xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision dbd34bee)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2014-2021 Broadcom
3  * All rights reserved.
4  */
5 
6 #include <inttypes.h>
7 #include <stdbool.h>
8 
9 #include <rte_dev.h>
10 #include <ethdev_driver.h>
11 #include <ethdev_pci.h>
12 #include <rte_malloc.h>
13 #include <rte_cycles.h>
14 #include <rte_alarm.h>
15 #include <rte_kvargs.h>
16 #include <rte_vect.h>
17 
18 #include "bnxt.h"
19 #include "bnxt_filter.h"
20 #include "bnxt_hwrm.h"
21 #include "bnxt_irq.h"
22 #include "bnxt_reps.h"
23 #include "bnxt_ring.h"
24 #include "bnxt_rxq.h"
25 #include "bnxt_rxr.h"
26 #include "bnxt_stats.h"
27 #include "bnxt_txq.h"
28 #include "bnxt_txr.h"
29 #include "bnxt_vnic.h"
30 #include "hsi_struct_def_dpdk.h"
31 #include "bnxt_nvm_defs.h"
32 #include "bnxt_tf_common.h"
33 #include "ulp_flow_db.h"
34 #include "rte_pmd_bnxt.h"
35 
36 #define DRV_MODULE_NAME		"bnxt"
37 static const char bnxt_version[] =
38 	"Broadcom NetXtreme driver " DRV_MODULE_NAME;
39 
40 /*
41  * The set of PCI devices this driver supports
42  */
43 static const struct rte_pci_id bnxt_pci_id_map[] = {
44 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
45 			 BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
46 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
47 			 BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
48 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
49 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
50 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
51 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
52 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
53 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
54 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
55 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
56 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
57 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
58 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
59 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
60 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
61 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
62 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
63 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
64 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
65 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
66 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
67 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
68 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
69 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
70 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
71 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
72 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
73 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
74 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
75 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
76 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
77 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
78 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
79 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
80 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
81 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
82 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
83 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58812) },
84 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58814) },
85 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818) },
86 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818_VF) },
87 	{ .vendor_id = 0, /* sentinel */ },
88 };
89 
90 #define	BNXT_DEVARG_ACCUM_STATS	"accum-stats"
91 #define BNXT_DEVARG_FLOW_XSTAT	"flow-xstat"
92 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
93 #define BNXT_DEVARG_REPRESENTOR	"representor"
94 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
95 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
96 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
97 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
98 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
99 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
100 #define BNXT_DEVARG_APP_ID	"app-id"
101 
102 static const char *const bnxt_dev_args[] = {
103 	BNXT_DEVARG_REPRESENTOR,
104 	BNXT_DEVARG_ACCUM_STATS,
105 	BNXT_DEVARG_FLOW_XSTAT,
106 	BNXT_DEVARG_MAX_NUM_KFLOWS,
107 	BNXT_DEVARG_REP_BASED_PF,
108 	BNXT_DEVARG_REP_IS_PF,
109 	BNXT_DEVARG_REP_Q_R2F,
110 	BNXT_DEVARG_REP_Q_F2R,
111 	BNXT_DEVARG_REP_FC_R2F,
112 	BNXT_DEVARG_REP_FC_F2R,
113 	BNXT_DEVARG_APP_ID,
114 	NULL
115 };
116 
117 /*
118  * accum-stats == false to disable flow counter accumulation
119  * accum-stats == true to enable flow counter accumulation
120  */
121 #define	BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)	((accum_stats) > 1)
122 
123 /*
124  * app-id = an non-negative 8-bit number
125  */
126 #define BNXT_DEVARG_APP_ID_INVALID(val)			((val) > 255)
127 
128 /*
129  * flow_xstat == false to disable the feature
130  * flow_xstat == true to enable the feature
131  */
132 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
133 
134 /*
135  * rep_is_pf == false to indicate VF representor
136  * rep_is_pf == true to indicate PF representor
137  */
138 #define	BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)	((rep_is_pf) > 1)
139 
140 /*
141  * rep_based_pf == Physical index of the PF
142  */
143 #define	BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)	((rep_based_pf) > 15)
144 /*
145  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
146  */
147 #define	BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)	((rep_q_r2f) > 3)
148 
149 /*
150  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
151  */
152 #define	BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)	((rep_q_f2r) > 3)
153 
154 /*
155  * rep_fc_r2f == Flow control for the representor to endpoint direction
156  */
157 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)	((rep_fc_r2f) > 1)
158 
159 /*
160  * rep_fc_f2r == Flow control for the endpoint to representor direction
161  */
162 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)	((rep_fc_f2r) > 1)
163 
164 int bnxt_cfa_code_dynfield_offset = -1;
165 
166 /*
167  * max_num_kflows must be >= 32
168  * and must be a power-of-2 supported value
169  * return: 1 -> invalid
170  *         0 -> valid
171  */
172 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
173 {
174 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
175 		return 1;
176 	return 0;
177 }
178 
179 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
180 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
181 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
182 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
183 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
184 static int bnxt_restore_vlan_filters(struct bnxt *bp);
185 static void bnxt_dev_recover(void *arg);
186 static void bnxt_free_error_recovery_info(struct bnxt *bp);
187 static void bnxt_free_rep_info(struct bnxt *bp);
188 
189 int is_bnxt_in_error(struct bnxt *bp)
190 {
191 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
192 		return -EIO;
193 	if (bp->flags & BNXT_FLAG_FW_RESET)
194 		return -EBUSY;
195 
196 	return 0;
197 }
198 
199 /***********************/
200 
201 /*
202  * High level utility functions
203  */
204 
205 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
206 {
207 	unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
208 					     BNXT_RSS_TBL_SIZE_P5);
209 
210 	if (!BNXT_CHIP_P5(bp))
211 		return 1;
212 
213 	return RTE_ALIGN_MUL_CEIL(num_rss_rings,
214 				  BNXT_RSS_ENTRIES_PER_CTX_P5) /
215 				  BNXT_RSS_ENTRIES_PER_CTX_P5;
216 }
217 
218 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
219 {
220 	if (!BNXT_CHIP_P5(bp))
221 		return HW_HASH_INDEX_SIZE;
222 
223 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
224 }
225 
226 static void bnxt_free_parent_info(struct bnxt *bp)
227 {
228 	rte_free(bp->parent);
229 	bp->parent = NULL;
230 }
231 
232 static void bnxt_free_pf_info(struct bnxt *bp)
233 {
234 	rte_free(bp->pf);
235 	bp->pf = NULL;
236 }
237 
238 static void bnxt_free_link_info(struct bnxt *bp)
239 {
240 	rte_free(bp->link_info);
241 	bp->link_info = NULL;
242 }
243 
244 static void bnxt_free_leds_info(struct bnxt *bp)
245 {
246 	if (BNXT_VF(bp))
247 		return;
248 
249 	rte_free(bp->leds);
250 	bp->leds = NULL;
251 }
252 
253 static void bnxt_free_flow_stats_info(struct bnxt *bp)
254 {
255 	rte_free(bp->flow_stat);
256 	bp->flow_stat = NULL;
257 }
258 
259 static void bnxt_free_cos_queues(struct bnxt *bp)
260 {
261 	rte_free(bp->rx_cos_queue);
262 	bp->rx_cos_queue = NULL;
263 	rte_free(bp->tx_cos_queue);
264 	bp->tx_cos_queue = NULL;
265 }
266 
267 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
268 {
269 	bnxt_free_filter_mem(bp);
270 	bnxt_free_vnic_attributes(bp);
271 	bnxt_free_vnic_mem(bp);
272 
273 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
274 	 * If the number of rings change across fw update,
275 	 * we don't have much choice except to warn the user.
276 	 */
277 	if (!reconfig) {
278 		bnxt_free_stats(bp);
279 		bnxt_free_tx_rings(bp);
280 		bnxt_free_rx_rings(bp);
281 	}
282 	bnxt_free_async_cp_ring(bp);
283 	bnxt_free_rxtx_nq_ring(bp);
284 
285 	rte_free(bp->grp_info);
286 	bp->grp_info = NULL;
287 }
288 
289 static int bnxt_alloc_parent_info(struct bnxt *bp)
290 {
291 	bp->parent = rte_zmalloc("bnxt_parent_info",
292 				 sizeof(struct bnxt_parent_info), 0);
293 	if (bp->parent == NULL)
294 		return -ENOMEM;
295 
296 	return 0;
297 }
298 
299 static int bnxt_alloc_pf_info(struct bnxt *bp)
300 {
301 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
302 	if (bp->pf == NULL)
303 		return -ENOMEM;
304 
305 	return 0;
306 }
307 
308 static int bnxt_alloc_link_info(struct bnxt *bp)
309 {
310 	bp->link_info =
311 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
312 	if (bp->link_info == NULL)
313 		return -ENOMEM;
314 
315 	return 0;
316 }
317 
318 static int bnxt_alloc_leds_info(struct bnxt *bp)
319 {
320 	if (BNXT_VF(bp))
321 		return 0;
322 
323 	bp->leds = rte_zmalloc("bnxt_leds",
324 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
325 			       0);
326 	if (bp->leds == NULL)
327 		return -ENOMEM;
328 
329 	return 0;
330 }
331 
332 static int bnxt_alloc_cos_queues(struct bnxt *bp)
333 {
334 	bp->rx_cos_queue =
335 		rte_zmalloc("bnxt_rx_cosq",
336 			    BNXT_COS_QUEUE_COUNT *
337 			    sizeof(struct bnxt_cos_queue_info),
338 			    0);
339 	if (bp->rx_cos_queue == NULL)
340 		return -ENOMEM;
341 
342 	bp->tx_cos_queue =
343 		rte_zmalloc("bnxt_tx_cosq",
344 			    BNXT_COS_QUEUE_COUNT *
345 			    sizeof(struct bnxt_cos_queue_info),
346 			    0);
347 	if (bp->tx_cos_queue == NULL)
348 		return -ENOMEM;
349 
350 	return 0;
351 }
352 
353 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
354 {
355 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
356 				    sizeof(struct bnxt_flow_stat_info), 0);
357 	if (bp->flow_stat == NULL)
358 		return -ENOMEM;
359 
360 	return 0;
361 }
362 
363 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
364 {
365 	int rc;
366 
367 	rc = bnxt_alloc_ring_grps(bp);
368 	if (rc)
369 		goto alloc_mem_err;
370 
371 	rc = bnxt_alloc_async_ring_struct(bp);
372 	if (rc)
373 		goto alloc_mem_err;
374 
375 	rc = bnxt_alloc_vnic_mem(bp);
376 	if (rc)
377 		goto alloc_mem_err;
378 
379 	rc = bnxt_alloc_vnic_attributes(bp);
380 	if (rc)
381 		goto alloc_mem_err;
382 
383 	rc = bnxt_alloc_filter_mem(bp);
384 	if (rc)
385 		goto alloc_mem_err;
386 
387 	rc = bnxt_alloc_async_cp_ring(bp);
388 	if (rc)
389 		goto alloc_mem_err;
390 
391 	rc = bnxt_alloc_rxtx_nq_ring(bp);
392 	if (rc)
393 		goto alloc_mem_err;
394 
395 	if (BNXT_FLOW_XSTATS_EN(bp)) {
396 		rc = bnxt_alloc_flow_stats_info(bp);
397 		if (rc)
398 			goto alloc_mem_err;
399 	}
400 
401 	return 0;
402 
403 alloc_mem_err:
404 	bnxt_free_mem(bp, reconfig);
405 	return rc;
406 }
407 
408 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
409 {
410 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
411 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
412 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
413 	struct bnxt_rx_queue *rxq;
414 	unsigned int j;
415 	int rc;
416 
417 	rc = bnxt_vnic_grp_alloc(bp, vnic);
418 	if (rc)
419 		goto err_out;
420 
421 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
422 		    vnic_id, vnic, vnic->fw_grp_ids);
423 
424 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
425 	if (rc)
426 		goto err_out;
427 
428 	/* Alloc RSS context only if RSS mode is enabled */
429 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
430 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
431 
432 		/* RSS table size in Thor is 512.
433 		 * Cap max Rx rings to same value
434 		 */
435 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
436 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
437 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
438 			goto err_out;
439 		}
440 
441 		rc = 0;
442 		for (j = 0; j < nr_ctxs; j++) {
443 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
444 			if (rc)
445 				break;
446 		}
447 		if (rc) {
448 			PMD_DRV_LOG(ERR,
449 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
450 				    vnic_id, j, rc);
451 			goto err_out;
452 		}
453 		vnic->num_lb_ctxts = nr_ctxs;
454 	}
455 
456 	/*
457 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
458 	 * setting is not available at this time, it will not be
459 	 * configured correctly in the CFA.
460 	 */
461 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
462 		vnic->vlan_strip = true;
463 	else
464 		vnic->vlan_strip = false;
465 
466 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
467 	if (rc)
468 		goto err_out;
469 
470 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
471 	if (rc)
472 		goto err_out;
473 
474 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
475 		rxq = bp->eth_dev->data->rx_queues[j];
476 
477 		PMD_DRV_LOG(DEBUG,
478 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
479 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
480 
481 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
482 			rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
483 		else
484 			vnic->rx_queue_cnt++;
485 	}
486 
487 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
488 
489 	rc = bnxt_vnic_rss_configure(bp, vnic);
490 	if (rc)
491 		goto err_out;
492 
493 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
494 
495 	rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
496 				    (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO) ?
497 				    true : false);
498 	if (rc)
499 		goto err_out;
500 
501 	return 0;
502 err_out:
503 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
504 		    vnic_id, rc);
505 	return rc;
506 }
507 
508 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
509 {
510 	int rc = 0;
511 
512 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
513 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
514 	if (rc)
515 		return rc;
516 
517 	PMD_DRV_LOG(DEBUG,
518 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
519 		    " rx_fc_in_tbl.ctx_id = %d\n",
520 		    bp->flow_stat->rx_fc_in_tbl.va,
521 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
522 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
523 
524 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
525 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
526 	if (rc)
527 		return rc;
528 
529 	PMD_DRV_LOG(DEBUG,
530 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
531 		    " rx_fc_out_tbl.ctx_id = %d\n",
532 		    bp->flow_stat->rx_fc_out_tbl.va,
533 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
534 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
535 
536 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
537 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
538 	if (rc)
539 		return rc;
540 
541 	PMD_DRV_LOG(DEBUG,
542 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
543 		    " tx_fc_in_tbl.ctx_id = %d\n",
544 		    bp->flow_stat->tx_fc_in_tbl.va,
545 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
546 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
547 
548 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
549 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
550 	if (rc)
551 		return rc;
552 
553 	PMD_DRV_LOG(DEBUG,
554 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
555 		    " tx_fc_out_tbl.ctx_id = %d\n",
556 		    bp->flow_stat->tx_fc_out_tbl.va,
557 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
558 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
559 
560 	memset(bp->flow_stat->rx_fc_out_tbl.va,
561 	       0,
562 	       bp->flow_stat->rx_fc_out_tbl.size);
563 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
564 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
565 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
566 				       bp->flow_stat->max_fc,
567 				       true);
568 	if (rc)
569 		return rc;
570 
571 	memset(bp->flow_stat->tx_fc_out_tbl.va,
572 	       0,
573 	       bp->flow_stat->tx_fc_out_tbl.size);
574 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
575 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
576 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
577 				       bp->flow_stat->max_fc,
578 				       true);
579 
580 	return rc;
581 }
582 
583 static int bnxt_alloc_ctx_mem_buf(struct bnxt *bp, char *type, size_t size,
584 				  struct bnxt_ctx_mem_buf_info *ctx)
585 {
586 	if (!ctx)
587 		return -EINVAL;
588 
589 	ctx->va = rte_zmalloc_socket(type, size, 0,
590 				     bp->eth_dev->device->numa_node);
591 	if (ctx->va == NULL)
592 		return -ENOMEM;
593 	rte_mem_lock_page(ctx->va);
594 	ctx->size = size;
595 	ctx->dma = rte_mem_virt2iova(ctx->va);
596 	if (ctx->dma == RTE_BAD_IOVA)
597 		return -ENOMEM;
598 
599 	return 0;
600 }
601 
602 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
603 {
604 	struct rte_pci_device *pdev = bp->pdev;
605 	char type[RTE_MEMZONE_NAMESIZE];
606 	uint16_t max_fc;
607 	int rc = 0;
608 
609 	max_fc = bp->flow_stat->max_fc;
610 
611 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
612 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
613 	/* 4 bytes for each counter-id */
614 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
615 				    max_fc * 4,
616 				    &bp->flow_stat->rx_fc_in_tbl);
617 	if (rc)
618 		return rc;
619 
620 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
621 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
622 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
623 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
624 				    max_fc * 16,
625 				    &bp->flow_stat->rx_fc_out_tbl);
626 	if (rc)
627 		return rc;
628 
629 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
630 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
631 	/* 4 bytes for each counter-id */
632 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
633 				    max_fc * 4,
634 				    &bp->flow_stat->tx_fc_in_tbl);
635 	if (rc)
636 		return rc;
637 
638 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
639 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
640 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
641 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
642 				    max_fc * 16,
643 				    &bp->flow_stat->tx_fc_out_tbl);
644 	if (rc)
645 		return rc;
646 
647 	rc = bnxt_register_fc_ctx_mem(bp);
648 
649 	return rc;
650 }
651 
652 static int bnxt_init_ctx_mem(struct bnxt *bp)
653 {
654 	int rc = 0;
655 
656 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
657 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
658 	    !BNXT_FLOW_XSTATS_EN(bp))
659 		return 0;
660 
661 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
662 	if (rc)
663 		return rc;
664 
665 	rc = bnxt_init_fc_ctx_mem(bp);
666 
667 	return rc;
668 }
669 
670 static int bnxt_update_phy_setting(struct bnxt *bp)
671 {
672 	struct rte_eth_link new;
673 	int rc;
674 
675 	rc = bnxt_get_hwrm_link_config(bp, &new);
676 	if (rc) {
677 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
678 		return rc;
679 	}
680 
681 	/*
682 	 * On BCM957508-N2100 adapters, FW will not allow any user other
683 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
684 	 * always returns link up. Force phy update always in that case.
685 	 */
686 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
687 		rc = bnxt_set_hwrm_link_config(bp, true);
688 		if (rc) {
689 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
690 			return rc;
691 		}
692 	}
693 
694 	return rc;
695 }
696 
697 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
698 {
699 	rte_free(bp->prev_rx_ring_stats);
700 	rte_free(bp->prev_tx_ring_stats);
701 
702 	bp->prev_rx_ring_stats = NULL;
703 	bp->prev_tx_ring_stats = NULL;
704 }
705 
706 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
707 {
708 	bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
709 					      sizeof(struct bnxt_ring_stats) *
710 					      bp->rx_cp_nr_rings,
711 					      0);
712 	if (bp->prev_rx_ring_stats == NULL)
713 		return -ENOMEM;
714 
715 	bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
716 					     sizeof(struct bnxt_ring_stats) *
717 					     bp->tx_cp_nr_rings,
718 					     0);
719 	if (bp->prev_tx_ring_stats == NULL)
720 		goto error;
721 
722 	return 0;
723 
724 error:
725 	bnxt_free_prev_ring_stats(bp);
726 	return -ENOMEM;
727 }
728 
729 static int bnxt_start_nic(struct bnxt *bp)
730 {
731 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
732 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
733 	uint32_t intr_vector = 0;
734 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
735 	uint32_t vec = BNXT_MISC_VEC_ID;
736 	unsigned int i, j;
737 	int rc;
738 
739 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
740 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
741 			DEV_RX_OFFLOAD_JUMBO_FRAME;
742 		bp->flags |= BNXT_FLAG_JUMBO;
743 	} else {
744 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
745 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
746 		bp->flags &= ~BNXT_FLAG_JUMBO;
747 	}
748 
749 	/* THOR does not support ring groups.
750 	 * But we will use the array to save RSS context IDs.
