,
4584  * it has to trigger a reset to recover from the error condition.
4585  * A “primary function” is the function who will have the privilege to
4586  * initiate the chimp reset. The primary function will be elected by the
4587  * firmware and will be notified through async message.
4588  */
4589 static void bnxt_check_fw_health(void *arg)
4590 {
4591 	struct bnxt *bp = arg;
4592 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4593 	uint32_t val = 0, wait_msec;
4594 
4595 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4596 	    is_bnxt_in_error(bp))
4597 		return;
4598 
4599 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4600 	if (val == info->last_heart_beat)
4601 		goto reset;
4602 
4603 	info->last_heart_beat = val;
4604 
4605 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4606 	if (val != info->last_reset_counter)
4607 		goto reset;
4608 
4609 	info->last_reset_counter = val;
4610 
4611 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4612 			  bnxt_check_fw_health, (void *)bp);
4613 
4614 	return;
4615 reset:
4616 	/* Stop DMA to/from device */
4617 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4618 	bp->flags |= BNXT_FLAG_FW_RESET;
4619 
4620 	bnxt_stop_rxtx(bp->eth_dev);
4621 
4622 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4623 
4624 	rte_eth_dev_callback_process(bp->eth_dev,
4625 				     RTE_ETH_EVENT_ERR_RECOVERING,
4626 				     NULL);
4627 
4628 	if (bnxt_is_primary_func(bp))
4629 		wait_msec = info->primary_func_wait_period;
4630 	else
4631 		wait_msec = info->normal_func_wait_period;
4632 
4633 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4634 			  bnxt_fw_reset_cb, (void *)bp);
4635 }
4636 
4637 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4638 {
4639 	uint32_t polling_freq;
4640 
4641 	pthread_mutex_lock(&bp->health_check_lock);
4642 
4643 	if (!bnxt_is_recovery_enabled(bp))
4644 		goto done;
4645 
4646 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4647 		goto done;
4648 
4649 	polling_freq = bp->recovery_info->driver_polling_freq;
4650 
4651 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4652 			  bnxt_check_fw_health, (void *)bp);
4653 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4654 
4655 done:
4656 	pthread_mutex_unlock(&bp->health_check_lock);
4657 }
4658 
4659 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4660 {
4661 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4662 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4663 }
4664 
4665 static bool bnxt_vf_pciid(uint16_t device_id)
4666 {
4667 	switch (device_id) {
4668 	case BROADCOM_DEV_ID_57304_VF:
4669 	case BROADCOM_DEV_ID_57406_VF:
4670 	case BROADCOM_DEV_ID_5731X_VF:
4671 	case BROADCOM_DEV_ID_5741X_VF:
4672 	case BROADCOM_DEV_ID_57414_VF:
4673 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4674 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4675 	case BROADCOM_DEV_ID_58802_VF:
4676 	case BROADCOM_DEV_ID_57500_VF1:
4677 	case BROADCOM_DEV_ID_57500_VF2:
4678 	case BROADCOM_DEV_ID_58818_VF:
4679 		/* FALLTHROUGH */
4680 		return true;
4681 	default:
4682 		return false;
4683 	}
4684 }
4685 
4686 /* Phase 5 device */
4687 static bool bnxt_p5_device(uint16_t device_id)
4688 {
4689 	switch (device_id) {
4690 	case BROADCOM_DEV_ID_57508:
4691 	case BROADCOM_DEV_ID_57504:
4692 	case BROADCOM_DEV_ID_57502:
4693 	case BROADCOM_DEV_ID_57508_MF1:
4694 	case BROADCOM_DEV_ID_57504_MF1:
4695 	case BROADCOM_DEV_ID_57502_MF1:
4696 	case BROADCOM_DEV_ID_57508_MF2:
4697 	case BROADCOM_DEV_ID_57504_MF2:
4698 	case BROADCOM_DEV_ID_57502_MF2:
4699 	case BROADCOM_DEV_ID_57500_VF1:
4700 	case BROADCOM_DEV_ID_57500_VF2:
4701 	case BROADCOM_DEV_ID_58812:
4702 	case BROADCOM_DEV_ID_58814:
4703 	case BROADCOM_DEV_ID_58818:
4704 	case BROADCOM_DEV_ID_58818_VF:
4705 		/* FALLTHROUGH */
4706 		return true;
4707 	default:
4708 		return false;
4709 	}
4710 }
4711 
4712 bool bnxt_stratus_device(struct bnxt *bp)
4713 {
4714 	uint16_t device_id = bp->pdev->id.device_id;
4715 
4716 	switch (device_id) {
4717 	case BROADCOM_DEV_ID_STRATUS_NIC:
4718 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4719 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4720 		/* FALLTHROUGH */
4721 		return true;
4722 	default:
4723 		return false;
4724 	}
4725 }
4726 
4727 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4728 {
4729 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4730 	struct bnxt *bp = eth_dev->data->dev_private;
4731 
4732 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4733 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4734 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4735 	if (!bp->bar0 || !bp->doorbell_base) {
4736 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4737 		return -ENODEV;
4738 	}
4739 
4740 	bp->eth_dev = eth_dev;
4741 	bp->pdev = pci_dev;
4742 
4743 	return 0;
4744 }
4745 
4746 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4747 				  struct bnxt_ctx_pg_info *ctx_pg,
4748 				  uint32_t mem_size,
4749 				  const char *suffix,
4750 				  uint16_t idx)
4751 {
4752 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4753 	const struct rte_memzone *mz = NULL;
4754 	char mz_name[RTE_MEMZONE_NAMESIZE];
4755 	rte_iova_t mz_phys_addr;
4756 	uint64_t valid_bits = 0;
4757 	uint32_t sz;
4758 	int i;
4759 
4760 	if (!mem_size)
4761 		return 0;
4762 
4763 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4764 			 BNXT_PAGE_SIZE;
4765 	rmem->page_size = BNXT_PAGE_SIZE;
4766 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4767 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4768 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4769 
4770 	valid_bits = PTU_PTE_VALID;
4771 
4772 	if (rmem->nr_pages > 1) {
4773 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4774 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4775 			 suffix, idx, bp->eth_dev->data->port_id);
4776 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4777 		mz = rte_memzone_lookup(mz_name);
4778 		if (!mz) {
4779 			mz = rte_memzone_reserve_aligned(mz_name,
4780 						rmem->nr_pages * 8,
4781 						bp->eth_dev->device->numa_node,
4782 						RTE_MEMZONE_2MB |
4783 						RTE_MEMZONE_SIZE_HINT_ONLY |
4784 						RTE_MEMZONE_IOVA_CONTIG,
4785 						BNXT_PAGE_SIZE);
4786 			if (mz == NULL)
4787 				return -ENOMEM;
4788 		}
4789 
4790 		memset(mz->addr, 0, mz->len);
4791 		mz_phys_addr = mz->iova;
4792 
4793 		rmem->pg_tbl = mz->addr;
4794 		rmem->pg_tbl_map = mz_phys_addr;
4795 		rmem->pg_tbl_mz = mz;
4796 	}
4797 
4798 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4799 		 suffix, idx, bp->eth_dev->data->port_id);
4800 	mz = rte_memzone_lookup(mz_name);
4801 	if (!mz) {
4802 		mz = rte_memzone_reserve_aligned(mz_name,
4803 						 mem_size,
4804 						 bp->eth_dev->device->numa_node,
4805 						 RTE_MEMZONE_1GB |
4806 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4807 						 RTE_MEMZONE_IOVA_CONTIG,
4808 						 BNXT_PAGE_SIZE);
4809 		if (mz == NULL)
4810 			return -ENOMEM;
4811 	}
4812 
4813 	memset(mz->addr, 0, mz->len);
4814 	mz_phys_addr = mz->iova;
4815 
4816 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4817 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4818 		rmem->dma_arr[i] = mz_phys_addr + sz;
4819 
4820 		if (rmem->nr_pages > 1) {
4821 			if (i == rmem->nr_pages - 2 &&