751 	 */
752 	if (BNXT_CHIP_P5(bp))
753 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
754 
755 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
756 	if (rc) {
757 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
758 		goto err_out;
759 	}
760 
761 	rc = bnxt_alloc_hwrm_rings(bp);
762 	if (rc) {
763 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
764 		goto err_out;
765 	}
766 
767 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
768 	if (rc) {
769 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
770 		goto err_out;
771 	}
772 
773 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
774 		goto skip_cosq_cfg;
775 
776 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
777 		if (bp->rx_cos_queue[i].id != 0xff) {
778 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
779 
780 			if (!vnic) {
781 				PMD_DRV_LOG(ERR,
782 					    "Num pools more than FW profile\n");
783 				rc = -EINVAL;
784 				goto err_out;
785 			}
786 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
787 			bp->rx_cosq_cnt++;
788 		}
789 	}
790 
791 skip_cosq_cfg:
792 	rc = bnxt_mq_rx_configure(bp);
793 	if (rc) {
794 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
795 		goto err_out;
796 	}
797 
798 	/* default vnic 0 */
799 	rc = bnxt_setup_one_vnic(bp, 0);
800 	if (rc)
801 		goto err_out;
802 	/* VNIC configuration */
803 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
804 		for (i = 1; i < bp->nr_vnics; i++) {
805 			rc = bnxt_setup_one_vnic(bp, i);
806 			if (rc)
807 				goto err_out;
808 		}
809 	}
810 
811 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
812 	if (rc) {
813 		PMD_DRV_LOG(ERR,
814 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
815 		goto err_out;
816 	}
817 
818 	/* check and configure queue intr-vector mapping */
819 	if ((rte_intr_cap_multiple(intr_handle) ||
820 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
821 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
822 		intr_vector = bp->eth_dev->data->nb_rx_queues;
823 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
824 		if (intr_vector > bp->rx_cp_nr_rings) {
825 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
826 					bp->rx_cp_nr_rings);
827 			return -ENOTSUP;
828 		}
829 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
830 		if (rc)
831 			return rc;
832 	}
833 
834 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
835 		intr_handle->intr_vec =
836 			rte_zmalloc("intr_vec",
837 				    bp->eth_dev->data->nb_rx_queues *
838 				    sizeof(int), 0);
839 		if (intr_handle->intr_vec == NULL) {
840 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
841 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
842 			rc = -ENOMEM;
843 			goto err_out;
844 		}
845 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
846 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
847 			 intr_handle->intr_vec, intr_handle->nb_efd,
848 			intr_handle->max_intr);
849 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
850 		     queue_id++) {
851 			intr_handle->intr_vec[queue_id] =
852 							vec + BNXT_RX_VEC_START;
853 			if (vec < base + intr_handle->nb_efd - 1)
854 				vec++;
855 		}
856 	}
857 
858 	/* enable uio/vfio intr/eventfd mapping */
859 	rc = rte_intr_enable(intr_handle);
860 #ifndef RTE_EXEC_ENV_FREEBSD
861 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
862 	if (rc)
863 		goto err_out;
864 #endif
865 
866 	rc = bnxt_update_phy_setting(bp);
867 	if (rc)
868 		goto err_out;
869 
870 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
871 	if (!bp->mark_table)
872 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
873 
874 	return 0;
875 
876 err_out:
877 	/* Some of the error status returned by FW may not be from errno.h */
878 	if (rc > 0)
879 		rc = -EIO;
880 
881 	return rc;
882 }
883 
884 static int bnxt_shutdown_nic(struct bnxt *bp)
885 {
886 	bnxt_free_all_hwrm_resources(bp);
887 	bnxt_free_all_filters(bp);
888 	bnxt_free_all_vnics(bp);
889 	return 0;
890 }
891 
892 /*
893  * Device configuration and status function
894  */
895 
896 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
897 {
898 	uint32_t link_speed = 0;
899 	uint32_t speed_capa = 0;
900 
901 	if (bp->link_info == NULL)
902 		return 0;
903 
904 	link_speed = bp->link_info->support_speeds;
905 
906 	/* If PAM4 is configured, use PAM4 supported speed */
907 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
908 		link_speed = bp->link_info->support_pam4_speeds;
909 
910 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
911 		speed_capa |= ETH_LINK_SPEED_100M;
912 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
913 		speed_capa |= ETH_LINK_SPEED_100M_HD;
914 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
915 		speed_capa |= ETH_LINK_SPEED_1G;
916 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
917 		speed_capa |= ETH_LINK_SPEED_2_5G;
918 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
919 		speed_capa |= ETH_LINK_SPEED_10G;
920 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
921 		speed_capa |= ETH_LINK_SPEED_20G;
922 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
923 		speed_capa |= ETH_LINK_SPEED_25G;
924 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
925 		speed_capa |= ETH_LINK_SPEED_40G;
926 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
927 		speed_capa |= ETH_LINK_SPEED_50G;
928 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
929 		speed_capa |= ETH_LINK_SPEED_100G;
930 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
931 		speed_capa |= ETH_LINK_SPEED_50G;
932 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
933 		speed_capa |= ETH_LINK_SPEED_100G;
934 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
935 		speed_capa |= ETH_LINK_SPEED_200G;
936 
937 	if (bp->link_info->auto_mode ==
938 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
939 		speed_capa |= ETH_LINK_SPEED_FIXED;
940 
941 	return speed_capa;
942 }
943 
944 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
945 				struct rte_eth_dev_info *dev_info)
946 {
947 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
948 	struct bnxt *bp = eth_dev->data->dev_private;
949 	uint16_t max_vnics, i, j, vpool, vrxq;
950 	unsigned int max_rx_rings;
951 	int rc;
952 
953 	rc = is_bnxt_in_error(bp);
954 	if (rc)
955 		return rc;
956 
957 	/* MAC Specifics */
958 	dev_info->max_mac_addrs = bp->max_l2_ctx;
959 	dev_info->max_hash_mac_addrs = 0;
960 
961 	/* PF/VF specifics */
962 	if (BNXT_PF(bp))
963 		dev_info->max_vfs = pdev->max_vfs;
964 
965 	max_rx_rings = bnxt_max_rings(bp);
966 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
967 	dev_info->max_rx_queues = max_rx_rings;
968 	dev_info->max_tx_queues = max_rx_rings;
969 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
970 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
971 	max_vnics = bp->max_vnics;
972 
973 	/* MTU specifics */
974 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
975 	dev_info->max_mtu = BNXT_MAX_MTU;
976 
977 	/* Fast path specifics */
978 	dev_info->min_rx_bufsize = 1;
979 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
980 
981 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
982 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
983 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
984 	if (bp->vnic_cap_flags & BNXT_VNIC_CAP_VLAN_RX_STRIP)
985 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_VLAN_STRIP;
986 	dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
987 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
988 				    dev_info->tx_queue_offload_capa;
989 	if (bp->fw_cap & BNXT_FW_CAP_VLAN_TX_INSERT)
990 		dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_VLAN_INSERT;
991 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
992 
993 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
994 	dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
995 			     RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
996 
997 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
998 		.rx_thresh = {
999 			.pthresh = 8,
1000 			.hthresh = 8,
1001 			.wthresh = 0,
1002 		},
1003 		.rx_free_thresh = 32,
1004 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
1005 	};
1006 
1007 	dev_info->default_txconf = (struct rte_eth_txconf) {
1008 		.tx_thresh = {
1009 			.pthresh = 32,
1010 			.hthresh = 0,
1011 			.wthresh = 0,
1012 		},
1013 		.tx_free_thresh = 32,
1014 		.tx_rs_thresh = 32,
1015 	};
1016 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
1017 
1018 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1019 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1020 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1021 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1022 
1023 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1024 		dev_info->switch_info.name = eth_dev->device->name;
1025 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1026 		dev_info->switch_info.port_id =
1027 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1028 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1029 	}
1030 
1031 	/*
1032 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1033 	 *       need further investigation.
1034 	 */
1035 
1036 	/* VMDq resources */
1037 	vpool = 64; /* ETH_64_POOLS */
1038 	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
1039 	for (i = 0; i < 4; vpool >>= 1, i++) {
1040 		if (max_vnics > vpool) {
1041 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1042 				if (dev_info->max_rx_queues > vrxq) {
1043 					if (vpool > vrxq)
1044 						vpool = vrxq;
1045 					goto found;
1046 				}
1047 			}
1048 			/* Not enough resources to support VMDq */
1049 			break;
1050 		}
1051 	}
1052 	/* Not enough resources to support VMDq */
1053 	vpool = 0;
1054 	vrxq = 0;
1055 found:
1056 	dev_info->max_vmdq_pools = vpool;
1057 	dev_info->vmdq_queue_num = vrxq;
1058 
1059 	dev_info->vmdq_pool_base = 0;
1060 	dev_info->vmdq_queue_base = 0;
1061 
1062 	return 0;
1063 }
1064 
1065 /* Configure the device based on the configuration provided */
1066 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1067 {
1068 	struct bnxt *bp = eth_dev->data->dev_private;
1069 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1070 	int rc;
1071 
1072 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1073 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1074 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1075 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1076 
1077 	rc = is_bnxt_in_error(bp);
1078 	if (rc)
1079 		return rc;
1080 
1081 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1082 		rc = bnxt_hwrm_check_vf_rings(bp);
1083 		if (rc) {
1084 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1085 			return -ENOSPC;
1086 		}
1087 
1088 		/* If a resource has already been allocated - in this case
1089 		 * it is the async completion ring, free it. Reallocate it after
1090 		 * resource reservation. This will ensure the resource counts
1091 		 * are calculated correctly.
1092 		 */
1093 
1094 		pthread_mutex_lock(&bp->def_cp_lock);
1095 
1096 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1097 			bnxt_disable_int(bp);
1098 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1099 		}
1100 
1101 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1102 		if (rc) {
1103 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1104 			pthread_mutex_unlock(&bp->def_cp_lock);
1105 			return -ENOSPC;
1106 		}
1107 
1108 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1109 			rc = bnxt_alloc_async_cp_ring(bp);
1110 			if (rc) {
1111 				pthread_mutex_unlock(&bp->def_cp_lock);
1112 				return rc;
1113 			}
1114 			bnxt_enable_int(bp);
1115 		}
1116 
1117 		pthread_mutex_unlock(&bp->def_cp_lock);
1118 	}
1119 
1120 	/* Inherit new configurations */
1121 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1122 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1123 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1124 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1125 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1126 	    bp->max_stat_ctx)
1127 		goto resource_error;
1128 
1129 	if (BNXT_HAS_RING_GRPS(bp) &&
1130 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1131 		goto resource_error;
1132 
1133 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1134 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1135 		goto resource_error;
1136 
1137 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1138 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1139 
1140 	if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1141 		rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1142 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1143 
1144 	if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1145 		eth_dev->data->mtu =
1146 			eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1147 			RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1148 			BNXT_NUM_VLANS;
1149 		bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1150 	}
1151 	return 0;
1152 
1153 resource_error:
1154 	PMD_DRV_LOG(ERR,
1155 		    "Insufficient resources to support requested config\n");
1156 	PMD_DRV_LOG(ERR,
1157 		    "Num Queues Requested: Tx %d, Rx %d\n",
1158 		    eth_dev->data->nb_tx_queues,
1159 		    eth_dev->data->nb_rx_queues);
1160 	PMD_DRV_LOG(ERR,
1161 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1162 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1163 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1164 	return -ENOSPC;
1165 }
1166 
1167 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1168 {
1169 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1170 
1171 	if (link->link_status)
1172 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1173 			eth_dev->data->port_id,
1174 			(uint32_t)link->link_speed,
1175 			(link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1176 			("full-duplex") : ("half-duplex\n"));
1177 	else
1178 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1179 			eth_dev->data->port_id);
1180 }
1181 
1182 /*
1183  * Determine whether the current configuration requires support for scattered
1184  * receive; return 1 if scattered receive is required and 0 if not.
1185  */
1186 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1187 {
1188 	uint16_t buf_size;
1189 	int i;
1190 
1191 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1192 		return 1;
1193 
1194 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1195 		return 1;
1196 
1197 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1198 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1199 
1200 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1201 				      RTE_PKTMBUF_HEADROOM);
1202 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1203 			return 1;
1204 	}
1205 	return 0;
1206 }
1207 
1208 static eth_rx_burst_t
1209 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1210 {
1211 	struct bnxt *bp = eth_dev->data->dev_private;
1212 
1213 	/* Disable vector mode RX for Stingray2 for now */
1214 	if (BNXT_CHIP_SR2(bp)) {
1215 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1216 		return bnxt_recv_pkts;
1217 	}
1218 
1219 #if (defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)) && \
1220 	!defined(RTE_LIBRTE_IEEE1588)
1221 
1222 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1223 	if (eth_dev->data->scattered_rx)
1224 		goto use_scalar_rx;
1225 
1226 	/*
1227 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1228 	 * asynchronous completions and receive completions can be placed in
1229 	 * the same completion ring.
1230 	 */
1231 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1232 		goto use_scalar_rx;
1233 
1234 	/*
1235 	 * Vector mode receive cannot be enabled if any receive offloads outside
1236 	 * a limited subset have been enabled.
1237 	 */
1238 	if (eth_dev->data->dev_conf.rxmode.offloads &
1239 		~(DEV_RX_OFFLOAD_VLAN_STRIP |
1240 		  DEV_RX_OFFLOAD_KEEP_CRC |
1241 		  DEV_RX_OFFLOAD_JUMBO_FRAME |
1242 		  DEV_RX_OFFLOAD_IPV4_CKSUM |
1243 		  DEV_RX_OFFLOAD_UDP_CKSUM |
1244 		  DEV_RX_OFFLOAD_TCP_CKSUM |
1245 		  DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1246 		  DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1247 		  DEV_RX_OFFLOAD_RSS_HASH |
1248 		  DEV_RX_OFFLOAD_VLAN_FILTER))
1249 		goto use_scalar_rx;
1250 
1251 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1252 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1253 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1254 		PMD_DRV_LOG(INFO,
1255 			    "Using AVX2 vector mode receive for port %d\n",
1256 			    eth_dev->data->port_id);
1257 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1258 		return bnxt_recv_pkts_vec_avx2;
1259 	}
1260  #endif
1261 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1262 		PMD_DRV_LOG(INFO,
1263 			    "Using SSE vector mode receive for port %d\n",
1264 			    eth_dev->data->port_id);
1265 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1266 		return bnxt_recv_pkts_vec;
1267 	}
1268 
1269 use_scalar_rx:
1270 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1271 		    eth_dev->data->port_id);
1272 	PMD_DRV_LOG(INFO,
1273 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1274 		    eth_dev->data->port_id,
1275 		    eth_dev->data->scattered_rx,
1276 		    eth_dev->data->dev_conf.rxmode.offloads);
1277 #endif
1278 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1279 	return bnxt_recv_pkts;
1280 }
1281 
1282 static eth_tx_burst_t
1283 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1284 {
1285 	struct bnxt *bp = eth_dev->data->dev_private;
1286 
1287 	/* Disable vector mode TX for Stingray2 for now */
1288 	if (BNXT_CHIP_SR2(bp))
1289 		return bnxt_xmit_pkts;
1290 
1291 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64) && \
1292 	!defined(RTE_LIBRTE_IEEE1588)
1293 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1294 
1295 	/*
1296 	 * Vector mode transmit can be enabled only if not using scatter rx
1297 	 * or tx offloads.
1298 	 */
1299 	if (eth_dev->data->scattered_rx ||
1300 	    (offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) ||
1301 	    BNXT_TRUFLOW_EN(bp))
1302 		goto use_scalar_tx;
1303 
1304 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1305 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1306 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1307 		PMD_DRV_LOG(INFO,
1308 			    "Using AVX2 vector mode transmit for port %d\n",
1309 			    eth_dev->data->port_id);
1310 		return bnxt_xmit_pkts_vec_avx2;
1311 	}
1312 #endif
1313 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1314 		PMD_DRV_LOG(INFO,
1315 			    "Using SSE vector mode transmit for port %d\n",
1316 			    eth_dev->data->port_id);
1317 		return bnxt_xmit_pkts_vec;
1318 	}
1319 
1320 use_scalar_tx:
1321 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1322 		    eth_dev->data->port_id);
1323 	PMD_DRV_LOG(INFO,
1324 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1325 		    eth_dev->data->port_id,
1326 		    eth_dev->data->scattered_rx,
1327 		    offloads);
1328 #endif
1329 	return bnxt_xmit_pkts;
1330 }
1331 
1332 static int bnxt_handle_if_change_status(struct bnxt *bp)
1333 {
1334 	int rc;
1335 
1336 	/* Since fw has undergone a reset and lost all contexts,
1337 	 * set fatal flag to not issue hwrm during cleanup
1338 	 */
1339 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1340 	bnxt_uninit_resources(bp, true);
1341 
1342 	/* clear fatal flag so that re-init happens */
1343 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1344 	rc = bnxt_init_resources(bp, true);
1345 
1346 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1347 
1348 	return rc;
1349 }
1350 
1351 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1352 {
1353 	struct bnxt *bp = eth_dev->data->dev_private;
1354 	int rc = 0;
1355 
1356 	if (!BNXT_SINGLE_PF(bp))
1357 		return -ENOTSUP;
1358 
1359 	if (!bp->link_info->link_up)
1360 		rc = bnxt_set_hwrm_link_config(bp, true);
1361 	if (!rc)
1362 		eth_dev->data->dev_link.link_status = 1;
1363 
1364 	bnxt_print_link_info(eth_dev);
1365 	return rc;
1366 }
1367 
1368 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1369 {
1370 	struct bnxt *bp = eth_dev->data->dev_private;
1371 
1372 	if (!BNXT_SINGLE_PF(bp))
1373 		return -ENOTSUP;
1374 
1375 	eth_dev->data->dev_link.link_status = 0;
1376 	bnxt_set_hwrm_link_config(bp, false);
1377 	bp->link_info->link_up = 0;
1378 
1379 	return 0;
1380 }
1381 
1382 static void bnxt_free_switch_domain(struct bnxt *bp)
1383 {
1384 	int rc = 0;
1385 
1386 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1387 		return;
1388 
1389 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1390 	if (rc)
1391 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1392 			    bp->switch_domain_id, rc);
1393 }
1394 
1395 static void bnxt_ptp_get_current_time(void *arg)
1396 {
1397 	struct bnxt *bp = arg;
1398 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1399 	int rc;
1400 
1401 	rc = is_bnxt_in_error(bp);
1402 	if (rc)
1403 		return;
1404 
1405 	if (!ptp)
1406 		return;
1407 
1408 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1409 				&ptp->current_time);
1410 
1411 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1412 	if (rc != 0) {
1413 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1414 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1415 	}
1416 }
1417 
1418 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1419 {
1420 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1421 	int rc;
1422 
1423 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1424 		return 0;
1425 
1426 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1427 				&ptp->current_time);
1428 
1429 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1430 	return rc;
1431 }
1432 
1433 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1434 {
1435 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1436 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1437 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1438 	}
1439 }
1440 
1441 static void bnxt_ptp_stop(struct bnxt *bp)
1442 {
1443 	bnxt_cancel_ptp_alarm(bp);
1444 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1445 }
1446 
1447 static int bnxt_ptp_start(struct bnxt *bp)
1448 {
1449 	int rc;
1450 
1451 	rc = bnxt_schedule_ptp_alarm(bp);
1452 	if (rc != 0) {
1453 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1454 	} else {
1455 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1456 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1457 	}
1458 
1459 	return rc;
1460 }
1461 
1462 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1463 {
1464 	struct bnxt *bp = eth_dev->data->dev_private;
1465 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1466 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1467 	struct rte_eth_link link;
1468 	int ret;
1469 
1470 	eth_dev->data->dev_started = 0;
1471 	eth_dev->data->scattered_rx = 0;
1472 
1473 	/* Prevent crashes when queues are still in use */
1474 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1475 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1476 
1477 	bnxt_disable_int(bp);
1478 
1479 	/* disable uio/vfio intr/eventfd mapping */
1480 	rte_intr_disable(intr_handle);
1481 
1482 	/* Stop the child representors for this device */
1483 	ret = bnxt_rep_stop_all(bp);
1484 	if (ret != 0)
1485 		return ret;
1486 
1487 	/* delete the bnxt ULP port details */
1488 	bnxt_ulp_port_deinit(bp);
1489 
1490 	bnxt_cancel_fw_health_check(bp);
1491 
1492 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1493 		bnxt_cancel_ptp_alarm(bp);
1494 
1495 	/* Do not bring link down during reset recovery */
1496 	if (!is_bnxt_in_error(bp)) {
1497 		bnxt_dev_set_link_down_op(eth_dev);
1498 		/* Wait for link to be reset */
1499 		if (BNXT_SINGLE_PF(bp))
1500 			rte_delay_ms(500);
1501 		/* clear the recorded link status */
1502 		memset(&link, 0, sizeof(link));
1503 		rte_eth_linkstatus_set(eth_dev, &link);
1504 	}
1505 
1506 	/* Clean queue intr-vector mapping */
1507 	rte_intr_efd_disable(intr_handle);
1508 	if (intr_handle->intr_vec != NULL) {
1509 		rte_free(intr_handle->intr_vec);
1510 		intr_handle->intr_vec = NULL;
1511 	}
1512 
1513 	bnxt_hwrm_port_clr_stats(bp);
1514 	bnxt_free_tx_mbufs(bp);
1515 	bnxt_free_rx_mbufs(bp);
1516 	/* Process any remaining notifications in default completion queue */
1517 	bnxt_int_handler(eth_dev);
1518 	bnxt_shutdown_nic(bp);
1519 	bnxt_hwrm_if_change(bp, false);
1520 
1521 	bnxt_free_prev_ring_stats(bp);
1522 	rte_free(bp->mark_table);
1523 	bp->mark_table = NULL;
1524 
1525 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1526 	bp->rx_cosq_cnt = 0;
1527 	/* All filters are deleted on a port stop. */
1528 	if (BNXT_FLOW_XSTATS_EN(bp))
1529 		bp->flow_stat->flow_count = 0;
1530 
1531 	return 0;
1532 }
1533 
1534 /* Unload the driver, release resources */
1535 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1536 {
1537 	struct bnxt *bp = eth_dev->data->dev_private;
1538 
1539 	pthread_mutex_lock(&bp->err_recovery_lock);
1540 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1541 		PMD_DRV_LOG(ERR,
1542 			    "Adapter recovering from error..Please retry\n");
1543 		pthread_mutex_unlock(&bp->err_recovery_lock);
1544 		return -EAGAIN;
1545 	}
1546 	pthread_mutex_unlock(&bp->err_recovery_lock);
1547 
1548 	return bnxt_dev_stop(eth_dev);
1549 }
1550 
1551 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1552 {
1553 	struct bnxt *bp = eth_dev->data->dev_private;
1554 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1555 	int vlan_mask = 0;
1556 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1557 
1558 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1559 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1560 		return -EINVAL;
1561 	}
1562 
1563 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1564 		PMD_DRV_LOG(ERR,
1565 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1566 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1567 
1568 	do {
1569 		rc = bnxt_hwrm_if_change(bp, true);
1570 		if (rc == 0 || rc != -EAGAIN)
1571 			break;
1572 
1573 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1574 	} while (retry_cnt--);
1575 
1576 	if (rc)
1577 		return rc;
1578 
1579 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1580 		rc = bnxt_handle_if_change_status(bp);
1581 		if (rc)
1582 			return rc;
1583 	}
1584 
1585 	bnxt_enable_int(bp);
1586 
1587 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1588 
1589 	rc = bnxt_start_nic(bp);
1590 	if (rc)
1591 		goto error;
1592 
1593 	rc = bnxt_alloc_prev_ring_stats(bp);
1594 	if (rc)
1595 		goto error;
1596 
1597 	eth_dev->data->dev_started = 1;
1598 
1599 	bnxt_link_update_op(eth_dev, 1);
1600 
1601 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1602 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1603 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1604 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1605 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1606 	if (rc)
1607 		goto error;
1608 
1609 	/* Initialize bnxt ULP port details */
1610 	rc = bnxt_ulp_port_init(bp);
1611 	if (rc)
1612 		goto error;
1613 
1614 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1615 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1616 
1617 	bnxt_schedule_fw_health_check(bp);
1618 
1619 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1620 		bnxt_schedule_ptp_alarm(bp);
1621 
1622 	return 0;
1623 
1624 error:
1625 	bnxt_dev_stop(eth_dev);
1626 	return rc;
1627 }
1628 
1629 static void
1630 bnxt_uninit_locks(struct bnxt *bp)
1631 {
1632 	pthread_mutex_destroy(&bp->flow_lock);
1633 	pthread_mutex_destroy(&bp->def_cp_lock);
1634 	pthread_mutex_destroy(&bp->health_check_lock);
1635 	pthread_mutex_destroy(&bp->err_recovery_lock);
1636 	if (bp->rep_info) {
1637 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1638 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1639 	}
1640 }
1641 
1642 static void bnxt_drv_uninit(struct bnxt *bp)
1643 {
1644 	bnxt_free_leds_info(bp);
1645 	bnxt_free_cos_queues(bp);
1646 	bnxt_free_link_info(bp);
1647 	bnxt_free_parent_info(bp);
1648 	bnxt_uninit_locks(bp);
1649 
1650 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1651 	bp->tx_mem_zone = NULL;
1652 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1653 	bp->rx_mem_zone = NULL;
1654 
1655 	bnxt_free_vf_info(bp);
1656 	bnxt_free_pf_info(bp);
1657 
1658 	rte_free(bp->grp_info);
1659 	bp->grp_info = NULL;
1660 }
1661 
1662 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1663 {
1664 	struct bnxt *bp = eth_dev->data->dev_private;
1665 	int ret = 0;
1666 
1667 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1668 		return 0;
1669 
1670 	pthread_mutex_lock(&bp->err_recovery_lock);
1671 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1672 		PMD_DRV_LOG(ERR,
1673 			    "Adapter recovering from error...Please retry\n");
1674 		pthread_mutex_unlock(&bp->err_recovery_lock);
1675 		return -EAGAIN;
1676 	}
1677 	pthread_mutex_unlock(&bp->err_recovery_lock);
1678 
1679 	/* cancel the recovery handler before remove dev */
1680 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1681 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1682 	bnxt_cancel_fc_thread(bp);
1683 
1684 	if (eth_dev->data->dev_started)
1685 		ret = bnxt_dev_stop(eth_dev);
1686 
1687 	bnxt_uninit_resources(bp, false);
1688 
1689 	bnxt_drv_uninit(bp);
1690 
1691 	return ret;
1692 }
1693 
1694 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1695 				    uint32_t index)
1696 {
1697 	struct bnxt *bp = eth_dev->data->dev_private;
1698 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1699 	struct bnxt_vnic_info *vnic;
1700 	struct bnxt_filter_info *filter, *temp_filter;
1701 	uint32_t i;
1702 
1703 	if (is_bnxt_in_error(bp))
1704 		return;
1705 
1706 	/*
1707 	 * Loop through all VNICs from the specified filter flow pools to
1708 	 * remove the corresponding MAC addr filter
1709 	 */
1710 	for (i = 0; i < bp->nr_vnics; i++) {
1711 		if (!(pool_mask & (1ULL << i)))
1712 			continue;
1713 
1714 		vnic = &bp->vnic_info[i];
1715 		filter = STAILQ_FIRST(&vnic->filter);
1716 		while (filter) {
1717 			temp_filter = STAILQ_NEXT(filter, next);
1718 			if (filter->mac_index == index) {
1719 				STAILQ_REMOVE(&vnic->filter, filter,
1720 						bnxt_filter_info, next);
1721 				bnxt_hwrm_clear_l2_filter(bp, filter);
1722 				bnxt_free_filter(bp, filter);
1723 			}
1724 			filter = temp_filter;
1725 		}
1726 	}
1727 }
1728 
1729 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1730 			       struct rte_ether_addr *mac_addr, uint32_t index,
1731 			       uint32_t pool)
1732 {
1733 	struct bnxt_filter_info *filter;
1734 	int rc = 0;
1735 
1736 	/* Attach requested MAC address to the new l2_filter */
1737 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1738 		if (filter->mac_index == index) {
1739 			PMD_DRV_LOG(DEBUG,
1740 				    "MAC addr already existed for pool %d\n",
1741 				    pool);
1742 			return 0;
1743 		}
1744 	}
1745 
1746 	filter = bnxt_alloc_filter(bp);
1747 	if (!filter) {
1748 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1749 		return -ENODEV;
1750 	}
1751 
1752 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1753 	 * if the MAC that's been programmed now is a different one, then,
1754 	 * copy that addr to filter->l2_addr
1755 	 */
1756 	if (mac_addr)
1757 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1758 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1759 
1760 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1761 	if (!rc) {
1762 		filter->mac_index = index;
1763 		if (filter->mac_index == 0)
1764 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1765 		else
1766 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1767 	} else {
1768 		bnxt_free_filter(bp, filter);
1769 	}
1770 
1771 	return rc;
1772 }
1773 
1774 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1775 				struct rte_ether_addr *mac_addr,
1776 				uint32_t index, uint32_t pool)
1777 {
1778 	struct bnxt *bp = eth_dev->data->dev_private;
1779 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1780 	int rc = 0;
1781 
1782 	rc = is_bnxt_in_error(bp);
1783 	if (rc)
1784 		return rc;
1785 
1786 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1787 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1788 		return -ENOTSUP;
1789 	}
1790 
1791 	if (!vnic) {
1792 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1793 		return -EINVAL;
1794 	}
1795 
1796 	/* Filter settings will get applied when port is started */
1797 	if (!eth_dev->data->dev_started)
1798 		return 0;
1799 
1800 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1801 
1802 	return rc;
1803 }
1804 
1805 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1806 {
1807 	int rc = 0;
1808 	struct bnxt *bp = eth_dev->data->dev_private;
1809 	struct rte_eth_link new;
1810 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1811 			BNXT_MIN_LINK_WAIT_CNT;
1812 
1813 	rc = is_bnxt_in_error(bp);
1814 	if (rc)
1815 		return rc;
1816 
1817 	memset(&new, 0, sizeof(new));
1818 
1819 	if (bp->link_info == NULL)
1820 		goto out;
1821 
1822 	do {
1823 		/* Retrieve link info from hardware */
1824 		rc = bnxt_get_hwrm_link_config(bp, &new);
1825 		if (rc) {
1826 			new.link_speed = ETH_LINK_SPEED_100M;
1827 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1828 			PMD_DRV_LOG(ERR,
1829 				"Failed to retrieve link rc = 0x%x!\n", rc);
1830 			goto out;
1831 		}
1832 
1833 		if (!wait_to_complete || new.link_status)
1834 			break;
1835 
1836 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1837 	} while (cnt--);
1838 
1839 	/* Only single function PF can bring phy down.