4822 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4823 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4824 			else if (i == rmem->nr_pages - 1 &&
4825 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4826 				valid_bits |= PTU_PTE_LAST;
4827 
4828 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4829 							   valid_bits);
4830 		}
4831 	}
4832 
4833 	rmem->mz = mz;
4834 	if (rmem->vmem_size)
4835 		rmem->vmem = (void **)mz->addr;
4836 	rmem->dma_arr[0] = mz_phys_addr;
4837 	return 0;
4838 }
4839 
4840 static void bnxt_free_ctx_mem(struct bnxt *bp)
4841 {
4842 	int i;
4843 
4844 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4845 		return;
4846 
4847 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4848 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4849 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4850 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4851 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4852 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4853 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4854 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4855 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4856 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4857 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4858 
4859 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4860 		if (bp->ctx->tqm_mem[i])
4861 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4862 	}
4863 
4864 	rte_free(bp->ctx);
4865 	bp->ctx = NULL;
4866 }
4867 
4868 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4869 
4870 #define clamp_t(type, _x, min, max) RTE_MIN_T(RTE_MAX_T(_x, min, type), max, type)
4871 
4872 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4873 {
4874 	struct bnxt_ctx_pg_info *ctx_pg;
4875 	struct bnxt_ctx_mem_info *ctx;
4876 	uint32_t mem_size, ena, entries;
4877 	uint32_t entries_sp, min;
4878 	int i, rc;
4879 
4880 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4881 	if (rc) {
4882 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4883 		return rc;
4884 	}
4885 	ctx = bp->ctx;
4886 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4887 		return 0;
4888 
4889 	ctx_pg = &ctx->qp_mem;
4890 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4891 	if (ctx->qp_entry_size) {
4892 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4893 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4894 		if (rc)
4895 			return rc;
4896 	}
4897 
4898 	ctx_pg = &ctx->srq_mem;
4899 	ctx_pg->entries = ctx->srq_max_l2_entries;
4900 	if (ctx->srq_entry_size) {
4901 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4902 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4903 		if (rc)
4904 			return rc;
4905 	}
4906 
4907 	ctx_pg = &ctx->cq_mem;
4908 	ctx_pg->entries = ctx->cq_max_l2_entries;
4909 	if (ctx->cq_entry_size) {
4910 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4911 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4912 		if (rc)
4913 			return rc;
4914 	}
4915 
4916 	ctx_pg = &ctx->vnic_mem;
4917 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4918 		ctx->vnic_max_ring_table_entries;
4919 	if (ctx->vnic_entry_size) {
4920 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4921 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4922 		if (rc)
4923 			return rc;
4924 	}
4925 
4926 	ctx_pg = &ctx->stat_mem;
4927 	ctx_pg->entries = ctx->stat_max_entries;
4928 	if (ctx->stat_entry_size) {
4929 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4930 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4931 		if (rc)
4932 			return rc;
4933 	}
4934 
4935 	min = ctx->tqm_min_entries_per_ring;
4936 
4937 	entries_sp = ctx->qp_max_l2_entries +
4938 		     ctx->vnic_max_vnic_entries +
4939 		     2 * ctx->qp_min_qp1_entries + min;
4940 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4941 
4942 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4943 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4944 	entries = clamp_t(uint32_t, entries, min,
4945 			  ctx->tqm_max_entries_per_ring);
4946 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4947 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4948 		 * i > 8 is other ext rings.
4949 		 */
4950 		ctx_pg = ctx->tqm_mem[i];
4951 		ctx_pg->entries = i ? entries : entries_sp;
4952 		if (ctx->tqm_entry_size) {
4953 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4954 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4955 						    "tqm_mem", i);
4956 			if (rc)
4957 				return rc;
4958 		}
4959 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4960 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4961 		else
4962 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4963 	}
4964 
4965 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4966 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4967 	if (rc)
4968 		PMD_DRV_LOG(ERR,
4969 			    "Failed to configure context mem: rc = %d\n", rc);
4970 	else
4971 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4972 
4973 	return rc;
4974 }
4975 
4976 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4977 {
4978 	struct rte_pci_device *pci_dev = bp->pdev;
4979 	char mz_name[RTE_MEMZONE_NAMESIZE];
4980 	const struct rte_memzone *mz = NULL;
4981 	uint32_t total_alloc_len;
4982 	rte_iova_t mz_phys_addr;
4983 
4984 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4985 		return 0;
4986 
4987 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4988 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4989 		 pci_dev->addr.bus, pci_dev->addr.devid,
4990 		 pci_dev->addr.function, "rx_port_stats");
4991 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4992 	mz = rte_memzone_lookup(mz_name);
4993 	total_alloc_len =
4994 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4995 				       sizeof(struct rx_port_stats_ext) + 512);
4996 	if (!mz) {
4997 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4998 					 SOCKET_ID_ANY,
4999 					 RTE_MEMZONE_2MB |
5000 					 RTE_MEMZONE_SIZE_HINT_ONLY |
5001 					 RTE_MEMZONE_IOVA_CONTIG);
5002 		if (mz == NULL)
5003 			return -ENOMEM;
5004 	}
5005 	memset(mz->addr, 0, mz->len);
5006 	mz_phys_addr = mz->iova;
5007 
5008 	bp->rx_mem_zone = (const void *)mz;
5009 	bp->hw_rx_port_stats = mz->addr;
5010 	bp->hw_rx_port_stats_map = mz_phys_addr;
5011 
5012 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
5013 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
5014 		 pci_dev->addr.bus, pci_dev->addr.devid,
5015 		 pci_dev->addr.function, "tx_port_stats");
5016 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
5017 	mz = rte_memzone_lookup(mz_name);
5018 	total_alloc_len =
5019 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
5020 				       sizeof(struct tx_port_stats_ext) + 512);
5021 	if (!mz) {
5022 		mz = rte_memzone_reserve(mz_name,
5023 					 total_alloc_len,
5024 					 SOCKET_ID_ANY,
5025 					 RTE_MEMZONE_2MB |
5026 					 RTE_MEMZONE_SIZE_HINT_ONLY |
5027 					 RTE_MEMZONE_IOVA_CONTIG);
5028 		if (mz == NULL)
5029 			return -ENOMEM;
5030 	}
5031 	memset(mz->addr, 0, mz->len);
5032 	mz_phys_addr = mz->iova;
5033 
5034 	bp->tx_mem_zone = (const void *)mz;
5035 	bp->hw_tx_port_stats = mz->addr;
5036 	bp->hw_tx_port_stats_map = mz_phys_addr;
5037 	bp->flags |= BNXT_FLAG_PORT_STATS;
5038 
5039 	/* Display extended statistics if FW supports it */
5040 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
5041 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
5042 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
5043 		return 0;
5044 
5045 	bp->hw_rx_port_stats_ext = (void *)
5046 		((uint8_t *)bp->hw_rx_port_stats +
5047 		 sizeof(struct rx_port_stats));
5048 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
5049 		sizeof(struct rx_port_stats);
5050 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
5051 
5052 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
5053 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
5054 		bp->hw_tx_port_stats_ext = (void *)
5055 			((uint8_t *)bp->hw_tx_port_stats +
5056 			 sizeof(struct tx_port_stats));
5057 		bp->hw_tx_port_stats_ext_map =
5058 			bp->hw_tx_port_stats_map +
5059 			sizeof(struct tx_port_stats);
5060 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
5061 	}
5062 
5063 	return 0;
5064 }
5065 
5066 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
5067 {
5068 	struct bnxt *bp = eth_dev->data->dev_private;
5069 	size_t max_mac_addr = RTE_MIN(bp->max_l2_ctx, RTE_ETH_NUM_RECEIVE_MAC_ADDR);
5070 	int rc = 0;
5071 
5072 	if (bp->max_l2_ctx > RTE_ETH_NUM_RECEIVE_MAC_ADDR)
5073 		PMD_DRV_LOG(INFO, "Max number of MAC addrs supported is %d, but will be limited to %d\n",
5074 			    bp->max_l2_ctx, RTE_ETH_NUM_RECEIVE_MAC_ADDR);
5075 
5076 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
5077 					       RTE_ETHER_ADDR_LEN * max_mac_addr,
5078 					       0);
5079 	if (eth_dev->data->mac_addrs == NULL) {
5080 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
5081 		return -ENOMEM;
5082 	}
5083 
5084 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
5085 		if (BNXT_PF(bp))
5086 			return -EINVAL;
5087 
5088 		/* Generate a random MAC address, if none was assigned by PF */
5089 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
5090 		bnxt_eth_hw_addr_random(bp->mac_addr);
5091 		PMD_DRV_LOG(INFO,
5092 			    "Assign random MAC:" RTE_ETHER_ADDR_PRT_FMT "\n",
5093 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
5094 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
5095 
5096 		rc = bnxt_hwrm_set_mac(bp);
5097 		if (rc)
5098 			return rc;
5099 	}
5100 
5101 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5102 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5103 
5104 	/*
5105 	 *  Allocate memory to hold multicast mac addresses added.
5106 	 *  Used to restore them during reset recovery
5107 	 */
5108 	bp->mcast_addr_list = rte_zmalloc("bnxt_mcast_addr_tbl",
5109 					  sizeof(struct rte_ether_addr) *
5110 					  BNXT_MAX_MC_ADDRS, 0);
5111 	if (bp->mcast_addr_list == NULL) {
5112 		PMD_DRV_LOG(ERR, "Failed to allocate multicast addr table\n");
5113 		return -ENOMEM;
5114 	}
5115 	bp->mc_list_dma_addr = rte_malloc_virt2iova(bp->mcast_addr_list);
5116 	if (bp->mc_list_dma_addr == RTE_BAD_IOVA) {
5117 		PMD_DRV_LOG(ERR, "Fail to map mcast_addr_list to physical memory\n");
5118 		return -ENOMEM;
5119 	}
5120 
5121 	return rc;
5122 }
5123 
5124 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5125 {
5126 	int rc = 0;
5127 
5128 	/* MAC is already configured in FW */
5129 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5130 		return 0;
5131 
5132 	/* Restore the old MAC configured */
5133 	rc = bnxt_hwrm_set_mac(bp);
5134 	if (rc)
5135 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5136 
5137 	return rc;
5138 }
5139 
5140 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5141 {
5142 	if (!BNXT_PF(bp))
5143 		return;
5144 
5145 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5146 
5147 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5148 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5149 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5150 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5151 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5152 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5153 }
5154 
5155 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5156 {
5157 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5158 
5159 	if (info) {
5160 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5161 			memset(info, 0, sizeof(*info));
5162 		return;
5163 	}
5164 
5165 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5166 		return;
5167 
5168 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5169 			   sizeof(*info), 0);
5170 	if (!info)
5171 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5172 
5173 	bp->recovery_info = info;
5174 }
5175 
5176 static void bnxt_check_fw_status(struct bnxt *bp)
5177 {
5178 	uint32_t fw_status;
5179 
5180 	if (!(bp->recovery_info &&
5181 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5182 		return;
5183 
5184 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5185 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5186 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5187 			    fw_status);
5188 }
5189 
5190 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5191 {
5192 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5193 	uint32_t status_loc;
5194 	uint32_t sig_ver;
5195 
5196 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5197 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5198 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5199 				   BNXT_GRCP_WINDOW_2_BASE +
5200 				   offsetof(struct hcomm_status,
5201 					    sig_ver)));
5202 	/* If the signature is absent, then FW does not support this feature */
5203 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5204 	    HCOMM_STATUS_SIGNATURE_VAL)
5205 		return 0;
5206 
5207 	if (!info) {
5208 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5209 				   sizeof(*info), 0);
5210 		if (!info)
5211 			return -ENOMEM;
5212 		bp->recovery_info = info;
5213 	} else {
5214 		memset(info, 0, sizeof(*info));
5215 	}
5216 
5217 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5218 				      BNXT_GRCP_WINDOW_2_BASE +
5219 				      offsetof(struct hcomm_status,
5220 					       fw_status_loc)));
5221 
5222 	/* Only pre-map the FW health status GRC register */
5223 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5224 		return 0;
5225 
5226 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5227 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5228 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5229 
5230 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5231 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5232 
5233 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5234 
5235 	return 0;
5236 }
5237 
5238 /* This function gets the FW version along with the
5239  * capabilities(MAX and current) of the function, vnic,
5240  * error recovery, phy and other chip related info
5241  */
5242 static int bnxt_get_config(struct bnxt *bp)
5243 {
5244 	uint16_t mtu;
5245 	int rc = 0;
5246 
5247 	bp->fw_cap = 0;
5248 
5249 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5250 	if (rc)
5251 		return rc;
5252 
5253 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5254 	if (rc) {
5255 		bnxt_check_fw_status(bp);
5256 		return rc;
5257 	}
5258 
5259 	rc = bnxt_hwrm_func_reset(bp);
5260 	if (rc)
5261 		return -EIO;
5262 
5263 	rc = bnxt_hwrm_vnic_qcaps(bp);
5264 	if (rc)
5265 		return rc;
5266 
5267 	rc = bnxt_hwrm_queue_qportcfg(bp);
5268 	if (rc)
5269 		return rc;
5270 
5271 	/* Get the MAX capabilities for this function.