1840 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1841 	 */
1842 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1843 		memset(&new, 0, sizeof(new));
1844 
1845 out:
1846 	/* Timed out or success */
1847 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1848 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1849 		rte_eth_linkstatus_set(eth_dev, &new);
1850 		bnxt_print_link_info(eth_dev);
1851 	}
1852 
1853 	return rc;
1854 }
1855 
1856 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1857 {
1858 	struct bnxt *bp = eth_dev->data->dev_private;
1859 	struct bnxt_vnic_info *vnic;
1860 	uint32_t old_flags;
1861 	int rc;
1862 
1863 	rc = is_bnxt_in_error(bp);
1864 	if (rc)
1865 		return rc;
1866 
1867 	/* Filter settings will get applied when port is started */
1868 	if (!eth_dev->data->dev_started)
1869 		return 0;
1870 
1871 	if (bp->vnic_info == NULL)
1872 		return 0;
1873 
1874 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1875 
1876 	old_flags = vnic->flags;
1877 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1878 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1879 	if (rc != 0)
1880 		vnic->flags = old_flags;
1881 
1882 	return rc;
1883 }
1884 
1885 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1886 {
1887 	struct bnxt *bp = eth_dev->data->dev_private;
1888 	struct bnxt_vnic_info *vnic;
1889 	uint32_t old_flags;
1890 	int rc;
1891 
1892 	rc = is_bnxt_in_error(bp);
1893 	if (rc)
1894 		return rc;
1895 
1896 	/* Filter settings will get applied when port is started */
1897 	if (!eth_dev->data->dev_started)
1898 		return 0;
1899 
1900 	if (bp->vnic_info == NULL)
1901 		return 0;
1902 
1903 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1904 
1905 	old_flags = vnic->flags;
1906 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1907 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1908 	if (rc != 0)
1909 		vnic->flags = old_flags;
1910 
1911 	return rc;
1912 }
1913 
1914 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1915 {
1916 	struct bnxt *bp = eth_dev->data->dev_private;
1917 	struct bnxt_vnic_info *vnic;
1918 	uint32_t old_flags;
1919 	int rc;
1920 
1921 	rc = is_bnxt_in_error(bp);
1922 	if (rc)
1923 		return rc;
1924 
1925 	/* Filter settings will get applied when port is started */
1926 	if (!eth_dev->data->dev_started)
1927 		return 0;
1928 
1929 	if (bp->vnic_info == NULL)
1930 		return 0;
1931 
1932 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1933 
1934 	old_flags = vnic->flags;
1935 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1936 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1937 	if (rc != 0)
1938 		vnic->flags = old_flags;
1939 
1940 	return rc;
1941 }
1942 
1943 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1944 {
1945 	struct bnxt *bp = eth_dev->data->dev_private;
1946 	struct bnxt_vnic_info *vnic;
1947 	uint32_t old_flags;
1948 	int rc;
1949 
1950 	rc = is_bnxt_in_error(bp);
1951 	if (rc)
1952 		return rc;
1953 
1954 	/* Filter settings will get applied when port is started */
1955 	if (!eth_dev->data->dev_started)
1956 		return 0;
1957 
1958 	if (bp->vnic_info == NULL)
1959 		return 0;
1960 
1961 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1962 
1963 	old_flags = vnic->flags;
1964 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1965 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1966 	if (rc != 0)
1967 		vnic->flags = old_flags;
1968 
1969 	return rc;
1970 }
1971 
1972 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1973 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1974 {
1975 	if (qid >= bp->rx_nr_rings)
1976 		return NULL;
1977 
1978 	return bp->eth_dev->data->rx_queues[qid];
1979 }
1980 
1981 /* Return rxq corresponding to a given rss table ring/group ID. */
1982 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1983 {
1984 	struct bnxt_rx_queue *rxq;
1985 	unsigned int i;
1986 
1987 	if (!BNXT_HAS_RING_GRPS(bp)) {
1988 		for (i = 0; i < bp->rx_nr_rings; i++) {
1989 			rxq = bp->eth_dev->data->rx_queues[i];
1990 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1991 				return rxq->index;
1992 		}
1993 	} else {
1994 		for (i = 0; i < bp->rx_nr_rings; i++) {
1995 			if (bp->grp_info[i].fw_grp_id == fwr)
1996 				return i;
1997 		}
1998 	}
1999 
2000 	return INVALID_HW_RING_ID;
2001 }
2002 
2003 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
2004 			    struct rte_eth_rss_reta_entry64 *reta_conf,
2005 			    uint16_t reta_size)
2006 {
2007 	struct bnxt *bp = eth_dev->data->dev_private;
2008 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2009 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2010 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2011 	uint16_t idx, sft;
2012 	int i, rc;
2013 
2014 	rc = is_bnxt_in_error(bp);
2015 	if (rc)
2016 		return rc;
2017 
2018 	if (!vnic->rss_table)
2019 		return -EINVAL;
2020 
2021 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
2022 		return -EINVAL;
2023 
2024 	if (reta_size != tbl_size) {
2025 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2026 			"(%d) must equal the size supported by the hardware "
2027 			"(%d)\n", reta_size, tbl_size);
2028 		return -EINVAL;
2029 	}
2030 
2031 	for (i = 0; i < reta_size; i++) {
2032 		struct bnxt_rx_queue *rxq;
2033 
2034 		idx = i / RTE_RETA_GROUP_SIZE;
2035 		sft = i % RTE_RETA_GROUP_SIZE;
2036 
2037 		if (!(reta_conf[idx].mask & (1ULL << sft)))
2038 			continue;
2039 
2040 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2041 		if (!rxq) {
2042 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
2043 			return -EINVAL;
2044 		}
2045 
2046 		if (BNXT_CHIP_P5(bp)) {
2047 			vnic->rss_table[i * 2] =
2048 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
2049 			vnic->rss_table[i * 2 + 1] =
2050 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
2051 		} else {
2052 			vnic->rss_table[i] =
2053 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2054 		}
2055 	}
2056 
2057 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2058 	return rc;
2059 }
2060 
2061 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2062 			      struct rte_eth_rss_reta_entry64 *reta_conf,
2063 			      uint16_t reta_size)
2064 {
2065 	struct bnxt *bp = eth_dev->data->dev_private;
2066 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2067 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2068 	uint16_t idx, sft, i;
2069 	int rc;
2070 
2071 	rc = is_bnxt_in_error(bp);
2072 	if (rc)
2073 		return rc;
2074 
2075 	if (!vnic)
2076 		return -EINVAL;
2077 	if (!vnic->rss_table)
2078 		return -EINVAL;
2079 
2080 	if (reta_size != tbl_size) {
2081 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2082 			"(%d) must equal the size supported by the hardware "
2083 			"(%d)\n", reta_size, tbl_size);
2084 		return -EINVAL;
2085 	}
2086 
2087 	for (idx = 0, i = 0; i < reta_size; i++) {
2088 		idx = i / RTE_RETA_GROUP_SIZE;
2089 		sft = i % RTE_RETA_GROUP_SIZE;
2090 
2091 		if (reta_conf[idx].mask & (1ULL << sft)) {
2092 			uint16_t qid;
2093 
2094 			if (BNXT_CHIP_P5(bp))
2095 				qid = bnxt_rss_to_qid(bp,
2096 						      vnic->rss_table[i * 2]);
2097 			else
2098 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2099 
2100 			if (qid == INVALID_HW_RING_ID) {
2101 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2102 				return -EINVAL;
2103 			}
2104 			reta_conf[idx].reta[sft] = qid;
2105 		}
2106 	}
2107 
2108 	return 0;
2109 }
2110 
2111 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2112 				   struct rte_eth_rss_conf *rss_conf)
2113 {
2114 	struct bnxt *bp = eth_dev->data->dev_private;
2115 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2116 	struct bnxt_vnic_info *vnic;
2117 	int rc;
2118 
2119 	rc = is_bnxt_in_error(bp);
2120 	if (rc)
2121 		return rc;
2122 
2123 	/*
2124 	 * If RSS enablement were different than dev_configure,
2125 	 * then return -EINVAL
2126 	 */
2127 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
2128 		if (!rss_conf->rss_hf)
2129 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2130 	} else {
2131 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2132 			return -EINVAL;
2133 	}
2134 
2135 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
2136 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2137 	       rss_conf,
2138 	       sizeof(*rss_conf));
2139 
2140 	/* Update the default RSS VNIC(s) */
2141 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2142 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2143 	vnic->hash_mode =
2144 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2145 					    ETH_RSS_LEVEL(rss_conf->rss_hf));
2146 
2147 	/*
2148 	 * If hashkey is not specified, use the previously configured
2149 	 * hashkey
2150 	 */
2151 	if (!rss_conf->rss_key)
2152 		goto rss_config;
2153 
2154 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2155 		PMD_DRV_LOG(ERR,
2156 			    "Invalid hashkey length, should be %d bytes\n",
2157 			    HW_HASH_KEY_SIZE);
2158 		return -EINVAL;
2159 	}
2160 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2161 
2162 rss_config:
2163 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2164 	return rc;
2165 }
2166 
2167 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2168 				     struct rte_eth_rss_conf *rss_conf)
2169 {
2170 	struct bnxt *bp = eth_dev->data->dev_private;
2171 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2172 	int len, rc;
2173 	uint32_t hash_types;
2174 
2175 	rc = is_bnxt_in_error(bp);
2176 	if (rc)
2177 		return rc;
2178 
2179 	/* RSS configuration is the same for all VNICs */
2180 	if (vnic && vnic->rss_hash_key) {
2181 		if (rss_conf->rss_key) {
2182 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2183 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2184 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2185 		}
2186 
2187 		hash_types = vnic->hash_type;
2188 		rss_conf->rss_hf = 0;
2189 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2190 			rss_conf->rss_hf |= ETH_RSS_IPV4;
2191 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2192 		}
2193 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2194 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2195 			hash_types &=
2196 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2197 		}
2198 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2199 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2200 			hash_types &=
2201 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2202 		}
2203 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2204 			rss_conf->rss_hf |= ETH_RSS_IPV6;
2205 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2206 		}
2207 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2208 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2209 			hash_types &=
2210 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2211 		}
2212 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2213 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2214 			hash_types &=
2215 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2216 		}
2217 
2218 		rss_conf->rss_hf |=
2219 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2220 
2221 		if (hash_types) {
2222 			PMD_DRV_LOG(ERR,
2223 				"Unknown RSS config from firmware (%08x), RSS disabled",
2224 				vnic->hash_type);
2225 			return -ENOTSUP;
2226 		}
2227 	} else {
2228 		rss_conf->rss_hf = 0;
2229 	}
2230 	return 0;
2231 }
2232 
2233 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2234 			       struct rte_eth_fc_conf *fc_conf)
2235 {
2236 	struct bnxt *bp = dev->data->dev_private;
2237 	struct rte_eth_link link_info;
2238 	int rc;
2239 
2240 	rc = is_bnxt_in_error(bp);
2241 	if (rc)
2242 		return rc;
2243 
2244 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2245 	if (rc)
2246 		return rc;
2247 
2248 	memset(fc_conf, 0, sizeof(*fc_conf));
2249 	if (bp->link_info->auto_pause)
2250 		fc_conf->autoneg = 1;
2251 	switch (bp->link_info->pause) {
2252 	case 0:
2253 		fc_conf->mode = RTE_FC_NONE;
2254 		break;
2255 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2256 		fc_conf->mode = RTE_FC_TX_PAUSE;
2257 		break;
2258 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2259 		fc_conf->mode = RTE_FC_RX_PAUSE;
2260 		break;
2261 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2262 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2263 		fc_conf->mode = RTE_FC_FULL;
2264 		break;
2265 	}
2266 	return 0;
2267 }
2268 
2269 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2270 			       struct rte_eth_fc_conf *fc_conf)
2271 {
2272 	struct bnxt *bp = dev->data->dev_private;
2273 	int rc;
2274 
2275 	rc = is_bnxt_in_error(bp);
2276 	if (rc)
2277 		return rc;
2278 
2279 	if (!BNXT_SINGLE_PF(bp)) {
2280 		PMD_DRV_LOG(ERR,
2281 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2282 		return -ENOTSUP;
2283 	}
2284 
2285 	switch (fc_conf->mode) {
2286 	case RTE_FC_NONE:
2287 		bp->link_info->auto_pause = 0;
2288 		bp->link_info->force_pause = 0;
2289 		break;
2290 	case RTE_FC_RX_PAUSE:
2291 		if (fc_conf->autoneg) {
2292 			bp->link_info->auto_pause =
2293 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2294 			bp->link_info->force_pause = 0;
2295 		} else {
2296 			bp->link_info->auto_pause = 0;
2297 			bp->link_info->force_pause =
2298 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2299 		}
2300 		break;
2301 	case RTE_FC_TX_PAUSE:
2302 		if (fc_conf->autoneg) {
2303 			bp->link_info->auto_pause =
2304 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2305 			bp->link_info->force_pause = 0;
2306 		} else {
2307 			bp->link_info->auto_pause = 0;
2308 			bp->link_info->force_pause =
2309 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2310 		}
2311 		break;
2312 	case RTE_FC_FULL:
2313 		if (fc_conf->autoneg) {
2314 			bp->link_info->auto_pause =
2315 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2316 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2317 			bp->link_info->force_pause = 0;
2318 		} else {
2319 			bp->link_info->auto_pause = 0;
2320 			bp->link_info->force_pause =
2321 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2322 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2323 		}
2324 		break;
2325 	}
2326 	return bnxt_set_hwrm_link_config(bp, true);
2327 }
2328 
2329 /* Add UDP tunneling port */
2330 static int
2331 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2332 			 struct rte_eth_udp_tunnel *udp_tunnel)
2333 {
2334 	struct bnxt *bp = eth_dev->data->dev_private;
2335 	uint16_t tunnel_type = 0;
2336 	int rc = 0;
2337 
2338 	rc = is_bnxt_in_error(bp);
2339 	if (rc)
2340 		return rc;
2341 
2342 	switch (udp_tunnel->prot_type) {
2343 	case RTE_TUNNEL_TYPE_VXLAN:
2344 		if (bp->vxlan_port_cnt) {
2345 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2346 				udp_tunnel->udp_port);
2347 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2348 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2349 				return -ENOSPC;
2350 			}
2351 			bp->vxlan_port_cnt++;
2352 			return 0;
2353 		}
2354 		tunnel_type =
2355 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2356 		bp->vxlan_port_cnt++;
2357 		break;
2358 	case RTE_TUNNEL_TYPE_GENEVE:
2359 		if (bp->geneve_port_cnt) {
2360 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2361 				udp_tunnel->udp_port);
2362 			if (bp->geneve_port != udp_tunnel->udp_port) {
2363 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2364 				return -ENOSPC;
2365 			}
2366 			bp->geneve_port_cnt++;
2367 			return 0;
2368 		}
2369 		tunnel_type =
2370 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2371 		bp->geneve_port_cnt++;
2372 		break;
2373 	default:
2374 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2375 		return -ENOTSUP;
2376 	}
2377 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2378 					     tunnel_type);
2379 	return rc;
2380 }
2381 
2382 static int
2383 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2384 			 struct rte_eth_udp_tunnel *udp_tunnel)
2385 {
2386 	struct bnxt *bp = eth_dev->data->dev_private;
2387 	uint16_t tunnel_type = 0;
2388 	uint16_t port = 0;
2389 	int rc = 0;
2390 
2391 	rc = is_bnxt_in_error(bp);
2392 	if (rc)
2393 		return rc;
2394 
2395 	switch (udp_tunnel->prot_type) {
2396 	case RTE_TUNNEL_TYPE_VXLAN:
2397 		if (!bp->vxlan_port_cnt) {
2398 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2399 			return -EINVAL;
2400 		}
2401 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2402 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2403 				udp_tunnel->udp_port, bp->vxlan_port);
2404 			return -EINVAL;
2405 		}
2406 		if (--bp->vxlan_port_cnt)
2407 			return 0;
2408 
2409 		tunnel_type =
2410 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2411 		port = bp->vxlan_fw_dst_port_id;
2412 		break;
2413 	case RTE_TUNNEL_TYPE_GENEVE:
2414 		if (!bp->geneve_port_cnt) {
2415 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2416 			return -EINVAL;
2417 		}
2418 		if (bp->geneve_port != udp_tunnel->udp_port) {
2419 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2420 				udp_tunnel->udp_port, bp->geneve_port);
2421 			return -EINVAL;
2422 		}
2423 		if (--bp->geneve_port_cnt)
2424 			return 0;
2425 
2426 		tunnel_type =
2427 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2428 		port = bp->geneve_fw_dst_port_id;
2429 		break;
2430 	default:
2431 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2432 		return -ENOTSUP;
2433 	}
2434 
2435 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2436 	return rc;
2437 }
2438 
2439 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2440 {
2441 	struct bnxt_filter_info *filter;
2442 	struct bnxt_vnic_info *vnic;
2443 	int rc = 0;
2444 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2445 
2446 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2447 	filter = STAILQ_FIRST(&vnic->filter);
2448 	while (filter) {
2449 		/* Search for this matching MAC+VLAN filter */
2450 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2451 			/* Delete the filter */
2452 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2453 			if (rc)
2454 				return rc;
2455 			STAILQ_REMOVE(&vnic->filter, filter,
2456 				      bnxt_filter_info, next);
2457 			bnxt_free_filter(bp, filter);
2458 			PMD_DRV_LOG(INFO,
2459 				    "Deleted vlan filter for %d\n",
2460 				    vlan_id);
2461 			return 0;
2462 		}
2463 		filter = STAILQ_NEXT(filter, next);
2464 	}
2465 	return -ENOENT;
2466 }
2467 
2468 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2469 {
2470 	struct bnxt_filter_info *filter;
2471 	struct bnxt_vnic_info *vnic;
2472 	int rc = 0;
2473 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2474 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2475 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2476 
2477 	/* Implementation notes on the use of VNIC in this command:
2478 	 *
2479 	 * By default, these filters belong to default vnic for the function.
2480 	 * Once these filters are set up, only destination VNIC can be modified.
2481 	 * If the destination VNIC is not specified in this command,
2482 	 * then the HWRM shall only create an l2 context id.
2483 	 */
2484 
2485 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2486 	filter = STAILQ_FIRST(&vnic->filter);
2487 	/* Check if the VLAN has already been added */
2488 	while (filter) {
2489 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2490 			return -EEXIST;
2491 
2492 		filter = STAILQ_NEXT(filter, next);
2493 	}
2494 
2495 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2496 	 * command to create MAC+VLAN filter with the right flags, enables set.
2497 	 */
2498 	filter = bnxt_alloc_filter(bp);
2499 	if (!filter) {
2500 		PMD_DRV_LOG(ERR,
2501 			    "MAC/VLAN filter alloc failed\n");
2502 		return -ENOMEM;
2503 	}
2504 	/* MAC + VLAN ID filter */
2505 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2506 	 * untagged packets are received
2507 	 *
2508 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2509 	 * packets and only the programmed vlan's packets are received
2510 	 */
2511 	filter->l2_ivlan = vlan_id;
2512 	filter->l2_ivlan_mask = 0x0FFF;
2513 	filter->enables |= en;
2514 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2515 
2516 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2517 	if (rc) {
2518 		/* Free the newly allocated filter as we were
2519 		 * not able to create the filter in hardware.
2520 		 */
2521 		bnxt_free_filter(bp, filter);
2522 		return rc;
2523 	}
2524 
2525 	filter->mac_index = 0;
2526 	/* Add this new filter to the list */
2527 	if (vlan_id == 0)
2528 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2529 	else
2530 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2531 
2532 	PMD_DRV_LOG(INFO,
2533 		    "Added Vlan filter for %d\n", vlan_id);
2534 	return rc;
2535 }
2536 
2537 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2538 		uint16_t vlan_id, int on)
2539 {
2540 	struct bnxt *bp = eth_dev->data->dev_private;
2541 	int rc;
2542 
2543 	rc = is_bnxt_in_error(bp);
2544 	if (rc)
2545 		return rc;
2546 
2547 	if (!eth_dev->data->dev_started) {
2548 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2549 		return -EINVAL;
2550 	}
2551 
2552 	/* These operations apply to ALL existing MAC/VLAN filters */
2553 	if (on)
2554 		return bnxt_add_vlan_filter(bp, vlan_id);
2555 	else
2556 		return bnxt_del_vlan_filter(bp, vlan_id);
2557 }
2558 
2559 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2560 				    struct bnxt_vnic_info *vnic)
2561 {
2562 	struct bnxt_filter_info *filter;
2563 	int rc;
2564 
2565 	filter = STAILQ_FIRST(&vnic->filter);
2566 	while (filter) {
2567 		if (filter->mac_index == 0 &&
2568 		    !memcmp(filter->l2_addr, bp->mac_addr,
2569 			    RTE_ETHER_ADDR_LEN)) {
2570 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2571 			if (!rc) {
2572 				STAILQ_REMOVE(&vnic->filter, filter,
2573 					      bnxt_filter_info, next);
2574 				bnxt_free_filter(bp, filter);
2575 			}
2576 			return rc;
2577 		}
2578 		filter = STAILQ_NEXT(filter, next);
2579 	}
2580 	return 0;
2581 }
2582 
2583 static int
2584 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2585 {
2586 	struct bnxt_vnic_info *vnic;
2587 	unsigned int i;
2588 	int rc;
2589 
2590 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2591 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2592 		/* Remove any VLAN filters programmed */
2593 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2594 			bnxt_del_vlan_filter(bp, i);
2595 
2596 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2597 		if (rc)
2598 			return rc;
2599 	} else {
2600 		/* Default filter will allow packets that match the
2601 		 * dest mac. So, it has to be deleted, otherwise, we
2602 		 * will endup receiving vlan packets for which the
2603 		 * filter is not programmed, when hw-vlan-filter
2604 		 * configuration is ON
2605 		 */
2606 		bnxt_del_dflt_mac_filter(bp, vnic);
2607 		/* This filter will allow only untagged packets */
2608 		bnxt_add_vlan_filter(bp, 0);
2609 	}
2610 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2611 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2612 
2613 	return 0;
2614 }
2615 
2616 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2617 {
2618 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2619 	unsigned int i;
2620 	int rc;
2621 
2622 	/* Destroy vnic filters and vnic */
2623 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2624 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2625 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2626 			bnxt_del_vlan_filter(bp, i);
2627 	}
2628 	bnxt_del_dflt_mac_filter(bp, vnic);
2629 
2630 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2631 	if (rc)
2632 		return rc;
2633 
2634 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2635 	if (rc)
2636 		return rc;
2637 
2638 	rte_free(vnic->fw_grp_ids);
2639 	vnic->fw_grp_ids = NULL;
2640 
2641 	vnic->rx_queue_cnt = 0;
2642 
2643 	return 0;
2644 }
2645 
2646 static int
2647 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2648 {
2649 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2650 	int rc;
2651 
2652 	/* Destroy, recreate and reconfigure the default vnic */
2653 	rc = bnxt_free_one_vnic(bp, 0);
2654 	if (rc)
2655 		return rc;
2656 
2657 	/* default vnic 0 */
2658 	rc = bnxt_setup_one_vnic(bp, 0);
2659 	if (rc)
2660 		return rc;
2661 
2662 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2663 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2664 		rc = bnxt_add_vlan_filter(bp, 0);
2665 		if (rc)
2666 			return rc;
2667 		rc = bnxt_restore_vlan_filters(bp);
2668 		if (rc)
2669 			return rc;
2670 	} else {
2671 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2672 		if (rc)
2673 			return rc;
2674 	}
2675 
2676 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2677 	if (rc)
2678 		return rc;
2679 
2680 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2681 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2682 
2683 	return rc;
2684 }
2685 
2686 static int
2687 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2688 {
2689 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2690 	struct bnxt *bp = dev->data->dev_private;
2691 	int rc;
2692 
2693 	rc = is_bnxt_in_error(bp);
2694 	if (rc)
2695 		return rc;
2696 
2697 	/* Filter settings will get applied when port is started */
2698 	if (!dev->data->dev_started)
2699 		return 0;
2700 
2701 	if (mask & ETH_VLAN_FILTER_MASK) {
2702 		/* Enable or disable VLAN filtering */
2703 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2704 		if (rc)
2705 			return rc;
2706 	}
2707 
2708 	if (mask & ETH_VLAN_STRIP_MASK) {
2709 		/* Enable or disable VLAN stripping */
2710 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2711 		if (rc)
2712 			return rc;
2713 	}
2714 
2715 	if (mask & ETH_VLAN_EXTEND_MASK) {
2716 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2717 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2718 		else
2719 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2720 	}
2721 
2722 	return 0;
2723 }
2724 
2725 static int
2726 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2727 		      uint16_t tpid)
2728 {
2729 	struct bnxt *bp = dev->data->dev_private;
2730 	int qinq = dev->data->dev_conf.rxmode.offloads &
2731 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2732 
2733 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2734 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2735 		PMD_DRV_LOG(ERR,
2736 			    "Unsupported vlan type.");
2737 		return -EINVAL;
2738 	}
2739 	if (!qinq) {
2740 		PMD_DRV_LOG(ERR,
2741 			    "QinQ not enabled. Needs to be ON as we can "
2742 			    "accelerate only outer vlan\n");
2743 		return -EINVAL;
2744 	}
2745 
2746 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2747 		switch (tpid) {
2748 		case RTE_ETHER_TYPE_QINQ:
2749 			bp->outer_tpid_bd =
2750 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2751 				break;
2752 		case RTE_ETHER_TYPE_VLAN:
2753 			bp->outer_tpid_bd =
2754 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2755 				break;
2756 		case RTE_ETHER_TYPE_QINQ1:
2757 			bp->outer_tpid_bd =
2758 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2759 				break;
2760 		case RTE_ETHER_TYPE_QINQ2:
2761 			bp->outer_tpid_bd =
2762 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2763 				break;
2764 		case RTE_ETHER_TYPE_QINQ3:
2765 			bp->outer_tpid_bd =
2766 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2767 				break;
2768 		default:
2769 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2770 			return -EINVAL;
2771 		}
2772 		bp->outer_tpid_bd |= tpid;
2773 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2774 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2775 		PMD_DRV_LOG(ERR,
2776 			    "Can accelerate only outer vlan in QinQ\n");
2777 		return -EINVAL;
2778 	}
2779 
2780 	return 0;
2781 }
2782 
2783 static int
2784 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2785 			     struct rte_ether_addr *addr)
2786 {
2787 	struct bnxt *bp = dev->data->dev_private;
2788 	/* Default Filter is tied to VNIC 0 */
2789 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2790 	int rc;
2791 
2792 	rc = is_bnxt_in_error(bp);
2793 	if (rc)
2794 		return rc;
2795 
2796 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2797 		return -EPERM;
2798 
2799 	if (rte_is_zero_ether_addr(addr))
2800 		return -EINVAL;
2801 
2802 	/* Filter settings will get applied when port is started */
2803 	if (!dev->data->dev_started)
2804 		return 0;
2805 
2806 	/* Check if the requested MAC is already added */
2807 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2808 		return 0;
2809 
2810 	/* Destroy filter and re-create it */
2811 	bnxt_del_dflt_mac_filter(bp, vnic);
2812 
2813 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2814 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2815 		/* This filter will allow only untagged packets */
2816 		rc = bnxt_add_vlan_filter(bp, 0);
2817 	} else {
2818 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2819 	}
2820 
2821 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2822 	return rc;
2823 }
2824 
2825 static int
2826 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2827 			  struct rte_ether_addr *mc_addr_set,
2828 			  uint32_t nb_mc_addr)
2829 {
2830 	struct bnxt *bp = eth_dev->data->dev_private;
2831 	char *mc_addr_list = (char *)mc_addr_set;
2832 	struct bnxt_vnic_info *vnic;
2833 	uint32_t off = 0, i = 0;
2834 	int rc;
2835 
2836 	rc = is_bnxt_in_error(bp);
2837 	if (rc)
2838 		return rc;
2839 
2840 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2841 
2842 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2843 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2844 		goto allmulti;
2845 	}
2846 
2847 	/* TODO Check for Duplicate mcast addresses */
2848 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2849 	for (i = 0; i < nb_mc_addr; i++) {
2850 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2851 			RTE_ETHER_ADDR_LEN);
2852 		off += RTE_ETHER_ADDR_LEN;
2853 	}
2854 
2855 	vnic->mc_addr_cnt = i;
2856 	if (vnic->mc_addr_cnt)
2857 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2858 	else
2859 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2860 
2861 allmulti:
2862 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2863 }
2864 
2865 static int
2866 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2867 {
2868 	struct bnxt *bp = dev->data->dev_private;
2869 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2870 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2871 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2872 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2873 	int ret;
2874 
2875 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2876 			fw_major, fw_minor, fw_updt, fw_rsvd);
2877 	if (ret < 0)
2878 		return -EINVAL;
2879 
2880 	ret += 1; /* add the size of '\0' */
2881 	if (fw_size < (size_t)ret)
2882 		return ret;
2883 	else
2884 		return 0;
2885 }
2886 
2887 static void
2888 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2889 	struct rte_eth_rxq_info *qinfo)
2890 {
2891 	struct bnxt *bp = dev->data->dev_private;
2892 	struct bnxt_rx_queue *rxq;
2893 
2894 	if (is_bnxt_in_error(bp))
2895 		return;
2896 
2897 	rxq = dev->data->rx_queues[queue_id];
2898 
2899 	qinfo->mp = rxq->mb_pool;
2900 	qinfo->scattered_rx = dev->data->scattered_rx;
2901 	qinfo->nb_desc = rxq->nb_rx_desc;
2902 
2903 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2904 	qinfo->conf.rx_drop_en = rxq->drop_en;
2905 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2906 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2907 }
2908 
2909 static void
2910 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2911 	struct rte_eth_txq_info *qinfo)
2912 {
2913 	struct bnxt *bp = dev->data->dev_private;
2914 	struct bnxt_tx_queue *txq;
2915 
2916 	if (is_bnxt_in_error(bp))
2917 		return;
2918 
2919 	txq = dev->data->tx_queues[queue_id];
2920 
2921 	qinfo->nb_desc = txq->nb_tx_desc;
2922 
2923 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2924 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2925 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2926 
2927 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2928 	qinfo->conf.tx_rs_thresh = 0;
2929 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2930 	qinfo->conf.offloads = txq->offloads;
2931 }
2932 
2933 static const struct {
2934 	eth_rx_burst_t pkt_burst;
2935 	const char *info;
2936 } bnxt_rx_burst_info[] = {
2937 	{bnxt_recv_pkts,		"Scalar"},
2938 #if defined(RTE_ARCH_X86)
2939 	{bnxt_recv_pkts_vec,		"Vector SSE"},
2940 #endif
2941 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2942 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
2943 #endif
2944 #if defined(RTE_ARCH_ARM64)
2945 	{bnxt_recv_pkts_vec,		"Vector Neon"},
2946 #endif
2947 };
2948 
2949 static int
2950 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2951 		       struct rte_eth_burst_mode *mode)
2952 {
2953 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2954 	size_t i;
2955 
2956 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2957 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2958 			snprintf(mode->info, sizeof(mode->info), "%s",
2959 				 bnxt_rx_burst_info[i].info);
2960 			return 0;
2961 		}
2962 	}
2963 
2964 	return -EINVAL;
2965 }
2966 
2967 static const struct {
2968 	eth_tx_burst_t pkt_burst;
2969 	const char *info;
2970 } bnxt_tx_burst_info[] = {
2971 	{bnxt_xmit_pkts,		"Scalar"},
2972 #if defined(RTE_ARCH_X86)
2973 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
2974 #endif
2975 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2976 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
2977 #endif
2978 #if defined(RTE_ARCH_ARM64)
2979 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
2980 #endif
2981 };
2982 
2983 static int
2984 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2985 		       struct rte_eth_burst_mode *mode)
2986 {
2987 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2988 	size_t i;
2989 
2990 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2991 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2992 			snprintf(mode->info, sizeof(mode->info), "%s",
2993 				 bnxt_tx_burst_info[i].info);
2994 			return 0;
2995 		}
2996 	}
2997 
2998 	return -EINVAL;
2999 }
3000 
3001 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3002 {
3003 	struct bnxt *bp = eth_dev->data->dev_private;
3004 	uint32_t new_pkt_size;
3005 	uint32_t rc = 0;
3006 	uint32_t i;
3007 
3008 	rc = is_bnxt_in_error(bp);
3009 	if (rc)
3010 		return rc;
3011 
3012 	/* Exit if receive queues are not configured yet */
3013 	if (!eth_dev->data->nb_rx_queues)
3014 		return rc;
3015 
3016 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
3017 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
3018 
3019 	/*
3020 	 * Disallow any MTU change that would require scattered receive support
3021 	 * if it is not already enabled.
3022 	 */
3023 	if (eth_dev->data->dev_started &&
3024 	    !eth_dev->data->scattered_rx &&
3025 	    (new_pkt_size >
3026 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
3027 		PMD_DRV_LOG(ERR,
3028 			    "MTU change would require scattered rx support. ");
3029 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
3030 		return -EINVAL;
3031 	}
3032 
3033 	if (new_mtu > RTE_ETHER_MTU) {
3034 		bp->flags |= BNXT_FLAG_JUMBO;
3035 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
3036 			DEV_RX_OFFLOAD_JUMBO_FRAME;
3037 	} else {
3038 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
3039 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
3040 		bp->flags &= ~BNXT_FLAG_JUMBO;
3041 	}
3042 
3043 	/* Is there a change in mtu setting? */
3044 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
3045 		return rc;
3046 
3047 	for (i = 0; i < bp->nr_vnics; i++) {
3048 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3049 		uint16_t size = 0;
3050 
3051 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
3052 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
3053 		if (rc)
3054 			break;
3055 
3056 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
3057 		size -= RTE_PKTMBUF_HEADROOM;
3058 
3059 		if (size < new_mtu) {
3060 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
3061 			if (rc)
3062 				return rc;
3063 		}
3064 	}
3065 
3066 	if (!rc)
3067 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
3068 
3069 	if (bnxt_hwrm_config_host_mtu(bp))
3070 		PMD_DRV_LOG(WARNING, "Failed to configure host MTU\n");
3071 
3072 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3073 
3074 	return rc;
3075 }
3076 
3077 static int
3078 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3079 {
3080 	struct bnxt *bp = dev->data->dev_private;
3081 	uint16_t vlan = bp->vlan;
3082 	int rc;
3083 
3084 	rc = is_bnxt_in_error(bp);
3085 	if (rc)
3086 		return rc;
3087 
3088 	if (!BNXT_SINGLE_PF(bp)) {
3089 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3090 		return -ENOTSUP;
3091 	}
3092 	bp->vlan = on ? pvid : 0;
3093 
3094 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3095 	if (rc)
3096 		bp->vlan = vlan;
3097 	return rc;
3098 }
3099 
3100 static int
3101 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3102 {
3103 	struct bnxt *bp = dev->data->dev_private;
3104 	int rc;
3105 
3106 	rc = is_bnxt_in_error(bp);
3107 	if (rc)
3108 		return rc;
3109 
3110 	return bnxt_hwrm_port_led_cfg(bp, true);
3111 }
3112 
3113 static int
3114 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3115 {
3116 	struct bnxt *bp = dev->data->dev_private;
3117 	int rc;
3118 
3119 	rc = is_bnxt_in_error(bp);
3120 	if (rc)
3121 		return rc;
3122 
3123 	return bnxt_hwrm_port_led_cfg(bp, false);
3124 }
3125 
3126 static uint32_t
3127 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
3128 {
3129 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
3130 	struct bnxt_cp_ring_info *cpr;
3131 	uint32_t desc = 0, raw_cons, cp_ring_size;
3132 	struct bnxt_rx_queue *rxq;
3133 	struct rx_pkt_cmpl *rxcmp;
3134 	int rc;
3135 
3136 	rc = is_bnxt_in_error(bp);
3137 	if (rc)
3138 		return rc;
3139 
3140 	rxq = dev->data->rx_queues[rx_queue_id];
3141 	cpr = rxq->cp_ring;
3142 	raw_cons = cpr->cp_raw_cons;
3143 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3144 
3145 	while (1) {
3146 		uint32_t agg_cnt, cons, cmpl_type;
3147 
3148 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3149 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3150 
3151 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3152 			break;
3153 
3154 		cmpl_type = CMP_TYPE(rxcmp);
3155 
3156 		switch (cmpl_type) {
3157 		case CMPL_BASE_TYPE_RX_L2:
3158 		case CMPL_BASE_TYPE_RX_L2_V2:
3159 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3160 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3161 			desc++;
3162 			break;
3163 
3164 		case CMPL_BASE_TYPE_RX_TPA_END:
3165 			if (BNXT_CHIP_P5(rxq->bp)) {
3166 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3167 
3168 				p5_tpa_end = (void *)rxcmp;
3169 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3170 			} else {
3171 				struct rx_tpa_end_cmpl *tpa_end;
3172 
3173 				tpa_end = (void *)rxcmp;
3174 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3175 			}
3176 
3177 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3178 			desc++;
3179 			break;
3180 
3181 		default:
3182 			raw_cons += CMP_LEN(cmpl_type);
3183 		}
3184 	}
3185 
3186 	return desc;
3187 }
3188 
3189 static int
3190 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3191 {
3192 	struct bnxt_rx_queue *rxq = rx_queue;
3193 	struct bnxt_cp_ring_info *cpr;
3194 	struct bnxt_rx_ring_info *rxr;
3195 	uint32_t desc, raw_cons, cp_ring_size;
3196 	struct bnxt *bp = rxq->bp;
3197 	struct rx_pkt_cmpl *rxcmp;
3198 	int rc;
3199 
3200 	rc = is_bnxt_in_error(bp);
3201 	if (rc)
3202 		return rc;
3203 
3204 	if (offset >= rxq->nb_rx_desc)
3205 		return -EINVAL;
3206 
3207 	rxr = rxq->rx_ring;
3208 	cpr = rxq->cp_ring;
3209 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3210 
3211 	/*
3212 	 * For the vector receive case, the completion at the requested
3213 	 * offset can be indexed directly.
3214 	 */
3215 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3216 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3217 		struct rx_pkt_cmpl *rxcmp;
3218 		uint32_t cons;
3219 
3220 		/* Check status of completion descriptor. */
3221 		raw_cons = cpr->cp_raw_cons +
3222 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3223 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3224 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3225 
3226 		if (bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3227 			return RTE_ETH_RX_DESC_DONE;
3228 
3229 		/* Check whether rx desc has an mbuf attached. */
3230 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3231 		if (cons >= rxq->rxrearm_start &&
3232 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3233 			return RTE_ETH_RX_DESC_UNAVAIL;
3234 		}
3235 
3236 		return RTE_ETH_RX_DESC_AVAIL;
3237 	}
3238 #endif
3239 
3240 	/*
3241 	 * For the non-vector receive case, scan the completion ring to
3242 	 * locate the completion descriptor for the requested offset.