5272 	 * This function also allocates context memory for TQM rings and
5273 	 * informs the firmware about this allocated backing store memory.
5274 	 */
5275 	rc = bnxt_hwrm_func_qcaps(bp);
5276 	if (rc)
5277 		return rc;
5278 
5279 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5280 	if (rc)
5281 		return rc;
5282 
5283 	bnxt_hwrm_port_mac_qcfg(bp);
5284 
5285 	bnxt_hwrm_parent_pf_qcfg(bp);
5286 
5287 	bnxt_hwrm_port_phy_qcaps(bp);
5288 
5289 	bnxt_alloc_error_recovery_info(bp);
5290 	/* Get the adapter error recovery support info */
5291 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5292 	if (rc)
5293 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5294 
5295 	bnxt_hwrm_port_led_qcaps(bp);
5296 
5297 	return 0;
5298 }
5299 
5300 static int
5301 bnxt_init_locks(struct bnxt *bp)
5302 {
5303 	int err;
5304 
5305 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5306 	if (err) {
5307 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5308 		return err;
5309 	}
5310 
5311 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5312 	if (err) {
5313 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5314 		return err;
5315 	}
5316 
5317 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5318 	if (err) {
5319 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5320 		return err;
5321 	}
5322 
5323 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5324 	if (err)
5325 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5326 
5327 	return err;
5328 }
5329 
5330 /* This should be called after we have queried trusted VF cap */
5331 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5332 {
5333 	int rc = 0;
5334 
5335 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5336 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5337 		if (rc)
5338 			PMD_DRV_LOG(ERR,
5339 				    "Failed to alloc switch domain: %d\n", rc);
5340 		else
5341 			PMD_DRV_LOG(INFO,
5342 				    "Switch domain allocated %d\n",
5343 				    bp->switch_domain_id);
5344 	}
5345 
5346 	return rc;
5347 }
5348 
5349 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5350 {
5351 	int rc = 0;
5352 
5353 	if (reconfig_dev) {
5354 		rc = bnxt_get_config(bp);
5355 		if (rc)
5356 			return rc;
5357 	}
5358 
5359 	rc = bnxt_alloc_switch_domain(bp);
5360 	if (rc)
5361 		return rc;
5362 
5363 	if (!reconfig_dev) {
5364 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5365 		if (rc)
5366 			return rc;
5367 	} else {
5368 		rc = bnxt_restore_dflt_mac(bp);
5369 		if (rc)
5370 			return rc;
5371 	}
5372 
5373 	bnxt_config_vf_req_fwd(bp);
5374 
5375 	rc = bnxt_hwrm_func_driver_register(bp);
5376 	if (rc) {
5377 		PMD_DRV_LOG(ERR, "Failed to register driver");
5378 		return -EBUSY;
5379 	}
5380 
5381 	if (BNXT_PF(bp)) {
5382 		if (bp->pdev->max_vfs) {
5383 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5384 			if (rc) {
5385 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5386 				return rc;
5387 			}
5388 		} else {
5389 			rc = bnxt_hwrm_allocate_pf_only(bp);
5390 			if (rc) {
5391 				PMD_DRV_LOG(ERR,
5392 					    "Failed to allocate PF resources");
5393 				return rc;
5394 			}
5395 		}
5396 	}
5397 
5398 	if (!reconfig_dev) {
5399 		bp->rss_conf.rss_key = rte_zmalloc("bnxt_rss_key",
5400 						   HW_HASH_KEY_SIZE, 0);
5401 		if (bp->rss_conf.rss_key == NULL) {
5402 			PMD_DRV_LOG(ERR, "port %u cannot allocate RSS hash key memory",
5403 				    bp->eth_dev->data->port_id);
5404 			return -ENOMEM;
5405 		}
5406 	}
5407 
5408 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5409 	if (rc)
5410 		return rc;
5411 
5412 	rc = bnxt_setup_int(bp);
5413 	if (rc)
5414 		return rc;
5415 
5416 	rc = bnxt_request_int(bp);
5417 	if (rc)
5418 		return rc;
5419 
5420 	rc = bnxt_init_ctx_mem(bp);
5421 	if (rc) {
5422 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5423 		return rc;
5424 	}
5425 
5426 	return 0;
5427 }
5428 
5429 static int
5430 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5431 			     const char *value, void *opaque_arg)
5432 {
5433 	struct bnxt *bp = opaque_arg;
5434 	unsigned long flow_xstat;
5435 	char *end = NULL;
5436 
5437 	if (!value || !opaque_arg) {
5438 		PMD_DRV_LOG(ERR,
5439 			    "Invalid parameter passed to flow_xstat devarg.\n");
5440 		return -EINVAL;
5441 	}
5442 
5443 	flow_xstat = strtoul(value, &end, 10);
5444 	if (end == NULL || *end != '\0' ||
5445 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5446 		PMD_DRV_LOG(ERR,
5447 			    "Invalid parameter passed to flow_xstat devarg.\n");
5448 		return -EINVAL;
5449 	}
5450 
5451 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5452 		PMD_DRV_LOG(ERR,
5453 			    "Invalid value passed to flow_xstat devarg.\n");
5454 		return -EINVAL;
5455 	}
5456 
5457 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5458 	if (BNXT_FLOW_XSTATS_EN(bp))
5459 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5460 
5461 	return 0;
5462 }
5463 
5464 static int
5465 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5466 					const char *value, void *opaque_arg)
5467 {
5468 	struct bnxt *bp = opaque_arg;
5469 	unsigned long max_num_kflows;
5470 	char *end = NULL;
5471 
5472 	if (!value || !opaque_arg) {
5473 		PMD_DRV_LOG(ERR,
5474 			"Invalid parameter passed to max_num_kflows devarg.\n");
5475 		return -EINVAL;
5476 	}
5477 
5478 	max_num_kflows = strtoul(value, &end, 10);
5479 	if (end == NULL || *end != '\0' ||
5480 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5481 		PMD_DRV_LOG(ERR,
5482 			"Invalid parameter passed to max_num_kflows devarg.\n");
5483 		return -EINVAL;
5484 	}
5485 
5486 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5487 		PMD_DRV_LOG(ERR,
5488 			"Invalid value passed to max_num_kflows devarg.\n");
5489 		return -EINVAL;
5490 	}
5491 
5492 	bp->max_num_kflows = max_num_kflows;
5493 	if (bp->max_num_kflows)
5494 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5495 				max_num_kflows);
5496 
5497 	return 0;
5498 }
5499 
5500 static int
5501 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5502 				 const char *value, void *opaque_arg)
5503 {
5504 	struct bnxt *bp = opaque_arg;
5505 	unsigned long app_id;
5506 	char *end = NULL;
5507 
5508 	if (!value || !opaque_arg) {
5509 		PMD_DRV_LOG(ERR,
5510 			    "Invalid parameter passed to app-id "
5511 			    "devargs.\n");
5512 		return -EINVAL;
5513 	}
5514 
5515 	app_id = strtoul(value, &end, 10);
5516 	if (end == NULL || *end != '\0' ||
5517 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5518 		PMD_DRV_LOG(ERR,
5519 			    "Invalid parameter passed to app_id "
5520 			    "devargs.\n");
5521 		return -EINVAL;
5522 	}
5523 
5524 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5525 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5526 			    (uint16_t)app_id);
5527 		return -EINVAL;
5528 	}
5529 
5530 	bp->app_id = app_id;
5531 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5532 
5533 	return 0;
5534 }
5535 
5536 static int
5537 bnxt_parse_devarg_ieee_1588(__rte_unused const char *key,
5538 			    const char *value, void *opaque_arg)
5539 {
5540 	struct bnxt *bp = opaque_arg;
5541 	unsigned long ieee_1588;
5542 	char *end = NULL;
5543 
5544 	if (!value || !opaque_arg) {
5545 		PMD_DRV_LOG(ERR,
5546 			    "Invalid parameter passed to ieee-1588 "
5547 			    "devargs.\n");
5548 		return -EINVAL;
5549 	}
5550 
5551 	ieee_1588 = strtoul(value, &end, 10);