3243 	 */
3244 	raw_cons = cpr->cp_raw_cons;
3245 	desc = 0;
3246 	while (1) {
3247 		uint32_t agg_cnt, cons, cmpl_type;
3248 
3249 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3250 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3251 
3252 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3253 			break;
3254 
3255 		cmpl_type = CMP_TYPE(rxcmp);
3256 
3257 		switch (cmpl_type) {
3258 		case CMPL_BASE_TYPE_RX_L2:
3259 		case CMPL_BASE_TYPE_RX_L2_V2:
3260 			if (desc == offset) {
3261 				cons = rxcmp->opaque;
3262 				if (rxr->rx_buf_ring[cons])
3263 					return RTE_ETH_RX_DESC_DONE;
3264 				else
3265 					return RTE_ETH_RX_DESC_UNAVAIL;
3266 			}
3267 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3268 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3269 			desc++;
3270 			break;
3271 
3272 		case CMPL_BASE_TYPE_RX_TPA_END:
3273 			if (desc == offset)
3274 				return RTE_ETH_RX_DESC_DONE;
3275 
3276 			if (BNXT_CHIP_P5(rxq->bp)) {
3277 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3278 
3279 				p5_tpa_end = (void *)rxcmp;
3280 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3281 			} else {
3282 				struct rx_tpa_end_cmpl *tpa_end;
3283 
3284 				tpa_end = (void *)rxcmp;
3285 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3286 			}
3287 
3288 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3289 			desc++;
3290 			break;
3291 
3292 		default:
3293 			raw_cons += CMP_LEN(cmpl_type);
3294 		}
3295 	}
3296 
3297 	return RTE_ETH_RX_DESC_AVAIL;
3298 }
3299 
3300 static int
3301 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3302 {
3303 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3304 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3305 	uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3306 	struct cmpl_base *cp_desc_ring;
3307 	int rc;
3308 
3309 	rc = is_bnxt_in_error(txq->bp);
3310 	if (rc)
3311 		return rc;
3312 
3313 	if (offset >= txq->nb_tx_desc)
3314 		return -EINVAL;
3315 
3316 	/* Return "desc done" if descriptor is available for use. */
3317 	if (bnxt_tx_bds_in_hw(txq) <= offset)
3318 		return RTE_ETH_TX_DESC_DONE;
3319 
3320 	raw_cons = cpr->cp_raw_cons;
3321 	cp_desc_ring = cpr->cp_desc_ring;
3322 	ring_mask = cpr->cp_ring_struct->ring_mask;
3323 
3324 	/* Check to see if hw has posted a completion for the descriptor. */
3325 	while (1) {
3326 		struct tx_cmpl *txcmp;
3327 		uint32_t cons;
3328 
3329 		cons = RING_CMPL(ring_mask, raw_cons);
3330 		txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3331 
3332 		if (!bnxt_cpr_cmp_valid(txcmp, raw_cons, ring_mask + 1))
3333 			break;
3334 
3335 		if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3336 			nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3337 
3338 		if (nb_tx_pkts > offset)
3339 			return RTE_ETH_TX_DESC_DONE;
3340 
3341 		raw_cons = NEXT_RAW_CMP(raw_cons);
3342 	}
3343 
3344 	/* Descriptor is pending transmit, not yet completed by hardware. */
3345 	return RTE_ETH_TX_DESC_FULL;
3346 }
3347 
3348 int
3349 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3350 		     const struct rte_flow_ops **ops)
3351 {
3352 	struct bnxt *bp = dev->data->dev_private;
3353 	int ret = 0;
3354 
3355 	if (!bp)
3356 		return -EIO;
3357 
3358 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3359 		struct bnxt_representor *vfr = dev->data->dev_private;
3360 		bp = vfr->parent_dev->data->dev_private;
3361 		/* parent is deleted while children are still valid */
3362 		if (!bp) {
3363 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3364 				    dev->data->port_id);
3365 			return -EIO;
3366 		}
3367 	}
3368 
3369 	ret = is_bnxt_in_error(bp);
3370 	if (ret)
3371 		return ret;
3372 
3373 	/* PMD supports thread-safe flow operations.  rte_flow API
3374 	 * functions can avoid mutex for multi-thread safety.
3375 	 */
3376 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3377 
3378 	if (BNXT_TRUFLOW_EN(bp))
3379 		*ops = &bnxt_ulp_rte_flow_ops;
3380 	else
3381 		*ops = &bnxt_flow_ops;
3382 
3383 	return ret;
3384 }
3385 
3386 static const uint32_t *
3387 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3388 {
3389 	static const uint32_t ptypes[] = {
3390 		RTE_PTYPE_L2_ETHER_VLAN,
3391 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3392 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3393 		RTE_PTYPE_L4_ICMP,
3394 		RTE_PTYPE_L4_TCP,
3395 		RTE_PTYPE_L4_UDP,
3396 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3397 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3398 		RTE_PTYPE_INNER_L4_ICMP,
3399 		RTE_PTYPE_INNER_L4_TCP,
3400 		RTE_PTYPE_INNER_L4_UDP,
3401 		RTE_PTYPE_UNKNOWN
3402 	};
3403 
3404 	if (!dev->rx_pkt_burst)
3405 		return NULL;
3406 
3407 	return ptypes;
3408 }
3409 
3410 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3411 			 int reg_win)
3412 {
3413 	uint32_t reg_base = *reg_arr & 0xfffff000;
3414 	uint32_t win_off;
3415 	int i;
3416 
3417 	for (i = 0; i < count; i++) {
3418 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3419 			return -ERANGE;
3420 	}
3421 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3422 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3423 	return 0;
3424 }
3425 
3426 static int bnxt_map_ptp_regs(struct bnxt *bp)
3427 {
3428 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3429 	uint32_t *reg_arr;
3430 	int rc, i;
3431 
3432 	reg_arr = ptp->rx_regs;
3433 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3434 	if (rc)
3435 		return rc;
3436 
3437 	reg_arr = ptp->tx_regs;
3438 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3439 	if (rc)
3440 		return rc;
3441 
3442 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3443 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3444 
3445 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3446 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3447 
3448 	return 0;
3449 }
3450 
3451 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3452 {
3453 	rte_write32(0, (uint8_t *)bp->bar0 +
3454 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3455 	rte_write32(0, (uint8_t *)bp->bar0 +
3456 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3457 }
3458 
3459 static uint64_t bnxt_cc_read(struct bnxt *bp)
3460 {
3461 	uint64_t ns;
3462 
3463 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3464 			      BNXT_GRCPF_REG_SYNC_TIME));
3465 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3466 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3467 	return ns;
3468 }
3469 
3470 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3471 {
3472 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3473 	uint32_t fifo;
3474 
3475 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3476 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3477 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3478 		return -EAGAIN;
3479 
3480 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3481 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3482 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3483 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3484 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3485 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3486 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3487 
3488 	return 0;
3489 }
3490 
3491 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3492 {
3493 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3494 	struct bnxt_pf_info *pf = bp->pf;
3495 	uint16_t port_id;
3496 	int i = 0;
3497 	uint32_t fifo;
3498 
3499 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3500 		return -EINVAL;
3501 
3502 	port_id = pf->port_id;
3503 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3504 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3505 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3506 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3507 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3508 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3509 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3510 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3511 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3512 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3513 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3514 		i++;
3515 	}
3516 
3517 	if (i >= BNXT_PTP_RX_PND_CNT)
3518 		return -EBUSY;
3519 
3520 	return 0;
3521 }
3522 
3523 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3524 {
3525 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3526 	struct bnxt_pf_info *pf = bp->pf;
3527 	uint16_t port_id;
3528 	uint32_t fifo;
3529 
3530 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3531 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3532 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3533 		return -EAGAIN;
3534 
3535 	port_id = pf->port_id;
3536 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3537 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3538 
3539 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3540 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3541 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3542 		return bnxt_clr_rx_ts(bp, ts);
3543 
3544 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3545 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3546 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3547 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3548 
3549 	return 0;
3550 }
3551 
3552 static int
3553 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3554 {
3555 	uint64_t ns;
3556 	struct bnxt *bp = dev->data->dev_private;
3557 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3558 
3559 	if (!ptp)
3560 		return -ENOTSUP;
3561 
3562 	ns = rte_timespec_to_ns(ts);
3563 	/* Set the timecounters to a new value. */
3564 	ptp->tc.nsec = ns;
3565 	ptp->tx_tstamp_tc.nsec = ns;
3566 	ptp->rx_tstamp_tc.nsec = ns;
3567 
3568 	return 0;
3569 }
3570 
3571 static int
3572 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3573 {
3574 	struct bnxt *bp = dev->data->dev_private;
3575 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3576 	uint64_t ns, systime_cycles = 0;
3577 	int rc = 0;
3578 
3579 	if (!ptp)
3580 		return -ENOTSUP;
3581 
3582 	if (BNXT_CHIP_P5(bp))
3583 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3584 					     &systime_cycles);
3585 	else
3586 		systime_cycles = bnxt_cc_read(bp);
3587 
3588 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3589 	*ts = rte_ns_to_timespec(ns);
3590 
3591 	return rc;
3592 }
3593 static int
3594 bnxt_timesync_enable(struct rte_eth_dev *dev)
3595 {
3596 	struct bnxt *bp = dev->data->dev_private;
3597 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3598 	uint32_t shift = 0;
3599 	int rc;
3600 
3601 	if (!ptp)
3602 		return -ENOTSUP;
3603 
3604 	ptp->rx_filter = 1;
3605 	ptp->tx_tstamp_en = 1;
3606 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3607 
3608 	rc = bnxt_hwrm_ptp_cfg(bp);
3609 	if (rc)
3610 		return rc;
3611 
3612 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3613 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3614 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3615 
3616 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3617 	ptp->tc.cc_shift = shift;
3618 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3619 
3620 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3621 	ptp->rx_tstamp_tc.cc_shift = shift;
3622 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3623 
3624 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3625 	ptp->tx_tstamp_tc.cc_shift = shift;
3626 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3627 
3628 	if (!BNXT_CHIP_P5(bp))
3629 		bnxt_map_ptp_regs(bp);
3630 	else
3631 		rc = bnxt_ptp_start(bp);
3632 
3633 	return rc;
3634 }
3635 
3636 static int
3637 bnxt_timesync_disable(struct rte_eth_dev *dev)
3638 {
3639 	struct bnxt *bp = dev->data->dev_private;
3640 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3641 
3642 	if (!ptp)
3643 		return -ENOTSUP;
3644 
3645 	ptp->rx_filter = 0;
3646 	ptp->tx_tstamp_en = 0;
3647 	ptp->rxctl = 0;
3648 
3649 	bnxt_hwrm_ptp_cfg(bp);
3650 
3651 	if (!BNXT_CHIP_P5(bp))
3652 		bnxt_unmap_ptp_regs(bp);
3653 	else
3654 		bnxt_ptp_stop(bp);
3655 
3656 	return 0;
3657 }
3658 
3659 static int
3660 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3661 				 struct timespec *timestamp,
3662 				 uint32_t flags __rte_unused)
3663 {
3664 	struct bnxt *bp = dev->data->dev_private;
3665 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3666 	uint64_t rx_tstamp_cycles = 0;
3667 	uint64_t ns;
3668 
3669 	if (!ptp)
3670 		return -ENOTSUP;
3671 
3672 	if (BNXT_CHIP_P5(bp))
3673 		rx_tstamp_cycles = ptp->rx_timestamp;
3674 	else
3675 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3676 
3677 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3678 	*timestamp = rte_ns_to_timespec(ns);
3679 	return  0;
3680 }
3681 
3682 static int
3683 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3684 				 struct timespec *timestamp)
3685 {
3686 	struct bnxt *bp = dev->data->dev_private;
3687 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3688 	uint64_t tx_tstamp_cycles = 0;
3689 	uint64_t ns;
3690 	int rc = 0;
3691 
3692 	if (!ptp)
3693 		return -ENOTSUP;
3694 
3695 	if (BNXT_CHIP_P5(bp))
3696 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3697 					     &tx_tstamp_cycles);
3698 	else
3699 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3700 
3701 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3702 	*timestamp = rte_ns_to_timespec(ns);
3703 
3704 	return rc;
3705 }
3706 
3707 static int
3708 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3709 {
3710 	struct bnxt *bp = dev->data->dev_private;
3711 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3712 
3713 	if (!ptp)
3714 		return -ENOTSUP;
3715 
3716 	ptp->tc.nsec += delta;
3717 	ptp->tx_tstamp_tc.nsec += delta;
3718 	ptp->rx_tstamp_tc.nsec += delta;
3719 
3720 	return 0;
3721 }
3722 
3723 static int
3724 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3725 {
3726 	struct bnxt *bp = dev->data->dev_private;
3727 	int rc;
3728 	uint32_t dir_entries;
3729 	uint32_t entry_length;
3730 
3731 	rc = is_bnxt_in_error(bp);
3732 	if (rc)
3733 		return rc;
3734 
3735 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3736 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3737 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3738 
3739 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3740 	if (rc != 0)
3741 		return rc;
3742 
3743 	return dir_entries * entry_length;
3744 }
3745 
3746 static int
3747 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3748 		struct rte_dev_eeprom_info *in_eeprom)
3749 {
3750 	struct bnxt *bp = dev->data->dev_private;
3751 	uint32_t index;
3752 	uint32_t offset;
3753 	int rc;
3754 
3755 	rc = is_bnxt_in_error(bp);
3756 	if (rc)
3757 		return rc;
3758 
3759 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3760 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3761 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3762 		    in_eeprom->offset, in_eeprom->length);
3763 
3764 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3765 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3766 						in_eeprom->data);
3767 
3768 	index = in_eeprom->offset >> 24;
3769 	offset = in_eeprom->offset & 0xffffff;
3770 
3771 	if (index != 0)
3772 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3773 					   in_eeprom->length, in_eeprom->data);
3774 
3775 	return 0;
3776 }
3777 
3778 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3779 {
3780 	switch (dir_type) {
3781 	case BNX_DIR_TYPE_CHIMP_PATCH:
3782 	case BNX_DIR_TYPE_BOOTCODE:
3783 	case BNX_DIR_TYPE_BOOTCODE_2:
3784 	case BNX_DIR_TYPE_APE_FW:
3785 	case BNX_DIR_TYPE_APE_PATCH:
3786 	case BNX_DIR_TYPE_KONG_FW:
3787 	case BNX_DIR_TYPE_KONG_PATCH:
3788 	case BNX_DIR_TYPE_BONO_FW:
3789 	case BNX_DIR_TYPE_BONO_PATCH:
3790 		/* FALLTHROUGH */
3791 		return true;
3792 	}
3793 
3794 	return false;
3795 }
3796 
3797 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3798 {
3799 	switch (dir_type) {
3800 	case BNX_DIR_TYPE_AVS:
3801 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3802 	case BNX_DIR_TYPE_PCIE:
3803 	case BNX_DIR_TYPE_TSCF_UCODE:
3804 	case BNX_DIR_TYPE_EXT_PHY:
3805 	case BNX_DIR_TYPE_CCM:
3806 	case BNX_DIR_TYPE_ISCSI_BOOT:
3807 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3808 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3809 		/* FALLTHROUGH */
3810 		return true;
3811 	}
3812 
3813 	return false;
3814 }
3815 
3816 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3817 {
3818 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3819 		bnxt_dir_type_is_other_exec_format(dir_type);
3820 }
3821 
3822 static int
3823 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3824 		struct rte_dev_eeprom_info *in_eeprom)
3825 {
3826 	struct bnxt *bp = dev->data->dev_private;
3827 	uint8_t index, dir_op;
3828 	uint16_t type, ext, ordinal, attr;
3829 	int rc;
3830 
3831 	rc = is_bnxt_in_error(bp);
3832 	if (rc)
3833 		return rc;
3834 
3835 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3836 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3837 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3838 		    in_eeprom->offset, in_eeprom->length);
3839 
3840 	if (!BNXT_PF(bp)) {
3841 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3842 		return -EINVAL;
3843 	}
3844 
3845 	type = in_eeprom->magic >> 16;
3846 
3847 	if (type == 0xffff) { /* special value for directory operations */
3848 		index = in_eeprom->magic & 0xff;
3849 		dir_op = in_eeprom->magic >> 8;
3850 		if (index == 0)
3851 			return -EINVAL;
3852 		switch (dir_op) {
3853 		case 0x0e: /* erase */
3854 			if (in_eeprom->offset != ~in_eeprom->magic)
3855 				return -EINVAL;
3856 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3857 		default:
3858 			return -EINVAL;
3859 		}
3860 	}
3861 
3862 	/* Create or re-write an NVM item: */
3863 	if (bnxt_dir_type_is_executable(type) == true)
3864 		return -EOPNOTSUPP;
3865 	ext = in_eeprom->magic & 0xffff;
3866 	ordinal = in_eeprom->offset >> 16;
3867 	attr = in_eeprom->offset & 0xffff;
3868 
3869 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3870 				     in_eeprom->data, in_eeprom->length);
3871 }
3872 
3873 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3874 				struct rte_eth_dev_module_info *modinfo)
3875 {
3876 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3877 	struct bnxt *bp = dev->data->dev_private;
3878 	int rc;
3879 
3880 	/* No point in going further if phy status indicates
3881 	 * module is not inserted or if it is powered down or
3882 	 * if it is of type 10GBase-T
3883 	 */
3884 	if (bp->link_info->module_status >
3885 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3886 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3887 			    dev->data->port_id);
3888 		return -ENOTSUP;
3889 	}
3890 
3891 	/* This feature is not supported in older firmware versions */
3892 	if (bp->hwrm_spec_code < 0x10202) {
3893 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3894 			    dev->data->port_id);
3895 		return -ENOTSUP;
3896 	}
3897 
3898 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3899 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3900 						   module_info);
3901 
3902 	if (rc)
3903 		return rc;
3904 
3905 	switch (module_info[0]) {
3906 	case SFF_MODULE_ID_SFP:
3907 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3908 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3909 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3910 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3911 		break;
3912 	case SFF_MODULE_ID_QSFP:
3913 	case SFF_MODULE_ID_QSFP_PLUS:
3914 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3915 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3916 		break;
3917 	case SFF_MODULE_ID_QSFP28:
3918 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3919 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3920 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3921 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3922 		break;
3923 	default:
3924 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3925 		return -ENOTSUP;
3926 	}
3927 
3928 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3929 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3930 
3931 	return 0;
3932 }
3933 
3934 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3935 				  struct rte_dev_eeprom_info *info)
3936 {
3937 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3938 	uint32_t offset = info->offset, length = info->length;
3939 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3940 	struct bnxt *bp = dev->data->dev_private;
3941 	uint8_t *data = info->data;
3942 	uint8_t page = offset >> 7;
3943 	uint8_t max_pages = 2;
3944 	uint8_t opt_pages;
3945 	int rc;
3946 
3947 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3948 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3949 						   module_info);
3950 	if (rc)
3951 		return rc;
3952 
3953 	switch (module_info[0]) {
3954 	case SFF_MODULE_ID_SFP:
3955 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
3956 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
3957 			pg_addr[2] = I2C_DEV_ADDR_A2;
3958 			pg_addr[3] = I2C_DEV_ADDR_A2;
3959 			max_pages = 4;
3960 		}
3961 		break;
3962 	case SFF_MODULE_ID_QSFP28:
3963 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
3964 							   SFF8636_OPT_PAGES_OFFSET,
3965 							   1, &opt_pages);
3966 		if (rc)
3967 			return rc;
3968 
3969 		if (opt_pages & SFF8636_PAGE1_MASK) {
3970 			pg_addr[2] = I2C_DEV_ADDR_A0;
3971 			max_pages = 3;
3972 		}
3973 		if (opt_pages & SFF8636_PAGE2_MASK) {
3974 			pg_addr[3] = I2C_DEV_ADDR_A0;
3975 			max_pages = 4;
3976 		}
3977 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
3978 			pg_addr[4] = I2C_DEV_ADDR_A0;
3979 			max_pages = 5;
3980 		}
3981 		break;
3982 	default:
3983 		break;
3984 	}
3985 
3986 	memset(data, 0, length);
3987 
3988 	offset &= 0xff;
3989 	while (length && page < max_pages) {
3990 		uint8_t raw_page = page ? page - 1 : 0;
3991 		uint16_t chunk;
3992 
3993 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
3994 			raw_page = 0;
3995 		else if (page)
3996 			offset |= 0x80;
3997 		chunk = RTE_MIN(length, 256 - offset);
3998 
3999 		if (pg_addr[page]) {
4000 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
4001 								   raw_page, offset,
4002 								   chunk, data);
4003 			if (rc)
4004 				return rc;
4005 		}
4006 
4007 		data += chunk;
4008 		length -= chunk;
4009 		offset = 0;
4010 		page += 1 + (chunk > 128);
4011 	}
4012 
4013 	return length ? -EINVAL : 0;
4014 }
4015 
4016 /*
4017  * Initialization
4018  */
4019 
4020 static const struct eth_dev_ops bnxt_dev_ops = {
4021 	.dev_infos_get = bnxt_dev_info_get_op,
4022 	.dev_close = bnxt_dev_close_op,
4023 	.dev_configure = bnxt_dev_configure_op,
4024 	.dev_start = bnxt_dev_start_op,
4025 	.dev_stop = bnxt_dev_stop_op,
4026 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4027 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4028 	.stats_get = bnxt_stats_get_op,
4029 	.stats_reset = bnxt_stats_reset_op,
4030 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4031 	.rx_queue_release = bnxt_rx_queue_release_op,
4032 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4033 	.tx_queue_release = bnxt_tx_queue_release_op,
4034 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4035 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4036 	.reta_update = bnxt_reta_update_op,
4037 	.reta_query = bnxt_reta_query_op,
4038 	.rss_hash_update = bnxt_rss_hash_update_op,
4039 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4040 	.link_update = bnxt_link_update_op,
4041 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4042 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4043 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4044 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4045 	.mac_addr_add = bnxt_mac_addr_add_op,
4046 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4047 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4048 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4049 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4050 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4051 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4052 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4053 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4054 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4055 	.mtu_set = bnxt_mtu_set_op,
4056 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4057 	.xstats_get = bnxt_dev_xstats_get_op,
4058 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4059 	.xstats_reset = bnxt_dev_xstats_reset_op,
4060 	.fw_version_get = bnxt_fw_version_get,
4061 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4062 	.rxq_info_get = bnxt_rxq_info_get_op,
4063 	.txq_info_get = bnxt_txq_info_get_op,
4064 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4065 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4066 	.dev_led_on = bnxt_dev_led_on_op,
4067 	.dev_led_off = bnxt_dev_led_off_op,
4068 	.rx_queue_start = bnxt_rx_queue_start,
4069 	.rx_queue_stop = bnxt_rx_queue_stop,
4070 	.tx_queue_start = bnxt_tx_queue_start,
4071 	.tx_queue_stop = bnxt_tx_queue_stop,
4072 	.flow_ops_get = bnxt_flow_ops_get_op,
4073 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4074 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4075 	.get_eeprom           = bnxt_get_eeprom_op,
4076 	.set_eeprom           = bnxt_set_eeprom_op,
4077 	.get_module_info = bnxt_get_module_info,
4078 	.get_module_eeprom = bnxt_get_module_eeprom,
4079 	.timesync_enable      = bnxt_timesync_enable,
4080 	.timesync_disable     = bnxt_timesync_disable,
4081 	.timesync_read_time   = bnxt_timesync_read_time,
4082 	.timesync_write_time   = bnxt_timesync_write_time,
4083 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4084 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4085 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4086 };
4087 
4088 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4089 {
4090 	uint32_t offset;
4091 
4092 	/* Only pre-map the reset GRC registers using window 3 */
4093 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4094 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4095 
4096 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4097 
4098 	return offset;
4099 }
4100 
4101 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4102 {
4103 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4104 	uint32_t reg_base = 0xffffffff;
4105 	int i;
4106 
4107 	/* Only pre-map the monitoring GRC registers using window 2 */
4108 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4109 		uint32_t reg = info->status_regs[i];
4110 
4111 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4112 			continue;
4113 
4114 		if (reg_base == 0xffffffff)
4115 			reg_base = reg & 0xfffff000;
4116 		if ((reg & 0xfffff000) != reg_base)
4117 			return -ERANGE;
4118 
4119 		/* Use mask 0xffc as the Lower 2 bits indicates
4120 		 * address space location
4121 		 */
4122 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4123 						(reg & 0xffc);
4124 	}
4125 
4126 	if (reg_base == 0xffffffff)
4127 		return 0;
4128 
4129 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4130 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4131 
4132 	return 0;
4133 }
4134 
4135 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4136 {
4137 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4138 	uint32_t delay = info->delay_after_reset[index];
4139 	uint32_t val = info->reset_reg_val[index];
4140 	uint32_t reg = info->reset_reg[index];
4141 	uint32_t type, offset;
4142 	int ret;
4143 
4144 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4145 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4146 
4147 	switch (type) {
4148 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4149 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4150 		if (ret < 0) {
4151 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4152 				    val, offset);
4153 			return;
4154 		}
4155 		break;
4156 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4157 		offset = bnxt_map_reset_regs(bp, offset);
4158 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4159 		break;
4160 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4161 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4162 		break;
4163 	}
4164 	/* wait on a specific interval of time until core reset is complete */
4165 	if (delay)
4166 		rte_delay_ms(delay);
4167 }
4168 
4169 static void bnxt_dev_cleanup(struct bnxt *bp)
4170 {
4171 	bp->eth_dev->data->dev_link.link_status = 0;
4172 	bp->link_info->link_up = 0;
4173 	if (bp->eth_dev->data->dev_started)
4174 		bnxt_dev_stop(bp->eth_dev);
4175 
4176 	bnxt_uninit_resources(bp, true);
4177 }
4178 
4179 static int
4180 bnxt_check_fw_reset_done(struct bnxt *bp)
4181 {
4182 	int timeout = bp->fw_reset_max_msecs;
4183 	uint16_t val = 0;
4184 	int rc;
4185 
4186 	do {
4187 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4188 		if (rc < 0) {
4189 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4190 			return rc;
4191 		}
4192 		if (val != 0xffff)
4193 			break;
4194 		rte_delay_ms(1);
4195 	} while (timeout--);
4196 
4197 	if (val == 0xffff) {
4198 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4199 		return -1;
4200 	}
4201 
4202 	return 0;
4203 }
4204 
4205 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4206 {
4207 	struct rte_eth_dev *dev = bp->eth_dev;
4208 	struct rte_vlan_filter_conf *vfc;
4209 	int vidx, vbit, rc;
4210 	uint16_t vlan_id;
4211 
4212 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4213 		vfc = &dev->data->vlan_filter_conf;
4214 		vidx = vlan_id / 64;
4215 		vbit = vlan_id % 64;
4216 
4217 		/* Each bit corresponds to a VLAN id */
4218 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4219 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4220 			if (rc)
4221 				return rc;
4222 		}
4223 	}
4224 
4225 	return 0;
4226 }
4227 
4228 static int bnxt_restore_mac_filters(struct bnxt *bp)
4229 {
4230 	struct rte_eth_dev *dev = bp->eth_dev;
4231 	struct rte_eth_dev_info dev_info;
4232 	struct rte_ether_addr *addr;
4233 	uint64_t pool_mask;
4234 	uint32_t pool = 0;
4235 	uint32_t i;
4236 	int rc;
4237 
4238 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4239 		return 0;
4240 
4241 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4242 	if (rc)
4243 		return rc;
4244 
4245 	/* replay MAC address configuration */
4246 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4247 		addr = &dev->data->mac_addrs[i];
4248 
4249 		/* skip zero address */
4250 		if (rte_is_zero_ether_addr(addr))
4251 			continue;
4252 
4253 		pool = 0;
4254 		pool_mask = dev->data->mac_pool_sel[i];
4255 
4256 		do {
4257 			if (pool_mask & 1ULL) {
4258 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4259 				if (rc)
4260 					return rc;
4261 			}
4262 			pool_mask >>= 1;
4263 			pool++;
4264 		} while (pool_mask);
4265 	}
4266 
4267 	return 0;
4268 }
4269 
4270 static int bnxt_restore_filters(struct bnxt *bp)
4271 {
4272 	struct rte_eth_dev *dev = bp->eth_dev;
4273 	int ret = 0;
4274 
4275 	if (dev->data->all_multicast) {
4276 		ret = bnxt_allmulticast_enable_op(dev);
4277 		if (ret)
4278 			return ret;
4279 	}
4280 	if (dev->data->promiscuous) {
4281 		ret = bnxt_promiscuous_enable_op(dev);
4282 		if (ret)
4283 			return ret;
4284 	}
4285 
4286 	ret = bnxt_restore_mac_filters(bp);
4287 	if (ret)
4288 		return ret;
4289 
4290 	ret = bnxt_restore_vlan_filters(bp);
4291 	/* TODO restore other filters as well */
4292 	return ret;
4293 }
4294 
4295 static int bnxt_check_fw_ready(struct bnxt *bp)
4296 {
4297 	int timeout = bp->fw_reset_max_msecs;
4298 	int rc = 0;
4299 
4300 	do {
4301 		rc = bnxt_hwrm_poll_ver_get(bp);
4302 		if (rc == 0)
4303 			break;
4304 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4305 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4306 	} while (rc && timeout > 0);
4307 
4308 	if (rc)
4309 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4310 
4311 	return rc;
4312 }
4313 
4314 static void bnxt_dev_recover(void *arg)
4315 {
4316 	struct bnxt *bp = arg;
4317 	int rc = 0;
4318 
4319 	pthread_mutex_lock(&bp->err_recovery_lock);
4320 
4321 	if (!bp->fw_reset_min_msecs) {
4322 		rc = bnxt_check_fw_reset_done(bp);
4323 		if (rc)
4324 			goto err;
4325 	}
4326 
4327 	/* Clear Error flag so that device re-init should happen */
4328 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4329 
4330 	rc = bnxt_check_fw_ready(bp);
4331 	if (rc)
4332 		goto err;
4333 
4334 	rc = bnxt_init_resources(bp, true);
4335 	if (rc) {
4336 		PMD_DRV_LOG(ERR,
4337 			    "Failed to initialize resources after reset\n");
4338 		goto err;
4339 	}
4340 	/* clear reset flag as the device is initialized now */
4341 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4342 
4343 	rc = bnxt_dev_start_op(bp->eth_dev);
4344 	if (rc) {
4345 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4346 		goto err_start;
4347 	}
4348 
4349 	rc = bnxt_restore_filters(bp);
4350 	if (rc)
4351 		goto err_start;
4352 
4353 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4354 	pthread_mutex_unlock(&bp->err_recovery_lock);
4355 
4356 	return;
4357 err_start:
4358 	bnxt_dev_stop(bp->eth_dev);
4359 err:
4360 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4361 	bnxt_uninit_resources(bp, false);
4362 	if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4363 		rte_eth_dev_callback_process(bp->eth_dev,
4364 					     RTE_ETH_EVENT_INTR_RMV,
4365 					     NULL);
4366 	pthread_mutex_unlock(&bp->err_recovery_lock);
4367 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4368 }
4369 
4370 void bnxt_dev_reset_and_resume(void *arg)
4371 {
4372 	struct bnxt *bp = arg;
4373 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4374 	uint16_t val = 0;
4375 	int rc;
4376 
4377 	bnxt_dev_cleanup(bp);
4378 
4379 	bnxt_wait_for_device_shutdown(bp);
4380 
4381 	/* During some fatal firmware error conditions, the PCI config space
4382 	 * register 0x2e which normally contains the subsystem ID will become
4383 	 * 0xffff. This register will revert back to the normal value after
4384 	 * the chip has completed core reset. If we detect this condition,
4385 	 * we can poll this config register immediately for the value to revert.
4386 	 */
4387 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4388 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4389 		if (rc < 0) {
4390 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4391 			return;
4392 		}
4393 		if (val == 0xffff) {
4394 			bp->fw_reset_min_msecs = 0;
4395 			us = 1;
4396 		}
4397 	}
4398 
4399 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4400 	if (rc)
4401 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4402 }
4403 
4404 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4405 {
4406 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4407 	uint32_t reg = info->status_regs[index];
4408 	uint32_t type, offset, val = 0;
4409 	int ret = 0;
4410 
4411 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4412 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4413 
4414 	switch (type) {
4415 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4416 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4417 		if (ret < 0)
4418 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4419 				    offset);
4420 		break;
4421 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4422 		offset = info->mapped_status_regs[index];
4423 		/* FALLTHROUGH */
4424 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4425 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4426 				       offset));
4427 		break;
4428 	}
4429 
4430 	return val;
4431 }
4432 
4433 static int bnxt_fw_reset_all(struct bnxt *bp)
4434 {
4435 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4436 	uint32_t i;
4437 	int rc = 0;
4438 
4439 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4440 		/* Reset through master function driver */
4441 		for (i = 0; i < info->reg_array_cnt; i++)
4442 			bnxt_write_fw_reset_reg(bp, i);
4443 		/* Wait for time specified by FW after triggering reset */
4444 		rte_delay_ms(info->master_func_wait_period_after_reset);
4445 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4446 		/* Reset with the help of Kong processor */
4447 		rc = bnxt_hwrm_fw_reset(bp);
4448 		if (rc)
4449 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4450 	}
4451 
4452 	return rc;
4453 }
4454 
4455 static void bnxt_fw_reset_cb(void *arg)
4456 {
4457 	struct bnxt *bp = arg;
4458 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4459 	int rc = 0;
4460 
4461 	/* Only Master function can do FW reset */
4462 	if (bnxt_is_master_func(bp) &&
4463 	    bnxt_is_recovery_enabled(bp)) {
4464 		rc = bnxt_fw_reset_all(bp);
4465 		if (rc) {
4466 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4467 			return;
4468 		}
4469 	}
4470 
4471 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4472 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4473 	 * (including MASTER FUNC). After receiving this Async, all the active
4474 	 * drivers should treat this case as FW initiated recovery
4475 	 */
4476 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4477 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4478 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4479 
4480 		/* To recover from error */
4481 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4482 				  (void *)bp);
4483 	}
4484 }
4485 
4486 /* Driver should poll FW heartbeat, reset_counter with the frequency
4487  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4488  * When the driver detects heartbeat stop or change in reset_counter,
4489  * it has to trigger a reset to recover from the error condition.
4490  * A “master PF” is the function who will have the privilege to
4491  * initiate the chimp reset. The master PF will be elected by the
4492  * firmware and will be notified through async message.
4493  */
4494 static void bnxt_check_fw_health(void *arg)
4495 {
4496 	struct bnxt *bp = arg;
4497 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4498 	uint32_t val = 0, wait_msec;
4499 
4500 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4501 	    is_bnxt_in_error(bp))
4502 		return;
4503 
4504 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4505 	if (val == info->last_heart_beat)
4506 		goto reset;
4507 
4508 	info->last_heart_beat = val;
4509 
4510 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4511 	if (val != info->last_reset_counter)
4512 		goto reset;
4513 
4514 	info->last_reset_counter = val;
4515 
4516 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4517 			  bnxt_check_fw_health, (void *)bp);
4518 
4519 	return;
4520 reset:
4521 	/* Stop DMA to/from device */
4522 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4523 	bp->flags |= BNXT_FLAG_FW_RESET;
4524 
4525 	bnxt_stop_rxtx(bp);
4526 
4527 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4528 
4529 	if (bnxt_is_master_func(bp))
4530 		wait_msec = info->master_func_wait_period;
4531 	else
4532 		wait_msec = info->normal_func_wait_period;
4533 
4534 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4535 			  bnxt_fw_reset_cb, (void *)bp);
4536 }
4537 
4538 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4539 {
4540 	uint32_t polling_freq;
4541 
4542 	pthread_mutex_lock(&bp->health_check_lock);
4543 
4544 	if (!bnxt_is_recovery_enabled(bp))
4545 		goto done;
4546 
4547 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4548 		goto done;
4549 
4550 	polling_freq = bp->recovery_info->driver_polling_freq;
4551 
4552 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4553 			  bnxt_check_fw_health, (void *)bp);
4554 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4555 
4556 done:
4557 	pthread_mutex_unlock(&bp->health_check_lock);
4558 }
4559 
4560 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4561 {
4562 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4563 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4564 }
4565 
4566 static bool bnxt_vf_pciid(uint16_t device_id)
4567 {
4568 	switch (device_id) {
4569 	case BROADCOM_DEV_ID_57304_VF:
4570 	case BROADCOM_DEV_ID_57406_VF:
4571 	case BROADCOM_DEV_ID_5731X_VF:
4572 	case BROADCOM_DEV_ID_5741X_VF:
4573 	case BROADCOM_DEV_ID_57414_VF:
4574 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4575 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4576 	case BROADCOM_DEV_ID_58802_VF:
4577 	case BROADCOM_DEV_ID_57500_VF1:
4578 	case BROADCOM_DEV_ID_57500_VF2:
4579 	case BROADCOM_DEV_ID_58818_VF:
4580 		/* FALLTHROUGH */
4581 		return true;
4582 	default:
4583 		return false;
4584 	}
4585 }
4586 
4587 /* Phase 5 device */
4588 static bool bnxt_p5_device(uint16_t device_id)
4589 {
4590 	switch (device_id) {
4591 	case BROADCOM_DEV_ID_57508:
4592 	case BROADCOM_DEV_ID_57504:
4593 	case BROADCOM_DEV_ID_57502:
4594 	case BROADCOM_DEV_ID_57508_MF1:
4595 	case BROADCOM_DEV_ID_57504_MF1:
4596 	case BROADCOM_DEV_ID_57502_MF1:
4597 	case BROADCOM_DEV_ID_57508_MF2:
4598 	case BROADCOM_DEV_ID_57504_MF2:
4599 	case BROADCOM_DEV_ID_57502_MF2:
4600 	case BROADCOM_DEV_ID_57500_VF1:
4601 	case BROADCOM_DEV_ID_57500_VF2:
4602 	case BROADCOM_DEV_ID_58812:
4603 	case BROADCOM_DEV_ID_58814:
4604 	case BROADCOM_DEV_ID_58818:
4605 	case BROADCOM_DEV_ID_58818_VF:
4606 		/* FALLTHROUGH */
4607 		return true;
4608 	default:
4609 		return false;
4610 	}
4611 }
4612 
4613 bool bnxt_stratus_device(struct bnxt *bp)
4614 {
4615 	uint16_t device_id = bp->pdev->id.device_id;
4616 
4617 	switch (device_id) {
4618 	case BROADCOM_DEV_ID_STRATUS_NIC:
4619 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4620 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4621 		/* FALLTHROUGH */
4622 		return true;
4623 	default:
4624 		return false;
4625 	}
4626 }
4627 
4628 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4629 {
4630 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4631 	struct bnxt *bp = eth_dev->data->dev_private;
4632 
4633 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4634 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4635 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4636 	if (!bp->bar0 || !bp->doorbell_base) {
4637 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4638 		return -ENODEV;
4639 	}
4640 
4641 	bp->eth_dev = eth_dev;
4642 	bp->pdev = pci_dev;
4643 
4644 	return 0;
4645 }
4646 
4647 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4648 				  struct bnxt_ctx_pg_info *ctx_pg,
4649 				  uint32_t mem_size,
4650 				  const char *suffix,
4651 				  uint16_t idx)
4652 {
4653 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4654 	const struct rte_memzone *mz = NULL;
4655 	char mz_name[RTE_MEMZONE_NAMESIZE];
4656 	rte_iova_t mz_phys_addr;
4657 	uint64_t valid_bits = 0;
4658 	uint32_t sz;
4659 	int i;
4660 
4661 	if (!mem_size)
4662 		return 0;
4663 
4664 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4665 			 BNXT_PAGE_SIZE;
4666 	rmem->page_size = BNXT_PAGE_SIZE;
4667 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4668 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4669 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4670 
4671 	valid_bits = PTU_PTE_VALID;
4672 
4673 	if (rmem->nr_pages > 1) {
4674 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4675 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4676 			 suffix, idx, bp->eth_dev->data->port_id);
4677 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4678 		mz = rte_memzone_lookup(mz_name);
4679 		if (!mz) {
4680 			mz = rte_memzone_reserve_aligned(mz_name,
4681 						rmem->nr_pages * 8,
4682 						bp->eth_dev->device->numa_node,
4683 						RTE_MEMZONE_2MB |
4684 						RTE_MEMZONE_SIZE_HINT_ONLY |
4685 						RTE_MEMZONE_IOVA_CONTIG,
4686 						BNXT_PAGE_SIZE);
4687 			if (mz == NULL)
4688 				return -ENOMEM;
4689 		}
4690 
4691 		memset(mz->addr, 0, mz->len);
4692 		mz_phys_addr = mz->iova;
4693 
4694 		rmem->pg_tbl = mz->addr;
4695 		rmem->pg_tbl_map = mz_phys_addr;
4696 		rmem->pg_tbl_mz = mz;
4697 	}
4698 
4699 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4700 		 suffix, idx, bp->eth_dev->data->port_id);
4701 	mz = rte_memzone_lookup(mz_name);
4702 	if (!mz) {
4703 		mz = rte_memzone_reserve_aligned(mz_name,
4704 						 mem_size,
4705 						 bp->eth_dev->device->numa_node,
4706 						 RTE_MEMZONE_1GB |
4707 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4708 						 RTE_MEMZONE_IOVA_CONTIG,
4709 						 BNXT_PAGE_SIZE);
4710 		if (mz == NULL)
4711 			return -ENOMEM;
4712 	}
4713 
4714 	memset(mz->addr, 0, mz->len);
4715 	mz_phys_addr = mz->iova;
4716 
4717 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4718 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4719 		rmem->dma_arr[i] = mz_phys_addr + sz;
4720 
4721 		if (rmem->nr_pages > 1) {
4722 			if (i == rmem->nr_pages - 2 &&
4723 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4724 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4725 			else if (i == rmem->nr_pages - 1 &&
4726 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4727 				valid_bits |= PTU_PTE_LAST;
4728 
4729 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4730 							   valid_bits);
4731 		}
4732 	}
4733 
4734 	rmem->mz = mz;
4735 	if (rmem->vmem_size)
4736 		rmem->vmem = (void **)mz->addr;
4737 	rmem->dma_arr[0] = mz_phys_addr;
4738 	return 0;
4739 }
4740 
4741 static void bnxt_free_ctx_mem(struct bnxt *bp)
4742 {
4743 	int i;
4744 
4745 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4746 		return;
4747 
4748 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4749 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4750 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4751 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4752 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4753 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4754 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4755 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4756 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4757 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4758 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4759 
4760 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4761 		if (bp->ctx->tqm_mem[i])
4762 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4763 	}
4764 
4765 	rte_free(bp->ctx);
4766 	bp->ctx = NULL;
4767 }
4768 
4769 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4770 
4771 #define min_t(type, x, y) ({                    \
4772 	type __min1 = (x);                      \
4773 	type __min2 = (y);                      \
4774 	__min1 < __min2 ? __min1 : __min2; })
4775 
4776 #define max_t(type, x, y) ({                    \
4777 	type __max1 = (x);                      \
4778 	type __max2 = (y);                      \
4779 	__max1 > __max2 ? __max1 : __max2; })
4780 
4781 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4782 
4783 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4784 {
4785 	struct bnxt_ctx_pg_info *ctx_pg;
4786 	struct bnxt_ctx_mem_info *ctx;
4787 	uint32_t mem_size, ena, entries;
4788 	uint32_t entries_sp, min;
4789 	int i, rc;
4790 
4791 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4792 	if (rc) {
4793 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4794 		return rc;
4795 	}
4796 	ctx = bp->ctx;
4797 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4798 		return 0;
4799 
4800 	ctx_pg = &ctx->qp_mem;
4801 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4802 	if (ctx->qp_entry_size) {
4803 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4804 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4805 		if (rc)
4806 			return rc;
4807 	}
4808 
4809 	ctx_pg = &ctx->srq_mem;
4810 	ctx_pg->entries = ctx->srq_max_l2_entries;
4811 	if (ctx->srq_entry_size) {
4812 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4813 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4814 		if (rc)
4815 			return rc;
4816 	}
4817 
4818 	ctx_pg = &ctx->cq_mem;
4819 	ctx_pg->entries = ctx->cq_max_l2_entries;
4820 	if (ctx->cq_entry_size) {
4821 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4822 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4823 		if (rc)
4824 			return rc;
4825 	}
4826 
4827 	ctx_pg = &ctx->vnic_mem;
4828 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4829 		ctx->vnic_max_ring_table_entries;
4830 	if (ctx->vnic_entry_size) {
4831 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4832 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4833 		if (rc)
4834 			return rc;
4835 	}
4836 
4837 	ctx_pg = &ctx->stat_mem;
4838 	ctx_pg->entries = ctx->stat_max_entries;
4839 	if (ctx->stat_entry_size) {
4840 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4841 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4842 		if (rc)
4843 			return rc;
4844 	}
4845 
4846 	min = ctx->tqm_min_entries_per_ring;
4847 
4848 	entries_sp = ctx->qp_max_l2_entries +
4849 		     ctx->vnic_max_vnic_entries +
4850 		     2 * ctx->qp_min_qp1_entries + min;
4851 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4852 
4853 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4854 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4855 	entries = clamp_t(uint32_t, entries, min,
4856 			  ctx->tqm_max_entries_per_ring);
4857 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4858 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4859 		 * i > 8 is other ext rings.