5552 	if (end == NULL || *end != '\0' ||
5553 	    (ieee_1588 == ULONG_MAX && errno == ERANGE)) {
5554 		PMD_DRV_LOG(ERR,
5555 			    "Invalid parameter passed to ieee_1588 "
5556 			    "devargs.\n");
5557 		return -EINVAL;
5558 	}
5559 
5560 	if (BNXT_DEVARG_IEEE_1588_INVALID(ieee_1588)) {
5561 		PMD_DRV_LOG(ERR, "Invalid ieee-1588(%d) devargs.\n",
5562 			    (uint16_t)ieee_1588);
5563 		return -EINVAL;
5564 	}
5565 
5566 	bp->ieee_1588 = ieee_1588;
5567 	PMD_DRV_LOG(INFO, "ieee-1588=%d feature enabled.\n", (uint16_t)ieee_1588);
5568 
5569 	return 0;
5570 }
5571 
5572 static int
5573 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5574 			    const char *value, void *opaque_arg)
5575 {
5576 	struct bnxt_representor *vfr_bp = opaque_arg;
5577 	unsigned long rep_is_pf;
5578 	char *end = NULL;
5579 
5580 	if (!value || !opaque_arg) {
5581 		PMD_DRV_LOG(ERR,
5582 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5583 		return -EINVAL;
5584 	}
5585 
5586 	rep_is_pf = strtoul(value, &end, 10);
5587 	if (end == NULL || *end != '\0' ||
5588 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5589 		PMD_DRV_LOG(ERR,
5590 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5591 		return -EINVAL;
5592 	}
5593 
5594 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5595 		PMD_DRV_LOG(ERR,
5596 			    "Invalid value passed to rep_is_pf devargs.\n");
5597 		return -EINVAL;
5598 	}
5599 
5600 	vfr_bp->flags |= rep_is_pf;
5601 	if (BNXT_REP_PF(vfr_bp))
5602 		PMD_DRV_LOG(INFO, "PF representor\n");
5603 	else
5604 		PMD_DRV_LOG(INFO, "VF representor\n");
5605 
5606 	return 0;
5607 }
5608 
5609 static int
5610 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5611 			       const char *value, void *opaque_arg)
5612 {
5613 	struct bnxt_representor *vfr_bp = opaque_arg;
5614 	unsigned long rep_based_pf;
5615 	char *end = NULL;
5616 
5617 	if (!value || !opaque_arg) {
5618 		PMD_DRV_LOG(ERR,
5619 			    "Invalid parameter passed to rep_based_pf "
5620 			    "devargs.\n");
5621 		return -EINVAL;
5622 	}
5623 
5624 	rep_based_pf = strtoul(value, &end, 10);
5625 	if (end == NULL || *end != '\0' ||
5626 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5627 		PMD_DRV_LOG(ERR,
5628 			    "Invalid parameter passed to rep_based_pf "
5629 			    "devargs.\n");
5630 		return -EINVAL;
5631 	}
5632 
5633 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5634 		PMD_DRV_LOG(ERR,
5635 			    "Invalid value passed to rep_based_pf devargs.\n");
5636 		return -EINVAL;
5637 	}
5638 
5639 	vfr_bp->rep_based_pf = rep_based_pf;
5640 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5641 
5642 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5643 
5644 	return 0;
5645 }
5646 
5647 static int
5648 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5649 			    const char *value, void *opaque_arg)
5650 {
5651 	struct bnxt_representor *vfr_bp = opaque_arg;
5652 	unsigned long rep_q_r2f;
5653 	char *end = NULL;
5654 
5655 	if (!value || !opaque_arg) {
5656 		PMD_DRV_LOG(ERR,
5657 			    "Invalid parameter passed to rep_q_r2f "
5658 			    "devargs.\n");
5659 		return -EINVAL;
5660 	}
5661 
5662 	rep_q_r2f = strtoul(value, &end, 10);
5663 	if (end == NULL || *end != '\0' ||
5664 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5665 		PMD_DRV_LOG(ERR,
5666 			    "Invalid parameter passed to rep_q_r2f "
5667 			    "devargs.\n");
5668 		return -EINVAL;
5669 	}
5670 
5671 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5672 		PMD_DRV_LOG(ERR,
5673 			    "Invalid value passed to rep_q_r2f devargs.\n");
5674 		return -EINVAL;
5675 	}
5676 
5677 	vfr_bp->rep_q_r2f = rep_q_r2f;
5678 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5679 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5680 
5681 	return 0;
5682 }
5683 
5684 static int
5685 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5686 			    const char *value, void *opaque_arg)
5687 {
5688 	struct bnxt_representor *vfr_bp = opaque_arg;
5689 	unsigned long rep_q_f2r;
5690 	char *end = NULL;
5691 
5692 	if (!value || !opaque_arg) {
5693 		PMD_DRV_LOG(ERR,
5694 			    "Invalid parameter passed to rep_q_f2r "
5695 			    "devargs.\n");
5696 		return -EINVAL;
5697 	}
5698 
5699 	rep_q_f2r = strtoul(value, &end, 10);
5700 	if (end == NULL || *end != '\0' ||
5701 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5702 		PMD_DRV_LOG(ERR,
5703 			    "Invalid parameter passed to rep_q_f2r "
5704 			    "devargs.\n");
5705 		return -EINVAL;
5706 	}
5707 
5708 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5709 		PMD_DRV_LOG(ERR,
5710 			    "Invalid value passed to rep_q_f2r devargs.\n");
5711 		return -EINVAL;
5712 	}
5713 
5714 	vfr_bp->rep_q_f2r = rep_q_f2r;
5715 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5716 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5717 
5718 	return 0;
5719 }
5720 
5721 static int
5722 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5723 			     const char *value, void *opaque_arg)
5724 {
5725 	struct bnxt_representor *vfr_bp = opaque_arg;
5726 	unsigned long rep_fc_r2f;
5727 	char *end = NULL;
5728 
5729 	if (!value || !opaque_arg) {
5730 		PMD_DRV_LOG(ERR,
5731 			    "Invalid parameter passed to rep_fc_r2f "
5732 			    "devargs.\n");
5733 		return -EINVAL;
5734 	}
5735 
5736 	rep_fc_r2f = strtoul(value, &end, 10);
5737 	if (end == NULL || *end != '\0' ||
5738 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5739 		PMD_DRV_LOG(ERR,
5740 			    "Invalid parameter passed to rep_fc_r2f "
5741 			    "devargs.\n");
5742 		return -EINVAL;
5743 	}
5744 
5745 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5746 		PMD_DRV_LOG(ERR,
5747 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5748 		return -EINVAL;
5749 	}
5750 
5751 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5752 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5753 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5754 
5755 	return 0;
5756 }
5757 
5758 static int
5759 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5760 			     const char *value, void *opaque_arg)
5761 {
5762 	struct bnxt_representor *vfr_bp = opaque_arg;
5763 	unsigned long rep_fc_f2r;
5764 	char *end = NULL;
5765 
5766 	if (!value || !opaque_arg) {
5767 		PMD_DRV_LOG(ERR,
5768 			    "Invalid parameter passed to rep_fc_f2r "
5769 			    "devargs.\n");
5770 		return -EINVAL;
5771 	}
5772 
5773 	rep_fc_f2r = strtoul(value, &end, 10);
5774 	if (end == NULL || *end != '\0' ||
5775 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5776 		PMD_DRV_LOG(ERR,
5777 			    "Invalid parameter passed to rep_fc_f2r "
5778 			    "devargs.\n");
5779 		return -EINVAL;
5780 	}
5781 
5782 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5783 		PMD_DRV_LOG(ERR,
5784 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5785 		return -EINVAL;
5786 	}
5787 
5788 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5789 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5790 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5791 
5792 	return 0;
5793 }
5794 
5795 static int
5796 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5797 {
5798 	struct rte_kvargs *kvlist;
5799 	int ret = 0;
5800 
5801 	if (devargs == NULL)
5802 		return 0;
5803 
5804 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5805 	if (kvlist == NULL)
5806 		return -EINVAL;
5807 
5808 	/*
5809 	 * Handler for "flow_xstat" devarg.