4860 		 */
4861 		ctx_pg = ctx->tqm_mem[i];
4862 		ctx_pg->entries = i ? entries : entries_sp;
4863 		if (ctx->tqm_entry_size) {
4864 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4865 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4866 						    "tqm_mem", i);
4867 			if (rc)
4868 				return rc;
4869 		}
4870 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4871 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4872 		else
4873 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4874 	}
4875 
4876 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4877 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4878 	if (rc)
4879 		PMD_DRV_LOG(ERR,
4880 			    "Failed to configure context mem: rc = %d\n", rc);
4881 	else
4882 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4883 
4884 	return rc;
4885 }
4886 
4887 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4888 {
4889 	struct rte_pci_device *pci_dev = bp->pdev;
4890 	char mz_name[RTE_MEMZONE_NAMESIZE];
4891 	const struct rte_memzone *mz = NULL;
4892 	uint32_t total_alloc_len;
4893 	rte_iova_t mz_phys_addr;
4894 
4895 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4896 		return 0;
4897 
4898 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4899 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4900 		 pci_dev->addr.bus, pci_dev->addr.devid,
4901 		 pci_dev->addr.function, "rx_port_stats");
4902 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4903 	mz = rte_memzone_lookup(mz_name);
4904 	total_alloc_len =
4905 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4906 				       sizeof(struct rx_port_stats_ext) + 512);
4907 	if (!mz) {
4908 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4909 					 SOCKET_ID_ANY,
4910 					 RTE_MEMZONE_2MB |
4911 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4912 					 RTE_MEMZONE_IOVA_CONTIG);
4913 		if (mz == NULL)
4914 			return -ENOMEM;
4915 	}
4916 	memset(mz->addr, 0, mz->len);
4917 	mz_phys_addr = mz->iova;
4918 
4919 	bp->rx_mem_zone = (const void *)mz;
4920 	bp->hw_rx_port_stats = mz->addr;
4921 	bp->hw_rx_port_stats_map = mz_phys_addr;
4922 
4923 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4924 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4925 		 pci_dev->addr.bus, pci_dev->addr.devid,
4926 		 pci_dev->addr.function, "tx_port_stats");
4927 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4928 	mz = rte_memzone_lookup(mz_name);
4929 	total_alloc_len =
4930 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4931 				       sizeof(struct tx_port_stats_ext) + 512);
4932 	if (!mz) {
4933 		mz = rte_memzone_reserve(mz_name,
4934 					 total_alloc_len,
4935 					 SOCKET_ID_ANY,
4936 					 RTE_MEMZONE_2MB |
4937 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4938 					 RTE_MEMZONE_IOVA_CONTIG);
4939 		if (mz == NULL)
4940 			return -ENOMEM;
4941 	}
4942 	memset(mz->addr, 0, mz->len);
4943 	mz_phys_addr = mz->iova;
4944 
4945 	bp->tx_mem_zone = (const void *)mz;
4946 	bp->hw_tx_port_stats = mz->addr;
4947 	bp->hw_tx_port_stats_map = mz_phys_addr;
4948 	bp->flags |= BNXT_FLAG_PORT_STATS;
4949 
4950 	/* Display extended statistics if FW supports it */
4951 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4952 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4953 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4954 		return 0;
4955 
4956 	bp->hw_rx_port_stats_ext = (void *)
4957 		((uint8_t *)bp->hw_rx_port_stats +
4958 		 sizeof(struct rx_port_stats));
4959 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4960 		sizeof(struct rx_port_stats);
4961 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4962 
4963 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4964 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4965 		bp->hw_tx_port_stats_ext = (void *)
4966 			((uint8_t *)bp->hw_tx_port_stats +
4967 			 sizeof(struct tx_port_stats));
4968 		bp->hw_tx_port_stats_ext_map =
4969 			bp->hw_tx_port_stats_map +
4970 			sizeof(struct tx_port_stats);
4971 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4972 	}
4973 
4974 	return 0;
4975 }
4976 
4977 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4978 {
4979 	struct bnxt *bp = eth_dev->data->dev_private;
4980 	int rc = 0;
4981 
4982 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4983 					       RTE_ETHER_ADDR_LEN *
4984 					       bp->max_l2_ctx,
4985 					       0);
4986 	if (eth_dev->data->mac_addrs == NULL) {
4987 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4988 		return -ENOMEM;
4989 	}
4990 
4991 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4992 		if (BNXT_PF(bp))
4993 			return -EINVAL;
4994 
4995 		/* Generate a random MAC address, if none was assigned by PF */
4996 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4997 		bnxt_eth_hw_addr_random(bp->mac_addr);
4998 		PMD_DRV_LOG(INFO,
4999 			    "Assign random MAC:" RTE_ETHER_ADDR_PRT_FMT "\n",
5000 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
5001 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
5002 
5003 		rc = bnxt_hwrm_set_mac(bp);
5004 		if (rc)
5005 			return rc;
5006 	}
5007 
5008 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5009 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5010 
5011 	return rc;
5012 }
5013 
5014 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5015 {
5016 	int rc = 0;
5017 
5018 	/* MAC is already configured in FW */
5019 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5020 		return 0;
5021 
5022 	/* Restore the old MAC configured */
5023 	rc = bnxt_hwrm_set_mac(bp);
5024 	if (rc)
5025 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5026 
5027 	return rc;
5028 }
5029 
5030 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5031 {
5032 	if (!BNXT_PF(bp))
5033 		return;
5034 
5035 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5036 
5037 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5038 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5039 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5040 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5041 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5042 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5043 }
5044 
5045 struct bnxt *
5046 bnxt_get_bp(uint16_t port)
5047 {
5048 	struct bnxt *bp;
5049 	struct rte_eth_dev *dev;
5050 
5051 	if (!rte_eth_dev_is_valid_port(port)) {
5052 		PMD_DRV_LOG(ERR, "Invalid port %d\n", port);
5053 		return NULL;
5054 	}
5055 
5056 	dev = &rte_eth_devices[port];
5057 	if (!is_bnxt_supported(dev)) {
5058 		PMD_DRV_LOG(ERR, "Device %d not supported\n", port);
5059 		return NULL;
5060 	}
5061 
5062 	bp = (struct bnxt *)dev->data->dev_private;
5063 	if (!BNXT_TRUFLOW_EN(bp)) {
5064 		PMD_DRV_LOG(ERR, "TRUFLOW not enabled\n");
5065 		return NULL;
5066 	}
5067 
5068 	return bp;
5069 }
5070 
5071 uint16_t
5072 bnxt_get_svif(uint16_t port_id, bool func_svif,
5073 	      enum bnxt_ulp_intf_type type)
5074 {
5075 	struct rte_eth_dev *eth_dev;
5076 	struct bnxt *bp;
5077 
5078 	eth_dev = &rte_eth_devices[port_id];
5079 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5080 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5081 		if (!vfr)
5082 			return 0;
5083 
5084 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5085 			return vfr->svif;
5086 
5087 		eth_dev = vfr->parent_dev;
5088 	}
5089 
5090 	bp = eth_dev->data->dev_private;
5091 
5092 	return func_svif ? bp->func_svif : bp->port_svif;
5093 }
5094 
5095 void
5096 bnxt_get_iface_mac(uint16_t port, enum bnxt_ulp_intf_type type,
5097 		   uint8_t *mac, uint8_t *parent_mac)
5098 {
5099 	struct rte_eth_dev *eth_dev;
5100 	struct bnxt *bp;
5101 
5102 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF &&
5103 	    type != BNXT_ULP_INTF_TYPE_PF)
5104 		return;
5105 
5106 	eth_dev = &rte_eth_devices[port];
5107 	bp = eth_dev->data->dev_private;
5108 	memcpy(mac, bp->mac_addr, RTE_ETHER_ADDR_LEN);
5109 
5110 	if (type == BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5111 		memcpy(parent_mac, bp->parent->mac_addr, RTE_ETHER_ADDR_LEN);
5112 }
5113 
5114 uint16_t
5115 bnxt_get_parent_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5116 {
5117 	struct rte_eth_dev *eth_dev;
5118 	struct bnxt *bp;
5119 
5120 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5121 		return 0;
5122 
5123 	eth_dev = &rte_eth_devices[port];
5124 	bp = eth_dev->data->dev_private;
5125 
5126 	return bp->parent->vnic;
5127 }
5128 uint16_t
5129 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5130 {
5131 	struct rte_eth_dev *eth_dev;
5132 	struct bnxt_vnic_info *vnic;
5133 	struct bnxt *bp;
5134 
5135 	eth_dev = &rte_eth_devices[port];
5136 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5137 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5138 		if (!vfr)
5139 			return 0;
5140 
5141 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5142 			return vfr->dflt_vnic_id;
5143 
5144 		eth_dev = vfr->parent_dev;
5145 	}
5146 
5147 	bp = eth_dev->data->dev_private;
5148 
5149 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
5150 
5151 	return vnic->fw_vnic_id;
5152 }
5153 
5154 uint16_t
5155 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
5156 {
5157 	struct rte_eth_dev *eth_dev;
5158 	struct bnxt *bp;
5159 
5160 	eth_dev = &rte_eth_devices[port];
5161 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5162 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5163 		if (!vfr)
5164 			return 0;
5165 
5166 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5167 			return vfr->fw_fid;
5168 
5169 		eth_dev = vfr->parent_dev;
5170 	}
5171 
5172 	bp = eth_dev->data->dev_private;
5173 
5174 	return bp->fw_fid;
5175 }
5176 
5177 enum bnxt_ulp_intf_type
5178 bnxt_get_interface_type(uint16_t port)
5179 {
5180 	struct rte_eth_dev *eth_dev;
5181 	struct bnxt *bp;
5182 
5183 	eth_dev = &rte_eth_devices[port];
5184 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
5185 		return BNXT_ULP_INTF_TYPE_VF_REP;
5186 
5187 	bp = eth_dev->data->dev_private;
5188 	if (BNXT_PF(bp))
5189 		return BNXT_ULP_INTF_TYPE_PF;
5190 	else if (BNXT_VF_IS_TRUSTED(bp))
5191 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
5192 	else if (BNXT_VF(bp))
5193 		return BNXT_ULP_INTF_TYPE_VF;
5194 
5195 	return BNXT_ULP_INTF_TYPE_INVALID;
5196 }
5197 
5198 uint16_t
5199 bnxt_get_phy_port_id(uint16_t port_id)
5200 {
5201 	struct bnxt_representor *vfr;
5202 	struct rte_eth_dev *eth_dev;
5203 	struct bnxt *bp;
5204 
5205 	eth_dev = &rte_eth_devices[port_id];
5206 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5207 		vfr = eth_dev->data->dev_private;
5208 		if (!vfr)
5209 			return 0;
5210 
5211 		eth_dev = vfr->parent_dev;
5212 	}
5213 
5214 	bp = eth_dev->data->dev_private;
5215 
5216 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
5217 }
5218 
5219 uint16_t
5220 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
5221 {
5222 	struct rte_eth_dev *eth_dev;
5223 	struct bnxt *bp;
5224 
5225 	eth_dev = &rte_eth_devices[port_id];
5226 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5227 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5228 		if (!vfr)
5229 			return 0;
5230 
5231 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5232 			return vfr->fw_fid - 1;
5233 
5234 		eth_dev = vfr->parent_dev;
5235 	}
5236 
5237 	bp = eth_dev->data->dev_private;
5238 
5239 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
5240 }
5241 
5242 uint16_t
5243 bnxt_get_vport(uint16_t port_id)
5244 {
5245 	return (1 << bnxt_get_phy_port_id(port_id));
5246 }
5247 
5248 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5249 {
5250 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5251 
5252 	if (info) {
5253 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5254 			memset(info, 0, sizeof(*info));
5255 		return;
5256 	}
5257 
5258 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5259 		return;
5260 
5261 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5262 			   sizeof(*info), 0);
5263 	if (!info)
5264 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5265 
5266 	bp->recovery_info = info;
5267 }
5268 
5269 static void bnxt_check_fw_status(struct bnxt *bp)
5270 {
5271 	uint32_t fw_status;
5272 
5273 	if (!(bp->recovery_info &&
5274 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5275 		return;
5276 
5277 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5278 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5279 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5280 			    fw_status);
5281 }
5282 
5283 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5284 {
5285 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5286 	uint32_t status_loc;
5287 	uint32_t sig_ver;
5288 
5289 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5290 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5291 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5292 				   BNXT_GRCP_WINDOW_2_BASE +
5293 				   offsetof(struct hcomm_status,
5294 					    sig_ver)));
5295 	/* If the signature is absent, then FW does not support this feature */
5296 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5297 	    HCOMM_STATUS_SIGNATURE_VAL)
5298 		return 0;
5299 
5300 	if (!info) {
5301 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5302 				   sizeof(*info), 0);
5303 		if (!info)
5304 			return -ENOMEM;
5305 		bp->recovery_info = info;
5306 	} else {
5307 		memset(info, 0, sizeof(*info));
5308 	}
5309 
5310 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5311 				      BNXT_GRCP_WINDOW_2_BASE +
5312 				      offsetof(struct hcomm_status,
5313 					       fw_status_loc)));
5314 
5315 	/* Only pre-map the FW health status GRC register */
5316 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5317 		return 0;
5318 
5319 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5320 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5321 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5322 
5323 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5324 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5325 
5326 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5327 
5328 	return 0;
5329 }
5330 
5331 /* This function gets the FW version along with the
5332  * capabilities(MAX and current) of the function, vnic,
5333  * error recovery, phy and other chip related info
5334  */
5335 static int bnxt_get_config(struct bnxt *bp)
5336 {
5337 	uint16_t mtu;
5338 	int rc = 0;
5339 
5340 	bp->fw_cap = 0;
5341 
5342 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5343 	if (rc)
5344 		return rc;
5345 
5346 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5347 	if (rc) {
5348 		bnxt_check_fw_status(bp);
5349 		return rc;
5350 	}
5351 
5352 	rc = bnxt_hwrm_func_reset(bp);
5353 	if (rc)
5354 		return -EIO;
5355 
5356 	rc = bnxt_hwrm_vnic_qcaps(bp);
5357 	if (rc)
5358 		return rc;
5359 
5360 	rc = bnxt_hwrm_queue_qportcfg(bp);
5361 	if (rc)
5362 		return rc;
5363 
5364 	/* Get the MAX capabilities for this function.
5365 	 * This function also allocates context memory for TQM rings and
5366 	 * informs the firmware about this allocated backing store memory.
5367 	 */
5368 	rc = bnxt_hwrm_func_qcaps(bp);
5369 	if (rc)
5370 		return rc;
5371 
5372 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5373 	if (rc)
5374 		return rc;
5375 
5376 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5377 	if (rc)
5378 		return rc;
5379 
5380 	bnxt_hwrm_port_mac_qcfg(bp);
5381 
5382 	bnxt_hwrm_parent_pf_qcfg(bp);
5383 
5384 	bnxt_hwrm_port_phy_qcaps(bp);
5385 
5386 	bnxt_alloc_error_recovery_info(bp);
5387 	/* Get the adapter error recovery support info */
5388 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5389 	if (rc)
5390 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5391 
5392 	bnxt_hwrm_port_led_qcaps(bp);
5393 
5394 	return 0;
5395 }
5396 
5397 static int
5398 bnxt_init_locks(struct bnxt *bp)
5399 {
5400 	int err;
5401 
5402 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5403 	if (err) {
5404 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5405 		return err;
5406 	}
5407 
5408 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5409 	if (err) {
5410 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5411 		return err;
5412 	}
5413 
5414 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5415 	if (err) {
5416 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5417 		return err;
5418 	}
5419 
5420 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5421 	if (err)
5422 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5423 
5424 	return err;
5425 }
5426 
5427 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5428 {
5429 	int rc = 0;
5430 
5431 	rc = bnxt_get_config(bp);
5432 	if (rc)
5433 		return rc;
5434 
5435 	if (!reconfig_dev) {
5436 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5437 		if (rc)
5438 			return rc;
5439 	} else {
5440 		rc = bnxt_restore_dflt_mac(bp);
5441 		if (rc)
5442 			return rc;
5443 	}
5444 
5445 	bnxt_config_vf_req_fwd(bp);
5446 
5447 	rc = bnxt_hwrm_func_driver_register(bp);
5448 	if (rc) {
5449 		PMD_DRV_LOG(ERR, "Failed to register driver");
5450 		return -EBUSY;
5451 	}
5452 
5453 	if (BNXT_PF(bp)) {
5454 		if (bp->pdev->max_vfs) {
5455 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5456 			if (rc) {
5457 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5458 				return rc;
5459 			}
5460 		} else {
5461 			rc = bnxt_hwrm_allocate_pf_only(bp);
5462 			if (rc) {
5463 				PMD_DRV_LOG(ERR,
5464 					    "Failed to allocate PF resources");
5465 				return rc;
5466 			}
5467 		}
5468 	}
5469 
5470 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5471 	if (rc)
5472 		return rc;
5473 
5474 	rc = bnxt_setup_int(bp);
5475 	if (rc)
5476 		return rc;
5477 
5478 	rc = bnxt_request_int(bp);
5479 	if (rc)
5480 		return rc;
5481 
5482 	rc = bnxt_init_ctx_mem(bp);
5483 	if (rc) {
5484 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5485 		return rc;
5486 	}
5487 
5488 	return 0;
5489 }
5490 
5491 static int
5492 bnxt_parse_devarg_accum_stats(__rte_unused const char *key,
5493 			      const char *value, void *opaque_arg)
5494 {
5495 	struct bnxt *bp = opaque_arg;
5496 	unsigned long accum_stats;
5497 	char *end = NULL;
5498 
5499 	if (!value || !opaque_arg) {
5500 		PMD_DRV_LOG(ERR,
5501 			    "Invalid parameter passed to accum-stats devargs.\n");
5502 		return -EINVAL;
5503 	}
5504 
5505 	accum_stats = strtoul(value, &end, 10);
5506 	if (end == NULL || *end != '\0' ||
5507 	    (accum_stats == ULONG_MAX && errno == ERANGE)) {
5508 		PMD_DRV_LOG(ERR,
5509 			    "Invalid parameter passed to accum-stats devargs.\n");
5510 		return -EINVAL;
5511 	}
5512 
5513 	if (BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)) {
5514 		PMD_DRV_LOG(ERR,
5515 			    "Invalid value passed to accum-stats devargs.\n");
5516 		return -EINVAL;
5517 	}
5518 
5519 	if (accum_stats) {
5520 		bp->flags2 |= BNXT_FLAGS2_ACCUM_STATS_EN;
5521 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature enabled.\n");
5522 	} else {
5523 		bp->flags2 &= ~BNXT_FLAGS2_ACCUM_STATS_EN;
5524 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature disabled.\n");
5525 	}
5526 
5527 	return 0;
5528 }
5529 
5530 static int
5531 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5532 			     const char *value, void *opaque_arg)
5533 {
5534 	struct bnxt *bp = opaque_arg;
5535 	unsigned long flow_xstat;
5536 	char *end = NULL;
5537 
5538 	if (!value || !opaque_arg) {
5539 		PMD_DRV_LOG(ERR,
5540 			    "Invalid parameter passed to flow_xstat devarg.\n");
5541 		return -EINVAL;
5542 	}
5543 
5544 	flow_xstat = strtoul(value, &end, 10);
5545 	if (end == NULL || *end != '\0' ||
5546 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5547 		PMD_DRV_LOG(ERR,
5548 			    "Invalid parameter passed to flow_xstat devarg.\n");
5549 		return -EINVAL;
5550 	}
5551 
5552 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5553 		PMD_DRV_LOG(ERR,
5554 			    "Invalid value passed to flow_xstat devarg.\n");
5555 		return -EINVAL;
5556 	}
5557 
5558 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5559 	if (BNXT_FLOW_XSTATS_EN(bp))
5560 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5561 
5562 	return 0;
5563 }
5564 
5565 static int
5566 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5567 					const char *value, void *opaque_arg)
5568 {
5569 	struct bnxt *bp = opaque_arg;
5570 	unsigned long max_num_kflows;
5571 	char *end = NULL;
5572 
5573 	if (!value || !opaque_arg) {
5574 		PMD_DRV_LOG(ERR,
5575 			"Invalid parameter passed to max_num_kflows devarg.\n");
5576 		return -EINVAL;
5577 	}
5578 
5579 	max_num_kflows = strtoul(value, &end, 10);
5580 	if (end == NULL || *end != '\0' ||
5581 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5582 		PMD_DRV_LOG(ERR,
5583 			"Invalid parameter passed to max_num_kflows devarg.\n");
5584 		return -EINVAL;
5585 	}
5586 
5587 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5588 		PMD_DRV_LOG(ERR,
5589 			"Invalid value passed to max_num_kflows devarg.\n");
5590 		return -EINVAL;
5591 	}
5592 
5593 	bp->max_num_kflows = max_num_kflows;
5594 	if (bp->max_num_kflows)
5595 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5596 				max_num_kflows);
5597 
5598 	return 0;
5599 }
5600 
5601 static int
5602 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5603 				 const char *value, void *opaque_arg)
5604 {
5605 	struct bnxt *bp = opaque_arg;
5606 	unsigned long app_id;
5607 	char *end = NULL;
5608 
5609 	if (!value || !opaque_arg) {
5610 		PMD_DRV_LOG(ERR,
5611 			    "Invalid parameter passed to app-id "
5612 			    "devargs.\n");
5613 		return -EINVAL;
5614 	}
5615 
5616 	app_id = strtoul(value, &end, 10);
5617 	if (end == NULL || *end != '\0' ||
5618 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5619 		PMD_DRV_LOG(ERR,
5620 			    "Invalid parameter passed to app_id "
5621 			    "devargs.\n");
5622 		return -EINVAL;
5623 	}
5624 
5625 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5626 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5627 			    (uint16_t)app_id);
5628 		return -EINVAL;
5629 	}
5630 
5631 	bp->app_id = app_id;
5632 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5633 
5634 	return 0;
5635 }
5636 
5637 static int
5638 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5639 			    const char *value, void *opaque_arg)
5640 {
5641 	struct bnxt_representor *vfr_bp = opaque_arg;
5642 	unsigned long rep_is_pf;
5643 	char *end = NULL;
5644 
5645 	if (!value || !opaque_arg) {
5646 		PMD_DRV_LOG(ERR,
5647 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5648 		return -EINVAL;
5649 	}
5650 
5651 	rep_is_pf = strtoul(value, &end, 10);
5652 	if (end == NULL || *end != '\0' ||
5653 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5654 		PMD_DRV_LOG(ERR,
5655 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5656 		return -EINVAL;
5657 	}
5658 
5659 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5660 		PMD_DRV_LOG(ERR,
5661 			    "Invalid value passed to rep_is_pf devargs.\n");
5662 		return -EINVAL;
5663 	}
5664 
5665 	vfr_bp->flags |= rep_is_pf;
5666 	if (BNXT_REP_PF(vfr_bp))
5667 		PMD_DRV_LOG(INFO, "PF representor\n");
5668 	else
5669 		PMD_DRV_LOG(INFO, "VF representor\n");
5670 
5671 	return 0;
5672 }
5673 
5674 static int
5675 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5676 			       const char *value, void *opaque_arg)
5677 {
5678 	struct bnxt_representor *vfr_bp = opaque_arg;
5679 	unsigned long rep_based_pf;
5680 	char *end = NULL;
5681 
5682 	if (!value || !opaque_arg) {
5683 		PMD_DRV_LOG(ERR,
5684 			    "Invalid parameter passed to rep_based_pf "
5685 			    "devargs.\n");
5686 		return -EINVAL;
5687 	}
5688 
5689 	rep_based_pf = strtoul(value, &end, 10);
5690 	if (end == NULL || *end != '\0' ||
5691 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5692 		PMD_DRV_LOG(ERR,
5693 			    "Invalid parameter passed to rep_based_pf "
5694 			    "devargs.\n");
5695 		return -EINVAL;
5696 	}
5697 
5698 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5699 		PMD_DRV_LOG(ERR,
5700 			    "Invalid value passed to rep_based_pf devargs.\n");
5701 		return -EINVAL;
5702 	}
5703 
5704 	vfr_bp->rep_based_pf = rep_based_pf;
5705 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5706 
5707 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5708 
5709 	return 0;
5710 }
5711 
5712 static int
5713 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5714 			    const char *value, void *opaque_arg)
5715 {
5716 	struct bnxt_representor *vfr_bp = opaque_arg;
5717 	unsigned long rep_q_r2f;
5718 	char *end = NULL;
5719 
5720 	if (!value || !opaque_arg) {
5721 		PMD_DRV_LOG(ERR,
5722 			    "Invalid parameter passed to rep_q_r2f "
5723 			    "devargs.\n");
5724 		return -EINVAL;
5725 	}
5726 
5727 	rep_q_r2f = strtoul(value, &end, 10);
5728 	if (end == NULL || *end != '\0' ||
5729 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5730 		PMD_DRV_LOG(ERR,
5731 			    "Invalid parameter passed to rep_q_r2f "
5732 			    "devargs.\n");
5733 		return -EINVAL;
5734 	}
5735 
5736 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5737 		PMD_DRV_LOG(ERR,
5738 			    "Invalid value passed to rep_q_r2f devargs.\n");
5739 		return -EINVAL;
5740 	}
5741 
5742 	vfr_bp->rep_q_r2f = rep_q_r2f;
5743 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5744 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5745 
5746 	return 0;
5747 }
5748 
5749 static int
5750 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5751 			    const char *value, void *opaque_arg)
5752 {
5753 	struct bnxt_representor *vfr_bp = opaque_arg;
5754 	unsigned long rep_q_f2r;
5755 	char *end = NULL;
5756 
5757 	if (!value || !opaque_arg) {
5758 		PMD_DRV_LOG(ERR,
5759 			    "Invalid parameter passed to rep_q_f2r "
5760 			    "devargs.\n");
5761 		return -EINVAL;
5762 	}
5763 
5764 	rep_q_f2r = strtoul(value, &end, 10);
5765 	if (end == NULL || *end != '\0' ||
5766 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5767 		PMD_DRV_LOG(ERR,
5768 			    "Invalid parameter passed to rep_q_f2r "
5769 			    "devargs.\n");
5770 		return -EINVAL;
5771 	}
5772 
5773 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5774 		PMD_DRV_LOG(ERR,
5775 			    "Invalid value passed to rep_q_f2r devargs.\n");
5776 		return -EINVAL;
5777 	}
5778 
5779 	vfr_bp->rep_q_f2r = rep_q_f2r;
5780 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5781 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5782 
5783 	return 0;
5784 }
5785 
5786 static int
5787 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5788 			     const char *value, void *opaque_arg)
5789 {
5790 	struct bnxt_representor *vfr_bp = opaque_arg;
5791 	unsigned long rep_fc_r2f;
5792 	char *end = NULL;
5793 
5794 	if (!value || !opaque_arg) {
5795 		PMD_DRV_LOG(ERR,
5796 			    "Invalid parameter passed to rep_fc_r2f "
5797 			    "devargs.\n");
5798 		return -EINVAL;
5799 	}
5800 
5801 	rep_fc_r2f = strtoul(value, &end, 10);
5802 	if (end == NULL || *end != '\0' ||
5803 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5804 		PMD_DRV_LOG(ERR,
5805 			    "Invalid parameter passed to rep_fc_r2f "
5806 			    "devargs.\n");
5807 		return -EINVAL;
5808 	}
5809 
5810 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5811 		PMD_DRV_LOG(ERR,
5812 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5813 		return -EINVAL;
5814 	}
5815 
5816 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5817 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5818 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5819 
5820 	return 0;
5821 }
5822 
5823 static int
5824 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5825 			     const char *value, void *opaque_arg)
5826 {
5827 	struct bnxt_representor *vfr_bp = opaque_arg;
5828 	unsigned long rep_fc_f2r;
5829 	char *end = NULL;
5830 
5831 	if (!value || !opaque_arg) {
5832 		PMD_DRV_LOG(ERR,
5833 			    "Invalid parameter passed to rep_fc_f2r "
5834 			    "devargs.\n");
5835 		return -EINVAL;
5836 	}
5837 
5838 	rep_fc_f2r = strtoul(value, &end, 10);
5839 	if (end == NULL || *end != '\0' ||
5840 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5841 		PMD_DRV_LOG(ERR,
5842 			    "Invalid parameter passed to rep_fc_f2r "
5843 			    "devargs.\n");
5844 		return -EINVAL;
5845 	}
5846 
5847 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5848 		PMD_DRV_LOG(ERR,
5849 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5850 		return -EINVAL;
5851 	}
5852 
5853 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5854 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5855 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5856 
5857 	return 0;
5858 }
5859 
5860 static int
5861 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5862 {
5863 	struct rte_kvargs *kvlist;
5864 	int ret;
5865 
5866 	if (devargs == NULL)
5867 		return 0;
5868 
5869 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5870 	if (kvlist == NULL)
5871 		return -EINVAL;
5872 
5873 	/*
5874 	 * Handler for "flow_xstat" devarg.
5875 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5876 	 */
5877 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5878 				 bnxt_parse_devarg_flow_xstat, bp);
5879 	if (ret)
5880 		goto err;
5881 
5882 	/*
5883 	 * Handler for "accum-stats" devarg.
5884 	 * Invoked as for ex: "-a 0000:00:0d.0,accum-stats=1"
5885 	 */
5886 	rte_kvargs_process(kvlist, BNXT_DEVARG_ACCUM_STATS,
5887 			   bnxt_parse_devarg_accum_stats, bp);
5888 	/*
5889 	 * Handler for "max_num_kflows" devarg.
5890 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5891 	 */
5892 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5893 				 bnxt_parse_devarg_max_num_kflows, bp);
5894 	if (ret)
5895 		goto err;
5896 
5897 err:
5898 	/*
5899 	 * Handler for "app-id" devarg.
5900 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5901 	 */
5902 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5903 			   bnxt_parse_devarg_app_id, bp);
5904 
5905 	rte_kvargs_free(kvlist);
5906 	return ret;
5907 }
5908 
5909 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5910 {
5911 	int rc = 0;
5912 
5913 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5914 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5915 		if (rc)
5916 			PMD_DRV_LOG(ERR,
5917 				    "Failed to alloc switch domain: %d\n", rc);
5918 		else
5919 			PMD_DRV_LOG(INFO,
5920 				    "Switch domain allocated %d\n",
5921 				    bp->switch_domain_id);
5922 	}
5923 
5924 	return rc;
5925 }
5926 
5927 /* Allocate and initialize various fields in bnxt struct that
5928  * need to be allocated/destroyed only once in the lifetime of the driver
5929  */
5930 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5931 {
5932 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5933 	struct bnxt *bp = eth_dev->data->dev_private;
5934 	int rc = 0;
5935 
5936 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5937 
5938 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5939 		bp->flags |= BNXT_FLAG_VF;
5940 
5941 	if (bnxt_p5_device(pci_dev->id.device_id))
5942 		bp->flags |= BNXT_FLAG_CHIP_P5;
5943 
5944 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5945 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5946 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5947 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5948 		bp->flags |= BNXT_FLAG_STINGRAY;
5949 
5950 	if (BNXT_TRUFLOW_EN(bp)) {
5951 		/* extra mbuf field is required to store CFA code from mark */
5952 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5953 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5954 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5955 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5956 		};
5957 		bnxt_cfa_code_dynfield_offset =
5958 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5959 		if (bnxt_cfa_code_dynfield_offset < 0) {
5960 			PMD_DRV_LOG(ERR,
5961 			    "Failed to register mbuf field for TruFlow mark\n");
5962 			return -rte_errno;
5963 		}
5964 	}
5965 
5966 	rc = bnxt_map_pci_bars(eth_dev);
5967 	if (rc) {
5968 		PMD_DRV_LOG(ERR,
5969 			    "Failed to initialize board rc: %x\n", rc);
5970 		return rc;
5971 	}
5972 
5973 	rc = bnxt_alloc_pf_info(bp);
5974 	if (rc)
5975 		return rc;
5976 
5977 	rc = bnxt_alloc_link_info(bp);
5978 	if (rc)
5979 		return rc;
5980 
5981 	rc = bnxt_alloc_parent_info(bp);
5982 	if (rc)
5983 		return rc;
5984 
5985 	rc = bnxt_alloc_hwrm_resources(bp);
5986 	if (rc) {
5987 		PMD_DRV_LOG(ERR,
5988 			    "Failed to allocate response buffer rc: %x\n", rc);
5989 		return rc;
5990 	}
5991 	rc = bnxt_alloc_leds_info(bp);
5992 	if (rc)
5993 		return rc;
5994 
5995 	rc = bnxt_alloc_cos_queues(bp);
5996 	if (rc)
5997 		return rc;
5998 
5999 	rc = bnxt_init_locks(bp);
6000 	if (rc)
6001 		return rc;
6002 
6003 	rc = bnxt_alloc_switch_domain(bp);
6004 	if (rc)
6005 		return rc;
6006 
6007 	return rc;
6008 }
6009 
6010 static int
6011 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
6012 {
6013 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
6014 	static int version_printed;
6015 	struct bnxt *bp;
6016 	int rc;
6017 
6018 	if (version_printed++ == 0)
6019 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
6020 
6021 	eth_dev->dev_ops = &bnxt_dev_ops;
6022 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
6023 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
6024 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
6025 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
6026 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
6027 
6028 	/*
6029 	 * For secondary processes, we don't initialise any further
6030 	 * as primary has already done this work.
6031 	 */
6032 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6033 		return 0;
6034 
6035 	rte_eth_copy_pci_info(eth_dev, pci_dev);
6036 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
6037 
6038 	bp = eth_dev->data->dev_private;
6039 
6040 	/* Parse dev arguments passed on when starting the DPDK application. */
6041 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
6042 	if (rc)
6043 		goto error_free;
6044 
6045 	rc = bnxt_drv_init(eth_dev);
6046 	if (rc)
6047 		goto error_free;
6048 
6049 	rc = bnxt_init_resources(bp, false);
6050 	if (rc)
6051 		goto error_free;
6052 
6053 	rc = bnxt_alloc_stats_mem(bp);
6054 	if (rc)
6055 		goto error_free;
6056 
6057 	PMD_DRV_LOG(INFO,
6058 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
6059 		    DRV_MODULE_NAME,
6060 		    pci_dev->mem_resource[0].phys_addr,
6061 		    pci_dev->mem_resource[0].addr);
6062 
6063 	return 0;
6064 
6065 error_free:
6066 	bnxt_dev_uninit(eth_dev);
6067 	return rc;
6068 }
6069 
6070 
6071 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
6072 {
6073 	if (!ctx)
6074 		return;
6075 
6076 	if (ctx->va)
6077 		rte_free(ctx->va);
6078 
6079 	ctx->va = NULL;
6080 	ctx->dma = RTE_BAD_IOVA;
6081 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
6082 }
6083 
6084 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
6085 {
6086 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
6087 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6088 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
6089 				  bp->flow_stat->max_fc,
6090 				  false);
6091 
6092 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
6093 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6094 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
6095 				  bp->flow_stat->max_fc,
6096 				  false);
6097 
6098 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6099 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
6100 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6101 
6102 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6103 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
6104 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6105 
6106 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6107 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
6108 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6109 
6110 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6111 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
6112 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6113 }
6114 
6115 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
6116 {
6117 	bnxt_unregister_fc_ctx_mem(bp);
6118 
6119 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
6120 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
6121 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
6122 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
6123 }
6124 
6125 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
6126 {
6127 	if (BNXT_FLOW_XSTATS_EN(bp))
6128 		bnxt_uninit_fc_ctx_mem(bp);
6129 }
6130 
6131 static void
6132 bnxt_free_error_recovery_info(struct bnxt *bp)
6133 {
6134 	rte_free(bp->recovery_info);
6135 	bp->recovery_info = NULL;
6136 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
6137 }
6138 
6139 static int
6140 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
6141 {
6142 	int rc;
6143 
6144 	bnxt_free_int(bp);
6145 	bnxt_free_mem(bp, reconfig_dev);
6146 
6147 	bnxt_hwrm_func_buf_unrgtr(bp);
6148 	if (bp->pf != NULL) {
6149 		rte_free(bp->pf->vf_req_buf);
6150 		bp->pf->vf_req_buf = NULL;
6151 	}
6152 
6153 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
6154 	bp->flags &= ~BNXT_FLAG_REGISTERED;
6155 	bnxt_free_ctx_mem(bp);
6156 	if (!reconfig_dev) {
6157 		bnxt_free_hwrm_resources(bp);
6158 		bnxt_free_error_recovery_info(bp);
6159 	}
6160 
6161 	bnxt_uninit_ctx_mem(bp);
6162 
6163 	bnxt_free_flow_stats_info(bp);
6164 	if (bp->rep_info != NULL)
6165 		bnxt_free_switch_domain(bp);
6166 	bnxt_free_rep_info(bp);
6167 	rte_free(bp->ptp_cfg);
6168 	bp->ptp_cfg = NULL;
6169 	return rc;
6170 }
6171 
6172 static int
6173 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
6174 {
6175 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6176 		return -EPERM;
6177 
6178 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
6179 
6180 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
6181 		bnxt_dev_close_op(eth_dev);
6182 
6183 	return 0;
6184 }
6185 
6186 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6187 {
6188 	struct bnxt *bp = eth_dev->data->dev_private;
6189 	struct rte_eth_dev *vf_rep_eth_dev;
6190 	int ret = 0, i;
6191 
6192 	if (!bp)
6193 		return -EINVAL;
6194 
6195 	for (i = 0; i < bp->num_reps; i++) {
6196 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6197 		if (!vf_rep_eth_dev)
6198 			continue;
6199 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6200 			    vf_rep_eth_dev->data->port_id);
6201 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6202 	}
6203 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6204 		    eth_dev->data->port_id);
6205 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6206 
6207 	return ret;
6208 }
6209 
6210 static void bnxt_free_rep_info(struct bnxt *bp)
6211 {
6212 	rte_free(bp->rep_info);
6213 	bp->rep_info = NULL;
6214 	rte_free(bp->cfa_code_map);
6215 	bp->cfa_code_map = NULL;
6216 }
6217 
6218 static int bnxt_init_rep_info(struct bnxt *bp)
6219 {
6220 	int i = 0, rc;
6221 
6222 	if (bp->rep_info)
6223 		return 0;
6224 
6225 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
6226 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
6227 				   0);
6228 	if (!bp->rep_info) {
6229 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6230 		return -ENOMEM;
6231 	}
6232 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6233 				       sizeof(*bp->cfa_code_map) *
6234 				       BNXT_MAX_CFA_CODE, 0);
6235 	if (!bp->cfa_code_map) {
6236 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6237 		bnxt_free_rep_info(bp);
6238 		return -ENOMEM;
6239 	}
6240 
6241 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6242 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6243 
6244 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6245 	if (rc) {
6246 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6247 		bnxt_free_rep_info(bp);
6248 		return rc;
6249 	}
6250 
6251 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6252 	if (rc) {
6253 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6254 		bnxt_free_rep_info(bp);
6255 		return rc;
6256 	}
6257 
6258 	return rc;
6259 }
6260 
6261 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6262 			       struct rte_eth_devargs *eth_da,
6263 			       struct rte_eth_dev *backing_eth_dev,
6264 			       const char *dev_args)
6265 {
6266 	struct rte_eth_dev *vf_rep_eth_dev;
6267 	char name[RTE_ETH_NAME_MAX_LEN];
6268 	struct bnxt *backing_bp;
6269 	uint16_t num_rep;
6270 	int i, ret = 0;
6271 	struct rte_kvargs *kvlist = NULL;
6272 
6273 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6274 		return 0;
6275 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6276 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6277 			    eth_da->type);
6278 		return -ENOTSUP;
6279 	}
6280 	num_rep = eth_da->nb_representor_ports;
6281 	if (num_rep > BNXT_MAX_VF_REPS) {
6282 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6283 			    num_rep, BNXT_MAX_VF_REPS);
6284 		return -EINVAL;
6285 	}
6286 
6287 	if (num_rep >= RTE_MAX_ETHPORTS) {
6288 		PMD_DRV_LOG(ERR,
6289 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6290 			    num_rep, RTE_MAX_ETHPORTS);
6291 		return -EINVAL;
6292 	}
6293 
6294 	backing_bp = backing_eth_dev->data->dev_private;
6295 
6296 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6297 		PMD_DRV_LOG(ERR,
6298 			    "Not a PF or trusted VF. No Representor support\n");
6299 		/* Returning an error is not an option.
6300 		 * Applications are not handling this correctly
6301 		 */
6302 		return 0;
6303 	}
6304 
6305 	if (bnxt_init_rep_info(backing_bp))
6306 		return 0;
6307 
6308 	for (i = 0; i < num_rep; i++) {
6309 		struct bnxt_representor representor = {
6310 			.vf_id = eth_da->representor_ports[i],
6311 			.switch_domain_id = backing_bp->switch_domain_id,
6312 			.parent_dev = backing_eth_dev
6313 		};
6314 
6315 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
6316 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6317 				    representor.vf_id, BNXT_MAX_VF_REPS);
6318 			continue;
6319 		}
6320 
6321 		/* representor port net_bdf_port */
6322 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6323 			 pci_dev->device.name, eth_da->representor_ports[i]);
6324 
6325 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6326 		if (kvlist) {
6327 			/*
6328 			 * Handler for "rep_is_pf" devarg.
6329 			 * Invoked as for ex: "-a 000:00:0d.0,
6330 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6331 			 */
6332 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6333 						 bnxt_parse_devarg_rep_is_pf,
6334 						 (void *)&representor);
6335 			if (ret) {
6336 				ret = -EINVAL;
6337 				goto err;
6338 			}
6339 			/*
6340 			 * Handler for "rep_based_pf" devarg.
6341 			 * Invoked as for ex: "-a 000:00:0d.0,
6342 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6343 			 */
6344 			ret = rte_kvargs_process(kvlist,
6345 						 BNXT_DEVARG_REP_BASED_PF,
6346 						 bnxt_parse_devarg_rep_based_pf,
6347 						 (void *)&representor);
6348 			if (ret) {
6349 				ret = -EINVAL;
6350 				goto err;
6351 			}
6352 			/*
6353 			 * Handler for "rep_based_pf" devarg.
6354 			 * Invoked as for ex: "-a 000:00:0d.0,
6355 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6356 			 */
6357 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6358 						 bnxt_parse_devarg_rep_q_r2f,
6359 						 (void *)&representor);
6360 			if (ret) {
6361 				ret = -EINVAL;
6362 				goto err;
6363 			}
6364 			/*
6365 			 * Handler for "rep_based_pf" devarg.
6366 			 * Invoked as for ex: "-a 000:00:0d.0,
6367 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6368 			 */
6369 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6370 						 bnxt_parse_devarg_rep_q_f2r,
6371 						 (void *)&representor);
6372 			if (ret) {
6373 				ret = -EINVAL;
6374 				goto err;
6375 			}
6376 			/*
6377 			 * Handler for "rep_based_pf" devarg.
6378 			 * Invoked as for ex: "-a 000:00:0d.0,
6379 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6380 			 */
6381 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6382 						 bnxt_parse_devarg_rep_fc_r2f,
6383 						 (void *)&representor);
6384 			if (ret) {
6385 				ret = -EINVAL;
6386 				goto err;
6387 			}
6388 			/*
6389 			 * Handler for "rep_based_pf" devarg.
6390 			 * Invoked as for ex: "-a 000:00:0d.0,
6391 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6392 			 */
6393 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6394 						 bnxt_parse_devarg_rep_fc_f2r,
6395 						 (void *)&representor);
6396 			if (ret) {
6397 				ret = -EINVAL;
6398 				goto err;
6399 			}
6400 		}
6401 
6402 		ret = rte_eth_dev_create(&pci_dev->device, name,
6403 					 sizeof(struct bnxt_representor),
6404 					 NULL, NULL,
6405 					 bnxt_representor_init,
6406 					 &representor);
6407 		if (ret) {
6408 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6409 				    "representor %s.", name);
6410 			goto err;
6411 		}
6412 
6413 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6414 		if (!vf_rep_eth_dev) {
6415 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6416 				    " for VF-Rep: %s.", name);
6417 			ret = -ENODEV;
6418 			goto err;
6419 		}
6420 
6421 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6422 			    backing_eth_dev->data->port_id);
6423 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6424 							 vf_rep_eth_dev;
6425 		backing_bp->num_reps++;
6426 
6427 	}
6428 
6429 	rte_kvargs_free(kvlist);
6430 	return 0;
6431 
6432 err:
6433 	/* If num_rep > 1, then rollback already created
6434 	 * ports, since we'll be failing the probe anyway
6435 	 */
6436 	if (num_rep > 1)
6437 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6438 	rte_errno = -ret;
6439 	rte_kvargs_free(kvlist);
6440 
6441 	return ret;
6442 }
6443 
6444 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6445 			  struct rte_pci_device *pci_dev)
6446 {
6447 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6448 	struct rte_eth_dev *backing_eth_dev;
6449 	uint16_t num_rep;
6450 	int ret = 0;
6451 
6452 	if (pci_dev->device.devargs) {
6453 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6454 					    &eth_da);
6455 		if (ret)
6456 			return ret;
6457 	}
6458 
6459 	num_rep = eth_da.nb_representor_ports;
6460 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6461 		    num_rep);
6462 
6463 	/* We could come here after first level of probe is already invoked
6464 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6465 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6466 	 */
6467 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6468 	if (backing_eth_dev == NULL) {
6469 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6470 					 sizeof(struct bnxt),
6471 					 eth_dev_pci_specific_init, pci_dev,
6472 					 bnxt_dev_init, NULL);
6473 
6474 		if (ret || !num_rep)
6475 			return ret;
6476 
6477 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6478 	}
6479 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6480 		    backing_eth_dev->data->port_id);
6481 
6482 	if (!num_rep)
6483 		return ret;
6484 
6485 	/* probe representor ports now */
6486 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6487 				  pci_dev->device.devargs->args);
6488 
6489 	return ret;
6490 }
6491 
6492 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6493 {
6494 	struct rte_eth_dev *eth_dev;
6495 
6496 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6497 	if (!eth_dev)
6498 		return 0; /* Invoked typically only by OVS-DPDK, by the
6499 			   * time it comes here the eth_dev is already
6500 			   * deleted by rte_eth_dev_close(), so returning
6501 			   * +ve value will at least help in proper cleanup
6502 			   */
6503 
6504 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6505 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6506 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6507 			return rte_eth_dev_destroy(eth_dev,
6508 						   bnxt_representor_uninit);
6509 		else
6510 			return rte_eth_dev_destroy(eth_dev,
6511 						   bnxt_dev_uninit);
6512 	} else {
6513 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6514 	}
6515 }
6516 
6517 static struct rte_pci_driver bnxt_rte_pmd = {
6518 	.id_table = bnxt_pci_id_map,
6519 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6520 			RTE_PCI_DRV_INTR_RMV |
6521 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6522 						  * and OVS-DPDK
6523 						  */
6524 	.probe = bnxt_pci_probe,
6525 	.remove = bnxt_pci_remove,
6526 };
6527 
6528 static bool
6529 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6530 {
6531 	if (strcmp(dev->device->driver->name, drv->driver.name))
6532 		return false;
6533 
6534 	return true;
6535 }
6536 
6537 bool is_bnxt_supported(struct rte_eth_dev *dev)
6538 {
6539 	return is_device_supported(dev, &bnxt_rte_pmd);
6540 }
6541 
6542 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6543 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6544 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6545 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6546