5810 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5811 	 */
5812 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5813 				 bnxt_parse_devarg_flow_xstat, bp);
5814 	if (ret)
5815 		goto err;
5816 
5817 	/*
5818 	 * Handler for "max_num_kflows" devarg.
5819 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5820 	 */
5821 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5822 				 bnxt_parse_devarg_max_num_kflows, bp);
5823 	if (ret)
5824 		goto err;
5825 
5826 err:
5827 	/*
5828 	 * Handler for "app-id" devarg.
5829 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5830 	 */
5831 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5832 			   bnxt_parse_devarg_app_id, bp);
5833 
5834 	/*
5835 	 * Handler for "ieee-1588" devarg.
5836 	 * Invoked as for ex: "-a 000:00:0d.0,ieee-1588=1"
5837 	 */
5838 	rte_kvargs_process(kvlist, BNXT_DEVARG_IEEE_1588,
5839 			   bnxt_parse_devarg_ieee_1588, bp);
5840 
5841 	rte_kvargs_free(kvlist);
5842 	return ret;
5843 }
5844 
5845 /* Allocate and initialize various fields in bnxt struct that
5846  * need to be allocated/destroyed only once in the lifetime of the driver
5847  */
5848 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5849 {
5850 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5851 	struct bnxt *bp = eth_dev->data->dev_private;
5852 	int rc = 0;
5853 
5854 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5855 
5856 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5857 		bp->flags |= BNXT_FLAG_VF;
5858 
5859 	if (bnxt_p5_device(pci_dev->id.device_id))
5860 		bp->flags |= BNXT_FLAG_CHIP_P5;
5861 
5862 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5863 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5864 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5865 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5866 		bp->flags |= BNXT_FLAG_STINGRAY;
5867 
5868 	rc = bnxt_map_pci_bars(eth_dev);
5869 	if (rc) {
5870 		PMD_DRV_LOG(ERR,
5871 			    "Failed to initialize board rc: %x\n", rc);
5872 		return rc;
5873 	}
5874 
5875 	rc = bnxt_alloc_pf_info(bp);
5876 	if (rc)
5877 		return rc;
5878 
5879 	rc = bnxt_alloc_link_info(bp);
5880 	if (rc)
5881 		return rc;
5882 
5883 	rc = bnxt_alloc_parent_info(bp);
5884 	if (rc)
5885 		return rc;
5886 
5887 	rc = bnxt_alloc_hwrm_resources(bp);
5888 	if (rc) {
5889 		PMD_DRV_LOG(ERR,
5890 			    "Failed to allocate response buffer rc: %x\n", rc);
5891 		return rc;
5892 	}
5893 	rc = bnxt_alloc_leds_info(bp);
5894 	if (rc)
5895 		return rc;
5896 
5897 	rc = bnxt_alloc_cos_queues(bp);
5898 	if (rc)
5899 		return rc;
5900 
5901 	rc = bnxt_init_locks(bp);
5902 	if (rc)
5903 		return rc;
5904 
5905 	rc = bnxt_get_config(bp);
5906 	if (rc)
5907 		return rc;
5908 
5909 	if (BNXT_TRUFLOW_EN(bp)) {
5910 		/* extra mbuf field is required to store CFA code from mark */
5911 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5912 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5913 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5914 			.align = alignof(bnxt_cfa_code_dynfield_t),
5915 		};
5916 		bnxt_cfa_code_dynfield_offset =
5917 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5918 		if (bnxt_cfa_code_dynfield_offset < 0) {
5919 			PMD_DRV_LOG(ERR,
5920 			    "Failed to register mbuf field for TruFlow mark\n");
5921 			return -rte_errno;
5922 		}
5923 	}
5924 
5925 	return rc;
5926 }
5927 
5928 static int
5929 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5930 {
5931 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5932 	static int version_printed;
5933 	struct bnxt *bp;
5934 	int rc;
5935 
5936 	if (version_printed++ == 0)
5937 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5938 
5939 	eth_dev->dev_ops = &bnxt_dev_ops;
5940 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5941 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5942 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5943 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5944 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5945 
5946 	/*
5947 	 * For secondary processes, we don't initialise any further
5948 	 * as primary has already done this work.
5949 	 */
5950 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5951 		return 0;
5952 
5953 	rte_eth_copy_pci_info(eth_dev, pci_dev);
5954 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5955 	eth_dev->data->dev_flags |= RTE_ETH_DEV_INTR_LSC;
5956 
5957 	bp = eth_dev->data->dev_private;
5958 
5959 	/* set the default app id */
5960 	bp->app_id = bnxt_ulp_default_app_id_get();
5961 
5962 	/* Parse dev arguments passed on when starting the DPDK application. */
5963 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5964 	if (rc)
5965 		goto error_free;
5966 
5967 	rc = bnxt_drv_init(eth_dev);
5968 	if (rc)
5969 		goto error_free;
5970 
5971 	rc = bnxt_init_resources(bp, false);
5972 	if (rc)
5973 		goto error_free;
5974 
5975 	rc = bnxt_alloc_stats_mem(bp);
5976 	if (rc)
5977 		goto error_free;
5978 
5979 	PMD_DRV_LOG(INFO,
5980 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
5981 		    DRV_MODULE_NAME,
5982 		    pci_dev->mem_resource[0].phys_addr,
5983 		    pci_dev->mem_resource[0].addr);
5984 
5985 	return 0;
5986 
5987 error_free:
5988 	bnxt_dev_uninit(eth_dev);
5989 	return rc;
5990 }
5991 
5992 
5993 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5994 {
5995 	if (!ctx)
5996 		return;
5997 
5998 	rte_free(ctx->va);
5999 
6000 	ctx->va = NULL;
6001 	ctx->dma = RTE_BAD_IOVA;
6002 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
6003 }
6004 
6005 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
6006 {
6007 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
6008 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6009 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
6010 				  bp->flow_stat->max_fc,
6011 				  false);
6012 
6013 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
6014 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6015 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
6016 				  bp->flow_stat->max_fc,
6017 				  false);
6018 
6019 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6020 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
6021 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6022 
6023 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6024 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
6025 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6026 
6027 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6028 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
6029 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6030 
6031 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6032 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
6033 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6034 }
6035 
6036 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
6037 {
6038 	bnxt_unregister_fc_ctx_mem(bp);
6039 
6040 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
6041 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
6042 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
6043 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
6044 }
6045 
6046 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
6047 {
6048 	if (BNXT_FLOW_XSTATS_EN(bp))
6049 		bnxt_uninit_fc_ctx_mem(bp);
6050 }
6051 
6052 static void
6053 bnxt_free_error_recovery_info(struct bnxt *bp)
6054 {
6055 	rte_free(bp->recovery_info);
6056 	bp->recovery_info = NULL;
6057 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
6058 }
6059 
6060 static int
6061 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
6062 {
6063 	int rc;
6064 
6065 	bnxt_free_int(bp);
6066 	bnxt_free_mem(bp, reconfig_dev);
6067 
6068 	bnxt_hwrm_func_buf_unrgtr(bp);
6069 	if (bp->pf != NULL) {
6070 		rte_free(bp->pf->vf_req_buf);
6071 		bp->pf->vf_req_buf = NULL;
6072 	}
6073 
6074 	rc = bnxt_hwrm_func_driver_unregister(bp);
6075 	bp->flags &= ~BNXT_FLAG_REGISTERED;
6076 	bnxt_free_ctx_mem(bp);
6077 	if (!reconfig_dev) {
6078 		bnxt_free_hwrm_resources(bp);
6079 		bnxt_free_error_recovery_info(bp);
6080 		rte_free(bp->mcast_addr_list);
6081 		bp->mcast_addr_list = NULL;
6082 		rte_free(bp->rss_conf.rss_key);
6083 		bp->rss_conf.rss_key = NULL;
6084 	}
6085 
6086 	bnxt_uninit_ctx_mem(bp);
6087 
6088 	bnxt_free_flow_stats_info(bp);
6089 	bnxt_free_switch_domain(bp);
6090 	rte_free(bp->ptp_cfg);
6091 	bp->ptp_cfg = NULL;
6092 	return rc;
6093 }
6094 
6095 static int
6096 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
6097 {
6098 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6099 		return -EPERM;
6100 
6101 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
6102 
6103 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
6104 		bnxt_dev_close_op(eth_dev);
6105 
6106 	return 0;
6107 }
6108 
6109 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6110 {
6111 	struct bnxt *bp = eth_dev->data->dev_private;
6112 	struct rte_eth_dev *vf_rep_eth_dev;
6113 	int ret = 0, i;
6114 
6115 	if (!bp)
6116 		return -EINVAL;
6117 
6118 	for (i = 0; i < bp->num_reps; i++) {
6119 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6120 		if (!vf_rep_eth_dev)
6121 			continue;
6122 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6123 			    vf_rep_eth_dev->data->port_id);
6124 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6125 	}
6126 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6127 		    eth_dev->data->port_id);
6128 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6129 
6130 	return ret;
6131 }
6132 
6133 static void bnxt_free_rep_info(struct bnxt *bp)
6134 {
6135 	rte_free(bp->rep_info);
6136 	bp->rep_info = NULL;
6137 	rte_free(bp->cfa_code_map);
6138 	bp->cfa_code_map = NULL;
6139 }
6140 
6141 static int bnxt_init_rep_info(struct bnxt *bp)
6142 {
6143 	int i = 0, rc;
6144 
6145 	if (bp->rep_info)
6146 		return 0;
6147 
6148 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
6149 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS(bp),
6150 				   0);
6151 	if (!bp->rep_info) {
6152 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6153 		return -ENOMEM;
6154 	}
6155 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6156 				       sizeof(*bp->cfa_code_map) *
6157 				       BNXT_MAX_CFA_CODE, 0);
6158 	if (!bp->cfa_code_map) {
6159 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6160 		bnxt_free_rep_info(bp);
6161 		return -ENOMEM;
6162 	}
6163 
6164 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6165 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6166 
6167 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6168 	if (rc) {
6169 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6170 		bnxt_free_rep_info(bp);
6171 		return rc;
6172 	}
6173 
6174 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6175 	if (rc) {
6176 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6177 		bnxt_free_rep_info(bp);
6178 		return rc;
6179 	}
6180 
6181 	return rc;
6182 }
6183 
6184 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6185 			       struct rte_eth_devargs *eth_da,
6186 			       struct rte_eth_dev *backing_eth_dev,
6187 			       const char *dev_args)
6188 {
6189 	struct rte_eth_dev *vf_rep_eth_dev;
6190 	char name[RTE_ETH_NAME_MAX_LEN];
6191 	struct bnxt *backing_bp = backing_eth_dev->data->dev_private;
6192 	uint16_t max_vf_reps = BNXT_MAX_VF_REPS(backing_bp);
6193 
6194 	uint16_t num_rep;
6195 	int i, ret = 0;
6196 	struct rte_kvargs *kvlist = NULL;
6197 
6198 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6199 		return 0;
6200 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6201 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6202 			    eth_da->type);
6203 		return -ENOTSUP;
6204 	}
6205 	num_rep = eth_da->nb_representor_ports;
6206 	if (num_rep > max_vf_reps) {
6207 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6208 			    num_rep, max_vf_reps);
6209 		return -EINVAL;
6210 	}
6211 
6212 	if (num_rep >= RTE_MAX_ETHPORTS) {
6213 		PMD_DRV_LOG(ERR,
6214 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6215 			    num_rep, RTE_MAX_ETHPORTS);
6216 		return -EINVAL;
6217 	}
6218 
6219 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6220 		PMD_DRV_LOG(ERR,
6221 			    "Not a PF or trusted VF. No Representor support\n");
6222 		/* Returning an error is not an option.
6223 		 * Applications are not handling this correctly
6224 		 */
6225 		return 0;
6226 	}
6227 
6228 	if (bnxt_init_rep_info(backing_bp))
6229 		return 0;
6230 
6231 	for (i = 0; i < num_rep; i++) {
6232 		struct bnxt_representor representor = {
6233 			.vf_id = eth_da->representor_ports[i],
6234 			.switch_domain_id = backing_bp->switch_domain_id,
6235 			.parent_dev = backing_eth_dev
6236 		};
6237 
6238 		if (representor.vf_id >= max_vf_reps) {
6239 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6240 				    representor.vf_id, max_vf_reps);
6241 			continue;
6242 		}
6243 
6244 		/* representor port net_bdf_port */
6245 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6246 			 pci_dev->device.name, eth_da->representor_ports[i]);
6247 
6248 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6249 		if (kvlist) {
6250 			/*
6251 			 * Handler for "rep_is_pf" devarg.
6252 			 * Invoked as for ex: "-a 000:00:0d.0,
6253 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6254 			 */
6255 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6256 						 bnxt_parse_devarg_rep_is_pf,
6257 						 (void *)&representor);
6258 			if (ret) {
6259 				ret = -EINVAL;
6260 				goto err;
6261 			}
6262 			/*
6263 			 * Handler for "rep_based_pf" devarg.
6264 			 * Invoked as for ex: "-a 000:00:0d.0,
6265 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6266 			 */
6267 			ret = rte_kvargs_process(kvlist,
6268 						 BNXT_DEVARG_REP_BASED_PF,
6269 						 bnxt_parse_devarg_rep_based_pf,
6270 						 (void *)&representor);
6271 			if (ret) {
6272 				ret = -EINVAL;
6273 				goto err;
6274 			}
6275 			/*
6276 			 * Handler for "rep_based_pf" devarg.
6277 			 * Invoked as for ex: "-a 000:00:0d.0,
6278 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6279 			 */
6280 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6281 						 bnxt_parse_devarg_rep_q_r2f,
6282 						 (void *)&representor);
6283 			if (ret) {
6284 				ret = -EINVAL;
6285 				goto err;
6286 			}
6287 			/*
6288 			 * Handler for "rep_based_pf" devarg.
6289 			 * Invoked as for ex: "-a 000:00:0d.0,
6290 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6291 			 */
6292 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6293 						 bnxt_parse_devarg_rep_q_f2r,
6294 						 (void *)&representor);
6295 			if (ret) {
6296 				ret = -EINVAL;
6297 				goto err;
6298 			}
6299 			/*
6300 			 * Handler for "rep_based_pf" devarg.
6301 			 * Invoked as for ex: "-a 000:00:0d.0,
6302 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6303 			 */
6304 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6305 						 bnxt_parse_devarg_rep_fc_r2f,
6306 						 (void *)&representor);
6307 			if (ret) {
6308 				ret = -EINVAL;
6309 				goto err;
6310 			}
6311 			/*
6312 			 * Handler for "rep_based_pf" devarg.
6313 			 * Invoked as for ex: "-a 000:00:0d.0,
6314 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6315 			 */
6316 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6317 						 bnxt_parse_devarg_rep_fc_f2r,
6318 						 (void *)&representor);
6319 			if (ret) {
6320 				ret = -EINVAL;
6321 				goto err;
6322 			}
6323 		}
6324 
6325 		ret = rte_eth_dev_create(&pci_dev->device, name,
6326 					 sizeof(struct bnxt_representor),
6327 					 NULL, NULL,
6328 					 bnxt_representor_init,
6329 					 &representor);
6330 		if (ret) {
6331 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6332 				    "representor %s.", name);
6333 			goto err;
6334 		}
6335 
6336 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6337 		if (!vf_rep_eth_dev) {
6338 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6339 				    " for VF-Rep: %s.", name);
6340 			ret = -ENODEV;
6341 			goto err;
6342 		}
6343 
6344 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6345 			    backing_eth_dev->data->port_id);
6346 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6347 							 vf_rep_eth_dev;
6348 		backing_bp->num_reps++;
6349 
6350 	}
6351 
6352 	rte_kvargs_free(kvlist);
6353 	return 0;
6354 
6355 err:
6356 	/* If num_rep > 1, then rollback already created
6357 	 * ports, since we'll be failing the probe anyway
6358 	 */
6359 	if (num_rep > 1)
6360 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6361 	rte_errno = -ret;
6362 	rte_kvargs_free(kvlist);
6363 
6364 	return ret;
6365 }
6366 
6367 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6368 			  struct rte_pci_device *pci_dev)
6369 {
6370 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6371 	struct rte_eth_dev *backing_eth_dev;
6372 	uint16_t num_rep;
6373 	int ret = 0;
6374 
6375 	if (pci_dev->device.devargs) {
6376 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6377 					    &eth_da);
6378 		if (ret)
6379 			return ret;
6380 	}
6381 
6382 	num_rep = eth_da.nb_representor_ports;
6383 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6384 		    num_rep);
6385 
6386 	/* We could come here after first level of probe is already invoked
6387 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6388 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6389 	 */
6390 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6391 	if (backing_eth_dev == NULL) {
6392 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6393 					 sizeof(struct bnxt),
6394 					 eth_dev_pci_specific_init, pci_dev,
6395 					 bnxt_dev_init, NULL);
6396 
6397 		if (ret || !num_rep)
6398 			return ret;
6399 
6400 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6401 	}
6402 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6403 		    backing_eth_dev->data->port_id);
6404 
6405 	if (!num_rep)
6406 		return ret;
6407 
6408 	/* probe representor ports now */
6409 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6410 				  pci_dev->device.devargs->args);
6411 
6412 	return ret;
6413 }
6414 
6415 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6416 {
6417 	struct rte_eth_dev *eth_dev;
6418 
6419 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6420 	if (!eth_dev)
6421 		return 0; /* Invoked typically only by OVS-DPDK, by the
6422 			   * time it comes here the eth_dev is already
6423 			   * deleted by rte_eth_dev_close(), so returning
6424 			   * +ve value will at least help in proper cleanup
6425 			   */
6426 
6427 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6428 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6429 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6430 			return rte_eth_dev_destroy(eth_dev,
6431 						   bnxt_representor_uninit);
6432 		else
6433 			return rte_eth_dev_destroy(eth_dev,
6434 						   bnxt_dev_uninit);
6435 	} else {
6436 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6437 	}
6438 }
6439 
6440 static struct rte_pci_driver bnxt_rte_pmd = {
6441 	.id_table = bnxt_pci_id_map,
6442 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6443 			RTE_PCI_DRV_INTR_RMV |
6444 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6445 						  * and OVS-DPDK
6446 						  */
6447 	.probe = bnxt_pci_probe,
6448 	.remove = bnxt_pci_remove,
6449 };
6450 
6451 static bool
6452 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6453 {
6454 	if (strcmp(dev->device->driver->name, drv->driver.name))
6455 		return false;
6456 
6457 	return true;
6458 }
6459 
6460 bool is_bnxt_supported(struct rte_eth_dev *dev)
6461 {
6462 	return is_device_supported(dev, &bnxt_rte_pmd);
6463 }
6464 
6465 struct tf *bnxt_get_tfp_session(struct bnxt *bp, enum bnxt_session_type type)
6466 {
6467 	return (type >= BNXT_SESSION_TYPE_LAST) ?
6468 		&bp->tfp[BNXT_SESSION_TYPE_REGULAR] : &bp->tfp[type];
6469 }
6470 
6471 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6472 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6473 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6474 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6475